Identify what changes when a material changes state
An ice cube placed in a warm room first warms and then partly becomes liquid water. While ice and water coexist, additional energy can mainly change their relative amounts rather than continually increase temperature. Other transitions are less obvious from shape: a previously transparent liquid mixture becomes milky near a particular temperature, or a heated magnet loses spontaneous magnetization. Phase-transition theory asks what these phenomena share. Why can interacting particles produce a sudden macroscopic change, or an unusually strong continuous response, near specific conditions?
A phase is a homogeneous form of organization with a common macroscopic state. Ice and water can have identical chemical composition but differ in molecular arrangement and motion. Two polymer-rich phases may differ primarily in composition while both remain liquid. Conversely, differently oriented grains in a polycrystal can belong to the same crystalline phase. Identifying a phase therefore involves composition, symmetry, structure and thermodynamic properties, rather than simply counting visible pieces. Interfaces often make phases recognizable, but near a critical point their differences and the interface between them can gradually disappear.
Phase equilibrium means there is no persistent net heat flow, volume change or transfer of components. Equal temperatures prevent sustained one-way heat transfer. Mechanical equilibrium requires the appropriate balance of stresses, often simplified to equal pressures for a flat interface without external fields. Each species able to exchange must also have equal chemical potential in both phases. Chemical potential measures how adding a small amount of that species changes free energy. It resembles a free-energy cost per amount of material, not a price consciously compared by molecules. A curved interface can support a capillary pressure difference, so the mechanical condition must include geometry.
Equilibrium does not require equal densities or equal amounts of the phases. A small piece of ice can coexist with much more water while their masses, volumes and densities differ. Temperature and pressure are intensive quantities: cutting a specimen in two does not simply halve them. Total energy, entropy and amount of substance are extensive and do depend on specimen size. Equilibrium constrains suitable intensive quantities, while total amount and composition conservation determine phase amounts. This distinction prevents “equilibrium” from being interpreted as equality of every number.
Temperature and pressure alone do not uniquely specify every complex specimen. A multicomponent material also needs composition; fields, strain and fixed-network constraints must be stated when relevant. On a coexistence line, additional constraints determine relative phase amounts. In polymers, glass formation and slow relaxation can preserve a memory of history. The equilibrium discussion initially assumes sufficient time to explore accessible states. Measurement timescales will be brought back into the interpretation later.
This language gives different phenomena clear roles. Melting compares crystalline and liquid free energies. Demixing compares distributions of composition. Magnetic ordering compares average orientation states. Their shared method is to identify a quantity that distinguishes states, determine how it changes with control conditions, and explain the collective organization behind it. The theory begins with those questions, before assigning a first-order or continuous-transition label to an experimental curve.
How free-energy slopes record a transition
For fixed temperature, pressure and amount of substance, equilibrium minimizes Gibbs free energy. Imagine separate curves for ice and liquid water as temperature changes. They cross at the equilibrium melting temperature; on opposite sides, different phases provide the lower value. The equilibrium free energy follows the lower portions of both curves. It can remain continuous while changing slope at their crossing. Here slope means how rapidly free energy changes when temperature is changed slightly. Connecting that geometry to measurable quantities explains latent heat.
| Symbol | Meaning | Unit |
|---|---|---|
| S | Total specimen entropy | J K⁻¹ |
| G | Gibbs free energy | J |
| T | Absolute temperature | K |
| P | Pressure | Pa |
| V | Specimen volume | m³ |
| n | Amounts of all components, held fixed | mol |
| ∂G/∂T, ∂G/∂P | Partial derivatives with respect to temperature and pressure; subscripts identify quantities held fixed | J K⁻¹ and m³, respectively |
Reading the relationshipFor fixed composition, the temperature contribution to a small Gibbs-energy change is −S times the temperature change, while the pressure contribution is V times the pressure change. Taking local rates gives the derivatives.
A worked exampleIf G decreases by 20 J per 1 K over a small interval at fixed pressure, S is approximately 20 J K⁻¹. If adding 1,000 Pa at fixed temperature increases G by 0.001 J, the local volume estimate is 10⁻⁶ m³.
