Ferromagnetic Dyson Models

• Ferromagnetic Dyson models are long-range ferromagnetic systems characterized by polynomially decaying interactions that enable one-dimensional phase transitions and distinct Gibbs states.
• They exhibit a dual framework, with a transfer-operator formulation leading to continuous g-measures and a bosonic spin-wave representation that clarifies non-Hermitian dynamics.
• The models serve as a benchmark for studying renormalization, boundary effects, and weak field phenomena, advancing insights into both statistical mechanics and quantum spin systems.

Searching arXiv for papers on ferromagnetic Dyson models and closely related transfer-operator/Gibbs results. arXiv search query: ferromagnetic Dyson model transfer operator Gibbs g-measure random-cluster Dyson hierarchical Ferromagnetic Dyson models are a class of long-range ferromagnetic systems associated with Freeman Dyson’s work, but the term has two distinct established usages. In the dominant statistical-mechanical usage, it denotes one-dimensional Ising ferromagnets with pair couplings decaying polynomially, typically $J(|i-j|)=|i-j|^{-\alpha}$ or $\beta |i-j|^{-\alpha}$, whose slow decay allows genuine phase transitions in one dimension. In a second, historically earlier usage, it refers to Dyson’s 1956 bosonic representation of the Heisenberg ferromagnet, in which spin-wave excitations are described by a non-unitary but physically self-adjoint quasiparticle formalism. These traditions are mathematically and physically distinct, and both must be separated from the “Dyson model” of interacting Brownian motions used in random-matrix theory, including elliptic Dyson models, which are not ferromagnetic spin systems (Bissacot et al., 2017, Jones-Smith, 2013, Katori, 2016).

1. Terminology and canonical formulations

In the long-range Ising tradition, the ferromagnetic Dyson model is defined on $\Omega=\{-1,+1\}^{\mathbb Z}$ or on the half-line $\{-1,+1\}^{\mathbb N}$, with pair interaction

$\Phi_A^D(\omega)= \begin{cases} -\dfrac{1}{|i-j|^\alpha}\,\omega_i\omega_j, & A=\{i,j\},\[1ex] 0, & \text{otherwise}, \end{cases}$

or, more generally,

$$H(\omega)=-\sum_{i>j}J(|i-j|)\omega_i\omega_j-\sum_i h_i\omega_i, \qquad J(n)=n^{-\alpha}.$$

Ferromagneticity is encoded by $J(k)\ge 0$, so aligned spins are energetically favored. In the standard Dyson case the key regime is $1<\alpha\le 2$, because the interaction is summable but still sufficiently long-ranged to sustain low-temperature ordering in one dimension (Bissacot et al., 2017).

The same models admit a one-sided thermodynamic-formalism representation. On $X=S^{\mathbb N}$ with $S=\{-1,+1\}$, the potential

$\beta |i-j|^{-\alpha}$0

generates the transfer operator

$\beta |i-j|^{-\alpha}$1

where $\beta |i-j|^{-\alpha}$2 is the left shift. The pair interaction and the one-point potential are linked by

$\beta |i-j|^{-\alpha}$3

This one-sided representation is central in the modern analysis of eigenfunctions, $\beta |i-j|^{-\alpha}$4-measures, and the comparison between half-line and full-line models (Johansson et al., 2023).

A different usage of “ferromagnetic Dyson model” arises from Dyson’s spin-wave theory for the Heisenberg ferromagnet. There the starting point is not a long-range Ising chain but a short-range quantum spin system rewritten in terms of bosonic quasiparticles. In that framework, “Dysonization” means a non-unitary bosonization in which the Hamiltonian is not Hermitian with respect to the standard bosonic inner product, yet is self-adjoint in the induced physical metric (Jones-Smith, 2013). The coexistence of these meanings is a persistent source of confusion.

