Kac - 1977 - Lie Superalgebras

ADVANCES IN MATHEMATICS 26, 8-96 (1977)
Lie Superalgebras
V. G. KAC
Massachusetts Institute of Technology,
Cambridge, Massachusetts 02139
DEDICATED TO THE MEMORY OF MY FATHER, G. M. KAC

(SEPTEMBER 15, 19 1 S-SEPTEMBER 25, 1974)
Contents. Introduction. 1. Basic definitions and preliminary remarks:

1.1. Superalgebras and Lie superalgebras-Supertrace. 1.2. Z-graded Lie
superalgebras. 1.3. Lie superalgebras with filtrations. 1.4. Information from
the theory of representations of semisimple Lie algebras. 2. Classical Lie
superalgebras: 2.1. Examples of classical Lie superalgebras. 2.2. Splitting of the
classification of classical Lie superalgebras into two cases. 2.3. Classification of
Lie superalgebras with nondegenerate Killing form. 2.4. Completion of the
classification of the classical Lie superalgebras. 2.5. Contragredient Lie super-
algebras. 3. Cartan Lie superalgebras: 3.1. The Lie superalgebras W(n).
3.2. Two algebras of differential forms 3.3. Special and Hamiltonian Lie super-
algebras. 4. The classification theorem: 4.1. Classification of certain Z-graded
Lie superalgebras. 4.2. The classification of the simple Lie superalgebras.
5. Continuation of the theory: 5.1. Description of semisimple Lie superalgebras
in terms of simple ones. 5.2. Irreducible finite-dimensional representations
of solvable and simple Lie superalgebras. 5.3. Simple Lie superalgebras over
nonclosed fields. 5.4. On the classification of infinite-dimensional primitive Lie
superalgebras. 5.5. Some unsolved problems. References.
“Graded Lie algebras have recently become a topic of interest in physics

in the context of ‘supersymmetries’ relating particles of different statistics”
(see the survey [22], from which this quotation is taken and which contains
an extensive bibliography).
In this paper, we attempt to construct a theory of Lie superalgebras or,
as the physicists call them, Za-graded Lie algebras. We prefer the term “super-
algebra,” which is also inspired by physicists, because speaking generally,
Lie superalgebras are not Lie algebras.
A superalgebra is a Z,-graded algebra A = Aa @ Ai (that is, if a E A, ,
bE-%, 01,/? E Z, = (0, I}, then ab E Aa+J. A Lie superalgebra is a superalgebra
G = G, @ Gi with an operation [ , ] satisfying the following axioms:
8
Copyright Q 1977 by Academic Press, Inc.
All rights of reproduction in any form reserved. ISSN 0001-8708
LIE SUPERALGEBRAS 9
[u, b] = -(-1)“8[6, a] for UE G., by Gs,

[a, [h 41 = EC@,
4, cl + (-1mh [a, cl] for aeG,, bEG,.
We mention that the Whitehead operation in homotopy groups satisfies

these axioms; Lie superalgebras also occur in several cohomology theories,
for example, in deformation theory (see [22, 241).
Lie superalgebras appear in [4] as Lie algebras of certain generalized groups,
nowadays called Lie supergroups, whose function algebras are algebras with
commuting and anticommuting variables. Recently, a satisfactory theory,
similar to Lie’s theory, has been developed on the connection between Lie
supergroups and Lie superalgebras [5].
We now give a brief account of the main features of the theory of finite-
dimensional Lie superalgebras. Let G be a finite-dimensional Lie superalgebra.
Then G contains a unique maximal solvable ideal R (the solvable radical).
The Lie superalgebra G/R is semisimple (that is, has no solvable ideals).
Therefore, the theory of finite-dimensional Lie super-algebras is reduced in a
certain sense to the theories of semisimple and of solvable Lie superalgebras.
(But note that Levi’s theorem on G being a semidirect sum of R and G/R
is not true, in general, for Lie superalgebras.)
The main fact in the theory of solvable Lie algebras is Lie’s theorem, which
asserts that every finite-dimensional irreducible representation of a solvable
Lie algebra over C is one-dimensional. For Lie superalgebras this is not true,
in general. In the paper we obtain a classification of finite-dimensional irreducible
representations of solvable Lie superalgebras (Section 5.2.2, Theorem 7).
In particular, we derive a necessary and sufficient condition for any finite-
dimensional irreducible representation to be one-dimensional (Section 5.2.2,
Proposition 5.2.4).
Next, it is well known that a semisimple Lie algebra is a direct sum of simple
ones. This is by no means true for Lie superalgebras. However, there is a
construction that allows us to describe finite-dimensional semisimple Lie
superalgebraa in terms of simple ones (Section 5.1.3, Theorem 6). It is similar
to the construction in [21].
So we come to the fundamental problem of classifying the finite-dimensional
simple Lie superalgebras. A solution of this problem in the case of an alge-
braically closed field of characteristic 0 is the main aim of the paper and occupies
the major part of it (Chapters 2-4). The principal difficulty lies in the fact
that the Killing form (see the definition in Section 2.3.1) may be degenerate,
which cannot happen in the case of simple Lie algebras. Therefore, the classical
technique Killing-Cartan is not applicable here, The classification is divided
into two main parts (presented in Chapters 2 and 4, respectively).
In the first part we give a classification of the classical Lie superalgebras.
A Lie superalgebra G = G @ Gi is called cZu.rsical if it is simple and the
representation of the Lie algebra Gr, on Gi is completely reducible. This &IS-
10 V. G. KAC
sification is divided into two parts, corresponding to the cases of a nondegenerate

and a zero Killing form. In Section 2.3 we give a classification of all finite
dimensional Lie superalgebras with a nondegenerate Killing form (Theorem 1).
This is the first key point of the classification. Here the usual technique is
applicable. In Section 2.4 we consider the second key point: the case of a zero
Killing form (proposition 2.4.1). The fact that the Killing form is zero is used
to obtain severe restrictions on the index of the representations of G, on Gi
(for the definition of the index see Section 1.4.3). Each of the parts corresponding
to Sections 2.3 and 2.4 is, in its turn, divided into two parts according to whether
or not the representation of G6 on Gi is irreducible (see Section 2.2). The
resulting classification of the classical Lie superalgebras that are not Lie algebras
is as follows (Theorem 2): (a) four series A(m, n), B(m, n), C(n), and D(m, n),
in many respects similar to-the Cartan series A,, , B, , C, , and D, ; (b) two
exceptional Lie superalgebras: a 40-dimensional F(4) and a 31-dimensional
G(3), and a family of 17 dimensional exceptional Lie superalgebras D(2, 1; a),
which are deformations of D(2, 1); (c) two “strange” series P(n) and Q(n). The
construction of all these classical Lie superalgebras is carried out in Section 2.1.
In the second part we give a classification of the nonclassical simple Lie
superalgebras. For this purpose we construct a filtration G = L-i 3 L, 1 L, 3 ...,
where L, is a maximal subalgebra containing GG , and Li = {a E LipI ( [a, L] C
LieI} for i > 0. Then we classify Z-graded Lie superalgebras with the
properties that the associated graded Lie superalgebra Gr G = oi4-i Gri G
necessarily has (Section 4.1.1, Theorem 4). This is the third key point. In
the proof we make essential use of the method developed in our paper [I l]
for the classification of infinite-dimensional Lie algebras. After this it only
remains to reconstruct the Lie superalgebra G with filtration from the Z-graded
Lie superalgebra Gr G.
The final classification of simple finite-dimensional Lie superalgebras is as
follows (Section 4.2.1, Theorem 5): (a) the classical Lie superalgebras (listed
above); (b) the Lie superalgebras of Cartan type W(n), S(n), H(n), s(n), where
the first three series are analogous to the corresponding series of simple infinite-
dimensional Lie algebras of Cartan type and s(n) is a deformation of S(n).
The construction of the Lie superalgebras of Cartan type is carried out in
Sections 3.1 and 3.3.
The finite-dimensional irreducible representations of the simple Lie algebras
are described by the theorem on the highest weight. A similar result holds
for simple Lie superalgebras (Section 5.2.3, Theorem 8). Full reducibility
of finite-dimensional representations is lacking, in general.
It is not hard to reduce the classification of simple Lie superalgebras over
nonclosed fields for the classical Lie superalgebras to the same problem for
simple Lie algebras and for Lie superalgebras of the Cartan type a complete
list can be made. This is done in Section 5.3, where we also list all finite-
dimensional simple real Lie superalgebras (Theorem 9).
LIE SUPERALGEBBAS 11
Finally, in Section 5.4 & attempt to extend Cartan’s results on the clas-
sification of complete i&rite-dimensional primitive Lie algebras to Lie super-
algebras. In this direction we have only obtained a partial result (Theorem 10).
This also makes clear the reason for the appearance of finite-dimensional
Lie superalgebras of Cartan type: Lie superalgebras of Cartan type are Lie
superalgebras of vector fields in commuting and anticommuting variables,
and also their subalgebras defined by the action on the volume, Hamiltonian,
and contact forms. If there are no commuting variables, then the superalgebra
is finite-dimensional, and so there is no finite-dimensional analog for the contact
Lie algebra.
Here is a brief account of the contents of the paper.
Chapter 1 is introductory. In.it we give the basic definitions (Section l.l),
establish the simplest properties of gradings and filtrations (Sections 1.2 and 1.3),
and quote the necessary information on finite-dimensional representations
of semisimple Lie algebras (Section 1.4).
Chapter 2 is devoted to a description (Section 2.1) and classification (Sections
2.2-2.4) of Lie superalgebras with a nondegenerate Killing form (Theorem 1)
and of the classical Lie superalgebras (Theorem 2). In Section 2.5 we describe
the root systems of the classical Lie super-algebras and find all up to equivalence
systems of simple roots. We classify the simple finite-dimensional contragredient
Lie superalgebras (Theorem 3); their properties are very close to those of
simple Lie algebras.
In Chapter 3 we introduce and study two algebras of differential forms
(Section 3.2) with anticommuting and commuting differentials; it is curious
that the second algebra has all the properties that one would naturally expect
of an algebra of differential forms. In Sections 3.1 and 3.3 we construct the
finite-dimensional Lie superalgebras of Cartan type and study their properties.
In Chapter 4 we classify Z-graded Lie super-algebras that arise in the con-
struction of filtrations in simple Lie superalgebras for which the representation
of Gd on C; is reducible (Section 4.1, Theorem 4), and then, on the basis of
this classification, we complete the classification of simple Lie superalgebras
(Section 4.2, Theorem 5).
Theorems 1,2,4, and 5, and also partially Theorems 6 and 7, were announced
by the author in the note [16] (Theorem 4 even earlier in [13]).
In Chapter 5 we discuss the following problems. In Section 5.1 we give a
description of the finite-dimensional semisimple Lie superalgebras in terms
of the simple ones (Theorem 6) and we find the Lie superalgebras of derivations
of all simple Lie superalgebras. As in [21], Theorem 6 is a consequence of a
general result on differentially simple superalgebras (Proposition 5.1 .I).
Section 5.2 is concerned with the theory of finite-dimensional irreducible
representations of solvable and simple Lie superalgebras (Theorem 7 and 8).
In Section 5.3 we treat the classification of simple finite-dimensional Lie super-
algebras over nonclosed fields (Propositions 5.3.1-5.3.3). We also give a
12 V. G. KAC
classification of the simple real Lie superalgebras (Theorem 9). In Section 5.4
we introduce infinite-dimensional Lie superalgebras of Cartan type and
formulate the theorem on Z-graded Lie superalgebras that arise in the clas-
sification of infinite-dimensional complete primitive Lie superalgebras
(Theorem 10). Finally, in Section 5.5 we discuss some unsolved problems.
All spaces and algebras are regarded over a ground field k, which is assumed
to be algebraically closed and of characteristic 0 unless the contrary is stated.
The symbol (M) denotes the linear span over K of a subset M of a linear space,
the symbol @ the direct sum of K-spaces, and @ the tensor product of k-spaces.
Here, I would also like to express my deep indebtedness to F. A. Berezin,
E. B. Vinberg, and D. A. Leites for numerous conversations and constructive
help. I also thank Professor I. Kaplansky for his interest in my work; having
become acquainted with his preprint on root systems of simple Lie super-
algebras with a nondegenerate invariant form I could remove some errors that
had slipped into the original version of the article.
Remark. The history of this article began in 1969 when, impressed by

Stavraky’s example of a simple Lie superalgebra A(l, 0) [19], the author was
led to employ the technique of [l 11 to prove the present Theorem 4. Two
years later, having read [4], I decided to publish this result [13]. The classification
of classical Lie superalgebras (Theorem 2) was obtained in 1974 under the
stimulation of the physicists’ interest in the subject. At the beginning of 1975
the key to the complete solution of the classification problem of simple Lie
superalgebras was found (filtration!) and Theorem 5 was proved. The results
were announced in [16]. Then the results of the Chapter 5 were obtained
and by September 1975 the work was completed. In October 1975 the manu-
script was submitted to the Soviet journal Uspehi Matematickskih Nauk, but
later was withdrawn and resubmitted to the present journal. In the beginning
of 1976 the paper was translated into English. I am grateful to Professor
Sternberg for his genuine interest in my work and for making the translation
of it possible. I am obliged to Professor Hirsch who translated the text.
In the English version some remarks on further results in the field have been
added.
1. BASIC DEFINITIONS AND PRELIMINARY REMARKS
1 .l . Superalgebras and Lie Superalgebras- Supertrace
1. I .l. Superalgebras. We recall that if A is an algebra and M an Abelian

group, then an M-grading of A is a decomposition of A into a direct sum of
subspaces A = eaeM A, for which A,A, C Aa+4. An algebra A equipped
LIE SUPERALGEBRAS 13
with an M-grading is called M-graded. If a E A,, then we say that a is

rhomogmeous of degree(Yand we write deg a = 0~.A subspaceB of an M-graded
algebra A is called M-graded if B = CJ&,,,,(B n A,). A s&a&bra (or Seal)
of an M-graded algebra is an M-graded subalgebra(or ideal). A homomorphism
@: A + A’ of M-graded algebras preserves the grading in the sense that
@(A,) C A;,,, 9 where C+Jis an automorphism of M.
Now let Zs = Z/22 be the residue classring mod 2, with the elements0 and i.
A superalgebrais a Zs-graded algebra A = A0 @ Ai. The elements of A6
are called men, those of Ai odd. Throughout what follows, if deg a occurs
in an expression,then it is assumedthat a is homogeneous,and that the expres-
sion extends to the other elementsby Iinearity.
The direct and semidirect sum of superalgebrasare defined in the usual
way. With the definition of the tensor product things are different. Let A and B
be superalgebras.Their tensorproduct A @ B is the superalgebrawhose space
is the tensor product of the spacesof A and B, with the induced Zs-grading
and the operation defined by
(a1 @ b,)(a, @ b,) = (- l)(degaB)(degbl)

ala2 @ b,b, , ajEA, b,EB.
There is a natural way of defining a bracket [ , ] in a superalgebraA, i.e.,

by the equality,
[a, b] = ab - (- I)(dega)(degb)ba. (1.1.1)
A superalgebra is called commutatiweif [a, b] = 0 for all a, b E A. Quite

generally, permutability in a superalgebrais understood in the senseof the
bracket (1.I .l). Associativity of superalgebrasis defined as for algebras.
For an associativesuperalgebraA we have the following important identity:
[a, bc] = [a, b]c + (-l)(deg*)(degW[a, c]. (1.1.2)
EXAMPLE 1. Let M be an Abelian group and V = eapM V, an M-graded

space. Then the associativealgebra End V is equipped with the induced M-
grading End V = GoleMEnd, V, where
End, V = {a E End V I a(V,) C Vs+.}.
In particular, for M = Zs we obtain the associative superalgebraEnd V =

End, V @ Endi V.
EXAMPLE 2. Let n(n) be the Grassmannalgebra in n variables & ,..., 5, .

Then n(n) becomesZs-graded if we set deg & = T, i = l,..., n. The result is
called a Grassmannsuperalgebra.It is commutative and associative.Evidently
A(m) @ A(n) = A(m + n).
14 V. G. KAC
A generalization of this example is the commutative superalgebra fl(m, n) =

4, ,..-, xnz] @ A(n), where the polynomial algebra k[x, ,..., x,] is regarded as a
superalgebrawith trivial Za-grading.
1.1.2. Definition of a Lie superalgebra. A Lie superalgebra is a superalgebra
G = G3 @ Gi with an operation [ , ] satisfying the following axiom:
[a, b] = -(-l)(dega)(dw)[b, a] (anticommutativity),

[a, [b, c]] = [[a, b], c] + (-l)fdesa)(degb)[b, [a, c]] (Jacobi identity).
Observe that Ge is an ordinary Lie algebra, that multiplication on the left by

elementsof Go determinesa structure of a G-,-module on Gi , and that multi-
plication of elements of Gi determines a homomorphism of G-,-modules
q: SaGi --+ Gr, . Thus, every Lie superalgebracan be specifiedby three objects:
the Lie algebra G0 , the Ga-module Gi , and the homomorphismof Go-modules
v: S2Gi + Gfi, with the solecondition
da, b)c + db, c)a + v(c, a)b = 0 for a, b, c E Gi . (1.1.3)
EXAMPLE 1. If A is an associativesuperalgebra, then the bracket (1.1.1)

turns A into a Lie superalgebra. (The Jacobi identity follows from (1.1.2).)
We denote the resulting Lie superalgebraby AL .
EXAMPLE 2. Let G be a Lie superalgebraand A(n) a Grassmann super-

algebra. Then G @ /l(n) is alsoa Lie superalgebra.
The definitions of a solvable and a nilpotent Lie superalgebraare the same
asfor Lie algebras.A Lie superalgebrais called simple(semisimple)if it contains
no nontrivial (no solvable) ideals.
1.1.3. The universal enveloping superalgebra. Let G = C;, @ Gi be a Lie
superalgebra.As usual, a pair (U(G), i), where U(G) is an associativesuper-
algebra and i: G -+ U(G), is a homomorphism of Lie superalgebras,is called
the univemal enveloping superalgebra of G if for any other pair (U’, i’) there
is a unique homomorphismf3: U + U’ for which i’ = 0 o i.
The universal enveloping superalgebraof G = Gr, @ Gi is constructed as
follows [24]. Let T(G) be the tensor superalgebraover the space G with the
induced Za-grading, and R the ideal of T(G) generated by the elementsof the
form:
[a, b] - a @ b + (-l)(desa)(desb)b@ a.
We set U(G) = T(G)/R. The natural map G + U(G) evidently induces a

homomorphismi: G - U(G), , and the pair (U(G), i) is the required enveloping
superalgebra.
In [24] the following theorem is verified.
THE Pomcm&Bmui~Wm THBORBM. Let G = 60 G be (1 Lie

superalgebra, a, ,..., a,,, be a basis of GO, and b1 ,..., b, be a bash of G . Tken
tked?mentsof thefm
form a basis of U(G).

Finally, we define the diagonal homomorphism. As it is easy to see, the map
a ti i(a) @ 1 + (- I)dqal @ i(a), a E G, is a homomorphism of Lie super-
algebras G + (U(G) @ U(G)) L and therefore determines a homomorphism
of associative superalgebras:
d: U(G) -+ U(G) @ U(G),
which is called the diagonal homomorphism.

1.1.4. Derivations and automorpkisms of a superalgebra. A derivation of
degree s, s ~2, , of a superalgebra A is an endomorphism D E End, A with the
ProPe*Y
D(ab) = D(a)b + (- l)“d~“aD(b).
We denote by der, A C End, A the space of all derivations of degree s, and

we set der A = der-, A @ der-, A. The space der A C End A is easily seen
to be closed under the bracket (l.l.l), in other words, it is a subalgebra of
(End A),, ; it is called the SuperaZgebra of derivations of A. Every element of
der A is called a deriuation of A.
EXAMPLE 1. Let G be a Lie superalgebra. It follows from the Jacobi identity

that ad a: b w [a, b] is a derivation of G. These derivations are called inner;
they form an ideal inder G of der G, because [D, ad a] = ad Da for D E der G.
EXAMPLB 2. Let n(n) = lia(n) @n,(n) be a Grassmann superalgebra.

Let us find der&). For this purpose it is convenient to represent n(n) in
the form &)/1, where CT(n) is the free associative superalgebra with the
generators & ,..., &, whose Zs-grading is given by deg & = T, i = l,..., n,
and I is the ideal generated by all the elements &1; + .$,& . Note that if P and Q
are homogeneous elements of J(B), then PQ - (-l)td@P)(eeso)QP~I.
Let D be a derivation of degree s of A(n). Then
from which it follows that I is invariant under D. Since, obviously, there is

one and only one derivation of x(n) with prescribed values D(&) E A(R), we
607/26/1-2
16 V. G. KAC
see that for any P1 ,..., P,, En(n) there is one and only one derivation D E der /l(n)
for which D(&) = Pi E A(n).
In particular, the relations a/a&(&) = & define the derivation a/a& ,
i=l ,..., 71. The derivation DE der d(n) for which D(&) = Pi can now be
written as a linear differential operator:
It is equally easy to find the automorphism group of /l(n) [3]. Observe that
there is a unique homomorphismv: cl(n) -+ K; we agree to write f(0) instead
of P(f), f E -q4. If now @ is an automorphism of n(n), then deg @(ti) = i
and det(a/a&(@(&))(O)) # 0; any map & t-+ @(&) ~fl(rz), i = I,..., n, having
these two properties extends uniquely to an automorphism of A(n).
Any automorphismCDof cl(n) induces an automorphismof der cl(n) according
to the formula
(@D)f = @(WW)), f E 44.

Note that if D is an even derivation of a superalgebraA = A, @ A,, then
exp tD, t E k, is a one-parameter group of automorphisms.In particular, if A
is a Lie superalgebra, then exp(ad a) for a E A, is an automorphism of A;
the group generated by these automorphisms is called the group of inner
automorphisms. The preceding remark leads to the following result.
PROPOSITION 1.1.l. Let 29 and S,, be the connected components of the identity
in the automorphism groups of superalgebras A and A, , and let S? be the subgroup
of ‘3 consisting of the automorphisms that act identically on A, . Then the restriction
induces an epimorphism 29 + Y,, with kernel &‘. In particular, if A is a Lie super-
algebra, then every inner automorphism of A, extends to an inner automorphism of A.
1.lS. The superalgebra l(V) and the supertrace. Let V = VG @ Vi be a

Za-graded space. Then (Section 1.l.l) the algebra End V is endowed with a
Za-grading and so becomesan associatesuperalgebra.Now (End V), is a Lie
superalgebra(Section 1.1.2) which we denote by l(V) or l(m, n), where m =
dim V, , n = dim Vi . In the theory of Lie superalgebrasl(V) = l(V), @ l(V)i
plays the samerole asthe generallinear Lie algebrain the theory of Lie algebras.
If we regard the same decomposition V = V,, @ V, as a Z-grading of V,
then it correspondsto a Z-grading of l(V), which is compatible with the Z,-
grading: l(V) = l-, @ l(V), @ lI .
Let el ,. .. . em , em+l ,..., em+n be a basisof V, formed from basesof VG and Vi .
It is natural to call such a basis homogeneous.In this basis the matrix of an
operator a from l(V) can be written in the form [F E], where a:is an (m x m)-,
6 an (n x n)-, p an (m x n)-, and y an (n x m)-matrix. The matrices of even
elements have the form [g 3, and those of odd ones [t 8. Here Z, consists
of the matrices of the form c 3 and Z-r of the form c 3. Hence it is clear that
the Zrmodules 4 and I-, are contragredient and the Is-module 4 isomorphic
to g&n @g&a *
Now we come to the definition of the supertrace. Forthe matrix a = [; $1E ltrnSn)
thii is the number
str(a) = tr (Y - tr 6.
Observe that the supertrace of the matrix of an operator a E Z(V) does not
depend on the choice of a homogeneous basis. Therefore, we have the right
to speak of the supertrace of a, meaning the supertrace of this operator in any
homogeneous basis.
To state properties of the supertrace (and for other purposes) it is useful
to introduce the following definitions. Let G = GG @ G be a Zs-graded space
and f be a bilinear form on G. Then f is called consistent if f (a, b) = 0 for
a E Go, b E G , and supersymmetric if f (a, b) = (- l)(desa)(d@)f(b, a). If G
is a Lie superalgebra, f is called inwatint if f ([a, b], c) = f (a, [b, c]).
PROPOSITION 1.1.2. (a) The bdinear form (a, b) = str(ab) on Z(V) is con-
sistent,supersymmetric,and inwariant.
(b) str([a, b]) = 0 for any a, b EZ(V).
Proof. The consistency follows from the fact that ab E I( V)i for a EZ(V)‘),-,
bEZ(V)i.
Supersymmetry for a, b EZ(V),- follows from the corresponding property of
the trace, and for a EZ(V), , b E I( V)i from consistency.It remains to consider
the casea, b E Z(V)i. Let a = (B”i) and ZJ= (i 3) be the matricesof a and b in a
homogeneousbasis. Then (a, 6) = tr &3 - tr &, (6, a) = tr y/I - tr &Y, from
which it follows that (a, b) = -(b, a), as required.
(b) is simply another way of writing down supersymmetry.
We still have to verify invariance. By (1.1.2) we have [b, ac] = [b, a]c +
(- l)(deg@(degb)a[b, c]. Therefore, by (b):
0 = str([b, UC])= ([b, a], c) + (-l)(dego)(degb)(a, [b, cl),
as required.
1.1.6. Linear representationsof Lie SuperaZgebras.Let V = V,j @ Vi be a

Zs-graded linear space. A linear representationp of a Lie superalgebra G =
G@@ Gi in V is a homomorphism p: G + Z(V).
18 v. G. KAC
For brevity we often say in this case that V is a G-module, and instead of
p(g)(w) we write g(v), g E G, o E V. Note that, by definition, GJVJ C V,+i ,
6.i EZ2, andkl pA(4 = gd&>> - (- 1)(deggl)(degsz)gs(g~(w)). Note also that
the map ad: G + Z(G) for which (adg)(a) = [g, a] is a linear representation
of G. It is called the adjoint representation.
A submodule of a G-module V is assumed to be &graded; a G-module V
is said to be irreducible if it has no nontrivial submodules. By a homomorphism
of G-modules Qb: V - V’ we mean one that preserves the &-grading in the
sense that @(Vi) = V,&, , where v is a bijection Za ---f Z, .
SCHUR’S LEMMA. Let V = VG@ Vi, & an irreduciblefamily of operators

from Z(V), and C(d) = {a EZ(V) 1[a, m] = 0, m E A’}. Then either C(A’) = (1)
or dim V, = dim Vi and C(4) = (1, A >, whereA is a nondegenerate operator
in V permuting V, and Vi , and A2 = I.
EXAMPLE. We consider the Lie superalgebra N = NC @ NT, where NG =