ConditionsEquilibrium thermodynamics at fixed composition with ordinary pressure–volume work. Additional fields or stresses must be held fixed or included through their conjugate terms.
Differentiating with respect to temperature gives minus entropy; differentiating with respect to pressure gives volume. A derivative takes a small change, divides it by the change that causes it, and considers progressively smaller intervals to obtain a local rate. Subscripts specify what stays fixed: pressure for the temperature derivative, temperature for the pressure derivative. A sudden change of temperature slope means an entropy discontinuity; a change of pressure slope means a volume discontinuity. A first-order transition has a discontinuity in an appropriate first derivative. It does not require entropy and volume both to have nonzero jumps in every special case.
Latent heat makes an entropy jump tangible. Reversibly converting a specified amount of material between phases at constant temperature requires heat equal to absolute temperature times the entropy change. For a hypothetical transition at 300 K with a molar transition entropy of 10 joules per mole per kelvin, the heat is 3,000 joules per mole. A mole counts approximately 6.022×10²³ specified entities. It allows comparison of equal amounts of material without counting individual molecules. These illustrative parameters show how temperature, entropy and heat combine with consistent units.
| Symbol | Meaning | Unit |
|---|---|---|
| Lm | Molar latent heat for the specified direction, positive for absorption | J mol⁻¹ |
| Ttr | Absolute equilibrium transition temperature | K |
| Δsm | Molar entropy of final phase minus initial phase | J mol⁻¹ K⁻¹ |
Reading the relationshipFor reversible heat exchange at the equilibrium transition temperature, entropy change is heat absorbed divided by temperature; rearranging gives the relation.
A worked exampleAt Ttr=300 K and Δsm=10 J mol⁻¹ K⁻¹, Lm=3,000 J mol⁻¹. Reversing the transition reverses both signs.
ConditionsEquilibrium reversible isothermal conversion at a specified pressure. Molar, mass-specific and total quantities require consistent bases.
At a continuous transition, an order parameter can approach zero continuously, while relevant first derivatives such as entropy and volume remain continuous and response functions become anomalous. Heat capacity measures how much heat accompanies a small temperature increase. Susceptibility measures how strongly an order parameter responds to a small applied field. Such responses involve higher derivatives of free energy and may show finite jumps, peaks, power-law divergences or logarithmic singularities. Restricting all continuous transitions to finite jumps of second derivatives would miss important critical behaviour. The term second-order transition is often used for continuous ordering, but the specific singularity still needs to be stated.
A broad experimental peak is not automatically evidence for a continuous transition. Finite specimen size, uneven temperature, composition distributions and measurement rate can broaden an intrinsically sharp transition. A continuous transition can also produce a very sharp response peak. A first-order transition may show superheating, supercooling and hysteresis because creating a new phase costs interfacial energy. Hysteresis reflects a path involving nucleation and relaxation; its temperatures need not equal the equilibrium transition temperature. Comparing heating and cooling, varying rate and measuring structure connects the classification to its physical cause.
Polymer glass formation makes this distinction particularly useful. During cooling, segmental motion slows until structure can no longer adjust fully within the experimental time. The material responds like a solid, and the observed glass-transition temperature depends on timescale and thermal history. Its measurement conditions differ from those of the equilibrium critical points discussed here. Melting, blend demixing and glass formation can all appear in one temperature scan, while involving crystalline order, composition and dynamical arrest respectively. Identify which information about the state is changing before interpreting the signal with a free-energy or response-function model.
An order parameter makes collective change measurable
Comparing order and disorder requires a quantity that reflects their difference. In a magnet, averaging local magnetic moments gives magnetization. In an ordered alloy, one can compare preferences for two species to occupy different equivalent sublattices. In a symmetric binary mixture, local deviation from equal composition can serve the same purpose. These quantities are order parameters. The criterion is whether they distinguish the phases of interest, not whether all materials use the same symbol or the same kind of average.
Consider a scalar order parameter m with two equivalent signs. Zero means an average symmetric state; nonzero positive and negative values represent two symmetry-related ordered states. With no external field, reversing all orientations leaves the physical energy unchanged. Free energy must then be the same at m and −m. This symmetry excludes odd powers: first and third powers change sign on reversal, while second and fourth powers do not. Scalar means that one number describes the local degree of order. Vector or tensor order parameters require the symmetries appropriate to their additional components.
