Pictures from super Chern-Simons theory

We study super-Chern-Simons theory on a generic supermanifold. After a self-contained review of integration on supermanifolds, the complexes of forms (superforms, pseudoforms and integral forms) and the extended Cartan calculus are discussed. We then introduce Picture Changing Operators and their mathematical properties. We show that the free equations of motion reduce to the usual Chern-Simons equations proving on-shell equivalence between the formulations at different pictures of the same theory. Finally, we discuss the interaction terms. They require a suitable definition in order to take into account the picture number. This leads to the construction of a series of non-associative products which yield an A∞ algebra structure, sharing several similarities with the super string field theory action by Erler, Konopka and Sachs.


JHEP03(2020)043 1 Introduction
Our main motivation is to provide a general method for constructing classical actions for quantum field theories on supermanifolds using the powerful methods of supergeometry. As is known for general relativity, using differential forms one constructs physical interesting quantities (actions, observables, globally defined quantities). Here we would like to set up an equivalent framework, but over a supermanifold. Namely we would like to formulate quantum field theory models on supermanifolds as we use to do it in general relativity.
Given a supermanifold SM (n|m) with n bosonic dimensions and m fermionic dimensions, we would like to construct an action of the form where L (n|m) is an integral form [1][2][3][4][5] with form degree n and picture number m. L (n|m) is a form which can be integrated over the supermanifold, i.e. it is a top form and any super-diffeomorphism leaves the action invariant. One strategy to build an action L (n|m) is to start from a conventional superform L (n|0) in terms of the classical fields and their differentials and then "complete" it to an integral form as where Y (0|m) is a Picture Changing Operator (PCO) mapping the superform L (n|0) to an integral form L (n|m) . If L (n|0) is closed, one can change Y (0|m) by exact pieces without changing the action S. The question is: can one always get a factorized form or are there other possibilities? We certainly know that there are special supermanifolds which are non-projected or non-split, for those the existence of a global PCO is not granted, and therefore we are looking for a more general description. Namely, given the fields in a given picture, is there a way to build an action consistently producing meaningful results? For example, given a gauge field A (1|0) which is the usual connection 1-form at picture equal to zero, can one use a picture one field as A (1|1) instead? Then, we would replace the Lagrangian (1.2) as L (n|0) (A (1|0) ) ∧ Y (0|m) −→ L (n|m) (A (1|1) ) , (1.3) such that the equations of motion are still dynamical equations.
A similar issue appears in string theory [6] and string field theory [7], where the ghost sector of RNS string theory model requires a choice of the vacuum due to the replicas of the same Hilbert space at different pictures. As is well known, the quantization of the β −γ ghost sector leads to a Fock space filtered according to the ghost number and with respect to the picture number. This translates into the definition of the vertex operators representing the target space fields. Those vertex operators can be chosen in different pictures such that the total sum of pictures of the vertex operators inserted into a correlation function saturates the required picture charge (see also [8][9][10]) at given genus and number of punctures. The result should be independent of the choice of the picture. In the case of string field theory, the situation is slightly different. In order to write a string field theory action, JHEP03(2020)043 one needs to take into account the saturation of the picture on a disk (tree level classical action) and for that some alternatives were proposed (see [11] and [7]). However, despite some interestring features for these models, they fail to give a complete interacting superstring field theory action. Only recently, by the work of [12], a complete interacting superstring field theory action has been proposed. The string fields are taken into a picture one Hilbert space leading to a meaningful kinetic term. Nonetheless, the interactions are constructed in terms of a non-associative product which multiplies two string fields without increasing the picture and which is the first non-trivial element of an A ∞ algebra. That algebra has been built completely [12,13]. As shown in [14,15], for any supermanifold, in terms of the PCO built in the complexes of forms, one can define a corresponding A ∞ -algebra [16][17][18][19][20] on the geometrical data and therefore we expect that we can follow the same pattern.
In the same way, for the construction of quantum field theories on supermanifolds (we recall that the picture in string theory is related to the superghosts zero modes which are in relations with the supermoduli space of the underlying super-Riemann surface), one needs to fix the total picture of the action, but that does not select a given picture for the fields involved. This means that one can choose different set of fields, defined as forms in the full complex, and construct the corresponding action (See [2,3]).
To provide an illustration of this constructing procedure, we focus on a specific model, namely super Chern-Simons theory on a (3|2) supermanifold. The classical action can be written in terms of the (1|0) connection A (1|0) . It is shown that by using the factorized form L (3|0) ∧ Y (0|2) all superspace formulations can be obtained. The choice of the PCO Y (0|2) ranging from the simplest example to more symmetric expressions (see [1]) leads to different actions with manifest supersymmetry or in components.
In the present work, we consider an action for super Chern-Simons theory (henceforth SCS) built in terms of the A (1|1) gauge fields, namely those at picture one. Their expansions in term of component fields are infinite dimensional, then the kinetic term is obtained by using repeated distributional properties and integrating over the supermanifold. The goal is to verify that the kinetic term yields the correct equations of motion, namely Chern-Simons flat connection and vanishing fermions. This is achieved by showing that, on-shell, all unphysical components can be gauged away except the Chern-Simons connection and its flatness condition. Here we consider those gauge transformations which are obtained by derivatives along the fermionic directions. Since the fermionic fibers do not have any topology, those gauge transformations can be reabsorbed (without any topological obstractions) leaving only usual gauge transformations along bosonic coordinates.
Once we have verified that the free action leads to the correct equations of motion, we consider the interaction terms. Thus, following the string field theory construction [12], we define a 2-product which multiplies two (1|1) gauge fields and decreases the picture by one [14]. This leads to the usual interaction term plus additional interactions due to the infinite components of the picture one connections. The product used for the interaction is non-associative, but its associator is cohomologically trivial and therefore can be compensated by a 3-product. This leads to an A ∞ algebra which consistently provides the complete set of equations of motion. Together with the cyclicity of the inner product, we finally derive the equations of motion from a consistent variational principle.

JHEP03(2020)043
At the end, we discuss the supersymmetry in this framework. We found that even in the picture one setting for the gauge fields, the rheonomic conditions do show the same set of physical fields. In addition, since the construction is based on supermanifolds, the super-diffeomorphisms are those transformations which preserve the entire structure.
The main motivation of this work relies on the search for a general formulation for quantum field theories on supermanifold. The richness of the geometry of forms over supermanifolds -involving differential, pseudo and integral forms at once -allows for exploring the possibility of new theories in the present framework. For that, we abandon the factorized form unlashing the full power of pseudoform geometry. However, this very first example needs to be rather conservative, namely, we would like to verify whether our mathematical tools are suitable to consider more general actions, but yet to be physically sound. For that reason, we use super Chern-Simons theory: it is significantly easy to treat as a topological model, sharing several analogies with string field theory and supersymmetrizable in several ways. The non-factorized form of the action can be guessed from string field theory and that provides a good guideline for our analysis. Still, the full comprehension of the factorized action (see [1]) gives us an important reference point to be compared with. Nonetheless, our aim is to explore this new framework for new theories which are not so simple, such as super-Yang-Mills and full-fledged supergravity.
We take our steps from the example of string field theory in the formulation given in [12], but one might wonder whether the two construction share other common features. In particular, as has been underlined in [12], the construction of the modified two product (M 2 ) is due to emerging divergences when two Picture Changing Operators collide (in the OPE framework). Here, a similar phenomenon takes place. This is illustrated in appendix C, with a simple four point function computation. Two types of divergences arise due to the action of Heaviside theta function (appearing in the PCO) on picture zero forms and on inverse forms. The complete analysis will be published elsewhere [21]. The divergences are visible using the integral representation of distributional operators.
The paper is organized as follows: in section 2, we collect all mathematical tools needed for the construction of the action and the derivation of the equations of motion. In particular, we describe the action of various operators on the space of forms for the supermanifold SM (3|2) . In section 3, we recall the action for super Chern-Simons theory in the factorized form. In section 4, we get to the main derivation of the equations of motion in the non-factorized form, interaction terms, gauge invariance and supersymmetry. In appendices, we collect some review material on A ∞ algebras and their automorphisms and some explicit computations omitted in main text.

