Microlocal Renormalization Ambiguities and Their Classification

Two locally covariant renormalization prescriptions satisfying the same causal, microlocal, scaling, symmetry, field-equation, and analytic-dependence axioms differ by finite local maps. In a four-dimensional scalar theory these maps are precisely redefinitions of composite operators and the finite shifts of matter and gravitational couplings allowed by dimension and covariance—not arbitrary nonlocal functions.

Required background. Time-ordered products and the renormalized stress tensor supplies the local prescriptions; renormalization freedom and the Stückelberg–Petermann group supplies their composition law.

Helpful background. Equivalence, uniqueness, and comparison notions fixes the equivalence claim; source authority, dated status, and specialist review supplies source discipline; entropy counterterms and renormalization ambiguities, curvature counterterms and composite-operator mixing, running couplings and curvature couplings, curved-space renormalization schemes, and generalized entropy and UV renormalization provide bounded applications.

Local equivalence of prescriptions

Let $S_T(F)$ be the formal local $S$-matrix constructed from time-ordered products $T$, and let $\widetilde T$ be a second prescription satisfying the same axioms. Then there is an invertible formal local map $Z$ with $Z(0)=0$ and first derivative the identity such that

$$S_{\widetilde T}(F)=S_T(Z(F)).$$

The map is local in the additivity sense, covariant under spacetime embeddings, compatible with the involution, and analytic as a formal power series. Composition of prescription changes gives the Stückelberg–Petermann renormalization group. This is an equivalence of formal perturbative constructions after coupling and field redefinitions; it is not equality of numerical running couplings at fixed coordinates on theory space.

For Wick powers, the component form is the finite triangular relation classified by Hollands and Wald 2001, Theorem 5.1, pp. 30–34. For time-ordered products, ambiguities are distributions supported on total diagonals, with derivatives of delta functions multiplying local covariant curvature polynomials; Hollands and Wald 2001, Theorem 5.2, pp. 36–40. Causal factorization confines the difference to coincident configurations, while scaling bounds make the set finite at each order.

Four-dimensional φ⁴ through quartic order

For a scalar interaction $F=\int f\lambda\phi^4/4!$, power counting and covariance permit finite shifts of the local Lagrangian density by

$$\delta Z\,(\nabla\phi)^2 +\delta m^2\phi^2 +\delta\xi R\phi^2 +\delta\lambda\phi^4,$$

together with identity terms

$$\delta\Lambda+\delta\kappa R +aR^2+bR_{ab}R^{ab}+cR_{abcd}R^{abcd}+d\Box R.$$

After accounting for the four-dimensional Euler density and boundary terms, one may choose a smaller equivalent curvature basis. At composite-operator level, $\phi^2$ mixes with $m^2$ and $R$, while $\phi^4$ mixes with $m^2\phi^2$, $R\phi^2$, and dimension-four c-number curvature polynomials. Hollands and Wald display the wavefunction, mass, curvature-coupling, quartic, and gravitational shifts in Hollands and Wald 2001, Eq. (81), pp. 40–41.

Under a change of subtraction scale, $Z$ becomes scale dependent and induces beta functions and anomalous dimensions. A different finite scheme changes their coordinate expressions beyond universal leading data, while observables and properly transformed RG-invariant relations agree. This is the precise handoff to scheme transformations and RG invariants.

An independent dimension check verifies every listed term has Lagrangian dimension four. Locality can be checked by varying a coupling supported in one region: the change in $Z(F)$ at a point depends only on the germ of fields, couplings, and background there, not on remote geometry.

The linearization of $Z$ around an interaction makes the equivalence operational. Its first nonlinear terms insert local counterterm vertices into retarded or time-ordered products; composing two changes reproduces the same result as first redefining one prescription and then the next. Because $Z'(0)=\mathrm{id}$, the free theory and first-order identification of fields are fixed, while higher derivatives encode finite renormalizations. At any fixed order in $\hbar$ and the couplings, scaling degree bounds allow only finitely many derivatives and curvature monomials. The group is therefore infinite as a formal series but finite-dimensional order by order after fixing the field content and power-counting degree.

Adversarial nonlocal counterterms

Propose $R\Box^{-1}R$ or $R^{-1}\phi^4$ as a finite ambiguity. The first depends on a Green operator and hence on global boundary/support choices; it violates polynomial locality and causal additivity. The second is singular at backgrounds with $R=0$, is not polynomial or analytic in the background, and violates the allowed scaling dependence. Neither belongs to the classification, even if it has a superficially correct total dimension.

The converse is also limited. Every allowed local term is a possible ambiguity before normalization conditions, but a particular theory or symmetry can set coefficients to zero. The classification does not prove perturbative convergence, nonperturbative equivalence, or that a finite scheme transformation preserves a truncation when higher orders are discarded.

Exercises

1. Operator mixing. List the dimension-two scalars that can mix with $\phi^2$ in four dimensions.

Solution

Under the polynomial local-covariant assumptions they are $m^2\mathbf1$ and $R\mathbf1$. A coordinate function or inverse-curvature expression is excluded.

2. Locality test. Why is $R(x)\int R(y)^2d\mathrm{vol}_y$ not an allowed coefficient of a local counterterm at $x$?

Solution

Changing the metric far from $x$ changes the integral and hence the coefficient at $x$. The term depends on global rather than germ data, violating locality and covariance of the prescription.

References

• Hollands, Stefan, and Robert M. Wald. “Local Wick Polynomials and Time Ordered Products of Quantum Fields in Curved Spacetime.” Communications in Mathematical Physics 223 (2001): 289–326. DOI. Open PDF.
• Hollands, Stefan, and Robert M. Wald. “On the Renormalization Group in Curved Spacetime.” Communications in Mathematical Physics 237 (2003): 123–160. DOI. Open PDF.