Statistical-mechanical open problems for the most compact regular lattices in large dimensions: the asymptotics of the maximal reduced packing radius $r_M$ and the lattice-versus-amorphous question (Parisi, 2008)

Parisi (2008) recast the search for the densest high-dimensional lattice sphere packings as the statistical mechanics of the ensemble of random covolume-one lattices introduced by Rogers: the maximal packing radius is to be extracted from $P(0,r)$, the probability that a random lattice has no nonzero vector shorter than the packing diameter, whose logarithm $S_L(r)$ plays the role of a lattice entropy. The paper left a coherent set of open problems: (i) the analytic-continuation conjecture, asserting that the connected-moment function $f_c(s)$ of the vector-count statistics, computable from Rogers' formulae only for $s \le D$, extends to an analytic function of $(s,D)$ so that $P(0,r)$ can be reconstructed from its moments in the packing regime $r>1$; (ii) whether the exponentially many neglected contributions (multi-linear-constraint classes, bridge diagrams, non-chain virial topologies) remain subleading as $D \to \infty$; (iii) whether the integral-equation entropy $S_E(r)$ (hypernetted-chain plus successive bridge diagrams) converges and equals $S_L(r)$; (iv) whether the maximal amorphous and lattice radii coincide, $r_A = r_M$, as $D \to \infty$; (v) whether a glass transition near $r_G \approx 1$ occurs in $S_L(r)$, fixing the ordering of $r_G, r_C, r_M, r_A$; and (vi) low- and moderate-dimension numerical tests. The hard-sphere prerequisite has since been solved exactly, Poisson statistics are proven only for ball volumes of order $D$ — far below the packing regime — and no component of the lattice-side core, in particular $\lim_{D \to \infty} r_M$ and the analytic-continuation conjecture, has been settled. The problem is open.

Background

How densely congruent spheres can fill $D$-dimensional space is a foundational and famously unsolved question when $D$ is large. A lattice packing restricts the sphere centers to the points of a lattice $\Lambda \subset \mathbb{R}^D$ (the integer combinations of $D$ linearly independent basis vectors). After rescaling $\Lambda$ to unit covolume (unit volume of its fundamental cell), a packing by spheres of diameter $R$ exists exactly when no nonzero lattice vector is shorter than $R$, and the packing density equals the volume $V_D(R/2)$ of a $D$-ball of radius $R/2$. The best known constructions give lattice densities of order $D \cdot 2^{-D}$, with the constant improved by Venkatesh (2013) and the order to $D^2 \cdot 2^{-D}$ by Klartag (2025); the best upper bound is $2^{-0.599D}$, obtained in 1978 by coding-theoretic methods and essentially unchanged since (Cohn and Zhao, 2014). Whether lattice packings exponentially denser than the $D \cdot 2^{-D}$ order exist is a central open question of the geometry of numbers.

Parisi (2008) approached it with the statistical mechanics of disordered systems. One endows the space of covolume-one (unimodular) lattices with its natural invariant probability measure — in practice the parametrization introduced by Rogers, in which a random lattice is generated from $D-1$ random parameters — and studies the random variable $K(R)_\Lambda$, the number of nonzero lattice points inside a ball $B(R)$ centered at the origin. Writing $P(k,R)$ for the probability that $K(R)_\Lambda = k$, the key quantity is $P(0,R)$: the probability that a random lattice avoids $B(R)$, i.e. that it carries a packing of diameter $R$. The maximal packing diameter $R_M$ is the largest $R$ with $P(0,R) > 0$. It is convenient to measure radii in reduced units: $R_c$ is defined by $V_D(R_c) = 1$ and $r = R/R_c$, so that a ball of radius $R$ has volume $r^D$ and the corresponding lattice-packing density is $(r/2)^D$. In these units the classical bounds read $1 \le r_M \le 1.322$, and the central question is the limit of $r_M$ as $D \to \infty$: the Minkowski-type lower bounds give only $r_M \to 1^{+}$, so any limit strictly above 1 would amount to an exponential improvement of the lattice-packing lower bound.