(e), NT = (ai ,..., a, , b, ,..., 6,) and [ui , bi] = e, i = I ,..., n, the remaining
brackets being zero. We construct a family of representations pa , 01E K*, of N
in A(n) by setting: p,,(a& = a~/@~, p&b& = oltiu, pJe)u = au. Clearly,
pa is a 2’“-dimensional irreducible representation of N.
We now consider the Lie superalgebra N’ = N @ (c), where [N, c] = 0,
[c,cl = e, and the superalgebra A’(n) = n(n) @ K[E], where deg l = i,
6s = 42, 01E k. We define a representation pa’ of N’ in /l’(n) by setting
p,‘(h)(u 0 4 = p,(h)u 0 a, pa’(c)(u @ w) = (1 @ e)(u @ w), 24 @ w E A’(n).
Clearly, pu’ is a 2Q+1-dimensional irreducible representation of N’.
Both N and N’ are nilpotent. They are called Heisenberg superalgebras.
Note that pu and pal’ fall under the two cases of Schur’s lemma.
This example shows that Lie’s theorem need not be true for Lie super-
algebras. However, Engel’s theorem remains valid, and the proof is the same
as for Lie algebras [lo].
ENGEL’S THEOREM. Let G be Q subalgebraof Z(V) and supposethat all the

operatorsof G are nilpotent. Then thereis a wectorw E V, w # 0, that is annihilated
by all the operatorsof G.
Let V = V, @ Vi be a Zs-graded space. By the symmetric (respectively,

exterior) algebra over V we mean the Z-graded superalgebra S(V) = S( V,) @
A(Vi) = @ S’(V) (respectively, A(V) = A(Vo) @ S(Vi) = @ A”(V)).
If V is a module of the Lie superalgebra G, then we have homomorphisms
G --f der S(V) and G -+ der A(V), so S(V) and rl( V) become G-modules.
The submodules Sk(V) (respectively, /lk( V)) are called the symmetric(respec-
tively, exterior) powersof the G-module V.
LIE SUPBRALGEBRAS 19
1.2. Z-GrahamLie Supera&ebnu
1.2.1. Z-gradings. A Z-grading of a superalgebraG is a decompositionof it

into a direct sum of finite-dimensional Zs-graded subspacesG = eiEz Gi
for which G,G, C G,,, . A Z-grading is said to be cottsistentif G6 = 0 Gai ,
G = CDG,,,, -
By definition, if G is a Z-graded Lie superalgebra,then G, is a subalgebra
and [G,, , Gil C Gi ; therefore, the restriction of the adjoint representation
to Gs induces linear representationsof it on the subspacesGi .
A Z-graded Lie superalgebraG = @ioz Gi is called irreducibleif the repre-
sentation of Go on G-, is irreducible.
A Z-graded Lie superalgebraG = Gjsz Gi is called transitive if for a E Gi ,
i > 0, it follows from [a, G-,] = 0 that a = 0, and bitnznsitive if in addition
for a E Gi , i ,( 0, it follows from [a, G1] = 0 that a = 0.
These properties are closely connected with G being simple, as is shown
by the following proposition.
PROPOSITION 1.2.1. If in a simpleZ-graded Lie superakebraG = eieZ Gi

the subspaceG-, @ G, @ GI generatesG, then it is bitransitive.
The proof is the sameas that of [ll, Proposition 11.
1.2.2. Local Lie superalgebras.Let G be a Za-graded space, decomposed

into a direct sum of Za-graded subspaces,G = G-, @ G, @ G1 . Supposethat
whenever 1i + j 1 < 1 a bilinear operation is defined Gl x G, + Gi+&, r) ++
[x, y]), satisfying the axiom of anticommutativity and the Jacobi identity for
Lie superalgebras,provided that all the commutatorsin this identity are defined.
Then G is called a local Lie superalgebra.
To a Z-graded Lie superalgebra G = @ Gi there correspondsa local Lie
superalgebraGml @ Go@ G1 , which we call the ZocuZ part of G.
Homomorphisms, transitivity, bitransitivity, etc., for local Lie superalgebras
are defined as for Z-graded Lie superalgebras.
In this subsectionwe consider only Z-graded Lie superalgebrasG = @ Gt
for which the subspaceGel @ G,, @ Gr generatesG.
A Z-graded Lie superalgebra G = @ Gt with local part G is said to be
maximal (respectively, timal) if for any other Z-graded superalgebraG’ an
isomorphismof the local parts G and &’ extends to an epimorphiamof G onto G’
(respectively, G’ onto G).
PROPOSITION 1.2.2. Let C?= G-, Q G, @ GI be a local Lie supereebra.

Then there is a maximal and a minimal Zgrad2d Lie superakebrawhoselocal
parts are &morph& to G.
20 V. G. KAC
PROPOSITION
1.2.3. (a) A bitransitive Z-graded Lie superalgebra is minimal.
(b) A minimal Z-graded Lie superalgebra with bitransitive local part is
bitransitive.
(c) Two bitransitive Z-graded Lie superalgebras are isomorphic if and only if
their local parts are isomorphic.
These two propositions are proved just as the corresponding assertions for
Lie algebras (see [l 1, Propositions 4 and 51).
1.2.3. Invariant bilinear forms. The following proposition is proved in the

same manner as [ll, Proposition 71.
PROPOSITION 1.2.4. Suppose that on the local part of a Z-graded Lie super-
algebra G = @ Gi a consistent supersymmetric invariant bilinear form ( , ) is
given (see Section 1.1.5) for which (Gi , Gi) = 0 when i + j # 0. If G-, @
G, @ G1 generates G, then the form can be extended uniquely to a consistent
supersymmetric invariant bilinear form with the same property on the whole G.
The following assertion is proved in the same manner as [7, Corollary 2 to

Theorem 41.
PROPOSITION1.2.5. Let G = &-, Gi be a simple finite-dimensional Z-

graded Lie superalgebra, with Gfk = G:, , k >, 0, G--d # 0, G, # 0. On G
there exists a nondegenerate consistent supersymmetric invariant bilinear form
( , ) if and only if th e rep resentations of G, on GPd and (G,)* are equivalent.
Forthisform(Gi,Gi)=Owheni+j#t-d.
Clearly, the kernel of an invariant form is an ideal. Therefore, we have the

following result.
PROPOSITION 1.2.6. If G is a simple Lie superalgebra, then an invariant form

on it is either nondegenerate or identically xero, and any two invariant forms on G
are proportional.
1.2.4. Conditions for simplicity. In this subsection we state some conditions

for the simplicity of Lie superalgebras. The proofs are standard.
PROPOSITION1.2.7. The following conditions are necessary for a Lie super-

algebra G = GG @ Gi to be simple:
(1) The representation of G, on Gi is faithful and [Gi , Gi] = Gti.
If, in addition,
(2) the representation of G, on Gi is irreducible,
then G is simple.
LIE SU-PJXALGEBRAS 21
P~POS~ON 1.2.8. The fo#owitg cond~tiolls are necessary for a Z-graded Lie
superalgebra G = @Q-~ Gt to be simple:
(1) G is transitive and irreducible; [G-, , GJ = G, .
If, in aa%Xon,
(2) the kernel of the G,-module Gl is 0 and Gt = Gli for i > 0,
then G is simple.
1.2.5. Someproperties of Z-graded Lie superalgebras of the form @o-l G<.
The following assertion facilitates the work with Z-graded Lie superalgebras.
PROPOSITION 1.2.9 (cf. [23]). Let G = @o-1 Gi be a tram&e irreducible

Lie superalgebrawith a consistentZ-grad&, and Gl # 0. Then [G,, , Go] _C
K-I , Gl-
Proof. Observe, first of all, that [G-, , [GWI, GJ] = GY1, because[G-, ,
[GWI , Gr]] is a nontrivial G,-submodule of G-r . Let C be the centralixer in
G,, of [G-, , GJ. Since the Lie algebra G, is reductive, it is sufficient to show
that the Lie algebra C of linear transformations of the space[Gml, [Gml, GJ
is Abelian. To do this we have to verify that for x, y EG-r , a, b E C the
expression d = [[[[t, x], y], a], b] is symmetric in a and b. Now d = [[[[[t, x],
[y, all, ~1,al, bl = -C[P, ~1, Lx,all, bl = [iIt, [x9 all, Cn 611= -[P, [Y, bll, I$, all,
which proves the assertion.
PROPOSITION 1.2.10. Let G = @I>-1 Gt be a Z-graded Lie super&ebra

satisf@qgthe c&Sons of Proposition 1.2.9 and supposethat, in addition, the
representationof G, on Gl is irreducible.Let H be a Cartan subakebraof G,, ,
FA theihighestwe&ht vector of the representationof Go m G-, , and EM the loroest
weightfivectorof the representationof Go on Gl .
(a) If the representationsof G, oreGel and Gl are eontragredient,then
(1) M = -A,
(2) [F,,E,1 =h #O, where hEH,
(3) [‘5 , ‘%I = 0,
(4) the Lie superakebraG+ 0 [GBl, GJ @ Gl is simple.
(b) If the r~esentatiom of Go m G-, and Gl are not contragredtit,
then
(1) PA 2EMI =e,#O,whereu =A+MisanonzerooootoftheLie
a&ha [Go, GoI,
(2) F-I ,41 = [Go 9Gl,
(3) [Go , G,,] is simple.
22 V. G. KAC
Proof. Since G-i = ([...[F,, e-J,..., e-,J> and G, = {[...[EM, e,J ,..., eJ>,
where yi ,,.., ylc > 0, we evidently have
I% , Gl = CL**-FA, -&Ml,e,, ,...,e,,I>. (1.2.1)
Since, by transitivity, [G-, , GJ # 0, we obtain from (1.2.1) that [PA , EM] # 0.

But [t, PA, J%II = (A + W(W’A >-&I f or any t E H. Since contragredience
of the representationsof G, on G-i and Gi meansthat n + M = 0, we have
now established(1) and (2) in (a) and (1) in (b).
Let us prove (3) in (a). We consider the graded subalgebraG of G generated
by the subspaceGel @ G,, @ G1. Clearly G is bitransitive. There is an obvious
automorphism 9 of its local part carrying the positive roots of G,, into the
negative ones and interchanging G-, with G1 . Since, according to Proposition
1.2.3, G is minimal, 9 extends to an automorphism of G. Therefore, it follows
from [G-r , G-J = 0 that [Gi , GJ = 0, asrequired.
(4) in (a) follows from Proposition 1.2.8.
Let us now prove (2) and (3) in (b). From (1.2.1) we seethat [GM1, GJ =
<[*.fe,, e,J,..., e/J) C p, where A is the simple component of the semisimple
Lie algebra [G,, , G,,] the root of which is 01.From Proposition 1.2.9 we seethat
[G-i , G-J = Z? = [G,, , Gs], which proves (2) and (3).
PROPOSITION 1.2.11. Let G = G, @ G, @ GI be a transitive Z-graded Lie

superalgebra satisfy’ng the conditions of Proposition 1.2.9. Then either the repre-
sentation of G,, on GI is faithful and irreducible, or dim Gi = 1.
Proof. Let Gi = G,’ @ G; be some nontrivial decomposition of G1 into

a direct sum of G,-submodules.Applying Proposition 1.2.9 to Gml @ G,, @ G,’
we seethat [G, , GO]C [G-i, G,‘]. On the other hand, clearly [[G-i , G,‘], G,] @
[G-x,4’10 G’ is an ideal in G. Hence it follows, in particular, that
[[G, , G,], G;] = 0. So we find that if the Ga-module Gr is reducible, then
[G, , G,] acts trivially on Gi . But in that caseit clearly follows from transitivity
that dim Gi = 1.
It remains to show that the following situation is impossible: H C G, ,
H # [G, , Ga] is a simple subalgebra of G,, whose representation on Gr is
trivial. In that case [x, G-i] C H for every x E Gi , which contradicts Proposi-
tion 1.2.9. This completesthe proof of Proposition 1.2.11.
PROPOSITION 1.2.12. Let G = @i)-1 Gi be a Z-graded irreducible transitive

Lie superalgebra. If the even part of the center C of G,, is nontrivial, then C, = (z),
and [z, g] = sg for g E G, ,
hof. By Schur’s lemma, Cs = (z)), where [s, g] = -g for g E G-r . Now

let g,+r E Gn+r , x E G-r . Then we have by induction:
b, [x9&c+lll = -4% &k+ll+ [x9r4 k%+Jl= w, &+ll,

from which we see that [x, [s, gb+d - (A + l)g,+,] = 0. By the transitivity
of G, what we need now follows.
PROPOSITION 1.2.13. Let G = @I>-1 Gi be a Z-graded irreducible transitive

Lie supera&ebra for which the representation of Go on GI is faithful. Then G is
bitmn.sitive.
Proof. Clearly, Y = {a E G-, 1[a, GJ = 0} is a submodule of the Gs-
module G-r . By the transitivity of G we have [G-, , Gr] # 0; therefore,
V # G-, ; consequently, V = 0.
1.3. Lie Sicpe*algebras

with Filtrations
1.3.1. Filtrations. A sequence of embedded Zs-graded subspacesin a
superalgebraL: L = L-, 1 L, r) Lr 3 *** is called a filtration if
LfLI CL,+, and nLi =0, i,jeZ.
A Lie algebraL with a filtration is called transitive if for any a EL,\L++, , i > 0,
there is an elementb EL for which [a, b] #L, . This condition can alsobe written
asfollows:
Lf = {a E Ltml I [a, L] C Ltml}, i > 0. (1.3.1)
Let L be a Lie superalgebraand L, be a subalgebraof L that contains no

nonzero ideals of the whole algebra L. Then (1.3.1) defines a filtration in L.
The first property of a filtration is easily verified by induction, using the
Jacobi identity, and the secondfollows from the fact that n Lt , clearly, is an
ideal of L and so, by hypothesis, n Lf = 0.
The filtration constructed in this way is called the transitive Jiltration of
the pair (L, Lo).
With a Lie superalgebraL we can associate,in the usual way, the Z-graded
Lie superalgebra
GrL = @ Gr,L, where Gr6 L = LJLI, .

g-1
Owing to the grading of the subspaceL, , the algebra Gr L is equipped with a

natural Zs-grading; however, the Z-grading of Gr L is not, in general, consistent.
A Z-graded superalgebra G = Qf>-r Gf is canonically equipped with a
filtration: Lf = @&;)(G, .
24 V. G. KAC
A super-algebra L with filtration is transitive if and only if GrL is transitive.

If GrL is simple, then so is L.
1.3.2. ConnectionbetweenL and Gr L. If in a Lie superalgebra with filtration

L =L-,~L,~L,~ ... subspaces Gi are given such that L, = G, @ Lsfl and
[Gi , Gj] C G,+j , then we say that L is equipped with a grading that is consistent
with its filtration. In that case, clearly, L N Gr L, provided that L is finite-
dimensional.
PROPOSITION 1.3.1. Let L = L-, r) L, r) L, r) ... be a transitive Jinite-

dimensionalLie superalgebrawith a filtration for which the representationof
Gr,, L on Gr-, L is irreducibleand evenpart of the center of Gr,, L is nontrivial.
ThenL c1 Gr L.
Proof. According to Proposition 1.2.12 there is an element z E Gr, L such
that [z, g] = sg for g E Gr, L. Let 5 be some inverse image of z under the map
L,, -+ Gr, L = L,/L, . As is easy to see, f is diagonalizable in L, so that L, =
G, O-L,, 3 where G, is the, eigenspace of 5 for the eigenvalue s. This gives
us the required grading, consistent with the filtration of L.
1.3.3. Propertiesof Jiltrations. Let L = L,- @Li be a Lie superalgebra and

L, be a maximal proper subalgebra containing L, . Suppose that Lo does not
contain nonzero ideals of L. Let us construct the transitive filtration of the pair
(L, L,) (see1.1.1):
Li = {a ELi-1 1[a, L] c L&l}, i > 0.
Let Gr L = &>-r Gri L be the associated Z-graded Lie superalgebra.
PROPOSITION 1.3.2. Gr L has thefollowing properties:
(a) Gr L is tramitive;
(b) the Z-grading of Gr L is consistentwith the Zz-grading;
(c) Gr L is irreducible;
(d) if the representationof Ls on Li is reducible,then Gr, L # 0.
Proof. (a) follows from the transitivity of L. The fact that L, contains Lc
implies that Gr-, L C (Gr L)i . By the transitivity of Gr L, we obtain (b) by
induction.
Let us now prove (c). Suppose the contrary; then there exists a Za-graded
subspace z of L containing L, but different from L and L, for which [L, , J?]_C z.
Then E = L, @ V, where V C Li and [V, V] CL, because L, > Lo . Therefore,
[z, z] = [L, @ V, L, @ V] = [L, , L,] + [L, , V] + [V, V] CL. But this con-
tradicts the maximality of L, .
Now we prove (d). If Gr, L = 0; then clearly Gr, L = Lb. Consequently,
by (c), the representation of L, on Li is then irreducible.
PR~P~~ITI~N 1.3.3. Lie superalgebra G = Gc @ 0, is solvabk $7 Lie akebra

G6 is solvable.
Proof. Let C=, be solvable, L,, be a maximal proper subalgebra containing Gs ,
and J be a maximal ideal among ideals of G, containing in L, . We have a filtered
Lie superalgebra G/J = z 3 &, 3 El 3 **., which satisfies all the conditions of
Proposition 1.3.2. In particular, Gr L is irreducible. But Gr,L is solvable
since c;i is solvable and so dim Gr, L = 0 or 1. Therefore dim G/J = 1 or 2
and G/J is a solvable Lie superalgebra. By induction, J is a solvable super-
algebra too. So G is a solvable Lie superalgebra.
1.4. Information f&m the Theory of Representations of Semisimple Lie Algebras

1.4.1. The theorem on the highest weight. Let G be a semisimple Lie algebra
and H be a Cartan subalgebra of it. We consider a representation p of G in a
finite-dimensional space V (or, as we usually say, a G-module V). For h E H*
we set V, = {v E V 1h(v) = h(h)v}. If VA # 0, then h is called a weight of p
and a nonxero vector v, in V, is called a weight vector. Let ZP be the set of all
weights; then V = &?ip V,, .
A weight a of the adj&nt representation of G is called a root of G. Then
G = 0, G. , where G, = H and dim G, = 1 for a # 0. A nonxero vector e,
in G, is called a root vector. GolV, # 0 is contained in VA+= if h + a E pD , and
G,V, =0 if A+a#90.
Let (a, b) = tr(ad a)(ad b) be the Killing form on G. Both the Killing form
and its restriction to H are nondegenerate; therefore, it induces on H* a non-
degenerate form. If a # 0, then [e= , e,] = (em, e-,)h, # 0, where the vector
h, E H is determined by the relation a(h) = (ha , h).
Let d’ be the set of *all nonxero roots, d+ be the set of positive roots (in some
fixed lexicographical ordering), and Z = {a, ,..., a,} be the set of simple roots
of G. Then d’ = d+ u --d+, the system 2 forms a basis of the space H*,
and every root a E d+ is of the form a = C kt% , where the ki are nonnegative
integers.
Let H,* be the linear span of d’ over Q. The Killing form is positive definite
on Ho*, and 5$ C H,*. Let a E A’, h E 9, ; then the set of weights of the form
~+saformsaprogression:X-pa,X-(p-l)a ,..., h---,&h+& ,..., h+qa,
where p and q are nonnegative integers and p - q = 2(h, a)/(a, a).
The numbers 2(X, a$)/( ad , ei) are called the numerical marks of the linear
function h E H*.
If h E ZO, then its numerical marks are integers. A function h E Ho* is said
to be dominant if its numerical marks are nonnegative integers.
Now let p be an irreducible finite-dimensional representation of a Lie algebra
G. A highest (respectively, lowest) we&ht of p is a weight rl E g0 (respectively,
A4 E SD) for which h + OL$5$ (respectively, M - 01$64,) for a E A+. The
highest and lowest weights are unique, and dim V, = dim VM = 1. Every
26 V. G. KAC
nonzero vector in V, (respectively, V,) is called a highest (respectively, lowest)

vector of p. The theorem on the highest weight asserts that the function (1
is dominant and that for any dominant linear function A there is a unique
irreducible finite-dimensional representation with the highest weight fl.
A representation p of a Lie algebra G in a space I’ induces a representation
p* of it in the dual space I’*; p and p* are said to be contragredient. X E PO if
and only if --X E 9+ . In particular, if /I is the highest weight of p, then (--A)
is the lowest weight of p*.
If a Lie algebra G has a faithful irreducible finite-dimensional representation p,
then G = G’ @ C, where G’ is a semisimple Lie algebra, C is the center of G,
and p(C) are scalar operators. The restriction of p to G’ is also irreducible and
its highest weight (vector) is defined as the highest weight (vector) of p.
1.4.2. Diagrams of highest and dominant roots. A semisimple Lie algebra can
be represented by a Dynkin diagram. Let .Z = (01~ ,..., a,} by the system of
simple roots; then aij = -2(ori , c#ai , aj) are nonnegative integers. The
Dynkin diagram of G consists of Y circles correspondnig to the simple roots,
and the ith circle is joined to the jth by an airaji segment with arrows pointing
to the ith circle when aij < aii . An irreducible representation of G is represented
by a Dynkin diagram equipped with the nonzero numerical marks 2((1, ~~)/(a~ , aJ
of the highest weight (1 standing against the corresponding circles.
If G is simple, then its adjoint representation is irreducible; its highest weight
0 is the highest root of G. The diagrams of the highest roots of all simple Lie
algebras are given in Table I. Apart from the highest roots, the simple Lie
algebras also have the dominant roots given in Table II.
1.4.3. The index. Let p be a finite-dimensional faithful linear representation

of a semisimple Lie algebra G in a space V. Then the bilinear form (a, b)y =
tr p(a) p(6) on G is nondegenerate. If G is simple, then
(a, 4” = &(a, 4, a, b E G,
TABLE I
A,(n > 2) &o-..~-o-h G, hao
B&z > 3) o-&.-o z. o F4 i--o-o--0

01
C,(n > 1) ~--o-..~--oco & o-o-b-o-o
0
?
D,(n > 4) o-&d-. ET 0-0-0-0-0-01
0
El3 LO-O-O-L- ~,
LIE SUPEBALGEBBAS 27
TABLE II
B,(n > 3) ~--o---.*-O* 0 G 03A
C,(n > 2) Lo---o G=0 F, o-OS-O-h

I
TABLE III
P-w t Diagram dimTi IV
1
o-o-...-O-0 1
A sl,, n-l n
2n
1
-n(n - 1)
o-o-...-O-0 n-2
A%l, n-l -
2 2n
I n(n + 1) n+2
S%l, n-1 o-(J-...-O-0 - -
2 2n
AV, , 5, 6 1 o-o-o-...-0 1 20,35,56 $3 $9 %
n = 6, 7, 8
-n-l 1 1
B so, o-*-.-.-o =7 0 n
2 n-2
vn, , 3,4, S,6 o-o-...--0 =I 6 8,16,32,64 6,+,+,-j+

n = 7,9, 11, 13
n 1 1
C SP* o-o-.*.-o-e 0 n
z n+2
n 1 --n(n - 1) n-2
fvsPn o-o---“‘-0 -s 0 l-
2 2 nt2
4?SP6 3 o-o-=;) 14 Q
n 1
@.-o-.*.-o-o ? 1
D so* n
5
? 1
spin, , 5, 6, 7 o-o--“.-o-o 16, 32, 64
n = 10,12,14
0
E J& 6 &-o-b-o-o 27
0
ET 7 &--o-o-b-o-o 56
F F, 4 LOGO-0 26
G G 2 LO 7
28 V. G. KAC
where 1, is a positive rational number, independent of a and b. It is called the

index of p. The index of a direct sum of representations is the sum of the indices
of the representations.
The index of a one-dimensional representation of a simple Lie algebra G
is 0, and that for irreducible representation is 1 only of the adjoint representation.
A list of the irreducible representations of the simple Lie algebras for which
0 < 1, < 1 is given, to within transition to a contragredient representation, in
Table III, which is taken from [l].
Here sl, , sp, , and so, stand for standard representations of these Lie algebras;
Sk and fl” denote kth symmetrical and exterior degrees, respectively, S,l; and
A,” their highest component, spin, stands for irreducible spinor representation
of sole , G, , F4 , E, , E, denote also the simplest representations of the cor-
responding Lie algebras.
1.4.4. A technical lemma. We now prove a lemma on representations that

is used in an essential way in the classification of the classical Lie superalgebras.
LEMMA 1.4.1. Let p be a faithful irreducible jinite-dimensional representation

of a semisimple Lie algebra G in a space V. Let A be the system of all roots of G,
9 the system of weights of p, and A the highest weight.
(a) If 2.4 E A, then the G-module V is isomorphic to sp, ;

(b) if A - p E A for any p E 9, then G-module V is isomorphic to sl, or sp, ;
to sl(c) if * - P E A f or any t.~E 9, TV # -A, then G-module V is isomorphic

129 SP, > so, 7 spin, , or G, .
Proof. Let M be the lowest weight of p. Observe that 2fl and /l - M can
lie in A only when G is simple. This is, therefore, the case in (a) and (b).
(a) By hypothesis, 2/l E A. However, from Tables I and II it is clear
that only C, has a dominant root with mark 2. Half this root is the highest
weight of the C.-module spar . This proves (a).
(b) By hypothesis, (1 - ME A; this root is dominant, and the sum of
its marks is not less than 2. From Tables I and II it is clear that /1 - M = 8
is the highest root of one of the Lie algebras A, or C, ; therefore, the G-module V
is isomorphic to sl,,, or spar , respectively.
(c) If /l # -M, then by hypothesis fl - ME A, and from the proof
of (b) it is clear that V is isomorphic to sl, . Now let
(1=-M (1.4.1)
If 0 is the highest root of one of the simple components of G, then clearly

il - 0 E 9. If /l - 0 = M, then from (1.4.1) and the proof of (b) it follows
that V is isomorphic to sp, . But if /1 - 0 # M, then there is a simple root OL
LIE SUPEBALGEBRAS 29
for which A - 6 - a E 9. Therefore, A - 0 - a = M, because otherwise,

by hypothesis and (1.4.1), d - (li - B - a) = 0 + a E d, which is impossible.
Hence,
2./l - 0 = a is a simple root. (1.4.2)
From (1.4.2) it follows that if G is not simple, then the number of simple com-
ponents is 2 and the highest root of each simple component is a simple root.
Therefore, G = A1 @ A, , and V is isomorphic to so, = sl, @ sl, .
Now suppose that G is simple. Now (1.4.2) means that (1 = i(e + U) is a
dominant linear function. If the circle corresponding to a is not at an end
of the diagram, then it has at least two negative marks, so that t9 has at least
two positive marks. From Table I we see that in this case G is of type A,,
moreover, that I = 3, a = aI, and V is isomorphic to /PSI., = so, . But if
the circle is at an end of the diagram and Y > 2, then the mark of (Yis negative,
therefore, the positive mark of 0 is not at an end. From Table I we see that
in this caseG is of type B, or D, , a = 011,and V is isomorphic to so, with
n > 6, or G is of type Bs , a = ua, and V is isomorphic*to spin, . Finally,
the caser < 2, asis easyto see,gives the G-modules so, with n = 3,5 and G, .
The proof of the lemma is now complete.
2. CLASSICAL LIE SUPERALGEBRA~
A finite-dimensional Lie superalgebra G = Gs @ Gi is called cksstial if

it is simple and the representation of Gn on Gi is completely reducible.
The aim of this chapter is the description and classification of the classical
Lie superalgebras.
2.1. Examplesof ClassicalLie Superalgebras