Landau theory expands free energy in powers of a sufficiently small order parameter and retains the low-order terms needed to describe its local shape. A quadratic term determines whether the origin curves upward or downward. A positive quartic term prevents an unlimited descent at larger order parameter. Temperature can change the quadratic coefficient, turning a single minimum into two. The expression does not require every atomic position, yet it puts symmetry and stability directly into a calculable landscape.
| Symbol | Meaning | Unit |
|---|---|---|
| f | Dimensionless free-energy function normalized by a fixed positive energy scale, omitting the m-independent background | 1 |
| m | Dimensionless scalar order parameter | 1 |
| a | Positive coefficient scale for the quadratic term | 1 |
| b | Positive quartic coefficient | 1 |
| t | Reduced temperature | 1 |
| T | Absolute temperature | K |
| Tc | Critical temperature of this mean-field model | K |
Reading the relationshipSign symmetry permits even powers. The coefficient at changes curvature at the origin; positive b raises the free energy at larger amplitude. Dividing temperature difference by Tc gives a dimensionless distance.
A worked exampleFor a=b=1, Tc=300 K and T=297 K, t=−0.01 and f=−0.005m²+0.25m⁴. At m=0.1, f=−0.000025, below the value at zero.
ConditionsUniform mean-field expansion for a sign-symmetric scalar at zero field, with a and b approximately constant near criticality. Spatial fluctuations are treated separately.
At a minimum, the slope vanishes because an infinitesimal move in either direction cannot lower the free energy. Differentiating the polynomial produces a factor m multiplied by a bracket. A product vanishes when at least one factor does, giving the origin and two possible nonzero candidates. Checking curvature or comparing their free energies identifies the stable minima. As reduced temperature changes from positive to negative, the original minimum becomes a maximum and two minima emerge continuously from zero.
| Symbol | Meaning | Unit |
|---|---|---|
| meq | Equilibrium order parameter; the signs identify equivalent states | 1 |
| t | Reduced temperature (T−Tc)/Tc | 1 |
| a, b | Positive quadratic scale and quartic coefficient from the preceding model | 1 |
| √ | Square root, taken nonnegative before applying either sign | 1 |
| T, Tc | Absolute and critical temperatures | K |
Reading the relationshipThe first derivative is atm+bm³. Setting it to zero and identifying minima gives zero above or at Tc and two nonzero solutions below Tc.
A worked exampleFor a=b=1, t=−0.01 gives meq=±0.1 and t=−0.0025 gives ±0.05. A fourfold reduction in distance to criticality halves the magnitude.
ConditionsUses the preceding uniform mean-field model. At t=0 the quadratic curvature vanishes, but the positive quartic term still makes zero a minimum.
This continuous emergence produces a square-root dependence of the mean-field order parameter on distance below the transition. Reducing that distance by a factor of four reduces the order parameter by a factor of two rather than four. The square root expresses a nonlinear loss of collective order. The derivation, however, replaces spatially varying local states with one mean value and omits long-range fluctuations. It gives a useful mechanism and quantitative predictions in suitable mean-field regimes. Closer to many real critical points, correlations between different regions must be restored.
Symmetry must also match the actual object. A nematic liquid-crystal molecule has equivalent head and tail directions, so reversing its director leaves the state unchanged. That does not mean positive and negative scalar degrees of order describe equivalent states. The complete orientational order parameter is a tensor; an allowed cubic invariant can produce a discontinuous isotropic–nematic transition. Treating even powers as a universal requirement for every transition would remove that mechanism. The first step in Landau reasoning is always to identify symmetry and then select the allowed terms.
How one minimum gives way to another
In a continuous transition, ordered minima emerge smoothly from the origin. A first-order transition can occur differently: the original minimum remains while another appears at a finite distance. Changing the control parameter reverses their relative depths, and equilibrium jumps between them. Free energy is continuous when the minima are equally deep, while the order parameter changes discontinuously. Drawing this landscape explains metastability, supercooling and nucleation more directly than memorizing a latent-heat label.
