Shifted Symplectic, Poisson, and BV Geometry

The BV pairing is an odd symplectic structure on the derived space of fields and equations. “Odd” is not merely a sign convention: the shift makes ghosts pair with ghost antifields and fields pair with antifields in the degrees required for the BV bracket. Nondegeneracy means a quasi-isomorphism of complexes, not pointwise invertibility of a matrix. On manifolds with boundary, the bulk pairing generally fails to be closed by a boundary flux, so boundary data must be supplied before the symplectic claim is valid.

Required background. Derived Critical Loci and Derived Gauge Quotients supplies the tangent complex of a gauge quotient. L∞ Algebras, Formal Moduli, and Field Equations supplies its nonlinear formal neighborhood. Helpful background. Classical Observables and Poisson Factorization explains how a bracket acts on observables. Stress–Energy Response and Background Variation supplies the distinction between metric dependence and dynamical fields.

Shifted nondegeneracy is a quasi-isomorphism

Let $\mathcal X$ be a derived space whose cotangent complex $\mathbb L_{\mathcal X}$ is perfect, with tangent complex $\mathbb T_{\mathcal X}=\mathbb L_{\mathcal X}^{\vee}$. A closed two-form $\omega$ of cohomological degree $n$ induces

$$\Theta_\omega:\mathbb T_{\mathcal X} \longrightarrow \mathbb L_{\mathcal X}[n].$$

It is $n$-shifted symplectic precisely when $\Theta_\omega$ is a quasi-isomorphism. This is the nondegeneracy condition of Pantev–Toën–Vaquié–Vezzosi 2013, Definition 1.18, pp. 23–24. It allows kernels in a chosen presentation, provided they are paired after passing to the full complex. An inverse, when it exists in the derived category, is an $n$-shifted Poisson bivector and induces a bracket of degree $-n$ on functions, subject to the appropriate homotopy Jacobi identities.

The canonical example is a derived critical locus. If $S$ is a function on a smooth finite-dimensional space, then $\operatorname{RCrit}(S)$ is $(-1)$-shifted symplectic. Its induced bracket on functions has degree $1$, matching the BV antibracket. In field theory, the same statement requires a topological vector-space category and support conditions under which the duality pairing is continuous and the inverse kernel can act on the selected observables.

First application: the free Maxwell BV complex

The field–antifield construction in BV Fields, Antifields, and Odd Symplectic Structure can be stated without hiding its domain. Let $M$ be a closed oriented Riemannian four-manifold. For the trivial real abelian gauge theory, place connection variations in degree zero and consider the elliptic deformation complex

$$\underset{\text{degree }-1}{\Omega^0(M)} \xrightarrow{d} \underset{\text{degree }0}{\Omega^1(M)} \xrightarrow{d{*}d} \underset{\text{degree }1}{\Omega^3(M)} \xrightarrow{d} \underset{\text{degree }2}{\Omega^4(M)}.$$

Here the middle map is $A\mapsto d(*dA)$, the linear Maxwell equation in form notation. Both consecutive composites vanish. The first does because $d^2=0$; the second does because $d(d(*dA))=0$. The four terms represent gauge parameters, fields, equations dual to fields, and Noether identities dual to gauge parameters.

Define a degree $-1$ pairing on compactly supported variations—automatically compactly supported because $M$ is compact—by

$$\omega(\delta_1,\delta_2) =\int_M\left( \delta_1A\wedge\delta_2A^+ -\delta_2A\wedge\delta_1A^+ +\delta_1c\wedge\delta_2c^+ -\delta_2c\wedge\delta_1c^+ \right),$$

with Koszul signs understood for homogeneous elements. The form degrees add to four, while the cohomological degrees add to one; hence the pairing has degree $-1$. Poincaré duality identifies $\Omega^1$ with the continuous dual of $\Omega^3$ and $\Omega^0$ with that of $\Omega^4$. Compatibility of the differential with integration by parts makes $\omega$ closed. In the derived sense it identifies the tangent complex with its shifted dual, including harmonic zero modes rather than deleting them.

For smooth compactly supported linear observables $F_f(A)=\int_M A\wedge f$ and corresponding antifield-linear observables, the inverse pairing gives the BV bracket directly. On gauge-invariant on-shell observables, a gauge-fixing homotopy and causal Green operator lead instead to the degree-zero Peierls bracket. These are related constructions but not the same bracket: the BV bracket acts on the enlarged field–antifield complex, while the physical Poisson bracket is obtained after the appropriate reduction.

An independent check is degree matching. $A$ has degree zero and $A^+$ degree one in the tangent-complex convention, so a degree-$-1$ bilinear form can pair them to degree zero. Likewise $c$ in degree $-1$ pairs with $c^+$ in degree two. Any proposal pairing $A$ directly with $A$ through this BV form would have the wrong cohomological degree.

Boundary flux and the failure of closedness

Now let $M$ have a boundary and repeat the integration-by-parts check. Variation of the Maxwell action produces the surface term

$$\int_{\partial M}\delta A\wedge *dA.$$

At the level of the symplectic current, the corresponding flux contains

$$\int_{\partial M} \left(\delta_1A\wedge *d\delta_2A -\delta_2A\wedge *d\delta_1A\right).$$

Unless boundary conditions make this vanish, or boundary fields absorb it, the bulk differential is not skew-adjoint and $\omega$ is not closed on the proposed bulk complex. The BV–BFV identity makes this precise: contraction of the bulk symplectic form with the cohomological vector field differs from the variation of the action by the pullback of a boundary one-form Cattaneo–Mnev–Reshetikhin 2014, §3.1, pp. 8–10.

This is the adversarial test. Take unrestricted boundary values and declare the closed-manifold pairing nondegenerate and closed. Choose variations with nonzero tangential $\delta A$ and normal electric field at the boundary; the displayed flux is nonzero, contradicting closedness. Dirichlet, Neumann, relative, or absolute conditions can repair the problem in different complexes, while a BFV extension keeps boundary degrees of freedom. There is no converse saying that a formally antisymmetric bulk integral determines a valid shifted symplectic field theory.

The conclusion is correspondingly limited. On a closed manifold, or with specified boundary data that cancel flux, the Maxwell deformation complex carries the stated $(-1)$-shifted pairing. This establishes the classical BV geometry. It does not establish a quantum measure, positivity, a preferred state, or equivalence of different gauge fixings beyond the class for which the required quasi-isomorphisms and analytic estimates have been proved.

Exercises

Verify that the Maxwell sequence above is a cochain complex.

Solution

For $c\in\Omega^0$, the first composite is $d(*d(dc))=d(*d^2c)=0$. For $A\in\Omega^1$, the second composite is $d(d(*dA))=d^2(*dA)=0$. Thus both neighboring composites vanish. The two identities are respectively gauge invariance of the equation and the linear Noether identity.

Give one boundary condition that kills the displayed Maxwell symplectic flux.

Solution

Fix the pullback of $A$ to $\partial M$. Allowed variations then have vanishing tangential pullback, so each term $\delta A\wedge *d\delta A'$ pulls back to zero. This is a relative or Dirichlet-type choice. It is only one possible choice; fixing the complementary electric data gives a different polarization.

References

• Cattaneo, Mnev, and Reshetikhin 2014, Classical BV Theories on Manifolds with Boundary, Communications in Mathematical Physics 332(2), 535–603. Open PDF
• Pantev, Toën, Vaquié, and Vezzosi 2013, Shifted Symplectic Structures, Publications Mathématiques de l’IHÉS 117, 271–328. Open PDF