Voronoi Foams: Models & Optimization

• Voronoi foams are cellular structures defined by partitioning space into cells based on seed points, critical for modeling disordered and hyperuniform materials.
• They capture both local disorder and long-range order using metrics like cell volume, surface area, and topological variances for quantitative analysis.
• Their applications span mechanical optimization, light transport simulation, and inverse design in metallic, liquid, and biological foam systems.

Voronoi foams are space-filling cellular structures defined by partitioning space into polyhedral (in 3D) or polygonal (in 2D) cells, with each cell containing all points closer (or, in the Laguerre-Voronoi case, with lower power distance) to a given seed point (or weighted center) than to any other. These constructs are central to the theoretical modeling, simulation, and characterization of both physical foams (e.g., liquid or metallic open-cell foams) and biological or bio-inspired cellular packings. Voronoi foams are critical in understanding local disorder, large-scale hyperuniformity, mechanical optimization, light transport, and the interplay between geometry, topology, and physical properties across various applications in physics, materials science, and biology.

1. Mathematical Foundation and Key Definitions

The canonical Voronoi tessellation in Euclidean space $\mathbb{R}^d$ associates to a set of points $\{P_i\}$ a family of convex cells:

$$V_i = \{\mathbf{x} \in \mathbb{R}^d : \|\mathbf{x} - P_i\| \leq \|\mathbf{x} - P_j\|~\forall j \neq i\}$$

In the weighted case (Laguerre or power diagram), a system of weighted points (e.g., spheres in $\mathbb{R}^3$, $S = \{s_i = (c_i, r_i)\}$) yields cells:

$$\Omega_i = \{x \in \mathbb{R}^3 : d_P(x, s_i) \leq d_P(x, s_j)~\forall j \neq i\},$$

where $d_P(x, s_i) = \|x - c_i\|^2 - r_i^2$ (Jung et al., 27 Jan 2025). This construction is vital for representing realistic, polydisperse, or statistically fitted cellular microstructures extracted from experimental data (e.g., micro-CT of foams).

Key cell features include volume $V_i$, surface area $S_i$, mean edge length $L_{i,e}$, number of faces $\{P_i\}$0, and face-edge statistics $\{P_i\}$1, with derived statistics such as isoperimetric quotient $\{P_i\}$2 and topological variances $\{P_i\}$3, $\{P_i\}$4 (Sadjadi et al., 2012). These descriptors provide a quantitative basis for comparing physical foams and corresponding Voronoi or Laguerre models.

2. Local and Long-Range Structural Metrics

Voronoi foams exhibit a separation between local disorder and long-range order. Local quantities such as cell area or volume distributions, side-count statistics (polygons in 2D: $\{P_i\}$5, $\{P_i\}$6), and elongation metrics $\{P_i\}$7 are used to quantify local heterogeneity. In quasi-2D foams, area distribution widths $\{P_i\}$8 indicate that physical foams (1.82) are more disordered locally than Poisson (1.28), Halton (1.06), or Einstein (1.05) point-seeded Voronoi patterns (Chieco et al., 2021).

Long-range uniformity is characterized using spectral density $\{P_i\}$9 and, in real space, the hyperuniformity disorder length $$V_i = \{\mathbf{x} \in \mathbb{R}^d : \|\mathbf{x} - P_i\| \leq \|\mathbf{x} - P_j\|~\forall j \neq i\}$$0:

$$V_i = \{\mathbf{x} \in \mathbb{R}^d : \|\mathbf{x} - P_i\| \leq \|\mathbf{x} - P_j\|~\forall j \neq i\}$$1

Defining $$V_i = \{\mathbf{x} \in \mathbb{R}^d : \|\mathbf{x} - P_i\| \leq \|\mathbf{x} - P_j\|~\forall j \neq i\}$$2 as the effective width from the window boundary within which area fluctuations occur,

$$V_i = \{\mathbf{x} \in \mathbb{R}^d : \|\mathbf{x} - P_i\| \leq \|\mathbf{x} - P_j\|~\forall j \neq i\}$$3

Strong hyperuniformity yields $$V_i = \{\mathbf{x} \in \mathbb{R}^d : \|\mathbf{x} - P_i\| \leq \|\mathbf{x} - P_j\|~\forall j \neq i\}$$4 constant at large $$V_i = \{\mathbf{x} \in \mathbb{R}^d : \|\mathbf{x} - P_i\| \leq \|\mathbf{x} - P_j\|~\forall j \neq i\}$$5; maximal disorder gives $$V_i = \{\mathbf{x} \in \mathbb{R}^d : \|\mathbf{x} - P_i\| \leq \|\mathbf{x} - P_j\|~\forall j \neq i\}$$6; intermediate scaling $$V_i = \{\mathbf{x} \in \mathbb{R}^d : \|\mathbf{x} - P_i\| \leq \|\mathbf{x} - P_j\|~\forall j \neq i\}$$7 reflects power-law suppression of density fluctuations (Chieco et al., 2021).

