rayleigh-plateau3d¶
英語 | 日本語
Rayleigh-Plateau: a liquid column breaks up because surface tension pays for it -- showcase case B10 (docs/design/showcase-cases-design.md). --------------------------------------------------------------------------- A cylinder of liquid at rest is UNSTABLE to long-wavelength varicose perturbations: pinching it in one place and bulging it in another lowers the total surface area whenever the wavelength exceeds the circumference. The classical statement is
Geometry¶

Animation¶
Fields¶
この計算が最後の出力ステップで書き出した場をそのまま描いたもの。3D ケースは中央断面。

Build & run¶
cmake -S native -B native/build && cmake --build native/build -j
./native/build/orheo-rayleigh-plateau3d
rayleigh_plateau3d <config-file> (KEY = VALUE lines)
このページの図を作った計算のコスト(Miyabi-G Grace CPU (72 cores, -stdpar=multicore)): 33.8 s wall。
Validation¶
| Tier | T1 — 自動ゲートあり |
| Gate | ✅ 直接ゲート |
Source & run details¶
| Item | Value |
|---|---|
| Target | orheo-rayleigh-plateau3d |
| Source | native/benchmarks/rayleigh_plateau3d.cpp |
| Lattice | — |
| GPU twin | no |
| Modules | two_phase.hpp |
| Gallery run | 2026-07-28 @ c9b0572c |
| Args | mode=single R0=12 kR=0.7 rho_l=0.1 steps=20000 |
| Wall time | 33.79 s |
| CPU time | 2414.85 s |
| Device | Miyabi-G Grace CPU (72 cores, -stdpar=multicore) |
Computation log¶
この計算の実際の出力(先頭と末尾)。
=== rayleigh_plateau3d [single] 108 x 72 x 72 = 0.56 M cells
R0=12.0 W=4.0 (W/R0=33% diffuse) rho 1.000/0.100 nu=0.0100 sigma_nominal=0.0100 t_cap=416 steps
Oh = mu/sqrt(rho sigma R) = 0.0289 (viscous correction to the inviscid rate is a few % at this Oh -- quoted, not removed)
kR=0.700 -> Lx = 2 pi R0 / kR = 108 cells (ONE mode admitted by periodicity; no windowing, no forcing)
step 0 a=0.2372 (window [0.480, 2.400]) r in [11.898, 12.372]
step 2000 a=0.5463 (window [0.480, 2.400]) r in [11.579, 12.672]
step 4000 a=2.1087 (window [0.480, 2.400]) r in [9.937, 14.154]
fit over 24 samples, t in [1800, 4100]
growth rate omega_meas = 6.743220e-04 [VERIFY]omega_Rayleigh = 8.259159e-04 (-18.4%)
fit rms 3.673e-03 phase mass drift 2.075e-13
STABILITY: kR = 0.700 is UNSTABLE -- a column is unstable iff lambda > 2 pi R (kR < 1), which is surface-area bookkeeping and needs no table
RESULT mode=single R0=12.0 kR=0.7000 sigma_used=1.000000000e-02 omega=6.743220111e-04 omega_rayleigh=8.259159210e-04 fit_rms=3.673e-03 npts=24 mass_drift=2.075e-13 t_end=4200