2.1.1. The Lie superalgebras
A(m, n). We recall somefacts from Section 1.1.
Let V = Vs @ Vi be a &-graded space, dim V6 = m, dim Vi = n. The
associativealgebra End V becomesan associativesuperalgebraif we let
Endi V = {a EEnd V 1aV, C Vi+s}, i,sE&.
The bracket [a, b] = ab - (-l)(des”)(dea*)ba makesEnd V into a Lie super-

algebra, denoted by l(V) or l(m, a). If we regard the same decomposition
V = V, @ VI as a Z-grading of V, then the sameconstruction gives a con-
sistent Z-grading: l(V) = G-, @ l(V), @ Gr . On l(V) we define the super-
trace, a linear function str: l(V) + K. Its basic property is str([a, b]) = 0,
a, b E l(V).
From this it follows that the subspace
sl(m, n) = {a E l(m, n) / str a = 0}
30 V. G. KAC
is an ideal in Z(m, n) of codimension 1. The resulting Z-grading sZ(m, n) =

G-i @ sZ(m, n)c @ Gl looks in some homogeneous basis of V as follows:
sZ(m, n)a is the set of matrices of the form (i z), where tr 01= tr 6, Gr is the set
of matrices of the form (i {) and G-, of the form (i “,) (where a is an (m x m)-,
6 an (n x n)-, /? an (m x n)-, and y an (n x m)-matrix).
Now sZ(n, n) contains the one-dimensional ideal consisting of the scalar
matrices hl,, . The Lie superalgebra sZ(1, 1) is three-dimensional and nilpotent.
We set
A(m,n) =sZ(m+ l,n+ 1) for m # n, m, 113 0,
A@, 4 = s4n + 1, n + l)l(Ln+2h 71> 0.
The Z-grading of sZ(m + 1, n + 1) induces a Z-grading of A(m, n) of the form
A(m, n) = G-, @ G,, @ GI .
2.1.2. The Lie superalgebras B(m, n), D(m, n) and C(n). Again, let V =
F’a @ Vi be a Zs-graded space, dim VG = m, dim Vi = n. Let F be a non-
degenerate consistent supersymmetric bilinear form on V, so that V, and Vi
are orthogonal and the restriction of F to Vti is a symmetric and to Vi a skew-
symmetric form (in particular, n = 2r is even).
We define in Z(m, n) the subalgebra osp(m, n) = osp(m, n)a @ osp(m, n)i
by setting
osp(m, n), = (a E Z(m, n), 1F(a(x), y) = -(-l)r(degs)F(x, a(y))}, SEZz.
We call osp(m, n) an orthogonal-symplectic superalgebra (for n = 0 or m = 0
it turns into an orthogonal or symplectic Lie algebra, respectively).
Let us find the explicit matrix form of the elements of osp(m, n). We treat
two cases separately.
m = 2Zf 1. In some basis the matrix of the form F can be written as
0 11 0
i Is
11 0 0
0 0 1
0 1,
-1, 0
from which we see that a matrix in osp(m, n) is of the form
a b u x x1 1
C -a= Y Yl
-$ -,$- ;; z z1
,
ylT xlT zlT d e
-yT -xT --xT f -dT
LIE suPERALGEBRAs 31
here h is any (Z x &matrix; b and c are skew-symmetric (Z x I)-matrices;

d is any (r x r)-matrix; u and f are symmetric (r x +atrkes; -n and c) are
(Z x I)-matrices; x and y are (Z x r)-matrices, and z is an (r x l)-matrix.
011
In particular, we see that osp(m, n)a is a Lie algebra of type BI @ C, , and
lz0.,^
the osp(m, n)a-module osp(m, a)~ is isomorphic to SO, @ sp, .
--,
m = 21. In some basis the matrix of F can be written as
[ 01,
1 -1, 0
from which we see that a matrix in osp(m, n) has the same form as in the first
case, with the middle row and column deleted.
In particular, we find that osp(m, n)a for 12 2 is a Lie algebra of type
D1 @ C, , and that the osp(m, n)a-module osp(m, a)~ is isomorphic to SO, @ sp,, .
By analogy with Cartan’s notation we set:
B(m, a) = osp(2m + 1,2n), m > 0, 11> 0;

D(m, n) = osp(2m, Zn), m>2, n>O;
c(n) = osp(2,2n - 2), II 3 2.
We now examine the Lie superalgebra C(n). Subalgebra C(n), consists of

matrices of the form
where a, b, and c are (a - 1 x tt - I)-matrices, b and c being symmetric,

and a E K. Furthermore, C(n) has the consistent Z-grading: C(n) = G-r @
C(n), @ Gr , where G-r and Gr consist of the matrices of the form (respectively):
The representations of C(n), on G-r and Gr are contragradient, and the C(n),
module Gi is isomorphic to cspan-s .
607/26/r-3
32 V. G. KAC
Supplement. We consider another realization of osp(m, n). Let V, be an

m-dimensional space with a nondegenerate symmetric bilinear form ( , )s
and Vi an n-dimensional space with a nondegenerate skew-symmetric bilinear
form ( , )r , n = 2r. Then osp(m, n) can be realized as follows:
osp(m, n)a = A”Vs @ S2Vf, osp(m, n)f = VG @ Vi .
The definitions of the operations are
[a A h cl = (4 c)ob- (k 4@, UAbEAV,, cev();

b ob,cl = (6 C)JJ+ (h 41% uobES2Vl, CE:V~.
These brackets define brackets on f.t”V,, and S2Vl in the usual way:
[ub, cd] = [ub, c]d + crab, d].
Finally, for a @ c, b @ d E V,, @ VI we set
[u @Jc, b @ d] = (u, b&c 0 d + (c, d),u A b.
In this realization there is a natural way of defining an interesting Z-grading

of osp(m, n):
osp(m, n) = G, @ G-, @ G, @ G1 0 G, .
To obtain this, we represent Vi as a direct sum of isotropic subspaces Vi =

VI, @ V; . The following decomposition is then a Z-grading:
osp(m, ?Z) = S2Vi’ @ ( VG@ Vi’) @ (Vi’ @I V! @ A”VJ @ (Vb @ Vf) @ S”Vf .
Clearly, G,, =gZ, @ so, , the representations of G, on Gi and G-, are con-
tragredient, the G,-module G1 is isomorphic to gl, @ so, , and G, to Sag&. .
2.1.3. The Lie superalgebra P(n), n >, 2. This is a subalgebra of sZ(n+ 1, n+ 1),
consisted of the matrices of the form:
where tr a = 0, b is a symmetric matrix, and c is a skew-symmetric matrix.

2.1.4. The Lie superalgebra Q(n), n > 2. First we denote a(n) a subalgebra
of sZ(n + 1, n + I), consisting of the matrices of the form
a b
--- ,
b I a 1
where tr b = 0. Lie superalgebraQ(n) hasone-dimensionalcenter C = (lzn+s).

We put Q(n) = o(n)/C.
2.1.5. Tke Lie superalgebras

F(4), G(3), and D(2, 1; a).
PROPOSITION 2.1.l. (a) There is one aud only one 40-dimmsimal ckzssical
Lie superalgebraF(4) fir whd~ F(4)G is a Lie &ebra of type BsQA, and its
represmtatbn on F(4)i is spin, @ ~1, .
(b) There is one and only one 31-dimensionalclassicalLie superalgebra
G(3) for which G(3)G is a Lie algebra of type G2 @ A1 and its representation
WI G(3)i is Gs @ SZze
(c) There is a one-parameterfamzly of l’l-dimensionalLie superalgebras
D(2, 1; a), a Ek*\{O, -l}, con.&ting of all simpleLie superalgebras
for which
D(2, 1; 0~)~is a Lie algebra of type A1 @ A1 @ A1 and its representationon
D(2, 1; a)1 is sZ, @ sZ2@ S& .
The proof can be obtained by a direct construction of epimorphisms of

G,-modules S”Gi + G, satisfying (1.1.3). However, in Section 2.5 (proposi-
tion 2.5.4) we give an alternative proof, by meansof contragredient Lie super-
algebras(cf. [8, 111).
2.1.6. Proper&v and uniqueness.From the description of the classicalLie

superalgebrasin Sections 2.1.l-2.1.5 above and Propositions 1.2.7 and 1.2.8,
we derive the following result.
PROPOSITION 2.1.2: (a) All the ‘Lie supera&ebrasA(m, n), B(m, n), C(n),
D(m, n), D(2, 1; a), F(4), G(3), P(n), Q(a) me ckzssical.
(b) For the Lie superakebrasB(m, n), D(m, n), D(2, 1; a), F(4), G(3),
and Q(n) the G+aodule Gi is irreducibleand isonwrphix to the modulesin the
followinglist:
(c) The Lie superalgebrasA@, n), C(n), and P(n) admit a unique con-
sistentZ-grading of the fnm G+ @ G, @ GI . Here the Go-modulesGI and Gmz
are itreducible and for A(m, n) and C(ts) contragredient; they are zbmorpkic to
the modulesin tke foL?owi;nglist:
34 v. G. KAC
G Gil Go I G-1 G Go Go I G-1 Go I G,
A@, 4 A, @ A, 0 k ~l,+~ 0 sL+~0 k C(n) Cl 0 k csp,,-, cs~2*n-~

A@, n) &@A,, 4z,lO &,l P(n) A, A%” n+1 Saul,
PROPOSITION
2.1.3. Let G = Go @ Gi be one of the Lie superalgebras
A@, n), B(m, 4, C(n), D(m, n), D(2, 1; 4, F(4), G(3), P(n), or Q(n). Then
the G,-module S2Gi containsGOwith multiplicity 1.
This is not hard to prove, by using the table in [9]. Here we can also exploit
the fact that in the tensor product of two irreducible G6-modules, of which
one has a simple spectrum, the multiplicity of any simple submodule is at
most 1.
PROPOSITION 2.1.4. Let G = G6 @ Gi be a simple Lie superalgebrafor

which the representationof Gfi on Gi is the sameasfor oneof theLie superalgebras
A@, 4, B(m, 4, C(n), D(m, 4 where (m, 4 f (2, 11, F(4), G(3), P(n), 0~
Q(n). Then G is isomorphicto this algebra.
Proof. Let CD:S2Gi -+ G6 be the homomorphism of Ggmodules defined

by the bracket on Gi , and let @’ be the same map for the corresponding super-
algebra, as listed in the proposition. On account of simplicity, CD and @’ are
epimorphisms. By Proposition 2.1.3, CD and @’ are projections of SaGi onto
the same subspace. Since the homomorphism @, to within a constant factor,
determines the superalgebra uniquely, we have to show that @ and W are
proportional projections onto G6 C S2Gi . If G, is simple, this is clear; therefore,
we assume that Gii is not simple. The projections #J and 4’ can be decomposed:
CD = @s + @r + ..., 0’ = CD,, + @r’ + ..., where CD,,and at, are projections
onto the center, and ai and ai’, i > 0, are projections onto the simple com-
ponents; Qi , ai # 0; by Schur’s lemma, ai = ciQi’, cI E k*.
Now we observe that the kernel of each bilinear map Qi (similarly, Qi’):
S2Gi + G, is trivial. For ker Gi C Gi is a G6-submodule; therefore, if G,
is irreducible on Gi , then ker Qi = 0; but if G, is reducible on Gi , then
Gi =: G-, @ G1 is a direct sum of irreducible Ga-modules, and if G-i 2 ker di, ;
then we see again that ker oi = 0 because ai(G, , G,) = 0.
Now suppose that CD and @’ are not proportional. Taking W = @ + c@’
for a suitable c E k, we may assume that @“(Gi , Gi) = HG is a nonzero ideal
in G, that does not contain simple component G:) of G, of maximal dimension.
From what we have said above it follows that H = H, @ Gi is a simple Lie
superalgebra. Evidently, in all cases listed in the proposition the Hc-module Gi
contains more than two irreducible components. This contradicts Proposition
2.2.2 below, and Proposition 2.1.4 is proved.
2.2. SpWing of the Classi$catti of CksssicalLie Stapera&bras into Two Cases
Let G = G-s@ C; be a classical Lie superalgebra. Then Go = G,’ @ C,

where q is a semisimpleLie algebra and C is the center of G. We treat two
casesseparately:
CaseI. The representation of G, on Gi is irreducible. Then G, is a semi-
simple Lie algebra. If this is not so, then there exists a center element z E G
for which [z, g] = 2g for a g E cfi , a contradiction.
CaseII. The representation of GG on Gi is reducible. Then we consider
in G a proper maximal subalgebraL, containing Go. We construct the appro-
priate transitive filtration (see Section 1.3.1) G = L-i 1 L,T>& 3 ..., where
Li = {a ELi-l I [a, L] CL++}, i > 0.
Let Gr L = @i>-l Gr, L, where Gri L = Li/Li+l is an associatedZ-graded

Lie superalgebra. This filtration induces one on GO:G6 = (L, n G,) 3
(L, n G,) 3 -.a. Since G6 is a reductive Lie algebra, we seethat L, n G, = 0.
Since the Z-grading of Gr L is consistentwith the Zs-grading (Proposition 1.3.2),
we have Gr G, = @&, Grst L, from which it follows that Grs L = 0, because
L, n G6 = 0.
Thus, Gr L = Gr-, L @ Gr, L 0 Gr, L. From Proposition 1.3.2 and
Proposition 1.2.11, we now obtain:
LEMMA 2.2.1. Let G = Go @ Gi be a class&alLie superalgebrafor which

the repreqmtationof GGon Gi is reducible. Then G has a filtration G = L-, 3
L, 3 L, with the following properties:
(a) transitivity;
(b) L,=G~IQL,;
(c) the representationsof G8 ok L-,/L, and on L, are irreducible;
(d) either the representationof GGOttL, is faithfur 01 dim L, = 1.
We now derive from Lemma 2.2.1 the following result:
PROPOSITION 2.2.2. Let G = GEQ Gi be a classixalLie superalgebrafor

which the representationof GG011 Gi ts reducible. Then G has a j&at&m G =
L-, 3 L, 3 L, for which Gr L = Gr-, L @ Gr, L @ Gr, L is a simpk Z-graded
Lie supera&ebra;tJu representationsof GrsL on Gr, L and Gr-, L are faithful
a& irreducibk, (Gr L)a = Gr,,L N GG, and the representationof GG cm Gi is
equivalent to that of Gr, L on Gr-, L @ Gr, L.
Proof. If the center of Gr,,L is nontrivial, then L N Gr L, according to
Proposition 1.3.1, and there is nothing to prove. But if it is trivial, then
[Gr-, L, Gr, L] = Gr, L, by Proposition 1.2.9. If, in addition, the representation
36 V. G. KAC
of Gr, L on Gr, L is faithful, then all the conditions of Proposition 1.2.8 are
satisfied, and therefore, GrL is simple.
According to Lemma 2.2.1 it only remains to show that the case dim L, = 1
is impossible. Now G = V @ G6 @L, is a decomposition into the direct
sum of Go-invariant subspaces, [V, L,] = Gn and [V, I’] C G,- , from parity
arguments, and [G6 , L,] = 0. Hence, it follows that V @ Gfi is an ideal in G,
which contradicts simplicity.
Thus, Case II leads us to a classification of the simple Z-graded Lie super-
algebras G = G-r @ G, @ Gr , with G,, = G, , where the representations of G,,
on Gr and G-, are faithful and irreducible.
2.3. Classification of Lie Superalgebras with Nondegenerate Killing Form
2.3.1. Definition and properties of the Killing form. The Killing form on a
Lie superalgebra G is the bilinear form
(a, b) = str((ad a)(ad b)).
From the properties of the supertrace (see Proposition 1.1.2) we obtain cor-
responding properties of the Killing form.
PROPOSITION 2.3.1. The Killing form on a Lie superalgebra G6 @ Gi has the

following properties:
(a, b) = 0 for a EGG, b E Gi (con.s~tency),

(a, b) = (-l)tdega)(dW)(b, a) (supersymmetry),
([a, 4,~) = (a, Lb,cl) (invariance).
From Proposition 2.3.1 we derive, in particular, the next result.
PROPOSITION 2.3.2. If the Killing f orm on G = G,- @ Gi is nondegenerate,

then its restriction to G6 is nondegenerate, and its restriction to Gi gives a non-
degenerate bilinear skew-symmetric form that is invariant under the representation
of Go on Gi .
Just as for Lie algebras (see [lo], for example), we can prove the following
two propositions.
PROPOSITION 2.3.3. A Lie superalgebra with a nondegenerate Killing form

splits into an orthogonal direct sum of Lie superalgebras (with nondegenerate
Killing forms).
PROPOSITION 2.3.4. Every derivation of a Lie superalgebra with nondegenerate

Killing form is inner.
: PROPOSITION 2.3.5. Let G = Q Gi 6e a 7;-graddd Lie snpwalge6ra with

nondegenerate Killing form. Then
(4 (G , G,) = 0 for i # -j,

(b) (u,G,) #Offora~G-i, a #O,
(c) the representations of Go on Gs and G-, are cmtragredient,
(d) thereisaneZementzEGoforwhich[z,g] =sgforgEG,.
proof. If a E’G* ,6 E G, , then (ad a)(ad 6) for i + j # 0 is clearly a nilpotent
operator on G, therefore, (a, 6) = 0, which gives (a). Now (b) follows from
(a) and the fact that the Killing form is nondegenerate; (c) follows from (b).
Let us prove (d). The endomorphism D for which D(g) = sg for g E G,
evidently gives rise to a derivation of degree0. According to Proposition 2.3.4,
this derivation is inner.
Now let G = G, @ G be a Lie superalgebra. On 6 we can define two
bilinear forms:
(a, 6)s = tr(ad a)(ad b)lo,- and (a, b), = tr(ad a)(ad 6)loi. (2.3.1)
By the definition of the Killing form:
(a, 6) = (a, 6). - (a, bh for a, 6 E Gfi . (2.3.2)
From (2.3.1) and (2.3.2) it follows that if Ga = G @ Gi is a direct sum of

Lie algebrasand Gs’ is simple, then
(a, 6) = (1 - Z)(a, 6)s for a, 6 E Gi, (2.3.3)
where 2 is the index of the representation of G-,’ on Gi (seeSection 1.4.3).
PROPOSITION 2.3.6. A -simpleLie superargebraG = G@@ Gi with non-

degenerateKilling form is classical.
Proof. The unipotent radical N of GG is known (see [9], for example) to
lie in a kernel of the form ( , )y of the representation of G, in V. Therefore,
if a EN, 6 E Gs , then by (2.3.2): (a, 6)s = (a, b), = 0. It follows that a lies
in the kernel of the Killing form of G. Hence N = 0 and G is a classicalLie
superalgebra,becauseof Proposition 1.3.2.
By means of the Killing form we can write the Jacobi identity for three
odd elements in a very convenient way. Let G = G @ G be a Lie super-
algebra with nondegenerateKilling form. We choosein Gb any basis ut and
the dual basis o( relative to the restriction of the Killing form to 5. Let
a, 6, c E Gi . Then: [a, 61 = XI, cy.~,. Multiplying both sides scalarly by u, ,
we obtain % = ([a, 61,ut). Making use of the invariance of the Killing form,
we have: CY~ = (a, [6, q]) ‘= -(a, [ui , 61). Thus, [a, 61 = -xi (a, [q , 6])v, .
38 V. G. KAC
Therefore, the Jacobi identity [[a, b], c] + [[b, c], Q] + [[c, a], b] = 0 gives us:
; ((6 bi 34)h 9cl + (b,hi , cm , a] + (c, [up., u])[q , b]) = 0. (*)
2.3.2. Class@ation of the simple Lie superalgebras with nondegenerate Killing

form in Case I. In this section we prove the following proposition:
PROPOSITION 2.3.7. Let G = Go @ Gi be a simple Lie superalgebra with

nondegenerate Killing form for which the representation of G0 on Gf is irreducible
and Gi # 0. Tha G is isomorphic to B(m, n), D(m, n) with m - n # 1, F(4),
or G(3).
Before the proof we give a lemma.
As we have shown in Section 2.2, G; is semisimple.Let H be a Cartan sub-
algebra of G, , d be the system of all roots, and d’ be that of nonzero roots.
Let 9 be the system of weights of the representation of GG on Gi , and
Gi = @ V, the weight decomposition. From Proposition 2.3.2 it follows that
(VA , Vu) = 0 for h # -p; (2.3.4)

if h E 9, then --h E9 and (We, V-,) # 0. (2.3.5)
Let Go = @ G,I”’ be the decomposition of Gr, into a direct sum of simple

components.Evidently, this decompositionis orthogonal relative to the bilinear
forms ( , ) and ( , ),, . We denote by ( , )F’ the restriction of ( , ),, to GF).
Let 1, be the index of the representation of GF’ in Gi . Let h, ,..., h, be a
basis of H formed from basesof the Cartan subalgebrasH n Gt’ of GF’.
Let ft, ,..., A,.be the dual basiswith respectto ( , ) and & ,..., A,.with respect to
( , )0 . From (2.3.3) it clearly follows that
h, = (1 - 1J hi (2.3.6)
In particular,
Killing form is nondegenerateiff 1, # 1 for somes. (2.3.7)
LEMMA 2.3.8. (a) If h E 9, 2X $ A, then
(A, A) = ; py = 0.
8
(b) IfX,pE9and/\-&$A, then
(A, CL) = T pp = 0.
s
LIE SUF’ERALGEBRAS
Proof. We consider the following basis of 3:

{ui> = k , a E A’; h, , i = l,..., Y}.
The dual basis with respect to ( , ) is
b4 = L , arEA’; hi,i = l,..., r}.
Let us prove (a). Let h E .V. We set a = c = wA , b = wpA , where (wA , w-,J = 1
(by (2.3.5) such a vector exists). Now we write the identity (*) for the chosen
bases {Us} and {We} and the vectors a, b, c. Taking (2.3.4) into account, we have
(wA, k4h , fhl)kb , ~1 = 2~ C WJ Wb).

i
Therefore, if 2A $ A, then
c A(&) A@,) = 0. (2.3.8)
t
Since (4 1-4 = Cd Q4 I&> and (A, do = C Wd) &> for any h P E H+,
when (2.3.6) is taken into account, it can be rewritten in the form
as required.
(b) is proved similarly, but in (*) we must put a = w, , b = wmA, c = wU .
This proves Lemma 2.3.8.
Proof of Proposition 2.3.7. If G is simple, and A is the highest weight, then
and by Lemma 2.3.8(a) it follows that 211 E A. From Lemma 1.4.1(a) it therefore
follows that Grmodule Gi is isomorphic to sp, . By Proposition 2.1.4, we now
see that G is isomorphic to B(0, n/2). Since Z, = l/(n + 2) (see Table III),
by virtue of (2.3.7) the Killing form on B(0, n/2) is nondegenerate.
Suppose now that Ga is semisimple, but not simple. We represent G, in
the form GG = G& @ Gf, where Gb* and # consist of all simple components
of G, for which 1 - 1, is positive and negative, respectively. As is clear from
Lemma 2.3.8(a), both these subalgebras are nontrivial. Let A = A1 + IIn be
the highest weight of the representation of GB on G (where I and II indicate
that the weight is restricted to the relevant direct summand). We consider a
weight of the form p = ~1 + A*], where ~1 # -&A’. Observe that, clearly,
A+P~A. (2.3.9)
40 V. G. ILK!
Next, (A, ,u) = ((11, ~1) + (An, An) = (A*, I(L*)+ (A, A) - (AI, AI). Since 24 4 A,
by Lemma 2.3.8(a),
(4 PI = (A’, CL’) - (A’, 4 (2.3.10)
This relation can be rewritten (see the proof of Lemma 2.3.8) in the form
(2.3.11)
where the summand is over the simple components occurring in G&. Since
A is the highest weight, (A, A)a(‘I > (CL,p)p’ for all s. Therefore, all the terms
in (2.3.11) are negative, by the Cauchy-Bunjakowskii inequality. Consequently,
(4 CL)f 0. (2.3.12)
Now we can use Lemma 2.3.8(b), according to which it follows from (2.3.9)
and (2.3.12) that A - p EAI. Thus, if p1 # -AI, then A1 - p1 EAI. Of
course, the same is true for Gf . Therefore, we find from Lemma 1.4.1(c)
that the linear representation of G, on Gi can only be equivalent to the tensor
product of two of the following linear Lie algebras: sp,, , n > 2; sl, , n > 3;
so, , n > 3; spin, ; Ga .
We recall now (Proposition 2.3.2) that the representation of Gr, on Gi admits
a nondegenerate skew-symmetric invariant bilinear form. This can only be
the case when one of the factors of the tensor product has a skew-symmetric
invariant and bilinear form and the other an invariant symmetric form.
Therefore, only the following possibilities remain:
(1) %n 0 SP, 3
(2) SP, 0 spin, ,
(3) SP, 0 6 -
In case (1) we obtain from Proposition 2.1.4 that G is isomorphic to
B(m - l/2, n/2) for odd m > 1, or D(m/2, n/2) for even m > 2. Since
I1 = n/m - 2 and I, = m/n + 2, (Table III), by (2.3.7) the Killing form is
nondegenerate on B(m, n) and also on D(m, n) when m - n # 1.
In cases (2) and (3) we use Lemma 2.3.8(a) again:
(2.3.13)
In cases (2) and (3), (2.3.13) yields that n = 2 and therefore we see from
Proposition 2.1.4 that G is isomorphic to F(4) and G(3), respectively.
This completes the proof of Proposition 2.3.7.
2.3.3. Ckzstiificatim of the simp~% Lie supera&ebraswith nma%generateKill&g

fm in CaseII. In this section we prove the following proposition.
PROPOSITION 2.3.9. Let G = G-, @ GO@ Gl be a simpleLie superaZgebra

with a consistentZ-grading for w&h the representatimsof G,, on Gl and GT1
are faithful and irreducible and the Killing jm k nondegenerate.Then G is
isomorphic(even as a Z-graded supera&ebra)to one of A(m, n), m # n, or C(n).
The proof of this proposition is based on the same arguments as that of
Proposition 2.3.7.
It follows from Proposition 2.3.5 that Gs is the direct sum of the one-dimen-
sional center C and the semisiiple Lie algebra G,,‘. Here C = <z), where
[z, g] = fg for g E G*l , and the representationsof G, on G-r and Gr are
contragredient. Let H be a Cartan subalgebraof Gs’, A its root system, .!Z&
the systemsof weights of the representationsof G,,’ on G*, . From Proposi-
tion 2.3.5 it follows that
9-1 = -21, (2.3.14)
(Gl , G) = (G-1 , G-1) = 0, (2.3.15)
(q, V-J # 0 for h EZr, --h E Z1. (2.3.16)
Let G,’ = @ @se’be the decomposition of G,’ into the direct sum of simple
components. We denote by ( , )i the restriction of ( , )s to Gk’ and by 1,
the index of the representation of Gt) in G1. Note that it is also the index
of the representation of Gt’ in G-r . Just as in Section 2.2, we choosea basis
h1 ,..., h, of H, its dual basis!ar ,..., A, with respect to ( , ) and K1,..., fi, with
respect to ( , )s .
From (2.3.3) it follows that
K, = (1 - 2Z,)K, for hi rz Gt’. (2.3.17)

In particular,
Killing form is nondegenerate,iff Z, # 4 for somes. (2.3.18)
LEMMA 2.3.10. (a) If h E gl , then
(JtA) =p& 2difnG =o.

