 Solution

In units $\hslash =c=1$, a dimensionless action fixes $[\mathcal{L}]=4$. Canonical kinetic terms then give $[\varphi ]=[A_{\mu }]=1$, $[\psi ]=3/2$ and $[{\partial }_{\mu }]=1$. If an interaction operator has dimension $d_v$, its coefficient has mass dimension $[g_v]=4-d_v$. Thus scalar quartic, Yukawa and gauge couplings are dimensionless; a scalar cubic coupling has dimension one, a four-fermion coupling has dimension minus two, and the Weinberg coefficient has dimension minus one. These are engineering dimensions; anomalous scaling is an additional quantum effect.

For a connected graph with $L$ independent loop momenta, $I_B$ boson lines, $I_F$ fermion lines and $r_v$ derivatives at vertex $v$, ordinary covariant-gauge ultraviolet counting gives the superficial degree of divergenceThe second form follows from $L=I_B+I_F-V+1$ and counting field ends at vertices. It presumes the canonical large-momentum propagators; unitary-gauge vector propagators can obscure the cancellations of a renormalizable gauge theory. A nonnegative $D$ signals a possible overall ultraviolet divergence, not proof of one. Symmetries can cancel it, while a graph with $D<0$ can still contain divergent subgraphs.

With only operators of dimension at most four, increasing the vertex count does not require counterterms of unbounded dimension. Subject to the necessary symmetry identities and absence of gauge anomalies, scalar quartic theory, Yukawa theories and the Standard Model are perturbatively renormalizable when all required interactions are included. In particular, coupling a scalar to other fields can require a scalar quartic counterterm even if its coefficient was initially set to zero. The Higgs mechanism does not destroy renormalizability of the underlying anomaly-free gauge theory.

By contrast, repeated insertions of the dimension-six Fermi interaction or a dimension-five operator allow increasingly high-dimensional counterterms. Such terms are useful in an effective field theory with a stated expansion in energy over a high scale; nonrenormalizable does not mean unpredictive at a fixed order. A scalar ${\varphi }^6$ interaction and perturbative Einstein gravity in four dimensions provide other examples. Renormalizability also does not assert small coupling at every scale or a complete ultraviolet theory.