How a quantum ferromagnet orders: Bloch’s law and the magnetization sphere

← All digests

Family 271 · Mathematical physics
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-25

1. The question

A ferromagnet favors neighboring spins pointing in the same direction. Yet its interaction may have no preferred direction at all. How can an equilibrium system choose one? And if a finite sample preserves rotation symmetry, what becomes of the magnetization that one expects in the ordered phase?

Family 271 addresses these questions for the quantum Heisenberg ferromagnet. Four manuscripts report increasingly detailed statements: existence of magnetized equilibrium states; the leading low-temperature reduction known as Bloch’s law; its first correction from the lattice; and the distribution of magnetization in a symmetric periodic sample.

The last conclusion is especially useful for intuition. A random vector can have mean zero while having a fixed nonzero length, if its direction is uniform. The claimed spherical law makes this picture precise in a quantum system without assuming that its noncommuting spin components admit a joint measurement.

2. Background and essential definitions

Spins and equilibrium

At each lattice site put a spin $S\in\{\tfrac12,1,\tfrac32,\ldots\}$, acting on a space of dimension $2S+1$. Its three self-adjoint components obey

$$[S^a,S^b]=i\sum_c\epsilon_{abc}S^c, \qquad \sum_a(S^a)^2=S(S+1)I.$$

The symbol $[A,B]=AB-BA$ is the commutator; $\epsilon_{abc}$ encodes the signs of cyclic permutations. Thus spin is a quantum vector whose components generally cannot be measured simultaneously.

For the nearest-neighbor model in a finite box $\Lambda\subset\mathbb Z^d$, the energy is

$$H_\Lambda=-\sum_{\{x,y\}}\mathbf S_x\cdot\mathbf S_y, \qquad M_\Lambda^z=\sum_{x\in\Lambda}S_x^z.$$

The sum counts each unordered nearest-neighbor edge once. Its negative sign favors alignment. Lattice spacing, the coupling, Boltzmann’s constant and $\hbar$ are set to one. Temperature is $T=\beta^{-1}$, where $\beta$ is inverse temperature.

A field $h$ adds $-hM_\Lambda^z$. The thermodynamic pressure and spontaneous magnetization are

$$p_\beta(h)=\lim_{\Lambda\uparrow\mathbb Z^d} \frac{\log\operatorname{Tr}e^{-\beta(H_\Lambda-hM_\Lambda^z)}}{\beta|\Lambda|}, \qquad m_S(\beta)=p_\beta'(0+).$$

The volume tends to infinity before the field tends to zero from above. Reversing this order would retain the zero mean of a symmetric finite-volume state. Convexity of pressure guarantees the right derivative.

An infinite-volume state assigns consistent expectations to every local observable. The KMS condition is the equilibrium condition expressed through analytic continuation of time correlations to imaginary time $i\beta$; it generalizes the finite Gibbs formula. Translation invariance says expectations do not depend on where an observable is placed.

Magnons and dispersion

A magnon is a spin-wave excitation that reduces magnetization by one unit. For the three-dimensional unit-coupling nearest-neighbor model its one-particle energy is

$$\varepsilon_S(k)=2S\sum_{j=1}^{3}(1-\cos k_j), \qquad k\in[-\pi,\pi]^3.$$

Near zero momentum it is $S|k|^2+O(|k|^4)$. Treating magnons as independent bosons gives their density

$$n_S^{\mathrm{sw}}(\beta)=\int_{[-\pi,\pi]^3} \frac{d^3k}{(2\pi)^3(e^{\beta\varepsilon_S(k)}-1)}.$$

Rescaling small momenta by $\sqrt{\beta S}$ suggests a $T^{3/2}$ deficit. The challenge is to justify this independent-particle prediction for the interacting finite-spin system and for its zero-field pressure derivative.

3. Development history and previous partial results

Heisenberg’s exchange model supplied a quantum mechanism for ferromagnetism. Bloch’s 1930 analysis predicted the $T^{3/2}$ reduction using long-wavelength spin waves. Holstein and Primakoff introduced a bosonic representation of spin, and Dyson developed the effects of spin-wave interactions.

The power series contains different mechanisms. Terms of orders $T^{5/2}$ and $T^{7/2}$ already arise from the nonquadratic lattice dispersion of ideal magnons. Dyson’s leading interaction contribution to magnetization is of order $T^4$. M2 addresses the first lattice correction, so it should not be described as establishing that interaction term.

