Statistical approach of random packing of polydisperse hyperspheres: with application to binomial and lognormal size distributions
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Condensed Matter. The day in condensed matter: materials, phases and the measurements under them.
Kai: I'm Kai.
Mira: And I'm Mira, here with Lev.
Kai: We're looking at the paper: "Statistical approach of random packing of polydisperse similar hyperspheres".
Mira: Here's the gist of the paper. Statistical approach of random packing of polydisperse similar hyperspheres: with application to binomial and lognormal size distributions is a study deriving a closed-form expression for the packing fraction of multi-sized particles using a statistical approach based on volume contraction, which is significant because it provides an explicit formula governed by the standard deviation, Euclidean dimension, and monosized packing fraction that matches empirical data across various conditions Derivation of the Closed-Form Expression The paper first derives a closed form expression for a mix consisting of multiple discrete particles in D-dimensional Euclidean space RD with small size differences A recently developed analytical model for binary packings with small size ratio u is used, which has its foundations in the excluded volume model by Onsager, here applying it to the statistically probable particle pairs in the multi-sized mix The new model is subsequently applied to discretely sized particles with a binomial number distribution and a constant size ratio u between each subsequent particle size By taking the limit of u → 1 and an infinite number of particles sizes, the binomial particle distribution is transformed into a continuous lognormal particle size distribution The resulting expression for the packing fraction can be solved in closed form, resulting in an original expression for the packing fraction of mixtures of hyperparticles with a lognormal distribution This explicit expression is governed by the standard deviation, the Euclidean dimension D, and the monosized packing fraction of the considered particle shape Modeling Binary Mixtures The model for binary particles with small size disparity is recapitulated in Section 2 The mean excluded volume of randomly mixing the small and large particles follows from the statistically probable combinations of these particle pairs Based on integral geometry it appeared that the excluded volume that underlies Eqs.
Kai: (10) holds for arbitrary binary similar convex particles, so not for and a (sphero)cylinders (D = 3) only The contraction function i j v D omega, (,) appears in Eq.
Mira: (10), defined as
30, 32, 33: Generalization to Multicomponent Mixes The binary mix model is extended to multicomponent mixes by considering the statistical frequencies of all particle pairs in the assembly The resulting polydisperse packing fraction n i in i = 1/n n i in i j jn ij i = 1 i = 1 j = i + 1 f X is given by Eq.
Kai: (25) This general closed-form equation provides the packing fraction as a function of standard deviation, the single-sized packing fraction of the particle shape considered, and the Euclidean dimension Application to Binomial and Lognormal Distributions The paper studies the binomial distribution of n particle sizes and probability p with a constant size ratio u between each subsequent size Adding more monosized particle sizes, all adjacent sizes having the same size ratio u, and letting u → 1, following the central limit theorem the binomial distribution turns into a continuous lognormal distribution The presented analytical model is valid for particles which size ratio is about 1/9 ≤ Ω ≤ 9 Comparison with Reported Data The model predictions are compared with a large collection of experimental data, simulations and theoretical approaches reported before For weak polydispersity, the scaled packing fractions as given by Eq.
Paper summary: Kai: To wrap up this paper, we're looking at "Statistical approach of random packing of polydisperse similar hyperspheres: with application to binomial and lognormal size distributions." The authors developed a closed-form expression for the packing fraction of multi-sized particles using a statistical approach based on volume contraction.
Mira: They successfully derived an analytical expression for binary packings with small size disparity, rooted in Onsager's excluded volume model. Then they used that to move to discrete binomial distributions and then took the limit to get the continuous lognormal distribution.
Lev: The main result is this final expression for the lognormal packing fraction, which matches their analytical model for small standard deviation but also has an asserted asymptote for large standard deviation, and they validated it specifically for Euclidean dimensions D equals two and three.
Kai: So what does this mean in simpler terms? It means they found a way to calculate how much space particles take up when you have many different sizes, using just the standard deviation, the dimension D, and the monosized packing fraction.
