Lu Lu

Lu Lu's research pioneers a paradigm shift in Scientific Machine Learning (SciML). By combining physics with deep learning, he co-developed "physics-informed machine learning" (PIML), a method that addresses the efficiency bottlenecks of traditional physics-driven scientific computing and the lack of generalizability and interpretability in purely data-driven models.

At the core of his work is the paradigm of "operator learning," which uses neural networks to learn nonlinear mappings between infinite-dimensional function spaces. The DeepONet framework he developed can solve complex partial differential equations (PDEs) orders of magnitude faster than traditional numerical methods. In recognition of this work's potential, he received a U.S. Department of Energy (DOE) Early Career Award in 2022 to support his continued research on physics-informed neural operators.

Between 2023 and 2025, he further expanded the theory, algorithms and applications of operator learning. To address practical data limitations, he created a one-shot learning method that requires only one PDE data and developed federated physics-informed machine learning to handle distributed data. He developed the Fourier-enhanced DeepONet, significantly improving model accuracy, efficiency, and generalization in geophysical applications like geological carbon sequestration and full waveform inversion. He published a scalable framework for learning geometry-dependent PDE solution operators in predicting electrical signal propagation on patient-specific hearts. He also developed methods for efficiently designing channel topological structures of reactors or separators in chemical processes, determining mechanical parameters of diseased heart valve tissue, improving hemodynamic analysis in aortic dissection and aortic aneurysms, and rapidly predicting steady-state solutions in astrophysical disk-planet systems, demonstrating his work's broad applicability. More recently, he leveraged quantum computing to accelerate DeepONet and developed a novel framework for physics-informed generative modeling over function spaces.

Looking to the future, his work continues to accelerate interdisciplinary scientific discovery by creating efficient simulation tools that fuse physics, artificial intelligence, and quantum computing.