Dissertation

submitted to the
Combined Faculty of Mathematics, Engineering and Natural Sciences
of Heidelberg University, Germany
for the degree of
Doctor of Natural Sciences

Put forward by
Fabian Hahner
Born in: Fulda (Germany)
Oral examination: 25.01.2024
The Pure Spinor Superfield Formalism
and
Twisted Supergravity

Referees:
Prof. Johannes Walcher
Prof. Owen Gwilliam
Abstract

This thesis discusses the pure spinor superfield formalism and its applications, specifi-
cally in the context of twisted eleven-dimensional supergravity. We start by developing
the pure spinor superfield formalism as a framework for the construction of supermul-
tiplets from a graded equivariant module over the ring of functions on the nilpotence
variety. This perspective establishes a connection between algebrogeometric properties
of the nilpotence variety and the physics of multiplets. Furthermore, it allows for efficient
computations by means of homological algebra. After exploring the formalism in various
examples, we extend it to the setting of derived geometry, show that this generalization
establishes an equivalence of categories, and relate it to Koszul duality. In particular, this
result establishes a method to construct superspace descriptions for any multiplet. As an
application, we provide an extensive case study of supermultiplets with six-dimensional
N = (1, 0) supersymmetry and classify all multiplets whose derived invariants for the su-
pertranslation algebra define a line bundle on the nilpotence variety. In the second part,
we consider eleven-dimensional supergravity and its twists. We compute the maximal
twist in the free perturbative limit starting from the L∞ action of the super Poincaré
algebra on the BV complex of component fields. Then, we use the pure spinor super-
field formalism to construct a generalization of Poisson–Chern–Simons theory, defined
on any supermanifold equipped with an appropriate odd distribution. This theory re-
covers Cederwall’s formulation of eleven-dimensional supergravity, Costello’s description
of the maximal twist, and gives a pure spinor lift of the interactions in the minimally
twisted theory. Compatibility between the pure spinor formalism and twisting implies
that all these theories are related by twists. Motivated by holographic duality, we use
these methods to explore (twisted) six-dimensional (2,0) supersymmetry. We give a pure
spinor construction of the decomposition of the minimally twisted eleven-dimensional
supergravity fields into E(3|6)-modules and provide an interpretation in terms of super-
geometry which hints towards a generalization in the untwisted case.

v
Zusammenfassung

Diese Arbeit behandelt den Pure-Spinor-Superfeld-Formalismus und seine Anwendungen,

insbesondere im Hinblick auf getwistete elfdimensionale Supergravitation. Zunächst ent-
wickeln wir den Formalismus als konzeptuellen Rahmen zur Konstruktion von Supermul-
tipletts aus äquivarianten Moduln über dem Ring der Funktionen auf der Nilpotenzva-
rietät. Mit dieser Perspektive werden Verbindungen zwischen den algebrogeometrischen
Eigenschaften der Nilpotenzvarietät und der Physik von Multipletts offengelegt. Zu-
sätzlich ermöglicht sie es effektive Berechnungen mithilfe von Methoden aus der homo-
logischen Algebra durchzuführen. Nachdem wir den Formalismus auf einige Beispiele
angewendet haben, erweitern wir ihn auf den Kontext der derivierten Geometrie, zeigen,
dass diese Verallgemeinerung eine Äquivalenz von Kategorien liefert, und verbinden sie
mit Koszul-Dualität. Insbesondere zeigen diese Resultate, dass der Formalismus Su-
perfeldbeschreibungen für jedes Multiplett konstruiert. Als Anwendung präsentieren
wir eine ausführliche Betrachtung zu Multipletts mit sechsdimensionaler N = (1, 0)
Supersymmetrie und klassifizieren hierbei alle Multipletts, deren derivierte Invarianten
bezüglich Supertranslationen Geradenbündel über der Nilpotenzvarietät definieren. Im
zweiten Teil wenden wir uns elfdimensionaler Supergravitation und deren Twists zu. Wir
berechnen den maximalen Twist im perturbativen freien Limes explizit ausgehend von
der L∞ -Wirkung der Super-Poincaré-Algebra auf den BV Komplex der Komponenten-
felder. Darüber hinaus verwenden wir den Pure-Spinor-Superfeld-Formalismus um eine
Verallgemeinerung von Poisson–Chern–Simons-Theorie zu konstruieren, welche auf jeder
Supermannigfaltigkeit ausgestattet mit einer passenden ungeraden Distribution definiert
ist. Diese Theorie vereinheitlicht Cerderwalls Beschreibung von elfdimensionaler Super-
gravitation, Costellos Formulierung des maximalen Twists und gibt einen Lift der Wech-
selwirkungen im minimalen Twist. Motiviert durch die holografische Korrespondenz,
nutzen wir diese Methoden um (getwistete) (2,0) Supersymmetrie in sechs Dimensio-
nen zu studieren. Wir beschreiben eine geometrische Konstruktion der Zerlegung der
Felder minimal getwisteter elfdimensionaler Supergravitation in E(3|6)-Moduln mithilfe
des Pure-Spinor-Formalismus. Das entstehende geometrische Bild liefert Hinweise für
eine Verallgemeinerung im ungetwisteten Fall.

vi
 Published Contents

This thesis is based on results which appeared in the following articles.

[EH23] R. Eager, F. Hahner. Maximally twisted eleven-dimensional supergravity. Com-

mun. Math. Phys. 398 (2023). arXiv:2106.15640.

[Eag+22] R. Eager, F. Hahner, I. Saberi, B. R. Williams. Perspectives on the pure spinor

superfield formalism. J. Geom. Phys. 180 (2022). arXiv:2111.01162.

[EHS23] C. Elliott, F. Hahner, I. Saberi. The derived pure spinor superfield formalism as
an equivalence of categories. SIGMA 19 (2023) arXiv:2205.14133

[Hah+22] F. Hahner, S. Noja, I. Saberi, J. Walcher. Six-dimensional supermultiplets from

bundles on projective spaces. (June 2022) arXiv:2206.08388

[HS23] F. Hahner, I. Saberi. Eleven-dimensional supergravity as a Calabi–Yau twofold

(April 2023) arXiv:2304.12371

[Hah+23] F. Hahner, S. Raghavendran, I. Saberi, B. R. Williams. in preparation (2023)

The presentation of the pure spinor superfield formalism in §2 and its derived generaliza-
tion in §3 is adapted from the material published in [Eag+22] and [EHS23]. Chapter §6 is
based on [EH23] and §7 is adapted from [HS23]. Finally, §8 is based on the forthcoming
work [Hah+23].

vii
viii
Acknowledgements

First, I would like to thank my supervisor Johannes Walcher for his guidance, advice,
and support throughout this project. I am especially grateful for the freedom and trust
I have enjoyed during the course of this work.

A huge thank you goes to Ingmar Saberi with whom I have collaborated on (almost)
all of the projects that make up this thesis. Your advice has not only been invaluable
to this thesis, but has inspired me in many different ways. I am especially grateful for
insightful discussions on topics too numerous to mention, often going far beyond math
and physics.

Further, I’d like to thank all the other people who collaborated with me on the papers
making up this thesis: Richard Eager, Chris Elliott, Simone Noja, Surya Raghavendran,
and Brian Williams. Working with you is a pleasure and each of you has taught me
amazing things.

Next, I want to thank Owen Gwilliam for agreeing to be the second referee for this
thesis, and Matthias Bartelmann and Hans-Christian Schultz-Coulon for serving on the
examination committee.

During the three years of my doctoral studies I had the privilege to travel to various
places and to discuss with amazing people. This was only possible due to the hospitality
and support of many institutions: LMU München, University of Massachusetts Amherst,
Boston University, Perimeter Institute for Theoretical Physics, University of Washington,
GGI in Florence, MITP in Mainz, and DESY in Hamburg. Further, I want to thank
the Cluster of Excellence Structures in general (and especially the YRC), as well as
the DFG for the funding which allowed me to write this thesis. While traveling, I
learned a tremendous amount from discussions with a number of amazing people. Here’s
a (probably incomplete) list: Ilka Brunner, Martin Cederwall, Ivan Contreras, Kevin
Costello, Niklas Garner, Owen Gwilliam, John Huerta, and Natalie Paquette. Thank
you all!

ix
Further, I want to thank all my fellow PhD students in the Physical Mathematics group
for their advice, support, and all the interesting discussions. Thanks to Michael Bleher,
Lukas Hahn, Sebastian Nill, Steffen Schmidt and Raphael Senghaas. I would also like to
thank Simon Heuveline and Jakob Ulmer for interesting discussions during their visits.
A special thanks goes to Nora Schrenk for helping with all the paperwork.

Finally, I would like to thank my friends Tristan Daus, Benjamin Haake, Thomas Junker-
mann, Anna-Sophie Schäfter, and Julia Wietzel for their wonderful companionship during
my time in Heidelberg. I want to express my deepest gratitude to Katrin Lehle who sup-
ported me at every step of this journey in a million different ways. Zu guter Letzt möchte
ich noch meiner Familie, besonders meinen Eltern, für all die Unterstützung danken, die
es mir überhaupt erst ermöglicht hat, diesen Weg zu beschreiten.

x
Contents

Abstract v

Zusammenfassung vi

Published Contents vii

Acknowledgements ix

Contents xi

1 Introduction 1
1.1 Main results of this thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 The pure spinor superfield formalism 9

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3 The pure spinor superfield formalism . . . . . . . . . . . . . . . . . . . . . 28
2.4 From multiplets to theories . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.5 Ten-dimensional super Yang–Mills theory . . . . . . . . . . . . . . . . . . 65
2.6 A bestiary of multiplets from modules . . . . . . . . . . . . . . . . . . . . 71
2.A Homotopy transfer for L∞ modules . . . . . . . . . . . . . . . . . . . . . . 83

3 Derived pure spinor superfields 87

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.2 The category of multiplets . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
3.3 Derived pure spinor superfields . . . . . . . . . . . . . . . . . . . . . . . . 97
3.4 An equivalence of categories . . . . . . . . . . . . . . . . . . . . . . . . . . 105
3.5 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

4 Six-dimensional supermultiplets from bundles on projective spaces 127

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
4.2 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
4.3 A family of multiplets from line bundles . . . . . . . . . . . . . . . . . . . 136
4.4 Antifield multiplets and duality . . . . . . . . . . . . . . . . . . . . . . . . 145
4.5 Short exact sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
4.6 The normal bundle exact sequence . . . . . . . . . . . . . . . . . . . . . . 155
4.7 The conormal bundle exact sequence . . . . . . . . . . . . . . . . . . . . . 160

xi
5 Zwischenspiel: twisting and holography 167
5.1 Twisting supersymmetric field theories . . . . . . . . . . . . . . . . . . . . 167
5.2 A panoramic view on twisted holography . . . . . . . . . . . . . . . . . . . 172

6 Maximally twisted eleven-dimensional supergravity 177

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
6.2 The two twists of eleven-dimensional supergravity . . . . . . . . . . . . . . 179
6.3 Eleven-dimensional supergravity in the pure spinor superfield formalism . 182
6.4 Twisting the free theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
6.5 Conclusions and future directions . . . . . . . . . . . . . . . . . . . . . . . 204

7 Eleven-dimensional supergravity as a Calabi–Yau twofold 207

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
7.2 Flag structures and generalized Dolbeault complexes . . . . . . . . . . . . 210
7.3 Poisson–Chern–Simons theories via derived brackets . . . . . . . . . . . . 220
7.4 Calabi–Yau twofolds from certain Gorenstein rings . . . . . . . . . . . . . 225
7.5 Eleven-dimensional supergravity, both twisted and not . . . . . . . . . . . 232

8 Differential operators and twisted (2,0) supersymmetry 241

8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
8.2 Differential operators on the canonical multiplet . . . . . . . . . . . . . . . 242
8.3 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
8.4 The minimal twist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
8.5 Untwisted physics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
8.6 Conformal supergravity from derivations . . . . . . . . . . . . . . . . . . . 258

9 Outlook 263

Bibliography 265

xii
Chapter 1

Introduction

Over the last fifty years, the construction and analysis of supersymmetric field theories
has been a rewarding endeavor for both theoretical physics and mathematics. Speak-
ing broadly, there are at least two a priori distinct aspects to this effort. The first
concerns the construction and classification of supersymmetric field theories and asks
for systematic procedures to come up with interesting supersymmetric models as well
as their interpretation, especially in terms of supergeometry. On the other hand, the
second aspect deals with the extraction of interesting mathematical quantities from a
given supersymmetric field theory, often by restricting to suitable subsectors. This thesis
explores both, eventually making the case for them to be addressed simultaneously.

Regarding the former, the essential difficulty lies in the fact that the supersymmetry
transformations typically only define a representation on the space of fields after impos-
ing the equations of motion for fermionic fields. Similarly, if gauge fields are present, the
algebra is only represented up to gauge transformations. For the purposes of quantiza-
tion, it is desirable for the symmetry to act on the full space of fields, without regard to
the dynamics; to attain this, one must often pass to a more complicated model, which
tends to involve additional “auxiliary fields” that do not change the physics of the theory.
Such auxiliary fields may or may not exist, depending on the example in which one is
interested. Relatedly, it is pleasing to think of supersymmetry as arising from the action
of particular geometric symmetries on an appropriate supermanifold, which amounts to
giving “superfield” formulations of supersymmetric theories. Since supersymmetry nec-
essarily acts on the space of superfields, giving such a formulation implicitly requires
extending the field content of the theory, in one fashion or another, by some set of aux-
iliary fields as described above. (The literature on supersymmetry is immense, and we
cannot hope to give an overview here; for foundational work in the subject, the reader

1
2 Introduction

is referred to the collection of reprints [Fer87], or the early review [Soh85] and references
therein.)

Finding systematic techniques to construct superfield formulations and extend the range
of theories for which they are available has thus been a subject of great interest. Numer-
ous approaches have been developed, including harmonic superspace [Gal+01] and the
“rheonomy” approach to supergravity theories [CdF91]. In the first part of this thesis
we address this challenge by means of the pure spinor superfield formalism. Techniques
based on the “space of pure spinors” for the construction of supersymmetric multiplets
date back more than thirty years ago to papers by Nilsson [Nil86] and Howe [How91a;
How91b]. Pure spinors were used to great effect in Berkovits’ formulation of the super-
string [Ber00], and their applications to superfield formulations have been developed in a
wide variety of examples, notably in work of Cederwall and collaborators. See [CNT01;
CNT02; BN05] for early papers, and [Ced14] for a review with references to further
literature.

There have been numerous related studies of ideas connected to the space of pure spinors,
for example by Krotov, Losev and collaborators [Ale+07; KL09] and, notably, in the work
of Movshev and Schwarz on ten-dimensional supersymmetric Yang–Mills theory [MS04;
Mov05a; Mov05b; MS06; Mov15]. See also the related work of Kapranov [Kap21].
This work was further developed in a mathematical context in [GKR07], in [GS09],
and in [Gál+16], where connections to the theory of Koszul duality were emphasized.
Relatedly, work by Movshev, Schwarz, and Xu on the Lie algebra cohomology of super-
symmetry algebras [MSX12; MSX14] made the appearance of the space of square-zero
elements in that context clear.

Remark 1.0.1. It is worth making an orienting remark on terminology at this point. Let
g = g+ ⊕ Πg− be a super Lie algebra. We refer to the space of odd square-zero elements

Y = {Q ∈ g− | [Q, Q] = 0} (1.1)

as the nilpotence variety of g [ESW21]. In the physics literature, the term “pure spinor”
is often used to refer to points in (or coordinates on) the nilpotence variety. On the
other hand, pure spinors, in the sense of Cartan and Chevalley [Che54], are defined
to be elements of the spin representation for which the dimension of the annihilator
under Clifford multiplication is maximal; the pure spinors form the minimal orbit in
the action of the spin group on the (projectivized) spin representation. If g is a super
Poincaré algebra, these terms are closely related though not identical: pure spinors have
vanishing self-brackets in the super Poincaré algebra [ESW21; ES19b], but the converse is
not always true. In ten-dimensional minimal supersymmetry, which was the first example
studied [Nil86] from this angle, the odd elements of the super Poincaré algebra consist
Introduction 3

of a single spin representation and the bracket is defined by Clifford multiplication such
that the nilpotence variety coincides with the space of pure spinors. This coincidence is
responsible for the confusing terminology; the usage of “pure spinor formalism” in this
thesis is historically, rather than logically, motivated.

Concerning the extraction of interesting mathematical quantities from a given supersym-

metric field theory, the crucial observation is that such theories posses protected subsec-
tors that preserve parts of the supersymmetry and whose observables measure interesting
mathematical quantities of the underlying spacetime manifold. Such subsectors can be
systematically extracted by twisting. To twist a given supersymmetric field theory, one
fixes an odd square-zero symmetry of the theory and then takes invariants by the one-
dimensional odd abelian super Lie algebra spanned by the chosen supercharge [Cos13a].
The twisted theory then describes a subsector of the full theory and, for twisting su-
percharges from the super Poincaré algebra, it is sensitive to underlying topological or
holomorphic structures on the spacetime manifold. This gives a way to access invariants
of manifolds as observables in twisted supersymmetric field theories, thereby achieving
the promised extraction. (The prime example of this procedure is Witten’s celebrated
expression of the Donaldson invariants of smooth four-manifolds using four-dimensional
N = 2 supersymmetry [Wit88]). Even more, as supersymmetric theories are related by
a fascinating web of dualities, the subsectors described by their twisted versions are in
correspondence too. With the twisted theories being mathematically tractable, twisting
on both sides of duality can establish new relations between the mathematical structures
on either side. This procedure has led to many fascinating insights originating from a
variety of different physical dualities (for example from mirror symmetry, phrased as a
duality between two topological phases of string theory extracted by twisting [Hor+03],
or the applications of S-duality to the geometric Langlands program [KW07], just to
name two); the AdS/CFT correspondence, however, until recently, has mostly been
fence-sitting regarding this program. In part this was due to a lacking understanding
of twisted supergravity theories as compared to twisted gauge theories. Crucially, the
definition of twisted supergravity theories is more subtle than the one of twisted su-
persymmetric field theories. These subtleties were discussed and resolved in [CL16].
As supergravity theories in ten dimensions arise as low energy effective field theories
from various string theories, their twisted cousins are closely related to topological string
theory which arises by a twist procedure on the worldsheet. Using this line of reason-
ing conjectures for twisted supergravity theories were provided and studied through the
lens of twisted holography [CL16; Cos17; Cos16; CG18]. Eleven-dimensional supergrav-
ity [CJS78] is the low energy limit of M-theory, a conjectural theory believed to unify
the different superstring theories [Wit95]. In [Cos16] investigated eleven-dimensional su-
pergravity in the omega background and conjectured a link between the maximal twist
4 Introduction

and Poisson–Chern–Simons theory.

In this thesis, we develop the pure spinor superfield formalism as a systematic tool for
the construction of supersymmetric multiplets. We apply it to a variety of problems, in
particular we use it to study twisted supergravity theories and provide powerful target
space techniques to directly study the twisted eleven-dimensional supergravity theory in
target space, without relying on any relation to the worldsheet models. We will see that
this closely links the superspace methods employed in the construction of these theories
and the computation of their twists.

1.1 Main results of this thesis

The first part of this thesis (§2-4) address the construction by means of the pure spinor
superfield formalism. In the second part (§6-8), we use these tools to study twisted
supergravity theories and their holographic duals from a target space perspective.

Let n be a supertranslation algebra acted on by a Lie algebra g0 and let g = g0 n n. We

will call such super Lie algebras of super Poincaré type (see Definition 2.3.2 for details)
and set the up pure spinor superfield formalism for such algebras.

The (derived) pure spinor superfield formalism. The first chapter of this thesis
develops the pure spinor superfield formalism as a systematic framework for the con-
struction of supersymmetric multiplet from modules over the ring of functions on the
nilpotence variety. We develop suitable computational techniques by means of homo-
logical algebra and homotopy transfer and demonstrate these in various examples. We
establish links between the algebraic geometry of the nilpotence variety and the physics of
supermultiplet; here the Gorenstein and Cohen–Macaulay properties play an important
role.

In §3, we construct a derived generalization of the formalism. We define a category

of multiplets and show how the derived formalism establishes an equivalence with the
category of C • (n)-modules. The main result of this chapter says the following.

Theorem (Theorem 3.4.3). There is an equivalence of dg-categories

A• : Multg  ModgC0• (n) : C • (1.2)

between g-multiplets and g0 -equivariant modules over the Chevalley–Eilenberg algebra

C • (n). Here A• denotes the pure spinor functor, and the inverse functor C • = C • (n, −)
is the functor taking derived invariants with respect to the supertranslation algebra n.
Introduction 5

This equivalence of categories is closely related to Koszul duality. We demonstrate the

formalism in several examples and use it to answer some questions on the ordinary
(underived) pure spinor superfield formalism.

Supermultiplets in six-dimensions: a case study. The projective nilpotence va-

riety for six-dimensional N = (1, 0) supersymmetry is isomorphic to P1 × P3 . As the
geometry of this space is very well understood, we use this example to provide an ex-
tensive case study for the link between algebraic geometry and multiplets provided by
the pure spinor superfield formalism. Line bundles on P1 × P3 are easy to understand.
Describing the associated multiplets, we provide the following classification result.

Theorem (Theorem 4.3.1). The multiplets described in §4.3.2 and §4.3.3 classify, up
to quasi-isomorphism, all mutiplets for six-dimensional N = (1, 0) superymmetry whose
derived invariants are graded global section modules of line bundles over the projective
nilpotence variety.

These multiplets also have a natural interpretation in terms of their twists: We will see
that one can think of these multiplets as being those whose holomorphic twists have rank
one over Dolbeault forms on spacetime.

In addition, we explicitly construct multiplets associated to natural higher-rank equiv-

ariant vector bundles, including the tangent and normal bundles as well as their duals.
Among the multiplets constructed are the vector multiplet and hypermultiplet, the fam-
ily of O(n)-multiplets, and the supergravity and gravitino multiplets. Along the way,
we tackle various theoretical problems within the pure spinor superfield formalism. In
particular, we give some general discussion about the relation of the projective nilpotence
variety to multiplets and prove general results on short exact sequences and dualities of
sheaves in the context of the pure spinor superfield formalism.

Eleven-dimensional supergravity and its twists. The chapters §6 and §7 deal

with (twisted) eleven-dimensional supergravity from a target space perspective.

First, we compute the maximal twist of eleven-dimensional supergravity directly in the

component field formulation and thereby verify a conjecture by Costello in the free limit.
To this end, we briefly review how the eleven-dimensional supergravity theory arises from
the structure sheaf on the nilpotence variety in the pure spinor superfield formalism and
describe the L∞ action of the super Poincaré algebra on the component fields. Starting
from these supersymmetry transformations we explicitly compute the maximal twist of
the free theory and show that the it matches with Poisson–Chern–Simons theory in the
free limit.
6 Introduction

In §7, we construct a generalization of Poisson–Chern–Simons theory, defined on any

supermanifold equipped with an appropriate filtration of the tangent bundle. This con-
struction recovers interacting eleven-dimensional supergravity in Cederwall’s pure spinor
formulation, as well as all possible twists of the theory, and does so in a uniform and
geometric fashion. Under the assumption of compatibility between the pure spinor super-
field formalism and twisting (for which we provide evidence) this proves that Costello’s
description of the maximal twist is the twist of eleven-dimensional supergravity in its
pure spinor description. It also provides a pure spinor lift of the interactions in the
minimally twisted theory (for which a conjectural form was given in [RSW23]).

The following picture gives an overview of the different twist calculations of eleven-
dimensional supergravity (both at pure spinor cochain and component field level) which
are carried out in the literature and in this thesis.

Pure spinor Minimal Maximal

[SW21]
supergravity pure spinor twist pure spinor twist
(free limit)
(Cederwall; §7.5.4) (§7.5.3) (§7.5.2)

Homotopy Homotopy
∼
=
Transfer Transfer

Supergravity Minimal twist Poisson–Chern–Simons

[RSW23]
on R11 on C5 × R theory on R7 × C2
(Cremmer–Julia–Scherk) (Rhagavendran–Saberi–Williams) (Costello)

§6 (free limit)

Figure 1.1: Twisting eleven-dimensional supergravity using pure spinor superfields.

In §7, we construct the three upper nodes of the diagram showing that they are ho-
motopy Poisson–Chern–Simons theories. Conjecture 7.4.3 implies that the upper three
nodes are twists of each other. The maximal twist in component fields is computed in
the free limit in §6.
.

(Twisted) (2,0) supersymmetry and holography. M-theory allows for M2 and M5

branes. The worldvolume theory on a (stack of) M5 brane is a six-dimensional super-
conformal field theory with (2,0) supersymmetry and in the framework of AdS7 /CF T6 -
correspondence one expects an equivalence to eleven-dimensional supergravity in back-
reacted geometry. Specifically, Kaluza-Klein modes of the supergravity fields are dual
to CFT operators in the worldvolume theory. In §8, we explore a schematic picture of
this correspondence by starting from well-established results in the maximal twist to-
gether with the structural insights that the pure spinor superfield formalism provides on
twisting.
Introduction 7

We provide a novel pure spinor construction for the decomposition of minimally twisted
supergravity into E(3|6)-modules both at component field level and at pure spinor
cochain level. The construction hints towards a uniform picture appearing in the maxi-
mally, minimally and untwisted cases which we sketch out schematically. In the process,
we find a systematic way to construct conformal supergravity theories.
Chapter 2

The pure spinor superfield

formalism

2.1 Introduction

Speaking broadly, a classical field theory concerns itself with the study of the sheaf of
solutions to particular partial differential equations on the spacetime manifold, or more
properly on the site of manifolds equipped with appropriate structure. Over an open set
U , one considers solutions to the equations of motion of the theory on U , considered up to
gauge equivalence; since the equations of motion that are of physical interest tend to arise
from variational principles, we will refer to it with the suggestive notation Crit(S)/G,
where S refers to the action functional and G to the group of local gauge transformations.

In general, this sheaf has several properties: First and foremost, its sections over U
can be thought of as a covariant version of the phase space associated to ∂U [Crn88],
and thus have the structure of a symplectic space. (We are passing over numerous
technical subtleties in silence; in particular, degeneracies of various kinds can and do
occur, notably in the theory of constrained systems. Such examples arise naturally in
our context [CNT02; SW23b], though we do not treat degeneracies in any detail here.) As
already indicated above, it may not consist just of the space of solutions to the equations
of motion, but of its quotient by gauge equivalences. Lastly, since the degrees of freedom
of many quantum field theories include fermions, it should most properly be understood
as a (possibly singular, stacky, or infinite-dimensional) supermanifold or graded space.

In studying field theories, symmetries play a crucial role. Let g be a sheaf of Lie algebras.
A classical theory has a symmetry by g when its sheaf of fields is equipped with a local
action of the sheaf g by infinitesimal automorphisms. (We will make this more precise

9
10 The pure spinor superfield formalism

in §2.2 below.) Usually, g is either a constant or a locally free sheaf (though other
examples are possible, notably in holomorphic field theories). In the former case, one
refers to a “global” symmetry, and in the latter to a “local” symmetry. By Noether’s
second theorem, local symmetries correspond to degeneracies in the variational problem
of precisely the kind we ruled out above; as such, local symmetries are usually only
relevant when gauged, and the terms “local symmetry” and “gauge symmetry” are often
used interchangeably.1 Examples of symmetries abound; for example, any field theory
on affine space should admit the Lie algebra of infinitesimal affine transformations (the
Poincaré algebra) as a symmetry, reflecting the coordinate invariance (homogeneity and
isotropy) of its dynamics.

Since fermions are typically present in the theory, Vect(Crit(S)/G) is most naturally not
a Lie algebra, but a graded or super Lie algebra. The most important examples of super
Lie algebras extend the Poincaré symmetry by odd spacetime symmetries transforming
in the spin representation of the Lorentz group; a field theory that admits an action of
such an algebra is called supersymmetric. The problem of constructing supersymmetric
field theories has a long history in physics, dating back to the first explorations of the
subject in the seventies [GL71; GS71; VA72].

It is common wisdom in physics that representations of supersymmetry algebras in typical

field theory models can be quite intricate. Often, the supersymmetry algebra closes only
on-shell or up to gauge transformations. In other words, while a symmetry of the theory
in the above sense can be defined, it does not arise in a straightforward manner from an
action on the larger space of fields inside of which the equations of motion are solved.
This leads, among other issues, to difficulties in quantizing the theory.

In typical field theory models the structure of supersymmetry transformations roughly

falls into four distinct cases:

— There is a set of fields on which the supersymmetry algebra is represented on the

nose. This is the case, for example, for the four-dimensional N = 1 chiral multiplet.

— The supersymmetry algebra is only represented after taking the quotient by the
action of the gauge group. This happens, for example, for the four-dimensional
vector multiplet.

— The supersymmetry algebra is represented only after imposing the equations of

motion. Here, the six-dimensional hypermultiplet is an example.
1
For a field theory whose physical fields are of the form Crit(S)/G as above, the Lie algebra of G is
a locally free sheaf of Lie algebras that acts on the degenerate variational problem Crit(S).
The pure spinor superfield formalism 11

— The supersymmetry algebra is represented only after taking the quotient by gauge
transformation and imposing the equations of motion. This most general case
appears in ten-dimensional super Yang–Mills theory, among other examples.

The first objective of this note is to formalize these considerations using the language
of homotopical algebra; we work in the context of the BRST and BV formalisms, which
seek to respectively replace the quotient by gauge symmetries and the imposition of
equations of motion by appropriate derived analogues. In §2.2, we set up some necessary
preliminaries for this context; in particular, we give a definition of a multiplet that is
designed to capture all these different aspects of symmetry in our context.

Once this terminology is established, we turn our attention towards the construction
of supermultiplets via the pure spinor superfield formalism; see [Ber01], and especially
the review [Ced14] and references therein. Our perspective is somewhat nontraditional.
In §2.3 we set up the formalism in a generalized setting (without restricting to supersym-
metry algebras of physical interest), clarify its relation to various standard constructions
in homological algebra, and give an explicit account of calculational techniques from
commutative algebra.

In our interpretation, which builds on that in [ESW21], the pure spinor superfield formal-
ism constructs a supermultiplet out of the datum of an equivariant module over the ring of
functions on the nilpotence variety Y of the relevant superalgebra. Speaking roughly, the
output of the formalism is a rather large cochain complex that is automatically equipped
with a strict action of the supersymmetry algebra—indeed, which is quasi-isomorphic to
a standard component-field description of the multiplet in the BRST or BV formalism,
but which is free over superspace rather than just over the spacetime manifold. We can
then recover the usual component-field description by moving from this large resolution
to a smaller, quasi-isomorphic cochain complex of vector bundles over spacetime, which
is in a certain sense the “minimal” resolution of this kind. A particular filtration on
the pure spinor cochain complex produces the component-field formulation in canonical
fashion; the set of component fields is identified with the vector bundle associated to the
representation of Lorentz and R-symmetry on the Koszul homology of the input module.

One can then transfer the various structures present on the large complex to the com-
ponent fields, using the homotopy transfer theorem. As we will see, this procedure links
the component field description of the multiplet closely to the minimal free resolution of
the equivariant module over the ambient polynomial ring. In particular, we find that all
supersymmetry transformations without spacetime derivatives can be read off directly
from the resolution differential. (This was conjectured by Berkovits in [Ber02].)
12 The pure spinor superfield formalism

Given our presentation of the pure spinor superfield formalism, it is natural to ask ques-
tions how algebraic properties of OY -modules are related to physical properties of the
resulting multiplet. In §2.4, we point out that the Gorenstein property ensures the ex-
istence of a pairing on the multiplet; this pairing, however, can admit various different
physical interpretations. We furthermore study dualizing modules and explain how the
Cohen–Macaulay property is related to antifield multiplets.

Throughout the text we illustrate the procedure with examples in different dimensions
and with various amounts of supersymmetry. In particular, we provide a detailed discus-
sion of ten-dimensional super Yang–Mills theory showing how all the different structures
present in the component field formulation arise via homotopy transfer.

2.2 Preliminaries

2.2.1 Gradings and basic definitions

Many objects appearing throughout this work (be they vector spaces, vector bundles,
associative algebras, or Lie algebras) will carry a grading by Z × Z/2Z as well as a
differential of bidegree (1, +). We will use the abbreviation “dgs,” for “differential graded
super,” to refer to objects of this sort, at least for emphasis.

Definition 2.2.1. A dgs vector space is a Z × Z/2Z-graded vector space E • , equipped

with a square-zero differential d of bidegree (1, +). Equivalently, E • is a cochain complex
in the category of super vector spaces. We can thus write
M 
k k
E= E+ ⊕ ΠE− [−k]. (2.1)
k∈Z

The total parity |v| ∈ Z/2Z of a homogeneous element v ∈ E n is defined by


n mod 2, v ∈ E+n;
|v| = (2.2)
n + 1 mod 2, v ∈ E n .
−

We remark that each of these gradings has a clear physical meaning: the integer grading
corresponds to the ghost number or cohomological degree, whereas the Z/2Z grading
corresponds to the intrinsic parity (fermion number modulo two). The total parity
denotes the Z/2Z-grading which arises by forgetting the Z-grading to Z/2Z and then
totalizing with the intrinsic parity. It is the total parity which governs all signs.

It will often be convenient to introduce an additional piece of data: an auxiliary action

of a one-dimensional abelian Lie algebra R on a dgs vector space. Most results will make
The pure spinor superfield formalism 13

sense for arbitrary R-equivariant dgs vector spaces, but our most common examples will
have the following behavior.

Definition 2.2.2. A lifted dgs vector space is an R-equivariant dgs vector space where
the R-action has integral weights, and the Z/2Z-grading coincides with the R-weight
modulo two. We will use round brackets to refer to shifts of the weight grading and
reserve square brackets for shifts in cohomological degree.

Let us now introduce some notation that we will use when we discuss lifted dgs vector
spaces (and later lifted dgs vector bundles on a smooth manifold X). Let us write
M
E= E w,k (−w)[−k] (2.3)
(w,k)∈Z2

for a lifted dgs vector space, where the first index indicates the decomposition of E
by R-weight, and the second index indicates cohomological Z-degree on E. In some of
our calculations, we will place the summands in a two-dimensional array, where the two
coordinates are given by w − k and k. For example,
 
· · · E −1,0 E 0,0 E 1,0 E 2,0 · · ·
 
E=· · · E 0,1 E 1,1 E 2,1 E 3,1 · · · . (2.4)
 
· · · E 1,2 E 2,2 E 3,2 E 4,2 · · ·

such that we recover w as the total degree with respect to this sheared bigrading. The
overall parity is then determined by the column. If the differential on our dgs vector
bundle decomposes as a sum of homogeneous differential operators of order k, say D =
P
k≥0 Dk , the summand Dk acts—with respect to the sheared grading—with bidegree
(2k − 1, 1). So the homogeneous summands all increase the vertical degree by 1, but they
may modify the horizontal degree by any odd integer ≥ −1.

Definition 2.2.3. A commutative dgs algebra, or cdgsa, is a dgs vector space A• equipped
with a bilinear multiplication

m 2 : A• ⊗ A• → A• . (2.5)

The multiplication is required to be a cochain map of bidegree (0, +); furthermore, it

should be commutative with respect to the Koszul sign rule determined by the total
parity. That is,
ab = (−1)|a||b| ba. (2.6)

We remark that a cdgsa is a commutative differential graded algebra in the category of

super vector spaces. There is also an obvious notion of a lift of a cdgsa, such that a
lifted cdgsa is a commutative differential graded algebra in the category of graded vector
14 The pure spinor superfield formalism

spaces. Finally, we can extend our definitions to encompass super A∞ algebras: a (lifted)
super A∞ algebra A• is an A∞ algebra in the category of super (or graded) vector spaces.
That is, it is a collection
M
A• = Ak [−k] (2.7)
k∈Z

of super (or graded) vector spaces, equipped with maps mn of arity n and bidegree
(2 − n, +) (or (2 − n, 0)) that satisfy the usual A∞ relations.

Example 2.2.4. Let V • be a dgs vector space. The polynomial algebra Sym(V • ) is the
free dgs-commutative algebra generated by V • . Concretely, it is the quotient

Sym(V • ) = T (V • )/(xy − (−1)|x||y| yx) (2.8)

of the tensor algebra by the ideal generated by all (anti)commutators of homogeneous

elements, where (anti)commutativity is determined by the Koszul sign rule for the total
parity.

Of course, all of the notions we have introduced for associative algebras have parallels for
Lie algebras, which we now quickly introduce. Let x1 , . . . , xn be homogeneous elements of
a dgs vector space V • , and σ ∈ Sn a permutation. Then the Koszul sign (x1 , . . . , xn ; σ)
of the permutation is defined by the relation

x1 · · · xn = (x1 , . . . , xn ; σ)xσ(1) · · · xσ(n) . (2.9)

in the algebra Sym(V • ). Furthermore define χ(σ) = (−1)sgn(σ) (x1 , . . . , xn ; σ).

Definition 2.2.5. Let g be a (lifted) dgs vector space. A (lifted) super L∞ algebra
structure on g is a collection of multilinear maps

µk : g×k → g (2.10)

for k ≥ 1, of bidegree (2 − k, +) (or (2 − k, 0), respectively), such that the following two
conditions hold:

(1) Graded skew symmetry. For all σ ∈ Sk , xi ∈ g one has

µk xσ(1) , . . . , xσ(k) = χ(σ)µk (x1 , . . . , xk ) . (2.11)

(2) Higher Jacobi identities. For all xi ∈ g one has

X X
(−1)i(j−1) χ(σ)µj µi xσ(1) , . . . , xσ(i) , xσ(i+1) , . . . , xσ(k) = 0 .

i+j=k+1 σ∈S(i;k)
(2.12)
The pure spinor superfield formalism 15

Here S(i; k) ⊂ Sk denotes all permutations such that σ(1) ≤ · · · ≤ σ(i) and σ(i + 1) ≤
· · · ≤ σ(k). We remark that a (lifted) super L∞ algebra is just an L∞ algebra in super
(respectively, in graded) vector spaces. We further remark that the datum of a (lifted)
super L∞ algebra structure is equivalent to a square-zero derivation of bidegree (1, +)
(or (1, 0) in the lifted case) on the free dgs commutative algebra Sym(g∨ [−1]). This
derivation dg defines the complex computing Lie algebra cohomology,

C • (g) := Sym(g∨ [−1]) , dg .

(2.13)

The shift is with respect to the cohomological degree.

There are some special cases of this definition that we point out. When g is supported
purely in even parity, we recover the ordinary notion of an L∞ algebra [HS93; LM95].
On the other hand, when g is supported in cohomological degree zero, we recover the
notion of a super Lie algebra (or, in the lifted case, a graded Lie algebra). When µk = 0
for all k > 2, we obtain the notion of a dg super Lie algebra.

Example 2.2.6. Let V • be a (lifted) dgs vector space. Then End(V • ) is a dg super Lie
algebra; the bracket µ2 is given by the commutator

µ2 (x, y) = [x, y] = xy − (−1)|x||y| yx (2.14)

whereas the differential arises via

dEnd(V • ) = [d, −] . (2.15)

We remark that End(V • ) is in fact naturally a dgs associative algebra; the dgs Lie
structure is obtained by applying the usual forgetful functor.

Definition 2.2.7. A L∞ map between super L∞ algebras

Φ: g h

is a map of graded super commutative algebras

Φ∗ : C • (h) → C • (g). (2.16)

that preserves the augmentation map to constants in degree zero.

Definition 2.2.8. Let g be a super L∞ algebra. An L∞ dgs module is a dgs vector

space V • , together with an L∞ map

g End(V • ). (2.17)
16 The pure spinor superfield formalism

2.2.2 Homotopy transfer

We will repeatedly make use of the homotopy transfer theorem in various contexts. We
refrain from giving a general review of homotopy algebraic structures here; the reader is
referred to [LV12a; Val14; LM95]. Nonetheless, we will quickly recall the general idea.

It is common knowledge that various mathematical objects—for example sheaves or

modules—admit interesting “higher structures.” This might include higher sheaf coho-
mology groups, for example, or more generally other derived functors such as Ext and
Tor. These higher structures originate, in some sense, from the “constraints” imposed on
these objects: for example, the failure of a module to be free.

To compute higher derived functors, one technique is to replace the object one wants
to study by a “resolution.” This is a cochain complex of simpler objects (for example,
free modules) that is quasi-isomorphic to the complicated object one wants to study. In
derived geometry, one views this cochain complex as a replacement of the underlying
object.

Just as the equations defining a non-free module lead to higher structures and need to
be resolved, many algebraic structures are defined by collections of structure morphisms
that satisfy certain strict equations. (For example, one requires associativity in the form
((ab)c) = (a(bc)), or the Jacobi identity for a Lie bracket.) When such equations are
imposed in a cochain complex, they do not play well with homotopy-theoretic operations
or notions of equivalence such as quasi-isomorphism. The remedy consists of “resolving”
the equations that are imposed on the defining maps of the algebraic structure. In tech-
nical language, one resolves the operad defining the algebraic structure one is interested
in by a free dg operad. (See [Mar98] for discussion of this perspective.)

There is then a collection of general results, which state that a homotopy algebraic
structure may be transferred along homotopy data between two quasi-isomorphic cochain
complexes. For our purposes, this data is typically provided by a deformation retract
(though more general situations are possible [LV12a]), i.e. a diagram

p
h (C • , dC ) (D• , dD ) (2.18)
i

with the maps satisfying

p ◦ i = idC • and i ◦ p − idD• = dC ◦ h + h ◦ dC . (2.19)

In many examples, the complex on the right hand side will simply be the cohomology of
the complex on the left. The transfer of the homotopy algebraic structure can then be
The pure spinor superfield formalism 17

obtained in a systematic way by by summing over marked trees in a consistent fashion.

Here, vertices are to be labeled with operations of the structure to be transferred, and
internal edges with the homotopy. We remark, by [LV12a, Theorem 10.3.15], that the
transferred structure is independent of the choice of homotopy data up to isomorphism.

The phenomenon of homotopy transfer is very broad, and encompasses many examples
from throughout mathematics, both more and less familiar. We mention some examples:

— A cochain complex is defined by a grading, together with a single endomorphism

D of degree +1, satisfying the equation D2 = 0. A cochain complex in cochain
complexes is a bicomplex: we give a second grading on (C • , d), together with a
square-zero cochain map D. Resolving the equation D2 = 0 gives rise to an operad
known as the D∞ operad: it encodes a sequence of maps Di of bidegree (1 − i, i),
which obey the relations
X
dDn + (−1)n Dn d = (−1)i Di Dj . (2.20)
i+j=n

Homotopy transfer of D to H • (C, d) generates a D∞ module structure whose con-

stituent maps encode the higher differentials of the spectral sequence of the bicom-
plex. This will play a role for us in describing the relation of pure spinor superfields
to their component-field descriptions; see §2.3.

— The operad governing associative algebras is resolved by the A∞ operad, which has
operations {mn } of arity n and degree 2 − n for all n ≥ 1. Similarly, the operad
governing Lie algebras is resolved by the L∞ operad, which has bracket operations
µn of arity n and degree 2 − n for all n ≥ 1 as we reviewed explicitly above.
For example, transferring the associative algebra structure on de Rham forms to
cohomology produces an A∞ structure with vanishing m1 and m2 the ordinary cup
product. Higher mn ’s correspond to the classical Massey product operations.

— In the BV formalism, a perturbative classical field theory is described by a cyclic

local L∞ algebra whose differential encodes the linearized equations of motion and
gauge invariances of the free theory. Homotopy transfer to the cohomology of the
differential is related to the interaction picture in quantum field theory; the dia-
grams that describe the transferred L∞ structure on on-shell states are precisely
tree-level Feynman diagrams, where the homotopy is the Feynman propagator.
The operations of the transferred L∞ structure correspond to tree-level ampli-
tudes [Kaj07; MSW19]. Homotopy transfer of L∞ structures will be relevant for
us when discussing interactions for pure spinor superfields and their relation to the
component-field formalism; see §2.5 for an example.
18 The pure spinor superfield formalism

— In the BV or BRST formalism, the symmetries of a field theory are encoded as L∞

module structures on the complex of fields. Moving to another quasi isomorphic
complex of fields (e.g by integrating out an auxiliary field), one can obtain the new
module structure via homotopy transfer. We will use this to derive the action of
the supersymmetry algebra on the component fields in the pure spinor superfield
formalism. An explicit account on the homotopy transfer for module structures is
given in Appendix 2.A.

2.2.3 Maurer–Cartan elements and nilpotence varieties

We recall that the Maurer–Cartan equation in an L∞ algebra (g, µk ) takes the form
X 1
µk (x, . . . , x) = 0. (2.21)
k!
k≥1

Here x ∈ g is an element of degree one; each of the terms in the above equation thus
carries degree two. We can clearly generalize this definition to super L∞ algebras by
asking for Maurer–Cartan elements x of bidegree (1, +). Maurer–Cartan elements of
this form define deformations of the super L∞ algebra structure; nontrivial deformations
are classified by Maurer–Cartan elements up to gauge equivalence. Nonetheless, we will
write MC(g) for the naive space of Maurer–Cartan elements; in other words, we do not
pass to the space of gauge equivalence classes, preferring to think of MC(g) as a space
equipped with a g0 -action by vector fields.

Now, given any super L∞ algebra, we can forget the Z × Z/2Z-grading down to a Z/2Z-
grading by remembering only the total parity. This is enough information to define the
appropriate Koszul signs, and µk is then simply a multilinear operation with appropriate
symmetry properties and parity (−1)k . We will call the resulting object a Z/2Z-graded
L∞ algebra. We can then ask about the space Yg of odd elements satisfying the Maurer–
Cartan equation (2.21). Elements of this space will correspond to deformations of g as
a Z/2Z-graded L∞ algebra and there will be an injective map MC(g) ,→ Yg . We call Yg
the nilpotence variety of g; when g is a super Lie algebra, this agrees with the notion
given in [ESW21]. (Whenever there is no danger of confusion we drop the subscript g.)

For our purposes, Y is an affine scheme; we take

Y = Spec R/I (2.22)

where R = Sym(g∨
− ) is a polynomial ring in commuting variables, and I is the ideal
generated by the Maurer–Cartan equations (2.21). In this work we will be mostly con-
cerned with the case where g is a super Lie algebra, thus I will be generated by quadratic
The pure spinor superfield formalism 19

equations and R/I is a graded ring. Since we view Y as an affine scheme, we will move
back and forth freely between discussing the geometry of Y and the graded ring R/I;
hopefully, no confusion should arise. Sometimes we may also consider the geometry of
the projective scheme Proj R/I (in particular in §4); in either case, the essential object
is the graded ring R/I. The distinction between a variety and a scheme will, in fact,
play a role in applications; see §2.3.9.

2.2.4 Multiplets and local modules

In this section, we move towards the setting of field theory by introducing the new
ingredient of locality.

Local modules. Let X be a manifold thought of as spacetime. By a dgs vector bundle

on X, we mean a Z × Z/2-graded vector bundle
M 
k k
E= E+ ⊕ ΠE− [−k] (2.23)
k∈Z

k → E k+1 such that D ◦ D = 0.2

equipped with a collection of differential operators D : E± ±
k = Γ(X, E k ) denotes the C ∞ -sections of E k .
Here, E± ± ±

Suppose that g is a super L∞ algebra. We will define a local g-module to be a dgs vector
bundle on X equipped with a sufficiently local homotopy action of g.

To give the precise definition we first need a small bit of background. Consider the
Z × Z/2-graded vector space E = Γ(X, E). As explained in Example 2.2.6, the endo-
morphisms End(E) naturally form a dg super Lie algebra: the structure maps consist
of the commutator and the differential [D, −]. Inside (End(E), [D, −]) there is a sub dg
super Lie algebra consisting of all endomorphisms which are differential operators. We
will denote it by (D(E), [D, −]).

Definition 2.2.9. A local (super L∞ ) g-module is a dgs vector bundle (E, D) equipped
with a super L∞ -map (see Definition 2.2.7):

ρ: g D(E) , [D, −] . (2.24)

We will refer to the data of a local g-module by a triple (E, D, ρ).

2
We assume that each graded piece E k is a finite rank vector bundle, but the total rank of E may
still be infinite.
20 The pure spinor superfield formalism

The space of sections of any dgs vector bundle is a dgs vector space. The space of
sections (over any open set) of a local g-module (E, D) is a dgs L∞ module for the super
Lie algebra g, see Definition 2.2.8.

Concretely, the data of ρ consists of a collection of maps

ρ(j) : g⊗j −→ D(E), j≥1 (2.25)

of cohomological degree 1 − j. These satisfy compatibility relations, the lowest of which

reads
[ρ(1) (x), ρ(1) (y)] − ρ(1) ([x, y]) = [D, ρ(2) (x, y)] . (2.26)

Note that if the left hand side were zero, then we would have a strict Lie algebra action.
Thus, ρ(2) provides a homotopy correcting the failure of ρ(1) to be strict.

One way to unravel this definition is in terms of the cochain complex computing the Lie
algebra cohomology of g. The map ρ is equivalent to an element
X
ρ= ρ(k) ∈ C • (g) ⊗ D(E), ρ(k) ∈ C k (g) ⊗ D(E) (2.27)
k

of bidegree (1, +) which satisfies the Maurer–Cartan equation

1
dg ρ + [ρ, ρ] = 0 . (2.28)
2

Here dg denotes the Chevalley–Eilenberg differential of g and [−, −] is the commutator

of differential operators.

We observe that ρ determines a super L∞ structure on g ⊕ E in such a way that there is

a short exact sequence of L∞ algebras

0→E →g⊕E →g→0 (2.29)

where E is thought of as an L∞ algebra with µk = 0 for k > 1.

Let us take some time to reflect on this definition from the physics point of view. It is
well known that the supersymmetry algebra is sometimes only realized on-shell or up
to gauge transformations. This is precisely captured in the fact that we used a super
L∞ -map g


D(E) , [D, −] to define a multiplet instead of a super Lie map. The
higher order terms ρ(j) for j ≥ 2 precisely correspond to closure terms correcting ρ(1) by
a gauge transformation or contributions proportional to an equation of motion.

This discussion explains how the supersymmetry algebra acts on the fields of the theory.
The operators of the theory consist of functionals of the fields and are denoted by O(E).
The pure spinor superfield formalism 21

For any point x ∈ X, we can define the local operators via

Ox (E) = Sym• (J ∞ E|x )∨ , (2.30)

where J ∞ E denotes the jet bundle of E. In other words, the local operators at x evaluate
polynomials in the fields and derivatives of fields at x. Given a map

ρ: g D(E) , [D, −] , (2.31)

the dual maps (ρ(j) )∨ define an action on the linear local operators, which extends to
O(E)x via the Leibniz rule. Fixing an element Q ∈ g we can define a map
X
δQ = ρ(j) (Q, . . . , Q)∨ : Ox (E) −→ Ox (E) , (2.32)
j

which defines the action of Q ∈ g on the operators of the theory.

Local algebras. For completeness, let us briefly remark that there is a natural way
to make the symmetry algebra g local as well. This is relevant if g encodes a gauge
symmetry.

Definition 2.2.10. A local super L∞ algebra on a manifold X is a dgs vector bundle

L → X, equipped with a collection of polydifferential operators

µk : (L)×k → L (2.33)

of bidegree (2 − k, +) that satisfy the relations of a super L∞ algebra structure. Here

L = Γ(X, L) are the smooth sections of L.

The definition of a local module structure generalizes in obvious fashion. We note that,
given a super L∞ algebra g, the constant sheaf g is not an example of a local super L∞
algebra for d > 0, since it is not given as the smooth sections of any dgs vector bundle.
However, we can remedy this by resolving the constant sheaf by the de Rham complex:
Ω• (X) ⊗ g is a local super L∞ algebra on X. (This example is relevant to Chern–Simons
theory.)

Furthermore the above definition is important in the general context of the BV formalism:
A perturbative classical field theory in the BV formalism will be equivalent to a local
super L∞ algebra on X, equipped with a trace map of degree −3. We will further review
this perspective in what follows.
22 The pure spinor superfield formalism

Multiplets. In the context of supersymmetry, we are interested in local modules that

satisfy an additional compatibility condition. For now, let X = VR = Rd be a d-
dimensional affine space and let V = Cd be its complexification.3 The Poincaré group is
the group of affine transformations of this space; it is of the form

Aff(V ) = Spin(V ) n V . (2.34)

The complexified Lie algebra aff(V ) is

spin(V ) n V ∼
= ∧2 V n V. (2.35)

A multiplet is a local module structure for a dgs L∞ algebra on an affine 4 dgs vector
bundle on Rd , where the g-action is required to be compatible with the action of the
affine algebra in a certain sense. We make this precise with the following definition.

Definition 2.2.11. Let E be an affine dgs vector bundle on X = VR , and g a super L∞

algebra equipped with a map
φ : aff(V ) → g . (2.36)

A g-multiplet is a local g-module structure on E, such that the pullback of the module
structure along φ agrees with the natural action on sections of the affine vector bundle.
Concretely, this means that the following diagram commutes.

ρ(1)
g D(E)
φ (2.37)

aff(V )

We think of a multiplet as a derived replacement for the (not necessarily locally free)
“supersymmetric sheaf” H • (E). Even though this sheaf could be regarded as the central
object of study in physics, it is more natural from either the BRST/BV perspective or
the perspective of derived geometry to just work at the cochain level. There is again
a generalization of this definition to local super L∞ algebras, where the global affine
algebra is replaced by a local L∞ algebra modeling local isometries. We do not pursue
this further here.

We briefly note that this definition implies that the image of φ is represented strictly
on the fields. Furthermore, since the natural action of the affine algebra is effective, the
3
For the pure spinor superfield formalism it will be useful for us to use complex Lie algebras.
4
A dgs vector bundle is called affine if the total space carries an action of the affine group such that
the projection is equivariant with respect to the action of the affine group on Rd .
The pure spinor superfield formalism 23

above definition requires implicitly that φ be injective. So multiplets naturally lead us

to study superalgebras that contain the affine algebra as a subalgebra.

We take note of the following examples:

— Let h be a Lie algebra, and consider the product g = h ⊕ aff(V ), equipped with the
obvious choice of φ. Then a g-multiplet contains a collection of fields transforming
in a local representation of h. Flavor symmetry multiplets are an example of this
kind.

— Let conf(V ) be the Lie algebra of conformal vector fields on V . There is a canonical
embedding of aff(V ) in conf(V ). Then a conf(V )-multiplet encodes the notion of a
conformally invariant multiplet of fields.

— Let g be the super Poincaré algebra. It contains aff(V ) as a subalgebra, and a

g-multiplet recovers the usual notion of a supermultiplet.

Historically speaking, the construction of interesting multiplets for algebras that were
not products was the motivation that led to the origin of supersymmetry; we return to
this point (and construct examples of the relevant algebras of physical interest) below.

To conclude this paragraph, we discuss a first example of a multiplet.

Example 2.2.12. Let us give one example of a non-strict multiplet for three dimensional
N = 1 supersymmetry. Recall that Spin(3) ∼ = SU(2); we denote the two dimensional
spinor representation by S and the three-dimensional vector representation by V . We
fix g to be the super Poincaré algebra, whose underlying Z/2Z-graded vector space is of
the form
g = (so(3) ⊕ V ) ⊕ ΠS (2.38)

where Π indicates the shift into odd parity. The symmetric bracket is induced from the
isomorphism Γ : Sym2 (S) ∼= V of SU(2)-representations. Let us define E to be the trivial
vector bundle over V = R3 with fibers

−1 0 0
E+ =C E+ =V E− =S, (2.39)

where we have used the subscript ± to indicate Z/2Z-degree. The field content consists
of a zero-form, a one-form and a fermion field. The differential D operates on sections as
the de Rham differential d : Ω0 −→ Ω1 , and vanishes elsewhere. This dgs vector bundle
0 has weight one, and E 0 has
can be lifted to a Z × Z-graded vector bundle where E+ −
weight two.
24 The pure spinor superfield formalism

We summarize this field content with the below array, using the conventions of §2.2.1:
 
Ω0
d (2.40)
 
 
Ω1 S

The even part of g acts in the standard geometric fashion. For Q ∈ g− we set

ρ(1) (Q) : S −→ Ω1 , ψ 7→ Γ(Q, ψ)

: Ω1 −→ S , A 7→ /
Q ∧ ∂A (2.41)
ρ(2) (Q, Q) : Ω1 −→ Ω0 , A 7→ ι[Q,Q] A .

Here ι denotes the contraction of a differential form by a vector field and we view [Q, Q]
as a constant vector field on X.

2.2.5 Further structures on multiplets

As we will see in the following sections, the pure spinor superfield formalism naturally
produces multiplets for the supersymmetry algebra. Some extra data is required to
produce a theory out of a multiplet; furthermore, depending on whether or not super-
symmetry closes off-shell, the resulting theory may be a BRST or a BV theory, so that
the additional data required may differ. There are also conditions on the additional data
that ensure that the theory is nondegenerate in an appropriate sense. We set up some
formalism for the required additional structure in this section.

BRST data. In the BRST formalism, a perturbative field theory is described by a

local super L∞ algebra L equipped with a BRST action functional S, which is invariant
for the L∞ structure. The L∞ structure describes the (higher) infinitesimal gauge trans-
formations and the variation of the BRST action gives rise to the equations of motion.

Definition 2.2.13. A BRST datum on the g-multiplet (F, D, ρ) consists of

— a local super L∞ structure {µk } on L := F [−1] such that µ1 = D, and whose

associated Chevalley–Eilenberg differential we denote by QBRST ; and

— a local functional S0 ∈ Oloc (F ) of bidegree (0, +), called the “BRST action func-
tional,” which is closed with respect to QBRST .

This data should be such that all maps in the short exact sequence

0→L→g⊕L→g→0 (2.42)
The pure spinor superfield formalism 25

are L∞ maps, and S0 is invariant for the L∞ action ρ.

For physicist readers, the shift by one appearing in L = F [−1] may deserve some com-
ment. Essentially, this arises from the fact that the ghosts for the theory (which are
valued in the Lie algebra of the gauge group, and thus carry a Lie algebra structure)
sit in cohomological degree −1. The observables of the theory, which are functions on
compactly supported sections of F , are Lie algebra cochains of the local L∞ algebra L;
the BRST differential on observables is the Chevalley–Eilenberg differential arising from
the gauge algebra structure.

BV data. The (classical) Batalin–Vilkovisky (BV) [BV81; BV84; BV85] formalism is a

generalization of the BRST formalism that encodes the equations of motion in a derived
way. For a comprehensive review of the classical BV formalism we refer to [CG17; CG21]
(see also [Jur+19; Mne17]). We recall the general idea briefly.

Perturbatively, a BV theory is described by a local super L∞ algebra LBV equipped

with an invariant, skew-symmetric, non-degenerate, local pairing of degree −3 (see the
definition below). The space of “BV fields” is the space of sections of the bundle E =
LBV [1] given by the shift in Z-degree of the L∞ algebra encoding the BV theory.

The degree-(−3) local pairing on LBV is equivalent to a local skew pairing of degree −1
on the space of BV fields E = LBV [1], which in turn can be thought of as a (−1)-shifted
symplectic structure on the BV fields. As above, the shift is needed so that observables of
the classical BV theory can be identified with the Lie algebra cochains of LBV . The (−1)-
shifted symplectic structure equips the observables with a degree-(+1) Poisson bracket,
often called the antibracket. In turn, the degree-(+1) Chevalley–Eilenberg differential
on observables that encodes the super L∞ algebra structure on LBV is a degree-(+1)
Hamiltonian vector field, that can be encoded in the datum of a BV action functional
SBV , so that
C • (LBV ) = O(E) , {SBV , −} .


(2.43)

The zeroth cohomology of this cochain complex is the space of functions on the critical
locus of the BRST action modulo gauge equivalence. The condition that {SBV , −} define
a differential is equivalent to the classical master equation

{SBV , SBV } = 0 . (2.44)

Of course, proper care must be taken to make rigorous sense of the BV complex above.
As we are working perturbatively, the space of BV fields will arise as the space of sections
26 The pure spinor superfield formalism

of some graded vector bundle on spacetime. Furthermore, the BV action will be given
as the integral of a Lagrangian density of the fields. More details on the BV formalism
can be found in [CG17; CG21].

For any multiplet, we will define a notion of “BV datum,” which consists of the set of data
necessary to construct a BV theory (a (−1)-shifted invariant symplectic pairing, with
respect to which the homotopy g-action is defined by Hamiltonian vector fields, and a BV
action functional that is compatible with the action of g). A BV theory will then consist
of a multiplet equipped with a BV datum that satisfies an additional nondegeneracy
condition.

Definition 2.2.14. A BV datum on a g-multiplet (E, D, ρ) consists of:

— a graded antisymmetric map

h−, −iloc : E ⊗ E −→ DensX (2.45)

of bidegree (−1, +), which is fiberwise non-degenerate; and

— a C • (g)-valued BV action

(k) (k)
X
SBV,g = SBV,g ∈ C • (g) ⊗ Oloc (E), SBV,g ∈ C k (g) ⊗ Oloc (E) (2.46)
k

of bidegree (0, +) of the form

Z
(0)
SBV,g (Φ) = hΦ, DΦiloc + IBV (Φ) (2.47)
X

where IBV (Φ) is a Lagrangian that is at least cubic in the fields and where
Z
(k)
SBV,g (x1 , . . . , xk ; Φ) = hΦ, ρ(k) (x1 , . . . , xk )Φiloc (2.48)
X

such that

(i) h−, −iloc is invariant for the L∞ action ρ;

(ii) the total action SBV,g satisfies the g-equivariant master equation

1
dg SBV,g + {SBV,g , SBV,g } = 0. (2.49)
2

(0)
If D is elliptic, then SBV,g (Φ) is a g-equivariant perturbative BV theory in the sense
of [CG21]. According to the terminology in loc. cit., this total action SBV,g endows
The pure spinor superfield formalism 27

(0)
SBV,g (Φ) with the structure of a g-equivariant theory. We will refer to a multiplet
equipped with a BV datum for which D is elliptic as a g-equivariant BV theory.

To go from a multiplet with BRST datum to a multiplet with BV datum, one considers

LBV = L ⊕ L∨ [−3] (2.50)

which is equipped with a canonical evaluation pairing of degree (−3). The BRST action
deforms the obvious L∞ structure on the direct sum of L with L∨ [−3], thus giving rise to
an L∞ structure on LBV for which the evaluation pairing is invariant (after an application
of the homological perturbation lemma, which can be thought of as solving the classical
master equation for SBV order by order).

We will say that a multiplet equipped with a BRST datum is a BRST theory when the
corresponding BV datum itself defines a BV theory, meaning that the kinetic term in
the BV action involves an elliptic operator.

Note that, in the way we have set things up, any multiplet can be equipped with a trivial
BRST datum, whereas a BV datum may not always exist. In §2.4 we will see that some
of the multiplets produced in the pure spinor formalism can be naturally equipped with
nondegenerate BV data, while this is not possible for others. Of course, a degenerate
BRST datum does not, in itself, define a BRST theory.

For a multiplet with BV datum (E, D, ρ, h., .i), the inner product always allows us to
write
E = F ⊕ F ∨ [−1] , (2.51)

k . This induces a splitting on the space of sections

where F = ⊕k≤0 EBV

E = F ⊕ F ! [−1] . (2.52)

Note that this is a splitting on the level of super vector spaces.

Definition 2.2.15. A BV multiplet (E, D, ρ) is off-shell if the above splitting exists on

the level of g-modules. Then F is naturally a BRST multiplet, and (F ∨ [−1], D|F ! , ρ|F ! )
is called the antifields multiplet for (F, D|F , ρ|F ).

Intuitively, this definition means that it is possible to consider the g-action separately
on the fields and antifields. Then, the equations of motions are not needed to close the
algebra and the only corrections for the action come from gauge transformations.
28 The pure spinor superfield formalism

2.3 The pure spinor superfield formalism

2.3.1 A universal construction

Let g be a super Lie algebra, and Y its nilpotence variety, viewed as an affine scheme as
discussed above. Let M be any (dgs) g-module, and Γ any graded module for the graded
ring R/I. (We can equivalently view Γ as defining a sheaf on Spec R/I.) Then there is
a map
ρ : g → End(M ) (2.53)

defining the g-module structure, and an obvious map

m : g∨
− → End(Γ) (2.54)

given by left multiplication (after recalling that g∨

− includes into R/I in weight one). If
we consider the tensor product M ⊗ Γ, the above two maps define a map

ρ · m : g− ⊗ g∨
− → End(M ⊗ Γ) (2.55)

as explained in the following diagram.

ρ·m
g− ⊗ g∨
− End(M ⊗ Γ)
ρ⊗m
· (2.56)
End(M ) ⊗ End(Γ)

End(M ⊗ Γ) ⊗ End(M ⊗ Γ)

That is, we apply ρ ⊗ m, include the resulting element into End(M ⊗ Γ) ⊗ End(M ⊗ Γ)
and finally multiply to obtain an endomorphism of M ⊗ Γ.

The map ρ · m equips M ⊗ Γ with a canonical square-zero differential D, defined to be

the image of the canonical element

1 ∈ g− ⊗ g∨ ∼
− = End(g− ). (2.57)

The square of this differential sits in the defining ideal of R/I, and thus is zero for any
R/I-module Γ.

Remark 2.3.1. In the case that Γ = R/I is the ring of functions, we note that this
construction is closely related to the following construction: As in derived geometry,
we define the “classifying space” of a super L∞ algebra g to be the derived scheme Bg
The pure spinor superfield formalism 29

whose ring of functions consists of the Lie algebra cochains C • (g). Then a version of the
associated bundle construction associates a sheaf on Bg to any g-module M ; the global
sections of this sheaf are C • (g; M ). In the cases we are interested in, there is a close
connection between OY and Lie algebra chochains. This is already a first hint towards
the derived formalsim we construct in §3.

2.3.2 The case of interest: from sheaves to multiplets

Let n be a two-step nilpotent super Lie algebra, defined by a central extension

0 → n2 → n → Πn1 → 0 (2.58)

of the odd abelian super Lie algebra Πn1 . We imagine such objects as generalizations of
supertranslation algebras. The automorphisms of n, which are all outer, will contain an
abelian factor generating scale transformations, with respect to which n1 has weight one
and n2 weight two. There is also a natural map

aut(n) → gl(n2 ). (2.59)

The kernel of this map can be thought of as the R-symmetry algebra; in physical exam-
ples, aut(n) will be the product of an orthogonal algebra so(d), scale transformations,
and the R-symmetry algebra.

All of our constructions will take place in reference to a fixed super Lie algebra g of the
following type.

Definition 2.3.2. A super Lie algebra g is of super Poincaré type if it can be written
as an extension
0 → n → g → g0 → 0, (2.60)

where n is a two-step nilpotent super Lie algebra of the form (2.58) and g0 is a Lie algebra
equipped with a Lie map g0 → aut(n).

This means that the Z/2Z grading on g can be lifted to a Z-grading concentrated in
degrees zero, one, and two:

g = g0 ⊕ n1 (−1) ⊕ n2 (−2). (2.61)

In keeping with the above discussion, we regard this as a lifted super L∞ algebra that
is concentrated in cohomological degree zero (and in degrees zero, one and two with
repsect to the lifted weight degree); as such, only the binary bracket operation can be
nonvanishing for degree reasons.
30 The pure spinor superfield formalism

The most important examples will be super Poincaré algebras; we review how these
are constructed below, but just remark here that n2 consists of translations and n1 of
supersymmetries in that case.

As we will see momentarily, n2 will play the role of the spacetime on which the multiplet
is constructed. We note that much of the construction would go through if g were any
nonnegatively graded Lie algebra. In such a case, however, the bosonic part of g>0 may
not be abelian, and an interpretation of the construction in terms of multiplets on an
affine supermanifold will not be immediate. As such, we do not study any examples of
this sort here.

To be very explicit, if we choose a basis dα of n1 and a basis eµ of n2 , we can express the

µ
bracket in terms of structure constants fαβ

µ
[dα , dβ ] = fαβ eµ . (2.62)

We denote by λ1 , . . . , λn coordinates on n1 dual to the basis dα . Then the defining ideal

I of the nilpotence variety is generated by the equations

µ β
I = (λα fαβ λ ), (2.63)

such that the quotient ring

R/I = C[λ1 , . . . , λn ]/I, (2.64)

is its ring of functions.

We are interested in a particular example of the construction above, where M is taken

to be the g-module consisting of smooth functions on N = exp(n) (viewed as a super-
manifold). Concretely,

M = C ∞ (N ) = C ∞ (X) ⊗C ∧• (n∨
1), (2.65)

where we already identified X = n2 . There are two commuting actions of g on M , on

the left and the right; we denote these by

R, L : g −→ End(M ). (2.66)

Now, for any graded R/I-module Γ that is equivariant for the g0 -action, applying the
construction above to M (with respect to the right action of g) produces a cochain
complex
A• (Γ) = (Γ ⊗C M, D = λα R(dα )). (2.67)
The pure spinor superfield formalism 31

Explicitly, let xµ be linear coordinate functions on n2 and θα be (odd) linear coordi-

nate functions on n1 , dual to the basis (eµ , dα ) above. Then the differential is given in
coordinates by  
∂ µ β ∂
D = λ R(dα ) = λ
α α
α
− fαβ θ , (2.68)
∂θ ∂xµ
where the differential operators act in M = C ∞ (N ) and λα acts on Γ via the OY -module
structure. Checking that D squares to zero explicitly is also straightforward:

1
D2 = λα λβ R(dα )R(dβ ) = λα λβ [R(dα ), R(dβ )]
2 (2.69)
1 α β 1 µ
= λ λ R([dα , dβ ]) = λα λβ fαβ R(eµ ) = 0.
2 2

A• (Γ) naturally has the structure of a dgs vector space: we assign bidegree (1, −) to
λα , (0, −) to θα , and (0, +) to xµ . Since g admits a natural lift, there is also a natural
candidate for a lifted dgs vector space structure, in which λα carries bidegree (1, −1),
θα bidegree (0, −1), and xµ bidegree (0, −2). However, this lift only defines a sensible
bigrading on polynomial functions on n2 , rather than on all smooth functions. This
bigrading is often referenced in the pure spinor superfield literature, often under the
names “ghost number” and “dimension.” We will not need it in what follows, and will
view A• (Γ) just as a dgs vector space. However, a filtration related to the dimension will
play an important role for us.

From this discussion, it is clear that A• (Γ) can be viewed as the global sections of an
affine dgs vector bundle E → X over X = n2 with typical fiber

Exk ∼
= ∧• n∨ k
1 ⊗C (Γ) . (2.70)

This is the underlying vector bundle of the multiplet we construct.

We note some properties of this construction below:

— By construction, the left action of n commutes with the differential D. As such, the
left action defines a strict n-module structure, which is equivariant with respect
to Aut(n) and as such can be extended to a strict action of g.

— There is an obvious sense in which (a subgroup of) the even part of g consists of
affine transformations acting on M . The g-action is compatible with this inclusion
map, and thus makes A• (Γ) into a g-multiplet.