* 1
(b) If~,pE91andh-P#A, then
(A/L) =p& 2di?G =o.

l 1
42 V. G. KAC
Proof. We consider the following basis of G,,:
+4> = k , 01E A’; hi , i = l,...) r; z}.
The dual basis with respect to ( , ) is
hi> = {e-, , a~d’;h~,i = l,..., Y; -1/(2dimG&}.
Let us prove (a). Let h E 9i . We set a = c = v,_* E G-, , b = ZIPE Gi ,

where (v~ , w-3 = 1. (According to (2.3.16) such vectors exist.) We now write
down the identity (*) in the chosen bases {ui} and {q} and the vectors a, b, c.
Taking (2.3.15) and (2.3.4) into account, we have
(‘,‘) = cI ‘(h)‘(‘i) - 2di; G = O, 1
from which, using (2.3.17), we obtain (a).

(b) is proved similarly, only in (*) we must put a = P/-~ E G-, , b = V~ E Gi ,
c = wM E Gl .
Proof of Proposition 2.3.9. We represent G,,’ in the form G,’ = GA’ @ Gin,
where Gi’ and Ga’ consist of those simple components for which 1 - 21,
is positive and negative, respectively. For definiteness, let Gi’ # 0. Let
(1 = /11 + (1” be the highest weight of the representation of G,’ in Gi . We
consider a weight of the form p = p1 + (1”. Just as in the proof of Proposi-
tion 2.3.1 we find that
From Lemma 2.3.10(b) it now follows that fir - pi E dr. So we see that if
p E 9, then (1’ - p E d’. Therefore, we obtain from Lemma 1.4.1(b) that the
G,,‘-module G1 can only be isomorphic to a linear Lie algebra sl, or sp, or to any
tensor product of them. So we have the following possibilities for the representa-
tion of G,, on G,:
(1) &L 0 4 3
(2) CSPn 3
(3) g&n 0 SP, , m z 2, n a 4,
(4) CsP, 0 sp, , * 3 2, n 3 4.
In case (l), Zi = n/(2m), I, = m/(2n) (Table III); hence, by (2.3.18) m # n.
But then we see from Proposition 2.1.4 that G is isomorphic to A(m - 1, n - 1).
In case (2), we derive from the same Proposition 2.1.4 that G is isomorphic
to C((n/2) + 1). From (2.3.18) it is clear that the Killing form for these Lie
superalgebras is nondegenerate.
LIE SUPERALGEBFlAS 43
That cases (3) and (4) are impossible we deduce from Lemma 2.3.10(a):
(2.3.19)
In case (3), (2.3.19) yields
m-1 1 1
m(2m - 2n) +z(n+2-2m2nr) =2mn’
so that either m = 1 or n = 2.
In case (4), (2.3.19) yields
1 1 1
2(n + 2 - 2m) + 2(m + 2 - 2n) =21tll19
which is impossible.
Thus, cases (3) and (4) cannot occur. This completes the proof of Proposi-
tion 2.3.9.
2.3.4. Conclusion of the t&.@c&ion of simple Lie superalgebras with nos-

degenerate Killing fin-m.
THEOREM 1. A simple $nite-dimensional Lie superakebra G = Ga @ Gi with

nondegenerate Killing fm is isomorphic to one of the simpk Lie a&ebras or to
one of the following classical Lie superakebras:
A(m, n) with m # n, B(m, n), C(n), D(m, n) with m - n # 1, F(4), or G(3).
Proof. From Proposition 2.3.6 it follows that G is classical. In accordance

with Section 2.2 we have to discuss two cases.
Case I. The representation of G6 on Gi is irreducible. Then Theorem 1
follows from Proposition 2.3.7.
Case II. The representation of 6 on q is reducible. In that case, according
to Lemma 2.2.1, G has a filtration G = Lb, 3 L, 1 L, for which Gr G is a
Z-graded Lie superalgebra satisfying all the conditions of Proposition 2.3.9,
(Gr G)6 3~ G5, and the (Gr Gkmodule (Gr G)i is isomorphic to the G,-
module Gi . By Proposition 2.3.9, Gr G is one of A(m, n), m # n, or C(n).
From Proposition 2.1.4 it now follows that G N Gr G, and the theorem is
proved.
2.4. Completion of the Classification of the Classical Lie Supera&ebrus

The classification of the classical Lie superalgebras is given by the following
theorem.
44 V. G. KAC
THEOREM 2. A classical Lie superalgebra is isomorphic either to one of the

simple Lie algebras A, , B, ,..., Es, or to one of A(m, n), B(m, n), C(n), D(m, n),
D(2, 1; 4 E(4), C(3), P(n), or Q(n).
By virtue of Theorem 1 and of Proposition 1.2.6, what remains to be proved
is the following proposition.
PROPOSITION 2.4.1. A classical Lie superalgebra G = G,- @ Gi with zero

Killing form is isomorphic to one of A@, n), D(n + 1, n), P(n), Q(n), or D(2, 1; a).
2.4.1. Beginning of the proof of Proposition 4.1.
LEMMA 2.4.2. Let G = G6 @ Gi be a Lie superalgebra with zero Killing

form for which the representation of G, on Gi is faithful and completely reducible.
Then the index 1; of the representation of any simple component Gij’ of G6 on Gi
is 1. In particular, the index of the representation of Gt’ on any irreducible com-
ponent of Gi does not exceed 1.
Proof. By (2.3.3), (a, b) = (1 - &)(a, b),, for a, b E Gc’. Since (a, b) = 0

and (a, b) 01 ‘s nondegenerate on Gp’,,weseethat&=l.
2.4.2. ClassiJcation of the classical Lie superalgebras with zero Killing form
in Case I.
PROPOSITION 2.4.3. Let G = G, @ Gi be a simple Lie superalgebra with

zero Killing form for which the representation of G,- on Gi is irreducible. Then G is
isomorphic to one of Q(n), D(n + 1, n), or D(2, 1; a).
Proof. As we remarked in Section 2.2, G,j is semisimple. Therefore, the

representation of G, on Gi is equivalent to the tensor product of some simple
irreducible linear Lie algebras:
GG(Gi) = Gc’(&) @ -.. @ G,“)(V,). (2.4.1)
Let !r ,..., L$ be their indices. Then, clearly, the index li of the representation
of GF’ on Gi is equal to
(2.4.2)
according to Lemma 2.4.2.

If Gs is simple, then lI = & = 1, and so (see Section 1.4.3) the representation
of Gr, on Gi is the adjoint one, and an epimorphism of GG-modules, S2Gi - GO
exists only when Gr, is of type A, (see [9], for example). From Proposition 2.1.4,
it follows that this case leads to Q(n).
Now suppose that G, is not simple. Then it follows from (2.4.2) that .& = t;l,
LIE SUF%RALGEBRAs 45
where tr > 2 is an integer. If tf < 4, then &+( dim V, < 4 by (2.4.2); hence,
dim V, < 4 for s # i. From Table III (see Section 1.4.3) it is clear that if
dim V, < 4, then t, < 8, and by (2.4.2), then dim V, < 8. Thus, from tf < 4
it follows that dim Vi < 8. From Table III we can now see that only sl,, ,
spdo, , spin, , and Gs can occur in (2.4.1). We claim that the last two cases
are impossible. If Gp(Vr) = spin, in (2.4.1), then s = 2 and dim V, = 5.
But then Gr)(Va) = sZ6or so6. In the first caseZ, = &, and in the second
Z, = Q. If Gt’(V& = Ge, then GdGi) = Ga@ sZp or Gs @ sls@ sl, or
G8 @Jsp, . In the first case Z, = 8, in the second Z2= 3, and in the third
4 = s . Hence, only sl, , sp, , and so, can occur in (2.4.1):
where 2 < n, < *** < n,, 4 < r, < .-* < Is, 5 < m, < *** < m,.
Relations (2.4.2) can be rewritten in the form
(2.4.3)
(2.4.4)
(2.4.5)
From these relations it is evident that OL< 3, fi < 2, y < 1. If OL= 3, then
it is clear from (2.4.3) that G(Gi) = sl, @ sZz@ sl, , and we have D(2, 1; 8).
If (Y= 2, the only possibility is G(Gi) = sZ,,=@ sZ,,,; but then n, = &,
na = 2nr , which is ‘impossible.If /I = 2, then by (2.4.4) the only possibility
k~GdG) = @r 0 %-+a, which clearly cannot be realized. If a! = /3 = y = 1,
then we have by multiplying (2.4.3), (2.4.4), and (2.4.5), ~Y~Z~Z =
WI + 2)(m, - 21, which is impossible. The casesCY= @= 1, y = 0 and
(Y= y = 1, p = 0, are also impossible. There remains the case GAGi) =
spr @ so, . Then of = Y + 2 (seeTable III).
Thus, the only remaining possibility for the representation of G on Gi is
%l+2 0 sP?i 9 n > 2. This is, in fact, realized for D(n/2 + 1, n/2). By Proposi-
tion 2.1.4, there can only be one superalgebrawith this representation of G6
on Gf . This proves the proposition.
2.4.3. CZa.wi$.cationof the cZassicaZ
Lie superalgebras
wilh zero KiZZiq form
in CaseII.
PROPOSITION 2.4.4. Let G = G-, @ Go (TJGl be a SimpZe Lie superakebra

with a con&tent Z-grading for which the represetttatiomof Go olt Gl and Gel
46 V. G. KAC
are faithful and irreducible, and the Killing form is zero. Then G is isomorphic
(even as a Z-graded superalgebra) to one of A(n, n) or P(n).
Before the proof we give two lemmas.
LEMMA 2.4.5. Under the conditions of Proposition 2.4.4, G, is semisimpb.
Proof. If the center C of G,, is nontrivial, then by Proposition 1.2.12, there

exists a z E C such that [z,g] = &g for g E G;tl . But then, clearly, (2, Z) =
-dim Gel - dim GI. , which is impossible, because the Killing form is zero.
LEMMA 2.4.6. If under the conditions of Proposition 2.4.4 the representations

of G, on GL and G-, are contragradient, then the highest weight of the representation
of G, on G1 must have more than one nonz~o numerical mark.
Proof. Let H be a Cartan subalgebra of GQ , 01~,..., ~1, its system of simple

roots, h, ,..., h, a basis of H, where hi = [eai, e+J, c+(hJ = 2. If E,+ is the
lowest weight vector of the representation of Go on G-, , and FmUOthe highest
weight vector of that of G,, on G1 , [E,, , F-,J = h, E H, h, # 0 (see Proposi-
tion 1.2.10(a)), then
dh,) = 0, (2.4.6)
q(hJ = 0 for q,(h,) = 0, (2.4.7)
det(ari(hj));,+,, = 0. (2.4.8)
(2.4.6) follows from 0 = [EUO, [EaO, F-, J] = -2q,(ho)FD0, (2.4.7) is obtained

by multiplying both sides of [EaO , FM, J = h, by eUi, and (2.4.8) follows from
the linear dependence of the vectors h, , h, ,..., h, .
Suppose now that cx,,(hJ # 0 for one s only. Then a,(h,) # 0; otherwise,
h,, = 0 by (2.4.6) and (2.4.7), and the remaining elements of the first column
are zeros. By hypothesis, &hJ # 0, but the remaining elements of the first
row are zeros. It then follows that det(ori(hj)) = q,(h,) a,(h,) det A, where A is
the Cartan matrix of the Dynkin diagram of GD , with the sth circle omitted.
Since det A # 0, we have reached a contradiction to (2.4.8).
We divide the proof of Proposition 2.4.4 into two cases corresponding to the
following two lemmas.
LEMMA 2.4.7. If under the conditions of Proposition 2.4.4 the representations

of Go on G-, are contragradient, then G is isomorphic to A(n, n).
P~ooj. By Lemma 2.4.5, the representation of G, on G, is equivalent to the

tensor product of some simple irreducible linear Lie algebras:
G,(G,) = Gt’( V,) @ *-. @ Gp’( V,).

LIE SUPJZRALGEBRAS 47
L@t
r,,-e-P
lb be their indices. Since contragredient representationshave equal
indices, the-index of the representation of Gi‘) on G1 is 8 (by Lemma 2.4.2).
Therefore,
Zi n dim V, = 4. (2.4.9)
aft
If Gs is simple, then we seefrom (2.4.9) that Z, = 4 . From Table III it is
clear that there are only the two possibilities for Gs(G,): As& and ~%ps.
Both cannot occur according to Lemma 2.4.6.
Suppose now that G,, is not simple. From Table III it is clear that
2(dim V& > 1 and that equality holds for sl, only. Therefore, we see from
(2.4.9) at once that the only possibility for Gs(G,) is sl,, @ sl, . By Proposi-
tion 2.1.4, G is then isomorphic to A(n - 1, n - l), and the lemma is proved.
LEMMA 2.4.8. Umberthe condalimsof Ropositim 2.4.4, if tke representations

of G,, on Gwl and Gl are not txmtragredient,tken G is &morphic to P(n).
Proof. It follows from Lemma 2.4.5 that G, is semisimple, and from
Proposition 1.2.10 (and Proposition 1.2.1) that Gs is simple. Thus, the relevant
pair of representationsof Gs on GW1and G1 can only be one from Table III,
for which the sum of the indices is 1. This leadsto the following cases:
(1) sZsand A*sZs;
(2) sZs*and A*sZs;
(3) AZ,, and SBsZ,, n > 4;
(4) A%l,* and S%l, .
Proposition 1.2.10(b) imposesyet another restriction: If II is the highest
weight of the representation of G, on G-i and M is the lowest weight of that
on Gi , then n + M is a root of G, . This rules out cases(l)-(3) at once. The
fourth case corresponds(on the basis of Proposition 2.1.4) only to P(n), and
the lemma is proved.
The conclusion of the proof of Proposition 2.4.1 proceedsverbatim on the
samelines asthat of Theorem 1 (seeSection 2.3.4), on the basisof Propositions
2.4.3 and 2.4.4.
This completesthe proof of Theorem 2.
AddSmaZ Remark. The following authors independently obtained clas-
sification results on classicalLie superalgebrasunder the following restrictions:
A. Pais and V. Rittenberg, J. Math. Phys. 16 (1975), 2062-2073, Gs is
simple and the Killing form is nondegenerate.
D. 2. Djokovic, f. Pare AppZ. A&bra 7 (1976), 217-230, Gx is simple.
P. G. 0. Freund and I. Kaplansky, J. Math. Pkys. 17 (1976), 228-231,
G-sis reductive and G admits bilinear invariant form.
607/26/r-4
48 V. G. KAC
W. Nahm, V. Rittenberg, and M. Scheunert, Phys. Lett. B 61 (1976),

383-385, Killing form is nondegenerate (and later all classical ones in Journal
of Math. Phys. 17 (1976), 1626-1640).
2.5. ContragredientLie Superalgebras

2.5.1. Definition of the superalgebras G(A, T). Let A = (ail) be an (r x r)-
matrix with elements from a field k and r be a subset of I = {1,2,..., r>. Let
G-, , G,, , and G1 be vector spaces over k with bases {fi}, {h,j, and {es}, i E I,
respectively. As is easy to see, the following relations determine the structure
of a local Lie superalgebra (?(A, T) on the space G-, @ G, @ G,:
[ei ,fjl = Wi , [hi , hjl = 0,

[hi , ej] = aijej , [hi ,fil = -aiifi ,
deg hi = 0, deg e, = degf, = 0 for i$7,
deg ea = deg fi = i for iE7.
According to Proposition 1.2.2, there exists a minimal Z-graded Lie super-
algebra G(A, T) with local part e(A, T). We call G(A, T) contrugredient Lie
superalgebra,A its Cartan matrix, and r its rank. Note that when 7 = $, we have
contragredient Lie algebras whose theory is developed in [ 111.
When hi is replaced by chi and fi by cfi , c E k*, then the ith row of A is
multiplied by c. Therefore, we may (and will) assume that if ai,: # 0, then
aii = 2. If we can obtain the pair (A, b) from (A, T) by multiplying several
rows by nonzero constants and renumbering the indices, then we regard (A, T)
and (a, ~7) as equivalent; the corresponding contragredient Lie superalgebras
are isomorphic.
Observe that if I1 C 1, A, is the corresponding principal minor of A, and
T1 = I1 n 7, then the subalgebra of G(A, T) generated by the elements e, ,
fi , and hi , i E I1 , is isomorphic to G(A 1 , T1). Note also that if A is decom-
posable,
then G(A, T) splits into the direct sum of the algebras with the Cartan matrices
A, and A,.
Let H = (h, ,..., h,), let 0~~,..., 01~ be the linear functions on H defined
by the relations ori = aji , j = l,..., r, and let M be the free Abelian group
with the generators 01~,..., 01~. We set G, = ([...[ei, , ezz] ,..., e<J), G-, =
<P-Lfj, ,fiJ,...,fiJ>, a = c %, .
As m [ll], it is easy to show that G(A, T) = H @ (0 G,).
Many assertions about contragredient Lie algebras in [8, 111 remain valid
for Lie superalgebras (with the same proofs). Here we state only those that
are needed in what follows.
PROPOSITION 2.5.1. The center C of G(A, T) cons& of the ekments of the

fm 2 rdhd , whe X atnf = 0.
PROPOSITION 2.5.2. L-et G(A, T) be finite-dimenkmal and C be its center.

Then G(A, r)/C is sim+ if and only if
for any i, je I there exists a sequence il ,..., it E I

(4
for which aif,ai,f, ,..., ai,, # 0.
PROPOSITION 2.5.3. Suppose that G(A, r) is &ite-dime&ma1 and satisjies

(m). Then otz G(A, 7)/C’ there is a nom&generate con&tent supersymmetric
invkt bilinear form. Thtk induces a form ( , ) on G(A, r) having the fohbwing
properties:
(1) The hernel of the form ( , ) is C;
(2) (G , G,J = 0 when 0~# -P;
(3) thefm(, )det ermines a nondegenerate pairing of G, with G, ;
(4) ka , e-d = (e, , e-3, , where h, is a nonzero vector in H fm which
(h, , h) = a(h), h E H.
Roof. The mapping 8: ei ++ -fi , hi F+ -hi , h ++ -(- l)deaarei evidently
induces an automorphism of G(A, T)/C. Hence, all the conditions for Proposi-
tion 1.2.5 are satisfied and the required form exists. The remaining properties
are proved as in [ll].
2.5.2. Existence of the exceptional Lie supera&ebras D(2, 1; a), F(4), and G(3).
PROPOSITION 2.5.4. We COB&Y the following matrices (a E h/(0, -I}):
(a) For G(D,, ill), G(F,, U)), and G(G, {I)) the Grmdule G is
isomorphic to sl, @ sl, @ slz , spin, @ ~1, , and GB @ sl, , respectively.
(b) The G(D, , (1)) wchmrst all simple Lie superakebras G = G @ Gi
for which the Gmodule Gi is isomorphic to sl, @ sl, @ s18. Two members
D(2, 1; a) and D(2, 1; p) of this family are isomorphic ;f and only if a! and p
lieinthcsanreorbitbftAegroupVoforder6g~atedbyor~-l -~,ol~l/or.
Proof. (a) and the condition for isomorphy of members of the family
D(2, 1; CX)are established exactly as for [8, Proposition 3.61.
Now let G = G$ @ Gf be a simple Lie superalgebra for which the +
50 V. G. KAC
module Gi is isomorphic to sl, @ sl, @ ~1, . Then GG = Ai” @ Ai2’ @ At3),

where A:i’ = (eoLi , hej ,&>. L e t e, and fr be the lowest and highest weight
vectors. Since G is simple, [e, ,fr] = h, # 0. We define aZ-grading G = @ Gi
on G by setting Cl = <fl , .fa, ,fa,>, Go = (hl , hE2, Q, and G = (e, , em2, e,,>.
Then G is equipped with the structure of a contragredient Lie superalgebra
G(A, T), where
1
0 ~2(hl) ~3@1)
A=-1 2 0) 7 = (1).
[ -1 0 2
Since G is simple, c&z,), a3(h,) f 0. So we have G N D(2, 1; am %(/z&l).
Remark. The family D(2, 1; a) becomes a family of Lie algebras if the

characteristic of the field is 2. This family of Lie algebras is studied in [8].
Note that group I/ has three exceptional orbits: (0, -1, a3}, (1, -2, -$},
and (-4 & i3112/2}. D(2, 1) corresponds to the second orbit. Superalgebra,
corresponded to the third orbit, admits an outer automorphism of order 3.
2.5.3. The root decomposition of the classical Lie superalgebras. Let G =

G, @ Gi be a Lie superalgebra and H be a Cartan subalgebra of Ga . We call
H a Cartan subalgebra of G. Since every inner automorphism of G, extends
to one of G (see Proposition 1.1.1) and Cartan subalgebras of a Lie algebra
are conjugate, so are Cartan subalgebras of a Lie superalgebra.
A Cartan subalgebra of a classical Lie superalgebra is diagonalizable. There-
fore, we have the root decomposition:
G = @ G,, where G, = {a E G 1 [h, a] = a(h) a for h E H}. (2.5.1)

JiCH*
The set d = {a E H* 1 G, # 0} is called the root system. Clearly, d =

d,, u d, , where d, is the root system of G6 and d, is the system of weights
of the representation of G, on Gi ; d, is called the system of even and d, that of
odd roots.
A straightforward inspection of examples of classical Lie superalgebms
together with standard arguments from the theory of Lie algebras yields the
following information on their root decompositions.
PROPOSITION 2.5.5. Let G be a classical Lie superalgebra and let G = 0 G,

be its root decomposition relative to a Cartan subalgebra H.
(a) G, = H in all cases except Q(n);

(b) dim G, = 1 for OL# 0, except for A(l, l), P(2), P(3), and Q(n);
(c) On G there is one and, up to a constant factor, only one nondegenerate
invariant supersymmetric bilinear form ( , ), except for P(n) and Q(n).
(d) A, and A, are invariant under the Weil group W of GG .
LIE SUPBRALGEBBA!3 51
(4 Ij G is A@, n), (m, 4 # (1, 11, B(m, 4, c(n), D(m, n), W, 1; 4

F(4), or G(3), then the jo&wing properties holdz
(1) [G,,G,1 #Oif-f~yif~,B,~+B~4
(2) (6 , '44 = 0 for a # -8;
(3) tkjmm( 3 l&t enmites a not&generate pair+ of G, with G-, ;
(4) [eel, 4 = (em , 44, , where h, is the nonzero vector &term&d
by (ha , h) = 44, h E H;
(5) if OLis in A (or d, , or Al,), then so is --a;
(6) kor~djoror#O,k#fl,ifandonZyz~u~d~and(u,u)#O;
herek=f2.
2.54. Expricit a%scr$tion of tk system of roots and of simple roots. A system
of roots n = {‘or ,..., a~,>C A is said to be simple if there are vectors ei E G@,,
jr E G-,$ , for which [ei , jj] = 6,,hd E H, the vectors e, and jt , i = I,..., Y,
generate G, and I7 is minimal with these properties. Below we describe for
A@, n), B(nt, n), C(n), D(nr, n), D(2, 1; LX), F(4), and G(3) the systems of
even nonxero roots d,’ and of odd roots d, , and all systems of simple roots,
up to W-equivalence.
In all the examples the Cartan subalgebra H is a subspace of the space D
of diagonal matrices; the roots are expressed in terms of the standard basis
ci of D* (more accurately, the restrictions of the l t to H).
A@, n). The roots are expressedin terms of linear functions t1 ,..., l ,,,+~,
8, = %+a 9-*-P
8n+1 = %+n+a-
A,’ = {q - 4 ; 8, - a,>, j #j; 4 = k!+t - w

Up to W-equivalence, all the systemsof simple roots are determined by two
increasing sequencesS = {sr < s, < *+e}and T = (tl < te < **.} and a sign:
J&J = It{%- ra, E*- c ,..*,Qal

- 6,, s,- 611
,...,s,,- hl+l)...).
The simplestsuch systemis
Gl - ‘8 ?% - 6 ,-**,%?a+1
- 81, s, - 8a ,a-*,48 - &l+ll.
B(m, n). The roots are expressed in terms of linear functions l 1 ,..., +,, ,
81 = %lnfl ,*-*,&I = Qam+o
*
d,’ ={Itcd f 4 ; IliZ-%; fci ; fsi f &)9 i#j; 4 = w, ; far f %

increasing sequencesS and T:
52 V. G. KAC
The simplest such system is
(6, - 6, ,..., 6, - 61 , fl - E$)..., G-1 - E, ) cm]
if m > 0, and
v4 - 6, ,***, Ll - %I >u
ifm =O.
C(n). The roots are expressed in terms of linear functions E1 , 6, = l a ,...,
&a-l = %+1*
A,’ = (k-26, ; fSi & S,}; 4 = k&l xt Si>.
Up to W-equivalence there are the following systemsof simple roots:
&I(% - 61,a, - 6, >*.a,L, - Ll ,2L3;

It{% - 6, ,*.a,4 - El, El - ai+1,**., L, - Ll 92L,);
&t(Sl - 6, ,**-, L - ha-1 >h-1 - 9 , L + 4.
D(m, n). The roots are expressedin terms of linear functions or ,..., E, ,
6, = %n+17**.>
%a= %m+n.
dO’ = {I!I% III Ej ; f2si ; fsi f sj}, i#j, 4 = {Itsi i Sj>.

increasing sequencesS and T, and a number:
n;., = {El - Q I..., Es1- 6, , 6 - 6, ,...) s,, - Es,+1)...) cm-1- %n9%a-1+ %I
(or 6, - cm,a, + 4>;
nf,, = (3 - 62,a.., es1 - 6, , 6, - 6, ,..., s,, - l al+1,...) 6,-l - 6, , 26,).
The simplestsuch systemsare
{S, - 6, )..., 6, - 99 El - E2 9***> G-1 - %I ,%n-1 + 'mh
{El - 9 ,..a, %a- s, , 6, - 6, ,...) 6,-l - 6, ) 2S,}.
D(2, 1; a). The roots are expressedin terms of linear functions or , c?P
and l a .
A,’ = {&2q}; 4 = k!51 i f2 i $1.
Up to W-equivalence there are four systemsof simple roots:
{E1 + E2 + E3 t -26i > -zEj}, i#=j, i,j= 1,2,3;

{El + 52+ E3 7 q - E2 - E3, -El - E2 + E3).
F(4). The roots are expressed in terms of linear functions l 1 , es , l s ,

corresponding to Ba , and S, corresponding to A1 .
Up to W-equivalence there are four systemsof simple roots:
G(3). The roots are expressed in terms of linear functions 9 , e2, ~a,
corresponding to G2 , Er + r, + l 3 = 0, and 8, correspondingto A1.
A,’ = {q - 4 ; fCf ; f2S); dl = (fEi & s; Its)-

