Skip to content

Anomaly-Induced and Nonlocal Actions

The trace anomaly fixes the Weyl variation of an effective action, not the entire action. Integrating that variation produces a particular nonlocal or Wess–Zumino functional; boundary conditions, zero modes, local scheme terms, and arbitrary Weyl-invariant functionals remain to be supplied before a stress tensor is determined.

Required background. Trace Anomalies and Convention Translation fixes the anomaly coefficients, Matter-Induced Nonlocal Form Factors supplies nonlocal kernels, and The 1PI Effective Action and Mean-Field Equations supplies metric variation.

Helpful background. Wess–Zumino Consistency and Descent explains integrability, while Conservation, Local Covariance, and the Backreaction Source supplies stress-tensor Ward identities.

Write the four-dimensional anomaly in a declared Euclidean convention as

2gEgμνδΓδgμν=1(4π)2(cC2aE4+bE2RE).\frac{2}{\sqrt{g_E}}g_{\mu\nu} \frac{\delta\Gamma}{\delta g_{\mu\nu}} =\frac1{(4\pi)^2} \left(cC^2-aE_4+b\nabla_E^2R_E\right).

The sign relating this Euclidean equation to the site’s Lorentzian stress definition must be performed with the full metric continuation. The coefficients aa and cc multiply nontrivial anomaly classes. The coefficient bb is shifted by a finite local RE2R_E^2 counterterm.

It is useful to combine

Q=E423E2RE\mathcal Q=E_4-\frac23\nabla_E^2R_E

with the fourth-order conformal operator

Δ4=E4+2Rμνμν23REE2+13(μRE)μ.\Delta_4=\nabla_E^4 +2R^{\mu\nu}\nabla_\mu\nabla_\nu -\frac23R_E\nabla_E^2 +\frac13(\nabla^\mu R_E)\nabla_\mu.

For gμν=e2σgˉμνg_{\mu\nu}=e^{2\sigma}\bar g_{\mu\nu},

gEQ=gˉ(Qˉ+4Δˉ4σ),gEC2=gˉCˉ2.\sqrt{g_E}\,\mathcal Q =\sqrt{\bar g}\left(\bar{\mathcal Q}+4\bar\Delta_4\sigma\right), \qquad \sqrt{g_E}\,C^2=\sqrt{\bar g}\,\bar C^2.

These identities integrate the nontrivial part of the anomaly.

Section titled “First application: a conformally related background”

Integrating along gt=e2tσgˉg_t=e^{2t\sigma}\bar g yields the Wess–Zumino difference

Γ[e2σgˉ]Γ[gˉ]=1(4π)2gˉ[cσCˉ2a(σQˉ+2σΔˉ4σ)]+Γb[e2σgˉ]Γb[gˉ],\begin{aligned} \Gamma[e^{2\sigma}\bar g]-\Gamma[\bar g] =\frac1{(4\pi)^2}\int\sqrt{\bar g}\bigl[ &c\,\sigma\bar C^2 -a\left(\sigma\bar{\mathcal Q} +2\sigma\bar\Delta_4\sigma\right) \bigr]\\ &+\Gamma_b[e^{2\sigma}\bar g]-\Gamma_b[\bar g], \end{aligned}

where Γb\Gamma_b is a chosen local R2R^2 representative for the scheme-dependent total derivative. Differentiating with respect to σ\sigma returns the anomaly, providing a direct check of every factor.

A covariant nonlocal representative for the nontrivial part is

Γpart[g]=18(4π)2gQΔ41(2cC2aQ).\Gamma_{\rm part}[g] =\frac1{8(4\pi)^2} \int\sqrt g\, \mathcal Q\,\Delta_4^{-1} \left(2cC^2-a\mathcal Q\right).

Its Weyl variation is gσ(cC2aQ)/(4π)2\int\sqrt g\,\sigma(cC^2-a\mathcal Q)/(4\pi)^2. This is the Riegert construction Riegert 1984, pp. 56–60. The inverse Δ41\Delta_4^{-1} requires a domain, treatment of its kernel, and boundary or Green-function data. A Lorentzian in–out inverse is not automatically retarded.

On a compact manifold, define the inverse only after projecting out every zero mode or imposing a normalization condition; otherwise Δ41Q\Delta_4^{-1}\mathcal Q is undefined. On a conformally flat background C2=0C^2=0, the anomaly still fixes only the conformal-factor dependence relative to gˉ\bar g: it does not determine Γ[gˉ]\Gamma[\bar g]. These two checks separate a genuine anomaly coefficient from information hidden in the reference functional or Green function. For backreaction, the in–in kernel must also make the metric response real, conserved, and retarded.

Let Sinv[g]S_{\rm inv}[g] be any Weyl-invariant functional, local or nonlocal:

Sinv[e2σg]=Sinv[g].S_{\rm inv}[e^{2\sigma}g]=S_{\rm inv}[g].

Then

Γ[g]=Γpart[g]+Sinv[g]\Gamma'[g]=\Gamma_{\rm part}[g]+S_{\rm inv}[g]

has the same trace anomaly. Nevertheless,

2gδSinvδgμν\frac{2}{\sqrt g}\frac{\delta S_{\rm inv}}{\delta g^{\mu\nu}}

is generally a nonzero conserved traceless stress contribution, and its second variation changes response. Different homogeneous solutions of Δ4\Delta_4 similarly encode boundary or state information. Mazur and Mottola formulate this cohomological nonuniqueness explicitly Mazur and Mottola 2001, §§ II–IV.

Therefore the anomaly licenses the trace and a particular integrated representative. It does not select a quantum state, determine all traceless stress components, or prove a cosmological backreaction solution.

The structure map shows anomaly integration branching from, but not exhausting, the nonlocal effective action. Inspect the explicit homogeneous-functional branch.

The trace anomaly integrates to a particular nonlocal action while Weyl-invariant functionals and Green-function data remain undetermined

Anomaly coefficients fix a Weyl variation; local scheme terms, kernel boundary data, and traceless Weyl-invariant response require additional input. Schematic; not to scale.

The displayed formula assumes a smooth four-dimensional geometry and a defined inverse of Δ4\Delta_4. Boundaries add anomaly and Wess–Zumino terms; zero modes obstruct a naive inverse; causal use requires an in–in Green function. See Domain and failure conditions.

The failure map stops the inference from a correct trace to a unique stress tensor. Adding SinvS_{\rm inv} is the decisive counterexample: the trace is unchanged while response changes.

An unchanged trace anomaly can coexist with different traceless stresses, boundary conditions, states, and causal kernels

Anomaly integration is nonunique up to local scheme terms and Weyl-invariant functionals; choosing a particular representative does not remove that freedom. Schematic; not to scale.

Anomaly coefficients and sign translation remain with Trace Anomalies and Convention Translation. Metric variation continues in Effective-Action Variation, Stress Tensors, and Consistency Checks.

  • Mazur, Paweł O., and Emil Mottola. “Weyl Cohomology and the Effective Action for Conformal Anomalies.” Physical Review D 64 (2001): 104022. DOI. Open PDF.
  • Riegert, Ronald J. “A Non-Local Action for the Trace Anomaly.” Physics Letters B 134 (1984): 56–60. DOI.