Perturbatively Local Poisson Sigma-Model

• The paper introduces the perturbatively local Poisson sigma-model, a 2D topological field theory where the Poisson tensor is expanded as a Taylor series at constant maps.
• It explains how formal geometry and gauge fixing yield local Feynman rules with propagators on the worldsheet and vertices labelled by derivatives of the Poisson tensor.
• The analysis shows that under conditions like the torus or unimodular Poisson structures, quantum corrections vanish, reducing the effective action to its classical part.

Searching arXiv for the cited Poisson sigma-model papers to ground the article. arXiv search: "(Bonechi et al., 2011) Poisson sigma model on closed surfaces" A perturbatively local Poisson sigma-model is a two-dimensional AKSZ/BV topological field theory whose perturbative expansion is organized so that the dependence on the target Poisson geometry is expressed locally, either through the jets of the Poisson tensor at a constant map or through local differential polynomials on the worldsheet. In the closed-surface setting, the perturbative construction around vacua uses formal geometry to reduce all target-space data to formal Taylor series at a point $x_0\in M$, yielding Feynman rules in which propagators live solely on the source surface $\Sigma$ while vertices are labeled by derivatives of the Poisson tensor $\partial^k\pi(x_0)$ (Bonechi et al., 2011). In a later higher-spin realization with an infinite-dimensional target, “perturbatively local” is used more explicitly to mean that the expansion of the Poisson bivector produces only genuine local differential polynomials on $\Sigma$, with no non-local kernels in worldsheet coordinates (Bekaert et al., 24 Sep 2025).

1. AKSZ formulation and target-space Poisson data

The underlying target datum is a smooth manifold $M$ of dimension $m$ equipped with a Poisson tensor

$$\pi=\tfrac12\,\pi^{ij}(x)\,\partial_i\wedge\partial_j, \qquad \{\pi,\pi\}_{\rm SN}=0.$$

The source is a connected closed oriented surface $\Sigma$. In AKSZ language the space of fields is

$$\mathrm{Fields}=\mathrm{Map}\bigl(T[1]\Sigma,\;T^*[1]M\bigr),$$

whose points are described by superfields

$$X^i\in\Omega^0(\Sigma)[0]\oplus\Omega^1(\Sigma)[-1]\oplus\Omega^2(\Sigma)[-2],$$

$\Sigma$0

with the usual total form-plus-ghost grading (Bonechi et al., 2011).

In local coordinates $\Sigma$1 on $\Sigma$2, the BV action splits into an unperturbed term and a Poisson interaction,

$\Sigma$3

so that

$\Sigma$4

This action satisfies the classical master equation $\Sigma$5, and if $\Sigma$6 has $\Sigma$7 or if $\Sigma$8 is unimodular, it also satisfies the quantum master equation (Bonechi et al., 2011).

This AKSZ presentation is the canonical local starting point for the perturbative theory. A central feature is that the nonlinear structure is entirely encoded by the Poisson tensor, so perturbative locality becomes a question of how $\Sigma$9 is expanded and how the resulting interaction vertices are organized.

2. Formal geometry and expansion around constant maps

The perturbative analysis on closed surfaces is carried out around a constant map $\partial^k\pi(x_0)$0. To do so, one chooses an affine connection $\partial^k\pi(x_0)$1 on $\partial^k\pi(x_0)$2 with exponential map

$\partial^k\pi(x_0)$3

introduces formal fiber coordinates $\partial^k\pi(x_0)$4 on $\partial^k\pi(x_0)$5, and writes

$\partial^k\pi(x_0)$6

The Poisson tensor is then pulled back and Taylor-expanded: $\partial^k\pi(x_0)$7 All target-space information is thus converted into a formal power series in the fiber variables $\partial^k\pi(x_0)$8 (Bonechi et al., 2011).

The fields are split into zero modes, interpreted as vacua, and fluctuations: $\partial^k\pi(x_0)$9 where $\Sigma$0 and $\Sigma$1 vanish on cohomology and $\Sigma$2 labels the constant map. This produces a perturbation theory in which the local model near each vacuum depends only on the formal neighborhood of $\Sigma$3 (Bonechi et al., 2011).

This construction is the precise sense in which locality in the target emerges in the closed-surface theory. No non-local feature of $\Sigma$4 enters the Feynman rules at a given vacuum: the propagator belongs entirely to $\Sigma$5, while the target contributes only through the jet data of $\Sigma$6 at $\Sigma$7.

