Pure spinors in AdS and Lie algebra

cohomology

Andrei Mikhailov

Instituto de Fı́sica Teórica, Universidade Estadual Paulista

R. Dr. Bento Teobaldo Ferraz 271, Bloco II – Barra Funda
CEP:01140-070 – São Paulo, Brasil

Abstract
We show that the BRST cohomology of the massless sector of the
Type IIB superstring on AdS5 × S 5 can be described as the relative
cohomology of an infinite-dimensional Lie superalgebra. We explain
how the vertex operators of ghost number 1, which correspond to
conserved currents, are described in this language. We also give some
algebraic description of the ghost number 2 vertices, which appears
to be new. We use this algebraic description to clarify the structure
of the zero mode sector of the ghost number two states in flat space,
and initiate the study of the vertices of the higher ghost number.

MSC classes: 83E50 (Primary) 17B55, 16S37 (Secondary)

Keywords: superstring theory, quadratic algebras,

pure spinor formalism, vertex operators,
AdS/CFT corresondence
Contents
1 Introduction 2

2 Pure spinor formulation of the SUSY Maxwell theory 4

2.1 Supersymmetric space-time and basic constraints . . . . . . . 4
2.2 Definition of the Lie superalgebra L. . . . . . . . . . . . . . . 5

3 Lie algebra cohomology and solutions of the SUSY Maxwell

theory 6
3.1 Vacuum solution . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2 Deformations of solutions and cohomology . . . . . . . . . . . 7
3.3 Koszul duality and its application to deformations . . . . . . . 7

4 Type IIB SUGRA in AdS5 × S 5 9

4.1 BRST complex . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4.2 Lie algebra formed by the covariant derivatives . . . . . . . . . 11
4.3 Lie algebra cohomology . . . . . . . . . . . . . . . . . . . . . . 12
4.4 An analogue of the Koszul resolution . . . . . . . . . . . . . . 15
4.5 Reduction to the cohomology of the ideal I ⊂ Ltot . . . . . . . 17
4.6 Ghost number 1: global symmetry currents . . . . . . . . . . . 18
4.7 Ghost number 2: vertex operators . . . . . . . . . . . . . . . . 22

5 Flat space limit 23

5.1 Ghost number 1. . . . . . . . . . . . . . . . . . . . . . . . . . 24
5.2 Ghost number 2. . . . . . . . . . . . . . . . . . . . . . . . . . 24
5.3 Higher ghost numbers . . . . . . . . . . . . . . . . . . . . . . 28
5.3.1 Super-Maxwell BRST complex . . . . . . . . . . . . . . 29
5.3.2 Type IIB BRST complex . . . . . . . . . . . . . . . . . 29
5.3.3 Dual picture . . . . . . . . . . . . . . . . . . . . . . . . 36
5.3.4 Conjecture about the vertices at the ghost number 3 . 36

6 Conclusion 36

A Exactness of (77) 37

1
1 Introduction
Pure spinor formalism [Ber00] is a generalization of the BRST formalism
with the ghost fields constrained to satisfy a nonlinear (quadratic) equation:
λα Γm β
αβ λ = 0 (1)
where Γmαβ are the Dirac’s Gamma-matrices. A natural question arizes, what
kind of nonlinear constraints can ghost fields satisfy in a physical theory?
What if we replace (1) by an arbitrary set of equations:
λα Cαβ
i
λβ = 0 , i∈I ? (2)
Of course, this would generally speaking have nothing to do with the string
theory. But the question is, besides coming from superstring theory, what
m
special properties of Cαβ = Γmαβ are important for physics? This would be
useful to know, for example when thinking about possible generalizations of
the pure spinor formalism.
It turns out that there is some special property of (1) which plays an im-
portant role in the string worldsheet theory. This is the so-called Koszulity
— see [GKR06] and references therein. The formalism of Koszul duality was
extensively used in the study of the algebraic properties of the supersymmet-
ric Yang-Mills theories in [MS04b, MS04a], and in the classification of the
possible deformations of these theories in [MS09].
In this paper we will study the BRST cohomology of the massless sector of
the Type IIB superstring in AdS5 × S 5 . We will use the formalism of Koszul
duality to gain better understanding of the massless BRST cohomology.
The BRST cohomology counts infinitesimal deformations of the back-
ground AdS5 ×S 5 , also called “linearized excitations” or “gravitational waves”.
From the point of view of the string worldsheet theory, they are identified
with the massless vertex operators. Understanding the properties of these
vertex operators is important already because of their role in the scatter-
ing theory. Indeed, the correlation function of vertex operators is the main
ingredient in the string theory computation of the S-matrix.

Main results
1. We show that the cohomology of the BRST complex of the Type IIB
SUGRA on AdS5 × S 5 is equivalent to some relative Lie algebra coho-
mology.

2
 2. We classify the vertex operators of the ghost number 1, which corre-
spond to the densities of the local conserved charges

3. We give a general Lie-algebraic description of the vertex operators of

the ghost number ≥ 2 and use this description to study the properties
of the zero momentum states (“discrete states”)

4. We show that there are nontrivial cohomology classes at the ghost num-
ber three; we conjecture that these nontrivial cohomology classes ob-
struct the nonphysical ghost number two vertices of [BBMR11, Mik12]

Previous results for ghost number 1 The classification of the vertex

operators in the ghost number 1 was done, at least partially, in the Appendix
of our previous paper [Mik11b]; the method which we develop here appears
more elegant.

Zero momentum states In a typical string theory computation one con-

siders the scattering of physical excitations (vertex operators) which depend
on the space-time coordinates exponentially:

V (x) ≃ eikx (3)

But we find it interesting to also consider vertex operators depending on x

polynomially. We will call them “zero momentum vertices” because their
wavefunction in the momentum space is supported at k = 0. It turns out
that this “zero momentum sector” carries one potentially unpleasant surprize:
there are some well-defined vertex operators which do not correspond to any
physical states [BBMR11, Mik12]. In this paper we use the Koszul duality to
obtain a dual description of such unphysical states in terms of fields satisfying
unusual equations of motion, similar to this one:

∂m An + ∂n Am = 0 (4)

Such equations imply that higher derivatives of A vanish.

This means that just the requirement of BRST invariance alone does
not yet provide a complete characterization of the physically relevant sigma-
models. But the picture becomes complete if one considers the sigma-model
at the one-loop level. It was shown in [BH02] that the holomorphicity condi-
tions of the BRST currents at the one-loop level require the dilaton superfield

3
to satisfy certain system of equations. We have shown in [Mik12] that these
equations become incompatible for the unphysical vertices. The one-loop
properties of the sigma-model in AdS5 × S 5 were studied in [Val02, Ber05b,
MV09].
If we do not discuss loop effects, then the dilaton coupling does not mat-
ter. At the purely classical level, the theory deformed by a nonphysical
operator is consistent. However, such nonphysical deformations are proba-
bly inconsistent with those physical deformations which break symmetries.
We can also say that “nonphysical deformations and physical deformations,
generally speaking, obstruct each other”. In other words, nonphysical in-
finitesimal deformations only exist in those backgrounds which have global
symmetries, such as flat space and AdS5 × S 5 [Mik14]. The nonphysical
states should be absent in the semirelative cohomology, i.e. if we request
that the cochains be annihilated by the b0 − b̄0 . See [Mik14] for the analysis
of vertices in flat space.

Plan of the paper We will start in Sections 2, 3 with the application

of Koszul duality to the ten-dimensional supersymmetric Maxwell theory.
In Section 4 we apply a similar method to the study of linearized Type IIB
SUGRA in AdS5 × S 5 . We introduce in Section 4.2 some infinite-dimensional
super-Lie algebra, and show in Section 4.5 that the BRST cohomology is
equal to the Lie-algebraic cohomology of some ideal I of this super-algebra.
In Section 5 we consider the flat space limit and in particular study the
zero momentum states. One unusual finding is the existence of nontrivial
cohomology at the ghost number three.

Note added in the revised version The approach developed in this

paper is useful for clarifying the construction of integrated vertex [CMV13].

2 Pure spinor formulation of the SUSY Maxwell

theory
2.1 Supersymmetric space-time and basic constraints
Here we will remind the superspace descirption of the classical supersymmet-
ric Maxwell theory in 10 dimensions. The superspace is formed by 10 bosonic

4
coordinates xm and 16 fermionic coordinates θα . This is the supersymmetric
space-time, we will call it M :
M = R10|16 (5)
The basic superfield is the vector potential Aα (x, θ). For every α ∈ {1, . . . , 16},
the corresponding Aα is a scalar function:
Aα : M → R (6)
The equations of motion of the theory are encoded in the following construc-
tion. Let us consider the “covariant derivatives”:
∂ β ∂
∇α = α
+ Γm
αβ θ + Aα (x, θ) (7)
∂θ ∂xm
It turns out [Nil81, Wit86] that the equations of motion of SUSY Maxwell
theory are equivalent to the constraint:
• There exists a differential operator ∇m = ∂
∂xm
+ Am (x, θ) such that:
{∇α , ∇β } = Γm
αβ ∇m (8)

The nontrivial requirement of the constraint is that the LHS of (8) is pro-
portional to Γm
αβ , because the most general structure would be:
m1 m2 m3 m4 m5
Γm
αβ ∇m + Γαβ Xm1 m2 m3 m4 m5 (9)
where Xm1 ...m5 = X(x, θ)m1 ...m5 some function on the superspace. Equiva-
lently, the constraint (8) can be written:
Γαβ
m1 m2 m3 m4 m5 {∇α , ∇β } = 0 (10)
With the constraint (10) satisfied, we consider (8) as the definition of ∇m .
The pure spinor interpretation of (10) is due to [How91].

2.2 Definition of the Lie superalgebra L.

Now let us forget Eq. (7) and consider the Lie superalgebra L generated
by the letters ∇α with the relation (8). This is an infinite-dimensional Lie
superalgebra. It turns out that some properties of the SUSY Maxwell theory
can be described in terms of this algebra L. In the next Section we will
describe an application of the cohomology of L.