For small balls, the nonzero vectors of a random lattice behave approximately like a Poisson point process of intensity $1/2$ (the factor $1/2$ because $v$ and $-v$ come in pairs). Classical results quoted in the rigorous literature give $P(0) = e^{-V/2} + O(e^{-cD})$ for ball volumes $V < CD$, and Kim (2018) proved a Poisson description with exponentially small error for ball volumes up to order $n$, beyond which linear dependencies among short vectors appear; he explicitly identifies the packing regime, where the expected number of short vectors is exponential in the dimension, as the unsolved region closely related to the sphere-packing problem. Rogers had computed the moments $\langle K^s \rangle$ exactly, but only for $s \le D$. Parisi's program reconstructs $P(0,r)$ from the full moment sequence: introducing the connected moments (cumulants) of $K(R)$ and their generating function $C(z)$, the void probability is isolated in the limit $z \to -\infty$. The technical crux is the analytic-continuation conjecture: the connected-moment function $f_c(s)$, given by Rogers' finite-$s$ formulae inside their convergence region, is conjectured to extend to an analytic function of $s$ and $D$ jointly, with controlled poles, so that moments outside the directly computable region can be obtained by continuation. The paper verifies the needed analyticity only inside the region where Rogers' formulae converge and states: "We do not know how such a conjecture could be proved in a rigorous way."

The same computation makes contact with liquid-state physics. Keeping only the terms with a single linear constraint among the lattice vectors reproduces a subset of the diagrams of the virial expansion (the systematic low-density expansion of the entropy of a fluid in powers of the density, whose terms are indexed by Feynman-like diagrams). The hypernetted-chain (HNC) approximation — a standard integral-equation closure for the pair correlations of a liquid that neglects the bridge diagrams, a specific class of multiply-connected virial diagrams — then yields a computable entropy $S_E(r)$, and the conjecture is that including successive bridge diagrams converges to the lattice entropy $S_L(r) := \ln P(0,r)$, which diverges as $r \to r_M$. On the non-lattice side, $S(r)$ denotes the entropy density of the unconstrained hard-sphere gas in the same reduced units, with a crystallization transition conjectured at $r_C \approx 1$, a glass transition predicted by replica methods (a technique averaging over copies of a disordered system to reach frozen non-equilibrium phases) at $r_G \approx 1$, and a maximal amorphous packing radius $r_A$ at which the entropy of disordered packings diverges. The equality $S(r) \approx S_L(r)$ — hence $r_A = r_M$ — is the naive conjecture that Parisi explicitly distrusts ("only small chance to be true"), because exponentially many subleading diagrams could pile up.

Problem Statement

As posed by Parisi (2008, Sections 6.3 and 7.3-7.4), the problem is to carry out the statistical-mechanical determination of the maximal lattice-packing radius $r_M$ in the limit $D \to \infty$ and its relation to the unconstrained hard-sphere system. All components below jointly serve this single objective, which the source states as finding the infinite-dimension limit of $r_M$ (supposed to exist), with $r_M$ the largest reduced radius such that $P(0,r) > 0$ for the random covolume-one lattice ensemble and $R_c$ fixed by $V_D(R_c) = 1$:

1. Analytic continuation of the connected moments. Prove or disprove that the connected-moment function $f_c(s)$, defined by Rogers' finite-$s$ formulae (valid for $s \le D$), admits the analytic continuation in $(s,D)$ needed to evaluate the connected-moment generating function $C(z)$ and its $z \to -\infty$ limit, and thereby to control $P(k,R)$ and the tail of the void probability $P(0,r)$ in the packing regime; quantify the domain of validity and the nature of the singularities (poles, cuts) involved, and decide whether the single-linear-constraint integral representation of $f_c(s)$ captures the dominant singularity structure.
2. Control of the neglected contributions. Determine whether the exponentially many subleading terms discarded in the leading-order analysis — classes with several linear constraints among the lattice vectors, bridge diagrams, and other non-chain virial topologies — remain collectively subleading as $D \to \infty$; that is, prove or refute that they do not pile up and alter the leading conclusions, such as the Poisson law for $P(k,R)$ and the term-by-term agreement between the lattice and hard-sphere entropies.
3. Computation and convergence of the integral-equation entropy. Implement the proposed resummation of the lattice entropy via integral equations (the hypernetted-chain approximation plus successive bridge diagrams), evaluate $S_E(r)$ for reduced radii up to the conjectured packing thresholds, determine whether the procedure converges, and test the refined conjecture $S_E(r) = S_L(r)$.
4. Lattice versus amorphous packings. Determine whether the maximal amorphous packing radius $r_A$ (where the entropy of the unconstrained hard-sphere system diverges) equals the maximal lattice radius $r_M$ as $D \to \infty$ — equivalently whether $S(r) \approx S_L(r)$ in the infinite-dimensional limit — or whether amorphous (non-crystalline) packings can exceed the lattice packing density in high dimensions.
5. Glass and crystallization phenomena in the lattice ensemble. Determine whether a glass transition near $r_G \approx 1$ occurs in $S_L(r)$, whether a first-order (crystallization-like) transition appears, and what the ordering of $r_G, r_C, r_M, r_A$ is.
6. Numerical checks at low and moderate dimension. Compute $P(k,R)$ and $S_L(r)$ numerically for small and moderate $D$ to test the domain of the Poisson approximation, the analytic-continuation conjecture via extrapolation of the first moments, and the evidence for or against $S_E = S_L$.

