Lectures On Sl2cmodules Illustrated Voldymyr

Mazorchuk download

https://ebookbell.com/product/lectures-on-sl2cmodules-
illustrated-voldymyr-mazorchuk-33122670

Explore and download more ebooks at ebookbell.com

Here are some recommended products that we believe you will be
interested in. You can click the link to download.

Lectures In Algebraic Combinatorics Youngs Construction Seminormal

Representations Sl2 Representations Heaps Basics On Finite Fields 1st
Ed Adriano M Garsia

https://ebookbell.com/product/lectures-in-algebraic-combinatorics-
youngs-construction-seminormal-representations-sl2-representations-
heaps-basics-on-finite-fields-1st-ed-adriano-m-garsia-22496990

Lectures On Numerical Radius Inequalities Pintu Bhunia Silvestru Sever

Dragomir

https://ebookbell.com/product/lectures-on-numerical-radius-
inequalities-pintu-bhunia-silvestru-sever-dragomir-47277802

Lectures On Quantum Mechanics With Problems Exercises And Solutions

3rd Edition Jeanlouis Basdevant

https://ebookbell.com/product/lectures-on-quantum-mechanics-with-
problems-exercises-and-solutions-3rd-edition-jeanlouis-
basdevant-47661760

Lectures On Quantum Mechanics And Attractors Alexander Komech

https://ebookbell.com/product/lectures-on-quantum-mechanics-and-
attractors-alexander-komech-48710812
Lectures On Differential Geometry Richard Schoen Shingtung Yau

https://ebookbell.com/product/lectures-on-differential-geometry-
richard-schoen-shingtung-yau-48750644

Lectures On Intelligent Systems 1st Edition Leonardo Vanneschi

https://ebookbell.com/product/lectures-on-intelligent-systems-1st-
edition-leonardo-vanneschi-48980358

Lectures On Phase Field 1st Edition Ingo Steinbach Hesham Salama

https://ebookbell.com/product/lectures-on-phase-field-1st-edition-
ingo-steinbach-hesham-salama-49416220

Lectures On The Philosophy Of Art The Hotho Transcript Of The 1823

Berlin Lectures 1st Edition Georg Wilhelm Friedrich Hegel Translated
By Robert F Brown

https://ebookbell.com/product/lectures-on-the-philosophy-of-art-the-
hotho-transcript-of-the-1823-berlin-lectures-1st-edition-georg-
wilhelm-friedrich-hegel-translated-by-robert-f-brown-49422546

Lectures On Field Theory And Topology Daniel S Freed

https://ebookbell.com/product/lectures-on-field-theory-and-topology-
daniel-s-freed-49433136
 Lectures on
sC2(CC)-modules

Volodymyr Mazorchuk
Uppsala University, Sweden

~f)r---~~~~-Im_pe_ria_lC_oll_eg_eP_re_ss_
Published by
Imperial College Press
57 Shelton Street
Covent Garden
London WC2H 9HE

Distributed by
World Scientific Publishing Co. Pte. Ltd.
5 Toh Tuck Link, Singapore 596224
USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601
UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.

LECTURES ON SL_2(C)-MODULES
Copyright© 2010 by Imperial College Press
All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means,
electronic or mechanical, including photocopying, recording or any information storage and retrieval
system now known or to be invented, without written permission from the Publisher.

For photocopying of material in this volume, please pay a copying fee through the Copyright
Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to
photocopy is not required from the publisher.

ISBN-13 978-1-84816-517-5
ISBN-10 1-84816-517-X

Printed in Singapore by B & JO Enterprise

 Preface

The representation theory of Lie algebras is an important and intensively

studied area of modern mathematics with applications in practically all
major areas of mathematics and physics. There are several textbooks which
specialize in different aspects of the representation theory of Lie algebras
and its applications, but the usual topics covered in such books are finite-
dimensional, highest weight or Harish-Chandra modules.
The smallest simple Lie algebra s[2 differs in many aspects from all
other semi-simple Lie algebras. One could, for example, mention that s[2
is the only semi-simple Lie algebra for which all simple (not necessarily
finite-dimensional) modules are in some sense understood. The algebra
s[ 2 is generated by only two elements and hence is an invaluable source of
computable examples. Moreover, in many cases the ideas which one gets
from working with s[ 2 generalize relatively easily to other Lie algebras with
a minimum of extra knowledge required.
The aim of these lecture notes is to give a relatively short introduction
to the representation theory of Lie algebras, based on the Lie algebra s[ 2 ,
with a special emphasis on explicit examples. Using this Lie algebra, we
can examine and describe many more aspects of the representation theory
of Lie algebras than are covered in standard textbooks.
The notes start with two conventional introductory chapters on finite-
dimensional modules and the universal enveloping algebra. The third chap-
ter moves on to the study of weight modules, including a complete classifica-
tion and explicit construction of all weight modules and a description of the
category of all weight modules with finite-dimensional weight spaces, via
quiver algebras. This is followed by a description and study of the primitive
spectrum of the universal enveloping algebra and its primitive quotients.
The next step is a relatively complete description of the Bernstein-Gelfand-

v
vi Lectures on sl2(C)-modules

Gelfand category 0 and its properties. The two last chapters contain a
description of all simple s[ 2 -modules and various categorifications of simple
finite-dimensional modules. The material presented in the last chapter is
based on papers which were published in the last two years.
The notes are primarily directed towards postgraduate students inter-
ested in learning the basics of the representation theory of Lie algebras. I
hope that these notes could serve as a textbook for both lecture courses
and reading courses on this subject. Originally, they were written and used
for reading courses which I gave in Uppsala in 2008.
The prerequisites for understanding these notes depend on the chapter.
For the first two chapters, one needs only some basic knowledge in linear
algebra and rings and modules. For the next two chapters, it is assumed
that the reader is familiar with the basics of the representation theory of
finite-dimensional associative algebras and basic homological algebra. The
last three chapters also require some basic experience with category theory.
At the end of each chapter are comments including some historical back-
ground, brief descriptions of more advanced results, and references to some
original papers. I tried to present these comments to the best of my knowl-
edge and I would like to apologize in advance for any unforeseen errors or
omissions.
There are numerous exercises in the main text and at the end of each
chapter. The exercises in the main text are usually relatively straightfor-
ward and required to understand the material. It is strongly recommended
that the reader at least looks through them. Answers and hints are supplied
at the end of the notes.
I would like to thank Ekaterina Orekhova and Valentina Chapovalova
for their corrections and comments on the earlier version of the manuscript.

Uppsala,. August 2009 Volodymyr Mazorchuk

 Contents

Preface v

l. Finite-dimensional modules 1
1.1 The Lie algebra s[ 2 and s[ 2 -modules 1
1.2 Classification of simple finite-dimensional modules 5
1.3 Semi-simplicity of finite-dimensional modules . 10
1.4 Tensor products of finite-dimensional modules . 15
1.5 Unitarizability of finite-dimensional modules. 17
1.6 Bilinear forms on tensor products . 21
1.7 Addenda and comments 23
1.8 Additional exercises . . . . 26

2. The universal enveloping algebra 33

2.1 Construction and the universal property . . . . . . . . 33
2.2 Poincare-Birkhoff-Witt Theorem . . . . . . . . . . . . 37
2.3 Filtration on U(g) and the associated graded algebra . 41
2.4 Centralizer of h and center of U(sC 2 ) 44
2.5 Harish-Chandra homomorphism. 48
2.6 Noetherian property . . 50
2. 7 Addenda and comments 52
2.8 Additional exercises 56

3. Weight modules 59
3.1 Weights and weight modules 59
3.2 Verma modules 62
3.3 Dense modules . . . . . 68

vii
viii Lectures on s[2(C)-modules

3.4 Classification of simple weight modules . 72

3.5 Coherent families . 75
3.6 Category of all weight modules with finite-dimensional
weight spaces 82
3.7 Structure of wl;,T in the case of one simple object 85
3.8 Structure of wl;,T in the case of two simple objects 86
3.9 Structure of wl;,T in the case of three simple objects 90
3.10 Tensoring with a finite-dimensional module 92
3.11 Duality 101
3.12 Addenda and comments 103
3.13 Additional exercises 108

4. The primitive spectrum 115

4.1 Annihilators of Verma modules 115
4.2 Simple modules and central characters 117
4.3 Classification of primitive ideals . 119
4.4 Primitive quotients . 121
4.5 Centralizers of elements in primitive quotients . 124
4.6 Addenda and comments 128
4.7 Additional exercises 132

5. Category 0 135
5.1 Definition and basic properties 135
5.2 Projective modules . 139
5.3 Blocks via quiver and relation . 144
5.4 Structure of a highest weight category 149
5.5 Grading 153
5.6 Homological properties . 160
5.7 Category of bounded linear complexes of
projective graded D-modules . 164
5.8 Projective functors on 0 0 170
5.9 Addenda and comments 178
5.10 Additional exercises 189

6. Description of all simple modules 195

6.1 Weight and nonweight modules 195
6.2 Embedding into a Euclidean algebra 197
6.3 Description of simple nonweight modules . 201
 Contents ix

6.4 Finite-dimensionality of kernels and cokernels 204

6.5 Finite-dimensionality of extensions 210
6.6 Addenda and comments 213
6. 7 Additional exercises . . . . . . . . 215

7. Categorification of simple finite-dimensional modules 219

7.1 Decategorification and categorification 219
7.2 Nai:ve categorification of y(n) . . . . . . . . . . 221
7.3 Weak categorification of y(n) . . . . . . . . . . 226
7.4 Categorification of y(n) via coinvariant algebras 232
7.5 Addenda and comments 234
7.6 Additional exercises . . . . . . . . 237

Appendix A Answers and hints to exercises 241

Bibliography 249
Index of Notation 255
Index 259
 Chapter 1

Finite-dimensional modules

1.1 The Lie algebra sb and sb-modules

In what follows we will always work over the field <C of complex numbers.
Unless stated otherwise, all vector spaces, tensor products and spaces of
homomorphisms are taken over <C. As usual, we denote by Z, Q and JR
the sets of integer, rational and real numbers, respectively. We also denote
by N the set of all positive integers and by No the set of all non-negative
integers.
The Lie algebra g = s[2 = sl2 (<C) consists of the vector space

s[2 = {( ~ ~) : a, b, c, d E <C; a + d = 0}
of all complex 2 x 2 matrices with zero trace and the binary bilinear oper-
ation [X, Y] = XY - Y X of taking the commutant of two matrices on this
vector space. Here XY denotes the usual (associative) product of the ma-
trices X = ( xn x 12 ) and Y = ( Yn Y 12 ) given by the following formula:
X21 X22 Y21 Y22

(
xu x12) (Yu Y12) = (X11Y11 + x12Y21 XuY12 + x12Y22).
X21 X22 Y21 Y22 X21Y11 + X22Y21 X21Y12 + X22Y22
To simplify the notation we will usually denote the Lie algebra s[ 2 simply
by g.

Exercise 1.1. Prove that for any two square complex matrices X and Y
of the same size, the matrix [X, Y] has zero trace.

From Exercise 1.1 it follows that the operation [·, ·] on g is well-defined.

The fact that g is a Lie algebra means that it has the following properties:
2 Lectures on .sl2(C)-modules

Lemma 1.2.

(i) For any X E g we have [X, X] = 0.

(ii) For any X, Y, Z E g we have [X, [Y, Z]] + [Y, [Z, X]] + [Z, [X, Y]] = 0.
Proof. We have [X, X] =XX -XX= 0, proving the statement (i). The
statement (ii) is proved by the following computation:
[X, [Y, Z]] + [Y, [Z, X]] + [Z, [X, Y]]
= X(YZ - ZY) - (YZ - ZY)X + Y(ZX - XZ)
- (ZX - XZ)Y + Z(XY - YX) - (XY - YX)Z
=XYZ-XZY-YZX+ZYX+YZX-YXZ
-ZXY+XZY+ZXY-ZYX-XYZ+YXZ
=0. D
Exercise 1.3. Show that the condition in Lemma 1.2(i) is equivalent to
the following condition: [X, Y] = -[Y, X] for all X, YE g.
The condition in Lemma 1.2(ii) is called the Jacobi identity. The asser-
tion of Exercise 1.3 is true over any field of characteristic different from 2
and basically says that the operation [·, ·] is antisymmetric.
From the definition we have that elements of the algebra g are given by
four parameters and one non-trivial linear relation. This means that this
algebra has dimension three. We now fix the following natural or standard
basis of g:

f = (0 0)
10 '
h= (1 0).
0 -1

By a direct calculation one gets the following Cayley table for the oper-
ation [·, ·] in the standard basis:
[·, ·lll e I r I h
e 0 -2e
h
f -h 0 2f
h 2e -2f 0
Another way to notate the essential information from the above Cayley
table (the diagonal of the table is fairly obvious and given by Lemma 1.2(i))
is the following:
[e, f] = ef - fe = h,
[h,e] =he - eh= 2e, (1.1)
[h, f] = hf - fh = -2f.
 Finite-dimensional modules 3

A module over g (or, simply, a g-module) is a vector space V together

with three fixed linear operators E = Ev, F = Fv and H = Hv on V,
which satisfy the right-hand side equalities in (1.1), that is

EF-FE=H, HE-EH=2E, HF-FH=-2F. (1.2)

It is worth noting that the last two relations can be rewritten as follows:

HE= E(H + 2), HF= F(H - 2). (1.3)

Example 1.4. Let V = C and E = F = H = 0. Then all equalities in

(1.2) obviously hold and we get the trivial g-module.

Example 1.5. Let V = C 2 . In the usual way we identify the set of all
linear operators on V with the set of all complex 2 x 2 matrices. Set E = e,
F = f and H = h. All equalities in (1.2) hold because of (1.1) and we get
the natural g-module.

Example 1.6. Take now V = g. Let E denote the linear operator on V

given by [e, - l (that is the linear operator of taking the commutator with
e, the latter standing on the left). Analogously define Fas [f, -land Has
[h, _ J. From Lemma 1. 7 below we have that these linear operators satisfy
(1.2) and we get the adjoint g-module.

Lemma 1. 7. For any X E g we have

[e, [f,X]] - [f, [e,X]] = [h,X],

[h, [e, X]] - [e, [h, X]] = [2e, X] ,
[h,[f,X]]-[f,[h,X]] = [-2f,X].

Proof. The equality [e, [f, X]] - [f, [e, X]] = [h, X] can be rewritten as
follows:

[e, [f, X]] - [f, [e, X]] - [h, X] = 0. (1.4)

Recall that h = [e, f]. Applying now Exercise 1.3 to the inner bracket of
the second summand and the outer bracket of the third summand reduces
the equality (1.4) to the Jacobi identity. Hence the first equality from the
formulation follows from Lemma 1.2(ii). The rest is proved similarly. D
4 Lectures on sl2(C)-modules

Given two g-modules V and W, a homomorphism from V to W (or a

g-homomorphism, or, simply, a morphism) is a linear map <I> : V----+ W that
makes the following diagram commutative for all X E { E, F, H}:
Xv
v----___,._v
~t Xw t~
w w
In other words, the linear map <I> intertwines the actions of e, f and h on
V and W in the following sense:
<I>Ev = Ew<I>, <I>Fv = Fw<I>, <I>Hv = Hw<I>. (1.5)
The set of all homomorphisms from V to W is denoted by Hom 9 (V, W).

Example 1.8. For any two g-modules V and W, the zero linear map from
V to W obviously satisfies (1.5). This is the so-called zero homomorphism.

From Example 1.8 it follows that the set Hom 9 (V, W) is always non-
empty.

