DIE GRUNDLEHREN DER

MATHEM ATISCHE N
WISSEN SCHAFT EN
IN EINZELDARSTELLUN GEN MIT BESONDERER
BERUCKSICHTIGUNG DER ANWENDUNGSGEBIET E

HERAUSGEGEBEN VON
R. GRAMMEL. E. HPF. H. HPF. F. RELLICH
F. K. SCHMIDT. B. L. VAN DER WAERDEN

VOLUME X
RICCI-CALCULUS
AN INTRODUCTION TO TENSOR ANALYSIS
AND ITS GEOMETRICAL APPLICATIONS

BY
J. A. SCHOUTEN

SECOND EDITION

SPRINGER-VERLAG BERLIN HEIDELBERG GMBH

1954
 RICCI-CALCULUS
AN INTRODUCTION TO TENSOR ANALYSIS
AND ITS GEOMETRICAL APPLICATIONS

BY

J. A. SCHOUTEN
EMERITUS PROFESSOR OF MATHEMATICS IN THE UNlVERS!TY
OF AMSTERDAM
DIRECTOR OF THE MATHEMATICAL CENTRE AT AMSTERDAM

SECOND EDITION

WITH 16 FIGURES

SPRINGER-VERLAG BERLIN HEIDELBERG GMBH

1954
 ISBN 978-3-642-05692-5 ISBN 978-3-662-12927-2 (eBook)
DOI 10.1007/978-3-662-12927-2

ALLE RECHTE,
INSBESONDERE DAS DER BERSETZUNG IN FREMDE SPRACHEN
VORBEHALTEN

OHNE AUSDRCKLICHE GENEHMIGUNG DES VERLAGES

IST ES AUCH NICHT GESTATTET, DIESES BUCH ODER TEILE DARAUS
AUF PHOTOMECHANISCHEM WEGE (PHOTOKOPIE, MIKROKOPIE) ZU VERVIELFALTIGEN

COPYRIGHT 1954
BY SPRINGER-VERLAG BERLIN HEIDELBERG
URSPRNGLICHERSCHIENENBEI SPRINGER-VERLAG OHG. IN BERLIN, GTTINGEN AND HEIDELBERG I954
SOFTCOVER REPRINT OF THE HARDCOVER 2ND EDITION I954
 This book is dedicated to the memory of

DR. GREGORIO RICCI CURBASTRO

in life Professor of Mathematics in the

University of Padua,
who laid the foundations
of tensor calculus.
 Preface to the second edition.
This is an entirely new book. The first edition appeared in 1923 and
at that time it was up to date. But in 193 5 and 1938 the author and Prof.
D. J. STRUIK published a new book, their Einfhrung I and li, and
this book not only gave the first systematic introduction to the kernel-
index method but also contained many notions that had come into
prominence since 1923. For instance densities, quantities of the second
kind, pseudo-quantities, normal Coordinates, the symbolism of exterior
forms, the LIE derivative, the theory of variation and deformation and
the theory of subprojective connexions were included. Now since 1938
there have been many new developments and so a book on RICCI cal-
culus and its applications has to cover quite different ground from the
book of 1923. Though the purpose remains to make the reader acquainted
with RICCI's famous instrument in its modern form, the book must
have quite a different methodical structure and quite different applica-
tions have to be chosen.
The first chapter contains algebraical preliminaries but the whole
text is modernized and there is a section on hybrid quantities (quantities
with indices of the first and of the second kind) and one on the many
abridged notations that have been developed by several authors.
In the second chapter the most important analytical notions that
come before the introduction of a connexion aredealt with in full. The
theory of integrability and PFAFF's problern are treated here and do
not get a chapter of their own this time as they did in the first edition
because ScHOUTEN and v. D. KuLK's book on PFAFF's problern of 1949
can be referred to. A special section pays attention to CARTAN's sym-
bolism.
In the first edition the chapters IV-VI contained the theory of
affine, riemannian and WEYL connexions and in these chapters pro-
blems of projective and conformal transformation and of imbedding
were also dealt with. In the light of modern developments this mixing
up of quite different topics could not be tolerated. So there is now a
chapter III dealing with linear connexions, curvature and normal Co-
ordinates, also with respect to anholonomic systems of reference and
in different notations. Then there is a chapter V on imbedding and
curvature, in the last section of which the theory of higher curvatures
VIII Preface to the second edition.

is developed. Chapter VI is dedicated to projective and conformal

transformations of connexions and in this chapter subprojective con-
nexions and concircular transformations of a vn find their natural place.
Between chapter III and these latter chapters it has been necessary
to insert a chapter IV on the theory of LIE groups. This theory not
only gives a very beautiful example of speciallinear connexions but it
also forms the necessary foundation for the following chapters, espe-
cially for chapter VII.
This chapter VII on the theory of variations, deformations and
movements leads in 4, 5 to the theory of the holonomy groups that
is used in 6 to give an application of CARTAN's method.
The last chapter in the first edition dealt with theinvariant decom-
position of tensors and in 1923 this was a new and interesting topic.
But now we can refer to D. E. LITTLEWoon's book of 1951 and it does
not seem any Ionger justifiable to treat such a special purely algebraic
problern in a book on Rrccr calculus and differential geometry.
Instead a chapter VIII has been added containing miscellaneous
examples, mostly taken from recent literature. Here the harmonic Vn
( 1), the many different connexions for hybrid quantities and their
properties with respect to imbedding and transformation ( 2-8) and
the spaces of recurrent curvature ( 9) are considered. The choice of
these subjects was entirely free. Of course this chapter could have
been made many times longer. But there must be a limit to the size
of a book of this kind and this is also the reason why all those topics
contained in the following long list, however interesting they are, could
at most be mentioned very shortly:
1. properties in the large as for instance properties of compact
manifolds;
2. theory of geometric objects; fibre bundles etc.;
3. theory of VrTALI;
4. geometries of FrNSLER, CARTAN, A. KAWAGUcHI;
5. extensors of CRAIG and generalizations of M. KAWAGUCHI;
6. linear elements and surface elements of higher order;
7. connexions of higher order;
8. projective and conformal geometry with supernumerary coor-
dinates either in the tangent spaces or in the whole space;
9. path spaces; K-spreads;
10. natural families of curves;
11. non-linear connexions;
12. special geometries for small values of the dimensions;
 Preface to the second edition. IX

13. contact transformations and their connexions;

14. treatment of differential equations by means of linear connexions
belonging to them;
15. spaces with a fundamental tensor of rank < n;
16. G. KRoN's applications of tensor analysis to electric circuits.
In the first edition there was an extensive literature list. With its
196 titles it covered nearly all the literature on many dimensional
differential geometry. The literature list of the Einfhrung of 193 5-1938
with its 488 titles represents only a selection of the literature at that
time. This is even more true of the literature list presented here which
is by no means exhaustive though it contains about 1400 titles and refers
to about 350 authors. By selecting the titles in such a way that the
reader interested in some topic will always find at least a few titles
that can lead him to more references, the author has tried to retain
something of the encyclopedic character of the first edition.
The author owes many thanks to the Mathematical Institute of
Amsterdam University and the Mathematical Centre at Amsterdam
for the valuable help of inany of their collaborators. In the first place
I mention Dr. NIJENHUIS (M. I. and M. C.) who has given much
time to the study of literature and the collection of exercises and who
proposed valuable improvements to the text in nearly every chapter.
Mr. BARNING (M. 1.) and Mrs. V. RoOTSELAAR (M. 1.) did important work
by studying literature and checking formulae. For the correction of
proofs I had the valuable help of Mr. BARNING (M. I.) and Mr. VER-
HOEFF (M. C.). Mr. Woon of the Clarendon Press, Oxford, has kindly
given advice on the language. This is the third book with which he
has helped me and I am very grateful to him for all the English idiom
he has taught me. Last but not least I have to thank my wife
Mrs. ScHOUTEN-BYLSMA who spent much time on the administration
of all literature references.
I wish to express my best thanks to the publishers for their most
agreeable collaboration.

Epe (Holland), im Juli 1954. J. A. SCHOUTEN.

 Contents.
I. Algebraic preliminaries (1).
1. The En (1).
affine space (1) - En (1) -affine group (1)- Ga (1)- minor (1)- summation
convention (1) - Af, A~. (1) - point space (1) - allowable coordinate trans-
formations (1) - rectilinear coordinates (1) - homogeneaus linear group (2) -
Gko (2) - centredEn (2)- point transformations (2)- * (2) - dragging along
of coordinate system (2) - !5): (2) - KRONECKER symbol (2, 7) - 5f (2) - gene-
ralized KRONECKERsymbol (2) - kemelletter (2) - running indices (2) - fixed
indices (2) - kemel-index method (3) - flat submanifold (3) -linear submani-
fold (3) - Cf (3) - rank (3) - B~ (3) - point (3) - straight Jine (3) - plane
(3)- hyperplane (3)- coordinate axes (4)- coordinate planes (4)- coordinate
Ep's (4) - net (4) - lie in (4) - contain (4) - translation (4) - parallel (4) -
p-direction (4) - improper Ep_ 1 (4) - Ep_ 1 "at infinity" (4) - direction
(4) - reduction (5) - tjp-parallel (4) - section (4, 5) - spanning (4) -
join (4) - projection (4, 5) - screwsense (5) -sense (5) -sense of rotation (5) -
opposite (5) - inner orientation (5) - outer orientation (5).
2. Quantities in En (6).
quantity (6) - components (6) - kind (6) - sum of quantities (5) - manner
of transformation (6) - scalar (6) - contravariant vector (6) - contragredient
(6, 7) - contravariant basis (7) - dead indices (7) - living indices (7) -
covariant vector (7) - covariant basis (8) - transvection (8) - reciprocal sets
(9) - strangling (10) - tensor (10) - affinor (10) - valence (10) - co- and
contravariant tensor (10) - cocontravariant (10) - mixed (10) - unity tensor
(10) - A~ (10) - intermediate components (11) - equiform (11) - pseudo-
tensor (11) - pseudo-scalar (11) - pseudo-vector (11) - class of pseudo-quan-
tity (11)- tensor density (12) -tensor .d-density (12) -weight (12)- W-tensor
(12) - w (12) - G8 a (12) - .connecting quantities (13).
3. Invariant processes and relations (13).
addition of quantities (13) -isomer (13) - general multiplication (13) - con-
traction (13) -transvection (14)- saturated indices (14)- dummy indices (14)-
free indices (14) - mixing (14) - round brackets (14) - symmetric (14) - sym-
metric part (14) - altemation (14) - square brackets (14) - altemating (14) -
altemating part (14) - concomitant (15) -invariant (15) -rational integral (15).
4. Section and reduction with respect to an En& in En (16).
decomposition with respect to a rigged E,., (16) - section (16) - reduction (1 7\ -
rigging (17)- projection (18)- E,..-part (19).

5. Rank, domain and support of domain with respect to one or more indices (20).
,u-rank (20) - ,u-domain (20) - support of domain (20) - double brackets [[)] (20).
6. Symmetrie tensors (21).
symmetric tensor (21) - symmetric multiplication (21) - divisor of a symmetric
tensor (21) - irreducible symmetric tensors (21).
 Contents. XI

7. Multivectors (22).
multivector (22) - p-vector (22) - bivector, trivector, quadrivector (23) -
altemating multiplication (23) - divisor of a multivector (23) - simple multi-
~"M {25)- <;~1 .111 {25) - cx: (25) - @:"""
vector (23) - n-vector {23) - (X}
{26) - e.1, .1,. {26) - identifications {27) - pseudo-scalar (27) - tables of
alternating quantities in E 3 (28).

s. Tensors of valence 2 {28).

Cocontravariant tensor of valence 2 (28) - trace (29) - spur (29) - eigenvector
{29) - eigenvalue (29) - characteristic equation {29) - first canonical form of
the matrix {33) - elementary divisors (33)- co- and contravariant tensors (34) -
s-vector tensor (34) - adjoint (34, 38) - symmetric co- and contravariant ten-
sors (34) - index {34) - signature (34) - rank tensors (34) - definite {35) -
indefinite {35) - semi-definite (35) - co- and contravariant bivectors {35)
blades {36) - classification of trivectors {36) - reduction number {37).

9. Introduction of a metric in an E,. (40).

length (40) - distance {40) - indicatrix (40) - fundamental tensor (41) -
euclidean R,. (41) - minkowskian R,. (41, 42)- time-like (41) - space-like (41) -
-region {41) - nullcone (41) - unitvector (41) -angle (41) - duration {41) -
isotropic Ep (42) - full isotropic Ep {42) -null Ep (42) - perpendicular {42)-
cartesian coordinate system {42)- minkowskian proper rotations {43)- LoRENTZ
transformations {43) - minkowskian reflexions {43) - improper minkowskian
rotations {44) - vector {44) - raising and lowering of indices (44) - I""n
w - w
{45) - i,1, ... .1,. {45) - i, i {45) - g {45) - principal multidirection {46) -
theorem of principal axes (46) - theorem of principal blades {46) -orthogonal
coordinate transformations (47ff.) - rotations (47) - reflexotations {47) -
principal angles (48) - reflexional orthogonal transformations {50) - proper
and improper rotations {50) - tfp-parallel (51) - tfp-perpendicular (51).

10. Hybrid quantities {51).

quantities of the first and second kind (51) - hybrid quantities (52) - weight
and antiweight {52) - hermitian (symmetric) tensor (52)- hermitian alternating
tensor (52) - hybrid bivector {52) - -invertible (53) - index {53) - signature
(53) - -definite {53) - indefinite {53) - semi-definite {53) - the auxiliary
E 2 ,. {53)- principal E,. (53)- first and second set of invariant E,.'s {54) -her-
mitian fundamental tensor {54)- R,. {54)- ordinary R,. {54)- kets and bras {54)
- unitary perpendicular vectors {54) - norm of vector (54) - unitvector {54) -
unitary cartesian coordinate systems (54) - unitary orthogonal group (55) -
fundamental figure in .R,. {55) - nullcone in R,. (55) - the auxiliary R 2 ,. {55) -
principal multidirections of a hermitian tensor (56) - theorem of principal axes
{56).
11. Abridged notations (57).
collecting indices {57) -representativeindices (57) -the notation of GIVENS (57)-
VITALI's method (57)- CRAIG's extensors {58)- idea:tvectors (58)- systems of
"direct" calculus {58)- skeleton of a formula {59)- the method of the radius-
vector in E,. and R,. {60) - CARTAN's method {61) - abbreviations ad hoc {61).
XII Contents.

II. Analytic preliminaries (61).

1. The arithmetic n-dimensional manifold 2ln (61).
arithmetic point (61) - arithmetic manifold W,. (61) - components of the arith-
metic point (61)- polycylinder (61) -box (61)- region of m,. (62) - neighbour-
hood (62).
2. The geometric n-dimensional manifold Xn {62).
coordinate system (62) - analytic functions (62) - dass of a function (62) -
coordinate transformation (63) - pseudo-group (64) - groupoid (64) - super-
numerary Coordinates (64) - allowable coordinate systems (64) - geometric
manifold (64) - Xn (64) - geometric point (64) - KLEIN's principle (65) -
cell (65) - region of Xn (65) - imbedding (66) - neighbourhood (66) - analy-
ticity in Xn (66)- geometry in the large or global geometry (66) -dass in Xn (66).

3. Geometrie objects and quantities in Xn (67).

geometric object at a point (67)- {} (67)- components of geometric object (68)-
kind of an object (68) - quantities (68) - connecting objects (68) - pseudo-
quantities (69) - pseudo-scalar (69) - dass of pseudo-quantities (69) - tangent
E,. (69) - local En (69) - object field (70) - analyticity of field (70) - dass of
field (70) - natural derivative (70) - gradient (70) - Grad (70) - D (70) -
convention about operators (70) - covariant and contravariant basis vectors in
Xn (70f.)- e", ~;. (71) -A~' (71)- 0~ (71) -region of constant rank (71)- contra-
;.
and covariant measuring vectors (71).

4. The X,.. in Xn (72).

functional matrix (72) -rank of a set of functions (72) - theorem of adaptation
(73) - rank of a system of equations (73) - null point of a system of equations
(73) - nullmanifold (73) -null form (73) - equivalent systems (73) - minimal
regular (73)- dimension of nullmanifold (73)- imbedding (74 f.)- rn-dimensional
surface (74) - curve, surface, hypersurface (74) - covariant connecting quantity
(74) - base (74) - first and second base theorem (74f.) - base transformation
(75) - parametric form or representation (75) -minimal regular (75)- contra-
variant connecting quantity (75) - coordinate Xm's (76) - normal system of
Xm's (76) - net of a coordinate system (76) - reduction (Zusammenlegung) of
Xn (76) - tangent Em (77) - section with Xm (77) - reduction with respect to
Xm (77) - rigged (77) - parts of a tensor (77) - prolongation of a field of Xm
over Xn (78).
5. The Ern-field or XU" in Xn. Systems of linear partial differential equations (78).
section with X;:' (78) - reduction with respect to X;:' (78) - rigged X;:' (79) -
parts of a tensor (79) - enveloping (79) - outer and inner problern (79) -
PFAFF's problern (79) - solution of a system of partial differential equations (79) -
nullpoint of a system (80) - totally integrable systems (80) - conditions of
integrability (80) - homogeneaus linear equations (80) - adjoint systems (80) -
integral function (80) -integral (80) - derived systems (81) - complete systems
(81) - Xm-forming Em-field (81) -non homogeneaus linear equations (82) -
reduced system (82) - complete non homogeneaus systems (82) - gradient
equation (83).
6. Theinvariantoperators Rot and Div (83).
alternated derivative (83) - natural derivative (83, 84) - rotation (83) - Rot
(83) - D (83) - divergence (84) - Div (84) - gradient product (84) - theorem
of rotation and divergence (85) - other conditions of integrability (86 ff.) - section
of a rotation (89).
 Contents. XIII

7. PFAFF'S problern {89).

class K {89} - rotation class 2(! (89} -similarity class k {89}- supports of rotation
(90) -Supports (90) - characteristics (90) - 51 , 5~, 5 4 , 5~ (90) - similarity
transformations (91) - gradient transformations (91) - table of results (92) -
canonical form (93) - w_. is x.-enveloping, v = n-e-e (94) - special canonical
forms {94).
8. Theorem of STOKES (95).
generalized theorem of STOKES (97) - differential forms (97 f.) - non invariant
forms (99).
9. Anholonomic Coordinates (99).
anholonomic coordinate system {h} (99) - net (100) - (100) - A1
h (100) -
(d~)h (100) - object of anholonomity Df> (100) - Xm in Xn with anholonomic
coordinates (100) - X;:' with anholonomic coordinates (101) - intermediate
components !J~ .. ofn;.(101).

10. The LIE derivative (102).

dragging along of a coordinate system (102) - dragging along of an object field
(102} - dragging along over v" dt (103)- operator e1X (104} - displacement
over TV" (104} -infinitesimal point transformation (104) - group generated by
an infinitesimal transformation (104) - LIE derivative, LIE differential (104) -
LIE derivative of sums, products, transvections, contractions (106} - LIE deri-
vative of scalars, vectors, tensors, p-vectors, densities {106)- .1-densities {107) -
geometric objects (107) - absolutely invariant fields ( 107} -absolute invariance
with respect to streamlines (107)- LIE derivative of II~A (108} - operator eU!.
working on linear geometric objects {108}- 5;~" and " (109)- commutativity
5;~
of 5; and D {110).
11. The LAGRANGE derivative (111).
LAGRANGE derivative (112)- LAGRANGE equation (112)- relations for LAGRANGE
derivative (114).
12. CARTAN's symbolical method (11 7).
alternating differential form (118) - symbolical method of CARTAN and KHLER
( 119) - symbolical method applied to geometric objects (119) - scheme of trans-
lation {120) - examples (120f.).

111. Linear connexions (121).

1. Paralleldisplacement in an En (121).
covariant differential and derivative (122) - parameters of parallel displacement
{122}- <5, v,..{122).
2. Paralleldisplacement in Xn (123}.
parallel displacement of quantities (123)- conditions for the covariant differential
(124} - r;.. (124) -linear displacement (125) -linear connexion (125) - geo-
metric interpretation (125}- covariant stationary (125ff.)- covariant constant
(125f.) - recurrent or parallel field {126)- s;.;," (126) - symmetric and semi-
symmetric connexion (126} - geometric interpretation (127} - displacement of
points (127) - CARTAN displacement (128f.) - covariant differentiation of
densities (129) - covariant differentiation of pseudo-quantities (130) -parallel
displaced field (130).
XIV Contents.

3. A linear connexion expressed in terms of s;;.~<, an auxiliary symmetric tensor

field Y;." of rank n and fj. Y;." (131).
CHRISTOFFEL symbol (132)- {;;.} (132) - {.u A x} (132)- metric connexion (132) -
u" (133)- fundamental tensor (133)- v" (133) - riemannian geometry (133)-
ordinary Un and V" (133) - semi-metric connexion (133) - W" (133) - WEYL's
geometry (133) -conformal transformation of g;." (134)- "Umeichung" (134)-
Wn dealt with by means of pseudo-tensor (135)- necessity of fundamental tensor
of valence 2 (135ff.)- rings of quantities (137).

4. Curvature ( 138).
R;;." (138) - RIEMANN-CHRISTOFFEL tensor (139) - curvature tensor (139)-
integrable connexion (139)- V[v V,.1 and the rule of LEIBNIZ (139f.)- application
to densities and pseudo-quantities (140f.)- V"p (140)- Rpl. Kp;. Ricci tensor
(141)- teleparallelism or absolute parallelism (142)- R;;," and integrability
conditions (142)- the ideal factors a;., b;. (143)- special semi-symmetric and
metric connexions (143) - volume preserving connexions (144).

5. The identities for the curvature tensor (144).

first and second identity (144) - third and fourth identity (145) - principal
directions of vn (146) - BIANCHI's identity (146) - K (148) - G";. (148) -
scalarcurvaturex (148) -Gaussian curvature (148)- RICCI space, special (148)-
EINSTEIN space, special (148) - theorem of HERGLOTZ (148) - space of constant
p p+l p
curvature (148) - S" (148) - V[!' ... ] (149) -- J7'r, V,. ... ] and the rule of LEIBNIZ
p
(149)- geometric interpretation of V[,u ... ] (150)- field in equilibrium (150) -
geometric interpretation by means of CARTAN displacements (151f.)- LIE deri-
vative of ~";. (152) -t t f(
expressed in terms of V instead of a(152)- V'"- V,. 152)-
riemannian curvature with respect to a 2-direction (153) - theorem of ScHUR
(153).
6. Integrability conditions in Ln (153).
integrability conditions with V (154) - possibility of covariant constant n-vector
fields (155).
7. Geodesics and normal coordinates (155).
geodesie (1 55) - affine parameter (15 5) - symmetric connexion determined by
geodesics and their affine parameters (156) - coordinate system geodesie at a
point (157)- ~; ... p,;. (157) -normal Coordinates in An (158)- normal tensors
(159f.) - extension (160)- relations between the R's and the N's (162)- A"
symmetric with respect to a point (163)- R;;" covariant constant in symmetric
An (163) - differential concomitants (164) - absolute and relative differential
invariants (164) - first and second reduction theorem (164).

8. FERMI Coordinates (166).

FERMI Coordinates (169) - generalization for Xm (169).

9. Linear connexions expressed in anholonomic coordinates (169) .

.lf1(169) - If (169) - Qiik (170) - Xjki (170) - If;h (170) - 'Yiki (171)
coefficients of rotation (171)- Ri,j;h (172)- Vkj (172).
 Contents. XV

10. CARTAN's symbolical method used for connexions (172).

A h,[Jh, I;", S". Rl h (172) -
structural equations of L,. (173) - tMoreme de con-
p
servation de la courhure et de la torsion (173)- [17 ... ] (174)- mapping of two
sets of n vectorfields on each other (174f.) -mapping of two L,.'s on each other
(176ff.).
11. Linear connexions depending on a non-symmetric fundamental tensor (179).
-1
non-symmetric fundamental tensor (179) - sh, S", s><, 5, "". h;.". k;." (180)-
the connexion r;:;. (180) - @~~" (180) - the ( - - + )-differentiation (180) -
@)[!'] and its divergence (181) - *F;:;. (181 ff.) - the condition for *5;. (181) -
the four necessary relations derived by means of the LIE derivative (183) - the
connexion *F;:;. in special cases (184) - first geometrical interpretation of the
condition (184) - EINSTEIN's interpretation (184) - discussion of results (184f.).

IV. LIE groups and linear connexions (185).

1. Finite continuous groups ( 185).
r-parameter finite continuous group (185) - unity element (185) - group germ
(186) - group space (186) - ()-equipollent (186) - ()-constant (187) -

Fra (187)- ()-connexion (187)- ()-parallel (187)- RJnJa= 0 (187)- an-
holonomic coordinate systems (a) and (A) ( 187 ff.) - geodesie (188) - I'ya (188) -
+-
(0)-connexion (188)- 17, 17, 17 (189)- Sj,f/ (189)- Cyi/ c~b C~B (190)- c~b =
const., c~B= const. (190)- RJ}.a (191)- 17, RJj,a= 0 (191).

2. The parameter-groups and the adjoint group of a finite continuous group (191).
continuous group (191) - parameter-groups (191f.) - special anholonomic Co-
ordinate systems in general X, (195ff.) - application to group space (196) -
structural formulae of the parameter-groups after LIE and after MAURER-CARTAN
(196) - structural constants (196) - t t (
197) - finite transformations of the
parameter-groups (197) - adjoint group (198) - centre (198) - homologaus
elements and subgroups (198) - invariant or normal subgroup (198) - linear
adjoint group (199) -normal coordinates r/l (199) - series for A'i! (200) -linear
transformations of the 'Y)a (200).
3. Finite continuous transformation groups (201).
essential parameters (201) - connected and dependent infinitesimal transforma-
tions (203) - theorem I. 1 (203) - structural formulae of a transformation group
after LIE and their generalization by means of the LrE derivative (205) - struc-
tural constants (205) - theorems II. 1, III. 1, I. 2 (206) - theorem II. 2 (207) -
generators of a group germ (208) - theorem III. 2 (208).
4. The geometry of group space (209).
relations between general Coordinates and the normal coordinates 'Y)a (209ff.)
+
series for A ~, AC, At and A ~ (211) - series for A '1, rcab, Fcab and c; (, a (212f.)
reflexions at the point 'Y)a (213).
0

5. Invariantsofatransformation group in the Xn of the !;" (214).

invariant functions (214) - invariant subspace (215) - relative invariant sets
of functions (215) - smallest invariant subspace through a point (215) - group
induced in invariant subspace (216) -transitive (216) - simply transitive (217 f.) -
invariant contravariant vector fields (217ff.)- primitive (218).
XVI Contents.

6. Invariantsofa group in group space (220).

gb, gba gcba (220) - Xp's representing subgroups (221 ff.) - the group of g 6 ,
the eentre (224) - the derived groups (224f.) - the group of gba (225) - the
group of c;i,d gda (225).
7. Properties of integrable groups {225).
integrable groups (225) - neeessary and suffieient eonditions (225) - eanonieal
form for an integrable group (226).

8. Simple and semi-simple groups {227).

the group spaee of a semi-simple group is a V,. (227).

V. lmbedding and Curvature (227).

1. The XI in l'n (227).
linear element (227) - first eurvature (228) - eurvatura geodetiea (228) - eur-
vature veetor (228) - first normal (228) - oseulating R 2 (228) - stationary point
(228) - seeond eurvature (229) - seeond normal (229) - oseulating R 3 (229) -
stationary oseulating R 2 (229) - higher eurvatures (229)- oseulating Rm (229) -
FRENET formulae (229 f.) -natural equations (230) - invarianee of the proportians
of the eurvatures (230) - eontact of order u of a eurve and a hypersurfaee (231).

2. The XI in Wn and Ln (231).

X 1 in Wn (231) - X 1 in L,. (231) - oseulating p-direetion (232) - affine para-
meter (232) - affine length (233).
3. The Xn-I in An (235).
Xn-I in A,. (235) - rigged X,._ 1 (235) - pseudo-normal veetor (235) - tangent
veetor (235) - first normalizing eondition (235) - indueed eonnexion (236) -
heb (236) - prineipal tangents (237) - asymptotie lines (237) - direetions of
prineipal eurvature (237) - equation of GAuss for A,._ 1 in A,. (237) - seeond
eondition of normalization (237) - equations of CoDAZZI for A,._ 1 in An (238) -
different eases of normalization in a volume preserving A,. (238ff.) - integrability
eonditions for an An-! in E,. (240).
4. The Vn-I in Vn (242).
V,._ 1 in Vn (242) - equations of GAUSS and CODAZZI for Vn_ 1 in Vn and R,. (242)-
seeond fundamental tensor (243)- integrability eonditions for Vn_ 1 in Rn (243)-
bending (244).
5. Congruences in Vn {244).
real eongruenees (244) - eanonieal eongruenees (245) - normal eongruenees
(245) - geodesie eongruenees (245) - orthogonal net (246)- orthogonal system
(246) - eongruenees belanging to an orthogonal system (246ff.) - eonditions
for the prineipal direetions of a symmetrie tensor to be vn-1-normal (247f.) -
eonditions for the eigenvectors of a eo-eontravariant tensor to be X,._ 1-forming
(248f.).
6. Properties of curvature of a Vn-I in V,. (250).
absolute, relative and enforeed eurvatures and eurvature veetors (250) - theorem
of MEUSNIER (250) - direetions of prineipal eurvature (250) - prineipal eurvatures
(250) - mean eurvature (250) - theorem of DUPIN (250) - prineipal tangents
(250) - asymptotie lines (250) - geodesie vn-1 in vn (251) - umbilieal point
(251) - umbilieal vn-1 (251) - SCHUR spaee (252) - v,,_l in Sn (252) - s,._l
in S,. (252) - indieatrix of DUPIN {252) - theorem of EuLER (253).
 Contents. XVII

7. The rigged X::' in Ln and Xn (253).

X;:' rigged in Ln (253ff.) - D-symbolism, Dw Dc, Dz (254) - full sets of com-
ponents (255) - skeleton of a formula (255) - first and second curvature tensors
m' m m' m' m m'
HJi/', Lc'~y H;;,a, L;':b (256) - L;:' (A;:') in Ln (An) (257) - za,x, z;;,a (257)-
m m' m m' m
'Q~b Q~Y' "Q:Y' Q:y (258) - rJJi/, r,;;;,x (260) -- ~<' ~z (260f.) - RJ;i/,
m' mm' m'm
R~;;,x (261) - equation of GAuss for L;:' in Ln (261) - RJ;;/, R~;i,a (261 f.)-
equations of RICCI for L;:' in Ln (261 f.) - equations of CoDAZZI for L;:' in Ln
(262f.) - geodesie L;:' (263) - L;:' (A;:') rigged in Xn (263f.) - generalization of
1lc and 1lz (264).
8. The rigged x ... in A,. (265).
X;:' rigged in An (265) - equations of GAuss, C:oDAZZI and RICCI for Am in An
(266f.)- equations for Am in En (267) - supposition of ScHLFLI (268) - class of
a vm (268).
9. The v::-
and Von in V,. and X,. (269).
vnm in Vn (269)- V,!" in Xn (270)- Vm in Vn (2it)- absolute, relative and enforced
curvature vectors (271) - principal tangent (271) - asymptotic lines (271) -
first curvature region (271) - curvature figure (Krmmungsgebilde) (271) -
axial, planar and spatial points (272) - umbilical point (272) - umbilical Vm (272)
- mean curvature vector (272) - minimal V.n (272) - transposition of a problern
of variation (273).
10. Higher curvatures of a v ... in Vn (275).
p-th curvature region (27 5) - asymptotic Vm of order p (27 5) - asymptotic lines
of order p on Vm (27 5) - higher curvature tensors of valence 3 (276) - generalization
of the FRENET formulae for Vm in Vn (277) - equations of GAUSS, CODAZZI and
RICCI involving higher curvatures (278ff.) - imbedding theorems (280ff.) -
Wm in Wn and Am in An (285).

11. Product spaces (285).

product spaces (285)- decomposable spaces (285)- breakable and decomposable
objects (285) - covariant constant symmetric tensor fields (286).

VI. Projective and conformal transformations of connexions (287).

1. Projective transformation of a symmetric connexion (287).
vector P;. of projective transformation (287) - preservation of parallelism (287) -
restricted projective transformation (288) - volume preserving connexion (288) -
projectively enclidean connexion (288ff.) - Dn (288) - projective curvature
tensor (289) - necessary projective transformation of an En into an En (290).

2. Projective transformation of the connexion in a Vn (292).

every S" a Dn (292) - projective transformation of a V" into a Vn (292)- equations
of LAME (296).

3. lmbedded spaces in An under projective transformations of the connexion (296).

induced connexion transformed projectively (297)- X"_ 1 in An (297f.)- curve
in A" (298) -quasi-plane curve (298) -affine parameter (299) -affine normal
(299) -affine conic (299).
Schonten, RicciwCalculus, 2. Aufl. Il
XVIII Contents.

4. Projective connexions (300).

projective parameters of T. Y. THOMAS (301 f.) of WHITEHEAD and of BORTO-
LOTTI (302)- connexion of CARTAN (302)- connexion of KNIG (303)- geometry
of paths (303) - homogeneaus coordinates of VAN DANTZIG (303)- K-spreads of
DouGLAS (303).
5. Conformal transformation of a connexion in Vn (304).
geodesie and conformal properties fix a vn (304) - conformally euclidean vn
(305) - C,. (305) - every V2 an 5 2 (306) - conformal curvature tensor (306f.) -
orthogonal systems in Cn (307) - umbilical V,._ 1's (308f.)- conformally geodesie
Vm 's (309ff.) - theorem of THOMSON and TAIT (310) - null geodesie invariant
(310)- umbilical pointsinvariant (310)- theorem of LIOUVILLE and LIE (312).

6. Conformal transformations of the connexion in an EINSTEIN space (312).

integrability conditions (313) - conformally invariant tensor density [";. (313) -
theorem of BRINKMAN (314).
7. Conformal connexions (315).
conformal parameters of J. M. THOMAS (315) - connexions of CARTAN, KNIG,
T. Y. THOMAS and of SCHOUTEN-HAANJTES (316).
8. Subprojective connexions (317).
k-fold projective An (317) - subprojective A,. (317) - pole direction (317) -
theorem of RACHEVSKI (317) - every subprojective connexion projectively in
variant (319) - auxiliary algebraic theorem (320) - subprojective D,. (321).
9. ADATI'S problern (321).
torse forming fields (322) - concircular, special concircular, concurrent, recurrent,
parallel-and covariant constant fields (322f.) - chief theorem of the subprojective
connexions (324) - special solutions (325f.).
10. Subprojective transformations of a connexion in An (327).
subprojective transformation of a connexion belanging to a direction field (327) -
subgeodesics belanging to a direction field (327).
11. The subprojective Vn (328).
every subprojective Vn, n > 2, a special C,. (329) special cases (329).
12. Concircular transformations of a Vn (330).
geodesie circles (330) - concircular transformations (330ff.) - relations with the
principal directions of Vn (321 f.) - umbilical Vm's (321) - necessary and sufficient
conditions for a Vn to admit a concircular transformation (322f.) - concircular
curvature tensor (334) - concircular euclidean V,. (334).

VII. Variationsand deformations (335).

1. Generaldeformationpro blems (335).
natural value of a field (335) - value after dragging along (335) - value after
pseudoparallel displacement (33 5) - operators invariant with respect to certain
objects (336)- covariant differential (336) - Lm differential ~ dt (336) - appa-
n
rent differential fl dt (336) - identities (337) -natural variation D dt (335) -
a g
absolute variation D dt (338) - geodesie variation D dt (338) - relations between
these operators (338) - a deformation problern (339) - product integrals (340ff.) -
 Contents. XIX
n a
transposition of a variation problern (341) - 'D = D (341) - BERTRAND curves
and their generalizations (342ff.).
2. Groups of "motions" in Vn and Ln (346).
affine, conformal, homothetic and projective motions (346) - details on affine
motions (346f.) - number of parameters of groups of motions in V,, and An (348) -
KILLING's equation (348) - translation (349) - translations in Vn (349) - groups
of motions in Vn and projective mapping (349) - number of parameters of the
group of all motions in a vn (350).
3. Deformation of subspaces (352).
details Oll Vm in Vn under infinitesimal transformations (352ff.) - deformation
of vnm in Vn, Lm in Ln and L;:' in Ln (352ff.)- method of transpositionimpossible
for Xm in Xn (354) - method of NIJENHUIS for the rigged X;:' in Xn (354ff.) -
the rigged L;:' in Xn (358ff.).
4. The holonomy group of an Ln (361).
the holonomy group (361) -non homogeneaus holonomy group (361) - the in-
finitesimal transformations of the holonomy group deduced by means of product
integrals (362ff.) - generators of the holonomy group (363 f.) - introduction of
an auxiliary tangent E, (364) - extension of covariant differentiation (364ff.) -
covariant constant fields and the holonomy groups (367) - special results for L 11
= Vn (367 f.) - the V" with a recurrent curvature tensor (368) - product spaces
(370).
5. Affinemotionsand the holonomy group in a symmetric An (370).
the holonomy group in a symmetric An (370) - equations for the affine motions
(371) - extension of the LIE operator (372) - group of isotropy (373) - the
group of affine motions is transitive (374) - number of parameters (374).
6. CARTAN'S method applied to the holonomy group and the symmetric An (375).
introduction of local coordinate systems (3 7 5) - the allowable anholonomic co-
ordinate system in a region (37 5) - the infinitesimal transformations of the holo-
nomy group (376) - the case of the symmetric An (377) - the case of the general
An (377ff.) - group of point transformations in a symmetric An leaving the con-
nexion invariant (381).

VIII. Miscellaneous Examples (381).

1. The harmonic Vn (381).
Vn harmonic at a point (381) - characteristic function (382) - completely har-
monic (382) - Centrally harmonic (382) - RusE s invariant (382) - simply
harmonic (383) - equation of LICHNEROWICZ (384) - equations of CoPSON and
RusE (385) - condition for a completely harmonic Vn to be an EINSTEIN space,
an Sn or an Rn (385 ff.) - ScHUR space and centrally harmonic space (385) -
a completely harmonic Vn that is not an S 11 (386) - the inequality of LICHNERO-
WICZ (388) - imbedding of an EINSTEIN Vn and a completely harmonic V" in
Sn+l (388).
2. Connexions for hybrid quantities (388ff.).
analytic and semi-analytic fields (389) - the new conditions for the connexion of
hybrid quantities (389) - most general form of connexion (389) - the auxiliary
X 2 n (390) -the two invariant sets of oon Xn's in X 2 n (390) -equipollent figures
(390) - the principal Xn in X 2 n (390) - most general connexion in X 2 n (391) -
first invariant conditions: the tangent n-directions are parallel (391) - second
invariant condition: the principal Xn is geodesie (391) - third invariant condi-
tion about equipollence and parallelism (392)- the Ln (393) - the auxiliary X 211
XX Contents.

is an L 2 n (393) - curvature tensors in Ln (393 ff.) - first and second identity

(393) - identity of BIANCHI (393) - V~ll and Vill (394) - identities found by
contraction (394f.) - the An (394) - necessary and sufficient conditions for a
semi-analytic scalar field tobe analytic (395) - necessary and sufficient conditions
for a semi-analytic transformation of Xn to be analytic (395).
3. Unitary connexions (395ff.).
hermitian tensor field a;.! (395) - unitary connexi?n (396) - the un
(396) -
fundamental tensor of a Un (396) - connexion in a Un derived from -a connexion
in an X 2 n with a symmetric fundamental tensor (397f.) -the auxiliary Xu is a
u2n (397) - the v,; and its auxiliary v2n (397) - the a;.;c in a vn
derived from a
scalar (397f.) - KXHLER space (397) - the curvature tensors of a Un (398ff.) -
Vvp vanishes and Vjll is hermetian (398) - the four identities in Un (398f.) -
BIANCHI's identity in n (399) - S;. a gradient in a semi-symmetric fln (399) -
R;;;. hermitian in a V;. (399) - the identities in vn
(399) - list of formulae in Ln,
An, fln and V" (400f.) - mutually perpendicular analytic fields of unitvectors in
n and the integrability and analyticity of the connexion (402f.) - a V,. with an
analytic connexion is a Rn (403).
_ 4. The Vn of _constant curvature (404ff.). _
Vn of co~stant curvature (404) - Sn (405) - projective tran_:;formation of an Sn
into an Rn (405)- the linear element of FuBINI and STUDY in Sn (406)- hermitian
non euclidean geometry (406).
5. Imbedding in an Ln (407ff.).
a rigged Xm in Ln (407) - B~, q (407) - the X", is an Lm (408) - induced con-
nexion (408) - a rigged XminAnis an Am (408) -H;;,"; Li~;.; z;~ l (408) - geodesie
Lm in Ln (409) - two GAuss equations (410) - five CoDAZZI equations (410) -
two RICCI equations (410) - the GAUSS, CODAZZI and RICCI equations in Un
(411 f.) - imbedding in V"
(412).
6. Curves in a Dn with a positive definite fundamental tensor (412ff.).
the U1 in n (412) - the dz of CoBURN (412) - the R1 in n (414) - curves with
a real parameter (415).
7. Conformal transformation of a connexion in Dn (413ff.).
a conformal transformation is also restricted projective transformation (415) -
a vn determined by its conformal properties only (416) - conformal euclidean
Vn (416) - Cn (416) - conformal curvature tensor (417) - Ci/.i" (417) - con-
formally symmetric fln (417).
8. Conformal unitary connexions (418ff.).
c c c c
21;.,. (418) - conformal connexion II~;. (419) - J.7 (419) - R;;;;," (419)- R;;'P,i"
(419).
9. Spaces of recurrent curvature (421ff.).
recurrent curvature tensor (421) - Kn, K! (421) - kw always a gradient (421) -
the decomposable K! (421 f.) - flat extension (422) - null-extension (422) -
the three cases for the simple K! (423) - classification of simple K! 's (423 f.) -
linear elements of WALKER (424).
Bibliography . 425
Index . . . . . . . . . . . . . 512
 I. Algehraic preliminaries.
1. The E,..
In this chapter we consider the "ordinary" n-dimensional affine
space En, i.e. the space in which the Coordinates of a point are subject
to transformations of the affine group Ga .1) If the ~"; u = 1, ... , n 2}
are the "old" Coordinates of a point, we can get "new" coordinates
~"; u' = 1', ... , n' by means of the coordinate transformation

(1.1}
where the A~ and a"' are constants. The coordinate systems are usu-
ally denoted by (u) and (u') respectively. From (1.1) it follows that there
exists a transformation of ~ into t', called the inverse transformation
(1.2}
with constant coefficients. For every value of (! and a' the coefficient A~.
equals the minor 4) of the element A~ in the matrix of the A~, divided
by ,1 and a"' and a" are related by the equation
(1.3) a"' = - A~' a".

No mention has been made so far of the meaning of the words "point"
and "space". We leave this to Ch. II where the En will be defined as
a special case of more general manifolds. For the moment it is suffi-
cient to know that each point has certain coordinates, i.e. a set of n
real or complex numbers assigned to it and that the linear transforma-
tions (1.1, 2) play an important role. The transformations (1.1, 2) will
be called allowable coordinate transformations and the coordinates obtain-
able by them rectilinear coordinates.
1) A set of transformations is said to form a group if 1 o any two transformations
of the set performed after each other yield a transformation belanging to the set,
2 the inverse of every transformation of the set belongs to the set, 3 the set
contains the identical transformation.
2 ) Unless otherwise specified the indices )!, , p, v, (!, a, -r: and sometimes w
always take the values 1, ... , n (in italics); )! 1 , ', p', v', e', a', -r:' and sometimes w'
take the values 1', ... , n' (in italics) etc.
3 ) We adopt the summation convention: if an index appears twice in the same
term, once as a subscript and once as a superscript, the sign 1: will be omitted.
4 ) The minor of At is the determinant remairring after dropping the row and
the column containing At in the matrix of A~ and multiplying by (-1)Ha.
Schonten, Ricci-Calculus, 2. Auf!. 1
2 I. Algebraic preliminaries.

If in (1. 1) we take a"' = 0 we get the transformations of the homo-

geneous linear group Gh 0 The En is then called a centred E... In a centred
En the origin has the Coordinates zero with respect to all allowable
coordinate systems.
There are also other transformations called point transformations,
for instance the linear transformation
(1.4)
with constant coefficients P~, and p>'. Here the point !;" is transformed
into the point rj" but the Coordinates are not changed.
To every coordinate transformation (x) _". (x') there belongs a point
transformation such that the old coordinates of every point are numer-
ically equal to the new coordinates of the trapsformed point. We may
express this by the equation
(1.5)
valid for the coordinate systems (x) and (x') only. The sign ~- (speak:
star equals) will always be used if we wish to emphasize the fact that
an equation is only valid or that its validity is only asserted for the
coordinate system or coordinate systems occuring explicitly in the
formula itself. From (1.4, 5) we get
(1.6)

valid for every choice of !;", and this is only possible if

(1.7) b) P"*--A~.a.-.
If a point transformation belongs to the coordinate transformation
(x) _". (x') we say that (x) is dragged along by this point transformation.
Hence the numerical values of the coordinates of a point remain in-
variant if any point transformation is applied and if at the same time
the coordinate system is dragged along by the same point transformation.
In all formulae there are kernelletters like A, !;, P 2), running indices
like x, x' and fixed indices like 1, ... , n; 1', ... , n'. With coordinate
transformations the kernel letters do not change but we get a new set of
running indices and fixed indices. But with point transformations the
kernelletters change and the running and fixed indices remain the same.
1) The symbol t5l used hereafter is the KRONECKER symbol. Its value is+ 1 if
"= A (not to summarize) and zero if "=!= A. It occurs also in the forms 15"' and
+
<5"'. The symbol <51' is the generalized KRONECKER symbol. Its value is 1 if"
and A' take corresponding values from 1, ... , n and 1', ... , n' and zero if they
do not.
2 ) A kernel can also be symbolized by a more complicated symbol, cf. for

instance (VII 1, 24).

 1. The En. 3

This is the principle of the kernel-index method used in this book andin
many other publications .1)
A kernelleHer can be changed in the following ways:
1. By changing the letter itself, e.g. ~ -> 'YJ;
2. by adjoining an accent or another sign, e.g. * preferably to the
left in order to avoid collision with upper running or fixed indices to
the right;
3. by adjoining an index over or under the kernel 2). The upper and
lower places to the right are reserved for running or fixed indices;
4. by putting upper or lower indices, without moving them, bet-
ween brackets as will be shown later.
A flat or linear m-dimensional submanifold or Emin En can be given
by a set of n- m linear equations
(1.8) q~A + C" = 0; x= m + 1, ... , n 3)

with the conditions that the Ci and the C" are constants and that
the matrix of the Cf has the rank 4) n-m. But it can also be given
by a set of m parametric equations
(1.9) ~"= B~r/ + B"; b= 1, ... ,m
with m parameters r/ and the condition that the B~ and the B" are
constants and that the matrix of B~ has rank m. From (1.9) we see
that the 'YJb can be used as rectilinear Coordinates in the submanifold
and that accordingly this submanifold is really an Em, because every
other set of parameters that could be used in (1.9) must be linearly
connected with the 'YJb From (1.8, 9) it follows that
( 1.1 0) crm=o; b= 1, ... ,m; x=m+1, ... ,n
and that
(1.11) C" =-GiB'-.
An E 0 is a point, an E 1 a straight line, an E 2 a plane and an En-l
a hyperplane. To every rectilinear coordinate system there belang n E 1 's
1 ) The kernel-index method was sketched for the first time in the authors
lectures on continuous groups at Leiden University 1926/27 (opalographic
syllabus 1927, 1). The first printed exposition of the method appeared in ScHOUTEN
and V AN KAMPEN 1930, 2 and GOLAB (thesis) 1930, 1.
2) If possible it is advisable to avoid using upper and lower indices of this kind
simultaneously.
3 ) In reasonings like this we always suppose that m and m and also n and n
stand for the same integer written in italics and vertical.
4 ) A matrix has rank p if it contains at least one non zero subdeterminant
with p rows but none with more rows.
1*
4 I. Algebraic preliminaries.

the coordinate axes, (;) coordinate planes and (;) coordinate Ep's,
p = 1, ... , n -1. These flat manifolds form the net of the coordinate
system. If a cordinate system is dragged along by a point transforma-
tion, this means that the point transformation is applied to its net.
Ep is said to lie in Eq and Eq is said to contain Ep if every point of Ep
is a point of Eq. In order to discuss the common points of an Ep and
an Eq we first define parallelism in En. A translation is a point trans-
formation of the form
(1.12) c" = constant
with respect to some rectilinear coordinate system and thus with
respect to all these systems. Two Ep's are said to be parallel if they
can be transformed into each other by a translation. The set of all Ep's
parallel to a given Ep is called a p-direction or an improper Ep_ 1 . 1 )
In a centred En there is a one to one correspondence between all im-
proper Ep_ 1's and all Ep's containing the origin. A 1-direction is called
a direction for short. 2) An Ep is said to contain an improper Eq and the Eq
is said to lie in the Ep if the Ep contains one of the Eq+l's of the Eq.
The equation
(1.13) x = m+ 1, ... ,n
with variable ''" represents all Em's parallel to the Em fixed by C1
through the origin. In the manifold of these Em's the C'" can be used
as rectilinear coordirrates. Accordingly this manifold is an En-m each
point of which represents an Em. We call this En-m the reduction
of En with respect to the Em through the origin or with respect to its
m-direction.
The following statements are valid both for proper and for improper
flat submanifolds.
1. An Ep and an Eq; 0 < p;:;;. q, are called tjp-parallel, t > 0, and
for t = p parallel if they contain the same t-direction but no common
(t + 1)-direction:
2. The section 3) of an Ep and an Eq consists of all common E 0 's.
This section is an E.; p :2:;s :2:P+q-n, q ;;;;;s.
J. An Ep and an Eq span an E, if they are both contained in E,
and if there is no E,_ 1 containing both Ep and Eq. E, is called the
join) of Ep and Eq. r and s are related by the equation r=P+q-s.
1) Formerly often called an Ep_ 1 "at infinity".
2) A direction in every day langnage is according to our definition a direction
with a sense (arrow).
3 ) Some authors use the terms intersection and meet.

') In P. P. 1949, 1 we used the term junction.

 1. The En.

If in En a proper EP and a q-direction, q = n- p, which have no

direction in common, are given, then every geometric figure in En can
be subjected to the following processes:
1. Section with Ep: All points of the figure not lying in Ep are
dropped. Only the Ep is used.
2. Reduction with respect to the q-direction: All points of the figure
lying in an Eq with this q-direction are identified. The resulting figure
lies in the Ep that arises from reduction of the En with respect to the
q-direction. Only the q-direction is used.
3. Projection on Ep in the q-direction: Through every point of the
figure an Eq is laid with the given q-direction. The section of this Eq
with Ep, is the projection of this point. Both the Ep and the q-direction
are used.
An n-dimensional screwsense in En is determined by n directions,
with a sense (arrow), which are not contained in the same (n -1)-
direction and are given in a definite order. For n = 2 the screwsense
is often called sense of rotation and for n = 1 sense. Two screwsenses
are equal if the defining figures can be transformed into each other by
a point transformation (1.4) with Det (P~;..) > 0 and opposite if they
can not. Hence there exist only two screwsenses in an En. The screw-
sense can also be given by a part of a general curve, not lying in an
En_ 1 , which is provided with a sense (arrow). If -~ n(n+1) is even,
e.g. for n = 3, the sense of the curve can be omitted. In this case it
makes no difference if the senses of all directions are changed and their
order is inverted at the same time. A screwsense is invariant for all
linear point transformations (1.4) with Det (P~;..) > 0. To every co-
ordinate system (u) there belongs a definite screwsense fixed by the
directions of the axes, their +-senses and the order 1, ... , n. This
screwsense is invariant for rechlinear coordinate transformations if
and only if LI > 0.
An Ep with a p-dimensional screwsense fixes a screwsense in every
Ep parallel to it. Hence a p-direction may have a screwsense. An Ep
or a p-direction with screwsense are said to be oriented with inner
orientation. If to a given Ep or p-direction an (n- p)-dimensional
screwsense is given in some En-p which has no direction in common
with Ep or the p-direction, this screwsense fixes a screwsense in every
En-p with this property. Then the Ep or the p-direction are said to
be oriented with outer orientation. For instance an E 1 in E 3 has inner
orientation if there is an arrow in it and it has onter orientation if
there is an arrow around it. 1 )
1) The idea of orientation has been dealt with from a very general and highly
interesting point of view by E. CARTAN 1941, 1.
6 I. Algebraic preliminaries.

Exercises.
I 1,1. Prove that (cf. Exerc. II 2,1)

I 1,1 IX)

I 1,2. An Ep in En can be fixed by (p + 1) (n- p) numbers. An Ep

through a given Eq, q < p, can be fixed by (p-q) (n-p) numbers.
I 1,3. If an Ep and an Eq, having no direction in common, are
given in a definite order, and if in each of them a screwsense is given,
a screwsense in the join EP+q is determined. If the order of Ep and Eq
is changed the screwsense in EP+q changes if and only if pq is odd.

2. Quantities in En.
A quantity in En is a correspondence between the rectilinear co-
ordinate systems and the ordered sets of N numbcrs, satisfying the
conditions:
1. To every coordinate system (x) there corresponds one and only
one set of N numbers;
2. if <PA; A =1, ... , N corresponds to (x) and <PA'; A' = 1', ... , N'
to (x') the <PA' are functions of the <PA and the A:', homogeneous linear
in <PA and homogeneous algebraic in A~'. The <PAare called the compo-
nents of the quantity with respect to (x) . 1) Quantities with the same
number of components and the same manner of transformation are
said to be of the same kind. According to the definition, by addition
of corresponding components of two quantities of the same kind we
get components of another quantity of the same kind called the sum
of both.
Quantities are distinguished by the manner of transformation of
their components. We give here a list of some of the most important
quantities in En. Other quantities will be defined later.
1. A scalar with one component, invariant for the transformations
(1.1).
2. A contravariant vector 2) with n components v" and the transforma-
tion
(2.1)
1 ) It is often convenient to denote a quantity in the text by its kerne! Ietter
only. This must not of course be done if the same kerne! Ietter is being used for
two or more different quantities.
2 ) In the classical theory of invariants two transformations of the kind (2.1)

and (2.4) were called contragredient with respect to each other. Then the two
transformations were distinguished by the words covariant and contravariant.
But the reasons for calling (2.1) contravariant and (2.4) covariant (and not the
other way round) are purely historical.
 2. Quantities in En. 7

A contravariant vector v" is represented by two points ;" and ;" + v"
in definite order fixed to within a common translation. The order
can be represented by numbering the points or by a curve or straight
line with an arrow joining the points. In a centredEn a contravariant
vector v" can be represented by one point with the Coordinates v",
the origin playing the role of first point. Addition of two contravariant
vectors can be illustrated by the wellknown figure called "parallelo-
gram of forces" in mechanics. Iv1 is the projection of v" on the 1-axis
1

in the (n -1)-direction of the other axes measured by the unit in the

1-axis. v1 is positive if the projected sense of v" on the 1-axis is the
same as the +-sense of the 1-axis and negative otherwise.
To every coordinate system (u) there belong n contravariant vectors

"_!_ .ll" _ { 1 for u = A.

(2.2) e -u,;-
A 0 for u =FA.
called the contravariant basis vectors of (u). 1 ) The components of these
vectors with respect to another coordinate system (u') in general no
Ionger have the valu~s 1 or 0 because
(2.3)

and to (u') there belongs another contravariant basis e". Here we have
A'
an example of kernels with indices under the letter. Such indices are
not subject to the transformation of coordinates and are called dead
indices in contradisti;o.ction to the indices to the right that are called
living. 2)
3. A covariant vector with n components wA and the transformation
(2.4)
The transformations (2.4) and (2.1) are said to be contragredient 3 ) to
each other. A covariant vector W;. is represented by two parallel hyper-
planes
(2.5) c = const.
1 ) In former publications we used the term measuring vectors instead of basis
vectors and called a basis a set of measuring vectors.
2) The KRONECKER symbol 15~ stands for n 2 scalars and accordingly its indices

are dead. So it ought to be written 15" but we shall not do so for historical reasons
A
and also to avoid such towerlike constructions as much as possible. The same
holds for 15i, 15",;, t5><A etc. All these symbols represent n 2 scalars with the values 1
or 0 and not, as might be expected, different components of some geometric object 15.
This is an exception to the general rule for the use of kernels.
3 ) See footnote 2 on p. 6.
8 I. Algebraic preliminaries.

given in this order and fixed to within a common translation, i.e. addi-
tion of the same arbitrary constant to c. 1) The order can be represented
by a (preferably curved) 2) arrow in a curve through the hyperplanes.
Fig.1 represents a covariant vector in E 3 In a centredEn a covariant
e
Vector W.l can be represented by one hyperplane W = 1, the parallel
En 1 through the origin playing the role of first hyperplane. Addition
of two covariant vectors is illustrated in Fig. 2, representing a section
with an arbitrary E 2 3)
I w 1 is the inverse of the segment on the 1-axis cut out by the
1

E,._ 1 's of w.. , measured by the unit on this axis. w 1 is positive if the
z

1
fi}
Fig. 1. Fig. 2.

sense of w-' corresponds to the +-sense of the 1-axis and negative other-
wise.
To every coordinate system (x) there belong n covariant vectors

(2.6) "e.<=U.<
* !l."

called the covariant basis vectors 4) of (x). Their components with respect
to (x') no Ionger have in general the values 1 and 0, and to (x') there
belongs another covariant basis ~~ .
The combination of a contravariant and a covariant vector v"w"
is an invariant
(2.7)

and is called the transvection of these vectors. Iv"w" I is the length

of v" measured by the segment cut out from the line of v" by the hyper-
planes of w_.. v"w" is positive if the senses of v" and w_. correspond and
1) Cf. ScHaUTEN, 1923, 1, p. 164.
2) A straight arrow could suggest that a special direction were given.
3 ) If v and w.l are parallel the construction can be effected by means of an

+ +
auxiliary vector u.l: v.l w.l = (v.< +u.<) (w.l- u;).
4 ) See footnote 1 on p. 7.
 2. Quantities in En. 9

negative otherwise. If v"w,. = 1 the vector v" fits exactly between

the hyperplanes of W.;. If v"w,. = 0 the direction of v" lies in the (n -1)-
direction of W.;.
In Fig. 3 we see the six basis vectors belanging to a coordinate
system in E 3 .
If n linearly independent contravariant vectors 'I!" are given, there
h
always exist n covariant vectors w,;; h, i = 1, ... , n suchthat

(2.8) 'I!" w,.=u;

h s.h

The sets 'I!" and /;;.; are said

tobe reciprocal to each other.
For each value of i,v" fits
i
between the hyperplanes of
~.; and lies in the (n - 1)-
direction of every other co- '
variant vector /;;" ; h =l= i. .
I
I

The co- and contravariant

basis vectors of a rectilinear
coordinate system are reci- e"
1
procal sets and conversely Fig. 3.
two reciprocal sets always
constitute the basis vectors of some rectilinear coordinate system.
If a vector v" is transvected by the n covariant basis vectors e;.,
we get n scalars
(2.9)

each representing the result of measuring the projection on an ax1s

by the unit on this axis. The equation
(2.10) v"=Je"
.:1

gives the construction of v" from these scalars and the contravariant
basis vectors. The two equations (2.9, 10) are numerically equal if
we use the coordinate system (x), but by introducing (x') we get the
equations
(2.11) a) v" = v .'"
e;.' b) v"I = ve
"'

v
that are no Ionger numerically equal. The x in is a dead index and ii
stands for a set of n scalars, but the x in 'IJ" is a living index transforming
10 I. Algebraic preliminaries.

with transformations of coordinates. Accordingly we must write

(2.12) v* v"
but V=I= v". Also e;. * ;.e" but e;. =I= Ae". This illustrates the fact that it
is not sufficient to give only the components of a quantity with respect
to some coordinate system, and that we must also give the manner
of transformation because the components of two quantities of different
kind with respect to some special coordinate system may happen to
be equal.
The process in (2.9) that leads from a living index to a dead one by
means of transvection with the basis vectors is called the strangling
of an index. It is clear that strangling of the same index or indices
at both sides of an equation leaves the equation invariant. But if the
strangled indices are not the same on both sides the sign = must be
changed into * . Moreover it is clear that in an equation with * that
has the invariant form, both sides have the same living running indices
and it is therefore allowable to introduce = instead of * .
4. A tensor 1 ) of contravariant valence p and covariant valence q with
nPH components P~ .. :P ;., ... Jg and the transformation

(2.13)

The place of the indices is essential, for instance we must not write
sometimes P~~l' and at another time P~~~ for the same quantity. If
only lower (upper) indices occur the tensor is called co-(contra-)variant.
If both occur it is often called cocontravariant (or mixed).
The tensor of valence 1, 1
(2.14)
is called the unity tensor. !ransvection of a quantity over one index
with the umty tensor does not change the quantity, for instance
(2.15) v--
A"A P"~' P"~'
. v

Tensors with more than one indexalso have components with respect
to two or more different coordinate systems, e.g.
(2.16) v = A"'P""
P"'" X V
=A":'P"'"',
V v

1) In former publications we used the word affinor instead of tensor and the
word tensorwas used in a more restricted sense (cf. footnote 1, p. 21).
2) A~{ stands for A:' Af. Generally we do not put upper and lower indices
in the same verticalline so as to avoid ambiguity if the process of raising and lower-
ing of indices (to be defined later on) is to be applied. But there are a few excep-
tions.
 2. Quantities in En. 11

These components are called intermediate. From this we see that A:'
and A1, in (2.1) and (2.4) are intermediate components of A~. This
justifies the use of the same kernelletter A in all three cases. The equa-
tions (2.1) and (2.4) now also express the fact that a vector is invariant
for transvection with the unity tensor. The same holds mutatis mutandis
for (2.13).
Two quantities of the same kind are said to be equiform if there
exist two coordinate systems (x) and (x') such that the components
of the first quantity with respect to (x) are equal to the corresponding
components of the second with respect to (x'), e.g.
(2.17)
5. A tensor may be determined to within a scalar factor. Then
it is called a pseudo-tensor (pseudo-scalar, pseudo-vector if the valence
is 0 or 1). If more pseudo-tensors occur it may be that the unknown
scalar factors in them are not entirely independent. For instance a
direction and an (n -1)-direction fix a contravariant and a covariant
pseudo-vector whose scalar factors are entirely free and independent.
But if the scalar factors in v" and w are free, the scalar factor in the
pseudo-tensor J: = v" w depends on the scalar factors in v" and w.
In the first case a pseudo-tensor can be distinguished if necessary by
the sign l J. For instance lP".J is the pseudo-tensor belanging to the
tensor P". and l J only means that there is an entirely unknown scalar
factor. In the second case it is most convenient to introduce next to
the ~" an auxiliary variable ~ 0 (or if necessary more of these variables) 1 )
with the transformation
(2.18)
A pseudo-tensor of class c is then defined by the transformation formula
of the form
(2.19)
For pseudo-tensors defined in this way we usually prefer gothic kernel
letters. Pseudo-scalars have only one component and no indices. It is
therefore necessary to introduce an extra notation in order to distin-
guish between the different values that the pseudo-scalar can take.
(0)

If we write ~ or ~ [0] for the component belanging to ~ 0 we get the

transformation formula
<0') (0)

(2.20) ~ = ac ~ or ~ [0'] = ac ~ [0].

The second notation is used if there is some reason for avoiding an
upper index, for instance if there is already another upper or lower index.
1) SCHOUTEN and HLAVATY 1929, 2.
12 I. Algebraic preliminaries.

6. If a quantity transforms like a tensor but with an extra factor T

dependent on LI it is called
a) tensor Ll-density of weight w and antiweight w' if T = Ll-w Lf-w';
J = complex conjugate of LI;
b) tensor density of weight w if T = !LI I-w (cf. I 10);
c) W-tensor if T = LIJ! LI I.
Ll-densities frequently occur in differential geometry but densities
and W-tensors are more used in physics. For densities we prefer a gothic:;
kernel and for Ll-densities a gothic kernel with the sign ,...... (snake)
above it, for instance b~~,.. For non-scalar W-tensors we prefer roman
kernelletters and for W-scalars a small greek letter, both with a snake.
But there are exceptions. For instance the mass or charge per par-
allelepiped of basis vectors is a scalar density of weight + 1 and is
generally denoted by a small greek letter.
If we consider real coordinate systems only, there are two kinds
of coordinate systems, transforming into themselves by coordinate
transformations with LI > 0 and into each other by transformations
with LI < 0. They can be distinguished by calling them right and lejt
systems if some geometric object suitable for comparison is introduced
in En.l) We call this the introduction of a screwsense inEn. The screw-
sense can be fixed algebraically by the W-scalar w,
whose component
is + 1 (-1) with respect to right (left) systems.
After the introduction of w there is a one to one correspondence
between Ll-densities and densities and between W-tensors and tensors.
For instance
(2.21} m"A -wffi"
1-'. I ' -
.
1-'. I''
If, instead of Ga, we use only the group of allreal affine transforma-
tions with LI > 0, corresponding quantities have the same manner of
transformation and can be identified. This is the first example of identi-
fication of quantities after introducing a subgroup of Ga. 2)
Analogaus to the case of pseudo-scalars we denote the value of
a scalar density or Ll-density or a W-scalar with respect to the coordinate
system c~) by writing c~) right over the kernel or c~J behind it, fm
instance
(><') (><)

(2.22) ~ = LI-W~ or ~ [~'] = LI-W j3 [~]

~') ..1 ~)
(2.23) P = prP or H~'J = 1~1 pc~J
1) This is just what we do in ordinary spacc if we call a system right or lef1
after comparison with our own right hand.
2) We get another example if instead of Ga we use the so called special affim
group Gs a consisting of all transformations of Ga with ..1 = + 1. Then the differ
ences between tensors, tensor densities and tensor ..1-densities vanish.
 3. Invariant processes and relations. 13

7. If a quantity has different kinds of indicesbelanging to different

spaces it is called a connecting quantity. These quantities occur fre-
quently. E.g. Bi: in (1.9) is a connecting quantity, " belongs to En
and b to Em. C1 in (1.8) is another connecting quantity. . belongs to
E,; and x belongs to the En-m that arises from reducing the En with
respect to the m-direction of Em. C1 and Bi: are called the covariant
and the contravariant connecting quantity of the Em in En.

Exercises.
I 2, 1. nP numbers P" "P are components of a contravariant tensor
of valence p if and only if for some definitely chosen value of q:;;;; p
I 2,1 a.) P"'"'"P Q.., ... ",
is a con travariant tensor for every choice of the tensor Q" ... ",.
I 2,2. P~ A= a.A~ if and only if P~ AvA =a.v" for every choice of v".
I 2,3. P~A= a.A~+PAq" if and only if P~AvA=a.v" for every v"
satisfying the equation PAvA=O.
I 2,4. Let P~ A be the tensor of the linear vector transformation
transforming the e" into the e"
A A'

I 2,4a.) e" = .l:P"

uA' "eAp .
A' 'r

Prove that
p~A *!51.
I 2,5. Prove that two equiform quantities can always be transformed
into each other by a homogeneaus linear vector transformation.

3. Invariant processes and relations.

The following processes and relations areinvariant for all rectilinear
coordinate transformations.
1. The addition of two quantities of the same kind (cf. I 2).
2. The construction of an isomer by changing the location of the
upper indices andjor of the lower indices, e.g.
(3 .1)
3 The (general) multiplication of two quantities, e.g.
(3.2) p><A
I'
Q'v!l. a-- R"!l A
aJJ.V"

4. The contraction (Faltung, Verjngung) with respect to an upper

and a lower index or with respect to several such pairs of indices, e.g.
(3.3) P"A . r1 = QA (!"
'XG(!
14 I. Algebraic preliminaries.

5. The transvection (berschiebung) of two quantities by applica-

tion of the process of contraction to the general product, e.g.
(3.4) I' Q''.vJ.-
P";. -R".v
The indices used for summation in contractions and transvections are
called saturated (also dummy), the other free indices. The transvection
is a kind of multiplication, i.e. it is distributive with respect to addition.
6. The process of mixing over p upper or p lower indices by con-
structing all p! isomers that can be obtained by permutation of these
indices, addition of these isomers and division by p!. The process is
denoted by round brackets ( ) .1) If indices have to be singled out the
sign II is used, e.g.
. ...V) w =
p(><II'I 1 (P><J.pv . . ."' + p V><pJ.
. .. w+ pJ.vp>< .... w +
(3.5) { 6

+ pw
><vpJ.
+Pw
vJ.p><
+P .. w)
J.><pv

where f-l stays put and the three other lower indices permute in the
remaining places. The effect of ( ) is not intercepted by any form of
ordinary brackets, e.g.
().6)
A quantity is called symmetric in p upper or p lower indices if it is
invariant for mixing over these indices. It is then also invariant if
two of these indices are interchanged. P(" "Pl is called the symmetric
part of P" "P
7. The process of alternation over p upper or p lower indices, effected
in the same way as the mixing process with the only difference that
all isomers obtained by odd permutations get a negative sign. This
process is denoted by square brackets [] 1), e.g.
p .. w -1.(pW + pW +PW _
(3.7) { [><J.Iplv] - 6 ><I' J.vp>< V><I'
_pW _pW _p W)
><VI' vJ.p>< J.><pv

As far as interception by other brackets is concerned [] has the same

property as ( ). A quantity is called alternating in p upper or p lower
indices if it is invariant for alternation over these indices. I t then
changes sign if two of these indices are interchanged. Alternation
over more than n indices always leads to zero. pC><,. >"p] is called the
alternating part of P""P.
Of these seven processes the first four are fundamental, the others
being only combinations of them. In order to see what can be done
1) The use of () and [] for mixing and alternating was proposed by BAcH
1921, 1, p. 113.
 3. Invariant processes and relations. 15

by these four fundamental processes we need a theorem concerning

concomitants. If a set of quantities is given, every quantity whose
components can be expressed as functions of the components of the
quantities of the set is called a (simultaneous) concomitant of the set.
Fora scalar the term invariant also is used. If all these functions are
rational integral the concomitant is called rational integral. The theorem
in question is:
All rational integral concomitants of a given set of quantities can be
derived from the quantities of the set by means of addition, construction
of isomers, multiplication and contraction.
This theorem is an immediate consequence of the well known first
fundamental theorem of the symbolic method used in the theory of
invariants. 1), 2)
Exercises.
I 3, 1. Prove that P"' "P vanishes if i t is symmetric in :v.1 ... "'
and if P"' "P w", ... w", = 0 for every choice of the vector u;..
I3,2. If u,"vAv"=O for every vector v" for which vAw;.=O, u(A><)
has the form
I 3,2 IX)

I 3,3. If w"P~;. =IXW;. for every vector W;. for which vAw;. = 0, P~;.
has the form
I 3.3 IX)
and IX is independent of the choice of W;..

I 3,4. If P~;. 1,w"vAv11 =0 for every choice of v" and wA satisfying

vAw;. = 0, P~().p) has the form

I 3.4 IX)

1) Cf. e.g. WEITZENBCK, 1923, 1, p. 93.

2)There is a !arge amount of Iiterature on the problern of finding algebraical
concomitants of a quantity or a set of quantities. Cf. for instance WEITZENBCK
1923, 1 ; 1939, 1 ; DROST 1931, 1; BERTI 1936, 1 ; WEYL 1939, 1 ; CrsoTTI 1940, 1 ;
1941, 1; 1943, 1; 1944, 1; WADE 1941, 1; 1943, 1; 1944, 1; 1945, 1; DoYLE 1941, 1;
DUBNOV 1941, 1; BRUCK and WADE 1943, 1; BRUCK 1944, 1; LITTLEWOOD 1944, 1;
1950, 1; RUTHERFORD 1948, 1; RACAH 1949, 1.
3 ) Propositions of this kind can always be proved by means of the following
weil known algebraic theorem. If a polynomial f(x 1 , ... , xn) does not contain
multiple irreducible factors and if the polynomial F (x1 , ... , xn) is zero for all
values for which f = 0, then F is divisible by f.
4) From a Ietter of A. FRIEDMANN.
16 T. Algehraic prcliminaries.

I 3,5. Verify that v[", "Pl can be written out in two ways

I 3,5 oc) v[". Yp] = ~ L (-1)s-1 v"

1, ... ,p
[><, "- Y-+ .. ><p].

I 3,5 ) v[><, . ><p] = L (-1)s-1

1, ... ,p
~j;- v[><, ... ><,1><,1 "~+ ... ><p].
s

I 3,6. If P~A"=P~[ApJ and if P~A~'w"v"oc.wA for every choice of v"

and wA satisfying vAwA = 0, P~ ;.p has the form
I 3,6rx)

4. Section and reduction with respect to an E_ in E".

Decomposition with respect to a rigged E_.
Let an EminE" through the origin be given by the equations

(4.1) {
a) ~" = B'br/ or b) qe
=0
b= 1, ... ,m; x=m+ 1, ... ,n.
The Em's parallel to it are given by the parametric equation (1.13}
with the parameters C"'. The r/ ate rectilinear coordinates in the Em
and the C'" rectilinear coordinates in the Em,; m' =n-m, arising from
the reduction of E" with respect to the m-direction of Em.
The points of the Em which lie in the hyperplanes of a vector wA
satisfy the equations [cf. (2.5)]
(4.2) wA Bir/ = c; WA Bir/ = c + 1.
Hence
(4.3)
is the section of wA with the Em. Conversely wA is not uniquely determined
by 'wb. It is only determined to within an additive vector whose (n -1)-
direction contains the m-direction of Em. The section of a covariant
tensor with valence > 1 is formed in the same way, e.g.
(4.4)
To every contravariant vector pa in Em there belongs one and only
one contravariant vector in E"
(4.5)
P" vanishes if it is transvected with C!. Conversely to every contra-
variant vector in E" which satisfies this condition there belongs one
and only one contravariant vector in Em. In fact pa and p" have the
 4. Section and reduction with respect to an Em in En. 17

same geometric interpretation, viz. an arrow lying in the m-direction

of Em. Therefore we consider P" and pa as two different sets of com-
ponents of the same quantity and accordingly we use the same kernel
letter. The En-components of a contravariant tensor of Em with valence
>1 are formed in the same way, e.g.
(4.6) a,b,c=1, ... ,m.
The Em's parallel to the given Em and passing through the points
;", ;" + v" have the coordinates C'" = C! ;", C'" = C! ;" + C! v".
From this
we see that to the contravariant vector v" of En there corresponds the
vector
(4-7) x=m+1, ... ,n
in the Em' We call "v'" the reduction of v" with respect to the given Em.
Conversely v" is not uniquely determined by "v'". It is only determined
to within an additive vector lying in the m-direction of Em. The reduc-
tion of a contravariant tensor with valence > 1 is formed in the same
way, e.g.
(4.8) "Vxyz def cxyz v"J.P
- ><Ap

To every covariant vector sy in Em' there belongs one and only one
covariant vector in En
(4.9) def CY
S;.= ;.Sy.

s;. vanishes if it is transvected with B~. Conversely to every covariant

vector in En satisfying this condition there belongs one and only one
covariant vector in Em' In fact sy is represented by the two Em,_ 1 's in
Em' resulting from reduction of the two En_ 1 's of s;.. Therefore we consider
sy and s;. as two different sets of components of the same quantity
and express this by using the same kernelletter. The En-components
of a covariant tensor of Em' with valence > 1 can be formed in the same
way, e.g.
(4.10)
If in En an Em' through the origin is given which has no direction
in common with the Em, the Em is called rigged (eingespannt). Conversely
the Em' is rigged by the Em. This Em' is not identical with the Em'
arising from reduction of the En with respect to the Em. But there is
a one to one correspondence between the points of these two Em,'s
because every point of the Em' arising from reduction represents an Em
and this Em has just one point in common with the rigging Em'. Hence
we may now, after having carried out the rigging, identify both Em:s.
From now on we have only one Em' in which the C'" in (1.13) can be
Schonten, Ricci-Calculus, 2. Auf!. 2
 18 I. Algebraic preliminaries.

used as rectilinear Coordinates and this Em' can be interpreted geometri-

cally in two ways. Firstly it may be interpreted as an Em' in En through
the origin with the equations [cf. (4.1a)]
(4.11)

and the contravariant connecting quantity c:.

Secondly it can be
interpreted as the manifold of all Em's parallel to the given Emin which
every Em is considered as a point. Substituting (4.11) in (1.13) we get
(4.12)

and this proves that c; ~t Cf C~ is the unity tensar of Em'.

The Em' is rigged by the Em. Hence the Em can be identified with
the Em arising from reduction of the En with respect to the Em' The
Em.'s constituting this latter Em are the Em.'s parallel to the rigging
Em' and are given by the equations [cf. (1.13)]
(4.13)

B~ is the covariant connecting quantity of the rigging Em'. Substitut-

ing (4.1 a) in (4.13) we see that ~def B~B~ is the unity tensor of Em.
After the rigging has been performed we are able to form
1. the sections of covariant quantities of En with Em and Em',
for instance
(4.14) b) "wy=c;.wA.

2. the reductions of contravariant quantities of En with respect to

Em and to Em', for instance
(4.15) a) "vx = C!v"; b) 'va = B~v".

These reductions can also be considered as projections on Em' (Em) in

the m-(m'-)direction of Em(Em.)
3 the En-components of covariant and contravariant quantities of
Em and Em'' for instance
(4.16) a) P" = B~pa; b) qA = B1qb
(4.17) a) sA= qsy; b) r" = c:rx.
In (4.14-17) the (a) formulae hold for every EminEn but the (b) for-
mulae are only valid if the Em is rigged. Applying the processes (4.16)
and (4.17) to B~ and c; respectively we see that B~, Bi: and B~ and
in the same way c;, Ci and c; are three different sets of components
 4. Section and reduction with respect to an E". in En. 19

of the same quantity and that

(4.18) B~ def B'; B~ and

form a fourth set. This justifies the use of the same kerne! letter B
and C respectively in all four cases.
In the following table results are gathered tagether for the rigged E"..

change of change of
pa components
-+ +-
components
rx

change of
va .._change ot v" with C~ v" = o; v" with B~ v" = 0 ----=--+ vx
components components

reduction reduction
~ V~ ~
(projection) (projection)

(4.19)

J'wb =Bgw._J +- section section

.....
"wy == Cy W

'w._ = B~ 'wb = B).w"

change of . change of
wb -<-
components
w._ wlth CY w._ = 0, w_, with Bbw_, = 0 components
-+wY

change of change of
qb component: components sy

If the Em is not rigged the Em' is the reduced En. Then Ol.l.ly the for-
mulae put in a frame remain valid and the reduction of v" can no Ionger
be interpreted as a projection. From (4.18), (1.10) and the correspond-
ing formulae for B~ and C~ it follows that for a rigged Em
(4.20) (B"J. + C") CJ.y =
"
C"y
and
(4.21)

Hence by means of B~ and C'). a vector can be split up into two parts
in Em and in Em' called the Em-part and the Em,-Part :respectively.
Besides its components in En, each part has also components in the
submanifold to which it belongs. Tensors with a valence > 1 split
up into more parts, for instance

(4.22)

The first is the Em-part and the last the Em,-part.

2*
20 I. Algebraic preliminaries.

5. Rank, domain and support of domain

with respect to one or more indices.
From a co- or contravariant or mixed tensor of valence p we choose
one index at some definite place and write fl, say, for this index. If
the tensor is then transvected over all other indices with all basis vectors
belanging to some arbitrary coordinate system (u) (that is, if 1, ... , n
are substituted for all other indices with respect to this system) we
get nP-l co- or contravariant vectors according as f.1 is an upper or a
lower index. The number r of linearly independent ones among them
is an arithmetic invariant and called the 11-rank of the tensor. This
rank is always ;;:;;; 0 and ;:;;;: n. The set of all vectors which are linearly
dependent on these r is called the 11-domain. If 11 is an upper (lower)
index the r vectors span an r-direction [(n~r)-direction] called the
support of the 11-domain. The 11-domain and its support are invariants
of the given tensor. The same holds mutatis mutandis for every quantity
of valence p. Instead of vectors we then have pseudo-vectors or vector
densities or W-vectors.
The rank and the domain with respect to more than one index is
defined in the same way. 1 ) Obviously the rank with respect to all
indices is always equal to one.
If the valence is 2 the rank with respect to one index is equal
to the rank with respect to the other one. Hence in this case there
exists only one non-trivial rank and this rank is equal to the rank of
the matrix of the components.

Exercises.
I 5, 1. If a symmetric tensor with valence p has 11-rank r, it can be
written as a sum of r products of a vector in the 11-domain with a sym-
metric tensor of valence p - 1. Hence every symmetric tensor can be
written as a sum of products of vectors. (Cf. I 6.)
I 5,2. Prove that the rank with respect to some indices is always
equal to the rank with respect to the other indices.
I 5,3. If a tensor v" "P is symmetric or alternating in all indices
its rank is the same with respect to every index. This rank is r if and
only if the equation v" "P w" = 0 has exactly n- r linearly independent
solutions w,;.
I 5,4. Prove that the v" "P from Exerc. I 5,3 has rank r if and
only if 2)
1)Cf. ALEXANDER 1925, 1; RrcE 1928, 1.
2)In (I 5.4oc) the alternation is only over the last index of each factor. In
(I 5,4) the last indices are alternated, and independently the last but one of
each factor are also alternated. The notation [[ ]] is very often used in this sense.
 6. Symmetrie tensors. 21

for s;;;;: r
I 5,4cx.) V~' [1', .. vh : l'sl
for s> r
and also if and only if
for s;;;;: r
I 5,4)
for s > r.

6. Symmetrie tensors.
A co- or contravariant tensor that is symmetric in alt indices is
called a symmetric tensor.l) Hence a symmetric tensor is invariant for
interchange of any two indices. Symmetrie pseudo-tensors, tensor
densities and W-tensors are defined in the same way. The number of
the linearly independent components of a symmetric tensor of valence p
is (n +; - 1 ). In a centred En, a covariant symmetric tensor P;., ... lv
is represented by the hypersurface with the equations
(6.1)
in point coordinates and a contravariant symmetric tensor Q~,. ~q 1s
represented by the hypersurface with the equations
(6.2)
in hyperplane coordinates. If we put zero instead of 1 in (6.1, 2)
we get the equation of the hypercone belanging to the symmetric
tensor. The hypercone determines only the symmetric pseudo-tensor
LP,~, ... ,~vj or LQ"' "v j respectively. From two covariant (contravariant)
symmetric tensors of valences p and q a symmetric tensor of valence
p + q can be derived by general multiplication and mixing over all
indices. This process is called symmetric multiplication of symmetric
tensors. It can also be applied to a greater nurober of symmetric
tensors. The polynominal at the left hand side of the equation of the
hypersurface of the symmetric product is the product of the polynomials
on the left hand side of the equations belanging to the factors. Hence
the notions division and divisor known from the theory of homogeneaus
polynomials can be extended to symmetric tensors. A symmetric
tensor Q is a divisor of a symmetric tensor P if and only if there exists
a symmetric tensor R suchthat Pis the symmetric product of Q and R. A
symmetric tensor without a divisor is called irreducible. It follows from
thc theory of homogeneaus polynomials that every symmetric tensor is the
1 ) In former publications we used the term tensor instead of symmetric tensor

(cf. footnote 1 on p. 10). The word "tenseur"' in CARTAN 1931'1, 1, I 30 has quite
another meaning.
22 I. Algebraic preliminaries.

symmetric product of a finite nurober of irreducible symmetric tensors

and that these factors are uniquely determined to within a scalar
factor. If all irreducible divisors have valence 1 the hypercone of a
covariant symmetric tensor of valence p consists of p hyperplanes and
the hypercone of a contravariant symmetric tensor of valence p con-
sists of p straight lines.
Exercises.
I 6, 1 ). The domain of a symmetric tensor is the same with respect
1

to all its indices. If the rank is r the tensor can be written as a sum of
symmetric products of vectors from this domain. The support of the
domain of a symmetric tensor W.t, .. "" of rank r is an (n- r)-direction.
Its hypersurface is a cylinder consisting of oo -J En_,'s with this (n- r)-
direction. The hypercone consists for r > 1 of oo- 2 En-r+/s all of
them containing the En-r with the equation
I6,1at) W.t, ... ""e"=o.
I 6,2 2). The rank of the symmetric tensor w .t, ... .t" is < n if and only if
I 6,2at) w(", ... .t, [p, [, ... w",. ... A..l p,.J.,.J = 0.
In this expression the mixing has to be effected over all indices except
the last two of every factor. The homogeneaus polynomial of degree
n (p- 2) belonging to the left hand side of (I 6,2at) is the hessian covariant
of the polynomial belonging to W.t, .. "".
I 6,3 3 ). If the symmetric tensors PA" and QA" satisfy the equation
P[" (A Ql'l l = 0 they differ only by a scalar factor.
I 6,4. A symmetric tensor of valence 2 is for n:;;:;; 2 always the sym-
metric product of two vectors.
I 6,5. A symmetric tensor W.t, ... "" is the symmetric product of p
equal vectors if and only if
I 6,5 at) W[.t,[A,IAa ... A"l w",] ><,] " "P = 0.

7. Multivectors.
A co- or contravariant tensor that is altemating in all indices is
called a multivector or alternating tensor. It is also called a p-vector
if its valence is p. 4) The valence can not be >n. Pseudo-multivectors,
multivector densities and W-multivectors are defined in the same way.
A multivector changes its sign if two indices are interchanged. The
nurober of linearly independent components of an altemating quantity
1) P. P. 1949. 1, p. 12.
2) SINIGALLIA 1905, 1, p. 371.
3 ) SCHOUTEN 1924, 1, p. 205;
EISENHART 1926, 1, p. 32.
') A multivector is also called antisymmetric or skew symmetric tensor by
some authors.
 7. Multivectors. 23

of valence p is (;). It is usual to call a 2-vector bivector, a 3-vector

trivector and a 4-vector quadrivector.
From two covariant (contravariant) multivectors of valences p and q;
p + q ~ n a multivector of valence p + q can be derived by general
multiplication and alternation over all indices. This process is called
alternating multiplication of multivectors. It can also be applied to a
greater number of multivectors, provided that the sum of the valences
is :::;;: n. For this alternating product we often use a notation without
indices. Instead of uC" ... "" v-< ... -<.J we write shortly [uv].I) Note that
(7.1) [uv]=(-J)Pq[vu].
A multivector Q is called a divisor of a multivector P if and only
if there exists a multivector R such that P is the alternating product
of Q and R. A vector v" is a divisor of v" "P if and only if vC" "P v"l = 0.
To prove this we take e" = v". Then v" "P cannot have non vanishing
1
components with aU indices =J=l. Hence e" is a divisor. If v" ... "P has s
1
linearly independent vector divisors this p-vector can be written as
the alternating product of these vectors and a (p- s)-vector. This can be
proved by taking the s vectors as the firstscontravariant basis vectors.
A p-vector that can be written as the alternating product of p vectors
is called simple. The necessary and sufficient condition that v" "P is
simple is
(7.2)
The same holds mutatis mutandis for covariant p-vectors. The ne-
cessity of the condition is obvious. In order to prove the sufficiency
we take the coordinate system in such a way that v1 .. P =l= 0. Then
(7.2) states that the p vectors v" 2.. P, v" 13 ... P, v" 12 4 ... P, ... , v"1 ... (p-1)
are divisors of v" "~>. But these p vectors can not be linearly dependent
because the i-th component v1 ... p of the i-th vector is =l= 0 and the
i-th components of all other vectors are zero. Hence v" .. "~> has p in-
dependent vector factors. 3)
Every n-vector is simple. Hence any two n-vectors differ only
by a scalar factor. But every (n -1)-vector is also simple. 4) In order
1 ) Cf. P. P. 1949, 1, I 7,10. This notation is intimately connected with CAR-
TAN's notations for differential forms (cf. II 12).
2 ) Givens proved1937, 1 thatv[><, ... ><pvl,A,j ... AP=O is anothernecessary and
sufficient condition and moreover that from v[><, .. "v vl, .. .<H,] ... Ap = 0 it always
follows that vC><, ... " vi. ..'. .l.s] .. v = 0 if s is odd. This proposition was proved by
WEITZENBCK 1908, 1 for P=3, n=5 and formulated 1923, 1, p. 87 for the
general case Cf. BoMPIANI 1952, 2.
3 ) If v"" "P is written as the product of these p vectors and a scalar, we get

the relations of V AHLEN. Cf. WEITZENBCK 1923, 1, p. 118.

') GRASSMANN 1862, 1, p. 61.
24 I. Algebraic preliminaries.

to prove this we suppose that v[", "n- 1 v;.,J ... Xn- 1 =I= 0. From this it
would follow that at least one of the vectors obtained by replacing
in vx, n-1 the last n - 2 indices by n- 2 of the indices 1, ... n were
not zero. By interchanging indices we can always arrange that v.1,z (n- 1 )
is this vector. But thm it would follow that
(7.3) v1 ... (n-1) vn 2 ... (n-1)- vn 2 ... (n-1) vl ... (n-1) =I= 0
and this is impossible.
The component v1 P of the simple p-vector
(7.4) v"'"P = p! 1v["' . .. pv"Pl

~~ \
I
I
I .!_ >0
I I w.ll
I

J Fig. 4.
~'~' Fig. 5. J

is the p-dimensional volume of the projection of the parallelotop of

v", ... , v" on the Ep of e", ... , e" in the (n- p)-direction of e" , ... , e",
1 p 1 P P+1 n
measured by the parallelotop of e", ... , e". Its sign is
1 p
+ if the screwsense
of the projections of v", ... , v" in this order is the same as the screw-
1 p
sense of e", ... , e" and - if it is not. From this it follows that a simple
1 p
contravariant p-vector is represented geometrically by adefinite p-dimen-
sional volume with an arbitrary form but adefinite screwsense in a defi-
nite Ep. Fig. 4 shows a simple contravariant bivector in E 3 with its
projection on the 31-plane.
A simple covariant p-vector is represented geometrically by a
cylinder of ooP- 1 parallel En_p's with an outer orientation, that is a
p-dimensional screwsense araund the cylinder. 1) The component w1 . P
of wJ, ... .lp is the inverse of the p-dimensional volume of the section of
this cylinder with the EP of e'<, ... , e" measured by the parallelotop
1 p
of these vectors. Its sign is +
if the outer orientation of the p-vector
corresponds to the screwsense of e", ... , e" and - if it does not. Fig. ~
1 p
Ehows a simple covariant hivecior in E 3 with its sections with the 31-
and 12-plane. 2) To every rechlinear coordinate system in En there belongs
1) SCHOUTEN-STRUIK 1922, 1. 2) w12 >0 should be w12 <0.
 7. Multivectors. 25

for everyvalue of p from 1 to n, a set of (;)simple contravariant p-vectors

and a set of (;) simple covariant p-vectors, uniquely determined for
p = 1 and p = n and determined except for the sign for 1 < p< n. These
p-vectors are altemating products of p basis vectors multiplied by p!.
Their p-directions and (n- p)-directions are the (;) coordinate-Ep's
and coordinate-En_p's respectively. The n-vectors
E"t"" det n! e["' ... e""l
(><) 1 n
(7.5) {
(><) def I 1 n
e.-, ... An = n. e[.l, eAn]
are most important among them. They are both represented by the
parallelotop of the basis vectors with the screwsense fixed by these
vectors in the order 1, ... , n. If the choice of the coordinate system is
free, they are determined to within a scalar factor. Their components
satisfy the identities (not summarize over x)
(><)
(7.6) E 1 "=el ... n =+1
(~<)

(1<)
(7.7) e p., p .. -
- n l

(7.8) EPt . Jlm "m+t r,.

(>-)

These can easily be verified.

From (7.5) it follows that
(7.9) E1' ... n' = Af""' EA, ... J.. = n! A1' ... n' El ... n -LI E1 ... n
(>.) , .<,. (~<) [1 . .. n] (><) - (x)

Hence El.. " is the component with respect to (x) of a scalar Ll-density
(><)

of weight -1. In the same way (~~ ... n is the component of a scalar
Ll-density of weight + 1. This is the first example of sets of numbers
that can be looked upon in two different ways as components of a
quantity.
The n-vectors E and (;) determine a one to one correspondence bet-
(><) ' .
ween con travariant pseudo-p-vectors and covarian t pseudo-(n -- p)-
vectors:

(7.10)

1) The sign oo means proportional to.

26 I. Algebraic preliminaries.

If v" "" is simple, Lv" "P J can be represented in a centred En by

an Ep through the origin.l) Now this Ep can be given by p (n- p)
numbers, hence v" "P has only p (n- p) + 1 independent components
and among the equations (7.2) there must be (;)-p(n-p)-1 in-
dependent ones. In order to obtain a convenient set of p (n- p) in-
dependent components of Lv" "P J we choose the coordinate system in
such a way that v1 P =!= 0 and put
B"
(7.11)
def vl. ({J-1) "({J+l) P
fJ = vl. .. p ' = 1, ... 'p.2)

Then it is easily proved that

(7.12) v"t"P = p! [><, "P] B fJ=UfJ

ot def ~"'
v1 ... p 1 p '
or
v, ... "'+"'P [ Bn. t t
(7.13) ( vl ... p =S.1 "': ... "'!, <;l, ... ,ro.=P+1, ... ,n,
ocv ... , ocp = even permutation of 1, ... , p.

The B~ are the p (n - p) components required. They are very often

used in problems concerning systems of partial differential equations. 3 )
The alternating quantities ~.......... and e,~, ... M. defined by

(7.14) ~l. .. n = +1 el. .. n = + 1

with respect to every rectilinear coordinate system in En are n-vector
L1-densities of weight + 1 and -1 respectively. They exist a priori
in E" just like A~ and they establish a one to one correspondence bet-
ween conha- (co-) valiant p-vectors and co- (contra-) variant (n- p)-

l
vector L1-densities of weight - 1 ( + 1) :

- (- 1)"' e)., ... Ag><, ... ><p v" .. "P

a) p I UA, ... Ag-
b) q! v""P = (- 1)-" bA, ... Ag ~A, ... Ag"'" "P
(7.15)
c) p ! fii" .. "P = (- 1)/Jp w A, . Ap ~' .. "" "' .. "
d) q I wJ.. ... Ap = (- 1) -/Jp eA, ... Ap><, ... ><g fii" .. "< ' q = n - p
1 ) A contra-(co-)variant pseudo-p-vector can also be represented by an
Ep-1 (En-p-1 ) in an E,._ 1 if it is simple. The components are then the homogeneaus
coordinates of this Ep-dEn-p-1). If it is not simple it can be represented by an
En-p-1-(Ep-dcomplex in En_1 . Hence for n = 4 a pseudo-bivector can be represented
by a line complex in E 3 that degenerates into all straight lines intersecting the
same straight line if and only if the pseudo-bivector is simple. Cf. e.g. WEITZEN-
BCK 1923, 1, p. 69, 83.
2) KHLER 1934, 1, p. 17.
3) Cf. e.g. KHLER 1934, 1; P. P. 1949, 1, Ch. X.
 7. Multivectors. 27

or in another form
a) iJA, ... Aq = ( -1)"v"""P;
( b) fii""P=(-1)flPw
(7.16) A, ... q'
1 ... q u1 ... :>tp = even permutation of 1, ... n.

The choice of the exponents oc and is free. We may take them

all equal to zero. But if subgroups of Ga are introduced and if we wish
to make use of the possibility of identification of quantities with respect
to such a group {cf. e.g. I 2) it is more convenient to give the oc's and
's values suitable for this special purpose. 1)
From (7.16) we see that the components of a contra- (co-) variant
p-vector, if Iahelied in another way, transform as the components of
a co- (contra-) variant (n-p)-vector Ll-density of weight -1 (+1).
Hence there is no geometric difference between these quantities. As
an example we prove explicitly that the components v23 , v31 , v12 of a
contravariant bivector in E 3 transform like the components b1 , b2 , b3
of a covariant vector Ll-density of weight (- 1). In fact

(7.17) { V
2'3' _
-
A2'3' "A _
" V -
(A2'3' -A2'3')
23 32 V
23+
cyc1. 1 ' 2' 3
= LI (A ~ v23 + A ~ v3 1 + A ~ v1 2)

according to the construction of the A from the Ar given in I 1.

The densities most frequently occurring in physics are contra- (co-)
variant multivectordensities of weight + 1 (- 1) which transform with
ILI! instead of LI. Hence for real coordinate transformations they
transform in the same way as Ll-densities if LI > 0 and change the sign
for LI< 0. This implies that their geometric interpretation is the same
as that of the corresponding Ll-densities to within the orientation.
Instead of an inner (outer) orientation they get an outer (inner) one.
e
If transvected over all indices with or ~ they give rise to W-multi-
vectors, transforming like ordinary multivectors but with a factor 2)
LI/JLIJ. The value of a W-scalar 3) is invariant for LI >0 and changes
its sign for LI< 0 (transformation of a right hand system into a left
hand system).
The alternating quantities in E 3 with their geometric interpretations
and their notations are to be seen in the following tables. 4)
1 )These identifications are elaborately dealt with in T. P. 1951, 1, Ch. III.
2) ScHOUTEN 1938, 1. The two different kinds of p-vectors were used in
DE RHAM and KODAIRA 1950, 1.
3) Often called pseudo-scalar in physical publications.
4 ) Cf. ScHOUTEN 1938, 1; ScHoUTEN and VAN DANTZIG 1940, 1 (note that in

these papers the sign -is used in another less satisfactory way); T.P. 1951, 1, Ch. II.
28 I. Algebraic preliminaries.

Inter- Second notation Weight Orientation

Ordinary notation
pretation

5,_,;.~ cov. triv. Ll-dens. -1

Number s scalar
~.u
sign
contr. triv. Ll-dens. +I

/ v~ contr. vect. b~ cov. biv. Ll-dens. -1 inner

$ w cov. vect. itJ~ J. contr. biv. Ll-dens. +1 outer

~ j'd contr. biv. f cov. vect. Ll-dens. -1 inner

~ h,~ cov. biv. r;~ contr. vect. Ll-dens. +1 outer

iJ scalar Ll-eens.
~ {
-1
}
p~.u contr. triv.
screw
q.U~ cov. triv. Cf scalar Ll-dens. +1

Inter- Ordinary nota tion Weight Second notation Orienta tion

pretation

Number+ ?, 1d~ cov. triv. dens. -1

li' W-scalar screw
screwsense 6~ J.,u contr. triv. dens. +1 r}
;6 t)~ cov. biv. dens. -1 v~ contr. W-vect. I outer

,
I

g
I
I
ro" contr. biv. dens. +1 WA cov. W-vect. inner
I

cj h cov. vect. dens. -1 Jd contr. W-biv.

I
outer

Ij" contr. vect. dens. +1 h~ cov. W-biv. inner

~ scalar dens. -1 ;pd.u contr. W-triv.

c() {I q scalar dens. +1 q,uu cov. W-triv.
sign

Exercise.
I 7, 1. Prove that for every m-vector v" "m:

I 7,1 ~) v'Kt;fm E l . . . J.mYm-!l"n === ( n) vPt Am E~<m+l . Xn] ){1 "'-m.

(~) m. (><)

8. Tensors of valence 2.
A. Cocontravariant tensors.
A tensor P~;. is represented by the homogeneaus linear vector trans-
formation (which is in a centred E,. also a point transformation)
(8.1.)
 8. Tensors of valence 2. 29

If the rank of P~ 4 is n this transformation is invertible

-1 -1 -1
(8.2) v"=P",). 'v 4 ' P"I' P'",}. =P"I' P 11,}. =A"
J., Det (P")
,). I--1-- 0.
-1
For every value of (! and a, P';" is equal to the minor of the element P~e
in the matrix of the P~, divided by Det (P~;.) (cf. Exerc. I 1,1 and
!8,1;2,.
The components of
8
( 3) SI p[x,
' [J.
px,] J.,] -- SI p[x, P''s] s -- SI P", [ pvss]
1 ' ' ' , 1 ' ' ' , 1 ' ' ' ,

are the s-rowed subdeterminants of the matrix of P~.< Hence the rank
is r if and only if
(8 .4) p[x, ... y.J { =F 0 for s = r (hence for s < r)
" -' = 0 for s > r.
The determinant of P is the scalar
(8.5)
Another concomitant of P is the trace (Spur, spur) P~,.
A vector v" is said to be an eigenvector of P if it satisfies an equation
of the form
(8.6) P~, v-' = A. v"
and A. is called the eigenvalue of P belanging to this eigenvector. Fr.om
(8.6) the rank of P~;.- A.A1 is < n, hence the possible values of A. are
the solutions of the algebraic equation of degree n
cp (A.) ~U -Jt Det (P~;.- A.A~)
(8.7) {
= (-Jt n! (Pr:u- A.Ag} ... (P~lnl- A.A:l) = 0
the so called characteristic equation of P. From this we see that if in
(8.6) we take the transvection with a covariant vector instead of wit h
a contravariant vector we get the same eigenval.ues. Writing (8.7) in
the form
(8.8) cp (A.) = A_n _ p A_n-1 + p A_n-2 __ ... + (-J)n p = 0
I 2 n
we have
( P= nn! Ag ... A~=1 p~ln] = P~"

l ~ ~ (7) nI A:; ... A;:; 1":,;:'! /, ~

. 1':1 1 P1: \ . /'"
j
(8.9)

P -_ n.1 p[l. [I.. .

pnl
. n] -
_ p[x, "... pn] __
" --
D t (JJ><)
e ..
n
30 I. Algebraic preliminaries.

Using the short notation of matrix calculus

(8.10) PQdetp~QQ~;,; P 2 =PP, etc.
we prove now that P itself satisfies the equation (8.8):
(8.11) cp(P) def pn_ ppn-1 + ppn-2 _ ... + (-1tPA = 0.
1 2 n

Equation (8.11) is easily verified by writing out the identity

(8.12) P [,., .
.[ ,
P"" A""+l
).] ""
-- 0
and making use of (8.9).
If the roots of (8.8) are determined, the eigenvectors belanging to
same raat ..1 can be found by salving
.
the equatian
(8.13) (.P:'A- fA) v" = 0.
But if we wish to campute eigenvectars far all raats it is mare canvenient

I
to determine the fallowing auxiliary quantities first.l)

l:
p defpA-P; pdeftraceP; p~efttracePP
1 1 1 2 1

P def pA-PP; pdef!tracePP

2 2 1 J 2

(8.14)
n-1
P ~pA-PP;
n-1 n-!3
pdet_!_tracePP
n n n-1

p detpA-P p.
n n n-1
It will be faund that P = 0 and this can be used as a check on the
n
carrectness af the operations. We naw define an auxiliary functian af ..
(-1)n-1 Q (.} def _ A ;.n-1 +
(8.15}
{
+P.n-2_p;.n-3+ .. + (-1)n-1 p J.+(-1)" p .2)
1 2 n-2 n-1
Then we have
(8.16) . Q(.) = A Det (P-..A) + PQ(.).
Hence
(8.17) PQ (.) = . Q (.)
1 1 1

and this proves that Q~A(1)v" is an eigenvectar af P belanging to the

eigenvalue . for every choice of v". This gives an easy methad for the
1
camputing af eigenvectars if the auxiliary quantities (8.14) and the
eigenvalues are known. But if a raat . has a multiplicity > 1 it may
1
1) FETTIS 1950, 1; Souriau 1950, 1.
2) FETTIS uses ). Q instead of Q but this can only be dorre for ). =j= 0.
1 1
 8. Tensors of valence 2. 31

happen that Q(A.) vanishes andin this case the method can not be used 1 ).
1
The matrix of a given quantity P~;. can be brought into a canonical
form by means of transformation of coordinates. We give here briefly
the results of the theory of elementary divisors 2).
a) For every eigenvalue A. with frequency rp there exist rp linear
p
independent vectors from which ap; are annihilated by (P-A. A )i but
p
not by (P-A.A)i- 1 . We call them vectors of order i with respect to A..
p p
i takes the values 1, ... , sp and the numbers ap; satisfy the conditions
CJp; ~ CJp(i-1)
(8.18) {
CJpl + ... + CJpsp = rp.
The numbers r P, ap; are arithmetic invariants of P. Also the Eap, spanned
by the first ap 1 vectors, the Eap,+ap, spanned by the first ap 1 -t-ap 2
vectors etc. are invariant but the directions in these flat spaces are
not individually uni-
quelydeterminedby P.
b) The rp-t-rq vec-
tors belonging to two 0

different eigenvalues
A. and A. are linearly
p q
independen t.
c) The coordinates
in En can be chosen in
such a way that the
matrix of P splits up
into a nurober of
square matrices ar-
ranged along the main 0
diagonal each of which
belongs to one eigen-
value. See fig. 6 for
n=14, r1 =5, r2 =4, Fig. 6.
r 3 =2, r 4 =2, r 5 =1.
If the rank of P is r, the last square has n- r rows and only zero's
in the main diagonal. The quantity P is now split up into a nurober of
1) Q (J.) is always zero if, in the canonical form
It can in fact be proved that
1
fig. 6 of the square belanging to A, there is at least one gap in the secondary
diagonal. 1
2) MuTH1899, 1; BeHER 1910, 1, p. 282; Pascal Repertorium I, 1, p. 102;
WEYL 1922,1, p. 88; ScHREIER in KLEIN 1926, 1, p. 379ff.; DICKSON 1926,1;
VAN DER WAERDEN 1931, 1, p. 120; EI, 1935, 1, p. 38ff.
32 I. Algebraic preliminaries.

quantities in an E,,, E,,, etc. and each of thcse quantities has only
one eigenvalue A, A, etc., in its own space. Hence, in order to get a
1 2
further re-arrangement of the matrix of P we may suppose that P
has only one eigenvalue /.. Then the result can be applied later to all
subspaces E,,, E,, etc.
d) Let P have only one eigenvalue A. For 11p; and sp we now write 11;
and s [cf. (8.18)]

(8.20)
{
11: +.'.~ +
11:::;;: 11 1
115 = n.

The basis vectors in En can now be chosen as follows.

Let e" be a vector of order s with respect to A. That means that it
1
is annihilated by (P - AA )' but not by (P - AA )s- 1 . Then we choose
the first s basis vectors as follows

e"
1
e"=c (P- I.A)~e.l
(8.21) 2 I

e" ={(P-I. A)'-1 }~ e'.

s I

The last vector e" is an eigenvector. For e" we choose another vector
s s+1
of order s which is linearly independent of e", ... , e" and another set
I s
of s vectors e", ... , e" is constructed. This process can be repeated 115
s+I 2s
times, the first vectors of every set and also the 115 s vectors of all sets
being linearly independent. In the Ea,_, of all vectors of order s -1
there are still 115 _ 1 - 115 vectors which are linearly independent of each
other and of the 115 vectors of this order already chosen. From these
115 _ 1 - 115 vectors arise 115 _ 1 - 115 sets of s -1 vectors and among them
there are exactly 115 _ 1 - 115 vectors of each order from 1 to s -1. Pro-
ceeding in this way we get at last n linearly independent vectors and
among them 115 of order s, 115 _ 1 of order s-1 and so on down to 111
vectors of order 1. It may happen that some of the differences 115 _ 1 - 115 ,
115 _ 2 - 115 _ 1 , etc. vanish. Now if z1 = s, z 2 , .. , za, are the numbers of
vectors in the sets constructed above, in the order in which they arise
we have
(8.22)

and the relations between the numbers 111 , ... , 112 , and z1 , .. , za, can be
made clear in a diagram, e.g. for 111 = 5, 112 = 4, 113 = 3, 114 = 1, n = 13:
 8. Tensors of valence 2. 33

0 0 0 0 0 <11 =5
0 0 0 0 <12 = 4

(8.23) 0 0 0 <1a = 3
0 <14 =1

The numbers z are of course also invariants of P. If the basis vectors

in En are chosen in this way the matrix of P takes the form

;.., 1
;.., 1
1
)., 1
).,

)., 1
)., 1
).,
- ~----

;..,11
I;., 1
).,
----------- ;.. 1
z'~- I-- I--
).,
t
Zs
-~-----------------
).,
1.
Fig. 7.

If this process is applied to each part of a general quantity P, which

comes from splitting up the quantity into parts in E,,, E,,, etc. then
every square matrix in a diagram like fig. 6 can be brought into a
form like fig. 7. We then get the first canonical form of the matrix
of P. 1)
The determinant Det (P- .A.A) can be written in the form
(8.24) Det (P- .A.A) =Il(.A.-.A.)'P ... (.A.-.A.)'P"v.
p p p

The factors on the right hand side, for instance (.A.- .A.)'v, are called
p
the elementary divisors of the matrix of P. Obviously two mixed tensors
of valence 2 are equiform (cf. I 2) if and only if their matrices have the
same canonical form, in other words if they have the same invariants z,
that is the same elementary divisors.
1) Another canonical form was given by WEYL 1922, 1, p. 99.
Schauten, Ricci-Calculus, 2. Auf!. 3
34 l. Algchraic prcliminarics.

B. Co- and contravariant tensors.

Every tensor P";. of rank r can of course be written in the form
(8.25) px?. = u"v;.
1 1
+ + u"v;..
r r

The vectors u" span the u-domain and the vectors v;. the J.-domain.
The same holds mutatis mutandis for the covariant case.
The components of
(8.26) S! p[><, [A, ... p><s] A,] = S! p[>< JA, I p><s] s = S! p>< [
1 1 p'><s' s]
1

are the s-rowed subdeterminants of the matrix of px?.. Hence the rank
of P";. is r if and only if the quantity (8.26) vanishes for s > r but not
for s = r. 1 )
If P"-' is symmetric the quantity (8.26) is invariant for the operation
of interchanging the set of indices :><:1 , ... , "s with the set of indices
A1 , ... , As. Such a quantity is called a symmetric s-vector-tensor. It
plays the same role with respect to s-vectors as a symmetric tensor
of valence 2 does with respect to vectors.
It is possible to define to a co- or contravariant tensor of valence 2
an adjoint (cf. Exerc. I 8,2). If the determinant of the matrix is not
zero this adjoint is equal to the product of the determinant and the
inverse. Relations like (Exerc. I 8,2 ) can be established for the ad-
-1 -1
joint.2) Note that P"~' PI';.= A:l and that generally P"~' P;.l' =!= A1.
C. S ymmetric co- and contravariant tensors.
If we consider real and complex symmetric tensors and real and
complex coordinate transformations the following theorem holds:
Jf the symmetric tensor P";.(Q"-') has rank r the coordinate system (:><:)
can always be chosen in such a way that the matrix of P".. (Q";') has the
diagonal form with r numbers + 1 and n - r numbers 0 in the main diagonal.
If only real symmetric tensors and real coordinate transformations
are considered we have the theorem:
If the symmetric tensor P".< (Q"-') is real and of rank r, there always
exists a real coordinate system such that the matrix of P".< (Q";') has the
diagonal form. with s numbers - 1, r - s numbers +
1 and n - r numbers 0
in the main diagonal. 3 )
The number s, called the index, is an invariant of the symmetric
tensor. The sequence - - - + + + with s- -signs and (r- s) +-
signs is called the signature. The signature is said to be even (odd) if s
1) WADE 1943, 1, calls therefore (8.26) the rank tensors.
2) P. P. 1949, 1, p. 28.
3) Cf. for the proof of these two theorems DICKSON 1926, 1, p. 70.
 ~ 8. Tensors of valence 2. 35

is even (odd). The symmetric tensor is called positive (negative) definite

if s = 0 (n) and r = n and indefinite in all other cases. It is called posi-
tive (negative) semi-definite if n- r > 0 and s = 0 (r).
It follows from these theorems that two symmetric tensors are
equiform for general transformations if and only if they have the same
rank and that two real symmetric tensors are equiform for real trans-
formations if and only if they have the same rank and the same index.

D. Co- and contravariant bivectors.l)

An antisymmetric matrix always has an even rank. Hence the rank
of a bivector is always even. If v"" has the rank r it can be proved that 2)
(8.27) r! v[><,[J., ... v"']J.,] = ( , r! ) 2 vr"'"' ... v"'_,",] v[J.,J., ... v-<,_,J.,J.
2''(tr)!
Hence, if r = n
(8.28) Det (v"J.) = ( ' n! v[lZ . .. vn-1 nJ). 2)
2"n(~n)!

The same holds mutatis mutandis for a covariant bivector.

Let F;." be a bivector of rank r and let u" and v" be two vectors such
that
(8.29)
Then, if
(8.30)
we will prove that the rank 'r of
(8. 31)
is r- 2. Because the rank of F;." is r, there exist n- r linearly in-
dependent contravariant vectors whose transvections with F;." vanish.
Because of (8.29 to 31) their transvections with 'F.;" also vanish. Hence
'r-:;;;, r. Now, accordingto (8.30, 31) thetransvections of u" and v"with 'F;."
vanish and these vectors are linearly independent of each other and
of the n- r vectors just mentioned. Hence 'r-:;;;, r- 2. N ow we have
IF[J., [><, .. F,] ><r] = 'F[A,[><, .. 'F,]>c,]- r w[,[><, 'F;.,", . 'FA,]><r] +
(8.32)
l + (;) w[,[><, WJ.,", 'F,><a .. 'F,] >c,]

and from this we see that the rank of F;." could not be r if the rank of
'F"'" were <r-2. Hence 'r=r-2. If this process is carried out
1) There is a great number of papers on special properties of tensors, especially

of symmetric tensors and multivectors. We mention only a few: RusE 1936, 1;

WONG 1940, 1; GUREWITSCH 1950, 2; 3; 4; MoREAU 1950, 1.
2) P.P. 1949,1, p.28.
3*
 36 I. Algebraic preliminaries.

ir -1 times we get a decomposition of F,." into tr simple bivectors.

This is called the splitting up of Fh into tr blades. 1 ) The same can be
done with a contravariant bivector. Note that the blades are by no
means uniquely determined by the bivector. If F;." is real, u" and v"
can be chosen to be real and the process can be carried out in such
a way that all blades are real.
If
1 Q
(8-33) F;."=/;."++fX; e=tr
is such a decomposition, the r-dimensional domain of F;." is spanned by
the tr two-dimensional domains of the blades. Hence these domains
can have no vector in common and from this it follows that
(8-34)
Hence for a bivector, besides the necessary and sufficient condition

F F { =1= 0 for s= r
(8-3 5) [A,[x, . .<,]x,J = 0 for s> r

for the ranktobe equal to r, we still have another necessary and sufficient
condition 2)
(8.36) F F { =!= 0 for s = r
[)., .<, . s-r s]
= 0 f or s > r.
This also holds mutatis mutandis for a contravariant bivector.
If we start from (8-33) we can always choose the coordinate system (x)
in such a way that
1 1 2 2 3 4
(8.37) /;."=2e[;.e"l; /;."=2e[;.e"J; etc.
That proves the theorem:
If a real bivector has rank r, there is always a real coordinate system
such that the matrix of the components with respect to this system consists
of exactly ir matrices of the form II- ~ ~ I , arranged along the main diagonal
and zero's at all other places. 3)

1)There is amistakein the proof in EI 1935, 1, p. 46 which is corrected here.

2)Cf. (8.27) and GRASSMANN 1844, 1, p. 206; 1862, 1, p. 56; ROTHE 1912, 1, p. 1039.
3 ) The results obtained under A, C and D of this section give a possibility of

classification for cocontravariant and symmetric tensors of valence 2 and for

bivectors. The problern of the classification of quantities with a valence >2 is only
solved in very special cases. For trivectors the classification was given for n = 6
by REICHEL 1907, 1, for n = 7 by SCHOUTEN 1931, 1 and for n = 8 by GUREWITSCH
1935, 1. Cf. W. KMMERER 1927, 1; GUREWITSCH 1933, 1; 2; 1934, 1; 2; 3; 1935, 2;
1950, 1; WEITZENBCK 1937, 1; 2; 1938, 1; PAPY 1946, 1; PAPY and TURNAY
1946, 2; HuTeHINsaN 1948, 1.
 So Tensors of valence 2o 37

If a covariant bivector F;.,. and p linearly independent covariant

vectors ~;., 000 .~;. are given, we may require the section of F;.,. (cfo I 4)
with the En-p spanned by these p vectorso Taking (x) in such a
q+1 n
way that the vectors are e;., 000, e;.; q=n-p, the ~; a.=1, 000, q can
be taken as coordinates in theEq (cfo I 4) and we have for the connecting
quantity
B(r~ ex; = 1, 000 , q

and for the section

(839) *F;.,.e x*p

'F.rx= e ~ rx.
rx

This means that with this special choice of the coordinate system we
have only to drop all components of F;." with indices q + 1, 000 , no
In order to find the rank of 'F we have to consider the set of multi-
vectors (8036)

(8.40)

r
the last non vanishing multivector of this set being I 0 The multivectors
2 4 2 4
'I, 'I, 00 0 belonging to 'F are the sections of I,.T, 000 with the En-P
and these sections can be found in the same way as 'F, ioeo by dropping
all components with indices q + 1, n with respect to the special 0 ,

coordinate systemo N ow consider for instance the multivector

4 1 p 4 q+I n
(8.41) [Iu.oou]=[I e .. oe].

Components of I with indices q + 1,

4
0 0. , n can not occur in this quantity.
4
Hence it is zero if and only if 'I vanishes. This proves the following
theorem:
The section of the covariant bivector F;.,. with the En-p spanned by the
covariant vectors ~;., 0 , /;,;. has rank 'r if and only if

2s 1 p { =f= 0 for 2 s = 'r

(8.42) [Iuooou]
= 0 for 2 s > 'r.
We call r- 'r the reduction number of the vectors ~!, 000 , t with respect
to the bivector F.
38 I. Algebraic preliminaries.

It can be proved that the reduction number of J, ... ,/l with respect
to F is u if and only if the matrix
-<--n------>- +---P-----+
+ 1 9
n F,.;. U;. Ui.

--z:,.
~ 1
0 0
(8.43) i
p
p
t -u,. 0 0
has the rank R=r+2p-u. 1 )

Exercises.
I 8,1. (Cf. Exerc. I 1,1.) Prove that for every tensor P~;. of rank n
(cf. Exerc. II 2,2)
I 8,1 cx)
p~ _ OlogDet (P~a)
.;.- oP~,.
I 8,2. The quantity
I 8,2 cx) P" def
,;.=-a~
oDet (P~ 0 )

is often called the adjoint of P. It exists always, also if Det (P) = 0.

The elements of the adjoint are the minors of the elements of P. The
adjoint of the adjoint of P does in generalnot equal P. Prove that for
every tensor P~;. of rank n (cf. Exerc. II 2,2)
I 8,2 ) P*",;. -- n.n.I A"[1 Au
[;.! p2.2 . . .
pnJ
. n]

I 8,3. Prove that Q (A.) defined in (8.15) is the adjoint (cf. Exerc. I 8,2)
of P- A.A.
I 8,4. If p, q and rare the ranks of P, Q and R=PQ, prove that
r:;;;; p, r:s;, q a.nd that r=P if q=n.
I 8,5. 'Prove that the trace of PQ- QP is always zero.
I 8,6. There exist always two coordinate systems (u) and (u') such
that the matrix of the intermediate components P" ;: has the diagonal
form. There exist always two tensors Q~;., R~;. such that the matrix of
Q~e R~a pea has the diagonal form.

I 8,7. If .P;.,. and Q;.,. satify the equation

I 8,7 cx) P[,.[i. Q,. 11 = 0
and if the rank of .P;.,. is 1 or 2, then the rank of Q;.,. is 0, 1 or 2. If the
rank of .P;.,. is > 2, Q;.,. is zero.
1) Cf. P.P. 1949,1, p.27.
 8. Tensors of valence 2. 39

I 8,8. If P;,x is a tensor of rank n and D the determinant of its

matrix, prove that

u><l -1
I 8,8 IX) = - px}.U V D.
0 , " .!

I 8,9. P" 2 is determined to within a factor +1 by the quantity V

(8.26) for s = 1, ... r -1 but not for s = r. The same holds mutatis
mutandis for covariant and mixed tensors.
I 8,10. The reduction number of one vector with respect to a bivector
is always 0, 1 or 2. It is zero if and only if the vector belongs to the
domain of the bivector.
I 8, 11. Prove the inequalities [cf. (8.43)]
2p;;;;,R-;;;r+2P
I 8,111X) R~r
R;;;;;, n+p.
I 8,12. The m;;;;,n vectors b;., ... , V;. are linearly independent and
the m vectors tb;., ... , w"' satisfy the equation
I 8,121X)
w"' =
U V
Prove that there exist m 2 coefficients IX such that IX V;. and that
uv uv
IX
uv
= IX ; U, V
vu
= 1, ... , m. 1)

I 8,13. If P~" hasrankn and if P~ ~ = 0; IX= 1, ... ,m; 17 =m+ 1, ... ,n

prove that
-1
I8,131X) prx. =0'
~

-1
I8,13) ~YP:=A~; 1X,,y=1, ... ,m
-1
I8,13y) P~. c P'. ~ =A~
~ '
;,'Yl,C=m+1,
'I
... ,n.
I 8,14. An Em is imbedded in En. For the tensor P, of En the
rank of B~ P'";. is r and the rank of B~t P'"" is r'. The support of the
.1.-domain of B~ ~,J. intersects Em in an Es. Prove that,
I8,141X) r- (n- m) ;;;;;, r' = m- s ;;;;;, r.
+
I 8,15 2). If c""' = a;.>< i b;.,.;, where a;." and b;.x arereal and symmetric,
eh has a rank < n and ah is positive definite or positive semi-definite,
then there exists a real vector s>< such that eh s>< = 0.
1) CARTAN 1919, 1, p. 151.
2) PEREMANS, DUPARC and LEKKERKERKER 1952, 1; QUADE 1953, 1.
40 I. Algebraic preliminaries.

9. Introduction of a metric in an E ...

In an En the magnitude of two contravariant vectors can only be
compared if they are parallel. If we wish to establish a comparison
of magnitudes of non-parallel vectors we have to introduce a metric,
i.e. a definition of "length". This definition must satisfy the con-
ditions that the length of a real vector is always positive and that
multiplication of a vector by a real nurober k alters its length by a
factor lkl. Hence, if F(~"-~") is the length of ~"-~" or the distance
2 1 2 1
of the points ~" and ~", we have
1 2

(9.1) F (k (~"- ~")) = lkiF(~"- ~").

2 1 2 1

We assume that the function F is single valued and continuous and

that it has continuous derivatives up to a certain order. The equation
(9.2)

represents the indicatrix of the metric with respect to the point ~".
0
In a centred En we have ~= 0 and ~" is the radius vector. According
0
to (9.1) the indicatrix is symmetric with respect to the point ~". To
0
every direction there belongs one and only one (n -1)-direction viz.
the tangent (n -1)-direction of the indicatrix at its section with the
straight line through ~" in the given direction. If the equation of the
0
indicatrix is algebraic, its degree is even because ~" must be the centre.
0
The projection of a vector u" on a vector v" can now be defined as
the projection in the (n -1)-direction that belongs to the direction
of v". If u is the length of u" and 'u the length of the projection, the
angle between u" and v" (in this order) can be defined by

(9.3) cos (u", v") def ~ 1 )

u

But in general cos (u", v") =\= cos (v", u").

If we wish to define a "rotation" in a centred En in a natural way
we will have to define it as a linear homogeneaus transformation which
leaves invariant both an n-dimensional screwsense and the indicatrix. 2 )
That excludes all forms of indicatrix which are not invariant for some
subgroup of Gho. This subgroup must be chosen big enough to guarantee
some freedom of "motion ". For instance it could be required that
1) FINSLER 1918, 1, p. 39.
2) Cf. WEYL 1921, 1, p. 12Sff.
 y. Introduction of a metric in an En 41

an E 1 through the centre, an E 2 throughE1 , an E 3 through E 2 , etc.

being given, there always exists a rotation turning E 1 in a given direction,
E 2 in a 2-direction containing this direction, etc. Then it can be proved 1)
that the only possible indicatrix is a non-degenerate real quadratic
hypersurface. Following another line of thought it would be possible
to require that the angle between two directions be independent of
their order. This condition also leads to a real quadratic indicatrix. 2 )
Accordingly, let us take the equation of the indicatrix in the form
C9.4) g'"" W- 0e) W- 0~") = 1
where g;_,. is a real symmetric tensor of rank n and index s:;;:;. n.
g;.,. and its inverse g";. are then called the covariant and contravariant
fundamental tensor. An En with a symmetric fundamental tensor is
called an Rn and is said to be ordinary if g;_,. is positive definite. The
geometry in Rn is called euclidean if g;_,. is definite and minkowskian
if g;.,. is indefinite. The length of a vector v"
(9.5)
is always positive or zero but g;.,. v'- v" can be > 0 or < 0 or zero. In
the first case v" is called time-like and in the second space-like. 3) The
time-like vectors through the origin fill the + -region and the space-
like vectors the - -region of a centred Rn. The vectors with zero
length fill the nullcone. A real nullcone only exists in the indefinite
case 0 < s < n. A vector with length + 1 is called a unitvector. The angle
between two real vectors, both of which lie in the + -region or both
in the - -region is defined by
g u'-v"
(9.6) COS (u", V") = COS (V", U") d f
e I for the -region.
I
><
i Vgeaue u" gTw vTvw
The possible values of the cos depend on the form of the section of the
plane of the vectors with the nullcone:
( Section cos

(9.7)
I~ 2 real E 1 's
{ or ~
+1
:2::
-1
1 real E 1

I
1
2 imaginary E 1 ' s { ::;;: +1
l and :2:: -1.
1) HELMHOLZ 1868, 1; LIE 1890, 1; WEYL 1921, 1, p. 2fi.
2) FINSLER 1918, 1, p. 40.
3) The space of special relativity is an R 4 with signature - - - and a min- +
kowskian metric. For time-like vectors the term duration (Dauer) is sometimes
used instead of the term length.
 42 I. Algebraic preliminaries.

If we take a unitvector as first vector we see from this that the length
of its projection on the other vector is :2; 1, = 1 and ;;::;; 1 in these three
cases. In the first case the geometry in the plane is minkowskian and
in the third case it is euclidean. In the second case the plane is not
an R 2 because it does not contain a symmetric tensor of rank 2. Such
an E 2 in Rn is called isotropic.
An Ep in Rn is called isotropic if the symmetric tensor of its metric
has a rank < p and full isotropic or a null Ep if this rank is zero.
Any two vectors u" and v" for which g," u" v" = 0 are said to be
mutually perpendicular. With this exception the angle between a vector
in the + -region and a vector in the - -region is not defined.
An Ep and an Eq in Rn are mutually perpendicular if every direc-
tion in Ep is perpendiculat to every direction of Eq. All directions
perpendicular to Ep span an En-p The section of Ep and this En-p
is an E 1 with the property that all its directions are mutually perpendi-
cular. Such an E 1 is a null-E1 . Its t-direction lies in the nullcone.
A null-E1 can never be an R 1 For t > 0 the Ep must be isotropic be-
cause it contains a direction perpendicular to all its directions.
By using-the second theorem of I 8.C, a real coordinate system (h);
h = 1, ... , n may be introduced such that

h
Such a coordinate system is called cartesian. We write ih, i; for its
t

basis vectors. These vectors are mutually perpendicular unitvectors.

We always use roman running indices and vertical fixed indices for
cartesian coordinate systems in Rn.
Fig. 8 shows the minkowskian geometry in two dimensions. The
signature is - +. The indicatrix consists of the hyperbolas
(9.9)
or in paranietric form
(9.10) x1 = coshcp; x 2 = sinhcp; x1 = sinh cp; x 2 = cosh cp .
The fact that, if the figure is looked at from a euclidean point of view,
the 1-axis seems to be perpendicular to the 2-axis and that i and i
l' 2'
seem to have the same length, has no significance at all. The whole
figure could be subjected to an affine transformation and would then
represent just the same thing. In a minkowskian metric, length is
fixed by the indicatrix, hence i, i, i, and i all have the same length.
l 2 1' 2'
 9. Introducbon of a metric in an En. 43

The area of any part of the plane happens to be the same for both metrics.
The angle cp is real and ~ cp is the shaded area. According to (9.6) we
have
(9.11) cos (i , i ) = cosh cp ;;;;; 1
1 1'
for the angle between 1-axis and 1'-axis. The vectors i and i are per-
l 2
Pendicular to each other and so are 1'i and 2'i . But the angle between
arbitrary directions that do not belang to the same region is not defined.

1'

LL~~~~~~~~~~~----------------1
~--------1------~
1---------cos h tp------~
Fig. 8.

The +- region and the -- region each consist of two parts. Every
orthogonal transformation leaves each region invariant as a whole, but
it may happen that it interchanges its parts. So we have four different
kinds of orthogonal transformations in a minkowskian R 2

r +-region --region L1 example

(9.12)
I A) not interchanged not interchanged +1 i-+ i;
B) not interchanged interchanged
1

-1 i-+-i;i-+i
1'
t-+ t
2 2'

~ 1 1' 2 2"

in terchanged not interchanged -1 i -+i; t-+-t

IC) 1 1' 2 2'
. . . .
l D) interchanged interchanged +1 1t-+-t;
1'
t-+-t
2 2'

The transformations (A) form a group and are called proper min-
kowskian rotations and in the theory of relativity, LoRENTZ transforma-
tions. (B) and (C) are minkowskian reflexions and (D) are improper
44 I. Algebraic preliminaries.

minkowskian rotations. They do not form a group neither separately nor

together. But (A), (B), (C) and (D) together form the group of all
rninkowskian orthogonal transformations.
For all values of n, the fundamental tensor establishes a one to one
correspondence between co- and contravariant vectors

(9.13)

We use the same kernel for corresponding vectors and consider v"
and V;. as two different sets of components of the same quantity, called
vector, that can be represented geometrically either by an arrow or by
an oriented set of two parallel hyperplanes. 1) The distance between the
two hyperplanes is the inverse of the length of v". Theorientations of v"
and V;. are equal if v" lies in the +
-region and opposite if v" lies in the
--region. The process (9.13) is calledraising and lowering of indices. 2 )
It can also be applied to one or more indices of a quantity of valence > 1,
e.g.
(9.14) P;).v = ge" P~;.; P,.;.a = P;;. gva
The relation
(9.15) g;.,.=g;.eg,.agea
justifies the use of the same kernel in g;.,. and g";..
i
Because i" and i;. have the same orientation for every value of f,
1
we have
1 s+l n
(9.16) i" = - i";
1
0 0 0 ; i" = - i"; s+l
s
i" = + i" ' 0 0

' '
i" = + i"
n

and accordingly
h = 1, s
(9.17)
0 0 0'

h=s+1, ... ,n.

The length of v" in cartesian coordinates is

(9.18)

and for the equation of the nullcone in these Coordinates we get

(9.19)
1)Identifications of this kind and their influence on the choice of the coefficients
IX, inI 7 are dealt with elaborately in T. P. 1951, 1, Ch. III.
2 ) This process will never be applied to geometric objects which.. are not quan-

tities. Cf. footnute 2 on p. 10.

3 ) If cartesian coordinates are used we often write x instead of ~ for the kernel

of the radius vector in a centred Rn.

 9. Introduction of a metric in an En. 45

For the n-vectors belonging to a cartesian coordinate system (h) we

(h)
use the kernels I and i (cf. 7.5)

!
(II)

!''"" ~5 n! i[><, ... i"n]; p ... n

(h)
= +1
(k) 1 n
(9.20) (h) 1 n (h)
def
~.t, ... .tn= n.1 ~[A,

~n]>

il. .. n = + 1.
They change sign if we pass from a right hand system to a left hand
system and are invariant if (h) and (h') have the same screwsense.
Geometrically they are represented by a unit volume with the screw-
(h)
sense of (h). The component i 1 ... n may be considered as the component
of a scalar L1-density f of weight + 1
. (~) 1
(9.21) t [u] = ~l. .. n = ~ .
(h)

w +
If is a W-scalar with the component 1 with respect to all rectilinear
coordinate systems with the same screwsense as (h), the product

(9.22)
is a scalar density of weight + 1 with a component + 1 with respect
to all cartesian coordinate systems. This latter density is intimately
connected with g,t". In fact
(9.23)
is a scalar density of weight + 2 and g [h] = + 1, hence
(9.24) i = 1glf.
This density has no screwsense and this is in accordance with the fact
that g;," fixes length and angles but not a screwsense.
After introduction of a fundamental tensor we may ask whether
any vectors exist which satisfy the equation
(9.25) 7;." v" = a V;.
where 7;." is some given real symmetric tensor and a a suitably chosen
scalar. If we write
(9.26)
we see that the rank of 7;."- a gA" must be < n. Hence a must be
a solution of
(9.27) Det (7;."- a g,t") = 0.
46 I. Algebraic preliminaries.

This algebraic equation of degree n in a is the so-called 5-equation,

weil known from the theory of quadrics in ordinary R3 If g;_" is definite
it can be proved that the solutions of (9.27) are all real. 1) If a and a
1 2
are two different solutions and v" and v" vectors satisfying (9.26) for
1 2
a = a1 and a = a 2 , we have
(9.28) v'" T v" = av'"g v" = av'"g v"
1 l< 2 1 1 '"" 2 2 1 '"" 2 '

hence v" and v" are perpendicular to each other. If a is a solution of

1 2 1
(9.27) with multiplicity m it can be proved that the vectors satisfying
(9.26) for a = a1 are the vectors that lie in a definite real m-direction
called the principal multidirection belonging to a, and that principal
1
multidirections belonging to different solutions are mutually perpendi-
cular. If the contravariant basis vectors of a cartesian coordinate
system (h) are chosen in these multidirections, we have
(9.29)
and this proves the
Theorem of principal axes of a symmetric tensor in Rn 2) :

If 1';.,. is a real symmetric tensor and if the fundamental tensor is

definite, the realways exists a real cartesian coordinate.system (h) suchthat
(9.30) T.~o-"11;__0; h=Fi.
The axes of this coordinate system are called principal axes of the
SJ[mmetric tensor. They are uniquely determined if and only if the
solutions of the 5-equation are all different.
There is a corresponding theorem for real bivectors. 1 )

Theorem of principal blades of a bivector in Rn 3):

If F"'" is a real bivector of rank r and if the fundamental tensor is
definite there always exists a real cartesian system (h) such that all
components pk vanish except

(9-31) pu = _ F21; pa4 = -F43; ... ; p-1,r = _ pr,r-1.

The t r simple bivectors with the 2-directions 1 2; 3 4; ... ; r - 1, r are

called the principal blades of the bivector. They are uniquely deter-
mined if and only if the components F12, F 3 4, ... , pr- 1, r all have dif-
ferent absolute values.
1) Cf. e.g. DICKSON 1926, 1.
2) See p. 34.
3) See p. 36.
 9. Introduction of a metric in an En. 47

The theorems of principal axes and of principal blades do not hold

if the fundamental tensor is indefinite because in this case a real sym-
metric tensor or bivector may possibly have a special position with
respect to the real nullcone. But if the symmetric tensor or bivector
does not have any such a special position the theorems remain valid.
Orthogonal coordinate transformations are transformations from
one cartesian coordinate system (h) to another (h'). Because
(9.32)
we have
(9.33)
and, because Ai,~- i;; Ar-*- ih' this leads to the reciprocal relation
i' h
between the covariant (contravariant) components of the contra-
variant (covariant) basis vectors of both systems

(9-34)

Orthogonal point transformations in a centred Rn are those that

leave the fundamental tensor invariant. Hence if T~ _. is such a trans-
formation we have
-1
(9.3 5) T-'.a g'"' = T".e ge-'
hence
(9.36)
from which we see that Det (T~ A) = 1. Orthogonal point transfor-
mations are called rotations (reflexotations) if Det (T~_.) is equal to
+ 1 (-1).
An infinitesimal point transformation is a transformation of the
form A + F~_. dt. Its inverse is A- F~A dt. Hence an infinitesimal
point transformation is orthogonal if and only if F";. is a bivector. Of
course such a transformation is always a rotation. If the fundamental
tensor is definite, F"-' can be split up into ! r mutually perpendicular
blades, r being the rank of F";.. From this it follows that in this case
the transformation consists of l r rotations in the i r R 2 's of the blades,
leaving these R 2 's and the Rn-r perpendicular to them invariant. The
Rn-r plays the part of an "axis of rotation". Hence, if n is odd, there
is always such an "axis".
In the indefinite case with index s we consider two real flat mani-
folds R 8 and 'R 8 through the origin 0 both of which lie wholly in the
- -region and who intersect in an R,. Let a hypersphere with radius 1
48 I. Algcbraic preliminaries.

and centre in 0 in R, be projected orthogonally on R,. Then we get a 1

quadratic hypersurface in R.. It always has a set of s principal axes

1

and these are the projections of s mutually perpendicular unitvectors

in R, exactly z of which lie in R,. The s angles between these unit-
vectors and their projections are called the principal angles between R.
and R,. Exactly z of them are zero. We now choose (h) in such a
1

way that i", ... , i" lie in the direction of the s unitvectors and i", ... , i"
1 s 1 z
lie in R,. Then if P", a = 1, ... , s is the projection of i" we have the
a a
equations
(9.37) i" = P" + q"; a = 1, ... ,s
a a a

where q", ... , q" are zero and q", ... , q" are perpendicular to R. and 1

1 z z+1
consequently lie in the + -region. Now all non-vanishing vectors P"
a
are mutually perpendicular and also perpendicular to all vectors q".
a
Hence, according to (9.37)

(9.38) o= (P" + q") (P .. + q") = q" q,.; a, o= 1, ... , s; a =l= 6

a a b b a b
and from this it follows that the s - z planes of projection of i" , ... , i"
z+1 s
are mutually perpendicular. All these planes have a minkowskian
geometry and the length of P" is I cos rp I;;;;: 1, where rp is the angle bet-
a a a
ween p" and i". Hence the principal angles between R, and R. are all 1

a a
imaginary or zero and R, can be transformed into R, by a continuous 1

rotation consisting of s - z real minkowskian rotations in the s - z

planes of projection of i" , ... , i". Note that during this rotation R,
z+1 s
remains in the - -region and that i" is transformed into p" cos- 1 rp.
a a a
If originally an s-dimensional screwsense is given in R. the process
of orthogonal projection fixes a screwsense in 1R 5 We call two screw-
senses in R. and 1R, the same if each is the orthogonal projection of the
other and opposite if not. But we still have to prove that if R,, 1R 5
and 11R, alllie in the -- region and each is provided with a screwsense
and if the screwsenses of R. and 1R 5 and also those of 1R, and 11R, are
the same, then the screwsenses of R, and 11 R, are also the same. The
screwsenses in R. and 11 R, are both orthogonal projections of the screw-
sense in 1R 5 But as we have seen already, R, can be transformed into
11 R. by a real rotation. Since this process is continuous the projection
of the screwsense in 1R, on the moving R. during its way to 11R, can
never change except if there is a moment where there is no screwsense
 9 Introduction of a metric in an En. 49

at all, i.e. where Rs contains a direction perpendicular to 1R 5 But at

that moment Rs would contain a direction in the +- region. This is
impossible. Hence it is possible to fix a screwsense in all real R;s in
the - -region simultaneously by fixing a screwsense in one of them.
The same can of course be proved for the real Rn_;s in the + -region.
N ow let 1 i", ... , 1i" be unitvectors in the direction and with the sense
1 s
of P", ... ,p":
1

(9.)9) 1
i" j cos rp I = P''; a = 1, ... , s .
<l <l <l

We know that
(9.40) s1
' 1
1 i[" ' ' 1 si"'l 11[><,'''
i 1i
s"']
= (-1) 5

and from this and (9.37) it follows that

s.I z'[>< ... z'><]1'

t[,., . . . Iz",J
'

{ 1 s 1 s
(9.41) -- S .1p[" . . . p"J~Z[><, ~t><,]- - (-1)' jcosrp ... cosrpl
1 s 1 s z+1 s

and in this formula instead of i" and t"' we may take two arbitrarily 1

<l <l
chosen sets of s mutually perpendicular unitvectors in Rs and 1R 5 ,
provided that the screwsenses fixed by their orders are the same in the
sense defined above.
Now let a coordinate system (h) be chosen in such a way that
i", ... , i" lie in an arbitrary way in R 5 and let there be another coordinate
1 s
system (h 1) suchthat i", ... , i"lie in an arbitrary wayin 1R 5 Then the
1' s'
transvection
(9.42) ( - 1) s S.I Z
'[><,
... t
"><]
Z[>< Z,.]
1 s 1' 1 s' 8

is ~ 1 (:::;;; - 1) if the two screwsenses fixed in R. and 1R. by the order

of these two sets of unitvectors are the same (opposite).
If P",. represents the real point transformation belonging to the
coordinate transformation (h) -+ (h 1), i.e. the transformation trans-
forming i", ... , i" into i", ... , i", we have
1 n 1' n'

(9.43) P".Aii;.=i"
i' '
i=1, ... ,n; i 1 =1 1 , ,n 1

or
(9.44) P~i j,h .* Af,.
Schonten, Ricci-Calculus, 2. Auf!. 4
 (l
so I. Algcbraic prcliminaries.

Accordingly (9.41) can now be written as

- 1) 5 sl, 1iU , , sii,J P"
J1 ,
P"' i
}s 1 hl
i
, s hs
1 s
(9.45) -- s.I i'[f, ... !'is] ph,.i, . . . phs. is 1'h, ... 1'h.
* s! P[.1 [ 1 ... P:ls 1 =Det(P~b); a,6=1, ... ,s.
Hence the screwsense in Rs and its transform in 'Rs are the same (oppo-
site) if and only if Det (P~b) ;:;::; +
1 (;;:;: - 1). But we still have to
prove that if, instead of Rs, we take another Rs in the -- region not
in any way connected with the coordinate system (h), Det (P~b) ;::::;1 is
also the necessary and sufficient condition for the screwsense in this
new Rs to be the same as its transform. Let all R/s in the - -region
be provided with a screwsense and let all these screwsenses be the same
in the sense defined before. Now let P~;. be applied to all these R/s
and let there be, among all transforms, one with the same and one with
the opposite screwsense. Because every Rs in the - -region can be
transformed into every other R 5 , with the same screwsense, in that
region by a continuous rotation, there must be a transform where the
screwsense changes and this transform could not have any screwsense
at all which is impossible. Hence we may state the theorem 1):
lf P~;. represents a real orthogonal Iransformation in an Rn with index s
and if a real coordinate system (h) is chosen in such a way that i", ... , i"
1 s
lie in the - -region and the other unitvectors in the + -region, P";. trans-
forms every real Rs, with a given screwsense in the -- region into an Rs
with the same ( opposite) screwsense in that region, if the subdeterminant
+
Det (P~b); a, 6 = 1, ... , s is ;:;::; 1 (;;:;: -1) and every real Rn-s with
+
a given screwsense in the -region, into an Rn-s with the same ( opposite)
screwsense in that region if the subdeterminant Det (P:!JJ; !, t) = s +1, ... , n
is;:;::; +1 (;;:;: -1).
If we call a real orthogonal transformation - -reflexional ( -re- +
flexional) if it changes the screwsense of a real Rs (Rn-sl in the - -region
(+ -region), we have the following possibilities for real orthogonal
transformations in Rn
A) non + -reflexional; non -- reflexional: proper rotation, L1 =+ 1
B) non + -reflexional; - -reflexional: reflexotation, L1 = - 1
C) + -reflexional; non - -reflexional: reflexotation, L1 = - 1
D) + -reflexional; - -reflexional: improper rotation, L1 = + 1.
The group of all real orthogonal transformations splits up into
four sets of transformations. Obviously a transformation of one set
cannot be transformed by a continuous proper rotation into a trans-
1) Cf. BRAUERand WEYL 1935, 1; GIVENS 1940, 1 gave an elementary proof.
 10. Hybrid quantities. 51

formation of another set. From these sets only the set (A) and the
four sets tagether form a group.
Exercises.
I 9, 1. Prove that for a definite fundamental tensor
I 9 ' 10t) I ""PI'II'n-p JA, . ).p1' l'n-p = p I (n - p) I g(><,[J., g,.p]i.p] 1)
I 9,2. An Ep and an Eq, p :;:;; q, in En were defined in I 1 to be
tjp-parallel if the Ep contains an E 1 and no E 1+1 parallel to Eq. 2 )
An Ep and an Eq, p :;:;; q, in Rn are said to be tjp-perpendicular, if
the Ep contains an E 1 and no E 1+1 perpendicular to Eq. 3 )
If v""P and w""f aresimple multivectors in Ep and Eq prove that
Ep and Eq are tjp-parallel if and only if

I 9,2 oc) v" .. "" [A, .. Ap-u w .. vq] { =f= 0 for u = t

=0 for u = t -1.
Prove that Ep and Eq are tjp-perpendicular if and only if
for U=t
I 9,2 )
for U=t-1.
I 9,3. The real symmetric tensors P;.,. and Q",. in an ordinary Rn
have the same principal multidirections if and only if
I 9,3 oc) Pr~" QpJJ. = 0.
I 9 4. 4) If g. is definite and v" ""real the equations v" "" vX 1 , ,,
I AX I Yp

= 0 and v" ",. = 0 are equivalen t.

10. Hybrid quantities.

Every set of N independent functions of the components of a quan-
tity with N components could be considered as a kind of components
of this quantity. But generallysuch components arenot used because
their transformation formulae are too complicated. There is one excep-
tion. In dealing with a quantity with not only real components, for
instance a vector v", the complex conjugates of the v" may be considered
as another set of components of the same vector. But we may also
look upon them as components of another quantity, a quantity of the
second kind in contradistinction to the ordinary quantities that will
be called of the first kind. If we write v" for these complex conjugates,
they transform as follows
(10.1)
1) LIPKA 1922, 1, p. 243.
2) ScHOUTEN 1902, 1, p. 34.
3) ScHOUTEN 1902, 1, p. 49.
4) N ARLIKAR 1938, 1 for p = 2.
4*
52 I. Algebraic preliminaries.

where the A~' stand for the complex conjugates of the A:'. In the same
way to every tensor for instance P~~JL a tensor of the second kind with
components P~:ii = P~~JL can be constructed. But it is not possible to
construct in this way components P~~JL' because P~:JL can be written
in several ways as a sum of products of vectors and these several ways
would lead to different values for P~~w Notwithstanding this it has a
definite meaning to consider for instance a quantity with components
Q~;.JL transforming as follows

(10.2) Q;;,, , = A;;;;p.

-,, ' Q;; ;- .
A J1. ><A J1. AJl.

Such quantities we call hybrid. 1) The indices with a bar are called indices
of the second kind in contradistinction to those of the first kind without
a bar. Hybrid ..1-densities can be defined in the same way as ordinary
..1-densities. If they get the factor ..1-w J-w', w is the weight and w'
the antiweight. If w = w' we have an ordinary density of weight 2w.
Hence ordinary densities as defined in 2 are really a kind of hybrid
quantities. Ordinary rensors and their complex conjugates may be
considered as special cases of hybrid quantities but as a rule we use
this latter term only if there are really two different kinds of indices.
To a co- or contravariant hybrid quantity of valence 2 for instance
T;;.A there belongs the complex conjugate T"J... Now if
(10.3)
T";. (and its conjugate) is called a hermitian symmetric tensor or her-
mitian tensor for short. But if
(10.4)
T;;.A (and its conjugate) is called a hermitian alternating tensor or a
hybrid bivector. The difference between hermitian tensors and hybrid
bivectors is not so great as in the ordinary case, because if pA is her-
mitian, i pA isahybrid .bivector. The property is obviously invariant for
coordinate transformationsandalso for multiplication with areal scalar.
A mixedhybrid quantity of valence 2, V~.. has the complex conjugate
-1 -1
V~;; If its determinant is not zero the inverses V~;;. V~ .:t exist. Now if

(10. 5)

Ar.
1) SCHOUTEN 1929, 1; ScHOUTEN and VAN DANTZIG 1930, 1; EI 1935, 1, p. 8.

