Ward Identities and Anomalous Obstructions

A Ward identity can be imposed on renormalized time-ordered products precisely when every local breaking term is removable by an allowed finite renormalization. Causal perturbation theory turns that statement into a local cohomology problem: causality and lower-order identities force the first possible breaking onto the total diagonal, consistency makes it a cocycle, and a counterterm removes it only if its cohomology class is trivial. A nontrivial class is an anomaly, not a poor choice of subtraction.

Required background. Causal Wick expansion and time-ordered products supplies the renormalized multilinear maps on which the identity acts. Stückelberg–Petermann renormalization freedom identifies the admissible local changes of prescription. Helpful background. Analyticity, CPT, and spin–statistics separates axiomatic symmetry theorems from perturbative Ward conditions. Counterexamples and hypothesis stress tests supplies the logic of the obstruction test. Interacting stress-tensor Ward identities treats the locally covariant gravitational analogue.

Local breakings and the consistency differential

Let $s$ be the infinitesimal differential encoding a classical symmetry on local functionals; for a gauge symmetry it is the BRST differential, with $s^2=0$ on the chosen off-shell complex. Write $T(e^F)$ for the formal generating functional of renormalized time-ordered products. The desired quantum identity has the schematic form

$$s,T(e^F)=\frac{i}{\hbar} T\!\left((sF)e^F\right),$$

with the field-equation, antifield, and contact terms included in $sF$ when the precise Master Ward identity requires them. This formula is not asserted on arbitrary operator products: its domain is the algebra of compactly supported local functionals, extended coefficientwise to formal power series, and its products are the already-renormalized $T_n$.

Suppose the identity has been fulfilled through total order $n-1$ in the fields, coupling, and loop filtration. Causal factorization makes the order-$n$ breaking vanish away from coincident configurations, because every causally ordered region there reduces to lower-order products. Hence the anomaly is a local diagonal distribution,

$$\mathcal A_n(F_1,\ldots,F_n) =\sum_{|a|\le N} C_a(jF_1,\ldots,jF_n)\,\partial^a\delta_D,$$

where $C_a$ depends on finite jets and $N$ is bounded by scaling degree. Covariance, ghost number, dimension, parity, and any retained discrete symmetries further restrict the coefficients. Applying $s$ once more and using $s^2=0$ yields the Wess–Zumino consistency condition $s\mathcal A_n=0$, modulo total derivatives and the equations already built into the complex. The original derivation of the consistency condition is Wess and Zumino 1971, pp. 95–97; its causal-renormalization realization as an anomalous Master Ward identity and renormalization cocycle is made precise in Brunetti, Dütsch, Fredenhagen, and Rejzner 2023, §10, Theorem 10.3, pp. 515–521.

The theorem is directional. Locality and the lower-order identities imply that the breaking represents a class in the appropriate local cohomology. They do not imply that the class vanishes. If $\mathcal A_n=sB_n$ for an admissible local $B_n$, changing the prescription by the corresponding Stückelberg–Petermann element cancels the breaking at that order, after which the induction continues. If $[\mathcal A_n]\ne0$, no local counterterm respecting all the declared conditions can restore the identity. A nonlocal functional might cancel a formula pointwise, but it leaves the renormalization class and destroys the locality statement that made the theorem meaningful.

First application: a four-dimensional chiral current

Consider a massless left-handed fermion of $U(1)$ charge $q$ coupled to a compactly supported external connection $A$. The classical gauge variation is $sA=\mathrm dc$, $s\psi=iqc\psi$, and $sc=0$. At one loop the triangle contribution can produce a consistent anomaly of the form

$$\mathcal A(c,A)=C_q\int_M c\,F\wedge F, \qquad F=\mathrm dA,$$

where $C_q$ is fixed by the charge and by the normalization chosen for the current and differential forms. Keeping $C_q$ symbolic avoids confusing the consistent-current coefficient with the covariant-current coefficient. Since $sF=0$, the displayed functional is $s$-closed. The descent calculation asks whether it is $s$-exact modulo a total derivative in the space of local, gauge-covariant counterterms. For an anomalous one-fermion representation it is not. Thus the one-loop localized Ward–Takahashi identity cannot be imposed together with gauge covariance and the consistency relation.

A Bardeen–Zumino polynomial can shift the anomaly between consistent and covariant currents. That changes the representative, not the nonzero class. In an anomaly-free multiplet the charge-weighted sum of classes can cancel; the conclusion then concerns that full representation, not each fermion separately. Likewise, a global-current anomaly is not automatically a logical inconsistency, whereas an uncancelled gauge anomaly obstructs the gauge construction.

The proof mechanism is independently checkable without evaluating the triangle integral. Ghost number requires one factor of $c$; four-dimensional form degree requires a four-form; Abelian gauge invariance leaves $F\wedge F$ as the parity-odd candidate at this dimension. Finally $s(cF\wedge F)=0$. These checks establish consistency and the possible class, while the loop calculation fixes $C_q$.

Adversarial test: cancelling only one identity

Assume a proposed local counterterm makes the divergence of one chosen current vanish. Recompute its gauge variation and the commutator of two infinitesimal transformations. If the modified current is not gauge covariant, or if the new breaking fails

$$\delta_{\alpha_1}\mathcal A(\alpha_2) -\delta_{\alpha_2}\mathcal A(\alpha_1) =\mathcal A([\alpha_1,\alpha_2]),$$

the proposal has merely moved the anomaly and is inadmissible. This is the decisive false-converse boundary: satisfying one divergence equation does not prove simultaneous fulfillment of the Ward system. The anomaly map itself obeys a cocycle law under finite transformations Brunetti, Dütsch, Fredenhagen, and Rejzner 2023, §5, Proposition 5.3, pp. 493–496.

Exercises

1. Abelian consistency. Show directly that $\mathcal A(c,A)=C\int cF\wedge F$ is BRST closed.

Solution

Use $sc=0$ and $sF=\mathrm d(sA)=\mathrm d^2c=0$. The graded Leibniz rule therefore gives $s(cF\wedge F)=0$. Closure is necessary for an anomaly; it does not establish exactness or nonexactness.

2. A removable breaking. Suppose a first-order breaking is $\mathcal A=sB$ with $B$ local and compatible with the scaling bound. Explain why the replacement $T\mapsto T'$ generated by the finite counterterm $-B$ removes it to that order.

Solution

The change of time-ordering contributes $-sB$ to the Ward variation at the first order where the two prescriptions differ. It cancels $\mathcal A=sB$. Locality and the scaling bound ensure that this change is an allowed Stückelberg–Petermann transformation; higher-order terms are then handled inductively.

References

• Brunetti, Romeo, Michael Dütsch, Klaus Fredenhagen, and Kasia Rejzner. “The Unitary Master Ward Identity: Time Slice Axiom, Noether’s Theorem and Anomalies.” Annales Henri Poincaré 24 (2023): 469–539. DOI; Open PDF.
• Wess, Julius, and Bruno Zumino. “Consequences of Anomalous Ward Identities.” Physics Letters B 37, no. 1 (1971): 95–97. DOI.