Binary Alloy Order-Disorder Transition

Cool a brass coin slowly and something remarkable happens invisibly inside it. Above about 741 K the copper and zinc atoms of $\beta$ -brass sit at random on the sites of a body-centred-cubic lattice — a disordered solid solution in which every site is equally likely to be Cu or Zn. Below that temperature the two species sort themselves onto two interpenetrating simple-cubic sublattices, building a perfectly regular CsCl-type superlattice with copper on one and zinc on the other. This order–disorder transition, sharp and reversible, is one of the oldest and cleanest applications of statistical mechanics to a real material, and it was the concrete problem (Bragg & Williams 1934; Bethe 1935) that motivated much of mean-field theory. The decisive simplification is that the atomic positions are frozen by the crystal lattice; only the chemical identity of each site fluctuates. Each site therefore carries a single binary degree of freedom — exactly the data of an Ising spin. Writing $s_i=+1$ when site $i$ holds an A atom and $s_i=-1$ when it holds a B atom, the configurational energy built from the three nearest-neighbour bond energies $\varepsilon_{AA}$ , $\varepsilon_{BB}$ , $\varepsilon_{AB}$ collapses algebraically onto the Ising Hamiltonian, with a single effective coupling $J$ , an external field $h$ fixed by the composition, and an irrelevant additive constant. Because the mapping is exact, every Ising tool transfers wholesale. The Bragg–Williams theory of ordering is Weiss mean-field theory; the Bethe approximation is Bethe–Peierls; the two-dimensional case is Onsager-solvable; and real three-dimensional alloys belong to the Ising universality class with its measured critical exponents. Only the sign of the effective coupling decides whether the alloy orders into a superlattice (unlike neighbours favoured) or unmixes into A-rich and B-rich domains (like neighbours favoured) — one model, two opposite faces. The observed $\lambda$ -shaped specific-heat anomaly and Ising critical scattering in CuZn are a textbook confirmation that universality reaches even into metallurgy.

⚡ Key Takeaways

A substitutional binary alloy AB on a rigid lattice — each site occupied by an A or a B atom — maps exactly onto the spin-½ Ising model by setting $s_i=+1$ for A and $s_i=-1$ for B.
The three pairwise bond energies $\varepsilon_{AA},\varepsilon_{BB},\varepsilon_{AB}$ collapse into one effective coupling $J=\tfrac14(2\varepsilon_{AB}-\varepsilon_{AA}-\varepsilon_{BB})$ plus a field and constant; only the contrast $w=\varepsilon_{AB}-\tfrac12(\varepsilon_{AA}+\varepsilon_{BB})=2J$ matters.
When unlike-neighbour bonds are favoured ( $2\varepsilon_{AB}<\varepsilon_{AA}+\varepsilon_{BB}$ ) the alloy orders into two interpenetrating sublattices below a critical temperature — the classic order–disorder transition of $\beta$ -brass (CuZn), continuous and in the 3D Ising universality class.
The order parameter is the long-range order parameter $\eta$ (the sublattice-occupancy difference); mean-field (Bragg–Williams) theory gives $k_BT_c=z|J|$ , which overshoots the true $T_c$ .
The same lattice maps onto the lattice gas, so alloy ordering, magnetism, and gas–liquid condensation are three faces of one Ising model.
Key idea: chemical identity → spin, three bond energies → one coupling $J$ , composition → field $h$ . Order–disorder of the alloy is simply the Ising order parameter $\eta$ turning on below $T_c$ .

💡 The Core Insight

A two-species occupation problem on a rigid lattice is identical to one Ising spin per site: chemical identity maps to spin, the three bond energies collapse to one effective coupling $J=\tfrac14(2\varepsilon_{AB}-\varepsilon_{AA}-\varepsilon_{BB})$ , and composition becomes an external field $h$ . Order–disorder of the alloy is the Ising order parameter turning on below $T_c$ .