— In the definition given above, the notion of a multiplet was designed to capture
the notion of a sheaf over spacetime admitting an action of supersymmetry. In
physical terms, this sheaf could be thought of as either on-shell or off-shell field
32 The pure spinor superfield formalism

configurations up to gauge equivalence. A multiplet, that is a cochain complex of

vector bundles with a homotopy action of supersymmetry, can be thought of as
a resolution of this sheaf. (This corresponds to studying off-shell supersymmetric
theories in the BRST formalism, and on-shell theories in the BV formalism.) The
multiplet A• (Γ) goes one step further: it resolves a supersymmetric sheaf not just
freely over spacetime, but freely over superspace. The action of the supersymmetry
algebra is thus just the obvious one on functions on superspace, which is both strict
and geometric in nature.

— For any super Lie algebra g of super Poincaré type, we can choose as input mod-
ule the ring of functions, Γ = R/I. We will sometimes call the the associated
multiplet A• (R/I), the canonical multiplet of g (this terminology was introduced
in [Ced+23]). It is apparent that A• (R/I) has the structure of a commutative
algebra, and therefore that its cohomology H • (A• (R/I)) is also an A∞ algebra in
a canonical way. A• (R/I) is a strict model of this A∞ structure.

— To sum up, we have constructed a canonical way of associating a multiplet to any

equivariant sheaf on Y . Schematically, we depict the construction as an assignment

Pure spinor formalism

{Graded equivariant R/I-modules} −−−−−−−−−−−−−→ {g-Multiplets} . (2.71)

In §3 we will upgrade this construction to a functor to a suitable category of

multiplets and, eventually, into an equivalence of dg-categories.

In many examples, there is further structure available, and A• (Γ) can be equipped with
a collection of higher brackets endowing it with the structure of an L∞ algebra. By
homotopy transfer this yields an L∞ structure on the cohomology. In physically relevant
examples, such L∞ structures precisely correspond to those appearing in the BV or
BRST description of the underlying field theory.

To give one example, the ten-dimensional super Yang–Mills multiplet is constructed

by considering A• (R/I) for the ten-dimensional N = 1 supersymmetry algebra. Since
A• (R/I) is a commutative dgs algebra, we can tensor with any finite-dimensional Lie
algebra h. Then A• (R/I) ⊗ h is a dgs Lie algebra that freely resolves the L∞ structure
of the BV description of interacting N = 1 super Yang–Mills theory. This description
is well-known from work of Berkovits and Cederwall, but we review it in our language
below in §2.5 and explicitly derive the standard structures using homotopy transfer.

The general construction we have outlined so far produces a “large” multiplet, which,
as outlined above, can be thought of resolving a sheaf over spacetime with an action of
supersymmetry. Of course we can just move to the cohomology of our multiplet to recover
The pure spinor superfield formalism 33

this sheaf; however, one might wonder whether and how a smaller multiplet resolving
the same sheaf can be extracted. For example, one wants to connect the pure spinor
multiplet A• (Γ) to typical component field multiplets, that is to a multiplet with finite
total rank over spacetime. In fact, there is a general technique for producing “minimal”
resolutions of this kind, which was discussed in [MSX12; KL09]. We review it in our
language below and give a proof that highlights the relation to standard constructions
in algebraic geometry and homological algebra. After that, we will construct our first
examples of physically relevant algebras and multiplets.

2.3.3 Filtrations and Koszul homology

The object A• (Γ) that we have constructed admits a natural filtration F • A• (Γ); un-
derstanding the spectral sequence associated to this filtration will allow us to relate the
multiplets we are constructing to finite-rank vector bundles over the spacetime X. The
filtration is associated to a second integer grading; we will find that, while not all of the
structures we are interested in preserve this second grading, they do play nicely with the
associated filtration. The filtration degree is defined by the assignments in the following
table:
homological degree intrinsic parity filtered weight
x 0 + 0
(2.72)
λ 1 − 1
θ 0 − 1

(These conventions for the filtration follow those used in [SW21].)

Since C ∞ (X) plays no role in the filtration, we are exhibiting A• (Γ) as a filtered dgs
vector bundle over X. Moreover, since the filtration plays well with the product structure
on the algebra A• (R/I), it gives rise to the structure of a filtered commutative dgs algebra
there. However, we observe that the tautological differential does not respect the integer
grading by filtration weight. Recall that, in coordinates,

∂ µ β ∂
D = D0 + D1 = λα α
− λα fαβ θ . (2.73)
∂θ ∂xµ

As the notation suggests, the differential is the sum of two terms, which have filtered
weight zero and two respectively. The associated graded complex is thus equipped only
with the differential D0 , which is independent of smooth functions on X. The associated
graded then takes the following form.
 
• ∞ ∂ ∼
Gr A (Γ) = α
C (X) ⊗C (Γ ⊗C C[θ ]) , D0 = λ α
= C ∞ (X) ⊗C K • (Γ). (2.74)
∂θα
34 The pure spinor superfield formalism

Here, we have defined the Koszul homology of any R-module in standard fashion:
 
• α α ∂
K (Γ) := Γ ⊗C C[θ ] , D0 = λ . (2.75)
∂θα

The fact that Γ is an R/I-module is of course vitally important for our construction, but
Koszul homology makes sense for any R-module. In the pure spinor superfield literature,
the cohomology of Gr A• is often referred to as “zero mode cohomology” [Ced14].

If we consider the spectral sequence associated to this filtration, we find that the E1 page
is just given by
H • (Gr A• (Γ)) = C ∞ (X) ⊗C H • (K • (Γ)). (2.76)

Since Γ is a graded module, the Koszul homology of Γ is a finite-dimensional bigraded

representation of the Lorentz group. As such, H • (Gr A• (Γ)) determines a vector bundle
over X = n2 ∼= Rd with fibers

(Ex0 )k ∼
= H • (K • (Γ))(k) . (2.77)

We emphasize that the homological degree of E 0 is determined by the internal (weight)

grading on Γ, whereas the parity is determined by the homological degree in Koszul
homology modulo two. D1 induces a new differential D0 acting on the sections of this
vector bundle via homotopy transfer of D∞ -algebras. In addition, the g-module structure
transfers as well such that (E 0 , D0 , ρ0 ) is again a multiplet. This multiplet precisely
corresponds to the component field description of multiplets as they are known from the
physics literature. The transferred differentials play the role of BRST or BV differentials.

Of course one could go on and consider the full cohomology of A• (Γ). If the transferred
differential D0 on the component field level does not already vanish, then the resulting
object will no longer be free over spacetime, i.e. it does not consist of vector bundles and
thus does not fit our definition of a multiplet. It is, however, still a sheaf on spacetime
which carries a g-action. Physically speaking this sheaf consists of the on-shell, gauge
invariant states of the multiplet.

Let us summarize these relations by the following diagram.

(A• (Γ), D) Free over superspace

HT

(H • (GrA• ), D0 ) Free over spacetime (2.78)

HT

(H • (A• (Γ)), 0) Not necessarily free

The pure spinor superfield formalism 35

The compatibility of the differential with the filtration in fact arises from a compatibility
of the left and right g-actions with the filtration, once g is filtered in an appropriate way.
Using the standard definition of a complete filtered Lie algebra [KN64; Koc77], we can
equip g with a filtered structure by setting

g = g(−1) ⊃ g(0) = g+ . (2.79)

We observe that this filtration corresponds to the one we defined above, viewing the
pure spinor superfield as constructed from functions on superspace together with the
degree-zero Lie algebra cohomology of g>0 (see §2.6.3).

The associated graded super Lie algebra Gr(g) is then the extension of g0 by the abelian
module consisting of n1 ⊕ n2 ; said differently, we set the bracket between odd elements
to zero. It is immediate that there is a Gr(g)-module structure on Gr A• (Γ). We will be
able to derive this module structure, which consists of “all supersymmetry transforma-
tions that are independent of spacetime derivatives,” efficiently in examples, using purely
algebrogeometric information about Γ.

2.3.4 Examples of interest: Supersymmetry algebras

We are mostly interested in multiplets for supersymmetry algebras on an affine spacetime

X = VR . Depending on the dimension, Spin(V ) will have either one or two spinor
representations, which we denote by S or S± respectively; furthermore, there will be
an equivariant map Γ that witnesses V as a submodule of the tensor square of the spin
representation.

We construct the space n1 by taking the tensor product of a spin representation with an
auxiliary vector space U , which (depending on dimension) may or may not be equipped
with either a symmetric or antisymmetric bilinear form. The bracket is constructed from
the pairing Γ; if Γ pairs one spin representation with the other (in dimension 0 mod 4),
we tensor one spin representation with U and the other with U ∨ . If Γ is a symmetric self-
pairing (as in dimensions 1, 2, and 3 mod 8), U should have a symmetric bilinear form;
similarly, if Γ is an antisymmetric self-pairing on a spin representation (as in dimensions
5, 6, and 7 mod 8), U must be a symplectic vector space. The “degree of extended
supersymmetry,” denoted N , is the dimension of U as a multiple of the smallest possible
dimension (two in the symplectic case and one otherwise). In cases where a self-pairing
exists on chiral spin representations (dimension 2 and 6 mod 8), two independent choices
of N are possible, one for each chirality. By abuse of notation, we will also write Γ for
the symmetric pairing on n1 .
36 The pure spinor superfield formalism

The supertranlation algebra n constructed in this way is extended to the super Poincaré
algebra g by adding in the automorphisms of n in degree zero; these consist of Lie(Spin(V )) =
so(V ), together with the automorphisms of U that preserve the pairing if present: either
gl(U ), so(U ), or sp(U ), depending on dimension. In physics, this additional automor-
phism is known as R-symmetry.

The nilpotence varieties of these algebras were studied systematically in [ESW21; ES19b];
most examples were already present in the previous pure spinor literature. It is worth
commenting briefly on the connection to the classical notion of a “pure spinor” given
by Cartan. Recall that the spin representation of Spin(VR ) is constructed by choosing
a maximal isotropic subspace L ⊂ VC . Then S = ∧• (L∨ ), and VC = L ⊕ L∨ acts
via Clifford multiplication just by wedging and contracting. (In odd dimensions, VC =
L⊕L∨ ⊕(L⊥ /L), and the single generator in L⊥ /L acts diagonally by the parity operator.)

Given the construction of the brackets in g, it is clear that an element lying in ∧0 (L∨ )
(tensored with any element of U ) is automatically square-zero, and that it will be a
“minimal” or holomorphic supercharge. Considered as a projective variety, the space
of such elements thus consists of the product of the projective space P (U ) and the
space OGr(n, d) of isotropic subspaces L = Cn ⊂ VC = Cd . (Here n = bd/2c.) The
latter is the space of Cartan pure spinors, the minimal nonzero Spin(d) orbit in the spin
representation. However, we emphasize that the nilpotence variety in general contains
many more strata, and may even include nonminimal orbits in the spin representation,
quite independently of R-symmetry (as in eleven dimensions).

We will not construct all supersymmetry algebras in detail here (for discussion that uses
similar style and notation, see [ESW21]). We will just introduce examples as we need
them, beginning with the four-dimensional N = 1 algebra in the next section.

2.3.5 Motivating example: the 4d chiral multiplet

As an explicit example, let us consider the N = 1 supersymmetry algebra in four dimen-

sions. A related discussion of the chiral multiplet already appeared in [ESW21].

Since the dimension is zero modulo four, U carries no pairing and can be taken to be one-
dimensional. n1 is then S+ ⊕ S− , and the bracket is constructed using the isomorphism

S+ ⊗ S− ∼
=V (2.80)

of Spin(4) representations. Because this is an isomorphism, the self-bracket of an element

Q ∈ n1 is zero precisely when either Q ∈ S+ or Q ∈ S− ; as such, Y consists of two
The pure spinor superfield formalism 37

coordinate planes of the form C2 ⊂ C4 , intersecting at the origin. More precisely,

Y = S+ ∪{0} S− . (2.81)

We repeat the same computation in coordinates for emphasis. A general supercharge Q

can be written in the form
Q = λα Qα + λ̄β̇ Q̄β̇ . (2.82)

Accordingly, the equation [Q, Q] = 0 reduces to the four quadratic equations

λα λ̄β̇ σαµβ̇ = 0 , (2.83)

where we used that the isomorphism S+ ⊗ S− ∼

= V can, in a basis, be described by the
Pauli matrices σ µ . Multiplying matrices gives the four equations

λ1 λ̄1 + λ2 λ̄2 = 0
λ1 λ̄1 − λ2 λ̄2 = 0
(2.84)
λ1 λ̄2 + λ2 λ̄1 = 0
λ1 λ̄2 − λ2 λ̄1 = 0 .

Adding and subtracting these equations one finally finds

λ1 λ̄1 = λ2 λ̄2 = λ1 λ̄2 = λ2 λ̄1 = 0 , (2.85)

which implies that λα or λ̄β̇ vanish and recovers our result from above.

To construct a multiplet, we have to choose an R/I-module. One possible choice is

Γ = C[λ̄α̇ ], which corresponds to the pushforward of the structure sheaf of S+ to Y
along the inclusion map. We form the pure spinor complex:
 
• ∞ ∂ α̇ α µ
(A (Γ) , D) = C (N ) ⊗ C[λ̄α̇ ] , D = λ̄α̇ − λ̄ θ σαα̇ ∂µ . (2.86)
∂ θ̄α̇

As emphasized above, we can relate this multiplet to the component field formulation
by computing the Koszul homology of Γ. Using n1 = S+ ⊕ S− , we see that the relevant
complex can be written as
 
• • ∂
∧ S+ ⊗ ∧ S− ⊗ C[λ̄α̇ ] , D0 = λ̄α̇ . (2.87)
∂ θ̄α̇
38 The pure spinor superfield formalism

Field Representative in the D0 -cohomology

φ φ
ψ ψθ
F F θ1 θ2

Table 2.1: Representatives for the N = 1 chiral multiplet in four dimensions organized
by θ-degree.

Here we introduced coordinates on S+ denoted by θα and on S− written as θ̄α̇ . Since θα

does not occur in the differential D0 , we find that the cohomology is a tensor product

∧• S+ ⊗ H • (∧• S− ⊗ C[λ̄α̇ ]) . (2.88)

However, it is easy to see that the second factor is acyclic, i.e. H • (∧• S− ⊗ C[λα ]) = C.
Thus, reinstalling the spacetime dependence, the D0 -cohomology reads

∧• S+ ⊗ C ∞ (V ) . (2.89)

We immediately see that we are dealing with two scalar fields in degrees 0 and 2 and a
Weyl fermion in degree 1. This is precisely the field content of the chiral multiplet. In
Table 2.1, we display the corresponding representatives and relate them to the component
fields of the chiral multiplet.

It is clear that the differential D10 acts trivially on these component fields. Hence, there
are also no further terms induced by homotopy transfer. We thus obtain a multiplet
described by a of super vector bundle

E 0 = V × ∧• S+ , (2.90)

with differential D0 = 0.

As the differential D vanishes, this is one of the rare cases where the supersymmetry
algebra acts strictly on the component fields. Expanding Q = α Qα and Q̄ = ¯α̇ Q̄α̇ we
have

∂
ρ(Q) = Q = α − i(σ µ θ̄)∂µ
∂θα (2.91)
∂
ρ(Q̄) = ¯Q̄ = ¯α̇ α̇ + i(θσ µ ¯)∂µ .
∂ θ̄
The pure spinor superfield formalism 39

The transferrred action only has a ρ0(1) component, which is given explicitly by

∂
ρ0(1) (Q) = p ◦ ρ(Q) ◦ i = α
∂θα (2.92)
ρ0(1) (Q̄) = p ◦ ρ(Q̄) ◦ i = i(θσ µ ¯)∂µ .

Now we can apply these to the representatives to find

ρ0(1) (Q)(φ) = 0 ρ0(1) (Q̄)(φ) = −i¯/

∂φθ
ρ0(1) (Q)(ψθ) = ψ ρ0(1) (Q̄)(θβ ψβ ) = i¯/ 1 θ2
∂ψθ (2.93)
ρ0(1) (Q)(F θ1 θ2 ) = F θ ρ0(1) (Q̄)(θ1 θ2 F ) = 0 .

Writing these relations in terms of operators we obtain the usual supersymmetry trans-
formation rules.

δφ = ψ
δψ = i¯/ + F
∂φ (2.94)
δF = −i¯/ .
∂ψ

2.3.6 Computational techniques: Koszul homology via free resolutions5

In the above example, we were able to compute the cohomology by hand and even could
write down explicit representatives easily. In general, such computations are much more
convoluted such that we have to rely on more advanced techniques. In this section,
we show how the cohomology can be computed from the minimal free resolution of the
module Γ and the corresponding Hilbert series. This allows for a fairly direct identifica-
tion of the ingredients of the multiplet. Further, using tools from the study of spectral
sequences, we can write down explicit formulas for the representatives.

Koszul homology and minimal free resolutions. Let us fix a nilpotence variety
Y and an R/I-module Γ. To understand the component field description of the multi-
plet A• (Γ), we are interested in the Koszul homology of Γ. The following proposition
shows that we can understand this by considering a free minimal resolution of Γ as an
R-module.

Proposition 2.3.3 ([MSX12; KL09]). Let L• → Γ → 0 be the minimal free resolution

of Γ in free R-modules. Then

H • (K • (Γ)) ∼
= L• ⊗R C . (2.95)
5
A preliminary version of this section was part of [Hah20].
40 The pure spinor superfield formalism

Proof. We denote the differential on the minimal free resolution L• by dL . By definition

we have 
Γ, if k = 0
H k (L• , dL ) = (2.96)
0, else.

This implies that there is a quasi-isomorphism

∧• n∨ • ∨ •



1 ⊗ Γ , D0 ' ∧ n1 ⊗ L , D0 + dL . (2.97)

Note that the right hand side is a bicomplex, such that we can use the associated spectral
sequence to compute its cohomology. We start by taking cohomology with respect to D0
and thus consider
∧• n∨ •


1 ⊗ L , D0 . (2.98)

It is easy to see that



C, if k = l
Hk ∧• n∨
1 ⊗ R[−l] , D 0 = (2.99)
0, else.

This means that we obtain a copy of C for each generator of L• . In total we get

H • ∧• n∨ • •


1 ⊗ L , D0 = L ⊗R C, (2.100)

where the R-module structure on C is obtained by applying the canonical augmentation

(quotienting out the maximal ideal). The differential on the first page is just the mor-
phism induced by dL . However, since L• is minimal, dL contains no constant terms and
therefore induces the zero map on the first page. Thus, we find the result as claimed.

The proposition reduces the task of computing Koszul homology to the task of finding
a minimal free resolution of Γ in R-modules. This can easily be done with commutative
algebra software such as Macaulay2 [GS]. In order to identify the field content of the
associated component field multiplet, we have to identify the Kosul homology not only
as a vector space, but as a representation for the Lorentz and R-symmetry groups. In
the next paragraph we show how this can be achieved by means of the Hilbert series;
later we will refine the technique (see §4.2.1).

L
Hilbert series. Let Γ = i≥0 Γi be a graded R-module generated by finitely many
elements in positive degree. The Hilbert series or graded dimension of Γ is the formal
power series
∞
X
grdim(Γ) = dim(Γn ) tn ∈ Z[[t]]. (2.101)
n=0
The pure spinor superfield formalism 41

Let R = C[λ] be the polynomial ring in a single variable λ. Since there is only a single
monomial in each degree the Hilbert series takes the form

∞
X 1
grdim(R) = tn = . (2.102)
1−t
n=0

As the dimension is multiplicative under the tensor product, the Hilbert series of a
polynomial ring in n variables R = C[λ1 , . . . , λn ] = C[λ1 ] ⊗ · · · ⊗ C[λn ] is just the
product
1
grdim(R) = . (2.103)
(1 − t)n
Now suppose we perform a shift R(−d) with respect to the polynomial degree such that
the constants are in degree d. We obtain for the Hilbert series

∞
X
grdim(R(−d)) = dim(R(−d)n ) tn
n=0
X∞
= dim(Rn−d ) tn
n=0
(2.104)
= td grdim(R)
td
= .
(1 − t)n

Thus, considering a free R-module Γ generated by elements in degree d1 , . . . , dk we find

td1 + · · · + tdk
grdim(Γ) = . (2.105)
(1 − t)n

The Hilbert series is additive with respect to short exact sequences. This means given a
sequence
0 −→ A −→ B −→ C −→ 0 , (2.106)

we find
grdim(B) = grdim(A) + grdim(C) . (2.107)

If L• is a free resolution of Γ, we have a sequence

Γ ←− L0 ←− L−1 ←− · · · ←− L−(k−1) ←− L−k ←− 0 . (2.108)

Then the additivity implies

k
X
grdim(Γ) = (−1)j−1 grdim(L−j ). (2.109)
j=1
42 The pure spinor superfield formalism

Using this together with (2.105), we can express the Hilbert series of Γ in terms of the
degrees of the basis vectors of the free resolution

k dj1 djnj
X
j−1 t + ··· + t
grdim(Γ) = (−1) . (2.110)
(1 − t)n
j=1

Coming back to our original question, we see that the Hilbert series of Γ as a R-module
contains all the information about the degrees of a basis of the minimal free resolution,
which in turn coincides with the Koszul cohomology. All we have to do is to store the
information about the transformation behavior under Lorentz and R-symmetry in the
grading. Therefore, we assign to λi the degree

deg(λi ) = (1, w1i , . . . , wli ) , (2.111)

where w1i , . . . , wli are the weights of the Lorentz and R-symmetry representation. The
first entry 1 remembers the weight degree. The Hilbert series then becomes a polynomial
j
in l + 1 variables t0 , . . . , tl . Equation (2.110) remains valid, but we have to replace tdi
by products of t0 , . . . , tl where each factor carries an exponent according to its weight.
Initializing such a grading in Macaulay2 and computing the Hilbert series, we can read
off the weights of a basis of the cohomology in each degree, allowing to identify the
cohomology as a representation of Lorentz- and R-symmetry.

Identifying representatives. Examining the proof of Proposition 2.3.3 closely, we

can deduce a procedure to write down explicit representatives for the cohomology classes
that correspond to the component fields of the multiplet. Recall that we used the quasi-
isomorphism
∧• n∨ • ∨ •



1 ⊗ Γ , D0 ' ∧ n1 ⊗ L , D0 + dL . (2.112)

On the right side we have a double complex of the form shown in Table 2.2. There
are two different spectral sequences computing the total cohomology: the horizontal
sequence starting with the differential dL and the vertical sequence starting with D0 . In
the proof of Proposition 2.3.3 we have seen that the latter already gives the exact result
for the zero mode cohomology on the first page. The computational procedure amounts
to coming to a better understanding of this bicomplex.

It is a fact that any bicomplex can be understood (non-canonically) as the sum of different
indecomposable pieces [Ste21]. These pieces are squares

• •
(2.114)
• •
The pure spinor superfield formalism 43

.. .. ..
. . .
D0 D0 D0

L0 ⊗ ∧2 n∨
1 dL
L−1 ⊗ ∧2 n∨
1 dL
L−2 ⊗ ∧2 n∨
1 dL
...
D0 D0 D0 (2.113)

L0 ⊗ ∧1 n∨
1 dL
L−1 ⊗ ∧1 n∨
1 dL
L−2 ⊗ ∧1 n∨
1 dL
...
D0 D0 D0

L0 ⊗ ∧0 n∨
1 dL
L−1 ⊗ ∧0 n∨
1 dL
L−2 ⊗ ∧0 n∨
1 dL
...

Table 2.2: The bicomplex obtained by using a free resolution to compute Koszul
homology

and stairs of different lengths

• •
• • • ... (2.115)
• • •

Here, the bullet denotes the underlying field • = K. Crucially, the decomposition can
be chosen such that all the arrows are just identity maps. The length of a stair is the
number of bullets • occurring.

One can understand the behavior of spectral sequences by thinking about the ways that
these indecomposable pieces contribute to cohomology. It is a matter of inspection to see
that stairs of even length are acyclic at the E1 page of one of the two spectral sequences
of the bicomplex, but contribute two generators to the E1 page of the other which cannot
be canceled by the differential on that page just for degree reasons. It is thus precisely
the (vertically or horizontally oriented) stairs of length 2k that contribute to differentials
on the Ek page of the corresponding spectral sequence. In contrast, stairs of odd length
contribute to the total cohomology of the complex, but do so in a bidegree that depends
on which spectral sequence is being considered. If we consider such a stair, we see that
the cohomology with respect to the horizontal differential is concentrated at the upper
end, while the cohomology with respect to the vertical differential lives at the lower end.
They are thus responsible for the breaking of the bigrading to the single homological
grading of the total complex.

Now note that the cohomology of our double complex (2.113) is concentrated in the
bottom row L• ⊗ ∧0 n∨ 0 • ∨
1 (for the vertical differential D0 ) and on the left column L ⊗ ∧ n1
(for the horizontal differential dL ). This implies that we have odd stairs contributing to
44 The pure spinor superfield formalism

the cohomology in the following manner:

• ...

• • ...
(2.116)
• • • ...

• • • • ...

Classes in the total cohomology can be represented by elements on either end of the
stair. However, if we want to view the representatives as elements in ∧• n∨
1 ⊗ Γ, we
have to apply the spectral sequence starting with dL , which amounts to choosing the
representatives on the upper end ∧• n∨ 0
1 ⊗ L and then projecting onto the quotient. On
the other hand, a basis of the vertical D0 -cohomology is clearly provided by the standard
basis {ei } ⊆ L−k ⊗∧0 n∨ nk k ∨ 0
1 = R . In order to get the desired basis in ∧ n1 ⊗L we have to
walk up the corresponding stair. Since we are now working only with Koszul complexes
of maximal ideals in polynomial rings, this can be done explicitly by defining a simple
“inverse” or “adjoint” differential to D0 by the formula6

∂
D0† = θα . (2.117)
∂λα

Then our discussion implies the following lemma.

Lemma 2.3.4. Let π : L0 −→ Γ be the projection. The elements π((D0† dL )k ei ) form a

basis of the cohomology H • (∧• n∨
1 ⊗ Γ) in θ-degree k.

2.3.7 Homotopy transfer to component fields

The new differential acting on the component fields, as well as the action of the su-
persymmetry algebra and, if present, an L∞ structure are obtained from the respective
structures on A• (Γ) via homotopy transfer. For this we need homotopy data

p
h (Gr A• (Γ), D0 ) (H • (Gr A• (Γ)), 0) . (2.118)
i

Using Lemma 2.3.4, we can define an inclusion map

i : H • (Gr A• (Γ)) ,→ (A• (Γ), D0 ) (2.119)

Note that this differential is well defined on ∧• n∨
6 • •
1 ⊗ L , since L consists of free R-modules. In
† • ∨
particular D0 will in general not descend to a well defined map on ∧ n1 ⊗ Γ.
The pure spinor superfield formalism 45

by sending a cohomology class to this representative. This inclusion is a quasi-isomorphism.

In addition, choosing a complementary subspace inside A• (Γ) gives the projection p. (We
always work equivariantly with respect to Lorentz and R-symmetry.)

The differential. Recall that we decomposed the differential on A• (Γ) as the sum of
two pieces, of filtered weight zero and two, respectively:

∂ µ ∂
D = λα α
− λα θβ fαβ = D0 + D1 . (2.120)
∂θ ∂xµ

We can thus view D1 as defining a deformation of the differential on Gr A• (Γ), which

in turn equips H • (Gr A• (Γ)) with a new differential D0 that is obtained by homotopy
transfer of D∞ structure [DSV15; LV12b; Lap01]. This uses the choice of a homotopy
datum to write all of the higher differentials of a spectral sequence as terms in a single
differential, acting on the E1 page, whose cohomology is E∞ . In formulas, we have

∞
X
0
D = Dn0 (2.121)
n=1

where the pieces are given by

Dn0 = p ◦ (D1 h)n−1 D1 ◦ i.

(2.122)

(Note that, due to our conventions for the filtration, only differentials on even pages are
non-trivial; the differential on page 2n is represented by Dn0 above.) Furthermore, we
can fix new homotopy data [Lap01]

p0
h0 (A, D) (H • (A, D0 ) , D0 ) , (2.123)
i0

where
∞
X ∞
X
0
i = i0n = (hD1 )n ◦ i
n=0 n=0
X∞ ∞
X
0
p = p0n =p◦ (D1 h)n (2.124)
n=0 n=0
X∞ X∞
h0 = h0n = h ◦ (D1 h)n .
n=0 n=0

We can use this homtopy data to transfer further structures, such as the action of the
supersymmetry algebra or an L∞ structure, from A• (Γ) to the component field descrip-
tion. Note that, in terms of sum-over-trees formulas, homotopy transfer with respect to
46 The pure spinor superfield formalism

the new homotopy data from (2.123) is expressed in terms of (2.118) simply by allowing
for unary vertices which are decorated by D1 .

The supersymmetry action. The supersymmetry action is obtained by a homotopy

transfer of L∞ module structures. As a result one obtains an map of super L∞ algebras

ρ0 : g D(E 0 ) , [D0 , −] ,


(2.125)

whose component maps can be obtained via sum over trees formulas. For example ρ0(2)
is given by

ρ0(2) (x1 , x2 ) = p0 ◦ ρ(x1 ) ◦ h0 ◦ ρ(x2 ) ± ρ(x2 ) ◦ h0 ◦ ρ(x1 ) ◦ i .

(2.126)

Interestingly, there is a close link between the resolution differential and the action of
the supersymmetry algebra. This connection was already conjectured in [Ber02], where
it was noticed that the non-derivative supersymmetry transformations and their closure
terms appear in the resolution differential of eleven-dimensional supergravity. Using our
knowledge on the representatives and the homotopy transfer description of the action of
the supersymmetry transformations we can make this observation precise and provide a
derivation. Later, we will see that this result is an easy consequence of the equivalence
of categories developed in §3 (see Corollary 3.4.8).

We note that the strict part of a non-derivative supersymmetry transformation acts by

∂
Q0 := ρ∂x =0 (Q) = α . (2.127)
∂θα

In addition it is easy to see that

∂
[Q0 , D0† ] = α (2.128)
∂λα

and obviously
[Q0 , dL ] = 0 . (2.129)

Now suppose Q0 acts on a representative in θ-degree k

0(1)
ρ∂x =0 (f ) = p ◦ Q0 ◦ i(f )
(2.130)
= p ◦ Q0 ◦ π(D0† dL )k (f i ei ) .
(k)

(k)
Here (ei ) denotes a basis of L−k ⊗R C and π : L0 −→ Γ the projection. Note that
Q0 ◦ π = π ◦ Q0 . In the following, we abbreviate the components of the resolution
differential by dk := (dL )k . Now we can use the relation (2.128) to bring Q0 to the right.
The pure spinor superfield formalism 47

We find  
k
0(1) ∂
D0† d1 . . . D0† dj−1  dj D0† . . . dk (f ) .
X
ρ∂x =0 (f ) = p ◦ π  (2.131)
∂λ
j=1

Since we already know the explicit form of the representatives, we can carry out the
projection to D0 -cohomology directly. The only remaining term is the following.
 
0(1) † k−1 ∂ i (k)
ρ∂x =0 (f ) = π (D0 dL )  dk (f ei ) . (2.132)
∂λ

Furthermore, only the part of (dL )k linear in λ can contribute in D0 -cohomology. Then
∂
 ∂λ simply replaces λ with  in the dk . Let us denote the resulting map by dk and its
components by (dk )i j . Then we find
 
0(1)
ρ∂x =0 (f ) = π (D0† dL )k−1 (dk )i j ej
(k−1) i
f
  (2.133)
= π (D0† dL )k−1 (ej
(k−1)
)(dk )i j f i .

Identifying the representative in degree k − 1 and writing the transformation rule dually
in terms of operators, we find
δg j = (dk )i j f i , (2.134)

where g j denotes the operator corresponding to the respective representative in θ-degree

k − 1. This shows that linear parts in the resolution differential precisely correspond to
the strict part of the non-derivative supersymmetry transformations.

This generalizes to the higher components of the supersymmetry action. For n ≥ 2, the
non-derivative part of ρ(n) acts is given by

ρ0(n) = p ◦ (Q0 ◦ h ◦ Q0 )n−1 ◦ i . (2.135)

For example one finds for ρ0(2)

0(2)
ρ∂x =0 (Q, Q)(f ) = p ◦ Q0 ◦ h ◦ Q0 ◦ i(f )
 
k (2.136)
∂
D0† d1 . . . D0† dj−1  dj D0† . . . dk (f ) .
X
= p ◦ Q0 ◦ h ◦ π 
∂λ
j=1

Now assuming that the homotopy h acts via h ◦ π = π ◦ D0† we find using (D0† )2 = 0
 
0(2) † ∂ †
ρ∂x =0 (Q, Q)(f ) = p ◦ π Q0 D0  d1 D0 . . . dk (f )
∂λ
  (2.137)
∂ † †
= p ◦ π  Q0 D0 d1 D0 . . . dk (f ) ,
∂λ

∂
where we used that  ∂λ commutes with both D0† and Q0 .
48 The pure spinor superfield formalism

Now we can again use the relation (2.128) to find

 
k
0(2) ∂ ∂
D0† d1 . . . D0† dj−1  dj D0† . . . dk (f ) .
X
ρ∂x =0 (Q, Q)(f ) = p ◦ π  (2.138)
∂λ ∂λ
j=1

Carrying out the projection p on D0 -cohomology, we see that only one term survives.
 
0(2) 2
ρ∂x =0 (Q, Q)(f ) = π (D0† d)k−1 (ej
(k−1)
)(dk )i j f i , (2.139)

2
where dk denotes the quadratic part of the resolution differential with λ’s replaced by
’s. Written in terms of operators this gives a transformation rule

2
δg j = (dk )i j f i . (2.140)

Using a similar calculation as above one sees that only the part of order n in the resolution
differential contributes to a supersymmetry transformation and we obtain supersymmetry
transformation rules of the form

n
δg j = (dk )i j f i . (2.141)

Interestingly this provides a direct link between the polynomial degree of the terms in
the resolution differential and the homotopy action of the supersymmetry algebra. That
is, if the resolution differential is at most quadratic, then the L∞ module structure will
contain at most ρ0(2) corrections.

L∞ structures. If (A• (Γ), D) carries an L∞ structure with differential D, this structure

can be transferred as well. For this one uses the usual sum over trees formulas. As we
will see below, the transferred L∞ structure on the component fields can encode the
structure of gauge transformations and in some cases also interactions. Note that the
new L∞ structure has µ01 = D0 the transferred differential. We will see this explicitly in
the case of ten-dimensional super Yang–Mills theory.

2.3.8 An example of the technique: the 4d gauge multiplet

To illustrate these techniques, we are going to perform all the necessary calculations for
the d = 4, N = 1 vector multiplet by hand. Let Y be the nilpotence variety of the N = 1
super Poincaré algebra in four dimensions. We choose the structure sheaf OY as input;
in the language of [Ced+23], we construct the canonical multiplet of the N = 1 super
Poincaré algebra in four dimensions. Using Macaulay2 we can compute the minimal free
The pure spinor superfield formalism 49

resolution. Its Betti numbers are displayed in the following table.

" #
1 − − −
(2.142)
− 4 4 1

Here, the horizontal axis denotes degree in θ, while the vertical axis counts powers in λ.
To analyze the field content of the multiplet as representations of the Lorentz group, we
assign gradings to the generators λ and λ̄, corresponding to their weights under

so(4) ∼
= su(2) × su(2) . (2.143)

Concretely this means that we assign the grading

deg(λ1 ) = (1, 1, 0) deg(λ2 ) = (1, −1, 0)

(2.144)
deg(λ̄1 ) = (1, 0, 1) deg(λ̄2 ) = (1, 0, −1) .

Then we examine the numerator of the Hilbert series. We organize the terms by degree
in the variable t0 , which indicates the total degree in the complex. In degree 0 we simply
obtain 1, which means the field in total degree 0 is a scalar. In degree 2 we find the term

− t20 (t1 t2 + t1 t−1 −1 −1 −1

2 + t1 t2 + t1 t2 ) . (2.145)

Reading off the highest weights we see that the corresponding representation of SU (2) ×
SU (2) is
[1, 1] = [1, 0] ⊗ [0, 1] , (2.146)

which shows that the field in degree 2 is a vector. In degree 3 we obtain

t30 (t1 + t−1 −2

1 + t2 + t2 ) . (2.147)

Correspondingly, the representation in degree 3 is a direct sum

[1, 0] ⊕ [0, 1] . (2.148)

Hence, the field in degree 3 is a Dirac fermion. Finally the term of order 4 is just −t40
indicating that the field in degree 4 is a scalar. This means that we recover the usual
field content of the d = 4, N = 1 vector multiplet.

To find representatives with the procedure explained above, we need the differential
on the free resolution. The minimal free resolution is of the form
 
(dL )1 (dL )2 (dL )3
R ⊗ C ←−−− V ←−−− S+ ⊕ S− ←−−− C . (2.149)
 50 The pure spinor superfield formalism

The differential can be described by the matrices

 
(dL )1 = λ1 λ̄1 λ1 λ̄2 λ2 λ̄1 λ2 λ̄2
 
0 −λ̄2 0 −λ2
 
 0 λ̄1 λ2 0 
(dL )2 = 
 

−λ̄2 0 0 λ1 
 
λ̄1 0 λ1 0 (2.150)
 
λ1
 
−λ 
 2
(dL )3 =  .
−λ̄1 
 
λ̄2

Choosing a basis eαα̇ of V and (sα , s̄α̇ ) of S+ ⊕ S− , these maps can be conveniently
packaged as follows.

(dL )1 : V −→ C , A 7→ λα λ̄α̇ Aαα̇

(dL )2 : S+ ⊕ S− −→ V , (ψ, ψ̄) 7→ (λα ψ̄ α̇ + ψ α λ̄α̇ )eαα̇ (2.151)
(dL )3 : C −→ S+ ⊕ S− , D 7→ (λα s α − λ̄α̇ s̄α̇ )D

Note that we can apply the isomorphism S+ ⊗ S− ∼

= V by a change of basis eµ =
(σµ )αα̇ eαα̇ . With this description, it is easy to identify representatives in D0 -cohomology.
For example, the vector is represented by

(dL )1 D†
A 7−−−→ (λσ µ λ̄)Aµ 7−−→
0
(λσ µ θ̄ + λ̄σ µ θ)Aµ . (2.152)

For the fermions we find

(dL )2 D† (dL )1 D†
ψ 7−−−→ ψ α λ̄α̇ eαα̇ 7−−→
0
ψ α θ̄α̇ eαα̇ 7−−−→ ψ α θ̄θ̇ λα λ̄α̇ 7−−→
0
ψ α θ̄α̇ (λα θ̄α̇ + θα λ̄α̇ ) (2.153)

A similar calculation gives the complex conjugate representative for ψ̄. Finally we can
apply the procedure to the auxiliary field.

D† ◦(dL )3 (dL )2 D† (dL )1 D†

D 7−−0−−−−→ (θs−θ̄s̄)D 7−−−→ (θα λ̄α̇ −λα θ̄α̇ )eαα̇ D 7−−→
0
2θα θ̄α̇ eαα̇ D 7−−−→ 2(θλ)(θ̄λ̄)D 7−−→
0
2(θ2 λ̄θ̄+θ̄2 λθ)D
(2.154)
We summarize these representatives in Table 2.3. Note that these representatives are not
unique. Other choices are possible; for example one can simplify these representatives
by eliminating terms in the image of D0 . For instance the antisymmetric expression

λα θ̄α̇ − λ̄α̇ θα (2.155)

The pure spinor superfield formalism 51

Field Representative in the D0 -cohomology

c c
A (λσ µ θ̄ + θσ µ λ̄)Aµ
ψ ψ α θ̄α̇ (λα θ̄α̇ + θα λ̄α̇ )
ψ̄ ψ̄ α̇ θα (λ̄α̇ θα + θ̄α̇ λα )
D (θ2 λ̄θ̄ + θ̄2 λθ)D

Table 2.3: Representatives for the d = 4, N = 1 vector multiplet organized by

θ-degree.

is clearly in the image of D0 . This implies that we could represent the vector equally
well by λα θ̄α̇ . Similar observations also hold for the other fields.

Let us now study the different structures arising on the component fields via homotopy
transfer.

The differential. By degree reasons, only the first order part D10 of the transferred
differential D0 can act non-trivially on the component fields. Recall

D10 = p ◦ D1 ◦ i , (2.156)

where
D1 = (λσ µ θ̄ + λ̄σ µ θ)∂µ . (2.157)

The only non-vanishing contribution arises by acting on the ghost. There we find

D1 c = (λσ µ θ̄ + λ̄σ µ θ)∂µ c . (2.158)

Identifying the representative of the gauge field, we see that the differential is simply the
de Rham differential
c 7→ dc . (2.159)

Written dually in terms of operators this gives the BRST differential

QBRST Aµ = ∂µ c . (2.160)
52 The pure spinor superfield formalism

The following picture summarizes the complex on the component field level.

Ω0 (R4 )
d
(2.161)
Ω1 (R4 ) Γ(R4 , S+ ⊕ S− ) Γ(R4 , C) .

The action of the supersymmetry algebra. As explained above, we can read off the
non-derivative part of the supersymmetry transformations directly from the resolution
differential. This gives transformation rules

δc = (σ µ ¯)Aµ
δAµ = σµ ψ̄ + ψσµ ¯
δψ = D (2.162)
δ ψ̄ = −¯
D
δD = 0 .

Note that there is one higher order component indicating that the action of the super-
symmetry algebra is not strict. We will come back to this in a moment.

First, let us investigate the contributions containing derivatives. By degree reasons there
cannot appear any higher order contributions containing derivatives, such that we can
focus on the strict part. The derivative part of ρ0(1) acts on the representatives by

Q1 = σ µ θ̄∂µ + θσ µ ¯∂µ . (2.163)

For example we can act on the fermions to find

σµ θ)∂µ ψ α θ̄α̇ (λα θ̄α̇ + θα λ̄α̇ )

Q1 (ψ) =(¯
(2.164)
β̇ σβµβ̇ ∂µ ψ α )(λα θβ θ̄2 + θα θβ λ̄θ̄) .
=(¯

Projecting to cohomology this equals

/
¯∂ψ(λθ θ̄2 + θ2 λ̄θ̄) , (2.165)

such that we can identify a transformation rule

/ .
δD = ¯∂ψ (2.166)

A similar calculation also holds for the complex conjugate ψ̄, as well as for the gauge
field and yield the usual supersymmetry transformation rules.
The pure spinor superfield formalism 53

This describes the entire L∞ module structure of the superymmetry algebra on the
four-dimensional, N = 1 vector multiplet. The ρ0(1) part resembles the well known
supersymmetry transformations from standard physics textbooks. In addition there is
one higher correction. Recall that we found a transformation rule

δc = (σ µ ¯)Aµ . (2.167)

This corresponds to a map ρ0(2) given by

ρ0(2) : n ⊗ n ⊗ Ω1 −→ Ω0 (Q1 ⊗ Q2 ⊗ A) 7→ ι[Q1 ,Q2 ] A . (2.168)

We can immediately check that ρ0(2) indeed defines a homotopy correcting for the failure
of ρ0(1) to be strict. We clearly have

ρ0(1) (Q)(c) = ρ0(1) (Q̄) = 0 . (2.169)

However, the bracket of Q and Q̄ gives a translation which acts via the Lie derivative

[Q, Q̄](c) = L[Q,Q̄] (c) . (2.170)

Thus, according to (2.26) we have to check

L[Q,Q̄] (c) = −[D, ρ0(2) (Q, Q̄)](c) . (2.171)

Plugging in D = d the de Rham differential, we obtain

L[Q,Q̄] (c) = − (d ◦ ι[Q1 ,Q2 ] − ι[Q1 ,Q2 ] ◦ d)(c)

(2.172)
=(ι[Q1 ,Q2 ] ◦ d)(c) ,

where the first term vanishes by degree reasons. We immediately see that this is indeed
satisfied due to Cartan’s magic formula. This discussion illustrates that the ρ0(2) -term
indeed provides a homotopy for the failure of ρ0(1) to be strict. In terms of physics
terminology, ρ0(2) is a closure term for the supersymmetry action, which closes only up
to gauge transformations.

L∞ structure. To treat the non-abelian vector multiplet we can tensor the entire
construction with a Lie algebra h. We notice that R/I is not only an R/I-module,
but a ring. Hence, A• (R/I) carries an algebra structure such that the tensor product
A• (R/I) ⊗ h comes equipped with an L∞ structure given by

µ1 = D ⊗ idh µ2 = m2 ⊗ [−, −] . (2.173)

54 The pure spinor superfield formalism

Here m2 denotes the multiplication in A• (R/I). Since the differential does not inter-
fere with the Lie algebra at all, the component fields of the multiplet take values in
H • (K • (R/I)) ⊗ h. This is just the field content of the abelian version only now taking
values in the Lie algbera h. The transfer of the L∞ algebra structure to the component
fields is very simple. The differential only acts on the ghost fields via the de Rham
differential.
µ01 = d ⊗ idh : Ω0 ⊗ h −→ Ω1 ⊗ h . (2.174)

In addition to the differential, only two-ary brackets arise.

µ02 : Ω 0 ⊗ h × Ω0 ⊗ h −→ Ω0 ⊗ h µ02 (c, c) = [c, c]

µ02 : Ω 0 ⊗ h × Ω1 ⊗ h −→ Ω1 ⊗ h µ02 (c, A) = [c, A]
µ02 : Ω0 ⊗ h × Γ(X, S+ ⊕ S− ) ⊗ h −→ Γ(X, S+ ⊕ S− ) ⊗ h µ02 (c, ψ) = [c, ψ] .
(2.175)
We can also write these dually as a BRST operator.

1
QBRST c = − [c, c]
2
QBRST A = dc + [A, c]
(2.176)
QBRST ψ = [ψ, c]
QBRST D = [D, c] .

Hence, we recover the usual BRST complex of the d = 4, N = 1 gauge multiplet. To

equip the multiplet with a BRST datum, we could write the usual component field action
for the gauge multiplet. In the terminology of §2.2 this action then makes the multiplet
a BRST theory.

2.3.9 Scheme-theoretic properties: three-dimensional N = 1 super-

symmetry

In three dimensions we have the isomorphism Spin(3) ∼

= SU (2). The vector representa-
tion V corresponds to the three-dimensional representation of SU (2), while the spinor
representation S is given by the two-dimensional representation. The bracket is provided
by the isomorphism
Sym2 (S) ∼
=V . (2.177)

Therefore, as a set, the nilpotence variety is simply a point

Y = {0} . (2.178)
The pure spinor superfield formalism 55

Even though the nilpotence variety, regarded as a set, is just a point it still may carry an
interesting structure as a scheme which allows for the construction of different multiplets.
Expanding the equation [Q, Q] = 0 in coordinates (λ1 , λ2 ) we obtain the equations

(λ1 )2 = λ1 λ2 = (λ2 )2 = 0 . (2.179)

Clearly, the only solution to these equations is λ1 = λ2 = 0. However, the quotient ring
R/I  C differs from just the constants which are the functions on the point considered
as an affine variety. This reflects the fact that the affine scheme Spec(R/I) is not just an
ordinary point, but what is called a fat point. As we will see momentarily, this allows us
to construct different multiplets from R/I-modules, even though there are no non-trivial
square-zero supercharges.

The gauge multiplet. First of all we can consider R/I itself as an equivariant module.
This gives rise to the gauge multiplet in three dimensions. The minimal free resolution
has the following Betti numbers. " #
1 − −
(2.180)
− 3 2

In terms of representations, the free resolution takes the form

 
(dL )1 (dL )2
R ⊗ C ←−−− V ←−−− S (2.181)

with the differentials being described by

(dL )1 : V −→ C , A 7→ (λσ µ λ)Aµ

(2.182)
(dL )2 : S −→ V , ψ 7→ (λσ µ ψ)eµ .

Thus, we find that the multiplet contains a one-form field together with its ghost as
well as a fermion. The only differential acting on the component fields is the de Rham
differential
c 7→ dc (2.183)

which encodes the gauge invariance of the one-form. The non-derivative supersymmetry
transformations can be read off from the resolution differential and take the usual form.

δc = (σ µ )Aµ
δAµ = σµ ψ (2.184)
δψ = 0 .
56 The pure spinor superfield formalism

The free superfield. In addition, we can also consider C = R/(λ1 , λ2 ) as an R/I-

module. This yields the free superfield whose Betti numbers we display in the following
table.
h i
1 2 1 (2.185)

Indeed, the Koszul homology of this module is just an exterior algebra ∧• S and we just
recover the usual superspace description of the free superfield.

2.4 From multiplets to theories

In §2.2.5, we introduced the notions of BV and BRST data for multiplets. Under certain
conditions on the module Γ, the Koszul homology is naturally equipped with a perfect
pairing that equips the corresponding multiplet with a BV datum. This provides an inter-
esting link between the physics of supersymmetric multiplets and the algebraic geometry
of OY -modules. In fact, the pure spinor formalism provides many such links between
algebrogeometric properties of the module Γ and physical properties of the multiplet.
We start exploring these in the following section.

2.4.1 Commutative algebra and dualizing complexes

Let us start with a short survey of the relevant notions from commutative algebra. As
motivated by the pure spinor superfield formalism, we are mostly interested in quotients
of polynomial rings R = C[λ1 , . . . , λn ] by (quadratic) ideals I. We think of R/I as the
ring of functions on a nilpotence variety. The main sources for our discussion are [Eis95;
Bas63; Hun99].

Definition 2.4.1. A quotient ring R/I is called a complete intersection, if I can be

generated by r = codim(R/I) = n − dim(R/I) elements, i.e. I = (f1 , . . . , fr ).

Intuitively, this definition means that there are no non-trivial relations among the fi such
that each equation cuts the dimension of the corresponding variety by one. Equivalently
we can say that f1 , . . . , fr forms a regular sequence on R. To be clear we recall the
definition.

Definition 2.4.2. Let S be a commutative ring and M a S-module. A sequence

(x1 , . . . , xk ) ⊂ S is called M -regular if xi is not a zero divisor in M/(x1 , . . . , xi−1 ) for all
i = 1, . . . , k.

One can define a notion of “size” for modules by asking for the maximal length of a
regular sequence in M . The resulting number is called the depth of M .
The pure spinor superfield formalism 57

Definition 2.4.3. The depth of a S-module M with respect to an ideal J ⊆ S is the

maximal length of an M -regular sequence in J and will be denoted by depthJ (M ). For
local ring (S, m) with maximal ideal m, one writes depthm (M ) = depth(M ).

We are interested in modules Γ over the polynomial ring R. Maximal ideals then cor-
respond to points x ∈ Spec R; there exists a unique maximal equivariant ideal in R,
corresponding to the skyscraper sheaf at the origin, and consisting of all polynomials
with zero constant term. We will always consider the depth with respect to this ideal.

On general grounds, one can show that for any module depth(Γ) ≤ dim(Γ). There is an
important class of modules for which equality holds. These are called Cohen–Macaulay
modules.

Definition 2.4.4. An R-module Γ is called Cohen–Macaulay if depth(Γ) = dim(Γ).7

In this context, two equivalent characterizations of the Cohen–Macaulay property will

be useful. The first one is in terms of the length of a minimal free resolution.

The Auslander–Buchsbaum formula states

depth(Γ) + lR (Γ) = n , (2.186)

where lR (Γ) is the length8 of the minimal resolution L• of Γ in free R-modules.

Proposition 2.4.5. An R-module Γ is Cohen–Macaulay if the length of its minimal free

resolution equals its codimension, i.e.

lR (Γ) = n − dimR (Γ) = codimR (Γ) . (2.187)

Equivalently we can characterize Cohen–Macaulay modules via their Ext-groups. We

define the dualizing complex by

ωΓ• = RHom•R (Γ, R) = HomR (L• , R). (2.188)

We note that the cohomology H • (RHom•R (Γ, R)) ∼

= Ext•R (Γ, R). If R/I is Cohen–
Macaulay, this cohomology is concentrated in a single degree.

Proposition 2.4.6. An R-module Γ of dimension q is Cohen–Macaulay if and only if

ExtkR (Γ, R) = 0 for all k 6= codimR (Γ) = n − q =: r.
7
Here the correct notion of dimension is the Krull dimension.
8
The length of a free resolution L• = (L0 ← L−1 ← · · · ← L−k ← 0) is k.
58 The pure spinor superfield formalism

Thus, for Cohen–Macaulay modules, the dualizing complex is in fact quasi-isomorphic

to a dualizing module (often also called the canonical module).

Let us now specialize to the ring of functions, i.e. Γ = R/I. If the canonical module
ωR/I is trivial (free of rank one), the scheme Spec(R/I) can be thought of as a singular
analog to a Calabi–Yau space. This property is called Gorenstein.

Definition 2.4.7. A quotient ring R/I is called Gorenstein if R/I is Cohen–Macaulay

−(n−d)
and the dualizing module ExtR (R/I, R) = R/I, where d = dim(R/I).9

Clearly, the Gorenstein property is stronger than the Cohen–Macaulay property. How-
ever, to be a complete intersection is an even stronger condition. We thus have the
following chain of implications.

Complete intersection =⇒ Gorenstein =⇒ Cohen–Macaulay (2.189)

The key property of Gorenstein rings that is relevant to us is that their minimal free
resolutions are self-dual: If R/I is Gorenstein and (L• , dL ) is a minimal free resolution,
then the dual complex ((L• )∨ , (dL )∨ ) is, by definition, a resolution of the dualizing
module, which, by assumption, is again R/I. Thus, (L• , dL ) and ((L• )∨ , (dL )∨ ) are both
minimal free resolutions for R/I and hence, due to the uniqueness of the minimal free
resolution, they must be isomorphic.

In fact one can recognize Gorenstein rings conveniently by examining their minimal free
resolution:

Proposition 2.4.8. R/I is a Gorenstein ring ⇐⇒ The length of the minimal free
resolution L• of R/I is lR (R/I) = codimR (R/I) =: k and L−k = R.

Note that this extends the above statement on the Cohen–Macaulay property. The self-
∼ (L−(k−i) )∨ and thus
duality of the minimal free resolution induces isomorphisms L−i =
a non-degenerate pairing
L−i × L−(k−i) −→ R . (2.190)

Tensoring both sides with C we obtain a pairing

(L−i ⊗R C) × (L−(k−i) ⊗R C) −→ C . (2.191)

As we explained in §2.3.6, L• ⊗R C can be identified with Koszul homology and thus

with the component fields of the multiplet. As such, if we feed a Gorenstein ring into
9
This is not the most general definition, but it suits our setting. In general a ring S is called
Gorenstein, if S has finite injective dimension as an S-module. There is also a notion of Gorenstein
modules in the literature, but we do not need this level of generality for our discussion.
The pure spinor superfield formalism 59

the pure spinor superfield formalism, we can equip the component fields of the resulting
multiplet with a local pairing (a density valued pairing on sections of a vector bundle on
spacetime). The parity and homological degree will depend on the properties of the free
resolution. These pairings are often of physical interest.

2.4.2 Supplemental structures on multiplets

In some cases these pairings can be used to equip multiplets obtained in the pure spinor
superfield formalism with a BV datum. Here, the prime example is ten-dimensional super
Yang–Mills theory which we will discuss below. However, this is not the only relevant
case. There are other examples of multiplets obtained from Gorenstein rings where the
pairing does not give rise to a BV structure; nevertheless, the natural pairings may still
be interesting.

As an easy example, let us once again come back to the chiral multiplet for N = 1
supersymmetry in four dimensions. Recall that we obtained the chiral multiplet from
the module Γ = C[λ̄α̇ ] = C[λα , λ̄α̇ ]/(λα ). This is obviously a complete intersection ring,
and thus in particular Gorenstein. The minimal free resolution is of the form

R ←− R2 ←− R, (2.192)

and it is clear what the pairing looks like: L0 = R pairs with L−2 = R, while L−1 = R2
pairs with itself. Since this is a perfect pairing on Koszul homology, we obtain a local
pairing on the component fields. Recall that the scalar field was represented simply by
φ, the fermion by ψ = ψα θα and the auxiliary by F = F θ1 θ2 . Thus, we get a pairing
which is simply induced by the algebra structure on ∧• S+ and the projection on the θ1 θ2
component
ha, bi = (ab)|θ1 θ2 . (2.193)

So this pairing gives rise to F-term Lagrangians for the chiral multiplet through the
following local pairing on component fields

hφ, F iloc = φF, hψ, ψiloc = ψ α ψα . (2.194)

Similar pairings of course exist for other chiral multiplets with more supersymmetry.
Furthermore, we could consider the free superfield, arising from the quotient of R by
the maximal ideal (or more geometrically from the skyscraper sheaf with value C on Y
supported at the origin). For four-dimensional N = 1 supersymmetry, this module is
C = C[λα , λ̄α̇ ]/(λα , λ̄α̇ ). Then one gets a pairing which projects on the θ2 θ̄2 component.
In physics, this pairing gives rise to D-terms.
60 The pure spinor superfield formalism

2.4.3 Constructing cotangent theories: six-dimensional N = (1, 0)

If a ring is not Gorenstein, there is no perfect pairing on Koszul homology, and the corre-
sponding multiplet cannot obviously be equipped with a BV structure. (We note that this
does not mean that such multiplets are never on-shell; in six-dimensional N = (2, 0) su-
persymmetry [CNT02; SW23b] and ten-dimensional type IIB supersymmetry [ESW21],
BV multiplets with degenerate pairings naturally arise. Details on the pairing are given
in [SW23b] at the level of the component fields; we do not study the cochain-level origin
of such degenerate pairings here, but hope to return to this question in future work.)

For a Cohen–Macaulay module Γ giving rise to a multiplet (E, D, ρ), however, another
interesting observation applies: We can consider the dualizing module ωΓ in the pure
spinor superfield formalism. If (L, dL ) is the minimal free resolution of Γ, then (L∨ , (dL )∨ )
is the corresponding minimal free resolution of ωΓ . With the obvious pairing between L
and L∨ we can equip the multiplet corresponding to the direct sum L ⊕ L∨ [k] with a BV
datum (for an appropriate shift k). In the terminology of Definition 2.2.15 the resulting
BV multiplet is off-shell and ωΓ gives rise to the antifield multiplet of (E, D, ρ).

On the other hand, if the input module is not Cohen–Macaulay, the cohomology of the
dualizing complex will not be concentrated in a single degree, such that we cannot take a
single dualizing module to produce an antifield multiplet. Rather, the antifield multiplet
will be represented by a dg module. We will see this below in the case of four-dimensional
minimal supersymmetry.

Let us now consider the example of six-dimensional N = (1, 0) supersymmetry. There

is an accidental isomorphism Spin(6) ∼
= SU (4), under which the spinor representation
S+ is identified with the fundamental representation of SU (4) and S− = (S+ )∨ with the
antifundamental representation. The supertranslation algebra takes the form

n = V ⊕ Π(S+ ⊗ U ) , (2.195)

where U = (C2 , ω) is a symplectic vector space. The R-symmetry group is thus Sp(1) ∼
=
SU (2); corresponding indices will be denoted by i, j. There is an isomorphism

∧2 S+ ∼
=V , (2.196)

which is used to express the bracket as

[−, −] = ∧ ⊗ ω . (2.197)
The pure spinor superfield formalism 61

Since ∧ is an isomorphism, an element is square-zero precisely when it is of rank one as

an element of S+ ⊗ U .

In a basis, the supertranslation algebra takes the form

[Qiα , Qjβ ] = γαβ

µ ij
ε Pµ . (2.198)

Using coordinates λαi , the defining equations of the nilpotence variety Y are given by the
2 × 2 minors of the matrix !
λ11 λ21 λ31 λ41
, (2.199)
λ12 λ22 λ32 λ42

which cut out the space of rank-one matrices. As such Y is a determinantal variety.
Taking the ring of functions R/I as the equivariant module in the pure spinor formalism,
we recover the d = 6, N = (1, 0) gauge multiplet. The Betti numbers of the free
resolution are displayed in the following table.
" #
1 − − −
(2.200)
− 6 8 3

Working equivariantly, one finds that these correspond to a one-form with zero-form
gauge invariance, fermions in S+ ⊕ S− , and a triplet of scalars in the adjoint of the
R-symmetry group SU (2). We immediately see that the Koszul homology corresponds
to the field content of the BRST complex of the gauge multiplet. Since the length of the
resolution equals the codimension, R/I is Cohen–Macaulay. This can also be seen as a
consequence of the following result on determinantal varieties.

Lemma 2.4.9 ([HE71; Sva74]). Let R = C[(xij )] for i = 1 . . . n and j = 1 . . . m and I

the ideal generated by the r × r minors of the matrix with entries xij . Then R/I is a
Cohen–Macaulay ring. Further R/I is a Gorenstein ring if and only if m = n or r = 1.

As we are dealing with 4 × 2 matrices, R/I is not Gorenstein; hence, we cannot expect
to equip the multiplet with a BV datum, but only with a BRST datum. However, R/I
is Cohen–Macaulay, which means that the dualizing complex is represented by a single
sheaf. Thus, we can produce the corresponding antifield multiplet from that sheaf by
applying the pure spinor formalism. The dualizing module is

−codim(Y )
ExtR (R/I, R) = Ext−3
R (R/I, R) . (2.201)
62 The pure spinor superfield formalism

Due to the Cohen–Macaulay property, this is the only non-vanishing Ext-module. Its
free resolution has the Betti numbers
" #
3 8 6 −
. (2.202)
− − − 1

Forming the direct sum of the structure sheaf and the dualizing sheaf and shifting ap-
propriately, we obtain a multiplet with the following Betti numbers.
 
− − − − −
1
 
− 6 8 3 − −
(2.203)
 
 
− − 3 8 6 −
 
− − − − − 1

This is the expected field content of the BV description for the six-dimensional gauge
multiplet. The component multiplet can be equipped with a BV datum by writing the
usual action as known from the component formalism. The resulting BV multiplet is
off-shell; in fact, it is constructed as the cotangent theory of the corresponding BRST
theory. The supersymmetry algebra closes without use of the equations of motion and the
antifields can be separated from the fields. Doing this, one recovers the BRST multiplet
in components.

One could also consider equipping the pure spinor superfield multiplet with a BRST
datum. This was done in [Ced18b], where Cederwall considered a differential operator
mapping pure spinor superfields for the structure sheaf to pure spinor superfields for the
canonical module. This operator allows one to write a quadratic action functional for
the structure sheaf multiplet, which defines a BRST datum for that multiplet.

2.4.4 Failure to be Cohen–Macaulay: the example of four-dimensional

N =1

As we have seen above, the pure spinor superfield formalism applied to the structure
sheaf of the d = 4, N = 1 nilpotence variety yields the BRST description of the gauge
multiplet. The absence of antifields and BV differential is not particularly surprising: The
failure of the supersymmetry action to be strict solely comes from gauge transformations;
the equations of motions do not need to be imposed. Nevertheless one can ask if and how
the corresponding antifield multiplet can be realized independently in the pure spinor
superfield formalism. For this purpose, let us compute the dualizing complex of R/I. A
model for the dualizing complex is given by

•
ωR/I = RHom•R (R/I, R) = HomR (L• , R). (2.204)
The pure spinor superfield formalism 63

To compute this complex explicitly, we can use the minimal free resolution L• → R/I
from above. By definition, the differential of the dualizing complex is the dual map d∨
L
of the resolution differential dL . In terms of matrices this means that d∨
L is represented
by the transposed matrices of (2.150). From these matrices, the cohomology can be
computed explicitly. We find that

coker (λ1 , −λ2 , −λ̄1 , λ̄2 ) ∼




 = C, i = 3;

•
H i (ωR/I ) = C[λ1 , λ2 ] ⊕ C[λ̄1 , λ̄2 ], i = 2; (2.205)



0, otherwise.


Note that the codimension of Y is two. If the dualizing complex were to resolve a single
module, then H • (ωR/I ) should be concentrated in degree two. Instead, we find a copy
of two disjoint C2 ’s; Y itself, of course, consists of two C2 ’s intersecting at the origin.
This discrepancy is accounted for homologically by the presence of a single copy of the
skyscraper sheaf in degree three. At the end of the day, this means that we cannot
find a single (non-dg) dualizing module for R/I to feed into the pure spinor superfield
formalism to obtain the antifield multiplet. This phenomenon will occur whenever R/I
is not a Cohen–Macaulay ring. We will come back to this example in the context of the
derived pure spinor superfield formalism in §3.

2.4.5 A partial dictionary

In this section we summarize some features of the correspondence between algebrogeo-

metric properties of R/I-modules and physical features of the corresponding multiplets.
This dictionary is of course by no means complete, but it should serve to provide a quick
overview.

— Γ = O(S 0 ) for some hyperplane S 0 ⊆ Y .

S 0 is a complete intersection of linear equations. The resulting multiplet is an

exterior algebra ∧• S 0 , concentrated in homological degree zero. No differentials are
transferred to the component field level. The representation of the supersymmetry
algebra is strict. Examples include chiral superfields (S 0 = S± ), which always exist
in dimension 0 (mod 4), and free superfields (S 0 = {0}), in any dimension and
with any amount of supersymmetry. We emphasize that the free superfield always
corresponds to the canonical augmentation of the graded ring R/I.

— Γ = R/I is a complete intersection of quadratic equations.

The Koszul homology is an exterior algebra generated by the elements λf µ θ in ho-

mological degree one. The resulting multiplet can be identified with the de Rham
64 The pure spinor superfield formalism

complex Ω• (Rd ) on spacetime. The transferred differential acts as the de Rham dif-
ferential on the component fields; as such, translations act homotopically trivially.
Tensoring with a Lie (d − 3)-algebra, one obtains the BV complex of higher Chern–
Simons theory. Odd elements in the supersymmetry algebra act by zero. Examples
include the structure sheaves of the three-dimensional N = 8 and four-dimensional
N = 4 supersymmetry algebras; see [Ced08] and [Ced18a], respectively. This sheaf
is used, together with another equivariant sheaf, in the construction of the pure
spinor resolution of the Bagger–Lambert–Gustavsson model in [Ced08].

— Γ = R/I is a Gorenstein ring, but not a complete intersection.

The resulting component multiplet is equipped with a local pairing, inherited from
the perfect pairing on Koszul homology. For appropriate values of the spacetime
dimension and the codimension of Y , this local pairing defines an odd symplectic
structure, which may be used to construct a BV datum on the multiplet. The
underlying cochain complex always starts with

d
Ω0 −→ Ω1 −→ . . . ; (2.206)

as such, it always contains at least a one gauge field. By duality, the multiplet ends
with the corresponding antifields,

d
· · · −→ Ωd−1 −→ Ωd . (2.207)

Examples include ten-dimensional super Yang–Mills theory and eleven-dimensional

supergravity [Ced10c; Ced10a]; see also [SW21; EH23] for treatments using a lan-
guage close to this work.

— Γ is Cohen–Macaulay, but not Gorenstein.

The resulting multiplet will not carry a pairing and thus cannot be equipped with
a nondegenerate BV datum. We can interpret the multiplet as a BRST multiplet
and obtain the corresponding antifield multiplet from the dualizing module. Here,
the structure sheaf of six-dimensional N = (1, 0) supersymmetry is an example.
To understand theories of physical interest, though, it may be necessary to con-
sider degenerate pairings (six-dimensional N = (2, 0) supersymmetry and type IIB
supergravity are examples).

— Γ is not Cohen–Macaulay.

The resulting multiplet usually looks like a BRST multiplet. However, there is
really only a dualizing complex instead of a dualizing module. As such, we cannot
obtain the antifield multiplet from a single (non-dg) OY -module via the pure spinor
The pure spinor superfield formalism 65

superfield formalism. An example is the gauge multiplet in four-dimensional N = 1

supersymmetry, as discussed above. It would be interesting to consider extending
the formalism to dg sheaves on Y .

— Γ is a Golod ring.

A ring is Golod if and only if all Massey products on Koszul homology van-
ish [Fra18]. Recall that, if Γ is assumed to be a ring, the tensor product A• (Γ) ⊗ h
carries an L∞ structure. Transferring the L∞ structure to the component fields
and then compactifying to a point yields an L∞ structure which is given by the A∞
structure on Koszul homology tensored with the Lie algebra h. The Golod prop-
erty of Γ implies that this L∞ structure is strict. For example, the presence of the
three-ary product in ten-dimensional super Yang–Mills theory, which after com-
pactification to a point gives rise to a corresponding product in the IKKT matrix
model [MS04], witnesses the fact that the ring of functions of the ten-dimensional
N = 1 nilpotence variety is not Golod.

2.5 Ten-dimensional super Yang–Mills theory

In this section, we present a detailed analysis of ten-dimensional super Yang–Mills theory

in the pure spinor superfield formalism. This is the initial example which sparked interest
in the formalism [Ber01; CNT02] and was also analyzed in [Ale+07]. As we will see,
the canonical multiplet for N = 1 supersymmetry in ten dimensions (which we obtain
from the structure sheaf OY ) can be naturally equipped with the full structure of a
perturbative interacting BV theory within the pure spinor superfield formalism.

2.5.1 Field content and representatives

We will denote the two 16-dimensional spin representations of Spin(10) by S+ and S− .

The vector representation is, as always, denoted by V . The defining ideal of the nilpo-
tence variety reads
I = (λγ µ λ) . (2.208)

One finds for the minimal free resolution of R/I the following Betti numbers.
 
1 − − − − −
 
− 10 16 − − −
(2.209)
 
 
− − − 16 10 −
 
− − − − − 1
 66 The pure spinor superfield formalism

More concretely, the minimal free resolution of R/I in R-modules takes the form
 
• (dL )1 (dL )2 (dL )3 (dL )4 (dL )5
L = R ⊗ C ←−−− V ←−−− S+ ←−−− S− ←−−− V ←−−− C . (2.210)

The resolution differential can be described explicitly. Let us choose a basis eµ of V and
sα of S+ . The corresponding dual basis of (S+ )∨ = S− is denoted by sα .

(dL )1 : V −→ C , A 7→ (λγ µ λ)Aµ

(dL )2 : S+ −→ V , χ 7→ (λγ µ χ)eµ
(dL )3 : S− −→ S+ , χ∨ 7→ (λγ µ λ)(χ∨ γµ s) − 2(χ∨ λ)(λs) (2.211)
(dL )4 : V −→ S− , A∨ 7→ (λγ µ s)A∨µ
(dL )4 : C −→ V , c∨ 7→ µ ∨
(λγ λ)c eµ

We can perform our procedure to find the representatives. For example, starting with
the gauge field,
(dL )1 D†
A 7−−−→ (λγ µ λ)Aµ 7−−→
0
(λγ µ θ)Aµ , (2.212)

so that the elements (λγ µ θ)Aµ represent the one-form in D0 -cohomology. For the gaugino
we obtain

(dL )2 D† (dL )1 D†
χ 7−−−→ (γ µ λ)α χα eµ 7−−→
0
(γ µ θ)α χα eµ 7−−−→ (λγµ λ)(γ µ θ)α χα 7−−→
0
(λγµ θ)(γ µ θ)α χα .
(2.213)
This means that the gaugino is represented by (λγµ θ)(γ µ θ)α χα in D0 -cohomology. This
procedure can also be applied to the antifields

(dL )3 ◦D† (dL )2 ◦D† (dL )1 ◦D†

χ∨ 7−−−−−−→
0
(λγ µ θ)(χ∨ γµ s) 7−−−−−−→
0
(λγ µ θ)(γ ν θ)α (γµ χ∨ )α eν 7−−−−−−→
0
(λγ µ θ)(λγ ν θ)(γν θ)α (γµ χ∨ )α
(2.214)
We can simplify the last term to find

(λγ µ θ)(λγ ν θ)(γν θ)α (γµ χ∨ )α = (λγ µ θ)(λγ ν θ)(γµν θ)α χ∨

α, (2.215)

where γµν = γ[µ γν] denotes the antisymmetrized product of two gamma matrices. Sim-
ilarly one can track down a representative for the antifield of the one-form field. The
result is
(λγ ρ θ)(λγ ν θ)(θγµνρ θ)A+µ . (2.216)

Finally, the antighost can be represented by

(λγ µ θ)(λγ ν θ)(λγ ρ θ)(θγµνρ θ)c∨ . (2.217)

The pure spinor superfield formalism 67

These representatives were already listed in [MSX14]. Let us summarize the results in
the following table.