Supermanifolds and superspaces
Let us briefly recall the most basic definitions in supergeometry. For a rigorous and thorough mathematical treatment of the subject we suggest the reader to refer to [22][23][24][25]. The most basic and most important example of superspace is given by R (p|q) , that is the pair JHEP03(2020)043 given by the manifold R p and the sheaf C ∞ R p θ 1 , . . . , θ q This means that the superspace is constructed over the usual space R p and the functions we consider are C ∞ functions over the coordinates of the space R p and they have polynomial dependence on the Grassmann coordinates θ i q i=1 . Since the θ coordinates are anticommuting, this is equivalent to consider the exterior algebra generated by q variables with value into C ∞ functions: This means that a general function can be expanded as where the Einstein's summation convention is understood.
where |SM| is a real manifold and O SM is a sheaf which is locally isomorphic to C ∞ θ 1 , . . . , θ q . In other words a real supermanifold of dimension (p|q) is a superspace which is locally isomorphic to open subsets of R (p|q) . In this paper we will only deal with real supermanifolds: in particular this means that we will not be concerned by the subtleties related to non-projected and non-split supermanifolds which arise only in the context of complex supermanifolds [26][27][28][29][30][31][32].
We consider the case of a real supermanifold SM (3|2) ; in terms of the coordinates, we define the following differential operators where the second and the third are known as superderivative and supersymmetry generator, respectively. They satisfy the superalgebra relations In 3d, for the local subspace we use the Lorentzian metric η ab = (−, +, +), and the real and symmetric Dirac matrices γ a αβ given by Numerically, we haveγ αβ a = γ a αβ andγ αβ a = η ab (Cγ b C) αβ = C αγ γ a,γδ C δβ . The conjugation matrix is αβ and a bi-spinor is decomposed as follows R αβ = R αβ + R a γ a αβ where R = − 1 2 αβ R αβ and R a = Tr(γ a R) are a scalar and a vector, respectively. In addition, it is easy to show that γ ab αβ ≡ 1 2 [γ a , γ b ] αβ = abc γ cαβ .

JHEP03(2020)043
Given a form Φ, its differential is expanded on a basis of forms as follows where V a = dx a + θγ a dθ and ψ α = dθ α which satisfy the Maurer-Cartan equations We can compute the supersymmetry variation as a Lie derivative L with = α Q α + a ∂ a ( a are the infinitesimal parameters of the translations and α are the supersymmetry parameters) and by means of the Cartan formula we have where the term dι Φ is simply a gauge transformation. It follows easily that δ V a = δ ψ α = 0 and δ dΦ = dδ Φ.

Superforms, integral forms and pseudoforms
As seen in [4,5,23,[33][34][35][36][37], the space of differential forms has to be extended in order to define a meaningful integration theory. We define Ω (•|•) (SM) as the complete complex of forms; they are graded w.r.t. two gradings as where q = 0, . . . , m, p ≤ n if q = m, p ≥ 0 if q = 0 and p ∈ Z if q = 0, m. The usual wedge product for form multiplication is defined as where 0 ≤ p, q ≤ n and 0 ≤ r, s ≤ m with (n|m) are the bosonic and fermonic dimensions of the supermanifold SM. 1 From this point on, we will omit the wedge product symbol to avoid cumbersome expressions. Locally, a (p|r)-form ω formally reads where g(x) denotes the differentiation degree of the Dirac delta function corresponding to the 1-form dθ x . The three indices l, h and r satisfy the relation . where the last equation means that each α l in the above summation should be different from any β k , otherwise the degree of the differentiation of the Dirac delta function can be reduced and the corresponding 1-form dθ α k is removed from the basis. The components Due to the anticommuting properties of the δ forms, this product is by definition equal to zero if the forms to be multiplied contain δ localized in the same variables dθ, since the δ's have to be considered as de Rham currents [32]. In figure 1, we display the complete complex of forms. We refer to the first line as the complex of superforms, to the last line as the complex of integral forms and to the middle lines as the complex of pseudoforms. We notice that the first line and the last line are bounded from below and from above, respectively. This is due to the fact that in the first line, being absent any delta functions, the form number cannot be negative, and in the last line, having saturated the number of delta functions we cannot admit any power of dθ (because of the distributional law dθδ(dθ) = 0). In our case, we have n = 3 and m = 2, hence the complex has three lines.
Let us consider the space Ω (1|1) in the middle complex, spanned (in the sense of formal series) by the following psuedo-forms where the number n is not fixed and it must be a non-negative integer. For example, consider the pseudoform spanned by the second element in (2.16) with n = 1: we have the implicit summation for the indices α and β, thus (2.17) becomes for any test function φ, which extends to the same rule for the 1-form dθ: Hence we get (2.21) Notice that the first and the last terms are elements that can be spanned by dx a (dθ α ) n δ (n) (dθ β ) for n = 0; this means that by a redefinition of the fields A (p) aαβ we can assume w.l.o.g. that α = β in the implicit sums. This reflects the property that elements spanned by (dθ α ) n+1 δ (n) dθ β are exactly equal to 0 if α = β, ∀n ≥ 0.
Due to 1-forms dx a and due to the fact that we are free to set α = β, the number of generators (monomial forms) at a given n is (8|8), but the total number of monomial generators in Ω (1|1) is infinite.

Integration
Once the integral forms are defined, we have to clarify how the integration is performed. For that we consider an integral form given by
The final expression contains a usual Riemann/Lebesgue integral on x's and the Berezin integral over θ's. The symbol [d 3 xd 2 θ] is only a reminder on which variables the integral has to be performed. For example, in the case of SM (3|2) = R (3|2) we have where the Berezin integration has been performed and we are left with the Riemann/ Lebesgue integral. We define a product (Serre's duality) between Ω (p|q) and Ω (r|s) forms as which is non-vanishing only if p + q = 3 and r + s = 2. Under these conditions, the spaces Ω (p|r) and Ω (q|s) are isomorphic and therefore there is a (super)form in Ω (p|0) corresponding to an integral form in Ω (3−p|2) . By partially computing the form integral (leaving undone only the Berezin integral over the coordinates θ and the Riemann/Lebesgue integral over x), we have where ω J (x, θ) are the coefficients (the index J stands for the collection of indices needed to define the form) of the form ω (p|r) , while ω J are the coefficients of the dual forms in Ω (3−p|2) . For the space R (3|2) , if we consider for example the spaces Ω (1|0) and Ω (2|2) we have: and Then, we can compute ω (1|0) , ω (2|2) as Notice that the product is a pairing and it does not need to be positive definite.

JHEP03(2020)043
If we use the same technique for Ω (1|1) and Ω (2|1) , we have to recall that the dimension of these spaces is infinite and therefore the sum over J must be substituted with formal series. In the same way as described in the previous subsection, for a general supermanifold SM (3|2) any form belonging to the middle complex Ω (p|1) is decomposed into an infinite number of components as in (2.16).

Covariance on Ω (p|r)
In this subsection, we clarify how the Lorentz symmetry is implemented in the space of pseudoforms. This is a crucial point in order to understand how the covariance is recovered at any picture number.
If we consider the complex of integral forms Ω (p|2) , and we perform an infinitesimal Lorentz transformation. We have to use the distributional relation to check that the product of δ(dθ 1 )δ(dθ 2 ) transforms as an inverse of a density (we avoid the absolute value of the determinant since we are keeping track of the orientation of the integration) and therefore, each monomial of the complex Ω (p|2) transforms according to a tensorial representation and the inverse of the determinant of a Lorentz transformation in the spinor representation (sections of the Berezinian bundle). This confirms the fact that the top form d 3 xδ 2 (dθ) is indeed invariant under Lorentz transformations. In addition, when the derivatives of the product δ(dθ 1 )δ(dθ 2 ) are taken into account, for example as in the Ω (−2|2) form the components ω αβ (x, θ) transform as in a linear tensor representation of the spinorial representation. This means that the spinorial indices in (2.35) are covariantly contracted. Therefore, for both the superforms Ω (p|0) and the integral forms Ω (p|2) , the usual covariant techniques can be used.
Let us now consider the infinite dimensional complex Ω (p|1) . As seen above, it is unbounded from above and from below and each space Ω (p|1) is (double)-infinite dimensional. Double means that we have two ways to construct a pseudo form, along δ(dθ 1 ) and along δ(dθ 2 ). However, under any transformation which mixes θ 1 with θ 2 (for example Lorentz transformations) the two directions indeed mix and the following situation arises.

JHEP03(2020)043
If we consider a single Dirac delta function δ(dθ 1 ), we cannot use the distributional identity (2.34), but we observe that, infinitesimally, where δ (1) (dθ α ) is the first derivative of δ(dθ α ) and we have neglected higher order terms. The first and the second terms come from the Taylor expansion of the Delta distribution, with dθ 1 and dθ 2 respectively. This fact implies that in order to implement the Lorentz symmetry in the space of pseudoforms Ω (p|1) , one necessarily needs an infinite dimensional space. Indeed, for a finite Lorentz transformation one needs all components in the n expansion of a generic pseudoform in Ω (p|1) . For example, let us consider a (0|1)-pseudoform, it can be written as where we collected the pieces with different powers of dx's (we use a little abuse of notation by omitting the dx s and writing as superscripts only the fermionic form number and the picture number). Since the first term ω (0|1) 0 does not contains powers of dx, it can be written as where the coefficients ω 21 (x, θ) are superfields. Since we have distinguished the terms with dθ 1 and dθ 2 , the covariance of the expression is not manifest. Indeed, it might be better to write (2.38) as where the indices α and β are summed, as conventionally. Notice that if α = β, we have (dθ α ) n multiplying δ (n) (dθ α ) and, by using the distributional property (dθ α ) n δ (n) (dθ α ) = (−1) n n!δ(dθ α ), the coefficient ω (n), α α (x, θ) is reabsorbed into a redefinition of ω (0) (x, θ) which multiplies δ(dθ α ).
If we perform an infinitesimal Lorentz transformation w ab , we have that

JHEP03(2020)043
where the coefficients ω (n), β α (x, θ) are suitably redefined using Then the coefficients ω (n), β α are shifted as This holds at the infinitesimal level, but for a finite transformation all the coefficients ω (n), α β are involved. Therefore, the covariance of the expressions is maintained only if the complete series is taken into account. For other pieces ω (−p|1) p with p = 1, 2, 3, we notice that the dependence upon dx a is polynomial and therefore they transform linearly as always, but in addition there is a complete reshuffling of the coefficients of the series. In the next sections we will adopt the notation of writing the Greek indices of the components fields of forms both below.

Geometric picture changing operators: some explicit results
Having clarified the form complexes and having outlined how usual differential operators of Cartan calculus (d, ι X , L X ) work on superspace, we point out that we can build a new set of differential operators 2 acting on general forms such as δ [5,14,15]. These operators are used to change the picture number of a given form (and eventually its form number as well) and are usually referred to as Picture Changing Operators (PCO's). The specific form of those operators is suggested by String Theory analogy [6,8,9] and their geometric interpretation [34]. In the present section we provide some results that will be used in the rest of the paper.
The first PCO we define is Y: given a (p|q)-form ω (p|q) ∈ Ω (p|q) , we define the Picture Raising Operator Y (0|s) as a multiplicative operator s.t.

JHEP03(2020)043
Since it is a multiplicative operator that raises the picture number by s, it follows that locally Again, given a (p|q)-form ω (p|q) ∈ Ω (p|q) , we define the Picture Lowering Operator Z D as where [·, ·] denotes as usual a graded commutator and the action of the operator Θ(ι v ) is defined by the Fourier-like relation of the Heaviside step function (2.45) where we have used the fact that e itιv is a translation operator. Hence the operator Θ i.e. it lowers either the form degree or the picture degree. As we will see in the following examples this operator does not give a pseudoform as a result, but rather an inverse f orm, i.e. an expression containing negative powers of dθ. We remark, as was discussed in [14], that the introduction of inverse form requires the definition of a new complex Ω corresponding to the Large Hilbert Space (LHS) used in string theory. In the following, we will denote simply by Ω (•|•) the space suitably enlarged. The relation between Large Hilbert Space and Small Hilbert Space (SHS) was clarified in [14] in the case of a single fermionic variable.
Here we list some examples, not only in order to explain how to manipulate the Θ(ι v ) operator, but also in order to prepare some results that will be used in the next sections. In particular we have opted to highlight some of the following results to stress their particular significance and because they will be directly employed.
We can also obtain the previous result in a slightly different way: where in the second passage we have performed the substitution y = dθ α + iv α t. 3

JHEP03(2020)043
Example 2. We have the following result: The result already stated follows after a direct calculation: In order to get more general formulas we consider other simple examples.
Example 4. We have the following result: (2.50) Again, the result follows from direct computation: where we have made use of the assumption q ≥ p when expanding the binomial: the term with highest power of y behaves like y p , but since it is multiplied by a y −1 term, we have a global y p−1 which is annihilated by d q dy q if q > p − 1, i.e. q ≥ p. The same happens for all the other terms of the expansion except for the (dθ α ) p term which is multiplied by y − dθ β −1 and does not give a trivial result after derivation.
The following examples are studied because they are explicitly needed in the following section.

JHEP03(2020)043
where the result arises from a straightforward calculation.
where again the result follows from direct calculation.
As a final example we evaluate the application of Θ to δ(dθ α )δ(dθ β ): Observe that δ(v · dθ) allows us to rewrite the result in two other equivalent ways: Starting from the operator Θ (ι v ) we directly define the PCO δ (ι v ) as the formal derivative w.r.t. the argument of Θ: such that it acts on a general (p|q)-form by using the Fourier representation We define now the operator η as the geometric partner of η of String Theory [38] in terms of its action on forms: given a (p|q)-form ω (p|q) ∈ Ω (p|q) , we define the operator η as where the action of e i ιv is defined as a translation operator acting on generalised functions of dθ and Π is the parity changing functor which allows us to convert bosonic/fermionic quantities into fermionic/bosonic ones. 4,5 Let us consider a few examples in order to understand better the action of this operator.
Example 8. Let us now consider the action of η on a generic fermionic p-form with picture number 0: thanks to the limit.
Example 9. Let us now consider the action of η on a Dirac delta form: since being f a test function, it is certainly continuous in 0. This result is then extended for x ≡ dθ.
Example 10. We have that the η operator acting on a general pseudoform with picture number 1 gives 0: The result follows after a direct calculation in the distributional sense, i.e. where it is involved the application to a generic C ∞ test function.
Example 11. Let us now consider the action of η on a (−1|0)-inverse form: where we have used the usual distributional formula (2.65) 4 Notice that in String Theory [38] there exists only one operator η associated to the zero-mode of the ηz field emerging from bosonisation of β, γ system. In our case we define an ηv operator for each fermionic direction. 5 The limit in (2.59) has to be intended as a distributional limit.