5
3 Lie algebra cohomology and solutions of
the SUSY Maxwell theory
3.1 Vacuum solution
(0)
Let us consider the vacuum solution Aα (x, θ) = 0. In this case ∇α = ∇α =
∂ β ∂
∂θα
+ Γm
αβ θ ∂xm . The vacuum solution is invariant under the supersymmetry
algebra susy generated by the operators Sα :
∂ β ∂
Sα = α − Γm αβ θ (11)
∂θ ∂xm
(0)
We observe that {Sα , ∇α } = 0, and in this sense the vacuum solution is
susy-invariant. It turns out that the operators ∇(0) themselves generate the
same (isomorphic) algebra susy as do Sα . This can be explained using the
interpretation of M as the coset space of susy. Let us consider the abstract
algebra susy generated by todd even
α and tm with the commutation relations:
m even
{todd odd
α , tβ } = Γαβ tm (12)
and other commutators all zero. Let us interpret xm and θα as coordinates
on the group manifold of the corresponding Lie group:
g = exp(θα todd m even
α + x tm ) (13)
Then ∇α acts as the multiplication by tα on the left, and Sα as the multipli-
odd

cation by todd
α on the right. We can consider the universal enveloping algebra
U susy as a representation of susy, by the left multiplication. Then the
regular representation can be considered as its dual, which will be denoted
(U susy)′ .

Relation between L and susy. There is an ideal I ⊂ L such that the

factoralgebra over this ideal is susy:
L/I = susy (14)
The basic constraint (8) actually implies the existence of W α such that1 :
[∇α , ∇m ] = Γm
αβ W
β
(15)
This W α is the element of I, because if ∇α were the generators of the 10-
dimensional supersymmetry algebra, then W α would be zero.
1
A thorough investigation of the consequences of the basic constraint (8) can be found
in [Maf09]

6
3.2 Deformations of solutions and cohomology
The deformation of the given solution Aα (x, θ) is:
Aα 7→ Aα + δAα (16)
where δAα should satisfy:
{∇α , δAβ } = Γm
αβ δAm (17)
The fact that the LHS is proportional to Γm αβ is a nontrivial constraint on
δAβ , and if it is satisfied than (17) becomes the definition of δAm .
Let us introduce pure spinors λα satisfying:
λα Γm β
αβ λ = 0 (18)
Using these pure spinors, Eq. (17) can be written:
Qv = 0 (19)
where Q = λα ∇α (20)
and v = λα δAα (21)
Therefore the problem of classifying the infinitesimal deformations of the
vacuum solution is reduced to the computation of the cohomology of Q.

3.3 Koszul duality and its application to deformations

Let us consider a representation V of the Lie algebra susy, and the following
version of the BRST complex:
Q Q Q
BRST
. . . −→ V ⊗C P n −→
BRST
V ⊗C P n+1 −→
BRST
... (22)
where P n is the space of polynomial functions of degree n on the pure spinors
λα . A representation V of susy is also a representation of L, because susy =
L/I.
Koszul duality2 implies that the cohomology of (22) coincides with the
Lie algebra cohomology of L:
H n (QBRST ; V ) = H n (L ; V ) (23)
2
A nice review can be found in the introductory part of [GKR06]; cohomology with
coefficients in a representation was not considered in [GKR06], but it was discussed in
[MS09]

7
 ∞
P n is a commutative algebra with quadratic relations. This
L
Notice that
n=0
algebra is Koszul dual to the universal enveloping of a Lie algebra U L.

Brief review of (23) The Koszul duality implies that the following se-
quence:
. . . −→ HomC (P 2 , U L) −→ HomC (P 1 , U L) −→ U L −→ C −→ 0 (24)
is exact, and therefore provides a free resolution of the U L-module C. This
fact depends on special properties of the quadratic constraint (1).
In (24) the action of U L on U L is by the left multiplication, and the
action of the differential involves the right multiplication by the ∇α :
dϕ(p) = ϕ(λα p)∇α (25)
Here on the right hand side we have the product of ∇α ∈ U L with ϕ(λα p) ∈
U L. In other words, for ϕ ∈ HomC (P n , U L) we have:
dϕ = µright α
U L (∇α ) ◦ ϕ ◦ µP (λ ) (26)
where µP (λα ) : P n → P n+1 is a multiplication of a polinomial by λα ∈ P 1 ,
and µright
U L (∇α ) is the right multiplication by ∇α in U L. (The composition
ϕ ◦ µ(λα ) is of the type P n → U L; we then multiply by ∇α ∈ U L.)
Since we have a projective resolution of C, we can now use it to compute
the Lie algebra cohomology of L with coefficients in V , i.e. ExtU L (C, V ). It
is the cohomology of the following sequence:
0 −→ HomU L (U L, V ) −→ HomU L (HomC (P 1 , U L), V ) −→ . . . (27)
n n+1
. . . −→ HomU L (HomC (P , U L), V ) −→ HomU L (HomC (P , U L), V ) −→ . . .
where the differential is induced by (26) and acts as follows. For f ∈
HomU L (HomC (P n , U L), V ), the df ∈ HomU L (HomC (P n+1 , U L), V ) is evalu-
ated on ϕ ∈ HomC (P n+1 , U L) as follows:
(df )(ϕ : P n+1 → U L) = f (µright α
U L (∇α ) ◦ ϕ ◦ µP (λ )) (28)
There is an isomorphism:
P n ⊗C V ≃ HomU L (HomC (P n , U L), V ) (29)
p ⊗ v 7→ [ϕ 7→ ϕ(p)v] (30)
Here “ϕ(p)v” means the action of ϕ(p) ∈ U L on the element v of the repre-
sentation V of U L. This isomorphism relates (27) to (22).

8
Special case The cohomology problem described in Section 3.2 corre-
sponds to the particular case of V = (U susy)′ . As we have just explained,
this is equivalent to the computation of the Lie algebra cohomology:
H • (L, (U susy)′ ) (31)
Notice that susy = L/I and therefore (U susy)′ is naturally a representation
of L, by the left multiplication. To calculate this cohomology, we notice that
the following complex:
. . . −→ U L ⊗C Λ2 I −→ U L ⊗C I −→ U L −→ U susy −→ 0 (32)
is a free resolution of U susy as a U L-module. This means that:
H n (QBRST ; (U susy)′ ) = H n (L, (U susy)′ ) = H n (I, C) (33)
(cf. Corollary 69 of [MS04a]). More specifically, the ghost number one vertex
operator λα δAα corresponds to the first cohomology:
 ′
1 I
H (I, C) = (34)
[I, I]
I
This has the following physical interpretation. The space [I,I] can be iden-
tified with the space of field strengths. Then (34) tells us that the classical
solutions are linear functionals on the space of field strengths. Indeed, given
a classical solution, we can compute the value of the field strenght on this
classical solution. Therefore, the space of classical solutions is expected to
be dual to the space of field strengths, as we indeed observe in (34).

I
Explicit description of [I,I] Elements W α of I were introduced in Eq.
(15). Consider the projection of W α to I/[I, I], i.e. W α mod [I, I]. We
conjecture that all the other elements of I/[I, I] can be obtained from W α
by commuting with ∇α , i.e. acting with susy. This means that all the gauge
invariant operators at the linearized level are W α and its derivatives.

4 Type IIB SUGRA in AdS5 × S 5

Note in the revised version The constructions of this paragraph can
be illustrated by explicit examples of vertex operator, corresponding to the
β-deformation These examples are constructed in [CMV13].

9
4.1 BRST complex
The BRST complex of Type IIB SUGRA in AdS5 ×S 5 [BH02, BC01, Ber05b]
is based on the coset space G/G0 where G is the Lie supergroup corresponding
to the Lie superalgebra g = psu(2, 2|4) and G0 is the subgroup corresponding
to g0̄ = so(1, 4) ⊕ so(5). A Z4 -grading of g plays an important role. The
generators of g are denoted:

t3α of degree 3, α ∈ {1, . . . , 16}

t1α̇ of degree 1, α̇ ∈ {1, . . . , 16}
t2n of degree 2, n ∈ {0, . . . , 9} (35)
t0[mn] of degree 0

The subalgebra g0̄ is generated by t0[mn] , g3̄ by t3α , g1̄ by t1α̇ , and g2̄ by t2m .
The index [mn] of t0[mn] runs over a union of two sets: the set of choices of
2 different elements m, n from {0, . . . 4}, and the set of choices of 2 different
elements m, n from {5, . . . , 9}. This corresponds to the split of g0̄ into the
direct sum of so(1, 4) and so(5). Both t3α and t1α̇ transform as spinors of both
so(1, 4) and so(5) under the adjoint action of g0̄ , and t2m transform as vectors.
The BRST complex computing supergravity excitations on the back-
ground AdS5 × S 5 is:
Q Q
QBRST
BRST
. . . −→ Homg0̄ (U g , P n ) −→
BRST
Homg0̄ U g , P n+1 −→ . . . (36)

where P n is the space of polynomials functions of the order n of two inde-

pendent pure spinors λL and λR :

λαL fαβ m λβL = 0 , λα̇R fα̇β̇ m λβ̇R = 0 for m ∈ {0, . . . , 9} (37)

where f•• • are the structure constants of g, and QBRST is given by:

QBRST = QLBRST + QR
BRST (38)
where QLBRST = λαL L(t3α ) (39)
and QRBRST = λα̇R L(t1α̇ ) (40)

Here L(t) is the left multiplication by t. We will use the notation P p,q
for theLspace of polynomials of the order p in λL and q in λR . Therefore
P n = p+q=n P p,q .