Rigorous ordering results also distinguish models. The Mermin–Wagner theorem excludes positive-temperature spontaneous order for finite-range isotropic Heisenberg systems in dimensions one and two. Infrared bounds established ordering for classical continuous spins and important quantum antiferromagnets in higher dimensions. Those theorems did not directly settle the corresponding nearest-neighbor quantum ferromagnet.

Probabilistic representations developed by Powers, Conlon–Solovej, Tóth and Nachtergaele converted quantum spins into permutations, walks and exchange cycles. They yielded pressure and free-energy bounds. Correggi, Giuliani and Seiringer established the leading low-temperature free-energy asymptotic in three dimensions at every fixed spin. This was a major partial result, but a zero-field free-energy estimate alone does not control the field derivative defining spontaneous magnetization.

The spherical distribution had antecedents too. Ueltschi derived its transform conditionally from a proposed law for macroscopic loop lengths. Björnberg, Fröhlich and Ueltschi proved it for the complete-graph quantum Heisenberg model. M3 concerns the three-dimensional nearest-neighbor lattice and does not assume a macroscopic-loop length distribution.

Within this family, M4 supplies the ordering and finite pin estimates; M1 develops the leading magnetization asymptotic; M2 sharpens its precision; M3 adds spatial and pressure arguments for the symmetric zero-field law. Their conclusions and hypotheses differ despite this shared foundation.

4. What the new manuscripts claim

Existence at a fixed positive temperature

M4 states that for every $d\geq3$ and fixed allowed spin $S$, there is a finite $\beta_0(d,S)$ such that every finite $\beta\geq\beta_0(d,S)$ admits a translation-invariant zero-field KMS state $\omega$ with

$$\omega(S_0^z)\geq S/4.$$

This is a statement about a selected ordered infinite-volume state. Averaging it over rotations may erase its mean. M4 also gives a spin-one-half consequence for the particle–hole-symmetric attractive hard-core hopping model: a rotated state has density one half and averaged off-diagonal long-range order. Its uniform-mode occupation is at least $1/32$ of the expected particle number in every finite restriction of that state. This is not a claim about finite-volume Gibbs states of that boson model.

Bloch’s law with spatially anisotropic couplings

M1 works in $\mathbb Z^3$ with a fixed finite-support interaction $J(z)\geq0$, satisfying $J(0)=0$, $J(z)=J(-z)$, whose positive support generates the whole lattice as an additive group. The Hamiltonian is $-\tfrac12\sum_{x,y}J(y-x)\mathbf S_x\cdot\mathbf S_y$.

Define the positive definite matrix

$$D_{S,J}=\frac S2\sum_z J(z)zz^{\mathsf T}.$$

It is the quadratic coefficient of the magnon dispersion. M1 claims

$$\lim_{\beta\to\infty}\beta^{3/2}[S-m_{S,J}(\beta)] =\frac{\zeta(3/2)}{8\pi^{3/2}\sqrt{\det D_{S,J}}}.$$

Here $\zeta(s)=\sum_{n\geq1}n^{-s}$. Spin and interaction remain fixed. The order is volume first, positive field removal second, temperature tending to zero last. For unit nearest-neighbor couplings, $D_{S,J}=SI$.

The first lattice correction

For the three-dimensional nearest-neighbor model, M2 claims that the deficit differs from the full ideal-magnon density by $o((\beta S)^{-5/2})$. Equivalently,

$$\begin{aligned} S-m_S(\beta) ={}&\frac{\zeta(3/2)}{8\pi^{3/2}}(\beta S)^{-3/2}\\ &+\frac{3\zeta(5/2)}{128\pi^{3/2}}(\beta S)^{-5/2} +o((\beta S)^{-5/2}). \end{aligned}$$

The little-$o$ error becomes negligible after division by $(\beta S)^{-5/2}$. The assertion is for every fixed spin, rather than a simultaneous large-spin limit.

A sphere of magnetization directions

M3 fixes sufficiently low but positive temperature in the three-dimensional nearest-neighbor model and takes even periodic cubes of side $L$, volume $V=L^3$. Writing $\mathbf M_L$ for total spin and $m=p_\beta'(0+)>0$, it claims

$$\lim_{L\to\infty,\ L\ \mathrm{even}} \left\langle e^{\mathbf t\cdot\mathbf M_L/V}\right\rangle_{\beta,L} =\frac{\sinh(m|\mathbf t|)}{m|\mathbf t|} \qquad(\mathbf t\in\mathbb R^3).$$

The value at $\mathbf t=0$ is one. The right side is the transform of $m\mathbf n$, with $\mathbf n$ uniform on the unit sphere. In addition, the volume-averaged two-point correlation converges to $m^2$. The formula uses each self-adjoint linear combination separately; it does not posit a joint finite-volume measurement of all three components.