Field Representative in the D0 -cohomology

c c
A (λγ µ θ)Aµ
χ (λγµ θ)(χγ µ θ)
χ∨ (λγ µ θ)(λγ ν θ)(γµν θχ∨ )
A∨ (λγ ρ θ)(λγ ν θ)(θγµνρ θ)A+µ
c∨ (λγ µ θ)(λγ ν θ)(λγ ρ θ)(θγµνρ θ)c∨

Table 2.4: Representatives for the d = 10, N = 1 vector multiplet organized by

θ-degree.

2.5.2 The differential

The first order part of the transferred differential is given by

D10 = p ◦ (λγ µ θ)∂µ ◦ i . (2.218)

We immediately see that D10 acts on the ghost as the de Rham differential.

Furthermore, the differential, D10 acts on the gaugino as the Dirac operator,

/
χ 7→ ∂χ (2.219)

encoding the field equation for the gaugino.

Interestingly, this multiplet contains a second order contribution to the differential aris-
ing from homotopy transfer. As we will see momentarily, this encodes the equation of
motion for the gauge field. Recall that the second order contribution to the transferred
differential D0 is given by
D20 = p ◦ (D1 ◦ h ◦ D1 ) ◦ i . (2.220)

To apply D20 to the gauge field we need to know how the homotopy h acts on expressions
of the form
(λγ µ θ)(λγ ν θ) . (2.221)

Note that the naive guess h ◦ π = π ◦ D0† does not work in this case since

D0† (λγ µ θ)(λγ ν θ) = 0 (2.222)

68 The pure spinor superfield formalism

by the symmetry of the bracket. However, the result is easily found by a representation
theoretic argument. As h acts as a scalar, we are looking for a representative inside

∧2 V ⊂ ∧3 S+ ⊗ S+ . (2.223)

It is easy to check that there is only one such summand in the decomposition of the right
hand-side into irreducibles. This representation is spanned by the elements

(λγ ρ θ)(θγµνρ θ) . (2.224)

We set,
h ((λγ µ θ)(λγ ν θ)) = (λγ ρ θ)(θγµνρ θ) . (2.225)

Equipped with this knowledge we find

D20 ((λγ µ θ)Aµ ) =p (D1 ◦ h ((λγ µ θ)(λγ ν θ)(dA)µν ))

(2.226)
=p ((λγ σ θ)(λγρ θ)(θγ µνρ θ)∂σ (dA)µν ) .

Projection to the cohomology gives

(λγν θ)(λγρ θ)(θγ µνρ θ)∂ σ (dA)σµ . (2.227)

This shows that the transferred differential D20 acts via

A 7→ ?d ? dA. (2.228)

The differentials appearing in the multiplet can be summarized by the following diagram.

Ω0 (R10 )
d

Ω1 (R10 ) Γ(R10 , S+ )
∂/
?d?d

Γ(R10 , S− ) Ω1 (R10 )
?d?

Ω0 (R10 )

(2.229)
The pure spinor superfield formalism 69

2.5.3 The supersymmetry action

We can read off the non-derivative supersymmetry transformations directly from the
resolution differential.

δc =(γ µ )Aµ
δAµ =γµ χ
δχ =(γ µ )χ∨ γµ − 2(χ∨ )
(2.230)
δχ∨ =γ µ A∨
µ

δA∨
µ =(γµ )c
∨

δc∨ =0

Note that there are two types of closure terms present. For the gauge field, there are
again transformation witnessing that the supersymmetry algebra is represented only up
to gauge transformations. We already encountered this type of transformation in our dis-
cussion of the four-dimensional gauge multiplet. In addition, there are now second order
transformations for the gaugino, signaling that the supersymmetry algebra is represented
only on-shell.

2.5.4 The L∞ structure

We can define a dgs Lie algebra structure by tensoring A• (R/I) with a Lie algebra h.
Homotopy transfer gives rise to an L∞ structure on the component field multiplet. As we
will see, this L∞ structure, together with the pairing, equips the ten-dimensional super
Yang–Mills multiplet with the usual structure as an interacting BV theory.

The binary bracket µ02 is given by

i0

µ02 = p0 .

i0 (2.231)

Expressing this in terms of the unprimed homotopy data, there will be obviously a
diagram of the form
i

p .

i (2.232)
70 The pure spinor superfield formalism

As we already explored in the case of the four-dimensional N = 1 vector multiplet, this

diagram encodes the structure of gauge transformations on the component fields. In
particular it yields brackets

µ02 : Ω0 × Ω0 −→ Ω0 µ02 (c, c) = [c, c]

µ02 : Ω0 × Ω1 −→ Ω1 µ02 (c, A) = [c, A] (2.233)
µ02 : Ω0 × Γ(X, S+ ) −→ Γ(X, S+ ) µ02 (c, ψ) = [c, ψ] .

Furthermore, considering degree bounds, we see that only two more diagrams can con-
tribute, namely

i i
h
p and h p .

i i (2.234)

Here we marked the unary vertices with a dot, signaling the application of D1 .

From the first type of diagram we obtain

p ((λγ σ θ)h((λγ µ θ)(λγ ν θ))) [Aσ , ∂µ Aν ] = p ((λγ σ θ)(λγρ θ)(θγ µνρ θ)) [Aσ , ∂µ Aν ] (2.235)

Using the antisymmetry in µ and ν and projecting onto D0 -cohomology this gives

(λγν θ)(λγρ θ)(θγ µνρ θ) [Aσ , (dA)µσ ] . (2.236)

The second diagram gives a contribution of the form

p ((λγ σ θ)∂σ h ((λγ µ θ)(λγ ν θ)) [Aµ , Aν ]) = p ((λγ σ θ)(λγρ θ)(θγ µνρ θ)) ∂σ [Aµ , Aν ]. (2.237)

Projection to the cohomology gives

(λγν θ)(λγρ θ)(θγ µνρ θ) ∂ σ [Aµ , Aσ ] . (2.238)

Together this gives a transferred binary product

µ02 : Ω1 × Ω1 −→ Ω1 µ02 (A, A)µ = [Aσ , (dA)µσ ] + ∂ σ [Aµ , Aσ ] . (2.239)

The pure spinor superfield formalism 71

By degree reasons, there are no D1 insertions allowed for µ03 . Hence, the only contributing
diagram is of the following form.

i
h
µ03 = i p

i (2.240)

This diagram gives a contribution of the form

p ((λγρ θ)(θγ µνρ θ)(λγ σ θ)[Aσ , [Aµ , Aν ]]) = (λγρ θ)(λγν θ)(θγ µνρ θ) [Aσ , [Aµ , Aσ ]] . (2.241)

This gives a product

µ03 : Ω1 × Ω1 × Ω1 −→ Ω1 µ03 (A, A, A)µ = [Aσ , [Aµ , Aσ ]] . (2.242)

Thus, we see that the transferred L∞ structure equips the multiplet with the usual
interactions as expected for ten-dimensional super Yang–Mills theory.

2.5.5 The pairing

The ring R/I is Gorenstein, which implies that the minimal free resolution, and hence
the component field formulation of the multiplet, is equipped with a local (in the sense
of Definition 2.2.14) pairing. At the level of Koszul homology, the pairing is induced by
multiplication and projection to the subspace spanned by the top class

(λγ µ θ)(λγ ν θ)(λγ ρ θ)(θγµνρ θ) . (2.243)

This equips the component field multiplet with a BV structure.

We thus obtained the usual description of ten-dimensional super Yang–Mills theory as

an interacting BV theory solely by homotopy transfer from the pure spinor superfield
description.

2.6 A bestiary of multiplets from modules

In this final section we construct a variety of equivariant R/I-modules and examine

the structure of the associated supersymmetric multiplets. We offer some observations
72 The pure spinor superfield formalism

connecting certain of these multiplets to constructions in the physics literature, along

with some other speculations of various kinds.

2.6.1 Presentations of modules and shift symmetry

Any module Γ over any ring S can be described using a free presentation, that is an
exact sequence
ϕ
0 ←−
− Γ ←−
− F0 ←−
− F1 , (2.244)

where F0 and F1 are free S-modules. The module can then be identified as the cokernel
of the map ϕ
Γ∼
= cokerϕ = F0 /Im(ϕ) . (2.245)

As F0 and F1 are free, we can think of ϕ as a matrix with entries in S, these entries give
the relations to obtain Γ as a quotient from F0 . In fact, a free presentation is just the
start of a free resolution. By resolving kernels we can extend a free presentation to a free
resolution
ϕ0 ϕ1 ϕ2
0 ←−
− Γ ←−
− F0 ←−
− F1 ←−
− F2 ←−
− ... . (2.246)

For R = C[λ1 , . . . , λn ] it is very easy to study such maps ϕ; these just correspond to
matrices whose entries are polynomials in λ. The cokernels of such maps are then R-
modules. For the pure spinor superfield formalism, it is crucial to use R/I-modules
as this ensures that the differential D squares to zero. Suppose we have an R-module
defined by a free presentation

ϕ : Rn −→ Rk Γ = coker(ϕ) . (2.247)

The R-module Γ descends to a R/I-module if the image of ϕ contains I ×k . Thus,

we can conveniently construct R/I-modules by studying suitable maps between free R-
modules. If the map ϕ is also equivariant with respect to the action of the Lorentz and
R-symmetry groups on R, then the resulting module is also equivariant. Hence, such
equivariant maps between free R-modules precisely give rise to the desired input for
the pure spinor superfield formalism. In the physics literature this procedure was used
to construct multiplets in the pure spinor superfield formalism under the name shift
symmetry (see [Ced10c; Ced10a; CK11; Ced14]).10
10
The name “shift symmetry” arises from writing out the equivalence relation (2.245) as

f0 ≈ f0 + ϕ(f1 )

with explicit representatives f0 ∈ F0 and f1 ∈ F1 .

The pure spinor superfield formalism 73

Example 2.6.1. We can immediately give a free presentation for the quotient rings R/I
which we previously considered. The map
 
ϕ : Rd −→ R, ϕ = λγ 0 λ . . . λγ d−1 λ . (2.248)

realizes the free presentation coker(ϕ) = R/I.

2.6.2 Motivating example of a nontrivial sheaf: the six-dimensional

hypermultiplet

As an example to demonstrate this technique, let us construct the six-dimensional hyper-

multiplet. We already constructed the six-dimensional vector multiplet from the struc-
ture sheaf of the nilpotence variety in §2.4.3. Recall that for six-dimensional N = (1, 0)
supersymmetry, the odd part of the supertranslation algebra is

S+ ⊗ U , (2.249)

where S+ is the fundamental representation of su(4) and U ∼

= C2 carries the fundamental
representation of su(2). The polynomial ring R is nothing but the symmetric algebra on
S+ ⊗ U and comes with the natural action of su(4) × su(2). There is a unique equivariant
map
S+ ⊗ R −→ U ⊗ R (2.250)

which is linear in λ. Choosing a basis for U and S+ , this map is represented by

!
λ11 λ21 λ31 λ41
ϕ : S+ ⊗ R −→ U ⊗ R ϕ= . (2.251)
λ12 λ22 λ32 λ42

It is easy to check that the image of ϕ indeed contains I ×2 , thus we can consider Γ =
coker(ϕ) as an equivariant R/I-module in the pure spinor superfield formalism.

We display the Betti numbers of the minimal free resolution in the following table.
" #
2 4 − −
(2.252)
− − 4 2

The representations appearing in the minimal free resolution can be computed using
Macaulay2 via the highest weight package. The minimal free resolution of Γ in R-
modules takes the form
 
• ϕ  3 ϕT 4
L = R ⊗ U ←−− S+ ←−− ∧ S+ ←−
− U ⊗ ∧ S+ , (2.253)
74 The pure spinor superfield formalism

and is a special case of the Buchsbaum–Rim complex [BE77] (see [Eis95, Appendix A.2.6]
for a textbook presentation and a description of the differential  in terms of the 2 × 2
minors of ϕ).

Choosing a basis ei for U and sα a basis for S+ , we can write out the differentials in the
complex as:

d1 : ∧1 S+ −→ U ψ 7→ λαi ψα ei
d2 : ∧3 S+ −→ ∧1 S+ ψ + 7→ (λαi λβj ij )ψαβγ
+
sγ (2.254)
d3 : U ⊗ ∧4 S+ −→ ∧3 S+ φ+ 7→ λiα φ+ α
i s .

In the last differential we identify sα with αβγδ sβ ∧ sγ ∧ sδ along the isomorphism

∧3 S+ ∼ β
= S− . The differential (λα λ ij ) is the differential  appearing in the Buchsbaum–
i j
Rim complex.

As expected, the hypermultiplet consists of two scalars that form a doublet under su(2)
as well as fermions in S+ that are neutral under su(2) and their corresponding antifields.
The two maps are expected to encode the respective equations of motions. We are thus
dealing with an on-shell representation of the supersymmetry algebra. The multiplet can
be equipped with a pairing which yields a BV structure.

We can use the zig-zag procedure to find representatives for the fields in the multiplet.
These are expressed in terms of the basis ei of U .

Field Representative in the D0 -cohomology

φ φi e i
ψ ψα θiα ei

ψ∨ λαi θjβ θkγ εij ψαβγ

∨ ek

φ∨ λαi θjβ ij θlδ θkγ εαβγδ φ+k el

Table 2.5: Representatives for the hypermultiplet in six dimensions organized by

θ-degree.

From the resolution differential, we can easily read off the non-derivative supersymmetry
transformations.

δφi = αi ψα
δψα = βi γj εij ψαβγ
∨
(2.255)
δψα∨ = iα φ∨
i

δφ∨
i =0
The pure spinor superfield formalism 75

Again, we see the quadratic transformation involving the fermion and its antifield showing
that the supersymmetry algebra only closes up to the equations of motion.

Consequently, the equations of motions are encoded in the transferred differential D0 .

There is a first order term D10 acting on the fermion. Given the representatives, it is easy
to see that D10 acts by the Dirac operator

/ .
ψ 7→ ∂ψ (2.256)

Further, there is a second order differential D20 induced via homotopy transfer which
encodes the field equation of the scalar field and which acts via

D20 = p ◦ (D1 ◦ h ◦ D1 ) ◦ i . (2.257)

Acting on the scalar, we find

 
[i j]
D20 φ = p D1 h (λ[α θβ] ∂ [αβ] φi ei ) . (2.258)

By degree reasons, applying the homotopy h to the element in the brackets yields an
expression in θ2 . On purely representation theoretic grounds, we can see that there is a
unique (up to a non-zero prefactor) expression which comes into question, namely

(i j)
θ[α θβ] ∂ [αβ] φ(i ej) . (2.259)

As a check, we may apply the differential D0 to that representative. There we obtain

(i j)
λ[α θβ] ∂ [αβ] φ(i ej) , (2.260)

which, at first sight, does not look like the original element we started with. However,
recall that we are working in the module Γ which is the quotient R2 /Im(ϕ). In particular
this means that λi ei = 0 and hence

0 = λi θj ei φj = λ[i θj] e[i φj] + λ(i θj) e(i φj) , (2.261)

such that we indeed get back our original expression (up to a non-zero prefactor). Moving
on, we then easily find

D20 φ = (λαi θjβ ij θlδ θkγ εαβγδ ) ∂ µ ∂µ φk el , (2.262)

such that the transferred differential indeed encodes the Laplace equation.
76 The pure spinor superfield formalism

Summarizing, the multiplet has the following structure.

Ω0 ⊗ U Ω0 ⊗ S+
∂/
?d?d
(2.263)
Ω0 ⊗ S− Ω0 ⊗ U

This multiplet was defined in Equation (3.2) of [Ced18b] using shift symmetry.

2.6.3 Lie algebra cohomology

Another natural source for equivariant modules are the Lie algebra cohomology groups
of the supertranslation algebra n. This was already noted in [ESW21]. Recall that the
Chevalley–Eilenberg complex takes the form

C • (n) = Sym• (n∨ [1]) , dCE .

(2.264)

The Chevalley–Eilenberg differential is induced by the dual of the bracket, which is

extended to the whole algebra according to the Leibniz rule. For the supertranslation
algebra, the Z × Z/2 grading of the Chevalley–Eilenberg complex lifts to a Z × Z grading
by viewing the supertranslations as a graded Lie algebra as we have done above. If we
totalize this bigrading, generators in n∨ ∨
2 sit in degree −1 and generators in n1 in degree
zero. Identifying, Symq (n∨ α
1 ) = R = C[λ ] we write

C −p (n) = ∧p (n∨
2 ) ⊗ R. (2.265)

Denoting a basis on n∨ µ
2 by v , the Chevalley–Eilenberg differential acts on the generators
by

µ β
dCE v µ = λα fαβ λ
(2.266)
dCE λα = 0 .

Now two observations turn out to be crucial. First, the zeroth Chevalley–Eilenberg
cohomology is nothing else then the ring of functions of the nilpotence variety

H 0 (C • (n)) = R/I . (2.267)

Second, as the Chevalley–Eilenberg complex comes with the structure of a cdgsa, the
cohomology is equipped with a multiplication which preserves the grading. Hence, all
The pure spinor superfield formalism 77

cohomology groups are H 0 (C • (n)) = R/I-modules and can thus be used as input data
for the pure spinor superfield formalism.

The analysis of examples suggests some speculations about dualities between the mul-
tiplets associated to Chevalley–Eilenberg cohomology groups in different degrees. For
a start, it seems to be the case that the Chevalley–Eilenberg cohomology groups are
concentrated in negative degrees up to n := dim(V ) − codim(Y ). In all examples we
have checked there is an isomorphism

Ext−codim(Y ) (R/I, R) ∼
= H −n (C • (n)) . (2.268)

In addition, for the example of ten-dimensional N = 1 supersymmetry, we further observe

dualities “up to a copy of the free superfield” for the multiplets associated to H i (C • (n))
and H −n−i (C • (n)). Here, we give a first overview on these phenomena; we will formalize
some of these findings later (see in particular §7.4.2).

Three-dimensional N = 1. As a motivating example, let us consider again N = 1

supersymmetry in three dimensions. Using Macaulay2 one can compute the Chevalley–
Eilenberg cohomology. Only H 0 and H −1 are non-vanishing. The zeroth cohomology is
R/I and thus gives rise to the gauge multiplet from §2.3.9. As the length of the minimal
free resolution is two—which equals the codimension of Y —we immediately see that R/I
is Cohen–Macaulay. The first cohomology group is represented as the cokernel of the
map !
3 2 λ1 0 λ2
ϕ : R −→ R ϕ= (2.269)
0 λ2 λ1

The resulting multiplet is the antifield multiplet of the gauge multiplet.

" #
2 3 −
(2.270)
− − 1

Note that, as discussed in §2.4 we could have also obtained the antifield multiplet from
Ext−2 (R/I, R).

Four-dimensional N = 1. The Chevalley–Eilenberg cohomology is concentrated in

degrees zero, minus one and, minus two. As the zeroth cohomology is just R/I, the
corresponding multiplet is the gauge multiplet. The first cohomology group yields a
multiplet with the following Betti numbers.
" #
4 7 −
(2.271)
− − 6 4 1
 78 The pure spinor superfield formalism

Decomposing the minimal free resolution equivariantly, we find

 
• (dL )1 2 (dL )2 3 2 (dL )3 (dL )4
L = R ⊗ S+ ⊕ S− ←−−− ∧ V ⊕ C ←−−− ∧ V ⊕ C ←−−− S+ ⊕ S− ←−−− C .
(2.272)
Thus we see that this multiplet contains a two-form. It would be interesting to interpret
this as a field-strength multiplet.

The second Chevalley–Eilenberg cohomology yields two copies of the chiral multiplet.
h i
2 4 2 (2.273)

Note that this precisely matches with Ext−2 (R/I, R) as described in §2.4.4.

Ten-dimensional N = 1. Let us further study the multiplets associated to the ten-

dimensional Chevalley–Eilenberg cohomology of the ten-dimensional N = 1 supertrans-
lation algebra. These cohomology groups were already computed equivariantly in [MSX12]
The multiplet associated to the first Chevalley–Eilenberg cohomology has the following
Betti numbers.
 
16 45 − − − − − − − − − − − − − − −
 
 − 16 250 720 1874 4368 8008 11440 12870 11440 8008 4368 1820 560 120 16 1
 
− − − − 16 10 − − − − − − − − − − −
(2.274)
We notice that the graded rank (with respect to the homological degree) of the associated
vector bundle over spacetime—which, in physical terms, corresponds to the number of
degrees of freedom—is given by

− 16 − 45
+ 16 + 250 + 720 + 1874 + 4368 + 8008 + 11440 + 12870
+ 11440 + 8008 + 4368 + 1820 + 560 + 120 + 16 + 1 (2.275)
− 16 − 10
= 65792 = (215 + 215 ) + (128 + 128) .

This precisely matches the number of degrees of freedom of the supercurrent multiplet
constructed in [HNv82]. Further, recall that the free superfield just corresponds to the
exterior algebra ∧• S on 16 generators. Hence, its Betti numbers are precisely binomial
coefficients 16


i . We note that (2.274) contains precisely such coefficients, except for a
missing 1 in degree (0, 1). However, we can add a trivial pair in degrees (0, 1) and (1, 0).
Then we can subtract the respective Betti numbers of the free superfield to obtain the
The pure spinor superfield formalism 79

following table.  
16 45 + 1 − − − −
 
−
 − 130 160 154 − 
 (2.276)
− − − − 16 10

This is precisely the dual of the Betti table of H −4 (C • (n)), which is displayed in (2.280).
We remark that this “almost-duality” phenomenon is closely analogous to the structure
sheaf of 4d N = 1; it reflects the failure of the module to be Cohen–Macaulay. We
further note that the fields in the first row are a spinor, a two-form, and a scalar; it is
tempting to interpret this as a field-strength multiplet, containing the gaugino χ and the
field strength F of the gauge field, and subject to certain constraints.

The multiplet associated to H −2 (C • (n)) is the stress-energy tensor multiplet or super-

current multiplet. Its Betti table is of the following form.
" #
120 720 2130 4512 8008 11440 12870 11440 8008 4368 1820 560 120 16 1
− − − 136 160 45 − − − − − − − − −
(2.277)

The supercurrent multiplet can be constructed as

Jµνρ = tr(χγµνρ χ). (2.278)

Here γµνρ just represents the isomorphism ∧2 (S+ ) ∼

= ∧3 (V ), and χ is the spinor superfield
describing on-shell Yang-Mills theory [HU87] that corresponds to H 1 (C • (n)). Alterna-
tively it can be described as an abstract superfield satisfying the constraints [HU87]

1
Dα Jabc = (γ[a Jbc] )α + (γ[ab Jc]1 )α + (γ[abc] J 1 )α

1 , J 1 , and J 1 are three superfields in the representation [0, 1, 0, 1, 0],

where the superfields Jbcα cα α
[1, 0, 0, 0, 1] and [0, 0, 0, 1, 0]. The total dimension of the constraints is 560 + 144 + 16 =
720. The leading component of Jabc is in the ∧3 V representation [0, 0, 1, 0, 0] of dimension
120.

Again, introducing trivial pairs and subtracting precisely yields the dual of the Betti
table of H −3 (C • (n)).

 
45 160 136 − − − −
 
 − − 144 310 160 − −
  (2.279)
− − − − − 16 1
80 The pure spinor superfield formalism

The Betti table associated to H −4 (C • (n)) takes the following form.

 
10 16 − − − −
 
 − 54 160 130 − − 
  (2.280)
− − − − 46 16

Finally, H −5 (C • (n)) ∼
= R/I again yields the vector multiplet. Note that Y is Gorenstein
and of codimension five, such that Ext−5 (R/I, R) ∼ = R/I.

2.6.4 Six-dimensional multiplets from line bundles

For six-dimensional N = 1 supersymmetry, the nilpotence variety can be identified with

P1 × P3 using the Segre embedding, hence there is an interesting family of multiplets
associated to line bundles. We will study this family in detail in §4 in detail; here we
just give a short preview.

Line bundles on Pn are classified by a single integer j ∈ Z and are denoted by O(j).
Using the projections
π3
P1 × P3 P3
π1 (2.281)

P1
we can define a family of line bundles

O(i, j) := π1∗ O(i) ⊗ π3∗ O(j) (2.282)

on P1 × P3 . This family has been investigated in the physics literature [KNT18a].

Let us here list the corresponding multiplets for some integers i and j. Clearly O(0, 0) is
just the structure sheaf of the nilpotence variety and hence the corresponding multiplet
is the vector multiplet. O(1, 0) is the hypermultiplet, which we studied above. O(2, 0)
is the antifield multiplet of the vector.

For O(3, 0) a multiplet with the following Betti numbers arises.

h i
4 12 12 4 (2.283)

The minimal free resolution of the module in R-modules takes the form
 
(dL )1 (dL )2 (dL )3
L• = R ⊗ C4 ←−−− C3 ⊗ S+ ←−−− C2 ⊗ ∧2 S+ ←−−− C1 ⊗ ∧3 S+ , (2.284)
The pure spinor superfield formalism 81

The multiplet for O(4, 0) is a building block in the construction of the “relaxed hyper-
multiplet” [HST83].
h i
5 16 18 8 1 (2.285)

The minimal free resolution of the module in R-modules takes the form
 
• 5 (dL )1 4 (dL )2 3 2 (dL )3 2 3 (dL )4 1 4
L = R ⊗ C ←−−− C ⊗ S+ ←−−− C ⊗ ∧ S+ ←−−− C ⊗ ∧ S+ ←−−− C ⊗ ∧ S+ ,
(2.286)
The minimal free resolutions are “twisted Lascoux” complexes which are described with
their differentials in [DS14].

2.6.5 Conormal modules

Denoting the defining ideal of the nilpotence variety by I, the conormal module is defined
as the quotient I/I 2 . This gives another interesting module to consider as an input for
the pure spinor superfield formalism. The resulting multiplets seem to often correspond
to supergravity theories. We demonstrate this in low dimensions.

Three-dimensional N = 1. The resulting multiplet has the following Betti numbers.

" #
3 2 −
(2.287)
− 5 4

Investigating the Hilbert series, we find that all occuring representations are irreducible
representations of the spin group Spin(3) ∼= SU (2). Thus, the first line contains a vector
and a spinor, while the second line can be identified with a symmetric traceless tensor
and the four-dimensional part of the decomposition

S⊗V ∼
= [1] ⊕ [3] . (2.288)

Four-dimensional N = 1. In four dimensions the conormal module yields a multiplet

with the following Betti numbers.
" #
4 4 1 −
(2.289)
− 9 12 4

Investigating the Hilbert series we find that the representations in the first line are a
vector, a Dirac spinor and a scalar. The nine-dimensional representation in the second
line is a symmetric traceless tensor. The twelve-dimensional representation has Dynkin
82 The pure spinor superfield formalism

labels [2, 1] ⊕ [1, 2]. Thus, the multiplet consists of one spin-2, two spin- 23 and a single
spin-1 field. In terms of Dynkin labels, the multiplet takes the following form.

[1, 1] [1, 0] ⊕ [0, 1] [0, 0]

(2.290)
[2, 2] [2, 1] ⊕ [1, 2] [1, 1]

Ten-dimensional N = 1. In this case, by a pleasing coincidence, the conormal module

coincides with the module H −4 (C • (n)) constructed above. The resolution was studied
in [Kuz18, Corollary 4.4].

2.6.6 Dimensional reduction and restriction to strata: the 4d N = 2

tensor multiplet

There are interesting relations between the nilpotence varieties of supersymmetry alge-
bras in different dimensions, for instance the nilpotence variety of a higher dimensional
supersymmetry algebra may sit inside the nilpotence variety of a lower dimensional
one. The resulting multiplets will then be related by dimensional reduction. We il-
lustrate this by considering the relation between six-dimensional N = (1, 0) and four-
dimensional N = 2 supersymmetry. Recall that we described the nilpotence variety for
six-dimensional N = (1, 0) supersymmetry by the 2 × 2-minors of a 2 × 4-matrix with en-
tries λαi . As explained in [ESW21] one obtains the nilpotence variety for four-dimensional
N = 2 supersymmetry by replacing

λαi −→ (λβi , λ̄β̇i ) , (2.291)

and throwing away the two minors which do not mix the different chiralities. Hence,
there is an inclusion
i : Y (6; 1, 0) ,→ Y (4; 2) , (2.292)

whose image we denote by Y0 . In fact the global structure of Y (4; 2) is easily described.
It consists of three strata; in addition to Y0 there are two copies of (S± ⊗ U ) ∼ = C4
corresponding to solutions where λ = 0 or λ̄ = 0 respectively:

Y (4; 2) = Y0 ∪ Y1 ∪ Y2 ∼
= Y (6; 1, 0) ∪ (S+ ⊗ U ) ∪ (S− ⊗ U ) . (2.293)

Pushing forward the structure sheaf OY (6;1,0) along i we thus obtain OY0 . As we already
discussed at multiple occasions, the structure sheaf OY (6;1,0) produces the vector multi-
plet. Clearly, considering OY0 in the pure spinor superfield formalism gives a multiplet
The pure spinor superfield formalism 83

with the same Betti numbers; only the weights have to be adapted to four dimensions.
Resolving OY0 equivariantly, we see that the six-dimensional vector splits up into a
four-dimensional vector and two scalars. The fermion gives two Dirac fermions in four
dimensions and the scalars remain scalars. Hence, the resulting multiplet is precisely
the N = 2 vector multiplet in four-dimensions as one can obtain it from dimensional
reduction. A similar phenomenon holds in general: given a multiplet in dimension d,
we can push the corresponding sheaf forward along the dimensional reduction map to
obtain the dimensionally reduced multiplet.

Interestingly, considering OY (4;2) as an input in the pure spinor superfield machinery

gives a multiplet with the following Betti numbers.
 
− − − −
1
 
− 4 − − −
  (2.294)
− − 9 8 2

Working equivariantly, the minimal free resolution gives

 
• (dL )1 (dL )2 2 3 (dL )3 (dL )4
L = R⊗ C ←−−− V ←−−− ∧ V ⊕ C ←−−− (S+ ⊗ U ) ⊕ (S− ⊗ U ) ←−−− C2 ⊕ C−2 ,
(2.295)
where C3 carries the adjoint representation of SU (2)R and has U (1)R -charge 0 while
the two scalars in the top degree have U (1)R -charges +2 and −2 as indicated by the
subscript. This is the field content of a tensor multiplet as described in [WS06; Jur+19].

Of course we can also restrict to the other strata. The minimal free resolutions are
then exterior algebras ∧• (S± ⊗ U ), the resulting multiplets are thus chiral multiplets as
described in [dRo+80].

2.A Homotopy transfer for L∞ modules

Let (L, µ̃k ) be a (super) L∞ algebra and (V, dV , ρ(j) ) an L∞ module for L. As was
explained in [Lad04], the L∞ module structure gives rise to an L∞ structure on L ⊕ V .
Explicitly we can define (setting ρ(0) = dV )

k
!
X
µk ((x1 , v1 ), . . . , (xk , vk )) = µ̃k (x1 , . . . , xk ), ±ρ(k−1) (x1 , . . . , x̂i , . . . , xk )vi .
i=1
(2.296)
84 The pure spinor superfield formalism

For example, if (L, [., .]) is a (super) Lie algebra and ρ is a strict action, we find

µ1 ((x, v)) = (0, dV v)

(2.297)
µ2 ((x1 , v1 ), (x2 , v2 )) = ([x1 , x2 ] , ρ(x1 )v2 − ρ(x2 )v1 )

All higher order operations vanish. Now suppose we have homotopy data

p
h (V, dV ) (W, dW ) (2.298)
i

and want to transfer an L∞ module structure on V to a new L∞ module structure on

W . The fact that these L∞ module structures can be thought of as L∞ structures on
L ⊕ V and L ⊕ W suggests to extend the above homotopy data to

id⊕p
id⊕h (L ⊕ V, dV ) (L ⊕ W, dW ) (2.299)
id⊕i

and then to use the usual homotopy transfer for L∞ structures. Let us denote the
transferred L∞ structure on L ⊕ W by µ0k . We can read off the transferred module
action ρ0(k) as follows. Let
π : L ⊕ V −→ V (2.300)

be the obvious projection. Then (2.296) implies

ρ0(k) (x1 , . . . , xk )w = π µ0k+1 ((x1 , 0), . . . , (xk , 0), (xk+1 , w)) .

(2.301)

As usual, the transferred L∞ structure µ0k can be calculated by sum over trees formulas.
Using this, one can also derive sum over tree formulas for the induced action ρ0 . For our
purposes we are only interested in the case where L = g is a super Lie algebra and ρ
is a strict action. As explained above, this means that (g ⊕ V, µk ) is a dgs Lie algebra.
In this case the L∞ structure on g ⊕ W is computed by the sum over all rooted binary
trees by decorating each leaf with the inclusion i, each internal line with the homotopy
h, and the root by the projection p. A vertex means the application of the product µ2 .
In the case of the binary product one writes:

µ02 = p .

i (2.302)
The pure spinor superfield formalism 85

In formulas this means

µ02 ((x1 , w1 ), (x2 , w2 )) = ([x1 , x2 ] , p(ρ(x1 )i(w2 ) ± ρ(x2 )i(w1 ))) . (2.303)

Accordingly we find for the L∞ module action ρ0

ρ0(1) = p ◦ ρ ◦ i . (2.304)

In the case of µ03 we can write

i i i
h h
µ03 = i p ± i h p ± i p .

i i i
(2.305)
This gives for ρ0(2)

ρ0(2) (x1 , x2 ) = p ◦ (ρ(x1 )hρ(x2 ) ± ρ(x2 )hρ(x1 )) ◦ i . (2.306)

In this manner we can also obtain a general sum over trees representation for ρ0(k) in
terms of ρ. Using equations (2.301) and (2.296) we see that ρ0(k) can be obtained from
binary rooted trees with k + 1 leaves by the following rules. Label the first k leaves by
elements x1 , . . . , xk and the last one by the inclusion i. Keep only those trees where
there are no vertices connecting two elements of g. As usual, each internal line carries
the homotopy h and the root is decorated by p. A vertex now means “apply ρ(xi )”. For
example we can write for ρ0(2) :

x1 x1
h
ρ0(2) (x1 , x2 ) = x2 h p ± x2 p .

i i (2.307)

Clearly this recovers (2.306).

Chapter 3

Derived pure spinor superfields

3.1 Introduction

As established in the previous chapter, the pure spinor superfield formalism constructs
a supersymmetric multiplet out of the datum of a graded module over the ring of func-
tions on the nilpotence variety of a supertranslation algebra. It is natural to ask for
a characterization of all multiplets which arise in this manner. In §3.5.2, we already
described an example of a multiplet which cannot be constructed using the technique.1
There, we linked the failure to the geometry of the underlying nilpotence variety: it is not
Cohen–Macaulay, so that the dualizing complex of its ring of functions has cohomology
in multiple degrees and is not quasi-isomorphic to a single homogeneous module. At the
level of multiplets, the structure sheaf gives rise to the four-dimensional N = 1 vector
multiplet. On general grounds, one expects that the dualizing module gives rise to the
dual (or antifield) multiplet. However, due to the failure of the Cohen–Macaulay prop-
erty, this does not work on the nose. Instead, the appearance of the dualizing complex
suggests that a generalization of the pure spinor superfield formalism to the world of
derived algebraic geometry is necessary.

In this chapter, we tackle these questions systematically, and show that a derived gener-
alization of the pure spinor formalism can be used to produce every supermultiplet in a
very general setting.

As before, let g be a super Lie algebra of super Poincaré type with supertranslation
subalgebra n. Building on the terminology of the previous chapter, we will introduce the
dg-category of g-multiplets which we denote by Multg . The main result of this chapter
(Theorem 3.4.3), stated somewhat informally, establishes an equivalence of categories
1
The obstruction to constructing this multiplet using the pure spinor formalism was previously ob-
served by Martin Cederwall.

87
88 Derived pure spinor superfields

between between g-multiplets and g0 -equivariant modules over the Chevalley–Eilenberg

algebra C • (n).
Multg  ModgC0• (n) (3.1)

We will view the functor from left to right as a natural derived enhancement of the pure
spinor construction. Indeed, taking Lie algebra cochains of (2.58), we obtain an exact
sequence
C → C • (Πn1 ) → C • (n) → C • (n2 ) → C (3.2)

of bigraded cdgas, witnessing C • (n) as an R-algebra, where R := C • (Πn1 ) ∼

= Sym• (n1 ) is
the free commutative algebra on n∨ •
1 . If we totalize the bigrading, C (n) is concentrated
in non-positive degrees and its degree-zero cohomology is then R/I, the ring of functions
on the nilpotence variety of g. Furthermore, there is a natural equivariant map

C • (n) → R/I, (3.3)

so that any R/I-module is a C • (n)-module in a natural way. When applied to modules

of this sort, our derived enhancement agrees on the nose with the standard pure spinor
formalism.

From this point of view, we can view C • (n) as a derived enhancement of the ring of
functions on the nilpotence variety: its spectrum is an affine derived scheme whose
underlying classical affine scheme is exactly the affine nilpotence variety, but whose
derived data remembers the higher cohomology of n with respect to the totalized grading.

Remark 3.1.1. While we will not need any technology from the theory of derived algebraic
geometry in this work, we refer the interested reader to the article [Toë14] of Toën for a
survey.

The functor in (3.1) from ModgC0• (n) to multiplets is defined by extending the construction
in §2. It is closely connected to more standard versions of Koszul duality in at least two
ways. First of all, it can be thought of as arising from the dual pair (C • (n), U (n)). The
standard Koszul duality functor is generated by the kernel

(K0 , D0 ) = (C • (n) ⊗ U (n) , X i ⊗ ni ), (3.4)

where {X i } is a basis for n∨ , therefore a set of generators for C • (n), and ni denotes the
corresponding basis of n (thus set of generators for U (n)).

If we now view C ∞ (N ) as a right module for U (n), where n acts by right-invariant vector
fields—in particular, ni by the vector field Di := ρ(ni )—we can modify this kernel to
Derived pure spinor superfields 89

obtain a (C • (n), C ∞ (N ))-bimodule of the form

(K, D) = (C • (n) ⊗ C ∞ (N ) , X i ⊗ Di ). (3.5)

Our functor is defined as the integral transform associated to this kernel. From this
perspective, we can interpret the functor heuristically in two steps:

1. First form the Koszul dual U (n)-module of a module over the Chevalley–Eilenberg
complex.

2. Then apply the associated bundle construction to obtain a dg-vector bundle over
the supergroup N , with a residual right N -action. In particular, we can forget
down to a dg-vector bundle on the even part V of N , with a geometric action of
the supersymmetry algebra n.

It is another version of Koszul duality which is perhaps most closely connected. Kapra-
nov [Kap91] established a version of Koszul duality that provides a Quillen equivalence
between the model categories of D-modules on a space and Ω• -modules over the same
space. Our construction can be understood as a version of this equivalence in the case
where that space is the nilpotent super Lie group N ; it relates translation-invariant nat-
ural vector bundles to modules over the translation-invariant differential forms, which
are precisely the Lie algebra cochains. The connection to the notion of “multiplet” in the
physics literature has, as far as we know, not been appreciated before.

While Koszul duality provides a natural language in which to understand the pure spinor
construction, our proof of Theorem 3.4.3, and all of our discussion of examples, will
proceed by direct computation using the kernel (K, D). This is meant to emphasize
that the technique provides not just an abstract equivalence, but a set of efficient and
practical computational techniques. We also emphasize that the version of Koszul duality
encapsulated in the pure spinor formalism naturally produces dg vector bundles whose
sections are sheaves on the site of manifolds equipped with appropriate structure: in the
case of standard super Poincaré algebras, this means that one knows how to place the
resulting multiplet on any Riemannian manifold. We see this as being connected to ideas
in Cartan geometry for the model space N ∼ = G/G0 , and hope to return to versions of
the formalism on non-flat spacetimes (for nontrivial or more general Cartan geometries)
in future work.

Finally, it is worth remarking that the functor from right to left in (3.1) is also easy to
understand: it is just taking the derived n-invariants of a multiplet. This is intuitively
satisfying in various respects. For example, the twist of a multiplet consists of the derived
invariants of some abelian odd subalgebra of n, whereas its dimensional reduction consists
90 Derived pure spinor superfields

of the invariants of an abelian even subalgebra of n. The fact that a multiplet can be
recovered from the datum of its derived n-invariants thus says, in a sense, that it is
equivalent to all of its possible twists, considered as a natural family over the derived
classifying space Bn.

The upshot of this result is that, after an appropriate derived upgrade, every supermul-
tiplet has a pure spinor superfield description. We can use the equivalence of categories
to shed some light on the motivating questions concerning the underived pure spinor
formalism discussed above. For example, the following is an immediate consequence:

Theorem (Corollary 3.4.6). A given multiplet (E, D, ρ) lies in the image of the under-
ived pure spinor formalism if and only if the Chevalley–Eilenberg cohomology H • (n, E)
is concentrated in a single degree.

We illustrate the derived formalism with some applications and examples; we also show
that the derived formalism gives rise to a construction of certain maps relating the
multiplets associated to the different Chevalley–Eilenberg cohomology groups of the su-
pertranslation algebra, associated to a filtration on C • (n). We give explicit calculations
for examples with minimal supersymmetry in dimensions three and four.

Further directions

The realization of the pure spinor formalism as an equivalence of dg-categories offers

potential insight to numerous more involved applications. We give a few indications of
possible directions here:

1. Given any multiplet M = (E, D, ρ), associated to any supersymmetry algebra g

with associated supertranslation algebra n, it is possible to realize a C • (n)-module
Γ generating M via the pure spinor formalism: one can simply set Γ = C • (n, E).
These modules have cohomology consisting of a graded sum of finitely generated
modules over the ring of functions on the nilpotence variety, together with ad-
ditional information encoding the action of the generators of C • (n) in nonzero
total degree. By the results of [SW21], they efficiently and fully encode the in-
formation of the various twists of the original multiplet M : one can compute the
stalk of the module at a classical point Q : C • (n, E) → C, obtaining descriptions
of different twists for different orbits under the action of Spin(n) and the group of
R-symmetries. We refer to the discussion in [ES19b; ESW20; ESW21] for details
of this classification.
It would, for example, be interesting to explicitly understand the modules associ-
ated to multiplets such as the N = (1, 0) tensor multiplet in six dimensions.
Derived pure spinor superfields 91

2. One can use the Batalin–Vilkovisky (BV) formalism to construct interacting su-
persymmetric classical field theories using the pure spinor formalism. In the BV
formalism, an interacting supersymmetric field theory is encoded by a cyclic L∞
structure on a multiplet M , that is, an L∞ algebra structure together with a pairing
of degree −3. This additional data can be carried along the pure spinor functor;
it is enough to define a Lie algebra structure and shifted symplectic pairing in-
ternal to equivariant C • (n)-modules. This will require a bit more input; it is not
straightforward to define a natural tensor product on the category of multiplets, so
one would need to work instead with the D-modules obtained by taking the sheaf
of sections of the multiplet with its natural tensor product. One would then aim
use the equivalence to establish a “convolution” monoidal structure ∗ making the
functor monoidal.

For example, as computed in [CNT02] and discussed in [SW21, §7.3] and [EH23], if
n is the eleven-dimensional N = 1 supertranslation algebra, the eleven-dimensional
supergravity multiplet arises from the zeroth cohomology of C • (n), placed in de-
gree −3 with respect to the natural weight grading, as a C • (n)-module. In order
to obtain an interaction, it would suffice to define, up to homotopy, a Lie struc-
ture on this module. It would be interesting to understand the action functionals
of [Ced10c; Ced10a] in this language, and to use them to connect to component
actions for perturbative supergravity, either twisted or untwisted. An interacting
BV theory conjecturally describing the minimal twist of eleven-dimensional super-
gravity was studied in [RSW23]. We come back to this question in §7.

3. There are several sheaves over the (derived) nilpotence variety that automatically
carry Lie structures. For example, if h is a finite-dimensional semisimple Lie al-
gebra, one can form the tensor product H 0 (n) ⊗ h.2 It has been known for some
time that this Lie algebra describes interacting ten-dimensional super Yang–Mills
theory in the BV formalism; the correct gauge algebras for lower-dimensional super
Yang–Mills theories are also obtained in this fashion.

In dimensions higher than eleven we expect C • (n) to have no higher cohomology,

so Spec(C • (n)) will be purely classical. Equivalently, the nilpotence variety is ex-
pected to be a complete intersection according to Hartshorne’s conjecture: roughly
speaking, the nilpotence variety is determined by a system of n equations, while
the number of variables is of order 2n/2 . Hartshorne’s conjecture states that any
smooth projective variety in Pn with codimension < n/3 is a complete intersection.
2
When we discuss Lie algebra cohomology H • (n), we will view n as being a Z-graded object with
respect to the weights of the natural rescaling action. As such, the complex C • (n) is naturally bigraded;
we refer here to the zeroth cohomology with respect to the total grading. We will discuss these degree
conditions further in §3.3.1.
92 Derived pure spinor superfields

The codimension condition applies to nilpotence varieties with minimal supersym-

metry in dimension ≥ 12, although these varieties are not smooth (but see for
instance the recent article [ESS21] for version of this result that would be appli-
cable to the example of nilpotence varieties in high dimensions). However, it is
often possible to obtain non-trivial cohomology by taking—rather than the entire
Chevalley–Eilenberg complex—a non-trivial quotient associated to an orbit closure
within the nilpotence variety.
To give a further example, the tangent sheaf Der(C • (n)) carries a Lie bracket for
all supertranslation algebras n. While the associated multiplet is not generally
associated to a BV theory—it does not typically carry a −1-shifted symplectic
pairing—one can always build a BV theory by applying the cotangent theory con-
struction and considering the multiplet M ⊗ M ! [−1], where M ! = M ∨ ⊗ Ωtop (V )
is the density-valued dual. This procedure allows for the construction of a very
general family of interacting classical supersymmetric field theories.

4. The equivalence of categories allows for a program to classify families of multiplets

starting from the algebraic geometry of the (derived) nilpotence variety. We explore
this direction in §4, where (among other things) a description of all six-dimensional
N = (1, 0) multiplets whose derived invariants form a single line bundle on the
projective nilpotence variety is given.

3.2 The category of multiplets

In §2, we introduced the notion of a g-multiplet for a super Lie algebra g, now we move
on to define a category of multiplets Multg . To do so, we proceed in two steps: First,
we discuss strict multiplets with strict morphisms, then we move on to the homotopy-
theoretic generalization.

3.2.1 Strict multiplets

As before, for a dgs vector bundle (E, D) over a base manifold X, we denote the space of
global smooth sections by E = Γ(X, E). The endomorphisms End(E) form a dgs Lie al-
gebra, where the bracket is given by the commutator and the differential is [D, −]. Inside
(End(E), [D, −]) there is a sub dgs Lie algebra consisting of all endomorphisms which
act on sections via differential operators. We denote this subalgebra by (D(E), [D, −]).

Definition 3.2.1. A strict local dgs g-module is a triple (E, D, ρ) where (E, D) is a dgs
vector bundle and
ρ : g −→ (D(E), [D, −]) (3.6)
Derived pure spinor superfields 93

is a map of dgs Lie algebras. Here the super Lie algebra g is viewed as a dgs Lie algebra
in cohomological degree zero with trivial differential.

Remark 3.2.2. This definition (as well as many of the following) has a natural gener-
alization for g a dgs Lie algebra. Since we are ultimately interested in the pure spinor
superfield formalism, we restrict our attention to super Lie algebras with no cohomolog-
ical grading.

Note that, since a super Lie algebra g has no differential, ρ commutes with the differential
on the dgs vector bundle,
[D, ρ(x)] = 0 ∀x ∈ g . (3.7)

It is standard to encode a g-module structure ρ on (E, D) as a dgs Lie algebra structure

on the direct sum g ⊕ E. Concretely we set for the unary and binary operations

µ1 (x1 , σ1 ) = (0, Dσ1 )

(3.8)
µ2 ((x1 , σ1 ), (x2 , σ2 )) = ([x1 , x2 ] , ρ(x1 )σ2 − (−1)|σ1 ||x2 | ρ(x2 )σ1 ) ,

where x1 , x2 ∈ g and σ1 , σ2 ∈ E.

There is an obvious notion of morphisms between strict local g-modules.

Definition 3.2.3. A strict morphism of strict local g-modules (E, D, ρ) and (E 0 , D0 , ρ0 )

is a map of cochain complexes
ψ : E −→ E 0 (3.9)

realized by differential operators such that

ψ ◦ ρ(x) = ρ0 (x) ◦ ψ (3.10)

for all x ∈ g.

A strict morphism ψ : (E, D, ρ) −→ (E 0 , D0 , ρ0 ) gives rise to a strict morphism of the

associated dgs Lie algebras by setting ψ̃ = idg ×ψ : g ⊕ E −→ g ⊕ E 0 . Conversely it is easy
to check that every strict morphism of dgs Lie algebras of that form gives rise to a strict
morphism of g-modules. We call ψ a quasi-isomorphism if it is a quasi-isomorphism of
dgs vector bundles; equivalently ψ̃ is a quasi-isomorphism of dgs Lie algebras.

Now, suppose that (E, D) is a dgs vector bundle over an affine space V (so, in the
notation used above, V = X). Let us additionally assume that g is equipped with a map
of super Lie algebras
φ : aff(V ) −→ g. (3.11)
94 Derived pure spinor superfields

In essence, when we refer to strict g-multiplets, we are referring to strict local g-modules
for which the affine transformations acts in a geometric way.

Definition 3.2.4. A strict g-multiplet (E, D, ρ) on V is an affine dgs vector bundle over
V equipped with a strict local g-module structure

ρ : g −→ D(E) (3.12)

such that the pullback of the module structure along φ agrees with the natural action on
sections of the affine vector bundle. Concretely, this means that the following diagram
commutes.
ρ
g D(E)
φ (3.13)
aff
aff(V )

Here aff denotes the natural action of the affine algebra on E. We define the category of
strict multiplets with strict morphisms to be the full subcategory of the category of strict
local g-modules with objects strict g-multiplets. We denote this category by Multstrict
g .

Remark 3.2.5. Note that a morphism of strict local g-modules is automatically compat-
ible with the action of aff(V ). For example let ψ : (E, D, ρ) −→ (E 0 , D0 , ρ0 ) be a strict
morphism. Then we have

ψ ◦ aff(x) = ψ ◦ ρ(φ(x)) = ρ0 (φ0 (x)) ◦ ψ = aff 0 (x) ◦ ψ . (3.14)

Therefore it is sensible to define Multstrict

g as a full subcategory of strict local g-modules.

3.2.2 Homotopy theory of multiplets

The appropriate homotopy theoretic generalization of strict g-module structures are L∞

g-modules. In the context of this work, we will refer to such homotopy module structures
simply as module structures and choose to emphasize whenever a module is strict instead.
In this spirit we can easily define (not necessarily strict) local g-modules and g-multiplets
by replacing Lie maps in the above definitions with L∞ maps.

Recall that a local g-module is just a dgs vector bundle (E, D) together with an L∞ map
of L∞ algebras
ρ: g D(E) . (3.15)

We can expand ρ in component maps

ρ(k) : g⊗k −→ D(E)[1 − k] k≥1 (3.16)

Derived pure spinor superfields 95

satisfying a series of compatibility relations. As usual, we recover strict modules if and

only if ρ(k) = 0 for all k ≥ 2.

Similarly to the strict case, we can conveniently encode a g-module structure ρ as an

L∞ structure on g ⊕ E. To this end we supplement the operations (3.8) by the following
brackets for k ≥ 3. For details we refer to [Lad04; All10].

k
!
X
µk ((x1 , v1 ), . . . , (xk , vk )) = 0, ±ρ(k−1) (x1 , . . . , x̂i , . . . , xk )vi (3.17)
i=1

One can define a morphism ψ : (E, D, ρ) −→ (E 0 , D0 , ρ0 ) of local g-modules by component

maps
ψn : g⊗n−1 ⊗ E −→ E 0 , (3.18)

which are given by differential operators and satisfy a series of compatibility relations
(see [All10] for details). We recover strict morphisms by restricting to those where the
only component map is ψ1 . Again, we can describe such morphisms by morphisms of
the associated L∞ algebras
ψ̃ : g ⊕ E g ⊕ E0 (3.19)

by supplementing ψ̃1 = idg ×ψ1 with

k
X
ψ̃k ((x1 , σ1 ), . . . , (xk , σk )) = (0 , ±ψk (x1 , . . . , x̂i , . . . , xk , σi )) (3.20)
i=1

for k ≥ 2. ψ is a quasi-isomorphism of local g-modules if ψ1 is a quasi-isomorphism of

cochain complexes, or equivalently if ψ̃ is a quasi-isomorphism of L∞ algebras. In general,
encoding module structures as L∞ structures is very convenient because it allows the
use of many known tools from the theory of L∞ algebras, like homotopy transfer for
instance.

Remark 3.2.6. An equivalent realization of the notion of a non-strict morphism between

a pair of local L∞ algebras g, g0 is provided by considering the Chevalley–Eilenberg chain
complex C• (g) of an L∞ algebra g. This is a cocommutative dg coalgebra whose underly-
ing graded coalgebra is Sym• (g[−1]), with differential induced from the L∞ structure on
g as a sum of terms given by the L∞ brackets on g. The data of a morphism ψ : g → g0
of local L∞ algebras is equivalent to that of a morphism of cocommutative dg coalgebras

ψ 0 : C• (g) → C• (g0 )

given by differential operators. This interpretation allows us to view local L∞ algebras

as objects of a dg-category, so that we can discuss (for example) homotopies between
L∞ algebra morphisms. When the source g is finite-dimensional, we can dually consider
96 Derived pure spinor superfields

morphisms of commutative dg algebras C • (g0 ) → C • (g). This is for example the case for
the local g-module structures ρ : g D(E) appearing in the definition of multiplets.

Replacing the strict morphisms ρ in the definition of strict multiplets by an L∞ -map, we

recover the definition of a multiplet as presented in Definition 2.2.11.

Definition 3.2.7. We define the dg-category of g-multiplets to be the full subcategory

of local g-modules with objects being g-multiplets and denote it by Multg .

We will sometimes wish to refer to the full subcategory of strict multiplets, but allowing
arbitrary morphisms.

Definition 3.2.8. Denote by Multstrict-ob

g the full sub dg-category of Multg generated by
strict multiplets.

We can obtain homotopy categories from the dg-categories Multg as well as Multstrict
g by
replacing the hom space HomMultg (E, E 0 ) by its zeroth cohomology H 0 (HomMultg (E, E 0 )).
We denoted the resulting categories by Ho(Multg ) and Ho(Multstrict
g ). Speaking physi-
cally, quasi-isomorphisms of g-multiplets correspond to perturbative equivalences of mul-
tiplets, where by “perturbative” here we mean equivalences of the derived formal neigh-
borhoods of a point in the classical moduli field space. Therefore isomorphism classes in
the homotopy category correspond to perturbatively distinct multiplets.

Remark 3.2.9. Because every L∞ algebra can be strictified [KM95, Part II Corollary
1.6], the natural inclusion

Ho(Multstrict-ob
g ) → Ho(Multg ) (3.21)

is an equivalence of categories, where Multstrict-ob

g denotes the full subcategory of multi-
plets spanned by strict objects.

3.2.3 Linear structure

We can define the direct sum of two g-multiplets as follows.

Definition 3.2.10. The direct sum of two g-multiplets (E, D, ρ) and (E 0 , D0 , ρ0 ) is de-
fined to be the multiplet

(E, D, ρ) ⊕ (E 0 , D0 , ρ0 ) = (E ⊕ E 0 , D ⊕ D0 , ρ ⊕ ρ0 ) . (3.22)

Remark 3.2.11. In contrast, defining a tensor product on the category of g-multiplets

is not straightforward. Considering the D-modules of global sections, one can take the
Derived pure spinor superfields 97

tensor product in the category of D-modules, but this does not take the additional
structures on a multiplet into account. As alluded to in the introduction, one can might
also try using the equivalence established in §3.4 to define a product on the category
which makes the functor monoidal. We will not discuss this issue further in this work.

For further reference, we also define the dual of a g-multiplet. In physics context, these
are usually called antifield multiplets

Definition 3.2.12. The dual of a g-multiplet (E, D, ρ), also referred to as the associated
antifield multiplet is the g-multiplet (E ! , D∨ , ρ∨ ). Here E ! is the linear dual vector bundle
to E twisted by the canonical bundle. The action ρ∨ denotes the map

ρ∨ : g D(E ! ) (3.23)

given by ρ∨(k) (Q1 , . . . , Qk ) = ρ(k) (Q1 , . . . , Qk )∨ .

Note that we will typically be working over the flat space Rn , so we may choose a
trivialization of the canonical bundle if we wish to identify E ! with E ∨ .

3.3 Derived pure spinor superfields

In this section, we define the derived generalization of the pure spinor superfield con-
struction, which will provide one of the two functors that witness the equivalence of
categories. We begin by setting up the general context in which we want to work.

Let us choose a basis dα for n1 and eµ for n2 such that we can expand the symmetric
bracket in structure constants3

µ
[dα , dβ ] = fαβ eµ . (3.24)

We denote by R = Sym• (n∨ α

1 ) = C[λ ] the ring of polynomial functions on n1 . For Q in
n1 , the equation [Q, Q] = 0 defines an ideal I in R, explicitly given by

µ β
I = (λα fαβ λ ). (3.25)

The pure spinor superfield formalism gives a systematic tool to construct g-multiplets
from the input datum of a graded g0 -equivariant R/I-module. Here we generalize the
pure spinor superfield construction to g0 -equivariant C • (n)-modules and show that it
defines a functor.
3
In the context of super Poincaré algebras, these structure constants are typically expressed in terms
of the matrix elements of the gamma matrices.
98 Derived pure spinor superfields

3.3.1 The category of C • (n)-modules

Recall that the Chevalley–Eilenberg complex of n takes the form

C • (n) = Sym• (n∨ [1]) , dCE

(3.26)

where the Chevalley–Eilenberg differential dCE is induced by the dual of the bracket.
Here the notation n∨ [1] means that we shift n∨ down in cohomological degree by one.
The Chevalley–Eilenberg complex has a Z×Z/2Z-grading endowing it with the structure
of a dgs algebra. In our present case, since the the Z/2Z grading of n lifts to a Z-grading
where ni has weight i, C • (n) can be given a Z × Z-grading. Totalizing this grading, the
generators of n∨ ∨
1 sit in degree 0 while the generators of n2 live in degree −1. We denote
these generators by λα and v µ respectively. We can thus identify

C −p (n) = ∧p n∨
2 ⊗R (3.27)

with respect to the totalized grading.

The Chevalley–Eilenberg differential acts on these generators according to

dCE λα = 0
(3.28)
µ β
dCE v µ = λα fαβ λ .

In coordinates we will often write the Chevalley–Eilenberg algebra in the form

 
• α µ α µ β ∂
(C (n) , dCE ) = C[λ , v ] , dCE = λ fαβ λ . (3.29)
∂v µ

Using this description, it is immediate to see H 0 (n) = R/I.

C • (n) is a dgs algebra, therefore we can consider dgs modules over it.

Definition 3.3.1. A C • (n)-module is a dgs vector space (Γ, dΓ ) together with a mor-
phism
(C • (n) , dCE ) −→ (End(Γ) , [dΓ , −]), (3.30)

of dgs algebras.

Definition 3.3.2. A morphism of C • (n)-modules (Γ, dΓ ) and (Γ0 , dΓ0 ) is a cochain map

f : (Γ, dΓ ) −→ (Γ0 , dΓ0 ) (3.31)

such that
f (x ·Γ γ) = x ·Γ0 f (γ) . (3.32)
Derived pure spinor superfields 99

In other words, f is just a morphism of dgs modules.

Note that g0 acts on both n1 and n2 and thus also on C • (n).

Definition 3.3.3. A C • (n) module Γ is called g0 -equivariant if Γ is also a representation

of g0 and the module structure map is g0 -equivariant. We further assume that each degree
Γk is finite dimensional as a complex vector space. We will denote the dg-category of
g0 -equivariant C • (n)-modules by ModgC0• (n) .

Our main results will take place in this category of g0 -equivariant modules.

3.3.2 The derived pure spinor superfield formalism

Recall that n is a two-step nilpotent super Lie algebra and let N = exp(n) be the
associated nilpotent super Lie group. Left and right translations induce two commuting,
aut(n)-equivariant actions of n on N by vector fields

L , R : n −→ Vect(N ). (3.33)

Let us choose coordinates xµ on the abelian group N2 and θα on N1 . In terms of these

coordinates, the vector fields given by the action of the odd elements can be described
as follows.

∂ µ β ∂
R(dα ) = α
− fαβ θ
∂θ ∂xµ (3.34)
∂ µ β ∂
L (dα ) = α + fαβ θ
∂θ ∂xµ

In addition, the even elements simply act by derivatives

∂
R(eµ ) = L (eµ ) = (3.35)
∂xµ

As before, the free superfield is the strict g-multiplet with E = C ∞ (N ), vanishing differ-
ential, and module structure given by the left translations L .

Let us now generalize the pure spinor superfield formalism to a derived setting.

Definition 3.3.4. We define the pure spinor functor A• : ModgC0• (n) −→ Multstrict
g by
setting
A• (Γ) = (C ∞ (N ) ⊗C Γ , D) , (3.36)

for an object (Γ, dΓ ). The differential D is constructed using the right action R and the
C • (n)-module structure on Γ. Explicitly, it takes the form

D = λα R(dα ) + v µ R(eµ ) + dΓ . (3.37)

100 Derived pure spinor superfields

For a morphism f : Γ −→ Γ0 we define

A• (f ) = idC ∞ (N ) ⊗f : A• (Γ) −→ A• (Γ0 ) . (3.38)

A few comments are in order.

— The differential D squares to zero precisely since Γ is a C • (n)-module.

— We can equip A• (Γ) with the structure of a dgs vector bundle over the spacetime
V := N2 by placing C ∞ (N ) in cohomological degree zero. Note that in particular
the differential is of bidegree (1, +).

— Further, g acts on C ∞ (N ) via left translations and on Γ by the trivial extension of

the g0 -module structure which was part of the input datum. The tensor product
of these two action makes A• (Γ) into a strict multiplet.

— It is immediate to check that A• (f ) is a strict morphism of strict multiplets. Since

f is a morphism of C • (n)-modules we have

A• (f ) ◦ D = D0 ◦ A• (f ) . (3.39)

In addition,
A• (f ) ◦ L = L ◦ A• (f ) (3.40)

obviously follows from the definition.

— A• is additive. The direct sum of two C • (n)-modules is mapped to the direct sum
of the respective multiplets, A• (Γ ⊕ Γ0 ) = A• (Γ) ⊕ A• (Γ0 ).

This construction is a direct generalization of the pure spinor superfield formalism as

described in [Eag+22]. To see this, we notice that the category of graded equivariant of
R/I-modules sits as a subcategory inside Modstrict
C • (n) , namely precisely as those modules
concentrated in cohomological degree zero. Indeed, every R/I-module is a C • (n)-module
by the map
C • (n) = C[λα , v µ ] −→ R/I (λα , v µ ) 7→ λα . (3.41)

The other way round, let (Γ, dΓ ) be a C • (n)-module concentrated in cohomological degree
zero. Then the differential dΓ vanishes and v µ acts trivially for degree reasons. Therefore
one has, for γ any element of Γ,

0 = dΓ (v µ · γ) = (dCE v µ )γ = (λf µ λ)γ , (3.42)

which endows Γ with the structure of an R/I-module.

Derived pure spinor superfields 101

Restricting the functor A• to graded equivariant R/I-modules, we obtain a functor

A•R/I : ModgR/I
0
−→ Multstrict
g (3.43)

where the output simplifies to

A•R/I (Γ) = (C ∞ (N ) ⊗C Γ , λα R(dα )) (3.44)

such that we precisely recover the pure spinor superfield formalism as presented in §2.

Further, this is a derived generalization of the pure spinor superfield construction in

the sense that C • (n) can be viewed as a derived replacement of the ring R/I. We will
therefore refer to this construction as the derived pure spinor superfield formalism.
Remark 3.3.5. We can alternatively view this geometrically as a derived enhancement of
the affine nilpotence variety Spec(R/I). We can view a multiplet as arising from a quasi-
coherent sheaf over the nilpotence variety, but this point of view requires forgetting some
of the data given by the C • (n) action. The philosophy of derived algebraic geometry
suggests instead retaining this information by viewing a multiplet as arising from a
coherent sheaf over the affine derived scheme Spec(C • (n))—the derived analogue of the
nilpotence variety.

3.3.3 The multiplet associated to C • (n)

As a first example we can plug C • (n) itself into the pure spinor functor A• and study
the associated multiplet.

Lemma 3.3.6. There is a natural equivalence

A• (C • (n)) ' Ω• (N ). (3.45)

Proof. First, we can describe

A• (C • (n)) = (C ∞ (N ) ⊗ C • (n) , D = λα R(dα ) + v µ R(eµ ) + dCE ) . (3.46)

Let us write V for the even part N2 of N , viewed as an affine space. We can identify
C ∞ (N ) = C ∞ (V ) ⊗ C[θα ] and C • (n) = C[λα , v µ ]. The differential takes the explicit
form
∂ µ β ∂ ∂ µ β ∂
D = λα − λα fαβ θ µ
+ v µ µ + (λα fαβ λ ) µ. (3.47)
∂θ ∂x ∂x ∂v
On the other hand, N is parallelizable, therefore its de Rham complex takes the form

Ω• (N ) ∼
= C ∞ (N ) ⊗ Sym• (n∨ • ∨
1 ) ⊗ ∧ n2 . (3.48)
102 Derived pure spinor superfields

Identifying dθα = λα and dxµ = v µ , the de Rham differential takes the form

∂ ∂
ddR = λα + vµ µ . (3.49)
∂θα ∂x

Further, g acts on the de Rham complex via left translation making it a strict multiplet.
The map defined in coordinates via

(A• (C • (n)) , D , L ) −→ (Ω• (N ) , ddR , L ) (xµ , λα , θα , v µ ) 7→ (xµ , λα , θα , v µ +λf µ θ)

(3.50)
is a quasi-isomorphism of g-multiplets. Therefore, we can identify the multiplet associ-
ated to C • (n) itself as the differential forms on the super Lie group N .

Remark 3.3.7. We can further compute cohomology with respect to D0 = λα ∂θ∂α and
obtain a deformation retract to the de Rham complex on the spacetime manifold V :

p
h (Ω• (N ) , D0 ) (Ω• (V ) , 0) (3.51)
i

Here, i is the embedding of the factor of polynomial degree 0 in the λ and θ variables,
and p is the obvious projection. The homotopy h is given by h = θα ∂λ∂α . The induced
differential via homotopy transfer is the de Rham differential on V . The module struc-
ture, however, is no longer strict. In fact the strict part now vanishes, but there are now
quadratic pieces appearing. These take the form

ρ(2) (Q, Q) = ι[Q,Q] : Ωk (V ) −→ Ωk−1 (X) . (3.52)

Here we view [Q, Q] as a constant vector field on V and ι denotes the contraction of a
differential form with a vector field.

There is, of course, a further quasi-isomorphism of g-modules to the trivial g-module C.

In this spirit, each of the multiplets A• (C • (n)), Ω• (N ), and Ω• (V ) can be viewed as free
resolutions of the trivial module—either free over spacetime (i.e. as C ∞ (V )-modules),
or even free over superspace (i.e. as C ∞ (N )-modules). Note, however, that the trivial
module does not form a g-multiplet since the translations do not act geometrically.
Therefore this last quasi-isomorphism is just a quasi-isomorphism of g-modules.

3.3.4 Component fields and homotopy transfer

The multiplet A• (Γ) given by applying the pure spinor functor does not at first glance
resemble the more standard component field formulations known from physics. These
Derived pure spinor superfields 103

component field multiplets are distinguished by the fact that they are given by dgs
vector bundles whose total rank (as vector bundles over spacetime) is finite. In other
words, component field multiplets are usually defined with the assumption that they
only contain a finite number of component fields. With some care, it is always possible
to choose a homotopy representative for A• (Γ) that satisfies this condition, as long as Γ
satisfies a finiteness condition.

Lemma 3.3.8. Let Γ be a C • (n)-module such that only finitely many cohomology groups
H • (Γ) are non-vanishing and each of these is finitely generated as an R-module. Then
A• (Γ) is quasi-isomorphic to a multiplet of finite rank.

In the rest of this section, we will explain an algorithm that provides explicit finitely
generated component field representatives for A• (Γ), thus establishing the lemma. We
already discussed a procedure to extract such a component field multiplet out of the
underived model A•R/I (Γ) in §2.3.6. In essence, this is done by homotopy transfer: one
identifies a retraction to another quasi-isomorphic complex and then applies homotopy
transfer to the structures present for the multiplet. By construction, this yields a quasi-
isomorphic and thus physically equivalent multiplet. Crucially, the component field
multiplets obtained in that way are not necessarily strict anymore. Higher-order terms
in the module structure can arise during the transfer.

Let us begin by summarizing the procedure for the underived pure spinor formalism A•R/I
and then describe the generalization to A• .

Minimal models for A•R/I . Recall that the differential on (A•R/I (Γ), D) admits an
obvious splitting
∂ µ β ∂
D = D0 + D1 = λα α
− λα fαβ θ , (3.53)
∂θ ∂xµ
which can be viewed by equipping A•R/I (Γ) with the filtration discussed in §2.3.3 and
splitting the differential into the terms that preserve and lower filtered degree. We can
take cohomology with respect to D0 and then perform homotopy transfer along the
diagram
p
h (A• (Γ) , D0 ) (H • (A• (Γ), D0 ), 0) . (3.54)
i

We observe, by [LV12a, Theorem 10.3.15], that although the transfer data depends on
a choice of section for the projection onto the cohomology, the multiplet obtained by
homotopy transfer is independent of this choice up to isomorphism. In this example, it
is always “minimal” in the sense that its differential does not contain terms of order zero
104 Derived pure spinor superfields

in differential operators. Intuitively, this means that we cannot take any further coho-
mology without leaving the category of multiplets. (In more general examples coming
from C • (n)-modules, “minimal” is not related to having no terms of order zero in the
differential; the massive Klein–Gordon field is a simple example.)