10
 More generally, we can consider the cohomology with coefficients in an
arbitrary representation V of g:
Q Q Q
BRST
. . . −→ V ⊗g0̄ P n −→
BRST
V ⊗g0̄ P n+1 −→
BRST
... (41)

The cohomology of this complex3 will be denoted H n (QBRST ; V ). With

this notation, the cohomology of the “standard” BRST complex (36) is
H n (QBRST ; (U g)′ ). These complexes were studied in [BC01, Mik11b,
Mik11a].
It is useful to consider a filtration F p on the space of vertex opera-
tors, corresponding to the powers of λR . We will consider an element of
Homg0̄ (U g , P n ) to be of the order p if it goes like O(λpR ) when λR → 0.
The space of such operators will be donoted F p Homg0̄ (U g , P n ). This is
a decreasing filtration, i.e. . . . ⊃ F p ⊃ F p+1 ⊃ F p+2 ⊃ . . . This allows
us to calculate the cohomology of QBRST using some approximation scheme,
starting from the cohomology of QLBRST and considering QR BRST as a small
correction. The first approximation is:

E2p,q = H p (QR q L
BRST ; H (QBRST ; V )) (42)

4.2 Lie algebra formed by the covariant derivatives

Now we will introduce some infinite-dimensional Lie algebra, which we will
use later to study the cohomology of the complexes (36) and (41).

Definition of the Lie algebra Ltot . We will consider the infinite-dimensional

super Lie algebra generated by the following letters:

∇Lα , ∇R 0
α̇ , t[mn] (43)
3
Frobenius reciprocity implies a relation between (36) and (41), see [Mik11b].

11
where the indices α, α̇ and [mn] run over the same sets as in (35), and with
the following relations:

{∇Lα , ∇Lβ } = fαβ m ∇Lm (44)

{∇R R m R
α̇ , ∇β̇ } = fα̇β̇ ∇m (45)
{∇Lα , ∇R
β̇
} = fαβ̇ [mn] t0[mn] (46)
[t0[mn] , ∇Lα ] = f[mn]α β ∇Lβ (47)
[t0[mn] , ∇R β̇ R
α̇ ] = f[mn]α̇ ∇β̇ (48)
[t0[kl] , t0[mn] ] = f[kl][mn] [pq] t0[pq] (49)

where Eqs. (44) and (45) are the definitions of ∇Lm and ∇R
m . The coefficients
f•• • are the structure constants of psu(2, 2|4) in the basis (35). We will
introduce the following notation for this Lie algebra:

Ltot = LL + LR + g0̄ (50)

where the sum is as linear spaces. More details are in [Mik13].

Grading. We will introduce on Ltot a Z-grading as follows:

deg(∇Lα ) = 1
deg(∇R
α̇ ) = − 1 (51)

Definition of the ideal I ⊂ Ltot . There is an ideal I ⊂ Ltot such that

Ltot /I = g. The structure of g is explained in Eq. (35). Modulo I the
generators t0[mn] become the generators t0[mn] of g0̄ ⊂ g, ∇Lα becomes t3α , ∇Rα̇
becomes t1α̇ , and both ∇Lm and ∇Rm become t2
m . The ideal I is not invariant
under the U (1) which defines the Z-grading (51), but only under Z4 ⊂ U (1).

4.3 Lie algebra cohomology

Let us consider the relative Lie algebra cohomology4 :

H • Ltot ; g0̄ ; V


(52)
4
For introduction into the Lie algebra cohomology, see [Kna88, FF88]

12
We claim that this cohomology coincides with the BRST cohomology:

H • Ltot ; g0̄ ; V = H(QBRST ; V )

(53)

We will prove a stronger statement. Let us introduce a decreasing filtration of

the Lie algebra cochain complex in the following way. We say that a cochain
c belongs to F p C q (Ltot ; g0̄ ; V ) if c(ξ1 , . . . , ξq ) is zero whenever there are
less than p letters ∇R 3 2
α̇ among ξ1 , . . . , ξq . For example, for c ∈ F C should be
R R R R
true that c(∇α̇ , ∇β̇ ) = 0, but c(∇m , ∇β̇ ) does not have to be zero (because
∇R R
m is defined in (45) as the commutator of two ∇α̇ , i.e. has degree 2).
R
In other words, the ghost dual to ∇α̇ is considered “small of the order
ε”; the ghost dual to ∇R 2
m is considered “small of the order ε ”, etc. But all
the “left” ghosts are of the order 1. The F p C consists of cochains which are
of the order εp and higher.
Similarly, the BRST complex has a filtration by the powers of λR .
We will construct a filtered quasi-isomorphism between the relative Lie
algebra complex C • (Ltot ; g0̄ ; ; V ) and the BRST complex. A filtered quasi-
isomorphism of two filtered complexes C1• and C2• is a map of complexes
which is a quasi-isomorphism grp C1• → grp C2• for every p. A filtered quasi-
isomorphism is a quasi-isomorphism of complexes in the usual sense, if one
forgets the grading [Sta, Lemma 05S3]. This can be understood from the
point of view of spectral sequences; filtered quasi-isomorphism becomes an
isomorphism at E1•,• .
In particular, it follows that the relative Lie algebra cohomology (52)
coincides with the BRST cohomology (41).

Construction of filtered quasi-isomorphism. Let C • (Ltot ; g0̄ ; V )

denote the space of cochains in the relative Lie algebra cohomology complex
(52). Let us introduce the operation of restriction from the space of relative
cochains to the BRST complex:

R : C • (Ltot ; g0̄ ; V ) −→ V ⊗ Fun(λL , λR ) (54)

which is defined as follows. Given the cochain c ∈ C q (Ltot ; g0̄ ; V ), we have

to define Rc ∈ V ⊗ Fun(λL , λR ). By definition c is a polylinear function of q
elements of L:

ξ1 ∧ ξ2 ∧ · · · ∧ ξq 7→ c(ξ1 ∧ ξ2 ∧ . . . ∧ ξq ) (55)

13
Elements of the linear space Ltot /g0̄ are, by definition in Section 4.2, nested
commutators of ∇L s plus nested commutators of ∇R s. We define Rc as the
following function of λL and λR :

⊗q 
Rc(λL , λR ) = c λαL ∇Lα + λα̇R ∇R
α̇ (56)
for c ∈ C q (Ltot ; g0̄ ; V )

We used the following notation: ξ ⊗q means ξ ⊗ ξ ⊗ · · · ⊗ ξ . We observe:

| {z }
q times

RQLie = QBRST R (57)

Lemma: R is a filtered quasi-isomorphism.

To prove this, we consider the action of QLie on the following space:
F p C p+q (Ltot ; g0̄ ; V )
grp C p+q (Ltot ; g0̄ ; V ) = =
F p+1 C p+q (Ltot ; g0̄ ; V )
p
M
= C q+r (LL ; V ) ⊗g0 grp C p−r (LR ; C) (58)
r=0

We observe that:
1. The action of QLie on grp C p+q (Ltot ; g0̄ ; V ) coincides with the ac-
[H • (LL ,V )] [H • (LR ,C)]
on pr=0 C q+r (LL ; V ) ⊗g0
L
tion of the operator QLie + QLie
grp C p−r (LR ; C)

2. The restriction map R is only nonzero on the r = 0 term. It inter-

twines this complex with the left BRST complex, which has the BRST
operator QL = λαL t3α . In other words, it is a morphism of complexes:
p+q
grp C p+q (Ltot ; g0̄ ; V ) −→ grp CBRST (59)

With these two observations, the Koszul isomorphisms:

H q (LL ; V ) ≃ H q (QLBRST ; V ) (60)

p R
H (L ; C) ≃ H p
(QR
BRST ; C) = Fun(λ⊗p
R ) (61)
p+q
imply that grp R : grp C p+q (Ltot ; g0̄ ; V ) −→ grp CBRST is a quasi-
isomorphism, i.e. R is a filtered quasi-isomorphism.

14
4.4 An analogue of the Koszul resolution
In fact, it is possible to glue two Koszul resolutions (one for LL and another
for LR ) along g0̄ , as we will now explain5 . Similarly to (24), consider the
following BRST-type complex:

0 −→ C −→ Homg0̄ (U Ltot , C) −→ Homg0̄ (U Ltot , P 1 ) −→ . . . (62)

. . . −→ Homg0̄ (U Ltot , P n ) −→ Homg0̄ (U Ltot , P n+1 ) −→ . . .

where the differential acts as follows:

dϕ = µP (λαL ) ◦ ϕ ◦ µright L α̇ right R

U Ltot (∇α ) + µP (λR ) ◦ ϕ ◦ µU Ltot (∇α̇ ) (63)

(notations as in (24)), and Homg0̄ means linear maps invariant under the
following action of g0̄ :

(η.ϕ)(x) = ϕ(xη) + η [mn] t0[mn] ϕ (64)

We will call the two terms on the right hand side of (63) dL ϕ and dR ϕ. We
will introduce the abbreviated notation for the terms of (62):

0 −→ C −→ X 0 −→ X 1 −→ . . . (65)
There is a bigrading: X n = p+q=n X p,q where X p,q = Homg0̄ (U Ltot , P p,q );
L

notice that dL : X p,q → X p+1,q and dR : X p,q → X p,q+1 .

We will now prove that (62) is a (U Ltot , U g0̄ )-injective (U Ltot , U g0̄ )-exact
resolution of C in the sense of [Hoc56].

Proof Being (U Ltot , U g0̄ )-injective follows from Section 1 of [Hoc56] (Lemma
1). Note that every term of (62) is a direct sum of finite-dimensional represen-
tations of g0̄ . This implies that the kernel and the image of every differential
is a direct g0̄ -submodule as required in [Hoc56]. It remains to prove the ex-
actness. We will prove the equivalent statement, that the cohomology of the
truncated complex:
0 −→ X 0 −→ X 1 −→ . . . (66)
is only nonzero in the zeroth term: H 0 = C. We will use the spectral
sequence of the bicomplex d = dL + dR . Let us first calculate the cohomology
5
Note in the revised version: we are greatful to the referee of [CMV13] for pointing out
an error in the original version of this subsection

15
of dL . We will “normal order” the elements of U Ltot by putting elements of
U LR to the left and elements of U LL to the right. This gives an isomorphism
of linear spaces:
Homg0̄ U Ltot , P n−p, p = HomC U LL , PLn−p ⊗ HomC U LR , PRp

(67)
The differential dL only acts on the HomC U L , PL , while HomC U LR , PRp
n−p
L

is “inert”. The action of the differential on HomC U LL , PLn−p is the

same as in the Koszul complex of U LL . Therefore the cohomology of dL

is HomC U LR , PRp . The action of dR on the cohomology of dL is the same

as the action of the differential in the Koszul complex of U LR . Therefore

H(dR , H(dL )) = C, corresponding to constant ϕ. This completes the proof.

