--- title: Dyson's Hierarchical Model (DHM) url: https://www.emergentmind.com/topics/dyson-s-hierarchical-model-dhm type: topic --- # Dyson's Hierarchical Model (DHM) Searching arXiv for relevant papers on Dyson’s Hierarchical Model and closely related recent work. Dyson’s Hierarchical Model (DHM) is a ferromagnetic Ising model on a recursively constructed hierarchical lattice in which pair couplings decay with hierarchical distance rather than remaining uniform as in Curie–Weiss theory. In the formulation emphasized in "From Dyson to Hopfield: Processing on hierarchical networks" [1407.5019], the model contains \(2^{k+1}\) spins \(\sigma_i=\pm1\) arranged into dyadic blocks, and its recursive construction induces a genuine metric, modular organization, and a non-mean-field free-energy landscape. Historically, the model was introduced by Dyson as a tractable lower-bound model for the one-dimensional long-range Ising problem, and more recent work has treated it both as a paradigmatic non-mean-field ferromagnet and as an approximation scheme for translationally invariant long-range systems [2510.02458]. ## 1. Definition, recursive geometry, and distance-dependent couplings In the standard recursive construction, one starts from two subsystems of size \(2^k\), denoted \(\vec{\sigma}_1\) and \(\vec{\sigma}_2\), and couples them weakly at level \(k+1\). A representative recursive Hamiltonian is \[ H_{k+1}(\vec{\sigma})= H_k(\vec{\sigma}_1)+H_k(\vec{\sigma}_2) -\frac{J}{2^{2\rho(k+1)}}\sum_{i0\) and \(\rho\in(1/2,1)\) in the non-mean-field regime [1407.5019]. Closely related parameterizations also appear in the literature, for example \[ H_N=-\sum_{p=1}^N\frac{J}{2^{p(1+\sigma)}}\sum_{r=1}^{2^{N-p}}(S_{p,r})^2, \] where \(S_{p,r}\) are hierarchical block spins and \(1/2<\sigma<1\) describes the nontrivial long-range critical regime of the one-dimensional long-range Ising chain approximation [2510.02458]. The defining geometric object is the hierarchical distance \(d_{ij}\): two spins \(i\) and \(j\) are at distance \(d\) if they first belong to the same block at the \(d\)-th recursive iteration. In this representation the Hamiltonian becomes \[ H_{k+1}(\vec{\sigma})=-\sum_{i0\), this gap vanishes as \(k\to\infty\), so the mixed state ceases to be merely metastable and becomes stable in the thermodynamic limit, sharing the same intensive free energy as the ferromagnetic state [1407.5019]. The same logic iterates down the hierarchy: the two largest branches can be split again, and the construction can be repeated up to \(\mathcal O(k)\) times, generating a hierarchy of internally ordered but mutually anti-aligned communities [1407.5019]. This phase structure is the sharpest contrast with mean-field ferromagnets. In Curie–Weiss theory all spins remain equally coupled and a single bulk order parameter is typically sufficient. In DHM, distance-dependent weakening of upper-level ties makes several block magnetizations thermodynamically relevant and produces a richer multiplicity of ordered states [1407.5019]. ## 3. Variational bounds, fluctuation control, and dynamical characterizations Several complementary analytical methods have been used to characterize DHM. One line of work derives interpolation-based lower bounds on the free energy. "Free-energy bounds for hierarchical spin models" [1312.2528] formulates both a mean-field-type lower bound and a stricter non-mean-field lower bound for DHM. In the notation of that paper, with decay parameter \(\sigma\), the mean-field lower bound is \[ \phi_{k+1}^{\mathrm{MF}}(m) = \log2+\log\cosh\!\left[\beta J\sum_{l=1}^{k+1}2^{l(1-2\sigma)}m+h\right] -\frac{\beta J}{2} \left[ \sum_{l=1}^{k+1}2^{l(1-2\sigma)}m^2+\sum_{l=1}^{k+1}2^{-2l\sigma} \right], \] whereas the improved non-mean-field bound is \[ \phi_{k+1}^{\mathrm{NMF}}(m) = \log 2+\log\cosh\!