Fate of the Parisi-Slanina hard-sphere toy model above the critical density $\rho_{1c}=1-\ln 2$ in the infinite-dimensional limit

The Parisi-Slanina toy model of the mean-field theory of hard-sphere liquids (Phys. Rev. E 62, 6554 (2000)) replaces the hypernetted-chain kernel $L_3(x)=-\ln(1+x)+x-x^2/2$ by a hard constraint $x\ge -1$, reducing the equilibrium of a hard-sphere liquid in spatial dimension $D=2N+3\to\infty$ to the constraint system $\rho\hat h(p)\ge -1$, $h(r)\ge -1$, $h(r)=-1$ for $r<1$, plus continuity of $h$ in the density. The model was solved analytically only for rescaled densities $0<\rho_1=\ln(\rho V_D)/N<\rho_{1c}=1-\ln 2\approx 0.3068$, where a Fourier cutoff momentum $\hat p_c<1$ exists; the authors left the regime $\rho_1>\rho_{1c}$ (where gaps in $g(r)$ are expected, continuity may fail, and the absolute density bound is $\rho_1<\ln 4$) explicitly open, together with the crossover window $\hat p_c=1-O(1/N)$ in which the contact amplitude $A$ must be solved jointly with $\hat p_c$. Follow-up literature identified $\rho_{1c}$ with the Kirkwood instability of the exact ring-resummed infinite-dimensional liquid, reported that no solutions could be found above it, declared the question mathematically interesting but physically irrelevant (the glass transition preempts it), and completed the associated glass program by the replica route. The open problem is to determine the solution set of the constraint system (and of the exact infinite-dimensional liquid) for $\rho_1>\rho_{1c}$: existence, structure (gaps, discontinuities, delta-function components), equation of state, the crossover, and the maximal density admitting a solution.

Background

Hard spheres - particles whose interaction energy is infinite on overlap and zero otherwise - are the reference model of classical liquid-state theory. The structure of a uniform liquid is encoded in the radial distribution function $g(r)$, the relative probability of finding a pair of particles at separation $r$, and in the total correlation function $h(r)=g(r)-1$. The hypernetted chain (HNC) approximation expresses the free energy as a variational functional $F[h]$ of $h$; in the version relevant here the momentum-space part of the functional involves the kernel $L_3(x)=-\ln(1+x)+x-x^2/2$ acting on $\rho\hat h(p)$, where $\hat h$ is the radial Fourier transform of $h$ and $\rho$ the number density. Stationarity of $F$ yields the HNC integral equations, which must be solved numerically; the competing Percus–Yevick (PY) approximation is exactly solvable in three dimensions but has no convenient variational form. The variational form is what the replica approach to the static glass transition needs: following the analogy between structural glasses and mean-field spin glasses, M. Mézard and G. Parisi's replica-HNC program represents the glass as an effective liquid of $m$-component bound states ($m$ being the replica index to be sent to 1 at the end), whose study requires a tractable two-particle correlation function as input.

What was missing was an analytically solvable 'mean-field' liquid. Parisi and Slanina proposed to take 'mean field' as spatial dimension $D\to\infty$ and to simplify the HNC functional by replacing $L_3(x)$ with $L_\infty(x)$, defined as $\infty$ for $x<-1$ and $0$ otherwise. The variational problem then reduces to the minimal physical requirements

$\rho\hat h(p)\ge -1,\qquad h(r)\ge -1,\qquad h(r)=-1\ \text{for}\ r<1,$

together with continuity of $h$ in the density (failure of such continuity would signal a phase transition, which is not found). Each constraint has a direct physical meaning: $h(r)\ge -1$ states that the pair probability $g(r)$ is nonnegative; $h(r)=-1$ for $r<1$ encodes the hard core; and $\rho\hat h(p)\ge -1$ is the nonnegativity of the structure factor $S(p)=1+\rho\hat h(p)$, the spectral weight of density fluctuations at wavevector $p$ that scattering experiments measure. In three dimensions the model is solved numerically by iterating $h(r)$ while slowly increasing $\rho$; the results qualitatively resemble HNC and PY up to $\rho\simeq 1$, while at higher densities strict zeros ('gaps') of $g(r)$ open between successive peaks - an acknowledged artifact of the $L_\infty$ replacement.

