---
title: 'Weyl Consistency: Anomaly Integrability in QFT'
url: https://www.emergentmind.com/topics/weyl-consistency-conditions
type: topic
---

# Weyl Consistency: Anomaly Integrability in QFT

Weyl consistency conditions are integrability relations imposed on quantum field theories by the commutativity of local Weyl rescalings. In the modern local renormalization-group formulation, the metric and couplings are treated as background fields, the trace anomaly becomes a local functional of these sources, and the requirement that two successive Weyl variations commute constrains both anomaly coefficients and RG data. In even dimensions this framework ties the Euler term and Weyl-invariant curvature densities to distinct cohomological classes, while in perturbation theory it yields gradient-flow-type relations for beta functions and candidate monotonic quantities such as \(c\)- and \(a\)-functions [0704.2472] [1308.1096].

## 1. Algebraic origin in Wess–Zumino consistency

The most basic formulation starts from a classically diffeomorphism- and Weyl-invariant theory coupled to a background metric \(g_{\mu\nu}(x)\). Quantum mechanically, the effective action \(\Gamma[g]\) acquires a Weyl anomaly defined by
\[
\delta_\omega \Gamma[g] \equiv A[\omega,g].
\]
Wess–Zumino consistency requires the anomaly to satisfy an integrability condition. In BRST language, with diffeomorphism ghost \(\xi^\mu(x)\), Weyl ghost \(\omega(x)\), and BRST operator \(s=s_p+s_w\), the condition is
\[
s\,A=0
\]
in the space of integrated local functionals at ghost number one. Because diffeomorphism invariance can be preserved by suitable counterterms, the relevant cohomology problem reduces to
\[
H^{1,d}(s_w|d),
\]
the cohomology of the Weyl BRST differential modulo exterior derivatives in \(d\) dimensions [0704.2472].

In even dimension \(d=2m\), the general solution splits into two classes. The type-A anomaly is unique and proportional to the Euler density \(E_{2m}\); it is the only class with a non-trivial descent, paralleling the Stora–Zumino descent of the non-Abelian chiral anomaly. Type-B anomalies are all other nontrivial solutions, of the form \(\int \omega\, I_i(g)\), where each \(I_i(g)\) is a strictly Weyl-invariant scalar density of dimension \(d\), typically built from Weyl tensors and their covariant derivatives. The resulting cohomology is
\[
H^{1,d}(s_w|d)=\langle \omega E_d\rangle \oplus \bigoplus_i \langle \omega I_i\rangle.
\]
This classification is purely algebraic and regularization-independent: it uses locality, covariance, the nilpotency of \(s_w\), and ghost-number grading, without reference to any specific regulator [0704.2472].

The algebraic viewpoint is foundational because it isolates which anomaly structures are genuinely nontrivial. A plausible implication is that subsequent local-RG constructions should be read as refinements of this cohomological backbone rather than as alternatives to it.

## 2. Local renormalization group and the Osborn form

The local-RG formulation promotes couplings \(g^i\) to spacetime-dependent sources \(g^i(x)\) on a curved background \(g_{\mu\nu}(x)\). The generating functional \(W[g,g^i]\) obeys the global Callan–Symanzik equation
\[
\Bigl(\mu\partial_\mu+\beta^i\partial_{g^i}\Bigr)W=0,
\]
while under a local Weyl rescaling one has an anomalous variation of the form
\[
\delta_\sigma W=\int d^dx\sqrt g\,\sigma(x)\Bigl[\beta^i(g)\,O_i + A(g)\,E_d + C_{IJ}(g)\,I^{IJ}+\cdots\Bigr].
\]
Here \(E_d\) is the Euler density, the \(I^{IJ}\) are Weyl-invariant scalar densities built from curvature, and the omitted terms include total derivatives and source-derivative structures [1308.1096].

The abelian property of Weyl transformations,
\[
[\delta_{\sigma_1},\delta_{\sigma_2}]\,W=0,
\]
implies differential constraints among \(\beta^i(g)\), the Euler coefficient \(A(g)\), the coupling-space tensor \(\chi_{ij}(g)\) multiplying \(\partial g\,\partial g\) terms in the anomaly, and a one-form \(w_i(g)\) associated with total derivatives. In arbitrary even dimension, the central relation takes the universal Osborn form
\[
\partial_i \tilde A(g)=\chi_{ij}(g)\,\beta^j(g)+\partial_{[i}w_{j]}(g)\,\beta^j(g),
\qquad
\tilde A(g)=A(g)+w_i(g)\beta^i(g).
\]
Equivalently,
\[
\partial_iA(g)-\chi_{ij}(g)\beta^j(g)=\partial_i w_j\,\beta^j-\partial_j w_i\,\beta^j.
\]
Contracting with \(\beta^i\) yields
\[
\frac{d}{dt}\,\tilde A(g(t))=-\,\chi_{ij}(g)\,\beta^i\beta^j.
\]
If one can choose a scheme in which \(\chi_{ij}\) is positive-definite, \(\tilde A\) decreases monotonically along RG trajectories; at fixed points \(\beta^i=0\), \(\tilde A\) reduces to the coefficient of the Euler term in the conformal anomaly, reproducing Zamolodchikov’s \(c\)-function in \(d=2\) and the \(a\)-coefficient in \(d=4\) [1308.1096].