\left[ \beta J\left(\sum_{l=1}^{k+1}2^{l(1-2\sigma)}-\sum_{l=1}^{k+1}2^{-2l\sigma}\right)m+h \right] -\frac{\beta J}{2} \left(\sum_{l=1}^{k+1}2^{l(1-2\sigma)}-\sum_{l=1}^{k+1}2^{-2l\sigma}\right)m^2. \] The conceptual distinction is that the non-mean-field scheme absorbs part of the order-parameter fluctuations into effective sub-block Hamiltonians instead of discarding all fluctuations at once [1312.2528]. This fluctuation absorption relies on the decomposition of the whole-block variance into intra-block and inter-block pieces. The intra-block terms are reabsorbed, while only the residual inter-block correlation must be bounded. In DHM this can be done rigorously using Griffiths inequalities, because the model is ferromagnetic [1312.2528]. The resulting critical inverse temperatures reported there are \[ \beta_c^{\mathrm{MF}}=2^{2\sigma-1}-1,\qquad \beta_c^{\mathrm{NMF}}=2^{1-2\sigma}-3+2^{2\sigma}, \] with the non-mean-field estimate closer to the exact transition point than the mean-field one [1312.2528]. A second line of analysis uses a graph-theoretic normalization of the couplings. Writing \[ T_{ij}=\frac{J_{ij}}{w_i},\qquad w_i=\sum_j J_{ij}, \] one obtains a Markov transition matrix \(T\). The associated stochastic process satisfies either \[ p(t+1)=Tp(t) \qquad\text{or}\qquad \dot p(t)=Tp(t)-p(t). \] The Perron–Frobenius eigenvalue is \(\lambda_0=1\), but the second eigenvalue obeys \[ \lambda_1\to 1-\mathcal O\!\left(2^{-(2\rho-1)(k+1)}\right), \] so the spectral gap closes as \(k\to\infty\). The corresponding eigenvector has opposite signs on the two main branches, and its degeneracy with the uniform mode implies stationary states localized on left and right halves [1407.5019]. Thermodynamically, this mirrors the vanishing coupling between macroscopic communities; stochastically, it is an ergodicity-breaking mechanism. A third characterization is dynamical. Writing the local field as \[ H_{k+1}(\vec{\sigma})=\sum_i h_i(\vec{\sigma}\mid \rho)\,\sigma_i, \] and considering stochastic alignment \[ \sigma_i(t+\delta t)= \operatorname{sign}\left\{ \tanh[\beta h_i(\vec{\sigma}(t)\mid\rho)]+\eta_i(t) \right\}, \] with \(\eta_i(t)\) i.i.d. uniform in \([-1,1]\), one obtains in the zero-noise limit \[ \lim_{\beta\to\infty}\sigma_i(t+\delta t)=\operatorname{sign}\{h_i(\vec{\sigma}(t)\mid\rho)\}. \] The dynamical stability condition is \[ \sigma_i h_i(\vec{\sigma}\mid\rho)>0\qquad \forall i. \] Under this criterion, the fully aligned ferromagnetic state is stable for \(\rho\in(0.5,1]\), and the mixed left/right anti-aligned state is also stable in the thermodynamic limit because \[ \lim_{k\to\infty} h_i(\vec{\sigma}\mid\rho) = \frac{1}{2^{1-2\rho}+4^\rho-3}>0 \] throughout the same interval [1407.5019]. ## 4. Renormalization-group structure and relation to the one-dimensional long-range Ising model DHM has long served as a proxy for the one-dimensional long-range Ising chain because its hierarchical interaction reproduces the same scale dependence while remaining exactly amenable to real-space renormalization. In the recent comparison "One-dimensional long-range Ising model: two (almost) equivalent approximations" [2510.02458], the model is treated not as a historical curiosity but as one of two central approximations for the critical behavior of the one-dimensional long-range Ising model with couplings \(J_{ij}=J/|i-j|^{1+\sigma}\). The crucial RG feature is the exact recursion \[ H_N = H_{N-1}^{\rm left}+H_{N-1}^{\rm right} -\frac{J}{2^{N(1+\sigma)}}S_{N,1}^2, \] which, after a Hubbard–Stratonovich transformation, yields the exact partition-function identity \[ Z_N(\beta,h) = \frac{1}{\sqrt{\pi}} \int_{-\infty}^{\infty}d\phi\,e^{-\phi^2} \,Z_{N-1}(\beta,\hat h_N)^2, \qquad \hat h_N=h+2\sqrt{A_N}\,\phi\,\beta^{-1}, \] with \(A_N=2^{-N(1+\sigma)}\beta J\) [2510.02458]. The important structural point is that after one RG step the Hamiltonian keeps the same hierarchical form: no generic new nonlocal couplings are generated. The same paper reformulates DHM as a local-potential field theory. After a