In $D=2N+3$ dimensions the natural variables are the rescaled density $\bar\rho=\rho V_D$ ($V_D$ being the volume of the unit-radius $D$-ball), the rescaled pressure $\bar P=V_D P/(kT)$, and $\rho_1=\ln\bar\rho/N$; the pressure follows from the contact value of $h$. The large-$D$ Fourier transform of the unit-ball indicator decays as $\Psi_0(\hat p)=\exp[-N\phi_0(\hat p)]$ with $\phi_0(\hat p)=1-\ln 2+\ln(1+\sqrt{1-\hat p^2})-\sqrt{1-\hat p^2}$ for rescaled momentum $\hat p=p/N<1$, and becomes oscillatory for $\hat p>1$. The analytic solution is: for $\bar\rho<1$, $\bar P=\bar\rho+\bar\rho^2/2$ (first virial correction); for $\bar\rho>1$ and $0<\rho_1<\rho_{1c}=\phi_0(1)=1-\ln 2\approx 0.3068$, the constraint $\rho\hat h\ge -1$ is saturated for $\hat p<\hat p_c$ where $\hat p_c$ solves $\rho_1=\ln(1+\sqrt{1-\hat p_c^2})-\sqrt{1-\hat p_c^2}+1-\ln 2$, the pressure correction scales as $P_c=e^{N P_1}$ with $P_1=2-2\ln 2+\ln\hat p_c^2$, and the amplitude $A$ of the contact discontinuity of $h$ is exponentially small in $N$ (so it can be neglected). An absolute close-packing bound is $\rho_1<\ln 4$.

Two facts delimit what is known at the boundary $\rho_{1c}$. First, inside the model: as $\rho_1\to\rho_{1c}^-$ the exponent controlling $h$ becomes $O(1)$ and a gap in $g(r)$ can appear; the authors state that the detailed investigation of this process and the behavior for $\rho_1>\rho_{1c}$ 'remains an open question', and that within $O(1/N)$ of $\hat p_c=1$ the coupled equations for $A$ and $\hat p_c$ must be solved together - an interval they deliberately omit. Second, outside the model: Frisch and Percus (Phys. Rev. E 60, 2942 (1999)) resummed the ring diagrams (the chain-like terms that dominate the virial series, i.e. the expansion of the free energy in powers of the density, at large $D$) of the true (full-$L_3$) hard-sphere fluid for $D\to\infty$ and obtained the same liquid entropy up to the rescaled density $\rho_1=1-\ln 2$, at which the resummed propagator develops a pole - the Kirkwood instability, i.e. a divergence of the structure factor at nonzero wavevector that terminates the liquid solution of integral-equation theories. Thus $\rho_{1c}$ is not merely an artifact of the toy model; it is where the exact infinite-dimensional liquid itself loses its analytic liquid branch, and the follow-up literature could not decide whether this reflects nonexistence of solutions or the need for more complex (possibly non-continuous, possibly measure-valued) solutions.

Problem Statement

Consider the Parisi-Slanina toy model of a hard-sphere liquid in spatial dimension $D=2N+3$ with $N\to\infty$: hard spheres of diameter 1 at density $\rho$, described by a correlation function $h(r)$ (with radial Fourier transform $\hat h$) that must satisfy the constraint system

