Order Parameter

Heat a bar magnet past its Curie point and its magnetization vanishes; the atomic spins, once aligned, now point every which way. Cool a fluid below its critical point and it separates into liquid and vapor with different densities. Chill helium below $2.17\,\text{K}$ and it acquires a macroscopic quantum wavefunction that flows without friction. These transitions look unrelated, but statistical mechanics describes them all with a single object: the order parameter — a quantity that is exactly zero in the disordered, symmetric phase and becomes nonzero once the system spontaneously picks out a direction, a density, or a phase. The order parameter is the central character of the entire theory of phase transitions. Choosing it correctly means identifying which symmetry the transition breaks, because the order parameter is precisely the quantity that transforms nontrivially under that symmetry: magnetization for a ferromagnet (rotational symmetry of spins), the density difference for a fluid, the complex condensate wavefunction $\psi$ for a superfluid, the staggered magnetization for an antiferromagnet. Once chosen, the order parameter organizes everything downstream — its onset ( $\phi\sim(T_c-T)^\beta$ at a continuous transition, or a jump at a first-order one) classifies the transition; its symmetry group fixes the universality class and the number of Goldstone modes; its spatial fluctuations define the correlation length. This article owns the concept and its defining property: minimizing an analytic Landau free energy $f=a\,t\,\phi^2+b\,\phi^4$ gives the mean-field onset $\phi\propto(-t)^{1/2}$ , showing exactly how a nonzero order parameter emerges continuously from zero as the symmetry breaks.

Key takeaways

An order parameter $\phi$ is a thermodynamic quantity that is zero in the symmetric (disordered) phase and nonzero in the broken-symmetry (ordered) phase — it measures how much symmetry has been spontaneously broken.
Its identity is dictated by the broken symmetry: magnetization (ferromagnet), density difference (liquid–gas), complex wavefunction $\psi$ (superfluid), staggered magnetization (antiferromagnet).
At a continuous transition $\phi$ grows continuously from zero as $\phi\sim(T_c-T)^\beta$ ; at a first-order transition it jumps discontinuously.
Its symmetry type — scalar, vector, complex, tensor — fixes the universality class and the spectrum of Goldstone modes.
$\langle\phi\rangle$ over the full symmetric ensemble is always zero; nonzero order appears only after breaking the symmetry, via $\lim_{h\to0^+}\lim_{N\to\infty}\langle\phi\rangle$ (the limits do not commute).
Key idea: pick the quantity that transforms nontrivially under the symmetry broken at the transition; its equilibrium average is zero where the symmetry is intact and nonzero where it is broken, and that single object — its symmetry, onset, and fluctuations — organizes the whole theory of the phase transition.

The core insight

Identify the symmetry that the transition breaks, then choose the order parameter to be a quantity that transforms nontrivially under it. Its thermal average is zero in the symmetric phase and nonzero in the ordered phase. Everything else — the exponent $\beta$ of its onset, the universality class, the Goldstone modes, the correlation length — follows from the order parameter's symmetry and its free-energy expansion.

Where this lives today

The order-parameter concept has spread far beyond its thermodynamic birthplace. In condensed-matter physics it now classifies exotic phases — superconducting gap symmetries ( $s$ -wave, $d$ -wave), spin-density waves, charge-density waves, and the tensor order parameters of liquid crystals used in every display screen. Topological phases of matter forced a genuine extension: some ordered states (the quantum Hall effect, topological insulators) have no local order parameter at all, ordered instead by global topological invariants — a discovery that reshaped the field and showed the limits of the Landau paradigm. In particle physics the same mathematics describes the Higgs field as the order parameter of electroweak symmetry breaking, and the chiral condensate as the order parameter of the QCD vacuum. In cosmology, order parameters and their defects (cosmic strings, domain walls) model symmetry-breaking phase transitions in the early universe. Wherever a system spontaneously chooses among symmetry-equivalent states, an order parameter is the tool that names the choice.

