---
title: Spherical Shell Lattice Models
url: https://www.emergentmind.com/topics/spherical-shell-lattice-model
type: topic
---

# Spherical Shell Lattice Models

Searching arXiv for the cited paper and closely related spherical-shell lattice work.
“Spherical shell lattice model” denotes a class of constructions in which discrete degrees of freedom are constrained to a spherical manifold and their organization is determined by curvature, topology, and the interaction law. In the recent literature, the term covers nonequilibrium assembly of identical Lennard-Jones particles on the unit sphere [2603.12360], packed soft-core particles on a rigid spherical shell with competing hexagonal and square order [2501.01270], XY spins on a spherical Fibonacci lattice [2109.00254], ultra-soft cluster crystals on $S^2$ [1807.11303], Coulomb-coupled point particles on a sphere in a Thomson-type setting [2410.04311], and normalized shells of the $D_4$ lattice on $S^3$ studied as spherical designs [2303.09000]. Across these usages, the model family is characterized by a finite spherical geometry, unavoidable topological constraints, and a shell-based notion of order.

## 1. Terminological scope and geometric setting

In the particle-assembly literature, the shell is a finite cluster whose constituents are constrained to the surface of a sphere. Golushko et al. consider $N$ identical point particles on the surface of a sphere of unit radius $R=1$, with shell growth proceeding sequentially until no new particle can be attached with negative binding energy; the final shell size is denoted $n^\*$, and the energy per particle is $u=E_{\mathrm{tot}}/n^\*$ [2603.12360]. Xie et al. likewise place particles on a spherical surface $S^2(R)$, but study competition between hexagonal and square patterns under a Hertzian interaction [2501.01270]. In the ultra-soft cluster-crystal setting, the sphere is again the substrate, with distances measured along the great circle and the shell viewed as a curved two-dimensional assembly [1807.11303].

In lattice-spin work, the shell may be a discrete spherical sampling rather than a self-assembled particle cluster. The spherical Fibonacci lattice is constructed by placing $N$ points on a sphere of radius $R$ through
$$
z_i = R\left(\frac{2i-1}{N}-1\right), \qquad \phi_i = 2\pi i \phi,
$$
with $\phi=(\sqrt{5}-1)/2$, followed by the corresponding Cartesian embedding; nearest neighbors are then defined by a cutoff radius $r_c$ [2109.00254]. In the $D_4$-lattice setting, a shell is instead the set of lattice points of fixed squared norm, normalized onto the unit three-sphere $S^3$; this is a shell in the arithmetic sense rather than a physical particle shell [2303.09000].

This range of usage suggests that “spherical shell lattice model” is not a single standardized formalism. Rather, it names a family of models sharing the constraints of spherical embedding and finite topology, while differing in whether the microscopic objects are particles, spins, or lattice points.

## 2. Interaction laws and energy functionals

The simplest shell-assembly realization uses an isotropic Lennard-Jones pair potential depending on the geodesic chord distance $d$,
$$
V_{LJ}(d)=\varepsilon\left[\left(\frac{\sigma}{d}\right)^{12}-2\left(\frac{\sigma}{d}\right)^6\right],
$$
with equilibrium separation $d=\sigma$ at which $V_{LJ}=-\varepsilon$. A weak second minimum can be introduced through the Lennard-Jones-Gauss potential to mimic anisotropy [2603.12360]. The sole control parameter in the minimal model is $\sigma/R$, with $R=1$.

The competing Hex–Sq model employs a purely repulsive Hertzian contact potential of range $\sigma$ and energy scale $\varepsilon$,
$$
V(r)=
\begin{cases}
(1-r/\sigma)^{5/2}, & r<\sigma,\\
0, & \text{otherwise},
\end{cases}
$$
with the total energy $E[\{r_i\}]=\sum_{i<j}V(r_{ij})$. The spherical metric is
$$
ds^2 = R^2(d\theta^2+\sin^2\theta\, d\phi^2),
$$
and distances entering the interaction are computed from the chord distance $r_{ij}=2R\sin(\Delta\psi_{ij}/2)$ [2501.01270].

The XY model on a spherical Fibonacci lattice replaces particle coordinates by tangent-plane spins and introduces a Gaussian-screened interaction on the neighbor graph,
$$
H=-J\sum_{\langle i,j\rangle} e^{-\alpha r_{ij}^2}(s_i\cdot s_j),
$$
with $\alpha=10/R^2$ in the reported calculations [2109.00254]. The ultra-soft cluster-crystal model uses the GEM-4 interaction
$$
U(r)=\varepsilon \exp[-(r/\sigma)^4],
$$
written on the sphere through the geodesic distance $r=R\arccos(x\cdot x')$ or equivalently $U(\theta)=\varepsilon \exp[-(\theta/(\sigma/R))^4]$ [1807.11303].