$\rho\hat h(p)\ge -1,\qquad h(r)\ge -1,\qquad h(r)=-1\ \text{for}\ r<1,$

together with continuity of $h$ in the density (this is the $L_\infty$ simplification of the HNC free energy, whose exact kernel $L_3(x)=-\ln(1+x)+x-x^2/2$ enforces the same positivity constraint softly). In the rescaled variables $\bar\rho=\rho V_D$, $\bar P=V_D P/(kT)$ and $\rho_1=\ln\bar\rho/N$, the known analytic solution covers $\bar\rho<1$ (where $\bar P=\bar\rho+\bar\rho^2/2$) and $0<\rho_1<\rho_{1c}=1-\ln 2\approx 0.3068$, where the solution is parametrized by a Fourier cutoff $\hat p_c<1$ solving $\rho_1=\ln(1+\sqrt{1-\hat p_c^2})-\sqrt{1-\hat p_c^2}+1-\ln 2$, with pressure correction $P_c=\exp(NP_1)$, $P_1=2-2\ln 2+\ln\hat p_c^2$, and an exponentially small contact amplitude $A=h(1^+)$, the value of $h$ just outside contact (since $h=-1$ inside the core, $A+1$ is the jump of $h$ across $r=1$; in the source's notation $A=\bar h(1^+)$, with $\bar h(r)$ the part of $h$ that vanishes for $r<1$). The absolute close-packing bound is $\rho_1<\ln 4$.

The problem, left open by the source paper and confirmed unresolved by the follow-up literature, is to determine the fate of this model at and above the critical density $\rho_{1c}$:

1. Existence and structure for $\rho_1>\rho_{1c}$. Does the constraint system admit a solution there? If yes, construct it - the correlation function $h(r)$, its Fourier-space support, the equation of state $\bar P(\bar\rho)$, and the qualitative changes expected as $\rho_1\to\rho_{1c}^+$: gaps (near-zeros) in $g(r)$, possible failure of continuity of $h$ at contact or in the density, and different correlation patterns. If no, prove nonexistence within the class of continuous-in-density solutions and determine the supremum of densities admitting such a solution; a separate question is the extension to wider classes of admissible $h$ that allow delta-function components (which correspond to lattice-like order and connect the question to the densest-packing problem in high dimensions).

2. The crossover window. Solve jointly the two coupled equations for the contact amplitude $A$ and the cutoff $\hat p_c$, $\phi_0(\hat p_c)=\rho_1-\frac{\ln(A+1)}{N},\qquad A=\left(\frac{N}{2\pi}\right)^D\frac{N}{D+1}\,V_D^2\,\hat p_c^{D+1}\,\phi_0'(\hat p_c)\,(A+1)\,\Psi_0^2(\hat p_c),$ in the window $\hat p_c=1-O(1/N)$, where $A$ ceases to be exponentially small, so that the below-threshold solution is matched to whatever lies beyond.

3. The exact-liquid counterpart. For the exact infinite-dimensional hard-sphere liquid (the ring-resummed, full-$L_3$ theory, whose thermodynamics coincides with the toy model below the threshold): does the Kirkwood instability at $\rho_1=\rho_{1c}$ terminate the liquid branch, or does a more complex continuation exist?

The three parts are independent sub-questions.