Mira: It’s about providing an explicit formula that is governed by those three parameters, which matches their empirical data across various conditions. This is significant because it replaces complex simulations with a direct calculation for packing fractions in these specific distributions.
Lev: For someone building real hardware, this means you have a predictable way to estimate the density of packed states based on the statistical properties of your particle sizes without needing massive computational overhead just to check every combination.
Kai: It’s a geometric approach that successfully links integral geometry to particle packing in dimensions two and three. That connection between the shape and the statistics is what makes this work useful.
Conclusion: Kai: So we've seen how they built this model for packing particles of different sizes using volume contraction statistics.
Mira: Yeah, looking at that title, "Statistical approach of random packing of polydisperse similar hyperspheres," it sounds really technical, but the core idea is just figuring out how much space things take up when they aren't all the same size.
Lev: The authors are applying this statistical approach specifically to particle size distributions that are binomial and lognormal. That's a specific area of physics they’re tackling.
Kai: What does that mean for us? It means we can get a direct calculation for packing fractions when we have mixtures of particles, whether the sizes follow a simple binomial pattern or something more complex like lognormal.
Mira: The implication is that instead of running huge simulations every time we want to know how dense a system is, we can use this closed-form expression. It's governed by the standard deviation, the dimension you're in, and that monosized packing fraction for a single size.
Lev: From an engineering standpoint, it’s about reducing the overhead of those simulations. If you can get a reliable number from these parameters, it makes designing systems with varying particle sizes much faster to iterate on.
Kai: And they validated this stuff in dimensions two and three. That’s where we actually have experimental data for packing things like spheres and cubes.
Mira: They show that this formula works well for small standard deviations, but the way it behaves changes when the standard deviation gets really large. It shifts from being governed by the standard deviation to being governed by something else entirely at high densities.
Lev: That shift is important because it tells you where your assumptions need to change when you push those systems toward their limits.
Kai: So, they've given us a mathematical tool that connects the shape of the particle, its statistical spread, and the space it occupies in a dimension. But what happens when we move beyond just spheres and cubes? We'll take a look at some of those comparisons next.
Eindhoven University of Technology
cond-mat.soft
Submitted: 2026-11-03
Updated: 2026-11-14
Comments: 14 pages, 7 figures, 18 tables
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 74/100
The gist: Statistical approach of random packing of polydisperse similar hyperspheres: with application to binomial and lognormal size distributions is a study deriving a closed-form expression for the packing…
Key concepts
- Volume Contraction
- This statistical approach uses the concept of volume contraction to derive a closed-form expression for how particles pack together in space. It is based on analyzing the statistically probable combinations of particle pairs, which helps determine the excluded volume effect when mixing different sizes.
- Lognormal Distribution
- The paper transforms a binomial distribution of particle sizes into a continuous lognormal size distribution by taking the limit as the number of monosized particles increases. This transformation allows for an explicit analytical solution for calculating packing fractions in mixtures with this continuous size variation.
- Packing Fraction ($oldsymbol{n_i}$)
- The packing fraction ($n_i$) represents the proportion of space occupied by a specific size of particle within the mixture. The general closed-form equation derived allows researchers to calculate this for multicomponent mixes based on the standard deviation, dimension, and monosized packing fraction.
- Euclidean Dimension ($oldsymbol{D}$)
- The Euclidean dimension ($D$) refers to the number of spatial dimensions in which the particles are packed (e.g., 2 for a plane, 3 for typical space). The derived formula is generalized to be valid for arbitrary $D$, making it applicable beyond simple 2D or 3D sphere packing problems.
