A General Theory of a Cosserat Surface

A. E . GREEN, P . M. NAGHDI ~ZW . L.WAINWRIGHT

Abstract
This paper is concerned with a general dynamical theory of a Cosserat surface, i.e., a
deformable surface embedded in a Euclidean 3-space to every point of which a deformable vector
is assigned. These deformable vectors, called directors, are not necessarily along the normals
to the surface and possess the property that they remain invariant in length under rigid body
motions. An elastic Cosserat surface and other special cases of the theory which bear directly
on the classical theory of elastic shells are also discussed.

Contents Page
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
2. Preliminaries. Kinematical results . . . . . . . . . . . . . . . . . . . . . . . . 289
3. Theory of a Cosserat surface . . . . . . . . . . . . . . . . . . . . . . . . . . 292
4. Alternative forms of the basic field equations . . . . . . . . . . . . . . . . . . . 296
5. An elastic Cosserat surface . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
6. Infinitesimal theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300
7. Special cases of the general theory . . . . . . . . . . . . . . . . . . . . . . . . 303
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308

1. Introduction
Consider a surface embedded in a Euclidean 3-space to every point of which
a vector - not necessarily along the normal to the surface - called a director,
is assigned. Such a surface with deformable directors will be called a Cosserat
surface, and the present paper is concerned with a general dynamical theory of
such a surface in continuum mechanics.
The idea of using directors in continuum mechanics evidently goes back to
D u r I ~ (1893), who regarded a body as a collection of points together with direc-
tions associated with the points. Theories based on such a model of an oriented
medium were further developed by E. & F. COSSERAT(1909). A modern account
of the kinematics of oriented bodies characterized by ordinary displacement and
the independent deformation of n directors in n-dimensional space, has been given
by ERICKSEN& TRUESDELL (1958). An exposition of the kinematics of oriented
bodies, together with references to other contributions on the subject up to 1960,
is given in the monograph by TRUESDELL& TOUPIN (1960). Related and more
recent contributions to the subject include the use of directors by ERICKSEN(1961)
in his theory of liquid crystals, the development of a general theory of multipolar
continuum mechanics by GREEN & RIVLIN (1964b), a discussion of theories of
elasticity with couple-stress by TOUPIN (1964), and a complete dynamical theory
of directors by GREEN, NAGHDI & RIVLIN (1965), where the relation of directors
to multipolar displacements is also discussed. Additional references may be found
in the papers already cited.
288 A.E. G~N, P.M. NAGI-IDI& W. L. WAINWmOn'r:

As noted previously by ERICKSEN& TRUESDELL(1958), the COSSERATS(1908)

had recognized the significance of the idea of oriented bodies in one and two
dimensions for the construction of theories of rods and shells and included, in
particular, a development of a theory of shells in their book (1909)which, however,
is incomplete. A general theory of strain for rods and shells, suggested by the work
of the COSSERATS, was developed more recently by ERICKSEN& TRUESDELL in
their (1958) paper, where a derivation of the differential equations of equilibrium
for shells is also supplied; ERICKSEN& TRUKSDELL(1958) do not, however, con-
sider the problem of constitutive equations. A systematic development of a general
and complete theory of elastic shells by direct methods (in contrast to the use of
the three-dimensional equations of elasticity theory) is not available. In a recent
paper SERmN (1963), however, considers an exact and complete (isothermal)
linear theory of elastic shells by a direct method; but in early stages of his analysis
he assumes a strain energy function for the linear theory which is too restrictive.
The present paper is concerned with a general theory of a COSSERATsurface
which is exact, complete, and fully consistent with dynamical and thermodynami-
cal principles of continuum mechanics. After some geometrical preliminaries, we
derive in Section 2 the necessary kinematical results due to the classical (mono-
polar) surface displacement and the director displacement of a Cosserat surface
and also consider, for future use, the nature of these kinematical results under
superposed rigid body motions. Next, following GREEN & RIVLIN (1964a, b) and
using a principle of balance of energy and an entropy production inequality, valid
for a surface embedded in a Euclidean 3-space, we develop in Section 3 a general
theory of a Cosserat surface whose deformation is characterized by both the
transformation of its coordinates as well as that of its directors [see Eq. (2.6)].
The derivation in Section 3, carried out in a neat vectorial form, leads to an appro-
priate (local) equation for conservation of mass and various (local) dynamical and
thermodynamical equations which hold for any Cosserat surface, as defined here.
These equations which (in vector form) display a remarkably simple structure,
conceal the relative complexity of the results and hence alternative forms of the
basic equations in terms of tensor components are collected in Section 4.
The remaining parts of the paper deal with an elastic Cosserat surface and
specialized cases of the general theory. Nonlinear constitutive equations for an
elastic Cosserat surface in terms of both the Helmholtz free energy and the internal
energy functions are constructed in Section 5. By considering the effect of material
symmetries, explicit results are also deduced in Section 5 for an initially isotropic
Cosserat surface in terms of joint invariants of temperature and 12 kinematical
quantities. Section 6 is concerned with the infinitesimal theory of an elastic Cosserat
surface, where the basic field equations and the constitutive equations are systema-
tically and consistently specialized to correspond with a suitable linearization of
the kinematical results. Finally, special cases of the general theory involving
further simplifying assumptions, motivated by classical treatments of the theories
of shells, are discussed in Section 7. These include a special case of the theory of
Sections 3 and 5, as well as that of the linearized theory of Section 6, in which the
directors are identified with the unit normals to the surface.
Although the use of more general kinematical ingredients would be possible,
the general theory of this paper for a Cosserat surface (with deformable directors)
 Theory of Cosserat Surfaces 289

is exact and complete from both dynamical and thermodynamical points of view.
Almost all of the field equations of Section 3 are either entirely novel or more
general than the corresponding previously known results. Even the differential
equations of equihbrium in Section 3 (or Section 4) are more general than those
derived previously by direct methods. Also, the specialized cases of both the non-
linear and the linear theory in which the director is identified with the normal
to the surface (Section 7) are rather illuminating and bear directly on the founda-
tions of the classical theory of shells developed under the so-called Kirchhoff-
Love hypothesis.
2 . P r e l i m i n a r i e s . K i n e m a t i c a l results
Let z ~(i= 1, 2, 3) refer to a fixed right-handed Cartesian coordinate system
and let x i denote an arbitrary (real) curvilinear coordinate system defined by the
transformation
zi=z'(xl, x2,x3, t), det ~ 0z'. ] > 0 (2.1)
L x' _l '
and its inverse. Let a surface ~, embedded in a Euclidean 3-space, be defined by
the equation x a = 0 and let x ~ be identified as convected normal coordinates with
x ~ (0t = 1, 2) on ~ and x 3 along the normal to ~. Also, let the initial values of z t at
time t = 0 be designated by Z ~ and refer to the (initial) undeformed surface by 6e
and to the deformed surface at time t by ~. We can show that the metric tensor
g~j of the coordinate system x ~, when evaluated on x 3 =0, is given by

