Lectures on supersymmetry

Joseph Bernstein

Lecture 2

1. Super liner algebra.

1.1. A super vector space (over a eld) is a Z2-graded vector space, standard notation:
V = V0  V1 . We let p denote the function of parity dened on homogeneous elements.
Super vector spaces form an abelian category endowed with a functor of change of parity
that we denote as . There is also a tensor structure on the category of super vector
spaces:

() (V0  V1 )
(W0  W1 ) := V0
W0  V1
W1  V0
W1  V1
W0
| {zx |
} {zx }
? ?
even component odd component
The braiding operation S : V
W ! W
V is dened by the rule:

() SV
W (X
Y ) = ( 1)p(X )p(Y )Y
X:

If V = V0 V1 is a super vector space, we set (m; n) := (dim V0; dim V1 ) to be its dimension.
All types of algebras in the super world are dened as algebras in the tensor category of
super vector spaces. We will give below a more down-to-Earth denition of basic algebraic
structures.
1
1.2.
Associative super algebra is a Z2-graded algebra (of course, the product operation
must respect the Z2-grading). Left and right modules over an associative super algebra
are dened in an obvious manner, and the corresponding categories will be denoted by
A mod and mod A, respectively. On each of these categories we have the parity
change functor .
An associative algebra is said to be commutative if 8 a; b a
b = ( 1)p(a)p(b)b
a.
Lemma-Construction. Let A be a commutative super algebra. Then every left A-module
M acquires in a functorial way a right A-action. Moreover, the right and the left A-actions
on M commute.
Proof. For a 2 A, m 2 M we let m
a := ( 1)p(a)p(m)am. This is a right action and we
have:
(b
m)a = ( 1)p(a)(p(b)+p(m))(a
b)(m) = ( 1)p(a)p(m)b
(am) = b
(m
a) 

Corollary. If M; N 2 A-mod, we can dene new left modules Hommod-A(M; N ); M

A N .
Therefore, super commutative algebras play the same role in the super world as usual
commuttive algebras do in the usual mathematics.
Finally, a Lie super algebra is a vector space L = L0  L1 with a bracket operation
[ ; ] : L
L ! L with:
[X; Y ] = ( 1)p(X )p(Y )[Y; X ]
[X;[Y; Z ]] + ( 1)p(Z)(p(X )+p(Y ))[Z; [X; Y ]] + ( 1)p(X )(p(Y )+p(Z))[Y; [Z; X ]] = 0
Note that, if X 2 L1, [X; X ] does not necessarily vanish. We have a functor:
Associative algebras ! Lie algebras.
This functor preserves the underlying vector space and
[X; Y ] = XY + ( 1)p(X )p(Y )Y X:
In particular, if X 2 A1 , [X; X ] = 2X 2
2
1.3.
Let V be a super vector space, or more generally a module over a commutative super
aglebra A.
We have a natural action of the symmetric group S n on
V
A V




A V (Remember, (X
Y ) := ( 1)p(X )p(Y )Y
X )
| {z }
n times
Symn(V ) is dened to be (V
n)Sn .
The space Sym(V ) = Symn(V ) has a natural structure of graded commutative super-
algebra.
If V = V0 , Symn (V ) in the super sense identies with the usual Symn(V ), whereas if
V = V1, Symn(V ) in the super sense coincides with n(V ).
We dene the exterior algebra (V ) as Sym((V )). This again is a commutative su-
peralgebra.
1.4.
Let A be a commutative super algebra and let M and N be two free A-modules, i.e.

|  A 
{z

 A  A}  |A  A{z


 A}

M =A
p q
|  A 
{z

 A  A}  |A  A{z


 A}
N =A
p0 q0
and let (e1 ; : : : ; ep ; ep+1; : : : ; ep+q ), (f1 ; : : : ; fp0 ; fp0+1; : : : ; fp0+q0 ) denote their bases.
Then every element in Hom(M; N ) (cf. Corollary 1.2) can be presented in a matrix
form p q
n 0 z}|{ z}|{ 1
p0 B A B C
T= n @ A
q C 0 D
T is even if all the entries of A, D are even and those of B and C are odd.
T is odd if the entries of A, D are odd and those of B and C are even.
When M = N , End(M ) will be identied with the matrix algebra which we denote as
Matp;q (A)
3
Lemma.
A B Consider an even morphism str : Matp;q (A) ! strA given by the formula
str C D = tr A ( 1)p(D)tr D. Then str([X; Y ]) = 0.