This framework extends explicitly to \(d=6\) and suggests a general even-dimensional pattern. In \(d=6\), for example, the consistency conditions again single out the Euler coefficient through a divergence-free Lovelock tensor, producing a candidate \(\tilde a\) with
\[
\frac{d\tilde a}{dt}=-\tfrac16\,H^1_{ij}\,\beta^i\beta^j.
\]
The unresolved issue is positivity of the relevant coupling-space metric. The existence of the consistency condition is exact; monotonicity is conditional [1308.1096].

## 3. Exact gradient flow from a local Wilsonian cutoff

A distinct realization of Weyl consistency conditions introduces a position-dependent UV cutoff \(\Lambda(x)\) that transforms with Weyl weight \(+1\). On a \(d\)-dimensional curved manifold, one may take
\[
\delta_\sigma g_{\mu\nu}=-2\sigma g_{\mu\nu},
\qquad
\delta_\sigma \Lambda(x)=+\sigma \Lambda(x),
\]
and define the Weyl-covariant Laplacian
\[
\square_g^{(\xi)}\equiv \nabla^2-\xi R,
\qquad
\xi=\frac{d-2}{4(d-1)},
\]
together with the dimensionless operator
\[
D_\Lambda\equiv \Lambda^{-2}(x)\,(\nabla^2-\xi R).
\]
Because \(\delta_\sigma D_\Lambda=0\), any scalar function \(F(D_\Lambda)\) is Weyl invariant; with \(F(D_\Lambda)=e^{-D_\Lambda}\), one obtains the Weyl-invariant kinetic action
\[
S_k[ \phi;g,\Lambda]=\frac12\int d^dx\sqrt g\,\phi\,(-\nabla^2+\xi R)\,e^{-D_\Lambda}\phi
\]
for a real scalar of classical weight \((d-2)/2\) [2101.07615].

Local interactions are then written as
\[
S_{\text{int}}=\sum_n\int d^dx\sqrt g\,g_n(A)\,O_n(x),
\]
with \(A\equiv \Lambda\), operator dimension \(d_n\), and anomalous dimensions defined by
\[
A\,\partial_A g_n(A)=\gamma_n(A,g)\,g_n(A).
\]
After introducing dimensionless local variables
\[
P_n(x)=A(x)^{-d_n}g_n(x),
\qquad
A\partial_A P_n=\beta_n(P),
\]
the full action obeys a local Weyl-RG equation,
\[
\Bigl\{\delta_\sigma-\int d^dx\,\sigma(x)\,\beta_n(P)\,\frac{\delta}{\delta P_n(x)}\Bigr\}S=0.
\]
Applying two commuting Weyl variations to the vacuum functional and expanding \(W[g,\Lambda,P]\) to fourth order in derivatives of
\[
X^i\equiv \{\sigma\equiv \ln\Lambda,\;P_n\},
\]
one finds an integrability condition on the antisymmetric part of the \(X\)-sector,
\[
\partial_i B_j-\partial_j B_i=0,
\qquad
B_i\equiv A\partial_A X^i=\{-1,\beta_n\}.
\]
Since \(\partial_i(-1)=0\), the nontrivial content is
\[
\partial_{P_m}\beta_n-\partial_{P_n}\beta_m=0,
\]
so the \(\beta_n\) form an exact gradient. Equivalently,
\[
B_i=\chi_{ij}\,\partial_j D,
\qquad
D(P)=b(P)+\text{const.}
\]
The local cutoff thus appears as an auxiliary coupling with trivial \(\beta\)-function \(\beta_0=-1\), while the physical bare couplings satisfy an exact gradient-flow condition in the enlarged coupling space \(\{\Lambda,P_n\}\) [2101.07615].

This construction is notable because the exact gradient structure is obtained before taking a renormalized infinite-cutoff limit. It therefore separates a genuinely local, regulator-aware statement from the more familiar perturbative statements about renormalized couplings.

