Rod Rave — watch hard rectangles spontaneously align

Controls

aspect ratio 6k = l / w, w = 1

target density 42%cells filled

simulation speed

slow

normal

fast

turbo

horizontal

vertical

The real result

A 2026 preprint by Qidong He (arXiv:2607.10510) proves that l×w hard rectangles on the square lattice, restricted to two allowed orientations, must form a nematic phase — most of them pointing the same way — once the aspect ratio k = l/w is large enough.

The proof uses a two-scale cluster expansion (building on Disertori–Giuliani's work for 2D hard rods and Disertori–Giuliani–Jauslin's for 3D hard plates), tracking every constant explicitly instead of leaving them implicit — the first fully rigorous quantitative bound of its kind.

A non-optimized evaluation of the resulting conditions gives k ≥ 10⁷². Numerical simulations, meanwhile, place the actual onset at k_min = 7. The theorem isn't wrong — cluster expansions have to survive worst-case bookkeeping at every scale — it's just staggeringly conservative. Drag the slider below and watch the real onset happen, live.

k = 6 · rod is 1×6

order S

0.00

rods

density

sweeps

spontaneous symmetry breaking, live

Alignment achieved at k ≈ 8

You just watched hard rectangles spontaneously align at an aspect ratio of 8. The theorem only guarantees this happens once k ≥ 10⁷². You didn't need anywhere near that.

click a rod to kick it out · space = pause · ←/→ = aspect ratio · r = reset

The gap between proof and reality

k≈7 numerical onset

k=6

⚡ axis intentionally snapped here — what follows is not remotely to scale ⚡

10⁷²

the theorem's rigorous, non-optimized aspect-ratio bound — the value the proof can actually guarantee ordering by.

for scale (not from the paper): roughly 10¹⁸–10²⁰ grains of sand on Earth
for scale (not from the paper): roughly 10⁸⁰ atoms in the observable universe