The function (str) can also be dened using the tensor structure mentioned in 1.1.
2. Super calculus.
2.1. Super domains.
A super domain of dimension (p; q) is an open subset U  Rp with a super alge-
bra of functions C 1(U )

(1 ; : : : ; q ). For each x 2 U we have the evaluation map
evx: C 1(U )

(1 ; : : : ; q ) ! R dened as
 evx(f
1) = f (X )
evx(1
) = 0 :
1

The triple (U; C (U )
 (1; : : : ; q ); evx ) will be denoted by calligraphic U and the algebra
C 1(U )

(1 ; : : : ; q ) will be denoted by C 1(U).
If U and V are two super domains, a morphism ': U ! V is a pair ('0 ; '), where
'0: U ! V is a map of the underlying topological spaces and ' : C 1(V) ! C 1(U) is a
super algebra map, s.t. 8 x 2 U , 8 f 2 C 1(V), evx(' (f )) = ev'0 (x)(f ).
Remarks. Note that neither a function 2 C 1(U) nor a morphism U ! V are determined
by their values at x 2 U .
Example.
Let U = (R1; C 1(R1)
(1; 2 )). Consider a morphism ': U ! U given by '0 = Id;
'(1 ) = 1, ' (2) = 2; ' (x) = x + 12. This is an example of an automorphism which
is trivial on the underlying topological space.
Lemma.
(a) A map of super domains ': U ! V is completely determined by the image of the
coordinate functions
y1; : : : ; yp0 ; 1; : : : ; q0 2 C 1(V):
(b) For any ye1 ; : : : ; yep0 ; e1 ; : : : ; eq0 2 C 1(U) 9! map ': U ! V with ' (yi ) = yei :
'(j ) = ej provided that 8 x 2 U , (ye1 (x); ye2 (x); : : : ; yep0 (x)) 2 V .

4
2.2. Supermanifolds.
A supermanifold M is a topological space M endowed with a sheaf of super algebras
C 1(M) which is locally isomorphic to a super domain.
Morhisms between two supermanifolds M and N are dened as in the case of super
domains.
Analogous denitions can be also given in the context of real analytic, complex, or
algebraic manifolds.
Again, we emphasize that neither of the geometric constructions on supermanifolds that
will be described is determined by its action on the points of M . Nevertheless, the following
construction is often convenient:
Lemma-Construction. For every supermanifold M of dimension (p; q ) there exists a
canonical sub-supermanifold M of dimension (p; 0) (i.e. a usual manifold) with a property
'
that every map N ! M with N being completely even must factor as
M
% x
'

N
? [
&'
M
Proof. We set M to have the same underlying topological space as M does. C 1(M) is the
quotient of C 1(M) by the ideal generated by all odd functions. 
M is called the underlying manifold of the supermanifold M.
We let S Man denote the category of supermanifolds. As in the case of usual manifolds,
S Man admits the direct product construction: M; N ! N  M.
As in the usual geometry, many important constructions for supermanifolds are car-
ried out in families, where by a family MS of supermanifolds over a base S which is a
supermanifold too, we mean a submersive morphism
MS
??
y (cf. 2.3):
S
We will often consider sheaves of C 1(M)-modules on a supermanifold M. A sheaf is called
locally free, if locally on M , we have F ' C 1(M)
V , where V is a super vector space.
R

