EMS Series of Lectures in Mathematics

Edited by Andrew Ranicki (University of Edinburgh, U.K.)

EMS Series of Lectures in Mathematics is a book series aimed at students, professional

mathematicians and scientists. It publishes polished notes arising from seminars or lecture series
in all fields of pure and applied mathematics, including the reissue of classic texts of continuing
interest. The individual volumes are intended to give a rapid and accessible introduction into
their particular subject, guiding the audience to topics of current research and the more
advanced and specialized literature.

Previously published in this series:

Katrin Wehrheim, Uhlenbeck Compactness
Torsten Ekedahl, One Semester of Elliptic Curves
Sergey V. Matveev, Lectures on Algebraic Topology
Joseph C. Várilly, An Introduction to Noncommutative Geometry
Reto Müller, Differential Harnack Inequalities and the Ricci Flow
Eustasio del Barrio, Paul Deheuvels and Sara van de Geer, Lectures on Empirical Processes
Iskander A. Taimanov, Lectures on Differential Geometry
Martin J. Mohlenkamp and María Cristina Pereyra, Wavelets, Their Friends, and What They
Can Do for You
Stanley E. Payne and Joseph A. Thas, Finite Generalized Quadrangles
Masoud Khalkhali, Basic Noncommutative Geometry
Helge Holden, Kenneth H. Karlsen, Knut-Andreas Lie and Nils Henrik Risebro, Splitting Methods
for Partial Differential Equations with Rough Solutions
Koichiro Harada, “Moonshine” of Finite Groups
Yurii A. Neretin, Lectures on Gaussian Integral Operators and Classical Groups
Damien Calaque and Carlo A. Rossi, Lectures on Duflo Isomorphisms in Lie Algebra and
Complex Geometry
Claudio Carmeli
Lauren Caston
Rita Fioresi

Mathematical
Foundations of
Supersymmetry
Authors:

Claudio Carmeli Lauren Caston Rita Fioresi

Dipartimento di Fisica RAND Corporation Dipartimento di Matematica
Università di Genova 1776 Main Street Università di Bologna
Via Dodecaneso, 33 Santa Monica, CA 90407-2138 Piazza Porta San Donato 5
16146 Genova USA 40126 Bologna
Italy Italy
E-mail: [email protected]
E-mail: [email protected] E-mail: [email protected]

2010 Mathematics Subject Classification: 58-02; 58A50, 58C50, 14M30, 17A70, 32C11, 81Q60

Key words: supermanifolds, supersymmetry, superschemes, Lie superalgebras

ISBN 978-3-03719-097-5

The Swiss National Library lists this publication in The Swiss Book, the Swiss national bibliography,
and the detailed bibliographic data are available on the Internet at http://www.helveticat.ch.

This work is subject to copyright. All rights are reserved, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, re-use of illustrations, recitation, broadcasting,
reproduction on microfilms or in other ways, and storage in data banks. For any kind of use permission
of the copyright owner must be obtained.

© 2011 European Mathematical Society

Contact address:
European Mathematical Society Publishing House
Seminar for Applied Mathematics
ETH-Zentrum FLI C4
CH-8092 Zürich
Switzerland
Phone: +41 (0)44 632 34 36
Email: [email protected]
Homepage: www.ems-ph.org

Typeset using the authors’ TEX files: I. Zimmermann, Freiburg

Printing and binding: Druckhaus Thomas Müntzer GmbH, Bad Langensalza, Germany
∞ Printed on acid free paper

987654321
Preface

Supersymmetry was discovered by physicists in the 1970s. The mathematical treatment

of it began much later and grew out of the works notably of Berezin, Kostant, Leites,
Manin, Bernstein, Freed, Deligne, Morgan, Varadarajan and others. These works are
all in what one may call the differential category and contain many additional references
to the subject.
This monograph has grown out of the desire to present a moderately brief and
focussed account of the mathematical foundations of supersymmetry both in the dif-
ferential and algebraic categories. Our view is that supergeometry and super Lie theory
are beautiful areas and deserve much attention.
Our intention was not to write an encyclopedic treatment of supersymmetry but to
supply only the foundational material that will allow the reader to penetrate the more
advanced papers in the wide literature on this subject. For this reason we do not treat the
differential and symplectic supergeometry and we are unable to give a comprehensive
treatment of the representation theory of Lie supergroups and Lie superalgebras, which
can be found in more advanced papers by Kac, Serganova, Penkov, Duflo, Cassinelli
et al. and so on.
Our work is primarily directed to second or third year graduate students who have
taken a one year graduate course in algebra and a beginning course in Lie groups and Lie
algebras. We have provided a discussion without proofs of the classical theory, which
will serve as a departure point for our supergeometric treatment. Our book can very
well be used as a one-semester course or a participating seminar on supersymmetry,
directed to second and third year graduate students.
The language used in this monograph is that of the functor of points. Since this lan-
guage is not always familiar even to second-year graduate students we have attempted
to explain it even at the level of classical geometry. Apart from being the most natural
medium for understanding supergeometry, it is also, remarkably enough, the language
closest to the physicists’ method of working with supersymmetry.
We wish to thank professor V. S. Varadarajan for introducing us to this beautiful
part of mathematics. He has truly inspired us through his insight and deep under-
standing of the subject. We also wish to thank Dr. L. Balduzzi, Prof. G. Cassinelli,
Prof. A. Cattaneo, Prof. M. Duflo, Prof. F. Gavarini, Prof. A. Kresch, Prof. M. A. Lledo,
Prof. L. Migliorini, Prof. I. M. Musson, Prof. V. Ovsienko, Dr. E. Petracci, Prof. A. Vis-
toli and Prof. A. Zubkov for helpful remarks. We also want to thank the UCLA De-
partment of Mathematics, the Dipartimento di Matematica, Università di Bologna, and
the Dipartimento di Fisica, Università di Genova, for support and hospitality during
the realization of this work.
Introduction

Supersymmetry (SUSY) is the machinery mathematicians and physicists have devel-

oped to treat two types of elementary particles, bosons and fermions, on the same
footing. Supergeometry is the geometric basis for supersymmetry; it was first dis-
covered and studied by physicists, Wess and Zumino [80], Salam and Strathdee [65]
(among others), in the early 1970s. Today supergeometry plays an important role in
high energy physics. The objects in super geometry generalize the concept of smooth
manifolds and algebraic schemes to include anticommuting coordinates. As a result,
we employ the techniques from algebraic geometry to study such objects, namely
A. Grothendieck’s theory of schemes.
Fermions include all of the material world; they are the building blocks of atoms.
Fermions do not like each other. This is in essence the Pauli exclusion principle which
states that two electrons cannot occupy the same quantum mechanical state at the same
time. Bosons, on the other hand, can occupy the same state at the same time.
Instead of looking at equations that simply describe either bosons or fermions sep-
arately, supersymmetry seeks out a description of both simultaneously. Transitions
between fermions and bosons require that we allow transformations between the com-
muting and anticommuting coordinates. Such transitions are called supersymmetries.
In classical Minkowski space, physicists classify elementary particles by their mass
and spin. Einstein’s special theory of relativity requires that physical theories must be
invariant under the Poincaré group. Since observable operators (e.g. Hamiltonians)
must commute with this action, the classification corresponds to finding unitary repre-
sentations of the Poincaré group. In the SUSY world, this means that mathematicians
are interested in unitary representations of the super Poincaré group. A “super” rep-
resentation gives a “multiplet” of ordinary particles which include both fermions and
bosons.
Up to this point, there have been no colliders that can produce the energy required
to physically expose supersymmetry. However, the Large Hadron Collider (LHC)
in CERN (Geneva, Switzerland) became operational in 2007. Physicists are plan-
ning proton–proton and proton–antiproton collisions which will produce energies high
enough where it is believed supersymmetry can be seen. Such a discovery will solidify
supersymmetry as the most viable path to a unified theory of all known forces. Even
before the boson–fermion symmetry which SUSY presupposes is proved to be physical
fact, the mathematics behind the theory is quite remarkable. The concept that space is
an object built out of local pieces with specific local descriptions has evolved through
many centuries of mathematical thought. Euclidean and non-Euclidean geometry, Rie-
mann surfaces, differentiable manifolds, complex manifolds, algebraic varieties, and
so on represent various stages of this concept. In Alexander Grothendieck’s theory of
schemes, we find a single structure that encompasses all previous ideas of space. How-
viii Introduction

ever, the fact that conventional descriptions of space will fail at very small distances
(Planck length) has been the driving force behind the discoveries of unconventional
models of space that are rich enough to portray the quantum fluctuations of space at
these unimaginably small distances. Supergeometry is perhaps the most highly de-
veloped of these theories; it provides a surprising application and continuation of the
Grothendieck theory and opens up large vistas. One should not think of it as a mere
generalization of classical geometry, but as a deep continuation of the idea of space
and its geometric structure.
Out of the first supergeometric objects constructed by the pioneering physicists
came mathematical models of superanalysis and supermanifolds independently by
F. A. Berezin [10], B. Kostant [49], D. A. Leites [53], and De Witt [25]. The idea
to treat a supermanifold as a ringed space with a sheaf of Z=2Z-graded algebras was
introduced in these early works. Later, Bernstein [22] and Leites [53] used techniques
from algebraic geometry to deepen the study of supersymmetry. In particular, Bern-
stein and Leites accented the functor of points approach from Grothendieck’s theory
of schemes. Interest in SUSY has grown in the past decade, and most recently works
by V. S. Varadarajan [76] and others have continued exploration of this beautiful area
of physics and mathematics and have inspired this work. Given the interest and the
number of people who have contributed greatly to this field from various perspectives,
it is impossible to give a fair and accurate account of all the works related to ours. We
have nevertheless made an attempt and have provided bibliographical references at the
end of each chapter, pointing out the main papers that have inspired our work. We
apologize for any involuntary omissions.
In our exposition of mathematical SUSY, we use the language of T -points to build
supermanifolds up from their foundations in Z=2Z-graded linear algebra (superalge-
bra). The following is a brief description of our work.
In Chapter 1 we begin by studying Z=2Z-graded linear objects. We define super
vector spaces and superalgebras, then generalize some classical results and ideas from
linear algebra to the super setting. For example, we define a super Lie algebra, discuss
supermatrices, and formulate the super trace and determinant (the Berezinian). We
also discuss the Poincaré–Birkhoff–Witt theorem in full detail.
In Chapter 2 we provide a brief account of classical sheaf theory with a section
dedicated to schemes. This is meant to be an introductory chapter on this subject and
the advanced reader may very well skip it.
In Chapter 3 we introduce the most basic geometric structure: a superspace. We
present some general properties of superspaces which lead into two key examples of
superspaces, supermanifolds and superschemes. Here we also introduce the notion of
T -points which allows us to treat our geometric objects as functors; it is a fundamental
tool to gain geometric intuition in supergeometry.
Chapters 4–9 lay down the full foundations of C 1 -supermanifolds over R. In
Chapter 4, we give a complete proof of foundational results like the chart theorem
and the correspondence between morphisms of supermanifolds and morphisms of the
superalgebras of their global sections. In Chapter 5 we discuss the local structure
 Introduction ix

of morphisms proving the analog of the inverse function, submersion and immersion
theorems. In Chapter 6 we prove the local and global Frobenius theorem on super-
manifolds. In Chapters 7 and 8 we give special attention to super Lie groups and their
associated Lie algebras, as well as look at how group actions translate infinitesimally.
We then use infinitesimal actions and their characterizations to build the super Lie sub-
group, subalgebra correspondence. Finally in Chapter 9 we discuss quotients of Lie
supergroups.
Chapters 10, 11 expand upon the notion of a superscheme which we introduce
in Chapter 3. We immediately adopt the language of T -points and give criteria for
representability: in supersymmetry it is often most convenient to describe an ob-
ject functorially, and then show that it is representable. We explicitly construct the
Grassmannian functorially, then use the representability criterion to show that it is a
superscheme. Chapter 10 concludes with an examination of the infinitesimal theory of
superschemes.
We continue this exploration in Chapter 11 from the point of view of algebraic
supergroups and their Lie algebras. We discuss the linear representations of affine
algebraic supergroups; in particular we show that all affine supergroups are realized as
subgroups of the general linear supergroup.
We have made an effort to make this work self-contained and suggest that the reader
begins with Chapters 1–3, but Chapters 4–9 and Chapters 10–11 are somewhat disjoint
and may be read independently.
Contents

Preface v

Introduction vii

1 Z=2Z-graded linear algebra 1

1.1 Super vector spaces and superalgebras . . . . . . . . . . . . . . . . . 1
1.2 Super Lie algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 Modules for superalgebras . . . . . . . . . . . . . . . . . . . . . . . 9
1.4 The language of matrices . . . . . . . . . . . . . . . . . . . . . . . . 10
1.5 The Berezinian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.6 The universal enveloping superalgebra . . . . . . . . . . . . . . . . . 16
1.7 Hopf superalgebras . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.8 The even rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.9 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

2 Sheaves, functors and the geometric point of view 28

2.1 Ringed spaces of functions . . . . . . . . . . . . . . . . . . . . . . . 28
2.2 Sheaves and ringed spaces . . . . . . . . . . . . . . . . . . . . . . . 30
2.3 Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.4 Functor of points . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.5 Coherent sheaves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.6 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

3 Supergeometry 45
3.1 Superspaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.2 Supermanifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.3 Superschemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.4 The functor of points . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.5 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

4 Differentiable supermanifolds 54
4.1 Superdomains and their morphisms . . . . . . . . . . . . . . . . . . . 54
4.2 The category of supermanifolds . . . . . . . . . . . . . . . . . . . . 59
4.3 Local and infinitesimal theory of supermanifolds . . . . . . . . . . . 65
4.4 Vector fields and differential operators . . . . . . . . . . . . . . . . . 70
4.5 Global aspects of smooth supermanifolds . . . . . . . . . . . . . . . 74
4.6 The functor of points of supermanifolds . . . . . . . . . . . . . . . . 79
4.7 Distributions with finite support . . . . . . . . . . . . . . . . . . . . 82
xii Contents

4.8 Complex and real supermanifolds . . . . . . . . . . . . . . . . . . . 86

4.9 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

5 The local structure of morphisms 90

5.1 The inverse function theorem . . . . . . . . . . . . . . . . . . . . . . 90
5.2 Immersions, submersions and the constant rank morphisms . . . . . . 92
5.3 Submanifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.4 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

6 The Frobenius theorem 103

6.1 The local super Frobenius theorem . . . . . . . . . . . . . . . . . . . 103
6.2 The global super Frobenius theorem . . . . . . . . . . . . . . . . . . 110
6.3 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

7 Super Lie groups 112

7.1 Super Lie groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
7.2 The super Lie algebra of a super Lie group . . . . . . . . . . . . . . . 114
7.3 The Hopf superalgebra of distributions . . . . . . . . . . . . . . . . . 118
7.4 Super Harish-Chandra pairs . . . . . . . . . . . . . . . . . . . . . . . 123
7.5 Homogeneous one-parameter supergroups . . . . . . . . . . . . . . . 137
7.6 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140

8 Actions of super Lie groups 141

8.1 Actions of super Lie groups on supermanifolds . . . . . . . . . . . . 141
8.2 Infinitesimal actions . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
8.3 Actions of super Harish-Chandra pairs . . . . . . . . . . . . . . . . . 146
8.4 The stabilizer subgroup . . . . . . . . . . . . . . . . . . . . . . . . . 149
8.5 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

9 Homogeneous spaces 154

9.1 Transitive actions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
9.2 Homogeneous spaces: The classical construction . . . . . . . . . . . 157
9.3 Homogeneous superspaces for super Lie groups . . . . . . . . . . . . 158
9.4 The functor of points of a quotient supermanifold . . . . . . . . . . . 162
9.5 The super Minkowski and super conformal spacetime . . . . . . . . . 166
9.6 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172

10 Supervarieties and superschemes 173

10.1 Basic definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
10.2 The functor of points . . . . . . . . . . . . . . . . . . . . . . . . . . 179
10.3 A representability criterion . . . . . . . . . . . . . . . . . . . . . . . 183
10.4 The Grassmannian superscheme . . . . . . . . . . . . . . . . . . . . 190
10.5 Projective supergeometry . . . . . . . . . . . . . . . . . . . . . . . . 192
10.6 The infinitesimal theory . . . . . . . . . . . . . . . . . . . . . . . . . 196
 Contents xiii

10.7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201

11 Algebraic supergroups 202

11.1 Supergroup functors and supergroup schemes . . . . . . . . . . . . . 202
11.2 Lie superalgebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
11.3 Lie.G/ of a supergroup functor G . . . . . . . . . . . . . . . . . . . 210
11.4 Lie.G/ for a supergroup scheme G . . . . . . . . . . . . . . . . . . . 213
11.5 The Lie superalgebra of a supergroup scheme . . . . . . . . . . . . . 215
11.6 Affine algebraic supergroups . . . . . . . . . . . . . . . . . . . . . . 221
11.7 Linear representations . . . . . . . . . . . . . . . . . . . . . . . . . . 223
11.8 The algebraic stabilizer theorem . . . . . . . . . . . . . . . . . . . . 227
11.9 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230

Appendices (with the assistance of Ivan Dimitrov) 231

A Lie superalgebras 231

A.1 Classical Lie superalgebras . . . . . . . . . . . . . . . . . . . . . . . 232
A.2 Root systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
A.3 Cartan matrices and Dynkin diagrams . . . . . . . . . . . . . . . . . 242
A.4 Classification of finite-dimensional irreducible modules . . . . . . . . 244
A.5 Representations of basic Lie superalgebras . . . . . . . . . . . . . . . 248
A.6 More on representations of Lie superalgebras . . . . . . . . . . . . . 254
A.7 Schur’s lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256

B Categories 259
B.1 Categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
B.2 Sheafification of a functor . . . . . . . . . . . . . . . . . . . . . . . . 262
B.3 Super Nakayama’s lemma and projective modules . . . . . . . . . . . 266
B.4 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268

C Fréchet superspaces 269

C.1 Fréchet spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
C.2 Fréchet superspaces . . . . . . . . . . . . . . . . . . . . . . . . . . . 272

Bibliography 277

Index 283
1
Z=2Z-graded linear algebra

The theory of manifolds and algebraic geometry are ultimately based on linear algebra.
Similarly the theory of supermanifolds needs super linear algebra, which is linear
algebra in which vector spaces are replaced by vector spaces with a Z=2Z-grading,
namely, super vector spaces. The basic idea is to develop the theory along the same
lines as the usual theory, adding modifications whenever necessary. We therefore first
build the foundations of linear algebra in the super context. This is an important starting
point as we later build super geometric objects from sheaves of super linear spaces.
Let us fix a ground field k, char.k/ ¤ 2; 3.

1.1 Super vector spaces and superalgebras

Definition 1.1.1. A super vector space is a Z=2Z-graded vector space
V D V0 ˚ V1
where elements of V0 are called “even” and elements of V1 are called “odd”.
Definition 1.1.2. The parity of v 2 V , denoted by p.v/ or jvj, is defined only on
non-zero homogeneous elements, that is elements of either V0 or V1 :
´
0 if v 2 V0 ;
p.v/ D jvj D
1 if v 2 V1 :

Since any element may be expressed as the sum of homogeneous elements, it suffices
to consider only homogeneous elements in the statement of definitions, theorems, and
proofs.
Definition 1.1.3. The superdimension of a super vector space V is the pair .p; q/
where dim.V0 / D p and dim.V1 / D q as ordinary vector spaces. We simply write
dim.V / D pjq.
From now on we will simply refer to the superdimension as the dimension when it
is clear that we are working with super vector spaces. If dim.V / D pjq, then we can
find a basis fe1 ; : : : ; ep g of V0 and a basis f1 ; : : : ; q g of V1 so that V is canonically
isomorphic to the free k-module generated by the fe1 ; : : : ; ep ; 1 ; : : : ; q g. We denote
this k-module by k pjq and we will call .e1 ; : : : ; ep ; 1 ; : : : ; q / the canonical basis of
k pjq . The .ei / form a basis of k p D k0pjq and the .j / form a basis for k q D k1pjq .
2 1 Z=2Z-graded linear algebra

Definition 1.1.4. A morphism from a super vector space V to a super vector space W
is a linear map from V to W preserving the Z=2Z-grading. Let Hom.V; W / denote
the vector space of morphisms V ! W .
Thus we have formed the category1 of super vector spaces that we denote by .smod/.
It is important to note that the category of super vector spaces also admits an “inner
Hom”, which we denote by Hom.V; W /; for super vector spaces V; W , Hom.V; W /
consists of all linear maps from V to W ; it is made into a super vector space itself by
the following definitions:

Hom.V; W /0 D fT W V ! W j T preserves parityg .D Hom.V; W //I

Hom.V; W /1 D fT W V ! W j T reverses parityg:

If V D k mjn , W D k pjq we have, in the canonical basis .ei ; j /:

² ³ ² ³
A 0 0 B
Hom.V; W /0 D and Hom.V; W /1 D
0 D C 0

where A, B, C , D are respectively .p  m/, .p  n/, .q  m/, .q  n/-matrices with

entries in k.
In the category of super vector spaces we have the parity reversing functor ….V !
…V / defined by
.…V /0 D V1 ; .…V /1 D V0 :
The category of super vector spaces admits tensor products: for super vector spaces
V; W , V ˝ W is given the Z=2Z-grading as follows:

.V ˝ W /0 D .V0 ˝ W0 / ˚ .V1 ˝ W1 /;
.V ˝ W /1 D .V0 ˝ W1 / ˚ .V1 ˝ W0 /:

The assignment V; W 7! V ˝ W is additive and exact in each variable as in the

ordinary vector space category. The object k functions as a unit element with respect
to tensor multiplication ˝; and tensor multiplication is associative, i.e., the two products
U ˝.V ˝W / and .U ˝V /˝W are naturally isomorphic. Moreover, V ˝W Š W ˝V
by the commutativity map

cV;W W V ˝ W ! W ˝ V

where v ˝ w 7! .1/jvjjwj w ˝ v.
The significance of this definition is as follows. If we are working with the category
of vector spaces, the commutativity isomorphism takes v ˝ w to w ˝ v. In super linear
algebra we have to add the sign factor in front. This is a special case of the general
1
We refer the reader not accustomed to category language to Appendix B.1.
 1.1 Super vector spaces and superalgebras 3

principle called the “sign rule” that one finds in some physics and mathematics litera-
ture. The principle says that in making definitions and proving theorems, the transition
from the usual theory to the super theory is often made by just simply following this
principle, which introduces a sign factor whenever one reverses the order of two odd
elements. The functoriality underlying the constructions makes sure that the definitions
are all consistent.
The commutativity isomorphism satisfies the so-called hexagon diagram:
cU;V ˝W
U ˝ V ˝ QW /V ˝W ˝U
QQQ
QQQU;V
c c U;W mmm
mm6
QQQ mm
QQ( mm
mmm
V ˝U ˝W
where, if we had not suppressed the arrows of the associativity morphisms, the diagram
would have the shape of a hexagon.
The definition of the commutativity isomorphism, also informally referred to as the
sign rule, has the following very important consequence. If V1 ; : : : ; Vn are super vector
spaces and  and  are two permutations of n elements, no matter how we compose
associativity and commutativity morphisms, we always obtain the same isomorphism
from V.1/ ˝    ˝ V.n/ to V.1/ ˝    ˝ V.n/ namely:

V.1/ ˝    ˝ V.n/ ! V.1/ ˝    ˝ V.n/ ;

v.1/ ˝    ˝ v.n/ 7! .1/N v.1/ ˝    ˝ v.n/

where N is the number of pairs of indices i , j such that vi and vj are odd and  1 .i / <
 1 .j / with  1 .i / >  1 .j /.
The dual, V  , of V is defined as

V  ´ Hom.V; k/:

Notice that, if V is even, that is V D V0 , we have V  is the ordinary dual of V ,

consisting of all even morphisms V ! k. If V is odd, that is V D V1 , then V  is also
an odd vector space and consists of all odd morphisms V 1 ! k. This is because any
morphism from V1 to k D k 1j0 is necessarily odd since it sends odd vectors into even
ones.
The category of super vector spaces thus becomes what is known as a tensor category
with inner Hom and dual. We start by recalling the universal property of the tensor
product.
Proposition 1.1.5. Let V and W be two super vector spaces and f a bilinear map
of V  W into a third super vector space Z. Then there exists a unique morphism
g W V ˝ W ! Z such that

g.v ˝ w/ D f .v; w/ .v 2 V; w 2 W /:
4 1 Z=2Z-graded linear algebra

Proof. See [51], Ch. XVI. 

Remark 1.1.6. The object V ˝n D V ˝    ˝ V (n times) for a super vector space V
is perfectly well defined. We can extend this notion to make sense of V ˝njm via the
parity reversing functor …. Define
V njm ´ „
V  V ƒ‚
     V…  ….V /  ….V /      ….V /;
„ ƒ‚ …
n times m times

from which the definition of V ˝njm follows by the universal property. In other words,
we have:
V ˝njm ´ V ˝ V ˝    ˝ V ˝ ….V / ˝ ….V / ˝    ˝ ….V /
where the parity is coming from the tensor product.
In the ordinary setting, an algebra is a vector space A with a multiplication which
is bilinear. We may therefore think of it as a vector space A together with a linear map
A ˝ A ! A. We now define a superalgebra in the same way:
Definition 1.1.7. A superalgebra is a super vector space A together with a multiplica-
tion morphism  W A ˝ A ! A.
We then say that a superalgebra A is (super)commutative if
 B cA;A D ;
that is, if the product of homogeneous elements obeys the rule
ab D .1/jajjbj ba:
This is an example of the sign rule mentioned earlier. Note that the signs do not
appear in the definition; this is the advantage of the categorical view point which
suppresses signs and therefore streamlines the theory.
Similarly we say that A is associative if
 B  ˝ id D  B id ˝ 
on A ˝ A ˝ A. In other words if .ab/c D a.bc/. We also say that A has a unit if there
is an even element 1 so that
.1 ˝ a/ D .a ˝ 1/ D a
for all a 2 A, that is if a  1 D 1  a D a.
The tensor product A ˝ B of two superalgebras A and B is again a superalgebra,
with multiplication defined as
.a ˝ b/.c ˝ d / D .1/jbjjcj .ac ˝ bd /:
As an example of associative superalgebra we are going to define the tensor super-
algebra.
 1.1 Super vector spaces and superalgebras 5

Definition 1.1.8. Let V be a super vector space. We define tensor superalgebra to be

the super vector space
M M M
T .V / D V ˝n ; T .V /0 D V ˝n ; T .V /1 D V ˝n ;
n0 n even n odd

together with the product defined, as usual, via the ordinary bilinear map r;s W V ˝r 
V ˝s ! V ˝.rCs/ ,
r;s .vi1 ˝    ˝ vir ; wj1 ˝    ˝ wjs / D vi1 ˝    ˝ vir ˝ wj1    ˝ wjs :
One can check that T .V / is a well-defined associative superalgebra with unit, which
is noncommutative except when V is even and one-dimensional.
From now on we will assume that all superalgebras are associative and with unit
unless specified. Moreover we shall denote the category of commutative superalgebras
by .salg/.
If we take a super vector space and mod out the odd part, we obtain a classical (that
is, purely even) vector space. In a superalgebra the corresponding object is defined
by taking the quotient by the ideal generated by the odd elements. This allows one to
always refer back to the classical setting.
We denote by JA the ideal in the commutative superalgebra A generated by the odd
elements in A.
Example 1.1.9 (Grassmann coordinates). Let
A D kŒt1 ; : : : ; tp ; 1 ; : : : ; q 
where the t1 ; : : : ; tp are ordinary indeterminates and the 1 ; : : : ; q are odd indetermi-
nates, i.e., they behave like Grassmann coordinates:
i j D j i :
(This of course implies that i2 D 0 for all i .) In other words we can view A as
the ordinary tensor product kŒt1 ; : : : ; tp  ˝ ^.1 ; : : : ; q /, where ^.1 ; : : : ; q / is the
exterior algebra generated by 1 ; : : : ; q .
As one can readily check, A is a supercommutative algebra. In fact,
˚ P 
A0 D f0 C jI j even fI I j I D fi1 <    < ir g
where I D i1 i2 : : : ir , jI j D r and f0 ; fI 2 kŒt1 ; : : : ; tp , and
˚P 
A1 D jJ j odd fJ J j J D fj1 <    < js g :

Note that although the fj g 2 A1 , there are plenty of nilpotents in A0 ; take for example
1 2 2 A0 .
This example is important since any finitely generated commutative superalgebra
is isomorphic to a quotient of the algebra A by a homogeneous ideal.
As one can readily check, JA D .1 ; : : : ; q / and A=JA Š kŒt1 ; : : : ; tp .
6 1 Z=2Z-graded linear algebra

Another important example of commutative superalgebra is the symmetric algebra

over a super vector space V .
Consider the natural action of the permutation group Sn on the tensor product V ˝n :

s  v1 ˝    ˝ vn D .1/p.s/ vs 1 .1/ ˝    ˝ vs 1 .n/ ; s 2 Sn ;

where
p.s/ ´ jf.k; l/ j k > l; vk ; vl odd; s.k/ < s.l/gj:
Let an be the subspace generated by the elements

v1 ˝    ˝ vn  s  v1 ˝    ˝ vn for all s 2 Sn and all vi 2 V:

L L
Observe that a D n0 an is an ideal in T .V / D n0 V ˝n .
Definition 1.1.10. Let V be a super vector space. We define the symmetric n-power
Symn .V / as the super vector space

Symn .V / D V ˝n =an :

We also define the symmetric superalgebra Sym.V / as the superalgebra

Sym.V / D T .V /=a:

The next proposition is a simple exercise.

Proposition 1.1.11. Let V be a super vector space.

Hom.salg/ .Sym.V /; A/ D Hom.smod/ .V; A/

for any commutative superalgebra A. In other words any linear morphism V ! A

extends uniquely to a superalgebra morphism Sym.V / ! A.
In a similar way we can also define the exterior algebra.
Definition 1.1.12. Let V be a super vector space. We define the exterior n-power
Vn
.V / as the super vector space
V
n
.V / D V ˝n =bn ;

where bn is the subspace of V ˝n generated by the elements

v1 ˝    ˝ vn  .1/p.s/ s  v1 ˝    ˝ vn for all s 2 Sn and all vi 2 V;

where p.s/ denotes

V again the parity of the permutation s. We also define the exterior
superalgebra .V / as the superalgebra
V
.V / D T .V /=b;
L
with b D n0 bn .
 1.1 Super vector spaces and superalgebras 7

Example 1.1.13. If V D k pjq ,

V
Sym.V / D kŒt1 ; : : : ; tp  ˝ Œs1 ; : : : ; sq ;
V V
.V / D kŒs1 ; : : : ; sq  ˝ Œt1 ; : : : ; tp ;
where the ti and sj are just (even) indeterminates.
Observation 1.1.14. We have the following:
V V
(1) Sym.V / Š Sym.V0 / ˝ .V1 /, where .V1 / denotes in this case just the usual
exterior algebra of the ordinary vector space V1 .
V V
(2) .V / Š .V0 / ˝ Sym.V1 /, where Sym.V1 / denotes in this case just the usual
symmetric algebra of the ordinary vector space V1 .
V V
(3) Sym.…V / D .V /, .….V // D Sym.V /.
Our next goal is to define derivations in the super context. An even derivation of
a superalgebra A is a super vector space homomorphism D W A ! A such that for
a; b 2 A, D.ab/ D D.a/b C aD.b/. We may of course extend this definition to
include odd linear maps:
Definition 1.1.15. Let D 2 Homk .A; A/ be a k-linear map. Then D is a derivation
of the superalgebra A if

D.ab/ D D.a/b C .1/jDjjaj aD.b/ (1.1)

for a; b 2 A.
This definition itself is an instance of the sign rule; since a and D are being inter-
changed, the sign factor appears.
The derivations in Hom.smod/ .A; A/ are even (as above) while those contained in
Hom.smod/ .A; A/1 are odd. The set of all derivations of a superalgebra A, denoted
Der.A/, is itself a special type of superalgebra called a super Lie algebra which we
describe in the following section.
Example 1.1.16. Consider the k-linear operators f@=@ti g and f@=@j g of the polyno-
mial superalgebra A D kŒt1 ; : : : ; tp ; 1 ; : : : ; q  to itself defined as

@=@ti .tk / D ıki ; @=@ti .l / D 0;

@=@j .tk / D 0; @=@j .l / D ılj ;

and extended to the whole of A according to (1.1). Then f@=@ti ; @=@j g 2 Der.A/, and
we have ² ³
@ @
Der.A/ D spanA ; :
@ti @j
@=@i is to be viewed as an odd derivation.
8 1 Z=2Z-graded linear algebra

1.2 Super Lie algebras

An important object in supersymmetry is the super Lie algebra.
Definition 1.2.1. A super Lie algebra L is an object in the category of super vector
spaces together with a morphism Œ ;  W L ˝ L ! L, often called the super bracket, or
simply, the bracket, which satisfies the following conditions.
(1) Anti-symmetry
Π;  C Π;  B cL;L D 0
which is the same as Œx; y C .1/jxjjyj Œy; x D 0 for x; y 2 L homogeneous.
(2) The Jacobi identity
Π; Π;  C Π; Π;  B  C Π; Π;  B  2 D 0
where  2 S3 is a three-cycle, i.e., it takes the first entry of Œ; Œ ;  to the second,
the second to the third, and the third to the first. So for x; y; z 2 L homogeneous,
this reads:
Œx; Œy; z C .1/jxjjyjCjxjjzj Œy; Œz; x C .1/jyjjzjCjxjjzj Œz; Œx; y D 0:

It is important to note that in the super category, these conditions are modifications
of the properties of the bracket in a Lie algebra, designed to accommodate the odd
variables.
We shall often use also the term super Lie algebra instead of Lie superalgebra since
both are present in the literature.
Remark 1.2.2. We can immediately extend this definition to the case where L is an
A-module for A a commutative superalgebra, thus defining a Lie superalgebra in the
category of A-modules that we shall discuss in detail in Section 1.3.
Example 1.2.3. Define the associative superalgebra End.V / as the super vector space
Hom.V; V /:
End.V / D Hom.V; V /0 ˚ Hom.V; V /1
with the composition as product. It is a Lie superalgebra with bracket
ŒX; Y  D X Y  .1/jXjjY j YX;
where the bracket as usual is defined only on homogeneous elements and then extended
by linearity.
Example 1.2.4. In Example 1.1.16 above,
² ³
@ @
Der.A/ D spanA ;
@ti @j
is a super Lie algebra where the bracket is defined, for D1 ; D2 2 Der.A/, to be
ŒD1 ; D2  D D1 D2  .1/jD1 jjD2 j D2 D1 .
 1.3 Modules for superalgebras 9

In fact, we can make any associative superalgebra A into a Lie superalgebra by

taking the bracket to be
Œa; b D ab  .1/jajjbj ba;
i.e., we take the bracket to be the difference    B cA;A where we recall  is the mul-
tiplication morphism on A. We will discuss other examples of super Lie algebras after
the following discussion of superalgebra modules. In particular we want to examine
the super version of a matrix algebra.

Remark 1.2.5. If the ground field has characteristic 2 or 3 in addition to the antisym-
metry and Jacobi conditions, one requires that Œx; x D 0 for x even if the characteristic
is 2 and Œy; Œy; y D 0 for y odd if the characteristic is 3. For more details on super-
algebras over fields with positive characteristic see [3].

1.3 Modules for superalgebras

Let A be a superalgebra, associative, but not necessarily commutative, in this section.

Definition 1.3.1. A left A-module is a super vector space M with a morphism A˝M !
M , a ˝ m 7! am of super vector spaces obeying the usual identities; that is, for all a,
b 2 A and x, y 2 M we have
(1) a.x C y/ D ax C ay,
(2) .a C b/x D ax C bx,
(3) .ab/x D a.bx/,
(4) 1x D x.

A right A-module is defined similarly. Note that if A is commutative, a left A-

module is also a right A-module if we define (the sign rule)

m  a D .1/jmjjaj a  m

for m 2 M , a 2 A. Morphisms of A-modules are defined in the obvious manner:

they are super vector space morphisms  W M ! N such that .am/ D a.m/ for
all a 2 A, m 2 M . So we have the category of A-modules. For A commutative, the
category of A-modules admits tensor products: for M1 ; M2 A-modules, M1 ˝ M2 is
taken as the tensor product of M1 as a right module with M2 as a left module.
Let us now turn our attention to free A-modules. We already have the notion of the
super vector space k pjq over k, and so we define Apjq ´ A ˝ k pjq where

.Apjq /0 D A0 ˝ .k pjq /0 ˚ A1 ˝ .k pjq /1 ;

.Apjq /1 D A1 ˝ .k pjq /0 ˚ A0 : ˝ .k pjq /1 :
10 1 Z=2Z-graded linear algebra

Definition 1.3.2. We say that an A-module M is free if it is isomorphic (in the category
of A-modules) to Apjq for some .p; q/.
This definition is equivalent to saying that M contains p even elements fe1 ; : : : ; ep g
and q odd elements f1 ; : : : ; q g such that

M0 D spanA0 fe1 ; : : : ; ep g ˚ spanA1 f1 ; : : : ; q g;

M1 D spanA1 fe1 ; : : : ; ep g ˚ spanA0 f1 ; : : : ; q g:

We shall also say that M is the free module generated over A by the even elements
e1 ; : : : ; ep and the odd elements 1 ; : : : ; q .
Let T W Apjq ! Arjs be a morphism of free A-modules and write epC1 ; : : : ; epCq
for the odd basis elements 1 ; : : : ; q . Then T is defined on the basis elements
fe1 ; : : : ; epCq g by
X
pCq
T .ej / D ei tji : (1.2)
iD1

Hence T can be represented as a matrix of size .r C s/  .p C q/:

 
T1 T2
T D ; (1.3)
T3 T4
where T1 is an r  p matrix consisting of even elements of A, T2 is an r  q matrix of
odd elements, T3 is an s  p matrix of odd elements, and T4 is an s  q matrix of even
elements. We say that T1 and T4 are even blocks and that T2 and T3 are odd blocks.
Note that the fact that T is a morphism of super A-modules means that it must preserve
parity, and therefore the parity of the blocks is determined. Note also that when we
define T on the basis elements, in the expression (1.2) the basis element precedes the
coordinates tji . This is important to keep the signs in order and comes naturally from
composing morphisms. In other words if the module is written as a right module with
T acting from the left, composition becomes matrix product in the usual manner:
X  X
.S  T /.ej / D S ei tji D ek sik tji :
i i;k
P
Hence for any x 2 Apjq , we can express x as the column vector x D ei x i and so
T .x/ is given by the matrix product T x.

1.4 The language of matrices

Let A be a commutative superalgebra.
Let us now consider all endomorphisms of M D Apjq , i.e., Hom(M; M ). This is
an ordinary algebra (i.e., not super) of matrices of the same type as T above. Even
 1.4 The language of matrices 11

though in matrix form each morphism contains blocks of odd elements of A, each
morphism is an even linear map from M to itself since a morphism in the super category
must preserve parity. In order to get a truly supergeometric version of the ordinary
matrix algebra, we must consider all linear maps M to M , i.e., we are interested in
Hom(M; M ). Now we can talk about even and odd matrices. An even matrix T takes
on the block form from above. But the parity of the blocks is reversed for an odd matrix
S ; we get  
S1 S2
SD ;
S3 S4
where S1 is a p p odd block, S4 is a q q odd block, S2 is a p q even block, and S3 is
a q  p even block. Note that in the case where M D k pjq , the odd blocks are just zero
blocks. We will denote this superalgebra of even and odd .pCq/.pCq/ D pjq pjq
matrices by Mat.Apjq /. This superalgebra is in fact a super Lie algebra where we define
the bracket Π;  as in Example 1.2.4:
ŒT; S  D T S  .1/jT jjSj S T
for S; T 2 Mat.Apjq /.
Remark 1.4.1. Note that Mat.Apjq / D Hom.Apjq ; Apjq /. We do not want to confuse
this with what we will later denote as Mpjq .A/, which will functorially only include
the even part of Mat.Apjq /, i.e.,
Mat.Apjq /0 D Mpjq .A/ D Hom.Apjq ; Apjq /:
We shall come back to this key model in Example 3.1.5.
We now turn to the supergeometric extensions of the trace and determinant. Let
T W Apjq ! Apjq be a morphism (i.e., T 2 .Mat.Apjq //0 ) with block form (1.3).
Definition 1.4.2. We define the super trace of T to be
str.T / ´ tr.T1 /  tr.T4 /;
where “tr” denotes the ordinary trace.
This negative sign is actually forced upon us when we take a categorical view of
the trace. We will not discuss this here, but we later motivate this definition when we
explore the supergeometric-extension of the determinant.
Remark 1.4.3. The super trace is actually defined for all linear maps. For S 2
Mat.Apjq /1 an odd matrix,
str.S / D tr.S1 / C tr.S4 /:
Note the sign change. Note also that the trace is commutative, meaning that for even
matrices A; B 2 Mat.Apjq /0 , we have the familiar formula
str.AB/ D str.BA/
12 1 Z=2Z-graded linear algebra

that we shall prove in Observation 1.5.8 after we have introduced the notion of Bere-
zinian.
Definition 1.4.4. If M is an A-module, then GL.M / is defined as the group of auto-
morphisms of M and we call it the super general linear group of automorphisms of
M . If M D Apjq the free A-module generated by p even and q odd variables, then
we write GL.M / D GLpjq .A/. We may also use the notation GLpjq .A/ D GL.Apjq /.

1.5 The Berezinian

We want to define the generalization of the determinant, called the Berezinian, on
elements of GL.Apjq /. We may say that this is the point where linear supergeometry
differs most dramatically from the ordinary theory.
Proposition 1.5.1. Let T W Apjq ! Apjq be a morphism with the usual block form
(1.3). Then T is invertible if and only if T1 and T4 are invertible.
Proof. Let JA  A be the ideal generated by the odd elements and let AN D A=JA .
There is a natural map Mpjq .A/ ! Mpjq .A/, N T 7! Tx , where Tx is obtained from the
matrix T by applying to its entries the map A ! A. N We claim that T is invertible if and
only if Tx is invertible. One direction is obvious, namely the case in which T is invertible.
Now assume that Tx is invertible. This implies that there exists Sx 2 Mpjq .A/ N such that
x x x x N
T S D S T D I , where I denotes the identity (both in Mpjq .A/ and in Mpjq .A/).
Hence there exists S 2 Mpjq .A/ such that T S D I C N , with N 2 Mpjq .A/ (we
consider only the case of a right inverse since the left inverse is the same). To prove T
is invertible it is enough to show that N is nilpotent, i.e., N r D 0 for some r. Since
the entries of N m are in Am 1 for m sufficiently large they are all zero. 
Definition 1.5.2. Let T be an invertible element in Mpjq .A/, i.e., T 2 GL.Apjq / with
the standard block form (1.3) from above. Then we formulate the Berezinian Ber:

Ber.T / D det.T1  T2 T41 T3 / det.T4 /1 (1.4)

where “det” is the usual determinant.

The Berezinian is named after Berezin, who was one of the pioneers of superalgebra
and superanalysis.
Remark 1.5.3. The first thing we notice is that in the super category, we only define
the Berezinian for invertible transformations. This marks an important difference with
the determinant, which is defined in ordinary linear algebra for all endomorphisms of a
vector space. We immediately see that it is necessary that the block T4 be invertible for
the formula (1.4) to make sense, however one can actually define the Berezinian on all
matrices with only the T4 block invertible (i.e., the matrix itself may not be invertible,
 1.5 The Berezinian 13

but the T4 block is). There is a similar formulation of the Berezinian which requires
that only the T1 block be invertible:
Ber.T / D det.T4  T3 T11 T2 /1 det.T1 /:
So we can actually define the Berezinian on all matrices with either the T1 or the T4
block invertible. Note that in the case where both blocks are invertible (i.e., when the
matrix T is invertible), both formulae of the Berezinian give the same answer as we
shall see after the next proposition.
Proposition 1.5.4. The Berezinian is multiplicative: For S; T 2 GL.Apjq /,
Ber.S T / D Ber.S / Ber.T /:
Proof. We will only briefly sketch the proof here and leave the details to the reader.
First note that any T 2 GL.Apjq / with block form (1.3) may be written as the product
of the following “elementary matrices”:
     
1 X Y1 0 1 0
TC D ; T0 D ; T D :
0 1 0 Y2 Z 1
If we equate T D TC T0 T , we get a system of equations which lead to the solution
X D T2 T41 ;
Y1 D T1  T2 T41 T3 ;
Y2 D T4 ;
Z D T41 T3 :
It is also easy to verify that Ber.S T / D Ber.S / Ber.T / for S of type fTC ; T0 g
for all T or T of type fT ; T0 g for all S. Let G  GLpjq .A/ be the set of elements
S such that Ber.S T / D Ber.S / Ber.T / for all T . One can check right away that G
is a subgroup of GLpjq .A/. To prove our result is it enough to show that matrices of
type TC , T , T0 2 G since they generate GLpjq .A/. By our previous discussion TC ,
T0 2 G, hence we only need to show that Ber.S T / D Ber.S / Ber.T / for S of type
T for all T . Notice that
Ber.S TC T0 T / D Ber.S TC / Ber.T0 / Ber.T /
as we have already seen. Hence, the last case to verify is for
   
1 0 1 X
SD and T D :
Z 1 0 1
We may assume that both X and Z each have only one non-zero entry since the product
of two matrices of type TC results in the sum of the upper right blocks, and likewise
with the product of two type T matrices. Let xij ; zkl ¤ 0. Then
 
1 X
ST D
Z 1 C ZX
14 1 Z=2Z-graded linear algebra

and Ber.ST / D det.1  X.1 C ZX /1 Z/ det.1 C ZX /1 . Since both X and Z have
only one non-zero entry, .1 C ZX /1 D .1  ZX /, hence

det.1  X.1 C ZX /1 Z/ D det.1  X.1  ZX /Z/ D det.1  XZ/:

This is because all the values within the determinants are either upper triangular or
contain an entire column of zeros (X; Z have at most one non-zero entry), the values
xij ; zkl contribute to the determinant only when the product ZX has its non-zero term
on the diagonal, i.e., only when i D j D k D l. Ber.S T / D det.1  XZ/ det.1 
ZX/ D .1  xi i zi i /.1 C xi i zi i / D 1. A direct calculation shows that Ber.S / D
Ber.T / D 1. 
Corollary 1.5.5. Let T 2 GLpjq .A/. Then

Ber.T / D det.T4  T3 T11 T2 /1 det.T1 /:

Proof. Consider the decomposition

     
T1 T2 1 0 T1 0 1 T11 T2
T D D :
T3 T4 T3 T11 1 0 T4  T3 T11 T2 0 1

By multiplicativity of the Berezinian, we obtain the result. 

Corollary 1.5.6. The Berezinian is a homomorphism

Ber W GL.Apjq / ! GL1j0 .A/ D A

0

into the invertible elements of A.

Proof. This follows immediately from the multiplicativity property. 
Remark 1.5.7. In the course of the proof of Corollary 1.5.5 we have seen the decom-
position
GLpjq .A/ D UH V;
where
² ³ ² ³ ² ³
1 0 H1 0 1 S
U D ; H D ; V D :
T 1 0 H2 0 1

This very much resembles the big cell decomposition in the theory of ordinary algebraic
groups, however here the decomposition holds globally.
Observation 1.5.8. The usual determinant on the general linear group GLn induces
the trace on its Lie algebra, namely the matrices Mn (see Remark 1.4.1). Let  be an
even indeterminate,  2 D 0 and let us compute the value Ber.1 C T / 2 AŒ:

Ber.1 C T / D det.1 C T1 / det.1  T4 / D .1 C  tr T1 /.1   tr T4 / D 1 C  str.T /:

 1.5 The Berezinian 15

We are now going to see that str.S T / D str.T S /.

Ber.1 C S T / D .1 C  tr.S T /1 /.1   tr.S T /4 /
D .1 C  tr.S1 T1 C S2 T3 //.1   tr.S3 T2 C S4 T4 //
D .1 C  tr.T1 S1 /   tr.T3 S2 //.1 C  tr.T2 S3 /   tr.T4 S4 //
D 1 C  tr.T1 S1 C T2 S3  T3 S2  T4 S4 /
D 1 C  str.T S /;
Hence
Ber.1 C ST / D 1 C  str.S T / D 1 C  str.T S /;
thus proving our claim.
This of course leads to the question of how the formula for the Berezinian arises.
The answer lies in the supergeometric version of integral forms on supermanifolds
called densities. In his pioneering work in superanalysis, F. A. Berezin calculated the
change of variables formula for densities on isomorphic open submanifolds of Rpjq
([10]). This led to an extension of the Jacobian in ordinary differential geometry; the
Berezinian is so-named after him.
We are ready for the formula for the inverse of a supermatrix.
Proposition 1.5.9. Let  
T1 T2
T D 2 GLpjq .A/
T3 T4
(hence T1 and T4 are invertible ordinary matrices). Then
0 1
.T1  T2 T41 T3 /1 T11 T2 .T4  T3 T11 T2 /1
T 1 D @ A:
1 1 1 1 1
T4 T3 .T1  T2 T4 T3 / .T4  T3 T1 T2 /
Proof. Direct check. 
We finish our summary of superlinear algebra by giving meaning to the rank of an
endomorphism of Apjq . We shall return to this concept and an important variation of
it in Chapter 5, Section 5.2.
Definition 1.5.10. Let T 2 End.Apjq /. Then the rank of T , rank.T /, is the superdi-
mension of the largest invertible submatrix of T (obtained by removing columns and
rows).
Proposition 1.5.11. Again, let T 2 End.Apjq / with block form (1.3). Then rank.T / D
rank.T1 /j rank.T4 /.
Proof. Assume that rank.T / D rjs. Then there is an invertible rjs  rjs submatrix
of T and it is clear that r  rank.T1 /, s  rank.T4 /. Conversely, if rank.T1 / D r 0 ,
rank.T4 / D s 0 it is also clear that there exists an invertible r 0 js 0  r 0 js 0 submatrix of T .
Therefore we must have r D r 0 , s D s 0 . 
16 1 Z=2Z-graded linear algebra

1.6 The universal enveloping superalgebra

Given a Lie superalgebra g, possibly infinite-dimensional, we would like to introduce
an associative superalgebra U.g/, called the universal enveloping superalgebra of g
(UESA for short) with a natural universal property with respect to the Lie superalgebra
structure of g. The universal enveloping superalgebra is the main tool to convert Lie
superalgebra problems into associative superalgebra ones. For example we will see
that a representation of g can be uniquely extended to a representation of its UESA
U.g/. The main results regarding the UESA are the existence and uniqueness of the
UESA and the Poincaré–Birkhoff–Witt Theorem and we will discuss them in full detail.
In order to keep the exposition simple, we shall treat the case when g has a linearly
ordered basis (e.g., when it is countable dimensional) as most applications fall in this
framework. By the well-ordering principle this hypothesis is unnecessary, but we shall
not press the point.

Definition 1.6.1. Let g be a Lie superalgebra, T .g/ the tensor superalgebra over the
underlying super vector space of g; i W g ! T .g/ the natural immersion. Let I  T .g/
be the two-sided ideal in T .g/ generated by

i.X/ ˝ i.Y /  .1/jXjjY j i.Y / ˝ i.X /  i.ŒX; Y / 2 T .g/ for all X; Y 2 g:

(As usual in the super setting we give relations only for homogeneous elements.)
We define U.g/ the universal enveloping superalgebra (UESA) of g as

U.g/ ´ T .g/=I:

We denote the product of two elements X; Y 2 U.g/ by X Y .

As any associative superalgebra, U.g/ has a natural Lie superalgebra structure:

ŒX; Y  D XY  .1/jXjjY j YX .
Let j W g ! U.g/ be the composition of the injective linear map i W g ! T .g/ and
the surjective superalgebra morphism  W T .g/ ! U.g/ D T .g/=I . We shall see later
that this j is injective.
We are now ready to state the universal property of U.g/.

Proposition 1.6.2. The UESA U.g/ has the following properties:

(1) j.g/ generates U.g/ (as an associative superalgebra).
(2) j.ŒX; Y / D j.X /j.Y /  .1/jXjjY j j.Y /j.X /; in other words, j is a Lie su-
peralgebra morphism.
(3) If A is any associative superalgebra and W g ! A is a Lie superalgebra mor-
phism, i.e., .ŒX; Y / D .X/ .Y /  .1/jXjjY j .Y / .X/, then there exists a
unique morphism  of associative superalgebras such that the following diagram
 1.6 The universal enveloping superalgebra 17

is commutative:
U.g/
{{=
j
{{
{{

{{  
g / A.

Proof. (1) j.g/ generates U.g/ since j D  B i and i.g/ generates T .g/.
(2) Since .I / D 0,

j.ŒX; Y /  j.X /j.Y / C .1/jXjjY j j.Y /j.X /

D .i.ŒX; Y /  i.X /i.Y / C .1/jXjjY j i.Y /i.X // D 0:

(3) Define the superalgebra morphism 0 W T .g/ ! A, 0 .i.X // D .X/ for all
X 2 g. Since by definition 0 .I / D 0 we have that 0 factors as  B  for some
 W U.g/ D T .g/=I ! A. The uniqueness of  comes from (1). 

We give now an important application of this result: we prove that a representation

of a Lie superalgebra extends uniquely to a representation of its UESA.

Definition 1.6.3. We define a representation of a Lie superalgebra g in a super vector

space V , as a linear morphism W g ! End.V / preserving parity and the bracket.
We say also that V is a g-module. A g-module V is irreducible if it does not have
non-trivial submodules. We also define a representation of an associative superalgebra
A in V as a superalgebra morphism 0 W A ! End.V /.

Theorem 1.6.4. Let g be a Lie superalgebra and U.g/ its UESA. Let be a represen-
tation of g in a super vector space V . Then extends uniquely to a representation 0
of U.g/ in V .

Proof. Take A D End.V / in Proposition 1.6.2. In fact a representation of g in V is a

morphism of Lie algebras W g ! End.V /. By the universality, extends uniquely to
a representation 0 W U.g/ ! End.V /. 

We turn now to the Poincaré–Birkhoff–Witt Theorem, which allows us to construct

explicitly a basis for the UESA U.g/ starting from a basis of g.
Let fXk ; k 2 Kg be a homogeneous basis of g, K a linearly ordered set. To
simplify the proof and the notation we assume without loss of generality that the
indices corresponding to even elements are smaller, in the given order, than the indices
corresponding to the odd elements. Moreover we will use the same symbol to denote
an element in g and its image via i in T .g/.
U.g/ is spanned by 1 and j.Xk1 /; : : : ; j.Xk t /, for all k1 ; : : : ; k t 2 K. Since

j.Xk /j.Xl / D .1/p.Xk /p.Xl / j.Xl /j.Xk / C Œj.Xk /; j.Xl /

18 1 Z=2Z-graded linear algebra

for Xk , Xl 2 g we expect U.g/ to be spanned by

1; j.Xk1 / : : : j.Xkr /j.XkrC1 / : : : j.XkrCs /;

k1      kr < krC1 <    < krCs ;
Xki even; 1  i  r; Xkj odd; r C 1  j  r C s:

The fact that an odd element X appears only with exponent 1 is due to the relation
j.ŒX; X/ D 2j.X /2 . The Poincaré–Birkhoff–Witt Theorem says that these monomi-
als form a basis for U.g/.
Let T 0  T .g/ denote the linear super space of the standard monomials, that is,
the monomials of the form
Xk1 ˝    ˝ Xkr ˝ XkrC1 ˝    ˝ XkrCs ;
k1      kr < krC1 <    < krCs ;
Xki even; 1  i  r; Xkj odd; r C 1  j  r C s:

Theorem 1.6.5 (Poincaré–Birkhoff–Witt). Let g be a Lie superalgebra, U.g/ its UESA,

fXk ; k 2 Kg a homogeneous basis of g, K a linearly ordered set as above. Then

1; j.Xk1 / : : : j.Xkr /j.XkrC1 / : : : j.XkrCs /;

k1      kr < krC1 <    < krCs ;
Xki even; 1  i  r; Xkj odd; r C1j r Cs

is a basis for U.g/.

Proof. The proof is very similar to the one in the ordinary setting. The only important
difference comes from the fact that in a Lie superalgebra for X odd ŒX; X  may not be
zero, hence the relation holding in the UESA 2X 2  ŒX; X  will have to be examined
separately.
The statement of the theorem is equivalent to the two statements:
(1) I C T 0 D T .g/.
(2) I \ T 0 D .0/.
(1) Let Tm denote the span of the monomial tensors of degree m and Tm0 D T 0 \Tm .
It is enough to prove that X
Tm  I C Tq0 :
0qm

We do this by induction on m. The cases m D 0; 1 are clear. If t D Xk1 ˝    ˝ Xku ,

we denote by ind.t / the number of pairs .p; q/ with 1  p; q  u such that p < q,
but kp  kq if Xkp and Xkq are both odd, p < q but kp > kq otherwise. Clearly the
P/ D 0. Let
elements t 2 T 0 have ind.t Tmd D ft 2 Tm j ind.t / D d g. It is enough
to show that Tm  I C 0nm Tn . We use induction on d . The case d D 0 is
d 0
 1.6 The universal enveloping superalgebra 19

clear. Fix t D Xk1 ˝    ˝ Xkm 2 Tmd . If d > 0, there exists v, 1  v  m  1,

such that kv  kvC1 , if Xkv and XkvC1 are odd, kv > kvC1 otherwise. Assume first
that Xkv ¤ XkvC1 . Let t 0 D Xl1 ˝    ˝ Xlm be defined as follows: li D ki for all
i ¤ v; v C 1 and lv D kvC1 , lvC1 D kv . Then t 0 2 Tmd 1  I C T 0 , by induction
hypothesis. Since
p.Xkv /p.XkvC1 /
Xkv ˝ XkvC1  .1/ XkvC1 ˝ Xkv D ŒXkv ; XkvC1  mod I
P
we have t  t 0 2 I C Tm1  I C 1nm1 Tn0 , which by induction concludes our
argument in case Xkv ¤ XkvC1 . Assume now that Xkv D XkvC1 odd. Since Xkv ˝
P
Xkv D .1=2/ŒXkv ; Xkv  mod I , we have that t 2 I C Tm1  I C 1nm1 Tn0 ,
which again by induction concludes our argument.
(2) Our strategy is the following. We shall construct an endomorphism L W T .g/ !
T .g/ such that:
(i) L.t/ D t for all standard monomials.
(ii) If 1  v  p  1, p  2, kv > kvC1 , then

L.Xk1 ˝    ˝ Xkv ˝ XkvC1 ˝    ˝ Xkp /

jXkv jjXkvC1 j
D .1/ L.Xk1 ˝    ˝ XkvC1 ˝ Xkv ˝    ˝ Xkp /
C L.Xk1 ˝    ˝ ŒXkv ; XkvC1  ˝    ˝ Xkp /:

If 1  v  p  1, p  2, kv D kvC1 and Xkv odd, then

L.Xk1 ˝    ˝ Xkv ˝ XkvC1 ˝    ˝ Xkp /

D 12 L.Xk1 ˝    ˝ ŒXkv ; XkvC1  ˝    ˝ Xkp /:

If we succeed in building such an endomorphism we are done since L is the identity

on T 0 and it is zero on I . We will define L by induction on p, the degree of the
monomials in T .g/. Define L to be the identity on T0 and T1 . Now assume that p > 1.
We want to define L on t D Xk1 ˝    ˝ Xkp in such a way that (i) and (ii) are satisfied.
We proceed by induction on d D ind.t /. If d D 0 we define L.t / D t . Assume that
d > 0 and that we have defined L on every monomial s of degree p and ind.s/ < d .
Since d > 0 there exists an integer r, 1  r  p  1 (not unique) such that kr > krC1
or kr D krC1 with Xkr odd. We can then use the right-hand side of (ii) (after choosing
the appropriate case) to define L.t / and clearly such L is what we want and is defined
by induction on d and on p. We have now to show that such an L is well defined,
i.e., the definition is independent from the choice of the integer r. Assume that we
first choose the integer r and then another integer l and obtain, following our recipe, a
certain expression a for L.t /. Then we proceed by first choosing the integer l and then
the integer r and obtain another expression b for L.t /. We want to show that a D b.
Whenever jr  lj  2 it is easy to convince ourselves that a D b, since there will be no
“interference” between the two definitions of L. The problems appear when jr lj D 1.
20 1 Z=2Z-graded linear algebra

Assume without loss of generality that r D l  1. To ease the notation let Xir D X ,
XirC1 D Y , XirC2 D Z, so that we have t D Xk1 ˝    ˝ X ˝ Y ˝ Z ˝    ˝ Xkp .
Let us first assume that no two of X , Y , Z are odd and equal. After some calculations
one finds that the two possible ways of defining L.t / give the following two results:

a D .1/jXjjY jCjZjjXjCjY jjZj L.: : : Z ˝ Y ˝ X : : : /

C .1/jXjjY jCjZjjXj L.: : : ŒY; Z ˝ X : : : /
C .1/jXjjY j L.: : : Y ˝ ŒX; Z : : : / C L.: : : ŒX; Y  ˝ Z : : : /;

b D .1/jXjjY jCjZjjXjCjY jjZj L.: : : Z ˝ Y ˝ X : : : /

C .1/jY jjZjCjZjjXj L.: : : Z ˝ ŒX; Y  : : : /
C .1/jZjjY j L.: : : ŒX; Z ˝ Y : : : / C L.: : : X ˝ ŒY; Z : : : /:

Hence we have

a  b D L.    ŒX; ŒY; Z C ŒŒX; Y ; Z C .1/jXjjY j ŒY; ŒX; Z : : : / D 0

because of the Jacobi identity. Hence there is no ambiguity in our definition of L.

Let us now assume that two of X, Y , Z are equal and odd. We can assume without
loss of generality to have t D Xk1 ˝    ˝ X ˝ X ˝ Y ˝    ˝ Xkp , the other cases
being the same. After some calculations we have

a D 12 L.: : : Y ˝ ŒX; X  : : : / C .1/jXjjY j L.: : : ŒX; Y  ˝ X : : : /

C L.: : : X ˝ ŒX; Y  : : : /;

b D 12 L.: : : Y ˝ ŒX; X  : : : / C 12 L.: : : ŒŒX; X ; Y  : : : /:

Hence
a  b D L.: : : ŒX; ŒX; Y   12 ŒŒX; X ; Y  : : : / D 0:

Again by the Jacobi identity it follows that ŒX; ŒX; Y  C .1/jXjjY j ŒX; ŒY; X  C
ŒY; ŒX; X D 2ŒX; ŒX; Y   ŒŒX; X ; Y  D 0. 

Corollary 1.6.6. The Lie superalgebra morphism j W g ! U.g/ described above is

an injection.

To ease the notation, from now on we identify g with its image j.g/ in U.g/.

Corollary 1.6.7. We have the linear isomorphism

V
U.g/ Š U.g0 / ˝ .g1 /;
V
where .g/ is the exterior algebra over the (ordinary) vector space g1 .
 1.6 The universal enveloping superalgebra 21

Next we show that U.g/ has a filtered algebra structure and we relate it to Sym.g/
the symmetric algebra over the super vector space g.
From now on assume that g is finite-dimensional. Let us first recall some basic
definitions.

Definition 1.6.8. An associative superalgebra A is graded if for all integers n  0, we

have a subspace A.n/  A such that
(1) 1 2 A.0/ ,
(2) A.m/ A.n/  A.mCn/ ,
L
(3) A D n0 A.n/ .
The elements in A.n/ are said to be Z-homogeneous of degree n or just homogeneous
when there is no ambiguity. WeLalso say that the gradingL is compatible with the
superalgebra structure if A0 D n even A.n/ , A1 D n odd A.n/ . All gradings we
consider have this property.
We say that an associative superalgebra B is filtered if for all integers n  0, we
have a subspace B n  A such that
(i) 1 2 B 0 ,
(ii) B n B m  B mCn ,
S
(iii) B 0  B 1      n0 B n D B.

L Any graded superalgebra A can be viewed as a filtered superalgebra defining A D

n

0in A.i/ , though not all filtered algebras arise in this way. Vice versa to any
filtered superalgebra B, we can associate a graded superalgebra
L Gr.B/ in the following
way. Define Gr.B/.n/ ´ B n =B n1 and Gr.B/ ´ n0 Gr.B/.n/ as vector super
spaces. Let n W B n ! Gr.B/.n/ be the natural projection. Given a 2 Gr.B/.m/ and
b 2 Gr.B/.n/ choose a0 2 B m and b 0 2 B n such that m .a0 / D a and n .b 0 / D b.
Then define the product ab ´ mCn .a0 b 0 / 2 Gr.B/.mCn/ . One can check this is well
defined and that Gr.B/ is a graded superalgebra.
There are two examples of graded and filtered algebras that are of special interest
to us: the tensor superalgebra T .g/ and the UESA U.g/. T .g/ is graded by taking
T .g/.n/ as the tensors
L of degree n. As any graded algebra, T .g/ is also filtered by
taking T .g/n D 0in T .g/.i/ .
Since the ideal I  T .g/ generated by the elements X ˝ Y  .1/jXjjY j Y ˝ X 
ŒX; Y  in the tensor superalgebra T .g/ is not homogeneous (i.e. is not generated by
homogeneous elements) we cannot expect U.g/ to be graded. However it is filtered
with U.g/n ´ .T .g/n / (recall  W T .g/ ! T .g/=I Š U.g/). The next proposition
clarifies the structure of U.g/ as filtered superalgebra.

Proposition 1.6.9. Let X1 ; : : : ; XM ; XM C1 ; : : : ; XN be a basis for the Lie superalge-

bra g, X1 : : : XM even, XM C1 : : : XN odd. Then:
(i) 1, Xk1 ; : : : ; Xkr ; XkrC1 ; : : : ; XkrCs , k1      kr < krC1 <    < krCs , Xki
22 1 Z=2Z-graded linear algebra

even for 1  i  r, Xkj odd for r C 1  j  r C s, is a basis for U.g/n ,

r C s  n.
(ii) Moreover Gr.U.g// is commutative and the natural map g ! Gr.U.g// extends
to an algebraic isomorphism of Sym.g/ onto Gr.U.g//.
Proof. (i) By induction on n D r C s. The cases n D 0; 1 are clear. Let n > 1. Since
Xj1 ˝    ˝ Xjl , l  n, generate T .g/n , we have that Xj1 : : : Xjl , l  n, generate
U.g/n . By PBW Theorem 1.6.5 the standard monomials are linearly independent, so
we only need to prove they generate U.g/n . While proving the PBW Theorem we
showed that X
T .g/n  I C Tr0 ;
1rn
Pn
hence U.g/ n 0
rD1 .Tr /,which is what we want.
(ii) To prove Gr.U.g// is commutative, notice that since
X Y  .1/jXjjY j YX D ŒX; Y  2 g  U.g/;
we have
Xk1 : : : Xkj  .1/u X .k1 / : : : X.kr / 2 U.g/r1
for a suitable u. Hence if a 2 U.g/p and b 2 U.g/q , then
ab  ba; mod U.g/pCq1 ;
and this proves commutativity.
Now we build the isomorphism Sym.g/ Š Gr.U.g//. Since U.g/1 D k  1 ˚ g, the
natural map U.g/1 ! Gr.U.g// induces an injection g ! Gr.U.g//. Since Gr.U.g//
is commutative, we have a superalgebra morphism  W Sym.g/ ! Gr.U.g//. The
elements Xk1 ; : : : ; Xkn are linearly independent in U.g/n modulo U.g/n1 (the order
on the k’s is specified above), hence .Xk1 ; : : : ; Xkn / D Xk1 : : : Xkn are linearly
independent and form a basis of Gr.U.g//.n/ . By comparing dimensions we have that
 is an isomorphism. 
An important application of Proposition 1.6.9 is the construction of the symmetrizer
map, when k is of characteristic 0.
Let char.k/ D 0.
Consider the natural action of the permutation group Sn on T .g/.n/ :

s  X1 ˝    ˝ Xn ´ .1/p.s/ Xs 1 .1/ ˝    ˝ Xs 1 .n/

where
p.s/ ´ jf.k; l/ j k > l; Xk ; Xl odd ; s.k/ < s.l/gj:
Notice that p.s/ has the following interpretation. It is the (unique) function p W Sn !
ZC such that if X1 ; : : : ; Xn 2 Sym.g/:
Xs 1 .1/ : : : Xs 1 .n/ D .1/p.s/ X1 : : : Xn ; s 2 Sn :
 1.6 The universal enveloping superalgebra 23

Definition 1.6.10. Let t 2 T .g/.n/ . We say that t is symmetric if s  t D t , for all

s 2 Sn . We denote by T .g/.n/ the homogeneous symmetric tensors of degree n and
L
with T .g/ D n0 T .g/.n/ .

Proposition 1.6.11. The linear morphism Qn W T .g/ ! T .g/ defined as

1 X
Q0 D 1; Qn .t / D st
nŠ
s2Sn

is a projection onto the symmetric tensors T .g/.

Proof. Direct check. 

Lemma 1.6.12. Let the notation be as above and let  W T .g/ ! U.g/.
(1)  W T .g/ ! U.g/ is a linear isomorphism preserving the filtration, that is,
.T .g/n / D U.g/n .
(2) T .g/ D I C T .g/.

Proof. By induction on the Z-grading of T .g/. For n D 0 we have that  clearly

induces an isomorphism of T .g/0 with U.g/0 . Let t D X1 ˝  ˝Xn ,  D .i; i C1/ 2
Sn . We have

t    t D X1 ˝ : : : Xi ˝ XiC1 ˝    ˝ Xn
.1/jXi jjXi C1 j X1 ˝    ˝ XiC1 ˝ Xi    ˝ Xn
D X1 ˝    ˝ ŒXi ; XiC1     ˝ Xn C I 2 T .g/n1 C I:

Since any s 2 Sn is the product of adjacent transpositions we have u  s  u 2

T .g/n1 C I for all u 2 T .g/n . Averaging over Sn we get

u  Qn .u/ 2 T .g/n1 C I:

Applying  to this relation we get

U.g/n  U.g/n1 C .T .g/n /:

By induction, comparing dimensions, we get the result.

(2) is a consequence of (1). 

Let gN be an abelian Lie superalgebra with the same underlying super vector space
as g. Clearly U.g/N D Sym.g/. Let N W T .g/
N ! U.g/. N J D ker N is the ideal in T .g/
generated by the elements

X ˝ Y  .1/jXjjY j Y ˝ X:
24 1 Z=2Z-graded linear algebra

Lemma 1.6.13. Let Sym.g/.n/ be the space of elements of degree L n in Sym.g/.

Consider the filtration of the symmetric superalgebra Sym.g/n D in Sym.g/.i/ .
Then N W T .g/ ! Sym.g/ is a linear isomorphism preserving the filtration, i.e.,
N .g/n / D Sym.g/n .
.T

Proof. As in Lemma 1.6.12. 

We are ready to define the symmetrizer map.

Definition 1.6.14. We define the symmetrizer of Sym.g/ onto U.g/ as the unique linear
map S making the following diagram commute:

T .g/
 / U.g/
u :
S uu
u
u
N
uu
 uu
Sym.g/:

Theorem 1.6.15. The symmetrizer map S is a linear isomorphism and preserves the
filtration; moreover
1 X
S..X N n // D
N 1 / : : : .X .1/p.s/ Xs.1/ : : : Xs.n/ :
nŠ
s2Sn

Proof. The fact that S is a linear isomorphism and preserves the filtration is clear by
definition. For its expression observe that
1 X
.Qn .X1 ˝    ˝ Xn // D .1/p.s/ Xs.1/ : : : Xs.n/
nŠ
s2Sn

and
N n .X1 ˝    ˝ Xn // D .X
.Q N 1 / : : : .X
N n /:
By the commutativity of the above diagram we are done. 

In conclusion of this section, we want to remark that in [22] the authors take a
radically different point of view in proving the statement of the PBW Theorem. They
use the symmetrizer linear isomorphism to transfer the product structure from U.g/ to
a new product structure in Sym.g/, and then they prove that .Sym.g/; / Š U.g/ as
superalgebras. Their proof holds over a field of characteristic zero, however it is more
general in the sense that it holds for a Lie algebra object in an arbitrary tensor category.
Another different and more general proof than the one that we present is found also
in [76]. There, the statement is given for a Lie superalgebra over k commutative ring
with 1, where 2 and 3 are invertible elements. In this case, the technique of the proof
makes essential use of the “Diamond Lemma” by G. Bergman [12].
 1.7 Hopf superalgebras 25

1.7 Hopf superalgebras

In this section we briefly discuss Hopf superalgebras and some examples taken from
ordinary geometry. As we shall see, Hopf superalgebras are important in the under-
standing of Lie supergroups and affine algebraic supergroups since they represent an
alternative way to approach and discuss them.

Definition 1.7.1. We say that the superalgebra A (not necessarily commutative) is a

Hopf superalgebra if A has the following properties.
1. A is a superalgebra, with multiplication W A ˝ A ! A and unit i W k ! A.
2. A is a super coalgebra, that is, we can define two linear maps called comultipli-
cation W A ! A ˝ A and counit  W A ! k with the following properties:

A ˝O A
id˝ /A˝k A ˝O A
˝id /k˝A A ˝O A
˝id /A˝A˝A
O O O
 Š  Š  id˝

A
id / A, A
id / A, A
 / A.

A morphism  W A ! B of two super coalgebras with comultiplication and counit A ,

B and A , B respectively is a linear map such that . ˝/ A D B , B  D A .

3. The multiplication and the unit i are super coalgebra morphisms.

4. The comultiplication and the counit  are superalgebra morphisms.
5. A is equipped with a bijective linear map S W A ! A called the antipode such
that the following diagrams commute:

A ˝O A
S˝id /A˝A A ˝O A
id˝S /A˝A
O O
   

A
i / A, A
i / A.

A superalgebra A satisfying the first four properties is called a super bialgebra.

A Hopf superalgebra morphism is a linear map  W A ! B which is a morphism

of both the superalgebra and super coalgebra structures of A and B, and in addition

SB   D   SA ;

where SA and SB denote respectively the antipodes in A and B.

We say that I is a Hopf ideal of a Hopf superalgebra A if I is a two-sided ideal,
and moreover

.I /  I ˝ A C A ˝ I; .I / D 0; S.I /  I:
26 1 Z=2Z-graded linear algebra

One can check immediately that the superalgebra A=I inherits naturally a Hopf super-
algebra structure from A.
I is a coideal of a coalgebra A if it is an abelian subgroup of A and
.I /  I ˝ A C A ˝ I:
There are many interesting examples of Hopf algebras; we refer the reader to [20],
[59], [72] for a comprehensive treatment.
In particular to any affine algebraic group in ordinary geometry, we can associate
two very important Hopf algebras: O.G/, the commutative Hopf algebra of algebraic
functions on G, and U.Lie.G//, the universal enveloping algebra of the Lie algebra
Lie.G/. In the case of O.G/ the comultiplication and the counit  are given as
follows:
.f /.x ˝ y/ ´ f .xy/; .f / D f .1G /:
In the case of U.Lie.G// the comultiplication and the counit  are given on the
generators g as follows:
.X / D X ˝ 1 C 1 ˝ X; .X / D 0 for all X 2 g:
Under suitable conditions, these two Hopf algebras are in duality with each other, as
is explained in [20], Ch. 7.

1.8 The even rules

We conclude our treatment of linear supergeometry by discussing the even rules.
The even rules are a device that enables us to work with purely even objects and
has a variety of applications in supergeometry. In order to use them, however, we have
to show the functoriality of the construction we are interested in, under extension of
scalars. We shall not take advantage systematically of the even rules, to go back and
forth between super (i.e., Z2 -graded) and purely even objects; however, since the even
rules are often employed in the physics literature, sometimes implicitly, we shall briefly
outline the key ideas behind them. For a more detailed account we invite the reader to
look at [22], p. 57.
Definition 1.8.1. Let V a super vector space and let A be a commutative superalgebra.
We define
VA ´ A ˝ V; V .A/ ´ .A ˝ V /0 ;
where VA is an A-module, while V .A/, the even part of VA , is an A0 -module. We also
say that VA is obtained by V .A/ by extension of scalars.
It is clear that if we have a linear morphism f W V1 ˝ V2 ! V , such a morphism
induces an A0 -linear map f .A/ W V1 .A/ ˝ V2 .A/ ! V .A/, (or if we prefer an A0 -
multilinear map .f1 .A/; f2 .A// W V1 .A/  V2 .A/ ! V .A/) which is functorial in A.
The next proposition establishes a converse for this fact.
 1.9 References 27

Proposition 1.8.2. Any A0 -multilinear map .f1 .A/; f2 .A// W V1 .A/  V2 .A/ ! V .A/
functorial in A comes from a unique linear morphism f W V1 ˝ V2 ! V .
Proof (Sketch). We start by proving uniqueness. We want to determine f .v1 ˝ v2 /
by the knowledge of the multilinear morphism f .A/ D .f1 .A/; f2 .A//, for A D
kŒ1 ; 2 . Define the element in V .A/:
´
vi if vi even;
wi D
i vi if vi odd:
For concreteness assume that v1 even and v2 odd. Then we have
f .A/.w1 ; w2 / D f .A/.v1 ; 2 v2 / D 2 f .v1 ˝ v2 /:
This formula determines uniquely f .v1 ˝ v2 /. This formula moreover also gives us
existence since it provides a way to define f .v1 ˝ v2 /. 
Remark 1.8.3. (1) There is an obvious generalization of the previous result to the case
of linear morphisms f W ˝i2I Vi ! V . The reasoning is similar, and we invite the
reader to consult [22] for more details.
(2) The previous result holds if A is just taken to be an exterior algebra, A D
kΠ1 : : : n , in other words, we can substantially weaken our hypothesis. This is clear
by looking at the proof.
(3) The same result holds also for modules over some commutative superalgebra
R replacing our field k while A runs over commutative R-superalgebras and V .A/ is
the A0 -module .A ˝R V /0 as before. The proof goes practically unchanged.
Proposition 1.8.2 has the following very useful corollary, which allows us to extend
in a straightforward manner algebraic structures from V .A/ to VA .
Corollary 1.8.4. Let the notation be as above. In order to give a superalgebra structure
on a super vector space V , it is enough to give a functorial structure of A0 -algebra on
V .A/. The superalgebra will be Lie, respectively associative or commutative and so
on, if and only if the V .A/ A0 -algebras are.
Proof. The first statement is clear since a superalgebra structure comes precisely from
the morphisms considered in Proposition 1.8.2. As for the Lie, associative and com-
mutative properties, they rely on commutative diagrams constructed starting from such
morphisms. Proposition 1.8.2 ensures the commutativity of such diagrams by the func-
toriality of the corresponding diagrams for V .A/. We leave to the reader all the checks
involved in this construction. 

1.9 References
All of the material appearing here is standard and can be found in one of the references
[56], [76], [10], [49].
2
Sheaves, functors and the geometric point of view

Classically, sheaf theory provides an alternative and very elegant way to look at the-
ories, like differentiable manifolds and algebraic varieties, which originally were in-
troduced using very different methods. On the other hand, in supergeometry, its use
is unavoidable since at the very start we need sheaves in order to define properly any
supergeometric object in differential or algebraic category.
We start our treatment by discussing sheaves of functions in the familiar setting of
differentiable manifolds and algebraic varieties and then we go on to the more abstract
definition of sheaves in general, arriving finally at ringed spaces and locally ringed
spaces, which is our starting point for the development of supergeometry, as we shall
see in the next chapter.
We also give a brief introduction to schemes and their functor of points, discussing
the examples of projective space and Grassmannian variety. The language of the functor
of points and its generalization to supergeometry are extremely important to us since
they provide geometric intuition to an otherwise very abstract setting. In the end, we
recall some results on coherent sheaves, that we shall need in the sequel.
We do not present here proofs for most of our statements, which are all classical and
well known; a treatment of sheaf theory and ringed spaces can be found. for example.
in [43], Ch. II, or in [29], Ch. I, while a comprehensive functorial treatment of algebraic
geometry, via the functor of points, is found in [23].

2.1 Ringed spaces of functions

A ringed space is a broad concept in which we can fit most of the interesting geometrical
objects. It consists of a topological space together with a sheaf of functions on it. Let
us examine some familiar and important examples.
Let M be a differentiable manifold, whose topological space is Hausdorff and
second countable. For each open set U  M , let C 1 .U / be the R-algebra of smooth
functions on U .
The assignment
U 7! C 1 .U /
satisfies the following two properties:
(1) If U  V are two open sets in M , we can define the restriction map, which is
an algebra morphism:
rV;U W C 1 .V / ! C 1 .U /; f 7! f jU ;
 2.1 Ringed spaces of functions 29

which is such that

(i) rU;U D id,
(ii) rW;U D rV;U B rW;V .
(2) Let fUi gi2I be an open covering of U and let ffi gi2I , fi 2 C 1 .Ui /, be a
family such that fi jUi \Uj D fj jUi \Uj for all i; j 2 I . In other words the elements of
the family ffi gi2I agree on the intersection of any two open sets Ui \ Uj . Then there
exists a unique f 2 C 1 .U / such that f jUi D fi .
We leave to the reader, as an exercise, to check that indeed the assignment satisfies
the two given properties. As for property (1), the existence of the restriction map
satisfying equations (i) and (ii) is clear while for the property (2), also-called the
gluing property, it is a simple exercise.
Such an assignment is called a sheaf. The pair .M; C 1 /, consisting of the topo-
logical space M , underlying the differentiable manifold, and the sheaf of the C 1
functions on M is an example of locally ringed space (the word “locally” refers to a
local property of the sheaf of C 1 functions we shall discuss in the next sections). We
shall examine in detail locally ringed spaces and their properties in the next section.
Remark 2.1.1. It is customary to denote a differentiable manifold and its underlying
topological space with the same symbol. In this section, we follow this convention,
however starting from the next section, we shall mark the difference between the
manifold M and its underlying topological space jM j. We shall also do the same for
algebraic varieties.
Given two manifolds M and N , and the respective sheaves of smooth functions
1
CM and CN1 , a morphism f from M to N , viewed as ringed spaces, is a morphism
jf j W M ! N of the underlying topological spaces together with a morphism of
algebras,
f  W CN1 .V / ! CM
1
.f 1 .V //; f  ./.x/ D .jf j.x//; V open in N;
compatible with the restriction morphisms.
Notice that, as soon as we give the continuous map jf j between the topological
spaces, the morphism f  is automatically assigned. This is a peculiarity of the sheaf
of smooth functions on a manifold. Such a property is no longer true for a generic
ringed space and, in particular, as we shall see, it is not true for supermanifolds.
A morphism of differentiable manifolds gives rise to a unique (locally) ringed space
morphism and vice versa.
Moreover, given two manifolds, they are isomorphic as manifolds if and only if they
are isomorphic as (locally) ringed spaces. In the language of categories, we say we
have a fully faithful functor from the category of manifolds to the category of locally
ringed spaces.
Before going to the general treatment, let us consider another interesting example
arising from classical algebraic geometry. As we shall see in the next chapters, the gen-
eralization of algebraic geometry to the super-setting comes somehow more naturally
30 2 Sheaves, functors and the geometric point of view

than the similar generalization of differentiable geometry. This is because the machin-
ery of algebraic geometry was developed to take already into account the presence of
(even) nilpotents and consequently, the language is more suitable to supergeometry.
Let X be an affine algebraic variety in the affine space An over an algebraically
closed field k and let O.X / D kŒx1 ; : : : ; xn =I be its coordinate ring, where the ideal
I is prime. This corresponds topologically to the irreducibility of the variety X . We
can think of the points of X as the zeros of the polynomials in the ideal I in An . X is a
topological space with respect to the Zariski topology, whose closed sets are the zeros
of the polynomials in the ideals of O.X / (see [43], Ch. I, for a complete discussion of
the Zariski topology). For each open U in X , consider the assignment
U 7! OX .U /;
where OX .U / is the k-algebra of regular functions on U . By definition, these are the
functions f W X ! k that can be expressed as a quotient of two polynomials at each
point of U  X .
As in the case of differentiable manifolds, our assignment satisfies the properties
(1) and (2) described above. The first property is clear while for the second one we
refer the reader to [43], Ch. II, Section 2.
The assignment U 7! OX .U / is another example of a sheaf and we shall call it the
structure sheaf of the variety X or the sheaf of regular functions. .X; OX / is another
example of a (locally) ringed space.

