AKSZ Sigma Models as BV–BFV Examples

The AKSZ construction turns finite-dimensional graded symplectic target data into an infinite-dimensional BV theory on a mapping space. On a source with boundary, transgression also produces the BFV one-form and exposes the exact condition on a boundary brane. It is a powerful source of topological BV–BFV examples, but it is not a universal quantization scheme and it does not remove zero-mode, anomaly, or convergence problems.

Required background. The BV complex and classical master equation supplies the target master function. BV–BFV structures, boundaries, and gluing supplies the relative equation. Bulk–boundary master equations and anomaly inflow separates classical compatibility from quantum obstruction cancellation.

Helpful background. BF theory as a topological gauge theory is the first field-theory example. Hamiltonian group actions and moment maps supplies the finite-dimensional Hamiltonian language.

Hamiltonian graded target

Let $M$ be an oriented $d$-manifold and let the target be a graded manifold $\mathcal X$ with a symplectic form $\omega_{\mathcal X}$ of degree $d-1$. Choose a primitive $\alpha_{\mathcal X}$ and a Hamiltonian function $\Theta$ of degree $d$ such that

$$\omega_{\mathcal X}=\delta\alpha_{\mathcal X}, \qquad \{\Theta,\Theta\}_{\mathcal X}=0.$$

The last equation makes $Q_{\mathcal X}=\{\Theta,-\}$ cohomological. AKSZ fields are maps

$$\mathcal F_M=\operatorname{Map}(T[1]M,\mathcal X).$$

Evaluation followed by integration over $T[1]M$ transgresses $\omega_{\mathcal X}$ to a degree $-1$ BV form. In local target coordinates $X^a$, the action is

$$S_M[X]=\int_{T[1]M} \left(\alpha_a(X)\,d_MX^a+\Theta(X)\right).$$

On a closed source, the target master equation and Stokes’ theorem imply $(S_M,S_M)=0$. On a source with boundary, the Stokes term is precisely the pullback of the transgressed boundary primitive. The mapping-space construction and relative Hamiltonian identity are Cattaneo, Mnev, and Reshetikhin 2014, §§6.1–6.3, pp. 41–45.

The degree hypotheses are essential. A symplectic target of the wrong degree gives the wrong BV bracket, and a function $\Theta$ with nonzero self-bracket makes the induced $Q_M$ fail to square to zero.

Chern–Simons as the bracket test

In dimension three, let $\mathfrak g$ carry a nondegenerate invariant pairing and take $\mathcal X=\mathfrak g[1]$. The pairing gives a degree-two symplectic form, while

$$\Theta(q)=\frac16\langle q,[q,q]\rangle$$

has degree three. The equation $\{\Theta,\Theta\}=0$ is equivalent to the Jacobi identity together with invariance of the pairing. Transgression gives the BV superconnection $\mathcal A\in\Omega^\bullet(M,\mathfrak g)[1]$ and

$$S_M[\mathcal A] =\int_M\left( \frac12\langle\mathcal A,d\mathcal A\rangle +\frac16\langle\mathcal A,[\mathcal A,\mathcal A]\rangle \right).$$

Its ghost-number-zero component is Chern–Simons theory. Restriction to the boundary produces the Atiyah–Bott symplectic form and the flatness BFV constraint. This example is a useful independent check: every sign in the mapping-space bracket must reproduce the Lie-algebra differential and curvature. The full classical calculation appears in Cattaneo, Mnev, and Reshetikhin 2014, §7.2, pp. 47–50.

The example also exposes the global ceiling of AKSZ data. The finite-dimensional target knows the infinitesimal Lie algebra and invariant polynomial, but level quantization, nontrivial principal bundles, large gauge transformations, and the global Chern–Simons line require additional geometry.