Corollary This means that for any U Ltot -module W , the Ext(U Ltot ,U g0̄ ) (W, C)
can be computed as the cohomology of the following complex:
. . . −→ HomU Ltot W , Homg0̄ (U Ltot , P n ) −→

−→ HomU Ltot W , Homg0̄ (U Ltot , P n+1 ) −→ . . .

(68)
As in Section 3.3, there is an isomorphism of complexes (68) and (41):
HomU Ltot W , Homg0̄ (U Ltot , P n ) ≃ Homg0̄ (W, P n )


(69)
f 7→ [w 7→ f (w)(1)] (70)
If W is semisimple as a representation of g0̄ , then this shows that Ext(U Ltot ,U g0̄ ) (W, C)
can be identified with the cohomology of (41) for V = W ′ .

Variation Similarly, we can consider the following projective resolution:

. . . −→ (P n+1 )′ ⊗g0̄ U Ltot −→ (P n )′ ⊗g0̄ U Ltot −→ . . . (71)
. . . −→ (P 1 )′ ⊗g0̄ U Ltot −→ C ⊗g0̄ U Ltot −→ C −→ 0
where the differential acts as follows:
∂(s ⊗ ξ) = (s ◦ µP (λαL )) ⊗ ξ∇Lα + (s ◦ µP (λα̇R )) ⊗ ξ∇R
α̇ (72)
This means that Ext(U Ltot ,U g0̄ ) (C, V ) can be computed as the cohomology of
the following complex:
. . . −→ HomU Ltot (P n )′ ⊗g0̄ U Ltot , V −→

−→ HomU Ltot (P n+1 )′ ⊗g0̄ U Ltot , V −→ . . .

(73)

16
As in Section 3.3, there is an isomorphism of complexes (73) and (41):

HomU Ltot ((P n )′ ⊗g0̄ U Ltot , V ) ≃ P n ⊗g0̄ V (74)

f 7→ [λ 7→ f (λ ⊗ 1)] (75)

The expression [λ 7→ f (λ ⊗ 1)] on the right hand side of (75) denotes an

element of P n ⊗g0̄ V , understood as a g0̄ -invariant polynomial function of
pure spinors of the order n, whose value on a pair of pure spinors λ = (λL , λR )
is defined as follows. Since λ can be interpreted as an element of (P n )′ , we
can consider λ ⊗ 1 an element of (P n )′ ⊗g0̄ U Ltot ; then we can act on it by
f ∈ HomU Ltot ((P n )′ ⊗g0̄ U Ltot , V ).
Eq. (74) is another proof of (53).

4.5 Reduction to the cohomology of the ideal I ⊂ Ltot

The following construction works for an arbitrary completely reducible rep-
resentation A of g0̄ . Given such an A, let us consider H n (QBRST ; V ) in the
special case:

V = HomC U g ⊗g0̄ A , C (76)
According to Section 4.3 H n (QBRST ; V ) is equivalent to H n (Ltot ; g0̄ ; V ),
which in the case (76) is the same as Extn(U Ltot , U g0̄ ) (U g ⊗g0̄ A ; C) [Hoc56].
Consider the following complex of U Ltot -modules:

. . . −→ U Ltot ⊗g0 (Λ2 I ⊗C A) −→ U Ltot ⊗g0 (I ⊗C A) −→

−→ U Ltot ⊗g0̄ A −→ U g ⊗g0̄ A −→ 0 (77)

Here the action of g0̄ on Λp I ⊗C A is the sum of the adjoint action on I

and the action on A. The complex (77) is a (U Ltot , U g0̄ )-projective and
(U Ltot , U g0̄ )-exact resolution of U g ⊗g0̄ A as a U Ltot -module, in the sense
of [Hoc56]; see Appendix A. Therefore:

H n Ltot ; g0̄ ; HomC U g ⊗g0̄ A , C = Homg0̄ (A , H n (I))

(78)

Geometrical interpretation Consider the case when A is a finite-dimensional

representation. With V defined by (76) the BRST complex of (41) is:

Homg0̄ U g ⊗g0̄ A , P •


(79)

17
Geometrically, this is the space of A′ -valued functions fa (g, λ3 , λ1 ) where the
index a enumerates a basis of A′ , such that for h ∈ G0̄ :

fa (hg, hλ3 h−1 , hλ1 h−1 ) = fa (g, λ3 , λ1 ) (80)

fa (gh, λ3 , λ1 ) = fb (g, λ3 , λ1 )ρba (h) (81)

More precisely, this is the space of Taylor series of sections of the pure
spinor bundle over AdS5 × S 5 ; the universal enveloping algebra is the space
of finite linear combinations, i.e. we do not care about the convergence of the
Taylor series f . Equation (80) says that f is a section of a bundle over the
homogeneous space. On the other hand, Eq. (81) requires that f transform
in a fixed representation A′ under the group G0̄ of global rotations around
g = 1.
The space of Taylor series, as a representation of the global rotations G0̄ ,
is the direct sum of infinitely many finite-dimensional representations:
M
Homg0̄ (U g , P • ) = A ⊗ Homg0̄ U g ⊗g0̄ A , P •


(82)
A

Therefore (78) implies that:

H n ( QBRST , (U g)′ ) = H n (I) (83)

Action of the global symmetries Notice that g naturally acts on H m (I).

This corresponds to the right action of g on the BRST complex (36), i.e. to
the global symmetries of the AdS5 × S 5 sigma-model.

4.6 Ghost number 1: global symmetry currents

The elements of H 1 (QBRST ; (U g)′ ) = H 1 (I) correspond to the global sym-
metry currents of the σ-model [Val04, Ber05a, Ber05b, BBMR11]. There are
finitely many global symmetries. We have:
 ′
1 I
H (I) = (84)
[I, I]
I
We will now show that [I,I] is a finite-dimensional representation of g, actu-
ally the adjoint representation of g.

18
Special notations for summation over repeating indices. As already
introduced in (35), the index m enumerates the basis of the vector repre-
sentation of g0̄ = so(1, 4) ⊕ so(5), and runs from 0 to 9; more precisely,
m ∈ {0, . . . , 4} enumerates vectors of so(1, 4), and m ∈ {5, . . . , 9} vectors of
so(5). For a vector v m we denote:
 m
m v if m ∈ {0, . . . , 1}
v = (85)
−v m if m ∈ {5, . . . , 9}

For two vectors v m and wm we denote:

9
X
v m wm = v 0 w0 − v i wi
i=1
X4 9
X
v m wm = v 0 w0 − v i wi + v i wi (86)
i=1 i=5

I
Proposition. As a representation of g, [I,I]
is generated by the following
objects6 :

Tm2 = ∇Lm − ∇R
m (87)
0
T[mn] = [∇Lm , ∇Ln ] − [∇R R
m , ∇n ] (88)
1
ZαL = ∇Lα − [ ∇Lm , [∇Lm , ∇Lα ] ] (89)
10
1
Zα̇R = ∇Rα̇ − [ ∇R , [∇R R
m , ∇α̇ ] ] (90)
10 m
Notice that [(∇Lm − ∇R L R
m ) , (∇n − ∇n )] ∈ [I, I] implies that:

[∇Lm , ∇Ln ] + [∇R R 0

m , ∇n ] − 2t[mn] = 0 mod [I, I] (91)

Similarly, [(∇Lm − ∇R L L L
m ), [(∇m − ∇m ), ∇α ]] ∈ [I, I] implies that:

1 mα̇ R R
∇Lα − fα [∇m , ∇α̇ ] = −ZαL mod [I, I] (92)
10
We will write “≡ 0” instead of “= 0 mod [I, I]”.
6 1
The coefficient 10 depends on the choice of normalization for ∇α ; in our conventions
fαβ = Γαβ , and the projection pr(∇m ) of ∇m to g satisfies: (adpr(∇m ) )2 |g3̄ = 1 — no
m m

summation over m.

19
 The (30|32)-dimensional linear space generated by Tm2 , T[mn]0
, ZαL , Zα̇R is
closed under the action of g. It must be the adjoint representation of g.
For example, let us consider {∇Lα , ZβL }. Modulo [I, I] this is same as
{[∇R R L
m , ∇α̇ ] , Zβ }, and using (46), (47) and (48) this is proportional to Tm .

Proof of the proposition. Let J denote the subspace of I/[I, I] generated

by the action of g on (87), (88), (89) and (90). We have to prove that J = I.
Let us consider some linear combination of commutators of ∇Lα , for example:
X
C α1 ...αq [∇Lα1 , {∇Lα2 , . . . [∇Lαq−2 , {∇Lαq−1 , ∇Lαq }] . . .}] (93)
α
⃗

Suppose that the coefficients C are such that this expression belongs to I. We
will prove that it also belongs to J, using the induction in q — the number
of commutators. Suppose that for q < n, all such expressions lie in J. We
will prove that for q = n, (93) is also in J.
Notice that:
 
X
α1 ...α5 L L L L 1 mβ R R
C [∇α1 , {∇α2 , . . . {∇αq−1 , ∇αq − fαq [∇m , ∇β̇ ] }] . . .}] ∈ J
10
α
⃗
(94)
L 1 mβ R R
because ∇α − 10 fα [∇m , ∇β̇ ] ∈ J. Therefore, it remains to prove that the
following expression belongs to J:
X
C α1 ...α5 [∇Lα1 , {∇Lα2 , . . . {∇Lαq−1 , fαq mβ [∇R R
m , ∇β̇ ]}]}] (95)
α
⃗

(notice that it automatically belongs to I). When we commute ∇R with ∇L ,

the number of commutators drops and we are left with q − 4 commutators.
This provides the step of the induction.