2.2 Sheaves and ringed spaces

We are now going to formulate more generally the notion of sheaf and of ringed space
that we have described in two specific examples of the previous section.
Definition 2.2.1. Let jM j be a topological space. A presheaf of commutative algebras
F on X is an assignment
U 7! F .U /; U open in jM j; F .U / a commutative algebra;
such that the following holds:
(1) If U  V are two open sets in jM j, there exists a morphism rV;U W F .V / !
F .U /, called the restriction morphism and often denoted by rV;U .f / D f jU , such
that
(i) rU;U D id,
(ii) rW;U D rV;U B rW;V .
A presheaf F is called a sheaf if the following holds:
(2) Given an open covering fUi gi2I of U and a family ffi gi2I , fi 2 F .Ui / such
that fi jUi \Uj D fj jUi \Uj for all i; j 2 I , there exists a unique f 2 F .U / with
f jUi D fi .
 2.2 Sheaves and ringed spaces 31

The elements in F .U / are called sections over U ; when U D jM j we call such

elements global sections.
The assignments U 7! C 1 .U /, U open in the differentiable manifold M and
U 7! OX .U /, U open in the algebraic variety X , described in the previous section,
are examples of sheaves of functions on the topological spaces jM j and jXj underlying
the differentiable manifold M and the algebraic variety X respectively.
Observation 2.2.2. In the language of categories, property (1) in Definition 2.2.1 says
that we have defined a functor, F , from top.M / to .alg/, where top.M / is the category
of the open sets in the topological space jM j, the arrows given by the inclusions of
open sets while .alg/ is the category of commutative algebras. In fact, the assignment
U 7! F .U / defines F on the objects while the assignment

U  V 7! rV;U W F .V / ! F .U /

defines F on the arrows.

Hence we can give the following equivalent definition of presheaf and sheaf of
commutative algebras.
Definition 2.2.3 (Alternative definition of presheaf and sheaf). Let jM j be a topological
space. We define a presheaf of algebras on jM j to be a functor

F W top.M /op ! .alg/:

The suffix “op” denotes as usual the opposite category; in other words, F is a con-
travariant functor from top.M / to .alg/. A presheaf is a sheaf if it satisfies the property
(2) of Definition 2.2.1.
If F is a (pre)sheaf on jM j and U is open in jM j we define F jU , the ( pre)sheaf F
restricted to U , as the functor F restricted to the category of open sets in U (viewed as
a topological space itself). In other words, we restrict our attention to the assignment
V 7! F .V / for just the open sets V in U .
We have defined sheaves as functors with values in the category of commutative
algebras, however, in the same way, we can also define sheaves with values in commuta-
tive rings, groups, sets or other algebraic structures and their super correspondents. Of
course, depending on the structure of the category of arrival, the restriction morphisms
have to be taken as the appropriate morphisms.
From now on, we shall speak only of “presheaf” and of “sheaf”, without further
specifications, whenever our statements are independent of the category of arrival.
A most important object associated to a given presheaf is the stalk at a point.
Definition 2.2.4. Let F be a presheaf on the topological space jM j and let x be a point
in jM j. We define the stalk Fx of F , at the point x, as the direct limit

lim F .U /;
!
32 2 Sheaves, functors and the geometric point of view

where the direct limit is taken for all the U open neighbourhoods of x in jM j. Fx
consists of the disjoint union of all pairs .U; s/ with U open in jM j, x 2 U , and
s 2 F .U /, modulo the equivalence relation: .U; s/ Š .V; t / if and only if there exists
a neighbourhood W of x, W  U \ V , such that sjW D t jW .
The elements in Fx are called germs of sections.

Definition 2.2.5. Let F and G be presheaves on jM j. A morphism of presheaves

 W F ! G is a collection of morphisms U W F .U / ! G .U /, for each open set U in
jM j, such that for all V  U the following diagram commutes:

F .U /
U
/ G .U /

rU;V rU;V
 
F .V /
V
/ G .V /.

Equivalently and more elegantly, one can also say that a morphism of presheaves is a
natural transformation between the two presheaves F and G viewed as functors (see
the alternative definition of sheaf in Definition 2.2.3).
A morphism of sheaves is just a morphism of the underlying presheaves.

Clearly any morphism of presheaves induces a morphism on the stalks: x W Fx !

Gx . The sheaf property, i.e., property (2) in Definition 2.2.1, ensures that if we have
two morphisms of sheaves  and such that x D x for all x, then  D .
We say that the morphism of sheaves  is injective (resp. surjective) if x is injective
(resp. surjective).
On the notion of surjectivity, however, one should exert some care, since we can
have a surjective sheaf morphism  W F ! G such that U W F .U / ! G .U / is not
surjective for some open sets U . This strange phenomenon is a consequence of the
following fact. While the assignment U 7! ker..U // always defines a sheaf, the
assignment
U 7! im..U // D F .U /=G .U /
defines in general only a presheaf and not all the presheaves are sheaves. A simple
example is given by the assignment associating to an open set U in R the algebra of
constant real functions on U . Clearly this is a presheaf, but not a sheaf.
We can always associate, in a natural way, to any presheaf a sheaf called its sheafi-
fication. Intuitively, one may think of the sheafification as the sheaf that best “approx-
imates” the given presheaf. For example, the sheafification of the presheaf of constant
functions on open sets in R is the sheaf of locally constant functions on open sets in R.
We construct the sheafification of a presheaf using the étalé space, which we also
need in the sequel, since it gives an equivalent approach to sheaf theory.

Let F be a presheaf on jM j. We define the étalé space of F to be

Definition 2.2.6. `
the disjoint union x2jM j Fx . For each open U  jM j and each s 2 F .U / define the
 2.2 Sheaves and ringed spaces 33

`
map sOU W U ! x2jU j Fx , sOU .x/ D sx . We give to the étalé space the finest topology
that makes the maps sOU continuous, for all open U  jM j and all sections s 2 F .U /.
We define Fet to be the presheaf on jM j:
`
U 7! Fet .U / D fOsU W U ! x2jU j Fx ; sOU .x/ D sx 2 Fx g:

As we shall see, the presheaf Fet is actually a sheaf and it provides an explicit
construction of the sheafification of the presheaf F .
Definition 2.2.7. Let F be a presheaf on jM j. A sheafification of F is a sheaf Fz ,
together with a presheaf morphism ˛ W F ! Fz , such that
˛
(1) any presheaf morphism W F ! G , G a sheaf, factors via ˛, i.e., W F ! 
Fz ! G ,
(2) F and Fz are locally isomorphic, i.e., there exists an open cover fUi gi2I of jM j
such that F .Ui / Š Fz .Ui / via ˛.
The next proposition establishes the existence of the sheafification of a presheaf.
Proposition 2.2.8. Let F be a presheaf on jM j.
(1) The sheafification Fz of F always exists and it is unique.
(2) The presheaf Fet is a sheaf and it is the sheafification of the presheaf F with
˛ W F ! Fet defined as ˛U .s/.x/ D sx , s 2 F .U /, x 2 U .
In particular F is a sheaf if and only if F Š Fz D Fet .
Proof. See [43], p. 64. 
The previous discussion enables us to properly define the quotient of sheaves.
Definition 2.2.9. Let F and G be sheaves of rings on some topological space jM j.
Assume that we have an injective morphism of sheaves G ! F such that G .U / 
F .U / for all U open in jM j. We define the quotient F =G to be the sheafification of
the image presheaf: U 7! F .U /=G .U /.
In general, .F =G /.U / ¤ F .U /=G .U /, however they are locally isomorphic by
Definition 2.2.7.
There is a most effective way to define a sheaf and a sheaf morphism, which proves
to be very useful for applications. We just state the results, referring to [29], Ch. I, for
more details.
Definition 2.2.10. Let B be a base for the open sets in the topological space jM j. The
assignment U 7! F .U /, for all U 2 B, is called a B-sheaf if it satisfies the condition
(1) as in Definition 2.2.1 for the open sets in B and the following modification of the
condition (2). For all open U in B, let fUi gi2I be a covering in B of U and ffi gi2I a
family such that fi 2 F .Ui /. If fi jV D fj jV for all V  Ui \ Uj , V 2 B, then there
exists a unique f 2 F .U / such that f jUi D fi .
34 2 Sheaves, functors and the geometric point of view

Proposition 2.2.11. Let B be a base for the open sets in the topological space jM j.
(1) Every B-sheaf extends uniquely to a sheaf on jM j.
(2) If G and H are two sheaves on jM j and for all U in B we have a collection of
morphisms
U W G .U / ! H .U /

commuting with restrictions, then there is a unique sheaf morphism  W G ! H such

that U D U for all U 2 B.

Proposition 2.2.12. Let fUi gi2I be an open covering of the topological space jM j.
Assume the following:
(1) We have defined sheaves FUi for all i .
(2) Ui Uj W FUi jUi \Uj ! FUj jUi \Uj are isomorphisms satisfying the compatibility
conditions

Ui Uj Uk Ui D Uk Uj on Ui \ Uj \ Uk for all i; j; k 2 I:

Then there exists a unique sheaf F on jM j such that F jU D FUi .

We are ready to define ringed spaces.

Definition 2.2.13. We define ringed space to be a pair M D .jM j; F / consisting of

a topological space jM j and a sheaf of commutative rings F on jM j. We say that the
ringed space .jM j; F / is a locally ringed space, if the stalk Fx is a local ring for all
x 2 jM j.
A morphism of ringed spaces  W M D .jM j; F / ! N D .jN j; G / consists of a
morphism jj W jM j ! jN j of the topological spaces (in other words, jj is a continuous
map) and a sheaf morphism   W ON !  OM where  OM is the sheaf on jN j defined
as follows: . OM /.U / D OM . 1 .U // for all U open in jN j. A morphism of ringed
spaces induces a morphism on the stalks for each x 2 jM j: x W ON;j j.x/ ! OM;x .
If M and N are locally ringed spaces, we say that the morphism of ringed spaces
 is a morphism of locally ringed spaces if x is local, i.e., x1 .mM;x / D mN;j j.x/
where mN;j j.x/ and mM;x are the maximal ideals in the local rings ON;j j.x/ and OM;x ,
respectively.

Observation 2.2.14. In the previous section we have seen differentiable manifolds

and algebraic varieties as examples of ringed spaces. Actually both are also examples
of locally ringed spaces, as one can readily verify. Moreover, one can also check that
their morphisms, in the differential or in the algebraic setting respectively, correspond
precisely to morphisms as locally ringed spaces.
At this point it is not hard to convince ourselves that we can take a different point of
view on the definition of differentiable manifold. Namely we can equivalently define
a differentiable manifold as a locally ringed space M D .jM j; OM / as follows.
 2.3 Schemes 35

Definition 2.2.15. Alternative definition of differentiable manifold. Let M be a topo-

logical space, Hausdorff and second countable, and let OM be a sheaf of commutative
algebras on M , so that .M; OM / is a locally ringed space. We say that .M; OM / is
a real differentiable manifold if it is locally isomorphic to the locally ringed space
.Rn ; CR1n /, where CR1n is the sheaf of smooth functions on Rn .

In the same way we can define analytic real or complex manifolds as locally ringed
spaces locally isomorphic to .Rn ; HRn / or .Cn ; HCn /, where HRn and HCn denote
the sheaves of analytic functions on Rn or Cn respectively (we leave to the reader as
an exercise their definition, see [37] for more details).
It is important to keep this point of view in mind since this is the path we are going
to take to define supermanifolds.
At this point, it is clear that we could also give an equivalent definition of algebraic
variety in order to fit it into the framework of locally ringed spaces. However, as we shall
see in the next section, we prefer to give a far reaching generalization of this picture,
namely the notion of scheme, which turns out to be the best for our supergeometric
applications.

2.3 Schemes
The concept of scheme is a step towards a further abstraction. We shall start by defining
affine schemes and then we proceed to the definition of schemes in general. Our
treatment is necessarily very short, for all the details and the complete story, we refer
the reader to [29], Ch. I, and [43], Ch. II.
Let us start by associating to any commutative ring A its spectrum, that is the
topological space Spec A. As a set, Spec A consists of all the prime ideals in A. For
each subset S  A we define as closed sets in Spec A:

V .S/ ´ fp 2 Spec A j S  pg  Spec A:

One can check that this actually defines a topology on Spec A called the Zariski topol-
ogy.
If X is an affine variety, defined over an algebraically closed field, and O.X / is
its coordinate ring, we have that the points of the topological space underlying X
are in one-to-one correspondence with the maximal ideals in O.X /. So we notice
immediately that Spec O.X / contains far more than just the points of the topological
space of X; in fact it contains also all the subvarieties of X , whose information is
encoded by the prime ideals in O.X /. This tells us that the notion of scheme, we are
about to introduce, is not just a generalization of the concept of algebraic variety, but it
is something deeper, containing more information about the geometric objects we are
interested in.
36 2 Sheaves, functors and the geometric point of view

We also define the basic open sets in Spec A as

Uf ´ Spec A n V .f / D Spec Af with f 2 A;

where Af D AŒf 1  is the localization of A obtained by inverting the element f .

The collection of the basic open sets Uf , for all f 2 A, forms a base for the Zariski
topology.
Next, we define the structure sheaf OA on the topological space Spec A. In order
to do this, it is enough to give an assignment U 7! OA .U / for each basic open set
U D Uf in Spec A, by Proposition 2.2.11.
Proposition 2.3.1. Let the notation be as above. The assignment

Uf 7! Af

defines a B-sheaf on the topological space Spec A and it extends uniquely to a sheaf of
commutative rings on Spec A, called the structure sheaf and denoted by OA . Moreover
the stalk at a point p 2 Spec A, OA;p is the localization Ap of the ring A at the prime p.
Proof. Direct check. (See also [29], I-18.) 
Hence, given a commutative ring A, Proposition 2.3.1 tells us that the pair
.Spec A; OA / is a locally ringed space that we call Spec A, the spectrum of the ring A.
By an abuse of notation we shall use the word spectrum to mean both the topological
space Spec A and the locally ringed space Spec A, the context making clear which one
we are talking about.
We are finally ready for a definition of scheme. While the differentiable manifolds
are locally modelled, as ringed spaces, by .Rn ; CR1n /, the schemes are geometric objects
modelled by the spectrums of commutative rings.
Definition 2.3.2. We define an affine scheme to be a locally ringed space isomorphic to
Spec A for some commutative ring A. We say that X is a scheme if X D .jX j; OX / is
a locally ringed space, which is locally isomorphic to affine schemes. In other words,
for each x 2 jX j, there exists an open set Ux  jX j such that .Ux ; OX jUx / is an affine
scheme. A morphism of schemes is just a morphism of locally ringed spaces.
Observation 2.3.3. (1) There is an equivalence of categories between the category
of affine schemes .aschemes/ and the category of commutative rings .rings/. This
equivalence is defined on the objects by

.rings/op ! .aschemes/; A 7! Spec A:

In particular a morphism of commutative rings A ! B corresponds contravariantly to

a morphism Spec B ! Spec A of the corresponding affine superschemes. For more
details see [43], Ch. II, Proposition 2.3, and [29], Ch. I, Theorem I-40. We are going
to discuss in detail a generalization of this result in Chapter 10.
 2.3 Schemes 37

(2) Since any affine variety X is completely described by the knowledge of its
coordinate ring O.X /, (the ring of regular functions on the whole variety), we can
associate uniquely to an affine variety X the affine scheme Spec O.X /. As we noted
previously, the two notions of X as algebraic variety or as a scheme are different,
however they describe the same geometrical object from two different points of view.
Similarly to any algebraic variety (not necessarily affine) we can associate uniquely
a scheme. Moreover a morphism between algebraic varieties determines uniquely a
morphism between the corresponding schemes. In the language of categories, we say
we have a fully faithful functor from the category of algebraic varieties to the category
of schemes. For more details see [43], Ch. II, Proposition 2.3, and [29], Ch. I. We shall
show this proposition in Chapter 10 in the more general setting of supergeometry.

In the next example we describe the simplest example of a non-affine scheme: the
projective space.

Example 2.3.4 (Proj S). Let S be a graded commutative k-algebra, k a field, i.e.,
L
S D i0 Si , Si Sj  SiCj , S0 D k. The elements in Si are called homogeneous
elements of degree i . We define Proj S as the set of all relevant homogeneous
L prime
ideals (i.e., prime ideals generated by homogeneous elements, p ¤ i>0 Si ). Proj S
is a topological space with the closed sets defined as

V .I / D fp 2 Proj S j p I g:

Define Wf as the open set

Wf D j Proj Sj  V .f /;

where .f / is the ideal generated by the homogeneous element f . The points of Wf

can be identified with the homogeneous primes of S Œf 1 , which in turn correspond to
the primes in the ring of the elements of degree zero in S Œf 1 , denoted by S Œf 1 0 .
Hence we can identify Wf with the topological space Spec S Œf 1 0 , which has an
affine scheme structureL by its very definition. If x0 ; : : : ; xn are elements of degree 1
that generate the ideal i>0 Si , then the open sets

.Proj S/i ´ Proj S  V .xi /

cover Proj S.
It is not hard to see, using Proposition 2.2.12, that the structure sheaves on each
.Proj S /i extend uniquely and compatibly to a sheaf on Proj S to give a locally ringed
space that we denote by Proj S. By its very construction the locally ringed space Proj S
is locally isomorphic to the affine schemes Spec S Œxi1 0 , hence it is a scheme.
If S D kŒx0 ; : : : ; xn  is the graded polynomial ring, we also write P n for the scheme
Proj S and we call it the projective space of dimension n. (For more details on this
construction see [29], p. 97.)
38 2 Sheaves, functors and the geometric point of view

2.4 Functor of points

When we are dealing with classical manifolds and algebraic varieties, we can altogether
avoid the use of their functor of points. In fact, both differentiable manifolds and
algebraic varieties are well understood just by looking at their underlying topological
spaces and the regular functions on the open sets.
However, if we go to the generality of schemes, the extra structure overshadows
the topological points and leaves out crucial details so that we have little information,
without the full knowledge of the sheaf. We shall see that the same happens for
supergeometric objects. For example the evaluation of odd functions on topological
points is always zero. This implies that the structure sheaf of a supermanifold cannot
be reconstructed from its underlying topological space. For now let us continue our
treatment of ordinary geometry.
The functor of points is a categorical device to bring back our attention to the points
of a scheme; however the notion of point needs to be suitably generalized to go beyond
the points of the topological space underlying the scheme.
Grothendieck’s idea behind the definition of the functor of points associated to a
scheme is the following. If X is a scheme, for each commutative ring A, we can define
the set of the A-points of X in analogy to the way the classical geometers used to define
the rational or integral points on a variety. The crucial difference is that we do not focus
on just one commutative ring A, but we consider the A-points for all commutative rings
A. In fact, the scheme we start from is completely recaptured only by the collection of
the A-points for every commutative ring A, together with the admissible morphisms.
Let .rings/ denote the category of commutative rings and .schemes/ the category
of schemes.
Definition 2.4.1. Let .jX j; OX / be a scheme and let T 2 .schemes/. We call the T -
points of X , the set of all scheme morphisms fT ! X g, that we denote by Hom.T; X /.
We define the functor of points hX of the scheme X as the representable functor defined
on the objects as

hX W .schemes/op ! .sets/; hX .T / D Hom.T; X /:

Hence hX .T / are the T -points of the scheme X . The restriction of hX to affine schemes
is not in general representable. However, since, as we noticed in Observation 2.3.3,
the category of affine schemes is equivalent to the category of commutative rings we
a
have that such restriction gives a new functor hX :
a
hX W .rings/ ! .sets/; a
hX .A/ D Hom.Spec A; X / D A-points of X :

Notice that when X is affine, X Š Spec O.X / and we have

a
hX .A/ D Hom.Spec A; O.X // D Hom.O.X /; A/:
a
In this case the functor hX is again representable.
 2.4 Functor of points 39

This is a consequence of the following proposition, which comes from Observa-

tion 2.3.3.
Proposition 2.4.2. Consider the affine schemes X D Spec O.X / and Y D Spec O.Y /.
There is a one-to-one correspondence between the scheme morphisms X ! Y and the
ring morphisms O.Y / ! O.X /.
Both hX and hX a
are defined on morphisms in the natural way. If  W T ! S is a
morphism and f 2 Hom.S; X / we define hX ./.f / D f B . Similarly if W A ! B
is a ring morphism and g 2 Hom.O.X /; A/ we define hX a
. /.g/ D B g.
a
The next proposition tells us that the functors hX and hX , for a given scheme X,
are not really different, but carry the same information.
Proposition 2.4.3. The functor of points hX of a scheme X is completely determined
by its restriction to the category of affine schemes or equivalently by the functor
a
hX W .rings/ ! .sets/; a
hX .A/ D Hom.Spec A; X /:

Proof. See [29], Ch. VI. 

Example 2.4.4 (The affine space). Let An be the affine space over a field k; its co-
ordinate ring is kŒx1 ; : : : ; xn , the ring of polynomials. Its functor of points is by
definition haAn W .rings/ ! .sets/, haAn .A/ D Hom.Spec A; An /. Since a morphism
of two affine varieties corresponds contravariantly to a morphism of their coordinate
rings (see Observation 2.3.3), we have Hom.Spec A; An / D Hom.kŒx1 ; : : : ; xn ; A/.
Any morphism  W kŒx1 ; : : : ; xn  ! A is determined by the knowledge of .x1 / D
a1 ; : : : ; .xn / D an , ai 2 A. The choice of such morphism  corresponds to the
choice of an n-uple .a1 ; : : : ; an /, ai 2 A. So, we can identify hAn .A/ with the set
of n-uples .a1 ; : : : ; an / with entries in A. If A D Z or A D Q, this is the notion we
already encounter in classical algebraic geometry.
The functor of points, originally introduced as a tool in algebraic geometry, can
actually be employed in a much wider context.
Definition 2.4.5. Let M D .jM j; OM / be a locally ringed space and let .rspaces/
denote the category of locally ringed spaces. We define the functor of points of the
locally ringed space M as the representable functor:

hM W .rspaces/op ! .sets/; hM .T / D Hom.T; M /:

As before, hM is defined on morphisms as follows: hM ./.g/ D g B .

If the locally ringed space M is a differentiable manifold, we have the following
important characterization of morphisms.
Proposition 2.4.6. Let M and N be differentiable manifolds. Then

Hom.M; N / Š Hom.C 1 .N /; C 1 .M //:

40 2 Sheaves, functors and the geometric point of view

We are going to see a proof of this result in the more general setting of supergeometry
in Chapter 4, Section 4.5.
We are now going to state Yoneda’s lemma, a basic categorical result. As an imme-
diate consequence, we have that the functor of points of a scheme (resp. differentiable
manifold) does determine the scheme (resp. differentiable manifold) itself.

Theorem 2.4.7 (Yoneda’s lemma). Let C be a category and let X , Y be objects

in C and let hX W C op ! .sets/ be the representable functor defined on the objects
as hX .T / D Hom.T; X / and, as usual, on the arrows as hX ./.f / D f  , for
 W T ! S, f 2 Hom.T; X /.
(1) If F W C op ! .sets/, then we have a one-to-one correspondence between the
sets:
fhX ! F g () F .X /:
(2) The functor
h W C ! Fun.C op ; .sets//; X 7! hX ;
is an equivalence of C with a full subcategory of functors. In particular, hX Š hY
if and only if X Š Y and the natural transformations hX ! hY are in one-to-one
correspondence with the morphisms X ! Y .

Proof. We briefly sketch it, leaving the details to the reader. Let ˛ W hX ! F . We
can associate to ˛, ˛X .idX / 2 F .X /. Vice versa, if p 2 F .X /, we associate to p,
˛ W hX ! F such that

˛Y W Hom.Y; X / ! F .Y /; f 7! F .f /p: 

Corollary 2.4.8. Two schemes (resp. manifolds) are isomorphic if and only if their
functors of points are isomorphic.

The advantages of using the functorial language are many. Morphisms of schemes
are just maps between the sets of their A-points, respecting functorial properties. This
often simplifies matters, allowing us to leave the sheaves machinery in the background.
The problem with such an approach, however, is that not all the functors from .schemes/
to .sets/ are the functors of points of a scheme, i.e., they are representable. The next
theorem establishes an important criterion. We shall state the theorem for functors from
.rings/ to .sets/ leaving to the reader, as a simple exercise, to write a similar statement
for functors from .schemes/ to .sets/.

Theorem 2.4.9. A functor F W .rings/ ! .sets/ is of the form F .A/ D Hom.Spec A; X /

for a scheme X if and only if:
(1) F is local or is a sheaf in the Zariski topology. This means that for each ring R
and for every collection ˛i 2 F .Rfi /, with .fi ; i 2 I / D R, so that ˛i and ˛j map to
the same element in F .Rfi fj / for all i and j , there exists a unique element ˛ 2 F .R/
mapping to each ˛i .
 2.4 Functor of points 41

(2) F admits a cover by open affine subfunctors. This means that there exists
a family Ui of subfunctors of F , i.e., Ui .R/  F .R/ for all R 2 .rings/, Ui D
hSpec Ui , with the property that for all natural transformations f W hSpec A ! F , the
functors f 1 .Ui /, defined as f 1 .Ui /.R/ D f 1 .Ui .R//, are all representable,
i.e., f 1 .Ui / D hVi , and the Vi form an open covering of Spec A.

Proof. See [29], p. 259 or [23], Ch. I. 

In Chapter 10 we are going to see a complete proof of this statement in the more
general setting of superschemes.
This theorem states the conditions we expect for F to be the functor of points of a
scheme. Namely, locally, F must look like the functor of points of a scheme (property
(2)), moreover F must be a sheaf, that is F must have a gluing property that allows
us to patch together the open affine cover we are given in the hypothesis (property (1)
and (2) together).
A similar criterion holds for the functor of points in the differential category and
more general for a locally ringed space and we are going to discuss it in Chapter 9.
We conclude this section by examining the two important examples of projective
space and Grassmannian variety using the functorial point of view.

Examples 2.4.10. (1) Projective space. Let us revisit Example 2.3.4 of projective space
as a scheme using the functor of points point of view. Define the functor h W .rings/ !
.sets/, where h.A/ is the set of projective submodules of AnC1 of rank n. Equivalently,
by duality, we have that h.A/ consists of the morphisms ˛ W AnC1 ! L, where L has
rank 1, modulo the equivalence relation ˛ ˛ 0 if and only if ker.˛/ D ker.˛ 0 /.
To complete the definition we need to specify h on morphisms W A ! B.
Given a morphism W A ! B, we can give to B an A-module structure by setting

ab D .a/b; a 2 A; b 2 B:

Also, given an A-module L, we can construct the B-module L ˝A B. So given and

the element of h.A/, f W Am ! L, we have an element of h.B/,

h. /.f / W B m Š Am ˝A B ! L ˝A B:

We want to show that h is the functor of points of the projective space, in other
words h D haP n . We briefly sketch how to check property (2) in Theorem 2.4.9; as for
property (1) and in general for more details, we refer the reader to [23], Ch. 1, §1.
Property (2) says that we need to cover h by open affine subfunctors vi . The functors
vi are defined as follows. For A local, vi .A/ is the set of ˛ such that ˛.0; : : : ; ai ; : : : ; 0/
is invertible. As an exercise one can show that vi corresponds to an open affine subfunc-
tor of h, and it corresponds to the functor of points of an affine space of dimension n. If
A is local we have the following nice characterization of the A-points of the projective
space (see [29], Ch. III, §2).
42 2 Sheaves, functors and the geometric point of view

Proposition 2.4.11. The A-points of P n , for A local, are in one-to-one correspondence

with the set of n C 1-uples Œa0 ; : : : ; an  2 AnC1 such that at least one of the ai is a unit,
modulo the equivalence relation Œa0 ; : : : ; an  Š Œ a0 ; : : : ; an  for any unit in A.
(2) Grassmannian scheme. Classically the Grassmannian variety is a natural gen-
eralization of the projective space: its points are the set of r vector spaces inside an
m-dimensional vector space (r < m). It is a projective variety and now we want to
describe the functor of points of its associated scheme.
Define the functor Gr W .rings/ ! .sets/ as follows. Gr.A/ is the set of projective
quotients of rank r of Am , that is,

Gr.A/ D f˛ W Am ! L; with L a projective A-module of rank rg=Š:

˛ is a surjective morphisms, L0 is the kernel of ˛ and Am D L ˚ L0 , Š means up to

isomorphism of L. Equivalently this is the set

Gr.A/ D fAm =L0 ; with L0 a projective submodule of Am of rank m  rg

D fprojective submodules L of Am of rank rg:

As before we have the following characterization of Gr.A/ in case A is a local ring

(projective A-modules are free). Gr.A/ is the set of all free A-modules of rank r in Am
modulo isomorphism. If we have a ground field k and A D k we can see immediately
that Gr.k/ are the points of the classical Grassmannian variety of r subspaces in k m .
To complete the definition we need to specify Gr on morphisms W A ! B the
same way we did before.
Given a morphism W A ! B, we can give to B the structure of an A-module by
setting
a  b D .a/b; a 2 A; b 2 B:
Also, given an A-module L, we can construct the B-module L ˝A B. So given and
the element of Gr.A/, f W Am ! L, we have an element of Gr.B/,

Gr. /.f / W B m D Am ˝A B ! L ˝A B:

The fact that Gr is the functor of points of a superscheme is again an application

of Theorem 2.4.9. For the property (1), namely the fact that Gr is local, we send the
reader to [23], Ch. I, §1. For the property (2), we have to give explicitly a cover by open
affine subfunctors. Consider the multi-index I D .i1 ; : : : ; ir /, 1  i1 <    < ir  m,
and the map I W Ar ! Am defined by setting I .x1 ; : : : ; xr / D .y1 ; : : : ; ym /, with
yik D xk for k D 1; : : : ; m and yj D 0 otherwise.
Let us define subfunctors vI of Gr for A local as follows:

vI .A/ D f˛ W Am ! L j ˛ B I is invertibleg:

We leave it to the reader, as an exercise, to verify that the vI correspond to open

affine subfunctors of Gr isomorphic to the functor of points of affine spaces. In any
 2.5 Coherent sheaves 43

case, we are going to study in detail this same example in the supergeometric setting in
Chapter 10. For more details on this and the previous example in the ordinary setting,
see [23], Ch. I, §1, and [29], Ch. VI.

2.5 Coherent sheaves

In this section we briefly give the definition and some basic properties of coherent
sheaves. This notion is introduced in ordinary algebraic geometry in order to charac-
terize sheaves which have good properties and in general are well behaved. As we shall
see in the next chapter, it is one of the building blocks for our definition of superscheme.
Let A be a commutative ring and M an A-module. We want to define a sheaf M z on
Spec A, which has an OA -module structure, i.e., for all open sets U in Spec A, we want
Mz .U / to have an OA .U /-module structure compatible with the restriction morphisms.
We are going to define the sheaf on the basic open sets Uf introduced in the previous
section; by Proposition 2.2.11, this will suffice.
Let us consider the assignment

Uf 7! Mf ;

where Mf D M Œf 1  is the Af -module obtained by M by inverting just the ele-

ment f 2 A. This assignment defines a B-sheaf that, by Proposition 2.2.11, extends
z . The next proposition summarizes
uniquely to a sheaf on Spec A that we denote by M
z.
the properties of M
z
Proposition 2.5.1. Let M be a module for a commutative algebra A. The sheaf M
defined above has the following properties:
(1) Mz is an OA -module;
(2) .Mz /p Š Mp for all p 2 Spec A, i.e., the stalk at any prime p of the sheaf Mz
coincides with the localization of M at p;
(3) .Mz /.Spec A/ D M , i.e., the global sections of the sheaf coincide with the A-
module M .

Proof. See Proposition 5.1, Ch. II in [43]. 

Definition 2.5.2. Let X D .jX j; OX / be a scheme, F a sheaf on jX j of OX -modules.

In other words, F .U / is an OX .U /-module for all U open in jX j and the restriction
morphisms behave nicely with respect to the OX -module structure. We say F is
quasi-coherent, if there exists an open affine cover fUi D Spec Ai gi2I of X such that
F jUi Š M z i for a suitable Ai -module Mi . F is coherent if the affine cover can be
chosen so that the Mi ’s are finitely generated Ai -modules.
44 2 Sheaves, functors and the geometric point of view

We are going to give an example of a quasi-coherent sheaf, which is most important

for our algebraic supergeometry applications.
Example 2.5.3. Let R be a commutative super ring. R0 is a commutative ring in the
ordinary sense. Since R1 is an R0 -module, the whole of R is also an R0 -module,
hence we can construct the quasi-coherent sheaf R z on the topological space Spec R0 .
z
One can easily check that .Spec R0 ; R/ is a locally ringed space. We shall see in more
detail in Chapter 10 that this is the local model for our definition of superscheme.
The next proposition establishes an important equivalence of categories.
Proposition 2.5.4. Let A be a commutative ring. The functor M 7! M z gives an
equivalence of categories between the category of A-modules and the category of
quasi-coherent sheaves of OA -modules. The inverse of this functor is the functor F 7!
F .Spec A/. If A is noetherian, the same functor gives an equivalence of categories
between the category of finitely generated A-modules and the coherent sheaves which
are OA -modules.
Proof. See Corollary 5.5, Ch. II in [43]. 

2.6 References
For the theory of sheaves and schemes see [37], [43], [29], [38] and the fundamental
paper by Serre that originated the theory [70]. For a complete treatment of the functor
of points of a scheme see [23]; however, a good summary of most of the properties
needed is found in [29]. As for all the algebraic statements (e.g. the definition of
localization, its properties and so on), we refer to the classical textbooks [51], [2].
3
Supergeometry

In this introductory chapter, we begin our discussion on the foundations of superge-

ometry. We give an overview on some fundamental supergeometric objects, namely
supermanifolds and superschemes, that we shall discuss in detail in the forthcoming
chapters. Here, we also want to stress how the functorial treatment allows the simul-
taneous treatment of both the differential and the algebraic categories at once.
The most basic supergeometric object, the superspace, is first introduced, from
which the concepts of supermanifolds and superschemes are then built. We leave
for later chapters the full development of all of these notions. We also give a brief
description of the functor of points approach in supergeometry, which will be mas-
sively employed, both in the differential and algebraic category, in all of the following
chapters. We illustrate the definitions and the basic concepts with some key exam-
ples, including supermatrices and the general linear supergroup, which turn out to be
fundamental in the later development.

3.1 Superspaces
A unified way to look at the categories of ordinary differentiable manifolds or algebraic
schemes, is to think of an object as a pair, consisting of a topological space together
with a sheaf of functions defined on it. Such a pair is often referred to as a ringed
space. For ordinary manifolds, for example, the sheaf of functions is the sheaf of the
C 1 functions while for ordinary algebraic varieties it is the sheaf of regular functions,
as we have seen in the previous chapter. We are going to generalize this point of
view, discussed in detail for the ordinary setting in the previous chapter, introducing
supermanifolds and superschemes in the framework of ringed spaces.

Definition 3.1.1. As in ordinary algebraic geometry, a super ringed space S is a

topological space jSj endowed with a sheaf of commutative super rings, called the
structure sheaf of S , which we denote by OS . Let S denote the super ringed space
.jSj; OS /.

Notice that S0 ´ .jS j; OS;0 / is an ordinary ringed space as in Definition 2.2.13,

where OS;0 .U / ´ OS .U /0 is a sheaf of ordinary rings on jSj. Notice also that
OS;1 .U / ´ OS .U /1 defines a sheaf of OS;0 -modules on jSj, i.e., for all open sets U
in jS j, we have that OS;1 .U / is an OS;0 .U /-module and this structure is compatible
with the restriction morphisms.
46 3 Supergeometry

Definition 3.1.2. A superspace is a super ringed space S with the property that the
stalk OS;x is a local super ring for all x 2 jS j.

As in the ordinary setting a commutative super ring is local if it has a unique

maximal ideal. Notice that any prime ideal in a commutative super ring must contain
the whole odd part since it contains all nilpotents.
As we have seen in the previous chapter, ordinary differentiable manifolds and
algebraic schemes are examples of superspaces, where we think of their sheaves of
functions as sheaves of commutative super rings with trivial odd part.
Let us now see an example of a superspace with non-trivial odd part.

Example 3.1.3. Let M be a differentiable manifold, jM j its underlying topological

1
space, CM the sheaf of ordinary C 1 functions on M . We define the sheaf of super-
commutative R-algebras as (for V  jM j open)
1
V 7! OM .V / ´ CM .V /Œ 1 ; : : : ;  q ;
1 1
where CM .V /Œ 1 ; : : : ;  q  D CM .V / ˝ ^. 1 ; : : : ;  q / and the  j are odd (anticom-
muting) indeterminates. As one can readily check, .jM j; OM / is a super ringed space;
moreover .jM j; OM / is also a superspace. In fact OM;x is a local ring, with maximal
1
ideal mM;x generated by the maximal ideal of the local ring CM;x and the odd elements
 ; : : : ;  . One can check immediately that all the elements in OM;x n mM;x are
1 q

invertible.
In the special case M D Rp , we define the superspace

Rpjq D .Rp ; CR1p Π1 ; : : : ;  q /:

From now on, with an abuse of notation, Rpjq denotes both the super vector space Rp ˚
Rq and the superspace .Rp ; CR1p Π1 ; : : : ;  q /, the context making clear which one we
mean. Rpjq plays a key role in the definition of supermanifold since it is the local model.
If t 1 ; : : : ; t p are global coordinates for Rp we shall speak of t 1 ; : : : ; t p ;  1 ; : : : ;  q as
a set of global coordinates for the superspace Rpjq .

We are going to study the example of Rpjq in detail in the next chapter.

Definition 3.1.4. Let S D .jSj; OS / be a superspace. Given an open subset jU j  jSj,

the pair .jU j; OS jjU j / is always a superspace, called the open subspace associated to
jU j.

The next example is very important for our applications.

Example 3.1.5 (Supermatrices Mpjq and the general linear supergroup GLpjq ). Let
2 2
Mpjq D Rp Cq j2pq . This is the superspace corresponding to the super vector space
of .pjq  pjq/-matrices, the underlying topological space being Mp  Mq , the direct
 3.1 Superspaces 47

product of p  p and .q  q/-matrices (see also Remark 1.4.1). As a super vector space
we have
² ³ ² ³ ² ³
A B A 0 0 B
Mpjq D ; .Mpjq /0 D ; .Mpjq /1 D ;
C D 0 D C 0

where A, B, C , D are respectively p  p, p  q, q  p, .q  q/-matrices with entries

in R.
Hence as a superspace Mpjq has p 2 C q 2 even global coordinates t ij , 1  i; j  p
or p C 1  i; j  p C q, and 2pq odd ones  kl , 1  k  p, p C 1  l  p C q or
p C 1  k  p C q, 1  l  p. The structure sheaf of Mpjq is the assignment

V 7! OMpjq .V / D CM1p Mq .V /Πkl  for all V open in Mp  Mq :

The superspace Mpjq D .Mp  Mq ; OMpjq / is called the superspace of supermatrices.

2 2
Now let us consider in the topological space Mp  Mq D Rp Cq , the open set U
consisting of the points for which det.tij /1i;j p ¤ 0 and det.tij /pC1i;j pCq ¤
0. We define the superspace GLpjq ´ .U; OMpjq jU /, the open subspace of Mpjq
associated to the open set U . As we shall see, this superspace has a Lie supergroup
structure and it is called the general linear supergroup.
Next we define morphisms of superspaces, so that we can talk about the category
of superspaces.
Definition 3.1.6. Let S and T be superspaces. Then a morphism ' W S ! T is a
continuous map j'j W jSj ! jT j together with a sheaf morphism '  W OT ! ' OS so
that 'x .mT;j'j.x/ /  mS;x , where mS;x is the maximal ideal in OS;x , while mT;j'j.x/
is the maximal ideal in OT;j'j.x/ and 'x is the stalk map.
Remark 3.1.7. Recall from the previous chapter that the sheaf morphism '  W OT !
' OS corresponds to the system of maps 'U W OT .U / ! OS .j'j1 .U // for all open
sets U  jT j, compatible with the restriction morphisms. To ease notation, we also
refer to the maps 'U as '  .
Essentially the condition 'x .mj'j.x/ /  mx means that the sheaf homomorphism
is local. Note also that '  is a morphism of supersheaves, so, as usual, it preserves
the parity. The main point to make here is that when we are giving a morphism of su-
perspaces the sheaf morphism must be specified along with the continuous topological
map, since sections are not necessarily genuine functions on the topological space as
in ordinary differential geometry. An arbitrary section cannot be viewed as a function
because commutative super rings have many nilpotent elements, and nilpotent sections
are identically zero as functions on the underlying topological space. Therefore we
employ the methods of algebraic geometry to study such objects. We will address this
in more detail later. Now we introduce two types of superspaces that we examine in
detail in the forthcoming chapters: supermanifolds and superschemes.
48 3 Supergeometry

3.2 Supermanifolds
A supermanifold is a specific type of superspace, which we describe via a local model.
Because we always keep an eye on the physics literature from which supersymmetry
springs, the supermanifolds of interest to us are the C 1 -supermanifolds over R. Nev-
ertheless, all the definitions we give here hold also in the context of analytic real or
complex supermanifolds, as we shall see more explicitly at the end of the next chapter.
Let CU1 be the sheaf of C 1 -functions on the domain U  Rp . We define the su-
perdomain U pjq to be the superspace .U; CU1 Π1 ; : : : ;  q / where CU1 Π1 ; : : : ;  q  D
V
CR1p jU ˝ . 1 ; : : : ;  q /. Most immediately, the superspaces Rpjq are superdomains
with sheaf CR1p Π1 ; : : : ;  q .
Definition 3.2.1. A supermanifold M D .jM j; OM / of dimension pjq is a superspace
that is locally isomorphic to Rpjq . In other words, given any point x 2 jM j, there
exists a neighborhood V  jM j of x with q odd indeterminates  j so that

V Š V0 open in Rp ; OM jV Š C 1 .t 1 ; : : : ; t p /Œ 1 ; : : : ;  q : (3.1)

„ ƒ‚ …
1j
CR p V0

We call t 1 ; : : : ; t p ;  1 ; : : : ;  q the local coordinates of M in V and pjq the su-

perdimension of the supermanifold M .
Morphisms of supermanifolds are morphisms of the underlying superspaces. For
M; N supermanifolds, a morphism ' W M ! N is a continuous map j'j W jM j !
jN j together with a local morphism of sheaves of superalgebras '  W ON ! ' OM ,
where local, as before (see Chapter 2), means that 'x1 .mM;x / D mN;j'j.x/ , where
'x W ON;j'j.x/ ! OM;x is the stalk morphism, for a point x 2 jM j, and mM;x , mN;j'j.x/
are the maximal ideals in the stalks. Note that in the purely even case of ordinary C 1 -
manifolds, the above notion of a morphism agrees with the ordinary one.
We may therefore talk about the category of supermanifolds. The difficulty in deal-
ing with C 1 -supermanifolds arises when one tries to think of “points” or “functions”
in the traditional sense. The ordinary points only account for the topological space and
the underlying sheaf of ordinary C 1 -functions, and one may truly only talk about the
“value” of a section f 2 OM .U / for U  jM j an open subset; the value of f at x 2 U
is the unique real number c so that f  c is not invertible in any neighborhood of x.
For concreteness, let us consider the example of M D R1j1 , with global coordinates t
and  . Let us take the global section f D t  2 OM .R/. For any non-zero real num-
ber c, we have that t   c is always invertible since t  is nilpotent, its inverse being
c 2 t  c 1 . Hence the value of t  at all points x 2 R D jR1j1 j is zero. What this
says is that we cannot reconstruct a section by knowing only its values at topological
points. Such sections are then not truly functions in the ordinary sense, however, now
that we have clarified this point, we may adhere to the established notation and call
such sections “functions on U ”. We shall return to the delicate notion of value of a
section at a point in Chapter 4, Section 4.1.
 3.2 Supermanifolds 49

Remark 3.2.2. Let M be a supermanifold, U an open subset in jM j, and f a function

on U . If OM .U / D C 1 .t 1 ; : : : ; t p /Π1 ; : : : ;  q  as in (3.1), there exist even functions
fI 2 C 1 .t/ (t D t 1 ; : : : ; t p /) so that

X X X
q
f .t;  / D f0 .t /C fi .t / i C fij .t / i  j C   D f0 .t /C fI .t / I ; (3.2)
i i<j jI jD1

where I D fi1 < i2 <    < ir gqrD1 .

So in some sense, we can expand f .t; / in power series, with respect to the odd
coordinates  j ’s. We are going to discuss this important point further in the next
chapter.

Let us establish the following notation. Let M be a supermanifold, then we write

the nilpotent sections (i.e., the sections in the nilpotent part of OM ) as

JM D OM;1 C OM;1
2
D hOM;1 iOM :

This is an ideal sheaf in OM and defines a natural subspace of M we shall call Mred
or Mz (not to be confused with the notation for coherent sheaves in Proposition 2.5.1),
the reduced manifold associated with M where
z D .jM j; OM =JM /:
M
z is a completely even superspace, and hence lies in the ordinary category
Note that M
1
of ordinary C -manifolds, i.e., it is locally isomorphic to Rp . The quotient map from
OM ! OM =JM defines the inclusion morphism M z ,! M . The subspace M z should
not be confused with the purely even superspace .jM j; OM;0 /, which is not a C 1 -
manifold, since the structure sheaf still contains nilpotents.
We now examine closed submanifolds in the super category.

Definition 3.2.3. Let M D .jM j; OM / be a supermanifold. We say that N D

.jN j; OM =/ is a submanifold of M if:
(1) N is a supermanifold,
(2) jN j is a closed subset of jM j,
(3)  is an ideal sheaf of OM with the following property: for all x 2 jN j, there
exists a neighbourhood Ux of x in jM j and an appropriate choice of coordinates
in M and N such that t 1 ; : : : ; t m ,  1 ; : : : ;  n are coordinates for x in the open
subspace .Ux ; OM jUx / of M while t 1 ; : : : ; t p ,  1 ; : : : ;  q are coordinates for x
in the open subspace .Ux \ jN j; .OM =/Ux \jN j / of N (where p  m, q  n).

One sees immediately that M z D .jM j; OM =JM / is a submanifold of M while

there is no natural way to realize .jM j; OM;0 / as a submanifold of M .
We shall return to the concept of submanifold in Chapter 5, Section 5.3.
50 3 Supergeometry

3.3 Superschemes
A superscheme is an object in the category of superspaces that generalizes the notion of
an ordinary scheme, which we have introduced and discussed in the previous chapter.

Definition 3.3.1. A superspace S D .jSj; OS / is a superscheme if .jS j; OS;0 / is an

ordinary scheme and OS;1 is a quasi-coherent sheaf of OS;0 -modules.

As we shall see in Chapter 10, superschemes are also characterized by local models,
called in this case affine superschemes. They are superspaces locally isomorphic to the
spectrum of superrings, in analogy with the ordinary setting.
Because any non-trivial supercommutative ring has non-zero nilpotents, we need
to redefine what we mean by a reduced superscheme.

Definition 3.3.2. We say that a superscheme S is super reduced if OS =JS is reduced.

In other words, in a super reduced superscheme, we want that the odd sections generate
all the nilpotents.

Example 3.3.3 (The affine superspace). Let Am be the ordinary affine space of di-
mension m over a field k. Am consists of the topological space k m , that is the vector
space k m with the Zariski topology, and the sheaf OAm of regular functions on k m . On
k m we define the sheaf OAmjn of superalgebras in the following way. Given U  k m
open, V
OAmjn .U / D Ok m .U / ˝ . 1 ; : : : ; n /
where 1 ; : : : ; n are odd (anticommuting) indeterminates. One may readily check that
.k m ; OAmjn / is a reduced superscheme, which we here denote by Amjn .

Let us now revisit the Example 3.1.5 taking the algebraic point of view.
alg
Example 3.3.4 (Algebraic supermatrices Mpjq and the algebraic general linear super-
2 2
group GLpjq ). Let Mpjq D Ap Cq j2pq be the affine superspace corresponding to the
alg alg

super vector space of .pjq  pjq/-matrices with entries in the field k. The underly-
alg alg alg alg
ing topological space of Mpjq is the product Mp  Mq , where Mp denotes the set
of .p  p/-matrices with entries in k, with the Zariski topology. The super ring of
alg
global sections of Mpjq is kŒt ij ;  kl , with 1  i; j  p or p C 1  i; j  p C q
and 1  k  p, p C 1  l  p C q or p C 1  k  p C q, 1  l  p. The
conditions det.tij /1i;j p ¤ 0 and det.tij /pC1i;j pCq ¤ 0 define a Zariski open set
alg alg alg
U in Mp  Mq , hence we have a superspace GLpjq D .U; OMpjq jU /. One can check
immediately that this is a superscheme. From now on we shall drop the suffix alg to
improve readability, the context making clear if we are considering the general linear
supergroup GLpjq in the algebraic or in the differential context.

Morphisms of superschemes are just morphisms of superspaces, so we may talk

about the subcategory of superschemes. The category of superschemes is larger than the
 3.4 The functor of points 51

category of schemes; any scheme is a superscheme if we take a trivial odd component

in the structure sheaf. We will complete our exposition of the category of superschemes
in Chapters 10–11.

3.4 The functor of points

The presence of odd coordinates steals some of the geometric intuition away from the
language of supergeometry. For instance, we cannot see an “odd point” – they are
invisible both topologically and as classical functions on the underlying topological
space. We see the odd points only as sections of the structure sheaf. To bring some of
the intuition back, we turn to the functor of points approach from algebraic geometry.

Definition 3.4.1. Let S and T be superspaces. A T -point of S is a morphism T ! S .

We denote the set of all T -points by S.T /. Equivalently,

S.T / D Hom.T; S /:

We define the functor of points of the superspace S to be the functor

S W .sspaces/op ! .sets/; T 7! S.T /; S./.f / D f B ;

where .sspaces/ denotes the category of superspaces and the superscript op as usual
refers to the opposite category (see Appendix B for more details).

By a common abuse of notation the superspace S and the functor of points of S are
denoted by the same letter. Whenever it is necessary to make a distinction, we shall
write hS for the functor of points of the superspace S.
We have defined the functor of points of a superspace; clearly we can also define the
functor of points of a supermanifold or a superscheme, just by changing the category
we start from.

Definition 3.4.2. Let .smflds/ and .sschemes/ denote respectively the categories of
supermanifolds and superschemes introduced above. We define the functor of points
of the supermanifold M to be the functor

M W .smflds/op ! .sets/; T 7! M.T /; M./f D f B :

Similarly we define the functor of points of the superscheme X to be the functor

X W .sschemes/op ! .sets/; T 7! X.T /; X./f D f B :

The importance of the functor of points is a consequence of the following lemma,

which is one of the many versions of Yoneda’s lemma (see also Theorem 2.4.7).
52 3 Supergeometry

Lemma 3.4.3 (Yoneda’s lemma). Let M and N be two superspaces (resp. superman-
ifolds or superschemes). There is a bijection from the set of morphisms ' W M ! N
to the set of maps 'T W M.T / ! N.T /, functorial in T . In particular M and N are
isomorphic if and only if their functors of points are isomorphic.

Proof. Given a map ' W M ! N , for any morphism t W T ! M , ' B t is a morphism

T ! N . Conversely, we attach to the system .'T / the image of the identity map from
'M W M.M / ! N.M /. For more details see Appendix B. 

Yoneda’s lemma allows us to replace a superspace (resp. a supermanifold or a

superscheme) S with its set of T -points, S.T /. We can now think of S as a representable
functor from the category of superspaces (resp. supermanifolds or superschemes) to the
category of sets. In fact, when constructing a superspace, it is often most convenient
to construct first its functor of points and then prove that the functor is representable
in the appropriate category. In fact in Chapter 9 we shall give a criterion for the
representability of functors from the category of supermanifolds to the category of sets
while in Chapter 10 we shall prove the same result for superschemes. As in the ordinary
case (see Theorem 2.4.9), it turns out that representable functors must be local, i.e.,
they should admit a cover by open affine subfunctors, which glue together in some
sense that we shall specify.
The following proposition, that we shall prove in the next chapter, is very useful
when we want to explicitly describe the functor of points of a supermanifold or a
superscheme. Its very formulation shows how the functorial treatment allows us to
deal at once with both the differential and algebraic categories.

Remark 3.4.4. To ease the notation we write OT .T / or simply O.T / for the global
sections of a superspace T .

Proposition 3.4.5. Let M D .jM j; OM / and T D .jT j; OT / be supermanifolds or

affine superschemes. Then

Hom.T; M / D Hom.O.M /; O.T //:

Let us give some examples of T -points.

Examples 3.4.6. (i) Let T be just an ordinary topological point viewed as a super-
manifold, i.e., T D R0j0 D .R0 ; R/. By definition a T -point of a manifold M is a
morphism  W R0j0 ! M .  consists of a continuous map jj W R0 ! jM j, which cor-
responds to the choice of a point x in the topological space jM j, and a sheaf morphism
  W OM !  .R/, which assigns to a section its value in x. Then a T -point of M is
an ordinary topological point of jM j.
(ii) Let M be the supermanifold Rpjq and let T be a supermanifold. By the previous
proposition we have that a T -point of M corresponds to a morphism

O.M / D C 1 .t 1 ; : : : ; t p /Π1 ; : : : ;  q  ! O.T /:

 3.5 References 53

Then, in this case, a T -point of M is a choice of p even and q odd global sections
on T . This is made more clear in Chapter 4 by Theorem 4.1.11. Thus Rpjq .T / D
p q
OT;0 .T / ˚ OT;1 .T /.
(iii) Let X be the superscheme Amjn as in Example 3.3.3 and let T be an affine
superscheme. By definition, a T -point of X is a morphism of schemes  W T !
Amjn , which again, by the previous proposition, corresponds to a super ring mor-
phism W O.Amjn / ! O.T /, that is W kŒx1 ; : : : ; xm ; 1 ; : : : ; n  ! O.T /, where
kŒx1 ; : : : ; xm ; 1 ; : : : ; n  denotes the polynomial algebra in the even (commuting) in-
determinates x1 ; : : : ; xm and in the odd (anticommuting) ones 1 ; : : : ; n . Hence, as
before, amounts to a choice of m even global sections in O.T / and n odd ones:

Amjn .T / D O.T /m 0 ˚ O.T /1

n

D f.a1 ; : : : ; am ; ˛1 ; : : : ; ˛n / j ai 2 O.T /0 ; ˛j 2 O.T /1 g:

We already see the power of T -points in these examples. The first example (T D
R0j0 ) gives us complete topological information while the second (M D Rpjq ) allow us
to talk about coordinates on supermanifolds. The third example shows how remarkably
the differential and the algebraic categories resemble each other under the functorial
treatment.
We plan to fully explore all these topics in the next chapters.

3.5 References
For our brief introduction and overview of supergeometry we send the reader to the
works by Berezin [10], Kostant [49], Manin [56] and the notes of Bernstein [22]. In
such papers, especially at the beginning, such exposition is done with more details. In
particular in [56] there is a discussion of what a superscheme is, and, though implicitly,
the problem of representability makes its appearance. In [49], in 2.15, there is a
discussion of Proposition 3.4.5.
4
Differentiable supermanifolds

In this chapter we come to a more systematic and detailed study of supermanifolds. We

shall be mainly interested in smooth supermanifolds, briefly discussing, only towards
the end of the chapter, some aspects of real and complex analytic supermanifolds.
As we have seen in the previous introductory chapter, a smooth supermanifold is a
superspace locally isomorphic to the superspace Rmjn , which is an example of super-
domain. We start by discussing various results and local properties of superdomains;
the most important is the Chart Theorem, which allows us to characterize morphisms
of superdomains.
We then proceed to a thorough description of the category of supermanifolds and
their morphisms, introducing the concepts of tangent space and differential of a super-
manifold morphism.
In analogy with ordinary differential geometry, we shall see that we can reconstruct
a supermanifold starting from the superalgebra of the global sections of its structure
sheaf. This is a fundamental result, and for this reason we devote a good part of
this chapter to proving it in all details. As a consequence of this result we have a
bijective (contravariant) correspondence between morphisms of supermanifolds and
morphisms of the superalgebras of their global sections. This is the main tool we are
going to employ to discuss supermanifold theory via the functor of points approach.

4.1 Superdomains and their morphisms

In this section we collect a few fundamental results on superdomains that we shall need
in the sequel to develop the theory of smooth supermanifolds. The main result here
is the Chart Theorem, which allows us to identify a morphism between superdomains
of dimension mjn and pjq with p even and q odd functions in m even and n odd
indeterminates. This is a very natural generalization of a similar result for ordinary
differentiable manifolds.
Let us start by recalling a few definitions from Chapter 3.

Definition 4.1.1. We say that S D .jS j; OS / is a super ringed space if jS j is a topo-

logical space and OS is a sheaf of super rings. If the stalk OS;x is a local super ring
for all points x 2 jSj we say that S is a superspace.

Definition 4.1.2. We define the smooth superdomain of dimension .pjq/ to be the super
ringed space U pjq D .U; OU pjq /, where U is an open subset of Rp and OU pjq .V / D
 4.1 Superdomains and their morphisms 55

V V V
C 1 .V / ˝ q for each V open subset of U , with q D . 1 ; : : : ;  q /. pjq is called
the superdimension of the superdomain. We denote by O.U pjq / the global sections of
the sheaf OU pjq .
A morphism between two superdomains V njm and U pjq is a morphism of super
ringed spaces.
Remark 4.1.3. It is a consequence of Lemma 4.1.9 or of a similar ordinary result
(see Example 3.1.3) that each superdomain is actually a superspace. Moreover we
will show that morphisms between smooth superdomains are automatically superspace
morphisms (see Lemma 4.2.3).
In the following we will make frequent use of the multi-index notation. An n-
tuple IPD .i1 ; : : : ; in / of non-negative integers is called a multi-index. We define
jI j ´ jnD1 ij and I Š ´ i1 Ši2 Š : : : in Š. Moreover we will use the shorthand

@jI j @jI j
´ :
@x I @x1i1 : : : @xnin

If f 2 OU pjq .V /, we can speak of the value f .x/ of f at a point x 2 V .

P
Definition 4.1.4. Let f D f0 C jI j1 fI  I be a section in OU pjq .V /, f0 , fI 2
C 1 .V /. If x 2 V , the value f .x/ of f at x is f0 .x/.
Remark 4.1.5. When we speak of the value of a section we mean a real number
f .x/ 2 R. We may however use the same notation to mean just a section in OU pjq .V /,
to stress its dependence on x, which hence in this case is not assumed to be a point,
but just a set of coordinates. This ambiguity is already present in the ordinary setting,
and we shall make an effort to warn the reader whenever confusion may arise.
The next lemma gives an intrinsic characterization of f .x/.
P
Lemma 4.1.6. Let f D f0 C jI j1 fI  I be a section in OU pjq .V /. If x 2 V , then
f .x/ is the unique number such that f  f .x/ is not invertible in any neighborhood
of x contained in V .
Proof. Classically, that is when f D f0 , the result is true, hence f0 .x/ is the unique real
number for which f0  f0 .x/ is not invertible in any neighbourhood of x. We have f 
f0 .x/ D f0  f0 .x/ C n, with n nilpotent. Since invertibility is not changed by adding
a nilpotent element, we have that f  f0 .x/ is not invertible in any neighbourhood of
x. Uniqueness also comes from the classical result. 
If U pjq is an ordinary domain, that is q D 0, the value of a section f , now a true
function on the topological space U , is f .x/, that is, the function f evaluated at the
point x as one usually intends this.
Rpjq denotes the superdomain whose reduced space is Rp and whose sheaf of
sections is given by ORpjq .U / ´ C 1 .U / ˝ ^q for each U open subset of Rp .
56 4 Differentiable supermanifolds

Remark 4.1.7. The reader should notice that here Rpjq denotes a completely differ-
ent object than the super vector space Rpjq introduced in Chapter 1. We will see in
Example 4.6.3, in what sense these two objects can be identified.
If ft i gpiD1 are coordinates in C 1 .U /, and f j gjqD1 is a system of linearly inde-
V
pendent algebraic generators of q , then the set ft i ;  j g is called a system of (super)
coordinates on U pjq . The assignment of a superdomain U pjq together with a system
of super coordinates is called a superchart or chart for short. We notice that on U pjq
there is a canonical chart, consisting of the canonical coordinates inherited from Rpjq .
We now want to discuss morphisms between superdomains in more detail. We start
with an example, that will lead us to the formulation of the Chart Theorem.
Example 4.1.8. Consider the supermanifold R1j2 with a morphism  W R1j2 ! R1j2 .
On R1j2 we have global coordinates t ,  1 ,  2 and so we may express any section
f as in (3.2):

f D f .t;  1 ;  2 / D f0 .t / C f1 .t / 1 C f2 .t / 2 C f12 .t / 1  2 :

Then f0 2 C 1 .R/ sits as a function on the C 1 -manifold R. By definition the

morphism  is described by a continuous map jj and a sheaf morphism   .
Let us first prescribe the images of the global coordinates under   :

t 7! t  ´ t C  1  2 ;
 1 7!  1 ´  1 ; (4.1)
 7! 
2 2
´ :2

We claim that knowing   on only these global coordinates is enough to completely

describe . Indeed, we first see that t 7! t C  1  2 tells us that jj is just the identity
map. Next, let f 2 C 1 .t /Œ 1 ;  2  be as above. Then f 7!   .f / ´ f  can be
written formally as

f  D f .t  ;  1 ;  2 /
D f0 .t C  1  2 / C f1 .t C  1  2 / 1 C f2 .t C  1  2 / 2 C f12 .t C  1  2 / 1  2 :

And so we must only make sense of fI .t C  1  2 /. The key is that we take a formal
Taylor series expansion; the series of course terminates due to the nilpotence of the the
odd coordinates:
fI .t C  1  2 / D fI .t / C  1  2 fI0 .t /:
It is easy to check that this in fact gives a homomorphism of superalgebras. For
g; h 2 C 1 .R/, .gh/ D gh C  1  2 .gh/0 D g  h . Notice moreover that in order to
determine the sheaf morphism, it is enough to specify the images of the global sections
since the full sheaf map is determined by restrictions of the global coordinates, as we
shall see in complete and detailed generality later. So in this example, in fact, the
morphism induced by equations (4.1) is unique via the above construction.
 4.1 Superdomains and their morphisms 57

This fact is indeed true in general. As we shall see, the Chart Theorem states that
a morphism  between superdomains is determined by images of local coordinates
under the sheaf morphism   . Its proof requires a preliminary result on polynomial
approximations of smooth sections, which is of interest by itself.

Lemma 4.1.9 (Hadamard’s lemma). Let U pjq be a superdomain, f 2 O.U pjq / and
x 2 jU pjq j D U . Then for each k 2 N there exists a polynomial Pk;x of degree k in
t 1  t 1 .x/; : : : ; t p  t p .x/,  1 ; : : : ;  q , such that

f  Pk;x 2 IxkC1

where Ix D ff 2 O.U pjq / j f .x/ D 0g.

Moreover if f and g are such that for a given k  q, f  g 2 IxkC1 for each
x 2 jU pjq j, then f D g.

Notice that Ix is the ideal generated by  1 ; : : : ;  q and by the maximal ideal of

sections in C 1 .U / vanishing at x.
P
Proof.
Vq Consider x 2 U with coordinates t01 ; : : : ; t0p . Let f D I fI  I 2 C 1 .U / ˝
. For each given integer r, fI 2 C 1 .U / can be expanded using Taylor series:

X 1  @j j fI 
fI .t/ D fI .t0 / C .t  t0 / C hI;rC1 .t /.t  t0 /rC1 :
 Š @t t0
j jr

Here hI;r 2 C 1 .U /, t D .t 1 ; : : : ; t p /, and we are using a multi-index notation.

Notice that in this expression fI .t0 / is the value of the C 1 function fI in t0 in the
ordinary sense.
Define
X 1  @j j fI 
PI;r .t / D fI .t0 / C .t  t0 / ;
Š @t t0
j jr

RI;rC1 .t / D hI;rC1 .t /.t  t0 /rC1 :

Clearly PI;r is a polynomial in t 1  t 1 .x/; : : : ; t p  t p .x/ of degree r, while RI;rC1 2

IxrC1 . Now define
X
Pk;x .t;  / D PI;kjI j .t / I :
I

One can readily check that this satisfies the requirements of the statement.
Now we turn to the second part of the statement. Let h D f  g 2 IxkC1 for all
x 2 jU pjq j with k  q.PWe want to show that h D 0. Since  I D 0 for jI j > q for all
x, it follows that h D I;jI jq hI  I , where hI 2 Ixr for some r > 0. By definition,
hI .x/ D 0 for all x 2 U , hence hI is identically zero and we have h D 0. 
58 4 Differentiable supermanifolds

Remark 4.1.10. Before turning to the proof of the Chart Theorem we stress that the
above lemma has a number of consequences that deserve attention, and that will be
discussed more deeply in Section 4.3.
Theorem 4.1.11 (Chart Theorem). Let U  Rpjq and V  Rmjn be open superdomains
(with canonical charts). There is a bijection between
(i) the set of morphisms  W V ! U and
(ii) the set of systems of p even functions t i and q odd functions  j  in O.V / such
that .t 1 .m/; : : : ; t p .m// 2 jU j for all m 2 jV j.
Proof. It is clear that if we have a morphism  we can uniquely associate to it a set of
p even and q odd functions in O.V /. In fact, we take t i ´   .t i /,  j  ´   . j /,
where ft i ;  j g is the canonical chart in O.U /.
Assume now that (ii) holds. We denote by ft i ;  j g and fx r ; s g coordinates on U
and V , respectively. It is clear that we have a continuous map jj W jV j ! jU j. We V now
need to show V that there exists a unique superalgebra morphism   W C 1 .jU j/˝ p !
C 1 .jV j/ ˝ n such that   .t i / D t i and   . j / D  j  . Let us start with the
existence of such   . It is enough V to show that there exists a superalgebra morphism
  W C 1 .jU j/ ! C 1 .jV j/ ˝ n , since we know the image of the polynomial odd
P
generators  j . Let t i D t i Cni with t i 2 C 1 .jV j/, ni ´ i I
jI j>1 tI and define
the pullback, through a formal Taylor expansion, by
X 1 @ f ˇ
ˇ

 .f / ´ ˇ n ; (4.2)
 Š @t ˇ z
t

where we are using a multi-index notation as in Lemma 4.1.9 and by jtz we mean to
substitute any instance of ti with tzi .
We show that it is a superalgebra morphism. Indeed,
X 1 @ .f  g/ ˇˇ
  .f  g/ D ˇ n
 Š @t ˇ z
t
X 1    @˛ f ˇ ˇ
ˇ @ ˛ g ˇ
D ˇ ˇ
 Š ˛ @t ˛ ˇ z @t ˛ ˇ z n
˛< t t

X 1 @ f ˇˇ ˛ ˇ
1 @ˇ g ˇˇ ˇ
D ˇ n˛ n :
˛Š @t ˛ ˇtz ˇŠ @t ˇ ˇtz
˛;ˇ

We
Vq now come to uniqueness.
Vn Suppose that 1 and 2 are two morphisms C 1 .jU j/ ˝
1
! C .jV j/ ˝ such that 1 .t i / D 2 .t i / D t i , 1 . j / D 2 . j /VD  j  .
They clearly coincide on polynomial sections. If x 2 jU j, f 2 C 1 .jU j/ ˝ q , due
to Lemma 4.1.9, we can write, for k > q, f D Pk;x C h with h 2 IxkC1 . Hence
.1  2 /.Pk;x C h/ D .1  2 /.h/. It is easily checked that i .Ixk / Iyk with y
such that jj.y/ D x, so that the result follows in view of Lemma 4.1.9.
 4.2 The category of supermanifolds 59

V V
Given the superalgebra morphism   W C 1 .jU j/ ˝ q ! C 1 .jV j/ ˝ n and an
 1
open
Vq subset jW j jU j, reasoning
Vn as before we can define a morphism W W C .jW j/˝
1 1
! C .jj .jW j// ˝ by
t i jW ! t i jj j1 .jW j/ ;  j jW !  j  jj j1 .jW j/ :

In this way it is easy to check that we have a sheaf morphism. We leave to the reader
the easy check of all details. 
Remark 4.1.12. Equation (4.2) in the proof of the previous proposition gives the
recipe for calculating the pullback of a generic section from the pullback of the super
coordinates.
Notice also that because the expansion (4.2) involves an arbitrary number of deriva-
tives, there is no straightforward way to make sense of C k -supermanifolds.

4.2 The category of supermanifolds

In this section we introduce the category of supermanifolds. A supermanifold is es-
sentially a super ringed space locally modelled by superdomains. Since, as we have
seen in the previous section, smooth superdomains are superspaces, we have that also
supermanifolds are superspaces.
We also define the important technical notion of partition of unity in the super
context, which we shall use in many proofs, and we discuss the reduced manifold
associated with a supermanifold.
Definition 4.2.1. A superspace M D .jM j; OM / is called a supermanifold if
(i) jM j is a (locally compact) second countable Hausdorff topological space
(ii) for each x 2 jM j there exists an open neighborhood U 3 x such that there exists
an isomorphism
'U
.U; OM jU / ! U mjn  Rmjn
for fixed (i.e., independent of x) mjn, where U mjn is a superdomain in Rmjn .
A morphism between supermanifolds is a morphism between the corresponding
superspaces. The pair mjn is called the superdimension (or dimension for short) of M .
We denote by .smflds/ the category of supermanifolds. Clearly each superdomain
is a supermanifold. The pair .U; 'U / is called a superchart (or chart) around x 2 U .
In analogy with Lemma 4.1.6, we can prove the following lemma that enables us
to give precise meaning to the evaluation of a section in the structure sheaf.
Lemma 4.2.2. Let M be a smooth supermanifold, U be an open subset of jM j, and
f 2 OM .U /. There exists a unique real number f .x/ such that f  f .x/ is not
invertible in any neighborhood U V containing x.PMoreover if .V 0 ; t i ;  j / is a chart
containing x and contained in V , and f jV 0 D f0 C jI j1 fI  I , then f .x/ D f0 .x/.
60 4 Differentiable supermanifolds

The next lemma shows that a morphism of supermanifolds as super ringed spaces
is automatically a morphism of superspaces, hence of supermanifolds. This result is
especially important for applications.
Lemma 4.2.3. Let M and N be supermanifolds, and let  W M ! N ,  ´ .jj;   /
be a super ringed spaces morphism. Then
(i)   .f /.x/ D f .jj.x// for all f 2 ON .U /,
(ii)  defines a supermanifold morphism.
Proof. Let us start with the proof of (i). Suppose that   .f /.x/ ¤ f .jj.x//. If
g D f  f .jj.x// 2 ON .U /, this means that   .g/.x/ ¤ 0. g is not invertible
in any neighbourhood of jj.x/, however, since   is a superalgebra morphism and
  .g/.x/ ¤ 0, we have a contradiction.
In order to prove (ii), we only need to prove that   is a local morphism in the sense
that if Jj j.x/ denotes the maximal ideal of the stalk OM;j j.x/ then

  .Jj j.x/ / Jx :

Since Jx identifies with the germs of sections that evaluated at x are zero, (ii) follows
directly from (i). 
Remark 4.2.4. We can state the following more general version of the Chart Theorem
that is proven in the same way as its local version 4.1.11.
Theorem 4.2.5 (Global Chart Theorem). Let U  Rpjq be a superdomain and M a
supermanifold. There is a one-to-one correspondence between the morphisms M ! U
and the set of p C q-uples of p even functions t i and q odd functions  j  on M such
that .t 1 .x/; : : : ; t p .x// 2 jU j for all x 2 jM j.
Proof. One way is clear, namely the fact that for each morphism W M ! U we have
the family of sections detailed above. Suppose now we are given a family of sections
t i ,  j  on M such that .t 1 .x/; : : : ; t p .x// 2 jU j for all x 2 jM j. We want to define
a morphism W M ! U . For each point x 2 jM j, if .Vx ; OM jVx / is a chart around
x, we have a morphism x W Vx ! U which is uniquely determined by the Chart
Theorem for superdomains. More precisely x is the morphism corresponding to the
assignment  .x i / D t i jVx ,  . j / D  j  jVx , where x i , j are local coordinates
for the chart Vx . Now, if f is a section in O.U /, we can consider the pullbacks x .f /
for each x 2 jM j. Clearly x .f /jVx \Vy D y .f /jVx \Vy for each x and y in jM j.
Due to the sheaf property there exists a unique section  .f /, hence we have defined
uniquely our morphism  . 
Remark 4.2.6. Any section f 2 OM .U / can now be interpreted as a morphism
f W U ! R1j1 , thus recovering the intuitive meaning of sections as functions on open
subsets of the supermanifold. In fact as one can readily see, by the Chart Theorem an
even function f W U ! R1j0 corresponds to the choice of an even section in OM .U /,
 4.2 The category of supermanifolds 61

while an odd function f W U ! R0j1 corresponds to the choice of an odd section in

OM .U /. For this reason we shall write jf j, for a section f 2 OM .U /, meaning the
topological function jf j W jU j ! R. Clearly jf j.x/ D f .x/ as in Definition 4.1.1.
We shall also use fQ to denote jf j whenever there is ambiguity with the absolute value.

If jU j is an open subset of jM j, the superspace .jU j; OM jU / is itself a superman-

ifold and it is called the open submanifold determined by jU j. We shall use the term
“submanifold” instead of the more cumbersome, though more precise, “subsuperman-
ifold”. In particular if M and N are supermanifolds, U is an open submanifold of
N , and jj W jM j ! jN j is a continuous map, then we denote by jj1 .U / the open
submanifold of M determined by jj1 .jU j/.
In the next sections we shall investigate the structure of a supermanifold at different
levels: infinitesimal, local and global. In order to do this, we need some preliminary
results. We start with a definition of partition of unity.
As in ordinary differential geometry, partitions of unity are very useful in passing
from global to local problems and vice versa.
Let M D .jM j; OM / be a supermanifold and let f be a section over U open in
jM j. Consider the set of points x 2 U , for which there exists an open neighborhood
V U such that f jV D 0. This is an open set, whose complement is called the
support of f , supp.f /.
We also recall that if jM j is a topological space and fUi gi2I is a cover of M ,
we say that fVj gj 2J is a refinement of fUi gi2I if for each Vj there exists Ui such that
Vj Ui .The refinement is said to be locally finite if each x 2 jM j has a neighbourhood
intersecting only a finite number of Vj .

Proposition 4.2.7. Let fUi gi2I be an open covering of jM j then there exists a family
fgj g of sections such that
(1) gj 2 O.M /0 ,
(2) fsupp gj gj 2J is a locally finite refinement of fUi gi2I and supp gj is compact for
each j ,
P
(3) gj D 1 and gzj  01 for each j .
There exists also a family fhi gi2I such that
(i) hi 2 O.M /0 ,
(ii) supp hi  Ui for all i 2 I of fUi gi2I ,
P
(iii) hi D 1 and hzi  0 for each i .

Proof. If fUi gi2I is an open cover of jM j, then there exists a locally finite refinement
fVj gj 2J where each Vj is an open subset with compact closure (see [78]). Without loss
of generality assume that both Ui and Vj are supercharts coverings. Hence for each Vj
there exists a superchart Ui.j / containing it such that VSj  Ui.j / . For each j define
1
See Remark 4.2.6 for the notation.
62 4 Differentiable supermanifolds

an even section fj 2 OM .Ui.j / /0 such that supp fj D supp fzj VSj and fzj > 0 for
each x 2 Vj . Each fj can be identified with a global section fP j 2 O.M / by defining
c
it to be zero in Ui.j /
. 2
By the local finiteness condition f D fj is a well-defined
section in O.M /0 such that fQ > 0. Hence f is invertible and gj ´ fj is the required
f

partition of unity.
The second part of the proposition P is proved as follows. For each i 2 I let Ji ´
fj 2 J j Vj Ui g and define gi0 D k2Ji gk . Due to local finiteness the sum is well
g0
defined and supp gi0 Ui . The sections hi ´ Pi 0
gi
do the job. 

Definition 4.2.8. Let M be a supermanifold. A family of global sections satisfying

the properties (i)–(iii) in Proposition 4.2.7 is called a partition of unity. It is customary
to use the same terminology also for a family satisfying (1)–(3) in Proposition 4.2.7,
but we shall make a comment whenever we need this stricter requirement.
Remark 4.2.9. It is worth noticing that in the previous definition, in points (1)–(3), we
require the sections forming a partition of unity to have compact support. Hence we
are forced to take for them a different index set from the one used for the open cover.
On the other hand, in points (i)–(iii) we can take the two index sets to be the same, but
then we have to abandon the compact support requirement.
Corollary 4.2.10. If U and V are open subsets of a supermanifold M such that
U Ux V , then there exists an even section f 2 O.M /0 such that 0  fQ.x/  1
for all x 2 jM j and
f jU D 1; f jVx c D 0:
Proof. The proof is immediate in view of the previous proposition considering the open
cover of M given by V , Ux c . 
z , whose un-
To each supermanifold M , we can associate an ordinary manifold M
z
derlying topological space is jM j, and a morphism j W M ! M , which (as we shall
see later) is a closed embedding.
Define the super ideal sheaf

JM .U / ´ JOM .U / ´ ff 2 OM .U / j f is nilpotentg:

A section f 2 OM .U / belongs to JM .U / if and only if f .x/ D 0 for all x 2 U . This

can be proven easily and we leave it to the reader as an exercise.
Definition 4.2.11. Let M D .jM j; OM / be a supermanifold. We define M z D
.jM j; OM =JM / to be the reduced ordinary manifold associated to M (for the defi-
nition of quotient of sheaves, we refer the reader to Chapter 2, Section 2.2).
z is an ordinary manifold.
As we can readily, see M
2
If A is a subset in a topological space X, then Ac denotes its complement in X , AN its closure in X .
 4.2 The category of supermanifolds 63

Remark 4.2.12. Even if the structure sheaf of a smooth supermanifold is isomorphic

to the sheaf of sections of an exterior bundle (see [7]), we cannot think of a super-
manifold simply as an exterior bundle over an ordinary manifold. Morphisms between
1
supermanifolds mix both even and odd coordinates and thus CM cannot be realized
as a subsheaf of OM for an open neighborhood U of a supermanifold M ; it follows
that there is no natural morphism M ! M z . The symmetries of interest in these ex-
tensions of classical manifolds are those which place even and odd on the same level.
Such symmetries are called supersymmetries and are at the foundation of the physical
supersymmetry theory which aims to treat bosons and fermions on the same footing.

Next we want to show that the quotient sheaf OM =JM possesses the property
.OM =JM /.U / D OM .U /=JM .U /; in other words, we do not need to take any sheafi-
fication to obtain the quotient sheaf. (See Definition2.2.7 for the notion of sheafifica-
tion.)

Proposition 4.2.13. Let M be a supermanifold of dimension pjq and let U jM j

be an open subset. The assignment U ! OM .U /=JM .U / is a sheaf on jM j locally
isomorphic to C 1 .Rp /.

Proof. The fact that U ! OM .U /=JM .U / is a presheaf on jM j locally isomorphic to

C 1 .Rp / is immediate. Hence we only have to show the sheaf property.
We need to prove that if we have a family f˛ S 2 OM .U˛ /=J S M .U˛ / such that
f˛ jUˇ D fˇ jU˛ , then there exists a unique f 2 OM . U˛ /=JM . U˛ / with f jU˛ D
f˛ . Let fg˛ g be a partition of unity in OM .U /, subordinated to the open cover fU˛ g
of U . Notice that the product g˛ f˛ is a well-defined element in OM .U /=JM .U /,
and define fO˛ ´ g˛ f˛ . Clearly each fO˛ is a smooth section over U whose support
P O
is contained in U˛ . If we now define f ´ ˛ f˛ , it is very easy to check that
f jU˛ D f˛ . 

We can now immediately define a natural morphism j W M z ! M as follows.

jj j W jM j ! jM j is the identity. The pullback jU W OM .U / ! C 1

z
.U / is sim-
M
ply the projection map OM .U / ! OM .U /=JM .U /. Notice that j  f is the (ordi-
nary) differentiable function on jM j whose value at each x 2 jM j is f .x/ 2 R, i.e.,
.j  f /.x/ D f .x/.
The correspondence between supermanifolds and their associated reduced mani-
folds can be extended to the morphisms in the following way. Let  ´ .jj;   / be
a morphism between the supermanifolds M and N . In order to define Q W M z ! Nz ,
 
notice that  .JN / JM . Hence  descends to a well-defined map between the
quotient sheaves and we have the following proposition that we leave to the reader as
an exercise.
z and  ! Q define a functor
Proposition 4.2.14. The assignments M ! M

.smflds/ ! .mflds/:
64 4 Differentiable supermanifolds

The next question we want to tackle is how to build supermanifolds by gluing

supermanifold structures on an open covering of a topological
S space.
Let .Ui ; Oi / be supermanifolds such that jM j D i Ui is a locally compact,
Hausdorff, second countable topological space, in which every Ui is an open subset.
We define Uij D Ui \ Uj and Uij k D Ui \ Uj \ Uk .
Suppose that we have a family of morphisms

fij W .Uj i ; Oj jUij / ! .Uij ; Oi jUij /

such that
(1) fij is an isomorphism of ringed superspaces,
(2) jfij j D idUij .
Definition 4.2.15. Let the notation be as above. We say that a supermanifold .jM j; OM /
is obtained by glueing the supermanifolds .Ui ; Oi /, if for each i there exists a sheaf
isomorphism
i W .Ui ; OM jUi / ! .Ui ; Oi /
with ji j D idUi and such that (with some abuse of notation)

fij D i j1

on Ui \ Uj . The fij ’s are called transition functions.

We have the following proposition.
Proposition 4.2.16. Let .Ui ; Oi / be as above. The fij ’s satisfy the cocycle conditions
(1) fi i D id on .Ui ; Oi /,
(2) fij fj i D id on .Uij ; Oi jUij /,
(3) fij fj k fki D id on .Uij k ; Oi jUij k /,
if and only if there exists a unique glueing of the .Ui ; Oi /.
Proof. See Proposition 2.2.12. 
Notice that if M is a supermanifold and fU˛ ; '˛ g is a superatlas, then M is iso-
morphic to the supermanifold obtained by glueing the U˛ with transition functions
'ˇ1 '˛ .
Definition 4.2.17. Let M and N be supermanifolds of dimensions pjq and rjs, respec-
tively. We define the product of the supermanifolds M and N to be the super ringed
space
M  N D .jM j  jN j; OM N /;
where the sheaf OM N is defined by OM N .U  V / D C 1 .x; t /Œ;  for coordinate
neighborhoods U.D .x;  //  jM j, V .D .t; //  jN j. One must show that gluing
 4.3 Local and infinitesimal theory of supermanifolds 65

conditions are satisfied, but this calculation mimics that in the ordinary category and
is left to the reader. So M  N is a .p C r/j.q C s/-dimensional supermanifold with
B
M N D M z  Nz . As in the ordinary category, OM N ¤ OM ˝ ON ; instead we
must take the completion of the tensor product to get an equality.
We shall return to the important concept of product of supermanifolds in Section 4.5.