References

1. G. Parisi and F. Slanina, 'Toy model for the mean-field theory of hard-sphere liquids', Phys. Rev. E 62, 6554 (2000), DOI 10.1103/PhysRevE.62.6554, https://doi.org/10.1103/PhysRevE.62.6554 (preprint arXiv:cond-mat/0003205, https://arxiv.org/abs/cond-mat/0003205)
2. H. L. Frisch and J. K. Percus, 'High dimensionality as an organizing device for classical fluids', Phys. Rev. E 60, 2942 (1999), DOI 10.1103/PhysRevE.60.2942, https://doi.org/10.1103/PhysRevE.60.2942
3. M. Mézard and G. Parisi, 'A tentative replica study of the glass transition', J. Phys. A: Math. Gen. 29, 6515 (1996), DOI 10.1088/0305-4470/29/20/009, https://doi.org/10.1088/0305-4470/29/20/009
4. G. Parisi and F. Zamponi, 'Amorphous packings of hard spheres for large space dimension', J. Stat. Mech. (2006) P03017, DOI 10.1088/1742-5468/2006/03/P03017, https://doi.org/10.1088/1742-5468/2006/03/P03017 (preprint arXiv:cond-mat/0601573, https://arxiv.org/abs/cond-mat/0601573)
5. G. Parisi, 'On the Most Compact Regular Lattices in Large Dimensions: A Statistical Mechanical Approach', J. Stat. Phys. 132, 207 (2008), DOI 10.1007/s10955-008-9539-6, https://doi.org/10.1007/s10955-008-9539-6
6. G. Parisi and F. Zamponi, 'Mean-field theory of hard sphere glasses and jamming', Rev. Mod. Phys. 82, 789 (2010), DOI 10.1103/RevModPhys.82.789, https://doi.org/10.1103/RevModPhys.82.789 (preprint arXiv:0802.2180, https://arxiv.org/abs/0802.2180)
7. J. Kurchan, G. Parisi, and F. Zamponi, 'Exact theory of dense amorphous hard spheres in high dimension. I. The free energy', J. Stat. Mech. (2012) P10012, DOI 10.1088/1742-5468/2012/10/P10012, https://doi.org/10.1088/1742-5468/2012/10/P10012 (preprint arXiv:1208.0421, https://arxiv.org/abs/1208.0421)
8. P. Charbonneau, J. Kurchan, G. Parisi, P. Urbani, and F. Zamponi, 'Exact theory of dense amorphous hard spheres in high dimension. III. The full replica symmetry breaking solution', J. Stat. Mech. (2014) P10009, DOI 10.1088/1742-5468/2014/10/P10009, https://doi.org/10.1088/1742-5468/2014/10/P10009 (preprint arXiv:1310.2549 under the shorter title 'The full RSB solution', https://arxiv.org/abs/1310.2549)
9. J. Kurchan, T. Maimbourg, and F. Zamponi, 'Statics and dynamics of infinite-dimensional liquids and glasses: a parallel and compact derivation', J. Stat. Mech. (2016) 033210, DOI 10.1088/1742-5468/2016/03/033210, https://doi.org/10.1088/1742-5468/2016/03/033210 (preprint arXiv:1512.02186, https://arxiv.org/abs/1512.02186)
10. G. Parisi, Y. G. Pollack, I. Procaccia, C. Rainone, and M. Singh, 'Robustness of mean field theory for hard sphere models', Phys. Rev. E 97, 063003 (2018), DOI 10.1103/PhysRevE.97.063003, https://doi.org/10.1103/PhysRevE.97.063003 (preprint arXiv:1709.01607 under the different title 'How Generic are the Robust Theoretical Aspects of Jamming in Hard Sphere Models?', https://arxiv.org/abs/1709.01607)
11. P. Charbonneau, P. K. Morse, W. Perkins, and F. Zamponi, 'Three simple scenarios for high-dimensional sphere packings', Phys. Rev. E 104, 064612 (2021), DOI 10.1103/PhysRevE.104.064612, https://doi.org/10.1103/PhysRevE.104.064612 (preprint arXiv:2109.03063, https://arxiv.org/abs/2109.03063)
12. A. Ramezanpour and S. Moghimi-Araghi, 'Loop corrections for hard spheres in Hamming space', J. Stat. Mech. (2025) 033401, DOI 10.1088/1742-5468/adba42, https://doi.org/10.1088/1742-5468/adba42
13. Y. Tomita and H. Yoshino, 'Replicated liquid theory in $1+\infty$ dimensions', J. Chem. Phys. 163, 224108 (2025), DOI 10.1063/5.0300245, https://doi.org/10.1063/5.0300245
14. G. Parisi, P. Urbani, and F. Zamponi, 'Theory of Simple Glasses', Cambridge University Press (2019), DOI 10.1017/9781108120494, https://doi.org/10.1017/9781108120494