Wilsonian RG and Effective Field Theory

The renormalization group as Callan–Symanzik is a statement about scheme-dependence. Wilson's formulation is deeper and more physical: build the theory with a cutoff $\Lambda$, then systematically integrate out high-momentum modes to obtain an effective description at lower energies. This reconceives renormalization entirely — divergences become the natural sensitivity of low-energy physics to a cutoff, "non-renormalizable" theories become respectable effective field theories, and the classification of operators by relevance falls out geometrically.

Conventions: $\hslash =c=1$, ${\eta }_{\mu \nu }=\mathrm{diag}(+1,-1,-1,-1)$.

Integrating out shells

Start with a path integral defined with a hard momentum cutoff $\Lambda$. Split the field into low- and high-momentum parts relative to a lower scale $b\Lambda$ ($b<1$):

$$\varphi ={\varphi }_{<}(|p|<b\Lambda )+{\varphi }_{>}(b\Lambda <|p|<\Lambda ),$$

and do the integral over ${\varphi }_{>}$ exactly (in principle):

$$e^{i{S}_{\text{eff}}[{\varphi }_{<}]}=\int \mathcal{D}{\varphi }_{>}{e}^{iS[{\varphi }_{<}+{\varphi }_{>}]}.$$

The result is an effective action $S_{\text{eff}}[{\varphi }_{<}]$ for the remaining low-energy modes, with a lower cutoff $b\Lambda$. Integrating out the high modes does not throw information away — it repackages it into shifted (and newly generated) couplings in $S_{\text{eff}}$.

The RG flow of couplings

Rescale momenta back to the original cutoff to compare theories at different scales. The net effect is a flow in the space of all possible couplings $\{g_i\}$:

$$\{g_i\}{|}_{\Lambda }⟶\{g_i^{\prime }\}{|}_{b\Lambda }.$$

This is the RG as a flow on theory space — the modern picture behind the Callan–Symanzik $\beta$-functions, which are the infinitesimal version $\mu d{g}_i/d\mu ={\beta }_i(g)$. Crucially, integrating out modes generates all operators consistent with the symmetries, even those absent from the starting Lagrangian; the RG flow lives in the infinite-dimensional space of all such operators.

Relevant, marginal, and irrelevant operators

Linearizing the flow near a fixed point classifies operators by how their couplings scale under $\Lambda \to b\Lambda$ — equivalently by the mass dimension of the operator $O_i$ (with coupling dimension $[g_i]=4-[O_i]$ in $d=4$):

The name is the physical content: as you flow to low energy, irrelevant operators die off as powers of $E/\Lambda$, relevant ones dominate, and marginal ones run logarithmically. This is exactly the power-counting classification of renormalizability — but now with a physical reason: a renormalizable theory is just what you automatically flow to at low energy, because the non-renormalizable (irrelevant) operators are suppressed.

Renormalizability reinterpreted

Wilson's viewpoint dissolves the old worry about "infinities":

• Every QFT is an effective theory with an implicit cutoff $\Lambda$ where new physics enters. There is no need for it to make sense to arbitrarily high energy.
• Renormalizable = the generic low-energy limit. Whatever the (unknown) high-energy theory, at energies $E\ll \Lambda$ only the relevant and marginal operators survive — the renormalizable ones. This explains why the Standard Model is renormalizable: not by design, but because that is all that survives to low energy.
• Non-renormalizable operators are predictions, not failures. They are present but suppressed by $(E/\Lambda {)}^n$, and measuring them probes the scale $\Lambda$ of new physics — the logic of effective field theory and SMEFT.

Matching and the decoupling theorem

When a heavy particle of mass $M$ is integrated out, the Appelquist–Carazzone decoupling theorem guarantees its effects at $E\ll M$ reduce to (i) renormalizations of the light couplings and (ii) irrelevant operators suppressed by $1/M$. Building the low-energy effective theory proceeds by matching: compute the same observable in the full and effective theories at the scale $\mu \sim M$ and fix the effective couplings so they agree. Below $M$ the effective couplings are run down with their own RG equations. This match-and-run procedure is the backbone of precision calculations spanning widely separated scales — from Fermi theory below $m_W$ to heavy-quark and chiral effective theories.

Triviality and the continuum limit

Read backward — flowing up in energy — the Wilsonian picture explains triviality. For a theory with $\beta >0$ (like ${\varphi }^4$ or QED), holding the low-energy coupling fixed while sending $\Lambda \to \infty$ forces the coupling to zero: no interacting continuum limit exists, so the theory must be regarded as effective, with a finite cutoff. Theories with a UV fixed point (asymptotically free or safe) instead admit a genuine continuum limit.

Summary

• Wilsonian RG: integrate out high-momentum shells → effective action with a lower cutoff → a flow on the space of all couplings.
• Operators are relevant / marginal / irrelevant by dimension; irrelevant ones die in the IR — reinterpreting renormalizability.
• Every QFT is an EFT; renormalizable is the generic low-energy limit; non-renormalizable operators probe the cutoff scale.
• Matching + decoupling builds multi-scale EFTs; triviality forbids some continuum limits.

Where this leads

• The EFT toolkit in practice: effective field theory.
• The differential RG and $\beta$-functions: the renormalization group.
• Fixed points and scale invariance: conformal field theory.
• Nonperturbative RG on a lattice: lattice field theory.

References

• Wilson & Kogut, Phys. Rep. 12, 75 (1974).
• Polchinski, Nucl. Phys. B 231, 269 (1984).
• Peskin & Schroeder, An Introduction to Quantum Field Theory, Ch. 12.1.
• Weinberg, The Quantum Theory of Fields, Vol. 1, Ch. 12.4–12.5.