3. Gauge fixing, propagators, and graph expansion

A Hodge-type decomposition

$\Sigma$8

is chosen with homotopy operator $\Sigma$9 satisfying

$M$0

The integral kernel $M$1 of $M$2 is the propagator, and in this gauge the only nonzero two-point function is

$M$3

Interaction vertices come from the Taylor coefficients of the Poisson tensor. The $M$4-th Taylor term produces a vertex with two $M$5-legs and $M$6 $M$7-legs,

$M$8

Hence each vertex is weighted by derivatives of $M$9 evaluated at the vacuum $m$0 (Bonechi et al., 2011).

The effective action on vacua $m$1 is expressed as a sum over connected oriented graphs $m$2 with leaves decorated by $m$3,

$m$4

Here $m$5 is the number of loops, $m$6 is the graph automorphism factor, and $m$7 is the Fulton–MacPherson compactified configuration space of vertices in $m$8 (Bonechi et al., 2011).

The same pattern reappears in boundary and BV formulations. For Poisson sigma-models on surfaces with boundary, the propagator is constructed as a smooth one-form $m$9 satisfying

$$\pi=\tfrac12\,\pi^{ij}(x)\,\partial_i\wedge\partial_j, \qquad \{\pi,\pi\}_{\rm SN}=0.$$0

and Feynman weights are again configuration-space integrals built from propagators and local multilinear vertices (Cui et al., 2020). This suggests that perturbative locality is compatible with compactified configuration-space technology: locality is retained in the integrands even though amplitudes are organized by global configuration spaces.

4. Vanishing theorems, quantum corrections, and the torus partition function

A central perturbative result for closed surfaces is the vanishing of quantum corrections in two important situations. First, for $$\pi=\tfrac12\,\pi^{ij}(x)\,\partial_i\wedge\partial_j, \qquad \{\pi,\pi\}_{\rm SN}=0.$$1 in the axial gauge, every graph with at least one loop or with more than one vertex vanishes. Second, for arbitrary $$\pi=\tfrac12\,\pi^{ij}(x)\,\partial_i\wedge\partial_j, \qquad \{\pi,\pi\}_{\rm SN}=0.$$2, if the Poisson structure is regular, meaning constant rank, and unimodular, one can choose a formal Darboux exponential such that higher vertices vanish by homotopy identities. In both cases all quantum corrections $$\pi=\tfrac12\,\pi^{ij}(x)\,\partial_i\wedge\partial_j, \qquad \{\pi,\pi\}_{\rm SN}=0.$$3 with $$\pi=\tfrac12\,\pi^{ij}(x)\,\partial_i\wedge\partial_j, \qquad \{\pi,\pi\}_{\rm SN}=0.$$4 vanish, and the full effective action reduces to the tree-level term

$$\pi=\tfrac12\,\pi^{ij}(x)\,\partial_i\wedge\partial_j, \qquad \{\pi,\pi\}_{\rm SN}=0.$$5

(Bonechi et al., 2011).

On the torus, one must still integrate over the remaining zero modes. In the nondegenerate symplectic unimodular case, after choosing a compatible Kähler polarization, the partition function is shown to satisfy

$$\pi=\tfrac12\,\pi^{ij}(x)\,\partial_i\wedge\partial_j, \qquad \{\pi,\pi\}_{\rm SN}=0.$$6

Thus the torus partition function equals the Euler characteristic of the target manifold in this case (Bonechi et al., 2011). The same work states that in the case of a Kähler structure or of a trivial Poisson structure, the partition function on the torus is the Euler characteristic of the target, and gives evidence that this may hold more generally (Bonechi et al., 2011).

These results sharply delimit the role of perturbative locality. Locality at the level of vertices and propagators does not automatically imply nontrivial loop corrections; in the torus and regular unimodular settings, the perturbative expansion collapses to the classical sector. A common misconception is that a nontrivial Poisson interaction necessarily yields nontrivial quantum contributions. In the cases above, the homotopy structure and gauge choice eliminate them.

5. Differential Poisson algebras and local deformation quantization

A distinct but closely related perturbative framework is provided by the two-dimensional topological sigma model whose target carries a differential Poisson algebra on differential forms. The target data are a Poisson bivector

$$\pi=\tfrac12\,\pi^{ij}(x)\,\partial_i\wedge\partial_j, \qquad \{\pi,\pi\}_{\rm SN}=0.$$7

a connection one-form $$\pi=\tfrac12\,\pi^{ij}(x)\,\partial_i\wedge\partial_j, \qquad \{\pi,\pi\}_{\rm SN}=0.$$8 respecting $$\pi=\tfrac12\,\pi^{ij}(x)\,\partial_i\wedge\partial_j, \qquad \{\pi,\pi\}_{\rm SN}=0.$$9, and the de Rham differential $\Sigma$0, subject to graded skew-symmetry, graded Leibniz, compatibility with $\Sigma$1, and the graded Jacobi identity. In local coordinates, with the choice $\Sigma$2, the unique covariant bracket is

$\Sigma$3

with graded Jacobi equivalent to the conditions

$\Sigma$4

(Arias et al., 2015).

