0% found this document useful (0 votes)
17 views6 pages

wild2006rmjm -- The fundamental theorem of projective geometry for an arbitrary length two module --

The document presents a theorem regarding the fundamental theorem of projective geometry for an arbitrary R-module of length two, establishing that every permutation of length one submodules is induced by an isomorphism if and only if the number of such submodules is 3 or 4. It discusses the implications of this theorem in the context of lattice isomorphisms and provides examples and proofs related to the structure of R-modules and their endomorphism rings. The author acknowledges contributions from Peter Pálfy and references various works related to the topic.

Uploaded by

2775402990
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
17 views6 pages

wild2006rmjm -- The fundamental theorem of projective geometry for an arbitrary length two module --

The document presents a theorem regarding the fundamental theorem of projective geometry for an arbitrary R-module of length two, establishing that every permutation of length one submodules is induced by an isomorphism if and only if the number of such submodules is 3 or 4. It discusses the implications of this theorem in the context of lattice isomorphisms and provides examples and proofs related to the structure of R-modules and their endomorphism rings. The author acknowledges contributions from Peter Pálfy and references various works related to the topic.

Uploaded by

2775402990
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

ROCKY MOUNTAIN

JOURNAL OF MATHEMATICS
Volume 36, Number 6, 2006

THE FUNDAMENTAL THEOREM OF


PROJECTIVE GEOMETRY FOR
AN ARBITRARY LENGTH TWO MODULE

MARCEL WILD

ABSTRACT. Let V be an arbitrary R-module of length 2


with n ≥ 3 submodules of length 1. Then every permutation
of the length 1 submodules is induced by an isomorphism

V → V if and only if n = 3 or 4.

1. Introduction. In this note all rings R have an identity and all R-


modules V are unital. We write L(V ) for the lattice of all submodules

of V . Every module isomorphism f : V → V clearly induces a lattice

isomorphism F : L(V ) → L(V ) where F (W ) := f (W ). Call V linearly

induced if conversely for each lattice isomorphism F : L(V ) → L(V )

there is a module isomorphism f : V → V such that F (W ) = f (W )
for all W ∈ L(V ). A variant of the fundamental theorem of projective
geometry can be phrased as follows:

Theorem 1 [1, p. 62]. Let K be a division ring such that every


automorphism is inner. Then each K-vector space of finite dimension
≥ 3 is linearly induced.

(In the classic fundamental theorem of projective geometry [1, p.


44] there is no restriction on the division ring but then the lattice

isomorphism F : L(V ) → L(V ) perhaps is only induced by a semilinear
bijection f : V → V . We do not wish to bother about semilinearity in
this article.)
In particular, in Theorem 1 division rings without proper automor-
phisms, such as K = R, comply. The lattice L(V ) of subspaces of
the K-vector space V is often called the projective geometry associated
with K. The dimension 1, 2, 3 subspaces are the points, lines, planes of

the projective geometry. Lattice isomorphisms L(V ) → L(V ) are called

Received by the editors on January 23, 2004, and in revised form on June 1,
2004.

c
Copyright 2006 Rocky Mountain Mathematics Consortium

2075
2076 M. WILD

projectivities in [1] (but in modern texts the meaning of collineation


and projectivity may be switched).
Theorem 1 fails for two-dimensional vector spaces. In this case von
Staudt type theorems take over. They essentially assert that it works
for permutations of the points that preserve cross ratios as defined in
[1, p. 71]. This condition is necessary in the sense that each K-linear

isomorphism K 2 → K 2 induces a cross ratio preserving permutation
L(K 2 ) → L(K 2 ). Here is an example of a von Staudt type theorem:

Theorem 2 [1, p. 87]. The division ring K is commutative if and


only if the identity is the only permutation of the K-projective line
which preserves cross ratios and possesses three fixed points.

As a consequence, note the following: Suppose K is commutative and


the K-projective line L(V ) has n ≥ 5 points (this amounts to |K| ≥ 4).
Then obviously there is a non-identity permutation F of L(V ) that

fixes three points. By Theorem 2 each linear isomorphism f : V → V
fixing three proper subspaces must be the identity. Therefore V cannot
be linearly induced.
Theorem 1 has been generalized in many ways in order to accommo-
date rings R other than K. For instance, the m-dimensional “projective
geometry” associated with a ring R is often defined as the set of all
direct summands of the module Rm+1 . Theorem 2 has been general-
ized to a lesser extent; usually the concept of cross ratio is somehow
adapted to the relevant ring.
Rather than looking at R-modules R2 , which have length bigger than
two unless R is a field, in this paper we let V be any R-module of length
two. Also, as opposed to the usual generalizations of Theorem 2, we
do not focus on special types of permutations F : L(V ) → L(V ), but
focus on those numbers n := |L(V )| − 2 for which Theorem 1 holds
unconditionally.
So let V be an arbitrary length two module. Then the lattice
L(V ) is isomorphic to the length two modular lattice Mn completely
characterized by the number n = n(V ) of atoms. When n is infinite
we write n = ∞ rather than distinguishing between infinite cardinals.