Calculation of {∇Lα , Zα̇R } and {∇R L

α̇ , Zα }. Here we will prove that both
{∇Lα , Zα̇R } and {∇R L
α̇ , Zα } are proportional to fαα̇
[mn] 0
T[mn] , and [∇m , Tn2 ] is
proportional to fmn [pq] T[pq]
0
. Let us define ∇R L
α and ∇α̇ so that:

[∇Lm , ∇Lα ] = fmα α̇ ∇Lα̇ (96)

[∇R R
m , ∇α̇ ] = fmα̇ ∇R
α
α (97)

20
That the RHS of (96) is proportional to fmα α̇ and the RHS of (97) is pro-
portional to fmα̇ α follows from (44) and (45).
To calculate {∇Lα , Zα̇R }, {∇R L 2
α̇ , Zα } and [∇m , Tn ] we start with the
following observation:

{∇Lα , Zα̇R } + {∇R L

α̇ , Zα } ≡ 0 (98)

This follows from:

0 ≡ {∇Lα − ∇R L R L L R R 0
α , ∇α̇ − ∇α̇ } = {∇α , ∇α̇ } + {∇α , ∇α̇ } − 2tαα̇ (99)

Also notice:

{[∇Lm , ∇Lβ̇ ] , ∇R L β L R L
α̇ − ∇α̇ } ≡ fmβ̇ {∇β , ∇α̇ − ∇α̇ } =

= [∇Lm , {∇Lβ̇ , ∇R L L L R L
α̇ − ∇α̇ }] − {∇β̇ , [∇m , ∇α̇ − ∇α̇ ]} ≡

≡ − fα̇β̇ n [∇Lm , ∇Ln − ∇R γ L R L

n ] − fmα̇ {∇β̇ , ∇γ − ∇γ } (100)

This implies:

fmβ̇ β {∇Lβ , Zα̇R } − fmα̇ γ {∇R

β̇
, ZγL } = −fα̇β̇ n [∇Lm , ∇Ln − ∇R
n] (101)

Similarly:

fmβ β̇ {∇R
β̇
, ZαL } − fmα γ̇ {∇Lβ , Zγ̇R } = −fαβ n [∇R R L
m , ∇n − ∇ n ] (102)

Taking into account (98), we get the following system of equations for Xαα̇ =
{∇Lα , Zα̇R } and Xmn = [∇Lm , ∇Ln − ∇Rn ]:

2fm(α̇| γ Xγ|β̇) + fα̇β̇ n Xmn = 0 (103)

2fm(α γ̇ Xβ)γ̇ + fαβ n Xmn = 0 (104)
0
This system of equations has the following solution, which defines T[pq] :

Xαα̇ = fαα̇ [pq] T[pq]

0
(105)
[pq] 0
Xmn = − fmn T[pq] (106)

We have to prove that there are no other solutions. Let us use the identity:
n
fm αβ fαβ n = 16 δm (107)

21
Contracting (103) and (104) with f α̇β̇ k and f αβ k we get:

2fmα̇ γ fk α̇β̇ Xγ β̇ + 16 Xmk = 0 (108)

2fmα γ̇ fk αβ Xβ γ̇ + 16 Xmk = 0 (109)

This implies:
fmα̇ γ fk α̇β̇ Xγ β̇ + fkα̇ γ fm α̇β̇ Xγ β̇ = 0 (110)
Let us assume that the pair (m, k) is such that:
either m ∈ {0, . . . , 4} and k ∈ {5, . . . , 9}
or m ∈ {5, . . . , 9} and k ∈ {0, . . . , 4};
then (110) implies that for such pairs (m, k) the expression fmα̇ γ fk α̇β̇ Xγ β̇ is
symmetric under the exchange m ↔ k. But Xmk is always antisymmetric
under such an exchange. Therefore Eq. (108) implies that Xmk is only
nonzero when either both m and k belong to {0, . . . , 4}, or both m and k
belong to {5, . . . , 9}. This means that Xmk is proportional to fmk • , and we
0
can define T[pq] from (105). Then (103) gives:
 
2fm(α̇| γ Xγ|β̇) − fγ|β̇) [pq] Y[pq] = 0 (111)

which implies that Xαα̇ = fαα̇ [pq] Y[pq] .

I
To summarize, [I,I]
is a finite-dimensional space, the adjoint representa-
tion of g.

4.7 Ghost number 2: vertex operators

The cohomology group H n (I) is a linear space dual7 to the homology Hn (I).
The vertex operators correspond to H 2 (I) = (H2 (I))′ . The linear space
H2 (I) consists of the expressions of the form:
X
a= xi ∧ yi (112)
i
X
[xi , yi ] = 0 (113)
i

7
This is the Poincaré duality, Section VI.3 of [Kna88].

22
where xi and yi are elements of I, with the equivalence relations:
a ≃ a + [x, y] ∧ z + [y, z] ∧ x + [z, x] ∧ y (114)
We do not have the complete analysis at the ghost number two. It must be
true that H2 (I) correspond to the space of gauge-invariant8 operators at a
marked point in AdS5 × S 5 . This is an infinite-dimensional representation of
g. The simplest element of H2 (I) is:
O = C αα̇ (∇Lα − WαR ) ∧ (∇R L
α̇ − Wα̇ ) (115)
where C αα̇ is the coefficient in the decomposition of the quadratic Casimir
operator of psl(4|4):
C = C mn t2m t2n + C [mn][pq] t0[mn] t0[pq] + C αα̇ t3α t1α̇ + C α̇α t1α̇ t3α (116)
This should correspond to the value of the dilaton. It should be possible to
obtain other fields by acting on (115) with ∇Lα and ∇Rα̇ .

Note in the revised version : To show that (115) is nonzero in coho-

mology, let us evaluate it on the beta-deformation vertex [CMV13]:
⟨ O , V ⟩ = B ab C αα̇ g −1 π(∇Lα + 3WαR )g a g −1 π(∇R L



α̇ + 3Wα̇ )g b g=1 =

= 16B α̇α Cαα̇ ̸= 0 (117)

This is generally speaking nonzero. Therefore O does represent a nontrivial
cohomology class.

5 Flat space limit

In this section we will study the cohomology of the BRST operator in flat
space.
In flat space Ltot = LL ⊕ LR . The limit of the BRST complex (36) is:
   
α ∂ m β ∂ α̂ ∂ m β̂ ∂
QSUGRA = λL + Γαβ θL m + λR + Γα̂β̂ θR m (118)
∂θLα ∂x ∂θRα̂ ∂x
acting on functions of θL , θR , x, λL , λR .
8
Gauge invariance means is the diffeomorphism invariance plus various gauge symme-
tries of the Type IIB SUGRA

23
5.1 Ghost number 1.
I
The space [I,I]
is generated by ∇Lm − ∇R L L α α̇
m , [∇m , ∇n ], WL and WR . We
observe:
[∇Lm , ∇Ln ] = − [∇R R
m , ∇n ] mod [I, I] (119)
As a representation of susy, this space should be the dual to susy+Lorentz.
We observe:
γ 1 n
{∇(α , Γm L
β)γ WL } ≡ Γαβ [∇n , ∇m ]
L
(120)
2
As explained in [Maf09], Eq. (120) implies that ∇α WLγ is proportional to
(Γmn )γα [∇Ln , ∇Lm ].

5.2 Ghost number 2.

We do not have the complete analysis at the ghost number two. The RR
should correspond to WLα ∧ WRα̇ . The NSNS 3-form field strength H = dB
should correspond to:
Hklm = (∇L[k − ∇R L L
[k ) ∧ [∇l , ∇m] ] (121)
The following expression
′
Rklmn = [∇Lk , ∇Ll ] ∧ [∇R R L L R R
m , ∇n ] + [∇m , ∇n ] ∧ [∇k , ∇l ] (122)
should correspond to a linear combination of the curvature tensor Rklmn
and the second derivatives of the dilaton — see Eq. (133). It satisfies the
relations:
′ ′ ′
Rklmn = Rmnkl = −Rlkmn (123)
′
R[klmn] =0 (124)
′
∇[j Rkl]mn =0 (125)
′
Notice that Rk[lmn] = 0 follows from (123) and (124). Eq. (123) follows
immediately from (122). Here is the proof of (124):
′
Rklmn = [∇Lk , ∇Ll ] ∧ [∇R L R L
m − ∇m , ∇n − ∇n ] + ((kl) ↔ (mn))
′
⇒ R[klmn] = 2 [∇L[k , ∇Ll ] ∧ [∇R L R L
m − ∇m , ∇n] − ∇n] ] ≡

≡ 4 [[∇L[k , ∇Ll ] , ∇R L R L
m − ∇m ] ∧ (∇n] − ∇n] ) = 0 (126)

24
— here [[∇L[k , ∇Ll ], ∇Lm] ] = 0 because of the Jacobi identity. To prove (125)
we observe that when calculating ∇j ϕ for any element ϕ of H2 (I), we can use
either ∇Lj ϕ or ∇R j ϕ. Since both terms on the right hand side of (122) are in
H2 (I), we are free to use ∇Lj when calculating ∇j ([∇Lk , ∇Ll ] ∧ [∇R R
m , ∇n ]) and
∇R L L R R
j when calculating ∇j ([∇m , ∇n ] ∧ [∇k , ∇l ]). Those are both zero because
of the Jacobi identity.

Mismatch. It turns out that the linearized SUGRA equations of motion

are not satisfied, because ∇k Hklm ̸= 0. Using the identities from Appendix
B of [Maf09], we derive using (121):
2 1
∇k Hklm = − [∇Lk , ∇Ll ] ∧ [∇Lk , ∇Lm ] + (∇Lk − ∇R L L L
k ) ∧ [∇k , [∇l , ∇m ]] +
3 3
1 β
+ (∇L[l − ∇R α
[l ) ∧ Γm]αβ {WL , WL } (127)
3
However, the derivatives of ∇k Hklm are all zero9 :

∇n ∇k Hklm = 0 (128)

therefore this is a “zero mode effect”. Moreover, we have:

∇k Hklm = ∇[l ALm] = ∇[l AR

m] (129)
2 1 β
where AL L R L L α
m = (∇n − ∇n ) ∧ [∇n , ∇m ] + Γαβm WL ∧ WL
3 3
2 R 1 β
AR L R R α
m = (∇n − ∇n ) ∧ [∇n , ∇m ] + Γαβm WR ∧ WR (130)
3 3
Notice that ALm and AR
m are both in H2 (I).