Retain a sign-symmetric scalar but make the quartic coefficient negative and add a positive sixth-order term. The negative quartic term permits distant wells, while the sixth-order term stabilizes the curve at large amplitude. A transition need not wait for the origin to become unstable. Equilibrium changes as soon as a distant well becomes lower. If the origin still curves upward, it can remain metastable until a sufficient disturbance or nucleation event occurs.
| Symbol | Meaning | Unit |
|---|---|---|
| f | Dimensionless free energy with a constant background omitted | 1 |
| m | Dimensionless continuously varying order parameter | 1 |
| r | Adjustable quadratic coefficient, independently controlled in the interaction | 1 |
| u | Quartic coefficient, chosen as +1 or −1 in the interaction | 1 |
Reading the relationshipThe positive sixth-order term stabilizes large amplitude. With u=+1, minima emerge continuously as r crosses zero. With u=−1, distant minima become equal in depth and then replace the origin.
A worked exampleFor u=−1, r=3/16 and m²=3/4, the three contributions are 9/128, −9/64 and 9/128. Their sum is zero, equal to f(0), with |m|=√(3/4)≈0.866.
ConditionsDimensionless uniform scalar model. r is not a temperature in kelvin; mapping it to temperature requires calibration. The interaction compares equilibrium minima without computing spatial nucleation or time evolution.
With the dimensionless coefficients chosen here and a negative quartic term, the ordered and disordered minima become equally deep at a quadratic coefficient of 3/16. The ordered magnitude is √(3/4), approximately 0.866. The origin loses local stability only when the quadratic coefficient reaches zero. Equilibrium transition and loss of stability therefore occur at different parameter values. Substitution checks the result: at 3/16 and the exact magnitude √(3/4), the positive quadratic, negative quartic and positive sixth-order contributions sum to zero, matching the free energy at the origin. Substituting the rounded value 0.866 instead leaves a small rounding difference.
The curve describes spatially uniform order and contains no surface of a small nucleus. Real nucleation creates a boundary in space, combining a bulk reward for the new phase with an interfacial-area cost. A sufficiently small nucleus can shrink because surface cost dominates, while a sufficiently large one can grow. A specimen can consequently remain in the old phase below the equilibrium transition temperature because of this barrier and a finite waiting time. Further cooling changes both the bulk driving force and nucleation rate.
Green: global minima. Orange: metastable minima. Curve and coefficients are dimensionless.
Change the quartic sign in one sixth-order free-energy model to compare transition types. The slider adjusts the quadratic coefficient; the button selects coexistence or the continuous bifurcation. Feedback lists minima and distinguishes global from local stability.
Select a positive quartic coefficient and slowly change the quadratic coefficient from positive to negative. Watch the minima leave the origin continuously. Then select the negative quartic coefficient and locate the range with three minima: a central disordered state and two equivalent ordered states. Global minima determine equilibrium; other local minima represent metastable states. The slider changes a free-energy parameter rather than simulating molecular jumps through time, so moving it faster does not alter the equilibrium decision.
The comparison distinguishes local curvature, global stability and transition order. Positive curvature means resistance to nearby infinitesimal perturbations, not superiority to every distant state. Equal depths establish equilibrium coexistence but do not give the time required to form an interface. A fuller theory progresses from comparing states through free energy, to describing spatial structure, and then to specifying dynamics. Each stage answers a different question, keeping a free-energy plot tied to what it actually calculates.
From two local choices to collective order
Landau theory starts from average order. The Ising model moves closer to many interacting local units. Imagine a regular grid whose sites each carry a variable restricted to +1 or −1. It can represent a simplified magnetic orientation or, under another mapping, a choice between two species. Neighbouring sites have lower energy when their signs agree, while thermal motion permits local flips. Every site has only two choices, yet the model retains the collective mechanism of neighbours influencing neighbours.
| Symbol | Meaning | Unit |
|---|---|---|
| ℋ | Total energy of a specified lattice configuration | J |
| J | Positive alignment interaction energy; the symbol has the same letter as the joule unit | J |
| h | Field-coupling energy favouring the positive sign, including the field–moment product | J |
| si, sj | Two-state variables at sites i and j | 1 |
| i, j | Lattice-site indices | 1 |
| Σ⟨i,j⟩ | Sum over distinct nearest-neighbour pairs, counting each bond once | 1 |
| Σi | Sum over all lattice sites | 1 |
Reading the relationshipEqual signs give product +1 and bond energy −J; opposite signs give −1 and energy +J. Summing bond energies and site-field contributions gives total energy.