In these publications we wrote vi and f instead of vi and This is useful

if at any time we wish to drop the indices as for instance in matrix calculus. Be-
cause otherwise, where the indices had been dropped, we could no longer make a
difference between a quantity and its complex conjugate. But as lang as we have
no intention of dropping indices and no ambiguity arises we can simplify all nota-
tions considerably by dropping the bar over the kernels. It can always be reintro-
 10. Hybrid quantities. 53

V~ A (and its conjugate) is called positive or negative invertible. 1) This

property is invariant for coordinate transformations and also for multi-
plication with a factor of the form e'"' where cp is real. 2)
For hermitian tensors the following theorem holds 3 ) (cf. I 8):
If r is the rank of TA;;.= T;;,;. (U;;,;. = uAx), the coordinate system can
always be chosensuchthat the matrix of TA;;.(U;;,;.) has the diagonal form with
s numbers -1, r- s numbers + 1 and n- r numbers 0 in the main diagonal.
The nurober s, called the index, is an invariant of the hermitian
tensor. The sequence - - - + + + with s-- signs and (r- s) +-
signs is called the signature. The signature is said to be even (odd) if s
is even (odd). The hermitian tensor is called positive (negative) definite
if s = 0 (n) and r = n and indefinite in all other cases. It is called posi-
tive (negative) semi-definite if n-r>O and s=O(r) (cf. I 8).
As a corollary we get that two hermitian tensors are equiform
(cf. I 2) if and only if they have the same rank and the same index.
In a centredEn a hermitian tensor TA;;. is represented by a subspace
satisfying the equations
( 10.6) T;.;;, x-' x;;, = 1.
The nullcone of T;.;;, has the equation
A -
(10.7) T;.;;, x x" = 0.
If T;.;;, is definite its nullcone degenerates into the origin because in
this case there are no other points satisfying (10.7).
In dealing with hybrid quantities it is often convenient to consider
an auxiliary E 2 n with coordinates x", x;; that are now supposed to be
linearly independent and each to take all complex values. The allow-
able coordinate transformations in this E 2 n are the linear homogeneous
transformations of the x" and x;; allowable in the originalEn, hence the
E 2 n has a restricted transformation group. It contains the original En
as a subspace given by the n equations
(10.8)
This subspace, called the principal En, is invariant for the allowable
coordinate transformations and the same holds for the subspaces re-
duced where necessary (cf. T. P. 1951, 1, Ch. X, also for the shorter notations of
matrix calculus arising by dropping all indices of quantities with valences 1 and 2).
1 ) Canonical forms for V~_. were given for the general case by HAANTJES 1935, 1
and for the invertible case by JACOBSTHAL 1934, 1. Cf. VEBLEN, V. NEUMANN and
GrvENS 1936, 1, p. 4. 23; ScHOUTEN 1949, 3, p. 218.
2) Invertible quantities were introduced by ScHaUTEN and HAANTJES 1935, 2,
p. 1 79 and by VEBLEN, v. NEUMANN and GIVENS 1936, 1, p. 4.23. For the projective
case these latter authors called them anti-involutions. They are important in
spinspace. Cf. ScHaUTEN 1949, 2; 3; 4; 1950,1.
3 l Cf. for the proof DrcKSON 1926, 1, p. 70.
54 I. Algebraic preliminaries.

presented by each of the two sets of n equations

(10.9) a) x" = c" = const. b) x" = c" = const.
1 2
Each set of equations represents a set of oo" E,.' s called theinvariant E,.'s
of the first and of the second set respectively. The x" can be used as
coordinates in every E,. of the first set and the x" play the same role
for the E,.'s of the second set. Every E,. of one set is mapped on each E,.
of the same set by means of the E,.'s of ihe other set. According to
KLEIN's principle instead of x" and x" we may now also introduce
general rectilinear coordinate systems subject to the full group G11 a
in E 2 ,., provided that we adjoin the two sets of invariant E,.'s and the
principal E,. (10.8).
Instead of an ordinary symmetric tensor gJ.,. a hermitian tensor aM,
of rank n can be introduced as fundamental tensor. Lowering and
raising of indices can then be effected by means of a;;< and its inverse
denoted by a"J.. The determinant of aJ.i< is a scalar density of weight + 2.
An E,. with a fundamental tensor aJ.i< is called an R,. 1) and it is said to be
ordinary if aJ.;; is positive definite. In an R,., v" and v" each get two kinds of
components, viz. v", vr and v", vJ.. So in R,. the difference between co-
and contravariant vectors does not vanish, as it does in R,.. There
remain in R,. two kinds of vectors, called kets and bras by DIRAC. The
ket (bra) is a contravariant vector of the first (second) kind, equivalent
to a covariant vector of the second (first) kind 2). Two vectors v" and w"
are said to be ( unitary) perpendicular to each other if vA w" aJ.>< = 0.
This is equivalent to wJ. v" aJ.i< = 0. The norm of a vector v" is the
expression vJ. v" aJ.,. that is always real. The vectors with a positive
(negative) norm fill the + (- )-region of the R,.. A vector is called
unitvector if its norm equals 1. A unitvector remains unitvector if
it is multiplied by e"' with real q;.
According to the theorem stated above there exists at least one
coordinate system (h) with basis vectors ff"; ~J. suchthat

aJ.,. = -uJ.ui<- -u.. u,. + uJ. ~t;; + +u.<u;;.
1 1 s s s+1 s+1 n n
(10.10)
II
The u.< are mutually perpendicular unitvectors and they satisfy the
equations i i -
U:" = -u" = -u;;a"" i = 1, ... ,s;
{
(10.11)
~" = +u" = + u;;aJ."
i i -
i=s+1, ... ,n.
They form a unitary cartesian coordinate system. All these systems
are transformed into each other by the transformations of the unitary
1) In EI 1935, 1, p. 60 we wrote U,. instead of R,. (cf. Ch. VIII).
2) DIRAC 1947, 1; cf. T. P. 1951, 1, p. 243.
 10. Hybrid quantities. 55

orthogonal group. This is the group of homogeneaus linear transfor-

mations characterized by the invariance of a;.;, (cf. 9-36):
-1 -1
(10.12) a) Ulf;.U~"aea=a;.-;; or b) Ue;;=U;;e or c) U~;.=U;.".

From this we see that Det (U~;.) has the form ei"' with real cp and
that if U~;. represents a unitary orthogonal transformation, the same
holds for ei"' U~;. with real "P 1)
The subspace represented by
(10.13)
is called the fundamental figure (the nullcone) in the Rn. But (10.13)
can also be interpreted as the equation of a hyperquadric in the auxiliary
E 2 n that makes this E 2 ,. to an R 2 ,.. In this R 2 ,. the a;.;; are the com-
ponents of a fundamental tensor whose components a;." and a;:" are
zero. Because of this the hyperquadric has a very peculiar position
with respect to the two sets of invariant En's. In fact, the section of
the hyperquadric (nullcone) with an En of the first set is an E,._ 1 with
+
the n 1 linear equations
(10.14) a) x"=c" b) 2a;."x;.c"=l (0)
1 1
in which n independent parameters occur. Hence the hyperquadric
is built up of two sets of oo" E,._ 1 's and each E,._ 1 of one set has at most
one point in common with an E,._ 1 of the other set. 2)
If 'I;." is an arbitrary hermitian tensor, we may ask whether any
vectors exist satisfying the relation
(10.15)
a is a solution of the equation
(10.16) Det (T;.;:;- aa;.;:;) = 0.
If a;." is definite, it can be proved that the solutions are all real
and that to every solution with multiplicity m there belongs an m-direc-
tion perpendicular to all multidirections belanging to other solutions.
1) Instead of a;.;; a positive or negative invertible hybrid quantity (cf. 10.5)
could be introduced as fundamental quantity in E,.. This would Iead to a geometry
with a group leaving this quantity invariant. This geometry seems to have
escaped attention and even its group seems not to have got a name.
(n+1)
2) It is weil known that a quadric in E 2 ,. contains oo 2 E,._ 1 's. But a quadric
in E 2 ,. does not fix two preferred sets of E,.'s. This means that the E 2 ,. is not
an ordinary R 2 ,. but an R 2 ,. that contains besides a quadric also two sets of in-
variant En's. If these E,.'s are fixed it contains moreover the principal E,. with the
equation (10.8). This is the real reason why there is something new in the geometry
of an R,. and why this geometry is not the ordinary geometry of an R 2 n described
in a queer way.
56 I. Algebraic preliminaries.

These multidirections are called the principal multidirections of the

hermitian tensor. 1 ) The unitvectors of (h) can be chosen in these multi-
directions, hence (cf. I 9)

Theorem of principle axes of a hermitian tensor in Rn:

Ij 1';.." is a hermitian tensor and ij the fundamental tensor is definite
there always exists a unitary cartesian system (h) such that
(10.17) T;h * 0; h =f= i.
The axes of this coordinate system are called principal axes of the
hermitian tensor. They are uniquely determined if and only if the
solutions of the equation (10.16) are all different. The theorem of
principal axes does not hold if the fundamental tensor is indefinite
because in this case the tensor T;,-x may possibly have a special position
with respect to the nullcone. If this is not the case the theorem remains
valid.
Exercises.
I 10, 1. The infinitesimal transformation
I 10,1 ~)

belongs to the unitary orthogonal group if and only if P";. is hermitian

alternating. Then iP;;;. is hermitian symmetric and every principal
direction of this tensor is invariant for the transformation.
I 10,2 2). If I;,." is hermitian symmetric (alternating) and Sj." arbitrary,
prove that
I 10,2 ~) sa T-aQ S:Jfx
is also hermitian symmetric (alternating).
If T~;. is positive (negative) invertible and 5~" arbitrary, prove that
-1
I 10,2 ) Sj_a T~a Sl
is also positive (negative) invertible. If 'x" = S~" x" is a transformation
of the unitary orthogonal group, in both formulae S~ii can be replaced
-1
by s~".
I 10,3 2 ). In an ordinary R 2 the most general unitary orthogonal
transformation can be written in the form

I 10,3 {
'x = xeirJ.coscp + yeisincp
x eir sin cp + y ei(y-CJ.+) cos cp
~)
'y = -

with four real parameters ~. , y, cp with respect to a unitary cartesian

coordinate system.
1) Cf. e.g. DICKSON 1926, 1.
2) Cf. EI1935, 1, 5.
 11. Abridged notations. 57

11. Abridged notations.

In tensor calculus there sometimes appears a great number of indices.
Many authors therefore have tried to get rid of them by introducing
abridged notations. Most of these attempts failed because an abridged
notation always means a new kind of stenography that has tobe learned
by the reader, and because most readers are overbusy and seldom willing.
It should be held in mind that as a rule abbreviations make the reading
more difficult and the calculus less foolproof. Authors should therefore
be very careful and they should introduce them only when necessary.
We mention here some of the most interesting abbreviations.
1. Collecting indices. Some of the indices of a quantity can be
replaced by one upper or lower index. For instance P~. or PA. may be
written for P~~"'.. Even transvections over collected indices can be
made possible, for instance P~. Q i!a for P~~"'. Q~ ii?"'a. This abbreviation
can be very useful if during an investigation or even during some part
of it only the indices not collected are important. It is often used and
there seems to be no objection provided its use is not exaggerated.
2. Representative indices. Let Ap Aq where q< p, always stand for
Ap ... Aq+I Aq ... A-1 . 1 ) Then we are able to write P[;vlA for P[.<p ... -'+ 1lA ... -'.
This notation is useful in investigations where repeated differentiations
occur often, for instance o_,v w_,. stands for o_,v ... 8-'+ 1 w_,... -'. But it
needs to be handled carefully and this may be the reason that it has not
been accepted. The same holds for the notation vx.1P-'Iv for v~ .. ~v-<.... .<q 2)
that has the same disadvantages and is less practical. A less developed
Version of this latter notation that has come out recently 3) is to write
P~ _,_ .. instead of P: "; " .. ;.. . If this version should find approval this
would demonstrate the fact that only very uncomplicated abbreviations
have any chance of being accepted.
3. The notation of GIVENS. 4) GIVENS proposed to write (A.) instead
of p alternated neighbouring indices [A-1 ... Ap] wherever they might
occur and I(A.) I for their number p. This notation makes it easy to
write down transvections of multivectors. Though GIVENs' results
were interesting bis notation was not accepted.
4. VITALI's method. VITAL! 5) introduced a kind of indices that can
take the values 1, ... , n, 11, 12, ... , 1n, 21, 22, ... , 2n, ... , nn,
111, ... , nnn, . . . and so on till all combinations (with repetition)
from the numbers 1, ... , n. The notation is to be used in connexion
1) SCHOUTEN 1925, 1.
2) EI 1935, 1, p. 33.
3) CRAIG 194 3, 1.
4 ) VEBLEN, V. NEUMANN and G!VENS 1936, 1; GIVENS 1937, 1.
5) VITAL! 1923, 1.
58 I. Algebraic preliminaries.

with higher differentiations. Though strongly advocated by BoRTO-

LOTTI1) it never came in general use. The method is outside of the realm
of this book.
5. CRAIG'S extensors. Another notation, also connected with higher
differentiations is the method of extensors introduced by CRAIG. 2)
Though the double-indices used are in the beginning somewhat be-
wildering the notation is often used especially by japanese authors
and it see~ to have a fair chance of being accepted for those investi-
gations where higher differentiations play an important role. A. KAWA-
GUCHI3) and M. KAWAGUCHI 4) gave generalizations. The method is out-
side of the realm of this book.
6. Ideal vectors (cf. III 4). A tensor, for instance P~;., is generally
not a product of vectors. But we may write P~ ;. = u,. v,. w and look
upon u,., v,. and w as symbolic factors. If this process is applied to
a symmetric tensor the factors can be chosen equal, for instance w;, ... Aq =
w(A, ... Iq) = w,., . w 1.q and the transvection wA, ... Iq x 1' ... x.<q can then be
written (wA xA)q. So we get the CLEBSCH-ARONHOLD symbolism well
known from the theory of algebraic invariants. If w,., ... q is a q-vector
we may write w;, ... Aq = wA, . w,q if we assume that w,., wA, = - wA, w,.,
for every choice of ~ and .A.2 Then we arrive at WEITZENBCK's "Kom-
plexsymbolik ''. 6 ) Of course in both cases in all terms that contain a
non-symbolical factor more than once, separate and distinguishable
ideal vectors must be introduced for every factor in order to avoid
ambiguity. This is all well known from algebra. Now both symbolisms
are very useful in purely algebraic investigations as so many publi-
cations on algebraic invariants prove and they can of course be equally
useful in tensor algebra. But as soon as differentiations are introduced,
as in differential geometry, many difficulties arise with respect to the
differentiation of ideal vectors. In some earlier publications 6) an attempt
was made to introduce ideal vectors in differential geometry. But the
simplifications to be got in this way did not appear important enough
to compensate for the introduction of symbolic differentiations of ideal
vectors and so ideal vectors never got a proper place outside of the
purely algebraic domain.
7. Systems of "direct" calculus. An invariant equation between
tensors does not depend on the choice of the coordinates or the letters
1) BoRTOLOTTI 1930, 1 ; 193 7, 1 ; also for more literature.
2) CRAIG 1943, 1.
3) A. KAWAGUCHI 1939, 1.
4) M. 1952, 1.
KAWAGUCHI
5) WEITZENBCK 1908, 1.
6) ScHOUTEN 1918, 1; 1921, 1; ScHOUTEN and STRUIK 1919, 1; 1921, 2; 3;
1922, 1; 2; STRUIK 1922, 1. Cf. III 4.
 11. Abridged notations. 59

used for the indices. It is a roessage froro the author to the reader con-
cerning certain properties of tensors independent of any coordinate
systero and this roessage is readable as soon as we know the skeleton
of the forroula that is the corobination of the kernel, the places of the
indices for every roanifold concerned and the places where the trans-
vections act. Hence it roust be possible to invent notations that represent
in soroe way this skeleton without the use of indices. Such a notation
is often called a "direct" systero of calculus. Well known exaroples
are GRASSMANN's "Ausdehnungslehre ", HAMILTON's quaternion calculus,
ordinary "vectoranalysis ", "tensoranalysis ", roatrix calculus etc. A
direct systero roust have a notation for the kernels, roostly a letter
of soroe special alphabet, in order to distinguish between tensors and
scalars, a special roethod of roarking the places of upper and lower
indices that roakes it possible to drop these indices and a nurober of
special signs for the roost frequently used roultiplications and trans-
vections. For instance in the roost developed direct systero 1) the vectors
v" and w,~ were written v' and w and the tensor P~ ~ pv was represented
by P''". The transvections v"w,., P~;." w., w" P~;." were written
v' 1w, P''" 1w, w 1P''" and other roultiplication signs were given for
a nurober of other transvections. This systero worked rather efficiently
and it was used not only for the translation of well known results (as
roany of these systeros were) but especially for the establishroent of
new theoreros. Froro the beginning there was a great difficulty, not-
withstanding the great nurober of roultiplication syrobols the use of
ideal vectors for the representation of all necessary transvections could
not be avoided. For instance w"P~ ~" could only be represented by
(w 1r') p' qs' after introduction of ideal vectors: P' " = p' qr' s'. Now in
view of this difficulty one could get the idea to iroprove the direct
calculus by introducing instead of the two signs and ' to roark the
places of the indices, as roany signs as there are places. Then for in-
stance a tensor P~ ;." could be written P:
xo * and it would be easy to
P:
write allpossibletransvections, forinstance W, X *; woP: ~ *; ~ * yX
0 0
P: 0

without the necessity of introducing ideal vectors. This would certainly

be an iroproveroent but then the next step would be the reroark: "why
not "'AflV instead of , x o *?" and this would lead us back to RicCI's
calculus. The real cause of this reroarkable situation is that tensor
calculus has two different aspects. The running indices are on the one
hand only roarks labeHing certain places and constituting tagether with
the kernels the skeleton of the forroula. Froro this point of view tensor
calculus is a highly efficient "direct" calculus because it enables us
1) ScHOUTEN 1918, 1; STRUIK 1922, 1. In ScHOUTEN and STRUIK 1922, 1
(the first edition of EI 1935, 1) all formulae were given in duplo, once indexfree
and once with indices.
60 I. Algebraic preliminaries.

to represent all kinds of multiplications and transvections without the

help of any auxiliary device. But on the other hand as soon as a definite
coordinate system is chosen, the running indices may be replaced by
fixed ones, for instance by 1, ... , n and if this is clone the equations
can be read as equations between components. The "directness" of
the calculus lies in the fact that the formulae can be read (and should
be read) as relations between quantities and not between components.l)
The process leading from the direct formulae to the formulae in compo-
nents (in most kinds of "direct" calculus rather a difficult process !)
happens automatically here because the same formulae can also be
read (and should be read were desirable) as relations between com-
ponents. Of course this is only exactly true if the kernel-index method
is used, that is, if every object gets its own kerneland if every coordinate
system gets its own kind of running indices,- for instance "' "'' h with
their corresponding fixed indices, 1, ... , n; 1', ... , n'; 1, ... , n.
The greatest difficulties for all systems without indices lie in the
construction of the necessary equivalent for the places of these indices
and in the representation of all possible multiplications and transvec-
tions. These difficulties increase strongly if the valences of the quantities
occurring become higher. Therefore indexfree systems only succeeded
for valences 0, 1 and 2. In fact vector analysis in R 3 with valences 0
and 1 only, and matrix calculus 2) in Rn with valences 0, 1 and 2 only,
arerather satisfactory. But the vector analysis that can be constructed
for E 3 or R 4 is already much more complicated and systems of this
kind for En and Rn are very amusing for the inventor hirnself and
perhaps for some of his friends but they never will find general appli-
cation.3)
8. The method of the radiusvector in En and Rn. For differential
geometry in En and Rn often a middlecourse between tensor calculus
and indexfree calculus is followed by writing r (instead of r") for the
radiusvector and conserving all other indices. In this way quantities
1) All students of tensor calculus should be encouraged from the beginning
to try to read all invariant formulae as formulae between geometric objects and
not as formulae between components only.
2 ) Matrix calculus in connexion with tensor calculus was dealt with elaborately
in EI 1935. 1, p. 35ff. andin T. P. 1951, 1, p. 39ff. and Ch. X (DIRAc's matrix
calculus).
3 ) The present author has some experience in this matter. He wrote his paper
1922, 3 on different kinds of connexions in differential geometry, in his own index-
free system. But following the advice of F. KLEIN who warned him that the paper
would never be read, he translated it in terms of Rrccr calculus. The advice was a
good one, the paper was much referred to and it had its influence on the develop-
ment of modern differential geometry.
II. Analytic preliminaries. - 1. The arithmetic n-dimensional manifold 2!,.. 61

of valence p derived from r may get p -1-indices. The method is

ccrtainly efficient and it is very often used, of course only in euclidean
spaces.
9. CARTAN's method. Though from the point of view of Rrccr cal-
culus CARTAN's method could be considered as an abbreviation we prefer
to consider it as a full sized method in its own right. It is not necessary
to discuss it here because it will be dealt with elaborately at many places
in this book.
10. Abbreviations ad hoc. Sometimes it is convenient to introduce
for a short time during an investigation some abbreviation that is
specially adapted to this part of the calculation but will be dropped
later on. For instance in II 8 we write Q with suppressed indices
for any quantity and -o for any operation of the kind considered there.
This is certainly justified because only in this way could such a general
formula be written down. So abbreviations ad hoc may have their
advantages. But if a book contains too many of them it becomes un-
readable. Y oung authors especially, who can not yet foresee all the
consequences, should abstain as far as possible from introducing ad hoc
abbreviations. It is inexusable to spend valuable time on the writing
of a paper that will not actually be read.

li. Analytic preliminaries.

1. The arithmetic n-dimensional manifold ~n. 1)

A set of n real or complex values of n ordered variables $" is called

an arithmetic point and the totality of all these points an arithmetic
manifold ~n. The l are called the components (Bestimmungszahlen)
" means "point with components $" ".
of the arithmetic point. "Point$"
The set of all arithmetic points satisfying the inequalities

(1.1) "
l$-~l
"
" <,
0
" are n arbitrarily given
where ;" are n arbitrarily given numbers and
0
positive numbers, is called a polycylinder (or box) of ~n
If a set of arithmetic points is given, satisfying the conditions:
1. every point of the set belongs to at least one polycylinder con-
sisting of points of the set;
1 ) For 1 and 2 of this Chapter cf. VEBLEN and WHITEHEAD 1932, 1; BEHNKE

and TRULLEN 1934, 1 ; BOCHNER and MARTIN 1948, 1.

62 II. Analytic prcliminaries.

2. for two points of the set there always exists a finite chain of
polycylinders, all consisting of points of the set, such that the first
point lies in the first and the second point in the last polycylinder and
that consecutive polycylinders have at least one point in common, this
set is called a region of 2rn. Every polycylinder is a region and the whole
2rn is a region. But not every region is a polycylinder. Every region
"
is called a neighbourhood of each of its points. For "neighbourhood of ~"
we write ~(i).

2. The geometric n-dimensional manifold Xn.

Let the elements of a set M be in one to one correspondence with
the points of a region ffi of 2rn. We make no presuppositions concerning
0
the "nature" of these elements. They may be, for instance, homogeneaus
linear forms in n variables or polynomials of degree n- 1 in one va-
riable but they may be also just elementsnot connected in any way with
other notions. We call the one to one correspondence between the
elements of M and the points of ffi a coordinate system defined over M.
" 0 "
If the point ~ corresponds to a certain element of M, we call the ~ also
coordinates of this element with respect to the coordinate system (x).
If the ~" are considered in this way we write ~" instead of ~" in order to
emphasize this fact.
We consider now a point transformation in 2{n
(2.1)
". "
satisfying the conditions that the f are analytic 1 ) at ~ and that the func-
o
tional determinant
(2.2) (), def~
A- ).

8~

is =!= 0 for ~"" Then it is a well established fact that there is an in-
= ~.
o
verse transformation
(2-3)
1) " is said to be analytic at ~" if there exists an
A function defined in an \n (~)
X O X X O
\n (~) where it can be expanded into a power series in ~- ~. convergent in this
o x o x'
latter \n (.;'). Instead we may only require that the functions f ace of class u in
0

" A function is said tobe of class u in some region if it is continuous and

some \n(~).
0
has continuous derivatives up to the order u at each point of that region. This
case has been dealt with by VEBLEN and WHITEHEAD 1932, 1, p. 36.
".
 2. The geometric n-dimcnsional manifold X". 63

with functions analytic at {

o
det J( )0
and that there exists a neighbour-
" ~
hood :li of ~ and a neighbourhood ffi' of ~ for the points of which (2.1)
0 0
and (2-3) establish a one to one correspondence. Suppose now that ffi
lies in ffi. Then there is a subset R of M whose points are in one to one
0
correspondence with the points of ffi and also with the points of ffi'.
But this latter correspondence is another coordinate system (x') over R.
N'

For the ~. considered as Coordinates of the elements of R, we write ~"'.

in SUn:

Fig. 9.

Fig. 9 1 ) illustrates what is going on in ~n and in M. In ~n we have

only one kind of variables and a transformation means a point trans-
"' x' x x'
formation ~~~; ~~~; ffi~ffi'. In M we have two kinds of variables,
0 0
the coordinates ~" and ~"' and the transformation

~"=lW');
" Det (A~') =!= 0;

corresponding to (2.1, 3) transforms the coordinates only and leaves

the eleme:nts of M unchanged. As we see, our notation agrees with
the principles of the kernel-index method (cf. I 1). In ~n the trans-
formation (2.1, 3) is considered as a point transformation and accordingly
the kernel changes. But in M the same transformation is a coordinate
transformation and accordingly in (2.4) the index changes and the
kernel is invariant.
Collecting results we see that a coordinate system over a set of elements R
is a one to one correspondence between the elements of R and the points
><'
1) It is not necessary that ~ lies in ffi.
0 0
64 II. Analytic preliminaries.

of a region of 12rn and that transformation of coordinates in R means passing

to another one to one correspondence between these elcmcnts and the points
of another region of 12rn.l)
Until now we had only the set R with a coordinate system over it
and the possibility of introducing other coordinate systems by means
of point transformations in 12(n Now everything depends on the choice
of these point transformations. If we allow all analytic transformations
(or transformations of dass u) with ffi = ffi' = ffi then these trans-
o
formations form a group (cf. I 1). But if we allow the set \l3n of all
transformations of points of 12rn that are analytic (or of dass u) each
in a certain region, but without the condition that these regions coin-
cide, the set does not constitute a group. Because, if ffi--+ ffi' and
ffi"--+ ffi"' are two transformations of \l3n, the second can only be per-
formed after the first if ffi' and ffi" have some region in common. Such
a set of transformations, satisfying the first condition for groups (cf.
I 1) only if the second transformation can really be effected after the
first is called a pseudo-group. 2), 3)
The set M, provided with the pseudo-group \l3n and with all allow-
able coordinate systems, viz. all systems that can be derived from one
originally specified system by transformations of \l3n is called a generat
geometric n-dimensional manifold or Xn. The elements of M are called
geometric points or points of Xn- In the following "point" means
always "geometric point" and "point ~~" means "point with Coordi-
nates ~" with respect to (u) ".
Instead of \l3n some subgroup or sub-pseudo-group of \l3n can be
taken. If for instance ffi = 12rn and if \l3n is replaced by the group Ga
0
or G110 we get instead of Xn, the En or the centredEn (cf. I 1). If we
take ffi = 12(n and the group G0 , of all orthogonal transformations or
0
the group G, 0 of all proper rotations we get the Rn or the Rn with a
fixed screwsense (cf. I 9). In all these cases the geometry is fixed
wholly by the choice of the group or pseudo-group of allowable coordinate
transformations. As was pointed out by F. KLEIN in 1872 4) the geometry
of a figure in some space is the theory of the invariants of this figure
1) VEBLEN and WHITEHEAD 1932, 1, p. 22. We consider here only ordinary
Coordinates. For supernumerary coordinates cf. R. S. 194 7, 1 ; P. P. 1949, 1,
Il 9.
2) VEBLEN and WHITEHEAD 1932, 1, p. 38.
3) For a discussion of the foundation of pseudo-groups cf. WAGNER 19 50, 1
PENSOV 1950, 1. NIJENHUIS 1952, 1, succeeded in replacing the pseudo-group
by a groupoid in the sense of BRANDT, by taking as the elements the transformations
of the neighbourhoods of points instead of the transformations of regions.
4) KLEIN 1872, 1.
 2. The geometric n-dimensional manifold Xn. 65

with respect to the underlying set of allowable coordinate transfor-

mations.1)
If we wish to make an En from a given Xn, we have to choose one
coordinate system and consider all coordinate systems that can be
derived from this chosen one by means of transformations of Ga. Then
the Xn provided with this preferred set of coordinate systems is an En
and the geometry of a figurein this En is, in Xn, the geometry of this
figure together with the set of all preferred coordinate systems. This
is an application of the principle of F. KLEIN 2 ):
If in any space with a group G1 the subgroup G2 is introduced, con-
sisting of alt transformations of G1 which leave a figure F1 invariant,
then the geometry of a figure F~ with respect to G2 is identical with the
geometry of the set of figures F1 , F 2 with respect to G1
Starting from Xn we took for F 1 the set of all preferred coordinate
systems in order to get En. Butthis is not always necessary, we could
also have chosen some figure from which these preferred coordinate
systems might be derived and that is uniquely determined by these
systems. For instance if we start from En in order to get Rn, we can
use for F 1 the set of all orthogonal coordinate systems, but \ve can as
well use (cf. I 9) thequadratichypersurface representing the fundamental
tensor.
The allowable coordinate systems in En are the rectilinear systems
used in Ch. I. This does not mean in any way that we arenot allowed
to introduce more general coordinate systems in En. But in doing this
we consider the En as an Xn in which the allowable coordinate systems
of En play the part of preferred coordinate systems. In the same way
in an Rn besides the allowable orthogonal coordinate systems, general
rectilinear or curvilinear systems can be used, but then we look upon
the Rn as an En or Xn with preferred coordinate systems.
In an Xn we can not use the notion of polycylinder because contrary
to mn there is no preferred coordinate system in Xn. Instead we use
cells, a celt being defined as a set of points given by the inequalities
(2.5) Wl <1
with respect to some allowable coordinate system. A pointset R in Xn
is called a region, if there exists an allowable coordinate system, deter-
mining a one to one correspondence between the points of R and the
arithmetic points of a region ffi in mn .
3) Every cell is a region, but not

1) Cf. for more general views on geometry ScHOUTEN and v. DANTZIG 1934, 1
and NIJENHUIS 1952, 1.
2) See footnote 4, p. 64.
3 ) A region in Xn can also be defined in the same way as a region in Wn by
using cells instead of polycylinders.
Schauten, RicciCalculus, 2. Auf!. 5
 66 II. Analytic preliminaries.

every region need be a cell. Also the Xn is itself a region and every
region in Xn is itself an Xn. Every Xn which is a region of another X,.
is said tobe imbedded in the latter. Every region in Xn is called a neigh-
bourhood of every one of its points. For "neighbourhood of ;"" we
write 91 W'). 0
0
If \n consists of analytical transformations only, a field p is said
to be analytic in the region R if for some choice of (") we have p = I(~")

and if l(i) is analytic at all arithmetic points of the region ffi of mn

corresponding to R by means of ("). Because in this case I is(/'(g))
analytic at all points of the region ffi' of mn corresponding to R by means
of ("'), the analytic nature of p is an invariant notion.l)
If \n consists of transformations of clas.s u an analytic function
in a region of Xn can not be defined. But with the same reasoning as
above we may call p a function of class u in some region R of Xn, if
with respect to some allowable coordinate system (") we have p =I(~")
and if l(g) is of class u in the region ffi corresponding to R by means
of ("). In. both cases the choice of the allowable coordinate system is
not important.
Exercises.
II 2,1. Prove that (cf. Exerc. I 1,1)
II2,11X)
-1
II 2,2 (cf. Exerc. I 8,1). If P" is a tensor of rank n, P" its inverse
and j its determinant (scalar density of weight +
2), prove that
-1
II 2,2 IX) P" o" P" = o" log j.
II 2,3 (cf. V 5). If h)."' is a tensor, prove that

II 2,3 IX)
is also a tensor. 2)
II 2,4. If the eigenvalues of h~" of Exerc. II 2,3 are all different
and if (h) is chosen in such a way that e", ... , e" are eigenvectors of h~",
1 n
prove that any p of them are Xp-forming (d. II 5) if and only if
H;;h=O; h,i,j-4:. .2)
1 ) In general in this book we only consider fields in a neighbourhood of a point
and not "in the large" though these fields play a very important role in "global"
differential geometry, cf. e.g. BocHNER 1950, 1; 1951, 1.
2) NIJENHU!S 1951, 1; 2.
 3. Geometrie objects and quantities in Xn. 67

II 2,5. If P";. and Q,.;. are tensors, prove that

II 2,5 IX) pp[><, 81' Q",><al + p[x, !I'! 81' Q"><,]- Q~'["1 81'P"",] -- Q[" 'pl 81' p><,><o]
1

is a tensor and that the same holds for () instead of [ ] . 1 )

3. Geometrie objects and quantities in Xn.

A geometric obfect 2), 3 ) at a point ~,. of Xn is a correspondence between
the allowable coordinate systems in regions containing ~,. and the ordered
sets of N numbers, satisfying the conditions:
1. To every coordinate system (n) there belongs one and only one
ordered set of N numbers.
2. If (/) A; A = 1, ... , N cotresponds to (n) and (/) A'; A' = 1', ... , N'
to (x'), the ([JA' are functions of the ([JA and fundionals of the functions /"'
in a neighbourhood of the point ~". Instead we may also require that
the ([JA' are functions of the ([JA and of the values of the ~,., and of the
derivatives A~' = 8,. ~"', 8" A~', 8v 8" A:', ... up to a certain order u at
the point ~". If the j"' are analytic and u = oo both definitions are
equivalent, but in all other cases the first is more general. But we prefer
here the second since it is sufficiently general for our purposes and
avoids the use of functionals. 4 ) Then the transformation formula of a
general geometric object takes the form
(3.1 a)

where g} means that the values of the whole foregoing expression must
be taken at the point ~". Note that in expressions as ([JA {n it would not
be better to write ~" instead of ~. because ([JA g"} would have the same
meaning as ([JA{~"'}. The form of the functions F,1 does not depend on
the values at ~" mentioned between the parentheses. Hence the in-
version of (3 .1 a)
(3 .1 b)

contains the same functions FA.

1) ScHOUTEN 1940, 2 also for general valences of P and Q.
2) Cf. for more general definitions ScHOUTEN and HAANTJES 1937, 1, and P. P.
1949. 1.
3 ) Further Iiterature: WuNDHEILER 193 7, 1; GoLAB 1938, 1; 1946, 1; 194 7, 1 ;

1948, 1; 1950, 1; 2; NAKAE 1944, 1; WAGNER 1945, 1; 1949, 1; 1950, 1; PENSOV

1946, 1; 1948, 1; GEORGHIU 1949, 1; LIBER 1951, 1; VASILIEV 1951, 1; EHRES-
MANN 1952, 1; HAANTJES and LAMAN 1953, 1; NIJENHU!S 1952, 1, gave a develop-
ment from the point of view of the representation theory of a groupoid.
4 ) Of course the pseudo-group \~Sn used in the definition of the Xn must only

contain transformations of class ::2; u.

5*
68 II. Analytic preliminaries.

The <P "1 are called the components of the geometric object with respect
to (").
If we wish to express that a scalar p is a function of the ;" we must
write p = fW') and not p = p(;"). In fact, if a new coordinate system
is in troduced we can write p = f' (;"') but we could not write p = p (;"')
because the same letter p can not be used as a function symbol for
two different functions. The same holds for more general geometric
objects as <P.1 and this is the reason why we introduced the notation
$"1{;} for the values of <P.1 at the point ;K.
For all geometric objects we may make use of a notation first intro-
duced by NrJENHurs. 1 ) If the components <PA of a geometric object
with respect to the coordinate system (u) are expressed as functions
of the coordinates ;"' he uses (x')<[JA as function symbol. Hence
(3 .2) l[J A = (x')(/J A (;"') = (><)l[J A (;"); l[J .1' = (x')<[J A' (;"') = (><)<[J .1' (;")

It should be remarked that is not a kernel in our sense because

(x')<[J .1

the change of a kernel always means a change of the object. If the

object is not a scalar it is convenient to write for shortage <PA instead
of (><lr/J A and (/JA' instead of (x'J<[J .1-, though by doing this we use a symbol
for components as a function symbol.
It may happen that two or more geometric objects have the same
kernel letter. If, in some investigation, this is not the case it is often
convenient to denote a geometric object in the text by its kernelletter
only. Objects with the same nurober of components and the same
manner of transformation are said to be of the same kind.
If, for every choice of (x) and ("'), the <PA' are functions of the <PA
and the A~' only, linear homogeneaus in the <PA and homogeneaus
algebraic in the A~' the geometric object is called a geometric quantity
or briefly a quantity.
Geometrie objects may be defined with respect to two or more
spaces 2 ), for instance an X" with the Coordinates ;x and an xm with
the Coordinates rt; a = 1, ... , m. The components may be functions
of ;'' and rt.Objects of this kind are often called connecting objects.
There may be relations between the Xm and the Xn, for instance the ;"
may be functions of the 1t. If a connecting object is a quantity it often
happens that its components have two different kinds of indices each
belanging to one of the underlying spaces. For instance B~ = o;''jor/.
In dealing with objects with components with respect to more than
one coordinate system, in most cases it is more convenient to introduce
lJ
NIJENHUIS 1952, 1, p. 44ff.
Such quantities were introduced incidentally by BoMPIANI 1921, 1 and more
2)

systematically by VAN DER WAERDEN 1927, 1. Cf. SCHOUTEN 1927, 2; DusCHEK

and MAYER 1930, 1, Ch. VII; TucKER 1931, 1; RusE 1931. 1; EI 1935, 1, p. 9.
 3. Geometrie objects and quantities in Xn. 69

special function symbols, for instance A~' = F'/ (<;") = er(.;"'). But if
symbols for components are used as function symbols in this more general
case, it must always be made very clear to which coordinate system
or systems the function symbols introduced in this way belong.
Pseudo-quantities are a special kind of connecting quantities. They
are defined with respect to the Xn of the ~,. and the X 1 of a variable ~0
with the transformation group (cf. I 2 under 5):
~"' = !"' (~")
(3-3)
{ ~o = a (~"Ho.
Their components may be functions of the ~,. and of ~0 A pseudo-
scalar of class c has one component, depending only on the ~" and has
the transformation
(0') -" (0)
(3.4) :p=u :p.
Hence ~0 itself is a pseudo-scalar of dass 1. Other pseudo-quantities
arise if a pseudo-scalar is multiplied by an ordinary quantity. All these
pseudo-quantities can be multiplied and transvected. Addition is only
possible if the pseudo-quantities are defined with respect to the same
(n + l)st variable and if the valences, the weights and the dass are the
same.
If we have some quantity or some set of quantities each defined to
within an entirely free scalar factor, we might consider them as pseudo-
quantities, each defined with respect to its own (n + l)st auxiliary
variable. But as there are no connections between these auxiliary
variables there is in general no use in introducing them. In this case
we use the sign LJ (cf. I 2). For instance Lv"j is a vector defined to
within a scalar factor. Note that it is no use asking whether Lv" j is a vector
or a vector density because the influence of a factor L1 is nullified by the
entirely free scalar factor. Lv"j is neither a vector nor a vector density
but a pseudo-vector though an auxiliary variable ~0 is not mentioned.
The ~,. represent a geometric object which is no quantity. But
according to (2.4) the d~" undergo a homogeneous linear transformation
d~"' =A:' d~". Hence the d~" at the point ~,. represent a quantity at
that point. But from the same equation we see that every allowable
coordinate transformation in an 91 Cf') induces one and only one trans-
formation of the group G,. 0 at ~,. and that to every transformation of
this group at least one allowable coordinate transformation in Xn can
be found inducing this very transformation of G,. 0 at ~". This means
that to every point ~" of Xn there belongs a centred En, the tangent
space or tangent En of the point ~". 1 }
1) In former publications we used the term local En.
 70 II. Analytic preliminaries.

From the definition it follows that a quantity of X,. at ;" is the

same as a quantity as defined in I 2 in the tangent E,. of ;". Hence
all quantities of E,. considered in Ch. I may occur as quantities of X,.
at every point of X,.. All equations of transformation like (I 2.1),
(I 2.4), (I 2.13), (I 2.19) and (I 2.22, 23) arevalid in Xn, but the A~', A1,
and LI are no Ionger constants.
If the components of a geometric object are defined in some region
of X,. as functions of the ;", the object is said to form an obfect field.
A field is said to be analytic (of class u) in a region if all its components
are analytic (of class u) in that region.
As a first example we take a scalar field p = fW'). If p is analytic
or at least of class 1 in the region considered we may form the derivatives
o;. p. Because of
(3 .5)
these derivatives are the components of a covariant vector, the natural
derivative or gradient of p. If we use the abridged notation of I 7,
we write Grad p or Dp for the gradient of p and also p if the vector
occurs in the abridged form (I 7.1) of an alternating product.
Now we take a vector field v" = f"(e) and form its derivatives o;.v".
Because of 1)
(3.6)

these derivatives are neither the components of a tensor nor of any

geometric object, since the 8;.' v"' depend not only on o;v", A:' and
o,A~' but also on v". But the n +
n 2 expressions v", o;. v" are the compo-
nents of a geometric object whose transformation involves the o,~A:'.
The same is true for the derivatives of all geometric quantities except
scalars and W-scalars.
To every coordinate system (x) in Xn there belong n scalar fields ;" .
The transformation ;" -+;"' is a coordinate transformation in X,.,
accompanied by the point transformations ;" -+~"' in lll,.. But this latter
transformation can be considered also as an object transformation in
x x'
Xn, viz. the transformation of the scalar fields ~ into the scalar fields ~.

The gradients of the scalar fields ~" are the vectors of the covariant basis
In dealing with Operators like
1) a,..
(and also v,.. in the next chapter, cf. foot-
note 3 on p. 132) we agree that the operator works by differentiation on all
quantities to the right of it in the same term till the first closing bracket, whose
corresponding opening bracket stands at the left of the operator. If these opera-
tors are applied several times in succession we usually prefer to write the kerne!
Ietter 8 (or V) only once, for instance 81'P .. 1'1 instead of 81'P ... a,..,.
 3. Geometrie objects and quantities in Xn. 71

belanging to (x).l) Their components with respect to (x) have the values
1 and 0 only
(3.7) e.! _
def 0\ ;3'_ ~"
u;.\0 __ u.!

but the components with respect to (x') may have quite other values

(3.8) e.~. = o;:i *-A~..

The vectors of the contravariant basis belanging to (x) arise from dif-
ferentiation of the Coordinates ~" in Xn with respect to the ~"

(3.9) e" def 8~" _":_ 0~;

). ).
a~

The equations (3.7-9) and

(3.10) A~ = a.~~";

"
(3.11) ~"-~
U;,-
).
a~

show clearly the difference between the tensor A2, the two bases each
consisting of n vectors e" and e;. and the set of n 2 Sealars o2. We also
).

see the difference between the living indices, transforming only with
coordinate transformations in xn' and the dead indices, transforming
only with the object transformations in xn corresponding to point
transformations in llln. 2 )
At every point every quantity field has a rank with respect to each
set of indices, as defined in I 5. If such a rank is rat~", this means
0
that a certain matrix formed from the components of the quantity has
rankrat that point .. Hence all subdeterminants with more than r rows
vanish at ~" and there is at least one non vanishing subdeterminant
0
with r rows. Now let the quantity be analytic in a region R containing
~" and let ~" be a point where r has a maximum value. Because all
0 0
subdeterminants are analytic, there exists an 91(~") such that the sub-
o
determinants not vanishing at ~" can not vanish at any point of 91(~").
0 0
Hence the rank has toberat all points of 91(~"). We call 9'l(~") aregion
0 0
of constant rank with respect to the index or the indices concerned.
1) The e" and ;;. were formerly called the contra- and covariant measuring
).
vectors belanging to (u).
2 ) All these differences should not be neglected. Neglecting them Ieads to
many difficulties especially after the introduction of special kinds of processes of
differentiation.
 72 II. Anal ytic preliminaries.

Now let there be two subregions of constant rank in R in which the rank,
always taken with respect to the same indices, takes the values r1 and
r 2 ::;;;; r1 . Then in the first region all subdeterminants with r 1 1 rows +
vanish. But this implies that they vanish at all points of R. Hence
r1 = r 2 . This proves that a quantity analytic in R has the same rank
with respect to some given indices in all subregions of constant rank
in R and that there is no point where it can have a higher rank. This
does however not imply that there are no points in R where the rank
is lower than in the subregions of constant rank. But these points can
not form a region. For instance a vector field v" analytic in R and not
vanishing at all points of R has rank zero at all points where v" = 0 and
rank 1 at all other points. Points where v" = 0 may occur but they
can not form a region because in that case v" would vanish all over R.
If v" = 0 at ~",in every 91(~") there is at least one point where v" =f: 0.
0 0

Exercise.
II 3, 1 1). Let (/JA be a tensor field with the collecting index A. If 'Pn
is a tensor field with the collecting index Q whose components are func-
tions of the ~" and the f/J A, prove that the fi'l1gj8f/J A, where the ~" are
supposed to be left constant, are components of a tensor field. The
same holds mutatis mutandis for tensor densities.

4. The Xm in Xn . 2 )
Let N functions Fa(~"); a = 1, ... , N be analytic in 91(~"). The
0
matrix
(4.1)
is the functional matrix of the set of functions and its rank r at ~" is
0
the rank of the set at that point. r is also the maximum number of
linearly independent differentials among the dFa, hence r ~n; r ~N.
It is well known that the functions Fa are independen t in 91 (~") if and
0
only if r = N at at least one point of 91( ~").
0
Let r = N:;:;; n. Then we can take m = n - N auxiliary functions
pN +1 (f'), ... , pn (~") analytic in 91(;"), such that the rank of the set pk;
0
h = 1, ... , n equals n at all poin ts of 91 (~"). Then the transformation
0
(4.2) ~h = ph (~"); h = 1, ... , n
IS an allowable coordinate transformation in 91 (~"). That proves the
0
1) Cf. HoRAK 1929, 1.
2) Cf. for a more elaborate treatment P. P. 1949, 1, Ch. II; R. S. 1947, 1, 2.
 4. The X.,. in Xn. 73

Theorem of adaptation.
If the functions Fa(t'); a = 1, ... , N, are analytic in W W') and if
0
r = N ;;;;;: n, there exists in WW') a1~ allowable coordinate system .;h;
0
h = 1, ... , n, suchthat .;a =F11 (t'); a = 1, ... , N.
Let the system of equations
(4.3) Fa(t') = 0

be consistent. We call r the rank of the system. Every point .;" of W(.;")
0
satisfying (4.3) is called a null point of (4.3) and the set of allnull points
is called the null manifold of (4. 3). (4. 3) is said to be a null form of this
manifold. Two systems of equations having the same null points in
W(.;") are said to be equivalent in that region. Two equivalent systems
0
need not have the same rank at all null points. E.g. _;I= 0; ,;2 = 0 and
ee = 0; .;z
= 0 at the point .;" = 0.
Now we consider a system of this kind
(4.4) C(t')=O; x=m+1, ... ,n; (C analytic)

with rank n- m at the null point .;". Then there exists an W(.;") such
0 0
that the rank is n- m at all its points. In this case (4.4) is called mini-
mal regular of dimension m at $". The nurober m is called the dimension
0
of the null manifold and we have 0 ;;;;;: m ;;;;;: n. From the definition we
see that every subsystem of a system which is minimal regular at .;"
0
is itself minimal regular at .;" and we seealso that there exists an W($")
0 0
such that the system is minimal regular and the null manifold rn-
dimensional at all null points in W(;"). The notions "minimal regular"
0
and "dimension" are invariant for all allowable coordinate transfor-

l
mations. Of the following four examples of systems of equations only
the last one is minimal regular at .;" = 0.

a) e .; 0; e .; = o
2 = 3

(4.5)
b) e .;z = 0; e=o
c) e.;z = 0; e=o; e=o
d) .;1 = 0; .;z = 0

and only (4.5 b) and (4.5 c) are equivalent to a minimal regular system,
viz. (4.5 d). The null manifold of (4.5 a) has no dimension at all at .;" = 0.
If the .;" are rectilinear coordinates in En, (4.5 a) represents a figure
consisting of an En-l and an En_ 2 through the origin.
74 II. Analytic preliminaries.

If (4. 3) is minimal regular of dimension m a t W(.;") , according to

0
the theorem of adaptation there exists a coordinate system (h); h =
1, ... , n suchthat (4.3) is equivalent to the system
(4.6) .;x = 0; x=m+1, ... ,n.
Accordingly the .;a; a = 1, ... , m can be used as coordinates in the
null manifold. Now the pseudo-group \!Sn contains as a sub-pseudo-group
the pseudo-group \!Sm of all analytic and invertible transformations
of the .;a only, the .;x being left invariant. Hence the null manifold
is an Xm. Such an Xm is said tobe imbedded in an Xn or an Xm in Xn.
An Xm in Xn is often called an m-dimensional surface and for m = 1, 2
and n -1 it is also called curve, surface and hypersurface respectively. 1 )
The quantity
(4:7) qdef oi. C'"; X= m + 1, ... ,n
with rank n- m is the covariant connecting quantity of the xm in xn
(cf. I 2 point 7 and II 3).
If a function s (.;") is analytic in W(.;") and zero at all null points
0
of (4.4) in W(.;"), there always exists an W(c;") suchthat s satisfies in
0 0
this W(.;'') an equation of the form
0
(4.8) s =tPxW) CX; x= m + 1, ... , n
with analytic functions t1Jx. This is the so-called first base theorem. 2)
+
A set of n- m functions C'"'; x' = (m 1)', ... , n' which take the value
zero at every null point of (4.4) and whose differentials dC"' are linearly
independent at .;", constitutes a base of the Xm at .;" and consequently
0 0
at all null points in some W(.;"). 3 ) Hence the C'" in (4.4) constitute a
0
base. As a consequence of the first base theorem we have equations
of the form
(4.9) CX'=C;'cx; x=m+1, ... ,n; x' = (m + 1)', ... , n'
valid in an W(.;") with coefficients C;' analytic in this region. Hence
0
(4.10) acx = C!' dCX; X= m + 1, ... 'n; x' = (m + 1)', ... , n'
1 ) Note that in consequence of the definition anXm inXn never has singularities.
An X 2 or surface in our sense in ordinary space is never a surface in the ordinary
sense with singular points and curves, but it can be apart of such a surface, free
from singularities.
2 ) Cf. foraproof of this well known theorem for instance P.P. 1949, 1, p. 37;

R. S. 1947, 1, 3.
3) We define a base in a different way as KHLER, who also allows sets of more
than n-m functions. Cf. 1934. 1, p. 13.
 4. The Xm in Xn. 75

at all null points in this region. We see from this that cx' = 0 is a null
form of the xm in some ~ (f') which is minimal regular in this region.
0
This is the second ~ase theorem. 1 ) It is easy to see that Det (C;') =f= 0 in
~ (~").
0
Every transformation of the fom1 (4.9) with Det (C;') =f= 0 is called
a base transformation. If a coordinate transformation and a base trans-
formation are effected simultaneously we have
(4.11)
valid at the null points of the region considered.
There is another way of defining an Xm in Xn. Let the rt; a = 1, ... , m
be allowable coordinates of some Xm which has no point in common
with xn and Jet there be given a system of equations
(4.12) ~"=B"(rt); a=1, ... ,m
where the B" are analytic in an ~ (rt) and the matrix of
0

(4.13)

has rank m in this region. Then (4.12) fixes a one to one correspondence
between the points of xm in an ~ (rt) and certain points of xn in an
0
~W); ~><def B"(rt). Hence every point of this ~(1t) can be identified
0 0 0 0
with its corresponding point in ~ (~"). This process is called the im-
o
bedding of the Xm in Xn. (4.12) is called the parametrieform (or para-
metrie representation) of the Xm in Xn, minimal regular of dimension m
at rt, and BI, is the contravariant eonn.eeting quantity of the Xm in Xn.
0
In order to prove that this second definition of an X". in Xn is in
agreement with the first one, we start from the null form (4.4) and use
the theorem of adaptation. According to this theorem the coordinate
system (u) can always be chosen in such a way that the null form takes
+
the form ~c = 0; C=m 1, ... , n. Now, writing these equations in the
form
(4.14) ;rx = ~; ~c = 0; rx = 1, ... , m; C= m + 1, ... , n
and considering the ~rr. as parameters we see that we have got a minimal
regular parametric form. Now let us start from the parametric form
(4.12). Since BI, has rank m, the indices 1, ... , n can always be rear-
ranged in such a way that B~; rx = 1, ... , m; b = 1, ... , m has rank m.
1 ) Cf. for a proof of thiswellknown theorem for instance P. P. 1949, 1, p. 37;
R. S. 1947, 1, 3.
 76 Ilo Analytic preliminarieso

Then the transformation ~" -+~h; h = 1, 000, n, defined by

(4015) ~" = B"W) + o~~x; a = 1, 000'm; x=m+1,o .. ,n
is an allowable coordinate transformation in the n!gion concemed. If

r
this transformation is applied to (4012) we get the system
B"'W) = B"'(rt) b) B'W) + o;~x = B'(rt);
(4016) a=1, .. o,m; x=m+1,o.o,n
IX=1,ooo,m; t;=m+1,o .. ,no
From (4o16a) it follows that e=1t and consequently that (4.16b) is
equivalent to the system e = 0, constituting a minimal regular null
form. 1 )
To every allowable coordinate system (~) there belang for every
value of m<n just (:) sets of oon-m Xm's with equations of the form

(4.17) ~"=c"; Z:=m+1, ... ,n

where the c" are n- m parameters. These Xm's are called the coordinate-
X"''s (-curves, -surfaces, -hypersurfaces form= 1, 2, n -1) of (x). Hence
the theorem of adaptation proves that every Xm in Xn is coordinate-X"'
of some suitable allowable coordinate system. oon-m Xm's with equa-
tions of the form (4.17) are said to constitute anormal system of Xm'so
The most important property of a normal system is that every point of
the region considered lies in one and only one Xm of the systemo From this
we see once more that an Xm in our sense must always be considered
in a certain region only. For instance the straight lines through the
origin of a centred En form anormal system of X 1 's for all regions not
containing the origin and for these regions only.
The coordinate-X",'s are said to constitute a net, the n(!t of the Co-
ordinate system. A net contains for instance n normal systems of X 1 's
whose directions at every point are linearly independent. Not every
net can belang to a coordinate system (cf. II 9).
The oon-m Xm's of anormal system themselves constitute an (n- m)-
dimensional manifold. If the equations have the form (4.17) the ~c
can be used as coordinates in this manifoldo Now the pseudo-group
\lSn in ~ (~") contains a sub-pseudo-group \lSn-m of all analytic and
0
invertible transformations of the ~c only, the ~"'; IX=1, ... , m, being
left invariant. Hence the manifold is an Xn-m. It is called the reduc-
tion (Zusammenlegung) of the Xn with respect to the normal system of
1 ) In Po Po 1949, 1, II 4 it has been proved that it is also possible to derive
a minimal regular parametric form from a minimal regular null form and vice
versa without using coordinate transformations.
 4. The Xm in X,.. 77

Xm's. This process of reduction is a generalization of the process de-

fined in Ch. I 1.
If an Xm in X" is given by its null form and its parametric form,
the connecting quantities B'b and C1 are known at every point of Xm.
In the tangent E,. of each of these points they fix an Em called the tangent
Em of Xm. It is a sub-space of the tangent E" at that point. It can be
identified with the tangent E ... of Xm, as defined in II 3. B'b and Ci
can now be used in the way described in I 4 to form
1 o the section of any covariant tensor of X", called here the section
with Xm, e.g. 'wb = Bg W;. (cf. I 4.3, 14a);
2 the reduction of any contravariant tensor of X", called here the
reduction with respect to Xm, e.g. "v"' = C;, v" (cf. I 4.7, 15 a);
3 the X"-compon.ents of a contravariant tensor of xm' e.g. P" = B; pa
(cf. I 4.5, 16a);
4 the X,.-components of a covariant tensor in the E"_m arising
from the tangent E" by reduction with respect to the tangent Em,
e.g. S;. = c~ Sy (cf. I 4.9, 17a).
An Xm in X" is called rigged if all its tangent Em's are rigged (cf.
I 4), that is if, at each of its points, an E"_m in the tangent E,. is given
having no direction in common with the tangent Em. In this case
which occurs frequently in differential geometry the unity tensors B~
and c; of the Xm and of the local En-m also have components B'b,
BL B~ and CL C~, C~. The quantities can be used in the way described
in I 4 to form:
1o the section with E"_m of any covariant tensor of X", e.g. "wr =
c~ w;. (cf. I 4.14b);
2 the reduction with respect to E,._m of any contravariant tensor
of X,., e.g. 'va=B:v" (cf. I 4.15b);
3 the X"-components of a covariant tensor of Xm, e.g. q;. = B~ qb
(cf. I 4.16b);
4 the X"-components of a contravariant tensor in the En-m e.g.
r"=C~r"' (cf.I4.17b);
5 the Xm-Part and the E,._m-Part of any tensor of X,., e.g. (cf. I 4.21)
=
1
'v" B~v'-; 'w;. = B;:w,.
(4.18) "v" = c~ v'-; "w;. = C~w..
v" = B1v; + c;: v;; W;.= B;:w,.+C1w,..
If a field, for instance a scalar field p, is defined in an W(1t) over the
0
Xm with the parametric equations ~" = B" (rt), it is often useful to con-
sider a field, analytic (or of a certain dass u) at all points of an 1)1 (~");
0
78 II. Analytic preliminaries.

~" = B" (rt), having at all points of 'in (rt) the same values as the given
0 0 0
field. This process is called a Prolongation of the field over xn and
it is convenient to denote the prolonged field with the same kernel.
Then 8;.p also has a sense and the equation 8b p = Bt 8~-' p holds. This
enables us to use the operator 8;. freely. Of course all those results
obtained in this way, that can be expressed in terms of quantities and
operators of Xm only, are independent of the manner of prolongation.

5. The E'"-field or X::" in Xn.

Systems of linear partial differential equations.
If we take the tangent Em's of the X"''s belanging to anormal system
we get an E"'-field. But this is a very special field because its Em's
can be arranged in oon-m sets of oom Em's each set consisting of the
tangent Em's of an Xm.
Instead of this special field we may now define a general Em-field
by giving, in each tangent En of a point of a region, an Em through
the contactpoint by means of its connecting quantities B'b and (or)
q. Such a general field is called an X;:' in X". 1 ) The Em at a point
is called tangent Em because it lies in the tangent En. If a b-transforma-
tion and an X-transformation is effected in all En's simultaneously we

l
have
a) B'b, =Bi, B'b; b) Cf= c;' q
(5.1) b = 1, ... , m; x = m + 1, ... , n
+
b' = 1', ... , m'; x' = (m 1)', ... , n'
but the Bi, and the c;' are now functions of the ~". The Em at each point
is spanned by the contravariant vectors B~, ... , B':n and also by the
covariant vectors C:f+ 1 , ... , C~. A b-transformation can be interpreted
as a linear homogeneaus transformation of the m vectors B'b; b = 1, ... , m
and an X-transformation as a transformation of the same kind of the
n- m vectors Cf.; x = m + 1, ... , n. But the b-transformations here
arenot connected in any way with a coordinate transformation in some
variables rt
(cf. II 4).
By means of B'b and Cf. we are now able to define the section with
X;:' of any covariant tensor of Xn, the reduction with respect to X;:' of
1 ) Cf. for the elder Iiterature on the X;:' and anholonomic coordinate systems

VRANCEANU 1926, 1; 2; 1927, 1; 2; 1928, 1; 2; 3; 4; 1929, 1; 2; 1930, 1; 1931, 1;

1932, 1; 1933, 1; 1934, 1; 1936, 1; HLAVATY 1924, 1; 1928, 1; 1930, 1; 2; T. Y.
THOMAS 1926, 1; HoRAK 1927, 1; 2; ScHOUTEN 1928, 1; 1929, 3; SYNGE 1928, 1;
ScHLESINGER 1929, 1; SCHOUTEN and V. KAMPEN 1930, 2; BoRTOLOTTI 1931, 1;
2; 1932, 1; 1933, 1; 1937, 2; 3; 1940, 1; DIENES 1932, 1. Cf. for the newer Iiterature
Ch. V 7.
 5. The Em-field or Xlf in Xn. 79

any contravariant tensor of Xn and the Xn-components of any contra-

variant tensor of X': and any covariant tensor in the local En-m arising
from reduction of the tangent En with respect to the X':. The formulae
are the same as (I 4.14a, 15 a, 16a, 17a). If, besides the X':( we give
an x~-m suchthat the tangent Em andEn-m have no direction in common,
the X': (and also the x~-m) are said to be rigged. In this case we have
B~, BI,, Bi:, B~ and CI, c;, c;, C7. at every point. By means of these
we are now able to construct also the section with x~-m of any covariant
tensor of Xn, the reduction with respect to x~-m of any contravariant
tensor of Xn, the Xn-components of every tensor of X': and of x~-m
and the X';;-part and x~-m-part of every tensor of X n.
Now the question arises whether a given Em-field belongs to some
normal system of Xm's. But this is only a special case of a far more
general question. An Xp is said to envelop an Em-field if its tangent
Ep contains, at every point, the tangent Em of the field and it is said
to be enveloped by an Em-field if its tangent Ep is contained at every
point in the tangent Em. Hence there are two problems, the outer
Problem concerning all enveloping Xp's and the inner problem concern-
ing all enveloped Xp's. 1 ) We require especially all enveloped normal
systems of Xp's for the maximum value of p and also all enveloping
normal systems of Xp's for the minimum value of p. According to the
definition an enveloping Xm is at the same time enveloped, hence the
two problems are identical for p = m. But for all other values of p
they are essentially different and require quite different mathematical
methods.
The inner problern is much more difficult than the outer problem.
Only the casem=n-1, known asPFAFF's Problem, can be dealt with
by fairly elementary means as will be shown in II 7.
The outer problern is closely connected with the theory of systems
of linear partial differential equations of the first order. We give first
a definition of totally integrable systems of general partial differential
equations.
If a system of N partial differential equations of the first order with
n independent variables ~" and M unknown variables z; b = 1, ... , M
b
i
(5.2) F(;",~,o;.~)=O; i=1, ... ,N; b=1, ... ,M
is given, every set of functions i satisfying (5.2) is called a solution of
(5 .2) and every set of values ~"' ~' n}. satisfying the equations
* b b

(5.3) i
F(~", * *)
n.rh =0
* b b
1) Cf. P. P. 1949, 1 Ch. III, VIII, also for literature.
 80 II. Analytic preliminaries.

is called a null point of (5.2). Now the system (5.2) is said tobe totally
integrable 1) in an ~ (e", ~, ~;.) if for every null point ~, ~. ~;. in this
0 I> I> * b b
region there exists at least one solution of (5.2) satisfying the condition
that for ;" =~"
* * *
(5 .4) Z='f} 8;.Z ='f/;_.
b I>' b I>

The necessary and sufficient conditions for a system to be totally

integrable are called its first conditions ol integrability.
Now let p be the smallest number for which a normal system of
Xp's exists enveloping the Em-field B'b, Cf. and let

(5.5) PW)=c~; ;"=W"('fJa,c~); .a=I. ... ,p;-

6=P+1, ... ,n
be the null form and the parametric form of the enveloping normal
system. Then the Fli are n- p independent solutions (i.e. solutions
with linearly independent gradients) of the homogeneaus linear equa-
tions
(5.6) B't 8~' I= 0; b = 1, ... , m

and every linear element d;" satisfying the equations

(5.7) x=m+ 1, ... ,n
also satisfies the equations
(5.8)
The equations (5.6) and (5.7) are said to be adjoint to each other. The
first gives the condition for the tangent Ep's of an enveloping Xp and
the second the condition for its linear element&. lf F is a solution of
(5.6), Fis called an integral lunction of (5.7) and the equationF = const.
an integral of (5.7).
lf the operator B't 81' is applied twice in succession we get, after
altemation,
(5.9)

because 8[ 81' 1I= 0. Hence every solution of (5.6) must satisfy the (;)
linear homogeneaus equations (5.9). If, among these equations, there
are exactly f-l1 linearly independent of (5.6) and of each other, we get
1) For the definition of total integrability for systems of higher order see for
instance P. P. 1949. 1, p. 89.
2 ) Cf. for the conditions for one single enveloping Xp. P. P. 1949. 1, p. 8Sf.
 5. The Em-field or X:;' in Xn. 81

a system of m1 def m + fl1 equations

(5.10) B(.8'"f=0; b1 =1, ... ,m1
called the jz:rst derived system of (5 .6). This system has the same solu-
tions as (5.6). fl1 = 0 if and only if
(5.11) C.cb" det
= - &p (8 lv, B'")
b] C"-
p. - c b 8[v C"
'" +
'p.] -
- 0

It is remarkable that C~i,X is a tensor though B[c 81 B'b1 does not have 1

this property. If flt = 0 the system (5.6) is said tobe complete. It can
be proved that a complete system of m equations is totally integrable,
that it has just n- m independent solutions and that its integrability
conditions 1) are identically satisfied. Hence if (5 .11) is satisfied and only
in that case there exists an enveloped and enveloping normal system
of Xm's. The Em-field is then called Xm-forming.
The solution of the complete system is equivalent to the determina-
tion of this set of oon-m Xm's. If the Xm's are known, every function
of the ;" that is constant on each of these Xm's is a solution.
If the first derived system is complete there exists anormal system
of Xm+ 11 ,'s. If it is not complete a second derived system with m flt fl 2 + +
equations can be found, and so on. At last we get a complete system
having the same solutions as (5.6). If this last system consists of n
equations, (5 .6) has no solutions except the trivial one f = const. The
numbers fl 1 ,fl 2 , are arithmetic invariants of the Em-field.
By
(5.12) v""m cx: Bf" ... B~';]
a contravariant pseudo-m-vector is given with the m-direction of the
tangent Em. In the same way by
(5 .13)
a covariant pseudo-(n- m)-vector is introduced with the same m-direc-
tion. The n. a. s. conditions (5.11) for the E",-field to be Xm-forming
can now be written as
15.14)
from which we may easily deri:ve the iollowing equivalent forms
(5.15) (8,., v""["m) vml = 0 3)
(5 .16) (8", wx,, ... n-ml v" "m = 0 4)' 2)
(5.16a) (8[", w"m+l"nl wx,], ... n-m = 0.
1) Cf. for the exact formulation and the proof of the existence theorem for in-
stance P. P. 1949, 1, p. 108.
2) ScHaUTEN 1918, 2; ScHaUTEN and STRUIK 1919, 1; STRUIK 1922, 1, p. 53ff.
3) v. WEBER 1900, 1, p. 73 ff. has an equivalent form.
4) v. WEBER 1900, 1, p. 103 has an equivalent form.

Schauten, RicciCalculus, 2. Aufl. 6

 82 II. Analytic preliminaries.

To every simple contravariant pseudo-vector field v" .. ""' that is

X".-forming there belongs a complete system with the equations
(5.17) v" .. ""'o,..f=O or w[A. ... ;.,._".oA1 1=0,
expressing the fact that the (n -1)-direction of oJ.I contains thetangent
m-direction, and the adjoint system with the equations
(5.18) v[""""'d~l=O or wA. ...;.,._".d~A.=O
expressing the fact that the direction of d ~" lies in the tangent m-direc-
tion.
If the E".-field is X".-forming and if
(5.19) PW)=cr; x=m+1, ... ,n
are the equations of the normal system o~ X".'s, the tangent E". at
everypoint lies in the tangent En-t of each of the n-m Xn_ 1's (5.19).
Hence
(5.20)
and this equation expresses the fact that wA.".;.,._". is a gradient pro-
duct (alternating product of gradients) to within a scalar factor. This
proves that (5.16a) is the necessary and sufficient condition for wA. ... ;.,._.,
to be a gradient product to within a scalar factor.
The theory of homogeneous linear equations of the first order can
be generalized for non homogeneous linear equations of the form
(5.21)
where BI: has rank m. lt can be proved 1) that such a system is totally
integrable and that it has just n- m + 1 independent solutions if and
only if (5.6), called the reduced system ol (5.21), is complete, that is,
if there exist functions O:b W) such that
(5.22) Bf" fJII'I B~ = O:b B:
and if moreover the (/Jb satisfy the equations
(5.23) Bfcoii'I(/Jb] = O:b (/Ja
If the integrability conditions (5.22, 23) are satisfied, (5.21) is called
I n-m
complete. If I, ... , I are n - m independent solutions of the reduced
0
system, and if lis one solution of the complete system (5.21), the general
0 I n-m
solution is I + 1p (1, ... , I ) where 1p denotes an arbitrary analytic
function. 2)
Cf. for instance P. P. 1949, 1, 111 14.
1)
Cf. for the exact formulation and the proof of the existence theorem for
2)

instance P. P. 1949, 1, p. 111.

 6. The invariant operators Rot and Div. 83

We apply this theory first to the gradient equation

(5.24)
If this system is written in the form

(5.25)
we see that the reduced .system is complete and that the coefficients a
are all zero. Hence here the only integrability conditions are
(5 .26)
0
In fact, if a solution p of (5.24) is known, the general solution is
p = p + c and th1s means that the value of p at any point ;'' can be
0

given arbitrarily. *
Exercise.
II 5, 1. If a vector field v" is given m 91 (~"), there always exists
0
a coordinate system (x) in an 91 (~") suchthat v" = e". Every covariant
0 1
(n -1)-vector in xn is product of a scalar and a gradient product. 1 )

6. Theinvariant operators Rot and Div.

As we have seen in II 3, the derivatives of the components of a
tensor with valence > 0 with respect to the f' in an Xn are not com-
ponents of a tensor. But if we form the alternated derivative of a co-
variant q-vector, the additional term in

(6.1) {
Clr.u' w_.; ... i.qJ = A~,1;::: ~~ Clr.u W;.... ;.J + w-' .. ' o[.u' Ai; ... Ai~l
-- A.u-' ... ;..
i'' ;.; ... i.q
o[.u w;,. '. 'g]

vanishes because o["'A;:, 1 = 0. We call this alternated derivative multi-

plied by q + 1 the natural derivative (ct. II 3) and for q > 0 also the
rotation of w.<.... -<. If we use the abridged notation of I 7, we write
Rotwand also (q+1) Dw. Rot w vanishes identically if q=n. As a
consequence of the definition the process of forming the natural derivative
applied twice in succession, always leads to zero:
Rot Grad p = 2 D D p = 0
(6.2) {
Rot Rot w = (q + 2) (q + 1) D D w = 0.

All this is also true for W-scalars and covariant W-q-vectors. If the
corresponding (n- q)-vector Ll-density tu""P; p = n- q is formed from
1) GoURSAT 1922, 1, p. 117.
6*
 84 II. Analytic preliminaries.

I
w by means of ~~, ... ~n (cf. I 7) it follows that for p;:;;; 1
~ 1 ~
8I' ltJI'~, ... ~p= (-1)----8 W
q! I' .l, ... .lq
~.l, ... Aqfl"'""P
(6.3)
= (- 1)q-q __l _ (q -f- 1) 8 W ~l'.l, ... .lq~, ... ~p.
(q +I)! [~t .<, ... .<.J

Hence, 81'fi5~'~- .. ~v is a (p-1)-vector Ll-density of weight +1. It is

called the natural derivative or the divergence of fi5x, xv and it is written
Div fi5 in abridged notation. Div fi5 is by definition zero if p = 0. From
(6.3) we see that if
(6.4) W=Rotw; llli=Divfii,

corresponds to ( -1)q-.++' W. From the definition it follows that

the rule concerning the result of two natural derivations in succession
holds also in this case
(6.5) Div DiviU = 0.
All this is also true for contravariant p-vector densities of weight + 1.
Let W.<, ... ?.q+t be a simple (q + 1)-vector whose rotation vanishes in
an irr(~"). Then according to (5.16a) the Ep_ 1 -field with the (p-1)-
o
direction of W is Xp_ 1 -forming and this means that there exists an
irr ( ~~) where W is a gradient product to within a scalar factor
0

(6.6) s].
W=s[s1 ... q+1
Taking the rotation we get
(6.7) s]
[ss1 ... q+1 = o
and this equation expresses the fact that s can be written as a function
of S, ... , S . If We take S, ... , S as the first q + 1 COOrdinates
1 q+1 1 q+1
~q+ 1 e, ... ,
of (x), we have in the region considered

(6.8)

where s is a function of e, ... ,

~q+ 1 only. Now in the Xq+l of these
variables we take a solution of the non homogeneaus equation

(6.9)
Then
(6.10)
and this proves that a simple covariant (q + 1)-vector is a gradient
product and consequently the rotation of some q-vector in an irr (~")
0
 6. The invariant operators Rot and Div. 85

if and only if its rotation vanishes in an 9C ( ;"). As a corollary it follows

0
that every n-vector given in an 9C ( ~") is a gradient product in an 9C ( ~).
0 0
If the (q + 1)-vector is not simple it is still true that it is the rota-
tion of a q-vector if and only if its rotation vanishes. This can be proved
by induction. Let the theorem be proved for every value of q,
(1 ;;:;: q +
1 :;;;; n - 1) in an X.,_ 1 . Then it can be proved that it is
+
also true in an Xn . 1) N ow if n;;;;;; 3 the (q 1)-vector is always simple
and the theorem has been proved for simple (q 1)-vectors. Hence it +
is true for all values of n. The q-vector is determined to within an addi-
tive term whose rotation vanishes and it can be found by quadratures
only. 2) We give here the exact formulation of the theorem for (q 1)- +
vectors and for (p -1)-vector Ll-densities: 3)
+
If W.:t1 ?.q+~ is a covariant (q 1)-vector, 1 ;;;;;; q 1:;;;; n, analytic in +
an 9C (;"), it can be written in some 9C (;") as the gradient of a scalar for
0 0
q = 0 and as the rotation of a q-vector for q > 0, if and onl'y if there exists
an 9C (;") where Rot W = 0. This condition is identically satisfied for
0
q + 1 = n.
If "1 "P-1 is a contravariant (p -1)-vector Ll-density of weight 1; +
0 ;;;;;; p -1 ;;;;;; n -1, analytic in an 9C (~"), it can be written in some 9C (~")
0 0
as thc divergence of a p-vector Ll-density of weight 1, if and only if +
there exists an 9C (;'') where Div = 0. This condition is identically
0
satisfied for p -1 = 0.
There are of course theorems of exactly the same form for covariant
+
W-(q 1)-vectors and (P -1)-vector densities of weight 1. +
Both theoremsexpress conditions of integrability. For the equations
(6.11)
1) P. P. 1949, 1, II 11. The whole theorem isproved there without using the
theory of linear partial differential equations.
2 ) GouRsAT 1922, 1, p. 105ff. Using the homotopy operator known from the

theory of graded rings (H. CARTAN 1949, 1, p. 107) a very elegant expression can
be found. If
W,t1 q = f.t1q W'J
and

U.t1 g-1 -def q"

1 I:A g ft=J
f A1 g (t"
'> t) st s-1 d t

t=O
it follows immediately that w = Rot u. If we write u = Op w it is proved easily that
RotOpw- Op Rotw = w
for every field w. (From a personal communication of N. H. KuiPER.)
3) VoLTERRA 1889, 1 for Rn; BROUWER 1906, 1, p. 22 for R.,.
86 II. Analytic preliminaries.

the conditions are

(6.12}
and for the equations
(6.13} 0I' iV"""P=""P
we have the conditions
(6.14}
We remark that these integrability conditions are independent of
the manner of transformation of the obiects concerned. For instance
if wpA, ... t[> alternating in f-l ~ ... Aq, is any geometric object whose com-
ponents with respect to (u) satisfy (6.12}, in general no such relation
holds for the components with respect to an y other coordinate system (u').
But nevertheless there exist functions U'.<, ... Aq. of the ~" such that (6.11}
holds.
Here the question arises whether there exist solutions of equations
of the form
(6.15}
where A is a collective index standing for any set of co- and contra-
variant indices and where nA, ... AqA and llpJ., ... .~.tA are alternating in the
first q and q + 1 indices respectively. The equations are not invariant
and n and Jl need not be tensors. But they are closely connected with
invariant equations to be dealt with in the next chapter. In order to
show what is happening we take the simple example
(6.16}
We may consider (6.16) as a set of n 2 equations by giving u and e all
values from 1 to n. For instance if we take u = 1, e= 2, the integrability
conditions of this special equation are
(6.17} B[.ll;;.l ~ 2 = 0
and this proves that the integrability conditions of (6.16) are
(6.18} a[.ll;;.J~e = o.
It can be proved in the same way that the equation
(6.19}
where A is a collective index and where n and ll are alternating in the
first p and p -1 indices respectively, has the integrability conditions
(6.2o} a.n""PA = o.
If q = 0 we can go a step further and consider the case
(6.21} 8"n8 =ll" 8 W,nc)
 6. Theinvariantoperators Rot and Div. 87

where the right hand side depends also on the n 8 (the n 8 need not
be components of a geometric object!).l), 2) (6.21) can be written in the
form of a system of total differential equations
(6.22) dn8 - II~tB (;", nc) d~~' = 0.
Now let Q (~", n 8 ) be an integral function (hence Q = const. an integral)
of this system, then Q is a solution III 5) of the homogeneaus linear
system
(6.23)

with the independent variables ~" and n 8 . This system is complete if

and only if
(6.24) (-8r II JB ) ~-+
o n r 1c1~-~~-
oll~tJB o
~' o:rcs v, o:rcc o:rcs

(where &. denotes differentiation with respect to ~, the n 8 being left

+
constant) depends linearly on the operators 81' II~tB ~ in the left
u:rcs
hand side of (6.23). Butthis is only possible if
oll~tJB - o
(6.25) - -8 [v II~tlB
8 [v II~tlB- + n [ICI-<>~--
u:rcc

Now let these equations, which do not contain any derivatives of n 8 ,

be symbolized by
(6.26)
If we apply 8"' to (6.26) and eliminate 8rv nA by means of (6.21) we get
a second set of the same kind, symbolized by

(6.27)
and this process can be continued ad infinitum. We need not consider
the case when a relation between the ~" can be derived from all these
equations, because in that case we could diminish the nurober of in-
dependent variables. So there are only two possibilities. Either after
i + 1 steps the (i + 1)th set contains an equation which is not consistent
with the former ones, in which case (6.21) has no solutions. Or the
+
(j 1) th set depends on the foregoing sets. Forthis latter case VEBLEN
and J. M. THOMAS 3 ) have proved that then all further sets depend on
the first i ones. Collecting results we get the rule:
In order to find the integrability conditions of equations of the form
(6.21), where the II~tA depend on the ~" and the n 8 , we have to form the
1) GuREWITCH considered 1933, 1 equations ofthe form O[l'w;., ... Aq] = Pr~tw.<, ... Aq]
2) Cf. for literature ScHOUTEN 1925, 1, p. 442ff.
3 ) VEBLEN and J. M. THOMAS 1926, 1; cf. T. Y. THOMAS and LEVINE 1934, 1;
T. Y. THOMAS 1934, 2, Ch. X.
88 11. Analytic preliminaries.

integrability conditions (6.26), (6.27), and so on. If these conditions are

inconsistent, (6.21) has no solutions. lf the jirst i sets do not contain in-
+
consistent equations and if the (i l)th set depends on- the foregoing sets,
the equation (6.21) together with the i first sets form a totally integrable
system . 1) This means here that if n 8 and ~, 8 satisfy Il"' 8 = Il"' 8 (nA, ;")
0 0 0 0 0
and the first i conditions of integrability for e" = 0;", there exists exactly
= II,.B at e"
one Solution of (6.21) such that 7/:B = 7/:B and 01-' 7/:B 0

00 0
lf besides (6.21) a Set of equations is given in 7/:B and e" which does
not contain derivatives of n 8 , this set must be adjoined to the first
integrability conditions. 2)
As we see the construction of the integrability conditions depends
always on the possibility of eliminating at each step the derivatives
of the n 8 Hence the same method can in general not be used for an
equation of the form (6.15) if the right hand side depends on the ;"
and the n, because the derivatives of n arising at every step do not
in general appear in the form or"'n.<, ... .<,] 8 Only if they happen to appear
in this form the elimination can be effected and in that case the inte-
grability conditions can be constructed in the way described. But in
the more general case the problern whether a set of equations has solu-
tions can only be solved either by the theory of CARTAN or by the theory
of RIQUIER. 3)
It often happens that the Il"' 8 in (6.21) are linear homogeneaus in
the n 8
(6.28)
Then for total integrability the equation
(6.29) {or,.II~ B + IIf"' c II B} nA
1 1 = o
must be satisfied identically. Accordingly the conditions of integra-
bility are in this case
(6.JO) or,.II~ B + IIt. c II B = o.
1 1

If an X".inXn is given, the section of a field w.~, ... A,; q~ m with X". is

(6.J1) 'wb1 b1 -- B"'Ar w .

b1 bg A, A1 '
B"b def ob I:A
S"

and the section of its rotation is

(632) 'W:cb, ... b1 -- (q + J) B"'A,
c b ... g
bg 0[!-' wA
1 1 .lg] -
- (q + 1) J-',
[cb 1 . g
b1 ]
01-' w1 .lg

1) Cf. II 5.
2) J. M. THOMAS 1927, 1.
3) Cf. also for Iiterature P. P. 1949, 1, Ch. X.
 7. PFAFF's problem. 89

We prove that the rotation of the section equals the section of the rotation.
In fact

Exercises.
II 6,'1. If w1., ... p-,n=0 and Dw=O the field w does not depend
Oll ~n .1}

II 6,2. 2} If for every choice of the q-vector P;., ... ;.. the rotation of P;., ... ;..
vanishes always if the rotation of Q[;. 1P;., ... Aq] is zero, it follows that
II 6,2 ~) Q;" = ~A~; ~ = constant.

7. PFAFF'S problem. 3 )
A field W;, in Xn, analytic in some 91(~"), fixes an En-rfield or x~- 1 .
0
For such a field the outer problern is trivial because it is immediately
clear that the En-1-field is xn-1-forming if and only if W;, is a gradient
to within a scalar factor. Only the inner problern is interesting, in fact
we have here the most simple form of this problem.
We form the rotation W.u 1. = 2 or.u w;.J and the multi vectors
1 2 3 4
(7.1a) I 1.=w1,; I,u 1.=W,u 1.; Iv,u.l=WrvW,u.l]; Iwvp.l=TV(w,,W,u.l]; etc.
or, using the abridged notation of I 7
1 2 3 4
(7.1 b) I = w; I = W; I = [w W]; I = [W W]; etc.
Then if
K K+l
(7.2) l =I= 0; I =0,
s s
and consequently I =I= 0 for s;:;;;: K and I = 0 for s > K, we call K
the class of w. K is the A.-rank of the set w1_, W.u The rank 2e of W
is called the rotation class and the A.-rank k of w, wr. W.ul 1. the similarity
class of w. Thcre arc two cases:
In case I w does not belong to the domain of W (cf. I 5). The
A.-domain of the set w", W.u.l contains both the domain of Wand the
ze 2e+1
vector w outside it, hence K = 2 e+ 1 and k = K. I and I are both
1) CARTAN 1922, 1, p. 72.
2) NAGABHUSHANAM 1949, 1.
We give here only a brief summary of results. Cf. for a detailed treatment
3)

for instance P. P. 1949, 1, Ch. IV, also for literature.

90 II. Analytic preliminaries.

simple multivectors and gradient products because their rotations

2o K
vanish (cf. li 6). Hence [ is Xn_ 211-forming and 1 is Xn-K-forming
and the corresponding complete systems are
(7.3) (5~): [}111] = 0
k
(7.4) (5~): [I TJ = 0; (k = K) 1).
The Xn_ 211 's are called the supports of rotation of w and the Xn-K's
2Q K
the supportsandalso the characteristics of w. Because I and I are both
gradient products and I lies in the domain of 1, there exists a scalar p
11

suchthat
K 2p 2Q
(7.5) I=[wi] =[pi].
In case li w belongs to the domain of W, hence K = 2 e and k = 2 e-1.
To prove this take the coordinate system at a definite point such that
w= ~. W = [~ ~] + + [~ "i 1]. The rotation od does not vanish because
2Q k k
I = 2DI. But now it follows from (5.16a) that I is Xn_,.-forming.
I is simple and a gradient product, hence Xn_ c.>-forming,
11
2 because its
rotation vanishes. The corresponding complete systems are
k_
(7.6) (5~): [I/] = 0
, 2g_
(7.7) (51 ):[I t] = o.
The Xn_ 211 's are called the supports of rotation and (because K = 2e)
also the supports of w and the xn-k 's are called the characteristics of w.
Because flies in the domain of }!! there must exist a vector uA such that
2(! "
I = [ui]. Moreover we can prove that it is always possible to make
uA a gradient. Let
a
(7.8) F(~") = const; a= 1, ... , k

"
be the equations of the Xn-k's formed by I. Then every tangent En-k
"
of them has the (n- k)-direction of I, Hence there exists an equation
of the form
" r 1<
(7.9) I=-r[F ... F]
from which it follows that
29 r 1: __ "
(7.10) I = 2 ['r F ... F] = 2 [log -r I].
1 ) As in P. P. 1949, 1, IV 2 we denote the complete systems (7.3, 4) by S~
and S~ and their adjoint systems by S 1 and s,.
 7. PFAFF's problem. 91

Note that in both cases 2e is the greatest even nurober ;:;;:; K and k
the greatest odd nurober ;:;;:; K.
Transformations of the form
(7.11)
are called similarity transjormations of the field w. Such a transforma-
tion transforms W into
(7.12) 'W = a W + 2 [a w1
2Q k
and accordingly I and I into
2e 2e ze-1
(7.13) 'I=a!li+2eae- 1 [a I]

(7.14)
Hence k and the characteristics are invariant for similarity transjorma-
tions.
K K+2 K+l
In case I 'I =f= 0 and 'I = 0. For 'I we get

(7.15)
K+l K+l
If a is a solution of (5~) we get 'l = 0, but if it is not a solution, 'l does
not vanish and we pass from case I to case II. Of course this latter
possibility can never occur for K = n.
k k+2 2Q
In case II 'I =f= 0 and 'I = 0. From (7.13) we see that '[ vanishes if
and only if a is a solution of the non homogeneaus linear equation
_ _ 2Q-1 2Q
(7.16) 2e[loga I J =-I .
This equation implies that a is a solution of (5~) and not a solution of (5~),
but these conditions are only necessary and not sufficient. Comparing
1
(7.16) with (7.10), it follows that a =-r-eis a solution of (7.16). Collect-
ing results we see that by means of a similarity transformation it is
always possible to pass from case II to case I with 'K =K -1 and for
K < n also from case I to case II with 'K = K + 1.
A transformation of the form
(7.17) (u not constant)
is called a gradient transjormation of the field w. Such a transformation
2Q k
leaves W invariant and accordingly transforms I and I into
2Q 2Q
(7.18) 'I =I
k k k-1
(7.19) 'I =I+ [ui ].
 92 II. Analytic preliminaries.

Hence 2 e and the supports of rotation are invariant for gradient trans-
formations.
2e 2e+2 K
In case I 'I =1= 0 and 'I = 0. Because K = k we get 'I = 0 if and
only if u is a solution of the non homogeneaus linear equation
2e K
(7.20) [uiJ=-I.
This equation implies that u is a solution of (5~) and not a solution of
(5~), but these conditions are only necessary and not sufficient. In
this case it follows 'K = 2e =K -1 so that we get from case I to case II.
Comparing (7.20) with (7.5) it follows that u = - e is a solution of (7.20).
K K+2 K+l
In case II 'I =I= 0 and 'I = 0. For 'I we get
K+l K
(7.21) 'I= [UIJ.
k+J k+J
If u is a solution of (5~) we get 'I = 0 but if u is not a solution, 'I does
not vanish and we pass from case II to case I. Of course this latter
possibility can never occur for K = n. Collecting results we see that
by means of a gradient transformation it is always possible to pass
from case I to case II with 'K = K -1 and, for K < n, also to pass
from case II to case I with 'K =K + 1.
Here is a table of results 1 ):

5imilarity transformations

(7.22) K=2 + 1{'K=K+1; aisnot a solution of (5~) (never for K=n)

(! 'K =K; a is a solution of (5~) (always for K =n)

(7.23) K=2e {'K =K; aisnot a solution of (7.16)

'K=K-1; a is a solution of (7.16)

Gradient transformations

(7.24) K =2e {'K=K+1; u is not a solution of (5~) (never for K=n)

'K =K; u is a solution of (5~) (always for K =n)

(7.25 ) K _ 2 + 1{'K =K; u is not a solution of (7.20)

- (! 'K=K-1; u is a solution of (7.20).

If a field w is given, K can be found by means of differentiations

and algebraic operations only. Then by altemately applying a gradient
transformation and a similarity transformation the dass can be reduced
1 ) FROBENIUS 1879, 1.
 7. PFAFF's problem. 93

to zero after K steps. But this proves that w can always be written in

l
the form
1 for K odd
W;.=s#;.+if;.+ +H;.; s- {
(7.26) O for K even
a def a
s;.=8;.s; a=1, ... ,(!.

This form is called a canonical form for w. There can not be a canonical
+
form with less then (! e terms, because in that case we see, by forming
the rotation, that the dass had to be smaller than K. The K variables
occurring in (7.26) must be independent functions of the ;". Because
otherwise w could be expressed in terms of less than K variables and
K
that would imply that I = 0. But these K variables are not uniquely
determined. For instance for K odd
(7.27)
is another canonical form for w.
Taking the K variables in (7.26) as the first K coordinates of (x)
we get

(7.28) (
w" = /" + ; ~ + +
2 ;K- 1 ~ for K odd
W;. = e J). + ... + ;K- 1 ~ for K even.
We now describe the process leading to a canonical form in more
detail. If K is odd we have first to find a gradient transformation
'w.! = w"- 8;. f, diminishing K by 1. According to (7.25) --# is an
arbitrary solution of the non-homogeneaus equation (7.20). If K is even
we have first to find a similarity transformation 'w" = i- 1 W;. diminishing
K by 1. According to (7.23) 1- 1 is a solution of the non-homogeneaus
equation (7.16). Now it seems that (7.20) and (7.16) are each a set of
(;) equations instead of one equation. But this is not true. Because
this is a property independent of the choice of the coordinate system
we may suppose that (x) is chosen such that (7.28) holds. Then for
K odd (7.20) takes the form
0 0
(7.29) 81 s=1; 8bs=0; b=K+1, ... ,n
with the general solution
(7.30) / = e +<PW, ... , ;K)
with the arbitrary function <P. If K is even (7.16) takes the form
1 -z1_1 (t:18
\> 1 + \>t:3 88 z1 +- ...
1 z
+ t:_K-18'K-lz1) = O,
\>
(7.31) ( 1
ob z = 0; b= K + 1, ... , n
 94 II. Analytic preliminaries.

whose general SOlution is any function of e, ... , ~K Which is homogeneaUS

of degree one in ~ 1 , e, ...
'~K- 1 . Of course the equations (7.29) and
(7.31) do not require any integrations. But we must remernher that in
order to get the special coordinate system we need a canonical form
and that this form can not be formed without integrations.
If w is Xm-enveloping, there exists in the region considered an en-
veloped normal system of Xm's with equations of the form

(7.32) PW) =c"; x=m+1, ... ,n

and because w,~ is, at every point, tangent to the local Xm, it can be
expressed linearly in the o,~F"

(7.33) W;.=ocxo;.F"; x=m+l, ... ,n.

From this equation it follows that K = 2e + s ;S;: 2(n- m) or

m :S:: n - e - s. But from (7.26) we see that
o o 1 1 Q e
(7.34) ss=sc; s=c; ... ; s=c

represents an enveloped normal system of xn-e-.'s. Hence V def n- e-s

is the maximum value of m. This proves
Every field W;. of class K = 2 e + s is x.-enveloping but not x.+l-envelop-
ing; v=n-e-s.
If (7.28) holds we have

s=1 s=O
K
I = 2P p! w... ~KJ }e = 2P p! w ... ~KJ
k
(7.35) 2e
I=~ p![~
9P -2 -K
... ~] I= 2P-1 (p-1)! {e [~2 ... ~KJ +
+ e c~1 ~2 ~4 . .. ~KJ + ...
+ ~K-1 w... ~K-2~KJ}
and
for s= 1
supports = characteristics: e=const.; ... ; ~K=const.
supports of rotation: e=const.; ... ; ~K =Const.
(7.36) for s = 0
supports=supports ofrotation: ~ 1 =const.; ... ; ~K =const.
e:e: ... :~K- 1 =const.;
characteristics : {
e=const.; ~ 4 =const.; ... ; ~K=COnst.
 8. The theorem of STOKES. 95

Exercise.
II 7, 11). Prove that the dass of
a) x 1 x 3 dx 2 + x 1 x 2 dx 3 + (x 1 + x 3 x 5 ) dx 4 + x 3 x 4 dx 5
b) x2 d x1 + x1 d x2 - x3 x5 d x4 - x3 x4 d x5 + x2 d x6
is 4 and 5 respectively.

8. The theorem of STOKES.

In X,. we consider an Xq+l with the equations

(8.1) ~q+2 = 0; ... ; ~" =0
with respect to some coordinate system (x) and in this Xq+l we consider
a (q + 1)-dimensional volume Tq+l and its boundary -rq, such that from
the curves on Xq+l
(8.2) e = const.; ... ; ~a- 1 = const.; ~0 + 1 = const.; ... ; ~q+ 1 = const.
for every value of a from 1 up to q + 1 each intersects Tq in at most
two points. Let an inner orientation be fixed on Tq+l by e", ... , e" in
1 q+1
this order. Then the volume element d /"' "+~ with this orientation
rnay be written in the form
(8.3) dj"I"H = (q + 1)! 1d ~[x, . q+1
d ~><wl.

If the q + 1 vectors d ~", ... , d ~" are chosen in such a way that
1 q+1

ae=ae;
1
ae=o;
1
... ;
(8.4)

we have
(8.5)

and all cornponents of d f"' " which have an index q + 2, ... , n vanish.
Let one of the curves (8.2) for a = 1 intersect the boundary Tq at the
two points

(8.6)

1) CARTAN 1899, 1, p. 259 and 265.

96 II. Analytic preliminaries.

The points of Tq+l satisfying the equations

(8.7) e-;;;,e-;;;,e;
1 2
~-;;;,~a-;;;,~a+d~a;
0 0
a=2, ... ,q+l

+
constitute a (q 1)-dimensional tube in Tq+l cutting of a q-dimensional
element from Tq at each of the points P1 and P2 .
2

FigiO.

Fig.10 illustrates the case n=3, q=,2. For convenience the coordinate-
X1's are drawn in this figure as straight lines but this has no real signi-
ficance. If we fix now an inner orientation all over Tq by e", ... , e"
2 q+l
in this order at P2 , then this orientation is opposite to the orientation
of e", ... , e" at ~ Theorientations in Tq+t and Tq are now chosen in
2 q+l
such a way that the direction from a point of "tq+t towards the boundary
followed by the orientation of Tq gives the orientation of Tq+t Accordingly
we have for the q-dimensional element of Tq at P2

(8.8)
and at ~

(8.9)

After these preliminaries we now consider a covariant q-vector field

vx, ... -' satisfying the following conditions:

a) the v-'-' are continuous in "tq+l and on Tq;

b) the derivatives which occur in 8r'" v-'-'l exist at all points of Tq+l;
 8. The theorem of STOKES. 97

c) these derivatives are continuous at all points of <q+l except at

the points of a finite number of Xq's.
We consider the integral of o1 v 2 ... q+ 1 over <q+ 1 (cf. 8.5)
(8.10)

As a consequence of the conditions for the field v,~, ... .~. the part of the

l
integral over the tube is
2 1
( V 2 ... q+1 - V 2 ... ) df:" 2 d~q+ 1
(8 .11 ) q+1
. 2
_>!:_V
- 2 ... q+1
2
dj2 ... q+1 + 1z ... q+1 dj2
V
1
... q+1

2 1
where v 2 ... q+1 and v 2 ... q+1 are the values of v2 ... q+1 at P 2 and at P1
respectively. Hence
(8.12) J 01 V
2 ... q+1
djl...q+ 1 _'!<_
-
J V
z ... q+1 dj 2 ... q+l
Tq

By using the other q sets of coordinate-X1's, equations similar to (8.12)

but with 82 , ... , oq+ 1 instead of 81 , can be derived. By addition we get

(8.13)
I J orl' v;., ...
Tg+t
.<9J djl'" ... .<q = J v,~, ... .<q df" .. ;.i 1).
Tq

In order to get rid of the condition of intersection imposed on the

form of <q we consider the case where <q+l can be split up into a finite
number of <q+l's such that the condition of intersection holds for the
boundary of each of them. Then (8.1)) holds for each of these Tq+ 1 's
and by summation of the integrals all integrals over boundaries between
two adjacent <q+l's cancel each other out, because the orientations in the
common boundary are opposite. Hence (8.13) is proved now for this
more general case.
(8.13) isaform of the generalized theorem of STOKES (in Germany
also named after GAuss and in France after STROGRADSKI). Many
other forms can be deduced from it. First 6" .. "P, .<p- and f...... .<p L. ...
(p = n - q) can be introduced instead of v.<, ... ;.., /~'" .. " and /"'"". Then
we get four other forms
(8.14)

(8.15)

1 ) PoiNCARE 1887, 1; 1895. 1; BROUWER 1906, 1; 2; 1919, 1; WEITZENBCK

1923, 1, p. 398ff., also for ]iterature; WEYSSENHOFF 1937, 1; SCHOUTEN and
V. DANTZIG 1940, 1; LEVI CrVITA 1940, 1.
Schouten, Ricci-Calculus, 2. Aufl. 7
98 11. Analytic preliminaries.

n- q
+
q 1
J fJ ij[A, ... A"-tl.ul df""'-""rl
.u
(8.16) { 'rt+l
= f fjp ij[A, ... Ap dfl.uf"t"fl = f ij[A, ... Ap df" ... "f]
Tg+l Tl

which all have the same geometrical signification as (8.13). If the

elements of -rg and Tq+I have outer orientation (this is the case that
occurs most frequently in physics) and if these orientations are chosen
in such a way that the orientation of -rg+l followed by the direction
from a point of ?:q+I towards the boundary gives the orientation of ?:q
We get the following five equations for a field VA, ... Aq

(8.17)

(8.18)

(8.19) P J fJ.u bA, ... J>-.U dfA, ... .ap_, = J bA, ... p dfA, ... Ap
Tg

q+l
n- q
.u
J fJ b[A, ... Ap-f.ul d jA~>"'-""fl
(8.20) { Tt+l
= J o,..b[A, ... J>dil.uf", ... ..,J = J b[A, ... .a"df"' ... ""J. 1)
Tg+l Tf

All integrals which have occurred up till now in this section have
been scalars or n-vector Ll-densities of weight 1. In fact only these
quantities can be added even if they belong to different points. But
other forms of STOKEs' theorem may arise if we restriet ourselves
to less general transformations of coordinates. For instance, if only
those coordinate transformations with LI > 0 are allowed, W-scalars
also can be added and thus besides (8.13) we have in that case the
invariant formula
(8.21) J fJ[,u VA, ... A,] df,uA, ... Aq = J VA, ..,.Aq dfA, ... Aq 0

Tg

If transformations of Ga only are allowed, that is, if we areinan E,.,

also quantities that belong to different points can be added, provided
that they have the same manner of transformation. Accordingly in E,.
the following equations hold
(8.22)

(8.23) P J [f [A, ... p-1 fJ p] -oQ-

- fdf A, ... p -oQ

1 Cf. T. P. 1951, 1, p. 71 f. for the formulae occurring most frequently in physics.

 9. Anholonomic coordinates 99

(8.24) f d f'"A, ... Aq 81' - 0 Q = f df A, ... Aq - 0 Q

(8.25) Pf df [A, ... Ap-1 8Ap] -oQ-fdf
- A, ... Ap -oQ
Tq

where Q is a symbol with suppres"ed indices standing for any quantity

and where -o symbolizes an operation performed on the indices A1 , ... , Ap
or A1 , ... , Aq and the indices of Q, which is composed of alternation,
mixing and transvection 1).
For instance if
(8.26)
we have in E 3
(8.27) f iJ A d f = f ~~ A d fI'
Ta T2

but this equation has no invariant meaning in X 3 .

The STOKES integrals may be looked upon from a quite different
point of view, which is very important for many investigations. Take
for instance (8.13). The vA, ... A. may be considered as (;) functions of
the ~" that need not be components of a q-vector or of any geometric
object. Then we are not able to give a corresponding formula foranother
coordinate system (x') but the formula (8.13) remains true. This enables
us to use STOKEs' formulae in all cases where the manner of transforma-
tion of a set of variables is not known or not yet known. Of course in
this case the integrals are not scalars because transformation of coordi-
nates is not allowed.

9. Anholonomic coordinates ( cf. II 5 ).

If at every point of Xn we introduce a set of n linearly independent
contravariant vectors e" and the reciprocal set tA; h, i = 1, ... , n, in such
a way that these fields 'are analytic (or of class u) in the region considered,
these vectors are contra- and covariant basis vectors of a coordinate
system ~h if and only if the rotations 8rl' ~l vanish at all points. In
fact 8rl' t;,l = 0 are the integrability conditions of the system 8;} = t.
If these conditions are not satisfied, the system of reference formed by
the fields e", S, is called an anholonomic coordinate system (h) 2) in
'
contradistinction to the coordinate systems used before, that we call
1) EI 1935,1, p.1JO.
2 ) Cf. for Iiterature the footnote 1 on p. 78 and Ch. V 7. Though we often

prefer latin running indices for anholonomic coordinate systems, this is not ob-
ligatory; also (x) can be taken anholonomic.
7*
 100 II. Analytic preliminaries.

holonomic. The t;" determine a net of X 1's (cf. II 4), called the net of the

anholonomic coordinate system. But in general an anholonomic coordi-
nate system has no coordinate-Xm's, 1 < m < n. The components of any
quantity with respect to (h) can be derived from the components with
respect to (u) by using the intermediate components A1, Ai of the
unity tensor
(9.1) A.'! def J.~oe" * e"
t
just in the same way as if (h) were holonomic. But the components
of a more general geometric object (as defined in II 3) with respect
to (h) can not be derived by using A1, Ai and o; def Af o,.. just as if (h)
were holonomic. In some way the components with respect to (h)
must first be fixed by the introduction of a condition. 1)
From now on we sometimes use the sign " in order to emphasize
the fact that an equation is only valid with respect to holonomic Co-
ordinate systems. We write (d~)" instead of d~" for the components
of the linear element d~" with respect to an anholonomic coordinate
system (h) because variables ~" do not exist.
The expression
(9.2) nh def AP " A" * p J. " h
;:,~ii = i i U[p ] = ~ ~ U[pe]
1

is called the object of anholonomity. It is a geometric object all of whose

components vanish with respect to all holonomic systems and only
with respect to such systems. For two holonomic or anholonomic
systems (h) and (h') it follows from this definition that
;:,~i'i' = Ai'i'h D;t + Au'Ai'J o1 A,
nh' i i h' h i i h'
(9.3)
(9.2) is a special case of this formula.
If any expression with respect to holonomic coordinates is trans-
formed with respect to an anholonomic system, correction terms appear,
all containing the bject of anholonomity. For instance 2)
(9.4) oc; oil p = Af; o 1 ~' 1 Af1 o;. p = (Afi o1,.. 1 At1) A1 o" p = - Dt, o" p.
If an Xm in X,. is given by its parametric equations
(9.5) ~" = f"(r/"); oc, , y, ~ = i, ... ,m
we may introduce besides (u) and (oc) the anholonomic coordinate systems
(h); h,i,j,k=1, ... ,n and (a); a,b,c,d=i, ... ,min X,. and Xm
respectively. If '{J~b is the object of anholonomity of the xm and if
the xm is rigged, we have
1 ) For an example cf. III 9. Here the condition for the geometric object F

is that the equations (III 9.1) hold for an anholonomic coordinate system.
2 ) Cf. Exerc. II 9, 1.
 9. Anholonomic coordinates. 101

(96)
Biia 4 ii = BP""8
cbh;:,nh cbh [p A"A] = B"8
h [c B"
b ] - BPB"8
[c lA PI B"b] = B"8
h [c B"
b] +'n"
il> 4 cb

where
(9 .6a) BPc
BfJc B~'
det
-
{J
Bidet
c-
BfJc BP{J Ai.
JJ'
B"Adet- B"a. Ba.
A
B~ det B" Ba. A~
- a. A
If in this case we choose at all points of xm the first m vectors '!"
coinciding with the vectors za. and the last n- m vectors ( in the En-m

of the rigging, we have, taking a, b, c, d = 1, ... , m and x, y, z, u =
m+1, ... ,n

(9.7)
A1 * #A; A~ * (; B * tp; B~ * ea.; Bi~ePAi = eiYL(i
( b c- c JJ c - ''
B~ * ef1
c
B~ = eP;
c
B1 * ta. B~ = t"; Bf * t" AZ = t_ * bf,
hence
(9.8)
Anholonomic coordinate systems are frequently used when dealing
with an X': in X,. 1 ). If the X:;' is rigged it is often convenient to take
at each point the e"; a, b, c, d = 1, ... , m in the tangent Em and the e";
b y
x, y, z, u=m+ 1, ... , n in the tangent En-m Forthis choice we prove
that the X:;' is tangent to oon-m Xm's if and only if Q';b * 0. In fact,
the fields e" are Xm-forming if and only if the eP 8IPI e" depend linearly
b [c b]
on the e", that is if
b
(9.9) 0 = [ceP (81PI b]e") !;_ = - [ceP b]e" 8/J J" *- Q';b.

All other applications of anholonomic coordinates will be postponed

until after the introduction of a linear connexion in Ch. 111.
Besides the {J~ components of another kind are often used, these
are the intermediate components ((u) holonomic)
(9.10 ) nh def Aii
;:,,:pA=
nh _ 8 A"A]=
p;,.;:..:ii- [p
* 8[peJ.]
h

The n objects D~l resulting from strangling the index h are bivectors.

Exercises.
II 9, 1. Prove the farmulae for (u) holonomic
119,1cc.) +
8r/J w;.1 = Ati (8[i w.1 w,.D~.);
119,1) 8P tJP =LI (8itJi- 2Dt, v); LI = Det (A1)
where w;. is a covariant vector and tJ" a vector Ll-density of weight + 1.
II 9,2 2). Prove the identity
II 9,2 cc.) 8[k {J~i] - 2Q~ki {J~]I = 0.

1 ) Cf. for applications to anholonomic dynamical systems and references to

literature T. P. 1951, 1, Ch. VIII.
2) Cf. (II 12.tb).
102 li. Analytic preliminaries.

10. The LIE derivative. 1), 2)

Let the points of a region R be subject to a point transformation
(10.1)
The functions f" are supposed to be analytic in the region considered
and to establish a one to one correspondence between the points of R
and those of some other region R'. Another coordinate system (u')
is now introduced such that the transform in R' has the same Coordi-
nates with respect to (u') as the original point in R had with respect
to (u). Hence if
(10.2) ~" = qf(r/)

is the inverse of (10.1) and if rj" and r( are the coordinates with respect
to (u) and to (u') of the transform of f', we have
(10.3) r( =Cf:,'~"= tJ:' rp"(r/).
Accordingly the transformation of ~" into ~"' and vice versa is given
by the equations
(10.4) ~"' = tJ:' rp" (~"); ~" = /"(~"') 0

This process is called the dragging along 3) of the coordinate system (u)
by the point transformation (10.1) and (u') is called the coordinate
system dragged along. The dragging along of a coordinate system (u)
can be interpreted geometrically in the following way. Let the region
considered be filled up with gelatine and let a coordinate net of (u)
with sufficiently small meshes be made visible by red curves in the ge-
latine. Then, if the material points of the gelatine suffer the point
transformation (10.1), the red curves in their new position form a Co-
ordinate net of the system dragged along (u').
Now let the components of a geometric object (/) A as defined in
"' in R' whose components
II 3 be analytic in R. Then we form a field (/JA
"'
<PA'with respect to (u') at each point of R' are equal to the (}) A at the
corresponding point of R. This process is called the dragging along of
a field by the point transformation (10.1) and the field "'
<PA is called the
1 ) This section was written in collaboration with A. NIJENHUIS. For literature
cf. SLEBODZINSKI 1931, 1; SCHOUTEN and V. KAMPEN 1933, 1; EI 1935, 1, p. 142;
E li 1938,2, p. 161; P. P. 1949, 1, Ch. Il; T. P. 1951, 1, Ch. IV; NIJENHUIS 1952, 1.
2 ) The LIE derivative is defined here for an Xn- It can be defined for more

general spaces for instance path-space, that is an X,. in which a system of curves
is given such that through any two points there passes one and only one curve
of the system. See YANO 1945, 1; KosAMBI 1952, 1.
3) "Mitschleppen" in SCHOUTEN and V. KAMPEN 1933, 1 and EI 1935, 1.
 10. The LIE derivative. 103

field dragged along 1). The dragging along of a geometric object can be
interpreted geometrically. Every geometric object as defined in II 3
can be represented by some figure or figures in Xn, for instance a contra-
variant vector in a point by an arrow on an infinitesimal scale. Let
the points of these figures be marked blue in the gelatine filling up the
region. Then, if the gelatine is transformed, the transformed figures
represent the object dragged along.
From the definition of dragging along it follows that (cf. II 3)
(1o.s) J)A, {n} = !51.(/)A {e},
hence, making use of (3.1 b)

+
lf the transformation is infinitesimal: e"-+ e" v" dt, we use the
term dragging along over v" d t. The process of dragging along of a
contravariant vector field u" ds over
v" dt is illustrated in Fig. 11. The
vector l:" d s in e"
+ v" d t results
from the displacement of beginning- .L----;;;-::-:----av"'dt
point and endpoint of the vector v"'dt ll"'ds
u" ds in e"over v" dt. In general
l:"ds and u" ds in e"
+ v" dt do
not coincide and so there arises a
pentagon.
We now consider a vector field
v" = 1p" W) and the system of n
ordinary differential equations
d'YI" u!'ds
(10.7) _._, = "P"(n")
dt Fig. 11.

with the independent variable t and the unknown variables 17", with
functions 1p" analytic in the region considered and with the initial con-
ditions
(10.8) n" = e" for t = 0.
The solution is

(10.9)

1) "Mitschleppen" in ScHOUTEN and v. KAMPEN 1933, 1 and EI 1935, 1.

 i04 II. Analytic preliminaries.

or symbolically
(10.10)
This solution represents a point transformation, conveniently in-
tv
dicated by ~"J.E~ rj" or r;" = T ~". Its geometrical significance follows from
(10.7). According to this equation t is a parameter fixed by the field v"
on every streamline of this field if the point t = 0 on each streamline
is chosen arbitrarily. Now ~~ indicates the point transformation that
on every streamline transforms a point with parameter t into the point
+
with parameter t -r. We call this transformation the displacement over
-r v". From this it follows immediately that

(10.11)
tv
and that accordingly the transformations 1 form a one-parameter
abelian group (cf. Ch. IV).
Neglecting second and higher powers of t and writing dt instead
of t we get the infinitesimal transformation
vdt -vdt
(10.12) r;"-*-~"+v"dt or r;"=T~"; ~"'--r;"-v"dt or ~"= T r;"
which is said to generate the group of finite transformations (10.10).
According to (10.1, 2, 4) the corresponding transformations of coordi-
nates are
~"'
-vdt
J * b:' W - v" d t) or ~" = T ~"
1~"
(10.13)
b:. ~"' +
vdt
* v" d t or ~"=T~"'
and (x') is the coordinate system dragged along over v" dt. By differentia-
tion of (10.13) we get

(10.14) {
a) A1' * b~ (A~- 8;. v" dt)
b) 8I' A"'-"'--
;. -
b"'
"
8I' 8.il v" dt

(10.15) {
a) A~. * b1 (A1 + 8.il v" dt)
b) 81', A1 * 0~1 81' 8;. v" dt.
The expression
m
(10.16) f ([J ~~ Lim l/1 - l/1
v - e__,.o t

will be called the LIE derivative of the field ([J with respect to the field v"
and f: ([J dt the LIE differential of ([J with respect to the infinitesimal dis-
v
Placement v" d t 1). This can be expressed in another way:
1) SCHOUTEN and VAN KAMPEN 1933, 1,; YANO 1946, 5; 1949, 1; YANO and
T AKANO 1946, 6; ScHOUTEN 1949, 5 and for more general geometric objects LAP-
 10. The LIE derivative. 105

lf a field (/) is dragged along over v" dt the variation of the field at
~,. is the negative LIE differential of (/) at that point.
If ([JA is a geometric object, also (f/J A, o,. (/) A) are components of a
geometric object. From the definition of dragging along it follows that
m
...--"-- m m
f/J A, o,. ([JA = (/J A o,. (/) A. But this implies that f (f/J A, o,. f/J A) = Cf f/J .1,
o,. f ([JA) Now on the other hand this same LIE derivative can also
be written ( f (/) A , f o,. f/J .1), and this proves that in the expression of
V

V V
the LIE derivative of the object consisting of a geometric object and
its first derivative, fo,.f/JA may be replaced by o,.f([JA;
V V

(10.17) r. o,.f/J.1 =
V
o"r ff/J.1
V
1).

Of course the expression f o,. (/) A can not be considered as the LIE deri-
vative of o,.f/JA but only as a set of components of the LIE derivative
of (f/JA, o,.f/JA), because o,.f/JA is not a geometric object. In the same
way f u 1 is not the LIE derivative of the component u 1 but only the
V

1-component of the LIE derivative f u". Obviously (10.17) is also true

if (") is anholonomic 2). v

According to the definition we have for.a contravariant vector field

u" and the field it'" dragged along over v" d t
(10.18a)
or
(10.18b) it'"' {~} * <5=' u"{~- vdt} = b:' (u"{~}- v"o,.u"{~}dt),
hence, according to (10.15)

(A + a, v" dt) (u-'- v" o,. u-'dt)

{10.19) ( it'"== u" - v"' o,. u" dt + u" o,. v" dt
and, according to (10.16)
(10.20a)
Returning now to Fig. 9 we see that f u" d s d t is the fifth side of
V
the pentagon and that f u" vanishes if and only if the fields u" and v"
V
are X 2-forming (cf. II 5).
TEW 1938, 1 ; DE DONDER and GEHENIA U 1942, 1 ; RACHEVSKI 1949, 1; T ASHIRO
1950, 1; WAGNER 1950, 1; NIJENHUIS 1952, 1, Ch. rr. The name LIE derivative
occurred first in v. DANTZIG 1932, 1; 2.
1952, 1, p. 95.
1 ) NIJENHUIS

Cf. for LIE derivatives in anholonomic coordinates

2) NIJENHUIS 1952, 1,
II 6 and Exerc. II 10,6; 7.
106 li. Analytic preliminaries.

From the definition of dragging along it follows that for any pro-
m
,_..-.._, m m
duct or transvection of geometric objects we have (/J lJI = (/J lJI and from
this we see that the rule of LEIBNIZ
(10.21)
holds for f applied to products and transvections provided that f/J, lJI
V
and f/J lJI are geometric objects. It can be proved in a similar way that
1) the LIE derivative of a sum of geometric objects equals the sum
of the LIE derivatives of the summands, provided that the sum is a
geometric object;
2) the LIE derivative of a contraction equals the contraction of
the LIE derivative.
m
If f/J is a quantity, (/J is a quantity of the same kind and therefore
f f/J is also a quantity of the same kind. The same is not true for every
V
geometric object.
According to the definition we have for a scalar
(10.22) fs=v~'o,..s

and from this and the rule of LEIBNIZ for transvections it can be proved
that for a covariant vector
(10.20b)
because v" W;. is a scalar.
Because every tensor can be written as a sum of products of vectors
we get for the LIE derivative of a tensor, for instance P~;,:
(10.23) f
P~,; = v~' 8,.. P~,;- P':.t o,.. v" + P~,.. 8,; v~'.
For a tensor with higher valences the LIE derivative is constructed
in the same way. Each upper (lower) index gives rise to a term with a
negative (positive) sign. As a special case we mention the contra-
and covariant p-vector field
a) fu" 1 '""" = v~' 8fJ u"1 " ' " " - p u~'["""v 8fJ v" J 1

(10.24) ( V

b) fw -v~'ofJ wt ... Ap +PwfJ [.I, ... p 8tl

v At ... p- v~'

From (10.24b) we get for p =n

(10.25) { f W,; 1 ... .a,. = v~' 8,.. W.a1 ... .t,. + n w,..[.t, ... .a,. 8,; 1) v~'
= v~' 8,.. W,;1 ... .a,. + W,; 1 ... .a,. 8,.. v~' = 8,.. v~' W,; 1 ... .a,.

and because w1 ... n may be considered as the component of a scalar

L1-density of weight + 1, it follows that for such a quantity q
(10.26) fq =
V
v~' o,..q + q o,..v~' = o,..qvl' = Div (qv).
 10. The LIE derivative. 107

Since every scalar density q of weight + 1 is a product of the W-scalar w

(cf. I 2) with a scalar Ll-density of weight + 1 we have
{10.27) fV q = v~' 81' q + q 81' v~' = 81' q v~' = Div (q v).
From this we get for a scalar density 5 of weight w
(10.28)
and for a tensor density of weight w for instance $~"'

(10.29) f \13~.< = v~' 81' \13~.<- $~;. 81' v" + \13~1' 8;. v~' + w \~;. 81' v~'.
Formulae of the same form hold for scalar Ll-densities and tensor
Ll-densities.
It can be proved for more general geometric objects as defined in
II 3 that the LIE derivative of a geometric object together with the
object itself form a geometric object and that the LIE derivative itself
is always a geometric object if and only if the transformation of the
given geometric object is linear (not necessary homogeneous) in the
components with coefficients depending on ~><, ~"', A:', 81"A~', .... 1)
A field is said to be absolutely invariant with respect to the jield v"
if its LIE derivative with respect to v" vanishes. The necessary and
sufficient condition is that the field does not change if it is dragged along
by the point transformations of the one-parameter groups generated by
v" dt. Obviously sums, products and transvections of absolutely invariant
fields are absolutely invariant.
A field is said to be absolutely invariant with respect to the streamlines
of v" if it is absolutely invariant with respect to a v" for every choice
of the scalar a.
The additional condition for this stronger invariance for an absolutely
invariant covariant p-vector field w;., ...?.p is
(10.30)
This follows immediately from (10.24b). 2 ) The LIE derivative of every
geometric object ([JA as defined in II 3 can be found from the defini-
m
tion (10.16) and the form ([JA in (10.6). But the computing may be
rather cumbersome if the transformation of the object is complicated.
As an example we deal here with a rather simple geometric object
without using {10.6). Let the transformation of the object be 3 )
II"' - A~';.
'J.'- "'JI"! ' ; . - A'""
!t'J.'" !t'J.' 8I' A"'
J.

1 ) NIJENHUIS 1952, 1, Ch. I I. Cf. ScHOUTEN and VAN KAMPEN 1933, 1; EI 1935, 1;
TASHIRO 1950, 1.
2 ) Cf. for more particulars concerning absolutely and relatively invariant fields

P. P. 1949, 1, p. 43ff. and T. P. 1951, 1, p. 78.

3 ) Objects of this kind occur frequently in differential geometry, cf. Ch. III.
 108 II. Analytic preliminaries.
m
For n; 4 we get
m
(10.32) n;:J.{e} * t5:'t:-in;J.{e- vdt} = t5:'::-i-(n;J.{e}-vvovn;J.{e}dt)
hence, according to {10.14, 15)
m
( ) {n;J. = (A~- op VT dt) (Af- o,, v" dt) (A~ +Oe v" dt) X
X (fl~a-Vwow/I~adt)-(A~-opvTdt) (A-oJ. V" dt) OTOaV"dt
10 33

{10.34) f.ll~ 4
V
= vv o.fl';, 4 - /I~ 4 oe v" + n;" o4 v" + IJ;4 o" ve + o" oJ. v".
We see from (10.31) that the difference of two objects of this kind is
always a tensor, because the second term to the right in (10.31) does not
m
contain fl';, 4 Now n;J. is an object of the same kind as li';.J. and this
implies that f, n~J. is a tensor.
v m
If a field v" is given, the LIE operator f. can be used to find the field (/) A
V

that results from a field of a quantity (/) A if this field is dragged along
tv
by a finite point transformation T belanging to the one-parameter
group generated by the infinitesimal point transformation e"-+ e" + v" d t.
Let (") be chosen in such a way that v"=f" (cf. Exerc. II 5,1). Then
f. e" = 0 for every value of and this has as a consequence that the trans-
v J.
tv
formation T takes the form
(10-35)
and that
{10-36)
Now for this special choice of (") we have
(10.37)
hence
{10.38)
Because this equation has the invariant form it is independent of the
choice of ("). Every term represents the values at e" of the components
of a quantity of the same kind as (/) A.
It can be proved that (10.38) also holds for all geometric obfects
whose transformation is linear 2).
1 ) This series occurs in KosAMBI 1946, 1 but no proof is given. SHABBAR's

justification of (10.38) in 1942, 1 really shows only the correctness of the first two
terms. In 1951, 1 KosAMBI gave for the firsttime a complete proof. The proof
given here is from NIJENHUIS 1952, 1.
2) NIJENHurs 1952, 1. GoLAB 1950, 1 called these objects "purely differential".
The II~;. in (10.31) are components of such an object.
 10. The LIE derivative. 109

In Ch. IV we need the following formula 1)

(10.39) w" def e-s,"v" I

valid for all geometric objects whose transformation is linear. First
we prove for objects of this kind
-s< +s<ll
(10.40) *1p <ief e u f e u 1p = f 1p .

l
V W

For the derivatives of *1p we get (cf. Exerc. II 10,9)

d* -s ,ll +s , -s , +s ,
_____1j!_
ds
= - e " (f f - f f) e " 1p = - e " f e " 1p
U V V U W
(10.41) J2* -s< +s<ll
1

__ "P = e "f e "1p; w" def f v"; w" <!e! f w" = f 2 v"
d s2 'fi 1 u 2 u 1

Hence we have
(10.42)
and because of
(10.43)

and the definition of w", it follows that *1p = f1p. Now writing
w
x for
the left hand side of (10.39) we have according to (10.40)
d -s , - I , +s , -s , +s ,
(10.44) __X=-e
dt
"fe "e "W=-e "fe "x=-fx
V V W

and if this differential equation is solved by means of a series, we get

(10.)9).
Exercises.
II 10, 1. Prove that f v" = 0 and that the LIE derivative of A1, ~"'"n
...
V

and e.~., n vanish for every choice of v".

II 10,2. Prove that the additional conditions for absolute invariance
of a p-vector density (or J-density) tv"' .. "P of weight 1 is +
II 10,2 ll)
II 10,)2). Prove the identities for the geometrical object ~" and for
the scalar fields ~"
II 10,3 ll) f~"
V
=0
X X X X
II 10,3 ) f~ =v~-'8 ~ =v~'e
V !' f.'
=V.
1) NIJENHUIS 1952, 1, II 8.
2) NIJENHUIS 1952, 1, II 4.
110 II Analytic preliminaries.

II 10,4 1 ). Prove that the operators f and D are commutative, for

instance "
II 10,4 IX) 0[1' W.<, .. .lq] = O[p W.l, .. .lq]

II 10,4 ) f 0
V I'
ttJI'" ><p = 0 f ll:JI'""P.
I' V

II 10,5 2 ). Prove that

II 10,5 IX) f w = v. Rot w + Rot v. w
V

II 10,5 ) fkw = (v. Rot)kw +(Rot v.)kw

V

II10,5y) f W = Div v. W - v. Div W

V

II 10,5 CJ) ;tkw

V
= (Divvlw + (- v. Divlw,
where w is a covariant q-vector and ltJ a cont_ravariant p-vector density
of weight + 1. Rot has to be changed into Grad if working on a scalar
and v. is defined by:
II 10,5 s) v.w:
II 10,5 i;) v.w:
II 10,6. Prove that3)
II 10,6 IX) fuh = vi a.uh- (o.vh- 2viQh.) ui
V 1 1 7>

II 10,6 ) f W; = vi Oi W; + (o; Vh - 2vi f27j) Wh

II 10,6 y) fQti
V
= 0.
v1 oi .):1 (8; v1 + 2v1 Qj;).
(h) . (h) . . .
II 10,6 CJ) f .):1 =
V '

II 10,7. Prove that 3 )

II 10,7 IX) feh_>!<_- 2Qh.
i i 1'

II 10,7 ) J;:e" *- 2Q7;A/,.

1'

where : denotes the LIE derivative with respect to e".

1 1

II 10,8. Prove that for any three vector fields t", u", v" 4)

II 10,8 IX) f f v" + f f t" + f f u" = 0.

t U U V V t

II 10,9. Prove that for any quantity tP (indices suppressed) and any
vector fields u", v" 5)
1) P. P. 1949, 1, p. 76. This is only a special case of the commutativity of f
and oll [cf. (10.17)].
2) E. J. Posr, personal communication.
3) NIJENHUIS 1952, 1, II 6.
4) DAVIES 1938, 1.
5) N IJENHurs 19 52, 1, li 8; Taschiro 19 52, 1.
 11. The LAGRANGE derivative. 111

II 10,9 cx) (!;!! - ! !) tP = !tl>; w" def ! v" = - ! u".

uv vu w u

II 10, 10. Prove that .t v" = 0 if and only ifl)

..
sutv tvsu
II 10,10 cx) TT~=TT~".

II 10,112). In an Xn a covariant q-vector field w,~, ... Aq and a set of

oo1 Xq's depending on a parameter t are given. Let a set of curves
intersect each Xq in one point, such that t is also a parameter on each
of these curves. If then -r:q is a q-dimensional part of one Xq there is
a one to one correspondence between the points of this -r:q and oo1 -r:q's
on the oo1 Xq's and
II 10,11 cx) (/> (t) def J W A, . q
d jA, . Ag

is a function of t. Prove that

II 10,11 ) -d(])
dt =
f .tw V , g
djAAq1
Tg

II 10,12. Prove that .t o; u" = o; .tu" for (h) anholonomic.

11. The LAGRANGE derivative.

Let tP A; A = 1, ... , N be a set of functions of the ~". They may be
scalars or components of geometric objects or just quite arbitrary func-
tions for which no manner of transformation is defined. Let the index A
be suppressed and the derivatives of the tPA with respect to the ~" be
denoted by tP""tP,,.., etc.
If ~ is a function of the tP, tP,.., tP.,_.. etc. up to a certain definite
order, then, in a certain region of Xn, ~ is a function of the ~". We
consider the volume integral (cf. II 8.5)
(11.1)

over an arbitrary region -r:n where the tP are analytic.

V
Let the field (/> be subjected to a variation d tP such that the Varia-
tions of tP and of all its derivatives occurring in ~ vanish at the boundary
"'Cn-1 of "'Cn.

Then we have for the variation of ~

(11.2)
1 ) A. NIJENHUIS, personal communication.
2) E. J. PosT, personal communication.
112 Il. Analytic preliminaries.

and consequently

{11.3)

Let now "in-l have a form such that the theorem of STOKES is valid
for "in and "in-t Then we have, integrating by parts and using this
theorem in the form (8.20) for p = 1 1 )

hence
V 1 V 1 n
(11.5) df2d~ ... d~=J[2]dl/>d~ ... d~

where
{11.6)

[2] is called the LAGRANGE derivative of 2. [2] has of course a

suppressed upper index A just as l/> has a suppressed lower index A.
If [2] =0 the variation (11.3) vanishes for every choice of "in, provided
that the variation of l/> satisfies the boundary conditions.
The equation [2] = 0 is called the LAGRANGE equation. In classical
dynamics LAGRANGE equations with n = 1 occur. The f' reduce then
to one variable t and the l/> A are the coordinates of the dynamical
system. 2)
If the l/>A are components of a geometric object with a given manner
of transformation, and if 2 is computed from the l/> A and their derivatives
in the same way for all coordinate systems it may happen that 2 is
a scalar Ll-density of weight +1. In that case the integral (11.3) is
a scalar and the process of forming the LAGRANGE derivative is invariant
for all coordinate transformations. If moreover l/> is a tensor with
the valences p, q, [2] is a tensor Ll-density of weight 1 with the +
valences q, p. If only coordinate transformations with L1 > 0 are con-
sidered we may replace Ll-density by density in the above statements.
If l/> is a tensor and 2 a scalar Ll-density (or density for L1 > 0)
of weight + 1, a very important relation exists between l/>, [2] and
their first derivatives. In order to obtain this relation we suppose
1) Cf. the remark at the end of II 8.
2) Cf. T. P. 1951, 1, p. 198.
 11. The LAGRANGE derivative. 113

V
that the variation d (/) is due to the dragging along of the field (/) over
v" dt, hence
V
{11.7) d (/) = - f,(/J dt
V

and that v", 8"' v", 8"' 81. v", . 0 0 are zero on r,._ 1 0 Let (/) be the tensor
P"_ .. ~PA, ... Ag Then we have

P';"o~pA, ... Ag = P":"'?;." ... J.g- v"' 8"' P":"~PA, ... Agdt +
{11.8) \ + P~" .. ~PA, ... Agof.' v"dt +
- p~ .. ~Pf.'A, ... Ag 0;. 1 VI' dt-
and thus

{11.9)
where * denotes terms each of which contains v" or a first or second
V V
derivative of v" as a factor. Hence dP and d 8"' P vanish on r,._ 1 and the
same can be proved for the variations of all higher derivatives of P.
By substituting (11.7) in (11.4) we get using STOKEs' theorem

;., ... Ag f,V P" .. "P, ... g dt d 5"tl .. od 5"t"

dfB d 5"tlo .. d 5"t"=-f[B].><, ... ><p o

Tn Tn

.J., ... Ag{v~-'8I' P" .. "P A, ... Ag -P"'"

=-f[B]""'"'"P A g 8I' v"'-
.. "P 1 ...

(11.10)
=- ff_[B] ,., ... ><p ;." ... ... ><p A, ... Ag +
"' P" .. "P A, ... Ag +A I'" 8e pe><,
Ag {8

- .pA, ... Ag- ...}+

0l" P"'""P
+ ><p A, ... Ag 8e [B]~'"
pe><, ... ;., ... ;..+ ..
.. "P
-P"'-""P ' L f), [B].Xt"P. ;., ... Aq- .lj v"' dt de ... d~"
/J"-B"'q' 11.1

but this variation must be zero because the field dragged along has
exactly the same components with respect to (x') as the original field
with respect to (x)o Hence

[B] ,., ... ><p A, ... Ag {8"' P"'-""P

A, ... Ag
+A" "' 8e pe"""P
A" ... J.g +
o

(11.11) \ -8 ;., P"'""P .pA, ... lg

- ..}+pe><,
... ><p . . A, ... J.+ ...
8 [B]I'"""P
l" ... g e
- P"''""P
8 [B] ><,. ... ><p. A, ... J.g- = 0
.pl, ... lg ;.,

For p= 0, q = 2 the identity takes the simple form

(11.12) [B];." (81, P;.,.- 8;. P"'"- 8,. P;."') - P"'" 8;. [BJ""- P;."' 8,. [B];." = 0
Schauten, Ricci-Ca!culus, 2. Auf!. 8
114 II. Analytic preliminar ies.

which is of special importance in the theory of relativity. (11.11) is

not the only relation between a quantity, its LAGRANGE derivative and
the derivatives of these two, as has already been pointed out by BucH-
DAHL1) and ScHRDINGER. 2) BucHDAHL derived identities of this kind
for a density as a function of the fundamental tensor and its derivatives
up to a certain order. ScHRDINGER considered a function of the para-
meters of a linear connexion and their first derivatives.
We give here only an example, the generalization being then ob-
vious. 3 ) Let P~;. be a tensor field and ~ a density of weight + 1 depend-
ing on the P~;. and their first and second derivatives
(l<)
(11.13) ~ = F (P~;.. 8~' P~;.. 8. 8~' P~;.).

)
Then we have

I
and in this equation the expressions

a P "'
,;. = A"';.
";.' P",;.

(11.15)
b) 8~', P~~ = (8~'. A:' i) P~;. + A:'f, ~ 8~' P~;.
c) 8," 8~', P~~ = (8. 8~', A=' i) P~;. + (8~', A:' i) A:. 8, P~;. +
+ (8. A~'i~) 8~' P~;. + A:'f~: 8. 8~' P~;.
can be substituted. After this substitution (11.14) is an identity in
the variables P~;.. 8~' P~;.. 8. 8~' P~;., A7', 8~'A7', 8,. 8~'AJ:' and these may
be considered as independent. If this identity is differentiated with
respect to 8. 8~' P~;. we get

(11.16)

and this proves that

(11.17) ~vp def iH3
"- o(o. oll P~;J
is a tensor density of weight + 1. Similarly by differentiating the identity
with respect to 81' P~;. and with respect to P~;. it can be proved once
more that [~]~;. is a tensor density of weight + 1. But new identities
arise if we first differentiate with respect to A7', 8~' Al' and 8. 81' A7' and
afterwards substitute A7' = br, 8~' A7' = 0, 8. 8~' A7' = 0. This treatment,
which is in reality equivalent to effecting an infinitesimal coordinate
transformation, is rather complicated and it is much simpler to replace
1 ) BUCHDAHL 1948, 1; 1951, 1; 2; 3.
2) SCHRDINGER 1948, 2; 1950, 1; 1951, 1.
3) NIJENHUIS 1952, 1.
 11. The LAGRANGE derivative. 115

the coordinate transformation by a point transformation. Then we

can use the LIE derivative. The LIE derivative of ~ can be computed

l
in two ways (cf. 10.27 and 10.17)

a) ~~ =
V
oI' ~v"
(11.18) b) ~~= ~~P"-< + 82 oI' ~P"-< + 0 (o 82 o oI' ~P"
V 0 p" V 0 (0 I' p" ) V
p
0I' P" ) p V

Equating these expressions we get

or [cf. (11 .17)]

(11.20) 1 -' + o o
[ ~] ~ P" + oI' {( <> ("8 2p" ) -
" V

V I' ~""
(I

>< ~
"I'

= 0
V P"
.<
2 oV ~"" ><) ~ P" - ~ v"} +
V

and after some calculation

(11.21)

This is an identity of the form

(11.22) 11
,..,.
v" + o q.:t " v" + o o t".:t
I' }. "
v" + o o o ?t".:t
I' "
v" = 0
8*
 116 II. Analytic preliminaries.

valid for all values of the field v" and its derivatives. Now such an
identity has as a consequence that

(lJ" + 8;. q~" + 81' 8;. r(~~l,, + a. 81' 8;. !J(~~~l,,) v" +
(11.23) { + (q~" + 2 81' r(~~l" + 3 a. 81' !J(~~~l") 8;. v" +
+ (r".< + 3 8 i;J(vl'.<)) 8 8 v" + i;J(~tA) 8 8 8
. x V x f.l "Vf-lA
v" = 0 '
hence
(11.24)

I
Applying this to (11.22) we get four identities
a) [~]~.< 81' P~;.- 8;. [~]~.< P~l' + 8;. P\ [~]~" = 0
b) - ~ A A + [~]. A P" - p.<
V X o V X
[~] " + (~- -
V ( >< )
2 8(] ~QAI'
X
) 8 P"fl, -
'P
8 8;.1". I'

- 8 P" (
I'
8B
.v 8(8;.P~I')
- 2 8 ~e "~' )
e ""
+ 8" P~'" (~-
8(8;.P~")
- 2 8 ~e.<x
e ... v
) = 0
(11.25)
c) (--8 ~ ... - 2 8 ~e(".<)) P" - A("
v
pJ.) (---82
'" 8(8vP,0 ")
-- 2 8
e
~evx)
... a,
+
8(81'P~!.) e '"" .a

+ ~(~~)~" 8a P~,,- 8. ~(~~)~" P~ a + 8. P~" ~t~~l~ a = 0

Of these (11.25 a) is a special case of (11.11). (11.25 d) is an algebraic

identity and the others are differential identities.
Similarly, if a density ~ of weight 1 is formed from a tensor P +
(with arbitrary valences) and its derivatives up to the order p, an
equation of the form (11.23) arises with p + 2 terms and thus we get
p+ 2 identities whose total nurober of equations is

(11.26) n + n 2 1 , n (n +2 1) + + n (nP+l _ (n + P+ l) .
+ P) -n
P+l
The first n equations can be found in the same way in which (11.11) is
found and the last n(nP+l
+ P) are algebraic identities linear homogeneaus
in P and in the derivatives of ~ with respect to 81'v ... 8{1, P.
If instead of P.\ a tensor g.<x of rank n is taken, the sccond and
third term of (11.25b) are
(11.27) - [~J" g";.- C~Y" g;." =- 2 [~J"g";..

This is also true when higher derivatives of g;." occur. Hence by

transvection with the inverse of g;." the LAGRANGE derivative can be
expressed in terms of ~. g." and their derivatives. 1)
1) BucHDAHL 1951, 3.
 12. CARTAN's symbolical method. 117

If P is a geometric object but not a quantity [for instance the

II~;.in (10.31)] the LIE derivative ofll contains not only first but also
second derivatives of v". If for instance II and its first derivatives
occur in Q, one finds n ( n t 3 ) equations of which the last n (n t 2)
are algebraic. 1 )
Exercise.
II 11, 1. Prove that the left hand side of (11.11) is a covariant
vector L1-density of weight + 1 if [Q] is replaced by an arbitrary tensor
L1-density 0 of valences q and p and weight +1. For q=1, P=O and
q = 0, p = 1 this gives the differential concomitants
II 11,11X) 20;. 8r"P_.1 - P" 8;.0"'
II 11,1 ) 0;. 8" P"'+ 8_.(0" P;.) .

In an R3 with signature ++ + these are the concomitants written in

ordinary vector notation:
II 11,1 y) Q X rot P - P div Q
II 11,1 b) grad (P . Q) - P X rot Q + Q div P.

12. CARTAN's symbolical method.

CARTAN's symbolical method is intimately connected with anholono-
mic coordinate systems. It is in fact an abridged notation specially
suited to dealing with these systems. If the A1 of an anholonomic
system (h) are fixed, these components may be considered as n sets
of components of the n basis vectors l_., ... , e~ of (h) or (if we like)
of the corresponding pfaffians t_. de.
Then any geometry introduced
into xn can be expressed in terms of these vectors or these pfaffians.
This is the point of view ofCARTAN. In his thesis of 1894 2) he developed
a very ingenious method to deal with pfaffians and their generalizations
and this method was used in connection with some kind of tensor
calculus in nearly all his papers on differential geometry. The introduc-
tion of a tensor calculus will always be necessary if non alternating
quantities are considered, because the symbolical method is made for
alternating quantities only. In 1934 KHLER 3 ) gave a systematic account
of the possibilities of the method as far as alternating quantities only
were concerned and under his influence some notations were slightly
changed.
1) ScHRDINGER 1951, 1.
2) CARTAN 1894, 1; 1901, 1.
3) KHLER 1934, 1.
 118 II. Analytic preliminaries.

In this section we try to give an exposition of the abridged notation

in such a way that it becomes fully adapted to the notations of tensor
calculus. 1) Using the kemel-index method this can be attained very
easily by using the same kernels in both sets of formulae. Then a simple
rule for translation can be established.
If f~", ... , l~" are linearly independent linear elements and w.<, . . .<p
a covariant p-vector, all at the same point, the expression
(12.1)

where ct.p is a constant that will be fixed later on, is a contravariant

p-vector and
W def W [d c;;t d c;;t.lp] -_ ct.p
1 1 d t.l d t.lp
- .<1 -<v w.<, ... -<v 1 c;; P c;;
(12.2) {
= C/..pW.<,' .l'd~ A'1 d~;.'P = C/..p W, 1 i ( d~ )"' 1 ( d~ )''P
P1 p P1 p

1s a generalization of a pfaffian, called an alternating differential form.

For P=l we take ct..1 =1 and

(12-3)

The product of two forms J and 4 of valence p and q may be defined

as follows

(12.4) [44]defct..p+qJ["t"P, 4", ... "011de' ... Pd~"vP+1

d ~"' ... P+q
d ~"; P+q:;;;,n.

Then we have (cf. I 7.1)

(12.5)

In all these expressions it is quite unimportant whether we consider w

as an alternating form or as an abridged notation of a multivector.
The results are the same and the interpretation is only a question of
personal predilection. 2)
1 ) This is very important. All authors who are using, where convenient, CAR-

TAN's abridged calculus, always use also notations of tensor calculus and they
seem to have no difficulties with the latter. But as a matter of fact most authors
using tensor calculus have the greatest difficulties with CARTAN's abridged cal-
culus, even in translating results. This is a great pity. Both methods have their
advantages (CARTAN hirnself performed miracles with his method !) and both should
be used by every author, each on the place where it suits best. But this ideal
can only be attained if there is a convenient bridge between them. We try to
build this bridge here.
2 ) Just the same thing occurs in linear algebra. We may Iook upon w as an

abridged notation for a covariant vector w or as a linear form w x-'. The old
theory of invariants preferred the latter point of view but nowadays many people
think more easily in terms of vectors.
 12. CARTAN's symbolical method. 119

If s is a scalar, an in variant differentiation ofthe p- vector s [d ~"' . .. d ~"P]

can be defined as follows
[d {S [ d~"' . .. d~"PJ}] ~~ Q(p+l (o,t s) d~[t. d~"' . .. d~><p]
{12.6) { 1 p
= (o,ts) [d~Ad~"' ... d~"P].
From this definition and (12.2) it follows that

(12.7) [dwJ = (o,twl, ... .lp) [ded~J.,_ .. dePJ =Q(p+Ia[,lw.l, ... ~.p 1 defe' .. f~Ap.
The differentiation of a product we easily find then to be
1
p = valence of w.
CARTAN always took Q(P = p! and he used the product notation

[44]. But instead of [dw], at first he wrote w' and more recently
under the influence of KHLER dw. Now this latter notation may lead
to ambiguity because [dw] is certainly not an ordinary differential.
KHLER wrote dW'', ... , ~"P) for what we have written Q(Pd~["' . .. d~"Pl
1 p
and [d~"' ... d~"P], and he did not fix the values of Q(P. In fact, if only
symbolic expressions are used, and if we never wish to translate into
notations of tensor calculus, the constants Q(P never occur. For [4 4]
KHLER wrote 4 4 without brackets. But in dealing with less simple
problems the notation with brackets should be preferred. As to the
choice of the Q(P we remark that, according to (12.4), [44] does not
correspond to the ordinary alternating product of the multivectors
0(;
corresponding to 4 and 4 but to Q(p+q 1 Q(-q 1 times this product. So,
if we wish to establish a reasonable adaptation we have to take Q(P = 1
for all values of p. Doing this and always using the same kernels in
both notations we get the following simple scheme of translation. Let
II;., ... .lpA be components of a geometric object, alternating in the first
p indices and carrying a collective index A standing for any nurober of
co- and contravariant indices, each belanging to any holonomic or
anholonomic coordinate system in Xn or any other space. Let the objects
II;,, ... .lp(A) with p living indices A.1 ... Ap, resulting from strangling all
indices symbolized by A, be covariant p-vectors. As an example we
take the intermediate components A1 and fJ!.l (cf. II 9). In fact
Ar;;= i.l are n vectors and Q~hj are n bivectors. Now IIA denotes the
p-vectors, resulting in this way:

(12.9) 1 and alw

II;., ... Ap (A)
120 II. Analytic prelirninaries.

1 2
The product [11A 11B] stands for

1 z
1 z \11[1,, ... J.p j(A)[ 11,., ... ><g] (B)
(12.10) [11A11B]: and also for
1 2

'--------
11[;., ... J.p (A)[ 11,., ... ><g] (B) 1e ~~\~/;"' p~q~g
and the expression [d11A] symbolizes

(12.11)

With this scheme every abridged formula can immediately be trans-

lated into a formula of tensor calculus. There is one point that is at
first bewildering for one accustomed to tensor calculus only. For in-
stance w does not stand for W;. only, but also for W;. or W; and even for
w;.d~J., W;: d~;: and wddW, and [dw] denotes not only 8r" w.~ 1 but also
+
8r" w.~. 1 and 8u w;1 wh ilf; (cf. Exerc. II 9, 1) and even the alternating
differential forms obtained by transvecting these expressions by d~" de,
1 2
d~" d~J.' and (d~)i (d~)i respectively. This means that in passing ta
1 2 1 2
the abridged notatian the caardinate system belanging ta the alternating
part af the abfect is set free and that at the same time the indicesnot belanging
ta this part are all strangled.
From now on we mark all formulae in abridged notation where they
appear together with the non abridged formula by the sign C (a sign
taken from music) and give both formulae the same number. We give
here two examples, concerning the object of anholonomity [cf. (9.10)
and Exerc. II 9,2] :
" Ah _ Ai i
(12.12) ur" J.] - pJ.
nh _ nh
.:..:ii- .:..: pJ.'. ((x) holonomic)

(12.12) c
(12.13) _
kii nh nt nh }
0- Av";. {8rk.:..:iiJ- 2.:..:[ki.:..:;Jl

(12.13) c
So far we have seen that certain (not all) formulae of tensor calculus
can be written in the abridged form and that by doing this a lot of
indices can be saved. On the other hand the abridged calculus requires
 III. Linear connexions. - 1. Paralleldisplacement in an En. 121

more attention during the calculation and so it is merely a question

of personal inclination whether one or the other calculus will be pre-
ferred. But in order to give abridged calculus fair play we draw the
attention to some cases where it has certainly advantages.
Let us take a vector field w_. and its rotation W.u-<:

(12.14)

(12.14) c 2[dw] = W

and let us suppose that we wish to consider w.< not only as a function
of the ~" but also of certain parameters 'YJ""; rx = i., ... , r. Then we intro-
duce in the xn+r of ~'', 'YJ"" a field wB; B=l, ... 'n, i, ... 'r with the
n+r components w.<, Wp=O. If this Xn+r is reduced with respect to
the normal system of x;s ~" = const. the field wB is reduced to the
original field w.< in the Xn, depending on the parameters 'YJr~.. In tensor-
calculus we have now

(12.15)

or
(12.16) B, C = 1, ... , n, i, ... , r
but in the abridged calculus this equation is symbolized by the same
equation (12.14C). Hence the abridged calculus gives the possibility of
introducing at any moment auxiliary variables without changing the form
of the formulae.
Same other examples of the use of the abridged calculus will be
found in Ch. III.

111. Linear connexions.

1. Parallel displacement in an Eu- 1)

Let v" be a vector field in an Xn. As we have already seen in II 3,

dv" is not a vector and 8.u v" not a tensor. But if the Xn is an En and if
(h) is a rectilinear coordinate system, dvh and 8ivh are a vector and
a tensor because if (h') isanother rectilinear system, the Ar are constants.
1 ) Tensor calculus was founded, developed and applicated by Rrccr in a great

number of papers. We mention here 1884, 1 and 1886, 1; 2 (all preliminary);

1887, 1; 2; 1888, 1; 2; 1889, 1; 2; 1892. 1; 1893. 1; 2; 1894. 1; 2; 1895. 1; 2; 1897.
122 III. Linear connexions.

dvh rs the difference between the vector vh + dvh at ;h + d;h and a

vector with the components vh at the same point. This latter vector
can be derived from the vector vh at ;h by the well known process of
parallel displacement.
Now let (x) be a general coordinate system in En. Then the Ai:
are not necessarily constants and we have

where
(1.2)

Hence the components of the vector dvh with respect to (x) are

(1.3)

We call /Jv" the covariant differential of v" and the r:A

the parameters
of the parallel displacement in En with respect to (x). If we write
/Jv" = V~" v" d;~", we get
(1.4)

v" is called the covariant derivative of v". If (x) is rectilinear, the

V~"
r;;.are all zero and we have /Jv" -"'- dv" and V~" v" -"'- 8~" v".
Instead of vh we can take a covariant vector field W;. Then for the
components of W; and 8i w,; with respect to (x) we get

(1.5) IJW;. def dw;.- r:;. W" d;~",

(1.6) J7I" W;. def o.
= Lf< W-
T"pJ. W".

