In 1967 B. D. Josephson derived the hyperscaling (Josephson) inequalities dν≥2−αd\nu \ge 2-\alpha and dν′≥2−α′d\nu' \ge 2-\alpha' for the critical exponents of the specific heat (c∼∣T−Tc∣−αc \sim |T-T_c|^{-\alpha}) and correlation length (ξ∼∣T−Tc∣−ν\xi \sim |T-T_c|^{-\nu}), by combining a rigorous equilibrium fluctuation inequality for the specific heat with certain plausible assumptions about the correlations in the neighbourhood of the critical point: integrable decay of the maximal normalized correlation function away from criticality, a relative-temperature-derivative bound on cell order-parameter fluctuations, a uniform fourth-moment bound, and a classical equilibrium setting. Under these hypotheses, systems with Goldstone modes (the massless long-wavelength fluctuations that accompany continuous symmetry breaking) or genuinely long-range forces lie beyond reach, and the constants in the bound were left undetermined. In the subsequent literature, Sokal (1981) rigorously proved both the high- and low-temperature Josephson inequalities for the ferromagnetic Ising/ϕ4\phi^4 lattice class following Josephson's own strategy, and the sharpness results of Aizenman-Barsky-Fernández (1987), Duminil-Copin-Raoufi-Tassion (2019) and Duminil-Copin-Goswami-Raoufi (2019) established the correlation-decay hypothesis for Ising-type models at all T≠TcT \ne T_c. The surviving open problem is the extension beyond this class: continuous-symmetry systems below TcT_c (the Goldstone obstruction, with the restricted-operator remedy proposed in the literature unproven), quantum systems, systems without ferromagnetic correlation inequalities (continuum fluids, non-additive long-range interactions), and ensembles with global constraints whose induced correlations may spoil the decay hypothesis.

Background

Near a continuous phase transition at temperature TcT_c, equilibrium quantities are singular. The specific heat per unit volume behaves as c∼∣T−Tc∣−αc \sim |T-T_c|^{-\alpha}, where the exponent α\alpha may be zero (logarithmic divergence) or negative, and the correlation length behaves as ξ∼∣T−Tc∣−ν\xi \sim |T-T_c|^{-\nu}, where ξ\xi measures the range over which fluctuations of the order parameter (e.g., the magnetization) are correlated. Primed exponents (ν′\nu', α′\alpha') refer to the approach from the ordered phase, T→Tc−T \to T_c^-. A small number of exponent combinations are constrained by rigorous inequalities: thermodynamic convexity gives Rushbrooke's α+2β≥2\alpha+2\beta \ge 2 (β\beta governs the spontaneous magnetization), Fisher (1969) proved γ≤(2−η)ν\gamma \le (2-\eta)\nu (γ\gamma governs the susceptibility, η\eta the anomalous decay of correlations at TcT_c), and Buckingham and Gunton (1969) proved related rigorous bounds coupling the dimension dd to the exponents γ′\gamma', β\beta, δ\delta and α′\alpha'.

The heuristic hyperscaling relation 2−α=dν2-\alpha = d\nu (dd = spatial dimension) follows from the singularity hypothesis: if the singular part of the free-energy density in a correlated volume ξd\xi^d is of order kBT/ξdk_B T/\xi^d, then differentiating twice suggests c∼ξ−d∣T−Tc∣−2c \sim \xi^{-d}|T-T_c|^{-2} and hence 2−α≤dν2-\alpha \le d\nu as a one-sided bound. Hyperscaling as an equality is expected below the upper critical dimension ducd_{\mathrm{uc}} (the dimension above which mean-field theory, with ν=1/2\nu=1/2, α=0\alpha=0, becomes exact), and fails above it, where dangerous irrelevant variables (couplings that are formally irrelevant at the renormalization-group fixed point yet still modify the scaling of the singular part of the free energy) cap the fluctuation volume.

