The Renormalization Group

Renormalization in a scheme like $\overline{\text{MS}}$ leaves the couplings depending on an arbitrary scale $\mu$. Physical observables cannot depend on this choice — and demanding so is not a triviality but a powerful constraint: it forces the couplings to run with energy in a calculable way, governed by $\beta$-functions. This is the renormalization group (RG), the single most important tool for extracting physics from a renormalizable QFT: it explains the running $\alpha$, asymptotic freedom, and the fate of a theory at high and low energies.

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

Scale independence of physics

In dimensional regularization the renormalized coupling $\lambda (\mu )$ and mass $m(\mu )$ carry the arbitrary scale $\mu$. A physical quantity $G$ (an S-matrix element, a correlator) computed to all orders cannot depend on $\mu$:

$$\mu \frac{d}{d\mu }G(p_i;\lambda (\mu ),m(\mu ),\mu )=0.$$

Expanding the total derivative gives the Callan–Symanzik equation — the master RG equation:

$$\boxed{\left[\mu \frac{\partial }{\partial \mu }+\beta (\lambda )\frac{\partial }{\partial \lambda }+{\gamma }_{m}m\frac{\partial }{\partial m}-n\gamma \right]G^{(n)}=0,}$$

with the RG functions defined by how the renormalized quantities respond to $\mu$:

$$\beta (\lambda )=\mu \frac{d\lambda }{d\mu },\gamma =\frac{1}{2}\mu \frac{d\ln Z_{\varphi }}{d\mu },{\gamma }_m=-\mu \frac{d\ln m}{d\mu }.$$

$\beta$ governs the running coupling, $\gamma$ is the anomalous dimension of the field (the quantum correction to its scaling), and ${\gamma }_m$ the anomalous dimension of the mass.

The running coupling

The $\beta$-function determines how the coupling changes with scale. Integrating $\mu d\lambda /d\mu =\beta (\lambda )$ gives $\lambda (\mu )$ as a function of energy. At one loop $\beta =b{\lambda }^2+\cdots$, with solution

$$\frac{1}{\lambda (\mu )}=\frac{1}{\lambda ({\mu }_0)}-b\ln \frac{\mu }{{\mu }_0}.$$

The sign of $b$ decides the theory's character:

This "running" is not formal: the QED coupling measured at the $Z$ mass, $\alpha (m_Z)\approx 1/128$, differs from the low-energy $\alpha (0)\approx 1/137$ exactly as the $\beta$-function predicts — a directly measured RG effect (see electroweak observables).

The QED and ${\varphi }^4$ beta functions

For QED, one-loop vacuum polarization gives

$$\beta (e)=\frac{e^3}{12{\pi }^2}>0,$$

so the effective charge grows at short distance (screening by virtual $e^{+}{e}^{-}$ pairs): the coupling formally diverges at the astronomically high Landau pole, signalling that QED alone is incomplete in the deep UV (it is embedded in the electroweak theory well before then). For ${\varphi }^4$, $\beta (\lambda )=3{\lambda }^2/16{\pi }^2>0$ — same story, a triviality issue.

Fixed points and scaling

A zero of the $\beta$-function, $\beta ({\lambda }_{\ast })=0$, is a fixed point: there the coupling stops running and the theory becomes scale-invariant, in fact typically a conformal field theory. Fixed points organize the long-distance behavior of QFTs:

• UV fixed point — the coupling flows to ${\lambda }_{\ast }$ at high energy. A free (Gaussian) UV fixed point is asymptotic freedom; an interacting one is asymptotic safety (a proposed route to quantum gravity).
• IR fixed point — governs the deep-infrared / long-distance physics, and controls critical phenomena in statistical mechanics (via the Euclidean analogy).

Near a fixed point, correlation functions exhibit power-law scaling with critical exponents built from the anomalous dimensions $\gamma$ — the same exponents measured at second-order phase transitions, the deep link between QFT and critical phenomena that won Wilson the 1982 Nobel Prize.

Improved perturbation theory

Even away from fixed points, the RG resums large logarithms. A one-loop amplitude carries $\lambda \ln (E/\mu )$; at high energy this log is large and naive perturbation theory fails term by term. Choosing $\mu \sim E$ sets the log to zero and moves all the scale dependence into the running coupling $\lambda (E)$, so the expansion is controlled whenever $\lambda (E)$ is small. This RG-improved perturbation theory is what makes QCD predictive at high energy (small ${\alpha }_s(E)$) despite being strongly coupled at low energy.

Summary

• Physics is $\mu$-independent ⇒ Callan–Symanzik equation, with $\beta$, $\gamma$, ${\gamma }_m$.
• $\beta (\lambda )=\mu d\lambda /d\mu$ makes couplings run; its sign fixes UV behavior (Landau pole vs. asymptotic freedom).
• Fixed points ($\beta =0$) are scale-invariant CFTs controlling UV/IR limits and critical phenomena.
• The RG resums large logarithms, giving RG-improved perturbation theory.

Where this leads

• The cutoff/effective-theory formulation of the RG: Wilsonian RG and EFT.
• Asymptotic freedom computed: asymptotic freedom in Yang–Mills.
• Fixed points as CFTs: conformal field theory.

References

• Peskin & Schroeder, An Introduction to Quantum Field Theory, Ch. 12.
• Weinberg, The Quantum Theory of Fields, Vol. 2, Ch. 18.
• Wilson & Kogut, Phys. Rep. 12, 75 (1974).
• Srednicki, Quantum Field Theory, Ch. 27–28.