Exercise 1.9. Show that Hom 9 (V, W) is closed with respect to the usual
addition of linear maps and multiplication of linear maps by complex num-
bers. In particular, show that the set Hom 9 (V, W) is a vector space.

Example 1.10. For any g-module V the identity map idv on V obviously
satisfies (1.5) (where V = W). This is the so-called identity homomorphism.

An injective homomorphism is called a monomorphism, a surjective

homomorphism is called an epimorphism and a bijective homomorphism is
called an isomorphism. As usual, it only makes sense to study g-modules
up to isomorphism. The fact that two modules V and W are isomorphic is
usually denoted by V ~ W.
Let V be a g-module. A subspace W c Vis called an submodule (or a
g-submodule) of V provided that it is invariant with respect to the action
of Ev, Fv and Hv, that is
Ev W c W, Fv W c W, Hv W c W. (1.6)
For example, the module V always has two obvious submodules; namely,
the zero subspace and the whole space V. Any submodule, different from
 Finite-dimensional modules 5

these two is called a proper submodule. A module which does not have any
proper submodules is called simple. For example, any module of dimension
one is simple.

Exercise 1.11. Show that all g-modules from Examples 1.4, 1.5 and 1.6
are simple.

Exercise 1.12. Let V = C 2 and E = F = H = 0. Show that this defines

on V the structure of a g-module that is not simple.

Exercise 1.13. Let V be a g-module and W a submodule of V. Show that

the quotient space V /W carries the natural structure of a g-module given by
E(v+ W) = E(v)+ W, F(v+ W) = F(v) +Wand H(v+ W) = H(v) + W.
The module V /W is called the quotient or the factor of V by W.

Lemma 1.14. Let V and W be two g-modules and <I> E Hom 9 (V, W). Then

(i) The kernel Ker( <I>) of <I> is a submodule of V.

(ii) The image Im( <I>) of <I> is a submodule of W.

Proof. Let v E Ker(<I>) and XE {E,F,H}. This gives us

<l>(Xv(v)) = <l>Xv(v) (~) Xw<l>(v) = 0,

implying Xv(v) E Ker(<I>). This proves (i). To prove (ii) is left as an
exercise to the reader. D

1.2 Classification of simple finite-dimensional modules

This section contains perhaps the most classical part of the s[2 -represen-
tation theory; namely, a classification of all simple finite-dimensional s[ 2 -
modules. As we will see later, such modules form only a very small family of
simple s[ 2 -modules. A description of all simple s[2 -modules is an ultimate
goal of this book, but that will require much more theory and effort. The
beauty of finite-dimensional modules is in the fact that their description is
absolutely elementary.
Let Vi= 0 be a finite-dimensional g-module. For ..\ E (('. set
V(..\) = {v EV : (H - ..\)kv = 0 for some k EN},
VA = {v E v : H v = AV}
(here, as usual, we identify (('. with multiples of idv ).
6 Lectures on sl2(1C)-modules

As we are working over the algebraically closed field of complex numbers,

from the Jordan Decomposition Theorem we have that
v ~ E}jv(>.). (1.7)
AEIC

Set W = EB VA c V and note that W i=- 0 as H must have at least one

AEIC
eigenvalue and hence at least one non-zero eigenvector in V.

Lemma 1.15. Let >. E CC.

(i) EV(>.) c V(>. + 2) and EVA c V.>-+2·

(ii) FV(>.) c V(>. - 2) and FVA c VA-2·
(iii) HV(>.) c V(>.) and HVA C VA·

Proof. For v E VA we have

H(E(v)) = HEv (l~) EHv + 2Ev =>.Ev+ 2Ev = (>. + 2)E(v),

which implies the second part of the statement (i). Similarly, for v EV(>.)
let k E No be such that (H - >.)kv = 0. Then we have

(H - (>. + 2))k(E(v)) = + 2))k Ev (l~)

(H - (>.
= E(H + 2 - (>. + 2))kv = E(H - >.)kv = 0,
which implies the first part of the statement (i).
The statement (ii) is proved similarly. The statement (iii) is obvious
from the definitions. D

Exercise 1.16. Generalizing the arguments from the proof of Lemma 1.15,
show that for any f(x) E IC[x] one has the equalities f(H)E = Ef(H + 2)
and f(H)F = Ff(H - 2).

From Lemma 1.15 we immediately obtain:

Corollary 1.17. The space Wis a submodule ofV, in particular, we have

W = V if the module V is simple.

If the module V is simple, we can use Corollary 1.17 to improve the

decomposition given by (1.7) as follows:
(1.8)
AEIC
Since V is finite-dimensional, the decomposition (1.8) must be finite in the
sense that only finitely many summands are non-zero. Thus we can fix
 Finite-dimensional modules 7

some µ E C such that Vµ -/= 0 and Vµ+ 2 k = 0 for all k E N. Let v E Vµ

be some non-zero element. As Vµ- 2 k must be zero for some k E N, from
Lemma 1.15(ii) it follows that there exists a minimal n E N such that
Fnv = 0. For i E {1, 2, ... , n - 1} set Vi = Fiv, and also set Vo= v. From
(1.8) it follows that the Vi's are linearly independent. From Lemma 1.15(ii)
we have H vi = (µ - 2i)vi for all i. Let N denote the linear span of all vi's.

Lemma 1.18. We have Ev 0 = 0 and Evi = i(µ - i + l)vi-l for all i E

{1,2, ... ,n-1}.

Proof. That Ev 0 = 0 is obvious. To prove the rest we proceed by induc-

tion on i. For i = 1 we have
(1.2)
= EFvo = FEvo + Hvo = 0 + µvo = µvo,
Ev1
which agrees with our formula. When i > 1 for the induction step we have:
Evi = EFvi-1
(1.2)
= FEvi-1 + Hvi-1
(inductive assumption) (i - 1)(µ - i + 2)Fvi-2 + (µ - 2(i - l))vi-1
i(µ - i + l)vi-1·
This completes the proof. D

Corollary 1.19. N is a submodule of V, in particular, N = V provided

that V is simple.

Proof. That N is invariant with respect to the action of H and F is

obvious. By Lemma 1.18 it is also invariant with respect to the action of
E. The claim follows. D

Lemma 1.20. µ = n - 1.

Proof. From the inductive argument used in the proof of Lemma 1.18 we
get EFvn-1 = n(µ-n + l)vn-1· However, Fvn-l = 0 by our assumptions,
hence n(µ - n + 1) = 0 implying µ = n - 1. D
Assuming that V is simple, let us sum up the information which we now
have about this module. It has the basis {v0 , v1 , ... , Vn-l}, in which the
action of the operators E, F and H can be depicted as follows:
n-5 n-3 n-1
an-1
1'- '. c an-2
1 ·. c an-3 1
r __..,;:... . r __..,;:... . r _____,.._
a3 . a2 '
r _...---.;,._.
~.
. 1', a1 1:_ '.
r _...---.;,._.
0
1:_
r..-----..,,,._
\;::::-- Vn-1 ~ Vn-2 ~ Vn-3 ~ · · · ~ V2 ~ V1 ~Vo
0 1 1 1 1 1 1
(1.9)
8 Lectures on sl2(C)-modules

Here a; = i( n- i). The double arrow represents the action of F, the regular
arrow represents the action of E and the dotted arrow represents the action
of H. The numbers over arrows are coefficients.

Exercise 1.21. Check that for any n E N the picture (1.9) defines on the
formal linear span of {Vo, ... , Vn-d the structure of a g-module. We will
denote this module by y(n).

Now we are ready to formulate the main result of this section.

Theorem 1.22 (Classification of simple finite-dimensional modules).

(i) For every n E N the module y(n) is a simple g-module of dimension
n.
(ii) For any n, m E N we have y(n) ~ y(m) if and only if n = m.
(iii) Let V be a simple finite-dimensional g-module of dimension n. Then
V ~ y(n)_

Proof. That y(n) is a module follows from Exercise 1.21. Let MC y(n)
be a non-zero submodule and v E M, v -=f. 0. From (1.9) we have that
Env = 0, in particular, En M = 0 and hence M must have a non-trivial
intersection with the kernel of E. Again, from (1.9) it follows that the kernel
of E is just the linear span of v 0 and is, in particular, one-dimensional.
Hence M contains v 0 . Applying to v 0 the operator F inductively we get
that M contains all the vi's. Hence M = y(n). This proves the statement
(i).
As dim y(n) = n, the statement (ii) is obvious. The statement (iii)
follows from the analysis leading to the picture (1.9). D

Exercise 1.23. Show that after rescaling the basis {vi} in the following
way: w; = .frv; the picture (1.9) transforms into the following symmetric
form:
1-n 3-n n-5 n-3 n-1

: 1 : .....: 2 n-3 ··· ..: n-2 :··· : n-1 : : O

y ---... . y ____,,,_ _______:..... ·. y-------s... ·. y-------s... ·. y---------...,..
W1 ..<-_____....-_ Wo
\;:::,-- Wn-1 ~ Wn-2 ~ . . .
~
W2
~ -=---
0 n-l n-2 3 2 1
(1.10)

Exercise 1.24. Show that one can rescale the basis {v;} so that in the new
 Finite-dimensional modules 9

basis { wi} the picture (1.9) transforms into the following symmetric form:
1-n 3-n n-5 n-3 n-1

· j, n-l . y n-2 3 . j, 2 . y 1 y 0
A ----;i... A ~ _______,,,_ A _________,._ A _________,._ A ~

\"=::----- Wn-1 ~ Wn-2 ~ · · · ~ W2 ~ W1 ~ Wo

0 1 2 n-3 n-2 n-l
(1.11)

Exercise 1.25. Let V be a simple finite-dimensional g-module which con-

tains a non-zero vector v such that E(v) = 0 and H(v) = (n - l)v. Show
that V ~ y(n).

In the basis { w 0 , w 1 , ... , Wn-l} from Exercise 1.23 the linear operators
E, F and Hare given by the following matrices:

0 n-1 0 ... 000 0 0 0 ... 0 0 0

0 0 n-2 ... 000 1 0 0 ... 0 0 0
0 0 0 000 0 2 0 ... 0 0 0
E= F=
0 0 0 020 0 0 0 ... 0 0 0
0 0 0 00 1 0 0 0 ... n-2 0 0
0 0 0 000 0 0 0 ... 0 n-1 0

n-1 0 0 0 0 0
0 n-3 0 0 0 0
0 0 n-5 ... 0 0 0
H=
0 0 0 5-n 0 0
0 0 0 0 3-n 0
0 0 0 0 0 1-n

We complete this section with a description of homomorphisms between

simple modules:

Theorem 1.26 (Schur's lemma).

(i) Any non-zero homomorphism between two simple g-modules is an iso-

morphism.
(ii) For any two simple finite-dimensional g-modules V and W we have

Hom 9 (V, W) ~ {C, V~W;

0, otherwise.
10 Lectures on sb (<C)-modules

Proof. Let <I> E Hom 9 (V, W) be some non-zero homomorphism. Apply-

ing Lemma 1.14(i) we have that the kernel of <I> is a submodule of V. As
V is simple and <I> -/=- 0, we thus get that Ker( <I>) = 0 and <I> is injective. By
Lemma 1.14(ii) the image of <I> is a submodule of W. As W is simple and
<I> -/=- 0, we thus get that Im( <I>) = W and <I> is surjective. Therefore any
non-zero element of Hom 9 (V, W) is an isomorphism, proving the statement
(i). In particular, Hom 9 (V, W) = 0 if V '¥. W.
Assume now that V ~ W and 0 -/=- \I! E Hom 9 (V, W). We then have an
obvious isomorphism from Hom 9 (V, V) to Hom 9 (V, W) given by <I> f-7 \I! o <I>.
Let us show that
Hom 9 (V, V) = lC idv ~ lC.
If <I> E Hom 9 (V, V) is non-zero, it must have a non-zero eigenvalue, >.. E re
say (as re is algebraically closed). Then <I> - >.. · idv E Hom 9 (V, V). However,
any eigenvector of <I> with eigenvalue>.. belongs to the kernel of <I> - >.. · idv.
Thus <I> - >.. · idv is not an isomorphism, and hence it must be zero by the
previous paragraph. This yields <I> = >.. · idv and completes the proof. D

1.3 Semi-simplicity of finite-dimensional modules

Given two g-modules V and W the vector space V EB W can be endowed

with the structure of a g-module as follows:
E(v EB w) = E(v) EB E(w),
F(v EB w) = F(v) EB F(w), (1.12)
H(v EB w) = H(v) EB H(w).
The module V EB W is called the direct sum of V and W. For n E N and
any g-module V we denote by n V the g-module

n summands

Exercise 1.27. Check that the formulae (1.12) indeed define on V EB W

the structure of a g-module.

A g-module V is called decomposable provided that V ~ Vi EB Vi for

some non-zero g-modules Vi and Vi. Those g-modules which are not de-
composable are called indecomposable. A module, which is isomorphic to a
direct sum of (possibly finitely many) simple modules is called semi-simple.
 Finite-dimensional modules 11

Example 1.28. Every simple module is indecomposable; in particular,

every one-dimensional module is indecomposable. Indeed, if V ~ V1 EB Vi
and both V1 and Vi are non-zero, then Vi is a proper submodule of V and
hence V cannot be simple.

In the general situation (for example if one considers all g-modules)

there exist many indecomposable modules which are not simple. We will see
many examples later on. However, the case of finite-dimensional modules
turns out to be very special. The main aim of the present section is to
prove the following statement:

Theorem 1.29 (Weyl's Theorem). Every indecomposable finite-dimen-

sional g-module is simple. Equivalently, every finite-dimensional g-module
is semi-simple.

To prove this theorem we will need some preparation. From now and
until the end of the proof we assume that Vis a finite-dimensional g-module.
Consider the Casimir operator C = Cv on V, defined as follows:

C = (H + 1) 2 + 4F E.

Lemma 1.30.

(i) C=(H-l) 2 +4EF=H 2 +l+2EF+2FE.

(ii) HG= CH, EC= CE, FC =CF.

Proof. The statement (i) follows from the definition of C and the equality
EF =FE+ H. To prove the statement (ii) we use (i) and Exercise 1.16
as follows:

HG= H((H + 1) 2 + 4FE)

= H(H + 1) 2 + 4HFE
(by Exercise 1.16) = (H + 1) 2 H + 4F(H - 2)E
(by Exercise 1.16) = (H + 1)2 H + 4F EH
= ((H + 1)2 + 4FE)H
=CH;
12 Lectures on sb(rc)-modules

EC= E((H + 1) 2 +4FE)

= E(H + 1) 2 + 4EFE
(by Exercise 1.16) = (H -1)2E+4EFE
= ((H - 1) 2 + 4EF)E
by (i) =CE.

The equality FC = CF is checked in the same way. D

To proceed, we need to recall the following result from linear algebra:

Exercise 1.31. Let W be a vector space, A and B two linear commuting

operators on the space W and >. E <C. Show that both the subspace

{ w E W : Aw = >.w}

and the subspace

{ w E W : (A - >.)kw= 0 for some k E N}

are invariant with respect to B.

Applying the Jordan Decomposition Theorem to the linear operator C

on V we find that

where

V(C,T) = {v EV (C -T)kv = 0 for some k EN}.

Lemma 1.32. For any TE <C the subspace V(C,T) is a g-submodule ofV.
In particular, if Vis indecomposable, then V = V(C,T) for some TE <C.

Proof. By Lemma l.30(ii), the operator C commutes with the operators

E, F and H. Hence all these operators preserve V(C,T) by Exercise 1.31.
The claim follows. D

Exercise 1.33. Check that CvcnJ = n 2 · idvcnJ for all n E N.

