💻 computer science

Efficient generation of large-scale non-equilibrium distributions of particles

This paper introduces the Swelling and Random Migration (SRM) algorithm, a highly efficient method capable of generating statistically representative, large-scale non-equilibrium microstructures of up to $10^7$ particles in both spherical and non-spherical forms to advance the modeling of composite materials.

Original authors: Sergejs Tarasovs

Published 2026-05-19

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Sergejs Tarasovs

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to build a model of a composite material, like a strong plastic filled with tiny fibers or particles. To understand how strong or conductive this material is, scientists need to create a perfect digital "snapshot" of how those particles are arranged inside.

Usually, computers try to arrange these particles like they are in a calm, peaceful state (equilibrium), where everything is spread out evenly. But in the real world, manufacturing processes (like injection molding or mixing) often leave particles clumped together, forming messy, uneven clusters. These "messy" states are called non-equilibrium distributions, and they are hard to simulate because standard computer methods just try to smooth everything out.

This paper introduces a new, super-fast computer method called SRM (Swelling and Random Migration) that can create these messy, realistic particle arrangements efficiently.

Here is how it works, using some everyday analogies:

1. The "Swelling Balloon" Game

Imagine you have a room full of deflated balloons (the particles).

The Goal: You want to fill the room until the balloons take up a specific amount of space (the target volume), but they can't pop or overlap.
The Problem: If you just blow them up all at once, they get stuck in a jam.
The SRM Solution:
- Swelling: You gently blow a little air into every balloon at the same time.
- Random Migration: If two balloons bump into each other, you give them a tiny, random nudge (a "migration") to see if they can find a new spot where they don't touch.
- The "Shake": If the balloons get stuck in a tight knot and can't move, the computer gives the whole room a big, random "shake" to loosen things up, then tries again.

By repeating this cycle of "blow up a little, nudge a little, shake if stuck," the algorithm can pack millions of particles together without them overlapping, even if they end up in messy, clumpy groups.

2. The "Smart Neighborhood" Trick

Checking if millions of balloons are touching each other is usually a slow, boring task. If you check every balloon against every other balloon, it takes forever (like checking every person in a stadium against every other person to see if they are standing too close).

The SRM algorithm uses a clever trick called Cell-Based Search:

Imagine dividing the room into a grid of small neighborhoods (cells).
The computer only checks if a balloon is touching its neighbors in the same neighborhood or the eight neighborhoods right next to it.
It ignores everyone else in the stadium because they are too far away to matter.
This makes the process incredibly fast, allowing the computer to handle up to 10 million particles (10^7) in a reasonable amount of time.

3. What Can It Do? (The Results)

The paper shows that this method is very flexible:

From Calm to Chaotic: You can tune the "nudge" and "shake" settings to create anything from a perfectly calm, evenly spread crowd (equilibrium) to a chaotic, clumpy crowd (non-equilibrium).
The "House of Cards": The researchers tested this with flat, coin-shaped particles (platelets). By adjusting the settings, they could make the platelets stack up like a messy "house of cards" (where edges touch faces) or line up in neat, parallel rows (like a school of fish).
Why It Matters: They found that these different arrangements change how electricity or heat moves through the material. For example, the "house of cards" structure allowed electricity to jump across gaps much easier than the neat, parallel arrangement.

Summary

In short, the author created a digital tool that acts like a very fast, very smart game of "musical chairs" for particles. It can generate huge, realistic models of how particles clump together in real-world materials, which helps scientists understand why some materials are stronger or more conductive than others.

Important Note: The paper emphasizes that this is a geometric tool. It creates shapes and arrangements based on rules of space, not based on real physics forces (like magnetism or gravity). It tells you what the arrangement looks like, but it doesn't simulate the actual physical forces that created it.

Technical Summary: Efficient Generation of Large-Scale Non-Equilibrium Distributions of Particles

Problem Statement
Representative Volume Elements (RVEs) with periodic microstructures are essential for linking microscale heterogeneity to macroscale effective properties in composite materials. While traditional algorithms (e.g., Monte Carlo, Lubachevsky–Stillinger event-driven molecular dynamics) excel at generating equilibrium or near-equilibrium particle configurations, they often fail to capture the non-uniform, anisotropic, and history-dependent distributions characteristic of real-world manufacturing processes (e.g., filament winding, injection molding). Existing methods tend to suppress the irregularities, defects, and controlled disorder found in actual composites. Furthermore, generating large-scale microstructures with high particle counts often suffers from prohibitive computational costs due to quadratic scaling in collision detection. There is a distinct need for an algorithm capable of efficiently producing periodic RVEs with tunable degrees of randomness, controlled spatial correlations, and realistic non-equilibrium features, including clustering and agglomeration.

