Renormalization Group

The renormalization group (RG) is a framework in physics for describing how a system’s effective description changes with the length or energy scale at which it is examined. It connects microscopic interactions with large-scale behavior and organizes the scale dependence of theoretical parameters. Its principal applications are in quantum field theory, statistical mechanics, and condensed matter physics. Unlike renormalization understood narrowly as the treatment of divergent expressions, RG concerns transformations between descriptions at different scales. (damtp.cam.ac.uk)

Historical development

RG methods originated in quantum field theory during the 1950s, where changes in the normalization scale were related to changes in renormalized parameters. During the 1960s and 1970s, Leo Kadanoff, Michael Fisher, and Kenneth Wilson developed the scale-based framework for critical phenomena. Wilson’s approach treated a difficult many-scale problem as a sequence of transformations, systematically accounting for fluctuations from atomic distances to macroscopic lengths. He received the 1982 Nobel Prize in Physics for his theory of critical phenomena in connection with phase transitions. (damtp.cam.ac.uk)

This development also connected problems in statistical physics with questions about fundamental interactions. Wilson’s work demonstrated how effective Hamiltonians containing many couplings could be constructed at progressively lower energies, including in his treatment of the Kondo problem, involving a magnetic impurity interacting with conduction electrons. (nobelprize.org)

Coarse-graining and effective descriptions

The Wilsonian procedure begins with coarse-graining: microscopic variables or short-wavelength fluctuations are averaged over or integrated out. In a lattice spin system, neighboring spins can be replaced by block variables. In a field theory, one can integrate over modes whose momenta lie in a shell near an ultraviolet cutoff. Coordinates and fields are then rescaled so that the resulting theory can be compared with the original one. (damtp.cam.ac.uk)

Eliminated fluctuations are not simply ignored: their effects modify the interactions among the remaining degrees of freedom. The result is an effective Hamiltonian or action, generally containing additional interactions allowed by the system’s symmetries. Repeating the procedure produces a trajectory through “theory space,” whose coordinates are interaction strengths and other parameters. (s3.cern.ch)

Successive changes of scale compose, giving RG its name. However, coarse-graining generally loses information and cannot be uniquely reversed. In this formulation, the transformations form a semigroup rather than a group in the strict sense of group theory. (damtp.cam.ac.uk)

Flow equations and fixed points

Continuous changes of scale are expressed through beta functions. For dimensionless couplings $g_i$ defined at an energy scale $\mu$,

These coupled equations describe RG flow. A fixed point satisfies $\beta_i=0$ for every coupling, so the dimensionless description remains unchanged under rescaling. A Gaussian fixed point describes a free theory; interacting fixed points are non-Gaussian. (s3.cern.ch)

Near a fixed point, small perturbations can be decomposed into scaling directions. With increasing length scale $b$, a perturbation behaves to linear order as

Directions with $y_a>0$ are relevant, those with $y_a<0$ are irrelevant, and those with $y_a=0$ are marginal. Marginal directions require higher-order analysis to determine whether they grow, diminish, or remain constant. Relevant perturbations drive a system away from criticality; irrelevant ones typically supply corrections to its leading scaling behavior. (damtp.cam.ac.uk)

Critical phenomena and universality

At a continuous critical point, the correlation length diverges in the ideal infinite-system limit. Fluctuations occur across many scales, making approximations that neglect their interactions inadequate. RG explains this behavior through flows toward a scale-invariant fixed point, where correlations exhibit power-law dependence rather than decay governed by a finite characteristic length. (nobelprize.org)

Universality arises when microscopically different systems approach the same fixed point and consequently share critical exponents and scaling forms. Spatial dimension, symmetry, and interaction range help determine this classification. Microscopic differences associated with irrelevant directions fade at large scales. Thus, an Ising model can describe universal aspects of systems whose microscopic constituents are quite different. (damtp.cam.ac.uk)

For example, near a critical temperature,

where $\xi$ is the correlation length. The exponent $\nu$ is related to the relevant thermal scaling direction by $y_t=1/\nu$. Such relations connect measurable singular behavior with the structure of RG flow. (damtp.cam.ac.uk)

Running couplings in quantum field theory

In particle physics, RG equations describe running couplings: renormalized interaction strengths depend on the scale at which they are defined. In quantum electrodynamics, the electromagnetic coupling increases with energy within its perturbative regime. In quantum chromodynamics, the strong coupling decreases at sufficiently high energies, producing asymptotic freedom. (arxiv.org)

The arbitrary renormalization scale is not itself an observable. In an exact prediction, explicit scale dependence cancels against that of renormalized parameters and operators. Calculations truncated in perturbation theory retain residual dependence. RG evolution also resums classes of large logarithms that would otherwise impair a fixed-order expansion. (arxiv.org)

Methods and limitations

RG is a framework rather than a single computational method. The epsilon expansion studies theories near a critical spatial dimension; numerical RG iteratively constructs low-energy descriptions; and functional RG evolves a scale-dependent effective action. Although functional flow equations can be exact, practical solutions usually require truncating the interactions or functional dependence retained. (damtp.cam.ac.uk)

Within effective field theory, RG separates effects associated with different scales and evolves the coefficients of effective operators. Its usefulness does not require knowing every microscopic interaction, but quantitative predictions depend on controlling approximations, matching conditions, and the range over which the effective description applies. (cds.cern.ch)