The dilaton The difference ALm − AR m should be identified with the first
derivative of the dilaton ∂m ϕ. Notice that:
4 L L
∇n (ALm − AR R R
m ) = [∇k , ∇(m ] ∧ [∇n) , ∇k ] (131)
3
This is in agreement with the statement that (122) is a linear combination
of the Riemann-Christoffel tensor Rklmn and the derivatives of the dilaton
9
Since the homology of I is I-invariant, we can calculate either ∇L k
n ∇ Hklm or
∇R k
n ∇ Hklm ;
R k
it is easier to calculate ∇n ∇ Hklm

25
∂[l gk][m ∂n] ϕ. Indeed, we have:

3
g lm [∇Lk , ∇Ll ] ∧ [∇R R L L R R L R
m , ∇n ] + [∇m , ∇n ] ∧ [∇k , ∇l ] − ∇n (Ak − Ak ) = 0
4
(132)

which is the Einstein’s equation Rkn = 0 for the Ricci tensor Rkn = g lm Rklmn ,
if we identify:

[∇Lk , ∇Ll ] ∧ [∇R R L L R R

m , ∇n ] + [∇m , ∇n ] ∧ [∇k , ∇l ] =
= Rklmn + ∂[l gk][m ∂n] ϕ (133)

where Rklmn is the Riemann-Christoffel tensor in the Einsten frame, and

∂n ϕ = 38 (ALn − AR L R
n ). Also observe that ∇n (An − An ) = 0 — the Klein-Gordon
equation for the dilaton. Indeed:

[∇Lk , ∇Ll ] ∧ [∇R R

k , ∇l ] =
= [(∇Lk − ∇R L R R R
k ), (∇l − ∇l )] ∧ [∇k , ∇l ] ≃
β
≃ 2(∇Lk − ∇R R R R L R α
k ) ∧ [∇l , [∇l , ∇k ]] = −(∇k − ∇k ) ∧ Γkαβ {WR , WR } ≃
β
≃ Γkαβ [∇R α
k , WR ] ∧ WR = 0 (134)

(We used the Dirac equation Γkαβ [∇R α

k , WR ] = 0.)

Unphysical operator We have seen that the difference ALm − AR m corre-

sponds to the derivative of the dilaton: ∂m ϕ. But the sum ALm + AR
m presents
a problem. Observe that:

∇l (ALm + AR L R
m ) = ∇[l (Am] + Am] ) (135)
∇k ∇l (ALm + AR
m) = 0 (136)

This means that the first derivative of (ALm + AR

m ) is a constant.

Relation to the results of [BBMR11, Mik12] This mismatch is not

surprizing. We know from [BBMR11] that the zero momentum states are not
correctly reproduced as the cohomology of the “naive” BRST complex (36).
Therefore we do expect a mismatch in the zero mode sector of the space of
local operators.

26
 A state on which ALm +AR m is nonzero is described in [Mik12]. It is obtained
as the flat space limit of the nonphysical AdS vertex of [BBMR11] with the
internal commutator taking values in g2̄ (using the notations of Section 4.1).
In this case ALm +AR m is constant — the gradient of the “asymmetric dilaton”.
Besides being constant, ALm + AR m can also be depending on x linearly. To
L R
obtain the state with Am + Am depending linearly on x, we have to consider
the flat space limit of the nonphysical vertex Bab j a ∧ j b with the internal
commutator f ab c Bab taking values in g0̄ [BBMR11, Mik12]. It depends on a
constant antisymmetric tensor Bmn . The leading term in the flat space limit
is a trivial constant NSNS B-field Bmn dxm ∧ dxn , which can be gauged away.
Discarding the terms with θ’s, the leading nontrivial term is:
4 4
!
X X
Bmn dxm ∧ xn (dxk xk ) − dxn (xk xk ) (137)
k=0 k=0

This does not solve the SUGRA equations ∂ n Hnml = 0, instead ∂ n Hnml is
proportional to Bmn dxm ∧ dxn — a constant 2-form.
In terms of the unintegrated vertex, the observable ALm + AR m should be
identified as follows. It is proportional to ∂ n Bmn in the gauge where the
vertex has ghost number (1, 1), i.e. only λL λR terms, no λL λL and λR λR
terms10 .

Nonphysical operator: summary Let us denote:

[∇Lk , ∇Ll ] ∧ [∇R R L L R R

m , ∇n ] + [∇m , ∇n ] ∧ [∇k , ∇l ] = Rklmn (138)
(∇L[k − ∇R
[k ) ∧ [∇Ll , ∇Lm] ] = Hklm = ∂[k Blm] (139)
A±m = ALm ± AR
m (140)

We get the following equations of motion:

3
g lm Rklmn = ∇(k A− n) (141)
4
−
0 = ∇[k An] (142)
∇k Hklm = ∇[l A+
m] (143)
0 = ∇(l A+
m) (144)
10
If we try to change the gauge Bmn → Bmn + ∂[m Λn] to get rid of ∂ n Bmn , this would
generate some λL λL and λR λR terms [BBMR11].

27
The gradient of the dilaton corresponds to A− +
n , while An does not have a clear
interpretation in the Type IIB supergravity. The “observable” A+ n is dual to
the unphysical vertex of [Mik12]. The unphysical vertex is not BRST trivial.
However, as we explained in [Mik12], it should be thrown away because it
leads to a quantum anomaly in the worldsheet sigma-model at the 1-loop
level.

Generic element of H2 (I) The “generic” element is:

O = xL ∧ xR (145)

where xL ∈ I ∩ LL and xR ∈ I ∩ LR . Notice that the following expression:

(∇m xL ) ∧ xR − xL ∧ (∇m xR ) (146)

is zero in homology, i.e. exact:

(∇m xL ) ∧ xR − xL ∧ (∇m xR ) = δ((∇Lm − ∇R

m ) ∧ xL ∧ xR ) (147)

Indeed, the generic gauge-invariant SUGRA operator can be understood as

the product of two gauge-invariant Maxwell operators OL and OR , with the
↔
condition that OL ∂x∂m OR = 0. The zero momentum special operators of
the form (129) are not of this form.

5.3 Higher ghost numbers

This section was added in the revised version of the paper. We have
previously claimed that the cohomology at the ghost number higher than 2
vanishes. We are greateful to the referee for insisting that we present a proof
of this statement. Upon careful examination, it turns out that the statement
is wrong. There is some nontrivial cohomology at least at the ghost number
3. Here we will only do a preliminary analysis:

• We prove that the cohomology at the ghost number > 4 vanishes.

• We give an example of the nontrivial cohomology class at the ghost

number 3.

28
We suspect that the cohomology at the ghost numbers 3 and 4 is a finite-
dimensional space, and is in some way related to the unphysical states of
[BBMR11, Mik12].
We will start by proving the vanishing theorem for the super-Maxwell co-
homology at the ghost number higher than 1. We will then point out that the
SUGRA BRST complex is amlost the tensor product of two super-Maxwell
complexes (the “left sector” and the “right sector”). If it were, literally, the
tensor product, that would indeed imply the vanishing theorem at the ghost
number > 2. But in fact, even in flat space there is some “interaction” be-
tween the left and the right sector, and this leads to a nontrivial cohomology
at least at the ghost number 3.

5.3.1 Super-Maxwell BRST complex

The cohomology of the super-Maxwell BRST complex:
 
α ∂ m β ∂
QSMaxw = λ + Γαβ θ (148)
∂θα ∂xm
is only nontrivial at the ghost numbers 0 and 1.

Sketch of the proof This fact is well-known in the pure spinor formal-
ism. At the ghost number 0, the cohomology is formed by the constants (no
dependence on λ, x and θ). At the ghost number 1, the cohomology is the
solutions of the free Maxwell equation and the free Dirac equation. The van-
ishing of the cohomology at the ghost number 2 is equivalent to the following
two statements: 1) for any current jm such that ∂m jm = 0 always exists the
gauge field Fmn satisfying ∂[k Flm] = 0 and ∂m Fmn = jn and 2) for any spinor
ψ exists a spinor ϕ such that Γm ∂m ϕ = ψ. The vanishing of the cohomology
at the ghost number 3 is equivalent to the statement that for any ρ exists
jm such that ∂m jm = ρ. All these facts are proven in any graduate course of
classical electrodynamics.

5.3.2 Type IIB BRST complex

The BRST complex of Type IIB in flat space is almost the tensor product of
two SMaxwell complexes:
   
α ∂ m β ∂ α̂ ∂ m β̂ ∂
QSMaxw⊗SMaxw = λL + Γαβ θL m + λR + Γα̂β̂ θR m (149)
∂θLα ∂xL ∂θRα̂ ∂xR

29
The cohomology of (149) is the tensor product of the cohomologies of two
super-Maxwell complexes. Therefore it is only nontrivial at the ghost num-
bers 0,1 and 2. However, in the Type IIB BRST complex there is no sepa-
ration of x into xL and xR . The actual BRST complex is therefore different
from (149):
   
α ∂ m β ∂ α̂ ∂ m β̂ ∂
QSUGRA = λL + Γαβ θL m + λR + Γα̂β̂ θR m (150)
∂θLα ∂x ∂θRα̂ ∂x

The difference is that the left and the right sector have a common x instead
of separate xL and xR . We also write:

QSUGRA = QL + QR (151)

where QL and QR are the first and second terms on the right hand side of
(150).

Vanishing theorem: HQn SUGRA = 0 for n > 4. Let us consider, for example,
a vertex of the ghost number 5.