Methodology: The Swelling and Random Migration (SRM) Algorithm
The paper proposes the Swelling and Random Migration (SRM) algorithm, a hybrid approach combining a modified Miller–Torquato "swelling" technique with a modified Random Sequential Addition (RSA) scheme. The method operates on a unit box containing monodisperse inclusions and proceeds through the following iterative steps:

Initialization: An initial non-overlapping configuration is generated at a low volume fraction using classical RSA.
Swelling: All particles are simultaneously increased in radius by a small increment ( $c_w$ ).
Random Migration: Each particle is displaced by a small random vector ( $c_m$ ).
Local Admissibility Check: Distances to neighbors are evaluated using an adaptive cell-based spatial partitioning scheme. If a particle overlaps, it is returned to its previous position, and a new random displacement is attempted. This loop continues until a valid position is found or a maximum number of attempts ( $N_l$ ) is reached.
Shake Step: If a particle cannot find an admissible position after several attempts, a global relaxation ("shake") is performed. This involves randomly displacing all particles and testing for overlaps to overcome local jamming, particularly in clustered configurations.
Termination: The cycle repeats until the target volume fraction is achieved.

Computational Efficiency
A critical component of the methodology is the implementation of a cell-based neighbor search. By subdividing the RVE into cells slightly larger than the particle diameter, the algorithm limits overlap checks to particles within the current cell and its immediate neighbors (a $3 \times 3$ block in 2D). This reduces the computational complexity of collision detection from quadratic to near-linear ( $O(N)$ ) with respect to the number of particles. With additional optimizations to particle access patterns to mitigate cache misses, the generation time scales approximately as $t \propto N^{1.15}$ , enabling simulations with up to $10^7$ particles in both two and three dimensions.

Key Results

Scalability: The optimized SRM algorithm successfully generated configurations with up to $10^7$ particles. For low to intermediate volume fractions, the method maintains high efficiency.
Non-Equilibrium Control: By tuning the swelling rate ( $c_w$ $c_{w}$ ) and migration rate ( $c_m$ $c_{m}$ ), the algorithm can produce a spectrum of microstructures ranging from equilibrium-like distributions to strongly clustered arrangements.
- Comparison with MD: Unlike event-driven molecular dynamics (MD), which naturally evolves toward equilibrium, SRM generates configurations with significantly broader local volume fraction (LVF) distributions and stronger clustering signatures. MD-generated "clustered" states remain closer to equilibrium than those produced by SRM.
- Volume Fraction Limits: In 2D, the method can exceed the classical Random Close Packing (RCP) limit (achieving ~87% for disks). In 3D, the practical limit is approximately 60% for spheres; beyond this, random migration becomes inefficient, and event-driven MD is preferred.
Extension to Non-Spherical Inclusions: The algorithm was successfully applied to thin circular platelets (spherodisks). By adjusting parameters, the authors generated:
- "House-of-Cards" (HoC) networks: Highly interconnected structures formed by high swelling and low migration rates.
- Quasi-nematic domains: Metastable, partially ordered regions formed by relaxing HoC structures with high migration rates.
Structure-Property Impact: Percolation analysis of the platelet microstructures revealed that HoC configurations exhibit a critical percolation distance roughly an order of magnitude lower than relaxed, quasi-nematic structures. This demonstrates that microstructural arrangement significantly influences transport properties.

Significance and Claims
The paper positions the SRM algorithm as a powerful, flexible tool for the computational modeling of composite materials. Its primary significance lies in its ability to generate realistic, diverse, and computationally accessible particle configurations that reflect the statistical variability and non-equilibrium features observed in practical manufacturing.

The authors emphasize that SRM is a purely geometrical method; particle rearrangements are not governed by physical forces or interparticle potentials. Consequently, while SRM is ideal for generating statistically representative initial microstructures for homogenization studies, uncertainty quantification, and virtual testing, it does not reproduce the true physical evolution of particulate systems or study phase transitions. The method's value is in providing a controlled means to vary microstructural disorder and mimic manufacturing-induced anisotropy, thereby enabling more accurate structure-property relationship analyses in heterogeneous media.

1. The "Swelling Balloon" Game

2. The "Smart Neighborhood" Trick

3. What Can It Do? (The Results)

Summary

More like this