For n(V ) = 1 the only lattice isomorphism L(V ) → L(V ) is the

identity, which is induced by the identity V → V . Let n(V ) = 2,
THEOREM FOR ARBITRARY LENGTH TWO MODULE 2077

so L(V ) = {0, U1 , U2 , V }. Such a V is necessarily cyclic. The


only nontrivial lattice isomorphism F switches U1 and U2 . Clearly

F is induced by a module isomorphism f : V → V if and only if
U1  U2 . Things get more interesting for n(V ) ≥ 3; here is our (not so
fundamental) theorem of projective lines.

Theorem 3. Let R be an arbitrary ring and let V be an R-module


of length two with n(V ) ≥ 3. Then V is linearly induced if and only if
n(V ) ≤ 4.

Proof. The following fact will be crucial.


(1) Let L(V ) = {0, U1 , U2 , U3 , . . . , Un , V }. Then the map T which
sends φ to {u + φ(u) | u ∈ U1 } is a bijection between the set of

isomorphisms φ : U1 → U2 and the set {U3 , . . . , Un }.
We omit the straightforward verification; notice that for given Ui (3 ≤
i ≤ n) the φ with T (φ) = Ui is the map which sends u ∈ U1 to the
unique u ∈ U2 with u + u ∈ Ui . It follows from (1) that

(2) |Aut (U1 )| = |{φ : U1 → U2 }| = n − 2 (note ∞ − 2 = ∞).

By Schur’s lemma End (U1 ) = Aut (U1 )∪{0} is a division ring. Thus, if
|End (U1 )| < ∞, then Wedderburn’s theorem yields End (U1 )  GF (q)
where the latter is the Galois field of cardinality q (= power of a prime).
Summarizing, either n = ∞ or q − 1 = n − 2. So n = q + 1.

(3) |Aut (V )| = n(n − 1)(n − 2)2



Indeed, the module automorphisms f : V → V are exactly the maps
∼ ∼
f1 ⊕f2 : U1 ⊕U2 → Ui ⊕Uj , i = j, where f1 : U1 → Ui and f2 : U2 → Uj
are module isomorphisms. The number of pairs (i, j) is n(n − 1) and
by (2) the number of f1 ’s, respectively f2 ’s, is n − 2.
One checks that n(n − 1)(n − 2)2 < n! for all n ≥ 6. This includes
infinite cardinals n since then n(n − 1)(n − 2)2 = n < 2n ≤ n!.
Thus, for n ≥ 6, the mere cardinality argument (3) guarantees lattice

automorphisms F : L(V ) → L(V ) which are not induced by any
f ∈ Aut (V ). Can one explicitly pinpoint such an F ? Provided the
division ring End (U1 ) has a nontrivial center and n ≥ 5 we shall
manage to do so. In particular this will settle the case n = 5 since
2078 M. WILD


then End (U1 )  GF (4). So suppose n ≥ 5 and let F : L(V ) → L(V )
be any fixed lattice isomorphism such that F (Ui ) = Ui , 1 ≤ i ≤ 3,
F (Ui ) = Ui , 4 ≤ i ≤ n. Such an F exists because n ≥ 5. (Recall
that for the very special case of a two-dimensional vector space V over
a commutative division ring this F does the job due to Theorem 2.)

Suppose f : V → V is a module isomorphism that induces F . We want
to derive a contradiction. According to (1) we have

(4) U3 = {u + φ(u) | u ∈ U1 }

for some unique isomorphism φ : U1 → U2 . Because by assumption
f (U1 ) = U1 , we have

(5) U3 = {f (u) + φ(f (u)) | u ∈ U1 }.

Using (4), one derives

(6) U3 = f (U3 ) = {f (u) + f (φ(u)) | u ∈ U1 }.