Up to W-equivalence there is a unique system of simple roots:
@ + l l 9 62 , % - c21-
2.5.5. Examplesof Jinite-dimensionalcontragredientLie super&ebras. Exam-

ples can be obtained in the following manner. Let G be one of the Lie super-
algebras sZ(m+ 1, n + l), B(m, n), C(n), D(m, n), D(2, 1; a), F(4), or G(3).
Let H be a Cartan subalgebra,n be one of its systemsof simple roots (listed
in Section 2.5.4 above), and e, and fi be the corresponding nonzero vectors
in G=, and G_,, , respectively, a, E II. Then the vectors [et ,fi] = Rdform a
basis of H. Setting deg ei = -deg fi = 1, deg hi = 0, we define a Z-grading
on G. Since G is simple modulo its center, G is the minimal Z-graded Lie
superalgebrawith the local part G-, @ Go@ G1 . In this way the structure
of a contragredient Lie superalgebrais introduced in G. Its Cartan matrix is
A = (q(hJ), and 7 = {iEIl %EAJ.
In the next section we show that these examplesexhaust all simple modulo
center finite-dimensional contragredient Lie superalgebras.
We now list all the resulting pairs (A, I), up to equivalence. For this it is
convenient, as usual, to introduce L$nkin d&grams. To begin with we extract
from our examplesin Tables IV and V all the contragredient Lie superalgebras
of rank 1 and 2 with indecomposable Cartan matrices, the corresponding
pairs (A, T), and Dynkin diagrams.The circles 0, 0, and l are called, reapec-
tively, white, gruy, and b&k. Contragredient Lie superalgebrasof rank I are
depicted by a diagram consistingof r white, gray, or black circles; the ith circle
is white if i 4 7 and gray or black if i E r and ati = 0 or 2, respectively. The
ith and the jth circles are not joined if afj = aji = 0; otherwise, they are
joined as shown in Table V (note, that isomorphic Lie superalgebrasmay
correspondto different Dynkin diagrams).
54 v. G. KAC
TABLE IV
GM, 4 A 7 Diagram dim G(A,T)
4 (2) 4 0 3
w, 1) (0) (11 0 3
WA 1) (2) (11 0 5
TABLE V
GW 7) A 7 Diagram dim G(l, T)
A, t-f -3 o-o 8
B2 ( 2-7-2 2
0-o 10
GZ 030 14
(2 3
W,O) O-O 8
c-: 3
AU, 0) O--o 8
( ; 2
BU, 1) O-0 12
c-i 3
B(1, 1) 0+-e 12
co-2 2‘1
W, 2) (4 3
0-e 14
Matrix A also satisfies the following restrictions. If a. = 0, then in every

submatrix of order 3 with aii in the center, the diagram of which is not a cycle
and does not contain two arrows, the sum of the elements of the second row
is 0. For the diagram @ + @ -+ 0, it is always uzl = -2qs , and for the
cycle,
0
I\
O--C3
itisalwaysa,= --2a,, . We introduce matrices
D..?(fIp 3, D:=(_lta i -‘ia).
The diagrams
0
I\
0+0-o and o--o
always correspond to matrices D, and D,‘, respectively, and a = 1 unless the

contrary is stated.
The following proposition is a consequence of the results in Section 2.4.
PROPOSITION 2.5.6. Let G be one of A(m, n), B(m, n), C(n), D(m, n), D(2,l; a),
F(4), or G(3) und let G cx G(A, T)/C, where C is the center.Then C # 0 om’y
forA(n,n)andinth&caseclimC=l. Thedkgmnrofthepairs(A,~)cmbe
demibed as folkbws (eachpoint can be u white w a gray &CL%):
A . _ . _ . _ ... _ . _ .
B . _ . _ ... _ . _ . 5 0, .-.-...-.-.*a
0
t
CD ._._..._._. + 0,
.-.-...-.-oeo, .-a -...-. +@+O,
0
I\
. _ . _ ... _ . _ . -69
D(2, 1; 4 o+o+o, A =D,, /3 = a, -(l + a), -a(1 + a)-’
A
o- 09 A =D,
F(4) o-o-=0-00, o-o+-c3+@,
0
I\
o*o+o+o, 0~0-0,
where for the last two diagrars, subdiagram without the Ji*st circle correspond
to matrices D-s and D> , respectively.
G(3) O-Of 0.
56 V. G. KAC
Below (Table VI) we list the “simplest” diagrams, the coefficients of the
decomposition of the highest root into simple roots, the index s of the only
nonwhite circle, and the number r of the circles.
2.5.6. The classiJcation of Jinite-dimensional contragredient Lie superalgebras.
THEOREM 3. Let G(A, T) b e a Jinite-dimensional contragredient Lie super-

algebra whose Cartan matrix satisfies condition (m) of Proposition 2.5.2, and let
C be its center. Then G’ = G(A, Q-)/C is classical, and (A, r) is equivalent either
to one of the pairs listed in Proposition 2.5.6 or to (A, $), where A is the Cartan
matrix of a simple Lie algebra.
Proof. According to Propositions 2.5.1 and 2.5.2, C _CH and G’ is simple.
Therefore, by Theorem 2 and Proposition 2.5.6, it is sufficient to show that
the linear Lie algebra GG’ acting on Gi’ is reductive.
According to Proposition 2.5.3, on G’ there exists a nondegenerate invariant
bilinear form ( , ). We consider a new form on G’: f (x, y) = (x, By),
where fl is the automorphism from the proof of Proposition 2.5.3 (B(G,) =
G-,). Clearly, the restriction off to G, is nondegenerate, and the operators
ad e, and -ad Oe, are dual with respect to this form. Hence, it follows that
[eel , Oe,] = h is a nonzero element of H. For otherwise we would have two
dual nonzero commuting nilpotent operators, which is impossible.
Now let R be the radical of G,‘. Evidently, R is graded relative to the root
TABLE VI
G Diagram s I
Ah 4
B(m, n), m > 0
W, 4
C(n), n > 2
D(m, n)
F(4)
G(3)
LIE SuPJfRALGEBRAS 57
decomposition. We claim that R C H, which will prove the theorem. If this

were not so, then e, E R for some a # 0, and [eel,0eJ = h # 0, b E R. Since
R is solvable, [h, e,J = 0; hence c@) = 0. Now we look at the adjoint repre-
sentation of the subalgebra (ee,, h, e,). By the Lie theorem, in some basis
the matrices of ad eea and ad e, are triangular. But then the matrix of ad h =
[ad e, , ad &J is also triangular with zeros along the main diagonal. Since
ad h is diagonalixable, we see that h = 0. This is a contradiction and proves
the theorem.
2.5.7. Z-grudings. It is not hard to show, just as in [12], that the relations
degei = -degf, = ki, degh, = 0, k,EZ, i = I,..., I, determine all possible
Z-gradings of finite-dimensional contragredient Lie superalgebras. In particular,
if (A, T) is a pair from Table VI, then for ki = 0, i # s, k, = 1 we obtain
the Z-gradings of A(nr, n), B(nr, n), C(a), and D(nr, n), as described in Section 2.1,
and for D(2, 1; a), F(4), and G(3) we obtain consistent Zgradings of the form
G-, @ Gml @ G, @ G1 @ Ga , where dim G&a = 1 and the G,,-modules G+1
are isomorphic, respectively, to so, @ k, spin, @ k, and Gs @ k. In the same
way, the Z-gradings are defined for Q(n) (as Q(n)a = A,, and we can naturally
identify spaces Q(n)i and Q(n)5 : gat) e, , hi t) hi).
3. CARTAN LIE SUPERALGEBRAS
3.1. The Lie Superakgebras

W(n)
3.1.l. Defnitiun of W(n). Let A(n) be the Grassmann superalgebrawith

the generators 61 ,..., 6,. We denote der A(n) by W(n). We recall (see Section
1.1.4) that every derivation D EW(n) can be written in the form
where a/a& is the derivation defined by
Letting deg & = 1, i = l,..., n, we obtain a consistent Z-grading of /l(n),

which induces one of W(n) = &+-i W(n), , where
deg Pr = k + 1, i = l,..., ?t .
I
58 V. G. KAC
In particular, W(n)-, = (a/a& ,..., a/a&). Hence it follows that

[a/afi , a/a&] = 0, that is,
a a a a
--=---s (3.1.1)
ah a& 6 6
Formula (3.1.l) is one of the standard facts of analysison a Grassmannalgebra,
as developed in [2].
We now list someproperties of W(n).
PROPOSITION 3.1.1.
(a) W(n) = @!I:, W(n)< is transitiwe.

(b) The W(n),,-moduleW(n)-l is isomorphicto gZ, .
(c) W(n), = w(n): ) h >, 1.
(d) W(n) is simplefor n > 2.
(e) If G = @Q-~ Gt is a transitive Z-graded Lie superalgebrafor which
the GO-modulesG*, are isomorphicto the W(n),-modulesW(n)*, , then G N W(n).
Proof. Properties (a)-(c) are easily verified directly. (d) follows from
Proposition 1.2.8 and (e) follows from (c) and Proposition 3.1.2 below.
3.1.2. The universality of W(n) as a Z-graded Lie superalgebra. Let G =

@i)-1 G$ be a transitive Lie superalgebrawith a consistent Z-grading, and
dim G-, = n. Then the map &: G, + Hom(D+l(G-J, G,), defined by
determinesa monomorphismof G,-modules &: Gk -+ A”+l(G_*,) @ G-, . The

x,&‘syield a canonical monomorphism of G,,-modules
4: G-t @ (Ak+‘(G?l) 0 G-,).

k>-1
From this we see by dimension arguments that if G = W(n), then # is an

isomorphismof W(n),-modules.
If we are now given a monomorphism of the Gs-module G-r into W(n),-
module W(n)-, , we obtain a chain of maps:
G --+ 9 (Ak+l(G_*l) 0 G-1) - W(n).
It is easy to verify that the composite map is a monomorphism of Z-graded

Lie superalgebras.So we have the following result.
LIE SWERALO&BRAS 59
PROPOSITION 3.1.2. Let G = &>+ Gt be u tmn.ri~~ ~~@&ge&a cvitir a

con&tent Z-grading, and dim GWI = n. Then there is an embedding of G in
W(n) preserwing the Z-gradiq.
3.1.3. The uniwersality of W(n) as a Lie superaigebra with a filtration. W(n)

is canonically equipped with a filtration. Now it turns out that there is also
an embedding theorem for Lie superalgebras with filtration.
PROPOSITION 3.1.3. Let L = L, r) L,, I& 3 **a be a tratuitive Lie super-

akebra with a Jiltration, dim L/LO = n, and suppose that L, contains L6. Then
there is an embedding u: L + W(n) preserving the $hztion. If p is any other
such embedding, then there is one and otdy one automorphism @ of W(n) that
is induced by an automorphism of A(n) for which OL= # o Is.
The proof carries over almost verbatim from [20], with the definitions
replaced by the relevant definitions in Section 1.1.3. True, the proof in [20]
only gives the existence of @ under the assumption that (a - B)(L) CL,, .
However, this assumption is easily dispensed with, by modifying /3 to an
automorphism of W(n) induced by a linear automorphism of n(n).
As a corollary to Proposition 3.1.3 we have the next result.
PROPOSITION 3.1.4. Let L = L-, I) L, 1 L, 1 *a* be a s&algebra of W(n)

m*th the induced Jiltration, and dimL/Ls = n. Thea every automorphism of L
preserving the filtration is induced by an automorphism of A(n).
Clearly, in W(n) with n 3 3 there is a unique subalgebra containing W(n), ,
n=ely, Okh Wh . Hence for n > 3 this subalgebra, and therefore, the
filtration in W(n), are invariant under all automorphisms. So we obtain the
next result from Proposition 3.1.4.
PROPOSITION 3.1.5. Every aatomorphism of W(n) with n > 3 is induced by

an automorphism of A(n).
3.2. Two Algebras of DifferentialForms

3.2.1. The superalgebra Q(n). Let /t(n) be the Grassmann superalgebra on
51 ,..., 5, . We denote by Q(n) the associative superalgebra over n(n) with the
generators d& ,..., d[, and the defining relations d& 0 d& = d& 0 d& ,
deg dti = Ti, i, j = l,..., n. Note that D(n) is commutative (in the sense of
the bracket); in particular, &d& = (d~,)l‘i .
Every element 9 E Q(S) can be written uniquely as a sum of elements of
the form
60 V. G. KAC
We define on Q(n) the differential d as the derivation of degree i for which
i = l,..., n.
It is easy to verify, as in Section 1.1.4, that this derivation exists and is

unique.
Q(n) is called the superalgebra

of dzjkential forms with commutingd~@rentials.
PROPOSITION 3.2.I. The d#erential d hasthefollowing proputies:
(a) d(p,0 4) = dp,0 4 + (-l>deB~~ 0 44 v, # EQ(n).

(b) df = Xi afP& d& , f E 44.
(c) d2 = 0.
(d) Every derivation D of A(n) extendsuniquely to a derivation D of Q(n)
for which [s, d] = 0.
(e) Every automorphismof A(n) extends uniquely to an automorphism
of Q(n) commutingwith d.
Proof. (a) is true, by definition.

(b) is proved by induction on the Z-grading of A(n) = @ Ai . Suppose
that f E A, ; it is enough to prove (b) for f = fitj , where fi E A,-, . By the
inductive hypothesis we have
df = (dfd & + (-I>“-‘fi d& = -$ (3) z d& + (-I)“-‘fi d&
= F & (fib) @i ,
as required.
(c) We now define a height h on Q(n) by putting h(dtJ = 1, h([J = 0,
i = l,..., n. We conduct the proof of (c) by induction on h.
If h(f) = 0, thenf EA(n) and
by (X1.1). Suppose, next, that h(v) = k; it is enough to prove (c) for v =

IJJ~o dti . By induction we have d2(q+0 d&) = d(drp,0 d&) = 0.
(d) Let D = C P,(a/a[J be a derivation of A(n) of degree k. It is easy
to verify, as in Section 1.1.4, that there is one (and only one) extension of D
LIE SWERALGEBRAS 61
to a derivation fi of Q(n) for which qd&) = (-1)’ d(D&), i = l,..., R. We

claim that Dd = (-l)K dl?. Letfe A, ; then we have
Now clearly we have
Wf G, o a*. o d&J = (-l)k dD(f d&, 0 *** o d&J
as required.
(e) is proved rather like (d) (b esi d es, as is easy to see, it follows from (d)).
The proposition is now proved.
Note that (d) gives us an action of W(n) on Q(a), commuting with d in the
sense of the bracket. There is also an analog to Poincare’s lemma.
PROPOSITION 3.2.2. If a d$%mntial fm # E Q(n) is closed, that is, d$ = 0,

then Q = d& for mm & E Q(n).
Proof. We consider the linear map (homotopy operator) K: Q(n) + QH)
and the homomorphism l : Q(n) + Q(n), defined by the formulas,
K( f dfi, 0 a** 0 d5ilml 0 d6pJ = IJ dti, 0 see0 dtt,-l*
K(f d5il 0 -es o d&,) = 0 when it # n, t = l,..., S, f E A(n),
~(6~) = & and @&) = 4 if i # n, 4%) = +$,J = 0.
It is easy to check that Kd + dK = 1 + E. Therefore, if d$ = 0, then 1,4=

WW - 44, w h ere l (I,!J) is closed and does not depend on 6,. Now we
can use the induction by n.
3.2.2. The superalgebra e(n). We denote by 9(n) the associative super-
algebra over A(n) with the generators f?& ,..., et, and defining relations
et1 A etj = -etj A e& , deg I& = i, i,j = l,..., n.
Note that e(n) is commutative (in the sense of the bracket); in particular,
ut, = -uvx, .
62 V. G. KAC
Every element w E S(n) can be written uniquely as a sum of elements of the

form
We define a differential 0 on O(n) as the derivation of degree 0 for which
Wi) = 6 7 e(efi) = O, i = l,..., n.
It is easy to verify that this derivation exists and is unique.
PROPOSITION 3.2.3. The d@rential 8 has the following properties:

(a) @J, A w2) = 44 A w2 + w1 A qw,).
@) e(f) = Xi WtX~f/W f E 44.
(c) Every derivation D of A(n) extendsuniquely to a derivation b of 8(n)

for which &If = @f, f E A(n); ;f e2(D(5‘,)) = 0, i = l,..., n, then be = @.
(d) Every automorphism@ of A( n) ext ends uniquely to an automorphism6
of S(n) for which c&f = d6f, f E A(n).
The proof is similar to that of Proposition 3.2.1. Observe that e2 # 0. For

example, e2(&[,) = 285, A et, . Also, it is not true that [D, 01 = 0 for
D E W(n). Nevertheless, (c) provides us with some action of W(n) on O(n).
3.3. Special and Hamiltonian Lie Superalgebras

3.3.1. Volumeforms and the Lie superalgebras
S(n) and s(n). A volumeform
is a differential form in O(n) like
w = fetl A ..’ A et,, f E 446 9 f (0) f 0.
To a volume form w there correspondsin W(n) the subalgebra
S(w) = {DE W(n) ) Dw = 0).
Among these subalgebraswe single out two: S(n) = S(L)& A ... A &) and
S(n) = S((1 + & ... 5,) et1 A ... A 08,) for n = 2K.
The condition for an operator C Pi(a/@,) to belong to S(w) can be written
asfollows:
Hence it is easyto seethat S(w) is the linear span of the elementslike

aa a aa 8
a E A(n).
f-lag,% +f-‘ag,agi’
On S(W) a filtration is induced from W(n), and on S(n) clearly even a Z-grading.
PROPOSITION 3.3.1.
(a) Gr S(w) N S(n) = @iI?1 S(t& .
(b) port of S(w)a is isolnorpliic to sl, .
The semiw’mple
(c) The S(n),-mod& S(n), is iwmorphic to sl,, .
(d) The S(n),,-mod&s S(n)k are irreducible and isomorphicto the h&hest
componentof the??mddsl, @An-k-lsl~ .
(e) S(n), = S(n): , K > 1.
(f) The Z-graded Lie suMalgebra S(n) is transitive. The S(w) are simple
for n 2 3.
(g) Ewy automorphismof S(w), n > 3, is induced by an automorph
of A(n) under which the d#erention four w is multiplied by an elementof k.
(h) If G = %>-I ‘5 is a transitive Z-graded Lit superalgebrafor which
the Go-moduleGml is &morph to sl, , then G -N S(n).
Proof. Properties (a)-(e) are easily derived from the description of the
elements of S(w). The fact that S(w) is simple now follows from Proposition
1.2.8; therefore, (f) is true by (a).
To prove (g), we note (aswas done in [15]) that if wr and ws are not propor-
tional, then S(UJJ # S(ws) and that the filtration in S(w) for n > 3 is invariant
under automorphisms. (For n > 3 this is proved as in Section 3.1.3, and for
n = 3 it is obvious.)
Finally, (h) is obtained by embedding G in W(n); clearly, then Gf = S(n)i
for i = -1, 0, and we can then use Proposition 3.3.2 below.
WesetDs =C &@/a&), T,(n) ={fD,,f eA(n),}C W(a), ,R = O,..., n - 1.
The next proposition is easyto obtain, for example, by dimensionarguments.
PROPOSITION 3.3.2. W(n), = S(& @ T,(n), K > 0, is a direct sum of

irreducibleW(n)s-modules,and the W(n),-module Tk(n) is isomorphicto A’+ISgl,,.
We now obtain a classificationof volume forms.
PROPOSITION 3.3.3. Every volumeform o = fogI A **aA et,, can be reduced

by an automorphismrof A(n) to the shape
(u+811...~n‘,)e51h...Ae~nn, u#O; /9= 0 whenni.soaX
Proof. We may sssumethat n > 3. The semisimplepart of the Lie algebra

S(w)s (which exists by Levi’s theorem) can be carried, by Maltsev’a theorem,
into S(n),, C W(n), by an inner automorphism Q, of W(n), . According to
Proposition 1.1.l, 0 extends to an automorphism of W(n) and is, therefore,
induced by an automorphism y of A(n) (seeProposition 3.1.5). Replacing w
607/26l1-5
64 V. G. KAC
by P(W), we may assume DW = 0 for D E S(n), . But S(n), 3 {[,(8/a&), i # j};

therefore, &(af/a&) = 0 for i # j, and hence, f = 01+ /3fl *a-&,, and p = 0
for odd n.
Taking Proposition 3.1.3 into account, we can derive the following result
from Proposition 3.3.3.
PROPOSITION 3.3.4. Every superalgebraS(W) is isomorphicto one of S(n)

OY s(n). Thesetwo Lie superalgebras
are not isomorphic.
PROPOSITION 3.3.5. Let L = L-, 3 L,I Ll r) e-ebe a Lie superalgebrawith

a @ration, and Gr L ‘v S(n). Then L N Gr L E S(n) for odd n, and L N S(n)
m S(n) for evenn.
Proof. Using Proposition 3.1.3, we embed L in W(n). Then La C W(n), ,
and by applying Maltsev’s theorem to this pair, we may assume that the semi-
simple part of L, (which is isomorphic to s&J lies in W(n),, , and hence, that
L I S(n), .
We now observe that W(n) and L are S(n),,-modules, that the S(n),-modules
L and S(n) are isomorphic and, by Proposition 3.3.2, that W(n) = S(n) @ T(n)
is a direct sum of S(n),,-modules, where T(n) = ok T,(n).
From Propositions 3.3.1(d) and 3.3.2 it is clear that the S(n),,-modules T(n)
and L can only contain a unique common simple component:
Gr-, L rv T,-,(n).
Therefore, L = V @ (Ok>,, S(n),), where V C S(n)-, @ T,-,(n). If I’ =

S(n)-, , then L = S(n). F or odd n there is no other possibility because V C Li .
But if V # S(n)-, for even n, then L N s(n), as is easy to see.
3.3.2. Hamiltonian forms and the Lie superalgebras H(n) and R(n). A
Hamiltonianform is a closed differential form in Q(n) of the kind
To a Hamiltonian form w there corresponds a subalgebra of W(n):
R(w) = {D E W(n) [ Dw = O}.
We set H(w) = [I%(W), R(w)], R(n) = fi((d[,)” + .** + (d[,J2), and H(n) =
lRn>, W41.
It is not difficult to see that the condition for C P,(a/a&) to belong to B(W)
can be written as follows:
By the analog to Poincare’s lemma, this condition shows that there is an

element f E n(n) (depending on 0) for which x:t w,,P, = 8f&$ . Therefore,
if (cz,,) is the inverse matrix to (CO& we see that R(w) consists of all the elements
of the form
fe 49, f(O)= 0,
and that [D, , D,] = Db,#) , where
In particular, Q(n) consists of the elements of the form
4=$& , 1 f E44, f(O) = 0,
and the bracket looks as follows:
A filtration is induced on R(w) from W(n), and on R(n) and H(n) even Z-
gradings.
PROPOSITION 3.3.6.
(4 Gr B(w) N R(n) = @Ei!, R(n), .

@I The s&mph part of $&)a is isomorphic to so, .
(4 H(n) = @;zl R(n), , that is a(n) = H(n) @ (D,l...c,).
(4 H(n),, = H(n); , K 2 1.
@%)a!, &~-II = H(&,f~ h < n - 2; l3-W , H(&l = H(n),-, ,
k<?
\ *
(f1 The &&,-moduZes B(n), are isomo~phzi to /P+*so,, , - 1 < h < n - 2.
k> The Z-graded Lie s+eraZgebras H(n) and R(n) are transitive.
04 H(n) is simpk for n 2 4.
(9 Every u&morphism of H(w) for n > 4 and of i&u) fmw n > 3 is
induced by an automorphism of A(n) under which w is mltiplied by an element of k.
(j) If G = @Q-~ Gt is a transitive z-graded Lie supera&bra for which
the G,,-module Gvl is isomorphic to so, , then G N H(n) or R(n) or P(3).
Proof. Properties (a)-(i) are established just like the corresponding assertions
in Proposition 3.3.1.
66 v. G. KAC
(j) is now a consequence of the following assertion: If G = &>-I Gi is a

transitive Lie superalgebra with a consistent Z-grading and if the Go-module
G-, has an invariant symmetric bilinear form ( , ), then there is an embedding
of G in R(n) preserving the Z-grading. To prove this we construct embeddings
z&: G, + Ak+2(G_*,) by the formula#,(g)(u,,..., uk+s) = ([...[g, a,],..., a,,,], ++s).
The subsequent arguments are the same as in Section 3.1.2.
Remark (cf. [22]). Let %n be a Clifford superalgebra with the natural

Zs-grading. The bracket turns this into a Lie superalgebra (%7& . The factor
algebra (V&/(l) is isomorphic to R(n).
PROPOSITION 3.3.7. Let L = L-, 3 L, 1 L, 3 ... be a Lie superalgebra with

ujltrution, and GrL ‘v H(n) or R(n). Then L = GrL.
Proof. By Levi’s theorem, L6 contains a subalgebra Go isomorphic to so,

(see Proposition 3.3.6(b)). Now L splits into aZ,-graded direct sum of irreducible
G,-submodules, and the G,,-module L is isomorphic to the Gr, L-module Gr L.
Hence we see that L = G-, @L, is a direct sum of G,-modules, and that
the Go-module G-, is isomorphic to so, . In particular, [G-i , G-,] C S2so, ;
hence, and from Proposition 3.3.6(f), it is clear that [G-i , G-i] CL,-, . It
follows that [G-i , G-i] can differ from zero only when n is even and Gr L c1
R(n). However, in that case [G-i, G-r] = L,-, = (Dfl...E,), and for a E G-,
we have [[a, a], a] = (a, a) [DE1.+, , a] = 0 (by the Jacobi identity); this
contradicts the transitivity of A(n).
Thus, [G-i , G-i] = 0. We now embed L in W(n) with preservation of the
filtration 01:L -+ W(n), using Proposition 3.1.3. Since [G-, , G-i] = 0, there
is an embedding /3: G-, ---f W(n) for which /3(G-,) = W(n), . Therefore,
OLcan be modified (by Proposition 3.1.3) to an automorphism of W(n) for which
4-l) = W% v..., WL>.
Now let GrL = H(n). Then W(n) 3 L = L-, 3 Lo3 ... 1 L,-,I 0, where
G-r = W(n)-, and the G,-module L,-, is isomorphic to so, , and L,-, C
W(n),-, @ W(n),-, ; hence, it follows evidently that L,-, C W(n),-, . Therefore,
(ad G-,)8-3 L,-, C W(n),-, .
In this way, in accordance with Proposition 3.3.6(e), a Z-grading is introduced
on L that is consistent with the filtration; therefore, L = Gr L.
The arguments for the case Gr L E R(n) are similar.
It is now easy to prove the analog to Darboux’ lemma.
PROPOSITION 3.3.8. Ezery Humiltoniun form w = C wij dti 0 d& can be

brought by an uutomorphism of A(n) to the shape
f. Md2.
i=l
Pmof. We consider R(U). Atxmding to Proposition 3.3.6(a), Gr R(w) N

&z). Therefore, by Proposition 3.3.7, &w) N R(rr), and from Proposition 3.1.3
it follows that there is an automorphism d of A(n) carrying @UJ) into &I).
Therefore a/a&(@(~)) = 0 (bec~u~ea/at< E R(n)) and @(uJ)= C cij d[i 0 dt; 3
c<jE K. It remainsto apply a linear automorphism.
3.3.3. DEFINITION. The Lie superalgebrasW(n) for n > 3, S(n) for n 2 4,

S(n) for even n > 4, and H(n) for n >, 5 are called Curti Lie ~up”aZg&cs.
(For other values of tr they are either not simple or isomorphic to classicalLie
superalgebras:W(2) N A(1, 0) N C(2), S(3) N P(2), H(4) N A(l, 1))
4. THE CLASSIFICATION THEOREM
In thii chapter we complete the classiticationof the simple Lie superalgebras

over an algebraically closedfield of characteristic 0.
4.1. ‘Classijkationof Certain Z-Graded Lie Superalgebras

4.1.l. The noncontragredientcase. Here we prove the following important
proposition, which is the most complicated technically.
PROPOSITION 4.1.1. Let G = @i=, Gi be a bitransitive Lie superalgebra

with a consistentZ-grading, for which d or t = 1, and supposethat
(a) G,, is semisinzple;
(b) the representationsof Go on Gml and Gl are irreducible;
(c) the representatkms of G,, on Gwl and Gl are not contragredient;
(d) G-i 0 G, @ G1generatesG.
Then G is isomorphicasa Z-graded superalgebrato oneof S(n), H(n) with n > 4,
OYP(n).
The proof of the proposition is basedon an analysisof the relations between
the highest vector FA of the representationof G,, on G-i and the lowest vector
EM of the representation of G,, on G1. We know (seethe proof of Proposition
1.2.10) that
PA, &I = e-, , (4.1.1)
where OL= -(A + M) is a nonzero root of Go . Interchanging if necessary

Gk with G-, , we may assumethat 01> 0, and hence, in particular, that
[FA ,e,3 = [& , 4 = 0. (4.1.2)
In the proof of the proposition we need a number of lemmas.