5
 If x is a point of a supermanifold M and if F is a sheaf on M, the ber Fx of F at
x is dened as a super vector space F=mxF, where mx is the maximal ideal in C 1(M)
corresponding to x.
2.3. Tangent sheaf.
Consider the sheaf of all (even and odd) derivations of C 1(M). This is a sheaf of locally
free C 1(M)-modules, called the tangent sheaf and we denote it by T(M).
The dual sheaf to T(M ) denoted by T(M) is called the cotangent sheaf. We have a
canonical even morphism D : O(M) ! T(M) dened by D(f )(X ) = ( 1)p(f )p(X )X (f ).
The sheaf
1(M) = T(M) is called the sheaf of 1-forms on M. We have canonical
odd morphism d : O(M) !
1(M).
Lemma-Denition. The sheaf i  0 i(
1 (M) def
' i  0
i (M) has a structure of a commu-
tative super algebra such that:
1) The dierential d:
i(M) !
i+1(M) is odd and satises the Leibnitz rule.
2) For f 2 C 1(M) def
'
0 (M) and X 2 (M; T(M)) we have

df (X ) = ( 1)p(x)X (f ): 

Note that, unlike the usual situation, the De Rham complex

0(M) !
1(M) ! : : :
is innite unless M is even.
Let ': M ! N be a morphism between two supermanifolds. If x 2 M and y = '0(x) 2
N we have a map of superspaces called the dierential of ' at x:

Tx': Tx(M) ! Ty(N):

As in the classical case, we have the \implicit function theorem":

Theorem.
Let ': M ! N be a submersion (i.e. Tx ' is onto 8 x). Then locally on M, it can be
identied with a projection M = N  N0 ! N

6
2.4. Distributions.
A distribution on a manifold M is a subsheaf a ,! T(M) which locally on M can be
realized as a direct summand. Such a leads to considering the Frobenius pairing:
Fr: a
a ! T(M)=a:
It is easy to see that Fr is a map of C 1(M)-modules. We have an analog of the Frobenius
theorem:
Theorem. Fr  0 if and only if the distribution a is integrable, i.e. if locally on M there
exists a submersive map  : M ! N such that a identies with the kernel

0 ! a ! T (M) !   T (N ) ! 0 

2.5.
A convenient way to think about supermanifolds is to view them as contravariant func-
tors on the category S Man. In other words, to a supermanifold M associate a functor
FM: SMan0 ! Sets
given by FM(S ) = Hom(S; M). By the general category theory, such a functor determines
M completely. (Sometimes, the appropriate supermanifold does not exist and such a
functor is a good substitute for it.) Let us consider several examples.
Example 1. R1;0, R0;1.
It is easy to see that

FR 1;0 (S ) ' C 1(S )0

FR 0;1 (S ) ' C 1(S )1:

Example 2.
Let V be a nite dimensional super vector space. Then one can canonically dene a
supermanifold which corresponds to V . This supermanifold corresponds to the functor
FV ( S ) = ( C 1 ( S )
V ) 0 :
7
More generally, let E be a locally free sheaf of C 1(M)-modules on a supermanifold M.
We dene the total space of the corresponding super vector bundle V (E ) on M by means
of the functor
FV (E)(S ) = fpairs : S ! M; g;
where  is a morphism of supermanifolds and  is an even section of  (E ) on S . (Here
 denotes the functor of pull-back. )
Example 3.
We will consider a particular case of the previous example for E = (T(M)). The
e.
supermanifold V (E ) will be also denoted as M
Lemma. FMf(S ) ' Hom(S  R0;1; M) 
e bijectively correspond to morphisms R0;1 ! M.
In particular, the points of M
Supermanifolds obtained in this way will be sometimes denoted as Mor(R0;1; M).

3. Lie supergroups and Lie superalgebras.

3.1. Lie supergroups.
Let G be a supermanifold, which we would like to endow with a structure of a Lie
supergroup on G. It is very convenient to phrase the corresponding denition in terms of
the corresponding functors.
Denition. 1) We say that G is a Lie supergroup if the functor FG : S Man0 ! Sets
possesses a group structure (i.e. if for every B 2 S Man the set FG(B) is a group structure
and for any morphism B1 ! B2 of supermanifolds the corresponding morphism FG(B2 ) !
Fg (B1) is a homomorphism of groups.)
2) We say that a Lie supergroup G acts on a supermanifold M if the functor FG acts
on the functor FM (i.e. if for every supermanifold B we are given an action of the group
FG(B) on the set FM(B) such that the obvious compatibility condition holds.)
Remark. One can give also the following (equivalent) denition of a Lie supergroup, or
of its action on a supermanifold:
8
 We can say that a structure of a Lie supergroup on a supermanifold G is a pair of
morphisms m: G  G ! G (multiplication) and i: G ! G (taking the inverse) such that
the obvious axioms hold. However, we will see below that in most examples the denition
via functors is more convenient.
Example. Let us dene the general linear Lie group GL(p; q), which is a super analog of
GL(n). It should be thought of as the group of linear automorphisms of Rp;q. We dene
a functor FG: S Man0 ! Sets by

