Mean-Field 3D Vertex Model

• Mean-field three-dimensional vertex models are approaches that reduce complex 3D interfaces and vertex configurations to self-consistent fields or effective elastic media.
• They integrate discrete spin and local vertex constraints through methods like hard-spin mean-field theory, tetrahedron-equation integrability, and representative cell approximations.
• Applications range from modeling Ising interface roughening to predicting elastic moduli in cell-based systems and defect-induced instabilities in epithelial tissues.

A mean-field three-dimensional vertex model is not a single canonical object but a family of constructions in which a three-dimensional vertex or interface problem is reduced to self-consistent fields, representative-cell variables, or an effective elastic medium. In the literature surveyed here, the term covers at least three technical settings: a hard-spin mean-field treatment of a three-dimensional Ising interface whose low-energy solid-on-solid limit is an effective two-dimensional SOS/vertex model (Çağlar et al., 2011), an algebraic three-dimensional two-color vertex model built from $R$-matrices satisfying the tetrahedron equation and explicitly formulated in terms of local vertex weights (Korepanov, 2016), and biological three-dimensional vertex models in which a truncated octahedron, a regular hexagonal cell, or a single defect cell embedded in a homogeneous shell plays the role of the mean-field degree of freedom (Kim et al., 2023, Drozdowski et al., 2024, Drozdowski et al., 2024).

1. Scope and meanings of the mean-field reduction

These works suggest that “mean field” enters three-dimensional vertex modeling in three distinct senses. In hard-spin mean-field theory, the local discrete variables are retained but neighbor correlations are approximated by an independent-product distribution, producing self-consistent equations for local magnetizations. In ordered cell-based vertex models, the many-cell packing is replaced by a single representative polyhedron whose reduced shape parameters encode the mechanical response of the whole tessellation. In spherical epithelial models, the tissue outside a distinguished defect cell is replaced by a homogeneous elastic shell with analytically derived stretching and bending moduli, while only the defect core remains discrete [(Çağlar et al., 2011); (Kim et al., 2023); (Drozdowski et al., 2024); (Drozdowski et al., 2024)].

Setting	Mean-field variables	Principal output
Anisotropic 3D Ising interface	Layer magnetizations $m_z$	Ordering and roughening phase diagram
3D two-color vertex model	Local vertex-state structure of $\mathcal R(a)$	Tetrahedron-equation-constrained 16-vertex weights
Ordered cell-based 3D vertex model	Shape parameters of a truncated octahedron	Elastic moduli and compatible–incompatible transition
Spherical epithelial vertex models	Continuum shell fields plus one explicit defect cell	Buckling, faceting, bulging, and extrusion thresholds

In all cases, the three-dimensionality is essential. The interface model distinguishes the existence of roughening in $d=3$ from its absence in $d=2$. The integrable construction replaces the Yang–Baxter equation by the tetrahedron equation. The biological models derive genuinely three-dimensional shape indices, bending moduli, and defect-induced shell instabilities that have no direct two-dimensional equivalent [(Çağlar et al., 2011); (Korepanov, 2016); (Kim et al., 2023); (Drozdowski et al., 2024)].

2. Hard-spin mean-field theory as a three-dimensional interface/vertex model

In "Interface-Roughening Phase Diagram of the Three-Dimensional Ising Model for All Interaction Anisotropies from Hard-Spin Mean-Field Theory" (Çağlar et al., 2011), the microscopic system is the uniaxially anisotropic three-dimensional Ising model

$$-\beta \mathcal H = J_{xy}\sum_{\langle ij\rangle}^{xy} s_i s_j + J_z \sum_{\langle ij\rangle}^{z} s_i s_j,$$

with $s_i=\pm1$, ferromagnetic couplings $J_{xy}>0$, $J_z>0$, and one special antiferromagnetic interplane bond $J_z^A=-J_z$ inserted between a single pair of adjacent $m_z$0-planes to force a bulk domain wall. The anisotropy parameter is $m_z$1, varied from the decoupled-planes limit $m_z$2 through the isotropic case $m_z$3 to the solid-on-solid limit $m_z$4.

The hard-spin mean-field closure is

$m_z$5

with $m_z$6. This differs from ordinary mean field $m_z$7 because the discrete neighbor-spin configurations are kept inside the nonlinear $m_z$8 factor and only their correlations are factorized. On a $m_z$9 cubic lattice with periodic boundary conditions, translational symmetry in the plane reduces the solution to layer magnetizations $\mathcal R(a)$0, obtained by numerical iteration of the self-consistent equations.