g~#=a~a, g3~=0, g33=1, (2.2)

where a~ a is the first fundamental form of a. We can also show that det [d zi/8 x J] :t:0
on ~ so that the conjugate tensor g~i, with components g~a=a ~a, g3~=0, g33= 1,
exists.
The position vector of a point on 6a will be denoted by R and its dual on ~ by
r. Thus, if a~ are the base vectors along the x%curves on ~, and a 3 is the unit normal
to ~, we have
a~=r,~, a~. aa=a~#, a ~. a#=6~#, (2.3)
a~- a3 =0, a3- a3 = I, as,~.a3=O, (2.4)

where 6~ is the Kronecker symbol in 2-space and a comma denotes partial differ-
entiation with respect to x ~. We also recall the well-known formulae from differ-
ential geometry, namely
a~la=b~pa3,
a s, # = - b~ a , , (2.5)

b~ air = b, rl#,

where b~# is the second fundamental form of the surface a and a single stroke
designates covariant differentiation with respect to a,#. Results similar to those
in (2.3) to (2.5) hold also for the surface ~ whose base vectors and the first and
second fundamental forms will be denoted by A s, A~# and B~#, respectively.
290 A.E. GREEN, P.M. NAGHDI • W.L. WAINWRIGHT :

Throughout the paper, Latin indices will have the range 1, 2, 3, whereas Greek
indices with the range 1, 2 are used for surface tensors or the components of
space tensors.
Let a vector D, not necessarily along the normal to the initial surface, be
assigned to every point of 6p. The vector D will be called a director and its dual
at time t on ~ will be denoted by d. In what follows, we regard the motion of a
Cosserat surface to be characterized by
zi= zS(x ~, t), d = d(x ~, t), (2.6)
where d has the property that its components, referred to the base vectors a s,
remain invariant when the motion is altered only by superposed rigid body motions.
Equivalently, given a surface 5a with fundamental forms A,a, B,a and director
D at Z s, we wish to determine a second surface ~ and the director d at z s, accord-
ing to (2.6), uniquely to within a rigid motion.
Let v, a three-dimensional vector field, denote the velocity of ~ at time t which,
when referred to the base vectors as={a~,, aa} of ~, is
v .= v sa i = v ~ a , + v 3 a s = v i a s. (2.7)
Since the coordinate curves on ~ are convected, it follows that*
h~=v , , v,~,=(vt~l~-bp~,v3) atJ +(vs,~,+b~v~)aa, (2.8)
and from (2.4)i and (2.4)2 we can show that
fia = - ( v 3 , , + b~ vp) a', (2.9)
where a superposed dot denotes the material derivative with respect to t, holding
x" fixed. The lowering and raising of superscripts and subscripts of space tensor
functions such as vs in (2.7) and Ck sin (2. I0) can be accomplished by using the metric
tensor gij defined by (2.2).
Since, for each i, ds may be expressed as a linear combination of a s, we may
write
ai=Cks ak, Cks=ak" fis. (2.10)
Let r/kS and ~bkSstand, respectively, for the symmetric and antisymmetric parts of
ck s. Then,
Cks=rlkS+l, kkS, (2.11)
where
2rlk i=ak " as=ak" ;'i+as" dk=ctki=2rhk, (2.12)
2~Oki= a k 9hi--a i 9hk = --21,OSk. (2.13)
From (2.12) and (2.13) together with (2.8)1, (2.4)1 and (2.4)2, we conclude that

2 q ~ = a ~ . v,~+atJ, v,~
=V~l~ + Vpl~,- 2 b,, p v3 , (2.14)
? ] 3 ~ : ?]~t3 : 0 , 1~3 3 ~--'0 ,
9 See, e.g., GREEN & ZERNA (1954).
 Theory of Cosserat Surfaces 291

and
2~k,a= - 2~Op,= v,,Ip- V~l,,,
~*~3---- --~/3a = --a3"v,~= -(Va,~-l-b~vp), (2.15)

~33=0,

where q,p and ~O,a may be referred to as the surface deformation rate tensor and
the surface spin tensor, respectively. We also note that in view of (2.11), (2.10)t
may also be expressed in the form
hi = (/~k i "{-~r i) a k. (2.16)

To obtain suitable expressions for the rate of change of the reciprocal base
vectors, we recall that the reciprocal base vectors are given by

a~=a'~a a, aa=a3 . (2.17)

Hence, with the use of (2.12) to (2.17), and the result

~= _ a ~ atJV ~ ~,
we obtain
~ ' : a ' a aa

=a" ;'(qk ;, + r a) ak-- 2a'aaaV rl~.vap (2.18)

= a" Z(~k Z-- ~k 4) ak,

and
a3 = a 3 =l~k 3 ak. (2.19)

We now introduce further kinematical results in terms of the director dis-

placement d and its derivatives. The vector d referred to the base vectors a i is

d = d i a i = d~ a ~ + d 3 a 3 (2.20)

and the director velocity d, with the aid of (2.18) and (2.19), can be put in the form

w = d = F + d i i t i = F + d i I[Iki ak +d~q~, (2.21)

where
q~= --?lka ak, F=d~a~. (2.22)
Also,
d,~=2i~a i
(2.23)
=+4[= a% g=
where
2p~=a# 9 d,~=d#l ~ - b a ~ d 3 , ~..P~=a#v 2v:,
2a~=a a 9d,~=da,~+b~dt~, 2.3~=23~, (2.24)
292 A.E. GREEN,P.M. NAGHDI& W.L. WAINWRIGHT:

For future reference, we also calculate the rate of change of an element of

area da of the surface a. Since

d t r = ( a l x a2- a3) d x l d x 2 = a 89d x l d x 2

and a = l a = a l , by (2.12), we have

~= 89189 itdxtdx 2
(2.25)
=rl~ d tr =(V~l~ - b~ va) d tr .
In the remainder of this section, we introduce further kinematical results which
will be utilized in Section 3. In particular, we consider a second motion of a which
differs from the previous motion only by a superposed rigid motion when the
medium has the same orientation as in the first motion. Let the velocity vector at
time t, corresponding to the second motion, be denoted by v*. Then

v* = v + [Vo + x ( r - to)]
(2.26)
=v+[b+toxr],

where the square bracket represents an arbitrary velocity due to rigid motion,
and b = ( % - t o x to), vo and to are vector functions which depend on t only.
If in (2.12) and (2.13) we replace ~ii by [~ii+ t a x as], then it can be easily verified
that
~lk~=qkt, ~k~i=~ki--f2k~, (2.27)
where
f2k i = ek i m COrn= a+ ek i m corn, (2.28)

and the e-system is related to the permutation symbols ek t,~, ek ira, through
eki,n=a§ ekim, e k i m = a - 8 9 ekira (2.29)

Also, by (2.26), (2.3)1 and (2.28), we have

(2.30)
= v, ct -- ~Qk et a k.