68 V. G. KAC
LEMMA 4.1.2. Let G be a finite-dimensionalZ-graded Lie superalgebra.Let

Et , Fi E G, i = 1,2, be Z-homogeneous
of nonzero degreeand Zz-homogeneous
elementsand HE G a nonxeroelementsuchthat
[H, Ei] = aiEi , [E, ,Fj] = &H, [H, FJ = -aiFi . (4.1.3)
Then a, = a2 = 0.
Proof. The subalgebra P of G generated by the Ei and Fi is clearly endowed

with a Z-grading when we set deg Ei = 1, deg Fi = - 1, deg H = 0. Suppose
that one of the ai is not zero. We consider the matrix
and set 7 = {i E { 1, 2) 1Ei is odd}. Then the contragredient Lie superalgebra

G(A, T) is infinite-dimensional. For if both a, # 0, this follows, for example,
from Theorem 3 (all the matrices of Table V are nondegenerate). But if a1 # 0,
a2 = 0, then we replace Ez by E,’ = [El , E,] and Fz by F,’ = [Fl , F,] and
arrive at the preceding case. The factor algebra PI = G(A, 7)/C, where C
is the (one-dimensional) center, is also infinite-dimensional.
Now, evidently, the map Ei + e, , Fi -+ fi induces an epimorphism of
Z-graded Lie superalgebras P ---f PI . Therefore, dim P = co, which contradicts
the fact that G is finite-dimensional.
LEMMA 4.1.3. Let G = @ Gi be a jinite-dimensionalZ-graded Lie super-

algebra, with G, semisimple.Supposethat there exist odd elementsX~ and x, ,
Z-homogeneous of nonxero degree,that are weight vectors of the adjoint repre-
sentationof Go on G, and a root vector em6of G, , linked by the relations
h ,.d = e-6, (4.1.4)

[xA , es] = [x, , e-J = 0. (4.1.5)
Then (h, 6) = 0.
Proof. Suppose that (X, 6) # 0. The same argument as in the proof of

Lemma 3.1 in [12] gives that 2(;\, 6) = (6, 6). W e ch oose a root vector ep6 such
that [eps , e,] = h, . We consider the elements
El = xA2, Fl = -(4(& S))-lb, , e612, H = h, ,

E2 = HxA2,4, e-d, F2 = @(A s)(s - 4 @(4h - 6, s))-Y[h , ed2,4, 4.
A direct calculation shows that these elements satisfy (4.1.3). Therefore, we
find from Lemma 4.1.2 that (X, 6) = 0, and we have a contradiction.
Henceforth we assume that all the conditions of Proposition 4.1.1 are satisfied
and, using Lemma 4.1.3, we look for restrictions on the weight n and the root a.
From (4.1.1), (4.1.2), and the lemma we deduce at once the next result.
LEMMA 4.1.4. (II, a) = 0.
LEMMA 4.15. Let ,8 and y be positive roots of G,, .
(a) If c4+ p is a root, then (A - /3, a + /?) = 0.

(b) Ifa+j3isar&tanda-/3isnot,then
(fl+a,fl)=O and w=-- 2(a7 B)

, (8, PI = la
(C) If a + P is a root and a - /3, a - y, /3 - y are not, then (A, y) = 0.
Proof. (a) We set x, = [e-s , F,J, x, = E,,, , 6 = a + fi. Then it follows

from (4.1.1) and (4.1.2) that relations (4.1.4) and (4.1.5) hold. By Lemma 4.1.3
we now see that (A - /3, a + Is) = 0.
(b) Suppose that ((i + a, 8) # 0. We set x, = [e+ , F,& x, = [es , EM],
6 = a. Then it follows from (4.1 .l) and (4.1.2) that relations (4.1.4) and (4.1.5)
hold. By Lemmas 4.1.3 and 4.1.4, @I - /I, a) = -(a, /?) = 0. However,
by hypothesis, (a, /?) < 0, which is a contradiction.
We have yet to show that 2(11, /I)/@, Is) = 1. We know that ((i, p) -
(a, /I) - (j?, p) = 0 and (ri, 8) + (a, /I) = 0. Adding, we see that 2(11,8) =
(p, ,9), as required.
(c) Suppose that (rl, r) # 0. Then it is easy to see that [[[[EM, e,], es],
4, PM y41 Z 0 and W n , e-J, e-,I, FJ # 0, which proves the lemma.
LEMMA 4.1.6. (a) Only one numerical murk of A is different from 0, and
that is equal to 1; in partkdar, G,, is simple.
(b) a is the h&hest root of one of the parts of the Dynkin diagram of G,
into which it is divided by the numerical mark of A.
Roof. (a) Since (/i, a) = 0, clearly there is a simple root p for which a + p
is a root, but a - B is not. If there is a simple root y # /3 for which (/1, r) # 0,
then a - y is not a root and by applying Lemma 4.1.5(c) we arrive at a con-
tradiction.
Thus, the only nonzero numerical mark of li corresponds to the simple

root /3. It is equal to 1 by Lemma 4.1.5(b).
70 V. G. KAC
(b) Suppose the contrary. Then there is a simple root /3 for which 01+ p
is a root and (A, 8) = 0. Multiplying both sides of (4.1 .l) by e+ we have
[[FA , e-J, EM] = eparp8, from which it follows that [FA , e-J # 0; therefore
(/l, /3) # 0, which is a contradiction.
We denote by s the number of the circle in the Dynkin diagram of G,, against
which the only nonzero numerical mark of fl is placed.
LEMMA 4.1.7. Either the sth circle of the Dynkin diagram of GO is at an end,
or it is joined to an end circle with the number t, and then OL= ut is a simple root.
Proof. Suppose that the sth circle is joined both to the (s - I)th and
(s + 1)th. Applying Lemma 4.1.5(c) to fi = 01~ and y = o~,-r + 01~+ OL,+~
(where oli is a simple root corresponding to the ith circle), we see that 01is a
simple root. Lemma 4.1.7 now follows from Lemma 4.1.6(b).
Conclusion of the proof of Proposition 4.1 .I. Unfortunately, I have not
succeeded in avoiding case distinctions.
Let /3 = 0~~be the unique simple root for which (A, cy,J # 0.
By Lemma 4.1.5(b):
--2(% BW, B) = 1. (4.1.6)
In accordance with Lemma 4.1.7, we consider two cases separately.

Case I. The sth circle of the Dynkin diagram is at an end. If the G,-module
G-r is isomorphic to one of the linear Lie algebras sl, with n > 2 or so, with
n > 4, n # 6, then we see evidently, by Lemma 4.1.6(b), that the local Lie
superalgebra G-i @ G, 0 Gi is isomorphic to the local part of S(n) and H(n),
respectively.
We claim that all other cases are impossible. Let t be the number of the
circle in the Dynkin diagram that is determined by the following properties:
t # s, t is an end circle and belongs to the longest of the possible “tails” of
the diagram. We denote by y the largest root for which in the decomposition
into simple roots the coefficient of CY~is zero. It is not hard to check (using
Table I) that in all cases satisfying (4.1.6), except the adjoint representation
of G, , neither cx - y nor ,B - y is a root and that (A, r) # 0. Therefore, by
Lemma 4.1.5(c), this case cannot occur. For G, we set /3’ = 01+ /3. Then
01+ ,6’ is a root; however, (/l - /Y, 01+ /3’) # 0, as is easy to see, and this
contradicts Lemma 4.1.5(a).
Case II. The sth circle of the Dynkin diagram is not at an end, but is
joined by an edge to an end circle with the number t, and 01= at . If the G,,-
module G-, is isomorphic to A2sl, , n > 3, then we see clearly that the local
Lie superalgebra G-r @ G0 @ Gi is isomorphic to P(n - 1).
We claim that all other cases are impossible. Let 0 be the highest root of G,, .
LIE SUPBFULGEBBAS 71
If (8, a,) = (6, q) = 0, then by setting y = 8 we arrive at a contradiction

to Lemma 4. I .5(c). In the remaining cases satisfying (4.1.6), except the repre-
sentation of C, with highest weight 0 - at , we denote by y the largest root
for which in the decomposition into simple roots the coefficient of cq is zero,
and we again use Lemma 4.1.5(c). In the case of C, we set /3’ = 8 - 2ar, and
arrive at a contradiction to Lemma 4.1.5(a).
Thus, the local part of G is isomorphic to the local part of one of S(n), H(n),
or P(n). The isomorphism of the Z-graded Lie superalgebras themselves now
follows from Propositions 1.2.3(c), 3.3.1(e), and 3.3.6(d).
This completes the proof of Proposition 4.1.1.
4.1.2. Class$cation of Z-g-m&d Lie supera&ebras of depth 1.
We now describe two constructions of transitive Z-graded Lie superalgebras.
Every Z-graded Lie superalgebra G = @ Gi can be extended by means of
an even derivation x defined by
[2,x] = hx for XE Gk.
So we obtain a Z-graded Lie superalgebra, which we denote by Ge = @ G/,

where G,* = G$ for i # 0 and G,* = G, @ (a). If G,, is transitive and the
center of G,, is trivial, then clearly Gz is also transitive.
The other construction goes as follows. Let H be a Lie algebra without
center. On it we construct a Lie superalgebra He = G-, @ G, @ GI with a
consistent Z-grading, by setting G-r = fH, G,, = H, GI = (d/d#, where the
commutators are defined as follows: [d/df, .fh] = h, [IhI , IIcd] = e[h, , hJ,
[d/df, h] = 0. Evidently, He is transitive.
Now we are in a position to state the main theorem of this section.
THEOREM 4. A transitive irreducible Lie superalgebra G = &a-1 Gi with a

consistent Z-grading and GI # 0 is isomorphic as Z-graded superalgebra to one
of thefOllawi~ list:
I. A@, 4, C(n), P@>;
II. W(4, SW, H(4, Ri(n);
III. He, where H is a simple Lie algebra;
IV. Gz, where G is of type I, II, 01 III mrd the center of G, is tkiaZ.
Proof. Since the representation of G, on G-r is faithful and irreducible,
Gs = G,’ @ C, where Go is semisimple, C is the center of G, , dim C < 1,
and if dim C = 1, then C = (I), with [z, g] = kg for g E Gk (see Proposi-
tion 1.2.12). Therefore, the representation of G,, on G, is completely reducible;
let
Gl .= @ G,‘“’ (4.1.7)
72 V. G. KAC
be the decomposition of G1 into G,,-irreducible components. We denote by

G(@ the Z-graded subalgebra of G:
G’“’ = G-, @ [G-, , G:“] @ G;“’ @ (G:“‘)’ @ -0. .
If we consider in G the Z-graded subalgebra G-, @ G, @ Gy) 0 (Gp’)2 @ ‘.‘,

we can infer from Proposition 1.2.9 that [G-, , Gp’] C G,‘. Therefore, G($)
is a Z-graded Lie superalgebra satisfying all the conditions of Theorem 4.
There are two possibilities.
(1) The representations of [G-, , Gy’] on G-, and Gp’ are contragredient.
According to Proposition 1.2.10(a), G(S) = G-, @ [G-, , Gp’] @ Gp’ is
classical. Propositions 2.3.9 and 2.4.4 now show that G(@ is isomorphic as
Z-graded algebra to one of A(m, n) or C(n).
(2) The representations of [G-, , Gp’] on G-, and Gp’ are not con-
tragredient. According to Proposition 1.2.10(b), [G-, , Gp’] = G,,’ is then
simple. If the representation of [G-, , Gp’] on Gp) is not faithful, that is,
dim Gp) = 1, then, as is easy to see, GcS) = G-, @ [G-, , Gp’] @ Gp’ is
isomorphic as Z-graded algebra to (G,,‘)E. If the representation of G,’ on Gy’
is faithful, then according to Proposition 4.1.1 (G(@ is bitransitive, by Proposi-
tion 1.2.13), G(S) is isomorphic as Z-graded Lie superalgebra to one of S(n),
H(n), or P(n).
Thus, when G1 = G:S’ is an irreducible Go-module, then Lemma 4.1.8
below and Propositions 3.3.1 and 3.3.6 show that G is one of A(m, n), C(n),
S(4, I-W), R(n), P(n), H', or of type IV. (In Lemma 4.1.8 the case A(1, 1)
is excluded; however, A(l, 1) N H(4).)
We claim that in (4.1.7) all the G,-modules are pairwise inequivalent. Let
FA be the highest weight vector of the Go-module G-, , and EM, the lowest
weight vector of Gp). Suppose that in (4.1.7) there are two isomorphic G,-
modules, say Gil’ and Gf’. If they are contragredient to G-, , then by Proposi-
tion 2.1.6, G(1) and Gt2) are isomorphic. Therefore, the vectors [FA , EMI]
and [FA , E,J are proportional; consequently, [FA , EMI - cEMJ = 0 for some
c E k. But then, clearly, [G-, , EM, - cEMt] = 0, which contradicts the fact
that G is transitive. If Gy’ and Gr’ are not contragredient to G-, , then [FA , EMJ
are root vectors of G,,’ corresponding to one and the same root and are, therefore,
proportional. Again, this contradicts the transitivity of G.
When we now compare the possibilities for G@) obtained above, we see that
the G,,-module G1 can be reducible only if the Go-module Gel is isomorphic
to gl, or the G,,‘-module G-, to A%l, E so6 . In the first case, Propositions
3.3.2 and 3.1.1(e) show that G is isomorphic to W(n). In the second case it
follows from Proposition 3.3.6(j) that G is isomorphic to one of H(6), R(6),
H(6)“, or R(6)“.
The proof of Theorem 4 is now complete.
4.1.3. on fxtera&ms of someLie 2Vsper&e&rias.
LEMMA 4.1.8. Let G = &am1 Gi be a transitive Z-graded Lie superalgebra

and G’ = G-, @ Go @ Gl imm#ic to one of
(4 H?
(b) A@, 4 for Cm, 4 f (1, 1) 0~ C(n);
(4 P(n).
Then G = G’.
Proof. In all three cases we have to show that Ga = 0.

(a) We recall that Ht = fH @ H @ (d/d&. Suppose that there is a
t E G, , t # 0. Then [t, fa] = a(a)(d/df), where a is a nonzero linear function
on H (because of transitivity). Therefore, 0 = [[t, [a], ta] = a(a) for every
a E H, which is impossible.
(b) Suppose the contrary. Then there is a nonzero weight vector tA
of the representation of G, on Gs . By the transitivity of G, there is a root
vector e-s E G-, for which
[tA , e-J = e,-, E Gl . (4.1.8)
By Proposition 2.5.5(e) there is a root vector e--h+8E G-1 for which [e-A+\+s, e,+.J =
hAPB. Multiplying both sides of (4.1.8) by eml+s , we have
[[e-A+8, 61,e-81= LI . (4.1.9)
Hence it follows that [e-A+s , tJ = e, , so that (4.1.9) showsahat h,+# is propor-

tional to hO ; and then h = cj? (also by Proposition 2.5.5(e)). Moreover, as we
have seen, X - p = (c - 1)/3 is a nonzero root of G’. However, as is clear
from Section 2.5.4, multiples of p can only be the roots 0 and -8, that is,
c = 0 or 2. In the first case we see that dim Ga > 2, which contradicts Proposi-
tion 2.5.5(b). Thus, any weight of the representation of G, on Ga is equal
to twice a root of G,, . But clearly this is impossible.
(c) We recall that the G,-module G-, is isomorphic to Aas& or Spsl, ,
and the Go-module G1 to S%l,* or Aas&,*. Let A*,, be the highest and Ma
the lowest weights of G*, . By transitivity, Ga is a G,,-submodule of G1 @ Gzl .
(A mapping #: G, + G1 @ G1; = Hom(Gwl, G,) can be constructed in the
obvious fashion: #(g)(a) = [g, a], g E Ga , a E Gml .) Thus, Ga is a Ga-sub-
module of A%l~+, @ S%l~+, . This module splits into two irreducible com-
ponents with the lowest weights M(l) = -A-, + MI and Mta’ = -A-, +
MI + q + as, where 4 and or, are the first two simple roots of the Lie algebra
A A.
74 V. G. KAC
We now assume that one of the weight vectors E,,,(i) is contained in G, .

By transitivity we then have, respectively,
[F/I-~, E,wl = hI , (4.1.10)
[FL, 3E,ml = Lb1 >eu,+aJ. (4.1.11)
Now (4.1.lO) also gives
[FAel , ~~1, E,d = [h+ , e,J. (4.1.12)
Next, [FA-, , EMI] = ewe1 (or emI). Therefore, from (4.1.10):
0 = [Eel I E,+l = WAel , &I = e+ .

This is a contradiction of EMu, E G, . Similarly, we find from (4.1 .l 1) (or
(4.1.12)) that
0 = F’?t-,, EMcad= eua (respectively,0 = [[FAel, c,J, EM(,)1

= OWL, j +I, [-CW~
l e,,ll = e,J;
hence EMczj $ G, . This proves the lemma.
4.2. The Classi$cationof the SimpleLie Superalgebras
4.2.1. The main theorem. The following theorem is the central result of
the paper.
THEOREM 5. A%mple finite-dimensionalLie superalgebraover an algebraically

closedjeld k of characteristic0 is isomorphiceither to oneof the simpleLie algebras
or to one of the Lie superalgebras A(m, n), B(m, n), C(n), D(m, n), D(2, 1; ol),
F(4), G(3), P(n), Q(n),W(n), S(n), S(n), or H(n).
Proof. Let L = L, @ Li be a simple finite-dimensional Lie superalgebra
over k. If the representation of L6 on Li is irreducible, then L is classical.
Therefore, by Theorem 2, L is isomorphic to one of B(m, n), D(m, n), F(4),
G(3), Q(n),or D(2, 1; 4
Suppose now that the representation of L6 on Li is reducible. Then, by
Proposition 1.3.2, L has a filtration: L = L-, 1 L, 1 L, 1 ..* for which the
associated Z-graded Lie superalgebra Gr L = @ Gri L satisfies all the condi-
tions of Theorem 4. Therefore, GrL can only be isomorphic to one of the Lie
superalgebras of type I-IV listed there.
From the proof of Proposition 2.2.2, it is clear that if Gr L ‘v Ht, then L
is not simple; hence type III does not occur.
Proposition 1.3.1 shows that if the center of Gr, L is nontrivial, then L N Gr L.

Hence, type IV does not occur either, because clearly rio superalgebra of this
type is simple; also L N W(n) if Gr L N W(n).
If Gr L N A(m, n), C(n), or P(tr), then evidently, the representation of L,
on Li is for L the same as for Gr L. Therefore, Proposition 2.1.4 shows that
L 31 GrL, so that L is one of A(m, n), C(n), or P(n).
If Gr L N H(n) or a(n), then L N GrL, by Proposition 3.3.7. But fi(tz)
is not simple, so that this case is impossible, and L is one of the H(n).
Finally, if GrL N S(n), then by Proposition 3.3.5, L is isomorphic to an
S(n) or S(n).
This completes the proof of the theorem.
4.2.2. I~~@~MJs. It is not &d to list all the isomorphisms between
simple Lie superalgebras. They are: A(m, n) N A(n, m); A(1, 0) N C(2) rr W(2);
A(l, 1) N H(4); P(2) N S(3).
In the remaining cases, except for D(2, 1; (u), S(n), and s(n), simple Lie
superalgebras are pairwise nonisomorphic, because for them the LG-modules Lx
are nonisomorphic. S(R) and s( n ) are also nonisomorphic, according to Proposi-
tion 3.3.4. Conditions for isomorphisms of superalgebras in the family D(2, 1; a)
were derived in Proposition 2.5.4(b).
The following is a list of the dimensions of all the simple Lie superalgebras.
Ah 4 (m + n + 7 - 1 - a,,, P(n) 2(n + l)e - 1

B(m, 4 2(m + n)8 + m + 3n Q(n) 2(n+1)2-2
C(n) 2n*+n-2 W(n) n * 2”
D(m, 4 2(m + n)” - m + n s(n) (n - 1)2” + 1
D(2, 1; a) 17 w4 (n - 1)2n + 1
F(4), G(3) 40,31 H(n) 2” - 2
4.2.3. CEassijication of jinite-die prim&be Lie super&eliras. La L

be a Lie superalgebra and Lo be a distinguished subalgebra. The pair (L,L,,)
is called even primitive if L, is a maximal proper subalgebra, it contains no
ideals of L, and Lo3 La.
The same arguments as in the proof of Theorem 5 give a classification of
the primitive even pairs (clearly, they are all finite-dimensiomil automatically,
because Gr L is embedded in W(n), where n = dim L/L,). To state and prove
the result, we need only make the following remarks.
As it is easy to see, if L is a Lie superalgebra with a filtration for which
Gr L N Hc, then either L N He or L N der Q(n) = Q(n) @ D is a semidirect
sum of Lie superalgebras, D being, up to a constant factor, the only odd outer
derivation of Q(n) (see Proposition 5.1.2(c)).
To an irreducible faithful representation of a Lie algebra H in a space V
76 v. G. KAC
there corresponds the primitive Lie superalgebra Hv = V @ H, where

[V, V] = 0, [h, V] = h(o) for h E H, v E V, and H,” = H, Hi’ = V.
THEOREM 5’. Let (L, Lo) be a primitive evenpair. Then L is isomorphicto a

Lie superalgebrain the following list:
I. A@, 4, C(n), P(n);

II. B(m, 4, D(m, n), P(4), G(3), Q(n), der Q(n), D(2, 1; 4;
111. W(4 S(n), @I, H(n), fi(n);
IV. H’, whereH is a simpleLie algebra;
V. G”, where G is one of the Z-graded Lie superalgebrasA(n, n), P(n),
S(n), H(n), R(n), or He.
VI. Hv, where H is a Lie algebra and V a faithful irreducible H-module.
Each of theseLie superalgebras,

exceptP(n) and P”(n), admitsa unique structure
of a primitive evenpair. There are two suchstructuresfor P(n) and P”(n).
Note that Theorem 5’ also gives a classification of the primitive transitive

supergroups of transformations of a supermanifold whose stabilizer contains
a maximal reduced subgroup.
5. CONTINUATION OF THE THEORY
5.1. Descriptionof Semisimple

Lie Superalgebras
in Termsof Simple Ones
5.1.1. DEFINITION. Let A = A, @ Ai be a superalgebra, der A the Lie

super-algebra of its derivations, and L a subset of der A. Then A is said to be
L-simpleif A contains no nontrivial ideals that are invariant under all the deriva-
tions in L. If A is der A-simple and A2 # 0, then A is called d@rentiabZy
simple.
We define operators 1, and rs, s E A, on A by the formulas
Z,(a) = sa, r,(a) = (-l)(dega)(degs)as, aEA.
It is easy to verify that if D E der A, then
P, rsl = r,(s) .
We denote by T(A) the associative subalgebra of l(A) (all the endomorphisms
of A) generated by all the 1, and Y,?, s E A.
LIE SUpERALGEBRAS 77
Finally, the centroid of A is the associative superalgebra r(A) = {g EL(A) 1

[g, h] = 0, h E T(A)). A is called central if P(A) = k.
51.2. LX#erentiably sainpk snwakebras. A verbatim repetition of the
arguments in [21], with the relevant definition replaced by those above, leads
to the following result.
PROPOSITION 5.1.1. Let G be a fmite-dimen&mal diffeentiably simpk (not

necessarily Lie) superakebra. Then G N S @ A(n), where S is a simpk and
A(n) is the Gnzwmam nsperalgebra.
51.3. Description of semisimpleLie superalgebras.We recall that a super-
algebra A is said to be semisimpleif A2 # 0 and A contains no nontrivial
solvable ideals. In [21] the description of differentiably simple algebras is
used to derive a description of semisimple Lie algebrasover any field. The
samearguments are suitable for Lie superalgebras.
THEOREM 6. Let S, ,..., S, be jinite-dime&anal Lie superalgebras,n, ,,.., n,

be nonnegativeintegers,and S = @I-, S, @ A&). T%en
S = inder S = 6 (inder S,) @IIl(n,) _Cder S

i-l
= i ((der W 0 44 + 1 0 der 44).
Let L be a subalgebraof der S containing S; we denoteby L, the set of com-

ponents of ekmentsof L in 1 @ der A(q). Then:
(a) L is semkimpkif and only if A(nt) is L&mpk for all i.
(b) All finite-dimen&nal semkimpkLie supera&ebrasarise in the manner
indicated.
(c) der L is the normalizer of L in der S, prov2ed that L is semisimple.
5.1.4. Descrt$tionof der G for the simpleLie superakebrasG.
PROPOSITION 51.2. Let G be a simpk Lie superalgebra,der G be the Lie

superalgebraof its derivatiuns, and inder G (E-G) be the i&al of der G consisting
of the inner derivations.
(a) If G is one of the classicalLie superalgebrasA(m, n) with m # n,

B(m, n), CM W, 4, F(4), G(3), or one of Ha&Lie superajgebras of Carts
type W(n) or s(n), then der G = inder G.
@) If G=OG is one of the Z-graa%dLie superalgebras
A(n, n) with
78 V. G. KAC
n > 1, S(n), or P(n), then der G = inder G @ (z) is a semidirect sum, where
z is an even derivation of G such that [z, g] = kg for g E Gk .
(c) If G N Q(n) = Q(n)6 @ Q(n)i , then der G = inder G @ (D) is a
semidirect sum, where D is the (within proportionality unique) endomorph&m of
Q(n) for which D(Q(n),) = 0, D(Q(n)i) = Q(n),=, , and D: Q(n)i -+ Q(n)a is an
isomorphism of Q(n)-,-modules.
(d) If G N H(n), n > 5, then der G = inder G @ T, where
is a two-dimensional solvable Lie superalgebra.