Abelian BF theory from AKSZ

Let $V$ be a finite-dimensional vector space and take

$$\mathcal X=T^*[d-1](V[1]).$$

Write its coordinates as $q\in V[1]$ and $p\in V^*[d-2]$, with $\omega_{\mathcal X}=\delta p\,\delta q$ and $\Theta=0$. A map from $T[1]M$ is a pair of superfields

$$\mathcal A\in\Omega^\bullet(M,V)[1], \qquad \mathcal B\in\Omega^\bullet(M,V^*)[d-2].$$

The AKSZ action and symplectic form reduce to

$$S_M=\int_M\langle\mathcal B,d\mathcal A\rangle, \qquad \omega_M=\int_M\langle\delta\mathcal B,\delta\mathcal A\rangle.$$

Therefore $Q_M\mathcal A=d\mathcal A$ and $Q_M\mathcal B=d\mathcal B$. The boundary primitive is the transgression of $p\,\delta q$. This reproduces Abelian BF theory and its BFV data, rather than merely matching its ghost-number-zero action Cattaneo, Mnev, and Reshetikhin 2014, §§5.4 and 6, pp. 36–45. The physical model is developed at BF Theory as a Topological Gauge Theory.

Boundary branes

A local AKSZ boundary condition may be obtained from a graded Lagrangian submanifold $\mathcal L\subset\mathcal X$ such that

$$\alpha_{\mathcal X}|_{\mathcal L}=0, \qquad \Theta|_{\mathcal L}=0.$$

Then $\operatorname{Map}(T[1]\partial M,\mathcal L)$ is Lagrangian in the boundary field space, the boundary primitive vanishes, and $Q_{\mathcal X}$ is tangent to the brane. For Abelian BF theory, setting the $p$ coordinate to zero or the $q$ coordinate to zero gives the two elementary complementary polarizations. More general linear Lagrangians are allowed when they respect the grading.

For the Poisson sigma model, admissible target branes are conormals to coisotropic submanifolds. This illustrates why “Lagrangian” alone is insufficient: tangency to the target $Q$ structure is a separate constraint Cattaneo, Mnev, and Reshetikhin 2014, §3.7, p. 22.

Failure boundary

If $\{\Theta,\Theta\}\ne0$, the induced vector field obeys $Q_M^2X^a=\tfrac12\{\{\Theta,\Theta\},X^a\}$, so there is no classical BV theory. If $\mathcal L$ is not Lagrangian, symplectic flux remains. If it is Lagrangian but $\Theta|_{\mathcal L}\ne0$, the boundary condition is not preserved by $Q$.

Even valid classical AKSZ data do not prove the quantum master equation. Configuration-space singularities, unimodularity or anomaly conditions, residual cohomology, and a gauge-fixing propagator remain to be controlled. The safe conclusion is a classical BV–BFV example until those quantum obligations are met.

Exercises

Check the degrees of the transgressed symplectic form.

Solution

The target form has degree $d-1$, while integration over $T[1]M$ lowers degree by $d$. Hence the mapping-space form has degree $(d-1)-d=-1$, exactly the BV degree.

Why does $\Theta|_{\mathcal L}=0$ imply tangency of $Q_{\mathcal X}$ to a Lagrangian $\mathcal L$?

Solution

For every tangent vector $v$ to $\mathcal L$, $d\Theta(v)=0$. Since $\iota_{Q_{\mathcal X}}\omega_{\mathcal X}=d\Theta$ and $\mathcal L$ is Lagrangian, the symplectic orthogonal of $T\mathcal L$ equals $T\mathcal L$. Thus $Q_{\mathcal X}$ lies in $T\mathcal L$.

References

• Alexandrov, Mikhail, Maxim Kontsevich, Albert Schwarz, and Oleg Zaboronsky. “The Geometry of the Master Equation and Topological Quantum Field Theory.” International Journal of Modern Physics A 12 (1997): 1405–1429. DOI; Open PDF.
• Cattaneo, Alberto S., Pavel Mnev, and Nicolai Reshetikhin. “Classical BV Theories on Manifolds with Boundary.” Communications in Mathematical Physics 332 (2014): 535–603. DOI; Open PDF.