∼ ∼
Because of f ◦ φ : U1 → U2 → U2 , both (5) and (6) are representations
of U3 of type (1). Hence

(7) φ ◦ f (u) = f ◦ φ(u) (u ∈ U1 )

by the uniqueness of this representation. By assumption End (U1 )


contains a central element ψ = 0, 1. By (1) we have

(8) {u + φ ◦ ψ(u) | u ∈ U1 } = Uj

for some j ∈ {4, . . . , n}. Now

(8)
f (Uj ) = {f (u) + f ◦ φ ◦ ψ(u) | u ∈ U1 }
(7)
= {f (u) + φ ◦ f ◦ ψ(u) | u ∈ U1 }
(8)
= {f (u) + φ ◦ ψ ◦ f (u) | u ∈ U1 } = Uj

which contradicts F (Uj ) = Uj .


Now we show that V is linearly induced when n = 3 or 4. The
case n = 3 being analogous, we only do n = 4, so L(V ) =
THEOREM FOR ARBITRARY LENGTH TWO MODULE 2079

{0, U1 , U2 , U3 , U4 , V }. Then End (U1 )  GF (3), so Aut (U1 ) = {id, ψ}.


Analogous to (4) and (8) above we have

U3 = {u + φ(u)| u ∈ U1 }
U4 = {u + φ ◦ ψ(u)| u ∈ U1 }


where φ : U1 → U2 is a unique isomorphism. Since the symmetric
group of degree 4 is generated by 2-cycles, it suffices to show that the

lattice isomorphisms L(V ) → L(V ) determined by the permutation

U1 −→ U1 , U2 −→ U2 , U3 −→ U4 −→ U3



is linearly induced. Put f := ψ ⊕ id : U1 ⊕ U2 → U1 ⊕ U2 . Using
ψ ◦ ψ = id we get

f (U3 ) = {ψ(u) + φ(u) | u ∈ U1 } = {ψ(u) + φ ◦ ψ(ψ(u)) | u ∈ U1 } = U4 .



Because f induces a bijection L(V ) → L(V ), this forces f (U4 ) = U3 .

What else can be said about an arbitrary length two module V ? As to


its endomorphism ring, if n(V ) ≥ 3 then V  U1 ⊕ U1 , and so End (V )
is isomorphic to the ring M2 (End (U1 )) of 2 × 2 matrices with entries
from the division ring End (U1 ). In particular, when n = n(V ) < ∞,
then n = q + 1 and End (V )  M2 (GF (q)). The reader may check
that the number of invertible 2 × 2 matrices over GF (q) is indeed
n(n − 1)(n − 2)2 in accordance with (3). What can be said about
the Abelian group (V, +)? Not much, but if n(V ) < ∞ and V = R V
is noncyclic, then (V, +) turns out to be (GF (q)2 , +). This does not
imply that R  GF (q). Whether V is cyclic or not, n(V ) < ∞ always
implies that n = q +1 for some prime power q. Now 7 is the first integer
≥ 3 not of type q + 1, and so there cannot be a length two module V
with n(V ) = 7.
This relates to a major unsolved problem of universal algebra: Which
finite lattices occur as congruence lattices of a finite algebra? A
breakthrough was made in [3] where the problem is reduced to intervals
in subgroup lattices of finite groups. In particular, which lattices Mn
occur as such an interval? It has, e.g., been shown in [2] that the
2080 M. WILD

answer is affirmative for n = q + 2. Thus, n = 7 works, but not with


modules.
When is a module V of length at least three linearly induced?
Theorem 3 suggests that this is unlikely, unless either L(V ) has no
interval sublattice M5 or the identity is the only lattice isomorphism
of L(V ). As mentioned in the introduction, a module isomorphism
∼ ∼
V → V trivially induces a lattice isomorphism L(V ) → L(V ). But
what if f : V → V is merely a homogeneous bijection, i.e., satisfying
f (λx) = λf (x) but not necessarily f (x + y) = f (x) + f (y)? Call V
hom-proj if such a f nevertheless always induces a lattice isomorphism

L(V ) → L(V ). It is easy to see that every length two module is hom-
proj, but many others are as well [4].

Acknowledgment. I am grateful to Peter Pálfy for helpful com-


ments.

REFERENCES

1. R. Baer, Linear algebra and projective geometry, Academic Press, New York,
1952.
2. A. Lucchini, Representations of certain lattices as intervals in subgroup
lattices, J. Algebra 164 (1994), 85 90.
3. P.P. Pálfy and P. Pudlák, Congruence lattices of finite algebras and intervals
in subgroup lattices of finite groups, Algebra Univ. 11 (1980), 22 27.
4. M. Wild, Homogeneous bijections that induce automorphisms of the submodule
lattice, Comm. Algebra 33 (2005), 2649 2661.

Department of Mathematics, University of Stellenbosch, 7602 Matieland,


South Africa
E-mail address: [email protected]

You might also like