A worked exampleFor two neighbouring sites at h=0, (+1,+1) has energy −J and (+1,−1) has +J, differing by 2J. With J=10⁻²¹ J, the difference is 2×10⁻²¹ J.
ConditionsTwo-state nearest-neighbour ferromagnetic-coupling model. Geometry, boundaries and field specify its results. A composition mapping requires the appropriate conservation constraint.
A summation adds the contributions over a specified set. The nearest-neighbour sum counts each bond once so that an interaction is not duplicated. Equal signs give a product of +1 and energy −J; opposite signs give −1 and energy +J. Changing a bond from agreement to disagreement therefore costs 2J. The field term favours one sign. Here h includes the field–moment coupling and has energy units, preventing an energy parameter from being confused with a magnetic field measured in tesla.
Low-energy states have greater statistical weight, but higher-energy states remain accessible at nonzero temperature. A flip can create more unlike neighbours and raise energy. There are often many more disordered arrangements, however, which supplies an entropy benefit. At low temperature, large aligned regions are favoured; at high temperature, numerous disordered arrangements dominate the statistics. A transition emerges from the competition throughout many connected sites, not from one site crossing an independent temperature threshold.
The model developed from Lenz’s 1920 proposal, and Ising studied a one-dimensional nearest-neighbour chain in 1925. For an infinite chain with finite-range nearest-neighbour interactions and no field, spontaneous magnetization is absent at every nonzero temperature. Intuitively, reversing an arbitrarily long interval creates boundaries at a finite energy cost, while there are many possible boundary positions. In two dimensions, the boundary length grows with the size of a region, changing the competition and allowing finite-temperature order. Dimensionality is therefore a physical condition; a one-dimensional result cannot be extended to all spatial models.
In 1944, Onsager obtained the exact zero-field free energy of the two-dimensional square-lattice Ising model, showing critical singularities different from simple mean-field predictions. For equal nearest-neighbour couplings in both directions on an infinite square lattice, critical thermal energy is approximately 2.269 times the interaction strength. That number belongs to the specified lattice, coupling and field conditions. Changing connectivity or interaction range changes the transition temperature. Yet different microscopic models can share critical exponents, creating the deeper question of universality.
A large same-sign region in a lattice picture does not mean that every site is frozen. Local flips continue and boundaries move. A finite zero-field specimen can even switch between positive and negative overall magnetization, making its very long-time signed average nearly zero. Discussing spontaneous order requires attention to the order of the infinite-system and vanishing-field limits, or observables suitable for a finite specimen such as absolute magnetization and its distribution. An average is a defined measurement; a zero value alone does not establish the absence of ordered structure.
How fluctuations correlate distant regions
A fixed temperature does not give every small region an identical instantaneous energy or composition. Moving molecules continually change local density, composition and magnetization around their averages. These changes are fluctuations. If two regions are independent, a high value in one tells us nothing about the other. If they are correlated, observing a deviation in one changes the expected state of the other. A correlation function records this tendency to deviate together as a function of distance, making the spatial reach of collective behaviour measurable.
| Symbol | Meaning | Unit |
|---|---|---|
| C(r) | Order-parameter correlation between locations separated by r | 1 |
| δm | Local m minus its mean | 1 |
| m | Dimensionless local order parameter | 1 |
| 𝐱, 𝐫 | Position vector and separation vector | m |
| r | Magnitude of separation | m |
| ⟨…⟩ | Statistical average over equilibrium samples or equivalent positions | 1 |
| C0 | Amplitude of the illustrative exponential decay | 1 |
| ξ | Correlation length | m |
| e | Base of the natural exponential, approximately 2.71828 | 1 |
| ∼ | Indicates the discussed approximate decay, not exact equality for every r | 1 |
Reading the relationshipTwo deviations of the same sign give a positive product; independent deviations have a product averaging toward zero. The exponential illustrates rapid loss of correlation after several correlation lengths.