Starting from a multiplet of the form A•R/I (Γ), we will refer to the corresponding minimal
multiplet as µA•R/I (Γ). One can identify the D0 -cohomology with the Koszul cohomology
of Γ, tensored with functions on spacetime,

H • (A•R/I (Γ)) = C ∞ (V ) ⊗ H • (K • (Γ)) . (3.55)

The Koszul cohomology is conveniently computed by a minimal free resolution L of Γ in

R-modules. The minimal multiplet takes the form

µA•R/I (Γ) = C ∞ (V ) ⊗ (L ⊗R C) , D0 , ρ0 ,


(3.56)

where D0 is the differential induced from D1 and ρ0 the module structure induced from
L via homotopy transfer. For Γ a finitely generated module, Hilbert’s syzygy theorem
states that the minimal free resolution exists, consists of finitely generated modules and
its length is less or equal than dim(n1 ) (see for example [Eis95, Theorem 1.13]; for a
discussion in the equivariant case we refer to [Gal16, Proposition 2.4.9, Remark 2.4.10]).
Therefore, µA•R/I (Γ) is indeed of finite rank over spacetime, i.e. it provides a reasonable
component field multiplet.

Minimal models for A• . Let us now discuss a generalization of this procedure to the
functor A• . We will assume in this section that Γ carries an action of the abelian Lie
algebra R compatible with the action of R on n where ni has R-weight i. This will be
used to construct a filtration at the very end of this section (however, this assumption
is not required for Lemma 3.3.8).

Recall that the differential on A• (Γ) takes the form

∂ µ β ∂ ∂
D = λα α
− λα fαβ θ µ
+ dΓ + v µ µ . (3.57)
∂θ ∂x ∂x

One can construct a finite-dimensional component field model in two steps. First, one
takes cohomology with respect to the internal differential of the module dΓ and performs
homotopy transfer along the diagram

p
h (A• (Γ) , dΓ ) (C ∞ (N ) ⊗ H • (Γ)), 0) . (3.58)
i
Derived pure spinor superfields 105

The differential and the C • (n)-module structure of the resulting multiplet may contain
additional pieces induced from homotopy transfer. Since H • (Γ) is a C • (n)-module with
vanishing differential, each homogeneous summand H k (Γ) carries the structure of an
R/I-module. Thus, we can proceed by taking cohomology with respect to the Koszul
differential D0 = λα ∂θ∂α , and applying the homotopy transfer along a retraction on to
this cohomology. As before, the D0 -cohomology is computed by minimal free resolutions
of the individual H k (Γ) in R-modules, and as before, it is independent of the choice of
transfer datum. The resulting multiplet is thus of the form
M
C ∞ (V ) ⊗ ( L•k ⊗R C) , (3.59)
k

where L•k is the minimal free resolution of H k (Γ). As long as the cohomology H • (Γ) is
bounded, this is already a finite rank vector bundle over spacetime. The field content
of this multiplet is just the field content of the direct sum of all the minimal multiplets
associated to the cohomology groups, i.e.
M
µA• (H k (Γ)) . (3.60)
k

At this stage we have already obtained a finite-dimensional model, as required by Lemma

3.3.8. However, additional acyclic differentials induced by homotopy transfer can still be
present. At this final stage, we will use the filtration on each multiplet µA• (H k (Γ)) by
weight with respect to the R-action on Γ. We define µA• (Γ) by taking the cohomology
with respect to the summand of the total differential of filtered degree zero (in other
words, we pass to the E1 page of the associated spectral sequence), and apply homotopy
transfer. We will illustrate this procedures using examples in §3.5.

3.4 An equivalence of categories

We now show that the derived pure spinor superfield formalism provides an equivalence
of categories between the dg categories of g0 -equivariant C • (n)-modules and g-multiplets.
This implies in particular that, up to quasi-isomorphism, every g-multiplet can be con-
structed using the derived pure spinor superfield formalism.

Remark 3.4.1. Recall that for a general Lie algebra there is a Koszul duality equivalence
between the dg-categories of C• (g)-comodules and U (g)-modules. If g is, for instance,
finite dimensional, we can instead consider C • (g)-modules. For a relevant discussion in a
similar context, see [Cos13b, §7 and §8]. Relatedly, Kapranov [Kap91] gives a formulation
of Koszul duality that establishes a Quillen equivalence between the categories of D-
modules and Ω• -modules on the same space. Our results show that the pure spinor
106 Derived pure spinor superfields

formalism admits a natural derived generalization that is closely related to Kapranov’s

construction; however, in our setting, one is working on a supermanifold, and asks for
appropriate equivariance conditions. Alternatively, our procedure can be viewed in two
steps, as an explicit form of commutative/Lie Koszul duality tailored to the examples in
question, combined with an associated bundle construction that realizes a U (n)-module
as an n-equivariant dg vector bundle over V = N2 .

3.4.1 The inverse functor: derived n-invariants

Any g-module is in particular a n-module, and we can thus take derived invariants with
respect to n. For multiplets, this defines a functor in the opposite direction to the functor
A• , assigning a strict C • (n)-module to a multiplet. The resulting C • (n)-module is also
g0 -equivariant. The functor takes the form

C • (n, −) : Multstrict
g −→ ModgC0• (n) , (3.61)

It maps the multiplet (E, D, ρ) to C • (n, E); this Chevalley–Eilenberg complex can be
written more explicitly as

C • (n, E) = (C • (n) ⊗ E , dCE + D + λα ρ(dα ) + v µ ρ(eµ )) . (3.62)

It is a strict C • (n)-module; the action is on C • (n) via multiplication and on E via the
identity. Morphisms are mapped according to the rule

ψ 7→ idC • (n) ⊗ψ. (3.63)

Remark 3.4.2. The assignment on the level of objects is also well defined for not neces-
sarily strict modules. Then the higher-order terms of the g-module structure enter the
differential such that the term λα ρ(dα ) is replaced by

∞
X
λ · ρ := λα1 . . . λαk ρ(k) (dα1 , . . . , dαk ) . (3.64)
k=1

Notice that the output is still a strict C • (n)-module.

There are several ways to intuitively understand the fact that the inverse functor is
given by the derived invariants of supertranslations. One can of course appeal to the
general structure of Koszul duality. On a more down-to-earth level, one can recall that
the component fields of a supermultiplet in the usual pure spinor superfield formalism
correspond to the generators of the minimal free resolution of that module over R =
C • (Πn1 ). The resolution differential agrees with those supersymmetry transformations
Derived pure spinor superfields 107

that are order zero in spacetime derivatives. Since spacetime derivatives act by zero
on translation-invariant sections of E, it is clear that one can think of the minimal
free resolution of the module as arising from the derived Πn1 -invariants of translation-
invariant sections: R consists of the Lie algebra cochains of Πn1 , the generators of the free
graded R-module arise from the component fields of the multiplet, and the differential
encodes the action of supersymmetry on translation-invariant sections. It is clear that
this story is just a two-step procedure to compute derived n-invariants by first taking
n2 -invariants, and then accounting for the action of supersymmetry.

To flesh this story out, we now describe the module C • (n, E) in some more detail and
sketch how its cohomology can be computed. Recall that, since E is a multiplet, the
action of n2 is just given by derivatives along the coordinate directions

∂
ρ(eµ ) = . (3.65)
∂xµ

Thus, taking cohomology with respect to the term v µ ρ(eµ ) in the differential means
restricting to translation invariant sections of E and eliminating v µ . Denoting the fiber
of E over 0 by E0 we find a quasi-isomorphism

C • (n, E) ' (R ⊗ E0• , λ · ρconstants ) . (3.66)

Note that E0• carries a Z × Z/2Z-grading since it comes from a dgs vector bundle. In
addition, there is an integer grading by polynomial degree in λ present. In the following,
we will label cohomology groups by the former degree. Then each cohomology group
is an g0 -equivariant R/I-module graded by polynomial degree in λ. If the cohomology
is concentrated in a single degree, the complex on the right hand-side can be viewed
as the minimal free resolution of the R/I-module forming the cohomology. Note that
this is indeed a minimal free resolution, since all terms in the differential are of nonzero
polynomial degree in λ. It will follow from the theorem below that this R/I-module is
precisely the algebraic input datum the multiplet can be constructed from in the pure
spinor superfield formalism, i.e. by applying the functor A•R/I .

3.4.2 Main theorem and proof

Let us now show that the functors A• and C • (n, −) induce an equivalence of dg-categories
between the homotopy categories of multiplets and equivariant C • (n)-modules.

Theorem 3.4.3. A• and C • (n, −) provide an equivalence of dg-categories between Multstrict-ob

g
and ModgC0• (n) .
108 Derived pure spinor superfields

Proof. We first show that there is an equivalence of dg-functors

idModg0• → C • ◦ A• . (3.67)
C (n)

We will naturally construct quasi-isomorphisms

Γ ' C • (n, A• (Γ)) (3.68)

for each equivariant C • (n)-module (Γ, dΓ ).

We can explicitly describe C • (n, A• (Γ)) by

 
0 0 ∞ 0 µ 0 ∂ 0µ ∂ 0α
C[λ , v ] ⊗ C (N ) ⊗ Γ , dΓ + (λ f λ ) 0µ + v + λ L (dα ) + λ R(dα ) + v R(eµ ) .
α µ
∂v ∂xµ
(3.69)
Here, we made a notational distinction between the generators of the C • (n) in the con-
struction (denoted by λ0 and v 0 ) and the action of C • (n) on Γ (denoted by λ and v).

The differential contains a piece of the form

∂ ∂
ddR = v 0µ µ
+ λ0α α . (3.70)
∂x ∂θ

Thus, we can identify

 
• • 0µ ∂ 0α ∂
C (n, A (Γ)) , v +λ = (Ω• (N ) , ddR ) ⊗ Γ . (3.71)
∂xµ ∂θα

Since the de Rham complex on N is acyclic, this complex is quasi-isomorphic to Γ. We

can fix homotopy data

p
h (Ω• (N ) ⊗ Γ, ddR ) (Γ, 0) . (3.72)
i

It is easy to see that the only induced differential on the right hand side is dΓ , so that
we obtain a quasi-isomorphism uniformly for all choices of Γ:

C • (n, A• (Γ)) ' (Γ, dΓ ). (3.73)

Now, for any morphism f : Γ → Γ0 of C • (n)-modules, we can identify

C • (A• (f )) = idC • (n) ⊗ idC ∞ (N ) ⊗f. (3.74)

Derived pure spinor superfields 109

So there is a commutative square

Γ C • (n, A• (Γ))
f A• (f ) (3.75)

Γ0 C • (n, A• (Γ0 ))

inducing an equivalence of hom complexes Hom(Γ, Γ0 ) → Hom(C • (n, A• (Γ)), C • (n, A• (Γ0 ))).

Conversely, we will construct an equivalence of dg-functors

A• ◦ C • → idMultstrict-ob . (3.76)
g

Let (E, D, ρ) be a g-multiplet. We can describe A• (C • (n, E)) explicitly by

 
∞ ∂
C[λ, v] ⊗ E ⊗ C (N ) , D + λf λ µ + λ · ρ + v ρ(eµ ) + λ R(dα ) + v R(eµ ) .
µ µ α µ
∂v
(3.77)
We denote coordinates on the base of the vector bundle E by xµ and on N2 by y µ . In
this notation, we find
 
∂ ∂
v (ρ(eµ ) + R(eµ )) = v
µ µ
+ . (3.78)
∂xµ ∂y µ

Taking cohomology with respect to this piece, we obtain a quasi-isomorphic complex of

the form
(C[λ, θ] ⊗ E , D + λ · ρ + λα R(dα )) . (3.79)

The differential contains a piece corresponding to the Koszul differential

∂
dK = D0 = λα . (3.80)
∂θα

Clearly, taking cohomology with respect to the Koszul differential we arrive at E. Con-
cretely let us consider homotopy data

p
D0† (C[λ, θ] ⊗ E, D0 ) (E, 0) , (3.81)
i

where the homotopy is given by D0† = θ ∂λ

∂
, while i is the obvious inclusion and p evaluates
at λ = θ = 0. It is easy to see that the induced differential is just D. Thus, we find
110 Derived pure spinor superfields

homotopy data
p0
D00† (C[λ, θ] ⊗ E, d) (E, D) , (3.82)
i0

providing a quasi-isomorphism of cochain complexes (here d denotes the full differential

in (3.79)). The induced maps are given by

∞
(D0† (D + D1 + λ · ρ))n ◦ i
X
0
i =
n=0
∞
(D0† (D + D1 + λ · ρ))n = p
X
0
p =p◦ (3.83)
n=0
∞
D00† = D0† ◦ ((D + D1 + λ · ρ)D0† )n .
X

n=0

By construction, p ◦ D0† = 0 and thus p0 = p. We note that these sums are all finite
by degree reasons, since the left hand side in (3.82) is concentrated in finitely many
degrees with respect to the grading by polynomial degree in the θ-variables, and in each
expression in equation (3.83), the operator being raised to the power n within the sum
raises this θ-degree. Therefore for n sufficiently large, all terms in the sum defining our
maps vanish.

We have to check that this not only provides a quasi-isomorphism of cochain complexes,
but of multiplets. Therefore we transfer the module structure induced by L to the right
hand side and check that it agrees with the original module structure ρ of the multiplet.

Explicitly, the transferred module structure can be described by

ρL (Q, . . . , Q) = pL (Q)(D00† L (Q))k−1 i0

(k)

∞ ∞
pL (Q)(D0† ρ)D0† )n L (Q))k−1 (D0† (D + D1 + λ · ρ))j ◦ i.
X X
= ((D + D1 + λ ·
n=0 j=0

(3.84)

Since p projects onto λ = θ = 0, only terms of order zero in λ and θ can contribute. It
is easy to see that the only such term is
 k−1
(k) ∂ † ∂
ρL (Q, . . . , Q) = p D0  D0† λ · ρ(k) i . (3.85)
∂θ ∂θ

Here we expressed Q in a basis, Q = α dα . Noting that

∂ ∂
[D0† ,  ]= , (3.86)
∂θ ∂λ
Derived pure spinor superfields 111

we deduce
(k)
ρL (Q, . . . , Q) = ρ(k) (Q, . . . , Q) . (3.87)

This shows that A• (C • (n, E)) ' (E, D, ρ) as multiplets.

As before, for any morphism ψ : E → E 0 we can realize

A• (C • (ψ)) = C • (ψ) ⊗ idC ∞ (N ) . (3.88)

So there is a commutative diagram

A• (C • (n, E)) E
C • (ψ) ψ (3.89)

A• (C • (n, E 0 )) E0

inducing an equivalence of hom spaces, and hence we have an equivalence of dg-categories.

Remark 3.4.4. The equivalence of dg-categories automatically induces an equivalence of

the underlying homotopy categories. Hence, each multiplet is perturbatively equivalent
to a multiplet constructed via the derived pure spinor superfield formalism.

3.4.3 Some consequences of the theorem

Let us now discuss some consequences of the equivalences of categories.

Corollary 3.4.5. Let (E, D, ρ) be any multiplet. A• (C • (n, E)) is a strictification.

Note that this does not imply that there exist a strict finitely generated component
field multiplet that is equivalent to E. In particular, E need not admit an auxiliary
field formulation in the usual sense. The strictification A• (C • (n, E)) typically contains
infinitely many component fields.

We can further use the equivalence of categories to derive some statements on the (un-
derived) pure spinor superfield formalism, i.e. the functor A•R/I .

Corollary 3.4.6. The essential image of the functor A•R/I consists of those multiplets
for which H • (n, E) is concentrated in a single Z-degree after totalization of the natural
Z × Z-grading.

This gives a description of all multiplets which can be constructed via the pure spinor
superfield formalism. We already argued in §2.4.4 that the antifield multiplet of the
112 Derived pure spinor superfields

four-dimensional vector multiplet cannot be constructed via A•R/I . We come back to

this example in §3.5.2 where we compute the relevant cohomologies and explain how the
multiplet is built in the derived formalism.

Corollary 3.4.7. The functors A•R/I and H • (n, E) provide an equivalence of categories
between the essential image of A•R/I and the category of graded g0 -equivariant R/I-
modules.

Suppose we have an R/I-module Γ with associated minimal multiplet µA•R/I (Γ). As

discussed earlier, the fields of µA•R/I (Γ) take values in the minimal free resolution of Γ.
In §2.3.7, we already observed a close link between the supersymmetry module structure
and the resolution differential. We can now see this result as a consequence of the
equivalence of categories. Specifically, we can pull back µA•R/I (Γ) along the inclusion

{0} ,→ V . (3.90)

This restricts the multiplet to the fiber µA• (Γ)0 . The multiplet µA•R/I (Γ)0 carries a
module structure for the odd abelian super Lie algebra Πn1 . This module structure
coincides with the resolution differential in the following sense.

Corollary 3.4.8. Let Γ be an R/I-module and let (L, dL ) be its minimal free resolution
in R-modules. Let us identify the fiber

µA•R/I (Γ)0 = L ⊗R C. (3.91)

The map generated over R by the Πn1 -module structure on µA•R/I (Γ)0
M
ρconstants : µA•R/I (Γ)0 ⊗ ( n⊗k •
1 ) −→ µAR/I (Γ)0 (3.92)
k

coincides with the resolution differential. In coordinates, we can express this as

(k)
X
λ · ρconstants = ρconstants (λα1 dα1 , . . . , λαk dαk ) = dL , (3.93)
k

where dα is a basis for n1 .

Proof. By construction we have H • (n, µA• (Γ)) = Γ. But by (3.66) we know that there
is a quasi-isomorphism

C • (n, µA• (Γ)) ' (µA• (Γ)0 ⊗ R, λ · ρconstants ) . (3.94)

Thus, the cochain complex on the right is the minimal free resolution of Γ in R-modules
and we obtain the desired result.
Derived pure spinor superfields 113

In practice this means that the resolution differential contains all the information on the
supersymmetry transformations which are of order zero in the derivatives. This result
was conjectured by Berkovits in [Ber02].

Let us now state some results on the duality operations in the category of multiplets.

Corollary 3.4.9. Let (E, D, ρ) be a g-multiplet and let (E ∨ , D∨ , ρ∨ ) be the respective

dual (or antifield) multiplet. If these are both quasi-isomorphic to a multiplet in the image
of A•R/I and we have
(E, D, ρ) ' A•R/I (Γ) , (3.95)

then there is also a quasi-isomorphism

(E ∨ , D∨ , ρ∨ ) ' A•R/I (Extn−q

R (Γ, R)) . (3.96)

Here, n = dim(n1 ) and q = dimR (Γ).

Proof. Move to the minimal multiplet µA•R/I (Γ) ' (E, D, ρ). Corollary 3.4.8 implies
that
C • (n, E) ' C • (n, µA•R/I (Γ)) ' (L, dL ) , (3.97)

where (L, dL ) is the minimal free resolution of Γ in R-modules. There is a quasi-

isomorphism
 ∨
(E ∨ , D∨ , ρ∨ ) ' µA•R/I (Γ) . (3.98)

This in turn implies that

 
C • (n, E ∨ ) ' C • n, µA•R/I (Γ)∨ ' (L∨ , d∨
L) , (3.99)

where (L∨ , d∨
L ) is the dual of the minimal free resolution (L, dL ). By assumption the
derived invariants of (E ∨ , D∨ , ρ∨ ) are concentrated in a single degree. Hence, (L∨ , d∨
L)
n−q
again resolves a single R/I-module, namely ExtR (Γ, R).

In general, however, taking dual multiplets can lead outside of the image of A•R/I . In
fact, the above proof implies that (A•R/I (Γ))∨ is in the essential image of A•R/I precisely
when Γ is a Cohen–Macaulay R-module. In general this may not be the case—the dual
of the vector multiplet in four-dimensional N = 1 supersymmetry is an example of this
type, that does not occur in the image of the underived pure spinor functor.

If Γ is not Cohen–Macaulay, we can still compute the derived invariants of the associated
multiplet to find

C • (n, A•R/I (Γ)∨ ) ' RHomR (Γ, R) ' Ext•R (Γ, R) . (3.100)
114 Derived pure spinor superfields

We can therefore deduce the following natural description for the dual of a multiplet
obtained using the (underived) pure spinor formalism.

Corollary 3.4.10. Let Γ be an R/I-module and µA•R/I (Γ) the associated component
field multiplet. Then there is a C • (n)-module structure on the Ext-algebra Ext•R (Γ, R)
such that
µA•R/I (Γ)∨ ' A• (Ext•R (Γ, R)) . (3.101)

3.4.4 First examples

Let us discuss the pure spinor formalism as an equivalence of categories in a few simple
examples.

Example 3.4.11. First, let us consider the situation where there is no supersymmetry at
all. Let V be an d-dimensional vector space. Let g denote its Poincaré Lie algebra of
infinitesimal isometries, and let n denote its Lie algebra of translations, viewed as a dg
Lie algebra concentrated in degree 2. So C • (n) ∼ = Sym• (n∨ [1]) is an exterior algebra
in d generators v1 , . . . , vd of degree −1. We are, therefore, interested in g0 ∼
= so(d)-
equivariant dg-modules over the exterior algebra. We restrict attention to those modules
with finite-dimensional cohomology in each degree.

On the other hand, the dg-category Multg of multiplets is nothing but the dg-category of
Poincaré equivariant dg-vector bundles on the affine space V : the g action is completely
determined (and all multiplets are automatically strict). Again, let us restrict attention
to bundles with finite-dimensional cohomology in each degree. Such multiplets are de-
termined by their restriction to a formal neighborhood of the origin, which is a Poincaré
equivariant dg-module over the completed Weyl algebra

D̂d = C[∂1 , . . . , ∂d ][[z1 , . . . , zd ]]/([∂i , zi ] − 1). (3.102)

Translation equivariance guarantees that all such modules are induced from so(d)-equivariant
modules over C[∂1 , . . . , ∂d ]. Our statement then reduces to ordinary Koszul duality as a
relationship between modules over an exterior and a commutative algebra.

Example 3.4.12. Let Σ be a finite-dimensional vector space. Let us now briefly discuss
the example where g = n = R ⊕ ΠΣ as a graded vector space, with Lie bracket given by
a non-degenerate inner product Σ⊗2 → R. This is the background for supersymmetric
classical mechanics. Now

C • (n) ∼
= (C[v, λ1 , . . . , λN ] , dCE ), (3.103)
Derived pure spinor superfields 115

where v is an odd generator, λi are even generators, dCE λi = 0 for all i, and dCE v =
(λ1 )2 + · · · + (λN )2 (so the λi are linearly dual to a choice of orthonormal basis of Σ).

Let (Γ, dΓ ) be a C • (n)-module. The derived version of the pure spinor formalism asso-
ciates to Γ the multiplet
 
∂ ∂
A• (Γ) = C ∞ (Rt ) ⊗ C[θ1 , . . . , θN ] ⊗ Γ , dΓ + (v − λa θa ) + λa a , (3.104)
∂t ∂θ

where θ1 , . . . , θN are, again, odd generators, and the expression λa θa implicitly uses the
choice of inner product on Σ.

On the other hand, we can study multiplets on R for the super Lie algebra n directly.
These are affine dgs-vector bundles on R equipped with N commuting odd symmetries,
each of which squares to the action of translation. One can see this structure more
concretely from the expression of equation (3.104) by applying homotopy transfer to
take the cohomology with respect to the final summand λa ∂θ∂a of the differential.

There is a rich theory underlying the classification of multiplets in supersymmetric me-

chanics; see for example [FG05]. It would be interesting to investigate the connections
between this existing work and the point of view described here.

3.5 Applications

In this final section, we discuss some examples of multiplets in the light of the derived
formalism. This also provides some insight to some of the curiosities of the underived
pure spinor superfield formalism. In particular, we connect the multiplets associated
to different Lie algebra cohomology groups by curvature maps and construct antifield
multiplets for four-dimensional N = 1 supersymmetry. In addition, we show how both
the on- and off-shell version of the chiral multiplet can be constructed.

3.5.1 Multiplets from Lie algebra cohomology

We noticed in §3.3.3 that the multiplet associated to C • (n) can be identified with the de
Rham complex of the super Lie group N . On the other hand, it was already appreciated
in the previous literature that the individual Lie algebra cohomology groups of n give
an interesting class of R/I-modules which yield multiplets via the underived pure spinor
construction A•R/I . We studied examples for various super Poincaré algebras in §2.6.3;
see also [MSX12]. The derived formalism gives a new tool to study these multiplets and
116 Derived pure spinor superfields

to highlight the relations among the multiplets associated to the different Lie algebra
cohomology groups.

To start with, recall that when the nilpotence variety Spec(R/I) associated to n is a
complete intersection, the Chevalley–Eilenberg cohomology of n is concentrated in degree
zero. The associated multiplet to H 0 (n) is then simply quasi-isomorphic to the de Rham
complex on the supertranslation group Ω• (N ).

In contrast, for cases with higher Lie algebra cohomology, there is no direct quasi-
isomorphism between the multiplets associated to C • (n) and H • (n). However—as ad-
vertised in §3.3.4—we can use the derived formalism to study the multiplets associated
to the individual Lie algebra cohomology groups and their relationships. As discussed,
there is a quasi-isomorphism identifying the multiplet associated to C • (n) with the de
Rham complex on spacetime,

A• (C • (n)) ' Ω• (N ) . (3.105)

On the other hand, we can first take cohomology with respect to the Chevalley–Eilenberg
differential. Let us filter the complex A• (C • (n)) by the total degree on the ring C • (n):
the internal Z-degree plus the weight: the Chevalley–Eilenberg cohomology has degree
one for this filtration, and the remaining piece of the differential has degree zero. The
homotopy transfer theorem leads to an equivalence
!
M  
A• (C • (n)) ' A• H k (n) , D0 , (3.106)
k

where the differential D0 on the right hand side is induced by homotopy transfer, and
consists of a sum of terms
   
0
Dj,k : A• H k (n) → A• H k−j (n) (3.107)

for j ≥ 1. An alternative way of understanding these differentials is to consider higher

differentials in the spectral sequence associated to the filtration discussed above, by total
degree on C • (n).

For example, let us consider the case j = 1. The process we have just discussed gives
rise to differential operators between the associated component field multiplets:

∇ : µA• (H k (n)) −→ µA• (H k−1 (n)) . (3.108)

As we will see in an example momentarily, these operators can intuitively be thought of

in the case k = 0 by viewing µA• (H −1 ) as the field strength multiplet of µA• (H 0 ), with
Derived pure spinor superfields 117

∇ acting as the field strength or curvature map. In particular, the multiplet associated
to H 0 always contains a summand corresponding to a p-form abelian gauge field for some
p. Since the deformation that we describe deforms the sum of the minimal multiplets
µA• (H k (n)) to (an object quasi-isomorphic to) the de Rham complex on spacetime,
one expects that the operator mapping µA• (H 0 ) to µA• (H −1 ) will include a de Rham
differential that carries the p-form to a (p + 1)-form “field strength” component field in
the multiplet µA• (H −1 ). In examples, we will see that this is the case.

We will see in §7 that these operators play a big role in the construction of interac-
tions. Broadly speaking, they provide a general coordinate-free description of operators
of the type “Ra ”, considered by Cederwall in constructing pure spinor superfield ac-
tions; see [Ced10b]. These operators appear, for example, in the discussion of interacting
eleven-dimensional supergravity in [Ced10c; Ced10a], and for the Dirac–Born–Infeld ac-
tion in [CK11].

Example 3.5.1. Consider N = 1 supersymmetry in three dimensions. Recall that the

supertranslation algebra is of the form

n = S(−1) ⊕ V (−2) (3.109)

where S is the two-dimensional spin representation of Spin(3) ∼

= SU(2), and V is the
three-dimensional vector representation of Spin(3), with the bracket being induced from
the equivariant isomorphism Sym2 (S) ∼ = V . We can choose bases for these representa-
tions to obtain generators and relations for the Chevalley–Eilenberg complex as follows:

C • (n) = C[λα , v µ ], dCE v 1 = (λ1 )2 , dCE v 2 = λ1 λ2 , dCE v 3 = (λ2 )2 . (3.110)

It will sometimes be convenient to write the differential more compactly as

dCE v (αβ) = λ(α λβ) . (3.111)

The Lie algebra cohomology is easily computed.




R/I if k = 0

k
(λ2 v 1 − λ1 v 2 )C[λ1 ] ⊕ (λ2 v 2 − λ1 v 3 )C[λ2 ] /I−1


H = if k = −1 (3.112)



0 otherwise.


The ideal I−1 appearing in the case where k = −1 is spanned by λ2 (λ2 v 1 − λ1 v 2 ) −

λ1 (λ2 v 2 − λ1 v 3 ). As explained in §2.3.9, µA• (H 0 ) is identified with the 3d N = 1 gauge
multiplet and µA• (H −1 ) with the corresponding antifield multiplet. The direct sum of
118 Derived pure spinor superfields

these multiplets takes the following form.

 
Ω0
d
 
 
 
 Ω1 S 
µA• (H 0 ) ⊕ µA• (H −1 ) =  (3.113)
 

 
 S Ω2 

 d 

Ω3

This direct sum is not yet isomorphic to the de Rham complex, but we can clearly see
how the differential can be deformed such that this is the case: one has to add an acyclic
piece which cancels the two spin representations and an additional differential of order
one forming the de Rham differential between Ω1 and Ω2 .

Let us now see how this deformation arises from the differential on A• (C • (n)) via homo-
topy transfer. We start by taking cohomology with respect to the Chevalley–Eilenberg
differential dCE and fix homotopy data

p
h (A• (C • (n)) , dCE ) (C ∞ (N ) ⊗ H • (n), 0) , (3.114)
i

such that we find induced differentials of the form D0 = pDi + pDhDi + . . . . By degree
reasons, only the first and the second summand can contribute non-zero terms. Taking
cohomology with respect to D0 and transferring to the minimal multiplets we finally
obtain the map
∇ : µA• (H 0 ) −→ µA• (H −1 ) , (3.115)

which we view as a field strength map. Let us now study this map explicitly using our
basis. Running the techniques described in §2.3.7 (see also [KL09] for an earlier account),
one finds the following representatives for the component fields in µA• (H 0 ).
" #
1 − −
. (3.116)
− λ(α θβ) λα θ2

For µA• (H −1 ) one finds

" #
λα v (αβ) λ(α θβ) v γδ −
, (3.117)
− − θ2 λα λβ v (αβ)

where now the rows indicate degree 2 and 3. We can fix a homotopy h for dCE by

h(λα λβ ) = v (αβ) . (3.118)

Derived pure spinor superfields 119

There will be terms in ∇ of the form

∂ ∂
pλ hλ i . (3.119)
∂θ ∂θ

Acting on representatives of the fermions in the multiplet µA• (H 0 ), we find

λα θ2 7→ λα λβ θβ 7→ v (αβ) θβ 7→ λβ v (αβ) . (3.120)

This is a map between µA• (H 0 ) and µA• (H −1 ), which induces an acyclic deformation
on their direct sum.

In addition, ∇ contains terms of order one given by

∂
pv i. (3.121)
∂x

Investigating the representatives, it is clear that this acts as a de Rham differential

between the one-form in µA• (H 0 ) and the two-form in µA• (H −1 ). This maps the one-
form gauge field present in µA• (H 0 ) to its field strength, which is part of µA• (H −1 ).

3.5.2 Antifield multiplets in four dimensions

Let us now come back to the multiplet which provided the original motivation for the
derived formalism. For four-dimensional N = 1 supersymmetry, µA• (R/I) can be iden-
tified with the vector multiplet. In §2.4.4, we argued that the corresponding antifield
multiplet, cannot be constructed in the underived pure spinor superfield formalism. Here,
we describe the antifield multiplet using component fields first and then calulate the de-
rived n-invariants. As expected we find that the cohomology is not concentrated in a
single degree such that the multiplet is not in the essential image of the functor A•R/I .

Let (E, D, ρ) denote the antifield multiplet to the vector multiplet on R4 . It is concen-
trated in weight 0 to 4 and cohomological degree 0 and 1, and can be concretely described
in array notation as
 
 Ω0 (R4 ) Γ(R4 , S+ ⊕ S− ) Ω1 (R4 ) 
µA• (R/I)∨ = 
 
?d?  . (3.122)
 
 
Ω0 (R4 )

We can also describe the L∞ action of the supertranslation algebra n concretely. We

will only need the action on constant sections. There, odd elements act by the following
120 Derived pure spinor superfields

formula:

(1)
ρconstants (Q) : S+ ⊕ S− −→ Ω0 , (ψ ∨ , ψ̄ ∨ ) 7→ Q+ ∧ ψ ∨ + Q− ∧ ψ̄ ∨
: Ω1 −→ S+ ⊕ S− , A∨ 7→ Q+ ∧ A∨ + Q− ∧ A∨
(2)
ρconstants (Q1 , Q2 ) : Ω0 −→ Ω1 , c∨ 7→ Γ(Q1 , Q2 ) ⊗ c∨ ,
(3.123)
where we have denoted the fields by (ψ ∨ , ψ̄ ∨ ) ∈ Γ(R4 , S+ ⊕ S− ), A∨ ∈ Ω1 (R4 ), and
c∨ ∈ Ω0 (R4 ) in degree one. We have similarly denoted the positive and negative helicity
summands of Q ∈ S+ ⊕ S− by Q+ and Q− respectively.

Let us write E0 for the fiber of E over 0 ∈ R4 (not to be confused with the summand of
degree zero), so concretely
 

E0 =  R S+ ⊕ S− R4 . (3.124)
R

We can use this to describe the Chevalley–Eilenberg complex of n with coefficients in

our multiplet, using the L∞ action of n on the sheaf E of sections of E. We find the
following.

Proposition 3.5.2.
C • (n, E) ∼
= (E0 ⊗ Sym(S+ ⊕ S− ), dρ ), (3.125)

where dρ is generated over Sym(S+ ⊕ S− ) by the sum of the terms

ρ(1) |constants : E0 ⊗ (S+ ⊕ S− ) → E0 (3.126)

ρ(2) |constants : E0 ⊗ Sym2 (S+ ⊕ S− ) → E0 (3.127)

obtained by restricting the L∞ action ρ to constant sections of E.

Proof. Note that

C • (n) ∼
= (Sym((R4 )∨ [1])) ⊗ Sym(S+ ⊕ S− ), dCE ), (3.128)

where we have identified S± with its dual using its canonical inner product. We obtain
the given description by taking the cohomology by the operator dual to the action of
the algebra of translations on E; the result is quasi-isomorphic to C • (n, E), no additional
homotopical correction terms appear.

Let us now discuss the cohomology of the Chevalley–Eilenberg complex.

Derived pure spinor superfields 121

Proposition 3.5.3. We have an isomorphism

H • (n, E) ∼
= C ⊕ (Sym(S+ ) ⊕ Sym(S− ))[−1]. (3.129)

Proof. This is a straightforward calculation using the description that we have just given.
In the weight zero term of E0 , all elements are dρ -closed, but all such elements other
than constants are also dρ -exact. In weight 1 in E0 , the dρ -closed elements are generated
over 1 ⊗ Sym(S+ ⊕ S− ) by ∧2 S+ ⊕ ∧2 S− ⊆ (S+ ⊕ S− ) ⊗ (S+ ⊕ S− ). When we quotient by
dρ -exact elements we are left with ∧2 S+ ⊗ Sym(S− ) ⊕ ∧2 S− ⊗ Sym(S+ ) . In weight

2 in E0 the closed and exact elements coincide, and in weight > 2 in E0 there are no
dρ -closed elements.

From our general results, we know abstractly that A• (C • (n, E)) is dual to the vector
multiplet. It is instructive to calculate the component field formulation for this multiplet
explicitly using the recipe presented in §3.3.4 in order to see how the is related to the
corresponding calculation in the underived pure spinor superfield formalism.

Our calculation will follow the same outline as the calculation we performed in the
previous section. We will first study the multiplet associated to each of the two coho-
mology groups of C • (n, E) in isolation, then we will compute the additional differential
in A• (H • (n, E)) relating these two individual terms, obtained by homotopy transfer.

First, recall that

A• (C) ' C ∞ (N )
(3.130)
' Γ(R4 , Sym(S+ [1] ⊕ S− [1])).

For the non-trivial summand of the cohomology, we can compute that

A• (Sym(S+ )) ' (Γ(R4 , Sym(S+ ) ⊕ Sym(S+ [1] ⊕ S− [1])), D)

(3.131)
' Γ(R4 , Sym(S− [1]),

where the differential D is generated by the degree one isomorphism S+ [1] → S+ . Simi-
larly
A• (Sym(S+ )) ' Γ(R4 , Sym(S+ [1])). (3.132)

So altogether, when we compute A• (H • (n, E)), we obtain a multiplet with the following
Betti numbers: " #
1 4 6 4 1
. (3.133)
- 2 4 2 -
122 Derived pure spinor superfields

Now, let us compute the correction terms that allow us to obtain A• (C • (n, E)) in full.
As discussed in the previous section, §3.5.1, there is an additional differential coming
from homotopy transfer. We will compute this differential in coordinates.

Running the procedure described in §2.3.7, we find the following local coordinates for
our multiplet:
(θα , θ̄α̇ ) (θ2 , θα θ̄α̇ , θ̄2 ) (θ2 θα , θ̄2 θ̄α̇ ) θ2 θ¯2
" #
1
. (3.134)
− (λs, λ̄s̄) (λsθα , λ̄s̄θ̄α̇ ) (λsθ2 , λ̄s̄θ̄2 ) −

Let us unpack the notation. Recall that we identified C • (n, E) with (E0 ⊗ Sym(S+ ⊕
S− ), dρ ). We use {sα , s̄α̇ } for a basis of S+ ⊕ S− in the first factor, and {λα , λ̄α̇ } for a
basis of S+ ⊕ S− in the second factor, and e.g. λs the obvious contraction. We then use
∨ ⊕ S ∨ ⊆ C ∞ (N ).
{θα , θ̄α̇ } for the linear odd coordinate functions S+ −

With this concrete basis in hand, let us now investigate the additional pieces in the
differential coming from homotopy transfer. For this purpose, let us fix a homotopy h
for the differential dρ . We have
dρ (sα ) = λα , (3.135)

and therefore we can set

h(λα ) = sα (3.136)

and similarly for λ̄ and s̄. There are two nontrivial terms contributing to the transferred
differential
D0 hD0 (θ2 ) = λs (3.137)

and again similarly for the complex conjugates. We see that this induces a differential
that cancels some of the representative basis elements in pairs. The remaining represen-
tatives are " #
1 (θ, θ̄) θθ̄ −
(3.138)
− − − λsθ2 + λ̄s̄θ̄2

It is immediate to see that these representatives span the Spin(4) representations occur-
ring in (3.122). In addition there is a differential of order one described by

D0 hD1 (θα θ̄β̇ Aαβ̇ ) = (λsθ2 + λ̄s̄θ̄2 )∂ µ Aµ . (3.139)

We thus recover the anticipated description of the vector antifield multiplet.

3.5.3 The chiral multiplet revisited

Let us discuss one further example in the derived formalism, which will illustrate the
relation between on- and off-shell multiplets within the formalism, i.e. the appearance
Derived pure spinor superfields 123

of non-trivial L∞ actions of the supersymmetry algebra. We will again work specifically

with four-dimensional N = 1 supersymmetry.

The minimal multiplet µA•R/I (Γ) associated to the module Γ = Sym(S+ ) is equivalent to
the BRST version of the chiral multiplet. Constructing the associated cotangent theory
yields the standard off-shell BV theory of the chiral multiplet whose component fields
include a scalar field φ, a chiral spinor field ψ, and an auxiliary scalar field F , as well as
their associated complex conjugates and antifields. Of course, one can integrate out the
auxiliary field F and obtain an equivalent BV theory, but the supersymmetry algebra
action is now no longer strict. This is referred to as the on-shell formulation of the chiral
multiplet. This is discussed in the L∞ -module language in [SW20], but the idea is much
older, for example, the related example of an N = 2 hypermultiplet is discussed in the
on-shell language in [Bau+90].

Let us again start from the component field formulation for these multiplets and compute
their derived n-invariants. Plugging in this module in the derived pure spinor formalism
we find that the minimal multiplet µA• (C • (n, E)) can be explicitly identified with the
on-shell formulation for the chiral multiplet described above. The off-shell formulation
including the auxiliary field is given by a quasi-isomorphic non-minimal multiplet.

The component fields of the chiral multiplet in the on-shell formulation take the following
form.
 
 Ω0 ⊗ C2 Ω0 ⊗ (S+ ⊕ S− ) 
∂/
 
E=
 ?d?d  (3.140)

 
Ω0 ⊗ (S+ ⊕ S− ) Ω0 ⊗ C2 .

In order to describe the n-action, we will denote the component fields by (φ, φ̄) in degree
zero, weight zero and (ψ, ψ̄) in degree zero, weight one and their respective antifields in
degree one by (φ∨ , φ̄∨ ) and (ψ ∨ , ψ̄ ∨ ). The odd elements of n act in the following way.

ρ(1) (Q) : S+ ⊕ S− −→ Ω0 ⊗ C2 , (ψ, ψ̄) 7→ p+ (Q) ∧ ψ + p− (Q) ∧ ψ̄)

: Ω0 ⊗ C2 −→ S+ ⊕ S− , (φ, φ̄) 7→ / + φ̄)
Q ∧ ∂(φ
ρ(2) (Q, Q) : S+ ⊕ S− −→ S+ ⊕ S− , (ψ + , ψ̄ + ) 7→ p− (Q) ⊗ p+ (Q) ∧ ψ ∨
+p+ (Q) ⊗ p− (Q) ∧ ψ̄ ∨
(3.141)

With this description we can compute the cohomology of the Chevalley–Eilenberg com-
plex. We will use the following notation. Let (e, ē) and (s, s̄) denote bases of C2 and
S+ ⊕ S− respectively. As in the previous section, let us use λα , λ̄α̇ to denote even ba-
sis elements for C • (n). One finds the following description for the Chevalley–Eilenberg
124 Derived pure spinor superfields

cohomology.

H • (n, E) ∼


= (Sym(S− )e ⊕ Sym(S+ )ē) + Sym(S+ )λs ⊕ Sym(S− )λ̄s̄ [−1]. (3.142)

Again, we see that the cohomology is not concentrated in a single degree.

We can calculate the multiplet associated to this module following a method similar
to the previous section, beginning with the multiplets associated to the summands of
the cohomology described above, and then computing the correction terms associated to
homotopy transfer. We obtain a quasi-isomorphic multiplet
!
M
• k 0 0
A (H (n, E)) , D , ρ , (3.143)
k

where the new differential D0 contains terms induced via homotopy transfer. Individ-
ually, both µA• (H 0 (n, E)) and µA• (H 1 (n, E)) contain the field content of a chiral and
an antichiral BRST multiplet. We have the following Betti numbers for the sum of the
multiplets induced from the individual cohomology groups:
" #
2 4 2 −
(3.144)
− 2 4 2

There are explicit elements representing the cohomology which take the forrm
" #
(e, ē) (θα e, θ̄α̇ ē) (θ2 e, θ̄2 , ē) −
. (3.145)
− (λs, λ̄s̄) (λsθ̄α̇ , λ̄s̄θ)α (λsθ̄2 , λ̄s̄θ2 )

Now let’s take a look at the induced differentials under homotopy transfer. For an explicit
homotopy h we can choose

h(λα e) = sα h(λ̄α̇ ē) = s̄α̇ . (3.146)

Applying this to the representatives, we find

pD0 hD0 i(F ) = pD0 hD0 (F θ2 e) = p((λs)F ) = F

/ = ∂ψ
pD1 hD0 i(ψ) = pD1 hD0 (ψθe) = p((λs)θ̄∂ψ) / (3.147)
pD1 hD1 i(φ) = pD1 hD1 (φe) = p((λs)θ̄2 ∂ 2 φ) = ∂ 2 φ .

Similar results hold for the complex conjugate fields. Thus, we find that there are
induced differentials of order zero, one and two. With respect to the weight grading,
these operators alter the weight grading by minus one, one and three respectively. We
Derived pure spinor superfields 125

can summarize the multiplet in the following diagram.

 
 Ω0 ⊗ C2 Ω0 ⊗ (S+ ⊕ S− ) Ω0 ⊗ C2 
 ∂/ id 

 ?d?d 
 (3.148)
 
Ω0 ⊗ C2 Ω0 ⊗ (S+ ⊕ S− ) Ω0 ⊗ C2 .

This is precisely the off-shell BV model for the chiral multiplet. Taking cohomology with
respect to the acyclic differential we obtain the minimal multiplet µA• (C • (n, E)) which
is precisely the original on-shell formulation that we started with.
Chapter 4

Six-dimensional supermultiplets
from bundles on projective spaces

4.1 Introduction

With the pure spinor superfield formalism and its derived generalization firmly estab-
lished, we turn our attention towards using the technique systematically in applications.
One natural possibility is to use the equivalence of categories to look for patterns as well
as possible classification results in the category of multiplets starting from the (derived)
algebraic geometry of C • (n)-modules. To get the best mileage out of the formalism, it is
natural to start in a setting where the category of C • (n)-modules, or at least the category
of equivariant sheaves on the nilpotence variety, is relatively easy to understand.

A promising candidate is N = (1, 0) supersymmetry in six dimensions, where the nilpo-

tence variety Y is the space of two-by-four matrices of rank one; as such, the corre-
sponding projective variety Y = Proj R/I is just P1 × P3 , sitting inside P7 via the Segre
embedding. There is a great abundance of geometrically interesting equivariant vector
bundles on Y. Taking the direct sum of the global sections of all twists of these bundles,
we obtain graded equivariant modules over the ring of functions on the nilpotence variety,
so that we can study the associated multiplets. In particular, we classify all multiplets
originating from line bundles over Y = P1 × P3 ; among others, this recovers the family of
so-called “O(n)-multiplets” studied in the literature [KNT17; KNT18b; GR+98; LTM12;
GIO87], and encompasses the vector multiplet and its antifield muliplet, as well as the
hypermultiplet. (These three examples have already been studied via the pure spinor
superfield formalism in [CN08; Ced18b].)

127
128 Six-dimensional supermultiplets from bundles on projective spaces

Roughly speaking, we provide a link between vector bundles on Y and g-multiplets

in two steps, by combining the connection between quasicoherent sheaves on Proj R/I
and graded R/I-modules (which is standard algebraic geometry) with the pure spinor
superfield construction. Concretely, we convert a sheaf on Y into a module by forming
its graded module of global sections (i.e. by taking the sum of the global sections of all
its twists). Equivariant vector bundles form a subcategory of equivariant quasi-coherent
sheaves; conversely, one can assign a sheaf on Y to each module, though it is important to
note that the two operations are not inverses in general. Following the results on twisting
pure spinor superfields in [SW21], we argue that modules whose associated sheaf is trivial
correspond to multiplets that are perturbatively trivial in every twist.

In turn, the category of graded equivariant R/I-modules sits as a subcategory inside

equivariant C • (n)-modules, which is equivalent to the category of multiplets. We can
summarize the situation with the following diagram:

Γ∗ A•
LineBundlesgY0 QCohgY0 ModgR/I
0
ModgC0• (n) Multg (4.1)
∼ C•

Since the inverse functor to the pure spinor superfield construction is given by taking
derived n-invariants, classifying all multiplets associated to line bundles thus amounts to
classifying all multiplets whose derived n-invariants are the graded global section module
of a line bundle on the projective nilpotence variety. An alternative characterization
of such multiplets is via their twists, which are necessarily holomorphic for minimal
supersymmetry in six dimensions; in keeping with the results of [SW21], one expects
that the holomorphic twist of such a multiplet is of rank one over Dolbeault forms on
the spacetime, and we verify this below.

In addition, in §4.4, we extend the results of §2 by relating the duality theory of multiplets
to sheaves on the nilpotence variety. To this end, we study the Cohen–Macaulay property
and prove that, in good situations, the antifield multiplet can be constructed using the
dualizing sheaf on the projective nilpotence variety.

Moving on, in §4.5 we develop some general methods regarding short exact sequences of
sheaves in the pure spinor superfield formalism. These can be used to tackle higher-rank
bundles; generalizations would allow for the construction of the multiplet associated to
any higher-rank bundle via a resolution into a chain complex of sums of line bundles,
though we do not pursue this in detail here. Our results show that the multiplet asso-
ciated to a nontrivial extension of two sheaves is a deformation of the direct sum of the
multiplets associated to each sheaf by a further differential, and we study such deforma-
tions explicitly at the level of component-field presentations of various multiplets.
Six-dimensional supermultiplets from bundles on projective spaces 129

In §4.6 and 4.7, we then use the Euler exact sequence, as well as the normal and conormal
bundle sequences, to explicitly construct the multiplets associated to the tangent bundle,
the normal bundle, and their duals. Several of these multiplets are of obvious physical
interest; in particular, we identify the supergravity multiplet with the conormal bundle,
and the gravitino multiplet with the pullback of the tangent bundle to the ambient space.

4.2 Preliminaries

4.2.1 Computational techniques for pure spinor superfields

We work in the setting established in the previous chapters, i.e. g is a Lie algebra of super
Poincaré type with super translation subalgebra n, corresponding nilpotence variety Y
and ring of functions R/I.

Recall that, given an R/I module Γ, a minimal component field formulation of the
multiplet A• (Γ) can be canonically constructed. As before, we denote the minimal
component field multiplet by µA• (Γ). The fields of this component field multiplet take
values in the Koszul homology of Γ, which can be computed by means of a minimal
free resolution (L, dL ) of Γ in free R-modules. When we want to refer to the underlying
vector bundle of µA• (Γ) — which is the associated bundle of the bigraded Lorentz
representation on H • (K • (Γ)) — we will write µA• (Γ)# . This notation will in general
apply to any multiplet and denote (a natural bigraded lift of) its underlying Z × Z/2-
graded vector bundle, considered without the data of the differential and the module
structure.

As usual, the minimal free resolution appearing in this context are bigraded, by cohomo-
logical degree and by the weight grading on R such that L• is nonpositively graded and
the differential dL has cohomological degree one and weight zero. We will often write
resolutions as
M
L• = Wk ⊗ R[k], (4.2)
k≥0

where Wk is the finite-dimensional weighted g0 -representation in which the generators

of L−k transform. Recall that Proposition 2.3.3 establishes an isomorphism of bigraded
g0 -modules between the generators of L• and the Koszul homology such that the fields
of the minimal multiplet take values in the representations Wk .

In §2.3.6, we described a technique to extract the representations appearing in the min-

imal multiplet from the Hilbert series by assigning suitable weights to the generators.
Here, we use a variant of the procedure by considering the equivariant Hilbert series as
130 Six-dimensional supermultiplets from bundles on projective spaces

a formal power series in the representation ring of g0 .1 Therefore, we define

∞
X
Hilb(Γ) = Γk tk ∈ Rep(g0 )[[t]]. (4.3)
k=0

We can then rewrite the Hilbert series in the form

"∞ #  
X X
Hilb(Γ) = Symd (n∨
1)t
d
⊗  (−1)` Wk` tk 
d=0 k,` (4.4)

= Hilb(R) · Hilb(χ(W • )).

Comparing coefficients order by order, one obtains a system of equations which allows
to identify χ(W • ), and thus (at least in favorable cases) W • itself:

χ(W • )0 = Γ0
χ(W • )1 = Γ1 − n∨ •
1 ⊗ χ(W )0
.. (4.5)
.
k
X
χ(W • )k = Γk − Symd (n∨ •
1 ) ⊗ χ(W )k−d .
d=1

This technique is used frequently in the work of Cederwall and collaborators, and we will
apply it in examples in what follows.

4.2.2 The projective nilpotence variety for six-dimensional N = 1

We already introduced the six-dimensional N = (1, 0) super Poincaré algebra and the
corresponding nilpotence variety in §2.4.3. Recall that there are exceptional isomor-
phisms so(6) ∼
= sl4 for the Lorentz symmetry and sp(1) ∼= sl2 for the R-symmetry such
that the nilpotence variety and all the bundles we consider on it carry an action by
sl2 × sl4 . Since the bracket of odd elements in the supertranslation algebra is given by
wedging on the spin representations and the symplectic form on the R-symmetry space
U = (C2 , ω), a supercharge Q ∈ n1 is square zero if and only if the rank of the associated
linear map (S+ )∨ −→ U is less or equal then one. In terms of coordinates this means
that the defining ideal I of the nilpotence variety is spanned by the 2 × 2 minors of the
matrix with entries λαi ,
!
λ11 λ21 λ31 λ41
. (4.6)
λ12 λ22 λ32 λ42
1
The representation ring of g0 is the free abelian group on the set of finite-dimensional irreducible
g0 -representations, with the multiplication induced by the tensor product of representations.
Six-dimensional supermultiplets from bundles on projective spaces 131

Accordingly, the nilpotence variety Y = Spec R/I can be thought of as the space of
rank one matrices inside M 2×4 (C). Its projective version Y = Proj R/I can be identified
with the product of two projective spaces via the Segre embedding. In more detail, the
square-zero supercharges are precisely those which can be written as

Q=ξ⊗r with ξ ∈ S+ , r ∈ U. (4.7)

Interpreting [r0 : r1 ] and [ξ0 : · · · : ξ3 ] as homogeneous coordinates on P1 and P3 respec-

tively identifies Y with the image of the Segre embedding

σ : P1 × P3 −→ P7 ([r0 : r1 ], [ξ0 : . . . ξ3 ]) 7→ [r0 ξ0 : · · · : r1 ξ3 ] . (4.8)

We can thus explore supermultiplets in six dimensions using the algebraic geometry of
projective spaces.

4.2.3 From sheaves on projective schemes to modules

Clearly, the R/I-modules serving as inputs for the pure spinor functor A• are closely
related to sheaves of OY -modules on Y : For any affine scheme X = Spec S there is an
equivalence of categories between quasi-coherent sheaves of OX -modules and S-modules.
Explicitly, this equivalence is given by taking global sections

QCohOX −→ ModS F 7→ Γ(Spec(S), F) , (4.9)

and conversely assigning

ModS −→ QCohOX M 7→ M̃ , (4.10)

where M̃ is defined by the requirement M̃ (Df ) = Mf for all f ∈ S.2 If S is graded, one
can think of the grading as defining a gl1 -action on Spec S; it is then possible to define an
equivalence between graded S-modules and quasicoherent sheaves of OX -modules on X
that are equivariant for rescalings.

One can thus always think of the input to the (underived) pure spinor superfield for-
malism geometrically as a (g0 ⊕ gl1 )-equivariant sheaf on the affine nilpotence variety.
It is tempting to ask if one can picture the situation using the geometry of sheaves
on Y = Proj R/I. Here, the situation is geometrically compelling, but a bit less unequiv-
ocal. From a graded S-module M , we can construct a quasi-coherent sheaf on Proj S by
setting M̃ (Df ) = (Mf )0 . By the definition of the Proj-construction we have S̃ = OProj S .
2
Here Df ⊆ Spec S denotes all prime ideals of S not containing f and Mf the localization of M at f .
132 Six-dimensional supermultiplets from bundles on projective spaces

The twisting sheaves are defined by

].
OProj S (n) = S(n) (4.11)

For a sheaf F on Proj S, we define the associated S-module to be

M
Γ∗ (F) = Γ(Proj S, F(n)) . (4.12)
n∈Z

We will call Γ∗ (F) the graded global section module of F. In general, these assignments
no longer give an equivalence of categories, but we can still use Γ∗ (−) to construct
large families of input data for the pure spinor superfield formalism from sheaves on
the projective version of the nilpotence variety. This is in particular useful in the case
of N = (1, 0) supersymmetry in six dimensions, since—as we explained above—the
projective version of the nilpotence variety can be identified with P1 × P3 and equivariant
sheaves on this space are very well understood geometrically.

What the projective perspective misses. Contrary to the affine case, the functors
∼ and Γ∗ do not yield an equivalence of categories. While it is true that

Γ^ ∼
∗ (F) = F (4.13)

for any quasicoherent sheaf F, it can happen that Γ∗ (M̃ ) is not isomorphic to the original
module M . Let us restrict to the case where S = R is a polynomial ring and M is a
finitely generated graded module. Consider the class C of modules M such that Mn = 0
for n large enough. One finds that these are precisely the modules which are in the kernel
of ∼. One has the following result:

Proposition 4.2.1 ([Ser55]). Let M be a graded S-module. Then

M̃ = 0 ⇐⇒ M ∈ C . (4.14)

For the pure spinor superfield formalism, this means that multiplets corresponding to
modules which are concentrated in finitely many degrees cannot be obtained from sheaves
on the projective nilpotence variety. One such example is the free superfield A• (C) which
is constructed from the trivial module C (thought of as the quotient of R by the maximal
ideal corresponding to the origin). The corresponding sheaf on the affine nilpotence
variety is the skyscraper sheaf with value C at the origin; the associated sheaf on the
projective nilpotence variety is trivial.
Six-dimensional supermultiplets from bundles on projective spaces 133

In general, such sheaves must have zero-dimensional support. The support of an equiv-
ariant sheaf must consist of a union of orbits of the P0 -action; since we only consider
sheaves that are equivariant for rescaling, the origin is the unique zero-dimensional orbit,
so that any module in the kernel of ∼ defines a sheaf supported entirely at the origin.

Remark 4.2.2. It is natural to wonder how conditions on the support of a sheaf translate
into properties of the corresponding multiplet. An intuitive answer is suggested by the
results of [SW21] on twisting in the pure spinor formalism. There, it was noted that
deforming a super Poincaré-type algebra by a square-zero supercharge commutes with
forming the pure spinor multiplet of the structure sheaf. When Y is smooth (as is the case
here), only holomorphic twists are available, and the computations in [SW21] imply that
the holomorphic twist of a given multiplet is freely generated over the Dolbeault complex
on spacetime by the stalk of the corresponding sheaf at the holomorphic supercharge.
We do not explain this in detail here, but will remark from time to time on the physical
interpretations of our results that it suggests.

In keeping with Remark 4.2.2, we expect that multiplets corresponding to sheaves in the
kernel of ∼ are precisely those that are perturbatively trivial in every possible twist. We
note that the free superfield falls into this class.

4.2.4 Some natural equivariant vector bundles

In the bulk of this chapter we are going to consider various vector bundles over the
nilpotence variety Y ∼
= P1 × P3 and construct the associated multiplets using the pure
spinor superfield formalism. For later reference and completeness, we now introduce the
bundles that will appear later on.

Line bundles. The product space geometry of the nilpotence variety Y ∼

= P1 × P3
makes it easy to describe all of its line bundles. Indeed, holomorphic line bundles are
classified up to isomorphism by the Picard group Pic(Y) = ∼ H 1 (Y, O∗ ), which can be
Y
easily computed using the exponential short exact sequence

0 ZP1 ×P3 OP1 ×P3 OP∗ 1 ×P3 0 (4.15)

and its related long exact sequence in cohomology. In particular, one finds the isomor-
phism Pic(P1 × P3 ) ∼= Z ⊕ Z, which tells that every line bundle on the product variety
P1 ×P3 arises from line bundles defined on its factors P1 and P3 . (Recall that Pic(Pn ) ∼
=Z
134 Six-dimensional supermultiplets from bundles on projective spaces

for any n ≥ 1.) In fact, given the structural projections

P1 × P3
π1 π3 (4.16)

P1 P3

from P1 × P3 to its cartesian components, the line bundles on P1 × P3 are all given by
the exterior tensor product of a pair line bundles defined over P1 and P3 respectively. In
other words,

OP1 ×P3 (n, m) = OP1 (n)  OP3 (m) = π1∗ OP1 (n) ⊗OP1 ×P3 π3∗ OP3 (m) , (n, m) ∈ Z⊕2 .
(4.17)
Note that the generators of the Picard group Pic(P1 × P3 ) are given by OP1 ×P3 (1, 0)
and OP1 ×P3 (0, 1); the connecting (iso)morphism δ2 : Pic(P1 × P3 ) → Z ⊕ Z thus carries
OP1 ×P3 (n, m) to (n, m), and tensor product yields an isomorphism (of OP1 ×P3 -modules)

OP1 ×P3 (n, m) ⊗ OP1 ×P3 (k, l) ∼

= OP1 ×P3 (n + k, m + l) (4.18)

for any (n, m), (k, l) ∈ Z⊕2 . We will often use the shorthand O(n, m) = OP1 ×P3 (n, m).
Finally, we will denote a k-twisting sheaf for the nilpotence variety Y by

OY (k) := OP1 ×P3 (k, k). (4.19)

Tangent and Cotangent Bundles. Similarly, tangent and cotangent bundles on

a product variety can be reconstructed by the tangent and cotangent bundles of its
Cartesian components. In fact, the tangent bundle of P1 × P3 is given by the exterior
direct sum
TP1 ×P3 ∼
= π1∗ TP1 ⊕ π3∗ TP3 =: TP1
TP3 . (4.20)

Note that TP1 is a line bundle and one has TP1 ∼

= OP1 (+2), while TP3 is an ample non-
decomposable vector bundle of rank three. The tangent bundle on any projective space
Pn sits in the Euler exact sequence

0 OP n OPn (+1) ⊗ Vn+1 T Pn 0, (4.21)

where Vn+1 is a (n + 1)-dimensional complex vector space that carries the fundamental
representation of sln+1 . The Euler exact sequence (4.21) is a short exact sequence of
sln+1 -equivariant sheaves; this will play a role in §4.6, when we will study the multiplet
associated to the tangent bundle TY of the nilpotence variety.
Six-dimensional supermultiplets from bundles on projective spaces 135

In a similar fashion, the cotangent bundle Ω1Y := HomOP1 ×P3 (TP1 ×P3 , OP1 ×P3 ) of the
nilpotence variety Y is given by the exterior direct sum

Ω1P1 ×P3 ∼
= π1∗ Ω1P1 ⊕ π3∗ Ω1P3 = Ω1P1
Ω1P3 , (4.22)

where now Ω1P1 ∼

= OP1 (−2). Note that, taking the dual of the Euler sequence (4.21), one
finds
0 Ω1Pn ∨
OPn (−1) ⊗ Vn+1 OPn 0, (4.23)

which in turn describes the cotangent bundle on any projective space Pn .

Normal and Conormal Bundles. Let us now consider Y via its Segre embedding
σ : Y ,→ P7 . (Recall that this embedding is canonically associated to the datum of the
supertranslation algebra n.) Having introduced the tangent bundle TY , one defines the
normal bundle NY/P7 of Y in P7 to be quotient bundle TP7 |Y /TY , where TP7 |Y := σ ∗ TP7 .
As such, the normal bundle sits in the exact sequence

dσ
0 TY TP7 |Y NY/P7 0, (4.24)

of vector bundles on Y, which will be referred to as the normal bundle exact sequence.
Dualizing (4.24), one obtains the exact sequence defining the conormal bundle:

∨ dσ ∨
0 NY/P 7 Ω1P7 |Y Ω1Y 0. (4.25)

∨
The conormal bundle NY/P ∨ :
7 is thus the kernel of the morphism of vector bundles dσ

∨
Ω1P7 |Y → Ω1Y . Another characterization of the conormal bundle NY/P 7 is possible using

the sheaf of ideals JY , which is defined as the kernel of the morphism of sheaves σ ] :
OP7 → σ∗ OY . In fact, there is a natural isomorphism of vector bundles on Y given by
σ ∗ (JY /JY2 ) ∼ ∨
= NY/P 7 . In the following, since no confusion regarding the ambient space

can arise, we will denote the normal and conormal bundles with respect to the Segre
embedding by NY and NY∨ .
136 Six-dimensional supermultiplets from bundles on projective spaces

4.3 A family of multiplets from line bundles

4.3.1 General procedure

Let us now classify all multiplets associated to the infinite family of line bundles O(n, m).
We will denote the multiplets by

µA• (n, m) := µA• (Γ∗ (O(n, m))). (4.26)

As a first observation, we note that the construction exhibits the following symmetry
under twists of line bundles:
M
Γ∗ (O(n + k, m + k)) = H 0 (O(n + k + d, m + k + d))
d∈Z
M
= H 0 (O(n + d, m + d))(k) (4.27)
d∈Z

= Γ∗ (O(n, m))(k).

This implies that the multiplets µA• (n, m) and µA• (n + k, m + k) agree up to a total
degree shift. Since the weight grading of a graded equivariant R/I-module becomes the
cohomological grading of the corresponding multiplet, we have that

µA• (n + k, m + k) = µA• Γ∗ (O(n, m))(k) = µA• Γ∗ (O(n, m)) [k] = µA• (n, m)[k].

(4.28)
It is thus sufficient to consider the line bundles O(n, 0) and O(0, m) for n, m ≥ 0.
(Equivalently, one could also consider the family O(n, 0) for n ∈ Z.)

We will identify the field content of the multiplets using the technique sketched above
in §4.2.1. We resum the equivariant Hilbert series, working in the ring of formal power
series with coefficients in the representation ring of sl2 × sl4 , and read off the equivariant
structure of the minimal free resolution from its numerator.

Recall that Γ∗ (O(n, m))d = C[x0 , x1 ]n+d ⊗ C[y0 , . . . , y3 ]m+d . The monomials of degree d
are the d-th symmetric power of the defining representation of the corresponding group
of linear transformations, so that we have

Γ∗ (O(n, m))d = [n + d|m + d, 0, 0]. (4.29)

Six-dimensional supermultiplets from bundles on projective spaces 137

in terms of Dynkin labels for sl2 × sl4 . Thus, the equivariant Hilbert series takes the
form
∞
X
Hilb(n, m) := Hilb(Γ∗ O(n, m)) = [n + d|m + d, 0, 0] td . (4.30)
d=−min(n,m)

Following §4.2.1, we rewrite the Hilbert series using the identity

Hilb(n, m) = Hilb(R) · Hilb χ(W • (n, m)) ,

(4.31)

and solve for χ(W • (n, m)). The equations (4.5) become

χ(W • (n, m))0 = [n|m, 0, 0]

χ(W • (n, m))1 = [n + 1|m + 1, 0, 0] − [1|1, 0, 0] ⊗ χ(W • (n, m))0
.. (4.32)
.
k
X
χ(W • (n, m))k = [n + k|m + k, 0, 0] − Symd ([1|1, 0, 0]) ⊗ χ(W • (n, m))k−d .
d=1

In what follows, we solve these equations case by case.

4.3.2 The bundles O(n, 0) for n ≥ 0

We begin with the case of the bundles O(n, 0) for nonnegative n. As we will see, these
bundles include the vector multiplet, its antifield multiplet, and the hypermultiplet, as
well as an infinite family of strict component-field multiplets associated to O(n, 0) with
n ≥ 3.

Computation of the Betti numbers. We specialize the Hilbert series (4.30) to the
case at hand. At the level of the graded dimension,

∞
X (d + 3)(d + 2)(d + 1) d
grdim(n, 0) = (n + d + 1) t , (4.33)
6
d=0

which can be rewritten as a derivative of a geometric series

∞
1 ∂ 3 3−n ∂ X d+n+1 1 ∂ 3 3−n ∂ tn+1
grdim(n, 0) = t t = t . (4.34)
6 ∂t3 ∂t 6 ∂t3 ∂t 1 − t
d=0

Performing the derivatives, the general result can be expressed in the following form.

(n + 1) − 4nt + 6(n − 1)t2 − 4(n − 2)t3 + (n − 3)t4

grdim(n, 0) = (4.35)
(1 − t)8
138 Six-dimensional supermultiplets from bundles on projective spaces

The coefficients of the numerator now correspond to the Betti numbers of the associated
multiplet.

Let us write out these Betti numbers concretely for all n. There are three special cases
when n ∈ {0, 1, 2}. For n = 0 one finds
" #
1 − − −
grdimµA• (0, 0)# = , (4.36)
− 6 8 3

which corresponds to the vector multiplet. For n = 1, we obtain

" #
• # 2 4 − −
grdimµA (1, 0) = , (4.37)
− − 4 2

which corresponds to the hypermultiplet. For n = 2, the result reads

" #
• # 3 8 6 −
grdimµA (2, 0) = , (4.38)
− − − 1

which corresponds to the antifield multiplet of the vector multiplet. Finally, for n ≥ 3,
the resulting Betti numbers take the general form
h i
grdimµA• (n, 0)# = n + 1 4n 6(n − 1) 4(n − 2) n − 3 . (4.39)

Equivariant decomposition. The above recursive relations are easily solved, either
by hand or with the help of a computer program such as LiE [LCL]. Let us again first
consider the three special cases where n ∈ {0, 1, 2}. For n = 0 we obtain

W0 = [0|0, 0, 0]
W1 = 0
W2 = −[0|0, 1, 0] (4.40)
W3 = [1|0, 0, 1]
W4 = −[2|0, 0, 0].

Thus the resulting multiplet takes the form

 
Ω0
µA• (0, 0)# =  (4.41)
 

Ω1 C2 ⊗ S− Ω0 ⊗ C3

where the three scalar fields live in the adjoint representation of the R-symmetry group.
(Here and in the following tables showing the field content of multiplets Cn will always
Six-dimensional supermultiplets from bundles on projective spaces 139

denote the unique irreducible n-dimensional representation of sl2 .) This corresponds to

the vector multiplet of six-dimensional N = (1, 0) supersymmetry. For n = 1, we find

W0 = [1|0, 0, 0]
W1 = −[0|1, 0, 0]
W2 = 0 (4.42)
W3 = −[0|0, 0, 1]
W4 = [1|0, 0, 0].

We can thus identify µA• (1, 0) as the hypermultiplet

 
Ω0 ⊗ C2 S+
µA• (1, 0)# = 
 


 (4.43)
S− Ω0 ⊗ C2

For n = 2

W0 = [2|0, 0, 0]
W1 = −[1|1, 0, 0]
W2 = −[0|0, 1, 0] (4.44)
W3 = 0
W4 = −[0|0, 0, 0].

The resulting multiplet µA• (2, 0) is the antifield multiplet of the vector multiplet.
 
Ω0 ⊗ C3 S+ ⊗ C2 Ω1
µA• (2, 0)# =  (4.45)
 

Ω0

Finally for n ≥ 3, the general form is

W0 = [n|0, 0, 0]
W1 = −[n − 1|1, 0, 0]
W2 = [n − 2|0, 1, 0] (4.46)
W3 = −[n − 3|0, 0, 1]
W4 = [n − 4|0, 0, 0].
140 Six-dimensional supermultiplets from bundles on projective spaces

Thus, µA• (n, 0) for n ≥ 3 are of the form

µA• (n, 0)# = (4.47)

h i
Cn+1 Cn ⊗ S+ Cn−1 ⊗ ∧2 S+ Cn−2 ⊗ ∧3 S+ Cn−3 ⊗ ∧4 S+

This family of multiplets was described in the physics literature under the name O(n)-
multiplets [KNT17; KNT18b; GR+98; LTM12; GIO87].

Supersymmetry module structure and interpretation. Given our results so far,

it is easy to give an explicit description of the module Γ∗ (O(n, 0)) as a cokernel of a
map between free R-modules as well as to describe their minimal free resolutions in R-
modules. We already discussed the cases n = 0 and n = 1 discussed in §2. For n ≥ 1,
we are looking for a map

ϕn : Cn ⊗ S+ ⊗ R −→ Cn+1 ⊗ R (4.48)

which is linear in λαi and equivariant under sl2 ×sl4 . Up to a non-zero constant prefactor,
there is a unique such map which can be described in components by

F 7→ λα(in Fi1 ...in−1 ) α . (4.49)

Resolving Γ∗ (O(n, 0)) = coker(ϕn ) one recovers the field content of the multiplets de-
scribed above. In addition, the resolution differential encodes the part of the g-module
structure acting by differential operators of degree zero.