When the motion is specified by (2.26), then the kinematical results correspond-
ing to (2.20) to (2.24) become
d*=d~, 2*~=;t~, (2.31)
r*=r, r*:=r::, (2.32)
W* = W+ta X d.---w+dlf2ik a t`, (2.33)
w * , = w , , +ta x d . , = w , + 2!,f2ik at`. (2.34)

3. Theory of a Cosserat surface

Let tr, the area of a at time t, be bounded by a closed curve c and let v = v~ a ~
be the outward unit normal to c lying in the surface. If N is a three-dimensional
vector field and if, for all arbitrary velocity fields v, the scalar N 9 v represents a
 Theory of Cosserat Surfaces 293

rate of work per unit length of c, then N is a curve force vector measured per unit
length*. Similarly, if M is a three-dimensional vector field and if, for all arbitrary
director velocity fields w, the scalar M 9 w is a rate of work per unit length, then
M will be called the director force vector, per unit length of c. Also, let F and L
stand, respectively, for the three-dimensional fields of assignedforce and director
force, per unit mass of ~, at time t, such that F 9 v and L. w, for all arbitrary v
and w, respectively, represent rate of work per unit area of a. When referred to the
base vectors a~, these vector fields can be expressed as

N= Ni ai= N~a~+ N 3 a 3 , (3.1)

M = M i a~ = M ~ a~ + M 3 a 3 , (3.2)
and
F=Fiai, L=I/,a i . (3.3)

If p is the mass density at time t per unit area of ~ and U is the internal energy per
unit mass, then the equation of balance of energy may be written as**

D S [89 o2 +p U]da=~p[r+F v+[,.w]da+~[N.v+M.w]dc-~hdc; (3.4)

ot t~ r C

where v denotes the magnitude of v, L is the difference of the assigned director

force per unit mass L and the inertia terms due to the director displacement
d, r is the heat supply function per unit mass per unit time, h is the flux of heat
across c per unit length per unit time, and D/Dt stands for material derivative.
After carrying out the indicated differentiation in (3.4) and recalling (2.25),
we obtain

[p v. v+ p 6 ] d a + ~ {89v- v + U} [~; + ; (v'l,- b~ va)] d

(3.5)
=Ip[r+e.v+l,.w] da+S[N.v+M.w-h]dc,
a c

where a superposed dot stands for D/Dt. Now assume that p, U, r, h, F - ~ , L, N, M

and w remain unchanged under superposed uniform rigid body translational
velocities***. Thus, in the notation of (2.26), if we replace v by (v+b) in (3.5),
then by subtraction we deduce

b. { S ( p v - p g ) d a - ~Ndc+ S v[Pq-P(V~l,-b~v3)] d a } +
r c

+89 b)~ [/J+p(v~l~-b~,v3)] d a = 0 (3.6)

ff

which holds for all arbitrary b. By replacing b by/3 b,/3 being a scalar, it then fol-
lows from (3.6) that
[fi + p (V'l, - b~ v3)] d a = 0 (3.7)
~r

* N is called a curve force vector, in analogy to a corresponding surface force vector in a

three-dimensional continuum.
** In this form of (3.4) the inertia term in ! contains also a contribution --89
q-p (V*l~--b~v3)] which eventually vanishes in view of (3.8).
*** The operator DIDt is unaltered by superposed rigid body motions.
294 A.E. GREEN,P.M. NAGHDIr W.L. WAINWRIGHT:

for all arbitrary areas a. From (3.7) we deduce that

Dp
D t ~-P(V'l'- b~v3) = p + p r/:=0 (3.8)

which is the (local) equation for conservation of mass. Also, using (3.8), we obtain
from (3.6) the integral form of the equations of motion, namely

S (p(~-p F) dtr- S N dc=O. (3.9)

ff C

Over a curve with a unit normal v = v, a', the (physical) curve force vector is
N. If n" are the (physical) force vectors over each coordinate line, then the appli-
cation of (3.9) to a curvilinear triangle yields

N=~v~,n'(a~'~89 N'=n'(a~") ~. (3.10)

O~

Thus, N" transforms as a contravariant surface vector and we can put*

N~=Nr~ ar + Na~ a3, (3.11)
and by (3.1)
Nr=Nr~v~, N3=N3~v~,, (3.12)

where N ~" and _N3 ~ are surface tensors under transformation of surface coordinates.
Substituting (3.10) into (3.9), making the usual smoothness assumptions and
applying Stokes' theorem to the line integral, we obtain

S [P(v-F)-N'I,] da=O (3.13)

Cr

which holds for arbitrary a. Hence, from the vanishing of the integrand follow
the equations of motion
N"l,+pF=p~. (3.14)
Returning to (3.5), since
S N . v dc= ~ I N ' I , . v + N ' . v.,] dtr,
c ~r

then with the use of (3.14) and the continuity equation (3.8), the energy equation
becomes
~p(Jda=~[p(r+l,.w)+N'.r,,]da+~[M.w-h]dc. (3.15)
ar tT r

If h" is the flux of heat across the x'-curves and m" is the (physical) director
force vector over these curves, then application of (3.15) to a curvilinear triangle
gives
,~- w - h = 0 , (3.16)
where we have set
-h=h-q~'v~,, q'*=h~(d'")
89 (3.17)
M=M-M~v~,, M~=ra~'(a~)§ (3.18)
* The order of indices in Nr~, N 3a differs from that used, e.g., by GREEN& NAGnDI(1965).
 Theory of Cosserat Surfaces 295

Also, after substituting (3.16) into (3.15) and applying the result to an arbitrary
surface area, we have
p r - q ~ l ~ - p U + m . w+ N ~ . v , ~ + M ~. w,~=O, (3.19)
where
m = M~I~ + p/~. (3.20)

We now use the usual argument about uniform superposed rigid body angular
velocities, the continuum occupying the same position at time t. We assume that
p, r, q', U, m, N ~, M ", h, are unaltered under such uniform superposed rigid body
motion and then with the use of (2.27), (2.33) and (2.34), equations (3.16) and
(3.19) yield
, ~ . F + M . t/~,d ' - h = O , (3.21)
pr-q'l,-p(J+m.(F+thd~)-N'.rl~+m~.(F:~+2B.~tlp)=O, (3.22)
together with
dxM=0, (3.23)
and
N ~ x a,-t-(M ~ x d)l,,+p L x d=O. (3.24)