The physical picture

Picture a chessboard where every square must hold a copper or a zinc atom. At high temperature thermal agitation scrambles them: each square is a 50/50 coin toss, no pattern survives — that is the disordered phase. As you cool, the small energetic preference for Cu–Zn contacts over Cu–Cu or Zn–Zn ones nudges the atoms toward a perfect black-Cu/white-Zn checkerboard. That emergent two-sublattice pattern is the ordered phase, and the sharpness with which it appears at $T_c$ is a genuine thermodynamic phase transition, not a gradual crossover. The crucial bookkeeping is that only the contrast between unlike and like bonds matters. Add the same constant to all three bond energies and nothing physical changes — precisely the freedom that lets you absorb $\varepsilon_{AA},\varepsilon_{BB},\varepsilon_{AB}$ into a single Ising coupling $J$ plus a field and a constant. The field encodes how far the composition strays from equiatomic; at 50:50 stoichiometry it can be set to zero and the transition is symmetric.

Figure

The order–disorder transition — random alloy above T_c versus the two-sublattice superlattice below

Why is it the same mathematics as magnetism? Both problems read: each site picks one of two states, and neighbours prefer to agree (or to disagree). The magnet's up/down is the alloy's A/B. Spontaneous magnetization becomes spontaneous chemical ordering, and universality then guarantees that the critical behaviour of $\beta$ -brass near $T_c$ matches that of a 3D Ising ferromagnet despite utterly different microscopics. Ordering and unmixing are mirror images: flip the sign of the effective coupling and the checkerboard becomes a phase-separated blob — what the alloy literature calls a miscibility gap. One Ising model, two signs of $J$ .

Figure

The long-range order parameter η(T) — Bragg–Williams mean field versus the sharper real transition

The core equations

The starting point is the configurational energy of the alloy, obtained by counting nearest-neighbour bonds of each of the three types. Only bonds contribute; the kinetic and vibrational energies are the same in both phases and drop out of the ordering problem.

E=\sum_{\langle ij\rangle}\Big[\varepsilon_{AA}\,n_i^A n_j^A+\varepsilon_{BB}\,n_i^B n_j^B+\varepsilon_{AB}\,(n_i^A n_j^B+n_i^B n_j^A)\Big] \qquad (1)

where: $\langle ij\rangle$ runs over nearest-neighbour bonds; $n_i^A,n_i^B\in\{0,1\}$ are occupation indicators for an A or B atom at site $i$ with $n_i^A+n_i^B=1$ ; and $\varepsilon_{AA},\varepsilon_{BB},\varepsilon_{AB}$ are the bond energies. Physically, an ordering alloy is one that lowers its energy by surrounding each atom with the other species. Each site holds exactly one atom, so its occupation is a single binary variable — an Ising spin. The substitution is the whole trick: it converts chemistry into spin language.

n_i^A=\frac{1+s_i}{2},\qquad n_i^B=\frac{1-s_i}{2},\qquad s_i=\pm1 \qquad (2)

where: $s_i=+1$ if site $i$ holds an A atom and $s_i=-1$ if it holds a B atom. The identity $n_i^A+n_i^B=1$ is automatic, and $n_i^A n_i^A=n_i^A$ (idempotence) is respected because $s_i^2=1$ . Substituting (2) into (1) and collecting powers of the spins yields the Ising Hamiltonian exactly (derivation below). The three bond energies survive only through one combination.

E=-J\sum_{\langle ij\rangle}s_i s_j-h\sum_i s_i+E_0,\qquad J=\tfrac14\big(2\varepsilon_{AB}-\varepsilon_{AA}-\varepsilon_{BB}\big) \qquad (3)

where: $J$ is the effective coupling; $h=-\tfrac{z}{4}(\varepsilon_{AA}-\varepsilon_{BB})$ (augmented by any imposed chemical-potential difference at fixed composition) is the field; $E_0$ is a configuration-independent constant; and $z$ is the coordination number. Reading it physically: everything about ordering is carried by the single contrast $2\varepsilon_{AB}-\varepsilon_{AA}-\varepsilon_{BB}$ , not by the individual energies. It is often clearer to write the contrast directly. Define the ordering energy $w$ ; then $J=w/2$ in the convention above.

w\equiv\varepsilon_{AB}-\tfrac12(\varepsilon_{AA}+\varepsilon_{BB})=2J \qquad (4)

where: $w<0$ (unlike bonds cheaper) drives ordering into a superlattice; $w>0$ (like bonds cheaper) drives phase separation; $w=0$ is an ideal solution that stays random at all temperatures. The sign of $w$ is the single switch between the alloy's two possible fates. Order is measured by the long-range order parameter, the difference between the two sublattices. It is the alloy's staggered magnetization.