A worked exampleWith ξ=10 nm, the exponential factor is e⁻¹≈0.368 at r=10 nm and e⁻³≈0.0498 at 30 nm. If ξ increases to 30 nm, the factor at 30 nm becomes 0.368.
ConditionsApproximately translationally homogeneous equilibrium system. A pure exponential illustrates the length; full critical correlations can include a power-law prefactor and become long-distance power laws at criticality.
Correlation length measures the distance over which deviations remain appreciably connected. In many systems away from criticality, correlation decays rapidly beyond that length. Near a critical point, progressively larger regions fluctuate together and the length grows. In an exponential decay, e is the base of natural logarithms; at one correlation length, the exponential factor falls to approximately 0.368 of its initial value. This gives an intuitive length measure. Near criticality the complete function can also contain a power of distance, so not every correlation curve should be fitted to one pure exponential.
Critical opalescence makes the effect visible. Composition fluctuations in a liquid mixture produce local refractive-index variations. When the relevant scales and response become sufficiently large, strong light scattering makes the specimen cloudy. The scattering objects need not be frozen particles: they can be time-dependent density or composition correlations. Different scattering angles probe different spatial scales, allowing estimation of correlation length and long-wavelength fluctuation strength. Increased scattering near criticality connects molecular statistics with a macroscopic optical observation.
Fluctuations are also linked to response. A system that readily produces large spontaneous deviations without a field generally responds strongly to a small imposed preference. Increased magnetic susceptibility, compressibility and composition response can be understood through appropriate fluctuation–response relations. These connect conjugate variables under specified equilibrium conditions. Experiments holding different quantities fixed can measure different response functions. Stating the variable and boundary conditions is therefore necessary before comparing heat capacity, compressibility or susceptibility within one statistical framework.
Mean-field theory replaces a site’s neighbours with an average environment. It is effective when fluctuations are sufficiently weak. Near criticality, a deviation at one site is correlated with many other regions; replacing the surroundings with an uncorrelated average background removes information that matters. The average remains useful, but correlations now influence the leading behaviour as well. Spatial dimension, interaction range and observation distance all affect when these fluctuations must be included.
Every real specimen has finite dimensions. When correlation length approaches film thickness or the observation window, geometry cuts off further growth and an ideal infinite-system divergence becomes a finite peak or rounded crossover. Critical regions also commonly relax slowly because larger correlated regions take longer to rearrange. Thin films, particles and short measurements therefore truncate critical behaviour in different ways. Temperature precision, specimen dimensions and waiting time together establish which part of the critical regime an experiment has actually examined.
Using power laws to compare critical behaviour
Materials with different transition temperatures can be compared using relative distance: subtract the critical temperature and divide by it. The resulting reduced temperature is dimensionless. Being 3 K from 300 K then represents the same relative distance as being 6 K from 600 K. Near criticality, the leading behaviour of many quantities is a power of that distance. An order parameter decreases as a positive power below the transition, while correlation length increases as a negative power of absolute distance. Exponents describe the approach; prefactors retain material-specific information and units.
| Symbol | Meaning | Unit |
|---|---|---|
| t | Reduced temperature; |t| is its absolute value | 1 |
| T, Tc | Absolute and critical temperatures | K |
| |m| | Magnitude of spontaneous order | 1 |
| B | Order-parameter amplitude | 1 |
| χ | Here the order-parameter susceptibility to a dimensionless conjugate field, not a mixing interaction parameter | 1 |
| χ0 | Susceptibility amplitude | 1 |
| ξ, ξ0 | Correlation length and its amplitude | m |
| β, γ, ν | Critical exponents of order, susceptibility and correlation length | 1 |
| ∼ | Leading asymptotic law near criticality, with amplitudes defined separately on either side | 1 |
Reading the relationshipRelative temperature distance compares materials. A positive exponent makes order vanish, while negative exponents make response or length grow. Amplitudes retain units; exponents control multiplicative changes.
A worked exampleWith ξ0=1 nm and ν=1, reducing |t| from 0.1 to 0.01 increases ξ from 10 to 100 nm. With B=1 and β=1/2, reducing −t from 0.04 to 0.01 lowers |m| from 0.2 to 0.1.