 Finite-dimensional modules 13

Now we are ready to prove Theorem 1.29:

Proof. Let V be a non-zero indecomposable finite-dimensional g-module.
This means that it has a non-trivial simple submodule and hence from
Lemma 1.32 and Exercise 1.33 we obtain that V = V (C, n 2 ) for some
nEN.
Consider the decomposition (1.7). First, we claim that E acts injectively
on any V(>.), >. -1- n-1, -n-1. Indeed, for any v E V(>.)nKer(E) we have

E(H(v)) = EHv c;;;l HEv - 2Ev = 0,

and hence V(>.) n Ker(E) is invariant under the action of H. If we have
the inequality V(>.) n Ker(E) -1- 0, then VA n Ker(E) -1- 0 and for any
v E VA n Ker( E) we have
Cv = ((H + 1) 2 + 4FE)v = (H + 1) 2 v + 4FEv = (>. + 1) 2 v.
At the same time Cv = n 2 v as V = V(C,n 2 ), which implies>.= n-1 or
>. = -n-1. Analogously, one can show that Facts injectively on any V(>.)
such that >. -1- 1 - n, n + 1.
Since V is finite-dimensional, the previous paragraph implies that the
inequality V(>.) -1- 0 is possible only if>. E {-n + 1, -n + 3, ... , n - 1} and
that Ker(E) = V(n - 1), Ker(F) = V(l - n). In particular, dim V(>.) =
dim V(µ) for any>.,µ E {-n + 1, -n + 3, ... , n - l}. Furthermore, for any
i E {1, 2, ... , n - 1} the restriction Ai of the operator Fi to the subspace
V(n -1) gives a linear isomorphism from V(n -1) to V(n -1- 2i). Hence
we can identify V (n - 1) and V (n - 1 - 2i) as vector spaces via the action
of Ai. Set A= An-1·
As C commutes with H, all V(>.)'s are invariant with respect to C.
Denote by C 1 and H 1 the restrictions of C and H to V (n - 1), respectively.
Denote by C 2 and H 2 the restrictions of C and H to V(l - n) respectively.
Restricting cpn- 1 = pn- 1 c to V(n - 1) we get
AC1 = C2A. (1.13)
Analogously, using F H = (H + 2)F (see Exercise 1.16) we get
AH1 = (H2 + 2(n - l))A. (1.14)
As Ker(E) = V(n -1) and C = (H + 1) 2 + 4FE, we have
C1=(H1+1) 2 . (1.15)
As Ker(F) = V(l - n) and C = (H - 1) 2 + 4EF, we have
C2 = (H2 -1) 2. (1.16)
14 Lectures on sb(<C)-modules

Thus we have:
(H1+1) 2 c1;;5 l C1
A- 1 AC1
(by (1.13)) A- 1 c2A
(by (1.16)) A- 1(H2 - 1) 2A
(by (1.14)) A- 1 A(H1 - 1- 2(n - 1)) 2
(H1 - 1 - 2(n - 1)) 2 .
Hence (H1 +1) 2 = (H1 -1- 2(n-1)) 2 , which reduces to H 1 = n -1. This
means that V(n - 1) = Vn-1· Since AiH = (H + 2i)Ai and Ai identifies
the space V(n -1) with the space V(n -1- 2i) for all i, we get V(>.) =Vi
for all>. E {-n + 1, -n + 3, ... , n - 1}.
Let { V1, ... 'vk} be a basis of Vn-1· For i E {1, ... 'k} denote by withe
linear span of {Vi, FVi, ... , pn- 1 vi}. From the above we have

and, by Corollary 1.19, each Wi is a submodule of V. Since V is indecom-

posable by our assumptions, we get k = 1 and thus dim Vn-1 = 1. In this
case Corollary 1.19 and (1.9) imply that V ~ y(n), which completes the
~~ D
Corollary 1.34. Let V be a finite-dimensional g-module. Then
V ~ EBmn y(nl,
nEN
where
m n =dim Hom g (V(n) ' V) =dim Hom g (V' y(n)) .

Proof. From Theorem 1.29 it follows that we can decompose V into a

direct sum of simple modules, say V ~ Vi E9 · · · E9 Vk, where all the Vi's are
simple. Now
dim Hom 9 (V(n), V) = dim Hom 9 (V(n), EEli Vi)
= LdimHom9 (V(nl, Vi)

(by Schur's lemma)= l{i : y(n) ~ Vi}I.

This proves the first equality for mn and the second equality is proved
similarly. D
 Finite-dimensional modules 15

1.4 Tensor products of finite-dimensional modules

Given two g-modules V and W the vector space V ® W can be endowed

with the structure of a g-module as follows:

E(v ® w) = E(v) ® w + v ® E(w),

F(v ® w) = F(v) ® w + v ® F(w), (1.17)
H(v ® w) = H(v) ® w + v ® H(w).
The module V ® W is called the tensor product of V and W. For n E N
and any g-module V we denote by V®n the g-module

V®V®···®V.
n factors

Exercise 1.35. Check that the formulae (1.17) do indeed define on V ® W

the structure of a g-module.

Exercise 1.36. Let V and W be two g-modules. Check that the map
v ® w f--+ w ® v induces an isomorphism between V ®Wand W ® V.

Exercise 1.37. Let Vi, V2 and W beg-modules. Prove that

(Vi EB V2) ® W ~ (Vi ® W) EB (V2 ® W).

Exercise 1.38. Let V, Wand Ube g-modules. Prove that

V ® (W ® U) ~ (V ® W) ® U.

If both V and W are finite-dimensional, the module V ® W is finite-

dimensional as well. Due to Corollary 1.34, it is natural to ask how V ® W
decomposes into a direct sum of simple modules (depending on V and W).
Exercises 1.36 and 1.37 mean that to answer this question it is sufficient to
consider the case when both V and 1¥ are simple modules. This is what
we will do in this section. Our main result is the following:

Theorem 1.39. Let m, n EN be such that m::; n. Then

y(n) ® y(m) ~ y(n-m+l) EB y(n-m+3) EB .. ·EB y(n+m-3) EB y(n+m-1). (1.18)
16 Lectures on .st2(<C)-modules

Proof. We proceed by induction on m. If m = 1, the module v< 1 J ~ C

is the trivial g-module and hence y(n) ® y(l) is isomorphic to y(n), for
example via the isomorphism v ® 1 f-+ v.
Let m = 2. Then v< 2 J is the natural g-module. Let { e 1 , e 2 } be the
natural basis of v< 2 J. Then the action of E, F and Hin this basis is given
by the following picture (see (1.9)):

-1

: 1 : : 0
y________,._. y.------..,,,._
'"===="' e2 ~ e 1 (1.19)
0 1

Assume that y(n) is given by (1.9). Then from the formulae (1.17) we
obtain that the vector v0 ® e 1 E y(n) ® v< 2 J satisfies E(v 0 ® e 1 ) = 0 and
H(v 0 ® e 1 ) = n(v0 ® e 1 ). The only y(i) which contains a non-zero vector
with such properties is y(n+l) (see Exercise 1.25). Hence y(n+l) is a direct
summand of y(n) ® v< 2J.
Let w = v1 ® e 1 - (n - l)v 0 ® e 2 -1- 0. Using the definitions, one can
easily check that E(w) = 0 and H(w) = (n - 2)w. The only y(i) that
contains a non-zero vector with such properties is y(n-l) (Exercise 1.25).
Hence y(n-l) is a direct summand of y(n) ® v< 2J. But

dim y(n) ® v< 2l = dim y(n) X dim v< 2l

= 2n
= (n-l)+(n+l)
=dim y(n-l) +dim y(n+l).

This implies that

y(n) @ y(2) ~ y(n-1) EB y(n+l). (1.20)

Now let us prove the induction step. We assume that k > 2 and that
(1.18) is true for all m = 1, ... , k-1. Let us compute y(n) ® y(k-l) ® vC 2)
in two different ways. On the one hand we have

(1.20)
y(n)@ y(k-1)@ y(2) ~ y(n)@ (V(k) EB y(k-2))
(by Exercise 1.37) C>< y(n) ® y(k) EB y(n) ® y(k- 2) (1.21)
(by inductive assumption) ~ y(n) @ y(k) EB y(n-k+ 3) EB ...
. . .. EB y(n+k-5) EB y(n+k-3).
 Finite-dimensional modules 17

On the other hand, using the inductive assumption we have

y(n) Q9 y(k-1) Q9 y(2) 9" (~ y(n-k+2+2i)) Q9 y(2)

k-2
(by Exercise 1.37) 9" EB y(n-k+2+ 2i) @ y(2)
i=O
k-2
(by (1.20)) 9" EB (y(n-k+3+2i) EB y(n-k+1+2i))
i=O
9" yCn-k+l) EB yCn-k+3) EB ... EB y(n+k-l)EB
EBV(n-k+3) EB y(n-k+5) EB ... EB y(n+k-3).
(1.22)
The statement of the theorem now follows, comparing (1.21) with (1.22)
and using the uniqueness of the decomposition of y(n) Q9 y(k-l) Q9 v< 2l into
a direct sum of irreducible modules (Corollary 1.34). D

1.5 Unitarizability of finite-dimensional modules

The correspondence
e*=f, f*=e, h*=h
uniquely extends to a skew-linear involution * on the vector space g in the
sense that (.A.x)* =Xx* for all x E g and A. E <C, where~ denotes the complex
conjugation. This involution satisfies
[x*, y*] = [y, x]*
for all x, y E g and hence is a (skew) anti-involution of the Lie algebra g.
The involution * induces an involution on the set { E, F, H}, which we will
denote by the same symbol.
A g-module V is called unitarizable with respect to the involution *
provided that there exists a (positive definite) Hermitian inner product
(-, ·) on V such that
(X(v),w) = (v,X*(w)) (1.23)
for all v, w E V and X E { E, F, H}. The aim of this section is to prove the
following result:

Theorem 1.40. Every finite-dimensional g-module is unitarizable.

Exercise 1.41. Show that a direct sum V EB W of two g-modules V and

W is unitarizable if, and only if, each summand is unitarizable.
18 Lectures on sb (IC)-modules

Proof. By Corollary 1.34, every finite-dimensional g-module decomposes

into a direct sum of modules y(n), n E N. Hence Exercise 1.41 implies that
it is enough to prove the statement of the theorem for the modules y(n),
nEN.
Assume that n E N and the module y(n) is given by (1.9). Note that
all ai > 0 and define
Ci-1 .
Co= 1, Ci=--, Z = 1, ... , n - 1.
,;a;
Then Ui =Ci Vi defines a diagonal change of basis in y(n). In the basis {ui}
the action of E, F and H is given by:
-n+l -n+3 n-5 n-3 n-1

Let (-, ·) be the inner product on y(n) with respect to which the basis
{uo, ... , Un-1} is orthonormal. From (1.24) it follows by a direct calculation
that in this basis the linear operators E, F and H satisfy (1.23). This
proves that y(n) is unitarizable. As mentioned above, the general statement
follows. D
The anti-involution * is not the only anti-involution on g. The corre-
spondence
e 0 = e, f 0 = f, h 0 = -h
uniquely extends to a linear involution o on g. This involution satisfies
[xo,yo] = [y,x]°
for all x, y E g and hence is an anti-involution of the Lie algebra g. The
involution o induces an involution on the set {E, F, ±H}, which we will
denote by the same symbol. Ag-module V is called a o-module provided
that there exists a non-degenerate symmetric bilinear form ( ·, ·) on V such
that
(X(v),w) = (v,X 0 (w)) (1.25)
for all v,w EV and XE {E,F,H}.

Exercise 1.42. Let V be a non-trivial simple finite-dimensional o-module

with the corresponding symmetric bilinear form (·, ·). Show that ( ·, ·) is
neither positive nor negative definite.
 Finite-dimensional modules 19

Exercise 1.43. Show that a direct sum V EB W of two o-modules is a

o-module.

Theorem 1.44. Every finite-dimensional g-module is a o-module.

Proof. By Corollary 1.34, every finite-dimensional g-module decomposes

into a direct sum of modules y(n), n E N. Hence Exercise 1.43 implies that
it is enough to prove the statement of the theorem for the modules y(n),
nEN.
Assume that n EN and the module y(n) is given by (1.9). Let (·,·)be
the symmetric bilinear form on y(n) which is given by the matrix

000 ... 001

000 ... 010
000 ... 100

001 ... 000

010 ... 000
100 ... 000

in the basis {vo, v 1 , ... , Vn- 1 }. From (1.9) it follows by a direct calculation
that in this basis the linear operators E, F and H satisfy (1.25). This
proves that y(n) is a o-module and completes the proof. D

Note that the proof of Theorem 1.44 can be seen as a kind of justification
of the basis {vo, v1, ... , Vn-1} of the module y(n).
After the above results, given some g-module V it is natural to ask
how many different forms (., ·) on V the module V is unitarizable for (or a
o-module). The answer turns out to be easy for simple finite-dimensional
modules.

Proposition 1.45. Let V be a simple finite-dimensional g-module. Then

the Hermitian inner product with respect to which the module V is unita-
rizable is unique up to a positive real scalar.

Proof. Let V ~ y(n), n E N, and (·, ·) be an Hermitian inner prod-

uct on V with respect to which V is unitarizable. Consider the basis
{u 0 , ... , Un-l} from the proof of Theorem 1.40. The vectors in this basis
are eigenvectors of the self-adjoint linear operator H corresponding to pair-
wise different eigenvalues. Hence (ui, Uj) = 0 for all i -f. j. Let (u 0 , u 0 ) = c.
Then c is a positive real number. Let us prove that (ui, ui) = c for all
20 Lectures on sf2(C)-modules

i E {O, ... , n - 1} by induction on i. The basis of the induction is trivial.

For all i E {1, ... , n - 1} we have

(by induction) c.

The claim follows. D

Proposition 1.46. Let V be a simple finite-dimensional g-module. Then

the non-degenerate symmetric bilinear form with respect to which the mod-
ule V is a <>-module is unique up to a non-zero complex scalar.

Proof. Let V ~ y(n), n E N, and (., ·) be a non-degenerate symmetric

bilinear form on V with respect to which V is a <>-module. Consider the
basis {vo,v1, ... ,Vn-1} from (1.9). For i,j E {O, ... ,n-1} by (1.25) we
have (H(vi), vj) =-(vi, H(vj)). As all elements of our basis are eigenvec-
tors to H with pairwise different eigenvalues, it follows that (vi, Vj) -I- 0
implies that the eigenvalues Ai and Aj of Vi and Vj, respectively, satisfy
Ai=-Aj· Hence (vi,vj)-=f-Oimpliesi=n-1-j. Let (vo,Vn- 1)=c. As
(-, ·) is non-degenerate, we have c -I- 0. Let us show by induction on i that
(vi,Vn-1-i) = c for all i E {O, 1, ... ,n-1}. For all such i > 0 we have

(vi, Vn-1-i) (~) (F(vi_i), - 1-E(vn-i))

an-i
1
(by (1.25)) -(vi-1, F(E(vn-i)))
an-i
1
(by (1.9)) -(vi-1,an-iVn-i)
an-i
(vi-li Vn-i)
(by induction) c.

The claim follows. D

 Finite-dimensional modules 21

For a direct sum of simple modules, the description of bilinear forms

analogous to Propositions 1.45 and 1.46 will be more complicated. In par-
ticular, as an obvious observation one could point out that it is possible
to independently rescale the restrictions of the bilinear form to pairwise
orthogonal direct summands.

