Renormalization and Counterterms

Regularization makes the divergent loop integrals finite by introducing a parameter (a cutoff $\Lambda$, or the pole $1/ϵ$ of dimensional regularization). Renormalization is the physics: the divergences are absorbed into a redefinition of a finite number of parameters — masses, couplings, and field normalizations — leaving predictions that are finite and regulator-independent. This page sets up bare vs. renormalized quantities, counterterms, the one-loop program, and the classification of theories by renormalizability. It replaces the overview in preliminaries § Regularization and Renormalization.

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

Bare vs. renormalized quantities

The parameters in the Lagrangian — the "bare" mass $m_0$, coupling ${\lambda }_0$, and field ${\varphi }_0$ — are not the measured quantities. They are formal, cutoff-dependent inputs. The physical (renormalized) mass, coupling, and field are defined by measurement (or by a renormalization condition). The two are related by divergent renormalization constants $Z$:

$${\varphi }_0=\sqrt{Z_{\varphi }}\varphi ,m_0^2=Z_m{m}^2/Z_{\varphi },{\lambda }_0=Z_{\lambda }\lambda Z_{\varphi }^{-2},$$

where $Z_{\varphi }$ is the field-strength renormalization (the same $Z$ that appears in the LSZ formula). The key idea of renormalization: the bare parameters are allowed to be cutoff-dependent (even divergent) in exactly the way needed to keep the renormalized parameters finite and equal to their measured values.

Renormalized perturbation theory and counterterms

Rewrite the Lagrangian in terms of renormalized quantities, splitting each bare parameter into a finite renormalized part plus a divergent counterterm $\delta =Z-1$:

$$\mathcal{L}=\underset{\text{renormalized}}{\underbrace{\frac{1}{2}(\partial \varphi {)}^2-\frac{1}{2}{m}^2{\varphi }^2-\frac{\lambda }{4!}{\varphi }^4}}+\underset{\text{counterterms}}{\underbrace{\frac{1}{2}{\delta }_{\varphi }(\partial \varphi {)}^2-\frac{1}{2}{\delta }_m{\varphi }^2-\frac{{\delta }_{\lambda }}{4!}{\varphi }^4}}.$$

The counterterms generate new Feynman-rule vertices ($\times$, "counterterm insertions") whose divergent coefficients are fixed order by order to cancel the loop divergences. Concretely, at one loop:

1. Compute the divergent 1PI self-energy $\Sigma$ and vertex ${\Gamma }^{(4)}$ from the loop integrals.
2. Choose ${\delta }_{\varphi },{\delta }_m,{\delta }_{\lambda }$ so that the counterterm vertices exactly cancel the ${\Lambda }^2$/$1/ϵ$ divergences.
3. What remains is finite — the physical, renormalized amplitude.

Renormalization conditions and schemes

The finite parts of the counterterms are fixed by renormalization conditions — definitions of what "the mass" and "the coupling" mean. Two common choices:

• On-shell scheme: define $m$ as the physical pole mass (${\Gamma }^{(2)}(p^2=m^2)=0$), $Z_{\varphi }$ as the residue there, and $\lambda$ by the amplitude at a fixed kinematic point. Physical and intuitive; standard in QED.
• $\overline{\text{MS}}$ scheme: absorb just the $1/ϵ$ poles (plus the universal constant), leaving couplings that depend on the arbitrary scale $\mu$. Standard in QCD and precision electroweak fits.

Different schemes give different intermediate numbers but identical physical predictions once related by finite renormalizations — the freedom that becomes the renormalization group.

Renormalizability and power counting

Whether the program terminates — finitely many counterterms suffice to all orders — is decided by the power counting of loop divergences. Classify a theory by the mass dimension of its couplings:

A renormalizable theory needs only the counterterms already present in the Lagrangian (mass, coupling, field): all divergences, to all loop orders, are absorbed by redefining the same finite set of parameters. This is the criterion that singles out the dimension-$\le 4$ operators of the Standard Model.

Non-renormalizable theories are not useless — they are effective field theories, predictive below a cutoff scale $\Lambda$ where the infinitely many counterterms are suppressed by powers of $E/\Lambda$. This is the modern reinterpretation of renormalizability, made precise by the Wilsonian RG.

The systematic all-orders statement: BPHZ

Beyond one loop, divergences nest and overlap (a subdiagram divergence inside a larger one). The BPHZ theorem (Bogoliubov–Parasiuk–Hepp–Zimmermann) proves that the recursive subtraction of subdivergences by counterterms — organized by Zimmermann's forest formula — renders every diagram finite to all orders, provided the theory is power-counting renormalizable. This is the rigorous foundation under the order-by-order procedure above.

What renormalization is not

Renormalization is often misdescribed as "sweeping infinities under the rug." The modern (Wilsonian) understanding is physical:

• The bare parameters were never observable; only the renormalized ones are.
• A QFT is defined with a cutoff — it is an effective theory — and renormalization expresses low-energy physics in terms of a few measured parameters, insensitive to the unknown high-energy completion.
• The residual scale dependence is not a defect but real physics: the running of couplings, measured directly (e.g. $\alpha (m_Z)\ne \alpha (0)$).

Summary

• Bare parameters ($m_0,{\lambda }_0,{\varphi }_0$) are cutoff-dependent; renormalized ones are physical, related by divergent $Z$ factors.
• Counterterms $\delta =Z-1$ cancel loop divergences order by order.
• Renormalizable ⇔ couplings of dimension $\ge 0$ ⇔ finitely many counterterms; proven to all orders by BPHZ.
• Non-renormalizable ⇒ effective field theory, valid below a cutoff.

Where this leads

• Scale dependence and running: the renormalization group.
• The effective-theory viewpoint: Wilsonian RG, effective field theory.
• Gauge-theory renormalization and $Z_1=Z_2$: Ward identities.

References

• Peskin & Schroeder, An Introduction to Quantum Field Theory, Ch. 10.
• Weinberg, The Quantum Theory of Fields, Vol. 1, Ch. 12.
• Collins, Renormalization.
• Srednicki, Quantum Field Theory, Ch. 14–18, 27–31.