(e) If GNA(l,l)=G-l@G,@G,, thenderG=inderG@Pisa
semidirect sum, where P = (Oh1 , z, DJ is a three-dimensional simple Lie algebra
[z, g] = kg for k E G, , D*, are the (up to a constant factor unique) endomorphisms
of G for which D+(G,) = 0, D&G*,) = 0, D*,(G,,) = G*, , and D+,: G,, -+
Gk, are isomorphisms of G,-modules.
Proof. G N inder G C der G is an ideal of der G. Let G, be the reductive

part of Gfi . Since the Go-module der G is completely reducible, we have that
der G = inder G @ T is a direct sum of G,,-modules and T is a Z2-graded
subspace; in particular, [G, , T] C T. On the other hand, [G, , T] C inder G,
because inder G is an ideal of der G. Therefore, [Go, T] = 0 and if D E T,
then ad D is an endomorphism of the G,,-module der G. Using this fact it is
now easy to compute T in all cases.
5.2. Irreducible Finite-Dimensional Representations of Solvable and Simple Lie

Superalgebras
5.2.1. Induced modules. Let G be a Lie superalgebra, U(G) its universal

enveloping superalgebra (see Section 1 .1.3), H a subalgebra of G, and V an
H-module. I’ can be extended to a U(H)-module. We consider the Z2-graded
space U(G) @u(H) V (this is the factor space of U(G) @ V by the linear span
of the elements of the form gh @ v - g @h(v), g E U(G), h E U(H)). This
space can be endowed with the structure of a G-module as follows: g(u @J v) =
gu @ v, g E G, u E U(G), v E V. The so constructed G-module is said to be
induced from the H-module V and is denoted by Ind,G V.
We list some of the simplest properties of induced modules, which follow
from the PoincarbBirkhoff-Witt theorem (see Section 1 .I .3).
PROPOSITION 5.2.1. (a) Let G be a Lie superalgebra, H be a subalgebra,

V be a simple G-module, and W be an H-submodule of V considered as an H-module.
Then V is a factor module of the G-module Ind,G W.
LIE SUPEBALGEBBAS 79
(b) If H, CHIC G are s&a&&as of G and W an H,-moduk, then

Indg,(Ind,“: W) N Ind$, W.
(c) Let H C G be a s&a&bra of G containing G , andg, ,..., g, oddekments
of G whose projections onto G/H form a basis. Let W be an H-module. Then
InhG W = Ol<t,<...<t,~t gt, *-* gi, W is a direct sum of subspaces; in pmttilar,
dim Ind,G W = 2: dim W.
The next result follows from Proposition 5.2.1(c) and Ado’s theorem for
Lie algebras.
ADO’S THEOREM. Every finite-di me&ma1 Lie superalgebra has a Jinite-

dimensional faithful representation.
5.2.2. Representations of solvabk Lk superalgebras. Let G = Gs @ Gi be a
Lie superalgebra. A linear form I E G* is said to be distinguished if l([G, , q) =
l(Gi) = 0. We denote by 9 the space of distinguished linear forms, by 9s
the subspace consisting of those I for which Z([G, q) = Z(G,) = 0, and by
9r the subgroup of 9s generated by the linear forms given by the one-
dimensional factors of the adjoint representation of G.
Let p be a representation of G in a space V, .M a subgroup of 9s , and /\ E A.
We define a representation p of G in V by the formula p(g)w = p(g)w + h(g)w
(i.e., p” is a tensor product of p and a one-dimensional representation). The
G-modules p and p” are said to be ~-equivaknt.
LEMMA 5.2.2. Let G be a Lie superalgebra, H be a subalgebra of codimension 1

contain&gG-,,andgbeanoo!dekmentforwhichG=HQ(g)isadirectnsm
of subspaces.
(a) If W is an irreducible H-module, ttren all the irreducible factors of the
H-module IndHG W = W Q g W are &equivaknt to W (.J& C H*).
(b) If V is an irreducible Emoduk and W an irreducible H-s&module
with W # V, then V N Ind,o W.
Proof. (a) For h E H we have [h, g] = h(h)g + h’, where h’ E H, X E SO.

Therefore, h(gv) = g(her + X(h)v) + h’v; hence, the H-modules W and
IndHG W/W are go-equivalent.
(b) follows from (a) and Proposition 5.2.1.
Let I E 9 be a distinguished linear form, considered modulo 9s ; we set

G, = (g E G 1Z([g, gJ) = 0 for g, E G}. Clearly, G, is a subalgebra of G con-
taining Ga , and Z([G, , GJ) = 0. A subalgebra PC G is said to be subordinate
to I if Z([P, P]) = 0 and G, C P. Clearly, this concept is well defined.
We single out an important class of solvable Lie superalgebras-the com-
pletely solvable ones-for which all irreducible factors of the adjoint representa-
80 V. G. KAC
tion are one-dimensional. By Engel’s theorem, a nilpotent Lie superalgebra

is completely solvable, and 6p = 0.
Finally, we denote by {H, Z} the one-dimensional H-module given by a
linear form 1~ 9s according to formula h(w) = I(h)w.
Now we are in a position to state a theorem that describes the finite-
dimensionalirreducible representationsof solvable Lie superalgebras.
THEOREM 7. Let G = G6 @ Gi be a solvable Lie superalgebra.

(a) If V is an irreducible$nite-dimensional G-module, then all the irreducible
factors of V considered as a GO-module are one-dimensional, and their corresponding
linear forms, extended by zero to Gi , lie in a single coset & E 2Z’/& .
(b) Let I E 9tp12Z0 , P be a maximal subalgebra subordinate to I, and {P, 1)
be the one-dimensional P-s&nodule given by the linear fm 1 E 1. Then the G-
module V = Ind,c(P, 1) is Jinite-dimensional and irreducible, and 1 = iv . Two
such G-modules VI and V, are 90-equivalent sf and only af II = l2 .
(c) Every jinite-dimensional irreducible G-module V is isomorphic to one
of the modules IndpG(G, I), where I E &, and P is a maximal subalgebra subordinate
to 1.
(d) If G is completely solvable, then 9o can be replaced everywhere by 5.X1 .
In particular, if G is nilpotent, we obtain a bijective correspondence V H 1,
between the set of classes of isomorphic finite-dimensional irreducible G-modules
and 9.
Proof. (a) is proved by induction on dim G. Let V be an irreducible

G-module, H be a subalgebra of codimension 1 containing [G, G], G =
H@(g), where gEG,, SE&, and W be an irreducible submodule of an
H-module V. Clearly, then V = &OgiW and since H is an ideal of G, the
proof of Lemma 5.2.2(a) shows that all the irreducible factors of V are
isomorphic to W. Therefore, if g E Gi , then V and W, as G6-modules,have
the samestock of irreducible factors, and (a) is true by induction. But if g E Gr, ,
then GG= Ho @ (g), and only as Hrmodules do V and W have the same
stock of irreducible factors. But then it follows that if I1 , 1sE G* are linear
forms that are zero on Gi and give irreducible factors of the G,-module V;
then by the inductive hypothesis, lr - 1, ]rxi,xil = 0, in particular,
II - Is j[ci,oil = 0. Since, of course,I1 - 1sI[og,o;] = 0 we seethat I1 - 1, E 9s)
as required.
(c) is proved by induction on dim Gi . Let V be a finite-dimensional
irreducible G-module and P be a maximal subalgebra subordinate to IV.
If dim Gi = 0, then (c) is true by Lie’s theorem. We may, therefore, assume
that dim Gi > 0. We analyze first the caseG = P, that is, lr, = 0. By induction
on dim Gi we showthat dim V = 1. Let G’ be a subalgebraof G, of codimension
1, containing G, , and G = G’ @ (g). By the inductive hypothesis, the
G-module V contains a one-dimensional submodule (v). Suppose that

dim V > .I; then, evidently, V = (0) @ (gw). Now g(gu) = @g,g)u =
+Z([g,g])u = 0. Since V is irreducible, there is an element h E Q for which
h(gw) = u. Replacing h by h + cg for a suitable c E k, we may assumethat
A(V) = 0. But then [k, g]u = u and Z([Zr,g]) = 1, which contradicts our
assumption.
Suppose now that G # P. Then G has a subalgebraH of codimension I
containing P, and G = H @ (g), g E @ . Let W be an irreducible submodule
of the H-module V. By the inductive hypothesis, W = IndpH{P, 1). If H = P,
then, as we have shown, W = (u) is a one-dimensionalH-module. In that
case V # W becauseG # P, and so V = Indpc{P, Z}, by Lemma 5.2.2(b)
and Proposition 5.2.1(b).
Suppose now that H # P. Then Z([h, h]) # 0 for some h EHi, and the
quadratic equation in (Y,
4rg + oJI*g + 4) = 4[g, gl) + 24[g, h]) + a2Z([h,

h]) = 0,
has a root 06 . Replacing g by g + a& we may assumethat Z([g, g]) = 0.
What we have to show (according to Proposition 5.2.1(b)) is that the G-
module Ind,c W = W + gW is irreducible. Suppose the contrary, i.e., that
it contains a nonzero irreducible G-submodule W’. By Lemma 5.2.2(a), the
G-modules W and w’ are 9+xluivalent. Hence, in particular, there is a one-
dimensional P-submodule {P, 4) = (q + gu2) E W, where Z- Z,E J&. Now
h(vl + p,) = 4(h)(vl + p2) = h(q) + [h, glv2 + ghv, for h E Ga . Since
[k, g] = c(R)g + h’, h’ E H, we seethat h(v,) = (k(h) - c(k))u, . In particular,
(u,) is a one-dimensional submodule of the G-module W; hence by (a),
h(u2) = Z,(k)u, for h E 6, where Z,- ZsE 9s . Therefore, in particular,
z2([g,gl) =, 0. If v1 # 0, theng(q + p.J = pl + H([g, gls = get,# 0; but
if ul = 0, then gn, # 0. Thus, W’ ngW # 0. It therefore follows from
Lemma 5.2.2(a) that W’ = gW. But then k(gw) = Z([h,g])u + ghu for h E Hi .
Since gW is a G-module, we infer that Z([Zz,g]) = 0 for h E Hi , and since,
furthermore, Z([g, g]) = 0, it follows that I(@ , g]) = 0, that is, g E G8C P.
This contradicts the choice of g.
(b) evidently follows from (a) and (c), and (d) follows from the fact that if G
is a completely solvable Lie superalgebra, then all the irreducible factors of
the G-module U(G) are one-dimensionaland are given by linear forms in T1.
This completesthe proof of the theorem.
The following propositions are consequencesof Theorem 7.
PROPOSITION 5.2.3. For an irreduciblefinite-dimemicmalrepresentationof a

solvableLie supera&ebraG = Gs@Gi in a space V = Vo@ Vi we have:
Either dim V, = dim Vi and dim V = 2”, wheres < dim Gi , or dim V = 1.
82 V. G. KAC
PROPOSITION 5.2.4. All the irreducible finite-dimensional representations of a

solvable Lie superalgebra G = G5 @ Gi are one-dimensional ;f and only sf
[Gi , GL] C [G, , Go].
EXAMPLE 1. It follows from Theorem 7 that the families of representations

pa and pa’ of the Heisenberg superalgebrasN and N’, which were constructed
in Section 1.1.6, contain all their finite-dimensional non-one-dimensional
irreducible representations,and each precisely once.
EXAMPLE 2. Let G = E(1, 1) be the completely solvable Lie superalgebra

with the basisz = (t i), h = (i z), e = (z ‘,), f = (y i). The set of representa-
tions of dimension>1 is parameterizedby the numbers 01= Z(h), /3 = Z(z) # 0:
For /3 = 0 we obtain all the one-dimensionalrepresentationsh w 01,z, e, f w 0.

The proof of Theorem 7 also works for infinite-dimensional representations
of completely solvable Lie superalgebras. Two representations are called
weakly equivalent if they have the samekernel in U(G). We set 2 = {ZE G* 1
Z(Gi) = O}. Let 3s be the Zariski closure of Aut Gs in GL(G6).
It is known (Dixmier) that there is a bijection between the set of gfi-orbits
in Gr,* and the set of classesof weakly equivalent representationsof G, .
THEOREM 7’. (a) If G is a completely solvable Lie superalgebra and V is an

irreducible G-module, then there is an uncondensed Jordan-Holder series relative
to G6 and all its irreducible factors correspond to a singte 9,,-orbit Q, in ~?,J9r .
(b) Let 1 E p/9I
, P be a maximal subalgebra subordinate to 1, and
dim P0 = $(dim Ga + dim(G,)&. Then the G-module V, = Ind,G(P, Z} is
irreducible. The correspondence l++ V, induces a bijective correspondence between
the set of classes of 91-weakly equivalent irreducible representations of G and
the set of go-orbits in p/Z1 . Here Qs,, = S,, . 1.
5.2.3. Representations of simple Lie superalgebras.
PROPOSITION 5.2.5. Let G = @i2-d Gi be a Z-graded Lie superalgebra of

depth d, N* = @6>0 Grti , Bk = Nh @ B, , where B, is a Bore1 subalgebra of
(G& . Suppose that
[I%*, &*I C [Bo*-, Boil, Po v(G&d = (%>a 2 and G,, generates Nk.
(a) Let V be a finite-dimensional irreducible G-module; zLe set V, =

{v E V 1 N+(v) = O}. Then V,, is an irreducible submodule of the G,,-module V.
(b) Two j%iitedistumio& G-modules V and y’ are isomr~hic if atzd only
;f the correspodwg Go-mod&s V, rmd V,l are immqhic.
(c) If the aGpth d = 1, then for any jinite-dimmional imducible Go-
mod& V,’ there is a J;nite-di men&ml +educibk Gmoduk V for which tke
G,-mduk V, is &mor#ic to V;.
Proof. Since B+ is a solvable Lie superalgebra, it follows by Proposition
5.2.4 that any irreducible factor of the B+-module V is one-dimensional.
Using the properties of iV+, we hence find that the N+-module V is nilpotent.
In particular, V,, # 0.xThe same is true for N-.
Now it only remains to go through the arguments in [7j almost verbatim.
First, we show that W = V,, A G-,V is equal to 0. Let U’(G-J be the sub-
algebra of the enveloping superalgebra U(G) generated by G-r and U(G-,) =
U’(Gbl) @ (1). Since U’(G-,) is nilpotent, U(G+)WC U’(G-,)V # V.
U(G-JW is a G-submodule of V, and since V is a simple G-module,
U(G-,)W = 0; in particular, W = 0.
We set V, = G-,V$,_, for i > 0. Clearly, V is the sum of the subspaces V, .
We show by induction that this sum is direct. Suppose the contrary, i.e., that
w E vwa,, r\ \@& V,), w # 0. Then G,w # 0 because W = 0. But G,v C
vm n cozl VA which is impossible. Thus, V = @r>o Vi . From this it
follows, obviously, that V, is an irreducible G,,-module.
(b) Let V be a finite-dimensional irreducible G-module; then V N
IndgO V,/l, where V, is an irreducible Gs-submodule and I is a maximal
submodule of the G-module Ind& V, . Since V = &a0 V, , this I is uniquely
determined as the sum of all graded submodules of the G-module IndzO V, ,
and this proves (b).
(c) We define an action of G, @ N+ on V,‘, setting N+V,‘ = 0. Since
d = I, the induced G-module Ind& V, is finite-dimensional (see Proposition
5.2.1(c)). The required G-module is a factor module of this G-module.
We apply Proposition 5.2.5 to the following Z-graded Lie superalgebras
G = @ Gi: (a) P(n), W(n), S(n), H(n) with the “standard” Z-grading; (b) Q(n)
with the Z-grading in Section 2.5.7 with 4 = .*. = k, = 1; (c) the con-
tragredient Lie superalgebras with “standard” Z-grading (from Section 2.5.7
with kI = .*a = k, = 1).
In cases (b) and (c) we set H = (G,), = (h, ,..., h,), N+ = @lx Gi,
B = H@N+. Let AEH*, ai = A(h,) E k, (on) be a one-dimensional B-
module for which N+(er,) = 0, hc(wA)= arwA. We set VA = Ind,G(wA)/lA,
where I,, is the (unique) maximal submodule of the G-module VA . A is called
the highest weight of the G-module VA. It follows from Proposition 5.2.5(b)
that the G-modules VA, and VA, are isomorphic if and only if A1 = As .
Numbers a, = d(h,) are called the nunrerriczZ ~mks of A.
We let Z, denote the set of nonnegative integers.
84 V. G. KAC
THEOREM 8. (a) Let G = @ Gi be one of the following Z-graded Lie super-

algebras of depth 1: P(n), W(n), S(n), H(n). Then the correspondence in Proposi-
tion 5.2.5 between finite-dimensional irreducible G-modules and finite-dimensional
irreducible Go-modules is bijective.
(b) For Q(n) the set of numerical marks of the highest weight of the jinite-
dimensional module VA is characterized by the following conditions: a, EZ+ and sf
q=O, thena,+2a,+~~~+(i-l)ai~l=a,+2a,~,+~~~+(n-i)a,+l.
(c) For the contragredient Lie superalgebras in Table VI (the sth YOW of
the Cartan matrix is normalized so that a,,+1 = 1 for ass = 0), the set {ai} of
numerical marks of the highest weight of the finite-dimensional module VA is
characterized by the following properties:
(1) ai EZ, for i f s;

(2) kEZ+, where k is given by the following table:
G k b
JW, n) 1Pa, 0
B(m, n), m > 0 a, - ant1 - ... - am+n-l - lPa,+, m
D(m, 4 a, - a,,, - ... - am+n-2 - ~/wn+,-l + %+n) m
D(Z 1; 4 (1 + oI)-1(2a, - u2 - olus) 2
F(4) 1/3(2a, - 3a, - 4a, - 2a,) 4
G(3) 1/2(a, - 2a, - 3a,) 3
(3) for k < b (in the table) there are the supplementary conditions:
B(m, n): a ntktl -- ... =a w-n = 0.
D(m, n): a niki1 -- ... = a mtn = 0, k < m - 2 arntn-l = am+,, ,
k=m-1.
D(2, 1; a): alla~=Ozfk=O;(a,+l)or=~(a~+1)zfk=l.
F(4): alla,=Oifk=O;k#1;a,=a,=Oifk=2;
a,=2a,+lifk=3.
G(3): allai=Oifk=O;kfl;a,==Oifk-2.
Proof. (a) follows from Proposition 5.2.5.

To prove (b) and (c), we let f10 denote the system of simple roots of the
(reductive) Lie algebra Gs determined by the induced Z-grading. In order
for the G-module VA to be finite-dimensional it is necessary and sufficient that
eTT1vA = 0 for ag4, where c = 2(A, a)/((~, a) (52.1)
(it is well known that these relations generate the annihilator of vA).
(b) If G 1 Q(n), then 17, = (01~ ,..., oln}, where etii = ei ; hence ai EZ, _
It is easy to see that the supplementary conditions only arise when ai = 0:
LIE SVPERALGEBRAS 85
This condition gives the equality 8*fpA = 0, which is equivalent to the second
equality in (b).
(c) If G = A( m, n) or C(n), then Q, = {ar, , i # s}, and condition
(5.2.1) is equivalent to condition (1). For the remaining contragredient Lie
superalgebraa, n, = {ari , i # s, /?}, where p is the maximal root among the
roots of the form xi>* &a, . By the same token, condition (52.1) shows that
conditions (1) and (2) are necessary. It is also clear that (1) is sufficient for
(5.2.1) when a = ai. However, condition (2) turns out not to be sufficient
for (5.2.1) when a = 8. When a = /I, using direct computations from (5.2.1),
we can show that condition (3) is necessary. It is also not hard to verify that
(1) and (2) are sufficient for (5.2.1), when G = B(0, a).
It remains to show that conditions (2) and (3) are sufficient for (5.2.1) when
a = #?. To do this, it suffices to find a set of highest weights A of finite-
dimensional modules I’, with the property 2(A, /Q/(/3, B) Q K which generates
the plane defined by the equations in (3). It is clear that for B(m, n) and D(m, a)
without loss of generality we may assume that II = 1; then for B(m, 1) and
D(m, 1) the desired set is the exterior powers of the standard representation.
For D(2, 1; a), F(4), and G(3), we must take the exterior powers of the adjoint
representation; then we need only verify that (3) is sufficient for (5.2.1) when
a =/I, c = 2.
The theorem is proved.
Let G be a simple finite-dimensional contragradient Lie superalgebra, and
let ( , ) be an invariant nondegenerate bilinear form on G. We let p denote
the difference between the half-sums of the positive even roots and the positive
odd roots. It is not hard to show that p(k,J = (a(, aq)/2. We define the Casimir
operator in the center of the enveloping superalgebra by the formula: r =
C (- l)de*uii(iu~, where {ai> and {u’> are dual bases of G relative to the form ( , ).
Let V be a finite-dimensional irreducible G-module with highest weight A.
The action of r on V can be written in the form: r(wJ = (A, h + 2p)o?, +
c ax e-da . In particular, I’@,) 4 (A, A + 2p)vA, and, by Schur’s lemma,
r is a scalar operator; hence, r(w) = (A, A + 2.p)v, v E I’. We define the
supertruce form in the usual way: (a,.6), = str(ab). Since invariant forms on G
are proportional, we have (a, b)” = Zy(u, b), where IV E k is the index of the
representation V. We have: str(r) = C (-l)des”c str(u#) = &(dim Gs -
dim q). On the other hand, SW(~) = (dim Ve - dii Vi)(A, A + 2~).
Thus, Z,(dim G-6 - dim (;) = (dim V, - dii V&l, A + 2p), from which we
obtain
PROPOSITION 5.2.6. The supertrace form of a jinite-di~l beducibk

representa& of a simple contragredknt Lie superakebra with dim Grj # dim Gi
in a space V with highest weight A is nomkgenerate Q and only if (dim V, -
dii v&l, A + 2p) # 0.
86 V. G. KAC
EXAMPLE (compare [22]). We consider the standard representation osp(l,2)

of the dispin superalgebra B(0, 1). Then V = (1” osp(l,2), k = 0, l,... are
all the irreducible finite-dimensional representations of B(0, 1). The highest
weight of V is 2k, dim I’” = 2k + 1, V$ and Vik are B(0, l)O-irreducible,
dim VGa = k, dim Vi’ = k + 1. The supertrace form is always nondegenerate
(Proposition 5.2.6).
5.3. Simple Lie Superalgebras Over Nonclosed Fields
In this section the ground field k is arbitrary, of characteristic 0.
5.3.1. Reduction of the classsfication of simple Lie superalgebras over k to finding

of the forms. Let k be the algebraic closure of k. We recall that a k-algebra G
is said to be a form of a k-algebra G if G @ k ru G. If G is a form of G and
Y a G-module, where Y is a vector space over k, then V is called a form of the
G-module V @ k. We recall that if G is a semisimple Lie algebra over k and
V is an irreducible G-module over k, then the G @ k-module V @ k splits
into the direct sum of irreducible submodules, which are equivalent up to a
“twist” under an outer automorphism of G @ k. We also remark that there
is at most one irreducible form of a G-module V for a given form G of a Lie
algebra G.
The next result is proved just as for Lie algebras [lo].
PROPOSITION 5.3.1. A simple finite-dimensional Lie superalgebra over k is

isomorphic either to G @ k’, where k’ is a finite extension of k and G is one of the
k-algebras A(m, n), B(m, n), C(n),..., S(n), s(n), H(n), or it is a form of one
of these k-algebras.
If G = G, @ Gi is a Lie superalgebra over k, then for an element 01E

k* mod k*2 we can construct another form G’ for G @k, by setting [a, b]’ =
ol[a, b] for a, b E Gi’ and [a, b]’ = [a, b] otherwise. This form we call equivalent
to the original one.
5.3.2. Forms of the classical Lie superalgebras.
PROPOSITION 5.3.2. (a) If a Lie superalgebra G = G,- @ Gi over k is a form

of G = GG @ Gi , then G, is a form of Go, and the G,-module Gi is a form of the
GG-module Gi .
(b) Suppose that G = Ge @ Gi is a classical Lie superalgebra oz’er k;
suppose also that GG is a form of Gb and a GG-module V is a form of the Go-module
Gi . Then there is one and only one up to equivalence Lie superalgebra G =
Go @ Gi over k that is a form of G, with the G-,-module Gi isomorphic to V.
(c) The Lie superalgebras B(m, n), D(m, n), D(2, 1; ol), F(4), G(3), P(n),
and Q(n), &+d over k, kave at most 01lc up to equivaknce jxm G = Gc @ G

with a given subaZgebra G6 .
(d) The Lie snpera&bras A(m, n) and C(R) over K have at moEt one form
G = Gs @ Gi CoitR u g;Uen subdgebra G jar whicir the +nwdule c;I is
irr&cibk (W reducible, tzspectivsly), up to equivalmce.
Pro@. (a) is obvious. (b) follows from the fact mentioned earlier (see
Section 2.1.6) that the Grmodule SaGi contains ad GG as a direct summand
with multiplicity 1. The condition for the map SeGi -+ ad G$ to define a Lie
superalgebra (see Section 1.1.2, (1.1.3)) 1s
* p reserved under a change of field.
(c) and (d) follow from remarks made above in Section 5.3.1.
5.3.3. Forms of Lie SuperaZgebras of Cartan type. In Sections 3.1 and 3.3
it was shown that for all Lie superalgebras of Cartan type, that is, W(n) with
n > 3, S(n) and s(n) with 1 > 4, and H(n) with n > 5, the filtration is
invariant under automorphisms, and the reductive part of the automorphism
group is isomorphic to GL, in the first two cases, SL, , and SO,, , respectively.
Hence by using the same arguments as in [18], we obtain the following result.
PROPOSITION 5.3.3. Let G be a Lie supera&ebra wer k and a jomr of me

of the Lie superalgebras of Cmtun type W(n), S(n), S(n), ur H(n), wer k. Then
G is isomorphic to one of the jolZowing Lie supera&ebras wer k, respectively W(n),
S(n), s(n), OTH(C at(d&)*), where a, E k*.
5.3.4. Clastiification of simpk finite-dimensional real Lie supera&ebras. Let us
begin by constructing some series of examples of Lie superalgebras over R.
We fix the standard embeddings of the fields of real and of complex numbers
in the quaternion field: R CC C H. Throughout what follows, the bar denotes
the standard conjugation in C and H.
(a) Tke spe&l linear Lie superaZgebras sZ(m, n; k), k = R or C or H.
We consider the space Z(m, n; k) of all square matrices of order m + n over k.
In it we single out the subspaces
where 01is an (m x m)-, S an (n x n)-, B an(m x n)-, and y an (B x m)-matrix.