We complete the general theory by stating an entropy production inequality

in the form
s s h c->-0
which holds for all arbitrary surfaces, and where S is the entropy per unit mass
and T(>0) is the temperature. If we apply (3.25) to an arbitrary curvilinear
triangle bounded by coordinate curves on ~, we obtain
h - v~ q~__>0. (3.26)
In the special case when h does not depend explicitly on the director velocity w,
it may be shown from (3.16) or (3.21) that h = 0 and then, by (3.17)1,
h -- q~ v~. (3.27)
Inspection of equations (3.21) and (3.22) suggests that for a complete theory
of agiven Cosserat surface constitutive equations are required for U, N ~, M ~, m,
M, h and q~ and these can be reduced to a canonical form with the aid of invariance
conditions for each equation which keeps the left-hand sides of (3.21) and (3.22)
unaltered by all superposed rigid body motions. It may be noted here that follow-
ing the discussion given by GREEN &; RIVLIN (1964b), we can assume that h is a
scalar and M is an (invariant) vector under transformation of surface coordinates.
It then follows from (3.17), (3.18) and (3.20) that q= are components of a contra-
variant vector, m is a vector, and M ~ transforms as a contravariant vector under
transformation of surface coordinates. Also, if we set
M~= Mi~ ai= MV~ ar + Ma~ a3 , (3.28)
then M r~ and M 3~ a r e surface tensors.
296 A.E. GREEN, P.M. NAGHDI& W. L. WAINWRIGHT:

4. Alternative forms oI the basic field equations

The elegance and the simplicity of the derivation in the previous section does
not display the relative complexity of the results. For this reason and for future
reference, we obtain here the basic equations of Section 3 in tensor components.
To obtain the equations of motion in component form, consider the scalar
product of (3.14) with a ~ and again with a 3 and deduce

(a t3 9 N~)I - N ~ 9 a#l=+ P(F-i~) . a # = 0 ,

(a 3 9N=)I,- N ~ . a31~-b p ( F - ~ ) 9a 3 = 0 ,
or equivalently
NP ~1~- b~ Na~+ p F p = p cp,
(4.1)
N3~I~ + b~p N p ~+p F a = p c a,

where (3.11) and (2.5) have been used and where c~= {cp, c 3} are the components
of acceleration. In an entirely similar manner, we obtain from (3.20) and (3.28) the
equations
M~ ~1~- b~ M a ~ + p LP = m p,
(4.2)
Ma~l~+b=t ~M t ~ ' + p L a = m 3 ,

where L ' = L . a ~ are the components of I,, the difference of the assigned director
force per unit mass and the inertia terms due to the director displacement d, and
mS= m 9 a ~for which, in general, constitutive equations are required.
We now deduce the component forms of (3.23) and (3.24). W i t h / ~ i = ~ . a t,
from (3.23) follows the result e~k d ~ Mk=0, or equivalently
e,p d" J ~ = 0 , -Mad~=M~d3, (4.3)

where 8,p=e,~ a = a 89e,o and ex 1 = e 2 2 =0, el 2 = - e 2 1 = 1. Similarly, from (3.24)

we have
U"+ dJ M"] =0,
or equivalently
ep ~[N p ~+ m p d" + M p r 2.~] = 0,
(4.4)
N3" +(ma d ~ - m ~ da)+ Ma~ 2~. - M ~ 2a. = 0 ,
where 2.~ and 2.av are given by (2.24).
In addition, equation (3.21) and the energy equation (3.22) in component forms
may be written as
(4.5)

(4.6)
where
(4.7)

Also, it is convenient to introduce here the Helmholtz free energy function per unit
mass as
A = U-- TS (4.8)
 Theory of Cosserat Surfaces 297

and express the energy equation (4.6) in the form

pr--q't,--p(TS+~i's)--pii+N'P'q,a+m'd,+M"2,,=O. (4.9)

Thus, in addition to the (local) equation (3.8) for conservation of mass and
the entropy production inequality (3.25), the remaining basic equations and sym-
metry conditions are given by (4.1), (4.2), (4.5), (4.6) or (4.7), and (4.3) and (4.4).
These equations are the counterparts of those given in vector form by (3.14),
(3.20), (3.20, (3.22), and (3.23) and (3.24), respectively. From the discussion con-
cerning the required constitutive equations at the end of Section 3, it is clear that
N ' "p is a tensor and the left-hand sides of (4.5)and (4.6), as well as (4.9) are scalars.
Also, in the construction of the constitutive equations we assume that A (or U),
N ' ~'P, m i, M ~', -h, and q" are all unaltered under superposed rigid body motions
so that, in view of the behaviour of the kinematic variables, the left-hand sides of
(4.5) and (4.9), or (4.6), remain unaltered by such motions.

5. An elastic Cosserat surtace

We define an elastic Cosserat surface as one for which the following constitutive
equations hold for all time t:
A = A ( T , e~p, ~i~, 3~, As~, Di),
(5.1)
S = S(T, e,p, xi,, 6i, A~,, D3,
q ~ = q ~ ( T , T.r , er6, x i v , 6i, A i r , Di) (5.2)
and
h=h(T,e,,a,xi~,6~,Aio,,Di, vo,),
(5.3)
M ' = M ' ( T, e~,p , toil,, 6 i , Ai ~,, Di , v~,) ,
as well as
N'"a=N'*'a(T, er6,xi~,,6i,Air,Di),
m i = m i ( T , e~,a, x j ~,, 6 i , A i,,, D j ) , (5.4)
M i*' = M i ~'(T, ey 6, x j r, 6 i , Ay r, D j ) ,
where we have set
2 e,, a = a,, a - A,, p , (5.5)
xi,, = Ai ~,-- A ~,, , (5.6)

3~ = d~ - Di, (5.7)

and where A ~ , , = A ~ . D,,, is the value of 2i~ in the reference state*. We note here
that
(5.8)
and e~a, tq~, 6~ are unaltered under arbitrary superposed rigid body motions.
* The variables e~/~,~cl~and Jt in the constitutive equations (5.1) to (5.4) are introduced for
convenience. Alternatively, the kinematic variables in the constitutive equations may be taken
as e,~a, ,~0,, d l together with the initial values Al~ and D I. Dependence on the initial metric tensor
A~a is also understood.
298 A.E. GREEN, P.M. NAGFIDI 8r W.L. WAINWRIGHT:

Recalling that equation (3.16) holds for all arbitrary values of w and since,
in view of (5.3), the coefficients h and M are independent of the director velocity,
it follows that*
=0, (5.9)
Hence, by (3.17),
h = q v~, (5.10)
and by (3.18) and (3.28),
M = M ~,v~,=M t al, M i = M i ~ v~,, (5.11)

so that (4.3)t and (4.3) 2 are identically satisfied.

Using (5.10) in equation (3.25), making the usual smoothness assumptions
and transforming the line integral into a surface integral, we obtain

pTS-pr+q , I, q'T
T, >- 0 , (5.12)

since (3.25) holds for all arbitrary surfaces a. From (5.12) and (4.9), we have

-PT"S--Pfl+N'~'thp+midi+M*~2i~ 9 q~ TT., _>0.

- (5.13)

Introducing (5.1)1 into (5.13) and taking account of (5.1) 2, (5.2), (5.4) and (5.8),
we deduce that

(5.14)
/ ~ aA \ q'T,_>0

which holds for all arbitrary values of 2r, ~ha, d~ and 2i~ at time t. By considering
a homogeneous temperature distribution varying with time, it follows from (5.14)
that**
dA
S = - 0---Y' (5.15)

aA i t3A Mi~,=p aA (5.16)

N'~=P ~e~p ' m =P-ff~i ' t~i~ '

-q~T~>0, (5.17)

and the heat supply function r is determined from (4.9).