\eta=\frac{w_A^{(\alpha)}-w_A^{(\beta)}}{w_A^{(\alpha)}+w_A^{(\beta)}} \qquad (5)

where: $w_A^{(\alpha)},w_A^{(\beta)}$ are the fractions of A atoms on sublattices $\alpha$ and $\beta$ . Then $\eta=1$ is perfect order (one sublattice pure A, the other pure B) and $\eta=0$ is complete disorder. Note $\eta$ is long-range order — distinct from short-range order, the local excess of unlike neighbours, which persists above $T_c$ . Treating the neighbours in mean field (each replaced by its average) gives a self-consistent equation for $\eta$ , the Bragg–Williams theory. It is Weiss mean-field theory wearing metallurgical clothes.

\eta=\tanh\!\Big(\frac{z\,|J|\,\eta}{k_BT}\Big),\qquad k_BT_c^{\mathrm{MF}}=z\,|J| \qquad (6)

where: $z$ is the coordination number, $k_B$ the Boltzmann constant, $T$ the temperature. A non-trivial solution first appears when the slope of the tanh at $\eta=0$ reaches one, giving $k_BT_c=z|J|$ . This mean-field $T_c$ overestimates the true value because it ignores the fluctuations that locally oppose ordering.

Derivation AFrom three bond energies to one Ising coupling

The goal is to prove that the bond-counting configurational energy of a two-species alloy is exactly an Ising Hamiltonian, and to read off the effective coupling, field, and constant in terms of the microscopic bond energies. We substitute the occupation–spin relations $n^A=(1+s)/2$ , $n^B=(1-s)/2$ into every bond term, expand, and collect the pieces proportional to $s_is_j$ (the coupling), to $s_i$ alone (the field), and to the identity (the constant). The essential lesson that falls out is that of the three parameters $\varepsilon_{AA},\varepsilon_{BB},\varepsilon_{AB}$ only one combination controls the physics of ordering, so the entire three-parameter alloy problem is a one-parameter Ising problem plus a trivial energy shift.

Step 1. Expand each bond in spins: $\varepsilon_{AA}n_i^An_j^A=\tfrac{\varepsilon_{AA}}{4}(1+s_i)(1+s_j)=\tfrac{\varepsilon_{AA}}{4}(1+s_i+s_j+s_is_j)$ , and likewise $\varepsilon_{BB}n_i^Bn_j^B=\tfrac{\varepsilon_{BB}}{4}(1-s_i-s_j+s_is_j)$ , while $\varepsilon_{AB}(n_i^An_j^B+n_i^Bn_j^A)=\tfrac{\varepsilon_{AB}}{2}(1-s_is_j)$ . Why: the cross-species term is symmetric in $i\leftrightarrow j$ , so its odd-in- $s$ pieces cancel, leaving only $1$ and $s_is_j$ .

Step 2. Collect the $s_is_j$ coefficient in a single bond: $\tfrac14(\varepsilon_{AA}+\varepsilon_{BB})-\tfrac12\varepsilon_{AB}$ . Comparing with the Ising form $-J\,s_is_j$ gives $-J=\tfrac14(\varepsilon_{AA}+\varepsilon_{BB})-\tfrac12\varepsilon_{AB}$ , i.e. $J=\tfrac14(2\varepsilon_{AB}-\varepsilon_{AA}-\varepsilon_{BB})$ . Why: this is the only combination of the three energies that multiplies the spin–spin product, hence the only one that can drive cooperative ordering.

Step 3. Collect the linear pieces: each spin $s_i$ appears with coefficient $\tfrac14(\varepsilon_{AA}-\varepsilon_{BB})$ per bond, and each site sits in $z$ bonds, giving a field term $-h\sum_i s_i$ with $h=-\tfrac{z}{4}(\varepsilon_{AA}-\varepsilon_{BB})$ . Why: this field vanishes for the symmetric case $\varepsilon_{AA}=\varepsilon_{BB}$ and, at fixed composition, is fixed by the chemical-potential (Lagrange) constraint that conserves the A:B ratio.

Step 4. Collect the constants: $E_0=\tfrac{N_b}{4}(\varepsilon_{AA}+\varepsilon_{BB}+2\varepsilon_{AB})$ with $N_b=zN/2$ the number of bonds. Why: a term independent of all spins only shifts the free-energy zero and cannot affect the transition.