1.6 Bilinear forms on tensor products

Let V and W be two vector spaces and (·, · )i and (·, · )2 be bilinear forms
on V and W respectively. Then the assignment

(v ® w, v' ® w') = (v, v')i · (w, w')2 (1.26)

extends to a bilinear form on the tensor product V ® W.

Exercise 1.47. Check that the form(-,·) is symmetric provided that both
(-, ·)i and (-, ·)2 are symmetric; that the form (-, ·) is non-degenerate pro-
vided that both (·, · )i and ( ·, ·)2 are non-degenerate; and that the form (-, ·)
is Hermitian provided that both(·, ·h and(·, ·)2 are Hermitian.

Proposition 1.48. Assume that V and W are unitarizable modules (resp.

<>-modules) for the forms (-, ·h and (-, ·h respectively. Then V ® W is
unitarizable (resp. a <>-module) for(·,·).

Proof. We prove the statement for unitarizable modules. For o-modules

the proof is similar. Due to Exercise 1.4 7 it is sufficient to check (1.23) for
XE {E,F,H}. For v,v' EV and w,w' E W we have

(X(v ® w), v' ® w') C\;7 ) (X(v) ® w + v ® X(w), v' ® w')

(by linearity) (X(v) ® w, v' ® w') + (v ® X(w), v' ® w')
(by (1.26)) (X(v), v')i · (w, w'h + (v, v')i(X(w), w')2
(by (1.23)) (v,X*(v'))i · (w,w')2 + (v,v')i(w,X*(w')h
(by (1.26)) (v ® w, X*(v') ® w') + (v ® w, v' ® X*(w'))
(by linearity) (v ® w, X*(v') ® w' + v' ® X*(w'))
(by (1.17)) (v ® w, X*(v' ® w')).

The claim follows. D

We know that the tensor product of two simple finite-dimensional g-

modules is not simple in general (see Theorem 1.39). Hence the bilinear
22 Lectures on sl2(<C)-modules

form making this tensor product module unitarizable or a <>-module is usu-

ally not unique (even up to some scalar). However, we would like to finish
this section with a description of one invariant, which turns out in the real
case.

Exercise 1.49. Consider the real Lie algebra s[ 2 (JR). Show that (1.9) still
defines on the real span V~n) of {Vo, ... , Vn-1} the Structure of a simple
s[2 (JR)-module. Check that the analogues of Theorem 1.39 and all the
above results from Sections 1.5 and 1.6 are true for s[ 2 (JR) with the same
proofs.

After Exercise 1.49 one could point out one striking difference between
the real versions of Proposition 1.45 and Proposition 1.46. It is the possi-
bility of the sign change in the assertion of Proposition 1.46 (note that two
forms which differ by a sign change cannot be obtained from each other
by a base change in the original module). Let us call the form on y(n),
described in the proof of Proposition 1.46, standard, and the form, obtained
from the standard form by multiplying with -1, non-standard. Our main
result in this section is the following:

Theorem 1.50. Let m, n EN, m :<; n; (., ·h and(., ·h be standard forms

on V~m) and V~n) respectively; and(-,·) be the form on V~n) @V~ml given
by (1.26). Then, up to multiplication with a positive real number, for i =
0, 1, ... , m the restriction of(.,·) to the direct summand V~n+m-l- 2 i) of
v~n) @ v~m) is standard for all even i and non-standard for all odd i.
Proof. As in the proof of Theorem 1.39 we use induction on m. For
m = 1 the statement is obvious. To proceed we will need the following
lemma:

Lemma 1.51. Assume that the form (., ·)' on V~n) makes V~n) into a<>-
module. Then(-,·)' is standard if, and only if, (v 0 ,Fn- 1 (v0 ))' > 0 and is
non-standard if, and only if, (vo, pn- 1 (vo))' < 0.

Proof. From the definition we have that the form (., · )' is standard if,
and only if, (vo, Vn-1)' > 0. From (1.9) we have pn- 1(vo) = Vn-1· The
claim follows. D
Let m = 2, n ?: 2 and assume that V~n) is given by (1.9) and V~2 )
is given by (1.19). As all coefficients in (1.17) are positive, we get that
Fn(vo@ ei) = CVn-1@ e2, where c > 0. As
(vo@ el, Vn-1@ e2) = (vo, Vn-1)i(e1, e2)2 = 1 > 0
 Finite-dimensional modules 23

(here we used that both (-, ·)i and (-, ·h are standard), from Lemma 1.51
we obtain that the restriction of (-, ·) to the direct summand V~n+l) of
v~n) ® V~2 ) is standard.
For the element w = v 1 ®e 1 -(n- l)v 0 ®e 2 -j. 0 we have E(w) = 0, sow
generates the direct summand V~n-l) of V~n) ® v~l. A direct computation
shows that
pn- 2 (w) = Vn-l ®el - Vn-2 ® e2.
Another direct computation then shows that
(v1 ®el - (n - l)vo ® e2,Vn-l ®el -Vn-2 ® e2) = -n < 0.
Hence from Lemma 1.51 we obtain that the restriction of(-,·) to the direct
summand V~n+i) of V~n) ® V~2 ) is non-standard. This completes the proof
of the theorem in the case m = 2.
Fork E N let us denote Vf,;,+l and V~k,-) the module V~k) endowed
with a standard and non-standard (up to a positive real scalar) form, re-
spectively. Then we have just proved that
y(n,+) f7\ y(2,+) co,; yCn+l,+) "'y(n-1,-) (1.27)
IR'. \Yffi'. -IR'. 'CDJR'..

Note that we obviously have

V (n,+) f7\ y(l,+) co,; y(n,+) y(n,+) f7\ y(l,-) co,; y(n,-)
(1.28)
IR'. \Yffi'. -IR'.' IR'. \Yffi'. -IR'..

In this notation the statement of our theorem can be written as follows:

V IR'(n,+)
.
f7\
\Y
y(m,+)
IR'.
co,;
-
y(n+m-1,+)
IR'.
LD
\l)
y(n+m-3,-)
IR'.
LD
\.D
y(n+m-5,+)
IR'.
LD
\J.J····
The induction step now follows using (1.27) and (1.28) and, rewriting in
this new notation, the calculations in (1.21) and (1.22). We leave the details
to the reader. D

1. 7 Addenda and comments

1.7.1
Alternative expositions for the material presented in Sections 1.1-1.4
can be found in a large number of books and articles, see for example
[37, 46, 49, 57, 106]. Many of the results are true or have analogs in
much more general contexts (which also can be found in the books listed
above). In particular, simple finite-dimensional modules are classified (see
Theorem 1.22) and Weyl's Theorem (Theorem 1.29) is true for all simple
finite-dimensional complex Lie algebras. For all such algebras there is also
an analog of Theorem 1.39, however its formulation is more complicated,
as higher multiplicities appear on the right hand side.
24 Lectures on sl2(<C)-modules

1.7.2
If A is an associative algebra with associative multiplication ., then one can
define on A the structure of a Lie algebra using the operation of taking the
commutator with respect to-: [a, b] = a·b-b·a. The Lie algebra (A,[·,·]) is
called the Lie algebra underlying the associative algebra (A,·) and is often
denoted by A (- ). In particular, if V is a vector space, one can consider the
associative algebra £(V) of all linear operators on V and the underlying
Lie algebra £(V)(-).
An st2-module is then given by a Lie algebra homomorphism from s[2
to £(V)(-), that is a linear map rp : st2 ----+ £(V), which satisfies

rp([x, y]) = [rp(x), rp(y)] (1.29)

for all x, y E st2. For such rp in the notation of Section 1.1 we simply have
H = rp(h), F = rp(f) and E = rp(e).
Substituting st2 with an arbitrary Lie algebra, one obtains the notion
of a module over any Lie algebra. The homomorphism rp is usually called a
representation of the Lie algebra. Hence the notions of module and repre-
sentation are equivalent, differing only in their emphasis on the underlying
vector space V (for modules) or the homomorphism rp (for representations).
Sometimes one can also say that a representation defines an action of the
Lie algebra on the underlying vector space V.

1.7.3
The Lie algebra s(2 is a subalgebra of the Lie algebra gt 2, the latter being
the underlying Lie algebra of the associative algebra of all complex 2 x 2 ma-
trices. Moreover, the algebra gt 2 is a direct sum of st2 and the commutative
Lie subalgebra of all scalar matrices.

1.7.4
Weyl's Theorem can be proved using the notion of unitarizability of finite-
dimensional modules. Let V be an arbitrary finite-dimensional g-module.
Using the exponential map one first could lift the g-action on V to the action
of the group SL(2) and further SU(2). In fact one can show that there is a
natural bijection between finite-dimensional g-modules, finite-dimensional
SL(2)-modules and finite-dimensional SU(2)-modules. As SU(2) is com-
pact, all finite-dimensional SU(2)-modules and hence all finite-dimensional
 Finite-dimensional modules 25

g-module are completely reducible. We refer the reader to Chapter III, §6

of [106] for details.

1.7.5

So far, all modules which we considered were left modules. There is also
the natural notion of a right module. The triple E', F' and H' of linear
operators on a vector space V defines on V the structure of a right g-module
provided that the operators E', F' and H' satisfy
F'E' - E'F' = H', E'H' - H'E' = 2E', F'H' - H'F' = -2F'.
This corresponds to an antihomomorphism from g to .C(V) C- ) , that is to a

linear map cp : g-+ .C(V), which satisfies

cp([x, y]) = [cp(y), cp(x)] (1.30)
for all x, y E .s(2 . In what follows, by "module" we will always mean a left
module.

1.7.6

If A is an algebra (associative or Lie), then for any left A-module V the dual
space V* = Hom(V, C) carries the natural structure of a right A-module
given by (a(f))(v) = f(a(v)) for a EA, v EV and f EV*.
However, if A is an algebra (associative or Lie) with a fixed anti-
involution q, then for any left A-module V the space V* carries the natural
structure of a left A-module given by (a(f))(v) = f(aq(v)) for a EA, v EV
and f EV*.
Any element cp E HomA(V, V*) defines a bilinear form on Vas follows:
(v,w)'P = cp(v)(w). This form obviously satisfies
(a(v), w) = (v, aq(w)). (1.31)
If V is finite-dimensional then every bilinear form on V satisfying (l.31)
has the form (v,w)'P for some cp E HomA(V, V*). The form (v,w)'P is non-
degenerate if, and only if, cp is an isomorphism. These general arguments
give an alternative proof of Propositions 1.45 and 1.46. More details and
some further related results can be found in [93].

1.7.7

Theorem 1.50 appears in [6] (in the form presented later on in Exercise 1.70)
in connection to the study of Hodge-Riemann relations for polytopes. Our
26 Lectures on st2(C)-modules

proof follows the general idea of [6]. There exists an alternative "brute
force" argument for Theorem 1.50 worked out in [71]. Here is its outline:
Let m, n E N and n ;:o: m. Assume that y(n) is given by (1.9) and
y(m) is similarly given by (1.9) in the basis { wo, w1, ... , Wm-d· For k =
0, 1, ... , m - 1 set lk = m + n - 2 - 2k. A direct calculation shows that the
element
k

Uk= 2...)-l)ivn-l+i 181 Wm-l+k-i

i=O
satisfies
H(uk) = -lkuk, F(uk) = 0.
In the same way as in Lemma 1.51, one shows that the form
(·, ·)' on v~n)' making v~n) into a <>-module, is standard if, and
only if, (vn-1,En- 1(vn-1)) 1 > 0 and non-standard if, and only if,
(vn-1,En- 1(vn_i))' < 0. Hence to prove Theorem 1.50 one simply has to
show that for any k = 0, 1, ... , m-1 the sign of the number (Uk, E 1k ( uk)) al-
ternates with k. This reduces to the computation of Eh (Vk), which is not
entirely straightforward. However, using the following Karlsson-Minton
identity for the hypergeometric series 3 F 2 : for all integers a, b, c, d such
that 0 :S b :S a :S min( c, d) we have
min(a,c-b)
(-l)i (c-i)!(d-a+i)! =(-l)a+b
L
i=max(0,2a-b-d)
i!(a-i)!(c-b-i)!(d+b-2a+i)! '
one proves by a direct calculation that

Elk(
Uk
)=l '~(-l)k+i(m-1-i)!(n-l-k+i)!
k· L__, . (k _ ")I
. .
Vk-i 181 Wi·
i.1 i .
i=O

Using the last formula, the computation of (uk, En+m- 2- 2k(uk)) is fairly
straightforward and yields the necessary result. We refer the reader to [71]
for details.

1.8 Additional exercises

Exercise 1.52. Let a denote the vector space with the basis {e_ 1 ,e 0 ,e 1 }.
Define the bilinear operation [·, ·] on a via

[e 2., e1·]-{(j-ih+j,
-
i+jE{-1,0,l};
0, otherwise.
 Finite-dimensional modules 27

Show that this makes a into a Lie algebra. Show further that a is isomorphic
to s[2.

Exercise 1.53. Let b denote the vector space with the basis {a, b, c}. De-
fine the antisymmetric bilinear operation [-, ·] on b via

[a,b] = c, [b,c] =a, [c,a] = b.

Show that this makes b into a Lie algebra. Show further that b is isomorphic
to s[2.

Exercise 1.54. Consider the vector space V = C[x, y] and the linear op-
erators
8 8 8
E=x·- H=x·--y·-
ay' ax ay
on V.

(a) Show that the operators E, F and H make V into a g-module.

(b) Show that for every n E No the linear span of all homogeneous polyno-
mials of degree n is a submodule of V, isomorphic to y(n+l).

Exercise 1.55. Consider the vector space V = Mat 3 x 3 (C) of all complex
3 x 3 matrices and the matrices

001)
X= ( 000 ,
000
000)
y = ( 000
100
Z=(~~~)-
0 0 -1

Show that the linear operators

E = [X, _], F = [Y, _], H = [Z, -l

(here E(A) = [X, A] = X A - AX for all A E Mat 3 x 3 (C) and similarly
for F and H) define on V the structure of a g-module and determine its
decomposition into a direct sum of y(n) 's.

Exercise 1.56. Let n EN. For every>. E {-n+ 1, -n+3, ... , n-3, n-1}
fix some non-zero element X.A Evin) (see (1.8)). Show that the set

x={x.A: .AE{-n+l,-n+3, ... ,n-3,n-1}}

is a basis of y(n).
28 Lectures on sb(IC)-modules

Exercise 1.57 (Gelfand-Zetlin model, [52]). Let c E C be a fixed

complex number. For n E N set c' = c - n and consider the set T c,n
consisting of all tableaux

where a E { c- i : i = 0, 1, ... , n -1 }. Let V = Vc,n denote the linear span

of all elements from Tc,n· Define the linear operators E, F and Hon Vas
follows:

F(tc(a)) = {tc(a - 1), tc(a - ~) E Tc,n;

0, otherwise.

E(tc(a)) = {-(c -
0,
a)(c' - a)tc(a + 1), tc(a + ~) E Tc,n;
otherwise.
H(tc(a)) = (2a - c - c' - 1) · tc(a).
Show that this turns V into a g-module, which is isomorphic to y(n).

Exercise 1.58. Write a Cayley table of the Lie algebra g[ 2 in the standard
basis {e 11 , e 12 , e 21 , e 22} consisting of matrix units.

Exercise 1.59 (Gelfand-Zetlin model for g[2 , [52]).

(a) Show that the s[2-module structure on the module Vc,n from Exer-
cise 1.57 can be extended to a glrmodule structure in the following
way:
e12(tc(a)) = E(tc(a)),
e21(tc(a)) = F(tc(a)),
en(tc(a)) =a· tc(a),
e22(tc(a)) = (c + c' - a+ 1) · tc(a).
(b) Show that the module Vc,n is a simple glrmodule.
(c) Show that Vc,n ~ Vd,m if, and only if, c = d and n = m.
(d) Show that every simple glrmodule is isomorphic to Vc,n for some c E C
and n EN.