JHEP03(2020)043
Example 12. Let us now consider the action of η on a general inverse form with picture number 0: where in the last passage we have left only the linear terms in v since they are the only ones contributing.
RMK 1. The operator η is, modulo the multiplicative constant i, the left inverse of the operator Θ acting on pseudoforms, i.e.
We can apply η to the definition of the operator Θ (ι v ) (2.45) in order to find and by passing the limit under the integral sign we get Example 13. We have the following results for general inverse forms of picture degree 0 and 1: The proof of (2.71) is a direct consequence of example 8, having used the result of example 12, i.e. the operator η passes through the numerator without any contribution.
The proof of (2.72) is again a direct consequence of example 9, having used the result of example 12.

JHEP03(2020)043
RMK 2. The operator Θ is, modulo the multiplicative constant i, the left inverse of the operator η on inverse forms of picture degree 0 and negative form degree, i.e.
The proof is a direct consequence of the previous example and of (2.50): By using the results from the previous propositions we want now to investigate the commutation relation between the operator η and the operator Z v . Before doing this, let us study the commutation relation between the operator η and the exterior derivative d: RMK 3. The operator η and the operator d anticommute: The proof follows after direct calculation on different types of inverse forms and pseudoforms.
RMK 4. The successive application of the operators η and Z v gives 0: The proof is simply an application of the definition of Z D and of the results (2.67), (2.73) and (2.75):

Super Chern-Simons actions (SCS)
In the present section, we review some of the ingredients needed for the construction of the main body of the paper. We first review D=3 N=1 super Chern-Simons theory in its classical derivation [39]. We start from the superspace construction, but we provide also the component action. Then, we reformulate the theory using the geometrical methods discussed in the previous section and we give the rules for a Chern-Simons theory on any supermanifold. We show that it leads to a very complicate non-factorized form, to be the basis for a theory on any supermanifold.

SCS in components and in superspace
We start from a (1|0)-superform A (1|0) = A a V a + A α ψ α , (where the superfields A a (x, θ) and A α (x, θ) take values in the adjoint representation of the gauge group) and we define the field strength where In order to reduce the redundancy of degrees of freedom of A a and A α of the (1|0)-form A (1|0) , one imposes (by hand) the conventional constraint and by expanding the superfields A a , A α and W α in components we have where a a (x) is the gauge field, λ α (x) is the gaugino and In terms of those fields, the super-Chern-Simons Lagrangian becomes (we denote by [d 2 θ] the Berezin integral over the θ's variables), which in component reads That coincides with the bosonic Chern-Simons action with free non-propagating fermions. JHEP03(2020)043

SCS on supermanifold
In order to obtain the same action by integration on supermanifolds we consider the rheonomic action and the corresponding action principle [40][41][42]. It requires the choice of a bosonic submanifold M (3) immersed into a supermanifold SM (3|2) and a (3|0)-form on it Here the choice of the (3|0)-form L (3) is a three-form Lagrangian constructed with the superform A, and its derivatives dA, without using the Hodge dual operator (that is, without any reference to a metric on the supermanifold SM (3|2) ). The action S rheo The rules to build the action (3.8) are listed and discussed in [40][41][42] in detail. An important ingredient is the fact that for the action to be supersymmetric invariant, the Lagrangian must be invariant up to a d-exact term and, in addition, if the algebra of supersymmetry closes off-shell (either because there is no need of auxiliary fields or because it exists a formulation with auxiliary fields), the Lagrangian must be closed: upon using the rheonomic parametrization. This amounts to set F αβ = 0, which is an algebraic equation and it can be directly used in the action. One of the rules of the geometrical construction for supersymmetric theories given in [40][41][42] is that by setting to zero the coordinates θ α and its differential ψ α = dθ α , the Lagrangian reduces to the component Lagrangian invariant under supersymmetry (up to a total derivative). Furthermore, the equations of motion in the full-fledged superspace implies the rheonomic constraints (which coincide with the conventional constraints of superspace formalism). In order to express the action (3.8) in a more geometrical way by including the dependence upon the embedding into the integrand, we use the Poincaré dual form (already named PCO) Y (0|2) dual to the immersion of M (3) into SM (3|2) . The Poincaré dual form Y (0|2) is closed, it is not exact and any of its variation is d-exact. The action can now be written on the full supermanifold as (3.11)

JHEP03(2020)043
Therefore, by choosing the PCO Y (0|2) = θ 2 δ 2 (dθ), its factor θ 2 projects the Lagrangian L (3|0) (A, dA) to L (3) (A, dA) θ=0 while the factor δ 2 (dθ) projects the latter to . Any variation of the embedding yields δY (0|2) = dΛ (−1|2) and leaves the action invariant if the Lagragian is closed. The rheonomic Lagrangian L (3|0) (A, dA) reads 6 Again, by choosing the PCO Y (0|2) = θ 2 δ 2 (dθ) we get the component action (3.7) and the third term in the action is fundamental to get the mass term for the non-dynamical fermions. This is the most general action and the closure of L (3|0) implies that any gauge invariant and supersymmetric action can be built by choosing a PCO Y (0|2) inside the same cohomology class. Therefore, starting from the rheonomic action, one can choose a different "gauge" -or better said a different immersion of the submanifold M (3) inside the supermanifold SM (3|2) -leading to different forms of the action with the same physical content. It should be stressed, however, that the choice of Y (0|2) new (defined in the following subsection), is a preferred "gauge" choice, which allows us to derive the conventional constraint by varying the action without using the rheonomic parametrization.

SCS in superspace revised
The choice of the PCO could be done observing that there are representatives respecting some isometries. For example the new operator is manifestly supersymmetric. Computing the expression in the integral (3.11), we see that Y new picks up al least two powers of ψ's and one power of V a and that forces us to expand L (3|0) as 3-form selecting the monomial ψγ a ψV a dual to Y (0|2) new . That finally gives the supersymmetric action described in (3.6).
The equations of motion derived from the new action are 14) The equations of motion correctly imply F (αβ) = 0 (which is the conventional constraint) and W α = 0 which are the super-Chern-Simons equation of motions. The second condition follows from F αβ = 0 and by the Bianchi identities which implies that F aα = γ aαβ W β . Notice that this formulation allows us to get the conventional constraint as an equation of motion. In particular we find that the equations of motion, together with the Bianchi JHEP03(2020)043 identity, imply the vanishing of the full field-strength.

SCS with semi-supersymmetric PCO
The choice of the PCO implies the form of the action and we present here another possibility. We consider the following expression It is closed because of δ 2 (ψ) and by using gamma matrices algebra. The presence of the explicit θ implies that it is not manifestly supersymmetric, but its variation is d-exact It is easy to show that this PCO is also not exact. Before computing the action, we discuss some other aspect of the geometry of the PCO (3.17). Consider the expression Since the right hand side is closed (since it is a top integral form) and since Y half is also closed, ω (3|0) has to be closed or its variation is the kernel of Y (0|2) half . Let us verify the first possibility. Computing the variation of ω (3|0) , we have which does not vanish. Nevertheless, we can add two new terms and get The additional terms are needed to make ω (3|0) closed, but it does not affect the relation (3.19) because of the powers of θ's and the powers of V 's.