If desired, expressions for N' ~~, rn i and M i" may also be obtained in terms of the
internal energy U in which case, through (5.15), T is expressed as a function of
S and the kinematic variables e,~, ~:~, 6~, A,~ and D~. Thus, with
U = U ( S , e~ 13, ~q ~, ~ , Ai ~, D~), (5.18)
* The results (5.9) may also be deduced from (3.21).
** See COLEMAN& NOLL(1963). TO avoid ambiguity in evaluating aA/ae~lJ, the tensor e~#
is understood to stand for 89 e/~).
 Theory of Cosserat Surfaces 299

instead of (5.15) and (5.16), we have

dU
T= O~' (5.19)

U '
N ' # ~ ' = P de~,# m ' .= p t~
~riU ' M i , = P Otq~,
~U " (5.20)

The above nonlinear constitutive equations are valid for an elastic Cosserat
surface which is anisotropic in some preferred state, usually taken to be the
initial undeformed state. In the remainder of this section, we omit a detailed dis-
cussion of the heat conduction vector q~, but consider briefly a more explicit
form of the free energy function for a Cosserat surface which initially has holohedral
isotropy, i.e., isotropy with a center of symmetry, to which the work of RIVLIN
(1955) and ADKINS (1960) can be adapted. A discussion of the effect of material
symmetries in restricting the form of constitutive equations is given by Gm~E~ &
ADKINS (1960). References to more general developments on symmetry restrictions
may be found in a recent paper by WIN~MAN & PIPKIN (1964) who have shown
that the results of RIVLrN, SPENCER, ADKINS and others can be used, with only
minor changes, for any type of dependence of the functions on their arguments
and not necessarily polynomial dependence.
Consider the tangent plane and the normal at each point R of 5e with which
point, regarded as the origin of a Cartesian coordinate system, we associate a
triad of mutually orthogonal unit vectors H,, such that H~ are in the tangent plane
and H 3 is directed along the normal. If a typical vector Tis decomposed in such a
coordinate system along the local unit vectors Hm into tangential components
and a normal component, then under a change from one coordinate system to
another the tangential components transform like the components of a 2-vector
under the orthogonal group and the third component remains invariant. The
symmetry of the material, defined by preferred directions in the initial undeformed
state is then characterized by H~ and the appropriate group of (2-dimensional)
transformations which specify the equivalent positions of the vectors H~ from one
system to another and the constitutive relations must then be form-invariant
under each transformation of this group. We consider here symmetries for which
the associated group of transformations is a sub-group of the full orthogonal
group.
When an elastic Cosserat surface is initially isotropic with a center of symmetry
and assuming that the energy function A is a polynomial in the arguments indica-
ted in (5.1)*, then we may use the works of RIVLIN (1955)and AOKINS (1960)**
to express A as a function of the joint invariants of the arguments in (5.1). Although,
in principle, there is no difficulty in writing down all the invariants, to avoid undue
complications we limit ourselves in the rest of this Section to the special case when
A does not depend explicitly on Ai~ and Di. For convenience, we introduce the

* Since the coordinate system used is convected, the invariance of (5.1), as well as (5.15)
and (5.16), under superposed rigid motions is automatically fulfilled.
** See, especially, table (5.5) and the statement (ii) on p. 268 of the paper by ADKn~S(1960).
Arch. Rational Mech. Anal., Vol. 20 21
300 A.E. GREEN, P.M. NAGHDI & W. L. WArSWSlGrrr:

notations
e•=A'r
ert~,
x:p=A'~tcrp, (5.21)
6" = A" r 6~,
and consider the 2 x 2 matrices
/, J, jT
(5.22)
g-~utt T, p=uv T, PT=vuT, Q=OVT,
where u and v are column vectors, the notation u r stands for the transpose of u,
and
I=e~, J=x~.a, u = x ~ ~, v=6 ~. (5.23)
Then, by forming the invariants of the seven matrices in (5.22) and omitting the
redundant elements, A may be expressed as a function of T, d3 and the following
twenty-four joint invariants:
tr I , tr J, tr K , tr P, tr Q,
tr 12, t r J 2, tr I J, trJJ r ,
tr l K , triP, tr l Q,
tr J K , tr J Q , tr J P, t r J P r, (5.24)
tr I J j r ,
trlJK, trlJQ, trlJP, t r l J P r,
trJJrK, trjjrQ, trjjrp.
The above invariants apply when A does not depend explicitly on Ai, and Di.
In the general case, in addition to (5.24), invariants arising from the functions in
(5.21) and A~p, A3p, D, should also be included.

6. Infinitesimal theory
The theory of an elastic Cosserat surface, where the displacements are infini-
tesimal, can be deduced as a special case of the general theory of Sections 4 and 5.
Let
r=R+~u, u=u~A~, v = ~ u , (6.1)
d=D+~*, 8 " = 6 " A i, w=~*, (6.2)
where e is a non-dimensional parameter. By (6.1) t (2.3) and (2.4), we can show
that
a3=A3 q-~-]-O(~2),
(6.3)
[J=~'A,=~A', 8,= -(u3,,+B~ur).
From (2.3), (2.4), (5.5) to (5.7), and (6.1) to (6.3) we have
e,/~= 89(a, tJ- A,p)=e [89(U,l~+ U~lO- B, p u3] + O (e2) , (6.4)
ir = e [6*- At~+ u, p. D, ,] + 0 (e2), (6.5)
x3, = e [8,*, 9A 3 + D , , . p] + O (e2), (6.6)
 Theory of Cosserat Surfaces 301

where
,~ A~=~I~-Br~r ,~ A 3 - ~ 3 , ~ + B ~ ,
u,a. O.,,=(Urla-Br~u3)(O~l~,-B~,rO3)+(u3, a+ B~ur)(D3,~,+ B',.O,), (6.7)

D,,. #=(D~t,- B~,~,D.,,)#~,

and in (6.4) and (6.7), and throughout the present section, a stroke denotes co-
variant differentiation with respect to A,p of the undeformed surface, in contrast
to the meaning of a stroke in earlier sections. The infinitesimal theory which we
wish to consider here is such that the displacements and the director displacements,
as well as their surface and time derivatives, remain small of the order of e. We
also assume that all kinematical quantities such as e,B and tq, and their rates are
of the order of 8; the curve and director forces such as N t" and M ~', expressed in
suitable non-dimensional form, are of the order of e; and that (T-To)IT o and
( S - S o ) / S o are of the order of e, where T O and S O refer to a standard temperature
and entropy of the initial undeformed surface.
In the following, we retain only terms of the order e and after the approxi-
mations, without loss in generality, we set e = 1 and obtain*

e~ ~,=89( U~,l~,W Url~)-- Ba ~,u 3 ,

K~~= ( ~*l , - B, ~~ ) + I-(uV I~-B~v u3)(n~l,-B,, 03) +
(6.8)
+(u3,~+B~up)(D3.~,+B~O~)],
~,+B~,~)+(DrI~,-B~,~,D3)ft , fl~,=-(u3,,+Br,,u~),

e=~=89
~r == (Wrl~ - B ~ w3) + [(V~l~ -- B~ v3) (Dvl=-- B~vD3)+
(6.9)
+ (v3,r + B~ vp) (03, ~+ B~ D,)],
k3, = (w3. ~+ B~ w~)+ (D,l~- B~~D3) ~ ,
and
v,=fq, w,= 6", / ~ = -(v3.,,+BY, vr). (6.10)