The Coulomb spherical-lattice model takes
$$
U=\sum_{i<j}\frac{k_e (Zq)^2}{|r_i-r_j|},
$$
supplemented during equilibration by the viscous drag force $F_{\mathrm{drag},i}=-\nu |v_i|v_i$, with all forces projected to the tangent plane of the sphere [2410.04311]. In the elastic-shell many-body model, the shell itself is deformable: the discrete per-patch elastic energy is
$$
U_i=\frac{K}{2}\sum_{k=1}^{N_p}\frac{w_k}{\sigma^2}(\delta r_k)^2,
$$
and many-body overlap is handled by the rule $\delta r_k=\max_j \delta r_k^j$ [2005.14580].

## 3. Dynamical construction and computational protocols

The nonequilibrium Lennard-Jones shell grows in discrete steps. At each step, approximately $5\,000$ approximately equidistant trial points are placed in a spherical cap of angular radius approximately $1.5$ around each already attached particle. For a trial point $r$, the binding energy of the would-be $(n+1)$th particle is
$$
E_{\mathrm{attach}}(r)=\sum_{i=1}^n V\bigl(d(r,r_i)\bigr).
$$
The particle is attached at the trial point for which $E_{\mathrm{attach}}$ is most negative, and the entire cluster then undergoes constrained energy minimization on the sphere. Growth stops when no trial point yields a negative binding energy [2603.12360].

The same framework admits a finite-temperature generalization. If the local minima of the attachment energy are $E_j$, then one samples attachment site $j$ with probability
$$
p_j=\frac{\exp[-E_j/(k_B T)]}{\sum_k \exp[-E_k/(k_B T)]}.
$$
The limit $T\to 0$ reproduces deterministic growth, while $T\to\infty$ yields a random choice among local minima [2603.12360].

Other spherical shell lattice models use standard large-scale simulation protocols. The Hex–Sq study employs overdamped Langevin dynamics in LAMMPS with dimensionless time step $\Delta t=10^{-3}\tau$, friction $\gamma=1$, $20$ independent simulated-annealing runs per $(N,\rho^\*)$ point, and cooling from $k_B T/\varepsilon \simeq 0.1$ down to $10^{-4}$ [2501.01270]. The Fibonacci-sphere XY study uses classical Monte Carlo annealing with local Metropolis updates, with $O(1.5\times 10^8)$ MC steps for stable vortex-pattern snapshots and a one-layer graph-convolutional-network classifier trained on low- and high-temperature configurations [2109.00254]. The GEM-4 cluster-crystal model combines classical density functional theory, based on minimization of a grand-potential functional on $S^2$, with Monte Carlo simulations at fixed $N$, area $A=4\pi R^2$, and temperature [1807.11303]. The Coulomb “Crystal Ball” framework integrates Newton’s law with adaptive time step, exact reprojection onto the sphere, and a phase-transition protocol in which $N_{\rm rm}$ particles are removed around a random seed after equilibration [2410.04311]. The elastic-shell many-body model accelerates simulation by fitting the costly shell-deformation energy to $18$ symmetry functions, reducing evaluation time by $10^2$–$10^3\times$ before Monte Carlo phase-diagram calculations in the two-dimensional $NPT$ ensemble [2005.14580].

## 4. Topology, defects, and order metrics

A defining feature of spherical shell lattice models is that order is topologically frustrated. In the Lennard-Jones shells, most vertices are $6$-coordinated, but Euler’s theorem enforces exactly twelve $5$-coordinated defects in hexagonal order. When square tiles occur in square–triangular shells, vertices with coordination $4$ or $5$ appear, corresponding to $+90^\circ$ defect angles [2603.12360]. In the Hertzian model, the Gauss–Bonnet theorem gives the net disclination charge on $S^2$ as $\sum_k w_k=\chi(S^2)=2$, implying twelve $+1/6$ disclinations in a purely hexagonal lattice or eight $+1/4$ disclinations in a purely square lattice [2501.01270].

Xie et al. formulate this competition through bond-orientational order parameters
$$
\psi_{m,j}=\frac{1}{n_j}\sum_{k=1}^{n_j} e^{i m\theta_{jk}}, \qquad
\phi_m=\frac{1}{N}\sum_{j=1}^N |\psi_{m,j}|,
$$
with $m=6$ for Hex and $m=4$ for Sq. They also introduce the incompatibility angle
$$
\delta \equiv \theta_{\mathrm{sph}}-\theta_{\mathrm{flat}},
$$
which measures the mismatch between the interior angle of a spherical polygon and its Euclidean counterpart, and use it as a geometric criterion for the nucleation of counter-domains inside larger domains [2501.01270].

