Answers are grounded in this paper and cite their sources.
Renormalized observables must not depend on the regularization prescription, but this is usually asserted rather than quantified. For one-loop QED vacuum polarization, we define admissible regulators by three checkable conditions (finiteness, shift invariance, dominated convergence) and prove they yield the same on-shell renormalized function and running coupling, equal to a regulator-free dispersion integral (verified to 24 digits). For regulators acting on the transverse scalar function, shift invariance holds by construction; the non-trivial content appears at the tensor level for dimensional regularization and Pauli–Villars. For smooth proper-time cutoffs with profile 𝑤, the finite-cutoff error is −𝛼𝑄2𝑐𝑤/(15𝜋𝛬2), with 𝑐𝑤 a single moment of the profile; extrapolation or improved Pauli–Villars weights removes it and improves the accuracy by nearly six orders of magnitude. A sharp momentum cutoff and a Gaussian cutoff applied after Feynman parametrization both leave a longitudinal term with leading coefficient −(𝛼/2𝜋)𝛬2, which we compute exactly. Independent checks include a direct four-dimensional integration, the one-loop leptonic running at the Z mass (3×10−8 with masses tuned to the tabulated values), muon 𝑔−2, muonic hydrogen, and OPAL data. A diagram-level two-loop computation with two Pauli–Villars profiles and a Gaussian proper-time cutoff on the photon reproduces the Källén–Sabry function to within 4×10−4; dimensional regularization at two loops was not run. Scripts are provided as supplementary material.
No takes yet. Share an insight, caveat, or question.
Jagadeesh Selvaraj (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: