Regularization and renormalization

Consider one real scalar in four dimensions with Euclidean interaction $\lambda\phi^4/4!$ and positive reference mass $m$. A spherical Euclidean momentum cutoff $\Lambda$ turns the tadpole into an ordinary finite integral. This regulator is convenient for seeing powers of momentum, although a hard cutoff does not automatically preserve every symmetry in gauge theories. Regularization is the choice of a well-defined intermediate problem; it is not yet a prediction about the physical world.

Use the four-dimensional angular volume $2\pi^2$ and then substitute $u=k^2$. The radial measure becomes $u\,du/(16\pi^2)$, so the integral contains both a quadratic term and a mass-dependent logarithm. The answer has mass dimension two, exactly what a correction to a mass squared requires. Dimensional regularization packages the same ultraviolet sensitivity differently; the absence of an explicit $\Lambda^2$ there does not mean that matching to heavy physics has disappeared.

In the $\lambda\phi^4/4!$ convention, the one-loop self-energy tadpole carries a factor $\lambda/2$. Introduce a regulator-dependent bare mass $m_0^2(\Lambda)$ and choose it so that the pole of the continued two-point function has a specified measured mass $m_R$. At this order the tadpole is momentum independent, so replacing its internal mass by $m_R$ changes only higher orders. A counterterm is the adjustment required by this condition, not an extra independently adjustable force.

Now vary the cutoff while keeping the mass measurement fixed. The bare mass must move in the opposite direction to the tadpole. Reporting only the loop correction without its associated parameter prescription would falsely make the measured mass depend on a bookkeeping choice. Other observables remain predictions only after the other necessary parameters, including the coupling and field normalization, have been fixed. A single mass subtraction cannot repair every diagram in every possible theory.

Two subtraction schemes can assign different numerical values to a renormalized coupling at the same named scale. They agree on observables when their parameters are matched and calculations are carried to the same accuracy. A finite-order result normally retains a residual scale dependence of the first omitted order. That dependence is useful diagnostic information, but varying a scale is not a proof that every missing correction lies inside the resulting band.

There are also physical limits to the continuum story. A low-energy effective theory can keep a finite ultraviolet cutoff and predict long-distance experiments with controlled truncation errors. Conversely, perturbative renormalizability alone does not prove a nonperturbative continuum theory exists at every energy. Treat four-dimensional scalar theory here as a perturbative low-energy model, and distinguish that controlled use from a claim to have constructed an interacting ultraviolet completion.