SCS with pseudoforms
We consider now a new PCO. This is motivated by string theory, but we do not discuss here its origin, since it can be also described in terms of the supermanifold structure. The fermionic space spanned by the coordinates θ α can be decomposed in terms of two commuting spinors v α and w α with the property that v α αβ w β ≡ det(v, w) ≡ v·w = 1 where (v, w)

JHEP03(2020)043
is the 2×2 matrix built with the spinors. Notice that any spinor θ α can be decomposed on that basis θ α = −v α (w·θ)+w α (v·θ). Notice also that θ α αβ θ β = 2(v·θ)(w·θ). Any PCO Y (0|2) can be decomposed into the product of two PCO's Y (0|1) as follows The piece Ω is a (−1|2) form which depends on v and w. The two PCO's are equivalent in the sense that they belong to the same cohomology class and they increase the picture by one unity. One can check by direct inspection that the product of the two PCO's inserted in the action does not lead to the conventional constraint F αβ = 0 and therefore the exact term in (3.22) relating the two actions is important to get the full-fledged action principle.
Let us analyse the action with the new choice of PCO: where the Ω-term is dropped. Let us put aside the interaction term for the momentinteraction terms will be discussed in the forthcoming sections -and let us distribute the two Y's on the two pieces of the action as follows Since the PCO's are closed, we can also bring them after each connection term A (1|0) and after the spinorial W (0|0) forms as converting the gauge connection to a (1|1) form as (3.26) In the same way, the (0|0)-form W α is converted into a (0|1)-pseudoform. Notice that, if we change the basis by linearly composing v and w, we can write the same formula, up to an unessential overall factor. We can now forget that the connection pseudoform A (1|1) was originated by the factorised form (3.26), and consider an action built starting from a general (1|1)-pseudoform, the same applies for the pseudoform W (0|1) . Finally and most importantly, passing from A (1|0) , which has a finite number of components, to A (1|1) , which has an infinite number of them, we have made an important assumption: we have moved to an infinite dimensional space. Therefore, we conclude that we have to take into account the generic action The wedge product is taken in the space of pseudoforms, therefore we have used the convention that two (0|1)-forms must be multiplied with the wedge product.

JHEP03(2020)043
In this way, we have succeeded to find the natural geometrical formulation of super Chern-Simons theory on a three-dimensional supermanifold. According to this observation, we notice that we have only explored the case N = 1. This means two θ's and therefore two PCO's for the factorization Y (0|2) new = Y v Y w + dΩ. However, the same conclusion can be achieved in any supermanifold with an even number of θ's and, if the even dimension of the fermionic sector is m = 2r, we have the action where A (1|r) is a (1|r)-pseudoform and W (0|r),α a (0|r)-pseudoform. One can wonder whether the Lagrangian is closed. For that, we need the rheonomic conditions and the observation that they are left unchanged by multiplying them by the factor Y w . We remark that in two previous works of one of the author, see [44,45], a non-factorized form of the action has been used. That has led, in the first case to new results and in the second case to a complete D=3 N=1 supergravity action in all possible superspace representations.

General super Chern-Simons theory
The following section is the main core of the present work. We demonstrate that the non-factorized action (3.27) yields the same non-interacting equations of motion of super-Chern-Simons theory. We first write the explicit action by expanding the A (1|1) form in components, and, by integrating according to the previous discussion on the various variables, we arrive at the action principle from which we compute the equations of motion. By an iterative procedure we verify that the physical content of these equations is exactly the same as free Chern-Simons theory. Then we introduce the interactions. Because of the picture number, the interaction term has to be introduced in a non-trivial way leading to a non-associative product. This product will be the starting point of an A ∞ -algebra generated by the gauge-invariance requirement. Finally we discuss the closure of the gauge algebra and the supersymmetric properties of the obtained Lagrangian.

The Lagrangian
Let us start from the pseudoform A (1|1) = A 0 + A 1 + A 2 + A 3 , where the subscript denotes the number of dx's in the expression. We then have the decomposition:

JHEP03(2020)043
Let us clarify the notation: we have to fix a convention for the field A 1|1 , i.e. we want to decide whether it is an even or odd field. However we are not free to choose a convention for the following reason: the field A and the field dA have opposite parity, since the exterior derivative d is an odd operator, i.e. they commute AdA = dAA. This implies that In the even case we get that the Lagrangian AdA is trivial, since it is equal to 1 2 d(AA), while in the odd case we don't get a trivial Lagrangian, as it happens in usual Chern-Simons theory; therefore we must chose A (1|1) to be odd. This implies the following parities for the fields appearing in A i , i = 0, 1, 2, 3: [mn]αβ are even fields and A [mnr]αβ are odd fields ∀p ∈ N. (4.5) Having these parities fixed we can track the signs when moving a field across another one. Now let us derive the expressions for dA i , i = 0, 1, 2, 3: Notice that we have decomposed the sum in γ in a part with γ = α and in a part with γ = β. Let us enumerate the other dA i 's: 1 ; (4.7) 2 ; (4.8) 3 . (4.9) In the final expression for the four terms we denote with a superscript the number of dx's appearing. In the Lagrangian AdA, not every combination of the factors in the decompositions of A and dA is allowed, indeed we only need the terms where d 3 x appears,

JHEP03(2020)043
in order to obtain the correct top-form for the bosonic integration. Thus the allowed combinations are We now want to evaluate the terms of (4.10) in order to obtain the explicit form of the Lagrangian. Let us start from the last term, it has the form Recall that by definition we have δ(dθ 1 )δ(dθ 1 ) = 0 = δ(dθ 2 )δ(dθ 2 ), this implies that in the previous product we have ν = α and µ = β. Moreover, in order to avoid vanishing terms, we need the power of dθ α in the first term to be smaller or equal to the derivation order of δ(dθ α ) in the second term and the same holds true for dθ β and δ(dθ β ) as well. This fixes uniquely q in terms of p and therefore we have the reduction to a single sum as (4.12) The factors in (4.12) are obtained by integrations by parts and by fixing q = q(p) as discussed above.
In an analogous way we can calculate the other terms so that finally we get the Lagrangian: Notice that we have manifestly collected the dx's and δ(dθ)'s to ease the computation of the action integration on dx's and dθ's: (4.14) JHEP03(2020)043

Equations of motion from the action
We can now investigate the equations of motion by varying the action w.r.t. the various fields; let us consider for example all the terms with A [mnr]αβ (p) in the Lagrangian (4.13): where we have inserted the totally antisymmetric symbols as reminders for the symmetry of bosonic and fermionic indices. We can recast the last two terms of the previous expression as follows (4.17) where we have used the graded Leibniz's rule according to the parity of the fields described in (4.5). The total derivative terms may be neglected since it would lead to null terms after the Berezin's integration of the θ's.
We can now insert (4.16) and (4.17) in (4.15) and we get The variation of the action w.r.t. the field A [mnr]αβ (p) leads to: In an analogous way we can obtain the other equations of motion by varying the action w.r.t. the fields A (4.20) [nr]βα + (p + 1)∂ α A (p) [nr]βα = 0 ∀ p ∈ N ; (4.21) where we stress that in (4.21) and in (4.22) if p = 0 the fields A (−1) [nr]βα and A (−1) [mnr]βα are both defined to be zero.

Equations of motion from the curvature F
In this subsection we briefly show that the equations of motion already derived from the variational principle are the same that emerge by the usual flatness condition of (noninteracting) Chern-Simons theory F = dA = 0 at picture 1. For the sake of clarity we recall the strategy used to get the EoM. We gather the equations by the number of dx's JHEP03(2020)043 appearing, in other words we consider the various parts which are homogeneous in dx's; the four homogeneous parts are then formal series into the dθ's, therefore by power counting we can set each single term of the series equal to zero. This yields [nr]βα + (p + 1)∂ α A (p) [nr]βα = 0 ; (4.25) [mnr]βα = 0 .

Reducing the equations of motion
In the previous subsections we have explicitly obtained the equations of motion for super Chern-Simons theory with pseudoforms. Since the Lagrangian (4.13) we started with contains an infinite number of fields, we therefore have an infinite number of EoM. We now want to use these equations to determine which fields, when on shell, can be expressed as d-exact terms, i.e. we want to find out explicitly the cohomology (w.r.t. those forms which are related to the θ-dependent gauge parameters) representatives of the free theory at picture number 1. In this subsection we omit the calculations and present only the strategy to apply and its result. The interested reader is suggested to refer to appendix B where the explicit calculations are presented. We use the following strategy: first we expand the fields in powers of θ's (recall that the expansion is truncated to the term θ 1 θ 2 since they are odd variables), then we insert the expansions in the EoM (4.19) ÷ (4.22). We then use the resulting equations in order to find algebraic relations between the fields of the expansion. The results are then inserted back in (4.1) ÷ (4.4) and then we look for the terms that give rise to d-exact terms. We find that a representative of the cohomology class is 27) and the relative equation of motion is Remarkably, notice that even if we started from a SCS Lagrangian with an infinite number of fields, we have shown that there is only one physical field, indeed all the other fields are d-exact θ-dependent terms. Moreover we have shown that starting from the free SCS action with a general A 1|1 pseudoform we obtain the factorisation We now define an interaction term which can be integrated on a supermanifold. Apparently, a problem arises. Indeed in order to define an interaction term, we need three gauge fields A (1|1) , but the wedge product of three fields vanishes by anticommutativity of three Dirac delta functions in dθ 1 or dθ 2 . 7 Let us start from the action (3.22). If we factorise Y (0|2) = 2 α=1 v (α) · θδ(v (α) · dθ), where we have chosen any two linear independent vectors v (α) , such that v (1) · v (2) = 0, ). In other words we have distributed the PCO Y (0|2) on the gauge fields A (1|0) which have now picture 1 each. In order to have one more gauge field, one needs one more PCO. This can be done by inserting the combination Z w (1) Y w (1) = 1, thus obtaining The interaction term, rewritten in terms of pseudoforms, has the following structure In (4.30) we have inserted the PCO Z v in a generic place in the interaction term. However, a priori, we have to consider all the possible places where to put the PCO. Therefore, following [12,15], we are led to define the 2-product with picture degree -1 as m (−1) 2 : This definition encodes the prescription of an "equally-weighted" application of the PCO Z v , thus reflects the generality discussed above. 8 Notice that, after introducing we can recast the definition of m Care must be used in defining the product of two pseudoforms since it might lead to divergencies in the Feynman diagrams [21]. 8 In first quantised String Theory, the PCO is independent of worldsheet coordinates and therefore it can be placed at any point into a correlation function. However, in order to formulate a String Field Theory action, that arbitrariness can not be used since a given choice might break gauge invariance [46]. In [12] the authors avoid this problem by suitably smearing the PCO on the disc on which the correlation functions are computed. This democratic choice preserves gauge invariance an leads to the 2-product discussed in the text.

JHEP03(2020)043
In (4.34) we have adopted the coproduct formulation [15]. Observe that this product has form degree 0. In an analogous way, we can define a product with form degree −1 as m (−1) 2 : where [·, ·] denotes as usual the graded commutator. Starting from the definition (4.32), we now find an explicit expression for the interaction term. Details can be found in appendix B. Let us start from the action term where the trace is to be taken with respect to the group indices. From (4.37), we can extract the two following terms: We first calculate the action of the operator Z v on the A i 's, i = 0, 1, 2, 3 and on the product AA: iv β δ αβ v·dθ . (4.45)

JHEP03(2020)043
Notice that (4.43) and (4.44) are consequence of the general property By using these results in (4.40) we get to the explicit interaction term: (4.47) Therefore the Lagrangian reads Notice that the interaction term depends on the constant vector v α through v 1 v 2 , namely their relative phase. That resembles the usual frame dependence of Superstring Field Theory actions. That dependence is supposed to disappear whenever a calculation of a correlation function is performed. In our case, this is a consequence of the fact that any variation of the PCO Z v , by means a variation of its reference vector v α , is d-exact. Therefore, the action might depend upon the reference vector v, but the correlation functions will turn out to be independent of v. 9 The meaning of this interaction term can be understood if we consider the result (4.27). Indeed it is a straightforward calculation to verify that, if (4.27) holds, then the m (−1) 2 product reduces to the usual wedge product, hence the interaction term is the usual Chern-Simons one. Our result (4.48) is coherent with this observation as we can readily verify: (4.49) (4.49) shows that for the cohomology representative field the interaction term reduces to the usual one. A few remarks are necessary: first, we see that the interaction term does not depend on the vector v α , as expected; second, we see that we have two copies of the interaction term corresponding to the two propagating fields obtained in (4.27), i.e.B mαβ andB mβα . JHEP03(2020)043

Cyclicity of ·, ·
In order to derive the equations of motion we need the interior product to be cyclic. For the sake of completeness, let us verify it explicitly: let A, B, C be three (1|1)-pseudoforms, we want to verify that tr(Am since |A| = |B| = |C| = 1. In order to avoid a cumbersome notation, we omit the integration and trace symbols; we have We observe that the second term of (4.51) and the third term of (4.52) are the same. We now write the extended expressions for the other four terms: (4.54)

(4.56)
It is now easy to observe that the terms from (4.53) and (4.54) arrange together with the terms of (4.55) and (4.56) in order to for total derivative terms; for example if we subtract from the first term of (4.54) the first from (4.55) we get an expression of the form where attention should be paid when considering the correct minus signs. The same thing is easily verified for all the other terms as well.

JHEP03(2020)043
The cyclicity of the inner product is crucial in what concerns the variational principle involved in order to derive the expected equations of motion; indeed, when varying the interaction term with respect to the field A we get three contributions:  (4.60) The gauge invariant EoM are consistent at the present level of m 2 product. In the forthcoming subsections we will show that as a consequence of non-associativity of the m 2 product it is necessary to modify the Lagrangian (hence the EoM) and the definition of gauge variation.

Gauge invariance and the emergence of the A ∞ algebra
In this subsection we study the gauge invariance of the action. It is a well known result that a section of the bundle of Lie algebra-valued 1-forms under the action of g transforms as which infinitesimally becomes so that we have δ c A = Ac − cA + dc. In our case, A ∈ Ω (1|1) and thus a few remarks are mandatory: in order to have the right matching of form and picture degrees, we have that the gauge parameter c is a (0|1)−pseudoform and the products Ac and cA must be considered as m  In this setting, we have that the algebra of gauge transformations closes with respect to the commutator [·, ·] operation, that is The closure of the algebra is a direct consequence of the Jacobi identity as it can be easily verified.

JHEP03(2020)043
In our case, the non-associativity of the product m (−1) 2 invalidate both the previous results: the action is no longer gauge invariant and the algebra of gauge transformations does not close any longer. Let us see these two facts explicitly: having used the cyclicity of the internal product; thus we have A) .  is not associative, these two terms do not sum to zero and therefore the action is no longer gauge invariant. This is the reason why a non-associative product leads to the emergence a A ∞ -algebra structure: in order to have a gauge invariant theory, we need to add a piece with a 3-product, then a piece with a 4-product and so on.
Before doing this, let us analyse the closure of the gauge algebra. Let us rewrite (4.63) with the definition used in (A.10): does not satisfy the Jacobi identity, the algebra does not close. Notice, once again, that Jacobi identity plays a crucial role for the algebra to close. Indeed, if l 2 were to satisfy Jacobi identity, we would have had where in this case l . The break down of gauge-invariance shown in (4.74) suggests that we should add to the Lagrangian other terms in order to have the cancellation of the terms arising from the gauge variation and, therefore, a gauge-invariant action. This translates mathematically into the introduction of an A ∞ -algebra as mentioned previously and as we are about to show explicitly.
We now proceed by constructing explicitly the first multiproduct of the A ∞ -algebra. Let us consider the action discussed so far: Another way to check the need to introduce other terms in the action is to study the "Bianchi identities", i.e. we have to check whether dF = l . This is equivalent to verify the gauge invariance of the action, but this turns out to be useful for constructing explicitly the higher products. From (4.76) the field strength reads

