Renormalization Freedom and the Stückelberg–Petermann Group

The main theorem of perturbative renormalization states that two local $S$-matrices satisfying the same causal, starting-element, field-locality, and field-independence axioms differ by a unique local analytic redefinition $Z$ of the interaction: $\widehat S=S\circ Z$. Such maps form the Stückelberg–Petermann group. They encode finite local counterterms and field redefinitions, not Wilsonian integration of momentum shells and not arbitrary nonlocal changes of amplitudes.

Required background. Causal Wick expansion and renormalized time ordering provides the prescriptions being compared; scaling-degree extension makes their difference local and finite at each order. Helpful background. Equivalence and comparison notions clarify what is being identified, while rigorous RG as a dynamical system is a distinct coarse-graining construction.

The local renormalization group

Let $\mathfrak F_{\mathrm{loc}}[[\hbar]]$ be formal local interactions. An admissible finite renormalization is an analytic formal map

$$Z:\mathfrak F_{\mathrm{loc}}[[\hbar]] \longrightarrow\mathfrak F_{\mathrm{loc}}[[\hbar]]$$

with

$$Z(0)=0, \qquad Z^{(1)}(0)=\operatorname{id}, \qquad Z=\operatorname{id}+O(\hbar),$$

and the locality/additivity law

$$Z(A+B+C)=Z(A+B)-Z(B)+Z(B+C)$$

when $\operatorname{supp}A\cap\operatorname{supp}C=\varnothing$. Field independence and covariance imply $\operatorname{supp}Z(V)\subseteq\operatorname{supp}V$ and restrict the Taylor coefficients to local covariant differential polynomials. Reality or unitarity imposes the corresponding $*$ relation.

If $S$ and $\widehat S$ satisfy the same defining axioms, there is a unique $Z$ with

$$\widehat S=S\circ Z.$$

Conversely, composing an admissible $S$ with such a $Z$ gives another admissible prescription. Composition and formal inversion make these maps a group. The exact statement and inductive proof are Brunetti, Dütsch, and Fredenhagen 2009, Theorem 4.1 and equations (4.11)–(4.19), pp. 1561–1563. The proof compares the first order where two prescriptions differ; causal factorization forces that difference onto the diagonal, and field independence turns it into the next local Taylor coefficient of $Z$.

At Taylor order $n$, the new difference is a symmetric map

$$Z^{(n)}(0): \mathfrak F_{\mathrm{loc}}^{\otimes n}\longrightarrow \mathfrak F_{\mathrm{loc}}.$$

If the arguments have mutually separated supports, repeated additivity expresses its value in terms of lower Taylor orders, already fixed by induction. The genuinely new part is therefore supported where all arguments meet. Peetre-type locality and field independence turn that diagonal-supported map into a finite-order differential polynomial in jets of the fields, couplings, and background data. Scaling degree bounds the jet order. This is why the theorem yields local counterterms rather than merely assuming them.

Second-order freedom in four-dimensional φ⁴

Let

$$V_g=\int g(x)\left( \frac{\lambda}{4!}\phi^4+ \frac12m^2\phi^2+ \frac12\partial_\mu\phi\partial^\mu\phi \right)\mathrm d^4x$$

serve as a coordinate chart on local interactions. Through second order in $\lambda$, locality, Poincaré covariance, field parity, and power counting allow $Z(V_g)-V_g$ to contain local terms of the form

$$\lambda^2\int\left( a\,g^2\phi^4+b\,g^2m^2\phi^2 +c\,g^2\partial_\mu\phi\partial^\mu\phi +d\,g^2m^4 +e\,g\Box g\,\phi^2 \right)\mathrm d^4x,$$

up to integrations by parts and the precise normalization basis. For a coupling constant region where $g=1$, these coefficients are finite coupling, mass, field-strength, and vacuum-energy redefinitions. Terms involving derivatives of $g$ record switching-boundary freedom and vanish in the constant interior.

Substituting the displayed $Z$ into $\widehat S(V)=S(Z(V))$ and expanding to $O(\lambda^2)$ matches the difference of the two Epstein–Glaser diagonal extensions. This gives the local comparison required before discussing scheme transformations and RG invariants. Which combinations are observable or invariant is a separate physical question.

An independent group check composes two maps $Z_1=\operatorname{id}+z_1+O(\lambda^3)$ and $Z_2=\operatorname{id}+z_2+O(\lambda^3)$. At second order, $Z_1\circ Z_2=\operatorname{id}+z_1+z_2+O(\lambda^3)$, so finite counterterm coefficients add. Higher orders contain the expected substitution terms and remain local because composition preserves the additivity law.

A scale transformation gives a useful but one-way connection to running couplings. Conjugating a fixed prescription by spacetime dilation generally produces another prescription, hence a $Z_\rho$ satisfying a cocycle relation rather than necessarily a one-parameter subgroup. Differentiating at $\rho=1$ defines the algebraic beta functional. This construction is state independent and local, but extracting a numerical beta function requires choosing interaction coordinates and quotienting redundant field redefinitions.

Nonlocal maps are excluded

Adversarial test. Propose

$$Z_{\mathrm{bad}}(V)=V+\int\mathrm d^4x\,\mathrm d^4y\, K(x-y)g(x)g(y)\phi(x)^2\phi(y)^2$$

with $K$ not supported at the origin. Choose disjoint $A$ and $C$ whose supports are joined by $K$. The cross term remains in $Z_{\mathrm{bad}}(A+B+C)$ but is absent from the local additivity combination. Hence $Z_{\mathrm{bad}}$ is not in the Stückelberg–Petermann group. Momentum dependence by itself is not forbidden—derivatives of delta become polynomials in momentum—but a genuinely nonlocal kernel is.

This also blocks a false identification with Wilsonian RG. A Wilsonian effective action may be nonlocal at finite cutoff and relates different resolution scales; $Z$ compares local renormalization prescriptions for the same coefficientwise causal theory.

Nor does the theorem say all coefficients in the displayed basis are independent. Integrations by parts, field equations, symmetry identities, and a chosen parameterization can relate them. The group theorem classifies the form and support of the freedom; a physical renormalization scheme fixes coordinates within it.

Exercises

1. Support from additivity. Let $V$ vanish in a neighborhood of $x$. Use a decomposition into separated pieces to explain why an admissible $Z(V)$ also vanishes near $x$.

Solution

Split $V$ by a partition of unity into pieces supported away from a small neighborhood of $x$. Repeated local additivity expresses $Z(V)$ using only those supports. No Taylor coefficient can create support at $x$, giving $\operatorname{supp}Z(V)\subseteq\operatorname{supp}V$.

2. Inverse to second order. If $Z=\operatorname{id}+z_2+O(\lambda^3)$, find its formal inverse.

Solution

$Z^{-1}=\operatorname{id}-z_2+O(\lambda^3)$, because substitution changes $z_2$ only at higher order and the two second-order terms cancel. Locality of $z_2$ is preserved.

References

• Brunetti, Romeo, Michael Dütsch, and Klaus Fredenhagen. “Perturbative Algebraic Quantum Field Theory and the Renormalization Groups.” Advances in Theoretical and Mathematical Physics 13 (2009): 1541–1599. DOI; Open PDF.
• Epstein, Henri, and Vladimir Glaser. “The Role of Locality in Perturbation Theory.” Annales de l’Institut Henri Poincaré A 19 (1973): 211–295. EuDML record and PDF.