Let us describe the minimal free resolution and the module structure for the cases n ≥ 3.
This will provide an intuitive interpretation of µA• (n, 0): it is a multiplet whose observ-
ables are generated by the degree-n monomials in the observables of the O(1, 0)-multiplet,
i.e. the hypermultiplet. One can thus imagine that the fields of the hypermultiplet map
to the fields of the O(n, 0) multiplet via a n-fold covering, dual to the inclusion map on
observables.

In components, the resolution differential

(dL )i : Cn−i ⊗ ∧i S+ ⊗ R −→ Cn−i+1 ⊗ ∧i−1 S+ ⊗ R i = 1...4 (4.50)

is described by contracting along S+ and symmetrizing along the sl2 -representation, for
example
[(dL )2 F ]i1 ...in−1 α = λβ(in−1 Fi1 ...in−2 ) αβ . (4.51)
Six-dimensional supermultiplets from bundles on projective spaces 141

This translates into supersymmetry transformation rules of the form

δFi1 ...im = α(in Fi1 ...in−1)α . (4.52)

Recall that we identified the µA• (1, 0) as the hypermultiplet. Let us denote the linear
observables in physical fields of the hypermultiplet by φi and ψα . This suggests to identify
the linear observables of the O(n, 0)-multiplet as polynomials of degree n in the linear
observables of the hypermultiplet, as follows:

Fi1 ...in = φi1 . . . φin

Fi1 ...in−1 α = φi1 . . . φin−1 ψα
.. (4.53)
.
Fi1 ...in−4 αβγδ = φi1 . . . φin−4 ψα ψβ ψγ ψδ .

Further, recall that for the hypermultiplet the module structure of the supersymmetry
algebra contains terms of the form

δφi = αi ψα . (4.54)

By the Leibniz rule, this precisely induces the supersymmetry transformations of the
O(n, 0)-multiplet we recorded above. Thus, we can view, for n ≥ 3, µA• (n, 0) as con-
sisting of polynomials of degree n in the linear observables of µA• (1, 0). Intuitively, this
can be viewed as a remnant of the statement O(n, 0) = O(1, 0)⊗n after applying the
pure spinor superfield formalism. We remark that a special case of this is already visible
in the action for supersymmetric Yang–Mills theory coupled to hypermultiplets studied
in [Ced18b]. There, an action is written that reproduces the minimal coupling of the
gauge sector to matter; the relevant term is cubic, containing two hypermultiplets and
one gauge field. From our perspective, this makes use of the identification of the O(2, 0)
multiplet both as the dual to the vector multiplet and as governing quadratic functionals
on the hypermultiplet.

It is straightforward to compute the holomorphic twist of these multiplets uniformly

for n ≥ 3, and we sketch this briefly here. Following Remark 4.2.2, we expect to find
that the twist is of rank one over Dolbeault forms on C3 . Choosing a holomorphic
supercharge fixes a complex structure on R6 and a polarization of the R-symmetry space
(or equivalently a choice of Cartan subalgebra of sl2 .) We decompose

S+ ∼
= C ⊕ V, S− ∼
=V∨⊕C (4.55)

as sl3 -representations. Here V = V3 is the three-dimensional fundamental representation

142 Six-dimensional supermultiplets from bundles on projective spaces

of sl3 . Using the non-derivative supersymmetry transformations indicated above, we see

that the complex of fields takes the following form.

C
C V
C V V∨
C V V∨ C
C V V∨ C (4.56)

C V V∨ C
C V V∨
C V
C

Here, the vertical axis represents the sl2 -weight with respect to the fixed Cartan; we
have drawn the diagram for n = 4, but the pattern is clear. For each n, the surviving
fields are precisely isomorphic to Dolbeault forms on C3 , and the remaining (derivative-
dependent) components of the holomorphic supercharge generate the ∂ operator. Due to
the twisting homomorphism, the twist naturally resolves holomorphic sections of K n/2 :

O(n, 0)Q ' Ω0,• (C3 ) ⊗ K n/2 . (4.57)

(This clearly generalizes the results for n = 0, 1, and 2, which are well-known.)

4.3.3 The bundles O(0, m) for m ≥ 0

Computation of the Betti numbers. The Hilbert series specializes to

∞
X (m + d + 3)(m + d + 2)(m + d + 1) d
grdim(0, m) = (d + 1) t , (4.58)
6
d=0

which can be rewritten as

1 ∂ 1−m ∂ 3 tm+3
grdim(0, m) = t . (4.59)
6 ∂t ∂t3 1 − t
Six-dimensional supermultiplets from bundles on projective spaces 143

Again, we can bring the Hilbert series into a form such that we can read off the Betti
numbers of the associated multiplet.
"
1 11 2 n3 10 4
8
(1 + m + m + ) − (m3 + 5m2 + 6m)t − ( m3 + 10m2 − m − 8)t3
(1 − t) 6 6 3 3
#
5 9 m3 m 6
+ ( m3 + 5m2 − m − 3)t4 − (m3 + m2 − 2m)t5 + ( − )t
2 2 6 6
(4.60)

It is immediate to see that for m = 0 we recover the result from above. Let us in addition
give the Betti tables for some small values of m. For m = 1, we find
h i
grdimµA• (0, 1) = 4 12 12 4 . (4.61)

For n = 2 one obtains

h i
grdimµA• (0, 2) = 10 40 65 56 28 8 1 . (4.62)

Equivariant decomposition. Solving the equations (4.32) one finds the following
representations appearing in µA• (0, m).

W0 = [0|m, 0, 0]
W1 = −[1|m − 1, 1, 0]
W2 = [0|m − 2, 2, 0] + [2|m − 1, 0, 1]
W3 = −[1|m − 2, 1, 1] − [3|m − 1, 0, 0] (4.63)
W4 = [0|m − 2, 0, 2] + [2|m − 2, 1, 0]
W5 = −[1|m − 2, 0, 1]
W6 = [0|m − 2, 0, 0]

Presentation and equivariant resolution. We can describe the module Γ∗ (O(0, 1))
explicitly as the cokernel of a map of free R-modules

ψ1 : (∧2 S+ ⊗ C2 ) ⊗ R −→ S+ ⊗ R . (4.64)

For degree reasons, the map should be linear in λ. It is easy to check that there is, up
to non-zero constant prefactors, a unique such map explicitly given by

G 7→ λαi Gi[αβ] sβ . (4.65)

144 Six-dimensional supermultiplets from bundles on projective spaces

Here sβ denotes a basis of S+ . The modules Γ∗ (O(0, m)) are obtained by taking sym-
metric products. It can be checked explicitly (for example using a computer program
such as Macaulay2 [GS]) that the minimal free resolutions of these modules reproduce
the multiplets described above.

4.3.4 A classification result

The results above describe all multiplets for six-dimensional N = (1, 0) supersymmetry
which can be obtained from line bundles on the nilpotence variety P1 × P3 . Based on
the equivalence of categories between multiplets and C • (n)-modules developed in §3 this
can be viewed as a classification result as follows. Given an R/I-module Γ, the derived
n-invariants of the associated multiplet are concentrated in degree zero and we have

H • (n, µA• (Γ)) = Γ . (4.66)

Conversely, given a multiplet (E, D, ρ) such that its derived n-invariants are concentrated
in degree zero one can identify

µA• (C • (n, E)) ' (E, D, ρ) . (4.67)

Therefore we obtain the following theorem.

Theorem 4.3.1. The above multiplets classify, up to quasi-isomorphism, all multiplets

for six-dimensional N = (1, 0) supersymmetry such that H • (n, E) is the graded global
section module of a single line bundle on the projective nilpotence variety.

As remarked above (Remark 4.2.2), the interpretation of the input module as the Chevalley–
Eilenberg cohomology with coefficients in the multiplet provides an interesting conceptual
link to the twists of the multiplet involved. Twisting by a supercharge Q takes invariants
of the multiplet with respect to the abelian subalgebra spanned by that supercharge.
The cohomology groups H • (n, E) define a sheaf on the nilpotence variety which, by the
result of [SW21], encodes all the information on the twists of the original multiplet. In
fact, one expects that the twist by a square-zero supercharge Q ∈ Y is determined by
the stalk of that sheaf at Q.

In our example, we see that—as the derived invariants of all the multiplets above are line
bundles—the stalk at any point is isomorphic to OY,x . Our nilpotence variety Y = P1 ×P3
only has one stratum corresponding to the holomorphic twist. Correspondingly, as we
have seen above, the holomorphic twists of the above multiplets always have rank one
over Dolbeault forms on C3 .
Six-dimensional supermultiplets from bundles on projective spaces 145

This intuition makes many aspects of the physical behavior of the multiplets and their
twists manifest. For example, we can take any of the above multiplets and dimension-
ally reduce to a four-dimensional N = 2 multiplet. In this case, the nilpotence variety
is reducible and has three different components, one of which is the image of the six-
dimensional N = (1, 0) nilpotence variety under the dimensional reduction map [ESW21].
The other two components correspond to the Donaldson–Witten twist, which does not
descend from a square-zero supercharge in six dimensions. The Chevalley–Eilenberg co-
homology with coefficients in the dimensionally reduced multiplets is obtained by push-
ing forward along the inclusion Y6D ,→ Y4D . Clearly, any supercharge corresponding
to a Donaldson–Witten twist is outside of the support of the resulting sheaf, so that
the respective stalks are trivial. Following Remark 4.2.2, one thus expects that the
Donaldson–Witten twists of all multiplets arising by dimensional reduction are pertur-
batively trivial. We hope to give a more complete account of extensions of the methods
developed in [SW21] to general multiplets in future work.

4.4 Antifield multiplets and duality

4.4.1 General observations

Given any multiplet µA• (Γ), one may form the dual (or antifield) multiplet µA• (Γ)∨ by
dualizing the underlying vector bundle, the differential and the supersymmetry module
structure. Via the pure spinor superfield formalism, the operation of taking the antifield
multiplet corresponds, in good cases, to taking the dualizing module of the input module
Γ. We already recognized this in §2.4; here we explore this direction further and link it
to statements in terms of sheaves on the nilpotence variety.

Given a Cohen–Macaulay module Γ, the multiplet µA• (Γ) is described by the minimal
free resolution of Γ in R-modules. The antifield multiplet µA• (Γ) is described by the dual
of that minimal free resolution, which is, by definition, a minimal free resolution of the
dualizing module ExtrR (Γ, R). Therefore we can identify for Cohen–Macaulay modules
Γ,
µA• (Γ)∨ = µA• (ExtrR (Γ, R)) . (4.68)

If Γ is not Cohen–Macaulay, this is no longer true. Then the dual of the minimal free
resolution of Γ is no longer a resolution of a single module, but in fact a model for the
dualizing complex of Γ. Its cohomology is the Ext-algebra Ext•R (Γ, R).
146 Six-dimensional supermultiplets from bundles on projective spaces

The relation to sheaves. Here, we are interested in modules which arise from sheaves
on the nilpotence variety Y via Γ∗ . In this case, we can link the above statements on
duality to more geometric notions for sheaves on projective schemes.

Therefore, let us consider a Cohen–Macaulay projective scheme ι : X ,→ Pn of codi-

◦ .
mension r. In this setting the dualizing sheaf of X is a vector bundle, denoted by ωX
Explicitly, it can be defined in terms of the ambient projective space as

◦
ωX = ExtrOPn (ι∗ OX , ωPn ) . (4.69)

Let us further assume that Γ∗ (OX ) is Cohen–Macaulay as an R = Γ∗ (OPn )-module.

Then the following holds.

Proposition 4.4.1. Let F ∈ Coh(X) be a coherent sheaf on X and Γ∗ (F) is its associ-
ated R = Γ∗ (OPn )-module. Then there exists a natural isomorphism

ExtrR (Γ∗ (F), R) ∼ ◦

= Γ∗ HomOPn (ι∗ F, ωX )(n + 1), (4.70)

so that the following diagram is commutative

Γ∗
F Γ∗ (F)

◦ )
HomOPn (ι∗ −,ωX ExtrR (−,R) (4.71)

◦ ) Γ∗
HomOPn (ι∗ F, ωX ExtrR (Γ∗ (F), R).

Proof. One has

ExtrR (Γ∗ (F), R) ∼

= ExtrR (Γ∗ (F)(−n − 1) , R(−n − 1))
∼
= ExtrR (Γ∗ (F)(−n − 1) ⊗R Γ∗ (OX ) , R(−n − 1)) , (4.72)

where the second isomorphism follows from the fact that the sheaf F is supported on X
and Γ^ ∼
∗ (OX ) = OX . By derived hom-tensor adjunction [Huy06] one has

ExtrR (Γ∗ (F), R) ∼

= HomR (Γ∗ (F)(−n − 1) , ExtrR (Γ∗ (OX ), R(−n − 1)))
(4.73)
= HomR (Γ∗ (F)(−n − 1) , ExtrR (Γ∗ (OX ), Γ∗ (ωPn ))) ,

where we used that Γ∗ (ωPn ) = R(−n − 1) in the second step. Notice that, by assumption
Γ∗ (OX ) is Cohen–Macaulay as an R-module, hence the only non-zero Ext-module in the
derived hom-tensor adjunction is indeed the dualizing module ExtrR (Γ∗ (OX ), R(−n − 1)).
Six-dimensional supermultiplets from bundles on projective spaces 147

Further, note that HomR in (4.73) denotes graded morphisms of all degrees. For mor-
phisms of degree zero, we have the adjunction [Vak; Har77]

Homdeg=0
R (M, Γ∗ (H)) = HomPn (M̃ , H) (4.74)

between the functors Γ∗ : QCohOX → ModR and (−)

g : ModR → QCohO . Upon using
X
^
Γ∗ (G) = G, this implies that

Homdeg=0
R (Γ∗ (G), Γ∗ (H)) = HomPn (G, H) . (4.75)

Shifting and summing on both sides, one reconstructs the graded morphisms:
M M
HomR (Γ∗ (G), Γ∗ (H)) = HomPn (G, H(k)) = Γ ◦ (HomPn (G, H(k)))
k∈Z k∈Z
M
= Γ ◦ (HomPn (G, H) ⊗ OPn (k)) (4.76)
k∈Z

= Γ∗ (HomPn (G, H)),

where we have used that Γ ◦ HomPn = HomPn , where Γ is the global section functor.
Deriving the above functors, one gets a local-to-global spectral sequence which, in our
case, yields the isomorphism

ExtrR (Γ∗ (OX ), Γ∗ (ωPn )) ∼

= Γ∗ (ExtrPn (i∗ OX , ωPn )) . (4.77)

Plugging this into (4.73), we finally obtain

ExtrR (Γ∗ (F), R) ∼

= Γ∗ (HomPn (F(−n − 1), ExtrPn (i∗ OX , ωPn )))
∼
= Γ∗ (HomPn (F, ExtrPn (i∗ OX , ωPn ))(n + 1)) (4.78)
∼ ◦
= Γ∗ (HomPn (F, ωX )(n + 1)) ,

which concludes the proof.

This result establishes that the dualizing module of Γ∗ (F) arises geometrically from the
◦ )(n + 1). The above proposition, together with (4.68), implies the
sheaf HomOPn (ι∗ F, ωX
following corollary.

Corollary 4.4.2. If Γ∗ (F) is a Cohen–Macaulay R-module, then we have

µA• (Γ∗ (F))∨ ∼

= µA• (Γ∗ (HomOPn (ι∗ F, ωX
◦
))) . (4.79)

It is possible to prove whether or not a sheaf F gives rise to a Cohen–Macaulay module

via Γ∗ by studying its sheaf cohomology. In particular the following result holds [Kol13].
148 Six-dimensional supermultiplets from bundles on projective spaces

Proposition 4.4.3. Let X be a Cohen-Macaulay projective scheme and let L be an ample

line bundle on it. Given a coherent sheaf F on X, then Γ∗ (F) is a Cohen-Macaulay R-
module if and only if H i (X, F ⊗ L⊗k ) = 0 for any 0 < i < dim(X) for any k ∈ Z.

4.4.2 Duality and line bundles

As a case study for the general results above, let us consider the multiplets µA• (n, 0) for
n ∈ Z. For a start, recall that the field content of the multiplet µA• (n, 0) takes values
in the minimal free resolution of the R-module Γ∗ (O( n, 0)). Therefore, given the results
in the previous section, we can easily read off lR (Γ∗ (O( n, 0))):



3 for n ∈ {−1, 0, 1, 2, 3}

lR (Γ∗ (O(n, 0))) = 4 for n > 3 (4.80)



6 for n < −2.


Notice that all the modules Γ∗ (O(n, 0)) come from line bundles supported on the nilpo-
tence variety Y ∼= P1 × P3 ⊂ P7 , which is of codimension 3 in P7 . From this, we can infer
the following lemma.

Lemma 4.4.4. The R-module Γ∗ (O(n, 0)) is Cohen–Macaulay if and only if n ∈ {−1, 0, 1, 2, 3}.

Therefore, for n in this range, we have

µA• (n, 0)∨ ∼

= µA• (Ext3R (Γ∗ (O(n, 0)), R)) . (4.81)

Lemma 4.4.4 can also be proved directly by studying the sheaf cohomology of the line
bundles O(n, m) and using Proposition 4.4.3. Indeed, we can choose L = O(1, 1) as an
ample line bundle and use the Künneth theorem to verify that the middle cohomologies
H i (Y, O(n + k, k)) vanish for i = 1, 2, 3 and for all k precisely when n ∈ {−1, 0, 1, 2, 3}.

Furthermore, since Y = P1 × P3 , the dualizing sheaf can be described explicitly as the

exterior tensor product of the respective dualizing sheaves on the factors. Explicitly,

ωY◦ = π1∗ ωP1 ⊗ π3∗ ωP3 = O(−2, −4) . (4.82)

Using this together with Theorem 4.4.1 gives

Ext3R (Γ∗ (O(n, 0)), R) = Γ∗ (O(n, 0)∨ ⊗ O(−2, −4))

= Γ∗ (O(−n − 2, −4)) (4.83)
= Γ∗ (O(2 − n, 0))(4) .
Six-dimensional supermultiplets from bundles on projective spaces 149

Thus, we obtain

µA• (2 − n, 0)[4] = µA• Ext3R (Γ∗ (O(n, 0)), R) .

(4.84)

In the range where Γ∗ (O(n, 0)) is Cohen–Macaulay, this implies

µA• (n, 0)∨ ∼

= µA• (2 − n, 0)[4] . (4.85)

This can be viewed as a remnant of Serre duality for line bundles on the multiplet side.

4.5 Short exact sequences

In this section, we discuss some general conclusions that can be drawn about short exact
sequences of vector bundles in the context of the pure spinor superfield formalism, and
then move on to study some concrete examples in our six-dimensional setting. The sec-
tions that follow will study the multiplets associated to the tangent and normal bundles
and their duals, and apply these results in the context of natural short exact sequences
in which those vector bundles appear. As such, we are motivated both by abstract
considerations—having understood that the pure spinor construction is a functor, it is
natural to ask about it not just on single objects, but on diagrams of objects—and, as
throughout this thesis, by concrete computational examples.

4.5.1 General observations

Let
0 −→ Γ0 −→ Γ −→ Γ00 −→ 0 (4.86)

be a short exact sequence of graded equivariant R/I-modules. Applying A• is an exact

functor; we thus obtain a short exact sequence of strict multiplets

0 −→ A• (Γ0 ) −→ A• (Γ) −→ A• (Γ00 ) −→ 0. (4.87)

Up to perturbative equivalence, this is the end of the story. However, we are often inter-
ested in component-field descriptions, and therefore specifically in the minimal multiplets
µA• (Γ), µA• (Γ0 ), and µA• (Γ00 ). To investigate the relationship at this level, we note the
following: Each homogeneous degree Γk is a finite-dimensional representation of sl2 × sl4 ;
restricting the above short exact sequence to degree k gives a short exact sequence of
sl2 × sl4 -representations.

0 −→ Γ0k −→ Γk −→ Γ00k −→ 0 (4.88)

150 Six-dimensional supermultiplets from bundles on projective spaces

Since sl2 × sl4 is semisimple, all finite-dimensional representations are completely decom-
posable, and the sequence splits for all k ∈ Z. We thus have

Γk ∼
= Γ0k ⊕ Γ00k (4.89)

as sl2 × sl4 -representations. This implies for the equivariant Hilbert series,

Hilb(Γ) = Hilb(Γ0 ) + Hilb(Γ00 ) (4.90)

and thus for the field content of the respective multiplets

χ(WΓ• ) = χ(WΓ•0 ) + χ(WΓ•00 ) . (4.91)

In practical terms this means that the direct sum of µA• (Γ0 ) and µA• (Γ00 ) admits a
deformation to µA• (Γ),

Deform  • 0 Deform
µA• (Γ) = µA• (Γ0 ) ⊕ µA• (Γ00 ) = µA (Γ ⊕ Γ00 )

. (4.92)

Recall that the differential is given by the right action together with the module structure
on Γ,
D = λα R(Qα ) . (4.93)

Thus, the deformation on the direct sum of the multiplets precisely corresponds to a
deformation of the module structure on Γ0 ⊕ Γ00 such that

Deform
∼
 0
Γ ⊕ Γ00 =Γ. (4.94)

This deformation of the module structure arises from the class of Γ inside Ext1 (Γ0 , Γ00 ).
Even more explicitly, we notice that Γ0 sits inside Γ as a submodule; therefore the
deformation of the module structure is characterized by a map

R/I × Γ00 −→ Γ0 . (4.95)

We can summarize these findings by the following lemma.

Lemma 4.5.1. Let 0 → Γ0 → Γ → Γ00 −→ 0 be a short exact sequence of graded

equivariant R/I-modules. Then the deformation of the module structure on Γ0 ⊕ Γ00
determined by this sequence induces a deformation on the respective multiplets

Deform
µA• (Γ) ∼
= µA• (Γ0 ) ⊕ µA• (Γ00 )

. (4.96)
Six-dimensional supermultiplets from bundles on projective spaces 151

In this chapter we often deal with short exact sequences of equivariant sheaves on Y. Let

0 −→ F 0 −→ F −→ F 00 −→ 0 (4.97)

be such a sequence. First, we observe that taking the tensor product with a line bundle
keeps the sequence exact, thus we obtain short exact sequences

0 −→ F 0 (k) −→ F(k) −→ F 00 (k) −→ 0 (4.98)

for all k ∈ Z. Second, a short exact sequence induces a long exact sequence in cohomology

δ
0 −→ H 0 (F 0 (k)) −→ H 0 (F(k)) −→ H 0 (F 00 (k)) −→ H 1 (F 0 (k)) −→ . . . (4.99)

Thus, if the map δ vanishes (for example due to H 1 (F 0 (k)) being zero) for all k, we
obtain a short exact sequence on the global sections

0 −→ H 0 (F 0 (k)) −→ H 0 (F(k)) −→ H 0 (F 00 (k)) −→ 0 , (4.100)

and therefore a short exact sequence of graded equivariant R/I-modules

0 −→ Γ∗ (F 0 ) −→ Γ∗ (F) −→ Γ∗ (F 00 ) −→ 0 . (4.101)

and we find ourselves in the situation described above.

Remark 4.5.2. It is worth recalling that extensions of sheaves are often interpreted as
related to interactions or bound states in mathematical physics. Just for example, in
topological string theory, B-branes are identified with coherent sheaves on the target
space, which is typically a Calabi–Yau threefold. As emphasized in early work on the
subject [Sha99; Dou01; Asp04, for example], a nontrivial extension sequence of the form

0→A→B→C→0 (4.102)

indicates that B should be thought of as a bound state of the branes A and C. (Making
this interpretation precise led to the identification of the category of B-branes with the
derived category of coherent sheaves.)

In our setting, as explained, the extension defines a deformation of the module structure,
which in turn deforms the differential on the multiplet. Thinking in the context of the
Batalin–Vilkovisky formalism, a deformation of the differential can in turn be thought of
as a deformation of the quadratic part of the BV action. As such, the new differentials
we consider on component fields can be interpreted, at least schematically, as (quadratic)
152 Six-dimensional supermultiplets from bundles on projective spaces

supersymmetric interactions between the multiplets µA• (Γ0 ) and µA• (Γ00 ), such that the
deformed multiplet has derived supertranslation invariants Γ.

4.5.2 Excursion: three-dimensional N = 1

Let us illustrate the findings from above in the case of three-dimensional N = 1 super-
symmetry. Thus, let R = C[λ1 , λ2 ] be the polynomial ring in two variables and

I = ((λ1 )2 , λ1 λ2 , (λ2 )2 ) = R≥2 . (4.103)

the defining ideal of the nilpotence variety

We are interested in the following short exact sequence

0 −→ S(−1) −→ R/I −→ R/R≥1 −→ 0 , (4.104)

where the first map is given by sending a basis sα to the generators λα and the sec-
ond map is the obvious projection. Note that the module structure on the direct sum
R/R≥1 ⊕ S(−1) is trivial; the deformation which makes it isomorphic to R/I is simply
given by
R/I × R/R≥1 −→ S(−1) (λα , 1) 7→ sα . (4.105)

Let us now study the associated multiplets. µA• (R/I) is a free superfield and µA• (S(−1))
is a free superfield with values in the spinor representation shifted to cohomological de-
gree 1. Their direct sum is described as follows.
 
 C S C 
µA• (S(−1))# ⊕ µA• (R/R≥1 )# =   (4.106)
S Ω1 ⊕C S

Using the procedure from §3.3.4, we find the following representatives for the component
fields " #
1 θα θ1 θ2
. (4.107)
sα θα sβ θ1 θ2 sα

Deforming the module structure by (4.105), induces a non-trivial differential. From

∂ ∂
D = λα α
+ λ(α θβ) (ab) , (4.108)
∂θ ∂x
Six-dimensional supermultiplets from bundles on projective spaces 153

and using the representatives, it is easy to see that the direct sum of multiplets is de-
formed to  
 C d
S C 
 , (4.109)
 
S Ω1 ⊕ C S

where every arrow directed down and left is an identity morphism. On the other hand,
the multiplet associated to R/I is the gauge multiplet.
 
Ω0
µA• (R/I) =  d (4.110)
 

Ω1 S

It is immediate to see that the above deformation is quasi-isomorphic to this multiplet.

4.5.3 The Euler sequence for P1

Let us now discuss a family of short exact sequences in six dimensions. Identifying
TP1 ∼
= OP1 (2), the Euler exact sequence for P1 reads

0 −→ OP1 −→ OP1 (1) ⊗ C2 −→ OP1 (2) −→ 0 . (4.111)

Note that this is a sequence of equivariant sheaves and that C2 carries the fundamental
representation of sl2 . Twisting by OP1 (n) and pulling back along π1 we obtain a family
of short exact sequences of equivariant sheaves on Y.

0 −→ O(n, 0) −→ O(n + 1, 0) ⊗ C2 −→ O(n + 2, 0) −→ 0 . (4.112)

Let us restrict to the case n ≥ 0 for the moment. Twisting by OY (k) = O(k, k) we obtain
the sequences

0 −→ O(n + k, k) −→ O(n + k + 1, k) ⊗ C2 −→ O(n + k + 2, k) −→ 0 . (4.113)

The relevant first cohomology group is H 1 (O(n + k, k)) which is easily seen to vanish for
all k ∈ Z by the Künneth theorem. Thus, we obtain for all n ≥ 0 a short exact sequence
of graded equivariant R/I-modules,

0 −→ Γ∗ (O(n, 0)) −→ Γ∗ (O(n + 1, 0)) ⊗ C2 −→ Γ∗ (O(n + 2, 0)) −→ 0 . (4.114)

Let us study the associated multiplets.

154 Six-dimensional supermultiplets from bundles on projective spaces

n = 0. Recall that µA• (0, 0) is the vector multiplet, µA• (1, 0) the hypermultiplet and
µA• (2, 0) = µA• (0, 0)∨ the antifield multiplet of the vector. Therefore, µA• (O(2, 0) ⊗
C2 ) = µA• (O(2, 0)) ⊗ C2 is a doublet of hypermultiplets with values in the fundamental
representation of the R-symmetry sl2 . Let us arrange the direct sum as follows.
 
 C ⊕ C3 S+ ⊗ C2 Ω1
µA• (0, 0)# ⊕ µA• (2, 0)# = 



 (4.115)
Ω1 S− ⊗ C2 C ⊕ C3

We can deform it by adding an acyclic differential relating the the two one-forms, the
Dirac operator for the fermions, and the Laplacian for the scalar fields.
 
 C⊕ C3 S+ ⊗ C2 Ω1
∂/

• • Deform
[µA (0, 0) ⊕ µA (2, 0)] =
 
?d?d 
 
id
Ω1 S− ⊗ C2 C ⊕ C3
(4.116)
Taking cohomology with respect to the acyclic part of the differential (i.e. integrating
out the auxiliary field) and recalling that for sl2 -representations C2 ⊗ C2 ∼
= C ⊕ C3 , we
immediately see that we recover the hypermultiplet with values in C2 .

Interestingly there is another BV theory which can be formed out of µA• (0, 0) and
µA• (0, 2). Adding both multiplets with an appropriate shift and deforming the resulting
complex one obtains the BV theory of the vector multiplet; we discussed this in §2.4.3.
Denoting the vector multiplet by E, these findings can be summarized by stating that the
cotangent theory T ∨ [−1]E corresponds to the BV theory describing the vector multiplet,
while the construction we presented above corresponds to (T ∨ [1]E)[−1] which is seen to
be equivalent to the hypermultiplet. Let us remark that all these considerations are
purely perturbative.

n = 1. Proceeding analogously, we can define a deformation on the direct sum of

µA• (1, 0) and µA• (3, 0) that renders it quasi-isomorphic to µA• (2, 0) ⊗ C2 :

[µA• (1, 0) ⊕ µA• (3, 0)]Deform =

 
 C2 ⊕ C4 S+ ⊗ (C ⊕ C3 ) ∧2 S+ ⊗ 2
Cid S− 



 (4.117)
?d?
S− C2

' µA• (2, 0) ⊗ C2 .

Here, we used the isomorphisms V ∼

= ∧2 S+ and S− ∼
= ∧3 S+ .
 Six-dimensional supermultiplets from bundles on projective spaces 155

n = 2. Similarly, there is a deformation of µA• (2, 0) ⊕ µA• (4, 0) giving µA• (3, 0) ⊗ C2 .

[µA• (2, 0) ⊕ µA• (4, 0)]Deform =

 
C3 ⊕ C5 S+ ⊗ (C2 ⊕ C4 ) ∧2 S + ⊗ (C ⊕ C3 ) ∧3 S + ⊗ C2 C 
(4.118)

 
id
C
' µA• (3, 0) ⊗ C2 .

n ≥ 3. For n ≥ 3 , we interpreted µA• (n, 0) as receiving an n-fold covering map from the
hypermultiplet, witnessed by the isomorphism between its observables and the subalgebra
of hypermultiplet observables with polynomial degree divisible by n. The short exact
sequence gives a relation between µA• (n + 1, 0) ⊗ C2 and µA• (n, 0) ⊕ µA• (n + 2, 0):

µA• (n, 0) ⊕ µA• (n + 2, 0) = (4.119)

h i
Cn+1 ⊕ Cn+3 (Cn ⊕ Cn+2 ) ⊗ S+ (Cn−1 ⊕ Cn+1 ) ⊗ ∧2 S+ (Cn−2 ⊕ Cn ) ⊗ ∧3 S+ Cn−3 ⊕ Cn−1

' µA• (n + 1, 0) ⊗ C2 .

Here, identifying µA• (n + 1) ⊗ C2 just amounts to the decomposition rule Cn ⊗ C2 ∼

=
Cn+1 ⊕ Cn−1 . In other words, considering pairs of hypermultiplet observables of poly-
nomial degree (n + 1) and regarding such pairs as transforming in the fundamental
representation of the R-symmetry sl2 , we can either symmetrize with respect to the R-
symmetry index (yielding µA• (n + 2, 0)) or antisymmetrize to land in µA• (n, 0). Note,
however, that the polynomial degree of the observables involved does not change; the
multiplet µA• (n, 0), rather than its realization via a map from the hypermultiplet, is the
fundamental object.

4.6 The normal bundle exact sequence

In the next two sections, we use short exact sequences of geometric bundles on P1 × P3
to extend our survey of multiplets to higher-rank bundles. We start with the tangent
and normal bundles, using the defining short exact sequence that relates them; in the
following section, we will consider the dual of this sequence and work out the multiplets
involved explicitly.

As recalled above, the tangent bundle of the projectivized nilpotence variety TY , the
restriction TP7 |Y of the tangent bundle of the ambient P7 to Y, and the normal bundle
156 Six-dimensional supermultiplets from bundles on projective spaces

NY sit in the normal bundle exact sequence

0 −→ TY −→ TP7 |Y −→ NY −→ 0 . (4.120)

Since H 1 (TY (k)) = 0 for all k ∈ Z, this short exact sequence induces a short exact
sequence on global sections. Thus, applying Γ∗ , we obtain a short exact sequence of
R/I-modules.
0 −→ Γ∗ (TY ) −→ Γ∗ (TP7 |Y ) −→ Γ∗ (NY ) −→ 0 (4.121)

We apply this short exact sequence to study the associated multiplets and their relations
to one another. Again, we will find that there is a deformation of µA• (TY ) ⊕ µA• (NY )
which is quasi-isomorphic to µA• (TP7 |Y ). (Here and in the following section, we often
will suppress the graded global section functor when we are talking about the associated
multiplets, i.e. for a sheaf F, we set A• (F) := A• (Γ∗ (F)).)

4.6.1 Tangent bundle

Cohomology and Hilbert Series. Recall that, as seen above, the tangent bundle to
the nilpotence variety is given by the exterior sum

TY = π1∗ TP1 ⊕ π3∗ TP3 = TP1
TP3 , (4.122)

where π ∗ TP1 ∼
= O(2, 0). Accordingly, the resulting multiplet will be given by a direct
sum,
µA• (TY ) = µA• (2, 0) ⊕ µA• (π3∗ TP3 ). (4.123)

We have already identified µA• (2, 0) as the antifield multiplet of the vector multiplet
in §4.3.2, so we are left with studying µA• (π3∗ TP3 ), which amounts to computing the
zeroth cohomology of

π3∗ TP3 (k) = π1∗ OP1 (k) ⊗ π3∗ TP3 (k) = OP1 (k)  TP3 (k). (4.124)

By the Künneth theorem, we have

H 0 (π3∗ TP3 (k)) = H 0 (OP1 (k)) ⊗ H 0 (TP3 (k)). (4.125)

which in turn reduces the problem to compute H 0 (TP3 (k)). Twisting the Euler exact
sequence (4.21) by OP3 (k), we find

0 −→ OP3 (k) −→ OP3 (k + 1) ⊗ S− −→ TP3 (k) −→ 0 , (4.126)

Six-dimensional supermultiplets from bundles on projective spaces 157

where S− ∼
= C4 . The long cohomology sequence, for the relevant cases k ≥ 0, reduces to
the following short exact sequence

0 −→ H 0 (OP3 (k)) −→ H 0 (OP3 (k + 1)) ⊗ C4 −→ H 0 (TP3 (k)) −→ 0, (4.127)

since H 1 (OP3 (k)) = 0 for any k ≥ 0. As a consequence, one has

   
0 0 0 k+4 k+3 1
h (TP3 (k)) = 4h (OP3 (k+1))−h (OP3 (k)) = 4 − = (k+2)(k+3)(k+5).
3 3 2
(4.128)

The resulting Hilbert series is easily resummed, giving

∞
X 1
grdim(π3∗ TP3 ) = (k + 1)(k + 2)(k + 3)(k + 5)tk
2
k=0 (4.129)
15 − 48t + 54t2 − 24t3 + 3t4
= ,
(1 − t)8

such that the Betti numbers of the associated multiplet are

h i
grdimµA• (TP3 ) = 15 48 54 24 3 . (4.130)

Equivariant Decomposition. Since the Euler exact sequence splits, we find in terms
of representations of sl2 × sl4

H 0 (TP3 (k)) = [0|k + 1, 0, 0] ⊗ S− − [0|k, 0, 0] = [0|k + 1, 0, 1] , (4.131)

and hence by (4.125)

H 0 ((π3∗ TP3 )(k)) = [k|k + 1, 0, 1] . (4.132)

Running our machinery, we obtain the representations

W0 = [0|0, 1, 0]
W1 = −[1|0, 1, 1] − [1|1, 0, 0]
W2 = [0|0, 1, 0] + [2|0, 0, 2] + [2|0, 1, 0] (4.133)
W3 = −[1|0, 0, 1] − [3|0, 0, 1]
W4 = [2|0, 0, 0].
158 Six-dimensional supermultiplets from bundles on projective spaces

Let us summarize the field content.

µA• (π3∗ TP3 )# = (4.134)

h i
Ω2 S− ⊗ V ⊗ C2 V ⊕ Ω3− ⊗ C3 ⊕ V ⊗ C3 (C2 ⊕ C4 ) ⊗ S− C3

4.6.2 Restriction of TP7 to the nilpotence variety

Cohomology and Hilbert series. Since restriction to a smooth subvariety is an exact

functor, it is easy to describe the vector bundle TP7 |Y (k) as the quotient bundle sitting
in the restriction of the ordinary k-twisted Euler exact sequence of the embedding space
P7 of Y, i.e.

0 −→ OY (k, k) −→ OY (k + 1, k + 1) ⊗ [1|0, 0, 1] −→ TP7 |Y (k) −→ 0 , (4.135)

where we have used that OP7 |Y (k) ∼

= OY (k, k). Observing that H 1 (OY (k, k)) vanishes
for any k ≥ 0, we find the short exact sequence in cohomology

0 −→ H 0 (OY (k, k)) −→ H 0 (OY (k + 1, k + 1)) ⊗ [1|0, 0, 1] −→ H 0 (TP7 |Y (k)) −→ 0.

(4.136)
This yields the formula
   
0 k+4 k+3
h (TP7 |Y (k)) = 8(k + 2) − (k + 1)
3 3
(4.137)
4 1
= (k + 2)(k + 4)(k + 3)(k + 2) − (k + 1)(k + 3)(k + 2)(k + 1)
3 6

for the dimensions of the spaces of global sections of TP7 |Y (k). Notice that this also
accounts for the special case k = −1, when H 0 (TP7 |Y (−1)) ∼ = H 0 (OY ) ⊗ [1|0, 0, 1] ∼
=
[1|0, 0, 1]. The Hilbert series of TP7 |Y is found to be

8 − t − 48t2 + 70t3 − 32t4 + 3t5

grdim(TP7 |Y ) = . (4.138)
t(1 − t)8

Equivariant decomposition. Since (4.136) splits, we find on the level of representa-

tions
H 0 (TP7 |Y (k)) = [k + 1|k + 1, 1, 0, 0] ⊗ [1|0, 0, 1] − [k|k, 0, 0] . (4.139)

The associated representations appearing in µA• (TP7 |Y ) are

W0 = [1|0, 0, 1], W1 = −[0|0, 0, 0], W2 = −[1|0, 1, 1] − [1|1, 0, 0],

W3 = [0|0, 0, 2] + [2|0, 0, 2] + 2[0|0, 1, 0] + [2|0, 1, 0], (4.140)
W4 = −2[1|0, 0, 1] − [3|0, 0, 1], W5 = [2|0, 0, 0].
Six-dimensional supermultiplets from bundles on projective spaces 159

Explicitly, the field content of µA• (TP7 |Y ) is summarized in the array

µA• (TP7 |Y )# =
 
 S− ⊗ C2 Ω0 
 .
 Ω1 ⊗ (C ⊕ C ⊕ C3 ) 
S− ⊗ V ⊗ C2 S− ⊗ (C2 ⊕ C2 ⊕ C4 ) C3
Ω3− ⊗ (C ⊕ C3 )
(4.141)

This multiplet looks like a gravitino multiplet, containing a spin-3/2 (Rarita–Schwinger)

field, but no metric or other degree of freedom corresponding to a particle of spin two.

4.6.3 Normal bundle

Cohomology and Hilbert series. Using our results on the cohomology of the bundles
TP7 |Y (k) and TY (k), it is easy to compute the cohomology of NY (k) by means of the
twisted normal bundle exact sequence

0 −→ TY (k) −→ TP7 |Y (k) −→ NY (k) −→ 0 . (4.142)

Since H 1 (TY (k)) = 0 for any k, as can be seen upon using and Künneth theorem in
combination with the twisted Euler exact sequence to evaluate the first cohomology
group of TP3 (k), then one finds a short exact sequence in cohomology

0 −→ H 0 (TY (k)) −→ H 0 (TP7 |Y (k)) −→ H 0 (NY (k)) −→ 0. (4.143)

This implies that for k ≥ −1 one finds the Betti numbers

h0 (NY (k)) = h0 (TP7 |Y (k)) − h0 (TY (k)) . (4.144)

Using our previous results for h0 (TP7 |Y (k)) and h0 (TY (k)) we can deduce

1
h0 (NY (k)) = (k + 3)2 (k + 2)(k + 5) . (4.145)
2

Finally, resumming the Hilbert series yields

8 − 19t + 8t2 + 10t3 − 8t4 + t5

grdim(NY ) = . (4.146)
t(1 − t)8

Equivariant decomposition. For the equivariant decomposition, we identify

H 0 (NY (k)) = [k + 2|k + 1, 0, 1] . (4.147)

160 Six-dimensional supermultiplets from bundles on projective spaces

Using this, we can find the representations appearing in µA• (NY ):

W0 = [1|0, 0, 1], W1 = −[0|0, 0, 0] − [0|1, 0, 1] − [2|0, 0, 0],

W2 = [1|1, 0, 0], W3 = [0|0, 0, 2], W4 = −[1|0, 0, 1], W5 = [0|0, 0, 0].
(4.148)
Explicitly, the multiplet takes the following form:

µA• (NY )# =
 
 S− ⊗ C2 C⊕ Ω2 ⊕ C3 S+ ⊗ C2 
 . (4.149)
 
Ω3− S− ⊗ C2 C

4.6.4 Deformation

As in the example of the Euler sequence for P1 , we find that the field contents of the
direct sums of the multiplets associated to tangent and normal bundle does not match
the field content of µA• (TP7 |Y ), i.e.

µA• (TY )# ⊕ µA• (NY )# 6= µA• (TP7 |Y )# . (4.150)

This is again related to the fact that the normal exact sequence does not split as a
sequence of R/I-modules. However, there is a deformation of the direct sum such that
the resulting multiplet recovers the field content of µA• (TP7 |Y ) up to quasi-isomorphism:

µA• (TP7 |Y ) ' [µA• (TY )# ⊕ µA• (NY )# ]Deform =

 
 S− ⊗ C2 C⊕ Ω2 ⊕ C3 S+ ⊗ C2 
 

 id id 

 
 S+ ⊗ C2 Ω3− ⊗ (C ⊕ C3 ) 
 Ω2 ⊕ C3 S− ⊗ (C2 ⊕ C2 ⊕ C4 ) C ⊕ C3  .
 

 S− ⊗ V ⊗ C 2 Ω1 ⊗ (C ⊕ C ⊕ C3 ) 

 
id
 
 
C
(4.151)

4.7 The conormal bundle exact sequence

The cotangent bundle, the conormal bundle, and the restriction of the cotangent bundle
of the ambient P7 to the nilpotence variety sit in the conormal exact sequence, which is
Six-dimensional supermultiplets from bundles on projective spaces 161

the dual of (4.120):

0 −→ NY∨ −→ Ω1P7 |Y −→ Ω1Y −→ 0. (4.152)

In the same fashion as above, we now study the cohomology of these sheaves and their
associated multiplets.

4.7.1 Cotangent bundle

Cohomology and Hilbert series. As explained above, the cotangent bundle of the
nilpotence variety Y is given by the exterior sum

Ω1Y = π ∗ Ω1P1 ⊕ π3∗ Ω1P3 = Ω1P1
Ω1P3 , (4.153)

where Ω1P1 ∼
= OP1 (−2). As a consequence, the associated multiplet is again a direct sum

µA• (Ω1Y ) = µA• (−2, 0) ⊕ µA• (π3∗ Ω1P3 ) . (4.154)

The multiplet µA• (OY (−2, 0)), arising from the cotangent bundle of P1 , was already
described in §4.3.3, therefore we are left with describing µA• (π3∗ Ω1P3 ). To this end, we
need to study the zeroth cohomology of

π3∗ Ω1P3 (k) = π ∗ OP1 (k) ⊗ π3∗ Ω1P3 (k) = OP1 (k)  Ω1P3 (k). (4.155)

The Künneth theorem implies that

H 0 (π3∗ Ω1P3 (k)) = H 0 (OP1 (k)) ⊗ H 0 (Ω1P3 (k)), (4.156)

reducing the problem to compute the dimension of the zeroth cohomology of Ω1P3 (k).
This can be obtained by Bott formulas [OSS80] or by explicitly studying the twist of the
dual of the Euler exact sequence for Ω1P3 ,

0 Ω1P3 (k) OP3 (k − 1) ⊗ C4 OP3 (k) 0. (4.157)

In order to obtain a short exact sequence of modules, we have to check that the connection
morphism in the corresponding long exact sequence in cohomology vanishes. Clearly, if
k < 0 then H 0 (Ω1P3 (k)) = 0. If k = 0, then this corresponds to the Hodge number of P3
and in particular one finds h0 (Ω1P3 ) = 0 = h1,0 (P3 ). For k = 1 it is easy seen that the
map
ϕk=1 : H 0 (OP3 ) ⊗ C4 → H 0 (OP1 (1)), (4.158)
 162 Six-dimensional supermultiplets from bundles on projective spaces

P3
given by C4 3 (c0 , . . . , c3 ) 7→ i=0 ci Xi , for {X0 , . . . , X1 } global sections of OP3 (1) is
an isomorphism and hence H (Ω1P3 (1)) = 0. On the other hand in the case k > 1 the
0

map ϕk>1 : H 0 (OP3 ) ⊗ C4 → H 0 (OP1 (1)) is only surjective so that H 1 (Ω1P3 (k)) = 0 and
ker(ϕk>1 ) = H 0 (Ω1P3 ).

It follows that for k > 1 one has

   
k+2 k+3 1
h0 (Ω1P3 (k)) = 4 − = (k + 2)(k + 1)(k − 1). (4.159)
k−1 k 2

In turn, this implies that

1
h0 (π3∗ Ω1P3 (k)) = (k + 1)2 (k + 2)(k − 1), (4.160)
2

and the related Hilbert series gives

18 − 64t + 89t2 − 64t3 + 28t4 − 8t5 + t6

grdim(π3∗ Ω1P3 ) = t2 . (4.161)
(1 − t)8

Equivariant decomposition. For the equivariant decomposition, we identify

H 0 (Ω1P3 (k)) = [0|k − 2, 1, 0] , (4.162)

and find the following representations in µA• (π3∗ Ω1P3 ).

W0 = [2|0, 1, 0]
W1 = −[1|0, 0, 1] − [1|1, 1, 0] − [3|0, 0, 1]
W2 = [0|0, 0, 0] + [0|0, 2, 0] + [0|1, 0, 1] + [2|0, 0, 0] + [2|1, 0, 1] + [4|0, 0, 0]
W3 = −[1|0, 1, 1] − [1|1, 0, 0] − [3|1, 0, 0] (4.163)
W4 = [0|0, 0, 2] + [2|0, 1, 0]
W5 = −[1|0, 0, 1]
W6 = [0|0, 0, 0]

The resulting multiplet takes the following form.

µA• (π3∗ Ω1P3 )# = (4.164)

 
C2 ⊗ V ⊗ S+ Sym2 (V ) ⊕ Ω2 C2 ⊗ S− ⊗ V Ω3−
 Ω1 ⊗ C3 C2 ⊗ S− C 
C4 ⊗ S− C3 ⊕ C3 ⊗ Ω2 ⊕ C5 S+ ⊗ C4 Ω1 ⊗ C3
Six-dimensional supermultiplets from bundles on projective spaces 163

4.7.2 Restriction of Ω1P7 to the nilpotence variety

Cohomology and Hilbert series. Dually to the case of TP7 |Y , the cohomology of
Ω1P7 |Y and its twists is studied by restricting the dual of the Euler exact sequence for the
ambient space P7 to Y . This gives

0 Ω1P7 |Y (k) O(k − 1, k − 1) ⊗ C8 OY (k, k) 0. (4.165)

Studying the related long exact cohomology sequence, it is easy to see that if k ≤ 1
then H 0 (Ω1P7 |Y (k)) = 0. In the remaining case, when k > 1, the space of global sections
H 0 (Ω1P7 |Y (k)) is actually non-zero and the long cohomology sequence splits since the
polynomial map

(Xi ,Yj )
H 0 (OY (k − 1, k − 1)) ⊗ C8 H 0 (OY (k, k)) (4.166)

is surjective, so one gets the short exact sequence

0 → H 0 (Ω1P7 |Y (k)) → H 0 (OY (k − 1, k − 1)) ⊗ C8 → H 0 (OY (k, k)) → 0. (4.167)

This says that

   
0 k+2 k+3
h (Ω1P7 |Y (k)) = 8k − (k + 1)
k−1 k
(4.168)
4 1
= k(k + 2)(k + 1)k − (k + 1)(k + 3)(k + 2)(k + 1).
3 6

The related Hilbert series of Ω1P7 |Y can be resummed easily to give

34 − 112t + 137t2 − 80t3 + 28t4 − 8t5 + t6

grdim(Ω1P7 |Y ) = t2 (4.169)
(1 − t)8

Equivariant decomposition. In terms of representations, the sequence gives

H 0 (Ω1P7 |Y (k)) = [k − 2|k, 0, 0] + [k|k − 2, 1, 0] + [k − 2|k − 2, 1, 0] . (4.170)

 164 Six-dimensional supermultiplets from bundles on projective spaces

Using these results, we can deduce the field content of µA• (Ω1P7 |Y ).

W0 = [0|0, 1, 0] + [0|2, 0, 0] + [2|0, 1, 0]

W1 = −2[1|0, 0, 1] − 2[1|1, 1, 0] − [3|0, 0, 1]
W2 = [0|0, 0, 0] + [0|0, 2, 0] + [0|1, 0, 1] + 2[2|0, 0, 0] + 2[2|1, 0, 1] + [4|0, 0, 0]
W3 = − [1|0, 1, 1] − [1|1, 0, 0] − 2[3|1, 0, 0] (4.171)
W4 = [0|0, 0, 2] + [2|0, 1, 0]
W5 = − [1|0, 0, 1]
W6 = [0|0, 0, 0]

The resulting multiplet takes the following form.

µA• (Ω1P7 |Y )# = (4.172)

 
2
Sym (V ) ⊕ Ω2
 Ω1 ⊗ (C ⊕ C3 ) (C2 ⊗ V ⊗ S− )⊕2 C2 ⊗ V ⊗ S− Sym2 (S− ) 
Ω2 ⊗ C3 ⊕ C5 S− ⊗ C2
 
C
Sym2 S+ (S+ ⊗ C4 )⊕2

 C4 ⊗ S− S− ⊗ C2 
Ω2 ⊗ C3 ⊕ C3

4.7.3 The conormal bundle and its supergravity multiplet

Cohomology and Hilbert series. Having available the cohomology of the cotangent
bundle and the restriction of Ω1P7 to the nilpotence variety Y, one can study the conormal
bundle and its related multiplet in a similar fashion as for the normal bundle above, i.e.
by considering k-twists of the conormal exact sequence (4.25):

0 NY∨ (k) Ω1P7 |Y (k) Ω1Y (k) 0. (4.173)

The issue one faces following this approach is that the related long exact cohomology
sequence starts with a four-term sequence in the relevant case k > 1

0 → H 0 (NY∨ (k)) → H 0 (Ω1P7 |Y (k)) → H 0 (Ω1Y (k)) → H 1 (NY∨ (k)) → 0, (4.174)

and it is not completely trivial to establish the vanishing of the group H 1 (NY∨ (k)) for
any k ≥ 1. We refer to the appendix of [Hah+22] for this verification.

Using the above results, one computes

     
k+2 k+3 k+2 1
h0
(NY∨ (k)) = 8k − 2k 2
− (k − 1) = (k + 1)(k − 1)2 (k + 2).
3 3 2 2

(4.175)
 Six-dimensional supermultiplets from bundles on projective spaces 165

The related Hilbert series is resummed to give

∞
X 1 6 − 8t − 17t2 + 40t3 − 28t4 + 8t5 − t6
grdim(NY∨ ) = (k + 1)(k − 1)2 (k + 2)tk = t2 .
2 (1 − t)8
k=2
(4.176)

Equivariant decomposition. In terms of representations the conormal exact se-

quence implies
H 0 (NY∨ (k)) = [k − 2|k − 2, 1, 0] . (4.177)

We find the following field content for µA• (NY∨ ).

W0 = [0|0, 1, 0], W1 = −[1|0, 0, 1], W2 = −[0|0, 2, 0] + [2|0, 0, 0], (4.178)

W3 = [1|0, 1, 1], W4 = −[0|0, 0, 2] − [2|0, 1, 0], W5 = −[1|0, 0, 1], W6 = [0|0, 0, 0].

In summary, the multiplet takes the form

 
 V S− ⊗ C2 C3 
•
µA (NY∨ )# =
 

Sym20 (V ) (V ⊗ S− ) 3 ⊗ C2 V ⊗ C3 ⊕ Ω3− S− ⊗ C2
 
C.
2

(4.179)
This is precisely the field content of six-dimensional N = (1, 0) supergravity, presented
as the “type-II Weyl multiplet” [LTM12].

4.7.4 Deformation

There is again a deformation of µA• (Ω1Y )# ⊕ µA• (NY∨ )# such that the result is quasi-
isomorphic to µA• (Ω1P7 |Y )# :

µA• (Ω1P7 |Y )# ' [µA• (Ω1Y )# ⊕ µA• (NY∨ )# ]Deform = (4.180)

 
 Sym2 (V ) ⊕ Ω2 
 2
C ⊗ V ⊗ S− Ω3− 
 1
 Ω ⊗ C3 Ω2 ⊗ C3 ⊕ C5 C2 ⊗ V ⊗ S− 
C4 ⊗ S−

Ω1 ⊗ C3 S− ⊗ C2
(S+ ⊗ C4 )⊕2 C 

 Sym2 S+
 Sym20 (V ) 
C2 ⊗ (V ⊗ S+ ) 3 S− ⊗ C2


 V 2 2 2 C2 ⊗ (V ⊗ S− ) 3 Sym2 (S− ) C 
Ω ⊗C

S− ⊗ C2 2
S− ⊗ C2
 
 
3 3

 C ⊕C 

 
 
 
 Sym20 (V ) C2 ⊗ (V ⊗ S− ) 3 Ω1 ⊗ C3 ⊕ C2 S− ⊗ C2 C 
2
Chapter 5

Zwischenspiel:
twisting and holography

We have already encountered twists of supersymmetric field theories at various places

in this thesis at an ad-hoc level; a proper introduction seems overdue. This chapter
is meant to set the stage for the second part of this thesis which treats the eleven-
dimensional supergravity theory and its twists. For this purpose, we first give an brief
introduction to twisted supersymmetric field theories and then provide a short review of
some relevant aspects from the twisted holography program. Among many other things,
these works established conjectures on the twists of supergravity theories in ten and
eleven-dimensions by using tools from topological string theory. In §6 and §7, we study
these conjectures then directly from a target space perspective.

5.1 Twisting supersymmetric field theories

5.1.1 The nilpotence variety as the moduli space of twists

Let G be a super Lie group with super Lie algebra g = g+ ⊕ Πg− and consider a
field theory T with symmetry G. Broadly speaking, twisting means taking invariants
with respect to the odd abelian subalgebra spanned by a square-zero element Q ∈ g− .
Concretely, this is achieved by adding the action of Q to the BRST or BV differential
of the theory (we’ll discuss what this means in more detail below). Let us denote the
twisted theory by T Q .1
1
In addition to deforming the differential, often a twisting morphism which modifies the action of G+
on the theory is applied. For the moment this is not essential for our discussion.

167
168 Zwischenspiel: twisting and holography

Thus, given a G-equivariant field theory T , the nilpotence variety of g is the natural
moduli space of twists for T . The nilpotence variety decomposes into orbits under the
action of the Lie group G+ . As G+ acts on T by symmetries, physically inequivalent
twists are labeled by the G+ -orbits of Yg .

Fixing a square-zero element Q ∈ Y , we can deform the super Lie algebra of symmetries
itself g to a differential super Lie algebra

(g , [Q, −]) . (5.1)

Its cohomology gQ := H • ((g, [Q, −])) is again a super Lie algebra and consists of residual
symmetries in the twisted theory. In particular, all elements in g which are in the image
of [Q, −] act trivially on the twisted theory. The residual symmetry algebra gQ again
has a nilpotence variety YgQ encoding the further twists of the theory.

5.1.2 Twists for the super Poincaré algebra

For the purpose of this thesis, we are interested in the case where g is a super Poincaré
algebra. Then, the relevant orbit stratification is induced by the Poincaré group and R-
symmetry. Since the translations act trivially, this reduces to the Lorentz group and R-
symmetry. The nilpotence varieties for super Poincaré algebras and their orbit structures
were studied and classified in [ES19b; ESW21]. In a first rough overview the properties
of the different twists by elements from the super Poincaré algebra can be understood in
terms of the number of translations which act non-trivial on the twisted theory. This is
precisely measured by the cohomology group

H 2 (g, [Q, −]) = V /Im([Q, −]) , (5.2)

consisting of those translations which are not in the image of [Q, −]. On general grounds (see
for example [ES19b]), for non-vanishing Q, the dimension of H 2 ((g, [Q, −])) is at most
half the dimension of the vector representation

1
dim H 2 (g, [Q, −]) ≤ dim V. (5.3)
2

Depending on the dimension of this cohomology group, one distinguishes different cases.

— If H 2 (g, [Q, −]) = 0, all translations act trivially on the twisted theory. The super-
charge Q is called topological.
Zwischenspiel: twisting and holography 169

— If the dimension of V is even and the inequality is saturated, precisely half of

the translations act trivially on the twisted theory. Choosing such a supercharge
induces a complex structure on V ; the supercharge Q is called holomorphic.

— In the general case, when more than half but not all translations act trivially, the
supercharge Q is called mixed.

In general, twisting a field theory on Rd by a supercharge with

dim(H 2 (g, [Q, −])) = k (5.4)

yields a topologic-holomorphic theory formulated on Rd−2k × Ck . We say that Q has

d − 2k topological directions (and 2k holomorphic directions).

Further terminology arises from the orbit stratification induced from the action of the
Lorentz and R-symmetry groups on the nilpotence varieties. Twists lying in the minimal
orbits are called minimal. In even dimensions, these are precisely the holomorphic twists;
in odd dimensions, they have precisely one topological direction. Twists in the top strata
are called maximal; they correspond to the smooth points of the nilpotence variety. More
generally, the stratification of the nilpotence variety tells us which twists can be obtained
as further deformations from other twists. The maximal twists are precisely those having
no further deformations. (See [ESW21] for more information on this terminology.)

5.1.3 Twisting in the BV formalism

In order to compute twists in practice, one has to specify a model for the theory T at
hand. One natural choice is to work within the BV formalism (where all relevant infor-
mation about the theory is already encoded in a differential such that the deformation
by a square-zero element is straightforward). Recall that this means to model T by
a local cyclic L∞ algebra (L, µk , h−, −i) with pairing of degree −3 thought of as the
space of fields. The classical observables are described by the factorization algebra of
Chevalley–Eilenberg cochains

Obs(T ) = (C• (L) , QBV = dCE ) , (5.5)

where the Chevalley–Eilenberg differential is called the BV differential. By assumption,

the L∞ algebra L carries an L∞ g-module structure. For each square-zero element
Q ∈ Yg , this module structure induces a differential

δQ : Obs(T ) −→ Obs(T ), (5.6)

170 Zwischenspiel: twisting and holography

such that we can consider the deformation2

(C• (L) , QBV + δQ ) = Obs(T )Q , (5.7)

constituting the factorization algebra of observables of the twisted theory (for more
details see [Cos13a]). In order to describe the twisted theory, one typically moves to a
smaller quasi-isomorphic version of Obs(T )Q by eliminating acyclic pairs. Further, there
is a new local L∞ algebra LQ whose Chevalley–Eilenberg cochains are isomorphic to
Obs(T )Q . This L∞ algebra describes the fields of the twisted theory.

Equivalently, twisting in the BV formalism can be described by means of the BV action.

This amounts to deforming,

Q
XZ
SBV [Φ] = SBV [Φ] + hΦ, ρ(i) (Q, . . . , Q)(Φ)i. (5.8)
i M

Physically, this can be interpreted as putting the field theory into a background where
the constant ghosts associated with the symmetry transformations of g take the value Q
(here we take the background values for all other fields as trivial, though more general
situations are possible). This point of view is in particular relevant to supergravity
theories, where the supersymmetry transformations are gauged. In this case, this point
of view becomes the defining feature [CL16] such that a twisted supergravity theory is
one placed in such a background. Some fields then decouple from the rest of the theory
such that they cease influencing the dynamics. Integrating out these fields corresponds
to the elimination of acyclic pairs in Obs(T )Q .

Remark 5.1.1 (Twists of multiplets). While twisting supersymmetric field theories is

ultimately what we are interested in, it is clear that the above description of twists is
compatible with the definition of multiplets as introduced in §2. Thus, one can easily
consider twists of multiplets which are not equipped with a BV datum (and in fact we
briefly did this in §4.3.2).

5.1.4 Aside: Twisting in representation theory

Independent of the field theoretic background, the concept of twisting plays an important
role in the representation theory of simple super Lie algebras (see [DS05; Gor+22]).
Twisting there comes under the name of the Duflo–Serganova functor

DSQ : Modg −→ ModgQ (M, ρ) 7→ H • (M, ρ(Q)) (5.9)

2
In addition, one typically performs a regrading in order to guarantee that the deformed differential
is of uniform degree. If this is not possible, the twisted theory is only Z/2Z-graded.
Zwischenspiel: twisting and holography 171

which assigns to every g-module a corresponding gQ = H • (g, [Q, −])-module. The mod-
ules DSQ (M ) form a family over the nilpotence variety; the support of this family is a
subvariety of Yg and is called the associated variety of M .

Associated to any supersymmetric field theory comes a super Hilbert space of states
where the application of such results is natural.

5.1.5 Twisting and curved backgrounds

While the field theories we are interested in exist in any appropriately structured back-
ground geometry M , twists—as defined in the previous section—typically do not. In
essence, this is due to most backgrounds breaking supersymmetry (see for example [FS11]).
For starters, M has to admit covariantly constant spinors for the twisting procedure to
make sense globally.

Starting from the opposite direction, one can consider the twist of the theory on a model
geometry preserving supersymmetry (in our case Rd equipped with the standard euclid-
ean metric) and globalize the twisted theory from there. However, since the presence of
the twisting supercharge breaks the Lorentz (and R-symmetry) of the untwisted theory
to H 0 (g, [Q, −]), this typically come with requirements on the holonomy group of the
underlying manifold. This is, in particular unfortunate for topological twists, where the
twisted theory is supposed to compute smooth or topological invariants. In some cases
this can be addressed by a twisting morphism.

Definition 5.1.2. Let Q ∈ Y be a square-zero supercharge. A twisting morphism for

Q is a morphism of Lie groups f : SO(d) −→ GR such that Q is scalar under the new
SO(d)-action defined by composing with id ×f : SO(d) −→ SO(d) × GR .

In essence, a twisting morphism thus mixes the action of the Lorentz on the fields with the
one of the R-symmetry group such that we recover an action of a full copy of the Lorentz
group on the fields of the theory which allows globalization to a general manifold M .
Whenever twisting morphisms exist, the supercharge is topological; the converse however
is not true [ES19b].

In the general case, the twist on Rd defines a theory on Rd−2k × Ck and this theory
does not globalize to an arbitrary manifold M , but only to a manifold equipped with an
appropriate transverse holomorphic fibration. Somewhat pictorially, we can summarize
this approach to the twisting procedure in the following picture.
172 Zwischenspiel: twisting and holography

Theory on Theory on
Rd M

twist

Theory on globalize Theory on

Rd−2k × Ck MT HF

Figure 5.1: The twisted theory on a flat background globalizes to a theory on a

manifold with a T HF structure. It would be interesting to see how this square could
be completed by twisting the physical theory in a more general background M .

5.2 A panoramic view on twisted holography

Twists are especially of interest in combination with dualities: twisting on either side
of one can both establish fascinating relations between different areas of mathematics
as well as provide mathematically rigorous instances of a physical duality principle.
To set the stage for the following chapters, we now give a schematic overview on the
twisted holography program as initiated by Costello, Li, Gaiotto and others (see for
example [Cos07; CL16; CG18; Cos17; Cos16; Gin+22; CP21]).

The top-down approach to holography typically knows three ingredients: the underlying
worldsheet string theory, the induced worldvolume gauge theory on a configuration of
branes, and a closed string field theory in the backreacted geometry. Holographic duali-
ties establish equivalences between the latter two of those, by viewing them as instances
of the first system.

For illustrational purpose, let us briefly sketch the original argument provided by Malda-
cena [Mal98]. Consider type IIB superstring theory in R10 with a stack of N D3 branes
situated along a subspace R4 ⊂ R10 . There are two fundamentally different perspectives
on these branes

1. Open strings end on D branes. There is a low energy effective theory on the
worldvolume of the brane, here given by N = 4 super Yang–Mills theory on R4 .

2. D branes are sources for the fields in the closed string sector and as such deform
the background geometry. The low energy effective description is a supergravity
theory in the backreacted geometry X back , here type IIB supergravity.

Taking the first perspective, one obtains an effective theory at low energies consisting
of three sectors: the worldvolume theory describing the dynamics of open strings, type
Zwischenspiel: twisting and holography 173

IIB supergravity on R10 corresponding to the dynamics of the closed strings, and a third
piece describing interactions between the two. In an appropriate limit (the decoupling
limit [Mal98]), open and closed string modes decouple such that the third piece vanishes.

On the other hand, in the appropriate limit, the theory describing closed strings in the
backreacted geometry, also decomposes into two decoupled sectors: closed strings near
the brane and closed strings far away from the brane. Examining X back , one finds that
the former is described by type IIB supergravity in AdS5 × S 5 , while the latter is given
by type IIB supergravity in R10 .

Identifying both perspectives and matching the sectors, one arrives at an equivalence
between N = 4 super Yang–Mills theory in R4 and type IIB supergravity in AdS5 × S 5 ,
historically the first instance of the AdS/CFT correspondence.

The twisted holography program aims to study such correspondences in the twisted
setting. Thus, in order to arrive at a comprehensive understanding of twisted holography,
three different kinds of twists are relevant: twists of the supersymmetric worldvolume
gauge theories, twists of supergravity theories in backreacted backgrounds, and, finally,
the twists of the underlying worldsheet string theory from which the former two are
expected to arrive as effective field theories.

From the worldsheet perspective, a twisted version of the theory is most naturally de-
scribed as a topological string theory. Results by Costello [Cos07] and Lurie [Lur09] show
that specifying such a theory is equivalent to fixing a Calabi–Yau A∞ category, which
can be interpreted as a category of branes. (We are working at a very impressionistic
level here and ignore many technical details.)

At a qualitative level, it is useful to picture the category of branes as follows. Its objects
are the branes itself, while the morphisms are open string configurations connecting
branes,
Hom(A, B) = {Open strings connecting from A to B}, (5.10)

with composition, A∞ structure, and the Calabi–Yau pairing trA : Hom(A, A) −→ C

given by the respective diagrams joining open strings.

Fixing a category of twisted branes C, one can reconstruct both the worldvolume gauge
theory on the brane (describing open strings), as well as the closed string sector in the
following way [Cos07].

1. Fix an object C ∈ C. The worldvolume theory is modeled by RHomC (C, C)

2. The closed string field theory is described by the cyclic cochains of C, CC • (C).
174 Zwischenspiel: twisting and holography

From the perspective of topological string theory, two distinct twists of the worldsheet
model are central, the A and the B twist. The corresponding A and B model can be
thought of as topological phases of underlying string theory. Correspondingly, there are
categories of A and B branes.

A. C = Fuk(X). The category of branes in the A twist is the Fukaya category of target
space, where branes are Lagrangian submanifolds.

B. C = Coh(X). The category of branes in the B twist is the derived category of

coherent sheaves on target space.

Let us review the following classic example.

Example 5.2.1 (Holomorphically twisted type IIB with D3 branes; see [CL16]). Let C =
Coh(C5 ) and consider a single branes along C2 ⊂ C5 . This brane is represented by the
object OC2 ∈ Ob(C).

We compute the worldvolume theory in two steps. First we resolve OC2 in free objects
inside C. For this, we use a Koszul resolution giving,

K • = OC5 ⊗ C[ε1 , ε2 , ε3 ] , dK zi = εi ,


(5.11)

Thus, we find

RHomC (OC2 , OC2 ) = HomC (K • , OC2 ) = OC2 ⊗ C[ε1 , ε2 , ε3 ]. (5.12)

In a second step, we resolve in smooth functions on C2 to get a field theory on the brane.
This results in
Ω0,• (C2 ) ⊗ C[ε1 , ε2 , ε3 ] , ∂¯ .


(5.13)

Hence, the worldvolume theory is holomorphic Chern–Simons theory. Slightly more

general, considering a stack of N branes represented by OC⊕N
2 , we obtain holomorphic

Chern–Simons theory with gauge algebra glN . This indeed matches with the holomor-
phic twist of N = 4 super Yang–Mills theory in four dimensions as first computed by
Baulieu [Bau11].

On the other hand, using the Hochschild–Kostant–Rosenberg theorem one finds that the
closed string sector is described by BCOV theory, modeled by the cochain complex

PV•,• (C5 )[[t]] , ∂¯ + t∂ ,

(5.14)

where t is a parameter placed in degree two. The backreaction was carried out in [CL16]
by including appropriate branes as sources in BCOV theory. They find that there is a
Zwischenspiel: twisting and holography 175

non-vanishing five-form flux, such that BCOV theory on C5 r C2 with this background
five form flux gives the candidate for holomorphically twisted type IIB supergravity on
AdS5 × S 5 .

Crucially, for the purpose of this thesis, this procedure produces conjectural description
for twisted supergravity theories starting from a category of twisted branes. In the case
of the above example, this yielded the conjectural description of the minimal twist of
type IIB supergravity in terms of BCOV theory [CL16].