2. Long-range Ising ferromagnets and phase-transition structure

The statistical-mechanical significance of ferromagnetic Dyson models is that they realize one-dimensional phase coexistence through sufficiently slow power-law decay. Dyson’s 1969 theorem established low-temperature phase coexistence for $\beta |i-j|^{-\alpha}$5 with $\beta |i-j|^{-\alpha}$6, and later work sharpened the picture, including the marginal case $\beta |i-j|^{-\alpha}$7. In the notation of the review literature, for $\beta |i-j|^{-\alpha}$8 there exists $\beta |i-j|^{-\alpha}$9 such that for $\Omega=\{-1,+1\}^{\mathbb Z}$0,

$\Omega=\{-1,+1\}^{\mathbb Z}$1

Thus the model has spontaneous magnetization and distinct plus and minus extremal Gibbs states despite being one-dimensional (Bissacot et al., 2017).

The DLR formalism is standard. For a finite $\Omega=\{-1,+1\}^{\mathbb Z}$2, the Gibbs specification is

$\Omega=\{-1,+1\}^{\mathbb Z}$3

and the extremal translation-invariant Gibbs states are obtained by monotone limits

$\Omega=\{-1,+1\}^{\mathbb Z}$4

Because the interaction is attractive/FKG, every Gibbs measure lies between them in stochastic order. This monotonicity is not merely technical; it underlies comparison arguments for decimation, random-cluster representations, and one-sided/two-sided criticality.

The phase diagram changes sharply when the decay becomes faster. If $\Omega=\{-1,+1\}^{\mathbb Z}$5, there is no phase transition. At $\Omega=\{-1,+1\}^{\mathbb Z}$6, the model exhibits the Thouless effect, with discontinuous magnetization but continuous energy density at the transition, together with an intermediate phase with slow decay of correlations. This places the Dyson chain at an interface between one-dimensional and higher-dimensional Ising phenomenology.

A common misconception is that one dimension precludes ferromagnetic ordering. Ferromagnetic Dyson models are the standard counterexample: the dimension remains one, but the effective range of the interaction changes the infrared structure enough to reproduce phenomena usually associated with short-range systems in $\Omega=\{-1,+1\}^{\mathbb Z}$7.

3. Transfer operators, eigenfunctions, and the $\Omega=\{-1,+1\}^{\mathbb Z}$8-measure problem

The one-sided transfer-operator formulation turns the Dyson interaction into a problem about eigenmeasures and eigenfunctions of $\Omega=\{-1,+1\}^{\mathbb Z}$9. For the potential

$\{-1,+1\}^{\mathbb N}$0

the relevant tail sequence is

$\{-1,+1\}^{\mathbb N}$1

The main regularity threshold is square summability of variations,

$\{-1,+1\}^{\mathbb N}$2

For the Dyson interaction $\{-1,+1\}^{\mathbb N}$3, one has $\{-1,+1\}^{\mathbb N}$4, so square summability is equivalent to

$\{-1,+1\}^{\mathbb N}$5

Under this condition, a continuous positive eigenfunction exists throughout the whole subcritical region: $\{-1,+1\}^{\mathbb N}$6 This is precisely the regime where the two-sided model is in its uniqueness phase, even though for $\{-1,+1\}^{\mathbb N}$7 the potential need not have summable variations (Johansson et al., 2023).

The transfer-operator significance of such an eigenfunction $\{-1,+1\}^{\mathbb N}$8 is substantial. If $\{-1,+1\}^{\mathbb N}$9 and $\Phi_A^D(\omega)= \begin{cases} -\dfrac{1}{|i-j|^\alpha}\,\omega_i\omega_j, & A=\{i,j\},\[1ex] 0, & \text{otherwise}, \end{cases}$0 is an eigenmeasure, then