JHEP03(2020)043
Clearly the Bianchi identity for F does not hold, in particular from (4.80) we have A, A) . Again, we have an extra term breaking the Bianchi identity. By following the prescription described above, we add to the action an extra term, which we denote by A, A, A) , in order to restore the identity. As we get so far, one can easily infer the algorithm to be used in order to construct the whole A ∞ Lagrangian. Before we go on to the closure of the gauge transformations, another issue is to be addressed: is m 3 in the Small Hilbert Space?
In order to answer this question, let us now look back at the definition of the 3-product given into the equation (4.82). If we neglect the d-exact term dΛ we have that the product m 3 is defined by a certain combination of m 2 andm 2 . Now, since the product m 2 is defined via the application of the operator Z, it maps pseudoforms into pseudoforms, as discussed in sections 2 and 4. This means that m 2 maps the SHS into itself. On the other hand, m 2 is defined via the operator Θ, that maps pseudoforms into inverse forms. Therefore, by contrast,m 2 maps the SHS into the LHS. This means, a priori, that m 3 gives values in the LHS. Here is where the d-exact term becomes relevant: it can be defined as a term that annihilates the LHS part resulting from them 2 part of m 3 .
Superstring Theory suggests a simple way to establish whether an objects lies in the SHS, this is based on the operator "η", whose definition was given in subsection 2.5.
If we want m having used extensively the properties of the operator η described in section 2. The second term is Finally, the fourth term reads  We can now manipulate this expression in order to find an explicit formulation for the multiproduct Λ, indeed we have: We can now define the formal expression so that the terms in (4.91) become   1 ≡ d, is exactly the one described in [47].

Supersymmetry at picture 1
In sections 2 and 3 we have discussed the supersymmetric action of free SCS and the rheonomic equations when working at picture 0. When working at picture 1, things work differently, nonetheless yield the same results. The gauge field A (1|1) can be decomposed in powers of V a exactly as we showed for the expansion in dx a . Hence the field strength will be decomposed consequently as In (4.103) we have a slight abuse of notation: in the superscripts (n|1), n is to be intended as the fermionic form number, and the pseudoforms F (n|1) I do not have a further decomposition in V 's. We now apply the Bianchi identity (recall formulas in (2.8)):

JHEP03(2020)043
By looking at the homogeneous parts in V , we get the system Now we can apply the conventional constraint prescription: in (4.103) we put equal to 0 the term with no V , i.e. F 0 = 0. By means of this prescription we can solve the previous system as follows: the last equation becomes where W is any function, because of the Fierz identity. This result gives us the correct way to identify the gaugino field strength W in terms of the gauge field A. We can substitute this result in the third equation of (4.105) in order to get where γ [ab] αβ was defined in section 2. Because of the trace properties of γ matrices in three dimensions, it follows that D α W α(0|1) = 0. Notice that, up to now, we have the same results as in the picture 0 case. From the known results at picture 0, we can infer that in the second equation of (4.105) we have having made, without loss of generality, a choice for the direction ψ 2 of the δ term; we have (4.114) We have therefore proved that the Bianchi identities at picture 1 are the same as the Bianchi identities at picture 0. Moreover we have an explicit way to find the expression of the gaugino field strength W at picture 1, i.e. (4.106).

Supersymmetry at picture 2
In this section we study the prescriptions that supersymmetry imposes at picture 2. Our analysis is meant to be compared with its analogous at pictures 0 and 1, as to find analogies and differences. Let us start from the decomposition of the field strength: [abc] . are the only possible because of the presence of the two δ's. In particular we have the decomposition: where ι µ is the usual compact notation to indicate a fermionic derivation on (one of) the two δ's. We can now apply the Bianchi identity: It is immediate to see that many terms in the expression are trivially 0 because of the explicit decomposition (4.116), and the Bianchi identity reduces to In particular we have where we have factorised the dependence on the bosonic indices with a totally antisymmetric tensor by writing F α [abc] = abc W α . As can be directly seen, in the picture 2 case it is not necessary to invoke a conventional constraint prescription in order to solve the abstract Bianchi identities, since there is not a term with no V 's from the beginning; however, by imposing this condition one recovers the usual form of the theory, i.e. D α W α = 0 . (4.120) Also, notice that under this constraint, we have the same field content of the previous two cases, i.e. a field with two antisymmetric bosonic indices satisfying the standard Bianchi identity and a field with a fermionic index satisfying the null superdivergence condition.

JHEP03(2020)043
4.10 Passing from a picture to another picture In this section we want to discuss how the informations described above are recovered in term of PCO's. To this end, let us rewrite the field strength contents in a "diagrammatic" fashion. With some abuse of notation, we omit the V 's and indicate only the fermionic form number and the picture number as to get: The PCO Z v described in the previous sections acts vertically, from the last line to the first. A priori, one might expect that the first and the second 0's of the last line get mapped to 0's in the first and second line, but there is a subtlety to be considered. Indeed, we can modify the PCO Θ as follows where X is an even vector field. This modification allows to "move diagonally" in the previous diagram when reducing the picture. In the following we will provide a justification for this modification, showing that it corresponds to a sort of "gauge transformation" of Θ. The operator Z v is said to be a Picture Lowering Operator because it is the left inverse of the Picture Raising Operator Y = θ α δ (dθ α ), which is a representative of the cohomology as discussed in the previous sections: We now show that it is also possible to make an analogous choice for the operator Θ, and in particular the transformation (4.121) is allowed. We begin with the general identity A general form for Λ is given by Clearly the application of the operator Z v reduces to This equation means that any modification Λ as in (4.123) that satisfies (4.127) amounts to the same modification as adding a term U to the PCO Z v compatible with (4.123). It is easy to see that, in particular, the additional piece described in (4.121) works: because neither Y nor dΛ contain dx pieces, thus the contraction ι E gives automatically 0. This shows that it is possible to add additional pieces in Θ that do not change the equivalence class of the cohomology and in particular that δ (ι v ) ι E does this. This re-definition of the operator Θ really allows to move diagonally as follows: Thanks to the diagonal arrows, this diagram is meant to show that the contributions at lower picture may come from various terms. For example, we see that F . This allows to better understand the nature of the conventional constraint, which, for example, is already implemented at picture 2.

Conclusions and outlook
We have discussed in detail the construction of the super-Cherns-Simons theory using the language of pseudoforms. We pointed out that the interaction term has to be built in terms of a non-associative product leading to a tower of interactions organized into a A ∞ algebra. Finally, the compatibility with supersymmetry is studied. This is the starting point to several applications and follow-ups. Let us list some of them.
1. As we have learnt from string theory, the introduction of PCO is related RNS formalism [6], but also in pure spinor framework [48]. The present analysis is in part directly related to pure spinor formalism as was addressed some time ago, by one of the author and G. Policastro in [49], where it is shown how the two frameworks are related. Since, at the moment, we have understood some of the aspects of classical field theories on supermanifolds, it would be very interesting to explore again the relation between PCO's in string theory and in quantum field theory.
2. One of the main motivations to build the 2-product in string field theory was to avoid the singularities emerging when two PCO's collide at the same point creating a potential divergence. We have seen from a preliminary work that the same phenomenon JHEP03(2020)043 appears also in the present context. Although Feynman diagrams computations will be presented somewhere else [21], we show in appendix C a simple case where the collision of two PCO generates such divergences.