Also, now in all of the basic field equations (4.1) to (4.7), b,p, d~ and 2~ must be
replaced to order 8 by B~p, D i and Ai,,=A~ 9 D,,,, respectively; all tensors in these
equations are referred to the initial undeformed surface, and covariant differentia-
tion is with respect to A~p. In particular, for the infinitesimal theory, the energy
equation (4.9) assumes the form

Por-Q~'l,-po(TS+'i'S)-pofl+N'P~e~,p+m'6,+M'~,~,=O (6.11)

* Recalling (5.7), we have

21"
302 A.E. GREEN, P.M. NAGHDI (~. W. L. WAINWRIGHT."

and equation (5.10) and the constitutive equations (5.16) and (5.17) become,
respectively,
h = ov" Q~, (6.12)

dA . dA 9 t3A (6.13)
N'a==P~ ~e~,,a' m'=po~-~, M ' ~ = P ~ OK~= '

-Q'T~>0, (6.14)

where Po is the initial mass density, Q~ are the components of the heat flux per
unit length (in the undeformed surface) per unit time, and o v~ are the components
of the unit outward normal to the x~-curves on 6e.
If the surface is initially homogeneous, free from curve and director forces,
and in the state of rest at a constant temperature To and entropy So, then to the
order of approximation considered, it is sufficient to express Po A in (6.13) as a
quadratic function of e~p, r ~ , 6~ and T, where T is now the temperature difference
from To. Thus, if So = 0 ,

+ 3C~tare=alc3r+ a.C=Fr6=Kl~r+
--[-1Cat a t~ot(~lj --[-2 C~t fl K 3 atK 3 a -~- a Cat fl (~~ x 3 fl + (6.15)
+ 4 C ~a e~ a 63 + sC ~a x~ a 63 + 4 C ' ~a e= a T + sC' ~a r~ a T +
-I'- 1C~tba63"Jc-2C=K.aab3 -{- iC'=6= T + 2C' =/r = T +
"-[-C((~3)2--[-C ' T2 q-C"c~3 T,
where the coefficients

C, C', .C ~, nC'~, nC~l~, .C '~l~, .C ~ , ~C~ (n=1,2,...)

are constants* and some of them satisfy certain symmetry conditions, e.g.,
1C~ p ~ = I C a ~ 6 = I C ~ a 6 y = I C ~p, (6.16)
2C~ Cy~tfl, 3CCtfl76-..~3 CfttTa, 2Cr Cfl~y. (6.17)

We now restrict our attention to an elastic Cosserat surface possessing isotropy

with a center of symmetry. In this case, a tensor basis is given by** A ~a and since
there are no holohedral isotropic tensors of odd order, it follows that all odd order
coefficients in (6.15) must vanish. Hence

1C~#'~=2CatfT=3CatflT=4.CCt#7~_O,
(6.18)
i C ~ Z 2 C ~ t 1C, a = 2C, ct 0"

* In general, these coefficients will be functions of the initial values Ai~ and D i . But, to avoid
undue length, we restrict our attention throughout this section to the special case when A does
not depend explicitly on A~=and D i.
** See SMrm & RIvLIN(1957).
 Theory of Cosserat Surfaces 303

Moreover, the remaining coefficients in (6.15) must be homogeneous, linear

functions of products of A ~p so that, for example, 1C ~p~ may be written as
IC ~p r~=~ I A~PA ~ +~ 2 A ~ AP6 +~ 3 A*'~A p ~. (6.19)
But by (6.16) 1, cc2= ~3 and the same conclusion may be reached by (6.16)2. Similar
arguments can be applied to other coefficients in (6.15). In particular, it can be
shown that 3C ~p~e, as in 1C ~a~e, has only two independent constants but 2C ~p~6
involves three independent constants. Thus, the free energy poA may be finally
written in the form
PoA=89 A~PA~6 +o~2(A~rAP~+ A~6 AP~)Je~ae~+
-{-89~X3a~ Jg~ct~# -4-89 (63)2 -[-~x463 T+89

+89 A~ a n3~ x3a +cr A~a e~p 63 + ~ A ~a eaaT+ (6.20)

+ [oqo A"l~Ara +oql( A~ A~'~+ A"~ Aa r)] e~,ljtcr~+
"F-0[12 A~P tc,,a 63 +~'12 A '~p tGa T +
+O~13AO~'O6=K3p,
and then by (6.13), we also have
N ' " a = l-~t A~aAr~+~2(A~rAaa+A~Aa~)] eve+
+[oqoA~aAr'~+~zll(A~'rAa~+A~'~Aar)]xr,~+ (6.21)
+0~9 A~#63 +~x~A ~# T,

rn*'= c~3A" r 6~ + ~13 A" ~'~c3r, (6.22)

m3=cx4 63 +ot'4T+ot9 A~P e~p +cx12A~ tcotp,
and
M ~ = [as A ~/~A~ ~+~x6 A~AP'~+o~7 A ~ A ~ ~] xr~+
+[oqoA~a A ~ +alt(A'r Aa~ + A'~ Aar)]er~+
+ cq 2 A ~a 63 + ~'~2 A~ a T, (6.23)

MZ~=o~s A~r tc3r+ot13 ACtr6r,

where the coefficients ~1, ~2 ... ~t 3, ~,,
' tx9
' and ~ 2 are constants.