Step 5. Assemble: $E=-J\sum_{\langle ij\rangle}s_is_j-h\sum_i s_i+E_0$ , exactly the Ising model. The alloy partition function equals the Ising partition function times $e^{-\beta E_0}$ , so every Ising result transfers. Why: an exact algebraic identity, not an approximation, is what makes the correspondence and its universality watertight. *Verified: verify_aq-sm-binary-alloy-order-disorder.py substitutes $n^A=(1+s)/2,n^B=(1-s)/2$ , expands each bond symbolically, and confirms the $s_is_j$ coefficient is $\tfrac14(\varepsilon_{AA}+\varepsilon_{BB})-\tfrac12\varepsilon_{AB}=-J$ , the per-bond linear coefficient is $\tfrac14(\varepsilon_{AA}-\varepsilon_{BB})$ , and $w=\varepsilon_{AB}-\tfrac12(\varepsilon_{AA}+\varepsilon_{BB})=2J$ .*

Derivation BBragg–Williams free energy and the continuous transition

The goal is to build the mean-field (Bragg–Williams) free energy $f(\eta)$ for the stoichiometric, zero-field alloy and show that it produces a continuous order–disorder transition at $k_BT_c=z|J|$ , with the mean-field order-parameter exponent $\beta=\tfrac12$ . We combine the mean-field interaction energy (each bond contributing its average $-|J|\eta^2$ ) with the configurational entropy of an arrangement of order $\eta$ , then minimise $f=E-TS$ over $\eta$ . Expanding the resulting free energy in powers of $\eta$ exposes a Landau functional whose quartic coefficient is positive and whose quadratic coefficient changes sign at $T_c$ — the signature of a second-order transition — and lets us read the exponent directly.

Step 1. Mean-field energy per site: $E/N=-\tfrac12 z|J|\eta^2$ , since there are $zN/2$ bonds and each carries average $-|J|\eta^2$ . Why: replacing each neighbour spin by its mean $\eta$ turns the two-spin correlation $\langle s_is_j\rangle$ into $\eta^2$ .

Step 2. Configurational entropy per site: $S/(k_BN)=-\big[\tfrac{1+\eta}{2}\ln\tfrac{1+\eta}{2}+\tfrac{1-\eta}{2}\ln\tfrac{1-\eta}{2}\big]$ , the binary mixing entropy for order $\eta$ ; it equals $\ln 2$ at $\eta=0$ and $0$ at $\eta=1$ . Why: a site is A-on- $\alpha$ with probability $(1+\eta)/2$ and B-on- $\alpha$ with probability $(1-\eta)/2$ .

Step 3. Minimise $f=E-TS$ : setting $\partial f/\partial\eta=0$ gives $z|J|\eta=k_BT\,\mathrm{arctanh}\,\eta$ , i.e. $\eta=\tanh(z|J|\eta/k_BT)$ — the self-consistency equation. Why: the entropy derivative $\tfrac{k_B}{2}\ln\tfrac{1-\eta}{1+\eta}=-k_B\,\mathrm{arctanh}\,\eta$ balances the energy derivative.

Step 4. Expand: $f/N\simeq\tfrac12(k_BT-z|J|)\eta^2+\tfrac{k_BT}{12}\eta^4+\cdots$ . The quartic coefficient $k_BT/12>0$ , and the quadratic coefficient changes sign at $k_BT_c=z|J|$ — a continuous transition. Why: no cubic term survives (by the $\eta\to-\eta$ symmetry of the equiatomic case), which is exactly the Landau condition for second order.

Step 5. Near $T_c$ : minimising $\tfrac12(k_BT-z|J|)\eta^2+\tfrac{k_BT}{12}\eta^4$ gives $\eta^2=3(z|J|-k_BT)/k_BT\propto (T_c-T)$ , so $\eta\propto(T_c-T)^{1/2}$ and $\beta=\tfrac12$ . Why: the true 3D alloy has $\beta\approx0.326$ (Ising universality) — the mean-field $\tfrac12$ is qualitatively right but quantitatively too large, because it omits critical fluctuations. *Verified: verify_aq-sm-binary-alloy-order-disorder.py series-expands $f(\eta)$ , confirms quadratic coefficient $\tfrac12(k_BT-z|J|)$ vanishing at $k_BT=z|J|$ , quartic coefficient $k_BT/12>0$ , $\eta^2=3(z|J|-k_BT)/k_BT$ giving $\beta=\tfrac12$ , and the entropy endpoints $\ln 2$ and $0$ .*