Josephson (1967) turned the one-sided bound into a derivation modulo explicit correlation hypotheses. Part I (Proc. Phys. Soc. 92, 269) established, for a classical system in equilibrium, the fluctuation inequality Vc≥kBT2∣∂T⟨A⟩∣2/⟨A~2⟩Vc \ge k_B T^2|\partial_T\langle A\rangle|^2/\langle\tilde A^2\rangle for any observable AA not depending explicitly on temperature (VV = volume, tilde = centered, ⟨⋅⟩\langle\cdot\rangle = equilibrium average), a consequence of the Cauchy-Schwarz inequality applied to the covariance of AA with the Hamiltonian. Part II (Proc. Phys. Soc. 92, 276) took AA to be a sum over cubical cells of side LL of squared local order-parameter fluctuations and imposed three hypotheses: (i) away from TcT_c the maximal normalized correlation function k(R)k(R) — the supremum of correlation coefficients between local observables in regions separated by at least RR — decays fast enough that ∫k(R)dR\int k(R)d\mathbf{R} is finite and bounded by a1ξda_1\xi^d, giving cell-sum bounds ∑jk(dij)≤a2+a3(ξ/L)d\sum_j k(d_{ij}) \le a_2+a_3(\xi/L)^d (the sum running over cells jj at distance dijd_{ij} from cell ii); (ii) the cell mean-square fluctuation ϕ(L,T)=⟨Q~i2⟩\phi(L,T)=\langle\tilde Q_i^2\rangle obeys the relative-derivative bound ∣∂Tϕ∣/ϕ>a4/∣T−Tc∣|\partial_T\phi|/\phi > a_4/|T-T_c| whenever L≥a5ξL \ge a_5\xi; (iii) a uniform fourth-moment bound ⟨Q~i4⟩≤a6⟨Q~i2⟩2\langle\tilde Q_i^4\rangle \le a_6\langle\tilde Q_i^2\rangle^2 excluding heavy-tailed cell fluctuations. Inserting these into the Part-I inequality makes amplitudes cancel, yielding c(T)≥a7kBξ(T)−d(T/(T−Tc))2c(T) \ge a_7 k_B \xi(T)^{-d}(T/(T-T_c))^2 with a7a_7 an explicit function of a2,…,a6a_2,\ldots,a_6, and comparison with the power-law definitions of α\alpha and ν\nu gives dν≥2−αd\nu \ge 2-\alpha and dν′≥2−α′d\nu' \ge 2-\alpha' (Josephson's paper denotes the spatial dimension by ss, so his statements read sν≥2−αs\nu \ge 2-\alpha; the symbol dd denotes the spatial dimension here).

The gaps in the derivation are as follows. In systems with a continuously broken symmetry below TcT_c (e.g., a Heisenberg ferromagnet, where rotation symmetry O(3)O(3) breaks to O(2)O(2)), massless Goldstone modes — long-wavelength spin-wave-like fluctuations that necessarily accompany the broken symmetry — render the unrestricted ∫k(R)dR\int k(R)d\mathbf{R} divergent; the remedy proposed in the literature is to restrict the fluctuation variables to operators commuting with the broken-symmetry generators (e.g., the modulus of the local magnetization), a proposal whose required bounds were never proved. Systems with genuinely long-range, non-additive interactions — interactions so long-ranged that the total energy is not proportional to the system volume — violate the decay hypothesis outright. The derivation is classical and does not cover quantum systems, where the specific heat is not governed by the classical variance of the Hamiltonian, nor ensembles with global constraints, which can induce anomalous correlations of order the inverse volume 1/V1/V that spoil the decay hypothesis; the Part-I fluctuation inequality itself is proved for a class of constraints that includes constancy of the volume, the pressure and, where applicable, the magnetization and the applied magnetic field, so this concern attaches specifically to the correlation-decay hypothesis of Part II. The constants a2,…,a7a_2,\ldots,a_7 were left undetermined, so the sharpness of the bound is unknown. The hyperscaling equality itself remains unproven in the physically central dimensions (Pelissetto-Vicari 2002), while above ducd_{\mathrm{uc}} mean-field exponents (ν=1/2\nu=1/2, α=0\alpha=0) satisfy the Josephson inequality strictly for d>4d>4 and saturate it at d=4d=4, consistent with Josephson's bound but not with hyperscaling as an equality (Berche et al. 2022).

Problem Statement

The problem, as left open by Josephson (1967) and still unsettled for classes beyond the ferromagnetic Ising/ϕ4\phi^4 family, is to establish rigorously — or disprove — the specific-heat-correlation-length bound

c(T)≥a kB ξ(T)−d(TT−Tc)2,c(T) \ge a\, k_B\, \xi(T)^{-d}\left(\frac{T}{T-T_c}\right)^{2},