Exercise 1.60. Construct a counterexample which shows that Weyl's the-

orem fails for finite-dimensional glrmodules.

Exercise 1.61. Let A and B be two linear operators on some finite-

dimensional vector space V.
 Finite-dimensional modules 29

(a) Prove that [A, B] = >. · A for some >. E C, >. =/= 0, implies that the
operator A is nilpotent.
(b) Prove that [A, B] = A 2 implies that the operator A is nilpotent.
(c) Prove that [A, [A, B]] = 0 implies that the operator [A, B] is nilpotent.

Exercise 1.62. Let V be a g-module and n EN. Consider then-th tensor

power V®n of V. Let Sn denote the symmetric group on { 1, 2, ... , n}.
Show that the linear span of all vectors of the form
L Va-(1) ® Va-(2) ® ... ® Va-(n-1) ® Va-(n)'
o-ESn

where V1' V2' ... 'Vn E v forms a g-submodule of v®n. This submodule is
called the n- th symmetric power of V and is denoted by Symn (V).

Exercise 1.63. Prove that Symn(VC 2l) ~ y(n+l).

Exercise 1.64. Let V and n E N be as in Exercise 1.62. Show that the

linear span of all vectors of the form
L sign(O'). Va-(1) ® Va-(2) ® ... ® Va-(n-1) ® Va-(n),
o-ESn

where vi, v2, ... , Vn E V forms a g-submodule of V. This submodule is

called then-th exterior power of Vandis denoted by/\ n(V).

Exercise 1.65. Prove that /\ n(vCnl) ~ y(ll.

Exercise 1.66. Prove that /\ n(V) ~ 0 provided that n >dim V.

Exercise 1.67.

(a) Show that the correspondence

eq = -e, fq = -f, hq = -h
uniquely extends to an anti-involution q on g.
(b) Prove that for every n E N there exists a unique (up to a non-zero
scalar) non-degenerate bilinear form (·, ·)n on y(n) such that
(X(v), w)n = (v, Xq(w))n
for all XE {±E,±F,±H}and all v,w E y(nl.
(c) Prove that the form (., · )n is symmetric for odd n and antisymmetric
for even n.

Exercise 1.68. Formulate and prove an analogue of Proposition 1.48 for

the anti-involution q from Exercise 1.67.
30 Lectures on sb(rc)-modules

Exercise 1.69. Formulate and prove an analog of Theorem 1.50 for the
form (-, ·)n from Exercise 1.67.

Exercise 1. 70 ( [6]). For n E N set

y(n,+)
y(n) = { ' n is odd;
y(n,-J, n is even
(see notation of Theorem 1.50). Show that for any k, m E N the tensor
product y(k) ® y(m) decomposes into a direct sum of modules of the form
y(nl, n EN.

Exercise 1. 71. Fix n E N. For i = 1, ... , n - 1 let Si denote the involution

(i, i + 1) of the symmetric group Sn. The group Sn is a Coxeter group in
the natural way. Denote by R the set of all reflections in Sn. In particular,
the s;'s are simple reflections. Denote by e the identity element of Sn. For
i = 0, ... , n - 1 and j = 0, ... , i set

j =0;
j > O;
and define xi = {Xi,O, Xi,1, ... 'Xi,i-1}. Let xi denote the formal linear span
of the elements from Xi. Define on xi the structure of a g-module via (1.9)
using the convention VJ= xi,i-1-J·
(a) Show that every a E Sn admits a unique decomposition of the form
a= aoa1 ... °'n-1, where °'i E Xi for all i.
(b) Show that the underlying space of the tensor product X0 ®X 1®· · ·®Xn-l
can be canonically identified with C[Sn] via the map

XO,jo @ X1,j1 @ ... @ Xn-l,Jn-1 f---7 XO,j0Xl,j1 ... Xn-l,Jn-1 ·

This equips C[Sn] with the structure of a g-module. (This structure
comes from the Hard Lefschetz theorem (see [53]) applied to the co-
homology algebra of the flag variety, which can be naturally identified
with C[Sn] as a vector space, see [56].)
(c) Show that every a E Sn is a eigenvector for h (with respect to the
g-module structure described in (b)) with the eigenvalue

l{r ER : m < a}l - l{r ER : m > a}I,

where< denotes the Bruhat order on Sn. Show that the same number
equals - n(n2- 1l + 2Z(a), where l(a) is the length of a with respect to
the set {s1, s2, ... Sn-d of simple reflections.
 Finite-dimensional modules 31

Exercise 1. 72. Generalize Exercise 1. 71 to other Coxeter groups.

Exercise 1. 73. For every finite-dimensional g-module V we define the

function chv : Z --+ N0 as follows:
chv(.A) =dim Vi,, .A E Z.
(a) Show that chv(.A) = 0 for all .A E Z such that I.Al is big enough.
(b) Show that chv(.A) = chv(-.A) for all .A E Z.
(c) Show that chv(.A) ~ chv(µ) for all elements .A,µ E Z of the same parity
such that 0 S I.Al S lµI.
(d) Show that chv = chw if and only if V ~ W.
(e) Show that for any function ch : Z -+ N0 , which has the properties,
described in (a)-(c) above, there exists a unique (up to isomorphism)
g-module V such that ch = chv.

Exercise 1.74 ([60]). Show that the elements x = h, y = 2e - h and

z = -2f - h form a basis of g and that we have
[x,y] = 2x + 2y, [y, z] = 2y + 2z, [z,x] = 2z + 2x
The basis {x, y, z} is called the equitable basis of g.
 Chapter 2

The universal enveloping algebra

2.1 Construction and the universal property

As we saw in 1.7.2, a g-module corresponds to a Lie algebra homomorphism

cp : g ----+ L'.(V)(-). In all nontrivial cases, cp has one very annoying property:
the image of cp is not closed with respect to the composition of linear op-
erators, but is only closed with respect to taking the commutator of linear
operators (since the latter is the Lie algebra structure on L'.(V)C-l). This
heavily restricts our possibilities to analyze the structure of V as we are
forced to look for some external objects which initially seem unrelated to
the algebra g. A good example is the Casimir operator Cv from Section 1.3.
This operator does not belong to the image of the homomorphism defining
the module structure on V in general, so its appearance should be a total
mystery at the moment. At the same time, it played a crucial role in the
proof of Weyl's theorem, so it is clear that this operator is very important.
The aim of the present chapter is to clarify the situation in the following
way: We will define a certain associative algebra U(g), called the universal
enveloping algebra of g and show that it has the following properties:

• The Lie algebra g is a canonical subalgebra of the underlying Lie

algebra U(.g)H.
• Any g-action on any vector space (that is, any g-module) canoni-
cally extends to a U(g)-action on the same vector space (that is,
to a module over the associative algebra U(g)).
• This extension, and the restriction from U(g) to g, are mutually
inverse isomorphisms between the categories of all g-modules and
all U(g)-modules.

Basically, this says that the study of g-modules is the same as the study of

33
34 Lectures on s[2 (IC)-modules

modules over the associative algebra U(g). Any U(g)-module corresponds

to a homomorphism 'l/; : U(g) --> ,C(V) and the image of this homomorphism
is always closed with respect to the composition of operators. Hence the
study of U (g) mostly reduces to the study of the internal structure of U (g).
In particular, the Casimir operator Cv from Section 1.3 will turn out to be
the image of a special central element from U(g). The only disadvantage
of U(g) is that this algebra is infinite-dimensional, while the Lie algebra g
has finite dimension. But the above benefits mean that this is a reasonable
price to pay.
Consider the free associative algebra R with generators e, f and h and
denote by U(g) the quotient of R modulo the ideal generated by the fol-
lowing relations:
ef - fe = h, he - eh= 2e, hf - fh = -2f. (2.1)
The algebra U(g) is called the universal enveloping algebra of g. Abusing
notation we will usually identify the elements of R with their images in the
quotient algebra U(g).

Exercise 2.1. Prove that the image of the element (h + 1) 2 + 4fe from R
in the algebra U(g) belongs to the center of U(g).

Exercise 2.2. Prove that the elements (h + 1) 2 + 4fe and (h - 1) 2 + 4ef

are different elements in R, while their images in U(g) coincide.

Exercise 2.3. Show that there exists a unique anti-involution a- on U(g)

such that a-(f) = e and a-(h) = h.

Lemma 2.4.

(i) There is a unique linear map E : g--> U(g) satisfying

c(e) = e, c(f) = f, c(h) = h.

(ii) The map E is a Lie algebra homomorphism from g to U(g)(-).

Proof. The statement (i) follows from the fact that { e, f, h} is a basis
of g. The statement (ii) follows from (1.1), the definition of U(g) and the
definition of the underlying Lie algebra. D
The map E is called the canonical embedding of g into U (g) (- ) . However,
it is not obvious that this map is injective. We will prove this in the
next section. For the moment we would like to present the following very
important result:
 The universal enveloping algebra 35

Theorem 2.5 (Universal property of U(g)). Let A be any associative

algebra and VJ : g __, A (- l be any homomorphism of Lie algebras. Then
there exists a unique homomorphism lj5 : U(g) __, A of associative algebras
such that VJ = lj5 o E. In other words, the following diagram commutes:

(2.2)

Proof. We start with the existence of (/5. For the free algebra R with
generators e, f and h we have a unique homomorphism 'ljJ : R __, A of
associative algebras such that

'lji(e) = VJ(e), 'lji(j) = VJ(f), 'lji(h) = VJ(h). (2.3)

Let K be the kernel of the natural projection R--» U(g). We have

'l/!(ef - fe) = 'lji(e)'lji(J) - 'lji(J)'lji(e)

(by definition)= VJ(e)VJ(f) - VJ(f)VJ(e)
= [VJ(e), VJ(f)]
(as VJ is a Lie alg. homom.) = VJ([e, f])
(by (1.1)) = VJ(h)
(by definition) = 'lji(h).

This yields 'lji(ef- f e-h) = 0. Similarly one shows that 'lji(he-eh-2e) = 0

and 'lji(hf - f h + 21) = 0. This means that 'lji(K) = 0 and hence 'ljJ factors
through R/ K ~ U(g). Denote by lj5 the induced homomorphism from U(g)
to A. Then VJ = lj5 o E follows from the definitions.
The uniqueness of lj5 follows from the uniqueness of 'ljJ as the equality
VJ= lj5 o E forces formulae (2.3). D

As usual, the universal property from Theorem 2.5 guarantees that the
universal enveloping algebra is defined uniquely up to isomorphism:

Proposition 2.6 (Uniqueness of U(g)). Let U(g)' be another associa-

tive algebra such that there exists a fixed homomorphism c:' : g __, (U(g)')(-)
of Lie algebras with the universal property as described in Theorem 2. 5.
Then there exists a unique isomorphism "E : U(g)' __, U(g) such that
E o E 1 = c:; in particular, U(g)' and U(g) are canonically isomorphic.
36 Lectures on sb(IC)-modules

Proof. First we note that for A = U(g) and cp = s in (2.2) we have

= idu(g) (because idu(g) works and is unique by the universal property).
Cj5
Now take A= U(g)' and cp = s'. The universal property for U(g) gives a
homomorphism?: U(g) ----+ U(g)'. The universal propertyfor U(g)' gives a
homomorphism€: U(g)'----+ U(g) (see (2.4)). From the previous paragraph
we obtain that the composition €0? is the identity map on U(g). Similarly,
the composition? o €is the identity map on U(g)'. The claim follows.

er &U(g)' v
·. c 1 os=idu(g)'

(2.4)

.· U(g)

soc'=idu(g)
D

Having the universal property it is now easy to relate g-modules and

U(g)-modules.

Proposition 2.7.

(i) Let V be a g-module given by a homomorphism cp : g ----+ L'.(V)(-) of

Lie algebras. Then the homomorphism Cj5 : U(g) ----+ L'.(V), given by
the universal property, endows V with the canonical structure of a
U (g )-module.
(ii) Let V be a U(g)-module given by the homomorphism 1jJ : U(g) ----+
L'.(V). Then the composition 1jJ o s is a Lie algebra homomorphism
from g to L'.(V) (- ) , which endows V with the canonical structure of a
g-module.
(iii) Let V and W be two g-modules with the induced structures of U(g)-
modules given by (i). Then Hom 9 (V, W) = Homu(g)(V, W).
(iv) Let V and W be two U(g)-modules with the induced structures of g-
modules given by (ii). Then Homu(g)(V, W) = Hom 9 (V, W).
(v) The operations described in (i) and (ii) are mutually inverse to one
another.

Exercise 2.8. Let A and B be two associative algebras and 1jJ : A ----+ B
be a homomorphism of algebras. Check that 1jJ : A(-) ----+ B(-) is also a
homomorphism of Lie algebras.
 The universal enveloping algebra 37

Proof. The statement (i) is obvious, as well as a large part of the state-
ment (ii). The only thing in (ii) that one must check is that 1jJ o E is a
Lie algebra homomorphism. However, this follows from Exercise 2.8. The
statements (iii) and (iv) follow from the fact that for any g-module and the
associated U(g)-module V, the image of gin £(V)C-) is generated by the
same elements as the image of U(g) in L.:(V). The statement (v) follows
from the definitions and the uniqueness of E (Lemma 2.4). D

Denote by g-mod the category of all (left) g-modules, and by U(g)-mod

the category of all (left) U(g)-modules. From Proposition 2.7 we immedi-
ately obtain:

Corollary 2.9. Operations described in Proposition 2. 7(i) and Proposi-

tion 2. 7(ii) give mutually inverse isomorphisms between the categories
g-mod and U(g)-mod.

Remark 2.10. From now on we will call every g-module, if necessary,

a U(g)-module, and vice versa. Moreover, if V is a g-module, v E V and
u E U (g), we will usually denote the action of u on v by u( v) or, simply, uv.
In particular, we always have e(v) = E(v), f(v) = F(v) and h(v) = H(v).

Exercise 2.11. Let V be a finite-dimensional g-module. Then V is also

a U(g)-module via the construction from Proposition 2.7(i). Denote by
1/Jv : U(g) --+ £(V) the corresponding homomorphism. Show that the
Casimir element Cv, considered in Section 1.3, is the image of the central
element (h + 1) 2 + 4fe E U(g) under 1/Jv.

Exercise 2.12. Let a be a one-dimensional Lie algebra with basis x. Show

that the polynomial algebra C[x] is the universal enveloping algebra of a
in the sense that C[x] together with the homomorphism E: a--+ C[x]C-l,
defined via E(x) = x, have the universal property, analogous to that from
Theorem 2.5. Prove also analogs of Proposition 2.6 and Proposition 2.7 in
this situation.

2.2 Poincare-Birkhoff-Witt Theorem

Corollary 2.9 says that the universal enveloping algebra U(g), defined in
the previous section, is extremely important for the study of g-modules.
However, the definition of U(g) does not give us much information about
this algebra. It is not at all clear, for example, whether U(g) is finite-
38 Lectures on sb(rc)-modules

dimensional or infinite-dimensional (or even whether it is non-trivial). Nor

is it clear at this stage, as mentioned in the previous section, whether the
canonical map E is injective. In this section we will clarify all this by giving
a very explicit basis of U(g).

Theorem 2.13 (Poincare-Birkhoff-Witt). The set

· · k Eno
{f ihj e k : i,J, 1'-T }

is a basis of U(g).