$\Phi_A^D(\omega)= \begin{cases} -\dfrac{1}{|i-j|^\alpha}\,\omega_i\omega_j, & A=\{i,j\},\[1ex] 0, & \text{otherwise}, \end{cases}$1

is continuous and satisfies $\Phi_A^D(\omega)= \begin{cases} -\dfrac{1}{|i-j|^\alpha}\,\omega_i\omega_j, & A=\{i,j\},\[1ex] 0, & \text{otherwise}, \end{cases}$2. In that case the equilibrium measure $\Phi_A^D(\omega)= \begin{cases} -\dfrac{1}{|i-j|^\alpha}\,\omega_i\omega_j, & A=\{i,j\},\[1ex] 0, & \text{otherwise}, \end{cases}$3 becomes a continuous $\Phi_A^D(\omega)= \begin{cases} -\dfrac{1}{|i-j|^\alpha}\,\omega_i\omega_j, & A=\{i,j\},\[1ex] 0, & \text{otherwise}, \end{cases}$4-measure. The modern program in this area is therefore to determine when the line Gibbs state induces a continuous one-sided stochastic process.

Recent work extends this picture beyond the $\Phi_A^D(\omega)= \begin{cases} -\dfrac{1}{|i-j|^\alpha}\,\omega_i\omega_j, & A=\{i,j\},\[1ex] 0, & \text{otherwise}, \end{cases}$5 continuity threshold. For a general Dyson interaction with $\Phi_A^D(\omega)= \begin{cases} -\dfrac{1}{|i-j|^\alpha}\,\omega_i\omega_j, & A=\{i,j\},\[1ex] 0, & \text{otherwise}, \end{cases}$6 and $\Phi_A^D(\omega)= \begin{cases} -\dfrac{1}{|i-j|^\alpha}\,\omega_i\omega_j, & A=\{i,j\},\[1ex] 0, & \text{otherwise}, \end{cases}$7, Gaussian concentration bounds hold throughout the full uniqueness region

$\Phi_A^D(\omega)= \begin{cases} -\dfrac{1}{|i-j|^\alpha}\,\omega_i\omega_j, & A=\{i,j\},\[1ex] 0, & \text{otherwise}, \end{cases}$8

provided

$\Phi_A^D(\omega)= \begin{cases} -\dfrac{1}{|i-j|^\alpha}\,\omega_i\omega_j, & A=\{i,j\},\[1ex] 0, & \text{otherwise}, \end{cases}$9

For the standard Dyson model $$H(\omega)=-\sum_{i>j}J(|i-j|)\omega_i\omega_j-\sum_i h_i\omega_i, \qquad J(n)=n^{-\alpha}.$$0, this condition holds for every $$H(\omega)=-\sum_{i>j}J(|i-j|)\omega_i\omega_j-\sum_i h_i\omega_i, \qquad J(n)=n^{-\alpha}.$$1. In the same regime, $$H(\omega)=-\sum_{i>j}J(|i-j|)\omega_i\omega_j-\sum_i h_i\omega_i, \qquad J(n)=n^{-\alpha}.$$2 has a nonnegative principal eigenfunction in $$H(\omega)=-\sum_{i>j}J(|i-j|)\omega_i\omega_j-\sum_i h_i\omega_i, \qquad J(n)=n^{-\alpha}.$$3 whenever $$H(\omega)=-\sum_{i>j}J(|i-j|)\omega_i\omega_j-\sum_i h_i\omega_i, \qquad J(n)=n^{-\alpha}.$$4, again covering all $$H(\omega)=-\sum_{i>j}J(|i-j|)\omega_i\omega_j-\sum_i h_i\omega_i, \qquad J(n)=n^{-\alpha}.$$5, जबकि positivity and continuity still require the stronger condition

$$H(\omega)=-\sum_{i>j}J(|i-j|)\omega_i\omega_j-\sum_i h_i\omega_i, \qquad J(n)=n^{-\alpha}.$$6

equivalent to $$H(\omega)=-\sum_{i>j}J(|i-j|)\omega_i\omega_j-\sum_i h_i\omega_i, \qquad J(n)=n^{-\alpha}.$$7 in the standard case (Makhmudov, 3 Aug 2025).