valid for T>TcT > T_c, together with its analogue below TcT_c (as T→Tc−T \to T_c^- in the ordered phase), and hence the exponent inequalities dν≥2−αd\nu \ge 2-\alpha and dν′≥2−α′d\nu' \ge 2-\alpha', for precisely specified classes of critical systems. Here c(T)c(T) is the specific heat per unit volume at temperature TT, ξ(T)\xi(T) the correlation length (the range of order-parameter correlations), dd the spatial dimension, TcT_c the critical temperature, kBk_B Boltzmann's constant, aa a positive constant, and the exponents are fixed by the power-law asymptotics ξ∼∣T−Tc∣−ν\xi \sim |T-T_c|^{-\nu} and c∼∣T−Tc∣−αc \sim |T-T_c|^{-\alpha} as T→TcT \to T_c from above (primed exponents: the same limits from below); a rigorous treatment must also fix the exponent convention (e.g., liminf or limsup of logarithmic derivatives), on which the precise content of the inequalities depends. The three correlation conditions of the 1967 derivation, whose validity or necessity for a given class of systems is at issue, are: (i) integrable decay of the maximal normalized correlation function k(R)k(R) — the supremum, over pairs of local observables in regions separated by at least RR, of their correlation coefficient — away from criticality, with ∫k(R) dR≤a1ξd\int k(R)\,d\mathbf{R} \le a_1\xi^d; (ii) the relative-derivative bound ∣∂Tϕ∣/ϕ>a4/∣T−Tc∣|\partial_T\phi|/\phi > a_4/|T-T_c| on the cell mean-square order-parameter fluctuation ϕ(L,T)=⟨Q~i2⟩\phi(L,T)=\langle\tilde Q_i^2\rangle (tilde = centered variable; cells of side L≥a5ξL \ge a_5\xi); (iii) the uniform fourth-moment bound ⟨Q~i4⟩≤a6⟨Q~i2⟩2\langle\tilde Q_i^4\rangle \le a_6\langle\tilde Q_i^2\rangle^2.

The prototypical instance — the nearest-neighbor ferromagnetic Ising model, above and below TcT_c — was settled affirmatively by Sokal (1981). The directions identified in later literature in which the problem remains open are: (a) systems with a continuously broken symmetry below TcT_c — e.g., the classical three-dimensional Heisenberg magnet, whose rotation symmetry O(3)O(3) breaks to O(2)O(2) in the ordered phase — where massless Goldstone modes (the long-wavelength fluctuations that necessarily accompany continuous symmetry breaking) make ∫k(R) dR\int k(R)\,d\mathbf{R} diverge, and where it is open whether the inequality holds in this regime or fails; (b) quantum systems, where the classical fluctuation inequality would need a quantum counterpart, and where the resulting bound would have to be related to the specific-heat exponent at a quantum critical point (a continuous transition reached at zero temperature by tuning a non-thermal parameter), at which hyperscaling is expected with effective dimension d+zd+z (zz the dynamical critical exponent); and (c) systems and ensembles outside the reach of the 1967 hypotheses — continuum fluids and other systems for which no ferromagnetic correlation inequalities are available, systems with genuinely long-range non-additive interactions (interactions so long-ranged that the total energy is not proportional to the volume), and equilibrium ensembles with global constraints, which can induce anomalous correlations of order 1/V1/V — where the inequality or its hypotheses may genuinely fail. The constants of the derivation were left undetermined, so the quantitative sharpness of the bound remains unknown for every model; sharpness is noted here as a property of the derivation, not as a separate mandatory deliverable.

These directions are candidate domains of one and the same open problem, not a jointly required checklist: the original derivation question is answered for a precisely specified class of critical systems beyond the settled ferromagnetic Ising/ϕ4\phi^4 family once the bound, and hence the exponent inequalities, are rigorously established — or refuted — for that class. The shared objective is a mathematically controlled determination of the domain of validity of Josephson's inequality: which critical systems obey a specific-heat-correlation-length bound, which hypotheses are responsible, and how sharp the bound is.

