Critical Exponent · aiwiki.page

A critical exponent is a number that characterizes how a physical quantity vanishes, diverges, or varies with distance near a critical point. In statistical mechanics, these exponents describe the leading power-law behavior associated with continuous phase transitions. Different substances can share the same exponents despite having different microscopic structures, a property known as universality. Critical exponents therefore characterize collective behavior rather than the detailed properties of individual particles. (damtp.cam.ac.uk)

Definitions and principal exponents

For a transition controlled by temperature, the reduced temperature is

t=\frac{T-T_c}{T_c},

where $T_c$ is the critical temperature. A singular observable commonly behaves as $Q_{\mathrm{sing}}\sim A_\pm |t|^p$ . The exponent $p$ describes its asymptotic behavior, while the amplitudes $A_+$ and $A_-$ , above and below the transition, generally depend on the material. The approach path and any externally applied fields must be specified when defining an exponent. (damtp.cam.ac.uk)

The standard equilibrium exponents can be defined using a magnetic system. Let $m$ be its order parameter, represented by the magnetization, and let $h$ denote the conjugate magnetic field:

Exponent	Defining critical behavior	Physical quantity
$\alpha$	$C_{\mathrm{sing}}\sim \vert t\vert ^{-\alpha}$	Singular part of the [[heat-capacity
$\beta$	$\vert m\vert \sim(-t)^\beta$ , for $t<0$ , $h=0$	Spontaneous order
$\gamma$	$\chi\sim \vert t\vert ^{-\gamma}$ , at $h=0$	[[magnetic-susceptibility
$\delta$	$\vert m\vert \sim \vert h\vert ^{1/\delta}$ , at $t=0$	Response on the critical isotherm
$\nu$	$\xi\sim \vert t\vert ^{-\nu}$	[[correlation-length
$\eta$	$G(r)\sim r^{-(d-2+\eta)}$ , at criticality	Connected order-parameter correlation function

Here $d$ is the spatial dimension and $r$ is separation. The exponent $\eta$ describes the anomalous spatial decay of correlations, rather than a temperature dependence. Analogous definitions apply to fluids, where a density difference supplies the order parameter and compressibility replaces magnetic susceptibility. (damtp.cam.ac.uk)

The singular contribution must be distinguished from a regular background. A negative $\alpha$ need not imply a divergent heat capacity. Moreover, $\alpha=0$ does not determine whether the heat capacity remains finite or has a logarithmic divergence; additional information is required. (damtp.cam.ac.uk)

Universality and representative values

A universality class groups systems sharing the same asymptotic critical behavior. Important determinants include spatial dimension, order-parameter symmetry, and the range of interactions. Microscopic details can alter $T_c$ and amplitudes without changing the exponents. For example, ordinary liquid–gas criticality and certain three-dimensional magnets belong to the Ising universality class. (damtp.cam.ac.uk)

The short-range, two-dimensional Ising model has exact exponents

\alpha=0,\quad \beta=\frac18,\quad \gamma=\frac74,\quad \delta=15,\quad \nu=1,\quad \eta=\frac14.

Its heat capacity diverges logarithmically. In three dimensions, representative approximate values are $\beta\approx0.326$ , $\gamma\approx1.237$ , $\nu\approx0.630$ , and $\eta\approx0.0363$ ; these are not known as exact closed-form numbers. (damtp.cam.ac.uk)

Mean-field theory, which neglects important spatial fluctuations, instead gives

\alpha=0,\quad\beta=\frac12,\quad\gamma=1,\quad \delta=3,\quad\nu=\frac12,\quad\eta=0.

For short-range Ising-type systems, these leading power-law values apply above the upper critical dimension, $d_c=4$ . At four dimensions, logarithmic corrections modify the simple power laws; below four dimensions, fluctuations generally change the exponents. (damtp.cam.ac.uk)

Scaling relations and renormalization

Critical exponents are not generally independent. The scaling hypothesis for the singular free-energy density yields relations including

\alpha+2\beta+\gamma=2,\qquad \gamma=\beta(\delta-1),\qquad \gamma=\nu(2-\eta).

When ordinary hyperscaling applies, one also has

2-\alpha=d\nu.

Under these assumptions, two independent static exponents determine the others. Hyperscaling requires qualifications above the upper critical dimension, where the usual relation between singular free energy and correlation volume fails. (damtp.cam.ac.uk)

The renormalization group explains these relations by examining how a system changes under coarse-graining. Critical behavior corresponds to a fixed point whose long-distance properties are insensitive to many microscopic parameters. If the temperature-like perturbation has scaling eigenvalue $y_t$ , then $\nu=1/y_t$ . The scaling dimension of the order parameter is $(d-2+\eta)/2$ . This connects measurable exponents to the structure of the critical theory. (damtp.cam.ac.uk)

Measurement and finite-size effects

Exponents are determined through experiments, analytical calculations, and numerical simulations. A measured slope over a limited interval need not equal the asymptotic exponent: regular backgrounds and corrections to scaling can produce apparent, or effective, exponents. Numerical studies therefore analyze systematic errors as well as statistical uncertainties. (arxiv.org)

True equilibrium singularities require the thermodynamic limit. A finite sample rounds the transition when its correlation length becomes comparable with its size $L$ . Finite-size scaling exploits this dependence: for example, conventional scaling predicts $\chi(T_c,L)\sim L^{\gamma/\nu}$ . Comparing several system sizes allows exponent estimates while accounting for corrections to scaling. (damtp.cam.ac.uk)

Dynamic critical exponents

Static exponents describe equilibrium behavior, whereas dynamic exponents characterize relaxation. The dynamic exponent $z$ is defined by

\tau\sim\xi^z,

where $\tau$ is a characteristic relaxation time. Its growth near criticality is called critical slowing down. Dynamic behavior depends on conservation laws and the coupling between slow variables, so identical static critical behavior does not necessarily imply identical dynamic exponents. (www1.phys.vt.edu)