The vertex-model interpretation is indirect but structurally important. In the solid-on-solid limit $\mathcal R(a)$1, vertical chains become rigid, the forced interface can be represented by a height field $\mathcal R(a)$2, and overhangs and bubbles are suppressed. A height model on a two-dimensional lattice can often be recast into a vertex model, so the hard-spin mean-field treatment of the original three-dimensional Ising spins functions as a mean-field theory for the effective interface degrees of freedom.

The interface roughness is measured through the bulk magnetization magnitude $\mathcal R(a)$3 and the averaged deviation

$\mathcal R(a)$4

A rough interface gives a nonzero averaged deviation; a smooth localized interface gives zero in the thermodynamic limit. The roughening transition temperature $\mathcal R(a)$5 is defined by the vanishing of this quantity on cooling, while the ordering transition $\mathcal R(a)$6 is defined by the onset of nonzero $\mathcal R(a)$7.

The resulting phase diagram contains a disordered phase, an ordered phase with rough interface, and an ordered phase with smooth interface. For $\mathcal R(a)$8, the ordering transition is found at $\mathcal R(a)$9, compared with the exact two-dimensional Ising value $d=3$0. For the isotropic three-dimensional case $d=3$1, the mean-field ordering temperature is $d=3$2, compared with the Monte Carlo result $d=3$3. The roughening line rises from effectively zero in the weak-coupling limit, reaches $d=3$4 in the isotropic case, and saturates in the solid-on-solid limit at $d=3$5, to be compared with Swendsen’s Monte Carlo SOS value $d=3$6. Repeating the calculation in $d=3$7 yields no roughening transition: the deviation does not vanish down to zero temperature, so the interface does not localize. This establishes a specifically three-dimensional roughening phenomenon within a mean-field interface/vertex framework.

3. Algebraic three-dimensional vertex models and tetrahedron-equation structure

A different branch of the subject is represented by "Three-dimensionalizing the eight-vertex model" (Korepanov, 2016), which does not itself solve a mean-field theory but provides an explicit three-dimensional vertex model whose local variables and constraints are suitable for mean-field or effective-field formulations. The construction starts from two-color permutation-type operators in $d=3$8-valued spaces and generalizes the two-dimensional combination

$d=3$9

to a three-leg operator

$d=2$0

The local degrees of freedom are binary color states on each leg, with basis $d=2$1. The three-dimensional permutation-type operators are

$d=2$2

$d=2$3

with all sums modulo $d=2$4. Hence

$d=2$5

For each incoming triple there are up to two possible outgoing triples, with weights $d=2$6 and $d=2$7, so the model is a three-dimensional two-color 16-vertex model.

Integrability is encoded by the tetrahedron equation

$d=2$8

which holds when the quadruple $d=2$9 lies on an algebraic set with five irreducible two-dimensional components. Among these, the condition

$$-\beta \mathcal H = J_{xy}\sum_{\langle ij\rangle}^{xy} s_i s_j + J_z \sum_{\langle ij\rangle}^{z} s_i s_j,$$0

is the direct three-dimensional analog of the two-dimensional relation $$-\beta \mathcal H = J_{xy}\sum_{\langle ij\rangle}^{xy} s_i s_j + J_z \sum_{\langle ij\rangle}^{z} s_i s_j,$$1. One can write $$-\beta \mathcal H = J_{xy}\sum_{\langle ij\rangle}^{xy} s_i s_j + J_z \sum_{\langle ij\rangle}^{z} s_i s_j,$$2, making the analogy with additive rapidity parameterizations explicit.

For mean-field purposes, the importance of this model lies in the fact that all local ingredients are explicit: binary edge variables, deterministic local constraints, and Boltzmann weights $$-\beta \mathcal H = J_{xy}\sum_{\langle ij\rangle}^{xy} s_i s_j + J_z \sum_{\langle ij\rangle}^{z} s_i s_j,$$3, $$-\beta \mathcal H = J_{xy}\sum_{\langle ij\rangle}^{xy} s_i s_j + J_z \sum_{\langle ij\rangle}^{z} s_i s_j,$$4, or $$-\beta \mathcal H = J_{xy}\sum_{\langle ij\rangle}^{xy} s_i s_j + J_z \sum_{\langle ij\rangle}^{z} s_i s_j,$$5. The paper notes that for $$-\beta \mathcal H = J_{xy}\sum_{\langle ij\rangle}^{xy} s_i s_j + J_z \sum_{\langle ij\rangle}^{z} s_i s_j,$$6 the weights are positive, making the construction physically reasonable as a statistical-mechanical model. It also states that a practical mean-field or effective-field treatment could identify local vertex-state probabilities or average height-gradient variables as order parameters and build self-consistency equations analogous to hard-spin mean-field closures. The algebraic tetrahedron-equation setting therefore supplies a rigorously defined three-dimensional vertex substrate on which mean-field approximations can be erected.