References

  1. G. S. Rushbrooke, 'On the thermodynamics of the critical region for the Ising problem', Journal of Chemical Physics 39, 842-843 (1963), DOI 10.1063/1.1734338, URL https://doi.org/10.1063/1.1734338
  2. B. D. Josephson, 'Inequality for the specific heat: I. Derivation', Proceedings of the Physical Society 92, 269-275 (1967), DOI 10.1088/0370-1328/92/2/301, URL https://doi.org/10.1088/0370-1328/92/2/301
  3. B. D. Josephson, 'Inequality for the specific heat: II. Application to critical phenomena', Proceedings of the Physical Society 92, 276-284 (1967), DOI 10.1088/0370-1328/92/2/302, URL https://doi.org/10.1088/0370-1328/92/2/302
  4. A. D. Sokal, 'Rigorous proof of the high-temperature Josephson inequality for critical exponents', Journal of Statistical Physics 25, 51-56 (1981), DOI 10.1007/BF01008478, URL https://doi.org/10.1007/BF01008478
  5. A. D. Sokal, 'More inequalities for critical exponents', Journal of Statistical Physics 25, 25-50 (1981), DOI 10.1007/BF01008477, URL https://doi.org/10.1007/BF01008477
  6. A. D. Sokal, 'A rigorous inequality for the specific heat of an Ising or ϕ4\phi^4 ferromagnet', Physics Letters A 71, 451-453 (1979), DOI 10.1016/0375-9601(79)90633-9, URL https://doi.org/10.1016/0375-9601(79)90633-9
  7. R. Schrader, 'New rigorous inequality for critical exponents in the Ising model', Physical Review B 14, 172-173 (1976), DOI 10.1103/PhysRevB.14.172, URL https://doi.org/10.1103/PhysRevB.14.172
  8. M. E. Fisher, 'Rigorous inequalities for critical-point correlation exponents', Physical Review 180, 594-600 (1969), DOI 10.1103/PhysRev.180.594, URL https://doi.org/10.1103/PhysRev.180.594
  9. M. J. Buckingham and J. D. Gunton, 'Correlations at the critical point of the Ising model', Physical Review 178, 848-853 (1969), DOI 10.1103/PhysRev.178.848, URL https://doi.org/10.1103/PhysRev.178.848
  10. M. Aizenman, D. J. Barsky and R. Fernández, 'The phase transition in a general class of Ising-type models is sharp', Journal of Statistical Physics 47, 343-374 (1987), DOI 10.1007/BF01007515, URL https://doi.org/10.1007/BF01007515
  11. H. Duminil-Copin, A. Raoufi and V. Tassion, 'Sharp phase transition for the random-cluster and Potts models via decision trees', Annals of Mathematics 189, 75-99 (2019), DOI 10.4007/annals.2019.189.1.2, URL https://doi.org/10.4007/annals.2019.189.1.2
  12. H. Duminil-Copin, S. Goswami and A. Raoufi, 'Exponential decay of truncated correlations for the Ising model in any dimension for all but the critical temperature', Communications in Mathematical Physics 374, 891-921 (2019), DOI 10.1007/s00220-019-03633-y, URL https://doi.org/10.1007/s00220-019-03633-y
  13. H. Tasaki, 'Hyperscaling inequalities for percolation', Communications in Mathematical Physics 113, 49-65 (1987), DOI 10.1007/BF01221396, URL https://doi.org/10.1007/BF01221396
  14. R. Fernández, J. Fröhlich and A. D. Sokal, 'Inequalities for critical exponents', chapter in Random Walks, Critical Phenomena, and Triviality in Quantum Field Theory, Springer, Berlin, pp. 297-366 (1992), DOI 10.1007/978-3-662-02866-7_14, URL https://doi.org/10.1007/978-3-662-02866-7_14
  15. A. Pelissetto and E. Vicari, 'Critical phenomena and renormalization-group theory', Physics Reports 368, 549-727 (2002), DOI 10.1016/S0370-1573(02)00219-3, URL https://doi.org/10.1016/S0370-1573(02)00219-3, arXiv cond-mat/0012164, URL https://arxiv.org/abs/cond-mat/0012164
  16. B. Berche, T. Ellis, Yu. Holovatch and R. Kenna, 'Phase transitions above the upper critical dimension', SciPost Physics Lecture Notes 60 (2022), DOI 10.21468/SciPostPhysLectNotes.60, URL https://doi.org/10.21468/SciPostPhysLectNotes.60, arXiv 2204.04761, URL https://arxiv.org/abs/2204.04761
  17. D. Stauffer, M. Ferer and M. Wortis, 'Universality of second-order phase transitions: The scale factor for the correlation length', Physical Review Letters 29, 345-349 (1972), DOI 10.1103/PhysRevLett.29.345, URL https://doi.org/10.1103/PhysRevLett.29.345
  18. A. Eberlein, I. Mandal and S. Sachdev, 'Hyperscaling violation at the Ising-nematic quantum critical point in two-dimensional metals', Physical Review B 94, 045133 (2016), DOI 10.1103/PhysRevB.94.045133, URL https://doi.org/10.1103/PhysRevB.94.045133
  19. J. Fröhlich, B. Simon and T. Spencer, 'Infrared bounds, phase transitions and continuous symmetry breaking', Communications in Mathematical Physics 50, 79-95 (1976), DOI 10.1007/BF01608557, URL https://doi.org/10.1007/BF01608557
  20. M. Aizenman and R. Fernández, 'On the critical behavior of the magnetization in high-dimensional Ising models', Journal of Statistical Physics 44, 393-454 (1986), DOI 10.1007/BF01011304, URL https://doi.org/10.1007/BF01011304