The physical picture

Before a transition, a system respects some symmetry of its Hamiltonian: the spins of a magnet can point equally in any direction, the atoms of a fluid are uniformly distributed, the phase of a superfluid wavefunction is undetermined. The order parameter is a measuring stick for that symmetry. Build it from a quantity that changes when the symmetry acts — flip all spins and the magnetization reverses; shift the superfluid phase and $\psi$ rotates in the complex plane. In the symmetric phase the system averages democratically over all symmetry-related configurations, so the order parameter averages to zero. Below the transition the system commits to one option — one magnetization direction, one density, one phase — and the order parameter becomes nonzero. That commitment is spontaneous symmetry breaking, and the order parameter is what makes it quantitative. The punchline figure is the defining curve: the magnitude of the order parameter as a function of temperature. Above $T_c$ it is flat at zero. At $T_c$ it lifts off, and just below it rises as a square root, $\phi\propto(T_c-T)^{1/2}$ in mean-field theory — a vertical tangent at the critical point. This continuous onset is the signature of a continuous transition; a first-order transition would instead show the order parameter jumping discontinuously from zero to a finite value. The exponent governing the rise, $\beta$ , is one of the critical exponents, and its mean-field value $\tfrac12$ is corrected by fluctuations to $\approx0.326$ in real three-dimensional systems — the true curve rises even more steeply.

Figure

Order parameter magnitude φ versus temperature: zero above T_c, lifting off continuously at T_c and rising as (T_c − T)^β below (blue = mean-field β = ½, the defining onset curve); a first-order transition would instead jump discontinuously (grey dashed) from zero to a finite value

The second picture shows that "order parameter" is a family, not a single quantity, and that its symmetry type is what distinguishes the cases. A scalar (one real number, $\pm$ ) describes the Ising magnet and the liquid–gas transition; a two-component vector — equivalently a complex number — describes a superfluid or an XY magnet, whose order parameter can rotate continuously; a three-component vector describes a Heisenberg magnet. The dimensionality of the order parameter, together with the spatial dimension, is exactly what sorts transitions into universality classes and dictates how many gapless Goldstone modes appear when a continuous symmetry breaks.

Figure

The order parameter's symmetry type sorts systems: a scalar ±φ (Ising magnet, liquid–gas), a complex/2-vector φ = |φ|e^(iθ) with a continuous phase (superfluid, XY magnet), and a 3-vector (Heisenberg magnet); the number of components n and the broken-symmetry manifold fix the universality class

The core equations

The order parameter is defined as the thermal average of a local field that transforms nontrivially under the broken symmetry. For a magnet it is the magnetization density; for a fluid the deviation of the density from its critical value; for a superfluid the complex condensate amplitude.

\phi=\langle \hat\phi\rangle,\qquad \phi=0\ (T>T_c),\qquad \phi\neq0\ (T<T_c)

where: $\hat\phi$ is the microscopic operator (local spin, density, or field), $\langle\cdot\rangle$ is the equilibrium (thermal) average, and $T_c$ is the transition temperature. The defining property is the change in $\phi$ across the transition: it is an exact zero above and grows below. Near a continuous transition the order parameter follows a power law in the reduced temperature. This defines the critical exponent $\beta$ and encodes the continuous onset seen in the figure.

\phi\sim(T_c-T)^{\beta}=(-t)^{\beta},\qquad t\equiv\frac{T-T_c}{T_c}

where: $t$ is the reduced temperature (negative below $T_c$ ), and $\beta$ is the order-parameter critical exponent. Its mean-field value is $\beta=\tfrac12$ ; real 3D systems have $\beta\approx0.326$ (Ising), $0.35$ (XY), $0.37$ (Heisenberg). The full table of critical exponents is owned by the dedicated article; here $\beta$ is introduced as the exponent of the order parameter's onset. Landau's central postulate is that near the transition the free energy is an analytic function of the order parameter, expandable in powers consistent with the symmetry. For a scalar order parameter with $\phi\to-\phi$ symmetry (no odd terms), the leading expansion is quartic.

