Emergent Transfer of a Physics Foundation Model from Simulation to Laboratory Turbulence
Abstract
Whether physics foundation models can be usefully deployed on laboratory experiments remains an open question for scientific machine learning (ML). We test this question on the Rayleigh-Taylor instability (RTI), a ubiquitous and demanding fluid instability seen from tabletop flows to supernova explosions, in which small perturbations at a density interface grow into chaotic, multiscale mixing as a lighter fluid accelerates into a heavier one. Standard ML models struggle with RTI, and despite over a century of theoretical, numerical, and experimental work, it carries an unresolved discrepancy between simulation and experiment: the late-time mixing growth rate, α, measured in most laboratory experiments (∼ 0.06-0.07), is roughly three times the value from idealized direct numerical simulations (DNS, ∼ 0.02). The gap's origin remains debated. These properties make RTI a stringent test for a question that matters well beyond RTI: can foundation models trained only on simulations generalise to sparse, messy, and noisy laboratory settings? We finetune Walrus, a foundation model for continuum dynamics, on three or fewer DNS realizations and recover key RTI physics over long rollouts. Applied zero-shot to sliding-barrier laboratory data, the finetuned model leaves the DNS-like regime and enters the observed growth band, having never seen a single experimental sample. These results provide independent, data-driven evidence that initial conditions play a crucial role in the longstanding sim-experiment gap in α. The model also generalises zero-shot to stable stratification, a buoyancy regime absent from training, correctly slowing mixing-layer growth. Together, our results show that foundation models can generalise well beyond their training data, predicting laboratory behavior and unseen physical regimes, opening new ways to probe longstanding simulation-experiment gaps.
Related
- Relative dispersion and eddy diffusivity in laboratory experiments of β-plane turbulence
- Wave-optics numerical simulation of atmospheric turbulence in 10,000-km-scale satellite–ground optical links with spatially adaptive phase screens
- A generative foundation model for 2D magnetized plasma turbulence
- Resolving Galactic and Cluster Dynamics Without Dark Matter: Tsallis Entropy as the Unique Foundation of Emergent Gravity
- Beyond capillarity: extended interfacial energy transfer in bubble-laden turbulence
BibTeX
@misc{mukhopadhyay2026emergent,
title = {{Emergent Transfer of a Physics Foundation Model from Simulation to Laboratory Turbulence}},
author = {Payel Mukhopadhyay and Stefan S. Nixon and Romain Watteaux and Michael McCabe and Alberto Bietti and Kyunghyun Cho and Cristiana Diaconu and Irina Espejo Morales and David Fouhey and Siavash Golkar and Tom Hehir and Shirley Ho and Jake Kovalic and Geraud Krawezik and Francois Lanusse and Tanya Marwah and Rudy Morel and Mariel Pettee and Helen Qu and Jeff Shen and Hadi Sotoudeh and Stuart B. Dalziel and Miles Cranmer},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2606.01470},
eprint = {2606.01470},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2606.01470},
}
RIS
TY - UNPB
TI - Emergent Transfer of a Physics Foundation Model from Simulation to Laboratory Turbulence
AU - Payel Mukhopadhyay
AU - Stefan S. Nixon
AU - Romain Watteaux
AU - Michael McCabe
AU - Alberto Bietti
AU - Kyunghyun Cho
AU - Cristiana Diaconu
AU - Irina Espejo Morales
AU - David Fouhey
AU - Siavash Golkar
AU - Tom Hehir
AU - Shirley Ho
AU - Jake Kovalic
AU - Geraud Krawezik
AU - Francois Lanusse
AU - Tanya Marwah
AU - Rudy Morel
AU - Mariel Pettee
AU - Helen Qu
AU - Jeff Shen
AU - Hadi Sotoudeh
AU - Stuart B. Dalziel
AU - Miles Cranmer
PY - 2026
DA - 2026/11/04
PB - arXiv
DO - 10.48550/arxiv.2606.01470
UR - https://arxiv.org/abs/2606.01470
AB - Whether physics foundation models can be usefully deployed on laboratory experiments remains an open question for scientific machine learning (ML). We test this question on the Rayleigh-Taylor instability (RTI), a ubiquitous and demanding fluid instability seen from tabletop flows to supernova explosions, in which small perturbations at a density interface grow into chaotic, multiscale mixing as a lighter fluid accelerates into a heavier one. Standard ML models struggle with RTI, and despite over a century of theoretical, numerical, and experimental work, it carries an unresolved discrepancy between simulation and experiment: the late-time mixing growth rate, α, measured in most laboratory experiments (∼ 0.06-0.07), is roughly three times the value from idealized direct numerical simulations (DNS, ∼ 0.02). The gap's origin remains debated. These properties make RTI a stringent test for a question that matters well beyond RTI: can foundation models trained only on simulations generalise to sparse, messy, and noisy laboratory settings? We finetune Walrus, a foundation model for continuum dynamics, on three or fewer DNS realizations and recover key RTI physics over long rollouts. Applied zero-shot to sliding-barrier laboratory data, the finetuned model leaves the DNS-like regime and enters the observed growth band, having never seen a single experimental sample. These results provide independent, data-driven evidence that initial conditions play a crucial role in the longstanding sim-experiment gap in α. The model also generalises zero-shot to stable stratification, a buoyancy regime absent from training, correctly slowing mixing-layer growth. Together, our results show that foundation models can generalise well beyond their training data, predicting laboratory behavior and unseen physical regimes, opening new ways to probe longstanding simulation-experiment gaps.
ER -