Hubbard–Stratonovich transformation and a decomposition \(A^{-1}_{ij}=K_{ij}+m^2\delta_{ij}\), one obtains \[ H_N^{\rm eff}[\phi] = \sum_{i\neq j}\phi_iK_{ij}\phi_j+\sum_iV(\phi_i). \] This is the hierarchical counterpart of the long-range functional RG local-potential approximation (LPA), \[ \Gamma_k[\phi] = \int dx\,\left\{ \phi(x)(-\Delta)^{\sigma/2}\phi(x)+V_k(\phi(x)) \right\}. \] In DHM, the quadratic term is a hierarchical Laplacian \(\lfloor-\Delta\rfloor^{\sigma/2}\) with stepwise “wedding cake” dispersion \(\omega(q)\sim c_\sigma q^\sigma\) at small \(q\), while the eigenmodes are hierarchical wavelets rather than plane waves [2510.02458]. This structural comparison leads to a precise numerical statement: the DHM and FRG-LPA estimates of the thermal exponent \(\nu\) agree to extremely high precision throughout \(1/2<\sigma<1\), with relative difference in \(y=\nu^{-1}\) below \(10^{-3}\), the largest deviation occurring near \(\sigma\to1\) [2510.02458]. The interpretation proposed there is that these are “two (almost) equivalent approximations”: formally very close and numerically nearly indistinguishable over most of the nontrivial long-range regime, but not identical because DHM lacks translational invariance, has a discrete block-scaling factor \(\ell=2\), and replaces \(q^\sigma\) by a stepwise hierarchical dispersion [2510.02458]. This same comparison also clarifies a limitation. Near \(\sigma=1\), where the translationally invariant \(1/r^2\) chain exhibits BKT-like behavior with Thouless effect, both DHM and FRG-LPA become unsatisfactory. The standard DHM does not reproduce the exact \(\sigma=1\) behavior, although a modified hierarchical decay \(2^{-2p}\log p\) can recover BKT-like behavior in a modified model [2510.02458]. ## 5. Extensions: associative memory, disorder, susceptibility, and spectral theory One important extension couples the hierarchical kernel to the Hebb rule, producing the hierarchical Hopfield model (HHM). In [1407.5019] the couplings become \[ \widetilde J_{ij} = \frac{4^{\rho-d_{ij}\rho}-4^{-k\rho}}{4^\rho-1} \sum_{\mu=1}^{p}\xi_i^\mu\xi_j^\mu, \] with i.i.d. patterns \(\xi_i^\mu=\pm1\). The relevant order parameters are Mattis overlaps, globally \[ m^\mu=\frac{1}{2^{k+1}}\sum_{i=1}^{2^{k+1}}\xi_i^\mu\sigma_i, \] and separately on the two main branches. The same branch decoupling that yields mixed ferromagnetic states now yields both serial retrieval, in which the full system retrieves one pattern, and parallel retrieval, in which different large communities retrieve different patterns simultaneously [1407.5019]. The price is reduced capacity: the paper concludes that at best \[ p\le \mathcal O(k), \] so storage is much lower than in the standard mean-field Hopfield model [1407.5019]. A second extension introduces quenched random fields on the Dyson hierarchical lattice. "Existence of long-range order in random-field Ising model on Dyson hierarchical lattice" [2410.11515] studies \[ H_N^{\mathrm{RF}}(\vec\sigma)=H_N(\vec\sigma)-h\sum_{i=1}^{2^N}h_i\sigma_i, \] with \(h_i\) i.i.d. Gaussian or binary random fields and pure interaction \(J(r)\sim r^{-\alpha}\). The paper proves that for \(1<\alpha<3/2\) and sufficiently small but nonzero random field, there is long-range order at sufficiently low temperature, including zero temperature, in the sense that \[ m^2:=\liminf_{N\to\infty}\frac{1}{2^{2N}}\, \mathbb E\big[\langle S_{N,1}^2\rangle_N\big] \] is positive [2410.11515]. The proof is an adaptation of Dyson’s recursive method combined with concentration inequalities; the scaling competition \(b_N\sim2^{(2-\alpha)N}\) versus disorder fluctuations \(2^{N/2}\) yields the threshold \(\alpha<3/2\) [2410.11515]. A third development concerns the borderline \(|x-y|^{-2}\)-type interaction. "Double-exponential susceptibility growth in Dyson's hierarchical model with \(|x-y|^{-2}\) interaction" [2302.01509] proves, in the hierarchical long-range percolation model, that