f(\phi,T)=f_0+a(T)\,\phi^2+b\,\phi^4-h\,\phi,\qquad a(T)=a_0\,t

where: $a(T)=a_0 t$ changes sign at $T_c$ (positive above, negative below), $b>0$ ensures stability, and $h$ is the field conjugate to $\phi$ (the magnetic field for a magnet). This polynomial is the workhorse of the theory of phase transitions; the dedicated Landau-theory article develops its full consequences, while here it is used to establish the order-parameter onset.

Deriving the order-parameter onset

Derivation AContinuous onset φ ∝ (−t)^½ by minimizing the Landau free energy

The goal is to show, from the analytic free energy $f=a(T)\phi^2+b\phi^4$ with $a(T)=a_0 t$ and $b>0$ , that the equilibrium order parameter is exactly zero above $T_c$ and rises continuously as $\phi\propto(-t)^{1/2}$ below it — establishing both the defining property of an order parameter (zero to nonzero) and the mean-field exponent $\beta=\tfrac12$ . The strategy is to minimize $f$ over $\phi$ at each temperature, distinguish the two regimes by the sign of $a(T)$ , and check that the nonzero solution is a genuine minimum. The key point to watch is that the transition is where $a(T)$ crosses zero: above $T_c$ the only minimum is $\phi=0$ ; below, that point becomes a maximum and two symmetric minima appear.

Step 1. Stationarity condition. Minimize $f=a\phi^2+b\phi^4$ (set $h=0$ ): $\dfrac{\partial f}{\partial\phi}=2a\phi+4b\phi^3=2\phi(a+2b\phi^2)=0$ . The roots are $\phi=0$ and $\phi^2=-a/2b$ . Why: the factored form separates the always-present symmetric solution $\phi=0$ from the pair $\pm\sqrt{-a/2b}$ , which is real only when $a<0$ .

Step 2. Above $T_c$ : only $\phi=0$ . For $T>T_c$ , $a=a_0 t>0$ , so $\phi^2=-a/2b<0$ has no real solution and the sole stationary point is $\phi=0$ . The curvature $f''(0)=2a>0$ confirms it is a minimum. Why: a positive quadratic coefficient makes $f$ a single upward bowl; the symmetric state is stable and the order parameter vanishes — the disordered phase.

Step 3. Below $T_c$ : symmetric minima appear. For $T<T_c$ , $a=a_0 t<0$ , so $\phi^2=-a/2b=-a_0 t/2b>0$ gives the real pair $\phi=\pm\sqrt{-a_0 t/2b}$ . Now $f''(0)=2a<0$ (the origin is a maximum) while at the nonzero roots $f''=8|a|>0$ (minima). Why: as $a$ turns negative the single bowl inverts into a double well; the system rolls into one of the two degenerate minima, spontaneously breaking the $\phi\to-\phi$ symmetry.

Step 4. Read off the onset and exponent. The magnitude of the ordered solution is $|\phi|=\sqrt{-a_0 t/2b}=\sqrt{a_0/2b}\,(-t)^{1/2}$ . Thus $\phi$ grows continuously from zero at $T_c$ (where $t=0$ ) as $\phi\propto(-t)^{1/2}=(T_c-T)^{1/2}/T_c^{1/2}$ , i.e. $\beta=\tfrac12$ . Why: the square root is the generic onset of any analytic free energy losing stability through a quadratic-to-quartic balance — the same $\tfrac12$ appears for van der Waals, Weiss, and every mean-field theory, which is why the exponent is universal within mean field. *Verified symbolically (sympy: stationary points, second-derivative signs, and the $(-t)^{1/2}$ scaling) in verify_order-parameter.py.*

The same free energy shows why a conjugate field rounds the transition — the distinction between the order parameter and its field is essential.