ConditionsSpecified universality class and critical regime, near-zero conjugate field and adequate equilibration. Finite size, background and crossover must be identified in data.
A power law means a fixed rule for multiplicative changes. If the correlation-length exponent is one, approaching ten times closer to the critical point increases the length tenfold. For an exponent of one half, the increase is approximately 3.16-fold. Logarithmic axes convert multiplication into addition, turning the exponent into a straight-line slope. A log–log plot is therefore useful, but an approximately straight segment over a finite range still requires analysis of background, uncertainty and crossover. A small error in the estimated critical temperature can produce a large relative error in distance very near the transition.
Six common static exponents refer to different measurements. β describes disappearance of spontaneous order; γ, growth of zero-field response; ν, growth of correlation length; δ, response to field exactly at critical temperature; α, the singular part of heat capacity; and η, the critical spatial decay of correlation. Here γ is an exponent and χ a response function: their relationship does not make them the same quantity. Exponents also differ from critical temperature. The latter states where the transition occurs, while the former describe how quantities change on approaching it.
| Symbol | Meaning | Unit |
|---|---|---|
| csing | Singular heat-capacity part normalized by a fixed heat-capacity scale | 1 |
| A | Normalized heat-capacity amplitude | 1 |
| t | Reduced temperature (T−Tc)/Tc | 1 |
| T, Tc | Absolute and critical temperatures | K |
| m | Dimensionless order parameter | 1 |
| h | Dimensionless conjugate field in this equation, normalized by a fixed energy scale | 1 |
| D | Critical-isotherm amplitude | 1 |
| C(r) | Order-parameter correlation function | 1 |
| K | Correlation amplitude; here a symbol rather than the kelvin unit | 1 |
| r, r0 | Separation and fixed reference length | m |
| d | Spatial dimensionality | 1 |
| α, δ, η | Heat-capacity, critical-isotherm and correlation-decay exponents | 1 |
| ∼ | Leading asymptotic relation; α=0 requires distinguishing a logarithmic singularity from a finite jump | 1 |
Reading the relationshipEach exponent has a distinct protocol: vary temperature for heat capacity, field at t=0 for the critical isotherm, or separation at criticality for correlation. A length ratio keeps the powered quantity dimensionless.
A worked exampleWith δ=3 and D=1, lowering h from 0.008 to 0.001 lowers |m| from 0.2 to 0.1. For two-dimensional Ising with d=2 and η=1/4, increasing separation sixteenfold halves the correlation power-law factor.
ConditionsStatic equilibrium critical asymptotes. Heat capacity requires a regular background subtraction; correlation applies above microscopic lengths and below finite-size cutoffs.
The six exponents are not arbitrary independent fitting parameters. In common short-range critical systems, the scaling structure of free energy relates them. Order, critical-isotherm response and susceptibility are linked; correlation length and spatial correlations constrain susceptibility; where hyperscaling applies, dimension enters as well. The two-dimensional short-range Ising model provides a checkable example: β=1/8, γ=7/4, δ=15, ν=1 and η=1/4. Its α is zero but heat capacity has a logarithmic singularity. A zero α alone therefore does not specify constant heat capacity: logarithmic behaviour or a finite jump must also be identified.
| Symbol | Meaning | Unit |
|---|---|---|
| α | Heat-capacity exponent | 1 |
| β | Spontaneous-order exponent | 1 |
| γ | Susceptibility exponent | 1 |
| δ | Critical-isotherm exponent | 1 |
| ν | Correlation-length exponent | 1 |
| η | Critical spatial-correlation exponent | 1 |
| d | Spatial dimension | 1 |
Reading the relationshipThese relations connect separate observables through a common singular-free-energy and correlation-scaling structure. The final relation is hyperscaling and additionally requires free-energy scaling by correlated volumes.
A worked exampleFor two-dimensional Ising: α=0, β=1/8, γ=7/4, δ=15, ν=1 and η=1/4. Check 0+2/8+7/4=2, (1/8)×14=7/4, 1×(2−1/4)=7/4 and 2=2×1.
ConditionsCommon static equilibrium scaling framework. Hyperscaling is normally used below the upper critical dimension and cannot be applied unchanged to ordinary mean-field exponents above it. Logarithmic corrections require separate treatment.