The problem is to determine $\lim_{D \to \infty} r_M$ — in particular whether it exceeds 1 — together with its relation to the unconstrained hard-sphere system. The six components are independently refutable sub-claims. Components 1 (analytic continuation), 2 (control of the neglected contributions) and 3 (convergence of the integral-equation resummation) bear on the mechanism for extracting $\lim_{D \to \infty} r_M$: a proof or refutation of any one of them either supplies that mechanism or establishes that the statistical-mechanical route is blocked, with the failure mode identified (for example, a provable pile-up of a neglected class of diagrams, or divergence of the bridge-diagram expansion). Components 4 ($r_A$ versus $r_M$), 5 (the glass transition in $S_L(r)$ and the ordering of $r_G, r_C, r_M, r_A$) and 6 (low- and moderate-dimension numerics) are independently resolvable questions in which either direction — including $r_A \neq r_M$, the absence of a glass transition, or numerical evidence against the conjectures — is open, and settling any of them does not by itself determine $\lim_{D \to \infty} r_M$.

References

1. Giorgio Parisi, 'On the Most Compact Regular Lattices in Large Dimensions: A Statistical Mechanical Approach', Journal of Statistical Physics 132 (2008) 207-234, DOI 10.1007/s10955-008-9539-6, https://doi.org/10.1007/s10955-008-9539-6; preprint arXiv:0710.0882, https://arxiv.org/abs/0710.0882.
2. Giorgio Parisi, Francesco Zamponi, 'Mean-field theory of hard sphere glasses and jamming', Reviews of Modern Physics 82 (2010) 789-845, DOI 10.1103/RevModPhys.82.789, https://doi.org/10.1103/RevModPhys.82.789; arXiv:0802.2180, https://arxiv.org/abs/0802.2180.
3. Jorge Kurchan, Giorgio Parisi, Francesco Zamponi, 'Exact theory of dense amorphous hard spheres in high dimension I. The free energy', Journal of Statistical Mechanics: Theory and Experiment (2012) P10012, DOI 10.1088/1742-5468/2012/10/p10012, https://doi.org/10.1088/1742-5468/2012/10/p10012.
4. Jorge Kurchan, Giorgio Parisi, Pierfrancesco Urbani, Francesco Zamponi, 'Exact Theory of Dense Amorphous Hard Spheres in High Dimension. II. The High Density Regime and the Gardner Transition', The Journal of Physical Chemistry B 117 (2013) 12979-12994, DOI 10.1021/jp402235d, https://doi.org/10.1021/jp402235d.
5. Patrick Charbonneau, Jorge Kurchan, Giorgio Parisi, Pierfrancesco Urbani, Francesco Zamponi, 'Exact theory of dense amorphous hard spheres in high dimension. III. The full replica symmetry breaking solution', Journal of Statistical Mechanics: Theory and Experiment (2014) P10009, DOI 10.1088/1742-5468/2014/10/p10009, https://doi.org/10.1088/1742-5468/2014/10/p10009.
6. Giorgio Parisi, Pierfrancesco Urbani, Francesco Zamponi, 'Theory of Simple Glasses', Cambridge University Press (online 2019, print 2020), DOI 10.1017/9781108120494, https://doi.org/10.1017/9781108120494; chapter 'Packing Spheres in Large Dimensions', DOI 10.1017/9781108120494.009.
7. Anders Södergren, 'On the Poisson distribution of lengths of lattice vectors in a random lattice', Mathematische Zeitschrift 269 (2011) 945-954, DOI 10.1007/s00209-010-0772-8, https://doi.org/10.1007/s00209-010-0772-8; arXiv:1001.3623, https://arxiv.org/abs/1001.3623.