But these formulae can also be derived by using the fact that the trans-
vection ve we is invariant if both vectors undergo the same parallel
displacement. Hence we have d (ve we) = (J (ve we) and the equation

(1.7)

can only be true for every choice of the field v" if (1. 5) holds.
1; 2; 3; 1898, 1; 2; 1902, 1; 2; 1903, 1; 1904, 1; 1905, 1; 1910, 1; 2; 1912, 1;2;
1918, 1; 2. It found its first systematic representation in Rrccr and LEVI CrvrTA
1901, 1. His "derivazione covariante" got a geometric interpretation in the
"pseudo-parallel" or "geodesie" displacement (now called parallel displacement)
introduced by LEVI CrvrTA 1917, 1 and independently by ScHOUTEN 1918, 1.
Cf. also BRaUWER 1906, 2, p. 121 who gave a first example of a displacement of
a vector in an Sn (cf. III 5).
 2. Parallel displaceroents in X,.. 123

2. Parallel displacements in Xn 1), 2).

We have seen in III 1 that the construction of the covariant diffe-
rential in E,. only depends on the existence of parallel displacements
in E,.. We do not need all displacements but only those from one
point to a "neighbouring" point. Hence if in an X,. a definition of a
displacement of quantities were given for every point of the region
considered and for every linear element in such a point, it must be
possible to construct some kind of a covariant differential.
Let (/> (indices suppressed) be a field of quantities defined in some
region of X,.. Then at ~" and ~" + d~" we have the field values (/) and
(/) + d (/>. Let a parallel displacement for these quantities be defined
for every choice of ~" and d ~" in the region considered and let (/> + d*(/>
be the field value at ~" + d~" after a parallel displacement from ~" to
~" + d ~".
Then d (/> - d*(/> is a covarian t differential {J r;/J of r;/J belanging
to this displacement. Of course we make the condition that if the com-
ponents of (/> + d*(/> are given for one coordinate system (u) its com-
1) Paralleldisplacements {formerly often called pseudo-parallel displacements)
are considered here only with respect to the tangent E,.'s of an X,.. KNIG ge-
neralized 1920, 1 this idea for a set of EN's each of which is associated with a
point of X,.. Cf. KNIG 1932, 1; SIBATA 1934, 1; HLAVATY 1934, 1; 1935, 1;
KNIG and PEscHL 1934, 1; KNIG and WEISE 1935, 1. As we proved 1924, 2; 3
the conformal and projective connexions introduced by CARTAN 1923, 2 and
1924, 1 can be looked upon as special cases of KNIG connexions. There is much
Iiterature on these general connexions. We mention here only a few papers where
also other literature can be found: ScHaUTEN 1924, 1, p. 62; 1926, 1 ; ScHLE-
SINGER 1928, 1; MaRINAGA 1934, 2; 3; TucKER 1935, 1; YANO and MuTo 1935, 1;
1936, 1; WAGNER 1943, 1; YANO 1945, 2; 3.
2 ) There is a large literature on connexions and parallelism. We mention here

only HESSENBERG 1916, 1 (preliminary); LEVICIVITA 1917, 1; ScHOUTEN 1918, 1;

1922, 3; 1923, 1; 1926, 1 (also non linear) 1928, 1; FRIESEKE 1925, 1 (non linear);
EISENHART 1927, 1; STRUIK 1928, 1 {literature); SCHLESINGER 1928, 1; 1929, 1;
WuNDHEILER 1929, 1 (non linear); DuscHEK 1930, 2 (axioroatic); BoRTOLOTTI
1930, 3 (literature); 1930, 2 and 1931, 3 (degenerate fundamental tensor); 1931, 4
(also non linear and parallel directions) 1935, 1; 1936, 1 (non linear); 1937, 3;
1939, 1 (also for literature) 1942, 1; CoNFORTO 1931, 1; 2 (in function space);
A. KAWAGUCHI 1932, 1 {also non linear) 1933, 1; 1933, 2 (also non linear); 1934, 1;
1935, 1 (in function space and literature); 1952, 1 (non linear, general theory and
literature); VANDERSLICE 1934, 1; HOKARI 1934, 1 {for bivectors); BORTOLOTTI
and HLAVATY 1936, 2 (also literature); URABI 1940, 1; MIKAMI 1941, 1; WAGNER
1943, 1 and 1945, 2 (geheral geom. obj.); NoRDEN 1945, 1 and 1950, 1 (space with
two connexions); DALLA VoLTA 1946, 1 and 2 (asymmetric); BoMPIANI 1945, 1;
1946, 1 (fortensors); 1952, 1 (conn. non posizionali); RozENFEL'D 1947, 1 and
1948, 1 (generalizations); JoHNSON 1948, 1; LIBER 1950, 1 (classification of con-
nexions in L 2 ); EHRESMANN 1950, 1 (in fibre space); 1952, 2 (higher order); KNEBEL-
MANN 1951, 1 (parallelism relative to a continuous curve); WINOGRADSKI 1951, 1
(axiomatic); TAKAsu 1953, 1.
124 III. Linear connexions.

ponents with respect to (u') can be derived by the rules valid for the
transformation of (/J + d (/J. That means that we introduce as a first
condition
C I) The covariant differential of a quantity (/J transforms in the same
way as (/>.
Now there are many possibilities for the definition of a displacement.
Here we consider only those displacements that satisfy (CI) and the
following four conditions (all quantities considered are supposed to be
given as fields in a certain region):
C II) The covariant differential of a sum of quantities is the sum of
the covariant differentials of the terms.
C III) F or the covariant differential of a product or a transvection
of quantities the rule of LEIBNIZ holds, for insjance (indices suppressed)

C IV) The covariant differential of a quantity is linear homogeneaus

in the d~":
lJP = V~'lJI d~"'.
C V) The covariant differential of a scalar or W-scalar is identical
with the ordinary differential.
* hence accord-
Fora contravariant vector v" we have lJv"=dv"-dv",
* is linear homogeneous in
ing to (C IV) dv" d~"

(2.1) * =q;;(vJ.) d~"'

dv"
But then it follows from (C II) that q;; (vJ.) is linear homogeneous in v".
Hence lJv" and V"' v" have the form
(2.2) lJv" = dv" + T"p.J. vJ.dt:"'
5" '

(2.3) vp. v" = f)p. v" + r:J. vl..

According to (C I) we get for the transformation (u) ~ (u')
(2.4)
from which we easily deduce the transformation formulae

(2 5) T"'
p'J.' = J. "' J.pJ.
A~'p.'J.'x Tl'< + A"'J. up.'
" AJ.' = A~'p.'J.'x " A"'
J. ": T"p.J.- A~'p.'J.'J. up. J.
showing that the r:J. do not constitute a quantity but a more general
geometric object of the kind already considered in II 10.
The n 3 parameters r;;.
can be given arbitrarily for any system (u).
Then by (2.5) the parameters with respect to every allowable coordinate
 2. Parallel displacernents in X". 125

system are known. They are said to fix a linear displacement or a linear
connexion in Xn. 1) An Xn provided with a linear connexion is called
an Ln. If a vector in an Ln is displaced in such a way that v" = 0
the displacement is said to be parallel.
Using (C V) it can be proved in the same way as in III 1 that
for a covariant field w_.
(2.6) w_.=dw_.- r;;;.w"d~",
(2.7) V" w_. = a" w_.- r;;. w~.
Every tensor can be written as a sum of general products of vectors
(cf. I 3). Hence as a consequence of (C III) we have for instance
for P"<
!'

(2.8) P""'
f'
= dP""'
f'
+ T."
VQ
pQJ.
o f"
di:V
<;
+ r.-'
VQ
P"Q
of'
di:V-
<; r.Q
Vf'
P""'Q di:V
o <; >

(2.9) V" P~~'" = a. P~~'" + r"~ P~~-" + r.~ P~;" - r"~ P~~e.
For every upper (lower) index there is a term with a + -(- -)sign.
Every W-tensor is the product of a tensor with a W-scalar. Hence,
according to (C V) the formulae for the covariant differentiation of
tensors hold also for W-tensors.
The covariant differential is the difference between the field value
at ~" + d~" and the field value of the field that has been displaced
parallel from ~" to ~" + d ~". This can be said in another way; the co-
variant differential is the ordinary differential from the point of view of
a local coordinate system that is displaced parallel from ~" to ~" d ~". +
A quantity is called allround covariant stationary at a point of an L,.
if its covariant differential vanishes at that point for every direction.
I t is called covariant stationary at a point in an m-direction if the co-
variant differential vanishes at that point for all directions lying in
this m-direction. It is said to be covariant constant over Ln or covariant
constant for short if it is allround covariant stationary at all points
of Ln. 2 )
If a quantity is defined over an Xm in Ln it is called covariant con-
stant over xm if at all points of xm it is covariant stationary in the
tangent m-direction of xm. If the quantity is also defined in some
1 ) GALVANI has shown that connexions often can be realized by taking for

points certain subspaces of a flat space, 1942, 1; 2; 1943, 1; 2; 1945, 1; 1946, 1.

Cf. also for other choices of the elernents of space RozENFEL'D 1947, 1; 1948, 1;
YANO and HrRAMATU (sets of hypersurfaces) 1951, 1.
2 ) The existence of covariant constant fields in the !arge is quite another problern

that does not belang to the realrn of this book. We rnention here only a few papers
where also other literatme can be found: BoRTOLOTTI 1928. 1; 1929, 1; T. Y. THOMAS
1936, 1; MAYER and T. Y. THOMAS, 1937, 1.
126 III. Linear connexions.

neighbourhood of xm it may happen that it is allround covariant

stationary in Ln at all points of Xm. In this case it is called allround
covariant stationary in Ln over Xm. Of course it is then also covariant
constant over Xm. The following scheme gives the relations between
these five cases
allround cov. stat. cov. stat. in m-direction
at point of Xm at point of Xm
cov. constant
over Ln n n
allround cov. stat. cov. constant over Xm 1)
in Ln over Xm

A field that is at every point and for every direction proportional

to its covariant differential is called recurrent. A recurrent vector field
is also called parallel.
If we wish to write out a covariant derivative in terms of its inter-
mediate components (cf. I 2) for instance between (u) and (u') it is
convenient to introduce the notations
(2.10) F" = APp' F"
p' .< def p'. r" pA' def p F"'
= AP' p' A'

Then for P~~l' for instance we have the formulae

(2.11) V.v P"'-'
I'
= ov P"'-'
I'
+ F"', pe'-'I' + T.-'ve P"'l1
11 I'
- r.e
vp
p~.<
Q'

(2.12) V.. P"-''

I'
= o, P""'
I'
+ T.'fe P . + r.~. P" I'
11 -''
~-< 11
11' - T.V
~-<
P"'-'
11

equivalent to (2.9). They each contain only intermediate components

of the same kind. 2) The i! andr; r;;.
might be considered as a kind of
intermediate components of the geometric object !;.".<. But in using
them we have always to remernher that they can only be formed by a
transvection over the index fl and that r;" =!= r;;'"' A:.i'. That is the
reason why, in general, we use intermediate components of quantities
only.
From (2.5) it follows that
(2.13)
is a tensor. It is called the tensor of asymmetry or torsion 3) of the Ln.
The connexion is said to be symmetric and the Ln is called an An if
S~}." = 0 at all points of the region considered. It is said to be semi-
symmetric if 5~}." has the form
(2.14) S t-<
"-- S [p A"J 4)

1) Cf. E I 1935. 1, p. 77 f. The term "allround" introduced here is new.
2) Formulae of this kind were introduced first by R. LAGRANGE 1926, 1, p. 10.
3 ) Mc CoNNELL used 1928, 2; 3 the word torsion in Vn in another sense connected
with a geodesic.
4 ) Cf. FRIEDMANN and SCHOUTEN 1924, 1.
 2. Parallel displacements in Xn. 127

From (2.14) we easily deduce

{2.15) 5 =-2-5 .. ;.
" n -1 pi.

and this equation will also be taken as the definition of the conco-
mitant 5,. if (2.14) does not hold. 5~;." is intimately connected with the
operationRot (cf. II 6). The covariant derivative of a q-vector wi.,. ... Ag is

(2.16) {
J7,. W;., ... .lg = o,. W;., ... Ag- ~\ W,.;., ... i.g - ' ' ' - ~".~g WA,_ ... g-"
= o,. w.<, ... .~g - q ~'[.~g wi.,. ... Ag-,J"'
hence
(2.17)
and for the semi-symmetric case

2.18)

from which it follows that only in an An the symbols o and 17 can be

interchanged in the expression 17[,. w.<, ... Ag].
If two linear elements d~" and d~" at ~" are displaced parallel along
+ 1d~" we get the vector 2d~"- r;;. 2d~;. 1d~" and at ~" + 2d~"
1 2
each other, at ~"
the vector d~" -F;;. d~;. d~". neglecting all terms of third and higher
1 1 2
orders. The figure obtained is in general not a parallelogram but a
pentagon with the closing vector

{2.19)

Hence in an Ln with an asymmetric connection, in general no infinite-

simal parallelograms exist. In the semi-symmetrical case the closing
vector is
{2.20)

and accordingly infinitesimal parallelograms only exist in the 2-direc-

tions contained in the (n -1)-direction of 5;.. In the symmetric case
there exist infinitesimal parallelograms in every 2-direction. 1)
There is another geometric interpretation of 5~;." due to CARTAN.
Thus far we have only displaced the vectors and not the points of the
tangent En. This comes to the same thing as if we had assumed that
1 ) Cf. CARTAN 1922, 2; FRIEDMANN and ScHOUTEN 1924, 1; J. M. THOMAS
1926, 1. Cf. for other geometric interpretations of S~i_", PINL 1951, 1; BOMPIANI
1951, 1.
 128 111. Linear connexions.

the contact-point always remained a contact-point. But another dis-

placement of points can be fixed by the assumption that the contact-
point of the tangent E" of f' has after its arrival in the tangent E" of
f' + d~" the radiusvector - df'. This we call a CARTAN displacement.
If the X" is an E" all tangent En's coincide and the CARTAN displace-
ment is the displacement zero. The covariant differential eh" of a
vector v" is the differential with respect to a parallel displaced local
coordinate system. But the covariant differential of a point is the
differential with respect to a local coordinate system that is under-
going a CARTAN displacement. Hence, if v" is the radiusvector of a
field of points, each of which lies in a tangent E", its covariant differential
is d~" + (Jv". Vice versa, if a vector is displaced parallel over d~", its
ordinary differential is - d~,.. F,..). vA, but if the point with radiusvector v"
undergoes a CARTAN displacement, its ordinary differential is - d~"
d~,.. F,..";. v" . 1 ) Now let us suppose that there is a closed curve in X"
through ~" and that the local E" at ~ undergoes a CARTAN displacement
0 0
along this curve. The points of the curve may be considered as a point
field having in each local coordinate system along the curve the radius-
vector zero. During the motion the moving E" coincides at every
moment with one of the tangent E"'s of the points of the curve. Hence
at every moment the place of this point can be marked in the moving E".
Then in this E" we get a curve and it can be proved that if the original
curve is contracted into a point and if terms of third and higher
orders in the differentials are neglected, the image in the moving E"
is always closed in the symmetric case and is closed in the semi-symmetric
case if and only if the 2-direction of the enclosed surface element is
lying in the (n -1)-direction of 5;.. In an E" the curve and its image
coincide. A rigorous proof of theorems of this kind always requires
a rather detailed discussion. Therefore we only give an illustration here,
that, without being a rigorous proof, may have some heuristic value.
Let the curve be the quadrilateral of the points ~. ~" + d~", ~" + d~" + d~"
1 1 2
and ~ + 2d~". Starting from the first contact-point ~" the radiusvector
of this point with respect to the second contact-point ~" + d~" is after
1
the arrival of the E" at this latter point equal to - d~". Going from
1
~" + 1d~" to the third contact-point ~" + d~,. + 2d~,. the point ~" + d~"
1 1

1)From the !arge amount of.literature on the CARTAN displacement in various

spaces we mention here only a few titles: DIENES 1934, 1 ; WAGNER 1942, 1 ;
TACHIBANA 1949, 1; SASAKI and YANO 1949, 1; 0TSUKI 1950, 1; KANITANI 1950, 1.
Of course the CARTAN displacement can also be defir.ed for the moregenerallinear
connexions mentioned in footnote 1, 2 on p. 123 and especially for conformal and pro-
jective connexions (cf. CARTAN 1937, 1, 'P 163ff., 278ff.).
 2. Parallel displacements in X". 129

gets the radiusvector - d~" with respect to this third contact-point

2
but at the same time the vector - d~" is displaced parallel over d~"-
1 2
Hence
1
(j d~" = d 1d~" + d~~'- r;;. de = 0 and this has as a consequence
2 1
that the point that had a radiusvector zero with respect to the first
contact-point and after the first CARTAN displacement a radiusvector
- d~" with respect to the second contact-point gets after the second
1
displacement a radiusvector- d~"- d~" + d~l'- r;;. d~ with respect to the
1 2 2 1
third contact-point. Going via ~" + d~"2
we get in the same way a
radiusvector - d~"- d~"
2 1
+ d~~'- r;~ de.
1 2
Hence, if we go from ~" round
the quadrilateral back to ~", the original contact-point has after its
return in ~" the radiusvector - 2 d~~'- des;;.".
1 2

Applying (2.16) for q=n we get

17p, w1... n = 8p, w1. n - n F"- ,u ln w1 ... n- 11 x

(2.21) {
= 81'- w1 ... n - Fl'-"" w1 ... n
Hence, if w;., ... ;.,. is displaced parallel we have.

(2.22) awl. .. n = J w1 ... n = r 1'- wl. .. n d~'' '

I;, is a geometric object with the transformation [cf. Exerc. II 2,1 and
(2. 5) J
(2.23)

Now w1 ... " may be considered as the component of a scalar .1-density

of weight +
1. Accordingly we have for such a density fJ

(2.24) (jfJ = ap- rl'- fJ d~'',

(2.25) 171'- fJ = 81'-fJ - r,, fJ
and for a .1-density q of weight w
(2.26) (j q= d q - W I;, qd ~I' ,
(2.27) 171'-q = 81'-q- wrl'- q.
If a .1-density is multiplied by a W-scalar we get an ordinary density.
Hence, according to (C V) the formulae for the covariant differentiation
of ordinary scalar densities and scalar .1-densities are the same. 1 )
1) Cf. for covariant differentiation of densities VEBLEN and T. Y. THOMAS
1924, 1; HLAVATY 1927, 2; ScHOUTEN and HLAVATY 1929, 2.
Schauten, Ricci-Calculus, 2. Auf!. 9
 130 III. Linear connexions.

Because every tensor density can be written as the product of a

tensor and a scalar density, in the formulae for the covariant differential
and the covariant derivative we getan extra term with r", for instance
[cf. (2.8, 9)]

{ 15\~~ p = d\~~ p + r.~ \~~ p d~ + r.~ \~~ p d~-

(2.28) - r.e ffi"-'
vp"t-1. e
d /:V -
~
w ffi".!
-1-"'. .u
r." d t ,
~

(2.29)

For a p-vector density of weight + 1 we have

( ~p ltt'""P =
.+ r;l
ap ro"""" + T"p.< w-<""" + T"p.< w"-<""P + ...

t
itJI'""P-l - ~~.! itJI'""P
(2.30)
= a" ro"""P + 2 s;~ ltt'""P +
+ (p-1) S~i["itJIPA'""l
from which we see that in the expression 17" it>~'""P in an An the
symbols oand 17 can be interchanged but that in a general Ln a correction
term containing SA occurs and for p > 1 also one containing S;t.
In order to establish the covariant differentiation of pseudo-quantities
as defined in II 3 with respect to an auxiliary variable ~0 , we consider
a pseudo-scalar ~ of class 1. Its ordinary differential is not a pseudo-
scalar because
+ ~ opad~". 1 )
(0') (0) {0) (0)
(2.31) d ~=da~ =O"d~
From the conditions (CI-IV) we see that the covariant differential
must have the form
(2.32)
with parameters A" subject to the transformation
(2.33) A", = A;,A" + a", loga.
This implies that in order to get the covariant differential of a pseudo-
quantity ([J (indices suppressed) of class c we have first to write down
all terms that would occur if (/) were a quantity and then to add the
term - c (/) A" d~" For instance the covariant derivative of a pseudo-
tensor density ~~.< of weight w and class c is
(2.34) 17" ~~-< = o" ~~-< + ~"e ~~-<- r:, ~~a- w ~~-< ~- c ~~;,Aw
If a vector field v" is given we may require the field '(/) A that arises
from a given field (/) A if this field is displaced parallel along the stream-
1in es of v" over v" t (cf. II 10). In order to find '(/) A we take the co-

1) If a also depends on ~ 0 there isanother term rf 8 cr d~ 0

0
 ~ 3. A linear conncxion expressed in terms of 5 1, .1". 131

ordinate system (x) such that v" = e" (cf. Exerc. II 5,1) and introduce
1
an anholonomic system (h) such that the e~; h, i = 1, ... , n are co-
variant constant along the streamlines of v".' Let the components of
(/> A with respect to (h) be symbolized by (/>L. Then we have

(2.35)
Since the e" are covariant constant along the streamlines, '(/>L (~") equals
(/>L ( ~" - t 'e"). Hence
1

(2-36) '(/>LW) =(/>LW - t e") = e-"h (/>LW) = exp (- 1 v 1ll't) (/>LW)

1

and, because this result has the invariant form

(2-37)

Exercises.
III 2,1. Prove that A~ is covariant constant.
III 2,2. Prove (2.5) starting from 17"' A1' = 0.
III 2,3. The vector S~~" d~"'
1 2
ae
lies in the 2-direction of d~" and
1
d ~" for every choice of these latter vectors if and only if the connexion
2
is semi-symmetric (cf. Exerc. I 3,6).
III 2,4. Prove that for every given point there exists always a CO-
ordinate system (x) suchthat E" .. "n is covariant stationary at this point.
(")
III 2,5. Prove that ~><, ... ><n and eA, ... An are covariant constant (cf.
Exerc. II 10,1).
III 2,6. In the identity (II 11.11) for [~] in An it is allowed to re-
place 8"' by Vw Prove this for the special case (II 11.12).

3. A linear connexion expressed in terms of s;;.",

an auxiliary symmetric tensor field Y-<x of rank n and V"'g"""
Let g"" be a symmetric tensor, g"" = g"" and let its inverse be written
g"-'. At first we do not consider g"" as a fundamental tensor in tangent
space, it is just an auxiliary symmetric tensor and nothing more. But
we will use it to raise and lower indices in the way described in I 9.
If we write
(3.1)
1 ) This formula was first found by MoRINAGA 1934, 2, the proof given here is
by NIJENHUIS (personal communication). Cf. also the footnote 1 on p. 340 on pro-
duct integrals VII 1 and DIENES 1934, 1; JoHNSON 1948, 1; NIJENHUIS 1952, 1,
p. 174. tP,1 and 'tPA are here used as function symbols, cf. II 3.
9*
132 III. Linear connexions.

it follows from g"e gQA = Ai that

(3.2)
By writing out these equations
(3-3)
permuting the indices p, u and taking the sum of the permutations
p, u, u p, diminished by the permutation u p, we get by addition

(3.4) { 2I(~p)gQ" = o"g;.,. ~Qa,.g";.- a".::" + Q""" + Q,.";.- Q;.,.,,-

-25";. gQ,.-25,.; gp(1
or

(3.5) ~J.= U"} + s#"- s~"+ s~" + t (Q#" + Qi~"- Q~";.)

where
(3.6)
is the so called CHRISTOFFEL symbol belanging to the tensor g;_,. . 1) In
fact (3.5) satisfies the equation (3.3).
The tg"a o" gAa S#" and lQ#l" occur in (3.5) in an analogaus way.
This can be made clearer by using the abbreviation (the place of the
indices is important !)
(3.7)
for any expression '1flpA>< Then, if we write
(3.8)
(3.5) takes the form

(3.9) I r;:;. = g"Q (x{f.IQ)- s{PQ} + ~ Q{pQ)) I

If Q; l'' = 0 the connexion is called metric 2) wi th respect to g7." . In
this case the covariant differential of g"" vanishes and this implies
that the processes of covariant differentiation and of raising and lower-
ing of indices are commutative, for instance
(3 .10)
1) CHRISTOFFEL 1869, 1.
2)The equation (3.9) can also be used to establish relations between two dif-
ferent non-metric connexions. Cf. SEN 1944, 1; 1945. 1; 1946, 1; 1948, 1; 1949. 1;
2; 1950, 1; 2; NORDEN 1945, 1.
3 ) Most authors accept the convention that the range of differentiation of V
stretches as far as the first closing bracket whose corresponding opening bracket
stands at the left hand side of V (cf. footnote 1 on p. 70).
 3. A linear connexion expressed in terms of Sj..).". 133

If the connexion is metric and if g;.,. is real and introduced as a

fundamental tensor (cf. I 9) in each of the local E,.'s, the X,. is called
a U,. and g;.,. is said to be the fundamental tensor of the U,.. A U,. with
a symmetric connexion is called a V,.. 1) This is the case of riemannian
geometry. A U,. or V,. is called ordinary if g;.,. is positive definite. In
a V,. (3. 5) takes the form
(3.11) r;. = {;..} 2)
and from this it follows that
(3.12)
If
(3.13) Q;.>< = Qpt"
the connexion is called semi-metric. From (3.2) it follows that in this
case
(3.14)
If the connexion is semi-metric and symmetric, and if g;.,. is real,
the X,. is called a W,.. This is the case of WEYL's geometry 3 ). Let a
vector v" in a W,. be displaced parallel. Then tJv" = 0 and
(3.15) u.1> g;_" V.. V " = - g;.,. V V"Qp. df:JJ
\>

Hence in a W,. the length of a vector is not invariant for parallel dis-
placement but the ratio of the length of two vectors at the same point
is invariant. If gA,. in a W,. undergoes a conformal transformation
(3.16)
we have
(3.17) 17/g.. ,. = c-aQJJ + a/J a) g.. ,. = c-Q/J + cVoga) 'g.. ,..
Hence if we take 'g;.,. as fundamental tensor instead of g;.,., and if at
the same time Q;. is transformed into 'Q;. = Q;.- 8;. log a we get the
1 ) An A,. may be considered as a V,. if and only if there exists a covariant
constant symmetric tensor g;.,. of rank n. The conditions were investigated by
T. Y. THOMAS and LEVINE 1934, 1; T. Y. THOMAS 1935, 1; 1936, 2; LEVINE 1948, 1.
Several authors investigated the geometries with a fundamental tensor of rank < n.
We mention here only BORTOLOTTI 1930, 2; MOISIL 1940, 1; VRANCEANU 1942, 1;
NORDEN 1945, 2.
2) Note that {;;.} does not change if g;.,. gets a constant factor. Introducing
such a factor in fact means a change of gauge in the whole space without changing
the geometry.
3 ) WEYL 1918,1, p.400; R. K. 1924, 1, Ch. 6; HLAVATY 1929,2. Thesemi-metric
semi-symmetric case appeared in the classification R. K. 1924, 1, p. 75 and was
applied to mechanics by LICHNEROWICZ 1941, 1.
 134 III. Linear connexions.

same connexion. That implies that in a Wn the fundamental tensor is

only fixed to within an arbitrary scalar factor. But a Wn is not the
same as a conformal space. In the latter only lg.,"j is given (cf. I 2)
but no linear connexion can be fixed by lg.,"j alone. Therefore we need
the vector field Q", given for one of the possible fields g.",. Of course
we must exclude the case when Q., is a gradientvector because in that
case it would be possible to fix 'g.", by chosing a in such a way that 'Q.,
vanishes, and that would mean that we did not really have a Wn but a V,,
with an undetermined gauge in which g." could have been given in a
more practical way (determined to within a constant scalar factor). In
one dimensional space every vector is a gradientvector. Hence every
W1 is a V1 with an undetermined gauge. In a Wn (3.5) takes the form

(3.18)

and for I;, we have the formula

(3.19) r I' = -~2 0I' log g + _11_2 Q .

I'

In a Wn the process of covariant differentiation does not commute

with the process of raising and lowering of indices. But, since the
fundamental tensor is fixed only to within an arbitrary scalar factor,
this latter process is not invariant and can be effected only if a pre-
ference is given to one of the possible fundamental tensor fields. If
this can be done in an invariant way we go back to the Vn and in the
other case it is better not to use the non invariant process at all.
The geometry in wn is basedOll two different transformation groups,
the group of the coordinate transformations and the group of the con-
formal transformations of g.," (the "Umeichung" ofWEYL). But in the
above these two groups are not dealt with in the same way; the co-
ordinate transformations are fundamental because they form the founda-
tion for the definition of all objects and the conformal transformations
come in after the introduction of a fundamental tensor to make this
tensor variable. Though quite correct mathematically, this method
seems incongruous from a methodical and aesthetic point of view. 1 )
It is just the same as if, in dealing with some geometry, we first worked
out everything with respect to one definite coordinate system and after-
wards introduced coordinate transformations and raised the question
of invariance. 2) Now it is possible to handle both groups in the same
1 ) My attentionwas drawn for the first time to this point by G. LYRA in his
letter of 27. 6. 1949. Cf. LYRA 1951, 1. The same point of view occurs in HLAVATY
1949, 1; 1952, 1.
2 ) In many 19th century publications this was actually done!
 3. A linear connexion expressed in terms of s;.;,". 135

way by introducing at the beginning a fundamental pseudo-tensor 9.;,.

of dass +
1, defined with respect to an auxiliary variable ~0 with the
transformation ~0' = a ~0 (cf. II 3).1) Then, using the formula [cf.
(2.34)] for the covariant differentiation of pseudo-quantities we get
(3.20)

Equating this expression to zero, taking Sj.j." = 0 and using g;.,. and
its inverse g".a for raising and lowering of indices we get

(3.21) r;"= U',}*- ! (A 11 A:;: + A, A; - A" 911 .a)

(3.22)
where

(3.23)

@ is a pseudo-scalar density of weight +

2 and dass n. From (3 .18,
19) and (3.21-23) we see that A;. plays now the same role as
- Q, before. The only difference is that we have now two definitely
fixed geometric objects g;.,. and A;. instead of the variable quantities
g;.,. and Q;. g..,. being fixed, it can be used for raising or lowering of
indices and according to vp g..,. = 0 this process is commutative with
the process of covariant differentiation. Note however that by raising
(lowering) of an index the dass increases with -1 ( 1). 2) +
It is very remarkable that in fixing a linear displacement the tensor
g;.,. plays such a fundamental part. Why not any quantity with another
valence and another condition of symmetry ? This has been made dear
by WEYL. First let an arbitrary linear displacement be given by its
parameters F11";,. At a point ~" we choose an arbitrary symmetric tensor
g;.,. of rank n. Now we try to form a field g;.,. in 9? ( ~") suchthat g;.,. =g;,,.
0

0
at ~" and that the given displacement from ;" to any point ~
0 0 0
+ d~"
is the result of
1o the displacement from ~" to ~"
0 0
+ d~" belonging to the Riemannian
geometry with g;,,. as fundamental tensor;
2 a rotation of all vectors at ~" + d~" with respect to g;,,. + dg;,,.
as fundamental tensor. 0

1) NEWMAN 1927, 1; ScHOUTEN and HLAVATY 1929, 2, p. 426; EI 1935, 1,

p. 87.
2) Note that the geometry of a Wn is not the conformal geometry. In a con-

formal geometry we have only the pseudo-tensor 9;.,. and no A;,. That implies
that there is no fixed linear connexion.
136 III. Linear connexions.

Such a displacement WEvL calls a "congruent transplantation"

(kongruente Verpflanzung). If v" is a vector at ~", it changes first into
0

(3.24)

where {;,~.}is the CHRISTOFFEL symbol belonging to g",. and afterwards into

(3.25)

where F~" d~" is the bivector of the rotation (cf. I 9). Hence at ~
0

(3.26)

Now F,.;.. is alternating in .v and from this it follows that

(3.27)
or
(3.28)

Now let a field g;." of rank n be given in IJl ( ~"). Then alllinear connexions
0
representing congruent transplantations with respect to g,~.,. satisfy
equations of the foon (3.26) where F~'" is an arbitrary tensor, alternat-
ing in . v. Among these connexions there exists one and only one symmet-
ric connexion because F,.". vanishes if .F[p;.J = 0. Collecting results we
have two theorems:
I. lf a linear connexion and a symmetric tensor g",.
of rank n at a
point ~" be given, it is always possible to find a field g",. in an IJl ( ~) such
0 0
that g,~.,. = g,~.,. at ~" and that the linear displacement from ~" to every point
+ d~ is
0 0
~" a congruent transplantation with respect to g,~.,..
0
II. lf a symmetric tensor field g",. of rank n be given there exists one
and only one symmetric connexion representing a congruent transplanta-
tion with respect to g;.,.
The first theorem was proved by WEYL 1). The second theorem
deals with a special case of the general formula (3.5).
Now one could try to generalize the notion of congruent trans-
plantation by using instead of the group of rotations some other sub-
group of the group Gho which left the volume invariant. Then, if the
following conditions are introduced:
1. If the subgroup and a linear connexion be given, it is always
possible to define the "generalized rotation" in an IJl ( ~") in such a
0
1) WEYL 1921, 1, p. 131; cf. WEYL 1922,2, p. 117.
 3. A linear connexion expressed in terms of Spf'. 137

way that the connexion represents a congruent transplantation from

~" to ~"
0
+
d~" for every d~";
0
2. if the generalized rotations are given at each point of 91 ( ~"),
0
there exists one and only one symmetric connexion representing a
congruent transplantation with respect to this definition of rotation,
it has been proved 1 ) that there exists one and only one subgroup
of G,.o that leaves the volume invariant and also satisfies these condi-
tions, and that this subgroup is the group that leaves invariant a
symmetric tensor of valence 2 and rank n. This theorem makes clear
why in the problern of fixing a symmetric connexion, a symmetric
tensor of valence 2 and rank n plays such an important part.

Exercises.
III 3,1 2). The symmetric part T;;. def !(~;.) depends not only on Q~/'
but also on S ~t. The 'I;.";. are parameters of another connexion for
which
III3,1oc)

III 3,2 2 ). If the connexion T'"'';. is metric, the symmetric connexion

'T'"). = !(~;.) is also metric, if and only if 5'";." is a trivector.
III 3,3 2 ). If g;." and 17'" g,;" are known, T'" is fixed without it being
necessary to give S~j_".
III 3,4. In a metric connexion length and angles are invariant for
all parallel displacements.
III 3,5. In a Wn angles are invariant for parallel displacement.
III 3,6. Quantities in Ln form a ring (in the sense of abstract algebra)
with respect to addition, multiplication and the operators 17'" and 15.
Prove that pseudo-quantities in Wn defined with respect to ~ 0 with a
dass equal to one half of the difference between covariant and contra-
variant valence form a ring with respect to addition, multiplication
and the operators W)-I 17'" and W)-~ 15.
III 3,7 3). A semi-symmetric linear connexion is for n>2 fixed by
the covariant derivative Q~";. of some symmetric tensor g"' of rank n
and the derivative I"= 17'" /'"" of some bivector /'"" of rank n. This
derivative can be replaced by the derivative F'";." = 6l7r'" /;."1 of some
bivector /;." of rank n.
1) WEYL 1922, 2; CARTAN gave another proof, 1923, 1; cf. also KosAMBI 195 2
1, p. 6.
2) EI 1935. 1, p. 84.
3) EI 1935. 1, p. 88.
 138 III. Linear connexions.

4. Curvature.
We consider a field v" or W;. in an 91( .;") and an infinitesimal simple
0
contravariant bivector df";. at .;" represented by a part of an E 2 (with
0
inner orientation) through .;" with a definitely fixed boundary curve
0
passing through .;". Now let the vector v" or W;. at .;" be displaced parallel
0 0
along this curve till it returns to .;". Then it can be proved that the
0
difference between the final value and the initial value can be written
in the form
(4.1) b) + 2 R"w
1
vp "
dj"
where R~~)." is a tensor depending on the T"";. and their first derivatives

(4.2) . "
RvpJ. ctet
= 28 [v T"
p]
+ 2 1T'><[ el
re
p]J.

The rigorous proof of this theoremwill be given after (4.9b). 1) Here

we give a short illustration that is not meant to be a proof. For the
curve we take the quadrilateral of the points .;", .;" + d.;", .;" + d.;" + d.;"
0 0 1 0 1 2
and .;" + d.;". Then, displacing v" from .;" to .;" + d.;" + d.;" via the point
0 2 0 0 1 2
.;" + d.;" we get at .;" + d.;" the field value
0 1 0 1

(4.3) v"- T"

p
v;. d.;"
1

neglecting terms of second and higher orders. At.;"+ d.;" + d.;" we get
0 1 2

(4.4) v" - T"

p
V;. d.;l'
1
- T" d.;l' (v -
p 2
r;. ve d.;) -
VQ 1
(8V T")
p
V;. d.;
1
d.;l'.
2

Going via .;" + d.;" we get in the same way

0 2

(4.5)

and the difference of these two values is

(4.6) - 2(8 [v r" +p] 1T'><
[vJel re )V"dtdtl-
p] '> '> - - 2 R" V ii.df""
1 2
1 vp

because d t" = 2 d .;c d .;"l. If (4.1) is proved to be true for every field v"
1 2
(or w;.) and every bivector /"", R;~j_" must be a tensor because every
1) Cf. TIETZE 1923, 1; SYNGE 1924, 1; T. Y. THOMAS 1925. 1; ScHLESINGER
1928, 1; 1931, 1; 1932, 1; Mc CONNELL 1928, 1; AGOSTINELLI 1933, 1; MaRINAGA
1934, 1; GUGINO 1935. 1; jOHNSON 1948, 1; GRAEUB 1950, 1. If instead of vectors
otherquantities are displaced, other quantities arise instead of R~ j.~". cf. for instance
HO KAR! 1934, 1; HOMBU 1936, 1; T AKENO 1942, 1; BOMPIANI 1946, 1; PETROV
1948, 1; BOCHNER 1951, 2.
 4. Curvature. 139

expression of this kind with four indices alternating in Yfl, whose trans-
vection with every bivector f"P and every vector v" (or w,~) is a vector,
is necessarily a tensor.
R;;t is called the RrEMANN-CHRISTOFFEL tensor or curvature tensor
of the connexion Fp,"A. If it is zero at all points, the connexion is said
to be integrable.
Connecting these results with those obtained in III 2 we see that if
a point with radiusvector v" at ;" undergoes a CARTAN displacement
0
along the boundary curve of the infinitesimal bivector dj>P in the sense
of this bivector, the point will after its return have the radiusvector
(4.7) v"- 2 vp,"v"df"P-
~R" Svp,""df"P
If the process of covariant differentiation is applied twice in succes-
sion we get, after alternation over the indices of differentiation, an
expression containing 5;;." and R;;;.". For instance
17[v 17p,] Wx = 17[v ( Op,] Wx - rp,1,! we)
{
(4.8) = O[v Op,] W- (o[v~~x) We- l[~!XI Ov] We- T[~p,] OQW-
- r[~l.!i Op,] WQ + I[~p,] reaA Wa + r[;l.!l ~Je Wa
or

(4.9a)

and in the same way

(4;.9b)

Using (4.9b) we can now give a rigorous proof of (4.1 a). Let t"
and u" be two vector fields with the parameterstand u and let these fields
satisfy the equation f u" = 0.
t
Then for any field v" it follows from (4.9b) that
" v,~ + (ft
(4 10) (t rv uP V.p, - u 17' tP V.p, ) v" = t uP R vp, ue) 17 v" =
Q
"v-<
t" uP R vp,

Now instead of displacing a vector at ;" parallel along an infinitesimal

0
quadrilateral aswas donein the illustration given above we give the whole
vector field v" a finite parallel displacement over t" t (cf. III 2). We follow
this with another displacement over u" u, then over - t" t and finally
over - u" u. Because f u" = 0 1), after these four displacements every
t
point is back at its original place, and according to (2-37), the new field
1) Cf. Exerc. II 10, 10.
140 III. Linear connexions.

I
value 'v" is
'v" = exp (u u'" 171.) exp (t t'"l7'") exp (- u u'"17'") exp (- t t'" 17'") vx
(4.11) = v" - t u (!" 17. u'" 17'" - u 17. t'" 17'") v" +
= v" - t u t" u'" R " vl +
V

which proves (4.1a). 1)

The operator V[. 17'"1 has the very important property that it satisfies
the rule of LEIBNIZ. In fact, if (/) and lJI are two quantities with sup-
pressed indices, then we have (cf. footnote 3 on p. 132).

17[. 17'" 1(/) P V[. W:u 1(/)) P + V[.(/) 17'" 1P

1
=

(4.12) = (V[. 17'" 1(/J) P + (1/['"(/J) 17.1P + (17[,.(/J) 17'"1P +(/)V[. 17'"1P
= (17[. 171"1 (/)) lJI + (/) 17[, 171"1lJ'.
Now every tensor can be written as a sum of products of vectors and
from this we get immediately a formula for the operator 17[. 17'"1 applied
to a tensor. In this formula there appears a term with R;~;." with the
+ -(- -)sign for each upper (lower) index. For instance, for the tensor
P",~
w

{
17,[V V./"1 P", 1,w
= ~2 R"
V/"Q
" pel
, ,w
+ ~2 R"V/"Q,;. P"e, ,w -
(4.13)
- 1
"2
RVJ..tW
.. a P";.
.a
- SVJ.l.e V.(! P";
.w

If 17[ 17'"1is applied to a scalar density or Ll-density l.J of weight +1

we may deduce in the same way as (4.9a)
(4.14)

But, although r'" is not a vector, the expression 8[.I:1is always a bivector
because
(4.15)

Hence, we get for a scalar density or Ll-density l.J of weight w

(4.16)

and for a tensor density or Ll-density of weight w, for instance j~;.,

according to the rule of LEIBNIZ

(4.17) V[. 17'" 1j~,~ =! R;~~" j;',~-! R;~;.a j~a- 5;/ 17e j~,.-! w V"'" j~,..
In dealing with pseudo-quantities for which a covariant differentia-
tion (that is a set of parameters A,~ besides the r;,.) has been defined
1) Apart from a slight change this is the proof given by MoRINAGA 1934, 1.
2) Cf. III 6 for the geometric meaning of the bivector V,. I".
 4. Curvature. 141

(cf. III 2) we get in the same way an additional term containing the
bivector
(4.18)
instead of V.~' and c (dass) instead of w (weight). For instance for a
pseudo-vector density b"
(4.19)
If a symmetric tensorfield g;," of rank n, its covariant derivative
- Q~'"" and S~t are given, the curvature tensor can be expressed in
terms of these quantities, their first ordinary derivatives and the second
ordinary derivatives of g;.". If we write [cf. (3.9)]
(4.20) Tl'";.={:;.}+ T~,i"; Tp;,'" def
= g"e (- 5 {Pe'-} + .12 Q{Pe'-} )
and
(4.21)
K;~;." is the curvature tensor of the riemannian connexion belonging
to g;,", and T~;." is a tensor representing the difference between the
connexioh r;;.
and this riemannian connexion. If we indicate the
covariant differentiation of the connexion {~'";.} by 17* we get from (4.2)
and (4.20)
(4.22a)
and also
(4.22b) R;~;_" = K;~,i" + 2 17[,, T~Ji"- 2 T[; 1; ( T~Jie + 2 5; ;e T;;".
1

From (4.22a) we may derive the relations

(4.23) l Rp -K
- p).
+ v"
~ T"
R p )ctef
.-
pl.

R"'"
XJ'A '
~ T"+
-VI'

K
";.

-
p-
K
xe . p). - T"Tl!
T"Te
ct~f K ""
p - "~''
l'e >< '

(4.24) V"p = * T~]~"

2 Vfv = 2 EJ[v T;] ~"; (cf. 4.15).

Rl';. (Kp;.) is called the RICCI tensor of the Ln (Vn)

(4.22) is a special case of a more general formula. If A~;." is an ar-
bitrary tensor, T;;. = r;;.
+ A~;." represents another connexion and the
relation between the curvature tensors of these two connexions is easily
proved tobe
(4.25) 'R;~;," = R;~;." + 217[.A~Ji" + 2S;;e A;;."- 2A[; e A~J/
1;. 1

In a WEYL connexion we have according to (3.18)

(4.26) T;,i" = i (Q~' A + Q;. A;- Q" g;,~')
142 III. Linear connexions.

and from this it follows after some calculation

(4.27) { R;;;." = K;;)." + Vr: Ql'1 A~a- g[[e (:171'1 Q;.J + Ql'1 Q.u-
- 2 Qjaj Q gp]]) f
In a metric and semi-symmetric connexion we have according tc
(2.14) and (3.5)
(4.28) T;;." = - 5;. A~ + 5" g;.p
and this leads without difficulties to

(4.29) R~;j." = K~;t + g" 0 gc.ce (4 171'1 5;.J- 451'1 5;.1 + 25Jal 5" gpJ.<1).
An L .. is said to possess teleparallelism 1 }, 2} or absolute parallelism
(Fernparallelismus) in in (~") if for any two points P and Q of this region
0
the parallel displacement of any quantity from P to Q along a curve
of in (~) connecting P and Q gives a result at Q which is independent
0
of the choice of the curve. According to (4.1) the curvature tensor
vanishes in this case. Naturally an E .. possesses teleparallelism. In
order to prove that every A., with teleparallelism is an E .. , we take n
linearly independent covariant vectors ~;.; h = 1, ... , n at ~" and by
0
parallel displacement we construct n covariant constant fields ~;. in
an in (~"). Because 8rl' ~;.1 = 17[1' ~;. 1 = 0, the ~.. are. gradientfields. If
. 0

~.. -~ 8" ~h, the ~" are independent because their gradients are linearly
independent. If we take them as new coordinates we have

(4.JO)
and this proves that (h) is a rectilinear coordinate system in an E ...
The same does not hold for an L... In fact an Ln can be endowed with
teleparallelism without being an Ew 3 )
Here is another proof showing the relations of R;;j." to integrability
conditions. We look for a holonomic coordinate system (h) in Ln suchthat
the If~ vanish. The differential equations are [cf. (2.5)]
a) 81'~" =A!
(4.31) {
b) 8PA1 =A:Fp".<
1 ) Cf. WEITZENBCK 1923, 1; VITAL! 1924, 1; GRISS 1925, 1; CARTAN and ScHOUTEN
1926, 3; 4; CARTAN 1927, 2; BORTOLOTTI 1927, 1; 2; 1928, 1; 1929, 1; EINSTEIN
1928, 1 ; WEITZENBCK 1928, 1 ; SCHOUTEN 1929, 4; E I 193 5, 1, p. 88; HAIMOVICI
1943. 1.
2 ) For Iiterature on absolute parallelism in the large, cf. footnote 2 on p. 125.

3 ) lt is proved in Chapter IV that every non-trivial group space is such an Ln.

 4. Curvature. 143

with the integrability conditions

(4.32) { a)
0= arv A~J = A~ 5~/
0 -a [v afl Ah
-!] -(a [v A :xJ )F"' " [v F"
pj J. +Aha p]i. --
b) 11 1 A"R"
- - 2 " v Jl J.

These conditions are satisfied identically if and only if S;l" = 0 and

R~;~"= 0. 1 )
If the fundamental tensor in a vn is written as a product of two
equal ideal factors (cf. I 11), g;." = a1 a" = b1 b" = , we have
(4.33)
The expression (1711 a1 ) a" has no meaning by itself, only the sum of
(17" a1 ) a,. and (17" a,.) a 1 is known tobe zero. This enables us to simplify
the calculations by taking (1711 a1 ) a" = 0. Then we get for instance
(4-34) v V " =V17fl V " a.l a " = a"aI' V" a.l =
17fl afl V"+ V ""a
a I' a,;,
hence
(4-3 5)
and
(4.36)
hence
(4.37)
Using this symbolism many calculations can be shortened considerably
(cf. I 11 under 6, footnote 6, p. 58).

Exercises.
III 4, 1 2). A man is moving on the surface of the earth always facing
one definite point, say Jerusalem or Mekka or the North pole. Prove
that this displacement is semi-symmetric and metric and compute 5.:~. 3 )
During the mathematical congress in Moscow in 1934 one evening some
of us invented the "Moscow displacement ". The streets of Moscow are
approximately straight lines through the Kremlin and concentric
circles around it. Now let a person walk in the street always facing the
Kremlin. Prove that also this displacement is semi-symmetric and
metric and compute S .<
III 4,2. Prove that for a symmetric connexion the A.-rank of R~~)."
can not be greater than the v-rank. Fora metric and symmetric connex-
ion the rank of K,""" is the same for all indices.
1 ) RIEMANN 1861, 1, p. 402, LIPSCHITZ 1869, 1, p. 94; 1874, 1, p. 109; RICCI
1884, 1, p. 142, all for vn.
2 ) Cf. EI 1935, 1, p. 127, Exerc. 11,9.
3) HESSENBERG 1924, 1 gave other examples of teleparallelism on a sphere.
144 III. Linear connexions.

III 4,3. Prove that {cf. 4.20)

li I 4,3 1X) t R~~i"- 17lv T ;J ;_,. - S~~g T~;_,. = ~ K;~;_" + 17lv
* T; ;_".
1

III 4,4. If the vector v,. is displaced parallel along the boundary of
the infinitesimal simple bivector df~' the difference between the final
value and the initial value lies in the plane of df~' if and only if
III 4,4 IX)
III 4,5. Prove that in an Ln teleparallelism for scalar densities exists
if and only if V.,.= 0 (cf. III 6). The connexion is called in that case
volume preserving 2 ).

5. The identities for the curvature tensor. 3)

From the definition {4.2) we get the first identity

(5.1) I R(~~)i" = 0 I
and by altemation of (4.2) over vp, we get the second identity

(5.2) I R[~~;_j = 217[v S~;_j- 4S[;~e S.i1 ~,. I

taking for semi-symmetric connexions the form
(5.3) Rl;;;_j = 2Ai.l7,.. 5"1 4)
and for the symmetric case the form

(5 .4)

If we write (cf. 4.15, 23)

def R ...
Rp.t=
(5.5) .,.;. '

we get from (5.2) by contraction

from which for the semi-symmetric case

(5.7)
1) Cossu 1949, 1.
2)VEBLEN 1923, 2; ScHOUTEN 1923, 1; EISENHART 1923, 3. Cf. GOLAB 1931, 1.
3 ) Cf. for complete sets of identities for different cases T. Y. THOMAS 1934, 2;

LEVINE 1936, 1; SuN 1937, 1; BLuM 1947, 1.

4) The case where (5.3) holds for a not semi-symmetric connexion was investi-

gated by V. D. KULK 1939, 1.

 5. The identities for the curvature tensor. 145

and for the symmetric case

(5.8) 12Rrp] = - v~'" I

By differentiation and alternation we get from (3.2) and (4.13)
(5.9)
from which follows the third identity

(5.10) I Rvp(><) = II[. Q!']>< + S;~e QQ>< 11).

This identity takes the following forms for special cases

semi-symmetric connexion
(5.11)

semi-metric connexion
(5 .12)

metric connexion

(5 .13) I Rvp(><) = 0 I
Every quantity Q.,"'", alternating in Yfl satisfies the identity

(5.14) { Q,<V!' = Qvp><- ~ + Q[><,uv] + Q[,uv]>< +

{Q[><v]p + Q[><p]v)

+ Qv><(ft}+ Qv(,u")+ Q,.p(v)+ Q,u(,.v)+ Qpv(><)+ Q><(pv)

If (5.2) and (5.10) are substituted in (5.14) for Q.I',~"=R.u"" we get
expressed in terms of s;~", 17r.s;~r. Qp;.,. and 17r.Q.uJ"'
Rv,u,t,.-R><V!'
This rather complicated formula is the fourth identity. For the metric
and symmetric case it takes a very simple form. For this special case
(riemannian connexion) we collect the four identities
I) K(vp),t"= 0

(5.15) 1 II)
III)
K[vp]x

Kvp(><) =
= 0

K.u"" IV) = K,~".~'.

From (II, III, IV) it follows that
(5 .16) II') Krvphl = 0.
* l) CASTOLDI 1947,
* Ql'l"" at
1 finds only one term, 17rv
the right hand side, where
17is the operator of the riemannian connexion belanging to g,~ ". This seems to
be a mistake.
Schonten, Ricci-Calculus, 2. Aufl. 10
146 III. Linear connexions.

It can be proved that (I, II, III, IV) is equivalent to (I, II', III, IV).
The identities (I, II', III, IV) can be used to evaluate the number of
independent components of Kvp ;.~. According to (I, III, IV) Kvp>< is
a symmetric bivector tensor (I 8 under B) i.e. a symmetric tensor in the

+(;) {(;)
(~)-dimensional space of all bivectors. 1) Now a symmetric bivector tensor
has + 1} = ~ (n 2 - n) (n2 - n + 2) components. But (II') stands
for (~) independent relations. Hence the number of independent com-
ponents is }2 n 2 (n 2 -1), that is 1 for n=2; 6 forn=3 and20for n=4
(the case of relativity).
From (5.15) it follows that V." is always zero in a Vn and that K";.
is always symmetric there. Kp). is the RI~CI tensor of the vn. The
principal directions of K";. are called the principal directions of the Vn. 2)
The equations (4.2)
(5 .17)
have the form (II 6.15) with a right hand member that depends on the
r;;.. As we have seen in II 6 the construction of the integrability
conditions depends on the possibility of being able to eliminate the
derivatives of the T"";. at each step. N ow for (5 .17) this possibility
exists and we get for the first set of integrability conditions

0 = T[::,lal T.lel T"~;.- tR[;,;.I;I~ T$p + Fc::gl I',~ a: ~~;.-

(5.18) 1 t 17[wR;;l~"
-
+ S[;,;aRI~I,:lt.
+
tFc: 1e R;,~J~e t 8[w R;;11"
=

1f 8e is applied to the right hand side and if we alternate over (!WVfl

the expression is identically zero. 3 ) Hence the only integrability con-
ditions are
(5 .19)

known as BIANCHI's identity. 4 ) The identity can be proved in many

other ways, for instance by computing Vrw V"l7"l W;. from 17[w V"ll7" W;.
1 ) The geometric properties of this quantity were for n = 4 investigated by

ClruRCHILL 1932, 1; RusE 1936, 1; 1944, 1; 2; 1946, 1; 2; 1948, 1; LANczos 1938, 1;

d. WRONA 1948, 1 for general values of n.
2 ) Rrccr 1903, 1; 1904, 1. Cf. EISENHART 1922, 1. Other prir:icipal directions

were defined by SYNGE 1922, 1.

3 ) This result will be obtained later in a much shorter way.
4 ) An identity of this kind was first considered by Voss 1880, 1. Then Rrccr

found it independently and sent it to PADOVA, who published it in 1889, 1. In

1902, 1 it was proved independently by BIANCHI. Therefore italien mathematiciens
 5. The identities for the curvature tensor. 147

and from l7w l7[v 17fl 1 W;_. For a semi-symmetric connexion it takes the
form
(5 .20)

and for a symmetric connexion we have

(5 .21)

By contraction of (5.19) over uA. we get for the general case:

(equivalent to b) 8[w V,fl 1 = 0),

for the semi-symmetric case:

(5.23)
and for the symmetric case:

(5 .24)

By contraction of (5.19) over wu we get for the general case:

(5.25)

for the semi-symmetric case:

(5 .26)
and for the symmetric case :

(5 .27) 117"R;;,i"- 2l7[vR,u].l = 0 I

often speak of the identity of Rrccr-BIANCHI. All these publications consider the
metric symmetric case only. BAcH gave 1921, 1 the proof for a WEYL connexion
and the author at the Jena congress 1921 did the same for a general symmetric
connexion, cf. 1923, 2. In 1922, 1 VEBLEN dealt with the symmetric case and in
1923, 1 WEITZENBCK published the formula (5.19). Cf. SCHOUTEN-STRUIK
1923, 3. As a more recent publication we mention LEVINE 1936, 1. WAGNER gave
194 5, 2 a generalization for general geometric objects.
1 ) For an An the second identity and BrANcHr's identity can be proved in a

very short way by using normal Coordinates (cf. III 7, Exerc. III 7,4). Another
method is used in III 9.
2) BLUM proved 1946, 1 that BIANCHI's identity in a Vn consists of .}4 n 2 (n2 -1) X
(n-2) independent equations. Cf. BLuM 1947, 1 for complete sets of identities
10*
148 III. Linear connexions.

For a riemannian connexion we get from (5.27) by transvection

with g~";.
(5 .28) 17" K;" - V. K + 17" K;" = 0;
or
(5.29) I VI" G~;. = 0; G,,;. ~ K";.- ~K g";, 11).
The scalar X def ~~1 - K is called the scalar curvature 2) of the
n (n-1)
vn.
For n = 2 we have G;." = 0, and x is equal to the gaussian curvature.
For n = 4, G;." is the tensor of momentum and energy in general
relativity and (5.29) is the well known equation of conservation of
momentum and energy.
An Ln in which V" R~";. = 0 is called a RICCI space. If R~";. = 0 it is
said to be a special Ricci space. A Vn, n > i in which K~";. differs from
gp;. only by a scalar factor is called an EINSTEIN space. 3 ) If Kp;. = 0 it is
said to be a special EINSTEIN space. According to the definition of K
and x we have in this case

(5-30)
For n > 2 it follows from (5 .29) that the scalar curvature of an EIN-
STEIN space is constant. But a Vn with a constant scalar curvature
need not be an EINSTEIN space. 4) This is a theorem of HERGLOTZ. 5)
An EINSTEIN space is for n > 2 a special case of a Riccr space. 6)
A Vn, n > 2 in which K. ~" ;." differs from g[ [;. g_ul "J only by a scalar
factor is called an Sn or a space of constant curvature. In this case we have
(5.31)
For n = 2 an 5 2 is defined as a V2 in which x is constant.
1) BrANCHI 1902, 1; Rrccr 1903, 1; for n = 3.
2) WEYL 1921, 2 used the term "scalar curvature" in another sense. He
called an Sn a "space of scalar curvature ". Cf. footnote 4 and VI p. 289, footnote 1.
3 } Cf. EI 1935, 1, p. 125; FIALKOW 1938, 1; 1939, 1; 1942, 1; T. Y. THOMAS

1938, 1 (invariants); HAANTJES and WRONA 1939, 1 (and Cn); YANO 1943, 1
(conf. and concirc); WRONA 1947, 1 (Cn); KurPER 1950, 1; 2 (conf. and proj.);
1951, 1 (conf.); TACHIBANA 1951, 1 (parallel fields).
4 ) The opposite assertion in LovELL 1934, 1 is erroneous and perhaps caused
by a wrong interpretation of WEYL's term "scalar curvature ". Cf. footnote 2
and footnote 1 Ch. VI, p. 289.
5 ) HERGLOTZ 1916, 1, p. 203. Cf. for generalizations WRONA 1941, 1; BOM-

PIANI 1950, 1; VARGA 1943, 1.

6 ) There are many other examples of spaces whose curvature tensor satisfies spe-

cial identities. We mention here only a paper of v. D. KuLK 1939, 1 where the case
+
(n 1) R;(;.~/' = (Vv(p- Rv(p) A~) and especially R;(;.~( = 0 or R;;_,~" = R[;;.~]"
was investigated.
 s. The identities for the curvature tensor. 149

An Sn is a special case of an EINSTEIN space.

The second identity and the identity of BIANCHI can be put into
another form by introducing another invariant differential operator.
Let P;., ... J.p A be a tensor alternating in the first p indices and let A
be a collecting index standing for any number of co- and contravariant
indices. Then we write
p
(5.32) Vr[p P;., ... J.p]A-
ct_e!v,[p P;., ... ).p]A +PS"P
[p-1 ,crj}., ... J.p]A
1

p
As we see the significance of Vrp ... J is that the alternating part 5;1" of
r;;. does not apply top the alternated indices
p
1 ... Ap. In an An, the only
..1
difference between V" and V" is that V" can be used only if we alter-
o 0
nate over the first p indices. In V" there is no alternation, hence V"
is identical with vp. For this new operator the rule of LEIBNIZ holds

(5-33) l
in the form

ptr: P;., ...

=
ApjAj
p
Q", ... ><q]B
(Vr" P;., ... ApjA) Q", ... ><q]B + Pr;., ... Ap:AJ
q
v" Q", ... ".]s
P+l P

l
and for the Operator flrv VP .. .] the rule of LEIBNIZ takes the form
P+l P
(5.34) VrV.P
[v p J., ... J.,jAI QAB+IJ.p]B
s+l s p-s+l p-s
=VrV.P
( [v p J., ... J.,jAj) Q;.,+, ... J.p]B + R[J., ... A,jAj V.v V.p Q;.s+I ... J.p]B .
These identities and the identities

(5.35)

that are valid in Ln (not only in An) and simplify the equations (4.9),
are easily verified and from them and (5.32) we get the identity

1) The new operator was introduced by CARTAN. The general idea was in the
first edition 1928, 1, p. 211 and the operator itself, symbolized by D, appeared in
the second edition 1946, 1, p. 209.
150 III. Linear connexions.

In order to find the geometrical meaning of the new operator we

consider an Xp+l through ~" and an infinitesimal (properly chosen)
0
closed Xp in this Xp+l surrounding ~". Taking for simplicity the field
0
P;,, ... ip~" the field value at a point ~" of Xp is

(5 .37) P
O t " + (~~"
... p.T - ~~")
O "
8!' Pt ... p,T'

If df"' .. "P is the infinitesimal volume element of the Xp with a screw-

sense that is the same for all points, then

(5 .38) dj-<, ... .<p (P ."

O 1 ... p T
+ (~~" - O~~") 8 !'
p1 .. , p. ", T )

is a well defined tensor at f'. But the formation of the integral over Xp
of this expression has no geometrical sense because addition of tensors
at different points is not an invariant process. Now if we displace each
of them parallel to ~" we get

l
0

dj-<, ... .<p {P "

O , .. ,p,T + (~~" - ~~") ."
8!' p, ... p.T +
(5.39) 0
+ (~~" - 0
~" F"
~) I'Q
Q
Pt ... p.T- (~ ~" -
0
~ ~" ) a
ri'T
"
P, ... p,a}

and since these tensors can now be added we can form the integral
over Xp. Then, using the theorem of STOKES we can form the integral
over the part Tp+l of Xp+l bordered by Xp. Neglecting terms con-

l
taining ~"- ~" in the result, we get
0

" + T."[v!ei p.<, ... p],r-

J {8[v P)., ... p].r r,a.
[Ir,
pQ "} djv.<, ... p
, ... p].a
(5 .40) Tp+l p
= J V. p p].r [v
. " dfv, ...
1 ...
p
Tp+l

if d r-<, ...
p is the infinitesimal volume element of Tp+l This result also
holds mutatis mutandus if there are more contravariant and covariant
indices u and -r. The integral is zero for every choice of the (p 1)- +
P
direction of XP+l at ~" if and only if 17[ P;,, ... Jc~l~ r = 0 at ~". In that
0 0
case we say that the field P;,, ... i./ r is in equilibrium at ~".
0
If we compare (5 -32) with (5 .2) and (5 .19) we see that these equations
can now be written in the simple form

(5.41) R[;~it = zv[.s~;,t J (second identity)

2
(5 .42) 17[w R;~]" = 0 (identity of BlANCHI).
 5. The identities for the curvature tensor. 151

Hence BIANCHI's identity expresses the fact that the curvature tensor
is in equilibrium at each point of Ln.
In order to obtain the geometrical meaning of (5 .41, 42) we have to
make use of CARTAN displacements. We choose an X 3 through ~" and
0
in this X 3 we choose a closed X 2 surrounding ~". Let v" be a vector
0 0
t>< and let T"
at % s" be field values at %
R " and o~';.
o"';.'ovp,J. t>< Let djVfl be the
bivector of an infinitesimal surface-element of x2 whose bordering xl
goes through a point ~" of X 2 The point with radiusvector v" at ~"
0 0
now undergoes a CARTAN displacement from ~" to ~", along the border
0
of df"' in the sense of df"' back to ~" and finally from ~" to ~". From
0
~ to ~" the radiusvector changes into
0

(5.43)

At ~" we have the field values ~~.i"+ (~w-r) owS~;." and f~~,i" +
(~w- ~w) ow R~~";. ". Hence the change of the moving point due to the

l{-
0
displacement around df"' is given by the vector (cf. 4.7)

S "-
0 Vfl
(~w- ~w) oW SVfl"- 2fR "v;.
O V p,}. O
+ 1R
:J " (~w -~w) +
O V !l W
(5 44) 0 0
. +lR"(~w-;w)T;. ve_1(~w-~w)o
:J 0 vp,J. 0 0 we 0 2 0 wR"v;.}dfv"'.
vp,}. 0

For the integration of this vector over X 2 we must first displace it

parallel to ~" in order to make addition possible. Then at ~" we get
0 0
the vector

According to the theorem of STOKES the integral of this vector over

X 2 is equal to

J{- s[~~e r::]e- ~R[~~i)/

r, 0 0 0
r::]e- 8[w s~~t +
VJ.
0 0

(5 .46) + 2IR "+IR

0 [wvp,]
"ra
2 0 [v!' Iai ve
0 w]e 0 - 2tvJ.8
0 [w R
vp,]J."}dfwvp,

J{- V[w S;~t + ~ R[~;.~{- ~V}.

2 2
= v[w R~~];."h"~e" arvp,'
Ta 0 0

and this integral is zero in consequence of the second identity (5.41)

and BIANcm's identity (5.42). Hence these two identities together
express the following geometric fact :
 152 III Linear connexions.

If an infinitesimal closed X 2 in an X 3 through ~,.in X,. surrounds ~"

0 0
and if a point with radiusvector v" in the local E,. of ~,. undergoes a CARTAN
0 0
displacement from ~,. to a point ~" of X 2 , after that along the border of
0
a surfaceelement df"~' of X 2 in the sense of df"~', and finally back from
~" to ~", the dilference between the end value and the beginning value at
0 .
~ be called the CARTAN-deviation of the point v" with respect to the
0 0
surface-element df"~'. The integral of the CARTAN-deviation over X 2 is zero
for every (proper) choice of X 3 and X 2 , to within infinitesimal quantities
of an order higher than 1 in ~" - ~,. .
0
There is an identity connecting the curvature tensor with the LIE
derivative of r:;.
(5.47) I fT" ;. + va R
v pJ. =VI' v" "
Ufl I' v"
;. def
- 17:;. v"- 2SJ.e
.. "vl!

that can be derived immediately from (II 10.34). The quantity v;." is
a covariant derivative of v" formed after changing the order of the
lower indices of r:;..
It frequently occurs in problems of deformation.
In the general formula for the operation of f on a quantity in A,.,
V
for instance (II 10.29), it is allowed to replace oP by V.u in all terms at
the same time. This is clear because in an A,. the coordinates can
always be chosen in such a way that r:;.
= 0 at some given point
(cf. III 7). But in an L,. the same rule holds provided that VP v" is
changed into v; ". For instance
a) fu"=v~'o u"-u~'o v"=v~'V u"--v~'F.";.u;.-u~'o v"=v~'V. u"-u~-'v"
(5.48) { V /-' /-' /-' /-' /-' /-' p
b) : W;.= v~' oP w;.+ wP o;. v~'= v~' VP w;.+ wP v;.~'.
V

Now, making use of the commutability of : and

V
o" (cf. II 10) we get
for :V VP u"
(5.49)
hence
(5.50)

is the commutation formula for ;t and VP. If this formula is written out
V
we get
ve V. V u" +(V u") vl!- (V. ue) v"
(5.51) { (] I'
+ + u;. fF."
(] I' I' (]
=(V.p ve) VI! u" ve V" VI! u"- (VP ul!) v"-
e ue V.P v"
e v p;.

from which (5.47) follows, according to (4.9b).

 6. Integrability conditions in L,.. 153

Exercises.
III 5,1. Prove (5.2) by writing out II[. ~ V;.1 p in two ways.
III 5,2. If a vector v" in V2 is displaced parallel along a closed curve,
the angle between the beginning position and the end position is equal
to the surface integral of x over the area bounded by the curve.
III 5,3. In a Vn the expression - t K.,.)." f"" f", where /"" is a
simple bivector with the area 1, is called the riemannian curvature of
the Vn with respect to the 2-direction of /",. at the point considered. For
n = 2 it is equal to x. The principal directions of the symmetric tensor
KpA are the principal directions of the Vn. If they are determined
uniquely and if (h) is an orthogonal system whose basis vectors are
the unitvectors 1:", . , i" in these directions, prove that K 11 is the sum
1 n
of the riemannian curvatures with respect to n -1 arbitrary mutually
perpendicular 2-directions through the direction of z"" and that - G11
(n; f'
1
is the sum of 1) arbitrary mutually and to the direction of per-
pendicular 2-directions. 1)
III 5,4. (Theorem of ScHUR.) A vn is an Sn if and only if at every
point the riemannian curvature with respect to a 2-direction is in-
dependent of the choice of this 2-direction. 2)
III 5,5. A Vn is an S,. if and only if the components Kkfih with
three different indices with respect to every local cartesian coordinate
system at every point vanish. 3)
III 5,6. Prove the identity for Vn 4 )
III 5,6 ot) llf:eal K.,.).,. + Vr."l K).,.ea + llf:A><J Keavp = 0 llf:eal det Vre VaJ
2
III 5,7. Prove (5.41) by applying V to the identity

III 5,7 ot)

2 1
Check (III 5,7ot). Prove (5.42) by computing llf:w V. V" 1w). in two ways.

6. Integrability conditions in L ....

In II 6 we have developed integrability conditions for sets of
partial differential equations in Xn. In an Ln differential equations
1) HERGLOTZ 1916, 1. Cf. EI 1935, 1, p.117.
2 ) ScHUR 1886, 1; cf. EI 1935, 1, p. 126; WRoNA 1941, 1 and 1947, 1, also for

generalizations: EISENHART 1949, 1, p. 83. Cf. footnote 5 on p. 148; BoMPIANI

1950, 1.
3 ) R. K. 1924, 1, p.171; EI 1935, 1, p.126.
4) Cf. WALKER 1950, 1 and (VII 4.45).
154 111. Linear connexions.

occur containing the covariant instead of the ordinary differentiation.

We consider equations of the first order formed by covariant differentia-
tion and alternation. A very general form is
p
(6.1) V[. P~.... ~,.J~i::. = Q;~.... ~,.~L
where Q depends on the ~" only and not on the components of P. If
this equation is written out, on the left hand side we get 8[.P~.... ~,.J~.L
and on the right hand side we get terms which are linear in the com-
ponents of P and contain a factor r;;..
But these are equations of the
form (II 6.15) ; the only difference being that the right hand side is
not independent of the components of P. As we have seen in Chapter Il
the integrability conditions can only be found easily if the derivatives
of P arising at every step can be eliminated. by means of (6.1). Now
this is just the case here, hence, proceeding in this way we must get
at the first step equations containing first derivatives of Q and no
derivatives of P. The equations are obtained by applying 8., to (6.1)
and alternating over w v 1-ll . .. 1-lp. But instead of 8., we can just as weil
us/v~ because the difference vanishes by virtue of (6.1). But if we use
this operator we have nothing to compute because the result is already
known from the preceeding section (5.36):
P+l .
(6 2) 17;[w Qvp,
... p,.].;,
" ... - 1 R [wlel
.. -'2 e + -'21 R [wvj).j
"Pp, ...p,.].). a Pp "
1 ... pp].a-.

For p = 0 we get
0
(6.3) V. P" .. -V. P"" - Q" ..
V ),, " - V ).., - V )..,

with the first integrability conditions

1
(6.4) V,[w Q" ... -_1R"p
l ;,.,, '2 wve ... +-1Rapu
... '2 wv). ... -
a ...

and if P is a scalar
(6.5)
(6.4) takes the form
(6.6)
Applying the general rule to the equation
(6.7) V. v" ...... = 0;
we get the first integrability conditions

(6.8)
 7. Geodesics and normal Coordinates. 155

Hence in an L,. a covariant constant n-vector field (or a scalar density-

field) is possible if and only if V." = 0. This can also be proved by using
(5.22b). It .is in accordance with the property of volume preserving
of an L,. with V."= 0, mentioned already in Exerc. III 4,5. 1)

7. Geodesics and normal coordinates 2).

A curve ;'' = f" (t) in L,. is called a geodesie if the tangent vector
d;"jdt at a point remains tangent if it is displaced parallel along the
curve.
The n. a. s. conditions are
(7.1) ~ d~" = ot(t) d~"
dt dt dt
or
(7.2)

from which we see that the geodesics only depend on I{;.<>. If a new
parameter z is introduced
(7.)) z = z(t); t = t(z)
the equation (7.2) takes the form

(7.4) d2 ~" + F." d~P d~A = _ d~" (d 2 z _ ot!:!__) (!:!_)-2

dz9 pA dz dz dz dt 9 dt dt

and the right hand side of these equations reduces to zero if z is a solu-
tion of the ordinary differential equation of second order
d z
2 ) dz
(7.5) TtB- ot(t dt = 0.

If this is the case, z is called an affine parameter on the geodesic. The

general solution of (7.5) has the form
(7.6) z = C1 f efa.dt dt + C2 ; C1 =l= 0; C1 , C2 constants
1) Cf. for a generalization URABE 1940, t.
2) General references: RIEMANN 1854, 1 (for V,.); VEBLEN 1922, 1; 1927, 1;

BIRKHOFF 1923, 1; VEBLEN and T. Y. THOMAS 1923, 1; 1924, 1; VEBLEN and

J. M. THOMAS 1925, 1; T. Y. THOMAS 1925, 2; 1926, 2 and 1927, 3 (projective);
1928, 1; 1936, 3; J. M. THOMAS 1926, 2 (projective); HLAVATY 1926, 1; 2; 3;
1927, 1; 1930, 3; T. Y.THOMAS and MICHAL 1927, 2; EISENHART 1927, 1, 41
and 1930, 1 (projective); DUBOURDIEU 1927, 1; 1928, 1; NALLI 1928, 1; BORTO-
LOTTI 1929, 2; 1930, 3 (asymmetric); 1940, 2; MrcHAL 1930, 1; 2; 1931, 1; 2;
MATT! OLl 1930, 3; LEVY 1930, 1 ; WHITEHEAD and WILLIAMS 1930, 1 ; RusE 1930, 1 ;
1931, 2; 3; DOUGLAS 1931, 1; WHITEHEAD 1931, 1; EI 1935, 1, p.100ff.; SASAKI 1941, 1
(conformal); BOMPIANI 1945, 1; BLUM 1947, 1; NARLIKAR and PRASAD 1948, 1
(relativity);VARGA 1950, 1;PINL1951,1(asymmetric); HARTMANN 1951,1.
 156 III. Linear connexions.

and accordingly the affine parameter is fixed to within the place of the
nullpoint and a constant factor. This implies that two segments on
the same geodesie have an invariant ratio of "length" that can be
measured by means of one parameter chosen arbitrarily from all possible
affine parameters.
In an ordinary Vn the length .s on a real geodesie is always an affine
parameter because d;"jds is a unitvector and the covariant differential
of a unitvector is always perpendicular to the vector. Hence, if we
write i" for d;"jds, the equation of a geodesie in vn takes the form

(7.7) i~'V. i"= az~" +F" d~l' ae = 01).

I' ds 2 I' ds ds

We may ask whether it is possible to transform a connexion F~'";. in

such a way that all geodesics remain geodesics. If
(7.8)

P~~" is necessarily a tensor (cf. II 10) and we may take P[;.i.{ = 0

because the alternating part does not affect the geodesics. According
to (7.2) we must have
(7.9)

for every choice of d;''jdt with a function (t) that depends on this
choice. But this is only possible (cf. Exerc. I 3,4) if P~/' has the form
(7.10)

If (7.10) is satisfied the new affine parameter 'z has the form

(7.11) C1 , C2 constants

and from this it follows that z is an affine parameter for the trans-
formed connexion if and only if P;.de = 0 at all points of the geodesie
considered, i.e. if at all points P;. is tangent to the geodesic. z remains
affine parameter for all geodesics if and only if P;. = 0, hence a symmetric
connexion is wholly determined by its geodesics and the affine Parameters
on them.
As we have already seen in III 4 a coordinate system in an A,. with
the r;;.
vanishing at all points exists if and only if the An is an En.
But it is easy to find a coordinate systeminAn for which the vanishr;;.
at one arbitrarily given point ;". Such a coordinate system is called
0
1) The expression r:j, i" has a sense only if the field i" is prolonged (cf. II 4)
but il' 171' i" always has a sense.
2 ) Cf. EISENHART 1926, 1, p. 132; T. Y. THOMAS 1934, 2, p. 9.
 7. Geodesics and normal Coordinates. 157

geodesie at ~". VEBLEN has given the following general method for the
0
constructiori of a coordinate system of this kind, that has moreover
many other useful properties. At ~" we take an arbitrary vector t"=j= 0
0 0
depending on n -1 parameters, in such a way that to every direction
at ~ there belongs one of the vectors t" with that direction. Then we
0 0
consider all geodesics through ~" and on each of these geodesics we fix
0
an affine parameter z such that at ~"
0

(7.12) d~"=t" z=O.

dz o '

Now the equation of a geodesie enables us to compute for each geodesie

the value of d2 ~"jdz2 at ~
0

(7.13) d2 ~" (;) =- F.";. (~) t~-' t).

dz 2 o ,_. o o o

and by repeated differentiation of the equation of a geodesie we can

find all derivatives of the ~" with respect to z at ~
0
dJ'+IJ:><
(7.14) --~ (~)
dzP+I 0
= - F." (~) [1-'P
f-'p ... p,A 0 0
fl-' 1 [
. 0 0

where 1 )
(7.15) F." det a F" - pl'/(/-'p/-'p-1 F"

" "(
f-'p ... p,A- (/-'p /-'p-1 .. 1-'I) f-'p-, ... p,A)z

l
Then ~" can be expanded in a series

d ~" )
~=~+- 1 ( d~ ~" )
z+-- z+
2
(7.16) 0 dz z=O 2! dz z=O 2

= ~" + Z 0t"- _!_r"

0
(~) z 0t~-' 0t l - _!_r"
2! pA 0
2 (~) z3 0t~' 0t~-' 0tA- .
.3! ,_.,,_.,;. 0

If this series is convergent the expressions z t" can be used in an !n (~")

0 0
as new coordinates. Taking f' = 0 for simplicity and writing ~h;
0
h = 1, ... , n for these new coordinates

(7.17)
we get
(7.18) ~" = ~"h ~h -
2! F"
_!_ p).
(0) ~!JA
,.
~i ~i- _!_ ,_.,,_.,;. (0) b'fl'~
3! F" 1 " .
~;, ~ ~i- .. '

1) BORTOLOTTI and HLAVATY remarked 1936, 2 that the construction of ~~ ... ,_.,
is analogous to the construction of a covariant derivative for the covariant in-
dices only.
158 JII. Linear connexions.

and the inverse

(7.19) ~~~ = <5: (~" + :, A;;. (OH~' ~J. + ~(A;,p,J. (0) ~~' ~"' ~A + .. )
where

By differentiation of (7.18) and (7.19) we get at ~"

0

A1(~) =<51
0

a~'A1(~) = b:J;."J.(~)
0 0

o. 8" A1(~)
0
= b:A:"A (~)
0

and accordingly at ~"

0

(7.22)

Now the choice of the system (") is entirely free. Hence, if we take
(") equal to (h) it follows from (7.18) that at ~"
0

(7.23)

andin consequence of (7.15) at ~

0

(7.24)

(7.25)

The coordinates (h) defined in this way are called by VEBLEN the
normal coordinatesinAn with respect to ~" and to the coordinate system (")
0
They are a generalization of the normal coordinates introduced in a
Vn by RIEMANN 1 ). If we use normal coordinates, not only the I';~ vanish
at ~" but so do the symmetrized derivatives of I';~ of every order also.
0
If the coordinates ~" are transformed into ~"' we get the new normal
coordinates ~~~ belanging to the same point ~
0

(7.26) ~~~ det <5~ f' z = <5~ A:' (!) f z =Ai' ~h

1) RIEMANN 1854, 1, p. 279
 7. Geodesics and normal coordinatcs. 159

where the AZ' are constants. Hence, if the ~" are transformed, the
normal coordinates with respect to ~" and (u) undergo a linear homo-
o
geneous transformation with constant coefficients and these coefficients
are the values of A;:' at ~".
0
By means of the normal coordinates ~h the points of an 91 ( ~") are
0
mapped on the points of the tangent E,. of ~"in such a way that the point
0
of 91 ( ~") that has normal coordinates ~h is mapped on the point of E,.
0
with rectilinear coordinates ~h with respect to the local coordinate
system of E,. belanging to the coordinates ~".
From (7.21 a) we see that the basis vectors of (u) and of (h) at ~"
0
coincide. But that implies that the components of a quantity (not of
every geometric object, consider for instance r;:;.) at ~" with respect
0
to (u) and to (h) are numerically equal. This proves the following pro-
position:
lf a set of numbers be given at ~ with respect to (h) and if we know
0
that they transform like the components of a quantity if (h) is transformed
into (h'), these numbers are the components of a quantity at ~" with respect
0
to (u) and (u') and numerically equal to the components of this quantity
with respect to (h) and (h') respectively.
This proposition has been used by VEBLEN and T. Y. THOMAs 1 ) for
a very elegant construction of a set of quantities called the normal
tensors. The (not symmetrized) derivatives of order p of the F;~ with
respect to the ~h at ~"
0
(7.27) N;~ ... ;: /! h ( ~)
0
def (0
ofp ... 8;, If~ ~)

transform obviously like the components of a tensor with valences

1, p + 2 if (h) is transformed into (h'). Hence the N 1p f,f> are components
of a tensor. These tensors are called the normal tensors belanging to
the connexion r;:;. and the point ~". They occur in the series
0

(7.28) r;~c~) =~kNi/ih(~)

0
+ 21., ~keN;,;J;h+: w=o)
0
2 ).

They are all symmetric in the first p and in the last 2 lower indices
and vanish identically by mixing over all lower indices according to
(7.25). The normal tensors belanging to the connexion Fp";. can be con-
structed at every point and thus form tensor fields. But if we do this
in the way described, at every point we get the components of these
fields with respect to a normal coordinate system belanging to this
1) VEBLEN and T. Y. THOMAS 1923, 1.
2 ) The convergence of this series was discussed by T. Y. THOMAS 1934, 2, p. 126.
160 III. Linear connexions.

point and we should like to have the components with respect to a

general coordinate system (x). These can of course be computed by
repeated differentiation of [cf. (2. 5)]
(7.29) Tfh _
.l ji -
A~"urx
j i" p -
A'";."'
j i ul"
Ah;.
using (7.21). In this way we find for the first normal tensor at (~")
0
(7.30) 1\Th _
lVkji -
~vph ("'
Ukjix Uv
T"p - F."
vp-
2r.e T")
v(p )e

and accordingly for the general components

(7-31)
valid at all points of 91 ( ~"). The same could be done for all normal
0
tensors but as we shall see later there is a much shorter way using the
curvature tensor and its covariant derivatives.
If we start from the If:. .
j.i (zero at ~") other sets of generalized
0
normal tensors can be formed since the ak. .. ok, If: ... j,i transform for
(h)--? (h') like the components of a tensor. All these tensor fields are
differential concornitants of the connexion F'"";. and there exists a great
number of relations between them and their covariant derivatives.
Much information can be found in the publications of VEBLEN and
T. Y. THoMAs 1).
The process applied here to the If~ etc. can also be applied to any
quantity. In fact if we take the (h)-components (/JL (L is a collecting
index)' the derivatives oj, (/JL' oj, Oj, (/JL at ~" transform for (h) --7 (h')
0
like the components of a quantity and this gives rise to the construction
of a series of quantities, called by VEBLEN the first, second, etc. extension
e e
of (/JL. We denote them by the operators 17'", 17'"''"', etc. The first
extension is of course identical with the first covariant derivative but
this is not true for the higher extensions. For instance for a vector field v"
we have at ~":
0

a) r:j.v"(~)=AtJ:~vh(~)=AtJ:oi
0
vh(~)=At;:~vh(~)=V'"v"(~).
0 0 0
J b)
0

l
(7 .32) r;.,'",v"(f) =At~t~i:fii:17i,vh(j) =At~;;,i:(oi,ai.vh+vioi,r;;;) (j)
i,>< (J7. . vh + v; N7211~
=Ai,l-l2ftlh . ~h) (~)
0
1z71
e
= V. v" (~) + v;. N'"''"'}.
""(~)
0
'"''"' 0
1) VEBLEN and T. Y. THOMAS 1924, 1; T. Y. THOMAS 1934,2. When dealing
with all these sets of quantities the way is open for "orgies of formalism" but it
must be said that these authors have restricted themselves in an exemplary way
to those relations that are really interesting.
 7. Geodesics and normal coordinates. 161

and because these results have the invariant form and ~" is arbitrary
0

(7.33)

valid at all points of 91 ( ~). Conversely all extensions can be expressed

0
in terms of normal tensors and the covariant derivatives of the same
and lower orders.
The most interesting relations are those between the siruplest normal
tensors defined by (7.27) and the curvature tensor R;~;." and its deri-
vatives. With respect to (h) we have at ~" according to (7.31)
0

(7.34)
hence everywhere
(7.3 5) IR .. " =
Vf" 27\T ..
lV[Vf"]"I

But, becauseofN;[;i/' = OandN(;;":( = 0, N;~,i''can besolved from (7.35):

(7.36)

If the higher covariant derivatives of R;~;." are written out we get

the general formula
(7.37)
where * stands for terms containing only the ~";. and their derivatives
up to the order p. Using the coordinate system (h) we get at ~"
0
(7.38) V,kp ... k, R:.h(~)
kJ 0 -2N
- .:h(~)
kp ... kl[kJ]' 0 +*
where * symbolizes terms containing only normal tensors with a valence
lower than p + 4. Note that for p = 0 and p = 1 theseadditional terms
vanish because lj1 ( ~) = 0. Going back to general coordinates we get
from (7.38) 0
,-------------------------,
(7.39) I Vf""-
V,Vp . V1 R. - 2NVp v...
1 [Vf"]'" +* I
where * symbolizes the same quantities as in (7.38), valid at all points
of 91 ( ~"). Using the symmetry of the normal tensor with valence p + 4
0
in the first p + 1 and in the last 2 lower indices and the fact that mixing
+
over all p 3 lower indices leads to zero, from (7. 38) we get an equation
of the form
(7.40)
Schauten, RicciCalculus, 2. Auf!. 11
162 III. Linear connexions.

where L () denotes a polynomial linear homogeneaus in the isomers

of Vvp ....,R;;i'' with constant scalar coefficients and where * stands for
an expression in terms of R;;i" and its covariant derivatives up to the
order p -1.
In fact the equation
(7.41) N
(vp ... v"-0
1 vp) -

contains three kinds of terms: 1. those with p, as the last two lower
indices; 2. those with one v-index among the last two lower indices and
3. those with two v-indices as the last two lower indices. Because of
the symmetry of Pf./ in the first p + 1 and in the last 2 lower indices
the place of the indices in each of these two sets of indices is not
important. Now according to (7.39) a term 0f the second kind can be
written as a sum of PJ?~ a term of the first kind and terms containing
only N's with lower valences. In the same way a term of the third
P+4
kind can be written as a sum of an isomer of R, a term of the second
kind and terms containing only N's with lower valences. If the term
of the second kind arising in this way is expressed in the way des-
cribed above we have finally in (7.41) only a sum of isomers of PJ?"
with constant coefficients, terms of the first kind, and terms con-
taining only N's with lower valences. If these N's are dealt with
in the same way, we get (7.40). It is possible to give some more in-
v
formation concerning the form of the equations (7.39, 40). Let R
V
and N symbolize the (v- 4)th covariant derivative of R;;;." and the
normal tensor of valence v respectively. Then it is clear that (7.39)
4 4 5 5
expresses R in terms of N, and R in terms of N, but that in the ex-
6 6
pression for R we have not only N but also a transvection of two fac-
4
tors N. The general scheme is as follows
4 4
R: N
5 5
R: N
6 6 4 4
R: N,NN
7 7 5 4
(7.42) R: N,NN
8 8 6 4 5 5 4 4 4
R: N, NN, NN, NNN
9 9 7 4 6 5 fj 4 4
R: N, NN, NN, NNN
etc.
 7. Geodesics and normal Coordinates. 163

There is a similar scheme belanging to (7.40) and derived from (7.42)

by interchanging the letters R and N.
V1 Vu V1 Vu
From these schemes we see 1 o that a term with N ... N(R ... R)
V V
occurs in R (N) if and only if the following conditions hold
(7.43) v=v1 ++vu-2(u-1);
2 that there are exactly u -1 transvections in this term, every N (R)
being transvected with every other N (R) over at most one index and
3o that every N (R) in a product is transvected at least once.
The components of the normal tensors at ~" with respect to (h)
0
are the derivatives of the Ji1 with respect to the ~h. Hence (7.40) gives
very valuable information about the If1 as functions of the ~h in 91( ~").
0
We consider the case when all covariant derivatives of R;~;." of odd
orders are zero at ~". Then according to (7.40, 43) all N's with an odd
0
valence vanish at ~" and it follows from (7.28) that all terms with an even
0
number of factors ~h vanish at that point and that consequently
(7.44) If1 W) = - If~ (- ~k).
Butthis means that the Anis invariant for the transformation ~h--+- ~h,
that is a reflection at the point ~- In fact a vector field with a field
0
value vh at ~h has a field value - vh at - ~h after the reflection. Now
a linear element d~h at ~h has for reflection the element -d~h at -e,
hence (7.44) has as a consequence that the reflection of bvh at ~h is a
vector with the same components but with opposite sign at - ~h. An
An with this property, is called symmetric with respect to f'. If an An
0
has this property all N's with an odd valence must vanish in (7.28)
and then it follows from (7.39, 43) that all covariant derivatives of R;~;."
of odd orders vanish at ~". Hence the vanishing at ~" of all derivatives
0 0
of R;~;." of odd orders is necessary and sufficient for the symmetry of the
An with respect to ~".
0
An An is called symmetric if it is symmetric with respect to all
points of an 91 ( ~"). It follows from the above that this occurs if and
0
only if the first and consequently all covariant derivatives of R;~;."
vanish in 91 ( ~) :
0
An An is symmetric if and only if R;;;_" is covariant constant. 1 )
1) (Cf. also VII 5.) CARTAN 1926, 1; 2; 1927, 1, p. 87; 1932, 1; LEVY 1926, 1; 2;
WHITEHEAD 1932, 1 ; WALKER 1944, 1 ; 2; RASCHEVSKI 19 50, 1 ; ROZENFEL 'n and
ABRAMOV 1950, 1; ScHIROKOV 1950, 1; BoREL and LICHNEROWICZ 1952, 1 (also
for Un)
11*
164 III. Linear connexions.

Any quantity whose components can be expressed in terms of the

F"";. and their derivatives up to a certain order by means of equations
that are invariant for coordinate transformations will be called a dif-
ferential concomitant of the connexion r;;..
For instance R; ~).", R";.,
V,.;., V,.Rp;. In general we only consider concomitants whose com-
ponents can be expressed algebraically in terms of the rl'";. and their
derivatives. Special cases are scalars and scalar densities (or Ll-densities),
often called absolute differential invariants and relative differential in-
variants of the connexion respectively. If we use first anormal coordinate
system (h) with respect to a point ~. wegetat this point equations con-
o
taining only the derivatives of the I';~ with respect to the ~h and these
derivatives are the components of the normal tensors. Since the ex-
pressions are invariant for coordinate transformations we can go back
now to a general coordinate system ("), and by using (7.40) we may
change these expressions into expressions in R;;;." and its covariant
derivatives. Hence we have proved the first reduction theorem for dif-
ferential concomitants (cf. I 3):
All differential concomitants of a symmetric connexion are ordinary
concomitants of R;;;." and its covariant derivatives. 1)
Any quantity whose components can be expressed in terms of the
their derivatives up to a certain order, the components of a set
r;,~.,
of quantities C/J, ... , (/J (indices suppressed) and their (ordinary) deriv-
1 M
atives up to a certain order by means of equations that are invariant
for coordinate transformations, are called differential concomitants of
the set C/J, ... , C/J and the connexion F"";.. As before in general we only
1 M
consider concomitants whose components can be expressed algebraically.
Absolute and relative invariants are defined as before. If we use anormal
coordinate system (h) for the point ~" first, we get at this point equations
0
containing only the derivatives of the I';~ with respect to the ~h, and
the (h)-components of C/J, . , C/J and their derivatives with respect to
1 M
the ~h. Now these latter derivatives are the components of the exten-
sions of C/J, ... , C/J and as we have seen, these can be expressed in terms
1 M
of the covariant derivatives of C/J, ... , (/J and the normal tensors. Since
1 M
the expressions are invariant for coordinate transformations we can
now go back to a general coordinate system (") and by means of (7.40)
we may change all normal tensors into expressions in R;~;." and its
covariant derivatives. This proves the second reduction theorem for
differential concomitants:
1) CHRISTOFFEL 1869, 1.
 7. Geodesics and normal coordinates. 165

All differential concomitants of a set of quantities f[J, ... , f[J (indices

1 M
suppressed) and the symmetric connexion r;:;. are ordinary concomitants
of f[J, .. , f[J, R;~;." and their covariant derivatives. 1 )
1 M

If the connexion r;:;. is not symmetric we have only to consider s~;." as

one of the set of given quantities. Then alldifferential concomitants of the
connexion r;:;. and the set $, ... , f[J are ordinary concomitants of f[J, ... , $,
0 1 M 1 M
S ~;."and R;~.i" (the curvature tensor for the connexion !(;_.,) and their
covariant derivatives with respect to the connexion I(;;.). But they
are of coursealso ordinary concomitants of f[J, .. , f[J, S~;." and R;~;_"
1 M
(the curvature tensor for the connexion r;_.)
and their covariant deriva-
tives with respect to the connexion r;:;.. 1)

Exercises.
III 7,1 2). If an Xm in Anis formed by oom- 1 geodesics through ~"
0
tangent to an Em in the tangent space of that point, its equations in
the normal Coordinates belanging to ~ are linear. (Note that an xm
0
in our sense can not be singular at ~", cf. Ch. II, footnote 1 on p. 74.)
0

III 7,2) 2 If the ~" are normal coordinates with respect to ~", prove
that 0

III 7,2 oc)

is true at all points of 91{~") (use 7.25).

0
III 7,3 3). If the ~h arenormal Coordinates in a vn with respect to ;",
prove that 0

(g.,. g} + ~;.,. {;Hk ~i + .. ) dr d~"

III 7,3 oc) 1
ds2 -*-
=
0
(g.,.{~}
0
0
+ iKkt;d0~Hk ~i + ) d~ d~".
III 7,4. Prove the second identity and the identity of BrANCHI
using normal coordinates.
1 ) Rrccr and LEvrCrvrTA 1901, 1; WEITZENBCK 1923, 1; T. Y. THOMAS 1926, 2.
Cf. T. Y. THOMAs and MrcHAL 1927, 2; T. Y. THOMAS 1929, 1; 1934, 2, Ch. V.
Cf. for the projective and the conformal case T. Y. THOMAS 1927, 3; LEVINE 1942, 1
and for a more general case NoETHER 1918, 1.
2 ) EI 1935. 1, p. 102.

3 ) EI 1935, 1, p.104, 138.

 166 III. Linear connexions.

III 7,5. Prove that

III 7,5 1X) N;,;;;.>< =% V(w R;l(;.il ><- gV,~-' R,il(;,,:l ><
using normal coordinates.
III 7,6 1 ). Prove that in normal coordinates with respect to ~"
0
(Jlk Vj P~;) 0 -* (f\ 8iP\ + N,;i,;. h P":;- N,;i;m P~ m)o

8. FERMI coordinates 2).

If a curve is given in An, there exist coordinate systems (h) such
that the lj~ vanish at every point of the curve. FERMI proved this for
a vn 3) and EISENHART gave the generalization for An. 4) Let
(8.1) ~>< = !" (t);
be the equations of a curve and let
(8.2)
be a set of oon-z vectors at ~"in an En_ 1 that does not contain the tangent
0
of the curve, such that there is one vector in every direction of En_ 1 .
If now the vectors t" are displaced parallel .along the curve we get a set
0
of oon-z vector fields
(8.3)
whose covariant differentials vanish for displacements along the curve:
(8.4)
We only consider points of the curve where the t>< span an En_ 1 that
does not contain the tangent direction. We now take all geodesics
that are tangent to one of the vectors t" at an arbitrary point ~" of the
1
curve in 91 ( ~><) , and on each of these geodesics we choose an affine para-
o
meterz suchthat d~"jdz=t" at ~><. Then according to (7.16)
1

(8.5) ~" = 1~>< + z t"- _!___ F";. {~} z2 t~-' t'"- _!___ T" ;. {~} z3 t~-' t'" t;.-
2! 1-' 1 3! Pfl 1
expressing ~" in terms of ~" and zt". Hence, taking the general point
1
/" (t) of the curve for ~", we get the expansion in 91 ( ~")
1 1
(8.6) ~" = f" (t) + z t"- :, I;,";. {I" (t)} z2 t~-' t;.- ...
1 ) E I 1935, 1, p. 105 also for other examples.
2 ) FERMI 1922, 1; EISENHART 1927, 1, p. 64; BORTOLOTTI 1929, 2; 1930, 4;
1934, 1; DIENES 1933, 1; 2; E I 1935, 1, P 106.
3) FERMI 1922, 1.
4) EISENHART 1927, 1, p.64.
 8. FERMI coordinates. 167

expressing f' in terms of the n + 1 variables t, zt" and valid in a neigh-

bourhood of the curve. If now a set of n linearly independent vectors
~"; i = 1, ... , n is introduced at ~",such that %", ... , g.. span the En-I of
0 2 n
the t", we have at ~"
0 0
t" * t" e ";
0
(8.7) 0 ...
x = 2, ... , n

where the t... are functions of the parameters ot1 , ... , ot,._ 2 If the vec-
tors $"
are displaced parallel along the curve we get at each point of

the curve a set e" such that

(8.8) t" * t... e"; x = 2, ... , n
...
where the t... are constant along the curve. The ~" are known as func-
' get

l
tions of t. Introducing the values (8.8) in (8.6) we

~" * f"(t) + ...e" z t... _ _!___ e"' e;. F";. {t} z2 t... tY-
2! y ... "'
(8.9)
_l___e"'e"'e;.F"
3' /12/-'t
{t}z 3t... tY"'-
b '
x ' y ' z-2
- ' ,
n
. z y "

Now we introduce the new coordinate system (h) of An by means of

the equations

(8.10) { ~~ * wi (t); w1 (0) * 0;

~" * w ... (t) + z t"; w"(O) * 0; x=2, ... ,n
and try to determine the as yet unknown functions wh (t) ; h = 1, ... , n
in such a way that the contravariant basis vectors of (h) at all points
of the curve are identical with the ~" already defined. The equations
of the curve are

(8.11)

in these new coordinates and the ~"

and r;;. along the curve may now
be considered as functions of ~1 . It follows from (8.9) and (8.10) that

J~" *I" (OW)) + e"{~I} W'-w"(O(~l)))-

(8.12) 1 -:, :"'{~~};W}F;;.W}W-w"(O(~l)))W-wY(O(~l))) - ..

and, by differentiation with respect to ~h for all points of the curve,
that
(8.13) ,. *
81 ~ =df"
- (dw
~d
1)-1 -e,.dw" (dw1
----- )-1, By~"-*-
-e,,., _ ... ,n.
x,y-2,
dt t " dt dt y
168 III. Linear connexions.

Hence the e" are, in fact, basis vectors of (h), and in order to make e"
y 1
also such a basis vector, w 1 (t) and w"' (t) must be chosen such that they
satisfy the equations
(8.14) -df" - e"dw"' -~-~~-
* e" -dw- .
=
1

dt "' dt 1 dt

Now according to our assumptions df"Jdt and e" are linearly independent
1
vectors and they are known as functions of t along the curve. That
means that there exists an equation of the form

(8.15) -
df" = a.e"
dt
1
1
+ a. e";
"'
"'

valid at all points of the curve and that J, ~ are known as functions
of t. But then it follows from (8.14, 15) that

(8.16)

from which the functions w 1 (t) and w"' (t) can be found by means of
quadratures. If the curve happens to be a geodesie and if for t we
choose an affine parameter, e" can be chosen such that J = 1, ~ = 0.
e * t, e"' -"'- zt"'.
1
Then we get as a Solution 1

In order to prove that the .m

vanish at all points of the curve we
remark that (8.12) holds for every choice of (x), hence it holds if we
take (h) for (x). From (8.14) we then get fh * wh at all points of the
curve, which is in accordance with (8.10). (8.12) changes into

(8.17) {
eh * wh (O(e!)) +eh
"
w- w"' {O(e 1) ) ) -

- ;, ;i{lf~W}(e"-w"(O(e1)))(~Y-wY(O(e1))) -

and from this equation we get by differentiation

(8.18) {
r5~ * oy eh~ :h- r;~ (e"~ w" (O(e~))) -
x, y - 2, ... , n, h - 1, ... , n,
hence Fyh,. * 0 at all points of the curve. Moreover we know that the
fields tj' are covariant constant along the curve. Hence we have on

the curve [cf. (8.15)]
17 h * 1 T,h
(8.19) 0 = dji
dt
",. e -- a. J '. + a.y r.h
1i - Y'
.
or
(8.20) a.1 r,h
u+a:y T,h *0
1y=; I T'h
IX.llx=-
* O' x, y = 2, ... ,n;
 9. Linear connexions expressed in anholonomic Coordinates. 169

and because of &=l= 0 this also proves that FA

=I;,hy * 0 and I;,\ * 0
at all points of the curve. The coordinates (h) defined in this way are
called FERMI coordinates for the curve considered. By repeated dif-
ferentiation it can be proved that [cf. (7.15)]

(8.21) {;! . .
(Zp+l
;~ *cO; ~~,z1 ,
Zp .. ,z1 y ) -
... ,zp+I=2, ... ,n; h=1, ... ,n

at all points of the curve.

If we know of an xm that there exist n linearly independent vector
fields e", allround covariant stationary in An 1) over Xm, it can easily be

proved that there exists a coordinate system (h) suchthat the Ij1 vanish
at all points of Xm. 2) In the case of an X 1 we have proved more, the
fields e" being only covariant constant over X 1 . It is not yet known
whether' a system (h) with Ij~ vanishing over Xm can be constructed
if only covariant constant fields e" are supposed to exist over Xm.

9. Linear connexions expressed in anholonomic coordinates

(cf. II 9).
If we write for the covariant derivative of a vector with respect
to an anholonomic coordinate system (h)

a)
(9.1) {
b)

and if (x) is a holonomic system, we get for the If~ the transformation
[cf. (2.5)]

(9.2) If1 = Af[= r;;.- Aff o~< A1,

hence (cf. II 9)

(9-3)

which implies that I[7iJ need not be zero if the connexion is symmetric.
From (9.2) follows
(9.4)

The equations (9.2) and (9.4) also hold if (h) and (x) are both an-
holonomic.
1) Cf. III 2.
2) EI 1935. 1, p. 106. Read on that page An instead of Lw
170 III. Linear connexions.

From (9.3) we get the general formulae for Rot and Div with J7 in
anholonomic components (cf. 2.17, 2.30)

(9.5)

= . ttJih, ... hp _ 2 s~~ittJiho ... hp + (p -1) S!~ [II, roliilha ... hp] +
1
p:.ttJih, ... hp
1 1 1 ,.

(9.6) + 2Q~.1 ttJih, ... hp- (p - 1) {J[~ roliil h, ... hp]o

1' '
ro 11 ... hp = p-vector density of weight + 1.
From these equations another very simple proof of the equations in
Exerc. II 9,1 can be derived.
If we try to express a linear connexion in terms of s;; 11 , gih and
lijg; 11 the only difference from III 3 is that (9.3) occurs instead of
(2.13). Hence, if we write
(9.7) [Jfih def[Jtgkh 1)

we get
.--------------------------------------------.
(9.8) Ij~ = ghk (X{iki}- s{fki} + [J{iki} + ~ Q{fki}); Xiki def ~Bi gki I
instead of (3.9).
This equation takes a very simple form in an ordinary V,. if we intro-
duce an anholonomic system consisting of mutually perpendicular unit-
vectors i",

" at every point.
iJ.. Then the components of gih are all1 ot 0
and s~~~~
1'
and Q~~~~
1
vanish ' hence 2)
(9.9)
and
(9.10)

In this case the Ij1 have a simple geometric meaning. If we write

(9.11)
it follows from lij g; 11 = 0 that
(9.12)

Now the covariant differential of 1,11 is for gJ..,. positive definite

(9.13)

1) Note that the Qiih are not cornponents of a geornetric object.

2) VRANCEANU 1932, 1.
 9. Linear connexions expressed in anholonomic coordinates. 171

and this means that for every fixed value of i the vector ih undergoes
an infinitesimal rotation with respect to a parallel moving' eoordinate
system, with Ijkh(d~)i as "bivector" of rotation (cf. I 9). As such the
If;h were diseovered by Rrccr 1} long before anholonomic coordinates
came into use. He wrote
(9.14}
and ealled the Yihi the coefficients of rotation of the orthogonal net con-
sisting of the n eongruenees of streamlines of the n fields f'.
The Yihi
alternate in the first two indices. 2)
In an Ln facts are quite analogous. Having introduced the anholono-
mie system (h) the covariant differential of ((h is
'
(9.15}
Hence,
. .
for every fixed value of i the vector ((h undergoes a homogeneaus
affine transformation with respect to a parallel moving coordinate
system, with I'/k (d~ji as "tensor" of the transformation. 3)
The rotation coefficients If;h eontain a lot of information about
the orthogonal net of (h). A eongruence i" is geodesie (cf. III 7, V 5) if
and only if ill-J7P. ih = 0. Henee the n. a. s. eonditions for the congruenee
f to be geodesie are
(9.16}
A eongruenee is Vn_ 1-normal if and only if (cf. II 5.16, V 5) i[ rj, i1= 0,
henee the n. a. s. eonditions for the congruence i" to be vn-1-normal
n
are
(9.17} I[cb] n _!_ 0; b,c=1, ... ,n-1.
Taking (9.16) and (9.17) tagether we get for a geodesie and Vn_ 1-normal
eongruenee i" the n. a. s. eonditions
n

(9.18} I[ji]n * 0.
All eongruences i" aregeodesie and Vn_ 1-normal if and only if I[;;Jh -* 0.
' are alternating in ih this is only possible if If;h * 0,
But since the Ifih
1 ) Rrccr 1895. 1, p. 303. Cf. ScHOUTEN-STRUIK 1919, 1; LEVY 1925, 1; HLA-
VATY 1928, 1; VRANCEANU 1929, 3; MATTIOLI 1930, 1; 2; DIENES 1933, 1; 2; E I 193 5,
1, p. 8 5; E II 1938, 2, p. 34. The coefficients of rotation are intimately connected
with the theory of the "triedre mobile" of DARBoux, cf. Rrccr and LEVI CrvtTA
1901, 1, p. 157.
2 ) Of course, if we have introduced the anholonomic system (h) and thus the
Ij1 and If;h it is incorrect from a methodical point of view to introduce besides
these the rotation 'Yihi also, as many authors still do.
3 ) Cf. SLEBODZINSKI 1930, 1. In this case also it would be incorrect to introduce

some kind of "generalized coefficients of rotation" besides the Ij~.

172 III. Linear connexions.

that is if the .m
vanish at all points. But this implies according to
(9.3) that (h) is holonomic and that consequently Ki.i;'h vanishes. Hence
an orthogonal net of geodesie and Vn_ 1-normal congruences is only possible
in an Rn.
In order to find the expression for the anholonomic components
Ri.i;'h of the curvature tensor we start from (9.1b) and use (9.3) and
(II 9.4). Then we find
(9.19) r[k ril w; = - {a[k r;~i +T[i:ti Ifl; + DL 1m wh- s"ih rh w;.
Comparing this result with (4.9a) we get
(9.20)

instead of (4.2) and, instead of (4.15) we get

(9.21)

The equations (9.20, 21) could also be derived from (4.2), (4.15) and (9.2).

10. CARTAN's symbolical method used for connexions.

Using the symbolic notation explained in II 12 we write 1 )

Ah for A1 and also for A1de

,Qh f ,Qh _ ,Qh Ai A; and also for ,Q" d ~I" d ~).
or "''- ii "" " p, 1 2

(10.1) Jih for T,.;h -- Ij;h A""i and also for Th.d ~""
"''
Sh for 5~;., h and also for s .h 1d~"' 2a~
p,

l R~h for
'
R~h
p,
and also for R~h d ~"" d ~'-.
p,J. 1 2

In II 12 we had already for "' holonomic (cf. II 12.9):

a)
(10.2) {
b) c;
l)C ARTAN wntes
. Ahji and A;kj
h or Rhiki f or 5 ii h an dRhb
kji ut w h, w;,
hnho.: an dnh
..:;
for what we have written A h' r;h, Sh and Rj h. This gives rise to a duplication of
formulae that could have been avoided. The same holds for all publications where
the abridged calculus is used, except those dealing with alternating forms only.
T. Y. THOMAS gave 1927, 1 a severe criticism and used such expressions as "su-
perabundance of geometrical terminology" and "a lack of content ". This is
certainly too severe. The first objection can be countered by the change of nota-
tion we have proposed here, that makes duplication of formulae unnecessary
and the second one is only true where the abridged calculus is used at places where
it is not appropriate to do so.
 10. CARTAN's syrnbolical rnethod used for connexions. 173

Now we know from (9-3) that

a) nh _
,:,~!L - -
Aii
.
(T'h
p 1 [ii] -
S"h)
ji
_
- -
T'h kl Ak]
1 [!l
+ s"
p '
{
(10.3)
b) C Qh = [A'l?J + S".
Hence, from {10.2) and {10.3)
a)
(10.4) {
b) c
Moreover we know from (9.20) that

{
a) O[v~~i + l[~'ll.r;lli = . R;p_;h
(10 5) b) C [dfih] +[I? nJ = tR;".
CARTAN calls (10.4 b) and (10.5 b) the structural equations (equations de
structure) of the L",. According to the definition, repeated applica-
tion of o[v ...J or [d .. .] leads to zero, hence from (10.4), giving the cal-
culation in full for the symbolical method

vp]i h
- - 21 A;[w R..
0- + S ; Fh
+o [w
h
S vp] [wv p]i

0= [[dAi]J;h]- [A;[d.T;"J] + [dS"J

= [A~_n I;"]+ [S; Jih] +[Ai~" nJ-! [Ai Rih] + [d S"J
=-HAi Rih] + [dSh] +[Si J;h],
or
(second identity,
a)
(10. 7) cf. 5.2, 41 and
b) c Exerc. III 5,7).

In the same way, we get from {10.5)

0-
- 1-g "
U[w
Rh
vp]i -21
1.T'l[wlil Rh+tT'h
vp]l 21 [will
Rl
vp]i

C 0 = t[dRi "] - [[dJJh]I;IJ + [Ijh [dli1J]

t [dR;h] + [Fk" F/ nJ - t cn Rih]-
(10.8)
=
- [Tz" F{Jik] + t [ljh R;'l]
or
2
(identity of BIANCHI,
a) 17rw R;p_ 1;" = 0
(10.9) 2 cf. 5.19, 42 and
b) c [17 R;h] = 0 1) Exerc. III 5,7).

1) CARTAN calls (10.7 b) and ( 10.9 b) "theorerne de conservation de Ia courhure

et de Ia torsion ".
174 III. Linear connexions.

p
In these formulae we have used the operator [17 ... J corresponding to
p
the operator 17[w ... J defined in III 5. The working of the first operator
on any quantity Pi, ... ip~~ ~i, ... i, or its corresponding alternating differen-
tial form (cf. II 12) is defined for p > 0 by the equation (cf. 5.32)
p
+ [T,h,plh, ,, ...,] +
{10.10) { [
17 ph,. ... .h..,, ... ,,]-
. _ dPh, ... h.. .
[ . . " ___ ,,] l
... h.. .

- [I'}t.1 ph,

... hql.t2 Jr. J- .. .
In order to make comparison between tensor calculus and CARTAN's
notation possible in a more complicated case we now translate the proof
of the geometrical meaning of the two identities (10.7, 9) given
in III 5. Writing d~x for ~x-~x, d~[d~Pl for df"", d~[w d~ d~Pl for
0 1 2 1 2 3
dr"" and passing to general anholonomic coordinates we get for the
change of the radiusvector vh from ~x to ~x
0 0

(10.11)
{5.43)

and for the change after moving araund d r"

(10.12)
(5.44)

After parallel displacement we get

{10.13) Jc
{5.45) 1
and after applying the theorem of STOKES

The abridged calculus has often been used by CARTAN to solve pro"
blems concerning point transformations transforming fields into each
other. We consider first the question whether n given linearly inde-
pendent vector fields ~ in an xn with Coordinates ~X can be trans-
formed by a transformation ~x -+ ~"' into n given linearly independent
 10. CARTAN's symbolical method used for connexiozas. 17 5

vector fields 1 -!;: in another Xn with COordinates ~"'

h d [:A - h d [:A' 1)
J a) w" "- w". " I

(10.15)
1 b) c h lh
W= W.

By differentiation we get
8 h i

(
Cl) h
[p W] = i .
Uji Wfp W,<],
(10.16 a)
h .h I j I i
) 8[p'
I
w.<'J =
I
uii w[p' W..t'];

( J;
h h i i
Cl) w] = i [w
(10.16 b) c [d
h
u; w
. .
) [ d 'w J= 1
Ujih ['zb '~];
where u;;h and 'u;;h are known functions of ~" and ~" respectively.
From (10.15) and (10.16) there follows as a first condition the set of
non-differential equations between ~" and ~"
(10.17)
If these equations are inconsistent, no transformation ~"-+ ~" trans-
forming the fields i into the fields 1 -/ is possible. If they are consistent
we get by differentiation (defining the operator 8i in both Xn's by
81' = 41' 8i; 81'. = '41'. 8i):

(
h k h.
Cl) 81' Uji = Wl' 8k Uji ,
(10.18 a)
) 8p,' ~u~~h
1~
= ,-/1;t/ 8k ~u~~h.
J~ '

Cl) du~~h = J 8k u~~h.

(10.18 b) c ( 7~

) d 1 U~~h='/];8
1' k
u~~h.
1' '
1 ~ '

and from this we get as a second set of non-differential equations bet-

ween ~" and ~"
{10.19) " ' u h - " ' luh
uk ji - uk ji

In the same way we can derive

(10.20)
and so on till we get inconsistent equations or a set depending on the
former sets. In this latter case all following sets depend on these former
sets and we have got a finite number of equations from which the ~"'
can be solved as functions of the ~" and a finite number of parameters.
These functions give the transformation ~"-+e required.
1) CARTAN 1946, 1, p. 315ff.
17G III. Linear conncxions.

The re,;ult just obtained can be applied to the question of whether

there exists a transformation ;"---o.;"' mapping an L" with Coordinates;"
and a connexion r;;.on another L.,, called hereafter 'Ln with coordinates
;"' and a connexion 'r;;;. .1) In Ln we take an arbitrary anholonomic
coordinate system (h) and in 'Ln an arbitrary anholonomic coordinate
system (h). Then at each point of Ln the A1 depend on r (= n 2) para-
meters r/", et. = i, ... , i and at each point of 'Ln the A1 depend on r para-
meters 'YJ"', et.' = i', ... , i'. 2 ) Then we have to investigate whether it is
possible to choose the systems (h) in Ln and 'Ln such that lf~ = 'lf~ at
corresponding points. We remark that, though r;;..
S~~" and R;;;."
are independent of the rJ", the components Jft Si;h and Ri.iih also
depend on these parameters, and that the same holds mutatis mutandis
in 'Ln. In Ln the equations (10.4) and (10.5) are valid but A1, Fl;,
S~~h and R;~;h now depend not only on the ;" but also on the 'Y/" That
means that we are now working in the Xn+r of the variables;", 'YJ" from
which the Ln can be formed by reduction (cf. II 4). If a field v" or
W;. in Ln depends on the parameters 'YJ", it may be considered as the
reduction of a field v", v" = undetermined or as the section of a field
W;., Wp = 0 in the xn+r Hence the covariant multivectors A1, S~/ and
R;~ih may be considered as the sections of the multivectors A~, S(;;/
andRvcih; A, B, C, D=l, ... ,n, i, ... , r satisfying the conditions

(10.21) {
Ah-0.
p- '
s;~h = - s;.;h = o; s;;/ = o; R.;;;h = o;
Ct., , y, <5 = i, ... ,r.
In the Xn+r of ;", 'YJ" the allowable coordinate transformations are
;"' = /"' (;") ; 'YJ"' = rp"' ('YJ"), hence
(10.22) A/{ = 0; A= 0; Af = 0; A~. = 0.
Accordingly the components of some connexion FcAB in Xn+r transform
as follows

(10.23)

1 } CARTAN has dealt with this problern for the case of two Vn's.
2} In the case of two Vn's considered by CARTAN the local coordinate system~

can be taken orthogonal and then r = ( ~).

 10. CARTAN's symbolical method used for connexions. 177

We see from (10.23) that the F,.J. transform just like the parameters
of a connexion in Ln and that of the other seven sets of indices six
undergo a linear homogeneoUS transformation. Hence in xn+r a connex-
ion can be partially fixed by taking the r;;. equal to the I',..";. of the
connexion in Ln and
(10.24) r,.a.;. = o; ry";. = o; J;. = o; I'y = o; r,.rp = o; I'y"J. = o
and leaving the l'y undetermined. This agrees with the fact that
there is no connexion given in the space of the parameters rt.
Passing now to the anholonomic system (h, 0t) in Xn+r (the rt remain

l
unchanged) we get
a) Jj~ = Aff: T,.J.- Aff B,.A1;
(10.25) b) r;. = - Af ByAt
c) Ij~=O; d) r:p=O; e) lff=O; f) I'y~=O; g) !J=O
and this means that the lj~ transform in the ordinary way, that the
only new components are the I'y~, and that the I'y remain undetermined.
Using (10.25) we get besides (10.4)
a) BryA~1 =AlyF~i+S,;..i\
(10.26) {
b) o= BryA31 = AlyFtJi + s,;;h;
or, taking (10.4) and (10.26) tagether

a) = Afcr;Ji + S(;j/;
BrcA~1 B, C = 1, ... , n, i, ... , r;
(10.27) { .
b) C [dAh] = [A' Jih] + sn,
in accordance with CARTAN's general principle that an abridged equa-
tion remains in some way valid if the nurober of independentvariables
is extended by the introduction of parameters. According to this
principle we should now have also next to (10.5)
a) BrDFtli +Fr~lllrili = tRvcih;
(10.28) {
b) c [dJ:"J+c.n"nJ=tR;h.
In fact the F;';. are independent of the parameters, hence according
to (10.25) after some calculation
(10.29) o= a[d r;J;. = 2Ai A;: (a[d r,..~i +Fr~ 1z1F,l1i)
and this proves the validity of (10.28a) for D=lJ, C=f-l, R;;,;h=O.
From (10.25b) we gct
(10.30)
Schauten, Ricci-Calculus, 2. Aufl. 12
178 JII. Linear connexions.

Hence (10.28a) also holds for D=IJ, C=y, RJ;;"=O. Since the equa-
tions (10.4 b) and (10.5 b) are now valid in the more general sense of
{10.27b) and {10.28b) it stands to reason that the same must hold
for the second identity (10.7b) and BIANCHI's identity (10.9b), but
a direct proof could be given.
In Xn+r we now have the n+r vectors (or Pfaffians) A", F;," and
their rotations given by (10.27, 28) and in the 'Xn+r we have in
the same way vectors (or Pfaffians) 'A", '!';," whose rotations satisfy
equations of the same form. The problern is how to find a transfor-
mation ~",rt"-?-~"',r/': suchthat A"='A 11 ,1';,11 ='F;,". If we write (10.27)
and (10.28) in the form

{
a) u[C B]i + s"Ai
"' A"B] = Ai[C.l.,..,,. ii [C AiB]
(10.31)
b) C [dA"]= [Ai!';,"]+ 5;"(11 [Ai Ai];

{
a) o[D T."C]i = - F."
[Dflf r.zC]i + "2"tRkii" Ak
[D AiC]
{10.32)
b) C [dl';,"J =- [_n" F;,Z] + jR,;;;" [Ak Ai]
we have got equations of the form (10.16ot) and equations of the same
form, corresponding to (10.16) can be derived for 'Xn+r Comparison
with (10.16) shows that many of the coefficients corresponding to the
u;"t" and the 'u/t" vanish or aresimple constants and that the remaining
ones are the anholonomic components of S, R, 'S and 'R.
Now the holonomic components of any tensor of Xn (' Xn) are in-
dependent of the rt" ('YJa.'). Hence, for instance for a tensor Pj." we have
dP" = P"+ d'YIY(o A~)A"" P"+ d.,Y A") P"
ra) dt:~'o A~(o

j
< '> fJ 'I y A 'I y " A
= dr::~'
'>
AiI' V..1 P:"

+ dr::~'
'>
F 1.P;"-
I'' I
dr::~'
'>
F" I' l
P: 1 +
(10.33) + d'YlYF 1P"-d'YlYF" P: 1
'I Y l 'I yI

b) c d P:"
1- =Ai 17.
... , P:" ~ + ['} Pz"- T,z" p:II)
z . 1-

Applying this formula to S;"i'", R;.ji 11 and their covariant derivatives in

Ln we get
(10.34)

1) :g, CARTAN 1946, 1,p. 320 introduced this formula for Kkiih Jij Kkiih in Vn as
"evident". This optimism we don't share but it has tobe granted that the general
formula (10.33) after having been proved rigorously gives excellent means to use
the abridged calculus for the calculations.
 11. Linear connexions depending on a non-symmetric fundamental tensor. 179

Hence, dealing with the n + r = n + n 2 vector fields A~, Tt; in Xn+r in

the same way as with the n vector fields W,t in Xn and using (10.34-36)
we get the n. a. s. conditions for the mapping of the Ln and Ln on each
1

other in the form

s~~h = ~s~~h.
J a)
l b) vk sii.h
J ", 11 '
(10.37) = 'Vk si/h
I

J a)
(10.38)
l b)
in which all quantities on the left hand side are expressed as functions
of ~", r{' and all on the rigbt hand side as functions of ~", r;"'. Either
after a finite number of steps we get inconsistent equations or after a
certain numbers of steps all following steps lead to equations dependent
on the equations already derived. Only in the latter case there may
exist transformations ~", r;"-+~"', r;"' giving the desired mapping and
they can be found by solving ~", r;"', as functions of ~", r;" and some
parameters, from the equations (10.37, 38). 1)

11. Linear connexions

depending on a non-symmetric fundamental tensor 2).
In III 3 we have seen that a fundamental tensor tagether with S~;_"
determines uniquely a linear connexion. In order to obtain a unified
field theory EINSTEIN started 1945 3) from a non-symmetric fundamental
tensor. 4)
1) This example is very instructive, because it reveals at the same time the
strength and the weakness of the abridged calculus. The strength lies in the possi-
bility of reading one and the same formula in different aspects, the weakness
in the necessity to go back often to non abridged calculus in order to derive the
formulae indispensable for further progress or for the formulation of the results.
2 ) This section deals with a fundamental tensor whose symmetric part is not

zero. There is another line of thought starting from a fundamental bivector.

We mention here only LEE 1943, 1; 1945, 1; 1947, 1; CHERN and WANG 1947, 1;
VRANCEANU 1949, 1; GuGGENHEIMER 1951, 1 (relations with the V2ml
3 ) EINSTEIN 194 5, 1.
4) Cf. EINSTEIN and STRAUS 1946, 1; KOSAMBI 194 7, 1; 1949, 1; SCHRDINGER
1947, 1; 1948, 1; 2; 1950, 1; 1951, 2; EINSTEIN 1948, 1; 1950, 1; JORDAN 1948, 1;
2; 1950, 1; STRAUS 1949, 1; TONNELAT 1949, 1; 1950, 1; 2; 3; 4; 1951, 1; 2; 3; 4;
UDESCHINI 1950, 1; 1951, 1; 2; 3; lNGRAHAM 1950, 1; 1951, 1; 2; EISENHART 1951,
1; 1952, 1; HLAVATY 1952, 2; 3; 4; 6; 1953, 1;2; 3;4; 5; 1954, 1; LICHNEROWICZ
1953, 1. The idea of the deduction given here goes back to ScHRDINGER but its
very simple form is due to NIJENHUIS (personal communication).
12*
180 III. Linear connexions.

Let this tensor be denoted by sA," the isomer of its inverse 51 " 2
(cf. I 8) by 5" 2 , hence sJ.,. S"~-' = A:l, and let 5 stand for Det (s,1"). Then
we introduce the symmetric and alternating part of s2 "
(11.1) h>< def S
=
,
(")'
k.<>< def
= s[-<><l

and use h2 ", supposed to be of rank n, and its inverse h"J. for raising
and lowering of indices. Let f.J stand for Det (hJ."). Of course generally
s" 2 =f= S"J. and f.J =f= 5.
Let there also be a connexion I',."J. and let S~;." not necessarily be
zero. Then we establish relations between sJ." and I',."J. by means of
a variational equation. We require that the variation

(11.2) dj R,..<f!f
Tn
2 dt;l ... d~n;

should vanish if S>< and r,.".< are varied independently. Hence

J 15~-' 2 dR,. 2 dt;l ... d~n =

(11.3) ~ a) Tn
V
0;

l b) fR,. 2 de~-'"de ... d~n=o.

Tn

Now we have

because d !',."2 is a tensor. Hence

( 11. 5)

where

(11.6)

The terms in brackets in the right hand side of (11.6) do not represent
exactlythe covariant derivative 17" 15~-'J.but they differ from this derivative
only because of the fact that the term belanging to the first index of
15~-' 2 and the term owing its origin to the density-character of ~-'" have
r,<",. used instead of r,.".<. Slightly generalizing a notation of ErNSTEIN
we therefore write

(11. 7)
 11. Linear connexions depending on a non-symmetric fundamental tensor. 181

Making the usual condition that the Variations of the ~",. vanish
at the boundary, we get from (11.5)
(11.8)

but this implies that (1- n) @~~a = 0 and that (11.8) is consequently
equivalent to

(11.9)

If the ~"" are given there are n 3 algebraic equations for the T,..)..
But they cannot be independent because if r;,.
is a Solution J;.",. + t,.. A
0 0

is also a solution for every choice of t,... This is seen very easily if we
put (11.9) into the form
(11.10) 0>< f51J). = p~~~a rT(}a
where
(11.1 1) ppha~J
><o(} ---
~wA(-AapT
I<(JW
+ _2_AiJ[GT])
n- 1 + ~pw (-AATa +Au")
1< (}W (}><W (J><W

In fact P~~~a = 0 and this means that the T;-rank of Pis :;;;;;n3 - n.
But then the ~-~,."-rank is also :S. n 3 - n and this is confirmed by the
identity prt:J;a=O. This implies that (11.10) can only have solutions
rT(!G if
(11.12)

Note that this is an invariant equation because ~"" is a bivector

density of weight + 1.
0
If a solution ~,.,. is known the vector t,.. mentioned above can be
0
chosen in one and only one way so that for *T,..",. det~). + tPA~ the equa-
tion (cf. 2.15)

(11.13) I *S,.. - 2 *S"=

=n-1 ,..,. 0 I
holds. If this identityis introduced as a condition, the equations (11.10)
have one and only one solution *~";. if and only if the T;-rank of Pis
exactly n 3 - n. It is an algebraic problern to find the n. a. s. conditions
for this special value of the T;-rank. But there is a special case where
these conditions are not needed. Because of (11.13) we get for *17
instead of (11.9) the simpler equation

(11.14)
182 111. Linear connexions.

from which it is easily deduced that

(11.15) *17,.. 5 = 0
and consequently

(11.16) x = 0;
a) *17,.. 5-+ b) *Vp s~~ = 0 .I
From (11.16b) it follows that

(11.17) l a)
b)
o,..s,_x- *Tf,..sex- *F,..exSe = 0;
o;.s",..- *F~"se,..- *Ti,.. s"e = 0;
c) - o,.s,..;. + *f':xse" + r~;.s,..e = 0;

For sr;."J = 0 this gives immediately

(11.18) a) r,..x" =-- 5"e(o,..se" + o"se,..- oes,..") b) *5p,i" = 0.
But this proves that the T:-rank of P in (11.11) is exactly n 3 - n if
s;." is symmetric. Asthis rank can never be >n3 - n it must be exactly
n 3 - n for small values of sr" ,.1 . 1) Hence we see that the equations (11.13, 16)
determine the r;;.
uniquely if the fundamental tensor s;." differs only
little from a symmetric tensor.
If * r:;.
is known every other Solution of (11.9) satisfies the relations
(11.19)
because *5,.. = 0.
The equation (11.3 b) leads to

hence this part of the variation leads to

(11.21) G,..;.=
defR
,..;.-'1 -0 .
IR s,.."-

Here the general tensor G,.." plays the role of the symmetric tensor
G,.." defined in (5.29) and it is identical with this latter quantity if
sr,_"1=0 and 5,_=0. For n=j=2 (11.21) is equivalent to
(11.22)-
1) This remark was made by Mr. ScHRDINGER in a Ietter to Mr. NIJENHUIS
of 12th November 1951.
 11. Linear connexions depending on a non-symmetric fundamental tensor. 183

and because Rp;.=*R,..;.-28c,..5;.1, as follows from (11.19), also with

(11.23)

Now we have for the 4+16+60=80 unknowns 5;., S>< and *r:). the
4+60+16=80 equations (11.12), (11.16b) and (11.23). This nurober
is by four to much because the equations must be invariant for coor-
dinate transformal:ions and a coordinate transformation contains four
arbitrary functions. Hence there must be four relations between the
left hand sides of th~!?e equations and their derivatives, and these re-
lations must hold independent of the vanishing or not vanishing of
these left hand sides. They can be obtained by taking in (11.5) and
(11.20) only those Variations of r;A
and 5"). that arise from dragging
along these fields over some arbitrary v" d t. Because by dragging
along a Solution we get another solution, provided that v" and its deri-
vatives vanish at the boundary. According to II 10, III 5 we get

(11.25) _!__
dt
s,.;. = - fV 5"). = - v" 8p 5"}. + 5"). 8p v" + 5"" 8p V)..
From these variations the four equations can be derived in the ordinary
way. All divergences have to be reduced to zero by means of STOKEs'
theorem and therefore it is profitable to introduce *V instead of V
because *V,.. 5"" = 0 and *V,..l:J" = 8,..\J" for any vector density l:J" of
weight +1.
If the r;A
are solved from (11.16b) as functions of the sA,.
and their
derivatives these solutions can be substituted in (11.23). Then we get
the following 4 + 10 + 4 = 18 equations
(11.12)
(11.26) *R<,..A>=O
(11.27) 8[,. *R,..;.1 = 0

for the 16 unknowns sJ.,. There must be six relations. Two of them are
trivial
(11.28) 8;. 8u ~[pA] = 0

(11.29) 8[w 8v *R,..A] = 0

and the other four can be obtained in the way described above. In
the symmetric case only the 10 equations (11.26) remain for the 10 un-
v
knowns sJ.,. In this symmetric case R,..J. d~"A vanishes because the
184 111. Linear connexions.

r:A are already considered as functions of the sp.A From (11.20, 25)
we get now for SA = 0, neglecting terms that vanish after integration

(11.30) { p.A def'

R = 13~ Gp.A (51! 2 17 vP.
Q
+ SP.I! 17 V;') dt
Q

= - 2(V(! 5~ GxA se).) v" dt = - 2(V(! @;'x) v" dt

from which
(11.31) 17p. @P." = 0.

The condition (11.13) that ensures at least in the siruplest cases the
uniqueness of the connexion *FP.";., means geometrically that this connex-
ion behaves as a symmetric one as far as scalar densities are concerned.
But it has yet another meaning as was pointed out by EINSTEIN. If
s,,. is changed into s,.,, that is, if the alternating part of the fundamental
tensor gets another sign while the symmetric part is unchanged, *r:A
will change into *I';.~ as is seen quite easily from (11.17a). Now let
this change induce a change of *Rp.l into *R:,. Then we find

(11.32) { a
) *R (p.l)- *R'(p.)--
- 2 U(p. .. V + 2 *sVQ.. V *.lT!Q(p.A)'
" *s lvi)
b) *R [p.A] +*R'[!'A] -- - 2"'U[p. *S+2*S
lvl AJ V(!.. v*s
~ p.)... e'

hence, *S, = 0 is sufficient (but not necessary) for the tensor *Rp..< to
change into *R;.p. if s,~" is changed into s,.;,. In the original form of
the theory given by EINSTEIN in 1945, s(l><) was taken real and s[A><l
imaginary. Then s;.,. could be considered in a way as "hermitian"
and the condition that *R"" also should be "hermitian " could be ful-
filled by taking *SA = 0. This condition however presents itself in a
quite natural way if we wish that *R,~" and s," should differ only by
a scalar factor just as K,~,. and g,~" do in the ordinary theory of relativity.
The connexion derived in this way can be used for n = 4 to build
a unified field theory and the authors mentioned above have got some
remarkable results in this line. But it should be remarked that in the
end there is nothing but an X 4 with only two fields s(J") and sc;."1 and
that according to the second reduction theorem (III 7) the differential
concomitants of these fields are ordinary concomitants of s(A><), s[A><l,
the curvature tensor of the symmetric connexion belanging to s(A><) and
the covariant derivatives of these quantities with respect to this con-
nexion. So, from a mathematical point of view the new connexion
derived from (11.10) can never produce anything that could not be
expressed in terms of s(;."), s[A><l and the ordinary symmetric connexion
belanging to s(;."). It may be useful to introduce this new connexion
in order to get some heuristic principles that may suggest for instance
a choice of the Hamiltonian or a desirable form of the field equations.
But nothing really new can ever arise from following this course. This
 IV. LrE groups and linear connexions. - 1. Finite continuous groups. 185

remark is not valid if the introduction of a non-symmetric fundamental

tensor is combined with a change of dimension. In projective relativity
in five coordinates a non-symmetric fundamental tensor does not seem
to have been used till yet but in conformal relativity in six coordinates
INGRAHAM 1) has recently tried to generalize ScHRDINGER's ideas.
By transvecting (11 .17 a) wi tli S"" we get
(11-33)
hence
(11.34) "' *TA1-']-
U[9
-"'0 [9 *S!"].;.
and according to (4.15)
(11.35) - *RV!"
*V91"- . ;. -- 2"'0 [v *T!-']-
- 2"'0 [9 *S!"JA.;.
This proves that the condition (11.13} implies that *V.,.. vanishes, that
is, that for the connexion *F;;. a covariant constant scalar density
field is possible (cf. 111 6).
From (5 .6) we get according to the condition (11.13)
(11.36)
and it is easy to show that this is in accordance with (11.32b).

IV. LIE groups and linear connexions.

1. Finite continuous groups 2).
An r-parameter finite continuous group in the sense of LIE is a set
of elements between which a "multiplication" is defined and that
satisfies the following conditions:
a) The elements are in one to one correspondence with the points
of an IJl(r{") in an X, with the coordinates r{"; rx = i, ... , r.
b) If the element T'1 belongs to 'YJa. and Tc to r;:z, the product T8 = T'1 Tc
belongs to the set and the oa. are analytic functions 3) of the 'YJrr. and r;:-.
c) There is an element I called "unity" corresponding to rt such
that I T'1 = T'1 I = T'1 for every choice of T'1. 0

1) L. c. p. 179 footnote 4.
2) General references also for literature: LrE and ENGEL 1888, 1; BIANCHI
1918, 1; EISENHART 1933, 1; VRANCEANU 1947, 1. According to the general cha-
racter of this book we only consider the "group germ", that means we only are
interested in local properties. There is a vast Iiterature on properlies in the large.
We mention here only also for literature CARTAN 1927, 2; 1930, 2; 1936, 1; MAYER
and T. Y. THOMAS 1935, 1; CHEVALLEY 1946, 1.
3 ) Seefootnote 1 on page 186.
186 IV. LIE groups and linear connexions.

d) To every T'1 there belongs an element T"~- 1 such that T"~ T"~- 1 =
T"~- 1 T'1 =I. Its Coordinates are analytic functions 1 ) of the rj"".

e) (TrJ6)T0 =T'1(T6 T0).

The elements are often transformations in n variables but this is
not necessary. From the definition we see that only elements of the
group are considered in a neighbourhood of unity. This neighbourhood
is called the group germ (Gruppenkeim).
We identify each element with its corresponding point in X,. This
X, is called group space.
To every pair of elements 5, T there belang two elements T 5-1
and 5- 1 T. Two pairs 5,T and 51 ,~ are called
(+ )-equipollent if T 5_ 1 = 7;_ 5;- 1
(- )-equipollent if 5-1 T = 5! 1 T1 .
From this we get the following deductions:
1. If 5 1 , 7;_; 5 2 , T2 are ()-equipollent , three of these elements
determine the fourth uniquely.
2. If 5 1 ,~; 5 2 , T2 and 5 2 , T2 ; 5 3 , T3 are both ()-equipollent ,
the same holds for 5 1 , ~; 5 3 , T3 (Fig. 12).

3 If 5, T; 5 1 , T1 and T, U; 7;_, U1 are ()-equipollent , the same

holds for 5, U; 5 1 , U 1 (Fig. 13).
4. If 5 1 , T1 ; 5 2 , T2 are ()-equipollent , then 5 1 , 5 2 ; T1 , T2 are
(=f)-equipollent (Fig. 14).

fP
r,~;;

S,o~( Fig. 14. :.52

1) Though in the following all functions will be supposed to be analytic, many
theorems remain valid if only it is assumed that all functions are of class u.
 1. Finite continuous groups. 187

S. An ill(T1) can always be mapped continuously on an ill(T2) such

that T2 is the image of ~ and all pointpairs of irr(~) are mapped
()-equipollent on the pointpairs of ill(T2 ).
6. If irr(~) is mapped in this way on ill(T2 ) and ill(T2 ) in the same
way Oll m(Ta) the result is the same as if m(Tl) were mapped directly
on ill(T3).
If a pointpair T'YJ, T17 +d'YJ is given, there exists for every choice
0 0 0
ofT" in an irr (rt) a ()-equipollent pointpair T"~, T"+dry Hence, using for
0
instance the (+)-equipollence we get for every choice of an infinitesimal
vector vrx d t at rt a vector field va. d t in irr (rt). That means that the
0 0
(+)-equipollence fixes a parallel displacement (cf. III 2) for contra-
variant vectors in irr (rJ''). According to the conditions 1, 2, 3, 5 and 6
0
the parameters of this displacement establish a connexion that is linear
in the sense defined in III 2. Let erx, ~; a, b = 1, ... , r 1) be fields of
basis vectors covariant constant for this connexion or shortly (+ )-con-
b

+
stant. Let the covariant differentiation symbol of the connexion be V
+ +
and the parameters Ty, then we have Vy erx = 0, hence
b

(1.1)

Let this connexion be called the (+ )-connexion. The (- )-equipollence

leads in the same way to the (- )-connexion with the symbol V and
-
the parameters I'y

(1.2)

where erx, f; A, B = 1,

... , r are fields of basis vectors covariant con-

B
stant for this connexion or shortly (- )-constant. 2)
Vectors at different points that can be transformed into each other
by ()-equipollence are said to be ()-parallel.
As both connexions are integrable (cf. III 4 and deduction 2) their
curvature tensors vanish

(1.3) .. "'
R ~y def
~
2 0" [~ T."'
y]
+2 T'o<
.1. [~ il r.
y) =
0
1 ) Or 1, ... , r if the indices 1, 2, ... are already used for another purpose, cf.

for instance VII 4 and 5.

2) These connexions were introduced by CARTAN and ScHOUTEN 1926, 3; 4,
cf. CARTAN 1927, 1; ScHOUTEN 1927, 1; 1929, 4.
188 IV. LIE groups and linear connexions.

but they are in generalnot symmetrie (cf. III, p. 142, footnote 3)

(1.4)

If a eurve eontains a pointpair R, 5 and the first point of a (+)-equi-

pollent pointpair T, U, it may happen that it also eontains U. If this
is the ease for every ehoiee of R, 5 and T the eurve is said to be geodesic.
Now R, T and 5, U are (-)-equipollent pointpairs, henee, if we use
in the definition (- )-equipollenee instead of (+ )-equipollenee, we get
the same geodesics. If a geodesie eontains T and V T it also eontains
V2 T, V3 T, ete. Taking V infinitesimal different from I we see that a
geodesie as defined here is identical with a geodesie of the (+)-eonnex-
ion and that for every ehoice of V the points T, V T; V T, V 2 T ete.
have the same distanee if measured by means of an affine parameter
on the geodesie (cf. III 7). The same reasoning holds for the geodesies
of the (- )-eonnexion. Moreover, if V T = TW it follows that V2 T =
T W 2 ete. Henee the (+)- and (- )-geodesies are not only identieal
but the affine parameter on them is also the same for both eonnexions.
Using this parameter, the equation of the geodesie must have the form

(1. 5)

and from this it follows that

(1.6) r.y"' =def .1+(y) = .1- (y)

T'<X T'<X

are the parameters of a symmetrie eonnexion that ean be obtained by

symmetrizing the (+)-eonnexion and the (- )-eonnexion. This (0)-
connexion is in general not integrable (cf. III 4) and its eurvature
tensor
(1.7) "'
R 6y def
= y] + 2 .1T'<X
2"0 [6 F."' [6 :el r.
y]

plays an important role.

+ -
Besides the operators V, V and V we also need the LIE derivative
(cf. II 10) with respeet to the fields e"' and e"', for instanee
b B

~ 1.8) fb P"'

~.,{ beY 8"' P"'

- pY Oy e"'
b
+ P"'y o be'Y.
The LIE derivative f JY: with respect to the (+)-eonstant field v"' = va A~
V
with constant va is then given by

(1.9)
 1. Finite continuous groups. 189

If in deduction 4 for 5 1 , T1 ; 5 2 , T 2 we take elements that differ from

J only in infinitesimals as illustrated in Fig. 15, the quadrangle is built
up by means of the vector fields erx dt and erx dt'. This means that the
b B

e"dt
h

Fig.15.

field erx is absolutely invariant (cf. II 10) with respect to the field erx
b B
and vice versa for every choice of B and b:
(1.10) a) f erx = eY o erx - eY o erx = 0 b) fea.=O.
bB b Ys B yb ' Bb

From this it follows that

- + - +
(1.11) 0 = eY V.. erx - eY V. erx = f erx + (r.rx - r,rx .) eY Be;
b 'B B yb bB y l b
hence
+
(1.12) rrrx = Fr
The symmetric part of (1.12) leads back to (1.6) and the alternating
part is
+
(1.13) 5;t=- 5;t
hence, for every quantity (/) (indices suppressed)

(1.14)

It follows from (1.1, 2, 10, 12) that for any fields vrx, w

{1.15)

and this implies that for every quantity (/) (indices suppressed)
- +
(1.16) a) f (/) = er V. (/); b) f (/) = eY V. (/)
b b y B B y

expressing the fact that dragging along over a ()-constant infinitesimal

field is the same as a (=f)-displacement over the same field. Note that this
is just what we see in Fig. 15 for an infinitesimal contravariant vector.
190 IV. LIE groups and linear connexions.

Because the fields t are (+)-constant, we have

+ a a + a
(1.17) 0 = Vr e = oy e- rr'X e"';

hence
+
(1.18) cy"' def-
-
25y

_*_ c;i/:
or, introducing n 3 scalar fields c~b def

(1.19)

Now since the field t is invariant for dragging along over j"'dt, the
rotation o[r e"l must have the sameproperty. Hence, accordingto (1.16b) 1)

(1.20)

for every value of b, c and B. This is only possible if

(1.21) I c~ b = const I
Since the fields :"' t are (+ )-constant it follows from (1.21) that

(1.22)
In the same way we get

(1.23) " A -
u[y el- + 2cr A
1 "' e"' I
;

(1.24) I c~s = const.l;

For convenience we always choose the basis vectors be"' and Be"' in
such a way that they are identical at r;"': e"' = (;~ e"'. Then at all points
of X, we have o B b

(1.25)
+ -
The Fcab are all zero because the fields e"' and e"' are (+)-
and rts b B
and (- )-constant respectively. Hence it follows from (1.21) and also
from (1.24) that c;"' is (+)- and (-)-constant and (0)-constant

(1.26)

1) The difference between the n 3 scalars c~b and the tensor c~i,a is important
here.
 2. The parameter-groups and the adjoint group. 191

+
From this and (1.3) and the second identity for RiJ;' (cf. III 5.2)
we get
(1.27)
hence
(1.28)

By (1.26) and (1.28) we are now able to compute the curvature

tensor R;,;j/' of the symmetric connexion Fr(1. (cf. III, 4.25)
+ +
Rt,;/;(1. =28c6Fr'J + 2/[h,F-!J = R;,;(1.- 28c65;J(1.-
+ + + + + +
(1.29)
- 25c;, 1;1(1. F-iJfJ- 2Fc~ll 5;J'- 25c;, 1;1(1.5;J'
+ + + + + +
= - 2 Vc 6 5;J(1.- 25;,; 5;(1._ 25c;, 1;1(1. 5;J'

=- tc;,;c;t.
Hence Rt,;/;(1. is an algebraical concomitant of c;/;(1. and according to
(1.26) this implies

(1-30)

from which we see that every group space is a symmetric A, with respect
to its (0)-connexion (cf. III 7). Moreover we see that all ditferential
concomitants of group space are ordinary concomitants of c;(1. (cf. III 7).
This has as a consequence that all thesedifferential concomitants have
constant components with respect to the anholonomic systems (a)
and (A).
Exercise.
IV 1,1. Prove (1.3) using (1.1, 2).

2. The parameter-groups and the adjoint group

of a finite continuous group.
The equation
(2.1) 'T=UT
represents a transformation of the elements of a group by which a
one to one correspondence between the elements and their transforms
is established. If for U all the elements are taken we get .a group of
element transformations, called the first parameter-group of the given
group, and the transformations of this new group are in one to one
correspondence to the elements of the original group. The following
192 IV. LrE groups and linear connexions.

propositions can easily be proved for the transformations of the first

parameter-group:
a1) If 5, T is transformed into '5, 'T, the pairs 5, T; '5, 'T arc
(- )-equipollent and the pairs 5, '5; T, 'T are (+ )-equipollent.
a 2) If a transformation of elements transforms every pair of ele-
ments into a (- )-equipollent pair, the transformation belongs to thc
first parameter-group.
b) If 5 1 , T1 ; 5 2 , T2 are ()-equipollent, their transforms are also
( )-equipollent.
In the same way, starting from the transformations 'T =TU the
second parameter-group can be defined. For this group the same pro-
positions hold as did for the first group if () is changed into (=f) (see
Fig. 16).

Fig. 16.

Given a pair of elements r;"', r;,"' there is exactly one transformation

of the first and also exactly one transformation of the second para-
meter-group transforming r;"' into r;,"', namely Tr; Tr;- 1 and Tr;- 1 Tr; respec-
tively. Now let there be a transformation of the first parameter-group
carrying the element r;a. intD r;a. + dr;a. and the element r;a. into r;a. + dr{'.
1 1 1 2 2 2

Then the point pairs r;'X, r;a. + dr;a. and r{', r;a. + dr/' are (+)-equipollent.
1 1 1 2 2 2
Hence, if dr;"' = ea A~ {r;"'} dt it follows that dr;"' = ea A~ {r;a.} dt and the
1 1 2 2
constants eadt are the components of dr/' at r;a. and dr/' at r;a. with re-
1 1 2 2
spect to the anholonomic system (a). This proves that the transforma-
tions
(2.2) r/' --+ r;a. + ea.b d t
are r linearly independent infinitesimal transformations of the first
parameter-group. The same reasoning holds for the second parameter-
group and the system (A). Hence
(2.3)
 2. The parameter-groups and the adjoint group. 193

are r linearly independent infinitesimal transformations of the second

parameter-group. Taking S='YJa., T='YJa.+ea.dt, U S='YJa.+ea.dtin Fig.16
B b
we get back the situation of Fig. 15 and again it is proved that the
fields ea. (ea.) remain invariant for the infinitesimal transformations (2.2)
B b
((2.3)) of the first (second) para;meter-group.
Apart from any considerations about groups we now ask whether it
is possible to have in an X, an anholonomic system (a) satisfying the
condition
(2.4) c(cb) = 0
with l r 2 (r -1) given (not necessarily constant) coefficients c~ b, and
the condition that the fields ~P are invariant for r linearly independent
infinitesimal transformations. (2.4) expresses the fact that (cf. li 9.2)
(2.5)

But from this equation and Exerc. li 9,2 we see that the first integrability
conditions of (2.4) can be written in the form

(2.6)

The invariance of the ~P for an infinitesimal transformation 'YJ"~'YJ" + va.dt

is expressed by [cf. Exerc. li 10,6 and (2.5)]

(2. 7) 0 = ;tvbe" * v" ocbe" - (oc v" - ) e" = - (8c v" - vd c"dcb
vd c"dcb ) e"

and the first integrability condition of this equation is (cf. li 9.4)

(2.8) nb"'"* " ' dCjdj

= - U[eV
;:.~ec ubV
ac],.

or
(2.9)

Because this must hold for r linearly independent choices of v" we get

(2.10)

Now (2.6) and (2.10) tagether are equivalent to

(2.11)
If the conditions are satisfied, to every set of values ~P, va. chosen
at a given point 'YJa. of X, there exist fields ~P, va. in an 9? ('YJa.) satisfying
* r linearly independent vectors va. are chosen
(2.4) and (2.7). If * as the
r contravariant basis vectors ea. of an anholonomic coordinate system (A)
B
Schouten, RicciCalculus, 2. Auf!. 13
194 IV. LIE groups and linear connexions.

we have
(2.12)
and accordingly
(2.13)
hence
(2.14)
where
(2.15) , CA def _
CB-
Ca
cb
gy ~ceY Beae !d
As the e"" are invariant for the transformations e""dt, the fields e"" and
b B B
#Ii
are also invariant for the transformations e""dt. But we have already
b
proved that r linearly independent fields ~ satisfying equations of the
form (2.14) andinvariant for r linearly independent infinitesimal trans-
formations can only exist if the 'c~B are constants satisfying the equa-
.
t IOnS E1 A 10
C[DC CB]E = .