Pictorially, the different field theories and their twists are related as summarized by the
following diagram.

twisted holography Ω background

CC • (C)
RHomC (C, C) Category of branes C
+ backreaction

twist 2d TCFT twist

twist

Worldvolume theory WS string theory Supergravity in X back

holographic duality

Figure 5.2: A schematic overview of the twisted holography program

The bottom row represents the physical (i.e. untwistd) holographic duality arising from
worldsheet string theory. The top row sketches the procedure employed by Costello and
Li to arrive at conjectural descriptions for the twisted worldvolume gauge theories and
supergravity theories. However, by definition, these twisted theories arise as twists from
the full physical worldvolume gauge and supergravity theories respectively. In order
to provide proofs for these conjectures and to give a more complete understanding of
Figure 5.2 providing direct calculations of the twists of supergravity theories in target
space is indispensable (in the diagram this corresponds to starting in the bottom right
corner and following the arrow upwards).
176 Zwischenspiel: twisting and holography

In addition, there are cases of interest where a worldsheet perspective based on topo-
logical string theory is not readily available. This is in particular the case for eleven-
dimensional supergravity which is the low energy limit of M-theory. (Still, using duali-
ties and results from ten-dimensions the maximal twist of eleven-dimensional supergrav-
ity was conjectured to be Poisson–Chern–Simons theory by Costello [Cos; Cos16] and
tested [RY19].) In the following two chapters, we aim to understand twisted supergravity
from a target space perspective.
Chapter 6

Maximally twisted
eleven-dimensional supergravity

6.1 Introduction

Eleven-dimensional supergravity [CJS78] is the low energy limit of M-theory, a conjec-

tural theory that is believed to unify type I, II, and heterotic superstring theories [Wit95].
It realizes the maximal dimension that has a supersymmetric representation with par-
ticles of spin at most two [Nah78], and the action of eleven-dimensional supergravity is
expected to be unique [CJS78]. M-theory compactifications on manifolds with G2 holon-
omy result in four-dimensional field theories with minimal supersymmetry and have been
intensely studied in relation to non-perturbative string dualities and phenomenology.

In this chapter, we consider the maximal twist of eleven-dimensional supergravity start-

ing from the component field BV complex of the theory. This twist is of mixed type
(topological in seven, and holomorphic in the remaining four directions) such that the
twist of supergravity in a flat background defines a theory on R7 × C2 . More generally,
we can put the twisted theory on manifolds M 7 × M 4 of G2 × SU(2) holonomy. Costello
conjectured the maximal twist to be given by Poisson–Chern–Simons theory [Cos; Cos16;
RY19]. As a free BV theory, Poisson–Chern–Simons theory on R7 × C2 is given by the
cochain complex
Ω• (R7 ) ⊗ Ω0,• (C2 ) , Dtw ,


(6.1)

where the differential Dtw decomposes into

Dtw = dR7 ⊗ 1 + 1 ⊗ ∂ C2 . (6.2)

177
178 Maximally twisted eleven-dimensional supergravity

Here dR7 is the de Rham differential on R7 and ∂ C2 is the Dolbeault differential on

C2 . In addition, there are interactions given by the Poisson bracket, for this chapter we
restrict our attention to the free theory. We will come back to the interactions in §7.
The generalization to M 7 × M 4 is straightforward.

The aim of this chapter is to compute the maximal twist explicitly in component fields
and thereby make contact with the formulation of the full supergravity theory as orig-
inally envisioned by Cremmer–Julia–Scherk [CJS78]. To this end, we will show how to
obtain the fields and BV differential by directly twisting the component fields of eleven-
dimensional supergravity in the BV formalism [BV81]. After the twist, the three-form
C (3) with its ghost system C (2) , C (1) , C (0) , the spin-3/2 Rarita–Schwinger field ψ, and all
of their corresponding antifields organize into a differential form A ∈ Ω• (R7 ) ⊗ Ω0,• (C2 ),
as conjectured by Costello. Its components are displayed in Table 6.1.

Ω0 (R7 ) Ω1 (R7 ) Ω2 (R7 ) Ω3 (R7 ) Ω4 (R7 ) Ω5 (R7 ) Ω6 (R7 ) Ω7 (R7 )

Ω0,0 (C2 ) C (0) C (1) C (2) C (3) ψ ψ∨ C (3)∨ C (2)∨
Ω0,1 (C2 ) C (1) C (2) C (3) ψ ψ∨ C (3)∨ C (2)∨ C (1)∨
Ω0,2 (C2 ) C (2) C (3) ψ ψ∨ C (3)∨ C (2)∨ C (1)∨ C (0)∨

Table 6.1: Fields in maximally twisted supergravity

We will derive the conjectured form of the twisted fields and differential starting from
the manifestly covariant formulation of eleven-dimensional supergravity [Ber02; Ced10c;
Ced10a; BG18] in the pure spinor superfield formalism [Ber00; Ber05; Ced14]. We use
this formalism to obtain the BV complex of the three-form multiplet in eleven dimensional
supergravity, including the full action of the supersymmetry algebra on the component
fields. These results are then used to carry out the actual twist on the level of component
fields. This gives an explicit understanding of the fields in the twisted theory as well as
the formation of trivial pairs in terms of the fields of the untwisted supergravity multiplet.

The traditional approach to eleven-dimensional supergravity in superspace [BH80; CF80;

Ced+00a; Ced+00b; Ced+05] starts with the supervielbein and imposes conventional
constraints [GSW80; GS80] on torsions and curvatures. We will make some speculative
remarks about the twist of the supervielbein at the end. A partially off-shell formulation
of eleven-dimensional supergravity adapted to manifolds of G2 ×SU(2) holonomy is given
in [Bec+17; Bec+18; Bec+21] and is closely related to the twisted theory.

We will work in Euclidean signature. We hope to return to the twist of the higher order
terms and the formulation in Lorentzian signature in subsequent work.
Maximally twisted eleven-dimensional supergravity 179

Organization. The rest of this chapter is structured as follows. In §6.2, we de-

scribe the types of available twists in eleven-dimensional supergravity in general and
the G2 × SU(2) invariant maximal twist in particular. In §6.3 we briefly review how
the eleven-dimensional supergravity arises in the pure spinor superfield formalism. We
introduce the BV complex and describe the action of supersymmetry on its component
fields. Finally, in §6.4 we describe the decomposition of the fields and supersymmetry
transformations with respect to G2 ×SU(2). We then use the decomposition to determine
the fields surviving the partial topological twist and the resulting action of the modified
BV differential. We conclude with some thoughts on further directions in §6.5.

6.2 The two twists of eleven-dimensional supergravity

Eleven-dimensional supergravity can be twisted in two distinct ways that correspond to

the two orbits of the nilpotence variety [ESW21]. Let us quickly review the relevant
structure of the nilpotence variety.

Recall that, in any dimension, the Dirac spinor representation S is obtained from a
maximal isotropic subspace L ⊂ V of the vector representation V by setting

S = ∧• L∨ . (6.3)

S forms a Clifford module for Cl(V ) and thus in particular a representation of so(V ). In
the case where d = dim(V ) is odd, this representation is irreducible. As we are interested
in eleven-dimensional supergravity, we restrict to this case for the moment.

For Q ∈ S, the annihilator with respect to Clifford multiplication

Ann(Q) = {v ∈ V |v · Q = 0} (6.4)

gives an isotropic subspace Ann(Q) ⊂ V . Q is called a Cartan pure spinor if Ann(Q) is

maximal isotropic. Every Cartan pure spinor is square zero, the converse, however, is in
general not true. More generally, one can define the varieties

PSk = {Q ∈ S | dim(L) − dim(Ann(Q)) ≤ k} , (6.5)

which define a filtration

PS0 ⊆ PS1 ⊆ . . . PSn = S . (6.6)
180 Maximally twisted eleven-dimensional supergravity

Let V11 denote the vector representation of Spin(11) and S11 the Dirac spinor represen-
tation. The dimension of S is 32 and its symmetric square decomposes as

Sym2 (S) ∼
= V ⊕ ∧2 V ⊕ ∧5 V. (6.7)

The N = 1 super Poincare algebra in eleven dimensions takes the form,

g = so(V ) ⊕ S(−1) ⊕ V (−2), (6.8)

where the bracket of two degree one elements is given by the projection onto the vector
representation in the decomposition (6.7).

In coordinates, the nilpotence variety can be explicitly described by the eleven equations

λα Γµαβ λβ = 0 . (6.9)

This variety is closely related but not identical with the variety of Cartan pure spinors;
in fact, one finds Y = PS3 [ES19a]. The variety of Cartan pure spinors sits inside Y as a
subvariety PS0 ⊂ PS3 = Y . It is the singular locus of Y and can be explicitly described
by imposing the additional equations

λΓµν λ = 0 . (6.10)

In addition, PS0 is also the minimal orbit of the even part of the super Poincaré algebra
and thereby its points correspond to the minimal twists of the eleven-dimensional super-
gravity theory. These are topological in one direction and holomorphic in the remaining
ten; the stabilizer of such a supercharge is SU(5).

The second orbit consists of points away from the singular locus. There, the stabilizer of a
supercharge is G2 × SU(2). This corresponds to the maximal twist of eleven-dimensional
supergravity that we will study in this chapter. The twist exists on manifolds with
G2 × SU(2) holonomy [Mov11; Cos16; ESW21].

Let us elaborate a little further on the maximal twist. The spinor representation in
eleven dimensions decomposes as

S11 = S7 ⊗ S4 . (6.11)

The Dirac spinor representation in four dimensions, S4 , decomposes into Weyl spinor
representations S+ and S− :

S4 = ∧• L∨
4 =∧
even ∨
L4 ⊕ ∧odd L∨
4 =: S+ ⊕ S− . (6.12)
Maximally twisted eleven-dimensional supergravity 181

Identifying the group Spin(4) ∼

= SU(2)+ × SU(2)− , S+ and S− are the fundamental
representations of SU(2)+ and SU(2)− , respectively. Restricting to G2 , the spinor rep-
resentation S7 further decomposes as

S7 = 1G2 ⊕ VG2 , (6.13)

where VG2 is the seven-dimensional representation of G2 . Thus, we have the decomposi-

tion
S11 = (1G2 ⊕ VG2 ) ⊗ (∧0 L∨ 2 ∨
4 ⊕ ∧ L4 ⊕ S− ) . (6.14)

As a representation of G2 × SU(2)− × U(1)L , where U(1)L is the Cartan subgroup of

SU(2)+ this gives
S11 = [(00) ⊕ (10)] ⊗ 1−1 ⊕ 1+1 ⊕ 20 .


(6.15)

Here, we introduced Dynkin labels for the G2 -representation. SU(2)×U(1)-representations

are labeled by the dimension of the SU(2)-representation, with the U(1)-charge as a su-
perscript. To study the maximal twist, we choose a square zero supercharge

Q ∈ 1G2 ⊗ ∧0 L∨
4 = (00)1
−1
. (6.16)

Thus, we immediately see that Q is invariant under the action of G2 × SU(2)− and has
U(1)L -charge −1,.

The normal space to the nilpotence variety is spanned by the supercharges

Qm ∈ (VG2 ⊗ ∧2 L∨
4 ), (6.17)
Qα̇ ∈ (1G2 ⊗ S− ). (6.18)

They satisfy the following relations

[Q, Qm ] = Pm (6.19)
[Q, Qα̇ ] = P−α̇ . (6.20)

Here we already used that the vector representation decomposes under G2 ×SU(2)×U(1)
as
V11 = (10) ⊕ 2−1 ⊕ 21 . (6.21)

Our conventions are that indices m, n, . . . are indices for the seven-dimensional vector
representation, while α̇, β̇, . . . correspond to SU(2)− .

The above relations explicitly show that, as announced earlier, the twisted theory is
indeed topological in seven and holomorphic in the remaining four directions.
 182 Maximally twisted eleven-dimensional supergravity

6.3 Eleven-dimensional supergravity in the pure spinor su-

perfield formalism

The canonical multiplet A• (R/I) associated to the nilpotence variety of the super Poincaré
algebra in eleven dimensions is the eleven-dimensional supergravity multiplet [How91a;
Ber02; CNT02]. In this chapter, we are interested in the computation of the maximal
twist in component fields. For this purpose, we need the full L∞ action of the super
Poincaré algebra on the component fields which we obtain using pure spinor superfield
techniques (in particular Corollary 3.4.8).

6.3.1 Representatives for component fields

In the following, let V11 = V and S11 = S be the vector and spinor representations of
Spin(11) respectively. As before, the component fields take values in the minimal free
resolution of R/I in free R-modules

µA• (R/I) ∼
= (L• ⊗R C) ⊗ C ∞ (V ) , (6.22)

In our case, the minimal free resolution L• takes the form

d1 d2 d3 d4

R⊗ C V ∧2 V ⊕ V ∧3 V ⊕ Sym2 (V ) ⊕ S S⊗V
d5

d6 d7 d8

S⊗V ∧3 V ⊕ Sym2 (V ) ⊕ S ∧2 V ⊕ V C .

(6.23)
The resolution differential was already described in [Ber02]. Let us choose a basis (eµ ) of
V and (sα ) of S. We will need the maps d1 , . . . d5 . In this basis they take the following
form.

(1)
d1 : V −→ C C (1) 7→ (λΓµ λ)Cµ

d2 : ∧2 V ⊕ V −→ V v 7→ (λΓµν λ)vµ eν
(2)
C (2) 7→ (λΓµ λ)Cµν eν

(3)
d3 : ∧3 V ⊕ Sym2 (V ) ⊕ S −→ ∧2 V ⊕ V C (3) 7→ (λΓµ λ)Cµνρ (eν ∧ eρ )
g 7→ ((λΓµ λ)eν + ηρσ (λΓσν λ)(eµ ∧ eρ )) gµν
(λΓµ )α eµ + 12 (λΓµν )α (eµ ∧ eν ) ω α


ω 7→
Maximally twisted eleven-dimensional supergravity 183

d4 : S ⊗ V −→ ∧3 V ⊕ Sym2 (V ) ⊕ S ψ 7→ −(λΓµ λ)ψµα sα + 12 (λΓµν )α (λΓµ )β ψνβ sα

+ 21 (λΓµ )α ψνα (e(µ ⊗ eν) )
+ 14 (λΓνρ )α ψµα eµ ∧ eν ∧ eρ

αβ
d5 : S ⊗ V −→ S ⊗ V ψ ∨ 7→ (λMµν λ)ψβ∨ν v µ ⊗ sα .
(6.24)
αβ
We do not specify the tensor Mµν here, but just remark that it is a rather complicated
expression in terms of Γ-matrices. The D0 -cohomology is bigraded by λ and θ. The
component fields organize according to degree in λ and θ according to Table 6.2. As
usual, the λ-degree is linked to the cohomological grading (or ghost degree). Here, they
are related by a shift of three, i.e. the physical fields in ghost degree zero sit in λ-degree
three.
θ
0 1 2 3 4 5 6 7 8 9
λ
0 C (0)
1 C (1)
2 C (2) , vµ ω
3 C (3) , gµν ψ
4 ψ∨ ∨
C (3)∨ , gµν

5 ω∨ C (2)∨ , vµ∨

6 C (1)∨
7 C (0)∨

Table 6.2: θ and λ degrees for the supergravity three-form BV multiplet

We can run the machinery developed in §2 to find explicit representatives for the com-
ponent fields. For example we find for the one-form

d D†
C (1) 7−→
1
(λΓµ λ)Cµ(1) 7−−→
0
(λΓµ θ)Cµ(1) , (6.25)

(1)
such that the one-form field is represented by (λΓµ θ)Cµ .

Similarly one finds for the two-form

d D† d D†
C (2) 7−→
2
(λΓµ λ)Cµν
(2) ν 0
e 7−−→ (λΓµ θ)Cµν
(2) ν 1
e 7−→ (λΓν λ)(λΓµ θ)Cµν
(2) 0
7−−→ (λΓν θ)(λΓµ θ)Cµν
(2)
,
(6.26)
(2)
such that the two-form is represented by (λΓν θ)(λΓµ θ)Cµν . Likewise, the three-form is
(3)
represented by (λΓν θ)(λΓµ θ)(λΓρ θ)Cµνρ .
184 Maximally twisted eleven-dimensional supergravity

Let us continue with the vector ghost v

d D† d D†
2
v 7−→ (λΓµν λ)vν eµ 7−−→
0
(λΓµν θ)vν eµ 7−→
1
(λΓµ λ)(λΓµν θ)vν 7−−→
0
(λΓµ θ)(λΓµν θ)vν .
(6.27)
Thus, the representative is (λΓµ θ)(λΓµν θ)vν . For the graviton we find with a similar
calculation (λΓµ θ)(λΓµ(ν θ)(λΓρ) θ)gρν .

Performing this procedure one can find representatives for the gravitino and its ghost.
The results are summarized in Table 6.3.
Field Representative in D0 -cohomology
C (0) C (0)
(1)
C (1) (λΓµ θ)Cµ
(2)
C (2) (λΓµ θ)(λΓν θ)Cµν
v (λΓµ θ)(λΓµν θ)vν
(λΓµ θ)(λΓµν θ)(θΓν )α + 12 (λΓµ θ)(λΓν θ)(θΓµν ) ω α
 
ω
(3)
C (3) (λΓµ θ)(λΓν θ)(λΓρ θ)Cµνρ
g (λΓµ θ)(λΓµ(ν θ)(λΓρ) θ)gρν
ψ [(λΓµ θ)(λΓν θ)(λΓρ θ)(θΓνρ )α − (λΓµ θ)(λΓνρ θ)(λΓν θ)(θΓρ )α ] ψµα

Table 6.3: Representatives for the fields in 11D supergravity organized by θ-degree.

6.3.2 The BV differential

The differential D acting on the component fields is obtained by transferring D1 to the

D0 -cohomology. In general, this is done by a homotopy transfer of D∞ -algebras but here
we are only interested in the lowest order term that acts on the representatives simply
by the usual formula of D1 ,
D1 = (λΓµ θ)∂µ . (6.28)

This gives part of the differential, that is first order in derivatives. For example, we can
act on the C (0) ghost
D1 (C (0) ) = (λΓµ θ)∂µ C (0) . (6.29)

Thus, we see that the differential corresponds to the de Rham differential. This obviously
generalizes to C (1) and C (2) such that we see that the ghost system of the three-form
indeed corresponds to the usual ghost system of a higher form field. Moving on to the
Maximally twisted eleven-dimensional supergravity 185

diffeomorphism ghost vµ for the graviton, we find

D1 ((λΓµ θ)(θΓµν θ)vν ) = (λΓµ θ)(θΓµν θ)(λΓρ θ)∂ρ vν . (6.30)

From our calculations of the representatives, we know that only the part where ρ and ν
are symmetrized corresponds to a non-trivial cohomology class. Thus, we find

D1 (v) = (λΓµ θ)(θΓµ(ν θ)(λΓρ) θ)(∂ρ vν + ∂ν vρ ) . (6.31)

Written dually in terms of operators, we find that the BV operator acts by

QBV gµν = ∂µ vν + ∂ν vµ = (Lv η)µν , (6.32)

which is indeed the expected gauge transformation for the graviton.

A similar story also holds for the gravitino and its ghost. There we find

D1 (ω) = (λΓρ θ) [(λΓµ θ)(λΓµν θ)(θΓν )α + (λΓµ θ)(λΓν θ)(θΓµν )α ] ∂ρ ω α . (6.33)

This gives a gauge transformation

QBV ψµα = ∂µ ω α . (6.34)

Thus, we see that D1 encodes the usual gauge transformations, expected for the field
content. Furthermore, one expects D1 to encode the Rarita–Schwinger equation between
the gravitino and its antifield. In addition, homotopy transfer is expected to induce a
second order differential giving the linearized equations of motions of the graviton and
the three-form field.

6.3.3 The action of supersymmetry

Using Corollary 3.4.8 as well as the explicit forms of the representatives, we are able to
deduce the L∞ action of the supersymmetry algebra on the component fields.

The three-form ghost system. We begin with the ghost system of the three-form.
From degree reasons, it is obvious that ρ(1) acts trivially on the ghost system for the
three-form. Thus, we have

ρ(1) (C (0) ) = ρ(1) (C (1) ) = ρ(1) (C (2) ) = 0 . (6.35)

186 Maximally twisted eleven-dimensional supergravity

However, this is corrected by higher order contributions. Examining the resolution dif-
ferential, we find maps

ρ(2) (Q, Q) = ι[Q,Q] : Ωi (M ) −→ Ωi−1 (M ) , (6.36)

for i = 1, 2, 3. Written dually for operators, this gives a supersymmetry transformation

rule
δCµ(i) = (Γµ )C (i) . (6.37)

However, these transformations will not cancel any components in the twist since there
the relevant supercharge satisfies [Q, Q] = 0 and thus the above maps all vanish.

The diffeomorphism ghost. The only non-derivative transformation for the diffeo-
morphism ghost appears in ρ(2) . It takes the form

ρ(2) (Q, Q)(v) = ρ(2) (Q, Q)((λΓµ θ)(θΓµν λ)vν )

(6.38)
= (λΓµ θ)(Γµν )vν

and thus gives a transformation rule

δCµ(1) = (Γµν )v ν . (6.39)

In addition, there is a ρ(1) -piece involving a derivative that can be seen to give rise to the
usual supersymmetry transformation between the diffeomorphism and supertranslation
ghost [Ber02]
1
δωα = − (Γµν )α ∂µ vν . (6.40)
2

The gravitino ghost. For the gravitino ghost, we obtain

1
ρ(1) (Q)(ω) = (λΓµ θ)(λΓµν θ)(Γν ω) + (λΓµ θ)(λΓν θ)(Γµν ω) . (6.41)
2

This gives two supersymmetry transformations

δvµ = Γµ ω
1 (6.42)
(2)
δCµν = Γµν ω .
2

In this way, one obtains the full higher order corrections to the supersymmetry trans-
formations and encode them in the differential δ. We summarize the full non-derivative
supersymmetry transformations in Table 6.4. These results first appeared in [Ber02]. In
Maximally twisted eleven-dimensional supergravity 187

Operator φ Transformation rule δφ

(1)
C (0) δC (0) = (Γµ )Cµ
(1) (2)
C (1) δCµ = (Γν )Cµν + (Γµν )v ν
(2) (3)
C (2) δCµν = 12 Γµν ω + (Γρ )Cµνρ + (Γ[µρ )g ρν]
v δvµ = Γµ ω + (Γν )gµν
ω δωα = (Γµ )ψαµ + 21 (Γµν )α (Γµ )β ψβν
(3)
C (3) δCµνρ = 14 Γ[µν ψρ]
g δgµν = 12 Γ(µ ψν)
αβ
ψ δψµα = (Mµν )ψβ∨ν

Table 6.4: Non-derivative supersymmetry transformations

addition, we list the transformations including derivatives for the gravitino and its ghost
in Table 6.5.
Operator φ Transformation rule δφ
ω δωα = (Γµν )α ∂µ vν
(4)
ψ δψµ = (Γνρστ
µ − 8Γρστ δµν )Gνρστ 

Table 6.5: Supersymmetry transformations with derivatives

6.4 Twisting the free theory

In this section, we will show that the fields of the twisted theory arrange into a differential
form
A ∈ Ω• (R7 ) ⊗ Ω0,• (C2 ) . (6.43)

The strategy to establish this result is clear: we restrict the supersymmetry transforma-
tions from Table 6.4 to our G2 × SU(2) invariant supercharge and look for fields that
form trivial pairs under δ. In the twisted theory these fields decouple and can be ne-
glected. To find such cancellations we have to decompose the field content as well as the
supersymmetry transformations equivariantly under G2 × SU(2) × U(1).

As a result, we will see that only certain components of the three-form, the three-form
ghost system, the gravitino, and the corresponding antifields play a role in the twisted
theory. These fields then arrange into the differential form described above. We will
188 Maximally twisted eleven-dimensional supergravity

further see that the twisted differential takes the form

Dtw = dR7 ⊗ 1 + 1 ⊗ ∂ C2 . (6.44)

Before we continue, let us briefly remark on the different gradings present in the un-
twisted and twisted theories. As a BV theory, eleven-dimensional supergravity comes,
by definition, with a Z × Z/2Z-grading by cohomological (ghost) degree and internal
parity. As described earlier, the ghost degree corresponds to the λ-degree up to a shift
by three. The maximal twist, viewed as an interacting BV theory, will only be graded by
Z/2Z. Nevertheless, it can be useful to consider Z-gradings on the fields of the twisted
theory for the purpose of the calculation (these are then broken by the interaction to
Z/2Z). This is mostly, because the fields of the twisted theory are naturally organized
by their form degrees (even though these are not compatible with interactions) and it is
instructive to see how these form degrees arise from the fields in the untwisted theory.

To this end, we can consider a Z × Z-grading on the fields of the untwisted theory (again,
ignoring the interactions) given by the λ-degree dλ and the U(1)L -charge dU(1)L . After
twisting, the new BV operator QBV + δQ breaks the Z × Z-grading on the space of fields
E to the Z-grading
dQ = dλ − dU(1)L , (6.45)

in the twisted theory. Note that Dtw is not homogenous with respect to this grading since
∂ C2 operator carries U(1)L -charge −1. The new degree of a component of A is simply
its de Rham form degree on R7 . Alternatively, note that the twisted BV differential
preserves the total form degree and we can assign a total form degree to the components
of A. We observe that for component fields in A the total form degree agrees with their
original θ-degree.

6.4.1 Decomposition of the field content

We now decompose the field content into representations of G2 × SU(2)− × U(1)L . To

do this, recall the following sequence of inclusions

Spin(11) ⊃ Spin(7) × SU(2)− × U(1)L ⊃ G2 × SU(2)− × U(1)L . (6.46)

The branching of the relevant representations from Spin(11) to Spin(7)×SU(2)− ×U(1)L

is described by Table 6.6. Here we are using Dynkin labels to identify the Spin(11) and
Spin(7) representations. We identify SU(2) × U(1)-representations by the dimension
of the SU(2)-representation and denote the U(1)-charge as a superscript. Recall that
the vector representation V has Dynkin label (10000) and its second and third exterior
Maximally twisted eleven-dimensional supergravity 189

Spin(11) Spin(7) × SU(2)− × U(1)L

(00000) (000)10
(10000) (000)(2−1 + 21 ) ⊕ (100)10
(00001) (001)(1−1 + 11 + 20 )
(01000) (000)(1−2 + 10 + 30 + 12 ) ⊕ (010)10 ⊕ (100)(2−1 + 21 )
(00100) (000)(2−1 + 21 ) ⊕ (002)10 ⊕ (010)(2−1 + 21 ) ⊕ (100)(1−2 + 10 + 30 + 12 )
(20000) (000)(3−2 + 10 + 30 + 32 ) ⊕ (100)(2−1 + 21 ) + (200)10
(10001) (001)(2−2 + 3−1 + 1−1 + (20 )⊕2 + 11 + 31 + 22 ) ⊕ (101)(1−1 + 20 + 11 )

Table 6.6: Branching of Spin(11) → Spin(7) × SU(2)− × U(1)L -representations.

powers are labeled by (01000) and (00100). The spinor representation S has Dynkin label
(00001). Furthermore, the gravitino representation already decomposes as a Spin(11)
representation according to

S⊗V ∼
= (00001) ⊕ (10001) . (6.47)

Finally, the graviton transforms in the representation

Sym2 V ∼
= (20000) ⊕ (00000) . (6.48)

We also need the branching rules for Spin(7) → G2 , which we collect in Table 6.7.

Spin(7) G2
(000) (00)
(100) (10)
(001) (10) ⊕ (00)
(010) (01) ⊕ (10)
(002) (00) ⊕ (10) ⊕ (20)
(101) (01) ⊕ (10) ⊕ (20)
(200) (20)

Table 6.7: Branching of Spin(7) → G2 -representations.

From these branching rules, we can already develop some expectation how the computa-
tion of the maximal twist could play out. This works on any product manifold M 7 × M 4
of G2 × SU(2)-holonomy. Clearly, the three-form and its ghosts C (p) split into forms
in Ωi (M 7 ) ⊗ Ωj1 ,j2 (M 4 ), where i + j1 + j2 = p is the total form degree. Thus, in the
190 Maximally twisted eleven-dimensional supergravity

light of the conjecture, we expect all components with non-zero holomorphic form degree
(j1 6= 0) to cancel in the twisted theory.

We now consider the decomposition of the gravitino field ψµα . It transforms in the prod-
uct of the Spin(11) vector and spinor representations. We first consider its decomposition
under Spin(11) → Spin(7) × SU(2)− . We will later see that the only components that
survive in the twisted multiplet have index µ transforming in a Spin(7)-vector represen-
tation whose components we denote by m.

On a manifold of G2 holonomy exterior powers of the cotangent bundle decompose into

irreducible G2 representations [Joy07]. This induces the following decomposition on
differential forms.

Ω01 Ω17 Ω27 Ω31 Ω41 Ω57 Ω67 Ω71

⊕ ⊕ ⊕ ⊕
Ω214 Ω37 Ω47 Ω514 .
⊕ ⊕
Ω327 Ω427

(6.49)
Here, we denote the sections of each irreducible piece by Ωkl , where the subscript denotes
the respective dimension of the G2 -representation.

The spin 1/2 and spin 3/2 fields on M 7 decompose as [CG+18; HS19]

Σ1/2 ∼
= Ω01 ⊕ Ω17 (6.50)
Σ3/2 ∼
= Ω17 ⊕ Ω214 ⊕ Ω327 . (6.51)

Using the above decomposition and the Spin(11) → Spin(7) × SU(2) × U(1)L branchings
in Table 6.6, and the isomorphisms

Σ3/2 ⊕ Σ1/2 ∼
= Ω17 ⊕ Ω214 ⊕ Ω327 ⊕ Ω01 ⊕ Ω17



(6.52)
∼
= Ω2 ⊕ Ω3 , (6.53)

we see that the gravitino, given by a pair of spin 3/2 and spin 1/2 fields on a G2 holonomy
manifold, can be identified with a pair of two- and three-forms on the manifold. We will
find that the components of the gravitino that survive the twist are contained in the
representation

(S+ ⊕ S− ) ⊗ (Σ3/2 ⊕ Σ1/2 ) ∼

= (S+ ⊕ S− ) ⊗ (Ω2 ⊕ Ω3 ) . (6.54)
Maximally twisted eleven-dimensional supergravity 191

However, not all of these components survive. We will find that the surviving compo-
nents are Ω3 ⊗ ∧0 L∨ 3 2 2 ∨
4 , Ω ⊗ S− , and Ω ⊗ ∧ L4 . The gravitino has λ-degree 3 in the
untwisted theory and the representations ∧0 L∨ 2 ∨
4 , S− , ∧ L4 have U(1)-charge −1, 0, and
1, respectively. Thus, their new twisted degree defined by (6.45) are 4, 3, and 2. The
components surviving the twist are therefore in Ω4 (M 7 )⊗Ω0,0 (M 4 ), Ω3 (M 7 )⊗Ω0,1 (M 4 ),
and Ω2 (M 7 ) ⊗ Ω0,2 (M 4 ), where we have used the isomorphism Ω3 ∼ = Ω4 to ensure that
the gravitino has its correct twisted degree.

The components of the three-form and its ghosts C (p) , p = 0 . . . 3 and the gravitino along
with their antifields that survive the twist therefore give exactly the right field content
to be described by a form
A ∈ Ω• (M 7 ) ⊗ Ω0,• (M 4 ). (6.55)

6.4.2 Decomposition of the supersymmetry transformations

We now determine the supersymmetry transformations for the twisting supercharge Q.

For the moment, we are only interested in the supersymmetry transformations without
derivatives since these are the ones responsible for the formation of trivial pairs. The
transformations with derivatives will later be used to determine the twisted BV differ-
ential. Recall that the spin representation S decomposes as

[(00) ⊕ (10)] (1−1 + 11 + 20 ) . (6.56)

This means that we can decompose the parameter  from Table 6.4 into

 → (− , + , α̇ , −m , +m , mα̇ ) . (6.57)

Here m is an index for the seven-dimensional representation of G2 . To act by Q, we

specify − = 1 and set all other components to zero.

On general grounds, these transformation take a very simple form. As explained above,
the supercharge Q is invariant under G2 × SU(2) and has U(1) charge −1. As a con-
sequence, δQ is an G2 × SU(2)-equivariant map. By decomposing the field content into
irreducible G2 × SU(2)-representations, δQ splits up as a map between these irreducibles.
However, since δQ is equivariant, we can apply Schur’s lemma and find, first, that there
can not be any non-trivial maps between non-isomorphic components and, second, trans-
formations between isomorphic G2 × SU(2)-representations are always of the form α · id
for some α ∈ C. Thus, to check whether there are any trivial pairs, we only have to
see if there is a non-vanishing map between isomorphic representations. In addition,
δQ carries a U(1)-charge that simply equals minus the number of ’s appearing in the
192 Maximally twisted eleven-dimensional supergravity

transformation, which can be used as a further criterion to establish that certain maps
vanish.

To check whether or not supersymmetry transformation yields a trivial pair we need to

decompose Γ-matrices.

Gamma matrix decomposition. In eleven dimensions the symmetric square of the

spin representation decomposes as

Sym2 S ∼
= V ⊕ ∧2 V ⊕ ∧5 V . (6.58)

Accordingly, there are maps denoted by Γµ , Γµν and Γµ1 ...µ5 given by projecting onto
the summands in this decomposition. So for example, Γµ is given by the composition

∼
=
Sym2 (S) V ⊕ ∧2 V ⊕ ∧5 V
Γµ
. (6.59)
V

Recall the spin representation S decomposes under G2 × SU(2) × U(1) as

S → 1−1 + 11 + 20 + (10)(1−1 + 11 + 20 ) . (6.60)

We are interested in − Γµ  and − Γµν , where − ∈ 1−1 in the above decomposition and
 is arbitrary. This means we are looking at a map 1−1 ⊗ S → V or 1−1 ⊗ S → ∧2 V,
respectively. The representations V and ∧2 V decompose as

V → 21 ⊕ 2−1 ⊕ (10)
(6.61)
∧2 V → (1−2 ⊕ 10 ⊕ 30 ⊕ 12 ) ⊕ (10)(2−1 ⊕ 21 ) ⊕ (10) ⊕ (01).

We can now compare this with the decomposition of 1−1 ⊗ S and read off the following
results for Γµ :
− Γµ − = 0
− Γµ + = 0
− Γµ α̇ ∈ 2−1
(6.62)
− Γµ +m ∈ (10)
− Γµ −m = 0
− Γµ mα̇ = 0.
Maximally twisted eleven-dimensional supergravity 193

For Γµν we find:

− Γµν − ∈ 1−2
− Γµν + ∈ 10
− Γµν α̇ = 0
(6.63)
− Γµν +m = 0
− Γµν −m = 0
− Γµν mα̇ ∈ (10) 2−1 .

For example, we immediately see that all terms of the form − Γµ − vanish and hence do
not affect the twist; this is of course nothing else but the condition for Q to be square
zero.

Let us start examining the supersymmetry transformations. Note that we are ignoring
any potential non-zero scalar coefficients α as we are only interested in the formation of
trivial pairs.

Furthermore, we are only considering cancellations between the fields of the multiplet as
well as between the gravitino and its antifield. Since the action of supersymmetry respects
the pairing on the BV complex, the same cancellations also occur for the respective
antifields.

The zero-form C (0) . For the zero-form ghost, we obviously have δQ C (0) = 0. Since
there is no supersymmetry transformation generating C (0) , it descends to a field in the
twisted theory.

The diffeomorphism ghost v. Next we consider the diffeomorphism ghost vµ . It

decomposes into components

vµ → (vm , v+α̇ , v−α̇ ) . (6.64)

We have a supersymmetry transformation of the form

δQ vµ = Γµ ω . (6.65)

The gravitino ghost ω lives in the spinor representation and hence decomposes according
to (6.57). From the Γ-matrix decomposition in (6.62), we know that − Γµ ω is only non-
vanishing for the components ωα̇ and ω+m of ω. Thus, we get up to potential non-zero
prefactors
δQ vm = ω+m (6.66)
194 Maximally twisted eleven-dimensional supergravity

and
δQ v−α̇ = ωα̇ . (6.67)

Finally we have,
δQ v+α̇ = 0 . (6.68)

Thus, we already find that some components of the diffeomorphism ghost v form trivial
pairs with parts of the gravitino ghost. In addition, it is interesting to note that δQ v+α̇ =
0. In light of the conjecture, we expect that v+α̇ will not be part of the twisted multiplet.
Hence, it should be in the image of δQ , forming a trivial pair with another field. Indeed,
we will momentarily find that v+α̇ cancels the holomorphic part of the one-form C (1) .

The one-form C (1) . For the field C (1) , we have a supersymmetry transformation rule

δQ Cµ(1) = (− Γµν − )v ν . (6.69)

From the Γ-matrix decomposition, we know − Γµν − ∈ 1−2 . Thus, we immediately find

(1)
δ Q Cm =0 (6.70)

and
(1)
δQ C+α̇ = 0 . (6.71)

In addition, we have
(1)
δQ C−α̇ = v+α̇ . (6.72)
(1)
This shows that C−α̇ and v+α̇ form a trivial pair and thus do not appear in the twisted
theory. Recall that the choice (− , + , α̇ ) = (1, 0, 0) defines a complex structure on
R4 ∼
= C2 . The four-dimensional vector representation decomposes as

V4 = S+ ⊗ S− = 21 ⊕ 2−1 . (6.73)

The representation 2−1 corresponds to holomorphic and 21 to the antiholomorphic com-

(1)
ponents. Thus, we see that, for this complex structure, the components C−α̇ form the
holomorphic parts of the one-form ghost C (1) . As expected, only the anti-holomorphic
part of the one-form plays a role in the twisted theory.

We can alternatively describe the cancellation using complex geometry. With respect to
the complex structure on C2 ,

Ω = (− Γµν − )dxµ ∧ dxν (6.74)

Maximally twisted eleven-dimensional supergravity 195

defines a holomorphic (2, 0)-form. Introducing coordinates (z α̇ , z̄ α̇ ) on V = 2−1 ⊕ 21 ,

the holomorphic (2, 0)-form simplifies to

Ω = dz 1 ∧ dz 2 . (6.75)

This allows us to rewrite the supersymmetry transformation of the one-form ghost as

δQ C (1) = ιv Ω = v+α̇ dz α̇ . (6.76)

Thus, we again see that the holomorphic components of C (1) cancel with the diffeomor-
phism ghost.

The two-form field C (2) . Let us continue with the supersymmetry transformation of
the two-form
(2) 1 ρ
δQ Cµν = − Γµν ω + − Γ[µρ − gν] . (6.77)
2
The two-form and the graviton decompose into components

(2) (2) (2) (2) (2) (2) (2) (2)

Cµν → (Cmn , Cm+α̇ , Cm−α̇ , C2 , C0 , C(α̇β̇) , C−2 )
(6.78)
gµν → (gmn , gm+α̇ , gm−α̇ , g2(α̇β̇) , g(α̇β̇) , g0 , g−2(α̇β̇) , h) .

Consulting the Γ-matrix decomposition in (6.63), we get

(2)
δQ Cmn =0
(2)
δQ C+mα̇ = 0
(2)
δQ C−mα̇ = ωmα̇ + g+mα̇
(2)
δ Q C2 = 0 (6.79)
(2)
δ Q C0 = ω+
(2)
δQ C(α̇β̇) = g2(α̇β̇)
(2)
δQ C−2 = ω− + g0 .

Thus, we find that the components

(2) (2) (2) (2)

C−mα̇ C0 C(α̇β̇) C−2 (6.80)

do not appear in the twisted multiplet, while

(2) (2) (2)

Cmn C+mα̇ C2 (6.81)
196 Maximally twisted eleven-dimensional supergravity

are in the kernel of δQ and thus, since there are no supersymmetry transformations that
could make these exact, part of the twisted multiplet. Note again that this matches with
the expectation that only (0, •)-forms on C2 play a role in the twisted multiplet.

Note that we can rewrite the piece of the supersymmetry transformation (6.77) involving
the graviton using the holomorphic (2,0)-form Ω as

δQ C (2) = ιgνρ ∂ρ Ω ∧ dxν . (6.82)

However, due to the symmetry properties of the graviton, this transformation alone
does not cancel all holomorphic component of the two-form. So one really needs the
(2)
supersymmetry ghost to cancel the singlet C0 .

The three-form field C (3) . For the three-form field, we have a supersymmetry trans-
formation of the form
(3) 1
δQ Cµνρ = − Γ[µν ψρ] . (6.83)
4
The three-form decomposes into components

(3) (3) (3) (3) (3) (3) (3)

Cµνρ → (Cmnp , Cmn+α̇ , Cmn−α̇ , Cm−2 , Cm0 , Cm(α̇β̇) ). (6.84)

To decompose this transformation, we write for the gravitino

ψµα = ξ α ⊗ χµ (6.85)

where ξ α takes values in S and χµ in V . From (6.63), we see that ξ α has to live in

1−1 ⊕ 11 ⊕ (10)20 (6.86)

to get a non-zero result. Decomposing (1−1 ⊕ 11 ⊕ (10)20 ) ⊗ V into irreducibles, we can

identify the decomposed transformations. The results are listed in Table 6.8.

The supersymmetry ghost ω. The non-derivative part of the supersymmetry trans-

formation of ωα reads
1
δQ ωα = (− Γµν )α (− Γµ ψν ) . (6.87)
2
Again decomposing the gravitino as we did for the three-form field and using the decom-
position (6.62), we find that ξ α has to take values in

20 ⊕ (10)11 . (6.88)
 Maximally twisted eleven-dimensional supergravity 197

Tensoring with the vector representation V and identifying matching representations

gives the result listed below.

The graviton gµν . The supersymmetry transformation

1
δQ gµν = − Γ(µ ψν) (6.89)
2

again only allows for ξ to come from 20 ⊕ (10)11 . As before, we just list the results in
Table 6.8.

In Table 6.8, we collect all decomposed non-derivative supersymmetry transformations.

There M is an index for the 14-dimensional representation (01) of G2 . It appears in the
variation
(3)
δQ Cmn−α̇ = ψM α̇ + ψmα̇ (6.90)

where the notation describes the decomposition ∧2 (10) → (10)⊕(01) of G2 -representations.

Operator φ Transformation rule δQ φ

C (0) 0
(1) (1) (0)
Cm , C+α̇ , C−α̇ 0, 0, v+α̇
(2) (2) (2) (2) (2) (2) (2)
Cmn , C+mα̇ , C−mα̇ , C2 , C0 , C(α̇β̇) , C−2 0,0, ωmα̇ + g+mα̇ , 0, ω+ , g2(α̇β̇) , ω− + g0
vm , v+α̇ , v−α̇ ω+m , 0, ωα̇
ω+ , ω− , ωα̇ , ω−m , ω+m , ωmα̇ 0, ψ+ , 0, ψ+m , 0, ψ2mα̇
(3) (3) (3) (3) (3) (3)
Cmnp , Cmn+α̇ , Cmn−α̇ , Cm−2 , Cm0 , Cm(α̇β̇) 0, 0, ψM α̇ + ψmα̇ , ψm− , ψm+ , ψm+(α̇β̇)
(3) (3) (3)
Cm2 , C−α̇ , C+α̇ 0, ψα̇ , ψ2α̇
gmn , gm+α̇ , gm−α̇ , g2(α̇β̇) , g(α̇β̇) , g0 , g−2(α̇β̇) , h ψmn+ , ψ2mα̇ , ψmα̇ , 0, ψ+(α̇β̇) , ψ+ , ψ−(α̇β̇) , ψ+
αβ
ψ δQ ψµα = (− Mµν − )ψβ∨ν

Table 6.8: Decomposed supersymmetry transformations

6.4.3 Supersymmetry variation of the gravitino

The non-derivative supersymmetry transformation of the gravitino reads

δψµα = (Mµν
αβ
)ψβ∨ν . (6.91)

This transformation reflects the fact that the supersymmetry algebra acts only up to
the equations of motions of the gravitino. Correspondingly, there is a quadratic term in
198 Maximally twisted eleven-dimensional supergravity

antifields appearing in the BV action [Bau+90; Ber02]

S (2) ∝ (M )ψ ∨ ψ ∨ . (6.92)

The transformation (6.91) is responsible for the remaining cancellations between of the
gravitino. To argue that indeed the correct components of ψ cancel, we change our
αβ
strategy. As the structure of Mµν is very complicated, we will not decompose it directly
under G2 × SU(2). Instead we give an indirect argument.

For this, recall that (6.91) precisely represents the homotopy correcting for the failure
of the linear supersymmetry transformations to define a strict representation. Denot-
lin and the quadratic
ing the linearized part of the supersymmetry transformation by δQ
quad
transformation of the gravitino by δQ , we have

quad
lin lin
[δQ lin
, δQ ]ψ = δ[Q,Q] ψ + δQ QBV ψ ∨
quad
= δQ QBV ψ ∨ (6.93)
= (− M − )QBV ψ ∨ ,

where we have used the fact that Q is square zero in the second equality.

Thus, we can try to learn something about the quadratic transformation by studying
two consecutive linear transformations applied to the gravitino. Recall that a linear
transformation applied to the gravitino transforms it to the field strength of three-form,

lin
δQ ψµ = (Γνρστ
µ − 8Γρστ δµν )G(4)
νρστ − , (6.94)

while the three-form transforms to the gravitino

lin (3) 1
δQ Cµνρ = − Γ[µν ψρ] (6.95)
4

Decomposing the gravitino and applying these tranformations, there are two distinct
cases: Whenever the result is non-zero, the linearized piece fails to be a Lie map and
a homotopy is present for such components. In other words, evaluating the quadratic
transformation (6.91) on such a component gives a non-zero result and a trivial pair
forms. On the other hand, when a component is in the kernel of two consecutive linear
transformations, then a homotopy is not strictly necessary and it is possible that the com-
ponent is also in the kernel of the quadratic transformation such that the corresponding
field descends to the twisted theory.

This reasoning suggests to view the cancellations between components of the gravitino
and its antifield as a two-step procedure. First, the linearized transformation identifies a
piece of ψ with a component of G(4) = dC (3) . Then we can act with another linearized
Maximally twisted eleven-dimensional supergravity 199

transformation to obtain a component of ψ ∨ . Clearly the U(1)-charges of components

connected in this way satisfy

dU(1) (ψ ∨ ) = dU(1) (G(4) ) + 1 = dU(1) (ψ) + 2 . (6.96)

To investigate the kernel of two consecutive linearized transformations, recall from Ta-
ble 6.8 that the components components

(3) (3) (3)

Cmnp , Cmn+α̇ , Cm2 (6.97)

are in the kernel of δQ . They correspond to the differential forms

Ω3 (R7 ) ⊗ Ω0,0 (C2 ) ⊕ Ω2 (R7 ) ⊗ Ω0,1 (C2 ) ⊕ Ω1 (R7 ) ⊗ Ω0,2 (C2 ). (6.98)

Investigating the complex of differential forms, it is easy to see that the components of
the field strengths living in

Ω4 (R7 ) ⊗ Ω0,0 (C2 ) ⊕ Ω3 (R7 ) ⊗ Ω0,1 (C2 ) ⊕ Ω2 (R7 ) ⊗ Ω0,2 (C2 ) (6.99)

can only arise from the three-form components above. In particular, these components
of the field strength are then also in the kernel δQ . Components of the gravitino who are
transformed to such a component of the field strength by the first linear supersymmetry
transformation are annihilated by the second one. These components thus are expected
to descend to the twisted theory.

With this information, we can analyze the remaining compoments of the gravitino. In
Table 6.9, we display the G2 × SU(2)-equivariant decomposition of the gravitino, its
antifield, and the field strength organized by U(1)-charges. All components of ψ and
ψ ∨ that form trivial pairs with other fields according to Table 6.8 are indicated with an
arrow.

We circle the components of the field strength which are in the kernel with red dashed
lines and the corresponding components of the gravitino in blue. We then expect these to
descend to the twisted theory. In addition, we also circle the surviving dual components
of the gravitino antifield in blue. For example, with U(1)-charge 1, there appears a
representation
(10)1 ⊕ (01)1 (6.100)

in the decomposition of the gravitino which can be identified with Ω2 ⊗Ω0,2 after applying
lin . Indeed, these components descend to the twist as they cannot form any trivial pairs
δQ
with any components from the gravitino antifield due to their U(1)-charges alone.
200 Maximally twisted eleven-dimensional supergravity

Similarly, we circle pieces in blue with U(1)-charge 0 and −1 which can be identified
with the differential forms Ω3 ⊗ Ω0,1 and Ω2 ⊗ Ω0,2 respectively.

On the other hand, we see that different pieces of the gravitino are mapped to components
lin . These then can have
of the field strength which are not part of the kernel of δQ
lin , δ lin ]ψ 6= 0, such that a cancellation is possible. In Table 6.9 we indicate such
[δQ Q
components, the corresponding intermediate components of the field strength and the
respective partners from ψ ∨ with green rectangles.

In this way, one can understand all cancellation except one subtlety. For U(1)-charge
zero, there is a leftover representation (00)2. We expect that this component of the
gravitino cancels with the respective component of the antifield with U(1)-charge 2.
However, since the only field strength component which could serve as intermediary is
lin , we cannot understand this cancellation in the above manner. For a
in the kernel of δQ
more complete understanding, a direct investigation of the homotopy seems necessary.

6.4.4 Summary of cancellations

We summarize the cancellations obtained in the previous sections in Table 6.10. The
fields that do not form trivial pairs are circled in blue. They form the multiplet A ∈
Ω• (R7 ) ⊗ Ω0,• (C2 ) and appear in Table 6.1. The bi-directional strike-through arrows
indicate cancellations that occur between ψ and its anti-field ψ ∨ found in §6.4.3.

Special care should be taken for the variations of the components of C (2) that cancel
with a linear combination of components of the graviton and supersymmetry ghost

(2)
δQ C−mα̇ = ωmα̇ + g+mα̇ (6.101)
(2)
δQ C−2 = ω− + g 0 (6.102)

that occur in (6.79). A subsequent variation yields

δQ ωmα̇ = −δQ g+mα̇ = ψ2mα̇ (6.103)

δQ ω− = −δQ g0 = ψ+ (6.104)

2 C (2) = 0. These extra cancellations are indicated by the

which is consistent with δQ
strike-through arrows with labels x and y.
 Field 2 1 0 -1 -2

ψ (00)2 ⊕ (10)2 (10)1 ⊕ (01)1 (00)2 ⊕ (10)2 ⊕ (01)2 (10)1 ⊕ (01)1 (00)2 ⊕ (10)2

(00)1 ⊕ (10)1 ⊕ (20)1 (00)2 ⊕ (10)2 ⊕ (20)2 (00)1 ⊕ (10)1 ⊕ (20)1

(00)3 ⊕ (10)3 (10)2 (00)3 ⊕ (10)3

(00)1 ⊕ (10)1 (00)2 ⊕ (10)2 (00)1 ⊕ (10)1

G(4) (10)1 ⊕ (01)1 (00)2 ⊕ (10)2 ⊕ (20)2 (00)1 ⊕ (10)1 ⊕ (20)1 (20)2 ⊕ (10)2 (10)1 ⊕ (01)1
(10)2 (00)1 ⊕ (10)1 ⊕ (10)3 (00)2 ⊕ (10)2
(01)1 ⊕ (01)3

ψ∨ (00)2 ⊕ (10)2 (10)1 ⊕ (01)1 (00)2 ⊕ (10)2 ⊕ (01)2 (10)1 ⊕ (01)1 (00)2 ⊕ (10)2

201
(00)1 ⊕ (10)1 ⊕ (20)1 (00)2 ⊕ (10)2 ⊕ (20)2 (00)1 ⊕ (10)1 ⊕ (20)1

(00)3 ⊕ (10)3 (10)2 (00)3 ⊕ (10)3

(00)1 ⊕ (10)1 (00)2 ⊕ (10)2 (00)1 ⊕ (10)1

Table 6.9: Decomposition of the non-linear gravitino supersymmetry variation

 Field Spin(11) 2 1 0 -1 -2
C (0) (00000) (00)1
a
C (1) (10000) (00)2 (10)1 (00)2

b c d, x d, y
C (2) (01000) (00)1 (10)2 (00)1 ⊕ (00)3 ⊕ (10)1 ⊕ (01)1 (10)2 (00)1

a e f
v (10000) (00)2 (10)1 (00)2

b e f g, x h, y h
ω (00001) (00)1 ⊕ (10)1 (00)2 ⊕ (10)2 (00)1 ⊕ (10)1
i k k k l
C (3) (00100) (10)1 (00)2 ⊕ (10)2 ⊕ (01)2 (00)1 ⊕ (10)1 ⊕ (20)1 (00)2 ⊕ (10)2 ⊕ (01)2 (10)1
j j

202
(10)1 ⊕ (10)3
c d d m m o p
g (20000) (00)3 (10)2 (00)1 ⊕ (00)3 ⊕ (20)1 (10)2 (00)3
n
(00000) (00)1

i g k k k l q q q
ψ (10001) (00)2 ⊕ (10)2 (10)1 ⊕ (01)1 (00)2 ⊕ (10)2 ⊕ (01)2 (10)1 ⊕ (01)1 (00)2 ⊕ (10)2
h h m
(00)1 ⊕ (10)1 ⊕ (20)1 (00)2 ⊕ (10)2 ⊕ (20)2 (00)1 ⊕ (10)1 ⊕ (20)1
m j o p q
(00)3 ⊕ (10)3 (10)2 (00)3 ⊕ (10)3
n j q q q q
(00001) (00)1 ⊕ (10)1 (00)2 ⊕ (10)2 (00)1 ⊕ (10)1

Table 6.10: Cancellations of fields under Q. Fields are decomposed into G2 × SU(2)− × U(1)L -representations.
Maximally twisted eleven-dimensional supergravity 203

6.4.5 The twisted differential

Recall that the BV differential of the twisted theory is the sum of two terms

Qtw
BV = QBV + δQ . (6.105)

We already examined how the non-derivative part of δQ leads to the formation of various
trivial pairs; now we turn towards the parts containing derivatives in order to see how
they act on the twisted multiplet.

The BV operator Qtw

BV is dual to a differential D
tw acting on the fields of the twisted

multiplet. We already know that D acts as the de Rham differential on the three-form
ghost system. Under G2 × SU(2) the de Rham differential decomposes

d = dR7 + ∂¯C2 + ∂C2 . (6.106)

As only (0, •)-forms are part of the twisted multiplet, this restricts to

dR7 + ∂¯C2 . (6.107)

In addition, D acts on the gravitino by the Rarita–Schwinger equation. Identifying the

gravitino as a spinor valued one-form, ψ ∈ Ω1 (M ) ⊗ S, the Rarita–Schwinger operator
can be understood as a composition of the exterior differential and Clifford multiplica-
tion [HS19]. From this, one can see that it also acts by dR7 + ∂¯C2 on the relevant pieces
of the gravitino.

Finally, there is a contribution to Dtw coming from the supersymmetry transforma-

tion (6.94). This transformation also acts by dR7 + ∂¯C2 and provides the missing differ-
ential between C (3) and ψ.

In summary, the twisted multiplet can thus be described by the cochain complex

Ω• (R7 ) ⊗ Ω0,• (C2 ) , Dtw = dR7 + ∂¯C2 ,

(6.108)

as conjectured by Costello.

Interestingly, the form of the differential can also be deduced directly from the explicit
formulas in the pure spinor formalism. Recall that D1 acts on the representatives by

D1 = (λΓµ θ)∂µ , (6.109)

(1)
and that the one-form was represented by the cohomology classes Cµ (λΓµ θ). As we
already know that the twisted multiplet forms the exterior algebra Ω• (R7 ) ⊗ Ω0,• (C2 ),
204 Maximally twisted eleven-dimensional supergravity

we see that D1 simply acts by taking derivatives and wedging with the corresponding
component of the one-form, i.e. precisely by dR7 + ∂¯C2 .

In addition the derivative part of the supersymmetry transformation acts by

Q∂x = (− Γµ θ)∂µ . (6.110)

From the Gamma matrix decomposition (6.57), we see

(− Γµ θ) ∈ 2−1 ⊕ (10) . (6.111)

Identifying the corresponding components with dz̄ α̇ and dxm , we once again see that Q∂x
acts as desired.

Another more roundabout way of understanding the appearance of the de Rham differ-
ential is as follows. Recall that the gravitino field on M 7 can be organized into Ω2 ⊕ Ω3
when M 7 has G2 holonomy. Since there are b2 (M 7 )+b3 (M 7 ) zero modes of the gravitino
on M 7 [Fon10; CG+18; Wan91; HS19], this suggests that the BV differential acts by the
de Rham differential
ddR : Ω2 ⊕ Ω3 → Ω3 ⊕ Ω4 . (6.112)

We note that the appearance of the de Rham and Dolbeault differential is similar to
the holomorphic twist of ten-dimensional abelian super Yang–Mills theory on C5 (see
[ES19a]). In that case, the analogous BV differential between the gaugino χ and its
antifield expresses the Dirac equation. The relevant part of the differential in the twisted
theory is
QBV (χmn )∨ = imnpqr ∂ p χqr , (6.113)

and only involves the Dolbeault operator on Ω0,• (C5 ).

6.5 Conclusions and future directions

While the above calculation establishes the maximal twist on the level of the free theory,
the interactions remain opaque. The main advantage of the component field approach is
its immediacy. We could explicitly see how the fields of the physical theory arrange to
the twisted theory and thereby obtains direct insights on the twisted degrees of freedom.
However, as the interactions of supergravity theories expressed in component fields can
be quite complicated (in particular they are non-polynomial in the metric), approaching
them in the same fashion seems out of reach at the moment. In the next chapters, we
fully leverage the pure spinor superfield formalism to address the twist of the interacting
theory at pure spinor cochain level rather than at component field level.
Maximally twisted eleven-dimensional supergravity 205

First hints can already be obtained by looking at “field strength” formulations of su-
pergravity; in eleven-dimensional supergravity there is a “super-vielbein” multiplet (as
opposed to the “three-form multiplet” we studied in this chapter). The super-vielbein
multiplet contains the graviton, gravitino, and 4-form field strength G(4) as its physical
fields. It is used in the traditional superspace formulation of supergravity. It is natural to
expect that the twisted fields of the super-vielbein multiplet organize into a differential
form
∂A ∈ Ω• (M 7 ) ⊗ Ω1,• (M 4 ), (6.114)

with leading component v+α̇ from the diffeomorphism ghost. In fact, we will see that
precisely this happens and that the super-vielbein multiplet is best thought of as being
associated to the first Chevalley–Eilenberg cohomology of the residual supertranslation
algebra and is indeed a field strength in the sense discussed in §3 (see also [CNT02]).

In general, addressing twists directly in component field could prove to be useful in linking
mathematical and physical approaches to the holographic duality. For instance, cojec-
tural twists of type IIB supergravity were described in [CL15; CL16] and for the minimal
twist of eleven-dimensional supergravity in [RSW23]. In a particular limit, holographic
duality relates weakly coupled type IIB supergravity on products of five-dimensional AdS
space AdS5 with arbitrary Sasaki-Einstein manifolds SE 5 to four-dimensional supersym-
metric gauge theories. A different form of the conjecture relates the weak coupling limit
of M-theory on the products AdS4 × SE 7 to three-dimensional supersymmetric gauge
theories. The cone over the Sasaki–Einstein manifold is a local Calabi–Yau manifold.
One corollary of the conjecture is the equivalence of the superconformal index [Rö06;
Kin+07] under gauge-gravity duality. The gravity superconformal index was computed
in terms of holomorphic invariants of the Calabi–Yau manifold in [EST14; ES15]. The
corresponding field theory index was later shown to be most directly computed in the
holomorphic twist [ES19a; SW23c]. In [RW22] the minimally twisted eleven-dimensional
supergavity theory described in [RSW23] was used to compute superconformal indices
and matched with corresponding results from the physics literature. Twist computations
in component fields naturally bridge the gap between these different versions of index
calculations and therey between physical and mathematical approaches to holography.

In addition, one expects that a further twist of the one considered in this chapter can
be used to derive twisted M-theory in the Ω-background [Cos16] following [OY19]. This
could provide a physical origin for the applications in [GO19; OZ21] by coupling a twisted
M5-brane [SW23b] to twisted M-theory. Finally, we hope that twisted M-theory can
shed new light on topological M-theory [Hit00; GS04; Dij+05; GV05; Bec+16], which is
believed to unify the Kähler [BS96] and Kodaira–Spencer theories of topological gravity.
Chapter 7

Eleven-dimensional supergravity as
a Calabi–Yau twofold

7.1 Introduction

Since the first supersymmetric field theories were constructed, it has been a goal to
understand their properties and simplify their construction using superspace techniques.
This motivation has perhaps been largest in the case of supergravity theories. The
geometric nature of the theory of Einstein gravity, which is constructed using a covariant
least-action principle on the space of metrics of Lorentzian signature, has motivated
much research that tries to give an equally pithy formulation of supergravity theories as
governing moduli problems of (deformations of) particular natural geometric structures
on superspace.

Among all supergravity theories of physical interest, perhaps the most exceptional is
eleven-dimensional supergravity, which was first constructed by Cremmer, Julia, and
Scherk in 1978 [CJS78], and which is expected to be the low energy limit of M-theory [Wit95].
M-theory has yet to be constructed, although expectations exist that a worldsheet
construction as a theory of fundamental membranes might be possible. While the
component-field formulation of this theory is relatively streamlined—in addition to the
metric, the theory contains only a gravitino and an abelian three-form gauge field with
Chern–Simons term—it proved difficult to even formulate the theory in superspace, and
a superspace least action principle remained out of reach. Part of the difficulty can be
attributed to attempts to find sets of auxiliary fields that could be used to represent
supersymmetry off shell, which was seen as a necessary prerequisite.

207
208 Eleven-dimensional supergravity as a Calabi–Yau twofold

A major leap forward was taken in work of Cederwall, who applied the pure spinor super-
field formalism to construct a superspace description of perturbative eleven-dimensional
supergravity using the BV formalism. The relation of eleven-dimensional pure spinors
to supergravity dates back at least to [How91b]. The connection had been sharpened
in [CNT02], which observed that a particular eleven-dimensional pure spinor superfield
reproduced the BV supergravity multiplet. In [Ced10c], a candidate cubic interaction
term for this multiplet was constructed; in [Ced10a], Cederwall went on to extend this
by a somewhat subtle quartic term in the BV action functional, and to prove that
the result satisfies the BV master equation, thus giving a consistent, manifestly super-
symmetric interacting theory that—since the theory is expected to be unique—must
be eleven-dimensional supergravity itself. (Pure spinor techniques were also used from
a first-quantized perspective to give new models of the supermembrane; see Berkovits’
work in [Ber02], generalizing his formulation of the superstring.) The pure spinor descrip-
tion thus not only formulates the theory on superspace, but also dramatically simplifies
the structure of its interactions: a non-polynomial action for the component fields is
replaced by a quartic polynomial. Nonetheless, it does not provide a geometric origin for
the quartic polynomial in question. Neither does it give an interpretation of the moduli
problem it describes in terms of deformations of the superspace geometry itself.

Later, and in disjoint fashion, further progress was made on twisted versions of eleven-
dimensional supergravity. Twists of supergravity theories were defined by Costello and Li
in [CL16], generalizing the standard notion of a twist of a supersymmetric field theory.
Using worldsheet techniques from topological string theory, they gave a proposed de-
scription of the holomorphic twist of type IIB supergravity. Costello and Li’s theory is a
version of BCOV theory [Ber+94], for which the moduli-theoretic interpretation is clear;
it is related to the Kodaira–Spencer theory of deformations of Calabi–Yau structure.
In [Cos16], Costello went on to investigate eleven-dimensional supergravity in the omega
background; his proposed description links the maximal twist of eleven-dimensional su-
pergravity to Poisson–Chern–Simons theory.

Poisson–Chern–Simons theory is simple to describe in the BV formalism. Its fields are

given by the Dolbeault complex of (0, •)-forms on a Calabi–Yau twofold, tensored with
the de Rham complex on R7 (or, more generally, a G2 -manifold; for nonperturbative
issues related to G2 -manifolds, see [Dij+05; DZOZ22] and references therein). The inter-
actions are determined by an L∞ structure on the fields, which is in fact strict: the Lie
bracket is the Poisson bracket of holomorphic functions induced by the Calabi–Yau form,
whose inverse is a holomorphic Poisson bivector. This theory has two essential features.
Firstly, it also has a moduli-theoretic interpretation. The Lie algebra of holomorphic
functions with the Poisson bracket is a one-dimensional central extension of holomorphic
Hamiltonian vector fields. Since the symplectic structure is the holomorphic volume
Eleven-dimensional supergravity as a Calabi–Yau twofold 209

form, these are also divergence-free vector fields, and can thus be also thought of as
related to the moduli space of deformations of Calabi–Yau structures. Secondly, the
central extension equips the fields of Poisson–Chern–Simons theory with a commutative
structure; the interactions define not just a dg Lie structure, but a dg Poisson algebra
structure. Recalling that the observables of a three-dimensional TQFT are equipped
with an E3 -algebra structure, which is equivalent to an even-shifted Poisson structure,
we see that this formulation is at least suggestive of a first-quantized origin. (Note,
though, that there are subtleties in defining E3 algebra structures on theories of this
type; see [EW21].)

Recent work has pushed our understanding of twisted eleven-dimensional supergravity

further; all approaches have either used dualities or target-space techniques, since no
worldsheet description is available. Using the component field formulation on target
space, we computed the maximal twist in the free limit in §6. Pure spinor techniques
were applied in [SW21] to give concise and computationally straightforward descriptions
of the twists of supergravity multiplets. This led to the first direct computations of the
minimally twisted eleven-dimensional and type IIB supergravity multiplets, the latter
confirming Costello and Li’s proposal at the free level. In [RSW23], a consistent inter-
acting Z/2Z-graded BV theory was defined on the minimally twisted eleven-dimensional
supergravity multiplet. Surprisingly, the cohomology of this theory on flat space is a
one-dimensional L∞ central extension of the exceptional infinite-dimensional simple su-
per Lie algebra E(5|10) [Kac77]. Other exceptional simple super Lie algebras also play
fundamental roles in holomorphic M-theory [RW22; SW23a].

In this chapter, we take a step towards bringing some of these lines of work together
by exploiting a powerful and seemingly underappreciated analogy between the geometric
structures in play on each case. Thinking of Poisson–Chern–Simons theory (after local-
izing six directions with omega backgrounds) as a theory in five dimensions, we note
that the theory must be equipped with a transversely holomorphic foliation that lets
us think of the geometry as locally isomorphic to C2 × R. The THF structure is an
(involutive) three-dimensional subbundle of the complexified tangent bundle. Similarly,
the minimally twisted theory is most generally defined on eleven-dimensional manifolds
equipped with a six-dimensional complex distribution.

Flat superspace itself is also canonically equipped with a distribution, spanned by the
left–invariant odd vector fields. However, since all bosonic translations are in the image
of brackets of supersymmetry transformations, this distribution is as far from being
integrable as possible. It is thus not possible to naively draw a connection between
these two structures. A clue to the resolution is provided by the theory of Dolbeault
cohomology for almost complex manifolds, recently developed in [CW21]. This theory
210 Eleven-dimensional supergravity as a Calabi–Yau twofold

uses the distribution T (0,1) to define a filtration of the de Rham complex. The differential
on the associated graded measures the nonintegrability of the distribution; passing to its
cohomology and transferring the D∞ structure defined by the remaining terms in the de
Rham differential provides a new filtered complex, which they use as a replacement for
the Hodge filtration. Passing to the associated graded of this new filtration defines their
analogue of the Dolbeault complex.

If we apply the same construction to the de Rham complex on superspace, we can identify
the term in the differential encoding the nonintegrability of the odd distribution with the
Chevalley–Eilenberg differential of the supertranslation algebra. The “generalized Dol-
beault complex” that appears is nothing other than the sum of the pure spinor multiplets
associated to the cohomology groups of the supertranslation algebra; the cohomology in
degree −k plays the role of the Dolbeault complex resolving holomorphic (k, 0)-forms.
In particular, the canonical supermultiplet of [Ced+23] appears playing the role of the
holomorphic functions, and we think of it—equipped with its commutative structure—as
the appropriate structure sheaf with which to equip the spacetime. In eleven dimensions,
this is eleven-dimensional supergravity.

The analogy with complex geometry allows one to find ready generalizations of many
interesting notions: the complex dimension is the degree of the highest Lie algebra
cohomology of the supertranslations; a Calabi–Yau structure is a trivialization of (the
multiplet of) top cohomology as a module over the structure sheaf. In this analogy,
eleven-dimensional supergravity, and all of its twists, are Calabi–Yau twofolds. We use
this to construct a family of theories we call homotopy Poisson–Chern–Simons theories.
The construction uses the derived bracket technique of [KS96], as generalized by [Vor05],
and is entirely analogous to the standard construction of the Poisson bracket. However,
because we work in a derived setting, the corresponding L∞ structure is in general not
strict. Applying our construction recovers Cederwall’s quartic interaction functional in
geometric fashion, as well as Costello’s maximal twist. Furthermore, it gives a pure
spinor lift of the interactions of the minimal twist. It then follows from the results
of [SW21], which state that the twist of a canonical multiplet is the canonical multiplet
of the twisted supersymmetry algebra, that these theories are all related by twisting,
proving Costello’s conjecture on the maximal twist at the full interacting level.

7.2 Flag structures and generalized Dolbeault complexes

Throughout, we work in the category of graded super vector spaces, often equipped with
a G-action. The grading and the parity are independent; thus, an object is graded by
Z × Z/2Z, and the Z/2Z factor determines the monoidal structure. We will also consider
Eleven-dimensional supergravity as a Calabi–Yau twofold 211

cochain complexes; these are then equipped with three integer gradings, called cohomo-
logical degree, weight grading, and intrinsic parity, and the Koszul sign is determined by
the totalization of cohomological degree and intrinsic parity. Our conventions are always
cohomological.

7.2.1 Weighted flag structures

We begin with some very general considerations, related to the type of geometric intu-
ition we will draw on in the sequel. The essential point is to notice that certain (super
or graded generalizations of) filtered structures, as studied by Tanaka, are present both
on (almost) complex manifolds and on the superspaces of interest in physics. (Other
examples abound, but these are the two that will interest us here.) The resulting anal-
ogy between superspaces and almost complex manifolds will let us construct a sheaf
of commutative differential graded algebras on such a manifold, which reproduces Dol-
beault cohomology for complex manifolds, as well as its generalization to almost complex
manifolds as defined in [CW21]. When we apply our techniques to superspaces, the con-
struction naturally reproduces a particular supermultiplet in the pure spinor formalism.
This is the multiplet assigned to the structure sheaf of the nilpotence variety, termed the
canonical multiplet in [Ced+23].

Geometrically, we will be interested in manifolds (including supermanifolds or graded

manifolds) that are equipped with distributions. Recall that a distribution on a manifold
M is a subbundle D ⊂ T M of the tangent bundle. A distribution is said to be involutive
if the space of vector fields lying in D is a subalgebra of vector fields on M with respect
to the Lie bracket. By Frobenius’ theorem, involutive distributions are integrable, i.e.
there exists a submanifold of M whose tangent bundle is D.