when \(\alpha=d\) the susceptibility grows as \[ \chi(\beta)=e^{e^{\Theta(\beta)}} \qquad (\beta\to\infty), \] whereas for \(\alpha>d\), \[ \chi(\beta)=\beta^{\frac d{\alpha-d}-o(1)}. \] By stochastic domination and Edwards–Sokal, the paper states that analogous susceptibility bounds hold for hierarchical random-cluster, Potts, and Ising models, including Dyson’s hierarchical Ising model with interaction \(J(x,y)=\|x-y\|^{-d-\alpha}\) [2302.01509]. The distinction is important: this is a rigorous consequence for DHM-type Ising susceptibility obtained through comparison technology rather than through a direct DHM spin-system analysis. A fourth line of work studies the operator theory of the underlying hierarchical lattice. "On the negative spectrum of the hierarchical Schrödinger operator" [1206.4019] defines the Dyson hierarchical Laplacian on a countable hierarchical lattice with branching parameter \(\nu\) and scale parameter \(p\), obtains an explicit spectral decomposition with eigenvalues \[ \lambda_k=-p^{k-1},\qquad k=1,2,\dots, \] and identifies the spectral dimension \[ s_h=\frac{2\ln\nu}{\ln(1/p)}. \] The corresponding Markov process is transient for \(s_h>2\) and recurrent for \(s_h\le2\), and the heat kernel satisfies \[ p(t,x,x)\asymp t^{-s_h/2} \] up to a log-periodic factor [1206.4019]. Although this is an operator-theoretic rather than ferromagnetic analysis, it formalizes the multiscale geometry that underlies DHM and shows how hierarchical models realize continuously tunable effective dimension [1206.4019]. ## 6. Conceptual status, misconceptions, and limitations DHM is sometimes described loosely as a long-range mean-field model. That description is misleading. It is fully connected, but it is not mean-field in the Curie–Weiss sense because the couplings are neither uniform nor permutation-invariant; they are organized by hierarchical distance, and the non-vanishing relevance of block magnetizations is precisely what gives the model its non-mean-field character [1407.5019]. An opposite misconception is that DHM is simply the one-dimensional long-range Ising chain in disguise. The recent FRG comparison rejects that identification as well. DHM is structurally close to the translationally invariant model and, over much of \(1/2<\sigma<1\), gives nearly identical estimates for \(\nu\), but it remains an ultrametric approximation with non-plane-wave eigenmodes, a stepwise hierarchical dispersion, and discrete block scaling [2510.02458]. A plausible implication is that DHM is best viewed as an analytically controlled approximation scheme rather than as an exact surrogate for Euclidean long-range criticality. The literature also distinguishes carefully between directly proved DHM statements and transferred consequences. The mixed-state thermodynamics, vanishing energy gap, and graph-theoretic ergodicity breaking are direct DHM results [1407.5019]. The improved lower free-energy bounds are direct DHM theorems that exploit ferromagnetic correlation inequalities [1312.2528]. By contrast, the double-exponential susceptibility growth at the borderline \(|x-y|^{-2}\)-type decay is inferred for hierarchical Ising models through random-cluster comparison and Edwards–Sokal rather than through a direct Ising RG proof [2302.01509]. Taken together, these results establish DHM as a central hierarchical framework for non-mean-field statistical mechanics. Its main structural lesson is that a recursive ultrametric geometry can generate distance, modularity, multiple ordered states, and exact real-space RG within a fully connected spin system. Its main methodological lesson is that this geometry allows techniques—interpolation bounds, block-spin recursions, graph spectra, hierarchical Laplacians, and local-potential flows—that are often unavailable or only approximate in translationally invariant models [1407.5019]. Source: https://www.emergentmind.com/topics/dyson-s-hierarchical-model-dhm