Now if at some arbitrary point rJ"" we choose the erx such that they
0 B
coincide with the e"", we have at that point according to (2.15)
b
(2.16)
and because the c~b and 'c~B are constants, this holds at all points.
+
T o the fields e"" ~/l belongs a connexion, say ry""p. for which these
b '
fields are covariant constant and in the same way a connexion r;p
can be constructed with respect to the fields e"", lp. Then it is easily
B
proved from (2.4) and (2.14) that

(2.17)

and from the invariance of the fields erx and e"" with respect to each
other that b B
+
(2.18) S;"" = - S;p"".
This means that the geometry in X, is just the same as in a group
space of an r-parameter group. But the X, is not yet a group space;
the r-parameter group itself is still missing.
Now we have in fact two groups of point transformations in X,,
viz. the invertible transformations leaving invariant the fields e"" and
B
those that leave invariant the fields erx. Taking the first group we see
b
that the transformation rJ""---7rt+ ebA~dt represents for all values of
the constants eb one of its infinitesimal transformations (cf. II 10).
 2. The parameter-groups and the adjoint group. 195

We also see that the set of finite transformations

(2.19)

depending on r parameters t eb and consisting of the transformations

of all one-parameter subgroups each generated by the infinitesimal
transformation for some values of eb (cf. li 10), is contained in the
set. But we do not yet know whether the set contains all the trans-
formations of the group.
If we take the n"' in (2.19) as fixed, say n"' = 1]"', the point 'n"' describes
a curve with the differential equation 1

(2.20)
+
and this is a geodesie in X, for the connexions r;, Fy"' and
+ -
Fy"' = t (Fr"' + Fr"') because

(2.21)

The t eb are normal coordinates in X, with respect to the point 17"'

1
and the connexion r; (cf. III 7). Hence there is always one and
only one choice of t eb Jor which 'nrx = 17"' if 17"' is some arbitrarily given
2 2
point in 91 (n"'). Moreover it follows that there is one and only one trans-
1
formation in the set that transforms 17"' into an arbitrary point 1Ja. of
1 2
91(n"'). This transformation belongs to the group but we have not yet
1
proved that the group does not contain another transformation carrying
17"" into 17"'. Now every transformation of the group leaves the fields e"'
1 2 B
invariant. Hence it leaves invariant the set of all geodesics and also
the natural parameters on them. Hence if 17"' is transformed into 17'7.
1 2
the set of all geodesics through 17"' is transformed into the set of all
1
geodesics through 17"' and this transformation is fixed completely by
2
the condition that the direction of a geodesie at 17"' is (+)-parallel to
1
the direction at 1Jrx of its transform. Moreover the normal coordinates
2
of a point with respect to 17"' must be the same as those of its transform
1
with respect to 17"' and this condition not only fixes the transformation
2
of the geodesics but also the transformation of all points in 91 (n"').
1
13*
196 IV. LIE groups and linear connexions.

This proves that every transformation in the group is also a transforma-

tion of the set and that accordingly the group is an r-parameter one
+
with the infinitesimal transformations 'YJa.~ r/" eb A~ d t.
If now we associate with every point 'YJa. of X, the transformation
of the group carrying the already chosen point 'YJa. into 'YJa., X, is the group
1
space of this group. The group is now a group of point transformations
in its own group space and accordingly it coincides with its first para-
meter-group.
After this digression on sets of vector fields in an X, we return to group
space and the first and second parameter-group and ask whether the
+
b
transformation 'YJ" ~ 'YJ" il ea. d t is an infinitesimal transformation of
b
the first parameter-group. Necessary and sufficient is the invariance
of the fields erx:
B

(2.22)
b
and this is only possible if the il are constants.
In group theory it is usual to denote an infinitesimal transformation
+
in some xn' f< ~ f< +X" d t = ~,. XP op. ~>< d t by a symbol, for instance X,
standing for the operator XPow Then we have for the transformation
of ~>< and of any function /(~")
(2.23) d~" =X ~"dt; df =X fdt.

If X and Y are two different symbols of this kind, we write (X Y) for

the operator X Y- Y X. This notation was introduced by LIE.
Using ob det A~ oa. in this way as a symbol for the infinitesimal trans-
formation 'YJ"~'YJa.+ e"dt of the first parameter-group we get from (1.19)
b

(2.24)
hence
(2.25)

In the same way it is proved for the second parameter-group

(2.26) I (oc oB) = - c(;A OA I;

The equations (2.25, 26) are called the LIE structural formulae of the
first and second parameter-group and the c;;,a its structural constants.
On the other hand (1.19, 23) are called the CARTAN structural formulat
 2. The parameter-groups and the adjoint group. 197

of these same groups. These latter equations were established first by

MAURER.l) LIE started from infinitesimal transformations and this
is the reason why he looked at the matter from the contravariant side.
But CARTAN started from systems of Pfaffians and his point of view
was therefore entirely covariant. The rather fine dualism between the
two points of view was pointed out by CARTAN.
Comparing the equations (1.19) and (2.25) it is obvious that (1.19)
contains an invariant operation, viz. Rot, but that the operator oc is
only invariant if it is applied to scalars. Now we have according to
(1.16)
(2.27)

and therefore, considering Exerc. II 10,9

(2.28) (f f) (/) = c~b f tP; (ff) def f f - ff
c b a cb cb bc

for every quantity (/) (indices suppressed). Hence

(2.29) (f f)
c b
= c~b f
a
12)
and in the same way
(2.30) (ff)
CB
= - c~B f
A I
(2.29) and (2.30) are equivalent to (2.25, 26) if the operators are applied
to a scalar. But in contradistinction to (2.25, 26), in (2.29, 30) only
those operators occur that can be applied to all quantities.
The finite transformations of the first and second parameter-group
can be written in the form (cf. II 10)
(2.31)
and also in the more general form valid for any quantity (/) (indices
suppressed) dragged along (cf. II 10)
m b m Bo
m -tem m -te "m
(2-32) w=e b'V; w===e By...r;

We gather tagether the following results:

An r-parameter group being given, there exist in group space an-
holonomic coordinate systems (a) and (A), satisfying the equations
1) MAURER 1888, 1, p. 117; cf. for the solution of these equations \VHITEHEAD

1932, 2 also for literature.

2 ) The identities (2.29, 30) are proved here only for quantities. But they are

valid for all geometric objects having second Lm derivatives (cf. II 10).
198 IV. LrE groups and linear connexions.

(1.19), (1.23), (1.21), (1.24), (2.25), (2.26). There are two groups of
point transformations in X,, the two parameter-groups, whose trans-
formations are in one to one correspondence with the transformations
of the given group. The first leaves invariant the fields of basis vectors
of (A) and the second those of (a). Conversely if r 2 (r -1) constants
c~b = c]'cb] be given that satisfy (1.28b), it is always possible to construct
in an arbitrary X, an anholonomic system (a) satisfying (1.19) and
(2.25) and another anholonomic system (A) satisfying (1.23) and (2.26).
The ob are the symbols of infinitesimal transformations of a group of
point transformations in X, that leave invariant the fields of the basis
vectors of (A) and the same holds for the oB with respect to the basis
vectors of (a). The transformations of these groups can be brought
into one to one correspondence to the points of the X,, thus making it
a group space.
Besides the two parameter-groups there is a third transformation
group connected with every group. The equation

(2-33) 'T = UTU- 1

represents a transformation of the elements T of this group. If for U

all the elements are taken we get the adjoint group of the given group.
Because of
(2.34) V U T u- 1 v- 1 =(V U) T (V U)- 1
this group and the given group are homomorphic but the homo-
morphism need not be an isomorphism (that is one to one). In fact,
two elements ul and u2 may exist such that

(2.35)

for every T. This is the case if and only if there exists besides I another
element that is commutative with respect to every element of the group.
All elements of this kind form an invariant subgroup called the centre
of the group. Let C be an element of the centre, then the transforma-
tion U T u- 1 of the adjoint group corresponds not only to U but also
to CU.
An element T and its transform U T u- 1 are called homologous.
A subgroup and its transform are also said to be homologous. If a sub-
group is identical with its transform for every choice of U it is called
an invariant or normal subgroup.
From the definition (2.33) it follows that a transformation of the
adjoint group arises if a transformation of the first parameter-group
carrying I into U is followed by a transformation of the second group
 2. The parameter-groups and the adjoint group. 199

carrying I into u- 1 If srx. is a vector at rt and irx. and srx. are the fields
0
we can get from srx. by ()-displacement, the infinitesimal transforma-
tion of the adjoint group corresponding to the point rt+srx.du arises
+)-
0
if the (-)-parallel displacement over !a. du is followed by the (
parallel displacement over - s"--du. A vector ea at rt is transformed
0
by these transformations first into the vector ea-J:~ eb sc du and
afterwards into the vector ea-c;i,a eb sc du. Hence the infinitesimal
+
transformation corresponding to the point rt e''- d t in group space,
0
with the symbol ea A ~ 811 , is transformed by an i nfini tesimaltransforma tion
of the adjoint group in the following way

(2.36)

This is an infinitesimal transformation of the ea with the symbol

(2.37)
Because
(2.38)
the Eh and ob satisfy equations with the same structural constants.
But the Eb are independent if and only if the c-rank of ca,a is equal to r.
Writing s;,a def scc;ba we get for a finite transformation of the Olle-
parameter group generated by sb Eh for some constant values of the sb
(2.39)

where S symbolizes the matrix of Sia The equation (2.39) represents,

for variable values of the sa, a linear homogeneaus group of transforma-
tions of the ea, called the linear adfoint group. It is the adjoint group
but only if considered as a group of transformations working on the
infinitesimal transformations of the given group.
If we introduce the t ea as normal coordinates 'YJ"- of X, with respect
to the point 'YJa.
0
(2-40) a = 1, ... ,r
the three coordinate systems (a), (A) and (a) have the same basis
vectors at rl and we see that the transformation of these new Coordinates
0
is also linear homogeneous. That proves the statement:
The transformations of the adfoint group are linear homogeneaus t'f
they are expressed in the normal coordinates 'YJo. with respect to 'YJ"-.
0
200 IV. LrE groups and linear connexions.

We write these transformations

J a) dr;"=- Si,"r;bdu; Sb a def S ba {r;"'}I)

0 0 0
(2.41)
1b) Ir;" = ( e-uS)a "
0 b r;V ; S = matrix of Si,".
0 0

(2.41 b) is the transformation u ... u- 1 if u corresponds to the point

r;" = <5~ usa of group space. This transformation arises if the ( +)-
equipollence that transforms U into I is followed by the (- )-equi-
pollence that transforms I into U. The first transforms the vectors e"'
b
at <5" usa into the e(/. at r;"' coinciding with the e"' at that point, and the
a b 0 B
second transforms these vectors into the e"' at <5~ usa. Hence
B

(2.42) a) ;"'=b~(e-uo5 )~t or b) A~=(e-ul)~;

or in words:
The vectors e"' at a point of group space that corresponds to the trans-
b
formation U of the group, transform into the vectors e"' at that point by
B
the same linear homogeneaus transformation that represents the trans-
formation UTU- 1 oj the adjoint group ij applied to the normal Coordinates 'rja.

Exercises.
IV 2, 1. If Operators fb are defined by ;tb def eb f, prove that
a
IV 2,11X) (fcfb) = 0.

This is the reason why we do not use the operators fb. The same dif-
ficulty does not arise with the operators ob because ob ec = 0 whereas
fb ec =l= 0.
IV 2,2. Prove that
IV 2,2 IX)
IV 2,3. Prove that the transformation r;"' -+r;"' v(/. d t represents +
an infinitesimal transformation of the first parameter-group if and only
if va = const.
IV 2,4. Prove that
IV 2,41X) ff= f f
c B B c
and
IV 2,4 )
1) Note that the components Si/ are constants butthat the Si, a equal the Si/
at 1J"' only.
0
 3. Finite continuous transformation groups. 201

+
IV 2, 5. Prove that a quantity is ( )-constant if and only if its
components with respect to (a) are constants.
IV 2,6 1). Prove that the (0)-connexion in group space is volume pre-
serving.
3. Finite continuous transformation groups.
So far we have not supposed that the group originally given was
in any way a group of transformations. 2)
Let us now take an r-parameter group of transformations m n
variables ~"; x = 1, ... , n
(3.1) '~" = f"(~, r/); IX= i, ... ,r.
First we have to make sure that the ~" can be solved from (3 .1).
The necessary and sufficient condition is that the matrix of the o,j"
has rank n. Secondly it is necessary that the rj"" should be essential,
that is, that there should be no equations of the form
(3.2) f" W, rj"") = F" (~.Ca (r/'")); a= 1, ... ,i:'; r'<r.
From (3 .2) it follows that
8
(3-3) opf" = (opCa) oaF"; oa = ez;;a
and because the rank of opCa is ;;;;,r' this implies that there exists at
least one set of functions Uf1 (r/'") such that
(3.4)
Conversely, if an equation of the form (3.4) holds, the f"W, r/'") must be
solutions of the equation
(3.5) Uf1(r/'") opq;(r/'") = 0
containing parameters ~". But because an equation of this form has
at most r -1 independent solutions (cf. II 5) the 'YJa. cannot be essential.
Hence the 'YJa. in (3.1) areessential if and only if no equation of the form
(3.4) exists.
From now on we suppose that the 'YJa. areessential and that 'YJa. cor-
o
responds to the identical transformation. Then there is a one to one
correspondence between the transformations of the group in a neighbour-
hood of the identical transformation and the points of the X, of the
'YJa. in an 91 (na.).
_____ o
1) CARTAN 1927, 1, p. 67.
2) HAIMOVICI has proved 1945,1; 1946,1; 2; 1947, 1; 1948, 1; 2 that a connexion
can also be established in the manifold of a set of transformations. Certain con-
comitants of this connexion vanish if and only if the set is a group.
202 IV. LIE groups and linear connexions.

We only consider transformations of this group germ. For every

definite choice rj"' = r/" the equation (3 .1)
1
(3.6) 1 ~"=/"(~v,r/")
1

represents a definite transformation T~ operating on the ~x. Other

1
values r/< = r/" give another transformation T.-,
2 2

(3.7) "~" = !" (e, r/")

2

of the ~". If r/" is changed into r/" + dr/" the change of '~" is
1 1 1

(3.8)

and if in (3.8) ~" is eliminated from the right hand side by means of
(3 .6), the resulting equation represents the transformation of 1 ~" = T~~"
1
into T~+d~ ~" = T~+d~ T17- 11 ~", that is the infinitesimal transformation
1 1 1 1 1

T17 +d~ r~- 1 operating on the 1 ~". This transformation corresponds to the
1 1 1

infinitesimal vector dr/" at the point r/" in group space. In the same way
1 1

(3.9) d"~" = (8p/") 17 rx~~rxdr/

B 2

after elimination of the ~" by means of (3.7) represents the transforma-

tion T~+d~ r~- 1 operating on the variables "~" and it corresponds to the
2 2 2
infinitesimal vector drt at the point r;'Y. in group space. If dr;rx is now
2 2 2
chosen in such a way that
(3.10)

then d r;rx at r;"' is (+)-equipollent to dr;"' at r;"'. Hence, introducing the

1 1 2 2
coordinate system (a) we have (d r;)a= (d r;)a, and the two equations
1 2

(3.11) d 1 ~" = (8p'~")r,a~_;;rxA~ (f) (1r;)h)

(3.12)

after elimination of the ~" represent one and the same transformation,
viz. the transformation corresponding to the point t/"+A~ (r;rx) (dr;)b
0 0 1
in group space, operating on the 1 ~" in (3.11) and on the "~"in (3.12).
But this is only possible if the expressions
(3 .13)
1) Cf. for the use of A~ as a function symbol, II 3.
 3. Finite continuous transforroation groups. 203

after elimination of the f' by means of (3.6), depend only on the ~~~ and
are independent of the r{'.
1
The infinitesimal transformation corresponding to the point
rt" +ea A ~ d t of group space is given by
0
(3.14) d~" =E~ (;") ea d"t; ea = const.
and to every choice of ea d t there belongs a definite infinitesimal trans-
formation with d~" =!= 0. In fact no relation of the form
(3 .15)
can exist with coefficients cb independent of the f' because in that case
there would be a relation
(3.16)
and as we have proved already this would imply that the parameters
were not essential [cf. (3.4)].
According toCAMPBELL 1) we call a set of infinitesimal transformations
connected if they satisfy a homogeneaus linear equation, and dependent
if in addition the coefficients in this relation are constants. It frequently
happens that a set is independent but connected.
Collecting results we have the first part of the first fundamental
theorem of LrE 2):
I.1. I f the invertible transformations
(3.17)

with r essential parameters rt

form a group, there exist r 2 functions A~ (r/);
Det (A~) =!= 0; a, b = 1, ... , r and nr functions Eb (e) for which no equa-
tions exist of the form cb Eb = 0 with coefficients cb independent of the ~",
such that
(3.18)

Starting from the inverse of (3 .17)

(3 .19)
it can be proved in the same way that

(3.20)

1) CAMPBELL 1903, 1.
2) In this and the following fundamental theoreros all functions are supposed
to be analytic in the region considered.
204 IV. LIE groups and linear connexions.

The finite transformations of the group are found as follows. The

equation of a geodesie through r/" is
0

(3.21) d!l"' = A"' ea. rt = r!"' for t = 0.

dt a ' 0

Hence at every point of this curve we have according to (3.18)

(3.22)

and the solution of this equation with the condition 1

~" = ~" for t = 0
is (cf. II 10)
(3.23) ~~"=e 1 x~"; X~~ebXb; Xbd~S~W)ow

From this, for any function f (~") we get

(3.24)
representing the transformation of the value at the point ~" into the
value at 1 ~". But if we have a scalar field p = f W) and if this field is
dragged along by the point transformation (3.23) the new value at ~"
is the original value at e-tx~" (cf. II 10), hence this new value is

(3.25)
and also
(3.26)

where f is the LIE operator belonging to the field eb 3~.

Unlike (3.25) the formula (3.26) can be generalized for any quan-
tityW (indices suppressed) as defined in II 3 (cf. II 10.38)

(3.27)

If we introduce in X, the normal Coordinates 'f)a; a =I, r, with 0 0 0'

respect to 'f)"' and to the coordinate system (a), then the transformation
0
(3 .23) corresponds to the point in group space with the normal coordinates

(3.28)

The system of equations (3.18) has the solutions

(3.29)

depending on the n parameters ~". Hence the system is totally integrable

(cf. II 5) and its integrability conditions must be satisfied identically.
 3. Finite continuous transformation groups. 205

or, because the choice of the variables 'e is free and consequently ';"
can be replaced by ;"

(3. 31)

This identity only contains operators operating on scalar functions.

But it can be generalized by generalizing the operator :t defined in (1.8).
f be the Lm operator in the (n + r)-
b
Let from now on in this section
b
dimensional space of the ;" and r/' belonging to the field ea
b
..s:, t(f.
b
for
every value of b. 1 ) Then

(3-32)

is true and equivalent to the combination of (2.25) and (3.31) so long

as scalar functions only are operated upon. According to (2.28) it is
also true for quantities in X 1 In order to prove that it also holds for
quantities in Xn it can first be proved for CO- and contravariant Vectors
and after that the rule of LEIBNIZ can be applied. 2 )
Of course besides (3.31, 32) we also have

(3.33) I (XcXB) = - ccA xA I

(3.34) l(ff)=-clB:tl.
CB A

(3.31, 33) and also (3.32, 34) are called the Lm structural formulae
of the group and the c;;,a its structural constants. So a group of trans-
formations and both its parameter-groups have the same structural
formulae and the same structural constants. This is of course trivial
because these groups are isomorphic.
1) In the classicalproof of (3.31) (cf. LIE-ENGEL 1888, 1, p. 158ff.) the equa-
tions (3.18) are first brought back to the linear homogeneaus system ;\; rp (;", 11(1.) = 0,
b
so the operator ;\; really occurs there in its most general form, except that it is
b
connected with scalar functions only.
2 ) It is also valid for all geometric objects having second Lm derivatives (cf.

II 10).
206 IV. LIE groups and linear connexions.

Collecting results we may now state the first part of the second
fundamental theorem:
II.1. Jf Xb are the symbols of r independent infinitesimal transforma-
tions of the group mentioned in 1.1

(3 -3 5)
equations of the form (3-31) with constant coefficients c;;,a, c(;it=O hold
and the first part of the third fundamental theorem:
III.1. The constants c;;/ satisfy the equations (1.28b).
The second part of the first fundamental theorem is the converse
of the first part :
1.2. If a set of transformations of the ~" is given

(3-36) ~~1<=/"(~,rt); ~=i, ... ,r;

with Det (8;. '~") =l= 0 in an 91 ( ~", r/'"), depending on r essential parameters r/'
0 0
and containing the identical transformation (r/' = r/'") and if there exist
0
r 2 functions A~ (r/'); Det (A~) =l= 0 and nr functions Eb (;'") not satisfying
an equation cb Eb = 0 with coefficients cb independent of ~", such that

(3-37)

these transformations form an r-parameter group germ.

Proof: though the Xr is not yet proved tobe a group space, the A
and their inverses A~ fix an anholonomic coordinate system (a) and
+
a connexion rya satisfying the condition that the ea, ~ are covariant
b
constant. The geodesics of this connexion satisfy the equations

(3-38) ----
d'YJ(f.
dt
= eb Ab"'
with constants ea.
rJ"' be a point of Xr in 91 (rt) corresponding to the arbitrary
Now let
1 0
transformation T~ of the set. Every point 'rJ"' of 91 (r!"') on a geodesie
I 0
through 'rJa is given by
1
(3-39)

for suitable values of t and ea. If '~"=/"(~, 'YJ"'), then, changing ''rJ"'
into the point of the same geodesie 'YJa+dYJa='YJ"'+eaA~dt we get

(3.40)
 J. Finite continuous transformation groups. 207

and accordingly the transformations of the set belonging to an arbitrary

point r;"" of this geodesie is given by
2
(3.41) ~"=rW.r;""); X b=
def ';;I'"
...... b 0 /A
1 1

for some suitable value of t.

For fixed values of the eb the transformations ~---+'~"form a one-
1
parameter group with the parameter t. Every point r;"" of 9? (rt) lies
0
on one of the geodesics through r;"". Hence every transformation of
1
the set can be obtained by the transformation TtJ followed by a trans-
1
formation of one of these one-parameter groups and conversely every
transformation constructed in this way belongs to the set. As r;"" is
1
entirely arbitrary we may take r;"" = r;"" and this proves that every
1 0
transformation of the set belongs to one of these groups. But that
implies that two transformations of the set applied one after the other
give rise to a new transformation of the set. As the set also contains
the identical transformation and the inverses, it is proved that the set
is a group and the X, its group space.
The second part of the second fundamental theorem is the converse
of the first part :
Il.2. lf Xb = E~ (f') ol' are the symbols of r independent infinitesimal
transformations of the f' satisfying the equations (3. 31) with constant
coefficients c~b, c(cb) = 0; these transformations are the infinitesimal trans-
formations of an r-parameter group germ.
In order to prove this we remark first that the identities (cf.
Exerc. IV 3.3)
(3 .42) (X[d (Xe Xb1)) = 0
can easily be verified and that they imply that the c~b satisfy the identities
(2.11 b). Now we have proved in IV 2 that according to these identities
and the constancy of the c~b there exist in the X, of the auxiliary
variables r;"", r linearly independent vector fields fffJ satisfying the equa-
tions (2.4). But then it follows from (3.31, 42) that the integra-
bility conditions of (3.37) are satisfied identically. Hence these equa-
tions are totally integrable and they have solutions of the form
(3.43) '~"=rW.r;""); ~"=f"(~".r;""); Det(o"'~")=f=Oforr;""=r;""
0 0
valid in an 9? ( r;"") and depending on n parameters ~". That finishes the
0
proof because according to the fundamental theorem I.2, (3.43) represents
an r-parameter group and the xb are infinitesimal transformations of
208 IV. LIE groups and linear connexions.

this group. As we see an r-parameter group germ is wholly determined

by r linearly independentinfinitesimal transformations satisfying (3.31)
with constants c~b. It is said to be generated by them.
The second part of the third fundamental theorem is the converse
of the first part :
III.2. lf ~ r 2 (r -1) constants c~b; c(cb) = 0 satisfy equations of the
form (2.11 b) there always exists an r-parameter group germ with r linearly
independent infinitesimal transformations xb satisfying (3-31).
In order to prove this we construct in the X, of the auxiliary variables
'YJa. the fields SfJ satisfying (2.4). In IV 2 we have already proved that it
is possible to construct in X, an r-parameter group of point transforma-
+
tions with the infinitesimal transformations 'YJa. -+'YJa. eb A ~ d t whose
symbols Xb satisfy (3-31). But as the fundamental theorem 11.2 is
already known, it is only necessary to remark that these Xb satisfy
(3-31) because then according to 11.2 they generate an r-parameter
group germ. 1), 2)
The results obtained so far show that the properties of an r-para-
meter finite continuous group germ of an abstract group correspond
to and are in fact identical with the algebraic properties of a tensor
c;i,a, which is alternating in cb and satisfies (1.28b) and also identical
with the geometric properties of a very special A, which is symmetric
with respect to all its points (cf. III 7).
As soon as the elements are supposed to be transformations in
n variables we also get properties in the space of these variables. But
it makes a difference whether the elements are interpreted as point
transformations or as coordinate transformations as the following
example will show.
Let us assume first that the elements are point transformations in
the Xn of the ~". Let T and U be two of these transformations

(3 .44)
Then we have
(3.45) 'l" = U'~" =ur~"= uru- 1 l".
1 ) In the proofs of the three fundamental theorems we made use of the existence
theorems of partial differential equations. For our purposes this is the natural
way because sets of these equations are just what we are interested in. FREUDEN-
THAL proved in 1933, 1 that it is possible to use only the existence theorem for
ordinary differential equations.
2 ) E.CARTAN has shown 1930, 1; 1937. 2, p. 189 (with LERAY) that the original
proofs of the fundamental theorems given by LIE fail if one does not confine one-
self to the group germ. Of course the same holds for the proofs given here.
 4. The geometry of group space. 209

Hence, if T transforms ;" into 1; " we get the transformation of U ;x

int0 u I;>< by applying the transformation u,,, u- 1 0f the adjoint group
to T. Now let us assume that T is a point transformation and U
a coordinate transformation

(3.46)
Then we have
(3.47)

and this means that T describes a point transformation with respect

to (u) and U T u- 1 describes the same transformation with respect to (u').
In group space the transformations of the adjoint group are of
course always point transformations as we have already seen in IV 2.

Exercises.
IV 3, 1. Prove that
IV 3,11X)

IV 3,2. Prove that the functionsEb(;") and AHn<%) can be obtained

by algebraic operations and differentiations from the equations (3.17).
(Use the transformations 1; " = f" (;", 171%); ";" = f" (' ;", Ca;)= f" (;", ()IX)).
IV 3,3. Prove that
IV 3.3 IX) (f(ff))=O.
[c b a]

4. The geometry of group space.

Ingroupspace we now have the general Coordinates 1]1%, the integrable
but not symmetric ( +)-
and (-)-connexion and the anholonomic
coordinate systems (a) and (A) belonging to them, the symmetric (0)-
connexion and the normal coordinates 17a with respect to 1]1%, the first
0
and second parameter-group with the transformations T ~ U T and
+
T ~TU consisting of all (- )-equipollences and ( )-equipollences, and
the adjoint group with the transformations T ~ U T u- 1 linear homo-
geneous if expressed in the 17a. In this section a number of important
relations will be established.
In III 7 the relations between general coordinates and normal
coordinates were derived. If this time we take IX, , y = 1, ... , r instead
of u, A,fl=l, ... , n and a, b, c= I, ... , r instead of h, i, f=1, ... , n we
Schouten, Ricci-Calculus, 2. Aufl. 14
210 IV. LrE groups and linear connexions.

get from (III 7.18, 19)

(4.1) 'Yla = (p 'Yja _ _!___ r.a

"I a 2! y 0
{'YJ} {JYcb 'Yjc 'Yjb _ _!___ r.a
3! y,y.
{'YJ} (JYc, Yc.b'Yjc, 'Yjc
2 1 1 'Yjb _
0

(4.2)

where
(4 . 3) r.a def
yp .. Yt-
0 (Yp r.rx
Yp-.... y.),.
_ p T,6(YPYP-t r.aYp-, ... y.)6

Instead of (III 7.23, 24, 25) we have here

(4.5) r.a
cp ... c1 b{ 'YJ } -- 0
0

(4.6)

(4.7)

Now A~ and A could be derived from (4.1, 2). But in group space
we already have A~ and A and since it is very easy to find expressions
for Ag and Ab the indirect way is better. We consider the geodesics
through 'YJa At every point of such a geodesie we have on one hand
0

(4.8)
and on the other hand
(4.9)

and (a) and (a) have the same basis vectors at rt. Hence, because d'Yja oc 'YJa
we have 0

(4.10)

though A~ =CJ~ at 'YJrx only. By differentiation of (4.10) we get

0
(4.11)

The identities (1.19) are true forevery choice ofthe coordinate system (cx),
hence
(4.12)
 4. The geometry of group space. 211

and if this is substituted in (4.11) we get

(4.13)

For differentiation along the geodesie we have

d d YJ d r;b 0 - Db b - 1
(4.14) fit=(ito-(it "-
-
be 8"- 1 r; b ob,
hence, according to (2.36, 40) and (4.13, 14)

(4 15) ~Aa+_!_
dt C t
(Aa-
c
tf!)
c
= ca Ac eb = - Ea Ac.
cb c c c
Ebadef ec ccba.

Because A~ {r;} = ~ this equation can be compared with the ordinary

0
differential equation

(4.16)

whose well known solution can be written as a series

(4.17) e"t-1
X=X--=X
(
o rxt o
J+-od+-IX
1
2! 3!
t +)
1 2 2

and from this we derive the solution of (4.15) in the form

(4.18) a
Ab=
~ ( e-Et_A )a
U(,
-Et b
= S!.b (
ub A -
1
-E
2!
t + -E
1 2 2
t - )a .
3! b

The inversion of (4.17)

(4.19)
with
(4.20) A. = (- J)q Bq_
q q! '

where Bq are the BERNOULLI numbers 1 ) B 0 =1; B 1 =!; B 2 =k; B 3 =0;

B 4 = - io; B 5 = 0; B 6 = -l2 ; B 7 = 0; B 8 = -l0 , gives the inversion
of (4.18)
(4.21)

In the same way we find

(4.22) AA = DbA ( eEt-
b Et
ba
A )a
b
= DbA
ba
(A + _!__2! E t + _!__3! E2 t2 + .. )a. b'

(4.23) A Ba = DaaBb ( eEt-A

E t )a = Da b (A
b aB
+ "1J E t + A.2 Ez t2 + .. )ab
1) Cf. for instance Pascal Repertorium I, p. 437.
14*
212 IV. LrE groups and linear connexions.

From (4.18, 23) and (4.21, 22) we find by transvection once more that

(4.24) Aa = (Jb (e-E 1-A . Et )a = (Jb (e-Et)a

B B -Et eEt_Ab B b

(4.25) A: = ~ (eEt)g
in agreement with (2.42b).
The matrix notation can be used to compute the r,ab From (III 7.28)
we get in group space
(4.26) cb
r.a
= 'l'lb
.,
,.a + 11'lb,'l'lb'N'
Nbeb
0 ~ ., .,
a + ...
b,b,cb 0

According to (III 7.36) and (1.29) the first term of the right hand side
takes the form
(4.27) i 'YJb !fi> (~b)a = i'YJb fi> (~ e fi>); a.
The normal tensor of valence 5 is zero and the normal tensor of valence 6
can be written as the transvection of two tensors R;,~;,a (cf. III 7)
from which four factors c~;,a result. The result is a series with an even
number of factors c~~a in each term. Butthis result can be obtained much
0
more easily in the following way.
Because the t are (+)-constant we have
+ +
(4.28) a) r,a" = J(~bl =A~ a,,A&l; b) c~;,a = - 2Sa,a = - 2A~ 8[,A&1.
Hence, writing P for the matrix ee c;i/ t = n' b~ c;i/ and Ce for the
matrix c;ba we have
(4.29)
and (4.18) can be written
(4.30) A ab = (Jbb (A - _!___ p + _!___ p2 - _!___ p3 + .. )ab.
2! 3! 4!
From this we get, taking into account the non-commutativity of C
andP
8, A& = ~t(- :! Ce+ :! Ce P + :! PCe-
(4-31) 1 _ _!__c pz _ _!__pc p _ _!__pzc + .. )
4! e 4! e 4! e
and after some calculation (note that the factor A~ in (4.28a) comes
in matrix notation at the right hand side)
(8,A&) A~ = bg~~ (-~Ce- f'ICe P + tPCe- -fiCeP 2 +
+ -.fr p Ce p - pz Ce+ .. )~ n
= bt!~ (-lc;i,a- r\c;i,d'YJe(J~c;,t+ i'YJe~c;i,d c;;;a-
(4.32)
 4. The geometry of group space. 213

Hence, making use of the fact that (a) and (a) have the same basis
vectors at r/'" we get
0

and by mixing and alternating over cb

+
(434)
racb = r,a(cb) = . b fc)b
1 efc(b
61] .. a+t erms of an oddd egree

I
in accordance with (4.27) and

ca + =
cb = _ 2r,a
[cb]
ca
0 cb
+ _L'}']c,'}']c, cb, c .o, c .a _
121 ., 0 cb 0 c,b 0 e,b,1

(4.35) __121
L'}']c,'}']c, c .b, c b, c .a
., c,[c
0 jc,jb]
0 b,b, 0

+ terms of an even degree ~4 in r;a.

By a reflexion at the point r;a = 0 a field with the values va at r;a is
0
changed into a field with the value ua= -va at -r;a. If va are the
(a)-components of the first field at r;a, it follows from (4.18, 22) that the
ua at - r;a are not equal to - va but that uA = - r5;i va. The reflection
of a (+)-parallel displacement of va over dr;a at r;a is a (-)-parallel
displace:ment of ua over - dr;a at - r;a because (+ )-equipollency trans-
forms into (- )-equipollency by a reflexion at the origin. Hence the
reflexion of a (0)-parallel displacement of va over dr;a at r;a is a (0)-
parallel displacement of ua over -dr;a at -r;a. Accordingly, taking
these (0)-displacements we get
(4.36) dva =- dr;c rtb{r;} vb = - dua = + dr;c rcad- r;} v 0
or
(4.3 7)
and this proves that only terms of an odd degree in r;a occur in the series
of rcab in accordance with (4.34).
Taking a (+)-displacement at r;a and its reflexion at -r;a we get
+ -
dva = - dr;c rcab {r;} vb = - dua = + dr;c rcab {- r;} vb
from which
+ -
ccba{r;} = - 2T[~b] {r;} = 2T[~b] {-r;} = c~[,a { -r;};
and this proves that only terms of an even degree in r;a occur in the
series of ca,a in accordance with (4-35).
214 IV. LIE groups and linear connexions.

5. Invariants of a transformation group in the Xn of the ;".

In the X,. of the ~" we may have invariant functions of the ~", in-
variant subspaces or sets of subspaces, invariant fields of quantities
or geometric objects, andinvariant systems of partial differential equa-
tions.
A subspace is called invariant if its points are interchanged by the
transformations of the group; a set of subspaces if the subspaces are
interchanged. A field is called invariant if it does not change when
dragged along by the transformations of the group. A system of dif-
ferential equations is called invariant if by dragging along all solutions
pass into solutions.
Weshalldeal here only with the most important cases of invariance.
1. Invariant functions 1 ). A function cp(~") is invariant if and only if

(5.1)

The infinitesimal transformations Xb are independent but they may be

connected. The number of linearly independent equations among (5.1)
is equal to the b-rank of E;. If this b-rank is q::;;, r, (5 .1) is equivalent
to a system of q linearly independent equations and because of

this system is complete (cf. II 5) and has n- q independent solutions

~' ... ,"(j/ of which every other solution is a fun'ction. The equations
1 n-q
cp = const., ... , cp = const. represent a set of oon-q X/s and each Xq
of this set is invariant for transformations of the group.
If the finite equations of the group

(5-3) '~" = f"W, r/'")

are given we can see that there must be invariant X/s by giving the
~" fixed values ~" and letting the 'YJrr. vary. Then an Xq is described
0
because opf" has the rank q. (5.3) represents a system of n equations
in 2n+r variables, which is regular of dimension n+r and whose rank
with respect to the variables 'Y/rr. is q. Hence according to the well-
+
known theorem of elimination 3 ) exactly 2n r- n- r- q = n- q equa-
tions not containing the 'YJrr. can be derived from (5.3) and from these
1) LIE-ENGEL1888, 1, p. 211 ff.
The invariance for finite transformations follows from the series (2.31) and
~)
also from the series (2.32) because (5.1) can also be written in the form :1:: rp=O.
b
3) P. P. 1949, 1, p. 42.
 s. Invariantsofatransformation group in the Xn of the ~". 215

equations n- q of the '~" can be solved

a=l, ... ,n-q; ~=n-q+l, ... ,n.

Nowwe know that there exist n- q invariant functions ~; a = 1, ... , n- q;

and this proves that the set of n- _q equations

(5.5) rw) = r('~")

is equivalent to the set (5.4). Hence, if we give the '~"in (5.4) constant
values c1 , ... , cn the equations

(5 .6) "
1p(c', ~") = const.

represent the oon-q invariant X/s and the left hand side gives a set
of n- q independent invariant functions for each choice of the q con-
stants c'.
2. Invariant subspaces 1 ). A set of independent functions T(~") is
called relative invariant 2) if the equations

(5.7) a=l, ... ,n-q;

are a consequence of T=
0. This means that the Xq represented by
~ = 0 is invariant. For every choice of the constants eb the Xq is built
up by streamlines of the field eb E;. Therefore the dimension q is
minimal if and only if q is the rank ofE/:. Such an Xq through a point ~"
0
is called the smallest invariant subspace through ~". Hence, in order to
0
find all invariant subspaces of a group we have only to construct all
points where E~ has a rank q for all values q ;;;;,r. 3 )
The group interchanges the points of Xq and we may require the
group of these transformations. Let
(5.8) ~" =W"(yh); h = 1, ... , q;

be the parametric equations of the Xq and let Yb ct~! Yj oi be the infinites-

imal point transformations in Xq corresponding to Xb. Then, if /(~") =
t (w" (yh)) is given at all points of xq we have for the subgroup generated
by eb xb
(5.9)

1) LIE-ENGEL 1888, 1, p. 222ff.

2 ) EISENHART 1933, 1, p. 63.
3 ) LIE-ENGEL 1888, 1, p. 228, 237.
216 IV. LIE groups and linear connexions.

and for the subgroup generated by eb Yb

(5 .10) !:}__
dt
= eb Y/b l:'"
1 u,. f' ><1 def
-
O1~~"'

hence
(5 .11) E;; = BJ.Yl'.
In general such an equation can not exist because the x-domain of
E; does not coincide with the x-domain of BJ.. But here the Xq is in-
variant and this means that its tangent Eq is the support of both domains.
This implies that the Yt,h can be solved from (5.11). The easiest way to
find this solution is to choose the coordinate systemsuch that ;q+l = 0,
... , ;n = 0 are the equations of the Xq and to take ;q (suitablye, ... ,
chosen) as parameters on the Xq. Then h=l, ... , q and
(5.12) Bi:=ClJ.; "Yt,"=Ei; E'/,+ 1 =0; ... ; Eb=O.
Hence we get the Yt o" by dropping in E~ o,. all differentiations with
respect to ;q+l, ... , ;n. From this it follows immediately that the
group of point transformations in Xq found in this way, and called the
group induced in Xq, has r infinitesimal transformations Yb satisfying
the equations
(5.13)
These r transformations need not be independent. Take for instance
the case where the Xb are independent but connected. Then there
exists at least one equation of the form "Pb(;,.) Xb = 0. It may happen
that the "Pb(;,.) are constant over Xq and in that case the Yi, are not
independent. Let the induced group be r' -parametrical, or in other words,
Iet there exist exactly r- r' relations
(5.14) cbYi, = 0; p = 1, ... , r- r'; cb = const.
p p

Then there must also exist exactly r - r' relations

(5.15) cbXb=O
p

valid at all points of Xq. But that implies that there exist r- r' in-
dependent infinitesimal transformations of the original group that leave
every point of Xq invariant. Of course these transformations form an
(r- r')-parameter subgroup:
The group induced in an invariant Xq is an r'-parameter group if
and only if there exists an (r- r')-parameter subgroup of the original
group, leaving each point of Xq individually invariant.
A group is called transitive if the rank of s; is n at all points of an
9l(;"). If the group is transitive, to every point pair P and Q in a
0
 5. Invariants of a transformation group in the Xn of the ;". 217

sufficiently small 91 (~") there is at least one transformation of the group

0
transforming P into Q. Of course for a transitive group r ~n. The
group is called simply transitive if r = n; then there exists exactly one
transformation transforming P into Q.
3. Invariant contravariant vector fields. A contravariant vector fieldZ"
can be looked upon as an infinitesimal transformation ~"--;..~" +Z"dt
with the symbol Z def Z~' 81'. Hence invariance of infinitesimal trans-
formations and invariance of contravariant vector fields is essentially
the same problem. Besides the complete invariance defined in the
beginning of this section, there are less restricted forms of invariance.
We distinguish:
a) invariance to within an arbitrary factor;
b) invariance to within a constant factor;
c) complete invariance.
a) If a field Z" is subjected to the infinitesimal transformation xb
of a group, the increase of the field is equal to the negative LIE differential
with respect to the field e" E::
b

(5.16) dZ"=- dtfZ" = - e"(E~ 8~'Z" dt- Z~' 81'E:dt)

b b

or in another form, writing Z for the operator Z~' 81'

(5.17)

Instead of one vector field we now take a set of p independent vector

fields Z"; y = 1, ... , p and consider the case where the ~ Z" depend linearly
y b y
on the Z". This is a generalization of invariance to within an arbitrary
factor: Y
(5 .18) ~ Z"
b Y
= rxxby (~") Z"
X

or in another form, introducing the LIE operators ~ belongmg to the

y
fields Z"
y
(5.19)

for every scalar s. The equations (5.18, 19) express the fact that the
Ep-field spanned by the p vectors Z"y
is invariant. This Ep-field is
Xp-forming if and only if the equations Z~' 81' f = 0 form a complete
y
system (cf. II 5):
(5.20) (ZZ)
y Z
=;.(~")Z
X

1 ) Note that (5.19) in general does not hold if s is replaced by an arbitrary

non-scalar because the rt.:y
are not constant.
218 IV. LIE groups and linear connexions.

or in another form
(5.21) (fy f)
z
s = p;.
.
(~") fs
X'

and the equations of the Xp's are obtained by equating the solutions
of Z'"8'"
y
I= 0 to constants. If I is such a solution, then we get from (5 .19)
(5.22) ffl=ffl-tXty(~")fl
y b b y X

and this proves that fI is also a solution. Every solution of the com-
b
plete system is a function of n- p independent solutions, hence (cf.
II 10.38) the group transforms the set of all solutions into itself and
also the set of all oon-p Xp's (but not every Xp individually!).
If a group has an invariant complete system (that is an invariant
set of ocn-p Xp's forming a normal system) it is called imprimitive
and if not it is said to be primitive. If a group has an invariant Xp,
p < n, it can not be transitive, but an imprimitive group can be transitive.
b) 1 ) If in (5.18, 19) all coefficients are constants we have a generali-
zation of invariance to within a constant factor:
(5 .23) a) fZ"=
b Y
kt y Z"
X
or kty = const.
For p = 1 not only the streamlines of Z" are invariant but the field Z"
gets a constant factor. For general values of p the infinitesimal trans-
formation lz with constant coefficients l generates a one-parameter
y
group. These transfrmations tagether form a linear set (Schar) of
infinitesimal transformations. If the Z are independent (though possibly
y
connected) it can be proved 2) that this linear set always generates ooP
finite transformations. The statement is evidently true if the yZ belong
to one group (though not generating this group). In fact in this case
each one-parameter group is represented by a geodesie through r/'
0
and these geodesics form an Xp in 91 (11"'). If the Z are arbitrary we have
0 y
y

only to prove that in the transformations '~" = e sf ~" the tl are essential
1

parameters (cf. IV 3). Now if these parameters were not essential there
would exist equations of the form

(5.24) S) 0 X t S Z /:X X z eur;=r:p

0 =r:pX (t s-sevr;=r:p
y
ts Z /:1< Y z 1 1:><
!I

r;
elf x y

and this is only possible if ~ = 0 because the Z are supposed to be in-

Y
dependent.
l) LIE-ENGEL 1888, 1, p. 246ff.
2) LIE-ENGEL 1888, 1, p. 62.
 5. Invariantsofatransformation group in the Xn of the $". 219

If the infinitesimal transformations ~ Z are subjected to the infinites-

:v
imal transformation eb xb of the group, we have

(5.25)

This equation expresses that the infinitesimal transformation cha-

racterized by the constants ~ is transformed into the transformation
characterized by the constants l- Uk'bxdt. Hence, accordingto (li 10.38)
we have for the transformation of the l under a finite transformation
of the group
(5.26) f(t) = (e-tK)~s(O)
where K stands for the matrix of Jk: :v. From this we see that the
linear set of infinitesimal transformations lz is invariant for all finite
y
transformations of the group germ. A finite transformation of the one-
parameter group generated by ~ (0) Z has the form
y

(5.27)
with some parameter u. Now if a finite transformation of the group
is applied, the transformation (5.27) is transformed into

(5.28)
and this is a finite transformation of the group generated by ! (t)Z.
y y
Hence the one-parameter groups generated by s Z for all different values
y
of ~ are interchanged by the transformations of the group. This proves
that (5 .23) is the n. a. s. condition for the invariance of the linear set ! Z
y
for all transformations of the group and also for the invariance of the set
of all one-parameter subgroups generated by ~ Z for different values of
y y
the s.
If the Z belang to the given group and if they generate a group,
y
(5 .23) is the n. a. s. condition for this group to be an invariant subgroup.
We get as a corollary that a linear set generates a group if and only if
it is invariant for all its own infinitesimal transformations.
c) In the case of complete invariance there is no need to consider
p > 1 because each vector field is dealt with separately. The equations
are
(5 .29) a) f Z" = 0 or b) (f f) = 0
b b z

and this proves that the one-parameter group generated by Z is an

invariant subgroup of the group generated by the xb and z.
220 IV. LIE groups and linear connexions.

6. Invariants of a group in group space. 1)

At the end of IV 1 we have seen that the differential concomitants of
group space are the ordinary concomitants of c; <X and that accordingly
they have constant components with respect to the anholonomic Co-
ordinate systems (a) and (A). In IV 2 we have seen that the infinitesimal
transformation of the linear adjoint group corresponding to the trans-
formation Sb Xb (sb = COnst.) Of the given group is

(6.1) b ~=Sb
Sb E b =Sb Ea Oe!' cb ~.
ec c"a (Jea ' Eb.a def ec Cc'b'a.

This is a linear transformation in the variables ea and these variables

may be considered as affine coordinates in the tangent E, of rt.
0
We have mentioned already (cf. 2.42) the transformation of the eOt
b
into the eOt in a tangent E, of a point of group space as an example of
B
a finite transformation of the linear adjoint group. Now the invariance
of a quantity for a linear homogeneous vector transformation is the
same as the invariance of its components for the corresponding CO-
ordinate transformation. Moreover we know from (2.42) that the com-
ponents c;;,a areinvariant for every coordinate transformation belonging
to the linear adjoint group. Hence we know already that the quantity
c;i,a is invariant for vector transformations belonging to this group.
This can be checked in the following way. The transformation of a
contra- and a. covariant vector by (6.1) is
a) va-+ (A~- Si/ dt) vb
(6.2) {
b) Wb-+ (Ag+ s,;adt) wa
hence (cf. 1.28)

(6.3)

or in other words:
The concomitants of c;;,a are invariant for the transformations of the
linear adfoint group.
There is a set of concomitants, that is of great importance for the
theory:

(6.4)

1) LIE-ENGEL 1888, 1, p. 270ff.

 6. Invariants of a group in group space. 221

They are all invariant for cyclic permutations of the free indices.
This implies that gba is a symmetric tensor and that gcba is the sum of a
symmetric tensor and a trivector.
If a concomitant of c;i,a is transvected with one or more factors ea
and if this expression is equated to zero we get an equation that is in-
variant for the linear adjoint groilp. This was proved by CARTAN.
We prove this here for the transvection P~~cd ec ed where P~~cd is a
concomitant. According to (6.2) we have
(6.S) _!:____
dt
(Pab
cd
ec ed) = _ 5 e.a peb
. cd ec ed _ 5 e.b pae
. cd ec ed.
d
Hence, because -dt pabcd

= 0 we have
(6.6) 0=-Pab
cd sceed_pab
e . cd sdece
e
and this shows that if ea is a solution of the equation P~~ cd ec ed = 0,
then ea- Si/ eb dt is also a solution. This proves that the vectors ea
satisfying this equation are interchanged by the transformations of the
adjoint group (6.2) and that consequently an equation of this kind
represents a subspace in the tangent E, of 'YJrx that consists of straight
0
lines through 'YJa. and is invariant for transformations of this group.
0
To every point ea in the tangent E, of 'YJa. there corresponds the point
0
of group space with the normal coordinates 'YJa * ~ ea. Because the ea
and the 'YJa transform in the same way, to an Xp in E, consisting of
straight lines through o'YJa. there corresponds .an Xp in group space con-
sisting of geodesics through 'YJa.. Hence equations of the form considered
0
represent in group space subspaces that consist of geodesics through
YJa. and that are invariant for the adjoint group.
0
Now consider an Ep through the origin in E,. Tothis Ep corresponds
an Xp in group space consisting of ooP-l geodesics through 'YJa., each of
0
them representing a one-parameter subgroup. In normal coordinates
this Xp is represented by r- p homogeneaus linear equations. Let the
normal coordinates be chosen such that these equations are
(6.7) 'YJ~ = 0; ~. t), 3 = p + 1, ... ' :y
so that the 'YJm; m, n, lJ =I, ... , p can be used as coordinates in the Xp. An
Xp is said to be geodesie in X, if and only if every contravariant vector
in Xp remains in Xp if it is displaced parallel in a direction Iying in Xp.
Let va be the vector and d 'YJa the displacement, a=I, ... , :Y. Then we
have vt=O; drJ~=O; ~=P+l, ... ,r and
(6.8)
222 IV. LrE groups and linear connexions.

and dv~ must be zero. Hence

(6.9) Ttn=O; 'J,J,n=I, ... ,p; '"{.=P+l, ... ,r;
at every point of Xp and this condition is n. a. s. N ow making use of
the expansion (4.34) we see that
(6.10) cm(p.bcn)b~-o
- 0 0

is a necessary condition. But (6.10) expresses the fact that c~vbc~b' is

0 0
alternating in 'fJ, n and because the same expression is also alternating
in m, 'fJ, (6.10) expresses the fact that it is alternating in m, 'fJ, n. According
to (1.28) this is only possible if it is zero. Hence, (6.10) is equivalent to
(6.11) c .b c >- 0
0 mp 0 nb -

Now we will prove that (6.11) is not only necessary but also sufficient.
Every term of the expression (4.34) [cf. (4.32)] contains one factor Ce
and an odd number of factors P, for instance P 2 CeP 3 If this term is
written out we get for its contribution to 11 : rt
'Ylq, C. a, 'Ylq, c . a, c a, 'Ylq, c. . a, 'Ylq, c. . a, 1'1'' c . 1.
'I 0 qt(n 'I 0 iq,a,l 0 p)a, 'I 0 q,a,'I 'I q,a,'
0 q,a, 0

(6.12) \
a1 , ... , a5 =I, ... ,r;
n, 'fJ, ql' ... ' qs =I, ... ' P;
!=P+I, ... ,r.

But in this expression 'fJ 2 can be written instead of a2 according to

(6.11). Then by the same argument 'fJ 4 can be written instead of a4
and if this is done we see that the whole expression is zero because
of the index '"{. in the last factor. The reasoning is independent of the
number of factors P (if odd) and of the place of the factor Ce in the pro-
duct. Hence (6.10) has as a consequence that not only the first term in
the expansion of rtn
vanishes but also all other terms.
The Ep at 'Y)a corresponding to the XP (6.7) is spanned by the vec-
o
tors ea at 'Y)a. Taking the basis vectors of the system (a) at 17r:x suchthat
n o 0
they coincide with the basis vectors of (a), the one-parameter subgroups
(geodesics) constituting the Xp are generated by the linear set of in-
finitesimal transformations emam; m=1, ... , p. Writing out the ex-
pression (an(amap)); m,n,P=1, ... ,p; we get
(6.13)

Hence (6.11) is the n.a.s. condition that (a"(amap)) depends only on

al, ... , ap. 1)
1) CARTAN 1927, 1, p. 72ff.
 6. Invariants of a group in group space. 223

It may happen that (8n8m) already depends on 81 , ... , 8P only.

The n.a.s.condition isthat c~;,."'=O; m,n=1, ... ,p; x=p+1, ... ,r
or in another form that c~r;/ = 0. That means that the Xp represents
0
a subgroup with the structural constants c~;,.P; m, n, p = 1, ... , p. Of
course the XP remains geodesic, but in this special case it has another
remarkable geometric property. If the vector va, v' = 0 is displaced
(+ )- or (-)-parallel along the Xp we have
(6.14) dva = - rcab Vb drJ' ! C~i,a Vb drJ' = - T~n Vn dr;l' t Cp~a Vn dr;l';
hence, according to (6.9)
(6.15)

Now we use the expansion (4.35) [cL (4.32)] andremarkthat all terms
contain a factor Ce and an even nurober of factors P, for instance PCcP.
If this term is written out, its contribution to c;,~" is
.",q, C '
., oq,[n
a, c a, .",q,
o"]a, C 1
., oq,a,'
(6.16) {

n' \J' qt' q2 = I' ... 'P; r.=P+l, ... ,r.

In this expression a1 can be replaced by \)1 according to c~~' = 0 and
0
by the same argument a2 can be replaced by \) 2 and a3 by \)3 . But then
the whole expression is zero because of the index r_ in the last factor.
The reasoning is independent of the nurober of factors P (if even) and
of the place of the factor C c Hence the Xp represents a subgroup if and
only if it is not only geodesic, that is (0)-parallel in itself, but also ( +)-
parallel and (-)-parallel in itself.
If the vectors e" spanning the Ep at n" are subjected to the trans-
m 0
formations sb Eb of the linear adjoint group corresponding to the trans-
formation Sb Xb of the group, we get
(6.17) dea=- Si/ebdt = - scc;i/ebdt.
m m m
Hence c~;,.x=O c~~=O is the n.a.s. condition that the Ep is. in-
or
o
variant for those transformations of the linear adjoint group that cor-
respond to the transformation of the subgroup represented by Ep.
It may happen that not only (8n8m) but also (8b8m); b=1, ... , r,
depends on 81 , ... , 8P only. That means that the EP is invariant for
all transformations of the linear adjoint group. But we have already
seen in IV 5 that this means that 81 , ... , 8P generate an invariant sub-
group. In this case the Xp in Xr has yet another remarkable property.
In X, we form a ( + )-constant field va, v' = 0. This is always possible
224 IV. LrE groups and linear connexions.

+
because the Xp is (+)-parallel in itself. Then bv 4 = 0 and

(6.18)
hence
~vm = c~~m v" dr/
(6.19) {
bv~ = c~~~ v" dr/.
Now in the same way as above it can be proved that c~~~ = 0 implies
- 0
that also c~~~ = 0. Hence bv~ = 0 and this proves that an Xp through
rt" represents an invariant subgroup if and only if it is geodesie and (+)-
o
parallel and (-)-parallel in itself and if moreover every Xp that is ( +)-
parallel to it is also (-)-parallel to it.
This property is clear from the point of view of group theory. A
(+)-parallel displacement drags along a (- )-constant field (cf. IV 1) and
this is a transformation of the second parameter-group. Hence the Xp's
(+ )- or (-)-parallel to the one through 'YJ"" represent the left hand
0
and right hand co-sets (Nebenklassen) of the invariant subgroup.
Every set of equations linear homogeneaus in the ea, formed by
means of concomitants of c;;,a represents a flat manifold in E, that is
invariant for all transformations of the linear adjoint group. Hence
this manifold and the corresponding manifold in X, represent an in-
variant subgroup. The following invariant subgroups are most im-
portant (cf. 6.4):
a) the group of gb
(6.20)
If gb =l= 0 this group is an (r -1)-parameter group.
b) The centre
(6.21)
This is the group of those transformations of the group that commute
with all transformations of the group (cf. IV 2). The centreis always
abelian and it vanishes if the c-rank of c;;,a is equal to r, that is if the
linear adjoint group is an r-parameter group (cf. IV 2).
c) The first derived group, represented by the support of the a-domain
of ccba. If q is the a-rank of ccba the equation is
(6.22) cCtbt [a, c ageal-
. Cgbq -
0

Its infinitesimal transformations are those transformations of the group

that can be written as LIE brackets. This group is contained in the
 7. Properties of integrable groups. 225

group of g6 In fact
(6.23)
It may happen that a group is identical with its first derived group.
This is the case if and only if the a-rank of c; 6a is equal to r. The first
derived group of a subgroup is a subgroup of the first derived group of the
original group, as follows from the definition. Let X 1 , ... , Xm be in-
dependent infinitesimal transformations of an m-parameter invariant
subgroup. Then (X6 Xp); p = 1, ... , m belongs to this subgroup and
(XpXq); p, q=1, ... , m to its first derived group. Now in the identity
(6.24)
the last two terms belang to the first derived group of the subgroup
and this proves that the first derived group of an invariant subgroup is
an invariant subgroup of the group.
The first derived group of the first derived group is called the second
derived group, etc. All derived groups are invariant subgroups of the
group under consideration.
d) The group of gba
(6.25) gba ea = 0
is an (r- q)-parameter group if q is the rank of gba and vanishes if gba
has rank r.
e) The group of c;i,d gda viz. c;,;d gda ea=O obviously contains the
group of gba Notice that c;i,d gda is always a trivector. In fact
cdg - cdcf c-- cdcf
be cd c- cdcfc
(6.26) { cb da- cb ed fa - fa ec bd fa
= - 2g[bc]a = - 2g[abc]

because gabc is the sum of a trivector and a symmetrictensor.

7. Properties of integrable groups.

If we form the successive derived groups of a givm group it may
happen that the process stops because the derived group of the i-th
derived group is identical with this latter group. If this is not the case
the last derived group is zero and the last but one is abelian. The group
is then said to be integrable. Every abelian group is integrable and
the same holds for every 2-parameter group.
We prove the well known theorem:
An r-parameter group , is integrable if and only if it is possible
to find a sequence of groups 1 , ... , ,_ 1 , , such that p is a p-para-
meter group and an invariant subgroup of p+I; p = 1, ... , r -1.
Schauten, Ricci-Calculus, 2. Auf!. 15
226 IV. LIE groups and linear connexions.

Proof. Let , be integrable and let its q-th derived group @(qJ be
an s-parameter group and its (q -J)-th derived group @(q- 1 ) be a t-para-
meter group, t > s. Let XI, ... , X 5 be independent infinitesimal trans-
formations of @(q) and xl' ... ' xt independent infinitesimal transforma-
tions of @(q- 1 J. Then @(q) is the derived group of @(q- 1 ) and accordingly
c~ba= 0 for c, b = 1, ... , t; a =s+ 1, ... , t. Hence XI, ... , Xs+I generate
an (s + J)-parameter invariant subgroup of @(q- 1 ) and this subgroup
contains @(q) as an invariant subgroup. In the same way XI, ... , Xs+2
generate an (s + 2)-parameter invariant subgroup of @(q- 1 ) and this
subgroup contains the (s + J)-parameter subgroup as an invariant sub-
group. Proceeding in this way for all values of q we get at last a sequence
of groups @I, .. ,, such that p is always an invariant subgroup of
p+t provided that no derived groups occur between p and p+t.
Conversely let there be a sequence @I, ... , @,, such that p is
always an invariant subgroup of p+l If XI, ... , XP are independent
infinitesimal transformations of p and X 1 , ... , XP, XP+l such trans-
formations of p+I we have
(7.1) c, b = 1, ... , p
because p is an invariant subgroup of p+I But it follows from these
equations that the derived group of p+I is contained in p Hence
the first derived group of , is a subgroup of @,_I, the second is a
subgroup of @,_ 2 , and so on. This proves that it can never happen
that one of the derived groups is its own derived group and this implies
that the group is integrable.
If the sequence @I, ... , @, is known the coordinate system can be
chosen suchthat XI belongs to @I, XI and X 2 to @ 2 and so on. Then
we have the following simple equations for the integrable group
a) (XIX2) = Ci2 1 XI
(7.2) {
b)

l
or in a general formula
(Xp X_:_+t) = c;p~t q Xq; t > 0;
(7.3) p-1, ... ,r 1
q=1, ... ,p+t-1 <r.
Notice that the centre, the group of gba and the group of c~i,d gaa
are always integrable.l) The proof of these propositions is rather long
and makes use of the theory of elementary divisors applied to the matrix
ec c~i/. So we can not show this proof here though we need the inte-
grability of the group of gba in the next section.
1) Cf. EISENHART 1933, 1, p. 173.
 V. Imbedding and Curvature. - t. The X 1 in V,.. 227

8. Simple and semi-simple groups.

A group is called simple if it does not contain any invariant subgroup.
It is called semi-simple if it does not contain any integrable invariant
subgroup. Hence a simple group is also semi-simple. As the group of gba
is an integrable invariant subgroup (cf. IV 7) it follows that for a
semi-simple group gba has rank r. s a corollary we get that the group
space of a semi-simple group has a riemannian geometry. But this
condition is not only necessary but also sufficient for the group to be
semi-simple. We prove that the rank of gba is always <r if the group
contains an integrable invariant subgroup. If such a subgroup exists,
either it is abelian itself or one of its derived groups is abelian. This
latter abelian group is an invariant subgroup of the group in question.
Let X 1 , ... , X, be chosen in such a way that Xp; p = 1, ... , m generate
this abelian group. Then we have
(8.1} c;,qa=O; cj,i/=0; p,q=1, ... ,m; x =m+ 1, ... ,r

and consequently

and this proves that the rank of gba is <r.

V. Imbedding and Curvature.

1. The X1 in v...
In a V,. the length ds of d~". called the linear element of the V,., is
given by
(1.1} ds 2 = g)." d~A d~".
Hence a curve in Vn need not be a V1 because its linear element may
have a length zero at some point or at all points. But a real curve in
an ordinary vn is always a J.i .1)
1 ) In this chapter a V,. is always supposed tobe ordinary and only real subspaces
are considered. This means that multi-directions with a singular position with
respect to the nullcone do not occur. Cf. for the more general case EISENHART
1926, 1; EI 1935, 1; E II 1938, 2. Also GurcHARD 1928, 1; LENSE 1936, 1. There
is a vast Iiterature on isotropic subspaces, that are subspaces with isotropic tangent
spaces. We mention here only LENSE 1926, 1; 2; 3; 1928, 1; 1932, 1; 1936, 1; 2:
1939, 1; 2; 1940, 1; 2; SYNGE and Mc CoNNELL 1928, 2; SCHRENZEL 1929, 1 ;
PINL 1932, 1; 2; 1935, 1; 1936, 1; 1937, 1; 2; 1940, 1; 2; 3; 1947, 1; 1948, 1; 1949, 1;
ANDREOLl 1935, 1; BOMPIANI 1940, 1; 1941, 1; MOISIL 1940, 1; STRUBECKER
15*
228 V. Imbedding and Curvature.

If ~~ is a definitely chosen point of the curve and ~" an arbitrary

0
point of it, the length of the curve from ~" to ~ is s =
0
Jds and the
parametric equations of the curve can be written in the form

(1.2) ~" = !" (s).

The vector
(1.3) ."
1 =J
." def
=-
d;"
1 ds

is a unitvector because ds is the length of d~". The sense of j" depends

on the (arbitrary) choice of ~". The unitvector j" is only given at the
0
points of the curve but by the process of prolongation (cf. II 4) we
may make it possible for the operator 171' to be applied to fields of the v1.
Because j" is a unitvector, the vector f!'l71'j" is either non zero and per-
pendicular to j" or zero. If it is non zero and if j" is a unitvector with
2
the same direction and the same sense, we have

(1.4) f!' 17. j"

I' 1
= kj";
1 2
k > 0.
1

If it is zero the same equation is valid with k = 0. k is the first curvature

1 1
and k f" is the curvature vector 1 ) of the curve. The direction of i" is the
1 2 2
first normal of the curve and j" and f" span its osculating R 2 The sense
1 2
of j" is independent of the choice of ~". At a point where k = 0 the
2 0 1
tangent is said to be stationary (cf. III 2) and j" and the first normal
2
can not be found by an equation of the form (1.4). If k=f= 0 in the
1
neighbourhood of the point, anormal may be fixed by a limiting process
from one side or from the other side of the curve. If k = 0 at all
points, (1.4) passes into 1

(1.5)

and this is the equation of a geodesie with the affine parameter s (cf.
III 7). If k is not zero at all points, the vector j~' 171' j" is perpendicular
1 2
to j", hence it has the direction of
2
f" or there exists exactly one direc-
1

1941, 1; 1942, 1; 2; 1944, 1; THALER 1941, 1; CHERN 1942, 1; MoGHE 1943, 1;

WoNG 1945, 2; PETKANTSCHIN 1948, 1; WALKER 1950, 2; 3; 4; PATTERSON and
WALKER 1952, 3. Cf. for spaces with a fundamental tensor of rank <n Ch. III
p. 133, footnote 1.
1) Rrccr 1895, 1. p. 298 used the expression "curvatura geodetica".
 1. The X 1 in V,.. 229

tion perpendicular to i" and i" such that it lies in the R 2 spanned by i"
1 2 1
and this direction. Because
(1.6)

there exists a unitvector i" perpendicular to

3
i"
1
and j" such that
2

(1.7) i''V.i"=-kf"+kf"; k;;;;.O.

1'2 11 23 2

k is the second curvature, the direction of

2
i"
3
is the second normal and
f", i" and i" span the osculating R 3 The sense of i" depends on the choice
2 3 3
of ~ on the curve. If k = 0 at a point, the osculating R 2 is said to be
0 2
stationary (cf. III 2) at that point. In this case j"' and the second
3
normal can not be found by an equation of the form (1.7). lf k = 0 at only
2
one point, an osculating plane may be fixed at this point by a limiting
process from one side or the other. But if k = 0 at all points the process
2
stops. 1 ) lf k is not zero at all points we can go on. Let us suppose
2
that the process stops after the construction of the (m -1)-th normal j".
Then we have the equations m

(1.8) i~' V11 i" k i" + k i";

= - s-1 s= 1, ... ,m; k=O
0 '
k=O;
m
s s-1 s s+1

in which the curvatures k, ... , k are all > 0 at ordinary points. 2) lt

1 m-1
follows immediately from them that
(1.9) r v" i[", ... mi""'l = o
1

and this means that the osculating Rm is displaced parallel along the
curve. Fora curve in an R,. it means that the curve lies in its osculating
R".. According to (1.9) it is possible to construct along the curve n-m
fields of unitvectors i" , ... , i" so that these vectors are perpendicular
m+1 n
at each point to the osculating R". and to each other, and that
<5 i" = 0, ... , <5j" = 0. Now taking k = 0, ... , k = 0, we get the
m+1 n m+1 n
general equations, valid in all cases

__i__ i" = t V" i" k i" + k i";

= - s-1 s = 1, ... , n;
{ ds s s s-1 s s+1
(1.10)
k = k = k = ... = k = 0.
0 m m+1 n
-----
1) CARTAN 1927, 3 has investigated this case.
2) We always take the signs of all curvatures positive. But it may be done
in another way. Cf. v. n. WounE 1921, 1; BLASCHKE 1921, 1, p. 9; SYNGE 1924, 2;
EISENHART 1926, 1, p. 61.
230 V. lmbedding and Curvature.

These are known as the equations of FRENET. 1) In these formulae the

curvatures are all ~ 0 and the senses of j", j", ... are all fixed. But the
2 4
senses of j", (, ... are only fixed if ~" is chosen and the senses of j", j", ...
1 3 0 3 5
change if the sense of i" changes. From this it follows that the screwsense
+ 1)
1
of j", ... , j" is fixed by the curve and the sense of j", and when! n (n
1 n 1
is even, by the curve alone (cf. I 1). k, ... , k can be computed as
1 n-1
functions of s from the FRENET formulae if the curve is given in the
form (1.2). Conversely the curve is fixed if k, ... , k are given as func-
1 n-1
tions of s and if one point of the curve and the vectors j", ... , j" at that
1 n
point are given. Accordingly the equations expressing k, ... , k as func-
1 n-1
tions of s are called the natural equations of the curve. 2) The proof
can be given as follows. 3 ) The equations of FRENET can be considered
as a system of simultaneaus ordinary differential equations of the first
order with the unknowns ~", j", ... , j" and the independent variable s,
1 n
in which k, ... ' k are known functions of s and the
1 n
r;;. known func-
tions of the ~". If these functions are continuous and have continuous
derivatives, these equations have one and only one solution such that
for s = s the ~" take some prescribed values ~" and the j" take some
0 0 s
prescribed values satisfying the conditions j" i" = 15st. Then these
conditions are satisfied at all points of the cu~v~ because __!:__ j" j,. = 0
ds s t
is a consequence of the FRENET formulae. As a corollary we find that
for V,.= R,. a curve is determined by its natural equations to within
translations, rotations and reflexions.
If in an R,. the fundamental tensor g;.,. is transformed by multi-
plication with a constant real factor 1jc2 ; c > 0, the new unit tangent
vector at the curve is 'i"=cj". From (1.10) it follows that 'j"=cj" and
1 1 s s
'k = c k. Hence the proportians of the n -1 curvatures are invariant
s s
for this transformation of the fundamental tensor. As a corollary we
1 ) JoRDAN 1874,1; KHNE 1903,1; BLAScHKE 1920, 1; JuvET 1921, 1; 1925, 1;
STRUIK 1922, 1, p. 76; SvNGE 1924, 2; DuscHEK and MAYER 1930, 1, II, p. 62;
HLAVATY 1934, 2, p. 12; DAVIES 1938, 1. GOLAB gave 1945, 1 another form, gene-
ralizing a formula of LANCRET. lf instead of j" an arbitrary vector field is taken
along the curve we get a generalization of the FRENET formulae considered first
by MEYER 1911, 1. Cf. STRUIK 1922, 1, p. 77; BIANCHI 1922, 1 ; EISENHART 1926,1,
p. 105; Mc CoNNELL 1928, 4; 5; BoRTOLOTTI 1930, 4; KRAUs 1934, 1; WoNG 1940, 3.
2) There are many publications on curves in R,. satisfying relations between

the n-1 curvatures. Cf. for instance HAYDEN 1930, 1; SYPTAK 1932, 1; 1934, 1;
E II 1938, 2, p. 21; WONG 1940, 3; 1941, 1; 2.
3 ) HLAVATY 1934, 2, p. 16; E II 1938, 2, p. 16.
 231

get that a curve in Rn is fixed to within translations, rotations, re-

flexions and similarity transformations if the proportians of its curvatures
are given as functions of s.
If a h ypersurface
(1.11) F(~") = 0

is given, the points of intersection with the curve (1.2) are determined
by the solutions of
(1.12) 0 =W(s) <1~ F(f"(s)).
If s = s is a solution of (1.12) and if moreover
0
d ll> d"ll> { } d,"+lrp
(1.13) -d { s} = 0; ... ;-d,. s = 0 ; -u+J
s 0
d{ s} =!= 0
s 0 s 0

there is said to be a contact of order u at the point of intersection ~" =

0
r (s).
0
If (1.12) is satisfied identically in s the curve lies on the hypersurface.
If u = 1, the direction of the curve at ~" lies in the (n -1)-direction
0
of the hypersurface at that point. But if u = 2 the osculating R 2 does
not necessarily lie in this (n -1)-direction and a similar remark can be
made for all higher values of u.

2. The X 1 in W.. and Lw 1)

Curves in a space with a generallinear connexion can only be dealt
with in a satisfactory manner if it is possible to define in some invariant
way a parameter and a tangent vector. If no connexion is given in
a space it is not possible to define an invariant parameter and this is
the reason why for instance an xl in xn has no osculating 2-direction.
For an X1 in Wn [cf. III 3J2) the factor in g;,." can be fixed at a
point ~" of the X 1 . Then this fixed g;,." can be displaced parallel
0
along the curve and at each point a tangent vector f" can be constructed
having unit length with respect to this tensor field. The field f" is then
fixed to within a constant scalar factor by the connexion in Wn and the
curve. Taking f" = d~"jds and s = 0 for f' = ~" we get a parameter s
0
on the curve which is also fixed to within a constant scalar factor. This
parameter s can now be used to establish FRENET formulae in exactly the
same way as was described in the preceding section. From these formulae
we see then that all vectors f" and all curvatures k are fixed to within
s s
the same constant scalar factor and that consequently the proportians
1) Cf. HLAVATY 1935, 2, I.
2) HLAVATY 1928, 3.
232 V. Imbedding and Curvature.

of the k and also the vectors k i;. and k- 1 i" are invariants of the curve
s 1 s 1 s
and the connexion. 1 )
The case of an X 1 in Ln is much more complicated. 2) Let such a
curve be given by its parametric equation
(2.1) !;" = f"(t).
Then we may form the vectors

(2.2) v" def d!;" . ><def d!;".

etc.
1 - dt ' ~ =dtdt'

Let v" be the first vector that depends linearly on the preceding vectors.
P+l
Then it follows from (2.2) that
(2.))

where oc: is some determinable function of t. This equation expresses

the fact that the p-direction of v[", ... v"p] is covariant constant along
1 p
the curve. This p-direction is called the osculating p-direction of the
curve. If the Ln happens to be an En it follows that the curve lies in
an Ep.
N ow let s be another parameter on the curve

(2.4) t=t(s); t' -def !-!__

ds'
. t" def !!!...!._
= ds 2

Then we have
f u" def d !;" = t' v"

l
1 ds 1

u" def ~ d !;" = (t')z v" + t" v"

(2.5) 2 ds ds 2 1

~" ~ef (-)P-1 df;x = (t')P v"

p ds ds p
+
1) The centre of the osculating circle has the radiusvector k- 1 j" in the tangent
1 2
space.
Imbedding in Ln is a generalization of the imbedding in En dealt with in
2)

the so called "affine geometry" (Affingeometrie). There is a vast literature on

this subject. We mention here only some papers on the imbedding in An and Ln
(cf. Ch. VI 3 for projective imbedding): WEYL 1922, 3; R." K. 1924, 1, Ch. IV
(also for liter.); HLAVATY 1926, 4; 1934, 2 (also for literature); ]RNEFELT 1928,
1 (Ln); MAYER 1928, 1 (Ln); 1935, 2 (Ln); BoRTOLOTTI 1929, 3; 1931, 2; DIENES
1932, 1 (Ln); 1934, 1; GOLAB 1934, 1 (Ln); 2 (curves in Xm in Ln); WEISE 1935, 1;
EII1938, 2, p.21ff.; T.Y.THOMAS 1938, 3; MAEDA 1939, 1; WAGNER 1946,1
(variation Xm in Xn); Cossu 1947, 1; COTTON 1948, 1; CASTOLDI 1948, 1; RZEHNINA
1950, 1; KLINGENBERG 1951, 1; 2; 1952, 1.
 233

hence
(P+l)
(2.6) u[", ... u"vJ
1 p
= (t') 2 v["' ... v"PJ
1 p
and
(2.7) dds ~[,., ... rp] = ((P ~ 1) (t') (Pf) -1 t" + (t') (Pt 1) +1 IX) f["' ... g"p] .
This proves that u[", ... u"vJ is covariant constant along the curve if
1 p
and only if
(2.8)
or
(2.9) (p +
2
1) !!__!_ -
dt 2
!_!__
IXdt
=0

from which s can be solved in the form

(2.10)

containing two arbitrary constants. A parameter satisfying (2.10) is

called an afline parameter of the curve. It is determined to within
arbitrary linear transformations with constant coefficients. 1 ) If p = 1,
the curve is a geodesie and the affine parameter is the same as that
defined in III 7. If p = n, the two n-vectors in (2.6) can be replaced
by their corresponding L1-densities i1 and b:

(P+l)
(2.11) i1 = (t') 2 ti
hence

(2.12)

where ti is the L1-density belonging to t and i1 an arbitrary L1-density

of weight -1, that is covariant constant along the curve. In general ii is
determined to within a constant scalar factor. But in an L,. with a
volume preserving connexion, V,." = 0, scalar density fields exist which
are covariant constant over the whole space (cf. III 6). Hence if in
this case (f. i. in an En) such a field is given, this field can be taken for il
and then s is determined to within an additive constant s. The difference
0
of the values of s for two different points is the "affine length" (Affin-
lnge) known from affine geometry (in En). 2 ) If the coordinate system
1) HLAVATY 1934, 2.
2) Cf. footnote 2 on p. 232.
234 V. Imbedding and Curvature.

is chosen in such a way that E''"" corresponds to the covariant con-

stant field , the affine length between the points s and s takes the
form 0

t ll
(2.13) s- s = J (v[l ... v"l)n(n+l) dt
0 t, 1 n

and it is invariant for transformations of coordinates with LI = + 1. 1)

We retum now to the general L,.. Once the affine parameter is

known, other invariants may be deduced. Let ~" be a point of the curve
0
with s = 0. Then s is determined to within a constant scalar factor.
For simplicity we take p = n. Then

(2.14) _i_ u[x, u""l = 0

ds 1 n

and from this equation and (2.5) it follows that

(2.15) _i_ u" + " u" + + " u" = 0

ds n 1 n-1 n-11

where x, ... , " are functions of s. If s acquires a constant scalar factor c,

1 n-1
u" gets a factor c-i and "a factor c- v+ 1 l. In an E,. the vectors ~" and
J J J
the " are invariants for the special-affine group (LI=+ 1, cf. I 2).
J
The x's could be called special-affine curvatures of the curve. In fact
j
it could be proved that a curve in E,. is determined to within a special-
affine transformation if x, ... , " are given as functions of s. But there
1 n-1
is one serious difficulty. To each point of the curve there belongs the
invariant set of n vectors u", ... , u". N ow if these vectors are expressed
1 n
in terms of the derivatives of ~" with respect to an arbitrary parameter t,
it follows that these vectors may contain derivatives up to the orders
+
n, n 1, ... , 2n -1 and that in the x, ... , " derivatives up to the
1
+
n-1
orders n 2, ... , 2n may occur. But it is possible to find sets of n vec-
tors and a set of "curvatures" belonging to them, that depend only
on derivatives of lower order as was proved by WINTERNITZ for n = 3. 2)
Of course for a curve in L,. or E,. there are no equations similar
to the equations of FRENET in a Vn, because perpendicularity does
not exist in an En.
1) Cf. for E,. BERWALD 1923, 1, p. 95; BLASCHKE 1923, 1, p. 73.
2) Cf. BLASCHKE 1923, 1, p. 76ff.; HLAVATY 1929, 1; 1934, 2, p. 21 ff.; E IJ 1938,
2, p. 26.
 3. The X,._ 1 in A". 235

Exercise.
V 2,1. Let the tangent vector f' and a normal vector nA of a real
curve in an ordinary W,. satisfy the conditions
V 2,1 ~X) t,.. n,.. = 1; nA t,.. V,.. t" = 0.
Prove that the vector u,. = t,.. V,.. nA is perpendicular to f' and invariant
if f' and n,. are multiplied with a scalar as long as (V 2,1 ~X) remains
valid. u,. is the curvature vector of the curve in W,..

3. The .Xn-1 in An .1 )
Let an X,._ 1 in A,. be given by its nullform
{3.1)
or by its parametric form
(3.2) f:' = B" (r/'); m def ob f:'; a, b = 1, ... , n- 1
(cf. li 4.4, 12) and let /?' be a point of X,._ 1 . Then at each point of
0
the X,._ 1 in an 91(/?') we have the connecting quantities B6 and q
0
(cf. li 4). qis a covariant vector tangent to the x,._l and determined
to within a scalar factor, because here the upper index takes only one
value n.
Let the X,._ 1 be rigged (cf. li 5). To this end we introduce at
each point of X,._ 1 anE1 (in the tangent E,.) not lyirtg in the tangent E,._ 1
Then in the way described in I 4 and li 5 we have at each point
of the X,._ 1 in 91 (/?') the connecting quantities C~, B1; a = 1, ... , n -1.
c: is a contravariant vector in the direction of the rigging, determined
0

to within a scalar factor. As follows from I 4 and li 5 we have in

this special case, writing n" for c: and tA for q,
(3.3) a) Btt,_ = 0 b) B~n" = 0;
(3.4) a) B'bB~ = B~ b) I n,..t,.. =11;
(3.5) a) B~ B~ = B1 b) n"t.t = q;
(3.6) A1= B1+ n"t,_.
We call n" the pseudo-normal vector and t,_ the tangent vector of X,._ 1
and (3.4b) the first normalizing condition. If t,_ acquires a factor a, n" gets
a factor a- 1 but the quantities B~, Bt B}., B~ and C'J. do not change.
1) General references onX,._ 1 inA,.: ScHOUTEN 1923, 1; R. K. 1924, 1, p. 133ff.;
No!rsKA 1950, 1 and many papers dealing also with Xm in A,. and L,..
236 V. Imbedding and Curvature.

After the rigging every tensor with valence ~1 of Xn_ 1 has com-
ponents with greek indices and also components with latin indices
and every tensor with valence ~1 of An at a point of Xn_1 has an
Xn-rpart and anormal part (cf. I 4 and II 5). If the valence is 1,
the vector is the sum of its two parts:

{
a) + qv" = 'v" + "v"
v" = B~vA
(3.7)
b) wA = B~w,. + C~w,. = 'wA + "w,..
If p is a scalar field of Xn_ 1 , the expression V,. p = a,. p has no meaning
if the fieldisnot prolonged (cf. II 4). But BgJ7,.p=B;oAP=obp
has a meaning independent of any prolongation. It is the gradient of
p as a field of Xn_ 1 and B! ob p are its A,.-components.
If a vector field v" of X,._ 1 is prolonged we may define a covariant
differentiation in xn-1 by means of the formula
'r7 Va def P.a r7 v"
(3.8) Ye - C><Yp.

This means that 'J78 va is the Xn_ 1 -part of Jl"v". But 'Jlev" is independent
of the manner of prolongation as will be seen if we write out the right
band side of (3.8):
().9) 'V., v" = B~ B: (8" v" + J;..";. v1) = oe va + B~;: I;_.";. vb - B~ (oe B:) vb.
Moreover this equation proves that the connexion in A,. induces in
the rigged X 11 _ 1 a connexion with the parameters
'F.~ = B~~:J;..'i- B~ BeB:
(3.10) l -- B"la p.i. + Ba
cb>< F." " 8e B"
b

= B~~:J;..";. + B~8e8b~"
which are independent of the normalization of tJ.. Of course for this
connexion we have 'V., p = oe p. From the third line of (3.10) we see
that T[~bJ = 0, hence:
A rigged Xn-l in An is an An_ 1 whose connexion is uniquely deter-
mined by the connexion of the An and the rigging.
Besides 'F.ab other geometric objects are also induced in the A11 _ 1
The two tensors of A 11 _ 1

(3.11) I heb~- B~~ VI' t).l = - B~~ 01' a,. cn- B~~r;:,. t,.;
(3.12) 1- 1c -
.,.def _ l'av. n" I
cx JJ

get a factor a and e1- 1 respectively if t,. acquires a factor a. It follows

from (3.11) that heb is a symmetric tensor. The directions in the null-
 237

cone of hc b are called the principal tangents of the rigged or non rigged
X,._ 1 , and curves of X,._ 1 whose tangents are everywhere principal
are called asymptotic lines. The directions of the eigenvectors of l~"
are called the directions of the principal curvature of the rigged X,._ 1 ,
and curves of X,._ 1 whose tangent lies everywhere in a principal
direction are called lines of curvature.
In order to derive 'V" 'Vb v" we have first to take the A,.-components
B~~ V11 va of 'V" v" and to apply a process analogous to (3.8) to this
quantity of valence 2. Then we get

(3.13) 'Vt.~'Vcv" = B:t~:h.~'nPV11 v" + B:t~:t;" hpava + B~:V.V~'v"

hence
(3.14)
or

(3.15) '''"' =
'Rdcb vp).'"
dcb>< R"
~tAa + 2l '"' hc]b
[d

This is the generalized equation of GAuss for a rigged A,._ 1 in A,..

From (3.11) and (3.12) we get by differentiation and alternation

2'll[ahclb=- 2B[d~JtV,.B~~V11 ta
(3.16} 1 = 2B[d~Ji (V,. t;. na) V~' ta- 2B[d~Ji V. V~' t;.
= - 2Bfd hc]b na VI' ta + B;t~g R;;.;." f,.;

(3.17)

Now in (3.15) all terms are invariant if the normalization of t;. is

changed. But in (3.16, 17) only the last terms of each of the right hand
sides acquire a factor a and a- 1 respectively and the other terms change
in a more complicated way because a need not be a constant. This is
very inconvenient and we therefore introduce as a second condition of
normalization the equations

(3.18) b) Bf: (V~' n") t,. = 0 I

which are consequences of each other. 1 ) If (3.18) is satisfied, t" and n"
are fixed to within a constant scalar factor and in (3 .16, 17) the first
1) BERWALD 1922, 1, p. 164 for A,.=E,.; ScHOUTEN 1923, 1, p. 169.
238 V. Imbedding and Curvature.

term of each right hand side vanishes:

(3.19) 12 'Ji[a hc]b = B:l~t R;;.;_" t,. I

(3.20) 1
2' 17
l'[d c] -
.... n),
dcx Rv,_.),
za-- f.'a I
.

These are the generalized equations of ConAzzr for a rigged A,._ 1 in A,..
Moreover we have now from (3.11, 12, 18).
(3.21) 2l[Ja hc]a = 2B[dl:l~l!(V,.n") V,_.t;. = 2B[d~] (V,.n;.) J71,t;. = Bd~R;;.;_"t,.n;.
and by substituting this value in (3.15) and making use of (3.6)

(3.22) 'Ra-
aca - B"~-'
dc R.t
vp.J.

Hence, if both normalizing conditions are satisfied, and if the connexion

in A,. is volume preserving, the induced connexion in A"_ 1 has the
same property. 1)
We now have to investigate first whether it is possible to satisfy
(3.18) if the direction of n" is given. Let t;. and n" be normalized in any
way so as to satisfy (3.4b). Then, if Bf:(V'"' t;.) n;. =f= 0 we try to choose a
such that
(3.23)
This gives the following differential equation for a

(3.24)
and after some calculation we find for the integrability conditions of
this equation
(3.25)
This is equivalent to (3.22) because of (3.6) and (3.15). Hence:
If the direction of rigging be given, t;. and n" can be normalized accord-
ing to (3.4b) and (3.18) if and only if (3.22) holds
and as a corollary
If the connexion in A,. is volume preserving and the direction of n"
given, t;. and n" can be normalized according to (3.4b) and (3.18) if and
only if the connexion in A .. _ 1 is volume preserving as well.
We ask now whether it is always possible to find for a given nor-
malization of t;. a vector n" such that (3.4b) and (3.18) both hold. It
1) ScHaUTEN 1923, 1, p. 172. On p. 187 this is generalized for A". in A,..
2) Cf. EISENHART 1927, 1, p. 153 (55-9).
 3. The X,._ 1 in A,.. 239

follows from (3.18) that the direction of n" must lie in the support
of the -domain of B~ 17,. tJ.., hence the n. a. s. condition is that this
support is not wholly contained in the tangent E,._ 1 or in other words,
that tJ.. does not lie in the -domain of B~ 17,. tJ... Now if 'r is the rank
of heb the rank r of B~ 17,. tA is 'r or 'r+ 1 and it is 'r+ 1 if and only
if tJ.. lies in the -domain of B~ 17,. tJ... Hence for 'r = n -1 the vector t;.
can never lie in this domain, because r can never be greater than n -1.
Collecting results we get:
If the normalization of tJ.. be given, there exists a vector n" that satisfies
(3 .4 b) and (3 .18) if and only if tJ.. does not lie in the -domain of B~ 17,. t;..
If the rank of heb is equal to n -1, tJ.. never lies in this domain and n" is
uniquely determined and has the direction of the support of this domain. 1 )
If a vector n" satisfying (3.4b) and (3.18) has been found for a
given normalization of tJ.. and if this normalization is changed, tJ.. -+at;.,
the new value of the normal vector may be written in the form
a- 1 (n" + pj; P" t,. = 0. Then we have

(3.26) 0 = B~ (17,. a tJ..) (nJ.. + p;.) = B~ 17,. a +aB~ (17,. t;.) pJ..
or
(3.27)
Hence, if hba has rank n -1, the vector P" can be solved

(3.28)
and, as was to be foreseen, the new normal vector is uniquely deter-
mined. But if the rank of hba is < n -1, it may happen that the vector
ob log a does not lie in the b-domain of hb ". and in this case there is
no normal vector for the new normalization of tJ...
The quantity
(3.29) t;.def (n~l)l e..,.,., .....',. BJ.,a, ...... i.,.
a,. ~a, ... a..

is a covariant vector Ll-density of weight -1 in A,. and weight + 1

in X,._ 1 . lt is a concomitant of the X,._ 1 , independent of any rigging
and its (n-1)-direction is tangent to the X,._ 1 . Now if heb has rank
n - 1, the quantity
(3. 30) lj def / Det (hc b) j
is a scalar density of weight + 2 in X,._ 1 Hence, if a screwsense is
given in X,._ 1 and a scalar Ll-density q of weight + 1 in A,., at least
at all points of x,._l' then densities and Ll-densities in x,._l can be
1) Cf. ScHOUTEN 1923, 1, p. 173ff.
240 V. Imbedding and Curvature.

identified and a normalization of t;. can be fixed by the equation

(3.31) t;. = q g-i t;..
If the connexion in An happens to be volume preserving, the condition
V;. q= 0 determines a field q and therefore also the field t;. to within a
constant scalar factor, hence
For an Xn-l in a volume preserving An for which heb has rank n-1
at all points, special fields t;. and n" satisfying both normalization con-
ditions are determined to within a constant scalar factor and its reciprocal.
In an En the equations of GAuss and CoDAZZI take the very simple
form
(3.32) 2![;/ he]b = 'R,;ba
(3.33) 'r:(ahe]b = 0
(3.34) 'r:(al;t = 0.

These equations are the n. a. s. integrability conditions for the con-

struction of an En in which the An-l with a given volume preserving
connexion 'I'c't and given fields heb (of rank n -1) and z;a can be im-
bedded. Let the coordinate system ~" in the En to be constructed be
rectilinear, r;:;. = 0. Then the unknowns are ~", t;. and n" as functions
of the 'rJa and the equations for these unknowns are (cf. 3.10, 11, 12)
(3.35) a) ob~"=B'b b) t;.n;.=1 c) (8et;.)n;.=0 d) Bgt;.=O
(3.36) a) Bgoct;.=-heb b) oen"=-B~z;a c) oeB'b='T!i,B~+hebn".
Taking into account that the given connexion 'Teab is volume preserving
we get the following scheme for the integrability conditions of (3.35, 36):

3.35 a 3.36c
3.J5b 3.35c, d; 3.36b
3.35 c 3.36a; 3.32
(3.37) 3.35d 3.35b; 3.36a, c
3.36a 3.33; 3.35c
336b 3.34; 3.J5c; 3.36a
3.36c 332; 333
In this scheme the integrability conditions of the equation to the left
are identically satisfied as a consequence of the equation or equations
to the right on the same line. Hence the conditions are all identically
satisfied and this implies that there is one and only one solution of
rt
(3 3 5, 36) satisfying the initial conditions for = 'rja, heb= heb' z;a = z;a,
'T,a-'r,a.
cb- cb
0 0 0
0
 3. The Xn-l in An 241

a) f'=~" f) n;. Oe t;. = 0

0 0
b) t;. = t;. g) oe n" =- B" 0 a0
z;a
0

(3 -38) c) n"=n" h) oe B'b = T:b B; + heb n"

0 0 0 0 0
d) B'b=B'b i) t" n" = 1
0 0 0

e) Btoe t;. =-heb

0 0
j) Bt t"
0 0
= 0.

This means that the imbedding and rigging of an 91 ('~t) of an A,._ 1

0
with given fields 'l;'L heb (of rank n -1) and z~a, satisfying (3.32, 33, 34)
and 'R.;;~a = 0, inanEn is determined ifwe give 1. the point ~" correspond-
. 0
ing to rJ"; 2. the two vectors t;., n" at ~", satisfying (3.38i); 3. n-1
0 0 0 0
linearly independent vectors B;
0
in the E,._ 1 of t". If instead we give
0
in an arbitrary way the point *0~" and the vectors *0t .<, *n" and *B;
0 0
satisfying *t-< *n" = 1 and *Bt *t;. = 0, obviously there always exist
0 0 0 0
n 2 + n constants C~, c" such that
-1
(3-39) *0~" = q ~.<
0
+ c"; *t;. =
0
c~ t,.;
0
*n"
0
= C"n"
}. 0 '

If now ~", t;., n" are the solutions satisfying the old initial conditions
-1
+
it is clear that q; ~;. c", C~ t,. and C n:;. are the solutions satisfying
the new conditions. Hence the imbedding and rigging of the given
A,._ 1 is determined to within a non homogeneous affine transformation
in E,..
In an X,._ 1 in a general An the field hba is fixed to within an arbitrary
scalar factor. This means that, if hb a has rank n- 1, a conformal
geometry with a fundamental pseudotensor LhbaJ is induced in the
X,._ 1 . It also means that the asymptotic lines of an X,._ 1 in A,. are
independent of the rigging.
Exercise.
V 3,1. Prove for a rigged X,._ 1 in An satisfying (3.4b) and (3.18):
a) The covariant differential of t~. in an asymptotic direction con-
tains this direction.
b) The covariant differential of n" in a direction of principal cur-
vature lies in this direction.
c) For every line of curvature on X,._ 1 the covariant differential of
the bivector n["d~t.ljdt in the direction of the curve is simple and contains
the direction of n".
Schonten, Ricci-Calculus, 2. Auf!. 16
242 V. Imbedding and Curvature.

4. The V,._1 in V,.. 1), 2)

Let i;. be the tangent unitvector of a real Xn-l in an ordinary Vn.
n
Then the Xn_ 1 is a Vn_ 1 and the contravariant vector i" determines a
n
rigging. Because of

(4.1)

(4.2)

i;. and ni" satisfy the normalization conditions of

n
t;. and n" in the
preceding section. From now on we write n" for i". Note that the
n
normalization in An is quite different from that in Vn because in Vn
it depends entirely on the fundamental tensor g;.,.. In a Vn the difference
between heb and leb vanishes because
(4 . 3) he b
_ - - P'-17
ebv 11 n ;_. - - B";. 17
ebg;.,.v "_ 1
11 n - - g b a
Pal7 "-za 1gba
e,.v 11 n - e

1
gb a is the section of g;." and hence g;." = g;." - n;. n,.. The induced
1

connexwn is Riemannian because

(4.4) 117 I -11A><17

Ve gba- ebaVp
I
g;.,.--
- /JA><I7 -0
ebav#n;.n,.-

The equations of GAuss and ConAzzr take the form

(4.5)

(4.6) -
V[d hc]b-
2 117 PJ.K
deb v11 ;.,.n"

(4.7)

(4.8)
1 ) Cf. for a more elaborate treatment of the properties of a Vn_ in Vn deal1
1
with in V 4, 5: ErsENHART 1926, 1, Ch. IV; DuscHEK and MAYER 1930, 1; EIl
1938, 2, Ch. II.
2 ) General references on Vn_ 1 in Vn: BERWALD 1922, 1; 2; WEYL 1922, 3

FIALKOW 1938, 2; 3 (Rn and Sn); GHOSH 1938, 1; E I! 1938, 2, p. 5 7ff.; CARTA~
1939, 1; 2 Wn- 1 isopar. in Sn); COBURN 1940, 1 (EINSTEIN Vn _ 1 ); CoTTON 1940,1
MuT 1940, 2; MAEDA 1940, 1; URBAN 1947, 1 and 1948, 1 (V1 on umb. Vn_ 1)
VERBICKrl 1949, 1 (hyperquadrics); GOLAB 1949, 1 and GOLAB and WR6BEI
1951, 1 (curves in Vn_ 1 in Rn)
 4. The Vn-l in Vn. 243

hba is often called the second fundamemal tensor of the V,._ 1 in V,..
If a vn-1 is given in which the curvature tensor can be written in the
form (4.7), where heb is a symmetric tensor satisfying (4.8) it is always
a
possible to construct an Rn in which sufficiently small part of this
Vn_ 1 can be imbedded. For n = 3 this is a fundamental theorem of
classical differential geometry. T.he proof for an arbitrary n gives
another example of the significance of the GAuss-CoDAZZI equations
as integrability conditions. The unknowns are now ;" and n;. and the
system (") in Rn is supposed to be cartesian. Then the equations are

a) ab;"= B'b
b) n;. nA = 1 ; n" def g" An;.
(4.9) c) (oen;.) nA = 0
d) Btn;. = 0
e) Bt; g;,,. = ,gba; g;,,. det b;.,.

(4.10) {
a) Bt oe n;. = -heb
b) oe B'b = '{c't} B; +heb n"
and the integrability conditions are identically satisfied according to
the scheme (cf. scheme 3.37)

4.9a 4.10b
4.9b 4.9c, d
4.9c 4.10a, 4.7
(4.11) 4.9d 4.9b; 4.10a, b
4.9e 4.9d; 4.10b
4.10a 4.8; 4.9c
4.10b 4.7; 4.8.

Hence the imbedding is fixed if in Rn we give the point ;" corresponding

0
to rt, at this point a unitvector n" and in the Rn_ 1 of n;. a set of vectors
0 0 0
B;, ... , B~_ 1 having the same lengths and the same angles between
0 0
them as the basis vectors ea, ... , ea at rt in Vn_ 1 This proves that the
1 n-1 0
imbedding in Rn is fixed to within translations, rotations and reflexions.
If 'gba is given we know 'Kdeba and according to a wellknown theorem
of tensor calculus (cf. Exerc. I 8,9) the symmetric tensor heb is deter-
mined by (4.7) to within the sign if such an equation holds and the
rank of 'Kdeba is > 2. Hence a sufficiently small part of a V,._ 1 , n ~ 4,
whose curvature tensor can be written in the form (4.7) with a symmetric
field heb satisfying (4.8), can always be imbedded in an Rn and if the
16*
244 V. Imbedding and Curvature.

rank of 'Kacba is > 2 the imbedding is determined to within translations,

rotations and reflexions. For n = 3 the curvature tensor has rank 0 or 2
and can always be written in the form (4.7) but in this case this equa-
tion determines Det (hba) only and still leaves two degrees of freedom
for the symmetric tensor hba A transformation of a V,._1 in R.. pre-
serving the linear element is called a bending and the bending is said
to be genuine if it is not a combination of a translation and a rotation
or reflexotation. Using this term we are now able to state the theorem
of ScHuR:
A V,._ 1 in R,. can for n = 3 always suffer a genuine bending and for
n > 3 can do so if and only if hba has rank 0, 1 or 2. 1 )
In order to prove that the condition is sufficient it is necessary
to prove that for a rank < 3 the tensor hba can always be changed
continuously in such a way that the equations of GAuss and CoDAZZI
remain valid.
Exercise.
V 4,1. Let a V,._ 1 be given in a V,.. We choose the coordinate
system in such a way that !;" = 0 is the equation of the v,._l' that the
!;; cx = 1, ... , n -1, are coordinates in the V,._ 1 and that !;" is the distance
from a point to the V,._ 1 along a geodesie perpendicular to the V,._ 1 .
Prove that with respect to this special coordinate system
V 4,1 cx)

5. Congruences in v... 2)

The symmetric tensor h,.,_ of the preceding section is a special case

of a quantity occurring in the theory of congruences. Let i" be the
unitvector tangent to a real congruence in an ordinary V,.. Then,
because (J7,. i;.) i'" = 0 the covariant derivative of the field i;. can always
be written in the form
(5.1} V,.i,.=l,.;.+i,.u;.; i"l,.,_=O; l,.,_i;.=O,
where u;. is the curvature vector of the congruence (cf. V 1). Now
the n. a.s. condition for i;. to be V,._ 1-forming is (cf. II 5 or 7)
(5.2}
1 ) BEEZ 1874, 1; 1875,1 (inR,.); SCHUR 1886, 1; 2; 1887,1 (inR4); BOMPIANI
1914,1 (inR,.); KILLING 1885, 1; 1893,1. Cf. for Vn-1 inS,. CARTAN 1916, 1;
E 11 1938, 2, p. 143ff., also for literature. Special forms of bending of a surface in
R 3 leaving the principal axes of hba invariant aredealt with by CARTAN 1941, 2;
1942, 1; 1943, 1.
2 ) RICCI 1895, 1; RICCI and LEVI CIVITA 1901, 1; SCHOUTEN-STRUIK 1919, 1;

1921, 2; LEVI CIVITA 1925, 1; EISENHART 1926, 1; E 11 1938, 2, p. 27ff.

 5. Congruences in V". 245

or
(5-3) hp idef
.=-
z (pJ.)=- pi.
z
Now for every V"_ 1 of this set we have

(5.4) B1 =A1- i;. i",

and aeeordingly the symmetrie tensor
(5.5)

is the seeond fundamental tensor of these vn-1's.

Retuming to the general ease lcp;.J =l= 0 we may require the principal
directions of hf';.. One of these is the direetion of i", belanging to the
eigenvalue zero. The others with unitvectors i", ... , i" form n -1 eon-
1 n-1
gruenees perpendieular to i". They are determined uniquely if and
only if the eigenvalues of hf';. are all different. Rrccr 1) ealled them the
canonical congruences of i".

l(
If one of these eongruenees, for instanee i", is V,._ 1-normal (cf. li 5,
1
III 9)we have

Z
11
.
"f' 17pZ;.-Zp 17 . ) ).
VpZ;. Z
a
=-
2 Zp (17
1
. ) ).
V(pzi.) Z
a
+2 Z
p (17 . ) ,\
V[pZ] Z
1a
(5.6)
-- - 2 zp z,I v17<P z;.)
. -- 0., a -- 2, ... , n - 1
1 a

and this proves (cf. li 5) that

Every Vn_ 1-normal canonical congruence ol i" is tagether with i" V2-
lorming.
In other words: the equations i"' 8"' I = 0; i"' 8"' I = 0 form a eom-
1
plete system.
The eongruenee i" is geodesie if u" = 0. A geodesie eongruenee need
not be V"_rnormal as is wellknown from eongruenees of straight lines
in ordinary spaee. But if a eongruenee is geodesie and normal to one
V"_ 1 it is Vn_ 1 -normal throughout. 2) To prove this we start from i"' V~-' i" = 0
and vf' i;. = lf'' valid everywhere and vf' i;. = - hp valid at all points of
a V,._ 1 . Then we have everywhere (cf. III 4.9a)

1)Rrccr 1895, 1; cf. SCHOUTEN-STRUIK 1919, 1.

2)This theorem is due to BELTRAMI 1869, 1. Cf. DARBOUX 1889, 1; STRUIK
1922, 1, p.51; EisENHART 1926,1, p.57; EII1938, 2, p.45. The theorem in
its simplest form had already been found by GAuss, 1827, 1, 15, 16.
246 V. lmbedding and Curvature.

h ence puttmg pJ. = - l[I'J.l
kdef

(5.8)

from which it follows that k"J. is zero at all points if it is zero at the
points of one Vn_ 1 . This proves that iA is a gradient vector iA = 8Af
and from this i"o,.f=dfjds=l. Hence f=const. is the equation of
the normal V,._ 1's and any two of these Vn_ 1's cut off segments of the
same length from all curves of the congruence.
A set of n mutually perpendicular real congruences in an ordinary
V,. is called an orthogonal net. 1 ) I t is said to be an orthogonal system
if the congruences are all Vn_ 1 -norma1.2) Let i"=j"; j=1, ... ,n be
1 i
the tangent unitvectors. Hence we have for the n-th congruence of an
orthogonal system an equation of the form (cf. 5.1, 3)
n n
(5.9) V,.iJ. =- h"" + i"uA;

The orthogonal net is an orthogonal system if and only if each set of

two congruences is V2-forming. The n.a.s. condition is (cf. II 5)

(5 .10)

or (cf. III 9)
(5.11) h, i,j -=f=.

But because Ij~ * - Ijih this condition is satisfied if and only if

(5 .12) h, i,j-=f=.

This equation expresses the fact that every set of n -1 of the con-
gruences is canonical with respect to the remaining congruence.
Let us now suppose that the principal directions of h"A are uniquely
determined and that the congruences i"; b = 1, ... , n - 1 are chosen in
b a a
these principal directions. Then we have h"A * 2: haa i" iJ.. The n con-
a
1 ) The case of a Vn with an indefinite fundamental tensor is dealt with by
many authors. Cf. for instance EISENHART 1926, 1; 1949, 1'; E I 193 5, 1; E II
1938, 2; WoNG 1945, 2.
2 ) Cf. also for literature ScHOUTEN and STRUIK 1919, 1 ; R. K. 1924, 1, p. 190 ff.;

EISENHART 1923, 2; 1926, 1, p. 117ff.; SCHOUTEN 1927, 3; WALBERER 1934, 1;

E II 1938, 2, p. 27ff.; WoNG 1945, 2.
 5. Congruences in Vn . 247

gruences form an orthogonal system if and only if

(5.13)
a, b, c =f:; a, b, c = 1, ... ,n -1

since the remaining equation il' i"

c b
r)"
= 0 of (5 .12) is already satisfied
because of the special choice of the i". Now by differentiation of
b

(5.14)

we get

(5 .15)
b, c, d +;
and in these equations hbb =f: hcc. This proves the proposition:
A real congruence in an ordinary V,, whose canonical congruences
are uniquely determined, belongs to an orthogonal system of n congruences
if and only if

(5 .16)
J a) b, c, d = 1, ... , n - 1;
l b) b, c, d=f= 0

The geometrical meaning of these conditions is that the covariant

differential of hl'_, for a direction perpendicular to any set of m prin-
cipal directions of hl'_, has a component in the R.m of these directions
whose principal multidirections (cf. I 9) contain these principal direc-
tions.1)
If the principal directions of hl'_, are not uniquely determined results
can be obtained but they are rather complicated. 2 ) It can be proved
in the same way that the principal directions of a symmetric tensorfield
vn
J;," in in the special case that they are uniquely determined are
v;,_ 1-normal if and only if
(5 .17) j,i,h=1, ... ,n; j, i, h + 0
3)

1) LEVY 1870, 1 for v2 in Ra.

2) ScHOUTEN 1927, 3.
3) ScHaUTEN 1927, 3, p. 723ff. also for the general case.
248 V. Imbedding and Curvature.

The conditions (5.16, 17) can only be used if the principal directions
are already determined. ToNoLo 1) gave for n = 3 a condition for ~A
that only involves ~A and its adjoint (cf. Exerc. I 8,2) and ScHOUTEN 2)
gave the n.a.s. conditions
-1 -1
(5.18) a) 1[;" 17" 7;.1,. = 0 b) Tr;" 17" 7;.1,. = 0 c) Tt;" 17"T"1" = 0
for the case when T"" has n different eigenvalues none of which is
zero. The proof runs as follows. Multiplying (5.17) with the eigen-
value . and using the fact that i" is an eigenvector we get
h h

(5 .19) i"i"i" T"V. T><

j h V /'
= O> h,i,f=1, ... ,n; h, i, i =f=.
Alternation over f, i, h gives (5.18a). The other equations (5.18) are
derived in a similar way. Hence (5.18) is necessary. Conversely, using
the anholonomic system (h), (5.18) can be replaced by

-2 -1 -1
(5.20) Th. det
-
Th.1 Ti.

and if these equations are written out using the special values of the
components of h"" with respect to (h) we get

(5 .21) h, i, i =f=
which is equivalent to (5.17).
NrJENHurs') has given yet another form connected with the question
of whether a tensor T~,; in X,. with n different eigenvalues has covariant
eigenvectors which .re X,._ 1 -forming. For this problern without a
metric he introduces the new concomitant (cf. Exerc. II 2,3)

(5.22)

of the field Ti." and gives several forms of n. a. s. conditions. For in-
stance it is n. a. s. that H~~" can be written in the form

(5 .23)

1 ) ToNoLo 1941, 2; 1949, 1; 2; 3; 1953, 1; cf. ToNOLO 1941, 1. A Va whose

principal directions are V2-normal was already discussed by SLEBODZINSKI 1927, 1.
2 ) ScHaUTEN 1951, 2.

3 ) NIJENHUIS 1951, 1.
 5- Congruences in V,.. 249
g
where PJ., ... , PJ. are suitably chosen and where Ti'' represents the q-th
0 n-2
power of T,i ". Going back to the metric case he found other forms
for the conditions (5.18). The most elegant form is

Hr~.~ g,l,. = 0
l
a)
(5.24) b) HfJiJ. T,.1,. = 0
2
c) Hr;.i" T"1,. = 0
also suitable if one of. the eigenvalues of T,.J. is zero.

Exercises.
V 5, An ordinary V,. is an R,. if and only if there exists an ortho-
1 1 ).
gonal system of real geodesie congruences.
V 5,2 2). Areal geodesie congruence with unit tangent vector i" is
V,._ 1-normal if and only if the integral f i,.. d~,.. is zero for every closed
curve in some non singular V,._ 1 that has a point in common with each
curve of the congruence.
V 5,3 3). A V,._ 1 in R,. belongs to an orthogonal system in R,. if
and only if there exist n different principal directions of heb that form
an orthogonal system in v.. -1.
V 5,4 4). If the congruences with the unit tangent vectors i";
i = 1, ... , n, in V,. form an orthogonal system and if the field cx 1i" + i"
1

2
is v.. -1-normal, then the field - cx i" + i" is also V..-1-normal. cx and
1 2
need not be constants.
V 5,5 5 ). The curvature vector of the congruence of a field v" in V,.
that satisfies KILLING's equation 17(,.. vA> = 0, is a gradient vector.
V 5,6 6 ). Every congruence in V,. perpendicular to a field v" that
satisfies KILLING's equation, has a curvature vector perpendicular to v".
V 5,7 7). lf a field v" of V,. is v,._l-normal and satisfies KILLING's
equation and if the length of v" is constant, its congruence is geodesie
and normal to a set of (X} geodesie V"_ 1's.
1) E Il1938, 2, p. 31.
1) ScHOUTEN-STRUIK 1921, 2.
8) BRAUNER 1951, 1.

4) DEMOULIN 1913, 1 for R 3

5) KrLLING 1892, 1, p. 167; Rrccr 1898, 1; 2; 1901, 1, p. 608.
&) Rrccr 1898, 1 ; 2.
7) STRUIK 1922, 1, P 157-
250 V. lmbedding and Curvature.

6. Properties of curvature of a V,._1 in V,..

Let i,. be the tangent unitvector of a real congruence of V.. _ 1 . Every
curve of this congruence may be considered as a curve of V,. and as
a curve of Vn_ 1 . Writing 'ua for the curvature vector (cf. V 1) in V,._ 1
we get
+
(6.1) u,. def i~-' VI-' i,. = B~ ie 'Ve ia + i~-' (VI-' ie) ne n" = B~ 'ua ie ib heb n,..
We call u" the absolute curvature vector of the curve, 'ua the relative
curvature vector,
(6.2) "u,. def ie ib heb n,. = u><- 'u,. 1)

the enforced curvature vector and ie ib heb the enforced curvature. The
sign of this latter scalar depends of course on the choice of the sign
of n". The vector "u" is the same for all curves that have the same
tangent at the point considered. 2) It is equal to u" if and only if the
tangent of the curve, considered as a curve of the Vn_ 1 , is stationary
(cf. VI 1) at the point and it is equal to u,. all along the curve if and only
if the curve is a geodesie of V.. _ 1 . This proves that the curves of Vn_ 1
through one of its points, looked upon as curves of the vn' have extreme
values of the enforced curvature if and only if the tangent lies in a
principal direction of heb at that point. These directions are therefore
called the directions of principal curvature and the corresponding values
of ie ib heb are called the principal curvatures of the Vn_ 1 . The scalar

(6.3) h 1_
def _ _ 'gba h
n -1 ba
is called the mean curvature. Principal curvatures and mean curvature
are differential invariants of the vn-1
A curve of V,,_ 1 whose tangent lies everywhere in a direction of
principal curvature is said to be a line of curvature. Hence the lines
of curvature belong to canonical congruences of the congruences normal
to Vn_ 1 . As a result of this the Vn_ 1 ' s of an orthogonal system intersect
in lines of curvature. This is the theorem of DuPIN.
If ie ib heb = 0 at some point, the direction of ia is called a principal
tangent and the curves of Vn_ 1 whose tangents are everywhere principal
are called asymptotic lines. 3 )
1 ) If instead of the direction of n>< the direction of an arbitrary congruence

given over Vn_ 1 is taken, we get instead of the relative curvature the union curvature
of SPRINGER 1950, 1 and instead of the geodesics in Vn_ 1 the union curves with
respect to this congruence. Cf. YANO 1948, 1 also for .a generalization in An;
MISHRA 1952, 1.
2 ) This is the quintessence of the theorem of MEUSNIER (n = 3).
3 ) KRONECKER 1869, 1 ; Voss 1880, 1. Cf. for asymptotic lines of higher order
STRUIK 1922, 1, p. 80; E li 1938, 2, p. 73 ff.; Go LAB 1949, 1.
 6. Properlies of curvature of a V,._ 1 in V,.. 251

If heb has rank r, n-1-r prineipal eurvatures are zero and the
eone of principal tangents eonsists of oo' R,._,_ 1's all eontaining the
(n-1-r)-direction belanging to these eurvatures. If heb=O at a
point, every eurve of V,._ 1 through this point has an enforeed eurvature
zero at that point. The V,._ 1 is then ealled geodesie at that point. A
geodesie V,._ 1 is a V,._ 1 that is geodesie at all its points. Every geodesie
of a geodesie V,._ 1 is also a geodesie of V,. and this property is eharacteristie
for a geodesie vn-1
If heb equals 'geb to within a sealar factor

(6.4)
the point is ealled umbilieal. At an umbilieal point all prineipal eur-
vatures are equal and all direetions are directions of principal eur-
vature.1) If all points of V,._ 1 are umbilieal it is ealled umbilieal. It
follows from (4.6) that for an umbilieal v,._1

(6.5)

Henee, if umbilieal V,._ 1's are possible through every point of V,. and
with every (n -1)-direction at that point, all eomponents K,. 1,,. with
respect to any orthogonal (in general anholonomie) eoordinate system (h)
with four different indices are zero. In the next ehapter we shall prove
that this eondition is also suffieient. 2) If the V,._ 1 is not only umbilieal
but if moreover the mean eurvature h is eonstant, the left hand side of
(6.5) vanishes. Now B:i = A- n_. n", henee by eontraction of (6.5)

(6.6)

and this proves that the normal of every umbilieal V,._ 1 with eonstant
mean eurvature lies at eaeh point in a prineipal direetion of V,. [cf.
Exere. III 5.3 ]. 3 ) V,._ 1's of this kind through every point and with every
(n -1)-direetion at that point are only possible in an EINSTEIN spaee
beeause aeeording to (6.6) K,._. must equal g,.,_ to within a sealar factor.
But in 5 of the next ehapter it will be proved that this eondition
is not sufficient butthat the spaee must be an S,. [cf. III 5].') As a
eorollary we get that a geodesie V,,_ 1 is at eaeh point alwaysperpendieular
to a principal direction of V,. and that only in anS,. geodesie V,._ 1's are
1) Cf. on umbilical points T. Y. THOMAS 1938, 2; SASAKI 1939, 1; MuT 1940, 1;
WoNG 1943, 3 {EINSTEIN space); ADATI 1951, 1; 2.
1) ScHOUTEN 1921, 1, p. 86 for n > 3 (cf. footnote 1, p. 309).
8) Rrccr 1903, 1, p. 415 for h= 0; RrMrNr 1904, 1, p. 35 for n= 3; STRUIK
1922, 1, p. 143 for the general case.
4) SCHOUTEN 1921, 1, p. 87.
252 V. Imbedding and Curvature.

possible through every point and with every (n -1)-direction through

this point. 1 )
Some interesting propositions may be derived for V.. = 5.. from the
integrability equations (4.5, 6). According to (III 5.31) in an 5,. these
equations take the form
(6.7) 'Kdeba = - 2x 'g[d[b 'ge]a] -- 2h[d[b he]a]
(6.8) 'V[d he]b = 0
where " is the (constant) scalar curvature of the 5,.. If the v,._l is
geodesie it follows from (6.7) that it is an 5,._ 1 with the same constant
scalar curvature x. In that case (6.8) is trivial. If the V,._ 1 has only
umbilical points we have
(6.9) 'Kdeba= -2(x-t-h2 )'g[d[b ge]a] 1

hence the V,._ 1 must be an 5,._ 1 with the constant curvature x h2 +

and its mean curvature must be constant. From (6.7) it follows that
a V,._ 1 in 5,. can be an 5"_ 1 if and only if
(6.10)
For n:;;:; 4 this implies that the 5,._ 1 is umbilical and that heb= 'geb X
X V'x- x but for n = 3 the equation (6.10) only determines the
determinant of heb and leaves the possibility of bending (cf. V 4).
From (6.7) we get by contraction
(6.11)
from which it follows that the principal directions of heb are also prin-
cipal directions of 'Keb. Hence the directions of principal curvature of
a V,._ 1 in 5,. are principal directions of V,._ 1 2)
The quadric in the tangent Rn-t
(6.12)
is the indicatrix of DuPIN. Let (a) be an anholonomic orthogonal coor-
dinate system in V,._ 1 with unit basis vectors irx and let oc be the angles
b b
between an arbitrary congruence ia and the congruences ia. Then the
enforced curvature of the congruence ia is b

(6.13)
1 ) A V,. in which geodesie Vn_ 1's with every (n-1)-direction are possible

through one definite point is called a ScHUR space and the point is called centre.
DuscHEK and MAYER gave 1930, 1, p. 167ff. the form of the linear element in
normal coordinates with respect to the centre.
2 ) Cf. CoBURN 1940, 1.
 7. The rigged X;:' in L,. and X.,. 253

This is the theorem of EuLER (n = 3). The absolute value of the enforeed
eurvature for the direetion of ia equals the inverse quadrate of the
length of the radiusvector xa of DUPIN's indicatrix in that direetion.

Exercises.
V 6,1 1). Prove for a real eurve in a real V.,_ 1 m an ordinary V,.:
If the eurve is geodesie in vn-1 and a line of eurvature, then its
seeond eurvature in V.. is zero.
If the eurve is geodesie in v.. _1 and if its seeond eurvature in V.. is
zero, then it is a line of eurvature.
V 6,2 2). In a real v.. _1 in an ordinary V,. there are two real eon-
gruenees with unit tangent vectors i" and (. Prove that
V 6,2 oc) i" Vr7I' 1"" = i"'V./l 1." + ic1b hcb n" .

7. The rigged X:" in Ln and Xn. 3)

Let a rigged X': be given in an L.. and let the basis vectors e";
b
a, b, ... = 1, ... ,m span theE.,. oftheX';; and the e"; x, y, ... =m+ 1, ... ,n
"
span the En-m of the rigging X':'; m' = n- m at eaeh point. Then at
eaeh point we have the basis vectors ~; ~J.; h, i = 1, ... , n of the an-

holonomie eoordinate system (h) and the eonneeting quantities

{ m~l,fA:: Cf=ex~.~.*A!; B1= ~.~.e";

a B~ * ~~;
(7.1)
C"= ~ e"~A"
" b
B1= eae.l. *At C1=~.~.e"; c~~-~~-
:V "x- "' b X

1) E li 1938, 2, p. 72.
2) E l i 1938, 2, p. 66; cf. GRAUSTEIN 1934, 1; RICCI 1902, 2; AousT 1864, 1.
8) General references on the X': (cf. also li 5): ScHOUTEN 1923, 1 (X.,. in An);
1928, 1; 1929, 3; VRANCEANU 1926, 1; 2; 1927, 1; 2; 1928, 1; 2; 5; 1929, 1; 2;
1930, 1; 1931, 1; 1932, 1; 1934, 1; 1936, 1; 1938, 1; 1942, 2; HoRAK 1927, 1; 2;
1928, 1; SYNGE 1928, 1; HLAVATY 1930, 1; 2; 1934, 3; MOISIL 1930, 1; SCHOUTEN
and v. KAMPEN 1930, 2; BoRTOLOTTI 1931, 1; 2; 1936, 3 (also for literature);
1937,2 (X';; project.); 1941, 1 (also for literature); DIENES 1932, 1; AGOSTINELLI
1933,1 (V..); MORINAGA 1934, 1; VANDERSLICE 1934, 1; FABRICIUS-BJERRE 1936, 1
{torsionfree); WAGNER 1936, 1; ~938, 1; 2; 1940, 1 (non linear in E 3); 1941, 1
{X~-1); 1943, 2 (non linear inXn); PAuc 1937. 1; 1938,1; 2; YANO 1937, 1; 1938, 1;
1939, 1; YANO and PETRESCU 1940, 2; BOMPIANI 1938,1; MAXIA 1939,1; 2; 1940,1;
MIKAN 1939, 1; HAIMOVICI 1940, 1; 1946, 3; Su 1943, 1 (X;:-I project.); PAN-
TAZZI 1943, 1; 1947, 1; PETRESCU 1943, 1 (V;:- 1 ); 2; 3 Wlt+I
vHt~); 1944,1
(L';; in E.,); 1945, 1; 1948,1 (V;'); WANG 1943, 1; 2; ROZENFEL'D 1947,2 (X';; in
E ..); RoGovol 1949, 1 and 1950, 1 (X~ in E 3); DAVIES 1953, 1; TAKAsu 1953, 1
(in the large).
254 V. lmbedding and Curvature.

In order to consider questions of curvature we could proceed in

the same way as in the case of an Xn-l in An (cf. V 3) but then the for-
mulae would become crowded with factors B and C and would be
very unreadable. Therefore we introduce here the so-called D-symbolism
due to VAN DER WAERDEN and BoRTOLOTTI.l)
If P" = B; pa and q;. = B1 qb are vectors of X';;; r" = c; r and s_. =CIs,.
vectors of X';;' and v" and W;. general vectors of Xn, the working of the
operators D1., De and D, on these quantities and on scalars is defined
by the equations

(7.2)
1p D, ~ v, p ~ a, p
Dbp: B!V,.p: B:8,.p
D,. p - C,. V,. p- C,. 8,. p

(7.3) a)
r)
D,.p" = V,.P"
De P" = Br V,. P"
y) D. P" =
b)
r )
D,.q;. = V,.q;.
Deq;. = BrV,.q;.
y) D, q;. =

r
C~V,.p" C~ V,. q;.

D,.pa = B:V,.p" D,. qb = BtV,. q;.

(7.4) a) 1"1
) De pa = Br~ V,. P" b) ) De qb = Brgv,.q).
y) D,pa = B:V,.p" y) D,qb = C~ BtV,.qA

r r
C~

D,. r" =V,. r" D,.s;.=V,.sA

(7.5) a) ) Der" = Br V,. r" b) ) DesA = BrV,.sA
y) D, r" = C~ V,. r" y) D,s;. =

r r
C~V,.sA

D,.r = c:V,.r" D,. s,. = c; v,. s;.

(7.6) a) ) Der = Br c: V,. r" b) ) Des,.= Brc;v,.sA
y) D. r = c~: v,. r" y) D. s,. = c~; v,. s).

(7.7) a)
r D,.v"= V,.v"
) De v" =
y) D. v" =
B~

C~
V,. v"
V,. v"
b) 1"1)
y)
D,.wA=V,.w;.
De w;. = Br V,. wA
D.w;. = C~V,.w_.
1) V. D. W AERDEN 1927, 1; BORTOLOTTI 1927, 3. All formulae containing in-
dices a, b, ... or x, y, ... should of course be written with ~ because they are
meant to hold only for the special choice of the basis vectors f f
with respect
to the X:;' and x:;o'. Since this is understood we nearly always write in this sec-
tion =. Cf. for historical notes conceming the D-symbolism, ScHOUTEN and v. KAM-
PEN 1930, 2, p. 774. Cf. BOMPIANI 1952, 2.
 7. The rigged X;:' in L,. and X,.. 255

and by the rule of LEIBNIZ. Hence, if a quantity of higher valence

is given, for instance v~~'", lying with the index ~in X;:', with the index A.
in X;:'', and with the index,u in X,. but neither in X;:' nor in X;:'', there
exist full sets of components of four kinds, c.q. v:~'", v~~'", v:~'" and v~:'", 1 )
and we have the twelve derivatives

D. v".t =V. v"" )

rx D"v.a~" = Ba" V. v..
1
><
I" I" 1,

Dd V,></" = Bvd V. V,><

V
b) ) Dd va.l
I"
= Ba V. v".l
d>< v P
DU v".t
fl
= c V. v".tJ.l
U V
Y) DU va.t
I'
= c Ba V. v".t!-'
U X V
(7.8)
DV v""
P
=C"V.v""
V I"

Dd v""
I" = Bd C"V.
v v""
p

D U v""
P
= cx
U
V.V v""
p

From this we see that the following rules hold:

1. With respect to the index of differentiation, D'" corresponds to
17'"; Dc to B~ 17'" and D. to C~ 17'";
2. The only possible other indices that can occur in the derivative
are those that appear in possible components of the quantity to be
differentiated. For instance D.v~~'" or D.v:\ can not appear became
there are no full sets of components v:~'" or v~~ .
3. For an index a, b, ... or x, y, ... in the quantity tobe differentiated
a transvection with B or C respectively occurs and for an index x, A., ...
there is no such transvection.
4. In order to get the meaning of a formula with 17 or b we have
only to look at the skeleton (cf. I 11). But the meaning of a formula
with D depends also Oll the kinds of indices used. For instance 17.v~~'"'
17kv::1 , 17'"v~~i and 17kv~:i are all components of the same quantity,
the covariant derivative of v~~w But D.v~~'"' Ddv~~'"' Duv~:'" and
Du v~~P are four different quantities.
In (7.4a , y) and (7.4b , y) the covariant derivatives 171' P" and
17'" q,. are split up into parts that may be considered as a kind of covariant
derivatives themselves. The same holds for 17'" r" and 17'" sA in (7.6a , y)
and (7.6b , y). The (two) parts of 17'" p", 17'" q,., 17'" r,. and 17'" sA not
occuring in these formulae are no derivatives of these fields, because
for instance c~ 17'" PIJ = - pe 17'" c~. This leads to the investigation of
1 ) By a full set of components we mean a set that fix es the quantity completely.

For instance the components v~~c exist but they fix the quantity v~~'" only if
combined with the v~~.
 256 V. Imbedding and Curvature.

the derivative 171-' B1 = -171-' CA. If we introduce the quantitiesl)

m
(7.9) HpA... =def BT:a 17 B" ETa 17 C"
p.ti'T a=- pA~'T a=
BT C" 17
p a~'T
Ba. t = - BTp C"alfT
17 ca
A
m
(7.10) L p" .t= 1-'11 V.T BQ-
def BT" B..,. 17 C 11 - BT C 1117 B"- BTJ.' CP.tlf17.. C"II'
, t - - 1-'1/JfT ),-,..),V"~~--

m'
(7 .11) H 1-'
" def
= c..a
17
a- - - c..a
1-'J.IfT C" pJ.I'T B"
17
a-- CTp B" 17
alfT ca. \- - - 1-' B"
CT 17
alfT Ba.
>

m'
(7 .12) " def
L 1-' = CT>< 17 Cll-
1-'1/ Jf T - -
CT>< 17
1-'1/ Jf T
ll-
-
CT 1117 C " -
J.' Jf T {/ - -
CT 1117
1-' ), Jf T
B".
QI

we see that each of these tensors lies with one index or two indices
in X': and with the remaining index or indices in x:t. The full sets of
components
(7.13)
exist; this follows from (7.9-12). From these full sets we can see which
of the indices lie in X:' and which in X'::'. The covariant derivative 171-'B~
can be expressed in terms of these four quantities
m m m' m'
(7.14) V~-'B1 = H;;," + L;~.t- H;;,"- L;~A
The relationship A1 = B';. +CA shows that a quantity with valence 3
in general splits up into eight parts, but here four of these parts are
zero, for instance B~):~ VT B~.
m m
The quantities H;";." and L;~;. are called the first and second cur-
m'
vature tensor of valence 3 of X:'. 2) In the same way to X:'' belong H;;."
m'
and L;"i . The following formulae are useful
m m m
(7.15) a) H;bs =. c:n.B~ = -B~D.C~; b) H~b"=D.B~; c) H;J.s=-D.Cf
m m m
(7.16) a) L~~" = c;n.B~= -B~n.c;; b) L~~;. =D.B~; c) L~~"= -D.C~
m' m' m'
(7.17) a) H;;a=B:DzC~=-C~DzB:; b) H;;"=DzC~; c) H;;.a=-D.B~
m' m' m'
(7.18) a) L;:b = Bt D, Cf= -CfD.B~; b) L;:A=D,CX; c) L;~b= -D,B'b.

1 f ScHOUTEN
1923, 1 for Xm in A,..
For V,. the curvature tensor of valence 3 (cf.V 9) was introduced by ScHaUTEN
2)
and STRUIK 1921, 3 and at the sametime by BoMPIANI 1921, 1. Its components
correspond to the D, 1; of Voss 1880, 1, to the bcx 1 , . in RicCI 1903, 1 and to the k}1
in KORNE 1904, 1. Cf. ScHaUTEN and STRUIK 1922, 1; 2; STRUIK 1922, 1; BoM-
PIANI 1951, 3. Some authors call it the "EULER-ScHOUTEN" tensor which is not
correct from a historical point of view.
 ~ 7. The rigged X:;' in Ln and X 11 257

Using the anholonomic coordinate system (h) in L" we have the

following formulae for the contravariant vectors P" and r"
a) Dcpa _'I'+ I'cabpb;
ocpa -- B"
8c def c 8f.l

(7.19) 1 b)
c)
D. pa *_ 8. pa + F.~ Pb;
Dc rx * Oe rx + !;,~. yY
"
8zcie-1 C"z u!-l

d) D. rx * 8. rx + r.~ rY
and corresponding formulae for the covariant vectors q;., s;.. These
equations fix a linear connexion for quantities of X:;' and also a linear
connexion for quantities of X:;>'. We express this by writing L':, L>;:'
for X':, X:;>' from now on, and also A:;', A:;>' if the Ln happens to be an An.
In Ln we know that (cf. III 9-3)
(7.20)
hence the expressions
(7.21)
depend only on the rigging and on 5~;." and not on the symmetric
part !(~.<) of the connexion in Ln. In an An they depend only on the
rigging and not on the connexion in An.

l
From (7.15b) we get

llr;;,t = - Btt] w" c:n c~ = - B[ct]v" (l" ff") c~

B~i WriJJ.]) (8[1' lA] - s~;." l")

l
( 7.22 ) = - ex = - B~i ex
y y
=Bttq(-Qf;+ S;ih)
-
--
zm X
cb + Bicbi cxh S"ii h -* rx[cb]. zm .. X
cb
def
=
Bicbi cx nh *nx
h~~ji=~~cb.
m
and this implies that Z ;;, x is a tensorbelangin g to the given L': and that
it lies with two indices in L': and with one index in L:;>'. In II 9 we
have already proved that the X': is holonomic (that is Xm-forming) if and
m
only if Z ;;,x = 0. This n. a.s. condition can now be expressed in the form

(7.23) H
m .. X
[cb] -
Bii cx
cb h
sii.. h-- 0
m'
In the same way the tensor Z ;;.a belongs to L':' and this L':' is holonomic
if and only if
(7.24) _ zm'zy.. a -c1e1 Hm'[zy]
.. a _ Ci;
zy
Bah S"h
I'
= 0

m m' m
In an An the two quantities Z;i,x and Z;;.a are identical with -Hr;;,lx
m'
and -Hr;;.t.
Schouten, Ricci-Calculus, 2. Anfl. 17
258 V. Imbedding and Curvature.

m m'
Besides the tensors Z ;i,X and Z ;;. a four geometric objects can be
formed from ilf;, viz. '[j~b ~~ Q~b with the transformation (cf. II 9-3)
(7.25)
*
m
and aa d~* [2':
cy cy
with the transformation
m m
(7.26) na' * Acc'y'a
;!u!c'y'-:-
y a' na
~&cy-2
1 Acc'j/v 0y Aa'
c
m'
and the two corresponding objects "il;y and a;b. The reason why !Jf;
splits up into six separate geometric objects is of course that here only
transformations of (h) are allowed with A~ = 0; A~. = 0; A{ = 0; At.= 0.
From now on we write every one of these separate objects with its
own kernel letter.
If only one xm is given in Ln, only the Operator Dc is important
as long as only fields in Xm are considered. But the other operators
can be used also if the fields are prolonged in the neighbourhoods of
all points of the Xm. This is often convenient and the results as far
as fields of X", are concerned are independent of the manner of prolonga-
tion (cf. II 4). The vectors ae" may be chosen in such a way that they
are basis vectors of some holonomic coordinate system (a) in Xm. But
this is not necessary and it should only be done if it is convenient for
some purpose.
The equations concerning the curvature of the L';: in Ln can now
be obtained in a very simple way using the operators D. We have
first to compute the alternated second derivatives of quantities of Ln,
L':: and L';:':
a) D[v Dl'] p = - 5~/ 8Q p
b) D [v D1'] V ,. -- 1
2 " V "
Rvp - s 17
1< Q V Q v,.
c) Dc D"1 w,. = - t R~;.J.,. w"- S~;.e 17e w,.
m m m m'
d) D [v Dp] pa = (laR
2 b~<
.. "+La
vp [v. x'
HX
p]b - [v.jxj Lx
La p].b-

- itl~a fj~]b~ + ii;~l~la L~]:b) Pb- s~~Q DQ pa

m m m m'

(7.27 A) e) Dc D"l qb = - (t Bt~ R~~}.,. + Lc~~lxl H~Ji/- Lc~~lxl L~J:b --

m' m m' m'
- H:[vjxj.aHx+R
p]b
.aL.x)q
[v]x; p].b a-
S""eD
vp
q
Q b

f) D [v D!'] r x_(tc-<xR
- 2 Y" VI' .a Yl H!']a.x+L"a
.. " - I-[v.
m
[IYI Lx
p].a
m m m
+
m' m m' m'
+ Ll[vjy; p]a - [vjyj p].a - S""eD
rr aH .x H .aL. x )rY VI' Q
rx
m m m m'
g) Dc. Dl'] Sy = - (! c~~ R~~;."- Lc~~IYI ~;]~X+ Lc~~IYI L;.J:a +
m'
+ rr a H .x
LZ[v/yj
m
p]a -
m' m'
H[v:yl. a Lp].a
X ) s
x-
s.. e D
VI' Q y
s
 7. The rigged X;:' in Ln and Xn. 259

m
b) D[d Dc] v" = ~ B~~ R~~t VA- (B~~ S~/- Hr~l) ve v"
m
c) Drr1 DcJ w,~ = - l B;t~ R~~)." w"- (Bd~ S~~e- Hr;Je) Ve w,~
m m
d) DrdDcJ pa = (! Bd~g~R~~;_"- H[rilb x L;J~x) Pb- 1

m
- (B;t~ S~/- Hr~Je) De pa
(7.27B) m m
e) DrdDcJ qb =- (! B'd~g~R~~)." -- H(r1,i/ L;J~x) q a -
m
- ( B'd~ S~~e - Hr~Je) De qb
" + 11
m m
f) D [d D c] r X = (1:r fl
dc
c-tx
yx
R.
vp,l
rr XL a ) rY
[rlral c].y -
m
- (B"
dc svp.. e- H[dc]..Q) De rx

m'
( a) Dru D.l p = - ( C~~ S~~"- Hr;.;t) o" p
m'
b) D [u D z] V " = 1
~ cI'R"
uz vp,l V A - (C"~'S
uz ..
vp Q - u ..
11[uz] e)v v" Q

m'
C) D [u Dz] w, t--lC~'R"w
- ~ uz vp,l " -(C~'S
uz .. f!-R!!)J7w
vp [uz] Q ,t
m' m'
d) D [u D z] Pa-(1C''!'AaR"
- ~ uz b>< vp,l
+ rr
H[urx,
aL.x)pb
z].b -
m'
- (C~~ S~~e- Hr;.;/) De pa
m' m'
(7.27C) D [u Dz]qb --(1C~taR
-- 2 uz b>< vp,l "+RaL.x)
[u!xr z].b qa-

-
(C Vf.t
UZ
s.. e
Vf.t -
m'
.. Q) DQ qb
H [uz]
m'
f) D [u D z] r x = (1
2
cuzyx
A R"
I' X ."
vp,l -
rr a
H[u!y!
m'
L z].a
. X ) rY - ,
m
- (C~~ s~~e- H[;.;/) De yX
m' m'
g) D [u Dz ]s -y - - (~1 cuzy><
f.l A R .. " . :c ) s:c-
a L z],a
vp,l - Jl,[u:YI
X

m'
- uz SVf.t.e
(C" -
R[uz] e) DQ sy

Using the anholonomic system (h) and the equations (7.20) and
(7.28) (cf. II 9.4)
we find another expression for D[d Dc] pa
m
(7.29 ) { D[d DcJ pa = H(d~JY Dy pa + .
+ {o[d I'cjb +I[~ e 1
l'c]b + .Q~c Ij'b} Pb- S~/ B~~ De pa.
17*
260 V. Imbedding and Curvature.

Hence, if we write

(7.30) m adef2"
Ydcb = U[d
ra
c]b
+2na
1 [d[e[
rc]b
+2Qidc ra
jb

it follows from (7.29) that

m m
(7.31) 2D[d D c] pa=rapb-2z
dcb
.. yDy pa_95
dc ~
.. eDe pa
dc

and from (7.27 B d) that

(7.32)

In the same way we get for L';:'

(7.33)

and

(7.34)
m'.. , x _ Cv p x R" . " _ 2 u,
r uzy
m' x
m' , a L. I.
- uzy>< vp. Ll[u[y[ z],a

(7.32) and (7.34) are a kind of generalized equations of GAuss for L';:
and L';;' in Ln. But 'F and 'F' can not be considered as the true curvature
tensors of the induced connexion in L';: and L';:'. In fact there is some-
thing queer in (7.31). The left hand side depends only on the rigging
and the I;,~, that is on the displacement of vectors of the L';: in direc-
tions of the L';:. But none of the terms of the right hand side has this
property. Now because of (7.26) the equations

J a) :!lz pa d~t az pa + 2 ~baz Pb

m

(7.35)
l b)
1)
:!)z qb def Oz qb - 2 !:Jtz qa

define a covariant derivative of pa and qb depending only on the rigging

of the L';: and not on any connexion in Ln- As is to be expected this
derivative is intimately connected with the LIE derivative. In fact,
if v" is a vector field of L';:' it is easy to prove that

(7.36)
This means that the covariant differential (d~)Y :!)Y pa for a direction
lying in L';;' is zero if the L';:-part of the LIE differential vanishes
or in other words if the field P" dragged along over d ;" differs from the
original value at ~" + d~" only by a vector in L':::'.
1) For (7,35c, d) see p. 261.
2) Cf. VII 3.
 7. The rigged X:;' in Ln and X n- 261

An operator ~b operating on r" and sy may be defined in the same

way: m'
c) ~er"= Ocr"+ 2Q:crY
(7.35) { m'
d) ~csy = 8csy- 2Q;csx.
Note that the operation of ~Y on quantities of L';;' and of ~b on quantities
of L': is not yet defined. The relation between ~Y and Dy is the identity
m
(7.3 7) ~zpa = Dzpa + L;~,pc + 2S;;apc
and if this is substituted in (7.31) we get

+ Rdcbapb
m m
(7.38) 2D [d D c] pa--
- 2zY<!\
dc -'Vy pa- 2S
dc D e pa
where
m m m m
(7.39) Radefra+2ZY(Ja
dcb "- dcb dc -b.y -L2Sa)
by 1

is a quantity depending only on the displacement of vectors of L': in

directions of L': and on the rigging and not extra on any connexion
in L,.. This quantity is the true curvature tensor of the induced connexion
m'
in the L': and there is just such a quantity R;.;;," for the L>;:'. 1 ) If the
L': is holonomic, R
is identical with ;' as follows from (7.39). The true
generalizations of the equations of GAuss for L': and L':' in Ln are
(cf. 3.15)
m
.. a-
R dcb - BvpaR " -- 2H [dibl
dcb" vp). "Lc].x
a
m m
+
(7.40) m m (GAuss)
+ 2ZY(La
b.y -1-. 2Sa)
~ac by

m'
.. x-
R uzy - cvpXR
uzy" vp" - 2H [ulyl
.. aL.z].a
X
m'
+ m'

(7.41) m' (GAuss).

+ 2 z uz.. (L.y.b + 2 s yb..")
m'
b X

The following expression can be deduced for D[a Dcl r"

mm' m
(7.42) 2D[a Dc] r" = R;;" rY- 2Zj;Y Dy r"- 2 S; Der"
where
mm' m m
(7.43) R dcy
... xdef
~
Bv'"C;."R"+2H
dc Y" vp "Lb
[dlbl c].y (Rrccr) 2)

1) ScHOUTEN 1929, 3. p. 163f. also for historical notes; ScHOUTEN and v. KAM-

PEN 1930, 2, p. 777; E II 1938, 2, p. 161. The case of V:;' in Vn was treated earlier
by VRANCEANU 1926, 1; 2; 1928, 1; 3.
2 ) For Vm in Vn FABRICIUS-BJERRE 1934, 1 ; 1936, 1 called this tensor the
mm'
"torsion" of the Vm and discussed the case RJ; y" = 0.
262 V. lmbedding and Curvature.

and similarly
m'm m'
(7.44) 2D [u Dl p"-Rap
- uzb b-2zbD
uz b P"-2SYD
uz y P"

where
m'm
vpA + 2H.
m' m'
(7.45) R"
uzb def
. cP
uz BAa
b>< R" " LJ.b
[u!yl y (RICCI).

(7.43, 45) are the generalized equations of RICCI for L': and L':' in L,..
It is easily proved that the rule of LEIBNIZ holds for the operators
DrdDel and Dr,.D.1 (cf. III 4). Hence for a product v"qb we get from
(7.27B b, e)
D~d Dei v" qb = - !nrd.;b" v" q.. + ~ B'dt R;;.;." vA qb +
(7.46) (
+ Hr;_;1Y Dy v" qb- Bd,t 5;;. e De v,. qb
and thus

Transvection of this equation with B:

leads back to (7.32) but a new
identity arises from transvection with c; (cf. 3-19):
m m ~
. ""'--PAC"
2D [d H e]b - deb " 'R"
vpA - 2HYL"'
[de] y.b-
(7.48) m m' (CODAZZI)
- 25d.~"H~i/-l:- 2S;;.eB'd~qL;;b

and in the same way we get for L':':

~ ~ m
2D [u Hz]y""-
- CP
uzy B"" R"
vpA - 2H [uz]
b L""
b. y-
(7.49) m' m (CDAZZI).
- 25.;;"' H;;," + 2S;;.ec~~B~Lb~y
m
Differentiati on of L;~A gives, using (7.27 Ac, 27Bc, d)

(7.50)
D[d tl~A = Drd Dei Bt = !?pd~b" B~- t B'd~R;p,;." B: +
(
+ Hrd.~1 "' D.., Bt- B'd~ S;~e D(l Bt.
Transvection with B~ leads back to (7.32) but by transvection with C~
we get the new identity (cf. 3,20)
m
(7.51) 2D [d L"
m m' m (Co-
e]. y =-P"C;
de>< y.R"+2Z
vpA de"H"-2S
"y bL"
de b. y DAZZI)
 7. The rigged X;:' in L,. and X,.. 263

and in the same way for L':'

(Co-
DAZZI).
(7.48, 49, 51, 52) are the generalized equations of CoDAZZI for L': and
n In L ,.. 1)
L m'
A curve of L':, that is a curve whose direction at every point lies
in L':, is called a geodesie of L': if the covariant differential of its tangent
vector d~"jdt along the curve for the in L': induced connexion has the
direction of the tangent. Accordingly the n. a. s. condition that every
geodesie of L': is at the same time a geodesie of L,. is that the vector
d~P ( d~" d~") d~P ( d~A) d~P d~ m
(7.53) d! 17~-'dt-'17~-'dt =71 17~-'dt C).=dldiH;.;."
has the direction of d~"jdt for every choice of d~"jdt in the tangent E..,.
m
But from the definition of H;.;." we see that this is only possible if
m m
Hc~t = 0. If H(~( = 0 at a point, the L': is said to be geodesie at this
point. It is called geodesie if it is geodesie at all points of L,.. A geodesie
A", in E,. is of course always an E",. But a geodesie A': in E,. need not
have a curvature zero. Here is a very simple example in an Ra. The
null system of a system of forces in Ra fixes at every point the 2-direction
of the vector
(7.54)
where r" is the radiusvector, a,_ a constant vector and I",_ a constant
bivector. If p,_ = pi,_ where i,_ is a unitvector, we have
2 1 2 1
(7.55) a) H ;.;. " = - p 'lpA i"; b) L p.A
'" p 'I'"
= -- a ITa
p i >' 'fp). = BTp).

and accordingly (cf. 7.32, 39)

... a = p2-
2 'I [d jbj 'la
Ja) rzdcb c] ,
(7.56)
1 ... a
b) R2 dcb = pi2 'I[d[bj 'l'a
c]
+ p22 'Idc 'fa
b

The straight lines of the null system are also geodesics in the L~ but
'her r2 nor R2 vams
nelt 'hes.
If a rigged X': is given in X,., a connexion for quantities of the X':
only, can be fixed by giving the 'T.ab as functions of the coordinates ~-
1) The GAuss-ConAzzr equations for V.., in V,. were first given by Voss in
1880, 1, the Rrccr equations for vm in R,. by Rrccr 1888, 1. KHNE 1903. 1 had
already all equations for Vm in V,.. Then followed ScHOUTEN 1929, 3 for L': in
L,., ScHOUTEN and v. KAMPEN 1930, 2 for Vn"' in Vn and DIENES 1932, 1 for L;:'
in Ln. Cf. also for Iiterature STRUIK 1922, 1, p. 136; EISENHART 1926. 1, p. 192;
}RNEFELT 1928, 1; MAYER 1928, 1; 1935,2.
264 V. Imbedding and Curvature.