FG(B) = Aut(C 1(B)

Rp;q) (cf. 1.4)

This functor is endowed with the obvious group structure and it is not dicult to see
that this functor is represented by a supermanifold.
More generally, let V be any superspace of dimension (p; q). Then one can dene a Lie
supergroup GL(V ) by putting

FG(B) = GLC1 (B)(V

C 1(B))

(here GLC1 (B)(V

C 1(B)) means the group of automorphisms of the module V
C 1(B)
over the superalgebra C 1(B)).
3.2. Lie superalgebras.
If G is a supergroup, it possesses a distinguished point e called the unit element. There-
fore we can consider the tangent space TeG to G at the point e. It is easy to see that the
superspace g = TeG has a canonical structure of a Lie superalgebra (as in usual Lie theory,
one can identify g with the space of left invariant vector elds on G and then consider
their commutator).
Let G denote G. Then the Lie algebra Lie G of G is canonically identied with G0 and
we see that the pair (G; g) forms a Harish-Chandra pair. Let us recall this notion.
Denition. A Harish-Chandra pair is a pair (G; g) where G is a Lie group and g is a Lie
superalgebra together with an action of G on g by automorphisms and a Lie-algebra map
i: Lie G ! g0 , such that the standard compatibility condition saying that the action of
9
Lie G on g obtained by deriving the G-action coincides with the adjoint action of Lie G = g0
on g holds.
It is easy to see that the assignment G ! (G; Lie G) = (G; g) is a correctly dened
functor from the category of Lie supergroups to the category of Harish-Chandra pairs.
Proposition. This functor is an equivalence between the category of Lie supergroups and
the category of Harish-Chandra pairs for which the map

i: Lie G ! g0
is an isomorphism.
This proposition allows to reduce any question about Lie supergroups to a question
about Harish-Chandra pairs, which are intuitively easier to handle. For example, let us
explain how to describe an action of a Lie supergroup on a supermanifold in terms of
Harish-Chandra pairs.
Denition. 1) An action of a Lie superalgebra g on a supermanifold M is a morphism
of Lie superalgebras g ! Vect M (here Vect M denotes the algebra of global vector elds
on M).
2) An action of a Harish-Chandra pair (G; g) on M is an action of the Lie group G
on M together with an action of the Lie superalgebra g on M, satisfying the obvious
compatibility condition (i.e. that the two actions of Lie G on M obtained by deriving the
G-action on M and by identifying Lie G with g0  g, coincide).
It is easy to see that to give an action of a Lie supergroup G on M is the same as to
give an action of the corresponding Harish-Chandra pair.
Example. Let G = R0;1. This is a Lie supergroup (with respect to addition) and as
any Lie supergroup it acts on itself by translations. More precisely, for any supermanifold
B the set Hom(B; R0;1) is canonically identied with the vector space C 1(B)1 ,, which
endows FG with a group structure.
For any supermanifold M consider M e = Mor(R0;1; M). M is a supermanifold (cf. 2.5)
which admits a natural action of R0;1 (since it acts on itself). The Lie algebra Lie R0;1
10
 e on
is generated by the standard odd vector eld @@ . The action of its image in Vect M

 (M)  C 1(M e ) coincides with the De Rham dierential.