4. Ordered cell-based mean-field vertex models and the three-dimensional shape index

In "Mean field elastic moduli of a three-dimensional cell-based vertex model" (Kim et al., 2023), mean field means an ordered tiling of space by identical cells, each represented by a truncated octahedron. The cell energy is the three-dimensional analog of the standard quadratic area–perimeter vertex functional: $$-\beta \mathcal H = J_{xy}\sum_{\langle ij\rangle}^{xy} s_i s_j + J_z \sum_{\langle ij\rangle}^{z} s_i s_j,$$7 For an ordered packing of identical cells, this reduces to the single-cell dimensionless energy

$$-\beta \mathcal H = J_{xy}\sum_{\langle ij\rangle}^{xy} s_i s_j + J_z \sum_{\langle ij\rangle}^{z} s_i s_j,$$8

where $$-\beta \mathcal H = J_{xy}\sum_{\langle ij\rangle}^{xy} s_i s_j + J_z \sum_{\langle ij\rangle}^{z} s_i s_j,$$9, $s_i=\pm1$0, $s_i=\pm1$1, and $s_i=\pm1$2. The control parameter $s_i=\pm1$3 is the three-dimensional shape index.

The representative cell is parameterized by a reduced set of shape variables rather than by all vertex coordinates. In one parameterization the truncated octahedron is described by $s_i=\pm1$4, and for the regular truncated octahedron with $s_i=\pm1$5 the resulting shape index is

$s_i=\pm1$6

This value is the mean-field compatible–incompatible threshold $s_i=\pm1$7. For $s_i=\pm1$8, the target volume and target area cannot be achieved simultaneously, the minimum energy is strictly positive, and the state is incompatible. For $s_i=\pm1$9, a zero-energy compatible state exists in which both springs are unstrained.

Elastic moduli are extracted from the curvature of the energy under small deformations. The analysis distinguishes constrained, purely affine response from relaxed response in which shape parameters are allowed to adjust nonaffinely at fixed macroscopic strain. This distinction is central. In the compatible regime, relaxed moduli can soften strongly and may vanish for certain deformation protocols because the cell can move along zero-energy shape directions. In the incompatible regime, finite prestress remains even at zero imposed strain.

The paper also shows that the rigidity transition and the elastic moduli depend on the parameterization of cell shape. Reparameterizations that allow the truncated octahedron to morph into elongated dodecahedra, rhombic dodecahedra, staggered hexagonal cells, or trigonal trapezohedra modify the accessible shape landscape and produce nontrivial, sometimes multi-peaked modulus curves. This makes the mean-field theory explicitly shape-landscape dependent: the same quadratic energy can yield qualitatively different linear-response behavior depending on which reduced variables are taken to span the admissible three-dimensional polyhedral family.

5. Continuum shell mean fields for spherical epithelia

"Morphological instability at topological defects in a three-dimensional vertex model for spherical epithelia" (Drozdowski et al., 2024) converts a discrete three-dimensional vertex monolayer into a continuum elastic shell by a mean-field derivation of stretching and bending moduli. The microscopic energy is

$J_{xy}>0$0

with conserved cell volume $J_{xy}>0$1. The mean-field derivation assumes a regular hexagonal lattice, homogeneous strain at the cell level, and nonaffine relaxation between two interpenetrating sublattices.

For the flat configuration, the continuum stretching energy density takes the standard isotropic form

$J_{xy}>0$2

and the effective Lamé coefficients satisfy

$J_{xy}>0$3

Hence

$J_{xy}>0$4

A notable consequence is that the in-plane elastic response depends only on the sum $J_{xy}>0$5 and is independent of $J_{xy}>0$6. The paper interprets $J_{xy}>0$7 as evidence that the sheet is compressible in two dimensions because deformations can be partially accommodated by changes of cell height.

The bending sector is written in Helfrich form,

$J_{xy}>0$8

with mean curvature $J_{xy}>0$9, Gaussian curvature $J_z>0$0, bending rigidity $J_z>0$1, saddle-splay modulus $J_z>0$2, and spontaneous curvature $J_z>0$3. The mean-field derivation yields the scalings

$J_z>0$4

Thus apico-basal tension asymmetry generates spontaneous curvature, and the scaling of $J_z>0$5 differs qualitatively from that of a simple elastic plate.