Go deeper · Derivation B · 1 minOrder parameter versus conjugate field: why a field rounds the transition

Derivation BOrder parameter versus conjugate field: why a field rounds the transition

The goal is to make precise the difference between the order parameter $\phi$ and the field $h$ conjugate to it, and to show that any nonzero $h$ picks out a single sign of $\phi$ and smooths the sharp onset into a gradual crossover — so that true spontaneous order requires the double limit $\lim_{h\to0^+}\lim_{N\to\infty}$ . The strategy is to reinstate the $-h\phi$ term, minimize, and examine the critical isotherm at $T=T_c$ . The subtlety is the non-commuting limits: sending $h\to0$ first at finite size gives $\phi=0$ , whereas taking the thermodynamic limit first gives $\phi\neq0$ .

Step 1. Minimize with the field on. With $f=a\phi^2+b\phi^4-h\phi$ , stationarity gives $2a\phi+4b\phi^3=h$ . For small $h$ above $T_c$ ( $a>0$ ), linearize: $\phi\approx h/2a$ , so $\phi$ is proportional to $h$ and vanishes as $h\to0$ . Why: the field tilts the free energy, always producing some induced order; but above $T_c$ that order disappears the instant the field is removed.

Step 2. Critical isotherm. At $T=T_c$ , $a=0$ , so the balance is $4b\phi^3=h$ , giving $\phi=(h/4b)^{1/3}\sim h^{1/3}$ . This is the exponent $\delta=3$ : at the critical point the order parameter responds to the field with a pure cubic law, not linearly. Why: with the quadratic term gone, the leading restoring force is cubic, so the response is a cube root — the critical isotherm is infinitely susceptible at $h=0$ .

Step 3. The non-commuting limits. Below $T_c$ , an infinitesimal $h>0$ selects the $+\phi$ minimum; letting $h\to0^+$ after the thermodynamic limit leaves $\phi=+\sqrt{-a/2b}>0$ . But for any finite system, $\langle\phi\rangle=0$ by symmetry once $h=0$ . Hence spontaneous order is $\lim_{h\to0^+}\lim_{N\to\infty}\langle\phi\rangle\neq0$ , with the limits in that order. Why: symmetry breaking is a property of the infinite system; the field is the theorist's device for selecting which broken state the system occupies, and the order of limits encodes that the macroscopic system never tunnels between the wells. *Verified: the $\phi\sim h^{1/3}$ critical isotherm and the linear $\phi=h/2a$ response above $T_c$ (verify_order-parameter.py).*

Limiting cases

case	order parameter	broken symmetry
Ferromagnet	magnetization $\vec M$	rotational (spin) $O(3)$ , or $\mathbb{Z}_2$ for Ising
Antiferromagnet	staggered magnetization $\vec M_s$	sublattice + spin rotation
Liquid–gas	density difference $\rho-\rho_c$	none (Ising-like, $\mathbb{Z}_2$ by analogy)
Superfluid / superconductor	complex condensate $\psi=\|\psi\|e^{i\theta}$	$U(1)$ phase (gauge)
Nematic liquid crystal	orientational tensor $Q_{ij}$	rotational $O(3)$ , tensor
Crystal (solidification)	density Fourier components $\rho_{\vec G}$	continuous translation
$T>T_c$ (any)	$\phi=0$	symmetry intact (disordered)

Worked examples

Common trapThe order parameter is not the conjugate field

The single most common confusion is to mix up the order parameter with the field conjugate to it. For a magnet the order parameter is the magnetization $m$ ; the field is the external magnetic field $h$ . Spontaneous order means $m\neq0$ even as $h\to0$ . A related trap: the average $\langle\phi\rangle$ over the full symmetric ensemble is always exactly zero, because the ensemble sums over all symmetry-related states. To see nonzero order you must break the symmetry — apply an infinitesimal field and take the thermodynamic limit first, or restrict the average to a single pure state. Reporting " $\langle\phi\rangle=0$ therefore no order" from a naive symmetric calculation is the classic error.