Then De pa (and Dc qb) can be defined just as in (7.19a) but with 'l;,'t
instead of l;,'t and besides these derivatives we have also the derivatives
'1:!. pa and '1:!. qb defined already in (7. 3 5 a, b). These derivatives tagether
fix a covariant differentiation for quantities of X': in every direction:

(7.57) f a) '1)e pa def De pa def Oe pa + 'J;,ab pb;

1b) '1)e qb clef Dc qb def Oe qb - 'J;,'t qa;
.
'1)z qb ~~ 0z qb - 2.Q~z qa
m

In the same way, if the "T.; were given as functions of the ~", the deri-
vatives (cf. 7.3 5c, d)

(7. 57) {
c) '1) r"
z
<!f D r"
z
def o r"
z
+ "F."z rY y '
'1)c r" def Oe r"
m'
+ 2Q~e rY
m'
d) -Vz Sy def
C'l'\
____:_ D z Sy def "
= ui Sy- "F."
zy s",. '1)e sy def Oe Sy - 2.Q~e sx

tagether would fix a covariant differentiation for quantities of X':' in

m m m
every direction. The quantities H~;." and L~ ~;. do not exist now but Hl~i,{
can be defined by (cf. 7.22)
m m
(7.58) H [eb]
.. x det _
~
z"
eb
without the term with s;;h because in xn there is no connexion and
no S;;h. For this connexion repeated differentiation leads to [cf. (7-38)]
m m
(7. 59) 2'1J[d '1Je] pa = - 2Zje" Y'1:!y pa + *Rj~i/ Pb - 2 ('F[de] + 1
.Q:le) De P"
with
+ 2' T'a[diel 'F.ee]b + 2'.Qede 'F.aeb + 4ZmdeY.Qamby
m
(7.60) *Rdeb
... a =def 2 U[d
-o 'F.a
e]b 1

and
(7.61)

l
with

(7.62)
*R:~ba def Ou'Fc'/, + 2oe Q~b + 2'J;,t Q:u + 2'T.'t Q~e +
m m' m
+ 2'Fe~.Q~b+4.Q~e.Q~y
A rigged X': in Xn provided in this way with a linear connexion
is called an L': in Xn and in the special case when 'Tc~bJ + '.Q~b = 0 it
is also called an A': in Xn.

Exercises.
V 7, 1. If a rigged X': is given in Ln and if we take for the connexion
in the X': the connexion induced from the Ln, prove that *R~i, a as
defined in (V 7.60) with respect to this induced connexion is identical
m
with R;;ba as defined in (V 7.39).
 8. The rigged Xm in An. 265

V 7,2. Prove that for a rigged L': in L"

m
V 7,2 oc) L;~Y ~~- f'c~.
V 7,3 1 ). Prove that for a rigged L';: in L"
m' m
+ L~~;, =
1n m'
V 7,3 oc) a) - H~~a D'" Bt b) - H~~x + L;,~;, = D'" q.
V 7,4 1 ). For every vector field v>< of a rigged L';: in L" the following
identities hold
m m'
a) c~ 1711- v>< = (H~). X-L~~;,) v"
{
b) B'"c cxx J71-' v>< = cb H x vb
m'
c) CPXJl
yx 1'-v X- - L"'
- b
y.bv.

m
V 7,5 2 ). Prove for a rigged Lm in L" that H[~bt is always zero if
the connexio n in L" is semi-sym metric.

8. The rigged Xm, in A,.. 3)

In A" is zero and therefore an X:;' in A" is holonomic, that is, it
S~;,x
m m
consists of oon-m Xm's if and only if (cf. II 9, V 7.22) H[;b]x = -Z;;,x= 0.
In this case the e>< in An can be chosen in such a way that they are
b
contrava riant basis vectors of a holonomic coordina te system (a) on
every Xm. If this has been done we have
(8.1) 'na
:o&cb =* e11- e;, 8 [p, e;,J
a
= ee ed 8 [e ea]
a
= 0.
c b c b
Of course the X';;' need not be holonomic.
Now we consider the case when only one rigged Xm is given in A".
Then this Xm is an Am (cf. V 7). Let the coordina te system (a) in Am
be holonomic and let the e>< be given over Am in an arbitrary way in
y
the m' -direction of the rigging. Then the e><, ~;, and the Ai, A1; h, i =
1, ... , n are known over Am and the same' holds for the f'cab, Fc'b, rcay
and f'c~:
(8.2) ~:i- c ix rx
r h -A'""h J.tA - A"i 8c AhA
The only derivativ es that are interestin g are those in a direction of Am.
m m
This means that only Dc will be used and that only H;i,X and L~~Y'
defined by (7.15, 16), will occur. The equation s (7.31), (7.32), (7.42)
1) E II 1938, 2, p. 157.
2) EI 1935, 1, p. 97.
3) Cf. footnote 2 on p. 232.
266 V. Imbedding and Curvature.

and (7.43) take the form

(8.3) 2D[d Dc] pa = 'Rd,;i,a Pb

(8.4) 'Ra_
dcb - Pa "- 2H
dcb>< Rvp m . ,x L
[dlbi
m a
c].x
I (GAuss)

mm'
(8.5) 2D[d Dc] rx = R,;;, x rY
mm' m m
(8.6) .. x =
R dcy ~'c"xR"
dc Y>< vp ' 2H [dlbl
--j--
. xL.b
c].J' (RICCI)

m
where 'R,~i,a is the curvature tensor of the Am which equals R,;i,a and

l
also ri:i,a because of (7.30) and (7.39). These equations are only valid
at points of the Am. The equations (7.48) and (7.51) take the form

(8.7) 2D [d Hc] b. x- - Vfl

d "b
C"" R"
v I';.

m (CDAZZI)

(8.8) 12D[d Lm.a

c]. y
-
- -
PaC"Rx
d c" y v I').
I
and also are only valid at points of the Am
All these equations can be expressed in terms of the tangent vec-
tors t;, , the rigging vectors e" and their derivatives. It follows from
y
(7.9, 10) that (cf. 3.11, 12)

(8.9)

(8.1 0)
m
" -
I~p.-- T><I7
I'QVT
X
exeQ- z"e ),,
-- x''
X
z" ctet- B'" V: .xeQ
x~' ----- ftfl r

hence the equations (7.40, 43) of GAuss and Rrccr can be written in
the form
X

(8.11) 'Ra-~'"aR"
dcb - dc b>< vp - 2h [dlbi'Xz.a
c] (cf. 3.15) (GAuss)

m~ x
(8.12) ... x*
R dcy ~
~'c.ixR"+2h
dc Y>< vp.l. [dlbiy c]
tb (Rrccr).

If we introduce the (m') 2 vectors of Am

(8.13) x dei
Vb=c B''b (17 x) e -
v1.,e;. _
- l'b (17 x *
vl'e}' ;.) e;.=- ]'X
bv
y )I

we have
mm'
(8.14) 2 '17 X
v'[dvc * R X
1 ~- .;~ + 2"v[dv]
z
y y z yc
 8. The rigged Xm in An. 267

and accordingly the equations of Rrccr can be written in the form

(8.15)

If (8.13) is substituted in (8.7, 8) we get another form for the equations

of ConAzzr

(8.16) (CODAZZI)

(8.17) 2 ' Y[d

17 za
y
dex RVJlA
e] = - l'a .. " e
y
-
2 vx [d z.a
,V X
e] I (CODAZZI).

The equations (8.11, 15, 16, 17) are especially interesting for the case
of an Am in En . Then they take the simple form

(8.18) I
'Ra
deb = -
z.a
2hx [d lbl xe] II (GAuss)

(8.19) (Rrccr)

(8.20) 'V(a he]b =

X V[d
X
y
Y he]b 1 (CoDAZZI).
(8.21) 117
y [d
za =
e] - X
V[d
za
e]
y y X
.______

We prove that an Am can be imbedded in an En if and only if there

X

exist, in this Am, 2m' +m' 2 fields heb z;a, 0b satisfying (8.18-21) and
X y y
h[eb] = 0. Let (x) be a rectilinear coordinate system in the En to be con-
structed. Then we have from (8.9, 10, 13)

c.n Bg =
X
(8.22) heb.>~<_ (8e - (8e Bg) q
(8.23)
y
z;a * (8e c;) B~ = - (8e B~) c;
(8.24) ve
y
* (8e q) C~ = - (8e C~) CI

and from 'Vewb_*_B~b"81'w;.

(8.25) r;,ab _>I<_ (8e B~) B~ = - (8e B~) B~.
From these equations we get the differential equations for the unknowns
B'b. c;
(8.26)

(8.27) <0
uc *
C"y -"= z a B"a -
yc yV,
X C"x
268 V. Imbedding and Curvature.

and similar equations for B~, q that do not need to be considered.

The integrability conditions of (8.26, 27) are

(8.28)

(8.29)

and these are satisfied identically because of (8.18-21). Hence there

exist solutions B'b, c; that take arbitrarily given values B'b, c; at some
0 0
point 'Y)a of Am. When the B'b have been found in this way the equations
0

(8-30)
X
are completely integrable because of the symmetry in bc of r"ab and heb
in (8.26). The ;" are then the rechlinear coordinates in the En that
were to be determined.
X X
In these considerations the number n and the fields hc b, Zc" a and J
y y c
are given beforehand. But there is another much deeper problem.
We may require a number n suchthat every Am can always be imbedded
(in the neighbourhood of one of its points) in an En. By counting
the number of equations and of variables for the case of a Vm in Rn
ScHLFLI 1871 suggested that the value n would be (m ;-
1 ). The rigorous

proof of ScHLFLI's suppositionwas given by JANET 1926 and CARTAN

1927. 1 ) The number n- m is called the class of the V",. 2 ) T. Y.
1 ) ScHLFLI 1871, 1; }ANET 1926,1; CARTAN 1927, 4; BuRsTIN 1931, 1; 1932, 1.

Cf. also WEISE 1934, 1 ; ToMPKINs 1941, 1. LAPTEV proved 1943, 1; 194 5, 1 that
every An can be imbedded in an EN; N~t (n 2 +2n-1). CHERN proved 1937, 1
that every projectively connected n-dimensional space can be imbedded in an
ordinary N-dimensional projective space with N~t(n 2 -1) +n dimensions. Cf.
for the projective case KANITANI 1947, 1; 1948, 1; 1949, 1; 1950, 2 and for the
conformal case RYZKOV 1950, 1; IcHINOHE 1951, 1. A general theory for imbedding
in an RN was given by RozENFEL'D 1945, 1. Cf. for imbedding in the !arge for
instance ALLENDOERFER 193 7, 1 ; 1939, 1.
2 ) There is much Iiterature on subspaces of special dasses and their properties.

We mention here on!y SCHOUTEN 1918, 2; VRANCEANU 1930, 2; WEISE 1934, 1;

T. Y. THOMAS 1936, 4 (dass 1, here is the definition of "type"); ALLENDOERFER
1937, 1; 2 (dass 1, EINSTEIN spaces); 1939, 1 (dass >1, rigidity); BURSTIN 1937, 1;
EisENHART 1937, 1 (dass >1); E II 1938, 2, p. 141; KocziAN 1939, 1 (dass 1);
LAURA 1940, 1 (dass 1); RosENSON 1940, 1 (dass 1); 1941, 1 (dass 1); 1943, 1
(class 1); PALATIN! 1941, 1 (class 1); MASINI VENTURELLI 1943, 1; 2 (dass 1);
SCHWARTZ 1946, 1 (second curvature); BLUM 1946, 2; 1947, 2; 3; MATSUMOTO
 269

THOMAs 1) has proved that for n- m;;;:;; 4 the equations of CoDAZZI are
consequences of the equations of GAuss. 2)

Exercise.
V 8,1. If A, is groupspace of an r-parameter group and if Xp is
geodesie in A,, prove that
(V 8,1 oc) Bdcb
v,_.;. C"a c
dc ca
eb
=0
(cf. IV 1.29 and IV 6.11 and use ConAZzr).

9. The V,:U and v_ in V.. and Xn. 3)

We consider a real X';: in an ordinary Vn. Such an X': is always
rigged by the m -direction perpendicular to the tangent m-direction and
1

these m -directions form an X':{ rigged by the X';:. The fundamental

1

tensor of the X': is, taking the anholonomic system (h) as in V 7

and 8,
(9.1) 1
gba = Bt:g;.,.
and from (7.9, 12) we get
(9.2) Dc gba = B cba
11 ;." 17 <1Q
;.,. g<1Q =
2Bcba
11 "'"
ga(>< VIPI Ba) = 2B"(a Hm !I b).a,ga., = 0
1 1
Yp

m'
(9 3) D ygba-
1 -
ba Vp <1Q
CPy B;." -
;.,.gaQ-- 2B"(a L. <1 1 -0
IYIb)ga.,-
1950, 1; 2; 3 (class 2); 1951, 1 (class 2); 2 (affine, class 1); 1952, 1 (project., class 1);
VERBICKII 1950, 1 (class 2); BRAUNER 1951, 1 (class 1, orth. syst.).
1) T. Y. THOMAS 1936. 4.
2 ) The GAuss-ConAzzr-Rrccr equations are not always independent. Cf.
MAYER 1935, 2; ALLENDOERFER 1939, 1; SCHWARTZ 1941, 1; BLUM 1946, 2; 1947, 2.
3 ) General references on imbedding in Vn, Rn and Sn (cf. VI 5 for conformal

imbedding): CARTAN 1919, 1 ; 1920, 3; 1939, 1; 2 (Rn and Sn); SCHOUTEN and
STRUIK 1921, 3; 1922, 1; 2; R. K. 1924, 1, Ch. V; BOMPIANI 1921, 1; 1924, 1
(geod. Vm in V,.); 1950, 1; 2 (V2 in Rn); 1951, 1; 2 (higher order curvature); 3 (V.. );
KMMERER 1922, 1; V. D. WAERDEN' 1927, 1; BORTOLOTTI 1927, 3; HAYDEN 1932, 1;
MAYER 1933, 1 (V,.); 1935, 2 (Rn); .FABRICIUS-BJERRE 1934, 1 (Sn); 1936, 1 (Rn);
KAPLAN 1934, 1 (Re); 2 (R,.H); 1936, 1 (Re); PEREPELKINE 1934, 1 (R,.); 1935, 1
Wnl; 1936, 1 (R,.); DuscHEK 1935, 1 Wm in R,., GAuss-BoNNET); DAVIES 1935. 1
(Vn); FUCHS 1935, 1 (geod. in V4 ); ANDERSON and !NGOLD 1936, 1 (R,.); ToMP
KINS 1939. 1 (flat isometric in R,.); 1941, 1 (R4 ); T. Y. THOMAS 1939, 3 (R,.); Co-
BURN 1939, 1 (S,.); 1940, 1 (S,.) ; WONG 1940, 1 (Vn); LENSE 1940, 1 (R,.) ; 1944, 1
(Rn); FINSLER 1940, 1; GHOSH 1940, 1 (Rn); 1941, 1 (R,.); WAGNER 1942,2 (anhol.
inS,.); CHERN 1944, 1 (R,.); WONG 1945, 1; CASTOLDI 1946, 1 (V,.); 1949, 1 (V,.);
LIBER 1947, 1 (Sm inS,.); PETRESCU 1948, 1 (anhol. in V,.); HLAVATY 1949, 1;
1952, 1 (W,.); YANENKO 1949, 1; 1953,1 (R,.); LOPSCHITZ 1950, 1 (EmH); 2 (centr.
Em+2l; ABRAMOV 1950, 1; 1951, 1; 2 (all topol. inv. V,.); TSUKI 1950, 2 and
1951, 1 (conformal).
270 V. Imbedding and Curvature.

m m
The difference between H~~" aEd L~~" vanishes because we can show
from (7.9, 10) that

m' m'
and it can be proved in the same way that L;'t =H;t.
For ~,. we get from (7.37) and (9.3)
m m m
(9.5) ~y'gba = Dy'gba- Li~/gca- L~~y'gbc = - 2H(ba)y
and from (7.35b) we get another form
m
(9.6) ~y'gba = Oy 'gba- 4.Q{b]y] 'ga)c
If (h) is an anholonomic orthogonal system with unit basis vectors we
have oy'gba * 0 and consequently in this special case
m
(9.7) C'r\
.Vy ' * 4.Qy(b ' ga)c
gba=
Similar formulae hold for the X';:', We call an x: rigged in this way
a v,.m in V,,. 1 )
Now the question arises whether a displacement of quantities of
x: in X,. can be fixed by giving only 'gba and a rigging. The parameters
'I;,a" of this displacement can be found from (9.2)
(9.8)
and (7.20)
(9.9)
m
We may proceed as in V 7, putting~cPa=Dcpa and ~.Pa=o,pa+2.Qba.pb.
Then we get for the parameters of this covariant differentiation
(9.10)
hence
m
(9.11) ~y'gba = oy'gba + 4.Q~(b 1 ga)c
in accordance with (9.6).
A rigged x: in X,. provided in this way with a metric linear connexion
m
is called a V,:" in X,.. Note that in this case H;;," does not exist because
m
there is no connexion in X,. but that H(bah is already known because
of (9.5). For repeated differentiation see V 7.
1) Cf. ScHOUTEN 1929, 3, p. 164, 167 ff.; BoRTOLOTTI 1930, 2; MorsiL 1940, 1 ;
VRANCEANU 1942, 1.
 271

We will now deduce some of the most important properties of a Vm

m v;,. 1 )
If i" is the tangent unitvector of a curve of vm we have
(9 .12) i"'l7I' i"=i"(l7I' ie)B"=i"l7
e I' i"+i"i;.l7I' C"=i"l7
;. P i"-i"i"H"
,.;.
m
writing H~ ~" instead of ~;;.", or
(9.13) ux = 'itx + ''ux
where u" is the absolute curvature vector, 'u" the relative curvature vector
and "u" the enforced curvature vector (cf. V 6). If "u" = 0, the direction
of i" is called a principal tangent and a curve whose tangent is principal
at all points is said to be an asymptotic line. 2 ) Now let ia be the unit
b
basis vectors of an orthogonal anholonomic system (a) of Vm. Then
we have
(9.14) i" = .E i" cos (X
b b b

where the oc are the angles between i" and the basis vectors. Hence
b
(cf. 6.13)
(9.15) "u" =.Ei" i" cos cx cos oc H "i"
cb c b c b I'

and this proves that, at every point, the vector "u" lies in an Rm,;
m 1 -s;;,n-m, m 1 -s;;,tm(m+1) in the tangent Rn, and perpendicular to
the tangent Rm. The Rm and Rm, determine an Rm+m, called the first
curvature region of the Vm at the point considered. 3 ) If m1 = tm (m + 1)
the vector "u" lies on a cone in the local Rm, and its endpoint describes
a Vm_ 1 of degree 2m-z which lies in an Rm,- 1 . This Vm_ 1 is the curvature
figure (Krmmungsgebilde) and its centre of gravity has the radius-
vector in tangent space
(9.16a) _ -1- 'gcb H"
M" - cb
m
In general the Vm_ 1 does not contain the origin, hence principal tangents
only occur in special cases. There is a one to one correspondence bet-
ween the oom-l directions in the vm at the point considered and the
points of the curvature figure. For m1 < tm (m + 1) there are several
subcases. 4)
1 ) Cf. for a more detailed treatmentEISENHART 1926, 1, Ch. IV, V; DuscHEK

and MAYER 1930, 1, Ch. VII; MAYER 1935, 2; E II 1938, 2, Ch. III, IV.
2 ) ScHOUTEN and STRUIK 1922, 1 ; 2; HA YDEN 1930, 2; E II 1938, 2, p. 89.
3 ) An osculating space of another kind was introduced by BERZOLARI 1897, 1

making use of an idea of KrLLING 1885, 1. Cf. E II 1938, 2, p. 90ff.; BoMPIANI

1950, 3.
4 ) Cf. for an elaborate treatment that deals also with other geometric figures

connected with the curvature and that contains many examples E Il 1938, 2,
p. 94ff.
272 V. lmbedding and Curvature.

If m1 = 1, 2 or 3 at some point of the Vm this point is called axial,

planar and spatial respectively. 1 )
If the enforced curvature vector is the same for every direction of
the Vm at a point, this point is said tobe umbilical. 2 ) N. a. s. conditions are

(9.16b) 0 = M"
cb -
def H"" _
cb
_!____
m
'gcb 'g:ta Hda"

The Vm is called umbilical [cf. V 6]3) if all its points are umbilical.
Every v1
is umbilical because H;;," =X je jb j". We prove that the
1 2

section of an umbilical Vp and an umbilical Vq is always umbilical. This

section is a V.; s?:, p + q- n. At a point of this section, to every direc-
tion of Vp there belongs a curvature vector and the projection of this
vector on the normal Rn-p is the enforced curvature vector that is
the same for all these directions. The same holds for the directions
of Vq. N ow let us consider the directions of V. and the proj ections of
the curvature vectors belanging to them on the normal Rn-s. If such
a projection is projected on the normal Rn-P and on the normal Rn-q
(both lying in Rn-s) we get the enforced curvature vectors belanging
to the direction of V., considered as direction of Vp and of Vq respectively.
Hence the enforced curvature vector of a direction of Vs is a vector
of Rn-s from which we know that its projection on Rn-p and on Rn-q
is the same for all directions of V.. Rn-p and Rn-q have an Rn-P-q+s
in common and n - p- q + s ?:, 0. Hence the enforced curvature vector
of a direction of V. is wholly determined by these two projections and
this proves that it is the same for all directions of Vs.
M" is the mean curvature vector of the Vm. For m = 1 it is identical
with the curvature vector of the curve. A Vm in Vn is called minimal
if for every part of it bounded by a closed vm-1' the m-dimensional
volume has an extreme value for variations that do not change the
boundary. Now let (a) be a holonomic coordinate system of Vm with
the coordinates rt. Then dr/ ... drr is the component of a contravariant
m-vector represented by an infinitesimal volume element. This com-
ponent is a scalar density of vm of weight -1. If (a') is an anholonomic
orthogonal coordinate system in Vm with unit basis vectors, the volume
of such an element is l(d17) 1 ' ... (d17)m'l Now 'g<!"!IDet('gball is a scalar
density of vm of weight + 2 and its value with respect to (a') is + 1.
Hence the volume is
(9.17) d Tm =def I{d 17 )1' (d 17)m'l --- Id 171 d 17m' g~~

1) Cf. for Vm's in Sn with planar points CoBuRN 1939, 1.

2) Cf. for Vn's admitting umbilical Vm's ADATI 1951, 1; 2.
3) Cf. E II 1938, 2, Ch. III; PEREPELKINE 1936, 1; WoNG 1943, 1; 3; 1946, 1.
 273

If the part of the vm considered undergoes the finite point transtorrna-

Iv
tion T belanging to the one-parameter group generated by the infinites-
imal point transformation ~"---+l;''+v"dt (cfo li 10) the change of the
volume can be computed most easily by performing afterwards the
-tv
inverse transformation T and dragging along the coordinate systems (x)
and (a) and the field g;." with this latter transformationo Then the Vm
returns to its original position and the changed volume we acquire is
the volume in this original position with respect to the field g;." dragged
along (cfo li 10)0 This is the method of transposiNan that will be used
often hereafter. Now we have (cfo III 5.48)

(9018)

(919)

and hence, according to (li 10038)

(9020) { g;." ~~ g;."- 2t II(;. v") + t2 v'" 17'" 1/(;. v") +

+ t2 (II('" v")) 17;. v'" + t2 (II(;. vJ")) 17" v'" - 0

Because both (x) and (a) are dragged along, Bi: is not changed and
conseq uen tl y

Hence forthat variation of the field gba at the point considered, which
1

is a consequence of the dragging along, we get

(922) d gba=- 2BA"

dt I
baV(J.v><)
17 f or t= 0

and accordingly (cfo Exerco I, 1,1)

(923) l _!,__ 1 g!i

dt
=

=-
!__ g!i _d_log g = _J_ g! 1gb a !L 1gb
2
1
dt
1

1g! 1gba Bg~ 17;. v"

2
1

=-
dt a
1g!t B~ 17;. v" 0

Now v" can always be split up into 1v" perpendicular to Vm and "v"
in Vm Then we get
0

(924)

Schouten, Ricci-Calcu]us, 2. Aufl. 18

274 V. lmbedding and Curvaturc.

and accordingly

(9.25) -di
d I d -r". ----
- - I "b "
117 vb d r". + m I 'v1, M"' d r"..
Tm Tm Tm

The first integral at the right hand side vanishes if "tl = 0 on the
boundary (cf. II 8). Hence the variation of the volume is zero for
allinfinitesimal point transformations v"dt for which the V".-part of v"
vanishes at the boundary if and only if the mean curvature vector is
zero at all points of the region considered. 1 ) For m = 1 this proves
that the variation of the length of a part of a curve in V,. vanishes for
allinfinitesimal transformations for which the component in the tangent
vanishes at the endpoints if and only if the curve is a geodesic. We can
express this in a less rigorous way: geodesics arenot only the straightest
but also the shortest (langest) curves. Whether the volume really has
an extreme value or not, can only be investigated by means of the
higher terms of the series (9.20). 2)

Exercises.
V 9,1 3 ). In a real V". in an ordinary v.. there are two congruences
with unit tangent vectors i" and f. Prove that
i"' V"'i" = i"' 'V"'i" + icib H;;,"
(cf. (9.12) and Exerc. V 6,2.)
V 9,2 4). The vectorial mean-value of the enforced curvature vec-
tors with respect to m mutually perpendicular directions of a real V".
in an ordinary V.. is independent of the choice of this set of directions
and equals the mean curvature vector.
V 9,3 6). A set of oo 1 real V.,_/s in an ordinary Vn contains only
minimal vn-l's if and only if the orthogonal trajectories make correspond
the points of the vn-l's in such a way that the (n -1)-dimensional
volume is invariant.
V 9,4 (cf. Exerc. VI 5,2) 6 ). According to HAANTJES and WRONA
the scalar curvature of a non singular m-direction at a point of a V,, is
the scalar curvature of a geodesie V". with this m-direction at this
1 ) LIPSCHITZ 1874, 2; EoMPIANI 1921, 1, p. 1141; STRUIK 1922, 1, p. 98; ScHou-
lEN and v. KAMPEN 1933, 1, p. 19 also for V;:'; E II 1938, 2, p. 165ff.
2) Minimal surfaces in an R 3 with an indefinite fundamental tensor are dealt

with by many authors. We mention here only the elder investigators LIE 1879. 1;
ElANCHI 1888, 1; CAYLEY 1890, 1 and Woons 1895, 1.
3 ) E II 1938, 2, p. 85.
4 ) E II 1938, 2, p. 87.
6 ) ElANCHI 1902, 2, Il, p. 577 for Va inR4; EOMPIANI 1921, 1, p. 1141 for vn-1

in vn.
6) HAANTJES and WRONA 1939, 1.
 s 10. Highcr curvatures of a V m in Vn. 275

point. If now n =2m, the V2 m is an EINSTEIN space if and only if

at every point two mutually perpendicular m-directions have always
the same curvature.
V 9,5 1 ). Prove that a geodesie Vm in an Sn (Rn) is an Sm (Rm)

10. Higher curvatures of a Vm in J~.

If i", j" and k" are the tangent unitvectors of three real congruences
in a real Vm in an ordinary Vn we have (cf. Exerc. V 9,1)

Jp V17f' Z." --
__ Jp '17 + (17 e) C"
V f' Z
." J
p
V f' Z Q
_
- J
p '17 ."
V f' Z
+ p <
J 1
Hp)."
(10.1) {

!
= i" 'V" i" + j" ib Dc B'b
kv V.V J'f' V.f' i" = kv '17: 'V. i" + kv (V.
V
Jp
f' V
Jp 'V. iV) C" + kv V. J'f' ii. H "
f' Q V pJ.
(10.2) =k'V. j" 'V" i"+ (k"j"'V" i<+i< k 'V" j"+j" k'V. i;) H~t+
+ kd j' ib Dd Dc B'b
and in the same way the p-th covariant derivative of i" with respect
to P directions of Vtn can be proved to be a vector lying in the X-region
of the set B'b, Db, B'b, ... , Dbp ... Db, B'b. This region is called the p-th
curvature region of Vm at the point considered. 2) Calling its dimension
m + m1 + + mp for all values of p for which mp=l= 0 we have the
inequalities
(10.3) mp ;;:;;tmP (m + 1);
The Vm is called asymptotic of order p if the p-th (and not the (p + 1)-th)
curvature region is contained in the tangent Rm. Then every curve of
Vm is an asymptotic line of order p, i.e. its osculating Rp+l is contained
in the tangent Rm. 4)
Since the Vm is real, none of the curvature regions has an extra-
ordinary position with respect to the nullcone. There is a number k
such that for each value of l;;:;;; 1 the (k + l)-th curvature region is con-
tained in the k-th curvature region. Then these regions determine
uniquely an Rm, = Rm, Rm,, ... , Rmk perpendicular to each other. For
the space containing all directions perpendicular to Rm,, ... , Rmk we
write Rmk+t; mk+l = n- m- m1 - - mk. The x-th curvature region
X ;;:;;k is spanned by Rm,, ... , Rmx. If the Vn is an Rn it follows immediately
from the foregoing that the Vm is imbedded in a flat Rn-mk+ in Rn.
1) E II 1938, 2, p. 133. Cf. p. 150 for generalizations.
2 ) Higher curvatures of curves in Vm in Vn are considered by DEL
PEzzo 1886, 1;
HAYDEN 1930, 2; 1934, 1; PEREPELKINE 1936, 1; ANDERSON and lNGOLD 1936, 1 (Rn)
3) An improvement was giverr by SCHWARTZ 1946, 1 ; cf. ALLENDOERFER 1939, 1.
') Cf. p. 250 footnote 3).
18*
276 V. Imbedding and Curvature.

Now we choose an anholonomic coordinate system with unit basis vec-

tors such that for each value of x; x = 0, ... , k + 1 the vectors i"; p.,
X ~
qx, ... = 1, ... , ~x span Rmx. Then if a quantity lies with one index
in Rmx this index can be greek or Px, qx, . . . . If we write gqxPx and B~~
for the fundamental tensor and the unity tensor in Rmx we have

(10.4)
f
gq,p, =
0
gq,p,; Po =Po
qo
0
qo'
X
gqxPx-
-B).qxPxgA>< ><

x y Px for
0, ... , k + 1
x=y
(10.5) " PY = ( X qx
X, y =
qx " 0 for x=f=y

and the D-symbolism can be extended with the operators D,x:

(10.6)
X
(10.7) D rx v" def
~
B"'rx V.11 v"'
X

J a) D I' rPx def Px V.I' r" '

0

-
(10.8) X

l b)
The curvature tensor H~;." appears in the form
1 0
(10.9) Hro.Po ~-= Df'o Po

and is called now the first curvature tensor of valence 3. Its u-region
coincides with Rm,. The second curvature tensor of valence 3 is defined by

(10.10)
Because
(10.11)
and

the u-region of H;,'q," is m 2-dimensional and perpendicular to Rm, and

Rm, and therefore coincides with Rm,. It can be proved in the same way
that the u-region of

(10.13) x=1, ... ,k;

coincides with Rmm and that this quantity, the (x 1)-th cutvaturt +
tensor of valence 3 of the vm
vanishes for X :;;;;; k. The formulae

1
X X x+ 1
(10.14a)
D ro B"qx -- -H"
ro. Qx + H ro Qx"
0 k+1 k+Z
ro.qo = O'
H" HroQk "= O' H;,q~.~ = 0; x=O, ... ,k;
 10. Higher curvatures of a Vm in Vn. 277

are a generalization for vm in vn 1 ) of the FRENET formulae (cf. V 1)

and become these formulae for m = 1. They can also be written in

l
the form
D BPx _ _ fL.r,.<.px +Hr,.Px
1
(10 14b) r, A - A
0 k+ 1 k+2
o' o' o' o' 1' ... ' k .
0

Hr, .<.p, = Hr, Pk

A
= Hr, Pk+
).
= x =

" lies with the second index in Rmx- and with the third
Note that H~~"
index in Rmx.
The equations (10.14a or b) can not be used to find out whether
a given Vm can be imbedded in an Rn or Sn because these equations
contain connecting quantities. The method that will be followed here,
using equations that only contain quantities of the Vm is due to BURSTIN
and MAYER and is developed hereintheinvariant form given by ScHau-
TEN and V. KAMPEN. 2)
The following quantities are formed from the "
H;,~"

(10.15)
X =2, ... ,k

where the indices b, c, b1 , ... , bx+l belong to a coordinate system (a)

in Vm. Then

X=1, ... ,k

are k quantities of Vm. It can be proved that a Vm in Sn is determined

X

by the quantities L to within rotations and reflexions. But the quan-

x
tities L can not be given at random 3 ), because they must satisfy certain
X

conditions. A method for building a possible set of quantities L was

given by ScHOUTEN and STRUIK. 4 )
Of course the theory of imbedding involving only the first cur-
vature, developed in V 8 for the rigged Am in An can also be used
for vm in vn If for this purpose we take quite another anholonomic
0

system (h), namely the system (a) orthogonal and anholonomic with
unit basis vectors i"; a, b = 1, ... , m, in Vm and fori"; X, y = m + 1, ... , n
b y

1 ) ScHOUTEN and v. 1931, 2, p. 1 51 ; E II 1938, 2, p. 120; cf. also

KAMPEN
WoNG 1940, 2; IWAMOTO 1944, 1; BoMPIANI 1951, 2.
2) BURSTIN and MAYER 1926, 1; MAYER 1928, 1; 1931, 1; 1935, 2; BURSTIN
1929, 1; DuscHEK and MAYER 1930, 1, II, ,p. 201-232; ScHOUTEN and v. KAMPEN
1931, 2, p. 153.
3 ) MAYER 1928, 1; 1935. 2.
4 ) E II 1938, 2, p. 125 ff., also for literature.
278 V. Imbedding and Curvature.

unitvectors mutually perpendicular and perpendicular to

m x m
vm' the
difference between H, h and L, %l as defined in V 7, 8 vanishes and
we get immediately from (8.9, 13)

(10.17)
%

a) heb~! -B:t VI' i;.;

X %

b) ~b def Bf: (VI' i;.) i;.; X, y=m+ 1, ... , n

and from (8.11, 15, 16)

= Bd~:; KI'H- 2f h[d[b hc]a]

X %
(10.18) 'Kdcba (GAuss)

(10.19)

(10.20) (CODAZZI).

For R., the terms withK;;;." drop out and for 5., the quantity K.~';."
has tobe replaced by - 2ug[v[A gl'l"l' Using the method of V 8 it can
be proved that a Vm in R., or 5., is determined to within translations,
%
rotations and reflexions by n- m symmetric tensor fields hob and
(n-;m) vector fields ffb provided that these quantities satisfy the equations
(10.18 to 20).2)
Another form of the GAuss-CoDAZZI-RICCI equations, involving also
higher curvatures can be derived in the following way. 3 ) Let the unit
basis vectors i"; x = 0, ... , k + 1 once more be chosen in Rmx. Then
qx
we have for any vector of Rmz

(10.21)

where (cf. III 9.3, 20)

(10.22)

1888, 1; Ktl'HNE gave 1903, 1 the whole set ( 10.18-20) for vm in V.,.
1) RICCI
ScHOUTEN and v. KAMPEN gave 1930, 2 the corresponding equations for V;:' in
Vn . An analogous set for L;:' in L., was given by DIENES 1932, 1. Cf. also J RNE-
FELT 1928, 1; MAYER 1928, 1; 1935, 2; E i l 1938,2, p.131.
2) E li 1938, 2, p. 121 ff. Cf. EISENHART 1926, 1, p. 192; LENSE 1940, 1; KAPLAN
1941, 1; YANO and MUTO 1942, 1 and YANO 1942, 2 (conformal).
3) ScHOUTEN and v. KAMPEN 1931, 2, 3; E li 1938, 2, p. 121ff.
 10. Higher curvatures of a Vm in Vn. 279
0
Of course K;o;oi/o is identical with the curvature tensor of the Vm,
viz. 'K;o;.i/o.
Differentiation of the FRENET formulae (10.14a) leads to

x=O, 1, ... , k-+1

or in another form

(10.24)

According to (10.14) the right hand side of this equation lies with the
index X in the space spanned by Rmx-' Rmx-I' Rmx' Rmx+l and Rmx+o .
Hence it can be split up into the following four equations 1)

0 x-g x
BvSof.o
BPx-g
x
Bi.qx ' '"
K'~~-tA = 0'' X = g, ... , k --l' 1 , g "-
~
3
0 x-2 x
X= 2 ' ... , k + 1
x-1 x
b) v BPx- Bi,
B Soro >< qx
" -_ 2 rz[so,
K'V u.Px- H' " .
"I ro],qx>

(10.25) c) X=1, ... ,k+1

X X X X
d) B;.qx KVA
" " = K ' ' .p"-2H' 1
Soroqx [so!.
Pxs;.
ro].qx
-
x+1 x+1
-
2 ll[s,,
U ' Px
;><I
' "
Hro]qx' X
=
0' ' ' ' ' k + 1

(10.25) are the GAUSS-CODAZZI-Rrcc r equations of a v';,. in vn for higher

curvatures. For Vm in Rn only three equations remain

:r-1 X
a) H [so. Px-
llx-11
H'ro] fx-I -- 0 '. X= 2, ... ,k
Qx -
X
b) D [so H'ro] Px-1
qx
= 0,' X =o 1, ... , k
(10.26)
X X x+1 x+1 X
KPx_2R' Pxfltx-1 +2RPx fltx+l
C) soroqx -- [sollx-II ro] , qx [so, ]lx+II ro]qx

X= 0, ... , k + 1.

I
In the equations (cf. 10.14a) in an Rn with rectilinear coordinates ~"
0
a) def Bb 8
8qo- qo b

(10.27) x x x-1 x+1 x+1

b) Dfo B" =
qx
- HYo Px-,
qx
B"
Px-1
Px+' B"Px+l'
-+-' Hro qx
X= 0, ... , k + 1;
1) The first equation is due to ScHWARTZ 1941, 1; cf. CuTLER 1931, 1; 2.
280 V. lmbedding and Curvature.
0 k+J
the e", Bq" , ... , 1Bq"k+1 mayk be considered as the unknowns and the
0

r.Poq
r0 0 , ,
r_Pk+t
roqk+<' Hr Po.q,, ... , H", Pk- 1qkasg1ven
0 0

funct"wnso f thecoordinat eS'YJ"
1
of Vm. Then (10.26) and Hr;oi.l1 = 0 are the integrability conditions. 1 )
If these conditions are satisfied identically by the given functions,
0 k+l
(10.27) always has a solution that takes initial values e", B;,
0 0
... , B;
0k+l

at an arbitrary point 'YJ" of Vm. But these initial values must satisfy
the conditions

l
0

a) g;.,. #:z ~Pz = bqzpz

(10.28) "';. y
b) g;.,.~qz!(f,.=O; x,y=O, ... ,k+1; x=j=y

where the gJ.,. are some given constants with gJ.,.=g,.J. andDet(gJ.,.) =l= 0.
0 k+1
Then siniilar equations are satisfied by B;,, ... , B;k+ at all points in
e"
1

an 9l (t). In the Rn of the with the fundamental tensor g;.,. the first
m + m1 + + mk vectors span a flat Rm+. +mk in which the Vm is im-
bedded. Hence:
lf in a Vm the F,~qo are the parameters of the connexion with respect
to an anholonomic coordinate system with unit basis vectors and if in this
V:mthe roq,,
P1 r.
,
rPk+1 d h Hl.ro.q,,
o qk+1an t e
Po /}. Pk-1 . - "' "'
,~...,.o. qk' p".,q".-1, ... ,m".,
X= 1, ... , k + 1 are given functions of the Coordinates 'YJ" of the Vm satisfy-
1
ing the conditions H[;,i/' = 0 and (10.26), in which the operator Ds, is
defined by
a) D 5 0 vPz def 0S 0 vPz r_Pz vqz +
(10.29) { S0 qz
X= 0, ... ,k 1 +
b) Ds, Wqz def Os, Wqz- r.~~z Wpz'
"'
and K;,;oi/z by (10.22), then an Rn; n=m+m1 ++mk+ 1 can be con-
structed andin this Rn a flat Rm+ +mk' in which the Vm is imbedded in
such a way that its x-th curvature region is an Rm+m,+. +mz; x = 1, ... , k,
can also' be constructed.
If for a Vm in R,. we have m1 = 1, l ~1 and if ftx is a unitvector in
I
Rm, the quantity H;0 p;_," can be written in the form
I I 1
(10.30) Hr 0 P1-1
"-H
- roPI-1 n~<

Then (10.26a) gives for x=l+1

(10.31)
1) ScHOUTEN and V. KAMPEN 1931, 2, 4.
 10. Higher curvatures of a Vm in Vn. 281

For l = k this is trivial but for l < k (10.31) expresses that for every
l z 1+1 A
value of p1_ 1 and q1H the vectors H.,p 1_, and n;.H,:.q1+1 alt have the
same direction. Hence for m = 1 and 1 :;;;.z < k the curvature tensors
I k
of valence 3 from H;,~" to H;,;." can be written as follows

I
( I l
a) H;,i" = H,, U;. n"
1+1 1+1 l 1+1
b) ~,:;." = oc H,,n;. n"
1
(10.32)

C) H ;." = kH k-1 k,.

OC r, n;. n
10

in which H,, is a definite Vector of Vm and U;. a definite unit'IJector of

Rm1_, and where J", ... , fi" are unitvectors in the one dimensional spaces
Rm1 , , Rmk. F or k = l the vector H,, does not exist and instead of
(10.32) we have (10.30) only. 1) Note that D,,;,q,.=O; x=l, ... ,k.

l
From (10.26b) and (10.32) we get for 1:::;;.l<k and x=l, ... , k

a) (D[s, H,,]) uPz-t nq,

l
+ H[r, (Ds,] uPI-t). nq,l = 0
(10.33) b) (J?[s, l(il Hd }lzl:fiL, = 0

c)

from which we see (by transvection of (10.3 3 a) with up1_,) that H,, is
a gradientvector, H,, = o,,H, and that the coefficients ... , ~ can all 1 4"~
be written as functions of H only. The equation H = const. represents
a normal system of Vm_ 1's in Vm. If vPo is a vector in these Vm_ 1's we
have vPHp,=O and from (10.33a) it follows that
(10-34)
but this equation is only part of a more important result that can be
derived from the equations of FRENET.
In fact, according to (10.14a), (10.32) and (10-34) we have

(10.35)

1) EISENHART 1937, 1. Although in most parts of this paper EISENHART only

considers the case l = 2, his point of view is more general because his fundamental
tensor in Vn is also taken to be indefinite. This gives some very interesting re-
sults that could not be included here.
282 V. Imbedding and Curvature.

But (10.26a) gives, for x=l,

1-1 1
(10.36) HPt-
[s 0 ii

Hrol u"nql = 0 '
hence
(10.37)

and this proves that for m 1 = 1 and 1 ;S; l < k the field u" is covariant
constant over each of the Vm_ 1 's of the normal system.l)
Returning now to the equations of FRENET (10.14) for x=l, ... , k

l
we get according to (10.32) for m 1 =1; 1-;;;:,l<k

a)

(10.38) b)

l
c)
or
a) DTon1., = - Hro ("
u - 1+11+!)
IX n

1+1 1+11 z+ol+2

(10.39) b) J?ro n" =- H,o (IX n"- IX" n")

k k k-J
c) D,0 n" = - H,o IX n",
hence
(10.40) v'oDro il," = O' X=[, ... , k;

and this proves that for m 1 = 1, 1 S.l < k the fields k, ... .J" also are
covariant constant over each of the Vm_ 1 's of the normal system H = const. 1 )
Substituting (10.32) in (10.26c) we get for m 1 =1,1-;;;;,l<k

(10.41) X= l, ... , k

and for m 1 = 1, 1-;;;;;, l = k

(10.42)
J a)
1 b)

We see from (10.26c, 32a, 42b) that for l = 1

(10.43) /c .. Po = ~ 2 .AE;,Po h,oJ qo

Sofoqo
for k = 1
0 for k > 1.

1) EISENHART 1937, 1 for 1=2.

 10. High er curvatures of a Vm in Vn. 283

Hence if m1 = 1 the Vm lies in a flat Rm+l in Rn or it is a bent R.,.. This

implies that if a Vm is imbedded in an Rn suchthat m 1 = 1, it can always
be imbedded in an Rm+I 1) We prove the following more general theorem:
A Vm in Rn with m 1 = 1; 2:;;;" l < k can always be imbedded in an
Rm+m, + .. +mz_ 1 such that its x-th curvature region is an RmTm, + .. +mx;
X = 1, ... , [-J. 2)

If we take (10.32) into account we have from (10.14a) for 2:;;;" l < k

0 1 1
b) D,, s;, = H;,q:P 1 Bf,1

c) D r, B qx-
" - - HPx-
x
1
r, qx B"Px-1
x x-1
+x+1 x+1
H"r,qx Px+t B"Px+t X= 1' ' l - 2
l-1 l-1 p l-2 l

(10.44)
d) D,, B':/z-1 = - H;, :- qt-1 B'Pt- + H,, Uqz-1 n"
l l-1 l+l l+1
e) D,, n" = - H,, uP1-1 Bf,1_1 -j- IX H,, n"
y y y-1 y+J y+1
f) D,,n"= -IXH,, n" +IX H,, n"; y=l+1, ... ,k-1
k k k-1
g) D,,n" = -IXH,, n"

and from (10.26) the integrability conditions of (10.44)

x x+1
a) nPx-1
[s0 JlxJ
Htx
r,]. qx+1 -
-0' x=1, ... ,l-2
l-1
b) r_r.Pt-
ll[s0 Jlt-d
H
r,]
utl-1-
-
o'
c) D [so H:t'o] Pz-1 qz = o, z = 1, ... ' l - 1
d) D[s, H,,J = 0
1+1 k
e) H[r, Os,] IX = 0; ... ; H[r, Os,] IX = 0
0 1 1
(10.45) f) Ks:r:i/' = 2H[;,~'Jt 1 J H,:Jq,1'
X X X x+1 x+1
g) KP-2R PxHtx-1 +2RPx
s,r0 qx - [s0 Jix-d r,] qx [s0 Jtx+d
H" .tx+t.
r0 ]qx '
X=1, ... ,l-2
l-1 l-1 l-1
h) K Pl-1 - 2R. Pt-1 H" lt-
s,r,ql-1 - [s, Jtt-d r,] ql-1

u = l, ... , k +1

1 ) EISENHART 193 7, 1.
2) EISENHART 193 7, 1 for l = 2.
284 V. lmbedding and Curvature.

. l-J
... , Bq,_ ,nl" k" k+J
In (10.44) the vanables " o,.
~,Bq,, 1 n, Bqk+I are the
unknownsan d t h e r..roqo'
Po
... ,
J'..Pl-I
roql-t'
H1 Pt
foqo ' ... ,
lH-! Pl-I H
foql-2 '
Pt
'o'u -,
I 1+1
0:. , ...
k
,rx
are known functions of the COOrdinates in Vm, satisfying the conditions
(10.45 a-h, j). Now this same Vm can be imbedded in a flat Rm+m.+ +mz~ 1
in Rn with curvature regions of dimension m + ~, m + m1 + m2 , ... ,
m + + m1_ 1 if and only if equations of the form

0
a) 'B;, = oq,~"

l
0 1 1
b) Dr, IB;, = IH,:i/ IBf,.

(10.46) c)
x
D fo 1B"qx = - 1HPx-I
Yo qx
1B"
Px-1
x x-1
+ 1x+l x+1
H" Px+I 1"
1'o Px+t '
qx

X= 1, ... , [ - 2
l-1 l-1p 1-2
d) Dr0 B"ql-1 1 = - IH ,_,
ro ql-1
1 B"Pt-'J
I
e) D,, ~B~, = 0

have solutions. Now the integrability conditions of these equations are

x x+1
a) IH[s Px-IH
.1txl
tx _ 0 X= 1, ... , l - 2
1'0 qx+t - ' 0 ].

z
b) D [so IHr,].Pz-1 qx =
0 1 z = 1, ... ' l - 1
0 1 1
c) Ks,r,q,-
.. p,- 2 1H.[s".lt.!
Po 1Hr ]q 1 0 0

(10.47) X

-
X

[s, '.Ix-tl
X
Px _ 2 1H Px 1JJ tx-
d) 1Ks 0 r0 qx r,]. qx
+ 21R.
x+1
Px
[s, llx+II
x+1
IH .tx+1.
r,] qx 1

X= 1, ... , f- 2
1-1 1-1 1-1
IK Pt- _ 2 1H p,_,IH t,_,
e) So fo ql-1 - [so ~tl-2; fo] ql-1

I
f) KSofa..qlP, = 0
1

and these equations are identical with (10.45 a, c, f, g, h, i) if we take

1JlPx-I
So qx
= JlPx-I .
So qx '
X= 1, ... , l-1
{
(10.48) r.Px = r.Px .
1
roqx YorJ.x' x=1, ... ,l-1
H,, = 0 1)

,, q1 , wh ere th e m
an d zero for a 11 lr..P' . d"Ices p1 , q1 now b e1ong t o a re gwn
of dimension m 1 + + mk+ 1
1) It is not necessary to take H,, = 0.
 11. Product spaces. 285

A theory of the imbedding of a W". in W,. 1) and of an Am in A,. 2)

involving higher curvatures was developed by HLAVATY. Problems of
higher curvature for V".. in V". in V,. were dealt with by LoNGO 1948 3 ).
Umbilical subspaces in umbilical subspaces were considered by YANo 4 ).

11. Product spaces. 5 )

If two riemannian spaces V". and V,._", are given with coordinates ~;
+
a., , y = 1, ... , m, and ;; ; ;, 'Yf!; = m 1, ... , n, and fundamental ten-
sors gpa. and g11 ;, the V,. with the coordinates ;''; u, J..,p,=1, ... , n, and
the fundamental tensor g.,. for which g11 a. = 0; gp; = 0, is called the
product of V". and V,._m. A V,. that is a product space is said to be
decomposable (or separable). That means that V,. is decomposable if
and only if therc can be found a coordinate system (u) suchthat g11 a. = 0;
gp; = 0 and the gpa. and the g11 ; are functions of the ;a. and the ;; respec-
tively.
If two spaces L". and L,._", are given with coordinates ~a. and ~;
and the connexions rva.P and I'l11 , the L,. with the coordinates ;" and
the connexion I'p"J. for which all components with two indices of dif-
ferent kind vanish, is called the product of L". and L,._",. An L,. that
is a product space is also called decomposable. Hence an L,. is decom-
posable if and only if there exists a coordinate system (u) such that all
parameters of the connexion with two indices of different kind vanish
and that the others only depend on the set of coordinates to which
they belong. Obviously an L,. that is decomposable according to thc
second definition and happens to be a V,. is also decomposable according
to the first definition.
An object in a decomposable L,. is called breakable if its components
with respect to a special coordinate system as defined above are always
zero when they have indices from both ranges. A breakable object is
said to be decomposable if and only if the components belanging to the
subspace L". (L,._",) depend on the ;a. (~~) only. The objects gJ.>< and
I'p"J. in a decomposable V,, or L,. are examples of decomposable objects.
A decomposable L,. contains a set of oon-m X".'s and a set of oo"'
x,._ ...'s with the equations
(11.1) a);i*O;;=m+1, ... ,n and b)~*O;oc=1, ... ,m

1 ) HLAVATY 1949, 1; 1952, 1.

2) HLAVATY 1949, 2; 3; 4.
3) LONGO 1948, 1.
4) Y ANO 1940, 3.
5) Generalreferences: EISENHART 1923, 1; LEVY 1925, 2; DusCHEK andMAYER
1930, 1; T. Y. THOMAS 1939, 1; 2; FICKEN 1939, 1; LICHNEROWICZ 1944, 1.
286 V. Imbedding and Curvaturc.

such that nowhere an Xm has a direction in common with an Xn-m.

Each of these sets is geodesie and parallel in Ln. In fact, if v""; v~ = 0
is a field in the Xm's, we have for each displacement in these spaces
(d~" = 0)
(11.2) ~v~ *- d~Y F./ v * 0

and for each displacement in the other spaces (d~ = 0)

(11-J) ~V;~ d~; rt v J_~ 0.

Conversely an Ln is decomposable if there exist a set of oon-m geodesie

and parallel Xm's and a set of oom geodesie and parallel Xn-m's having
nowhere a direction in common. In an ordinary Vn it is sufficient
that there is a set of oon-m real geodesie and parallel Vm's because in
this case the set of orthogonal Vn-m's exists and is geodesie and parallel.
But if the fundamental tensor of Vn is indefinite it may happen that
the tangent m-direction and the (n- m)-directions perpendicular to it
have a direction in common.
If a vn
is decomposable, g;." can be split up into two symmetric
tensors
g;." = 'g;." + "g;."; 'g; * 0; 'g'I<X * 0; 'gfl< * 0
(11.4) {
"g$ * 0; "gf/<7.~-0; "grt.=* 0
and each of these tensors is covariant constant.
(11. 5) a) VI' 'g;." = 0; b) VI' "g;." = 0.
Moreover the 'g<>. ("g";) depend on the ~ W) only. But this implies
that also the symmetric tensor 'c 'g;."-t- "c "g;." with two constants 'c
and "c, is covariant constant. Conversely, if there exists in an ordinary
Vn a real covariant constant symmetric tensor h;." that is not a multiple
of g;.", the principal multidirections of h;." form parallel fields. Hence:
An ordinary Vn is decomposable if and only if there exists a real co-
variant constant symmetric tensor field h;." that is notamultiple of g;.". 1 )
In a decomposable Ln any figure can be "projected" by the Xn-m's
on one of the X".'s and by the Xm's on one of the Xn-m's. These pro-
jections have many interesting properties. For instance it can be
proved that a curve in a decomposable An is a geodesie if and only if
its projections are both geodesics.
There are several generalizations possible. WoNG 2 ) considered a V,,
whose linear element can be written in the form
(11.6)
1 ) T. Y. THOMAS 1939, 1; 2; LICHNEROWICZ 1944, 1.
2) WONG 1943, 1.
 1. Projective transformations of a symmetric connexion. 287

where the g(j, and g11 c are general functions of the ;"- and ~.; both. Such
a Vn need not be decomposable. In this case the Vm's: ~.; = 0 and the
vn-m's : ~(j, = 0 are called complementary subspaces.
If in (11.6) g(j, and g11 ,; have the form

(11.7)

where the I gcx. (" g'l,) depend Oll the ~(j, (~.;) only but e and (J are general
functions of the e and ~.; both, the v;. is said to be conformally de-
composable (separable). 1)

Many recent investigations deal with the geometry in the large of

decomposable and indecomposable compact spaces.

Exercise.
V 11 '1 Every symmetric vn that is not decomposable is an EIN-
2 ).

STEIN space.

VI. Projective and conformal transformations

of connexions. 3)
1. Projective transformations of a symmetric connexion.
In Ch. III 7 we proved that the transformations of the I;.";. in An
which preserve the geodesics have the form

( 1.1)

where P;. is an arbitrary vector. We call this transformation projective

and P;. the vector of the transformation. From (1.1) and (III 2. 3, 7)
we get
(1.2) + +
{ a) ,'17'" v": 17'" v" A~ P;. v;. v" P'"
b) 17'"w;. -17'"w;.- 2w<'"P;.)
1) YANO 1940, 1; WoNc 1943, 2.
2) LICHNEROWICZ 1944, 1.
3 ) In this chapter a Vn is always ordinary and imbedded spaces are always
real if another Supposition is not made explicitly.
4 ) A transformation of the r;;" in Ln of the form r;;"-->-T;;"+ p'"Al preserves
not only the geodesics but also parallelism of directions (cf. VI 9). The simplest
object that can be formed of a connexion given to within transformations of this
kind is r;;"- _!_~Alas was proved by J. M. THOMAS 1926, 1. Cf. FRIESECKE
n
1925, 1; BORTOLOTTI 1930, 3, p. 75ff.; 1931, 4; SCHEIBE 1952, 1 and footnotes 5,
p. 322 and 6, p. 334.
288 VI. Projective and conformal transformations of connexions.

If (1.1) is substituted in (III 4.2) it follows that

(1.3)

and accordingly

(1.4) { a) 'R'"=R'"+nP'"-P'"
b) 'V"'"=- 2'R1.'"1=V"'"- 2 (n + 1) P1.'"1= V"'"+ 2(n +1) Vt. p'"1
(1.4b) implies that a volume preserving connexion (Exerc. III 4,5) is
transformed into a connexion with the same property if and only if PA
is a gradient. The transformation is then called restricted projective. 1 )
If we wish to make 'V"'" zero, the integrability conditions of (1.4b) are
Vrw V"'"1= 0. But these are identically satisfied (cf. III 5.24), hence
every connexion can be transformed into a volume preserving connexion
by a projective transformation. 2 )
An An and its connexion are said to be projectively euclidean 3 ) and
the An is called a Dn if it can be transformed projectively into an En- 4 )
The n. a. s. condition is that there exists a vector field PA such that

{
a) R;~~" = 2 .F[.'"1 A~- 2A[v P'"H
(1. 5)
b) p'"A def- J7'" PA+ p'" PA.

Now in an An the curvature tensor satisfies only the first and the second
identity (III 5.1, 4) and therefore the number of its independent com-
ponents is n 2 ( ; ) - n (;) = { n 2 (n 2 -1). This r_ umher is > n 2 for n >2,
hence (1. 5a) can always be satisfied for n = 2, but for n > 2 only if
R~ ~~" satisfies certain algebraic conditions. In this case we get from
(1.4a) for P'"A=PpA and 'R'"A=O

a) Pp = _ _!!__R -n2_
- 1 -R
n2_ 1 p Ap 1
1 1 1
(1.6) -- R A- -2- V A = - - - R< A) -- -1- Rr Al
n-1 '" n -1 I" n-1 '" n+l I"

b) P.[pAJ = 1 VpA
2(n+l}

1)The term was introduced by CoBURN 1941, 1.

2)Cf. EISENHART 1922, 2, p. 236.
3 ) Many authors use the term projectively flat. But an An_
1 in En can be
euclidean, that is an En_ 1 , without being flat in the ordinary sense, for instance
a cone has this property. There is no objection to the term locally projectively flat.
4 ) In BaRTOLOTTI 1931, 4 a Dn is an Ln whose connexion is given to within

a transformation of the form r;:A --'>-r;A +

p'"A~. Cf. footnote 4 an page 287.
 1. Projective transformations of a symmetric connexion. 289

The first integrability condition of (1. 5 b) is

(1.7) R;,;~"' P", = 2 V[,. ~,JJ,- 2 (V[v P"l) P;. -2FL, 17vl P;.
or, according to (1.5 a, b)
(1.8)

Now, if the identity of BIANCHI (III 5.21) is applied to (1.5a) we get

(1.9)
and from this equation by contraction over x).

(1.10)
and by contraction over xv
(1.11) 17[w PJp] = i (n- 2) J7[wpp]).
Hence, for n > 2 the integrability condition of (1.5 b) is a consequence
of (1. 5a). This proves the theorem 1 ) :
An An is projectively euclidean if and only if
(1.12) ... xctetR"'
P vp). ~- vp). -
2n
r[vp]
A"+2A"P
). [v p]). = 0
where
(1.13) p;. cte_f- __ 1't_ R ' - ___1_ R,
I' - n2 - 1 I' n2 - l I'

and if for n = 2 moreover

(1.14)
For n=2, (1.12) is identically satisfied and for n >2 (1.14) is a conse-
quence of (1.12).
P;;~"' is called the projective curvature tensor. 2 ) From (1.13) and
(1.4a) it follows that
(1.15) 'P";.=Pf.',-Pf.'t.
and by substituting this in (1.12) we see that P;;~" is invariant for
projective transformations of the connexion. For n = 2 it vanishes identic-
ally and for n > 2 it satisfies the identities

{
a) l(~;)~"' = 0; b) P[;;~t = 0;
(1.16)
c) P;;~ = 0; d) P;~/ = 0.
1 ) WEYL 1921, 2, p. 105. In EISENHART 1927, 1, p. 97 there is amistakein the

condition for a Vn of being a Dn. It is probably caused by a wrang interpretation

of WEYL's term "scalar curvature ". The same mistake occurs in LoVELL 1934. 1.
Cf. footnote 4, Ch. III, p. 148. For n = 2 there is a mistake in ScHOUTEN 1953,
1, p. 74.
2) Cf. for the semi-symmetric case ScHOUTEN 1925, 2.
Schauten, Ricci-Calculus, 2. Auf!. 19
290 VI. Projective and conformal transformations of connexions.

As a eorollary we get from (1. 5 a) that for a projectively euelidean

eonnexion the eurvature tensor has at the most n 2 independent eom-
+
ponents in the general ease and ~n (n 1) if the connexion is volume
preserving.
From (1.12) we see that in a Dn all the eomponents of R;;.;," vanish
for whieh " is not equal to at least one of the indiees v, t-t or A.. It ean
be proved that for n > 2 this eondition, if valid for every coordinate
system, is also suffieient :1 )
An An, n >2, is a Dn il and only il the equation
(1.17) R;;.;," = 0 for "=l= V,f-t, A.
holds lor every coordinate system at every point.
Another n.a.s. eondition ean be derived as follows. The Xn_ 1 with
the equation I W) = c is geodesie (V 7) if and only if v" v;. V" V.< I= 0
is a eonsequenee of v" V" I= 0 and I= c. Hence a n. a. s. condition is
that (cf. Exere. I 3,2)
(1.18) V"w;.=2q<Pw;.l; w;.detv.. l.
at all points of the Xn_ 1 . Now the first integrability eonditions of this
equation are

(1.19) {
-R;;_.;,"w" = 2(J7(.qp]) W.< + 2q[p V"1W,< + 2w[p V"1 q;.
= 2w 17[. qp] + 2wr. (qp] q.. - vp] q.. ) .
If (1.18) is totally integrable, that is if (1.19) ean be identieally satisfied
for every value of w.<, then for every point of An and every (n -1)-
direction through this point, there exists a geodesie Xn_ 1 having at
this point this given (n -1)-direction. Now taking w.< = l.< it follows
from (1.19) that R; ;.;, 1 = 0 for A., f-l v =l= 1 and this holds also mutatis
mutandis for w.< = l.. , ... , W;. = 3.< and for every ehoice of the eoordinate
system. Hence, for n > 2, the An must be a Dn and beeause in an En
and therefore also in a Dn there exist geodesie Xn_ 1 's through every
point in every (n -1)-direction, the condition is also neeessary: 2)
An An, n > 2, is profectively euclidean il and only il there exist through
every point geodesie X,._ 1 's with every (n-1)-direction at that point.
In the ease when the An is already an En we may require all pro-
jective transformations of the r;;.
for which the En remains an En.
From (1.4, 5) it follows that p"" = 0 and that accordingly p" is a solu-
tion of the equation
(1.20) o" p.. - p" p.. = o
1) E II 1938, 2, p. 181.
2) E II 1938, 2, p. 182.
 1. Projective transfonnations of a symmetric connexion. 291

if the coordinate system is chosen such that r;;;. = 0. From (1.20) it

follows that
(1.21)

and this means that P;. is a parallel field (cf. III 1). Taking the recti-
linear coordinate system such that P;. = pl;. we get for p the equations
(1.22)

leading to the general solution for P;.

(1.23) P;. = Pl;. = - W + C)- 1 l;.; C = const.

For every choice of l;. and C there is one solution. If for any definitely
chosen solution we lay the first E,._ 1 of the P;. belanging to the point !;"
through this point, the second E,._ 1 goes through the point =- C e
on the 1-axis. 1) Hence this second E .. _ 1 is common to the P;.'s of the
whole field. In fact it istheimproper E.. _1 for the transformed connexion.
To prove this we take the affine parameterz on the geodesie 1-axis
for the given connexion equal to e.
Then, according to (III 7.11) we

l
get for the affine parameter belanging to the transformed connexion
'z = + C2
C1 f e+ZfPp.de'"' dz
(1.24) =Cl f e-2log(<l+C) de + c2
= - cl W + C)- 1 + c2

which takes the value oo for ~1 = - C. Hence

If the connexion of an E .. is transformed profectively so that the E,.
remains an En, the vector field P;. consists of parallel covariant vectors
which all have the same second E .. _ 1 and this En-l is the improper E,._ 1 of
the transformed En- 2)
Exercises.
VI 1,1a). Prove that V.,..=O in a D,. in which a covariant constant
field vJ< exists.
VI 1 ,2. An A,. is a E,. if and only if for every value of P;. at an arbi-
trary point !;" there exists a projective transformation that leaves
0 0
the curvature tensor invariant and whose vector takes the value p;.
at !;".
0
1) Cf. for the geometric representation of a covariant vector I 2.
2) R. K. 1924, 1, p. 133; E II 1938, 2, p. 184.
3) LEVINE 1949, 1. The term parallel is used there in the sense of covariant
constant. Cf. III 2, VI 9. Cf. LEVINE 1950, 1 for the c,..
19*
2-92 VI. Projective and conformal transformations of connexions.

VI 1,3 1). The projective transformations with vectors p;., 2p;. and
1 2 3
3p;. transform R;~;." into R;~;.", R;;;." and R;;;.". Prove that
+ 3R;~;." -R;;;." = 0.
1 2 3
VI 1,3 oc) R;~;."- 3R;~;."

VI 1,4 2 ). Prove the identity for An, n >2

VI 1,4 oc) r7
l'[w
. ."
Pvp]A ---
1 A"[w~'lal
r7 . .G
Pvp].\ = 0
n-2

2. Projective transformation of the connexion in a V,.. 3)

Every Sn is a Dn because in consequence of (cf. III 5)

(2.1)

the equations (1.12) ate identically satisfied for ~;.=-ugpJ. and P;. is
a solution of the equations
(2.2)
whose integrability conditions are identically satisfied. Because two
Dn's can always be mapped on each other so that geodesics remain
geodesics, the same holds for two Sn's and for an Sn and an Rn-')
If, in a Vn, a new (symmetric) fundamental tensor 'g;.,. is introduced,
a new connexion arises. As follows from (1.1) this change of connexion
preserves the geodesics if and only if a vector p4 exists such that
(2.3)
-1

l
By transvection with 'g" 4 we get (cf. III 3.12)

(2.4) 2(n
-1
+ 1) P,.. = 'g"). v,.. 'g;." =
-1
'g"). 81, 'g4 , . - 2 rl' = 81' log f;
I

'g cl~ JDet ('g;.,.)j.

As a corollary we get that P;. is a gradient vector, which could have

been foreseen because the connexion remains riemannian and therefore
remains volume preserving. From (2.3) and (2.4) it follows that the
transformation of the connexion is projective if and only if

(2.5)
1) BERWALD gave a similar proposition for conformal transformations in V2
(Arch. Math. u. Phys. 1918, 27, p. 81; Jahr. D. M. V. 1923, 22, p. 48).
2) R. K. 1924, 1, p. 132; E II 1938, 2, p. 182.
3 ) Cf. Eil 1938, 2, p.185ff. also for Sn.

') Note that we do not consider spaces in the large.

 2. Projective transformation of the connexion in a V,.. 293

The differential equation of the geodesics of an A,.

(2.6) _i_ d~" =0

dz dz

is said to have a firstintegral of degree p 1 ) if a tensor field q;., ... exists ).p

such that
d~}., d~p
(2.7) qJ., .. ;,P -dz- -dz- = const.

along every geodesic, and this condition is satisfied if and only if

(2.8)

Now it follows from (2.5) that

(2.9) f-l-(n+l) = _'[_
g
and this proves that in a V,. which can be transformed projectively
into another V,., the equation of the geodesics admits besides g;.,. d~J. d~"
d~J. d~" ds ds
= const. another first integral f-t 21 g;.,.----
ds ds
= const.
If the principal directions of g;." are uniquely determined, it follows
1

from (2.3) and (V 5.17) that they are all V,._ 1-normal. The integrability
conditions of (2.3) are
(2.10)

because P;. is a gradientvector. This equation could also be derived from

(1 3) by substituting K /'
'
.. " and K ...
V ~tJ.
" for R ~tJ.
.. " and IR ...
1
~t
V
" because
' V V

Prv~tJ = 0 and 1 K;;<;,~ g><)Q = 0. 1

If the V,. is an S,. (2.10) takes the form (cf. III 5)

(2.11) 0 = + 4:~e g[v(J. A~] 1

g,.)Q + 4P[v(J. 1 g~t]><) = 4(P[v(J. + :le g[v(J.) 1 g~t]><;

but this equation expresses the fact that P.;. + :~eg,;. and 1
gvJ. differ only
by a scalar factor. Let this factor be IX
(2.12)

then we get from (1. 3)

(2.13) 1
K;;;." = - 2:~e g[v lAI A=l - 2:~e A[v g~tlJ. + 21X A[v 1 g~tJ.< = 21X A[. 1 g~tJA

and this proves that for n >2 after the transformation once more we
have an S,. but with the scalar curvature Hence +IX.
1) Cf. EISENHART 1924, 1; 1927, 1, p. 83-86; LIBER 1941, 1.
294 VI. Projective and conformal transformations of connexions.

If the connexion of an Sn, n > 2, is transformed projectively and if

the transform happens to be the connexion of a Vn, this Vn is an Sn. 1 )
As a corollary we get that a Vn is projectively euclidean if and only
if it is an Sn 2 ).
We return now to the general Vn. By transvecting (2.10) with g";.
we get
( ?~ 14) KQI
v ge" +K.a.e1 -pe1
v." gea- v ge"- peagea1 +P 1g ea pe 1
gv" v" eag - " gev
hence from (2.14) and (2.10) transvected with ("
(2.15)
But this equation expresses the fact that the symmetric tensors 1g";.
and IK";. have at least one set of n mutually perpendicular principal
directions in common (cf. Exerc. I 9,3). Transvection of (2.10) with g;."
leads to
(2.16)
and this expresses the fact that also the symmetric tensors g";. and p";.
1

have at least one set of n mutually perpendicular principal directions

in common. But (2.10) contains still more information about these
principal directions. If a local coordinate system (h) with unit basis
vectors is laid in principal directions common to 1g"J. and p"J.' (2.10)
can be written as
(2.17) Kkjih (1ghh- 1gii) * Pki 1gjh- Pji 1gkh + pkh 1gJi- pjh 1gki
For k =j both sides vanish and the same holds for i = h. For k, j, i =F
we get 3 )
(2.18a) Kkiih *0
provided that h and i do not belong to the same principal multidirection.
For k=i, h=f we get
(2.18b)
Hence, if 1gii= 1gii' it follows that hi=P;; and this means that every
principal multidirection of 1g;." is contained in a principal multidirection
of P;.". But this has as a consequence that every principal direction
of 1gJ." is also a principal direction of pJ.". Hence the mutually perpendi-
cular principal directions common to g;." and K;." are also principal
1

directions of pJ.x, and thus from (1.4a) in tne form

(2.19) IK~"). = K,.J. + (n- 1) P,.).
1 ) BELTRAl\II 1868, 1; cf. EISENHART 1926, 1, p. 134.
2) WEYL 1921, 2, p. 110.
3) FUBINI 1905, 1, p. 306.
 2. Projective transformation of the connexion in a V,.. 295

it follows that they are also principal directions of K;",. This proves 1

the theorem
The symmetric tensors g;.,., p;",, KA>< and K;.,. have always at least
1 1

one set of n mutually perpendicular principal directions in common. 1 )

We now consider the special case where the principal directions of
gb are uniquely determined.
1
1'aking the anholonomic coordinate
h
system (h) with unit basis vectors ~"', i;. in these directions, (2.3) takes
the form
(2.20 ) { 8; gih- 2 Ij 1(; ghll
1 1
*- 1
g; 11 ~ ;og P-! 1
g11 ; 8; log p -! 1
g;; 811 log p;
p, e - ! 8; log p

I
and from this we derive (cf. III 9.11) for all possible assumptions with
respect to h, i and j
a) lj; 11 ('g11 ~o- 1 g;;)-*-O; (h, i, i =f=)
b) 8; I gii *- I gii 0; log p; (h = i =f=i)
(2.21)
C) IJ i h ( gii - g,. 11 ) * - ! gii 811 log p;
1
I I (h =f= i = i)
d) 8/g;; * -2 g;;8;logp; 1
(h = i =i).
From (2.21 a) and (V 5.12) it follows once more that the n congruences i"

are V,._ 1 -normal. Accordingly the coordinate system (") can be chosen
suchthat its contravariant basis vectors are tangent to those congruences.
This means that the linear element can be written in the form
hdef.~th"
(2.22) e;.=u,.e"
and
(2.23)
Then we have for h =f= i

]J,
h *
=- l7i z;~ =* -
p ' ~ ! ~ f ! ~ i i
z ~ V" f;. = z" z V" z;. = z" z V"'H e;
(2.24) (
iih i
i i h i i i i
= H i"' iA V,u e;. =HiP i" VPH- 1 i" * H o,.H- 1

l
and accordingly from {2.21 b, c, d), writing (!; for the g;; from (2.21)
1

a) o;Pe;""- 0; . i =f= i
(2.25) b) (e11 - e;) 811 log H-*-! 811 e;; h=f=i
c) 8; p (!; + (!; 8; p '!'_ 0
1) R. K. 1924, 1, p. 206.
296 VI. Projective and conformal transformations of connexions.

in which 8i can be interpreted as Af 8~" but also as bf 8~", because

i
Af '!_H- (J~" (no summation over j). The integrability conditions of
1

(2.25) are the well known equations of LAMitl)

Li 1 i k i i h
(2.26) 8i8iH~*-H- (8iH) 8iH +H- 1 (8iH) (8iH); h, i, j=i=;
(no summation over i and j)

which are equivalent to

(2.27) k, j, i =i=
and this is a special case of (2.18a).

Exercises.
VI 2,1 A Vn can be transformed projectively into a Vn if and only
2 ).

if there exist a gradient vector P;. =i= 0 and a symmetric tensor 'g;." of
rank n satisfying (2.3).
VI 2,2 3 ). In a Vn a covariant constant symmetric tensor field 'g;,"
of rank n can exist if and only if
a) the differential equation of the geodesics admits a first integral
' d~). d~"
g;." Tz -dz- = const. ,.
b) the connexion of the vn
can be transformed projectively into
the riemannian connexion belonging to 'g;.".

3. Imbedded spaces in An
under projective transformations of the connexion. 4)
Let rt; a = 1, ... , m be a set of Coordinates in an Xm imbedded
and rigged in An and let f'c~a be the parameters of the connexion in-
duced in Xm (cf. V 3). Then we have for each vector field va of Xm

(3.1)
1) Cf. BIANCHI-LUKAT 1899. 1, p. 485. Cf. for the integration of (2.25) LEVI
C!VITA 1896, 1; FUBINI 1905, 1; WRIGHT 1908, 1, p. SOff.
2 ) E II 1938, 2, p. 186.
3) LEVY 1925, 2; EISENHART 1927, 1, p. 80; EIl 1938, 2, p. 193. Cf. CARTAN
1927, 3 for another property of Vn's of this kind.
') General references on projective imbedding: ScHOUTEN and HAANTJES
1936, 1; BoRTOLOTTI 1941, 2; BoMPIANI 1943, I; KIMPARA 1943, 1; KANITANI
1943, 1; 2; 1947. 1; 1948, 1; 1949, 1; 1950, 2; NoRDEN 1945, 3: 1947, 1; 1948, 1;
1949, 1 ; BoL 19 so, 1.
 3. Imbedded spaces in An under projective transformations of the connexion. 297

hence (cf. V 3.10) 1)

(3.2) r c'ab_- Bt-<Aar,><

cb>< p;,-
B";." Ba
cbup }.

If now the connexion in An suffers the projective transformation (1.1)

we get for the new connexion in xm

(3-3) { 'F;t = B~t~ (F:;;. + p~ AJ: + P;, A:) - B~; 8" B'f.
= I'/t + p~ B~ + p~ B~; p~ d~ Bg p;,
and this proves that the induced connexion is also transformed profectively.
From this it follows at once that a geodesie Xm in An remains geodesic.
This is also one of the consequences of the fact that the quantities
m m'
H;";," and H;"i" defined in (V 7.9, 11) areinvariant for projective trans-
m m'
formations of the connexion in An. The quantities L~~;. and L;"~;. are
in general not invariant, for instance

(3.4) ~ 'LJ"~;. = ~~}. + ~~ (-PT Af B-~- P;. A~ B~ +PT A~ Bf +Pa A~ Bf)

"
= I p.J.- B" p".
},r
{I
p") .def= C"
},
p,..

If for an Xn_ 1 in An the rigging is chosen in the way described in

V 3, it depends on the connexion in An. Because it was assumed
in that section that the connexion was volume preserving we need only
consider here those projective transformations for which p1 is a gra-
dient, P;. = 8;. p. In V 3 the vector t;. was fixed by (V 3-31) and in
that formula qwas a covariant constant scalar density field of weight 1. +
Let now 'q = (! q be covariant constant for the transformed connexion

(3.5) 'V;. 'q =(V.. e) q + e 'V.~ q = qV.~ e- (n + 1) e P.~ q,

then
(3.6) V.~ log e= (n + 1) P.~
and according to (V 3-31), if 't.~ =at.~ is the new normalized tangent
vector, we have
,,... _n _J_ _tJ:
(3.7) a t.~ = q a 2 f) 2 i;, = (! a 2 t;.
or
(3.8) (n + 2) log a = 2log (!
at all points of xn-1' hence

(3.9)
1) We use an accent on the right here to denote the induced connexion in
order to preserve the accent on the left for the transformed objects.
298 VI. Projective and conformal transformations of connexions.

For the new vector 'n" we find from (V 3.28)

+ 2: j 2 Bi; hab'Pa).
-1 1 -1
(3.10) 'n" = a- 1 (n" +Bi; hab 8a log a) = a- 1 (n"

According to (V 3.11, 12) we get for the new values 'heb and 'l/
(3.11)

(3.12)

where
-1
(3.13) Va def 2 n +1_ hab pb'
- n+2

From (3 .11) we see that the asymptotic lines are invariant for the
transformations under consideration (cf. V 3). The lines of curvature
have not this independence as can be seen from (3.12).
If a curve is given in An with an arbitrary parameter t, the covariant
derivative ~ r}t;~ depends on the choice of this parameter. If z = z (t)
dt dt
is introduced as a new parameter, we get

(3.14) ~
dt
df;"
dt
= z' ~z' rjf;" = z' 2 ~_>:_
dz dz 2 dz
+ z" dt;"
dz '

hence the 2-direction spanned by d ~"jd t and (/ f'jd t2 is independent of

the choice of the parameter. Now if we write

(3.15) t" def _tl__["_ CJ 2 t;-l~

-- dt dt 2 '

this quantity transforms into

(3.16)

A curve in An will be called quasi-plane 1 ) if t"' is parallel along the curve:

(3.17)
In this case we ha ve
(3.18)

and this proves that

(3.19)
1) YANO and TAKANO 1944, 1; 2; cf. YANO, TAKANO and ToMONAGA 1948, 2;
Y ANO 1949, 2.
 3. Imbedded spaces in An under projective transformations of the connexion. 299

provided that z is a solution of the equation

(3.20)

The general solution of (3.20} has the form

(3.21) z = C1 f efl!J.dt dt + C2 ;
where C1 and C2 are constants and C1 =!= 0. z is called the affine para-
meter of the quasi-plane curve. It is fixed to within an affine trans-
formation. C2 fixes the nullpoint of the scale and C1 fixes a kind of
gauge. If z is chosen in this way we call n" = ~l ~"Jdz2 the affine normal
of the curve. It is fixed to within a constant factor. According to (3.19)
we have along the curve

(3.22)
f53;H
dz 3 + k (z) -d.Z = 0
d~

where k is a function of z. From (3.17) we get

(3.23) !5 3 ~ -oc(t) t5'~ -(t) d;"-=o

t5J3 dt' dt

and after the introduction of z

z' 3 -"- + (z"' - oc z" -

r~
z') -"-
d~
(3.24} d~ dz
=0
from which
(3.25) k = z'- 3 (z'"- oc z"- z') = c;- 2 e- ifrJ.dt (-~ oc'- ! oc2 - ).
If k = const., the curve is called an affine conic. Hence the n. a. s.
conditions for a quasi-plane curve to be an affine conic are

(3.26) 31 oc " - 32 ococ 1 + 27

4 oc + 2 oc - ' = 0 .
3 J

If the connexion is transformed projectively we have for any vector v"

along the curve (cf. 1.1}

(3.27}

hence, for any functions 'oc(t) and '(t)

'!53;" ' 't5';" ' d;"
dt3- oc-dtz- dt
cj3;H I f52;>< 1 2 d~"
(3.28) = dt3- ( oc- 3p) dt2 - ( + 3p1 --2ocp -2p -q) dt;

P-def d;~' p. q def d;~' ~}1'

dt I'' dt dt .
300 VI. Projective and conformal transformations of connexions.

Equating this to zero we get on account of (3.23)

1
0C=rx+3P
(3.29)
= + 3p
{
1 1
- q- 2p 2 - 2rxp.

From (3.23) and (3.29) it follows that (cf. 1.1)

.!_ 1 II _ 2_ I
_ I I+ _ 4 I J-+- ~I (1.., l _ l l _ .!__ II _ ~ I I
3 rx 3 rx rx 27 rx , 3 - 3 rx 3 rx rx '
I
(3 .30) + ~rx 3 + ~rx- + 2 d~,.. 1 !5 2 ~;.2 p _!_rxiJ +Il 1

27 3 dt dt I' A 3

Il def d~l' d~).

II- - P,..r----a:t----;[1 (cf. 1.3).
Hence k remains a constant for all conics if and only if P,..;. = 0. This
proves the theorem 1 ) (cf.1.J):
A projective transformation (1.1) of the connexion in An changes all
affine conics into affine conics if and only if P,..;. = 0, that is for n :?:;2,
if and only if the transformation leaves the curvature tensor invariant.
That means that an affine conic is in general not invariant for all
projective transformations. 2)

Exercises.
VI 3, 13 ). If a connexion in Ln suffers a transformation that preserves
parallelism of directions, the transformation of the connexion induced
in a rigged xm has the same property.
VI 3,2 4). Among all parameters T,..";. in Ln with the same parallelism
of directions there is at most one set that is symmetric in fl /..

4. Projective connexions. 5)

T. Y. THOMAS was the first to remark that the expressions

(4.1) rP"' ctet- F"' -

!lA - f.'A
2
+
-n- - - -
1
A"(
f.'
n
1 A')

1) YANO and TAKANO 1944, 2, p. 424.

2 ) HoMBU and MIKA MI 1941, 1 and MIKAMI 1941, 2 defined projective conics.

Cf. HoMBU and MIKAMI 1942, 1 ; Y ANO and T AKANO 1944, 1 and 1949, 3 also
for literature. It can be proved that an affine conic is transformed into a projective
conic by every projective transformation of the connexion.
3 ) BoRTOLOTTl 1931, 2, p. 19ff.

4) BoRTOLOTTI 1931, 4.
6 ) References. General: VEBLEN 1922, 1; 2; 1928, 1; 1929, 1; T. Y. THoMAs

1922, 1; 1925, 2; 3; 1926, 2; 1927, 3; 1934,2, Ch. III; VEBLEN and T. Y. THOMAS
1923, 1; J. M. THOMAS 1925, 2; VEBLEN and J. M. THOMAS 1926, 1; WEYL 1929, 2;
WEYL and RoBERTSON 1929, 1; WHITEHEAD 1929, 1; GOLAB 1930, 1 ; ScHOUTEN
 4. Projective connexions. 301

are invariant for the projective transformations (1.1) of a symmetric

connexion. They are called the projective parameters ofT. Y. THOMAs. 1 )
Their transformation is

(4.2)
P",
J;').' =
p><' P" ><'
A";:,. J;;. +Ae 8", A",
e
+ ~--
2
n+l
><'
A(p' 8).') log Ll.
p
Hence the J;";. are components of a geometric object of a more com-
P
plicated kind than the J;";.. Therefore the J;J. do not fix one linear
connexion but only an infinity of symmetric linear connexions that
transform into each other by projective transformations and that all
belong to the same system of geodesics.
If we first prefer one coordinate system and then consider coordinate
p
transformations with Ll = const. only, the F"";. transform in the same way
p
as the J;";. and in this case the J;";. represent the only connexion that
has the same geodesics as the connexion I;,";. and whose parameters
give zero if transvected with A~.
p
The I;,";. can be used to investigate an affine geometry that is given
to within projective transformations. On every geodesie a projective
parameter can be defined that is fixed to within an affine or a projective
transformation with constant coefficients. The equation of the geodesics
can be written in the form (cf. III 7.2)
d 2 ,;>< P" d,;P d,;t. d,;>< 2 " d,;P ae
(4.3) ~ +T;.--- =cx(t)-- --AI';.-----
dt2 " dt dt dt n +1 " dt dt

with an arbitrary parameter t. If a new parameter C is introduced

and if we write C' def dCJdt; C" ~ei d2 Cjdtlz, (4.3) takes the form

(4 .4)
d2,;"
dC 2 +
r" ---ar- df - 1'
p
d,;P de _ 1
(X
d/;"
-d, - n
2
+1
F, df; d,;"
).
C"
---ar- df- (C'l 2
d,;><
d(
and the right hand side of this equation is zero if and only if Cis a solu-
tion of
(4.5)

and GOLAB 1930, 3; EISENHART 1930, 1; BORTOLOTTI 1931, 6; 1932, 1; 1933, 2;

V.DANTZIG 1932, 1; 2; 3; 1934, 1; HLAVATY 1933, 1; ScHOUTENandv.DANTZIG
1933, 2; CARTAN 1934, 1; CARTAN 1935, 1; 1937, 1; 3; BERWALD 1936, 1; ScHOUTEN
and HAANTJES 1936, 1; BORTOLOTTI and HLAVATY 1936, 2 (Jiterature); HAANTJES
1937, 1; KANITANI 1941, 1; YANO 1942, 4; CHI TA YEN 1948, 1. Imbedding:
BORTOLOTTI 1941, 2 (Jiterature); BOMPIANI 1943, 1; KIMPARA 1943, 1; KANITANI
1943, 1; 2; 1947, 1; 2; 1948, 1; 1949, 1; 1950, 2; NoRDEN 1945, 3; 1947, 1; 1948, 1;
1949, 1; BoL 1950, 1. Mapping: EISENHART and KNEBELMAN 1927, 2; WHITE-
HEAD 1929, 1 (Dn); KANITANI 1947, 2; CZECH 1949, 1; 1950, 1; 2.
1 ) T. Y. THOMAS 1925, 3. The special case LI=+ 1 was called by him equi-

profective.
302 VI. Projective and conformal transformations of connexions.

hence
(4.6)

2 r. dlo
Now it follows from (1.1) and (III 7.2) that oc dt- n + 1 A s- 1s rn-
variant for projective transformations of J;.'A, hence l; is a projective
invariant. It is the projective parameter on a geodesie introduced by
p
T. Y. THOMAs. 1) If J;.'A is given, l; is determined on each geodesie to
within an affine transformation. It is remarkable that l; is not a scalar.
If the coordinates are transformed we get
_2_Jr;:d<-'' - 2 -Jr.<d<-' - -2-Jdiogd
(4.7) en+l = en+l e n+l

hence
("') 2 (")
(4.8) dl;=Lfn+l dl;

and this proves that dl; is a scalar Lf-density of weight - - 2- . If a

n+l
projective parameter on a curve is fixed in any way this parameter
can be used to build up the projective differential geometry of the
curve, generalized FRENET formulae etc. 2)
p
But the r;A
can also be used in a more general way. The most
simple method is to construct a scalar density that is a concomitant
p
of the r;...
3) Such a density is invariant for projective transformations

of the connexion. It fixes a set of coordinate systems with respect

to which it has the component +
1. These coordinate systems trans-
P
form into each other with L1 = + 1
and accordingly the now fix r;A
a linear connexion that is projectively invariant. The disadvantage of
the method is not that it would be difficult to form a projectively in-
variant density, in fact there are many of them, but there is no one
specially preferred and they all depend on higher derivatives. CAR-
TAN4) proposed in 1924 to use a flat projective space instead of an affine
space as local space. ScHOUTEN 5 ) proved in the same year that this
amounts to introducing a linear connexion for local projective spaces
and that it is intimately connected with the linear connexions of local
T. Y. THOMAS 1925, 3; cf. VEBLEN and J. M. THOMAS 1926, 1.
1)
WHITEHEAD 1931, 2 has introduced a scalar projective parameter deter-
2)
mined to within a "projective" transformation and BoRTOLOTTI 1932, 1 has used
another one. Cf. HLAVATY 1931, 1; BERWALD 1936, 1; HAANTJES 1937, 1; SA-
SAKI 1937, 1; YANO 1937. 2; 3; 1944, 3; 4; Eil 1938,2, p.194.
3 ) WEITZENBCK 1932, 1.

') CARTAN 1924, 1.

5 ) ScHOUTEN 1924, 2.
 *4. Projective connexions. 303

EN's; N=J=n already introduced by KNIG in 1920. 1) EISENHART, VEBLEN

and T. Y. THoMAS first considered the generalized projective geometry as
"geometry of paths" 2), 3 ) but soon after the introduction of the pro-
jective parameter of T. Y. THOMAS several authors developed theories
that could all be considered in some way as theories of linear connexion
of local En+l's with a restricted tiansformation group. 4) Another line
of thought found its origin in the idea of v. DANTZIG 5) of using in Xn
a set of n + 1 homogeneous coordinates subjected to homogeneous trans-
formations of degree one. Then local space is automatically a flat
projective space and for the vectors in this space (called profectors) a
linear displacement can be determined uniquely. 6), 7 )

Exercise.
* with the transformation
VI 4, 1. For every choice of an object I';,

VI) 4,1 ot)

* defined by
the rl'";.
VI 4,1 )

transform like the parameters of a symmetric linear connexion and

*;.
J;.;. * (cf. Exerc. VI 7,1).
=Fp
1) KNIG 1920, 1.
2) EISENHART and VEBLEN 1922, 3; VEBLEN 1922, 1; 1928, 1; VEBLEN and
T. Y. THOMAS 1923, 1.
3 ) DouGLAS gave two generalizations. In 1928, 1 he considered systems of
curves such that one and only one curve passes through every point in every
direction. A general system of this kind can be defined as solutionsofadifferential
equation of the second order. This theory is intimately connected with FINSLER
geometry. Cf. T. Y. THOMAS 1934, 2, p. 7, also for literature. In 1931, 1 DouGLAS
considered systems of XK's (K-spreads) suchthat one and only one passes through
every point in every K-direction. There are many recent investigations on the
geometry in the space of K-spreads especially by chinese and japanese authors.
These theories are beyond the realm of this book.
4 ) T. Y. THOMAS 1925, 3; 1926, 2; 1934, 2; WEYL 1929, 2; WEYL and Ro-
BERTSON 1929, 1 ; GOLAB 1930, 1 ; SCHOUTEN and GOLAB 1930, 3; BORTOLOTTI
1932, 1 (Jiterature); HLAVATY 1933, 1 (Jiterature); BERWALD 1936, 1.
6) V. DANTZIG 1932, 1; 2; 3.
6} V. DANTZIG 1932, 3; SCHOUTEN and V. DANTZIG 1932, 1; 2; ScHOUTEN
and HAANTJES 1936, 1 (literature).
7) Cf. for the relations between the different points of view also CARTAN 1924, 1;
SCHOUTEN 1924, 2; 3; WEYL 1929, 2; WEYL and ROBERTSON 1929, 1; V. DANTZIG
1932, 1; 2; 1934, 1; YANO 1937, 3; 1938, 4; 5; 6; 1940, 4; 1941, 6; 1942, 4; 1944, 3;
8; 9; 10; 11; 1945, 4; 1947, 1; HLAVATY 1939, 1; YANO and SASAKI 1949, 4.
304 VI. Projective and conformal transformations of connexions.

5. Conformal transformation of a connexion in v...

If the fundamental tensor g.:~,. in V,. is transformed into
-1
(5.1) 1g.:lx = v,. g><'. 1g"A = a-1 g".:t 1)

l
where a is an arbitrary scalar, the parameters I;..".:~ transform into

T;A = i 1(-a (81' 1g).a + 8.:~ 1gpa- 8a 1gpA)

(5.2) = r:.:~ + t (AJ: SI'+ A~ S;t- g"a gpA sa);
S;t def 8Alog a. 2)

We call this transformation conformal. From (5.2) and (III 2.3, 7)

we get
"JA v" = V.JA v" + J2 sp v" +
a) 117 . 1'2" s.I vA A"p - 21 g"a sa gpA v.:t
(5-3) {
b) '171' w.~ = 17"' w.~- w(p S)+ t gp). Sa gaT WT.

In dealing with these transformations the process of raising and lowering

should be used very carefully because there are now two fundamental
tensors g.~,. and 'g.." and the process can only be defined with respect
to one of them. From (5.3) it follows that

(5 .4)

If a transformation of the r;"

in V,. is at the same time projective
and conformal we have from (1.1) and (5.2)

(5.5) 2pl' AJ: + 2p.~A~ = AJ:sp + A~ S.<- g"a gpA Sa

from which by transvection with g~'" and by contraction with respect
to u. two equations can be derived which are consistent if and only if
s.~ = 0. Hence

A riemannian connexion is wholly determined if its geodesics are

given and the fundamental tensor to within a scalar factor. Then the
fundamental tensor is also fixed to within a constant scalar factor 3)
or in other words
The metric of a Vn is determined to within a constant scalar factor
by the projective and the conformal properties of the V,. together.
1) Since raising and lowering of indices is performed by means of g"A and gA><
-I
1
g"A can not be written as g""1

2) FUBINI 1909, 1, p. 144; ScHOUTEN 1918, 1; WEYL 1918, 1.

8) WEYL 1921, 2, p. 100.
 5. Conformal transformation of a connexion in Vn. 305

If (5 .2) is substituted in (III 4.2) it follows that

(5.6)
where
(5 .7)
and accordingly
(5.8) 'K~'-;. = Kp;.-! (n- 2) s~'-;.- i g~'-;, sarg'1T

(5.9) 'K = a- 1 K- t (n- 1) a- 1 spl. g~'-;.

(5.10) 'Gp;. = Gp;.- t (n- 2) sp;. + t (n- 2) gp;. sar g"r.

A Vn and its connexion are said to be conformally euclidean and the

Vn is called a Cn if it can be transformed conformally into an Rn. 2 )
N. a. s. condition is for n > 2 that there exists a vector field L;. such that

a) K.p;.,. 4 '
1
= n _ 2 g[v[J. LPI><J; L[p) = 0
(5.11)
b) - -4- L ;.=2V. L_.-L L,.+-1 g ;.L La
n-2 1'- 1'- 1'- 2 1'- a

(the factor n ~ 2 is introduced only for convenience). The number of

the independent components of K.p;.,. is -f'l n 2 (n 2 -1) and this is
> (n ;-
1) for n
>3, hence (5 .11 a) can always be satisfied for ~ 3 but n
for n >3 only if satisfies certain algebraical conditions. In this
K~'-'""

case we get from (5.8) for sp;. = - n ~ 2 L 1..;. and 'Kp;. = 0

1
(5.12) Lp;. =- Kp;. + 2(n-1) K gpi.

The first integrability condition of (5.11 b) is

(5.13)

from which we get by substituting (5 .11 a)

(5.14)
1 ) Notethat a- 1 'Kvpi.><=a- 1 'g,.Q'K;;,}.e=Kvpi.o<+g[v[.<Sp]><]; cf. STRUIK
1922 1, p. 151; FrNzr 1923, 1; ScHaUTEN and STRUIK 1921, 4; R. K. 1924, 1,
p. 168, footnote 3).
2) Cf. for the geometry of a Cn: HAANTJES and WRONA 1939, 1; HAANTJES
1940, 1; 1941, 1; 1942, 1; 2; 3; 1943. 1; YANO 1945. 4; 5; 6; 1946, 2; 3; WRONA
1948, 1; 2; 1949, 1; 1950, 1; LEVINE 1950, 1 (parallel fie]ds); MATSUMOTO 1951, 3
(dass one).
Scheuten, Ricci-Calculus, 2. Aufl. 20
306 VI. Projective and conformal transformations of connexions.

For an S,., n ~2 we have (cf. III 5.31) a solution of (5.11 a)

(5 .15) L,_.;. = -! (n- 2) x g,_.;.
and for n >2 this is the only solution. Because x is constant (5.14) is
satisfied and this implies that every S,. is a C,.. For a V2 which is not an
5 2 (5.15) is also a solution of (5 .11 a) but it does not satisfy (5 .14) because
x is no Ionger constant. But in this case (5.11a) has an infinite nurober
of solutions and it is well known from elementary differential geometry
that every V2 can be conformally mapped on an R 2 1 ) For a V,. the
identity of BIANCHI (III 5.21) can be applied to (5.11a). Then we get
(5.16) (n- 3) 17[.L,_. 1;. = 0.
Hence for n >3 the integrability condition of (5 .11 b) is a consequence
of (5.11 a). That. proves 2)
A V2 is always a C2 A V,., n >2. is conformally euclidean if and
only if
(5 .17) .. ; .. ~
C.,_.;. defK" 4 g L nll"
.,_.;. - - - [['- !-']Q] 15 =0
n-2
where
(5.18)

and if for n = 3 moreover

(5 .19)
For n=3, (5.17) is identically satisfied and for n>3, (5.19) is a con-
sequence of (5 .17).
C;;)." is called the conformal curvature tensor. From (5.8,12) it
follows that
(5.20) 'LP;. = L ,_.;. +-n-2
4- s,_.;.
and by substituting this in (5.17) it follows that c;~;." (but not c.,_.;.")
is invariant for conformal transformations of the connexion. 3 ) For
n:;;;; 3 it vanishes identically and for n >3 it satisfies the identities

(5.21) {
a) C "-0
(vp)A - 1
b) C[;~;.t = 0;
C) =o
C.vpA d) C;~;.'"=O; e) C;;;."= -C;;~;.-
1) Cf. for instance WEATHERBURN 1927, 1, p. 168; HAACK 1948, 1.
2) SCHOUTEN 1921, 1, p. 82ff.; COTTON 1899, 1, p. 412 and FINZI 1902, 1
for n=3. Cf. FINZI 1921, 1. WEYL proved 1918, 1, p. 404 that C;~,i." is zero in
a C,.. FINZI proved 1921, 1 that (5.17) and (5.19) together are n.a.s. for the Vn
tobe a C,.. Cf. also FINZI 1922, 1; CARTAN 1922, 3 for n = 4 and LAGRANGE 1923, 1,
p. 43 for n =!= 3. DouGLAS gave a synthetic criterion 1925, 1.
S) Cf. for the semi-symmetric case ScHOUTEN 1925, 2.
 5. Conformal transformation of a connexion in v... 307

As a corollary we get that for a conformally euclidean connexion the

curvature tensor has at most ! n (n + 1) independent components.
From (5.11a) we see that for a c.. all components of K.,_.A" with
respect to an (holonomic or anholonomic) orthogonal coordinate system,
with four different indices vanish. It can be proved 1) that for n> 3
this condition is also sufficient:
A n> 3, is a c.. if and only if with respect to every holonomic
v.. ,
or anholonomic orthogonal coordinate system (h)
(5.22)
at every point. 2)
Another form of the n.a.s. conditions is found as follows. We
consider here only an ordinary V.. and. real subspaces. Let a V,. be such
that it contains orthogonal systems (cf. definition V 5) in every direc-
tion, i.e. that for every set of n directions at any arbitrarily chosen
point there exists at least one orthogonal system having just these
directions at that point. Then writing i"; b = 1, ... , n -1 and i" for
b n
the unitvectors in the directions of this system we have (cf. V 5.16b)

(5.23) 0 -- 2 t..t ,_. tA 17 h - ...,_. K .,_..;.,tA t." =* K dcbn.

"[ p.].;-t t b, c, d =l=
dcb dc bn

from which we see that all orthogonal components of K.,_..;" with four
different indices vanish and that accordingly for n> 3 the V.. is a C.,.
Now an R., contains of course orthogonal systems in every direction
and orthogonal systems remain orthogonal under conformal trans-
formations. Hence
A V,., n > 3 is a C" if and only if for every choice of a point and an
(n -1)-direction at this point, it contains at least one V.. _. 1 belanging to
an orthogonal system and having fust this (n -1)-direction at this point. 3 )
In V 6 it was proved that if in a V,. there exist umbilical V.. _ 1 's
through every point and with every (n-1)-direction at this point, all
orthogonal components Kkiih with four different indices vanish. But
we just proved that for n > 3 this means that the V.. is a C,.. N ow let an
anholonomic coordinate system with mutually perpendicular unit-
vectors i," be chosen such that i" is normal to a set of cxi V.,_ 1 's. Then
1 n
we have for the second fundamental tensor heb of these vn-1's
(5.24) -
hcb- -
P.J7
cb p. t.;, b, c = 1, ... , n - 1
n

1 ) SCHOUTEN 1921, 1, p. 84; cf. EISENHART 1926, 1, p. 122ff.

2) Cf. Exerc. III 5.5.
3 ) ScHOUTEN 1927, 3, p. 719; E II 1938, 2, p. 204 also for n = 3 and for literature.

20*
308 VI. Projective and conformal transformations of connexions.

cllld after the transformation g;."-+ag"" according to (5.3 b), denoting

the components of the new heb with respect to the old unitvectors bi"
by 'heb

This proves that the directions of principal curvature are conformally

invariant and that all umbilical points remain umbilical points after
the conformal transformation. Now in an Rn, umbilical V,._ 1 's are
possible through every point with every (n-1)-direction at that point
and this implies that the same holds in a Cn. So we have proved for
n> 3 that the existence of these Vn_ 1 's is a necessary and sufficient
condition for the Vn to be a Cn. STELLMACHER 1) has proved that the
condition is also n.a.s. for n=3. We prove this in the following way:
Let rj""; I)( = i, 2 be Coordinates in an umbilical v2 in Va. Then we
have on the one hand from (V 5.9)

(5.26)

if j" is a congruence normal to the V2 On the other hand we have

because all points are umbilical

(5.27)

But in a V3 the equation (5 .17) is identically satisfied, hence from

(5 .26, 27)
(5.28) B6~~ (g,;. LJJ."- gJJ.;. L,") j" = 2 'g[y 861h
or
(5 .29)

and this implies that B~ L;." f is a gradient vector in V2 :

(5.30)

The rotation of this vector is zero and equal to the V2-part of the rota-
tion of L;."f

(5-3 1) l 0 = B Jl.Ar? L ."

yV[Jl. !.J><J =
Jl.A "><l7 L Jl.A L"
Y! V[Jl. !.]><- Y ,[!.
hJJ.]><

= B~~ f r:(Jl. L;. 1" - h B~~ L~ [.< gJJ.l" = B~~ f JlrJl. L;.1".
1) STELLMACHER 1951, 1.
 5. Conformal transformation of a connexion in V,.. 309

Hence, if in V3 there exist umbilical V2's through every point with

every 2-direction, all orthogonal components of ~,.L.~1 ,. with three dif-
ferent indices vanish. Frorn this we see that from the quantity (cf. I 9.20)

(5.32) det I"" 17 L

Nv><= v V[p ]><

all orthogonal components with two equal indices vanish. But this is
only possible with respect to all orthogonal systems if N",. = N[xl
Returning now to the identity of BIANCHI (III 5.21) we have according
to (5.17)
(5.33) 0 = g;,[ V., L,_.1,.- g,.[.V., L,_.p.
or, transvecting with !""'"
(5.34)
which implies that in a V3 of the kind considered 17[,_. L.~ 1 ,. = 0.
Gathering results we have 1 )
A V,., n >2, is conformally euclidean if and only if there exist um-
bilical v,._l's through every point with every (n -1)-direction at that point.
If a Vm in V,. can be transformed into a geodesie Vm by a conformal
transformation of the connexion in V,. it is called conformally geodesic.
If m = 1 we have for the tangent unitvector (d >0)

(5.35) 'i" = a-lj";

hence (cf. V 1)
(5.36) 'u" = 'i" 'V,_. 'i" = a-1 (u" - i z"),
where u" is the curvature vector and z" is defined by

(5.37)
is the projection of s" on the R,._ 1 perpendicular to i". If the transformed
curve is geodesie it follows that u" is the projection of a gradient vector.
Conversely, if we know that u" is the projection of a vector g"" 8.1 log a,
the transformation g.~,.~ag.~,. transforms the curve into a geodesic.
1) For n > 3 this was proved by ScHOUTEN 1921, 1, p. 86. But for n = 3 it
was stated there erroneously that every V3 contains umbilical V8 's through every
point in every 2-direction. This wrong statement occurred also in the first edition
of this book R. K. 1924, 1, p. 180. But it was not republished in E li 1938, 2 19
because at that time we smelled a rat without knowing exactly its lare. STELL-
MACHER gave the solution in a letter of 19. 11.' 49 to the author and published
this with many other results concerning conformal geometry in 1951, 1; cf. SA-
SAKI 1939, 1 ; 1940, 1.
310 VI. Projective and conformal transformations of connexions.

Hence 1)
Every single eurve is eonformally geodesie. A real eongruenee of
eurves in an ordinary V,. is eonformally geodesie if and only if the eur-
vature veetor is the projection of a gradientveetor.
Orthogonality being invariant for conformal transformations, the
theorem of BELTRAMI (cf. V 5) leads to the theorem of THOMSON
and TAIT 2)
I f a real eonformally geodesie eongruenee in an ordinary Vn is normal
to one vn-1 it is vn-1-normal throughout.
But such a congruence is not in any way special because it can be
proved that every vn-1-normal congruence is conformally geodesic. 3 )
Because of
(5.38)

every null geodesie (geodesie whose tangent lies everywhere in a null diree-
tion) remains a null geodesie after eonformal transformation of g;.".
For m>1 we get [cf. (5.37) and V 9] 4 )

(5.39) l 'H"
cb =- B'";.
cb

b,e=1, ... ,m;

('Vp 'i)
J. 'i"
" = H"-
cb
1
2

x=m+1, ... ,n;

'gcb z"

hence
An umbilical point of a Vm in Vn remams umbilical after conformal
transformation of g;,".
This proves that a conformally geodesie Vm is for m >1 the same
as an umbilical Vm. But this does not imply that a system of oon-m
umbilical Vm's; m=1, ... , n-1, can be transformed at the sametime
by one conformal transformation of the g;," into oon-m geodesie Vm's.
According to (5-39) this is possible if and only if the mean curvature
vector (cf. V 9) is the projection of a gradientvector. 5 )
According to (5.39) the tensor M;i" defined in (V9.16b) is con-
formal invariant. The same holds for the vectors .ffb defined in (V 8.13).
--------- y
1) ScHOUTEN 1928, 2.
2) THOMSON and TAIT 1879, 1, p. 353 for n = 3 and SCHOUTEN and STRUIK
1921, 2 for general values of n.
3) ScHOUTEN and STRUIK 1921, 2; 1922, 4; E II 1938, 2, p. 51; ScHOUTEN
1928, 2. Cf. for families of curves that may be considered as conformal geodesics
PAINLEVE 1894. 1; KASNER 1910, 1; 1913, 1; DouGLAS 1924, 1; ScHaUTEN 1928, 2;
BLASCHKE 1928, 1; FIALKOW 1939, 2; 1940, 1; 1942, 2; YANO 1950,1.
4) ScHOUTEN and STRUIK 1923, 4; R. K. 1924, 1, p. 202.
5) E II 1938, 2, p. 211.
 5. Conformal transformation of a connexion in v... 311

For a V". they take the form

(5.40)
hence
(5 .41) '"~b-
-B"'('V.''!),.-'_"
b p t.< ; - ~b
1)

In (V 8) it was proved that an A", can be imbedded in an E .. if there

exist fields heb z;a and ~b satisfying the equations (V 8.18-21) of GAuss,
:Y :Y "
ConAZZI and Riccr. For a V". in Rn the fields heb and leb coincide and
:Y "
accordingly a V". can be imbedded in an Rn if there exist fields heb
and tb satisfying the GAuss-CoDAZZI-RICCI equations. YANO and
:Y
Muro 2) proved that a V". with the fields .1\(b(x) and ~b can be conformally
imbedded in a c.. provided that these fields satisfy three equations
:Y

that may be considered as the conformal generalizations of the GAuss-

ConAZZI-RICCI equations. These equations were derived before by
YAN0. 3 )
In V 6 we have found that an umbilical v.. _1 with constant mean
curvature (h = const.) is at every point perpendicular to a principal
direction of V,. and that accordingly V,._ 1 's of this kind are only possible
through every point and with every (n -1)-direction at that point
if the v.. is an EINSTEIN space, KJJJ. cx gJJJ. But then also LpJ. equals
gJJ-' to within a scalar factor and according to (5.17) the V.. is an s.. 4 )
(cf. V 6):
A V.,, n>2, is an 5., if and only if there exist umbilical V"_ 1's with
constant mean curvature through every point with every (n -1)-direction
at this point.
A geodesie V.. _ 1 is a special case of an umbilical V,._ 1 (h = 0), hence
A V,., n >2, is an Sn if and only i/ there exist geodesie V"_ 1 ' s through
every point with every (n -1)-direction at this point.
According to (5.7, 12) an R,. is transformed conformally into another
Rn if s.o~ = a..~ log a satisfies the equations

(5 .42)
This equation is totally integrable 5).
1) SASAKI 1936, 1.
2) YANO and MuTO 1942, 1; 1946. 1.
3) YANO 1939, 2; 3; 1940, 3; 1943, 2.
4) ScHOUTEN 1921, 1, p. 87.
6) Cf. for R., transformed conformally into R.,, E li 1938. 2, p. 205ff.
312 VI. Projective and conformal transformations of connexions.

LIOUVILLE proved the theorem

For n >2 every conformal mapping of an R,. on itself is always a
similarity transformation with respect to a point or a combination of an
inversion at a point and a similarity transformation with respect to that
point
for n = 3 and LIE gave the generalization for n >3. 1) The conformal
transformations of an R 2 that leads to another R 2 are well known from
the theory of functions. 2 )

Exercises.
VI 5,1 ). A v2 is transformed conformally into an R2. If gA_ .. = (1 g.,"
3 I

and sA = oA log a, prove that

VI 5,1 oc)

VI 5,2 [cf. Exerc. V 9,4] 4). A V2 m is a C2 m if and only if the sum

of the scalar curvatures of 2 arbitrary mutually perpendicular m-direc-
tions is always independent of the choice of these m-directions.
VI 5.3 5). Anormalsystem of oo"-m Vm's in a V,. can be transformed
conformally into a system of oon-m minimal V,,.'s if and only if the
mean curvature vector is the projection of a gradient vector on the
local R.,_m perpendicular to the Vm.
VI 5,4 6 ). If a congruence in Vm in V,. is given, there is always a
conformal transformation in V,. that transforms the congruence into
a congruence of asymptotic lines of the vm.

6. Conformal transformations of the connexion

in an EINSTEIN space. 7)
A V.,, n>2, is transformed into an EINSTEIN space by a conformal
transformation of the connexion if and only if

(6.1) 'K = a- 1K- ~ (n -1) a- 1 s,.Ag,..A = const.

1) LrE 1872, 1; BIANCHI 1902, 2, p. 375. The proof is not difficult but rather
tiresome and uninteresting. A generalization was given by HAANTJES 1937, 2.
Cf. E II 1938. 2, p. 209.
2) Cf. E II 1938, 2, p. 210.
3 ) E II 1938, 2, p. 205.
4) HAANTJES and WRONA 1939, 1.
5 ) ScHOUTEN and STRUIK 1923, 4; E II 1938, 2, p. 212.
6) ScHOUTEN and STRUIK 1923, 4; R. K. 1924, 1, p. 202.
7 ) General references: BRINKMAN 1923, 1; 1924, 1; 1925, 1; HAANTJES and
WRONA 1939, 1; FIALKOW 1939, 1; 1942, 1; SASAKI 1942, 1; YANO 1943, 1; WONG
1943, 2; 3; WRONA 1948, 2; KurPER 1950, 1; 2; 1951, 1.
 6. Conformal transformations of the connexion in an EINSTEIN spacc. 313

On account of (5 .12) this is equivalent to

(6.2)

hence, according to (5.20)

(6.3)

This gives for a the differential equation (cf. 5.7)

(6.4) 2 V,,r s_.- s"r s_. + .!__2 g";. sa sa =

r
- ~ 4
n _-2- L";.-
r
2
n (n _ J) 'K a g"_.
r

with the first integrability conditions (cf. 5.13)

(6.5)

These conditions can also be derived immediately from (6.2):

(6.6)

because the left hand sides of (6.5) and (6.6) are equal.l)
Now from (III 5.27) it can be derived that2)

(6.7) C " = - 2(n-3)

V"vp.i. 17 L
-2-V[v p.].<
n-

and from (6.5) and (6.7) after some calculation

This equation is a part of the second integrability condition of (6.4).

It is remarkable that the right hand side is zero for n = 4. That im-
plies that for n = 4 the left hand side must be a conformal concomitant
to within a scalar factor. In fact it can be proved that for n = 4 the
symmetric tensor density of weight +!

(6.9)

1) E II 1938, 2, p. 203.
2) E II 1938, 2, p. 204.
314 VI. Projective and conformal transformations of connexions.

is invariant for conformal transformations of the connexion. 1 ) For

n = 4 the vanishing of ~.A is a necessary though not a sufficient con-
dition for the V4 to be conformally transformable i.nto an EINSTEIN V4 .
Returning to the first integrability conditions (6.5) for n >2 we
first remark that (6.4) is totally integrable if and only if C;~ ). " and
17[. L!-']A vanish, that is if the vn is conformally euclidean. But in
that case the EINSTEIN space is conformally euclidean as well, and
from (5 .17) it follows that a conformally euclidean EINSTEIN space is
always an Sn' 2 ) Now let us suppose that the vn is a special EINSTEIN
space, Lf-'A=O. Then (6.5) reduces to

(6.10) K;~;,"s" = 0; s" =j= 0

because in this case C;~;," is equal to K;~;.'' according to its definition.
Now (6.10) implies that K.~-';." can not have the "-rank n. But the rank
of K.~-';." is the same with respect to all indices, hence at every point
Kvf-IA" lies in an Rn-l perpendicularto s". Forn =4 this has an important
consequence. According to its symmetry properties K.~-';." can always
be written in the form 'a[[A Qf-ll><] where 'a;." is the fundamental tensor
of the Ra perpendicular to s", a;." = a;."- e S;. s"; e S;. S= 1. But because
I

K;~l" = C;~l", the RICCI tensor K!-'A must vanish and this is only possible
if Q;."=O. This proves the theorem of BRINKMAN: 3 )
A conformally euclidean special EINSTEIN space is always an Rn.
I f a special EIN STEIN. space is not conformally euclidean then for n = 4
it is impossible to map it conformally on another specialEINSTEIN space 4 ).
This is very important for general relativity. Time-space is a special
EINSTEIN space in all regions without matter. In these regions the
metric can only be investigated by means of light signals and these
signals fix only the null geodesics. That would imply that in empty
time-space a metric could only be determined to within conformal
transformations. But according to the theorem of BRINKMAN the con-
dition GI-';.= 0 (usually derived from a variational principle) is sufficient
to fix the metric 5 ).
1) This quantity was discovered by ScHOUTEN and HAANTJES 1936, 2. It
played an important role in the imbedding of an EINSTEIN V4 in a five-dimensional
space with a general projective geometry and it was rediscovered by HESSELBACH
1949, 1. He proved that ~ [~A = 0.
2) SCHOUTEN and STRUIK 1921, 4.
3 ) BRINKMAN 1923, 1; 1924, 1; 1925, 1 also for many other properties of con-

formal mapping of ErNSTEIN spaces; cf. E II 1938, 2, p. 213; KurPER 1950, 1; 2

also for Iiterature on ErNSTEIN spaces.
4 ) It is also impossible to rnap it conforrnally on a non special ErNSTEIN space.
6) This remark is due to KASNER 1921, 1; 1922, 1 who proved the theorem
of BRINKMAN for a special case.
 7. Conformal connexions. 315

7. Conformal connexions. 1)
If the (symmetric) fundamental tensor g"" is given to within a
2
scalar factor the (symmetric) tensor density of weight
n

(7.1)

is uniquely determined. Its determinant (or in the indefinite case the

absolute value) is + 1. This has as a consequence that on each curve
a conformal parameter C is defined whose differential

(7.2)

is a scalar density of weight - _!__ 2 ) If the process that leads from thc
n
g"" to the ~";. belanging to g""' is applied to the @"" we get

(7.3)

These are the conformal parameters of J. M. THOMAS 3 ), which are

invariant for conformal transformations of g;.". Their transformation is

(7.4)

1) CARTAN 1923, 3; 1937. 3; ScHaUTEN and STRUIK 1923, 4;

References. General:
T. Y. THOMAS 1926, 3; 1932, 1; 1934. 2, Ch. IV; VEBLEN 1928, 3; 1930. 1; 1935. 1;
WHITEHEAD 1929, 1; VANDERSLICE 1934, 2; HLAVATY 1935, 2; 1936, 1; SCHOUTEN
and HAANTJES 1936, 2; YANO 1938, 2; 1939. 4; 5; SASAKI 1940, 2; 1941, 2; 1948, 1;
SCHIROKOW 1940, 1; VRANCEANU 1940, 1; 1943, 1; YANO and MUTO 1941, 1; 2;
ScHMIDT 1949, 1. Curves: AGOSTINELLI 1933, 3; HLAVATY 1935, 2, Ill; MoDE-
SITT 1938, 1; YANO 1938, 3; 1944, 4; 1950, 1; FIALKOW 1939, 2; 1940, 1; 1942, 2;
SASAKI 1939, 2; MUTO 1939, 1; 1942, 1; HAANTJES 1941, 1; 1942, 1; LAGRANGE
1941, 1; 2; YANO and MuTO 1941, 1; 2; 3; 4. lmbedding: BRINKMAN 1923, 2;
PEREPELKINE 1935, 2; HLAVATY 1935, 2, Il; SASAKI 1936, 1; 1939, 3; 1940, 2;
3; HAIMOVICI 1938, 1; FIALKOW 1939, 1; 1942, 1; 1944, 1; Y ANO 1939, 2; 3 (GAUSS
and CODAZZI); 1940, 3; 1942, 2; 1943, 2; 4; MUTO 1939, 2; 1940, 3; 1942, 1; YANO and
MuTO 1940, 9; 1941, 5; 1942,1; 1946, 1; HAANTJES 1942, 2; 3; 1943, 1; BECKENBACH
and BING 1943, 1; NoRDEN 1945, 3; PETRESCU 1946, 1; 1948, 1; TSUKI 1950, 2;
VERBICKII 1951, 1; 2 (class); IcHILIONE 1951. 1; AKIVIS 1952, 1. Product spaces:
YANO 1940, 1; WoNG 1943, 2. Mapping: GoLAB 1935, 1; HAANTJES 1937. 2;
ZAREMBA 1937, 1; COBURN 1941, 2.
2) T. Y. THOMAS 1926, 3.
3) J. M. THOMAS 1925, 1; 1926, 3.
316 VI. Projective and conformal transformations of connexions.

c
Hence the ~";. are not the components of a geometric object because
c
their transformation formula contains not only ~"J. and A;:' but also
-1 c
@pJ. (or gpJ. because @pJ. @"!.' = gpJ. g"!.>). But r;J. together with @pJ.
form a geometric object. This geometric object does not fix one metric
and symmetric connexion but an infinity of such connexions trans-
forming into each other by conformal transformations. If we first prefer
one coordinate system and then consider coordinate systems with
c
LI = const. only, the ~"J. transform in the same way as the ~"J. andin
c
this case the r;J. represent the only symmetric connexion that leaves
@_.,. invariant and whose parameters give zero if transvected with A!.
c
The J;.'A can be used in several ways to fix some kind of connexion.
The simplest method is, as in the projective case, to construct an in-
variant scalar density. 1 ) In 1924 CARTAN 2) proposed to use a flat pro-
jective (n + 1)-dimensional space with a quadric instead of an affine
space as local space. ScHOUTEN 3) proved in the same year that this
amounts to introducing a linear connexion for local conformal spaces
and that it is intimately connected with the linear connexions of local
EN's; N =l= n already introduced by KNIG in 1920. 4) T. Y. THOMAs 6 )
c
succeeded in constructing from the ~"J. a connexion for certain quantities
in an X.,+l with a restricted group of coordinate transformations.
It is well known that the conformal geometry of an R., is identical
with the projective geometry on a quadric in a flat projective (n 1)- +
dimensional space. This can be proved for instance by introducing
+
n 2 polyspherical coordinates. 6 ) This can be generalized for the con-
formal geometry of a V.,. ScHOUTEN and HAANTJES 7 ) have proved
that for n odd an X., with a general conformal geometry can be imbedded
in an X.,+l with a general projective connexion, and that this connexion
is uniquely determined. This means that n-dimensional general con-
formal geometry can for n odd be treated with n + 2 homogeneaus
Coordinates that satisfy one condition. For n even the theorem holds
only if special conditions are satisfied. For n = 4 the only condition is
that <;r_.,. = 0 (cf. 6.9).
1)
COTTON 1899, 1 for n=3; HLAVATY 1935, 2; LEVINE 1935, 1.
2)
CARTAN 1924, 1; cf. 1923, 3.
3)
ScHOUTEN 1924, 2; 3.
4)
KNIG 1920, 1.
6)
T. Y. THOMAS 1932, 1; 1934, 2 also for literature; cf. VEBLEN 1928, 2; 3;
1935, 1; VANDERSLICE 1934, 2; HLAVATY 1935, 2; 1936, 1; YANO 1938, 2; 1940, 4.
6) Cf. for instance Woons 1922, 1, p. 251; KLEIN 1926, 1, p. 193; ScHOUTEN
and HAANTJES 1936, 2.
7) ScHOUTEN and HAANTJES 1935, 3; 1936, 2.
 8. Subprojective connexions. 317

Exercise.
VI 7,1. For every choice of an object F,.* with the transformation

VI 7,1 oc) *
F,.. = A1- F,.* - 8;: log LI
* defined by
the ~";.
V I 7,1 ) r*"~tA-- rc"~tA + ~:, "- !__n gld.g"Qr*
n .L(~tAA) Q

transform like the parametersofalinear connexion leaving ;,,.invariant

. * *
and r:;. =~ (cf. Exerc. VI 4,1) .

8. Subprojective connexions.
An An is said to be k-fold proiective 1 ) if there exists a coordinate
system with respect to which every geodesie can be given by means
of k linear equations and n- k -1 equations that need not be linear.
Hence a Dn is (n -J)-fold projective. For k = n- 2 it may happen
that there exists a coordinate system such that every geodesie is given
with respect to this system by n- 2 homogeneaus linear equations
and one other equation that need not be linear. Then the An is called
subproiective 2 ) and the origin of the special coordinate system a pole. 3 )
At any point the direction of the geodesie through this point and a
pole is called a pole direction. The special coordinate systems with
respect to which n- 2 equations of any geodesie are homogeneaus and
linear transform into each other by the transformations of the group
(8.1) P:~ = const.
if they have the same origin, as was proved by RACHEVSKI and will be
proved here later on. 4)
n- 2 homogeneaus linear equations in one of these special coor-
dinate systems fix an X 2 that is totally geodesie in An because every
geodesie having two points in common with such an X 2 lies wholly
in it. Hence through the origin totally geodesie Xn_ 1 's can be laid
with every (n -J)-direction and every one of tliem is given by one
homogeneaus linear equation. In the following we do not consider the
case n = 2 because according to the definition every J-fold projective A 2
is a D 2 and every A 2 is subprojective.
1 ) KAGAN 1930, 1; 1933, 1; RACHEVSKI 1930, 1; 1933, 1; SCHAPIRO 1930, 1 '

1933. 1.
2) KAGAN 1930, 1; 1935, 1.
3) The case k<n-2 was recently discussed by VRANCEANU 1947, 2; 1950,
1; 2. A linear element was given for a k-fold projective Vn.
4) RASCHEVSKI 1933. 1, p. 128. Cf. for this and some of the following results
also E II 1938, 2, p. 215ff.
31R VI. Projectivc and conformal transfm:mations of connexions.

Now let (h) be such a special coordinate system and let v11 be an
arbitrary vector field with constant components v". The equations
(8.2)
represent the parameter form of a geodesie X 2 at every point of which
the vector v" has a tangent direction. But because the X 2 is geodesie
this implies that
(8.3)
for every choice of the direction of v11 Because the field v11 is arbitrary,
we may now look upon the vh and ~~~in (8.3) as independent variables.
Then by differentiation with respect to v11 we get

(8 4)
.
2E h
ii
*
=
Cf).
ovf ov1 V
II + a!if
(J). A" + (J). A" + th 0 p.
i 7iiJl i
2
ovl ov1 '
10

which is only possible if a:::v' = 0 because the lj~ depend on the ~"
only and not on the v". Hence evr
(J).
and ovlaz~.~
ovl do not depend on the V
"
and we get for lj~ an expression of the form
(8.5) Ifi * 2cp<;Ail +cp1d 11 ; cp[i<l =0
valid with respect to all special coordinate systems.l) If we apply the
coordinate transformation (8.1) we have

(8.6) { a)
Ai' * e p~. + ~~~ a, log e
b) *h'
o;A =(o;e) h' h'
P.d-eP.Aloge+gh'
(o;loge) a.loge+gh' oAloge
and
(8.7)
with 2 )
a) cp,, * A! cpi - oi' log e
(8.8) {
b) cpr * (1 + ~~~ o11 log e) (Aj.t. cp;i- e- 1 or a e).
From these equations we see that indeed a transformation of the form
(8.1) transforms a special coordinate system into another special system
with the same pole. The other part of RACHEVSKr's theorem follows
from the condition that the transformation must transform every
homogeneous linear equation into an equation with the same property.
The cpi and cp; 1 do not behave like a vector field or a tensor field respectively
and each of them forms a geometric object for transformations of the
1) KAGAN 1933, 1, p. 41.
2) Cf. E II 1938, 2, p. 219.
 8. Subprojective connexions. 319

form (8.1) only. More general transformations disturb the form of

(8.5). It is often convenient to consider a vector field and a symmetric
tensor field defined by u; * rp; and U;; _>!<_ lf!;; but these fields depend on
the choice of (h) and they change if another special coordinate system
is introduced instead of (h). From (8.18a) we see that tp; can be made
zero if and only if the field U; is a gradient field. That implies that
2 8u rp;1 is an invariant bivector and a concomitant of the connexion.
Another consequence of (8.7) is that a subprojective connexion remains
subprojective for every projective transformation (1.1). From a geometrical
point of view this is quite obvious because a projective transformation
of the connexion does not change the geodesics.
The form (8.5) leads to the following expression for the curvature
tensor
(8.9) Ri,j;"h = 2A?k ~li- 2Af1[kn+ 2Ukii xh 1),
where Uk i;, ~; and xh are tensor fields and a vector field, defined by
(8.10) U,.;; ct~ 8rk IPn; + ~ 1 1PI[k lf!n;,
(8.11) ~ict:!-8;1f!;+ IP;; (1+ ~ 1 1P1) + lf!;rp;,
h _'!<_ th
(8.12) X def <;

For R;;, Vki P;; and P;.;;h (cf.VI 1) we have

(8.13) R;; :f<_ (n--1) ~;-28r;IP;J + 2 xh Uh 1 ;,
(8.14) V;;-"' 2 (n +1) 8u rp; 1 + 2 ~ih xh,
(8.15) P;;=-I;;+ ~i
(8.16) Pkjih = 2 U,.;; xh + 2 Atk 0;1;-2 Af Urkil
where
(8.17)

(8.18)

From (8.16, 17) it is seen that for n >2 the connexion is projectively
euclidean if U,.;; = 0. In order to prove that this condition is also
necessary we consider an expression of the form 2)
Ahrk .1 . + AhCk. + 1zh z.k. + ... + pzh z.k.;
(8.19) 1 1' ' 1

C<k i) = 0;
1' 1

Z;
p' 1
<k n = 0; ... ; Z; <k n = 0
1 p

1 ) In this equation the sign * can be dropped because the equation has the
invariant form.
2) SCHOUTEN 1953, 1.
320 VI. Projective and conforrnal transforrnations of connexionso

with zh, .. 0, zh linearly independento For Bi; =t= 0, Cki =t= 0 the h-rank of
1 p "
the first term is :2; n -1 and the second term has h-rank n 1 )0 In order
to find the h-rank of the first two terms together we try to find a
solution W; of
(8020)

Such a solution is only possible if Bi; has the form pi r; + rd; and if
Cki = p[k rilo 2) For P; =f= q; the solution is w, cx: r;o Hence the h-rank of
the first two terms together is n - 1 in this exceptional case and n in
the ordinary caseo The h-rank of the last p terms of (8019) together
is -;;;,p, hence
A uxiliary theorem:
F or n > p + 2 an expression of the form (8019) is zero if and only if
Bki' Cki and Ziki' .. o,Ziki are all zeroo For n = p + 1 the expression
1 p
can possibly also be zero if Bi, has the form Pi r; + ri q; and ij Ck i = p[k rilo
I f an expression can be written in the form (8019) and also in the form

(8021)

and if there are p + q- s linearly independent ones among the zh and 'zh,
it follows that
(8022) 'Bi;= Bi;; 'Cki = Cki

for n>P+q-s+20 This is also true for n=P+q-s+1 but it need

not be true if Bi;-'Bi; has the form piri +riqi and if Cki- 'Cki = p[krilo
Now let us suppose that there exists another decomposition of Piii h
(823)

then we have p = 1, q = 1 and s = 1 or s = 00 Then the exceptional

case can for n > 2 only occur for s = 0, n = 3 (necessary conditions
for this case are that uii- zii has the form pi r; + ri q; and that
2 p[k ril = q[k rn)o This proves that, for n > 2, Uki and Ukii are uniquely
determined by Pi/;" h provided that the direction of xh is known and
that, for n > 3, the direction of xh is uniquely determined by ~"/;" h
provided that Pk"/;"h='t,=Oo For n >2, ~i;= 0 if Pk,/;"h= Oo
Gathering results we have
1 ) The expressions Bii Ckj etco need not be tensorso Notwithstanding it is
possible to speak of a h-rank as long as the coordinate system is not transforrnedo
2 ) In ScHOUTEN 1953. 1 only the case qi = 0 was consideredo I owe this cor-

rection to a personal communication of J HAANTJESo

0
 9. AnATI's problem. 321

A subprojective connexion is for n > 2 projectively euclidean if and only

if U,. i i = 0. 1 ) I f it is not projectively euclidean the pole direction at every
point is for n > 3 uniquely determined and can be computed from ~//'.

9. AoATI's problem 2).

ADATI considered the question whether a connexion that is not
projectively euclidean and that takes the form

(9.1) E)i
h* 2 W(i A"
i) + Wi;V," wliil = 0;

withrespect to some coordinate system (h) is subprojective if some suitable

conditions are introduced for the field v". In (9.1) wi and the product
wii v" are supposed to be given as functions of the coordinates. Of
course it can not be said of wi, v" and wii that they transform in
the way of vectors or of a tensor because a coordinate transformation
generally disturbs the form of (9.1). But it is allowed and also con-
venient to consider a vector field and a symmetric tensor field that
have just the components wi, v" and wii with respect to (h). Of course
these fields belong then to the coordinate system (h) used in (9.1) and
to those systems that arise from (h) by a homogeneous linear trans-
formation with constant coefficients.
N. a. s. condition for the connexion (9.1) to be subprojective is that
there exists a coordinate system (h') such that (cf. 8.5)
h' *
(9.2) E''i' h'
= 2q;w Ai'l + 'Pi'i' ~h'
From (9.1) we get for the curvature tensor

{9-3) {
Ri,ji" * 2Af o1k wil + 2A?k (- oil wi + wil w; + wili w v + 1 1)

+ 2v" (81 wn; + 2w1k 111 wil; v + 2wi li okl v"

k 1)

and this expression could be equated to (8.9) because (8.9) is an invariant

equation valid with respect to every coordinate system. Of course (h')
is now the system for which (9.2) holds and for x" we have now

(9.4)

But comparing (9.3) and (8.9) we see that simple results can only be
obtained if oi v" has the form

(9.5)
1) ADATI 1951, 8, p. 136.
2) AnATI 1951, 3; 4; 5; 6; 7; 8; ScHaUTEN 1953, 1.
Schonten, Ricci-Calculus, 2. Aufl. 21
322 VI. Projective and conformal transformations of connexions.

This is equivalent to the invariant condition introduced by ADATI for

the vector field v"
(9.6) VI' v" = oc. A~ +
l' v".
A vector field in Ln satisfying an equation of the form (9.6) is called
torse forming because in an En the lines of the vectors of such a field
at the points of any curve form a torse. 1 ) By multiplying v" with a
scalar a another torse forming field arises and if oc. =j= 0, a can always
be chosen such that the new factor oc. equals + 1.
Important special cases are:
the concircular field 2 ): l' = gradient; oc. arbitrary. If v" is concircular,
av" has the same property and l' = 81' can always be reduced to
zero by taking a = C e-; C = const.;
the special concircular field: l' = 0; oc. arbitrary. The property is
only invariant if a = const.;
the concurrent field 3): l' = 0; oc. = const. By multiplying v" with oc.- 1
the constant oc. is reduced to + 1;
the recurrent or parallel field (cf. III 2) 4 ): oc. = 0; l' arbitrary;
the covariant constant field (cf. III 2): oc. = 0; l' = 0.
lt is easily proved that the only transformations of the that r:}.
leave every torse forming field torse forming are the transformations
of the form (cf. Exerc. VI 9,1)
(9.7)
These are at the same time those transformations in Ln that preserve
all geodesics and transform every semi-symmetric connexion into a
connexion with the same property. Here is a survey of the invariance
of the six vector fields just defined under the transformations (9.7)
1 ) If for a field vh in an An or Ln satisfying (9.5) a CARTAN displacement (cf.
III 2) is effected along a curve the images of the lines of the vectors in the moving
En form a torse. Fields of this kind were introduced by Y ANO 1944, 5. The pro-
perty of being torse forming is in fact not a property of the vector field but of its
field of directions. Obviously the field of the pole directions of a subprojective
connexion is torse forming.
2 ) YANO 1940, 5; 6; 7; 8; 1943, 3; ADATI 1951, 3, p. 161.
3 ) Concurrent fields were first introduced by MYLLER 1924, 1; 1928, 1 for vm
1ll Rn, cf. BORTOLOTTI 1931, 5; HAIMOVICI 1938, 1; SCHIROKOW 1939, 1; YANO
1943. 3.
4 ) Cf. for instance EISENHART 1922, 4; 1925, 1; 1926, 1, p. 67ff.; 1938, 1;
LOPSCHITZ 1936, 1; Y ANO 1943, 3; LEVINE 1948, 1; 1949, 1; RUSE 1949, 1; 1950, 1;
WALKER 1949, 1; 1950, 1; 3; 4; PATTERSON 1951, 1.
5 ) Transformationsofthis kind were considered by HLAVATY 1927, 3; 1928, 2;
1933, 1; ScHOUTEN and GoLAB 1930, 3; HoMBu and KADA 1941, 2; YANO 1944,
12; MIKAMI 1949, t.
 9. ADATI's problem. 323

p.
"' Transformation

Torse forming arbitrary arbitrary cx -+ cx + qp. vP.;

"'-+ "' P"'+
concircular arbitrary gradient invariant pp. = gradient
for
spec. concircular arbitrary 0 invariant p1, = 0
for
concurrent constant .0 invariant PP.= 0; qp.= 0
for
recurrent (parallel) 0 arbitrary invariant qp. = 0 1)
for
cov. constant 0 0 invariant for PP.= 0; qp.= 0

Torse forming fields are not possible in every An.

From (9.1, 5, 6) we get for n >2
(9.8)

If (9.5) is satisfied, its integrability conditions

(9.9)

must also hold. But if in the auxiliary theorem (cf. VI 8) i is strangled

and cki dropped it follows that for n >2 (9.9) is equivalent to

a) 8irx'- rx' yi!_ 0

(9.10) {
b) 8fi Y;J-"'- 0.

(9.10b) 1s a consequence of (9.10a), and (9.10a) can also be written

in the form
(9.11) 8irx-rxi* -vi(rxwii-8iw ;+wiw;).

If (9.5) is substituted in (9.3) it follows that Ri,ji" has the special

form (8.9). If then the auxiliary theorem is applied to (8.9) and (9-3)
it follows for n > 3 that

(9.12)
l
r :~
c)
d) Ukji
~;_'1<-rzwii- 8iwi
Triil =

v" * e x"
*
- 0ri w;J

(! (o[k Wj]i
+ wiwi

+ [k Wj]i- W[k Wj];)

Consequently (9.11) takes for n>3 the form

(9.13)
1) Cf. footnote 4 on p. 287.
2) In order to derive this equation we have only used
(9.1) and (9.5) and these
conditions are sufficient to ensure that Rj.ji" has the special form (8.9) without
it being known whether the connexion is subprojective or not.
21*
324 VI. Projective and conformal transformations of connexions.

The lf1 transform into the lj~;, as follows

Ai; h' 1T'h Ai; Ah'
(9.14) 1
Tlh'
i'i' = i'i'h ii- i'i' ai i .
Hence, in order that Ij~;, takes the form (9.2) it is necessary and suf-
ficient that oiAt' takes the form
(9.15)

Now we have from (9.4, 5, 12c, 15)

(9.16) { Ar* oie' * oi (2- 1 A;' vi

+ (2-
*Ar e-
1
V
h'
1 + 1p;Vi) +
(rx'
('!Pi+ "Pii v' + Yi- ai log e)
.

hence for n >3

(9.17) * (2 -1p; V i
IX I ~

(9.18) *
Yi ~-"Pi -'!Pii v'. + oi log e.
The integrability conditions of (9.15) are

(9.19)

and according to the auxiliary theorem these are for n > 3 equivalent to
a) (}.w.-
Jn '"'"-
"r7 T< 11l
1
_:!<_ 0
(9.20) { T1'-
b) "' ,
U[k 'lj!j] i -
, * 0"
'lj![k 'lj!j] i =

Now the integrability condition of (9.20a) is (9.20b) and the integrability

condition of (9.20b) is identically satisfied. Moreover if (9.17, 18) is
substituted in the integrability conditions (9.10) these equations are
identically satisfied on account of (9.17, 18, 20). Hence the system of
differential equations (9.4, 12c, 15, 17, 18, 20a, 20b) for the unknowns
~h', (2, A7', "Pi, "Pii is totally integrable. Every solution of (9.20a) if
substituted in (9.17) gives a value of e that satisfies (9.18) on account
of (9.20a). That proves the theorem 1 )
If the connexion of a non projectively euclidean An can be written in
the form
(9.21) w[iil = 0
1) ScHOUTEN 1953, 1; ADATI proved this theorem for vn but for An he intro-

duced (9.13), equivalent to (9.10a) as an additional condition and he did not

observe that this condition is in fact the integrability condition of (9.5) or (9.6)
and that it can be derived using (9.1) and (9.5) only [see footnote 2 on p. 323].
Cf. ADATI 1951, 7. p. 127 where he gives the final form of his theorem for An.
 9. ADATI's problem. 325

with respect to some coordinate system (h) and if the vector field which has
the components vh with respect to (h) is torse forming the connexion is
subprofective. The pole direction is the direction of vh. 1 )
The system (9.20a) has some very simple solutions. There are two
cases:
1. 01.' =F 0. Because the prod uct wi i vh is gi ven, we may fix the un-
known factor in vh in such a way that 01.' = 1. Then
(9.22) 8i Vh ~ * Ahi + Yi Vh .
But the integrability conditions of this equation are
(9.23)

and these can only be satisfied if Yi = 0. Then from (9.22) we get

(9.24) 8.1 vh --*- 8.Jr;.;eh
hence
(9.25)
N ow if we take
(9.26)
we have AZ' .!_ i5Z' and accordingly
(9.27)
as was demanded.
2. 01.' = 0. In this case we have
(9.28) 8i vh -*- Yi vh
for any choice of the scalar factor in vh. The integrability conditions are
(9.29)
hence by changing the scalar factor in vh we can always obtain
(9.30) -*- 0.
8.1 vh - '
vh -*-eh= const.

Now we take for the 'lfJ; n arbitrary constants c; such that c; Ci= 1
and for 'lfJfi we take -ci c;. Then (9.20a, 20b) are satisfied and from
(9.17) we get e=1, satisfying also (9.18). That leaves for the Af the
equations (9.15) in the form
(9.31)
1 ) The condition for vh is sufficient but not necessary. Necessary and sufficient

conditions are not yet found. The theorem holds also for n = 3 but for that case
it is not proved that the direction of vh is the only pole-direction at every point.
326 VI. Projective and conformal transformations of connexions.

with constants c, and C". Because the c0 and C" are constants the
coordinate system (h) can be transformed by a homogeneaus linear
transformation with constant coefficients such that after the trans-
formation c,
coincides with J, and C" with f"
and this transformation
does not disturb the form of (9.1) and (9.30).
Then (9.31) takes the form

(9.32) b, c = 2, ... , n

er= const.; b=2, ... ,n

with constants Ci' that can be chosen arbitrarily provided that

Det (Ci'} =F 0. By a second integration we get
(9.34) ~~~ * (Ci' f/' + C1') e~ 1 = 1fl'' = v11'
and this leads to the following expression for the IJ!';.
(9.35) IJ!';. * (wr +~)Ai:+ (w,. + e:.) At:+ (wr- ~ J..H"'
as was demanded. (9.35) is independent of the choice of the In er.
fact a change of these constants is equivalent to a homogeneaus linear
transformation of the ~~~ with constant coefficients and such a trans-
formation does not disturb the form of (9.35).
From (9.34) we see that the case oc' = 0 is not at all abnormal and
that it only arises because the coordinates are chosen such that they
get infinite at the pole. VRANCEANU 1 ) used 1947 this special choice
of the Coordinates in order to simplify his calculations.

Exercises.
VI9,1 2). The transformations ~}.--'>-~).+P;;." in Ln transforms
every torse forming vector field into a field with the same property if
and only if P;;." has the form
VI 9,1 oc) P;;."= p"A}. + A~q;..
VI 9,2. The transformation of Exerc. VI 9,1 leaves the tensor
R;;;."+A}.R." invariant for q;.=O.
1) VRANCEANU 1947, 2.
2) Cf. footnote 5 on p. 322.
 10. Subprojective transfonnations of a connexion in A,.. 327

VI 9,3 1). The object

VI 9,3 ct) A;A = J;). + n 2 ~ 1 {A; (JA~- nJ'e').) + A1 (f'e't- nJ;.Q0 )}.
is invariant for the transformations of Exerc. VI 9, 1.

10. Subprojective transformations of a connexion in An.

A transformation of a connexion of the form
(10.1)
with arbitrary vectors p1 and V" and an arbitrary symmetric tensor q1,
is called a subprofective transformation belanging to the direction Lv"J. 2)
All subprojective transformations belonging to the same direction LV"J
form a group.
The curves satisfying equations of the form

(10.2) d2 t;" +E~ dl;i dt;i = u dt;" +v"

dt 2 ,. dt dt dt

are called subgeodesics with respect to the field Lv"J. 3) The subprojective
transformations belonging to Lv11J leave the form of these equations in-
variant. If Lv"J is torse forming we know that subprojective connexions
exist withthe pole-direction LV"J andin this case the subprojective trans-
formations belonging to Lv"J are those transformations that transform
these subprojective connexions into each other. The set of all sub-
projective connexions belonging to Lv"J and the set of all streamlines
of Lv"J determine each other uniquely. Hence a transformation ~"-+~+
u" dt leaves this set of connexions invariant if and only if it leaves the
.
streamlines invariant, that is if ~ v11 cx; v". This can be proved also
from (II10.34). If this equation is written in the form 4)

(10.3)

!- +
it must be proved that lf7 has the form w1Ai w1 Af w11 v" 5) if and +
.
only if ~v"cx;v11 Now if (9.1) is substituted in (10.3) and if all terms

1) HLAVATY 1927, 3, p. 85.

2 ) YANO 1944, 6; ADATI 1951, 6, p. 106.
3 ) Equations of the form (10.1) and (10.2) were first considered by v. DANTZIG
1932, 3; cf. ScHaUTEN and HAANTJES 1936, 1; HAANTJES 1937, 1; YANO 1944, 3;
6, he introduced the name "subpaths ".
4 ) Cf. EI 1935, 1, p. 142.
6 ) YANO 1944, 6 has formulated this condition put he seems to have over-
looked that it is satisfied if and only if the streamlines are invariant.
328 VI. Projective and conformal transformations of connexions.

are dropped that have alreadythe right form, the onlyremainingtermsare

(10.4) (cf. 9.5)

and these take the right form if and only if f v" cx: v".
"
Exercise.
VI 10,1 1 ). Every subprojective An can be transformed into an En
by a subprojective transformation of its connexion.

11. The subprojective V...

If the subprojective Anis a Vn, (8.9) takes the form

(11.1) Kkiih -*- 2grklhl1fJ;- 2g;,. I[kil + 2Ukii x.,.

and this expression must be alternating in ih:
(11.2)

For n >2 we may take two vectors y", z", perpendicular to x" and to
each other. Then by transvection of (11.2) with y" i we get
(11.3)

and this means that 7j; yi cx: Yi for every vector y" perpendicular to x".
But this is only possible if 7j; has the form
(11.4)

with suitably chosen p, and li. If (11.4) is substituted in (11.1) we get

(11.5) Kkiih * 4p,g[kfklgn;+2g[kfkllnx;-2g;.,.lrkxil+2Dkiixh.

As this expression must be alternating in ih we get

(11.6)

and by transvection with yi y" it follows that lrk xil = 0. Hence 1f; is
symmetric
(11.7) 1f; * 2p,gi;+Axixi

and this implies that q;i is a gradient and that there exist Special CO-
ordinate systems for which q;i vanishes (cf. VI 8). If (11.7) is
substituted in (11.2) we get
(11.8) a) ukih =;, xlk gil,.; b) ukik x" = 0
1) YANO 1944, 6, p. 104.
 11. The subprojective Vn. 329

and consequently
4
J a) Kkiih = n _ 2 g[k[iLiJhJ
(11.9)
1 b) Lih d:f - (n- 2) (p gih + A xi xh).
For n >3, (11.9) is the n. a. s. condition for the Vn to be conformally
euclidean. For n = 3, Kkiih can always be written as the double alter-
nating product of gki with another symmetric tensor. But in that
case 17rkLilh=0 is the n.a.s. condition (cf. VI 5). Now we can prove
that this latter condition [that is a consequence of (11.9) for n >3] is
always satisfied for a subprojective Vw 1 ) We take the coordinate
system such that cp; = 0. Then cx = 1 and from (8. 5, 10, 11, 12) it follows
that
(11.10) Ij; * Cf!ii -*- 2pgii + A xi X;,
(11.11) T'h *
h
1j;=Cf!j;X,

(11.12)

But on account of (11.8) this leads to

(11.13)

where Cis a scalar. If (11.13) is substituted in (11.9), differentiating

gives 17rkLilh = 0. Moreover RACHEVSKI 1) proved that the condition
is also sufficient and can be put in the following form
A Vn, n > 2 is subprojective if and only if it is a Cn satisfying the
special condition
(11.14)

where a is an arbitrary scalar and T a function of a only.

From
(11.15) 17j X h
V * Ahj + Cf!ji X; X h ~* Ahj + (2 fl + A, X; X;) Xj X h
=

and (11.13) it follows immediately that i is a gradient. Hence x" is

not only torse forming but also concircular. 2 ) ADATI gave also the forms
of LiTt for the cases that xh is parallel or the unitvector of x" concurrent
and he gave also other forms of the linear element of subprojective Vn's
for different cases. 3)
1 ) RACHEVSKI 1933, 1, p. 137; cf. E li 1938, 2, p. 22ff.; cf. ADATI 1951, 5,
p. 343 ff.
2 ) ADATI 1951, 3, p. 172.
3 ) ADATI 1951, 4; 5, p. 333, 357f.
330 VI. Projective and conformal transformations of connexions.

Exercises.
VI 11,1 If in a subprojective V"; n>2, the eoefficient p, in (11.4)
1).

is a eonstant, the V" is an 5,. (use 11.9 and 11.13).

VI 11,2 2). The Cn with the linear element

is subprojeetive.

12. Concircular transformations of a Vn.

From (5.36) we get for the eurvature vector u"=t"~'Vpt"" of a eurve
with tangent unitvector f" under eonformal transformations of gA><

(12.1) 'u"=a- 1 (u"-tz"); z"dfs"-i~-'spi"; s"dfo"loga.