Terminology
Summary
Statistical approach of random packing of polydisperse similar hyperspheres: with application to binomial and lognormal size distributions is a study deriving a closed-form expression for the packing fraction of multi-sized particles using a statistical approach based on volume contraction, which is significant because it provides an explicit formula governed by the standard deviation, Euclidean dimension, and monosized packing fraction that matches empirical data across various conditions
Derivation of the Closed-Form Expression
The paper first derives a closed form expression for a mix consisting of multiple discrete particles in D-dimensional Euclidean space RD with small size differences A recently developed analytical model for binary packings with small size ratio u is used, which has its foundations in the excluded volume model by Onsager, here applying it to the statistically probable particle pairs in the multi-sized mix The new model is subsequently applied to discretely sized particles with a binomial number distribution and a constant size ratio u between each subsequent particle size By taking the limit of u → 1 and an infinite number of particles sizes, the binomial particle distribution is transformed into a continuous lognormal particle size distribution The resulting expression for the packing fraction can be solved in closed form, resulting in an original expression for the packing fraction of mixtures of hyperparticles with a lognormal distribution This explicit expression is governed by the standard deviation, the Euclidean dimension D, and the monosized packing fraction of the considered particle shape
Modeling Binary Mixtures
The model for binary particles with small size disparity is recapitulated in Section 2 The mean excluded volume of randomly mixing the small and large particles follows from the statistically probable combinations of these particle pairs Based on integral geometry it appeared that the excluded volume that underlies Eqs. (10) holds for arbitrary binary similar convex particles, so not for and a (sphero)cylinders (D = 3) only The contraction function i j v D omega, (,) appears in Eq. p. 2 (10), defined as [30, 32, 33]
Generalization to Multicomponent Mixes
The binary mix model is extended to multicomponent mixes by considering the statistical frequencies of all particle pairs in the assembly The resulting polydisperse packing fraction n i in i = 1/n n i in i j jn ij i = 1 i = 1 j = i + 1 f X is given by Eq. (25) This general closed-form equation provides the packing fraction as a function of standard deviation, the single-sized packing fraction of the particle shape considered, and the Euclidean dimension
Application to Binomial and Lognormal Distributions
The paper studies the binomial distribution of n particle sizes and probability p with a constant size ratio u between each subsequent size Adding more monosized particle sizes, all adjacent sizes having the same size ratio u, and letting u → 1, following the central limit theorem the binomial distribution turns into a continuous lognormal distribution The presented analytical model is valid for particles which size ratio is about 1/9 ≤ Ω ≤ 9
Comparison with Reported Data
The model predictions are compared with a large collection of experimental data, simulations and theoretical approaches reported before For weak polydispersity, the scaled packing fractions as given by Eq. (26) are used (Appendix), and compared with Eq. p. 7 (48) In Figure 5, scaled packing data up to much larger standard deviation Lnσg are shown The study reveals that for small Lnσg, the effect of polydispersity on packing fraction (and glass transition) is governed by DLnσg, whereas for large DLnσg, it is governed by Lnσg
Final Result and Validation
The final expression for the lognormal packing fraction is given by Eq. (51) This expression matches the analytical model for small Lnσg (Eq. p. 7 (47)), but is applicable to arbitrary space dimension D, and the asserted asymptote for large Lnσg, and is validated for Euclidean dimensions D = 2 and 3 The governing closed-form expression contain the monosized packing fraction f, the standard deviation Lnσg, and the Euclidean dimension D only
APPENDIX
