Skip to content

Obstruction–Deformation Complexes and Quantization Classes

A deformation problem is controlled not just by its parameters but by a cochain complex. Degree-one cohomology contains obstructions, degree-zero cohomology parameterizes inequivalent corrections after an obstruction vanishes, and degree-minus-one cohomology records infinitesimal equivalences between those corrections. In perturbative BV quantization the complex must be the local deformation complex. Exactness in an unrestricted complex is insufficient, because a nonlocal primitive is not an admissible counterterm.

Required background. BV Quantization and Obstruction–Deformation Complexes supplies the effective BV equation. L∞ Algebras, Formal Moduli, and Field Equations supplies the Maurer–Cartan interpretation. Deformation Quantization, Formality, and Star Products supplies the order-by-order associativity analogy. Helpful background. Ward Identities and Anomalous Obstructions explains the physical meaning of a nonzero class. Microlocal Renormalization Ambiguities and Their Classification supplies the distributional extension problem behind local counterterms.

The obstruction class at one perturbative order

Section titled “The obstruction class at one perturbative order”

Let I0I_0 be a classical interaction satisfying the classical master equation, and let

dI0=Q+{I0,}d_{I_0}=Q+\{I_0,-\}

act on the complex ClocC^\bullet_{\mathrm{loc}} of admissible local functionals, with any symmetry and scaling restrictions included in the definition of that complex. Seek an effective quantum interaction

I=I0+I1+2I2+I=I_0+\hbar I_1+\hbar^2I_2+\cdots

satisfying the renormalized quantum master equation. If terms through order n1\hbar^{n-1} have been chosen, the coefficient at order n\hbar^n has the form

dI0In+Θn=0.d_{I_0}I_n+\Theta_n=0.

The lower-order master equation and the graded Jacobi identity imply dI0Θn=0d_{I_0}\Theta_n=0. Therefore [Θn]H1(Cloc,dI0)[\Theta_n]\in H^1(C^\bullet_{\mathrm{loc}},d_{I_0}) is the obstruction. A lift exists exactly when this class vanishes. If Θn=dI0Bn\Theta_n=d_{I_0}B_n, one may take In=BnI_n=-B_n; any two choices differ by a closed degree-zero cochain. Modulo degree-minus-one equivalences, the set of lifts is a torsor for H0H^0. This is the field-theoretic version of the standard obstruction–deformation pattern.

Costello’s inductive construction makes the statement precise with simultaneous loop and polynomial-degree filtrations: the obstruction is closed and a lift is exactly a cochain whose differential is its negative Costello 2007, Lemma 13.1.2, pp. 53–54. The proof uses the master equation at the previous filtration stage; it does not assume that every closed local cochain is exact.

At first quantum order one can write schematically

Θ1=ΔrenI0,\Theta_1=\Delta_{\mathrm{ren}}I_0,

where Δren\Delta_{\mathrm{ren}} denotes the finite local remnant of the regulated BV contraction. The notation is schematic because the unregulated BV Laplacian on local functionals is not defined. A regulator, subtraction prescription, and removal limit are part of the construction. Changing the prescription changes Θ1\Theta_1 by an allowed coboundary when the comparison theorem’s hypotheses hold, so the cohomology class—not a chosen representative—is the invariant obstruction.

First application: perturbative Chern–Simons theory

Section titled “First application: perturbative Chern–Simons theory”

Connect this deformation problem to Chern–Simons Actions and Level Quantization. Let MM be a closed oriented three-manifold, GG compact with an invariant nondegenerate pairing, and A0A_0 a flat connection such that

H(M,adPA0)=0.H^\bullet(M,\operatorname{ad}P_{A_0})=0.

Acyclicity removes perturbative zero modes around A0A_0. The BV fields are Ω(M,adP)[1]\Omega^\bullet(M,\operatorname{ad}P)[1], and the cubic Chern–Simons interaction supplies I0I_0. With heat-kernel regularization, the candidate one-loop obstruction is the local degree-one cocycle obtained from the short-distance boundary of the configuration-space integral. In flat space, Costello proves that there are no counterterms and that the uncorrected Chern–Simons interaction satisfies the renormalized master equation Costello 2007, Theorem 15.2.1, pp. 69–70. Thus the field-dependent local obstruction class represented by Θ1\Theta_1 vanishes in that scheme and under those hypotheses.

The global theorem is deliberately weaker than a statement about the complete partition function: it constructs a solution canonically up to contractible choice modulo field-independent constants Costello 2007, Theorem 15.0.6, pp. 66–67. Those constants matter for framing dependence. Around an acyclic flat connection, Axelrod and Singer compute the metric variation of the two-loop contribution and show that subtracting a specified multiple of the gravitational Chern–Simons functional, defined after choosing a framing, removes that metric dependence Axelrod–Singer 1992, Theorem 5.5 and Corollary 5.6, PDF pp. 29–30. A later paper supplies the full compactified-configuration-space treatment of the perturbation theory.

Consequently there are two related but distinct classifications. Field-dependent QME lifts are controlled by the local BV cohomology and are canonical modulo constants in Costello’s construction. Restoring the field-independent perturbative invariant requires a framing-dependent gravitational counterterm; after a framing is fixed, remaining choices are the permitted closed degree-zero local terms and an overall normalization. This does not derive the integer level condition from perturbation theory. Large-gauge invariance and the global form of GG impose that separate nonperturbative requirement.

An independent check is degree and locality. Θn\Theta_n has ghost number one, as required for an anomaly, whereas InI_n and an allowed counterterm have ghost number zero. The gravitational Chern–Simons correction is local in the background metric but depends on a framing; it therefore belongs to the field-independent background sector that the modulo-constants theorem excludes.

Suppose Θ1\Theta_1 is closed in the local complex but becomes exact only after applying a Green operator GG:

Θ1=dI0(GΘ1).\Theta_1=d_{I_0}(G\Theta_1).

The kernel of GG has support away from the diagonal, so GΘ1G\Theta_1 is generally nonlocal. Adding it to the action would correlate separated regions and violate the admissible counterterm condition. The obstruction therefore remains nonzero in H1(Cloc)H^1(C^\bullet_{\mathrm{loc}}) even though it vanishes in a larger complex of all cochains. This is the requested adversarial failure: unrestricted exactness does not license a local quantization.

There is also no converse from a vanishing first obstruction. Higher Θn\Theta_n can be nonzero, a regulator-removal estimate can fail, or the resulting formal series can have no nonperturbative meaning. The theorem at each order establishes one lift in a specified local formal problem, not a complete quantum field theory.

Assume [Θn]=0[\Theta_n]=0 and let InI_n and InI'_n be two solutions of dI0In=Θnd_{I_0}I_n=-\Theta_n. Show that their difference defines a class in H0H^0.

Solution

Subtracting the two equations gives dI0(InIn)=0d_{I_0}(I_n-I'_n)=0. Both terms have degree zero, so their difference is a degree-zero cocycle. Changing either solution by dI0d_{I_0} of a degree-minus-one cochain is an equivalence, leaving a class in H0(Cloc,dI0)H^0(C^\bullet_{\mathrm{loc}},d_{I_0}).

Why does acyclicity of A0A_0 simplify, but not prove, perturbative Chern–Simons quantization?

Solution

Acyclicity removes harmonic zero modes and makes the gauge-fixed kinetic operator invertible on the relevant complement, so a propagator can be defined without residual finite-dimensional fields. It does not by itself control ultraviolet collisions, prove the local obstruction class vanishes, fix framing dependence, or impose large-gauge level quantization.