7. Special eases of the general theory

We consider in this section some special cases of the general nonlinear theory
of Sections 3 (or 4) and 5, as well as that of Section 6, by introducing simplifying
assumptions which are motivated by the existing classical theories of elastic
shells and plates. It may be recalled here that the known complete theories of shells
and plates, both linear and nonlinear, are developed from the equations of elasti-
city theory in 3-space under certain assumptions and by integration of these equa-
tions along the x3-coordinate and this, in general, requires additional approxima-
tions*.
* For a discussion of classical theories of shells and plates, see NA6rmi (1963a).
304 A.E. GP.~N, P.M. NAGHDI& W. L. WAINWRIGHT :

When the director is absent and only curve forces are admitted, i.e., when both
M = L = 0 in (3.4), the resulting theory may be called the membrane theory. The
basic field equations of the membrane theory follow from the results of the general
theory and include, in particular, the kinematical results (2.7) to (2.9), (2.14) and
(2.15), and the equation (3.8) for conservation of mass. Also, in this case, since
M ~ = 0 and M ~ = 0 , it follows that (4.3) is identically satisfied and, by (4.2) and
(4.5), we have m s= 0 and h = 0 , respectively. Equations (4.4) and (4.7) then yield
N'P"=N'~tJ=NtJ~, NS~=0, (7.1)

so that (4.1) and (4.9), the remaining relevant equations of Section 4 reduce to

NP~I~+PFP=pcp, b~NP~+pF3=pc 3, (7.2

p r-q~l~-p(TS+ TS)-p ,~i+ NP ~r/~ B= 0. (7.3)

The constitutive equations of the elastic membrane theory in terms of A are
given by (5.15) and (5.16)1 and A depends only on T a n d %p. F o r a surface which
is initially isotropic with a center of symmetry,
A = A (T, 11, I2), where I t = tr I and I 2 = tr 12 .
It may be noted here that the theory of elastic membranes as given in GREEN &
ADKINS (1960), and developed from the equations of nonlinear elasticity in 3-space,
is slightly more general than the above nonlinear membrane theory.
We now specialize the general theory to the case for which D = A a and d = a a ,
so that*
D~=0, Da = 1, (7.4)
d~ = 0 , d 3 = 1.
Then, from (2.22) and (2.24) and their duals follow the results
Ap~= - B p ~ , Aa~=0,
(7.5)

and by (2.21), (2.22), (2.9) and (5.6), we have

w = a s = -(Vs, ~ + b~ vp) a ~, (7.6)
rp ~= - (bp ~ - Ba ~), x s , = 0. (7.7)
In view of (7.4)2, (4.3)t is identically satisfied and (4.3)2 gives
M'=0. (7.8)
By (7.8), (7.4)2 and (4.5), we also have h = 0 so that (3.27)holds**. Thus, under the
assumption (7.4), the kinematical results are modified only by (7.6) and (7.7),
* We can also examine a more general case in which only the initial value of the director is
along the normal to the undeformed surface, i.e., when D----A3 only, but we do not consider
this here.
** The vanishing of ~rl and/~, in the case of an elastic Cosserat surface (Section 5) follow
without any special assumption such as (7.4). Here, the results ~r=----0, h----0, as well as others
between (7.4) and (7.20) are deduced directly from the general theory of Sections 3 and 4 by
specialization and without reference to constitutive equations.
 Theory of Cosserat Surfaces 305

and the basic equations of Section 4 are given by (4.1) and (4.2) while equations
(4.4), (4.7) and (4.9) become
N a ~= m ~+ M 3 ~ b~, (7.9)

N'*a= N' a~=Na~ + M ~ bar, (7.10)

and
pr-q'l~-p(TS+TS)-p,4+N'P'tl~p+MP*~p=O, (7.11)

where, by (7.7), k~p = - / ~ p . If we combine (7.9) with (4.2)1, we obtain

N3P = MB~I~+ P Lfl. (7.12)

Then, using (7.12), we have from (4.1)1 and (4.1)2, respectively,

NP ~1~- bzaMZ ~1~+ P Fp = P ~ , (7.13)

and
MP~t~p+ b ~ NP" + p Fa=pd a, (7.14)
where we have set
- p=d + be,L
(7.15)
p ~a = p c a _ ( p Ep)tp.
Hence, instead of the equations of motion (4.1) and (4.2), we have (7.12), (7.13)
and (7.14), as well as the three equations given by (4.2) 2 and (7.9) or equivalently

rtl a = M 3 ~1~q- b~p M a ' + p ~a, (7.16)

and
m ~ = N a ' - M a~ b~. (7.17)

It is convenient now to decompose N "p and M =p into their respective symmetric

and anti-symmetric parts. Thus,
N~P=N(~) + N t~p], N('P)=89 + NP~), Nt~P]=89 (7.18)
and with a similar notation for M t~p) and M t'a]. As is evident from the energy
equation (7.11), since x a , = 0 , ~ i = 0 and tGp is now a symmetric tensor, under the
assumption (7.4), there are no constitutive equations for M 3a, m i, M t~p] and these,
as well as ~ 3 , will remain indeterminate.
In the case of an elastic Cosserat surface, in addition to (7.8), we also have
~ a = 0 so that (5.11) holds. Moreover, under the assumption (7.4), the constitutive
equations (5.16) for an elastic Cosserat surface reduce to*

N,~=p aA M(~p)= p aA (7.19)

~e~tJ ' dx~a 9
In order to provide a determinate theory, we now assume that
Mt~P]=0 (7.20)
9 The tensors e~p and x~ in (7.19) are understood to stand for 89 e#~) and 89 r#~),
respectively.
306 A.E. GREEN, P. M. NAGI-.IDI& W. L. WAINWRIGHT:

and then N ~p is determined from (7.10) and (7.19)1. We also note here that by
(7.10) and (7.20), the anti-symmetric part of N "p may be determined through

Nr'P]= 89{ b ~ M (~'>- bl Mix P)}. (7.21)

If we introduce (7.18) and (7.21) into (7.13) and (7.14), then these equations may
also be put in the form*

NiPs) Ip-~t.1 rb,, 314+a-rba~ MCP~)]I;,-b~ (7.22)

b~p N (" p~+ M (~p>I~a + P (F3 - ~3) ---_0. (7.23)

The system of equations (2.7), (2.14), (5.5), (7.5), (7.7), (7.12), (7.20) to (7.23)
together with (7.11), (5.1) and (5.2) form a determinate system for N ~p, M ~a and
N 3~. The five unknowns M 3~, m ~, m a remain indeterminate; however, they are
related by the three equations (7.16) and (7.17). We complete our present discus-
sion of an elastic Cosserat surface, by observing that under the assumption (7.4)
and by (5.1), the free energy A assumes the form
A = A ( T , e,,p, x~,~, B~,p). (7.24)

When the surface is initially isotropic with a center of symmetry, following the
discussion concerning material symmetry at the end of Section 5, A may be ex-
pressed as a function of T and the joint invariants

tr I , tr J, tr L
t r I 2, tr j2, tr L2,
(7.25)
trIJ, trlL, trJL,
tr I J L ,

where the matrices I and J are defined as in (5.23) and (5.21), and the 2 x 2
symmetric matrix L is defined by
L = B~ = A ~~By p. (7.26)
If the free energy A does not depend explicitly on B,p, then the invariants corre-
sponding to those in (7.25) may also be obtained as a special case of the joint
invariants listed in (5.24).
The above special theory, deduced under the assumption (7.4), may be regarded
as comparable to the classical theory of shells founded under the so-called Kirch-
hoff-Love hypothesis. Previously, a nonlinear theory of shells under the Kirchhoff-
Love hypothesis, using the equations of nonlinear elasticity in 3-space, was given
by NAGHDX& NORDGREN(1963) and their results may be brought into correspond-
ence with those given here between (7.4) and (7.24), provided that the components
M "p of the director force M are identified with the so-called "stress-couple"
resultants.
* This form of the equations of motion is equivalent to those obtained previously by
NAGnD1(1963b, 1965)in another context. It may be of interest to note that Eq. (7.23) holds even
without the use of (7.20), since MP~ I~# = MP~ IIJ~"
 Theory of Cosserat Surfaces 307

The linearized theory corresponding to the nonlinear theory for which (7.4)
holds may be obtained directly as a special case of the infinitesimal theory of
Section 6. Thus, with
D -A3, d=a3, (7.27)
it follows from (6.1) to (6.3) that
~*=p, * =/L, 61=0, (7.28)
W=]~ct, w3=O.