Table A.1 shows Lognormal packing fraction η versus standard deviation logσg for OS 0. p. 93 as measured by Sohn and Moreland [13], and the computed λ using Eq. p. 9 (26) Table A. p. 1 p. 1 p. 1 p. 1 p. 1 p. 1 p. 1 p. 1 p. 1 p. 1 p. 1 p. 1 p. 12 shows Lognormal packing fraction η versus standard deviation logσg for MR 0. p. 95 as measured by Sohn and Moreland [13], and the computed λ using Eq. p. 9 (26) Table A.3 shows Lognormal packing fraction η versus standard deviation logσg for MR 1 as measured by Sohn and Moreland [13], and the computed λ using Eq. p. 9 (26) and an estimated f = 0. p. 9574 Table A. p. 14 shows Lognormal packing fraction η versus standard deviation logσg for MR 2 as reported by Sohn and Moreland [13], and the computed λ using Eq. p. 9 (27) Table A. p. 1 p. 1 p. 15 shows Lognormal packing fraction η versus standard deviation logσg for spheres as reported by Dexter and Tanner [14], and the computed λ using Eq. p. 6 (26) Table A.6 shows Lognormal packing fraction η versus standard deviation logσg for tapped spheres as reported by Suzuki et al. p. 9 [16], and the computed λ using Eq. p. 9 p. 9 p. 9 p. 9 (26) Table A.7 shows Lognormal packing fraction η versus standard deviation logσg for prodded spheres as reported by Suzuki et al. p. 9 [16], and the computed λ using Eq. (26) Table A.8 shows Lognormal packing fraction η versus standard deviation logσg for tapped crushed glass as reported by Suzuki et al. p. 9 [16], and the computed λ using Eq. (26) Table A.9 shows Lognormal packing fraction η versus standard deviation logσg for prodded crushed glass as reported by Suzuki et al. p. 9 [16], and the computed λ using Eq. (27) Table A.10 shows Lognormal packing fraction η versus standard deviation logσg for spheres as reported by Yang et al. p. 9 [17], and the computed λ using Eq. p. 9 p. 9 (26) Table A.11 shows Lognormal packing fraction η versus standard deviation logσg for spheres as reported by Yang et al. p. 9 [17], and the computed λ using Eq. (26) Table A.12 shows Lognormal packing fraction η versus standard deviation logσg for spheres as reported by He et al. p. 9 [18], and the computed λ using Eq. p. 9 (26) Table A.13 shows Lognormal packing fraction η versus standard deviation logσg for spheres as reported by Farr and Groot [19], and the computed λ using Eq. p. 9 (26) Table A.14 shows Lognormal packing fraction η of spheres at lowest possible packing density (“LT”) versus standard deviation logσg as reported by Baranau and Tallarek [22], and the computed λ using Eq. p. 6 (26) Table A.15 shows Lognormal packing fraction η of spheres at densest possible packing density (“RCP”) versus standard deviation logσg as reported by Baranau and Tallarek [22], and the computed λ using Eq. p. 6 (26) Table A.16 shows Lognormal packing fraction η of spheres at densest glass close packing limit (“GCP”) versus standard deviation logσg as reported by Baranau and Tallarek [22], and the computed λ using Eq. p. 6 (27) Table A.17 shows Lognormal packing fraction η of cubes versus standard deviation logσg as reported by Malmir et al. p. 9 [24], and the computed λ using Eq. p. 9 (26) Table A.18 shows Lognormal packing fraction η versus standard deviation logσg for disks in a plane as reported by Weeks [39], and the computed λ using Eq. p. 9
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Analytical Derivation of Packing Fraction for Polydisperse Systems: The improved system can derive a closed form expression for packing fraction as a function of standard deviation, Euclidean dimension D, and monosized packing fraction without needing adjustable parameters.
-
Handling Continuous Size Distributions: The system can model and solve the packing fraction analytically for continuous lognormal particle size distributions by taking the limit of discrete binomial distributions where the size ratio approaches unity.
-
Generalization to Multicomponent Mixtures: The system can extend models to multicomponent mixes by considering
the statistical frequencies of all particle pairs in the assembly,
allowing it to calculate packing fraction for systems with more than two particle sizes. -
Robust Asymptotic Behavior Prediction: The improved AI can predict the large standard deviation asymptote for packing fractions, which is stated as
not to depend on the Euclidean dimension
when large size differences are considered. -
Dimensional Independence Verification: The system can confirm that for binary and lognormal distributions,
binary and lognormal distributions feature the same characteristics with respect to the Euclidean dimension
for both small and large polydispersity.
Related papers
- Coarse-graining dense, deformable active particles
- Three-body corrections to the Brownian viscosity of hard spheres without hydrodynamic interactions
- Thin active nematohydrodynamic layers: asymptotic theories and instabilities
- Origins of suppressed self-diffusion of nanoscale constituents of a complex liquid
- Soft material mechanical property degradation at high strain rates
- Avalanche criticality emerges by thermal fluctuation in a quiescent glass