Common scalar mean-field values are β=1/2, γ=1, δ=3, ν=1/2 and η=0, with a finite heat-capacity jump in the simple double-well model. The upper critical dimension for the short-range scalar model is four. Below it, critical fluctuations generally alter exponents; four dimensions bring logarithmic corrections; above it, mean-field-type exponents can apply. Dimension here counts independent spatial directions, not time added as a fourth coordinate. It also differs from the number of order-parameter components. A three-component magnetic moment on a two-dimensional film illustrates those two distinct numbers.
Data often show crossover. Mean-field behaviour can work farther from criticality, fluctuation-dominated behaviour can emerge closer in, and finite size or equilibration time can cut off the closest region. Selecting a fit interval is therefore part of the physics. Specify whether the fitted quantity is an order parameter, a background-subtracted response or a correlation length, together with temperature range and geometry. Those choices determine whether a measured slope reflects asymptotic criticality, crossover or another physical process.
Why different materials can share critical behaviour
Water, magnets and polymer blends have different molecules, yet under suitable conditions they can share critical exponents. The reason becomes clearer when observation scale is enlarged repeatedly. Near a critical point, correlation length greatly exceeds molecular size. Many local variables within a region can be replaced by a collective variable, and those collective variables can describe still larger regions. Atomic details often change numerical coefficients without determining the dominant long-distance form. Symmetry, dimensionality and the properties of interactions that survive this enlargement become the basis for classification.
Kadanoff’s 1966 block-spin argument made the process concrete: group lattice sites into blocks, represent each block by collective magnetization, and ask how blocks interact. Blocks must be larger than the microscopic scale but smaller than correlation length to retain the relevant information. Wilson subsequently developed the scale-dependent evolution of parameters into the renormalization-group method. Renormalization is not simply blurring an image and declaring details irrelevant. It changes observation units and effective interactions together, tracking which differences grow and which diminish at larger scales.
If repeated rescaling approaches an unchanged form, that form is a fixed point. Different materials can begin with different parameters yet approach the same fixed point and share critical exponents: they belong to one universality class. Their critical temperatures, densities and amplitude factors remain dependent on chemistry and microscopic interactions. Universality therefore preserves material differences. It identifies shared exponents and suitable dimensionless ratios, rather than identical complete phase diagrams.
Spatial dimension and order-parameter component count are important, alongside symmetry, interaction range, disorder and constraints that can alter leading behaviour. A short-range one-component scalar order parameter often leads to Ising behaviour; two-component planar rotation connects to XY models; rotationally symmetric three-component moments connect to Heisenberg models. The relation between self-avoiding polymers and O(n) models uses the mathematical limit of component number approaching zero, not a four-component model. It links a path-counting problem with critical statistics; the component number does not count visible polymer chains.
Composition fluctuations in polymer blends often provide an approximately scalar order parameter. With short-range interactions in a three-dimensional bulk sample, critical behaviour can approach the three-dimensional Ising class. Long chains also affect the width of the fluctuation-dominated temperature region, making some accessible intervals appear closer to mean field. Thin-film geometry, crosslinks, long-range electrical interactions and dynamical arrest can change the problem. Choose a model from the composition variable, geometry and interactions, rather than assigning an exponent because the material happens to be a polymer.
A complete experimental interpretation follows a connected logic. Establish how the states differ using structure and composition, then select an order parameter. Use reversible paths and adequate waiting to locate the equilibrium regime and distinguish latent heat from a broadened response. Add correlation-length information from scattering or spatial measurements, and test exponent relations within the same temperature range. If closer approach produces departures from a previous power law, examine finite size, equilibration time and crossover. Each departure then becomes a measurable physical question rather than a reason to force every curve onto one straight line.
The central interest of phase transitions is how relationships among local choices produce macroscopic laws. A local orientation or a preferred segment contact supplies a microscopic rule. Correlations extend its influence, free energy and order parameters record collective organization, and scaling compares how materials approach criticality. Connecting these levels explains why some changes occur suddenly, why others remain continuous yet become extremely sensitive, and why very different materials can share a set of long-distance laws.