In view of (7.27) and (7.28), equations (6.8) and (6.9) simplify and may be written
as*
e~,:,= e:, ~= ~1 ( U~l~+ U~l~)-- B~ ~,u 3 ,
A ;~ 2 A
~c~,:, = ~cr ~,= - u 3 I~~,- Bvl~, u ; , - B~ Uzl,,- B,, u;,I~+ B,, a By u 3, (7.29)
Kaa=0,
and
e~ ~= 89(V~lT+ V~l~)- B~ ~ v3 ,
2 2 2 2
(7.30)
~---- -- v31~~-B~l~ v x - B~ Val~- B~ Vzl~-{- B~ a Br v3 ,

where in (7.29) and (7.30), and all equations that follow in the remainder of this
section, covariant differentiation is understood to be with respect to A~p. By an
argument entirely analogous to that discussed for the nonlinear theory subject to
the assumption (7.4), we can show that M~=0, h = 0 and hence (6.12) holds.
Also, the energy equation (6.11) now becomes

por-Q'i,-po(TS+ ~i'S)-poft+N'~P e, a + M ~ ' a ) x , a = O . (7.31)

For the linearized theory of an elastic Cosserat surface under consideration,

M a = 0 and again, in order to obtain a determinate theory, we set Mt~a]=0 and
conclude that N [~p] is given by (7.21) with b~a replaced by B~B. Similarly, the rele-
vant equations of motion are (7.22) and (7.23) except that b~p must be replaced
by B~a. The unknowns M 3a, m a, m 3 will remain indeterminate, but they are
related by equations corresponding to (7.16) and (7.17) with b~a replaced by
B~p. The free energy Po A, in this case, is a function of T, B~a, e~a and K~a given
by (7.29) and the constitutive equations (6.13) reduce to

OA Mt, t~)=po OA (7.32)

N'~P=P~ deep ' cgtGp"

The stress N ~a is obtained from (7.10) with b~ replaced by B~. We also note that
for a surface which initially is isotropic with a center of symmetry, Po A and the
explicit forms of (7.32) may be obtained from (6.20) to (6.23) by setting a3, a~,
as, c%, a t 2 and ax 3 equal to zero and since tGp=Ka~ and Mt~a]=0, without loss
* It is perhaps interesting to note that these expressions are the same as those used previously
by NAGFmI(1963b, 1965) and GREEN& NAGHDI(1965) in entirely different developmentsof the
linear theory of elastic shells.
308 A.E. GREEN, P. M. NAGHDI& W. L. WAINW~Wr : Theory of Cosserat Surfaces

in generality, we m a y also p u t ~6 = ~7. F u r t h e r d e v e l o p m e n t of the a b o v e linear

t h e o r y is possible, b u t we d o n o t p u r s u e this further. Results ( 6 . 2 0 ) - ( 6 . 2 3 ) h o l d
o n l y when A does n o t d e p e n d explicitly on B ~ . M o r e generally, A is a q u a d r a t i c
function of e~a, r~p, T a n d a function of B~p. This function can be o b t a i n e d f r o m
the invariants in (7.25) when the surface is isotropic with a centre of symmetry.

Acknowledgements. The results reported here were obtained in the course of research supported
by the U.S. Office of Naval Research under Contract Nonr 222(69) with the University of
California, Berkeley (U.C.B.). Most of the results in an earlier form were obtained during 1964
when all authors were at Berkeley and one of us (AEG) held a visiting professorship at U.C.B.

References
ADKr~S, J.E., Arch. Rational Mech. Anal. 5, 263 (1960).
COLEMAN,B.D., & W. NOEL, Arch. Rational Mech. Anal. 13, 167 (1963).
COSSERAT,E., & F., C. r. acad. sci. Paris 146, 169 (1908).
COSSERAT,E., & F., Thdorie des Corps Deformables. Pads Hermann 1909.
Dtn-mM, P., Ann. ]~cole Norm. 10, 187 (1893).
E~CKSEN, J.L., Trans. Soc. Rheol. 5, 23 (1961).
ERICKSEN,J.L., & C. TRUESDEEE,Arch. Rational Mech. Anal. 1, 295 (1958).
GREEN, A.E., & W. ZERNA,Theoretical Elasticity. Oxford: Clarendon Press 1954.
GREEN, A.E., & J. E. ADKINS,Large Elastic Deformations. Oxford: Clarendon Press 1960.
GREEN, A.E., & R.S. RrvLr~, ZAMP 15, 290 (1964a).
GREEN, A.E., & R.S. RtVLIN,Arch. Rational Mech. Anal. 17, 113 (1964b).
GREEN, A.E., NAGItDI,P.M., & R.S. RlvLra, Intern. J. Engng. Sci. 2, 611 (1965).
GREEN, A.E., & P.M. NAGHDI,Quart. J. Mech. Appl. Math. (1965)(to appear).
NAGHDI, P.M., "Foundations of Elastic Shell Theory" in Progress in Solid Mechanics, vol. 4
(edited by I.N. SNEI)DON& R. HILL), 1. North-Holland 1963a.
NAGHO[, P.M., Intern. J. Engng. Sci. 1, 509 (1963b); 2, 269 (1964).
NAGHDI,P.M., Proc. l l t h Intern. Congr. Appl. Mech. (Munich 1964). Berlin-Heidelberg-
NewYork: Springer 1965 (to appear).
NAGHDI,P.M., & R.P. NORDGREN, Quart. Appl. Math. 21, 49 (1963).
RrVLIN, R.S., J. Rational Mech. Anal. 4, 681 (1955).
SERnn~, H., J. Math. Phys. 4, 838 (1963).
SMITH, G.F., & R.S. RIVLIN,Quart. Appl. Math. 15, 308 (1957).
ToUPlN, R.A., Arch. Rational Mech. Anal. 17, 85 (1964).
TRUESDELL,C., & R.A. Totn'~, The Classical Field Theories. In: Handbuch der Physik, Bd.III/1
(edited by S. FLi3GGE),226. Berlin-G6ttingen-Heidelberg: Springer 1960.
WIrCEraAN,A.S., & A.C. PtrrdN, Arch. Rational Mech. Anal. 17, 184 (1964).

The University
Newcastle upon Tyne
University of California
Berkeley
University of Colorado
Boulder
(Received July 14, 1965)