The ordered phase behaves differently. For sufficiently low temperature and $$H(\omega)=-\sum_{i>j}J(|i-j|)\omega_i\omega_j-\sum_i h_i\omega_i, \qquad J(n)=n^{-\alpha}.$$8, with

$$H(\omega)=-\sum_{i>j}J(|i-j|)\omega_i\omega_j-\sum_i h_i\omega_i, \qquad J(n)=n^{-\alpha}.$$9

the plus and minus Gibbs measures are not $J(k)\ge 0$0-measures: the one-sided conditional probability is essentially discontinuous at the alternating configuration. The mechanism is mesoscopic interface localization under Dobrushin boundary conditions and the resulting entropic repulsion or wetting effect. This establishes a sharp distinction between the two-sided Gibbs property and the one-sided $J(k)\ge 0$1-measure property in long-range one-dimensional ferromagnets (Bissacot et al., 2017).

4. Renormalization, weak fields, and boundary-induced phenomena

Ferromagnetic Dyson models also serve as a laboratory for non-Gibbsianness under renormalization. Under decimation,

$J(k)\ge 0$2

a low-temperature Dyson Gibbs measure becomes non-Gibbsian for $J(k)\ge 0$3 when

$J(k)\ge 0$4

The bad configuration is the alternating visible-spin configuration

$J(k)\ge 0$5

Conditioning on this configuration cancels the induced field on the hidden odd spins, leaving an effective hidden Dyson model on the odd sublattice, again at zero field but with a rescaled temperature. The transformed measure therefore develops an essential discontinuity through a hidden phase transition, in close analogy with decimated short-range Ising models in $J(k)\ge 0$6 (Bissacot et al., 2017).

A related question concerns inhomogeneous fields decaying as

$J(k)\ge 0$7

For $J(k)\ge 0$8, phase coexistence persists at low temperature if

$J(k)\ge 0$9

The energetic heuristic is explicit: the ferromagnetic droplet cost is $1<\alpha\le 2$0, while the field gain is $1<\alpha\le 2$1, so the field is asymptotically too weak to destroy phase coexistence when $1<\alpha\le 2$2. The authors further state that the physically expected threshold is $1<\alpha\le 2$3, and regard the appearance of $1<\alpha\le 2$4 as technical rather than fundamental (Bissacot et al., 2017).

Boundary effects are also central in the comparison between half-line and full-line models. For $1<\alpha\le 2$5, the one-sided and two-sided critical inverse temperatures coincide: $1<\alpha\le 2$6 This equality is nontrivial because removing the negative half-line eliminates a polynomially decaying family of couplings rather than a local boundary perturbation. The proof uses FK/random-cluster methods, sprinkling, and multiscale renormalization. The same work conjectures that equality should continue to hold at $1<\alpha\le 2$7, but that case is not proved (Berger et al., 19 Dec 2025).

These results correct another common misunderstanding: in long-range models, “boundary” and “bulk” cannot be separated by short-range intuition. Half-line versus full-line criticality, hidden phase transitions under decimation, and the persistence of ordering under weak fields all depend on genuinely nonlocal effects.

5. Hierarchical, dynamical, and quantum variants

A major analytically tractable proxy for the Dyson chain is the Dyson hierarchical ferromagnet. In the classical hierarchical Ising model with $1<\alpha\le 2$8 spins, the couplings are organized by generations,

$1<\alpha\le 2$9

so that the effective pair strength at scale $X=S^{\mathbb N}$0 behaves as

$X=S^{\mathbb N}$1

This reproduces the power-law scaling $X=S^{\mathbb N}$2. In this setting the macroscopic domain-wall cost is

$X=S^{\mathbb N}$3

which is the hierarchical analogue of the long-range ferromagnetic barrier exponent (Monthus et al., 2012).

For near-zero-temperature stochastic dynamics satisfying detailed balance, the equilibrium time obeys explicit asymptotics. For $X=S^{\mathbb N}$4,

$X=S^{\mathbb N}$5

while for $X=S^{\mathbb N}$6,

$X=S^{\mathbb N}$7

These leading barriers are the same for the simple and Glauber classes of dynamics, although finite corrections depend on the rate choice (Monthus et al., 2012).