In the XY model, the analogous topological objects are vortices rather than disclinations. The vortex winding number is
$$
n=\frac{1}{2\pi}\oint \nabla \theta \cdot dl,
$$
and the Poincaré–Hopf theorem enforces total topological charge $+2$ for a continuous tangent-vector field on $S^2$, forcing at least two vortices in the ground state [2109.00254]. In the ultra-soft cluster-crystal model, topological frustration appears at the level of cluster occupancy and local stability: fivefold disclination clusters have smaller occupancy $N_c$, smaller nearest-neighbor spacing $d_{\rm nn}$, and higher local grand potential $\Omega_c$ than non-defective clusters [1807.11303].

The arithmetic shell model of $D_4$ encodes an algebraic analogue of spherical order. The normalized shell
$$
X_m=\frac{1}{\sqrt{2m}}(D_4)_{2m}\subset S^3
$$
is an antipodal spherical $\{10,4,2\}$-design for every $m\ge 1$, and the root shell $(D_4)_2$ is a tight $\{10,4,2\}$-design of size $24$ [2303.09000]. Here the “defect” language is replaced by harmonic vanishing conditions and linear-programming bounds, but the shell is still organized by spherical symmetry constraints.

## 5. Nonequilibrium Lennard-Jones shell assembly

Scanning the single parameter $\sigma\in[0.85,1.18]$ with $R=1$, in steps down to $\Delta\sigma=10^{-4}$, yields $43$ distinct shells with $n\le 72$ under the deterministic sequential protocol. These shells fall into three classes: icosahedral shells, intermediate-symmetry shells with a single $m$-fold axis $m>2$, and low-symmetry shells with no axis $m>2$ [2603.12360].

The icosahedral class contains $n=12$ and $n=32$ shells with symmetry $I_h$, corresponding to the classic $T=1$ and $T=3$ capsids. A deformed $T=7$ shell at $n=72$ appears with $D_3$ symmetry and, over a small $\sigma$ range, with $D_{5h}$ rather than perfect $I_h$. The classic $T=4$ shell at $n=42$ and the perfect $T=7$ icosahedron at $n=72$ do not assemble in the identical-LJ model [2603.12360].

A major result is the spontaneous formation of square–triangular surface order. The reported shells include $n=24$ with symmetry $O$, identified as the snub-cube Archimedean net; $n=44$ with symmetry $O_h$, described as a cubic S–T net with vacancies on the four-fold axes; $n=48$ with symmetry $O$; $n=60$ with symmetry $T$; and $n=64$ with symmetry $T$ [2603.12360]. These spherical polyhedral graphs correspond closely to experimental protein nanocages, including alpha-tocopherol transfer protein cages, ferritin mutants, and allophycocyanin assemblies, after superposition on PDB coordinates.

The energy landscape is organized into $\sigma$-bands. For each shell size one obtains a band in $u(\sigma)$; between bands, no shell of that size is stable, and band overlap can produce multiple isomers with the same $n$ but different symmetries. Deterministic growth yields energy-minimized shells for $n\le 72$. In stochastic assembly, $60$ runs at $T=0.16\varepsilon$, $T=\varepsilon$, and $T=10\varepsilon$ show more than $90\%$ perfect yield for high-symmetry shells such as $n=12,32,60,64$ even at $T=10\varepsilon$; most intermediate-symmetry shells with $n\le 50$ remain above $90\%$ yield for $T\le \varepsilon$; and low-symmetry shells with $n>50$ drop in yield but remain at least $20\%$ even at $T=10\varepsilon$ [2603.12360].

## 6. Other spherical-shell lattice realizations

The Hertzian Hex–Sq model identifies six distinct regimes as the reduced density $\rho^\*=N\sigma^2/(4\pi R^2)$ is varied: pure Hex with point disclinations or scars, Hex background with Sq domains, nested domain/counter-domain patterns, interwoven bands of both tilings, Sq background with Hex domains, and pure Sq with eight $+1/4$ defects. The rich defect morphologies occur in the narrow window $2.20<\rho^\*<2.40$, where Hex and Sq free energies are nearly equal, and the excess disclination number follows the scaling $N_d\sim \alpha N^{1/2}$ with $\alpha_{\rm hex}/\alpha_{\rm sq}\approx 0.38$ [2501.01270].

