Order Parameter and Spontaneous Magnetization

The spontaneous magnetization is the measurable order parameter of real ferromagnets — iron, nickel, cobalt — which carry a net moment in zero applied field below their Curie temperature and lose it continuously on heating through $T_c$. Careful measurements of $m_0(T)\sim(T_c-T)^\beta$ near $T_c$ return $\beta\approx0.33$, matching the 3D Ising universality class rather than mean field and confirming the decisive role of fluctuations. The same order parameter, suitably generalized, is read directly in the laboratory: neutron diffraction sees the staggered magnetization of an antiferromagnet as magnetic Bragg peaks at the Brillouin-zone boundary, exposing "hidden" order even when the net moment is zero; X-ray and neutron scattering track sublattice order in binary alloys such as $\beta$-brass; and coexistence-curve measurements on fluids near their critical point trace the lattice-gas order parameter $\rho_{\rm liq}-\rho_{\rm gas}\sim(T_c-T)^\beta$, the shrinking density difference whose exponent Guggenheim first collated across many substances. Spontaneous symmetry breaking, first sharpened here, has since become the organizing idea of condensed-matter and particle physics alike — superfluids, superconductors, liquid crystals, and the Higgs mechanism are all order-parameter stories.

Picture the free energy as a landscape over the magnetization $m$. Above $T_c$ it is a single bowl centred at $m=0$: the average spin is zero and the phase is symmetric. As $T$ drops below $T_c$ the bottom flattens and then splits into two symmetric wells at $\pm m_0$. The system must fall into one of them; whichever it picks, the reflection symmetry of the landscape is broken by the state it occupies, not by the landscape itself. This is spontaneous symmetry breaking made visual.

An infinitesimal field tilts the double well, lowering one minimum slightly. In a finite system thermal fluctuations still let the magnetization tunnel between the two wells, so the long-time average is zero. In an infinite system the barrier between wells grows with the volume and becomes insurmountable: the magnetization freezes into one well, and an arbitrarily small bias decides which. That is exactly why the thermodynamic limit must be taken before the field is switched off — the single most important subtlety of the whole subject. Equivalently, think in terms of correlations. In the disordered phase distant spins decorrelate, $\langle s_0s_r\rangle\to0$; in the ordered phase they remain correlated at infinite separation, $\langle s_0s_r\rangle\to m_0^2$. Long-range order is the spontaneous magnetization, defined without ever touching a field. Dimensionality then decides everything: in one dimension a single domain wall costs only the finite energy $2J$ but earns entropy $\sim k_B\ln N$, so order is destroyed at any $T>0$ and $m_0=0$; in two dimensions and above the cost of disordering wins at low $T$, and a genuine spontaneous magnetization survives below $T_c$. The punchline of the field is the shape of $m(h)$ itself. Above $T_c$ the magnetization passes smoothly through the origin — no memory of sign. Below $T_c$ it jumps: approaching $h=0$ from above leaves you at $+m_0$, from below at $-m_0$. The spontaneous magnetization is precisely the size of that jump, and its very existence is what a phase transition means.

The order parameter of a magnet is the magnetization per spin — the thermal average of a single Ising variable, averaged over the lattice.

where: $m$ is the magnetization per spin (the order parameter); $s_i=\pm1$ the Ising spin at site $i$; $N$ the number of sites; $\langle\cdot\rangle$ the thermal ensemble average. Physically: $m$ measures how far the state has broken the up/down symmetry. The spontaneous magnetization is the zero-field order parameter, but only if the two limits are taken in the correct order — infinite volume first, vanishing field second.

where: $m_0$ is the spontaneous magnetization; $h$ the external field conjugate to the spin; $\langle s_i\rangle_h$ the single-spin average in field $h$. Physically: the tiny field selects one of the two degenerate broken states, and infinite volume traps the system there so the state survives $h\to0$. The field-free signature of the same order is a nonzero plateau in the two-point correlation function at infinite separation.

where: $\langle s_0s_r\rangle$ is the correlation between spins a distance $r$ apart; $m_0^2$ the square of the spontaneous magnetization. Physically: distant spins that stay correlated forever are the definition of long-range order — no external field is needed to detect it. Mean-field (Weiss) theory replaces the fluctuating neighbour sum by its average, closing a single self-consistency equation for $m$.

where: $q$ is the coordination number (number of nearest neighbours); $J>0$ the ferromagnetic exchange; $k_B$ Boltzmann's constant; $T$ the temperature. Physically: each spin feels an effective field $qJm+h$ set by the average of its neighbours, and self-consistency $m=\langle s_i\rangle$ closes the loop. This is derived, not re-derived, in the companion article Mean-Field (Weiss) Theory of the Ising Model. Linearizing (4) for small $m$ locates the mean-field critical temperature where a nonzero solution first appears.

where: $T_c$ is the mean-field (Curie) temperature at which $m_0$ becomes nonzero. Physically: at $T_c$ the slope of $\tanh(qJm/k_BT)$ at $m=0$ equals one, so the flat minimum splits into two. Landau's symmetry-constrained free energy encodes the whole picture: an even polynomial in $m$ whose quadratic coefficient changes sign at $T_c$.

where: $f$ is the free-energy density; $a(T)$ the quadratic coefficient (negative below $T_c$); $b>0$ the stabilizing quartic; $h$ the ordering field. Physically: the $\mathbb{Z}_2$ symmetry $m\to-m$ forbids odd powers at $h=0$, so only even terms survive; $b>0$ keeps $f$ bounded when $a<0$. The same order-parameter idea, with a sign that alternates between sublattices, describes an antiferromagnet through its staggered magnetization.

where: $m_{\rm st}$ is the staggered (Néel) order parameter; $(-1)^i$ the sublattice sign on a bipartite lattice. Physically: the uniform magnetization stays zero while neighbouring sublattices point oppositely, so a sign-alternating average is the correct order parameter.