## 4. Renormalizable theories and conditions for exactness

For renormalizable theories, one passes to the \(\Lambda\to\infty\) limit by introducing finitely many counterterms, after which the physical couplings \(g_i\) satisfy the conventional RG equations
\[
\mu\,\frac{d}{d\mu}g_i \equiv B_i(g).
\]
The local RG equation for the generating functional remains of the same structural form, but the consistency conditions become initially approximate rather than manifestly exact in the gradient-flow sense. To leading two-derivative order, the relevant relation is schematically
\[
\partial_i B_j-\partial_j B_i
=
G_{ik}(g)\,\partial_k B_j
-
G_{jk}(g)\,\partial_k B_i,
\]
where \(G_{ij}(g)\) is an anomaly-induced metric on coupling space extracted from diagrams with insertions of \(\nabla g(x)\) [2101.07615].

The same analysis also specifies when this approximate structure can be promoted to an exact one. In perturbation theory, \(G_{ij}\) appears symmetric. Under the condition that the loop integrals are symmetric under permutations of like-insertions, Appendix B of the paper shows that
\[
G_{ij}=\partial_i V_j=\partial_j V_i,
\qquad
\partial_i G_{jk}-\partial_j G_{ik}=0,
\]
so the consistency equation can be rewritten as
\[
B_i(g)=G_{ij}(g)\,\partial_j \hat a(g)
\]
for a single scalar function \(\hat a(g)\). In that case, the renormalized beta functions also derive from a gradient once the local cutoff has been introduced as an auxiliary coupling and its associated components have been shown to decouple in the infinite-cutoff limit [2101.07615].

A one-loop illustration is provided by a single massive \(\phi^4\) theory in \(d=4\), with local bare mass \(m(x)\) and quartic coupling \(g(x)\). The analysis computes the coupling-space metric in the \(\{m(x),\Lambda(x)\}\) sector from one-loop diagrams with exponentiated cutoff propagators, determines the corresponding bare potential \(b(m,\Lambda)\), and verifies the resulting integrability relation \(\partial_m\beta_\Lambda-\partial_\Lambda\beta_m=0\) [2101.07615].

One common misconception is that Weyl consistency conditions automatically imply a globally exact gradient flow for renormalized couplings. The Wilsonian-cutoff analysis shows instead that exactness is straightforward for anomalous dimensions of bare couplings in the enlarged space, whereas for renormalizable theories additional symmetry properties of the anomaly-induced metric are required.

## 5. Extensions beyond relativistic even-dimensional flows

In three dimensions, local RG consistency survives even though the anomaly structure differs sharply from the even-dimensional Euler-type pattern. In a power-counting renormalization scheme for \(d=1+2\), the most general local Weyl anomaly is parity-odd and contains exactly two structures,
\[
C_{IJK}(g)\,\epsilon^{\mu\nu\rho}D_\mu g^I D_\nu g^J D_\rho g^K,
\qquad
C_I(g)\,\epsilon^{\mu\nu\rho}f^I_{\mu\nu}D_\rho g^I.
\]
The local RG generator includes scalar sources \(g^I(x)\) and vector sources \(a_\mu^a(x)\), and Wess–Zumino consistency yields both operator-level and anomaly-level constraints. The operator-level condition is the transversality relation
\[
B^I\,P_I^a=0,
\]
while anomaly integrability gives
\[
3\,B^I C_{IJK}+P_J^a(C_{a\,IK})-P_K^a(C_{a\,IJ})=0,
\qquad
B^I C_I=0.
\]
The same framework classifies gauge/equation-of-motion ambiguities, scheme ambiguities, and local-counterterm ambiguities, showing that the anomaly tensors are constrained functionals of the RG vectors rather than independent data [1307.8048].

Non-relativistic theories exhibit an even richer anomaly algebra. For generic \(2+1\)-dimensional theories with anisotropic scaling exponent \(z=2\), the anomaly density decomposes into a four-spatial-derivative sector with \(39\) terms, a two-time-derivative sector with \(6\) terms, and a mixed one-time/two-space sector with \(32\) terms, for a total of \(77\) independent anomaly coefficients. Commutativity of local Weyl rescalings produces approximately \(77\) integrability relations. These can be rearranged into gradient-flow candidates of the form
\[
\beta^\alpha \partial_\alpha C(g)=H_{\alpha\beta}(g)\,\beta^\alpha\beta^\beta,
\]
including explicit candidates such as
\[
\beta^\alpha\partial_\alpha\bigl(4a+\tfrac z2\,c\bigr)=-2\,a_3^{\alpha\beta}\beta^\alpha\beta^\beta.
\]
However, the sign-definiteness of \(H_{\alpha\beta}\) is generally unknown, and known examples with limit cycles show that no universal non-relativistic \(C\)-theorem can hold in all models [1605.02748].