3.3. Theory of connections.
We want to dene the notion of connection in the superworld. Let us rst explain what
we mean by principal G-bundle over M, where G is a Lie supergroup. A principal G bundle
over M is a supermanifold P together with a free G-action and a natural isomorphism
M ' P=G (the notions of quotient and free actions are again easily formulated in terms of
the corresponding functors).
As in the usual dierential geometry, one can dene a connection on P to be a dis-
tribution a  T(P ) of codimension equal to dim M and transversal to the bers of the
projection P ! M which is, moreover, G-invariant.
A connection on a vector bundle E on M is by denition a rst-order dierential operator
r: E

 (M) ! E

 (M)
which is odd, satises the Leibnitz rule and has degree 1 with respect to the natural
Z-grading on
 (M).
A connection on a principle G-bundle P on M can be dened as a compatible (w.r. to
tensor products) family of connections on every vector bundle associated with a represen-
tation : g ! GL(V ) of G.
4. Integration and determinants.
4.0. So far we have discussed only dierential geometry in the supercase. Now we want to
treat integration theory. It is easy to see that in supercase one cannot integrate dierential
forms over a supermanifold (since there is no natural notion of highest-degree forms).
Instead, we can integrate sections of some vector bundle over M. In order to dene it we
rst have to discuss the notion of superdeterminant (or Berezinian). So, we are going back
to linear algebra.
4.1. Let C be a commutative superalgebra and let M be a free C -module of rank (p; q).
We want to dene an analog of the notion of highest exterior power of M. We have the
following denition:
11
 Let S (M) denote the symmetric algebra of M.
Lemma.
(0 if i 6= p
ExtiS(M)(C; S (M)) =
free C -module of rank 1 and parity q if i = p
We dene BerC (M) := ExtpS(M)(C; S (M)) .
Properties of Ber
1) If M is of rank (p; 0) then
Ber M = pC M
2) Suppose that 0 ! M1 ! M ! M2 ! 0 is a short exact sequence of C -modules.
Then
Ber M = Ber M1
Ber M2:
3) Suppose that M is (0; q)-dimensional. Then
Ber M = S q (M) 1 :

Let now g 2 AutC (M) | an automorphism of M. Then one can dene Ber(g) to be

the invertible element of C0 by which g acts on Ber(M). We obtain a homomorphism
AutC (M) ! C0.
Berezin actually wrote a formula for Ber(g) (this was his original denition of Berezinian).
 A B  let us choose a basis of M over C . Suppose that in this basis g looks like
Namely,
g = C D . Then
Ber g = det(A B D 1 C )(det D) 1
Looking at this formula, it is not easy (although possible) to prove that it is multiplicative.
Let us emphasize that, unlike the usual situation, Berezinian is dened only on even
invertible matrices.
There is yet another deniton of the Berezinian: by proposition 3.2 there exists a
canonical morphism of Lie supergroups
Ber: GL(p; q) ! GL(1; 0)
12
given by the usual determinant GL(n) ! GL(1) and by the supertrace
str : Lie(GL(p; q)) ' Mat(p; q; R) ! R ' Lie(GL(1; 0))
and we can dene Ber on even invertible B-valued matrices as the corresponding homo-
morphism
FGL(p;q)(B) ! FGL(1;0)(B):
Let us now list some further properties of Ber(M).
4) Ber((M)) = Ber(M) 1
5) Ber(M ) = Ber(M) 1
6) (very important in integration theory).
Fix a basis e1 : : : ep , ep+1 : : : ep+q of M. Then it denes a canonical element
e1 e2 : : : ep+q ) 2 Ber(M). Moreover, if e11 : : : e1p+q is another basis of M and g is the
transition matrix from one basis to the other, then
(e11 : : : e1p+q ) = Ber(g)
(e1 : : : ep+q ).
In fact, this property may serve as a denition of Ber(M).
Let now F be a locally free sheaf of C 1(M)-modules on a supermanifold M. We can
associate to it an invertible sheaf Ber(F) on M.
4.2 Integration.
Let M be a supermanifold. We dene an invertible sheaf Vol(M) on it as Ber((T(M))).
When M is even, T(M) is even and Vol(M) '
top(M) coincides with the sheaf of usual
volume forms. Let us also x an orientation on the underlying manifold.
R
Theorem. We have a unique map c(M; Vol(M)) ! R of parity p satisfying the following
properties:
R
1) If M is even is the usual integration of a volume form ! ! !
R
M
2) If M ' M1  M2; !i 2 c(Mi ; Vol(Mi )), we have
Z Z Z
!= !1  !2
M M1 M2
3) If M ' R0;1 and !0 2 Vol(R0;1) is given by 
d,
R ! = 1.
0
;
R0 1