Now we eonsider a geodesie circle 3 ), that is a eurve whose eurvature

vector has eonstant length and whose seeond eurvature is zero

(12.2)

Then we have after the transformation

(12.3) ds ('u" 'u" 'g;.,. ) = aJp 'u" (-V.P sA + .!_

__!__ 2 sP s)
;.

and this proves that a eonformal transformation transforms every

geodesie eircle into a geodesie eircle if and only if an equation holds of
the form4)
(12.4)
or (cf. 5.7)
(12.5)

Conformal transformations satisfying this eondition are ealled con-

circular. They leave invariant the equation of all geodesie eircles, that
1)SCHAPIRO 1933, 1, p. 112; E l i 1938, 2, p. 222 and 225.
2)Cf. BucHHOLZ 1899. 1; KAGAN 1933. 1, p. 94; E II 1938, 2, p. 225.
3) FIALKow 1939. 2. Y ANO introduced 1938, 3 other generalized circles and
proved 1940, 6 that these curves are geodesie circles if their tangent direction is
everywhere a principal direction of V,..
4 ) BRINKMAN 1925, 1; FIALKOW 1939, 2, p. 461; cf. also FIALKOW 1940, 1;

1942, 2; YANO 1940, 5; 6; 7; 8; YANO and ADATI 1944, 7.

5) The factor 4 is introduced only in order to get the same t:p and tp as in the
paper of Y ANO. His r!pA is equal to t sp;..
 12. Concircular transformations of a Vn- 331

can be proved to be 1 )
(12.6)

Transvection of (12.4) with s"' leads to

(12.7)

hence the streamlines of s" are geodesics and the congruence of these
geodesics is normal to the vn-l's tangent to the gradient S;.. The inte-
grability conditions of (12.4) are
(12.8)
from which by transvection with g"''-

(12.9)

But by transvection of (12.8) with s;. it follows that V. 'tfJ is a multiple

of s. Hence s" lies in a principal direction (cf. III 5) of Vn 2) and the
function 1fJ is COllStant Olleach of the Vn_ 1 's orthogonal to the Congruence s".
If i" is the unitvector of s" and i' =Ci" we have from (12.4)

(12.10)
and
(12.11)

But this implies that V"' i;. has the form

(12.12)

Hence the second fundamental tensor of the Vn_ 1 's normal to i" is
(12.13) - - - B"';.V.
hc b . - h .gcb
cb p.t;.-
1

and this proves that these V,._ 1 's are umbilical. From the ConAzzr equa-
tion (V 6.5)
(12.14) 2 C'Vra h) g,1b = B'd~~ K."';." i"
1

we get by transvection with g'b 1

(12.15) (n - 2) V.d h = d (g"';. - i"' i;.) K vp.>< i" = d K "" i"

1

1) YANO 1940, 5, p. 195. He also eonsiders eurves that ean be transformed

eonformally into a geodesie eircle and ealls them eonformally geodesie eircles. But
beeause every single eurve is eonformally geodesie (cf. VI 5), it is also a eonformally
geodesie eircle and the term is useful only in eonnexion with systems of eurves.
2) Y ANO 1940, 6, p. 3 56.
332 VI. Projective and conformal transformations of connexions.

but, because ix lies in a principal direction of Kvx this proves that h

is constant over vn-1' i.e. that the vn-l's have each constant mean
curvature. So we see that the V,,_ 1 's normal to a geodesie and normal
congruence are always umbilical if the tangent direction of the con-
gruence is torse forming 1 ) and that the constancy of the mean cur-
vature is due to the fact that in the special case under consideration
the tangent lies in a principal direction of the vn.
Now let us suppose that there exists in a V" a congruence with unit-
vector i'-' satisfying (12.12) with '17bh=0. Then we prove that there
exists a scalar 'f} such that the field defined by s~ = 'f} i;. satisfies an
equation of the form (12.4). From
(12.16)
follows
a) 2 o" 'fJ = ('f} 2 - 2'f} h) i"
(12.17) {
b) 'f}h=-21p.
The integrability condition of (12.17a) is

(12.18)

and because i[l' a.] h = 0, this implies that 'f} is a function of u only, if
we write i" = o"u. Now hisalso a function of u only, hence if 'fJ det f(u)
and f' (u) <Iei !1_ the differential equation for 'fJ takes the form
du

(12.19) 2f'=f2-2hf

and this is a RICATTI equation. Gathering results we have 2 )

A Vn admits a concircular transformation if and only if there exists
at least one geodesie congruence, normal to a set of umbilical Vn_ 1 's with
constant mean curvature.
The concircular transformations form a subgroup of all conformal
transformations. If a V", in Vn is given, every concircular transforma-
tion gi'-+ (]" gi' in vn induces a conformal transformation 'gcb -+(]" 'gcb
in vmand we may ask 3 ) whether the induced transformation is also
concircular. Now if 'sb def Bts;. we have

(12.20) {
2'17c 'sb = 2'17c Bg S;. = + 2s;. Dc m
2B~g 171' S;.
= 'sc 'sb + 41p 'gcb + 2~i/ S;_,
1) 1944, 5.
YANO
2) 1940, 6, p. 359; cf. also FrALKOW 1939, 2. The proof given here is
YANO
much simpler than Y ANo's proof that is based on a special form of the linear
element and makes use of a function suggested by A. KA w AGUCHI.
3 ) Y ANO 1940, 8, p. 506.
 12. Concircular transformations of a V". 333

hence the induced transformation is concircular if and only if Hc'i/ s

is a multiple of 'gcb or, in another form, if and only ifl) (cf. V 9.16b)

{12.21)

As a corollary we get that a concircular transformation induces a con-

circular transformation in every umbilical Vm. This implies that every
geodesie and every geodesie eircle in an umbilieal Vm is a geodesie eircle
in V". 2 )
If a field u" is subjected to a parallel displacement over v" dt we
+
have for the field value of the new field at ~ v" d t

(12.22)

In order that this value has the same direction as the field dragged along
over v" dt
{12.23)

it is necessary and sufficient that

{12.24)

where oc: is a function of the /;". Hence, a point transformation v" dt

leaves invariant the tangent direction of every curve if and only if v"
is a special concircular field (cf. VI 9):

(12.25)

Because V"'v=oc:g"'' the vector v is a gradient. If we write v=oe- 1

it follows that
(12.26)

and this proves that the V" admits a concircular transformation g,..-+
rl g"' with tp = - oc:e (cf. 12.4). Con:versely, if we have a V" with a field s
satisfying (12.4) and if s~ 8 log 0', the field v" =I" 8 a-~ is a special
concircular field. Hence 3 )
A V" admits a eoneireular transformation if and only if it admits
at least one speeial eoneireular veetor field and this is the ease if and only
if it admits at least one infinitesimal point transformation that displaees
the tangent direetions of every eurve parallel.
1) See fo:>tnote 3, page 332.
2) Y ANO 1943, 3, p. 193.
3 ) Y ANO and ADATI 1944, 7. This paper deals also with the case where v" is a
concurrent field, cx = const.
334 VI. Projective and conformal transformations of connexions.

FRENET formulae for eurves in eoneireular geometry ean be deve-

loped by means of the vector (cf. 12.6)

(12.27)

that transforms into a-ll V" for a eonformal transformation of g,.;, if

and only if this transformation is eoncireular.l)
From (12.5) and (5.6) it follows that for a eoneireular transformation
(12.28) 'K;~~" = K;~~" + 4<p g[v[J. gp]a] g"'"
Henee, a Vn that ean be transformed into an Rn by a eoneireular trans-
formation must be an Sn. Moreover it follows from (12.28) that a
eoneircular transformation transforms every EINSTEIN spaee into an
EINSTEIN spaee. 2) Another eonsequenee of (12.28) is that3) 4)

(12.29)

is invariant for eoneircular transformations. Y ANo 3 ) proved that every

eonformal transformation that leaves Z;~~" invariant is eoncireular.
Z;~~" is ealled the concircular curvature tensor. In the same paper YANO
proved that its vanishing is for n > 2 n. a. s. for the to be trans- vn
formable into an Rn by a eoneireular transformation. The Vn eould be
ealled concircular euclidean in this ease but because for n > 2 there is
no difference between an Sn and a eoneireular euelidean spaee it is
better not to introduce a new name. ForZ;~~" the following identities
hold
(12.30)
{ a) ... " --
z"";, K ,.;. - t
2
K K,.;.
'
b) g"AZ,;~~"=O.

In the last of YANo's papers on eoneircular transformations and in a

paper of 1943 5) he studied eonformal and eoneireular properties of
EINSTEIN spaces. Reeently TACHIBANA 6 ) dealt with the relations bet-
ween eoneireular transformations and the investigations of SASAKI 7 ) on
the holonomy group of eonformal spaees.
Exercise.
VI 12,1 8 ). If every geodesie eircle of a Vn_ 1 in Vn isageodesie eircle
of V,., the Vn_ 1 is umbilical and its mean eurvature is eonstant.
1 ) YANO 1940, 4; 7
2) Y ANO 1940, 5, p. 200.
3) Y ANO 1940, 5.
4) TACHIBANA 1951, 2, p. 150.
5) YANO 1942, 3; 1943, 1.
6) TACHIBANA 1951, 2.
7) SASAKI 1943, 1; 2; 3.
8) YANO 1940, 7; 1943. 3.
VII. Variationsand deformations. - 1. General deformation problems. 335

VII. Variations and deformations.

1. General deformation problems. 1 )
In this chapter several problems concerning variations and deforma-
tions are dealt with. The general form of such a problern is as follows:
certain geometrical objects suffer a finite or infinitesimal transforma-
tion and the behaviour of some other objects depending on them is
required. In deformation problems we deal with the special case where
the variation is due to displacements of some kind, for instance a drag-
ging along or a parallel displacement. The case occurring most frequently
isthat some objects are left at rest, others are dragged along and others
are displaced parallel. The theory of variation and deformation is
very important because a great number of problems in differential
geometry can be treated in a very elegant way by using the methods
of this theory.
If a geometric object field is defined in an 91 (~") we have at ~" + v"dt. 2 )
1. the natural value of the field or the value of the field at rest, for
instance for a contravariant vector

( 1.1)

2. the value arising from dragging along the field over v"dt, for
instance (cf. II 10) for vectors
m

(1.2)
a) P" + dmp" = P" + P" a" v" dt
\
b) q;,+dq;,=q;,-qp,o;,v"dt.

If a parallel displacement in the direction of v" is defined for the quanti-

ties considered we also have
3. the value arising from parallel displacement of the field over
v" dt, for instance (cf. III 2)

(1.3)

1 ) Cf. McCONNELL 1929, 1 (curves in Vn); HAYDEN 1931, 1; ScHOUTEN and

v.KAMPEN 1933, 1; DAVIES 1933,1 (in Un); 1937, 1; 1938, 1; DIENES 1933,3
(in Ln); EI 1935, 1, 12; E II 1938, 2, 16; CoBURN 1940, 2 (on SCHOUTEN and
V. KAMPEN); SHABBAR 1942, 1 (finite deform.); YANO 1945, 7; 8; 9; 10 (on Co-
BURN); 1949, 2 (literature); ToNOLO 1950, 1 (in V3 and 5 3).
2 ) We follow here the line of thought developed in ScHaUTEN and v. KAMPEN

1933, 1 and EI 1935, 1, p. 140ff.

336 VII. Variations and deformations.

The three differentials arising in this way are all non invariant!)
but their differences are invariant:
1. the covariant differential b (cf. III 2), for instance for vectors

a) bp" =
*
dp"- dp" = v'" 17'" p" dt
(1.4) \
b) bq;. =
*
dq;.- dq;. = v'" 17'" q;. dt.

2. the LIE differential f dt (cf. II 10 and III 5.48), for instance

V

m
r a) f P" dt = dp"- d P" = (v'" 811 p><- P'" o'" v") dt

( 1. 5)
I
{ b) f q;. d t = d q;.- d q;.
m
=

=
v'" 17'" r
(v'" 8'" q;.
d t - P'" V~" d t.
+ q'" 8;. v'") dt

l v"
).

3. the apparent differential 2) Ds d t, for instance

- - 8). v"
cl~
= v'" 17'" q;.dt
+ F" p v'"
+ q'" v;'" dt.

(1.6) {
a) J? P" d t = dP" - dP" = - P'" v~" d t
* m
b) J?q;.dt=dq;.-dq;.=+q'"v~'"dt.

Between these three differentials applied to quantities we have the

relation
(1. 7) b-fdt+Ddt=O.
V S

The geometrical interpretation is as follows. Let the field v" be given

in an IJ1 ( ~") and let the field of a quantity Q (indices suppressed) undergo
0
on the one hand a parallel displacement over v" d t. Then the value Q
0
at ~" changes into .0- bQ. On the other hand let the field Q be dragged
0 0
along over v" dt. Then the value at ~ changes into Q- f Qdt. The
0 0 V

difference between the two new values at ~" is - D Qdt, hence we get
0 s
exactly the field value Q
0
+ Ds Q d t at ~
0
if the field Q is first dragged
along over - v" d t and afterwards displaced parallel over + v" d t. This
illustrates the fact that D Q dt is really not a true differential of the
s
field Q but only a linear transformation of the components of Q at ~
0
1) lf by applying a differential operator to the components of a geometric
object of a certain kind the components of another geometric object of the same
kind arise, we call the operator invariant with respect to objects of this kind.
2 ) "Scheinbares Differential" in ScHOUTEN and v. KAMPEN 1933, 1 and

EI 1935, 1, p. 142.
 1. General deformation problems. 337

depending on the values of Q and of the field v_i'' at the point ~ only,
0
and not on the values of these fields at neighbouring points.
From (II 10.34) or (III 5.47) we get

In (III 5.50) we obtained the finit of the following three identities

a) (f V11 - V11 f) P" = (fV F"";.) p" = (J711 v;."+ v R;;;.") p;.
1
V V

(1.9) b) (fV V11 - V11 f) q;. = - (f.F"";.) q,. =-(V~' v;." + v R;;;.") q,.
V V

c) (f.V V11 - V11 f)

V
q = - w (f.F,.)
V
q=- w (J711 v;.;.
0
+ v"V" 11 ) q
for the vectors p", q,_ and the density q of weight w, from which the last
two can be derived easily from the first. From (III 4.9a, b) we get for

l
these same quantities

a) (!5 V11 - V~' !5) P" = (v R;;;."p;.- v;e Ve P") dt

(1.10) b) (!5 V11 - V11 !5) q,_ = - (v R;;;." q,. + v;e Ve q;.) dt
c) (!5 V11 - V11 !5) q = - w (v V. 11 q + v;e Ve q) dt

and from these equations and (1.7) it follows that 1 )

(1.11)

All these identities are important because they give us an opportunity

of changing the order of the operators f., !5 or D and the operator V11
. any f ormua.
m 1 V S

If the operators !5, f dt and D dt are applied to a number of objects,

V S

other objects, depending on these will in general suffer variations of a

more complicated kind, that could not arise by the mere dragging along
or parallel displacement of these latter objects. Let such a variation
transform an object field Q (indices suppressed) with fieldvalues Q
0
+ v" dt.
V V
at ~ into a field with fieldvalues Q + dQ at ~ Then d is not
0 0 0
an invariant operator but the following three invariant operators can
be defined
1) ScHOUTEN and v. KAMPEN 1933, 1, p. 9.
Scheuten, Ricci-Calculus, 2. Auf!. 22
338 VII. Variations and deformations.
n
1. The natural variation D!Jdt, that is the difference between the
+
new value at !;" v" dt and the old value at that same point
0

(1.12)

The natural variation does in generat not depend on a connexwn.

a
2. The absolute variation D!Jdt, that is the difference between the
new value at ~" + v>< dt and the value arising at that point from dragging
0
along from ~>< over v" d t
0

(1.13)

The absolute variation does in generat not depend on a connexion.

g
3. The
geodesie variation D!Jdt, that is the difference between the
+
new value at !;" v" dt and the value arising at that point from parallel
0
displacement from ~" to !;>< + v" d t
0 0

D-
V

(1.14) def _!
dt

Of course these operators do not always exist. Between them and the
operators v~-' J7f-l, f and D the following relations hold
V S
n a g

f D= = D - f = D - v~-' J7ft

j
V
a n g
(1.15) D=D+f =D+D
s
n a
+ v~-' 17f-l =
g
D = D D - D
s

If (/J is a field (indices suppressed) in an 9C ( ~") we also have the

0
following geometrical interpretation. After the variation the field value
v n
+ d(/J = (/J + DC/Jdt.
n
at ~" is (/J- v~-'81-' C/Jdt So the natural variation DC/Jd t
0
is just the change of the field at ~" that we observe looking at the field
0 g a
during the variation. The changes DC/Jdt and DC/Jdt appear if we com-
pare the new fieldvalues at ~" not with the original fieldvalues at that
0
point but with the values arising at ~" after a parallel displacement
0
or a dragging along respectively of the field over v"dt.
Accordingly we have for a field left at rest:
n a
(1.16) D=O; D=f;
V
 1. General deformation problems. 339

for a field dragged along

n a g
(1.17) D=~f; D=O; D=~D
V s

and for a field displaced parallel

a g
(1.18) D=+D; D=O.
s

As a first example of a deformation problern we try to compute

the variation of the expression u'" V11 P~.< if 1) the vector field ux is left
unchanged, 2) the tensor field P~.< is displaced parallel over v"dt and
3) the object field I;,".< is dragged along over vxdt. Then we have

(1.19)

hence after some calculation

n n n
+ u'" (D F") FQ .<
n
u'" V11 D P" .1 ~ u'" (D F")
1
Du'" V11 P" .1 = 11-' P"
.a I'Q

(1.20) = ~ vv Vv u'" V11 P~.< + (f u'") V11 P~.< +

V

+ (u'" V v;.") P~" ~ (u'" V v;") P!.<;

11 11

and, according to (1.15)

g
(1.21) D u 11 VI" P",.1 = (fv u'") VI" P" .1 + (u'" VI' v.1 ") P" a ~ (u'" VI" v ") P ,.1 .
Q

(1.22)

Formula (1.22) can also be found by starting from

f fJu" = fu"
(1.23)
l ~ p~.< :rp~-1
DJ;,.<~O.

The formulae (1.15-18) are very convenient for computational pur-

poses because theymake it possible to choose always the easiest operators.
g
But it should be remernbered that v'" V11 , D and D can only be applied
s
if a parallel displacement is given and then only to quantities.
22*
340 VII. Variations and deformations.

A more important application, connected with ScHLESINGER's pro-

duct integrals, is the following. Let C be a closed curve ;" = f" (t) in L.,
through ;" and let t be a parameter on this curve, t 0 :;;,t:;;.t1 , such that
0
f" (t0) = f" (t1) = ;". Now let the vectors in the tangent
En at ;" be dis-
o 0
placed parallel along C till they return to ;". Then they have suffered
0
a linear homogeneaus transformation, depending on the form of C and
on the connexion. This transformation may be written in the form
(1.24)

in which the kernel of the tensor of the transformation is symbolized

by [;~]. This transformation can be expressed by means of a product
integral

(1.25)

A product integral is defined in matrix notation by

(1.26)

Product integrals have many properties that can be found in the papers
of ScHLESINGER and other authors.I) Herewe need only the generaliza-
tion of LEIBNiz' rule for the variation 2):

(1.27)

and the differentiation rule

(1.28)

that can be derived easily.

Starting from a vector u" at ;", at an arbitrary point ;" of C we get
0 0

(1.29) u" = lt,to'IA" 0u ' t 0 :;;;;, t:;;;;, t1 .

Here the tensor [;J~ is a connecting quantity lying with the lower
index in the tangent En of ;" and with the upper one in the tangent E.,
0
1) ScHLESINGER1928, 1; 1931, 1; 2; 1932, 1; cf. VaLTERRA 1887, 1; MaRINAGA
1934, 2; EI 1935, 1, p. 134; NIJENHUIS 1952, 1, Ch. li 13.
2) Cf. MrcHAL 194 5, 1; J OHNSON 1948, 1.
 1. General deformation problems. 341

of ~- Because u" is covariant constant along C we have along this

curve according to (1.28)

(1.30) ~ l't,).
0 -- dt ]" -- dtd ['t,]"). + fit
a!;~' F",..~ l1t, ]e).
andin the same way
(1.31) 0- ~
- dt ll]"- dtd lt1"
I .1. - t ,; - fdf;l' ,_.;. [11"
i t ra t a

Now let C be subjected to the variation e v". Then a point ~ of C

+
is transformed into ~ e v" and the variation d" l;:J~ must be computed
from the values r;.;
+ ev a.I',..",; at the points of the displaced curve.l)
Instead we transpose the problern in the way described in V 9 by
dragging back the curve over - e v" to its original position and at the
sametime dragging along the field I',..";. over - e v". If now we compute
the natural variation of [::]~ for the transposed problern this variation
is the difference between the new value and the original value both
at ~. But this new value arises from the new value of the original
0
(not transposed) problern at ~
0
+ e v" by dragging along over - e v".
This means that the natural variation for the transposed problem equals
the absolute variation for the original problem.

I
Hence, if we denote the natural variation for the transposed problem
" " a
by 'D, we have 'D=D. Now we have according to (1.25, 27)

J{A1- (r;" + e~I',.."") a:; dt}

n~

l!:l~ +e'D l~1~ =

(1.32) t. t,
= [e]"-
t, A
ej lt 1a t, A d t
df;l' (fFil) [t,]"
V t Q
(.1<1
dt '
t.
hence, using (1.8)

(1.33)

According to (1.30) we get, integrating by parts

t,
f l Ja~ ve [I]" = (lt 1a ve [1,1"\
~

(1.34) t
10 A a t 1/ t, A a t (l}
I= le,la
t A a
0
v"- ve [t]".
A t, e '
t0 t0

hence, returning to the original problern

1) Cf. ]OHNSON 1948, 1, who also considers finite shifts of C.

342 VII. Variations and deformations.

The natural variation of [::1~ for the original problern does not exist
because [::1~ is not a field defined in an 91 (f') and gaccordingly has no
field value at f + e v". But the geodesie variation D l::1~ exists because
it is the difference between the new value at ~
0
+ e v" and the value
arising from the parallel displacement of [::1~ from ~ to ~ + e v". From
{1.15) and (1.35) we get 0 0

!J [t't,1.t" = -J [tt, 1a.t v d/;J.t

t,
(1.36) dt R. ~ [t1" dt'
v,_.a t e
t,

and this equation does not contain any derivatives of the field v". The
case where v" = 0 at ~" is very important. Since ~ is then fixed, d is
0 0

zero at that point and Dl::1

g
equals the n[::1
"
that exists in this special
case and represents the variation at ~ that really appears. The formula
0
(1-36) will be used for the discussion of the holonomy group of a con-
nexion in VII 4.
As another interesting application we consider the generalization
of BERTRAND curves for V,. . In R 3 a curve is called a BERTRAND curve
if there exists another curve such that the first normals of both curves
are the same. 1 ) Let a curve in R,. be given by
{1.37) (" = r(s)

as in V 1 with s as parameter of length but with the restriction that

(x) is a rectilinear coordinate system. Then a point on the first normal
r
of a point (s) has the COordinates
(1.38) rp" (s) = r (s) + 1p (s) i" (s).
2

We try to choose 1p (s) in such a way that the curve ~ = rp" (s) has the
same first normals as the curve (1.37). According to the FRENET for-
mulae (V 1.10) we have

(1.39) d "
_({!_
ds
= -df"
ds
+ __tt'_
d
ds J
.,. + - =
dj"
1p ds
2 (1 - k ) .,.
1 1p J
+ __..!_
d
ds J
... + k
2
...
1p J

and because drp"fds must be perpendicular to J"', this means that 1p is

2
constant: 1p = c. Now s need not be a length on the second curve, hence,
1
if 's is the parameter of length on this second curve we have for its
tangent unitvector
(1.40) k" =!:..!.... dqJ". d's=J=O
1 d's ds ' ds

1) Cf. for instance WEATHERBURN 1927, 1, p. 34ff.; HAACK 1948, 1, p. 31f.

 1. General deformation problems. 343

From this it follows that the vector

(1.41) !!_s_ d cp" + (.!_s_)2 !!_cp"__

d 's 2 ds 2
d 's ds

has the direction of f'. But this is only possible if

2

d s )2 -
(1
- f k) d2 s
1 d 's 2 - f ~~ ( d'S
d
-
0

5 )z
az-s + cd-~ ( -d -
(1.42) ck =0
1 2 2 d 's 1 ds d 's

c k k (_!_5__) 2
= 0.
1 2 3 d 's

There are two trivial solutions: c = 0 and k

1 2
= 0 (hence k
3
= 0). If we

l
exclude these, (1.42) leads to the n. a. s. conditions

(1.43)
a) ck
11
+ c k = 1;
22
c and c constants
1 2
b) k = 0
3

for a curve to be a BERTRAND curve in Rn. If these conditions are satis-

fied it is easily proved that
(1.44) - -kV---:r---+~
d's- - c c
ds 2 1 2

and that cfc is the tangent of the (constant) angle between the tangents
1 2
of the curves at corresponding points. c is the (constant) distance bet-
1
ween corresponding points and from this we see that to a BERTRAND
curve there belongs in general one and only one corresponding curve.
But if k and k are constants, there is a second curve for every choice
1 2
of c as is well known from the theory of circular helices in R 3 .
1

Now this problern can be generalized for curves in Vn by requiring

that a curve passes into another curve by an infinitesimal deformation
in the direction of its first normal and such that the first normals of
both curves are parallel at corresponding points. 1) The curve may be
given by (1.37) but (x) is no langer a rechlinear system. Instead of
(1.38) we get
(1.45) cp"(s) = f"(s) + ei" (s) 2

and it can be proved as before that c; is an infinitesimal constant. 2)

But from here we must proceed in quite another way because the co-
variant derivatives are no langer ordinary derivatives and the I'"";.
1 ) PEARS 1935, 1; YANO, TAKANO and ToMONAGA 1948, 2. The latter paper
contains many theorems on deformation of curves.
2 ) lt is convenient to use FERMI coordinates, cf. 111 8.
 344 VII. Variations and deformations.

do not have the same values at corresponding points of the two curves.
Again we transpose the problem, dragging back the second curve and
the field g;,.,. over - ef". We note that although the curve dragged
2
back coincides with the original curve, s need no Iongerbe a length para-
meter on it.
For the fields g;,.,. and r;;.
dragged back over - s v" we have accord-
ing to (1.17)
(1.46)
n
(1.47) DF"
1
J' = 5:-F"
J' = 17J' 17 v"+ v K
V " V

but, because j" is defined over C only, we must put v" = j" on this curve
2 2
and prolang this field over a neighbourhood of C in order that the
derivatives in {1.46, 47) get a meaning. Of course all results will be
independent of the manner of prolongation.
From (1.46) it follows that

l
(1.48)
hence
n
1 Dds
(1.49) a) d-s = - ~;
n a d!;" d!;" a d!;"
b) 1
Df" = D dS = - (ds)" D ds = (rii)Z ~ ds = ~j".

For the first curvature vector we get

n n n n
ID j~' VI' i" = (I D j~") VI' i" + j~' VI' ID i" + j~' / ID I;.";.
= k 1'I' 171'1... + 'I' (17 k) .,. + 'I' k 17 ." '
{1.50) 1 1 I' 1 1 1 1'1 --r- 1

+i~"/VI'V;.v"+f"ji'jJ,. K;;;."
= 2 1k 1.1' 17I' 7'" + 1." _!:____ k -1- ~ 1."- k 1'" 17 v" + 1v 1'1' 1J. K "
d s 1 ' d s2 2 1 2 I' I';,. '
hence, returning to the original untransposed problern

(1. 51) (IJk) j" + k D i" = 2 k2 i" + i" _!-____ k +~ i" - k j~' 17P v" + f j~' {' K; 1';.";
1 2 1 2 1 2 ds 1 ds 2 1 2

l
or, according to {1.15) (cf. V 1.10)

(1. 52)
k b i" = (2 k 2 - Dk) i" + i"_}-k + _"2 i"+ j"j~'jJ,. K;;;."
1 2 1 a 1 2 d s 1 ds 2 2

= (k 2 - D k - k 2 ) 1'" 1 f" _!-____ k -1- k k J'" _j_ J. J.I'J' K. ".

I
1 1 2 2 :i
1 ds2'234 2 vp.J..
 1. General deformation problems. 345
g
Now D j" = 0 according to our Suppositions, hence, taking an an-
holonomic coordinate system (h) whose unit basis vectors coincide with
the f", r.
2
i" at all points of c we have
o I
n

I
a
D 1k = 1k 2 - k2 2 + K 2112
a)
d
b) yk=-K21u
(1.53)
I
s 2
kk=-K 2114
c) 2 3

l d) 0 = K2up; p = 5... ,n.

This is in accordance with the results (1.42) derived for R., (c1 in-
finitesimal). The problern of two curves in a general V., with common
normal geodesics having on both curves the direction of the firstnormal
can not b.e tackled in this way. In an R., the components of K."""
occurring in (1.53 b-d) are zero, hence either 3k = 0 and 2k = const. or
k = 3k = 0.
2
The direction of j" can be compared with the tangent direction at
the corresponding point of the other curve if this tangent is displaced
back parallel. This displaced tangent has the direction of the unit-
+
g
vector j" eD j". This gives for' the angle drt. between both directions
according to (1.15) and (1.49b)

Hence the angle doc is constant if k is constant. This is always the case
. an R.,.
1n
2

Exercises.
VII 1,1. Prove that the rule of LEIBNIZ holds for the operators
0 1 0 2 -02 0 1 if it holds for both operators 0 1 and 0 2 and if in the formulae
considered all operations can be effected.
VII 1,2. If the Variations of a quantity f/J (indices suppressed) and
of the connexion F,.";. are given, the quantity (D" V"- V"D)
" (/) is known.
Prove that this quantity does not change if the variation of (/) is changed,
as long as the variation of F,.";. remains unchanged. Prove the same
g g a a
for the operators D V"- V"D and D V"- V"D. 1)

1) SCHOUTEN and V. KAMPEN 1933, 1, p. 8ff.

346 VII. Variations and deformations.

VII 1,3. Prove that

VII 1,3 cx)

if the connexion is not changed.l)

2. Groups of "motions" in V.. and Ln. 2)

Adeformation in V,. is called a motion if it leaves the metric invariant.

Hence an infinitesimal deformation r' --+r' + v" dt is a motion if and
only if !V gA,. = 0. This idea can be generalized. A deformation in U,.
is a motion if f gJ.,. = 0; f S~;_" = 0. In L .. we may call a deformation
an affine motion if the connexion is left invariant. Hence an infinitesimal
affine motion in L,. is characterized by !v F"A p
= 0. Further generaliza-
tions are the conformal motions in V,. characterized by ! g,." = oc g,.",
V
the homothetic motion in V,., being a conformal motion for which cx is
a constant and the projective motion in A" characterized by ! J;,",. =
rp,..A~+rp,.A;. V

The conformal and projective motions are also characterized by

c p c p
! F",.
V P
= 0 and '-f F).
P
= 0 where F..",.
r
and F..",.
r
are the conformal and
projective parameters respectively of J. M. THOMAS (cf. VI 7) and of
T. Y. THOMAS (cf. VI 4). Hence every motion is characterized by the
absolute invariance (cf. II 10) of one or more fields. 3 )
The differential equations of the affine motions in L,. are according
to (1.8)
a) 17,.. V,i" = - v R;~;_"
(2.1) {
b) VA v" = v_i" + 2 S~,i" v~"

with the unknowns v" and v_i". The first integrability conditions (cf.
II 6, III 6) can be derived by differentiation, alternation and using
ScHOUTEN and v. KAMPEN 1933 1. p. 8 ff.
1)

SLEBODZINSKI 1932, 1. Cf. EISENHART 1935, 1; 1936, 1 (simp!y trans. gr.

2)

in V,.); LEVINE 1937, 1 and 1939. 1 (in C,.); 1950,2 (in A 2 and project.); 1951, 1
(in W2); 1951, 2 (in A 2); KNEBELMAN 1945, 1 (in A,. and V,.); YANO 1946, 5;
YANO and TAKANO 1946, 6 and YANO 1946, 7; 1947, 3 (in L,.); 1949, 1; 1951, 2
(homoth. in V,.); Y ANO and TOMONAGA 1946, 8; 9 (in L,.); EGOROV 1947, 1 (in A,.);
1948, 1 (project.); 1949, 1 (in V,.); 1949, 2 (in L,.); 1950, 1 (in L,.); Su 1949, 1 (con-
form); SHANKS 1950, 1 {homothet. in V,.); YANO and lMAI 1950, 2; MUTO 1951, 1
(simply trans. gr. in V,.).
3 ) Many results on these different "motions" are to be found in YANo's

booklet 1949, 1. Cf. also KosAMBI 1951, 1; 1952, 1.

 2. Groups of "motions" in Vn and Ln. 347

the second identity and the identity of BIANCHI. This gives

J a)
+ v,i" R;~~,.- v~" R;~;,e = 0
l
(2.2)
b) : 5"" = v V. s" + v" s ... + v" s "- v"
v I' pl p al l e sQ =
pd pl 0
The second and following integrability conditions are found by co-
variant differentiation of (2.2). As 171' and : can be interchanged on
account of (1.9) and f ~,.l = 0, we get "

(2.3)
{:~ ~;;~k..~~~~~::~; k = 0, 1, 2, 3, ...
These equations express that all covariant derivatives of the fields
R;~;." and s;;." are absolutely invariant for the field v" (cf. II 10).
If the set (2.3) for k = m + 1 depends algebraically on the sets for
k = 0, ... , m, the process stops because all other sets depend for k >m
on the foregoing ones (cf. II 6). The equations (2.1) have solutions
if and only if at some point r' there exist values v", vt' satisfying the
0 0 0
algebraic equations (2.3). Then these values can be taken as the initial
values at ~" and there is one and only one solution of (2.1) taking these
0
values at this point. As the number of independent sets of values
v", v;," satisfying (2-3) is finite, the v" generate a set of one-parameter
0 0
groups of transformations, depending on a finite number of parameters.
L.et u" and v" be two solutions:
(2.4)
then we have (cf. Exerc. II 10,9)
(2.5)
and this proves that : v" is also a solution. This shows that the one-
u
parameter groups together form a group.
The other cases mentioned above can be treated in the same way.
The equations of motion can always be found by equating the LIE
derivative of the fundamental geometric object or objects to zero. The
integrability conditions of these equations only express that the LIE
derivatives of all differential concomitants vanish. 1 )
1) KosAMBI 1951, 1 has hinted at the possibility of a theorem of this kind,
valid for every geometry whose fundamental geometric objects possess a LIE
derivative. But a proof of such a general theorem has not yet been given. We
only know that it is found to be true in all cases so far investigated. The same
holds for the supposition that ~ commutes with allinvariant differential operators
if and only if v" defines a motion.
348 VII. Variations and deformations.

The maximal number of parameters of a group of motions in a V,.

that is not an S,. was proved to be ~ n (n + 1) -1 for n > 2 by FUBINI.l)
EGOROV proved 2) that for a V,. that is not an EINSTEIN V,. the maximum
+ +
is ! n (n -1) 1 and that it is ! n (n -1) 2 for a V,. that is not an S,..
VRANCEANU reduced this latter number in 1951 3) to l (n-1) (n-2)+5.
In 1947 EGOROV proved 4) that for an L,. that is not an A,. the maximum
is n 2 and that an A,. with an n 2-parameter group f motions is neces-
sarily projective euclidean (cf. VI 1). The maximum has been established
by EGOROV and VRANCEANU for a great number of other cases. 5) Y ANO
proved 195 36) that the number of parameters for n > 4, n =I= 8 is (;) +1
if and only if the V,. is either the product space of an R 1 and an 5,._ 1
or an S,. with negative curvature.
For the motions in a V,. we have because of f g;". = 0; f gKA = 0
V V

l
and the commutativity of f and 8,..
V

a) f1J {,.."A} = ~ g" (8,.. fV gAv + 8 f g,... -

V
8. fV g,..A) = o;
(2.6) (also valid with 17 instead of 8)
b) V,. fV g,..A = (fV {.~}) ga- (f{.'J.})
V
gpa = 0

From (2.6) we see that the group of motions in V,. is a subgroup of

the group of affine motions and that the former group is determined
with R"""=K"
by those solutions of (21) vp vp Sp"=0 which also
satisfy the equations

(2.7)

Now from (2.1 a) it follows that 17" v(A><) = 0 because K.,..A .. is altemating
in A. x, hence, if v(A><) is zero at one point of V,., this quantity vanishes
at all points. Therefore an affine motion in V,. is an ordinary motion
in that space if and only if v(A><) = 0 at one point. Then v(A><) is also zero
(cf. Exerc. V 5, 6, 7) at all other points. The equation v(A><)=O is known
as KILLING's equation. 7 )
1) FUBINI 1903, 1 ; 1904, 1.
2)EGOROV 1949, 1.
3)VRANCEANtJ 1951, 2; 1953, 1.
4 ) EGOROV 1947, 1.
5 ) EGOROV 1948, 1; 1949, 1; 2; 1950, 1; 1951, 1; 1952, 1; 2; VRANCEANU 1947, 1,
1949, 1; 2; 1951, 1; 2; 3; this last publication gives a good summary of results
in french.
8 ) YANO 1953, 1.
7 ) KILLING 1892, 1 ; cf. BIANCHI 1918, 1; EISENHART 1926, 1 and for the more

general case where SJJ;.,. = SrJJA><l =I= 0 HAYDEN 1934, 2; E II 1938, 2, p. 163.
 2. Groups of "motions" in Vn and .Ln. 349

We prove here only a few theorems on motions m Vn. 1 ) From

fV g~... = 0 and fV v" = 0 it follows:
Along the streamlines of a one-parameter group of motions in Vn the
length of the displacement vector is constant.
A translation is a motion whose streamlines are geodesics. Because of
(2.8)
we have:
A motion in Vn is a translation if and only if the length of the displace-
ment vector is constant. 2)
We consider two Vn's that can be mapped on each other so that
geodesics correspond to geodesics. Then we wish to prove that the
groups of motions in both Vn's have the same number of parameters.
This is not trivial because the notion of a motion in Vn is based on the
metric and the mapping .is projective and will in general disturb the
metric. Let corresponding points have the same Coordinates ~ and
let g",., L"Jl}, v,.
belong to one vn and 'gjl," 'L.,"Jl}, 'V"' to the other V,.,
called 'Vn hereafter. Because the mapping is projective we have (cf.
III 7)
(2.9) '{;!l} = {;!l} + 2p<,. A):>
and (cf. 1113.19)
(2.10) 8"' log 'g = 8"' log g + 2(n + 1) P,.

which proves that P" is a gradientvector

(2.11)

N ow let d ~ = v" d t be an infinitesimal motion in Vn. Then we consider

in 'Vn the transformation d~ = 'v"dt with
(2.12) 'v" = 'g"!l'v!l = rp 'g""' g"'A vA.

l
For this transformation we get

~ 'g!l,. = 2'P(!l'v,.) = 2v<,. 8!l}rp + 2rp 'P(A v,.)

(2.13) = 2v(,.8!l)rp + 2rpr(!lvx)- 4rpA(Ap,.)vf.'
= 2rpr(Av,.> =rpfg!l,.
V
= 0;
1 ) Cf. for more facts and more literature e.g. EISENHART 1926, 1 or 1949, 1,
Ch. VI; 1933. 1, p. 208ff.; YANO 1949, 1. Ch. IV.
2) EISENHART 1933, 1, p. 212.
350 VII. Variations and deformations.

and this proves :

If two Vn's can be mapped on each other profectively, their groups of
motions have the same number of parameters. 1 )
Of course the groups need not be isomorphic because for two mo-
tions u" and v" in V,, we have in general
(2.14) f 'u" =!= cp 'g"" gpJ. fv u".
'v

It is well known that the group of all motions in an Sn has exactly

~ n (n+ 1) parameters. If we know that the group of all motions in a
V,. has exactly l n (n + 1) parameters, v" and v[-'"l can be chosen arbitrarily
at one point and the fields v", v,i" can then be computed at all points.
Hence the equations
(2.15) k = 0, 1, 2, 3, ...

must be satisfied for all values of v;." satisfying v(J.") = 0 and they should
not contain v". This implies that in (2.2a) the first term in the right
hand side must vanish
(2.16)

and that accordingly this equation can be written as

(2.17)

valid for all values of v[Tal. But this is only possible if the expression
between brackets is zero, hence, transvecting with g"T
(2.18)
from which it follows for n >2 immediately, after transvecting with g"",
that the Vn is an S,.. For n = 2 the constancy of K does not follow from
(2.18) but in this case it follows from (2.16). Hence 2)
+
The group of all motions of a vn has t n (n 1) Parameters if and
only if the Vn is an Sn.

Exercises.
VII 2,1 3 ). An Ln with an integrable non-symmetric connexion is
group space of a transitive n-parameter group if and only if V. S;.i." = 0.
VII 2,2 3 ). An An has the (0)-connexion of an n-parameter group
space if and only if there is a tensor field s;~" alternating in ,u suchthat
VII 2,2 a.:)
1 ) KNEBELMAN 1930, 1; Y ANO 1949, 1.
2) FUBINI 1903, 1; 1904, 1; EISENHART 1933, 1, p. 215; YANO 1949, 1, p. 36.
3 ) EISENHART 1933, 1, p. 197; YANO 1949, 1, p. 21.
 2. Groups of "motions" in Vn and Ln. 351

VII 2,3 1 ). Let $"-+;"+v"dt be a motioninan Ln with an integrable

connexion and let e" be n covariant constant vectorfields. Prove that the

V /1
"
vectors f e" are linear combinations of the e" with constant coefficients.
i

VII 2,4 2). The fields v" and e v" determine both a motioninAn if and
only if
VII 2,4 rx) v" 17,.17, e + 2 (II(,~ e) 17,.> v" = 0.
VII 2,5 3 ). An Ln admits a one-parameter group of motions if and
only if there exists a coordinate system such that either the ~",~ are
independent of e
or that the r;;. are homogeneaus of degree -1 with
respect to the coordinates.
VII 2,6 [cf. 1.8 or III 5.47]. 4 ) The field vj," in L,. is covariant constant
along every streamline of v". Streamlines, along which vj," = 0, are geo-
desics but this condition is only sufficient and not necessary. A necessary
and sufficient condition is that v" lies at all points of the curve in an
eigendirection of vj,''.
VII 2,7. The motion in Vn defined by the field v" is a translation
if and only if the angle between v" and any geodesie is constant along
this geodesic. 5)
VII 2,8 (cf. IV 8). The translations in the group space of a semi-
simple group are the transformations of the first and second parameter-
group. 6)
VII 2,9. Let v" be a displacement vector of a motion in an 5 3 .
The equations of motion can be written in the form (cf. I 9.20)

Prove that this system is totally integrable and that u" is the displace-
ment vector of another motion. 7 )
1) RoBERTSON 1932, 1; YANO 1949, 1, p. 15.
2) YANO 1949, 1, p. 17.
3 ) EISENHART 1933, 1, p. 232; YANO 1949, 1, p. 18. The booklet of YANO
contains in Ch. III a great nurober of interesting propositions on affine motions
in Ln and An. Note that An is used there in the sense of Ln and that an An is called
an An with a symmetric linear connexion.
4 ) A. NIJENHUIS, personal communication.
5 ) EISENHART 1933, 1, p. 212; YANO 1949, 1, p. 32. As NIJENHUIS pointed out
in 19 52, 1, theses No. 4, the proof at these places of the lemma that the trajectories
of two translations meet always under a constant angle is not sufficient and it
seems probable that the lemma is not true.
6 ) CARTAN and SCHOUTEN 1926,3, p. 811; EISENHART 1933, 1, p. 213; YANO

1949,1, p.33.
7 ) TONOLO 1950, 1.
352 VII. Variations and deformations.

3. Deformation of subspaces. 1)

The deformation of a subspace can often be dealt with, using the

method of transposition already explained in V 9 and VII 1. Here
we give a detailed treatment of a vm in vn subj.ected to the infinitesimal
transformation $" -+$" + v" dt. Let a holonomic coordinate system (a);
a=1, ... ,m in vm be completed by basis vectors e"; y=m+1, ... ,n
y
perpendicular to vm so that there results an anholonomic system (h);
h = 1, ... , n, defined over Vm. We suppose that the coordinate system (a)
in vmwith its basis vectors ea, and also the basis vectors
b
are e;.,
dragged along over v"dt. Further we suppose that the e" and e~ remain
y
perpendicular to the e" and
b
e;. respectively and that the equations
!;. ye;. = 15;; t;. e;. = 6: remain valid. This implies that in general the ex
b y

and t;. are not dragged along, because the deformation may disturb
orthogonality. In order to transpose the problem, the Vm and all basis
vectors e", l;. are dragged back over - v" d t and at the same time the

field g;." is dragged along over -v"dt. Then the e" and
b
e;. return to

their original positions but the fields g;."' e" and t;. over V". will have
Y a
been changed. As we have seen before, the absolute differentialsD dt of all
quantities connected with V". in the original problern equal the natural
n
differentials 'D dt in the transposed problem. Hence

(3.1) v,i"=fi':tv"
a n
(3.2) DB/, = 'DB/,= 0
a n
(3.3) DC~ = 'DC~ = 0
a n n
I _ I I _ p). I _ IJ. _
(3.4) { D gcb- D gcb-Bcb Dg"";.-2Bcbv(p.l)-2B(bDc)v;.
(cf. V 7).
1 ) General references: CARTAN 1916, 1 (in Rn); 1917, 1 (in Cn, n';:2 5); 1920, 1
(project.); 2; 1927, 5 (second var.); LEVI CrvrTA 1926, 1 (second var.); SYNGE
1926, 1; 2 and 1934, 1 (second var.); ScHOUTEN 1927, 2 Wm in Vn); Mc CoNNELL
1929, 1 (curves in Vn); HLAVATY 1929, 3 (curves in Un); 1950, 1; 1952, 5 (families
of Vm in Vn); BURSTIN 1931, 1; 2; HAYDEN 1931,1 (curves in Vn); 1934,2 (Xm in
Un); 1936, 1 (Lm in Ln); DIENES 1933. 4 and 1934, 2 (Xm in Ln); BoMPIANI 193 5. 1;
LAND 1936, 1; DAVIES 1936, 1 Wm in Vn); 1938, 1 (curves in Vn); 1938, 2 {V,:');
1942, 1 (second var. Vm in Vn); YANO and MuTo 1936, 2 (second var.); DIENES
and DAVIES 1937. 1 (Xm in vn and An, literature); E li 1938, 2, p. 164ff.; YANO
and ADATI 1944, 7 (par. tang. and concirc.); Y ANO 1945, 9; 10; 1949, 2 (literature);
YANO, TAKANO and TOMONAGA 1946, 10 and 1948, 2 (curves in Ln); CASTOLDI
1948, 2 (finite deform. of V". in Vn); YANENKO 1949, 2 (Vm in Vn)
 3. Deformation of subspaces. 353

If v" is split up into 'v" in V", and "v" perpendicular to V",, (3 .4)
leads to (cf. V 7.15)
a
a) D'g,b=2D<,'vb>-2"v,.H;b'"
(3.5) a
b) D 'gab= -2 'g''" D, 'vb) + 2 "v,. Habz
From (3.1-4) and B~=g;,,.B'b 'la we get

(3.6)

and accordingly
(3.7) IDB). =- DC). = 2 'g"e c~ v(I'-Q) I
a
Because of Dg;,,.=f.g;,,. we have (cf. 1.8)
V

+ vrr K "
Q,

(3.8) Dl:"
uA = f.l:"
v 11-A =V.1'- v"
), <11'-

l
and from this we get for D 'I;~ (cf. 3.2, 6 and V 3.10)

D'I;~ = D(B~t: !;";. + B: 8, B'b)

a a a
(3.9) = B~t f;";,D B: ~ B~t: Df;";, + (D E:) 8, Bt;
= B~t: f. f;";, + (D B:) (D, B'; + 'I;~ B;t)
V

hence
(3.10)

It can be proved in the same way that

(3.11}

By means of (3.8) and the ConAzzr equations (V 8.7) this equation can
be transformed into

In the exercises (VII 3,1 ; 2; 3) we give some other formulae concerning

the V", in Vn. 2 )
1) DAVIES 1936, 1; cf. BRTOLOTTI 1928, 2; 3 (literature); HAYDEN 1934, 2.
2) Cf. for finite deformations of V". in V,. CASTOLDI 1948, 2.
Schauten, Ricci-Calculus, 2. Aufl. 23
354 VII. Variations and deformations.

The V:' in V,., the rigged L". in L,. and the rigged L': in L,. can be
dealt with in the same way. 1 )
The method of transposition cannot be used for a subspace imbedded
in an X,. because in the X,. there is no field that could be dragged back.
But a more general method that can also be used in this case, was devel-
oped by NIJENHUis. 2) Let a rigged X': in X,. be subjected to the
infinitesimal transformation .~-+ ~ + v" dt defined over X,. and let
there be introduced an anholonomic coordinate system (h) such that
before the transformation the e!'; b = 1, ... , m lie in X': and the e";
y = m + 1, ... , n in the rigging. Note that the tangent E". and E"..;
b "

m' = n- m, at ~" after having been dragged along over v" d t do not coin-
cide with the tangent E". and E"., at ~" + v" dt before the transformation.
This implies, that for a vector P" of X':, having components P" =Bi: pb
with respect to ("), a LIE derivative is as yet not defined because after
the transformation the vector is no Ionger a vector in the tangent E"..
We now define the LIE derivatives for the fields pa, qb as the X':-parts
of the LIE derivatives of the fields p", q" considered as fields of X,.
(cf. V 7.36):

(3 .13)

The definition (3 .13) implies that the LEIBNIZ rule remains valid at
least for quantities. We assume that it holds always. If the X': is
holonomic (i.e. consisting of oo"-m X".'s) and if v" happenstobe a vector
of X':, the vector ! P" defined in the ordinary way with coordinates rj"
V
in the X".'s, that is if ob P"- pb ob v", is the X".-part of !p". This
V
proves that !V P" as defined in (3.13 a) has the ordinary meaning in
this special case. By the definition-formulae (3.13) the operator !V
gets the properlies of a v. D. WAERDEN-BORTOLOTTI operator (cf. V 7).
From (3.13) we get immediately (cf. II 9.10)

(3.14) { ~ P" = B: (v~' o~' Bi: pb - pb Bf: o"' v")

* v~' o~' P" - (ob 'v" + 2v~' !J=" Bt) pb;
hence (cf. V 7.25, 26, 35) .

(3 .15)

1) ScHOUTEN and v. KAMPEN 1933, 1; E II 1938, 2, p. 16Sff.

2) Personal communication.
 3. Deformation of subspaces. 355

where :tly is the invariant operator defined in V 7 that is independent

of any connexion in Xn. Note the resemblance of (3.15 a) to the expres-
sion for the LIE derivative in anholonomic coordinates in exerc. II 10,6.
In the same way we can derive

a) f:
'v qb = 'vc 8c qb + (8b'va + 2'vc 'Qa)
cb qa
m
b) ,~qb = "vY (oy qb- 2Q~Y qa) = "vY :tly qb
(3 .16)
c) f qb _,. _ v"' 8"' qb + (ob 'va + 2v"' Q~J. Bt) qa = .~ qb + .% qb
= f:
'V
qb + "vY :tly qb.

l
If we define
a) f r _cl~ c: f r"
(3.17) V V

b) fsy~C~f:sA
V V

for the fields r and sY in the rigging (cf. 3.13) we get in the same way
a) f:r = 'vb :tlb r;
'V

b) fr
"v = "vY 0y r - (8y "v" + 2"v' "Q"')
zy yY
(3.18) m'
c) :tlbr = obr + 2Q~brY (cf. V 7.35c)
d) fr
V
* v"'8p. r"- (8y "v" + 2v"'!J"p CA)rY
y
= fr
'V
+"v:r

(3.19)

All these formulae can also be derived easily ffGlm (Exerc. II 10,6). It
is now easy to derive the general formula for a quantity lying with
some indices in Xn, with other indices in X': and with the remaining
ones in the rigging, for instance

(3 .20) { fv P"a
y
* v" 8I' P"a
Y
- P"a 8 v" - P"b (8 'va
Y !-' Y b
+ 2v" Qa p.
BA)
b
+
+ P~~x (8y "v" + 2v" Q:A C~).
For a density (~) or Wof X': or X';:' with weight 'w or "w respectively
we find (.cf. Exerc. II 10, 6 <5)

a)
(3.21) {
b)
23*
356 VII. Variations and deformations.

For any quantity f1> (IJ') (indices suppressed) of X': (X':') we have (cf.
V 7)
(3.22) ! IJ' = 'vb 'l)b IJ'.
v

Of course 'l)b ('l),.) can not yet be applied to a quantity of X': (X::"')
because so far no connexion has been given (cf. V 7).
An X::' is invariant for a transformation v"dt if and only if every
vector P" of X::' remains a vector of X;:", that is if
(3.23)
is a vector of X;:". Now the connecting quantity depending on v"

(3.24) H:"
V b
det r,V B"b

lies with the index " in x: because

(3 .25) B" H:"-
X V b - B""!B"--
xb V J. - B";.!C"--
Xb V J. - B";.!C"C"-
xb V y .il - 0 '

Hence the necessary and sutficient condition for the invariance of X::'
is that the LIE derivative of Bi: is zero: H;," = 0.
V
In the same way the connecting quantity

(3.26) L"
V y
det r,V C"y

lies with the index x in X::' and it is zero if and only if the rigging X;:"'
is invariant. From (3.24, 26) it follows easily that

(3.27) !Bt=-
V
L,t";
V
b)
and
(3.28) !II BA = - !V CA = Hj,"
V
- L;,
V
".
For Hi," and L;" it is easily proved that
V V

a) Ifi" = C! (v 11 o11 Bi:- Bf: o11 v")

{
(3.29) b) Lv Y" = B"" (v 11 oII C"-
Y
C11Y oII v")

and these equations express that H;," depends on the X::' only and not
V
on the rigging, andin the same way L;" depends on the X::'' only.
V

If v" lies in X::', v" = 'v", we get (cf. V 7.22)

H:"
'v b = C"" ('v' oc B"-
b (ob B") c ob 'v')
c 'v' - B"
(3.30) {
"'
= 2C! 'v' o[ Bi;1 * - 2'v' D;b * - 2 'v' Z~;,"
 3. Deformation of subspaces. 357

and from this identity we see that an X': is invariant for all tangent
deformations if and only if it. is holonomic.
lf 'II' lies in X:', 'II' = "'II', we get from (3.29) and (V 7.35 c)

H."
'''II b
= C"" ("v' 8J B"-
b
(8b C")
,
"v'- C"J 8b "v')
(3.31) { m
=- 8. "v"- 2"v Q;b =- ~b "v"

hence, from (3.30, 31) and the analogaus reasoning for X':'

a) lfi" = - ~b "v"- 2'v' Z~;,"

(3.32) { m
b) LV y" = - ~~ 'v"- 2"v' zZ)l...

lf u" and'll' are both vectorsof X':: u" = 'u"; 'II' = ''II', we get from (3.30)
m
(3.33) f. 'v"
, 11
= f. B"b 'vb =
'U
B"b 'uf. 'vb + 'vb H."
b 'U
= B"b ,f. 'vb- 2'u' 'v6 z"
21 cb '

...
hence f. ''II' lies in X': for every choice of 'u" and ''II' if and only if X':
is holonomic.
In Ch. II 10 we found (Exerc. II 10,9)

(3.34) 2!! = f. ; (u, v)" def !'II' = - f.u"

[u v] (to,v) u v

for all quantities of X,.. But this simple rule does not hold for the
operators f.U and f.V as defined in (3 .13, 17), if applied to quantities of X':
or X';:'. Instead we get from (3.28)

a) 2 r. r. P"
[u v]
r. P" + 2H."
= (v,v) L"" pb
[v b v] ..

b) 2! f.q 6 = f. q6 - 2H;," ~;"q,.

[u v] (u,v) [u vJ
(3. 35)
c) 2! f. r" = f. r"
[u v] (to,v)
+ 2L;"
[u
H;" r"
v]

d) 2!!
[u v]
s" = (u,v)
f. s" - 2L;
[u
H;"
v]
s,.

hence, by applying (3.34, 35) to Bb and C~

a) 2! f. Bb = H;,"- 2H;," L;"

[u v] (u,v) [u v]
(3.36) {
b) 2 f. f. C~ = Lj"- 2Lj" H~"
[u v] (u,v) [u v]
358 VII. Variations and deformations.

and by transvection of these equations with C! and B: respectively

1
a) Hi,''= 2C! ;!;:Bi:= 2;i;Hi,'
(II,V) [11 v] [11 v]
(3 -3 7)
b) L;"= 2B: ;!;;!; C~ = 2;i;L;".
(11,v) [11 v] [11 u]

There is also a commutation formula for the operators ;!; and ~:v
(or ~b) that can be derived from (3.15, 16, 32, 35) u

(~ : -;!; ~ ) pa = ;!; L .a + 2Z: .a Ht" pb

(3 38)
1 a)

b)
:Y V

(~bf:-;!; ~b)
V
U

V
:Y
r
p
=;!; H;"
1'
U

V
:Y : :< U

+ 2Z;,~'" L;"r:Y.
fl
b

So far no connexion or metric has been introduced either in X,. or

in X';:. By the introduction of a connexion or a metric in the rigged X':
only, as was described in V 7 and V 9 respectively we get the rigged
L': and V,."' in X,.. Introduction of a connexion or a metric in Xn leads
to the rigged L': in L,. and the V,."' in V,. respectively. For all these
special cases the formulae derived here for the rigged X': in X,. remain
valid, only new additional invariant operators and formulae arise.
Herewedeal only with the rigged L': in X,. arising by the introduc-
tion of a connexion 'F.~ in the X';:. Then, as we saw in V 7 the ope-
rator ~b can be extended to quantities of X': (cf. V 7,35, 57)

a) ~. P" = D. P" = oe pa + 'F.~ pb

(3.39) {
b) ~e qb = De qb =Oe qb- 'J;~ qa

and the operator ~~' can be defined as

(3.40)
Then we get from (3.15, 16) (cf. III 9.3 and III 5.48)
a) : pa = v~' ~I' pa- 'v;,a pb; 'v;,a = Db 'v 4 - 2' S;,;a 'v
{
(3.4 1) b) ~ qb = v~' ~~' qb + 'v;," q,.; '5;;," =-= 'I[~bl + '.Q~b
For t r; ;!; sy there are no formulae of this kind as long as there is no
V V
connexion defined in xr;:'.
We wish to find :U 'F.~. Although :U and o~' commute, :V and Oe
have this property if and only if the X': is invariant for the transforma-
tion v"dt, because, using the rule of LEIBNIZ (cf. p. 354) we have

(3 .42) -Eo
V C
P"=-EB~'o
V C J1. pa=(-El3~')
V C
oJ.' P"+B~'oJ.'v:EP4 =H~'oJJ p4 +o -Ep4
C VC Cfl
 3 Deformation of subspaces. 359

Hence, if the X';: is holonornic and if v" lies in X';:, it follows frorn
(3-30) that ~ and 8c are always cornrnuting whether (c) is holonornic
or anholonornic, as wastobe expected (cf. II 10.17).
Applying (3.42) to f 'Ilc pa we get
~ 'Ilc P" = f (8c pa +'I;~ frb)
(3 .43) {
= H" 8iJ pa + 8 : pa + 'F" : pb +Pb: 'Facb
V C C V Cb V V

hence (cf. V 7-35 a)

m
(3.44) Pb f 'I;~= (~ 'Ilc- 'Ilc f) pa- IJ; Y 'Ily pa + 2 IJ; y Q~y Pb
Now it can be proved frorn (V 7.59-62) and (3.32a) that

(:v 'Ilc - 'Il'v:) P" - H

v'
Y 'Il P"
(3 .45) { Y
m mm'
= ('vd*R"""+
dcb "vY*Ra+
ycb er\.Vc 'va)
b Pb '
hence
m m
... a + "vY *R"
ycb +er\ b + 2HY
m~
(3.46) :v 'F"cb = 'vd *Rdcb .Vc 'va v c Qaby

m
In order to find ~ Q~ Y we start frorn ~ 8Y pa:
(3.47) : 8Y P" = : C"Y 8I' P" = LvY " 8I' P" + 8Yv: P"
V V

and apply this result to ~ 'Ily P" (cf. V 7. 3 5 a) :

m
(3.48) : ',!)y P"= LC
V y
8 P"+ V C
',!)Y :P"+
V
2pc;tQ"
V cy

or
m
(3.49) (:'.!)
v Y
-'IlYv:)P"=LC'I)
vY c
P"-LC'F"Pb+2PCf:Q".
vY cb v cy

Now frorn (3.15, 19) (cf. 3.45 and V 7.24, 58-61) it can be proved that

(:v 'IlY - 'IlYv:) P"- LvY.c 'Ilc P"

{
(3.50)
= (- 'IlbvY
L"+ 2'5"" Lc_
bcvy 2Z:Y.a"H")
vb pb
'
hence
m
= - '.!) bL y a + L y c ('F"-
~

(3.51) 2 :V Q"by V cb 2 Scb

V
") -2 Z yx
"a H b
1
V

or

(3.52)
360 VII. Variations and de formations.

Note that in the anholonomic case neither :E 'F:;, nor ;E

V V
D:,.
m
is a quantity,
m
as could have been foreseen, and that ~ D:" does not depend on the
connexion in L';:.
If an L';: is rigged in an L,., there is induced a connexion in L';: and
m
we have seen in (Exerc. V 7,1) that in this case *R,i;i/ as defined in
m
(V 7.60) with respect to this induced connexion is identical with RJ;i, a
as defined in (V 7.39). :EV 'F"~ can now also be computed by means of
(cf. III 9.2; V 3.10)
(3 .53) 'F."cb -- B"J.a pA + B"
cb>< F" " 8c B"
b
using (3.24, 26, 27; III 5.47) with the result

:Ev 'F."
cb---"'- PAa
cb" (17.p v"+
).
v" R")-
ap).
(3.54) m m
- H" c.y - L"
vb L" v" H"+
cb vc F"yb
H"

From (3.46) and (3.54) we get after some calculations the following
mm'
relation between *R;;i," and R;~;.",

mm'
ycb - cy Pi.a
*R"i pA"-
cb>< R
(3. 55) m m m' m m'
- ~c (Li,~y+ 2Si,;")- H;i," H;;"+ L;~,.L:Y~b

Exercises.
VII 3, 1. Prove that for a Vm in V,.

VII 3,1 cx)

II
D C"" -
_ - 1g><c (D c " v" + H ctly 1
V
tl) - B"l CP" V.p V .\

VI13,2. Prove that for a Vm in V,.

VII 3,2 cx) { D'F."

cb --'vd'K"+D
tlcb c Db'v+2D (c H"6)." "v"-
- 'g"tl Da Heb:< "v".

VII 3,3. Prove that for a Vm in V,.

VII 3,3 cx) { D'Kacba = ~ 1

Ktlcba + 4D[tl D(b Hc]a]:< "v"-
- 2'KJ;[i," Ha]ex"v".
 4. The holonomy group of an L,.. 361

VII 3,4. Prove that for an X::' in X,.

VII 3,4 cx)
VII 3, 5. Prove that for an X::' in L,.
m'
VII 3,5 cx) vfH"=D
cb H"+B"''"C"fF"
cvb cb "v p.A -HYL"
v v.b

VII 3,6 1 ). If the first and second fundamental tensor of a V,._ 1

in R,. admit the same infinitesimal transformation in V,._ 1 there exists
at least one relation between the principal curvatures (cf. V 6).
VII 3.7. (3.52) can also be derived by taking the LIE derivative of
.Q:., _,. _ BC C~ 8r,.B~1
m
VII 3,7 cx) (cf. V 7).
VII 3,8. Prove that (3.54) is in accordance with (3.46} if for the
'.I;;:, in (3.46) we take the parameters of the induced connexion (use
3.38 and V 7-37).

4 The holonomy group of an Ln.

Let C be a closed curve, in an 91(~). through ~ that can be con-
o 0
tracted into a point and let u" be a vector at ~- Then the components
0 0
of u" after having been displaced parallel along the curve are (cf. VII 1)
0
c
(4.1} u"

= T".A u'-
0
c
where r:A is a tensor depending on the choice of C but not on the
choice of u" at ~- The homogeneous linear transformations (4.1) for
0 0
all choices of C form a group, the holonomy group of the L,.. If instead
of parallel displacements we use CARTAN displacements (cf. III 2) we
get a non-homogeneous linear transformation of the points of the
tangent E,. forming the non-homogeneaus holonomy group of the L,.. 2)
1)ToNoLo 1943, 1.
2)The homogeneaus holonomy group in V,. was already used by ScHOUTEN 1918,
2 who remarked that the dimension of the RN in which the V,. can be imbedded
is n more than the number of parameters of this group. But the systematical
treatment of the holonomy groups is due to CARTAN, 1924, 2; 1925, 1; 1926, 5. He
introduced the name "groupe d'holnomie" though the group vanishes if the E,.
moves "holonomically ", that is if it always retums to its original position. Moreover
the word holonomic or anholonomic gives the false impression that the group
has something to do with an anholonomic coordinate system or with an X': in L,..
The term "circuit group" would certainly be more suitable but the name given
by CARTAN is so widely spread in Iiterature that it seems better to leave it at that
and to warn only against misinterpretation. Some american authors recently
used the term "holonomic group" but this gives the false impression that the
group itself has some property of holonomy. Of course holonomy groups can also be
defined for the more generallinearconnexions mentionedinlll 2, footnote 1), p. 123
and especially for conformal and projective connexions, cf. CARTAN 1937, 1, p. 278ff.
362 VII. Variations and deformations.

C'
If C' is a closed curve in ~ ( ~) through another point ~" and if T
0 1
(matrix notation) is the transformation belanging to it, we may form
a curve through ~ by going first from ~ to ~ along some curve C"
0 0 1
in ~ (~"), then along C' and finally back to ~ along C". If then U is the
0 0
transformation along C" from ~ to ~. we get for the whole transforma-
c o 1
tion u- 1 TU and this must be a transformation of the holonomy group
defined at ~. Hence the groups defined at ~ and ~ are isomorphic
0 0 1
and this means that there is in fact only one holonomy group for all
C' C'
points of Ln. In fact, if T and U- 1 TU are described with respect to two
local coordinate systems that can be transformed into each other by
parallel displacement along C", the numbers in these descriptions are
identical.
If for C we take the boundary of an infinitesimal facet d !""' at ~.
the vector u" is (cf. III 4) 0
*
(4.2) u"
* = u"-
0 :r RV J-l),
1 "u;.df"'.
0

For every choice of the facet this is an infinitesimal transformation

of the holonomy group. But this does not imply that these at most
(;) linearly independent infinitesimal transformations generate this
group. For the non-homogeneaus holonomy group we must use the
transformations of the vectors tagether with the transformation of the
point that originally coincided with the contactpoint and that after
the displacement has the radiusvector (cf. III 2)
(4.3) -S;~"df""'.

In order to find all infinitesimal transformations of the homogeneaus

holonomy group we may use the results of VII 1. 1 ) With the notations
introduced in that section we have for a curve arising from C by a
displacement e v" (t) with v" (t0 ) = v" (t 1 ) = 0
g
(4.4) 1{ = ([::1~ + e D l::1~) ~;.
and accordingly the infinitesimal transformation belanging to these two
curves is given by the tensor (cf. 1-36)
t,
(4.5) A"), + e lt1" b lt't 1a =
t1 <1 0 ),
A"-
),
eJ lt ]T v .!:_e_ R" .alt]" dt.
t0 ). dt VJ-lT t <7

t,
1) There are many other methods. Cf. for instance KANITANI 1952, 1 where
the transformations are developed in series.
 4. The holonomy group of an Ln. 363

Now we may take all curves through ~ and all deformations sv", hence
every tensor of the form 0

(4.6) lt1'p d/;"'

c"

= df R .. "ll]"
10 a V{!T I

is a generator of the group (cf. IV 3), i.e. A~ + C,i"dt is an infinitesimal

transformation of the group, and the same holds for all derivatives of
C.i'' with respect tot. In (4.6) d~"fdt and R;~;" are values at the point
~"(t) and p is an arbitrary vector at that point. Using (1.30, 31) we
find for these derivatives

C," ll ]' {ddff;w J7 p , .. "} [11"

1
~
d f , ).
= d /;1'
df
R V{!T
lo). W t G
(4.7)
~
d tk
C'"
i.
= lt ]' {(~;;w_
1o dt
V, )k p d/;P_
w dt
R. "} lt1"a I' T I

and in particular for t = t0

c"

= p d/;11
dt
R""
vpi.

(4.8)

Conversely all generators depend linearly on the expressions (4.8) because

the expressions (4.7) can be expanded in powers of t- t0 and the coeffi-
cients in these series are the expressions (4.8) multiplied by constants.
Hence the generators of the group are the quantities

(4.9) p q" R;~;. "; qw Vw p q"' R;~;, "; ... ; (qw Vw)k p q"' R;~;,"; ...
for all fields q" and all fields P" with P" = 0 at ~- That proves that the
0
generators lie in the ~-domain of the quantities
(4.,11)

Moreover, as was proved by NIJENHUis 1 ) both domains coincide. Hence

The generators of the holonomy group of an Ln span the ~-domain
of the curvature tensor R;;';." and its covariant derivatives.
Let r ;;;:;; n 2 be the rank of this ~-domain and let the domain be
spanned by the r linearly independent tensors C(i,l;,"; b = 1', ... , i:. Here
b is a dead index but it may be considered as a living one belanging
to an auxiliary E, in which every tensor of the domain is represented
by a contravariant vector.l) Then Cb;." is a connecting quantity
1 ) NIJENHUIS 1952, 1, Ch. II 13.
364 VII. Variations and deformations.

between this E, and the -domain. The infinitesimal transformations

of the holonomy group are
. _l 0
X b = Cu.... x---
(4.12)
ox"
where the x" are rectilinear coordinates in the tangent E,. and from this
we see that the E, is identical with the tangent E, at the point of group
space corresponding to the identical transformation (cf. IV 2). Now

l
R;~;." and its covariant derivatives can be written in the form

a) f!;~;." = B;/ c;,;."

(4.13) : k
b) "-
17,w~"'> Rvp. 1 vp
- B wkw 6 c"
bA ' k =1,2, ...
k
with uniquely determined coefficients B and B.
To every point of L,. there now belongs also besides the ordinary
tangent E,. a tangent E, independent of this E,., and covariant dif-
ferentiation can be extended to the quantities of these E,'s by intro-
ducing the condition that the covariant derivative of the connecting
quantity C;,;." be zero for displacements in L,. . 1} In fact, if we put
(4.14} 0 =V~' C;,;," = o~' C;,;."- r;:6 C~;."- F".llA C;,~" + T"."a C;,;.a
the r::b
are determined uniquely because c;,~" has the maximum b-rank.
If the F;:6 are found it follows from (4.13) that

17,W B 4 = h
1
4
PJ.I WVJ.I

(4.15}
k k+1
Vw B";J: ... c:.,;~" = BC:.wi ... :";;"; k = 1, 2, ...
It is easily proved that the coefficients r;:b
transform in the following
way if the coordinates in L,. and the (rectilinear) coordinates in E,
are transformed:
b'a J.pb + a Up
= A~'p' Ab"'
(4.16} nt~:
J.p'b'
Tltl, A"'., A"b'

For the part R;~;," of the extended curvature tensor (defined in the
ordinary way) concerning the E, we find
(4.17}
Because the covariant derivative of C;,;." vanishes, the following identity
holds
(4.18} 0 = 2 Vr J7~'1 C;,,t" = - R;~;," C~,t" - R;;;.a C;,;" + R;~;" C;,,t .. ,
1) Cf. ScHOUTEN 1926, 1.
 4. The holonomy group of an L,.. 365

hence, according to (4.13)

(4.19)
Since Cb}." has maximum b-rank, this implies that there exist uniquely
determined coefficients c;b" such that
a) c"C"-C"C"=c"C"
cA ba bJ. ca cb a).
(4.20) {
b) "4
R~pb = B c"
~JJ cb

It follows from (4.12, 20a) that

(4.21) (XC X)
b
"--x--1 8x'J
= 2c [c/J./ 8 8
Cb]p.TXJJ 8xT = c"X
Cb a

and this proves that the c~ba are at each point of L,. the structural
constants of the holonomy group at that point. In fact they satisfy
the identities (cf. IV 2)

(4.22)

as can be checked easily by means of (4.20a). By covariant differentia-

tion of (4.20a) we get because of (4.14)

(4.23)
The c;b" are not in general constants throughout L,. as follows from the
fact that the coordinates in the tangent E;s are chosen quite arbitrarily
at every point. But if, starting from ~ we displace the local coordinate
0
system in E, parallel from ~ along some curve to an arbitrary point ~'
0
the c;b" remain constant. If the system retums to ~ by some way or
0
other, always moving parallel, the local coordinate system has suffered
in the end a linear homogeneous transformation but the initial and
the final values of c;b" are the same. Hence the coordinate systems in
the tangent E;s can always be chosen in such a way that the c;ba are
constants over L,., but it is not necessary to do so.
Now let C be a closed curve through ~ and let the tangent E,. and
0
the tangent E, at ~ be displaced parallel along C until they are back
0
at ~. Let a general vector u" at ~ be changed by this process into
0 0

(4.24)

then T is a general transformation of the holonomy group and every

transformation of this group can be obtained in this way. If C,i" is any
366 VII. Variations and deformations.

generator of the group at !;", this quantity changes into

0

(4.25)
and this equation thus represents a transformation of the linear adjoint
group, that is the adjoint group applied to the infinitesimal trans-
formations of the holonomy group only (cf. IV 2). Now for every
choice of Cj_" we always have Cj_" = eb Ci,)." with uniquely determined
constants eb and the infinitesimal transformation represented by Cj_"
is at the same time represented by the vector eb in the tangent E,.
The quantity Cb)." is covariant constant, i.e. it remains unchanged if
it is displaced parallel along C. Hence we may write 'C;."='ebCb;."
and then we have
(4.26)

The b-rank of Cb.i." has the highest possible value r and therefore at
-1
least one quantity c~:;. must exist such that

(4.27)

By means of this quantity we get from (4.26)

-1 -1
(4.28) 'ea=T.aceT"CaJ.
). ba e "' eb
and this equation expresses how the vector ea is transformed if the
infinitesimal transformations of the holonomy group suffer a trans-
formation T of the linear adjoint group.
In the same way as above it could now be proved that the generators
of the linear adjoint group span the g-domain of the set

(4.29) R ... a.
'Vflb '

and, according to (4.14, 15, 20b) also the g-domain of the set
1 k
(4.30) B;,;;c c;ba; ... ; B;,k ... :o,;;c c;ba; ....
But because the set of all B's has the maximum c-rank r, this proves
that the generators of the linear adjoint group span the g-domain of c;ba
exactly, a result well known from IV 2.
The holonomy group of an Ln is intimately connected with covariant
constant fields of quantities. We prove the theorem 1 ):
1) \VAGNER 1943, 1; 1946, 2. This paper deals with general geometric objects.
The proof is from N IJ ENHurs 19 52, 1, Ch. II 13.
 4. The holonomy group of an L,.. 367

A covariant constant jield of a quantity with a given kind of trans-

formation in an L,. is possible if and only ij there exists a quantity of
this kind which is invariant for the holonomy group.
The proof is given here for a tensor P~;. of valence two. The generali-
zation is obvious. The equation
(4.31)
has the first integrability condition
(4.32)
and the following integrability conditions are obtained by successive
covariant differentiation of (4.32) (cf. III 6):
(4 33) ") P",}. - (17."'k"' R"
(17."'k"' Rvpa vpA
Q) P" I I-- O k =1, 2 ....

Because of (4.13) the conditions (4.J2, 33) are equivalent to

(4.34)
and this equation expresses the invariance of P~;. for the holonomy
group. 1 )
If the L., is a V,. special results can be obtained. Instead of (4,13a)
we get
.. bc .
(4.3 5) K pAx = B vp b><'

hence, according to the fourth identity (cf. III 5)

(4.36) .. bcb.l.><
B vp = B"bC
Ax bvp

1) There are a great number of investigations on holonomy groups that leave

certain geometric objects invariant. We give here some references. Symmetrie
tensor of valence two in V,.: LEVY 1925, 2; SAMUEL 1945, 1; LICHNEROWICZ 1950, 1;
Bivector in V,.: !WAMOTO 1950, 1; LICHNEROWICZ 1950, 2. Multivector in V,.:
LICHNEROWICZ 1951, 1; 2; lWAMOTO 1951, 1; Ep-field in A., and V.,: ABE 1944, 1; 2;
RusE 1949, 1; 1950, 1; WALKER 1949, 1; 1950, 3; 4; YANO and SASAKI 1948, 3;
WILLMORE 1951, 1; Pointfield in V.,: TACHIBANA 1949, 1; Hyperquadric in anormal
projective connexion: SASAKI and YANO 1949, 1; 0TSUKI 1950, 1; 1952, 1; Point
or hypersphere in anormal conformal connexion: SASAKI 1943, 1; 2; 3; YANO and
SASAKI 1947. 2. Connected with these.investigations are a number of papers on
classification of connexions by means of the holonomy group: SASAKI 1942, 1
(conform.); 1943. 1; 2; 3 (conform.); 1951, 1 (one-parameter groups); 1951, 2 (global);
YANO 1943. 5 (conform.); YANO and SASAKI 1944, 13 (conform.); 1946, 4 (con-
form.); 1947, 2 (conform.); 1948, 3; WAGNER 1949, 2 (L.,); 0TSUKI 1949, 1; 2
(V,); 1950, 2 (conform.); 1950, 3 (W4 ); 1951, 1 (conform.); LIBER 1949, 1 (one-
parameter groups); DEBEVER 1950, 1 (conform.); KANITANI 1950, '; 1952, 1 (pro-
jective); KUIPER 1951, 2; TAKIZAWA 1951, 1; HASHIMOTO 1951, 1; KURITA 1952, 1;
WAKAKUWA 1952, 1; NIJENHUIS 1953, 1. Further Iiterature is given in these papers.
368 VII. Variations and deformations.

But this implies that the v,u-domain of B;;b and of Cb.,.. are identical.
Hence there must exist equations of the form
(4.37)
where a"b is a uniquely determined symmetric tensor that can be com-
puted from (4.35). Since the b-rank of Cb.,. is r, the rank of aab is equal
to the a-rank of B;;a, and to the .x-rank of K.,..AK
The rank of the symmetric tensor a"b has the maximum value r if and
only ij the Ax-rank of K.,.i. .. equals r, or in other words, if the Ax-domain
o/V..,K,,..A>< lies in the Ax-domain of K.,..",.. 1 )
From (4.37) and (4.13) it follows that
(4.38)
and
k
(4.39) . a = CbPJJ V.WJ: Wt a"b
BQ)J:COttiJ'

and from this latter equation we see that the a-rank of the set consisting
of aab and all its covariant derivatives equals r.
As an example we consider a V,. with recurrent curvature tensor
(cf. VIII 9). This is a V,. with a vector field k.., such that

(4.40) V..,K.,..A>< = k..,K,,.A><; k..,=l= 0.

From (4.38, 40) it follows that
(4.41)
and from BIANCHI's identity, because a"b has rank r
{4.42)
But this equation expresses that for every choice of b the bivector Cb
is simple and has a 2-direction containing the direction of k..,. This
implies that cbp can be written in the form
(4.43)
in which ub is uniquely determined to within an additive term of the
form ub k, with an arbitrary vector u~. Now we may derive from {4.40)
(4.44)
but on the other hand we have (cf. 111 4.9 a)
(4.45)
1) NrJENHurs, personal communication.
 4. The holonomy group of an Ln 369

Hence 17["k01 lies in the ae-domain of K"e""' and thus in the ae-domain
Of CbUQ alSO, and thiS implieS that there iS an equati0n Of the form
(4.46)
This proves that k;. is a gradient to within a scalar factor (cf. II 7).
Substituting (4.46) and (4.38) in (4.44, 45) we get

(4.47)
{ 'kacaaeacaccp cd,;.,.=-aabc
ab
aa11 cb[v acac ICTI!Jl cd;.,.-

,T

T cd
-a CaaQCb[J. a Clcv,.CdTI><l
from which it follows immediately by writing out
(4.48) 'k(a acd) Caae Ccvp CdJ.><-
- 0

which is only possible if

(4.49)
that is, if 'ka is zero. This proves that k;. is a gradientvector. 1) If we write
k;. =-V;. log 0' we get from (4.41)
(4.50)
The Ci,i"; b= 1, ... , i: are the generators of the holonomy group,
hence, from (4.20a) and (4.43)

N ow the r vectors ub;.; b = 1, ... , i and k;. are linearly independent

because Cb.,. has b-rankr (cf. 4.43). Hence (4.51) is equivalent to
a) k" k" u[c IJ.I u;1 ~ = 0
(4.52) {
b) 2ur~" u(,t k;.k" = c;(,au~" k;..

(4.53)

and if this substituted in (4.52b) we get c;;a = 0 because u(," has b-rank r.
Now for r >1, (4.53) would nply that Ui;. and u2;. were linearly
dependent, hence r = 1. This proves 2)
1) WALKER 1950, 1, the proof given here is from NIJENHUIS (personal communi-
cation). Cf. RusE 1946, 3; 1948, 2; 1949, 2. If the scalar curvature is not zero
the proposition can also be proved by transvecting (4.40) with g" g~'J..
2) NIJENHUIS, personal communication.

Schouten, RicciCalculus, 2. Auf!. 24

370 VII. Variations and deformations.

The holonomy group of a v..

with a recurrent curvature tensor is a
one-parameter group if k" is not a nullvector.
Now let us suppose that k" is not a nullvector and that the 2-direc-
tion of the simple bivector Cb";,.; b = 1 is not tangent to the nullcone.
Then the holonomy group consists of all rotations in the tangent R,.
around this 2-direction. This means that the 2-direction and the
(n- 2)-direction perpendicular to it are covariant constant because
both are invariant for the holonomy group. But this implies that the
V,. is a product V2 X V,._ 2 (cf. V 11). The V2's aregeodesie and parallel
and their tangent R 2 is spanned by ub and k (cf. 4.43). The V.. _ 2 's
are also geodesie and parallel and their curvature tensor is the v..
-2-
part of the curvature tensor in v.. (cf. V 10, GAuss). But according
to (4.38) this part is zero. Hence
I f in a V.. with a recurrent curvature tensor the vector k;,. is not a null-
vector and if the simple bivector of the rotations of the one-parameter
holonomy group is not tangent to the mtllcone, the V.. is a product V2 X R.. _ 2 1 )
Note that accordingly every ordinary v.. with a recurrent curvature
tensor is a product V2 X Rn_ 2 2 )

Exercise.
VII 4,1 The non-homogeneaus holonomy group of a v.. leaves
3 ).

a point invariant if and only if there exists a concurrent contravariant

vector field (cf. VI 9).
It leaves a direction invariant if and only if there exists a covariant
constant contravariant vector field.

5. Affinemotionsand the holonomy group in a symmetricAn.

In a symmetric A .. (cf. III 7) the covariant derivative of R;~l''
vanishes and consequently the holonomy group is generated by the
1-domain of R;~;." only. This means that in the factorization (4.13 a)
(5.1)
the a-rank of B;~a and the a-rank of C~;." both are maximum and equal
to the nurober r of parameters of the holonomy group. Hence the quan-
1) WALKER 1950, 1, p. 46. In this paper the author gave a detailed discussion
of all other cases. He proved that the Vn can also be a product V3 X R,._ 3 or
V4 X Rn_ 4 where V3 and V4 have also recurrent curvature tensors. The proof
given here is from NIJENHUIS (personal communication).
2 ) We mention some further recent publications on holonomy groups: LAPTEV

1950, 1; SASAKI 1951, 2 (symmetric spaces); KANITANI 1952, 1; BoREL and

LICHNEROWICZ 19 52, 2 (Vn).
3 ) Y ANO 1943, 3.
 s. Affine motions and the holonomy group in a symmetric An. 371

k
tities Bin (4.13b) vanish. The equations (4.13a), (4.14) and (4.16-23)
remain valid but instead of (4.15) we have

(5.2)

and there is another identity

5.3)

that can be derived from 17[wl7vJ B~;,a = 0 because the a-rank of B;~a is r.
The equations for the group of affine motions are now [cf. (2.1)]

a) 17'" v,i" = - v R;~;,"

(5 .4)
b) VA v" = v.t

and the integrability conditions [cf. (2.2, 3)] reduce to

This equation expresses the fact that in the tangent En the tensor
+
R;~;," is invariant for the infinitesimal transformation x"-+x" Vi" x"dt.
If (5 .1) is substituted in (5. 5) we get

(5 .6) (v"
v B"~".a + v" Bva.a) C""
I" aA + B .a (v"
I"A Caa"- v"
e CaAe) = 0

But because C~;," and B;~a both have maximum a-rank, (5.6) can only
be true if there exist uniquely determined coefficients v;,a such that

a)
(5.7)
b)

The v;,a are the components of a tensor in the auxiliary tangent E,'s intro-
duced in VII 4 and (5.7) expresses the fact that in the tangent En+r
spanned by the tangent En and the tangent E, the quantities B~;,a and
Ca" are invariant for the infinitesimal transformation
(5.8)

in the tangent En+r Moreover (5.7) isanother form of the integrability

conditions (5.5). As to C~;," this means that the holonomy group is
invariant for all affine motions. Of course this is also true in a general A,.
because a motion does not change the geometry in the space and con-
sequently can not change the holonomy group. If pa is arbitrary,
24*
372 VII. Variations and deformations.

pa C~,i" is an arbitrary generator of the group. Taking the LIE derivative

of this quantity we get according to (5.7b)

(5.9) fV P" ca.l." = ca.l." (v~-'17. t',J.,a- pe ve

f.l
4)

and this proves in fact that the totality of allinfinitesimal transforma-

tions of the holonomy group is invariant for affine motions.
If r,.. is applied to (5.7b) it follows, using (5.4a) and (4.18), that

(5 .1 0)

From (5.7) and (4.20a) it can be deduced that

(5.11)

and this equation expresses the fact that in the tangent E, the quantity
c;ba is invariant for the infinitesimal transformation X 4 -7X4 +vbaxbdt.
The equations (5.4, 10) and (5.7) are equivalent to (5.4, 5) and thus
(cf. VII 2) form a totally integrable system with the unknowns v", v.i"
and vb" In order to prove that the solutions of this system generate a
group we can use the same way as was followed in VII 2. Thus we
extend the LIE Operator f to the xn+r formed by the oon tangent E;s.
V

Of course this can be done in several ways, but if we put the condition
that f C;;." = 0 we get uniquely for a vector of E, [for instance for the
V
contravariant vector pa from (5.9)]

{
a) ~ P" = v~-' V,..pa- Vi/Pb= v~-'o,..P" + v~-' ~abpb- vb"Pb
(5 .12)
b) f qb = v~-' V,.. qb
V
+ Vi," qa = v~-'o,.. qb- v~-' ~ab qa + vbaqa.

It is interesting to look at these equations from the point of view of the

Xn+r. The general formula for the LIE-derivative of a vector uA;
A = 1, ... , n, 1, ... , r in this space with respect to the field vA, resulting
from v,. and V4 , is
a) fV u,. = v~-' of.l u,. + '/Je oe u,. - u~-' of.l v,. - ue oe v"
(5.13) {
b) f u" = v~-' o,.. u" + ve oe ua- u~-' o,.. V 4 - ue oe v".
V

Fora field uA with ua= 0, only dependent on the ~". this gives the ordinary
expression for f u... But for a field with u" = 0 and ua depending on
V
 5. Affine motions and the holonomy group in a symmetric An 373

the f' only, we get

f a) !u"=O
(5.14)
l b) f ua
V
= v~' oll ua- uc ac va

because v" depends on the ~" only. Now (5.14b) is in accordance with
(5.12a) if
(5.15) vi/=obva+v~'r:b

This means that the LIE derivative f

V
pa in (5.12a) is the same as
would have been found if the vector field v" were extended with a field va,
dependent on the ~" and the xa, to v", va, in such a way that ob va
were satisfying (5 .15). This implies that
(5.16)

for all quantities of xn+r but this equation can also be derived directly
from (4.14, 22) and (5.1, 7, 11). It is very remarkable that va itself
does not occur in (5.14b) but only its derivative obva. Hence, if va is an
arbitrary field not depending on the xa, the field va = t
+(via- v~'* ~ab) xb
satisfies the condition (5 .15).
The group property can now be proved by showing that w" d_e! fu v"
is a solution of (5.4, 7, 10, 11) if u" and v" are solutions. This is done
by substituting in these equations
a) w" cJ:f f v" = u~' 171' v" - v~' 171' u" = u~' v;" - v~' u;"

(5 17) \ b) w;." -def ~;. w" = 17.;.uf v" = uf v;." = u v~' R vpJ.
" + v"" u ;."- u a" v;. 0
c) wa<ieJ
b -
f,. va
b
= u v~' Ra
vpb
+ vau.c-
c b
uav.c
c b

The equations (5 .17) give the composition of two infinitesimal affine

motions. If we drop the curvature tensor in (5.17b, c) these equations
give the well known composition of affine motions in En where u" d t,
v" dt and w" dt are the displacements and Ui" dt, Vi" dt and Wi" dt the
affine transformations of the tangent En's.l) From (5.17a) it follows
that the affine motions leaving one arbitrary point ~invariant, form
0
a group. This group is called following CARTAN the group of isotropy
with respect to ~".
0
We prove now that the group of affine motions in a symmetric An
is transitive and that it has at least n + r parameters, where r is the number
of Parameters of the holonomy group.
1) Some authors call them "affine rotations ".
374 VII. Variations and defonnations.

The equations (5.4, 7, 10, 11) are totally integrable, hence there
'exists a solution for every set of initial values v", vi" and v;,a at t', that
0 0 0 0
satisfies (5.7, 11). But because v" does not occur in these conditions,
v" can be chosen arbitrarily and this proves that the group is transitive.
0
Now according to (4.20a, 22b; 5.3) the set defined by

(5 .18)

satisfies (5.7, 11) for every choice of the constants va. This proves
0
that the group has at least n + r parameters. The group with the
parameters v", va defined by (5.18) is an (n+r)-parameter subgroup
0 0
of the group of affine motions. First we remark that (5.18) implies
that equations of the same form

(5 .19) a) v"-vc"
A - cA '
b) va-
b - vca
cb

with suitable values of if, hold at every point of A,.. In fact, according
to (5.1, 4, 10) and (4.14, 20b) we have

(5 .20) a) c"r7
c A Y p v= -v~ Bc"
~I' cA '

or
(5 .21)

and the integrability conditions of this equation are identically satisfied

because according to (4.20b) and (5.3)

2 Ji[. 17,.] Va = - 2(Ji(. vw) Bl,;,l~la = - 2v C;[;w Bl,;,l~la

(5.22) {
= v c;i, a B;~b = vb ci,;" B;~ = R;~;," vb.
That proves that there is one and only one field v" that satisfies (5.19, 21)
and takes at t' the values v" satisfying (5.18).
0 0
Now we prove that the subgroup defined by (5.18) is an invariant
subgroup of the group of affine motions (cf. VII 2). Therefore we
substitute (5.19) in (5.4, 10) and (5.7, 11). Then (5.7, 11) are satisfied
identically as a consequence of (4.20a, 22b; 5-3), and (5.4, 10) reduces
to the set
(5 .23)

which is of course totally integrable. If now u" is the vector of

an affine motion not necessarily belanging to the subgroup it follows
 6. CARTAN's method applied to the holonomy group and the symmetric A,.. 375

. .
from (5.17, 23) that f 'II', f v" also satisfy (5.23) and this means that
the subgroup is invariant.
Exercise.
VII 5,1. Prove (5.22} by writing out 2~.171' 1 v_i" in two different
ways and using (4.18) (use the ruleof LEIBNIZ for ~.171'1 ).

6. CARTAN's method applied to the holonomy group

and the symmetric A ...
In order to give further examples of CARTAN's symbolical method
we apply this method here independently of VII 4, 5 to the homo-
geneous holonomy group. The equations of VII 4, 5 will not be used in
this section but results will be compared.
Starting from any point ~ of an L,. we fix an anholonomic coor-
o
dinate system (h) at all points of an 91 ( ~) by parallel displacement of
0
some given local coordinate system in the tangent E,. of ~ along any
0
definitely chosen curves to all points of 91 ( ~). If this local system at ~
0 0
suffers a linear homogeneous transformation, the local coordinate
systems of (h) at all points are transformed in the same way and all
these transformations are even numerically the same if they are ex-
pressed at all points with respect to (h). When this has been done we
consider now at every point of 91 ( ~) not only the local coordinate
. 0
systems of (h) but all coordinate systems derivable from them by means
of all transformations of the holonomy group. The whole set of all
local coordinate systems obtained in this way will be called the allowable
local coordinate systems.
If from this set we choose in an arbitrary way one local coordinate
system at every point of 91 ( ~) we get an allowable anholonomic coordinate
system in that region. 0

It is clear that the set is independent of the choice of the curves

used above. If the holonomy group is an r-parameter group, the local
coordinate systems depend at each point on r parameters. If ~ and
1
~ are two points of 91 ( ~) and if the tangent En at ~ is displaced parallel
2 0 1
from ~ to ~ along some arbitrary curve, the local coordinates of the
1 2
points of E,. at ~ with respect to some allowable local coordinate system
1
and the local coordinates of the displaced points in the E,. at ~ with respect
2
to some other allowable local coordinate system always transform into
each other by a transformation of the holonomy group.
376 VII. Variations and deformations.

Now let xh be the radiusvector of a point of the tangent E,. of a

point f' in 9C ( ~) with respect to some allowable local coordinate system.
0
The general infinitesimal transformation of the r-parameter holonomy
group has the form (cf. IV 3)

(6.1) dxh=Ei(x')(drJ)b; h=1, ... ,n; b=L ... ,r.

But because this group contains linear homogeneaus transformations

only, dxh must be linear homogeneaus in xh, hence there exist equa-
tions of the form
(6.2)

in which the c;;" are constants.

If a vector v" is displaced parallel from ~ to ~,. + d~ we have for
the anholonomic components v" (cf. III 2.10; 9.1; 10.1)

(6.3)

and because of the choice of the allowable local coordinate systems

this transformation is a transformation of the holonomy group in the
variables v". Hence there must be a relation of the form

(6.4)

in which the o; are not in general constants.

The infinitesimal transformations of the holonomy group are

(6.5) X b-- c-" a.

bi X ()xk

and as we know tbat (Xe Xb) = c;;" X .. , it follows by Substitution of

(6.5) in this equation that (cf. 4.20a)

(6.6)

Now we substitute (6.4, 6) in the structural equations (III 10.4, 5) of

an A,.. Then S~j" = 0 and we get

{
a) " Ah
U[p 4).] = 1C"A'
~ ai lla
[p V).]
(6.7)
C b) [dA"]= tC~;" [A'Oa]
and
.. k"U[v VI']
a) Cai lla
-
.lc"C"llC
2 cb
llb -Rh
ai U[v U!i] - pi
(6.8)
b) C~i 11 [d oa] - i c;i," C.;; 11 [OC Ob]= R;n
 6. CARTAN's method applied to the holonomy group and the symmetric An. 3 77

or [cf. (4.13)]

(6.9)

where
a)
(6.10) {
c b)

As an example we consider first the case of a symmetric An (cf. III

7, VII 5). Then we have V"'R;;.i''=O and the components Ri,jl"
have the same values at all points with respect to all allowable coordinate
systems (h).
The infinitesimal transformation of the radiusvector x" in the tan-
gent En (cf. III 4.1)
(6.11) x"-+x"- ~Ri.f/" xidjki
is the transformation of the holonomy group belanging to the facet
d jki. Hence there must exist relations of the form (cf. 4.13 a)

(6.12)

in which now not only the C~;" but also the coefficients Bi.t are constants.
Substituting (6.12) in (6.8a) we get
(6.13) Cai.. h("'ur. eaJl] -
lcaeceb
2 cb [ p] -
BaAkAi)-0
kj [v l'l - '

and this is equivalent to

a)
(6.14) {
c b)

because the a-rank of C~;" is r.

We retum to the case of a general An. The allowable local coordinate
systemsinAn may be considered as depending on the general coordinates
r;""; rx. = i, ... , .Y in the group space X, of the holonomy group. To this
end we have only to fix one allowable local coordinate system at every
point of 91( ~x) and to give these systems the Coordinates 'rJa. belanging
0 0
to the identical transformation. Of course an allowable anholonomic
system (h) can then be fixed by giving the 'rja. as functions of ~x. In
the same way as in III 10 the A1 may be considered as the section
of the A~; B=l, ... , n, i, ... , i in the Xn+r of ~", 'rJ"' with A~=O. But
in contradistinction to III 10 we have now r:;;, n 2 and in the X, we
have not only the holonomic coordinates 'rja. but among others also the
378 VII. Variations and deformations.

anholonomic system (a). Hence in Xn+r we have besides the holonomic

r
system (x, oc); 1, ... , n, i, ... , the three anholonomic systems (", a);
1, 000' n, 1, 000' r and (h, ot); 1, 000' n, i, 000'; and (h, a); 1, 000' n, 1, 000' i:.
According to (6.2, 5) the infinitesimal transformation of A1 by the
holonomy group is given by

(6.15)

and from this equation we get

(6.16) A""'
i ur Ah
A = A"C~o
y ai

Hence, if we write
(6.17) () a def _ A"
y- )'

we get besides (6.7a) the equations

" Ah] -1ChAi
ur" - ~ ai
()"
[y ]
(6.18) {
8r"A~l = tC~;h A~"Ol; (Note that A~ = 0)

that can be gathered tagether with (6.7a) into the equations

B, C = 1, ... , n, i, ... , i;
h, i = 1, ... , n; a = 1, ... , i:.
Note that (6.19b) has exactly the sameform as (6.7b), though (6.19b)
contains much more.
Now we will prove that the equations (6.14) for the symmetric A,.
can be generalized in the same way to

a) " ()"
U[C B] = 1 alle Ob
:!" Ccb V[C B]
+ BaAk
ki [C
AiB]
(6.20) {
c b) [d0 4 ] =! c;,;a [()' ()b] + Bj,j" [Ak Ai].
We note that (6.20b) has the same form as (6.14b) but contains much
more. We look at the I;~ with respect to a definitely chosen allowable
anholonomic coordinate system (h). If a transformation of the holonomy
group is applied simultaneously to all local coordinate systems of (h)
we get a new allowable anholonomic coordinate system (h') and the
Af, A~, are constants. That means that in this case the transformation
I;~-r,.~: is linear homogeneaus (cf. 4.16). For an infinitesimal trans-
formation of this kind it follows from (6.15) that

(6.21) dF..~
P
= (Cbt: h r..i.
1-"
_ C":b i F..~)
PI
(d'YI)b
"I
 6. CARTAN's method applied to the holonomy group and the symmetric An. 379

hence, according to (6.4, 6, 17)

(6.22) { "' r" cdi.. "J.tii-

u~ f.d= rr; c .. i E."l'i di

= 1
~
(C 61~" C~i-
a Cd:i Ca1~") 6"I' = - 1
~
c" I' (Jb~
cb Ca"(Je
But on the other hand we have from (6.4)

(6.23) "' r" - c ""' O"

u6 pi - 1
~ ai ud I'

and because the a-rank of C~;" is r, (6.22, 23) have as a consequence

(6.24)
As ~ depends here on the rt" only, we have o;. ()~ = 0 and
(6.25) "' ()"
u[r J.] = 1
~
callC
cb V[r ()bJ.] + B"Ak
ki [r
AiJ.]

where the last term could be added because A~ = 0. Now we have

already two parts of (6.20) viz. (6.14) and (6.25). The last part is [cf.
(6.17)]
(6.26) "' A"] -- - ~1 c"Ac
u[r cb [r AbPl

but this equation is equivalent to CARTAN's structural equations of the

parameter groups of the holonomy group (IV 1,19), already proved
in IV 1.
In the Xn+r of ~,rf the A~; h=1, ... ,n and O'JJ; a=1, ... ,i are
n + r covariant vector fields. They are linearly independent because
an equation of the form
(6.27) IXJoA~ + IXa{)'JJ = 0
would be equivalent to
a) IX~o~1 + 1Xa()~ = 0
(6.28) {
b) 1Xa0p=0
and these equations can only be true for oca = 0; oc11 = 0 because of
(6.17). Hence the n+r covariant vectors determine an anholonomic
coordinate system in the Xn+r and the intermediate components A~ are
A".
B A'JJdet -O'JJ; X=1, ... ,n; 1, ... ,r;
(6.29) { h=1, ... ,n; a=L ... ,r;
B = 1, ... ,n; i, ... ,f.
With this notation the equations (6.19, 20) of the symmetric An can
be written together
a) occ A~1 = t Cfy Afc A};1
(6.30) {
c b) [dAx] = ~Cfy [Az AY]; X, Y,Z = 1, ... , n, 1, ... , i
380 VII. Variations and deformations.

with constant coefficients Cjy

c~b def 0; a, b, c = 1, ... , f
(6.31) c~. def - c~.; h,i,j=1, ... ,n
,.
C ~. -def 0.

But as was proved in IV 2 this is only possible if

Now if these latter equations are written out making use of (6.31), we
see that they are in fact satisfied as a consequence of (III 5.4), (IV 1.28b),
(6.6,12) and VwR;;i''=O.
As we have seen in IV 2, (6.32) has as a consequence that (6.30)
are the structural equations after C-\.RTAN of an (n + r)-parameter group
of point transformations in Xn+r. The structural equations of the
same group after LIE are
(6.33) (AzAy) =- CjyAx

l
or, written out,
(A. Ab) = c~b A,.; Ab= A~ oJ< + A~ oy;
(6.34) (A.Ai) = - C~tA~o; A 1 = Af oJ' +Ar oy;
(A;At) = BJZ" A,. = - Cf,A,..

This proves that if a system of constants c~b C~;. B;;" is given, satisfy-
ing (6.31, 32) there always exists a symmetric An whose holonomy group
depends on r parameters and has the c~b for structural constants (cf.
IV 2). In this An there exists an anholonomic coordinate system (h)
suchthat r:,=-tA;c:. and Rifi"=C=Bi.;a. 1 )
As was proved in IV 2 there exists in Xn+r an (n+r)-parameter
group leaving the n + r covariant vector fields A ~ (with AZ = 0) and A ~
invariant. Hence, if .;x~~"+v" dt combined with rl'"~'YJa.+va.dt is an

l
infinitesimal transformation of this group, the LIE derivative of these
n + r fields must vanish:

a) vJ< oJ< A1 + v'~' 8y A1 + A! 8;. vJ< + (A~ 8;. vY) =0

b) (vJ< oJ< AZ) + (v'~' ay AZ) + A! Op vJ< + (A~ Op v'~') = 0
(6.3 5)
c) vJ< oJ< A1 + (v'~' 8" A1) + A; 8;. vJ< + A~ 8;. v'~' = 0
d) (vJ< oJ< A) + v'~' o" A+ A; 8p vJ< + A~ op v'~' = 0

1) This theorem was proved by E. CARTAN in 1926, 1, p. 264ff. for the case
of a Vn-
 VIII. Miscellaneous examples. - 1. The harmonic Vn. 381

(the terms in brackets vanish already because 1 . A~ = 0; 2. the A

are independent of the ~ and 3 . the A~ are independent of the rt).
From (6.35b) it follows that the v" are functions of the ~" only. Hence
v" dt represents an ordinary infinitesimal transformation of the Xn.
With this transformation the fields ~;. are dragged along and accordingly
the changes of these fields are at every point of X,. equal to their nega-
tive LIE differentials. Apart from this transformation, va. dt represents
an extra infinitesimal transformation of these local vectors ~;. and
because in total the local fields ~;. must be invariant, these two infinitesi-
mal transformations must annihilate each other. This is just what the
equation (6.35a) expresses. Moreover the A~ are invariant, hence the
r~ = -l A~ C!, are invariant, and this means that the connexion
in An is invariant. This proves that there exists in a symmetric A,. an
+
(n r)-parameter group of point transformations, which leaves the
connexion invariant, and that this group has the - Cfy for structural
constants.l) If from this group we take the subgroup that leaves in-
variant a point ~. this subgroup transforms the vectors in the tangent En
0 '
of ~ in the same way as the holonomy group does.
0

VIII. Miscellaneous examples.

1. The harmonic JT,..
Let x" be rectilinear coordinates in an ordinary Rn. Then the
distance s of an arbitrary point x" from the origin is
(1.1)
and by differentiation we get
(1.2) V" sP = p sP-Z g1._ x"'
(1.3) V"V"sP=p(n+P-2)sP-2; p<!e!g""'V"'.

Hence v v.
s is a function of s only, and s2 -n is for n =l= 2 a Solution
of the equation of LAPLACE
(1.4)
In a V,. (1.4) is an invariant equation but we can not be certain
that solutions always exist. The V.. is called harmonic 3 ) at the point $"
0
1) This was proved by E. CARTAN 1926, 1 for the case of a V.,.
2 ) From now we write in this section da 9 instead of ds 9 for the linear element.
3 ) CoPS ON and RusE 1940, 1, p. 11 7. Cf. for the affine generalization RusE

19 52, 1 ; PATTERSON 19 52, 1.

382 VIII. Miscellaneous examples.

if (1.4} has at least one solution q;(s) which is a function of s only. s is

here the distance of the point ~" with respect to the point ~. If this
solution exists we have 0

{
0 = vv.q; (s) = q;' v V.. s + q;" (17s) V.. s;
(1.5} m' def !!.!P_ , m" def d 2 qJ
T ds T - ds 2
hence
(1.6}

and this is a function of s only. Conversely, if it is known that 1717. s

is a function of s only:
(1.7) 1717. s = X (s)
and if we write V (s) for a still unknown solution of (1.4}, we have

(1.8) 17 V,. V= V" (17s) V,.s +V' x(s) = 0

from which V can be found by quadratures as a function of s. Hence
A Vn is harmonic at a point ~ if and only if the distance s with respect
0
to ~ satisfies an equation of the form (1.7). 1)
0
It is convenient to introduce the function .Q def~ s2 2) and to write

(1.9) f (.Q) det v V,. .Q = s x (s) + 1 = (2.Q)l x {(2.Q)l} + 1.

According to COPSON and RusE a vn that is harmonic at all points
of the region considered is called completely harmonic. If it is harmonic
at one point only, the term centrally harmonic will be used. In a com-
pletely harmonic vn the functions X (s) and I (.Q) are each the same for
all points, as was proved by LrcHNERowrcz. 3 )
RusE 4) has proved that an elementary solution of (1.4} is given by
the integral
s
(1.10} q;(s) = A f e- 1 s1 -" ds + B
a

where A, B and a > 0 are constants, s is the distance from ~" to ~ and
0
where the integration has to be effected along the geodesie through
1) CopsaN and RusE 1940, 1, p. 121.
2) Q is a function of ~" and ~", and is often called the characteristic function
0
of the V,.. Its covariant derivatives for fixed ~" were derived by SvNGE 1930, 1.
0
Cf. CoPsaN and RusE 1940, 1; LrcHNERowrcz 1944, 1, p. 158.
3 ) LICHNEROWICZ 1944, 1, p. 150; cf. WALKER 1945, 1.
4) RusE 1930, 2; 3; 1939, 1 for S,.; cf. CoPsaN and RusE 1940, 1.
 1. The harmonic V,. . 383

~x and ~x. e is RuSE 1 S invariant

0

(1.11) /) def
"'~
g~ (~X) g~ (~)
o
I Det ( az Q
a,;~'c,;A
) .
I
0

it is a function of s onlyl if and only if the V,. is centrally harmonic at ~.

0
Its development in a power series of s and its geometric interpretation
were given by WALKER. 1) The solution (1.10) behaves like s2 -" at ~
0
and on all null geodesics through this pointl and it passes into s2 -" if
the Vn is an R,.. 2)
A completely harmonic V,. with e = 1 is called simply harmonic. 3)

WALKER proved that this condition is equivalent to X (s) = (n -1) s-1 4)

If the V,. is centrally harmonic at ~ and if the geodesie joining ~x

0
and ~ has a unit tangent vector at ~ with components t"'; h' = 1' I ... In'
0 0
with respect to some fixed local orthogonal coordinate system (h') in
the tangent R,. of ~x~ we may introduce the coordinates
0

(1.12) ~a def ta'; ~n' def s; a' = 1'1 ' ' ' I (n -1)'.

Then the linear element takes the form

(1.13)

and for x(s) we get

(h') (h')
+ g-! ai' gt ai' s
1
X (s) = V~' V~' s -*- (8,., {"') 8;, s
(1.14) (h') (h') (h')
= g-an'
0
~ 9" 8ll, s 1
= 1 0 1og
~-
2 05
g I

from which by integration

(1.15)
(h')
tlog g -*-I x(s) ds +iPW'); a' = 1'~ ... 1 (n- 1)'.

Instead of (h') we now introduce the coordinate system (h); h = 1 I I nl

defined by
(1.16) ~a def 6~, s ta'; ~ll <!ef s tll'; a = 1> ' ' ' In - 1

1) WALKER 1942, 1, p. 16.

2 ) LrcHNERowrcz 1944, 1, p. 148 f. Mean value theorems for harmonic func-

tions in a completely harmonic Vn were dealt with by WILLMORE 1950, 1. The

more general equation 17v V.V=F(V) was discussed by NoRDON 1944, 1.
3) WALKER 1942, 1; cf. PATTERSON 1951, 2; 1952, 1.
4) WALKER 1945, 1.
384 VIII. Miscellaneous examples.

where F' is a known function of the ta'. Then

(1.17) log Det (AZ.) ~- (n - 1) log s +log F' W'')

and
(1.18) ! log(g) * J X (s) ds- (n- 1) log s + (1) c~a') -log F' W').

But, for s = 0, ({i is independent of the ~a' and has the value 1. More- +
over in III 7 we have seen that the transformation of two normal
coordinate systems with respect to the same point is linear homo-
geneous with constant coefficients. Hence
(h)
In a centrally harmonic Vn with centrum ~ the components g of g
0
with respect to any arbitrary normal coordinate system (h) belanging to
~ is a function of s only and satisfies the equation
0

(1.19)

From (1.9) and (1.19) we get

(A)

(1.20) vV.Q * (n-1)+ ~s dl~;g +(Vs)V.s2).

According to (III 3 .19) it follows from (1.19) that for a positive definite

I
fundamental tensor

g =* (X(s)- -n-
Eiii = 21 oilog (h) 5-
1 ) o;s
(1.21)
= sx(s)-n+l () . .Q= f(Q)-n () . .Q3).
52 1 2 Q 1

From the series (III 7.28) for lf1 we get for lj~

(1.22)

Now in (1.21), f(.Q) can be developed in the series

(1.23) f(.Q) =f(O) +!Jf'(O) + ~.Q 2 f"(O) + ... ; f(O) =n.

1 ) LICHNEROWICZ 1944, 1, p. 155; the converse is also true. The proof given
there is not quite satisfactory because the result seems to depend on the choice
of the normal system, which in fact it does not.
2) LICHNEROWICZ 1944, 1, p. 150.
3) LICHNEROWICZ 1944, 1, p. 156.
 1. The harmonic Vn. 385

Making use of the following identities in normal coordinates

(1.24)

(1.25)
we get by equating corresponding terms in the development of both
sides of (1.21)

(1.26)

known as the equations of CoPsON and RusE.l)

Using the relations (III 7.36, 40; Exerc. III 7,5) between the normal
tensors and the curvature tensor and its covariant derivatives, and
the identities III (5.27, 30), the equations (1.26a, b) appear tobe equi-
valent to

(1.27) l a)
b)
tK;;;!(~) +tK;;:/(~)
0
r:(wKvp) (~) = 0;
0
0
= -!K.I'(~) =!f'(O)g.l' (~)
0 0

hence (cf. III 5)

(1.28) f' (0) = - __!__ K ( ~) = - ~ (n - 1) u ( ~) .
3n o 3 o

If the vn is completely harmonic, (1.27a, b) hold for every choice

of !;". Then (1.27a) expresses that the V,. is an EINSTEIN space. 2)
0
(1.27b) is a consequence of (1.27a) for n>2 and expresses for n=2
that the v2 is an 52. As every EINSTEIN Va is an Sa this proves that a
completely harmonic vn is always an Sn if n = 2 or 3. 2 ) Of Course
every Sn is completely harmonic because for every choice of !;", J7~' 171' s
0
is a function of s only. In the beginning of the investigations it seemed
1 ) CoPsaN and RusE 1940, 1, p. 130; cf. LrcHNERowrcz 1944, 1, p. 156; RusE
194 5. 1. CoPsaN and RusE went via the covariant derivatives of Q in order to
make nse of SYNGE's relations in Synge 1930, 1.
2 ) CoPSON and RusE 1940, 1, p. 132. In the same paper it was proved that
every ScHURspace [cf. V, p. 244, footnote 2)] is centrally harmonic with respect to
the origin but that for n >2 not every centrally harmonic space is a ScHUR space.
Schonten, Ricci-Calculus, 2. Auf!. 25
386 VIIL Miscellaneous examples.

quite probable that every completely harmonic V,. should be an S,..

Every conformally euclidean EINSTEINspace is for n >2 anS,. (cf. VI 6)
hence every completely harmonic space that is conformally euclidean is
an S,.. But as WALKER proved by a counterexample 1) a space can for
n >3 be completely harmonic without being an S,..
(1.26b) contains derivatives of the components of the curvature
tensor. But for n = 2 it follows by differentiation of (1.26a) that (1.26b)
is identically satisfied. In an analogous way the second derivatives of
6
the curvature tensor that occur in (1.26c), after expressing N in terms
4 6
of K and K (cf. III 7), can be eliminated after differentiating the
foregoing formulae of (1.26). Then we get after some calculation 2)

(1.29)

LICHNEROWICZ and WALKER 3) used this equation to prove that a

completely harmonic V,. is always an S,. if its index (cf. I 8) is 1 or
n -1. We give here a survey of results:
A completely harmonic V,. is an S,. if one of the following conditions
is satisfied
a) n = 2 or 3;
b) the V,. is conformally euclidean;
c) the index is 1 or n -1.
lf the V,. is simply harmonic it is an R,. if either one of these three
conditions is satisfied or the condition 4 )
d) the index is 0 or n. 5)

For any n:;::;:; 4 there is a completely (simply) harmonic V,. which is

not an S,. (R,.), with any index not satisfying c (c or d). 6)

1) WALKER 1945, 1; 2.
2) LICHNEROWICZ 1944, 1, p. 159. RUSE 1945, 1, p. 156, gave the geometric
interpretation of (1.29) for n = 4 and he determined all algebraically possible v,s
satisfying (1.27) and (1.29). Cf. LrcHNEROWICZ 1953, 2 also for literature on the
harmonic V,..
8) WALKER 1942, 1 for n=4; LICHNEROWICZ and WALKER 1945, 1 for the
general case.
') WALKER 1945, 1.
6) This is in accordance with T. Y. THOMAS and TITT 1939, 4, who proved that
for the definite case s8-n for n >2 and log s for n = 2 can only be solutions of LA-
PLACE's equation if the V,. is an R,.. Cf. WALKER 1942, 1, p. 26; LICHNEROWICZ
1944, 1, p. 160.
8) WALKER 1945, 1; 2.
 t. The harmonic V". 387

WALKER gave the following examples.l) The v4 with the linear

element with index 2

{ da = cw ae- ~ d~ ) + 2de dt;J + 2de ae;

2 1 2 2
(1.30)
C = const. =l= 0

is simply harmonic. It has a curvature tensor with the only non-zero

component K 1212 . Hence K".,~=O, that means that the V4 is a special
EINSTEIN space. But it can not be an 5 4 because an 5 4 with K".A = 0
would be an R 4 The V4 is symmetric because 17., K."..~ .. = 0.

l
The V4 with the linear element with index 2
da2 = 21-z cae
ae + dt;J d; 4 ) +
(1.31) + 2CI-zwae- e ae) wae- ;z dt;J)
I = 1 + c W;z + t;J e); I > o; c = const.
is completely but not simply harmonic. The curvature tensor satisfies
the equation

(1.32) { K."..~":
5."..~ .. -
=: g[.r.< g~"J_+ 5.".,~"~
5"..;."- 5.". .... - 5,~"."..
.
where 5.".,~" is a tensor with the only non-zero components

The V4 is symmetric but no 5 4 Aspace with the same properties but

adefinite fundamental tensor can be derived from (1.31) by a non-real
coordinate transformation.
In a later paper 2 ) WALKER proved that a completely harmonic v4
is always symmetric if its fundamental tensor is definite.
In the same paper he proved that the equations of a completely
harmonic v4 with a fundamental tensor of index two, only containing
g,~,., K.".,~" and llw K.".;." [viz. the equations of CoPSON and RusE
(1.27, 29)] admit solutions that do neither satisfy llwKv"..<.. = 0 nor
VwK.".;.,.= kwK.".,~"; kw =l= 0. But, as WALKER remarks on p. 22 this
does not yet prove that there exists a harmonic V,. that is neither sym-
metric nor of recurrent curvature.
LICHNEROWICZ 3) proved from (1.27) and (1.29) that in a completely
harmonic V.. the following inequality is valid
(1.34) f' 2 ( 0) ;S; - 2 0 (n - 1) I" (0)
1) WALKER 1945. 1; 2; 1946, 1; cf. RusE 1946,3.
2) WALKER 1949, 2.
3) LICHNEROWICZ 1944, 1, p. 161 ff; 2.

25*
388 VIII. Miscellaneous examples.

and that the sign = holds if and only if the V" is an 5,.. Moreover he
proved in the same paper that a completely harmonic V,. is an R,. if
there exists a number k >1 such that
(1.35) f' 2 (0) '2.- 20k (n- 1) f" (0).

This is a generalization of the result of THOMAS and TITT mentioned

before. The proof of (1.34, 35) is not difficult but rather elaborate.
In 193 8 1) FIALKOW proved that an EINSTEIN V,. in an 5,. +1 with scalar
curvature x is either an 5,. or a product (cf. V 11) of an 5p and an 5,._p
1y. L ICHNERO-
.
wlth scalar curvatures -n-2- x an d n-2 x respectlve
p-1 n-p-1
WICZ proved that a non-flat completely harmonic V,. can never be a
product space and that according to the theorem of FIALKOW every
completely harmonic V,. in 5,.+1 must be an 5,.. 2 )
Global properties of harmonic spaces were discussed by WILLMORE. 3 )
RENAUDIE 4) dealt with properties of the group of isotr~py (cf. VII 5)
in harmonic spaces.
Exercises.
VIII 1,1 5 ). Every completely harmonic V"_ 1 in 5,. is an 5"_ 1 .
VIII 1,2 6). The V4 with the linear element
da 2 = c wae- ~ 1 ae) 2 + 2de ae + 2d~ 2 ae; c = const. =l= 0
is simply harmonic.
VIII 1,3 6). In a Centrally harmonic V", log e is written as apower
series in s:
VHI 1,3 oc) 1og n
0:
= -2!1 w1 s + -3!1 w2 s2 + .
Prove that
VIII 1,3 ) W2 = 1
0
K pi.X"' X ~,
s

2. Connexions for hybrid quantities.

In I 10 we defined hybrid quantities in E". In order to consider
fields of such quantities in X,. it is necessary to make an agreement
about functional dependency. Ordinary quantities were called analytic
1) FrALKOW 1938, 2, p. 782 also for literature.
2) LICHNEROWICZ 1944, 1, p. 163ff; 1944, J.
3) WILLMORE 1953, 1.
4) RENAUDIE 1954, 1.
6) LICHNEROWICZ 1944, J.
6) WALKER 1945, 1.
 2. Connexions for hybrid quantities. 389

if their components with respect to some allowable coordinate system (x)

were analytic functions of the ~" (cf. II 2). It seems convenient to
call a quantity of the second kind (cf. I 10) analytic if its components
with respect to (u) are analytic functions of the ~x. Then a quantity
of the first or second kind is analytic if its complex conjugate is analytic.
Accordingly a scalar is called analytic if it is an analytic function either
of the ~" or of the ~x. N ow it would be impossible to define analyticity
of hybrid quantities in the same way, because such a definition would
not be invariant for allowable coordinate transformations. Therefore
we introduce for all fields the term semi-analytic. A field is called semi-
analytic if its components with respect to some allowable coordinate
system (u) are analytic in the 2n variables ~" and ~x .1) Only in this
case they are also analytic in the 2n variables 'fJ", !;" defined by ~" =
+
'YJ" i !;"; ~x = 'YJ"- i !;". Obviously the definition of semi-analyticity has
the desired invariance. In the following all fields are supposed to be
semi-analytic (and perhaps but not necessarily analytic in ~" or in ~x)
and oil stands for ojo~il. Note that the ~ always remain analytic func-
tions of the ~. 2)
In order to establish a linear connexion 3 ) for all quantities, hybrid
quantities included, we start with the four conditions C I, II, III, V
of Ch. III 2. But instead of C IV we take the condition
C IV' The covariant differential of a quantity lJ' (indices suppressed)
is linear homogeneaus in d~" and d~x:

(2.1)

and as a sixth condition we introduce

C VI The covariant differentials of complex confugate quantities are
complex confugate.
From these conditions we get in the same way as in III 2 the most
general form of the covariant differentials of contra- and covariant
vectors of both kinds

(2.2) {
a) ~v" = dv" + J;.J. if d~"' + Jfi";, vJ. d~il
b) ~w;, = dw;,- r;J. w,.d~"'- JfiJ. w,.d;il

(2.3) {
+ +
a) ~vx = dvx Jfi1, vr. d~il J7.1 vr. d~"'
b) ~wr. = dwx.- Jfi\wxd~il- J;.\wxd~"'

1) ScHOUTEN 1929, 1.
2) GHOSH 1950, 1 considers the more general case where the ~><' are semi-
analytic.
3) ScHOUTEN and v. DANTZIG 1929, 5; 1930, 1; ScHOUTEN 1929, 1.
390 VIII. Miscellaneous examples.

from which the covariant differentials of all other quantities can be

derived. In these formulae I;."1. and lj."1. are arbitrary but semi-analytic
and Ip1 and I;.1 are their complex conjugates. In the following we
often suppress formulae that are the complex conjugates of other for-
mulae already written out and write "conj." instead. The parameters
of the connexion transform in the following way

conj.
(2.4}
conj.

from which we see that the I;."J. transform in the ordinary way (cf.
III 2.5) but that the lp"1. are components of a hybrid tensor.
It is sometimes convenient to introduce an auxiliary X 2 ,. whose
coordinates are the ~ and ~x considered as 2n independent variables
that can each take all complex values. Of course the 'YJ", ~ could also
be used as (non allowable) Coordinates. In this x21l not all analytic
transformations of the ~. ~x are allowable but only those that can be
written in the form
(2.5} ~,., =f"'W); conj.

But this implies that the X 2 ,. contains two invariant sets of oo" X,.'s
with the equations ~x=const. and ~"=const. respectively. 1) In the X,.'s
of the first (second) invariant set the ~", (~x) can be used as coordinates.
Through every point of the X 2 ,. there goes one X,. of the first and one X,.
of the second set and two X,.'s of different sets intersect exactly in one
point. From this it follows that the points of each X,. of one set are
in one to one correspondence with the points of every other X,. of the
same set and that corresponding points lie in the same X" of the other
set. A figurein one X,. and the corresponding figurein another X" of the
same set are said to be equipollent.
The tangent space of the X 2 " is an E 2 ,. andin this E 2 ,. the tangent
n-directions of the two invariant X,.'s through the point considered
form with their parallels the two sets of invariant E,.'s defined in I 10.
A general contravariant vector in X 2 ,. has 2n components v" and vx
and here the vx are in general not the complex conjugates of the v".
The n-dimensional subspace of the X 2 ., with the equations
(2.6)
is invariant and is called the principal x .. of the x2 ... It may be con-
sidered as the image of the original X,.. If the vector v", vx belongs to
1) In ScHOUTEN and v. DANTZIG 1930, 1, p. 327 we called these X., 's "iso-
tropic". Here the word isotropic has another meaning, cf. I 9 and VIII 3.
 2. Connexions for hybrid quantities. 391

a point of the principal X" it lies in this X" if and only if vx = v". In
this way the original xn appears as a subspace of the auxiliary x2"
and the quantities of the original Xn, hybrid quantities included, are
those quantities of the X 2 ,. at points of the principal X" that satisfy
the condition, that dropping a bar where there is one and inserting a
bar where there is none changes every component into its complex
conjugate.
Apart from the connexion in the original X" defined in (2.2, 3) we
now consider the most general linear connexion in the X 2 ". This con-
nexion has the form (cf. III 2)
a) (Jv'' = + CJ;.";. v). +I;.";; vX) d~" + (J;). v). + J;"x. vl.) d~P.
dv"
C2.7) {
b) bv;c = dv;c + CJ;.x;. v). + J;.;c;;vJ.) d~" +CI;~ v;. + 1//'x.vJ.) d~P.

where the v", v;c are 2n arbitrary components of a generalcontravarian t

vector of x2 .. and where the rare 8n3 independentparame tersnot yet
in any way connected with each other or with theF in C2.2, 3). Because
the two sets of Xn's are invariant we may now first introduce the in-
variant condition that their tangent n-directions are parallel for every
parallel displacement in X 2 ". This gives immediately
C2.8)
The geometrical meaning is that the Xn's of either set are geodesie
and parallel. Secondly we introduce the invariant condition that the
principal Xn is geodesic. Because for every displacement in this X"
we have d~;c=d~". the n.a.s. conditions are

(2.9) { a) J.;:_x = I;.".. ;.l for ~x = ~".

b) I;.x =I;).
A

Hence the linear connexion in X 2 n for which the X"'s of either invariant
set are geodesie and parallel and for which the principal X,. is geodesic,
is at the same time a connexion in the original X" of the kind (2.2, 3), that
is a connexion satisfying the conditions CI, C II, C III, C IV', C V and
C VI. 1)
The covariant derivatives belanging to this connexion are

C2.10) { a)
b)
17" v" = 8" v" + J;.";. v;.;
V"w" = 8"w" -r;;.w,.;
conj.
conj.
17-p v" = +
(2.11) { a)
b)
8-p v"
Vp. w). = 8p. w).-
E"_.
p
v;. '
J;"J. w,.;
conj.
conj.
1) Cf. footnote 3) on p. 389.
392 VIII. Miscellaneous examples.

They can be looked upon as parts of the covariant derivatives of the

vectors v", vx and w,., w;; in the a uxiliary X 2 ,. .
The curvature tensors belonging to the connexion (2.2, 3) may be
considered as the only non-vanishing parts of the curvature tensor in

l
X 2 n. There are eight of them:

a) R;~;," = 2o[vJ;,];. + zi[:Iel J;,);.; conj.

b) R~ ~;, " = ov I;,";. - o I';;~ + I';;~ J;,e;. - J;,"e I';;). ;
I' con j.
(2.12)
c) R;(ij"=- R(i;;."; conj. (isomer of (b))
d) R~p,;," = 2orv.l/J];. + 2I[;Iel .l/J];.; conj.

the other eight, R;~/', R~~i.x, R;;;.x, R~[ii.x, conj. being zero. The tensors
of asymmetry can in the same way be looked upon as the only non-

l
vanishing parts of the tensor of asymmetry in X 2 n. There are six of them:

a) S~i."=l[:;.l; conj.
(2.13) b) Sp,;."= }z.l/J";.; conj.
c) S~i."=-Si~"; conj. (isomer of (b))

the other two, S~;,x, conj. being zero.

A contravariant vector of an Xn of the first invariant set is charac-
terized by v"=l= 0; vx = 0. If this vector is displaced parallel in a direc-
tion lying in an Xn of the second invariant set, we have d~" = 0, d~x =I= 0,
hence
(2.14)

But this proves that the transformation is an equipollence for every

choice of v", if and only if /fi";. = 0. Hence, if we introduce as a third
invariant condition that the parallel displacement of a vector of an xn
of one invariant set in a direction lying in an Xn of the other set is
always an equipollence, this leads to the equations

(2.15)

Because .l/J";. and ~x;; are hybrid tensors as we have seen, the condition
(2.15) is also invariant in the original Xn. Another geometrical meaning
of (2.15) is that it is the n. a. s. condition for the existence of infinitesimal
parallelograms in X 2 n with two parallel sides in an arbitrary direction
of an Xn of one invariant set and the other two in an arbitrary direction
of an Xn of the other set. This is quite clear because I;.~ and J'x.~ are
 2. Connexions for hybrid quantities. 393

already zero according to (2.8) and (2.15) is therefore equivalent to

1

the vanishing of the components FrftA], F~X] of the tensor of asymmetry

in X 2 ,.. So (2.15) is in fact a kind of condition of symmetry of the
connexion in X 2 ,. but only halfway.
If the third condition is also satisfied we have 1 )

(2.16) {
a) 1711 v" = o11 v" + I;."A v"; 17.a v" = o.a v"; conj.
b) 1711 WJ. = 011 WA- I;_."J. w,.; 17.a WJ. = Op WA; conj.

and this connexion differs from an ordinary connexion in L,. as dealt

with in III 2 only by its special equations for fields that are only
semi-analytic.
From (2.16) it can be derived in the ordinary way via the n-vectors
that for a density with the weights w and W there is for 1711 a term 1

- w r,:..and for 17.a a term with - w' IJ,..

An X,. with the connexion
(2.16) is called an L,. . I t is the most generallinear connexion for hybrid
quantities that satisfies allinvariant conditions andin this way the r ..
is analogous to the L,. in III 2. The auxiliary X 2 ,. of an L,. is an L 2 ,..
In an L,. only the curvature tensors R;~;."; conj. R;~;."; conj. and 1

R;p"A_"; conj. remain and (2.12b) takes the simpler form

(2.17)

From the tensors of asymmetry only 5;,;."; conj. remain.

In an L,. the quantities R;~;." and 5~;." satisfy theidentities (III 5.1)~
(III 5.2) and (III 5.19) and similar identities hold for their conjugates.
The quantities R;~;."~ R;~;."~ R;p;."~ 5~;." and their conjugates satisfy
the following identities that can be easily derived from (III 5.1 2 19} 1 1

if these equations are applied to the x2,.:

(2.18) R;~"A" = - R;~;."; conj. (first identity cf. III 5.1)

1

R V. [p....
A] =
17
I' ii 5p A" = Up
" 5p A"I con1
(2.19) {
(second identity cf. III 5.2) 1

a) 2 17
l'[w R:"+o-R""=25eR:""
JpA I' wvA a>v pQA conJ. 1

(2.20) { b) 17 R "- 5 Q R " conJ

"[w vJpA - u;;; I'QA 1

(identity of BIANCHI, cf. III 5.19)

1) SCHOUTEN 1929, 1, p. 460.

394 VIII. Miscellaneous examples.

The RICCI tensor R and the bivector V in L 2 ,. have the non vanishing
components
a) R.
!'A
=R+R~v
V!' V!'
=R
V!' 1 conj.
b) R;;;. = R~~;.'"' + R[i~;.il = R~~;.'"'; conj.
(2.21) c) conj.
d) conj.;

butfrom (2.18) it does not follow that v;'"' = -V~;;, because v;;; is according
to our definitions the complex conjugate of v;.
and thus equal to R~~l"
and not equal to R~~;.;.. From (2.19) we get the equation

(2.22)

which will be used hereafter. From (2.21) we see that V;;'"' (and not v;'"')
is always a hybrid bivector and that v;l' is hermitian if and only if
V;;'"'=O. From (2.20) we derive by contraction over ~A

conj.
(2.23)

from which, by means of (2.21 d)

(2.24)

in accordance with (III 5.22). By contraction over ~v and ~# we get

from (2.20a) and (2.20b)

(2.25) VwR1u + V.R[i,;,;."- 8;;w;. = 2S;,;e Rfi,;;.; conj.

(2.26) V[;;Ra;J;. = - S~{/J Rii;.; conj.

that could also be derived from (III 5.25). (2.23 b) and (2.26) are m
accordance with (2.22).
An I,. with a symmetric connexion is called an X,.. It is analogous
to the A,. of III 2. The auxiliary X 2 ,. of an X,. is an A 2 ,.. In an X,.
we have instead of (2.19, 20, 22, 23, 24, 25, 26)

(2.27) R~[~;.j = 0; conj.

(2.28) {
a) 2Vj-wR;Jp)." + 8pR,;,;;." = 0; conj.
b) V[a;R~l~;." = 0; conj.
 3. Unitary connexions. 395

(2.29) R;n. = -V~,_= - o;;Ff~'; conj.

(2.30) { a) 211["' Vjp 11- o V"'.=- 2Ji[wRitzll + 2otzR[wl = 0; conj.

b) Vrw v;1 ~' = - J?iwRvJI' = 0; conj.
(2.31) 2Ji[u;V;;11' + ol' Vu;;;= 0; conj.
(2.32) VwRtzA + V.R,i.;,t- otzRwA = 0; conj.
(2.33) Ji'[;;RtzJ'- = 0; conj.
Of course the equations (III 5.2-4) remain valid, hence we have in
an l,.:
(2.34) Rr;;;3' = 211[. S,_;.i]''- 4Sr;; 11 Sj.1 ~ "; conj.,
in a semi-symmetric r. :
(2.35) Rr;;r=2A[.V,.S,.1; conj.,
andin an .A,.:
(2.36) Rr;;1t = 0; conj.

Exercises.
VIII 2,1 1 ). Let v" be a semi-analytic field and vx its complex con-
jugate. This field is analytic if and only if the field v", 0 (0, vx) in Xu
is constant over each X,. of the second (first) invariant set.
A semi-analytic scalar field is analytic in ~" (r) if and only if it is
constant in every X,. of the second {first) invariant set. It is the product
of a scalar analytic in ~" with a scalar analytic in ~x if and only if
VIII 2,1 oc)
VIII 2,2 2). An analytic transformation of X 2 ,. is an analytic trans-
formation of X,. if and only if it leaves invariant both invariant sets of
X,.'s and the principal X,..

3. Unitary connexions.
If a hermitian tensor field aAx of rank n is given in X,. we may ask
for a connexion of the kind (2.2, 3) for which
(3.1)
or
(3.2)
1) ScHOUTEN and v. DANTZIG 1930, 1, p. 328 f.; E II 1938, 2, p. 236.
1) ScHaUTEN and v. DANTZIG 1930, 1, p. 329.
396 VIII. Miscellaneous examples.

From (3.2) we get immediately

(3-3) conj.
which equation expresses J;.";. in terms of a;.;c and its first derivatives
and a hybrid tensor J;.!J that could be chosen arbitrarily and that is
zero if we impose the condition that the X,. be an I,.. Then we getl)

(3.4) {
a) I;.: = a~" 8" a;.!J; conj.
b)
I,~.=O, conj.
Hence:
A hermitian tensor field a;,;c of rank n in an X,. determines one and
only one L,.-connexion for which a;.;c is covariant constant.
Such a connexion is called unitary and an X,. with a unitary con-
nexion is called a ft. 2) a;.;c is called the fundamental tensor of the ft.
It is analogaus to the U,. of III 3.
In the auxiliary X 2 ,. of ~" and ;x we now use indices A, B, ... taking
the values 1, ... , n, 1, ... , n. Then the a;.;c are the components of a sym-
metric tensor aBA with a)." = 0, a;.;c = 0. Hence the invariant X,.'s are
isotropic (cf. I 9) with respect to this fundamental tensor. The most
general linear connexion in x2n leaving aBA invariant is (cf. III 3-5)

(3.5) FcAB = i aAD (cb avB +oB avc- ov acB) + SciJA- S~c- Sc~B

where SciJA is an arbitrary tensor alternating in C B, or

(3.6)

and four expressions for I1, J;.x;., l~ and r;.x,~.. Now these latter four
expressions must be the complex conjugates of the expressions (3.6) in
consequence of the condition that the principal X,. in the X 2 ,. be geodesic.
But this is possible if and only if Sp,:/, S~i.;c and S~;.;c are the complex
conjugates of 5~;.", Sp,t and Sp,i," respectively. If now we introduce
the condition (2.8) expressing that the invariant X,.'s of either set are
geodesie and parallel and the condition (2.15) for the equipollence of
1) Cf. footnote 3) on p. 389.
1) In 1~0, 3 we wrote U,. a!;d i~ E]I !_?3~ 2 we !:sed the term unitary K,.

instead of Un. The notations Ln, An, U,., V,., S,. and C,. used now are in strict
analogy to the L,., A,., U,., V,., Sn and Cn of Ch. III and VI.
 3. Unitary connexions. 397

parallel invariant X,.'s we get

(3.7) sx=
pJ.
o' Sp.i"= 0; conj.
and
(3.8) S;;z;;;. = 8c;;z;;J;.; conj.
or
(3.9) s"-
pJ. - aQ" 8[p aA]iJ. conj.
hence
(3.10) { a) I;";. = aiJ" 8,_. a;.iJ; conj.
b) I';;z";.=O; conj.
in accordance with (3.4). This proves that the connexion in a 0,. can
be derived as a connexion in X 2 ,. with the symmetric fundamental
tensor aBA with a;.,.=O, ax;;=O, satisfying instead of the condition of
symmetry other conditions concerning the invariant X,.'s. Hence the
auxiliary x2 .. of a v.is a u2 ... 1) In general the connexion of V,. is
not symmetric. If it is symmetric we call the fJ,. a V,.. The V,, is anal-
ogaus to the V.. of III 3 and its auxiliary X 2 ,. is a V2 ,.. We prove that
5;.;." = 0; conj. if and only if there exists a semi-analytic scalar field cp
such that
(3.11)
Wehave only to prove that (3.11) holds if 5;.;."= 0; conj. because the
inverse is trivial. 3) In order to prove this we remark that the equation
(3.12)
with the unknowns cp,. is totally integrable because of (3 .2) and S;i x = 0;
conj. if the /;" are considered as parameters in (3.12). Hence, if rp,. is a
solution of (3.12) the general solution has the form
(3.13)
1) Cf. RozENFEL'D 1949, 1.
2) SCHOUTEN and V. DANTZIG 1930, 1 for a special case and KHLER 1933, 1
for the general case; cf. LEE 1942, 1; SASAKI 1949, 2; IWAMOTO 1950, 1, p. 125.
8) The symmetric unitary connexion or V,. appeared first in ScHOUTEN and
v. DANTZIG 1930, 1, p. 333. In 1933, 1 BERGMAN gave an application of this connexion
to the theory of functions of two complex variables by means of his "kerne! func-
tion ". This application has been dealt with since in many publications of BERG-
MAN and others. In 1933, 1 KHLER independently found the same connexion
and formulated the condition of symmetry by means of CARTAN's alternating
differential forms. A V,. is sometimes called "KHLER space" in the literature.
Cf. for instance EcKMANN and GuGGENHEIMER 1949, 1; 2. HoDGE 1951, 1; BocH-
NER 1953, 1; GUGGENHEIMER 1953, 1. There are a great number of publications
of these and other authors dealing with global properties of the r...
398 VIII. Miscellaneous examples.

where the f,. are arbitrary functions. This proves that cp is a solution of
(3.14)
The integrability conditions of (3.14) are
(3.15) o111 cp.;0 1 +or11 /;.1 =0; conj.
From (3.9, 12) and the vanishing of 5;.;." it follows that

(3.16)
and this proves that o111 ~Al depends on the 1;" only and that accordingly
/.; can always be chosen suchthat (3.15) is satisfied identically.
If cp is a solution of (3 .11), (jj is also a solution and this proves that
if (3.11) has solutions, there always exists a real solution.
In a V.. we have
(3.17) o[.J;j;. + l(:[a[ J;];. = 0[ af:i" op] a;.fJ + (af:i" 0[ a[afJ[) a'ia op] a).i = 0
hence
(3.18) R;~;." = 0; conj.
and the only remaining curvature tensors are Rp;.;_"; conj. and R;;i;,";
conj. As in Riemannian geometry it is convenient to lower the last
index of Rp;.;,". Then we get from (2.17) and (3.4a)
f a) R;; 11 ;.;c=o;;o11 a;.;c-af:ia(o;;a11 ;c)o11 a;.fJ; conj.
b) Ru=- o;; ax 11 o11 a;.;c
(3.19) = - o;;F/J. = - o;;o;.log a + (n -1) o;; 5.;; conj.
c) v;p = o;;J;.).;. = 0;; op log a =V~;;; a def IDet (a;.;c) I
d) V;; 11 = 0; coni.
The vanishing of V;; 11 is a consequence of the fact that the auxiliary X 2 ..
is a U2 ... But as we have already seen this vanishing implies that V; 11
is hermitian and this is in accordance with (3.19c).
The following identities hold in a fJ.. 1 )
(3.20) R;; 11 ;.;c=-R11 ;;;.;c; conj.
(first identity, cf. III 5.1; VIII 2.18)

(3.21) R;;r11 ;.1;c= 179 S11 .;;c; conj.

(second identity, cf. III 5.2; VIII 2.19)
) E li 1938, 2, p. 234-239.
 3. Unitary connexions. 399

(3.22) Rv".ut=- R;;".;c;.; conj.

(third identity, cf. III 5.13)

(3.23) RiipAx- RAxiip = 217ii S".u<- 2 ~ Ss<ii".; conj.

(fourth identity, cf. III 5.15 IV)

(3.24) li[wRv]iUx= S.;,;ll RpgAs<; conj.

(identity of BIANCHI, cf. III 5.19, VIII 2.20) 1)

In consequence of (3.22) the equations (2.20a, b) lead only to one

equation (3.24). The same holds for the equations (2.23 a, b) which lead
to one equation
(3.25)

In (2.25) the term - o11 R"'" drops out, (2.26) remains valid without
simplifications and (2.34, 35) take the form

(3.26)

for the general fJ,. and

(3.27)
for the semi-symmetric fJ,., n > 2. Hence in a semi-symmetric fJ,., n > 2, SA
always is a gradient vector.
In a V,. we have instead of (3.19b) [(cf. III 5.8) but consider the
difference between V;;). and v;_.]

(3.28) R;;_. = - o;; a_. log a =RH=- V;;~; conj.

!
hence Ru is now also hermitian. 2) The identities (3.20-25) take the
simpler form
a) R;;,.As<:- R,.;;.As<; c~nj. (1o id.; cf. III 5.15 I)
b) R;;,.J..s<-R;;J..pSi. COnJ. (2 id.; cf. III 5.151I)
(3.29)
c) R;;".As<=- R;;".u; conj. (3id.; cf. III 5.151II)
d) R;;".J..x=Ru;.;;,.; conj. (4id.; cf.III5.15IV)

(3.30) li[wR]iiJ..x = 0; conj.

(3.31) Ii("' Vjlzl ,.1= 0; conj.
1) HoMBU gave 1935, 1 a complete list of all identities containing first deriva-
tives of R and S.
2) SCHOUTRN and V. DANTZIG 1930, 1, p. 337; KHLER 1933, 1; FucHS 1937, 1.
 400 VIII. Miscellaneous examples,

T.. complex
coajugales ol
all equatiou I,. A,.
arevalid

Definition r;A=o; r:x=O; r:A=o S~)."=O;

- -
(2.12a) R'. - [v F"
vpA"-28 p]A+ 2F."
(viel Fp]A
11 see i,.

Corvature
(2.17) R:p;;."=a;;r;A see L,.
quaolilies
(2.21bl R;;A= -Ri~j,~-'= -a;;r;:A (2.29) R;;A= -o;;Ffl-'
(2.21e) v,'iip= R'iipA
A= aii r"pA see L,.

(2.21d) V;;~-'=v;~-'-v;;; see I,.

(2.34) R'''"
[vpi.] = 217,[v S"" [vp 11 S']p"
pi.] - 4S"
} (2.36) R[;~;_t=O
(2.35) R[;;;_{=2A(v ':S;.J for S~j,"=S[pA~l
(2.18) R~;i"= -R~pj," see I,.
ldeolilies (2.19) Ri[;it=o;;S~;_" (2.27) Ri[;)/=0
- -

- -
(2.22) R;;A+V;;.= (n-1) o;;SA (2.29) R;;A+v;A=O
I
(cf.III) 17,[ru R"
vp]i."=2S [wv ""
""R'p]<1 (cf. III) 17,[ru R'vp]
" -_ O
5.19
",.--- - 5.21

(2.20a) 2li(w R;1j, "+rp R,;,;;_ "=2 s,;,; 11 Rp~;_" (2.28a) 2li("'R;lj,"+ rpR.;,;;_"=O

(2.20b) 17,[w R'v]pJ.

""-S' ""
- wvCi R'pfiA (2.28b) li(wR~1 ~;_"=0

{ 2li(w Vl~ll-rp Vw=-2li(wRI,lv]+

(2.23a) 2li[w Vi~lv]-o,a vwv= -2 s.;,:;e v;e (2.30a)
+2rpR[wv]=O
Ideotilies with
R [w ii]p=- s.wv,fi v;'Ci!-'
dif!ereolialed (2.23b) 17, v;' (2.30b) li[w V~~-'=-li[wRvJp=O

(2.24) 21i(w V;;1 ~-'+o~-' Vw;;= -2S.i;~Ci viil-' (2.31) 2li(w Vii]p+VI-' Vw;;= 0

(2.25) VwR,aA+V,.RC:,.iv -rpRwA=2 s.;,; 11R~j,v (2.32) VwR,a,_+ V,.Rp~ -rpRAw= 0

(2.26) li(;;Rw]i.=-Sp,i;CiRCi;. (2.33) li(;;R.aJ,_=O

- -
- - I
 3. Unitary connexions. 401

iJn Vn

J:j,a;.;;=O; 5~"/'*-o Vl'a;.;;=O; 5~)."=0

(3.4a) r"I'A = aCi" 81' a;.ij see fjn

(3.18) R;~t=O see fjn

(3.i9a) R";;!';.
" = o;;aii"al'a;.!J see fjn

(3.19b) R;;;. = - 8;; 8;.log a + 28;;5).~~' (3.28) R;;;. = - 8;;8;.log a =R;.;;=- v;;.
(3.19c) v;l' = 8;; a~' log a = v;;; see fjn

(3.19d) V;; I'= 0 see an

(3.26) 2 r'(.5;).1"- 4 5[;/ 5).1 ~" =0 -

(3.27) r'(l' 5;.1=0 forn>2 if 5;(=5[1'A~l -

(3.20) R;;I';.;;=-RI';;;.;; (l.29a) R;;I';.;;=-RI';;;.;;

(3.21) R;;[!'J.]x= f-i;51';.;; (3.29b) R;;I';.;;=R;;;.I'x

(3.22) R;;I';.;;=-R;;I';;;. (3.29c) R;;I';.;;=-R;;I';;;.

(3.23) R;;I';.;;-R;.;;;;I' = 2 f-i; 51';.;;-2 ~ 5;;;;1' (3.29d) R;;l';.;; = R;.;;;;l'

see Ln see An

- -

} (3.24) r'(wRv]p;.;;= 5C:,; eR!leJ.x (3.30) r'(wRv]pJ.x=O

}~ v.[w v:'1111] =-
e v'
5 wv !le (3.31) Vcw Vj~ll = O

- -

VwRp;.+V,.RfiC:,,i V =2 5C:,;e Rfi;.i. 17wRp;. + V,.RfiC:,_i." = 0

see Ln see An
- (3.32) J:j,G;~'=O; c~'def R"~'-RA"fl

- R det R;;;.l;.
I
Schouten, Ricci-Calculus, 2. Auf!. 26
402 VIII. Miscellaneous examples.

In (2.32) the term - o;wJ. drops out but (2.33) remains valid
without simplifications. In a tJn a second contraction is possible. For
a V,. this leads in (2.32) to 0=0 but from (2.33) it follows that

which could be compared with (III 5.29).

