Boundary Phase Spaces, Constraints, and the BFV Charge

The BFV boundary phase space separates gauge degeneracy from genuine boundary motion. Its constraint surface is coisotropic, its characteristic distribution contains the transformations declared gauge, and its degree-one charge resolves that reduction. A transformation with nonzero boundary Hamiltonian is not automatically a redundancy: quotienting it can erase a physical surface degree of freedom.

Required background. BV–BFV structures, boundaries, and gluing supplies the bulk–boundary equation. Constraints, Dirac brackets, and symplectic reduction supplies coisotropic reduction. Proper and improper gauge transformations fixes which boundary transformations are quotiented.

Helpful background. Surface charges, integrability, and ambiguities explains when a boundary generator exists as a function rather than only as a one-form on field space.

Boundary reduction and the BFV charge

Variation of a first-order action has the form

$$\delta S_M=\mathrm{EL}_M+\pi_M^*\widetilde\alpha_{\partial M}.$$

The two-form $\delta\widetilde\alpha_{\partial M}$ can be degenerate on the unrestricted boundary traces. Quotienting its kernel, when the quotient is smooth, gives the BFV phase space $\mathcal F^\partial_{\partial M}$ and its degree-zero symplectic form. The projected cohomological vector field $Q^\partial$ is Hamiltonian:

$$\iota_{Q^\partial}\omega^\partial=\delta S^\partial, \qquad \{S^\partial,S^\partial\}_{\partial}=0.$$

The zero locus $\mathrm{EL}_{\partial M}$ of $Q^\partial$ is locally coisotropic. Its reduction identifies solutions of the boundary constraints modulo the characteristic directions, provided constant-rank and smoothness assumptions hold Cattaneo, Mnev, and Reshetikhin 2014, §§3.1.3–3.2, pp. 11–13. Singular stabilizers require a derived or stratified quotient instead of this smooth picture.

The ghost-number-zero restriction recovers ordinary Hamiltonian reduction. The full BFV complex retains ghosts for the constraint algebra and antifields or ghost momenta for its resolution. Its degree-zero cohomology is the algebra of gauge-invariant functions on the reduced constraint surface only under the regularity hypotheses behind that identification.

There are therefore three spaces that should not be conflated: the unrestricted trace space before presymplectic reduction, the BFV constraint surface, and its reduced phase space. The first still contains kernel directions of the boundary two-form; the second imposes the moment-map equations; the third divides only by the characteristic distribution declared gauge. A smooth symplectic quotient theorem applies only when the rank is controlled and the quotient is well behaved. Otherwise the BFV complex, rather than a coarse orbit space, is the safer mathematical object.

Yang–Mills boundary data

Use first-order Yang–Mills theory on a compact oriented Riemannian $d$-manifold. The classical fields are a connection $A$ and an adjoint-valued $(d-2)$-form $B$, with

$$S_{\mathrm{YM}}[A,B] =\int_M\operatorname{tr}\left(B\wedge F_A+\frac12 B\wedge *B\right).$$

The minimal BV extension adds a ghost $c$ and antifields. On the boundary, the surviving coordinates include $(A,B,c,A^\dagger)$, and the BFV primitive is

$$\alpha^\partial_{\partial M} =\int_{\partial M}\operatorname{tr} \left(B\wedge\delta A+A^\dagger\wedge\delta c\right).$$

Thus

$$\omega^\partial_{\partial M} =\int_{\partial M}\operatorname{tr} \left(\delta B\wedge\delta A+\delta A^\dagger\wedge\delta c\right).$$

The minimal BFV charge may be written, in the convention inherited from the bulk action,

$$S^\partial_{\partial M} =\int_{\partial M}\operatorname{tr} \left(B\wedge d_Ac+\frac12 A^\dagger\wedge[c,c]\right).$$

Its Hamiltonian vector field gives $Q^\partial A=d_Ac$, $Q^\partial B=[B,c]$, $Q^\partial c=\tfrac12[c,c]$, and the corresponding transformation of $A^\dagger$. Nilpotence follows from the Jacobi identity and covariance of $d_A$. These formulas and their codimension-two continuation are derived in Cattaneo, Mnev, and Reshetikhin 2014, §5.2.1, pp. 31–33.

At ghost number zero, varying the gauge parameter and integrating by parts exposes the Gauss constraint. In canonical notation on a spatial region $\Sigma$,

$$\mathcal G[\lambda] =-\int_\Sigma\operatorname{tr}(\lambda\,D_AE) +\int_{\partial\Sigma}\operatorname{tr}(\lambda E_\perp).$$

If $\lambda|_{\partial\Sigma}=0$, the surface term vanishes and the transformation is a null gauge direction on the constraint surface. If $\lambda$ is allowed to be nonzero, the surface term is its Hamiltonian charge. Whether it is treated as a symmetry, enlarged gauge direction, or datum paired with an edge field is part of the boundary problem, not a universal convention. The physical charge calculation belongs at Surface Charges, Integrability, and Ambiguities.

What reduction may remove

Let $\mathcal G_0$ be the subgroup of gauge transformations trivial on the boundary. Reduction by $\mathcal G_0$ removes bulk redundancy while retaining boundary electric flux and the action of the quotient boundary group. Reducing by the full group instead is legitimate only if the boundary condition, extended phase space, or observable question declares those transformations redundant and treats their moment map consistently.

As a failure test, take a Maxwell region with nonzero normal electric flux and quotient by gauge parameters whose boundary value is arbitrary, without adding edge variables or fixing the flux. The charge $\int_{\partial\Sigma}\lambda E_\perp$ is then set to zero by fiat. The resulting phase space has lost distinguishable charged boundary directions. What survives is the reduced theory for the smaller class of observables insensitive to those directions, not the original boundary system.

Exercises

Verify that $Q^\partial F_A=[F_A,c]$.

Solution

Since $F_A=dA+\tfrac12[A,A]$ and $Q^\partial A=d_Ac$, one has $Q^\partial F_A=d_A(d_Ac)=[F_A,c]$. This covariance is the step that makes the Yang–Mills constraint surface invariant under the BFV differential.

When is the Gauss generator differentiable without an added boundary term?

Solution

It is differentiable on a phase space where either the gauge parameter vanishes at the boundary or the allowed variations force the relevant normal electric-flux variation to vanish. Otherwise the variation of the bulk constraint produces a surface term, and a compatible charge or boundary extension is required.

References

• 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.
• Donnelly, William, and Laurent Freidel. “Local Subsystems in Gauge Theory and Gravity.” Journal of High Energy Physics 2016, no. 9 (2016): 102. DOI; Open PDF.