13
Lemma. For every X 2 (M; T(M)); ! 2 c(M; Vol(M)) we have R (X!) = 0. 
M
As in the classical case, everything said above can be put also in the relative framework.
Namely, consider a submersive morphism  : M ! N and let ! be a volume form on M
whose support is proper w.r. to . We can then \integrate it along the bers", i.e. 9 a
well dened volume form ! ! 2 (N; Vol(N)) and if ! 2 c(M; Vol(M)),
Z Z
! = ! !:
M N

In particular, in order to compute

R ! for a supermanifold M, one must choose a
M
?
M
y R R
projection p: which reduces ! to the usual integration p! (!).
M M M
Note that the integration ! ! p!(!) is \non-analytic".
4.3. Integral forms.
The following formalism will be useful for us in the sequel. We introduce sheaves i
P
on a supermanifold M as
P i = Hom(
i(M); Vol(M)). We have an action of vector
P P P
elds on M on i 's by Lie derivatives. The direct sum (M) :=  i is a Z-graded
i
module over the sheaf of superalgebras
(M) := 
i (M). It is easy to check that 8 X 2
i P
(M; F(M)) extends uniquely to a derivation of degree 1 on i i compatible with the
module structure.
Lemma. 9! odd dierential of degree 1 d : P ! P which makes it a DG-module over

(M) such that 8 X 2 (M; T(M)) we have [iX ; d] = LieX . 

5. SUSY Manifolds.
5.1.
Denition. A pre-SUSY manifold is a supermanifold M of dimension (p; q) equipped with
a distribution a ,! T(M) of dimension (0; q). It is called a SUSY manifold if the Frobenius
pairing
Fr: a
a ! T(M)=a
14
is suciently non-degenerate.

Let us consider rst the case of SUSY curves, i.e. (p; q) = (1; 1). In this case the non-
degeneracy condition reads as: Fr: a
a ! T(M)=a is an isomorphism. In what follows,
our assertions apply both to the case of usual supermanifolds and complex analytic ones.

Denition. A Spin curve (complex or real) is a curve C with a line bundle S and an
isomorphism S
S = T(C ).

Proposition. The category of SUSY curves is equivalent to the category of Spin curves.
Proof. Let (C; a) be a SUSY curve. Consider the restriction ajC . By denition,

ajC
ajC ' T(C )jC=ajC '  (C ):

Therefore, (C; ajC ) is a Spin curve and we have constructed a functor SUSY curves !
Spin curves.
Since locally on a curve all Spin structures are isomorphic, it is sucient to prove the
local faithfulness of this functor.
So, let C = R1;1 be with coordinates (x; ) and let a be given by C 1(C )
(@ + x@x ).
It is easy to see then that there are no non-trivial automorphisms of this SUSY structure
that induce the trivial automorphism of (C; ajC ). 

Let S Mg be the moduli space of Spin curves of genus g. The proposition proved above
implies that it admits a natural super extension, namely the supermanifold of moduli of
SUSY curves of genus g.
S Mg ,! SUSY Mg :

Moreover, S Mg identies with SUSY Mg .

5.2. In the sequel we will consider the example of a SUSY (2; 2)-manifold.
Let V be a SUSY (2; 2)-manifold and let us assume, moreover, that V is locally a product
15
of two SUSY curves. This condition is equivalent to the existence of a decomposition:

a = a1  a2
T (V)=a = (T (V)=a)1  (T (V)=a)2
Fr(a1
a1) ' (T(V)=a)1
Fr(a2
a2) ' (T(V)=a)2
Fr(a1
a2) = Fr(a2
a1) = 0:

Such a decomposition is canonical if it exists.

Lemma. In the above situation, Vol(V) canonically identies with a1
a2. 

16