4.3 Local and infinitesimal theory of supermanifolds

In Chapter 2, we have introduced the notion of stalk OM;x associated to a sheaf OM at
a point x 2 jM j. The elements in OM;x are called germs of sections at x. Formally

OM;x D lim OM .U /
!
where the direct limit is taken over all open sets U containing x. As for the ordinary
case, also for supermanifolds the stalk can be characterized as follows:

OM;x Š O.M /= :

Here
f g if there exists an open set U M such that f jU D gjU :
as one can readily check from the definitions (see [29], p. 14 for more details). This
characterization explains why the concept of a stalk at x is well suited for the study of
those properties of supermanifolds depending on the behavior of sections on arbitrarily
small neighborhoods of the point x.
If Œf  is a germ at x, it makes sense to evaluate Œf  at x in the same way as we
evaluate the sections. We need also to define the natural morphism

x W O.M / ! OM;x ; f 7! Œf x :

The next lemma provides another useful characterization of the stalk at a point of
a supermanifold.
Lemma 4.3.1. If mx D ff 2 O.M / j there exists an open neighbourhood U of x
such that fU D 0g, then
OM;x ' O.M /=mx :
Proof. Consider the morphism x W O.M / ! OM;x defined above. x is surjective.
In fact, if f is a representative of the germ Œf x , we can think of it as a section in
OM .U /, where U is an appropriate neighborhood of x. If V is an open subset such
that V Vx U and  2 O.M / is a section such that supp  U and jV D 1, then
  f is a global representative of Œf x . It is clear that the kernel of x consists of the
elements of O.M / that are zero in a neighborhood of x, i.e., of mx . 
66 4 Differentiable supermanifolds

In Section 4.1 we have discussed Hadamard’s Lemma 4.1.9, which holds also in
the context of germs of functions. We simply restate it, leaving the proof to the reader
as a simple exercise.
Lemma 4.3.2 (Local Hadamard lemma). Let M D .jM j; OM / be a supermanifold,
dim.M / D .mjn/, x a point in jM j, t i ,  j a local coordinate system around x. Let f
and g be sections defined in a neighbourhood U of x.
(1) If Œf  2 OM;x , then for each k 2 N there exists a polynomial Pk;x of degree k
in Œti  t i .x/, Œ j such that
Œf   Pk;x 2 JxkC1
where Jx D fŒf  2 OM;x j Œf .x/ D 0g.
(2) Moreover, if Œf  and Œg are germs at x such that, for a given k  n, Œf   Œg 2
JykC1 for each y in a neighbourhood of x, then Œf  D Œg.
OM;x is a local ring, that is, OM;x contains a unique homogeneous maximal ideal
Jx . Now we want to clarify the structure of Jx .
Proposition 4.3.3. Let Œf  2 OM;x and dim.M / D .mjn/.
(1) If Œf .x/ D 0, then Œf  2 .Œt i  t i .x/; Œ j /  OM;x . In particular we have
Jx D .Œt i  t i .x/; Œ j /  OM;x ; OM;x D R ˚ Jx :
Hence OM;x is a local superalgebra with maximal ideal mx D Jx .
(2) If f 2 O.M /, k > n and Œf x 2 Jxk for all x in an open set U , then f jU D 0.
Proof. This follows immediately from Lemma 4.3.2. 
At this point one may be tempted to expand in formal power series the germs at a
point x. This is fact makes sense, provided we exert some care (as one should also do
for the ordinary case). We are going to come back to this point in Remark 4.3.12.
Remark 4.3.4. This gives another, and independent, proof of the fact that each su-
perdomain is a superspace. Indeed the lemma essentially establishes that each stalk
OU pjq ;x has a unique maximal ideal given by the sections whose value at x is zero.
More precisely we have the decomposition
OU pjq ;x D R ˚ Jx

with Jx D .Œt i  t i .x/; Œ j /  OU pjq ;x , where Jx is the ideal of germs that are zero
at x.
This allows us to characterize the value of a section f 2 OU pjq .V / at a point x 2 V
as the image of f under the natural morphism
OU pjq .V / ! OU pjq ;x ! OU pjq ;x =Jx Š R;
where Jx denotes the maximal ideal in the stalk OU pjq ;x .
 4.3 Local and infinitesimal theory of supermanifolds 67

The next results characterize the ideals of finite codimensions in the stalks. We
shall use them in Section 4.7.

Proposition 4.3.5. Each ideal I of finite codimension of a stalk OM;x contains an

ideal of the form Jxk for some k 2 N.

Proof. First let us notice that, due to Proposition 4.3.3, Jx is a finitely generated OM;x -
module. Consider then the chain of ideals

OM;x Jx C I Jx2 C I  :

Since I has finite codimension and Jxk © JxkC1 , we have that there exists k such that

Jxk C I D JxkC1 C I:

Hence
Jxk JxkC1 C I:
So we have
Jxk D .JxkC1 C I / \ Jxk D JxkC1 C I \ Jxk :
Due to Nakayama’s lemma (see Appendix B) we have I \ Jxk D Jxk . 

We are ready to define tangent spaces. Let M D .jM j; OM / be a supermanifold.

Definition 4.3.6. We define a tangent vector at x 2 jM j to be a (super) derivation of

the stalk OM;x , i.e., a linear map

v W OM;x ! R

such that
v.f  g/ D v.f /g.x/ C .1/jvjjf j f .x/v.g/:

Remark 4.3.7. Notice that the sign .1/jvjjf j f is immaterial. In fact the sign appears
only when f and v are odd, but in this case f .x/ D 0.

Definition 4.3.8. The super vector space of all tangent vectors at a point x 2 jM j is
called the super tangent space at x and is denoted by Tx .M /.
Any supermanifold morphism  W M ! N induces a stalk morphism

x W ON;j j.x/ ! OM;x ;

which in turn defines a linear morphism of the tangent spaces

.d/x W Tx M ! T .x/ N; v 7! v B x :

The linear map .d/x is called the differential of  at x.

68 4 Differentiable supermanifolds

The differential .d/x is an even linear map of super vector spaces and for this
reason, once we fix homogeneous bases for such vector spaces, it corresponds to a
diagonal block matrix; in other words, it will not contain much information about
the behaviour of the odd variables. For example, if  is the morphism discussed in
Example 4.1.8, one readily checks that .d/x is the identity for all x. In order to better
study infinitesimally the odd directions, we need the concept of Jacobian that we shall
introduce in the next section and fully discuss in Chapter 5.
In the next two propositions we provide very useful characterizations of the tangent
space to a supermanifold.

Proposition 4.3.9. Let x 2 jM j. Suppose that v 2 Tx .M / and t i ,  j are a super

coordinate system around x. Then:
(1) v is completely determined by v.Œt i /; v.Œ j /.
˚ ˇ m ˚ @ ˇ n
(2) Let @i ˇ
@t
,
x iD1 j
ˇ
@
be the derivations
x j D1

@ ˇ
ˇ @ ˇ
ˇ ˇ k ˇ l
i .Œt k / D ıik ; .Œ j / D 0; @ ˇ Œt  D 0; @ ˇ Œ  D ıj l :
@t x @t i x @ j x @ j x

Then they form a basis of Tx M , hence dim M D dim Tx M .

Proof. Let g be a germ in OM;x . Notice that if g is a constant germ or it is in Jx2 then
v.g/ D 0. Due to Proposition 4.3.3, we can write
X X 
gD gI .x/ C Œt i  t i .x/gi;I C hI .t i /Œt i  t i .x/2 Œ I 
I i

with gi;I 2 RP
and hI .t i /Œt i  t i .x/2 2 Jx2 . Hence, with a mild abuse of notation, we
have v.g/ D gi;0 v.Œt i / C gi .x/v.Œ i /. Hence if we consider the difference
X ˇ ˇ
@ ˇ @ ˇ
V ´v v.Œt / i ˇˇ C v.Œ j / j ˇˇ ;
i
@t x @ x

it is immediate that V is a derivation that is zero on all the polynomials in Œt i  and

Œ j . Since v.J ˇ due to Proposition 4.3.3 we can conclude that v D
k
ˇ x/ Jxk1 and
P @ ˇ @ ˇ
v.Œt / @t i x C v.Œ / @ j x .
i j


Let x 2 jM j. Define the super vector space

Der x .O.M /; R/ D fv W O.M / ! R j v.fg/ D v.f /g.x/ C .1/jvjjf j f .x/v.g/g:

Proposition 4.3.10. The linear morphism

˛ W Tx M ! Der x .O.M /; R/; v ! v B x ;

is an isomorphism, where x W O.M / ! OM;x is the natural map.

 4.3 Local and infinitesimal theory of supermanifolds 69

Proof. Since x is surjective, the map ˛ is injective. To show that it is surjective, let
us take w 2 Der x .O.M /; R/. We want to show that w.g/ D 0 for all g 2 mx , so
that w induces a derivation v W O.M /=mx Š OM;x ! R, with ˛.v/ D w. So let us
take g 2 mx . By definition of mx , there exists an open subset U jM j such that
gjU D 0. We can find a section f 2 O.M / such that supp f U and f .x/ D 1, by
Corollary 4.2.10. So we have

0 D w.f  g/ D w.f /g.x/ C f .x/w.g/ D w.g/:

This shows that w descends to a derivation v of the stalk OM;x and that ˛.v/ D w.


Observation 4.3.11. From the Propositions 4.3.9, 4.3.10 we immediately have the
following facts:
(1) For any tangent vector v 2 Tx M and any neighbourhood U of x, there exists a
unique derivation that we still denote by v W OM .U / ! R.
(2) If .t i ;  j / are local coordinates in U , any derivation v W OM .U / ! R is deter-
mined once we know v.t i / and v. j /.

Remark 4.3.12. Consider the filtration

OM;x Jx Jx2  :

Elements of Jxr will be denoted by O.r/. Due to the Proposition 4.3.3, Jxr is generated
as an ideal by the products

Œ.t 1  t 1 .x//i1 : : : Œ.t m  t m .x//im Œ j1  : : : Œ js 

where i1 C    C im C s D r. Proceeding inductively, every Œf  2 OM;x can be written

as

f D f .x/ C O.1/
X
D f .x/ C ai1 :::im ;j1 ;:::;js Œ.t 1  t 1 .x//i1 : : :
1<i1 CCim Csr

: : : Œ.t m  t m .x//im Œ j1  : : : Œ js  C O.r C 1/

@s @i1 :::im f
with ai1 :::im ;j1 ;:::;js D 1
i1 Š:::im Š @ j1 :::@ js @.t 1 /i1 :::@.t m /im
.
T
One should be aware that there are germs in the intersection k Jxk . These are
the germs of sections whose germ is in Fx ˝ ƒ where Fx is the ideal of germs of flat
functions at x, i.e., those functions with zero derivatives at all orders. Of course we
cannot obtain any information about flat functions with power series expansion.
70 4 Differentiable supermanifolds

4.4 Vector fields and differential operators

Many concepts and results from ordinary differential geometry extend naturally to the
category of supermanifolds. If we keep the categorical language we have developed,
there is hardly any difference in fundamental differential geometry between the ordinary
and the super categories. For example, we can define the super tangent bundle on M ,
where we find super extensions of the constant rank mapping theorem and the local
and global Frobenius theorem, which we will prove in the following chapters.
Definition 4.4.1. A vector field V on a supermanifold M is an R-linear derivation of
OM , i.e., it is a family of super derivations VU W OM .U / ! OM .U / that is compatible
with restrictions.
In a similar way, we can define vector fields over open submanifolds U of M .
Observation 4.4.2. When the supermanifold M is a superdomain, one can easily
check that the vector fields on M are in one-to-one correspondence with the derivations
O.M / ! O.M /. Moreover, one could also check, using partitions of unity, that this
is true for generic smooth supermanifolds.
Definition 4.4.3. We define the tangent bundle VecM of the supermanifold M to be
the sheaf consisting of all vector fields on M . This sheaf associates to each open set
U in jM j the vector superspace consisting of all derivations of OM jU .
The sheaf VecM is actually locally free as a sheaf over OM , which we establish
with the following lemma. In other words we have that there exists an open cover of
jM j such that VecM jU Š OM jpjq
U for each U in the cover for suitable p and q. The
lemma also helps us to understand the local structure of a vector field.
Lemma 4.4.4. Let .t i ;  j / be coordinates on some open submanifold U  Rpjq . Then
the OU -module of R-linear derivations of OU is a rank pjq free sheaf over OU with
basis f@=@t i ; @=@ j g where @=@t i , @=@ j are the vector fields defined in U as
@ @fI .t / I @
i
.fI .t / I / D  ; .fI .t / j  I / D fI .t / I ;
@t @t i @ j
where j … I .
Proof. Assume that X is a vector field over the superchart .U; t i ;  j /. It is a simple
p p1
P X.Ii x /@ IxP , where
check that as usual Ix D ff 2 OM .U / j f .x/ D 0g. Define
Xy ´ X.t / @t i C X. j / @ @j . Clearly Xy is also a superderivation. Consider
the difference D D X  Xy . This is still a superderivation and, moreover, is zero on
the polynomials in the t i ,  j . If f 2 ORpjq .U / then, due to Lemma 4.1.9, for each
x 2 jU j and for k > q C 1, we can write f D Pk;x C h with h 2 IxkC1 . Hence
D.f / D D.h/ 2 Ixk . Since x is arbitrary in U , we can conclude using Lemma 4.1.9
that D D 0. 
 4.4 Vector fields and differential operators 71

Definition 4.4.5. Let M D .jM j; OM / be a supermanifold. A rank pjq locally free

sheaf on M is called a super vector bundle on M .
Since U  Rpjq is the local model for any dimension pjq supermanifold M , the
lemma implies that VecM is a vector bundle of rank pjq. If V is a vector field on
U , then in a coordinate neighborhood U 0  U with coordinates .t;  /, there exist
functions fi , gj on U 0 so that V has the unique expression
X
p
@ X
q
@
V jU 0 D fi .t;  / i C gj .t;  / j : (4.3)
@t @
iD1 j D1

We have defined tangent vectors at x as R-linear derivations Ox ! R of the stalk

at x; we may also think of a tangent vector v 2 Tx .M / as a (not uniquely determined)
vector field on U , a neighborhood of x, composed with evaluation at x. If the open
subset U from Definition 4.4.1 is a coordinate neighborhood around x, the vector v
takes the expression ˇ ˇ
X @ ˇ @ ˇ
vD ai i ˇˇ C bj j ˇˇ
@t x @ x
for ai ; bj 2 R, where
ˇ ˇ
@ ˇˇ @ @ ˇˇ @
´ evx B i ; ´ evx B j
@t i ˇx @t @ j ˇx @
correspond to the tangent vectors we defined in Proposition 4.3.9. Notice that, contrary
to the ordinary case, a vector field V on U is not determined by the family of tangent
vectors: evx B V .
For M and N supermanifolds, we can extend a vector on M to a ON -linear deriva-
tion on M  N , and likewise we may trivially treat any vector field on M as a vector
field on M  N . We will call these extensions extended vectors and extended vector
fields, respectively.
Definition 4.4.6. Let v be a tangent vector of M at m and Um  jM j an open
neighborhood of m. We view v as a derivation OM .Um / ! R and identify ON with
ORN . Then v extends uniquely to a ON -linear derivation

vN W OM N .Um  V / / ON .V /
SSS pp
SSS p
SSS pp
SSS
SS) ppppp
wp
ORN .R  V /

for any open V  jN j (this is easily seen locally by using coordinates, and then by
patching using local uniqueness) so that
vN .a ˝ b/ D v.a/b
where a and b are local functions of M and N , respectively.
72 4 Differentiable supermanifolds

One may similarly “extend” vector fields: let V be a vector field on M , and denote
by 1 the identity operator on the sheaf of the supermanifold. Then we extend V to a
derivation .V ˝ 1/ on M  N by forcing V to act trivially on N . If .t;  / and .x; / are
local coordinates on M and N respectively, V has the coordinate expression as in (4.3).
Then the extension .V ˝ 1/ has the same coordinate expression on M  N described
by coordinates .t; x; ; /, i.e., it is identically zero on .x; /. Again the extension is
made to be unique by patching and using local uniqueness.
The fact that we can define vN just by describing vN .a ˝ b/ will be fully justified
in Section 4.5.
We now want to consider the effect of a derivation on a pullback.

Proposition 4.4.7 (Chain rule). Let U pjq and V mjn be superdomains and denote col-
lectively by ua and v b the supercoordinates over U pjq and V mjn , respectively. If
W U pjq ! V mjn is a morphism, we have
 X @  .v b /  
@ .f /  @f
a
D : (4.4)
@u @ua @v b
b

Proof. The proof makes use of Lemma 4.1.9 and goes along the same line as that of
 P  b  @f
Lemma 4.4.4. Let us write D.f / ´ @ @u.f a
/
and D 0 .f / ´ b @ @u.va /  @v b and
0 0
consider the derivation D  D . Clearly D  D annihilates the supercoordinates v b
and hence all polynomials in them. If f is a generic section in O.V mjn / and x 2 V
then, in view of Lemma 4.1.9, there exists a polynomial Pk;x with k D n C 1 and
h 2 IxnC2 such that f D Pk;x C g. Hence .D  D 0 /.f / D .D  D 0 /.g/. On the
other hand it is easy to check that .D  D 0 /.g/ belongs to IxnC1 . Since x is arbitrary
in V , we can conclude that D D D 0 . 

In the ordinary setting equation (4.4) is rewritten in matrix form in terms of the
Jacobian matrix and we have that the composition of morphisms of ordinary manifolds
translates into matrix multiplication of their associated Jacobians. In the super setting
this requires some care due to the appearance of signs. Let us start with a definition of
the Jacobian supermatrix or Jacobian for short.

Definition 4.4.8. Let W M ! N be a morphism of supermanifolds. Let x 2 jM j be a

point with a local coordinate system t i ,  j and .x/ with local coordinate system s k , l .
We define the Jacobian supermatrix associated with a morphism in a neighbourhood
of x as  b a  b
J ´ .1/.jv jC1/ju j @ @u.va /
where .ua / D .t i ;  j / and .v b / D .s k ; l /. Explicitly we have
!
@  .s k / @  .s k /
@t i  @ j
@ . l / @ . l /
:
@t i @ j
 4.4 Vector fields and differential operators 73

Proposition 4.4.9. Let  W M ! N , W N ! P be two supermanifold morphisms

and let ua , v b , w c be local coordinate systems around points m, .m/ and ..m//.
Then
J B DJ J :
Proof. Let  ´ B . By Proposition 4.4.7 we have
@  .w c / X @  .v b / @  .w c /
D 
@ua @ua @v b
b
X b c a c a @ 
.w c / @  .v b /
D .1/jv j.jw jCju j/C1Cjw jju j   :
@v b @ua
b

(Recall that the parity of @f

@x
is jf j C jxj.)
So we can rewrite the expression, with some abuse of notation, as J B ´ J  J .
Reintroducing even and odd coordinates t i ;  j and s k ; l on U and V respectively, we
get our result. 
We end this section by briefly discussing the superalgebra of differential operators
on a supermanifold M .
Let M D .jM j; OM / be a supermanifold.
Definition 4.4.10. Let U be an open subset of jM j. We define the differential operators
of degree k in U as the subset of Hom.OM .U /; OM .U // defined inductively as follows.
The differential operators of degree zero Diff 0 .U / are the elements of OM .U / acting
multiplicatively. The differential operators of degree k are
Diff k .U / ´ fD 2 Hom.OM .U /; OM .U // j ŒD; f  2 Diff k1 .U /
for all f 2 OM .U /g;
where ŒD; f  denotes the commutator in Hom.OM .U /; OM .U // (see Chapter 1 for
its definition).
Clearly
Diff k1 .U / Diff k .U /
and each Diff k .U / inherits from Hom.OM .U /; OM .U // the obvious gradation.
We define the differential operators over U as the subalgebra
1
[
Diff.U / ´ Diff k .U /:
kD0

of Hom.OM .U /; OM .U //. In particular

Diff 1 .U / D OM .U / ˚ VecM .U /
and, as one can easily check, if D 2 Diff 1 .U /, then D.1/ 2 OM .U / and .D D.1// 2
VecM .U /.
74 4 Differentiable supermanifolds

Proposition 4.4.11. The assignment U 7! Diff.U / defines a sheaf on jM j. For each

positive integer k, U 7! Diff k .U / is a subsheaf. For each x 2 jM j, there exists a
chart .U; h/ containing x such that
M @jI j @j j
Diff k .U / ' O.U / :
@t I @
jI C jk

Proof. The proof goes along the same lines as the classical one and is based on
Lemma 4.4.4 together with an induction argument. We leave this to the reader as
an exercise. 

4.5 Global aspects of smooth supermanifolds

The purpose of this section is to show that a smooth supermanifold structure on a
topological space is fully encoded by the superalgebra of the global sections of the
structure sheaf of the supermanifold.
This is a technical result and we shall achieve it in several steps. We first need
the notion of Fréchet superalgebra and Fréchet supersheaf, which are very natural
generalizations of the ordinary corresponding notions and will be our technical device
for reconstructing the supersheaf. We have provided in Appendix C a brief summary
of the definitions and results concerning Fréchet spaces and superspaces that we will
need in this section.
We then study the maximal spectrum of the superalgebra O.M / of the global
sections, consisting of all the maximal ideals. Such ideals are in one-to-one corre-
spondence with the points of the topological space jM j underlying the supermanifold
and we shall see how to define a topology on the maximal spectrum so that we get a
homeomorphism with the topological space jM j.
Finally our crucial result, Proposition 4.5.9, builds the sections of the supersheaf
OM over an open set U as certain quotients of global sections, thus allowing us to
retrieve the full sheaf OM , just by the knowledge of O.M /.
This section is independent from the rest and may be skipped for a first reading,
with the one warning that the result in Proposition 4.5.4 is going to be used implicitly
in many places. The reader willing to take it for granted can very well go to the next
section.
We start by recalling two results which are proven in Appendix C using Fréchet
spaces terminology.
Proposition 4.5.1. Let M be a supermanifold of dimension mjn.
(1) OM is a Fréchet supersheaf with respect to the family of seminorms

B
pK;D .f / ´ sup j.D.f //.x/j; f 2 OM .U /;
x2K
 4.5 Global aspects of smooth supermanifolds 75

where D is a differential operator and K a compact in U .

Furthermore, OM is a sheaf of Fréchet superalgebras; in other words, OM .U / is
a Fréchet superalgebra for all U with respect to the submultiplicative seminorms

q˛;Kni ´ 2˛C2n max .pKni ; @ /;

j j˛;jI j @t  @ I

where fKni gi;n2N is a countable family of compact subsets covering U with each Kni
contained in a superchart.
(2) For all open coverings fUi g of U , OM .U / has the coarsest topology that makes
the restriction OM .U / ! OM .Ui / continuous.

Proposition 4.5.2. (1) If D is a super differential operator on M , then for each U ,

D W OM .U / ! OM .U / is continuous.
(2) If W M ! N is a supermanifold morphism then  W O.M / ! O.N / is
continuous.

As a consequence we have an important proposition which allows to construct

morphisms on cartesian products of supermanifolds in a simple way. Let us first recall
few well-known facts about the projective tensor topology.
Given two locally convex topological vector spaces V and W , it is possible to
endow the algebraic tensor product V ˝ W with several, in general inequivalent,
locally convex topologies. We do not want to enter into the details and we refer the
reader to the original references [40], [39] by Grothendieck or to [73]. The important
point is that if U Rm and V Rn are two open subsets, then there exists a unique
topology, called the projective tensor topology on C 1 .U / ˝ C 1 .V / such that

C 1 .U / y̋ C 1 .V / ' C 1 .U  V /;

where C 1 .U / y̋ C 1 .V / denotes as usual the corresponding completion. Nevertheless

it is important to stress that the precise nature of the projective topology is not essential
to our arguments, and that the topology induced by the family of seminorms in our
examination is equivalent to it. This is due to a nice property of the spaces of smooth
functions called nuclearity. We refer again to [73] for further details on this topic.
If Vi and Wi (i=1,2) are locally convex topological vector spaces and i D Vi ! Wi
are continuous linear maps, then the tensor product 1 ˝ 2 uniquely extends to a
continuous map 1 Ő 2 W V1 Ő V2 ! W1 Ő W2 .
We are ready to give a more intrinsic definition of product of supermanifolds than
the one we have encountered in Definition 4.2.17. As one can readily see, the two
definitions are equivalent.

Definition 4.5.3. Let M and N be supermanifolds.

We define the product of the supermanifolds M and N to be the super ringed space

M  N D .jM j  jN j; OM N /;
76 4 Differentiable supermanifolds

where the sheaf OM N is defined as follows. For each rectangular open set U  V
M  N define the superalgebra

OM N .U  V / ´ OM .U / Ő ON .V /:

Due to Proposition 2.2.11, since the rectangular sets are a base for the product topology,
it extends to a sheaf over jM j  jN j that is clearly locally trivial.
Involving again Proposition 2.2.11 (but this time item (2)), a similar reasoning can
be used for the morphisms.
The next proposition formalizes the important facts about the product of two su-
permanifolds.
Proposition 4.5.4. (1) O.M / ˝ O.N / is dense in O.M  N / when they are endowed
with the projective tensor topology.
(2) If i W Mi ! Ni , i D 1, 2 are supermanifold morphisms, then the pullback of
the map 1 2 W M1 M2 ! N1 N2 is given by 1 y̋ 2 which is in turn completely
determined by 1 ˝ 2 .
Notice that this proposition allows us to simplify dramatically the definition of
the pullback morphism on the sheaf on the product of two supermanifolds. We have
already used this in Definition 4.4.6 and we are going to use it again many times in the
text, mostly without mention.
We now turn to the problem of how to recover the topological space jM j of a
supermanifold .jM j; OM / by the superalgebra of its global sections. We start with a
simple proposition.
Proposition 4.5.5. Every x 2 jM j defines the superalgebra morphism evx W O.M / !
R given by evaluation at x, and ker evx D Jx is a maximal ideal in O.M /, with

Jx ´ ff 2 O.M / j f .x/ D 0g:

Proof. This is immediate since O.M /=Jx is a field. 

Definition 4.5.6. Let A be a commutative real superalgebra. The real spectrum of A
is defined as the set of all maximal ideals in A.

MaxSpecR .A/ ´ fM  A is a maximal ideal in Ag:

Classically, if M is an ordinary manifold, we have a one-to-one correspondence

between the points of M and the maximal spectrum of the algebra of smooth functions
on M :
MaxSpecR .C 1 .M // $ jM j; ker.evx / $ x:
This is a non-trivial result, playing the role of the Hilbert nullstellensatz in the context
of smooth supermanifolds and goes under the name of Milnor exercise (see [58] for
more details).
 4.5 Global aspects of smooth supermanifolds 77

Proposition 4.5.7 (Super Milnor exercise). All maximal ideals in O.M / are of the
form Jx for some x 2 jM j.

Proof. We have already noticed that each Jx is a maximal ideal in O.M /. Conversely,
let I be a maximal ideal in O.M / and denote IQ the corresponding ideal in C 1 .M z /,
Q 1 z Q
i.e., I ´ ff 2 C .M / j there exists g 2 I such that jgj D f g. I is a maximal ideal
in C 1 .M /. It cannot be the whole C 1 .M / since otherwise I would contain the unit.
Due to the classical Milnor exercise (see [58]) we have that there exists x 2 jM j such
that IQ D Jzx , where Jzx denotes the ideal of smooth functions vanishing at x. Let Jx
be the preimage of Jzx . Clearly I Jx , hence due to maximality I D Jx . 

From now on, due to the previous proposition, we can identify the real spectrum
MaxSpecR .O.M // with Hom.O.M /; R/, hence throughout this section we shall iden-
tify a point x with a maximal ideal Jx and with the morphism evx .
We now want to give to MaxSpecR .O.M // a topological space structure.
For each point xN 2 jM j, for each n 2 N, each n-tuple of elements f1 ; : : : ; fn 2
O.M / and each real number , we define the subset

B .evxN I f1 ; : : : ; fn / ´ fevx 2 MaxSpecR .O.M // j jfi .x/  fi .x/j

N <  for all i g:

of MaxSpecR .O.M //. As one can readily check, these subsets define a base for a
topology on MaxSpecR .O.M //.

Proposition 4.5.8. The map

W jM j ! MaxSpecR .O.M //; x 7! evx ;

is a homeomorphism.

Proof. This is a classical result; we nevertheless include it for completeness. The fact
that this map is a bijection follows immediately from the previous proposition. To
show that it is a homeomorphism, we need to show that it is open and the preimage of
an open set is open. It is enough to perform this check on the open sets of the basis.
Let U D B .evxN I f1 ; : : : ; fn / be an open subset in MaxSpecR .O.M //; since each fi
is a smooth section, we have 1 .U / is open.
Let now V be an open subset in jM j. We want to show that .V / is open. Let
evx 2 .V / (i.e., x 2 V ). Choose U to be an open subset of jM j such that U
Ux V . Let f 2 O.M / be a section such that f jU D 1 and f jVx c D 0. It is clear
that B 1 .evx I f / .V /, hence any point evx in .V / has an open neighbourhood
2
entirely contained in .V /. 

We now want to see how it is possible to reconstruct the whole sheaf OM starting
from the supercommutative algebra of global sections O.M /.
We will use the localization technique, borrowed from the algebraic setting. We
briefly recall a few facts about it, referring the reader to [2], Ch. 3, for all the details.
78 4 Differentiable supermanifolds

The idea is the following: we want to invert some elements in the superalgebra of
sections of the structure sheaf of a supermanifold. Since our rings are not commutative
but supercommutative, we must exert some care.
Consider an open set U  jM j and define the subset of sections in O.M / that are
invertible over U . More precisely, put

U ´ fs 2 O.M /0 j sjU .x/ ¤ 0 for all x 2 U g:

The set U is a multiplicatively closed subset of O.M /0 , i.e., 1 2 U , and it is closed

under multiplication. Hence we can localize O.M / as an O.M /0 -module with respect
to U . This amounts to defining the set

U1 O.M / ´ .U  O.M //= ;

where the equivalence relation is given by: .s; f / .s 0 ; f 0 / if and only if there exists
s 00 2 U such that
s 00 .s 0 f  sf 0 / D 0:
By construction U1 O.M / is an O.M /0 -module, however U1 O.M / is also a super-
algebra if we define addition and multiplication by

.s; f / C .s 0 ; f 0 / ´ .ss 0 ; s 0 f C sf 0 /; .s; f /  .s 0 ; f 0 / ´ .ss 0 ; ff 0 /:

The next result is crucial for the reconstruction of the structure sheaf of the super-
manifold from the superalgebra of its global sections.
Proposition 4.5.9. The map
f jU
` W U1 O.M / ! OM .U /; .s; f / 7! ;
sjU
is a superalgebra isomorphism.
Proof. Let us first prove that ` is injective. Suppose hence that fsjjUU is zero. This
means that f jU D 0. Let s 2 U be such that sjUx c D 0, then sf D 0, and injectivity
is proved.
We now come to surjectivity. Let f 2 OM .U /. We want to determine k 2 O.M /
and h 2 SU such that f D kjU = hjU . Let S fUi g be a collection of open sets with
Si UiC1 and
compact closure such that U i Ui D U . Let fgi g be sections in O.M /
such that gi jUi D 1, supp gi UiC1 and 0  jgi j  1 (see Corollary 4.2.10).
Consider
1
X 1
X
1 gi f 1 gi
kD and hD ;
2 1 C qO i .gi / C qO i .gi f /
i 2 1 C qO i .gi / C qO i .gi f /
i
iD1 iD1

where
qO i ´ maxfqi1 ; : : : ; qi i g:
 4.6 The functor of points of supermanifolds 79

(For the definition of the submultiplicative seminorms qi ’s refer to Proposition 4.5.1.)

We claim that they are both well-defined sections. Let us prove this fact, for example,
P
for the section k. We prove that the series 1 1 gi f
iD1 2i 1CqO i .gi /CqO i .gi f / converges by
showing that it is Cauchy. Nevertheless this is immediate from
X
s  X
s X
s
1 gi f 1 qO u .gi f / 1
qO u  
2 1 C qO i .gi / C qO i .gi f /
i 2 1 C qO i .gi / C qO i .gi f /
i 2i
iDr iDr iDr

for r; s  u. Hence the map is surjective. 

Corollary 4.5.10. Let M D .jM j; OM / be a supermanifold. Then the superalgebra

of global sections O.M / determines the sheaf OM :

jM j U 7! U1 O.M / Š OM .U /:

Remark 4.5.11. (1) It is important to remark that, as it happens already in the ordinary
case, OM .U / is in general larger than O.M /jU , that is, there are sections on U that do
not come as the restriction of global sections (e.g., 1=x 2 C1 .R / is not the restriction
of any global section on R).
(2) The superalgebra O.M / does not embed into its localization U1 O.M / Š
OM .U /. This has nothing to do with the odd nilpotents, but it is a phenomenon we
already observe at the ordinary level. Thanks to the partition of unity, we can have
non-zero global sections which are zero on an open set.

4.6 The functor of points of supermanifolds

In the previous section we have shown that, starting from the superalgebra of global
sections of a smooth supermanifold M , it is possible to reconstruct both the underlying
topological space jM j, and the sheaf OM .
In this section we want to give, as a fundamental application of the results of the
previous section, an effective way to compute the T -points of a supermanifold. In fact
we will show that a morphism between the superalgebras of global sections of two
supermanifolds M and N completely determines a morphism of the corresponding
smooth supermanifolds, thus identifying the T -points of a supermanifold M with the
morphisms between O.M / and O.T /, which are in general much more manageable.

Proposition 4.6.1. Let M and N be smooth supermanifolds. There is a bijective

correspondence
Hom.M; N / $ Hom.O.N /; O.M //:

Proof. One direction is completely straightforward, so let us concentrate on the other

one.
80 4 Differentiable supermanifolds

Suppose that W O.N / ! O.M / is a superalgebra morphism. We define a super-

manifold morphism .j j;   / in the following way. The reduced map is

j j W MaxSpecR .O.M // ! MaxSpecR .O.N //; evx ! evx B :

The sheaf morphism   is defined, using Proposition 4.5.9, by

  ;U W ON .U / ! OM .j j1 .U //; .s; f / 7! . .s/; .f //:

Let us now check that .j j;   / is really a supermanifold morphism .

The reduced map j j is well defined since evx B is a superalgebra morphism.
It remains to prove that it is continuous. Let

U D B .evxN I f1 ; : : : ; fn / D fevz 2 MaxSpecR .O.N // j jfi .z/  fi .x/j

N < g

be an open subset in MaxSpecR .O.N //. We have

j j1 .U / D fevy 2 MaxSpecR .O.M // j jhevy  evxN ; fj ij  g

D fevy 2 MaxSpecR .O.M // j jhevy ; .fj /i  fj .x/j
N  g;

which is open due to the smoothness of the sections .fj /.

The fact that   defines a sheaf morphism is easy. 

Summarizing all of the results of this and the previous section, we can state the
following theorem, which is one of the main results of this chapter.

Theorem 4.6.2. The functor

F W .smflds/ ! .salg/

that assigns to each supermanifold M the supercommutative algebra O.M / and to


each morphism .jj;   / the superalgebra map M is a full and faithful embedding.

Proof. The fact that F is a functor is straightforward. The bijectivity on the objects is
a consequence of Proposition 4.5.8 or in any case is coming from the classical result.
The fully faithfulness amounts to Proposition 4.6.1. 

One of the striking consequences of Theorem 4.6.2 is that morphisms of super-

manifolds can be assigned at the level of global sections. In other words, a morphism
M ! N is determined by the knowledge of just the morphism induced on the su-
peralgebras of global sections O.N / ! O.M /. We shall use this result many times
in the remaining chapters. We are now going to see in the next examples how simple
the description of the functor of points of a supermanifold becomes with the help of
Theorem 4.6.2.
 4.6 The functor of points of supermanifolds 81

Examples 4.6.3. (1) The super vector space Rmjn as superdomain. As we have seen
in Chapter 3, Section 3.4, if T is any supermanifold, we have that the T -points of the
supermanifold Rmjn are given by

Rmjn .T / D Hom.T; Rmjn / D Hom.O.Rmjn /; O.T //:

This means that a supermanifold morphism f W T ! Rmjn corresponds to a superal-

gebra morphism f  W O.Rmjn / ! O.T / and by Hadamard’s lemma such a morphism
is known once we know the image of the elements in the canonical chart: f  .t i /,
f  . j /. Vice versa, the Chart Theorem tells us that for any choice of m even elements
and n odd elements in O.T / there is a unique corresponding morphism T ! Rmjn .
Hence

Rmjn .T / Š f.f  .t 1 /; : : : ; f  .t m /; f  . 1 /; : : : ; f  . n // j f W T ! Rmjn g

D O.T /m
0 ˚ O.T /1 D .O.T / ˝ R
n mjn
/0 ;

where the first Rmjn is the functor of points of the supermanifold Rmjn , while the second
Rmjn is the real super vector space of superdimension mjn, Rmjn D Rm ˚ Rn . In other
words, a T -point of the supermanifold Rmjn consists of an m C n-uple of m even and n
odd global sections of O.T /. We realize that the notation Rmjn is used here to denote
three different objects: the superdomain, its functor of points and the real super vector
space; however this abuse of notation is justified by the above identification, it is widely
spread, and we shall make sure that the context clarifies at each time which one we are
referring to.
(2) The general linear supergroup. Let T be a supermanifold. In Example 3.1.5, we
2 2
have described the super vector space of matrices Mmjn Š Rm Cn j2mn . Reasoning as
in (1), we have that Mmjn .T / can be identified with the endomorphisms of the O.T /0 -
module Rmjn .T / Š .O.T / ˝ Rmjn /0 , by rearranging the m2 C n2 even sections and
the 2mn odd ones in a matrix form with diagonal block matrices with even entries and
off diagonal matrices with odd entries:
² ³
A B
Mmjn .T / D D .Mmjn /0 ˝ O.T /0 ˚ .Mmjn /1 ˝ O.T /1 :
C D

Here A D .aij /, B D .ˇil /, C D .kj /, D D .dkl / with aij ; dkl 2 O.T /0 , and
ˇil ; kj 2 O.T /1 . .Mmjn /0 and .Mmjn /1 denote respectively the even and odd part
of the super vector space of matrices Mmjn (again the same symbol Mmjn has several
different meanings).
Notice that in doing this, we reobtain Mat.Amjn /0 as discussed in Chapter 1, Sec-
tion 1.4, with A D O.T /.
Define F .T / as the group of automorphisms of Rmjn .T /. Clearly F .T /  Mmjn .T /.
We want to show that F D GLmjn , i.e., F is the functor of points of the general linear
supergroup, the supermanifold defined in Chapter 3, Sections 3.1–3.2. In that example,
82 4 Differentiable supermanifolds

GLmjn is defined as the open submanifold of the supermanifold Mmjn , whose reduced
space consists of diagonal block mCnmCn real matrices with non-zero determinant:
²  ³
P 0
j GLmjn j D j det.P / ¤ 0; det.Q/ ¤ 0 :
0 Q
1
By Proposition 4.5.9 we have O.GLmjn / D jGL O.Mmjn /.
mjn j
Hence
1
GLmjn .T / D Hom.SjGL mjn j
O.Mmjn /; O.T //
 Hom.O.Mmjn /; O.T // D Mmjn .T /:
1
The elements in Hom.jGL O.Mmjn /; O.T // correspond to .mjnmjn/-matrices
mjn j
with entries in O.T / whose diagonal blocks are invertible once reduced. We leave it to
the reader as a simple exercise to check that those correspond to .mjn  mjn/-matrices
with entries in O.T / whose diagonal blocks are invertible (recall that t C n 2 O.T /,
n nilpotent, is invertible if and only if t is invertible).
Notice that while in general, as we pointed out in Remark 4.5.11, O.M / does not
embed into U1 O.M /, it does in this special case, and this is because here U D jGLmjn j
is dense in jMmjn j.
Hence
²  ³
A B
GLmjn .T / D j A; D invertible  Mmjn .T /:
C D
This is what we have defined as GLmjn .A/ in Definition 1.4.4 for A D O.T /.

4.7 Distributions with finite support

This section generalizes the previous one and we include it since, in Kostant’s original
approach to supermanifolds, Sweedler duals play a central role. Finite support distri-
butions play an important role also in classical differential geometry (see, for example,
[26], [17]). We want, however, to remark that we shall not take this point of view
in examining the theory of supergeometry, so this section is not going to be used in
the sequel, except for those instances in which we relate our treatment with Kostant’s
original one.
Suppose that v 2 Tx M . Then ker v contains Jx2 , which, due to Proposition 4.3.3,
is an ideal of finite codimension in OM;x . Alternatively, using the characterization in
Observation 4.3.10, a direct computation shows that ker v Jx2 , where
Jx ´ ff 2 O.M / j f .x/ D 0g:
Using partitions of unity and Proposition 4.3.3, it is easy to prove that also Jxp is a
finite codimension ideal for each x 2 jM j and p 2 N.
 4.7 Distributions with finite support 83

Definition 4.7.1. Let A be a superalgebra and denote by A the corresponding algebraic

dual. The Sweedler dual of A is defined as

Aı D fX j X 2 A ; ker X contains an ideal of finite codimensiong:

It is well known that the importance of the Sweedler dual stems from the fact that it
inherits a coalgebra structure from the algebra structure of O.M /. Indeed if an algebra
A is not finite-dimensional then the adjoint of the multiplication

h X; a ˝ bi ´ hX; a  bi:

does not turn A into a co-algebra since

.A ˝ A/ © A ˝ A :

On the other hand, Aı is a coalgebra with respect to the comultiplication defined

above and counit
 W Aı ! R; X 7! hX; 1i:
The proof essentially follows the classical one (see [72]) and we are going to return to
this with more details in Chapter 7.
Following the classical notation, let O.M /0 denote the topological dual of O.M /,
that is the set of continuous linear operators O.M / ! R. We shall call such operators
distributions in analogy with the classical terminology.

Definition 4.7.2. Let  be a distribution in O.M /0 . A point x in jM j is said to be in

the support of  if for each neighborhood U of x there exists a section f 2 O.M /
supported in U such that h; f i ¤ 0. If the support of  consists of a finite number
of points, we say that  is a finite support distribution; if the support of  consists
of the point x, we say that  is a finite support distribution at x or a point supported
distribution.

The reader can easily check that the support of a distribution is a closed subset of
jM j. As we are going to see presently, the finite supported distributions are identified
with the Sweedler dual. Before this, we need the characterizations of the finite support
distributions and the finite codimension ideals.

Proposition 4.7.3. (1) Let  be a finite support distribution at x and let .t i ;  j / be

local coordinates around x. Then
X @ @
D aIJ I
jx J jx :
@t @
P
(2) Let  be a finite support distribution. Then  D i xi , where xi is a finite
support distribution at xi .
84 4 Differentiable supermanifolds

Proof (Sketch). (1) Let U be a domain where the local coordinates .t i ;  j / make sense
and temporarily denote by OM .U /? the space of distributions with finite support. Since
supp   U , we have that  is completely determined by jOM .U / . Since U is a
domain, OM .U / Š C 1 .U / ˝ ^. j /, hence OM .U /? ' C 1 .U /? ˝ ^. j /? Since
^. j / is finite-dimensional, its topological dual coincides with its dual as a vector
space. Moreover, since by Schwarz’s theorem (see [64], p. 150) we have that finitely
supported distributions at x are identified with differential operators evaluated at x, we
immediately obtain the result.
(2) Locally at each point of its support,  can be written as a point supported
distribution. Using a suitable partition of unity we obtain the global expression of  as
the sum of differential operators (that is, point supported distributions) at each point of
the support. 

There is a topological interpretation of the Sweedler dual, providing its “concrete”

realization. This is the content of the next proposition. Before this however we need a
lemma characterizing ideals of finite codimension.

Lemma 4.7.4. Let I be an ideal of finite codimension in O.M /. Then there exists a
finite set x1 ; : : : ; xn in jM j and an n-tuple of integers p1 ; : : : ; pn such that

Jxp11 \    \ Jxpnn I Jx1 \    \ Jxn :

Proof. Let U be an open subset of jM j and write OM .Ux / ´ O.M /jU . In particular
C 1 .Ux / D C 1 .M /jU identifies with the restriction of smooth functions to the closed
subset Ux .
If J is a finite codimension ideal in OM .Ux /, let JQ CM 1 x
.U / denote the associated
Q
reduced ideal. Clearly J is a finite codimension ideal in CM .U /, and JQ D CM
1 x 1 x
.U / if
and only if J D OM .Ux /. If f 2 CM .Ux /, let Zf ´ f .0/, and let
1 1

T
ZJ ´ ZJQ ´ ZfQ :
f 2J

It is well known (see, for example, [58]) that ZJ is a finite subset fx1 ; : : : ; xn g of jM j,
that ZJ D ; if and only if JQ D CM 1 x
.U /, and that

JQ JzZJ ´ Jzx1 \    \ Jzxn :

Taking U D M , it then follows that J Jx1 \    \ Jxn .

We now claim that if V is an open subset of jM j such that ZJ V , and f 2 O.M /
is such that f jV D 0, then f 2 I . Indeed let U be another open subset of jM j
such that ZJ U Ux V , and denote by OM .M n U / the supersubalgebra of
O.M / whose elements are restrictions of global sections to Ux c . Clearly the restriction
W O.M / ! OM .M n U / is surjective, moreover .I / is a finite codimension ideal
of OM .M n U / with ZJ D ;. Then .I / D OM .M n U /. Hence let g 2 I be such
 4.7 Distributions with finite support 85

that gjUx c D f jUx c and let h 2 O.M / be such that hjVx c D 0 and hjU D 1. Then
f D f  h D g  h 2 I.
We are now ready to end the proof by showing that there exist integers p1 ; : : : ; pn
such that Jxp11 \    \ Jxpnn I . Indeed notice that Ix i ´ x i .I / is a finite codimen-
p
sional ideal of OM;x i and hence, due to Proposition 4.3.5, contains Jx ii for some integer
pi . Let now f 2 Jxp11 \    \ Jxpnn and let gi 2 I be such that Œf x i D Œgi x i . Denote
by Vi the open subsets of jM j such that f jVi D gi jVi , and let Ui be another family of
open sets such that Ui U Si Vi . There exists a partition of unity fh1 ; : : : ; hn ; hg sub-
P
ordinated to the open cover fV1 ; : : : ; Vn ; US1 \    \ USn g. Then f D f  hi C f  h
c c

and f  h is in I since it is zero in a neighborhood of ZJ , while f  hi D gi  hi is

obviously in I . 

The next proposition identifies the finite support distributions with the Sweedler
dual.

Proposition 4.7.5. Let O.M /0 be the topological dual of O.M /, then

O.M /ı D f 2 O.M /0 j supp  consists of a finite number of points g:

Proof. (). Let  2 O ı .M /. Then ker  contains an ideal I of finite codimension

and by the previous lemma we have ker  I Jxp11 \    \ Jxpnn . We now claim
that supp   fx1 ; : : : ; xn g. Let x 62 fx1 ; : : : ; xn g and choose open mutually disjoint
neighbourhoods U , Ui of x and the xi ’s, respectively. Let f 2 O.M / with supp f 
p
U . Then f 2 Jxii for all i and consequently f 2 ker , hence x is not in the support
of .
( P). If supp  D fx1 : : : xn g, then, due to point (1) of Proposition 4.7.3, we have
 D xi , where xi is a distribution with support at xi . We can hence restrict
ourselves to consider the case in which  is a point supported distribution (i.e., its
support consists of just one point). In this case the result follows easily using (2) of
Proposition 4.7.3. In fact, ker  Jxi . 

Lemma 4.7.6. O.M /ı separates the points of O.M /, that is,

hf  h; x i D 0 for all x 2 O.M /ı H) f D h;

where x is a point supported distribution at x 2 jM j.

Proof. Let a D f  h. We are going to show that if a ¤ 0, then there exists a point
supported distribution x with x .a/ ¤ 0. Let U be an open coordinate neighborhood
such that X
ajU D aI  I ¤ 0:
I

Clearly there exist aJ and x 2 U such that aJ .x/ ¤ 0. Consider the element x of
O.M /ı given by @ @J x , then x .a/ ¤ 0. 
86 4 Differentiable supermanifolds

Proposition 4.7.7. Each supermanifold morphism W M ! N uniquely determines

a morphism of super coalgebras ı W O.M /ı ! O.N /ı given by ı ./ D  B  .
Proof. ı is well defined. In fact, suppose that  2 O.M /ı . Then ker  Jxp11 \
pn
   \ Jxn , the latter being an ideal of finite codimension. Since j  .f /j.x i / D
f .j j.x i // D 0, we have ker  B  Jjp1j.x1 / \    \ Jjpnj.xn / . 

The above proposition thus establishes that the map


 Hom.O.M /ı ; O.N /ı /
Hom.M; N / !

is injective. However, in general, this map is not bijective.

4.8 Complex and real supermanifolds

In our description of the foundations of the theory of supermanifolds, we have limited
our discussion to the category of differentiable supermanifolds. Most of our defini-
tions, however, hold also in other contexts and in this section we are interested in a
supergeometric theory generalizing real or complex analytic (or holomorphic) mani-
folds.
Definition 4.8.1. Let k be the real or the complex field and let Hk p denote the sheaf
of real or complex analytic functions on k p . We define the analytic superdomain k pjq
as the topological space k p endowed with the following sheaf of superalgebras. For
any open subset U  k p ,
V
Ok pjq .U / ´ Hk p .U / ˝ .1 ; : : : ; q /;
V
where . 1 ; : : : ;  q / is the exterior algebra generated by the q variables 1 ; : : : ; q .
Notation. In this section only, we use the symbol Rpjq to denote the real analytic
superdomain instead of the C 1 superdomain as in the rest of the text.
A real analytic or complex analytic supermanifold of dimension pjq is a superspace
M D .jM j; OM / which is locally isomorphic to k pjq , i.e., for all x 2 jM j there exists
an open neighbourhood of x Vx  jM j and U  k p such that

OM jVx Š Ok pjq jU :

A morphism of real or complex analytic supermanifolds is simply a morphism of their

underlying superspaces.
Most of the results discussed in Chapters 4 and 5 hold also for real and complex
analytic supermanifolds. For example, we have the chart theorem, we can give the
 4.8 Complex and real supermanifolds 87

exact same definition for tangent vector and tangent bundle to a real or a complex
supermanifold and we can prove in the same way all of the theorems relative to the local
structure of morphisms, like the inverse function theorem, the submersion, immersion
and the constant rank morphism theorems.
There are however some important differences at the very core of the theory: par-
titions of unity are not available for real or complex analytic (super)manifolds, so one
must exert extreme care in generalizing all the results that make use of them. Moreover,
contrary to the smooth case, the superalgebra of the global sections of the structure
sheaf of a (super)manifold tells us very little, in general, about the structure of the
(super)manifold and in general does not allow us to retrieve all the information about
the (super)manifold itself.
The main goal of this section is to understand how it is possible to define real
structures and real forms in supergeometry. A major character in this game is the
complex conjugate of a super manifold.
Let us first review quickly the classical setting.
Let M D .jM j; HM / be a complex manifold. The complex conjugate of M is the
manifold M D .jM j; HM / where HM is the sheaf of the antiholomorphic functions
on M , which are immediately defined once we have HM and the complex structure on
M . We have the C-antilinear sheaf morphism
HM ! HM ; f 7! fN: (4.5)
In the super context it is not possible to speak directly of antiholomorphic functions
and for this reason we need the following generalization of complex conjugate super
manifold.
Definition 4.8.2. Let M D .jM j; OM / be a complex super manifold. We define a
complex conjugate of M as a complex super manifold M D .jMj; OM /, where now
OM is just a supersheaf, together with a super ringed space C-antilinear isomorphism
with M , which is (4.5) on the reduced supermanifold. This means that we have an
isomorphism of topological spaces jM j Š jMj and a C-antilinear sheaf isomorphism
OM ! OM ; f 7! fN: (4.6)
By an abuse of notation we shall also write f 7! fN for the inverse of the morphism
(4.6).
Example 4.8.3. A complex conjugate of M D C1j1 D .C; OC Π/ is M D .C; OC N
x Π/,
where OC and OC x denote respectively the sheaf of holomorphic and antiholomorphic
functions on C. In fact the C-antilinear isomorphism is
OM ! OM ; z 7! z;
N  7! N :
Notice that while zN has the meaning of being the complex conjugate of z, N is simply
another odd variable that we introduce to define the complex conjugate.
Practically one can think of the complex conjugate super manifold as a way of
giving a meaning to fN, the complex conjugate of a super holomorphic function.
88 4 Differentiable supermanifolds

Remark 4.8.4. The underlying real manifold MR of a complex manifold M is defined

via the identification of C Š R2 and its real dimension is double the complex dimen-
sion of M . Notice that the underlying real manifolds MR and MR of M and M are
isomorphic as real manifolds. If z D .z1 ; : : : ; zn / are complex local coordinates for
M , by a slight abuse of notation we can give real local coordinates for both M and M
N (where zN is the complex conjugate).
as .z; z/

We are ready to define a real structure on a complex supermanifold.

Definition 4.8.5. Let M D .jM j; OM / be a complex super manifold and M a complex

conjugate of M . We define a real structure on M as an involutive isomorphism of super
ringed spaces W M ! M , which is C-antilinear on the sheaves  W OM !  OM ,

. f / D N  .f /. We define the real form M of M defined by as the supermanifold
.jM j ; OM / where jM j are the fixed points of W jM j ! jMj D jM j and OM are
all the functions f 2 OM jjM j such that  .f / D f .

Remark 4.8.6. Observe that classically M D .jM j ; OM / has a real manifold struc-
ture. In coordinates, the map is z 7! z.
N Let us look at as a real differentiable map
from M to M , seen as real manifolds. Since this is a local question, we look at in a
neighborhood with coordinates z and zN (see Remark 4.8.4). We have

W M ! M; N 7! .z;
.z; z/ N z/:

We are looking at the fixed points m of , .m/ D m or .id  /.m/ D 0. So

M D .id  /1 .0/. The differential of .id  / is
 
1 1
:
1 1

By the ordinary constant rank theorem we have that M is a submanifold of dimension

half the real dimension of M .

We are ready to discuss examples of real structures and forms in the super context.

Example 4.8.7. We take M D C. In this case we have j j W C ! C, j j.z/ D zN and


.f .z// D f .zN /. It is clear that R D C and that the functions are f 2 OC jR (where
z D zN ) such that fN D f .

Example 4.8.8. M D C1j1 . j j is the same as above and  W OC Œ  ! OC ŒN ,

f .z;  / 7! f .z; N Again jM j is the same as above while the functions are those
N /.
f 2 OM jR such that fN.z; N D f .z; /. For example zjR and  are global sections
N /
satisfying such a requirement. Actually this example is particularly instructive since
it shows that the odd variables come out practically unchanged from this procedure.
For this reason it is immediate to see that M is a real super manifold since it is a real
classical manifold and the local splitting is preserved.
 4.9 References 89

If M is a complex supermanifold, one can always construct the complex conjugate

Mx in the following way using the functor of points notation. Take jMj D jM j and as
OM the sheaf OM with the complex conjugate C-algebra structure (that is f D N f ).
In order to obtain a real structure on M , we need a ringed spaces morphism M ! M
with certain properties. By Yoneda’s lemma this is equivalent to give an invertible
natural transformation between the functors of points:

W M.R/ ! M.R/

for all super ringed spaces R satisfying the C-antilinear condition.

4.9 References
The idea of supercalculus goes back to Berezin’s seminal work [10]. The concept
of supermanifold was however introduced later by Berezin and Leites in [11] and in
the beautiful work [49] by Kostant. In particular [49] can be considered as the first
systematic and rigorous treatment of the foundations of smooth supergeometry. The
approach there is very algebraic in nature with a particular attention to the coalgebra
structure of the structure sheaf of a supermanifold. Section 4.7 provides a bridge
between Kostant’s approach and ours.
Our treatment of superdomains, in particular Lemma 4.1.9 and the Chart Theo-
rem 4.1.11, follows quite closely [53]. The content of Section 4.5 and our treatment of
the localization procedure in the super-setting is found also in [6]. In Section 4.7, we
generalize the proof of Proposition 4.7.5 given in [58].
For further reading and a more modern treatment of smooth supergeometry the
reader can consult [56], [6], [22], [76].
5
The local structure of morphisms

The goal of this chapter is to study the local structure of supermanifold morphisms. As
in the ordinary setting we have the inverse function theorem and locally we can classify
the morphisms according to the rank of their differentials, so that we can formulate the
supergeometric versions of the immersion and submersion theorems. There is however
an important difference with the ordinary differentiable manifolds theory, which arises
when we discuss the constant rank morphisms, which are the natural generalization
of the immersion and submersion morphisms. There is in fact a difficulty with the
definition of rank of the Jacobian of a morphism: as we shall see this notion is not
always well defined. This is a supergeometric peculiarity, ultimately linked to the
graded nature of super vector spaces. The constant rank morphisms are crucial for a
complete treatment of submanifolds and it is the key tool for their explicit constructions
as we shall see in the section on submanifolds.
All of these are well-known results that appeared in several papers, including [53],
[56], [76]. Nevertheless we provide complete proofs of them in the effort to make the
text self-contained.

5.1 The inverse function theorem

We start by proving the super analogue of the inverse function theorem. As in the
ordinary case it is of fundamental importance. Its proof heavily relies on the classical
version of the theorem, for which we refer to [52].

Proposition 5.1.1 (The inverse function theorem). Let  W M ! N be a supermanifold

morphism and let m 2 jM j such that .d/m is bijective. Then there exist charts U and
V around m and jj.m/ respectively such that jj.U / V and jU W U ! V is an
isomorphism of U onto V .

Proof. Since the statement is local we can assume both M and N to be superdomains
U pjq and V pjq respectively. The superdimensions must be equal since the differential
is bijective.
Denote by x i , j and t r ,  s supercoordinates on U pjq and V pjq , respectively. By
the Chart Theorem 4.1.11,   W O.V pjq / ! O.U pjq / is given by
´ P
  .t r / D jP j0 Pr .x 1 ; : : : ; x p / P ;
P
  . s / D jQj1 ˆsQ .x 1 ; : : : ; x p / Q :
 5.1 The inverse function theorem 91

Our assumption on the differential being bijective amounts to saying that

0@ 1 @ 01
1 0 1
0
 @x p
ˆ1.1;0;:::;0/  ˆ1.0;0;:::;1/
B : @x 1
A´B :: :: C
C B
B´@ :: :: :: C
@ :: : : A; : : : A
p p p p
@ 0

@ 0 ˆ.1;0;:::;0/  ˆ.0;0;:::;1/
@x 1 @x p

are non-singular. We can hence apply the P classical inverse function theorem to find
an open subset W pjq
U pjq
where f0 ; k Bj k j g define a new supercoordinate
i
i Nj
system that we denote by fxN ; g. If we denote by  the corresponding change of
coordinates we have
´ P
.  B   /.t i / D xN i C jP j2 NPi .xN 1 ; : : : ; xN p / N P ;
P
.  B   /. j / D N j C jQj3
x̂ j .xN 1 ; : : : ; xN p / N Q :
Q

Hence   B  has the form id CN , where N is a matrixP

with nilpotent entries. This
implies at once that   B   has an inverse  given by .1/j N j where the sum
terminates due exactly to the nilpotency of N . Hence   is invertible. Consequently,
by the Chart Theorem,  is an isomorphism. 

Definition 5.1.2. Let  W M ! N be a supermanifold morphism. If .d/m is bijective,

we say that  is a local superdiffeomorphism at m 2 jM j. If .d/m is bijective for all
m 2 jM j, we say that  is a local superdiffeomorphism.
If  is invertible and its inverse is a supermanifold morphism, we say that  is a
superdiffeomorphism. To simplify matters we shall often drop the suffix “super” and
speak of “diffeomorphism” whenever it is clear in which category we are working.

The next proposition clarifies the relation between local superdiffeomorphism and
superdiffeomorphism.

Corollary 5.1.3. Suppose that  W M ! N is a supermanifold morphism such that

(1) jj is bijective,
(2) .d/x is an isomorphism for each x 2 jM j.
Then  is a superdiffeomorphism.

Proof (Sketch). We shall produce a supermanifold morphism D .j j;  / W N !

M , which is the inverse of . We have by j j D j 1 j by (1). As for the sheaf
morphism,   is locally invertible, so we can define U D   1 jU on any U in a
suitable open cover of N . Such morphisms agree on the overlap of open sets, thus we
have defined the required inverse. 
92 5 The local structure of morphisms

5.2 Immersions, submersions and the constant rank morphisms

When M and N are classical manifolds and  W M ! N is a supermanifold morphism,
there exists a relationship between the properties of the differential .d/x at a point
x 2 jM j and the local structure of the map  in a neighborhood of x. This leads to a
classification of smooth mappings in terms of the rank of their differentials.
The purpose of this section is to replicate this same discussion in the supergeometric
setting. As we already pointed out, there are important differences with the ordinary
case in the definition of constant rank morphism.
Let us recall that given a morphism  W M ! N , we can speak both of its differential
.d/m at m 2 jM j and of the Jacobian J in a neighborhood of m 2 jM j. Choosing
charts .U; t;  / and .V; x; / containing m and j j.m/, respectively, the differential
and the Jacobian can be written as
ˇ ! !
@  .x/ ˇ @  .x/ 
0  @ @ .x/
.d/m D @t m ˇ
@  ./ ˇ ; J D @  ./
@t
@  ./ :
0 @ m @t @

Clearly we have the relation Jz .m/ D .d/m .

In Chapter 1, Section 1.5, we have defined the rank of a supermatrix as the di-
mension of the largest invertible supermatrix contained in it. With this definition one
has immediately that the rank of a supermatrix coincides with the rank of its reduced
matrix, so in particular rk.J / D rk.d/m .
We are now going to introduce a definition that is a variation of the concept of rank
and is subtly related to it.
 Q
Definition 5.2.1. Let Z D P R S
2 Mpjqmjn .A/ for a commutative superalgebra
A; in other words, Z is a pjq  mjn matrix with diagonal blocks entries in A0 and
off diagonal block entries in A1 (our prototype for A is OM .U //. We say that Z has
constant rank rjs if there exist G1 2 GLmjn .A/ and G2 2 GLpjq .A/ such that G1 ZG2
has the form 0 1
idr 0 0 0
B 0 0 0 0C
G1 ZG2 D B@ 0 0 ids 0A :
C

0 0 0 0
Remark 5.2.2. It is important to notice that the constant  rank is not defined for all
supermatrices. For example, for the supermatrix Z D 0 00 2 M1j11j1 we cannot
define any constant rank; in fact, as one can readily check, it is not possible with a
basis change, that is, with left and right multiplication by elements of the general linear
supergroup, to transform Z into a diagonal matrix with 1 or 0 on the diagonal, as the
definition of constant rank requires. However, once the constant rank is defined, it
coincides with the rank as we defined it in Chapter 1, Section 1.5. We leave this check
to the reader as an exercise.
We can now give the following definitions.
 5.2 Immersions, submersions and the constant rank morphisms 93

Definition 5.2.3. Let the notation be as above.

(1) A morphism  W M ! N is said to be an immersion at m 2 jM j if .d/m is
injective.
(2) A morphism  W M ! N is said to be a submersion at m 2 jM j if .d/m is
surjective.
(3) A morphism  W M ! N is said to have constant rank rjs in a neighborhood U
of m 2 jM j, if J is a matrix of constant rank rjs (and entries in OM .U /).
Notice that in the first two definitions we refer to the differential of the morphism,
while in the last one the Jacobian of the morphism enters.
Each one of the three kinds of morphisms has its corresponding structure theorem
and, as we shall see, there is an important result relating all three of them. We now
begin our discussion of each of these three kinds of morphisms.

Immersions. Let  W M ! N be a morphism of supermanifolds, dim M D mjn,

dim N D m C pjn C q and let x 2 jM j. Suppose that there exist charts U , ft i gm iD1 ,
a p b q
f j gjnD1 around x, and V D V1  V2 , fy i gmiD1 , fz g aD1 , f g
j n
j D1 , f gbD1
around
jj.x/ such that the restriction of  to U has the form

y i 7! t i ; j 7!  j ; z a ;  b 7! 0:

By a common abuse of notation, whenever there is no danger of confusion, we shall

denote from now on the coordinates in the domain and their images by the same letter,
so we shall write y i and j instead of t i and  j . Clearly  is an immersion at x. The
converse is also true and this is the content of the following proposition.
Proposition 5.2.4. Let  W M ! N be a supermanifold morphism, with dim.M / D
mjn  dim.N / D m C pjn C q. The following facts are equivalent:
(1)  W M ! N is an immersion at x.
(2) .d/x has rank .mjn/.
(3) There exist charts U , ft i gm
iD1 , f gj D1 around x and V D V1  V2 , ft giD1 ,
j n i m

ftN gaD1 , f gj D1 , fN gbD1 around jj.x/ such that the restriction of the map to
a p j n b q

U and V has the form

t i 7! t i ;  j 7!  j ; tNa ; N b 7! 0:

Proof. The proof mimics the classical one. We sketch it briefly. The equivalence
between (1) and (2) is true by definition, while (3) implies (1) comes from our previous
discussion.
Let .W , fy i g, fj g/ and .Q; fx r g; f s g/ be supercharts near x and jj.x/, respec-
  r m
tively. Possibly relabeling the supercoordinates we can suppose that @ @y.xi / r;iD1
  s n
and @ @ .j / s;j D1 are non-singular at x.
94 5 The local structure of morphisms

Hence if we consider the superchart .Rpjq ; z a ;  b /, we can define the map

˛ W W  Rpjq ! Q
through ˛  .x r / D   .x r / for 1  r  m, ˛  .x r / D z rm C   .x r / for m C 1 
r  m C p, ˛  . s / D   . s / for 1  s  n, and ˛  . s / D  sn C   . s / for
n C 1  s  n C q. It is clear that  D ˛ B iW ,!W Rpjq . It is also clear that .d˛/.x;0/
is a bijection so that, due to Proposition 5.1.1, it is a local diffeomorphism. Thus the
proposition is proved. 

Submersions. Let us proceed in analogy with immersions.

Proposition 5.2.5. Let  W M ! N be a supermanifold morphism, with dim.M / D
m C pjn C q  dim.N / D mjn. The following facts are equivalent:
(1)  W M ! N is a submersion at x of rank mjn.
(2) .d/x has rank .mjn/.
(3) There exist charts U D U1  U2 , ft i gm Na p Nb q
iD1 , ft gaD1 , f gj D1 , f gbD1 around x,
j n

and V , ft i gm
iD1 , f gj D1 around jj.x/ such that the restriction of the map to U
j n

and V has the form

t i 7! t i ;  j 7!  j :
Proof. The proof is completely similar to that for immersion and is left out. 

Constant rank morphisms. Due to the fact that the definition of the constant rank
morphisms involves the Jacobian matrix rather than the differential, the discussion is
more involved.
Proposition 5.2.6. Let M and N be supermanifolds, dim M D pjq, dim N D mjn.
Suppose that  W M ! N is a constant rank morphism of rank ˛jˇ in a neighbor-
hood of x 2 jM j. Then there exist charts U ' U ˛jˇ  U p˛jqˇ with coordinates
.y i ; z j ; r ;  s / and V ' V ˛jˇ V m˛jnˇ with coordinates .y i ; t k ; r ;  l / containing
x and jj.x/, respectively, such that   has the form
y i 7! y i ; r 7! r ;
t k 7! 0;  l 7! 0:
Proof. Since the statement is local, we can work on superdomains. Suppose that we
have a morphism  W U pjq ! V mjn and let .u; /, .v; / be local coordinates in
U pjq and V mjn . By the Chart Theorem we have that  is described by the pullbacks
v i ´   .v i / and  j  ´   . j /. It is not restrictive to suppose that 0 belongs both
to U pjq and to V mjn and that v i .0/ D  j  .0/ D 0. Since J has constant rank,
possibly relabeling the coordinates, we can assume that the matrices
 i ˛  j  ˇ
@v @
r
and
@u i;rD1 @ s j;sD1
 5.2 Immersions, submersions and the constant rank morphisms 95

are non-singular. In order to keep the notation minimal let w and generically denote
the supercoordinates in a superdomain. Hence the morphism

W U pjq ! R˛jˇ  Rp˛jqˇ ;

defined by the pullbacks  .w i ˝ 1/ D v i for 1  i  ˛,  .1 ˝ w j / D u˛Cj for

1  j  p  ˛,  . r ˝ 1/ D  r for 1  r  ˇ, and  .1 ˝ s / D ˇ Cs for
1  s  q ˇ, is a local superdiffeomorphism. An easy check shows indeed that has
invertible differential at x. Hence there exist U1 U pjq and U2 R˛jˇ  Rp˛jqˇ
such that W U1 ! U2 is an isomorphism. The morphism  B 1 is determined by
the pullback

v i 7! w i ˝ 1; 1  i  ˛;
v !
i
7 g .w ˝ 1; 1 ˝ w ;
j 1 j r
˝ 1; 1 ˝ s
/; ˛ C 1  j  m;
 !
r
7 r
˝ 1; 1  r  ˇ;
 s 7!  s .w 1 ˝ 1; 1 ˝ w j ; r
˝ 1; 1 ˝ s
/1; ˇ C 1  s  n:

The Jacobian matrix is hence given by

0 1
I 0 0 0
B @g @g @g @g C
B @ 2C
J B 1 D B @w1 @w2 @ 1
C;
@ 0 0 I 0 A
@ @ @ @
@w1 @w2 @ 1 @ 2

where we have denoted collectively by w1 (resp. w2 , 1 , 2 ) the variables of the form

w i ˝ 1 (resp. 1 ˝ w i , i ˝ 1, 1 ˝ i ).
It is clearly possible to rearrange rows and columns so that the matrix takes the
more tractable form
0 1
I 0 0 0
B 0 I 0 0 C
B C
J B 1 ´ B @g @g @g @g C :
@ @w1 @ 1 @w2 @ 2 A
@ @ @ @
@w1 @ 1 @w2 @ 2

Notice that correspondingly also the usual form of the matrices in the general linear
supergroup will change. In other words, when we write matrices in GL.V .R//, where
we choose for V a homogeneous basis where the even elements do not all come at the
beginning, then we no longer have matrices with even entries in the diagonal blocks
and odd entries in the off diagonal blocks. So these matrices represent elements in
GL.V .R//, but they are not in GLkjl .R/, for a given commutative superalgebra R.
The claim is now that, in order for the matrix J B 1 to have constant rank ˛jˇ, it
 @g @g 
is necessary and sufficient that the submatrix @w @
2 @2
@ is zero. Suppose indeed that
@w2 @2
96 5 The local structure of morphisms

 
I˛;ˇ 0
M ´ is a matrix of constant rank ˛jˇ, where I˛;ˇ is the identity matrix with
A B
 
˛ C ˇ rows and columns. Hence there exists G and G x such that GM x G D I˛;ˇ 0 . In
  0 0
G1 G2
particular there exists G D G3 G4 such that M G has only the first ˛ C ˇ columns
non-zero. In our case we have
 
G1 G2
MG D :
AG1 C BG3 AG2 C BG4

So if G2 D 0, then G4 is invertible and consequently B D 0.

We thus have that g i and  j do not depend on the coordinates w2 , 2 on Rp˛jqˇ .
We can hence simplify the notation and write

g i .w i ˝ 1; r
˝ 1/; ˛ C 1  i  m;
 .w ˝ 1;
j i r
˝ 1/; ˇ C 1  j  n:

Consider now the superdomain U3 V mjn given by

U3 ´ f.v 1 ; : : : ; v m ;  1 ; : : : ;  n / j .v 1 ; : : : ; v ˛ ; 0; : : : ; 0;  1 ; : : : ;  ˇ ; 0; : : : ; 0/ 2 U2 g

and define the morphism

' W U3 ! R˛jˇ  Rm˛jnˇ

via

w i ˝ 1 7! v i ; r
˝ 1 7!  r ;
1 ˝ w j 7! v ˛Cj  g .˛Cj / .v i ;  j /; 1˝ s
7!  ˇ Cs   .ˇ Cs/ .v i ;  j /:
1
It is a simple check to show that ' is invertible and that ' B  B has the required
form. 

We have immediately the following interesting corollaries.

Corollary 5.2.7. Let  W M ! N be a morphism of supermanifolds and let m 2 jM j.

Then the following are equivalent:
(1) In a neighbourhood of m,  splits as a submersion 1 and an immersion 2

1 2
jU W U ! V ! W:

(2)  is a constant rank morphism in a neighbourhood of m.

Corollary 5.2.8. Immersions and submersions are constant rank morphisms.

 5.3 Submanifolds 97

5.3 Submanifolds
In this section we shall use the results on the local structure of morphisms discussed
in the previous section in order to formulate a theory of smooth submanifolds of su-
permanifolds.1 As we stressed from the very beginning, even though many classical
results and constructions carry over to the super setting, the arguments present some ex-
tra subtleties and we invite the reader to go to [54] for some insight on all the problems
that can arise.
As in the classical theory, submanifolds of a given supermanifold M are defined as
pairs .N; j / where N is a supermanifold and j W N ! M is an injective morphism with
some regularity property. We will distinguish two kinds of submanifolds according to
the properties of the morphism j .
Definition 5.3.1. We say that .N; j / is an immersed submanifold if j W N ! M is
an injective immersion, in other words, if jj j W jN j ! jM j is injective and .dj /m is
injective for all m 2 jM j.
As in the ordinary setting, we can strengthen this notion by introducing the notion
of embedding.
Definition 5.3.2. We say that j W N ! M is an embedding if it is an immersion and
if jj j W jN j ! jM j is a homeomorphism onto its image. We say that .N; j / is an
embedded submanifold if j is an embedding. We say that .N; j / is a closed embedded
submanifold if it is an embedded submanifold and jj j.jN j/ is a closed subset of jM j.
Remark 5.3.3. Notice that the morphism j W M z ! M , where M
z is the reduced
manifold associated with M , is a closed embedding.
Closed embedded submanifolds have the following nice characterization in terms
of the properties of the pullback.
Proposition 5.3.4. Let N and M be supermanifolds. A map j W N ! M is a closed
embedding if and only if j  W O.M / ! O.N / is surjective.
Proof. Suppose first that j is a closed embedding and let f 2 O.N /. Denote by
fU˛ g˛2A an open covering of jN j. Since jj j is a homeomorphism onto its image,
there exist open sets fU˛0 g˛2A in jM j such that jj j.U˛ / D U˛0 \ jj j.jN j/. Possibly
passing to a refinement of the open cover, it is not restrictive to suppose each U˛0 to
be a superchart with coordinates x i , y j ,  p , q such that j  .y j / D j  .q / D 0
(see Proposition 5.2.4). In particular the map jU 0 W OM .U˛0 / ! ON .U˛ / is surjective.
˛
Let g˛ be a section in OM .U˛0 / such that jU 0 .g˛ / D f˛ ´ f jU˛ . Since j.jN j/ is
˛
a closed subset of jM j, we can consider the open cover of jM j given by fU˛0 ; ˛ 2
A; jM j n jj j.jN j/g and let fh˛ g denote a partition of unity subordinated to the given
1
To ease the terminology we prefer to use the term “submanifold” instead of the more appropriate, but
cumbersome “subsupermanifold”.
98 5 The local structure of morphisms

cover. PClearly g˛  h˛ belongs to O.M / and, due to the local finitenessPof the cover,
g ´ ˛ g˛  h˛ is a well-defined section in O.M / so that j  .g/ D j  .h˛ /f˛ .

On the other hand it is very easy to check that thePfamily fj .h˛ /g defines a partition
of unity on N subordinated to fU˛ g, hence f D j  .h˛ /f˛ . So we have proved the
surjectivity of j  .
Suppose now that the pullback j  W O.M / ! O.N / is surjective. It is immediate
that both (see notation of Chapter 4, Section 4.5)
jj j W jN j ! jM j; .dj /x W Tx N ! Tjj j.x/ M;
are injective. The fact that jj j is a homeomorphism onto its image is a consequence of
the same result being true for the ordinary setting (see [78], Ch. I). 
Proposition 5.3.4 characterizes closed embeddings among all possible supermani-
fold morphisms  W N ! M in terms of surjectivity of   . As a consequence we have
O.N / ' O.M /=JN where JN ´ ker   . Hence the ideal JN O.M / completely
characterizes the closed embedded submanifold N . The next proposition singles out
some important properties of JN .
Proposition 5.3.5. Let j W N ! M be a closed embedding and let JN D ker.j  /.
Then if m 2 jM j and Jm is the corresponding maximal ideal in O.M /, JN Jm if
and only if m 2 jj j.jN j/. Moreover:
(1) If m 2 jM j is such that JN Jm , then there exist homogeneous f1 ; : : : ; fn in
JN such that Œf1 ; : : : ; Œfn  generate JN;m and .df1 /m ; : : : ; .dfn /m are linearly
independent at m, where JN;m denotes the ideal generated by the image of JN
in the stalk OM;m .
(2) If ffi gi2N is a family in JN suchPthat any compact subset of M intersects only
a finite number of supp fi , then i fi belongs to JN .
Proof. We give a sketch leaving the details to the reader. Since j is a closed embedding,
due to Proposition 5.3.4, O.N / ' O.M /=JN so that there is a bijective correspondence
between maximal ideals in O.N / and maximal ideals in O.M / containing JN . We
now consider .1/. Let x 2 N . Since j is an immersion, there exist supercharts
.U; x i ;  j / 3 x and .U  W; x i ; y r ;  j ; s / 3 jj j.x/ such that j is expressed as an
injection, where the coordinates y r and s are sent to zero. Let U 0 W 0 3 x be an open
subset of U  W and consider a section g such that gjU 0 W 0 D 1 and gjU W c D 0.
The global section g allows us to extend the local coordinates to obtain global sections;
in fact the global sections fy r g; s gg have the required properties to be the fi ’s.
We now go about the proof of (2). We leave it to the reader to check that if
ffi g are sections in O.M / such P that fsupp fi g is a locally finite covering, then the
possibly infinite sum f D i fi is a well-defined element in O.M /. Now notice

that JN .U / ´ ker jjU j
is an ideal sheaf, hence to check if f 2 JN D JN .jN j/ it
is enough to check that f jUi 2 JN .Ui / for an open cover fUi g of jN j. Since any
point has a compact neighbourhood and since the hypothesis in (2) holds, we obtain
the result. 
 5.3 Submanifolds 99

Definition 5.3.6. Let I be an ideal in O.M /. I is called a regular ideal if the following
holds.
(1) If m 2 jM j is such that I Jm , then there exist homogeneous f1 ; : : : ; fn
in I such that Œf1 ; : : : ; Œfn  generate Im and .df1 /m ; : : : ; .dfn /m are linearly
independent at m, where Im denotes the ideal generated by the image of I in the
stalk OM;m .
(2) If ffi gi2N is a family in I such
Pthat any compact subset of M intersects only a
finite number of supp fi , then i fi belongs to I .
Remark 5.3.7. If the sum ffi gi2N is finite, the second condition is trivial since it is
clear that a finite sum of elements in I still belongs to the ideal I . When the sum
is infinite, the sum is still well defined since at any point we are summing a finite
number of fi ’s. However the fi ’s that we are summing can vary from point to point
and consequently in general we cannot assume that their sum still lies in I .
The next proposition says that each regular ideal I in O.M / is of the form JN for
a closed embedded subsupermanifold N . This is a converse to Proposition 5.3.5.
Proposition 5.3.8. Let M be a supermanifold and suppose that J is a regular ideal in
O.M /. Then there exists a unique closed embedded supermanifold .N; j / such that J
is the associated regular ideal, i.e., O.N / D O.M /=J .
Proof. Define the set
jN j ´ fx 2 jM j j J Jx g:
T
Clearly jN j D fjf j1 .0/ j f 2 J g so that jN j is closed in jM j. Since J is a regular
ideal, for each x 2 jN j there exist homogeneous sections f 1 ; : : : ; f p ;  1 ; : : : ;  q in
J , depending on x, of even and odd parity, respectively, such that f.df i /x , .d j /x g
are linearly independent and Œf i , Œ j  generate the space of germs Jx . Because of the
inverse function theorem, there exist sections y r , s in O.M / such that ff i ; y r ;  j ; s g
is a coordinate system in a neighborhood U x of x in jM j.
Define now U 0x ´ U x \ jN j and

O.U 0x / ´ C 1 .y 1 ; : : : ; y n / ˝ ƒ.1 ; : : : ; m /:

Then U 0x is a subset of jM j of the form

U 0x ´ fz 2 jM j j f i .z/ D 0g:

We now show that jN j can be endowed with a supermanifold structure. Clearly we

have the submersion
˛x W O.U 0x / ,! OM .U x /
given by y r 7! y r and s 7! s , and there is also the immersion

ˇx W OM .U x / ! O.U 0x /; f i ;  j 7! 0; y r 7! y r ; s 7! s :
100 5 The local structure of morphisms

We can thus define transition functions N xy between the various O.U x0 / through the
diagram
N
O.U 0x / _ _ _ _/ O.U 0y /
xy

O
˛x ˇy

OM .U x /
xy
/ OM .U y /.

It is evident that N xx D id and it is easy to prove that they obey the other glueing
conditions as specified in Chapter 2, Section 2.2. We call N the supermanifold obtained
in this way. It is clear that it is a closed embedded submanifold. 

We now introduce the concept of quasi-coherent sheaves on supermanifolds. As

we shall see, this concept is strictly related to closed embedded submanifolds, and
this terminology is widely used in the literature. We warn the reader that the word
“quasi-coherent” is used to stress the analogy with quasi-coherent sheaves of ideals
in the algebraic setting, but that this analogy has differences, which should not be
overlooked.

Definition 5.3.9. Let M D .jM j; OM / be a supermanifold and let J an ideal sheaf.

We say that J is a quasi-coherent sheaf of ideals if for each point m 2 jM j, there exists
an open neighbourhood U such that J.U / is a regular ideal.

Proposition 5.3.10. Let M D .jM j; OM / be a supermanifold. There is a one-to-one

correspondence between the following sets:
(1) The set of closed embedded submanifolds of M .
(2) The set of regular ideals of O.M /.
(3) The set of quasi-coherent sheaves on M .

Proof. The correspondence between (1) and (2) is given by Propositions 5.3.8 and
5.3.5. As for (2) and (3), it is clear that if J is a quasi-coherent ideal sheaf, then its
global sections define a regular ideal. Vice versa if I is a regular ideal in O.M /, then for
each point m 2 jM j there exists an open set U containing x and homogeneous sections
.y i ; s / in OM .U / that generate a regular ideal I.U / in OM .U /. These sections are
those computed during the proof of Proposition 5.3.5. The assignment U 7! I.U / is
a sheaf of ideals on jM j. 

The next concept we want to introduce is transversality. The main way to define a
submanifold of a given manifold is through equations. We want to give a criterion to
establish when we can give a supermanifold structure to the set of points which are the
zeros of a certain set of equations.