Lemma Given a vertex at the ghost number 5, we can always modify it

by adding Q-exact terms so that the new vertex has only terms of the type
λ1L λ4R .
We have to prove that the terms with λ5R , λ2L λ3R , λ3L λ2R , λ4L λR and λ5L can
be gauged away. The term with λ5R is QR -closed. Suppose that the term with
p
the lowest power of θR is proportional to λ5R θR . We observe that this term
is closed under λR ∂θ∂R and therefore is equal to λR ∂θ∂R of some expression
p+1
proportional to λ4R θR . This means that we can add Q-exact terms so that
p+2
the new vertex has terms of the order λ5R starting with λ5R θR . An induction
5
by p implies that the terms containing λR can be all gauged away. Similarly,
we can gauge away terms proportional to λ5L , then terms proportional to
λ4L λR , then λ3L λ2R , then λ2L λ3R . This proves the Lemma.
Now we are left with the terms proportional to λ1L λ4R . In this gauge
the vertex operator is both QR -closed and QL -closed. Let us look at the
expansion in powers of θR . Schematically:
k+1
V = λ4R θR k


ϕk (λL , θL , x) + θR ϕk+1 (λL , θL , x) + . . . (152)

30
were every ϕj is linear in λL . We observe that all these ϕj s are annihilated
by QL (because QL V = 0 and QL does not act on θR ):
QL ϕj = 0 (153)
We also observe that in the leading term, the coefficient of ϕk is annihilated
by λR ∂θ∂R . This implies:
k+1
V = QSUGRA λ3R θR


ϕk (λL , θL , x) +
 
4 k+1 k+2
+ λR θR ϕk+1 (λL , θL , x) + θR ϕ̃k+2 (λL , θL , x) + . . . (154)

This means that we are able to increase the order of the leading term by
adding a QSUGRA -exact expression. The induction in k proves the Theo-
rem.
But is it true that HQn SUGRA = 0 for n = 3 and n = 4? It turns out that at
least for n = 3 the cohomology is nontrivial. The fact that the cohomology
at the ghost number higher than 2 is nontrivial is (for us) unexpected. We
will leave this for future research, giving here only an example.

Example of a vertex at the ghost number 3 For any constant 5-form

F , let us denote F̂ = Fklmnp Γklmnp . Consider the following coboundary of
QSMaxw⊗SMaxw :
Φ[F ] = QSMaxw⊗SMaxw Ψ[F ] (155)
where
 
p
Ψ[F ] = (θL Γ λL ) θL Γp (xm n
+ 5||xL || )F̂ Γq θR (λR Γq θR ) +
L Γm xR Γn
2

 
+ (θL Γp λL ) θL Γp xm
L Γ m f [λ R θR
4
] + gn [λ L θ 4 n
]x
L R F̂ Γq θR (λR Γq θR )
(156)
4
where f [λR θR ] is chosen so that:
 
∂ l ∂ 
λR + (θR Γ λR ) l xnR Γn F̂ Γq θR (λR Γq θR ) + f [λR θR
4
] =0 (157)
∂θR ∂xR
and g[λL θL4 ] is chosen so that:
 
∂ l ∂
(θL Γp λL )θL Γp (xm n n 4


λL + (θL Γ λL ) l L Γm Γn − 10x L ) + g [λL θ L ] =0
∂θL ∂xL
(158)

31
 4
Such f [λR θR ] and g n [λL θL4 ] exist because the expression xnR Γn F̂ satisfies the
“right” Dirac equation:
∂  n 
x Γn F̂ Γk = 0 (159)
∂xkR R
and the expression (xm n
L Γm Γn − 10xL ) satisfies the “left” Dirac equation:

∂
Γk (xm n
L Γm Γn − 10xL ) = 0 (160)
∂xkL
We will now prove that Φ[F ] depends on xL and xR only in the combination
xL + xR . Indeed, for a constant cm let us introduce Ξ[c, F ] as follows:
 
m ∂ ∂
Ξ[c, F ] = c − m Ψ[F ] =
∂xmL ∂xR
 
= (θL Γp λL ) θL Γp cm Γm xnR Γn F̂ Γq θR (λR Γq θR ) +
+ (θL Γp λL ) θL Γp cm Γm f [λR θR 4


] −
 
− (θL Γp λL ) θL Γp (xm L Γm c n
Γn − 10(x L c))F̂ Γq θR (λR Γq θR ) −
 
− g n [λL θL4 ]cn F̂ Γq θR (λR Γp θR ) (161)

and we observe that:

 
∂m ∂
QSMaxw⊗SMaxw c − m Ψ[F ] = 0 (162)
∂xmL ∂xR
 
Since QSMaxw⊗SMaxw commutes with cm ∂x∂m − ∂x∂m , Eq. (162) implies that
L R
Φ[F ] depends on xL and xR only in the combination xL + xR , and is therefore
a cocycle of QSUGRA . We will now prove that Φ[F ] is not a coboundary of
QSUGRA . We know that Φ[F ] is a coboundary of QSMaxw⊗SMaxw , i.e. once we
introduce separate xL and xR we have (155). The question is:

can we modify Ψ[F ], by adding to it something QSMaxw⊗SMaxw - (163)

closed, so that the modified Ψ[F ] is annihilated by ∂x∂L − ∂x∂R ?

In order to answer this question, it is useful to consider  c as a ghost

 and
m ∂ ∂
interpret Ξ[c, F ] as a cocycle of the nilpotent operator c ∂xm
− ∂xm acting
L R

32
on the cohomology of QSMaxw⊗SMaxw . The answer to the question (163) is
positive only if Ξ[c, F ] is a coboundary in this complex. The cohomology of
QSMaxw⊗SMaxw is the tensor product of two super-Maxwell solutions. We will
now prove that Ξ[c, F ] represents a nonzero element of:
   
1 m ∂ ∂
H c − m , SMaxw(xL ) ⊗ SMaxw(xR ) (164)
∂xm L ∂xR

Remember that super-Maxwell is a direct sum of a solution of the free

Maxwell equations and a solution of the free Dirac equation. Looking at
(161), the corresponding cocycle corresponds to the tensor product of two so-
lutions of the free Dirac equation. Such an element of SMaxw(xL ) ⊗SMaxw(xR )
can be represented as a bispinor field ψ αβ̂ (xL , xR ) satisfying:

∂ α′ β̇
Γm
αα′ ψ (xL , xR ) = 0 (165)
∂xm
L
∂ αβ̇ ′
ψ (xL , xR )Γm
β̇ ′ β̇
=0 (166)
∂xm
R

The element of (164) corresponding to Ξ[c, F ] is:

 αβ̇
ψ(c; xL , xR )αβ̇ = ĉx̂R F̂ − (x̂L ĉ − 10(xL · c))F̂ (167)

where hat over letter stands for the contraction with the gamma-matrices,
e.g. x̂R = Γm xm
R .Let us analize the possibility of (167) being in the image
m ∂ ∂
of c ∂xm
− ∂xm :
L R

 αβ̇
?
ĉx̂R F̂ − (x̂L ĉ − 10(xL · c))F̂ =
 
? m ∂ ∂ αβ̇ m n αβ̇ m n αβ̇ m n

=c − ϕ x
mn L L x + χ x x
mn L R + σ x
mn R Rx (168)
∂xm L ∂xm R

with all three ϕαmn

β̇ m n
xL xL , χαmn
β̇ m n αβ̇ m n
xL xR and σmn xR xR satisfying both (165) and
(166). Looking at the part linear in xR , this implies:
 αβ̇
αβ̇ n αβ̇ n
Γm x̂R F̂ = − 2σmn xR + χmn xR (169)

33
 ′
The left Dirac equation on χ implies Γm α β̇
αα′ χmn = 0, therefore:

 β̇
α′ β̇ n
10 x̂R F̂ = −2Γm
αα′ σmn xR (170)
α

This implies that σ is of the form:

αβ̇
σmn = − 5δmn F̂ αβ̇ + smn
αβ̇
(171)
α′ β̇
where Γm
αα′ smn = 0 (172)

for some sαmn

β̇
symmetric in m ↔ n. As we have already mentioned, σ should
satisfy the right Dirac equation:
αβ̇ n ′
σmn Γβ̇ ′ β̇ = 0 (173)

Equations (172) and (173) imply that the traces of σ and s are zero:
αβ̇ αβ̇
σmm = smm =0 (174)

but this contradicts (171) because the

 trace of δmn F̂ is not zero. This shows
that (167) is not in the image of cm ∂x∂m − ∂x∂m , and therefore it represents
L R
a nonzero element of the cohomology group (164). This implies that Φ[F ] is
a BRST-nontrivial vertex operator at the ghost number three.

Generalization The cohomology of QSMaxw⊗SMaxw at the ghost number 3 is

trivial, i.e. any cocycle with three λ’s can be represented as QSMaxw⊗SMaxw Ψ.
But sometimes Ψ cannot be chosen to depend on xL and xR through xL + xR
only. The obstacle for that is in H 1 (R10 , SMaxw ⊗ SMaxw) where R10 is
theabelian group of translations, the Lie cohomology differential is QLie =
m ∂ ∂
c ∂xm
− ∂xm . Notice that SMaxw ⊗ SMaxw splits into components:
L R

SMaxw ⊗ SMaxw = (175)

= (Maxw ⊗ Maxw) ⊕ (Maxw ⊗ Dirac) ⊕ (Dirac ⊗ Maxw) ⊕ (Dirac ⊗ Dirac)

Consider the cohomology in the sector Dirac ⊗ Dirac, and more specifically
those elements of it which have linear x-dependence. It turns out that this
cohomology is identified with the quadratic in x solutions f of the “double

34
Dirac equation” modulo solutions presentable as a sum of a solution of the
left Dirac equation and a solution of the right Dirac equation:
∂ m ∂ n α′ α̇′
Γ ′ Γ ′f (x) = 0
∂xm αα ∂xn α̇α̇
but ∄ s and σ such that: f αα̇ = sαα̇ + σ αα̇ (176)
∂ m α′ α̇ ∂ αα̇′ n
m
Γαα′ s = 0 and σ Γα̇′ α̇ = 0
∂x ∂xn
Indeed, given such an f αα̇ with the quadratic x-dependence, we construct
ψ(c) in the following way:
∂
ψ(c) = ĉΓn f (xR ) + ξ(xL , c) (177)
∂xnR
where ξ is some solution of the left Dirac equation, chosen so that QLie ψ =
0; such a solution always exists because H 2 (R10 , Dirac) = 0. Suppose
that ψ is in the image of QLie acting on the quadratic (in xL|R ) elements of
Dirac ⊗ Dirac, i.e.:
 
? m ∂ ∂
ψ(c) = c − m (σ⟨xR ⊗ xR ⟩ + χ⟨xR ⊗ xL ⟩ + ϕ⟨xL ⊗ xL ⟩) (178)
∂xmL ∂xR
The part of ψ(c) linear in xR would be:
∂ m ∂
−cm m σ⟨x⊗2 R ⟩+c χ⟨xR ⊗ xL ⟩ (179)
∂xR ∂xm
L
This implies:
∂ ∂ ∂
Γm m ψ(c)⟨xR ⟩ = 10Γn n f (xR ) = −Γm m σ⟨x⊗2 R ⟩ (180)
∂c ∂xR ∂xR
in other words f = s + σ where σ satisfies the right Dirac equation and s the
left Dirac equation. This contradicts (176).
Eq. (171) has f αα̇ = ||x||2 F̂ αα̇ with a 5-form F̂ ; there are also solutions
corresonding to a 3-form or 7-form Ĝ:
1
f = Ĝ||x||2 − x̂Γp ĜΓp x̂ (181)
52
and a 1-form or 9-form Â:
1
f = Â||x||2 − x̂Γp ÂΓp x̂ (182)
28
This means that the cohomology at the ghost number 3 at least includes
states with the quantum number of a bispinor.