5. Why the advance matters

Importance: 5/5, provisional. These claims address the central ordering problem for the quantum ferromagnet and quantify both the size and distribution of its magnetization. They connect a fixed-temperature equilibrium state, a low-temperature expansion and a symmetric finite-volume observable.

Each stage answers something the previous one leaves unresolved. Nonzero order does not determine its exact thermal reduction. Bloch’s leading coefficient does not determine the first lattice correction. Rotation symmetry does not force a single magnetization radius: it could permit a mixture of several radii. The spherical law excludes that latter possibility in its stated setting.

The general finite-range coefficient also explains how spatial anisotropy enters through $\det D_{S,J}$. Spin-rotation symmetry is retained even when bonds differ by spatial direction. This distinction helps separate physical anisotropies that are often grouped together informally.

6. The main idea

Turn spin information into return probabilities

Represent a spin $S$ by $2S$ spin-one-half slots, projected onto their symmetric subspace. Random exchanges between slots trace cycles around an imaginary-time circle of circumference $\beta$. Each cycle receives an up or down color. A pin requires a cycle passing through a chosen point to be up.

M4 estimates how often interior cycles avoid boundary pins. Adding pins creates an induction: information under a more constrained law controls a less constrained one. After exposing certain cycles, a one-particle walk moves through the available slots. Diffusive estimates and finite-energy flows around obstacles bound its returns. A boundary pinning cost of surface order then permits selection of a magnetized equilibrium state by a field tending to zero.

Recover the coefficient and its correction

For M1, the ideal walk’s return probability behaves as $(4\pi t)^{-3/2}/\sqrt{\det D_{S,J}}$. Returns after one, two and more time periods generate the sum $\sum n^{-3/2}$, explaining the zeta factor. The argument must control late returns in the same repeatedly encountered environment, rather than silently replacing each period with an independent environment.

M2 needs greater accuracy. An obstacle both removes possible transitions and transports the remaining walk when it moves. Keeping these signed effects together produces a cancellation governed by obstacle displacement. A capacity comparison with a free trial potential retains the full Bose sum and pushes the error below the first lattice correction. Expanding the lattice heat kernel then supplies its coefficient.

Establish one radius

M3 combines spatial control across many scales with a pressure argument. The spatial estimates prevent local magnetizations from canceling through increasingly large patterns. The pressure analysis excludes translation-ergodic equilibrium states with different magnetization lengths, using zero-free traces and auxiliary replicated systems.

Finally, rotation representation theory converts concentration of the total-spin magnitude into the spherical transform. The randomness of direction survives; randomness of the macroscopic radius disappears. This extra argument is why the symmetric distribution is more than a consequence of merely constructing one ordered state.

7. Scope, limitations, and open questions

The four settings should remain distinct. M4 covers nearest neighbors in every dimension at least three. M1 permits general nonnegative finite-range couplings in three dimensions. M2 and M3 concern the three-dimensional nearest-neighbor model, with M3 taking even periodic cubes at fixed positive temperature.

The results do not identify a critical temperature, classify all equilibrium states, or establish Dyson’s $T^4$ interaction coefficient. The low-temperature expansion is not asserted uniformly as spin or the interaction changes. The spherical law is a thermodynamic moment statement, not a claim that the finite quantum components commute.

Quantitative temperature thresholds, further expansion terms and extensions of the spherical law to other interactions remain natural questions beyond the claims presented here.

8. Sources and further reading

The historical links in section 3 give Bloch’s prediction, Dyson’s expansion, rigorous fixed-spin free-energy progress and the complete-graph spherical-law predecessor. The four source manuscripts below specify the respective limits, normalizations and companion dependencies. Mathematical claims are reported as stated.

Source manuscripts

• M1. OpenAI, Bloch’s Law for Finite-Range Heisenberg Ferromagnets in Three Dimensions (October 5, 2026).
• M2. OpenAI, The first lattice correction to Bloch’s law (October 5, 2026).
• M3. OpenAI, The spherical magnetization law for the three-dimensional quantum Heisenberg ferromagnet (October 5, 2026).
• M4. OpenAI, Spontaneous magnetization in the quantum Heisenberg ferromagnet (September 24, 2026).

Sources consulted on 25 October 2026 at repository revision 1581d4c. Download the manuscript citations.