More generally, we can consider a flag of distributions, which is defined to be a finite

sequence
0 ⊂ D1 ⊂ · · · ⊂ Dk = T M (7.1)

of subbundles of the tangent bundle, each contained in the next. We require that this
flag is chosen to be compatible with the Lie bracket of vector fields, in the sense that

[Γ(Di ), Γ(Dj )] ⊂ Γ(Di+j ). (7.2)

When k = 2, this condition is vacuous; all of our examples will be of this type, so that
only a single distribution D1 is relevant. In any case, the flag of distributions gives
Vect(M ) the structure of a filtered Lie algebra.
212 Eleven-dimensional supergravity as a Calabi–Yau twofold

Given a flag of distributions in the tangent bundle, we can ask what corresponding
structure appears on the de Rham forms of M . To do this, we can filter Ω• (M ) in the
following way. We observe that the cotangent bundle is equipped with a dual series of
quotients of the form

T ∗ M = Dk∨ → Dk−1
∨
→ · · · → D1∨ → 0. (7.3)

Since we want to filter the de Rham forms by subalgebras, rather than by successive
quotients, we define a negatively graded filtration on the cotangent bundle by taking

F −i T ∗ M = ker(T ∗ M → Di∨ ), (7.4)

with respect to the map defined by (7.3). We can extend this multiplicatively to a non-
positive filtration F • Ω• (M ) of the de Rham forms, which is then automatically preserved
by the differential.

As an example, consider the flag of distributions on an almost complex manifold defined

by taking
0 ⊂ D1 = T (0,1) X ⊂ D2 = TC X. (7.5)

The filtration F • can be thought of as assigning weight −1 to dz̄ and weight −2 to dz.
As is clear from (7.19) below, this filtration is compatible with the de Rham differential
for any almost complex structure.

Note that this filtration, although it is compatible with the de Rham differential, is
not the standard Hodge filtration. Nor is it particularly convenient in applications. To
recover the Hodge filtration, one needs to construct a new filtration F+• Ω• (M ), defined
by taking
M
F+i Ω• (M ) = F j Ωk (M ). (7.6)
k+j=i

In the example of an almost-complex manifold, we then have that

F+−i Ω• (M ) = Ω≥i,• (M ). (7.7)

When the complex structure is not integrable, the de Rham differential does not preserve
F+• ; see §7.2.2 below.

Compatible weight gradings. Matters are simplified when we have a decomposition

of the tangent bundle via a positive integer grading that induces the flag of distributions
we are interested in. We will refer to such a grading as a weight grading. It consists of a
Eleven-dimensional supergravity as a Calabi–Yau twofold 213

direct sum decomposition of the tangent space of the form

M
TM = Tj M, (7.8)
1≤j≤k

such that the flag of distributions we are interested in is recovered by taking

M
Dk = Tj M. (7.9)
1≤j≤k

In fact, both for almost complex manifolds and superspaces, there is a canonical choice
of such a splitting: in the first case, we take the eigenspaces of J, and in the second, we
take the eigenspaces of the parity operator (−)F .

Motivated by the previous considerations we now give definitions which are meant to
abstractly model the structures that are present on the de Rham complex of a manifold
equipped with a flag of distributions (and perhaps with a compatible weight grading).

Definition 7.2.1. Let (Ω• , d) be a cdga. A flag structure on Ω• is a decreasing filtration

F+• Ω• of finite length that is compatible with the differential. A weighted flag structure on
Ω• consists of a weight grading for Ω• in non-positive degrees, with respect to which the
differential decomposes into pieces of non-positive weight. In other words, the differential
preserves the decreasing filtration associated to the weight degree.

From our perspective, there are (at least) three important and natural examples of flag
structures. The first of these is an essential motivating example: the Hodge filtration on
the de Rham complex of an (almost) complex manifold. The second is more obviously
related to the examples related to supersymmetric field theory that we have in mind as
applications: any flat superspace is equipped with a canonical distribution, defined by
considering the span of all translation-invariant odd vector fields. More generally, the su-
permanifolds that are valid backgrounds for supersymmetric field theories or supergravity
theories are equipped with a maximally non-involutive odd distribution, modelling the
local supersymmetry transformations. (This is well-known; consider, for example, the
definition of a super Riemann surface [Fri+86; RSV88; Wit19]. The idea goes back at
least to Manin in [Man85; Man84].) The third centers around the observation that our
definition is closely connected to a set of structures that appear in the theory of Tanaka
prolongation for filtered structures.1 We will not delve deeply into connections to that
theory, or to parabolic geometry more broadly, here, though these are certainly of great
interest. We will return to them in future work; for now, the interested reader is referred
to [Tan70; Zel09; CS09].
1
We owe deep thanks to John Huerta for calling our attention to the relevance of Tanaka’s work.
214 Eleven-dimensional supergravity as a Calabi–Yau twofold

In some sense, the usefulness of the definition lies in the fact that it brings the three
classes of examples under one roof. In particular, our main application—to eleven-
dimensional supergravity—will rely on exploiting the analogy between instances of the
first two types. To get to these examples, we need to construct the generalization of
Dolbeault cohomology to this more general setting. This will be done in the next section.
We then move on to discuss examples in §7.2.3.

7.2.2 D∞ algebras from weighted flag structures

Given a weighted flag structure, we can regrade Ω• with respect to the sum of the
weight grading and the cohomological grading. (The filtration associated to this totalized
grading recovers F+• .) Having done this, the differential d decomposes as a sum of terms

d = d1 + d0 + d−1 + · · · (7.10)

with respect to the totalized grading. (All terms have cohomological degree one.)

We observe that d1 itself defines a differential of square zero. (This is the differential
on Gr F • (Ω• ).) We will now choose to regard this differential as “internal,” and the
additional terms d0 + d−1 + · · · as defining a further structure on Gr F • (Ω• ).

Recall that a square-zero endomorphism can be thought of as the defining data of an al-
gebra structure over the operad D governing square-zero differentials. (See, for example,
[Val14].) This operad has a single operation d0 of arity one, subject to the relation that
its concatenation with itself vanishes. We view it as a dg operad in totalized degree zero.
(From the perspective of the D∞ structure, the cohomological degree is the totalized
degree.)

A D-algebra in cochain complexes is thus almost the same thing as a bicomplex, except
for the fact that the second grading has been forgotten. We could restore it by giving an
action of U (1) on the operad D with respect to which the nontrivial operation has weight
one, and asking for an equivariant D-algebra structure on a weighted cochain complex.

Due to the relation d20 = 0, the operad D is not free, and does not play well with quasi-
isomorphisms. As is standard in homotopical algebra, we must replace D by a freely
generated dg operad that resolves it. This operad D∞ is generated by one operation di
for each nonpositive i, all of which have arity one. The conditions defining a D∞ algebra
structure in cochain complexes amount to the condition that the total differential

d = d1 + d0 + d−1 + · · · (7.11)
Eleven-dimensional supergravity as a Calabi–Yau twofold 215

is of square zero, where d1 is the internal differential of the cochain complex and di for
i ≤ 0 encode the D∞ algebra structure. A weighted flag structure thus defines a D∞
algebra structure on Gr F • (Ω• ) with respect to the totalized degree.

Homotopy transfer. Since D∞ is a good homotopy replacement for D, one can use
homotopy transfer of D∞ algebra structures to pass between different quasi-isomorphic
models. This encodes, in particular, the higher differentials of the spectral sequence of a
bicomplex. We thus consider the cohomology

W • := H • (Gr F • (Ω• )) = H • (Ω• , d1 ) . (7.12)

Since d1 is homogeneous for the weight grading, W • is again bigraded, by weight and
cohomological degree—or equivalently, by cohomological degree and totalized degree.
We will find it more convenient to work with the totalized degree in the sequel.

We can apply the homotopy transfer theorem for D∞ algebras [LV12b] in order to obtain
a new D∞ algebra structure on W • . In concrete terms, this is done by fixing a retraction

p
h (Ω• , d1 ) (W • , 0) . (7.13)
i

The structure sheaf A• ; geometric interpretation. When applied to a weighted

flag structure, the output of the above construction is a cdga W • with zero internal
differential, equipped with a bigrading and a D∞ structure. We will denote the terms of
the D∞ structure by d0i for i ≤ 0; the term d0i has cohomological degree one and totalized
degree i. As such, d0 =
P 0
di is a square-zero differential of cohomological degree one,
which now does respect the filtration F+• W • associated to the totalized degree.

If we like, we can therefore repeat the procedure from above. Gr F+• W • will be a bigraded
cdga with a differential of totalized degree zero. If we were to shift the totalized grading
up by the cohomological degree again, we would get a D∞ structure on Gr F+• W • with
respect to that new grading. However, we will not have cause to do this. Instead, we
will regard Gr F+• W • = (W • , d00 ) as the fundamental object. We will allow ourselves to
refer to this object as the generalized Dolbeault complex.

We have seen above that, for an integrable complex structure, F+• is nothing other than
the Hodge filtration. As was worked out in [CW21], F+• W • is the correct object to
replace the standard Hodge filtration (and thus the standard Dolbeault cohomology) for
non-integrable complex structures. We will discuss this in detail in examples in the next
section.
216 Eleven-dimensional supergravity as a Calabi–Yau twofold

The complex geometry of a complex manifold is governed by its sheaf of holomorphic

functions; a good derived replacement for this sheaf is the sheaf Ω0,• of Dolbeault forms
that smoothly resolves it. There is an obvious generalization of this structure sheaf in
our setting as well. W • is negatively graded with respect to the totalized grading, so
that we can decompose it as a sum
M
W• = W i,• (7.14)
i≤0

of homogeneous subspaces. This splitting is compatible with the differential on Gr F+• W • .

As such, we can consider the cdga A• := (W 0,• , d00 ) sitting in totalized degree zero; this
should be viewed as the structure sheaf of the geometry we are considering. As we will
see in the next section, applying this construction to examples arising from superspaces
produces the canonical supermultiplet—and therefore, among other physically important
examples, the eleven-dimensional supergravity multiplet. Pursuing this analogy with
complex geometry further will allow us to produce the interactions of eleven-dimensional
supergravity from a holomorphic Poisson structure on this ringed space, reproducing and
generalizing work of Cederwall [Ced10c; Ced10a].

7.2.3 Examples of weighted flag structures

Complex manifolds. Let X be a complex manifold; locally, we can equip X with

corresponding coordinates (z i , z̄ i ). We consider the de Rham complex on X,

Ω• (X) , d = ∂ + ∂¯ .


(7.15)

The de Rham differential d splits into holomorphic and antiholomorphic pieces, the
¯ The cohomological grading is by form degree; to define the weight
operators ∂ and ∂.
grading, we assign dz̄ weight zero and dz weight −1. This coresponds to the filtration

0 ⊂ D1 = T (0,1) X ⊂ D2 = TC X (7.16)

of the complexified tangent bundle, which we have refined to give a weighted flag structure
by choosing
T1 X = T (0,1) X, T2 X = T (1,0) X. (7.17)

(Note that, for integrable complex structures, D1 is as far as possible from being bracket-
generating.) In this example, it is clear that the terms of the decomposition of the
differential are
d1 = 0, ¯
d0 = ∂, d−1 = ∂, (7.18)
Eleven-dimensional supergravity as a Calabi–Yau twofold 217

with all higher terms vanishing. As a result, W • can be identified with Ω• , and A• is
the Dolbeault complex Ω0,• (X).

Almost complex manifolds. Nothing in the construction of the weighted flag struc-
ture above depended on the integrability of the complex structure. In fact, the construc-
tion generalizes immediately to almost complex manifolds, with the difference that D1
is no longer involutive. Correspondingly, the internal differential d1 no longer vanishes.
We recover the theory of Dolbeault cohomology for almost complex manifolds, as worked
out in [CW21].

On an almost complex manifold, the de Rham differential decomposes as

d = µ + ∂¯ + ∂ + µ, (7.19)

where µ and its complex conjugate µ are related to the Nijenhuis tensor. No other terms
are present. Defining the weighted flag structure considered above, we see that

d1 = µ, ¯
d0 = ∂, d−1 = ∂, d−2 = µ. (7.20)

Crucially, the Dolbeault differential ∂¯ no longer squares to zero, such that standard
Dolbeault cohomology is no longer well defined. But we can nevertheless construct W •
by first passing to the cohomology of µ̄:

W • = H • (Ω• (X), µ̄). (7.21)

This reproduces the construction of the Dolbeault cohomology of an almost complex

manifold, as defined in [CW21]. Homotopy transfer as D∞ algebras then produces a D∞
structure on W • , which plays the role of the Hodge-to-de-Rham spectral sequence in this
case.

We note that the first term in the differential, d1 = µ, can be thought of as encoding
the failure of the corresponding flag of distributions to be integrable. (In the theory of
filtered structures, one would say that the symbol of the flag of distributions fails to be
abelian.) This is further illustrated by the next examples.

Superspaces and the canonical supermultiplet. Let n be a supertranslation alge-

bra in the sense of §2.2: a consistently Z-graded super Lie algebra supported in degrees
one and two. In our conventions here, which differ slightly, n1 has weight one and odd
internal parity, whereas n2 has weight two and even internal parity. Let N = exp(n) be
218 Eleven-dimensional supergravity as a Calabi–Yau twofold

the corresponding flat superspace. The de Rham complex

 
• ∞ ∂ ∂
(Ω (N ), ddR ) = C (T+ )[θ, dθ, dx] , dx + dθ (7.22)
∂x ∂θ

is then a cdga equipped with a weight grading.

We can define a flag of distributions in T N by choosing D1 to be spanned by the odd

left-invariant vector fields Vect(N )N − , and D2 to be just T N . In physical examples

in three or more dimensions, D1 is always bracket-generating, since every translation

is the square of some supercharge. Thus, the distribution we consider is maximally
noninvolutive.

This flag of distributions defines a weighted flag structure on Ω• (N ). Concretely, we can

express the de Rham complex in a left-invariant basis

λ = dθ, v = dx + λθ. (7.23)

Then, the de Rham differential takes the form

 
∂ 2 ∂ ∂ ∂
ddR =λ +λ −θ +v . (7.24)
∂v ∂θ ∂x ∂x

Note that we suppress the contractions in the notation when there is no ambiguity. The
weight grading on the de Rham complex is just given by the polynomial degree in v. The
differential splits according to

∂
d1 = λ2 ,
∂v
∂ ∂
d0 = λ − λθ , (7.25)
∂θ ∂x
∂
d−1 = v .
∂x

As we wil see later, the generalized Dolbeault complex,

W • = H • (Ω• (N ), d1 ) , d00 ,


(7.26)

has a natural interpretation within the pure spinor superfield formalism. In particular
the degree zero piece W 0,• coincides with the canonical multiplet of n [Ced+23]; the
analogue of the Dolbeault resolution of holomorphic p-forms is given by the multiplet
associated to the (−p)-th Lie algebra cohomology of the supertranslation algebra n, with
respect to the totalized degree. We already investigated examples of such multiplets
in §2.6.3; the acyclic deformation of the differential arising from the strictly negative
terms in d was defined, and worked out concretely in examples, in §3.5.1.
Eleven-dimensional supergravity as a Calabi–Yau twofold 219

These multiplets were discussed in detail in physical examples in [Eag+22]; the acyclic
deformation of the differential arising from the strictly negative terms in d was defined,
and worked out concretely in examples, in [EHS23].

Further examples; (flat) distributions of constant symbol. In the previous sec-

tions, we have already gone through the examples that will interest us in detail in the
remainder of this work. Our main aim here is to set up the analogy between almost com-
plex geometry and superspace by viewing them both as weighted flag structures, and
to exploit this to give a geometric construction of interacting eleven-dimensional super-
gravity and its twists. However, numerous other structures could be viewed through this
lens, and we feel it would be profitable to do so. We give a partial list of such examples,
to which we hope to return in future work.

— Any manifold equipped with a Tanaka structure [AD17, Definition 1] has a flag
structure on its de Rham complex.

— Let n be a super Lie algebra equipped with a positive weight grading. Follow-
ing [Zel09], we can consider the flat Tanaka structure with constant symbol n. By
definition, this is the simply connected super Lie group N = exp(n), equipped with
the flag of distributions spanned by the left-invariant vector fields in n≤j . We ob-
serve that flat superspace is a particular example of such a flat Tanaka structure,
with symbol the supertranslation algebra. It should be possible to consider non-
strict examples (super L∞ algebras with positive weight gradings), using results of
Getzler [Get09].

— Any Lie algebra equipped with a finite-length positive filtration gives rise to a flag
structure on its Chevalley–Eilenberg cochains.

— Any filtered Lie algebroid gives rise to a flag structure on its Lie algebroid cochains.
This is a clear generalization, both of the previous example and of a flag of distri-
butions in the tangent bundle of a manifold. It should be possible to extend this
definition to Courant algebroids, following [Roy99], and then to understand poten-
tial connections to exceptional generalized geometry. In particular, connections of
Tanaka prolongation to tensor hierarchy algebras [Pal14] should be interesting to
explore.
220 Eleven-dimensional supergravity as a Calabi–Yau twofold

7.3 Poisson–Chern–Simons theories via derived brackets

7.3.1 Holomorphic Poisson–Chern–Simons theory

In this section, we briefly review the construction of the standard Poisson–Chern–Simons

theory, defined on a product of a Calabi–Yau twofold and an odd-dimensional smooth
manifold. The theory is Z-graded only when the smooth manifold is one-dimensional.
Poisson–Chern–Simons theory was related to the maximal twist of eleven-dimensional
supergravity in a particular omega background by Costello in [Cos16].

Let X be a Calabi–Yau twofold with holomorphic volume form Ω. In complex dimen-

sion two, Ω is also a holomorphic symplectic structure. We denote the corresponding
holomorphic Poisson bivector by π = Ω−1 .

Recall from §7.2.3 above that the totalized grading places dz in degree −1 and dz in
degree zero. Our construction above recovers the standard Dolbeault complex (equipped
with a nonstandard grading):

W • = Ω• (X), ¯
d0 = ∂, d−1 = ∂. (7.27)

Contracting with π defines an isomorphism of Ω0,• (X)-modules

π : Ω2,• (X) , ∂¯ −→ Ω0,• (X) , ∂¯ ,

α 7→ π ∨ α. (7.28)

One can now use this data to equip the Dolbeault complex Ω0,• (X) with the structure
of a cyclic L∞ algebra. This can be done in two steps:

1. Turn Ω• (X) into a BV algebra.

2. Define the Poisson bracket on Ω0,• (X) as a derived bracket of the BV bracket.

For the first step, note that the commutator ∆ = [π, ∂] defines a second-order differential
operator acting on Ω• (X), satisfying ∆2 = 0 and ∆(1) = 0. Hence, we can define the
Koszul bracket on Ω• (X) by

{α, β} = (−1)|α| (∆(αβ) − ∆(α)β) − α∆(β) , (7.29)

making (Ω• (X), 1, ∆, {−, −}) into a BV algebra. This construction is due to Koszul [Kos85].
Eleven-dimensional supergravity as a Calabi–Yau twofold 221

For the second step, we employ the derived bracket construction with respect to the
differential ∂, as described by [KS96]. The derived bracket is defined by

[−, −]∂ := {∂(−), −}. (7.30)

Crucially, this bracket does not turn all of Ω• (X) into a Lie algebra; only after restricting
to an abelian subalgebra (with respect to the underived bracket {−, −}) does [−, −]∂ have
the right symmetry properties. It is easy to check that the Dolbeault complex Ω0,• (X)
is indeed such a subalgebra; from this, it follows that

Ω0,• (X) , ∂¯ , [−, −]∂

(7.31)

is a dg Lie algebra.

Evaluating [−, −]∂ on α, β ∈ Ω0,• (X), we find

[α, β]∂ = {∂α, β} = π(∂α ∧ ∂β) , (7.32)

recovering the well known formula for the Poisson bracket. Together with the pairing
induced by wedging with the holomorphic volume form Ω and integration, this makes
¯ [−, −]∂ ) into a cyclic L∞ algebra—indeed, a local L∞ algebra with a cyclic
(Ω0,• (X), ∂,
structure of degree −2. Tensoring with an odd-dimensional smooth manifold gives a
quasi-isomorphic cdga that is a local L∞ algebra with an odd-shifted cyclic structure on
the product manifold. The corresponding Z/2Z-graded BV theory is called holomorphic
Poisson–Chern–Simons theory.

7.3.2 Homotopy Poisson–Chern–Simons theory

We now generalize the above setting to the context of §7.2 in order to construct a
“homotopy” version of Poisson–Chern–Simons theory.

Let (Ω• , d) be a cdga equipped with a weighted flag structure, and let (W • , d0 ) be the
corresponding generalized Dolbeault complex. Let us assume that, with respect to the
totalized grading, W • is concentrated in degrees 0, −1, and −2. For degree reasons, the
differential then splits into three pieces

d0 = d00 + d0−1 + d0−2 . (7.33)

222 Eleven-dimensional supergravity as a Calabi–Yau twofold

Explicitly, these terms arise via homotopy transfer along the diagram (7.13).

d00 = i ◦ d0 ◦ p
d0−1 = i ◦ (d0 hd0 + d−1 ) ◦ p (7.34)
d0−2 = i ◦ (d0 h)2 d0 + d0 hd−1 + d−1 hd0 ◦ p

Note that the square zero condition for d0 implies the following identities:

(d00 )2 = 0
[d0−1 , d00 ] = 0
(d0−1 )2 + [d00 , d0−2 ] = 0 (7.35)
[d0−1 , d0−2 ] = 0
(d0−2 )2 = 0.

Here, the bracket [−, −] denotes the commutator of endomorphisms. As all terms are of
cohomological degree one, these are all symmetric. We further assume that there is an
isomorphism
π : (W −2,• , d00 ) −→ (W 0,• , d00 ) (7.36)

of W 0,• -modules.

In summary, the D∞ structure and the pairing π act on W •,• as indicated by the following
diagram.
π
d00 d00 d00

d0−1 d0−1
W 0,• W −1,• W −2,• (7.37)
d0−2

From this data, we now construct an L∞ structure on W 0,• . For this purpose we perform
the appropriate generalizations of the steps described in §7.3.1.

1. Turn W • into a BV∞ algebra.

2. Define an L∞ structure on A• = W 0,• using a derived bracket construction.

We will see that both steps can be viewed as instances of the derived bracket construction
described by [Vor05; BV16].

We begin by recalling the definition of a BV∞ algebra.

Eleven-dimensional supergravity as a Calabi–Yau twofold 223

Definition 7.3.1. A BV∞ algebra (A, ∆, 1) is a unital graded commutative algebra over
C together with a degree one linear map ∆ : A −→ A[[t]] which can be expanded as

∞
1X k
∆= t ∆k , (7.38)
t
k=1

such that ∆k is a differential operator of order at most k and

∆2 = 0 and ∆(1) = 0. (7.39)

One can equip both A[[t]] and A with L∞ structures in the following way. By identi-
fying an element a ∈ A by the endomorphism given by left multiplication with a, we
can embed A as an abelian subalgebra into its graded Lie algebra of endomorphisms,
(End(A), [−, −]). The other way round, evaluating an endomorphism at the unit gives a
right inverse to this embedding. One can define a series a series of brackets on A[[t]] by
the following formulas [Vor05].

{a1 , . . . , an }t = [. . . [∆, a1 ], . . . , an ](1) (7.40)

This makes A[[t]] into an L∞ algebra. Note that the unary bracket is just given by ∆,
while the binary bracket is then given by the well known formula for BV algebras

{a1 , a2 } = ∆(a1 a2 ) − ∆(a1 )a2 − (−1)|a1 | a1 ∆(a2 ). (7.41)

In general, the n-ary bracket can be thought of as measuring the failure of the (n−1)-ary
bracket to be a multiderivation with respect to the algebra structure.

Further, we can extract an L∞ algebra structure on A by taking an appropriate limit for

the parameter t. We define

1
{a1 , . . . , an } = lim {a1 , . . . , an }t . (7.42)
t→0 tn−1

The limit makes sense because ∆k is a differential operator of order at most k. Note
that, for this L∞ structure, the n-ary operation is generated by ∆n , i.e.

{a1 , . . . , an } = [. . . [∆n , a1 ], . . . , an ](1) . (7.43)

Coming back to our setting, we define the operator

∆ = ∆1 + t∆2 + t2 ∆3 = d00 + t[π, d0−1 ] + t2 [π, [π, d0−2 ]] (7.44)

224 Eleven-dimensional supergravity as a Calabi–Yau twofold

on W • [[t]]. A direct calculation shows the following proposition.

Proposition 7.3.2. (W • , ∆, 1) is a BV∞ algebra. Furthermore, W 0,• is an abelian

subalgebra, and W <0,• is a subalgebra with respect to the bracket {−, −}.

Proof. These statements can be shown by direct calculations. For example, we can
examine ∆2 = 0 order by order in t. Recall the identities (7.35) for the D∞ -algebra
structure on W • . At order t0 , ∆2 = 0 is just the square-zero condition for d00 , while the
t1 -term vanishes since d0−1 and d00 anti-commute. For the t2 -piece we find

[π, d0−1 ]2 + [d00 , πd0−2 π]. (7.45)

Recall that (d0−1 )2 = −[d00 , d0−2 ]. For degree reasons, the only term contributing to the
first summand is π(d0−1 )2 π, for which we find

π(d0−1 )2 π = −π[d00 , d0−2 ]π = −[d00 , πd0−2 π], (7.46)

using compatibility between the the pairing and d00 . All higher order pieces vanish
for degree reasons. The other claims are verified by similar calculations and degree
arguments.

Proposition 7.3.2 sets the stage for the second step. We now apply the derived bracket
construction to the differential

dt = d00 + td0−1 + t2 d0−2 . (7.47)

Again, this first endows W 0,• [[t]] with an L∞ structure

µtn (a1 , . . . , an ) = {. . . {dt , a1 }, . . . an } (7.48)

and then finally W 0,• by taking the limit

1
µn = lim µtn (7.49)
t→0 tn−1

The L∞ structure then takes the following form

µ1 (α) = d00 α
µ2 (α, β) = {d0−1 α, β} (7.50)
µ3 (α, β, γ) = {{d0−2 α, β}, γ}.

It is useful to express this L∞ structure in terms of the pairing π.

Eleven-dimensional supergravity as a Calabi–Yau twofold 225

Proposition 7.3.3. For α, β, γ ∈ W 0,• we have

µ2 (α, β) = π(d0−1 α · d0−1 β)

(7.51)
µ3 (α, β, γ) = π(d0−2 α · π(d0−1 β · d0−1 γ)).

Proof. For µ2 we have

{d0−1 α, β} = (−1)|α| πd0−1 (d0−1 α · β) − (π(d0−1 )2 α) · β ,

(7.52)

where we already used that [π, d0−1 ]β = 0 by degree reasons. Using that d0−1 is a deriva-
tion for the multiplication, we find the desired result.

For µ3 , note that

{d0−2 α, β} = (−1)|α| d0−1 π(d0−2 α · β) − (d0−1 πd0−2 α) · β

(7.53)
= (πd0−2 α) · d0−1 β ∈ W −1,• ,

where we used that π is an isomorphism of W 0,• -modules in the second step. Thus, we
find

{{d0−2 α, β}, γ} = {(πd0−2 α)d0−1 β, γ}

= (−1)|α|+|β| πd0−1 (πd0−2 α)(d0−1 β)γ − πd0−1 (πd0−2 α)(d0−1 β) · γ




= π (πd0−2 α) d0−1 β · d0−1 γ

(7.54)

Again, using that π is a map of W 0,• -modules, we find the desired result.

In the examples we are interested in and which we will discuss in the following sections,
W 0,• is local, i.e. arising as a sheaf of L∞ algebras on some manifold, and equipped with
a pairing making it a cyclic L∞ algebra. In these instances, (W 0,• , d00 , µ2 , µ3 ) defines a
perturbative interacting BV theory, perhaps after tensoring with the de Rham complex
of a smooth manifold to correct for the parity of the cyclic structure. Since the L∞
structure describing the interactions is no longer strict, we refer to such a theory as a
homotopy Poisson–Chern–Simons theory.

7.4 Calabi–Yau twofolds from certain Gorenstein rings

The construction of interactions in homotopy Poisson–Chern–Simons theory can be ap-

plied to supersymmetric field theories and their twists just by working with the example
of §7.2.3—that is, with the standard odd distribution on superspace. As we will show
226 Eleven-dimensional supergravity as a Calabi–Yau twofold

more explicitly below, this automatically places us in the context of the pure spinor su-
perfield formalism. It remains only to check which superspaces give rise to weighted flag
structures satisfying the conditions of §7.3. Of the standard superspaces that appear in
physics, there are precisely three examples, corresponding to eleven-dimensional minimal
supersymmetry and its two distinct twists.

We begin by discussing the compatibility between the pure spinor superfield formalism
and twisting; the observations here extend [SW21]. Then we remark on the algebraic
conditions required for the generalized Dolbeault complex (W • , d0 ) of a superspace to
have the properties of the Dolbeault complex of a Calabi–Yau twofold, and thus to give
rise to a homotopy Poisson–Chern–Simons theory using the techniques of §7.3. In §7.5
below, we will show that the resulting theories are eleven-dimensional supergravity and
its maximal and minimal twists.

7.4.1 Pure spinor superfields for twisted field theories

Let g be a super Lie algebra of super Poincaré type and n the corresponding super-
translation subalgebra. As witnessed in various places throughout this thesis, there is a
correspondence between supertranslation algebras and generating sets of quadratic ideals
in polynomial rings. Let R = Sym• (n∨
1 ) denote the ring of polynomial functions of n1
and I the quadratic ideal generated by the equations [Q, Q] = 0 for Q ∈ n1 . As usual,
the quotient ring R/I is the ring of functions of the nilpotence variety. Conversely, we
can produce a super Lie algebra of supertranslation type from any finite sequence of
quadratic equations. Let R = C[λ1 , . . . , λn ] be the polynomial ring in n variables and I
an ideal generated by the equations,

µ β
I = (λα fαβ λ ), µ = 1 . . . d, α, β = 1 . . . n. (7.55)

We define n to be the two-step nilpotent super Lie algebra

n = ΠS(−1) ⊕ V (−2), (7.56)

equipped with the indicated weight grading. Here S ∼

= Cn , V ∼
= Cd , and the only
non-trivial bracket is the map

[−, −] : Sym2 (S) −→ V, (7.57)

µ
generated by the equations (7.55)—in other words, with structure constants fαβ .
Eleven-dimensional supergravity as a Calabi–Yau twofold 227

Recall that applying the pure spinor superfield functor to C • (n) itself recovers Ω• (N ),
expressed in the left-invariant frame discussed in §7.2.3

A• (C • (n)) = (C ∞ (N ) ⊗ C • (n) , D) ∼
= (Ω• (N ), ddR ), (7.58)

where the differential splits according to (7.25); the internal differential d1 now coincides
with the Chevalley–Eilenberg differential dCE on C • (n). Taking cohomology with respect
to d1 , we thus recover that
W • = A• (H • (n)); (7.59)

the differential d0 is the standard pure spinor superfield differential, so that the gener-
alized Dolbeault complex in totalized degree k—the analogue of the holomorphic (−k)-
forms—consists of the supermultiplet associated by A•R/I to the Lie algebra cohomology
group H k (n), again in the totalized grading. In particular, the role of the would-be
structure sheaf is played by the canonical multiplet associated to the ring R/I itself.
This justifies our notation A• = A•R/I (R/I) = (W 0,• , d00 ) from above.

Pure spinor superfields and twisting. Fixing an element Q ∈ Y , we can twist the
algebra itself by defining a dg Lie algebra (g , [Q, −]). Its cohomology gQ = H • (g, [Q, −])
is again a graded Lie algebra in degrees zero to two and should be viewed as the residual
symmetry algebra of any theory twisted by Q; we denote its nilpotence variety (which
encodes the possible further twists of the Q-twisted theory) by YQ .

We call the positively graded piece of the cohomology

nQ = H >0 (g, [Q, −]) (7.60)

the twisted supertranslation algebra. Sometimes it is convenient to work with a quasi-

isomorphic dg model for nQ which keeps all the even translations in degree two. To this
end we define a dg Lie algebra ñQ by throwing away the degree zero piece of (g, [Q, −])
while simultaneously replacing its degree one piece by the cokernel of the adjoint action
of Q,
ñQ = (n1 /Im([Q, −])(−1) ⊕ n2 (−2) , [Q, −]) . (7.61)

Given any C • (n)-module Γ, the twist by Q of the multiplet associated to Γ,

A• (Γ)Q = (A• (Γ) , D + L (Q)) , (7.62)

is a multiplet for the dg Lie algebra (g, adQ ). On the other hand, we can also apply
the pure spinor superfield formalism directly to the residual supersymmetry algebra
(g, adQ ) (or equivalently its cohomology gQ ). As the formalism provides an equivalence
228 Eleven-dimensional supergravity as a Calabi–Yau twofold

of categories, there is a C • (nQ )-module (or equivalenty C • (ñQ )) ΓQ such that

A• (ΓQ ) ' A• (Γ)Q . (7.63)

Explicitly, we can take derived ñQ -invariants on both sides of this equation to find

ΓQ ' C • (ñQ , A• (Γ)Q ). (7.64)

Before the derived formalism was available, Saberi and Williams already recognized in
examples that the twist of the canonical multiplet is equivalent to the canonical multiplet
of the twisted supersymmetry algebra [SW21],

A• (OY )Q ' A• (OYQ ). (7.65)

With the above observations at hand, we can now show this in general.

Theorem 7.4.1. For the structure sheaf one has (OY )Q ' OYQ , i.e. the twist of the
canonical multiplet is equivalent the canonical multiplet of the twisted algebra.

Proof. The proof is a short cohomology calculation similar to those in §3. Recall that
the degree one piece (ñQ )1 = n1 /Im([Q, −]) Let us choose a splitting (as vector spaces)

n1 = (ñQ )1 ⊕ Im([Q, −]) (7.66)

and let us correspondingly organize the generators of C • (n) as (λ1 , . . . , λk , λk+1 , . . . , λn )

such that the first k correspond to the generators of ñQ .

We then have

(OY )Q = C • (ñQ , A• (OY )Q )

  (7.67)
∞ ∂
= C (N ) ⊗ C[λ̃, ṽ] ⊗ R/I , L (Q) + λ R(dα ) + dCE + λ̃ L (di ) + v
α i ˜µ .
∂xµ

Here, we denote the generators of C • (ñQ ) by λ̃i and ṽ µ . Note that, while the index i for
λ̃ only runs from 1 to k, the index µ runs over all spacetime coordinates. The generators
of the ring R are (as usual) denoted by λα . We now use the filtration by polynomial
degree in ṽ and take cohomology with respect to the degree one piece of the differential,
∂
given by ṽ ∂x . This yields
 
i α 2 α α ∂ i ∂
(OY )Q ' C[θ, λ̃ , λ ]/(λ ) , (λ +  ) α + λ̃ , (7.68)
∂θ ∂θi
Eleven-dimensional supergravity as a Calabi–Yau twofold 229

where we expanded the twisting supercharge Q into the basis Q = α dα . Finally, we see
that the cohomology is given by

(OY )Q ' C[λ̃i ]/((λ̃i + i )2 ) = OYQ . (7.69)

Whenever Q is a maximal twist, the twisted nilpotence variety YQ is just a point such
that the associated canonical multiplet is given by a tensor product of the de Rham and
Dolbeault complexes such that the above implies the following corollary.

Corollary 7.4.2. Let Q be a maximal twist with k surviving translations´, then

   
A• (OY )Q ' Ω• (Rd−2k ), d ⊗ Ω0,• (Ck ), ∂¯ . (7.70)

These considerations mean that the operation of twisting is, in a sense, fully internal to
the superspace: any construction which relies only on the “(almost) complex geometry”
of the weighted flag structure of a superspace, as encoded in its generalized Dolbeault
complex, should behave in the same way in any twist. (Recall, for example, that the
full Dolbeault complex can be reconstructed algebraically from Ω0,• by considering the
module of Kähler differentials. We can thus think of the acyclic D∞ structure we con-
struct on W • as related to the algebraic de Rham cohomology of the affine dg scheme
Spec A• .)

In light of the above considerations, we can bootstrap information about this maximal
twist: if we have a description of an interacting theory that uses only information about
the complex geometry of Cn (or, more precisely, the THF structure on Cn × Rd−2n ),
then the same construction (appropriately generalized to take into account the non-
involutiveness of the underlying distribution) should give a pure spinor model for the
untwisted interacting theory—or for any other twist—when applied to the corresponding
generalized Dolbeault complex.

More specifically, compatibility between the pure spinor superfield construction and
twisting at the interacting level can be formulated in the following way. Consider a
multiplet A• (Γ) and assume that it is further equipped with an L∞ structure making
it an interacting BV theory. As discussed above, there is a quasi-isomorphism in the
category of multiplets
A• (Γ)Q ' A• (ΓQ ). (7.71)

Explicitly, such a quasi-isomorphism can be obtained by a cohomology computation using

the techniques presented in [SW21].
230 Eleven-dimensional supergravity as a Calabi–Yau twofold

We can use this to formulate homotopy data

p
h (A• (Γ))Q A• (ΓQ ) (7.72)
i

and perform the homotopy transfer of the L∞ structure along this diagram to A• (ΓQ ).

Conjecture 7.4.3 (Saberi). When the L∞ structure on A• (Γ) is local on Spec A• (i.e.
there is a BV action in terms of the pure spinor superfield), the interactions are compat-
ible with twisting in the sense that the homotopy transfer (7.72) is formal.

7.4.2 The defect, the effective dimension, and the maximal twist

To apply the construction of §7.3.2 in the pure spinor superfield formalism, we thus
need to specify conditions that guarantee the existence and appropriate properties of the
pairing π. In particular, we would like the generalized Dolbeault complex W • to exhibit
the properties of the Dolbeault complex of a Calabi–Yau twofold.

Let us fix a supertranslation algebra n with corresponding polynomial ring R = Sym• (n∨
1)
together with dim(n2 ) generators for the quadratic ideal I and nilpotence variety Y . We
call the number

def(n) = dim(Y ) − (dim(n1 ) − dim(n2 )) = dim(n2 ) − codim(Y ) (7.73)

the defect of n.2 Roughly, it measures how far the generators of the ideal I are from
forming a regular sequence. The following proposition shows that the defect governs the
support of the Chevalley–Eilenberg cohomology of n, and thus the “complex dimension”
of Spec A• .

Proposition 7.4.4. Let R/I be a Cohen–Macaulay ring. Then, def(n) is the smallest
non-negative number such that H −i (n) 6= 0 for all i ≥ def(n).

Proof. Recall that the Chevalley–Eilenberg complex of n is the Koszul complex on our
set of generators for the ideal I. Let H −n (n) be the top cohomology group. By depth
sensitivity (see for example [Eis95, Theorem 17.4]) of the Koszul complex one has

depth(I, R) = dim(V ) − n. (7.74)

The Cohen–Macaulay condition implies depth(I, R) = codim(Y ) implies the claim.

2
If Y is not equidimensional, we take dim(Y ) to denote the maximum of the dimensions of its
irreducible pieces.
Eleven-dimensional supergravity as a Calabi–Yau twofold 231

We can further define a local version of the defect for any orbit in the nilpotence variety.
For Q ∈ Y we set
def(Q) = dim(V ) − codim(P0 · Q). (7.75)

The following lemma shows that the defect of Q is equal to the number of surviving
translations in a twist by Q.

Lemma 7.4.5. def(Q) = dim(H 2 (n, [Q, −])).

Proof. Recall that

H 2 (n, [Q, −]) ∼
= V /Im([Q, −]) . (7.76)

The map [Q, −] induces an isomorphism

n1 / ker([Q, −]) −→ Im([Q, −]) ⊆ V. (7.77)

Let P0 · Q denote the orbit of Q inside Y . Recall that P0 · Q sits inside n1 by the inclusion
i : (P0 · Q) ,→ n1 . The ambient space splits into directions tangent and normal to the
orbit:
n1 ∼
= TQ (P0 · Q) ⊕ NQ (P0 · Q). (7.78)

We can identify the tangent space with ker([Q, −]) and the normal space with the quotient
n1 / ker([Q, −]). Thus, we find in particular

codim(P0 · Q) = dim(NQ (P0 · Q)) = dim(n1 / ker([Q, −])). (7.79)

and therefore
def(Q) = dim(V ) − dim(n1 / ker([Q, −]))
(7.80)
= dim(V /Im([Q, −])) = dim(H 2 (n, [Q, −])),

proving the claim.

It follows from the proposition that the defect of n is the local defect evaluated at a
maximal twist lying in an orbit of maximal dimension. (Note that this is neither the
maximum, nor the minimum, value of the local defect; Y need not be—and often is
not—equidimensional.)

Gorenstein rings of defect two. Let us fix a supertranslation algebra n of defect

two such that the quotient ring R/I is both Gorenstein and strongly Cohen–Macaualay.3
3
A quotient ring R/I is called strongly Cohen–Macaulay, when all Koszul homology groups (for R/I
viewed as an R-module) are Cohen–Macaulay [Gol05].
232 Eleven-dimensional supergravity as a Calabi–Yau twofold

By construction, the zeroth Chevalley–Eilenberg cohomology of n yields,

H 0 (n) = R/I. (7.81)

Further, H • (n) is concentrated in degrees 0, −1 and −2. Since R/I is strongly Cohen–
Macaulay, H • (n) is a Poincaré duality algebra [AG71; Gol05]. In particular, we have

− codim(Y )
H −2 (n) ∼
= ExtR (R/I, R) ∼
= R/I, (7.82)

where we used the Gorenstein property for the last identification. Thus, there is an
isomorphism of A• (H 0 (n))-modules

π : A• (H −2 (n)) , d00 −→ A• (H 0 (n)) , d00 .

(7.83)

As we assumed that the defect of the supertranslation algebra equals two, transfer of the
D∞ along (7.13) yields an induced D∞ structure given by (7.34).

Hence, we are in the situation described in §7.3.2 and can construct an L∞ structure on
A• (H 0 (n)). Furthermore, the Gorenstein property implies that there is another pairing
on A• (H 0 (n)), (see §2.4), making it a cyclic L∞ algebra and hence an interacting BV
theory (after taking the product with an odd-dimensional smooth manifold to adjust the
parity of the cyclic structure, if necessary).

7.5 Eleven-dimensional supergravity, both twisted and not

As mentioned above, there are three significant examples of “Calabi–Yau twofolds” that
arise from superspaces relevant to physics. They are all connected to eleven-dimensional
supergravity: either the full theory, or one of its two twists. In this section, we review
the construction of these Gorenstein rings of defect two, and then construct the cor-
responding homotopy Poisson–Chern–Simons theories. These recover Cederwall’s pure
spinor formulation of eleven-dimensional supergravity, Costello’s description of the max-
imal twist in terms of holomorphic Poisson–Chern–Simons theory, and a (conjectural)
pure spinor lift of the interactions of minimally twisted eleven-dimensional supergravity
described in [RSW23]. We also recall how the rings are related to one another by twists
of the corresponding super Poincaré algebras, which, under the assumption of compati-
bility between the pure spinor construction and twisting, shows that the three interact-
ing theories are also obtained from one another—in particular, from eleven-dimensional
supergravity—by taking the corresponding twist
Eleven-dimensional supergravity as a Calabi–Yau twofold 233

7.5.1 Eleven-dimensional supersymmetry and its twists

Let V denote the vector representation for Spin(11) and S the unique spinor represen-
tation of dimension 32. The super Poincaré algebra in eleven dimensions is of the form

g = so(V ) ⊕ S(−1) ⊕ V (−2). (7.84)

The nilpotence variety Y ⊂ S is of dimension 23, so that def(Y ) = 2. Furthermore,

its coordinate ring, which is the quotient of polynomial functions on S by the quadratic
ideal generated by the eleven gamma matrices, is a Gorenstein ring. In this sense, the
generalized Dolbeault cohomology of eleven-dimensional superspace describes a Calabi–
Yau twofold. Furthermore, the structure sheaf of this space is nothing other than the
eleven-dimensional supergravity multiplet, described with a pure spinor superfield in
the BV formalism [How91b; CNT02]. As was emphasized in [SW21; Ced+23], eleven-
dimensional supergravity is a canonical supermultiplet, and is thus equipped with a
commutative structure on the space of fields.

Twists. As we have discussed in §6.2, the nilpotence variety decomposes into two orbits
for Spin(V ), as such, there are two distinct twists available. Recall that a maximal twist
is a smooth point of Y , whereas the minimal twist corresponds to a singular point. As is
well-known [BN05], the singularities take the form of the cone over the projective variety
Gr(2, 5). The stabilizer of a minimal supercharge is SU(5), whereas the stabilizer of a
maximal supercharge is G2 × SU(2).

Applying the pure spinor functor to the coordinate rings of the twisted nilpotence vari-
eties YQ , one obtains the BV complexes of the free twisted theories. We can now apply
our results to construct interactions for these theories in all these cases in a uniform way,
realizing them as homotopy Poisson–Chern–Simons theories.

We will begin by describing the maximal twist, and work up to the full theory.

7.5.2 The maximal twist

In [Cos16] a description of the maximal twist in terms of Poisson–Chern–Simons theory

was proposed. We computed the twist in the free limit using component fields in §6; it
was also realized as a further twist of the minimal twist in [RSW23]. Now, we finally
address the interacting theory with pure spinor methods.
234 Eleven-dimensional supergravity as a Calabi–Yau twofold

Twisting the supersymmetry algebra. The maximal twist on flat spacetime is de-
fined on R7 ×C2 . As in §6, we begin by decomposing all relevant Spin(11)-representations
to G2 × SU(2) × U(1). Recall that, under this subgroup, the vector representation of
Spin(11) decomposes as
V = V7 ⊕ L ⊕ L∨ , (7.85)

where V7 is the seven-dimensional irreducible representation of G2 and L ∼

= 21 (as well
as L∨ ∼
= 2−1 ) as SU(2) × U(1)-representations. The spin representation gives

S = (1G2 ⊕ V7 ) ⊗ (20 ⊕ 11 ⊕ 1−1 ). (7.86)

We immediately see that S contains two copies of the trivial representation of G2 ×

SU(2), coming with U(1) weights ±1. These correspond to the maximal square-zero
supercharges. For definiteness, we choose

Q ∈ 1G2 ⊗ 1−1 . (7.87)

Remembering that so(V ) ∼

= ∧2 V and that, as G2 -representations,

∧2 V7 ∼
= V7 ⊕ g2 , (7.88)

we can decompose the dg Lie algebra (g, [Q, −]) as shown in Table 7.1. Here, the arrows

V7 V7 ⊗ 1−1

g2 V7 ⊗ 11 V7

V7 ⊗ 21 V7 ⊗ 20 21

V7 ⊗ 2−1 20 2−1
(7.89)
10 1−1

12 11

1−2

30

Table 7.1: Decomposition under the stabilizer

represent the map [Q, −]. By Schur’s lemma, all non-vanishing arrows are multiples of the
identity; thus it is immediate to compute the cohomology. Identifying the holomorphic
translations as 21 = L, we find a purely even Lie algebra of the form

gQ = H • (g, [Q, −]) = g2 ⊕ sl(L) ⊕ V7 ⊗ 2−1 ⊕ 1−2 ⊕ L(−2).

(7.90)
Eleven-dimensional supergravity as a Calabi–Yau twofold 235

We note that the positively graded piece nQ is just the abelian even algebra L. The dg
model ñQ is of the form
V7 ⊗ 11 V7

20 2−1 . (7.91)

21
Correspondingly, OYQ = C, and the nilpotence variety is just a point. Note that both the
dimension as well as the codimension are zero. As there are two surviving translations,
the defect is thus def(nQ ) = 2.

We can now apply the formalism of §7.2.3 to the twisted supertranslation algebra nQ . The
weighted flag structure takes D1 to be the zero section and D2 to be the full (holomorphic)
tangent bundle. Doing so, we recover the negatively graded algebraic de Rham complex
of C2 :
Ω• = C[z1 , z2 ][dz1 , dz2 ], (7.92)

with dzi in totalized degree −1. The differential d1 is trivial, and W • = Ω• ; the “structure
sheaf,” which is the canonical multiplet of nQ , just consists of holomorphic functions
on C2 .

In order to give a representation as a multiplet living on V = R7 × C2 , we can resolve in

smooth functions over V ; this recovers the Dolbeault complex of (0, •) forms on C2 . We
note that this can be obtained directly by considering the canonical multiplet of the dg
model ñQ :
A• (OYQ ) ' Ω0,• (C2 ) ⊗ Ω• (R7 ) , ∂¯C2 + dR7 .


(7.93)

In either case, this corresponds to the field content of the maximal twist of eleven-
dimensional supergravity.

We thus find ourselves in the setting of Z/2Z-graded holomorphic Poisson–Chern–Simons

theory. Constructing the L∞ structure recovers the interactions of Poisson–Chern–
Simons described in §7.3.1.

We note that the vanishing of the Chevalley–Eilenberg differential on the twisted super-
translation algebra (which directly follows from maximality of the twist) ensures that
we end up with Poisson–Chern–Simons theory instead of its homotopy version. This is
a general feature of maximal twists. Nonetheless, applying our construction to a non-
integrable complex structure would have given rise to a non-strict Poisson–Chern–Simons
theory with nonvanishing 3-ary bracket.
236 Eleven-dimensional supergravity as a Calabi–Yau twofold

7.5.3 The minimal twist

The minimal twist was computed in the free limit at the pure spinor cochain level
in [SW21]. Interactions for the component fields were proposed (and numerous con-
sistency checks perfomed) in [RSW23].

Twisting the supersymmetry algebra. The stabilizer of a minimal square-zero su-

percharge Q ∈ Y is isomorphic to SU(5). Choosing such a Q is equivalent to the choice
of a maximal isotropic subspace L ⊂ V . The vector representation then decomposes as

V = L ⊕ L∨ ⊕ C. (7.94)

The twisted super Poincaré algebra (g, [Q, −]) and its cohomology gQ were analyzed
in [SW21]. The positively graded piece of the cohomology is found to be

nQ ∼
= Π ∧2 L(−1) ⊕ ∧4 L(−2), (7.95)

where the bracket of two odd elements is given by the wedge product. (The parentheses
refer to shifts in the weight grading.) The nilpotence variety YQ is isomorphic to the
affine cone over the the Grassmannian Gr(2, 5) of two-planes inside a five-dimensional
vector space. One can equivalently think of this as the space of bilinear skew forms of
rank two on L∨ . As an affine variety, we have dim(YQ ) = 7, and therefore

def(nQ ) = 7 − (10 − 5) = 2. (7.96)

OYQ is also Gorenstein, so that we can apply our procedure to construct interactions
for A• (OYQ ). By [SW21], the pure spinor multiplet A• (OYQ ) is equivalent to the min-
imal twist of the supergravity multiplet. Our procedure thus constructs interactions
for minimally twisted supergravity on the pure spinor cochain level, corresponding to a
suggestion in [Ced21]. We expect that the interacting theory with this field content con-
structed in [RSW23] can be obtained from this cochain-level description via homotopy
transfer, thus rigorously proving that the twisted eleven-dimensional supergravity the-
ory of [RSW23]—which is intimately related to the exceptional simple linearly compact
super Lie algebra E(5|10)—is in fact the twist of eleven-dimensional supergravity.

From above, we know that W • can be constructed by considering the pure spinor mul-
tiplets associated to the Lie algebra cohomology groups of nQ . The cochains are given
by
C • (nQ ) ∼
= ∧• L∨ ⊗ R, (7.97)
Eleven-dimensional supergravity as a Calabi–Yau twofold 237

where we identified
Sym• (n∨ ab
1 ) = R = C[λ ]. (7.98)

We think of λab as a basis on (nQ )∨ 2 ∨

1 = (∧ L) for a, b = 1, . . . , 5, and make use of the
isomorphism ∧4 L ∼
= L∨ . Further, we can think of L∨ as constant holomorphic one-forms
on L = C5 with basis {dz a }. The Chevalley–Eilenberg differential is of the form

∂
dCE = λab λcd εabcde . (7.99)
∂(dze )

As expected for a Gorenstein ring of defect two, the cohomology is concentrated in

degrees 0, −1 and −2: 


R/I k ∈ {0, −2}

H (nQ ) ∼
k
= M k = −1 (7.100)



0 else,


where M is the cokernel of the map

φ : R ⊗ ∧2 L −→ R ⊗ L ea ∧ eb 7→ εabcde λcd ee . (7.101)

Here {ea } is a basis of L. As R/I-modules, H 0 (nQ ) is freely generated by the unit 1,

while H −2 (nQ ) is freely generated by λab dz a dz b .

After tensoring with de Rham forms on R in order to resolve freely over C5 × R, we can
describe Ω• with the quasi-isomorphic complex
 
Ω•dR (C5 ) ⊗ C[λab , θab ] , ∂C5 + ∂¯C5 + R + dCE ⊗ (Ω• (R) , dR ) , (7.102)

where  
∂ ∂
R=λ −θ (7.103)
∂θ ∂x
is the standard pure spinor differential. Here, the spatial coordinate x is one of (z, z).
Note that, with respect to the description in §7.2.3, θ is an odd function on the superspace
N , whereas the one-forms are λ, dz, and dz. The weighted flag structure places dz in
totalized degree −1 and everything else in degree zero. We can identify

d1 = dCE , d0 = R + ∂¯ + dR , d−1 = ∂. (7.104)

We construct the generalized Dolbeault complex W • according to the standard proce-

dure, using the formulas for the transferred D∞ structure above (7.34). The weighted
pieces of the generalized Dolbeault complex are the pure spinor multiplets associated to
the modules of (7.100). In contrast to the maximal twist, a piece of degree −2 arises,
238 Eleven-dimensional supergravity as a Calabi–Yau twofold

such that there is a non-vanishing map

d0−2 : A• (H 0 (nQ )) −→ A• (H −2 (nQ )), (7.105)

signaling that the induced L∞ structure will not be strict.

The Gorenstein property guarantees that there is an isomorphism

π : W −2,• , d00 −→ W 0,• , d00 .

(7.106)

Explicitly, π is induced from the isomorphism between H −2 (nQ ) and H 0 (nQ ); thus, in
terms of representatives, we have

π(λab dz a dz b ) = 1. (7.107)

Hence, we obtain an L∞ algebra structure on A• (H 0 (nQ )) by the formulas in Proposi-

tion 7.3.3.

7.5.4 Eleven-dimensional supergravity

Recall that the canonical multiplet associated to the eleven-dimensional supertranslation

algebra is the supergravity multiplet. In [Ced10c] and [Ced10a], Cederwall constructed
a consistent quartic BV action functional, recovering interacting eleven-dimensional su-
pergravity in the pure spinor superfield formalism. We now recover these interactions as
an instance of homotopy Poisson–Chern–Simons theory.

Lie algebra cohomology and W • . The Chevalley–Eilenberg cochains of the un-

twisted supertranslation algebra take the form

C • (n) = ∧• V ∨ ⊗ R , dCE ,


(7.108)

where R = Sym• (S ∨ ) = C[λα ] is the polynomial ring in {λα } with α = 1, . . . , 32. Fixing
a basis {vµ } of V ∨ , the Chevalley–Eilenberg differential takes the form

∂
dCE = λα Γµαβ λβ . (7.109)
∂v µ

Again, Chevalley–Eilenberg cohomology is concentrated in degrees 0, −1 and −2, with

H 0 (n) and H −2 (n) both being isomorphic to the ring of functions on the nilpotence
variety OY = R/I. The cohomology in degree −2 is spanned by the class

(λα Γµν β
αβ λ )vµ vν . (7.110)
Eleven-dimensional supergravity as a Calabi–Yau twofold 239

Eleven-dimensional interactions. Applying the pure spinor superfield construction,

we construct the generalized Dolbeault complex as the sum of the multiplets associated
to the modules from the previous section. (We note that the multiplet W −1,• physically
corresponds to a field-strength multiplet for W 0,• ; this fact was already appreciated
in [CNT02].)

As always, the weighted flag structure on the de Rham complex of superspace induces a
D∞ -module structure on W • , where d00 is the standard pure spinor differential and d0−1
and d0−2 are both nontrivial. Restricting these differentials to W 0,• recovers Cederwall’s
differential operators constructed in [Ced10c; Ced10a], where

d0−1 : W 0,• −→ W −1,• , (7.111)

corresponds to “R” and

d0−2 : W 0,• −→ W −2,• (7.112)

corresponds to “T ”. Together with π induced from

π(λα Γµν β
αβ λ vµ vν ) = 1, (7.113)

this yields an L∞ -structure on W 0,• .

Chapter 8

Differential operators and twisted

(2,0) supersymmetry

8.1 Introduction

In the previous chapter, we established an analogy between superspace geometry as

employed by the pure spinor superfield formalism and almost complex geometry. In this
perspective, the canonical multiplet A• (R/I) plays the role of the structure sheaf on
superspace. It is natural to ask for the analogs of standard geometric constructions in
this language. For example, we can consider derivations of the cdgsa A• (R/I); these
then model “holomorphic vector” fields on superspace. More generally, it makes sense to
consider differential operators on the canonical multiplet.

As explored above, one advantage of this perspective is that it allows for uniform descrip-
tions of the full theory and all of its twist. This line of thought allowed us to construct
the eleven-dimensional supergravity theory and its twists in a unified way as homotopy
Poisson–Chern–Simons theories. M-theory allows for M5 branes and M2 branes. The
effective theory on the worldvolume of a stack of N indistinguishable M5 branes is a six-
dimensional superconformal field theory with (2, 0) supersymmetry. This theory, often
simply called the (2, 0) theory (of type AN −1 ), is of somewhat elusive nature; famously
it contains a two-form gauge field with self-dual curvature and it is expected to not ad-
mit a formulation in terms of a variational principle. Holographic duality states that the
eleven-dimensional supergravity theory in backreacted geometry (in this case AdS7 × S 4 )
is dual to such a worldvolume theory in the limit where N is large. This duality is best
understood in the maximal twist [Cos16]. Further, the maximal twist of the AN −1 theory
has been investigated and linked to WN algebras [Yag12; BRR15]. Given compatibility
between twisting and the pure spinor superfield formalism, it makes sense to ask for a

241
242 Differential operators and twisted (2,0) supersymmetry

lift of these findings to the minimal and eventually untwisted cases. In this chapter, we
take first steps towards such a program.

In recent work [RSW23], the component fields of minimally twisted eleven-dimensional

supergravity were linked to the infinite-dimensional super Lie algebra E(5|10). In [RW22],
a decomposition of E(5|10) into E(3|6)-modules was investigated and used to compute
superconformal indices in the six-dimensional (2,0) theory. We construct a pure spinor
lift of this decomposition, relate it to differential operators on the canonical multiplet,
and further to line bundles over the twisted nilpotence variety. In addition, we lift the
comparison of these pieces to minimally twisted supergravity to pure spinor cochain level.
Finally, we offer some speculations on the untwisted case.

8.2 Differential operators on the canonical multiplet

As usual, let g = g0 nn be super Lie algebra of super Poincaré type with supertranslation
algebra n. We start by studying the derivations of the cdga A• (R/I) and view the result-
ing multiplet Der(A• (R/I)) as an analog of holomorphic vector fields on Spec A• (R/I).
Moving on, we consider differential operators of any order k to obtain a family of multi-
plets Diff k (A• (R/I)) which for k = 0 recovers the canonical multiplet itself and for k = 1
coincides with the derivations. We now describe the construction of these multiplets.

Recall that, for a cdgs A a linear map D : A −→ A is called a differential operator of

order k if for all a0 , . . . , ak ∈ A we have [a0 , [a1 , . . . [ak , D] . . . ]] = 0. Differential operators
are filtered by order

A• (R/I) ⊆ Diff ≤1 (A• (R/I)) ⊆ Diff ≤2 (A• (R/I)) ⊆ · · · ⊆ Diff(A• (R/I)), (8.1)

such that differential operators of degree precisely k can be defined as the quotient

Diff k (A• (R/I)) = Diff ≤k /Diff ≤k−1 (A• (R/I)). (8.2)

Note that the derivations Der(A• (R/I)) differential operators of arbitrary degree Diff(A• (R/I))
are naturally equipped with a dg Lie structure induced by the commutator.

It turns out that the multiplet Der(A• (R/I)) is arises via the pure spinor functor from
a simple sheaf on the nilpotence variety Y . Thinking of the points of Y as possible
twists of a theory with n-supersymmetry, the stalk of this sheaf at Q ∈ Y consists of
those bosonic spacetime translations that survive in the Q-twist. We sum this up in the
following theorem.
Differential operators and twisted (2,0) supersymmetry 243

Theorem 8.2.1. Let Γ be the R/I-module, defined as the cokernel of the map

(λγ)µβ : n1 ⊗ R/I −→ n2 ⊗ R/I. (8.3)

Then there is an equivalence of multiplets

Der(A• (R/I)) ' A• (Γ). (8.4)

Note that the support of the sheaf coker(λγ), if viewed as an R-module, does not nec-
essarily lie within the support of the sheaf R/I. This means that the use of coefficients
R/I in (8.3) is essential.

Proof. We can expand a general derivation in coordinates as

µ ∂ α ∂ α ∂
δ = X(x) + X(θ) + X(λ) (8.5)
∂xµ ∂θα ∂λα
i
with coefficient functions X(A) ∈ C ∞ (N ) ⊗ R/I for A ∈ {x, θ, λ} and X(λ) subject to the
additional constraint
α µ β
X(λ) γαβ λ = 0 ∀µ. (8.6)

The differential acts as follows

 
i ∂ α β µ ∂ α ∂ µ β ∂
[D, δ] = D(X(A) ) i
± X(θ) λ γαβ µ ± X(λ) α
− γαβ θ . (8.7)
∂A ∂x ∂θ ∂xµ

Note that the first term in the differential is internal in the sense that it only acts on
the coefficient functions of the derivation. We first take cohomology with respect to the
second and the third term in the differential. Denoting with K ⊆ R/I ⊗ n1 the subset
satisfying the constraint (8.6), these terms act via

(λγ µ )β
C ∞ (N ) ⊗ R/I ⊗ n2 C ∞ (N ) ⊗ R/I ⊗ n1 C ∞ (N ) ⊗ K (8.8)

(γ µ θ)β

such that its cohomology indeed coincides with A• (Γ) as graded vector spaces. Further,
the induced differential from the first “internal” piece in (8.7) precisely coincides with
the pure spinor differential on A• (Γ) making it an equivalence of multiplets.

From the module Γ = coker(λγ), we can construct the family of modules S k Γ such that

A• (S k Γ) ' Diff k (A• (R/I)) (8.9)

244 Differential operators and twisted (2,0) supersymmetry

by taking symmetric powers. Explicitly, with (8.3) defining Γ as a cokernel, the cokernel
of the map
(S ⊗ S k−1 V ) ⊗ R/I −→ S k V ⊗ R/I, (8.10)

is S k Γ.

Derivations and conformal supergravity. The multiplet Der(A• (R/I)) has a mean-
ingful interpretation in terms of conformal supergravity. Conformal supergravity theories
have been constructed in various dimensions and with various amounts of supersymmetry
(see for example [BSVP86; BRW83]); typically by the following recipe. One starts with
a supersymmetric gauge theory and computes the conserved currents associated to the
supersymmetry transformations. This gives three currents: The energy-momentum ten-
sor associated to the even translations, the supercurrent associated to supertranslation,
and a current associated to the R-symmetry. These current operators can be represented
by quadratic polynomials in the fundamental field operators of the underlying gauge
multiplet. Applying the supersymmetry transformation rules to these expressions gener-
ates a subrepresentation inside all local operators of the gauge theory, the supercurrent
multiplet. The conformal supergravity theory is defined as the theory which couples to
this supercurrent multiplet. It contains spin-2 degrees of freedom coupling to the energy
momentum tensor, a spin-3/2 field coupling to the supercurrent and so on. Since the
currents obey conservation equations, the fields in the conformal supergravity theory are
gauged by the supertranslation algebra and R-symmetry.

Whenever the gauge theory is realized as a canonical multiplet, we find that Der(A• (R/I))
precisely recovers the field content of these conformal supergravity theories. We list these
results in §8.6.

8.3 Preliminaries

8.3.1 (Twisted) six-dimensional (2, 0) supersymmetry

The super Poincaré algebra and its nilpotence variety. In six-dimensions there
is an exceptional isomorphism identifying Spin(6) ∼
= SU (4); under this identification,
the two spinor representations S+ and S− correspond to the fundamental and antifunda-
mental representations. There are Spin(6)-equivariant isomorphisms ∧2 S± ∼
= V , where
V denotes the six-dimensional vector representation.
Differential operators and twisted (2,0) supersymmetry 245

The six-dimensional (2, 0) supertranslation algbebra is

n = (S+ ⊗ U2 )(−1) ⊕ V (−2) , (8.11)

where U2 = (C4 , ω) is a four dimensional symplectic vector space and the bracket is
provided by the isomorphism ∧2 S+ ∼= V and the symplectic form. The R-symmetry
group is hence Sp(2) such that the super Poincaré algebra is of the form

g = (so(6) ⊕ sp(2)) n n. (8.12)

The nilpotence variety decomposes into two orbits corresponding to the two distinct
twists [ESW21; ES19b; SW23b]. The orbits are distinguished by the rank of the elements
under the tensor product decomposition in n1 . The maximal twists correspond to rank
two elements which are of the form

Q = ξ1 ⊗ r1 + ξ2 ⊗ r2 with ξ1 , ξ2 ∈ S+ r1 , r2 ∈ U2 . (8.13)

They are topological in four directions and holomorphic in the remaining two. Corre-
spondingly, the defect of the supertranslation algebra is def(n) = 1. The minimal twists
correspond to rank one elements
Q = ξ1 ⊗ r1 . (8.14)

As a supercharge of rank one is automatically square-zero, the orbit of minimal super-

charges corresponds to the space of rank one matrices four-by-four matrices. As such the
minimal orbit is a Segre variety whose projectivization is P3 × P3 . Further, the minimal
supercharges are holomorphic; their stabilizer is SL(3).

The canonical multiplet and its twists. The canonical multiplet of the (2, 0) super
Poincaré algebra is the abelian (2,0) tensor multiplet [CNT02; ESW21] whose component
fields consist of a self-dual two-form gauge field, a scalar field with values in the five-
dimensional vector representation of the R-symmetry group Sp(2) ∼ = Spin(5), and chiral
fermions taking values in the fundamental representation U2 of the R-symmetry group.
The theory is a presymplectic BV theory in the sense of [SW23b].

Let Q ∈ Y be a maximal square-zero supercharge. Again, the twisted nilpotence variety

YQ is a point and no further twists are possible. The maximal twist was computed using
component fields techniques in [SW23b] and is described by the following dg Lie algebra

Ω0,• (C) ⊗ Ω• (R4 ) , ∂¯C + dR4 .

(8.15)
246 Differential operators and twisted (2,0) supersymmetry

Let Q ∈ Y now be a minimal square-zero supercharge. The choice of such a supercharge

is equivalent to the choice of a maximal isotropic subspace L ⊂ V and a line ρ ⊂ U2 . The
twisted super Poincaré algebra was described in [SW21]; for the twisted supertranslation
algebra we have
nQ = (L ⊗ U ◦ )(−1) ⊕ ∧2 L(−2) , (8.16)

where U ◦ denotes the symplectic reduction U2 //ρ. The bracket is given by the wedge
product and the symplectic form on U ◦ .

Odd square-zero elements in nQ are of rank one with respect to the tensor product
decomposition of (nQ )1 ; thus, the twisted nilpotence variety YQ can be identified as the
space of two-by-three matrices with rank less or equal the one (the ideal I is spanned by
the two-by-two minors of a two-by-three matrix). As such, we can identify the projective
version of the nilpotence variety as

YQ ∼
= P1 × P2 (8.17)

via the Segre embedding. The orbit structure on YQ corresponds to the natural sl(2) ×
sl(3)-action on P1 × P2 . Clearly, the twisted nilpotence variety only has a single orbit
which corresponds to the maximal twist realized as a further twist of the minimal one.

The canonical multiplet of gQ , and thus the minimal twist of the abelian (2,0) theory,
was described in [SW21]. The twist was previously computed using component field
methods in [SW23b]. Explicitly its component fields take the form,
 
0,• 3
 Ω (C )
µA• (OYQ ) = 

∂ . (8.18)
 
Ω1,• (C3 ) U◦

8.3.2 The maximally twisted example

Let us briefly review the maximally twisted case from our perspective. These results
have been worked out in great detail in the literature [Cos16; BRR15; Yag12].

Let Q be a maximal square-zero supercharge in the (2,0) supertranslation algebra. The

canonical multiplet consists just of the Dolbeault-de Rham complex on C × R4 ; its
minimal model is just holomorphic functions on C such that the superspace underlying
our discussion is just C. Therefore, the constructions of derivations and differential
operators are immediate: the derivations of the canonical multiplet can be modeled by
holomorphic vector fields Vect(C) and the differential operators as Diff(C) accordingly.
Differential operators and twisted (2,0) supersymmetry 247

WN -algebras and the boundary theory In [Yag12] it was argued that the algebra
of operators on N M5 branes in the omega background is the WN algebra. Further,
in [BRR15] a subsector of the six-dimensional (2,0) theory associated to a Lie algebra g
was identified which can be described by the W-algebra Wg ; this subsector is constructed
by a twist with a square-zero supercharge of the superconformal algebra of “mixed type”
(i.e. of the form Q + S, where Q is an ordinary supercharge in the super Poincaré
algebra and S is a conformal supercharge). By the arguments described in [OY19], this
construction corresponds to the omega deformation of the theory and thereby contains
the information on the maximal twist. In the large N limit, the WN algebra becomes
the W1+∞ algebra.

Maximally twisted holography Recall that the maximal twist of the bulk theory is
described by Poisson–Chern–Simons theory on C2 × R7 ,

Ω0,• (C2 ) ⊗ Ω• (R7 ) , ∂¯C2 + dR7 , {−, −}P B .

(8.19)

The minimal model is described by holomorphic functions on C2 equipped with the

Poisson bracket. Introducing coordinates (z, w) on C2 the Poisson bracket acts as

{z, w}P B = 1. (8.20)

This theory, placed in the omega background, and its relation to a stack of M5 branes
was discussed in [Cos16] (see also [Cos17] for the corresponding story with M2 branes).
For our purposes, we can introduce M5 branes along

Cz × R4 × {0} ⊂ Cz × Cw × R7 , (8.21)

such that w is the coordinate for the holomorphic direction transverse to the brane.

On the other hand, considering differential operators on C (now thought of as the part
of the brane’s worldvolume with holomorphic dependence), the relevant commutation
relation is
[z, ∂z ] = 1. (8.22)

Already at this level, we can see how identifying the scaling in the transverse direction
with the order of the differential operator makes the comparison between (Diff(C), [−, −])
and (O(C2 ), {−, −}P B ) apparent. This can be made much more precise, crucially one
has to include backreaction and work out the Kaluza-Klein compactification (see [Cos16]
where this is worked out in the omega background).
248 Differential operators and twisted (2,0) supersymmetry

8.4 The minimal twist

Let us now fix a minimal square-zero supercharge Q in the six-dimensional (2,0) super-
translation algebra; as usual nQ denotes the twisted algebra, and A• (OYQ ) is the twist
of the canonical multiplet.

8.4.1 Differential operators, E(3|6), and line bundles

Derivations and O(0, 1). Theorem 8.2.1 shows that Der(A• (OYQ )) ' A• (Γ(0,1) ),
where Γ(0,1) (the notation will become clear in a moment) is the cokernel of the map
induced by the bracket,
(L ⊗ U ◦ ) ⊗ R/I −→ L∨ ⊗ R/I. (8.23)

For the component fields of the multiplet, one recovers the content of E(3|6):
h i
µ Der(A• (OYQ ))# = Ω2,• (C3 ) U ◦ ⊗ Ω1,• (C3 ) S 2 U ◦ ⊗ Ω0,• (C3 ) . (8.24)

We note that Der(A• (OYQ )) is naturally equipped with a dgs Lie structure, which—by
homotopy transfer—gives rise to an L∞ structure on the component fields. We expect
this L∞ structure to coincide with the Lie bracket on E(3|6).