Definition 5.3.11. Let N1 , N2 and M be supermanifolds. Suppose that morphisms

1 W N1 ! M and 2 W N2 ! M are given such that, if m D j1 j.n1 / D j2 j.n2 /,
 5.3 Submanifolds 101

then
Tm M D .d1 /n1 Tn1 N1 C .d2 /n2 Tn2 N2 : (5.1)
The maps 1 and 2 are then said to be transversal at m.
Notice in particular that if m is not in j1 j.jN1 j/ or not in j2 j.jN2 j/, then 1 and
2 are automatically transversal at m. Notice also that the sum in equation (5.1) needs
not to be direct.
Proposition 5.3.12. Let  W L ! M be a supermanifold morphism and let .N; j / be
a closed embedded submanifold of M , O.N / D O.M /=JN . Suppose that  and j
are transversal and denote by J the ideal in O.L/ generated by   .JN /. Then J is a
regular ideal corresponding to a closed submanifold N 0 L.
Proof. Let us first notice that x 2 jLj is such that J Jx if and only if jj.x/ 2 jN j.
Indeed   .f /.x/ D 0 for all f 2 JN if and only if f .jj.x// D 0 for all f 2 JN . If
x 2 jN j then there exists a superchart U; x i ; xN j ;  r ; N q such that N \ U is determined
by xN j D N q D 0. By using an appropriate global section, we can suppose that xN j ; N q
are defined on the whole M . We claim that   .xN j /;   .N q / satisfy the properties
of .1/ of Proposition 5.3.5. They are clearly in J and they also generate Jx . It
remains only to prove that their differentials are linearly independent, but this is an
easy consequence
 @ of the transversality
P condition. Indeed let Xi 2 Tx L be such that
 i
.d/x Xi D @xN i j j.x/ . If ci .d .xN // D 0 for some non-null sequence of numbers
P P P  @
.ci / then hXp ; ci .d  .xN i //i D ci h .Xp /; xN i i D ci h @xN p j j.x/ ; x i i D 0,
which is clearly impossible. 
Proposition 5.3.13. Let  W L ! M be a supermanifold morphism and suppose that
m 2 jM j. Suppose that for each x in jj1 .m/ there exists a neighborhood where 
is a constant rank morphism.
(1) If J denotes the ideal in O.L/ generated by   .Jm /, then J distinguishes a
closed supersubmanifold L0 of L (Jm the ideal in O.M / corresponding to the point m).
(2) If .L0 ; j / denotes the closed embedded submanifold distinguished by J, then

Tx L0 ' ker.d/x

for each x 2 jL0 j D jj1 .m/.

Proof. (1) Suppose that J Jx for some x 2 jLj. This means that, for all f 2 J,
f .x/ D 0 and since f D   .g/, with g 2 Jm , g.jj.x// D 0, in other words x 2
jj1 .m/. Since  has constant rank, for each x 2 jj1 .m/ we can find coordinates
x 1 ; : : : ; x r1 , y 1 ; : : : ; y r2 , 1 ; : : : ; s1 , 1 ; : : : ; s2 around m such that   .x i /,   . j /
generate Jx . The verification of .2/ in Proposition 5.3.5 is an easy check. We thus
have a closed embedded submanifold .L0 ; j /.
(2) Suppose now that v 2 Tx L0 is given, i.e., a derivation O.L0 / ! R. Since, due
to Proposition 5.3.4, O.L0 / ' O.L/=I , there is a bijective correspondence between
102 5 The local structure of morphisms

Tx L0 and derivations v 2 Tx L such that ker v I . Since I is the ideal generated

by   .Jm /, a simple check shows that ker v I if and only if .d/x v D 0, thus
proving (2). 
Example 5.3.14. Consider the morphism T W GLmjn .T / ! R1j0 .T /, .X / D
Ber.X/, for T a supermanifold. This natural transformation between the functor of
points of these two supermanifolds corresponds to the morphism of the superalgebras
of global sections:   W O.R1j0 / ! O.GLmjn /,   .t / D Ber, where t is the global
canonical coordinate in R1j0 and Ber is the Berezinian function, that is, if xij and kl
are the usual global canonical coordinates in O.GLmjn /, then

X i; j D 1; : : : ; m;
Ber D det.xkl /1 det.xij  ik x
kl
lj /;
k; l D m C 1; : : : ; m C n;
k;l

and x kl denotes the element of the inverse of the matrix .xkl /. One can readily check
that this is a submersion, hence a constant rank morphism at all points of jj1 .1/. By
Proposition 5.3.13 we have that jj1 .1/ has a supermanifold structure and one can
readily check that this is SLmjn , the supermanifold whose T -points are the matrices in
GLmjn .T / with Berezinian equal to 1.
The differential at the identity .d/id can be calculated and it is the morphism

.d/id W Tid .GLmjn / Š Mmjn ! T1 R1j0 Š R; M 7! str.M /;

A B
where str denotes the supertrace of a matrix, i.e., str C D D tr.A/  tr.D/. Hence
the tangent space to SLmjn at the identity consists of those matrices in Mmjn , the real
super vector space detailed in Example 3.1.5 with supertrace zero. We shall see in the
next chapters that this super vector space has the extra structure of a super Lie algebra.
Using the method described in the previous example, one can easily calculate the
tangent space at the identity for other classical supergroups that we shall define in the
next chapter. However we shall give later a better tool to handle such calculations,
namely the stabilizer theorem.

5.4 References
With respect to the classical material we refer the reader to [47], [1], [78] for an
exhaustive treatment. Almost all the supergeometry material presented in this chapter
is a reworking of [53]. More recent treatments of the same material can be found in
[56], [76].
6
The Frobenius theorem

In ordinary geometry the Frobenius theorem is extremely important: it provides a

constructive and effective way to build a submanifold of a given manifold starting from
a subbundle of its tangent bundle, verifying some natural conditions. In this chapter
we want to prove the extension to supergeometry of the Frobenius theorem, both in its
local and global versions.
Though the results are stated in the super context almost unchanged, with respect
to the ordinary setting, the arguments turn out to be more delicate and in particular, as
we shall see, it is necessary to study separately the behaviour of odd vector fields and
their associated distributions.

6.1 The local super Frobenius theorem

We want a mechanism by which we can construct locally a submanifold of a given
supermanifold M .
Let M D .jM j; OM / be a supermanifold with tangent bundle VecM .
Definition 6.1.1. A distribution on M is an OM -submodule D of VecM which is
locally a direct factor, namely, given x 2 jM j, there exists an open neighbourhood U
of x and another subsheaf D 0  VecM so that
VecM;y D Dy ˚ D 0y
for all y 2 U .
We first want to show that any distribution D is locally free, in other words, Dx is
a free Ox -module. Before this, we need a lemma, whose proof relies heavily on the
super version of Nakayama’s lemma, detailed in Appendix B.
Lemma 6.1.2. Let x 2 jM j and let fXi ; ‚j g be vector fields defined in a neighbour-
hood of x and such that their associated tangent vectors are linearly independent. Then
Xi;x , ‚j;x 2 VecM;x are linearly independent.
Proof. In Lemma B.3.3 take E D spanOM;x fXi;x ; ‚j;x g and AN D VecM;x . Then the
result follows. 
Lemma 6.1.3. Let D be as above. Then Dx is a free module for all x 2 jM j. Moreover,
if M is connected, then the dimension of the vector space Dx =mx Dx is independent
from x, where mx is the maximal ideal in the local ring OM;x .
104 6 The Frobenius theorem

Proof. Let x 2 jM j. We set ourselves up to use the previous result, namely we will find
vector fields fXi ; ‚j g which generate D in an open neighborhood U of x so that the
vectors corresponding to evaluating fXi ; ‚j g at x are a basis for Tx .M /, the tangent
space of M at x. Lemma 6.1.2 then tells us that fXi ; ‚j g are in fact OM;x -linearly
independent in VecM;x .
Let fXi ; ‚j g be vector fields on Dx defined on an open neighborhood U of x so
that their tangent vectors at x are a basis for Dx =mx Dx Tx .M /. Because D is
a distribution, it is locally free, and there exists D VecM .U / so that VecM;y D
Dy ˚ Dy for all y 2 U .
Now let fYk ; „l g 2 Dx defined on U be such that their tangent vectors at x are a
basis for Dx =mx Dx Tx .M /. Since Tx .M / D Dx =mx Dx ˚Dx =mx Dx , it follows
that fXi ; Yk ; ‚j ; „l g evaluated at x is a basis for Tx .M /. In fact, this is true for all
points y 2 U by construction, hence by Lemma 6.1.2 they are linearly independent.
To see that fXi ; ‚j g generate D, let Dy0 D hfXi ; ‚j gi be the submodule of VecM;y
generated by these vector fields. Similarly let Dy0 D hfYk ; „l gi. Then Dy0 Dy and
Dy0 Dy , but we also know that Dy0 ˚ Dy0 D VecM;y D Dy ˚ Dy , hence the
generation.
This proves also that dim.Dy =my Dy / is invariant for y 2 U , if M is connected.

The next definitions are crucial for the statement of the local Frobenius theorem.
Definition 6.1.4. Let M D .jM j; OM / be a connected supermanifold, that is, jM j is
connected, and let D be a distribution on M . We define the rank of a distribution as
rank.D/ ´ dim.Dx =mx Dx /:
The previous lemma ensures the rank is well defined (see also Appendix B, Sec-
tion B.3).
Definition 6.1.5. We say that a distribution D is involutive if it is stable under the
bracket on VecM , i.e., for vector fields X , Y in D the bracket ŒX; Y  is also a vector
field in D.
Definition 6.1.6. We say that a distribution D is integrable if, for any x 2 jM j, there
exists (locally) a submanifold N of M whose tangent bundle can be identified locally
with D, that is,
VecN jU D DjU ; x 2 U;
for a suitable neighbourhood U of x.
It is clear that an integral distribution is involutive; the Frobenius theorem in its
local and global versions establishes a converse for this fact.
Lemma 6.1.3 and the definition of rank of a distribution allow us to make some
crucial change of coordinates calculations at the coordinate chart level to shape an
involutive distribution. We now prove a series of lemmas which will demonstrate, by
construction, the local Frobenius theorem on supermanifolds.
 6.1 The local super Frobenius theorem 105

Remark 6.1.7. The following lemmas pertain to the local Frobenius theorem and are
local results, thus it suffices to consider the case M D Rpjq in a coordinate neighbor-
hood of the origin, i.e., U pjq .
Next is an explicit local calculation of the previous lemma.
Lemma 6.1.8. Let D be an involutive distribution. Then there exist linearly indepen-
dent supercommuting vector fields which span D.
Proof. Let fX1 ; : : : ; Xr ; 1 ; : : : ; s g be a basis for D, in other words, fX1;x ; : : : ; Xr;x ;
1;x ; : : : ; s;x g form a basis for the free OM;x -module Dx . Let .t;  / D .t 1 ; : : : ; t p ,
 1 ; : : : ;  q / be a local set of coordinates. Then we can express the vector fields
X @ X @ X @ X @
Xj D aj i i
C ˛j l l ; k D ˇki i
C bkl l : (6.1)
@t @ @t @
i l i l

The coefficients form an rjs  pjq matrix T ,

 
a ˛
T D ;
ˇ b

with entries in OM .U / (for a suitable domain U ) and whose rows are the vector fields
X1 ; : : : ; Xr , 1 ; : : : ; s generating the distribution D. T has rank rjs since the fXi ; j g
are linearly independent. This is to say that the submatrix .a/ has rank r and rank.b/
= s. Then by renumerating coordinates .t;  /, we may assume that

T D .T0 j /;

where T0 is an invertible rjs  rjs matrix. Multiplying T by any invertible matrix on

the left does not change the row space of T (i.e., the distribution D), so we can multiply
by T01 and assume that  
Ir 0
T D ;
0 Is
which amounts to saying that we may assume that

@ Xp
@ X
q
@
Xj D C aji C ˛j l l ;
@t j @t i @
iDrC1 lDsC1
(6.2)
@ X
q
@ Xp
@
k D k C bkl l C ˇj i i :
@ @ @t
lDsC1 iDrC1

We then claim that ŒXj ; Xk  D 0. By the involutive property of D, we know that

X
r X
s
ŒXj ; Xk  D fi Xi C 'l ; l
iD1 lD1
106 6 The Frobenius theorem

where the fi are even functions and the 'l are odd functions. Then, by (6.2), fi is
the coefficient of the @t@i term in the vector field ŒXj ; Xk . However, again by (6.2), it
is clear that ŒXj ; Xk  has only @t@i terms for i > r, and so we have fi D 0 for all i .
Similarly, ŒXj ; Xk  has only @@ l terms for l > s, hence also 'l D 0 for all l.
The cases ŒXj ; k  D 0 and Œl ; k  D 0 follow by using the same argument above.

We now prove the local super Frobenius theorem in the case of an even distribution
of rank 1j0.
Lemma 6.1.9. Let X be an even vector field. There exist local coordinates so that
@
XD :
@t 1
Proof. Let J be the ideal generated by the odd functions on Rpjq . Then, since X
is even, X maps J to itself. Thus X induces a vector field, and hence an integrable
distribution, on the reduced space Rp . Then we may apply the classical Frobenius
theorem to get a coordinate system where X D @t@1 (mod J). So we may assume that

@ X @ X @
XD 1
C ai i C ˛j j ;
@t @t @
i2 j

where the ai are even, the ˛j are odd, and ai ; ˛j 2 J. That the ai are even
P implies that
ai 2 J 2 . Moreover, we can find an even matrix .bj k / so that ˛j D k bj k  k (mod
J 2 ), and so modulo J 2 we have
@ X @
XD 1
C bj k  k j :
@t @
j;k

Let .t; / 7! .y; / be a change of coordinates where y D t and  D g.t / for

g.t/ D gij .t/Pa suitable invertible matrix of smooth functions depending on t1 only,
that is, j D i gj i .t / i . Then
X  
@ @gj k X @
XD 1
C k 1
C gj l blk ; (6.3)
@y @t @j
j;k l

and we choose g.t / so that it satisfies the matrix differential equation and initial con-
dition
@g
D gb; g.0/ D I:
@t 1
By (6.3) we may then assume that, modulo J 2 ,
@
XD :
@y 1
 6.1 The local super Frobenius theorem 107

Next we claim that if X D @t@1 (mod J k ), then X D @t@1 (mod J kC1 ). Since J is
nilpotent, this claim will imply the result for the 1j0-case.
Again, let .t;  / 7! .y; / be a change of coordinates so that y i D t i C ci and
 D  j C j for ci ; j 2 J k suitably chosen in such a way that they depend on t1
j

only. In the .t;  / coordinate system, let

@ X @ X @
XD 1
C hi i
C 'u u
@t @t u
@
i2

for hi ; 'u 2 J k . In the new coordinate system, this becomes

X  X   
@ @ci @ @l @
XD C hi C C 'l C 1 CY
@y 1 @t 1 @y i @t @l
i l

for some Y D 0 (mod J kC1 ) since 2k  1  k C 1 for k  2. So choose the ci and

l so that they satisfy the differential equations

@ci @l
D hi ; D 'l ;
@t 1 @t 1

and we get that X D @

@y 1
(mod J kC1 ) as we wanted. 

The above Lemma 6.1.9 sets us up to prove the following.

Lemma 6.1.10. Let D be a distribution generated by the set fX1 ; : : : ; Xr g of super-
commuting even vector fields. Then there exist local coordinates .t;  / so that
j 1
X
@ @
Xj D j
C aij i
@t @t
iD1

for some even functions aij .

Proof. We proceed by induction. The case r D 1 is presented above.
We may now assume that we can find coordinates which work for r  1 supercom-
muting vector fields, and we want to prove the lemma for r. Again, assume that there
P 1
are coordinates so that Xj D @t@j C jiD1 aij @t@i for j < r. Then

X
p
@ X q
@
Xr D fi C 'k k
@t i @
iD1 kD1

for some even functions fi and odd functions 'k . The assumption ŒXr ; Xj  D 0 gives
X @ X @ X @ X @
fi i
; Xj C 'k k
; Xj  .Xj fi / i
 .Xj 'k / k D 0:
@t @ @t @
108 6 The Frobenius theorem

We know that Π@t@i ; Xj  is a linear combination of @t@l for l < r, which means that
Xj fi D 0 for all j > r  1. Because the coefficients of the Xj are “upper triangular”
for j  r  1, we see that fi depends only on .t r ; : : : ; t p ;  1 ; : : : ;  q / for i  r. We
also have Π@ @k ; Xj  D 0 for all k, and so Xj 'k D 0 for all j as well. We can then
again conclude that the 'k depend only on .t r ; : : : ; t p ;  1 ; : : : ;  q / as well.
Now we can rewrite Xr as follows:
X
r1
@  X
p
@ Xq
@
Xr D fi i C fl l C 'k k :
@t @t @
iD1 lDr kD1
„ ƒ‚ …
D Xr0

Here Xr0 depends only on .t r ; : : : ;  q /, and so by an application of Lemma 6.1.9 on

Xr0 , we may change the coordinates .t r ; : : : ;  q / so that Xr0 D @t@r and so

@ Xr1
@
Xr D r
C fi0 i
@t @t
iD1

(where the fi0 are the fi above under the change of coordinates prescribed by Lem-
ma 6.1.9). 
In fact, the above lemma proves the local Frobenius theorem in the case when D
is a purely even distribution (i.e., of rank rj0). For the most general case we need
one more lemma, which establishes the local Frobenius theorem in the case of an odd
distribution of rank 0j1.
Lemma 6.1.11. Say  is an odd vector field so that 2 D 0 and that spanfg is a
distribution. Then there exist coordinates so that locally  D @ @1 .
Proof. As we have previously remarked, since we want a local result, it suffices to
assume that  is a vector field on Rpjq near the origin. Let us say .y; / are coordinates
on Rpjq . Then
X @ X @
D ˛i .y; / i C aj .y; / j ;
@y @
i j

where the ˛i are odd, the aj are even, and we may assume that a1 .0/ ¤ 0.
Now consider the map

 W R0j1  Rpjq1 ! Rpjq

given by   W O.Rpjq / ! O.R0j1  Rpjq1 /:

O
y i D t i C ˛i .t; 0; /; 1 D a1 .t; 0; O /; j 2 D  j C aj .t; 0; O /:

Here  is the coordinate on R0j1 and .t 1 ; : : : ; t p ;  2 ; : : : ;  q / are the coordinates on

Rpjq1 , and O denotes the -indices 2; : : : ; q. The ˛.t; 0; O / and a.t; 0; O / are the
 6.1 The local super Frobenius theorem 109

functions ˛i and aj where we substitute t for y, let  1 D 0, and substitute O for

2 ; : : : ; q . We claim that the map  is a diffeomorphism in a neighborhood of the
origin. Indeed, the differential of  at 0 is
0 1
Ip 0
d D @ 0 a1 .0/ 0 A:
0 Iq1

Since it is non-singular, we may think of .t; ; O / as coordinates on Rpjq (near the

origin) with  being a change of coordinates. Under this change of coordinates, we
have
@ X X
D O @ C
˛i .t; 0; /
@
aj .t; 0; O / j :
@ @y i @
i j

The ˛i .t; 0; O / and aj .t; 0; /

O terms must be expressed as functions of .y; /. We shall
do this by following the method we developed in the proof of the Chart Theorem 4.1.11.
Notice that by a simple Taylor series expansion, ˛i .y; / D ˛i .t i C ˛i ; a1 ;  k2 C
O C ˇi for some even function ˇi . Similarly we get aj .y; / D
ak / D ˛i .t i ; 0; /
aj .t; 0; O / C bj for some odd function bj . Thus we can write

@
D  C Z
@
for some even vector field Z. Recall that 1 D  aO 1 where aO 1 is an even invertible
section. Hence  D 1 A from some invertible even section A.
Then we see that under the change of coordinates given by ,
@
 1 A  Z  D ;
„ƒ‚…
@ 0
DZ

where Z  denotes the pullback of Z by  and Z 0 is some even vector field (since both
A and Z are even). Now,

2 D 0 H) . @
@
 1 Z 0 /2 D 0
H) . @
@ 2
/  @ .1 Z 0 /  .1 Z 0 / @
@
C .1 Z 0 /2 D 0
„ƒ‚… @ „ ƒ‚ …
D0 D0
0
H) aO 1 Z C 1 @
@
Z0  1
Z 0 @
@
D0
H) aO 1 Z 0 D 0
H) Z 0 D 0;

so we really have @
@
D  under the change of coordinates. 
Now we can prove the full local Frobenius theorem.
110 6 The Frobenius theorem

Theorem 6.1.12 (Local Frobenius Theorem). Let D be an involutive distribution of

rank rjs. Then there exist local coordinates so that D is spanned by

@ @ @ @
;:::; r ; ;:::; s:
@t 1 @t @ 1 @
Proof. Let fX1 ; : : : ; Xr ; 1 ; : : : ; s g be a basis of vector fields for the distribution D.
By Lemma 6.1.8 we may assume that these basis elements supercommute, so then
D 0 D spanfX1 ; : : : ; Xr g is a subdistribution, and by Lemma 6.1.10 we get that there
exist coordinates so that Xi D @t@i .
We then use the fact that Œ1 ; Xi  D 0 for all i to see that 1 depends only on
coordinates .t rC1 ; : : : ;  q / (as in the proof of Lemma 6.1.10). In fact, this is not
completely accurate. If we express 1 as in (6.1), we see that it is only the ˇik and
blk which depend only on the coordinates .t rC1 ; : : : ;  q /. However, we can always
kill off the first r @=@t i terms by subtracting appropriate linear combinations of the
fX1 D @=@t 1 ; : : : ; Xr D @=@t r g.
Since 21 D 0, by Lemma 6.1.11, we may change only the coordinates .t rC1 ; : : : ;  q /
and express 1 D @ @1 . For 2 we apply the same idea: that Œ2 ; Xi  D 0 and Œ2 ; 1  D
0 again shows that 2 depends only on coordinates .t rC1 ; : : : ; t p ;  2 ; : : : ;  q /, and ap-
plying Lemma 6.1.11 once again gives 2 D @ @2 . The same argument then applies
successively to 3 ; : : : ; s . 

We are now in a position to state and prove the global Frobenius theorem on super-
manifolds.

6.2 The global super Frobenius theorem

Theorem 6.2.1 (Global Frobenius theorem). Let M be a supermanifold, and let D be
an involutive distribution on M . Then given any point of M there is a unique maximal
supermanifold corresponding to D which contains that point; in other words, D is
integrable.

Proof. Let D D spanfX1 ; : : : ; Xr ; 1 ; : : : ; s g as in the previous section (again the

Xi are even and the j are odd). Let D0 D spanfX1 ; : : : ; Xr g; this subdistribution
maps odd sections to odd sections, and so descends to an integral distribution D z0 on
Mz . Let x 2 jM j. Then by the classical global Frobenius Theorem, there is a unique
maximal integral manifold M zx  M z of Dz0 containing x. We want to build a sheaf of
commutative superalgebras on M zx .
By the local Frobenius theorem, given any point y 2 jM j, there exists an open
coordinate neighborhood around y, Uy  jM j, so that Uy is characterized by coordi-
nates .t; z; ; / (i.e., OM .Uy / D C 1 .t; z/Œ; ), where D is given by the TM -span
 6.3 References 111

˚@ 
of @t
; @@ . Now let U  jM j and define the following presheaf  on jM j:

z x \ U there exists
.U / D hff 2 OM .U / j for all y 2 M
Vy  U such that f jVy 2 C 1 .z/Œgi:

We claim that  is a subsheaf of OM . Again let U  jM j be an open subset and

let fU˛ g be an open covering of U so that for a family of s˛ 2 .U˛ / we have
s˛ jU˛ \Uˇ D sˇ jU˛ \Uˇ . We know that there exists a unique s 2 OM .U / so that
sjU˛ D s˛ . Let y 2 Mx \ U . Then y 2 U˛ for some ˛. There exists Vy  U˛ where
sjVy 2 C 1 .z/Œ since sjVy D s˛ jVy . Hence  is a subsheaf of OM . Moreover,  is
an ideal sheaf by construction.
z x , then p D OM;p since we can find some neighborhood
It is clear that if p … M
z
of p, Wp \ Mx D ;, where .Wp / D OM .Wp /. Thus the support of  is M z x , and
we have specified a sheaf of ideals with support M z x which defines a unique closed
submanifold of M (see Section 5.3). By going to coordinate neighborhoods it is clear
that this closed subspace is in fact a closed subsupermanifold which we shall now
call Mx .
The maximality condition is clear. From the classical theory we have that the
reduced space is maximal, and locally we can verify that we have the maximal number
of odd coordinates that D allows. 

6.3 References
The local Frobenius theorem for even distributions was proved by Giachetti–Ricci in
[35], while the version for odd distributions was considered in [50] by Koszul. The
first proof of the local and global super Frobenius theorem we are aware of is the one
contained in [19]. For other more modern treatments we refer the reader to [56], [22],
[76].
7
Super Lie groups

In this chapter we want to take a closer look at supermanifolds with a group structure:
Lie supergroups or super Lie groups. As in the ordinary setting, a super Lie group is
defined as a supermanifold together with the multiplication and inverse morphisms,
that satisfy the usual properties expressed in terms of certain commutative diagrams.
To any Lie supergroup, we can naturally associate a Lie superalgebra, consisting of the
left-invariant vector fields, As in the ordinary setting, the Lie superalgebra is identified
with the tangent superspace to the supergroup at the identity.
We can equivalently approach this theory using the language of the super Harish-
Chandra pairs (SHCP for short), which is historically how it was originally developed
by Kostant and Koszul [49], [50]. The SHCP allows us to recover the supergroup
structure of a Lie supergroup G only by knowing its reduced ordinary Lie group G z and
its associated Lie superalgebra g. In fact an even stronger statement is true: there is
an equivalence of categories between the SHCP and the super Lie groups. As we are
going to see in the next chapter, this equivalence extends to the categories of actions
of the SHCP and the super Lie groups, respectively.

7.1 Super Lie groups

A Lie group is a group object in the category of manifolds. Likewise a super Lie group
is a group object in the category of supermanifolds. More precisely:
Definition 7.1.1. A real super Lie group G (SLG for short) is a real smooth super
manifold G together with three morphisms

W G  G ! G; i W G ! G; e W R0j0 ! G;

called multiplication, inverse, and unit respectively satisfying the commutative dia-
grams
id
GGG /GG G;  GG G;  GG
O www
hidG ;ei GG  hidG ;ii www GG 
w GG w GG
w GG w GG
ww # ww eO #
 G GG
idG
/ G, / G,
id
w ; G GG w ;
GG ww GG ww
GG ww GG ww
  O G i G#
he;id ww  hi;idG i G# ww 
GG / G, GG GG

 7.1 Super Lie groups 113

where eO denotes the composition of the identity e W R0j0 ! G with the unique map
G ! R0j0 . Moreover, h ; i denotes the map .  / B dG , dG W G ! G  G being
the canonical diagonal map.
We may of course interpret all these maps and diagrams in the language of T -points,
which gives us (for any supermanifold T ) morphisms T W G.T /  G.T / ! G.T /,
etc. that obey the same commutative diagrams. In other words, if G is an SLG then
the functor T ! G.T / D Hom.T; G/ takes values in the category of set theoretical
groups. Conversely, Yoneda’s lemma says that if the functor T ! G.T / takes values
in the category of set theoretical groups, then G is actually a super Lie group. This
leads us to an alternative definition of a super Lie group.
Definition 7.1.2. A supermanifold G is a super Lie group if for any supermanifold T ,
G.T / is a group, and for any supermanifold S and morphism T ! S, the corresponding
map G.S / ! G.T / is a group homomorphism.
In other words, a supermanifold G is a super Lie group if and only if its functor of
points T 7! G.T / is a functor into the category of groups.
Remark 7.1.3. Let us notice that to each super Lie group G is associated a Lie group
z It is defined as the underlying manifold G
G. z with the “reduced morphisms”

z G
j jW G z ! G;
z z ! G;
jij W G z z
jej W R0 ! G:

Since the map  7! jj that associates to any supermanifold morphism  W M ! N

the morphism jj W M z ! Nz between the associated reduced manifolds is functorial,
z j j; ji j; jej/ is a Lie group. Notice also that in the ordinary
it is immediate that .G;
setting the reduced morphisms j j, ji j and jej fully determine the Lie group structure
z
on the manifold G.
Example 7.1.4. Let us consider the super Lie group R1j1 through the symbolic language
of T -points. The product morphism W R1j1  R1j1 ! R1j1 is given by

.t;  /  .t 0 ;  0 / D .t C t 0 C   0 ;  C  0 / (7.1)

where the coordinates .t;  / and .t 0 ;  0 / represent two distinct T -points for some su-
permanifold T . It is then clear by the formula (7.1) that the group axioms are satisfied.
We are going to return with more details in Example 7.2.4.
Remark 7.1.5. Notice that the properties required in Definition 7.1.1 translate into
properties of the morphisms on the global sections,  W O.G/ ! O.G  G/,
i  W O.G/ ! O.G/, that make O.G/ “almost” a Hopf superalgebra (see Chapter 1,
Section 1.7). One word of caution: since O.G G/ Š O.G/ Ő O.G/, strictly speaking
O.G/ is not a Hopf superalgebra but a topological Hopf superalgebra, meaning that
since we are taking a completion of the tensor product, we are allowed to consider
infinite sums (for a definition of topological Hopf algebras see [20]). In fact an SLG
114 7 Super Lie groups

can also be equivalently defined as a supermanifold G, where O.G/ has a topological

Hopf superalgebra structure. The product in G.T / is recovered via the comultiplication

in the following way:

x  y D mO.T / B x  y B ; x; y 2 G.T / D Hom.O.G/; O.T //:
Notice that we are making implicit use of Proposition 4.5.4, which tells us that specify-
ing a morphism on O.G/ ˝ O.G/ uniquely determines it on its completion O.G  G/.
We leave to the reader the easy check of the fact that the properties of multiplication
, unit e and inverse i of G will correspond to the (dual) properties of  , e  , i  that
make them respectively comultiplication, counit and antipode in the topological Hopf
superalgebra sense. We are going to come back to this with more details when we talk
about the algebraic setting in Chapter 11, where completion is no longer necessary.
In the language of T -points, as we already remarked, the definition of Lie super-
group is equivalent to saying that a super Lie group as a functor from the category of
supermanifolds to the category of groups is representable. In this vein, let us further
examine the GLmjn example.
Example 7.1.6. As one can readily see, the supermanifold GLmjn described in Exam-
ple 4.6.3 (2) is a Lie supergroup. In fact GLmjn .T / is the group of automorphisms of
the O.T /0 -module Rmjn .T / D .O.T /  Rmjn /0 .
Let us now consider another example of a super Lie group, SLmjn . We define
SLmjn in a way that mimics the classical construction. For each supermanifold T , the
Berezinian gives a morphism from the T -points of GLmjn to the T -points of GL1j0 :
Ber T W GLmjn .T / ! GL1j0 .T /; X 7! Ber.X /:
The super special linear group SLmjn is the kernel of Ber T . Certainly this functor
is group-valued, however it is not immediately clear that it is the functor of points of a
supermanifold, in other words that it is representable. We have already discussed the
representability in the previous chapter, however we are going to address this issue more
generally in the next chapter, when we discuss the Stabilizer Theorem that will give us
the representability of this and the other functor of points of the classical supergroups.

7.2 The super Lie algebra of a super Lie group

For an ordinary Lie group H , we can define a morphism `h , the left multiplication by
h 2 H , as
`h
H ! H; a 7! ha
(for a 2 H ). The differential of this morphism gives
.d `h /a
Ta .H / ! Tha .H /
 7.2 The super Lie algebra of a super Lie group 115

and for a vector field X on H , we say that X is left-invariant if

d `h B X D X B `h :

We want to interpret this in the super category by saying that a left-invariant vector
field on G is invariant with respect to the group law  “on the left”.
Let G be a super Lie group with group law W G  G ! G and let us denote by 1
the identity at the level of sheaf morphisms.
Let VecG denote the vector fields on G (see Chapter 4, Section 4.4).

Definition 7.2.1. A vector field X 2 VecG is said to be left-invariant if

 
.1 ˝ X / B D B X:

Similarly a vector field X 2 VecG is said to be right-invariant if

 
.X ˝ 1/ B D B X:

Since the bracket of left-invariant vector fields is left-invariant, as one can readily
check, the left-invariant vector fields are a super Lie subalgebra of VecG , which we
denote by g.

Definition 7.2.2. Let G be a super Lie group. Then

 
g D fX 2 VecG j .1 ˝ X / D Xg

is the super Lie algebra associated with the super Lie group G, and we write g D Lie.G/
as usual.

The next proposition says that g D Lie.G/ is a finite-dimensional super vector

space canonically identified with the super tangent space at the identity of the super
Lie group G.

Proposition 7.2.3. Let G be a super Lie group.

(1) If Xe denotes a vector in Te G, then

X ´ .1 ˝ Xe /

is a left-invariant vector field. Similarly X R ´ .Xe ˝ 1/ is a right-invariant vector
field.
(2) The map

Te G ! g; Xe ! X ´ .1 ˝ Xe / ; (7.2)
is an isomorphism of super vector spaces. Similarly for right vector fields.
116 7 Super Lie groups

Proof. To prove (1) for the left-invariant vector fields, we need to show that
   
Œ1 ˝ ..1 ˝ Xe / / B D B Œ.1 ˝ Xe / :
  
This is a simple check that uses the coassociativity of , that is, .1 ˝ / D
.  ˝ 1/  . In fact
 
Œ1 ˝ .1 ˝ Xe / B B D .1 ˝ 1 ˝ Xe /Œ.1 ˝  / B 

D .1 ˝ 1 ˝ Xe /Œ.  ˝ 1/ B 

D  B Œ.1 ˝ Xe /  :

As for (2) we notice that the injectivity of the map (7.2) is immediate. Let us
thus focus on the surjectivity. Suppose that X is a left-invariant vector field, i.e.,
.1 ˝ X/  D  X . Apply 1 ˝ e  to this equality to obtain

.1 ˝ e  /.1 ˝ X / 
D .1 ˝ e  / 
X;

from which we get X D .1 ˝ Xe / since .1 ˝ e  / 
D 1. So we are done. 

We can use the previous proposition to endow Te G with the structure of a super
Lie algebra and to identify it with g. From now on we shall use such an identification
freely without an explicit mention.
Example 7.2.4. We want to calculate the left-invariant vector fields on R1j1 by means
of the group law (7.1),

.t;  /  .t 0 ;  0 / D .t C t 0 C   0 ;  C  0 /;

from Example 7.1.4. In terms of W O.R1j1 / ! O.R1j1  R1j1 / the group law reads
 
.t / D t ˝ 1 C 1 ˝ t C  ˝ ; . / D  ˝ 1 C 1 ˝ :

From Proposition 7.2.3, we know ˚ that the Lie

 algebra of left-invariant vector fields can
be extracted from Te G D span @t@ je ; @@ je . We use the identity (7.2) to calculate the
corresponding left-invariant vector fields:
 ˇ  ˇ
1 ˝ @t@ ˇe B  ; 1 ˝ @@ ˇe B  : (7.3)

To get coordinate representations of (7.3), we apply them to coordinates .t;  /:

ˇ ˇ
.1 ˝ @t@ ˇe / B  .t / D .1 ˝ @t@ ˇe /.t ˝ 1 C 1 ˝ t C  ˝  / D 1;
ˇ ˇ
.1 ˝ @t@ ˇe / B  . / D .1 ˝ @t@ ˇe /. ˝ 1 C 1 ˝  / D 0;
ˇ ˇ
.1 ˝ @@ ˇe / B  .t / D .1 ˝ @@ ˇe /.t ˝ 1 C 1 ˝ t C  ˝  / D ;
ˇ ˇ
.1 ˝ @ ˇ / B  . / D .1 ˝ @ ˇ /. ˝ 1 C 1 ˝  / D 1:
@ e @ e
 7.2 The super Lie algebra of a super Lie group 117

Thus the left-invariant vector fields on .R1j1 ; / are

@ @ @
XD ; Y D  C : (7.4)
@t @t @
A quick check using the definition shows that (7.4) are in fact left-invariant. The
Lie superalgebra structure can now be easily computed: Lie.R1j1 / D spanfX; Y g with
brackets
ŒX; X  D 0; ŒX; Y  D 0; ŒY; Y  D 2X:

Proposition 7.2.5. Let G and H be super Lie groups and let  W G ! H be a morphism
of super Lie groups. The map
.d/e W g ! h
is a super Lie algebra homomorphism.

Proof. The only thing to check is that .d/e preserves the super Lie bracket. We
leave this to the reader as an easy exercise, recalling that the bracket has always to be
computed on the left-invariant vector fields. 

Corollary 7.2.6. The even part of the super Lie algebra Lie.G/ canonically identifies
z
with the ordinary Lie algebra Lie.G/.
z ! G and the
Proof. This is immediate considering the canonical inclusion j W G
previous proposition. 
z acts on G in a natural
We end this section showing that the reduced Lie group G
way.

Definition 7.2.7. If G is a super Lie group and M is a supermanifold, we say that G

acts on M if we have a morphism

aW G  M ! M

defined using the functor of points as

aT W G.T /  M.T / ! M.T /; .g; m/ 7! g  m;

such that
(1) eT  m D m,
(2) g1  .g2  m/ D .g1 g2 /  m.
In other words,

O 1M i D 1M ;
a B he; (7.5a)
a B .  1M / D a B .1G  a/; (7.5b)
118 7 Super Lie groups

where 1M W M ! M denotes the identity morphism of a supermanifold M and eO W G !

G is a super morphism defined, in the functor of points notation, as eOT .g/ D eT , with
eT the identity element in the group G.T /.
The morphism a is called an action of G on M . We will return more extensively
to actions in Chapter 8.
Remark 7.2.8. We will encounter also right actions (as opposed to the previous ones
that are called left actions). For them, condition .2/ is replaced by g1  .g2  m/ D
.g2 g1 /  m, while condition (7.5b) is replaced by

a B .  1M / D a B .1G  a/ B .  1M /;

where  W G  G ! G is the flip morphism defined by T .g1 ; g2 / D g2 g1 for gi 2

G.T /.
Since G is a super Lie group, it acts on itself via group multiplication W G  G !
G. Fix an element gN W R0j0 ! G in jGj (recall that a topological point is identified
with an element in G.R0j0 /, see Example 3.4.6 (i)). Moreover a topological point gN
N
can always be viewed naturally as a T -point of G. Define the left translation by g:
N
gid 
lgN W G ' R0j0  G ! G  G 
! G:

This induces an action

z  G ! G;
aW G
z At the level of sections we have
which we call left multiplication by G.

lgN f D .evgN ˝ 1/ B 
.f /; a f D .j ˝ 1/ B 
.f /;

where j denotes the embedding of Gz in G.

Similarly we can define the right translation by g 2 jGj by

rgN f D .1 ˝ evgN / B 
.f /:

z ! G defines a right action.

The corresponding morphism G  G

7.3 The Hopf superalgebra of distributions

In this section we are going to briefly describe Kostant’s original approach to the theory
of supergroups and we are also going to prove that this is essentially equivalent to our
treatment.
As we have seen in Section 4.7, if M is a supermanifold, the superalgebra of
distributions with finite support O.M /ı is a super coalgebra. We can use the group
 7.3 The Hopf superalgebra of distributions 119

structure of G to define a superalgebra structure on O.G/ı , which is compatible with

its coalgebra structure. As we shall see, the presence of an antipode makes O.G/ı a
Hopf superalgebra, not to be confused with the topological Hopf superalgebra O.G/
described above.
Definition 7.3.1. Let Xg , YgN be point supported distributions (resp. supported at the
N on the Lie supergroup G. Define the convolution product of Xg and
points g and g)
YgN by
.Xg YgN /.f / D hXg ˝ YgN ;  .f /i;
where f 2 O.G/ and is the super Lie group multiplication. Given a distribution Xg
we also define
.dlgN /g Xg D Xg B lgN :
Despite the analogy in notation and properties, distributions must not be confused with
vector fields. We shall establish how distributions and vector fields are related later in
this section.
Remark 7.3.2. Notice that the convolution product Xg YgN of the distributions Xg
N respectively, is supported at g  g.
and YgN , supported at g and g, N This is an immediate
consequence of the statement for ordinary convolutions, that can be found for example
in [26], Ch. XVII, n. 11.
The next proposition provides us with a generalization of well-known formulas for
distributions in the ordinary setting.
Proposition 7.3.3. Let Xg be a distribution on a Lie supergroup.
(1) .dlgN /g Xg D evgN Xg .
(2) .drgN /g Xg D Xg evgN .
Proof. We shall prove only (1), (2) being the same.
h.dlgN /g Xg ; f i D hXg ; lgN f i

D hXg ; .evgN ˝ 1/ B .f /i
D .evgN Xg /.f /: 
The above discussion establishes the following proposition.
Proposition 7.3.4. The convolution product defines a superalgebra structure on the
supercoalgebra O.G/ı of distributions with finite support.
Proof. By its very definition is associative and, by Proposition 7.3.3, eve is its identity
element. 
We now come to a closer examination of the algebraic structure of O.G/ı . We know
that O.G/ı is a super-coalgebra from Section 4.7 and a superalgebra with identity eve
from the previous proposition. It is only natural to ask whether the two structures are
compatible.
120 7 Super Lie groups

Proposition 7.3.5. O.G/ı is a super Hopf algebra with comultiplication , counit 

and antipode S given by

h .Xg /; f ˝ gi ´ hXg ; f  gi; h.Xg /; f i ´ hXg ; eve .f /i;

hS.Xg /; f i ´ hXg ; i  .f /i;

where i W G ! G denotes the inverse morphism.

Proof. First we show that the comultiplication and the counit  are superalgebra
morphisms. Let us start with the comultiplication . Consider

h .Xg Yh /; s ˝ ti D hXg ˝ Yh ;  .s  t /i
D hXg ˝ Yh ;  .s/  .t /i
D h Xg Yh ; s ˝ t i:

Consider now the counit . We have

h.Xg Yh /; si D hXg ˝ Yh ;  B eve .s/i

D hXg ˝ Yh ; eve .s/1 ˝ 1i
D eve .s/Xg .1/Yh .1/
D h.Xg / .Yh /; si:

The fact that S has the property of the antipode is a similar check. 
We end this section by pointing out an interesting decomposition of O.G/ı which
parallels a general structure theorem for Hopf algebras (see again [72] for the classical
case).
Let us examine more closely the product law in O.G/ı . Fix a distribution Xg in
O.G/ı . Each such element can be canonically written as the convolution product of a
distribution at the identity e of the super Lie group with the “element” g of jGj:

Xg D evg .evg 1 Xg / D evg .dlg 1 /g Xg : (7.6)

Thus each element in O.G/ı can be written as a product evg Xe , where Xe 2 O.G/ı
denotes the distribution at e given by

Xe ´ .dlg 1 /g Xg :

Notice that here Xe is used in a more general sense than in Section 11.2.4, where it
denoted an element in Te G.
The next proposition characterizes the distributions with support at e.
Proposition 7.3.6. The space of distributions supported at the identity e of the super
Lie group G is isomorphic, as a super Hopf algebra, to the enveloping algebra U.g/
of the super Lie algebra g associated with the supergroup G.
 7.3 The Hopf superalgebra of distributions 121

Proof. It goes exactly as in the classical case, but we shall nevertheless briefly outline
it to make the text self-contained. As we have seen in Chapter 4, Section 4.7, the
finitely supported distributions can be identified with differential operators. Hence a
distribution Xe supported at the identity can be identified with a derivation of type
X ˇ ˇ
@ ˇˇ @ ˇˇ
Xe D cI;J I ˇ ;
@x e @ J ˇe
I;J

where .xi ; j / are local coordinates around the identity e 2 jGj and I and J are multi-
indices. Every distribution Xe described above corresponds to a unique left-invariant
differential operator via the linear map

Xe 7! .1 ˝ Xe / ;
very much in the same way as tangent vectors at the identity correspond to left-invariant
vector fields. The linear map described above is a linear isomorphism; the proof
of this fact resembles very closely what we did in Proposition 7.2.3. Finally a left-
invariant differential operator can be uniquely identified with an element of the universal
enveloping superalgebra of g, a conclusion from dimension considerations and using
the PBW theorem. 
Using decomposition (7.6), we can now define a linear map
z ˝ U.g/;
W O.G/ı ! RG Xg 7! g ˝ .dlg 1 /g Xg ;
where we notice that .dlg 1 /g Xg is in fact a distribution supported at the identity, hence
identified with an element in U.g/ (see previous proposition). RG z denotes the group
algebra of Gz and consists of the formal finite sums of elements in G z with coefficients
in R. By Proposition 7.3.6 and equation (7.6), the above map is a bijection.
All the algebraic structures defined over O.G/ı can thus be transported to corre-
sponding structures over RG z ˝ U.g/, as the next proposition formalizes.
Proposition 7.3.7. RG z ˝ U.g/ is a super Hopf algebra. The coalgebra structure is
z and U.g/, namely
the one induced by the coalgebra structures of RG

zW
RG g 7! g ˝ g; gW Xe 7! Xe ˝ 1 C 1 ˝ Xe for all Xe 2 g;
while the algebra structure is given by
.g ˝ X /.gN ˝ Y / D .g gN ˝ .gN 1 X /Y /; (7.7)
where gN 1 X ´ evgN 1 X evgN .
Proof. For the coalgebra structure it is enough to check on the generators:
.g ˝ X / D .g/ .X /
D .g ˝ g/.1 ˝ X C X ˝ 1/
D .g ˝ 1/ ˝ .g ˝ X / C .g ˝ X / ˝ .g ˝ 1/:
122 7 Super Lie groups

Notice that
1
. .g ˝ X // D .evg 1/ ˝ .evg X / C .evg X / ˝ .evg 1/:

On the other hand,

1
. .g ˝ X // D .Xg /;
where
h .Xg /; f ˝ hi D hXg ; f  hi D Xg .f /h C f Xg .h/:
As for the algebra structure, we have

.g ˝ X /.gN ˝ Y / ´ B Œ 1 .g ˝ X / 1 .gN ˝ Y /
D .Xg YgN /
D .evg X evgN Y /
D .evg gN evgN 1 X evgN Y /
D .evg gN ..gN 1 X / Y //;

where gN 1 X ´ evgN 1 X evgN . Therefore we can rewrite our expression as

.g ˝ X /.gN ˝ Y / D .g gN ˝ .gN 1 X /Y /: 

Notice that we have defined a morphism that to each g 2 jGj associates the super
linear morphism X ! gX. This is of fundamental importance and in the ordinary case
z on U.g/. We are thus led to the following
reduces to the adjoint representation of G
definition.

Definition 7.3.8. Let G be a super Lie group and g the corresponding super Lie algebra.
The morphism
z ! Aut.g/;
Ad W G Ad.g/X ´ .evg ˝ X ˝ evg 1 /.1 ˝ 
/ 
;

with g 2 jGj and X 2 g, is called the adjoint representation of G z on g. As usual

z on U.g/ that is still called adjoint
this representation induces a representation of G
representation.

We now want to see an equivalent formulation of the adjoint representation using

the formalism of the functor of points.

Proposition 7.3.9. For each g 2 jGj let us define a morphism cg W G ! G, cg .x/ D

gxg 1 (recall that any topological point of G can be viewed naturally as a T -point of
G for all T ). Then Ad.g/ D .dcg /e .

Proof. The proof is an instructive application of Yoneda’s lemma. By definition we

have .dcg /e X D X B cg , X 2 g. If x 2 G.T / D Hom.O.G/; O.T //, we have
 7.4 Super Harish-Chandra pairs 123

g  x  g 1 D .g ˝ x ˝ g 1 /.  ˝ 1/  . By Yoneda’s lemma, cg corresponds to the

image of the identity morphism 1 W O.G/ ! O.G/, hence

cg .f / D .g ˝ 1 ˝ g 1 /. 
˝ 1/ 
.f /;

where g W O.G/ ! O.G/ has to be understood as the evaluation at g 2 jGj. Thus

we have proved our claim (notice that by coassociativity .  ˝ 1/  D .1 ˝  /  ).


Proposition 7.3.10. Let the notation be as above. Then d Ad D adjg0 , in other words,

d Ad W g0 ! End.g/; X 7! fY 7! ŒX; Y g:

Proof. This is a direct check. 

Since formula (7.7) resembles the one for semi-direct products, it is customary
z ˝ U.g/, endowed with the Hopf algebra structure just
in the literature to denote RG
described, by
RG z Ì U.g/: (7.8)
This is also called the smash product of RG z with U.g/, but we shall not define smash
products in general, referring the interested reader to [72] for more details.
So far we have associated with each super Lie group G the super Hopf algebra
O.G/ı of finite support distributions over G, and we have seen that it can be written as
z Ì U.g/:
O.G/ı ' RG

The assignment G ! O.G/ı is a faithful embedding from the category of su-

permanifolds to the category of Hopf superalgebras, however, this embedding is not
full. In order to have this, we need to give more structure to O.G/ı , which is why we
introduce the notion of super Harish-Chandra pairs in the next section.

7.4 Super Harish-Chandra pairs

Super Harish-Chandra pairs (SHCP for short) give an equivalent way to approach the
theory of Lie supergroups. A SHCP essentially consists of a pair .G0 ; g/, where G0
is an ordinary Lie group and g a Lie superalgebra such that g0 D Lie.G0 /, together
with some natural compatibility conditions. The name comes from an analogy with
the theory of Harish-Chandra pairs, that is, the pairs consisting of a compact Lie group
K and a Lie algebra g, with a Cartan involution corresponding to the compact form
k D Lie.K/. Harish-Chandra pairs are very important in the theory of representation of
Lie groups and we shall see in the next chapter that SHCP provide an effective method
to study the representations of Lie supergroups.
124 7 Super Lie groups

Definition 7.4.1. Suppose that .G0 ; g/ are respectively a Lie group and a super Lie
algebra such that
(1) g0 ' Lie.G0 /,
(2) G0 acts on g via a representation  such that  .G0 /jg0 D Ad and the differential
of  acts on g as the adjoint representation, that is,

d .X /Y D ŒX; Y :

Then the pair .G0 ; g/ is called a a super Harish-Chandra pair (SHCP). We shall
write .G0 ; g;  / when we want to stress the action  .

A morphism of SHCPs is simply a pair of morphisms D. 0; / preserving

the SHCP structure.

Definition 7.4.2. Let .G0 ; g;  / and .H0 ; h; / be SHCP. A morphism between them
is a pair . 0 ; / such that
(1) 0W G0 ! H0 is a Lie groups homomorphism,
(2) W g ! h is a super Lie algebra homomorphism,
(3) 0 and are compatible in the sense that

jg0 ' .d 0 /e ; . 0 .g// B D B  .g/:

Example 7.4.3. If G is a super Lie group, the pair .G; z g/ given by the reduced Lie
group of G and the super Lie algebra g is a super Harish-Chandra pair with respect to the
z on g as defined in Definition 7.3.8. Moreover, given a morphism
adjoint action of G
 W G ! H of super Lie groups,  determines the morphism of the corresponding
super Harish-Chandra pairs
.jj; .d/e /:

In the next observation we relate the SHCP with the Kostant construction described
in the previous section.

Observation 7.4.4. Given an SHCP pair, we can construct the super Hopf algebra
given by the semidirect product RG0 Ì U.g/ in which G0 is endowed with a Lie group
structure and the product is given by

.g1 ; Xe / .g2 ; Ye / ´ .g1 g2 ; . .g2 /1 Xe /Ye /:

This is a straightforward generalization of (7.8).

Such objects were introduced by Kostant in [49] under the name of Lie–Hopf
superalgebras. The category of Lie–Hopf superalgebras is clearly isomorphic to the
category of SHCP, and given a super Lie group G we can associate to it the Lie–Hopf
superalgebra O.G/ı ' RG z Ì U.g/ in which G z is the associated classical Lie group.
 7.4 Super Harish-Chandra pairs 125

We summarize our previous considerations by saying that we have defined a functor

H W SGrp ! .shcps/; z g; Ad/;

G 7! .G;

from the category of super Lie groups to the category of super Harish-Chandra pairs.
The most important result of this chapter is the following:
Theorem 7.4.5. The category of super Lie groups is equivalent to the category of
super Harish-Chandra pairs.
Roughly speaking this theorem says that each problem in the category of super
Lie groups can be reformulated as an equivalent problem in the language of SHCP.
Before embarking on the proof, let us outline the path that we shall follow. We show
the following.
(i) Given a SHCP .G0 ; g/ there exists a super Lie group G whose associated SHCP
is isomorphic to .G0 ; g/.
(ii) Given a morphism of SHCP . 0 ; / W .G0 ; g/ ! .H0 ; h/ there exists a unique
morphism of the corresponding super Lie groups from which . 0 ; / arises.
iii) Due to points (i) and (ii) we have a functor

K W .shcps/ ! SGrp:

In order to prove the theorem we have to show that K B H ' 1SGrp and H B K '
1.shcps/ . This means that, for each G 2 SGrp and .G0 ; g/ 2 .shcps/, .K B H /.G/ ' G
and .H B K/..G0 ; g// ' .G0 ; g/, and moreover the diagrams
 /G  / .G0 ; g/
.K B H /.G/ .H B K/..G0 ; g//
.KBH /. / .H BK/./  (7.9)
 
  

.K B H /.H / / H, .H B K/..H0 ; h// / .H0 ; h/

commute for each  W G ! H and  W .G0 ; g/ ! .H0 ; h/.

We start with the reconstruction of a super Lie group from an SHCP. Suppose that
an SHCP .G0 ; g;  / is given and notice that:
(1) U.g/ is naturally a left U.g0 /-module.
(2) For each open set U G0 , CG10 .U / is a left U.g0 /-module. In fact (see for
example, [75]) each X 2 g0 acts from the left on smooth functions on G0 as the
left-invariant differential operator1
ˇ
d ˇ
zXL f /.g/ ´
.D f .ge tX /ˇˇ :
dt tD0
1
Notice that, as already remarked, here and in the following we do not mention explicitly the isomorphism
Lie.G0 / ' g0 appearing in the definition of an SHCP.
126 7 Super Lie groups

Hence, for each open subset U G0 , we can define the assignment

U 7! OG .U / ´ HomU.g0 / .U.g/; CG10 .U //;
where the right-hand side is the subset of Hom.U.g/; CG10 .U // consisting of U.g0 /-
linear morphisms. (Notice that, for the moment, G is just a letter and we have not
defined any supergroup structure on G0 .)
Remark 7.4.6. If g D g0 we have
HomU.g0 / .U.g/; CG10 .U // Š CG10 .U /:
In fact a U.g0 /-linear map is uniquely determined by the image of 1 2 U.g/ and
consequently we can uniquely associate to any morphism an element of CG10 .U /.
OG .U / has a natural structure of unital, commutative superalgebra. The multipli-
cation OG .U / ˝ OG .U / ! OG .U / is defined by
f1  f2 ´ mC 1 .G0 / B .f1 ˝ f2 / B U.g/ (7.10)
and the unit can be identified (with a mild abuse of notation) with the counit  of U.g/.
We leave to the reader the simple check that f1  f2 2 OG .U /.
If U and V are open subsets of G0 such that U V , we define the superalgebra
morphism
V;U W OG .V / ! OG .U /; f 7! QV;U B f; (7.11)
where QV;U W C 1 .V / ! C 1 .U / is the usual restriction map. We will often abbreviate
V;U .f / with f jU .

Lemma 7.4.7. OG , together with the restriction maps (7.11), is a sheaf of superalge-
bras.
Proof. The fact that OG is a presheaf is a routine check. As for the glueing property,
let fU˛ g be open sets covering a fixed open set U and f˛ elements in OG .U˛ / such
that f˛ jU˛ \Uˇ D fˇ jU˛ \Uˇ for each ˛, ˇ. We want to define f 2 OG .U /. For
each X 2 U.g/, the f˛ .X / 2 C 1 .U˛ / glue together to give gX 2 C 1 .U /. Define
f .X/ D gX . Then U.g0 /-linearity is immediate. 
Lemma 7.4.8. The antisymmetrizer
1 X
 W ƒ.g1 / ! U.g/; X1 ^    ^ Xp 7! .1/jj X.1/    X.p/ ;
pŠ
2Sp

is a super coalgebra morphism, where Sp denotes the group of permutations of p

elements.
The map
O W U.g0 / ˝ ƒ.g1 / ! U.g/; X ˝ Y 7! X  .Y /;
is an isomorphism of super left U.g0 /-modules.
 7.4 Super Harish-Chandra pairs 127

Proof. Both assertions are consequences of the PBW theorem and we leave their proofs
to the reader. 

Proposition 7.4.9. .G0 ; OG / is a supermanifold that is globally split, i.e., for each
open subset U G0 there is an isomorphism of superalgebras
V V
OG .U / ' Hom. .g1 /; C 1 .U // ' C 1 .U / ˝ .g1 / :

Hence OG carries a natural Z-gradation.

Proof. In view of Lemma 7.4.7, it only remains to prove the local triviality of the sheaf.
For this purpose we define the map
V
U W OG .U / ! Hom. .g1 /; C 1 .U //; f ! f B :

Since  is a supercoalgebra morphism, U is a superalgebra morphism. In fact,

U .f1 f2 / D mBf1 ˝f2 B U.g/ B D mBf1 ˝f2 B. ˝ / U.g/ D U .f1 /U .f2 /:

That U is a superalgebra isomorphism follows at once from U.g0 /-linearity. 

So far we have used only the fact that CG10 .U / is a left U.g0 /-module.
The next proposition uses more heavily the structure of G0 and the representation
. It exhibits explicitly the structure of a super Lie group in terms of the corresponding
SHCP.

Proposition 7.4.10. .G0 ; OG / is a super Lie group with respect to the morphisms
multiplication W G  G ! G, inverse i W G ! G and unit e W k ! G, which are
defined in terms of their pullbacks  , i  , e  as

Œ .f /.X; Y /.g; h/ D Œf ..h1  X /Y /.gh/;
Œi  .f /.X / .g 1 / D Œf .g 1  Xx /.g/; (7.12)

e .f / D Œf .1/.e/;

where X; Y 2 U.g/, g; h 2 G0 , f 2 OG .U /, e is the unit of G0 , g  X ´  .g/X and

Xx denotes the antipode of X in U.g/.

Proof. The proposition will follow if we prove the following statements:


i) is well defined, a superalgebra morphism and associative;
 
ii) e is well defined, a superalgebra morphism and a unit for ;
 
iii) i is well defined, a superalgebra morphism and an antipode for .
The verification of these assertions consists in a quantity of long and tedious calcula-
tions. We hence select the most significant of them, leaving the others to the reader.
Let us begin with the various steps.
128 7 Super Lie groups

i) Comultiplication. In order to show that  is well defined, we have to prove that

if f 2 OG .U /,  .f / belongs to HomU.g0 ˚g0 / .U.g ˚ g/; C 1 .W //, for a suitable
W open in G0  G0 . Due to PBW theorem, it is enough to prove g0 -linearity. Let
Z 2 g0 . Then

.f /.ZX; Y /.g; h/ D f .h1 .ZX /Y /.gh/
D f ..h1  Z/.h1  X /Y /.gh/
DD z L1 Œf ..h1  X /Y /.gh/:
h Z

On the other hand,

ˇ
d ˇ
zZ
Œ.D L
˝ 1/. 
.f /.X; Y //.g; h/ D f ..h1 X /Y /.ge tZ h/
dt tD0
ˇ
d ˇ 1
D dt tD0
f ..h1 X /Y /.ghe t.h Z/ /
D z L1 Œf ..h1  X /Y /.gh/:
D h Z

Similarly for the left entry, one finds


.f /.X; ZY /.g; h/ D f ..h1 X /ZY /.gh/
D f .Z.h1 X /Y C Œh1 X; ZY /.gh/
DD zZ
L
.f ..h1 X /Y //.gh/ C f .Œh1 X; ZY /.gh/
and ˇ
zZ  d ˇ 
Œ.1 ˝ D L
/. .f /.X; Y //.g; h/ D dt jtD0
.f /.X; Y /.g; he tZ /
ˇ
d ˇ
D dt tD0
f ...he tZ /1 X /Y /.ghe tZ /
D ŒDzZ
L
f ..h1 X /Y /.gh/ C f .Œ.h1 X /; ZY /.gh/
where
ˇ
d ˇ
ˇ
d ˇ
dt tD0
..he tZ /1 /X D dt tD0
 .e tZ / .h1 /X D d .Z/h1 X D Œh1 X; Z:

We leave it to the reader to check that  is a superalgebra morphism.

We now show that  is associative. Indeed,
  
.. ˝ 1/ B /.f /.X; Y; Z/.g; h; k/ D .f /..h1 X /Y; Z/.gh; k/
D f Œk 1 ..h1 X /Y /Z.ghk/
D f ..k 1 h1 X /.k 1 Y /Z/.ghk/
and
  
..1 ˝ /B /.f /.X; Y; Z/.g; h; k/ D .f /.X; .k 1 Y /Z/.g; hk/
D f ..k 1 h1 X /.k 1 Y /Z/.ghk/:

ii) Counit. We need to check the counit property, that is,

.e  ˝ 1/ 
.f / D .1 ˝ e  / 
.f / D f:
 7.4 Super Harish-Chandra pairs 129

This is completely straightforward and is left to the reader.

iii) Antipode. First we show that it is well defined. Again due to the PBW theorem
we can take Z 2 h0 . Then

i  .f /.ZX /.g 1 / D f .g 1 .ZX //.g/

x Xx C ŒXx ; Z//.g/
D f .g 1 .Z x
x 1 Xx / C g 1 ŒX; Z/.g/
D f ..g 1 Z/.g
z L1 x f .g 1 Xx /.g/ C f .g 1 ŒX; Z/.g/
D ŒD g Z
and
ˇ
d ˇ
ˇ
d ˇ x
dt tD0
i  .f /.X /.g 1 e tZ / D dt jtD0
i  .f /.X /.e t Z g/1
ˇ
d ˇ x
D dt tD0
f ..g 1 e tZ /  Xx /.e t Z g/
D z L1 x f .g 1 Xx /.g/ C f .g 1 ŒX; Z/.g/:
D g Z

It is a superalgebra morphism since

x
i  .f1  f2 /.X /.g 1 / D .f1  f2 /.g 1 X/.g/
D Œ.f1 ˝ f2 / .g 1 Xx /.g; g/
and
.i  .f1 /  i  .f2 //.X /.g 1 / D Œ.i  .f1 / ˝ i  .f2 // X .g 1 ; g 1 /
D Œ.f1 ˝ f2 /. .g 1 / ˝  .g 1 //
 .SU.g0 / ˝ SU.g0 / / .X /.g; g/
D Œ.f1 ˝ f2 / .g 1 Xx /.g; g/:

We now show that i  is a right antipode.

Œ.m B .1 ˝ i  / B 
/.f /.X /.g/ D Œ  .f /.1 ˝  .g//.1 ˝ SU.g0 / / .X /.g; g 1 /
D Œf .m. .g/ ˝  .g//.1 ˝ SU.g0 / / .X //.e/
D Œf . .g/m..1 ˝ SU.g0 / / .X //.e/
D Œf  .g/.X /.e/:

A similar computation shows that i  is also a left antipode. 

Let us now collect a glossary of some frequently used operations in our real-
ization, completing those given in equations (7.10) and (7.12). Notice that since
.1/jX j.jf jCjY j/ f .YX / D .1/jXj f .YX /, it is possible to slightly simplify the form
of some expressions and this is a consequence of the trivial fact that f .Z/ D 0 unless
jf j D jZj.
Lemma 7.4.11. Let the notation be as above, and let f 2 OG .U /, g; h 2 jGj,
X; Y 2 U.g/. Then we have the following operations:
130 7 Super Lie groups

(1) Evaluation map: jf j D f .1/, evg .f / D f .1/.g/ D jf j.g/.

(2) Left translation: Œ`h .f /.X / D j`h j.f .X //.
(3) Right translation: Œrh .f / .X / D jrh j.f .h1  X //.
(4) Left invariant vector fields: .DXL f /.Y / D .1/jXj f .YX /.
(5) Right invariant vector fields: Œ.DXR f /.Y /.g/ D .1/jXjjf j f ..g 1  X /Y /.g/.

Proof. (1) is clear from the definitions.

As for (2) we have

Œ`h .f / .X /.g/ D Œ.evh ˝ 1/ B  .f /.X /.g/

X
D Œ.evh ˝ 1/ .f .1/ ˝ f .2/ /.X /.g/
X
D f .1/ .1/.h/f .2/ .X /.g/ D  .f /.1; X /.h; g/
D f .X /.hg/ D j`h j.f .X //:

(3) is very similar; in fact we have

Œrh .f / .X / D Œ.1 ˝ evh / B  .f /.X /.g/

X
D Œ.1 ˝ evh / .f .1/ ˝ f .2/ /.X /.g/
X
D f .1/ .X /.g/f .2/ .1/.h/ D  .f /.X; 1/.g; h/
D f .h1 X /.gh/ D jrh j.f .h1  X //:

(4) We first check that the vector field .DXL f /.Y / ´ .1/jXj f .YX / is left-
invariant, that is, .1 ˝ DXL /  D  DXL :


Œ.1 ˝ DXL / .f /.Y; Z/.g; h/ D .1/jXj 
.f /.Y; ZX /.g; h/
jXj
D .1/ f .h1 Y ZX /.gh/
and

ΠDXL f .Y; Z/.g; h/ D DXL .f /.h1 Y Z/.gh/
D .1/jXj f .h1 Y ZX /.gh/:

Hence we have a well-defined linear map from the Lie superalgebra g and the left-
invariant vector fields X 7! DXL . In order to show that it is an isomorphism it is
enough to prove injectivity for dimension considerations. If X ¤ Y are two elements
in g, with DXL .f / D DYL .f / for all f , then f .X / D f .Y / for all X and Y , reaching
a contradiction.
(5) For right-invariant vector fields the arguments are very similar and the proofs
are left to the reader. 
 7.4 Super Harish-Chandra pairs 131

Notice that we have introduced the notation DXL , DXR for the actions of left and right-
invariant vector fields, which is different from the one used in the previous sections.
With this approach it is very natural to recover a super Lie group morphism from
an SHCP one. Suppose that . 0 ; / is a morphism from .G0 ; g/ to .H0 ; h/, and that
f 2 OH .U /. Define  .f / via the diagram

U.g/ / U.h/

 .f /  f


CG0 . 0 .U // o
1 1

CH10 .U /.
0

It is not difficult to prove that this defines a super Lie group morphism with associated
SHCP morphism . 0 ; /. Indeed we have the following proposition.
Proposition 7.4.12. The map

W HomU.h0 / .U.h/; C 1 .H0 // ! HomU.g0 / .U.g/; C 1 .G0 //;
f 7!  .f / ´ 0 B f B ;

defines a morphism of super Lie groups whose reduced morphism is 0 and whose
differential at the identity is .
Proof. It is immediate that  is well defined. We check that the above defined map is
a map of superalgebras and that it is a super Lie group morphism. It is a superalgebra
morphism since, by Remark 7.4.6,

.f1  f2 /.X /.g/ ´ ŒmC 1 .G0 / .f1 ˝ f2 / .X /. 0 .g//
D Œ.f1 ˝ f2 /. ˝ / .X /. 0 .g/; 0 .g//
D Π .f1 /   .f2 /.X /.g/:

It is a super Lie group morphism; indeed compare

  
. .f //.X; Y /.g; h/ D Π.f /..h1 X /Y /.gh/
1
D Œf . .. 0 .h/ X // .Y //. 0 .gh//
with
   
Œ. ˝ / .f /.X; Y /.g; h/ D Œ .f /. .X /; .Y //. 0 .g/; 0 .h//
1
D Œf . . 0 .h /X / .Y //. 0 .gh//

and taking into account the obvious properties of 0 and .

It is also well behaved with respect to the inverse:

.i  .f //.X /.g 1 / D i  .f /. .X //. 0 .g//
1
D f. 0 .g/ .X //. 0 .g//
132 7 Super Lie groups

and
i . 
.f //.X /.g 1 / D f . .g 1 Xx //. 0 .g//

D f. 0 .g/
1 x
.X //. 0 .g//:

We now compute the differential .d /e at the identity and show that .d /e D .

Let  be an element of the Lie superalgebra h D Te H . Then, using various properties
of the previous lemma, we obtain

.d /e ./.f / D ΠB 0 .f /

DB 0 Bf B
D eve B DL B 0 B f B
D .DL B 0 B f B /.1/.e/
D .1/jj . 
0 Bf B /./.e/
jj
D .1/ Œf . .//. 0 .e//
jj
D .1/ Œf . .//.e/
D .1/jjCj ./j
ŒD L ./ f .1/.e/
D ./.f /: 

We can finally end the proof of Theorem 7.4.5, proving the following proposition.
Proposition 7.4.13. Let the notation be as above. Define the functors

H W SGrp ! .shcps/; G ! .G; z g; Ad/;  ! .jj; .d/e /;

K W .shcps/ ! SGrp; x D .G0 ; HomU.g / .U.g/; CG1 //;
.G0 ; g;  / ! G 0 0

! f 7! 0 f  ;

where G and .G0 ; g;  / are objects and , are morphisms of the corresponding
categories ( for the notation relative to see Proposition 7.4.12).
Then H and K define an equivalence between the categories of super Lie groups
and super Harish-Chandra pairs.
x moreover
Proof. We first check that .H B K/.G0 ; g/ ' .G0 ; g/. Clearly G0 D jGj,
the two equalities

ŒDXL DYL  .1/jXjjY j DYL DXL  D DŒX;Y

L
;

rg DXL rg1 D DgX

L

for each X; Y 2 g and g 2 G0 tell us that the Lie superalgebras Lie.G/ and g are the
same and so is the action of G0 on g.
We now turn to check that G x D .K B H /.G/ ' G. In order to do this we shall
x
build a morphism  W G ! G and prove that it is an SLG isomorphism.
 7.4 Super Harish-Chandra pairs 133

• Let  be defined by the pullback

x D HomU.g / .U.g/; C 1 .G0 //;
 W O.G/ ! O.G/ 0

s 7! .Ns W X ! .1/jXj j.DXL s/j/:

Here DXL denotes the left-invariant differential operator on G associated with
X 2 U.g/, that is, DXL D .1 ˝ X /  .
• The definition is well posed. Indeed,
zXL j.DXL s/j
sN W X0 X 7! .1/jXj j.DXL0 DXL s/j D .1/jXj D 0

for each X0 2 g0 and X 2 U.g/, so sN is actually U.g0 /-linear, and moreover,

 .s/ .t / D  .st /
for each s; t 2 O.G/ and X 2 U.g/ since
.Ns  tN/.X / D mCG1 .Ns ˝ tN/ .X /
X 0
D .1/jtjjX.1/ j .1/jX.1/ j .1/jX.2/ j j.DXL.1/ s/jj.DXL.2/ t /j
X
D .1/jtjjX.1/ j .1/jsjjX.2/ j .1/jXj j.DXL st /j
D sSt .X /;
P
where X.1/ ˝ X.2/ D .X /.
•  is an SLG morphism, i.e.,
B x
G D G B .  /:
Indeed, for each s 2 O.G/, X; Y 2 U.g/, and g; h 2 G0 ,
Œ.. ˝  / 
G .s//.X; Y /.g; h/ D .1/jXjCjY j j.DXL ˝ DYL / 
G .s/j.g; h/

D .1/jXjCjY j jDhL1 :X DYL sj.gh/


D Π.s/ .h1 :X /Y /.gh/
D Œ. Gx  .s//.X; Y /.g; h/:

•  is an isomorphism. This is due to Corollary 5.1.3 since jj is clearly bijective

and, for each g 2 G0 , the differential .d/g is bijective as
xXL jg /.s/ D D
Œ.d/g .D xXL jg  .s/
D evg .D xXL  .s//
xXL  .s/.1/.g/
D ŒD
D .1/jXj  .s/.X /.g/
D j.DXL s/j.g/
D DXL jg .s/;
134 7 Super Lie groups

where we denote by D x L a left-invariant differential operator on G

x corresponding
X
to X 2 U.g/ while DX denotes a left-invariant differential operator on G. Notice
L

also that if fXi g is a basis of g, then DXLi jg ´ evg DXLi , g 2 jGj D jGj, x forms
x
a basis of Tg .G/.

It only remains to check the commutative diagrams (7.9). If

.0 ; / W .G0 ; g/ ! .H0 ; h/

x !H
is an SHCP morphism, then .0 ; / induces N W G x and we want to check that
N N N
jj D 0 and d D . Clearly jj D 0 and, for each X 2 h and f 2 O.H x /,

N /.f / D eve DXL .N  .f //

Œd .X
D .1/jXj Œ0 f .X /.e/
D .D L
 .X/
f /.e/:

On the other hand, if

W G ! H
is an SLG morphism and N D .K B H /./, we want to check that N   D    . For
each s 2 O.H /, X 2 U.g/, and g 2 jGj, we have

Œ.N   .s//.X/.g/ D .1/jXj Œjj .jD.d

L
/e .X/ sj/.g/
jXj 
D .1/ j L
.D.d /e .X/ s/j.g/ (since jj .jsj/ D j  .s/j
for each s 2 O.H /)
jXj  
D .1/ Œ .1 ˝ .d/e .X // H .s/.g/

D .1/jXj Œ.1 ˝ X /.  ˝   / 

H .s/.g/

D .1/jXj Œ.1 ˝ X /  
G  .s/.g/ (since  is an SLG morphism)
jXj
D .1/ ŒDXL   .s/.g/
D Œ.   .s//.X /.g/:
L
(Notice that D.d /e .X/
is on H and DXL on G.) 

In the next example we show explicitly how to recover the group structure of a
super Lie group starting from its associated SHCP.

Example 7.4.14. We consider the super Lie group G D GL.1j1/. In the language of
T -points, GL.1j1/.T / is the set of invertible matrices x21 x12 with multiplication
     
x1 1 y x y C 1 x1 1 C 1 y2
 1 1
D 1 1 2
; (7.13)
2 x2 2 y2 2 y1 C x2 2 x2 y2 C 2 1
 7.4 Super Harish-Chandra pairs 135

with xi 2 O.T /0 , i 2 O.T /1 for a supermanifold T . The corresponding reduced

group is jGj D .R n f0g/2 . A basis for the left-invariant vector fields gl.1j1/ is easily
recognized to be

@ @ @ @
X1 D x1 C 2 ; X2 D x2 C 1 ;
@x1 @2 @x2 @1
@ @ @ @
‚1 D x1  2 ; ‚2 D x2  1 ;
@1 @x2 @2 @x1

with commutation relations (for all i; j D 1; 2)

ŒXi ; Xj  D 0; Œ‚i ; ‚i  D 0;
ŒXi ; ‚j  D .1/ iCj
‚j ; Œ‚1 ; ‚2  D X1  X2 :
y10
The element h D 0 y2 2 jGj acts through the adjoint representation on gl.1j1/1 as
follows:
h  ‚1 D y1 ‚1 y21 ; h  ‚2 D y2 ‚2 y11 :

Using the theory developed in the previous section, we now want to reconstruct
the multiplication map of G in terms of the corresponding SHCP. Introduce the linear
operators
V
fi W z
.gl.1j1/1 / ! C 1 .G/; 1 7! yi ; ‚1 ; ‚2 ; ‚1 ^ ‚2 7! 0
and
V
'i W z
.gl.1j1/1 / ! C 1 .G/; ‚i 7! 1; 1; ‚j ¤i ; ‚1 ^ ‚2 7! 0:
V
The maps ffi ; 'i g are going to be our (global) coordinates on O.G/ D Hom. .gl.1j1//,
z They extend in a natural way to U.g0 /-linear maps from U.g0 / ˝ ƒ.g1 / to
C 1 .G//.
C .G/, which we denote by the same letter. We write fO (resp. ')
1 z
O for the composition
f B  1 (resp. 'i B  1 ).
We want to calculate the pullbacks:

. 
.fi //.X; Y /.g; h/ ´ fOi .h1  .X /.Y //.gh/
(7.14)
D fi .O 1 .h1  .X /.Y ///.gh/;

.  .'i //.X; Y /.g; h/ ´ 'Oi h1  .X /.Y //.gh/
(7.15)
D 'i .O 1 .h1  .X /.Y ///.gh/:

In order to perform the computations we first need to compute the elements O 1 .h1 
.X/.Y //. The next table collects them.
136 7 Super Lie groups

H
HHY 1 ‚1 ‚2 ‚1 ^ ‚2
X H
1 1 ‚1 ‚2 ‚1 ^ ‚2
y21 y1 .‚1 ^ ‚2 y11 y2
‚1 y11 y2 ‚1 0 2
.X1 C X2 /‚1
 21 .X1 C X2 //
y21 y1 .‚1 ^ ‚2 y21 y1
‚2 y21 y1 ‚2 0  2
.X1 C X2 /‚2
 21 .X1 C X2 //
‚1 ^ ‚2 ‚1 ^ ‚2  12 .X1 C X2 /‚1 1
2
.X1 C X2 /‚2 1
4
.X1 C X2 /2

From this and using Definitions (7.14) and (7.15), we can calculate easily the
various pullbacks. Let us do it in detail in the case of f1 . The pullback table of
.  .f1 //.X; Y /..x1 ; x2 /; .y1 ; y2 // is the following.
HH Y
HH 1 ‚1 ‚2 ‚1 ^ ‚2
X H
1 x1 y1 0 0 0
‚1 0 0  12 x1 y1 0
‚2 0  12 y21 x1 y12 0 0
‚1 ^ ‚2 0 0 0 1
x y
4 1 1

The link with the form of the multiplication morphism as given in equation (7.13)
is established by the isomorphism
 
x1 D f1 1 C '12'2 ; x2 D f2 1  '12'2 ; i D fi 'i :
The SHCP give us immediately the correspondence between Lie subgroups of a
supergroup G and Lie subalgebras of g D Lie.G/.
Proposition 7.4.15. Suppose that G is a connected super Lie group with super Lie
algebra g, and let h g be a Lie subalgebra of g. There exists a unique immersed
subgroup .H; j / whose Lie superalgebra is h.
Proof. Since G z is a connected Lie group, there exists a unique immersed Lie subgroup
z Q Q
.H0 ; j / with j W Hz ! G. z Let H denote the super Lie group associated with the super
Harish-Chandra pair .H z ; h/, and define

j  W O.G/ ! O.H /; f ! jQ .f jU.h/ /:

It is immediate that j  so defined is a superalgebra morphism, thus determining a
morphism j W H ! G. It is a super Lie group morphism since jQ is a classical Lie
group morphism. It is an immersion at e (and hence everywhere), in fact we have
Œ.dj /e Xe ./ D .Xe Bj  /./ D eve DXL j  ./ D .DXL j  /./.1/.e/ D j  ./.X /.e/:

 7.5 Homogeneous one-parameter supergroups 137

7.5 Homogeneous one-parameter supergroups

In this section we want to discuss homogeneous one-parameter supergroups for ma-
trix Lie supergroups. This is not the most general notion of one-parameter subgroup
for a Lie supergroup since we can also have non-homogeneous ones that have to be
introduced with a careful treatment of the exponential morphism. We will, however,
not pursue this here (see [50] for more details).
In the ordinary setting, any element in a Lie algebra g generates a one-dimensional
Lie subalgebra of g, which is the tangent space to the one-parameter subgroup asso-
ciated with such an element. This fact is no longer true for Lie superalgebras since
we know that the bracket Œx; x for an odd element x may not be zero, hence the Lie
subalgebra generated by such an element is no longer in general one-dimensional. So
it makes sense to introduce the notion of cyclic subalgebra.
Definition 7.5.1. Let g be a real Lie superalgebra. A Lie subalgebra k of g is called
cyclic if it is generated by a single element x 2 g. We then write k D hxi.
As we already remarked, in the classical framework, the similar notion of cyclic
Lie subalgebra is trivial: one always has hxi D R  x because of the identity Œx; x D 0.
We single out three special cases, for a given x homogeneous in g:
x 2 g0 H) hxi D R  x;
x 2 g1 ; Œx; x D 0 H) hxi D R  x;
x 2 g1 ; Œx; x 6D 0 H) hxi D R  x ˚ R  Œx; x: (7.16)
In particular, notice that in the third case the sum is direct because Œx; x 2 g0 , and
g0 \ g1 D f0g. Moreover, hxi D R  x ˚ R  Œx; x because Œx; Œx; x D 0 by the
(super) Jacobi identity.
Notice that while the Lie superalgebra structure in the first two cases is trivial, in
the third case instead, setting y ´ Œx; x, we have
jxj D 1; jyj D 0; Œx; x D y; Œy; y D 0; Œx; y D 0 D Œy; x:
A one-parameter subgroup of an ordinary Lie group G is the unique (connected) sub-
group K which corresponds, via the Frobenius theorem, to a specific one-dimensional
Lie subalgebra k of Lie.G/. To describe such K one can use the exponential map.
When g is linearized and expressed by matrices,P the exponential map is described by
the usual formal series on matrices: exp.X / ´ C1 n
nD0 X =nŠ.
We shall now adapt this approach to the context of Lie supergroups.
Let G be a Lie supergroup and let g D Lie.G/ be its Lie superalgebra. As-
sume further that G is embedded as a supergroup into GL.V / for some suitable
finite-dimensional super vector space V ; in other words, G is realized as a matrix
Lie supergroup.
Recall that – see Definition 7.5.1 – in the super context the role of one-dimensional
Lie subalgebras is played by cyclic Lie subalgebras.
138 7 Super Lie groups

Definition 7.5.2. Let X 2 g D Lie.G/ be a homogeneous element. We define the

one-parameter subgroup associated with X to be the Lie subgroup of G corresponding
to the cyclic Lie supersubalgebra hX i, generated by X in g, via the Frobenius theorem
for Lie supergroups (see Chapter 6).
For the definition it is clear that in general one-parameter supergroups may depend
on more than just one parameter, however we prefer to keep this terminology in analogy
with the ordinary case.
Now we describe these one-parameter subgroups. Fix a supermanifold T and set
A ´ O.T / (the superalgebra of global sections). Let t 2 A0 ,  2 A1 , and X 2 g0 ,
Y 2 g1 , Z 2 g0 such that ŒY; Z D 0. We define
P
C1
t2 2
exp.tX / ´ t n X n =nŠ D 1 C tX C 2Š
X C    2 GL.V .T //; (7.17)
nD0
exp.#Y / ´ 1 C #Y 2 GL.V .T //;
exp.tZ C #Y / ´ exp.t Z/  exp.#Y /
D exp.#Y /  exp.t Z/
D exp.t Z/  .1 C #Y /
D .1 C #Y /  exp.t Z/ 2 GL.V .T //:
All these expressions single out well-defined elements in GL.V .T //. In particular,
exp.tX/ in (7.17) belongs to the subgroup of GL.V .T // whose elements are all the
block matrices with zero off diagonal blocks. This is the standard group of matrices
GL.A0 ˝ V0 /  GL.A0 ˝ V1 /, and exp.tX / is defined here as the usual exponential
of a matrix: in particular, no convergence problems arise.
The set exp.A0 X / D fexp.tX / j t 2 A0 g is clearly a subgroup of GL.V .T //, once
we define, very naturally, the multiplication as
exp.tX /  exp.sX / D exp..t C s/X /:
On the other hand, if we consider the same definition for exp.A1 Y / ´ fexp.#Y / j
# 2 A1 g, we notice that in general it is not a subgroup. In fact, if we compute
exp.#1 Y /  exp.#2 Y / D .1 C #1 Y /.1 C #2 Y / D 1 C #1 Y C #2 Y C #1 #2 Y 2 ;
it follows that it is not the same as
exp..#1 C #2 /Y / D 1 C .#1 C #2 /Y D 1 C #1 Y C #2 Y:
So, recalling that Y 2 D ŒY; Y =2, we find that exp.A1 Y / is a subgroup if and only if
ŒY; Y  D 0 or #1 #2 D 0 for all #1 ; #2 2 A1 . This reflects the fact that the R-span
of X 2 g0 is always a Lie supersubalgebra of g, but the R-span of Y 2 g1 is a Lie
supersubalgebra if and only if ŒY; Y  D 0.
Thus, when ŒY; Y  6D 0 we must take exp.hY i.T // D exp.A1 Y C A0 Y 2 / as the
one-parameter subgroup corresponding to the Lie supersubalgebra hY i.
We summarize our discussion in the following result.
 7.5 Homogeneous one-parameter supergroups 139

Proposition 7.5.3. There are three distinct types of one-parameter subgroups asso-
ciated with a homogeneous element in g of dimension 1j0, 0j1, and 1j1, respectively.
Their functors of points are:
(a) for any X 2 g0 we have
xX .T / D fexp.tX / j t 2 O.T /0 g D R1j0 .T / D Hom.C 1 .x/; O.T //I
(b) for any Y 2 g1 , ŒY; Y  D 0, we have
xY .T / D fexp.#Y / D 1C#Y j # 2 O.T /1 g D R0j1 .T / D Hom.RΠ; O.T //I
(c) for any Y 2 g1 , Y 2 ´ ŒY; Y =2 6D 0, we have
xY .T / D fexp.t Y 2 C #Y / j t 2 O.T /0 ; # 2 O.T /1 g
D R1j1 .T / D Hom.C 1 .x/Π; O.T //:
In cases (a) and (b) the multiplication structure is trivial, while in case (c) it is
given by .t; #/  .t 0 ; # 0 / D .t C t 0  # # 0 ; # C # 0 /.
Proof. The case (a), namely when X is even, is clear. When instead X is odd we have
two possibilities: either ŒX; X  D 0 or ŒX; X  ¤ 0. The first possibility corresponds,
by the Frobenius theorem, to a 0j1-dimensional subgroup whose functor of points is,
as is easily seen, representable and of the form (b). Let us now examine the second
possibility.
The Lie subalgebra hX i generated by X is of dimension 1j1 by (7.16). Hence by the
Frobenius theorem it corresponds to a Lie subgroup of the same dimension, isomorphic
to R1j1 .
Now we compute the group structure on this R1j1 , using the usual functor of points
notation to give the operation of the supergroup. For any commutative superalgebra
A, we have to calculate t 00 2 A0 , # 00 2 A1 such that
exp.tX 2 C #X /  exp.t 0 X 2 C # 0 X / D exp.t 00 X 2 C # 00 X /
where t; t 0 2 A0 , #; # 0 2 A1 . The direct calculation gives
exp.tX 2 C #X /  exp.t 0 X 2 C # 0 X /
D .1 C #X / exp.tX 2 /  exp.t 0 X 2 /.1 C # 0 X /
D .1 C #X / exp..t C t 0 /X 2 /.1 C # 0 X /
D exp..t C t 0 /X 2 /.1 C #X /.1 C # 0 X /
D exp..t C t 0 /X 2 /.1 C .# C # 0 /X  # # 0 X 2 /
D exp..t C t 0 /X 2 /.1  # # 0 X 2 /.1 C .# C # 0 /X /
D exp..t C t 0 /X 2 / exp.# # 0 X 2 /.1 C .# C # 0 /X /
D exp..t C t 0  # # 0 /X 2 /.1 C .# C # 0 /X /
D exp..t C t 0  # # 0 /X 2 C .# C # 0 /X /;
140 7 Super Lie groups

where we use the property

exp.U C V / D exp.U / exp.V / if ŒU; V  D 0;

and also the fact that ŒX; X 2  D ŒX 2 ; X 2  D 0. 