On the spherical Fibonacci lattice, the near-uniformity of the graph discretization is central. For $N=1000$ and $r_c/R=0.11395$, the lattice has $3998$ neighbor pairs; $850$ sites have $4$ neighbors, $76$ have $3$, and $74$ have $5$. At $T/J=5\times 10^{-4}$, $N=1000$, and $r_c/R=0.11395$, two $+1$ vortices remain after full annealing and sit approximately $131.5^\circ$ apart; for $r_c/R=0.2$, they become nearly antipodal at $174.3^\circ$. The graph-convolutional-network predicts $T_c/J=1.06$ for $r_c/R=0.11395$, $1.40$ for $r_c/R=0.12$, and $1.60$ for $r_c/R=0.13$ [2109.00254].

The GEM-4 cluster-crystal model generalizes the bulk clustering criterion to the sphere by expanding the potential in Legendre modes and identifying a $\lambda$-line through
$$
1+\beta \rho_\lambda \tilde U_{\ell^\*}=0.
$$
At fixed $T^\*=k_B T/\varepsilon=1$ and $\sigma/R=0.5$, one finds $\rho_\lambda R^2 \simeq 20$–$25$, above which a $32$-cluster crystal is stable. At fixed $\rho R^2\simeq 87$, varying $\sigma/R$ yields cluster phases with $10,16,20,32,\ldots$ clusters. The occupancy per site grows linearly with $\rho$, while $d_{\rm nn}$ remains nearly constant, which is the reported hallmark of cluster crystals [1807.11303].

The Coulomb spherical-lattice model studies rearrangement rather than static pattern formation. For $N=100,200,\ldots,1000$ and $N_{\rm rm}=1,2,\ldots,10$, with $24$ Monte Carlo trials per pair, the analytic fit
$$
T(N,N_{\rm rm})=
c\,k_e Z^2 q^2\,N_{\rm rm}\,[3+\sqrt{3(4N_{\rm rm}-1)}]/d(N)
$$
gives $c=0.0203\pm 0.0004$ at $95\%$ confidence and reduced $\chi^2\approx 2.37$. At $N=500$, the reported average peak kinetic energy rises from approximately $10$ eV for $N_{\rm rm}=1$ to approximately $120$ eV for $N_{\rm rm}=7$ [2410.04311].

The elastic-shell many-body model yields a simpler phase diagram. For all $\beta K\in[1,500]$, only two stable phases are reported: a low-density disordered fluid with $\chi_6\approx 0$ and a high-density hexagonal crystal with $\chi_6\approx 1$. At $\beta K=200$, coexistence occurs at $\rho \sigma^2\approx 3.05$–$3.35$, corresponding to $a/\sigma\approx 0.78$–$0.85$ [2005.14580].

## 7. Significance, limitations, and broader implications

The nonequilibrium Lennard-Jones study shows that a minimal model with only one geometric parameter, $\sigma/R$, reproduces both small icosahedral viral capsids and a broad set of octahedral, tetrahedral, and square–triangular shells. It further shows that nonequilibrium sequential attachment into the local deepest binding site often leads directly to global or near-global energy minima, especially for high symmetry, and that finite-temperature stochastic growth remains robust for the most symmetric shells [2603.12360]. This suggests that kinetic growth rules alone can be structurally selective on curved manifolds.

The competing-lattice and cluster-crystal studies emphasize a complementary principle: curvature does not merely perturb flat-space order, but changes the admissible defect content, domain morphology, and stability criteria. In the Hex–Sq system, the incompatibility angle provides a simple geometric predictor for when a domain must nucleate a counter-domain [2501.01270]. In the ultra-soft cluster-crystal system, the spherical harmonic decomposition of the interaction identifies the $\lambda$-line and links preferred cluster number to the first negative harmonic mode $\ell^\*$ [1807.11303].

The arithmetic shell perspective extends the topic beyond soft matter. The $D_4$ results show that shells normalized onto $S^3$ can be classified through spherical design theory, linear-programming bounds, and modular forms, with the root shell uniquely realizing the antipodal spherical $\{10,4,2\}$-design of size $24$ [2303.09000]. A plausible implication is that “shell lattice” can refer either to self-assembled physical shells or to highly symmetric norm shells of a Euclidean lattice.

Limitations are model-specific and explicit in the cited work. The Coulomb model is purely classical and ignores quantum tunneling, electron screening, magnetic fields, and relativistic effects [2410.04311]. The elastic-shell many-body model replaces a direct shell-deformation calculation by a symmetry-function regression surrogate [2005.14580]. The Lennard-Jones assembly model uses identical particles and isotropic pair interactions unless the Lennard-Jones-Gauss extension is invoked [2603.12360]. Even with these simplifications, the models collectively provide a blueprint for rational shell design in protein nanocages, colloidosomes, micelles, spherical spin systems, and mathematically defined spherical designs.

Source: https://www.emergentmind.com/topics/spherical-shell-lattice-model