One of the famous work by Witten on the relation between topological strings and
Chern-Simons gauge theories [50] can be finally repeated in the context of supermanifolds. Until now, these aspects of string theory and topological strings on supermanifolds have never been explored and we hope that the present framework might be suitable to address these problems. [45] was the case of D = 3, N = 1 supergravity. It was shown that the non-factorized form of the action leads to the component action matching the superspace constructions. Nonetheless, it has not been explored the same situation for higher dimensional supergravity models. A crucial aspect for supergravity models is the fact that PCO depends upon the dynamical fields and therefore a deeper analysis must be carried out.

5.
A long standing issue is the problem of auxiliary fields for higher dimensional and extended supersymmetry gauge theories and supergravity. We established a completely new framework to recast and re-think to that old problem, which might also serve to build off-shell supersymmetric models with extended supersymmetry.
6. Based on geometrical analysis ( [30] and [46]), we know that there are interesting supermanifolds which are not projected. For those manifolds, the factorized form cannot be globally defined since the existence of PCO Y (0|m) is granted only locally. The non-factorized form might be useful to consider both non-projected supermanifolds and non-split ones.
What is rather striking is the comparison between the factorised form of the action and the non-factorised form. Apparently, all the complications arise from the infinite-dimensional nature of pseudoforms, therefore it is natural to wonder whether a suitable field redefinition might immediately prove the equivalence of the two actions. Furthermore, we must investigate the theory at the quantum level where some potential singularities and divergences might jeopardise the classical equivalence relation. Nevertheless, we have shown that a very simple classical theory with some basic assumptions on the worldvolume supermanifold leads to a consistent algebraic structure of an A ∞ Chern-Simons theory. That construction parallels the EKS construction without referring to any string theory, conformal field theory, Riemann surfaces or using other mathematical ingredients.

JHEP03(2020)043
A A ∞ and L ∞ algebras Let us first recall the definitions of A ∞ and L ∞ algebras following [51] but by using a modern language: • an A ∞ -algebra, or a strongly homotopy associative structure on V , is a collection of linear maps m i : V ⊗ n → V such that they satisfy p+1 (a d , . . . , a p+q+1 , m p (a p+q , . . . , a q+1 ) , a q , . . . , a 1 ) = 0 . (A.1) • An L ∞ -algebra, or a strongly homotopy Lie structure on V , is a collection of skewsymmetric linear maps l i : V ⊗ n → V such that they satisfy j+k=i 0σ∈Sh(j;i) χ(σ; a 1 , . . . , a i )(−1) k l k+1 l j a σ(1) , . . . , a σ(j) , a σ(j+1) , . . . , a σ(i) = 0 , where Sh(j; i) are the permutations σ of {1, . . . , i} such that σ(1) < . . . < σ(j) and σ(j + 1) < . . . < σ(i) and χ(σ; a 1 , . . . , a i ) is the graded Koszul sign defined by In [51] the author stated that it is possible to obtain a SHLS starting from a SHAS by a "commutation operation". This can be rephrased in a more modern language by following [52] where it is proved the following proposition: the functor consists in antisymmetrising the A ∞ operations m n to get the L ∞ operations l m . This amounts to say that given an A ∞ relation (A.1) and by antisymmetrising it, we get the L ∞ relation (A.2). Let us try to understand this result by considering the first steps of the relations. First of all, we can observe that an hint to this result is given by the fact that by antisymmetrising the associative property we get the Jacobi identity: i.e. l 1 is a nilpotent operation. The second A ∞ relation is m 2 (a 2 , m 1 (a 1 )) + (−1) |a 1 |+1 m 2 (m 1 (a 2 ), a 1 ) + m 1 (m 2 (a 2 , a 1 )) = 0 , (A.8)

B.1 Reducing the equations of motion
In this subsection we explain in deep details how to reduce the equations of motion n order to determine the cohomology representative fields of the presudoform A (1|1) . In subsection 4.5 we have announced the strategy to be used and the result obtained, here we show the useful passages. First of all let us consider the expansion in powers of θ of any field: We start by applying this expansion to (4.19): We can separately equal to 0 the different coefficients of the monomials in θ, thus obtaining By inserting this result back in (B.1) we get Now we insert (B.1) in (4.20): Since these expressions are valid ∀p ∈ N, the second and the third equations are the same. The second equation is used in order to find a formal expression ofD rβα . Therefore when inserting these results back in (B.1) we get This will prove of fundamental importance as we will see shortly. For p = 0 the second and third equations in (B.11) are the same. We can insert these results back in (B.1) but now we keep the p = 0 case separated:   It is now a matter of rearranging all the terms correctly; for example, consider an expression like αβ (dθ α ) p+1 δ(dθ β ) (p) , (B.26) where we have used the fact thatÃ (p) αβ is even. This means that the first term in (B.22) and the third term in (B.23) arrange in a d-exact term. This means that we can omit them in order to get the right cohomological field. Let us now consider an expression like d −θ β θ αB (p) αβ (dθ α ) p δ (p) (dθ β ) = pθ αB (p) αβ (dθ α ) p δ (p−1) (dθ β ) + θ βB (p) αβ (dθ α ) p+1 δ (p) (dθ β )+ − dx m θ β θ α ∂ mB

JHEP03(2020)043
This means that we can arrange the second and third terms from (B.22) and the last term from (B.23) in a d-exact term. Observe that in order to arrange the terms correctly, we have to shift the first term in (B.27) p → p + 1 and the expression is valid for p = 0 as well.
Let us now consider an expression like m]αβ (dθ α ) p−1 δ (p) (dθ β ) ; (B.28) this means that we can arrange the first term in (B.23) and the third term in (B.24) as a d-exact term. Even in this case we have to make a shift on the second term of (B.28). Moreover the previous statement is not valid for the p = 0 term; in that case we have to consider an expression like which is exactly the separated p = 0 term. Let us now consider an expression like This allows us to fix as a d-exact term the second and fourth terms from (B.23) together with the last term from (B.24), but we have a fundamental observation to do: the previous relation allows us to fix the described terms after a shift p → p + 1. Notice that the term dx m θ βB (0) mαβ (x)δ (0) (dθ β ) remains free. Let us now consider an expression like d Ã (p) [mn]αβ θ α (dθ α ) p−1 δ(dθ β ) (p+1) = dx r θ α ∂ [rÃ (p) mn]αβ (dθ α ) p−1 δ(dθ β ) (p+1) +Ã (p) [mn]αβ (dθ α ) p δ(dθ β ) (p+1) ; (B.31) this means that we can arrange the first term in (B.24) and the third term in (B.25) as a d-exact term. As we previously noticed, we have to make a shift on the first term of (B.31). Moreover the previous statement is not valid for the p = 0 term; in that case we have to consider an expression like [mn]αβ δ (1) (dθ β ) − this allows as to arrange the second and third terms of (B.25) as d-exact. Even in this case it is necessary to do a shift p → p + 1.
Thus we have found that, modulo d-exact terms, a pseudoform A 1|1 of the Chern-Simons Lagrangian which is a representative of the cohomology class is Moreover we have obtained that the free Super Chern-Simons action with a general A 1|1 pseudoform leads to

JHEP03(2020)043
Firstly, let us consider the second term, it gives rise to This integral requires a regularisation, which can be formally done through an analytic continuation in the complex plane, even though some intricacies are involved (one has to interpret dθ as a "real number", which can be positive or negative). However, if we skip such problems, we can assign to (C.8) a finite value. Analogously, for the first term of (C.7) we have Again, this object is not well defined unless some formal regularisation is taken into account.
Of course, these arguments should be taken as a hint towards a complete analysis.
Open Access. This article is distributed under the terms of the Creative Commons Attribution License (CC-BY 4.0), which permits any use, distribution and reproduction in any medium, provided the original author(s) and source are credited.