7.6 References
Historically the concept of supermanifold was introduced in order to formalize the no-
tion of super Lie group. In fact, although super Lie algebras were already an object of
study by the physicists who originally introduced them to describe super-time infinites-
imal symmetries (see [36], [80], [34], [65]), the geometrical global objects encoding
such infinitesimal structure were not introduced until 1975 in [11].
The basic reference where super Lie groups are systematically studied for the first
time is again [49]. In particular, in [49], Kostant introduced Lie–Hopf algebras and
stated the analogue of Theorem 7.4.5. In the notes by Deligne and Morgan [22] there is
a brief discussion of super Harish-Chandra pairs that we fully develop in Section 7.4.
The realization of the structure sheaf of a super Lie group we give in terms of the
corresponding super Harish-Chandra pair is due to Koszul and appeared in [50].
The material in Section 7.5 appeared first, to our knowledge, in [31].
8
Actions of super Lie groups

In this chapter, we want to focus our attention on the concept of action of a super
Lie group G on a supermanifold M . When G acts on M , if we fix a topological
point p 2 jM j, the orbit morphism G.T / 3 g 7! g  p 2 M.T / is a constant rank
morphism. This nice property gives us the representability of the stabilizer functor and
hence allows us to show right away the representability of all the supergroup functors
for the classical supergroups like SLmjn and Ospmjn , the orthosymplectic supergroup
(see Example 8.4.8).
In the previous chapter we have seen how the concept of super Lie group is essen-
tially captured by the super Harish-Chandra pair associated with it, so that we have an
equivalence of categories between super Lie groups and SHCP. It is only natural to ask
how the concept of action of a super Lie group translates in the language of SHCP, to
give an equivalent approach to the theory of homogeneous spaces.

8.1 Actions of super Lie groups on supermanifolds

Let us start by briefly recalling the notion of action of an ordinary Lie group on an
ordinary manifold M .
Definition 8.1.1. Let G be a Lie group, M a manifold. We say that G acts on M or
equivalently that M is a G-space if there is a morphism G  M ! M , .g; x/ 7! g  x,
such that
(1) 1  x D x for all x 2 M ,
(2) .g1 g2 /  x D g1  .g2  x/ for all x 2 M and all g1 ; g2 2 G.
In Chapter 7 we have briefly discussed the definition and a few properties of actions
of Lie supergroups on a supermanifold. Let us quickly recall them.
Definition 8.1.2. Let G be a super Lie group, with multiplication, inverse and unit
given by , i and e, respectively. A morphism of supermanifolds

aW G  M ! M

is called a (left) action of G on M if it satisfies for all supermanifolds T :

(1) 1  x D x for all x 2 M.T /, 1 the unit in G.T /,
(2) .g1 g2 /  x D g1  .g2  x/ for all x 2 M.T / and all g1 ; g2 2 G.T /,
142 8 Actions of super Lie groups

or equivalently:
a B .  idM / D a B .idG  a/; (8.1a)
O idM i D idM :
a B he; (8.1b)
If we have an action a of G on M , then we say that G acts on M , or that M is a
G-supermanifold.
Remark 8.1.3. Equations (8.1a) and (8.1b) correspond to the commutativity of the
diagrams
idM
GGM /GM M ' R0j0 
eid /GM
MMM
MMM ww
idG a MMM www
ww a
a
MMM
  id
& w{ w
GM
a / M, M
and formalize the usual properties satisfied by classical actions. As we already pointed
out in Remark 7.2.8, we could also define right actions by introducing the flip morphism
in equation (8.1a).
Let p 2 jM j. We can interpret, as usual, a point p in the topological space jM j as
an element of M.R0j0 /, i.e., as morphism pR0j0 W R0j0 ! M with jpR0j0 j W R0 ! jM j
assigning to the only element in R0 the point p and pR 0j0 being the evaluation at p.
Since we always have the unique morphism T ! R0j0 for any supermanifold T , by
functoriality we also have a morphism M.R0j0 / ! M.T / and we can define pT the
image of pR0j0 under such a morphism. We also define pO W M ! M as the morphism
pOT W M.T / ! M.T /, m 7! pT .
Definition 8.1.4. If p 2 jM j and g 2 jGj define the morphisms
ap W G ! M; ag W M ! M
in the functor of points notation as
ap W G.S / ! M.S/; g 7! g  pS ;
ag W M.S/ ! M.S/; m 7! gS  m:
Equivalently:
ap ´ a B hidG ; pi;
O ag ´ a B hg;
O idM i;
idG p gidM
G 'GR 0j0 /GM M ' R0j0  /GM
LLL w MMM w
LLLap ww MMM g ww
LLL www MaMM www
LL& w a MMM w a
{w
w & w{ w
M, M.
The maps ap and ag satisfy the relations
 8.1 Actions of super Lie groups on supermanifolds 143

1
• ag B ag D idM for all g 2 jGj,
• ag B ap D ap B `g for all g 2 jGj and p 2 jM j.
We call ap the orbit morphism.
The next proposition states some nice properties of the maps ag and ap . In particular
we have that ap is a constant rank morphism. In the super context, this is a more delicate
result than its classical counterpart since, as we have seen in Chapter 5, the concept
of constant rank itself is more subtle. As we shall see, this result is of fundamental
importance in the proof of the existence of the stabilizer subgroup (see Section 8.4).
Proposition 8.1.5. Let G be a Lie supergroup acting on a supermanifold M via the
action a. Then
(1) ag is a superdiffeomorphism for all g 2 jGj,
(2) ap has constant rank for all p 2 jM j.
1
Proof. The first item follows at once from ag B ag D idM for all g 2 jGj.
Let us consider the second point. Suppose that M is a supermanifold of dimension
O n/
.m; O and G is a super Lie group of dimension .k;O q/.
O
Let g be the super Lie algebra of G and let Jap be the Jacobian matrix of ap in a
neighbourhood of a point g 2 jGj. Since
jJap j.g/ D .dap /g D .dag /p .dap /e .d`g 1 /g
and ag and `g 1 are diffeomorphisms, jJap j.g/ has rank dim g  dim ker .dap /e for
each g 2 jGj. Recall that if X 2 g we denote by DXL ´ .1 ˝ X /  the left-invariant
vector field associated with X. Using equation (8.1a) we have, for each X 2 ker .dap /e ,
DXL ap D .1 ˝ X / 
.1 ˝ evp /a
D .1 ˝ X ˝ evp /.  ˝ 1/a (8.2)
D .1 ˝ .dap /e .X //a D 0:
If ft i ;  j g and fx k ; l g are coordinates in a neighbourhood U of e, and in a neigh-
bourhood V jap j.U / of p, respectively, then
0  k  .x k /
1
@ap .x / @ap

Jap D @ @a@t i
 . l /
@ j A2M
@a . l / O nj
m; O qO .OG .U //:
O k;
p p
@t i @ j

We want to find a matrix A 2 GLkj O qO .OG .U // such that Jap A has a certain set of
column equal to zero. We are going to use equation (8.2).
Let mjn D dim ker .dap /e and let fXu g and f„v g be bases of g0 and g1 such that
Xu ; „v 2 ker .dap /e for u  m and v  n. Let
X @ X @ X @ X @
DXLu D au;i i
C ˇu;j j ; L
D„ v
D v;i i
C dv;j j
@t @ @t @
i j i j
144 8 Actions of super Lie groups

 au;i  v;i
(with au;i ; dv;j 2 OG .U /0 and ˇu;j ; v;i 2 OG .U /1 ) and A D ˇu;j dv;j . Clearly,
the vector fields Xu ; „v being linearly independent at each point, we have that the
reduced matrix jAj is invertible so that A 2 GLkj O qO .OG .U //.
Now use equation (8.2) to conclude that the matrix
!
DXLu ap .x k / D„ L  k
a .x /
Jap A D v p
DXLu ap . l / D„ L  l
a . /
v p

has m C n zero columns.

Suppose hence that we have an even matrix Jap in Mm; O nj O qO such that jJap j has
O k;
rank equal to kO  mj O
 qO  n with entries in OG .U / and only the first k  m C qO  n
columns non-zero: wz 00 . It is not restrictive to assume that z invertible so that the
 z 1 0 I 0
matrix G D wz 1 I is such that GJ D 0 0 . More precisely, it has the form

0 1
0 ˛1 0 ˇ1
B0 ˛2 0 ˇ2 C
B C:
@0 1 0 ı1 A
0 2 0 ı2

Since jJap j has rank .kO  m; qO  n/, we can suppose that ˛1 and ı1 are invertible.
 As
in Proposition 5.2.6, we can rearrange the matrix so that it takes the form wz 00 , with
 z 1 0 I 0
z invertible and the matrix G D wz 1 I such that GJ D 0 0 .
We can then conclude that Jap has constant rank in U and, by translation, in all
of jGj. 

8.2 Infinitesimal actions

In this section we discuss the infinitesimal interpretation of a super Lie group action
on a supermanifold. In particular we will show that any action (resp. right action) of a
Lie supergroup G on a supermanifold M induces an anti-morphism (resp. morphism)
from the Lie superalgebra of G to the tangent bundle of the supermanifold.
Let G be a super Lie group, M a supermanifold, and let

aW G  M ! M

be an action. If v 2 Te G and Ue is a neighbourhood of the identity e, then the

composition
a .v˝1/
OM .U / ! OGM .Ue  U / ! OM .U /
is a derivation of OM .U / for any open U  jM j. The Leibniz property can be verified
directly by calculating in local coordinates. The next definition is thus well posed.
 8.2 Infinitesimal actions 145

Definition 8.2.1. The composition .v ˝ 1/ B a defines a vector field on M which we

denote by a .v/:

a .v/.f / D .v ˝ 1/.a .f // (8.3)
for f a local section on M .
It is clear that as v varies we get a well-defined morphism of super vector spaces
aW Te .G/ ! VecM ; v 7! a .v/;

from Te .G/ into the super vector space of vector fields on M .

Remark 8.2.2. Let M D G and let the map a D  be the anti-group law defined by
 W G.S /  G.S/ ! G.S/, .g.S /; h.S // ´ h.S /g.S /. It is a right action. Then
 .v/ D .v ˝ 1/ D .1 ˝ v/ 
D DvL
is the unique left-invariant vector field on G which defines the tangent vector v at e. If
we take a D , where .gg 0 / D gg 0 is the ordinary group law, then

 .v/ D .v ˝ 1/ D DvR ;
where DvR is the unique right-invariant vector field defining the tangent vector v at e.
We know that v 7! DvL is a linear isomorphism of Te .G/ with Lie.G/, and one can
check that DvL 7! DvR is an anti-isomorphism of super Lie algebras. We leave this to
the reader as an exercise.
The next theorem asserts that the association
aW DvL 7! a .v/

is an antimorphism of super Lie algebras from Lie.G/ to VecM .

Theorem 8.2.3. Let a be an action of the super Lie group G on the supermanifold M .
The map
a W Lie.G/ ! VecM ; v 7! .v ˝ 1/a ;
(see (8.3)) is an antimorphism of super Lie algebras. Moreover,
.DvR ˝ 1O.M / /.a f / D a . a .v/f / (8.4)
for v 2 Te .G/, DvR its corresponding right-invariant vector field, and f a function
on M .
Proof. In order to prove the first statement it is enough to prove the second one. Indeed,
suppose that we have proved (8.4). Then
R R
a .ŒDv ; Dw e / D Œ.eve ˝ 1/ B .ŒDvR ; Dw
R
 ˝ 1/ B a
D Œ.eve ˝ 1/..DvR ˝ 1/.Dw
R
˝ 1/  .Dw
R
˝ 1/.DvR ˝ 1//a
D .eve ˝ 1/Œ.DvR ˝ 1/ B a a .w/  .Dw
R
˝ 1/ B a a .v/
D Πa .v/; a .w/;
146 8 Actions of super Lie groups

and the first statement follows from the fact that the assignment DXL ! DXR is a super
Lie algebra anti-isomorphism.
It thus remains to prove equation 8.4. This is proved by

.DvR ˝ 1/a .f / D .v ˝ 1O.G/ ˝ 1O.M / / B .  ˝ 1O.M / / B a .f /

D .v ˝ 1O.G/ ˝ 1O.M / / B .1O.G/ ˝ a / B a .f /
D .v ˝ a / B a .f / D .1 ˝ a /.v ˝ 1/ B a .f /
D a . a .v/f /: 

Corollary 8.2.4. The anti-morphism a extends to an associative algebra anti-mor-

phism (which we also call a )

aW U.Lie.G// ! U.VecM /:

Proof. We use the universal property of the universal enveloping algebra and extend
the anti-morphism by mapping basis to basis. We can characterize the extension also
by the relation (8.4): for v1 ; v2 ; : : : ; vk 2 Te .G/ and f a local section of OM , we have

.v1 v2 : : : vk ˝ 1O.G/ /.a f / D a . a .v1 v2 : : : vk /f /

D a . a .v1 / a .v2 / : : : a .vk /f /: 

8.3 Actions of super Harish-Chandra pairs

In this section we want to translate the concept of action of a super Lie group on a
supermanifold into the language of SHCP. As we shall see, the following definition
comes naturally.
Definition 8.3.1. We say that an SHCP .G0 ; g/ acts on a supermanifold M if there is
a pair .a; / such that the following holds:
(i) a W G0  M ! M is an action of G0 on M ;
(ii) W g ! Vec.M / is a super Lie algebra anti-morphism
such that

jg0 .X / ' .X ˝ 1O.M / /a for all X 2 g0 ;

g 1
.g  Y / D .a / .Y /.ag / for all g 2 jGj; Y 2 g:

Since the category of super Lie group is equivalent to the category of SHCP, one
could ask whether assigning an SHCP action on a supermanifold M is equivalent to
assigning an action of the corresponding super Lie group on M . The answer is positive
as we shall presently see.
 8.3 Actions of super Harish-Chandra pairs 147

Proposition 8.3.2. Let a W G  M ! M be an action of the super Lie group G on the

supermanifold M , then the morphisms
z  M ! M , with a ´ a B .j  idM /, where j W G
(1) a W G z ! G is the natural
z
immersion of the reduced Lie group G into G,

(2) aW g ! Vec.M /op ; X 7! .X ˝ 1O.M / /a (8.5)

z g/ on M .
define an action of the associated SHCP .G;

Proof. This is a simple check that we leave to the reader as an exercise. 

We are going to discuss the converse of the previous proposition. In other words, we
want to understand how to recover the action of a super Lie group G on a supermanifold
M from the knowledge of the action of its associated SHCP.
In analogy with the classical case, one could use the super Frobenius theorem to
reconstruct a local action of G from its infinitesimal action (8.5). However, we want
to take a different path and avoid the use of the super Frobenius theorem, proceeding
instead to the explicit reconstruction of the global action.
Assume that we have an action a of G on M , and let .a; a / be the corresponding
action of its SHCP as in Proposition 8.3.2. If f 2 O.M /, then by definition

z y̋ O.M / ' HomU.g / .U.g/; C 1 .G/

a .f / 2 ŒHomU.g0 / .U.g/; C 1 .G// z y̋ O.M //:
0

Using equation (8.1a) and the form of the left-invariant vector fields as given in the
previous chapter, we have

a .f /.X / D .1/jXj Œ.DXL ˝ 1/a .f /.1/

D .1/jXj Œ.1 ˝ X ˝ 1/ B . 
˝ 1/ B a .f /.1/
D .1/jXj Œ.1 ˝ X ˝ 1/ B .1 ˝ a / B a .f /.1/
D .1/jXj .1 ˝ a .X //.a

.f /.1//
jXj 
D .1/ .1 ˝ a .X //a .f /:

This suggests the definition of the action a and proves its uniqueness, as we shall
see in the next proposition.

Proposition 8.3.3. Let .G;z g/ be the SHCP associated with the super Lie group G and
z g/ on a supermanifold M . Then there is a unique action
let .a; / be an action of .G;
a W G  M ! M of the super Lie group G on M whose reduced and infinitesimal
actions are .a; /. a is explicitly given by

z y̋ O.M //;
a W O.M / ! HomU.g0 / .U.g/; C 1 .G/
(8.6)
f 7! ŒX 7! .1/jXj .1C 1 .G/ 
z ˝ .X //a .f /:
148 8 Actions of super Lie groups

Proof. Let us check that a .f / is U.g0 /-linear. For all X 2 U.g/ and Z 2 g0 we
have, using the fact that a is an anti-homomorphism,

a .f /.ZX / D .1/jXj .1 ˝ .ZX //a .f /

D .1/jXj .1 ˝ .X //.1 ˝ Ze ˝ 1/.1 ˝ a /a .f /
D .1/jXj .1 ˝ .X //.1 ˝ Ze ˝ 1/.j j ˝ 1/a .f /
D .DL ˝ 1/Œa .f /.X /:
Z

We now check that a is a superalgebra morphism:

Œa .f1 /  a .f2 /.X / D mC 1 .G/  

z y̋ O.M / Œa .f1 / ˝ a .f2 / .X /

D .1/jXj mŒ.1 ˝ .X.1/ //a .f1 / ˝ .1 ˝ .X.2/ //a .f2 /

D .1/jXj .1 ˝ .X //.a .f1 /  a .f2 //
D a .f1  f2 /.X /;

where fi 2 O.M / and X.1/ ˝ X.2/ denote synthetically .X /. Concerning the

“associative” property, we have, for X; Y 2 U.g/ and g; h 2 jGj,

Œ. ˝ 1/a .f /.X; Y /.g; h/ D Œa .f /.h1  X Y /.gh/

D .1/jXjCjY jCjX jjY j .Y / .h1 :X /.agh / .f /

D .1/jXjCjY jCjX jjY j .Y /.ah / .X /.ag / .f /
D Œ.1 ˝ a /a .f /.X; Y /.g; h/

and, finally, .eve ˝ 1/a .f / D .1/.ae / .f / D f . 

We end this section by revisiting Example 7.4.14.
Example 8.3.4. Consider again the super Lie group G D GL1j1 introduced in Exam-
ple 7.4.14. Then G acts on itself by left multiplication, and, using the same notation
as in the previous example, we have:
(1) Left action of jGj on G:
1   1   
t 0 x 1 t 1x1 t1 1
 D :
0 t2 2 x 2
t2 2 t 2x2

(2) Representation of gl.1j1/ on the super Lie algebra Vec.G/op :

@ @ @ @
X1 7! x 1 C 1 ; X2 7! x 2 C 2 ;
@x 1 @ 1 @x 2 @ 2
@ @ @ @
‚1 !
7 x2 C 2 ; ‚2 !
7 x1 C 1 :
@ 1 @x 1 @ 2 @x 2
 8.4 The stabilizer subgroup 149

In this case the representation sends each element of gl.1j1/ into the correspond-
ing right-invariant vector field.
The action can be reconstructed using equation (8.6); a simple calculation shows that
 
.t 1 / D t 1 x 1 .1 C  1  2 / C t 1 2  1 ; .t 2 / D t 2 x 2 .1 C  1  2 / C t 2 1  2 ;
 
. 1 / D t 1 1 .1 C  1  2 / C t 1 x 2  1 ; . 2 / D t 2 2 .1   1  2 / C t 2 x 1  2 :

The usual form of the multiplication map (as given in Example 7.4.14) is obtained
using the isomorphism

1 2
t 1 7! t 1 .1 C  1  2 /;  1 7! I t 2 7! t 2 .1 C  1  2 /;  2 7! :
t1 t2

8.4 The stabilizer subgroup

Let G be a super Lie group acting on a supermanifold M . The aim of this section is
to define the notion of stabilizer subgroup at a point in jM j and to characterize it from
different perspectives. Let us start by recalling such a concept in the ordinary setting.
Definition 8.4.1. Suppose that G is a Lie group acting on a manifold M , and let
x0 2 M . We define Gx0 the stabilizer subgroup at x0 to be the subgroup of G given
by
Gx0 ´ fg 2 G j g  x0 D x0 g:
This definition has a natural generalization to the super context. Let p 2 jM j be a
topological point. As we already remarked in Section 8.1, such p can be interpreted as
an element pT 2 M.T / for all supermanifolds T .
Definition 8.4.2. Let G be a super Lie group acting on a supermanifold M , and let
p 2 jM j. We define F Gp , the stabilizer functor at p, to be the functor from .smflds/
to .sets/ given by

F Gp .T / ´ fg 2 G.T / j g  pT D pT g  G.T / for all T 2 .smflds/:

(We leave to the reader its definition on morphisms.)

Notice that we have defined just a functor F Gp W .smflds/ ! .sets/, so we are not
sure that it corresponds to a supermanifold; in other words, we do not know if this is
the functor of points of a supermanifold.
We want to reformulate this definition using the language of categories in order to
prove the representability of such functor in the category of super Lie groups. We first
need to establish some terminology.
150 8 Actions of super Lie groups

Definition 8.4.3. Given two objects X and Y in a category and two arrows ˛ and ˇ
between them, an equalizer is a universal pair .E; / that makes the following diagram
commute:
 / ˛ /
E X /Y:
ˇ

This means that if  W T ! X is such that ˛ B  D ˇ B , then there exists a unique

 W T ! E such that  B  D  . If an equalizer exists, it is unique up to isomorphism.

One can also define the notion of coequalizer in the same way, by reversing all the
arrows.
Let p 2 jM j and denote by pO the morphism

Š ip
pO W G !
 R0j0 
! M;

where Š W G ! R0j0 and ip W R0j0 ! M denote respectively the unique morphism from
G to a point supermanifold and the canonical injection of p in M .

Definition 8.4.4. Let G be a super Lie group and a W G  M ! M an action of G on

the supermanifold M . The stabilizer subgroup at p 2 jM j is the supermanifold Gp
equalizing the diagram
ap
G // M :
pO

As one can readily check, in the functor of points notation we have

Gp .T / D fg 2 G.T / j g  p D pg; T 2 .smflds/:

We notice that it is not a priori clear that such an equalizer exists. We are going to
establish a proposition relating the two Definitions 8.4.2 and 8.4.4 of stabilizer functor
and stabilizer subgroup, showing that the stabilizer subgroup always exists and its
functor of points is precisely the stabilizer functor.
Before this, we need a category-theoretic lemma.

Lemma 8.4.5. Let C be a category. The Yoneda embedding, i.e., the injection C 3
X ! Hom.; X/, preserves the equalizers.

Proof. For this proof, we temporarily adopt the following notational convention. If X
is an object in C , the corresponding representable functor is denoted by hX .
e
Suppose that E !  X  Y is an equalizer in a category C . Then we need to
show that for any equalizing diagram F ! hX  hY , the natural transformation
a W F ! hX factors uniquely via hE ! hX .
For any object Z of C and any x 2 F .Z/, aZ .x/ is an arrow Z ! X that
composes equally with X  Y and therefore factors uniquely via e. Therefore define
 8.4 The stabilizer subgroup 151

bZ .x/ W Z ! E, bZ .x/ 2 hE .Z/ to be the unique map such that e B bZ .x/ D aZ .x/.
Such b is functorial in Z. Certainly he B b D a by construction.

F @
a / hX
@@ | >
@@b e |||
@@ |
@ |||
hE .

For uniqueness, if b 0 is some other natural transformation with he Bb 0 D a, then for any
0
object Z and any x 2 F .Z/ we have e B bZ .x/ D aZ .x/ D e B bZ .x/. This implies
0
that bZ .x/ D bZ .x/ since e is a monomorphism. Hence b D b 0 and uniqueness is
established. 

Lemma 8.4.6. If A and B are two (super)algebras and ˛, ˇ are morphisms between
them, their coequalizer is the algebra C D B=J , where J is the ideal .˛.a/  ˇ.a/ j
a 2 A/.

Proof. The coequalizer makes the following diagram commute:

ˇ
C o B oo A:
˛

The result is then clear. 

We are ready for the main result of this section.

Proposition 8.4.7. Let G be a Lie supergroup acting on the supermanifold M and let
p 2 jM j.
(1) The diagram
ap
G // M
pO

admits an equalizer Gp 2 .smflds/.

(2) Gp is a super Lie subgroup of G.
(3) The functor F Gp W T 7! F Gp .T / assigning to each supermanifold T the
stabilizer of pT of the action of G.T / on M.T / is represented by the supermanifold
Gp .
(4) Let .Gzp ; gp / be the SHCP associated with the stabilizer Gp . Then G
zp G z is
the classical stabilizer of p with respect to the reduced action and gp D ker dap .

Proof. (1) According to Proposition 8.1.5 ap W G ! M is a constant rank morphism.

Let Jp D ff 2 O.M / j f .p/ D 0g and I be the ideal in O.G/ generated by a .Jp /.
By Proposition 5.3.13 there exists a closed embedded submanifold .Gp ; jGp / of G
distinguished by the ideal I .
152 8 Actions of super Lie groups

Since the embedding jGp W Gp ! G is regular and closed, j  is surjective. Hence

Gp
O.Gp / ' O.G/= ker j  , and moreover,
Gp

ker jG D hap .f / j f 2 Jp i

p

D hap .f  f .p// j f 2 O.M /i

D hap .f /  pO  .f / j f 2 O.M /i:

Therefore we have the co-equalizing diagram


ap j
O.M / // O.G/ Gp
/ O.G / :
p
pO 

Hence Gp is the equalizer of

jG ap
p
/G // M :
Gp
pO

By the uniqueness of the equalizer we have Gp D Gp .

(2) In order to prove (2) we have to show that Gp is a super Lie subgroup of G.
Due to Yoneda’s lemma, it is enough to prove (3).
(3) This can be proved easily by noticing that the functor F Gp equalizes the natural
transformations
ap
h // h :
G M
pO

(Again here hG and hM denote the functor of points of the supermanifolds G and
M .) Since (see Lemma 8.4.5) the Yoneda embedding preserves equalizers and due to
uniqueness, it follows that F Gp ' hGp .
(4) The first statement is clear since jGj ' G.R0j0 / as set-theoretical groups.
Moreover, since jGp B ap .f / is a constant for all f 2 O.M /, we have gp ker dap ,
and equality holds for dimension considerations. 
We are ready for some important examples.
Examples 8.4.8. (1) Consider the action (expressed with the functor of points notation):

a W GLmjn R1j0 ! R1j0 ; .g; c/ 7! Ber.g/c; g 2 GLmjn .T /; c 2 R1j0 .T /:

The stabilizer of the point 1 2 jR1j0 j coincides with all the matrices in GLmjn .T / with
Berezinian equal to 1, that is, the special linear supergroup SLmjn .T /. By Proposi-
tion 8.4.7 we have immediately that SLmjn is representable and that it is a super Lie
subgroup of GLmjn .
 8.5 References 153

(2) Consider the action

a W GLmjn B ! B; .g; .  ;  // ! .g  ; g  /;

where B is the super vector space of all the symmetric bilinear forms on Rmjn . Consider
the point in B that corresponds to the standard bilinear form in B:
0 1
0 Ip 0 0 0
BIp 0 0 0 0C
B C
ˆDB B 0 0 1 0 0 C if m D 2p C 1; n D 2q;
C
@0 0 0 0 Iq A
0 0 0 Iq 0
or 0 1
0 Ip 0 0
BIp 0 0 0C
ˆDB @0 0
C if m D 2p; n D 2q:
0 Iq A
0 0 Iq 0

The stabilizer of the point ˆ is the supergroup functor Ospmjn . Again this is a Lie
supergroup by Proposition 8.4.7.
For more details on the Lie superalgebras and the bilinear forms see Appendix A.

8.5 References
The action of the super Harish-Chandra pairs appears in [22]. The representability of
the stabilizer functor in Proposition 8.4.7 is stated in [22]; however a complete proof
of this statement appears only in [4].
9
Homogeneous spaces

In this chapter we examine the construction of the quotient of supergroups in the

ordinary and the super geometric context. Classically we have that if G is a topological
group and H is a closed subgroup, we can give to the quotient G=H the quotient
topology, which is the finest topology for which the natural projection  W G ! G=H
is open. If the groups G and H have an additional structure, for example if they are
Lie groups or algebraic groups or the corresponding super objects, it is very natural to
ask if the topological space G=H inherits the extra structure in a unique way.
We start our discussion by describing transitive actions of super Lie groups on su-
permanifolds. We then proceed with a review of the classical construction of quotients
of groups in the differential category. We do not include the proofs for the statements,
referring the reader to [75], Ch. 2, for a comprehensive treatment of this subject. Fi-
nally we discuss the problem of the construction of quotients in the category of super
Lie groups in full detail, including the description of the functor of points of quotients
together with some physically relevant examples.

9.1 Transitive actions

In this section we want to describe transitive actions in the super setting and to look at
them from different perspectives. Let us start by recalling the classical definition.
Let G be an ordinary Lie group.
Definition 9.1.1. We say that Lie group G acts transitively on an ordinary manifold
M , or that M is a homogeneous space for G if there is an action

G  M ! M; g; m 7! g  m;

of G on M and a point x0 2 M such that the morphism x0 W G ! M , x0 .g/ D g  x,

is surjective.
We now turn to the supergeometric setting.
Let now G be a Lie supergroup, and e its multiplication and identity element
respectively.
Definition 9.1.2. We say that a Lie supergroup G acts on a supermanifold M if there
exists a morphism of supermanifolds a W G  M ! M , a.g; x/ ´ g  x for all
g 2 G.T /, x 2 M.T /, where T is a generic supermanifold, such that
 9.1 Transitive actions 155

(1) 1  x D x for all x 2 M.T /,

(2) .g1 g2 /  x D g1  .g2  x/ for all x 2 M.T / and all g1 ; g2 2 G.T /.
We say that the action a is transitive, if there exists p 2 jM j such that ap W G ! M ,
ap .g/ ´ g  p, is a surjective submersion. In this case we also say that M is a
homogeneous superspace or a G-superspace.

Remark 9.1.3. Let ag:p D ap B rg , with rg W G ! G defined by rg D .1 ˝ evg / 

or equivalently, in the functor of points notation, rg W G.T / ! G.T /, rg .x/ D xg. If
ap is submersive for one p 2 jM j then it is submersive for all p 2 jM j.

The next proposition characterizes transitive actions.

Proposition 9.1.4. Suppose that M is a G-superspace, for a fixed action a. Then the
following facts are equivalent:
(1) a is transitive.
(2) jaj W jGj  jM j ! jM j is transitive,
.dap /e W g ! Tp .M / is surjective for one p (hence for all p 2 jM j).
(3) If q denotes the odd dimension of G, then

ap;R0jq W G.R0jq / ! M.R0jq /

is surjective.
(4) The sheafification of the functor (see Appendix B) im ap W .smflds/op ! .sets/

.im ap /.T / ´ fap B g j g 2 G.T /g D fg  pT j g 2 G.T /g  M.T /

is the functor of points of M , where pT 2 G.T / is the topological point p 2 jGj

viewed as an element of G.T /.

Proof. (1) () (2). This is an immediate consequence of Proposition 8.1.5 and our
previous remark.
(1) H) (3). If  2 M.R0jq / D Hom.R0jq ; M /, let jj 2 jM j be the image of the
reduced map associated with  (with a small, but usual, abuse of notation we denote
by the same symbol jj both the morphism  W R0 ! jM j and the point p which is its
image jj.R0 / 2 jM j). The pullback   depends only on the restriction of the sections
of OM to an arbitrary neighbourhood of jj. This is an easy exercise, which we leave
to the reader. If ap is a surjective submersion, there exists a local right inverse s of ap
defined in a neighbourhood of jj. By the locality of , s B  is a well-defined element
of G.R0jq / and, moreover,

ap;R0jq .sR0jq B / D ap;R0jq B sR0jq B  D 

so that ap;R0jq is surjective.

156 9 Homogeneous spaces

(3) H) (2). Suppose that ap;R0jq surjective. Looking at the reduced part of each

morphism in ap;R0j0 G.R0jq / , we have that ap;R0j0 D jap j W jGj ! jM j is surjective.
As a consequence of the ordinary manifold theory result in [48], Theorem 5.14, jaj is a
classical transitive action and jajp is a submersion. Let now m 2 jM j and ft i ;  j g be
coordinates in a neighbourhood U of it. Consider the element  2 M.R0jq / defined
by
V
  W OM .U / ! O.R0jq / D .1 ; : : : ; q /; t i 7! jt i j.m/;  j 7! j :
By surjectivity of ap;R0jq , there exists 2 G.R0jq / such that ap;R0jq . / D :

B ap .t i / D jt i j.m/; 
B ap . j / D j :
This implies that Tm .M /1 is in the image of .dap /j j . Since, by our previous con-
siderations, jap j is a submersion, Tm .M /0 is in the image as well. Hence, due to
Proposition 8.1.5 we are done.
(1) H) (4). Let us suppose that ap is a surjective submersion. Let m 2 jM j
and g 2 jap j1 .m/ (jap j is surjective, so it exists). Since ap is a submersion there
exists V jGj with coordinates X1 ; : : : ; XpCq (dim G D pjq) and W jM j with
coordinates Y1 ; : : : ; YmCn (dim M D mjn) such that
ap .Yi / D Xi :
Let t 2 U jT j for a generic supermanifold T and ˛ W U ! M such that m D j˛j.t /.
We can assume that j˛j.U / W . If ˛  .Yi / D fi 2 OT .U /, then ˇ W U ! V defined
by ´
 fi if i  m C n;
ˇ .Xi / D
0 otherwise;
satisfies ap B ˇ D ˛. Then Œ˛ 2 .im ap;T / t , hence .im ap;T / t D M t . This shows that
A
im ap D M (see Proposition B.2.9).
A
(4) (H (1). Let us suppose that im ap D M . Taking T D R0j0 we have that jap j
must be surjective. Let us now assume that T D M and m 2 jM j. There exists U 3 m
and W U ! G such that ap B D idU . Then ap must be a submersion at j j.m/,
and this is true everywhere since ap has constant rank. Indeed for all g 2 jGj,
.dap /g B .dlgG /1 D .dlgM /x B .dap /1
0

where the isomorphisms lgG and lgM are the left actions of g on G and M , respectively.

Remark 9.1.5. Notice that in the statement of .4/ in Proposition 9.1.4, it is too restric-
tive to require the transitivity of aT for each T , i.e., im ap .T / D M.T /. In fact, in
the ordinary setting, this would imply that we can lift every morphism T ! M to a
morphism T ! G. As a consequence, we have the existence of a global section of the
fibration G ! M (take T D M and the identity map), which is not true in general. It
is hence necessary to take the sheafification of the image functor, as we do in point (4).
 9.2 Homogeneous spaces: The classical construction 157

9.2 Homogeneous spaces: The classical construction

In this section we review the classical construction of the quotient of a complex or real
Lie group G by a closed subgroup H . For the complete discussion we refer the reader
to [75], p. 74.
As we shall see, this construction is made possible by the existence of a local section
of the natural projection  W G ! G=H . Such a section in general does not exist in
the algebraic category; this is due to the different nature of the topology involved. For
this reason the arguments cannot be replicated in the algebraic setting, which requires
a much deeper analysis of the problem. We shall not pursue this question further.
We have the following result.
Proposition 9.2.1. Let G act transitively on a manifold M and let Gx0 D fg 2 G j
gx0 D x0 g be the stabilizer subgroup at x0 2 M . Then Gx0 is a closed subgroup of
G and the morphism x0 W G ! M , x0 .g/ D g  x0 is a submersion of G onto M .
Proof. See Lemma 2.9.2 in [75]. 
Let us assume that we have a Lie group G acting transitively on a manifold M and,
for a fixed x0 2 M , let us consider Gx0 the stabilizer subgroup at x0 , which is a closed
subgroup of G by the previous proposition. The set G=Gx0 is a topological space
with the quotient topology with respect to the natural projection  W G ! G=Gx0 . We
would like to give to G=Gx0 a manifold structure compatible with the natural action
of G. Also, we ask if every homogeneous space arises in this way and if the manifold
structure on G=Gx0 is unique. All these questions have positive answers, summarized
in the following theorem.
Theorem 9.2.2. Let G be a Lie group, H a closed Lie subgroup. Then there exists
a unique manifold structure on G=H such that the natural action of G on G=H ,
g; xH 7! gxH is a morphism. Moreover, if M is any manifold on which G acts
transitively and x0 is a fixed point in M with Gx0 D H , then the map

G=Gx0 ! M; gGx0 7! g  x0 ;

is a diffeomorphism of G=Gx0 onto M .

The manifold G=H is called the quotient of G by H .
Proof. See Theorem 2.9.4 in [75]. 
We would like to make some comments on the construction of the manifold structure
on the quotient G=H in the classical context, without going into the technicalities of
Theorem 9.2.2.
We begin by defining a sheaf O D OG=H on G=H as follows. For all open subsets
U in G=H , we define O.U / as the algebra of functions on U , consisting of all the
f W U ! R such that f   is C 1 on  1 .U /, where  W G ! G=H is the natural
158 9 Homogeneous spaces

projection. It is easy to check that this is a sheaf. We need to show that this defines a
manifold structure on G=H and that G acts on the corresponding ringed space naturally.
To this end, it is enough to show that, for some open subset V in G=H containing .1/,
there is a homeomorphism of V with a smooth manifold W such that OjV goes
over OW under . We obtain such W as a submanifold of G containing 1 such that
V D .W / and  1 .V / Š W  H in the following way. The map w; h 7! wh of
W  H into G is a diffeomorphism of W  H onto  1 .V / (which is open in G)
commuting with the right actions of H (y 7! yh) on G and .w; k/ 7! .w; kh/ on
W  H.
The crucial existence of W , based ultimately on the local Frobenius theorem, is
equivalent to the existence of a local section V  G=H ! W  H  G. Moreover,
as mentioned above, one has W  H Š W H . The sheaf OjV can hence be identified
with the sheaf of C 1 functions on W . This proves that we have defined a manifold
structure on G=H . The uniqueness and the universal nature of this structure are not
difficult to prove (see [75], Ch. 2).
What we have described above is the key idea to the proof of Theorem 9.2.2, and
we shall see in Section 9.3 that, despite the different context, the existence of a local
section plays a crucial role also in the construction of a quotient of Lie supergroups in
very much the same way.

9.3 Homogeneous superspaces for super Lie groups

We are now interested in the construction of homogeneous spaces for super Lie groups.
In Section 9.2 we have seen that we have a unique manifold structure on the quotient
of a Lie group by a closed subgroup, preserving the natural action of the group on its
quotient. We now want to obtain a similar result in the super setting.
Let G be a Lie supergroup and H a closed Lie subgroup.1 We want to define a
supermanifold structure on the topological space jGj=jH j. This structure will turn
out to be unique once we impose some natural conditions on the action of G on its
quotient. In order to do this we first define a supersheaf OX on the topological space
jXj D jGj=jH j, in other words, we define a superspace X D .jGj=jH j; OX /. We then
prove the local splitting property for X , that is we show that X is locally isomorphic to
domain in Rpjq for some p and q. We start by defining the supersheaf OX on jGj=jH j.
Let g D Lie.G/ and h D Lie.H /. For each Z 2 g, let DZ be the left-invariant
vector field on G defined by Z (see Chapter 7 for more details). For x0 2 jGj let `x0
and rx0 be the left and right translations of G by x0 . We denote by ix0 D `x0 B rx1
0
the
inner automorphism defined by x0 . It fixes the identity and induces the transformation
Adx0 on g (see Definition 7.3.8).
1
We say “subgroup” instead of the more cumbersome “subsupergroup”. However, the reader must be
aware that our H has a supergroup structure, inherited naturally from the one on G.
 9.3 Homogeneous superspaces for super Lie groups 159

Definition 9.3.1. For any subalgebra k of g we define the subsheaf Ok of OG as

Ok .U / D ff 2 OG .U / j DZ f D 0 on U for all Z 2 kg:

On the other hand, for any open subset W  jGj, invariant under right translations by
elements of jH j, we define

Oinv .W / D ff 2 OG .W / j f is invariant under rx0 for all x0 2 jH jg:

If jH j is connected we have

Oinv .W / D Oh0 .W /;

as one can readily check by looking infinitesimally at the condition rx0 f D f for all
x0 2 jH j. For any open set jW j  jX j D jGj=jH j with jW j D 01 .jW j/ we put

OX .jW j/ D Oinv .W / \ Oh .W /  OG .W /:

Clearly OX .jW j/ D Oh .W / if jH j is connected. The subsheaf OX is a supersheaf on

jXj. We have thus defined a ringed superspace X D .jX j; OX /. Our aim is to prove
that X is a supermanifold with OX as its structure sheaf.

It is clear that the left action of the group jGj on jX j leaves OX invariant and so
it is enough to prove that there is an open neighborhood jW j of the topological point
jj.1/  1N in jGj=jH j with the property that .jW j; OX jjW j / is a super domain, i.e.,
isomorphic to an open submanifold of Rpjq .
We will do this using the local Frobenius theorem (see Chapter 6). Also, we identify
as usual g with the space of all left-invariant vector fields on G, thereby identifying the
tangent space of G at every point canonically with g itself.
On G we have a distribution spanned by the vector fields in h. We denote it by Dh .
On each jH j-coset x0 jH j we have a supermanifold structure which is a closed
submanifold of G. It is an integral supermanifold of Dh , i.e., the tangent space at any
point is the subspace h at that point. By the local Frobenius theorem there is an open
neighborhood U of 1 and coordinates xi , 1  i  n, and ˛ , 1  ˛  m, on U such
that Dh is spanned on U by @=@xi , @=@˛ (1  i  r, 1  ˛  s). Moreover, from the
theory on jGj we may assume that the slices L.c/ ´ f.x1 ; : : : ; xn / j xj D cj ; r C1 
j  ng are open subsets of distinct jH j-cosets for distinct c D .crC1 ; : : : ; cn /. These
slices are therefore supermanifolds with coordinates xi , ˛ , 1  i  r, 1  ˛  s.
We have a submanifold W 0 of U defined by xi D 0 with 1  i  r and ˛ D 0 with
1  ˛  s. The map jj W jGj ! jX j may be assumed to be a diffeomorphism of
jW 0 j with its image jW j in jX j, and so we may view jW j as a superdomain, say W .
The map jj is then a diffeomorphism of W 0 with W . What we want to show is that
W Š .jW j; OX jjW j /.
160 9 Homogeneous spaces

Lemma 9.3.2. The map

W0H !
 G; w; h ! wh;

is a super diffeomorphism of W 0  H onto the open submanifold of G with reduced

manifold the open subset jW 0 jjH j of jGj.

Proof. The map  in question is the informal description of the map B .iW 0  iH /
where iM refers to the canonical inclusion M ,! G of a sub-supermanifold of G into
G, and W G  G ! G is the multiplication morphism of the Lie supergroup G. We
shall use such informal descriptions without comment from now on.
It is classical that the reduced map jj is a diffeomorphism of jW 0 j  jH j onto the
open set U D jW j0 jH j. This uses the fact that the cosets wjH j are distinct for distinct
w 2 jW 0 j. It is thus enough to show that d is surjective at all points of jW 0 j  jH j.
For any h 2 jH j, right translation by h (on the second factor in W 0  H and simply rh
on G) is a super diffeomorphism commuting with  and so it is enough to prove this
at .w; 1/. If X 2 g is tangent to W 0 at w and Y 2 h, then

d.X; Y / D d.X; 0/ C d.0; Y / D d .X; 0/ C d .0; Y / D X C Y:

Hence the range of d is all of g since from the coordinate chart at 1 we see that the
tangent spaces to W 0 and wjH j at w are transversal and span the tangent space to G
at w which is g. This proves the lemma. 

Lemma 9.3.3. We have

  OX jjW j D OW 0 ˝ 1;
where   W OG !  OW 0 H .

Proof. To ease the notation we drop the open set in writing a sheaf superalgebra, that
is, we will write OX instead of OX .U /.
We want to show that for any g in OX jU ,   g is of the form f ˝ 1 and that the
map g 7! f is bijective with OW 0 . Now   intertwines DZ (Z 2 h) with 1 ˝ DZ and
so .1 ˝ DZ /  g D 0. Since the DZ span all the super vector fields on jH j, it follows
using charts that for any p 2 jH j we have   g D fp ˝ 1 locally around p for some
fp 2 OW 0 . Clearly fp is locally constant in p. Hence fp is independent of p if jH j
is connected. If we do assume that jH j is connected, the right invariance under jH j
shows that fp is independent of p. In the other direction it is obvious that if we start
with f ˝ 1, it is the image of an element of OX jU . 

Theorem 9.3.4. The superspace .jXj; OX / is a supermanifold.

Proof. By the previous lemmas we know that .jX j; OX / is a super manifold at 1. N The
left invariance of the sheaf under jGj shows this to be true at all points of jX j. 
 9.3 Homogeneous superspaces for super Lie groups 161

We now want to describe the action of G on the supermanifold X D .jGj=jH j; OX /

we have constructed. Notice that in the course of our discussion we have also shown
that there is a well-defined morphism  W G ! X .
Proposition 9.3.5. There is a unique morphism ˇ W G  X ! X such that the diagram

GG /G

1 
 ˇ 
GX /X

commutes.
Proof. Let ˛ ´  B W G  G ! X . The action of jGj on jX j shows that such a map
jˇj exists at the reduced level. So it is a question of constructing the pullback map
ˇ  W OX ! OGX
such that
.1  / B ˇ  D ˛  :
Now   is an isomorphism of OX onto the sheaf OG restricted to a sheaf on X
(W 7! OG .jj1 .W //), and so to prove the existence and uniqueness of ˇ  it is a
question of proving that ˛  and .1  / have the same image in OGG . It is easy to
see that .1  / has as its image the subsheaf of sections f killed by 1 ˝ DZ .Z 2 h/
and invariant under 1  rh .h 2 jH j/. It is not difficult to see that this is also the image
of ˛  . 
We tackle now the question of the uniqueness of X .
Proposition 9.3.6. Let X 0 be a super manifold with jX 0 j D jX j and let  0 be a
morphism G ! X 0 . Suppose that
(1)  0 is a submersion,
(2) the fibers of  0 are the supermanifolds which are the cosets of H .
Then there is a natural isomorphism X ' X 0 .
Proof. Indeed, from the local description of submersions as projections it is clear
that, for any open jW j  jX j, the elements of  0 .OX 0 .jW j// are invariant under
rh ; .h 2 jH j/ and killed by DZ .Z 2 h/. Hence we have a natural map X 0 ! X
commuting with  and  0 . This is a submersion, and by dimension considerations it
is clear that this map is an isomorphism. 
We have proved the following result:
Theorem 9.3.7. Let G be a Lie supergroup and H a closed Lie subgroup. There exist
a supermanifold X D .jGj=jH j; OX / and a morphism  W G ! G=H such that the
following properties are satisfied:
162 9 Homogeneous spaces

(1) The reduction of  is the natural map jj W jGj ! jX j.

(2)  is a submersion.
(3) There is an action ˇ from the left of G on X reducing to the action of jGj on jX j
and compatible with the action of G on itself from the left through :

GG /G

1 
 ˇ 
GX / X.

Moreover, the pair .X; / subject to the properties (1), (2), and (3) is unique up to
isomorphism. The isomorphism between two choices is compatible with the actions,
and it is also unique.
Proof. This is an immediate consequence of the previous lemmas and propositions.


9.4 The functor of points of a quotient supermanifold

In this section we want to understand how to write the functor of points for the quotient
of a Lie supergroup G by a closed Lie subgroup H . Our intuition suggests that we
write the functor T 7! G.T /=H.T / that associates to each supermanifold T the coset
space G.T /=H.T /. As we shall see, this is not far from the correct answer, however, as
we already know giving a functor is by no means sufficient to define a supermanifold
unless we can prove the functor is representable, i.e., it is the functor of points of a
supermanifold.
Our goal is first to prove a representability criterion that enables us to single out
among all the functors from the category of supermanifolds to the category of sets,
those which are representable. This criterion is very formal, hence very similar in
statement and in proof to the same result in the algebraic category, which we are going
to describe in detail in Chapter 10. However, since the context and the notation here
are slightly different, we feel that it is worth giving a proof in both cases, although the
proofs of Theorem 9.4.3 and Theorem 10.3.7 are essentially the same.
This section is very much independent from the rest of our work and can be skipped
in a first reading. We suggest that the reader compare it with Section 10.3 of Chap-
ter 10, which deals with the representability issues in the algebraic context. Since our
treatment is very formal and relies on very general categorical results and definitions,
like Yoneda’s lemma and the sheaf property of functors, the reader will realize that
statements and proofs are very much the same in the differential and algebraic settings.
In dealing with representability issues we need to distinguish between a superman-
ifold X and its functor of points hX , so in this section we shall use this convention. We
start with the notion of local functor (see Appendix B for more details).
 9.4 The functor of points of a quotient supermanifold 163

Definition 9.4.1. Let F W .smflds/op ! .sets/ be a functor (not necessarily repre-

sentable). We say that F is local if it has the sheaf property, in other words if we
have the following. For any supermanifold T and any open covering fTi g of T let
i W Ti ,! T , ij W Ti \ Tj ,! Ti be the natural immersions. If we have a family
˛i 2 F .Ti / such that F .ij /.˛i / D F .j i /.˛j /, then there exists a unique ˛ 2 F .T /
such that i .˛/ D ˛i .
Notice that this implies that when F is restricted to the category of the open sets of
a fixed supermanifold, F is a sheaf in the ordinary sense.
Definition 9.4.2. Let U and F be functors from .smflds/op to .sets/. We say that a
functor U is a subfunctor of F if U.T /  F .T / for all T 2 .smflds/.
We further say that a subfunctor U  F is an open subfunctor if for all super-
manifolds S, T and morphisms ˛ W hT ! F we have ˛ 1 .U.S // D hV .S /, for V
open in T . We say that a collection of open subfunctors Ui cover F if ˛ W hT ! F ,
˛ 1 .Ui .T // D hVi for all T 2 .smflds/, and the Vi cover T .
U is an open supermanifold subfunctor of F if U is a representable open subfunctor
of F .
Theorem 9.4.3 (Representability criterion). Let F W .smflds/op ! .sets/ be a functor.
Then the functor F is representable if and only if
(1) F is local,
(2) F is covered by the collection of open supermanifold functors fUi g.
Proof. If F is representable, F Š hX for a supermanifold X . One can check directly
that hX has the two properties listed above, in particular a cover by open supermanifold
functors is given by hX˛ where the X˛ are open submanifolds covering X .
Now the other direction. We shall assume that Ui D hX˛ instead of Ui Š hX˛
leaving to the reader the slightly more general case as an exercise.
Define hX˛ˇ ´ hX˛ F hXˇ (for the definition of fibered product refer to Ap-
pendix B).
The functor hX˛ F hXˇ is representable, in fact by its very definition hX˛ F hXˇ D
iˇ1 .X˛ / D i˛1 .Xˇ /. By our hypothesis (2) this is representable, hence we can
write it as hX˛ˇ , for X˛ˇ an open submanifold of X˛ . Notice that X˛ˇ is also an
open submanifold of Xˇ (by a slight abuse of notation we shall use the same symbol
X˛ˇ to denote isomorphic supermanifolds). As we shall see, X˛ˇ corresponds to the
intersection of the two open submanifolds X˛ and Xˇ of the supermanifold X , that we
shall construct, whose functor of points is F .
By the definition of fibered product we have the commutative diagram
jˇ;˛
hX˛ˇ D hX˛ F hXˇ / hXˇ

j˛;ˇ iˇ
 i˛ 
hX˛ / F.
164 9 Homogeneous spaces

We now proceed and build the supermanifold X by gluing the supermanifolds X˛ .

We start by defining the underlying topological space.
As a set we define `
jXj ´ jX˛ j= ;
˛

where is the following equivalence relation: x˛ xˇ if and only if there exists

x˛ˇ 2 jX˛ˇ j with jj˛;ˇ j.x˛ˇ / D x˛ , jjˇ;˛ j.x˛ˇ / D xˇ for all x˛ 2 jX˛ j, xˇ 2 jXˇ j.
Here we use j˛;ˇ to denote also the superscheme morphism j˛;ˇ W X˛ˇ ! X˛ . One
can check this is an equivalence relation and that the map ˛ W jX˛ j ,! jX j is an
injective map. jX j becomes a topological space through the topology induced by the
(open) topological spaces jX˛ j.
We now want to define a sheaf of superalgebras OX on jX j by using the sheaves
OX˛ and “gluing” them. Let U be open in jX j and let U˛ D ˛1 .U /. Put
` 
OX .U / ´ fff˛ g 2 ˛2I OX˛ .U˛ / j jˇ; .fˇ / D j ;ˇ .f / for all ˇ;  2 I g:

The condition jˇ; .fˇ / D j ;ˇ .f / simply states that to be an element of OX .U /, the
collection ff˛ g must be such that fˇ and f agree on the intersection of Xˇ and X
for any ˇ and  . Here we are again abusing the notation using j˛ˇ for both the functor
of points morphism and the supermanifold morphism. One can check directly that OX
is a sheaf of superalgebras by its very construction.
We have defined a supermanifold X D .jX j; OX /; in order to finish the proof,
we need to show that hX Š F . We are looking for a functorial bijection between
hX .T / D Hom.T; X / and F .T /, for all T 2 .smflds/.
We first construct a natural transformation T W F .T / ! hX .T /.
Let t 2 F .T / D Hom.hT ; F /, by Yoneda’s lemma. Consider the diagram

hT˛ ´ hX˛ F hT / hT

t˛ t
 i˛ 
hX˛ / F.

Since fhX˛ g is an open cover of F , the fT˛ g form an open cover of T . Since by
Yoneda’s lemma: Hom.hT˛ ; hX˛ / Š Hom.T˛ ; X˛ / we obtain a family of morphisms:
t˛ W T˛ ! X˛  X . The morphisms t˛ glue together to give a morphism t 0 W T ! X ,
hence t 0 2 hX .T /. So we define T .t / D t 0 .
Next we construct another natural transformation T W hX .T / ! F .T /, which
turns out to be the inverse of .
Assume that we have f 2 hX .T /, i.e., f W T ! X . Let T˛ D f 1 .X˛ /. We
immediately obtain morphisms g˛ W T˛ ! X˛  F . By Yoneda’s lemma, g˛ corre-
sponds to a natural transformation g˛ W hT˛ ! hX˛ . Since F is local, the morphisms
i˛  g˛ W hT˛ ! hX˛ ! F glue together to give a morphism g W hT ! F , i.e., an
element g 2 F .T /. Define T .f / D g.
 9.4 The functor of points of a quotient supermanifold 165

One can directly check that and  are indeed natural transformations inverse to
each other, hence F Š hX . 

Remark 9.4.4. The representability criterion we just proved, holds as it is, also for the
category of analytic supermanifolds discussed in Section 4.8. The proof is the same,
and the reader is invited to check that at no point did we make any use of the fact that
a sheaf has any local C 1 property. In fact Theorem 9.4.3 is a categorical result, the
reader may also consult [77], Ch. 1, for a more general setting of this statement. We
are going to revisit this same proposition for the case of superschemes in Chapter 10.
However, we are going to employ a slightly different category and for this reason the
statement is only apparently not the same. All we say in Theorems 9.4.3 and 10.3.7 is
contained in [77], where the treatment is the most general possible.

We now turn to examine the functor of points of the quotient of a Lie supergroup
G by a closed subgroup H . We are interested in a characterization of the functor of
points of X D G=H in terms of the functor of points of the supergroups G and H .
A
Theorem 9.4.5. Let G and H be supergroups as above and let G=H be the sheafifica-
A
tion of the functor: T ! G.T /=H.T /. Then G=H is representable and is the functor
of points of the homogeneous space supermanifold X D G=H constructed above.

Proof. In order to prove this result, we shall use the uniqueness property which char-

A A
acterizes the homogeneous space G=H (see Theorem 9.3.7). So we only need to prove
that G=H is representable, i.e., G=H D hX for a supermanifold X , and that X satisfies

A
the three properties detailed in Theorem 9.3.7.

A
To prove that G=H is representable we use the criterion in Theorem 9.4.3. The fact

A
that G=H has the sheaf property is clear by its very definition. So it is enough to prove
there is a open supermanifold subfunctor of G=H around the origin (by translation we
can transport such open supermanifold subfunctor to obtain a neighbourhood at every
point). But this is given by hW , with W Š W 0  H constructed as in the previous
section.
We now turn to the properties (1), (2) and (3). (1) and (3) are left to the reader as
an exercise. The fact  is a submersion, that is property (2), comes by looking at it in
the local coordinates given by W . 

We now examine some interesting examples.

Example 9.4.6 (Coadjoint orbits of SLmjn ). Consider the following action of G D

SLmjn on its Lie superalgebra g D slmjn :

W G.T /  g.T / ! g.T /; g; X 7! gXg 1 :

Let us fix a topological point X0 D diag. 1 ; : : : ; mCn / 2 jgj, where the real
numbers 1 ; : : : ; mCn are all distinct. Then one sees immediately that the stabilizer
166 9 Homogeneous spaces

subgroup H at X0 is given by
80 19
ˆ
< a1 ::: 0 >
=
B :: C  G.T /:
H.T / D @ ::: : A>
:̂ ;
0 ::: amCn

We now want to describe the orbit of X0 as a submanifold of g. Consider the

ideal I in O.g/ generated by the polynomial functions str n , where str.Z/ returns the
supertrace of an element Z 2 g. I is a regular ideal (see Proposition 5.3.5), hence it
defines a closely embedded supermanifold X in g. We leave to the reader the routine
check of properties (1), (2) and (3) in Theorem 9.3.7 so that we have X D G=H . One
A
also immediately verifies that hX D G=H . Moreover, hX is the functor of points of
the regular coadjoint orbits of SLmjn . A similar calculation can be done to compute
the regular orbits of the other classical supergroups.

9.5 The super Minkowski and super conformal spacetime

In this section we want to examine the example of F ` D F `.2j0; 2j1I 4j1/: the an-
alytic supermanifold of 2j0, 2j1 flags in the space C4j1 (for the definition of analytic
supermanifolds refer to Section 4.8) and one of its real forms. This example has a
special importance in physics since it is the complexification of the super conformal
space. This supermanifold is the compactification of the Minkowski superspace in 4j4
dimensions, where the Minkowski is realized as the “big cell”, which is a dense open
set inside the conformal superspace.
We define F `.T / as the set consisting of all pairs F1  F2 , where F1 and F2
are locally free subsheaves of OT4j1 ´ OT;0 ˝ C4j0 ˚ OT;1 ˝ C0j1 of rank 2j0 and
2j1, respectively. In other words, for each point t 2 jT j, .F1 / t and .F2 / t are free
4j1
submodules of OT;t of rank 2j0 and 2j1, respectively (recall that projective modules
over local superrings are free, see Appendix B). The pairs F1  F2 are called flags.
Notice that if T is a point so that O.T / D C, we have that F1  F2 are super
vector spaces of dimension respectively 2j0 and 2j1 in C4j1 , thus recovering the usual
definition of flag.
On O.T /4j1 there is a natural action of G D GL4j1 .T / which is inherited by F `.T /:

GL4j1 .T /  F `.T / ! F `.T /; g; F1  F2 7! g  F1  g  F2 ;

where g  Fi are defined as follows. Since Fi is a locally free subsheaf of OT4j1 , it

follows that Fi .T˛ / is free for a suitable base fT˛ g of open sets in T so that g  Fi .T˛ /
makes sense. For instance, for F1 it is

g  hv1 ; v2 i ´ hgv1 ; gv2 i:

 9.5 The super Minkowski and super conformal spacetime 167

Thus we have defined .g  Fi /.T˛ / for a base of open sets. Since the compatibil-
ity conditions in Definition 2.2.10 are satisfied, g  Fi is a B-sheaf and we can use
Proposition 2.2.11 to obtain a sheaf on T that we denote by g  Fi .
Let us fix the flag F D fOT2j0  OT2j1 g 2 F `.T /. The stabilizer subgroup at F is
the subgroup H of G given in terms of the functor of points by
80 19
ˆ
ˆ g11 g12 g13 g14 15 > >
ˆB
ˆ >
<Bg21 g22 g23 g24 25 C C=
>
H.T / D B B 0 0 g33 g34 0 C C> 2 G.T /:
ˆ
ˆ
ˆ@ 0 0 g43 g44 0 A> >
>
:̂ ;
0 0 53 54 g55
In fact, one can check directly that H stabilizes F and that the functor T 7! H.T /
is the functor of points of a closed subgroup of G. Hence locally, we can identify the
set F `.T / with the set G.T /=H.T / so that the sheafification of both functors coincide
(see Proposition B.2.9). Since F ` is already a sheaf, by its very definition, we have
that F ` is the functor of points of the quotient supermanifold G=H .
Next we want to describe the big cell inside F `. This is an open submanifold, which,
as we shall see, has a special importance in physics since it is the complexification of
the super Minkowski space.
Locally at a point t 2 jT j, an element F1  F2 in F `.T / consists of two free
4j1
submodules of OT;t or rank 2j0 and 2j1, and is thus described by a pair of subspaces
4j1
hv1 ; v2 i  hw1 ; w2 ; w3 i with vi ; wj 2 OT;t of suitable parity. Let U1 .T / and U2 .T /
be the set of subsheaves of OT4j1 of ranks 2j0 and 2j1 expressed locally as
8 0 1 0 1 9
ˆ
ˆ v11 v12 >
>
ˆ*B C B C+
ˆ   >
>
< Bv21 C Bv22 C ˇ =
Bv31 C ; Bv32 C ˇ detˇ v 11 v12
U1 .T / t D B C B C invertible ;
ˆ
ˆ @v41 A @v42 A v21 v22 >
>
ˆ >
>
:̂ ;
v51 v52
8 0 1 0 1 0 1 9
ˆ
ˆ w11 w12 w13 0 1 >
>
ˆ
ˆ * w C Bw C Bw C+ >
< B B 21 C B 22 C B 23 C ˇˇ w11 w12 w13 >
=
U2 .T / t D Bw C ; Bw C ; B w C ˇ Ber @ w w w A invertible :
ˆ B 31 C B 32 C B 33 C 21 22 23
>
ˆ
ˆ @w41 A @w42 A @w43 A w51 w52 w53 >
>
:̂ >
;
w51 w52 w53
In order to simplify the notation, from now on we shall write everything locally
without further mention.
Using the identification F `.T / Š .G=H /.T /, which locally amounts to having
F `.T / Š G.T /=H.T /, we can write an element in U1 .T /  U2 .T / uniquely as
00 1 0 11
I I 0
@@AA ; @B ˇ AA 2 U1 .T /  U2 .T /;
˛ 0 1
168 9 Homogeneous spaces

where I is the identity, A and B are .22/-matrices with even entries, and ˛ D .˛1 ; ˛2 /,
ˇ t D .ˇ1 ; ˇ2 / are rows with odd entries in OT;t .
We define the big cell U.T / inside F `.T / as the pairs .u; v/ 2 U1 .T /  U2 .T /.
Notice that an element of U1 .T / is inside U2 .T / if

A D B C ˇ˛: (9.1)

So a flag in the big cell U is completely described by the triplet .A; ˛; ˇ/. We see also
that U is an affine 4j4 superspace, i.e., the functor U is representable and is the functor
of points of a 4j4 superspace. Equation (9.1) is also known as the twistor relation in
the physics literature. Since there are no relations among A, ˛ and ˇ, we can take them
as local coordinates in a neighbourhood of the identity in jGj=jH j.
In these coordinates, the flag F corresponding to the identity that we fixed at the
beginning becomes
00 1 0 11
I I 0
@@ 0 A ; @ 0 0AA  .0; 0; 0/:
0 0 1
The big cell U1  U2 is obtained by requiring the invertibility of the determinant
.1; 2/ and the Berezinian .1; 2; 5/ (obtained by taking respectively columns .1; 2/ and
rows .1; 2/ and columns .1; 2; 5/ and rows .1; 2; 5/) in the generic expression of a flag
in F `. Clearly one can repeat the same argument and obtain a total of six different big
cells by requiring the invertibility of the determinant and Berezinians:

.1; 2/ .1; 2; 5/; .1; 3/ .1; 3; 5/; .1; 4/ .1; 4; 5/;

.2; 3/ .2; 3; 5/; .2; 4/ .2; 4; 5/; .3; 4/ .3; 4; 5/:

We leave to the reader the easy check that these big cells cover the whole of F `. Hence
we can apply the representability criterion in Theorem 9.4.3, thus proving that F ` is
an analytic supermanifold.
We now want to write explicitly the morphism  W G ! F `, .g/ D g  F in these
coordinates and see it is a submersion. In a suitable open subset near the identity of
the group we can take an element g 2 G.T / as
 
gij i5
gD ; i; j D 1; : : : ; 4:
5j g55

Then we can write an element g  F 2 U1 .T /  U2 .T /  F `.T / as

0 1 0 1
g11 g12 g11 g12 15 0 1 0 1
Bg21 g22 C Bg21 g22 25 C I I 0
B C B C
Bg31 g32 C ; Bg31 g32 35 C  @W Z 1 A ; @V Y 1 .2  W Z 1 1 /aA ;
B C B C
@g41 g42 A @g41 g42 45 A 1Z
1
0 1
51 52 51 52 g55
 9.5 The super Minkowski and super conformal spacetime 169

where
  
 g31 g32 g11 g12
1 D 51 52 ; W D ; ZD ;
g41 g42 g21 g22
   
 
1 D 15 ; 2 D 35 ; d D .g55  Z 1 1 /1 ;
25 45
1 1
V D W  g55 2 1; Y D Z  g55 1 1:

Finally the map  in these coordinates is given by

g 7! .W Z 1 ; 1Z
1
; .2  W Z 1 1 /d /:

At this point one can compute the super Jacobian and verify that at the identity it is
surjective, so  is a submersion. This gives an equivalent and independent proof of the
fact that F ` is the quotient G=H (see Theorem 9.3.7).
The subgroup of G leaving invariant the big cell is the set of matrices in G of the
form 0 1
L 0 0
@NL R RA ;
d' 0 d
with L, N , R being 2  2 even matrices,  and odd 1  2 matrix, ' a 2  1 odd matrix
and d a scalar. This is then what the physicists call the complex Poincaré supergroup
and its action on the big cell can be written as

A ! R.A C ˛/L1 C N; ˛ ! d.˛ C '/L1 ; ˇ ! d 1 R.ˇ C /:

If the odd part is zero, then the action reduces to the one of the classical Poincaré group
on the ordinary Minkowski space.
We see that the big cell of the flag supermanifold F `.2j0; 4j0I 4j1/ can be inter-
preted as the complex super Minkowski space time, the flag being its superconformal
compactification.
We now turn to the construction of the real Minkowski superspace which is ex-
tremely important in physics. We start with a real form for the supergroup G. In order
to obtain such a form (see Section 4.8 for the definition) we need a natural transforma-
tion from G to its complex conjugate G. x Define the natural transformation (T being a
supermanifold) as
    
x D  D i 
G.T / ! G.T /; g D !g D :
d i   dN

Lemma 9.5.1. We have .hg/ D g h .

Proof. Direct calculation. 

170 9 Homogeneous spaces

Remark 9.5.2. It is important to notice that we are under the following convention: if
 and are odd variables, then
 D N N : (9.2)
This convention is opposed to the one used most commonly in physics, namely

 D N N ;

but as it is explained in [22], it is the one that makes sense functorially. According to
this convention, then for matrices X , Y with odd entries

.X Y /T D .Yx /T .Xx /T :

We are ready to define the involutionwhich gives the real form of G:

   

x /; 1 F 0 0 1
G.T / !
 G.T g!
7 g ´ L.x / L; L D

; F D :
0 1 1 0

We have .hg/ D h g  and .g  / D g, so it is a conjugation.

As one can readily check from the definitions, we have the following lemma and
proposition.

Lemma 9.5.3. The map x 7! x  is a natural transformation. It defines a ringed space

involutive isomorphism W G ! G x which is C-antilinear.

Proposition 9.5.4. The topological space G  consisting of the points fixed by has a
real supermanifold structure and the supersheaf is composed of those functions f 2 OG
such that  .f / D f .

The involution we chose has important physical properties and we invite the reader
to consult [32] for a complete treatment of them.
Also, it is easy to check that it reduces to a conjugation on the Poincaré supergroup
and we can compute explicitly such conjugation as follows:
0 1 0 1
L 0 0 L1 0 0
g D @ M R RA ; g 1 D @R1 ML1 C 'L1 R1 d 1 A ;
d' 0 d 'L1 0 d 1
0 1
1
R 0 0
B C
g  D @L 1 M  R 1  L 1 '   L 1 j L 1 '  A :
j dN 1  0 dN 1

It follows that the fixed points are those that satisfy the conditions
1 1 
L D R ;  D j'  ; ML1 D .ML1 /  j L ' 'L1 : (9.3)
 9.5 The super Minkowski and super conformal spacetime 171

To get a more familiar form for the reality conditions, we observe that the last
equation in (9.3) can be cast as
1 1
M 0 L1  ML1 C j L '  'L1 ; M 0 D M 0 :

2
This is just an odd translation and amounts to multiplying g on the right by the group
element 0 1
I 0 0
1
g 0 D @ 12 jR1 L '  'L1 I 0A :
0 0 1
We want now to compute the real form of the big cell. The first thing to observe is
that the real form is well defined on the quotient space G=H (the superflag), where H
is the group described previously.
Notice that a point of the big cell .A; ˛; ˇ/ can be represented by an element of the
group 0 1
I 0 0
g D @A I ˇ A
˛ 0 1
since 0 1 0 1 0 1 0 1 0 1
I I I 0 I 0 I 0
g @ 0 A D @AA ; g @0 0A D @A ˇ A  @A  ˇ˛ ˇA :
0 ˛ 0 1 ˛ 1 0 1
We first compute the inverse
0 1
I 0 0
g 1 D @A C ˇ˛ I ˇ A ;
˛ 0 1
and then 0 1
I 0 0
g  D @A  ˛  ˇ  I j˛  A :
jˇ  0 1

The element g  is already in the desired form, so the real points are given by
A D A  j˛  ˛; ˇ D j˛  :
We can make a convenient change of coordinates,
1
A0  A C j˛  ˛;
2
so the reality condition is
A0 D A0 ;


and we recover the usual form for the (purely even) Minkowski space time.
172 9 Homogeneous spaces

9.6 References
This treatment of quotients of supergroups, via their functor of points and, equivalently,
through the direct construction of the sheaf of invariant functions, is found in [32], [4],
[5]. As for Example 9.4.6 see [76], [56].
10
Supervarieties and superschemes

The aim of this and the next chapter is to lay down the foundations of algebraic su-
pergeometry, using the machinery of sheaves and schemes together with their functor
of points that we have described in detail in the classical setting in the previous chap-
ters. The presence of nilpotents in the structural sheaf of an ordinary scheme makes its
generalization to supergeometry very natural and allows us to carry most definitions
and results from the ordinary to the super setting, almost with no changes. There are,
however, some remarkable differences that we shall point out along with our treatment.
The most striking one is the rigidity of the projective superspace, which does not con-
tain as many subvarieties as its classical counterpart. We shall, in fact, see that the
Grassmannian superscheme is not in general a projective supervariety. This fact has
some deep consequences, for instance, the strategy for the construction of the quotient
of algebraic supergroups has to be suitably modified. We shall however not pursue this
point further in the text.
We start by giving the definition of superscheme and its functor of points and we
examine some important examples, including affine and projective superspaces and
Grassmannian superschemes. We then describe a representability criterion, which al-
lows us to single out among functors from the category of superalgebras to the category
of sets those which are the functors of points of a superscheme. This criterion is espe-
cially useful in supergeometry since the functor of points is often the only way we can
handle supergeometric objects. Taking this point of view, we discuss the infinitesimal
theory in the algebraic super-setting. In particular we study superderivations and their
relations with the tangent space at a rational point of a superscheme over a field, making
some concrete calculations to exemplify our definitions. Our study of the infinitesimal
theory is of capital importance for the theory of algebraic supergroups that we shall
treat in the next chapter.

10.1 Basic definitions

In this section we give the basic definitions of algebraic supergeometry. Because we
are in need of a more general setting in the next chapters we no longer assume that the
ground field to be R or C and actually we are going to replace it with a commutative
ring.
Let k be a commutative ring. Assume that all superalgebras are associative, com-
mutative (i.e., xy D .1/p.x/p.y/ yx) with unit and over k, unless otherwise specified.
174 10 Supervarieties and superschemes

We denote their category by .salg/. For a superalgebra A, let JA denote the ideal gen-
erated by the odd elements, i.e., JA D hA1 iA . Write Ar for the quotient A=JA . We say
that Ar is reduced or super reduced even if it may contain some (even) nilpotents.
In Chapter 3 we have introduced the categories of superspaces and of superschemes.
Let us briefly recall these notions.
A superspace X D .jX j; OX / is a topological space jX j together with a sheaf of
superalgebras OX such that OX;x is a local superalgebra, i.e., it has a unique two-sided
maximal homogeneous ideal.
The sheaf of superalgebras OX is a sheaf of OX;0 -modules, where OX;0 is the sheaf
over jXj defined as OX;0 .U / ´ OX .U /0 for all U open in jX j. Notice that also the
sheaf OX;1 , defined as OX;1 .U / D OX .U /1 , is also a sheaf of OX;0 -modules.
Given two superspaces X D .jXj; OX / and Y D .jY j; OY /, a morphism f W X !
Y of superspaces is given by a pair f D .jf j; f  / such that
(1) jf j W X ! Y is a continuous map;
(2) f  W OY ! f OX is a map of sheaves of superalgebras on jY j, that is, for U
open in jY j there exists a family of morphisms fU W OY .U / ! OX .jf j1 .U //
compatible with restrictions;
(3) the map of local superalgebras fp W OY;jf j.p/ ! OX;p is a local morphism, i.e.,
sends the maximal ideal of OY;jf j.p/ to the maximal ideal of OX;p .
A superscheme X is a superspace .jX j; OX / such that OX;1 is a quasi-coherent sheaf
of OX;0 -modules. OX is called the structure sheaf of X. A morphism of superschemes
is a morphism of the corresponding superspaces.
For any open U  jX j we define the superscheme U D .jU j; OX jjU j /, called an
open subscheme in the superscheme X .
The most important example of a superscheme is given by the spectrum of a super-
algebra A. It consists of the spectrum of the even part A0 together with a certain sheaf
of superalgebras on it. Let us see this construction in detail.
Definition 10.1.1 (The superscheme Spec A). Let A be an object of .salg/. We have
Spec.A0 / D Spec.Ar / since the algebras Ar and A0 differ only by nilpotent elements.
Let us consider OA0 the structure sheaf of Spec.A0 / (see Chapter 2 for more details).
The stalk of this sheaf at the prime p 2 Spec.A0 / is the localization of A0 at p. As for
any superalgebra, A is a module over A0 . So, according to the classical construction
detailed in Chapter 2, Section 2.5, we have a sheaf AQ (that we shall denote also by OA )
of OA0 -modules over Spec A0 , with stalk
˚ 
Ap D fg j f 2 A; g 2 A0  p ;
the localization of the A0 -module A over each prime p 2 Spec.A0 /. This localization
has a unique two-sided maximal ideal which consists of the maximal ideal in the local
ring .Ap /0 and the generators of .Ap /1 as A0 -module.
As one can easily check, AQ (or OA ) is a sheaf of superalgebras and .Spec A0 ,
OA / is a superscheme that we denote by Spec A. In fact we just showed that it is a
 10.1 Basic definitions 175

superspace since OA;p is local; moreover, by its very definition, we have that OA;1 is a
quasi-coherent sheaf of OA;0 -modules since it is a sheaf of OA0 D OA;0 -modules.
Notice that on the basic open sets

Uf D fp 2 Spec A0 j .f / 6 pg; f 2 A0 ;

we have OA .Uf / D Af D fa=f n j a 2 Ag.

Definition 10.1.2. An affine superscheme is a superspace that is isomorphic to Spec A
for some superalgebra A in .salg/. Such superalgebra A, by the construction of Spec A,
is isomorphic to the superalgebra of the global sections of the structure sheaf of X ,
which we shall denote by O.X / (instead of the more cumbersome notation OX .jX j/).
An affine algebraic supervariety is a superspace isomorphic to Spec A for some
affine superalgebra A, i.e., a finitely generated superalgebra such that A=JA has no
nilpotents. We will call A the coordinate ring of the superscheme or the supervariety.
The next proposition tells us that superschemes have affine superschemes as local
models, as it happens for the ordinary setting.
Proposition 10.1.3. A superspace S is a superscheme if and only if it is locally iso-
morphic to Spec A for some superalgebra A, i.e., for all x 2 jSj there exists Ux  jS j
open such that .Ux ; OS jUx / Š Spec A. (Clearly A depends on Ux .)
Proof. Since S is a superscheme, by definition S 0 D .jS j; OS;0 / is an ordinary scheme,
that is, for each point x 2 jSj there exists an open set U  jSj such that .U; OS;0 jU / D
Spec A0 for an ordinary commutative algebra A0 .
By definition of superscheme, OS;1 jU is a quasi-coherent sheaf of OS;0 jU -modules,
i.e., of OA0 -modules. Hence there exists an A0 -module A1 such that OS;1 jU .U / D A1
(see Chapter 2, Section 2.5). So if we look at global sections we have

OS jU D OS;0 jU ˚ OS;1 jU D OA0 ˚ OA1 D OA ;

where A ´ A0 ˚ A1 . Since OS is a sheaf of superalgebras, A is also a superalgebra

and one can readily check that .U; OS jU / D Spec A. 
Given a superscheme X D .jX j; OX /, let OXr denote the sheaf of algebras

OXr .U / D .OX =JX /.U /;

where JX is the ideal sheaf U 7! JOX .U / with JOX .U / the ideal generated by the odd
nilpotents in OX .U /.
We will call X r D .jX j; OXr / the reduced space associated to the superspace
X D .jX j; OX /. This is a locally ringed space in the classical sense. The scheme
.jX j; OXr / is called the reduced scheme associated to X . Notice that the reduced
scheme associated to a given superscheme may not be reduced, i.e., OXr .U /, U open in
jXj, can contain nilpotents. This is because OXr .U / is obtained by taking the quotient
of OX .U / by the ideal generated only by the odd nilpotents.
176 10 Supervarieties and superschemes

In Section 2.5 of Chapter 2 we have discussed quasi-coherent sheaves in the ordinary

setting and we examined the equivalence of categories between the category of quasi-
coherent sheaves of OA -modules over Spec A and the category of A-modules for an
ordinary commutative algebra A. We want to generalize this picture to the super
setting.
Let A be a commutative superalgebra and let M be an A-module. We can regard M
as an A0 -module and construct the sheaf M z on Spec A0 as in Section 2.5 of Chapter 2.
However, since M is a module for the superalgebra A, we have that the sheaf M z
z
inherits naturally an OA -module structure, i.e., M .U / has an OA .U /-module structure
for all U  Spec A0 and such a structure is compatible with restriction morphisms. We
summarize all the properties of the sheaf M z in a theorem, which is the supergeometric
counterpart of Propositions 2.5.1 and 2.5.4. The proof is a direct check, very similar
to the classical setting, and can be found for example in [43], Ch. II; we leave it to the
reader.

z be as above.
Theorem 10.1.4. Let M be an A-module for a superalgebra A, and let M
Then:
(1) Mz has a natural structure of OA -module.
(2) .Mz /p Š Mp for all p 2 Spec A0 , i.e., the stalk at any prime p of the sheaf M
z
coincides with the localization of the A0 -module M at p.
(3) .Mz /.Spec A0 / D M , i.e., the global sections of the sheaf coincide with the
A-module M .

Again we have as in the classical setting an equivalence of categories:

Proposition 10.1.5. The functor M 7! M z gives an equivalence of categories between

the category of A-modules and the category of quasi-coherent OA -modules. The inverse
of this functor is the functor F 7! F .Spec A0 /.

As in the classical setting we can define the notion of closed subsuperschemes.

Definition 10.1.6. We say that Y D .jY j; OY / is a closed subscheme of a superscheme

X D .jX j; OX / if:
(1) jY j is a closed subset of jX j and Y is a superscheme.
(2) OY D OX = where  is a quasi-coherent sheaf of ideals in OX . In other words,
for each x 2 jXj there exists an open affine set U , x 2 U , such that .U / is an
ideal in OX .U /.

Remark 10.1.7. If X D Spec A, the closed subschemes of X are in one-to-one corre-

spondence with ideals in A as it happens in the ordinary case. Such a correspondence
 10.1 Basic definitions 177

is realized in the following way:

² ³ ² ³
quasi-coherent sheaves two-sided homogeneous
!
of ideals in OA ideals in A

 ! .Spec A0 /
IQ  I:

Here IQ is the quasi-coherent sheaf associated to the A0 -module I .

