Skip to main content

Physics simulation: DLA

Diffusion-limited aggregation (DLA) is a model of how randomly walking particles stick together into branching clusters. It shows up in physics, materials science, and graphics.

Problem​

I needed a clear, repeatable DLA sim for coursework or demos: correct random-walk and sticking rules, but still fast enough to finish on ordinary hardware.

Solution​

Classic algorithm on a 2D lattice (or a continuous variant): seed a cluster, release walkers from a boundary, and freeze them on first contact. Render growth frame by frame or as a final aggregate.

Impact​

  • Makes nonlinear growth and fractal-looking structures easy to show in class or demos.
  • Gives a baseline for comparing parameters (stick probability, lattice vs off-lattice, noise).

Lessons learned​

  • Visualize early. Neighbor-check bugs show up immediately on screen.
  • Where new walkers spawn changes both runtime and cluster shape a lot.
  • For big clusters, move to sparse maps or distance fields before just adding more particles.

Tech stack​

  • Language and viz tools used for the assignment or demo build
  • Optional: numeric libraries for offline analysis of cluster stats