With these moduli, the vertex monolayer is mapped onto a thin-shell theory that reproduces both flat-sheet buckling and defect-induced faceting. For spherical epithelia, the relevant control parameter is the Föppl–von Kármán number

$J_z>0$6

Using the classical result for twelve five-fold disclinations on a sphere, the faceting transition occurs at $J_z>0$7, with an empirical nonlinearity factor $J_z>0$8 giving excellent collapse of simulation data. The corresponding critical radius scales as $J_z>0$9. Localized apico-basal tension asymmetry around defect cells lowers the transition threshold to smaller system sizes, showing how a mean-field shell description can incorporate both passive topological frustration and localized active bias.

6. Defect-centered mean-field bubbly vertex models

"Cell bulging and extrusion in a three-dimensional bubbly vertex model for curved epithelial sheets" (Drozdowski et al., 2024) refines the shell picture by treating one defect cell explicitly while the rest of the tissue is replaced by a mean-field elastic shell. The bubbly vertex model retains the same tension-based energy,

$J_z^A=-J_z$0

but allows apical, basal, and lateral faces to be curved rather than planar. Each cell volume is fixed to $J_z^A=-J_z$1 after nondimensionalization and $J_z^A=-J_z$2. In this setting, tissue-scale curvature and cell-scale interfacial curvature coexist.

The mean-field construction embeds a single explicit defect cell, typically a pentagon, into a homogeneous shell characterized by effective elastic constants $J_z^A=-J_z$3. The defect cell is parameterized as a regular $J_z^A=-J_z$4-gon pyramid with spherical apical and basal caps, and its state is described by a reduced energy landscape in variables such as the opening angle $J_z^A=-J_z$5 and basal edge length $J_z^A=-J_z$6. The total energy combines the discrete core surface energy, the mean-curvature bending energy of the surrounding cap, the Gaussian-curvature term obtained through Gauss–Bonnet, the conical outer-shell contribution, the annular stretching energy, and a boundary correction that removes double counting between the explicit defect and the continuum shell.

Within this mean-field landscape, bulging and extrusion are energetically preferred at topological defects because Gaussian curvature is redistributed into the defect. The stretching energy contains the disclination charge explicitly, so allowing the defect cell to adopt a larger opening angle and curved interfaces reduces the elastic cost of accommodating a pentagonal deficit angle. The bubbly model amplifies this effect relative to a standard flat-face vertex model because the defect can lower its core energy by becoming more spherical at fixed volume.

The model identifies several control mechanisms. Extrusion can be driven by a decrease in apico-basal tension or by contractile line tensions. For surrounding tissue with $J_z^A=-J_z$7, the critical defect-tension reduction required for extrusion is lower for pentagons than for hexagons, and the difference $J_z^A=-J_z$8 is of order $J_z^A=-J_z$9. Contractile basal line tension favors smaller basal perimeter and can initiate bulging, although it does not remove the final barrier between a fully bulged and a fully extruded state. Conversely, luminal pressure and interfacial bending rigidity suppress bulging and narrow the distribution of opening angles, stabilizing a more homogeneous shell.

7. Conceptual synthesis and dimensional structure

Taken together, these constructions show that a mean-field three-dimensional vertex model is best understood as a methodological category rather than a single Hamiltonian. In one variant, the mean-field variable is a self-consistent magnetization profile $m_z$00 encoding the width and localization of an Ising interface whose solid-on-solid limit is equivalent to a vertex model. In a second, the local vertex weights are defined exactly by $m_z$01-matrices satisfying the tetrahedron equation, and mean field enters as a prospective effective-field approximation to a rigorously specified three-dimensional 16-vertex structure. In a third, the many-cell geometry is replaced either by a representative truncated octahedron or by an elastic shell with analytically derived moduli, while local shape or defect variables remain explicit [(Çağlar et al., 2011); (Korepanov, 2016); (Kim et al., 2023); (Drozdowski et al., 2024); (Drozdowski et al., 2024)].

A persistent theme is that three-dimensionality changes the phase structure and the mechanical response. The hard-spin interface theory finds a roughening transition in $m_z$02 and none in $m_z$03. The integrable model replaces Yang–Baxter structure by tetrahedron-equation structure. The biological vertex models introduce a genuine three-dimensional shape index $m_z$04, a compatible–incompatible transition at $m_z$05, and shell instabilities governed by $m_z$06, topological disclinations, and apico-basal asymmetry. This suggests that mean-field reduction in three-dimensional vertex systems is most effective when it preserves the discrete or geometric ingredient that carries the essential physics: spin discreteness for interface roughening, exact local vertex constraints for integrable models, or cell-shape and defect geometry for epithelial mechanics.