35
5.3.3 Dual picture
We conjecture that the dual element of H3 (I) is of the form:

Oαβ̇ = [∇Lm , WLα ] ∧ WRβ̇ ∧ (∇Lm − ∇R

m) +
β̇
+ WLα ∧ [∇R L
m , WR ] ∧ (∇m − ∇m ) −
R

1 ′
− WLα ∧ WRβ̇ (Γmn )β̇β̇ ′ ∧ [∇R R
m , ∇n ] +
2
1 ′
+ WLα (Γmn )αα′ ∧ WRβ̇ ∧ [∇Lm , ∇Ln ] (183)
2

5.3.4 Conjecture about the vertices at the ghost number 3

Generally speaking, the physical interpretation of vertex operators is:

• Ghost number 1: global symmetries of the space-time

• Ghost number 2: infinitesimal deformations of the space-time

• Ghost number 3: obstructions to continuing the infinitesimal deforma-

tions of the space-time to the second order in the deformation param-
eter

It is natural to conjecture that the vertices at the ghost number 3 obstruct

those and only those infinitesimal deformations which are unphysical in the
sense of [Mik12].

The cohomology at the ghost numbers 3 and 4 deserves systematic investi-

gation. We hope to return to this subject in the future work.

6 Conclusion
In this paper we presented a relation between the cohomology of the pure
spinor BRST complex in AdS space and the relative Lie algebra cohomology.
We used this relation to develop a “dual” point of view on the vertex
operators in Type IIB. In this approach, instead of looking at the vertex
operators, we look at the dual linear space which is identified with the gauge-
invariant local operators of the Type IIB SUGRA. This works both in flat
space and in AdS. We observe that some elements of the BRST cohomology

36
do not correspond to any physical states, e.g. the A+ of (140). It turns
out that there are also vertex operators at the ghost number three. They
correspond to the obstructions for nonlinear deformations in the actions.
Physically, these obstructions should not be present.
Such “unphysical” elements should go away if we restrict the BRST com-
plex to the operators annihilated by the Virasoro constraints. We do not
know what this restriction means from the point of view of the Lie algebra
cohomology.
We conclude that the BRST complex (36) in AdS5 × S 5 and its flat space
limit (118) both have rich mathematical structure. But at the same time
the cohomology does not give a complete description of the supergravity
excitations. The difference is in some unphysical states. These unphysical
states have polynomial x-dependence, as opposed to the usually considered
exponential x-dependence. This polynomial (or “zero-momentum”) sector
could be important in the calculation of the scattering amplitude, because the
momentum conservation implies that the product of the scattering vertices
has zero total momentum.

A Exactness of (77)
This is similar to the proof of the exactness of the standard Koszul resolution
of the Lie algebra in [Kna88]. For any Lie algebra L, the universal enveloping
U L is filtered so that grp U L = F p U L/F p−1 U L = S p L. The differential in
our complex acts in such a way, that we can consistently define:

. . . −→ F p−2 U Ltot ⊗g0 (Λ2 I ⊗C A) −→ F p−1 U Ltot ⊗g0 (I ⊗C A) −→

−→ F p U Ltot ⊗g0̄ A −→ F p U g ⊗g0̄ A −→ 0
(184)
p
This defines a series of complexes d : Xnp → Xn−1 parametrized by an integer
p p
p, where X−1 = F U g ⊗g0̄ A, X0 = F U L ⊗g0̄ A, and Xnp = F p−n U Ltot ⊗g0
p p tot

(Λn I ⊗C A) for n > 0. At p = 0 we get the exact sequence:

0 −→ A −→ A −→ 0 (185)

37
On the other hand, the factor-complex X p /X p−1 is:
. . . −→ S p−2 Ltot /g0̄ ⊗C Λ2 I ⊗C A −→ S p−1 Ltot /g0̄ ⊗C I ⊗C A −→

−→ S p Ltot /g0̄ ⊗C A −→ S p (g/g0̄ ) ⊗C A −→ 0

(186)
This is exact, being the de Rham complex of the linear space I times functions
of additional “inert” variables corresponding to a complement to g0̄ + I in
Ltot . By induction, the complexes X p are exact for all values of p, and
therefore the complex (77) is exact.

Acknowledgments
I would like to thank Nathan Berkovits for discussions and the anonymous
referee for useful suggestions. This work was supported in part by the Min-
istry of Education and Science of the Russian Federation under the project
14.740.11.0347 “Integrable and algebro-geometric structures in string the-
ory and quantum field theory”, and in part by the RFFI grant 10-02-01315
“String theory and integrable systems”.

References
[BBMR11] O. A. Bedoya, L. I. Bevilaqua, A. Mikhailov, and V. O. Riv-
elles, Notes on beta-deformations of the pure spinor super-
string in AdS(5) x S(5), Nucl.Phys. B848 (2011), 155–215,
arXiv/1005.0049.
[BC01] N. Berkovits and O. Chandia, Superstring vertex operators in an
ads(5) x s(5) background, Nucl. Phys. B596 (2001), 185–196,
hep-th/0009168.
[Ber00] N. Berkovits, Super-Poincare covariant quantization of the su-
perstring, JHEP 04 (2000), 018, hep-th/0001035.
[Ber05a] , BRST cohomology and nonlocal conserved charges,
JHEP 02 (2005), 060, hep-th/0409159.
[Ber05b] , Quantum consistency of the superstring in AdS(5) x S(5)
background, JHEP 03 (2005), 041, hep-th/0411170.

38
[BH02] N. Berkovits and P. S. Howe, Ten-dimensional supergravity con-
straints from the pure spinor formalism for the superstring, Nucl.
Phys. B635 (2002), 75–105, hep-th/0112160.

[CMV13] O. Chandia, A. Mikhailov, and B. C. Vallilo, A construction

of integrated vertex operator in the pure spinor sigma-model in
AdS5 × S 5 , JHEP 1311 (2013), 124, arXiv/1306.0145.

[FF88] B. L. Feigin and D. B. Fuchs, Cohomology of lie groups and al-

gebras (in russian), VINITI t. 21, 1988.

[GKR06] A. L. Gorodentsev, A. S. Khoroshkin, and A. N. Rudakov, On

syzygies of highest weight orbits, arXiv/math/0602316.

[Hoc56] G. Hochschild, Relative homological algebra, Trans. Amer. Math.

Soc. 82 (1956), 246–269. MR 0080654 (18,278a).

[How91] P. S. Howe, Pure spinors lines in superspace and ten-dimensional

supersymmetric theories, Phys.Lett. B258 (1991), 141–144.

[Kna88] A. W. Knapp, Lie groups, lie algebras, and cohomology, Prince-

ton University Press, 1988.

[Maf09] C. R. Mafra, Superstring Scattering Amplitudes with the Pure

Spinor Formalism, arXiv/arXiv:0902.1552.

[Mik11a] A. Mikhailov, Finite dimensional vertex, JHEP 1112 (2011), 5,

arXiv/1105.2231.

[Mik11b] , Symmetries of massless vertex operators in AdS(5)

x S**5, Adv.Theor.Math.Phys. 15 (2011), 1319-1372,
arXiv/0903.5022.

[Mik12] , Cornering the unphysical vertex, JHEP 082 (2012),

arXiv/1203.0677.

[Mik13] , A generalization of the Lax pair for the pure spinor su-
perstring in AdS5 x S5, arXiv/1303.2090.

[Mik14] , Vertex operators of ghost number three in Type IIB su-

pergravity, arXiv/1401.3783.

39
[MS04a] M. Movshev and A. S. Schwarz, Algebraic structure of Yang-Mills
theory, arXiv/hep-th/0404183.

[MS04b] , On maximally supersymmetric Yang-Mills theories,

Nucl.Phys. B681 (2004), 324–350, arXiv/hep-th/0311132.

[MS09] , Supersymmetric Deformations of Maximally Supersym-

metric Gauge Theories, arXiv/0910.0620.

[MV09] L. Mazzucato and B. C. Vallilo, On the Non-renormalization of

the AdS Radius, JHEP 0909 (2009), 056, arXiv/0906.4572.

[Nil81] B. E.W. Nilsson, Simple ten-dimensional supergravity in super-

space, Nucl.Phys. B188 (1981), 176.

[Sta] The Stacks Project Authors, Stacks Project, http://math.

columbia.edu/algebraic_geometry/stacks-git.

[Val02] B. C. Vallilo, One loop conformal invariance of the super-

string in an AdS(5) x S(5) background, JHEP 12 (2002), 042,
hep-th/0210064.

[Val04] , Flat currents in the classical AdS(5) x S(5) pure spinor

superstring, JHEP 03 (2004), 037, hep-th/0307018.

[Wit86] E. Witten, Twistor - Like Transform in Ten-Dimensions,

Nucl.Phys. B266 (1986), 245.

40