Examples 10.1.8. (1) Affine superspace Amjn . Consider the polynomial superalgebra
kŒx1 ; : : : ; xm ; 1 ; : : : ; n  over an algebraically closed field k, where x1 ; : : : ; xm are
even indeterminates and 1 ; : : : ; n are odd indeterminates (see Chapter 1). We define
Spec kŒx1 ; : : : ; xm ; 1 ; : : : ; n  to be the affine superspace of superdimension mjn and
denote it by Amjn .
The topological space underlying Amjn is Spec kŒx1 ; : : : ; xm ; 1 ; : : : ; n 0 and con-
sists of the even maximal ideals

.xi  ai ; j k /; i D 1; : : : ; m; j; k D 1; : : : ; n;

and the even prime ideals

.p1 ; : : : ; pr ; j k /; j; k D 1; : : : ; n;

where .p1 ; : : : ; pr / is a prime ideal in kŒx1 ; : : : ; xm . In other words the prime ideals
in kŒx1 ; : : : ; xm ; 1 ; : : : ; n 0 are generated by the prime ideals in kŒx1 ; : : : ; xm  and
the even nilpotent ideal . i j ; i  j /.
At the prime ideal p 2 Spec kŒx1 ; : : : ; xm ; 1 ; : : : ; n 0 the stalk of the structure
sheaf of Amjn is
˚ 
OAmjn ;p D fg j f 2 OAmjn .Am /; g 2 OAmjn .Am /0 ; g … p :

(2) Supervariety over the sphere S 2 . Consider the polynomial superalgebra gener-
ated over an algebraically closed field k, kŒx1 ; x2 ; x3 ; 1 ; 2 ; 3 , and the ideal

I D .x12 C x22 C x32  1; x1  1 C x2  2 C x3  3 /:

Let kŒX D kŒx1 ; x2 ; x3 ; 1 ; 2 ; 3 =I and X D Spec kŒX . Then X is a supervariety

whose reduced variety X r is the sphere S 2 . A maximal ideal in kŒX 0 is given by

m D .xi  ai ; i j/ with i; j D 1; 2; 3; ai 2 k and a12 C a22 C a32 D 1:

The local ring of kŒX 0 at the maximal ideal m0 is the ring of fractions
˚ 
.kŒX 0 /m D fg j f; g 2 kŒX 0 ; g … m :
178 10 Supervarieties and superschemes

The stalk of the structure sheaf at m is the localization of kŒX  as a kŒX 0 -module, that
is,
kŒX m D f mg
j m 2 kŒX ; g 2 kŒX 0 ; g … mg:
Notice that if a1 ¤ 0 (not all ai are zero simultaneously), then x1 is invertible in the
localization and we have
1
1 D .x2 2 C x3 3 /;
x1
so f 2 ; 3 g generate kŒX m as an OkŒX0 -module.
X is an example of a closed subscheme of the affine superspace A3j3 . Notice also
that the reduced scheme corresponding to X is the unitary sphere S 2  A3 .
We are now going to see that the category of affine superschemes is equivalent to
the category of superalgebras; in other words, the superscheme X D Spec A and the
superalgebra A contain the same information. This is a generalization of the equiva-
lence between the category of affine schemes and the category of algebras (for a proof
of the classical result see Chapter 2).
Let .aschemes/ denote the category of affine superschemes. Let us define the functor
F W .salg/op ! .aschemes/ on the objects as F .A/ D Spec A. In order to define F
on the morphisms, let  W A ! B and 0 D jA0 . We need to give a morphism
f D F ./ W Spec B ! Spec A. On the topological space we have immediately
jf j W Spec B0 ! Spec A0 defined as jf j.p/ D 01 .p/. For the sheaf morphism, we
need to give a family of morphisms,
fU W OA .U / ! OB .jf j1 .U //; U open in Spec A0 ;
`
commuting with restrictions. If a 2 OA .U /, i.e., a W U ! x2U OA;x .U / (see
Definition 2.2.6), define fU .a/ W p 7! p .a.jf j.p/// and p W A 1 .p/ ! Bp , p 2
0
Spec B0 . One can check that this determines a sheaf morphism f  W OA ! f OB and
that such f  is local.
We now claim that F gives an equivalence between the category of superalgebras
and the category of affine superschemes. In order to show that F realizes an equivalence
of categories, we need to prove that there is a one-to-one correspondence between
objects and morphisms in the categories. We are going to do this by establishing an
inverse functor G. On the objects the inverse functor is given by
G W .aschemes/ ! .salg/; Spec A 7! OA .A0 / Š A:
On the morphisms G is given as follows. Let f W Spec B ! Spec A be a morphism.
Since the global sections of the structure sheaves coincide with the rings B and A
respectively, we obtain immediately a morphism from A to B, so we set

G.f / D fSpec A0 W OA .Spec A0 / Š A ! OB .Spec B0 / Š B:

We leave it to the reader to verify that F and G are functors and moreover that
F  G Š id and G  F Š id.
We have proven the following proposition.
 10.2 The functor of points 179

Proposition 10.1.9. There exists an equivalence of categories between the category of

commutative superalgebras and the category of affine superschemes.
When we restrict the functor F to the category of affine superalgebras, it gives an
equivalence of categories between affine superalgebras and affine supervarieties.
We now would like to give an example of a not affine superscheme which is of
particular importance: the projective superspace.
Example 10.1.10 (Projective superspace). Let S D kŒx0 ; : : : ; xm ; 1 ; : : : ; m . Then
S is a Z=2Z and Z graded algebra and the two gradings are compatible. Define the
topological space Proj S0 as the set of Z-homogeneous non-irrelevant primes in S0 (i.e.,
those primes not containing .x0 ; : : : ; xm ; i j /), with the Zariski topology. Proj S0 is
covered by open affine sets Ui consisting of those primes not containing .xi /. That is
to say that Ui consists of the homogeneous primes in kŒx0 ; : : : ; xm ; 1 ; : : : ; n Œxi1 
which in turn can be identified with the primes in the ring of elements of degree zero
in kŒx0 ; : : : ; xm ; 1 ; : : : ; n Œxi1 . One can readily see that such ring is isomorphic
to kŒu0 ; : : : ; uO i ; : : : ; um ; 1 ; : : : ; n . Hence, as in the classical setting (see Example
2.3.4), we have
Ui D Spec kŒu0 ; : : : ; uO i ; : : : ; um ; 1 ; : : : ; n 0 ; i D 0; : : : ; m:
We can define the sheaves
OUi D OkŒu0 ;:::;uO i ;:::;um ;1 ;:::;n 
so that .Ui ; OUi / are affine superschemes corresponding to these open affine subsets.
One can check that the conditions of Proposition 2.2.12 are satisfied, hence these
sheaves glue together to give a sheaf OS on all the topological space Proj S. So we
define the projective superspace P mjn , denoted also by Proj S , as the superscheme
.Proj S0 ; OS /.
The same construction can be easily repeated for a generic Z-graded superalgebra.

10.2 The functor of points

As for supermanifolds, we employ the functor of points approach from algebraic ge-
ometry to better handle the nilpotent elements and to recover the geometric intuition.
It is very important to realize that in ordinary algebraic geometry one could introduce
this subject at an elementary level, avoiding altogether this level of abstraction and
deal only with algebraic varieties described by prime ideals in a polynomial algebra
and their regular functions over an algebraically closed field. On the contrary here,
even when the superscheme has a reduced irreducible underlying ordinary scheme (that
is, it corresponds to an algebraic variety), this machinery is unavoidable and often the
functor of points and the representability criterion are the only means we have to handle
these objects in the algebraic setting, even when the field is algebraically closed.
180 10 Supervarieties and superschemes

Definition 10.2.1. For a superscheme X , the functor of points of X is the representable

functor
hX W .sschemes/op ! .sets/; hX .Y / D Hom.Y; X /;
defined as usual on the morphisms as hX ./. / D B ,  W Y ! Z. The elements
in hX .Y / are called the Y -points of the superscheme X.
In the previous chapters, we have mostly used the same notation to denote both a
supergeometric object, say a supermanifold, and its functor of points. In this chapter,
however, we want to make a distinction since we will also deal with non-representable
functors and with representability issues.
As in the ordinary setting, the functor of points of a superscheme is determined by
looking at its restriction to the category of affine superschemes. This result is obtained
in the same way as the corresponding classical one. We briefly sketch the proof and
describe the ideas involved (for more see [29], Ch. VI).
Definition 10.2.2. Let T be a superscheme. We say that fTi g is anSopen cover of T ,
if each Ti is an open subscheme of T and the Ti ’s cover T , that is i jTi j D jT j. We
further say that fTi g is an open affine cover of T , if it is an open cover and each Ti is
affine, that is to say Ti D Spec Ai for superalgebras Ai .
Lemma 10.2.3. Let X and T be superschemes and let fTi g be an affine open cover
of T . Consider a family of morphisms .i /i2I , i 2 hX .Ti /, such that hX .ij /.i / D
hX .j i /.j /, where ij W Ti \ Tj ,! Ti and j i W Ti \ Tj ,! Tj . Then there exists a
unique  2 hX .T / such that hX .i / D i , where i W Ti ,! T .
Notation. In algebraic geometry it is customary to summarize the statement of the
lemma by saying that the following sequence is exact:
Y hX .ij / Y
hX .T / ! hX .Ti / 





 hX .Ti \ Tj /:
hX .j i /

Proof. We have to look for a morphism  W T ! X , which amounts to finding a pair

consisting of a continuous map jj W jT j ! jX j and a sheaf morphism   W OX !
 OT . The hypothesis guarantees that since the continuous maps ji j W jTi j ! jX j
agree on the intersections jTi j \ jTj j, there exists a unique continuous map jj W jT j !
jXj, which is ji j when restricted to the jTi j. For the sheaf morphism the reasoning is
similar. Consider the commutative diagram

Ti \ Tj   / Ti
GG ij
GG
GG
GG i
ij G# 
X.
Clearly there is a corresponding commutative diagram, with the arrows reversed, in-
volving the sheaf morphisms (for clarity we remove the jj that indicates the continuous
 10.2 The functor of points 181

maps between the topological spaces underlying our superschemes):

OTi jTi \Tj .i1 .U // D OTi \Tj .ij1 .U // o OTi .i1 .U //

jUUUU O
UUUU
UUUU
UUUU
U
OX .U /.
We want to define   W OX !  OT , in other words, we want to define for all open
U  jXj U W OX .U / ! OT . 1 .U //.
Clearly the diagram detailed above, for f 2 OX .U /, gives us a family ffi g, fi 2
OTi .i1 .U // D OT jTi .i1 .U // D OT .i1 .U //. The commutativity of the above
diagram ensures that the fi ’s glue together to give f 0 2 OT . 1 .U //, so that we can
define U .f / D f 0 . We leave to the reader the routine checks that show that   so
defined has all the necessary properties.
Uniqueness is clear. 
Proposition 10.2.4. Let X and T be superschemes and let  W T ! X be a morphism.
Then  is determined by its restrictions to an open affine cover of T .
Proof. Let fTi g be an open affine cover of T and let i D jTi . As an exercise one
can check that the i satisfy the hypothesis of the previous lemma, hence we have that
there exists a unique morphism T ! X , whose restriction to each Ti is i . Such a
morphism hence coincides with . 
Theorem 10.2.5. Let X be a superscheme, hX its functor of points. Then hX is
determined by its restriction to the category of affine superschemes.
Proof. Assume that we know hX .Z/ for all affine superschemes Z. Any morphism 
in hX .T / corresponds to a family of morphisms i 2 hX .Ti / for an open cover of T .
Hence it can be uniquely extended to a morphism in hX .T /, necessarily . 
Since we have the contravariant equivalence of categories detailed in Proposi-
tion 10.1.9, we can view the restriction of hX to the category of affine superschemes
also as a functor:
a
hX W .salg/ ! .sets/; a
hX .A/ D Hom.sschemes/ .Spec A; X /:
When the superscheme X is affine, i.e., X D Spec R, hX
a
is representable. In fact
by Proposition 10.1.9,
a
hX .A/ D Hom.sschemes/ .Spec A; Spec R/ D Hom.salg/ .R; A/:
Observation 10.2.6. Since we have the equivalence of categories between affine su-
perschemes and superalgebras, we can define an affine superscheme X D Spec R
equivalently as a representable functor
.salg/ ! .sets/; A 7! Hom.salg/ .R; A/:
a
With an abuse of notation we shall use hX (X D Spec R) to denote this functor as well.
182 10 Supervarieties and superschemes

Remark 10.2.7. To simplify notation we drop the suffix a in hX a

; the context will
make clear whether we are considering hX or its restriction to affine superschemes.
Moreover, whenever we want the restriction of hX to affine superschemes, we shall use
the same symbol, hSpec R , for the two functors

.salg/ ! .sets/; A 7! Hom.R; A/;

.aschemes/ ! .sets/; A ! 7 Hom.Spec A; Spec R/:

Observation 10.2.8. Let X 0 be an affine variety over an algebraically closed field k.

Consider an affine supervariety X whose reduced part coincides with X 0 . Then one
can immediately check that the k-points of X correspond to the points of the affine
variety X 0 . For example, if X D Amjn we have X 0 D Am and, as we shall see in the
next example, hX .k/ D k m D hX 0 .k/.

Examples 10.2.9. (1) Affine superspace revisited. Let A 2 .salg/ and let V D V0 ˚ V1
be a free supermodule (over k). Let .smod/ denote the category of k-modules. Define

V .A/ D .A ˝ V /0 D A0 ˝ V0 ˚ A1 ˝ V1 :

Notice that following a very common abuse of notation we are using the same letter V
to denote both the super vector space V and its functor of points. The abuse will soon
become worse when we shall also use V to denote the superscheme whose functor
of points is V . In general this functor is not representable. However, if V is finite-
dimensional, the functor is indeed representable and in fact we have

.A ˝ V /0 Š Hom.smod/ .V  ; A/ Š Hom.salg/ .Sym.V  /; A/;

where Sym.V  / denotes the symmetric algebra over the dual space V  . Recall that V 

V maps V !
is the set of linear

V k not necessarily preserving the parity, and Sym.V / D
Sym.V0 / ˝ V1 , where V1 denotes the exterior algebra over the ordinary space
V1 .
Let us fix a basis for V and let dim V D pjq. The functor V is represented by

kŒV  D kŒx1 ; : : : ; xp ; 1 ; : : : ; q ;

where xi and j are respectively even and odd indeterminates.

Hence the functor V is the functor of points of the affine supervariety Amjn intro-
duced in Example 10.1.8 (1).
We also want to remark that the functor DV defined as

DV .A/ ´ Hom.smod/ .V; A/

is representable for any V (not necessarily finite-dimensional), and it is represented by

the superalgebra Sym.V /. Clearly V Š DV when V is finite-dimensional.
 10.3 A representability criterion 183

(2) Supermatrices revisited. Let A 2 .salg/. Define Mmjn .A/ as the set of endo-
morphisms of the A-supermodule Amjn . Choosing coordinates we can write
² ³
a ˛
Mmjn .A/ D ;
ˇ b
where a, b are m  m, n  n blocks of even elements in A and ˛, ˇ are m  n, n  m
blocks of odd elements in A.
This is the functor of points of an affine supervariety represented by the commutative
superalgebra: kŒM.mjn/ D kŒxij ; kl  where xij ’s and kl ’s are respectively even
and odd variables with 1  i; j  m or m C 1  i; j  m C n, 1  k  m,
m C 1  l  m C n or m C 1  k  m C n, 1  l  m.
Notice that Mmjn Š hAm2 Cn2 j2mn .
Now consider the invertible matrices in Mmjn .A/. This clearly gives us another
functor
.salg/ ! .sets/; A 7! invertible matrices in Mmjn .A/:
As we shall see in the next chapter, this is the functor of points of the superscheme
GLmjn introduced in Example 3.3.4, though this fact is far from being immediate.
This example raises a very natural question: how can we determine whether a func-
tor F W .salg/ ! .sets/ is the functor of points of a superscheme? We can reformulate
this question in an equivalent way: can we give a representability criterion for functors
.sschemes/ ! .sets/? These questions have a positive answer in the classical setting
and a comprehensive treatment can be found in [23], Ch. I, and in [29], Ch. VI. In the
next section we are going to see how the classical argument can be replicated, with
small changes, in the supergeometric environment.

10.3 A representability criterion

We now want to single out among all the functors F W .salg/ ! .sets/ those that are the
functor of points of superschemes. As we shall presently see, they are characterized by
two key properties. The first property is locality, which is the analogue of the gluing
property in the sheaf definition. Local functors are in fact also-called Zariski sheaves in
the literature. The word Zariski refers to the fact that essentially we are considering the
Zariski topology on the spectrum of the rings.1 The second property is more delicate
and corresponds to the fact that these functors locally look like the functor of points of
affine superschemes.
Let us recall some notation from Chapter 2, Section 2.3.
Let A be a superalgebra. Given f 2 A0 , let Af denote
Af D AŒf 1  ´ fa=f n j a 2 Ag:
1
There are other relevant topologies for this theory, but we are unable to treat them in our text.
184 10 Supervarieties and superschemes

The sets Uf D Spec.Af /0 D fp 2 Spec A0 j.f / 6 pg are the basic open sets in
the topological space X D Spec A0 (see Chapter 3). In fact by definition the open sets
in the Zariski topology of Spec A0 are

UI D fp 2 Spec A0 j I 6 pg

for all the ideals I in A0 .

Let us start with the notion of local functor.

Definition 10.3.1. Let F W .salg/ ! .sets/ be a functor. Let ffi gi2I be a family of
elements in A0 , .fi ; i 2 I / D A0 and let i W A ! Afi and ij W Afi ! Afi fj be the
natural morphism. We say that the functor F is local or that it is a sheaf in the Zariski
topology if for any family f˛i gi2I , ˛i 2 F .Afi /, such that F .ij /.˛i / D F .j i /.˛j /
for all i and j , there exists a unique ˛ 2 F .A/ with F .i /.˛/ D ˛i .

As we shall presently see, this is a far reaching generalization of the concept of

sheaf on a topological space, as we have described it in Chapter 2.
Let F be a local functor. By Observation 10.2.6, we can regard it as a functor
F W .aschemes/op ! .sets/, via the equivalence of categories described in Proposi-
tion 10.1.9. Let FA denote its restriction to the affine open subschemes of Spec A.
Since these subsets form a basis for the topology on Spec A, we are going to see that
FA is a B-sheaf in the sense specified in Chapter 2, where we forget the scheme struc-
ture on the affine open subschemes of Spec A and treat them just as open sets in the
topological space Spec A, their morphisms being the inclusions.
Since FA is a functor it is automatically a presheaf in the Zariski topology. If
U D Spec O.U / and V D Spec O.V / are two open affine sets in the topological space
Spec A, the very definition of FA , as a functor from the category of the open affine sets
in Spec A to .sets/, tells us that FA is a presheaf of sets. In fact, if i W U ,! V is the
inclusion of two open sets, one sees immediately that FA .i / W FA .V / ! FA .U / satisfies
the properties of the restriction morphism, customarily denoted as ˛jU D FA .i /.˛/,
for ˛ 2 FA .V /.
We are now going to see that the locality of F guarantees the gluing of any family of
local sections which agree on the intersection of any two open sets of an open covering.
S
Take an open affine covering fUfi gi2I of Spec A, that is, Spec A0 D Spec Afi ;0 ,
Ufi D Spec Afi ;0 . We can assume that for our problem any open covering to be of
this kind without any loss of generality since in the topological space SpecS A0 , fUf D
Spec Af gf 2A0 is a base for the open sets. The condition Spec A0 D Spec Afi ;0 is
equivalent to saying that .fi ; i 2 I / D A0 . The condition that F is local can hence
be translated in this context in the following way. Given any family f˛i gi2I with ˛i 2
FA .Ufi /, such that ˛i jUfi \Ufj D ˛j jUfi \Ufj , there exists a unique ˛ 2 FA .Spec A0 /
such that ˛jUfi D ˛i (where, again, we denote by ˛i jUfi \Ufj the image of ˛i under
FA .ij /, where ij W Ufi \ Ufj ,! Ufi ). This is precisely the condition to impose on
the presheaf FA in order to have a B-sheaf (see Chapter 2).
 10.3 A representability criterion 185

The examples of local functors are many; the one that interests us most is the
following.
Proposition 10.3.2. The functor of points hX of a superscheme X is local.
Proof. We briefly sketch the proof since it is the same as in the ordinary case. Let the
notation be as above. Consider a collection of maps ˛i 2 hX .Afi / which map to the
same element in hX .Afi fj /. Each ˛i consists of a continuous map j˛i j W Spec Afi 0 !

jXj and a family of ˛i;U W OX .U / ! OAfi .j˛j1 .U // respecting restrictions. The fact
the j˛i j glue together is clear. The gluing of the ˛i ’s to give ˛ W Spec A ! X depends
on the fact that OX and OAfi are sheaves. 
We now want to define the second ingredient for our representability criterion,
namely the notion of open subfunctor of a functor F W .salg/ ! .sets/. If we assume
F to be the functor of points of superscheme X, there is a sensible notion of an open
subfunctor of F D hX , namely, we could simply define it as the functor of points
of an open superscheme U  X . However since we are precisely interested in a
characterization of those F that come from superschemes, we have to carefully extend
the notion of open subfunctor to subfunctors of functors which do not come necessarily
as the functor of points of a superscheme.
Definition 10.3.3. Let U be a subfunctor of a functor F W .salg/ ! .sets/; this means
that we have a natural transformation U ! F such that U.A/ ! F .A/ is injective for
all A. We say that U is an open subfunctor of F if for all A 2 .salg/, given any natural
transformation f W hSpec A ! F , the subfunctor f 1 .U / coincides with hV for some
open V in Spec A, where

f 1 .U /.R/ ´ fR1 .U.R//; fR W hSpec A.R/ ! F .R/:

We say U is an open affine subfunctor of F if it is open and representable, in other

words U Š hSpec R for some superalgebra R.
Very naturally, an open subfunctor of the functor of points of a superscheme X is
hU where U is open in X, as we shall see in the next observation.
Observation 10.3.4. Let X D .jX j; OX / be a superscheme and U  X open affine
in X. Then hU is an open affine subfunctor of hX .
By Yoneda’s lemma f W hSpec A ! hX corresponds to a map f 0 W Spec A ! X .
Let V D f 01 .U / open in Spec A. We claim

fR1 .hU .R// D hV .R/ for all R 2 .salg/:

Let  2 hSpec A .R/, then fR ./ D f 0   2 hX .R/.

Hence if fR ./ 2 hU .R/ immediately we have that  factors out as:

 W Spec R ! V D f 1 .U / ! Spec A:
186 10 Supervarieties and superschemes

So fR ./ 2 hU .R/ if and only if  2 hV .R/. We leave to the reader all the routine
checks.
We want to define the notion of an open cover of a functor.
Definition 10.3.5. Let F W .salg/ ! .sets/ be a functor. F is covered by the open
subfunctors .Ui /i2I , if and only if for any affine superscheme Spec A and natural
transformation f W hSpec A ! F we have that the fibered product hSpec A F Ui Š hVi
and .Vi /i2I is an open cover of Spec A. (For the definition of fibered product see the
Appendix B).
Notice that, by the very definition of open subfunctor, the functor hSpec A F Ui is
always representable. In fact, it is equal to f 1 .Ui / which is by definition the functor
of points of an open and affine Vi in Spec A. Moreover, as before, we have that if F is
the functor of points of a superscheme X , i.e., F D hX , the family of open subfunctors
S
.Ui /i2I covers F if and only if Ui D hWi and the Wi cover X , that is jX j D jWi j.
S
Remark 10.3.6. Notice also that asking F .A/ D Ui .A/ for all superalgebras A is
far too restrictive. This is already a phenomenon we observe at the ordinary level. For
example let us consider F D hSpec Z . F admits fU1 D hSpec ZŒ1=2 ; U2 D hSpec ZŒ1=3 g
as open affine cover, in fact fSpec ZŒ1=2; Spec ZŒ1=3g cover Spec Z. However
F .Z/ ¤ U1 .Z/ [ U2 .Z/ since F .Z/ D id, while U1 .Z/ D U2 .Z/ D ;.
We are ready to state the main result of this section that allows us to characterize,
among all the functors from .salg/ to .sets/, those which are the functors of points of
superschemes.
Theorem 10.3.7. A functor
F W .salg/ ! .sets/
is the functor of points of a superscheme X , i.e., F D hX if and only if
(1) F is local,
(2) F admits a cover by affine open subfunctors.
Proof. The proof of this result is similar to that in the ordinary case detailed in [23],
Ch. I, §1, 4, 4.4, but given its importance we shall rewrite it. We first observe that if
hX is the functor of points of a superscheme, by Proposition 10.3.2 it is local and by
Definition 10.3.5 it admits a cover by open affine subfunctors.
Let us now assume that F satisfies the properties (1) and (2) of Theorem 10.3.7. We
need to construct a superscheme X D .jX j; OX / such that hX Š F . The construction
of the topological space jX j is the same as in the ordinary case. Let us sketch it.
Let fhX˛ g˛2A be the affine open subfunctors that cover F .2 Define hX˛ˇ D hX˛ F
hXˇ , (X˛ˇ will correspond to the intersection of the two open affine X˛ and Xˇ in
2
We are now assuming that the open affine subfunctors of F are equal and not just isomorphic to hX˛ .
We leave to the reader the checks necessary to verify all the statements in the slightly more general setting
required by the theorem.
 10.3 A representability criterion 187

the superscheme X , that we are now constructing). Notice also that hX˛ F hXˇ is
representable, as we remarked after Definition 10.3.5.
We have the commutative diagram

jˇ;˛
hX˛ˇ D hX˛ F hXˇ / hXˇ

j˛;ˇ iˇ
 i˛ 
hX˛ / F.

As a set we define `
jX j ´ jX˛ j= ;
˛2A

where is the following relation: if x˛ 2 jX˛ j, xˇ 2 jXˇ j then x˛ xˇ if and

only if there exists x˛ˇ 2 jX˛ˇ j such that jj˛;ˇ j.x˛ˇ / D x˛ and jjˇ;˛ j.x˛ˇ / D
xˇ . Notice that we are (improperly) using the symbol j˛;ˇ also for the superscheme
morphism j˛;ˇ W X˛;ˇ ! X˛ . As one can check, this is an equivalence relation, jX j is
a topological space and ˛ W jX˛ j ,! jX j is an open continuous injective map.
We now need to define a sheaf of superalgebras OX . We shall use the sheaves in
the open affine X˛ ’s, “gluing” them appropriately.
Let U be open in jX j and let U˛ D ˛1 .U /. Put
˚ ` 

OX .U / ´ .f˛ / 2 ˛2A OX˛ .U˛ / j jˇ; .fˇ / D j ;ˇ .f / for all ˇ;  2 A :

The condition jˇ; .fˇ / D j ;ˇ .f / simply states that to be an element of OX .U /, the
collection ff˛ g must be such that fˇ and f agree on the intersection of Xˇ and X
for any ˇ and  . One can check that OX is a sheaf of superalgebras.
Hence we have defined a superspace X D .jX j; OX /. X is a superscheme since,
as a superspace, it admits a cover by open affine subschemes whose functor of points
are the X˛ ’s.
To finish the proof, we need to show that hX Š F . We are looking for a functorial
bijection between hX .A/ D Hom.sschemes/ .Spec A; X / and F .A/ for all A 2 .salg/. It
is here that we use the hypothesis of F being local.
We first construct a natural transformation A W F .A/ ! hX .A/.
Let t 2 F .A/ D Hom.hSpec A ; F /, by Yoneda’s lemma. Consider the diagram

hT˛ ´ hX˛ F hSpec A / hSpec A

t˛ t
 i˛ 
hX˛ / F.

Since the hX˛ are an open affine cover of F , the fT˛ g form an open affine cover
of Spec A. Since by Yoneda’s lemma Hom.hT˛ ; hX˛ / Š Hom.T˛ ; X˛ /, we obtain a
188 10 Supervarieties and superschemes

morphism t˛ W T˛ ! X˛  X . One can check that the morphisms t˛ glue together to

give a morphism t 0 W Spec A ! X , hence t 0 2 hX .A/. So we define A .t / D t 0 .
Next we construct another natural transformation A W hX .A/ ! F .A/, which turns
out to be the inverse of .
Assume that f 2 hX .A/, i.e., f W Spec A ! X . Let T˛ D f 1 .X˛ /. By our
hypothesis, the T˛ ’s form a cover of Spec A that we can assume to be affine. We
immediately obtain morphisms g˛ W T˛ ! X˛ . By Yoneda’s lemma, g˛ corresponds
to a natural transformation g˛ W hT˛ ! hX˛ . Since F is local, the morphisms i˛ B
g˛ W hT˛ ! hX˛ ! F glue together to give a morphism g W hSpec A ! F , i.e., an
element g 2 F .A/. Define A .f / D g.
One can directly check that and  are indeed natural transformations, moreover
B  and  B are the identity so that and  are inverse to each other. Hence F Š hX .

An important consequence of this theorem is the existence of fibered products in
the category of superschemes. We start with a lemma for the affine case.
Lemma 10.3.8. Let X D Spec O.X /, Y D Spec O.Y /, Z D Spec O.Z/. Then the
fibered product X Z Y is an affine superscheme and

X Z Y D Spec O.X / ˝O.Z/ O.Y /:

Proof. The proof is a simple exercise that makes use of Proposition 10.1.9 and the
universal properties of the tensor product and the fibered product. 
Corollary 10.3.9. Fibered products exist in the category of superschemes. The fibered
product X Z Y , for superschemes X , Y , Z,

X Z Y /Y

g
 f 
X / Z,

is the superscheme whose functor of points is hX hZ hY .

Proof. We sketch the argument, since it is the same as in the ordinary case, found for
example in [23] I §1, 5.1. Let F W .salg/ ! .sets/ be the functor F .A/ D hX .A/hZ .A/
hY .A/. We want to show that F is the functor of points of a superscheme. We shall
use Theorem 10.3.7. The fact that F is local, comes from its very definition and we
leave it to the reader as an exercise. It is based on the fact that hX , hY and hZ are all
local.
We now want to show that it can be covered by open affine subfunctors. Let
fZi g be an affine open covering of Z. Define Xi D X Z Zi D f 1 .Zi / and
Yi D Y Z Zi D g 1 .Zi /. Notice that we know that Xi and Yi are superschemes
since they are open subschemes of X and Y , respectively. Let Xi˛ and Yjˇ be open
 10.3 A representability criterion 189

affine covers of Xi and Yj respectively. A straightforward check shows that both

hXi ˛ hZ hY and hX hZ hYjˇ are open subfunctors of F . Moreover, since

hXi ˛ .A/ hZi .A/ hYjˇ .A/ D .hXi ˛ .A/ hZ .A/ hY .A// \ .hX .A/ hZ .A/ hYjˇ .A//;

also hXi ˛ hZi hYjˇ is an open subfunctor of F . Hence fhXi ˛ hZi hYjˇ g form an open
covering of F by open subfunctors. These functors are affine since if Xi˛ D Spec Ri˛ ,
Yjˇ D Spec Sjˇ and Zi D Spec Ti , hXi ˛ hZi hYjˇ D hSpec Ri ˛ ˝Ti Sjˇ
Hence by the representability criterion, Theorem 10.3.7, F is the functor of points
of a superscheme, which is X Z Y , as one can readily check. 
We end this section with some remarks on an equivalent approach to the problem
discussed in Theorem 10.3.7.
As we have seen in the previous section, the functor of points of a superscheme X
can be also equivalently defined as the functor hX W .sschemes/op ! .sets/, hX .T / D
Hom.T; X/. Hence the question whether a generic functor F W .sschemes/op ! .sets/
is (isomorphic to) the functor of points of a superscheme, can be reformulated by asking
the conditions for F to be representable.
In such a setting, we can give an equivalent definition for F to be local, formulated
as a proposition, whose proof is a straightforward check, based on the results of the
previous section.
Let F W .sschemes/op ! .sets/ be a functor. Since the category of affine super-
schemes is equivalent to the category of commutative superalgebras, we can regard
F restricted to the category of affine superschemes as a functor from the category of
superalgebras to the category of sets. We shall call such functor F a .
Proposition 10.3.10. Let F W .sschemes/op ! .sets/ be a functor. Then F a is local if
and only if the following condition is satisfied. Let X be a superscheme, Xi an open
cover of X. Assume that there is a family of elements fi 2 F .Xi / such that fij D fj i
for all i; j , where fij is the image of fi under the natural map F .Xi / ! F .Xi \ Xj /
induced by the inclusion Xi \Xj  Xi . Then there exists a unique f 2 F .X / mapping
to each fi under the natural map F .X / ! F .Xi /.
The analogue of the notion of an open affine cover of such F (see Definition 10.3.5)
makes sense as it is in this setting, hence we are ready for the Representability Theorem.
Theorem 10.3.11 (Representability criterion for superschemes). A functor

F W .sschemes/op ! .sets/

is representable if and only if F is local, and F is covered by open affine subfunctors.

The proof is very similar to the proof of Theorem 10.3.7 and actually we have seen
it in the context of differentiable supermanifolds in Theorem 9.4.3. Since the checks
involved are very formal, that proof holds practically with no changes here and for this
reason we leave it to the reader as an exercise.
190 10 Supervarieties and superschemes

10.4 The Grassmannian superscheme

In this section we want to discuss the Grassmannian superscheme corresponding to the
rjs-dimensional superspaces inside a super vector space of dimension mjn, r < m,
s < n. We will show that it is a superscheme using the representability criterion
Theorem 10.3.7. This is a particularly important example since it is the first non-trivial
example of a non-affine superscheme.
Consider the functor Gr W .salg/ ! .sets/, where for any superalgebra A, Gr.A/
is the set of projective A-submodules of rank rjs of Amjn (for the definition of the
rank of a projective A-module see Appendix B.3). When A is a local superalgebra,
i.e., it contains a unique maximal ideal, one sees immediately from the definitions, that
Gr.A/ consists of the free A-submodules of dimension rjs of the free module Amjn . For
example if A is a field, say A D k, Gr.k/ are the super vector subspaces of dimension
rjs inside k mjn , thus recovering the more familiar notion of Grassmannian, as the set
of subspaces of a given dimension inside a vector space.
Equivalently, Gr.A/ can also be defined as the set Gr.A/ D f˛ W Amjn ! Lg,
where ˛ is a surjective morphism, L is a projective A-module of rank rjs, modulo the
following equivalence relation: ˛  ˛ 0 if and only if they have the same kernel. Notice
that in this definition L also varies.
We need also to specify Gr on morphisms W A ! B.
Given a morphism W A ! B of superalgebras, as usual we can provide B with
the structure of an A-module in a natural way by setting
ab D .a/b; a 2 A; b 2 B:
Also, given an A-module L, we can construct the B-module L ˝A B. So given and
the element of Gr.A/, f W Amjn ! L, we can define an element of Gr.B/ as follows:
Gr. /.f / W B mjn D Amjn ˝A B ! L ˝A B
where L ˝A B is clearly a projective B-module of rank rjs. We want to show that Gr
is the functor of points of a superscheme.
We will start by showing it admits a cover of open affine subfunctors. Consider
the multi-index I D .i1 ; : : : ; ir j 1 ; : : : ; s / and the map I W Arjs ! Amjn , where
I .x1 ; : : : ; xr j 1 ; : : : ; s / with 1  i1 <    < ir  m, 1  1 <    < s  n is the
mjn-uple with x1 ; : : : ; xr occupying the positions i1 ; : : : ; ir and 1 ; : : : ; s occupying
the positions 1 ; : : : ; s and all the other positions are occupied by zero. For example,
let m D n D 2 and r D s D 1. Then 1j2 .x; / D .x; 0j0; /.
Now define the subfunctors vI of Gr as follows. The vI .A/ are the morphisms
˛ W Amjn ! L, i.e., the elements in Gr.A/ such that ˛ B I is surjective. Notice that
if ˛ B  W Arjs ! Amjn ! L is surjective, since Arjs and L are projective modules
of the same rank, ˛ B  is an isomorphism, hence L is free. This is crucial for the
representability of vI and ultimately of Gr.
The functor vI is in fact representable: vI Š hM.mjn/.mrjns/ , that is, vI is
isomorphic to the functor of points of ..mjn/  .m  rjn  s//-matrices. In fact
 10.4 The Grassmannian superscheme 191

vI .A/ consists of morphisms ˛ W Amjn ! L Š Arjs , where we specify the images of

rjs elements in the canonical basis of Amjn . We leave to the reader the easy checks
involved.
We want now to show that the vI are open affine subfunctors of Gr. The condition
that vI is an open subfunctor is equivalent to asking that f 1 .vI / be open for any
morphism f W Spec A ! Gr. So we fix one of such morphisms f .
By Yoneda’s lemma, a map f W Spec A ! Gr corresponds to a point f in Gr.A/.
We need to find an open subscheme VI in Spec A such that

hVI .B/ D f 1 .vI .B// D f W A ! B j fB . / 2 vI .B/g  hSpec A .B/

for all B 2 .salg/, where fB W hSpec A .B/ ! Gr.B/. Notice that fB . / D f B # ,

with # W Spec B ! Spec A is the morphism of affine schemes induced by (we
prefer # instead of the more cumbersome notation Spec ).
To show that VI is open, we fix the canonical basis in Arjs and we define

ZA;I .f / ´ f  I 2 Hom.Arjs ; L/:

We want to determine which conditions must be satisfied to have f 2 vI .A/.

Since vI is local (it is representable), we can assume without loss of generality that A
is local (see also Appendix B), so that ZA;I .f / 2 Mrjs .A/. In this case we have that
f 2 vI .A/ if and only if ZA;I .f / is invertible. By Chapter 1, Section 1.5, we know
that this is equivalent to having the determinants, say d1 and d2 , of the diagonal blocks
of ZA;I .f /, invertible. This is an open affine condition, but let us look at it in more
detail.
We have f 2 vI .A/ if and only if d1 and d2 are invertible in A. Let now
2 hSpec A .B/, i.e., W A ! B. We have 2 f 1 .vI .B// if and only if factors via
AŒd11 ; d22 , in other words W A ! AŒd11 ; d21  ! B. So 2 hSpec AŒd 1 ;d 2  .B/.
1 2
This shows that the vI ’s are open and affine. It remains to show that these subfunc-
tors cover Gr.
Given f 2 Gr.A/, that is a function from hSpec A ! Gr, we have that since f is
surjective, there exists at least an index I so that ZA;I .f / is invertible, hence f 2 vI .A/
for this I . The above argument shows also that we obtain a cover of any Spec A by
taking vI Gr hSpec A .
Next we want to show that Gr is local. We shall do this by identifying Gr.A/ with
locally free coherent sheaves of rank rjs. By Appendix B.3, we have that any projective
module is locally free and moreover, by Theorem 10.1.4, one obtains the functorial cor-
respondence between (locally free) coherent sheaves and finitely generated (projective)
modules in the super-setting. Hence we have:

Gr.A/ Š fF  OAmjn j F is a subsheaf of locally constant rank rjsg;

where OAmjn D OA ˝ k mjn .

192 10 Supervarieties and superschemes

By its very definition this functor is local. This is perhaps best seen by using the
correspondence detailed in Proposition 10.3.10:

Gr.T / Š fF  OTmjn j F is a subsheaf of locally constant rank rjsg:

In other words, we are considering the functor Gr extended to the category of su-
perschemes: Gr W .sschemes/op ! .sets/ (by a common abuse of notation we use
the same symbol to denote it). If fTi g is an open cover of the superscheme T , as-
sume that we have a family of subsheaves Fi 2 Gr.Ti /, Fi  OTmjn i
D OTmjn jTi ,
with Fi jTi \Tj D Fj jTi \Tj . Then it is clear the Fi ’s glue together to give a subsheaf
F  OTmjn on T , still locally constant of rank rjs, hence F 2 Gr.T /.
We have shown that Gr is the functor of points of a superscheme that we will call
the Grassmannian superscheme (or superGrassmannian for short) of rjs subspaces in
an mjn-dimensional space.

10.5 Projective supergeometry

The aim of this section is to give a brief introduction to projective supergeometry and
to show that the Grassmannian superscheme described in the previous section is not
projective, that is, it is not a subscheme of the projective superspace. This section is
independent from the rest of our work and can be skipped in a first reading. Also:
in this section we assume that the reader is comfortable with some basic algebraic
geometry constructions, such as line bundles and cohomology groups. Such notions
will not appear elsewhere in our work.
Let S be a Z-graded superalgebra, with the Z-grading compatible with the super
grading. Denote by S n the subalgebra of elements with degree n. Assume further
that S is generated by S 1 , the subalgebra the elements of degree 1. S is a Z-graded
S -module in a natural way. We can define, very much in the same way as the ordinary
setting, the twisted module S.n/ as

S.n/d D S nCd :

Clearly also S.n/ is an S -module (recall that upper indices refer to the Z-grading while
lower indices to the Z2 -grading).
Let M be a graded S-module. Then M is also an S0 -module, hence we can follow
z . One
the classical recipe (see Chapter 2, Section 2.5) and build the sheaf on Proj S0 , M
can check right away that M z .U / is an OS .U /-module for all open sets U in Proj S0 .
We summarize the properties of M z in the following theorem.

Theorem 10.5.1. Let M be an S-module, for a Z-graded superalgebra S . Then:

z has a natural structure of OS -module.
(1) M
 10.5 Projective supergeometry 193

(2) .Mz /p Š Mp for all homogeneous prime ideals p 2 Proj S0 , i.e., the stalk at any
prime p of the sheaf Mz coincides with the localization of the S0 -module M at p.
(3) For all homogeneous f 2 S n , M f
A
z jU D M.f / , where Uf is the basic open
set in Proj S0 corresponding to f , M.f / are the elements of degree zero in the
localization Mf (this is commonly called projective localization).

Proof. We leave the proof to the reader as an exercise: it is done precisely as in the
ordinary setting that is found in [43], Ch. II, Section 5. 

At this point one could define supercoherent sheaves as sheaves modeled after M z,
in analogy to what we have done in Capter 2, but we shall not take this direction.
We now turn to the generalization of the Serre’s twisting sheaf in order to charac-
terize the morphisms from a superscheme to the projective superspace.
e
Definition 10.5.2. We define OS .n/ as the sheaf S.n/. Note that OS .1/ is called
Serre’s twisting sheaf.

OS .n/ is a locally free OS -module of rank 1. The proof is essentially the same as
the classical one (see [43], Ch. II, Section 5).

Definition 10.5.3. Let X be a superscheme. A super line bundle on X is a locally free

OX -module of rank 1. Moreover we say that a super line bundle L is generated by the
global sections s1 ; : : : ; sm , 1 ; : : : ; n if the images of such global sections in the stalk
Lp generate the stalk for all p 2 jX j.

For example, for S D kŒx0 ; : : : ; xm ; 1 ; : : : ; n  consider the line bundle O.1/ ´

OS .1/ on P mjn D Proj S. As one can readily check, this super line bundle is generated
by the global sections x0 ; : : : ; xm , 1 ; : : : ; n .
The next proposition relates the morphisms of a superscheme X into the projective
superspace P mjn , with the set of global sections of super line bundles.

Proposition 10.5.4. Let X be a superscheme.

(1) For any morphism  W X ! P mjn ,   .O.1// is a super line bundle on X and is
generated by the global sections   .xi /,   . j /, i D 1; : : : ; m, j D 1; : : : ; n.
(2) Vice versa, if L is a super line bundle on a superscheme X and if s0 ; : : : ; sm ,
1 ; : : : ; n are global sections generating L, then there exists a unique morphism
 W X ! P mjn such that L D   .O.1// and si D   .xi /, j D   . j /, i D 0; : : : ; m,
j D 1; : : : ; n.

Proof. This is an exercise, once one follows the proof for the ordinary setting; see [43],
Ch. II, Section 7. 

For a generic sheaf of rings F , let F  denote the sheaf of invertible sections, i.e.,
F .U / are the invertible sections in F .U /. The first cohomology group H 1 .jX0 j; OX0 /


for an ordinary scheme X0 classifies the equivalence classes of line bundles on X0 ([43],
194 10 Supervarieties and superschemes

Ch. III, Section 4). The next proposition tells us that the same happens also for super
line bundles. We leave the proof of this result to the reader since it is a precise replica
of the classical one.

Proposition 10.5.5. Let X D .jX j; OX / be a superscheme. There is a one-to-one

correspondence between the equivalence classes of super line bundles on jX j and the
elements in the cohomology group H 1 .jX j; OX /.

We now turn to the main goal of this section: we want to show that the Grassmannian
superscheme cannot in general be embedded into any projective superspace. This
statement appears for the analytic category in [56]; the reasoning here is the same, we
include it for completeness. We start with some general considerations on super line
bundles on superschemes.
Let X be a superscheme and let X0 be its reduced scheme, i.e., X D .jX j; OX /,
X0 D .jXj; OX0 / D .jX j; OX =/ where  is the ideal sheaf generated by the odd
nilpotents.
Consider the exact sequence of sheaves

0 !  ! OX ! OX0 ! 0:

We can construct the usual long exact sequence:

0 ! H 0 .jX j; / ! H 0 .jX j; OX / ! H 0 .jX j; OX0 / ! H 1 .jX j; /

! H 1 .jX j; OX / !
 H 1 .jX j; OX0 / 
! H 2 .jX j; / !    :

As we remarked previously, H 1 .jX j; OX / is in one-to-one correspondence with the

super line bundles on jX j, while H 1 .jX j; OX0 / is in one-to-one correspondence with
the line bundles on jX0 j D jX j. The map  tells us which super line bundles on jX j
restrict properly to line bundles on jX0 j. We are going to show that, for a specific
Grassmannian superscheme X , the super line bundles in H 1 .jX j; OX / restrict via
 only to line bundles in H 1 .jX j; OX0 / which have no global sections. Since the
projective morphisms are given via global sections of super line bundles, this will
show that X does not admit any projective embedding.
Let us take X D Gr.1j1; 2j2/ so that X0 D P 1  P 1 with projective coordinates
Œx0 ; x1  and Œy0 ; y1  for the first and second copy of P 1 . Then jX j D jX0 j has the
following open covering:

jU00 j D f.Œ1; u0 ; Œ1; v0 /g; jU10 j D f.Œu1 ; 1; Œ1; v0 /g;

jU01 j D f.Œ1; u0 ; Œv1 ; 1/g; jU11 j D f.Œu1 ; 1; Œv1 ; 1/g:

Consider the functor of points of X on local superalgebras (that is, superalgebras

which have only one maximal homogeneous ideal) since by the arguments in Ap-
pendix B.2, we know such superalgebras completely determine the functor of points.
 10.5 Projective supergeometry 195

X is the quotient of GL2j2 by the parabolic subgroup

0 1
B0 0 C
P .A/ D B
@
C  GL2j2 .A/;
A
0 0

which is the stabilizer of the element he1 ; 1 i 2 X.A/. This leads to the usual identifi-
cation of points in X.A/ with matrices in GL2j2 .A/=P .A/.
The open covering of jX j D P 1  P 1 described above induces the following open
covering of X by the open affine subfunctors Uij :
80 19 80 19
ˆ
ˆ 1 0 >
> ˆ
ˆ u1  >
>
<B C= <B C=
B u0  C B 1 0 C
U00 .A/ D @ A> ; U10 .A/ D ˆ@ 0 A> ;
ˆ 0 1 > 1 >
:̂ ; :̂ ;
 v0  v0
80 19 80 19
ˆ
ˆ 1 0 >
> ˆ
ˆ u1  >
>
<B C = <B C =
B u0  C B 1 0 C
U01 .A/ D @ A> ; U .A/ D A> :
ˆ v1 >
11
ˆ@ v1 >
:̂ ; :̂ ;
0 1 0 1
Assume that we have a line bundle on X and let us see how sections transform
under coordinate changes. Generic sections on jU00 j and jU11 j respectively are given
by

U00 ! U00  C; .u0 ; v0 ; ; / 7! ..u0 ; v0 ; ; /; s.u0 ; v0 ; ; //;

U11 ! U11  C; .u1 ; v1 ; ; / 7! ..u1 ; v1 ; ; /; t .u1 ; v1 ; ; //:

Let us see the behaviour of the transition function on the intersection jU00 j \ jU11 j. A
straightforward calculation shows that (see also Chapter 1, Section 1.5):
 1  
u0  .u0  v01 /1 u1 .v0  u1 /1 /
D 0 0 :
 v0 v01 .u0  v01 /1 .v0  u1
0 /
1

This allows us to view a generic element in U00 .A/ as an element in U11 .A/:
0 1 0 1
1 0 .u0  v01 /1 u1 1 1
0 .v0  u0 /
Bu0  C B 1 0 C
B C 7! B 1 C:
@0 1 A @v0 .u0  v01 /1 .v0  u1
0 /
1 A
 v0 0 1
We have that sections transform in the following way:

t.u1 ; v1 ; ; / 7!x  t ..u0  v01 /1 ; v01 .u0  v01 /1 ;

 u1 1 1 1 1
0 .v0  u0 / ; .v0  u0 / / D s.u0 ; v0 ; ; /;
196 10 Supervarieties and superschemes

where x is an invertible element in jU00 j \ jU11 j. This means that in order to have a
polynomial function the invertible element x must contain a power of the Berezinian.
When we consider this transition function, together with all the requirements of the
other variables, which we do not compute here, one sees that u0 and v0 appear with
exponents with opposite sign. This means that one copy of P 1 can get embedded (the
one corresponding to the positive sign) while the other cannot.
With this particular affine covering, X cannot have any projective embedding since
its reduced variety X0 does not admit such embedding.
In the course of our calculation, we have computed the generic element in
H 1 .U; OX / using the covering U D fU00 ; U01 ; U10 ; U11 g of jX j and Čech coho-
mology. Since H 1 .U; OX / D H 1 .jX j; OX /, the argument we give for the special
covering is actually a generic argument and shows that the image of the map  consists
only of line bundles with no global sections, hence they will not correspond to line bun-
dles that classically projectively embed into X0 D P 1  P 1 . Hence X D Gr.1j1; 2j2/
does not admit any projective embedding.

10.6 The infinitesimal theory

In this section we want to discuss the infinitesimal theory of superschemes, that is
we define the notion of tangent space to a superscheme and to a supervariety at a
point of the underlying topological space. We then use these definitions in explicit
computations.
Let k be a field. We want to restrict our attention to the algebraic superschemes.
Definition 10.6.1. Let X D .jXj; OX / be a superscheme (supervariety). We say that
X is algebraic if it admits an open affine finite cover fXi gi2I such that OX .Xi / is a
finitely generated superalgebra for each Xi .
Unless otherwise specified, in this section all superschemes are assumed to be
algebraic.
Given a superscheme X D .jX j; OX / each point of x in the topological space
jXj belongs to an open affine subsuperscheme Spec A, x Š p 2 Spec A0 , so that
OX;x Š Ap . Recall that Ap is the localization of the A0 -module A at the prime ideal
p  A0 and that
˚ 
Ap D fg j f 2 A; g 2 A0  p :
The local ring Ap D OX;x contains the maximal ideal mX;x generated by the
maximal ideal in the local ring .Ap /0 and the generators of .Ap /1 as an A0 -module.
We want to define the notion of a rational point of a superscheme. We will then
define the tangent superspace to a superscheme at a rational point.
We want to remark that it is possible to define (as in algebraic geometry) the notion
of tangent space at a generic point of a superscheme, not necessarily rational. We
 10.6 The infinitesimal theory 197

however shall restrict our attention to the rational points since, as we shall see, they
are in one-to-one correspondence with the k-points of the superscheme and in most
applications these are the only interesting points to consider.

Definition 10.6.2. Let X D .jX j; OX / be a superscheme. A point x 2 jXj is said to be

rational if OX;x =mX;x Š k. A point x 2 jX j is said to be closed if it corresponds to a
maximal ideal in Spec A0 , where .Spec A0 ; OA /  X is any affine open neighbourhood
of x.

Remark 10.6.3. As in the commutative case we have that if k is algebraically closed,

then all closed points of jX j are rational. This is because the field OX;x =mX;x is a finite
algebraic extension of k (see [2], 7.9, and [23], §3, n. 6, for more details).

It is important not to confuse the points of the topological space jX j with the ele-
ments in hX .A/ for a generic A 2 .salg/. These are called A-points of the superscheme
X. The next observation clarifies the relationship between the points of jX j and hX ,
the functor of points of X .

Observation 10.6.4. There is a bijection between the rational points of a super-

scheme X and the set of its k-points hX .k/. In fact, an element .jf j; f  / 2 hX .k/,
jf j W Spec k ! jXj, f  W OX;x ! k, determines immediately a point x D jf j.0/,
which is rational since OX;x =mX;x Š k. Vice versa a rational point x 2 jX j cor-
responds to a morphism , where   W OX;x ! OX;x =mX;x Š k, and we obtain
jj W Spec k ! jX j simply by assigning to the only prime .0/ the point x 2 jX j.

Definition 10.6.5. Let A be a superalgebra and M an A-module. Let D W A ! M be

an additive map with the property D.a/ D 0 for all a 2 k. We say that D is a super
derivation if

D.fg/ D D.f /g C .1/p.D/p.f / f D.g/; f; g 2 A;

where p as always denotes the parity.

Definition 10.6.6. Let X D .jX j; OX / be a superscheme and let x be a rational point

in jXj. We define the tangent space of X at x to be

Tx X D Der.OX;x ; k/;

where k is viewed as an OX;x -module via the identification k Š OX;x =mX;x , with
mX;x the maximal ideal in OX;x .

The next proposition gives an equivalent definition for the tangent space.

Proposition 10.6.7. Let X be a superscheme. Then

Tx X D Der.OX;x ; k/ Š Hom.smod/ .mX;x =mX;x

2
; k/:
198 10 Supervarieties and superschemes

2
Note that mX;x =mX;x is an OX;x -supermodule which is annihilated by mX;x , hence
it is a k D OX;x =mX;x -supermodule i.e., a super vector space.
Proof. Let D 2 Der.OX;x ; k/. Since D is zero on k and OX;x D k ˚ mX;x , we have
that D is determined by its restriction to mX;x , DjmX;x . Moreover, since mX;x acts as
zero on k Š OX;x =mX;x , one can check that
W Der.OX;x ; k/ ! Hom.smod/ .mX;x =mX;x
2
; k/; D 7! DjmX;x ;
is well defined.
Now we construct the inverse. Let ˛ W mX;x ! k, ˛.mX;x
2
/ D 0. Define
D˛ W OX;x D k ˚ mX;x ! k; D˛ .a; f / D ˛.f /:
This is a well-defined superderivation.
1
Moreover one can check that the map ˛ 7! D˛ is . 
The next proposition provides a characterization of the tangent space that is useful
for explicit calculations.
Proposition 10.6.8. Let X D .jX j; OX / be a supervariety x 2 jX j a rational closed
point. Let U be an affine neighbourhood of x, mx  OX .U / the maximal ideal
corresponding to x. Then
Tx X Š Hom.smod/ .mX;x =mX;x
2
; k/ Š Hom.smod/ .mx =m2x ; k/:
Proof. The proof is the same as in the ordinary case and is based on the fact that
localization commutes with exact sequences. 
Let us compute explicitly the tangent space in an example.
Example 10.6.9. Consider the affine supervariety represented by the coordinate ring
CŒx; y; ; =.x C y/:
Notice that the reduced variety is the affine plane.
Since C is algebraically closed, all closed points are rational. Consider the closed
point P D .1; 1; 0; 0/ Š mP D .x  1; y  1; ; /  CŒx; y; ; =.x C y/, where
we identify .x0 ; y0 ; 0; 0/ with maximal ideals in the ring of the supervariety, as we do
in the commutative case. By Proposition 10.6.8, the tangent space at P consists of all
the functions ˛ W mP ! k, ˛.mP2 / D 0.
A generic f 2 mP lifts to the family of f D f1 C f2 .x C y/ 2 CŒA2j2  D
CŒx; y; ;  with f1 .1; 1; 0; 0/ D 0 and where f2 is any function in CŒA2j2  D
CŒx; y; ; . Thus f can be formally expanded in power series around P :
@f1 @f1
f D @x
.P /.x  1/ C @y
.P /.y  1/ C . @f
@
1
.P / C f2 .P //
C . @f
@
1
.P / C f2 .P // C higher order terms:
 10.6 The infinitesimal theory 199

Define
@f1 @f1 @f1 @f1
XD @x
.P /; Y D @y
.P /; „D @
.P /; ED @
.P /:

These are coordinates for the super vector space MP =MP2 , MP D .x 1; y 1; ; / 
CŒx; y; ; . A basis for the dual space .Mp =Mp2 / , which is the tangent space to A2j2 ,
consists of functions sending the coefficient of one of the x 1, y 1, ,  to a non-zero
element and the others to zero. We get equations for the tangent space .mP =mP2 / as
a subspace of .MP =MP2 / :
„  E D 0:
So we have described the tangent space .mP =mP2 / as a subspace of .MP =MP2 / ,
the tangent space to the affine superspace Amjn .

There is yet another way to compute the tangent space, in the case X is an affine
supervariety. Before we examine this construction, we must understand first the notion
of differential of a function and differential of a morphism. We start by defining the
value and the differential for the germs of sections.

Definition 10.6.10. Let X D .jX j; OX / be a superscheme, x a rational point.

Consider the projections

 W OX;x ! OX;x =mX;x Š k; p W mX;x ! mX;x =mX;x

2
:

Let f 2 OX;x . We define the value of f at x to be

f .x/ ´ .f /:

Notice that f  f .x/ 2 mX;x , where we interpret f .x/ 2 k  OX;x . We also define
the differential of f at x to be

.df /x ´ p.f  f .x//:

We now want to define the value and differential of a section at a point.

If U is an open neighbourhood of x and f 2 OX .U /, we define the value of f at
x to be
f .x/ ´ ..f //;
where  W OX .U / ! OX;x is the natural morphism.
Finally, the differential of f at x is

.df /x ´ .d.f //x :

We leave to the reader the simple check that this definition is independent from the
chosen open neighbourhood U of x.
200 10 Supervarieties and superschemes

Example 10.6.11. If P D .x10 ; : : : ; xm 0

; 0; : : : ; 0/ is a closed rational point of the affine
superspace A mjn
with coordinate ring kŒx1 ; : : : ; xm ; 1 ; : : : ; n , a basis of mP =mP2 is
fx  xi0 ; j gfiD1;:::;m;j D1;:::;ng . Hence one can readily see that

.dxi /P D x  xi0 ; .d j /P D j; i D 1; : : : ; m; j D 1; : : : ; n:

Definition 10.6.12. Let .j˛j; ˛  / W X ! Y be a morphism of superschemes and

x a rational point in jX j. Then j˛j.x/ is also rational and ˛ induces a morphism
d˛x W Tx X ! Tj˛j.x/ Y by

d˛x .D/f D D.˛x .f //; D 2 Tx X D Der.OX;x ; k/; ˛x W OY;j˛j.x/ ! OX;x ;

called the differential of ˛ at the point x 2 jX j.

Definition 10.6.13. We say that ˛ W X ! Y is a closed embedding if j˛j is a homeo-
morphism of jX j onto a closed subset of jY j and ˛  is a surjective sheaf morphism.
If X and Y are affine superschemes, X D Spec A, Y D Spec B, then ˛ is a closed
embedding if and only if ˛  is surjective on the global sections, that is, ˛  W A ! B Š
A=I for some ideal I in A.
Proposition 10.6.14. If .j˛j; ˛  / is a closed embedding, then d˛x is injective.
Proof. Direct check. 
If X is a subsupervariety of Amjn it makes sense to ask for equations that determine
the tangent superspace to X as a linear subsuperspace of Tx Amjn Š k mjn , as we did
in Example 10.6.9 for a special case.
Proposition 10.6.15. Let X be a subvariety of Amjn defined by the ideal I  O.Amjn / D
kŒx1 ; : : : ; xm ; 1 ; : : : ; n . Let x be a rational closed point of X . Then

Tx X D fv 2 k mjn j .df /x .v/ D 0 for all f 2 I g:

Proof. The closed embedding ˛ W X  Amjn corresponds to a surjective morphism

 W O.Amjn / ! O.X /, hence O.X / Š O.Amjn /=I . Let mx and Mx denote respec-
tively the maximal ideal associated to x in X and Spec O.Amjn /0 , respectively. The
map  induces a surjective linear map

W Mx =M2x  mx =m2x :

between superspaces. Let us recall the following simple fact of linear algebra.
If a W V1 ! V2 is a surjective linear map between finite-dimensional vector spaces
V1 , V2 and b W V2  V1 is the injective linear map induced by a on the dual vector
spaces, then s 2 im.b/ if and only if sjker.a/ D 0.
We apply this observation to the maps and the differential .d˛/x ,

.d˛/x W Tx .X / D .mx =m2x / ,! T˛.x/ .Amjn / D .Mx =M2x / ;

 10.7 References 201

and we see that

Tx .X / D fv 2 Tj˛j.x/ .Amjn / j v.ker. // D 0g:

Observe that ker. / D f.df /x j f 2 I g. By identifying Amjn D k mjn with its double
dual .k mjn / we obtain the result. 
Remark 10.6.16. In the notation of the previous proposition, if I D .f1 ; : : : ; fr / one
can readily check that

Tx X D fv 2 k mjn j .dfi /x .v/ D 0 for all i D 1; : : : ; rg;

thus obtaining a quick and effective method to calculate the tangent space to a super-
variety.
Let us revisit Example 10.6.9 and see how the calculation is made using Proposi-
tion 10.6.15.
Example 10.6.17. Consider again the supervariety represented by the superalgebra

CŒx; y; ; =.x C y/:

We want to compute the tangent space at P D .1; 1; 0; 0/ D .x0 ; y0 ; 0 ; 0 /:

d.x C y/P D x0 .d /P C 0 .dx/P C y0 .d/P C 0 .dy/P

D .d /P C .d/P
Š .0; 0; 1; 1/:

Hence by Proposition 10.6.15 the tangent space is the subspace of k 2j2 given by the
equation
  D 0:

10.7 References
The definition of superscheme together with its functor of points appears in the work
of Manin [56], where also the examples of the Grassmannian and flag superschemes
are described in detail in the complex analytic category. The representability criterion
Theorem 10.3.7 appears in [23], Ch. I, for the ordinary setting and is used, however
not formally proved, in the work by Manin [56].
11
Algebraic supergroups

In this chapter we restrict our attention to the superschemes that have an additional
structure, namely the group structure, and thus are called supergroup schemes or sim-
ply supergroups for short. The simplest way to introduce this extra structure is the
requirement for the functor of points to be group-valued, that is, if X is a superscheme,
we are asking that hX .A/ be a group for each superalgebra A.
As a general rule in supergeometry, the functor of points is valued in the ordinary
categories like sets, groups, vector spaces or Lie algebras and the supernature of the
geometrical object stems from the category we start from, in our case superalgebras
or superschemes. As it happens for the ordinary setting, when the superscheme G is
affine, G is a supergroup if and only if its representing superalgebra O.G/ is a Hopf
superalgebra. We shall discuss in detail the example of the general and special linear
supergroups together with their Hopf superalgebras.
As for the ordinary setting, we can associate very naturally to any supergroup its
Lie superalgebra, which is a Lie algebra-valued functor, identified with the functor of
points of the super tangent space of the supergroup at the identity. We also introduce
the concept of representation of a supergroup and we prove the following two important
results: we show that any affine supergroup can be embedded into some GL.V / for
a suitable V , and then we prove the representability of the stabilizer functor for the
action of a supergroup on a superscheme. This gives the representability of the classical
algebraic supergroups corresponding to the classical Lie superalgebras (see Appendix A
for their list).
Our treatment follows very closely [23], Ch. II; most of the classical statements go
unchanged to the super-setting and we shall point out the differences when they arise.

11.1 Supergroup functors and supergroup schemes

Let k be a commutative ring. All superalgebras are assumed to be commutative and
over k unless otherwise specified. Their category is denoted by .salg/.
A supergroup scheme is a superscheme whose functor of points is group-valued,
that is to say, it associates a group to each superscheme or equivalently to each super-
algebra. In fact, as we know, the functor of points of a superscheme is determined by
its restriction to affine superschemes, whose category is equivalent to the category of
superalgebras.
As often happens in algebraic geometry, in order to study supergroup schemes we
need first to define and understand the weaker notion of supergroup functor, which
 11.1 Supergroup functors and supergroup schemes 203

is simply a group-valued functor, from .salg/ to .sets/, without any representability

requirement that characterizes the functor of points of supergroup schemes and more
in general superschemes.
Definition 11.1.1. A supergroup functor is a group-valued functor

G W .salg/ ! .sets/;

where by group-valued functor we mean a functor valued in the category of groups.

This is equivalent to having the following natural transformations:
(1) Multiplication W G  G ! G such that B .  id/ D .  id/ B , i.e.,
id
GGG /GG

id 
  
GG / G.

(2) Unit e W ek ! G, where ek W .salg/ ! .sets/, ek .A/ D 1A (in other words,

ek D hSpec k ) such that B .id  e/ D B .e  id/, i.e.,

ide /
GG o
eid
G  ekK ek  G
KK ss
KK s
KK ss
ss

KK
K%  ysss
G.

(3) Inverse i W G ! G such that B .id; i / D e B id, i.e.,

.id;i/
G /GG

 
ek e / G.

The supergroup functors together with their morphisms, that is the natural trans-
formations that preserve , e and i, form a category.
If G is the functor of points of a superscheme X , i.e., G D hX , in other words
G.A/ D Hom.Spec A; X /, we say that X is a supergroup scheme. An affine su-
pergroup scheme X is a supergroup scheme which is an affine superscheme, that is
X D Spec O.X / for some superalgebra O.X /. To make the terminology easier we
will drop the word “scheme” when speaking of supergroup schemes, whenever there
is no danger of confusion.
As we shall presently see, the functor of points of an affine supergroup is represented
by a superalgebra, which has the additional structure of a Hopf superalgebra. We refer
the reader to Chapter 1, Section 1.7, for their definitions and main properties.
204 11 Algebraic supergroups

Proposition 11.1.2. Let G be an affine superscheme. Then G is a supergroup if and

only if O.G/ is a super Hopf algebra. Moreover, we can identify the category of affine
supergroups with the category of commutative Hopf superalgebras.
Proof. We first observe that, if G is a superscheme, and O.G/ is a Hopf superalgebra
with comultiplication , counit  and antipode S, hG .A/ has a natural group structure.
In fact we can define the product of two morphisms in hG .A/ by
 x˝y A
xy D A Bx˝yB W O.G/ 
! O.G/ ˝ O.G/ ! A ˝ A ! A;

where A is the multiplication in the superalgebra A. One can immediately check that
the multiplication is a morphism, that is,

.x  y/.ab/ D .x  y/.a/.x  y/.b/ for all a; b 2 A

(though hidden, the sign rule plays a crucial role here). This multiplication in hG .A/
gives rise to in O.G/ as its associated comultiplication, as one can readily see.
The unit eA and the inverse iA in hG .A/ are defined by
 A
eA D A B  W O.G/ !
 k ! A; iA .x/ D x B S;

where A is the unit in A. We leave to the reader the routine checks of Definition 11.1.1.
Vice versa, if G is a supergroup scheme, then we can define the comultiplication
W O.G/ ! O.G/ ˝ O.G/ as the dual of the multiplication 2 Hom.G  G; G/
using the identification

Hom.G  G; G/ Š Hom.O.G/; O.G  G//

(one can readily check that O.G  G/ Š O.G/ ˝ O.G/). Similarly one defines the
counit and the antipode  and S as the duals of unit e and inverse i .
A careful look shows that formally the diagrams defining a supergroup functor
are essentially the same as those defining a Hopf superalgebra, with arrows reversed.
We leave to the reader the routine verification that hG satisfies all the diagrams in
Definition 11.1.1 if and only if O.G/ satisfies the diagrams in Definition 1.7.1.
The equivalence between the categories of affine supergroups and commutative
Hopf superalgebras is an immediate consequence of the previous discussion and Propo-
sition 10.1.9. 
Let us now examine some important examples of supergroup schemes and their
associated Hopf superalgebras.
Examples 11.1.3. (1) Supermatrices Mmjn . In Remark 1.4.1 we have introduced the
functor of points of supermatrices
 
a ˛
Mmjn W .salg/ ! .sets/; A 7! ;
ˇ b
 11.1 Supergroup functors and supergroup schemes 205

where a and b are m  m, n  n block matrices with entries in A0 , while ˛ and ˇ are
mn, nm block matrices with entries in A1 . As we have seen in Example 3.3.4, Mmjn
is a representable functor, represented by the superalgebra of polynomials kŒxij ; kl 
for suitable indices i , j , k, l. The functor Mmjn is group-valued, in fact any Mmjn .A/
has an additive group structure, where the addition is simply defined as the addition of
matrices. Hence by the previous proposition kŒxij ; kl  is a Hopf superalgebra, where
the comultiplication , the counit  and antipode S are given by

.xij / D xij ˝ 1 C 1 ˝ xij ; . kl / D kl ˝ 1 C 1 ˝ kl ;

.xij / D 0; . ij / D 0; S.xij / D xij ; S. ij / D  ij :

We leave to the reader the verification that kŒxij ; kl  together with ,  and S is a
Hopf superalgebra.
(2) The general linear supergroup GLmjn . Let A 2 .salg/. Let us define GLmjn .A/
as GL.Amjn / (see Chapter 1) to be the set of automorphisms of the A-supermodule
Amjn . Choosing the standard basis we can write
² ³
a ˛
GLmjn .A/ D  Mmjn .A/:
ˇ b

As we have previously remarked, GLmjn .A/ are the invertible transformations of

k mjn .A/ preserving parity.
This is the functor of points of an affine supergroup GLmjn represented by the Hopf
superalgebra

O.GLmjn / D kŒxij ; kl ŒU; V =.Ud1  1; Vd2  1/

where xij ’s, U , V and kl ’s are respectively even and odd variables with 1  i; j  m
or m C 1  i; j  m C n, 1  k  m, m C 1  l  m C n or m C 1  k  m C n,
1  l  m and
X
d1 D .1/l.s/ x1;s.1/ : : : xm;s.m/ ;
s2Sm
X
d2 D .1/l.t/ xmC1;mCt.1/ : : : xmCn;mCt.n/ :
t2Sn

It is customary to write d11 and d21 in place of U and V , so we shall note

1 1
O.GLmjn / D kŒxij ; kl Œd1 ; d2 :

Notice that the Berezinian function is well defined in O.GLmjn /, in fact

Ber D d21 det.a  ˇb 1 ˛/:

Since GLmjn is a representable group-valued functor, Proposition 11.1.2 ensures

that there is a well-defined Hopf superalgebra structure on O.GLmjn /, corresponding
206 11 Algebraic supergroups

(dually) to the multiplicative group structure on each GLmjn .A/. Nevertheless, we

want to explicitly describe such a Hopf superalgebra structure.
Let us first start with the bialgebra structure of O.GLmjn /.
The comultiplication and the counit  are given by the following formulas. To
ease the notation, let aij D xij if p.i / C p.j / is even and aij D ij otherwise (an
index is even if it is between 1 and m). We define

X
mCn
.aij / D aik ˝ akj ; .aij / D ıij :
kD1

We need to specify ,  and S also on the generators d11 and d21 and then to
check that they are well defined with respect to the ideal of the relations d11 d1 D 1,
d21 d2 D 1:

X
2mnC2
i i
.d11 / D .1/i1 .d11 / ˝ .d11 / . .d1 /  d11 ˝ d11 /i1 ;
iD1
X
2mnC2
i i
.d21 / D .1/i1 .d21 / ˝ .d21 / . .d2 /  d21 ˝ d21 /i1 ;
iD1
.d11 / D .d21 / D 1:

We leave to the reader the tedious verification that ,  are well defined and satisfy
respectively the properties of the comultiplication and counit as in Definition 1.7.1 and
they correspond to the group structure of the functor GLmjn .
As for the antipode, the definition turns out to be more complicated and this is
a consequence of the fact that the inverse for a supermatrix has a far more compli-
cated formula than the inverse of an ordinary matrix. For such formulas, we refer the
interested reader to [30].
(3) The special linear supergroup SLmjn . For a superalgebra A, let us define
SLmjn .A/ to be the subgroup of GLmjn .A/ consisting of matrices with Berezinian
equal to 1. This is the functor of points of an affine supergroup and it is represented by
the Hopf superalgebra
1 1
O.SLmjn / D kŒxij ; kl Œd1 ; d2 =.Ber 1/;

where the comultiplication, counit and antipode are inherited naturally from the ones
in GLmjn .

We end our introduction on supergroup schemes with an observation concerning

the unit of a supergroup scheme that we shall need later.

Observation 11.1.4. When k is a field, we may interpret the unit e of a supergroup

scheme G D .jGj; OG / as a rational point, denoted by 1G , of G. This is a point in jGj
 11.2 Lie superalgebras 207

for which OG;1G =m1G Š k, where m1G is the maximal ideal in the local superring
OG;1G , in the following way. By definition, the unit e is a morphism of superschemes
e W ek ! G, hence corresponds to a pair of morphisms e D .jej; e  /, jej W jek j ! jGj,
e  W OG ! e Oek D k. Define 1G ´ jej.jek j/ 2 jGj. This is a rational point, in
fact OG;1G =m1G Š k. Moreover notice that, by the very definition of e, 1G has the
property of a unit for the topological group jGj.

11.2 Lie superalgebras

In Chapter 1 we defined a Lie superalgebra as a super vector space with a bracket
satisfying the antisymmetry property and the Jacobi identity, of course with suitable
signs. In this section we want to reformulate functorially the same notion.
If g is a finite-dimensional super vector space and hg is its functor of points, we
shall see that g is a Lie superalgebra if and only if hg is Lie algebra-valued, in other
words, hg .A/ is a Lie superalgebra for all superalgebras A. As we remarked at the
beginning of this chapter, the reader should not be confused by the fact that hg is valued
in the category of ordinary Lie algebras. This is in complete analogy with the fact that
the functor of points of a supergroup is a group-valued functor. The supergeometric
nature of these objects is expressed by the category we start from, namely the category
of superalgebras.
Consider the functor of points of the superscheme A1j0 , the affine line, Ok W .salg/ !
.sets/, Ok .A/ D Hom.kŒx; A/ Š A0 For notational purposes we use the symbol Ok
to denote it, instead of hA1j0 .

Definition 11.2.1. Let M be a functor M W .salg/ ! .sets/, with an operation, that

is for each A 2 .salg/ we define functorially the operation M.A/  M.A/ ! M.A/,
.x; y/ 7! xy, sometimes also denoted additively as .x; y/ 7! x C y. We say that M
is an Ok -module if we have a natural transformation

Ok .A/  M.A/ ! M.A/; .a; x/ 7! ax;

such that
• a.xy/ D .ax/.ay/,
• .ab/x D a.bx/
for all a; b 2 Ok .A/ and x; y 2 M.A/,

For example if M is the functor of points of a super vector space V , that is, M.A/ D
.A ˝ V /0 , we can define the operation .x; y/ 7! x C y that associates to a pair of
elements in the A0 -module M.A/ their sum. Then M has a natural structure of an
Ok -module, which is the multiplication of elements in M.A/ by scalars in A0 .
208 11 Algebraic supergroups

Definition 11.2.2. Let g be a super vector space. We say that the functor

Lg W .salg/ ! .sets/; Lg .A/ D .A ˝ g/0 ;

is Lie algebra-valued if it is an Ok -module and there is an Ok -linear natural transfor-

mation, called the bracket,
Π;  W Lg  Lg ! Lg
that satisfies commutative diagrams corresponding to the ordinary antisymmetric prop-
erty and the Jacobi identity. In other words, for each superalgebra A, there is a well-
defined bracket Π; A that defines functorially a Lie algebra structure on the A0 -module
Lg .A/ D g.A/. We will drop the suffix A from the bracket and the natural transfor-
mations to ease the notation.

Remark 11.2.3. If the super vector space g is finite-dimensional, then the functor Lg
is representable and we have Lg D hg . In fact

Lg .A/ D .A ˝ g/0 D Hom.smod/ .g ; A/ D Hom.salg/ .Sym.g /; A/;

where .smod/ denotes the category of supermodules (over k in this case) and Sym.g /
the symmetric algebra over g . Notice also that in this case Lg is the functor of points
of an affine superscheme represented by the superalgebra Sym.g /.

We now want to see that the usual notion of Lie superalgebra, as we defined in
Chapter 1, Section 1.2, is equivalent to this functorial definition. We want to show that
g is a Lie superalgebra if and only if the functor Lg is Lie algebra-valued.
Let us first recall the definition of Lie superalgebra given in Chapter 1, Section 1.2.

Definition 11.2.4. Let g be a super vector space. We say that g is a Lie superalgebra
if there exists a bilinear map Π;  W g  g ! g called a superbracket such that
(a) Œx; y D .1/p.x/p.y/ Œy; x,
(b) Œx; Œy; z C .1/p.x/p.y/Cp.x/p.z/ Œy; Œz; x C .1/p.x/p.z/Cp.y/p.z/ Œz; Œx; y
for all x; y; z 2 g.

Proposition 11.2.5. Let g be a super vector space. Then g is a Lie superalgebra if

and only if Lg W .salg/ ! .sets/, Lg .A/ D .A ˝ g/0 , is a Lie algebra-valued functor.

Proof. This is an immediate consequence of the even rules principle, detailed in Chap-
ter 1, Section 1.8. Nevertheless, given the importance of this construction, we want to
see explicitly how the superbracket on g and the bracket on Lg .A/ correspond to each
other.
If we have a Lie superalgebra Lg there is always, by definition, a super vector
space g associated to it. Moreover by the even rules, there is a unique Lie superalgebra
structure on the A-module LA g D A ˝ g, whose even part is the Lie algebra Lg .A/.
 11.2 Lie superalgebras 209

Given v; w 2 g, since the Lie bracket on LA

g is A-linear, we can define the element
fv; wg 2 g as

Œa˝v; b˝w D .1/p.b/p.v/ ab˝fv; wg 2 .A˝g/0 2 Lg .A/; a˝v; b˝w 2 .A˝g/0 :

Clearly the bracket fv; wg 2 g does not depend on a; b 2 A. One can verify that it is
a superbracket. Let us see, for example, the antisymmetry property. Observe first that
if a ˝ v 2 .A ˝ g/0 , then p.v/ D p.a/ since .A ˝ g/0 D A0 ˝ g0 ˚ A1 ˝ g1 . So
we can write
Œa ˝ v; b ˝ w D .1/p.b/p.v/ ab ˝ fv; wg:
On the other hand,

Œb ˝ w; a ˝ v D .1/p.a/p.w/ ba ˝ fw; vg
D .1/p.a/p.w/Cp.a/p.b/ ab ˝ fw; vg
D .1/2p.w/p.v/ ab ˝ fw; vg
D ab ˝ fw; vg:

By comparing the two expressions we get the antisymmetry of the superbracket. For
the super Jacobi identity the calculation is the same.
A similar calculation also shows that, given a super Lie algebra g, the functor Lg
is Lie algebra-valued. 
Hence the previous proposition shows that a Lie algebra-valued functor Lg ac-
cording to Definition 11.2.2 is equivalent to a super Lie algebra g. With an abuse of
language we will refer to both g and Lg as “Lie superalgebra”.
Remark 11.2.6. Given a super vector space g one may also define a Lie superalgebra
to be the representable functor Dg W .salg/ ! .sets/ so that

Dg .A/ D Hom.smod/ .g ; A/ D Hom.salg/ .Sym.g /; A/

with an Ok linear natural transformation Π;  W Dg  Dg ! Dg satisfying the com-

mutative diagrams corresponding to antisymmetry and Jacobi identity. When g is
finite-dimensional this definition coincides with Definition 11.2.2, however we have
preferred the one given there since its immediate equivalence with the definition is the
one mostly used in the literature.
Example 11.2.7 (The supermatrices as a Lie algebra-valued functor). Consider again
the functor of points of supermatrices
² ³
P Q
A 7! Mmjn .A/ D ;
R S

where P , Q, R, S are respectively .m  m/, .m  n/, .n  m/, .n  n/-matrices with

entries: P and S in A0 , R and Q in A1 .
210 11 Algebraic supergroups

This is a Lie algebra-valued functor once we define a Lie bracket on each Mmjn .A/:

ŒX; Y  D X Y  YX for all X; Y 2 Mmjn .A/:

In Example 3.3.4 we have seen that Mmjn is a representable functor corresponding to

a super vector space M.mjn/ (by an abuse of notation we may at times use the same
letter). The super vector space M.mjn/ is a super Lie algebra with bracket

ŒX; Y  D X Y  .1/p.x/p.y/ YX for all X; Y 2 M.mjn/:

One word of warning: the super vector space M.mjn/ is not Mmjn .k/. In fact Mmjn .k/
consists only of the even part of the super Lie algebra M.mjn/ and contains just the
diagonal block matrices with entries in k, while M.mjn/, as a vector space, consists of
.m C n  m C n/-matrices with entries in k and has superdimension m2 C n2 j2mn.
The purpose of the next two sections is to naturally associate a Lie superalgebra
Lie.G/ to a supergroup G.

11.3 Lie.G / of a supergroup functor G

In ordinary geometry we can associate to any group scheme G a Lie algebra, commonly
denoted by Lie.G/, which is identified with the tangent space to the group scheme G at
the identity. This is an extremely important construction since it allows us to linearize
problems, by transferring our questions from the group to its Lie algebra.
We want to proceed and repeat the same construction in the supergeometric setting
and we shall achieve this in two steps: first we define the super vector space Lie.G/
associated to the supergroup G and we show that it is isomorphic to the tangent space
to G at the identity element. Then, in the next section, we show that Lie.G/ is a Lie
superalgebra functor. Our treatment follows closely the classical one as detailed in
[23], Ch. II.
As customary in algebraic geometry and in supergeometry, we start by giving the
more general notion of Lie.G/ associated to a supergroup functor G (not necessarily
representable).
Let G be a supergroup functor. Let A be a commutative superalgebra and let
A./ ´ AŒ=. 2 / be the algebra of dual numbers, where  is an even indeterminate.
We have A./ D A ˚ A, and there are two natural morphisms

i W A ! A./; i.1/ D 1; p W A./ ! A; p.1/ D 1; p./ D 0; p B i D idA :

Definition 11.3.1. Consider the homomorphism G.p/ W G.A.// ! G.A/. For each
G there is a supergroup functor

Lie.G/ W .salg/ ! .sets/; Lie.G/.A/ ´ ker.G.p//:

 11.3 Lie.G/ of a supergroup functor G 211

If G is a supergroup scheme, we denote Lie.hG / by Lie.G/.

If f W G ! H is a natural transformation of supergroup functors, we have the
following commutative diagram (where the vertical arrows form exact sequences):

1O /1
O

fA
G.A/ / H.A/
O O
G.p/ H.p/
fA./
G.A.// / H.A.//
O O

fA./ jLie.G/.A/
Lie.G/.A/ / Lie.H /.A/
O O

1 / 1.

We hence define: Lie.f /.A/ D fA./ jLie.G/.A/ . The following proposition is im-
mediate.

Proposition 11.3.2. Lie is a functor from the category of supergroup functors to the
category of groups.

We are going to show that, when G is an algebraic supergroup scheme, Lie.G/ is a

Lie algebra-valued functor, thus associating to any algebraic supergroup scheme a Lie
superalgebra. Let us first examine some important examples.