We define the bracket in the usual way: [u, b] = ub - (- l),$ ba if a E Z(m, n; k), ,
b E Z(m, n; k)t . This makes Z(m, n; k) into a real Lie superalgebra. The special
Zineur superakebra SZ(m,n; k) is the real subalgebra of Z(m, n; k) distinguished
by the conditions:
sZ(m, n; k) = (a E Z(m, n; k) 1&r(a) = 0} for k = R or C;

sZ(m, n; H) = {a E i(m, n; H) [ Re str(u) = O}.
607/26/r-7
88 V. G. KAC
For m = n these superalgebrascontain a one-dimensionalcenter, which has

to be factored out.
(b) Unitary and orthogonal-symplectic Lie superalgebras. Again let
K = R or C or H. We consider the Za-graded space km+n = k” @ R”; let
f = f (x, y), be a function on km+n with values in k, which is linear relative
to the first variable, superhermitian, i.e., f (x, y) = (-l)(desz)(deay)f( y, x),
nondegenerateand consistent, i.e., f (x, y) = 0 if x Ekm, y E k”. We put
su(m, n; f )s = {a E sl(m, n; C), If (a(x), y) = -(-lYdegzf (x, a(y)>,

osp(m, n; f )s = {a E sl(m, n; & I f (a(x), Y) = -(- lYdeg3f (x, a(y)>,
hosp(m, n; f )s = {a E sJ(m, n; Ws If (a(x), y) = -(-l)sdegEf (x9 a(y)>, s EZ2
The Lie superalgebrassu(m,n; f), osp(m, n; f) and hosp(m,n; f) are called

unitary, orthogonal-symplectic, and quaternion orthogonal-symplectic, respectively.
(c) The Lie superalgebras UQ(rz, p) and HQ(n). Let
@(n; P) = I(-+&), a Eu(P, n - p), b Esu(p, n - p)/ C sl(n, n; C)
@i(n)= I(+/+), a, b E l(n), Re tr b = 0 C Z(m,n; H).

1
We put UQ(n,$9 = $(n, ~)l(l,,h HQ(n) = %@)/W2, , XE RI.

(d) The Lie superalgebras D(2, 1; ar;p). For each of the representations
of so(4,4 - p; R) @ sZ(2;R), p = 0, 1, 2, there is a family of real Lie super-
algebrasD(2, 1; ol; p), 01ER\{O, -l}, that are forms of D(2, 1; a).
(e) The Lie supera2gebras F(4; p). Each of the Lie algebrasso(p, 7 - p),
p = 0, 1,2, 3, has a spinor representation spin,,,-, , which is a real form of
the Bs-module spin, . For each of the four linear Lie algebrasspin,,,+, there is,
by Proposition 5.3.2(b) one and only one real Lie superalgebra F(4; p),
p = 0, 1, 2, 3, which is a form of the complex Lie superalgebraF(4).
(f) The Lie superalgebras G(3; p). Each of the real forms G,,, , p = 0, I,
of the complex Lie algebra G, has a 7-dimensional representation G,,, . For
each of the two linear Lie algebrasG,,, @ sl, there is, by Proposition 5.3.2(b),
one and only one real Lie superalgebraG(3; p), p = 0, 1, which is a form
of the complex Lie superalgebraG(3).
(g) The Lie SuperaZgebras H(n; p; R):
H(n; p; R) = /,D EW(n; R) ( D (f (dtd’ - tj+l (4%)‘) = 01.

i=l
(h) The Lie superalgebras P(n; R), Q(n; R), W(n; R), S(n; R), and
S(n; R). These are P(n), Q(n) ,..., defined for k = R.
Real Lie superalgebras obtained from one another by the construction in

Section 5.3.1 (for u = -1) are called dual. From the classiiication of simple
real Lie algebras and from Propositions 5.3.1-5.3.3 we derive the following
theorem.
THEOREM 9. A simple jinite-dime&o& real Lie SuperaJgebra that is not a

Lie akebra is iwmwpk& eitker to one of the c~mpkx ‘Lie supera&ebras A(m, n),
B(m, n), c(n), D(m, n), D(2,l; 4, P(4), G(3), W), QW, W), W), S(n),
or H(H), regarded as real superalgebras, o*, up to transition to its dual, to one of
the forms of these Lie supera&ebras listed below:
A s&n, n; R), =(m, n; p, q), m, n > 1, m + n > 2; s&m, n, H),

m,n > l;H(4;p;R).
B osp(m, fl; p; R), m odd, m >, 1, n > 2.
C OS@, n; P; R),, hosp(l, n; P), n > 2.
D osp(m, n; p; R), m ewen, m 3 4, n > 2; hosp(m, n; p), m 3 2,
D(Z 1; a; P).
F F(4, P), P = 0, 1,2,3.
G G(3, P), P = 0, 1.
P P(n, R), n 2 3.
Q Q(n, R); UQ(n; P), n > 3; HQ(n), n a 2.
W W(n, R), n 2 3.
S S(n, R), n > 4.
s s(n, R), n > 4.
H H(n; p; R), n > 5.
5.4. On tke Classification of Infinite-Dimensional Primitive Lie Superalgebras

In Section 4.2.3, we have given a classification of the primitive Lie super-
algebras (L, L,) for which L, 1 L, ; they are all finite-dimensional.
In this section we state without proof some partial results on the classification
of infinite-dimensional primitive Lie superalgebras. We recall that Lie super-
algebra L with a distinguished subalgebra L, is called primitive if L, is a maximal
subalgebra and it does not contain nontrivial ideals of L.
54.1. Two &ebras of da&et&l forms. Let &,(m) be the superalgebra of

differential forms with coefficients from the polynomial algebra k[xl ,..., x,J;
in other words, J&,(m) is the associative superalgebra over k[x, ,..., x,,,] (with
trivial &grading) with the generators dx, ,..., dx, with the defining relations
dxi A dxj = -dx* A dxi 3 deg dxi = 1, i,j= 1,***, m.

90 V. G. KAC
On Q,(m) the differential d is defined in the usual way, as a derivation of degree i

for which d(xJ = dxi and d2(xi) = 0, i = l,..., m, with the standard properties.
We now define the following superalgebras:
Q(m, 4 = Q,(m) 0 Q(n) and @(m, n) = Q,(m) @ O(n).
The differentials d and 0 are extended from Q(n) and O(n) to Q(m, n) and
O(m, n) in the natural manner, namely, d = d @ 1 + 1 @ d, 8 = d @ 1 +
1 @ 0. It is not hard to establish their properties, which are similar to those
in Chapter 3.
We set A(m, n) = k[x, ,..., x,,] @ A(n). The relations deg & = deg xj ‘= 1
determine on A(m, n) a Z-grading (which is not consistent with the Za-grading).
Every derivation D of degree s of A(m, n) extends uniquely to a derivation of
O(m, n) and Q(m, n), subject to the conditions [D, 0]f = [D, d]f = O,~E A(m, n).
5.42. Six series of injinite-dimensional Lie superalgebras. We introduce the

following differential forms:
v = dx, A ... A dx,,, A tit, A -a. A et, E O(m, n),
h = 2 i dxi A dx,+i + i (d[i)2 E Q(m, n), m = 2k,

i=l i=l
k = dx,,+l + i (xi dx,+i - xk+i dxi) + i Es dti E Q(m, n), m = 2k + 1.

i=l i=l
We now define six series of infinite-dimensional Lie superalgebras (m > 0):

I. W(m, n) = der A(m, n).
The other five series consist of Lie algebras inside W(m, n), which are
characterized by the following action on the differential forms
II. S(m, n) = {DE W(m, n) 1 Dv = 0},

II’. CS(m, n) = {D E W(m, n) / Dv = Av, X E k},
III. H(m, n) = {DE W(m, n) 1 Dh = 0},
III’. CH(m, n) = (D E W(m, n) / Dh = Ah, h E k},
IV. K(m, n) = {D E W(m, n) 1 Dk = uk, u E A(m, n)}.
Note that for n = 0 we obtain the six standard series of infinite-dimensional

Lie algebras of Cartan type.
The Z-grading of A(m, n) induces an (inconsistent) Z-grading in W(m, n).
We write it down in more detail. Every element D E W(m, n) can be expressed
as a linear differential operator
D==fP,&f&& Pi 1 Qj E A(m, n). (54.1)

i=l 2 3=1 J
LIE SUI’EBALGEBBAS 91
The relations deg x, = deg t, = 1, deg B/ax, = deg a/at, = - 1 also deter-
mine on W(m, n) a Z-grading W(m, n) = &-r W(w, n), , which corresponds
in the canonical way to a filtration; the appropriate died subalgebra
is ad>0 W(w #If. The filtration and the distinguished subalgebra induce on
every subalgebra L a filtration and a distinguiied subalgebra L, = L n
Of&o W(m, 4 *
The Lie superalgebra S(nr, n) consists of the operators of the form (5.4.1)
satisfying the condition
Hence we see that S(m, n) is the linear span of the elements of the form
aa a aa a
&g+&$
i f -----9
ax, ax, ax* ax,
E& + (-l)d$$&, a E A(m, n).
The Lie superalgebra H(m, n) consists of the operators of the form
aEA(m,n).
Here [Da , G,] = D{a,a) , where
{a, b} = (-l)des=
Next,
CS(m 4 = Sh n) 0 ($ -5 & + 5 xi -g-)9

f-1
CJ-W ~1 = H(w 4 0 (cl 5t -& + 2 f xi -&)a

f-1 '
Finally, the Lie superalgebra K(m, ti) consists of elements of the form
92 V. G. KAC
The Z-grading of W( m, n ) a1so induces aZ-grading of the form G = &z-r G,

on the Lie superalgebras of series II, II’, III, and III’. This is not so in the
case of the last series K(m, n). However, if we set deg xi = deg & = 1,
dega/ax,=dega/a~j=-lfor1<i<m-l,1<j<nanddegx,=2,
deg a/ax, = -2, then the resulting Z-grading of W(m, n) also induces a
Z-grading of the form G = &>-a G, on K(m, n).
Note that the G,,-modules G-r for these six series are isomorphic to the
following Lie superalgebras:
Z(m, rz) for W(m, n) and CS(m, n), sZ(m, n) for S(m, n), osp(n, m) for H(m, n),
cosp(n, m) for CH(n, m) and cosp(n, m - 1) for K(m, n).
The superalgebras of all six series are transitive and irreducible, and those of
series I, II, III, and IV are even simple.
5.43. On the classification of primitive Lie superalgebras. Let L be an
infinite-dimensional primitive Lie superalgebra and L, be the distinguished
subalgebra. Let L-i be some minimal (Zs-graded) subspace of L that contains
L, and is different from L, and ad Lo-invariant. We construct a filtration in L
of the form
L = L-d’I)L--d+lI *..r)L_,T)L,ILL,r) .“, by setting (see [q):

L-(,+1)= [L-l , Ll + L-8 > L, = Ia EL1 I [a,L-J CL), s > 0.
The corresponding associate Z-graded superalgebra GrL = @i>-d Gi has
the following properties:
1”. Gr L is transitive and irreducible,

2”. G-, = G”_, for s > 0.
3”. Gl # 0.
We may also assume that
4”. &Co Gi does not contain nonzero ideals of Gr L (because we can

factor out such an ideal if it exists).
If the Z-grading is consistent, then
5”. [G,, , G,,] is a contragredient Lie superalgebra.

Apparently, 5” holds in general, but I have not been able to prove this.
Now we can state (without proof) the main result of this section.
THEOREM 10. Let G = @i>--dGi be an in$nite-dimensionalZ-graded Lie

superalgebra having properties I”-5”. Then G is isomorphic as Z-graded super-
algebra to one of W(m, n), S(m, n), CS(m, n), H(m, n), CH(m, n), or K(m, n)
with m > 0.
LIE suPERALGEBRAs 93
The proof usesthe samemethods as in Chapter 4 and relies on Theorem 3.

A primitive Lie superalgebraL with distinguished subalgebraLs is called
completeif it is complete in the topology defined by the subspacesof the tran-
sitive filtration of the pair (L, L,). (see Section 1.3.1). The super-algebra
if(m, n) = h[[xl ,..., xv,]] @n(n) is complete in the topology defined by its
natural filtration.
We denote by w(m, n) the L re superalgebra of all continuous derivations
of if(m, tt). Then w(m, n) is a complete primitive Lie superalgebrawith the
natural distinguished subalgebra. Complete and primitive are also s(m, n),...,
K(m, n), which are characterized by the sameaction on the differential forms
q h, h, as for S(m, n),..., K(m, n).
A well-known result of Cartan assertsthat w(m, O),..., R(m, 0) are the only
infinite-dimensional complete primitive Lie algebras.
CONJECTURE 1. An infkite-dimensional completeprim&be Lie su.&ebra

is bmorphic to w of w(m, n), s(m, n), ?%(m, n), fi(m, n), @@mm, n), or
$(m,n)form>O.
5.4.4. Rem&s. (a) In Chapter 4 we have, in fact, proved that if G =
@<>-aGt is an infinite-dimensional Lie superalgebrawith a consistentZ-grading
having properties 1”-4”, then G is isomorphic to K(1, n).
(b) There are general embedding theorems, which generalize standard
theorems for Lie algebrasand Propositions 3.1.2 and 3.1.3.
If G = @t-+1 Gt is a transitive Zgraded Lie superalgebra, and m =
dh(G&, , n = dim(G-r)i , then there is an embedding G + W(m, n)
preserving the Z-grading.
LetL =L,3L03h3 .** be a transitive Lie superalgebrawith a filtration,
m = dim(L/L& , n = di@L/L& . Then there is ZUIembedding a: L + tff(nt,n)
preserving the filtration. If /3 is another such embedding and (a - fljL CL,,
then there exista one and only one (continuous) automorphism F of w(m, n)
for which a = q 0/3; ‘p can be induced by an automorpbism of ii(m, n).
(c) By the samemethod asProposition 3.3.8, it can be proved that every
nondegenerate closed difkrential form from &a, n) of degree 2 is reduced
to the form h by an appropriate automorphism of A(m, n).
5.5. SomeUnsolvedProblems
5.5.1. Chjlcatti Lie mperalgebas. Apropos
of injFnite-dimmciorurlprindr;p/c
this topic, see Conjecture 1 and Theorem 10.
5.5.2. Form&s for the charactersand dimenkmsof iweducihlerepresentations.
The moat urgent task is to prove a formula for the characters in the caseof
contragredient Lie superalgebras.For contragredient Lii algebras(including
94 V. G. KAC
these of infinite dimension) this is done in [14]. However, the proof in [14]
only works for B(0, n).
5.5.3. Cohomology. For the definition of the cohomology group Hk(G, V)
of a Lie superalgebra V with coefficients in a G-module V, see [17]. As usual,
it is shown that if V is a finite-dimensional irreducible G-module and r is the
Casimir operator (the existence of an invariant bilinear form is assumed),
then Hk(G, V) = 0 for T(V) # 0. In the case of contragredient Lie super-
algebras, the latter condition is equivalent to (A, rl + 2~) # 0, where (1 is
the highest weight (see Section 5.2.3), and it is not violated in any nontrivial
representation only for B(0, n).
Now some questions arise at once: the cohomology of the simple finite-
dimensional Lie superalgebras with trivial coefficients, and the cohomology
of the infinite-dimensional complete primitive Lie superalgebras.
Closely connected with the problem of the triviality of H1(G, V) is the full
reducibility of representations and the theorems of Levi and Mal’tsev. A counter-
example to Levi’s theorem is sZ(n, n), and one to full reducibility is the adjoint
representation of A(n, n). As we have already mentioned, full reducibility
always holds for B(0, n). It is not hard to show that if G is a classical Lie super-
algebra, then Hl(G, V) = 0 for all irreducible representations, with the excep-
tion of a finite set S. It would be interesting to find this S and also to classify
all indecomposable representations of the classical Lie superalgebras.
5.5.4. Infinite-dimensional representations. Undoubtedly, Kirillov’s orbits
method extends to Lie superalgebras. (In particular, Theorem 7’ on infinite-
dimensional representations of solvable Lie algebras points to this.) We mention
that Kirillov’s differential form W(X, y) = I([%, y]) on an orbit of the co-adjoint
representation of a Lie superalgebra is a form in dx and df (see Section 5.4).
On infinite-dimensional representations of the simple Lie superalgebras almost
nothing is known. First in line is, of course, the dispin algebra B(0, 1).
5.5.5. GeneralizedLie superalgebras. We consider the ring M = 2, 0 .. @Z,
(s times). An M-graded algebra is called a generalized superalgebra. If 01=
(011,***, a,) EM, we set (--l)o: = (-I>“1 ... (- 1)~. Now all the definitions and
assertions of Section 1.1 carry over to generalized superalgebras, in particular,
the definitions of a Lie superalgebra, of the supertrace, and the Killing form.
Just as in Section 2.1, we can define series of generalized Lie superalgebras
d(n, ,..., nd, osp(n, ,..., 7ts”), Q(n), and as in Chapter 3, the series W, S, s, H.
The same problems arise here as for Lie superalgebras, first and foremost,
the problem of classifying the simple generalized Lie superalgebras.
Additional remarks. To Section 5.2. In my recent article, “Characters
of Typical Representations of Classical Lie Superalgebras” (Commun. Algebra 5,
NO. 8,889-897(1977)), the formulas for the character and supercharacter of finite-
dimensional irreducible representations in “general position” (so-called typical
representations) are obtained. For example, all the repreaentations of B(0, n)

are typical and all the representations of A( 1,O) are typical except for SV(2, ‘1)
and ita dual.
To Se&on 5.3. D. Z. Djokovic and G. Hochshild proved, in their article

“Semi-simplicity of Z-graded Lie algebras, II” (I&&r J. M&J. 20 (1976),
134143), that every finite-dimensional representation of a Lie superalgebra
G is fully reducible if and only if G is a direct sum of a semisimple Lie algebra
and several copies of B(0, nJ.
From my article, mentioned in Section 5.2, it follows that Ip(G, V) = 0
for a typical representation V. By the way, HyA(1, 0), V) = 0 only for one
irreducible representation-standard representation.
To Section 5.4. B. Kostant in his recent paper, “Graded manifolds, graded

Lie theory and prequantization,” gave a more correct definition of a super-
manifold than the one in [s]. This d&&ion allowed him to develop the theory
of homogeneous super-manifolds, and “orbits method” for supergroups.
REPERENCES
1. E. M. ANDRBBV, E. B. VI-Q, AND A. S. ELASHVILI, Orbits of largest dimension

of semi-simple linear Lie groups, Fun&ma1 Ad Appl. 1 (1967), 257-261.
2. F. A. BBREZIN, “The method of Second Qua&ration,” Academic Press, New York,
1966.
3. F. A. B-IN, Automorphisms of the Grassmann algebra, M&I. Notes 1 (1967),
180-184.
4. F. A. BERFZIN AND G. I. KATZ, Lie groups with commuting and anticommuting
parameters, Math. USSR. Sb. 11 (1970), 311-326.
5. F. A. BEREZIN AND D. A. LEITES, Supervarieties, Sow. Math. Dokl. 16 (1975), 1218-
1222.
6. B. Yu. WEISFEILER, Infinite-dimensional filtered Lie algebra and their connection
with graded Lie algebras, FunctionaZ Anal. Appl. 2 (1968). 88-89.
7. B. Yu. WEISPEILER AND V. G. KM, Irreducible representations of p-Lie algebras,
Fru~~tionul And Ap~l. 5 (1971), 111-117.
8. B. Yu. WBISP~ILER AND V. G. mc, Exponentiala in Lie Algebras of characteristic p,
Moth. USSR Iav. 5 (1971). 777-803.
9. E. B. VINBERO AND A. L. ONISHIK, ‘Seminar on Algebraic Groups and Lie Groups,”
Moscow, 1969 [in Russian].
10. N. J~~onao~, “Lie Algebras.” Wiley-Interscience, New York, 1962.
11. V. G. KAC, Simple irreducible graded Lie Algebras of &rite growth, Math. USSR
law. 2 (1968), 1271-1311.
12. V. G. KAC, On the classification of simple Lie algebras over a field of non-zero
characteristic, Math. USSR Iru. 4 (1970), 391-413.
13. V. G. KAC, “Some Algebraa Related to the Quantum Theory of Fields,” XIth
All-Union Algebr. COB., pp. 140-141, Kishinev, 1971 [in Ruseian].
14. V. G. KAC, Infinite-dimensional Lie algebras and Dedekind’s r)-function, Function&
Anal. AppZ. 8 (1974). 68-70.
96 V. G. KAC
15. V. G. KAC, Description of filtered Lie algebras associated with graded Lie algebras
of Cartan type, Math. USSR Izo. 8 (1974), 801-835.
16. V. G. KAC, Classification of simple Lie superalgebras, FunctionaE Anal. Appl. 9
(1975), 263-265.
17. D. A. LEITES, Cohomology of Lie superalgebras, Functional Anal. AppZ. 9 (1975),
340-341.
18. A. N. RUDAKOV, The automorphism groups of infinite-dimensional simple Lie
algebras, Math. USSR Izv. 3 (1969), 707-722.
19. G. L. STAVRAKI, Some non-local model of field selfinteractions and the algebra of
field operators, in “High Energy Physics and the Theory of Elementary Particles,”
Naukova Dumka, Kiev, 1966.
20. R. J. BLATTNER, Induced and produced representations of Lie algebras, Trans. Amer.
Math. Sot. 144 (1969), 457-474.
21. R. E. BLOCK, Determination of the differentiably simple rings with a minimal ideal,
Ann. Math. 90, No. 2 (1969), 433-459.
22. L. CORWIN, Y. NE’EMAN, AND S. STERNBERG, Graded Lie algebras in mathematics
and physics (Bose-Fermi symmetry), Rev. Mod. Phys. 47 (1975), 573-604.
23. S. KOBAYASHI AND T. NAGANO, On filtered Lie algebras and their geometric structure,
III, 1. Math. Mech. 14 (1965), 679-706.
24. J. MILNOR AND J. MOORE, On the structure of Hopf algebras, Ann. Math. 81 (1965),
21 l-264.

Kac - 1977 - Lie Superalgebras

Uploaded by

Copyright:

Available Formats

Kac - 1977 - Lie Superalgebras

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Kac - 1977 - Lie Superalgebras

Uploaded by

Copyright:

Available Formats

ADVANCES IN MATHEMATICS 26, 8-96 (1977)

Massachusetts Institute of Technology,

Cambridge, Massachusetts 02139

DEDICATED TO THE MEMORY OF MY FATHER, G. M. KAC

Contents. Introduction. 1. Basic definitions and preliminary remarks:

“Graded Lie algebras have recently become a topic of interest in physics

[u, b] = -(-1)“8[6, a] for UE G., by Gs,

We mention that the Whitehead operation in homotopy groups satisfies

sification is divided into two parts, corresponding to the cases of a nondegenerate

Remark. The history of this article began in 1969 when, impressed by

1. BASIC DEFINITIONS AND PRELIMINARY REMARKS

1 .l . Superalgebras and Lie Superalgebras- Supertrace

1. I .l. Superalgebras. We recall that if A is an algebra and M an Abelian

with an M-grading is called M-graded. If a E A,, then we say that a is

(a1 @ b,)(a, @ b,) = (- l)(degaB)(degbl)

There is a natural way of defining a bracket [ , ] in a superalgebraA, i.e.,

A superalgebra is called commutatiweif [a, b] = 0 for all a, b E A. Quite

[a, bc] = [a, b]c + (-l)(deg*)(degW[a, c]. (1.1.2)

EXAMPLE 1. Let M be an Abelian group and V = eapM V, an M-graded

End, V = {a E End V I a(V,) C Vs+.}.

In particular, for M = Zs we obtain the associative superalgebraEnd V =

EXAMPLE 2. Let n(n) be the Grassmannalgebra in n variables & ,..., 5, .

A generalization of this example is the commutative superalgebra fl(m, n) =

[a, b] = -(-l)(dega)(dw)[b, a] (anticommutativity),

Observe that Ge is an ordinary Lie algebra, that multiplication on the left by

da, b)c + db, c)a + v(c, a)b = 0 for a, b, c E Gi . (1.1.3)

EXAMPLE 1. If A is an associativesuperalgebra, then the bracket (1.1.1)

EXAMPLE 2. Let G be a Lie superalgebraand A(n) a Grassmann super-

We set U(G) = T(G)/R. The natural map G + U(G) evidently induces a

THE Pomcm&Bmui~Wm THBORBM. Let G = 60 G be (1 Lie

form a basis of U(G).

d: U(G) -+ U(G) @ U(G),

which is called the diagonal homomorphism.

We denote by der, A C End, A the space of all derivations of degree s, and

EXAMPLE 1. Let G be a Lie superalgebra. It follows from the Jacobi identity

EXAMPLB 2. Let n(n) = lia(n) @n,(n) be a Grassmann superalgebra.

from which it follows that I is invariant under D. Since, obviously, there is

(@D)f = @(WW)), f E 44.

1.lS. The superalgebra l(V) and the supertrace. Let V = VG @ Vi be a

0 = str([b, UC])= ([b, a], c) + (-l)(dego)(degb)(a, [b, cl),

1.1.6. Linear representationsof Lie SuperaZgebras.Let V = V,j @ Vi be a

SCHUR’S LEMMA. Let V = VG@ Vi, & an irreduciblefamily of operators

EXAMPLE. We consider the Lie superalgebra N = NC @ NT, where NG =

ENGEL’S THEOREM. Let G be Q subalgebraof Z(V) and supposethat all the

Let V = V, @ Vi be a Zs-graded space. By the symmetric (respectively,

1.2. Z-GrahamLie Supera&ebnu

1.2.1. Z-gradings. A Z-grading of a superalgebraG is a decompositionof it

PROPOSITION 1.2.1. If in a simpleZ-graded Lie superakebraG = eieZ Gi

1.2.2. Local Lie superalgebras.Let G be a Za-graded space, decomposed

PROPOSITION 1.2.2. Let C?= G-, Q G, @ GI be a local Lie supereebra.

1.2.3. Invariant bilinear forms. The following proposition is proved in the

The following assertion is proved in the same manner as [7, Corollary 2 to

PROPOSITION1.2.5. Let G = &-, Gi be a simple finite-dimensional Z-

Clearly, the kernel of an invariant form is an ideal. Therefore, we have the

PROPOSITION 1.2.6. If G is a simple Lie superalgebra, then an invariant form

1.2.4. Conditions for simplicity. In this subsection we state some conditions

PROPOSITION1.2.7. The following conditions are necessary for a Lie super-

PROPOSITION 1.2.9 (cf. [23]). Let G = @o-1 Gi be a tram&e irreducible

PROPOSITION 1.2.10. Let G = @I>-1 Gt be a Z-graded Lie super&ebra

B,(n > 3) ~--o---.-O 0 G 03A