Example 1 — Order-parameter onset for a mean-field magnet near $T_c$ .

Step 1. Set the Landau coefficients. Take $a(T)=a_0 t$ with $a_0=1$ (units where $k_B=1$ ) and $b=1$ , at a reduced temperature $t=-0.01$ (i.e. $1\%$ below $T_c$ ). Then $a=a_0 t=-0.01$ .

Step 2. Solve for the order parameter. $\phi=\sqrt{-a/2b}=\sqrt{0.01/2}=\sqrt{0.005}=\mathbf{0.0707}$ . So $1\%$ below $T_c$ the order parameter is about $7\%$ of its unit scale — the square-root onset means it rises much faster than linearly in the temperature offset.

Step 3. Check the scaling. At $t=-0.04$ ( $4\%$ below), $\phi=\sqrt{0.04/2}=\sqrt{0.02}=0.1414$ , exactly $\sqrt{4}=2$ times larger. Quadrupling the distance below $T_c$ doubles the order parameter — the hallmark of the $\beta=\tfrac12$ square-root law.

Example 2 — Critical isotherm exponent $\delta$ .

Step 1. Sit at $T_c$ . Set $t=0$ , so $a=0$ ; the free energy is $f=b\phi^4-h\phi$ with $b=1$ .

Step 2. Minimize. $\partial f/\partial\phi=4b\phi^3-h=0\Rightarrow\phi=(h/4)^{1/3}$ . For $h=0.032$ : $\phi=(0.032/4)^{1/3}=(0.008)^{1/3}=\mathbf{0.2}$ .

Step 3. Read the exponent. Doubling the field to $h=0.064$ gives $\phi=(0.016)^{1/3}=0.252$ , a factor $2^{1/3}=1.26$ . The order parameter grows as the cube root of the field: $\phi\sim h^{1/\delta}$ with $\delta=3$ . This is the mean-field critical-isotherm exponent, complementing $\beta=\tfrac12$ for the temperature onset.

Origins · 1 minorder parameter and how this was discovered

order parameter

the central variable of phase transitions, introduced by Landau (1937) · The order parameter is the organizing concept of the modern theory of phase transitions, introduced by Lev Landau in 1937. Landau's insight was that near a transition one need not know the microscopic details; it suffices to identify the single quantity — the order parameter — that distinguishes the two phases, determine which symmetry it breaks, and expand the free energy in its powers. The choice is dictated entirely by symmetry: whatever transforms nontrivially under the broken operation. This one idea unified phenomena that had seemed unrelated — ferromagnetism, superfluidity, superconductivity, alloy ordering, crystallization — under a common framework, and it remains the starting point for both Landau theory and the renormalization group. The order parameter's symmetry fixes the universality class; its onset exponent

\beta

is measured across systems; its fluctuations define the correlation length and, when a continuous symmetry breaks, the Goldstone modes.

History

1937

— Landau introduces the order parameter and the free-energy expansion for phase transitions

1950

— Ginzburg and Landau add spatial gradients, describing superconductors via a complex order parameter

1960

— the order-parameter concept is extended to superfluids and Bose condensates via the macroscopic wavefunction

1971

— Wilson's renormalization group grounds the order-parameter symmetry in universality classes

1982

— Wilson awarded the Nobel Prize for the theory of critical phenomena built on order parameters

Going further

Texts. Pathria & Beale, Statistical Mechanics, 3rd ed., ch. 12–13, introduces the order parameter within the general theory of phase transitions. Kardar, Statistical Physics of Fields, ch. 1–3, builds the whole subject around it. Sethna, Statistical Mechanics: Entropy, Order Parameters, and Complexity, ch. 9, is devoted to order parameters and their topological defects. Landau & Lifshitz, Statistical Physics Part 1, ch. XIV, is the original framework.

