Algorithms For Multiphysics Models In The Post-Moore's Law Era

The performance of supercomputers is no longer increasing exponentially, which presents a fundamental challenge to computational scientists. One cannot simply wait for the next computational platform’s arrival in order to tackle larger and more complex multiphysics problems using traditional approaches - instead, we have to develop fundamentally new algorithms and approaches to solving these challenges in each physics implementation and, in particular, multi-physics implementations. The goal of this workshop is to bring together a multidisciplinary audience of domain scientists, computational scientists, applied mathematicians, computer scientists, and the developers of next-generation computing platforms to share recent advances as well as brainstorm ideas about potential new paths forward.

This two-week workshop will be held in Los Alamos, New Mexico (at the SALA Los Alamos Event Center) during the weeks of June 2nd and 9th, 2025. The overarching theme of the first week will be “Multiphysics modeling beyond exascale” and the theme of the second week will be “Scale-bridging in models of physical systems,” though there will be cross-cutting topics (such as scientific machine learning) that will span the entire workshop.  

Topics will include (but are not limited to):

• Scientific machine learning as a tool for simulation acceleration and scale-bridging

• Computational platforms beyond exascale

• Co-design of algorithms and applications

• Novel algorithms for multi-scale and multiphysics models

• Randomized algorithms

• Verification, validation, and uncertainty quantification at scale

• Low-rank tensor decomposition

• Model reduction methods

• Linear and nonlear surrogates for subgrid models and closures

Meeting format: The format of the workshop will include talks each morning with time for discussion and collaboration in the afternoon, with panel discussions and poster sessions later in the afternoon. There will also be organized social activities during both weeks as well as over the weekend, driven by participants' interests.

Meeting goals: 

• Share recent advances in multiscale and multiphysics-modeling, with a focus on large-scale modeling in a broad range of disciplines

• Discuss how tools such as scientific machine learning might be most effectively coupled with large-scale computational models

• Discuss possible post-exascale directions for hardware and algorithms

• Identify areas for future resource investment, which will be communicated to funding agencies

Organizing committee:

• Brian O'Shea (Michigan State/LANL; committee chair)

• Chris Fryer (LANL)

• Timothy Germann (LANL)

• Abigail Hunter (LANL)

• Daniel Livescu (LANL)

• Qi Tang (Georgia Tech)

• Timothy Wildey (Sandia)

• Qian Yang (UConn)