Theorem 2.13 is usually called the PEW Theorem. The monomials of

the form f i h1 ek are usually called standard monomials. Note that standard
monomials also form a basis of the polynomial algebra C[f, h, e]. Hence the
PEW Theorem says that the non-commutative algebra U(g) is "of the same
size" as the commutative algebra C[J, h, e]. To prove the PEW Theorem
we will need some preparation.

Lemma 2.14. Standard monomials generate U(g).

Proof. The standard basis of the free algebra R with generators e, f

and h is given by arbitrary monomials x 1 x 2 · · · Xk, where k E N0 and
Xi E { e, f, h} for all i = 1, ... , k. Hence to prove the statement of our
lemma we just have to check that each x 1 x 2 · · · Xk can be written as a
linear combination of standard monomials.
We proceed by induction on k. For k = 1 the statement is obvious. Let
us now prove the induction step. For k > 1 consider some x 1 x 2 · · · Xk as
above. A pair (i,j), 1 :::; i < j :::; k, will be called an inversion provided
that one of the following holds:
Xi =hand Xj = f; Xi= e and x 1 = f; Xi = e and x 1 = h.
We proceed by induction on the number of inversions in x 1 x 2 · · · Xk· If
there are no inversions, the monomial x 1 x 2 · · · Xk is standard and we have
nothing to prove. Otherwise, we can fix some inversion (i, i + 1). We have

(2.1)
= X1 ... Xi-IXi+IXiXi+2 ... Xk + X1 ... Xi-I [xi, Xi+1lxi+2 ... Xk·
As [xi, xi+1] E {±h, ±2e, ±2f}, the second summand is a linear combina-
tion of monomials of degree k - 1 and hence is dealt with by induction on
k. The first summand, in turn, has one inversion less than x 1 x 2 · · · Xk and
hence is dealt with by induction on the number of inversions. Thus these
two inductions complete the proof. D
 The universal enveloping algebra 39

Consider the vector space V = C[a, b, c]. Define, using the induction on
the degree of a monomial, the following linear operators F, H and E on V:
(2.5)

(2.6)

i,j = 0,
i = O,j -1- 0, (2.7)
i -1- 0.
where i,j, k E No.

Exercise 2.15. Check that the formulae (2.5)-(2. 7) do give well-defined

linear operators on C[a, b, c].

Exercise 2.16. Check that the formulae (2.5)-(2.7) can be rewritten as

follows:
F(aibj ck) = ai+ 1 ~ ck;
j+l k
i = 0,
H(aibjck) = { b c '
F(H(ai-lbjck)) + [H, F]ai-lbjck, i -I- O;

{~;~(bj-lck)) + [E,
i,j = 0,
E(aibjck) = H](bj-lck), i = O,j -I- 0,
F(E(ai-l~ck)) + [E, F](ai-l~ck), i -1- 0.

Lemma 2.17. The formulae (2.5)-(2.7) define on V the structure of a

g-module.

Proof. We have to check the three relations from (1.2). Let us start with
the relation [H, F] = -2F. For i,j, k E No we have
H(F(aibjck)) (~) H(ai+ 1 ~ck)
(by (2.6)) = F(H(aibjck)) - 2ai+ 1 ~ck
(by (2.5)) = F(H(aibjck)) - 2F(ai~ck)
and the relation [H, F] = -2F is proved.
The relation [E, F] =His proved using the following computation: for
i, j, k E N0 we have
E(F(aibjck)) (~) E(ai+ 1 bjck)
(by (2.7)) = F(E(aibjck)) + H(aibjck).
40 Lectures on s[2 ( IC)-modules

Finally, let us prove the relation [H, E] = 2E, which we first write in
the form EH - HE= -2E. For any j, k E No we have

E(H(bick)) (~) E(bi+ 1 ck)

(by (2.7)) = H(E(bick)) - 2E(bick)

and the relation [H, E] = 2E is proved on monomials of the form bi ck.

The really tricky thing is to prove this relation on monomials aibj ck, where
i EN and j, k E N0 . We do this by induction on i. The case i = 0 is already
established, so we prove the induction step.
We rewrite [H,E] = 2E asHE-EH-2E = 0. ApplyingHE-EH-2E
to aibick, where i EN and j, k E N0 , and using Exercise 2.16 we obtain

(HE - EH - 2E)(aibick)
= (HFE + H[E,F] - EFH - E[H,F] - 2FE - 2[E,F])(ai-lbjck).
(2.8)

By induction we have - 2F E = F [E, H]. Using the definition of the com-

mutator and the relation [H, F] = -2F, which we proved above, we also
have

H[E,F] = HEF- HFE,

E[H,F] = EHF- EFH,
-2[E, F] = [E, [H, F]].

This reduces the equality (2.8) to

As we have already proved that [E, F] = H, we can add the following zero
term:

0 = [H,H] = -[H,H] = [H, [F,E]]

to the equality (2.9) and obtain

(HE - EH - 2E)(aibjck)
= ([F, [E, HJ]+ [E, [H, F]] + [H, [F, E]])(ai- 1 b5ck).
The right-hand side of the latter is equal to zero because of the Jacobi
identity for .C(V)(-). This completes the proof. D
 The universal enveloping algebra 41

Now we are ready to prove the PBW Theorem 2.13.

Proof. To prove that standard monomials form a basis in U(fJ), we have
to check that they generate U(fJ) and that they are linearly independent.
The fact that they generate U(fJ) was proved in Lemma 2.14.
To prove that standard monomials are linearly independent, consider
the U(fJ)-module V from Lemma 2.17. Note that for all i,j,k E No for the
constant polynomial 1 E V we have
FiHj Ek(l) = ait?ck.
Now it is left to observe that the elements aibjck EV are linearly indepen-
dent. Hence the linear operators Fi Hj Ek are also linearly independent.
Since these linear operators are exactly the images of standard monomi-
als under the homomorphism defining the U(fJ)-module structure on V, we
conclude that standard monomials are linearly independent as well. This
completes the proof. D

Exercise 2.18. Let x 1 , x 2 and x 3 be the elements e, f and h written in

some order. Show that the standard monomials xtx~x~, i,j,k E No, also
form a basis of U (fJ).

Corollary 2.19. The canonical embedding F: of fJ into U(fJ)(-) is injective.

Proof. This follows from the fact that the elements e, f and h form a
basis of fJ and the fact that the elements c( e) = e, c( f) = f and c(h) = h
are linearly independent in U(fJ) by Theorem 2.13. D
After Corollary 2.19 it is natural to identify fJ with c(fJ).

Remark 2.20. There exists an alternative (and somewhat easier) argu-

ment for Corollary 2.19. The elements e, f and h, which form a basis of
fJ, act linearly independently on the natural module (since it is given by
the identity map). From Proposition 2. 7 we have that this action coincides
with the action of s(e) = e, c(f) = f and s(h) = h. Hence the latter
elements must be linearly independent in U(fJ) and thus the map E must
be injective.

2.3 Filtration on U(g) and the associated graded algebra

As usual, for a monomial x 1 x 2 · · · Xk E U(fJ) (where Xi E {f, h, e} for all

i) the number k is called the degree of the monomial. The degree of the
42 Lectures on .s[2(C)-modules

monomial u is usually denoted by deg(u). For i E N0 denote by U(g)Cil the

linear span of all monomials of degree at most i (we also set U (g) (-l l = 0).
This gives us the following filtration on U(g):

U(g) = LJ U(g)Cil.
iENo

Note that U(g)C 0 l = <C. We obviously have

U(g)(ilU(g)(j) c U(g)(i+j), (2.10)

which means that U(g) is a filtered algebra.

Exercise 2.21. Let k E N0 . Show that standard monomials of degree at

most k form a basis of U(g)Ck).

Lemma 2.22. Let i, j E N 0 , u E U(g)Ci) and v E U(g)Ul. Then [u, v] E

U(g)Ci+j-1).

Proof. By Exercise 2.21 the space U(g)Ck), k E N0 , has a basis con-

sisting of standard monomials. Hence, by the linearity of [-, - l, it is
enough to prove the statement in the case when both u and v are stan-
dard monomials. We prove the statement by induction on deg( u) +deg( v).
If deg(u) + deg(v) ~ 1, then at least one of u and v must be a scalar and
the statement becomes obvious. If both u, v E g, then [u, v] E g as well
and the statement is true again. Now, to prove the induction step we may
assume that u tf_ U(g)C 1l and write u = xu', where x E {f, h, e} and u' is a
standard monomial. Using the definition of the commutator we have

[u,v] = uv - vu
= xu'v -vxu'
= xvu' + x[u', v] - vxu'
= vxu' + [x, v]u' + x[u', v] - vxu'
= [x, v]u' + x[u', v].
By induction we have [u', v] E U(g)Ci+j- 2 J and hence x[u', v] E U(g)(i+j-l).
As u tf_ U(g)(ll by our assumption, u' is not a scalar monomial and thus
x E U(g)(ll c U(g)Ci-l). Hence, applying the inductive assumption to [x, v]
we get [x, v]u' E U(g)CH 1- 1lu 1 c U(g)(i+j-l) and the claim follows. 0

For i E No put
Discovering Diverse Content Through
Random Scribd Documents
 Features of the hypothetical morphotype, that would permit
derivation of the modern genera would include the following: (1)
Skull generalized, neither excessively long and narrow or short and
broad; (2) skull smoothly rounded, without pronounced angularity,
rugosity or cresting (sagittal crest probably lacking, even in old
individuals); (3) zygomata slender, without lateral platelike
expansions; (4) rostrum moderately broad; (5) upper incisors
bisulcate, two grooves in pattern found in Pliogeomys, Zygogeomys
and Geomys; (6) lateral re-entrant angles of premolars obtuse; (7)
p4 having four enamel plates (one on anterior wall, one on posterior
wall, and two lateral plates) and lower molars having one enamel
plate on the posterior wall of tooth (anterior plate is lacking); (8) P4
having four enamel plates, in same pattern as described for p4, M1
having two enamel plates (one anterior and one posterior), M2 same
as M1, M3 having three plates (one anterior, two lateral on sides of
posterior loph, none posterior); (9) M3 subtriangular in cross-
section, distinctly bicolumnar, two columns marked by shallow re-
entrant folds and connected by broad isthmus; (10) masseteric ridge
large, forming high crest bordering masseteric fossa; (11)
basitemporal fossa shallow; (12) angular process of mandible short,
its lateral projection barely exceeding that of zygomatic arch.

Specializations in Genera
In relation to the primitive morphotype, increase in size,
simplification of dentition, and changes in shape of skull are
regarded as specializations. Considerable parallelism between the
four lineages is seen. But each lineage is distinguished by a
combination of specialized features, and three by a few unique
specializations.
Among trends resulting in simplification of the dentition,
reduction of enamel on the posterior wall of the upper cheek teeth
has occurred in various degrees in all lineages of the Geomyini even
to loss of all enamel on the posterior wall of the premolars and
molars in two genera. Loss of some enamel is more common on P4
than on M1-2, and has occurred in all genera (see Figs. 7 and 9.)
In evolutionary sequence loss of enamel from M1 and M2 usually
occurs after, but never preceding, the reduction of enamel on P4.
Loss of enamel plates from the posterior face of M1 and M2 is
associated with the evolution of an efficient anterotransverse
shearing action of the teeth.
On the anterior wall of those teeth no reduction of the cutting
blade has been observed; a complete anterior plate is retained in all
living Geomyini.
Presence of both the posterior and anterior plates decreases the
efficiency of transverse shearing, by providing two upper plates
(anterior plate of one tooth and posterior plate of the preceding
tooth) over which the lower cutting blade simultaneously must pass
with each movement. The advantages of shearing over the more
common mechanics of planing are largely lost unless the posterior
plates are eliminated. Also, none of the living Geomyini have
retained a definitive posterior enamel plate on M3, the last upper
molar; but two well-developed lateral plates, that extend almost all
of the way back to the posterior apex of M3, have been retained,
and, together function as a posterior plate. Loss of either or both of
the lateral plates of M3 is rare, and occurs only in old individuals.
Their loss in the final stages of wear may represent the beginning of
a new trend in those species where it occurs (the castanops-group
of the subgenus Cratogeomys). In any case, reduction of enamel
takes place by transverse shortening of the plate through the
complete loss of enamel on one end, the diminution beginning first
on the labial end and proceeding by progressive atrophy to the
lingual end of the plate. Evidently, when enamel has been eliminated
from the labial end of a plate, the rate of loss decreases markedly,
and the last stages of evolution, terminating in complete loss of an
enamel plate, occurs more slowly. Evolution may be arrested before
complete loss has occurred, and that part of the enamel that
remains forms a short, vestigial plate restricted to the lingual one-
fourth or one-third of the wall. The enamel pattern of the lower
dentition is the same in all of the diverging lineages, with no
evidence of additional loss of enamel from that which had already
occurred in their common ancestor (see Figs. 7 and 9). Reduction
and loss of enamel plates began and was terminated in the lower
dentition before reduction began in the upper dentition.
Other dental specializations have occurred in the shape of the
third upper molar and in the pattern of grooving in the upper incisor.
Unlike M3 of the Thomomyini, that of the Geomyini differs in shape
from M2, and its enamel investment differs from that of M2.
Primitively, M3 was probably subtriangular in cross-section, and the
posterior loph evidently projected posteriorly as a short, rudimentary
heel that formed the apex of the triangle. Other shapes of M3 are
considered to be specializations that have been derived from the
primitive form. In addition to the primitive subtriangular pattern, the
M3 of living Geomyini may be suborbicular, quadriform, elongate, or
obcordate in shape. Usually each lineage is characterized by only
one pattern, but in one genus (Pappogeomys) all patterns occur. Of
the different forms, the elongate and obcordate seem to be the most
highly specialized deviations from the triangular-shaped tooth. The
bicolumnar pattern is accentuated in the elongate type (Fig. 7D, F,
H) by deep lateral re-entrant folds, on both the lingual and labial
sides, and by the elongation of the posterior loph into a pronounced
heel. Teeth having this pattern have been illustrated by Merriam
(1895:76-82) in Figures 27 (6 and 7), 28 (c and d), 34 (7 through
15), and 35 (8).
 Fig. 7. Molariform dentitions of the Tribe Geomyini. Drawings
illustrating enamel patterns characteristic of Pliogeomys,
Zygogeomys, and the subgenera of Orthogeomys (Orthogeomys,
Heterogeomys and Macrogeomys). × 5.

A. Pliogeomys buisi, No. 29157 (UMMP), holotype, Buis

Ranch (Upper Middle Pliocene), Beaver Co.,
 Oklahoma. Right lower, p4-m2 (m3 unknown).
B and C. Zygogeomys trichopus trichopus, adult female,
No. 51971 (FMNH), Mt. Tancítaro, 10,500 ft.,
Michoacán. Left upper (B), P4-M3; right lower (C),
p4-m3.
D and E. Subgenus Orthogeomys. Orthogeomys grandis
guerrerensis, adult female, No. 39807 (KU), 1/2
mi. E La Mira, 300 ft., Michoacán. Left upper (D),
P4-M3; right lower (E), p4-m3.
F and G. Subgenus Heterogeomys. Orthogeomys
hispidus hispidus, adult female, No. 23975 (KU), 4
km. W Tlapacoyan, 700 ft., Veracruz. Left upper
(F), P4-M3; right lower (G), p4-m3.
H and I. Subgenus Macrogeomys. Orthogeomys
heterodus cartagoensis, adult female, No. 60664
(KU), Rancho Redando, Volcán Lrozá, Prov. San
José, Costa Rica. Left upper (H), P4-M3; right
lower (I), p4-m3.