How the order parameter behaves across the temperature range and dimensions:

Example 1 — the exact zero average and its field-induced response for a finite ring. Take a periodic 3-spin Ising ring at zero field with $K\equiv\beta J=0.9$ and confirm the single-spin average is exactly zero; then switch on a tiny field and show the response is small and linear.

Step 1. Enumerate all $2^3=8$ configurations. At $h=0$ the flip $s\to-s$ pairs each configuration with an equal-energy partner of opposite $s_0$, so the weighted sum of $s_0$ cancels identically: $\langle s_0\rangle_{h=0}=\mathbf{0}$ (exactly, for every finite $N$).

Step 2. Now add $\beta h=0.02$. Summing $s_0\,e^{K\sum s_is_{i+1}+\beta h\sum s_i}$ over the eight states and dividing by $Z$ gives $\langle s_0\rangle=\mathbf{0.0559}$ — small, positive, and (to leading order) proportional to $h$.

Step 3. Halving the field to $\beta h=0.01$ halves the response to $\mathbf{0.0280}$: the finite ring has a finite susceptibility and $\langle s_0\rangle\to0$ smoothly as $h\to0$. There is no spontaneous magnetization at finite $N$ — exactly as the order-of-limits argument demands.

Example 2 — comparing the onset exponents just below $T_c$. Evaluate the mean-field and exact-2D order parameters at a reduced temperature $t\equiv(T_c-T)/T_c=0.01$ and show how differently they rise.

Step 1. Mean field: $m_0\simeq\sqrt3\,t^{1/2}=\sqrt3\times(0.01)^{1/2}=1.732\times0.1=\mathbf{0.173}$.

Step 2. Exact 2D near $T_c$: $m_0\simeq (c\,t)^{1/8}$. Using the amplitude that makes the two curves meet the definition of $t$, the one-eighth power of a number of order $t$ is much larger — e.g. $(0.01)^{1/8}=\mathbf{0.562}$, more than three times the mean-field value at the same $t$.

Step 3. The lesson: the exact order parameter climbs steeply ($\beta=\tfrac18$) while the mean-field one climbs gently ($\beta=\tfrac12$). Mean-field theory is not just quantitatively off near $T_c$ in 2D — it gets the very shape of the onset wrong, because it discards the fluctuations that dominate below $d_c=4$.

:::physicist[spontaneous magnetization|concept — Weiss 1907, Onsager 1944, Yang 1952|Spontaneous magnetization — a net magnetic moment appearing with no applied field — is the historical prototype of an order parameter. Pierre Weiss's 1907 molecular-field theory first explained it as a self-consistent internal field, predicting a sharp Curie temperature and the square-root onset m₀∼(T_c−T)^{1/2}. The concept was sharpened by Lev Landau in the 1930s into the general order-parameter framework, where spontaneous magnetization is the m≠0 minimum of a symmetry-constrained free energy. Its most dramatic vindication came in 1944, when Lars Onsager solved the two-dimensional Ising model exactly and (in a famous 1949 conference remark, published in full by C. N. Yang in 1952) gave the exact magnetization m₀=[1−sinh⁻⁴(2βJ)]^{1/8}, revealing the exponent β=1/8 — proof that fluctuations below the upper critical dimension change the answer qualitatively, not just quantitatively. Spontaneous magnetization thus links the everyday magnet, the deepest exact result in statistical mechanics, and the modern theory of critical phenomena.]

References. Pathria & Beale, Statistical Mechanics, 3rd ed., Ch. 11–13 · Kardar, Statistical Physics of Fields, Ch. 1–3 · Reif, Fundamentals of Statistical and Thermal Physics, Ch. 10 · Sethna, Statistical Mechanics: Entropy, Order Parameters, and Complexity, Ch. 8–9 · Landau & Lifshitz, Statistical Physics Part 1, Ch. XIV.

Provenance. Equations and numbers are standard textbook results. The Weiss self-consistency $m=\tanh(\beta qJm)$ with $k_BT_c=qJ$ follows from a single-spin mean-field decoupling (Pathria & Beale Ch. 12; Reif Ch. 10); the Landau minimization gives $m_0\propto(T_c-T)^{1/2}$, $\beta=\tfrac12$ (Landau & Lifshitz §142–143); the Onsager/Yang result $m_0=[1-\sinh^{-4}(2\beta J)]^{1/8}$ with $\sinh(2\beta_cJ)=1$, $k_BT_c/J=2/\ln(1+\sqrt2)\approx2.269$, and $\beta=\tfrac18$ (Onsager 1944; Yang 1952; Kardar SPoF Ch. 6). The 3D Ising exponent $\beta\approx0.326$ is the conventional value (Sethna Ch. 12). The full derivation of the 2D magnetization is owned by the Onsager-solution article; here it is recalled by name.

Builds on: Order Parameter · The Ising Model: Definition and Hamiltonian · Mean-Field (Weiss) Theory of the Ising Model · Transfer Matrix Formalism for Lattice Models. Leads to: Onsager Solution of the Two-Dimensional Ising Model · Peierls Argument for Long-Range Order · Domain Walls and Interfacial Excitations · Binary Alloy Order-Disorder Transition.