Physics-Informed AI

Simulation tools for studying the behaviors of granular materials — sand, beads, grains, and powders — built with physics-based modeling and AI-assisted development.


About this site

Granular materials are collections of discrete solid particles that together behave unlike ordinary solids, liquids, or gases: they jam, flow, compact under tapping, and remember how they were prepared. This site hosts an interactive Granular Mechanics Lab — a discrete-element method (DEM) simulator that runs entirely in your web browser, with no installation and no internet dependency once loaded.

You can pour elastic spheres into boxes and cylinders, compress them with pistons, tap them to study compaction, bend the container walls, and measure the statistical mechanics of the packing: void distributions v(r), Voronoi specific volumes, and the full Edwards-ensemble analysis — free-volume distributions P(v), density of states, compactivity, partition function, and entropy — following Chang, Chang & Chao, Phys. Rev. E 114, 015410 (2026).


The labs

2D

Granular Lab — 2D

Disks in side-view containers. Fast and clear: pistons, bendable sidebars, tapping & Edwards statistics, v(r) void distribution. Handles thousands of particles smoothly.

Open the 2D lab →
3D

Granular Lab — 3D

True spheres in 3D boxes and cylinders with an orbiting camera and cutaway view. Elastic balls, bendable walls, pistons, exact 3D Voronoi statistics, built-in lab notes and test report.

Open the 3D lab →
THE SCIENCE

Why do granular packings fluctuate the way they do?

Key findings, significance, and applications of Chang, Chang & Chao, Phys. Rev. E 114, 015410 (2026): the intrinsic density of states of jammed packings is a Gamma distribution in free volume — the physics the labs on this site let you explore hands-on.

Read the paper summary →

What you can do

Everything is SI-united, documented in-page (see the Lab notes button inside each lab), and validated against analytic results — the 3D lab includes its full test report. A companion Python engine with a parallel C core is available for large offline runs with the same physics.