Further, the projective nilpotence variety is equipped with two natural projections,

P1 × P2
π1 π2 . (8.25)

P1 P2

such that all line bundles arise via pullbacks

O(n, m) := π1∗ OP1 (n) ⊗OP1 ×P2 π2∗ OP2 (m). (8.26)

It is easy to see that the module Γ(0,1) can be viewed as the graded global section module
of the line bundle O(0, 1),
M
Γ(0,1) = Γ∗ (O(0, 1)) = H 0 (O(k, k + 1)). (8.27)
k∈Z

Differential operators and O(0, m). We can take symmetric powers of the module
Γ(0,1) to obtain the modules which give rise to differential operators of higher degrees via
the pure spinor functor. Geometrically, these then are graded global section modules of
Differential operators and twisted (2,0) supersymmetry 249

the line bundles O(0, m) ∼

= O(0, 1)⊗m ,

Γ(0,m) = S m Γ(0,1) = Γ∗ (O(0, m)). (8.28)

The corresponding multiplets

A• (0, m) := A• (Γ(0,m) ) ∼
= Diff m (A• (OYQ )). (8.29)

describe differential operators of degree m on the canonical multiplet. We find the

following field content for their component fields.

µA• (0, m)# = (8.30)

h i
[0|m, 0] [1|m − 1, 1] [0|m − 2, 2] ⊕ [2|m − 1, 0] [1|m − 2, 1] [0|m − 2, 0] . (8.31)

More multiplets from line bundles. Taking a step back, we can use the pure spinor
superfield formalism to construct the family of multiplets for the twisted (2, 0) algebra
corresponding to the family of line bundles O(n, m). This is done by the same techniques
as in §4 (where the nilpotence variety was P1 × P3 ). As before, we denote the component
field multiplets associated to these modules bundles by

µA• (Γ∗ (O(n, m))) = µA• (n, m). (8.32)

Recall that every homogeneous piece of the graded global section module is a finite
dimensional representation of sl2 × sl3 . In terms of Dynkin labels, we can write for our
line bundles:

Γ∗ (O(n, m))d = H 0 (O(n + d, m + d)) = [n + d|m + d, 0]. (8.33)

Again, the multiplets µA• (n, m) and µA• (n + k, m + k) are identical up to a degree shift.
Thus, it is sufficient to consider the line bundles O(n, 0) and O(0, m) for m, n ≥ 0 and
it remains to describe the multiplets A• (n, 0) for n > 0.

Investigating the component fields, we find that the multiplet µA• (1, 0) is the antifield
multiplet of the canonical multiplet.
 
◦ Ω1,• (C3 )
• #
 U 
µA (1, 0) = 
 
 (8.34)
Ω0,• (C3 )
250 Differential operators and twisted (2,0) supersymmetry

For µA• (2, 0) one finds

h i
µA• (2, 0)# = S 2 U ◦ ⊗ Ω0,• (C3 ) U ◦ ⊗ Ω1,• (C3 ) Ω2,• (C3 ), , (8.35)

which is the dual of the derivations (8.24) which we identified with E(3|6).

For n ≥ 3 one finds the following field content.

h i
µA• (n, 0)# = S n U ◦ ⊗ Ω0,• S n−1 U ◦ ⊗ Ω1,• S n−2 U ◦ ⊗ Ω2,• S n−3 U ◦ ⊗ Ω3,•
(8.36)
We can also describe the graded global section modules of these line bundles explicitly
• 6d
as cokernels. Let us denote the ring of polynomial functions on n6d
Q by R = Sym (nQ ) =
C[λµi ] with µ = 1, . . . , 3 and i = 1, 2. Then, for example Γ∗ (O(1, 0)) is the cokernel of
the map
ϕ : C3 ⊗ R −→ U ◦ ⊗ R (8.37)

that is given in components by the matrix

!
λ11 λ21 λ31
ϕ= . (8.38)
λ12 λ22 λ32

In general, for n ≥ 1, Γ∗ (O(n, 0)) is the cokernel of the map

S n−1 U ◦ ⊗ C3 ⊗ R −→ S n U ◦ ⊗ R , F 7→ λµ(in Fi1 ...in−1 )µ (8.39)

and the multiplets described above can be obtained by considering their minimal free
resolutions.

8.4.2 Duality and the Cohen–Macaulay condition

Based on the component field multiplet constructed above (which take values in the
minimal free resolutions of the input modules), we can immediately deduce the following.

Lemma 8.4.1. The R-modules Γ∗ (O(n, 0)) are Cohen–Macaulay if and only if n ∈
{−1, 0, 1, 2}.

In fact, we can compute the Ext-groups explicitly for all line bundles. Recall that the
dualizing sheaf on P1 × P2 is O(−2, −3) and that the codimension of the nilpotence
variety is two. On general grounds, one therefore finds

Ext−2 ∼
R (Γ(n,0) , R) = Γ(−n−2,−3) = Γ(1−n,0) (3) = Γ(0,n−1) (n + 2). (8.40)
Differential operators and twisted (2,0) supersymmetry 251

For n ∈ {−1, 0, 1, 2}, this is the only Ext-module; outside of this range, there is an addi-
tional contribution. For n ≥ 3 the only other non-vanishing Ext-group is Ext3R (Γ∗ (O(n, 0)), R),
which has the following equivariant decomposition

n−3
M
Ext−3
R (Γ(n,0) , R) = [n − 3 − k|k, 0](−k + 3). (8.41)
k=0

In addition for the Γ(0,m) with m ≥ 2, there is an additional contribution

m−2
M
Ext−4
R (Γ(0,m) , R) = [k|m − k − 2, 0](−k + 2). (8.42)
k=0

For n in the Cohen–Macaulay range, we have

µA• (n, 0)∨ = µA• (1 − n, 0)[3] = µA• (0, n − 1)[n + 2]. (8.43)

Outside this range, the dualizing complex is not quasi-isomorphic to a single module,
but only to the above Ext-algebras.

In summary, we constructed four series of multiplets each parametrized by a natural

number: the multiplets A• (n, 0), A• (0, m), and their respective antifield multiplets. Only
within the Cohen–Macaulay range these overlap. We will now see how these appear in
the comparison to minimally twisted eleven-dimensional supergravity.

8.4.3 The comparison to minimally twisted supergravity

Minimally twisted eleven-dimensional supergravity and E(5|10). The com-

ponent fields of minimally twisted eleven-dimensional supergravity at free level were
described in [SW21; Ced21], interactions were conjectured (and numerous consistency
checks performed) in [RSW23]. Explicitly, the free component fields on C5 × R are the
Z/2Z graded BV theory described by the following cochain complex.
 
 ¯
(Ω0,• (C5 ), ∂) 
 
 ∂ 
 
 ⊗ (Ω• (R), d)
 

 ¯
(Ω1,• (C5 ), ∂) ¯
(PV1,• (C5 ), ∂) 
 ∂Ω

 
 
 
¯
(PV0,• (C5 ), ∂)
(8.44)
252 Differential operators and twisted (2,0) supersymmetry

Further, in [RSW23] it is also shown that the minimal model of the eleven-dimensional
supergravity multiplet is an L∞ central extension of the infinite-dimensional simple super
Lie algebra E(5|10).

Recall that E(5|10) can be described as follows. Its even piece consists of divergence free
vector fields on C5 ,
E(5|10)+ = Vect0 (C5 ), (8.45)

while its odd piece is closed two forms

E(5|10)− = Ω2cl (C5 ). (8.46)

The bracket of two even elements is just the bracket of vector fields, even elements act
on the odd piece via the Lie derivative, and the bracket of two odd elements is defined
by
[α, β] = ιΩ−1 (α ∧ β), (8.47)

where Ω−1 is the holomorphic volume form on C5 . It is immediate to see that he

cohomology of the complex (8.44) coincides with E(5|10) up tp a copy of C; this copy is
responsible for the central extension [RSW23].

Relating twisted supersymmetry in six and eleven dimensions. Recall that

choosing a minimal square-zero supercharge Q ∈ g11d fixes a decomposition V11 = L ⊕
L∨ ⊕ C and that the twisted supertranslation algebra n11d
Q takes the form

nQ ∼
= ∧2 L(−1) ⊕ L∨ (−2). (8.48)

It is clear that this algebra embeds into E(5|10) as constant two-forms and vector fields.

We now decompose the five-dimensional isotropic subspace of holomorphic translations

L∨ as
L∨ = Z ∨ ⊕ W ∨ , (8.49)

where Z is of dimension three and W is of dimension two. It is intuitive to think of this

decomposition in terms of the worldvolume of an M5 brane with Z corresponding to the
holomorphic directions along the brane and W to the holomorphic direction transverse
to the brane.

Then the exterior square decomposes as

∧2 L = ∧2 Z ⊕ (Z ⊗ W ) ⊕ ∧2 W. (8.50)
Differential operators and twisted (2,0) supersymmetry 253

Using this decomposition, we can equip n11d

Q with the following Z × Z-grading.

1 2
−1 ∧2 W
(8.51)
0 Z ⊗W Z∨
1 ∧2 Z W∨

By identifying the Z ∨ with the space of holomorphic translations determined by the

minimal twist in six dimensions and W with the the residual R-symmetry representation
U ◦ , we see that the subalgebra sitting in degree zero corresponds to the minimally twisted
six-dimensional (2, 0) algebra n6d
Q.

Similar to the symmetry enhancement of n11d

Q to E(5|10), the residual supersymmetry
algebra n6d
Q also enhances to the infinite dimensional super Lie algebra E(3|6) [RW22].
The even piece of E(3|6) consists of holomorphic vector fields and sl2 -valued holomorphic
functions on C3
E(3|6)+ = Vect(C3 ) ⊕ O(C3 ) ⊗ sl2 , (8.52)

while its odd piece is holomorphic one forms with values in the fundamental representa-
tion of sl2 ,
E(3|6)− = Ω1 (C3 ) ⊗ C2 . (8.53)

Further, the relation between n11d 6d

Q and nQ is compatible with these symmetry enhance-
ments, i.e. there is an additional grading on E(5|10) such that E(3|6) is a subalge-
bra sitting in degree zero and such that every graded piece of E(5|10) is an E(3|6)-
module [KR01].

The comparison at component field level. In [RW22] the aforementioned decom-

position of E(5|10) into E(3|6)-modules was investigated in the context of minimally
twisted eleven-dimensional supergravity and used to propose a holographic approach to
the six-dimensional superconformal index. This decomposition, called the fivebrane de-
composition, is concentrated in degrees above −1, and has E(3|6) sitting in degree zero.
The individual graded components were explicitly described in [RW22].

Observation 8.4.2. The degree n piece of the fivebrane decomposition constructed

in [RW22] is identical to µA• (n + 2, 0)∨ .

In terms of holography, it is instructive to think of the different pieces appearing in

the decomposition as analogous to the Kaluza-Klein modes of the bulk theory (as we
haven’t included the backreaction in any meaningful way, this comparison is more at
a schematic level). These modes are dual to CFT operators living on the boundary
254 Differential operators and twisted (2,0) supersymmetry

via a coupling prescription. Mathematically, this coupling prescription is realized by

Koszul duality [CL16; CP21]. One can speculate that the appearance of the duals in the
comparison 8.4.2 is related to this picture.

In the maximal twist, we were able to directly identify the different orders of differential
operators with powers in the coordinate transverse to the brane. Here, however, the
situation is slightly more complicated. Recall that the differential operators on the
canonical multiplet correspond to the line bundles O(0, m) with m ≥ 0. However, the
multiplets appearing the comparison 8.4.2 are the antifield multiplets of those associated
to O(n, 0). We can think of these as being generated by the dualizing complexes of
O(n, 0). In the Cohen–Macaulay range, the dualizing complex is quasi-isomorphic to
a dualizing sheaf and can be identifed (up to a global shift) with a line bundle of the
form O(0, m). In general, however, an additional Ext-group is present, such that the
multiplets in the decomposition 8.4.2 differ from differential operators on superspace. It is
tempting to speculate that forming the procedure of computing the dualizing complex on
the nilpotence variety acts as a proxy for Koszul duality in this holographic comparison.

We note that this complication is not visible in the maximal twist as the maximally
twisted nilpotence variety is just a point and all modules are automatically Cohen–
Macaulay.

The comparison at pure spinor cochain level. Recall that the twisted nilpotence
variety of eleven-dimensional minimal supersymmetry is the cone over the affine Grass-
mannian Gr(2, 5). In terms of the decomposition (8.49), we can describe its ring of
functions as follows. We identify polynomial functions on (n11d
Q )1 as

R = C[λww , λzw , λzz ], (8.54)

with
λww ∈ ∧2 W λzw ∈ Z ⊗ W λzz ∈ ∧2 Z. (8.55)

The defining ideal of the nilpotence variety takes the form

I = (λ2zw + λww λzz , λzz λzw ). (8.56)

In this way, the canonical multiplet can be described as

A•11d (OGr(2,5) ) = (C ∞ (Z × W ) ⊗ C[θww , θzw , θzz ] ⊗ R/I , D) , (8.57)

Differential operators and twisted (2,0) supersymmetry 255

where the differential is of the form

∂ ∂ ∂ ∂
D =λzz + λzw + λww + (λzz θzw + λzw θzz )
∂θzz ∂θzw ∂θww ∂w
(8.58)
∂
+ (λzz θww + λww θzz + 2λzw θzw ) .
∂z

Recall that the ring of functions on the minimally twisted six-dimensional nilpotence
variety is obtained from the above variety by intersecting with the plane where λww =
0 = λzz such that its ring of functions is given by C[λzw ]/(λ2zw ).

We now start from the canonical multiplet in eleven dimensions and apply a spectral
sequence to find a quasi-isomorphism to the multiplets obtained from line bundles over
the minimally twisted nilpotence variety in six dimensions. This can be seen as a lift of
the identification 8.4.2 to pure spinor cochain level.

Theorem 8.4.3. There is a quasi-isomorphism

M
A•11d (OGr(2,5) ) ' A•6d (n, 0)∨ . (8.59)
n≥1

Proof. We observe that we can filter the complex A• (OGr(2,5) ) by assigning weight zero
to any combination of λ variables and weight one to the remainder. Considering this
filtration together with the canonical filtration allows us to construct a spectral sequence
beginning with the cohomology of any individual term λ∂/∂θ.

We want to compute the cohomology with respect to the differential

∂ ∂
D̃ = λww + λzz . (8.60)
∂θww ∂θzz

This can be done in the following way. First, one replaces R/I with a minimal free
resolution (L, dL ) in free R-modules, such that one obtains a quasi-isomorphic complex
   
A• (OGr(2,5) ), D̃ ' C ∞ (Z × W ) ⊗ C[θww , θzw , θzz ] ⊗ L• , D̃ + dL . (8.61)

The precise form of the minimal free resolution can easily be computed using Macaulay2.
Working with the complex on the right hand side, we now consider the spectral sequence
that first takes cohomology with respect to D̃. As we have resolved in free R-modules,
the λww and θww as well as λzz and θzz form trivial pairs. Thus, one obtains for the
complex on the first page

(C ∞ (Z × W ) ⊗ C[θzw ] ⊗ (L• ⊗R C[λzw ]) , dL |λww =0=λzz ) , (8.62)

256 Differential operators and twisted (2,0) supersymmetry

where the differential is just obtained by restricting the resolution differential. In a

second step, we take cohomology with respect to the remaining pieces of the resolution
differential. One finds that the cohomology is concentrated in degrees zero and one. In
degree one we find,

C ∞ (Z × W ) ⊗ C[θzw , λzw ]/(λ2zw ) = A•6d (0, 0). (8.63)

The degree one piece is identified as

C ∞ (Z × W ) ⊗ C[θzw ] ⊗ Γ(1,0) . (8.64)

In terms of representatives in the total complex, we can think of the degree one piece as
being generated by the elements λzw θzz .

It now suffices to remember that the total complex carries an additional differential of
∂
the form λzw θzz ∂w . This induces a map between the piece in resolution degrees zero and
one. Expanding in polynomial degree in the variable w, this map is of the form

C ∞ (Z)[θzw , λzw ]/(λ2zw ) ⊗ Symk (W ) −→ C ∞ (Z)[θzw ] ⊗ Γ(1,0) Symk−1 (W ). (8.65)

This map can be explicitly analyzed and the cohomology computed using computer
algebra software. As a result, one finds
M
C ∞ (Z)[θzw ] ⊗ Ext•R (Γ(n,0) , R). (8.66)
n≥1

In the Cohen–Macaulay range, i.e. for n = 1 and n = 2, this already gives the desired
result (see (8.43)). For n ≥ 3 there are two Ext-groups present and there is an additional
differential arising via homotopy transfer which provides the equivalence to the antifield
multiplet µA• (n, 0)∨ . Explicitly, let h be a homotopy for the differential λzw θzz ∂w
∂
; then
the differential is given by terms of the form
   
∂ ∂
λzw ◦ h ◦ λzw (8.67)
∂θzw ∂θzw

which provide maps

C ∞ (Z)[θzw ] ⊗ Ext−3 ∞ −2
R (Γ(n,0) , R) −→ C (Z)[θzw ] ⊗ ExtR (Γ(n,0) , R). (8.68)
Differential operators and twisted (2,0) supersymmetry 257

8.5 Untwisted physics

Working with the full nilpotence variety of (2,0) supersymmetry, one can hope to spell
out the untwisted version of this story. Here, we take first steps in this direction by com-
puting the derivations of the abelian tensor multiplet (which is the canonical multiplet of
six-dimensional (2,0) supersymmetry), investigating the some symmetric powers of the
corresponding modules and their Ext-groups.

It is straightforward to compute the multiplet corresponding to the tangent sheaf of

Spec A• . Its component fields in degree one are those of the N = (2, 0) conformal super-
gravity multiplet of Bergshoeff, Sezgin, and van Proeyen [BSVP86]. In our conventions,
the underlying dg vector bundle takes the following form:

Vect S+ ⊗ [01] Ω0 ⊗ sp(2)

(8.69)
Ω1 ⊗ sp(2)
Sym2 (T R6 )0 Σ+
3/2 ⊗ [01] S− ⊗ [11] Ω0 ⊗ [20]
Ω3+ ⊗ [10]

Here, we use B2 Dynkin labels, so that [10] is the five-dimensional vector representation
of Spin(5) and [01] the four-dimensional spin representation of Spin(5)—equivalently, the
defining representation of sp(2). Σ+
3/2 denotes the spin-3/2 piece in the tensor product
Ω1 ⊗ S+ .

In principle, one can now move on and mimic our procedure from the minimal twist. Let
Γ denote the module such that A• (Γ) ' Der(A• (R/I)) as specified in Theorem 8.2.1.
Forming symmetric powers, one can compute the multiplets of higher order differential
operators. Based on the appearance of the dualizing complexes in the minimal twist,
one should then move on to investigate the Ext-groups Ext•R (S k Γ, R).

For k = 1, we immediately that the module Γ is Cohen-Macaulay such that the only non-
vanishing Ext-group is Ext5R (Γ, R). The associated multiplet is then the dual of (8.69)
and represents the currents to which the conformal supergravity multiplet couples.

For k ≥ 2, the modules S k Γ are no longer Cohen–Macaulay. In analogy to the minimally

twisted case, one expects the multiplets A• (Ext5R (S k Γ, R))∨ to appear in an untwisted
version of the comparison theorem 8.4.3, or respectively µA• (Ext5R (S k Γ, R))∨ at compo-
nent field level. Unfortunately, the computation of these Ext-groups and the associated
multiplets can be quite difficult in general. For k = 2, we find a multiplet with the
following Betti numbers
h i
grdim(µA• (Ext5R (S 2 Γ, R))∨ ) = 10 80 250 400 350 160 30 . (8.70)
258 Differential operators and twisted (2,0) supersymmetry

For k = 3, the Betti numbers are the following

h i
grdim(µA• (Ext5R (S 3 Γ, R))∨ ) = 1 16 110 400 840 1056 786 320 55 . (8.71)

We plan to investigate these multiplets and their relation to the Kaluza-Klein modes of
eleven-dimensional supergravity in future work. Here, we simply notice the following:
Recall that the abelian tensor multiplet contains a scalar field φ with values in the five-
dimensional vector representation of the R-symmetry group; the Dynkin label of this
representation is [1, 0]. The top component of the conformal supergravity multiplet (8.69)
is a scalar field with values in the representation [2, 0]. As explained in [BSVP86], this is
the field coupling to a quadratic current in the scalar field φ. The top components of the
multiplets in (8.70) and (8.71) are of dimensions 30 = dim([3, 0]) and 55 = dim([4, 0])
such that one is lead to suspect that these couple to cubic and quartic expressions in φ.

8.6 Conformal supergravity from derivations

In the following, let us collect the field contents of the multiplets Der(A• (R/I)) for various
dimensions and amounts of supersymmetry. For this purpose, we use the description
by Theorem 8.2.1 together with the techniques for the extraction of component fields
described described in §2. The calculations were performed using Macaulay2.

8.6.1 Dimension one

In one dimension and N supercharges, the nilpotence variety is defined by the single
quadratic equation,
λ21 + · · · + λ2N = 0. (8.72)

The map λγ, is simply given by the matrix

(λ1 , . . . , λN ), (8.73)

whose cokernel is just C. Hence, Der(A• (R/I)) can be identified with the free superfield,

Der(A• (R/I)) ' C ∞ (R) ⊗ ∧• (CN ). (8.74)

8.6.2 Dimension three

Recall that Spin(3) = SU(2); we denote the two-dimensional spin representation by S

and the three dimensional vector representation by V . The supertranslation algebra is
Differential operators and twisted (2,0) supersymmetry 259

of the form
S(−1) ⊗ U ⊕ V (−2), (8.75)

where U ∼= (CN , (−, −)) is equipped with a non-degenerate symmetric bilinear form and
the bracket is induced by the isomorphism Sym2 (S) ∼= V and (−, −).

N = 1. For the field content of Der(A• (R/I)) we find,

 
 Vect(R3 ) Ω0 ⊗S 
 
 
. (8.76)
 


 Sym20 (T R3 ) Σ3/2 

 

Here Σ3/2 = [3] denotes the spin 3/2 representation of SU(2). We note that this multiplet
coincides with the multiplet associated to the conormal module I/I 2 which we discussed
in §2.6.

N = 2. For N = 2 one finds the following field content for Der(A• (R/I)).
 
3 Ω0 ⊗ S ⊗ C2 Ω0
 Vect(R ) 
 
 
(8.77)
 
 

 Sym20 (T R3 ) Σ3/2 ⊗ C2 1
Ω 
 

8.6.3 Dimension four

We identify Spin(4) ∼
= SU (2) × SU (2) and denote the two two-dimensional spin repre-
sentations by S± . The supertranslation algebra is of the form

(S+ ⊗ U ) ⊕ (S− ⊗ U ∨ ) (−1) ⊕ V (−2),

 
(8.78)

where the bracket is induced by the isomorphism V ∼

= S+ ⊗ S− and the natural pairing
between U and its dual. We denote the Dirac spin representation by S+ ⊕ S− = S and
−
the spin 3/2-pieces in the tensor product V ⊗ S as Σ3/2 = Σ+
3/2 ⊕ Σ3/2
260 Differential operators and twisted (2,0) supersymmetry

N = 1. One finds the following field content, which matches the conformal supergravity
multiplet discussed in [FVP12].
 
 Vect(R4 ) Ω0 ⊗ S+ ⊕ S− Ω0 
 
 
(8.79)
 
 
−

 Sym20 (T R4 ) Σ+
3/2 ⊕ Σ3/2 Ω1 

 

N = 2. Again, we find the following field content, matching the ‘Weyl multiplet’ of
conformal supergravity as discussed in [DVV80].
 
4 Ω0 ⊗ S ⊗ U Ω0 ⊗ u(2)R
 Vect(R ) 
 
 
 
 

 Sym20 (T R4 ) Σ3/2 ⊗ U Ω2 ⊕ Ω1 ⊗ u(2)R S⊗U Ω0 

 

(8.80)

8.6.4 Dimension six

N = (1, 0). One finds the field content of the Weyl multiplet as presented in [BSVP86;
CVP11]. Note that this is the same field content we obtained from the conormal module
in §4.
 
6 Ω0 ⊗ S+ ⊗ U
 Vect(R ) sp(1) 
 
 
 
 
Sym20 (T R6 ) Σ+
3/2 Ω1 ⊗ sp(1) ⊕ Ω3− Ω0 ⊗ S+ ⊗ U Ω0
(8.81)

N = (2, 0). See (8.69).

Differential operators and twisted (2,0) supersymmetry 261

8.6.5 Dimension ten

N = 1. One finds a multiplet with the following field content.

 
 Vect(R10 ) Ω0 ⊗ S+ 
 
 
 
 
Sym20 (T R10 ) Ω1 ⊗ S+ Ω3 ⊕ Ω1
 
 
 
 
 
 
Ω2 ⊕ Ω0 Ω0 ⊗ S−
(8.82)
This multiplet is closely related to the gravity multiplets constructed in [BRW83] and [BR82].
Chapter 9

Outlook

In the following, we outline some directions for future research that fit naturally with
the work presented in this thesis.

Pure spinor superfields on coset spaces. As presented in this thesis, the pure
spinor superfield formalism constructs multiplets on flat spacetime Rd . Of course, once
constructed, we can then put these multiplets on any appropriately structured manifold,
however, this procedure will break supersymmetry in most cases. The fact that multiplets
constructed via pure spinors primarily live on Rd stems from the requirement on the super
Lie algebra to be of super Poincaré type. The construction then associates a multiplet
on the supertranslation group given by N = exp(n) = exp(g/g0 ) (whose even piece is
Rd ). However, supersymmetric theories also exist in other backgrounds, and there the
natural super Lie algebras in consideration are not of super Poincaré type. Here, the
most interesting examples are Anti de Sitter spaces. For instance, in ten dimensions
AdS5 × S 5 arises as the even subspace of the quotient

AdS5 × S 5 ∼
= (PSU(2, 2|4)/(SO(1, 4) × SO(5)))even . (9.1)

Generalizing the pure spinor superfield formalism to such pairs of super Lie algebras
seems to be a promising way to construct supersymmetric theories in curved backgrounds.
Given the relation between pure spinors and twisting this could also pave the way to
direct twist calculations in backreacted geometries.

Superconformal nilpotence varieties and twists. In some supersymmetric field

theories, the action of the super Poincaré algebra enhances to an action of the supercon-
formal algebra. The natural moduli space of twists for such superconformal field theories
is the nilpotence variety of the superconformal algebra. Let g denote a super Poincaré
263
264 Outlook

algebra and U the corresponding superconformal algebra. Clearly, there is an injective

map of the nilpotence varieties,
Yg ,→ YU , (9.2)

signaling that every twist of the underlying supersymmetric theory also defines a twist
of the superconformal theory. In general, however, this map is not surjective, i.e. the
superconformal theory admits additional twists. Typically, one denotes supercharges
in the super Poincaré algebra by Q and the additional superconformal supercharges
by S. Mixed twists of the form Q + S played a big role in their relation to chiral
algebras [Bee+15] as well as to the omega background [OY19]. A thorough investigation
of these nilpotence varieties and the corresponding twist with an eye towards the pure
spinor superfield formalism (in particular in negatively curved backgrounds as suggested
in the previous paragraph) seems worthwhile. This is in particular true in the context of
twisted holography, where the additional twists in the super Poincaré algebra are dual
to twists of the supergravity theory in the backreacted geometry.

The pure spinor superstring: Relating worldsheet and target space. In this
thesis, we developed the pure spinor superfield formalism as a method for the construction
and analysis of supersymmetric theories in target space. Of course pure spinor methods
also play a big role in worldsheet superstring theory [Ber00]. Understanding worldsheet
twists in this formalism, their relation to techniques from topological string theory as
well as to the target space perspective developed in this work seems to be a crucial for
developing a more holistic understanding of twisted holography.
Bibliography

[AD17] D. Alekseevsky and L. David. “Prolongation of Tanaka structures: an al-

ternative approach”. Annali di Matematica Pura ed Applicata (1923-) 196
(2017), pp. 1137–1164. url: https://doi.org/10.1007/s10231- 016-
0610-7.
[Ale+07] V. Alexandrov, D. Krotov, A. Losev, and V. Lysov. “On Pure Spinor Su-
perfield Formalism”. JHEP 10 (2007), p. 074. arXiv: 0705.2191 [hep-th].
[All10] M. P. Allocca. L∞ Algebra Representation Theory. Thesis (Ph.D.)–North
Carolina State University. 2010. url: https : / / repository . lib . ncsu .
edu/bitstream/handle/1840.16/7035/etd.pdf.
[Asp04] P. S. Aspinwall. “D-branes on Calabi-Yau manifolds”. Theoretical Advanced
Study Institute in Elementary Particle Physics (TASI 2003): Recent Trends
in String Theory. Mar. 2004, pp. 1–152. arXiv: hep-th/0403166.
[AG71] L. Avramov and E. Golod. “Homology algebra of the Koszul complex of
a local Gorenstein ring”. Mathematical notes of the Academy of Sciences
of the USSR 9.1-3 (1971), pp. 30–32. url: https://doi.org/10.1007/
BF01405047.
[BV16] D. Bashkirov and A. A. Voronov. “The BV formalism for L∞ -algebras”.
Journal of Homotopy and Related Structures 12.2 (2016), pp. 305–327. arXiv:
1410.6432 [math.QA]. url: https://doi.org/10.1007%2Fs40062-016-
0129-z.
[Bas63] H. Bass. “On the ubiquity of Gorenstein rings”. Mathematische Zeitschrift
82.1 (1963), pp. 8–28. url: https://doi.org/10.1007/BF01112819.
[BV81] I. A. Batalin and G. A. Vilkovisky. “Gauge Algebra and Quantization”.
Phys. Lett. B 102 (1981), pp. 27–31. url: https://doi.org/10.1016/
0370-2693(81)90205-7.
[BV84] I. A. Batalin and G. A. Vilkovisky. “Closure of the gauge algebra, general-
ized Lie equations, and Feynman rules”. Nucl. Phys. B 234 (1984), pp. 106–
124. url: https://doi.org/10.1016/0550-3213(84)90227-X.
265
Bibliography Outlook

[BV85] I. A. Batalin and G. A. Vilkovisky. “Existence theorem for gauge algebra”.

J. Math. Phys. 26 (1985), pp. 172–184. url: https://doi.org/10.1063/
1.526780.

[Bau11] L. Baulieu. “SU(5)-invariant decomposition of ten-dimensional Yang-Mills

supersymmetry”. Phys. Lett. B 698 (2011), pp. 63–67. arXiv: 1009.3893
[hep-th].

[Bau+90] L. Baulieu, M. P. Bellon, S. Ouvry, and J.-C. Wallet. “Balatin-Vilkovisky

analysis of supersymmetric systems”. Phys. Lett. B 252 (1990), pp. 387–394.
url: https://doi.org/10.1016/0370-2693(90)90557-M.

[Bec+18] K. Becker, M. Becker, D. Butter, and W. D. Linch. “N = 1 supercurrents

of eleven-dimensional supergravity”. JHEP 05 (2018), p. 128. arXiv: 1803.
00050 [hep-th].

[Bec+16] K. Becker, M. Becker, S. Guha, W. D. Linch, and D. Robbins. “M-theory

potential from the G2 Hitchin functional in superspace”. JHEP 12 (2016),
p. 085. arXiv: 1611.03098 [hep-th].

[Bec+21] K. Becker, D. Butter, W. D. Linch, and A. Sengupta. “Components of

Eleven-dimensional Supergravity with Four Off-shell Supersymmetries” (Jan.
2021). arXiv: 2101.11671 [hep-th].

[Bec+17] K. Becker et al. “Eleven-dimensional supergravity in 4D, N = 1 superspace”.

JHEP 11 (2017), p. 199. arXiv: 1709.07024 [hep-th].

[BRR15] C. Beem, L. Rastelli, and B. C. van Rees. “W symmetry in six dimensions”.

JHEP 05 (2015), p. 017. arXiv: 1404.1079 [hep-th].

[Bee+15] C. Beem et al. “Infinite chiral symmetry in four dimensions”. Comm. Math.
Phys. 336.3 (2015), pp. 1359–1433.

[BR82] E. Bergshoeff and M. de Roo. “The Supercurrent in Ten-dimensions”. Phys.

Lett. B 112 (1982), pp. 53–58.

[BRW83] E. Bergshoeff, M. de Roo, and B. de Wit. “Conformal Supergravity in Ten-

dimensions”. Nucl. Phys. B 217 (1983). Ed. by A. Salam and E. Sezgin,
p. 489.

[BSVP86] E. Bergshoeff, E. Sezgin, and A. Van Proeyen. “Superconformal Tensor Cal-

culus and Matter Couplings in Six-dimensions”. Nucl. Phys. B 264 (1986).
Ed. by A. Salam and E. Sezgin. [Erratum: Nucl.Phys.B 598, 667 (2001)],
p. 653.

[Ber00] N. Berkovits. “Super Poincare covariant quantization of the superstring”.

JHEP 04 (2000), p. 018. arXiv: hep-th/0001035.
Bibliography 267

[Ber01] N. Berkovits. “Covariant quantization of the superparticle using pure spinors”.

JHEP 09 (2001), p. 016. arXiv: hep-th/0105050.

[Ber02] N. Berkovits. “Towards covariant quantization of the supermembrane”. JHEP

09 (2002), p. 051. arXiv: hep-th/0201151.

[Ber05] N. Berkovits. “Pure spinor formalism as an N=2 topological string”. JHEP

10 (2005), p. 089. arXiv: hep-th/0509120.

[BG18] N. Berkovits and M. Guillen. “Equations of motion from Cederwall’s pure

spinor superspace actions”. JHEP 08 (2018), p. 033. arXiv: 1804 . 06979
[hep-th].

[BN05] N. Berkovits and N. Nekrasov. “The character of pure spinors”. Lett. Math.
Phys. 74 (2005), pp. 75–109. arXiv: hep-th/0503075.

[Ber+94] M. Bershadsky, S. Cecotti, H. Ooguri, and C. Vafa. “Kodaira-Spencer the-

ory of gravity and exact results for quantum string amplitudes”. Comm.
Math. Phys. 165.2 (1994), pp. 311–427. url: http://projecteuclid.org/
euclid.cmp/1104271134.

[BS96] M. Bershadsky and V. Sadov. “Theory of Kähler gravity”. Int. J. Mod. Phys.
A 11 (1996), pp. 4689–4730. arXiv: hep-th/9410011.

[BH80] L. Brink and P. S. Howe. “Eleven-Dimensional Supergravity on the Mass-

Shell in Superspace”. Phys. Lett. B 91 (1980), pp. 384–386. url: https:
//doi.org/10.1016/0370-2693(80)91002-3.

[BE77] D. A. Buchsbaum and D. Eisenbud. “Algebra structures for finite free reso-
lutions, and some structure theorems for ideals of codimension 3”. American
Journal of Mathematics 99.3 (1977), pp. 447–485.

[CS09] A. Cap and J. Slovák. Parabolic geometries I. 154. American Mathematical

Soc., 2009.

[CdF91] L. Castellani, R. d’Auria, and P. Fré. Supergravity and superstrings: a geo-

metric perspective (in three volumes). World Scientific Publishing Company,
1991.

[Ced08] M. Cederwall. “N=8 superfield formulation of the Bagger-Lambert-Gustavsson

model”. JHEP 09 (2008), p. 116. arXiv: 0808.3242 [hep-th].

[Ced10a] M. Cederwall. “D = 11 supergravity with manifest supersymmetry”. Mod.

Phys. Lett. A 25 (2010), pp. 3201–3212. arXiv: 1001.0112 [hep-th].

[Ced10b] M. Cederwall. “Operators on pure spinor spaces”. AIP Conference Proceed-

ings. Vol. 1243. 1. American Institute of Physics. 2010, pp. 51–59.
Bibliography Outlook

[Ced10c] M. Cederwall. “Towards a manifestly supersymmetric action for eleven-

dimensional supergravity”. JHEP 01 (2010), p. 117. arXiv: 0912 . 1814
[hep-th].

[Ced14] M. Cederwall. “Pure spinor superfields – an overview”. Springer Proc. Phys.

153 (2014). Ed. by S. Bellucci, pp. 61–93. arXiv: 1307.1762 [hep-th].

[Ced18a] M. Cederwall. “An off-shell superspace reformulation of D = 4, N = 4 super

Yang–Mills theory”. Fortschritte der Physik 66.1 (2018), p. 1700082. url:
http://dx.doi.org/10.1002/prop.201700082.

[Ced18b] M. Cederwall. “Pure spinor superspace action for D = 6, N = 1 super

Yang–Mills theory”. JHEP 05 (2018), p. 115. arXiv: 1712.02284 [hep-th].

[Ced21] M. Cederwall. “SL(5) supersymmetry”. Fortsch. Phys. 69.11-12 (2021), p. 2100116.

arXiv: 2107.09037 [math.RT].

[Ced+00a] M. Cederwall, U. Gran, M. Nielsen, and B. E. W. Nilsson. “Generalized

11-dimensional supergravity”. International Conference on Quantization,
Gauge Theory, and Strings: Conference Dedicated to the Memory of Profes-
sor Efim Fradkin. Oct. 2000. arXiv: hep-th/0010042.

[Ced+00b] M. Cederwall, U. Gran, M. Nielsen, and B. E. W. Nilsson. “Manifestly su-

persymmetric M theory”. JHEP 10 (2000), p. 041. arXiv: hep-th/0007035.

[Ced+05] M. Cederwall, U. Gran, B. E. W. Nilsson, and D. Tsimpis. “Supersymmetric

corrections to eleven-dimensional supergravity”. JHEP 05 (2005), p. 052.
arXiv: hep-th/0409107.

[Ced+23] M. Cederwall, S. Jonsson, J. Palmkvist, and I. Saberi. “Canonical super-

multiplets and their Koszul duals” (2023). arXiv: 2304.01258 [hep-th].

[CK11] M. Cederwall and A. Karlsson. “Pure spinor superfields and Born–Infeld

theory”. JHEP 11 (2011), p. 134. arXiv: 1109.0809 [hep-th].

[CNT01] M. Cederwall, B. E. W. Nilsson, and D. Tsimpis. “The Structure of maxi-

mally supersymmetric Yang-Mills theory: Constraining higher order correc-
tions”. JHEP 06 (2001), p. 034. arXiv: hep-th/0102009.

[CNT02] M. Cederwall, B. E. W. Nilsson, and D. Tsimpis. “Spinorial cohomology

and maximally supersymmetric theories”. JHEP 02 (2002), p. 009. arXiv:
hep-th/0110069.

[CN08] M. Cederwall and B. E. Nilsson. “Pure spinors and d = 6 super Yang–Mills”

(2008). arXiv: 0801.1428 [hep-th].

[Che54] C. C. Chevalley. The algebraic theory of spinors. Columbia University Press,

New York, 1954, pp. viii+131.
Bibliography 269

[CW21] J. Cirici and S. O. Wilson. “Dolbeault cohomology for almost complex man-
ifolds”. Advances in Mathematics 391 (2021), p. 107970. arXiv: 1809.01416
[math.DG].

[CVP11] F. Coomans and A. Van Proeyen. “Off-shell N=(1,0), D=6 supergravity

from superconformal methods”. JHEP 02 (2011). [Erratum: JHEP 01, 119
(2012)], p. 049. arXiv: 1101.2403 [hep-th].

[Cos] K. Costello. “Twisted M theory, the Maulik-Okounkov Yangian, and the

AdS dual of the Beem-Rastelli twist of the (2,0) theory”. url: http : / /
www.birs.ca/events/2015/5-day-workshops/15w5154/videos/watch/
201505251537-Costello.html.

[Cos07] K. Costello. “Topological conformal field theories and Calabi-Yau cate-

gories”. Adv. Math. 210 (2007), pp. 165–214. arXiv: math/0412149.

[Cos13a] K. Costello. “Notes on supersymmetric and holomorphic field theories in

dimensions 2 and 4”. Pure Appl. Math. Q. 9.1 (2013), pp. 73–165. arXiv:
1111.4234 [math.QA].

[Cos13b] K. Costello. “Supersymmetric gauge theory and the Yangian” (Mar. 2013).
arXiv: 1303.2632 [hep-th].

[Cos16] K. Costello. “M-theory in the Omega-background and 5-dimensional non-

commutative gauge theory” (Oct. 2016). arXiv: 1610.04144 [hep-th].

[Cos17] K. Costello. “Holography and Koszul duality: the example of the M 2 brane”
(May 2017). arXiv: 1705.02500 [hep-th].

[CG18] K. Costello and D. Gaiotto. “Twisted Holography” (Dec. 2018). arXiv: 1812.
09257 [hep-th].

[CG17] K. Costello and O. Gwilliam. Factorization algebras in quantum field the-

ory. Vol. 1. Vol. 31. New Mathematical Monographs. Cambridge University
Press, Cambridge, 2017, pp. ix+387.

[CG21] K. Costello and O. Gwilliam. Factorization algebras in quantum field the-

ory. Vol. 2. Vol. 41. New Mathematical Monographs. available at http :
/ / people . mpim - bonn . mpg . de / gwilliam. Cambridge University Press,
Cambridge, 2021.

[CL15] K. Costello and S. Li. “Quantization of open-closed BCOV theory, I” (May

2015). arXiv: 1505.06703 [hep-th].

[CL16] K. Costello and S. Li. “Twisted supergravity and its quantization” (June
2016). arXiv: 1606.00365 [hep-th].
Bibliography Outlook

[CP21] K. Costello and N. M. Paquette. “Twisted Supergravity and Koszul Duality:

A case study in AdS3 ”. Commun. Math. Phys. 384.1 (2021), pp. 279–339.
arXiv: 2001.02177 [hep-th].

[CF80] E. Cremmer and S. Ferrara. “Formulation of Eleven-Dimensional Super-

gravity in Superspace”. Phys. Lett. B 91 (1980), pp. 61–66.

[CJS78] E. Cremmer, B. Julia, and J. Scherk. “Supergravity Theory in Eleven-

Dimensions”. Phys. Lett. B 76 (1978), pp. 409–412.

[Crn88] Č. Crnković. “Symplectic geometry of the covariant phase space”. Classical

and Quantum Gravity 5.12 (1988), p. 1557.

[CG+18] T. C. da C. Guio, H. Jockers, A. Klemm, and H.-Y. Yeh. “Effective Action

from M-Theory on Twisted Connected Sum G2 -Manifolds”. Commun. Math.
Phys. 359.2 (2018), pp. 535–601. arXiv: 1702.05435 [hep-th].

[dRo+80] M. de Roo, J. W. van Holten, B. de Wit, and A. Van Proeyen. “Chiral

Superfields in N = 2 Supergravity”. Nucl. Phys. B 173 (1980), pp. 175–188.
url: https://doi.org/10.1016/0550-3213(80)90449-6.

[WS06] B. de Wit and F. Saueressig. “Off-shell N = 2 tensor supermultiplets”.

JHEP 09 (2006), p. 062. arXiv: hep-th/0606148.

[DVV80] B. De Wit, J. Van Holten, and A. Van Proeyen. “Transformation rules of

N = 2 supergravity multiplets”. Nuclear Physics B 167.1 (1980), pp. 186–
204. url: https : / / www . sciencedirect . com / science / article / pii /
055032138090125X.

[DZOZ22] M. Del Zotto, J. Oh, and Y. Zhou. “Evidence for an algebra of G2 instan-
tons”. JHEP 08 (2022), p. 214. arXiv: 2109.01110 [hep-th].

[Dij+05] R. Dijkgraaf, S. Gukov, A. Neitzke, and C. Vafa. “Topological M-theory as

unification of form theories of gravity”. Adv. Theor. Math. Phys. 9.4 (2005),
pp. 603–665. arXiv: hep-th/0411073.

[DS14] W. Donovan and E. Segal. “Window shifts, flop equivalences and Grass-
mannian twists”. Compos. Math. 150.6 (2014), pp. 942–978. url: https:
//doi.org/10.1112/S0010437X13007641.

[DSV15] V. Dotsenko, S. Shadrin, and B. Vallette. “De Rham cohomology and homo-
topy Frobenius manifolds”. Journal of the European Mathematical Society
17.3 (2015), pp. 535–547. url: http://dx.doi.org/10.4171/JEMS/510.

[Dou01] M. R. Douglas. “D-branes, categories and N=1 supersymmetry”. J. Math.

Phys. 42 (2001), pp. 2818–2843. arXiv: hep-th/0011017.

[DS05] M. Duflo and V. Serganova. “On associated variety for Lie superalgebras”
(2005). arXiv: math/0507198 [math.RT].
Bibliography 271

[EH23] R. Eager and F. Hahner. “Maximally Twisted Eleven-Dimensional Super-

gravity”. Commun. Math. Phys. 398.1 (2023), pp. 59–88. arXiv: 2106.15640
[hep-th].

[Eag+22] R. Eager, F. Hahner, I. Saberi, and B. R. Williams. “Perspectives on the

pure spinor superfield formalism”. J. Geom. Phys. 180 (2022), p. 104626.
arXiv: 2111.01162 [hep-th].

[ES19a] R. Eager and I. Saberi. “Holomorphic field theories and Calabi–Yau alge-
bras”. International Journal of Modern Physics A (2019). arXiv: 1805.02084
[hep-th].

[ESW21] R. Eager, I. Saberi, and J. Walcher. “Nilpotence varieties”. Annales Henri

Poincaré 22.4 (2021), pp. 1319–1376. arXiv: 1807.03766 [hep-th].

[ES15] R. Eager and J. Schmude. “Superconformal Indices and M2-Branes”. JHEP

12 (2015), p. 062. arXiv: 1305.3547 [hep-th].

[EST14] R. Eager, J. Schmude, and Y. Tachikawa. “Superconformal Indices, Sasaki-

Einstein Manifolds, and Cyclic Homologies”. Adv. Theor. Math. Phys. 18.1
(2014), pp. 129–175. arXiv: 1207.0573 [hep-th].

[Eis95] D. Eisenbud. Commutative algebra. Vol. 150. Graduate Texts in Mathemat-

ics. With a view toward algebraic geometry. Springer-Verlag, New York,
1995, pp. xvi+785. url: https://doi.org/10.1007/978-1-4612-5350-1.

[EHS23] C. Elliott, F. Hahner, and I. Saberi. “The Derived Pure Spinor Formalism as
an Equivalence of Categories”. SIGMA 19 (2023), p. 022. arXiv: 2205.14133
[math-ph].

[ES19b] C. Elliott and P. Safronov. “Topological twists of supersymmetric algebras

of observables”. Commun. Math. Phys. 371.2 (2019), pp. 727–786. arXiv:
1805.10806 [math-ph].

[ESW20] C. Elliott, P. Safronov, and B. Williams. “A taxonomy of twists of super-

symmetric Yang–Mills theory” (2020). arXiv: 2002.10517 [math-ph].

[EW21] C. Elliott and B. R. Williams. “Holomorphic Poisson Field Theories”. Higher

Structures 5 (Dec. 2021). arXiv: 2008.02302 [math-ph].

[ESS21] D. Erman, S. V. Sam, and A. Snowden. “Strength and Hartshorne’s conjec-

ture in high degree”. Math. Z. 297.3-4 (2021), pp. 1467–1471. url: https:
//doi.org/10.1007/s00209-020-02564-y.

[FG05] M. Faux and S. J. Gates. “Adinkras: A graphical technology for supersym-

metric representation theory”. Phys. Rev. D 71 (6 2005), p. 065002. url:
https://link.aps.org/doi/10.1103/PhysRevD.71.065002.
Bibliography Outlook

[Fer87] S. Ferrara, ed. Supersymmetry: Lectures And Reprints (in two volumes).
North Holland/World Scientific, 1987.
[FS11] G. Festuccia and N. Seiberg. “Rigid Supersymmetric Theories in Curved
Superspace”. JHEP 06 (2011), p. 114. arXiv: 1105.0689 [hep-th].
[Fon10] A. Font. “Heterotic strings on G2 orbifolds”. JHEP 11 (2010), p. 115. arXiv:
1009.4422 [hep-th].
[Fra18] R. Frankhuizen. “Massey products and the Golod property for simplicially
resolvable rings” (2018). arXiv: 1806.07887 [math.AT].
[FVP12] D. Z. Freedman and A. Van Proeyen. Supergravity. Cambridge, UK: Cam-
bridge Univ. Press, May 2012.
[Fri+86] D. Friedan et al. “Notes on string theory and two-dimensional conformal
field theory”. Unified string theories 162 (1986).
[GO19] D. Gaiotto and J. Oh. “Aspects of Ω-deformed M-theory” (July 2019). arXiv:
1907.06495 [hep-th].
[Gal16] F. Galetto. “Propagating weights of tori along free resolutions”. Journal of
Symbolic Computation 74 (2016), pp. 1–45. arXiv: 1406.1900 [math.AC].
[GIO87] A. S. Galperin, E. A. Ivanov, and V. I. Ogievetsky. “Duality Transformations
and Most General Matter Selfcoupling in N = 2 Supersymmetry”. Nucl.
Phys. B 282 (1987), p. 74.
[Gal+01] A. S. Galperin, E. Ivanov, V. I. Ogievetsky, and E. S. Sokatchev. Harmonic
superspace. Cambridge University Press, 2001.
[Gál+16] I. Gálvez, V. Gorbounov, Z. Shaikh, and A. Tonks. “The Berkovits complex
and semi-free extensions of Koszul algebras”. Ann. Fac. Sci. Toulouse Math.
(6) 25.2-3 (2016), pp. 363–384. url: https://doi.org/10.5802/afst.
1497.
[GS80] S. J. Gates Jr. and W. Siegel. “Understanding Constraints in Superspace
Formulations of Supergravity”. Nucl. Phys. B 163 (1980), pp. 519–545. url:
https://doi.org/10.1016/0550-3213(80)90414-9.
[GSW80] S. J. Gates Jr., K. S. Stelle, and P. C. West. “Algebraic Origins of Superspace
Constraints in Supergravity”. Nucl. Phys. B 169 (1980), pp. 347–364. url:
https://doi.org/10.1016/0550-3213(80)90037-1.
[GS04] A. A. Gerasimov and S. L. Shatashvili. “Towards integrability of topological
strings. I. Three-forms on Calabi-Yau manifolds”. JHEP 11 (2004), p. 074.
arXiv: hep-th/0409238.
[GS71] J.-L. Gervais and B. Sakita. “Field theory interpretation of supergauges in
dual models”. Nuclear Physics B 34.2 (1971), pp. 632–639.
Bibliography 273

[Get09] E. Getzler. “Lie theory for nilpotent L∞ algebras”. Annals of Mathematics

(2009), pp. 271–301. arXiv: math/0404003 [math.AT].

[Gin+22] G. Ginot, O. Gwilliam, A. Hamilton, and M. Zeinalian. “Large N phenom-

ena and quantization of the Loday-Quillen-Tsygan theorem”. Advances in
Mathematics 409 (2022), p. 108631. arXiv: 2108.12109 [math.QA].

[GL71] Y. A. Gol’fand and E. P. Likhtman. “Extension of the algebra of Poincaré

group generators and violation of P invariance”. JETP Lett. 13 (1971),
p. 323.

[Gol05] E. Golod. “On duality in the homology algebra of a Koszul complex”. J Math
Sci 128 (2005), pp. 3381–3383. url: https://doi.org/10.1007/s10958-
005-0276-y.

[GR+98] F. Gonzalez-Rey, M. Rocek, S. Wiles, U. Lindstrom, and R. von Unge.

“Feynman rules in N=2 projective superspace: 1. Massless hypermultiplets”.
Nucl. Phys. B 516 (1998), pp. 426–448. arXiv: hep-th/9710250.

[GS09] V. Gorbounov and V. Schechtman. “Homological algebra and divergent se-

ries”. SIGMA 5 (2009), p. 034. arXiv: 0712.3670 [math.AG].

[Gor+22] M. Gorelik, C. Hoyt, V. Serganova, and A. Sherman. “The Duflo-Serganova

functor, vingt ans après” (2022). arXiv: 2203.00529 [math.RT].

[GKR07] A. L. Gorodentsev, A. S. Khoroshkin, and A. N. Rudakov. “On syzygies

of highest weight orbits”. Moscow Seminar on Mathematical Physics. II.
Vol. 221. Amer. Math. Soc. Transl. Ser. 2. Amer. Math. Soc., Providence,
RI, 2007, pp. 79–120. url: https://doi.org/10.1090/trans2/221/05.

[GV05] P. A. Grassi and P. Vanhove. “Topological M theory from pure spinor for-
malism”. Adv. Theor. Math. Phys. 9.2 (2005), pp. 285–313. arXiv: hep -
th/0411167.

[GS] D. R. Grayson and M. E. Stillman. Macaulay2, a software system for re-

search in algebraic geometry. Available at http://www.math.uiuc.edu/
Macaulay2/.

[Hah20] F. Hahner. “The BV Formalism, L∞ -algebras, and Their Applications to

Supersymmetry and Chern–Simons theory” (2020). Master’s thesis, Univer-
sität Heidelberg.

[Hah+22] F. Hahner, S. Noja, I. Saberi, and J. Walcher. “Six-dimensional supermul-

tiplets from bundles on projective spaces” (June 2022). arXiv: 2206.08388
[math-ph].

[Hah+23] F. Hahner, S. Raghavendran, I. Saberi, and B. R. Williams. “” (2023). in

preparation.
Bibliography Outlook

[HS23] F. Hahner and I. Saberi. “Eleven-dimensional supergravity as a Calabi-Yau

twofold” (Apr. 2023). arXiv: 2304.12371 [math-ph].

[Har77] R. Hartshorne. Algebraic geometry. Graduate Texts in Mathematics, No. 52.

New York: Springer-Verlag, 1977, pp. xvi+496.

[HS93] V. Hinich and V. Schechtman. “Homotopy Lie algebras”. Adv. Soviet Math
16.2 (1993), pp. 1–28.

[Hit00] N. J. Hitchin. “The Geometry of Three-Forms in Six Dimensions”. J. Diff.

Geom. 55.3 (2000), pp. 547–576. arXiv: math/0010054.

[HE71] M. Hochster and J. A. Eagon. “Cohen-Macaulay Rings, Invariant Theory,

and the Generic Perfection of Determinantal Loci”. American Journal of
Mathematics 93.4 (1971), pp. 1020–1058. url: https : / / doi . org / 10 .
2307/2373744.

[HS19] Y. Homma and U. Semmelmann. “The Kernel of the Rarita–Schwinger Op-

erator on Riemannian Spin Manifolds”. Commun. Math. Phys. 370.3 (2019),
pp. 853–871. arXiv: 1804.10602 [math.DG].

[Hor+03] K. Hori et al. Mirror symmetry. Vol. 1. Clay mathematics monographs.

Providence, USA: AMS, 2003.

[HNv82] P. Howe, H. Nicolai, and A. van Proeyen. “Auxiliary fields and a super-
space Lagrangian for linearized ten-dimensional supergravity”. Physics Let-
ters B 112.6 (1982), pp. 446–450. url: https://doi.org/10.1016/0370-
2693(82)90845-0.

[How91a] P. S. Howe. “Pure spinor lines in superspace and ten-dimensional super-

symmetric theories”. Phys. Lett. B 258 (1991), pp. 141–144. url: https:
//doi.org/10.1016/0370-2693(91)91221-G.

[How91b] P. S. Howe. “Pure spinors, function superspaces, and supergravity theories

in ten and eleven dimensions”. Physics Letters B 273.1-2 (1991), pp. 90–94.
url: https://doi.org/10.1016/0370-2693(91)90558-8.

[HST83] P. S. Howe, K. S. Stelle, and P. K. Townsend. “The Relaxed Hypermultiplet:

An Unconstrained N = 2 Superfield Theory”. Nucl. Phys. B 214 (1983),
pp. 519–531. url: https://doi.org/10.1016/0550-3213(83)90249-3.

[HU87] P. S. Howe and A. Umerski. “Anomaly multiplets in six dimensions and ten
dimensions”. Phys. Lett. B 198 (1987), pp. 57–60. url: https://doi.org/
10.1016/0370-2693(87)90158-4.

[Hun99] C. Huneke. “Hyman Bass and ubiquity: Gorenstein rings”. Contemporary

Math. vol 243 (1999), pp. 55–78. arXiv: math/0209199.
Bibliography 275

[Huy06] D. Huybrechts. Fourier-Mukai transforms in algebraic geometry. Oxford

mathematical monographs. Oxford University Press, 2006.

[Joy07] D. D. Joyce. Riemannian holonomy groups and calibrated geometry. Vol. 12.
Oxford Graduate Texts in Mathematics. Oxford University Press, Oxford,
2007, pp. x+303.

[Jur+19] B. Jurčo, L. Raspollini, C. Sämann, and M. Wolf. “L∞ algebras of classical

field theories and the Batalin–Vilkovisky formalism”. Fortsch. Phys. 67.7
(2019), p. 1900025. arXiv: 1809.09899 [hep-th].

[Kac77] V. G. Kac. “Lie superalgebras”. Advances in Math. 26.1 (1977), pp. 8–96.

[KR01] V. G. Kac and A. Rudakov. “Representations of the Exceptional Lie Super-

algebra E(3,6) II: Four Series of Degenerate Modules”. Comm. Math. Phys.
222.3 (2001), pp. 611–661. arXiv: math-ph/0012050.

[Kaj07] H. Kajiura. “Noncommutative homotopy algebras associated with open strings”.

Reviews in Mathematical Physics 19.01 (2007), pp. 1–99. url: http://dx.
doi.org/10.1142/S0129055X07002912.

[Kap91] M. Kapranov. “On dg-modules over the de Rham complex and the vanishing
cycles functor” (1991), pp. 57–86.

[Kap21] M. Kapranov. “Supergeometry in mathematics and physics”. New spaces in

physics—formal and conceptual reflections. Cambridge Univ. Press, Cam-
bridge, 2021, pp. 114–152. arXiv: 1512.07042 [math.AG].

[KW07] A. Kapustin and E. Witten. “Electric-Magnetic Duality And The Geometric

Langlands Program”. Commun. Num. Theor. Phys. 1 (2007), pp. 1–236.
arXiv: hep-th/0604151.

[Kin+07] J. Kinney, J. M. Maldacena, S. Minwalla, and S. Raju. “An Index for 4

dimensional super conformal theories”. Commun. Math. Phys. 275 (2007),
pp. 209–254. arXiv: hep-th/0510251.

[KN64] S. Kobayashi and T. Nagano. “On filtered Lie algebras and geometric struc-
tures I”. Journal of Mathematics and Mechanics 13.5 (1964), pp. 875–907.

[Koc77] R. M. Koch. “On filtered Lie algebras”. Indiana University Mathematics

Journal 26.1 (1977), pp. 115–124. url: http://www.jstor.org/stable/
24891326.

[Kol13] J. Kollár. Singularities of the Minimal Model Program. Cambridge Tracts

in Mathematics. Cambridge University Press, 2013.

[KS96] Y. Kosmann-Schwazbach. “From Poisson algebras to Gerstenhaber alge-

bras”. Annales de l’Institut Fourier 46.5 (1996), pp. 1243–1274. url: http:
//eudml.org/doc/75211.
Bibliography Outlook

[Kos85] J.-L. Koszul. “Crochet de Schouten-Nijenhuis et cohomologie”. fr. Astérisque

S131 (1985). url: http://www.numdam.org/item/AST_1985__S131__257_
0/.

[KL09] D. Krotov and A. Losev. “Quantum field theory as effective BV theory

from Chern–Simons”. Nuclear Physics B 806.3 (2009), pp. 529–566. url:
http://dx.doi.org/10.1016/j.nuclphysb.2008.07.021.

[KNT17] S. M. Kuzenko, J. Novak, and S. Theisen. “New superconformal multi-

plets and higher derivative invariants in six dimensions”. Nucl. Phys. B 925
(2017), pp. 348–361. arXiv: 1707.04445 [hep-th].

[KNT18a] S. M. Kuzenko, J. Novak, and S. Theisen. “Non-conformal supercurrents in

six dimensions”. JHEP 02 (2018), p. 030. arXiv: 1709.09892 [hep-th].

[KNT18b] S. M. Kuzenko, J. Novak, and S. Theisen. “Non-conformal supercurrents in

six dimensions”. JHEP 02 (2018), p. 030. arXiv: 1709.09892 [hep-th].

[Kuz18] A. G. Kuznetsov. “On linear sections of the spinor tenfold. I”. Izv. Ross.
Akad. Nauk Ser. Mat. 82.4 (2018), pp. 53–114. url: https://doi.org/10.
4213/im8756.

[KM95] I. Kříž and J. P. May. “Operads, algebras, modules and motives”. Astérisque
233 (1995), iv+145pp.

[Lad04] T. Lada. “L∞ algebra representations”. Applied Categorical Structures 12.1

(2004), pp. 29–34.

[LM95] T. Lada and M. Markl. “Strongly homotopy Lie algebras”. Communications

in algebra 23.6 (1995), pp. 2147–2161.

[Lap01] S. V. Lapin. “Differential perturbations and D∞ -differential modules”. Sbornik

Mathematics 192 (2001).

[LTM12] W. D. Linch III and G. Tartaglino-Mazzucchelli. “Six-dimensional Super-

gravity and Projective Superfields”. JHEP 08 (2012), p. 075. arXiv: 1204.
4195 [hep-th].

[LV12a] J.-L. Loday and B. Vallette. Algebraic operads. Vol. 346. Grundlehren der
mathematischen Wissenschaften [Fundamental Principles of Mathematical
Sciences]. Springer, Heidelberg, 2012, pp. xxiv+634. url: https://doi.
org/10.1007/978-3-642-30362-3.

[LV12b] J.-L. Loday and B. Vallette. Algebraic operads. Vol. 346. Springer Science
& Business Media, 2012.

[Lur09] J. Lurie. “On the Classification of Topological Field Theories” (May 2009).
arXiv: 0905.0465 [math.CT].
Bibliography 277

[MSW19] T. Macrelli, C. Sämann, and M. Wolf. “Scattering amplitude recursion

relations in Batalin-Vilkovisky–quantizable theories”. Phys. Rev. D 100.4
(2019), p. 045017. arXiv: 1903.05713 [hep-th].

[Mal98] J. M. Maldacena. “The Large N limit of superconformal field theories and

supergravity”. Adv. Theor. Math. Phys. 2 (1998), pp. 231–252. arXiv: hep-
th/9711200.

[Man84] Y. I. Manin. “New directions in geometry”. Russian Mathematical Surveys

39.6 (1984), p. 51.

[Man85] Y. I. Manin. “Geometry of supergravity and Schubert supercells”. Journal

of Soviet Mathematics 31 (1985), pp. 3362–3373.

[Mar98] M. Markl. “Homotopy algebras via resolutions of operads”. Proceedings of

the 19th Winter School Geometry and Physics (1998). arXiv: 9808101 [math].

[Mne17] P. Mnev. “Lectures on Batalin-Vilkovisky formalism and its applications in

topological quantum field theory” (July 2017). arXiv: 1707.08096 [math-ph].

[MS04] M. Movshev and A. S. Schwarz. “On maximally supersymmetric Yang-Mills

theories”. Nucl. Phys. B 681 (2004), pp. 324–350. arXiv: hep-th/0311132.

[MS06] M. Movshev and A. S. Schwarz. “Algebraic structure of Yang-Mills theory”.

Prog. Math. 244 (2006), pp. 473–523. arXiv: hep-th/0404183.

[Mov11] M. V. Movshev. “Geometry of a desingularization of eleven-dimensional

gravitational spinors” (May 2011). arXiv: 1105.0127 [hep-th].

[Mov05a] M. Movshev. “On deformations of Yang–Mills algebras” (2005). arXiv: 0509119

[hep-th].

[Mov05b] M. Movshev. “Yang–Mills theories in dimensions 3, 4, 6, and 10 and bar

duality” (2005). arXiv: 0503165 [hep-th].

[MSX12] M. V. Movshev, A. Schwarz, and R. Xu. “Homology of Lie algebra of super-

symmetries and of super-Poincaré Lie algebra”. Nucl. Phys. B 854 (2012),
pp. 483–503. arXiv: 1106.0335 [hep-th].

[MSX14] M. V. Movshev, A. Schwarz, and R. Xu. “Integral invariants in flat super-

space”. Nucl. Phys. B 884 (2014), pp. 28–43. arXiv: 1403.1997 [hep-th].

[Mov15] M. V. Movshev. “Geometry of pure spinor formalism in superstring theory”.

Adv. Math. 268 (2015), pp. 201–240. url: https://doi.org/10.1016/j.
aim.2014.09.014.

[Nah78] W. Nahm. “Supersymmetries and their Representations”. Nucl. Phys. B135

(1978), p. 149.
Bibliography Outlook

[Nil86] B. E. W. Nilsson. “Pure spinors as auxiliary fields in the ten-dimensional

supersymmetric Yang–Mills theory”. Classical and Quantum Gravity 3.2
(1986), p. L41.
[OY19] J. Oh and J. Yagi. “Chiral algebras from Ω-deformation”. JHEP 08 (2019),
p. 143. arXiv: 1903.11123 [hep-th].
[OZ21] J. Oh and Y. Zhou. “A domain wall in twisted M-theory” (May 2021). arXiv:
2105.09537 [hep-th].
[OSS80] C. Okonek, M. Schneider, and H. Spindler. Vector Bundles on Complex
Projective Spaces. Springer New York, NY, 1980. url: https://doi.org/
10.1007/978-1-4757-1460-9.
[Pal14] J. Palmkvist. “The tensor hierarchy algebra”. J. Math. Phys. 55 (2014),
p. 011701. arXiv: 1305.0018 [hep-th].
[RSW23] S. Raghavendran, I. Saberi, and B. R. Williams. “Twisted Eleven-Dimensional
Supergravity”. Commun. Math. Phys. 402.2 (2023), pp. 1103–1166. arXiv:
2111.03049 [math-ph].
[RW22] S. Raghavendran and B. R. Williams. “A holographic approach to the six-
dimensional superconformal index” (Oct. 2022). arXiv: 2210.07910 [math-ph].
[RY19] S. Raghavendran and P. Yoo. “Twisted S-duality” (2019). arXiv: 1910 .
13653.
[RSV88] A. A. Rosly, A. S. Schwarz, and A. A. Voronov. “Geometry of supercon-
formal manifolds”. Communications in Mathematical Physics 119 (1988),
pp. 129–152.
[Roy99] D. Roytenberg. Courant algebroids, derived brackets, and even symplectic
supermanifolds. University of California, Berkeley, 1999.
[Rö06] C. Römelsberger. “Counting chiral primaries in N = 1, d=4 superconformal
field theories”. Nucl. Phys. B 747 (2006), pp. 329–353. arXiv: hep - th /
0510060.
[SW20] I. Saberi and B. R. Williams. “Twisted characters and holomorphic sym-
metries”. Lett. Math. Phys. 110.10 (2020), pp. 2779–2853. url: https :
//doi.org/10.1007/s11005-020-01319-4.
[SW21] I. Saberi and B. R. Williams. “Twisting pure spinor superfields, with appli-
cations to supergravity” (2021). arXiv: 2106.15639 [math-ph].
[SW23a] I. Saberi and B. R. Williams. to appear. 2023.
[SW23b] I. Saberi and B. R. Williams. “Constraints in the BV formalism: Six-dimensional
supersymmetry and its twists”. Adv. Math. 412 (2023), p. 108806. arXiv:
2009.07116 [math-ph].
Bibliography 279

[SW23c] I. Saberi and B. R. Williams. “Superconformal Algebras and Holomorphic

Field Theories”. Annales Henri Poincare 24.2 (2023), pp. 541–604. arXiv:
1910.04120 [math-ph].
[Ser55] J.-P. Serre. “Faisceaux algébriques cohérents”. Annals of Mathematics 61
(1955), p. 197.
[Sha99] E. Sharpe. “D-branes, derived categories, and Grothendieck groups”. Nu-
clear Physics B 561.3 (1999), pp. 433–450.
[Soh85] M. F. Sohnius. “Introducing supersymmetry”. Physics Reports 128.2-3 (1985),
pp. 39–204.
[Ste21] J. Stelzig. “On the structure of double complexes”. Journal of the Lon-
don Mathematical Society 104.2 (2021), pp. 956–988. arXiv: 1812 . 00865
[math.RT].
[Sva74] T. Svanes. “Coherent cohomology on schubert subschemes of flag schemes
and applications”. Advances in Mathematics 14.4 (1974), pp. 369–453. url:
https://www.sciencedirect.com/science/article/pii/0001870874900395.
[Tan70] N. Tanaka. “On differential systems, graded Lie algebras, and pseudo-groups”.
Journal of Mathematics of Kyoto University 10.1 (1970), pp. 1–82.
[Toë14] B. Toën. “Derived algebraic geometry”. EMS Surv. Math. Sci. 1.2 (2014),
pp. 153–240. url: https://doi.org/10.4171/EMSS/4.
[Vak] R. Vakil. Foundations of Algebraic Geometry. url: http://math.stanford.
edu/~vakil/216blog/FOAGnov1817public.pdf.
[Val14] B. Vallette. “Algebra + homotopy = operad”. Symplectic, Poisson, and non-
commutative geometry 62 (2014), pp. 229–290.
[LCL] M. A. A. van Leeuwen, A. M. Cohen, and B. Lisse. LiE, A package for Lie
group computations. Available at http://www-math.univ-poitiers.fr/
~maavl/LiE/.
[VA72] D. V. Volkov and V. P. Akulov. “Possible universal neutrino interaction”.
JETP Lett. 16 (1972), p. 367.
[Vor05] T. Voronov. “Higher derived brackets and homotopy algebras”. Journal of
Pure and Applied Algebra 202.1-3 (2005), pp. 133–153. arXiv: math/0304038
[math.QA].
[Wan91] M. Y. Wang. “Preserving parallel spinors under metric deformations”. Indi-
ana Univ. Math. J. 40.3 (1991), pp. 815–844. url: https://doi.org/10.
1512/iumj.1991.40.40037.
[Wit88] E. Witten. “Topological Quantum Field Theory”. Commun. Math. Phys.
117 (1988), p. 353. url: https://doi.org/10.1007/BF01223371.
Bibliography Outlook

[Wit95] E. Witten. “String theory dynamics in various dimensions”. Nucl. Phys. B

443 (1995), pp. 85–126. arXiv: hep-th/9503124.

[Wit19] E. Witten. “Notes On Super Riemann Surfaces And Their Moduli”. Pure
Appl. Math. Quart. 15.1 (2019), pp. 57–211. arXiv: 1209.2459 [hep-th].

[Yag12] J. Yagi. “Compactification on the \Omega-background and the AGT corre-

spondence”. JHEP 09 (2012), p. 101. arXiv: 1205.6820 [hep-th].

[Zel09] I. Zelenko. “On Tanaka’s prolongation procedure for filtered structures of

constant type”. SIGMA 5 (2009), p. 094. url: https://doi.org/10.3842%
2Fsigma.2009.094.