The subcordate form is characterized by pronounced

anteroposterior compression, and retention of a distinct labial re-
entrant fold. The posterior loph apparently has been rotated in such
a way that what was previously its posterior border now lies on the
outer margin of the tooth; therefore, the axis of the posterior loph is
strongly oblique in relation to the anteroposterior bearing of the
maxillary tooth-row, and the median enamel plate also has been
rotated and so lies transversely across the posterior wall of the
tooth. Owing to the rotation of the posterior loph, the apex of the
obcordate tooth is at its lingual side. The subcordate type is
illustrated by Merriam (loc. cit.) in Figures 27 (3 and 4), 28 (a and
b), 34 (3 and 4), and 35 (5, 6, and 7). The suborbicular and
quadriform types are less specialized than the two described above.
Both are characterized by reduction, often obliteration, of the
bicolumnar pattern of the subtriangular ancestral form, especially
marked by the decrease in depth of the lateral re-entrant folds and
the decrease in length of the posterior projection of the posterior
loph. With these changes, the tooth becomes essentially
monocolumnar, its occlusal surface oval in outline in one and
squarish in shape in the other. Occlusal views of the suborbicular
form are presented by Merriam (loc. cit.) in Figure 33 (1, 5, 6, 7, 11,
and 12) and the quadriform tooth is depicted in Figure 29. Grooved
upper incisors are characteristic of the living Geomyini, but variation
occurs in the number of grooves, and, if only one groove is present,
its position on the anterior face of the tooth varies. Except for the
previously mentioned (p. 480) abnormal tooth having three grooves,
incisors with no more than two grooves are found in these pocket
gophers, and this number of grooves is taken to be primitive. Loss of
one or the other of the two grooves of the bisulcate pattern,
therefore, is regarded as specialization. However, complete loss of
both grooves never occurs in the Geomyini. Each of the four major
lineages is characterized by one of the three patterns of grooving,
and the particular groove-pattern is remarkably stable in each group.
Shape of skull varies from dolichocephalic to platycephalic. The
morphology of each has been described in foregoing accounts. The
dolichocephalic skull is highly specialized for planing, a grinding
action of the teeth; whereas, the platycephalic skull is highly
specialized for shearing, a slicing action of the teeth. Of course,
concomitant specializations of the dentition, as described above, are
closely associated with both specialized trends in the skull. Most
kinds of living Geomyini have generalized skulls that show no
tendency toward either of the specialized conditions.
Increase in size of body and skull is seen in most Pleistocene
lineages of the Geomyini. Judging from the smallness of the skull in
late Pliocene species, representing the base of three of these
lineages, the ancestral species of the living assemblage were no
larger than the living species of the subgenus Pappogeomys or the
smaller subspecies of Geomys bursarius. The recorded range of
variation in condylobasal length is 36.1 to 45.5 in Pappogeomys
bulleri, including both adult males and females. Probably the skulls
of the ancestral species were not significantly larger. Maximum
dimensions of males in living species are 74.5 (subgenus
Cratogeomys) and 75.0 (subgenus Orthogeomys). These are more
than twice the minima observed in Pappogeomys bulleri.
 Zygogeomys
This is the least specialized and most primitive of the four
lineages, has a generalized type of skull, two grooves on the anterior
face of each upper incisor, an enamel plate on the posterior wall of
P4, open or divergent lateral re-entrant angles on the premolars, and
a bicolumnar and elongated M3. All of these features are primitive
and essentially as in the ancestral morphotype. No other modern
genus retains so much of the primitive structure. Phyletic trends in
Zygogeomys are not well documented in the fossil record; and only a
few fossils are known and they are fragmentary as discussed before.
The genus is represented in the late Pliocene (Z. minor), middle
Pleistocene (Z. persimilis), and Recent (Z. trichopus). The living
species is a relict population in the mountains of Central México.
Judging from the known material, the phyletic trends in the genus
have been increase in size, reduction of enamel on the posterior face
of P4 (occurring only in the living species) where a short enamel
plate is retained on the lingual side of the tooth (see Fig. 7B), loss of
the outer fourth of the enamel blade on the posterior wall of M1 and
M2 (also occurring only in the living species), development of a more
pronounced heel on the M3 by progressive elongation of the
posterior loph, reduction in size of the jugal and its displacement
ventrally, which allows the maxillary and squamosal bones to meet
along the dorsal border of the zygomatic arch. The last specialization
is seen in at least one taxon of Orthogeomys (Orthogeomys cherriei
costaricensis). In my opinion, too much weight has been given to
this feature in past classifications. Reduction of enamel in the upper
dentition evidently occurred in the late Pleistocene, since the
posterior plates on the upper cheek teeth were complete in
specimens from the middle Pleistocene (Z. persimilis).

Geomys
Geomys, slightly more specialized than Zygogeomys, must also
be regarded as one of the most primitive of the living genera.
Primitive features that have been retained are the generalized type
of skull, the bisulcate pattern of grooves on the upper incisor, and
the retention of enamel plates on both the anterior and posterior
walls of M1 and M2 (see Fig. 9A). All of these primitive features are
shared with Zygogeomys. In addition, three other trends, or
specializations, in evolution characterize the phyletic development of
Geomys. One major trend is toward loss of the enamel plate from
the posterior wall of P4. No trace of enamel remains on the posterior
wall of this tooth in late Pleistocene or Recent species of Geomys,
and at least one of the earlier species (quinni) was also
characterized by loss of this enamel plate. Secondly, M3 retains only
a vestige of the primitive bicolumnar pattern after the initial stages
of wear. In most Recent specimens, especially of the species G.
bursarius, the lateral re-entrant fold and the heel of M3 are small,
and the re-entrant inflection is hardly evident. The lateral fold is
more frequently well-developed in Irvingtonian species than in living
species (White and Downs, 1961:13), illustrating progressive loss of
the bicolumnar pattern in Pleistocene evolution. A third trend
involves the modification of the lateral folds of the premolars.
Primitively the angles of these folds are broadly open or divergently
V-shaped, and some of the earliest species of Geomys, for example
G. quinni, have retained this feature throughout life. Nevertheless,
the main trend is toward progressive compression of the folds
resulting in their walls being more nearly perpendicular, and parallel,
to the long axis of the tooth. Obtuse re-entrant angles persist in
premolars of young individuals of Irvingtonian species, but the adults
are characterized by well-compressed folds, as in Recent species.
Remains of Geomys are abundant, especially from Pleistocene
deposits of the Great Plains, but in most instances specific
assignment is difficult or impossible since only isolated teeth or
fragments of skulls have been preserved. Estimates of phyletic
relationships of the known species of Geomys are depicted in Figure
8; those estimates are useful in discussing the phyletic development
of the genus. One of the earliest known species, Geomys quinni,
ranges from Upper Pliocene to the later stages of the Lower
Pleistocene (Aftonian interglacial deposits). The dentition of G.
quinni is essentially the same as in the living species except that
open lateral re-entrant angles are retained in the premolars. Geomys
paenebursarius, also of the early Pleistocene, is a smaller species
and seems to be more directly in the line of evolution of the modern
species. As yet unnamed smaller species of Geomys from the
Rexroad fauna (late Pliocene) and Saunders fauna (latest Aftonian)
may also be on the main line of evolution. Surprisingly, Geomys
tobinensis and Geomys garbanii of later Irvingtonian provincial age
are less specialized than either Geomys quinni or Geomys
paenebursarius. It is likely that G. tobinensis and the unnamed
species from the Dixon are closer to the main line of descent than G.
paenebursarius suggesting that the direct ancestral lineage of the
living species of Geomys was more conservative and less specialized
than Geomys paenebursarius of the Lower Pleistocene. Geomys
quinni and G. paenebursarius seem to have acquired specialized
dental features in the early Pleistocene. Geomys quinni was
successful on the Great Plains, and persisted into the late Blancan.
The main line may be represented in the early Pleistocene by
Geomys paenebursarius from the Hancock formation of the Texas
Trans-Pecos. The structure of G. paenebursarius indicates that it is in
or close to the main line of descent, and probably evolved from one
of the more primitive late Pliocene species of Geomys from the
Rexroad fauna.
 Fig. 8. Tentative arrangement of species of the genus Geomys,
depicting phylogenetic trends and probable relationships within the
genus.

Isolated teeth, to which the name Geomys bisulcatus probably

applies, from Illinoian deposits on the Great Plains, show that the
dentition characteristic of the living Geomys had been developed by
that time. Actually, the Illinoian material is too fragmentary to show
clearly its taxonomic or phyletic affinities with the species of the later
Pleistocene. Even so, the two main stocks of living Geomys, G.
bursarius and G. pinetis, had certainly been differentiated by
Sangamon time. The other living species evidently evolved from one
or the other of these two stocks in a period of isolation from the
main population, probably in either the Wisconsin or post-Wisconsin.
For example, Geomys arenarius clearly differentiated from
populations of Geomys bursarius that were isolated by the eastward
retreat of the main population from the southwestern United States
as that region became more arid in the post-Wisconsin.
In review, it seems that the Recent species, represented basically
by bursarius and pinetis, evolved from Illinoian species (Geomys
bisulcatus?), which descended in turn from the more primitive
species of the early Pleistocene, possibly Geomys paenebursarius or
possibly from descendants of the Saunders species. Actually the
Saunders species may prove to be Geomys paenebursarius. At any
rate, three trends that took place during the Pleistocene stage of
evolution, in the direction of the modern species, were an increase
in size, progressive loss of the posterior enamel plate on P4, and a
decrease in the vertical depth of the enamel cap as a result of which
the dentine is reached in the initial phases of attrition on the tooth of
a juvenile. Geomys garbanii, occurring at the periphery of the range
of the genus, is regarded as a sterile offshoot of the primitive
tobinensis-line of evolution.

Orthogeomys
This is one of the more specialized genera of the Geomyini. Save
for one record in the late Pleistocene (Orthogeomys onerosus), there
is no fossil history of the genus upon which to reconstruct its
phylogeny; therefore, its phyletic development must be estimated by
comparing it and the primitive morphotype of the tribe. Results of
that comparison suggest that Orthogeomys has closer affinities with
Zygogeomys than with any of the other genera, and that
Orthogeomys may have originated in an early dichotomy of primitive
Zygogeomys stock instead of descending from the ancestral stock of
the tribe. Except for the unisulcate incisors and the longer posterior
loph on the third upper molars, the teeth of the two genera do not
differ significantly. As in Zygogeomys, the enamel blade on the
posterior wall of P4 has been reduced to a short plate restricted to
the lingual third of the tooth (see Fig. 7F and H). In Orthogeomys,
the trend in reduction of enamel is carried to its extreme only in the
subgenus Orthogeomys, where this plate has been completely lost in
most taxa (see Fig. 7D). The most significant trends in
Orthogeomys, and the principal basis for recognizing the genus, are
the dolichocephalic specializations of the skull, as described
elsewhere, and the adaptive traits that have equipped the genus for
living in tropical environments. The dolichocephalic features are
more sharply defined in the subgenera Orthogeomys and
Macrogeomys, and are less developed in the subgenus
Heterogeomys. Aside from the general dolichocephalic
specializations, trends in Orthogeomys include: Increase in size; loss
of the median one of the two grooves on the anterior face of the
upper incisor in the ancestral stock; increase in the anteroposterior
length of each of the cheek teeth, as well as the aforementioned
elongation of the posterior loph of M3; compression of the lateral
angles of the premolars; and the remarkable increase in the size of
the rostrum.

Pappogeomys
The genus Pappogeomys, as it is conceived of in this study, is
comprised of two subgenera; one, Pappogeomys, is generalized and
primitive, and the other, Cratogeomys, is specialized, and includes
the most highly specialized of the modern pocket gophers. The
subgenus Pappogeomys is regarded as the ancestral lineage, and
the subgenus Cratogeomys is regarded as an early offshoot,
probably in the early Pleistocene, that became progressively more
specialized in the course of its subsequent evolution. In the same
period of time, the subgenus Pappogeomys changed little. It is
known only from late Pliocene fragments and from the living species.
The ancestral morphotype is preserved in Pappogeomys. Primitive
characters are: (1) Small size; (2) skull generalized and smoothly
rounded; (3) temporal ridges separate (not uniting into a sagittal
crest); (4) enamel plates retained on both anterior and posterior
walls of M1 and M2; (5) M3 bilophate, its posterior loph short. Basic
specializations are few and include loss of the inner groove from the
anterior face of the upper incisor; anteroposterior compression of
the lateral re-entrant folds of the premolars; and loss of enamel from
the posterior wall of P4. All three features have been perpetuated in
the advanced subgenus Cratogeomys, suggesting that they were
already developed in the early evolution of the subgenus
Pappogeomys before Cratogeomys diverged. Agreement with
Geomys is demonstrated by the lack of enamel on the posterior wall
of P4 (see Fig. 9) and by retention of the posterior enamel plate on
M1 and M2. In Pappogeomys (Pappogeomys) alcorni the enamel
from the posterior face of M1 has been lost from all but the lingual
fourth or so of the posterior wall (Fig. 9E). Reduction of enamel in
M1 provides an example of parallelism with the more advanced
subgenus Cratogeomys, discussed below.
There is no record as yet of the early evolution of the subgenus
Cratogeomys. The features that characterize the subgenus were
already well developed in the first known fossils which are from
Wisconsin deposits of the late Pleistocene. Cratogeomys is not a
homogenous assemblage; instead it is composed of two groups of
living species, the generalized castanops group and the specialized
gymnurus group. The castanops group may be survivors of the
ancestral lineage that diverged in two different stages in the phyletic
development of the main line. Even so, the castanops group has
acquired its peculiar specializations. Indeed, P. merriami of the
castanops group differs from the hypothetical stem more than does
P. castanops. Judging from the structure of the living species of the
subgenus Cratogeomys and from the primitive subgenus
Pappogeomys, the subgenus Cratogeomys featured five major
trends: (1) Increase in size; (2) formation of sagittal crest by union
of the temporal impressions; (3) increase in rugosity and angularity
of the skull; (4) progressive development of platycephalic
specializations, including the elongation of the angular process of the
mandible; (5) complete loss of enamel plates from the posterior wall
of M1 and M2. Each trend is thought to be adaptive.

Fig. 9. Molariform dentitions of the Tribe Geomyini. Drawings

illustrating enamel patterns characteristic of Geomys and
 Pappogeomys (including the subgenera Pappogeomys and
Cratogeomys). × 5.

A and B. Geomys bursarius bursarius, adult female, No.

46275 (KU), Elk River, Sherborne Co., Minnesota.
Left upper (A), P4-M3; right lower (B), p4-m3.
C and D. Subgenus Pappogeomys. Pappogeomys bulleri
albinasus, adult female, No. 31002 (KU), W side
La Venta, 13 mi. W and 4 mi. N Guadalajara,
Jalisco. Left upper (C), P4-M3; right lower (D), p4-
m3.
E and F. Subgenus Pappogeomys. Pappogeomys alcorni,
adult female, No. 31051 (KU), holotype, 4 mi. W
Mazamitla, 6600 ft., Jalisco. Left upper (E), P4-M3;
right lower (F), p4-m3.
G and H. Subgenus Cratogeomys. Pappogeomys
gymnurus tellus, adult female, No. 31051 (KU), 1
mi. NE Tala, 4400 ft., Jalisco. Left upper (G), P4-
M3; right lower (H), p4-m3.