On pp. 400, 401 there is a list of the most important formulae for
Ln, A,., tJ,. and V,..
lf 'lf" are n fields of unitvectors in a tJ,. that are mutually (unitary)
perpen'dicular, if the reciprocal set (cf. I 2) is .f'tJ. and if the funda-
mental tensor aJ.x is positive definite (cf. I 10), this tensor can be
written in the form
h h
(3-33) a.-=l:u,u-
Al< h X A

but the fields /lJ. need not to be analytic. In fact, it is not necessary
that there exist analytic fields .f'tJ. satisfying (3-33). We prove that such
fields exist if and only if
(3-34)
or in an equivalent form

(3-35)
(3-34) expresses the fact that the I',.";. are analytic in ~". The condition
is necessary. This follows immediately by differentiation of (3-33) if
it is supposed that the /lJ. are analytic. In order to prove that it is
also sufficient we consider the system of differential equations

a)
(3-36) {
b)

The integrability conditions are

(3-3 7) {
a) /J,T a;jj-r (o(v alewl) ae Oll] a). + .f'te (o(v ae) Oll] a). = 0
b) !J,e (o;; ae) oll a). +!le ae 0;; oll a). =0
(3-37a) is satisfied identically and (3.36b) is a consequence of (3-35).
Hence, if (3-35) is satisfied, (3.36) is totally integrable and this means
that for every choice of a vectort1J. at some point, there exists a covariant
constant field having at this point exactly the value JJ.. Thus, if we
1) ScHOUTEN and v. DANTZIG 1930, 1; cf. E II 1938, 2, p. 236.
 3. Unitary connexions. 403

take n mutually perpendicular unitvectors at some point we get n fields

of unitvectors that are mutually perpendicular at every point. The
equation (3.34) implies that every analytic field remains analytic if it
is displaced parallel. If we call such a connexion analytic, then (3.34)
is the n. a. s. condition for the connexion to be analytic.
From (2.17) and (3.34) it follows that the last curvature tensors
still remaining vanish and that the connexion is therefore integrable
(cf. 111 4). Gathering results we have 1):
In a Un the fotlowing three conditions are equivalent:
a) the connexion is integrable: R;p_;,"= 0; conj.
b) the connexion is analytic: 8;; ~,.A = 0; conj.
c) there exist sets of mutually perpendicular fields of unitvectors
that are analytic.
If the connexion is integrable and moreover symmetric, we have
from (3.9) and (3.36a)
(3.38)
and this means that the !1;, are analytic gradientvectors, depending on
the ~,. only. Hence in this case the scalars belanging to those fields
are analytic in~ and if they are taken as Coordinates, the linear element
takes the form
(3.39)
because the /1J. are mutually perpendicular unitvectors. From this form
every linear coordinate transformation with constant coefficients leads
to a form
(3.40)
Hence (cf. I 10) 2)
I I the connexion of a vn is analytic it is also integrable and the vn
is an Rn.
Exercises.
VIII 3,1. Prove the third identity in Un
VII13,1cx)
VIII 3,2. Prove that R det Ril;. ailA is always real in a Vn but not
always in an un.
1) ScHOUTEN and v. DANTZIG 1930, 1, p. 333, cf. E li 1938, 2, p. 237.
2) SCHOUTEN and V. DANTZIG 1931, 3; E li 1938, 2, p. 237f.
26*
404 VIII. Miscellaneous examples.

VIII3Jl). In a Vn with a;.x=8;.8xrp is given that e'~'fW)gW) 1s

an ordinary density of weight +
2. Prove that there exists a scalar
field 'IP such that
VIII 3,3 oc) a;.x - Vix = 8;. 8x log 'IP.

VIII 3,4. Let LPx be defined in a Vn by

VIII 3,4oc)

l
Prove that
) LPx = LxP
VIII 3,4) y) L -def L pP = L:il
p =real

b) V.L~=VpL (use3.32).

VIII 3,5. Let F~;t be defined in a V" by

VIII 3,5 oc) F:. vpi." + _2_
vpi." -det R:. a- L "
n - l vp i.

where LPx is defined in (Exerc. VIII 3,4oc). Prove that

VIII 3,5 )

4. The f~ of constant curvature. 2)

For a symmetric unitary connexion, R;;p;.x is symmetric in JlA (cf.

3.21, 29b). Now if
(4.1) R;;pi.x = oca;;(palxli.J = oca(ii(pax)i.)
and therefore
(4.2) R;;;,. = - ~ (n + 1) oc a;;;.
the vn
is said to be of constant curvature and is called an oc is always sn.
constant and real. 3) The scalar curvature (cf. III 5) is also constant
and real because

(4.3)

1) BERGMAN 1933, 1, p. 310.

2) ScHOUTEN and V. DANTZIG 1931, 3; FUCHS 1937, 1 also for a P'n with COU-
stant R; BocHNER 1947, 1; E II 1938, 2, p. 239.
3} SATO 1950. 1.
 4. The V,. of constant curvature. 405

If the connexion of a V,. is transformed projectively (cf. VI 1)

we get a symmetric (but not necessarily unitary) connexion
a) 'I;.'A=I;.''".+2PcPA~>; conj.
(4.4) {
b) 'liz";. = liz'A = 0

l
and the new curvature tensors are
a) 'R;;;,"=-2P[JJIA~+2A[.P~'1 ;.; conj.

(4.5) b) :R~~~: _ ~:'~'.~~~~ 2(V~Pcp) A~);defconj. .

c) Rp;;;. - R;;p;., conJ. PpA= VJJp"+p,..p", conj.
In order to make the new connexion integrable it is n. a. s. that for
n~2:

{
a) V,.. p" =PP P;.; conj.
(4.6)
b) V.aP;. = o.aPJ. = - toca_a;.; conj.
but this implies that P;. is a gradientvector:
(4.7)
The integrability conditions of (4.6) are identically satisfied in conse-
quence of (4.1) and the constancy of ot. That proves
An 5,. can always be transformed into an R,. by a projective transforma-
tion of the connexion.
In this R,. there exist coordinate systems with respect to which
'I;.";.= 0. If we use one of these coordinate systems in the we get s"
(4.8) { a) I;.:;. : ~ 2 A(;. o:> log ,u; conj.
b) liz;. = 0, COnJ.
where ,u is a solution of (4.7). For this coordinate system we get
a) oP ax_,. * - 2ax_,,. pp); conj.
b) o~'a"X * a"XPP+A;px; conj.
c) o~'p" * - ppp;.; conj.
(4.9) d) oppx * pxpp; conj.
e) oJJPx * -iocax_JJ; conj.
f) oJJ P"_!__p"p
- JJ _ _!_ocA"(l-!___pxp)
2 JJ ot x. conj.
Now it can easily be verified that

(4.10)
406 VIII. Miscellaneous examples.

is a solution of (4.7) because cx is real. From (4.10) we see that ,u is also

real. If we introduce the vector

(4.11)

we get from (4.9c, d) and (4.11)

(4.12)

Hence the z" are analytic and by integration of (4.12) it follows that

(4.13) ~" * z" + c"; c" * const.

But this means that the coordinate system can be chosen such that

(4.14)

From (4.6b) and (4.9a, c) it follows that

(4.15)

is a hermitian tensor with constant components:

4.16)
hence:
In an Sn the coordinate system can always be chosen such that in the
region where the equation

(4.17) ,u * IX "2
1- -;-aJ.;c~J.~x

has a real Solution (that is where 1 + 2cx a).;c e ~;c ;;;;; 0) the fundamental
tensor can be written in the form

(4.18) * 2 b;.;;
a;.;c=- + --aJ.iiaa;cs-
IXf-1,2 l:a I:Q
s-
IXf.J, 2

where b;.;c is a hermitian tensor with constant components.

But this form of a.a;c gives just the linear element of the hermitian
non-euclidean geometry that was investigated by FuBINI and STUDY.l)
Hence hermitian non-euclidean geometry in the sense of these authors
is realized in anS n, in the same way as ordinary non-euclidean geometry
is realized in an Sn.
1) FUBINI 1903, 2; STUDY 1905, 1; cf. CooLIDGE 1924, 1; CARTAN 1931, 1;
E li 1938, 2, p. 243; BocHNER 1947, 1, p. 185ff.
 5. Imbedding in an-i,.. 407

Exercise.
VIII 4,1 1). Let the gradient vector PA be the vector of the trans-
formation (4.4a) that transforms an S,. into an R,.. p, is defined by
(4.10). Prove that (use 4.9 and 4.15)
logp
VIII 4,1 oc) aJ.-=- 2 oJ.o---
" " Ot
and
VIII 4,1 )
with respect to the coordinate system (4.14).

5. Imbedding in an l ..
Let an X". in I,. be given by its parameter form
(5.1) a=1, ... ,m; conj.
where the functions f' are analytic in the domain considered. 2) If we
define the connecting quantity Bi: in the ordinary way
(5.2)
the analyticity of f' ensures that the Bi: are analytic:

(5.3) o0 Bi:= 0; conj.

We suppose that the X". is rigged and that the vectors t!'; y = m + 1, ... , n
,.
span the E,._". of the rigging. These vectors need not be analytic. Besides
the ,.t!' we have the vectors of L,. derived from the basis vectors geb in X".
{5.4)

These vectors are analytic and they span the tangent E". of X".. All
these vectors are given at points of X". only. From the vectors bt!', ,.t!';
conj. we form the reciprocal system #J.; iJ.; conj. in the ordinary way.
These covariant vectors need not be analytic. Now the formulae (V 7.1)
and their conjugates hold over X".. From them we need especially
the definition of the second connecting quantity C1:
(5.5) conj.
In contradistinction to Bi: this quantity need not be analytic.
1) Cf. FUCHS 1937, 1; BOCHNER 1947, 1.
8) COBURN also considered X".'s in U,. defined by semi-analytic functions,
1942, 1.
408 VIII. Miscellaneous examples.

Now let p be a semi-analytic scalar field of Xm. Then we have

the derivatives 8b p and 80 p and if p is prolonged somehow in a neigh-
bourhood of Xm we have also 8"'p and 8pp over Xm. The relations
(5.6)
hold between them. If v" is a vector field of Xm, the components of
this field considered as a field of I,. over Xm are v" = B~ vb. Hence a
connexion in Xm can be defined by the equations
a) '17c vadef B':_j: 171" v"; conj.
(5.7) {
b) '17.; v" def B~ B: 17p v"; conj.

where B: is defined as in (V 7.1). From (5.7) and the analyticity of B~

we get the following equations for the parameters of this connexion:

a) 'I;~= B~t:r;:A- B~f 8"" Bi; conj.

(5.8) {
b) 'Tc~= B~t:Ip"A- B1~ 8p Bi= 0; conj.
and this proves that the Xm with this induced connexion is an Im. From
(5.8a) we see that if the connexion in I,. is analytic, 8;;F,.."A = 0, it need
not necessarily follow that the connexion in Im is also analytic, because
B: need not be analytic. Alternation of (5.8a) gives
(5.8) c) '5"- cb" 5"-
cb- B"'Aa pA 8 B"b ]-- B"Aa
B""[c cb" 5"
p). conj.

and this means that every Xm imbedded and rigged in an A,. is itself
an Am whatever the rigging may be.
Instead of two curvature tensors with valence 3 as in (V 7.15, 16)
here we get four with their complex conjugates
(5.9) a) H"def
cb- B""r7
cbVp B"Jl b) h:"def
cb- Bi!'-17-B"
cbViJ A conj.
(5.10) conj.

but from these hii," and its complex conjugate are zero. In fact, accord-
ing to the analyticity of B~ we have
(5.11) cb - BilA"'
h""- C b Up B"-
- - Bil"'
C Up B"-
b - O' conj.
From (5.9a) it follows that

(5.12)
u.... = - B""A"'
{LZ[cb] cb U[p C" + B""A
cb ne
Jl] cb Fr"
C"11 - B"'A
.1 [p] [p Iai C"]
C""' BA + B 5" C"
= Au[c b] cb
11 '- p.il 11 11= B"'"
cb 5"e
p C"e conj.

and this proves that Hl;i,j is zero for every choice of the Xm and its
rigging if the I,. is an .A,. (cf. V 7.2J).
 5. Imbedding in an Ln. 409

An Lm in Ln will be called geodesie if every vector of Lm remains

in Lm if it is displaced parallel in a direction of Lm. N.a.s. conditions
are that for every field v" = Bi: vb the following equations should hold

(5.13) a) v11 (V11 v')C=0; conj.;

But these conditions are equivalent to

5.14) a) H(~b)=.,--Bfci)V11 C=0; conj. b) vvbBfl"VpCA=O; conj. 1).

From these equations the first is equivalent to H;;" = 0 if the Ln is

an An and the second is satisfied identically because hii,"=O (cf. 5.11).
It is easily proved that the transformation of the connexion in Ln

(5 .15) Pr;.J.j = 0; conj.

leaves invariant the set of all Lm's which aregeodesie according to this
condition, if and only if P,;.i" has the form 2p<11 AAJ that is, if the trans-
formation is projective (cf. VI 1) 2).
For an rm
in Ln a D-symbolism may be introduced in the same way
as for an Lm in Ln (cf. V 7). Because the fields considered are all
defined only over the Lm here we need only D11 and Dc and their con-
jugates. Using these symbols, the tensors H;1,", Li~;. and l;,";. get the
form (cf. V 7.15 b, 16b)
(5.16) a) H;'b"=D,Bi:; conj. (b) O.=DzBi:; conj.)
(5 .17) a) Li~;.= D, Bt conj. b) l;~;. = Dc B~; conj.
We see from this that the greek index in these three formulae always
lies in the rigging and that (cf. V7.15a,16a)

(5 .18) H~6s = C~DcBi: =- Bi:D.C~; conj.

{ a) Li~ y = c; Dz B~ =-,- B~Dc c;; conj.

(5.19)
b) l;~Y = c;n. B~ = - B~Dcc;; conj.
1 ) If instead of (5.13) we take (vl'~vA+villi(Iv") C~=O; eonj. as the definition

of the geodesie lm, we get the same res~lt (5.14) beeause hii," = 0. Butthis latter
definition represents just the eondition for the L 2 m eorresponding to the lm tobe
geodesie in the auxiliary L 2 n. .
2 ) Hereisa differenee with BocHNER 1947, 1. He eonsiders allgeodesie sub-

spaees in the auxiliary L 2 n and requires that they all remain geodesie. Then it
ean be proved that this stronger eondition leads to P;.i"=O. But from our point
of view the L 2 n is only an auxiliary spaee in whieh we are only interested in what
happens to subspaces in the prineipal Xn, and this leads to P~i_"=2P< 11 A~>
 410 VIII. Miscellaneous. examples.

If v" is a field of Lm we have (cf. 111 4.9b; V 8.4)

'R,m,"vb- 2' S.;; 'V.,v" = 2'11[,/V.,lv" = 2DrdD-1v" = 2D[dB~1':,17,.v'
(5.20) { [dc] " fJ v"+2B~'l"
=2HP"l7: [c d], y fl""Yvb+
pb
+ dc>< ~fJ

l
~fJ" R"" "VA- 2VfJa
dc,. Sl/17
~fJ "e
v" ,
hence
('R.;;b" + 2B[cl.i]~yH~b'Y + BJ~::R;;,;,") vb
(5.21) = 2 S;,;.ll B~:;: 17. v" + 2B~: s;,;.11 C~ B: V" v"-
- 2BJ~: S;j/ 17(! v" = 0
or (cf. V 7.}2; 8.4)

(5.22) 'R"""-
dcb - tJa vp"
dcbx R + 21" [d.lyl Hc]b'Y, conj. (GAUSS I)

In the same way we get (cf. V 7.32; 8.4)

(5.23) 'R"""-;;P""R"+L""
leb - i!cbx iipA i!.:r H"''"
cb , conj. (GAuss II)

So we have in r ..
two GAUSS equations and their conjugates. 1)
Proceeding in the same way we get after some calculations fiv
CODAZZI equations (cf. V 7.48, 51; 8.7, 8):

(5.24)

(5 .25) Dii H""'"-C'"tJ;;R ......'

cb - " cbi! iipJ. conj. I (ConAZZI II)

(5.26) " = - B~P"CAR"

2D [d lc].y dcx y VfJ - 2 dc 'Y' conJ
'St" (CODAZZI 111)

(5.27) \ D [i! L"

c]. y - i!c 1i. y,
'SL" conj. I (CODAZZI IV)

(5.28) l .a +Dc L"

- D i!c.y -C"B"'";;R"
i!.y- y c><ii iipA, con]' (CODAZZI V)

with their conjugates and two Rzcci equations (cf. V 7.43; 8.6):
mm'
(5.29) '" = B~P
R dcy dc C'"" " + 2H [dlbl
><'Y Rvp '"lc],y
b conJ. (RICCI I)

mm'
(5.30) i!cy i!c 'Y>< vpJ. cb ii. 'Y'
R""'"=;;PChR!"-H""'"Lb conj. (RICCI II)

1) Cf. HoMBU 1935, 1.

 s. Imbedding in an l ... 411

mm' mm'
with their conjugates. Here R;;;" and RJ;;" are the quantities defined
by (cf. V 7.42; 8.5)
mm'
(5.31) 2DcdDcJ r" = RJ;j," rY- 2 'S;; D.r"

(5.32)

with an arbitrary vector r" in the (n- m)-direction of the rigging.

Only the three CoDAZZI equations (5.24, 26, 27) take a simpler form
if the I .. is an A.,.
In a tJ., the induced fundamental tensor is

(5.33)

and the imbedding is fixed by B~. Concomitants of B~ and aA;c are

(5.34) B~ = 'a'6a B~a;,;c; conj.
and
(5. 35)

We easily derive from these equations that in a Un

(5.36) Dc 'ab = 0; conj.
and
a) L!"
c .;, =H!!Ila-a"
ca conj.
(5.37) { Q '

b) l;~A = h;;/1 a" a;,x= 0; conj.

From (5.36) we see that the imbedded Xm is a Um with 'ab as funda-

mental tensor. (5. 3 7) expresses the facts that the quantities l; ~i. and
hij," both vanish and that Ha_" and L~~x are different components of
one and the same quantity.
The first equation of GAuss (5.22) drops out because both sides
are zero and the second (5.23) takes the form

(5 .38) I 'R ..
ilcb a-- Biipl.a
ilcb>< R"
;;,_.;, + Hax
il -
Hcbx' conJ. I (GAuss).
The third equation of CoDAZZI (5.26) drops out, the first (5.24) 1s
equivalent to the fourth (5 .27):

(5.39) I D [d Hc] b "- - - 'Sa

dc H"
ab ' conJ'I (CoDAZZI I)
L-------------------------~
412 VIII. Miscellaneous examples.

and the second (5.25) is equivalent to the fifth (5.28):

(5.40) I DH .. "
il cb = cbil R~"
C"" B";.;; vp ' conJI (CoDAZZI li).
~--------------------------~

The equations of Rrccr take the form

I x~;" = 0; conj.l (RICCI I)

(5.41)
\ m~ I
Rj~j" = B~~ C~~R~~~"- H:b" Hi~y;
L __ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ __ _
conj. (Rrcci li}.

Among all these equations only the CoDAZZI equation (5 -39) contains
S,t~" and it is only this equation that is simplified if the
I is a V,. .1) an
Exercise.
VIII 5,1 2 ). Prove that for an Um in Un

if ~ is analytic in r{. There is no analogous relation for a Um in U,..

6. Curves in a iJ" with a positive definite fundamental

tensor.
Form= 1 we write 'fJ, Tj, d~, dti instead of rJ 1 , n1, 81 , 81 and we suppose
that nowhere d~ ~" is a nullvector. Then we have
(6.1) B~ = d~~"; conj.

(6.2) 'a1 y = a;.;< (d'l ~;.) (dti ~) =d~~rj =I~~ \z; ' 11
a
-Ids ~-Z.
-di/ ' conj.

(6.3) 'I'/
11
= 'ailo'1 'a11- = 28T) log~~~
drJ ' conJ.

Here ds is defined by ds 2 = a;.x de d~x and it is not a differential of a

function of 'YJ or of 'YJ and Tj. In fact is two-dimensional and Jds4
depends on the way of integration. From (5:16) and (6-3) it follows that

(6.4)

1 ) Cf. for imbedding in fln ScHaUTEN 1929, 1; ScHaUTEN and v. DANTZIG

1930, 1, p. 335ff.; FucHs 1935, 1; 1937, 1; RozENFEL'D 1949, 2; CALABI 1953, 1.

Imbedding in R., was dealt with by BocHNER 1947, 1, p. 190ff.
2) ScHauTEN and v. DANTZIG 1930, 1; E II 1938, 2, p. 246.
 6. Curves in a fT,. with a positive definite fundamental tensor. 413

Ifinstead of ds we introduce a differentialdz defined by dsz =dz dz 1),

dz is determined to within a factor e'P with real f/J In general dz is
not a ditferential of a function and the expression z without a letter "d" has
no meaning. Using dz we have instead of (6.2-4) equations of exactly
the same form but with dz instead of ds.
The tangent unitvector can now be defined by
def d;" -1 d~"
(6.5) 1"' = --=a --;
dz d'YJ

It is defined to within a factor e'P with qJ real. This is not essentially

new because the tangent unitvector of a real curve in V,. is also fixed
to within a factor 1 because its sign depends on the choice of the
sense on the curve in which s increases.
If (6.5) is substituted in (6.4) (for dYJ-+dz) we get

(6.6) l Hii" =
.,
a1"'- 2af" 811 log Ia I + aZ J;."" j~'j"
dd'YI

= az j~' V,.f"- az jl'f" 8,. log ; conj.

because 8/811 = a j~' 8,.. If the curvature vector of the curve is defined
asin V 9
(6.7)

it follows that for a U1 in a U,. (cf. V 9)

(6.8)

in contradistinction to the case of a V1 in V,. where u" equals j~' V,. 7"'.
Note that u" gets a factor eZirp if j" gets a factor e'P. The vector jP V,. f"
r
l
is not perpendicular to because
au,j~'jxV,.jA= -~j;;V,.jx= -a-1j;;,8'1jx
(6.9)
=- a- 11'- d~" 8 a--J =1'~'8 loga- conJ
" dij '1 IJ '

but u" is the part of j~' V,. f" perpendicular to f' as follows from (6.8, 9)

(6.10) jA u" = f~' 8,. log- j~' 8,. log = 0; conj.

and also from the fact that the "-region of Hii" is perpendicular to i"
A geodesie U1 in V;. is characterized by

(6.11)
1) COBURN 1941, J,
414 VIII. Miscellaneous examples.

N ow the left hand side of this equation is equal to

(6.12)
(J j" 81' (J j" + a2 I;";. j" / = a2 j" 171' j" + a2 j" r 81' log (J

Hii" + 2a2 j"fP 8" log JaJ;

{
= conj.
hence the n.a.s. condition for a curve to be a geodesie is (cf. 5.14)
(6.13) Hii" = 0; conj.
[equivalent to u" = 0 according to (6.7)] and every geodesie satisfies the
equation
(6.14)

Note that the right hand side of this equation does not depend on the
choice of the free factor in j" because a j" and Jai are independent of
this choice.
We now consider the special case where the free factor in dz can
be chosen in such a way that dz is a differential of an analytic function
z(17). Because of (6.5) we then have
dl;" .,. dz . .
(6.15) -7 d;J'
d;J- COllJ.

and
(6.16) dz
- = J --=J
. d!;" . B" . conj.
1=7J"
d'Y} " d'YJ " '
But (6.16) expresses the fact that the vector jb in U1 is analytic. Conver-
sely, if jb is analytic, dzfdrJ is analytic and z can be found by integration
as an analytic function. It has already been proved in VIII 3 that jb is
a
analytic if and only if the 1 is an R1. Because j 1 is determined to within
a factor e;"' with rp real and analytic in 1], it follows that q; must be a
constant. Hence z is determined to within an affine transformation
'z=az+b with constant coefficients and with Jal =1. We call z a
natural Parameter on the R1 in an.
Collecting results we have 1) :
If an X 1 in a an
is given by the analytic functions ;" (17) of a complex
parameter 1], this xl is always a fJl. If and only if this f)l is an Rl'
that is, if and only if the induced connexion is analytic and thus integrable,
it is possible to find a parameter z = z (rJ) on this curve analytic in rJ such
that ds 2 = dz dz. This natural parameter is fixed to within an affine
+
transformation 'z = a z b with constant coefficients and with I aJ = 1.
Only these R1's in an
have an analytic tangent unitvector d;"fdz.
They are in many respects analogaus to the V1's in a V". 2) For instance
u" =j" 17pj"; (conj.) comes instead of (6.8).
1 ) ScHOUTEN and v. DANTZIG 1930, 1, p. 343; E II 1938, 2, p. 251. The theorem
was proved there without the auxiliary differential dz of COBURN.
2) Cf. COBURN 1941, 3.
 7. Conformal transformation of a connexion in U,.. 41 5

We see from (6.14) that the equation of a geodesie with respect to

a natural parameterz (if existing) takes the simple form

(6.17) conj.

There is still another kind of curves in U,. that arises if we consider

equations of the form ;" = ~" (t) where t is a real parameter .1) These
curves are one-dimensional. It is possible to get them from the fts
in U,. that are R1's, by considering only the points of such an R1 where
some definitely chosen natural parameter takes real values. Of course
the natural parameter on such a curve is fixed to within transformations
of the form 'z=z+b with real constant b. 2)

7. Conformal transformation of a connexion in Un.

If the fundamental tensor a;.;c of a U,. is transformed into
(7.1)
where a is an arbitrary redl semi-analytic scalar, the parameters r;.~.,
I{.";. transform into

{
a) 'J;";.= 1JQ"o,..'a;.Q=J;.";.+a,..A'J.; conj.
(7.2)
b) 'J{.";.=J{.";.= 0; conj. a,.. def o,..loga; conj.

We call this transformation conformal (cf. VI 5). From (7.2) we get

(7.3)
and
a) '17,.. v" = 17" v" + a,.. v"; '17p v" = 17p v" = Bp v"; conj.
(7.4) {
b) '17,.. W;. = 171' W;.- (Jf' W;.; '17pWo~= 17pW;.= OpWA; conj.

The transformation (7.2a) is projective (cf. VI 1) and leaves parallel

fields parallel (cf. VI 1). If and only if a;. = 0, J;";. and 'I;";. can for
n_>l be symmetric at the same time. Hence
A conformal transformation of a fJ,. is at the same time a restricted
projective transformation 3 ) and a;. is the gradient of a real. scalar. If a
1 ) ScHOUTEN and v. DANTZIG 1930, 1, p. 343; 1931, 3; CoBURN 1941, t; 3. Cf.
COBURN 1942, 2; SUGURI 19 51, 1 for FRENET formulae for these curves. V ARGA 1939, 1
dealt with curves in the elliptic 112 and GoLIFMAN 1941, 1 considered one-dimensional
curves in the hyperbolic 112
2 ) Congruences of curves of both kinds were studied by CoBURN 1943, 1.
3 ) CoBURN 1941, 1, p. 32f. Cf. VI 1; cf. for projective transformations also
DE MIRA FERNANDES 1950, 1.
416 VIII. Miscellaneous examples.

conformal transformation leaves the connexion of a V,., n> 1, symmetric,

a.il;c gets a constant factor. 1 )
This latter statement implies that the only conformal transforma-
tion that transforms an S,. into an S,. is a transformation with a = const.
We have seen in VI 5 that an S,. can be transformed conformally
into every other S,.. This is quite different in a unitary space. In
VI 5 we also saw that the connexion of a V,. is determined by its
conformal and projective properties together. Now Iet
(7.5)

be a connexion derived from the F,.".il of a ff,. by a conformal transforma-

tion. Then 8~' log a can be computed from 'T,.".il if SI'= _!___ S;;..il is
n- 1
given. Hence
The connexion of a V,. is determined by its conformal properties only,
but for the determination of the connexion of a V,. we also need the field s,...
s;;," and SI' transform as follows {cf. 7.2)
(7.6) a) 'Spi."= S""+
p.il a [p A".il]
{
b) 'Sp= SP+a~'
hence
The connexion of a U,. can be transformed conformally into a symmetric
connexion if and only if it is semi-symmetric and if SI' is a gradient of
a real scalar. 2)
The transformation of the curvature quantities can be derived from
(2.17, 21) and (7.2):
a) 'Ri;i." = Ri;i." +A~ o;;ap; conj.
b) 'R;ii."= R;j;." -A~ o;;ap; conj.
(7.7) c) 'Ru= R;;.il- o;;a;.; conj.
d) +
'V;I' = v;l' n o;;ap; conj.
e) 'R = a- R- a- av.il o;;a"; conj.
1 1

A fJ,. is called conformally euclidean or a C,. if it can be transformed

conformally into an R,.. From (7.6a, 7a) it follows:
A fJ,. is a C,. if and only if there exists a real semi-analytic gradient
field a;. such that
(7.8) a) Ri;i"= -A~ o;;ap; conj.
1) CoBURN 1941, 1, p. 31; 1942, 3, p. 140.
) A gradient of a real scalar need not have real components.
 7. Gonformal transformation of a connexion in ,.. 41 7

For n > 1 it follows from A['l' O'J.J = 0 that O'J.. = 0, hence

A V,. is for n > 1 a C,. if and only if it is an R,. .1 )
As a corollary we get that no S,., n>1, with non vanishing cur-
vature is a C,. .1)
From (7.8) we see that a C,. is always semi-symmetric and that

(7.9)

In VIII 3 it has been proved that for a semi-symmetric U,., n>2, SJ..
is always a gradient vector. Now if SJ.. equals o,. q; it is also the gradient
of the real scalar Hq; + qi) if and only if o;; S A = oJ.. S;;. Hence 2)
A U,. can be transformed conformally into a V,. if and only if for
n>2 the connexion is semi-symmetric and opS,.=oJ..Sp, and for n=2
S J.. is a gradient of a real scalar. A U,. is a C,. if and only if these conditions
are satisfied and if moreover

(7.10) 0= c~ vpA"-
"det R~
vpJ..- A"A 8-v SI'

c;;";." is called the conformal curvature tensor of the ff,..

CoBURN has dealt with several problems concerning conformal trans-
formations in fJ,.. He mostly considers the transformation not as a
transformation of the connexion in one and the same space but as a
transformation of a space ff,. into another space 'U,.. But of course
this interpretation is not essential. Since such a conformal transfor-
mation is always at the sametime a restricted projective transformation
he requires all those restricted projective transformations of this kind
that transform a fJ,. into a V,.. 3) In the same paper 4) he considers a
conformally symmetric U,., that is a fJ,. with the fundamental tensor

(7.11) =real; q; =real,

subjected to conformal transformations of the form

(7.12) oc = real and const.

and deduces invariants for these transformations.

1 ) CoBURN 1942, 3, p. 140, only for n >2. Remark that every S,. is a C,.

(cf. VI 5).
2 ) COBURN 1942, 3, p. 138.
8) CoBURN 1941, 1, p. 32ff.
4 ) p. 37ff.

Schonten, Ricci-Calculus, 2. Aufl. 27

418 VIII. Miscellaneous examples.

Exercises.
VIII 7, 1 1 ). In a Un that can be transformed into a Un with R~~;." = 0
by the conformal transformation au1 -'>-a a;.;; the following identities hold

VIII 7,1 tZ) a = Det (a 1 ;;)J

J

VIII 7,1 ) R; = V.a;

VIII 7,2. A Vn, n >1, can be transformed conformally into a un
with R~~;." = 0, if it is an Rn (cf. the special case mentioned in the text).
VIII 7,3. If in a semi-symmetric fJn the vector S J. is a gradient
of a real scalar and if there exists a coordinate system (x) such that
VIII 7,3 tZ)
it can be proved that the Un is a Cn and that there exists an Rn such
that the geodesics of an
and Rn are in restricted correspondence. 2)
VIII 7,4. In a Un the tensor Vd;. + n Ril J. is a conformal concomitant
that vanishes if the (]n is a cn.
8. Conformal unitary connexions 3).
The quantity (cf. VI 7)
1
cw def -,.
(8.1) u;.;; = a a;.;;, conj.

is a hermitian tensor density of weight - _!__. It is conformally invariant

n
+
and its determinant is 1. Analogaus to VI 7 and (VIII 3.4) we may
define from it the conformal parameters

(8.2) rc.u\=~1 "o.uuJ.=

cw 1 A"" 1
T.u"-'-n -T"
;.u.u oga- .u;.-n TA"
1
1, ;., conj.
They transform as follows

conj.
(8-3)

Now in contradistinction to the conformal parameters belanging to

an ordinary metric dealt with in VI 7 it is possible to derive a connexion
1) Cf. Exerc. VI 5, 1.
2 ) CoBURN 1941, 1 calls the correspondence between geodesics restricted if

the vector of the projective transformation is a gradient of a real scalar.

3 ) CoBURN 1942, 3, p. 129ff. This section is a free interpretation and at some
points a simplification of the most important of his results.
 8. Conformal unitary connexions. 419
c
from J;,";" without using any auxiliary objects or invariant objects of
higher order. In fact, if we define for n > 1

(8.4) conj.
c
the II;;" transform just like the J;,";" becauseA~ S'" is a tensor, and these
new parameters are conformally invariant because of (7.5, 6b). De-
c
noting the covariant derivative belanging to them by V we have
c

(8.5)
J a) V'"a;";c= V'"a;";c+ SPA~aa;c= S'"a;";c; conj.
1b) c c
17'"a=aV'"loga=aa
;c;.
c
r::,a;.;c=naS'"; conj.
hence
c c __ 1_
(8.6) V'"~;.;c=V'"a na;.;c=O; conj.
c
We call II;;. the parameters of the conformal connexion belanging
to La;.;cj. From (8.4) it follows that the conformal connexion is identical
with the connexion belanging to a;.;c if and only if S'" = 0, that is (cf. ).9)
if and only if a;.;c satisfies the differential equation

(8.7)

This is always true in a V,,, hence the conformal connexion of a Vn is

the connexion of the vn
itself. In fact we have already Seen that for
a conformal transformation of a V,. into a always (J = const. As avn
generalization we may mention the fact that for a conformal trans-
formation of a ffn with S'" = 0 into a Un with S'" = 0 also a = const.
as follows immediately from (7.2a; 6b).
c
The following conformally invariant tensors can be deduced from II; ,_

The conformal invariance of the first one follows from (7.6b) and the
second one equals the conformal curvature tensor defined in (7.10).
Both vanish in a Cn.
c
The connexion II;;. leaves invariant the hermitian tensor density
~;";c but not in general a hermitian tensor. If in an exceptional case a
27*
420 VIII. Miscellaneous examples.

tensor b;.;c with rank n is invariant

c c
(8.9) 17" b;.;c = 8" b;.;c- II1~;. bex = 0; conj.
the first integrability conditions must be satisfied:
c

(8.10) Ja) be;cR;~;.e=-b;.;c8[.s"l=0; conj.

l b) R~~;. e bex- R~~~ Ii b;.fi- b;.;c 811 S" + b;.;c 8" S 11 = 0; conj.
-1
From (8.10a) and (8.10b) transvected with b;c;. (cf. 3.19d) it follows that
S" must be the gradient of a real scalar (cf. VIII 7). But then the last
two terms in (8.1 0 b) vanish and there remains only the algebraic equation
(8.11) . Q bex- Rp,:ix
R iil';. -
fibfi- o, conj.
Because of (3 .22) we have
(8.12) conj.

and this proves that (8.11) is satisfied if b;.;c=aa;.;c, where a isareal

coefficient. In fact there exists such a solution if S" is a gradient of
a real scalar because if in this case we take a" = - S 1, we get
c
(8.13) + a S" a;.;c = 0; conj.
V"aa.it;c = aa"a;.;c
c
and aa;.;c is the fundamental tensor of a un with parametersii;;. in accor-
~

dance with VIII 3.

But there may exist other solutions. In order to find them we have
to differentiate (8.11) and to eliminate the derivatives of b;.;c by means
of (8.10). Proceeding in this way we get at last a set of algebraic equa-
tions for b;.;c that can be no more extended and that may have solutions
with rank n (cf. the theory of sets of partial differential equations in
c
II 5). If there is no solution with rank n the II~;. can not be the para-
meters of the connexion of a Un.

Exercise.
VIII 8,1. If in a Un (cf. Exerc. VIII 7,1)

VIII 8,1 oc)

and if 't = f (~") g (~x) is an ordinary density of weight - 2, the tJn with
the fundamental tensor
1
'a;.;c = -r-n-2l;.;c
has 'R~~i" = 0.
 9. Spaces of recurrent curvature. 42t

9. Spaces of recurrent curvature. 1)

In III 2 a recurrent field was defined as a field that at every point
and for every direction is proportional to its covariant differential and
in VII 4 we considered, from the point of view of their holonomy
group, spaces with a recurrent curvature tensor. 2) A V,. is called a K,.
if K.pJ.K is recurrent but not covariant constant
(9.1}
and it is said to be a K! if it is either a K,. or symmetric
(9.2)
and if in this latter case there exists a vector kw such that
(9.3)
If (9.2, 3) hold, kw need not be uniquely determined. For instance
an R,. is a K! with undetermined kw and an S,. with K =!= 0 is neither
a K,.nor a K!.
For n = 2 we have (cf. III 5.31):
(9.4)
hence
(9.5)
and this proves that a V2 is either an 5 2 with u =l= 0 or a K 2.
In VII 4 we proved that kw is a gradient in a K,.. In a symmetric
K! the integrability conditions of
(9.6)
are identically satisfied in consequence of (9.2). Hence kw can always
be chosen in such a way that it is a gradient. 3 )
If for n>2 a K! is a product Vm xV..-m (cf. V 11) the linear ele-

l
ment can be written in the form
d.s 2 * gA,p, W) d~" d~" + gA,p, W'') de d~~';
(9.7) "1 .1.U1-1, ... ,m
"2 .2,.U 2 = m + 1, ... , n
1 ) This subject was introduced by RusE 1948, 2; 1949, 2. Hisfirst investiga-
tions were connected with harmonic spaces and dealt especially with the cases
n = 2, 3, 4. WALKER 1950, 1 introduced the J4:
and the simple K! by means of
which he and RusE in 1951, 1; 2 succeeded in giving a classification.
1) Spaces with recurrent RrcCI tensor were considered by PATTERSON 1952, 2.
3) WALKER 1950, 1, p. 46.
422 VIII. Miscellaneous examples.

and from this we see that r;;.

and K.,_.;.,. have no components with
indices from both ranges. But then it follows from (9.3) that

(9.8)

and this implies that either V", or V"_m or both are euclidean. The only
non-trivial case is where only one of them, for instance V..-m is euclidean.
Because the rank of g;.,p, must be n- m, V"_m must be an R"_m. Because
kw,=l= 0 and independent of the ~" we have k[w,K.,,_., 1;.,,., = 0 and this
implies that the Vm is a K!. Hence 1 )
I f a K! is decomposable it is a product of a K! and an R"_m.
K! is called the flat extension of K!. Of course every flat extension
of a K! is some K! and the same holds for Km and K,.. 2)
If a K! contains a parallel field of m-directions and a field of perpen-
dicular (n- m)-directions, the first field forms a normal system of
geodesie Vm's and the second a normal system of geodesie V"_m's per-
pendicular to the first. But then the linear element can be written in
the form (9.7). Hence, the V..-m's are R"_m's and the Vm's are K!'s or
vice versa. If the V"_m's are R"_m's there exist in them n- m mututally
perpendicular non-null vector fields that are not only parallel but also
covariant constant. 3 ) Hence 4 )
If a K! admits exactly n-m ;;::,.1 mututally perpendicular non-null
covariant constant vector fields it is a product of a K! and an R,. -m.
This leads to the consideration of a K! admitting exactly r mutually
perpendicular covariant constant null vector fields. Now ErsENHART 5 )
proved that in a V,. of this kind n-r is always ;;;;;r and that the linear
element can be written in the form

ds 2 * g{Ja. d~fJ d~ + 21: d~X d~x+n-r; _

(9.9) { cx, -1, ... , n r
x=1, ... ,r
where the gfJct. depend on the ~ct. only. Such a V,. is called a null-extension
of the V"_, with the linear element gfJct. de d~. This does not imply
that the V,. is decomposable. If this result is applied to a K! with n- m ;;;;;; 1
1)WALKER 1950, 1, p. 38.
2)RusE 1949, 2.
8 ) We differ here from RusE and WALKER. We call a vector field covariant
constant if its covariant derivative vanishes at all points, and parallel if at every
point and for every direction it is proportional to its covariant differential (cf.
III 2).
4) WALKER 1950, 1, p. 39.
5 ) EISENHART 1938, 1 ; cf. WALKER 1949, 1 ; 19 50, 3; 4.
 9. Spaces of recurrent curvature. 423

mutually perpendicular covariant constant vector fields from which r

arenull and n-m-r non-null, we get the linear elemenP)

ds 2 * g(/. a; d;"' + 2.E a;x a;x+m + .E e a;w a;w

w

r::t.,=1, ... ,m
(9.10) x=1, ... ,r
w= r + m + 1, ... , n
e= 1.
w

A Vn that admits n- 2 independent covariant constant vector fields

can be proved to be a K! . Such a K! is called simple. 2 ) The classifica-
tion of all simple K!'s makes use of the following propositions due to
RusE
Every V2 is a Kt. Every V3 which is a null-extension of a V2 is a K:.
Every V4 which is a null-extension of a V2 is a Kt and harmonic.
From this it can be proved that there are only three cases for the
simple K! . 2 ) lf there are exactly n- 2 mutually perpendicular co-
variant constant vector fields of which exactly rare null vectors, we have
a) r = 0: K! is a flat extension of a Kt;
b) r = 1: K! is a flat extension of a K: which is a null-extension
of a Kt;
c) r = 2: K! is a flat extension of a K~ which is a null-extension
of a Kt.
From the linear element (9.10) it can be proved by direct computa-
tion that K.I'A>e in a simple K! is the general product of two equal
simple bivectors:
(9.11)

Conversely, if we know that the curvature tensor of a K! has the form

(9.11) the E"_ 2 spanned by u;. and v;. forms a parallel field. Hence a
K! is simple if and only if its curvature tensor has the form (9.11). 3)
RusE 4 ) has given the following classification of the simple K!'s:
I. k;. non-null; K! simple
A. K =!= 0; K! = Kt X R"_2,
B. K = 0; K~';. =!= 0; K! = k: X R"_ 3 ;
1) WALKER 1950, 1.
2) WALKER 1950, 1, p. 40ff.
3) WALKER 1950, 1, p. 44.
4) RusE 1951, 1; 2.
424 VIII. Miscellaneous examples.

II. k;. null

A. K 9= 0; K! = K: X Rn-2;
B. K = 0; Kil;. 9= 0; K! = k: X Rn_s;
C. K'";. = 0; K! = kt X Rn-4;
where k: and kt are defined as K*-spaces with the linear element
(9.12)
and
(9.13) ds 2 -~1p(x, z) dx 2 + 2dx dy + 2dz dt
respectively. WALKER 1) proved that the linear element of a non-simple
K! for n >3 can always be written in the form
(9.14) ds 2 !Ltpde de + 2de d~2 + kapd~""d;P; IX, = 3, ... , n

where
(9.15) 1p * () aap ~a ~fJ + Xrx ~a; aap * const.; kar~ const.; [kapl 9= 0

and where () 9= 0 and the x,. are functions of e.

The K! is symmetric
if and only if () = const.
In the same paperWALKERalso gave particulars about the symmetric
K!, the EINSTEIN K! 2) and the K! which is a C,.. Moreover he proved
some remarkable theorems on harmonic K!'s, for instance that for n >2
every harmonic K! is simply harmonic (cf. VIII 1) and that for every
n >3 there exists a simply harmonic K! which is non-decomposable.

Exercises.
VIII 9,1. Prove that the V,.'s with the linear elements
VIII 9,1 rl) ds 2 ~1p(x, z) dx 2 + 2dxdy + dz 2
and
VIII 9,1 ) ds 2 * 1p(x, z) dx 2 + 2dxdy + 2dzdt
are both K!'s.
VIII 9,2. Prove that the Kt with the linear element

ds 2 *- 1p(x, y) dx 2 + 2g;(x, y) dxdy +C(x, y) dy 2 + 2dxdz + 2dydt

is simply harmonic (cf. Exerc. VIII 1,2).
VIII 9,3 3 ). Every K!, n >2, that is an EINSTEIN space, is a special
EINSTEIN space.
1) WALKER 1950, 1, p. 51.
2) Cf. also MoGI 1950, 1.
3) WALKER 19 50, 1.
 Bibliography 1 ).
1944, 1, 2
ABE, M.: Sur la reduetibilite du groupe d'holonomie I, Les espaees a eonnexion
affine. Proe. Imp. Aead. Tokyo 20, 56--60; 177-182 . . . . . . . . . . 367
1950, 1
ABRAMOV, B. M. (see also B. A. RozENFEL'D): Topologieal invariants of Rie-
mann spaees. Uspeehi Mat. Nauk 5, 162-163 (russian) . . . . . . . . . 269
1951, 1
- On topologieal invariants of riemannian spaees obtained by the integration
of tensor fields. Akad. Nauk SSSR. 81, 125-128 (russian) . . . . . . . 269
1951, 2
- On topologieal invariants of riemannian spaees obtained by the integration
of pseudotensor fields. Akad. Nauk SSSR. 81, 325-328 (russian) 269
1951, 1, 2
ADATI, T. (see also K. YANo): On riemannian spaees admitting a family of
totally umbilieal hypersurfaees I, II. Proe. Japan Aead. 27, 1-6,49-54 251, 272
1951, 3-8
On subprojeetive spaees I-VI. Thoku Math. J. 3, 159-173, 330-342,
343-358; Tensor I, 105-115,116--129,130-136 . . . . 321,322,324,327,329
1933, 1
AGOSTINELLI, C.: Sui sistemi di velocita e le famiglie naturali di linee in uno
spazio eurvo. Atti Pont. Aeead. Sei. Nuovi Lineei 86, 287-314 . 138, 253
1933, 2
*- Parallelismo in un sistemi di S k euclidei tangenti a una varieta Vn . Gili,
Torino 31 pp.
1933, 3
Sullo spostamento eonforme in una varieta V". Atti Ist. Veneto. Sei. 92,
523-525 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
1952, 1
AKIVIS, M. A.: Invariant eonstruetion of the geometry of a hypersurfaee of a
eonformal spaee. Akad. Nauk SSSR. 82, 325-328 (russian) . . . . . . 315
1925, 1
ALEXANDER, J. W.: On the deeomposition of tensors. Ann. Math. 27, 421-423 20
1937, 1
ALLENDOERFER, C. B.: Einstein spaees of class one. Bull. Am. Math. Soc. 43,
265-270 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
1937, 2
- The imbedding of Riemann spaees in the large. Duke Math. J. 3, 317-333 268
1939, 1
- Rigidity for spaees of class greater than one. Am. J. Math. 61, 633-644
268, 269, 275
1 ) Titles marked with * are books or monographies.
426 Bibliography.

1952. 1
* ANDELIC, T. P.: Tensorcalculus. Naucna Knjiga, Belgrade (yugoslavian)
1936, 1
ANDERSON, N. L., and L. lNGOLD: Normalstoa space V,. in hyperspace. Bull.
Am. Math. Soc. 42, 429-435 . . . . . . . . . . . . . . . . . . 269, 275
1935. 1
ANDREOLI, G.: Elementi di una geometria-differenziale di una V1 in una v;,
isotrope cayleyane. Ace. Sc. Napoli Rend. (5) 4, 2o-23. . . . . . . . . 227
1864, 1
AousT, L.: Theorie des coordonnees curvilignes quelconques. Ann. di Mat.
6, 64-87 . . . . . . . . . . . . . . . . . . . . . . . . . 253
1921, 1
BAcH, R.: Zur Weylschen Relativittstheorie und der Weylschen Erweiterung
des Krmmungsbegriffs. Math. Zeitschr. 9, 110-135 . . . . . . . . . 14, 147
1943. 1
BECKENBACH, E. F., and R. H. BING: Conformal minimal varieties. Duke
Math. J. 10, 63 7-640 . . . . . . . . . . . . . . . . . . . . 315
1874. 1
BEEZ, R.: ber das Krmmungsma von Mannigfaltigkeiten hherer Ordnung.
Math. Ann. 7, 387-395 . . . . . . . . . . . . . . . . . . . . . . . 244
1875. 1
- Zur Theorie des Krmmungsmaes von Mannigfaltigkeiten hherer Ord-
nung. Zeitschr. f. Math. u. Phys. 20, 423-444; 21, 373-401 (1876) . . . . 244
1934, 1
*BEHNKE, H., and P. THULLEN: Theorie der Funktionen mehrerer komplexer
Vernderlichen. Springer, Berlin. 115 pp.; Chelsea, New York . . 61
1868, 1
BELTRAMI, E.: Teoria fondamentale degli spazii di curvature costante. Ann.
di Mat. (2), 232-255; Opere III, 406-409. . . . . . . . . . . . 294
1869, 1
- Zur Theorie des Krmmungsmaes. Math. Ann. 1, 575-582; Opere II, 119
to 128 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
1933. 1
BERGMAN, S.: ber eine in der Theorie der Funktionen von zwei komplexen
Vernderlichen auftretende unitre Geometrie. Proc. Kon. Ned. Akad.
Amsterdam 36, 307-313 . . . . . . . . . . . . . . . . . . . . 397,404
1922, 1
BERWALD, L.: Zur Geometrie einer n-dimensionalen Riemannschen Mannig-
faltigkeit im (n + 1)-dimensionalen euklidisch-affinen Raum. J. D. M. V.
31. 162-170 . . . . . . . . . . . . . . . . . . . . . . . . . 237, 242
1922, 2
- Die Grundgleichungen der Hyperflchen im euklidischen Raum gegenber
den inhaltstreuen Affinitten. Monatshefte f. Math. u. Phys. 32, 89-106 242
1923, 1
*- Differentialinvarianten in der Geometrie. Enz. der Math. Wiss. III, Teil 3,
D 11, 108pp. . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
 Bibliography.

1936, 1
BERWALD, L.: On the projective geometryof paths. Ann. of Math. 37, 8

1897. 1
BERZOLARI, L.: Sulla curvature delle varieta sopra una varieta qualunque. Ace.
Accad. Torino 33. 693-700, 7 59-778 . . . . . . . . . . . . . . . . . 271
1888, 1
BrANCHI, L.: Sulle forme differenziali quadratiche indefinite. Mem. Ace. Linc.
(4), 5. 539-603 . . . . . . . . . . . . . . . . . . . 274
1899, 1
*- Vorlesungen ber Differentialgeometrie. Autorisierte deutsche berset-
zung von M. LuKAT, 1. Aufl. Teubner, Leipzig. 659 pp. . . . . . . . . 296
1902, 1
Sui simboli a quattro indice e sulla curvature di Riemann. Rend. d. Linc.
(V), 11II, 3-7 . . . . . . . . . . . . . . . . . . . . . . . . . 146,148
1902, 2
* Lezioni di geometria differenziale, 2 ed. Spoerri, Pisa I (1902), II (1903)
274,312
1918, 1
*- Lezioni sulla teoria dei gruppi continui finiti di trasformazioni. Spoerri
Pisa. VI+ 590 pp. . . . . . . . . . . . . . . . . 185, 348
1922, 1
Sul parallelismo vincolato di Levi-Civita nella metrica degli spazi curvi.
Rend. Accad. Napoli (3), 28, 150-171 . . . . . . . . . . . . . . 230
1930, 1
*BILIMOVIC, A.: Geometrical principles of the computation with dyads I.:
Dyad and affinor (serbian), Belgrad.
BING, R. H. (see E. F. BECKENBACH).
1923,
*BIRKHOFF, G. D.: Relativity and modern physics. Harvard Univ. Press,
Cambr. Mass. 283 pp. . . . . . . . . . . . . . 155
1920, 1
BLASCHKE, W.: Frenet's Formeln fr den Raum von Riemann. Math. Zeitschr.
6, 94-99 . . . . . . . . . . . . . . . . . . . . . . . . . 230
1921, 1
*- Vorlesungen ber Differentialgeometrie I. Springer, Berlin. 1. Aufl.,
230 pp.. . . . ..... 229
1923, 1
*- Vorlesungen ber Differentialgeometrie und geometrische Grundlagen
von Einsteins Relativittstheorie. II. Affine Differentialgeometrie, bear-
beitet von K. REIDEMEISTER. Springer, Berlin. IX +
259 pp. . . . . . 234
1928, 1
Eine Umkehrung von Knesers Transversalensatz. Nieuw Archief v. Wisk.
(2), 15, 202-204 . . . . . . . . . . . . . . . . 310
1950, 1
* Conferenze di geometria. Geometria di Riemann, geometria affine. La
Edita Universitaria, Messina. 102 pp.
428 l3ibliography.

1950, 2
*BLASCHKE, W.: Geometria diferencial moderna (Modern differential geo-
metry). Conferencias de Matematica, voll. Instituto deMatematicas "Jorge
Juan", Madrid. I+ 43 pp.
1946, 1
BLUM, R.: Anzahl der Identitten von Bianchi in einer Vn. Acad. Roum. Bull.
28, 351-353. . . . . . . . . . . . . . . . . . . . . . . . . . . 147
1946, 2
- ber die Bedingungsgleichungen einer riemannschen Mannigfaltigkeit,
die in einer euklidischen Mannigfaltigkeit eingebettet ist. Bull. Math. Soc.
Roumaine 47, 144-201 . . . . . . . . . . . . . . . . . . 268, 269
1947, 1
Sur les identites de Bianchi et Veblen. C. R. 224, 889-890 144,147,155
1947, 2
Sur les tenseurs derives de Gauss et Codazzi. C. R. 224, 708-710 . . 268, 269
1947, 3
Sur la classe des varietes riemanniennes. Bull. Math. Soc. Roumaine 48,
88-101 . . . . . . . . . . . . . . . . . . . . . . . . 268
1910, 1
*BaCHER, M.: Einfhrung in die hhere Algebra. Teubner, Leipzig. 348 pp. 31
194 7. 1
BocHNER, S.: Curvature in hermitian metric. Bull. Am. Math. Soc. 53, 179
to 195 . . . . . . . . . . . . . . . . . . . . 404, 406, 407, 409, 412
1948, 1
*- and W. T. MARTIN: Several complex variables. Math. Ser. 10. Princeton,
Univ. Press. IX+ 216 pp. . . . . . . . 61
1950, 1
- Vector fields on complex and real manifolds. Ann. of Math. 52, 642-649 66
1951, 1
- Tensor fields with finite basis. Ann. of Math. 53, 400--411 66
1951, 2
- A new viewpoint in differential geometry. Canadian J. Math. 3, 460--470 138
1953. 1
*- Structure of complex spaces. Contributions to the theory of Riemann sur-
faces. Princeton Un. Press; Oxford Un. Press . . . . . . . . . . . . . 397
1950, 1
BoL, G.: Zur tensoriellen Behandlung der projektiven Flchentheorie. Math.
Ann. 122, 279-295 . . . . . . . . . . . . . . . . . . . . . . . 296, 301
1914, 1
BoMPIANI, E.: Forma geometrica delle condizioni per la deformabilita delle
ipersuperficie. Rend. d. Linc. (V), 23I, 126-131 . . . . . . . . . . . 244
1921, 1
Studi sugli spazi curvi. Atti Istituto Veneto (9), 5 (80), 355-386, 839-859,
1113-1145 . . . . . . . . . . . . . . . . . . . . . . .68, 256, 269, 274
1924, 1
Spazi riemanniani luoghi di variet. totalmente geodetiche. Rend. d. Pa-
lermo 48, 121-134 . . . . . . . . . . . . . . . . . . . . . . . . . 269
 Bibliography. 429

1935. 1
BoMPIANI, E.: Geometrie riemanniane di specie superiore. Mem. Akad. ltal. 6,
269-520 . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
1938. 1
- Sulle varieta anolome. Atti d. Linc. (6), 27, 37-45; 45-52. 253
1940, 1
- Le superficie ernisotrope nello spazio euclideo a quattro dimensioni. Atti
Accad. Italia 12, 1-23 . . . . . . . . . . . . . . . . . . . . . . . 227
1941, 1
Intorno alla varieta isotrope. Ann. Mat. Pura Appl. 20, 21-58 227
1943, 1
Geometria proiettiva di elementi differenziali. Ann. Mat. Pura Appl. (4)
22, 1-32 . . . . . . . . . . . . . . . . . . . . . . . . . 296, 301
1945. 1
- Geometria degli spazi a connessione affine. Ann. Mat. Pura Appl. 24, 257
to 282 . . . . . . . . . . . . . . . . . . . . . . . . 123, 155
1946, 1
- Le connessioni tensoriali. Atti d. Linc. 1, 478-482 123,138
1950, 1
- ber das Theorem von F. Schur in der Riemannschen Geometrie. Math.
Zeitschr. 52, 623-626 . . . . . . . . . . . . . . . . . . . 148, 153, 269
1950, 2
ber die Gausehe Krmmung der berraumflchen. Math. Zeitschr. 53,
131-132 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
1950, 3
Sopra una nozione di spazio osculatore ad una varieta introdotta da L. Ber-
zolari. Boll. Un. Mat. Ital. 5, 213-218 . . . . . . . . . . . . . . . . 271
1951, 1
Significato del tensore di torsione di una connessione affine. Boll. Un. Mat.
Ital. 6, 273-276 . . . . . . . . . . . . . . . . . . . . . . . . 127, 269
1951, 2
Geometries riemanniennes d'espece superieure. Coll. Gemn. Diff. Louvain.
1951, 123-156 . . . . . . . . . . . . . . . . . . . . . . . . 269,277
1951, 3
- Proprieta d'immersione di una varieta in uno spazio di Riemann. Rend.
Sem. Mat. Milano 22, 3-26 . . . . . . . . . . . . 256, 269
1952, 1
Sulle connessioni affini non-posizionali. Arch. Math. 3, 183-186 . . 12.1
1952, 2
- Sulle coordinate di Grassmann. Rend. d. Linc. (8) 13, 329-335 . . 23, 254
1952, 1
BoREL, A., and A. LICHNEROWICZ: Espaces riemanniens et hermitiens syme-
triques. C. R. 234, 2332-2334 . . . . . . . . . . . . . . . . . . . . 163
1952, 2
- - Groupes d'holonoinie des varietes riemanniennes. C. R. 234, 1835-1837 370
430 Bibliography.

1927, 1
BoRTOLOTTI, E.: Parallelismi assoluti nelle V,. riemanniane. Atti Veneto 86,
2, 455-465 . . . . . . . . . . . . . . . . . . . . . . . . 142
1927, 2
On metric connexions with absolute parallelism. Proc. Kon. Akad. Amster-
dam 30, 216--218 . . . . . . . . . . . . . . . . . . . . . . . 142

1927, 3
Spazi subordinati: equazioni di Gauss e Codazzi. Boll. Un. Mat. ltal. 6,
134-137 . . . . . . . . . . . . . . . . . . . . . 254,269

1928, 1
Parallelismo assoluto nella varieta a connessione affine e nuove veduti sulla
relativita. Mem. Acad. Bologna (8) 6, 45-58 . . . . . . . . . . . 125,142
1928, 2
- V arieta minime infinitesimale in une Vn riemanniana. Mem. Acad. Bologna
8, 5, 43-48 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
1928, 3
Spostamento geodetico e sue generalizzatione. Giom. M. Battaglini 3, 66,
153-191 . . . . . . . . . . . . . . . . . . . . . . . . . 353

1929, 1
Stelle di congruenze e parallelisme assoluto: basi geometriche di una recente
teoria di Einstein. Rend. d. Linc. (6) 9, 53G-538 . . . . . . . . . . 125, 142

1929, 2
- Sulle coordinate geodetiche lungo una linea I, II. Rend. d. Linc. (6) 10,
486-492; 544-548 . . . . . . . . . . . . . . . . . . . . . . . 155, 166
1929, 3
- Sulle varieta subordinate negli spazi a connessione affine e su di una espres-
sione dei simboli di Riemann. Boll. Un. Mat. ltal. 7, 86--94 . . . . . . 232

1930, 1
- Calcolo assoluto generalizzato di Pascal-Vitali etc. Atti. Soc. ltal. 19, 2,
15-16; Atti Ist. Ven. 90, 471-478. . . . . . . . . . . . 58
1930, 2
Sulle forme differenziali quadratiche spezializzate. Rend. d. Linc. (6) 12,
541-547 . . . . . . . . . . . . . . . . . . . . . 123, 133, 270

1930, 3
Sulla geometria della varieta a connessione affine. Teoria invariantiva delle
trasformazione ehe conservano il parallelismo. Ann. di Mat. 8, 53-101
123,155,287
1930, 4
- Leggi di trasporto nei campi di vettori applicati ai punti di una curva o di
una Vm in V,. riemanniana. Mem. Bologna (8) 7, 11-20 . . . . . . 166, 230

1931, 1
- Trasporti rigidi e geometria della varietl anolonome. Boll. Un. Mat. ltal.
10, 5-12 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78, 253

1931, 2
Sulle varieta subordinate. Rend. Ist. Lombardo (2) 64, 441-463 78, 232, 253, 300
 Bibliography. 431

1931, 3
BoRTOLOTTI, E.: Calcolo assoluto rispetto a una forma differenziale quadratica
specializzata. Rend. d. Linc. (6) 13, 19-25; 104-108 . . . . . . 123
1931, 4
- Differential invariants of direction and point transformations. Ann. of
Math. 32, 361-377 . . . . . . . . . . . . . . . . . . 123, 287, 288, 300
1931, 5
- Directions concourantes et connexions dans les espaces courbes. Bull. Soc.
Math. 59, 70-74 . . . . . . . . . . . . . . . . . . . . . . . 322
1931, 6
- Connessioni proiettive. Boll. Un. Mat. Italia 9 (1930), 288-294; 10, 28-34;
83-90 . . . . . . . . . . . . . . . . . . . . . . . 301
1932, 1
Sulle connessioni proiettive. Rend. d. Palermo 56, 1-57 . . 78, 301, 302, 303
1933, 1
- Connessione nella varieta luoga di spazi; applicazione alla geometrla me-
trica differenziale della congruenze di rette. Rend. Univ. Cagliari 3, 11 pp. 78
1933, 2
- Spazi proiettivamente piani. Ann. d. Bol. 4, 11, 111-134 301
1934, 1
- Riferimenti geodetici lungo piu linee nelle varieta a connessione affine.
Rend. d. Linc. 19, 556-560 . . . . . . . . . . . . . . . . . 166
1935, 1
- Trasporti di specie superiore. Rend. d. Linc. 21, 225-230 123
1936, 1
- Trasporti non lineare: geometria di un sistema di equazioni alle derivate
parziale del2 8 ordine. Rend. d. Linc. 22, 16-21; 104-110; 176-180 . . . 123
1936, 2
- and V. HLAVATY: Contributi alla teoria delle connessioni I. Ann. di Mat.
(4) 15 (1936/37), 1-71 . . . . . . . . . . . . . . . . . . . 123,157,301
1936, 3
Superficie anolonome complimentari. Ser. mat. off. a. L. Berzolari 553-576 253
1937, 1
- La methode de G. Vitali; dans la recherche geometrique; representation
fonctionelle et calcul absolu generalise. Abh. Sem. Vector u. Tensor. Moskau
4, 269-288 . . . . . . . . . . . . . . . . . . . . . . . . . 58
1937, 2
Geometria proiettiva differenziale delle superficie anolonome. Atti
1 congr. Un. Mat. Italia 305-311 . . . . .78,253
1937, 3
- Varieta subordinata: connessione di specie superiore, il problema dell
immersione. Atti 18 congr. Un. Mat. Italia 312-317 . . . . . . . . . 78, 123
1939, 1
- Contributi alla teoria della connessioni. II. Connessioni di specie superiore,
fondamenti analitici calcolo del Vitali generallizzato. Mem. Ist. Lombardo
Mat. Nat. 24, 1-39 . . . . . . . . . . . . . . . . . . . . . . . . 123
432 Bibliography.

1940, 1
BoRTOLOTTI, E.: Sulla geometria proiettiva differenziale di una superficie
anolonoma nello spazio ordinario. Ann. Mat. Pura Appl. (4) 19, 315-325 78
1940, 2
- Coordinate normali ed "estensioni" nella geometria degli spazi a connes-
sione lineare. Atti Accad. Italia (7) 2, 106-116 . . . . . . . . . . . . 155
1941, 1
- Duale Verwandtschaften, anholonome Flchen im projektiven und im af-,
finen Raume. J. D. M. V. 51, 151-169 . . . . . . . . . . . . . . . . 253
1941, 2
- Varieta a connessione proiettiva e loro immagini tangenziali. Atti Accad.
Italia (7) 2, 251-267 . . . . . . . . . . . . . . . . . . . . . . 296, 301
1942, 1
- Vedute e problemi della teoria delle connessione. Univ. Roma e Ist. Naz.
Alta Mat. 3. 241-281. . . . . . . . . . . . . . . . . . . . 123
1924, 1
*BOULIGAND, G.: Lec;ons de geometrie vectorielle, preliminaires a l'etude
de la theorie d'Einstein. Vuibert, Paris. XIII + 365 pp.
1949, 1
*- Les principes de l'analyse geometrique I. Lec;ons de geometrie vectorielle,
preliminaires a l'etude de la theorie d'Einstein. Vuibert, Paris. X+ 436pp.
1947,
*BRAND, L.: Vector and tensor analysis. Wiley a. Sons, New York; Chap-
man a. Hall, London. XVI + 439 pp.
1935.
BRAUER, A., and H. WEYL: Spinars in n dimensions. Am. Journ. of Math.
57. 425-449 . . . . . . . . . . . . . . . . . . . . . . . . 50
1951, 1
BRAUNER, H.: Orthogonalsysteme von riemannschen Hyperflchen der
Klasse I. Anz. ster. Akad. 1951, 29-36 . . . . . . . . . . . . . 249,269

1946, 1
*BRILLOUIN, L.: Les tenseurs en mecanique et en elasticite. Dover Publications,
New York. XI +364 pp.
1923, 1
BRINKMAN, H. W.: On Riemann spaces conformal to Einstein spaces. Proc.
Nat. Acad. 9, 172-174 . . . . . . . . . . . . . . . . . . . . . 312, 314
1923, 2
On Riemann spaces conformal to euclidean spaces. Proc. Nat. Acad. 9, 1-3 315
1924, 1
Riemann spaces conformal to Einstein spaces. Math. Ann. 91, 269-278 312, 314
1925, 1
- Einstein spaces which are mapped conformally on each other. Math. Ann.
94. 119-145 . . . . . . . . . . . . . . . . . . . . . 312, 314, 330
1906, 1
BROUWER, L. E. J.: Polydimensional vectordistributions. Proc. Kon. Ned.
Akad. Amsterdam 9, 66-78 . . . . . . . . . . . . . . . . . . . . 85, 97
 Bibliography. 433

1906, 2
BROUWER, L. E. J.: The force field of the non-euclidean spaces with negative
curvature. Proc. Kon. Ned. Akad. Amsterdam 9, 116--133 . . . . . . 97,122
1919, 1
- Opmerkingen over meervoudige integralen. Vers!. Kon. Ned. Akad. Am-
sterdam 28, 116--120 . . . . . . . . . . . . . . . . . . . . . . . . 97
1943, 1
BRUCK, R. H., and T. L. WADE: The number of independent components of
the tensors of given symmetry type. Bull. Am. Math. Soc. 49, 47.-472 15
1944, 1
- The number of absolute invariants of a tensor. .Am. J. Math. 66, 411-424 15
1948, 1
BUCHDAHL, H. A.: The hamiltonian derivatives- of a class of fundamental in-
variants. Quart J oum. of Math. Oxford 19, 15G-159 . . . . . . . . . 114
1951, 1
- ber die Variationsableitung von Fundamentalinvarianten beliebig hoher
Ordnung. Acta Math. 85, 63-72 . . . . . . . . . . . . . . . . . . 114
1951, 2
- On the hamiltonian derivatives arising from a class of gauge-invariant
action principles in a Wn. Joum. London Math. Soc. 26, 139-149 . . . 114
1951, 3
- An identity between the hamiltonian derivatives of certain fundamental
invariants in a ur,.. Joum. London Math. Soc. 26, 15G-152 . . . . 114,116

1899. 1
*BucHHOLZ, A.: Ein Beitrag zur Mannigfaltigkeitslehre. Mannigfaltigkeiten,
deren Linienelemente auf die Form:
ds = f(I:Xl) VI:dXi
gebracht werden knnen. Cohen, Bonn. 264 pp. . . . . . . . . . . . 330

1926, 1
BuRSTIN, C., and W. MAYER: Das Formenproblem der I-dimensionalen Hyper-
flchen in n-dimensionalen Rumen konstanter Krmmung. Mon. f. Math.
u. Phys. 24, 89--136 . . . . . . . . . . . . . . . . . . . . . . . . 277
1929, 1
- Beitrge zur mehrdimensionalen Differentialgeometrie. Mon. f. Math. u.
Phys. 36, 97-130 . . . . . . . . . . . . . . . . . . . . . . . . . 277
1931, 1
- Ein Beitrag zum Problem der Einbettung der Riemannschen Rume in
euklidischen Rumen. Receuil. Math. Soc. math. (Math. Sbomik) Moscou
38, 74-85 . . . . . . . . . . . . . . . . . . . . . . . . . . 268, 352
1931, 2
- Beitrge zum Problem der Verbiegung von Hyperflchen in euklidischen
Rumen. Rec. Math. Soc. Math. (Math. Sbomik) Moscou 38, 86-93 . . 352

1932, 1
- Zum Einbettungsproblem. Commun. Soc. Math. Kharkov et Inst. Sei.
math. Ukraine (4) 5, 87-95 . . . . . . . . . . . . . . . . 268
Schonten, Ricci-Ca!culus, 2. Auf!. 28
434 Bibliography.

1937, 1
BuRSTIN, C.: Ein Beitrag zur Theorie des Klassenbegriffes der quadratischen
Differentialformen. Abh. Sem. Vektor u. Tensor. Moskau 4, 121-136 . . 268
1953. 1
CALABRI, L.: Isometrie imbedding of complex manifolds. Ann. of Math. 58,
3-23 . . . . . . . . . . . . . . . . . . . . . 412
1903, 1
*CAMPBELL, J. E.: Introdction treatise on Lie's theory of finite continuous
groups. OKford . . . . . . . . . . . . . . . . . . . . . . . . . . 203
1894, 1
*CARTAN, E.: Sur Ia structure des groupes de transformations finis et continus.
These, Nony, Paris; second edition Vuibert, Paris 1933 . 117
1899, 1
Sur certaines expressions differentielles et Je problf~me de Pfaff. Ann .
.Ecole norm. sup. (III) 16, 239-332. . . . . . . . . . . . 95
1901, 1
Sur l'integration des systemes d'equations aux differentielles totales. Ann.
Ecole norm. 18, 241-311 . . . . . . . . . . . . . . . . . . . 117
1916, 1
La deformation des hypersurfaces dans l'espace euclidien a n dimensions.
Bull. Soc. Math. 44, 65-99 . . . . . . . . . . . . . . . . . . . 244, 352
1917, 1
- Ladeformation des hypersurfaces dans l'espace conforme del an :2:5 dimen-
sions. Bull. Soc. Math 45, 57-121 . . . . . . . . . . . 352
1919, 1
Sur !es varietes de courbure constante d'un espace euclidien ou non
euclidien. Bull. Soc. Math. d. France 47, 125-160 . . . . . . . . . . 39, 269
1920, 1
Sur Ia deformation projective des surfaces. C. R. 170, 1439; 171, 27; Ann.
Ec. Norm. 37. 259-356 . . . . . . . . . . . . . . . . . . . . . . 352
1920, 2
Sur le problerne general de la deformation. C. R. Congres Strasbourg 397
to 406 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
1920, 3
Sur les varietes de courbure constante d'un espace euclidien ou non eucli-
dien. Bull. Soc. Math. d. France 48, 132-208 . . . . . . . . 269
1922, 1
*- Le<;ons sur les invariants integraux. Hermann, Paris. X + 210 pp. . . 89
1922, 2
Sur une generalisation de la notion de courbure de Riemann et !es espaces
a torsion. c. R. 174, 593-595 . . . . . . . . . . . . . . . . . . . . 127
1922, 3
*- Sur les equations de la gravitation d'Einstein. Journ. d. math. p. et a.
(9) 1, 141-203. Separate: Gauthier-Villars, Paris. 65 pp. . . . . . . . 306
1923, 1
Sur Im theoreme fondamental deM. H. Weyl. J ourn. d. Math. p. et a. (IX),
2, 167-192; cf. also C. R. 175, 82 85 . . . . . . . . . . . . . . . . . 137
 Bibliography. 435

1923, 2
CARTAN, E.: Sur les varietes a connexion affine et la theorie de la relativite
generalisee. Ann. :Ec. norm. 40, 325-412 . . . . . . . . . . . . . . . 123

1923, 3
- Les espaces a connexion conforme. Ann. Soc. polon. d. math. 2, 171-221
315,316
1924, 1
- Sur les varilUs a connexion projective. Bull. d. I. Soc. Mat. d. France 52,
205-241 . . . . . . . . . . . . . . . . . . . . . . . 123,302,303,316
1924, 2
- La theorie des groupes et les recherches recentes de geometrie differentielle.
Ens. Math. 1-18 . . . . . . . . . . . . . . . . . . . . . . . . . . 361
1925, 1
- Sur les varietes a connexion affine et la theorie de la relative generalisee.
Ann. Ec. norm. 42, 1 7-88 . . . . . . . . . . . . . . . . . . . 361
1925, 2
*- La geometrie des espaces de Riemann. Mem. d. Sc. Math., fase. 9.

1926, 1
- Sur une classe remarquable d'espaces de Riemann. Bull. Soc. Math. 54,
214-264; 55, 114-134 . . . . . . . . . . . . . . . . . . 163,381
1926, 2
- Sur les espaces de Riemann dans lesquels le transport par parallelisme con-
serve la courbure. Rend. d. Linc. (6) 3, 544-547 . . . . . . . . . . . 163

1926, 3
- and J. A. ScHOUTEN: On the geometry of the group manifold of simple and
semi-simple groups. Proc. Kon. Ned. Akad. Amsterdam 29, 803-815
142,187,351
1926, 4
- - On riemannien geometries admitting an absolute parallelism. Proc.
Kon. Ned. Akad. Amsterdam 29, 933-946 . . . . . . . . . . . . 142, 187

1926, 5
- Les groupes d'holonomie des espaces generalises. Acta Math. 48, 1-42 . . 361

1927, 1
- La geometriedes groupes de transformations. Journ. de Math. 6 (9), 1-119
163,187,201,222
1927, 2
- La geometrie des groupes simples. Ann. di Mat. 4, 209--256 . . . . . 142, 185
1927, 3
- Sur les courbes de torsionnulle et les surfaces developpables dans les espaces
de Riemann. C. R. 184, 138-140 . . . . . . . . . . . . . . . . 229, 296
1927, 4
- Sur la possibilite de plonger un espace riemannien donne dans un espace
euclidien. Ann. d. I. Soc. Polon. de Math. 6, 1-7 . . . . . . . . . . . 268
1927, 5
Sur l'ecart geodesique et quelques notions connexes. Rend. d. Linc. (6a)
5 I, 609--613 . . . . . . . . . . . . . . . . . . . . . . 352
28*
436 Bibliography.

1928, 1
*CARTAN, E.: Le~ons sur la geometrie des espaces de Riemann. Gauthier-
villars, Paris. 270 pp. . . . . . . . . . . . . . . . . . . . . . . . 149
1930, 1
- Le troisieme theoreme fondamental de Lie. C. R. 190, 914-916; 1005-1007 208
1930, 2
*- La theoriedes groupes finis et continus et l'analysis situs. Mem. Sc. Math.
42. Gauthier-Villars, Paris. 60 pp. . . . . . . . . . . . . . . 185
1931, 1
*- Le~onssur la geometrie projective complexe. Gauthier-Villars, Paris.
322pp. . .... 406
1932, 1
- Les espaces riemanniens symetriques. Verh. Int. Math. Kongr. Zrich I,
152-161 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
1934, 1
- Les espaces a connexion projective. Mem. du Semin. Geom. Diff. Tens.
Moscou 4, 147-173- . . . . . . . . . . . . . . . . . 301
1935, 1
- Le calcul tensoriel projectif. Rec. Soc. Math. Moscou 42, 131-148 . . . 301

1936, 1
*- La topologiedes groupes de Lie. Act. Sc. et Ind. 358. Hermann, Paris.
28 pp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
1937, 1
*- Le~ons sur la theoriedes espaces a connexion projective. Gauthier-Villars,
Paris. 308 pp.. . . . . . . . . . . . . . . . . . . . . . . 128, 301, 361
1937, 2
*- and J. LERAY: La theoriedes groupes finis et continus. Gauthier-Villars,
Paris. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
1937, 3
- L'extension du calcul tensorial aux geometries non affines. Ann. of Math.
38, 1-13 . . . . . . . . . . . . . . . . . . . . . . . . . . . 301, 315
1938, 1
*- Le~ons sur la theorie des spineurs I 85 pp., II 96 pp. Hermann, Paris . . 21
1939, 1
Sur quelques familles remarquables de hypersurfaces. C. R. Congr. Sc.
Math. Liege 30-41 . . . . . . . . . . . . . . . . . . . . . . . 242, 269
1939, 2
Sur des familles remarquables d'hypersurfaces isoparametriques dans les
espaces spheriques. Math. Zeitschr. 45, 335-367 . . . . . . . . . 242, 269

1941, 1
- La notion d'orientation dans les differentes geometries. Bull. Soc. Math.
France 69, 47-70 . . . . . . . . . . . . . . . . . . . . . . 5
1941, 2
Sur les surfaces admettant une seconde forme fondamentale donnee. C. R.
212, 825-828 . . . . . . . . . . . . . . . . . . . . . . . . 244
 Bibliography. 437

1942, 1
CARTAN, E.: Sur les couples de surfaces applicables avec conservation des cour-
bures principales. Bull. Sc. Math. (2) 66, 55-72, 74-85 . . . . . . . . . 244

1943. 1
- Les surfaces qui admettent une seconde forme fondamentale donnee. Bull.
Sc. Math. (2) 67, 8-32 . . . . . . . . . . . . . . . . . . . . 244
1946, 1
*- Le<;ons sur la geometriedes espaces de Riemann, 2nd t'd. Gauthier-Villars,
Paris. VIII+ 378 pp. Reprinted 1951 . . . . . . . . . . . . 149,175,178
1949, 1
*CARTAN, H.: Algebraic topology. Springer and Pollak, Cambridge, Mass.
VI+169 pp. . . . . . . . . . . . . . . . . 85
1946, 1
CAsTOLDI, L.: Sopra alcune proprieta caratteristiche delle Vn totalmente geo-
detiche rispetto ad una Vn ambiente. Atti Ace. Naz. Linc. 1064-1069 . . 269

1947. 1
- Sopra gli spazi a connessione semimetrica. Univ. Roma Ist. Naz. Alta.
Mat. 6, 395-409 . . . . . . . . . . . . . . . . . . . . . . . . . . 145
1948, 1
Giaciture associate ad una serie di direzioni lungo una curva immersa in uno
spazio a connessione affine. Atti Ace. Sc. Torino 81-82; 205-214 . . 232
1948, 2
- Dislocazioni e deformazioni finite nella varieta sostanziali di Riemann. Ist.
Naz. AltaMat. 7. 373-392 . . . . . . . . . . . . . . . . . . . . 352,353
1949. 1
Vm "canoniche" in V" e loro caratterizzazione intrinseca. Ist. Lombardo
Sc. Mat. 13, 267-279 . . . . . . . . . . . . . . . . . . . . . . . 269
1890, 1
CAYLEY, A.: Sur les surfaces minima. C. R. 111, 953-954; Coll. Papers XIII,
41-42 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
1925, 1
*CERDANZEV, J. A.: Grundlagen der Vektor- und Tensoranalysis. Moskau,
Staatsverlag. 143 pp. (russian).
1937. 1
CHERN, S. S.: Sur la possibilite de plonger un espace a connexion projective
donne dans un espace projectif. Bull. Sc. Math. (2) 61, 234-243 . . 268
1942, 1
- The geometry of isotropic surfaces. Ann. of Math. (2) 43, 545-559 . 228
1944, 1
On a theorem of algebra and its geometrical application. J. Indian Math.
Soc. (N. S.) 8, 29-36 . . . . . . . . . . . . . . . . . . . . . . 269
1947, 1
- and H. WANG: Differential geometry in symplectic space I. Sei. Rep. Nat.
Tsing Hua Univ. 4, 453-477 . . . . . . . . . . . . . . . . . . . . 179
1946, 1
*CHEVALLEY, C.: Theory of Lie groups I. Princeton Math. Ser. vol. 8. Princeton
Univ. Press. IX+ 217 pp. . . . . . . . . . . . . . . . . . . . . . 185
438 Bibliography.

1869, 1
CHRISTOFFEL, E. B.: ber die Transformation der homogenen Differential-
ausdrcke zweiten Grades. Crelle's Journal 70, 46-70; Ges. math. Abh. I,
352-377 0 0 0 0 0 132, 164
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1932, 1
CHURCHIL, R. V.: On the geometry of the Riemann tensor. Trans. Am. Math.
Soc. 34, 126-152 . . . . . . . . . . . . . . . . . . . . . . . . . 146
1932, 1
*CrsoTTI, U.: Cenni sui fondamenti del calcolo tensoriale con applicazioni alla
teoria dell'elasticita. Milano, Hoepli. 46 pp.

1940,
- Una notevole scomposizione dei tensori. Boll. Un. Mat. Ital. (2) 2, 210-215 15
1941, 1
- Invarianti quadratici dei tensori. Atti Accad. Italia 2, 511-516 . . . . 15
1943, 1
- Invarianti ed eminvarianti cubici dei tensori. Atti Accad. Italia 4, 8-11 15
1944, 1
- Invarianti di vettori e di tensori doppi. Ist. Lombarde. Sc. Mat. Nat.
Rend. 8, 253-258 . . . . . . . . . . . . . . . . . . . . . . . . . 15
1939, 1
COBURN, N.: v", in Sn with planar points (m ~ 3). Bull. Am. Math. Soc. 45,
774-783 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 269, 272
1940, 1
Generalized Einstein hypersurfaces of spaces of constant curvature. J.
Math. Phys. Mass. Inst. Tech. 19, 140-152 . . . . . . . . 242,252,269

1940, 2
- A characterization of Schouten's and Haydens' deformation methods. J.
London Math. Soc. 15, 123-136 . . . . . . . . . . . . . . . . . . . 335
1941, 1
- Conformal unitary spaces. Trans. Am. Math. Soc. 50, 26-39 288, 415, 416, 417

1941, 2
- A note on conformal geometry. Proc. Nat. Acad. Sei. U.S.A. 27, 57-60 315

1941, 3
- Unitary curves in unitary space. Univ. Nac. Tucuman Revista A 2, 159-167
413,414,415
1942, 1
Semi-analytic unitary subspaces of unitary space. Am. J. Math. 64, 714-724 407

1942, 2
- Frenet formulae for curves in unitary space. J. Math. Phys. Mass. Inst.
Tech. 21, 10-18 . . . . . . . . . . . . . . . . . . . . . . . . . . 415
1942, 3
- Conformal geometry of unitary space. Univ. Nac. Tucuman Revista A 3,
125-140 0 0 0 0 416, 417, 418
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1943, 1
- Congruences in unitary space. Trans. Am. Math. Soc. 53, 25-40 . . . 415, 418
 Bibliography. 439

1931. 1
CoNFORTO, F.: Parallelismo negli spazi funzionali continui. Rend. d. Linc. (6)
13, 173-178 . . . . . . . . . . . . . . . . . . . . . . 123
1931, 2
Formalismo matematico in uno spazio funzionale continuo retto da un
elemento lineare di seconda spezie. Rend. d. Linc. (6) 13, 562-568 . . 123
1928, 1
CoNNELL, A. J. Jll c.: Il trasporlo parallelo di un vettore lungo un circuito finito.
Rend. d. Linc. (6) 7, 208-213; 306-309 . . . . . . . . . . . . . . 138
1928, 2
- The torsion of riemannian space. Atti Congr. lnt. Mat. Bologna 6, 321-338 126
1928, 3
- Strain and torsion in ri~mannian space. Ann. di Mat. (4) 6, 207-231 . . 126
1928, 4
- The derived directions of a vector defined along a curve in riemannian space.
Proc. London Math. Soc. (2) 28, 501-509 . . . . . . . . . . . 230
1928, 5
Schmidt's orthogonal ennuple and the Frenet formulas for a ..curve. Bull.
Am. Math. Soc. 713-714 . . . . . . . . . . . . . . . . . . . 230
1929. 1
The variation of curvatures in the deformation of a curve in riemannian
space. Proc. Roy. Ir. Acad. 39 A, 1-9 . . . . . . . . . . . . . . 335, 352

1931, 1
* Applications of the absolute differential calculus. Blackie and Son, Lon-
don. 330 pp.
1924, 1
*CooLIDGE, J. L.: The geometry of the complex domain. Clarendon Press, Ox-
ford. 242 pp. . . . . . . . . . . . . . . . . . . . . . . . . . . . 406

1940, 1
CoPSON, E. T., and M. S. RusE: Harmonie riemannian spaces. Proc. Roy. Soc.
Edinburgh 60, 117-133 . . . . . . . . . . . . . . . . . . . . 381,382,385

1947, 1
Cossu, A.: Aleune osservazioni sulle varieta Subordinate di una varieta a con-
nessione affine asimmetrica. Atti Ace. Naz. Linc. (8) 3, 303-311 . . . . 232

1949. 1
Proprieta di curvatura di una particolare classe di varieta a connessione af-
fine. Atti Ace. Naz. Linc. (6) 702-707 . . . . . . . . . . . . . . 144

1899.
*CoTTON, E.: Sur les varietes a trois dimensions. Ann. Fac. d. Sc. Toulouse (II)
1, 385-438; also These Paris. 54 pp. . . . . . . . . . . . . . . . 306, 316

1940, 1
Sur le calcul des invariants diffrentiels euclidiens d'une surface. J. Math.
Pures et Appl. (9) 19, 211-220 . . . . . . . . . . . . . . 242
1948, 1
Sur un systeme de bivecteurs associe a un cycle. C. R. 226, 1564-1 566 232
440 Bibliography.

1943, 1
*CRAIG, H. V.: Vector and tensor analysis. McGraw-Hill, New York. XIV +
434 pp.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57, 58
1931, 1
*CUTLER, E. H.: On the curvature of a curve in Riemann space. Trans. Am.
Math. Soc. 33, 832-838. Thesis Harvard 1930 . . . . . . . . . . . . . 279

1931, 2
Frenet formulas for a general subspace of a Riemann space. Trans. Am.
Math. Soc. 33, 839-850. . . . . . . . . . . . . . . . . . . . . . . 279

1949, 1
CzEcH, E.: Geometrie projective differentielle des correspondances entre deux
espaces I. Casopis Pest. Mat. Fys. 74, 32-48 (french, czech.) . . . . . . 301

1950, 1, 2
Geometrie projective differentielle des correspondances entre deux espaces
II, III. Casopis Pellt. Mat. Fys. 7 5, 123-136; 13 7-158. (french, czech.) 301

1932, 1, 2
DANTZIG, D. VAN (see also J. A. ScHOUTEN}: Zur allgemeinen projektiven
Differentialgeometrie I, II. Proc. Kon. Akad. Amsterdam 35, 524-534;
535-542 . . . . . . . . . . . . . . . . . . . . . . . . . 105, 301, 303
1932, 3
- Theorie des projektiven Zusammenhangs n-dimensionaler Rume. Math.
Ann. 106, 40-454 . . . . . . . . . . . . . . . . . . . . . 301, 303, 327
1934, 1
- On the general projective differential geometry III. Proc. Kon. Acad. Am-
sterdam 37, 15-155 . . . . . . . . . . . . . . . . . . . . . . 301, 303
1889, 1
DARBoux, G.: Lec;ons sur Ia theorie generale des surfaces I I. Gauthier-Villars,
Paris 522 pp. . . . . . . . . . . . . . . . . . . . . . . . . . . 245
1933, 1
DAVIES. E. T. (see also P. DIENES}: On the infinitesimal deformations of a
space. Ann. di Mat. (4) 12, 145-151 . . . . . . . . . . . . . . . . 335
1935, 1
- On (r, r) Subordination of a subspace in a Riemann space Vn. J oum.
London Math. Soc 10, 216-232 . . . . . . . . . . . . . 269
1936, 1
On the deformation of a subspace. Journ. London Math. Soc. 11, 295-301
352,353
1937, 1
On the second and third fundamental forms of a subspace. Joum. London
Math. Soc. 12, 290-295. . . . . . . . . . . . . . . . . . . 335

1938, 1
- Analogues of the French formulae determined by deformation operators
Joum. London Math. Soc. 13, 21-216; 320 . . . . . . . 110, 230, 335, 352

1938, 2
- On the deformation of the tangent m-plane of a Vnm Proc. Edinburgh.
Math. Soc. (2) 5, 202-206. . . . . . . . . . . . . . . . . . . . . . 352
 Bibliography. 441

1942, 1
DAVIES, E. T.: The first and second variations of the volume integral in
riemannian space. Quart. J. Math. Oxford Ser. 13, 58-64 352
1953.1
- On the invariant theory of contact transformations. Math. Zeitschr. 57.
415-427 . . . . . . . . . . . . . . . . . . . . 253
1950, 1
DEBEVER, R.: Le groupe d'holonomie des varietesde groupe de Lie a connexion
conforme normale. Congr. Nat. des Sc. Bruxelles 111, 56-58 . . . . . . . 367
1950, 1
*DELACHET, A.: Calcul vectoriel et calcul tensoriel. Presses Universit. de
France, Paris. 128 pp.
1913, 1
DEMOULIN, A.: Sur une propriete caracteristique des familles de Lame. C. R.
157, 105G--1053 . . . . . . . . . . . . . . . . . . . . . . . 249
1926, 1
*DICKSON, L. E.: Modem algebraic theories. Chicago (also german by E. Bode-
wig, Leipzig 1929) . . . . . . . . . . . . . . . . . . . . 31, 34, 46, 53,56
1932, 1
DIENES, P.: On the fundamental formulae of the geometry of tensor submani-
folds. Joum. d. Math. (9) 11, 255-282 . . . . . . . . . 78,232,253,263,278
1933. 1
- Sur le deplacement d'un n-uple et sur une interpretation nouvelle des
coefficients de rotation de Ricci. Rend. d. Linc. (6) 17, 119-122 166, 171
1933. 2
Sur un theoreme deM. Fermi. Rend. d. Linc. (6) 17, 369-372 . . 171
1933. 3
Sur la deformation des espaces a connexion lineaire generale. C. R. 197,
1084-1086 . . . . . . . . . . . . . . . . . . . . . . . 335
1933. 4
Sur la deformation des sous-spaces dans un espace a connexion lineaire
generale. c. R. 197, 1167-1169 . . . . . . . 352
1934, 1
- On curves in a space of generallinear connection. J oum. London Math. Soc.
9. 259-266 . . . . . . . . . . . . . . . . . . . . . 128, 131, 232
1934, 2
On the deformation of tensor manifolds. Proc. London Math. Soc. 37,
512-519 . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
1937, 1
- and E. T. DAVIES: On the infinitesimal deformations of tensor submanifolds.
J ourn. de Math. 24, 111-150 . . . . . . . . . . . . . . . . . . . . 352
1947. 1
*DIRAC, P. A. M.: The principles of quantum mechanics, 3rd ed. Oxford, Cla-
rendon Press. 311 pp. . . . . . . . . . . . . . . . . . . . . . . . 54
1942, 1
DoNDER, T. DE, and J. GEHENIAU: Sur la derivee covariante des tenseurs ge-
neralises. Acad. Roy. Belgique Bull. 28, 630-633 . . . . . . . . . . . 105
442 Bibliography.

1924. 1
DouGLAS, J.: Normal congruences and quadruply infinite systems of curves in
space. Trans. Am. Math. Soc. 26, 68-100. . . . . . . . . 310
1925. 1
- A criterion for the conformal equivalence of a riemannian space to a
euclidean space. Trans. Am. Math. Soc. 27, 299-306 . . . . 306
1928, 1
- The general geometry of paths. Ann. of Math. (2) 29, 143-168 303
1931. 1
- Systems of K-dimensional manifolds in an N-dimensional space. Math.
Ann. 105, 707-733 . . . . . . . . . . . . . . . . . . . . . . . 155, 303
1941, 1
DoYLE, T. C.: Tensor decomposition with applications to the contact and com-
plex .groups. Ann. of Math. (2) 42, 698-722 . . . . . . . . . . 15
1931, 1
*DROST, A. J.: Reeksontwikkelingen van algebraische vormen. Thesis.
Groningen. 68 pp. . . . . . . . . . . . . . . . . . . . . 15
1941, 1
DuBNOV, Y. S.: Complete system of invariants of two affinors in centro-affine
space of two or three dimensions. Abh. Sem. Vektor u. Tensor. 5. 25D-270
(russian) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1927. 1
DuBoURDIEU, J.: Sopra le coordinate cartesiane lungo una curva. Rend. d.
Linc., Rcma (5) 6, 654-655 . . . . . . . . . . . . . . . . . . . . . 155
1928, 1
Quelques applications de la theorie des coordonnees geodesiques le long
d'une courbe. Rend. d. Linc. (6) 7, 707-709 . . . . . . . . . . . . 155
DuPARE, H. J. A. (see W. PEREMANS).
1930,
*DuscHEK, A., and W. MAYER: Lehrbuch der Differentialgeometrie II: Rie-
mannsche Geometrie. Teubner, Leipzig und Berlin. 245 pp.
68,230,242,252,272,277,285
1930, 2
- Parallelverschiebung in allgemeinen Rumen. J. D. M. V. 39. 42-44 . . 123
1935. 1
- Das Variationsproblem der Fm im Riemannschen R 70 und eine Verallgemei-
nerung des Gau-Bormetsehen Satzes. Math. Zeitschr. 40, 279-291 . . 269
1948, 1
*- and A. HocHRAINER: Grundzge der Tensorrechnung in analytischer
Darstellung I. Springer, Wien. II 1950; III in printing.
1949, 1
EcKMANN, B., and H. GUGGENHEIMER: Formes differentielles et metrique her-
mitienne sans torsion. I. Structure complexe, formes pures. C. R. 229,
464-466 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397
1949. 2
and H. GUGGENHEIMER: Formes differentielles et metrique hermitienne
sans torsion. II. Formes de classe K, formes analytique. C. R. 229, 489
to 491 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397
 Bibliography. 443

1947. 1
EGOROV, I. P.: On the order of the group of motions of spaces with affine
connection. Akad. Nauk. SSSR 57, 867-870 (russian) . . . . . . . 346, 348
1948, 1
- On collineations in spaces with projective connection. Akad. Nauk. SSSR
61, 605-608 . . . . . . . . . . . . . . . . . . . . . . . . . 346, 348
1949. 1
On a strengthening of Fubini's theorem on the order of the group of mo-
tions of a riemannian space. Akad. Nauk. SSSR 66, 793-796 (russian) 346, 348
1949. 2
- On the groups of motion of spaces with asymmetric affine connection.
Akad. Nauk. SSSR. 64, 621-624 (russian) . . . . . . . . . . . 346, 348
1950. 1
- On groups of motion of spaces with general asymmetrical affine connec-
tion. Akad. Nauk. SSSR. 73, 265-267 (russian) . . . . . . . . . . 346, 348
1951. 1
Collineations of projectively connected spaces. Akad. Nauk SSSR. 80, 709
to 712 (russian) . . . . . . . . . . . . . . . . . . . . . . . . . . 348
1952, 1
A tensor characteristic of An of nonzero curvature with maximum mobility.
Akad. Nauk SSSR. 84, 209-212 (russian) . . . . . . . . . 348
1952. 2
- Maximally mobile Ln with a semi-symmetric connection. Akad. Nauk
SSSR. 84, 433-435 (russian) . . . . . . . . . . . . . . . . . 348
1950, 1
EHRESMANN, C.: Les connexions infinitesimales dans un espace fibre differen-
tiable. Colloque de Topologie Bruxelles 29-55 . . . . . . 123
1952. 1
- Structures locales et structures infinitesimales. C. R. 234, 587-589 67
1952, 2
- Les prolongements d'une variete differentiable V. Covariants differentiels
et prolongements d'une structure infinitesimale. C. R. 234, 1424-1425 123
1928, 1
EINSTEIN, A.: Riemann Geometrie mit Aufrechterhaltung des Begriffes des
Femparallelismus. Sitz. Preu. Akad. 217-221 . . . . . . . . . . . . 142
1945. 1
- A generalization of the relativistic theory of gravitation. Ann. of Math.
(2) 46, 578-584 . . . . . . . . . . . . . . . . . . . . . . . . . . 179
1946, 1
and E. G. STRAUS: A generalization of the relativistic theory of gravita-
tion II. Ann. of Math. 47. 731-741 . . . . . . . . . . . . 179
1948, 1
A generalized theory of gravitation. Rev. Modern Physics 20, 35-39 179
1950, 1
The Bianchi identities in the generalized theory of gravitation. Canadian
J. Math. 2, 120-128 . . . . . . . . . . . . . . . . . . . . . . . . 179
444 Bibliography.

1922, 1
EISENHART, L. P.: Ricci's principal directions for a Riemann space and the
Einstein theory. Proc. Nat. Acad. 8, 24-26 . . . . . . . . . . . . . . 146
1922, 2
- Spaces with corresponding paths. Proc. Nat. Acad. 8, 233-238 288
1922, 3
- and 0. VEBLEN: The Riemann geometry and its generalisation. Proc. Nat.
Acad. 8, 19-23 . . . . . . . . . . . . . . . . . . . . . . . . . . 303
1922, 4
- Fields of parallel vectors in the geometry of paths. Proc. Nat. Acad. 8,
207-212 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
1923, 1
- Symmetrie tensors of the seconcf order whose first covariant derivatives are
zero. Trans. Am. Math. Soc. 25, 297-306 . . . . . . . . . . . . . . 285
1923, 2
- Orthogonalsystems of hypersurfaces in a general Riemann space. Trans.
Am. Math. Soc. 25, 259-280 . . . . . . . . . . . . . . . . . . . . 246
1923, 3
- Affine geometry of paths possessing an invariant integral. Proc. Nat. Acad.
9, 4-7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
1924, 1
- Geometries of paths for which the equations of the paths admit a quadratic
first integral. Trans. Am. Math. Soc. 26, 378-384 . . . . . . . . . . . 293
1925, 1
- Fields of parallel vectors in a riemannian geometry. Trans. Am. Math.
Soc. 27, 563-573 . . . . . . . . . . . . . . . . . . . . . 322
1926, 1
*- Riemannian geometry. Princeton Univ. Press. 262 pp.
22,156,227,229,230,242,244,245,246,263,271,278,294,307,322,348,349
1927, 1
*- Non riemannian geometry. Am. Math. Soc. Coll. Publ. VIII. 184 pp.
123,155,166,238,289,293,296
1927, 2
- and M. S. KNEBELMAN: Displacements in a geometry of paths which carry
paths into paths. Proc. Nat. Acad. 13, 38-42 . . . . . . . 301
1930, 1
- Projective normal Coordinates. Proc. Nat. Acad. 16, 731-740 . 155,301
1933, 1
*- Continuous groups of transformations. Princeton Univ. Press. 302 pp.
185,215,226,349,350,351
1935, 1
Groups of motions and Ricci directions. Ann. of Math. 36, 823-832 . 346
1936, 1
Simply transitive groups of motions. Mon. Math. Phys. 43, 448-462 346
1937, 1
- Riemannian spaces of class greater than unity. Ann. of Math. 38, 794-808
268,281,282,283
 Bibliography. 445

1938. 1
EISENHART, L. P.: Parallel vectors in riemannian space. Ann. of Math. 39.
316-321 . . . . . . . . . . . . . . . . . . . . . . . . . . . 322, 422
1940, 1
*- An introduction to differential geometry. Princeton Math. Series, V, 3.
Princeton Univ. Press. X + 304 pp.
1949. 1
*- Riema:tm geometry, 2nd printing. Princeton Univ. Press. VII + 306 pp.
153,246,349
1951. 1
Generalized Riemann spaces ~~ Proc. Nat. Acad. 37, 311-315 . . . . 179
1952. 1
- Generalized Riemann spaces II. Proc. Nat. Acad. 38, 505-508 . . . . . 179
ENGEL, F. (see S. LIE).
1934, 1
*FABRICIUS BJERRE, FR.: Differentialgeometrische Untersuchungen, torsions-
freie Flchen in Rumen konstanter Krmmung. Thesis, Kopenhagen.
88 pp. (dansk.) . . . . . . . . . . . . . . . . . . . . . . . . 261, 269
1936. 1
Sur les varietes a torsion nulle. Acta Math. 66, 49-77 253,261,269
1922, 1
FERMI, E.: Sopra i fenomeni ehe avvengono in vicinanza di una linea oraria.
Rend. d. Linc. I, 21-23, 51-52 . . . . . . . . . . . . . . . . . . . 166
1950, 1
H. E. FETTIS: A method for obtaining the characteristic equation of a matrix
and computing the asso.ciated modal columns. Quart. Appl. Math. 8,
206-212 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1938. 1
FrALKOW, A.: Einstein spaces in a space of constant curvature. Proc. Nat.
Acad. 24, 3-34. . . . . . . . . . . . . . . . . . . . . . 148
1938, 2
- The riemannian curvature of a hypersurface. Bull. Ann. Math. Soc. 44.
253-257 . . . . . . . . . . . . . . . . . . . . . . . . . . . 242, 388
1938. 3
Hypersudaces of a space of constant curvature. Ann. of Math. Princeton
(2) 39. 762-785 . . . . . . . . . . . . . . . . . . . . . . . . . . 242
1939. 1
- Totally geodesie Einstein spaces. Bull. Am. Math. Soc. 45, 423-428 148, 312, 315

1939. 2
- Conformal geodesics. Trans. Am. Math. Soc. 45, 443-473 .. 310, 315, 330, 332
1940, 1
- The conformal theory of curves. Proc. Nat. Acad. 26, 437-439 . 310, 315, 330

1942, 1
- Correction to "Totally geodesie Einstein spaces ". Bull. Am. Math. Soc.
48, 167-168 . . . . . . . . . . . . . . . . . . . . . . . 148, 312, 315
446 Bibliography.

1942, 2
FIALKOW, A.: The conformal theory of curves. Trans. Am. Math. Soc. 51,
435-501 . . . . . . . . . . . . . . . . . . . . . . . 310, 315, 330
1944, 1
Conformal differential geometry of a subspace. Trans. Am. Math. Soc. 56,
309-433 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
1939, 1
FICKEN, F. A.: The riemannian and affine differential geometry of product
spaces. Ann. of Math. 40, 892-913 . . . . . . . . . . . . . . 285

1918, 1
*FINSLER, P.: ber Kurven und Flchen in allgemeinen Rumen. Thesis.
Gttingen. 121 pp. . . . . . . . . . . . . . . . . . . . . . 40, 41
1940, 1
ber die Krmmung der Kurven und Flchen. Reale Ace. d'Italia 9, Con-
vegno Volta 18, 1-18 . . . . . . . . . . . . . . . . . . . . . . . 269
1902, 1
FINZI, A.: Le ipersuperficie a tre dimensioni ehe si possono rappresentare con-
formemente sullo spazio euclideo. Atti R. Istit. Veneto (VIII) 5 (= 62),
1049-1 062 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306
1921, 1
Sulla representabilita conforme di due varieta ad n dimensioni l'una sull'al-
tra. Atti R. Istit. Veneto 8on, 777-789 . . . . . . . . . . . . . . 306

1922, 1
Sulle varieta in rappresentazione conforme con la varieta euclidea a piu
di tre dimensioni. Rend. d. Linc. (V) 31I, 8-12 . . . . . . . . . . . 306

1923, 1
Sulla curvatura conforme di una varieta. Rend. d. Linc. (V) 32I, 215-218 305

1949, 1
*FINZI, B., and M. PASTORI: Calcolo tensoriale e applicazioni. Zanichelli,
Bologna. VII + 427 pp.
1933, 1
FREUDENTHAL, H.: Ein Aufbau der Lieschen Gruppentheorie. J. D. M. V. 43,
26-39 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208

1924, 1
FRIEDMANN, A., and J. A. ScHOUTEN: ber die Geometrie der halbsymmetri-
schen bertragungen. Math. Zeitschr. 21, 211-223. . . . . . . . . 126, 127

1925, 1
FRIESECKE, H.: Vektorbertragung, Richtungsbertragung, Metrik. Math.
Ann. 94, 101-118 . . . . . . . . . . . . . . . . . . . . . . . 123,287

1879, 1
FROBENIUS, G.: ber homogene totale Differentialgleichungen. J ourn. f. d.
r. u. a. Math. 86, 1-19 . . . . . . . . . . . . . . . . . . . 92
1903, 1
FuBINI, G.: Sugli spazi ehe amettono un gruppo continuo di movimenti. Ann.
di Mat. (3) 8, 39-81 . . . . . . . . . . . . . . . . . . . . . . 348, 350
 Bibliography. 447

1903, 2
Fun!!\ I, G.: Sulle metriche definite da una forma Hermitiana. Atti Istit.
Veneto (8) 6 ( = 63), 501-513 . . . . . . . . . . . . . . . . . . . . 406
1904, 1
Sugli spazi a quattro dimensioni ehe ammettono un gruppo continuo di
movimenti. Ann. di Mat. (3) 9, 33-90 . . . . . . . . . . . . . . 348, 350
1905, 1
Sulle coppie di varieta geodeticamenta applicabili. Rend. d. Linc. (V) 141,
678-683; 14n, 315-322 . . . . . . . . . . . . . . . . . . . . 294, 296
1909, 1
Sulle rappresentazioni ehe conservano le ipersfere. Ann. di Matern. (111)
16, 141-160 . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
1935. 1
FucHs, B.: ber vollstndig geodtische analytische Mannigfaltigkeiten einer
vierdimensionalen. Mitt. Math. und Mech. Tomsk. 1, 168-174 . . . 269, 412
193 7, 1
- ber geodtische Mannigfaltigkeiten einer bei pseudokonformen Abbil-
dungen invarianten Riemannschen .Geometrie. Rec. Math. Moscow 2, 567
to 594 . . . . . . . . . . . . . . . . . . . . . . . . 399, 404, 407, 412
1942, 1
GALVANI, 0.: Sur Ia realisation de certains espaces a parallelisme absolu pardes
congruences de droites. C. R. 214, 337-339 . . . . . . . . . . . . . . 125
1942, 2
Sur !es connexions euclidiennes a courhure non nulle realisable par des con-
gruences de droites. C. R. 214, 733-735 . . . . . . . . . . . . . . . 125
1943,
Sur Ia realisation des connexions euclidiennes ponctuelles a deux dimensions
!es plus generales. C. R. 216, 23-25 . . . . . . . . . . . . . . . . 125
1943, 2
- Sur Ia realisation des connexions ponctuelles euclidienne et affines a
n dimensions. C. R. 216, 519-521 . . . . . . . . . . . . . . . . . 125
1945, 1
- Sur Ia realisationdes espaces ponctuels a torsion en geometrie euclidienne.
Ann. Ecole Norm. Sup. (3) 62, 1-92 . . . . . . . . . . . . 125
1946, 1
La realisation des connexions ponctuelles affines et Ia geometrie des
groupes de Lie. J. Math. Pures et Appl. 25, 209-239 . . . . . . . . 125
1827, 1
*GAuss, C. F.: Disquisitiones generales circa superficies curvas. Werke IV;
german translation in Ostwald's Klassiker 5 (1889). 62 pp . . . . . . . 245
G:Eri:ENrAu, J. (see T. DE DoNDER).
1949, 1
GEoRGHIU, 0. E.: Les lois de transformationdes objets geometriques speciaux
lineaires de classe v, avec une composante en X 1 C. R. 229, 611-613 67
1938, 1
GHoSH, N. N.: New methods in the geometry of a hypersurface. Bull. Calcutta
Math. Soc. 30, 73-84 . . . . . . . . . .. . . . . . . . . . . . . . . 242
448 Bibliography.

1940, 1
GHosH, N. N.: The tortuosity of a variety. Bull. Calcutta Math. Soc. 32, 51-60 269
1941, 1
- The tortuosity of submanifolds of a variety. Bull. Calcutta Math. Soc. 33.
187-195 0 0 0 0 0 0 0 0 0 0 269
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1950, 1
On complex tensor calculus. Bull. Calcutta Math. Soc. 42, 65-72. 389
1937. 1
GrvENS, J. W. (see also 0. VEBLEN): Tensor coordinates of linear spaces.
Ann. of Math. 38. 355-385 . . . . . . . . . . . . . . . . . . . . 23, 57
1940, 1
Factorization .and signatures of Lorentz matrices. Bull. Am. Math. Soc.
46, 81-85 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 50
1930, 1
GoLAB, S. (see also J. A. ScHOUTEN): ber verallgemeinerte projektive Geo-
metrie. Prace mat. fiz. Warszawa 37. 91-153. (Diss.) . . . . . . 3, 300,303
1931, 1
Sopra le connessioni lineari generali. Estenzione d'un teorema di Bompiani
nel caso piu generale. Ann. di Mat. 4 (8) 283-291 . . . . . . 144
1934, 1
Sur !'ordre de planeite des espaces plonges. Ann. di Mat. (4) 13, 119-125 232

1934, 2
ber die affinen Invarianten einer Kurve der Xp. die in einem Ln eingebet-
tet ist. Abh. Sem. Vektor u. Tensor. Moskau 4, 36-362 . . . . . . . 232

1935. 1
Ein Beitrag zur konformen Abbildung von zwei Riemannschen Rumen
aufeinander. Ann. Soc. Pol. 13, 13-19 . . . . . . . . . 315
1938, 1
- ber die Klassifikation der geometrischen Objekte. Math. Zeitschr. 44,
104-114 0 0 0 0 0 0 0 67
1945. 1
- Sur la generalisation d'une formule de Lancret concemant l'uniformisation
des equations de Frenet. Ann. Soc. Polon. Math. 18, 129-133 . 230
1946, 1
Sur la tMorie des objets geometriques. Ann. Soc. Polon. Math. 19, 7-35 67

1947. 1
Sur la theoriedes objets geometriques. (Reduction des objets geometriques
speciaux de premiere classe aux objets du type D.) Ann. Soc. Polon. Math.
26, 1-27 . 0 0 0 0 0 0 0 0 67
1948. 1
- Aleuni teoremi della teorema degli oggetti geometri. Atti Accad. Naz. Lir..c.
s. 12-122 . . . . . . . . . . . . . . . . . . 0 67
1949. 1
Generalisation des equations de Bonnet-Kowalewski dans l'espace a un
nombre arbitraire de. dimensions. Ann. Soc. Polon. Math. 22, 97-156 242, 250
 Bibliography. 449

1950, 1
Go LAB, S.: La notion de similitude parmi les objets geometriques. Bull. Int.
Acad. Polon. 1-7 . . . . . . . . . . . . . . . . . . . . . . . . 67, 108
1950. 2
Sur les objets geometriques a une composante. Ann. Soc. Polon. Math. 23,
79-89 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
1951, 1
- and T. H. WRBEL: Courbure et torsion geodesique pour les courbes situees
sur les hypersurfaces a n-1 dimensions plongees dans l'espace an dimen-
sions. Ann. Soc. Polon. Math. 24, 25-51 . . . . . . . . . . . . . . . 242
1941, 1
GoLIFMAN, R.: Sur les courbes a une dimensionreelle dans l'espace hermitien
hyperbolique. Bull. Soc. Roy. Sei. Liege 10, 57-67 . . . . . . . . . 415
1942, t
*GoNQALVEs-MIRANDA, M.: Tensor calculus. Anais Fac. Ci. Porto 27 (portu-
guese). 54 pp.
1922, 1
*GouRSAT, E.: Le~ons sur le problerne de Pfaff. Hermann, Paris. 386 pp. 83, 85
1950. 1
GRAEUB, W.: Geometrische Deutung des Krmmungstensors. Arch. Math.
2, 148-151 . . . . . . . . . . . . . . . . . . . . . . 138
1844, 1
*GRASSMANN, H.: Dielineale Ausdehnungslehre. Ein neuer Zweig der Mathe-
matik. Ges. Schriften I, 139 pp. Teubner, Leipzig 1894 . . . . . . . . 36
1862,
*- Die Ausdehnungslehre. Gesammelte Schriften II, 511 pp. Teubner,
Leipzig 1896 . . . . . . . . . . . . . . . . . . . . . . . . . . 23,366
1934.
GRAUSTEIN, W. C.: The geometry of Riemann spaces. Trans. Am. Math. Soc.
36, 542-585 . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
1925, 1
*GRISS, G. F. C.: Differentialinvarianten von Systemen von Vektoren, Diss.
P. Noordhoff, Groningen. 61 pp. . . . . . . . . . . . . . . 142
1951, 1
GuGGENHEIMER, H. (see also B. EcKMANN} : Sur les varietes qui possedent une
forme exterieure quadratique fermee. C. R. 232, 470-472 . . . . 179
1953. 1
Geometria pseudo-khleriana. Ace. Naz. Lincei {8) 14, 22-222 397
1935.- 1
GUGINE, E.: Sul trasporto ciclico de un tensore di ordine qualunque. Rend.
d. Linc. (6) 22, 295-299 . . . . . . . . . . . . . . . . . . 138
1928, 1
*GuiCHARD, C.: Les courbes de l'espace an dimensions. Mem. d. Sc. Math.
XXIX. Gauthier-Villars, Paris. 63 pp.. . . . . . . . . . . . 227
1933. 1
GuREVIC, G. B.: ber einige Integralaufgaben der Tensoranalysis. Abh. Sem.
Vektor u. Tensor. Moskau 1, 143-187 . . . . . . . . . . . . . 36, 87
Schonten, Ricci-Calculus, 2. Auf!. 29
450 Bibliography.

1933, 2
GuREVI<J. G. B.: Sur les formes canoniques d'un trivecteur dans l'espace a six
dimensions. C. R. 197, 384-385 . . . . . . . . . . . . . . . . . . . 36

1934, 1
Sur les trivecteurs dans l'espace a sept dimensions. C. R. Acad. Sc., URSS.
3, 564-569 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
1934, 2
Sur qiielques invariants d'un trivecteur et d'une ferme cubique. C. R. Acad.
Sc. USSR. 3, 317-319 . . . . . . . . . . . . . . . . . . . . 36
1934, 3
- Reduction d'un trivecteur a une forme canonique dans un cas special. C. R.
Acad. Sc. USSR. 3, 32-321 . . . . . . . . . . . . . . . . . 36
1935, 1
- Classification des trivecteurs ayant le rang huit. C. R. Acad. Sc. USSR.
2, 353-356 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

1935, 2
- L'algebre du trivecteur. Abh. Sem.Vektoru. Tensor.Moscow II, 51-113 36

1950, 1
- Complete systems of symmetric and skew-symmetric tensors. Mat. Sbornik
27, 103-116 (russian). . . . . . . . . . . . . . . . . . . . . . . . 36
1950, 2
- On some linear transformations of symmetric tensors or polyvectors. Mat.
Sbornik 26, 463-470 (russian) . . . . . . . . . . . . . . . . . . . . 35
1950, 3
- Canonization of a pair of bivectors. Trudy Sem. Vektor, Tenzor. Analizu
8, 355-363 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
1950, 4
- On some affinors connected with trivectors of the eighth rank. Trudy Sem.
Vektor, Tenzor. Analizu 8, 296-300 (russian) . . . . . . . . . . . . . 35
1940, 1, 2
*HAACK, W.: Differential-Geometrie I, II. Wolfenbtteler Verlagsanstalt,
Wolfenbttel and Hannover. 136 and 131 pp . . . . . . . . . . . . 306,342

1935, 1
HAANTJES, J. (see also J. A. ScHOUTEN) : Klassifikation der antilinearen Trans-
formationen. Math. Ann. 112, 131-148 . . . . . . . . . . . . . . 53

1937, 1
- On the projective geometry of paths. Proc. Edinburgh Math. Soc. 5, 103
to 115 . . . . . . . . . . . . . . . . . . . . . . . . . . 301,302,327
1937, 2
- Gonformal representations of an n-dimensional euclidean space with a non-
definite fundamental form on itself. Proc. Kon. Ned. Akad. Amsterdam
40, 70-705 . . . . . . . . . . . . . . . . . . . . . . . . . 312, 315
1939, 1
- and W. WRONA: ber konformeuklidische und Einsteinsehe Rume gerader
Dimensionen. Proc. Kon. Ned. Akad. Amsterdam 42, 626-636 148,274,305,312
 Bibliography. 451

1940, 1
HAANTJES, J.: Eine Charakterisierung der konformeuklidischen Rume. Proc.
Kon. Ned. Akad. Amsterdam 43, 91-94 . . . . . . . . . . . . . . . 305
1941, 1
Conformal differential geometry I. Curves in conformal euclidean spaces.
Proc. Kon. Ned. Akad. Amsterdam 44, 814-824 . . . . . . . . . . 305, 315
1942, 1
- Conformal differential geometry II. Curves in conformal two-dimensional
spaces. Proc. Kon. Ned. Akad. Amsterdam 45, 249-255 . . . . . . 305, 315
1942, 2
- Conformal differential geometry III. Surfaces in three-dimensional space.
Proc. Kon. Ned. Akad. Amsterdam 45, 836-841 . . . . . . . . . 305, 315
1942, 3
- Conformal differential geometry IV. Surfaces in three-dimensional space.
Proc. Kon. Ned. Akad. Amsterdam 45, 918-923 . . . . . . . . 305, 315
1943, 1
- Conformal differential geometry V. Special surfaces. Proc. Kon. N ed. Akad.
Amsterdam 52, 322-331 . . . . . . . . . . . . . . . . . . . . 305, 315
1953, 1
- and G. LAMAN: On the definition of geometric objects. Indagationes Math.
Proc. Kon. Ned. Akad. Amsterdam 15, 208-222 . . . . . . . . . . . . 67
1938, 1
*HAIMOVICI, A.: Directions concourantes et directions paralleles sur une va-
riete d'un espace quelconque. Thesis. Univ. Yassy. 70pp . . . . . . 315,322
1940, 1
Sur la definition intrinseque d'un element plan normal a une variete non
holonome. C. R. 210, 595-596 . . . . . . . . . . . . . . . . . . . 253
1943, 1
Les espaces a connexion affine a parallelisme absolu et les systemes homo-
graphiques de courbes. Arm. Sc. Univ. Jassy 29, 90-100 . . . . . . . 142
1945, 1
Sur les familles de transformations ponctuelles simplement transitives.
Acad. Roum. Bull. Sc. 28, 5 pp. . . . . . . . . . . . . . . . . . . 201
1946, 1
- Sur une nouvelle espece de connexions affines dans les espaces des familles
simplement transitives de transformations de variables. Bull. Math. Soc.
Roumaine 47, 35-48 . . . . . . . . . . . . . . . . . . . . . . . . 201
1946, 2
- La geometrie des fainilles de transformations de variables dependant de pa-
rametres. c. R. 223, 969-971 . . . . . . . . . . . . . . . . . . . . 201
1946, 3
- La geometrie intrinseque des varietes Riemanniennes non holonomes. J.
Math. Pures Appl. 25, 291-305 . . . . . . . . . . . . . . . . . . . 253
1947, 1
Sur les espaces des fainilles simplement transitives de transformations de
variables a courhure de III-e espece nulle. Bull. Ecole Polytech. Jassy 2,
46-70 . . . . . . . . . . . . . . . . . . . . . . . . . 201
29*
452 Bibliography.

1948, 1
HAIMOVICI, A.: Sur la geometrie des familles de transformations ponctuelles
simplement transitives. Ann. Univ. Jassy 1, 30 (1944--1947), 1-36 201
1948, 2
- La geometrie des familles de transformations de variables dependant de
parametres. Disquisit. Math. Phys. 6, 81-128 . . . . . . . . . . . 201
1951, 1
HARTMANN, P.: On geodesie Coordinates. Am. J. of Math. 73, 949--954 . 155
1951, 1
HASHIMOTO, S.: A new proof of Liber's theorem. Kodai. Math. Sem. Rep. 118
to 119 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367
1930, 1
HAYDEN, H. A.: On a generalized helix in a riemannian space. Proc. London
Math. Soc. (2) 32, 337-345 . . . . . . . . . . . . . . . . . . . 230
1930, 2
- Asymptotic lines in a Vm in Vn. Proc. London Math. Soc. (2) 33, 22-31 271, 275
1931, 1
- Deformations of a curve in a riemannian n-space, which displace certain vec-
tors parallely at each point. Proc. London Math. Soc. (2) 32, 321-336 335, 352
1932, 1
- Subspaces of a space with torsion. Proc. London Math. Soc. (2) 34, 27-50 269
1934, 1
- On some congruences associated with a sub-variety in a Vn. Quart. J oum.
Math. Oxford (4) 33, 117-125 . . . . . . . . . . . . . . . . . 275
1934, 2
- Infinitesimal deformations of subspaces in a general metrical space. Proc.
London Math. Soc. (2) 37, 416-440 . . . . . . . . . . . . . 348,352,353
1936, 1
- Infinitesimal deformations of an Lm in an Ln. Proc. Lond. Math. Soc. (2)
41, 332-336 . . . . . . . . . . . . . . . . . . . . . . 352
1868, 1
HELMHOLTZ, H.: ber die Tatsachen, die der Geometrie zugrunde liegen.
Gtt. Nachr. 193-221; Wissensch. Abhandl. II, 618-639 . . . . . 41
1916, 1
HERGLOTZ, G.: Zur Einsteinsehen Gravitationstheorie. Sitz. Leipzig 68, 199
to 203 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148, 153
1949, 1
HESSELBACH, B.: ber die Erhaltungsstze der konformen Geometrie. Math.
Nachr. 3, 107-126 . . . . . . . . . . . . . . . . . . . . . . . . . 314
1916, 1
HESSENBERG, G.: Vektorielle Begrndung der Differentialgeometrie. Math.
Ann. 78, 187-217 . . . . . . . . . . . . . . . . . . . . . . . . . 123
1924, 1
- Beispiele zur Richtungsbertragung. J. D. M. V. 33, 93-95 143
HIRAMATU, H. (see K. YANo).
 Bibliography. 4 53

1924, 1
HLAVATY, V. (see also E. BoRTOLOTTI, J. A. ScHOUTEN): Sur le deplacement
lineaire du point. Vestnik K. C. SP. N: 13, 8 . . . . . . . . . . . . 78
1926, 1
- Les parametres locaux dans une variete de Riemann. Rend. d. Linc. (6)
4, 5 PP . . . . . . . . . . . . . . . . . . . . . . . . . 155
1926, 2
- Application des parametres locaux. Ann. d. l. Soc. Polon. d. Math. 5. 19 155
1926, 3
- Deux notes sur la metrique. Rozpravy C. Ak. 16, 16 . . . . . . . . . 155
1926, 4
Contribution au calcul differentiel absolu. Vestnik Kral. Nauk. Praha 2, 1
to 12 . . . . . . . . . . . . . . . . . . . . . . . . . . 232
1927.
- Contact de deux courbes dans une Vn. Rend. d. Linc. (6) 5. 415-420 . . 155

1927, 2
Theorie des densites dans le deplacement general. Ann. di Mat. 5, 73-83 129

1927, 3
Deplacements isohodoiques. Ens. Math. 26, 84-97 . . . . . . . . 322, 327

1928, 1
- Sur les coefficients de Ricci. C. R. 186, 1691-1693 . . . . . . . . 78,171
1928, 2
- Bemerkung zur Arbeit von Herrn T. Y. THOMAS. Math. Zeitschr. 28, 142
to 146 . . . . . . . . . . . . . . . . . . . . . . . . . 322
1928, 3
- Ein Beitrag zur Theorie der Weylschen bertragung. Proc. Kon. Ned.
Akad. Amsterdam 31, 878-881 . . . . . . . . . . . . . . 231

1929, 1
- Proprieta differenziali delle curve in uno spazio a connessione lineare gene-
rale. Rend. d. Palermo 53. 356-388. Ancora sulle proprieta differenziali
delle curve in uno spazio a connessione lineare generale. Rend. d. Pa-
lermo 53, 389-410 . . . . . . . . . . . . . . . . . . . . . . . . . 234

1929, 2
- Le parallelisme de la connexion deM. Weyl. Ann. Ec. Norm. Sup. 46, 73
to 103 . . . . . . . . . . . . . . . . . . . . . . . . . 133

1929, 3
Sur ladeformationinfinitesimale d'une courbe danslavariete metrique avec
torsion. Bull. Soc. Mathem. d. France 56, 18-25 . . . . . . . . . . . . 352

1930, 1, 2
Sur la courbure des varietes non-holonomes. Rend. d. Linc. (6) 12, 566
to 574; 647-654. . . . . . . . . . . . . . . . . . .78,253
1930, 3
- Sulle coordinate geodetiche. Rend. d. Linc. (6) 12, 502-509 . . . . . . 155
454 Bibliography.

1931, 1
HLAVATY, V.: Projektive Invarianten der Kurvenkongruenz und einer Kurve.
Math. Zeitschr. 34, 58-73 . . . . . . . . . . . . . . . . . . . . . . 302
1933, 1
- Connexion projective et deplacement projectif. Ann. di Mat. 12, 217-294
301,303,322
1934, 1
- Espaces de Knig. Abh. Sem. Vekt. u. Tensor. Moskau 4 (publ. 1937),
117-120 . . . . . . . . . . . . . . . . . . . . . . . 123
1934, 2
- Les courbes de lavariete generale an dimensions. Mem. d. Sc. Math., fase.
LXIII. Paris, Gauthier-Villars. 73 pp. . . . . . . . . . . 230, 232, 233, 234
1934, 3
- Induzierte und eingeborene Konnexion in den (nicht) holonomen Rumen.
Math. Zeitschr. 38, 283-300 . . . . . . . . . . . . . . . . . 253
1935, 1
- Espaces abstraits, courbes de Knig. Rend. d. Palermo 13, 1-39 . 123
1935, 2
- Zur Konformgeometrie I, II, III. Proc. Kon. Akad. Amsterdam 38, 281
to 286, 738-743, 1006-1011 . . . . . . . . . . . . . . . . . 231, 315, 316
1936, 1
Systemes des connexion deM. Weyl. Bull. Acad. des Sc. Boh. 37, 181-184
315,316
1939, 1
- Contributo alla teoria delle connessioni. Reale Accad d'Italia, Fondazione
Ales. Volta 479-506 . . . . . . . . . . . . . . . . . . . . 303

1939, 2
*- Differentialgeometrie der Kurven und Flchen und Tensorrechnung.
Noordhoff, Groningen-Djakarta. XI + 569 pp.
1949, 1
-Theorie d'immersion d'une Wm dans Wn. Ann. Soc. Polon. Math. 21, 196
to 206 . . . . . . . . . . . . . . . . . . . . . . . . . . 134, 269, 285
1949, 2
- Affine embedding theory. I. Affine normal spaces. Proc. Kon. Ned.
Akad. Amsterdam 52, 505-517; Indagationes Math. 11, 165-177 285
1949, 3
- Affine embedding theory. li. Frenet formulae. Proc. Kon. Ned. Akad.
52, 714-724; Indag. Math. 11, 244-254 . . . . . . . . 285
1949, 4
- Affine embedding theory. III. Integrability conditions. Proc. Kon. Ned.
Akad. Amsterdam. 52, 977-986; Indag. Math. 11, 365-374 . . . . . . . 285
1950, 1
- Deformation theory of subspaces in a Riemann space. Proc. Am. Math.
Soc. 1, 60o-617 . . . . . . . . . . . . . . . . . . 352
1952, 1
- Embedding theory of a Wm in a Wn. Rend. d. Palermo 2, 403-438 134, 269, 285
 Bibliography. 455

1952. 2
HLAVATY, V.: The Einstein connection of the unified theory of relativity.
Proc.Nat.Acad. 38,415-419 . . . . . . . . . . . . . . . . . . . . 179
1952, 3
- The elementary basic principles of the unified theory of relativity. Proc.
Nat. Acad. 38, 243-247 . . . . . . . . . . . 179

1952. 4
- The elementary basic principles of the unified theory of relativity A. J. Ra-
tional Mech. Anal. 1, 539-562 . . . . . . . . . . . . . . . . . . . . 179

1952, 5
- Intrinsic deformation theory of subspaces in a Riemann space. J. Rational
Mech. Anal. 1, 49-72 . . . . . . . . . . . . . . . . . . . . . . . . 352
1952, 6
- The Schrdinger final affine field laws. Proc. Nat. Acad. 38, 1052-1058 179

1953. 1
- The elementary basic principles of the unified theory of relativity B. J.
Rational Mech. Anal. 2, 1-52 . . . . . . . . . . . . . . 179

1953. 2
-- Connections between Einstein's two unified theories of relativity. Proc.
Nat. Acad. 39. 507-510 . . . . . . . . . . . 179

1953. 3
- and A. W. SAENZ: Uniquenes theorems in the unified theory of relativity.
J. Nat. Mech. An. 2, 523-536 . . . . . . . . . . . . . . . . . 179

1953, 4
The spinor connections in the unified Einstein theory of relativity. Proc.
Nat. Acad. USA. 39. 501-536 . . . . . . . . . . . . . 179

1954, 1
- The elementary basic principles of the unified theory of relativity; C1
Introduction; C2 Applications I. J. Rat. Mech. Anal. 3, 103-146, 147-179 179
HocHRAINER, A. (see A. DuscHEK).
1951, 1
HoDGE, W. V. D.: A special type of Khler manifold. Proc. London Math.
Soc. (3) 1, 104-117 . . . . . . . . . . . . . 397

1934. 1
HoKARI, S.: ber die Bivektorbertragung. Journ. Hokkaido Univ. 2, 103
to 117 . . . . . . . . . . . . . . . . . . . . . . . . . 123, 138

1935. 1
HoMBU, H.: Zur Theorie der unitren Geometrie. Journ. Fac. Sc. Hokkaido
Univ. (1), 3. 27-42 . . . . . . . . . . . . . . . . . . . . . . . 399,410

1936, 1
Zur Theorie der Affinorbertragung. Proc. Imp. Acad. Japan 12, 109-111 138

1936, 1
and M. MIKAMI: Paraboias and projective transformations in the genera-
lized spaces of paths. Jap, J. Math. 17, 307-335 . . . . . . . . . . . 300
456 Bibliography.

1941, 2
HoMBU, H. and Y. OKADA: On the projective theory of asymmetric Connec-
tions. Proc. Phys. Math. Soc. Japan 23, 357-362 . . . . . . . . . . . 322
1942, 1
and M. MIKAMI: Conics in the projectively connected manifolds. Mem. Fac.
Sc. Kyilsyu Imp. Univ. 2, 217-239 . . . . . . . . . . . . . . . . . 300
1927, 1
HoRAK, Z.: Die Formeln fr allgemeine lineare bertragung bei Benutzung
von nichtholonomen Parametern. Nieuw Archief v. Wisk. (2) 15, 193-201
78,253
1927, 2
- Eine Verallgemeinerung des Mannigfaltigkeitsbegriffes. Spisy Brno 86,
20 pp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78,253
1928, 1
Sur !es courbures des varietes non holonomes. C. R. 187, 1273-1275 253
1929, 1
Sur les derivees par rapport aux affineurs. Ens. Math. 28, 212-214 . . . 72
1948, 1
HuTCHINSON, L. C.: Notes on the member of leaves of an alternating tensor.
Presented to the Am. Math. Soc. Dec. 28, 1948. The author was so kind
to send me a manuscript . . . . . . . . . . . . . . . . . . . . . . 36
1951, 1
IcHINOHE, A.: Sur la possibilite de plonger un espace a connexion conforme
donne dans un espace conforme. Tohoku Math. J. 2, 193-204 . . . . 268,315
IMAI, T. (see K. YANo).
1950, 1
INGRAHAM, R. L.: Contributions to the Schrdinger non-symmetric affine
unified field theory. Ann. of Math. 52, 743-752 . . . . . . . . . 179
1951, 1
Relativite conforme. C. R. 232, 1072-1074 179
1951, 2
Sur une theorie de la "relativite conforme". C. R. 232, 938-940 179
1944, 1
IwAMOTO, H.: On Frenet's formulae. Tensor J. 44-49 (japanese) 277
1950, 1
On the structure of riemannian spaces whose holonomy groups fix a null-
system. Tohoku Math. J. 1, 109-135 . . . . . . . . . . . . . . . 367,397
1951, 1
On the relation between homological structure of riemannian spaces and
exact differential forms which areinvariant under holonomy groups I. To-
huku Math. J. 3, 59-70 . . . . . . . . . . . . . . . . . . . . 367
1934, 1
J ACOBSTHAL, E.: Zur Theorie der linearen Ableitungen. Sitz. Berl. Math. Ges.
33. 15-34 . . . . . . . . . . . . . . . . . . . . 53
1926, 1
J ANET, M.: Sur la possibilite de plonger un espace riemannien donne dans
un espace euclidien. Ann. d. 1. Soc. Polon. d. Math. 5. 38-43 . . . . . 268
 Bibliography. 457

1928, 1
* JRNEFELT, G.: Zur Affinoranalysis. Diss. Helsinki, 91 pp. Acta Sei. Fenni-
ciae (A) 28, No. 9 . . . . . . . . . . . . . . . . . . . . . 232, 263, 278
1948, 1
J OHNSON, P. B. : The role of the directrix in Levi-Civita parallelism. Duke Math.
J. 15, 389-405 . . . . . . . . . . . . . . . . . . 123, 131, 138, 340, 341
1874, 1
JoRDAN, c.: Sur la theorie des courbes dans l'espace an dimensions. c. R.
79. 795-797 . . . . . . . . . . . . . . . . . . . . . . . . 230
1948, 1
JORDAN, P.: ber den Riemannschen Krmmungstensor I. Einsteinsehe Theo-
rie. Z. Physik 124, 602-607 . . . . . . . . . . . . . . . . . . . . 179
1948, 2
- ber den Riemannschen Krmmungstensor II. Eddingtonsche und Schr-
dingersche Theorie. Z. Physik 124, 608-613. . . . 179
1950, 1
- Einsteins neue Untersuchungen. Elektrotechn. Z. 71, 615-618 179
1921, 1
JuvET, G.: Les formnies de Frenet dans un espace generalise de Weyl. Bull.
Soc. Neuchteloise 46. Cf. also C. R. 172, 1647-1650 . 230
1922, 1
*- Introduction au calcul tensoriel et au calcul differentiel absolu. Blan-
chard, Paris. 100 pp.
1925, 1
- Sur le deplacement parallele le plus general et sur l'etude des courbes
tracees dans une multiciplite quelconque. Bull. Soc. Math. 53. 60-74 230
1930, 1
KAGAN, B.: Sur les espaces sous-projectifs. C. R. 191, 548-550. 317
1933, 1
- ber eine Erweiterung des Begriffes vom projektiven Raume und dem zu-
gehrigen Absolut. Abh. Sem. Vektor u. Tensor. Moskau 1, 12-101
317,318,330
1935, 1
- Der Ausnahmefall in der Theorie der subprojektiven Rume. Abh. Sem.
Vektor u. Tensor. Moskau 2-3. 151-170 . . . . . . . . . . . . . . . 317
1933, 1
KHLER, E.: ber eine bemerkenswerte Hermitesche Metrik. Abh. Math. Sem.
Hamburg 9, 173-186 . . . . . . . . . . . . . . . . . . 397,399
1934, 1
*- Einfhrung in die Theorie der Systeme von Differentialgleichungen.
Hamb. Math. Einzelschr. Heft 16 . . . . . . . . . . . . . . . 26, 74,117
1922, 1
*KMMERER, F.: Zur Flchentheorie im n-fach ausgedehnten Raume. Thesis.
Gieen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
1927, 1
*KMMERER, W.: Die trilineare alternierende homogene Form in acht Ver-
nderlichen. Mitt. Math. Sem. Gieen (Thesis) . . . . . . . . . . . . 36
458 Bibliography.

KAMPEN, E. R. VAN (see J. A. SCHOUTEN).

1941,
KANITANI, J.: Spaces with a linear connection reduced from a space with a pro-
jective connection. Tensor 4, 1-12 (japanese) . . . . . . . . . 301
1943, 1
Sur les espaces representables geodesiquement sur l'espace projectif. Proc.
Phys. Math. Soc. Japan 25, 87-100 . . . . . . . . . . . . . 296, 301
1943, 2
Sur un espace a connexion projective renfermant des hyperquadriques.
Proc. Phys. Math. Soc. Japan 25, 617-621 . . . . . . . . . 296,301
194 7, 1
Sur l'espace a connexion projective majorante I. Jap. J. Math. 19, 343
to 361 . . . . . . . . . . . . . . . . . . . . . . . 268, 296, 301
1947, 2
On a generalization of the projective deformation. Mem. Coll. Sc. Univ.
Kyoto 25, 23-26 301
1948, 1
Sur l'espace a connexion projective majorante II. Jap. J. Math. 19,
395-403 . . . . . . . . . . . . . . . . . . 268,296,301
1949, 1
On a method of plunging of R 2 with symmetric projective connexion into
a projective space 5 4 Mem. Coll. Sei. Univ. Kyoto 25, 87-91 268, 296, 301
1950, 1
Sur le developpement d'une courbe dans un espace a connexion projective.
Mem. Univ. Kyoto 26, 31-43 . . . . . . . . . . . . . . 128, 367

1950, 2
Sur l'espace a connexion projective majorante III. J ap. J. Math. 20,
45-54 . . . . . . . . . . . . . .... . 268,296,301
1952, 1
Sur la transformation infinitesimale du groupe d'holonomie. Mem. Coll. Sc.
Univ. Kyoto Math. 27, 75-93 . . . . . . . . . . . . . 362,367, 370
1934, 1
KAPLAN, N.: V. in R 6 Proc. Nat. Acad. 20, 489-499 . . . . . . . . . . 269
1934, 2
- Vm in Rm+l possessing an m-tuply orthogonal net along the lines of curva-
ture. Joum. Math. Phys. Mass. 13, 426-432 . . . . . . . . . . . . . 269
1936, 1
On certain three-dimensional manifolds in euclidean sixspace (V3 in R 6).
Journ. Math. Phys. Mass. 15, 219-248 . . . . . . . . . . . . . . . . 269
1941, 1
- The problern of equivalence. Abh. Sem. Vektor. u. Tensor. 5, 284-289
(russian) . . . . . . . . . . . . . . . . . . . 278
1910, 1
KASNER, E.: The theorem of Thomson and Tait and natural families of
trajectories. Trans. Am. Math. Soc. 11, 121-140 . . . . . . . . . 310
 Bibliography. 459

1913, 1
*KASNER, E.: Differential geometric aspects of dynamics. Princeton Collo-
quium lectures (delivered 1909). 117 pp. . . . . . . . . . . . . . . . 310
1921, 1
- Einsteins theory of gravitation: determination of the field by light signals.
Amer. Joum. 43, 20-28 . . . . . . . . . . . . . . . . . 314
1922, 1
- The solar gravitational field completely determined by its light rays.
Math. Ann. 85, 227-236 . . . . . . . . . . . . . . . . . . . . . . 314
1932, 1
KAWAGUCHI, A.: Theory of connexions in the general Finsler manifold I.
Proc. Imp. Acad. Tokyo 8, 340-343 . . . . . . . . . . . . . . . . . 123
1933, 1
- The foundation of the theory of displacement I. Proc. Imp. Acad. 9, 351-354 123
1933, 2
- Theory of connexions in the general Finsler manif~ld II. Proc. Imp. Acad.
Tokyo 9, 351-354 . . . . . . . . . . . . . . . . . . . . . . . . . 123
1934, 1
- The foundation of the theory of displacements II, III. Proc. Imp. Acad.
10, 45-48; 133-136 . . . . . . . . . . . . . . . . . . . . 123
1935, 1
- Differentialgeometrie in den verschiedenen Funktionalrumen. J oum.
Hokkaido Univ. 3, 43-106 . . . . . . . . . . . . . . . . 123
1939, 1
- Eine Verallgemeinerung von Extensoren. Monatsh. Math. Phys. 48, 329
to 339 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
1952, 1
- On the theory of nonlinear connections I. Introduction to the theory of
generalnonlinear connections. Tensor 2, 123-142 . . . . . . . . . . . 123
1952, 1
KAWAGUCHI, M.: A generalization of the extensor. Tensor 2, 59-66 58
1885, 1
*KILLING, W.: Die nicht-euklidischen Raumformen in analytischer Behand-
lung. Teubner, Leipzig. XII + 264 pp. . . . . . . . . . . . . . 244, 271
1892, 1
- ber die Grundlagen der Geometrie. Joum. f. d. r. u. a. Math. 109, 121
to 186 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249, 348
1893, 1
*- Einfhrung in die Grundlagen der Geometrie. Schnig, Paderbom. I,
357 pp.; II (1898), 361 pp . . . . . . . . . . . . . . . . . . . . . . 244
1943, 1
KIMPARA, M.: Theory of surfaces in a three-dimensional space with a projective
connection. Tensor 6, 33-44 (japanese) . . . . . . . . . . . . . . 296, 301
1872, 1
KLEIN, F. :Vergleichende Betrachtungenberneuere geometrische Forschungen.
Erlangen. Reprinted Math. Ann. 43 (1893) 63-100 . . . . . . . . . 64
460 Bibliography.

1926, 1
*KLEIN, F.: Vorlesungen ber hhere Geometrie, 3. Aufl., bearbeitet von
W. BLASCHKE. Springer, Berlin. 405 pp. . . . . . . . . . 31,316
1951, 1
KLINGENBERG, W.: Zur affinen Differentialgeometrie. I. ber p-dimensionale
Minimalflchen und Sphren im n-dimensionalen Raum. Math. Zeitschr.
54, 65-80 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
1951, 2
- Zur affinen Differentialgeometrie. II. ber zweidimensionale Flchen im
vierdimensionalen Raum. Math. Zeitschr. 54, 184-216 . . . . . . . . . 232
1952, 1
- ber das Einspannungsproblem in der projektiven und affinen Differen-
tialgeometrie. Math. Zeitschr. 55, 321-345 . . . . . . . . . . . . . . 232
1930, 1
KNEBELMAN, M. S. (see also L. P. EISENHART): On groups of motions in related
spaces. Am. J. Math. 52, 28-282 . . . . . . . . . . . . . . . . . . 350
1945, 1
- On the equations of motion in a Riemann space. Bull. Am. Soc. 51, 682-685 346
1951, 1
- Spaces of relative parallelism. Ann. of Math. 53, 387-899 . . 123
1939, 1
KocziAN, E.: Sopra i ds 1 di classe uno. Atti Ist. Veneto 98, 27-42 268
KODAIRA, K. (see G. DE RHAM).
1920, 1
KNIG, R.: Beitrge zu einer allgemeinen Mannigfaltigkeitslehre. J. D. M. V.
28, 213-228 . . . . . . . . . . . . . . . . . . . . 123,303,316
1932, 1
- Zur Grundlegung der Tensorrechnung. J. D. M. V. 41, 169-189 . . . . 123
1934, 1
- and C. PESCHL: Axiomatischer Aufbau der Operationen im Tensorraum
I, II, III. Ber. Schs. Akad. 86, 129-154; 267-298; 383-410 . . . . . . 123
1935, 1
- and K. H. WEISE: Axiomatischer Aufbau der Operationen im Tensorraum
IV. Ber. Schs. Akad. 87, 223-250 . . . . . . . . . . . . . . 123
1946, 1
KosAMBI, D. D.: Sur la differentiation covariante. C. R. 222, 211-213 108
1947, 1
- Les invariants differentials d'un tenseur covariant a deux indices. C. R.
225, 790-792 . . . . . . . . . . . . . . . . . . . . . . . . 179
1949, 1
- The differential invariants of a two-index tensor. Bull. Am. Math. Soc.
55, 90-94 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
1951, 1
Series expansions of continuous groups. Quart. J. Math. Oxford 2, 244-257
108,346,347
 Bibliography. 461

1952, 1
KosAMBI, D. D.: Path-spaces admitting collineations. Quart. J. Math. Oxford,
Ser. 3, 1-11 . . . . . . . . . . . . . . . . . . . . . . . 102,137, 346
1934, 1
KRAUS, A.: Hyperflchenstreifen im Riemannschen Raume. Mon. Math. Phys.
41, 424-438 . . . . . . . . . . . . . . . . . . . . . . . 230
1869, 1
KRONECKER, L.: ber Systeme von Funktionen mehrerer Variablen. Monats-
her. Akad. Wiss. Berlin 159-193; 688-698; Werke I, 175-212; 213-226 250
1903. 1
KHNE, H.: Die Grundgleichungen einer beliebigen Mannigfaltigkeit. Arch.
d. Math. u. Phys. (III) 4, 30Q-311 . . . . . . . . . . . . . . 230, 263, 278
1904. 1
- ber die Krmmung einer beliebigen Mannigfaltigkeit. Arch. d. Math. u.
Phys. (III) 6, 251-260 . . . . . . . . . . . . . . . . . . . . . . . 256
1950, 1, 2
KuiPER, N. H.: Einstein spaces an dconnections I, I I. Kon. Ned. Akad. Amster-
dam, Proc. 53. 156o-1567; 1568-1576 = Indagationes Math. 12, 505-512;
513-521 . . . . . . . . . . . . . . . . . . . . . . . . . 148,312, 314
1951. 1
Sur les proprietes conformes des espaces d'Einstein. Coll. Geometrie Dif-
fErentielle, Louvain 1951, 165-166 . . . . . . . . . . . . . . . . 148,312
1951, 2
- On the holonomic groups of the vector displacement in riemannian spaces.
Kon. Ned. Akad. Amsterdam. Proc. 54 = Indagationes Math. 13, 445-451 367
1939, 1
KULK, W. VAN DER (see also J. A. ScHaUTEN): bertragungen mit alternie-
rendem Krmmungsaffinor. Kon. Ned. Akad. Amsterdam. Proc. 42, 753
to 763 . . . . . . . . . . . . . . . . . . . . . . . . . 144, 148
1952, 1
KuRITA, M.: Riemann space with two-parametric homogeneaus holonomy
group. Nagoya Math. J. 4, 35-42 . . . . . . . . . . . . . 367
1923, 1
*LAG RANGE, R.: Sur le calcul differential absolu. These, Paris. 69 pp. 306
1926, 1
*- Calcul differentiel absolu. Mem. d. Sc. Math., fase. 19 . . . . . 126
1941, 1
- Sur les invariants conformes d'une courbe. C. R. 242, 1123-1126 . 315
1941, 2
- Proprietes differentielles des courbes de l'espace conforme an dimensions.
c. R. 213, 551-553. . . . . . . . . . . . . . . . . . . 315
LAMAN, G. (see J. HAANTJES).
1938, 1
LANczos, C.: A remarkable property of the Riemann-Christoffel tensor in four
dimensions. Ann. of Math. (2) 39. 842-850 . . . . . . . . . . . . . . 146
462 Bibliography.

