A finite prediction can be expressed through different parameter coordinates

A momentum cutoff restricts integration momenta, while dimensional regularization analytically continues loop integrals away from four dimensions. Each makes otherwise ill-defined expressions manageable, but neither is by itself a rule for assigning measured values to model parameters. A renormalization condition fixes finite parts by specifying an amplitude, pole, or subtraction prescription. The regulator can then be removed where a continuum limit exists, or retained within an effective theory with a stated cutoff and energy domain.

Use natural units ħ=c=1 and a scalar theory in d=4−ε dimensions. A quartic coupling has mass dimension ε, so a reference energy μ supplies its required dimension while the renormalized λ is dimensionless. This μ is not automatically a physical ultraviolet cutoff. In dimensional regularization it remains after the regulator ε is removed. A divergent bare parameter in that limit is a feature of the chosen parameterization; it does not mean that an observable mass or scattering probability must diverge.

Start with a bare kinetic term, mass term, and quartic interaction. After replacing the field, the kinetic coefficient is Zφ, the mass coefficient is ZφZm times m squared, and the quartic coefficient contains Zφ squared times Zλ. Split each into a chosen renormalized term and a counterterm. This algebra explains why a vertex renormalization factor and a coupling renormalization factor need not be the same: changing the field normalization also changes every vertex containing that field.

Counterterms are local interactions constrained by symmetries and the chosen expansion. They are inserted into diagrams at the appropriate perturbative order. They are not independent adjustable corrections for each external momentum or each experiment. Once enough renormalization conditions fix the available parameters, the remaining amplitudes are predictions. In an effective field theory, more local operators appear at higher order in the energy-to-heavy-scale ratio, but a finite set suffices at a specified accuracy when the power counting is controlled.

A sharp-cutoff scalar mass correction can contain a term proportional to the cutoff squared. Its coefficient is model and regulator dependent; dimensional regularization organizes the same matching problem differently and does not require displaying that particular power divergence. The illustrative relation M²=mB²+cλΛ² makes the bookkeeping transparent: if M is fixed while Λ changes, mB² must change. This statement alone does not resolve naturalness questions or establish a UV completion. It only separates an unphysical input from a fitted low-energy output.

The function lab illustrates the cutoff-square contribution. Reset and set both controls to a=0.05 and x=1, predicting y=0.05. Reset and set a=0.05 and x=2, predicting y=0.20. Assign x to a normalized cutoff and y to a normalized loop mass correction. Doubling the cutoff quadruples this illustrative contribution. The compensating bare input must be calculated separately; the lab has no field theory, regulator removal, or measured mass. Each setting is an independent calculation rather than a physical evolution of the sample.