On a closed two-manifold $\Sigma$5, the worldsheet fields are $\Sigma$6, $\Sigma$7, $\Sigma$8, and $\Sigma$9, and the manifestly target-space covariant Hamiltonian action is

$$\mathrm{Fields}=\mathrm{Map}\bigl(T[1]\Sigma,\;T^*[1]M\bigr),$$0

After Lorenz-type gauge fixing $$\mathrm{Fields}=\mathrm{Map}\bigl(T[1]\Sigma,\;T^*[1]M\bigr),$$1, $$\mathrm{Fields}=\mathrm{Map}\bigl(T[1]\Sigma,\;T^*[1]M\bigr),$$2 and expansion around a constant background $$\mathrm{Fields}=\mathrm{Map}\bigl(T[1]\Sigma,\;T^*[1]M\bigr),$$3, $$\mathrm{Fields}=\mathrm{Map}\bigl(T[1]\Sigma,\;T^*[1]M\bigr),$$4, the free action

$$\mathrm{Fields}=\mathrm{Map}\bigl(T[1]\Sigma,\;T^*[1]M\bigr),$$5

yields propagators

$$\mathrm{Fields}=\mathrm{Map}\bigl(T[1]\Sigma,\;T^*[1]M\bigr),$$6

where $$\mathrm{Fields}=\mathrm{Map}\bigl(T[1]\Sigma,\;T^*[1]M\bigr),$$7 is the scalar Green’s function on $$\mathrm{Fields}=\mathrm{Map}\bigl(T[1]\Sigma,\;T^*[1]M\bigr),$$8 (Arias et al., 2015).

The path integral is expanded in $$\mathrm{Fields}=\mathrm{Map}\bigl(T[1]\Sigma,\;T^*[1]M\bigr),$$9 around this Gaussian theory, and each Feynman graph is a worldsheet integral of a product of propagators and local interaction vertices. The work states explicitly that, since both propagators and vertices are supported on small neighborhoods on $$X^i\in\Omega^0(\Sigma)[0]\oplus\Omega^1(\Sigma)[-1]\oplus\Omega^2(\Sigma)[-2],$$0, locality is manifest and preserved order by order in $$X^i\in\Omega^0(\Sigma)[0]\oplus\Omega^1(\Sigma)[-1]\oplus\Omega^2(\Sigma)[-2],$$1 (Arias et al., 2015). Boundary insertions of differential-form observables then define a graded star product

$$X^i\in\Omega^0(\Sigma)[0]\oplus\Omega^1(\Sigma)[-1]\oplus\Omega^2(\Sigma)[-2],$$2

with

$$X^i\in\Omega^0(\Sigma)[0]\oplus\Omega^1(\Sigma)[-1]\oplus\Omega^2(\Sigma)[-2],$$3

and a second-order term containing iterated covariant derivatives, $$X^i\in\Omega^0(\Sigma)[0]\oplus\Omega^1(\Sigma)[-1]\oplus\Omega^2(\Sigma)[-2],$$4, and $$X^i\in\Omega^0(\Sigma)[0]\oplus\Omega^1(\Sigma)[-1]\oplus\Omega^2(\Sigma)[-2],$$5 (Arias et al., 2015). On zero-forms this reduces to the usual Kontsevich star product, while the $$X^i\in\Omega^0(\Sigma)[0]\oplus\Omega^1(\Sigma)[-1]\oplus\Omega^2(\Sigma)[-2],$$6-$$X^i\in\Omega^0(\Sigma)[0]\oplus\Omega^1(\Sigma)[-1]\oplus\Omega^2(\Sigma)[-2],$$7 sector extends the construction to arbitrary form degrees (Arias et al., 2015).

6. Boundaries, $$X^i\in\Omega^0(\Sigma)[0]\oplus\Omega^1(\Sigma)[-1]\oplus\Omega^2(\Sigma)[-2],$$8 structures, and higher-spin perturbative locality

For surfaces with boundary, the BV master action can be written in terms of a curved $$X^i\in\Omega^0(\Sigma)[0]\oplus\Omega^1(\Sigma)[-1]\oplus\Omega^2(\Sigma)[-2],$$9-algebra

$\Sigma$00

with multilinear operations $\Sigma$01 and pairing $\Sigma$02, and BV fields

$\Sigma$03

Writing $\Sigma$04, the classical BV action takes the AKSZ/Chern–Simons-type form

$\Sigma$05

and satisfies the classical master equation $\Sigma$06 (Cui et al., 2020).

This framework recovers deformation quantization through boundary observables. For two boundary insertions $\Sigma$07, the resulting associative product is

$\Sigma$08

and for the upper half-plane with the standard angular propagator it reduces exactly to Kontsevich’s universal formula (Cui et al., 2020). The same construction yields bulk $\Sigma$09-operations, boundary $\Sigma$10-operations, and a Swiss–Cheese algebra $\Sigma$11 of local observables (Cui et al., 2020).

The higher-spin realization pushes perturbative locality into an infinite-dimensional target setting. There the classical action is again of Poisson sigma-model form,

$\Sigma$12

with target coordinates $\Sigma$13 indexed by generators $\Sigma$14 of the associative algebra

$\Sigma$15

where $\Sigma$16 is the Weyl algebra generated by $\Sigma$17 with $\Sigma$18 (Bekaert et al., 24 Sep 2025). The coordinates arise by expanding a zero-form $\Sigma$19 as

$\Sigma$20

In this model,

$\Sigma$21

with $\Sigma$22 given by the structure constants of the commutator Lie algebra $\Sigma$23, and the higher tensors obtained from higher $\Sigma$24-vertices such as $\Sigma$25 (Bekaert et al., 24 Sep 2025). The paper defines “perturbatively local” by requiring that, in this expansion, all vertices remain integrals of genuine local differential polynomials on $\Sigma$26, with no non-local kernels. The quadratic correction,

$\Sigma$27

is integrated over a compact simplex in auxiliary parameters, and the same structure is stated for all higher vertices (Bekaert et al., 24 Sep 2025).

The gauge transformations retain the standard Poisson sigma-model form,

$\Sigma$28

with closure guaranteed by the Poisson Jacobi identity, and the BV/AKSZ master action satisfies $\Sigma$29 order by order (Bekaert et al., 24 Sep 2025). The paper further states that there are no gauge anomalies and that perturbative locality together with the absence of obstructions in the Moyal–Weyl algebra guarantees solvability of the quantum master equation by standard Fedosov–Kontsevich–Cattaneo–Felder arguments (Bekaert et al., 24 Sep 2025).

7. Locality, globalization, and conceptual scope

The closed-surface formalism makes the target-space locality statement especially precise. Every ingredient in the Feynman rules arises from the formal Taylor expansion of $\Sigma$30 around $\Sigma$31 and the local exponential map. The propagator is purely a worldsheet object and is independent of $\Sigma$32; all vertices are labeled by the jets $\Sigma$33; and each Feynman integral

$\Sigma$34

is multiplied by a polynomial in the derivatives $\Sigma$35 (Bonechi et al., 2011). Globalization over $\Sigma$36 is achieved by patching the formal neighborhoods through the Grothendieck connection of formal geometry, producing a global effective action

$\Sigma$37

whose Taylor expansion at each $\Sigma$38 reproduces the local perturbative construction (Bonechi et al., 2011).

For manifolds with boundary, gauge fixing requires additional auxiliary choices, including a Riemannian metric, a connection on $\Sigma$39, a representative $\Sigma$40 of the diagonal class, and a cutoff function $\Sigma$41 near the diagonal. A family version of the quantum master equation implies that the cohomology class of the effective action and of correlation functions is independent of these gauge choices (Cui et al., 2020). The same work states that the vacuum graph sum depends only on the topology of $\Sigma$42 and on the Poisson class $\Sigma$43, providing a candidate “Poisson–worldsheet invariant” (Cui et al., 2020).

Taken together, these constructions show that “perturbatively local Poisson sigma-model” does not designate a single universal formalism, but rather a tightly related family of perturbative realizations of Poisson sigma-models. In one sense, it means locality in the target jets around constant maps (Bonechi et al., 2011). In another, it means that the full perturbative expansion continues to be built from local worldsheet differential polynomials, even for infinite-dimensional higher-spin targets (Bekaert et al., 24 Sep 2025). A plausible implication is that the phrase marks a methodological criterion rather than a new class of classical actions: locality is preserved not by changing the Poisson sigma-model ansatz, but by controlling how its perturbation theory is organized.