Provenance. Pathria & Beale, Statistical Mechanics, 3rd ed., ch. 12–13; Kardar, Statistical Physics of Fields, ch. 1–3; Sethna, Statistical Mechanics: Entropy, Order Parameters, and Complexity, ch. 9; Landau & Lifshitz, Statistical Physics Part 1, ch. XIV; Reif, Fundamentals of Statistical and Thermal Physics, ch. 8.

Prerequisites. Continuous Phase Transitions · First-Order Phase Transitions · Spontaneous Symmetry Breaking.

Leads to. Landau Theory of Phase Transitions · Critical Exponents · Goldstone Modes · Universality and Universality Classes.

How do I know which quantity to choose as the order parameter?

Identify the symmetry that the transition breaks, then pick a quantity that transforms nontrivially under it — one that is forced to average to zero when the symmetry is intact but can be nonzero once it is broken. For a magnet the spin-rotation symmetry is broken, so magnetization works; for a superfluid the $U(1)$ phase symmetry is broken, so the complex condensate amplitude works; for a crystal continuous translation is broken, so density Fourier components work. If a candidate quantity is nonzero even in the symmetric phase, it is not an order parameter.

Why is ⟨φ⟩ zero in the symmetric phase — doesn't the ordered state exist below T_c?

The ordered state exists, but the full symmetric ensemble averages over all its symmetry-related copies, which cancel. A ferromagnet below $T_c$ can be magnetized up or down; summing over both gives zero net magnetization. To expose the order you must break the symmetry — apply an infinitesimal field, or take the thermodynamic limit so that the system is dynamically locked into one state. The correct definition of spontaneous order is $\lim_{h\to0^+}\lim_{N\to\infty}\langle\phi\rangle$ , with the limits in that order.

What is the difference between a scalar and a vector order parameter?

The number of components. A scalar order parameter ( $n=1$ ) has two states related by a sign flip, as in the Ising magnet or the liquid–gas transition — a discrete $\mathbb{Z}_2$ symmetry. A vector or complex order parameter ( $n=2$ for XY/superfluid, $n=3$ for Heisenberg) can rotate continuously, reflecting a broken continuous symmetry. The distinction matters enormously: continuous symmetry breaking produces gapless Goldstone modes and, in low dimensions, can be forbidden altogether (Mermin–Wagner), while discrete breaking has neither feature.

Does every phase transition have an order parameter?

Most conventional ones do, but not all. Topologically ordered phases — the fractional quantum Hall states, topological insulators, spin liquids — have no local order parameter; they are distinguished by global topological invariants and long-range entanglement instead. Their discovery was a genuine extension beyond the Landau paradigm, which had assumed every phase could be characterized by a local order parameter. So the order parameter is the right tool for symmetry-breaking transitions, but the modern classification of phases is broader.

How does the order parameter connect to the correlation function?

The order parameter is the one-point average $\langle\phi(\vec r)\rangle$ ; the correlation function is the two-point average $\langle\phi(\vec r)\phi(0)\rangle-\langle\phi\rangle^2$ , measuring how order-parameter fluctuations at different points are correlated. In the disordered phase where $\langle\phi\rangle=0$ , the correlation function still carries all the information — it decays over the correlation length $\xi$ , which diverges at the critical point. So the order parameter and its correlation function are the one- and two-point functions of the same field, and together they organize the critical behavior.

Why is the onset always a square root in mean-field theory?

Because mean-field theory replaces the true free energy with an analytic polynomial $f=a_0 t\,\phi^2+b\phi^4$ , and minimizing any such smooth function that loses stability through a sign change of its quadratic coefficient gives $\phi\propto(-t)^{1/2}$ . The square root is a property of the quadratic-to-quartic balance, not of any specific system, which is why van der Waals, Weiss, and Landau all yield $\beta=\tfrac12$ . Real systems deviate ( $\beta\approx0.326$ in 3D) precisely because fluctuations make the free energy non-analytic near $T_c$ — the failure that the renormalization group repairs.