Limiting cases

The effective ordering energy $w=\varepsilon_{AB}-\tfrac12(\varepsilon_{AA}+\varepsilon_{BB})=2J$ controls the phase behaviour:

Case	Result	Meaning
Ordering tendency ( $2\varepsilon_{AB}<\varepsilon_{AA}+\varepsilon_{BB}$ , $w<0$ )	two-sublattice superlattice below $T_c$	Unlike neighbours favoured: $\beta$ -brass CuZn orders into the CsCl (B2) structure.
Phase separation ( $2\varepsilon_{AB}>\varepsilon_{AA}+\varepsilon_{BB}$ , $w>0$ )	A-rich / B-rich domains, a miscibility gap	Like neighbours favoured: clustering, spinodal decomposition (ferromagnetic mapping).
Ideal solution ( $2\varepsilon_{AB}=\varepsilon_{AA}+\varepsilon_{BB}$ , $w=0$ )	random mixing at all $T$ , $\eta=0$	$J=0$ : no transition; configurational entropy alone fixes a disordered solid solution.
Zero temperature, stoichiometric	$\eta\to1$ , $S\to0$	Perfect long-range order: one sublattice all A, the other all B.
High temperature $T\gg T_c$	$\eta\to0$ , $S\to k_BN\ln2$	Complete disorder; each site equally likely A or B, maximal mixing entropy.

⚠️ Not all three bond energies matter — only their contrast

A frequent error is to think a strong A–B attraction alone drives ordering. What matters is whether unlike bonds are favoured relative to the average of like bonds: only the combination $2\varepsilon_{AB}-\varepsilon_{AA}-\varepsilon_{BB}$ (equivalently $w=\varepsilon_{AB}-\tfrac12(\varepsilon_{AA}+\varepsilon_{BB})$ ) enters as the coupling $J$ . Adding the same constant to all three energies changes nothing physical. And ordering ( $w<0$ , antiferromagnetic mapping on a bipartite lattice) and phase separation ( $w>0$ , ferromagnetic mapping) are opposite limits of one model — do not conflate them.

Worked examples

Example 1 — inferring the effective coupling of $\beta$ -brass from its transition temperature. $\beta$ -brass (CuZn) is BCC with coordination number $z=8$ and orders at $T_c\approx741$ K. Estimate the effective coupling $|J|/k_B$ implied by the mean-field (Bragg–Williams) relation, and note why the real coupling is larger.

Step 1. Bragg–Williams gives $k_BT_c=z|J|$ , so $|J|/k_B=T_c/z=741\ \text{K}/8=\mathbf{92.6\ \text{K}}$ — that is, $|J|\approx92.6\,k_B\approx1.28\times10^{-21}$ J per bond.

Step 2. Check the self-consistency: with this coupling the mean-field ordering temperature is $k_BT_c=z|J|=8\times92.6\,k_B=740.8\,k_B$ , i.e. $T_c=\mathbf{740.8\ \text{K}}$ , recovering the input to rounding.

Step 3. Because mean field overestimates $T_c$ for a given coupling (it ignores fluctuations that oppose order), the value $92.6$ K is a lower bound on the true $|J|/k_B$ : the real coupling needed to sustain order to $741$ K against fluctuations is somewhat larger. A Bethe–Peierls or series estimate, which capture short-range order, would infer a bigger $|J|$ from the same measured $T_c$ .

Example 2 — mean field overshoots on the 2D square lattice. Compare the Bragg–Williams $T_c$ with the exact Onsager value for the two-dimensional square-lattice Ising model, where $z=4$ .

Step 1. Mean field: $k_BT_c^{\mathrm{MF}}=z|J|=4|J|$ , i.e. $k_BT_c^{\mathrm{MF}}/|J|=\mathbf{4.000}$ .

Step 2. Exact (Onsager): $k_BT_c/|J|=2/\ln(1+\sqrt2)$ . Since $\ln(1+\sqrt2)=\ln(2.41421)=0.88137$ , this is $2/0.88137=\mathbf{2.2692}$ .