Examples 11.3.3. (1) The general linear Lie superalgebra. We want to determine the
functor Lie.GLmjn / for k a field. Consider the morphism
   
p C p 0 q C q 0 p q
GLmjn .p/ W GLmjn .A.// ! GLmjn .A/; 7 ! ;
r C r 0 s C s 0 r s

with p, p 0 , s, s 0 having entries in A0 and q, q 0 , r, r 0 having entries in A1 ; the blocks p

and s are invertible matrices. One can see immediately that
² ³
Im C p 0 q 0
Lie.GLmjn /.A/ D ker.GLmjn .p// D ;
r 0 In C s 0

where In is an nn identity matrix. The functor Lie.GLmjn / is clearly group-valued and
can be identified with the (additive) group functor Mmjn defined as (see Example 11.2.7)

Mmjn .A/ D Hom.smod/ .M.mjn/ ; A/ D Hom.salg/ .Sym.M.mjn/ /; A/;

212 11 Algebraic supergroups

where M.mjn/ is the super vector space

² ³
P Q 2 2
M.mjn/ D Š k m Cn j2mn
R S

with P , Q, R, S respectively .m  m/, .m  n/, .n  m/, .n  n/-matrices with entries

in k. An element X 2 M.mjn/ is even if Q D R D 0 and is odd if P D S D 0.
As we already noticed in Example 11.2.7, M.mjn/ is a Lie superalgebra with
superbracket
ŒX; Y  D X Y  .1/p.X/p.Y / YX:
So Lie.GLmjn / is a Lie superalgebra. In the next section we will see that in general we
can give a Lie superalgebra structure to Lie.G/ for any group scheme G.
(2) The special linear Lie superalgebra. A similar computation shows that
˚   
Cp 0 q 0
Lie.SLmjn /.A/ D W D Imr 0 I Cs 0
j Ber.W / D 1 :
n

The condition on the Berezinian is equivalent to

det.In  s 0 / det.Im C p 0 / D 1;

which gives
tr.p 0 /  tr.s 0 / D 0:
Hence
Lie.SLmjn /.A/ D fX 2 Mmjn .A/ j str.X / D 0g;

where str is the supertrace, i.e., str a dˇ D tr.a/  tr.d /.

In the previous examples we have discovered that Lie.G/ comes with a built-in
Ok -module structure. This is actually true in general, that is, for all superalgebras
A, we have that Lie.G/.A/ is an A0 -module in a functorial manner. In fact, let
ua W A./ ! A./ be the endomorphism, ua .1/ D 1, ua ./ D a, for a 2 A0 . ua
induces G.ua / W G.A.// ! G.A.//, that by compatibility conditions gives a well-
defined morphism Lie.G/.ua / W Lie.G/.A/ ! Lie.G/.A/. Hence there is a natural
transformation Ok  Lie.G/ ! Lie.G/ such that

.a; x/ 7! ax ´ Lie.G/.ua /x; a 2 Ok .A/; x 2 Lie.G/.A/;

for any superalgebra A. One can immediately check for GLmjn .A/ and its subgroups
that the morphism .a; x/ 7! ax corresponds to the multiplication of all the entries of
the matrix x by a 2 A0 .
We summarize our discussion with a proposition, the proof of which is a simple
check that we leave to the reader.
Proposition 11.3.4. Let G be a supergroup functor. Then Lie.G/ is an Ok -module
with respect to the morphism .a; x/ 7! ax detailed above. That is:
 11.4 Lie.G/ for a supergroup scheme G 213

(1) a.xy/ D .ax/.ay/,

(2) .ab/x D a.bx/ for all a; b 2 A0 , x; y 2 Lie.G/.A/.
Moreover if f W G ! H is a morphism of supergroup functors, then Lie.f /.ax/ D
a Lie.f /.x/, that is, a morphism of supergroup functors induces a morphism
Lie.f / W Lie.G/ ! Lie.H / of Ok -modules.
We are now going to introduce the adjoint morphism, which plays an important
role in the theory.
Definition 11.3.5. Let G be a supergroup functor. We define the adjoint morphism Ad
as the natural transformation Ad W G  Lie.G/ ! Lie.G/ given by
G.A/  Lie.G/.A/ ! Lie.G/.A/; .g; x/ 7! G.i /.g/xG.i /.g/1 :
Notice that since Lie.G/.A/ D ker.G.p// is a normal subgroup in G.A.//, we have
G.i/.g/xG.i/.g/1 2 Lie.G/.A/  G.A.//.
The following proposition establishes the naturality of our definition of Ad; its
proof, amounting to trivial checks, is left to the reader.
Proposition 11.3.6. The adjoint morphism Ad is compatible with the Ok -module struc-
ture of Lie.G/, in other words:
(1) Ad.g/.xy/ D Ad.g/.x/ Ad.g/.y/,
(2) Ad.g/.ax/ D a Ad.g/.x/ for all a; b 2 A0 , x; y 2 Lie.G/.A/.

11.4 Lie.G / for a supergroup scheme G

Let us now assume that G is a supergroup scheme over a field k.
We now want to show that Lie.G/ W .salg/ ! .sets/ is a representable functor and
its representing superscheme is identified with the tangent space at the identity T1G G
of the supergroup G.
We start by showing that Lie.G/ is isomorphic as an Ok -module (i.e., as a super
vector space) to T1G G, the tangent space to G at 1G . Before this we need some general
preliminaries.
Definition 11.4.1. Let X D .jX j; OX / be a superscheme, x 2 jX j. We define the first
neighbourhood of X at a point x 2 jX j, Xx , to be the superscheme Spec OX;x =mX;x2
,
where as usual mX;x is the maximal ideal in the local superring OX;x . The topological
space jXx j consists of the one point mX;x which is the maximal ideal in OX;x .
Observation 11.4.2. There exists a natural morphism f W Xx ! X , from the first
neighbourhood of X at x to the superscheme X . In fact we can write immediately the
topological space map
jf j W Spec OX;x =mX;x
2
D fmX;x g ! jX j; mX;x 7! x;
214 11 Algebraic supergroups

and the sheaf morphism

fU W OX .U / ! OXx ! OX;x =mX;x

2
D OXx .jf j1 .U //;

where fU is the composition of a natural map from OX .U / to the direct limit OX;x and
the projection OX;x ! OX;x =mX;x2
(with x 2 U ).

We now want to make some observations on the unit or identity element of a

supergroup G. By definition we have that the identity is a superscheme morphism
e W Spec k ! G. This corresponds to a natural transformation of the functor of points:
hSpec k ! hG assigning to the only morphism 1A 2 hSpec k .A/ a morphism that we will
denote by 1G.A/ 2 hG .A/ D Hom.Spec A; G/. The topological space map j1G.A/ j
sends all the maximal ideals in Spec A to 1G 2 jGj. The sheaf morphism OG ! k
is the evaluation at 1G that is OG .U / ! OG;1G ! OG;1G =mG;1G Š k (the identity
is a rational point by Observation 11.1.4). Hence it is immediate to verify that 1G.A/
factors through G1 (the first neighbourhood at the identity 1G ), i.e.,

1G.A/ W Spec A ! G1 D Spec OG;1G =m2G;1G ! G:

This fact will be crucial in the proof of the next theorem.

Theorem 11.4.3. Let G be a supergroup scheme. Then

Lie.G/.A/ Š Hom.smod/ .mG;1G =m2G;1G ; A/

as Ok -modules.

Proof. We start by defining the Ok -linear natural transformation

W Hom.smod/ .mG;1G =m2G;1G ; A/ ! Lie.G/.A/

and then we will show that is invertible by providing an explicit inverse.

Let d W mG;1G =m2G;1G ! A be a linear map. Let d 0 be the linear map

d 0 W OG;1G =m2G;1G Š k ˚ mG;1G =m2G;1G ! A./; .s; t / 7! s C d.t /:

So we have d 0 2 hG1 .A.// since G1 is the superscheme represented by OG;1G =m2G;1G .

This shows that we have a correspondence between hG1 .A/ and the elements of
Hom.smod/ .mG;1G =m2G;1G ; A/. Let  W G1 ! G be the morphism described in Obser-
vation 11.4.2. By Yoneda’s lemma,  induces A./ W hG1 .A.// ! hG .A.//, hence
we have an A0 -linear map

W Hom.smod/ .mG;1G =m2G;1G ; A/ ! hG .A.//; d 7! A./ .d 0 /:

 11.5 The Lie superalgebra of a supergroup scheme 215

The following commutative diagram shows that .d / 2 ker.hG .p// D Lie.G/.A/:

hG1 .p/
hG1 .A.// / hG .A/
1
 / 1G.A/
d0
A./ A
 hG .p/ 
hG .A.// / hG .A/
.d /  / 1G.A/ .

We now want to build an inverse for . Let z 2 ker.hG .p//  hG .A.//, i.e.,
hG .p/z D 1G.A/ , where

hG .p/ W hG .A.// D Hom.Spec A./; G/ ! hG .A/ D Hom.Spec A; G/; z 7! 1G.A/ :

We have the commutative diagram

Spec A./ z /G
;
O www
ww
w
ww
Spec p
ww 1G.A/
Spec A,

where Spec p is the morphism induced by p W A./ ! A.

Here z factors via G1 because 1G.A/ splits via G1 . Since z factors via G1 , that is,
z W Spec A./ ! G1 ! G, this provides immediately a map Spec A./ ! G1 , i.e., an
element in hG1 .A.//, corresponding to an element in Hom.smod/ .mG;1G =m2G;1G ; A/.
So we have defined a map Lie.G/.A/ ! Hom.smod/ .mG;1G =m2G;1G ; A/. One can
check it is functorial and Ok -linear. We leave to the reader the check that this is the
inverse of . 
Corollary 11.4.4. Let G be a supergroup scheme. Then

Lie.G/ Š T1G G;

that is, Lie.G/ is identified as a super vector space with the tangent space to G at the
identity.
Proof. Immediate from Theorem 11.4.3 and Proposition 10.6.8. 

11.5 The Lie superalgebra of a supergroup scheme

We now want to show that Lie.G/ is a Lie superalgebra for any supergroup scheme G
over a field k. In other words we want to show that the functor Lie.G/ W .salg/ ! .sets/,
216 11 Algebraic supergroups

defined in the previous section, is Lie algebra-valued and is represented by a super affine
space, which carries a Lie superalgebra structure.
Our treatment of these topics again follows closely the classical discussion that is
found in [23], Ch. II. We nevertheless find it necessary to repeat all of the arguments,
given our different scope and in order to make the text self-contained.
We now want to define a natural transformation Π;  W Lie.G/  Lie.G/ ! Lie.G/
which has the properties of a bracket.
Let GL.Lie.G//.A/ be the (multiplicative) group of linear automorphisms of the
super vector space Lie.G/ and let End.Lie.G//.A/ be the (additive) group of linear
endomorphisms of Lie.G/.A/:
GL.Lie.G//.A/ D f W Lie.G/.A/ ! Lie.G/.A/ j  invertible g  End.Lie.G//.A/:
The natural Ok -module structure of Lie.G/ makes the two supergroup functors
(one multiplicative the other additive)
GL.Lie.G// W .salg/ ! .sets/; End.Lie.G// W .salg/ ! .sets/
valued in the A0 -linear morphisms of Lie.G/.A/.
As in the ordinary setting, we have an identification between the Lie superalge-
bra of the general linear supergroup Lie.GL.Lie.G/// and the Lie superalgebra of
endomorphisms End.Lie.G//, as we shall see in the next proposition.
Proposition 11.5.1. Let V be a super vector space. Then
Lie.GL.V // D End.V /:
Proof. Let us first define a natural transformation W End.V / ! Lie.GL.V // and
then we show that it is an isomorphism. Let f 2 End.V /. If  is an even indeterminate
 2 D 0, we have 1 C f 2 End.V /.A.// and actually 1 C f 2 GL.V /.A.// since it
is invertible, its inverse being 1  f . Clearly 1 C f 2 ker GL.p/ D Lie.GL.V //.A/,
so we define .f / D 1 C f . We have immediately that is functorial and A0 -linear.
Now we show that is an isomorphism. Recall in general that if V is a super vector
space we have
End.V /.A.// ´ End.V .A./// Š End.V .A// ˝ A./ D End.V /.A/ ˝ A./:
Hence
End.V /.A.// Š End.V /.A/ ˚  End.V /.A/;
GL.V /.A.// Š GL.V /.A/ ˚  End.V /.A/;
which gives us the result. 
As in the ordinary Lie theory, we can define the adjoint morphisms Ad and ad, which
again play a key role. Though we have already defined Ad for supergroup functors, we
prefer to repeat the definition in the context of supergroup schemes, since the notation
will be different.
 11.5 The Lie superalgebra of a supergroup scheme 217

Definition 11.5.2. The adjoint action Ad of G on Lie.G/ is defined as the natural

transformation

Ad W hG ! GL.Lie.G//;
Ad.g/.x/ D hG .i /.g/xhG .i /.g/1 ; g 2 hG .A/; x 2 Lie.G/.A/:

Notice that Ad.g/x 2 hG .A.//; however, since ker.p/ is a normal subgroup, we

have Ad.g/x 2 Lie.G/.A/.
The adjoint action ad of Lie.G/ on Lie.G/ is defined as

ad ´ Lie.Ad/ W Lie.G/ ! Lie.GL.Lie.G/// D End.Lie.G//:

We are ready to define a bracket on Lie.G/ by

Œx; y ´ ad.x/y; x; y 2 Lie.G/.A/:

Our goal is to prove that Π;  is a Lie bracket for all superalgebras A in a functorial
manner or, equivalently, that Lie.G/ is a Lie algebra-valued functor.
Observation 11.5.3. Let c.g/ W hG .A/ ! hG .A/ be the conjugation by an element
g 2 jGj: c.g/.x/ D gxg 1 . Then

Ad.g/ D Lie.c.g//:

This equality amounts to the commutativity of the diagram

Lie.G/.A/
Lie.c.g//
/ Lie.G/.A/

 c.g/A./ 
hG .A.// / hG .A.//
O O

c.g/A
hG .A/ / hG .A/,

which is granted by the functoriality of our constructions.

We want to examine in detail the case in which V D k mjn so that GL.V / D GLmjn .
Example 11.5.4. We want to see that in the case of GLmjn the Lie bracket Œx; y ´
ad.x/y coincides with the bracket defined in Example 11.3.3 (1). We have

Ad W GLmjn .A/ ! GL.Lie.GLmjn //.A/ D GL.Mmjn .A//; g 7! Ad.g/:

Since hGLmjn .i/ W GLmjn .A/ ! GLmjn .A.// is an inclusion if we identify GLmjn .A/
with its image, we can write

Ad.g/x D gxg 1 ; x 2 Mmjn .A/:

218 11 Algebraic supergroups

By definition we have Lie.GL.Mmjn //.A/ D f1 C ˇ j ˇ 2 End.Mmjn /.A/g. Thus,

for a; b 2 Mmjn .A/ Š Lie.GLmjn /.A/ D f1 C a j a 2 Mmjn .A/g we have

ad.1 C a/b D .1 C a/b.1  a/ D b C .ab  ba/ D b C Œa; b:

Therefore, ad.1 C a/ D id C ˇ.a/, with ˇ.a/ D Œa; .

It is important to observe that in G.A.// it is customary to write the product of two

elements x and y as xy. However as elements of Lie.G/.A/, their product is written
as x C y (hence the unit is 0 and the inverse of x is x). In order to be able to switch
between these two ways of writing, it is useful to introduce the notation e x . This is
mainly a notational device and should not be interpreted as the exponential morphism
in a strict sense.

Definition 11.5.5. Let  W A./ ! B.˛/ be a superalgebra morphism such that ./ D
˛ and , ˛ are two even indeterminates with square zero. Let x 2 Lie.G/.A/ 
hG .A.//. Define e ˛x D hG ./.x/ 2 hG .B.˛//.

We have the following properties:

(1) e x D x for all x 2 Lie.G/.A/. In fact, e x D hG .id/.x/ D x.
(2) e ˛.xCy/ D e ˛x e ˛y . In fact, e ˛.xCy/ D hG ./.xy/ D hG ./.x/hG ./.y/ D
e ˛x e ˛y .
(3) e .ax/ D e ax for all a 2 A0 .

Proposition 11.5.6. Let G and H be two supergroups and let f W hG ! hH be a

natural transformation. Then

f .e ˛x / D e ˛ Lie.f /x :

Proof. This is immediate from the commutativity of the diagram:

Lie.f /
hG .A.// Lie.G/.A/ / Lie.H /.A/  hH .A.//
x / Lie.f /.x/

hG . / hH . /
 fB.˛/ 
hG .B.˛// / hH .B.˛/
 / e ˛ Lie.f /x .
e ˛x

Remark 11.5.7. One can readily check that if G D GL.V /, Lie.G/ D End.V /, we
have
e x D 1 C x:
 11.5 The Lie superalgebra of a supergroup scheme 219

Proposition 11.5.8. Let the notation be as above. Then

Ad.e x /y D y C Œx; y D .id C  ad.x//y for all x; y 2 Lie.G/.A/:

Proof. By Proposition 11.5.6, we have

Ad.e x / D e  Lie.Ad/x D e  ad.x/ D id C  ad.x/ 2 GL.Lie.G//.A/: 

Lemma 11.5.9. Let the notation be as above and let ,  0 be two even elements with
square zero. Then
0 0 0
e x e  y e x e  y D e  Œx;y 2 hG .A.;  0 //:

Proof. Reasoning as in Proposition 11.5.6 we have

0 0 Ad.e x /y 0
e x e  y e x D e  D e  .yCŒx;y/

by Proposition 11.5.8. Therefore,

0 0 0 0 0
e x e  y e x D e  y e  Œx;y D e  Œx;y e  y ;

which gives the result. 

Proposition 11.5.10. The bracket Π;  is antisymmetric.

Proof. In the proof of the previous proposition we saw that

0 0 0 0 0 0
e  Œx;y D e x e  y e x e  y D e  y e x e C y e x D e  Œy;x :

The result now follows from the next lemma. 

0 0
Lemma 11.5.11. If e  u D e  v , then u D v.

Proof. Clearly if e u D e v , then u D v. Let us consider  W A./ ! A. 0 /,

./ D  0 .  induces D Spec  W Spec A. 0 / ! Spec A./. Clearly j j is a
homeomorphism of the topological spaces since the spectrum of a ring is not influenced
by the presence of the nilpotents. Moreover  is a monomorphism of sheaves. We
claim that hX ./ W hX .A.// ! hX .A. 0 //, hX ./˛ D ˛  , is one-to-one for all
superschemes X, that is, ˛  D ˇ  implies that ˛ D ˇ. The fact j˛j D jˇj is
clear. The statement ˛  D ˇ  follows from the fact that we have a monomorphism of
sheaves. 

Proposition 11.5.12. Let W G ! GL.V / be a morphism of supergroup functors.

Then
Lie. /Œx; y D ŒLie. /.x/; Lie. /.y/:
220 11 Algebraic supergroups

Proof. By Proposition s 11.5.6 and 11.5.8 we have

.e x / D e  Lie. /x
D id C  Lie. /x:

Using Lemma 11.5.9 we have

0 0 0
.e  Œx;y / D .e x / .e  y / .e x / .e  y /:

Hence

id C  0 Lie. /Œx; y
D .id C  Lie. /x/.id C  0 Lie. /y/.id   Lie. /x/.id   0 Lie. /y/;

which immediately gives

Lie. /Œx; y D ŒLie. /.x/; Lie. /.y/: 

Proposition 11.5.13. The bracket Π;  satisfies the Jacobi identity.

Proof. In the previous proposition take D Ad. Then we have

Œad.x/; ad.y/ D ad.Œx; y/ for all x; y 2 Lie.G/.A/;

which gives us

ad.x/ ad.y/z  ad.y/ ad.x/z D ad.Œx; y/z;

ŒxŒy; z  Œy; Œx; z D ŒŒx; y; z;

which is the Jacobi identity. 

Corollary 11.5.14. The natural transformation Π;  W Lie.G/  Lie.G/ ! Lie.G/

defined as
Œx; y ´ ad.x/y; x; y 2 Lie.G/.A/;

is a Lie bracket for all A. Lie.G/ is a Lie algebra functor and is represented by a Lie
superalgebra (still denoted by Lie.G/).

Proof. Immediate from previous propositions. 

Corollary 11.5.15. If W G ! GL.V / is a morphism of supergroups, then Lie. /:

Lie.G/ ! Lie.GL.V // Š End.V / is a morphism of Lie algebra-valued functors and
corresponds to a Lie superalgebra morphism of their representing Lie superalgebras.
 11.6 Affine algebraic supergroups 221

11.6 Affine algebraic supergroups

Let k be a field.
We now want to restrict our attention and assume that the supergroup scheme G
is an affine algebraic supergroup. In this case our supergroup G can be effectively
replaced by its (finitely generated) Hopf superalgebra that we shall denote by kŒG.
Let , , S denote respectively the comultiplication, counit and antipode of kŒG.
Definition 11.6.1. We define the additive map D W kŒG ! kŒG to be a super deriva-
tion of kŒG if it satisfies the following properties:
(1) D is k-linear, i.e., D.a/ D 0 for all a 2 k.
(2) D satisfies the Leibniz identity, D.fg/ D D.f /g C .1/p.D/p.f / f D.g/.
D is called left-invariant if further:
(3) B D D .id ˝ D/ B , where denotes the comultiplication in kŒG.
Similarly we say that the additive map d W kŒG ! k is a superderivation if it is
k-linear and satisfies the Leibniz identity

d.fg/ D d.f /.g/ C .1/p.d /p.f / .f /d.g/:

Notice that in the definition of superderivation d W kŒG ! k the sign .1/p.d /p.f /
can be omitted: in fact, whenever p.f / D 1, we have .f / D 0 since  is a morphism
and therefore it preserves the parity.
We denote by Der.kŒG; kŒG/ the set of superderivations D W kŒG ! kŒG and b
Der.kŒG; k/ the set of superderivations d W kŒG ! k.
Proposition 11.6.2. The set L.G/ of left-invariant derivations of kŒG is a Lie super-
algebra with bracket

ŒD1 ; D2  ´ D1 D2  .1/p.D1 /p.D2 / D2 D1 :

Proof. The fact that Π;  is a superbracket is a straightforward calculation. We only

want to check that for all D1 , D2 left-invariant derivations, ŒD1 ; D2  is a left-invariant
derivation, i.e., it satisfies the properties (1), (2), (3) as in Definition 11.6.1. (1) is clear,
(2) is a tedious calculation that we leave to the reader as an exercise. As for (3), since
D1 and D2 are left-invariant derivations, we can write

. B D1 D2 /f D . B D1 /.D2 f /
D .id ˝ D1 B B D2 /.f /
D ..id ˝ D1 / B .id ˝ D2 / B /.f /
D ..id ˝ D1 D2 / B /.f /:

Similarly,

B .1/p.D1 /p.D2 / D2 D1 D .id ˝ .1/p.D1 /p.D2 / D2 D1 / B :

222 11 Algebraic supergroups

Hence
B ŒD1 ; D2  D .id ˝ ŒD1 ; D2 / B ;
as we wanted to show. 
In Chapter 7 we have seen the identification between the left-invariant vector fields
and the Lie superalgebra of a Lie supergroup. In the algebraic context the next theo-
rem provides an analogy establishing a one-to-one correspondence between the left-
invariant derivations and the Lie superalgebra of an affine algebraic supergroup.
Theorem 11.6.3. Let G be an affine supergroup scheme. Then we have natural bijec-
tions among the sets:
(1) L.G/ the left-invariant derivations in Der.kŒG; kŒG/,
(2) Der.kŒG; k/,
(3) Lie.G/.
Proof. Let us examine the correspondence between (1) and (2). We want to construct
an invertible map  W Der.kŒG; k/ ! L.G/. Let d 2 Der.kŒG; k/. Define .d / D
.id ˝ d / . Such .d / is left-invariant; in fact,
. B .d //.f / D .
B id ˝ d B /.f /
P .1/
D . f d.f .2/ //
P
D .id ˝ id ˝ d /. .f .1/ / ˝ f .2/ /
D ..id ˝ id ˝ d / B . ˝ id/ B /.f /;
P
where we use the Sweedler notation: .f / D f .1/ ˝ f .2/ . On the other hand we
have
.id ˝ .d // .f / D ..id ˝ id ˝ d / B .id ˝ / B /.f /
D .id ˝ .id ˝ d B / B /.f /
which is the same as before since by the coassociativity axiom in a Hopf superalgebra
. ˝ id/ B D .id ˝ / B .
Moreover, .d / is a derivation. In fact, since d is a derivation, it follows that
.d /.fg/ D .id ˝ d / B .fg/
P
D f .1/ g .1/ d.f .2/ g .2/ /
P
D f .1/ g .1/ .d.f .2/ /.g .2/ / C .1/p.d /p.f / .f .2/ /d.g .2/ //:
We can rewrite the last expression as (m denotes the multiplication)
P
m. .f .1/ ˝ d.f .2/ //.g .1/ ˝ .g .2/ ///
C .1/p.d /p.f / m..f .1/ ˝ .f .2/ //.g .1/ ˝ d.g .2/ ///
D .id ˝ d / .f /.id ˝ / .g/ C .1/p.d /p.f / .id ˝ / .f /.id ˝ d / .g/
D .d /.f /g C .1/p.d /p.f / f .d /.g/;
 11.7 Linear representations 223

since we recall that, by the Hopf superalgebra axioms, .id ˝ / B D id. Then
.d / 2 Der.kŒG; kŒG/. Vice versa, if D 2 L.G/ define .D/ D  B D. We leave
to the reader, as an easy exercise, the check that is a derivation. One can check
also that is the inverse of . We now want a correspondence between (b) and (c).
By Theorem 11.4.3 we have Lie.G/ D Hom.smod/ .T1G .G/ ; A/ D Der.OG;1G ; k/.
Observe that, as in the ordinary case,
Der.OG;1G ; k/ D Der.kŒG; k/;
that is, the derivation on the localization of the ring kŒG is determined by the derivation
on the ring itself. 

11.7 Linear representations

We now want to discuss linear representations. In particular we will show that, as in
the classical case, every affine algebraic supergroup G can be embedded into some
GLmjn .
We start by recalling the notion of closed embedding discussed in the previous
chapter.
Definition 11.7.1. Let X D .jX j; OX / and Y D .jY j; OY / be two superschemes and
let f W X ! Y be a superscheme morphism. We say that f is a closed embedding if
the topological map jf j W jX j ! jY j is a homeomorphism of the topological space jX j
onto its image in jY j and the sheaf map f  W OY ! f OX is a surjective morphism of
sheaves of superalgebras.
This means that we may identify X with a closed subscheme of Y , so its sheaf is
identified with OY = for some quasi-coherent sheaf of ideals . If both X and Y are
affine superschemes we have by Remark 10.1.7 that f is a closed embedding if and
only if O.X/ Š O.Y /=I for some ideal I .
We now want to introduce the notion of linear representation of a supergroup. Let
hG W .salg/ ! .sets/, hG .A/ D Hom.kŒG; A/ be the functor of points of our affine
algebraic supergroup G, with Hopf superalgebra kŒG.
Definition 11.7.2. Let V be a super vector space. We define linear representation of
G in V to be a natural transformation , which is a morphism of supergroup functors:
W hG ! GL.V /:
Here GL.V / is, as usual, the functor
GL.V / W .salg/ ! .sets/; GL.V /.A/ D GL.A ˝ V /;
and GL.A ˝ V / denotes the automorphisms of the A-module A ˝ V preserving the
parity. We will also say that G acts on V .
224 11 Algebraic supergroups

Definition 11.7.3. Let V be a super vector space. Then V is said to be a left G-comodule
if there exists a linear map
V W V ! kŒG ˝ V;

called a comodule map, with the properties

(1) .idG ˝ V/ V D. ˝ idV / V,

(2) . ˝ idV / V D idV ,

where idG W kŒG ! kŒG is the identity map.

One can also define a right G-comodule in the obvious way.

Observation 11.7.4. The two notions of G acting on V and V being a (left) G-

comodule are essentially equivalent. In fact, given a representation W G ! GL.V /,
it defines immediately a comodule map

V .v/ D kŒG .idG /v; idG 2 hG .kŒG/ D Hom.salg/ .kŒG; kŒG/;

where we are using the natural identification (for A D kŒG)

GL.V /.A/  End.V /.A/ Š Hom.smod/ .V; A ˝ V /:

Vice versa if we have a comodule map V we can define a representation

AW hG .A/ ! GL.V /.A/  Hom.smod/ .V; A ˝ V /; g 7! v 7! .g ˝ id/. V .v//;

where g 2 hG .A/ D Hom.salg/ .kŒG; A/.

Let us look at this correspondence in a special but important case.

Example 11.7.5. Let us consider the natural action of GLmjn on k mjn :

X
AW GLmjn .A/ ! GL.k mjn /.A/; g D .gij / 7! ej 7! gij ˝ ei ;

Here fej g is the canonical homogeneous basis for the super vector space k mjn . We
identify the morphism g 2 GLmjn .A/ D Hom.salg/ .kŒGLmjn ; A/ with the matrix with
entries gij D g.xij / 2 A, where xij ’s are the generators of kŒGLmjn .
This corresponds to the comodule map
P
k mjn W k ! kŒGLmjn  ˝ k mjn ; ej 7! xij ˝ ei :
mjn

P
Vice versa, the comodule map ej 7! xij ˝ ei corresponds to the representation

AW GLmjn .A/ ! GL.k mjn /.A/;

P P
g D .gij / 7! ej 7! .g ˝ id/. xij ˝ ei / D gij ˝ ei :
 11.7 Linear representations 225

Definition 11.7.6. Let G act on the superspace V via a representation corresponding

to the comodule map V . We say that the subspace W  V is G-stable if V .W / 
kŒG ˝ W . Equivalently W is G-stable if A .g/.A ˝ W /  A ˝ W .
Definition 11.7.7. The right regular representation of the affine algebraic group G is
the representation of G in the (infinite-dimensional) super vector space kŒG corre-
sponding to the comodule map

W kŒG ! kŒG ˝ kŒG:

Proposition 11.7.8. Let be a linear representation of an affine algebraic supergroup

G. Then each finite-dimensional subspace W of V is contained in a finite-dimensional
G-stable subspace of V .
Proof. It is the same as in the commutative case. Let us sketch it for the case of a right
representation, the case of left being the same. It is enough to prove for W generated
by one element x 2 V . Let V W V ! V ˝ kŒG be the comodule structure associated
to the representation . Let
X
V .x/ D xi ˝ ai
i

where fai g is a basis for kŒG.

We claim that spank fxi g is a G-stable subspace.
By definition of comodule we have

. V ˝ idG /. V .x// D .idV ˝ /. V .x//;

that is, X X X
V .xj / ˝ aj D xj ˝ .aj / D xj ˝ bij ˝ ai :
j j i;j

Hence X
V .xi / D xj ˝ bij :
j

The finite-dimensional stable subspace is given by the span of the xi ’s.

Moreover we have x 2 spank fxi g. In fact by property .2/ of Definition 11.7.3 we
have X
.id ˝ / V .x/ D idV .x/; that is, x D xj .bij /: 

Theorem 11.7.9. Let G be an affine supergroup variety. Then there exists a closed
embedding of algebraic supergroups

G  GLmjn

for suitable m and n.

226 11 Algebraic supergroups

Proof. By Definition 11.7.1 we need to find a surjective superalgebra morphism

kŒGLmjn  ! kŒG for suitable m and n. Let kŒG D kŒf1 ; : : : ; fn , where fi are
homogeneous and chosen so that W D spanff1 ; : : : ; fn g is G-stable, according to the
right regular representation. This choice is possible because of Proposition 11.7.8. We
have X X
kŒG .fi / D fj ˝ aij ; kŒG .aij / D aik ˝ akj :
j j
Define the morphism
 W kŒGLmjn  ! kŒG; xij 7! aij ;
where xij are the generators for kŒGLmjn . This is the required surjective superalgebra
morphism. In fact, since kŒG is both a right and left G-comodule, we have
P P
fi D . ˝ id/ .fi / D . ˝ id/. j fj ˝ aij / D j .fj /aij ;
which proves the surjectivity. This shows that we have an embedding of G into GLmjn
as superschemes. In order to see that this is also a supergroup morphism it is enough to
check that  is a Hopf superalgebra morphism, but this is a straightforward verification.

We thus have proved the following.
Corollary 11.7.10. G is an affine supergroup scheme if and only if it is a closed
subgroup of GLmjn for suitable m and n.
Proposition 11.7.11. Let the G be an affine algebraic supergroup and H a closed
subsupergroup, i.e., kŒH  D kŒG=I . Then there exists a finite-dimensional repre-
sentation of G, with super vector space V and a subspace W  V , which is fixed
by H .
Proof. Let Vy D kŒG and W D I . We have immediately defined two comodule maps
Vy
D G , W D G jI where G is the comultiplication in the Hopf algebra kŒG:

Vy
W Vy ! Vy ˝ kŒG; W
W W ! W ˝ kŒH :
The second map is induced by the first one, in fact
G .I / D I ˝ kŒG C kŒG ˝ I
since I is a Hopf ideal. It is simple to check from the definitions that such a comodule
map corresponds to the following action of G on the super vector space kŒG:
.g  f /.x/ D f .xg/; g; x 2 G.A/; f 2 kŒG;
where customarily we denote x.f / by f .x/ and as usual xg D m B .x ˝ g/ B (see
Proposition 11.1.2).
By Proposition 11.7.8 we can find a finite-dimensional subspace V invariant under
the action of G and containing all the generators of kŒG and a subspace W D I \ V
containing the generators of I , hence fixed by H . 
 11.8 The algebraic stabilizer theorem 227

11.8 The algebraic stabilizer theorem

In this section we want to prove the stabilizer theorem, which asserts the representability
of the stabilizer functor for the action of an affine supergroup on an affine supervariety.
This is important since it allows us to prove that all the classical algebraic supergroups
are representable and to compute explicitly their Lie superalgebras.
Definition 11.8.1. We say that an algebraic supergroup G acts on a superscheme X
if we have a morphism W G  X ! X corresponding to the functorial family of
morphisms

AW hG .A/  hX .A/ ! hX .A/; .g; x/ 7! g  x for all g 2 hG .A/; x 2 hX .A/;

satisfying the following properties:
(1) 1  x D x for all x 2 hX .A/.
(2) .g1 g2 /  x D g1  .g2  x/ for all g1 ; g2 2 hG .A/ and x 2 hX .A/.
Let u be a rational topological point of X , that is, u 2 jX j, or equivalently u 2
hX .k/ D Hom.O.X /; k/. Let mu be the maximal ideal corresponding to u in OX;u .
Notice that u can be viewed naturally as an A-point uA for all superalgebras A since
k  A. So we have a morphism
 W G ! X; g 7! g  uA ; g 2 hG .A/; uA 2 hX .A/:
If G and X are affine, we can write equivalently a morphism
Q W O.X / ! O.G/:
Definition 11.8.2. We call stabilizer supergroup functor of the point u 2 jX j with
respect to the action the group-valued functor Stabu W .salg/ ! .sets/ defined by
Stabu .A/ D fg 2 hG .A/ j A .g/ D g  uA D uA g;
where A W hG .A/ ! hX .A/, or equivalently if G and X are affine,
Stabu .A/ D fg 2 hG .A/ D Hom.O.G/; A/ j g B Q D uA g:
We want to prove that this functor is representable by an affine supergroup.
Theorem 11.8.3. Let G be an affine supergroup acting on an affine supervariety X
and let u be a topological rational point of X . Then Stabu is an affine supergroup.
Proof. The stabilizer can be described in an equivalent way as
Stabu .A/ D fg 2 hG .A/ j .g B Q /jmu D 0g;
where mu  O.X / is the ideal of the topological point u. Let I be the ideal in
O.G/ generated by .x/
Q for all x 2 mu . One can immediately check that g 2
hG .A/ D Hom.O.G/; A/ is in Stabu .A/ if and only if g factors via O.G/=I , that is,
g W O.G/ ! O.G/=I ! A. So we have Stabu .A/ D Hom.O.G/=I; A/. 
228 11 Algebraic supergroups

We now want to describe some important applications of this result, namely, the
representability of the supergroup functors corresponding to the classical Lie superal-
gebras (for a complete description of such superalgebras we refer to Appendix A).

1. A.mjn/ series. A.mjn/ consists of the matrices in the super vector space
M.m C 1jn C 1/ with supertrace zero. This is a Lie superalgebra with the bracket
induced by the one in M.m C 1jn C 1/.
Consider the morphism

W GLmjn k 1j0 ! k 1j0 ; .g; c/ 7! Ber.g/c:

The multiplicative property of Ber ensures this is a well-defined action of GLmjn on

k 1j0 . The stabilizer of the point 1 2 k 1j0 coincides with all the matrices in GLmjn .A/
with Berezinian equal to 1, that is, the special linear supergroup SLmjn .A/. From
Theorem 11.8.3 it follows immediately that SLmjn is representable. Moreover in Ex-
ample 11.3.3 we checked that A.mjn/ D Lie.SLmC1jnC1 /.

2. B.pjq/, C.p/, D.pjq/ series. Consider the morphism

W GLmjn B ! B; .g; .; // 7! .g; g/;

where B is the super vector space of all the symmetric bilinear forms on k mj2n . Con-
sider the point in B:
0 1
0 Ip 0 0 0
BIp 0 0 0 0C
B C
ˆDB
B0 0 1 0 0C C if m D 2p C 1, n D 2q;
@0 0 0 0 Iq A
0 0 0 Iq 0

or
0 1
0 Ip 0 0
BIp 0 0 0C
ˆDB
@0
C if m D 2p, n D 2q:
0 0 Iq A
0 0 Iq 0

We define the stabilizer of the point ˆ to be the supergroup functor Ospmj2n . Again
this is an algebraic supergroup by Theorem 11.8.3.
In order to compute its Lie superalgebra, let us consider a generic matrix M D
I CX
W I CZ in M.mjn/ Š Lie.GLmjn /. We are assuming m D 2p even since for
Y

m odd the calculation is very much the same. The condition for M to belong to
 11.8 The algebraic stabilizer theorem 229

Lie.Osp2pj2q / is M t ˆM D ˆ, which is
0 1
  0 Ip 0 0  
I C X t BIp 0C
W t B 0 0 C I C X Y
Y t I C Z t @0 0 0 Iq A W I C Z
0 0 Iq 0
0 1
0 Ip 0 0
BIp 0 0 0C
DB
@0
C:
0 0 Iq A
0 0 Iq 0

After a tedious calculation one finds that the generic matrix M in Lie.Osp2pj2q / has
the form 80 19
ˆ
ˆ a b x x1 > >
<B =
B ct at
t
y y1 C C ;
ˆ@ y1 x1 d e A> >
:̂ ;
y x f d
t t

where the matrices b and c are skewsymmetric and f and e are symmetric and the
entries of all the matrices a, b, c, d , e, f , x, x1 , y, y1 , u, v are in k. This shows that
Lie.Osp2pj2q / D osp.2pj2q/, in fact the matrices in osp.2pj2q/ have precisely the
form prescribed above.
So we have

B.pjq/ ´ osp.2p C 1j2q/ D Lie.Osp2pC1j2q /;

C.q/ ´ osp.2j2q  2/ D Lie.Osp2j2q2 /
and
D.pjq/ ´ osp.2pj2q/ D Lie.Osp2pj2q /:

3. P .n/ series. Define the algebraic supergroup Sp njn , as we did for Ospmjn by
taking antisymmetric bilinear forms instead of symmetric ones. Consider the action

Spnjn  k 1j0 ! k 1j0 ; .g; c/ 7! Ber.g/c:

By Theorem 11.8.3 we have that Stab1 is an affine algebraic supergroup, hence it is a

Lie supergroup. It corresponds to the P .n/ series of classical Lie superalgebras (see
Appendix A).
4. Q.n/ series. Let D D kŒ=.2 C 1/. This is a noncommutative superalgebra.
Define the supergroup functor GLn .D/ W .salg/ ! .sets/, with GLn .D/.A/ the group
of automorphisms of the left supermodule A ˝ D. In [22] is proven the existence of a
morphism, called the odd determinant,

odet W GLn .D/ ! k 0j1 :

230 11 Algebraic supergroups

Reasoning as before, define

GLn .D/  k 0j1 ! k 0j1 ; .g; c/ 7! odet.g/c:

Then G D Stab1 is an affine algebraic supergroup, and for n  2 we define Qg.n/

as the quotient of G and the diagonal subgroup GL1j0 . This is an algebraic and Lie
supergroup and its Lie superalgebra is Q.n/.

11.9 References
In [44] Kac proved a classification theorem for simple Lie superalgebras that we have
summarized in Appendix A. The description of the supergroup functors, corresponding
to the classical super series of Lie superalgebras introduced by Kac, appeared in [22],
p. 70; however no representability statement of the supergroups was proved there. In
[14], [15] all the classical supergroup functors together with the Hopf superalgebras
representing them are described in detail.
A

Lie superalgebras

The purpose of this appendix is to give a brief introduction to the theory of Lie superal-
gebras and, in particular, to the theory of representations of classical Lie superalgebras.
This material is well known and is mostly found in the works of Kac, [44], [45]. This
appendix is self-contained and does not require knowledge of any material of the pre-
vious chapters. However, we assume basic knowledge of semisimple Lie algebras.
We start by describing the classification of simple finite-dimensional complex Lie
superalgebras. We do not provide any details on the proof of the main Theorem A.1.7,
but we give a full description of the classical families of simple Lie superalgebras,
including the root systems, Cartan matrices, and Dynkin diagrams.
We then discuss the finite-dimensional representations of classical Lie superalge-
bras. In the ordinary setting, the finite-dimensional modules for the special linear Lie
algebra sl2 play a key role. Ultimately, for a Lie algebra g, the conditions to impose
to have a finite-dimensional g-module, boil down to the conditions to impose in order
to have finite-dimensionality for the sl2 -modules, corresponding to simple roots. The
situation for classical Lie superalgebras is different. In the ordinary setting, the root
vectors X˛ and X˛ , corresponding to a pair of opposite roots ˛ and ˛, always gen-
erate an sl2 subalgebra inside the given simple Lie algebra. In contrast, given a simple
Lie superalgebra g and a root ˛ of g, the root vectors corresponding to roots propor-
tional to ˛ may generate a subalgebra of g isomorphic to sl2 , osp.1j2/, sl.1j1/, sq.2/,
or even a nilpotent subalgebra. Hence in order to fully understand the representation
theory of Lie superalgebras it is necessary to study in detail and classify all irreducible
representations of all these Lie superalgebras. We shall describe the irreducible rep-
resentations for sl2 , osp.1j2/, sl.1j1/, and then we will give the main theorem on the
classification of finite-dimensional representations of basic classical Lie superalgebras,
describing in detail the case of osp.2m C 1j2n/. This example is very illuminating
since we can see how the extra conditions for the finite-dimensionality of the modules
make their appearance once we take into account all Borel subalgebras at once. In the
ordinary setting this is not necessary since if we fix a Borel subalgebra, all other Borel
subalgebras are conjugate to the fixed one under the action of the Weyl group. For
Lie superalgebras, the Weyl group, still a classical object, appears to be “too small”
and not all Borel subalgebras are conjugate. For this reason, in order to proceed to the
classification of finite-dimensional representations, we need to use the odd reflections
to compensate for the lack of enough symmetries in the Weyl group. In Section A.5
we will see explicitly in the example of an orthosymplectic Lie superalgebra how this
happens.
232 A Lie superalgebras

A.1 Classical Lie superalgebras

We assume the ground field k is algebraically closed and of characteristic zero. For
concreteness the reader may think of k as C.
We recall a few definitions from Chapter 1, since here our notation is slightly
different to adhere to the conventions in representation theory of Lie superalgebras.
Throughout this appendix, by a Lie superalgebra, we mean a super vector space
g D g0 ˚ g1 with a bracket Œ ;  W g  g ! g which is a bilinear even map and satisfies
the following properties:
(1) Anti-symmetry: Œx; y C .1/jxjjyj Œy; x D 0 for homogeneous elements x, y
in g.
(2) Jacobi identity:

Œx; Œy; z C .1/jxjjyjCjxjjzj Œy; Œz; x C .1/jyjjzjCjxjjzj Œz; Œx; y D 0

for homogeneous elements x; y; z 2 g. j  j as usual denotes the parity.

The most important example of Lie superalgebra is the algebra of endomorphisms
of a super vector space.
Let gl.V / be the super vector space of endomorphisms of the super vector space V .
If V D k mjn , we denote gl.V / by gl.mjn/. The even part gl.mjn/0 of gl.mjn/ consists
of the matrices with entries in k corresponding to endomorphisms preserving the parity,
while the odd part gl.mjn/1 consists of the matrices corresponding to endomorphisms
reversing parity:
² ³ ² ³
A 0 0 B
gl.mjn/ D gl.mjn/0 ˚ gl.mjn/1 D ˚ :
0 D C 0

Here A and D are .m  m/ and .n  n/-matrices, and B and C are .m  n/ and

.n  m/-matrices, all with entries in k.
Then gl.mjn/ is a Lie superalgebra with bracket

ŒX; Y  D X Y  .1/jXjjY j YX:

We now want to define the analogue of the special linear Lie algebra.

Definition A.1.1. We define the special linear Lie superalgebra, sl.mjn/ and the pro-
jective special linear Lie superalgebra psl.mjm/ as

sl.mjn/ ´ fX 2 gl.mjn/ j str.X / D 0g;

A B D tr Atr D, psl.mjm/ ´ sl.mjm/=kI
where str C D 2m . Here str is the supertrace
(see also Chapter 1) and I2m is the identity .2m  2m/-matrix. One can easily check
that these are Lie superalgebras with brackets induced by the bracket in gl.mjn/.
 A.1 Classical Lie superalgebras 233

As for the even and odd parts, we have

sl.mjn/0 D slm ˚ sln ˚ k; sl.mjn/1 D Vm ˝ Vn0 ˝ k ˚ Vm0 ˝ Vn ˝ k;
psl.mjm/0 D slm ˚ slm ; psl.mjm/1 D Vm ˝ Vm0 ˚ Vm0 ˝ Vm ;
where Vm and Vm0 denote respectively the defining representation of slm and its dual.
We now seek the analogue for the orthogonal and the symplectic Lie algebras.
Classically, given a non-degenerate bilinear form f on a vector space V , we can define
the orthogonal and the symplectic Lie algebras in the following way:
so.V / ´ fA 2 gl.V /jf .Av; u/ D f .v; Au/g if f is symmetric;
sp.V / ´ fA 2 gl.V /jf .Av; u/ D f .v; Au/g if f is skew-symmetric:
In the super case, as we shall presently see, the two notions come together, since
the condition of symmetry in the odd variables for a bilinear form brings in a minus
sign, hence becomes a skew-symmetric condition.
Definition A.1.2. We say that a bilinear form f on a super vector space V D V0 ˚ V1
is super symmetric (or symmetric for short) if
f .u; v/ D .1/juj jvj f .v; u/
for any homogeneous elements u; v 2 V . We say also that f is consistent if f .u; v/ D
0 for u 2 V0 and v 2 V1 .
Hence, as one can readily see, a super symmetric bilinear form f on V gives a
symmetric form f0 on V0 and a skew-symmetric form f1 on V1 .
Definition A.1.3. Let f be a non-degenerate consistent super symmetric bilinear form
on V , dimV D mjn. We define the orthosymplectic Lie superalgebra as
osp.V / ´ fX 2 gl.V / j f .Xu; v/ D .1/jXj juj f .u; Xv/g:
Notice that n has to be even since f defines a non-degenerate skew-symmetric form
on V1 .
The standard super symmetric form  in V D k mjn corresponds to the following
matrices:
0 1
0 Ip 0 0 0
BIp 0 0 0 0C
B C
DB B0 0 1 0 0C C if m D 2p C 1, n D 2q;
@0 0 0 0 Iq A
0 0 0 Iq 0
0 1
0 Ip 0 0
BIp 0 0 0C
DB @0 0
C if m D 2p, n D 2q:
0 Iq A
0 0 Iq 0
234 A Lie superalgebras

Fixing  as above, the matrix form for the orthosymplectic Lie superalgebra, which
we shall denote by osp.mjn/, becomes the following:
1. for m D 2p C 1, n D 2q,
80 19
ˆ
ˆ a b u x x1 > >
ˆB
ˆ >
<B c at v y y1 C
C=
>
osp.mjn/ D B B vt
t
ut 0 z C
z1 C I
ˆ
ˆ >
ˆ@ y1 x1t z1t d e A> >
>
:̂ t ;
y t x t z t f d
2. for m D 2p, n D 2q,
80 19
ˆ
ˆ a b x x1 > >
<B =
c at y y1 C
osp.mjn/ D B C ;
ˆ@ y1t x1t d e A> >
:̂ ;
y t x t f d t
where a, b, c, d , e, f , x, x1 , y, y1 , u, v are matrices of appropriate size and the
matrices b and c are skew-symmetric and e and f are symmetric; all matrices
have entries in k.
Notice that if n D 0, then osp.mj0/ is the orthogonal Lie algebra Bp or Dp de-
pending on the parity of m, while if m D 0, osp.0jn/ is the symplectic Lie algebra
Cq .
The decomposition of osp.mjn/ into even and odd parts is
osp.mjn/0 D Bp ˚ Cq if m D 2p C 1, n D 2q;
osp.mjn/0 D Dp ˚ Cq if m D 2p, n D 2q;
osp.mjn/1 D Vm ˝ Vn ;
where again Vm denotes the defining representation of so.m/ and Vn the defining
representation of Cq .
Next we have the Lie superalgebras in the strange series P .n/ and Q.n/.
Definition A.1.4. We define the strange series P .n/ as
² ³
A B
P .n/ D  gl.n C 1jn C 1/;
C At
where A 2 sl.n C 1/, B is symmetric and C skew-symmetric.
The strange series Q.n/ is defined as follows. Set
² ³
A B
q.n/ D I
B A
sq.n/ are the matrices in q.n/ with tr.B/ D 0 and Q.n  1/ D psq.n/ D sq.n/=kI2n ,
i.e., ²  ³
A B
Q.n  1/ D j B 2 sln =kI2n :
B A
 A.1 Classical Lie superalgebras 235

Again one can check that these are well-defined Lie superalgebras with the bracket
induced from gl.njn/.
We have
V 0
P .n/0 D sl.n C 1/; P .n/1 D Sym2 .VnC1 / ˚ 2 .VnC1 /;
Q.n/0 D sl.n C 1/; Q.n/1 D ad.sl.n C 1//;
where VnC1 and ad.sl.n C 1// are respectively the defining representation and the
adjoint representations of sl.n C 1/.
The simple finite-dimensional Lie superalgebras over k fall into several classes. To
introduce these classes we need a few more definitions.
Definition A.1.5. Let g be a Lie superalgebra (always finite-dimensional). We say
that g is simple if g is not abelian and it admits no non-trivial ideals. g is classical if
it is simple and g1 is completely reducible as a g0 -module, where the action is given
by the bracket. g is basic if it is classical and it admits a consistent, non-degenerate,
invariant bilinear form, that is to say, there exists a consistent, non-degenerate, bilinear
form h ; i W g  g ! g such that hX; ŒY; Zi D hŒX; Y ; Zi.
Observation A.1.6. If h ; i is a bilinear invariant form on a simple Lie superalgebra
g, then h ; i is either identically zero or non-degenerate. In fact if X 2 kerh ; i, i.e., if
X is such that hX; Y i D 0 for all Y 2 g, we have
0 D hX; ŒY; Zi D hŒX; Y ; Zi:
Hence the kernel of h ; i is an ideal, hence it is all of g or zero.
The simple Lie superalgebras divide into two main types: the classical type, when
the action of g0 on g1 is completely reducible, and the Cartan type. We make a list
of such Lie superalgebras, referring the reader to [44] for all proofs regarding the
classification.

1 Classical type. The classical type subdivides further into type 1 and type 2. Type 1
classical superalgebras are those for which g1 is not irreducible as g0 -module and
type 2 are those for which g1 is an irreducible g0 -module.

1.1 Classical type 1. The type 1 superalgebras are

A.mjn/ ´ sl.m C 1jn C 1/; m ¤ n;
A.mjm/ ´ psl.m C 1jm C 1/;
C.n/ ´ osp.2j2n  2/; P .n/:
For these superalgebras g1 decomposes into two components as a g0 -module so that
g admits a compatible Z-grading (where “compatible” in this case means with respect
to the Z2 -grading):
g D g1 ˚ g0 ˚ g1 ; g0 D g0 ; g1 D g1 ˚ g1 :
236 A Lie superalgebras

Notice that the upper indices refer to the Z-grading, while the lower indices to the
Z2 -grading.
For A.mjn/ and C.n/ we have that g1 and g1 are dual to each other; in fact,

A.mjn/0 D slmC1 ˚ slnC1 ˚ k;

0
A.mjn/1 D VmC1 ˝ VnC1 ˝ k;
A.mjn/1 D VmC1
0
˝ VnC1 ˝ k;
C.n/0 D Cn1 ; C.n/1 D Vn ; C.n/1 D Vn0 Š Vn ;

where Vi always denotes the natural representation of the Lie algebra under consider-
ation and Vi0 its dual.
For g D P .n/, g1 and g1 are not dual to each other. We have
V2
g1 D Sym2 VnC1 ; g1 D 0
VnC1 :

1.2 Classical type 2. The type 2 superalgebras are those for which g1 is irreducible,
so that there is no compatible Z-grading. These Lie superalgebras are

B.mjn/ D osp.2m C 1j2n/;

D.mjn/ D osp.2mj2n/;
D.2; 1I ˛/; F .4/; G.3/; Q.n/:

D.2; 1I ˛/ is a family with a continuous parameter ˛ 2 k n f0; 1g. Two elements

D.2; 1I ˛/, D.2; 1I ˇ/ of this family are isomorphic if and only if ˛ and ˇ lie in the
same orbit under the action of the group of order 6 generated by ˛ 7! 1˛, ˛ 7! 1=˛
(see [44], 2.5).
We have g0 D sl2 ˚ sl2 ˚ sl2 , g1 D V2 ˝ V2 ˝ V2 . Without going into detail
we want to remark that the bracket restricted to g1  g1 depends on the parameter ˛,
hence different superalgebras D.2; 1I ˛/ cannot be recognized by the structure of g1
as a g0 -module alone.
For the definition of the exceptional Lie superalgebras F .4/ and G.3/ see [44].

2 Cartan type. Let Sym.V / denote the symmetric algebra of the super vector space
V . If V has dimension mjn and we fix a basis, we can identify Sym.V / with A D
kŒx1 ; : : : ; xm ; 1 ; : : : ; n , the polynomial algebra with m even indeterminates and n
odd ones. We define W .mjn/ D Der.A/ as the superalgebra of derivations of A. In
general W .mjn/ is infinite-dimensional, however when m D 0, it is finite-dimensional.
We write W .n/ for W .0jn/:
n X o
W .n/ D aI;j i1 : : : il @j :
aI;j 2k
 A.2 Root systems 237

W .n/ D W .n/0 ˚ W .n/1 , where W .n/0 corresponds to those elements with l odd,
while W .n/1 to those with l even. Define ‚.n/ as the associative superalgebra over
A generated by  1 ; : : : ;  n with relations  i ^  j D  j ^  i (i ¤ j ). This
is a superalgebra with grading induced by deg. i / D 1. Furthermore, W .n/ acts on
‚.n/ by derivations. We define S.n/ and Sz.n/ as the subalgebras of W .n/ annihilating
certain elements of ‚.n/ called volume forms. Namely,

S.n/ ´ fD 2 W .n/ j D. 1 ^    ^  n/ D 0g;

z
S.n/ ´ fD 2 W .n/ j D..1 C 1 2 : : : n / 1 ^  ^  n/ D 0g for even n:

Finally, we define H.n/ as the commutator of a subalgebra of W .n/ preserving a certain

metric, i.e.,
z .n/; H
H.n/ D ŒH z .n/ where H
z .n/ ´ fD 2 W .n/ j D.d 2
1 C  C d 2
n/ D 0g:

We end our discussion on simple Lie superalgebras with a theorem:

Theorem A.1.7. Every simple finite-dimensional Lie superalgebra over k is isomor-
phic to one of the following:
(1) the classical Lie superalgebras, either isomorphic to a simple Lie algebra or to
one of the following classical Lie superalgebras:

A.mjn/; B.mjn/; C.n/; D.mjn/; P .n/; Q.n/;

for appropriate ranges of m and n,

F .4/; G.3/; D.2; 1I ˛/ for ˛ 2 k n f0; 1g;

(2) the Lie superalgebras of Cartan type:

W .n/; S.n/; Sz.n/ for even n; H.n/:

In the next section we will discuss the structure of simple Lie superalgebras in more
detail with a particular attention to the families A, B, C , and D.

A.2 Root systems

Similarly to the ordinary setting, for Lie superalgebras we have the notion of Cartan
subalgebras and the corresponding root decomposition.
Definition A.2.1. A subalgebra h  g is a Cartan subalgebra if h is a nilpotent,
self-normalizing Lie subalgebra of g. If ˛ 2 h0 , we define the super vector space

g˛ W D fX 2 g j Œh; X  D ˛.h/X for all h 2 h0 g:

238 A Lie superalgebras

If g˛ ¤ f0g for ˛ 2 h0 nf0g, we say that ˛ is a root and g˛ is its root space. We say
that a root ˛ is even if g˛ \ g0 ¤ f0g, and odd if g˛ \ g1 ¤ f0g. Notice that with this
definition a root can be both even and odd; this can actually happen, as we shall see in
Example A.2.2. We denote by the set of all roots.
In the same way as in the ordinary case, we have
X
gDh˚ g˛ :
˛2

The set of roots  h0 n f0g is the union of even and odd roots: D 0[ 1, where

0 D f˛ 2 h0 n f0g j g˛ \ g0 ¤ f0gg; 1 D f˛ 2 h0 n f0g j g˛ \ g1 ¤ f0gg:

If g is simple, we have h D h0 with the sole exception of Q.n/. Furthermore, it can

happen that 0 D 1 , as we shall see in the next example.
Example A.2.2. According to Definition A.1.4 we have
80 19
ˆ
ˆ h1 e hN 1 eN > >
<B =
B f h2 fN hN 2 C C
q.2/ D @ N  gl.2j2/:
ˆ h1 eN h1 e A> >
:̂ N N ;
f h2 f h2
As a super vector space we have that

q.2/ D spank fh1 ; h2 ; hN 1 ; hN 2 ; e; f; e;

N fNg;

where by a slight abuse of notation we denote a matrix with only two non-zero entries
by the corresponding letter above. Clearly, q.2/0 D spank fh1 ; h2 ; e; f g and q.2/1 D
spank fhN 1 ; hN 2 ; e;
N fNg. The space h D spank fh1 ; h2 ; hN 1 ; hN 2 g is a Cartan subalgebra of
q.2/. One can readily check that we have two root spaces of dimension 1j1:

g˛ D spank fe; eg;

N g˛ D spank ff; fNg;

where ˛.h1 / D 1, ˛.h2 / D 1. Therefore,

g D h ˚ g˛ ˚ g˛ ; 0 D 1:

This somewhat awkward behaviour may suggest excluding Q.n/ from the treat-
ment, however we believe that the right philosophy is to ask questions for all classical
Lie superalgebras. Before we proceed and give some examples of root systems, we
want to make some observations on the Cartan–Killing form.
Observation A.2.3. Let g be a classical Lie superalgebra. In analogy with the Cartan–
Killing form in the ordinary setting, define the bilinear form

.x; y/ D str.ad.x/ ad.y//;

 A.2 Root systems 239

where x; y 2 g. As one can easily check, this form is symmetric and consistent.
However, quite differently from what happens in the classical setting, it is not always
non-degenerate, hence zero. In particular its restriction to a Cartan subalgebra of g
may be degenerate.
For example, on sl.mC1jnC1/, if h is the Cartan subalgebra consisting of diagonal
matrices, we have
.x; y/ D 2.m  n/ str.xy/
for x D diag.a1 ; : : : ; amCn / and y D diag.a10 ; : : : ; amCn
0
/ in h. When m D n the
form is identically zero. Furthermore, it factors to a form on the corresponding Cartan
subalgebra of psl.mjm/, which is also identically equal to zero.
The fact that the Cartan–Killing form of a classical Lie superalgebra may be de-
generate prompts the definition of basic classical Lie superalgebras.
Definition A.2.4. A Lie superalgebra g is basic classical if g is simple, g0 is reductive,
and g admits a non-degenerate invariant symmetric consistent bilinear form.
For example, A.mjn/, C.n/, and B.mjn/ are basic simple Lie superalgebras, while
P .n/ is not. Notice that even though A.mjm/ has a degenerate Cartan–Killing form,
there exists a non-degenerate invariant symmetric bilinear form on A.mjm/.
The following table summarizes the classification of simple Lie superalgebras to-
gether with information about the existence of an invariant non-degenerate symmetric
bilinear form.

Simple Lie superalgebras

Classical Cartan type
Basic Strange
A.mjn/, B.mjn/, C.n/, and D.mjn/
D.1; 2I ˛/, G.3/, F .4/ P .n/, Q.n/ W .n/, S.n/, Sz.n/, H.n/

We now give a brief description of the root systems for the basic classical Lie
superalgebras in the families A, B, C , D. As in the ordinary setting every root system
admits a simple system … D f˛1 ; : : : ; ˛r g  . The defining property of … is that
every root in is a linear combination of elements of … with integral all non-positive,
or all non-negative, coefficients, and … is minimal among such subsets of . The
elements ˛i of … are called simple roots. Unlike the ordinary setting, the elements of
… are not necessarily linearly independent, e.g., when g is of type A.mjm/. So, once
we fix a simple system, we can write D C t  , where C is the set of the positive
roots, that is the roots ˛ D m1 ˛1 C    C mr ˛r , mi  0, while C is the set of the
negative roots, that is the roots ˛ D n1 ˛1 C    C nr ˛r , ni  0. From a representation
point of view, one may avoid the technicalities related to the superalgebras of type A
by replacing them by gl.mjm/. In the remainder of this appendix we will not discuss
240 A Lie superalgebras

the fine details distinguishing gl.mjm/ from the rest of the superalgebras of type A.
The interested reader will be able to fill in the details or consult [44], [45] for the whole
story.
For all of the families A, B, C , D, the Cartan subalgebras are even h D h0 , and
we can choose h D h0 to be the subalgebra of diagonal matrices. For each family
we shall give the root system and a choice of a simple system among the various
possibilities. Then for such a choice, in the next section we shall build the Cartan
matrix and the Dynkin diagram associated with the simple system. For a complete
treatment, comprehending all of the simple systems, we refer the reader to [44], 2.5.2,
and [33], part 3.

A.mjn/ D sl.m C 1jn C 1/ for m ¤ n. Let i ; ıj 2 h , 1  i  m C 1, 1 

j  n C 1, defined as i .diag.a1 ; : : : ; amCnC2 // D ai , i D 1; : : : ; m C 1, and
ıj .diag.a1 ; : : : ; amCnC2 // D amC1Cj , j D 1; : : : ; n C 1.
Root system:
D fi  j ; ık  ıl ; ˙.i  ık /g;
0 D fi  j ; ık  ıl g;
1 D f˙.i  ık /g; 1  i ¤ j  m C 1; 1  k ¤ l  n C 1:
Simple root system:
… D f˛1 D 1  2 ; ˛2 D 2  3 ; : : : ; ˛mC1 D mC1  ı1 ;
˛mC2 D ı1  ı2 ; : : : ; ˛mCnC1 D ın  ınC1 g:
For A.njn/, n > 1, the root system and the simple root system are given by the same
formulas as above. It is useful to remember that the elements i ; ıj are not linearly
independent. If m ¤ n, there is one relation between them: 1 C  CmC1 D ı1 C  C
ınC1 , while for m D n there are two relations: 1 C    C mC1 D ı1 C    C ınC1 D 0.

B.mjn/ D osp.2 m C 1j2n/. The Cartan matrix is the subalgebra of the diagonal
matrices:
h D fh D diag.a1 ; : : : ; am ; a1 ; : : : ; am ; 0; b1 ; : : : ; bn ; b1 ; : : : ; bn /g:
Define i ; ıj 2 h as follows: for h 2 h, let i .h/ D ai , i D 1; : : : ; m, and ıj .h/ D bj ,
j D 1; : : : ; n.
Root system, m ¤ 0:

0 D f˙i ˙ j ; ˙i ; ˙ık ˙ ıl ; ˙2ık g;

1 D f˙i ˙ ık ; ˙ık g; 1  i ¤ j  m; 1  k ¤ l  n:
Root system, m D 0:

0 D f˙ık ˙ ıl ; ˙2ık g; 1 D f˙ık g; 1  k ¤ l  n:

 A.2 Root systems 241

Simple root system, m ¤ 0:

… D f˛1 D ı1  ı2 ; : : : ; ˛n1 D ın1  ın ; ˛n D ın  1 ;

˛nC1 D 1  2 ; : : : ; ˛mCn1 D m1  m ; ˛mCn D m g:

Simple root system, m D 0:

… D f˛1 D ı1  ı2 ; : : : ; ˛n1 D ın1  ın ; ˛n D ın g:

C.n/ D osp.2j2n  2/. The Cartan matrix is again the subalgebra of the diagonal
matrices:

h D fh D diag.a1 ; a1 ; b1 ; : : : ; bn1 ; b1 ; : : : ; bn1 /g:

Define 1 ; ı1 ; : : : ; ın1 2 h as follows: for h 2 h, let 1 .h/ D a1 , ı1 .h/ D b1 , …,

ın1 .h/ D bn1 .
Root system:

0 D f˙2ık ; ˙ık ˙ ıl g; 1 D f˙1 ˙ ık g; 1  k ¤ l  n  1:

Simple root system:

… D f˛1 D 1  ı1 ; ˛2 D ı1  ı2 ; : : : ; ˛n1 D ın2  ın1 ; ˛n D 2ın1 g:

D.mjn/ D osp.2 mj2n/. The Cartan matrix is again the subalgebra of the diagonal
matrices:

h D fh D diag.a1 ; : : : ; am ; a1 ; : : : ; am ; b1 ; : : : ; bn ; b1 ; : : : ; bn /g:

Define 1 , : : : , m , ı1 , : : : , ın 2 h as follows: for h 2 g, let 1 .h/ D a1 , …,

m .h/ D am , ı1 .h/ D b1 , …, ın .h/ D bn .
Root system:

0 D f˙i ˙ j ; ˙2ık ; ˙ık ˙ ıl g;

1 D f˙i ˙ ık g; 1  i ¤ j  m; 1  k ¤ l  n:

Simple root system:

… D f˛1 D ı1  ı2 ; : : : ; ˛n1 D ın1  ın ; ˛n D ın  1 ;

˛nC1 D 1  2 ; : : : ; ˛mCn1 D m1  m ; ˛mCn D m1 C m g:

We end this section with a summary of the properties of root systems for classical
Lie superalgebras. For the proofs we refer the reader to [44], Sections 2.5.3–2.5.4.
P
Proposition A.2.5. Let g be a classical Lie superalgebra and let g D h ˚ ˛2 g˛ be
the root decomposition of g with respect to a Cartan subalgebra h. Then the following
hold:
242 A Lie superalgebras

(1) 0 \ 1 D f0g unless g D Q.n/.

(2) dim g˛ D 1 unless g D A.1j1/, P .3/, Q.n/.
(3) On g there is one and only one, up to a constant factor, non-degenerate invariant
symmetric consistent bilinear form . ; /, except for P .n/ and Q.n/ (where no
such form exists).
(4) If g ¤ A.1j1/; P .n/; Q.n/, then . ; / determines a non-degenerate pairing of g˛
with g˛ . Furthermore, for every ˛ 2 there exists an element X˛ 2 g˛ such
that ŒX˛ ; X˛  D .X˛ ; X˛ /h˛ , where h˛ 2 h is determined by ˛.h/ D .h˛ ; h/,
X˛ 2 g˛ , X˛ 2 g˛ .
(5)  0 D 0, and for g ¤ P .n/, 1 D 1.

(6) Let ˛ 2 and c˛ 2 . Then c D ˙1 unless ˛ 2 1 and .˛; ˛/ ¤ 0 when

c D ˙1 or c D ˙2.

A.3 Cartan matrices and Dynkin diagrams

Let us now turn to the construction of the Cartan matrix and the Dynkin diagram
associated to a classical Lie superalgebra g with a simple root system … D f˛i gi2I . For
each simple root ˛i 2 …, fix elements ei 2 g˛ , fi 2 g˛ and set hi D Œei ; fi  2 g0 .
Clearly hi is defined up to a constant. If ˛i .hi / ¤ 0, we fix it by imposing that
˛i .hi / D 2.
Definition A.3.1. The Cartan matrices A associated to the simple Lie superalgebra g
and the simple root system … is defined as
A D .aij / D .˛i .hj //:
We can write the Cartan matrix for all the classical Lie superalgebras whose roots,
together with a choice of a simple system, we have described in the previous section.
• A.mjn/. In the case of A.mjn/ we can choose ei D Ei;iC1 and fi D EiC1;i
(where as usual Eij denotes an elementary matrix). One can check immediately
that hi D Ei i  EiC1;iC1 for i ¤ m C 1, while hmC1 D ŒemC1 ; fmC1  D
emC1 fmC1 C fmC1 emC1 D EmC1;mC1 C EmC2;mC2 . The Cartan matrix has
the form
0 1
2 1 0 : : : 0 : : : 0 ::: 0
B1 2 1 : : : 0 : : : 0 ::: 0 C
B C
B :: :: :: :: :: :: :: :: :: C
B : : : : : : : : : C
B C
ADB B 2 1 0 C;
C
B 0 ::: 1 0 C1 : : : : : : : : :C
B C
B 0 ::: 0 1 2 1 : : : : : :C
@ A
:: :: :: :: :: :: ::
: : : : : : : 1 2
 A.3 Cartan matrices and Dynkin diagrams 243

where the zero appears in row .m C 1/ because ˛mC1 .hmC1 / D 0.

• B.mjn/, m ¤ 0.
0 1
2 1 0 ::: 0 ::: 0 ::: 0
B1 2 1 : : : 0 ::: 0 ::: 0C
B C
B :: :: :: :: :: :: :: :: :: C
::
B : : : : : : : : : C
:
B C
B 2 1 0 C
B C
ADB
B0 ::: 1 0 C1 : : : ::: : : :C
C
B0 ::: 0 1 2 1 ::: : : :C
B C
B :: :: :: :: :: :: :: :: :: :: C
B : : : : : : : : : : C
B C
@0 ::: 1 2 1A
0 ::: 0 2 2

• B.0jn/.
0 1
2 1 0 ::: 0 ::: 0 :::
B1 2 1 : : : 0 ::: 0 : : :C
B C
B :: C
A D B ::: ::
:
::
:
::
:
::
:
::
:
::
: : C
B C
@ 0 ::: 1 2 1A
0 ::: 0 2 2

• C.n/.
0 1
0 C1 0 : : : 0 ::: 0 :::
B1 2 1 : : : 0 ::: 0 : : :C
B C
B :: C
A D B ::: ::
:
::
:
::
:
::
:
::
:
::
: : C
B C
@ 0 ::: 1 2 2A
0 ::: 0 1 2

• D.mjn/.
0 1
2 1 0 ::: 0 ::: 0 ::: 0
B1 2 1 : : : 0 ::: 0 ::: 0C
B C
B :: :: :: :: :: :: :: :: :: :: C
B : : : : : : : : : :C
B C
B0 ::: 1 0 C1 : : : ::: ::: 0C
ADB
B0
C
B ::: 0 1 2 1 ::: ::: 0C C
B :: :: :: :: :: :: :: :: :: :: C
B : : : : : : : : : :C
B C
@0 :::   1 2 0A
0 :::   1 0 2
244 A Lie superalgebras

As it happens in the classical theory, we can associate a Dynkin diagram to any

Cartan matrix (except for the case D.2; 1I ˛/1 ) following the rules:
(1) Put as many nodes as simple roots.
(2) Connect the i-th node with the j -th node with jaij aj i j links.
(3) The i -th node is white if ˛i is even, is black if ˛i is odd and ai i ¤ 0 and it is
grey if ˛i is odd and ai i D 0.
(4) The arrow goes from the long to the short root.
Below we give a list of the Dynkin diagrams for the Lie superalgebras we have
described above with respect to the simple system we have written down. It is im-
portant to stress that, contrary to the ordinary setting, different Dynkin diagrams may
correspond to isomorphic Lie superalgebras. This is ultimately linked to the fact that
the Weyl group of g, defined as the Weyl group of g0 , does not provide “enough”
symmetries for the Lie superalgebra since it is a purely even object.

A.mjn/  :::  

  ::: 

B.mjn/; m > 0  :::  

  :::  +3

B.0jn/  :::   +3 

C.n/; n > 2 
  ::: ks 


pp
 :::  
  ::: pNpNp
D.mjn/ NNN


We leave it to the reader, as a simple exercise, to verify that these are indeed the
Dynkin diagrams associated to the Cartan matrices listed above.

A.4 Classification of finite-dimensional irreducible modules for

sl2 , osp.1j2/, gl.1j1/ and q.2/
In the theory of finite-dimensional representations for a classical Lie algebra g, the
representations of sl2 play a key role. This is due to the fact that, given a root ˛, the
root spaces g˛ and g˛ corresponding to ˙˛ generate a subalgebra of g isomorphic
to sl2 . The conditions for the finite-dimensionality of an irreducible g-module are
essentially derived from the conditions for finite-dimensionality of all the sl2 -modules
sitting inside the g-module. In the super category this is no longer true. Given a root
˛ of g, the root spaces gk˛ corresponding to all roots proportional to ˛ can generate
1
As for D.2; 1I ˛/ we invite the reader to consult [44], p. 55 for the convention that has to be used.
 A.4 Classification of finite-dimensional irreducible modules 245

different subalgebras of g. Subalgebras of this type were introduced by Penkov and

Serganova in [62]. More precisely, a line ` in h0 is defined as a one-dimensional real
subspace of h0 such that ` \ ¤ ;. The corresponding line subalgebra g` of g is
defined as the subalgebra of g generated by all root spaces g˛ for ˛ 2 `. Penkov and
Serganova proved a theorem characterizing all possible line subalgebras of an arbitrary
Lie superalgebra, see [62]. One can check immediately that every line subalgebra of a
classical simple Lie superalgebra is isomorphic to one of the following: sl2 , osp.1j2/,
sl.1j1/, q.2/, or an odd one-dimensional nilpotent superalgebra. The last two occur
for g of type P or Q. Below is a list of all finite-dimensional irreducible modules of
sl2 , osp.1j2/, gl.1j1/, and q.2/ (the modules for sl.1j1/ come from a straightforward
generalization from those for gl.1j1/).

The representations of the special linear Lie algebra sl2 . The finite-dimensional
representations of sl2 are well known, we just quote the theorem, referring the reader to
[75], Ch. IV, for more details. Let sl2 D spank fh; e; f g, with Œe; f  D h, Œh; e D 2e,
Œh; f  D 2f .

Theorem A.4.1. Let V be an irreducible finite-dimensional sl2 -module of dimension

n C 1. Then there exists a basis v0 ; : : : ; vn of V such that

hvi D .n  2i /vi ; 0  i  n;
ev0 D 0; evi D .n  i C 1/vi1 ; 1  i  n;
f vn D 0; f vi D .i C 1/viC1 ; 0  i  n  1:

Conversely, for each positive integer n, there exists exactly one equivalence class of
irreducible representations of sl2 of dimension n C 1, defined by the action described
above.

The vector v0 is called a highest weight vector and spans the one-dimensional
subspace of V annihilated by e.

The representations of the orthosymplectic Lie algebra osp.1j2/. Let us start with
an explicit description of this Lie superalgebra.
80 1 9
< 0 ˛ ˇ =
osp.1j2/ D @ ˇ A j B 2 sl2  gl.1j2/:
: B ;
˛

A basis is given by
0 1 0 1 0 1
0 0 0 0 0 0 0 0 0
e D @0 0 1A ; f D @0 0 0A ; h D @0 1 0 A ;
0 0 0 0 1 0 0 0 1
246 A Lie superalgebras

0 1 0 1
0 0 1 0 1 0
x D @1 0 0A ; y D @ 0 0 0A ;
0 0 0 1 0 0

where spank fe; f; hg Š sl2 and the other non-zero brackets are

Œe; y D x; Œf; x D y; Œx; x D 2e; Œy; y D 2f;

Œx; y D h; Œh; x D x; Œh; y D y:

The next proposition classifies the irreducible finite-dimensional representations of

osp.1j2/.
Theorem A.4.2. Let V be an irreducible finite-dimensional osp.1j2/-module. Then V
has dimension 2n C 1 and there exists a basis v0 ; : : : ; vn , w0 ; : : : ; wn1 of V such that

hvi D .n  2i /vi ; hwi D .n  1  2i /wi ;

evi D .n  i C 1/vi1 ; ewi D .n  i /wi1 ;
f vi D .i C 1/viC1 ; f wi D .i C 1/wiC1 ;
xvi D wi1 ; xwi D .n  i /vi ;
yvi D wi ; xwi D .i C 1/viC1 ;

where we have assumed that v1 D w1 D vnC1 D wn D 0. Up to a change of parity,

V0 D spanfv0 ; : : : ; vn g and V1 D spanfw0 ; : : : ; wn1 g and they are irreducible sl2 -
modules of dimension n C 1 and n, respectively.
Conversely, for every odd positive integer 2n C 1, up to a change of parity, there
exists exactly one equivalence class of irreducible representations of osp.1j2/ of di-
mension 2n C 1, defined by the action described above.
Proof. We sketch the proof leaving the details to the reader as an exercise. The parity
change in a super vector space commutes with actions. If v0 is the highest weight vector,
it could be v0 2 V0 or v0 2 V1 , hence without loss of generality we may assume that
v0 2 V0 . Applying Theorem A.4.1 we conclude that ev0 D 0 and hv0 D nv0 for some
non-negative integer n. Now set

vi D 1 i
iŠ
f v0 for 0  i  n; wi D 1 i
iŠ
f yv0 for 0  i  n  1:

One can now verify that the basis of osp.1j2/ acts on the set v0 , : : : , vn , w0 , : : : , wn1
as claimed and that this action provides V D span fv0 , : : : , vn , w0 , : : : , wn1 g with a
structure of an irreducible osp.1j2/-module.
Schematically we have the following picture:

V0 vn F e / vn1 e / vn2 :::

FF v v: MMM nn7
FF vvv MMM
M nnnnn
x FFF vv x x MMM nn x
# vv & nnn
V1 wn1 e / wn2 :::
 A.4 Classification of finite-dimensional irreducible modules 247

and similarly

f f
V0 vn cFo vn1 fMo vn2 :::
FF y
FF y vvv
v MMM
nnnn
v MMyM y
nnn
FF vv MMM nn
F zvv f wnnn
V1 wn1 o wn2 :::

The representations of the general linear supergroup gl.1j1/. As a super vector

space we have
²        ³
1 0 0 0 0 1 0 0
gl.1j1/ D span h1 D ; h2 D ; xD ; yD
0 0 0 1 0 0 1 0

with brackets

Œh1 ; x D x; Œh2 ; x D x; Œh1 ; y D y; Œh2 ; y D y;

Œx; y D h1 C h2 ; Œx; x D Œy; y D 0:

Theorem A.4.3. Let V be the irreducible representation of gl.1j1/ with highest weight
. 1 ; 2 /. Then:
(1) If 1 C 2 ¤ 0, V is 1j1-dimensional and spanned by v, the highest weight
vector, and yv.
(2) If 1 C 2 D 0, V is one-dimensional.
Proof. Let v be the highest weight vector of V . We have

x.yv/ D yxv C Œx; yv D .h1 C h2 /v D . 1 C 2 /v

and
y 2 v D .1=2/Œy; yv D 0; x 2 v D .1=2/Œx; xv D 0;

which proves the theorem. For 1C 2 ¤ 0 the statement is illustrated by the diagram
x
(
yv j v: 
y

The representations of q.2/. The irreducible finite-dimensional modules over g D

q.2/ have a more complicated structure. Recall that D f˙˛g, and each of gf˙˛g is
1j1-dimensional. Furthermore, h D h0 Ch1 is 2j2-dimensional. If we choose the usual
basis fh1 ; h2 g of h0 , we conclude that every finite-dimensional irreducible g-module
V has a highest weight . 1 ; 2 /. The highest weight space of V.1 ;2 / is an irreducible
h-module, and it is no longer necessarily one-dimensional. We leave it to the reader
248 A Lie superalgebras

to verify that for generic highest weight . 1 ; 2 /, the highest weight space V.1 ;2 /
is an irreducible module of a (suitably chosen) Clifford algebra. In general it is of
dimension 1j1. Since g contains a subalgebra isomorphic to gl2 , a necessary condition
for finite-dimensionality of V is 1  2 to be a non-negative integer. Surprisingly, it
is not enough. The following theorem is proved in [63]:
Theorem A.4.4. Let D . 1 ; 2 / be a weight of q.2/ such that 1  2 is a positive
integer or 1 D 2 D 0. There exists a unique (up to a change of parity) finite-
dimensional irreducible q.2/-module V . / with highest weight . The weights of
V . / are . 1 ; 2 /; . 1  1; 2 C 1/; : : : ; . 2 ; 1 /.
Conversely, every irreducible finite-dimensional q.2/-module V is isomorphic to
V . / for some as above.
For the proof as well as further study of q.n/-modules we refer the reader to the
work of Penkov, [63].

A.5 Representations of basic Lie superalgebras

Before starting our description of the finite-dimensional representations of basic Lie
superalgebras, we quickly review the classical theory of representations of Lie algebras.
Let g be a Lie algebra, h a fixed Cartan subalgebra, D C [  the root system
and V a g-module. If 2 h we define weight space of weight as the non-zero
vector space:
V D fv 2 V j h  v D .h/v for all h 2 hg:
The elements in V are called weight vectors.
The following theorem is crucial in understanding the finite-dimensional represen-
tations of a complex simple Lie algebra.
Theorem A.5.1. Let g be a simple Lie algebra and let V be an irreducible finite-
dimensional representation. Then we have the following.
(1) There exists a non-zero vector v C such that it is a weight vector for some weight
and X˛ v C D 0 for all X˛ , ˛ > 0, where X˛ is a root vector for the root ˛.
The weight is called highest weight and v C highest weight vector.
P
(2) We have dim V D 1, and every weight L has the form D  ˛>0 k˛ ˛ with
k˛ 2 Z0 for every ˛. Moreover V D 2h V . (Every representation with
this property is called a weight representation.)
(3) The weight and the span of v C are uniquely determined.
(4) We have .h˛ / 2 Z0 , where h˛ is as in Proposition A.2.5; such a weight is
called a dominant integral.
(5) Given any dominant integral weight 2 h there exists a unique irreducible
finite-dimensional module V in which is the highest weight.
 A.5 Representations of basic Lie superalgebras 249

Proof. See [71], p. 14. 

We now outline an approach to the representation theory of Lie algebras via induced
modules. This approach is alternative to the approach with generators and relations and
provides an elegant way to prove the existence of a finite-dimensional representation
associated to a dominant integral weight.
Let 2 h and define bC ´ nC ˚ h to be the Borel subalgebra associated to
n and h, where nC is the Lie subalgebra generated by the positive roots, once we fix
C

a simple system. The weight defines a one-dimensional representation k of bC on

which h acts via and nC acts trivially:
.h C n/  v D .h/v; v D 1 2 k; for all h 2 h and n 2 nC :
Definition A.5.2. We define the induced module
M. / D IndgbC D U.g/ ˝U.bC / k:
M. / is also called the Verma module associated to the weight .
Observation A.5.3. The induced module M. / Š U.n / as n -modules. The iso-
morphism is given by
U.n / ! M. /; u 7! u ˝ 1:
This follows immediately from the PBW theorem.
Proposition A.5.4. Let 2 h , f˛1 ; : : : ; ˛n g a simple system. Then we have:
L
(1) M. / D M. / , i.e., M. / is a weight representation,
X
M. / D X˛p1
1
pn
: : : X˛ n
˝ k:
p1 ;:::;pn 2Z0 ;p1 ˛1 CCpn ˛n D

(2) M. / D 1 ˝ k, M. / D U.n /M. / , U.nC /M. / D 0.