Area-weighted centroid patterns in foams achieve $$V_i = \{\mathbf{x} \in \mathbb{R}^d : \|\mathbf{x} - P_i\| \leq \|\mathbf{x} - P_j\|~\forall j \neq i\}$$8, demonstrating maximal hyperuniformity at large scales. A single “anneal-once” relocation of Voronoi seeds to their centroids can reproduce this uniformity in artificial patterns, while true foams remain more locally disordered but are globally the most uniform configurations.

3. Topological Evolution and Phyllotactic Order

A subset of Voronoi foams—phyllotactic or spiral-based variants—exhibits quasi-crystalline order encoded by Fibonacci-sequenced neighbor and side-number distributions (Rivier et al., 2016). Seeds are placed according to the Fermat–Spiral algorithm,

$$V_i = \{\mathbf{x} \in \mathbb{R}^d : \|\mathbf{x} - P_i\| \leq \|\mathbf{x} - P_j\|~\forall j \neq i\}$$9

yielding annular grains separated by grain boundaries of pentagons, hexagons, and heptagons in Fibonacci-block sequences:

$\mathbb{R}^3$0

Topological transformations ($\mathbb{R}^3$1 edge-flips and $\mathbb{R}^3$2 cell disappearances) mediate local and global rearrangements while preserving the overall parastichy grammar. This construction achieves provable homogeneity (minimum area variance under disk-packing) and demonstrates how cellular structures can respond to shear or growth by local transformations without disrupting global order.

4. Morphology, Optimization, and Numerical Models

Voronoi foams serve as a flexible design space for mechanical and structural optimization. Open-cell foams are constructed by associating beams (struts) to the 1-skeleton (edges) of a Voronoi tessellation clipped to a geometric domain $\mathbb{R}^3$3 (Li et al., 2023). Each beam is modeled as a cylinder with radius the average of the adjacent seeds’ radii. The implicit density field

$\mathbb{R}^3$4

is defined using smooth maximum (KS) summation over signed distance functions for each beam, enabling finite-element discretization and differentiability with respect to seed positions and radii.

Optimization proceeds by minimizing structural compliance under volume constraints, using gradient-based algorithms with explicit sensitivities. Material-aware numerical coarsening reduces computational cost by coarse-graining local stiffness contributions while preserving fine-scale mechanical detail. This framework achieves load-bearing, fully connected foams at low volume fractions, with compliance reductions of 10–30% compared to voxel-based or infill-based methods (Li et al., 2023).

Realistic, experimentally validated models employ 3D image-based fitting of random Laguerre tessellations (weighted power diagrams) to micro-CT data, extracting both the sphere system (cell centers and radii) and the strut-graph (edges) (Jung et al., 27 Jan 2025). The inferred stochastic-geometric models enable direct finite-element or FFT-based homogenization, with effective moduli matching experimental compression tests within 5–10% error.

5. Physical Properties and Light Transport

Voronoi foams accurately model the bulk light-transport properties observed in real foams, especially in the context of dry foams where the Plateau border network is minimal. Ray-optics simulations with constant reflectance and thin-film reflectance models consistently find the photon transport-mean-free path $\mathbb{R}^3$5 set primarily by film reflectance rather than detailed geometric or topological disorder:

$\mathbb{R}^3$6

($\mathbb{R}^3$7 is the cell diameter), with corrections $\mathbb{R}^3$8 as disorder increases (Sadjadi et al., 2012). For real film parameters ($\mathbb{R}^3$9), $S = \{s_i = (c_i, r_i)\}$0 varies monotonically with film thickness and remains insensitive to disorder, confirming the universality of light-diffusive behavior across honeycomb, Voronoi, Kelvin, and disordered foams.

6. Curved Geometries and Confinement Effects

Voronoi foams confined between curved surfaces are generated by (i) evenly seeding points on a reference manifold (e.g., sphere, torus, Schwarz P surface) using energy minimization of geodesic repulsions, (ii) constructing and clipping the Euclidean Voronoi diagram to the confining walls (defined by level sets), and (iii) relaxing the polyhedral structure in the Surface Evolver under surface-tension and fixed-volume constraints (Mughal et al., 2016). This approach enables study of geometric statistics (mean area, area variance, polygon-side distribution) as functions of local Gaussian curvature, revealing, for example, a tendency for negative Gaussian curvature to favor heptagonal cells, while positive curvature favors pentagonal cells. Such models are essential for understanding biological and synthetic cellular assemblies on curved substrates and for engineering foam morphologies in non-Euclidean domains.

7. Practical Implications and Microstructure Optimization

The combined use of stochastic geometry, image-based reconstruction, and robust optimization protocols empowers direct microstructure–property mapping. For metallic open-cell foams, adjusting the parameters of the sphere packing (Laguerre tessellation) changes cell-size distribution, anisotropy, and elastic moduli, enabling inverse design for target mechanical properties (Jung et al., 27 Jan 2025). Inclusion of physically motivated relaxations (e.g., minimization under Plateau’s laws) can further refine predicted moduli. Mechanical and transport properties (e.g., Young’s modulus, compliance, anisotropy ratio, transport mean-free path) can be targeted systematically, establishing Voronoi foams as a foundational model for both the scientific understanding and engineering of cellular materials.