Step 3. The mean-field value overshoots by a factor $4.000/2.2692=\mathbf{1.763}$ — a $76\%$ error. Fluctuations, entirely neglected by Bragg–Williams, cut the true ordering temperature almost in half in two dimensions. The discrepancy shrinks as dimensionality (and $z$ ) grow, vanishing above the upper critical dimension $d=4$ .

william shockley

1910–1989 · William Bradford Shockley was an American physicist whose 1938 review with Frank Nix, "Order–disorder transformations in alloys" (Reviews of Modern Physics 10, 1), organised and advanced the statistical theory of the very transition treated here — the ordering of binary alloys such as β-brass onto sublattices below a critical temperature. Their article synthesised the Bragg–Williams mean-field picture, the Bethe–Peierls cluster corrections, and the experimental calorimetric and diffraction evidence into a coherent framework, and it remains a landmark in the physics of metals. Shockley worked at Bell Telephone Laboratories, where his interest in the electronic and configurational order of solids led him to solid-state device physics; in 1947 he, John Bardeen, and Walter Brattain invented the transistor, for which the three shared the 1956 Nobel Prize in Physics. His order-disorder work exemplifies how the Ising mapping — chemistry recast as spins — turned a metallurgical puzzle into a solvable problem in statistical mechanics.

⏳ History

1920

— Lenz proposes the lattice model of two-state spins later named for Ising

1925

— Ising solves the one-dimensional chain and finds no ordering at finite temperature

1934

— Bragg and Williams give the mean-field theory of alloy order–disorder

1935

— Bethe introduces the cluster (Bethe–Peierls) approximation for the same problem

1938

— Nix and Shockley review order–disorder transformations in alloys

1944

— Onsager solves the two-dimensional Ising model exactly, fixing the true T꜀

1952

— Lee and Yang formalise the lattice-gas correspondence, unifying alloy, magnet and fluid

Where this lives today

The order–disorder transition is a workhorse of modern materials science and computational thermodynamics. The CALPHAD method, which underpins alloy design across the aerospace and electronics industries, models ordering free energies with cluster-expansion generalisations of the Bragg–Williams and Bethe–Peierls schemes fitted to first-principles (density-functional) bond energies, so that phase diagrams of multicomponent alloys can be predicted before a sample is ever cast. Intermetallic superalloys built on ordered $\gamma'$ (Ni $_3$ Al) phases, the ordering of lithium in battery-electrode lattices, and the site occupancy of high-entropy alloys are all analysed with the same Ising-mapped statistical mechanics. And the deep lesson — that a metallurgical order–disorder transition shares its critical exponents with a magnet and a fluid — remains one of the clearest experimental demonstrations of universality, checked by neutron and X-ray critical scattering in CuZn and its cousins.

Going further

References. Pathria & Beale, Statistical Mechanics, 3rd ed., §12.4 (Ising model and the alloy/lattice-gas correspondence) · Kardar, Statistical Physics of Particles, Ch. 7; Statistical Physics of Fields, Ch. 1–3 · Reif, Fundamentals of Statistical and Thermal Physics, Ch. 10 (order–disorder) · Sethna, Statistical Mechanics: Entropy, Order Parameters, and Complexity, Ch. 8 · Landau & Lifshitz, Statistical Physics Part 1, §142–144 (ordering of alloys) · W. L. Bragg & E. J. Williams, Proc. R. Soc. A 145, 699 (1934) · H. A. Bethe, Proc. R. Soc. A 150, 552 (1935) · F. C. Nix & W. Shockley, Rev. Mod. Phys. 10, 1 (1938).

Provenance. The exact mapping and $J=\tfrac14(2\varepsilon_{AB}-\varepsilon_{AA}-\varepsilon_{BB})$ are standard (Pathria & Beale §12.4; Landau–Lifshitz Part 1 §142–144; Reif Ch. 10). The Bragg–Williams free energy, self-consistency $\eta=\tanh(z|J|\eta/k_BT)$ , and $k_BT_c=z|J|$ are the Weiss mean-field results. The exact 2D benchmark $k_BT_c/|J|=2/\ln(1+\sqrt2)\approx2.2692$ is Onsager (1944). The $\beta$ -brass $T_c\approx741$ K ( $\approx468\,^\circ$ C) and its $\lambda$ -anomaly are standard calorimetric data; the 3D Ising exponent $\beta\approx0.326$ contrasts with the mean-field $\tfrac12$ . All algebra (spin re-expansion, Landau expansion, entropy limits, the Onsager and CuZn numbers) is recomputed in verify_aq-sm-binary-alloy-order-disorder.py.