1936, 1
LAND, F. W.: On certain properties of non holonomic submanifolds of a non-
riemannian space and the effect on them of a general deformation. Proc.
London Math. Soc. (2) 41, 497-516 . . . . . . . . . . . . . . . . . 352
1938, 1
LAPTEV, B.: La derivee de Liedes objets geometriques qui dependent de point
et de direction. Bull. Soc. Phys. Math. Kazan (3) 10, 3-39 (french summary) 105
1943, 1
- Sur une classe des geometries intrinseques induites sur une surface dans un
espace a connexion affine. C. R. Acad. Sei. URSS. 41, 315-317 . 268

1945, 1
Sur l'immersion d'un espace a connexion affine dans un espace affine. C. R.
Acad. Sei. URSS. 47, 531-534 . . . . . . . . . . . . . . . . 268

1950, 1
- Differential connections of manifolds and their holonomy groups. Doklady
Akad. Nauk SSSR. 71, 597-600 (russian) . . . . . . . . . . . . . . . 370
1950, 1
*LAss, H.: Vector and tensor analysis. McGraw Hili, New York. XI + 347pp.
1940, 1
LAURA, E.: Sopra un gruppo di condizioni necessarie affinehe un ds 2 sia di
classe h. Ist. Veneto Sc. Mat. Nat. 99, 97-113 . . . . . . . . . . 268
1942, 1
LEE, E. H.: On unitary geometry. Acad. Sinica Science Record 1, 49-54 397
1943, 1
- A kind of even-dimensional differential geometry and its application to ex-
terior calculus. Am. J. Math. 65, 433-438 . . . . . . . . . . . . . . 179
1945, 1
- On even dimensional skew-metric spaces and their groups of transforma-
tions. Am. J. Math. 67, 321-328 . . . . . . . . . . . . . . 179
1947, 1
- On skew-metric spaces and function groups. Am. J. Math. 69, 79-800 179
LEKKERKERKER, C. G. (see W. PEREMANS).
1926, 1
LENSE, J.: ber ametrische Mannigfaltigkeiten und quadratische Differential-
formen mit verschwindender Diskriminante. J. D. M. V. 35, 281-294 227
1926, 2
- ber spezielle ametrische Mannigfaltigkeiten. Sitz. Akad. Wien (2a) 135,
29-32 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
1926, 3
- ber ametrische Mannigfaltigkeiten und quadratische Differentialformen
mit verschwindender Diskriminante. J. D. M. V. 35, 28-294 . . . . . 227
1928, 1
- ber die Tangentialrume ametrischer Mannigfaltigkeiten. Math. Zeitschr.
29, 87-95 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
 Bibliography. 463

1932, 1
LENSE, J.: ber die Ableitungsgleichungen. der ametrischen Mannigfaltig-
keiten. Math. Zeitschr. 34, 721-736 . . . . . . . . . . . . . . . 227
1936, 1
- ber Kurven mit isotropen Normalen. Math. Ann. 112, 139-154 227
1936, 2
- ber vollisotrope Flchen. Monatsh. f. Math. u. Phys. 43, 177-186 227
1939, 1
- ber isotrope Mannigfaltigkeiten. Math. Ann. 116, 297-309 . . 227
1939, 2
- Beitrge zur Theorie der isotropen Mannigfaltigkeiten. Monatsh. Math.
Phys. 48, 121-128 . . . . . . . . . . . . . . . . . . . . . . . . . 227
1940, 1
- ber die Ableitungsgleichungen einer Mannigfaltigkeit im mehrdimensio-
nalen komplexen euklidischen Raum. Math. Zeitschr. 47, 78-84 227, 269, 278
1940, 2
- Lngentreue Abbildung, isotrope Mannigfaltigkeiten vom Rang Null, Ein-
bettungssatz. J. D. M. V. 50, 1-6 . . . . . . . . . . . . . . . . . . 227
1944, 1
- ber einige Determinanten aus der Theorie der mehrdimensionalen Mannig-
faltigkeiten. Math. Ann. 119, 216--220 . . . . . . . . . . . . . . . . 269
LERAY, J. (see E. CARTAN).
1896, 1
LEVI-CrvrTA, T. (see also G. Rrccr): Sulle trasformazioni delle equazioni dina-
miche. Ann. d. Mat. (II) 24, 255-300 . . . . . . . . . . . . . . . . 296
1917, 1
- Nozione di parallelismo in una varieta qualunque e consequente specifi-
cazione geometrica della curvature Riemanniana. Rend. Palermo 42, 173
to 205 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122, 123
1925, 1
*- Lezioni di calcolo differenziale assoluto. Stock, Roma 244
1926, 1
- Sur l'ecart geodesique. Math. Ann. 97, 291-320 . . 352
1940, 1
- Formula di Green e di Stokes. Univ. Nac Tucuman Revista 1, 23-33 97
1935. 1
LEVINE, J. (see also T. Y. THoMAS): Conformal-affine connections. Proc. Nat.
Acad. 21, 165-167 . . . . . . . . . . . . . . . . . . . . . . 316
1936, 1
Same new complete sets of identities for affine and metric spaces. Bull.
Am. Math. Soc. 41, 497-501 . . . . . . . . . . . . . . . . . . . 144, 14'7

1937. 1
Groups of motions in conformally flat spaces I. Bull. Am. Math. Soc. 42,
418-432 . . . . . . . . . . . . . . . . . . . . . . . . . 346
464 Bibliography.

1939. 1
LEVINE,].: Groups of motions in conformally flat spaces II. Bull. Am. Math.
Soc. 45, 766-773 . . . . . . . . . . . . . . . . . . . . . . . . . 34
1942, 1
A replacement theorem for conformal tensor invariants. Tohoku Math. J.
49, 69-86 . . . . . . . . . . . . . . . . . . . . . . . . 16
1948, 1
Invariant characterizations of two dimensional affine and metric spaces.
Duke Math. ] . 15, 69-77 . . . . . . . . . . . . . . . . . . . . 133, 32
1949, 1
- Fields of parallel vectors in projectively flat spaces. Duke Math. J. 16,
23-32 . . . . . . . . . . . . . . . . . . . . . . . . 291, 32
1950, 1
- Fields of parallel vectors in conformally flat spaces. Duke Math. ] . 1 7,
15-20 . . . . . . . . . . . . . . . . . . . . . . . . 291,30
1950, 2
- Classification of collineations in projectively and affinely connected spaces
of two dimensions. Ann. of Math. 52, 465-477 . . . . . . . . . . . 34

1951, 1
Collineations in Weyl spaces of two dimensions. Proc. Am. Math. Soc. 2,
264-269 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
1951, 2
Motions in linearly connected two-dimensional spaces. Proc. Am. Math. Soc.
2, 932-941 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

1925, 1
LEVY, H.: Ricci's coefficients of rotation. Bull. Am. Math. Soc. 31, 142-145 17

1925, 2
Symmetrie tensors of the second order whose covariant derivatives vanish.
Ann. of Math. (2) 27, 91-98. . . . . . . . . . . . . . . . . 285, 296, 36

1926, 1
Forma canonica dei ds 2 per iquali si annulano i simboli di Riemann a cinque
indici. Rend. d. Linc. (6) 3, 67-69 . . . . . . . . . . . . . . . . . . 16

1926, 2
Sopra alcune proprieta degli spazi per i quali si annullano i simboli di Rie-
mann a cinque indici. Rend. d. Linc. (6) 3, 126-129 . . . . . . . . . 16

1930, 1
Normalcoordinates in the geometry of paths. Proc. Nat. Acad. 16, 492 to
496 . . . . . . . . . . . . . . . . . . . . . . . . . . 15

1870, 1
L:EvY, M.: Memoire sur les coordonnees curvilignes orthogonales. ] ourn. de
l'Ec. Polyt. 26, 15 7-200 . . . . . . . . . . . . . . . . . . . . . . 24

1941, 1
LIBER, A. E.: The first algebraic integrals of the equations of geodesy. C. R.
Acad. Sei. URSS. (N.S.) 31, 84G-841 . . . . . . . . . . . . . 29
 Bibliography. 465

1947, 1
LIBER, A. E.: On the immersion of riemannian spaces of constant curvature
in one another. C. R. Acad. Sei. URSS. 55, 291-293 . . . . . . . . . 269
1949, 1
On spaces of linear affine connection "V'th one-parameter holonomy groups.
Akad. Nauk SSSR. 66, 1045-1046 (russian) . . . . . . . . . . . . . 367
1950, 1
On the classification of an affine connection in a space of two dimensions.
Mat. Sbornik 27, 249--266 (russian) . . . . . . . . . 123
1951, 1
On comitants of geometric differential objects. Akad. Nauk SSSR. 80,
529--532 . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
1941, 1
LICHNERowrcz, A. (see also A. BoREL): Les espaces a connexion semimetrique
et la mecanique. C. R. 212, 328-331 . . . . . . . . . . . . . . . . . 133
1944, 1
Sur les espaces riemanniens completement harmoniques. Bull. Soc. Math.
France 72, 146-168 . . . . . . 285, 286, 287, 382, 383, 384, 385, 386, 387, 388
1944, 2
Sur une inegalite relative aux espaces riemanniens completement harmoni-
ques. C. R. 218, 436-437 . . . . . . . . . . . . . . . 387
1944, 3
Sur les espaces riemanniens completement harmoniques. C. R. 218, 493
to 495 . . . . . . . . . . . . . . . . . . . . . . . . 388
194 5, 1
and ~. G. WALKER: Sur les espaces riemanniens harmoniques de type
hyperbolique normal. C. R. 221, 394-396 . . . . . . . . . . 386
1950, 1
- Derivation covariante et nombres de Betti. C. R. 230, 1248-1250 . . . 367
1950, 2
Sur les varietes riemanniennes admettant uneforme quadratique exterieure
a derivee covariante nulle. C. R. 231, 1413-1415 . . . . . . . 367
1950, 3
* Elements de calcul tensoriel. Colin, Paris. 216 pp.
1951, 1
Formes a derivee covariante nulle sur une variete riemannienne. C. R.
232, 146-147 . . . . . . . . . . . . . . . . . . . 367
1951, 2
Sur les varietes riemanniennes admettant une forme a derivee covariante
nulle. C. R. 232, 677-679 . . . . . . . . . . . . . . 367
1953, 1
Compatibilite des equations de la theorie unitaire d'Einstein-Schrdinger.
c. R. 237, 1383-1386. . . . . . . . . . . . . . . . . 179
1953, 2
Equations de Laplace et espaces harmoniques Premier colloque sur les
equations aux derivees partielles Louvain 17-19 Decembre 1953. 9-23 386
Schouten, Ricci-Calculus, 2. Aufl. 30
466 Bibliography.

1872, 1
LrE, S.: Komplexe, insbesondere Linien und Kugelkomplexe mit Anwendung
auf die Theorie partieller Differentialgleichungen. Math. Ann. 5, 145-246 312
1879. 1
- ber Minimalflchen. Math. Ann. 1~, 331-416. Gesamm. Abh. II 1, 122
to 218 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
1888, 1
*- and F. ENGEL: Theorie der Transformationsgruppen, Vol. I. Teubner.
185,205,214,215,218,220
1890, 1
- ber die Grundlagen der Geometrie I, II. Ber. Schs. Ges. d. Wiss. Leip-
zig 42, 284-321; 355-418 . . . . . . . . . . . . . . . . . . . . . . 41
1922, 1
LIPKA, J.: Sui sistemi E nel calcolo differenziale assoluto. Rend. d. Linc. 311,
242-245 . . . . . . . . . . . . . . . . . . . . 51
1869, 1
LIPSCHITZ, R.: Untersuchungen in Betreff der ganzen homogenen Funktionen
von n Differentialen. Joum. f. d. r. u. a. Math. 70, 71-102; 72, 1-56 {1870)
Excerpt in Monatsber. Akad. Berlin 1869, 44-53 . . . . . . . . . . . 143
1874. 1
- Extrait de six memoires dans le joumal de mathematiques de Borchardt.
Bull. d. Sc. Math. 4, 97-110; 142-157; 212-224; 297-307; 308-320 . . . . 143
1874. 2
- Ausdehnung der Theorie der Minimalflchen. Joum. f. d. r. u. a. Math. 78,
1-4S; cf. Monatsber. Akad. Berlin 1872, 361-367 . . . . . . . . . . . 274
1944, 1
LITTLEWOOD, D. E.: Invariant theory, tensors and group characters. Philos.
Trans. Roy. Soc. London 239, 305-365 . . . . . . . . . . . . . 15
1950, 1
*- The theory of group characters and matrix representation of groups.
Oxford. 310 pp. . . . . . . . . . . . . . . . . . . . . . 15
1948, 1
LoNGO, C.: Sopra una classe di varieta ehe ammettono varieta subordinate
quasi asintotiche. Atti Accad. Naz. Linc. 5, 19-21 . . . . . . . . . . 285
1936, 1
LoPCHITZ, A. M.: Sugli spazi riemanniani contenenti un campo di giaciture
parallele. Proc. Sem. Vector a Tens:>r. Moscow 2-3, 327-333 . . . . . . 322
1950, 1
- On the theory of a hypersurface in (n + 1)-dimensional equiaffine space.
Sem. Vektor Tensor An. 8, 273-285 (russian) . . . . . . . . . . . . . 269
1950, 2
- On the theory of a surface of n dimensions in an equicentroaffine space of
n + 2 dimensions. Sem. Vektor Tensor An. 8, 286-295 (russian) . . . . 269
1934, 1
*LovELL, C. A.: On certain associated metric spaces. Thesis Philadelphia.
19 pp. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148, 289
 Bibliography. 467

1951. 1
LYRA, G. : ber eine Modifikation der Riemannsehen Geometrie. Math.
Zeitsehr. 54, 52-64 . . . . . . . . . . . . . . . . . . . . . . . 134
1939. 1
MAEDA, J.: Einige Stze in der Kurventheorie im n-dimensionalen affinen
Raume. Thoku. Math. J. 45, 130-133 . . . . . . . . . . . . . . . . 232
1940, 1
- Sur la podaire d'une hypersurfaee dans l'espaee euelidien a n dimensions.
Tohoku. Math. J. 46, 360-384. . . . . . . . . . . . . . . . . . . . 242
MARTIN, W. F. (see S. BocHNER).
1943. 1
MASINI VENTURELLI, L.: Sopra i ds 2 di Liouville di classe uno. Ist. Veneto,
Cl. Sei. Mat. Nat. 102, 145-163 . . . . . . . . . . . . . . . . . . . 268
1943. 1
- Le ipersuperfieie di rotazione del tipo di Liouville. Ist. Veneto, Cl. Sei. Mat.
Nat. 102, 323-327 . . . . . . . . . . . . . . . . . . . . . . . . 268
1950, 1, 2, 3
MATSUMOTO, M.: Riemann spaces of class two and their algebraic characteri-
zation I, II, III. J. Math. Soc. Japan 2, 67-76; 77-86; 87-92 . . . . . 269

1951. 1
On the special Riemann spaees of class two. Mem. Coll. Univ. Kyoto Math.
26, 149-157 . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
1951. 2
- Affinely connected spaees of class one. Mem. Coll. Sei. Univ. Kyoto. Math.
26, 235-249 . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
1951, 3
- Conformally flat Riemann spaees of class one. J. Math. Soc. Japan 3, 306
to 309 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
1952, 1
- The class number of embedding of the spaee with projective conneetion.
J. Math. Soe. Japan 4, 37-58 . . . . . . . . . . . . . . . . . . . 269
1930, 1
MATTIOLI, G.: Su le ennuple di congruenze ortogonali a rotazioni costanti. Atti
Istituto Veneto 89, 97-105 . . . . . . . . . . . . . . . . 171
1930, 2
Sulla determinazione della varieta riemanniane ehe ammettono gruppi
semplicemente transitivi di movimenti. Rend. d. Line. (6) 11, 369-371 171

1930, 3
- Sopra eerti sistemi coordinati associati ad un ennupla di eongruenze e appli-
cazioni. Rend. Semenario Padova 1, 11 -132 . . . . . . . . . . . 155
1888, 1
MAURER, L.: ber gewisse Invariantensysteme. Bayr. Akad. d. Wiss. 18, 103
to 150 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
1939. 1
MAXIA, A.: Varieta anolonome immerse in una varieta a connessione affine
(X::- 1 in En affine). Mem. Soc. Roy. Sei. Lett. Boheme. 18 pp. . . 253
30*
468 Bibliography.

1939. 2
MAXIA, A.: Varieta anolonome associate ed una trasformazione dualistica.
Ist. Lombardo Rend. 72, 437-454 . . . . . . . . . . . . . . . . . . 253
1940, 1
Geometria proiettiva differenziale dei complessi anolonomi dirette. Atti
Sec. Congresso Un. Mat. Italia. Bologna 1940, 303-304 . . . . . . . . 253
1928, 1
MAYER, W. (see also C. BuRSTIN, A. DuscHEK): ber das vollstndige Formen-
system der F 1 im R,.. Mon. f. Math. u. Phys. 35. 87-110 . . 232, 263, 277, 278
1931' 1
Beitrag zur Differentialgeometrie, !-dimensionaler Mannigfaltigkeiten die
in euklidischen Rumen eingebettet sind. Sitz. Preu. Akad. 27, 606-615 277
1933. 1
Zum Tensorkalkl in Vektorrumen Riemannscher Mannigfaltigkeiten.
Mon. Math. u. Phys. 40, 283-293 . . . . . . . . . . . . . . 269
1935. 1
and T. Y. THOMAS: Foundations of the theory of Lie groups. Ann. of Math.
36, 77(}-822 . . . . . . . . . . . . . . . . . . . . . . . 185
1935. 2
Die Differentialgeometrie der Untermannigfaltigkeiten des Rn konstanter
Krmmung. Trans. Am. Math. Soc. 38, 267-309 . . 232,263,269,271, 277, 278
1937, 1
and T. Y. THOMAS: Fields of parallel vectors in non-analytic manifolds in
the large. Comp. Math. 5. 198-207 . . . . . . . . . . . . . . 125
1911' 1
*MEYER, W. F.: ber die Theorie benachbarter Geraden und einen verall-
gerneinerten Krmmungsbegriff. Teubner, Leipzig. 152 pp. 230
1930, 1
MrcHAL, A. D. (see also T. Y. THOMAs): Geodesie Coordinates of order r. Bull.
Amer. Math. Soc. 36, 541-546 . . . . . . . . . . . . . . 155
1930, 2
Scalar extensions of an orthogonal ennuple of vectors. Am. Math. Monthly
37. 529-533 . . . . . . . . . . . . . . . . . . . . . . . 155
1931, 1
N otes on scalar extensions of tensors and properties of local coordinates.
Proc. Nat. Acad. 17, 132-136 . . . . . . . . . . . . . . . . . . . 155
1931, 2
An operation that .generates absolute scalar differential invariants from
tensors. Thoku Math. J ourn. 34, 71-77 . . . . . . . . . . . . . . 155
194 5. 1
First order differentials of functions with arguments and values in topo-
logical Abelian groups. Revista Ci. Lima 47, 389-422 . . . . . . . . 340
194 7, 1
*- Matrix and tensor calculus with applications to mechanics, elasticity and
aeronautics. Wiley and Sons, New York. XIII + 132 pp. Chapman and
Hall, London.
 Bibliography. 469

1941. 1
MIKAMI, M. (see also H. HoMBU): On the theory of non-linear direction displace-
ments. J ap. J. Math. 17, 541-568 . . . . . . . . . . . . . . . . . 123

1941, 2
On parabclas in the generalized spaces. Jap. J. Math. 17, 185-200 . . . 300
1949, 1
- Projectively connected space and curves of the second order. Sugaku
Math. 1, 274-286 (japanese). . . . . . . . . . . . . . 322
1939. 1
MIKAN, M.: Die Rume von R. Knig, die den eingebetteten Mannigfaltigkeiteil
zugeordnet-sind, Acad. Tcheque Sei. Bull. Math. Nat. 40, 218-221 . . . 253

1950, 1
MIRA FERNANDES, A. DE: Le geodetiche degli spazi unitarii. 'Cniv. Lisboa
Revista Fac. Ci. A. Ci. Mat. 1, 173-186 . . . . . . . . . 415
1952, 1
MrsHRA, R. S.: Sur certaines courbes appartenant a un saus espace d'un espace
riemannien. Bull. Sei. Math. (2) 76, 77-84. . . . . . . . . . . . 250
1938, 1
MoDESITT, V.: Conformal transformations between riemannian spaces, AmeL
J ourn. of Math. (2) 60, 325-336 . . . . . . . . . . . . . . . 315
1943, 1
MoGHE, D, N.: On isotropic spaces. J. Univ. Bombay (N.S.) 12, 1-3 . . . . 228

1950, 1
MoGI, J.: A remark on recurrent curvature spaces. Kodai Math. Sem. Rep,
73-74 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 424
1930, 1
MorsrL, G. C.: Sur les proprietes affines et conformes des varietes nonholo-
nomes. Bull. Cern. 4, 15-21 . . . . . . . . . . . . . . . . . . . . 253
1940, 1
Sur les geodesiques des espaces de Riemann singuliers. Bull. Math. Soc.
Roumaine Sei. 42, 33-52 . . . . . . . . . . . . . . . . . . 133,227, 270
MooRE, C. L. E. (see E. B. WrLsoN).
1950,
MoREA u, J. J.: La symetrie de revolution en calcul tensoriel. C. R. 231, 1028
to 1029 . . . . . . . . . . . . . . . . . . . . . . . . . 35
1934. 1
MoRINAGA, K.: An extension of the parallelism in x~-m in X n. J ourn. Hiro-
shima Univ. {A) 4, 34,177-183 . . . . . . . . . . . . 138,140,253

1934. 2
- An extension of parallel displacements by matrices. J ourn. of Sc. Hiro-
shima A 2, 127-140 . . . . . . . . . . . . . . . . 123, 131, 340

1934, 3
On parallel displacementinan n-dimensional space to which N-dimensional
-vectorspaces are attached. Journ. of Sc. Hiroshima A 5. 15-30 . . . . 123
470 Bibliography.

1899. 1
*MUTH, P.: Theorie und Anwendung der Elementarteiler. Teubner, Leipzig.
236 pp. . . . . . . . . . . . . . . . . . . . . . . . . . 31
1939. 1
MuT6, Y. (see also K. YANO): On the equations of circles in a space with a con-
formal connection. Tensor 2, S-52 (japanese) . . . . . . . . . . . . . 315
1939. 2
On some properties of hypersurfaces in a conformally connected manifold.
Proc. Phys. Math. Soc. Japan 21, 615-625 . . . . . . . . . . . . . 315
1940, 1
On some properties of umbilical points of hypersurfaces. Proc. Imp. Acad.
Tokyo 16, 79--82 . . . . . . . . . . . . . . . . . . . . . . . . . 251
1940, 2
On a family of curves on a hypersurface in the riemannian space. Tensor
3. 39--40 (japanese) . . . . . . . . . . . . . . . . . . . . . . . . 242
1940, 3
On some properties of subspaces in a conformally connected manifold.
Proc. Phys. Math. Soc. Japan 22, 621-636 . . . . . . . . . . . . . . 315
1942, 1
Theory of subspaces in a space with a conformal connection. Tensor 5, 31
to 46 (japanese) . . . . . . . . . . . . . . . . . . . . 315
1951, 1
Some properties of a riemannian space admitting a simply transitive group
of translations. Tohoku Math. J. 2, 20 5-213 . . . . . . . . . . . . . 346
1924, 1
MYLLER, A.: Direzioni concorrenti sopra una superficie spieeanti dai punti di
una curva. Rend. d. Lincei 5, 33, 339--341 . . . . . . . . . . 322
1928, 1
Directions concourantes dans une variete metrique a n dimensions. Bull.
Soc. Math. 56, 1-6 . . . . . . . . . . . . . . . . . . . . . . 322
1949, 1
NAGABHUSHANAM, K.: An extension of a theorem of Wintner. Proc. Nat.
Inst. Sc. India 15, 35-36 . . . . . . . . . . . . . . . . . 89
1944, 1
NAKAE, T.: Das geometrische Objekt. Tensor 7. 1-5 (japanese) 67
1928, 1
NALLI, P.: Sopra le coordinate geodetiche. Rend. d. Linc. (6) 7, 195-198 155
1938. 1
NARLIKAR, V. V.: A note on riemannian tensors. Phil. Mag. VII, 26, 390-394 51
1948, 1
- and A. PRASAD: Canonical coordinates in general relativity. Bull. Calcutta
Math. Soc. 40, 123-128. . . . . . . . . . . . . . . . . . 155
NEUMANN, J. V. (see 0. VEBLEN).
1927, 1
NEWMAN, M. H. A.: A gauge-invariant tensorcalculus. Proc. Roy. Soc. 116,603
to 623 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
 Bibliography. 471

1918, 1
NoETHER, E.: Invarianten beliebiger Differentialausdrcke. Gttinger Nachr.
37--44 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
194 5. 1
NORDEN, A.: On pairs of conjugate parallel translationsinn-dimensional spaces.
C. R. Acad. Sc. URSS. 49, 625-628 . . . . . . . . . . . . . . . 123, 132
1945. 2
On an interpretation of a space'With degenerate metric. Akad. Nauk SSSR.
50, 57-60 (russian) . . . . . . . . . . . . . . . . . . . . . . . . . 133

1945, 3
The affine connectivity on the surfaces of a projective and conformal space.
C. R. Acad. Sc. URSS. 48, 539-541 . . . . . . . . . . 296, 301, 315

1947. 1
La connexion affine sur les surfaces de l'espace projectif. Rec. Math. 20,
263-281 (russian, french summary) . . . . . . . . . . . . . . . 296, 301
1948, 1
- The riemannian metric on surfaces of a projective space. Akad. Nauk SSSR.
60, 345-347 (russian) . . . . . . . . . . . . . . . . . . . . . . 296, 301
1949, 1
On the inner geometries of surfaces in a projective space II. Trudy Sem.
Vektor, Tensor. An. 7, 31-64 (russian) . . . . . . . . . . . . . 296,301
1950, 1
On conjugate connections. Trudy Sem. Vektor, Tensor. An. 8, 93-105
(russian) . . . . . . . . . . . . . . . . . . . . . . . 123
1950, 2
*- Prostranstva affinnoi svyaznosti. (Affinely connected spaces.) Gosudarstv.
Izdat. Tehn.-Teor. Lit., Moscow-Leningrad. 463 pp.
1944, 1
NoRDON, J.: Sur la Solution elementaire d'une equation aux derivees partielles
associee a un espace riemannien harmonique. C. R. 219, 436-438 . . . . 383
1950, 1
NoZISKA, F.: Le vecteur affinanormal et la connexion de l'hypersurface dans
l'espace affin. Casopis Pest. Mat. Tys. 7 5, 179-209 (czech summary) . . 235
1951, 1
NYENHUIS, A.: Xn-l forming sets of eigenvectors. Proc. Kon. Ned. Akad. Am-
sterdam 54= Indagationes Math. 13, 200-212 . . . . . . . . . . . 66, 248
1951, 2
- An application of anholonomic coordinates. Math. Centrum Amsterdam.
Rapp. Z. W. 1951,017. 6.pp. . . . . . . . . . . . . . . . . . . . 66
1952, 1
*- Theory of the geometric object. Thesis. Univ. Amsterdam. 238 pp. (dutch
sum1nary) 64,65,67,68,102,105,107,108,109,110,114,131,340,351,363,366

1953. 1
- On the holonomy groups of linear connections JA, I B. Proc. Kon. Ned.
Akad. Amsterda1n = Indagationes Math. 15, 233-249 . . . . . . . . . 367
472 Bibliography.

1936, 1
OBER TI, G.: Su una decomposizione carratteristica dei tensori doppi. Ist.
Lombardo, Rend. II, 69, 397-402 . . . . . . . . . . . . . . . . . 15
OKADA, Y. (see H. HoMBU).
1949, 1, 2
OTSUKI, T.: On some 4-dimensional riemannian spaces I, II. Mem. Fac. Sc.
Kysy Univ. 4, 193-212; 213-217 . . . . . . . . . . . . . 367

1950, 1
On projectively eonneeted spaees whose groups of holonomy fix a hyper-
quadrie. J. Math. Soe. Japan 1, 251-263 . . . . . . . . . . . . . 128,367

1950, 2
On the spaees with normal eonformal connexions and some imbedding pro-
blern of riemannian spaees I. Tohoku Math. J. 1, 194-224 . . . . 269, 315, 367

1950, 3
- Classifieation of 4-dimensional analytie Weyl spaces by their holonomy
groups. Mem. Fae. Sei. Kyusyu Univ. A5, 1-39 . . . . . . . . . . . . 367

1951, 1
On the spaces with normal eonformal eonnexions and some imbedding pro-
blems of riemannian spaees II. Tohoku Math. J. 2, 220-274. . . . . 269, 367

1952, 1
- On the spaees with normal projeetive eonnexions and some imbedding pro-
blems of riemannian spaees. Math. J. Okayama Univ. 1, 69-98. 367
1889, 1
PADOVA, E.: Sulle deformazioni infinitesimi. Rend. d. Line. (IV) 51, 174-178 146

1884, 1
PAINLEVE, P.: Memoire sur la transformationsdes equations de la dynamique.
Joum. de Math. (4) 10, 5-92 . . . . . . . . . . . . . . . . . . . . 310
1941, 1
PALATIN!, A.: Sopra le varieta di classe uno. Ist. Lombardo, Mat. Nat. Rend.
5, 293-304 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
1943, 1
PANTAZI, A.: Sur la deformation projeetive des surfaces non holonomes de
l'espace E 8 Bull. Math. Soe. Roumaine 45, 33-47 . . . . . . . . . . . 253

1947, 1
Sur les varietes nonholonomes a asymptotiques reetilignes et leur defor-
mation projeetive. Bull. Math. Soc. Roumaine 48, 3-26 . . . . . . . . 253
1946, 1
PAPY, G.: Sur les formes eubiques altemees de rang inferieur a 8. Bull. Soc.
Roy. Liege 15, 77-83 . . . . . . . . . . . . . . . . . . . 36
1946, 2
- and F. TouRNAY: Classifieation des formes eubiques altemees a six indeter-
minees par rapport au groupe eanonique de degre trois. Bull. Soe. Roy. Sei.
Liege 15, 513-523 . . . . . . . . . . . . . . . . . . . . . . . . . 36
PASTOR!, M. (see B. FINZI).
 Bibliography. 473

1951, 1
PATTERSON, E. M.: On symmetric recurrent tensors of the second order. Quart.
J. Math. Oxford 2, 151-158 . . . . . . . . . . . . . . 322
1951, 2
An existence theorem on simply harmonic spaces. J. London Math. Soc.
26, 238-240 . . . . . . . . . . . . . . . . . . . . . . . . . . . 383

1952, 1
Simply harmonic Riemann extensions. J. London Math. Soc. 27, 102-107
381,383
1952, 2
Some theorems on Ricci-recurrent spaces. J. London Math. Soc. 27, 287 to
295 . . . . . . . . . . . . . . . . . . . . . . . . . . . 421

1952, 3
- and A. G. WALKER: Riemann extensions. Quart. J. Math. Oxford ser. (2}
3, 1~28 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228

1937, 1
PAuc, C.: Etude geometrique d'un faisceau de transformations infinitesimales.
C. R. 204, 1027-1029; 205, 101-102 . . . . . . . . . . . . . . . . . 253

1938, 1
Extension aux varittes non holonomes v;:o- 1 , de quelques proprietes des
surfaces et des Va2 Atti Acad. Mat. Nat. (6) 27, 155-165 . . . . . . . . 253

1938, 2
- Interpretation geometrique dans les varietes non holonomes des theories
d'integration des systemes d'equations de Pfaff. C. R. 206, 885-888 . . . . 253

1935, 1
PEARS, L. R.: Bertrand curves in riemannian space. J oum. London Math. Soc.
10, 18o-183 . . . . . . . . . . . . . . . . . . . . . . . . . 343

1946, 1
PENSOV, Y. E.: Classification of differential geometric objects of the class v with
one component. C. R. Acad. Sc. URSS. 54, 563-566 . . . . . . . . . 67
1948, 1
- The classification of geometric differential objects with two components.
Akad. Nauk SSSR. 80, 537-540 . . . . . . . . . . . . . . . . 67
1950, 1
- The classification of continuous pseudo-groups of Lie transformations in
X 8 according to their characteristic objects. Trudy Sem. Vektor, Tenzor,
An. 8, 382-413 (russian) . . . . . . . . . . . . . . . . . . 64
1952, 1
PEREMANS, W., H. J. A. DUPARC and C. G. LEKKERKERKER: A property of posi-
tive matrices. Kon. Ned. Akad. Amsterdam Proc. 55 = Indagationes
Math. 14, 24-27 . . . . . . . . . . . . . . . . . . . . . 39

1934, 1
PEREPELKINE, D.: Sur les directions de Courbure d'une v". dans Rn. c. R.
199, 108~1091 . . . . . . . . . . . . . . . . . . . . . . . 269
474 Bibliography.

1935, 1
PEREPELKINE, D.: Sur les varietes paralleles dans un espace euclidien (ou
riemannien). C. R. Ac. Sc. URSS. 593-598; french text 596-598 269
1935, 2
- Sur la transformation conforme et la courhure riemannienne normale
intrinseque d'une Vm dans V,.. C. R. 200, 513-515 . . . . . . . . . . . 315
1936, 1
- Sur la courhure et les espaces normaux d'une Vm dans R,.. Rec. Math.
Moscow 42, 81-120 . . . . . . . . . . . . . . . . 269, 272, 275
PESCHL, E. (see R. KNIG).
1948, 1
PETKANTSCHIN, B.: Allgemeine isotrope Regelscharen im euklidischen Raum.
Ann. Univ. Sofia, Math. Phys. 44, 357-399 (bulgarian, german summary) 228
1943, 1
PETREscu, S. (see also K. YANO): Sur les invariants des hypersurfaces non
holonomes V,."- 1 . Disquisit. Math. Phys. 3, 3-84 . . . . . . . . . . . 253

1943, 2
:+
Sur les hypersurfaces non holonomes V22 1 Acad. Roum. Bull. Sect. Sc.
25, 249-253 . . . . . . . . . . . . . . . . . . . . . . . . . 253
1943, 3
:1:
Sur les hypersurfaces non holonomes V22 21 Acad. Roum. Bull. Sect. Sc.
25, 313-317 . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
1944, 1
- Quelques proprietes des espaces non honolomes A:', consideres dans un
espace A,., a connexion affine, admettant un parallelisme absolu. Bull.
Math. Soc. Roumaine 46, 145-154 . . . . . . . . . . . . . . . . . . 253

1945, 1
- Sur un problerne d'existence des espaces non holonomes A:' totalement
geodesiques dans un espace A,. a connexion affine symetrique. Disquisit.
Math. Phys. 4, 117-130 . . . . . . . . . . . . . . . . . . . . . . . 253

1946, 1
Quelques propriets conformes des sous espaces J.l, dans un Vm de V,..
Bull. Sc. Ecole Polytech. Timisoara 12, 167-174 . . . . . . . . . . . 315
1948, 1
Sur quelques proprietes conformes des espaces non holonomes V,:". Mathe-
matica, Timisoara 23, 108-122 . . . . . . . . . . . 253, 269, 315

1948, 1
PETROV, A. Z.: On the curvature of Riemann space. Akad. Nauk SSSR. (N. S.)
61, 211-214 (russian). . . . . . . . . . . . . . . . . . . . . . . . 138

1886, 1
DELL-PEzzo, P.: Sugli spaz tangenti ad una superficie o ad una varieta im-
mersa in uno spazio di piu dimensioni. Rend. Ace. Napoli 25, 176-180 . . 27fJ

1932, 1
PrNL, M.: Quasimetrik auf totalisotropen Flchen I, II. Proc. Kon. Akad.
Amsterdam 35, 1181-1188; 36, 550-557 . . . . . . . . . . . . . . 22'
 Bibliography. 475

1932. 2
PINL, M.: ber Kurven mit isotropen Schiegrumen im euklidischen Raum von
n Dimensionen. Mon. f. Math. u. Phys. 39, 157-172 . . . . . . . . . . 227
1935. 1
Quasimetrik auf totalisotropen Flchen Ill. Proc. Kon. Ned. Akad.
Amsterdam 38, 171-180 . . . . . . . . . . . . . . . . . . . 227

1937. 1
- Zur dualistischen Theorie isotroper und verwandter Kurven im euklidischen
Raum von n Dimensionen. Mon. f. Math. u. Phys. 44, 1-12 . . . . . . 227
1937. 2
- Zur Existenztheorie und Klassifikation totalisotroper Flchen. Compositio
Math. 5. 208-236 . . . . . . . . . . . . . . . . . . . . . . . 227
1940, 1
- W-Projektionen totalisotroper Flchen II. asopis Pest. Mat. Fys. 69,
23-35 (german, czech summary). . . . . . . . . . . . . . . . . 227
1940, 2
- Zur Theorie der halbisotropen Flchen in R,. J. D. M. V. 50, 65-78 . 227
1940, 3
- Zur dualistischen Theorie isotroper Kurven. Mon. f. Math. u. Phys. 49,
261-278 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
1947. 1
- ber Flchen mit isotropem mittlerem Krmmungsvektor. Ber. Math.
Tagung Tbingen 121-122 . . . . . . . . . . . . . . . . . . . . . 227
1948, 1
- ber Flchen mit isotropem mittlerem Krmmungsvektor. Monatsh. Math.
52, 301-310 . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
1949. 1
- Binre orthogonale Matrizen und integrallose Darstellungen isotroper Kur-
ven. Math. Ann. 121, 1-20 . . . . . . . . . . . . . . . . . . 227
1951, 1
Geodesie coordinates and rest systems for general linear connections.
Duke Math. J. 18, 557-562 . . . . . . . . . . . . 127,155
1887, 1
Po!NCARE, H.: Sur les residus des integrales doubles. Acta mathematica 6,
321-380 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
1895. 1
- Analysis situs. Joum. de I'Ec. Polyt. II, 1, t-123 97
PRASAD, A. (see V. V. NARLIKAR).
1953.
QuADE, W.: Beweis eines Satzes ber positive Matrizen. Proc. Kon. Nederl.
Akad. Amsterdam 56 (= Indag. Math. 15). SQ-St . . . . . . . . . . 39
1949. 1
RAcAH, G.: On the decomposition of tensors by contraction. Rev. Modem Phy-
sics 21, 494-496 . . . . . . . . . . . . . . . . . . . . . . . . . . 15
476 Bibliography.

1930, 1
RAcHEVSKY, P.: Sur les espaces sous-projectifs. C. R. 191, 547-548 317
1933. 1
Caracteres tensoriels de l'espace sous-projectif. Abh. des Seminars fr Vek-
tor und Tensoranalysis, Moskau 1, 126-140 . . . . . . . . . . . . 317, 329
1949, 1
Galois theory in fields of geometric objects. Trudy Sem. Vektor, Tenzor.
An. 7, 167-186 (russian) . . . . . . . . . . . . . . . . . . . . . . 105
1950, 1
Symmetrie spaces of affine connection with torsion I. Trudy Sem. Vektor,
Tenzor An. 8, 82-92 (russian) . . . . . . . . . . . . . . . . . . . . 163
1950, 2
*- Kurs differencial'noi geometrii (Acourse of differential geometry). 3rd ed.
Gosudarstv. lzdat. Tehn.-Teor. Lit. Moscow-Leningrad. 428 pp.
1954. 1
RENAUDIE, J.: Un theoreme sur les espaces harmoniques. C. R. 238, 199-201 388
1907, 1
*RErcHEL, W.: ber trilineare alternierende Formen in 6 und 7 Vernder-
lichen. Diss. Greifswald. 59 pp. . . . . . . . . . . . . . . 36
1950, 1
*RHAM, G. DE, and K. KoDAIRA: Harmonie integrals. lnst. for Adv. Study.
Princeton. N. J. 114pp. . . . . . . . . . . . . . . . . . . . . . . 27
1884, 1
Rrccr, G.: Princip di una teoria delle forme differenziali quadratiche. Ann. di
Mat. (II) 12, 135-167 . . . . . . . . . . . . . . . . . . . . . 121,143
1886, 1
- Sui parametri e gli invarianti delle forme quadratiche differentiale. Ann. di
Mat. (II) 14, 1-11 . . . . . . . . . . . . . . . . . . . . . 121
1886, 2
Sui sisterni di integrali independenti di una equazione lineare ed omogenea
a derivate parziali di 1 ordine. Rend. d. Linc. (IV) 2II, 119-122, 190-194.
Summary of 1887, 2 . . . . . . . . . . . . . . . . . . . . . . . . 121
1887, 1
Sulla derivazione covariante ad una forma quadratica differenziale. Rend.
d. Linc. (IV) 31, 15-18 . . . . . . . . . . . . . . . . . . . . . . . 121
1887, 2
Sui sistemi di integrali independenti di una equazione lineare ed omogenea
a derivate parziali di 1 ordine. Ann. di Mat. (2) 15, 127-159 . . . . . . 121
1888, 1
Sulla classificazione delle forme differenziali quadratiche. Rend. d. Linc.
(4) 41, 203-207 . . . . . . . . . . . . . . . . . . . . . . 121, 263, 278
1888, 2
*- Delle derivazione covariante e contravariante e delloro uso nell' Analisi ap-
plicata. Studie editi della Universita di Padova a commemorare l'ottavo
Centenario delle origina della Universita di Bologna, 111, Padova. 30 pp. 121
 Bibliography. 477

1889, 1
Ricci, G.: Sopra certi sistemi di funzioni. Rend. d. Linc. (IV) 51 , 112-118. 121
1889, 2
- Di un punto della teoria delleforme differenziali quadratiche temarie. Rend.
d. Linc. (IV) 51, 643-651 . . . . . . . . . . . . . . . . . . . . . . 121
1892, 1
- Resurne de quelques travaux sur les systemesvariables des fonctions associes
a une forme differentielle quadratique. Bull. Sc. Math. II 16, 167-189 . . 121
1893, 1
- Di alcune applicazioni del calcolo differenziale assoluto alla teoria delleforme
differenziale quadratiche binarie e dei sistemi a due variabili. Atti R. Istit.
Veneto (VII) 4, 1336-1364 . . . . . . . . . . . . . . . . . . . . . 121
1893. 2
- Dei sistemi di coordinate atti a ridurre la espressione del quadrato dell'ele-
mento lineare di una superficie alla forma ds 1 = (U + V) (du + dv 2 ). Rend.
d. Linc. (V) 21 , 73-81 . . . . . . . . . . . . . . . . . . . . . . . 121
1894, 1
Sulla teoria delle linee geodetiche e dei sistemi isotermi di Liouville. Atti
R. Istituto Veneto (VII) 5 (=53), 643-681 . . . . . . . . : . . . . . 121
1894, 2
Sulla teoria intrinseca delle superficie ed in ispecie di quelle di 2 grado.
Atti R. Istituto Veneto (VII) 5 (=53), 445-488 . . . . . . . . . . . 121
1895, 1
- Dei sistemi di congruenze ortogonali in una varieta qualunque. Mem. Ace.
Linc. (5) 2, 276-322 . . . . . . . . . . . . . . . . 121, 171, 228, 244, 245
1895, 2
- Sulla teoria degli iperspazi. Rend. d. Linc. (V) 4II, 232-237 . . . . . . 121
1897, 1
*- Lezioni sulla delle superficie. Frat. Drucker, Verona-Padova. VIII +
416 pp. (Lithographie 1898) . . . . . . . . . . . . . . . . . . . 122
1897, 2
Sur les systemes completement orthogonaux dans un espace quelconque.
c. R. 125, 810-811. 0 0 0 0 0 0 0 0 0122 0 0 0 0 0 0 0 0 0 0 0

1897, 3
- Dei teorema di Stokes in uno spazio qualunque a tre dimensioni ed in coordi-
nate generale. Atti Istituto Veneto (7) 8 (=56), 1536-1539, 1896-1897 122
1898, 1
Sur les groupes continus de mouvements d'une variete quelconque a trois
dimensions. C. R. 127, 344-346 . . . . . . . . . . . . . . . . . 122, 249
1898, 2
Sur les groupes continus de mouvements d'une variete quelconque. C. R.
127, 360-361 0 0 0 0 0 0 0 0 0 0 122, 249
0 0 0 0 0 0 0 0 0 0 0

1901, 1
*- and T. LEvi-CiviTA: Methades de calcul differentiel absolu et leurs appli-
cations. Math. Ann. 54, 125-201, Berichtigungen 608. Reprinted without
corrections. Collection de monographies scientifiques etrangeres, publiee
sous la direction M. G. Juvet, No 5. Blanchard, Paris 122, 165, 171, 244, 249
478 Bibliography.

1902, 1
Rrccr, G.: Sui gruppi continui di movimenti in una varieta qualunque a tre
dimensioni. Mem. d. Soc. Ital. (III) 12, 61-92 . . . . . . . . . . . . . 122
1902, 2
Formole fondamentali nella teoria generale di varieta e della loro curvatura.
Rend. d. Linc. (V) ur, 355-362 . . . . . . . . . . . . . . . . . 122,253

1903, 1
Sulle superficie geodetiche in una varieta qualunque e in particolare nella
varieta a tre dimensioni. Rend. d. Linc. (V) 12I 409-420 122, 146, 148, 251, 256

1904, 1
- Direzioni e invarianti principali di una varieta qualunque. Atti R. Istituto
Veneto (VIII) 6 ( = 63), 1233-1239 . . . . . . . . . . . 122, 146
1905, 1
Sui gruppi continui di movimenti rigidi negli iperspazzi. Rend. d. Linc.
(V) 14II, 487-491 . . . . . . . . . . . . . . . . . . 122
1910, 1
Sulla determinazione di varieta dotate di proprieta intrinseche date a priori.
Rend. d. Linc. (V) 191, 181-187; 19II, 85-90 . . . . . . . . . . . . . 122
1910, 2
Sulla determinazione di varieta ehe godone di proprieta intrinseche presta-
bilite. Atti Soc. Ital. 3, 477-480 . . . . . . . . . . . . . . 122
1912, 1
Di un metodo per la determinazione di un sistema completo di invarianti per
un dato sistema di forme. Rend. Palermo 33, 194-200 . . . . . . . . 122
1912, 2
Della trasformazione delle forme differenziali quadratiche. Rend. d. Linc.
(V) 21 1, 527-532 . . . . . . . . . . . . . . . . . . . . . . . . . . 122

1918, 1
Sulle varieta a tre dimensioni dotate di terne principali di congruenze geo-
detiche. Rend. d. Linc. (V) 27 I, 21-28, 75-87 . . . . . . . . . . . . 122

1918, 2
Della varieta a tre dimensioni con terne ortogonali di congruenze a rota-
zioni. Rend. d. Linc. (V) 27II, 36-44 . . . . . . . . . . . . . . . . . 122

1928, 1
RrcE, L. H.: Couche ranks in the general matrix. ] ourn. Math. a. Phys. Mass.
7. 93-96 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1854, 1
*RIEMANN, B.: ber die Hypothesen, welche der Geometrie zugrunde liegen
(Habilitationsschrift). Gtt. Abh. 13 (1868) 1-20; Ges. Werke, 2. Aufl.
(1892), 272-287. Neue Ausgabe mit Erluterungen von H. WEYL. Springer
1921. 47 pp. . . . . . . . . . . . . . . . . . . . . . . . . . 155,158

1861, 1
- Commentatio mathematica, qua respondere tentatur quaestioni ab Inma
Academia Parisiensi propositae etc. Gesamm. Werke, 2. Aufl. 1892, 391 to
423, mit Anmerkungen von DEDEKIND . . . . . . . . . . . . . . . 143
 Bibliography. 479

1904, 1
RIMINI, G.: Sugli spazi a tre dimensioni ehe ammettono un gruppo a quattro
parametri di movimenti. Ann. R. Scuola Norm. Pisa, 9, 57 pp. . . . . 251
1932, 1
RoBERTSON, H. P. (see also H. WEYL): Groups of motions in spaces admitting
absolute parallelism. Ann. of Math. 33, 496-520 . . . . . . . . . . . 351

1949, 1
RoGovoi, M. R.: On the projective differential geometry of nonholonomic sur-
faces in a three-dimensional space. Akad. Nauk SSSR. 66, 1055-1057
(russian) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253

1950, 1
- On the projective differential geometry of nonholonomic surfaces in a three-
dimensional space. Ukrain. Mat. Zumal 2, 102-116 (russian) . . . . . 253

1940, 1; 1941, 1; 1943, 1

RosENSON, N.: Sur les espaces riemanniens de classe 1, I, II, III. Bull. Acad.
Sc. URSS. Ser. Math. 4, 181-192; 5, 325-351; 7, 253-284 (russian, french
summary) 268
1912, 1
RoTHE, H.: ber Komplexgren 2-ter und (v- 2)-ter Stufe in einem Haupt-
gebiet v-ter Stufe und die durch sie bestimmten linearen Komplexe. Sitz.
Akad. Wien Ila, 121, 1015-1050 . . . . . . . . . . . . . . 36
1921, 1
*- Einfhrung in die Tensorrechnung. Seidel & Sohn, Wien. 180 pp.

1945, 1
RoZENFEL'D, B. A.: Theory of surfaces in symmetrical spaces. Bull. Acad. Sei.
URSS. Math. 9, 371-386 (russian, english summary) . . . . . . . . . 268

1947, 1
- The metric and affine connection in spaces of planes, spheres or quadrics.
Akad. Nauk SSSR. 57, 543-546 . . . . . . . . . . . . . . . . . 123, 125

1947, 2
- Geometrie differentielle des familles de plans a plusieurs dimensions. Bull.
Acad. Sei. URSS. Ser. Math. 11, 283-308 (russian, french summary) . . . 253

1948, 1
- Differential geometry of figures of symmetry. Akad. Nauk SSSR. 59, 1057
to 1060 (russian) . . . . . . . . . . . . . . . . . . . . . . . 123, 125
1949, 1
- On unitary and stratified spaces. Trudy Sem. Vektor Tenzor. An. 7, 260 to
275 (russian) . . . . . . . . . . . . . . . . . . . . . . . . . . . 397

1949, 2
- Unitary-differential geometry of families of Km in Kn. Mat. Sbomik 24,
53-74 (russian) .. . . . . . . . . . . . . . . . . . . . . . . . . . 412

1950. 1
- and A. A. ABRAMov: Spaces with affine connection and symmetric spaces.
Uspechi Matem. Nauk 5, 72-147 (russian) . . . . . . . . . . . . . . 163
480 Bibliography.

1950, 1
*RuBINOWICZ, W.: Wektory i tensory (vectors and tensors). Monogr. Matern.
Tom XXII, Warszawa-Wroclaw. III + 170 pp.
1930, 1
RusE, H. S. (see also E. T. CoPSON): Normal covariant derivatives. Proc.
London Math. Soc. (2) 33, 66-76 . . . . . . . . . . . . . . . 155
1930, 2
On the "elementary" solution of Laplace's equation. Proc. Edinb. Math.
Soc. (2) 2, 135-139 . . . . . . . . . . . . . . . . 382
1930, 3
Generalized solutions of Laplace's equation. Proc. Edinb. Math. Soc. (2)
2, 181-188 . . . . . . . . . . . . . . . . . . . . 382
1931, 1
On absolute partial differential calculus. Quart. J ourn. Math. Oxford,
Ser. 2, 190-202 . . . . . . . . . . . . 68
1931, 2
Normal covariant derivatives. Proc. London Math. Soc. (2), 33, 66-76 155
1931, 3
- Taylor's theorem in the tensor calculus. Proc. London Math. Soc. (2) 32,
87-92 . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
1936, 1
On the geometry of the electromagnetic field in general relativity. Proc.
London Math. Soc. (2) 41, 302-322 . . . . . . . . . . . . . . . . 35,146
1939, 1
Salutions of Laplace's equation in an n-dimensional space of constant cur-
vature. Proc. Edinb. Math. Soc. (2) 6, 24-45 . . . . . . . . . . . . 382
1944, 1
On the line-geometry of the Riemann tensor. Proc. Roy. Soc. Edinburgh
62, 64-73 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
1944, 2
Sets of vectors in a V4 defined by the Riemann tensor. J. London Math.
Soc. 19, 168-178 . . . . . . . . . . . . . . . . . 146
1945, 1
- The Riemann tensor in a completely harmonic V4 Proc. Roy. Soc .
Edinburgh 62, 156-163 . . . . . . . . . . . . . . . 385,386
1946, 1
- A. G. D. Watson's principal directions for a riemannian V4 Proc. Edin-
burgh Math. Soc. 7, 144-152 . . . . . . . . . . . . . . 146
1946, 2
- The five-dimensional geometry of the curvature tensor in a riemannian V4
Quart. J. Math. Oxford 17, 1-15 . . . . . . 146
1946, 3
On simply harmonic spaces. J. London Math. Soc. 21, 243-247 369,387
1948, 1
The self-polar Riemann complex for a V4 Proc. London Math. Soc. 50,
7 5-106 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
 Bibliography. 481

1948, 2
RusE, H. S.: On simply harmonic "kappa-spaces" of four dimensions. Proc.
London Math. Soc. 50, 317-329 . . . . . . . . . . . . . . . . . 369,421
1949, 1
On parallel fields of planes in a riemannian space. Quart. J. Math. Oxford
Ser. 20, 218-234 . . . . . . . . . . . . . . . . . . . . . . . 322,367
1949, 2
- Three-dimensional spaces of recurrent curvature. Proc. London Math. Soc.
so, 438-446 . . . . . . . . . . . . . . . . . . . . . . . . 369,421,422
1950, 1
Parallel planes in a Riemann Vn. Proc. Roy. Soc. Edinburgh 63, 78-92 322, 367
1951, 1
- The Riemann complex in a four-dimensional space of recurrent curvature.
Proc. London Math. Soc. 53, 13-31 . . . . . . . . . . . . . 421,423
1951, 2
A classification of K*-spaces. Proc. London Math. Soc. 53, 212-229 421, 423
1952, 1
Simply harmonic affine spaces of symmetric connection. Publ. Math. De-
brecen 169-174 . . . . . . . . . . . . . . . . . . . 381
1948, 1
*RUTHERFORD, D. E.: Substitutional analysis. Edinburgh Univ. Press.
103 PP . . 15
1950, 1
RYZKOV, V. V.: An imbedding theorem for riemannian geometries of higher
order. Akad. Nauk SSSR. 75, 503-506 (russian) . . . . . . . . . . . 268
1950, 1
RzEHNJNA, N. F.: On the theory of fields of local curves in Xn. Akad. Nauk
SSSR. 72, 461-464 (russian) . . . . . . . . . . . . . . . . . . . . 232
SAENZ, A. W. (see V. HLAVATY).
1945, 1
SAMUEL, P.: Sur les tenseurs a derivees covariantes nulles. C. R. 220, 16-162 367
1936, 1
SASAKI, S. (see also K. YANo): Some theorems of conformal transformations
of riemannian spaces. Proc. phys. math. Soc. Japan (3) 18, 572-578 311, 315
1937. 1
Contributions to the affine and projective differential geometries of space
curves. Jap. J. Math. 13, 473-481 . . . . . . . . . . . . . . . . . . 302
1939. 1
- A new proof of the theorem of J. A. Schonten on an umbilic surface in a
riemannian space. Tensor 2, 25-29 (japanese) . . . . . . . . . . 251, 309
1939, 2
On the theory of curves in a curved conformal space. Sc. Rep. Thuku Imp.
Univ. I, 27, 392-409 . . . . . . . . . . . . . . . . . . . 315
1939, 3
On the theory of surfaces in a curved conformal space. Tensor 2, 52-53
(japanese) . . . . . . . . . . . . . . . . . . . . 315
Schauten, Ricci-Calculus, 2. Auf!. 31
482 Bibliography.

1940, 1
SASAKI, S.: On a remarkable property of umbilical hypersurfaces in the
geometry of the normal conformal connexion. Sc. Rep. Thoku Imp.
Univ. 29, 412-422 . . . . . . . . . . . . . . . . . . . . . . . . . 309
1940, 2
Geometry of the conformal connexion. Sc. Rep. Thoku Imp. Univ. 29,
219-267 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
1940, 3
- On the theory of surfaces in a curved conformal space. Sc. Rep. Thoku
Imp. Univ. 28, 261-285 . . . . . . . . . . . . . . . . . . . 315
1941, 1
- On conformal normal Coordinates. Sc. Rep. Thoku Imp. Univ. 30, 71-80 155
1941, 2
- Geometry of conformal Connections. Tensor 4, 13-24 (japanese) . 315
1942, 1
~ On a relation between a riemannian space which is conformal with Einstein
spaces and normal conformally connected spaces whose groups of holonomy
fix a point or a hypersphere. Tensor 5, 66-72 (japanese) . . . . . 312, 367
1943, 1
- On the spaces with normal conformal connexions whose groups of holonomy
fix a point or a hypersphere I. Jap. J. Math. 18, 615-622 . . . . 334,367
1943, 2
- On the spaces with normal conformal connexions whose groups of holonomy
fix a point or a hypersphere Il. Jap. J. Math. 18, 623-633 . . . . 334, 367
1943, 3
On the spaces with normal conformal connexion whose groups of holonomy
fix a point or a hypersphere III. J ap. J. Math. 18, 791-795 . . . . 334, 367
1948, 1
*- Geometry of conformal connection. Kawade-shob, Tokyo. 3 + +
3
265 pp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
1949, 1
- and K. YANO: On the structure of spaces with normal projective connexions
whose groups of holonomy fix a hyperquadric or a quadric of (n- 2)-di-
mensions. Tohoku Math. J. 1, 31-39 . . . . . . . . . . . . . 128, 367
1949, 2
On the real representation of spaces with hermitian connexion. Sc. Rep.
Tohoku Univ. 33, 53-61 . . . . . . . . . . . . . . . . . . . . . 397
1951, 1
- An alternative proof of Liber's theorem. Proc. Japan Acad. 27, 73-80 . 367
1951, 2
- On a theorem conceming the homological structure and the holonomy
groups of closed orientable symmetric spaces. Proc. Japan Acad. 27, 81 to
85 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367,370
1950, 1
SATO, 1.: Unitary connected spaces with constant curvature. Bull. Yamag.
Univ. 2, 31-32 . . . . . . . . . . . . . . . . . . . . . . . . . . 404
 Bibliography. 483

1871, 1
ScHLFLI, L.: Nota alla memoria del sig. Beltrami. Ann. di Mat. (2) 5, 170-193 268
1930. 1
ScHAPIRO, H.: Sur !es espaces sous-projectifs. C. R. 191, 551-552 . . . . . 317
1933. 1
- ber die Metrik der subprojektiven Rurr:e. Abh. des Sem. Vektor u.
Tensor., Mcskau 1, 102-124 . . . . . . . . . . . . . . . 317, 330
1952, 1
ScHEIBE, E : ber einen verallgemeinerten affinen Zusammenhang. Math.
Zeitschr. 57. 65-74 . . . . . . . . . . . 287
ScHILD, A. (see J. L. SYNGE).
1939. 1
ScHIROKOV, P.: Sur les directions concourantes dans les espaces de Riemann.
Bull. Soc. Phys. Math. Kazan (3) 7, 77-87 . . . . . . . . . . . . . . 322
1940, 1
Symmetrisch-konformeuklidische Rume. Bull. Soc. Phys. Math. Kazan
(3) 11, 19-28 . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
1950, 1
Projectively euclidean symmetric spaces. Trudy Sem. Vektor Tenzor, An.
8, 73-81 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
1928, 1
ScHLESINGER, L.: Parallelverschiebung und Krmmungstensor. Math. Ann.
99. 413-434 . . . . . . . . . . . . . . . . . . . . . . . 123, 138, 340
1929, 1
ber Parallelverschiebung in der Weltgeometrie. Journ. f. d. r. u. a. Math.
161, 14-20 . . . . . . . . . . . . . . . . . . . . . . . . . . . 78,123
1931. 1
Ein Produktintegral mit Stetigkeitssprung. Journ. f. d. r. u. a. Math. 167,
73-78 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138, 340
1931. 2
- Neue Grundlagen fr einen Infinitesimalkalkl der Matrizen. Math.
Zeitschr. 33. 33-61 . . . . . . . . . . . . . . . . . . . . . . . . 340
1932, 1
- Weitere Beitrge zum Infinitesimalkalkl der Matrizen. Math. Zeitschr.
31, 485-501 . . . . . . . . . . . . . . . . . . . . . . . . . 138, 340
1949. 1
ScHMIDT, H.: Winkeltreue und Streckentreue bei konformer Abbildung Rie-
mannscher Rume. Math. Zeitschr. 51, 700-701 . . . . . . . 315
1902, 1
*ScHOUTE, P. H.: Mehrdimensionale Geometrie I. Gschen, Leipzig. 295 pp. 51
1918, 1
ScHOUTEN, J. A. (see also E. CARTAN): Die direkte Analysis zur neueren Rela-
tivittstheorie. Verh. Kon. Akad. Amsterdam, 12, Nr. 6. 95 pp.
58,59,122,123,304
1918, 2
On the nurober of degrees of freedom of the geodetically moving systems
and the enclosing euclidean space with the least possible nurober of dimen-
sions. ~oc. Kon. Akad. Amsterdam 21, 607-613 . . . . . . . . 81, 268,361
31*
4&4 Bibliography.

1919, 1
ScHCIUTEN, J. A. and D. J. STRUIK: On n-tuple orthogonal systems of (n- 1)-
dimensional manifolds in a general manifold of n dimensions. Proc. Kon.
Akad. Amsterdam 22, 596-605; 6&4-695 . . . . . . 58, 81,171,244,245,246
1921, 1
- ber die konforme Abbildung n-dimensionaler Mannigfaltigkeiteil mit
quadratischer Mabestimmung auf eine Mannigfaltigkeit mit euklidischer
Mabestimmung. Math. Zeitschr. 11, 5&-88 . . . . . 58, 251,306,307,309,311
1921, 2
and D. J. STRUIK: ber das Theorem von Malus-Dupin und einige ver-
wandte Theoreme in einern-dimensionalen Mannigfaltigkeit mit beliebiger
quadratischer Mabestimmung. Rend. Palermo 45, 313-331 . 58,244,249,310
1921, 3
and D. ]. STRUIK: On curvature and invariants of deformation of a Vm
in Vn. Proc. Kon. Akad. Amsterdam 24, 146-161 . . . . . 58,256, 269
1921, 4
- and D. ]. STRUIK: On some properties of general manifolds relating to Ein-
stein's theory of gravitation. Am. Journ. of Math. 43, 213-216 . . . 305, 314
1922, 1
- and D. ]. STRUIK: Einfhrung indieneueren Methoden der Differential-
geometrie. Christiaan Huygens I, 333-353; II, 1-24; 155-171; 291-306
(1923) . . . . . . . . . . . . . . . . . . . . . . 24, 58, 59,256,269,271
1922, 2
and D. J. STRUIK: ber Krmmungseigenschaften einer rn-dimensionalen
Mannigfaltigkeit, die in einer n-dimensionalen Mannigfaltigkeit mit belie-
biger quadratischer Mabestimmung eingebettet ist. Rend. Palermo 46,
165-184 . . . . . . . . . . . . . . . . . . . . . . . . 58, 256,269,271
1922, 3
ber die verschiedenen Arten der bertragung, die einer Differentialgeo-
metrie zugrunde gelegt werden knnen. Math. Zeitschr. 13, 56-81. Adden-
dum 15, 168 . . . . . . . . . . . . . . . . . . . . . . . . . . 60,123
1922, 4
and D. ]. STRUIK: Berichtigung zur Mitteilung ber das Theorem von Ma-
lus-Dupin. Rend. Palermo 46, 346 . . . . . . . . . . . . . . . . . . 310
1923, 1
ber die Einordnung der Affingeometrie in die Theorie der hheren ber-
tragungen. Math. Zeitschr. 17, 161-182, 183-188
8,123,144,235,237,238,239,253,256
1923, 2
ber die Bianchische Identitt fr symmetrische bertragungen. Math.
Zeitschr. 17, 111-115. . . . . . . . . . . . . . . . . . . . . . . . 147
1923, 3
and D. J. STRUIK: Note on Mr. Harward's paper on the identical relations
in Einstein's theory. Philos. Mag. 47, 584-585 . . . . . . . . . . . . 147
1923, 4
and D. J. STRUIK: U n theoreme sur la transformation conforme dans la geo-
metrie differentielle a n dimensions. C. R. 176, 1597-1600 . . . 310, 312, 315
 Bibliography. 485

1924, 1
*ScHOUTEN, J. A.: Der Ricci-Kalkl. Springer, Berlin. 312 pp. Here referred to
as RK 1924, 1 22,123,133,153,232,235,246,269,291,292,295,305, 309,310, 312

1924, 2
- On the place of conformal and projective geometry in the theory of linear
displacements. Proc. Kon. Akad. Amsterdam 27, 407-424 . 123, 302, 303, 316

1924, 3
Sur les connexions conformes et projectives de M. Cartan et Ia connexion
lineaire generale de M. Knig. C. R. 178, 2044-2046 . . . . . 123, 303, 316

1925, 1
- On the conditions of integrability of covariant differential.equations. Trans.
Am. Math. Soc. 27, 441-473. . . . . . . . . . . . . . . . 57,87

1925, 2
- Projective and conformal invariants of half-symmetrical connexions. Proc.
Kon. Akad. Amsterdam 29, 334-336 . . . . . . . . . . . . . . . 289, 306

1926, 1
Erlanger Programm und bertragungslehre. Neue Gesichtspunkte zur
Grundlegung der Geometrie. Rend. Palermo 50, 142-169 . . . . . 123, 364

1927, 1
- Vorlesungen ber die Theorie der halbeinfachen kontinuierlichen Gruppen.
Univ. of Leiden. 285 pp. (opolographic) . . . . . . . . . . . . . . 3, 187

1927. 2
- On infinitesimal deformations of Vm in Vn. Proc. Kon. Ned. Akad. Amster-
dam 31, 208-218 . . . . . . . . . . . . . . . . . . . . . . . . 68, 352

1927, 3
- bern-fache Orthogonalsysteme in Vn. Math. Zeitschr. 26, 706-730 246,247,307

1928, 1
- On non holonomic Connections. Proc. Kon. Akad. Amsterdam 31. 291-299
78,123,253
1928, 2
- ber die Umkehrung eines Satzes von Lipschitz. Nieuw Archief v. Wisk.
(2) 15. 97-102 . . . . . . . . . . . . . . . . . . . . . . . 310

1929. 1
- ber unitre Geometrie. Proc. Kon. Akad. Amsterdam 32, 457-465
52,389,391,393,396,412
1929. 2
- and V. HLAVATY: Zur Theorie der allgemeinen linearen bertragung. Math.
Zeitschr. 30, 414-432 . . . . . . . . . . . . . . . . . . . .11,129,135

1929. 3
- ber nicht-holonome bertragungen in einer Ln. Math. Zeitschr. 30, 149 to
172 . . . . . . . . . . . . . . . . . . . . . . . . 78,253,261,263,270

1929, 4
- Zur Geometrie der kontinuierlichen Transformationsgruppen. Math. Ann.
102, 244-272 . . . . . . . . . . . . . . . . . . . . . . . . . 142,187
486 Bibliography.

1929, 5
ScHaUTEN, J.A. and D. v. DANTZIG: ber die Differentialgeometrie einer her-
miteschen Differentialform und ihre Beziehungen zu den Feldgleichungen
der Physik. Proc. Kon. Akad. Amsterdam 32, 60-64 . . . . . . 389, 391,396
1930, 1
- and D. VAN DANTZIG: ber unitre Geometrie. Math. Ann. 103, 319-346
52,389,390,395,397,399,402,403,412,414,415
1930, 2
- and E. R. VAN KAMPEN: Zur Einbettungs-und Kriimmungstheorie nicht-
holonomer Gebilde. Math. Ann. 103, 752-783 3, 78, 253, 254, 261, 263, 278
1930, 3
- and S. GoLAB: ber projektive bertragungen und Ableitungen I.
Math. Zeitschr. 32, 192-214; Il. Ann. Math. Pure Appl. (4) 8, 141-157
301,303,322
1931, 1
- Klassifizierung der alternierenden Gren dritten Grades in 7 Dimensionen.
Rend. d. Palermo 55, 137-157. . . . . . . . . . . . . . . . . . . . 36
1931, 2
- and E. R. VAN KAMPEN: ber die Krmmung einer Vm in Vn; eine Revision
der Krmmungstheorie. Math. Ann. 105, 144-159 . . . . . . 277,278,280
1931, 3
- and D. VAN DANTZIG: ber unitre Geometrie konstanter Krmmung.
Proc. Kon. Akad. Amsterdam 34, 1293-1314 . . . . . . . . 403, 404, 415
1932, 1
- and D. VAN DANTZIG: Generelle Feldtheorie I: Zum Unifizierungsproblem
der Physik. Skizze einer generellen Feldtheorie. Proc. Kon. Akad. Am-
sterdam 35, 642-656 . . . . . . . . . . . . . . . . . . . . . . . . 303
1932, 2
- and D. VAN DANTZIG: Generelle Feldtheorie III. Z. f. Phys. 78, 639-667 303
1933, 1
- and E. R. v AN KAMPEN: Beitrge zur Theorie der Deformation. Prac. Mate-
matyczno-Fizycznych, Warszawa 41, 1-19
102,103,104,107,274,.335,336,337,345,346,354
1933, 2
- and D. VAN DANTZIG: General field theory VI: On projective connexions
and their applications to the general field theory. Ann. of Math. 34, 271-312 301
1934, 1
- and D. VAN DANTZIG: Was ist Geometrie? Abh. Sem. Vektor u. Tensor.,
Moskau 2-3, 15-50 . . . . . . . . . . . . . . . . . . . . . . . . 65
1935, 1
*- and D. ]. STRUIK: Einfhrung indieneueren Methoden der Differential-
geometrie, Vol. I by J. A. ScHaUTEN, NooRDHOFF, Groningen, here re-
ferred to as EI 1935, 1. Russian translation Moscow 1939
31,36,52,54,56,57,59,60,68,99,102,103,107,126,135,
137,142,143,148,153,155,165,166,169,171,227,246,265,327,335,336,340
1935, 2
- and J. HAANTJES: Konforme Feldtheorie und Spinraum. Ann. di Pisa 2,
4. 175-189 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 53
 Bibliography. 487

1935, 3
ScHOUTEN, J. A. and J. HAANTJES: ber allgemeine konforme Geometrie in
projektiver Behandlung I, II. Proc. Kon. Akad. Amsterdam 38. 706-708;
39. 27 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316
1936, 1
- and J. HAANTJES: Zur allgemeinen projektiven Differentialgeometrie.
Compositio Math. 3, 1-51. . . . . . . . . . . . . . . . 296, 301, 303, 327
1936. 2
and J. HAANTJES: Beitrge zur allgemeinen (gekrmmten) konformen Dif-
ferentialgeometrie!, II.Math.Ann. 112, 594-629; 113,568-583 (cf. 1935. 3)
314,315,316
1937. 1
*- and J. HAANTJES: On the theory of the geometric object. Proc. London
Math. Soc. (2) 42, 356-376 . . . . . . . . . . . . . . . . . . . . . 67
1938, 1
- ber die geometrische Deutung von gewhnlichen p-Vektoren und W- p-
Vektoren und den korrespondierenden Dichten. Proc. Kon. Akad. Am-
sterdam 41, 709-716 . . . . . . . . . . . . . . . . . . . . . . . . 27
1938, 2
*- and D. J. STRUIK: Einfhrung in dieneueren Methoden der Differential-
geometrie, Vol. II by D. J. STRUIK, NooRDHOFF, Groningen, here re-
ferred to as E II 1938, 1. Russian translation Moscow 1948
102,171,227,230,232,234,242,244,245,246,249,250,253,261,265,268,271
272,274,275,277,278,290,291,292,296,302,307,309,310,311,312,313,314
317,318,329,330,335,348,352,354,395,396,398,402,403,404,406,412,414
1940, 1
- and D. VAN DANTZIG: On ordinary quantities and W-quantities. Classifi-
cation and geometrical application. Comp. Math. 7, 447-473 . . . . . 27, 97
1940, 2
- ber Differentialkomitanten zweierkontravarianter Gren. Proc. Kon.
Ned. Akad. Amsterdam 43, 449-452 . . . . . . . . . . . . . . . . 67
1947, 1
- Regular systems of equations and supernumerary coordinates. Scripturn VI.
Math. Centre of Amsterdam. Here referred to as R. S. 1947, 1 64, 72, 74, 75
1949, 1
*- and W. v. D. KuLK: Pfaffs problern and its generalizations. Clarendon
Press, Oxford. Here referred to as P. P. 1949, 1 . . . . . . . 4, 22, 23, 26,
34, 35,38, 64, 67, 72, 74, 75, 76, 79, 80, 81, 82, 85, 88,89,102,107,110,214,291
1949, 2; 3; 4; 1950. 1
- On the geometry of spin spaces I, II, III, IV. Proc. Kon. Ned. Akad. Am-
sterdam 52, 597-609; 687-695; 938-948; 53, 261-272. Indagationes Math.
11, 178-190; 217-225; 336-346; 12, 41-52 . . . . . . . . . . . 53,291
1949. 5
*- Lie's differential operator. Math. Centrum. Amsterdam. Rapport Z.W.
1949-010. 7 pp. (dutch) . . . . . . . . . . . . . . . . . . . . . . 104
1951, 1
*- Tensor analysis for physicists. Clarendon Press, Oxford. X + 275 pp.
27,44,53,54,60,98,101,102,107,112
488 Bibliography.

1951, 2
ScHOUTEN, J. A.: Sur !es tenseurs de vn aux directions principales vn-rnor-
males. Colloque de Geometrie Differentielle, Louvain, 67-70 . . . . . . 248
1953, 1
On subprojective affine connections. Proc. Kon. Nederl. Akad. Amsterdam
56 (= Indag. Math. 15) 72-79 . . . . . . . . . . . . . . 319,320,321,324
1929, 1
ScHRENZEL, E.: ber Kurven mit isotropen Normalen. Sitz. Akad. Wien 138,
439-446 . . . . . . . . . . . . . . . . . . . . . . . . . 227
1947, 1; 1948, 1; 2
ScHRDINGER, E.: The final affine field laws I, II, III. Proc. Roy. Irish Acad.
51, 163-171; 205-216; 52,1-9 . . . . . . . . . . . . . . . . . . 114,179
1950, 1
*-Space-time structure. Cambr. Univ. Press. 119 pp. . . . . . . . 114, 179
1951, 1
On the differential identities of an affinity. Proc. Roy. Irish Acad. 54, 79 to
85 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114,117
1951, 2
- Studies in the non-symmetric generalization of the theory of gravitation I.
Com. Dublin Inst. Adv. Studies 6, 28 pp. . . . . . . . . . . . . . . . 179
1886, 1
ScHUR, F.: ber den Zusammenhang der Rume konstanten Krmmungs-
maes mit den projektiven Rumen. Math. Ann. 27, 537-567 . . . 153, 244
1886, 2
ber die Deformation der Rume konstanten Riemannschen Krmmungs-
maes. Math. Ann. 27, 163-176 . . . . . . . . . . . . . . . . . . . 244
1887, 1
- ber die Deformation eines dreidimensionalen Raumes in einem ebenen
vierdimensionalen Raume. Math. Ann. 28, 343-353 . . . . . . . . . 244
1941, 1
ScHWARTZ, A.: The Gau-Codazzi-Ricci equations in riemannian manifolds.
J. Math. Phys. Mass. Inst. Tech. 20, 3-79 . . . . . . . . . . 269, 279
1946, 1
On higher normal spaces for Vm in Sn. Am. J. Math. 68, 66-666 268, 275
1944, 1; 1945, 1; 1946, 1
SEN, R. N.: On parallelism in riemannian space, I, II, III. Bull. Calcutta Math.
Soc. 36, 102-107; 37, 153-159; 38, 161-167 . . . . . . . . . . . 132
1948, 1; 1949, 1; 2
- Paralleldisplacement and scalar product of vectors I, li, 111. Proc. Nat.
Inst. Sei. India 14, 45-52; Bull. CalcuttaMath. Soc. 41, 41-46; 113-120 . 132
1950, 1
- Corrections to my paper on "Parallel displacement and scalar product of
vectors III ". Bull. Calcutta Math. Soc. 42, 56 . . . . . . . . . . . . 132
1950, 2
On an algebraic system generated by a single element and its application
in riemannian geometry. Bull. Calcutta Math. Soc. 42, 1-13 . . . . . . 132
 Bibliography. 489

1942, 1
SHABBAR, M.: One parameter groups of deformations in riemannian spaces. J.
Indian Math. Soc. 6, 186-191 . . . . . . . . . . . . . . . . . . 108,335
1950, 1
SHANKS, E. B.: Homothetic correspondences between riemannian spaces. Duke
Math. ].17, 299-311 . . . . . . . . . . . . . . . . . . . . . . . . 346
1934, 1
SIBATA, T.: An extension of the definition of vector and parallel displacement.
J. Sc. Hirosh. Univ. 5. 83-97 . . . . . . . . . . . . . . 123
1905, 1
SINIGALLIA, L.: Sugli invarianti differenziali. Rend. Palermo 19, 161-184 22
1927. 1
SLEBODZINSKI, W.: Sur une classe d'espaces riemannien a trois dimensions.
Ann. Soc. Polon. 6, 54-82. . . . . . . . . . . . . . . . . . . . . . 248
1930, 1
Sur une generalisation des coefficients de Ricci. C. R. Varsovie 23, 14 7-150 171
1931. 1
Sur les equations de Hamilton. Bull. Acad. Roy. d. Belg. (5) 17, 864-870 102
1932. 1
Sur les transformations isomorphiques d'une variete a connexion affine.
Prac. Matern. Fiz. Warszawa 39. 55-62 . . . . . . . . . . . . . . . . 346
1951. 1
* SoKOLNIKOFF, I.
S.: Tensoranalysis theory and applications. Wiley and Sons,
New York; Chapman and Hall, London. IX + 335 pp.
1950, 1
SouRIAU, J. M.: Le calcul spinoriel et ses applications. Recherche Aeronauti-
que 1950, 14, 3-8 . . . . . . . . . . . . . . . . . . . . . . . . . 30
1950, 1
SPRINGER, C. E.: Union curves of a hypersurface. Canadian J. Math. 2, 457 to
460 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
1951. 1
STELLMACHER, K. L.: Geometrische Deutung konforminvarianter Eigenschaf-
ten des Riemannschen Raumes. Math. Ann. 123, 34-52 . . . . . . 308, 309
1949, 1
STRAUS, E. G. (see also A. EINSTEIN): Some results in Einstein's unified field
theory. Rev. Modem Physics 21, 414-420 . . . . . . . . . . . . . . . 179
1941, 1
STRUBECKER, K.: Differentialgeometrie des isotropen Raumes I. Theorie der
Raumkurven. Sitz. Akad. Wiss. Wien 150, 1-53 . . . . . . . . . . . 228
1942, 1
Differentialgeometrie des isotropen Raumes. II. Die Flchen konstanter
Relativkrmmung K = rt-s. Math. Zeitschr. 47. 743-777 . . . . . . . 228
1942, 2
- Differentialgeometrie des isotropen Raumes. III. Flchentheorie. Math.
Zeitschr. 48, 369-427. . . . . . . . . . . . . . . . . . . . . . . . 228
490 Bibliography.

1944, 1
STRUBECKER, K.: Differentialgeometrie des isotropen Raumes. IV. Theorie der
flchentreuen Abbildungen der Ebene. Math. Zeitschr. 50, 1-92 . . . . . 228
1922, 1
*STRUIK, D. J. (see also J. A. ScHOUTEN): Grundzge der mehrdimensionalen
Differentialgeometrie in direkter Darstellung. Springer, Berlin. 198 pp.
58,59,81,230,245,249,250,251,256,263,274,305
1928, 1
On the geometry of linear displacement. Publ. Mass. Irrst. of Techn. II 135,
522-564 . . . . . . . . . . . . . . . . . . . . . . . . . 123
1934, 1
* Theory of linear connections. Erg. der Math. III, 2. Springer, Berlin.
68 pp.
1950, 1
*- Lectures on classical differential geometry. Addison-Wesley Press. Inc
Cambridge Mass. VIII + 221 pp.
1905, 1
STUDY, E.: Krzeste Wege im komplexen Gebiet. Math. Ann. 60, 321-378 406
1943, 1
Su, B.: The projective differential geometry of a nonholonomic hypersurface.
Duke Math. J. 10, 575-586 . . . . . . . . . . . . . . . . . . . . . 253
1949, 1
Lie derivation in the geometry of conformal connections. Acad. Sinica Sc.
Record 2, 331-339 . . . . . . . . . . . . . . . . . 346
1951, 1
SuGURI, T.: Theory of curves in the unitary Kn-connected spaces. Mem. Fac.
Sc. Kyusyu Univ. 6, 31-40 . . . . . . . . . . . . . . . . . . . . . 415
193 7, 1
SuN, T.: On the complete set of identities for affine metric space. Thoku Math.
Journ. 43, 261-266; 44, 32-38 . . . . . . . . . . . . . . 144
1922, 1
SYNGE, J. L.: On principal directions in a riemannian space. Proc. Nat. Acad.
8, 198-207 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
1924, 1
Parallel propagation of a vector around an infinitesimal circuit in an affine-
connected manifold. Ann. of Math. 25, 181-184 . . . . . . . . . . . 138
1924, 2
Normals and curvatures of a curve in a riemannian manifold. Proc. Int.
Math. Congr. Toronto I (publ. 1928), 857-862 . . . . . . . . 229,230
1926, 1
The first and second variations of the length integral in riemannian space.
Proc. London Math. Soc. (2) 25, 247-264 . . . . . . . . . . . . . . . 352
1926, 2
On the geometry of dynamics. Trans. Roy. Soc. London A, 226, 33-106 352
1928, 1
Geodesics in non-holonomic geometry. Math. Ann. 99, 738-7 51 . . . 78, 253
 Bibliography. 491

1928, 2
SYNGE, ] . L., and A.]. McCoNNELL: Riemannian null-geometry. Phi!. Mag. (7)
5. 241-263 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
1930. 1
A characteristic function in riemannian space and its application to the
solution of geodesie triangles. Proc. London Math. Soc. (2) 32, 241-258 382, 385

1934. 1
- On the deviation of geodesics and nullgeodesics, particularly in relation to
the properties of spaces of constant curvature and indefinite line-element.
Ann. of Math. 35. 705-713 . . . . . . . . . . . . . . . . . . . . . 352
1949. 1
*- and A. ScHILD: Tensor calculus. Math. Expos. No. 5. University of Toronto
Press, Toronto. XII + 324 pp.
1932, 1
SvPTAK, M.: Sur les hypercirconferences et hyperhelices dans les espaces eucli-
diens a p dimensions. C. R. 195. 298-299 . . . . . . . . . . . . . . 230

1934, 1
Sur les hypercirconferences et hyperhelices generalisees dans les espaces
euclidiens a p dimensions. c. R. 198, 1665-1667 . . . . . . . . . . . 230

1949. 1
TAcHIBANA, S.: On normal coordinates of a Riemann space, whose holonomy
group fixes a point. Thoku Math. J. 1, 26-30 . . . . . . . . . . 128, 367

1951, 1
On pseudo-parallelism in Einstein spaces. Thoku Math. J. 2, 214-219 148

1951, 2
On concircular geometry and Riemann spaces with constant scalar curva-
tures. TOhoku Math. J. 3, 149-158 . . . . . . . . . . . . . . . . . 334
TAIT, P. G. (see W. THOMSoN).
TAKANO, K. (see K. YANo).
1953. 1
TAKAsu, T.: Connexion spaces in the large I-VI. Yokohama Math. Journ. 1,
1-87 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123, 253
1942, 1
TAKENO, H.: On the geometry of m-vectors. ]. Sc.Hiroshima Univ.12,109-123 138
1951, 1
T AKIZAWA, S.: On generalized spaces which adrnit given holonomy groups.
Mem. Coll. Sc. Uni'v. Kyoto Math. 26, 199-209 . . . . . . . . . . . . 367
1950, 1
TASHIRO, Y.: Sur Ia derivee de Liede l'etre geometrique et son groupe d'inva-
riance. T6hoku Math. J. 2, 166-181 . . . . . . . . . . . . . . . 105,107
1952. 1
Note sur Ia derivee de Lie d'un etre geometrique. Math. ] . Okayama Univ.
1, 125-128 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
1941, 1
*THALER, H.: Isotrope Projektion isotroper Kurven. These Innsbruck, 27 pp. 228
492 Bibliography.

1925, 1
THOMAS, J. M. (see also 0. VEBLEN): Conformal correspondence of Riemann
spaces. Proc. Nat. Acad. 12, 352-359 . . . . . . . . . . . . . . . 315
1925, 2
- Note on the projective geometry of paths. Proc. Nat. Acad. 11, 207-209 300
1926, 1
- Asymetric displacement of a vector. Trans. Am. Math. Soc. 28, 658-670
127,287
1926, 2
- On normal coordinates in the geometry of paths. Proc. Nat. Acad. 12, 58 to
63 . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
1926, 3
- Conformal invariants. Proc. Nat. Acad. 12, 389-393 315
1927, 1
- On systems of total differential equations. Ann. of Math. 28, 379-385 88
1922, 1
THOMAS, T. Y. (see also W. MAYER, 0. VEBLEN): On the equi-projective geome-
try of paths. Proc. Nat. Acad. 8, 592-594 . . . . . . . . . . . . . . 300
1925, 1
- On Weyl's treatment of the parallel displacement of a vector around an
infinitesimal closed circuit in an affinely connected manifold. Ann. of
Math. 27, 25-28 . . . . . . . . . . . . . . . . . . . . . . . . . 138
1925, 2
- Note on the projective geometry of paths. Bull. Am. Math. Soc. 31, 318 to
322 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
1925, 3
- On the projective and equi-projective geometries of paths. Proc. Nat. Acad.
11,198-203 . . . . . . . . . . . . . . . . . . . . . 300,301,302,303
1926, 1
- The identities of affinely connected manifolds. Math. Zeitschr. 25, 714-722 78
1926, 2
- A projective theory of affinely connected manifolds. Math. Zeitschr. 25,
723-733 . . . . . . . . . . . . . . . . . 155,165,300,303
1926, 3
- On conformal geometry. Proc. Nat. Acad. 11, 722-725 315
1927, 1
- Appell on tensor calculus. Bull. Am. Math. Soc. 33, 493-495 . 172
1927, 2
- and D. MICHAL: Differential invariants of affinely connected manifolds.
Ann. of Math. (2) 28, 196--236 . . . . . . . . . . . . . . . . . . 155,165
1927, 3
- The replacement theorem and related questions in the projective geometry
of paths. Ann. of Math. (2) 28, 549-561. . . . . . . . . . . 155, 165, 300
1928, 1
- Atheorem concerning the affine Connections. Am. Joum.of Math. 50, 518 to
520 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
 Bibliography. 493

1929, 1
THOMAS, T. Y.: Determination of affine and metric spaces by their differential
invariants. Math. Ann. 101, 713-728. . . . . . . . . . . . . . . . . 165
1931, 1
*- The elementary theory of tensors with applications to geometry and me-
chanics. McGraw-Hill, London. 122 pp.

1932, 1
- Conformal tensors. Proc. Nat. Acad. 18, 103-112; 189--193 . . . . . 315,316
1934. 1
- and J. LEVINE: On a class of existence theorems in differential geometry.
Bull. Am. Math. Soc. 40, 721-728 . . . . . . . . . . . . . . 87,133
1934. 2
*- The differential invariants of generalized spaces. Cambr. Univ. Press.
251 pp. . . . . . . . . . . . . . 87,144,156,159,160,165,300,303,315,316
1935. 1
- Algebraic characterization of complex differential geometry. Trans. Am.
Math. Soc. 38, 501-514; Bull. Am. Math. Soc. 41, 349 . . . . . . 133

1936, 1
Fields of parallel vectors in the large. Compositio Math. 3, 453-468 . . . 125

1936, 2
On the metric representation of affinely connected spaces. Bull. Am. Math.
Soc. 42, 77-78 . . . . . . . . . . . . . . . . . . . 133
1936. 3
On normal coordinates. Proc. Nat. Acad. 22, 309--312 . . . . . . 155
1936, 4
Riemann spaces of class one and their characterization. Acta Math. 67,
169--211 . . . . . . . . . . . . . . . . . . . . . . . . . . . 268, 269
1938. 1
- New theorems in Riemann-Einstein spaces. Rec. Math. Moscow 3. 331 to
340 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
1938. 2
Extract from a letter by E. Cartan concerning my note: On closed spaces
of constant mean curvature. Am. J oum. Math. 59. 793-794 . . . . . . 251

1938, 3
- Ares in affinely connected spaces. Ann. o. Math. Princeton (2) 38. 120-130 232

1939. 1
Reducible Riemann spaces and their characterization. Mon. f. Math. u.
Phys. 48, 228-292 . . . . . . . . . . . . . . . . . . . 285, 286

1939. 2
-- The decomposition of Riemann spaces in the large. Mon. f. Math. u. Phys.
47, 388-418 . . . . . . . . . . . . . . . . . . . . . . . . . 285, 286

1939. 3
Imbedding theorems in differential geometry. Bull. Am. Math. Soc. 45,
841-850 . . . . . . . . . . . . . . . . . . . . . . . . . 269
494 Bibliography.

1939. 4
THoMAs, T. Y., and E. W. TITT: On the elementary solution of the general
linear differential equation of the second order with analytic coeffients.
J. Math. Pures et Appl. 18, 21 7-248 . . . . . . . . . . . . . . . . . 386
1879. 1
*THOMSON, W., and P. G. TAIT: Treatise on natural philosophy. Cambr. Univ.
Press . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
TRULLEN, P. (see H. BEHNKE).
1923, 1
TIETZE, H.: ber die Parallelverschiebung in Riemannschen Rumen. Math.
Zeitschr. 16, 308-317. . . . . . . . . . . . . . . . . . . . . . . . 138
TITT, E. W. (see T. Y. THOMAs).
ToMoNAGA, Y. (see K. YANO).
1939. 1
ToMPKINS, C.: Isometrie imbedding of flat manifolds in euclidian space.
Duke Math. J ourn. 5, 58-61 . . . . . . . . . . . . . . . 269

1941, 1
A flat Klein bottle isometrically embedded in euclidean 4-space. Bull.
Am. Math. Soc. 47, 508. . . . . . . . . . . . . . . . . 268,269
1949, 1
ToNNELAT, M. A.: Theorie unitaire du champ physique. I. Les tenseurs fonda-
mentaux et la connection affine. C. R. 228, 368-370 . . . . . . . . . 179

1950, 1
- Resolution des equations fondamentales d'une theorie unitare purement
affine. C. R. 230, 182-184 . . . . . . . . . . . . . . . . . . . . . 179
1950, 2
- Theorie unitaire affine. I. Choix des tenseurs de base et obtention de l'equa-
tion fondamentale. C. R. 231, 47D-472 . . . . . . . . . . . . . . . . 179

1950, 3
Theorie unitaire affine. II. Resolution rigoureuse de l'equation fondamen-
tale. C. R. 231, 487-489 . . . . . . . . . . . . . . . . . . 179
1950, 4
- Theorie unitaire affine. III. Les equations du champ. C. R. 231, 512-514 179

1951, 1
- Theorie unitaire affine du champ physique. J ourn. Phys. Radium (8) 12,
81-88 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 179
1951, 2
- Compatibilite des equations de la tMorie unitaire des champs. C. R. 232,
2407-2409 0 0 0 0 0 0 0 0 0 179
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1951, 3
- Les equations du champ unitaire et leurs approximations. C. R. 233, 513 to
516 0 0 0 0 0 0 0 0 0 0 0 0 0179 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1951, 4
Les equations electromagnetique deduites d'une tMorie unitaire des champs.
c. R. 233. 555-557 0 0 0 0 0 179
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
 Bibliography. 495

1941, 1
ToNOLO, A.: Determinazione di una classe di varieta riemanniane normali a
tre dimensioni. Ist. Lombardo Sei. Mat. N at. Rend. 5. 113-124 . . . . 248
1941, 2
- Sulle equazione di Weingarten relative ai sistemi tripli ortog:mali di
superficie isostatiche. Univ. Roma e Ist. Naz. Alta Mat. 2, 170-192 . . 248
1943, 1
- Sulle ipersuperficie V,. le cui due forme fondamentaU amettono una stessa
trasformazione infinitesima X(!). Boll. Un. Mat. Ital. (2) 5, 137-140 . . 361
1949. 1
Sulla varieta riemanniane normali a tre dimensione. Pont. Acad. Sc. Acta
13, 29-53 248
1949. 2
Sulla varieta riemanniane normali a tre dimensione. Atti d. Linc. (8) 6,
438-444 0 0 0 0 0 0 0 0 0 248
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1949. 3
Sopra una classa di deformazione finite. Ann. Mat. Pura Appl. (4) 29, 99 to
114 0 0 0 0 0 0 0 0 0 0 0248 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1950, 1
Sopra un sistema di equazione differenziali relativo ai moti rigidi delle va-
rieta riemanniane a tre dimensioni a curvatura costante. Rend. Sem. Mat.
Univ. Fadova 19, 25G-272 . . . . . . . . . . . . . . . . . . . 335,351
1953. 1
- Sopra un problema di Darboux relativo alla meccanica dei mezzi continui.
Atti IV Congr. Un. Mat. Ital. 2, 578-582 . . . . . . . . . . . . . . . 248
TouRNAY, F. (see G. PAPY).
1931, 1
TucKER, A. W.: On generalized covariant differentation. Ann. of Math. (2)
32, 451-460 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 68
1935. 1
- Non-riemannian subspaces. Ann. of Math. 36, 965-968 123
1950, 1
UnESCHINI, P.: Le equazioni di prima approssimazione nella nuova teoria rela-
tivistica unitaria di Einstein. Atti d. Linc. Fis. Mat. Nat. 9, 256-261 179
1951, 1
- Le equazioni di seconda approssimazione nella nuova teoria relativistica
unitaria di Einstein I. Atti d. Linc. 8, 21-24 . . . . . . . . . . . . . 179
1951, 2
- Le equazioni di seconda approssimazione nella nuova teoria relativistica
unitaria di Einstein II. Atti d. Linc. 10, 121-123 . . . . . . . . . . 179
1951, 3
- Sulle mutue azioni fra campo gravitazionale e campo elettromagnetico.
Atti d. Linc. (8) 10, 39G-394 . . . . . . . . . . . . . . . . . . . . 179
1940, 1
V RABE, M.: Geometry of parallel displacement making a volume invariant.
J. Sc. Hiroshima Univ. 10, 95-108. . . . . . . . . . . . . . . 123,155
496 Bibliography.

1947. 1
URBAN, A.: L'equation differentielle des courbes sur une Vn-l speciale dans Vn.
Acad. Tcheque. Mat. Nat. 48, 57-61 . . . . . . . . . . . . . 242
1948, 1
Differential equations of curves Oll a Special vn-1 in vn. Rozpravy II Tredy
Ceske Akad. 57. no. 9, 12 pp. (czech) . . . . 242
1934. 1
VANDERSLICE, J. L.: Non holonomic geometries. Am. Journ. 56, 153-192123,253
1934, 2
- Conformal tensorinvariants. Proc. Nat. Acad. 20, 672-676 . . . . 315, 316
1939. 1
VARGA, 0.: ber die Integralinvarianten, die zu einer Kurve in der hermitischen
Geometrie gehren. Acta Hungar. 9. 88-102 . . . . . . . . . . . . 415
1943, 1
ber eine Charakterisierung der Riemannschen Rume konstanter Krm-
mung. Mat. Fiz. Lapok SO, 34-39 (hungarian, german summary) . . . . 148
1950, 1
- Normalkoordinaten in allgemeinen differentialgeometrischen Rumen und
ihre Verwendung zur Bestimmung smtlicher Differentialinvarianten (hun-
garian and german). C. R. first congress of hungarian mathem. 131-162 155
1951, 1
VASILIEV, A. M.: General invariant methods in differential geometry. Akad.
Nauk SSSR. 79. S-7 (russian) . . . . . . . . . . . . . . . . . . . 67
1922, 1
VEBLEN, 0. (see also L. P. EISENHART): Normal Coordinates for the geometry
of paths. Proc. Nat. Akad. 8, 192-197 . . . . . . . . 147,155, 300,303
1922, 2
Projective and affine geometry of paths. Proc. Nat. Acad. 8, 233-238 . . 300
1923, 1
and T. Y. THOMAS: The geometry of paths. Trans. Am. Math. Soc. 25, 551
to 608 . . . . . . . . . . . . . . . . 155, 159, 300, 303
1923, 2
Equiaffine geometry of paths. Proc. Nat. Acad. 9. 3-4 144
1924, 1
and T. Y. THOMAS: Extensions of relative tensors. Trans. Am. Math. Soc.
26, 373-377 . . . . . . . . . . . . . . . . 129,155,160
1925, 1
and J. M. THOMAS: Projective normal coordinates for the geometry of
paths. Proc. Nat. Acad. 11, 204-207 . . . 155
1926, 1
and J. M. THOMAS: Projective invariants of affine geometry of paths.
Ann. of Math. 27, 278-296 . . . . . . . . . . . . . . . . . . 87, 300,302
1927, 1
*- Invariants of quadratic differential forms. Cambr. Tracts No 24, 102 pp. 155
1928, 1
- Projective tensors and Connections. Proc. Nat. Acad. 14, 154-166 300,303
 Bibliography. 497

1928, 2
VEBLEN, 0.: Differential invariants and geometry. Atti del Congress Int.
Bologna 6, 181-189 . . . . . . . . . . . . . . . . 316
1928, 3
- Conformal tensors and connexions. Proc. Nat. Acad. 14, 735-745 315, 316
1929, 1
- Generalized projective geometry. Journ. London, Math. Soc. 4, 140-16o 300
1930, 1
- A generalization of the quadratic differential form. Quart. J ourn. Oxford
1, 60-76 . . . . . . . . . . . . . . . . . . . . . . 315
1932, 1
*- and J. H. C. WHITEHEAD: The foundations of differential geometry.
Cambr. Tracts. No 29, 96 pp. Reprinted 1953 . . . . 61, 62, 64
1935. 1
Formalism for conformal geometry. Proc. Nat. Acad. 21, 168-173 315,316
1936, 1
*- and J. v. NEUMANN and J. W. GrvENs: Geometry of complex domains.
Princeton Seminar, stencilled. . . . . . . . . . . . . . . . . . . 53, 57
1949, 1
VERBICKII, L. L.: On the metric differential geometry of hypersurfaces of the
second order. Trudy Sem. Vektor Tenzor., An. 7, 319-340 (russian) 242
1950, 1
On the equations for imbedding riemannian spaces of class 2 in euclidean
spaces. Trudy Sem. Vektor, Tenzor, c\n. 8, 425-429. . . . . . . . . 269
1951, 1
- A tensor criterion of conformally euclidean spaces of class 1. Akad. Nauk
SSSR. 81, 133-136 . . . . . . . . . . 315
1951, 2
- Construction of a conformally euclidean space of class 1 within a con-
taining euclidean spacc. Akad. Nauk SSSR. 81, 333-336 (russian) . . . 315
1923, 1
VITAL!, G.: I fcndam::nti del calcolo assoluto generalizzato. Giorn. d. Matern.
61, 159-202 . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
1924, 1
Una derivazione covariante formata coll'ausilio di n sistemi covarianti del
1 ordine. Atti Pavia 248-253 . . . . . . . . . . . . . . . . . . . 142
1946, 1
VoLTA V. DALLA: Sulla torsione di una varieta a tre dimensioni dotati di una
connessione affine. Univ. Roma Ist. Naz. Mat. e Appl. (5) 5, 271-282 . . 123
1946, 2
Sulle connessioni affine asimmetriche. Univ. Roma Ist. Naz. Mat. e Appl.
(5) 5, 315-326 . . . . . . . . . . . . . . . . . . . . . . . . . . 123
1887, 1
VoLTERRA, V.: Sui fondamenti della teoria delle equazioni differenziali lineari.
Mem. Soc. Ital. 40, III 6, No 8, 107 pp. . . . . . . . . . . . . . . . 340
1889, 1
Sulle funzioni conjugate. Rend. d. Linc. (IV) 51, 599-611 85
Schouten, Ricci-Calculus, 2. Auf!. 32
498 Bibliography.

1880, 1
Voss, A.: Zur Theorie der Transformation quadratischer Differentialausdrcke
und der Krmmung hherer Mannigfaltigkeiten. Math. Ann. 16, 129-178
146,250,256,263
1926, 1
VRANCEANU, G.: Sur les espaces non holonomes. C. R. 183, 852-854 78, 253, 261
1926, 2
Sur le calcul differentiel absolu pour les varietes non holonomes. C. R. 183.
1083-1085 . . . . . . . . . . . . . . . . . . . . . . . . . 78, 253, 261
1927, 1
Studio geometrico meccanico dei sistemi anolonomi. Ann. di Mat. (4) 5.
275-278 . . . . . . . . . . . . . . . . . . . . . . . 78, 253
1927, 2
Sur l'ecart geodesique dans les espaces non holonomes. Ann. Sei. Univ.
Jassy 15, 7-24 . . . . . . . . . . . . . . . . . . . .78,253
1928, 1
Sur quelques tenseurs dans les varietes non holonomes. C. R. 186, 995-997
78,253,261
1928, 2
Seconda forma quadratica fondamentale di una varieta anolonoma ed appli-
cazioni. Rend. d. Linc. (6) 8, 669-673 . . . . . . . . . . . . . . . 78, 253
1928, 3
Sullo scostamento geodetico nella varieta anolonome. Rend. d. Linc. (6)
7. 134-137 . . . . . . . . . . . . . . . . . . . . . . . . . . . 78, 261
1928, 4
- Note sur l'ecart geodesique dans les espaces non-holonomes. Ann. Univ.
Jassy 15, 307-310 . . . . . . . . . . . . 78
1928, 5
- Parallelisme et courhure dans une variete non holonome. Atti Congr. Int.
Mat. Bologna 6, 61-67 . . . . . . . . . . . . . . . . . . . . . . . 253
1929, 1
- Les trois points de vue dans l'etude des espaces non-holonomes. C. R. 188,
973-976 . . . . . . . . . . . . . . . . . . . . . . . . 78, 253
1929, 2
Studio geometrico dei sistemi anolonomi. Ann. di Mat. (4) 6, 9-43 . . 78, 253
1929, 3
Sur les espaces de Riemann ayant leurs coefficients de rotation constants.
c. R. 189. 386-389 . . . . . . . . . . . . . . . . . . . . . . . . 171
1930. 1
- Sur les groupes d'applicabilite des varietes non-holonomes. C. R. 190, 1100
to 1102 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78, 253
1930, 2
- Sulle condizioni di rigidita di una Vm in un Sn. Rend. d. Linc. (6a) 11, 385
to 389 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
1931, 1
- Sur quelques theoremes relatifs aux varietes non-holonomes et aux systemes
de formes de Pfaff. C. R. 192, 721-724 . . . . . . . . . . . . . . 78, 253
 Bibliography. 499

1932, 1
VRANCEANU, G.: Sur quelques points de la theorie des espaces non-holonomes.
Bull. fac. sei. Cernauti 5, 177-205 . . . . . . . . . . . . . . 78, 170, 253
1933, 1
Connexions affines liees aux systemes de Pfaff. Bull. fac. sei. Cernauti 6,
188-196 . . . . . . . . . . . . . . . . . . . . . . . 78
1934, 1
Etude des espaces non holonomes. J ourn. de Math. (9) 13, 113-174 78, 253
1936, 1
* Les espaces non holonomes. Mem. d. Sc. Math., fase. LXXVI, Gauthier-
Villars. Paris. 70 pp. . . . . . . . . . . . . . 78, 253
1938, 1
lnvariants de deux surfaces non holonomes complementaires. C. R. Raum.
459--465 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
1940,
Sur les espaces a connexion conforme. Disquisit. Math. Phys. 1, 63-81 315

1942,
Sur les invariants des espaces de Riemann singuliers, Disquisit. Math.
Phys. 2, 253-281 . . . . . . . . . . . . . . . . 133, 270
1942, 2
Sur les espaces a connexion affine et projective. Bull. Math. Soc. Rou-
maine Sei. 44, 85-139 . . . . . . . . . . . . . . 253
1943, 1
Sur la theoriedes espaces a connexion conforme. Bull. Math. Soc. Roumaine
Sei. 45, 3-31 . . . . . . . . . . . . . . . . . . . . . . . 353

1947, 1
* Leyons de geometrie differentielle, Vol. I. Congruences, formes de Pfaff,
groupes continus, invariants et equivalence, espaces a connexion affine,
espaces de Riemann, espaces a connexion projective. Bucarest. 422 pp.185, 348

1947, 2
Sur les espaces partiellerneut projectifs. Bull. Math. Soc. Roumaine Sei.
48, 43-64 . . . . . . . . . . . . . . . 317, 326
1949, 1
La geometrisation des espaces sympletiques. C. R. 229, 336--338 . . . 179, 348

1949, 2
Sur les espaces a connexion a groupe maximumdes transformations en eux-
mcmes. C. R. 229, 543-545 . . . . . . . . . . . . . . . . . 348
1950, 1
On Kagan spaces (romanian, russian and french summaries). Rom. Acad.
2, 299-304 . . . . . . . . . . . . . . . . . . . . . . . . 317
1950, 2
On metric Kagan spaces. (romanian, russian and french summaries). Rom.
Acad. 2, 6 pp. . . . . . . . . . . . . . . . . 317
1951, 1
Lectii de geometrie diferentiala, Il (romanian) . . 348
Schouten, Ricci~Calculus, 2. Aufl. 32a
soo Bibliography.

1951. 2
VRANCEANU, G.: Asupra grupurilor etc. Com. Ae. R.P.R. 1, 137; cf. Stud.
S. Cerc. Mat. 2, 57 (romanian) . . . . . . . . 348
1951. 3
Groupes de mouvement des espaces a connexion. Ae. R. P. R. 2, 387-433
(romanian with a french summary) . . . . . . . . . . . . 348
1953. 1
Sur !es groupes de mouvement d'un espace de Riemann a quatre di-
mensions. Stud. S. Cerc. Mat. 4, 121-153 (romanian with a french summary) 348
1941, 1
WADE, T. L. (see also R. H. BRucK): Tensor algebra and Young's symmetry
operators. Am. J. Math. 63, 64 5-65 7 . . . . . . . . . . . 15
1943. 1
The factorization of rank tensors. Duke Math. J. 10, 493-498 . . . . .15, 34
1944, 1; 1945. 1
- Tensoralgebra and invariants I, II. Nat. Math. Mag. 19, 3-10; 20, 5-10 15
1927. 1
WAERDEN, B. L.vAN DER: Differentialkovarianten von n-dimensionalenMannig-
faltigkeiten in Riemannschen m-dimensionalen Rumen. Abh. Math. Sem.
Harnburg 5. 153-160 . . . . . . . . . . . . . . . . . . . . . 68, 254, 269
1931. 1
*- Modeme Algebra II. Springer, Berlin. 216 pp. 31
1936. 1
WAGNER, V.: Sur Ia geometrie differentielle des multiplicites anholonomes.
Proc. Sem. Vect. a. Tens. Moscow 2-3. 269-318 . . . . . . . . . . . . 253
1938. 1
ber Vi von der Krmmung Null in R 3 Rec. Math. Moscow 4 (46) 2, 333
to 338 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
1938. 2
On the geometric interpretation of the curvature vector of a non-holonomic
v; in the three-dimensional euclidean space. Rec. Math. Moscow 4 (46) 2,
339-356 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
1940, 1
Differential geometry of non-linear non holonomic manifolds in the three-
dimensional euclidean space. Rec. Math. (Mat. Sbornik) N. S. 8 (SO), 3-39
(english, russian summary) . . . . . . . . . . . . . . . . . . . . . 253
1941, 1
The geometry of an (n- 1) -dimensional nonholonomic manifold in an n-
dimensional space. Proc. Sem. Vect. and Tensor 5. 173-225 (russian) 253
1942, 1
On the Cartan group of holonomicity for surfaces. C. R. Acad. Sc. URSS.
37. 6-8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
1942, 2
Differential geometry of the family of Rk's in Rn and of the family of totally
geodesie Sk_ 1 's in Sn-l of positive curvature. Rec. Math. (Mat. Sbornik)
10 (52) 165-212 (english, russian summary) . . . . . . . . . . . . . . 269
 Bibliography. 501

1943, 1
WAGNER, V.: The absolute derivative of field of local geometric object in a
compound manifold. C. R. Acad. Sc. URSS. 40, 94-97 . . . . . . . 123, 366
1943. 2
- The inner geometry of non-linear non-holonomic manifolds. Rec. Math.
(Mat. Sbornik) 13 (55) 135-167 (english, russian summary) . . . . . . 253
1945. 1
- The theory of geometric objects and the theory of finite and infinite conti-
nuous groups of transformations. C. R. Acad. Sc. URSS. 46, 347-349 67
1945. 2
- The generalization of Ricci's and Bianchi's identities for a connexion in
the compound manifold. C. R. Acad. Sc. URSS. 46, 303-305 . . . . 123, 147
1946, 1
On a sufficient condition in the problern of Lagrange for multiple inte-
grals. C. R. Acad. des Sc. Moscow 54, 479-482 . . . . . . . . . . . 232
1946, 2
Constant fields of local geometric objects in compound manifolds with
a linear connexion. C. R. Acad. Sc. URSS. 53. 183-186 . . . . . . . 366
1949. 1
Classification of simple differential-geometric objects. Akad. Nauk SSSR.
69, 293-296 (russian). . . . . . . . . . . . . . . 67
1949. 2
Classification of linear Connections in a composite manifold Xn+(t) accord-
ing to their holonomy group. Trudy Sem. Vektor Tenzor, An. 7. 205-226
(russian) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367
1950. 1
on the theory of pseudogroups of transformations. Akad. Nauk SSSR.
72, 453-456 . . . . . . . . . . . . . . . . . . . . . . . . 64,67,105
1952. 1
W AKAKUWA, H.: On Riemann spaces whose homogeneaus holonomy groups are
integrable. TChoku Math. J. 4, 96--98 . . . . . . . . . . . . . . . . 367
1934, 1
*WALBERER, P.: Orthogonale Kurvengewebe in Riemannschen Rumen. Dis-
sertation Hamburg. The three parts also appeared in Abh. Math. Sem.
Harnburg 10, 136--151 ; 152-168; 169-1 79 . . . . . . . . . . . . . . 246
1942, 1
WALKER, A. G. (see also E. M. PATTERSON): Note on a distance invariant and
the calculation of Ruse's invariant. Proc. Edinburgh Math. Soc. (2) 7, 16 to
26 . . . . . . . . . . . . . . . . . . . . 38J, 386
1944. 1
- Completely symmetric spaces. ]. London Math. Soc. 19, 219-226 163
1944, 2
- Complete symmetry in flat space. J. London Math. Soc. 19, 227-229 163
1945. 1
A particular harmonic riemannian space. J. London Math. Soc. 20, 93-99
382,383,386,387,388
502 Bibliography.

1945, 2
WALKER, A. G.: On completely harmonic spaces. J. London Math. Soc. 20,
159-163 . . . . . . . . . . . . . . . . . . . . . 386, 387

1946, 1
- Symmetrie harmonic spaces. J. London Math. Soc. 21, 47-57 . . . . . 387

1949, 1
- On parallel fields of partially null vector spaces. Quart. J. of Math. Oxford
Ser. 20, 135-145 . . . . . . . . . . . . . . . . . . . . . 322,367, 422

1949, 2
- On Lichnerowicz's conjecture for harmonic 4-spaces. J. London Math. Soc.
24, 21-28 . . . . . . . . . . . . . . . . . . . . 387

1950, 1
- On Ruse's spaces of recurrent curvature. Proc. London Math. Soc. 52, 36 to
- 64 . . . . . . . . . . . . . 163,322,369,370,421,422,423,424

1950, 2
- On parallel fields of partially null vector spaces. Quart. J. of Math. Ox-
ford 20, 135-145 . . . . . . . . . . . . . . . . . . . . . . 228

1950, 3
- Canonical form for . riemannian space with a parallel field of null planes.
Quart. J. of Math. Oxford (2) 1, 69-79 . . . . . . . . . . 228,322,367,422

1950, 4
- Canonical forms II. Parallel partially null planes. Quart. J. Math. Oxford
(2) 1, 147-152 . . . . . . . . . . . . . 228, 322, 367, 422

1943, 1
W ANG, H. CH.: On a projective invariant of a nonholonomic surface. Ann. of
Math. (2) 44, 562-571 . . . . . . . . . . . . . . . . . 253
1943, 2
- A projective invariant of a nonholonomic surface. Duke Math. J. 10,
565-574- (This paper appeared also Ann. of Math. (2) 44, 562-571.) . 253

1927, 1
*WEATHERBURN, C. E.: Differential geometry of three dimensions. Cambr.
Univ. Press. 268 pp. . . . . . . . . . . . . . . . . . . . . . . 306, 342

1933, 1
*- The development of multidimensional differential geometry. Report
Australian and New Zealand Assoc. for Advancement of Science 21, 12-28.

1938, 1
*- An introduction to Riemann geometry and the tensor calculus. Cambr.
Univ. Press. Cambridge. 191 pp.
1900, 1
*WEBER, E. v.: Vorlesungen ber das Pfaffsche Problem und die Theorie der
partiellen Differentialgleichungen erster Ordnung. Teubner, Leipzig. XI
+ 622pp. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
 Bibliography. 503

1934. 1
WEISE, K. H. (see also R. KNIG): Beitrge zum Klassenproblem der quadra-
tischen Differentialformen. Math. Ann. 110, 522-570 . . . . . . . . . 268
1935. 1
- Autoparallele Mannigfaltigkeiten im affin zusammenhngenden Raum.
Math. Zeitschr. 40, 208-220 . . . . . . . . . . . . . . . . . . . 232
1908, 1
*'VEITZENBCK, R.: Komplex-Symbolik. Leipzig 23, 58
1922, 1
*- Neuere Arbeiten der algebraischen Invariantentheorie. Differential-
invarianten. Encycl. Mathem. Wissensch. III D 10 (E 1). 71 pp.
1923. 1
*- Invariantentheorie. Noordhoff, Groningen. 408pp. 15, 23, 26, 97,142,147,165
1928, 1
-- Differentialinvarianten in der Einsteinsehen Theorie des Fernparallelismus.
Sitz. Preu. Akad. 466-474 . . . . . . . . . . . . . . . . . . . . . 142
1932, 1
ber den Reduktionssatz bei affinem und projektivem Zusammenhang.
Proc. Kon. Ned. Akad. Amsterdam 35, 1220-1229 . . . . . . . . . . . 302
1937. 1
Bemerkungen ber Trivektoren. Proc. Kon. Ned. Akad. Amsterdam 40,
312-315 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
1937. 2
- ber Trivektoren I, II, III. Proc. Kon. Ned. Akad. Amsterdam 40, 676 to
681; 704-710; 815-857 . . . . . . . . . . . . . . . . . . . . . . . 36
1938. 1
ber Trivektoren IV-VII. Proc. Kon. Ned. Akad. Amsterdam 41, 108-116;
238 245; 362-369; 454-461 . . . . . . . . . . . . . . . . . 36
1939. 1
- Zur Theorie der Komplexgren aijk Monatsh. Math. Phys. 48, 129-140 15
1918, 1
WEYL, H. (see also A. BRAUER): Reine Infinitesimalgeometrie. Math. Zeitschr.
2, 384-411 . . . . . . . . . 133, 304, 306
1921, 1
*- Raum, Zeit, Materie, 4. Aufl. Springer, Berlin. 300 pp. . 40,41,136
1921, 2
Zur Infinitesimalgeometrie: Einordnung der projektiven und der konformen
Auffassung. Gttinger Nachrichten 1921, 99--112 . . . . . 148, 289, 294, 304
1922, 1
*- Mathematische Analyse des Raumproblems. Springer, Berlin. 117 pp. 31, 33
1922, 2
Die Einzigartigkeit der Pythagoreischen Mabestimmung. Math. Zeitschr.
12, 114-146 . . . . . . . . . . . . . . . . . . . . . . . . . 136, 137
1922, 3
- Zur Infinitesimalgeometrie; p-dimensionale Flche im n-dimensionalen
Raum. Math. Zeitschr. 12, 154-160 . . . . . . . . . . . . . . . 232,242
504 Bibliography.

1929, 1
WEYL, H. and H. P. RoBERTSON: On a problern in the theory of groups arising
in the foundations of infinitesimal geometry. Bull. Am. Math. Soc. 35.
686--690 . . . . . . . . . . . . . . . . . . . . . . . . . . . 300, 303
1929, 2
On the foundations of generalinfinitesimal geometry. Bull. Am. Math. Soc.
35. 716--725 . . . . . . . . . . . . . . . . . . . . . . . . . 300, 303
1939. 1
*- The classical groups, their invariants and representations. Princeton Univ.
Press. Princeton N. J. 302pp. . . . . . . . . . . . . . . . . . . . 15
1937. 1
WEYSSENHOFF, J. VAN: Duale Gren, Grorotation, Grodivergenz und die
Stokes-Gauschen Stze in allgemeinen Rumen. Ann. Soc. Pol. 16, 127 to
144 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
1929, 1
WHITEHEAD, J. H. C. (see also 0. VEBLEN}: On a dass of projectively flat affine
connexions. Proc. London Math. Soc. 2, 32, 93-114 . . . . . 300, 301, 315

1930, 1
and B. V. WrLLIAMS: A theorem of linear connect10ns. Ann. of Math. (2)
31, 150--157 155
1931, 1
- On linear connexions. Trans. Am. Math. Soc. 33. 191-209 . . . . . . . 155

1931, 2
- The representation of projective spaces. Ann. of Math. 32, 327-360 . . . 302

1932, 1
- Affine spaces of paths which are symmetric about each point. Math. Zeit-
sehr. 35. 644-659 . . . . . . . . . . . . . . . . . . . 163
1932, 2
- Note on Maurer's equations. Journ. London Math. Soc. 7. 223-227 . . . 197
WILLIAMS, B. V. (see J. H. C. WHITEHEAD).
1950, 1
WrLLMORE, T. J.: Mean value theorems in harmonic riemannian spaces. J.
London Math. Soc. 25, 54-57 . . . . . . . . . . . . . . . . . . . . 383
1951, 1
- Les plans paralleles dans les espaces riemanniens globaux. C. R. 232, 298 to
299 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367
1953. 1
Some properties on harmonic riemannian manifolds. Int. Congress V enice
Sept. 1953 . . . . . . . . . . . . . . . . . . . . . . 388
1951, 1
WrNOGRADSKI, J.: Sur la connexion des espaces affines. C. R. 232, 936--938 123

1940, 1
WONG, Y. CH.: On the principal directions of a tensor. J. London Math. Soc.
15, 172-183 . . . . . . . . . . . . . . . . . . . . . . . . . . 35, 269
 Bibliography. SOS

1940, 2
WONG, Y. (H.: On the Frenet formulae for a vm in a vn. Quart. J. of Math.
Oxford 11, 146-160 . . . . . . . . . . . . . . 277
1940, 3
On a certain matrix occuring in the theory of helices. J. London Math. Soc.
15, 168-172 . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
1941, 1
Generalized helicesinan ordinary Vn. Proc. Cambridge Philos. Soc. 37, 14
to 28 . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
1941, 2
On the generalized helices of Hayden and Syptak in an N-space. Proc.
Cambridge Philos. Soc. 37, 229--243 . . . . . . . . . . . . . . 230
1943, 1
- A note an complementary subspaces in a riemannian space. Bull. Am. Math.
Soc. 49, 12o-125 . . . . . . . . . . . . . . . . . . . . . . . 272, 286
1943, 2
Same Einstein spaces with conformally separable fundamental tensors.
Trans. Amer. Math. Soc. 53, 157-194 . . . . . . . . . . . . . 287,312,315
1943, 3
- Family of totally umbilical hypersurfaces in an Einstein space. Ann. of
Math. 44, 271-297 . . . . . . . . . . . . . . . . . . . 251, 272,312
1945, 1
- A note an the first normal space of a Vm in an Rn. Bull. Am. Math. Soc.
51, 997-1000 . . . . . . . . . . . . . . . . . 269
1945, 2
- Quasi-orthogonal ennuple of congruences in a riemannian space. Ann. of
Math. (2) 46, 158-173 . . . . . . . . . . . . . . . . . . . 228, 246
1946, 1
- Sametheorems an Einstein 4-space. Duke Math. J. 13, 601-610 . 272
1895, 1
*Woons, F. S.: ber Pseudominimalflchen. Diss. Gttingen. 73 pp. 274
1922, 1
*- Higher geometry. Ginn & Co., Boston. 423 pp . . . . . . 316
1929, 1
WouDE, W. VAN DER: Over de invoering van het begrip kromtestraal. N. Ar-
chief v. Wisk. (2) 13, 396-397 . . . . . . . . . . . . . . . . . . . 229
1908, 1
*WRIGHT, J. E.: Invariants of quadratic differential forms. Cambridge Tracts
9- 90pp. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296
WRBEL, F. H. (see S. GoLAB).
1941' 1
WRONA, W. (see also J. HAANTJES): Eine Verallgemeinerung des Schursehen
Satzes. Proc. Kon. Ned. Akad. Amsterdam 44, 943-946 . . . . . 148,153
1947, 1
- Conditions necessaires et suffisantes qui detenninent !es espaces einsteiniens
conformement euclidiens et de courhure constante. Ann. Soc. Polon. Math.
20, 28-80 . . . . . . . . . . . . . . . . . . . . . . . . . . 148, 153
506 Bibliography.

1948, 1
WRoNA, W.: On multivectors in a Vn I. Proc. Kon. Ned. Akad. Arnsterdam 51,
1291-1301 = lndagationes Math. 10, 435-445 . . . . . . . . . . . 146, 305
1948, 2
- Conditions necessaires et suffisantes qui deterrninent les espaces einstein-
. iens, conformement euclidiens et de courhure constante. Ann. Soc. Polon.
Math. 20, 28-80 . . . . . . . . . . . . . . . . . . . . . . . 305, 312
1949, 1
On multivectors in a Vn II. Proc. Kon. Ned. Akad. Amsterdam 52, 29-36
= lndagationes Math. 11, 61-68 . . . . . . . . . . . . . . . . . . 305

1950, 1
Sur les multivecteurs dans Vn 111. Casopis Pest. Mat. Fys. 74, 273-280
(polish, french summary) . . . . . . . . . . . . . . . . . . . . . . 305
1929. 1
WuNDHEILER, A.: Sur un deplacement generalise dans les espaces riemanniens.
Rend. d. Linc. 9, 387-389 . . . . . . . . . . . . . 123
1937. 1
- Objekte, Invarianten und Klassifikation der Geometrien. Abh. Sem.
Vektor u. Tensor., Moscow 4, 366--375 (Congress Diff. Geom. Moscow 1934) 67
1949. 1
Y ANENKO, N. N.: The geometric structure of surfaces of small type. Akad.
Nauk SSSR. 64, 641-644 (russian) . . . . . . . . . . . . . . . . . 269
1949. 2
On some necessary conditions for the deformability of surfaces in a multi-
dimensional euclidean space. Akad. Nauk SSSR. 65, 449-452 (russian) 352
1953. 1
- Some questions of the theory of imbedding of riemannian metrics in eucli-
dean spaces. Uspedi Matern. Nauk (N. S.) 8, 21-100 (russian) 269
1935. 1; 1936, 1
YANO, K. (see also S. SASAKI) and Y. MuT: On the connexion in Xm asso-
ciated with the points of Ym. Proc. Phys. Math. Soc. Japan 111 17, 379 to
390; 18, 1-9 . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
1936. 2
and Y. MuT: Notes on the derivation of geodesics and the fundamental
scalar in a riemannian space. Proc. Phys. Math. Soc. Japan 111 18, 142-152 352
1937. 1
Sur !es espaces non holonomes totalement geodesiques. C. R. 205, 9-12 253
1937. 2
Sur le changement des coefficients d'une connexion projective. C. R. 205,
63 7-639 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
.1937. 3
Sur les fquations des geodesiques ans une variete a connexion projective.
c. R. 205, 829-831 . . . . . . . . . . . . . . . . . . . . . . 302, 303
1938, 1
- La relativite non holonome et la theorie unitaire d'Einstein et Mayer. Math.
Cluj. 14, 124-132 . . . . . . . . . . . . . . . . . . . . . . . . . 253
 Bibliography. 507

1938, 2
Y ANO, K.: Remarques relatives a la theorie des espaces aconnection conforme.
c. R. 206, 560-563 . . . . . . . . . . . . . . . . . . . . . . 315, 316
1938, 3
Sur les conferences generalisees dans les espaces a connection conforme.
Proc. lmp. Acad. 14, 329-332 . . . . . . . . . . . . . . . . . . 315, 330
1938, 4
*- Les espaces a connection projective et la geomEtrie projective des" paths ".
These Univ. Yassy. . . . . . . . . . . . . . . . . . . . . . 303
1938, 5
- Sur l'espace projectif de M. D. van Dantzig. C. R. 206, 1610-1612 303
1938, 6
- The non holonomic representation of projective spaces. Proc. Phys. Matl1.
Soc. Japan (3) 20, 442-4 50 . . . . . . . . . . . . . . . . . . . . 303
1939, 1
- Theory of non-holonomic spaces. Tensor 2, 13-24 (japanese) 253
1939. 2
Sur les equations de Gauss dans la geometrie conforme des espaces de Rie-
mann. Proc. Imp. Acad. Tokyo 15, 247-252 . . . . . . . . . . . . 311,315
1939. 3
Sur les equations de Codazzi dans la geometrie conforme des espaces de
Riemann. Proc. Imp. Acad. Tokyo 15, 340-344 . . . . . . . . . . 311,315
1939. 4
A projective treatment of a conformally connected manifold. Proc. Phys.
Math. Soc. Japan 21, 270-286 . . . . . . . . . . . . . . . . . . . 315
1939. 5
Sur la theoriedes espaces a connexion conforme. J. Imp. Univ. Tokyo 4,
1-59 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
1940, 1
- Conformally separable quadratic differential forms. Proc. Imp. Acad. Tokyo
16, 83-86 . . . . . . . . . . . . . . . . . . . . . . . 287,315
1940, 2
- and S. PETREscu: Sur les espaces metriques non holonomes complemen-
taires. Disquisit. Math. Phys. 1, 191-246 . . . . . . . . . . . 253
1940, 3
Sur quelques proprietes conformes de v;_ dans V", dans Vn. Proc. Imp. Acad.
Tokyo 16, 173-177 . . . . . . . . . . . . . . . . . . . . 285, 311, 315
1940, 4
Sur la connexion de Weyl-Hlavaty et ses applications a la geometrie con-
forme. Proc. Phys. Math. Soc. Japan (3) 22, 595-621 . . . . 303, 316, 334
1940, 5
- Concircular geometry I. Concircular transformations. Proc. Imp. Acad.
Tokyo 16, 195-200 . . . . . . . . . . . . . . . . 322,330,331,334
1940, 6
- Concircular geometry II. lntegrability conditions of pt-< =(/Jg,. . Proc.
Imp. Acad. Tokyo 16, 354-360 . . . . . . . . . . . . . 322,330,331,332
508 Bibliography.

1940, 7
Y ANO, K.: Concircular geometry III. Theory of curves. Proc. Imp. Akad.
Tokyo 16,442-448 . . . . . . . . . . . . . . . . . 322,330,334
1940, 8
- Concircular geometry IV. Theory of subspaces. Proc. Imp. Acad. Tokyo
16, 505-511 . . . . . . . . . . . . . . . . . . . . . . . 322, 330, 332
1940, 9
and Y. Mur: Sur Ia theoriedes hypersurfaces dans un espace a connexion
conforme. Proc. Imp. Acad. Tokyo 16, 266--273 . . . . . . . . . . . . 315

1941, 1
and Y. Mur: Sur Ia theoriedes espaces a connexion conforme normale et
Ia geometrie conforme des espaces de Riemann. Proc. Imp. Acad. Tokyo
17, 87-94 . . . . . . . . . . . . . . . . . . . . . . . . 315

1941, 2
- and Y. Mur: Sur Ia tht'orie des espaces a connexion conforme normale et
Ia geometrie conforme des espaces de Riemann. J. Imp. Univ. Tokyo 4,
117-169 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315

1941' 3
- and Y. Mur: On the conformal arc length. Proc. Imp. Acad. Tokyo 17,
318-322 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315

1941, 4
- and Y. Mur: On the generalized loxodromes in the conformally connected
manifold. Proc. Imp. Acad. Tokyo 17, 455-460 . . . . . . . . . . . 315

1941' 5
and Y. Mur: Sur Ia theoriedes hypersurfaces dans un espace a connexion
conforme. Jap. Journ. ofMath. 17,229--288 . . . . . . . . . . . . . 315

1941, 6
The variable u 0 of T. Y. Thomas and the homogeneaus coordinates of
D. van Dantzig. Tensor 4, 38-48 (japanese) . . . . . . . . . . . . . . 303

1942, 1
and Y. Mur : Sur le theoreme fondamental dans Ia geometrie conforme des
sous-espaces riemanniens. Proc. Phys. Math. Soc. Japan (3) 24, 437-449
278,311,315
1942, 2
On the fundamental theorem of conformal geometry. Tensor 5, 51-59
(japanese) . . . . . . . . . . . . . . . . 278, 315
1942, 3
- Concircular geometry V. Einstein spaces. Proc. Imp. Acad. Tokyo 18,
446--4 51 . . . . . . . . . . . . . . . . . . . . . 334
1942, 4
- Les espaces d'elements lineaires a connexion projective normale et la geo-
metrie projective generale des paths. Proc. Phys. Math. Soc. Japan 24,
9--24 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301' 303
1943, 1
- Conformal and concircular geometries in Einstein spaces. Proc. Imp. Ac.
Tokyo 19, 444-453 . . . . . . . . . . . . . . . . . . . . 148, 312,334
 Bibliography. 509

1943, 2
y ANO, K.: Sur les equations fondamentales dans la geometrie conforme des
sous-espaces. Proc. Imp. Acad. Tokyo 19, 326--334 . . . . . . . . 311, 315

1943, 3
Sur le parallelisme et la concourance dans l'espace de Riemann. Proc. Imp.
Acad. Tokyo 19, 189--197 . . . . . . . . . . . . . . . 322,333,334,370
1943, 4
Sur une application du tenseur conforme Cfk et de la scalaire conforme C.
Proc. Imp. Acad. Tokyo 19, 335-340 . . . . . . . . . . . . . . . . . 315
1943, 5
On Sasaki's study of holonomy groups of spaces with normal conformal
Connections. Tensor 6, 68--85 (japanese) . . . . . . . . . . . . . . . 367

1944, 1, 2
- and K. TAKANO: Sur les coniques dans les espaces a connexion affine ou pro-
jective I, II. Proc. Imp. Acad. Tokyo 20, 41o-417; 418--424 . . . . 298,300

1944, 3
Equations of paths in 0. Veblen's projective space. Tensor 7, 65-72 (ja-
panese) . . . . . . . . . . . . . . . . . . . . . . . . . 302, 303, 327
1944, 4
Projective parameters in projective and conformal geometries. Proc. Imp.
Acad. Tokyo 20, 45-53. . . . . . . . . . . . . . . . . . . . . 302, 315
1944, 5
On the torse-forming directions in riemannian spaces. Proc. lmp. Acad.
Tokyo 20, 34o-345 . . . . . . . . . . . . . . . . . . . . . . 322,332
1944, 6
Subprojective transformations, subprojective spaces and subprojective col-
lineations. Proc. Imp. Acad. Tokyo 20, 701-705 . . . . . . . . 327, 328, 330

1944, 7
- and T. ADATI: Parallel tangent deformation, concircular transformation
and concurrent vectorfield. Proc. Imp. Acad. Tokyo 20, 123-127 330, 333, 352

1944, 8, 9, 10
- Sur les espaces a connexion affine qui peuvent representer les espaces pro-
jectifs des paths I, II, III. Proc. Imp. Acad. Tokyo 20, 631-639; 21, 16--24;
97-103 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
1944, 11
Projective parameters on paths in D. van Dantzig's projective space. Proc.
Imp. Acad. Tokyo 20, 21o-215 . . . . . . . . . . . . . . . . 303
1944, 12
ber eine geometrische Deutung der projektiven Transformationen nicht-
symmetrischer affiner bertragungen. Proc. lmp. Acad. Tokyo 20, 284
to 287 . . . . . . . . . . . . . . . . . . . 322

1944, 13; 1946, 4

and S. SASAKI: Sur les espaces a connexion conforme normale dont les grou-
pes d'holonomie fixent une sphere a un nombre quelconque de dimensions
I, II. Proc. Imp. Acad. Tokyo 20, 525-535; 22, 225-232 . . . . . . . .'367
510 Bibliography.

194 5, 1
YANO, K.: Lie derivatives in general space of paths. Proc. Japan Acad. 21,
363-371 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
1945, 2, 3
Sur la theoriedes espaces a hyperconnexion euclidienne I, II. Proc. Japan
Acad. 21, 156--163; 164-170 . . . . . . . . . . . . . . . . . . . . . 123

1945, 4
On the fundamental differential equations of flat projective geometry. Proc.
Japan Acad. 21, 392-400 . . . . . . . . . 303,306
1945, 5, 6; 1946, 2, 3
On the flat conformal differential geometry I, II, III, IV. Proc. Japan
Acad. 21, 419--429; 454-465; 22, 9--19; 20-31 . . . . . . . . 306
1945, 7
- Bemerkungen ber infinitesimale Deformationen eines Raumes. Proc.
Japan Acad. 21, 171-178 . . . . . . . . . . . 336

1945, 8
Sur la deformation infinitesimale des sous-espaces dans un espace affine.
Proc. JapanAcad. 21,248-260 . . . . . . . . . . . . . . . . . . . 336

1945, 9
Sur la deformation infinitesimale tangenhelle d'un sous-espace. Proc.
JapanAcad. 21,261-268 . . . . . . . . . . . . . . . . . 336,362
1945, 10
Quelques remarques sur un article de M. N. Coburn intitule "A charac-
terization of Schouten's and Hayden's deformation methods ". Proc.
Japan Acad. 21, 330-336 . . . . . . . . . . . . . . . . . . . 335, 352
1946, 1
arid Y. MuT: Note sur Je theoreme fondamental dans la geometricconforme
des sous-espaces riemanniens. Proc. J ap. Acad. 22, 338-342 . . . . 311, 316

1946, 5, 6, 7, 8, 9; 1947, 3
- Quelques remarques sur les groupes de transformations dans les espaces a
connexion lineaires I, II, III, IV, V, VI. Proc. Japan Acad. 22, 41-4 7; 69 to
74 (with K.TAKANo); 167-172; 173-183 (with Y.TOMONAGA); 275-283
(with Y. ToMONAGA); 23, 143-146 . . . . . . . . . . . . . . . 104, 346

1946, 10
- , K. TAKANO and Y. ToMONAGA: On the infinitesimal deformations of curves
in the spaces with linear connexion. Proc. Japan Acad. 22, 294-309 352

1947, t
On the flat projective differential geometry. Jap. J. Math. 19, 385-440 30:i
1947, 2
and S. SASAKI: On the structure of spaces with normal conformal connec-
tion whose holonomy group leaves invariant a sphere of arbitrary dimen-
sion. Sugaku (Math.) 1, 18-28 (japanese) . . . . . . . . . . . . . . 367

1947, 4
*- Setsuzoku no kikagaku (Geometry of connections). Kawade shob, Tokyo.
2 + + +
4 2 185 pp.
 Bibliography. 511

1948, 1
YANO, K.: Union curves and subpaths. Math. Japonicae 1, 51-59 . . . . . 250
1948, 2
- , K. TAKANO and Y. ToMONAGA: On infinitesimal deformations of curves in
spaces with linear connection. Jap. J. Math. 19, 433-477 . . . . 298, 343, 352
1948, 3
- and S. SASAKI: Sur Ia structure des espaces de Riemann dont Je groupe
d'holonomie fixe un plan a un nombre quelconque de dimensions. Proc.
Japan Acad. 24, 7-13 . . . . . . . . . . . . . . . . . . . . . . . 367
1949. 1
*- Groups of transfonnations in generalized spaces. Akad. Press. Comp. Ltd.
Tokyo. 70pp . . . . . . . . . . . . . . . . . . . 104,346,349,350,351
1949. 2
Sur Ia theorie des deformations infinitesimales. J. Univ. Tokyo 6, 1-75
298,335,352
1949. 3
and K. TAKANO: Conics in D. van Dantzig's projective space. Proc. Japan
Acad. 21,179-187 . . . . . . . . . . . . . . . . . . . . . . . . . 300
1949. 4
- and S. SASAKI: On the structure of spaces with normal projective connex-
ions whose groups of holonomy fix a hyperquadric or a quadric of (N- 2)-
dimension. Thoku Math. J. {2) 1, 31-39. . . . . . . . . . . . . 303
1950, 1
- Note on the conformal theory of curves. Tensor 1, 6-13 310, 315
1950, 2
- and T. !MAI: On affine collineations in projectively related spaces. J.
Math. Soc. Japan 1, 287-288 . . . . . . . . . . . . . . . . . . 346
1951, 1
and H. HIRAMATu: Affine and projective geometries of systems of hyper-
surfaces. J. Math. Soc. Japan 3. 116-136 . . . . . . . . . . 125
1951, 2
On groups of homothetic transformations in riemannian spaces. Journ.
Ind. Math. Soc. 15 B, 105-117. . . . . . . . . 346
1953. 1
On n-dimensional riemannian spaces admitting a group of motions of
order n (n- 1)/2 + 1. Trans. Amer. Math. Soc. 74, 26-279 . . . . . . 348
1948. 1
YEN, CHIH-TA: Sur une connexion projective normale associee a un systeme
de varietes a k dimensions. C. R. 227, 461-462 . . . . . . . . . . . 301
1937. 1
ZAREMBA, S. K.: Demonstration elementaire d'un theoreme de M. Golab sur
la representation conforme. Ann. Soc. Polon. Math. 14, 142-144 . . . . 315
 Index.
An 126 BERTRAND Curves 342ff.
An 394 BIANCHI'S identity 146ff.
21n 61 bivector 23, 35
A:;' 257, 264 blades 35, principal - 46
abbreviations ad hoc 61 box 61
absolute differential invariant 164 bra 54
- invariance w. r. to the streamlines 107 breakable objects 285
- parallelism 142 BRINKMAN, theorem of 314
adaptation, theorem of 73
ADATI, problern of 321 ff.
Cn 305
addition of quantities 13
adjoint 34, 38
en 416
canonical congruences 245, - forms
- group 198
33, 93 ff.
- group linear 199
CARTAN's symbolical method 1 72
- system 80
cartesian 42
affine conic 299
cell 65
- group Ga 1
centre 2, 224
- length 233
centred En 2
- motions 346f., 370
characteristic equation 29
- parameter 155, 233, 299
characteristics 90
- space 1
CHRISTOFFEL symbol 132
alternated derivation 83
class, of field 70, of function 62, 66, of
alterna ting 14
pseudo-quantity 11, 69, of vectorfield
differential form 118
89, of V". 268
multiplication 23
co-, contra-, cocontravariant 6f., 10, 28,
- part 14
34, 35
- tensor 22
CoDAZZI equation 238, 242, 262f., 266f.,
analytic connexion in Un 393, 403
311, 410f.
- functions 62
collecting indices 57
analyticity of a field 66, 70
complementary subspaces 287
anholonomic coordinate system 99ff.
complete systems 81
- object !J~1 100
components 6, 11, 61, 68
angles 44
concircular 322, 330ff., 334
- principal 48
concurrent 322
asymptotic lines 237, 250, 271
concomitant 15, 164
- vm of order p 275 conformal connexion 315 ff., - unitary
axes, principal 46, 56
connexion 418ff., - curvature tensor
axial point 272
306, 417, - parameters 346
conformal transformations 134, 304,
base 70, 71, 74, - theorems 74, 75, - 312ff., 415ff.
transformation 7 5 conformally euclidean 305ff., 416, -
basis vectors 7 f. geodesie 309 f.
bending 244 congruences in V,. 244ff.
 Index. 513

connexion 125f., 132f., 139, 187f. dummy indices 14

connecting objects 13, 68, 74f. DuPrN, indicatrix of 252, theorem of 250
"conservation de courhure et de torsion" duration 44
173
contact of order u 231
contain 4 En 1. centred - 2
contraction 13 Em-field 78. Em-part 19
Contragradient 6f. eigenvalue 29
coordinate axes, planes, Ep's 3. - sy- eigenvector 29
stems 42, 54, 62, 64, - Xm's 76, ErNSTEIN space 148, conformal transfor-
transformations 47. 64 mation of - - 312ff.
CoPSON and RusE, equations of 385 elementary divisors 33
covariant constant 125, 187, 322 enveloping 79
covariant differentiation 122ff., 129f., 336 "equations de structure" 173
covariant stationary 125 f. equiform 11
curvature 228f., absolute, relative, en- equilibrium, field in 1 50
forced- 271, - figure 271, higher- equipollences in group space 18G, in
27 5ff., lines of - 237. mean - 250, auxiliary X 2 ,. 390
principal - 250, - regions 271. 27 5. equivalent systems 73
riemannian- 153.- tensor 139. con- essential parameters of transformation
formal - tensor 289. projective - groups 201
tensor 289, recurrent - tensor 359ff., euclidean R,. 44
- tensor of valence 3: 256. 276. - EuLER, theorem of 253
vector of curve 228f., - vector, ab- extension 160, flat - 422, null - 422
solute, relative, enforced 250
curves 227ff., 412ff.
FERMI Coordinates 166ff.
fixed indices 3
Dn 290 flat submanifold 3
dead indices 7 free indices 14
decomposable spaces, objects 285, FRENET formulae for a curve in Vn 229.
dito conformally 287 for a Vm in V,. 277
definite, indefinite 35. 53 functional matrix 72
deformation 335ff.,- of subspaces 352ff. fundamental figure in Rn 41, - pseudo-
density, Ll-density 12 tensor 135.- tensor 41, 133ff., second
derived groups 224, - systems 81 - tensor 243. non symmetric - ten-
differential concomitant 164,- invariant, sor 179ff., hermitian - tensor 54,
absolute and relative 164 396ff., - theorems of transformation
dimension of null manifold 73 groups 203, 26ff.
direction 4
" direct" calcul us 58
displacement, CARTAN- of points 127f., Ga Gko G5 a 1, 2, 12
linear - 125, parallel - 122ff., - GAuss, equations of 237, 240, 242, 261,
over r v" 104 266f., 278ff., 311, 41f.
distance 40 general multiplication 13
Div, divergence 84f., theorem of 85 generating of a group 104
divisor of symmetric tensor 21, of geodesie 155. - coordinate system 156.
multivector 23 -in group space 188,- L'::; in L,. 263,
domain with respect to indices 20 - V"_ 1 in V,. 251, - lm in l,. 409,
dragging along of coordinate system 2, - variation 338
102, - - of field 102, 335. - - over geometric manifold 64, - object 61 f.,
v" dt 103 - point 64
D-symbolism 254 geometry of paths 303
 514 Index.

Grad, gradient 70, - equation 83, Ln 125, Ln 393, L'ft 257, 264
- product 84, - transformation 91 f. LAGRANGE, derivative 112ff., -equa-
group, affine 1, homogeneaus linear - 2, tion 112
- germ 186, r-parameter - 185, - LAME equations 296
space 1 86, 209 ff. length 40
LICHNEROWIC zequation 384,- inequality
388
.~ 100 LIE derivation 104, 106ff., 152, 336
harmonic 381 ff., 386 LIE operator, extension of 372
hermitian 52 lie in 4
holonomy group 361 ff., 367 ff. linear element of un 227,- submanifold 3
homogeneaus equations 80, - linear living indices 7
group Gho 2 LrouvrLLE problem, generalization of 312
homologaus elements and subgroups 198 LORENTZ transformation 43
homothetic motions 346
hybrid quantities 51, connexions for--
388ff. manner of transformation 6
hyperplane 3 mapping of spaces 174 ff., 349 f.
mean curvature 250, 272
MEUSNIER, theorem of 250
ideal vectors 58 minimal regular 73. 7 5. - - vm 44
identifications 27 minkowskian R" 44, - reflexions and
identities for the curvature quantities rotations 43
144ff., 393ff., 398ff. minor 1
imbedding 66, 74f., 279ff., 407 mixing 10, 14
improper rotations 43, SO motions, affine 3 70, conformal 346, homo-
indices 3. 7, 14, 53. 57 thetic 346, projective 346, in vn and
indicatrix 40 Ln 346ff.
indefinite 3 5. 53 multiplication, general 13, symmetric 21,
infinitesimal transformations 104f., 203 alternating 23
inner orientation 5, - problern 79 multivector 22f.
intermediate components 80, 87, 153 multivector-tensor 34
integrable connexion 139, 393, 403,
- groups 225 f.
integral, - function 80 n-vector 23
invariant 5, 15, - of group 214ff., natural derivative 70, 83 f., - equations
220ff., - subgroup 198, 224 230, - field value 335, - parameter
invertible 53 414, - variation 338
irreducible 21 neighbourhood 62, 66
isomer 13 net 3, 76, 100
isotropic Ep 14, 42 non homogeneaus equations 82
isotropy groups 373 norm 54
normal coordinates 155 ff., 199, - system
76, - tensors 159
join 4 normalizing conditions 235, 237
normals 228 f.
nullcone 44, 55
Kn, K! 421 null Ep 44, - manifold 73, - point 73. 80
KHLER space 397
kernel-index method 3
ket 54 object field 70
kind of an object 6, 68 osculating p-direction 232
KRONECKER symbol 3 opposite 5
 Index. 515

ordinary Un, Vn 13 3 reciprocal sets 9

orien ta tion 5 rectilinear coordina tes 1
outer orientation 5, - problern 79 recurrent curvature 359ff., 421 ff., --
osculating R's 228f. fields 126, 322
orthogonal coordinate transformation 47, reduced system 82
- group 47, 54,- net 246,- system reflexional orthogonal transformation 50
246 reflexota tion 4 7
region 44, 62, 65, 71
representative indices 57
parallel 4, 122ff., 126, 187, 332 RICCI equations 261 f., 266f., 278ff., 311,
parameters of displacement 122, pro- 410, 412, - space 148, - tensor 141
jective - 301 f, conformal - 346 RIEMANN-CHRISTOFFEL tensor 139
parameter-groups 191 ff. riemannian curvature 153
parametric form 7 5 rigging 17, 77, 79
part 19, 77, symmetric and alternating- Rot, rotation 83, theorem of - 85
14 rotations 47, proper and improper - 50
p-direction 4 rotation class 89, coefficients of - 171,
perpendicular 42, 51 minkowskian - 4 3
PFAFF's problern 79, 89ff. running indices 3
planar point 272
plane 3
point 1, 3, - transformation 2 Sn 148, Sn 404
pole 317 satura ted indices 14
polycylinder 61 scalar 6
primitive groups 218 SCHLFLI, theorem of 268
principal angles 48, - axes 46, 56, - ScHuR space 252, theorem of - 153
blades 46, - curvature 237, - direc- screwsense 5
tions 146, 153, - multidirections 46, section 4, 5, 16, 77 f., 89
56, - tangents 237, - Xn in auxiliary semi-analytic 389, - -definite 3 5, 53
X2n 390 sense 5, - of rotation 5
product integral 340, - spaces 285 ff. signature 34, 53
projection 5, 18 similarity class 89, - transformation 91 f.
projective connexions 300ff., - curva- simple multivector 23
ture tensor 289, - parameters 301 f., simple and semi-simple groups 227
- transformations 287 ff., 296, 415, skeleton 59, 255
restricted - transformations 288 space of constant curvature 148
projective euclidean 288 spanning 4
proper rotations 50 spatial point 272
pseudo-group 64, -normal vector 235, special affine group Gsa 12
-quantities 11, 69, -scalar 11, -tensor 11, spur 29
-vector 11 stationary 228f.
STOKES, theorem of 97
strangling 10
quadrivector 23 subgroup 198, 223, invariant - 198, 224,
quantities 6, 13, 54, 68 representation of - 221 ff.
quasi-plane curves 298 subgeodesie 327
submanifold, flat 3
subprojective connexion 317ff.,- trans-
Rn 41, Rn 54 formation 327 ff., - V,. 328ff.
raising of indices 44 support of domain 20, - of rotation 90
rank 3, 72f. Summation convention 1
reduction theorem 164 surface 74
516 Index.

structural constants 196, 205, - equa- Un 133, n 396

tions of Ln 173, - formulae of trans- umbilical point 251. 272, 310
formation groups after LIE and MAu- unified field theory 1 79 ff.
RER-CARTAN 196, 205 unitary connexion 395ff., - orthogonal
symbolical method of CARTAN 11 7 ff. group 54, - perpendicular 54
symmetric 14, - tensor 21, - multi- unitvector 44, 54
plica tion 21 , - part 14 unity tensor 10

tangent Em 77, principal - 250, - space

69, - vector 23 5
V" 133, V" 397. vnm 270
valence 10
teleparallelism 141
varia tion 3 3 7 f.
tensor 10, 21 f., - density 12, unity -
vector 6, - of projective transformation
10, - of valence 2: 28ff.
287
torse-forming field 3 22
volume preserving 144, 155, 288
totally integrable 80
trace 29
transformation groups 201 ff.
transitive group 216f. WEYL geometry 133
translation 4, 349 W-tensor 12
transposition of variationproblern 273, 341
transvection 8, 14
trivectors 23, classification o f - 35f. Xn 64, X;;' 253

Symbols.
_>!<_ 3 [] 14 Ll 340
17 122, 139, 149, 174ff.,
oc 25 {} 67 {{}} 132
187f .. 336, 419.
() 14 U'. .}
132 f, 336
V