A related non-relativistic formulation uses Newton–Cartan or DLCQ sources. In the simplest \(N_n=0\) sector, the Wess–Zumino system is formally identical to the four-dimensional relativistic case and yields
\[
8\,\partial_i a-\chi^g_{ij}\beta^j=-\,\mathcal L_\beta w_i,
\qquad
\partial_i\tilde a=\tfrac18\,\chi^g_{ij}\beta^j,
\qquad
\frac{d\tilde a}{d\ln\mu}=\tfrac18\,\chi^g_{ij}\beta^i\beta^j.
\]
If \(\chi^g_{ij}\) is positive-definite, \(\tilde a\) is monotonic. By contrast, when the Newton–Cartan time slices satisfy the Frobenius condition \(dn\wedge n=0\), the only scheme-independent anomaly at a fixed point is of type B and no true \(a\)-anomaly survives [1610.00123].

These extensions show that Weyl consistency conditions are not restricted to relativistic even-dimensional Euler anomalies. What persists across settings is the integrability logic; what changes is the anomaly basis and the strength of the resulting monotonicity statement.

## 6. Perturbative applications and conceptual boundaries

In four-dimensional perturbation theory, Weyl consistency conditions impose concrete cross-relations among beta functions of different sectors. In the Standard Model at energies \(E\gg v\), where the theory is classically conformal after neglecting the nominally dimensionful operator \(H^\dagger H\), the couplings multiply marginal operators and one can derive relations of the form
\[
\partial_i\beta^j=\partial_j\beta^i
\]
for appropriately raised indices. At lowest nontrivial order this enforces the “3–2–1” counting: gauge beta functions must be taken to three loops, Yukawa beta functions to two loops, and scalar quartic beta functions to one loop. The rationale is that the one-loop quartic beta function must match specific two-loop Yukawa and three-loop gauge contributions under the consistency relations. In vacuum-stability studies, this changes quantitative outputs: for \(M_H\simeq125.7\,\mathrm{GeV}\) and \(M_t\simeq173.5\,\mathrm{GeV}\), both the Weyl-consistent \(3\!-\!2\!-\!1\) scheme and the conventional \(3\!-\!3\!-\!3\) scheme place the zero crossing of \(\lambda(\mu)\) around \(\mu\sim10^{10}\,\mathrm{GeV}\), while \(\lambda_{\text{eff}}(\mu)\) crosses zero around \(\mu\sim10^{11}\,\mathrm{GeV}\); the \(3\!-\!2\!-\!1\) crossing is shifted slightly lower by \(\sim0.5\) dex, and metastability boundaries in the \((M_t,M_H)\) plane differ at the \(\sim1\!-\!2\,\mathrm{GeV}\) level [1306.3234].

The same consistency machinery also resolves specific high-loop ambiguities. A notable example is the treatment of \(\gamma_5\) in dimensional regularization. For a particular term in the four-loop Standard-Model gauge beta function, the ambiguous coefficient \(c_g\) is related by Weyl consistency to a corresponding three-loop Yukawa coefficient \(c_y\) through
\[
c_g=\frac16\,c_y.
\]
Because the three-loop Yukawa quantity is unambiguous under a semi-naïve \(\gamma_5\) prescription, the relation fixes the four-loop gauge ambiguity and yields \(R=3\), hence
\[
c_g=16+96\,\zeta_3.
\]
This is a scheme-independent use of the four-dimensional Osborn equation rather than a diagram-by-diagram resolution of the Dirac-algebra problem [1901.02749].

Weyl consistency conditions also delimit which curvature couplings are compatible with quantum Weyl symmetry. In four dimensions, rewriting the trace anomaly as
\[
\delta_\sigma W_{\rm eff}
=
-\int d^4x\sqrt g\,\sigma\bigl(aE_4+cW^2+bR^2\bigr),
\]
the commutativity condition forces \(b=0\), while the \(\Box R\) term is removable by a local counterterm \(\int R^2\). More broadly, for all unitary theories in spacetime dimension \(d\le 10\), it has been argued that conformal invariance in flat spacetime implies Weyl invariance in a general curved background, up to the usual anomaly terms proportional to the identity operator; possible anomalous operator transformations are severely constrained, and for sufficiently low-dimensional operators the canonical Weyl transformation law is the only consistent one [1702.07079].

The conceptual boundary is therefore twofold. First, consistency conditions are exact integrability statements, but their conversion into monotonic theorems depends on positivity properties of coupling-space metrics. Second, they constrain rather than replace anomaly classification: the type-A/type-B distinction, the possibility of scheme redefinitions, and the role of auxiliary structures such as local cutoffs remain essential to their interpretation.

Source: https://www.emergentmind.com/topics/weyl-consistency-conditions