Builds on: The Ising Model: Definition and Hamiltonian · Order Parameter and Spontaneous Magnetization · Mean-Field (Weiss) Theory of the Ising Model · Lattice Gas Model and Ising Mapping. Leads to: Bethe-Peierls Approximation · Onsager Solution of the Two-Dimensional Ising Model · High- and Low-Temperature Series Expansions.

Why is a fixed-lattice alloy exactly an Ising model, not just approximately?

Because the atomic positions are frozen: the only degree of freedom left at each site is a binary chemical label, which is precisely an Ising spin. Substituting $n^A=(1+s)/2$ , $n^B=(1-s)/2$ into the nearest-neighbour bond energy is an exact algebraic identity, not a truncation — it maps the three-parameter bond-energy model onto $-J\sum s_is_j-h\sum s_i+E_0$ with no remainder. The kinetic and vibrational energies are essentially the same in the ordered and disordered phases and cancel from the ordering problem. That is why every Ising result, including its critical exponents and universality, applies verbatim to the alloy.

What is the difference between long-range and short-range order?

Long-range order, measured by $\eta$ , is the systematic difference in composition between the two sublattices — it is the alloy's staggered magnetization and it vanishes exactly at $T_c$ . Short-range order is the local, statistical excess of unlike over like nearest neighbours; it reflects the same energetic preference but survives above $T_c$ , decaying smoothly with temperature. Just above $T_c$ an alloy can still have substantial short-range order (atoms locally prefer unlike neighbours) even though the global sublattice pattern has melted. Diffuse (Huang) scattering measures short-range order; superlattice Bragg peaks measure long-range order.

\beta

-brass mapped to a ferromagnet or an antiferromagnet?

An ordering alloy on a bipartite lattice maps most naturally onto an Ising antiferromagnet: unlike neighbours are favoured, exactly as antiferromagnetic coupling favours anti-aligned neighbouring spins, and the alloy's long-range order parameter is the antiferromagnet's staggered magnetization. Equivalently, one can perform a sublattice spin-flip that turns the antiferromagnet into a ferromagnet with a redefined order parameter — the two descriptions are related by a gauge transformation and give the same thermodynamics. A phase-separating (clustering) alloy, by contrast, maps onto a ferromagnet, with the miscibility gap playing the role of the coexistence curve.

Why does the transition become first order away from stoichiometry?

At exactly equiatomic composition and zero effective field, the $\eta\to-\eta$ symmetry forbids a cubic term in the Landau free energy, so the transition is continuous. Away from stoichiometry the composition acts as a field that breaks this symmetry, and on many lattices additional order parameters or a cubic invariant appear; a nonzero cubic term generically produces a first-order transition with a discontinuous jump in $\eta$ and a two-phase coexistence region. This is the alloy analogue of how a magnetic field or a non-symmetric order parameter turns a continuous transition first order. The detailed outcome depends on lattice geometry and the range of interactions.

How is this related to the lattice gas and to condensation?

Identically. The lattice gas assigns each site an occupation $n_i\in\{0,1\}$ (particle or hole), with the same $n=(1+s)/2$ substitution mapping it to the Ising model; the particle–hole symmetry is the spin-flip $Z_2$ symmetry, and the gas–liquid critical point is the magnetic critical point. So the binary alloy (A/B), the lattice gas (particle/hole), and the magnet (up/down) are three realisations of one Ising model. This is why the critical exponents of a condensing fluid, an ordering alloy, and a ferromagnet all coincide — the deep content of universality, anticipated by Lee and Yang and by Landau.

How accurate is the Bragg–Williams estimate of

T_c

for real alloys?

Qualitatively good, quantitatively rough. Bragg–Williams gives $k_BT_c=z|J|$ , which overestimates the true value because it ignores the correlations (short-range order) that fight global ordering — on the 2D square lattice it overshoots by $76\%$ ( $4|J|$ versus the exact $2.269|J|$ ). In three dimensions with large coordination number the error is smaller, and the Bethe–Peierls cluster method, high- and low-temperature series, or Monte Carlo bring quantitative agreement with calorimetric $T_c$ values. The mean-field exponent $\beta=\tfrac12$ is also wrong: real 3D alloys show the Ising $\beta\approx0.326$ .