8. Seungki Kim, 'On the distribution of lengths of short vectors in a random lattice', Mathematische Zeitschrift 282 (2016) 1117-1126 (published online 2015), DOI 10.1007/s00209-015-1580-y, https://doi.org/10.1007/s00209-015-1580-y; arXiv:1410.2005, https://arxiv.org/abs/1410.2005.
9. Seungki Kim, 'Random Lattice Vectors in a Set of Size O(n)', International Mathematics Research Notices, Volume 2020, Issue 5 (published online 2018) 1385-1416, DOI 10.1093/imrn/rny060, https://doi.org/10.1093/imrn/rny060; arXiv:1611.02880, https://arxiv.org/abs/1611.02880.
10. Andreas Strömbergsson, 'On the probability of a random lattice avoiding a large convex set', Proceedings of the London Mathematical Society 103 (2011) 950-1006, DOI 10.1112/plms/pdr021, https://doi.org/10.1112/plms/pdr021; arXiv:1008.3805, https://arxiv.org/abs/1008.3805.
11. Matthew Jenssen, Felix Joos, Will Perkins, 'On the hard sphere model and sphere packings in high dimensions', Forum of Mathematics, Sigma 7 (2019) e1, DOI 10.1017/fms.2018.25, https://doi.org/10.1017/fms.2018.25; arXiv:1707.00476, https://arxiv.org/abs/1707.00476.
12. Patrick Charbonneau, Peter K. Morse, Will Perkins, Francesco Zamponi, 'Three simple scenarios for high-dimensional sphere packings', Physical Review E 104 (2021) 064612, DOI 10.1103/PhysRevE.104.064612, https://doi.org/10.1103/PhysRevE.104.064612; arXiv:2109.03063, https://arxiv.org/abs/2109.03063.
13. Boaz Klartag, 'Lattice packing of spheres in high dimensions using a stochastically evolving ellipsoid', Inventiones mathematicae 244 (2026) 1251-1279, DOI 10.1007/s00222-026-01412-w, https://doi.org/10.1007/s00222-026-01412-w; preprint arXiv:2504.05042 (2025), https://arxiv.org/abs/2504.05042.
14. Henry Cohn, Yufei Zhao, 'Sphere packing bounds via spherical codes', Duke Mathematical Journal 163, no. 10 (2014) 1965-2002, DOI 10.1215/00127094-2738857, https://doi.org/10.1215/00127094-2738857; arXiv:1212.5966, https://arxiv.org/abs/1212.5966.
15. Carlo Vanoni, Alan Guo, Salvatore Torquato, 'Analysis of Conjectural Improvements to Minkowski's Lower Bound on the Sphere Packing Density', arXiv:2608.00920 (2026), https://arxiv.org/abs/2608.00920.
16. S. Torquato, F. H. Stillinger, 'Jammed hard-particle packings: From Kepler to Bernal and beyond', Reviews of Modern Physics 82 (2010) 2633-2672, DOI 10.1103/RevModPhys.82.2633, https://doi.org/10.1103/RevModPhys.82.2633.
17. A. Andreanov, A. Scardicchio, 'Random perfect lattices and the sphere packing problem', Physical Review E 86 (2012) 041117, DOI 10.1103/PhysRevE.86.041117, https://doi.org/10.1103/PhysRevE.86.041117.
18. A. Andreanov, A. Scardicchio, S. Torquato, 'Extreme lattices: symmetries and decorrelation', Journal of Statistical Mechanics: Theory and Experiment (2016) 113301, DOI 10.1088/1742-5468/2016/11/113301, https://doi.org/10.1088/1742-5468/2016/11/113301.
19. Veit Elser, 'Packing spheres in high dimensions with moderate computational effort', Physical Review E 108 (2023) 034117, DOI 10.1103/PhysRevE.108.034117, https://doi.org/10.1103/PhysRevE.108.034117; preprint arXiv:2305.13492 (2023), posted under the title 'How densely can spheres be packed with moderate effort in high dimensions?', https://arxiv.org/abs/2305.13492.
20. Akshay Venkatesh, 'A Note on Sphere Packings in High Dimension', International Mathematics Research Notices 2013, no. 7 (2013) 1628-1642 (published online 2012), DOI 10.1093/imrn/rns096, https://doi.org/10.1093/imrn/rns096.