Loss of enamel is a trend common to all living genera of the tribe

Geomyini, but the greatest loss has occurred in Cratogeomys. It has
lost the plates on the posterior walls of M1 and M2 (Fig. 9G). If the
lateral plates of M3 are considered as one functional plate and the
lateral plates on either side of P4 together as two transverse plates,
then, the transverse cutting blades in Cratogeomys number seven in
the upper and seven in the lower cheek teeth compared with 10 in
the upper and seven in the lower in the primitive morphotype.
Indeed, in some species of the subgenus, one or both of the lateral
plates on M3 is also lost, usually in old age, resulting in even greater
reduction of enamel. Loss of enamel from the posterior walls of the
upper molars may be associated with changes in the mechanics of
mastication from anteroposterior planing to anterotransverse
shearing, as discussed elsewhere. Merriam (1895:95-96) argues
convincingly that the posterior cutting blades of the upper molars
would hinder efficient shearing action of the teeth; hence, selection
would favor their reduction and eventual loss. Changes in the shape
of the skull also seem to be correlated with the shift from a planing
to a shearing type of mastication. More efficient shearing action,
which depends upon lateral movement of the jaw, can be developed
if the functional muscles insert farther laterally than is possible in the
generalized type of skull. Therefore, platycephalic specializations
involved lateral expansion of the braincase and mandible.
Pronounced lateral expansion has been developed only in the
gymnurus group of species, suggesting that the dental
specializations evolved earlier in the evolution of the subgenus than
did the platycephalic specializations of the skull, and that the
castanops group separated from the gymnurus group before the
common ancestor had developed the more extreme trends in
platycephaly. It is interesting to note that the subtriangular M3 (Fig.
9G) postulated for the ancestral morphotype and that characterizes
the subgenus Pappogeomys is retained also in the gymnurus group.

LITERATURE CITED
Alston, E. R.
1876. On the classification of the order Glires. Proc. Zool.
Soc. London, 1876:61-98, 1 pl., June.
Alvarez, T.
1964. Nota sobre restos oseos de mamiferos del Reciente,
encontrados cerca de Tepeapulco, Hidalgo, Mexico.
Publ. Inst. Nac. Antro, e Hist., 15:1-15.
1965. Catálago Paleomastozoológico Mexicano. Publ. Inst.
Nac. Antro, e Hist., 17:1-70.
Axlerod, D. I.
1950. Studies in late Tertiary paleobotany. Carnegie Inst.
Washington Publ., 590:1-322.
 1958. Evolution of the Madro-tertiary geoflora. Bot. Rev.,
24:433-509.
Bader, R. S., and Techter, D.
1959. A list and bibliography of the fossil mammals of Illinois.
Nat. Hist. Misc., Chicago Acad. Sci., 172:1-8, October
30.
Baird, S. F.
1858. Mammals. Part I. General report upon the zoology of
the several Pacific railroad routes. Repts., explorations
and surveys for a railroad route from the Mississippi
River to the Pacific Ocean. Washington, D. C., vol. 8,
pp. xlviii + 757, 35 figs., 43 pls., July 14.
Barbour, H., and Schultz, C. B.
1937. An early Pleistocene fauna from Nebraska. Amer. Mus.
Nat. Hist. Novit., 942:1-10, September 10.
Berry, E. W.
1937. Tertiary floras of North America. Bot. Rev., 3:31-46.
Black, C. C.
1961. Rodents and lagomorphs from the Miocene Fort Logan
and Deep River Formations of Montana. Postilla, Yale
Peabody Museum, 48:1-20, 6 figs., January 16.
Brandt, J. F.
1855. Beiträge zur mähern Kenntiss der Säugethiere
Russlands. Acad. Imp. des Sciences St. Petersburg,
7:79-336, 11 pls.
Brown, B.
 1908. The Conard fissure, a Pleistocene bone deposit in
northern Arkansas with description of two genera & 20
new species of mammals. Mem. Amer. Mus. Nat. Hist.,
9:157-208, 2 pls., 3 figs.
1912. Brachyostracon, a new genus of glyptodont from
Mexico. Bull. Amer. Mus. Nat. Hist., 31:167-177, August
2.
Chaney, R. W.
1947. Tertiary centers and migration routes. Ecol. Monog.,
17:139-148.
Cook, H. J., and Cook, M. C.
1933. Faunal lists of the Tertiary Vertebrata of Nebraska and
adjacent areas. Nebraska Geol. Surv., Paper No. 5:49.
Cope, E. D.
1878. Description of new extinct Vertebrata from the upper
Tertiary and Dakota formations. Bull. U. S. Geol. and
Geog. Surv., 4:379-396.
1884. The White River and John Day faunae, pp. 759-1002,
pl. 64, figs. 1-9. In The Vertebrata of the Tertiary
formations of the west. Book I. Part second. Rept. U.
S. Geol. Surv. Terr. (F. V. Hayden), vol. 3.
1889. The vertebrate fauna of the Equus beds. Amer. Nat.,
23:160-165.
Coues, E.
1877. Monographs of North American Rodentia. No. 8,
Saccomyinae, pp. 483-542, and No. 10, Geomyinae,
pp. 601-629, pl. 7, August.
Dalquest, W. W.
 1962a. A record of the giant bison (Bison latifrons) from
Cooke County, Texas. Texas Jour. Sci., 13:41-44, March.
1962b. The Good Creek Formation of Texas, and its fauna.
Jour. Paleont., 36:568-582.
1964. A new Pleistocene local fauna from Motley County,
Texas. Trans. Kansas Acad. Sci., 67:499-505, 4 figs.,
December 11.
1965. New Pleistocene Formation and local fauna from
Hardeman County, Texas. Jour. Paleont., 39:63-72, 2
figs., January.
Dice, L. R.
1925. Rodents and lagomorphs of the Rancho La Brea
Deposits. Carnegie Inst. Washington Publ., 349:119-
130.
Davis, W. B.
1937. Variations in Townsend pocket gophers. Jour. Mamm.,
18:145-158, May 12.
Ellerman, J. R.
1940. The families and genera of living rodents. Vol. I.
Rodents other than Muridae. British Museum (Nat.
Hist.), London, pp. xxvi + 689, 189 figs., June 8.
Elliott, D. G.
1903. A list of mammals obtained by Edmund Heller from the
coast region of northern California and Oregon. Field
Columb. Mus. Publ., 76, zool. ser. vol. 3(11):175-197,
July 25.
Elftman, H. O.
 1931. Pleistocene mammals of Fossil Lake, Oregon, Amer.
Mus. Novit., 481:1-21, 10 figs., July 14.
Franzen, D. S.
1947. The pocket gopher, Geomys quinni McGrew, in the
Rexroad fauna, Blancan age, of southwestern Kansas.
Trans. Kansas Acad. Sci., 50:55-59.
Freudenberg, W.
1921. Geologie von Mexiko. Berlin, pp. viii + 232.
Galbreath, E. C.
1848. An additional specimen of the rodent Dikkomys from
the Miocene of Nebraska. Trans. Kansas Acad. Sci.,
51:316-317.
Gazin, C. L.
1935. Annotated list of Pleistocene Mammalia from American
Falls, Idaho, Jour. Washington Acad. Sci., 25:297-302.
1942. The late Cenozoic vertebrate fauna from the San Pedro
Valley, Arizona. Proc. U. S. Nat. Mus.,
92(3155):475:518, 2 pls., 9 figs.
Gervais, P.
1849. Rongeurs. In Dictionnaire universel d'historie naturelle,
Dirigé par M Ch. d'Orbigny, Paris, M. M. Renard,
Martinet et Cie, vol. 11:198-204.
Gidley, J. W.
1922. Preliminary report on fossil vertebrates of the San
Pedro Valley, Arizona, with descriptions of new species
of Rodentia and Lagomorpha. U. S. Geol. Surv. Prof.
Papers, 131:119-130, pls. 34 and 35.
Gidley, J. W., and Gazin, C. L.
1933. New Mammalia in the Pleistocene fauna from
Cumberland Cave. Jour. Mamm., 14:343-357, 9 figs.
Gill, T. H.
1872. Arrangement of the families of mammals with
analytical tables. Smithsonian Misc. Coll., 11, art. 1, pp.
vi + 98.
Gilmore, R. M.
1947. Report on a collection of mammal bones from
archeologic cave-sites in Coahuila, Mexico. Jour.
Mamm., 38:147-165, June 1.
Gray, J. E.
1868. Synopsis of the species Saccomyinae or pouched mice
in the collection of the British Museum. Proc. Zool. Soc.
London, 1868:199-206, May.
Green, M.
1956. The lower Pliocene Ogallala-Wolf Creek vertebrate
fauna, South Dakota. Jour. Paleo., 30(1):146-169, 12
figs., January.
Gut, H. J., and Ray, C. E.
1963. The Pleistocene vertebrate fauna of Reddick, Florida.
Quart. Jour. Florida Acad. Sci., 26:315-328.
Hall, E. R., and Kelson, K. R.
1959. The mammals of North America. 2 vols., xxx + 1083
pp., 553 figs., 500 maps, March 31.
Harris, A. H., and Findley, J. S.
 1964. Pleistocene—Recent fauna of the Isleta Caves,
Bernalillo County, New Mexico. Amer. Jour. Sci.,
262:114-120, January.
Hay, O. P.
1920. Description of some Pleistocene vertebrates found in
the United States. Proc. U. S. Nat. Mus., 57:83-146, 8
pls., 4 figs.
1921. Description of species of Pleistocene vertebrate types
of specimens of most of which are preserved in the U.
S. National Museum. Proc. U. S. Nat. Mus., 59:599-642.
1923. The Pleistocene of North America and its vertebrated
animals from the states east of the Mississippi River
and from the Canadian provinces east of longitude 95°.
Carnegie Inst. Washington, Publ. 322: 1-499, February
24.
1924. The Pleistocene of the middle region of North America
and its vertebrated animals. Carnegie Inst.
Washington, Publ. 322A:1-346.
1927. The Pleistocene of the western region of North
America and its vertebrate animals. Carnegie Inst.
Wash., Publ. 322B, pp. v + 346, 12 pls., 21 maps, 10
text figs.
Hibbard, C. W.
1938. An Upper Pliocene fauna from Meade County, Kansas.
Trans. Kansas Acad. Sci., 40:239-265, 5 pls., 2 figs.
1941a. The Borchers fauna, a new Pleistocene interglacial
fauna from Meade County, Kansas. Bull. State Geol.
Surv. Kansas, 38:197-220, 2 pls., July 14.
1941b. Mammals of the Rexroad Fauna from the Upper
Pliocene of southwestern Kansas. Trans. Kansas Acad.
Sci., 44:265-313, 4 pls.
1943. The Rezabek fauna, a new Pleistocene fauna from
Lincoln County, Kansas. Univ. Kansas Sci. Bull., 29:235-
247, October 15.
1944. Stratigraphy and vertebrate paleontology of
Pleistocene deposits of southwestern Kansas. Geol.
Soc. Amer. Bull., 55:707-754, 3 pls., 20 figs.
1950. Mammals of the Rexroad Formation from Fox Canyon,
Meade County, Kansas. Univ. Michigan Paleo. Contrib.,
8(6):113-192, 5 pls., 23 figs.
1951. Thomomys talpoides (Richardson) from a late
Pleistocene deposit in Kansas. Jour. Mamm., 32:229-
230.
1952. Vertebrate fossils from late Cenozoic deposits of
central Kansas. Univ. Kans. Paleont. Contrib.,
Vertebrata, 2:1-14, March 20.
1953. The Saw Rock Canyon fauna and its stratigraphic
significance. Papers Michigan Acad. Sci., Arts, and
Letters, 38:387-411.
1954. A new Pliocene vertebrate fauna from Oklahoma.
Papers Michigan Acad. Sci., Arts, Letters, 39:339-359.
1955a. Pleistocene vertebrates from the upper Bercerra
(Bercerra Superior) Formation, valley of Tecquixquiac,
Mexico, with notes on other Pleistocene forms.
Contribution from Mus. Paleo., Univ. Michigan, 12:47-
96, 9 pls., 5 figs., February 18.
1955b. The Jinglebob interglacial (Sangamon?) fauna from
Kansas and its climatic significance. Contrib. Mus.
 Paleo., Univ. Michigan, 12: 179-228, 2 pls., 8 figs.,
September 1.
1956. Vertebrate fossils from the Meade Formation of
southwestern Kansas. Papers Michigan Acad. Sci., Arts,
and Letters, 41:145-200.
1958. Summary of North American Pliestocene mammalian
local faunas. Papers of Michigan Acad. Sci., Arts, and
Letters, 43:1-32.
1963. A late Illinoian fauna from southwestern Kansas and its
climatic significance. Papers Michigan Acad. Sci., Arts
and Letters, 48:187-221, 8 figs.
Hibbard, C. W., and Keenmon, K. A.
1950. New evidence of the Lower Miocene age of the
Blacktail Deer Creek Formation in Montana. Contrib.
Univ. Michigan Mus. Paleo., 8:193-204, 3 figs., 1 map,
July 28.
Hibbard, C. W., and Mooser, O.
1963. A porcupine from the Pleistocene of Aguascalientes,
Mexico. Contrib. Mus. Paleo. Univ. Michigan, 18:245-
250, November 22.
Hibbard, C. W., Ray, D. E., Savage, D. E., Taylor, D. W., Guilday, G.
E.
1965. Quaternary mammals of North America. In Quaternary
of the United States. Eds. H. E. Wright, Jr., and D. G.
Frey. Princeton Univ. Press, Princeton, pp. x + 922.
Hibbard, C. W., and Riggs, E.
1949. Upper Pliocene vertebrates from Keef Canyon, Meade
County, Kansas. Geol. Soc. Amer. Bull., 60(5):829-860,
11 figs., 5 pls.
Hibbard, C. W., and Taylor, D. W.
1960. Two late Pleistocene faunas from southwestern
Kansas. Contrib. Univ. Michigan Mus. Paleo., 16(1):1-
223, 16 pls., 18 figs.
Hibbard, C. W., and Wilson, J. A.
1950. A new rodent from subsurface stratum in Bee County,
Texas. Jour. Paleo., 24:621-623, September.
Hooper, E. T.
1946. Two genera of pocket gophers should be congeneric.
Jour. Mamm., 27:397-399, November, 1965.
James, G. T.
1963. Paleontology and nonmarine stratigraphy of the
Cuyama Valley Badlands, California. Part I. Geology,
faunal interpretations, and systematic descriptions of
Chiroptera, Insectivora, and Rodentia. Univ. California
Publ. Geol. Sci., 45:iv + 170, 8 pls., 53 figs., June 26.
Johnson, C. S., and Savage, D. E.
1955. A survey of various Late Cenozoic vertebrate faunas of
the Panhandle of Texas, Part I. Univ. California Publ.
Geol. Sci., 31:27-50.
Kendeigh, C. S.
1961. Animal Ecology. Prentice Hall, Inc., Englewood Clifts,
New Jersey, pp. x + 468.
Kurten, B.
1965. The Pleistocene Felidae of Florida. Bull. Florida State
Mus., 9(6):215-273, June 8.
Leidy, J.
Welcome to our website – the perfect destination for book lovers and
knowledge seekers. We believe that every book holds a new world,
offering opportunities for learning, discovery, and personal growth.
That’s why we are dedicated to bringing you a diverse collection of
books, ranging from classic literature and specialized publications to
self-development guides and children's books.

More than just a book-buying platform, we strive to be a bridge

connecting you with timeless cultural and intellectual values. With an
elegant, user-friendly interface and a smart search system, you can
quickly find the books that best suit your interests. Additionally,
our special promotions and home delivery services help you save time
and fully enjoy the joy of reading.

Join us on a journey of knowledge exploration, passion nurturing, and

personal growth every day!

ebookbell.com