Proof. See [27], Ch. 7, Section 1. 
Observation A.5.5. If V is an irreducible finite-dimensional representation of a semi-
simple g, there exist a unique 2 h and a unique morphism  W M. / ! V with
.1 ˝ 1/ D v. Hence every irreducible representation V of g occurs uniquely as a
quotient of a Verma module.
Proposition A.5.6. (1) Any submodule F  M. / is a weight submodule, i.e., F D
L
F , F D F \ M. / .
(2) Every proper submodule of M. / is contained in
X
M. /C D M. / :
¤

(3) There exists K  M. / largest proper submodule such that M. /=K is irre-
ducible (when finite-dimensional M. /=K is isomorphic to the V . / in TheoremA.5.1).
250 A Lie superalgebras

Proof (Sketch). (1) is a direct check. (2) comes from the fact that if 1 ˝ 1 2 F , then
F D M. /. For (3) take K as the sum of all proper submodules. For the complete
proof see [27], Ch. 7. 
We have described an alternative way to construct the representations V . /. This is
the path taken by Kac in the construction of the irreducible representations of classical
Lie superalgebras and as we shall see the construction is very much the same as the
ordinary one.
Let g be a classical Lie superalgebra. Fix a Cartan subalgebra and a simple system
and let nC be as above. Let bC D nC ˚ h be a fixed Borel subalgebra and 2 h . As
before we have a one-dimensional representation of bC :

.h C n/  v D .h/v; v D 1 2 k; h 2 h; n 2 nC :

We define
M. / D IndgbC D U.g/ ˝U.bC / k:
Reasoning as in Proposition A.5.6, we have that M. / contains a unique maximal
submodule K such that M. /=K D V . / is irreducible. Since change of parity is not
an isomorphism of g-modules, we will always assume that the highest weight space is
even, and all statements of uniqueness below are up to a change of parity.
Proposition A.5.7. Let the notation be as above.
(1) The vector v is the unique vector in M. / (up to a constant) such that nC v D 0.
(2) V . 1/ Š V. 2/ if and only if 1 D 2.

(3) If V is a quotient of M. /, then V has a weight decomposition

L
V D V :

(4) Any finite-dimensional irreducible representation of g is isomorphic to V . /, for

some .
The proof of this proposition is similar to the one in the ordinary setting and more
details can be found in [45].
We are ready for the main result on the representation of classical Lie superalgebras.
Let g be a basic Lie superalgebra. Fix a Cartan subalgebra and a simple system
… D f˛1 ; : : : ; ˛r g. Let hi D h˛i as in Proposition A.2.5, i.e., .hi ; h/ D ˛.h/.
Definition A.5.8. If is in h we define the numerical marks ai D .hi /.
Theorem A.5.9. Let the notation be as above. In particular we fix a simple root system
for each Lie superalgebra as in Section A.2 or in [44]. Then V . / is a finite-dimensional
representation if and only if :
(1) ai 2 Z0 , i ¤ s, where s D m C 1 for A.mjn/, s D m for B.mjn/, D.mjn/
and s D 1 for C.n/, F .4/, G.3/, D.2; 1I ˛/.
 A.5 Representations of basic Lie superalgebras 251

(2) k 2 Z0 where k is given by

• For B.0jn/, k D .1=2/an .
• For B.mjn/, m > 0, k D an  anC1      amCn1  12 amCn .
• For D.mjn/, k D an  anC1      amCn2  12 .amCn1 C amCn /.
• For D.2; 1I ˛/, k D .1 C ˛/1 .2a1  a2  ˛a3 /.
• For F .4/, k D .1=3/.2a1  3a2  4a3  2a4 /.
• For G.3/, k D .1=2/.a1  2a2  3a3 /.
(3) Let b be an integer as follows:
B.0jn/ B.mjn/ D.mjn/ D.2; 1I ˛/ F .4/ G.3/
b 0 m m 2 4 3
There are the following supplementary conditions if k < b:
• For B.mjn/, akCnC1 D    D amCn D 0.
• For D.mjn/, if k < m  1, akCnC1 D    D amCn D 0; if k D m  1,
akCnC1 D amCn .
• For D.2; 1I ˛/, if k D 0, ai D 0 for all i; if k D 1, .a3 C1/˛ D ˙.a2 C1/.
• For F .4/, if k D 0, k ¤ 1, ai D 0 for all i ; if k D 2 a2 D a4 D 0; if
k D 3, a2 D 2a4 C 1.
• For G.3/, if k D 0, k ¤ 1 ai D 0; if k D 2, a2 D 0.
Conditions (1) and (2) are very natural – these are simply the conditions for to
be a dominant integral weight of g0 . Condition (3) is more interesting. One way of
understanding it is the following: V . / is a finite-dimensional module if and only if it
is a highest weight module with respect to every Borel subalgebra of g. In the classical
setting all Borel subalgebras of a Lie algebra are conjugate via the elements of the
Weyl group W and, if b0 D wb for w 2 W , then Vb . / D Vwb .w /. Furthermore,
the weight is dominant and integral with respect to wb if and only if is dominant
integral with respect to b. For Lie superalgebras it is no longer true that all Borel
subalgebras are conjugate. Penkov and Serganova introduced in [62] the so-called
“odd reflections” which transform Borel subalgebras and act on weights; these take
the place of the missing symmetries of the Weyl group, which is an even object and
for this reason does not contain enough reflections to act transitively on the set of
Borel subalgebras. The action on weights, however, is no longer a group action. In
the remainder of this section we present the main idea and illustrate it on a relevant
example.
Let b be a Borel subalgebra of g and let ˛ be a simple root of b. Consider the
line ` D R˛ \ , (see the definition of a line at the beginning of Section A.4). The
Borel algebra b determines a partition D C t  . Set `˙ D ` \ ˙ and let
˙ 0 ˙ ˙ 0
. / D . n` / [ ` . Let b be the Borel subalgebra of g corresponding to the
252 A Lie superalgebras

partition D . C /0 t .  /0 . If ˛ is odd and the subalgebra generated by X˛ and X˛

is isomorphic to sl.1j1/, the map b 7! b0 is the odd reflection along ` that we denote
by r` . Notice that, if ˛ is an even root, then b0 D s˛ b, s˛ 2 W , otherwise b0 D r` b. If
Vb . / is finite-dimensional, then Vb . / D Vb0 . 0 / for some weight 0 . (If ˛ is even,
then 0 D s˛ .) We write 0 D r` . The weight 0 is easily computable - it is simply
the lowest weight of the g` C h-module with highest weight , where g` is the line
subalgebra corresponding to the line ` (see Section A.4). In the previous section we
described all irreducible modules over the possible line subalgebras g` . We can now
rephrase these results as follows.
Theorem A.5.10. Let ` D R˛ be a simple line of b  g, where g is a classical
simple Lie superalgebra. Assume that 2 h0 is a weight such that the irreducible
g` C h-module with highest weight is finite-dimensional. Then r` D s˛ if g` is
isomorphic to sl2 , osp.1j2/, or sq.2/. If g` is one-dimensional nilpotent, then r` D .
If g` is isomorphic to sl.1j1/, then
´
 ˛ if .h˛ / ¤ 0;
r` D
if .h˛ / D 0:

We have now a stronger necessary condition for the finite-dimensionality of the

module Vb . /. Namely, the weight r`1 r`2 : : : r`k has to be dominant with respect to
the Borel subalgebra r`1 r`2 : : : r`k b for any sequence of reflections r`1 ; r`2 ; : : : ; r`k
applied subsequently to b. This condition is also sufficient. We refer the reader to [62]
for a detailed proof. This statement provides an effective but not very efficient method
for determining the weights for which Vb . / is finite-dimensional. Serganova in [67]
describes a way of finding the minimal number of verifications necessary in the case
when g is a Kac–Moody superalgebra. We will not give the detailed statement here.
Instead, we will use Theorem A.5.10 above to explain condition (3) of Theorem A.5.9
in the case when g is of type B.mjn/.
Example A.5.11. We want to show that the conditions (1), (2), (3) in Theorem A.5.9
are necessary conditions in order to have finite-dimensional representations for the Lie
superalgebra B.mjn/. Recall that the root system is (see Section A.2)

0 D f˙ıi ˙ ıj ; ˙2ıi ; ˙i ˙ j ; ˙i g;

1 D f˙ıi ˙ i ; ˙ıi g:

Let us choose a Borel subalgebra, i.e., a simple system

˛1 D ı1  ı2 ; ˛2 D ı2  ı3 ; : : : ; ˛n1 D ın1  ın ;
˛n D ın  1 ; ˛nC1 D 1  2 ; : : : ; ˛nCm D m :

Assume that we have a finite-dimensional representation of highest weight ƒ:

ƒD 1 ı1 C  C n ın C 1 1 C  C m m :
 A.5 Representations of basic Lie superalgebras 253

Let ai D ƒ.hi / be the numerical marks. We have

ai D i  iC1 ; 1  i  n  1;
an D n C 1;
anCj D j  j C1 ; 1  j  m  1;
amCn D 2 n :

Since V . / is a finite-dimensional B.mjn/0 representation, we have

1  2    n  0; 1  2    m  0;

and these are integers or half integers. Hence

a1  0; a2  0; : : : ; an1  0; anC1  0; anC2  0; : : : ; anCm  0

are non-negative and integral. This gives the condition (1) in Theorem A.5.9. Now
observe that

n D an  anC1  anC2      anCm1  .1=2/anCm :

The fact n 2 Z0 gives condition (2). Now we want to see how to get condition (3).
In order to clarify the construction we work in B.2j1/, but the reader can immedi-
ately see how easily this can be generalized to B.mjn/.
In this case our weight ƒ and our fixed Borel subalgebra b (i.e., simple system) are

ƒD 1 ı1 C 1 1 C 2 2 ; fı1  1 ; 1  2 ; 2 g:

In the ordinary setting we have that the Weyl group acts transitively on the set of Borel
subalgebras, hence the weight w ƒ is automatically dominant and integral with respect
to the Borel subalgebra w  b. Here however, we have fewer symmetries since the Weyl
group is not a super object, but remains ordinary. In particular not all Borel subalgebras
are conjugate with respect to the Weyl group. We can compensate the lack of enough
symmetries of the Weyl group, by using odd reflections. In this way we can reach
every Borel subalgebra starting from a fixed one. For this reason we are going to use
an action of gl.1j1/, whose representations we have studied in Section A.4.
Our goal is to use odd reflections to move the chosen Borel subalgebra to the one
with the associated simple system: f1  2 ; 2  ı1 ; ı1 g, while tracking down how the
weight ƒ transforms. The dominant integral condition on the new weight will give us
the additional conditions (3).
Let us start with an odd reflection along the root ı1  1 . This corresponds to a
gl.1j1/ representation of highest weight 1 C 1 . By Theorem A.4.3 we have only
two cases:
(1) 1 C 1 D 0, i.e., 1 D 1 D 0 (since they are both positive), corresponding to
a 1-dimensional representation,
254 A Lie superalgebras

(2) 1 C 1 ¤ 0 corresponding to a 1j1-dimensional representation.

In case 1 C 1 D 0, we have that ƒ does not change, while in case 1 C 1 ¤ 0,
ƒ becomes ƒ0 D ƒ  .ı1  1 / D . 1  1/ı1 C . 1 C 1/1 C 2 2 .
We now apply another odd reflection: along ı1  2 . Again we are faced with two
possibilities as before. We are going to see what happens separately in the two cases
discussed above.

• Case 1 D 1 D 0, ƒ D 1 ı1 C 1 1 C 2 2 .
i) If 2 D 1 D 0, then ƒ D 0, which is the trivial case.
ii) If 2 C 1 ¤ 0, we reach a contradiction since 1  2  0 and 1 D 1 D 0.

• Case 1 C 1 ¤ 0, ƒ0 D . 1  1/ı1 C . 1 C 1/1 C 2 2 .

iii) If 2 C 1  1 D 0, the weight ƒ0 is unchanged and since 1  1  0,
2 C 1  1  0, we have 1 D 1, 2 D 0.
iv) If 2 C 1  1 ¤ 0 from ƒ0 we obtain a new weight ƒ00 D . 1  2/ı1 C
. 1 C 1/1 C . 2 C 1/2 . Hence we obtain the condition 1  2.

Let us now take a closer look at condition (3) in Theorem A.5.9.

We have k D a1  a2  12 a3 D 1 and b D m D 2. If k D 1  2 then there are
no extra conditions and we are in the last case.
If k D 1 < 2, then we are in one of the three previous cases, i.e., at some point
at least once we have a trivial gl.1j1/ representation occurring. If k D 1 D 0 then
1 D 2 D 1 D 0, i.e., a2 D 1  2 D a3 D 2 2 D 0. If k D 1 D 1 we see
a3 D 2 2 D 0.

This proves only of course that the condition (3) is necessary. However, using
Theorem 10.5 in [67] we can conclude that it is also sufficient.

A.6 More on representations of Lie superalgebras

The aim of this short section is to mention some properties of superalgebras and their
representations which differ from the classical setting. We do not aim to give a complete
and broad view of the subject and our references are extremely limited.
The structure of Lie superalgebras beyond the simple finite-dimensional ones ex-
hibits some unexpected properties. The first example of such a difference is the fact
that a semisimple Lie superalgebra, i.e., a superalgebra whose radical is trivial, is not
necessarily a direct sum of simple ones, see [44]. Another surprising fact, related to
Schur’s lemma below, is that the tensor product V1 ˝ V2 of two irreducible modules
over g1 and g2 respectively is not necessarily irreducible as a module over g1 ˚ g2 .
Following the classical theory of Lie algebras, the simple finite-dimensional real Lie
superalgebras have been classified, [44].
 A.6 More on representations of Lie superalgebras 255

Not every simple finite-dimensional Lie superalgebra exhibits properties analogous

to the properties of Kac–Moody Lie algebras. Even the superalgebras that do have
such properties admit several Cartan matrices and corresponding Dynkin diagrams.
This feature of superalgebras has made the problem of identifying good classes of
infinite-dimensional Lie superalgebras non-trivial. Kac and Wakimoto introduced cer-
tain affine Lie superalgebras and wrote conjectural character formulas for their highest
weight modules. Then they showed how these character formulas imply denominator
identities which generalize Macdonald’s identities for affine Lie algebras. Despite the
significance of such identities, little progress has been made towards proving the re-
spective character formulas. Serganova in her recent paper [67] proposed a definition
of a Kac–Moody superalgebra which addresses the differences between superalgebras
and their ordinary counterparts. Apart from being an important foundational paper,
[67] is also an excellent review of the current state of the art results on Kac–Moody
superalgebras and initiates the study of their representations.
The representation theory of Lie superalgebras is much richer than the one of Lie
algebras and, despite the great progress made since the pioneering paper [45], many
important questions remain open. Kac noticed, [45], that unlike the ordinary case,
the category of finite-dimensional representations of a simple Lie superalgebra is not
semisimple. That is, not every finite-dimensional module is a direct sum of irreducible
modules. Kac also singled out a class of irreducible representations, the so-called typ-
ical representations, whose properties closely resemble the properties of irreducible
finite-dimensional modules over simple Lie algebras. Typical representations always
split in finite-dimensional representations and it is not difficult to determine their char-
acters, etc. The rest of the irreducible representations, the atypical ones, are far more
difficult and understanding even their characters is a very challenging problem. This
problem has now been solved for the cases of gl.mjn/, q.n/, and osp.mj2n/. There
are two approaches.
The first one, geometric in nature, is a very sophisticated generalization of its
ordinary counter part. It took more than twenty years until a complete solution for
the above mentioned superalgebras was obtained. The main contributors towards this
approach were Penkov and Serganova who in the late 1980s and 1990s developed the
foundations of the theory and proved many particular cases. For gl.mjn/ the character
formula was obtained by Serganova, [68], followed by a work of Penkov and Serganova
on the characters of q.n/. Finally, Gruson and Serganova, [42], proved the character
formula for osp.mj2n/. For more detail on this approach we refer the reader to [42].
The second approach uses super analogs of Schur and Howe duality to reduce the
problem of determining the characters of irreducible finite-dimensional representations
of the Lie superalgebras gl.mjn/, q.n/, and osp.mj2n/ to a related problem about
representations of infinite-dimensional Lie algebras. The solution of the latter problem
is given in terms of the celebrated Kazhdan–Lusztig polynomials. This approach was
first introduced by Brundan for gl.mjn/ and later developed further by, among others,
Brundan, Stroppel, Cheng, Lam, Wang, and Zhang. We refer the interested reader to
[21], and the references therein, for more detail on this approach.
256 A Lie superalgebras

Any detailed discussion of the intricacies of the representation theory of Lie super-
algebras is beyond the scope of this Appendix. However, we would like to illustrate
some features on the case of gl.mjn/. In the ordinary setting, the Verma module M. /
for dominant integral has a unique finite-dimensional quotient. This is no longer true
for gl.mjn/. If is an atypical dominant integral weight, i.e., if V . / is an atypical
finite-dimensional module, M. / has more than one finite-dimensional quotient. In
fact, there is a universal finite-dimensional quotient K. / called the Kac module K. /,
first introduced by Kac. The relationship between V . / and K. / somewhat resembles
the relationship between V . / and M. / for infinite-dimensional modules V . / over a
simple Lie algebra. Hence the problem of finding the character of V . / for atypical
is somewhat close to the problem of finding the characters of the irreducible modules
in the category O of highest weight modules over Lie algebras.

A.7 Schur’s lemma

We end our brief treatment of representation theory with the super version of Schur’s
lemma, which is a fundamental result.
Theorem A.7.1 (Schur’s lemma). Let M be a subset of gl.mjn/ acting in an irreducible
way on a super vector space V . Let C.M/ be the subalgebra of gl.mjn/ consisting of
the endomorphisms supercommuting with M:

C.M/ ´ fa 2 gl.mjn/ j Œa; b D 0 for all b 2 Mg:

Then one and only one of these two possibilities arises:

(1) C.M/ D h1i,
(2) dim V0 D dim V1 , C.M/0 D h1i, C.M/1 D hAi, where A ¤ 0 permutes V0
and V1 and A2 D 1.
Proof. Let F be an even endomorphism commuting with M. By the classical Schur’s
lemma, we have F D cI , where c 2 k. Let F be an odd morphism commuting with
M. Then F 2 is even, hence we apply the classical Schur lemma to obtain F 2 D cI .
If G is another such morphism, it follows that F G is even, hence F G D cI . Thus
G D F 1 D F up to a scalar. 
We now want to show that the two possibilities which arise in the super version of
Schur’s lemma, do really occur even in very simple examples. We want to stress the
fact that the hypothesis of irreducibility of the representation means that there are no
graded g-submodules of V .
In our example we shall take M D .n/, where is a representation of the Heisen-
berg superalgebra n, which is an interesting object in itself, given its importance in
physics.
 A.7 Schur’s lemma 257

Example A.7.2. Let us define the complex Heisenberg superalgebra n with even (cen-
tral) generator e and odd generators a1 ; : : : ; an , b1 ; : : : ; bn . The only non-zero brackets
are
Œai ; bi  D e:
V
Let V D . 1 ; : : : ; n / denote the complex exterior algebra with generators 1 ; : : : ; n .
This is a super vector space, where the even part is generated by products of an even
number of i ’s while the oddVpart, by products of an odd number of the i ’s. We define
a representation ˛ of n on . 1 ; : : : ; n / in the following way:

ai  u D @i u; bi  u D ˛ i u; e  u D ˛u;

where ˛ is a fixed non-zero complex number. One can easily check that ˛ is well
defined, in other words, the brackets are preserved. We now want to show that ˛ is
irreducible and C. .n// the endomorphisms of V supercommuting with the endomor-
phisms induced by this action are multiples of the identity. We will do this in the case
of n D 2, leaving the generalization to the reader as an exercise. For the irreducibility,
notice that the even and odd parts have dimension 2. Hence a proper invariant subspace
must have even or odd dimension equal to 1 and one can check immediately by direct
inspection that this is not possible.
V
Let  2 gl.V /, V D . 1 ; 2 /. By Schur’s lemma, in order to prove  D I it
is enough to prove that  preserves parity. By contradiction assume that  does not.
Again by Schur’s lemma,
 
0 B
D ;
C 0
where B and C are two by two invertible matrices and we have chosen the (graded)
basis 1, 1 ^ 2 , 1 , 2 . By hypothesis we have

@1 ..1// D .@1 .1// D 0; @1 .b11 1 C b21 2 / D b11 D 0:

Similarly by commuting  with @2 one can see that b21 D 0. This gives us a contra-
diction since by Schur’s lemma,  must be invertible.
Let us now turn to another example, which realizes the second possibility of Schur’s
lemma. Consider n0 D n ˚ hci where the odd element c satisfies

Œn; c D 0; Œc; c D e:
V
Consider the vector space V D . 1 ; : : : ; n / ˝ k./, where k./ D T .k 0j1 /=. ˝
  ˛=2/, and T .k 0j1 / denotes the full tensor algebra over k of the super vector space
k 0j1 with (canonical) basis . As a super vector space k./ D fa C bg Š k 1j1 . Define
on V the action ˛0 :
0 0
˛ .h/.u ˝ v/ D ˛ .h/u ˝v for all h 2 n; ˛ .c/.u ˝ v/ D .1 ˝ /.u ˝ v/:
258 A Lie superalgebras

We leave it to the reader to check that this is a well-defined action and that it is irre-
ducible. If we take n D 2 we can directly check that the odd morphism
 
0 I2
D
I2 0
0
commutes with the action ˛ for ˛ D 2, thus realizing (2) in Schur’s lemma.
B
Categories

In this appendix we collect some basic facts on categories, sheafification of functors

and commutative algebra that are needed in our notes. All of this material is well
known; we have made an effort to make the appendix self-contained and to provide
precise references for those results that we are unable to prove here.

B.1 Categories
We want to make a brief summary of formal properties and definitions relative to
categories. For more details one can see for example [51].

Definition B.1.1. A category C consists of a collection of objects, Ob.C /, and sets

of morphisms between objects. For all pairs A; B 2 Ob.C /, we denote the set of
morphisms from A to B by HomC .A; B/ so that, for all A; B; C 2 C , there exists an
association
HomC .B; C /  HomC .A; B/ ! HomC .A; C /;
called the “composition law” (.f; g/ ! f B g), which satisfies the properties:
(i) the law “B” is associative,
(ii) for all A; B 2 Ob.C /, there exists idA 2 HomC .A; A/ so that f B idA D f for
all f 2 HomC .A; B/ and idA B g D g for all g 2 HomC .B; A/,
(iii) HomC .A; B/ and HomC .A0 ; B 0 / are disjoint unless A D A0 , B D B 0 , in which
case they are equal.

If a morphism f 2 HomC .A; B/ is invertible; in other words, there exists another

morphism g 2 HomC .B; A/ such that f B g and g B f are the identities, respectively,
in HomC .B; B/ and HomC .A; A/, we say that f is an isomorphism.
Once the category is understood, it is conventional to write A 2 C instead of
A 2 Ob.C / for objects. We may also suppress the “C ” from HomC and just write
Hom whenever there is no danger of confusion.
Essentially a category is a collection of objects which share some basic structure,
along with maps between objects which preserve that structure.

Examples B.1.2. (1) Let .sets/ denote the category of sets. The objects are the sets,
and for any two sets A; B 2 Ob..sets//, the morphisms are the maps from A to B.
260 B Categories

(2) Let G denote the category of groups. Any object G 2 G is a group, and for any
two groups G; H 2 Ob.G /, the set HomG .G; H / is the set of group homomorphisms
from G to H .
Definition B.1.3. A category C 0 is a subcategory of category C if Ob.C 0 /  Ob.C/
and HomC 0 .A; B/  HomC .A; B/ for all A; B 2 C 0 , so that the composition law “B”
on C 0 is induced by that on C .
Example B.1.4. The category A of abelian groups and group morphisms is a subcat-
egory of the category of groups G .
Definition B.1.5. Let C1 and C2 be two categories. Then a covariant resp. contravari-
ant functor F W C1 ! C2 consists of:
(1) a map F W Ob.C1 / ! Ob.C2 / and
(2) a map (denoted by the same F ) F W HomC1 .A; B/ ! HomC2 .F .A/; F .B//
resp. F W HomC1 .A; B/ ! HomC2 .F .B/; F .A// so that
(i) F .idA / D idF .A/ and
(ii) F .f B g/ D F .f / B F .g/ resp. F .f B g/ D F .g/ B F .f / for all A; B 2
Ob.C1 /.
By “functor” we always mean covariant functor. A contravariant functor F W C1 !
op op
C2 is the same as a covariant functor from C1 ! C2 , where C1 denotes the opposite
category, i.e., the category where all morphism arrows are reversed.
Definition B.1.6. Let F1 ; F2 be two functors from C1 to C2 . We say that there is a
natural transformation of functors ' W F1 ! F2 if for all A 2 C1 there is a set of
morphisms 'A W F1 .A/ ! F2 .A/ so that for any f 2 HomC1 .A; B/ (B 2 C1 ), the
following diagram commutes:
'A
F1 .A/ / F2 .A/

F1 .f / F2 .f / (B.1)
 'B 
F1 .B/ / F2 .B/.

The family of functions 'A is called functorial in A.

We say that two functors F; G W C1 ! C2 are isomorphic if there exist two natural
transformations  W F ! G and W G ! F such that  B D id and B  D id. In
other words, F .A/ is isomorphic to G.A/ via the two maps A and A for all objects
A in a functorial way, i.e., in such a way that the diagram (B.1) holds.
The functors from C1 to C2 for any two given categories together with their natural
transformations form a category.
The notion of equivalence of categories is important since it allows us to identify
two categories which are apparently different.
 B.1 Categories 261

Definition B.1.7. We say that two categories C1 and C2 are equivalent if there exist
two functors F W C1 ! C2 and G W C2 ! C1 such that F G Š idC2 , GF Š idC1 ,
where idC denotes the identity functor of a given category, defined in the obvious way,
while F Š F 0 means that the two functors are isomorphic.
If F is a functor from the category C1 to the category C2 for any two objects A,
B 2 C1 , by its very definition, F induces a function (that we denoted previously
with F )
FA;B W HomC1 .A; B/ ! HomC2 .F .A/; F .B//:
Definition B.1.8. Let F be a functor. We say that F is faithful if FA;B is injective, we
say F is full if FA;B is surjective and we say that F is fully faithful if FA;B is bijective.
Next we want to formally define what it means for a functor to be representable.
Let us first define the representation functors.
Definition B.1.9. Let C be a category, A a fixed object in C . We define the two
representation functors HomA , HomA as

HomA W C op ! .sets/; B 7! HomC .B; A/;

HomA W C ! .sets/; B 7! HomC .A; B/;

where .sets/ denotes the category of sets. On the arrow f 2 Hom.B; C / we have

HomA .f / D  B f;  2 HomA .B/; HomA .f / Df B ; 2 HomA .B/:

Definition B.1.10. Let F be a functor from the category C to the category of sets. We
say that F is representable by X 2 C if for all A 2 C, F Š HomA or F Š HomA .
We end our small exposition of categories by constructing the fibered product which
is very important in our supergeometric constructions especially in Chapter 10.
Definition B.1.11. Given functors A, B, C from a category C to the category of .sets/,
and given natural transformations f W A ! C , g W B ! C , the fibered product AC B
is the universal object making the following diagram commute:

T
x
.x;y/
# %/
y
A C B p A
q f
 g 
B / C.

One can see that

.A C B/.R/ D A.R/ C.R/ B.R/ D f.a; b/ 2 A  Bjf .a/ D g.b/g:

262 B Categories

If g is injective, that is, gR W B.R/  C.R/, we have .A C B/.R/ D f 1 .B.R//.

The language of categories allows us to make (and prove) some sweeping gen-
eralizations about geometric objects without too much “forceful” computation. In
particular, it also allows us to generalize the notion of a “point” to a T -point; this
allows us to make more intuitive calculations with supergeometric objects. The main
categories we discuss in this exposition are the categories of C 1 -supermanifolds, super
Lie groups (a subcategory of C 1 -supermanifolds), superschemes, and superalgebraic
groups.

B.2 Sheafification of a functor

In this section we discuss the concept of sheafification of a functor in supergeometry.
Most of this material is known or easily derived from known facts. We include it here
for completeness and lack of an appropriate reference.
Hereafter we shall make a distinction between a superspace X and its functor of
points, that we shall denote by hX or, if X D Spec.A/, by hA .
We start by defining local and sheaf functors. For their definitions in the classical
setting see for example [23], p. 16, or [29], Ch. VI.
Definition B.2.1. Let F W .salg/ ! .sets/ be a functor. Fix A 2 .salg/. Let ffi gi2I 
A0 , .ffi gi2I / D A0 and let i W A ! Afi , ij W Afi ! Afi fj be the natural morphisms,
where Afi ´ AŒfi1 . We say that F is local if for any A 2 .salg/ and for any family
f˛i gi2I , ˛i 2 F .Afi / such that F .ij /.˛i / D F .j i /.˛j / for all i and j , there exists
a unique ˛ 2 F .A/ with F .i /.˛/ D ˛i for all possible families ffi gi2I described
above.
We want to rewrite this definition in more geometric terms in order to show that
this is essentially the gluing property appearing in the usual definition of sheaf on a
topological space.
We first observe that there is a contravariant equivalence of categories between the
category of commutative superalgebras .salg/ and the category of affine superschemes
.aschemes/, i.e., those superschemes that are the spectrum of some superalgebra (see
Chapter 10 for more details). The equivalence is realized by A 7! Spec.A/ and it is
explained in full details in Proposition 10.1.9. Hence a functor F W .salg/ ! .sets/ can
also be equivalently regarded as a functor F W .aschemes/op ! .sets/. With an abuse
of notation we shall use the same letter to denote both functors.
Let F be a local functor, regarded as F W .aschemes/op ! .sets/, and let FA be
its restriction to the affine open subschemes of Spec.A/. Then FA is a B-sheaf in
the usual sense (see Chapter 2); we must just forget the subscheme structure of the
affine subschemes of Spec.A/ and treat them as open sets in the topological space
Spec.A/, their morphisms being the inclusions. Then FA being a functor means that
it is a presheaf in the Zariski topology, while the property detailed in Definition B.2.1
 B.2 Sheafification of a functor 263

ensures the gluing of any family of local sections which agree on the intersection of
any two parts of an open affine covering.
We now want to give the definition of local functor in a more general setting, so
that it applies to the functor of points of supermanifolds and superschemes.
Definition B.2.2. Let F W .sspaces/op ! .sets/ be a functor. We say that F is local
or we also say it is a sheaf if it has the following property. For any superspace T
and any open covering fTi g of T let i W Ti ,! T , ij W Ti [ Tj ,! Ti be the natural
morphisms. If we have a family ˛i 2 F .Ti / such that F .ij /.˛i / D F .j i /.˛j /, then
there exists a unique ˛ 2 F .T / with F .i /.˛/ D ˛i .
One can readily check that this implies that, when F is restricted to the category of
the open sets of a fixed superspace, F is a sheaf in the ordinary sense. As we already
noticed, if we restrict F to the category of affine superschemes, this definition agrees
with the previous one.
We leave the following proposition as an exercise to the reader. It is very similar
to the proof in Chapter 10, Section 10.3.
Proposition B.2.3. If X is a superspace, its functor of points is local.
We now turn to the following problem. If we have a presheaf F on a topological
space in the ordinary sense, we can always build its sheafification, which is a sheaf
Fz together with a sheaf morphism ˛ W F ! Fz . This is the (unique) sheaf, which is
locally isomorphic to the given presheaf and has the following universal property: any
presheaf morphism  W F ! G , with G a sheaf, factors via ˛ (for more details on this
construction, see Chapter 2, Section 2.2). We now want to give the same construction
in our more general setting.
The existence of sheafification of a functor from the category of algebras to the
category of sets is granted in the ordinary case by [23], Ch. I, §1, no. 4, which is also
nicely summarized in [23], Ch. III, §1, no. 3. The proof is quite formal and one can
carry it to the supergeometric setting. We however prefer to introduce Grothendieck
topologies and the concept of site and to construct the sheafification of a functor through
them. In fact, as we shall see, very remarkably Grothendieck’s treatment is far more
general and it comprehends supergeometry. For more details we refer the reader to the
classical account by Grothendieck [38], [41] and the more modern treatment by Vistoli
[77].
Definition B.2.4. We call a category C a site if it has a Grothendieck topology, i.e., to
every object U 2 C we associate a collection of so-called coverings of U , that is, sets
of arrows fUi ! U g such that the following holds:
(1) If V ! U is an isomorphism, then the set fV ! U g is a covering.
(2) If fUi ! U g is a covering and V ! U is any arrow, then the fibered products
fUi U V g exist and the collection of projections fUi U V ! V g is a covering.
(3) If fUi ! V g is a covering and for each index i we have a covering fVij ! Ui g,
then the collection fVij ! Ui ! U g is a covering of U .
264 B Categories

The category of superschemes and its subcategory of affine superschemes are sites,
by taking for each object the collection of its (affine) coverings, as in the ordinary
setting. Similarly also the category of commutative superalgebras is also a site (for
the existence of fibered products in such categories and for more details on coverings
see Sections 10.3, 9.4). Such Grothendieck topologies, with an abuse of terminology,
are commonly referred to as Zariski topologies (one should prove that all of these
topologies are essentially equivalent). We shall not dwell upon the technicalities of
Grothendieck topologies, referring the reader to [77] where all the many subtleties are
discussed in the fullest detail.
Definition B.2.5. Let C be a site. A functor F W C op ! .sets/ is called a sheaf if for all
objects U 2 C, coverings fUi ! U g and families ai 2 F .Ui /, we have the following.
Let pij1 W Ui U Uj ! Ui , pij2 W Ui U Uj ! Uj denote the natural projections and
assume that F .pij1 /.ai / D F .pij2 /.aj / 2 F .Ui U Uj / for all i , j . Then there exists
a unique a 2 F .U / whose pullback to F .Ui / is ai for every i .
Again one can check that the functor of points of superschemes and supermanifolds
are sheaves in this more general setting for the corresponding sites.
We are ready for the sheafification of a functor in this very general setting.
Definition B.2.6. Let C be a site and let F W C op ! .sets/ be a functor. A sheafification
of F is a sheaf Fz W C op ! .sets/ with a natural transformation ˛ W F ! Fz such that:
(1) For any U 2 C and ;  2 F .U / such that ˛U . / D ˛U ./ in Fz .U /, there is a
covering fi W Ui ! U g with F .i /. / D F .i /./ in F .Ui /.
(2) For any U 2 C and any 2 Fz .U /, there is a covering fi W Ui ! U g and
elements i 2 F .Ui / such that ˛Ui . i / D Fz .i /. / in Fz .Ui /.
The next theorem states the fundamental properties of the sheafification.
Theorem B.2.7 ([77], p. 42). Let C be a site, F W C op ! .sets/ a functor.
(1) If Fz is a sheafification of F with ˛ W F ! Fz , then any morphism W F ! G,
with G a sheaf, factors uniquely through Fz .
(2) F admits a sheafification Fz , unique up to a canonical isomorphism.
We shall use this construction for the following supergeometric categories: C D
.smflds/, C D .sschemes/, C D .aschemes/, or equivalently C op D .salg/.
Observation B.2.8. Let F W C op ! .sets/ be a functor, Fz its sheafification, where C
is one of the supergeometric categories specified above. Let A be an object of C and
FA the restriction of the functor F to the category of the open subobjects of A (see
above). Then FzA is the sheafification of FA in the usual sense, that is, the sheafification
of F as a sheaf defined on the topological space underlying A. In particular, since a
sheaf and its sheafification are locally isomorphic, we have FA;p Š FzA;p , i.e., they
have isomorphic stalks (via the natural map ˛ W F ! Fz ) at any point p and for all
objects A. To simplify the notation we shall drop the suffix A and write just Fp instead
of FA;p .
 B.2 Sheafification of a functor 265

Proposition B.2.9. Let F; G W .sspaces/ ! .sets/ be local functors and let ˛ W F ! G

be a natural transformation. Assume that FA Š GA via ˛, where FA and GA denote
the ordinary sheaves corresponding to the restrictions of F and G to the category of
open subspaces in A (morphisms given by the inclusions). Then ˛ is an isomorphism,
hence F Š G.
Proof. We can certainly write an inverse for ˛A for every object A, the problem is to
see if it is well behaved on the arrows. However, this is true because ˛ is a natural
transformation. 
The rest of this section is devoted to proving the following result.
Theorem B.2.10. Let F; G W .salg/ ! .sets/ be two functors, with G a sheaf. Assume
that we have a natural transformation F ! G, which is an isomorphism on local
superalgebras, i.e., F .R/ Š G.R/ (via this map) for all local superalgebras R. Then
Fz Š G. In particular, F Š G if also F is a sheaf.
Lemma B.2.11. Let F W .salg/ ! .sets/ be a functor. For p 2 Spec.A/, let Fp D
lim F .R/, where the direct limit is taken for the rings R corresponding to the open
!
affine subschemes of Spec.A/ containing p. Then Fp D F .Ap /.
Proof. By Yoneda’s lemma, we have

Fp D lim F .R/ D lim Hom.hR ; F / D Hom.hlim R ; F / D Hom.hAp ; F / D F .Ap /

! ! !

as lim and Hom commute (see [51], p. 141) and Ap D lim R (see [2], p. 47). 
! !
Lemma B.2.12. Let A 2 .salg/, p 2 Spec.A0 /. Then Ap (the localization at p of A
as an A0 -module) is a local superalgebra, whose maximal ideal is m D .m0 ; .A1 /p /,
where m0 is the maximal ideal in the algebra .A0 /p D .Ap /0 .
Proof. From A D A0 ˚ A1 we get Ap D .A0 /p ˚ .A1 /p , and clearly this is a
superalgebra with .Ap /0 D .A0 /p , .Ap /1 D .A1 /p . Now let us consider m ´

m0 ; .A1 /p D m0 C .A1 /p . By the above, m ¤ Ap D .A0 /p ˚ .A1 /p . Take x 62 m:
then x D x0 C x1 with x0 2 .A0 /p , x1 2 .A1 /p , so x0 is invertible in .A0 /p .A1 /p
and x1 is nilpotent, hence x is invertible. 
We are ready for the proof of Theorem B.2.10:
Proof of Theorem B.2.10. Assume first that F and G are local. Since F .R/ Š G.R/
for all local algebras R, by Lemma B.2.11 this implies that Fp Š Gp for all p 2
Spec.A/ and all superalgebras A. Hence FA Š GA by [43], Ch. II, Section 1.1. By
Proposition B.2.9, we have F Š G (all isomorphisms have to be intended via the
natural transformation ˛ W F ! Fz ).
Now assume that F is not a sheaf. We have ˛ W F ! Fz ! G by Theorem B.2.7.
If A 2 .salg/, then by restricting our functors to the open affine sets in Spec.A/ we get
266 B Categories

FA ! FzA ! GA . By Observation B.2.8, FA and FzA are locally isomorphic via ˛, so

Fp Š FzA;p . By hypothesis, F .R/ Š G.R/, so Fp Š Gp by Lemma B.2.11, hence
Fzp Š Gp . Arguing as before, we obtain the result. 

Along the same lines, the reader can prove the following proposition:

Proposition B.2.13. Let  W F ! G be a natural transformation between two local

functors from .salg/ to .sets/. Assume that we know R for all local superalgebras R.
Then  is uniquely determined.

B.3 Super Nakayama’s lemma and projective modules

Let A be a commutative superalgebra.

Definition B.3.1. A projective A-module M is a direct summand of Amjn . In other

words, it is a projective module in the classical sense respecting the grading: M0 
Amjn
0 , M1  A1 .
mjn

Observation B.3.2. As in the classical setting, being projective is equivalent to the

exactness of the functor Hom.M;  /.

We want to show that a projective A-module has the property of being locally free,
that is its localization Mp into primes p of A0 is free as an Ap -module. This result
allows us to define the rank of a projective module as it happens in the ordinary case.
We start with a generalization of Nakayama’s lemma.

Lemma B.3.3 (Super Nakayama lemma). Let A be a local supercommutative ring with
maximal homogeneous ideal m. Let E be a finitely generated module for the ungraded
ring A.
(1) If mE D E, then E D 0; more generally, if H is a submodule of E such that
E D mE C H , then E D H .
(2) Let .vi /1ip be a basis for the k-vector space E=mE where k D A=m. Let
ei 2 E be above vi . Then the ei generate E. If E is a supermodule for the super ring
A, and vi are homogeneous elements of the super vector space E=mE, we can choose
the ei to be homogeneous too (and hence of the same parity as the vi ).
(3) Suppose that E is projective, i.e., there is an A-module F such that E˚F D AN ,
where AN is the free module for the ungraded ring A of rank N . Then E (and hence
F ) is free, and the ei above form a basis for E.

Proof. The proofs are easy extensions of the ones in the commutative case. We begin
the proof of (1) with the following observation: if B is a commutative local ring with n
a maximal ideal, then a square matrix R over B is invertible if and only if it is invertible
modulo n over the field B=n. In fact if this is so, det.R/ … n and so is a unit of B.
 B.3 Super Nakayama’s lemma and projective modules 267

P let ui , .1 < i < N / generate E. If E D mE, we can find mij 2 m so that

This said,
ui D j mij uj for all i . Hence, if L is the matrix with entries ıij  mij , then
0 1
u1
B u2 C
B C
L B : C D 0:
@ :: A
uN

It is now enough to prove that L has a left inverse P . Then multiplying the above from
the left by P , we get ui D 0 for all i and so E D 0. It is even true that L is invertible.
To prove this, let us consider B D A=JA where JA is the ideal generated by A1 . Since
JA  m we have
A ! B D A=JA ! k D A=m:
Let LB (resp. Lk ) be the reduction of L modulo JA (respectively modulo m). Then B
is local and its maximal ideal is m=JA , where Lk is the reduction of LB mod m=JA .
But B is commutative and Lk D I , and so LB is invertible. But then L is invertible.
If more generally we have E D H C mE, then E=H D m.E=H / and so E=H D 0,
which is to say that E D H .
To prove (2), let H be the submodule of E generated by the ei . Then E D mE CH
and so E D H .
We now prove (3). Clearly F is also finitely generated. We have k N D AN =mN D
E=mE ˚ F=mF . Let .wj / be a basis of F=mF and let fj be elements of F above wj .
Then by (2), the ei , fj form a basis of AN , while the ei (resp. fj ) generate E (resp. F ).
Now there are exactly N of the ei ; fj , and so if X denotes the .N  N /-matrix with
columns e1 ; : : : ; f1 ; : : : , then for some .N  N /-matrix Y over A we have X Y D I .
Hence XB YB D I where the suffix “B” denotes reduction modulo B. However, B is
commutative and so YB XB D I . Thus X has a left inverse over A, which must be Y
so that YX D I . If there is a linear relation among the ei and the fj , and if x is the
column vector whose components are the coefficients of this relation, then Xx D 0.
But this implies that x D YXx D 0. In particular E is a free module with basis .ei /.


We now wish to give a characterization of projective modules.

Theorem B.3.4. Let M be a finitely generated A-module, with A finitely generated

over A0 where A0 is noetherian. Then:
(1) M is projective if and only if Mp is free for all p primes in A0 .
(2) M is projective if and only if M Œfi1  is free for all fi ’s such that .f1 ; : : : ; fr / D
A0 .

Proof. (1) If M is projective, by part (3) of Nakayama’s lemma, we have that Mp is

free since it is a module over the supercommutative ring Ap .
268 B Categories

Now assume that Mp is free for all primes p 2 A0 . Recall that

HomAŒU 1  .M ŒU 1 ; N ŒU 1 / D HomA .M; N /ŒU 1 

for U a multiplicatively closed set in A0 . Recall also that given A0 -modules N , N 0 ,

N 00 , we have that 0 ! N 0 ! N ! N 00 is exact if and only if 0 ! Np0 ! Np ! Np00
is exact for all primes p in A0 . So given an exact sequence 0 ! N 0 ! N ! N 00 ,
since Mp is free, we obtain the exact sequence

0 ! Hom.Mp ; Np0 / ! Hom.Mp ; Np / ! Hom.Mp ; Np00 / ! 0

for all the primes p. Hence by the previous observation,

0 ! Hom.M; N 0 /p ! Hom.M; N /p ! Hom.M; N 00 /p ! 0

and
0 ! Hom.M; N 0 / ! Hom.M; N / ! Hom.M; N 00 / ! 0:

Thus M is projective.
(2) That Mp is free for all primes p is equivalent to M Œfi1  being free for
.f1 ; : : : ; fr / D A0 is a standard fact of commutative algebra and can be found in
[28], p. 623, for example. 
Definition B.3.5. Let M be a finitely generated projective A-module. We say that M
has rank rjs if Mp Š Arjs for all primes p in A. In the light of the previous proposition
one can show that the rank is locally constant, that is, if Mp Š Arjs , then there exists
an open neighbourhood U of p 2 Spec A for which Mp0 D Arjs , p 0 2 U .
Remark B.3.6. As in the ordinary setting we have a correspondence between projec-
tive A-modules and locally free sheaves on Spec A0 . In this correspondence, given a
projective A-module M , we view M as an A0 -module and build a sheaf of modules
OM on A0 . The global sections of this sheaf are isomorphic to M itself, and locally,
i.e., on the open sets Ufi D fp 2 Spec A0 j.fi / 6 pg, fi 2 A0 ,

OM .Ufi / D M Œfi1 :

More details on this construction in the ordinary setting can be found, for example, in
[43], Ch. II. As for its generalization to the super context, it is straightforward.

B.4 References
For a complete introduction to categories we refer the reader to [55], however a good
review of the main properties can also be found in [51]. As for the sheafification of a
functor we refer to [77] and of course to [38] by Grothendieck. Finally for the super
Nakayama lemma see [76] and for the ordinary setting [57].
C
Fréchet superspaces

The purpose of this appendix is to provide a quick introduction to Fréchet spaces,

superspaces and supersheaves, that is crucial for the understanding of a few key su-
pergeometric facts detailed in Chapter 4, Section 4.5. We first give an overview of the
ordinary setting, by providing the definition of Fréchet space and summarizing the main
results of the theory. It is of course impossible to give a fair account of this subject in
a few pages and for this reason we have provided precise references. We then proceed
to describe in more detail the theory of Fréchet superspaces and supersheaves.

C.1 Fréchet spaces

We want to start with the classical definition of Fréchet space. We first recall some
preliminary topology definitions, referring the reader to [64], Ch. I, for all the results
and further reading on these topics.
Let R be our ground field.

Definition C.1.1. A topological vector space V is both a topological space and a vector
space such that the vector space operations are continuous. A topological vector space
is locally convex if its topology admits a basis consisting of convex sets (a set A is
convex if .1  t /x C ty 2 A for all x; y 2 A and t 2 Œ0; 1).
We say that a locally convex topological vector space is a Fréchet space if its
topology is induced by a translation-invariant metric d and the space is complete with
respect to d , that is, all the Cauchy sequences are convergent.

There is an important characterization of Fréchet spaces that we are going to need

in the sequel. Let us start with the definition of seminorm and a proposition relating
seminorms and a translation-invariant metric.

Definition C.1.2. A seminorm on a vector space V is a real-valued function p such

that for all x; y 2 V and scalars a we have:
(1) p.x C y/  p.x/ C p.y/,
(2) p.ax/ D jajp.x/,
(3) p.x/  0.
Notice that the difference between the norm and the seminorm comes from the last
property: we do not ask that if x ¤ 0, then p.x/ > 0, as we would do for a norm.
270 C Fréchet superspaces

Proposition C.1.3. If fpi gfi2Ng is a countable family of seminorms on a topological

vector space V , separating points (that is, if x ¤ 0, there is an i with pi .x/ ¤ 0), then
there exists a translation-invariant metric d inducing the topology, defined in terms of
the fpi g:
X1
1 pi .x  y/
d.x; y/ D :
2i 1 C pi .x  y/
iD1

Proof. See [64], p. 27. 

The next proposition characterizes Fréchet spaces, giving an effective method to

construct them using seminorms.

Proposition C.1.4. A topological vector space V is a Fréchet space if and only if it

satisfies the following three properties:
• it is complete as a topological vector space;
• it is a Hausdorff space;
• its topology is induced by a countable family of seminorms fpi gi2N , i.e., U  V
is open if and only if for every u 2 U there exists K  0 and " > 0 such that
fv j pk .u  v/ < " for all k  Kg  U .

Definition C.1.5. We say that a sequence .xn / in V converges to x in the Fréchet space
topology defined by a family of seminorms if and only if it converges to x with respect
to each of the given seminorms. In other words, xn ! x if and only if pi .xn  x/ ! 0
for each i .
Two families of seminorms defined on the locally convex vector space V are said
to be equivalent if they induce the same topology on V .

To construct a Fréchet space, one typically starts with a locally convex topological
vector space V and defines a countable family of seminorms fpk g on V inducing its
topology and such that:
(I) if x 2 V and pk .x/ D 0 for all k  0, then x D 0 (separation property);
(II) if .xn / is a sequence in V which is Cauchy with respect to each seminorm, then
there exists x 2 V such that .xn / converges to x with respect to each seminorm
(completeness property).
The topology induced by these seminorms (as explained above) turns V into a
Fréchet space; property (I) ensures that it is Hausdorff, while the property (II) guaran-
tees that it is complete. A translation-invariant complete metric inducing the topology
on V can then be defined as above.
The most important example of Fréchet space, at least for the application we have
in mind, is the vector space C 1 .U /, the space of smooth functions on the open set
U Rn or more generally the vector space C 1 .M /, where M is a differentiable
 C.1 Fréchet spaces 271

manifold. For each open set U  Rn (or U  M ), for each K  U compact and for
each multi-index I , we define
ˇ @jI j  ˇ
ˇ ˇ
kf kK;I ´ sup ˇ I
.f / .x/ˇ; f 2 C 1 .U /:
x2K @x

Each k  kK;I defines a seminorm. The family of seminorms obtained by considering

all of the multi-indices I and the (countable number of) compact subsets K covering U
satisfies the properties (I) and (II) detailed above, hence makes C 1 .U / into a Fréchet
space. The sets of the form

ff 2 C 1 .U / j kf  gkK;I < "g

with fixed g 2 C 1 .U /, K U compact, and multi-index I are open sets and together
with their finite intersections form a basis for the topology.
All these constructions and results can be generalized to smooth manifolds, as we
now briefly outline.
Let M be a smooth manifold and let U be an open subset of M . If K is a compact
subset of U and D is a differential operator over U , then

pK;D .f / ´ sup jD.f /j

x2K

is a seminorm. The family of all the seminorms pK;D with K and D varying among
all compact subsets and differential operators respectively is a separating family of
1
seminorms endowing CM .U / with the structure of a complete locally convex vector
space. Moreover there exists an equivalent countable family of seminorms, hence
1
CM .U / is a Fréchet space. Let indeed fVj g be a countable open cover of U by open
S and let, for each j , fKj;i g be a countable family of compact subsets
coordinate subsets,
of Vj such that i Kj;i D Vj . We have the countable family of seminorms
ˇ @jI j  ˇ
ˇ ˇ
pK;I ´ sup ˇ .f / .x/ˇ; K 2 fKj;i g;
x2K @x I
1
inducing the topology. Notice that CM .U / is also an algebra: the product of two
smooth functions being a smooth function. It is thus natural in this context to introduce
the notion of Fréchet algebra.
Definition C.1.6. A Fréchet space V is said to be a Fréchet algebra if its topology
can be defined by a countable family of submultiplicative seminorms, i.e., a countable
family fqi gi2N of seminorms satisfying

qi .fg/  qi .f /qi .g/ for all i 2 N:

If we come back to our prototype of Fréchet space C 1 .Rn /, one can check that it
is also a Fréchet algebra, the countable family of submultiplicative seminorms given
272 C Fréchet superspaces

by
ˇ @jI j  ˇ
ˇ ˇ
qKi;m ;j .f / ´ 2 j
sup ˇ I
.f / .x/ˇ:
x2Ki;m ;jI jj @x

We end this section with a definition of a Fréchet sheaf.

Definition C.1.7. Let F be a sheaf of real vector spaces over a manifold M . F is a

Fréchet sheaf if:
(1) for each open set U M , F .U / is a Fréchet space;
(2) for each open set U M and for each open cover fUi g of U , the topology of
F .U / is the initial topology with respect to the restriction maps F .U / ! F .Ui /,
that is, the coarsest topology making the restriction morphisms continuous.
As a consequence, we have that each restriction map F .U / ! F .V / (V U ) is
continuous.
A morphism of sheaves W F ! F 0 is said to be continuous if the map F .U / !
F 0 .U / is continuous for each open subset U M .

C.2 Fréchet superspaces

We are now ready to introduce the notion of Fréchet space in the super context.

Definition C.2.1. (1) We say that a super vector space V D V0 ˚ V1 is a super Fréchet
space if there exist two families of homogeneous seminorms fp0i g and fp1i g defined
on V0 and V1 , respectively, with respect to which V0 and V1 are Fréchet spaces. We
will usually denote by pi a generic seminorm from one of the two families.
(2) Let A D A0 ˚ A1 be a super Fréchet space with respect to the family of
seminorms fpi gi2I . Suppose also that A is a superalgebra with multiplication m. We
say that A is a super Fréchet algebra if the topology is defined by an equivalent family
of submultiplicative seminorms fqi g: qi .ab/  qi .a/qi .b/.
(3) Suppose now that F is a sheaf such that F .U / is a super Fréchet algebra for
each U . We say that F is a super Fréchet sheaf if for each open set U and for any
open cover fUi g of U the topology of F .U / is the initial topology with respect to the
restriction maps F .U / ! F .Ui /.

We now turn to the example most interesting to us, namely the superalgebra of
sections on a supermanifold.
Let M D .jM j; OM / be a supermanifold. Fix now an open subset U jM j. For
each compact subset K  U and each differential operator D over U , define

B
pK;D .f / ´ sup j.D.f //.x/j; f 2 OM .U /:
x2K
 C.2 Fréchet superspaces 273

As before, one can readily check that each pK;D defines a seminorm. The family
of seminorms obtained by considering all the differential operators and the compact
subsets K covering U endows OM .U / with a Hausdorff locally convex topology (as
before), where the open sets that form a basis for the topology are
ff 2 OM .U / j pK;D .f  g/ < "g;
with fixed g 2 OM .U /, K U compact, D 2 Diff.U /, the differential operators on
U , and " > 0, together with their finite intersections.
In complete analogy with the ordinary setting one can prove the following propo-
sition.
Proposition C.2.2. OM is a super Fréchet sheaf.
Before proving it we need some preliminary results that we collect in the next
lemma and proposition.
Lemma C.2.3. Let U be a chart with coordinates t i ,  j and let ffn g be a sequence in
OM .U /: X
1
fn D fnI  I ; fnI 2 CM .U /:
I
(1) ffn g is a Cauchy sequence in OM .U / if and only if the sequence ffnI g is a
Cauchy sequence in C 1 .U / for each I .
(2) OM .U / is complete, moreover it is a super Fréchet algebra.
P j jCjI j
Proof. (1) Let K be a compact subset of U and let D D a ;I @@t  I be a super

F
differential operator. We have

X ˇ X X  ˇ
ˇ @ ˇ
pK;D . fJ  J / D sup ˇ a ;I fJ  J
.x/ˇ
x2K @t @ I
J ;I J
X
 max.a ;I /pK; @ .fI /;
x2K @t 
;I

which finishes the proof of (1).

1
(2) It is well known that CM .U / is complete with respect to such a locally convex
topology (see, for instance, [73]).
In order to show that OM .U / is a Fréchet space it is then enough to produce
a countable family of seminorms satisfying the properties detailed above. This is
obtained, as in the classical case, by considering the family fpKn ; @ gn; ;I , where
@t  @ I
each Kn isS a family of compact sets in U such that Kn is contained in the interior of
KnC1 and Kn D U (see [16], [26]).
We now show that OM .U / is a Fréchet superalgebra. This follows considering the
family of seminorms
q˛;Kn ´ 2˛C2n max .pKn ; @ /; ˛ 2 N:
j j˛;jI j @t  @ I
274 C Fréchet superspaces

It is not difficult to check that the family fq˛;Kn g˛;n is equivalent to the family
fpKn ; @ gn; ;I and that they are submultiplicative. 
@t  @ I

Proposition C.2.4. Let fUi g be an open cover of an open set U  jM j, let also fsn g
be a sequence in OM .U /.
(1) fsn g converges to s in OM .U / if and only if fsn jUi g converges to sjUi in OM .Ui /
for each Ui .
(2) fsn g is a Cauchy sequence in OM .U / if and only if fsn jUi g is a Cauchy sequence
in OM .Ui / for each Ui .
(3) OM .U / is complete for each open subset U .
(4) OM .U / is a super Fréchet algebra for each open subset U .
Proof. (1) Clearly if fsn g ! 0, then sn jUi ! 0 for each i . Suppose vice versa that for
each integer i, sn jUi ! 0 and let K be a compact subset of U . There exists a finite
open cover Vj of K such that each Vxj is compact and for each j there exists i such that
Vj  Ui (see [26]). Then .1/ follows from the inequality
X
pK;D .sn /  pK\Vxj ;Dj x .sn /:
K\Vj
j

(2) If fsn g is a Cauchy sequence in OM .U / and fUi g are charts covering U , then
there exists a family fsUi g such that sn jUi converges to sUi for each i . An easy check
shows that the various sUi glue together and define a section s in OM .U /. Clearly sn
converges to s.
(3) The fact that OM .U / is complete follows by Lemma C.2.3, by considering a
countable open cover fUi g of U by open supercharts and points (1) and (2).
(4) The fact that OM .U / is a super Fréchet algebra follows by considering the
family of seminorms

q˛;Ki;n ´ 2˛ max .pKi;n ; @ /;

j j˛;jI j @t  @ I

where fKi;n gn2N is a countable family of compact subsets covering Ui . 

Proof of Proposition C.2.2. The fact that OM .U / is a Fréchet algebra for each open
subset U is the content of the previous proposition. The fact that the topology of
OM .U / is the initial topology induced by the restriction maps is the content of item
(1) of Proposition C.2.4. 
With such a topology, all the important geometrical operations that can be performed
on the sheaf turn out to be continuous. We have in fact the following proposition.
Proposition C.2.5. (1) If D is a super differential operator on M , then D W OM .U / !
OM .U / is continuous for each U .
(2) If W M ! N is a supermanifold morphism, then  W O.N / ! O.M / is
continuous.
 C.2 Fréchet superspaces 275

Proof. (1) is clear because of the way the seminorms, hence the topology, are defined.
(2) Due to Proposition C.2.4 we can suppose that M and N are superdomains
with coordinates t i ,  j and x r , s , respectively. We have to prove that if ffn g is a
sequence in O.N / converging to zero, then the sequence f  .fn /g converges to zero
j˛jCjI j
in O.M /, i.e., for all compact subsets K and all differential operators D D @@t ˛ @ I ,
pD;K .  .fn // tends to zero. Let a 2 Diff 0 .M /, the differential operators of degree
zero, namely a 2 O.M /. Then

pa;K .  G
.fn // D sup j.a  
.f //.x/j n (C.1)
x2K
 . sup ja.x/j/. sup jfn .j j.x//j/; (C.2)
x2K x2K

and the result follows easily. If D D @y@

with y denoting indifferently an even or odd
variable, then using the chain rule (4.4), we have
  r 
@ .x /  @fn
pD;K .  .fn // ´ sup
x2K @y @x r
j˛j
and the result follows from equation (C.1). The case of a differential operator D D @@t ˛
is treated similarly. 
Bibliography

The numbers at the end of each item refer to the pages on which the respective work is cited.

[1] R. Abraham, J. E. Marsden, and T. Ratiu, Manifolds, tensor analysis and applications.
Appl. Math. Sci. 75, Springer-Verlag, New York 1993. 102
[2] M. F. Atiyah and I. G. MacDonald, Introduction to commutative algebra. Addison-Wesley
Publishing Co., Reading, Mass., 1969. 44, 77, 197, 265
[3] Yu. A. Bahturin, A. A. Mikhalev, V. M. Petrogradsky, and M. V. Zaicev, Infinite dimensional
Lie superalgebras. De Gruyter Exp. Math. 7, Walter de Gruyter, Berlin 1992. 9
[4] L. Balduzzi, C. Carmeli, and G. Cassinelli, Super G-spaces. In Symmetry in mathematics
and physics, Contemp. Math. 490, Amer. Math. Soc., Providence, RI, 2009, 159–176. 153,
172
[5] L. Balduzzi, C. Carmeli, and R. Fioresi, Quotients in supergeometry. In Symmetry in
mathematics and physics, Contemp. Math. 490, Amer. Math. Soc., Providence, RI, 2009,
177–187. 172
[6] C. Bartocci, U. Bruzzo, and D. Hernandez-Ruiperez, The geometry of supermanifolds.
Math. Appl. 71, Kluwer Academic Publishers, Dordrecht 1991. 89
[7] M. Batchelor, Graded manifolds and vector bundles: A functorial correspondence. J. Math.
Phys. 26 (1985), 1578–1582. 63
[8] M. Batchelor, Two approaches to supermanifolds. Trans. Amer. Math. Soc. 258 (1980),
257–270.
[9] M. Batchelor, The structure of supermanifolds. Trans. Amer. Math. Soc. 253 (1979), 329–
338.
[10] F. A. Berezin, Introduction to superanalysis. Math. Phys. Appl. Math. 9, D. Reidel Pub-
lishing C., Dordrecht 1987. viii, 15, 27, 53, 89
[11] F. A. Berezin and D. A. Leites, Supermanifolds. Dokl. Akad. Nauk SSSR 224 (1975),
505–508. 89, 140
[12] G. Bergman, The diamond lemma for ring theory. Adv. Math. 29 (1978), 178–218. 24
[13] A. Borel, Linear algebraic groups. Grad. Texts in Math. 126, Springer-Verlag, New York
1991.
[14] H. Boseck, Affine Lie supergroups. Math. Nachr. 143 (1989), 303–327. 230
[15] H. Boseck, Classical Lie supergroups. Math. Nachr. 148 (1990), 81–115. 230
[16] N. Bourbaki, General topology. Chapters 1-4, Elem. Math. (Berlin), Springer-Verlag,
Berlin 1989. 273
[17] N. Bourbaki, Éléments de mathématique. Fasc. XXXIII. Variétés différentielles et analy-
tiques. Fascicule de résultats (Paragraphes 1 à 7), Actualités Scientifiques et Industrielles
1333, Hermann, Paris 1967. 82
278 Bibliography

[18] N. Bourbaki, Lie groups and Lie algebras. Chapters 1–3, Elem. Math. (Berlin), Springer-
Verlag, Berlin 1998.
[19] U. Bruzzo and R. Cianci, Differential equations, Frobenius theorem and local flows on
supermanifolds. J. Phys. A Math. Gen. 18 (1985), 417–423. 111
[20] V. Chari and A. Pressley, A guide to quantum groups. Cambridge University Press, Cam-
bridge 1994. 26, 113
[21] S.-J. Cheng and W. Wang, Dualities for Lie superalgebras. Preprint 2009.
arXiv:1001.0074v2 [math.RT] 255
[22] P. Deligne and J. W. Morgan, Notes on supersymmetry (following Joseph Bernstein).
In Quantum fields and strings; a course for mathematicians, Vol. 1, Amer. Math. Soc.,
Providence, RI; Institute for Advanced Study (IAS), Princeton, NJ, 1999, 41–97. viii, 24,
26, 27, 53, 89, 111, 140, 153, 170, 229, 230
[23] M. Demazure and P. Gabriel, Groupes algébriques. Tome I: Géométrie algébrique, général-
ités, groupes commutatifs, Masson & Cie, Éditeur, Paris; North-Holland Publishing Co.,
Amsterdam, 1970. 28, 41, 42, 43, 44, 183, 186, 188, 197, 201, 202, 210, 216, 262, 263
[24] M. Demazure and A. Grothendieck, Schémas en groupes. I, II, III. Séminaire de géométrie
algébrique du Bois Marie 1962/64 (SGA 3). Dirigé par M. Demazure et A. Grothendieck.
Lecture Notes in Math. 151, 152, 153, Springer-Verlag, Berlin 1970.
[25] B. DeWitt, Supermanifolds. Cambridge Monogr. Math. Phys., Cambridge University
Press, Cambridge 1984. viii
[26] J. Dieudonné, Éléments d’analyse. Tome III: Chapitres XVI et XVII, Cahiers Scientifiques
XXXIII, Gauthier-Villars Éditeur, Paris 1970. 82, 119, 273, 274
[27] J. Dixmier, Enveloping algebras. Grad. Stud. Math. 11, Amer. Math. Soc., Providence,
RI, 1996. 249, 250
[28] D. Eisenbud, Commutative algebra with a view toward algebraic geometry. Grad. Texts
in Math. 150, Springer-Verlag, New York 1995. 268
[29] D. Eisenbud and J. Harris, The geometry of schemes. Grad. Texts in Math. 197, Springer
Verlag, New York 2000. 28, 33, 35, 36, 37, 39, 41, 43, 44, 65, 180, 183, 262
[30] R. Fioresi, On algebraic super-groups and quantum deformations. J. Algebra Appl. 2
(2003), 403–423. 206
[31] R. Fioresi and F. Gavarini, Chevalley supergroups. Mem. Amer. Math. Soc., to appear; see
also arXiv:0808.0785v5 [math.RA]. 140
[32] R. Fioresi, M. A. Lledó, and V. S. Varadarajan, The Minkowski and conformal superspaces.
J. Math. Phys. 48 (2007), no. 11, 113505-1–27. 170, 172
[33] L. Frappat, A. Sciarrino, and P. Sorba, Dictionary on Lie algebras and superalgebras.
Academic Press, Inc., San Diego, CA, 2000. 240
[34] D. Z. Freedman, P. van Nieuwenhuizen, and S. Ferrara, Progress toward a theory of su-
pergravity. Phys. Rev. D 13 (1976), 3214–3218. 140
[35] R. Giachetti and R. Ricci, R-actions, derivations, and Frobenius theorem on graded man-
ifolds. Adv. Math. 62 (1986), 84–100. 111
 Bibliography 279

[36] Yu. A. Gol’fand and E. P. Likhtman, Extension of the algebra of Poincaré group generators
and violation of P invariance. JETP Lett. 13 (1971), 323–326. 140
[37] P. Griffiths and J. Harris, Principles of algebraic geometry. Wiley Classics Lib., John Wiley
& Sons, Inc., New York 1994. 35, 44
[38] A. Grothendieck, Fondaments de la géometrie algébrique. Secretariat Mathematique, Paris
1962. 44, 263, 268
[39] A. Grothendieck, Produits tensoriels topologiques et espaces nucléaires. In Séminaire
Bourbaki, Vol. 2, Exp. No. 69, Soc. Math. France, Paris 1995, 193–200. 75
[40] A. Grothendieck, Résumé de la théorie métrique des produits tensoriels topologiques.
Reprint of Bol. Soc. Mat. São Paulo 8 (1953), 1–79; Resenhas 2 (1996), 401–480. 75
[41] A. Grothendieck, Technique de descente et théorèmes d’existence en géometrie algébrique.
I. Généralités. Descente par morphismes fidèlement plats. In Séminaire Bourbaki, Vol. 5,
Exp. No. 190, Soc. Math. France, Paris 1995, 299–327. 263
[42] C. Gruson and V. Serganova, Cohomology of generalized supergrassmannians and char-
acter formulae for basic classical Lie superalgebras. Proc. London Math. Soc. (3) 101
(2010), 852–892. 255
[43] R. Hartshorne, Algebraic geometry. Grad. Texts In Math. 52, Springer-Verlag, New York
1977. 28, 30, 33, 35, 36, 37, 43, 44, 176, 193, 265, 268
[44] V. G. Kac, Lie superalgebras. Adv. Math. 26 (1977), 8–26. 230, 231, 235, 236, 240, 241,
244, 250, 254
[45] V. G. Kac, Representations of classical Lie superalgebras. In Differential geometrical
methods in mathematical physics, II, Lecture Notes in Math. 676, Springer-Verlag, Berlin
1978, 597–626. 231, 240, 250, 255
[46] V. G. Kac, A sketch of Lie superalgebras theory. Comm. Math. Phys. 53 (1977), 31–64.
[47] A. W. Knapp, Advanced real analysis. Cornerstones, Birkhäuser, Boston, Mass., 2005.
102
[48] I. Kolář, P. W. Michor, and J. Slovák, Natural operations in differential geometry. Springer-
Verlag, Berlin 1993. 156
[49] B. Kostant, Graded manifolds, graded Lie theory, and prequantization. In Differential
geometrical methods in mathematical physics, Lecture Notes in Math. 570, Springer-
Verlag, Berlin 1977, 177–306. viii, 27, 53, 89, 112, 124, 140
[50] J.-L. Koszul, Graded manifolds and graded Lie algebras. In Proceedings of the interna-
tional meeting on geometry and physics, Pitagora Editrice, Bologna 1983,71–84. 111,
112, 137, 140
[51] S. Lang, Algebra. Addison Wesley, Reading, Mass., 1984. 4, 44, 259, 265, 268
[52] S. Lang, Real and functional analysis. Grad. Texts in Math. 142, Springer-Verlag, New
York 1993. 90
[53] D. A. Leites, Introduction to the theory of supermanifolds. Russian Math. Surveys 35
(1980), 1–64. viii, 89, 90, 102
[54] D. A. Leites, Quantization and supermanifolds. Appendix 3 in F. Berezin, M. Shubin The
Schrödinger equation, Math. Appl. (Soviet Ser.) 66, Kluwer Academic Publishers Group,
Dordrecht 1991. 97
280 Bibliography

[55] S. Maclane, Categories for the working mathematician, Grad. Texts in Math. 5, Springer-
Verlag, New York 1998. 268
[56] Y. I. Manin, Gauge field theory and complex geometry. Translated by N. Koblitz and J. R.
King, Grundlehren Math. Wiss. 289, Springer-Verlag, Berlin 1997. 27, 53, 89, 90, 102,
111, 172, 194, 201
[57] H. Matsumura, Commutative ring theory. Cambridge Stud. Adv. Math. 8, Cambridge
University Press, Cambridge 1989. 268
[58] P. W. Michor, Topics in differential geometry. Grad. Stud. Math. 93, Amer. Math. Soc.,
Providence, RI, 2008. 76, 77, 84, 89
[59] S. Montgomery, Hopf algebras and their actions on rings. CBMS Regional Conf. Ser. in
Math. 82 Amer. Math. Soc., Providence, RI, 1993. 26
[60] D. Mumford, The red book of varieties and schemes. Lecture Notes in Math. 1358,
Springer-Verlag, Berlin 1988.
[61] J. Nestruev, Smooth manifolds and observables. Grad. Texts in Math. 220, Springer-Verlag,
New York 2003.
[62] I. Penkov and V. Serganova, Generic irreducible representations of finite-dimensional Lie
superalgebras. Internat. J. Math. 5 (1994), 389–419. 245, 251, 252
[63] I. Penkov, Characters of typical irreducible finite-dimensional q.n/-modules. Funct. Anal.
Appl. 20 (1986), 30–37. 248
[64] W. Rudin, Functional analysis. Internat. Ser. Pure Appl. Math., McGraw Hill, New York
1991. 84, 269, 270
[65] A. Salam and J. Strathdee, Super-gauge transformations. Nucl. Phys. B 76 (1974), 477–
482. vii, 140
[66] M. Scheunert, The theory of Lie superalgebras. Lecture Notes in Math. 716, Springer-
Verlag, Berlin 1978.
[67] V. Serganova, Kac-Moody superalgebras and integrability. Preprint. 252, 254, 255
[68] V. Serganova, Kazhdan-Lusztig polynomials and character formula for the Lie superalge-
bra gl.mjn/. Selecta Math. (N.S.) 2 (1996), 607–651. 255
[69] V. Serganova, On generalizations of root systems. Comm. Algebra 24 (1996), 4281–4299.
[70] J.-P. Serre, Faiseaux algebrique coherent. Ann. of Math. 61 (1955), 197–278. 44
[71] R. Steinberg, Lectures on Chevalley groups. Notes prepared by John Faulkner and Robert
Wilson, Yale University, New Haven, Conn., 1968. 249
[72] M. E. Sweedler, Hopf algebras. Math. Lecture Note Ser., W. A. Benjamin, Inc., New York
1969. 26, 83, 120, 123
[73] F. Trèves, Topological vector spaces, distributions and kernels. Academic Press, NewYork
1967. 75, 273
[74] G. M. Tuynman, Supermanifolds and supergroups. Math. Appl. 570, Kluwer Academic
Publishers, Dordrecht 2004.
[75] V. S. Varadarajan, Lie groups, Lie algebras, and their representations. Grad. Texts in Math.
102, Springer-Verlag, New York 1984. 125, 154, 157, 158, 245
 Bibliography 281

[76] V. S. Varadarajan, Supersymmetry for mathematicians: an introduction. Courant Lect.

Notes Math. 11, New York University, Courant Institute of Mathematical Sciences, New
York; Amer. Math. Soc., Providence, RI, 2004. viii, 24, 27, 89, 90, 102, 111, 172, 268
[77] A. Vistoli, Grothendieck topologies, fibered categories and descent theory. In Fundamental
algebraic geometry, Math. Surveys Monogr. 123, Amer. Math. Soc., Providence, RI, 2005,
1–104. 165, 263, 264, 268
[78] F. W. Warner, Foundations of differentiable manifolds and Lie groups. Grad. Texts in Math.
94, Springer-Verlag, New York 1983. 61, 98, 102
[79] W. C. Waterhouse, Introduction to affine group schemes. Grad. Texts in Math. 66, Springer-
Verlag, New York 1979.
[80] J. Wess and B. Zumino, Supergauge transformations in four dimensions. Nucl. Phys. B 70
(1974), 39–50. vii, 140
Index

action, 141, 223, 227 closed, 197

of a Lie group, 141 closed embedding, 62, 200, 223
of a super Lie group, 141 closed submanifolds, 49
Ad, 213 closed subscheme, 176
adjoint morphism, 213 coadjoint orbits, 165
adjoint representation, 122 coalgebra, 25
affine algebraic supervariety, 175 coequalizer, 150
affine algebraic supergroups, 221 coherent sheaf, 43
affine scheme, 36 coideal, 26
affine supergroup scheme, 203 complex analytic supermanifold, 86
affine superscheme, 50, 175 complex conjugate of a complex super-
affine superspace, 50, 177, 182 manifold, 87
algebraic superscheme, 196 composition law, 259
analytic superdomain, 86 comultiplication, 25
anti-symmetry, 232 conformal superspace, 166
antipode, 25 consistent super symmetric bilinear form,
antisymmetrizer, 126 233
constant rank, 92, 93
basic, 235, 239 constant rank morphism, 143
basic open sets, 36 contravariant functor, 260
Berezin, F. A., v, viii, 12, 53, 89 convolution product, 119
Berezinian, 12, 212 coordinate ring, 175
Bernstein, J., v, viii, 53 correspondence Lie subgroups, Lie super-
Borel subalgebra, 249 algebras, 136
bosons, 63 counit, 25
bracket, 208 covariant functor, 260
bracket, super, 8 covering, 263
cyclic subalgebra, 137
canonical basis, 1
canonical chart, 56 Deligne, P., v
Cartan matrices, 242 Der, 7
Cartan type, 239 derivation, 67
category, 259 derivation, of a superalgebra, 7
chain rule, 72 De Witt, B., viii
chart, 56, 59 Diamond Lemma, 24
Chart Theorem, 58 differential, 92
Chart Theorem, global version, 60 of a function, 199
classical, 235 of a germ, 199
classical type, 235 of a morphism, 199
284 Index

of a morphism of supermanifolds, 67 representation, 261

of a section, 199 functor of points, 38, 51, 180
differential operators, 73 of a quotient, 162
dimension of supermanifolds, 79
of a supermanifold, 59 of the affine space, 39
dimension, super, 1 of the Grassmannian scheme, 42
distribution, 83, 103 of the projective space, 41
distributions, 119 functorial, 260
dominant integral, 248
dual numbers, 210 general linear supergroup, 12, 50, 81, 205
Dynkin diagrams, 242 germs of sections, 32, 65
global coordinates, 46
embedded submanifold, 97 global sections, 31
embedding, 97 global super Frobenius Theorem, 110
End(V), 8 GLpjq , 46
equalizer functor, 150 graded superalgebra, 21
étalé space, 32 Grassmannian scheme, 42
evaluation map, 130 Grassmannian superscheme, 190
even rules, 26, 208 Grothendieck topology, 263
exponential, 218 Grothendieck, A., 263
extended vector fields, 71 G-space, 141
extended vectors, 71 G supermanifold, 142
extension of scalars, 26
Hadamard’s lemma, 57
fermions, 63 Heisenberg superalgebra, 256
fibered product, 163, 186, 261 highest weight, 248
filtered, superalgebra, 21 highest weight vector, 248
finite support distribution, 83 homogeneous, 1
first neighbourhood, 213 homogeneous one-parameter supergroups,
flag supermanifold, 169 137
Fréchet algebra, 271 homogeneous space, 154
Fréchet sheaf, 272 homogeneous superspace, 155, 158
Fréchet spaces, 269 Hopf ideal, 25
Fréchet superalgebra, 75 Hopf superalgebra, 25, 203
Fréchet supersheaf, 74
Fréchet Superspaces, 269 immersed submanifolds, 97
Freed, D., v immersion, 93, 97
functor, 260 induced module, 249
faithful, 261 integrable distribution, 104
full, 261 inverse, 112
fully faithful, 261 inverse function theorem, 90
local, 163 involutive distribution, 104
representable, 261 isomorphism (category), 259
 Index 285

JA , 5 of ringed spaces, 34
Jacobi identity, 232 of schemes, 36
Jacobian, 68, 72, 92 of superdomains, 55
JM , 49 of supermanifolds, 48
of superspaces, 47, 174
Kostant, B., v, viii, 53, 82, 89, 112, 118 sheaf, 32
Koszul, J.-L., 112 super vector space, 2
morphisms
left translation, 118 between objects in a category, 259
left-invariant, 221 Mred , 49
left-invariant vector fields, 115 multiplication, 112
Leibniz identity, super, 221
Leites, D. A., v, viii, 89 natural transformation, 260
L.G/, 221
Lie algebra, super, 8 Ob.C /, 259
Lie algebra-valued functor, 208 objects in a category, 259
Lie superalgebra, 207, 232 open affine subfunctor, 185
line in root space, 245 open subfunctor, 163, 185
linear representation, 223 open submanifold, 61
local coordinates, 48 open subscheme, 174
local functor, 163, 183, 262 open subspace, 46
local super Frobenius theorem, 103 open supermanifold subfunctor, 163
local superdiffeomorphism, 91 opposite category, 260
locality, 183 orbit morphism, 143
localization, 78 orthosymplectic Lie superalgebra, 233
locally convex, 269 osp.V /, 233
locally finite refinement, 61
locally ringed space, 34 parity, 1
parity
ˇ reversing
ˇ functor, 2
z , 49, 62 @ ˇ @ ˇ
M ,
@t i x @ j x
, 68
Manin, Yu., v, 53 partition of unity, 61
maximal spectrum, 74 P .n/, 234
Milnor exercise, 76 Poincaré–Birkhoff–Witt Theorem, 18
Minkowski space, vii Poincaré group, vii
Minkowski superspace, 166 Poincaré supergroup, 169
module, 9 point supported distribution, 83
Morgan, J. W., v presheaf, 30
morphism product of supermanifolds, 65, 75
Hopf superalgebra, 25 projective localization, 193
of complex analytic supermanifolds, projective module, 266
86 projective space, 37
of locally ringed spaces, 34 projective special linear Lie superalgebra,
of real analytic supermanifolds, 86 232
286 Index

Projective superspace, 179 sheaf, 30

projective tensor topology, 75 sheaf functor, 262
Proj S , 37 sheaf of OX -modules, 43
sheafification, 33
Q.n/, 234 sheafification of a functor, 262
quasi-coherent OA -modules, 176 sign rule, 3
quasi-coherent sheaf, 43, 100 simple, 235
quasi-coherent sheaf of ideals, 176 simple roots, 239
quotient, 157 simple system, 239
Spec A, 35
rank of a projective module, 268 Spec A, 36
rational, 197 special linear group, 206
real analytic supermanifold, 86 special linear Lie superalgebra, 232
real spectrum, 76 spectrum, 35
real structure, 88 spectrum of a superalgebra, 174
reduced, 174 spectrum of superring, 50
reduced manifold, 62 stabilizer, 227
reduced scheme, 175 stabilizer functor, 149
reduced space, 175 stabilizer subgroup, 149
reduced supermanifold, 49 stabilizer supergroup functor, 227
refinement (of a cover), 61 stalk, 31
regular ideal, 99 standard monomials, 18
representability criterion, 162, 163, 183 strange, 234
representability theorem, classical, 40 structure sheaf, 36, 174
representation super ringed space, 45
of a Lie superalgebra, 17 subfunctor, 163
representations of Lie superalgebras, 248 open, 185
restriction, 30 open affine, 185
right translation, 118 submanifolds, 97
right-invariant vector fields, 130 submersion, 93
ringed space, 34 submultiplicative seminorm, 75
root system, 239 super derivation, 221
Rpjq , 56 super Fréchet algebra, 272
Mpjq , 46 super Fréchet sheaf, 272
Rpjq , 46 super Fréchet spaces, 272
super Frobenius theorem, 103
(salg), 5 super Harish-Chandra pairs, 112, 123
scheme, 36 super Lie group, 112
Schur’s lemma, 256 super Nakayama’s lemma, 266
sections of a sheaf, 31 super reduced, 174
seminorm, 74, 269 super ringed space, 45, 55
Serre’s twisting sheaf, 193 super symmetric bilinear form, 233
Serre, J-P., 44 super tangent bundle, 70
 Index 287

super tangent space, 67 Taylor series expansion, 56

super vector bundle, 71 topological dual, 83
superalgebra, 4 topological vector space, 269
associative, 4 T -point, 51
commutative, 4 tpointT -point, 81
exterior, 6 transition functions, 64
filtered, 21 transitive action, 154
graded, 21 transversal, 101
Hopf, 25 transversality, 100
of matrices Mat, 11 twisted module, 192
symmetric, 6
superchart, 56, 59 U.g/, 16
supercoherent sheaves, 193 unit, 112
superdiffeomorphism, 91 universal enveloping superalgebra, 16
superdimension, 48, 55, 59 U pjq , 56
superdomain, 48, 54, 59
supergroup functor, 203 value, 55, 199
supergroup scheme, 203 of a section, 48
supermanifold, 48, 59 Varadarajan, V. S., v, viii
supermatrices, 50, 183, 204, 209 VecM , 70
superscheme, 50, 174 vector field, 70
superspace, 46, 54, 174 vector space, super, 1
supersymmetries, 63 Verma module, 249
supersymmetry, vii Vistoli, A., 263
support, of a distribution, 83 weight representation, 248
Sweedler dual, 83 weight space, 248
Sweedler, M. E., 82 Wess J., vii
symmetrizer, 24
system of (super) coordinates, 56 Yoneda’s lemma, 40, 52
tangent bundle, 70 Zariski sheaf, 183
tangent space, 196 Zariski topology, 35, 183
tangent space at the identity, 213 Zumino B., vii
tangent vector
to a supermanifold, 67