At finite temperature, a real-space RG for the hierarchical ferromagnet closes explicitly on the dimensionless coupling $X=S^{\mathbb N}$8: $X=S^{\mathbb N}$9 The unstable fixed point is

$S=\{-1,+1\}$0

and the slowest relaxation time shows three regimes: finite in the paramagnet, critical power-law scaling

$S=\{-1,+1\}$1

and activated ferromagnetic scaling

$S=\{-1,+1\}$2

In this formalism the barrier exponent equals the domain-wall exponent,

$S=\{-1,+1\}$3

The same RG also yields $S=\{-1,+1\}$4 and $S=\{-1,+1\}$5 in the non-mean-field region $S=\{-1,+1\}$6 (Monthus, 2016).

The hierarchical construction also extends to the quantum transverse-field Ising case with couplings $S=\{-1,+1\}$7. For the pure model, the RG yields explicit recursions, a nontrivial unstable fixed point $S=\{-1,+1\}$8, and continuously varying exponents $S=\{-1,+1\}$9, $\beta |i-j|^{-\alpha}$00, $\beta |i-j|^{-\alpha}$01, together with

$\beta |i-j|^{-\alpha}$02

For the random case, numerical block RG in the hierarchical setting produces a conventional random critical point with finite $\beta |i-j|^{-\alpha}$03 and finite $\beta |i-j|^{-\alpha}$04, in contrast with SDRG expectations for the ordinary long-range chain. A plausible implication is that hierarchical geometry alters the balance between disorder and quantum fluctuations in a way that suppresses the infinite-disorder scenario (Monthus, 2015).

6. Dyson’s ferromagnet, non-Hermitian spin waves, and the alternate meaning of “Dyson model”

In the second major usage, a ferromagnetic Dyson model is not a long-range Ising chain but Dyson’s bosonic representation of a Heisenberg ferromagnet. For a single spin-$\beta |i-j|^{-\alpha}$05, Dyson maps the spin algebra to bosonic operators by

$\beta |i-j|^{-\alpha}$06

The bosons represent local deviations from the fully polarized ferromagnetic vacuum, hence magnons or spin waves. The decisive conceptual point is the existence of two inner products: the kinematic inner product, under which $\beta |i-j|^{-\alpha}$07 and $\beta |i-j|^{-\alpha}$08 are canonical adjoints, and the dynamical inner product induced from the spin Hilbert space, under which the Hamiltonian is self-adjoint. Accordingly, the Dyson Hamiltonian can satisfy

$\beta |i-j|^{-\alpha}$09

in the standard bosonic Fock metric, but

$\beta |i-j|^{-\alpha}$10

in the physical metric (Jones-Smith, 2013).

For the nearest-neighbor Heisenberg ferromagnet on the square lattice,

$\beta |i-j|^{-\alpha}$11

this mapping yields a bosonic Hamiltonian with a quadratic spin-wave part and quartic interaction terms. The quadratic term describes noninteracting magnons; the quartic terms are weak at low magnon density, which is the regime in which Dyson’s weakly interacting bosonic quasiparticles emerge naturally.

This short-range ferromagnetic usage remains historically important because later work revisited Dyson’s original analysis through effective field theory. In the isotropic nearest-neighbor “ideal ferromagnet,” the low-temperature spontaneous magnetization has the form

$\beta |i-j|^{-\alpha}$12

Dyson’s classic result is the first genuine interaction term, the $\beta |i-j|^{-\alpha}$13 contribution. The EFT analysis shows that the next interaction-induced correction appears already at order $\beta |i-j|^{-\alpha}$14, and that this coefficient is determined entirely by the leading effective Lagrangian, hence is independent of cubic-lattice anisotropies (Hofmann, 2011).

The two Dyson traditions are therefore complementary rather than interchangeable. The long-range Ising Dyson model concerns phase transitions, Gibbsianity, transfer operators, and nonlocal conditioning in one dimension. Dyson’s ferromagnetic bosonization concerns magnons, quasi-Hermiticity, and low-temperature spin-wave thermodynamics in short-range quantum magnets. The shared name reflects common historical origin, not common mathematical structure.