PSWF-LR Framework Breaks Long-Range Bottleneck in ML Interatomic Potentials, Accelerating Materials Discovery for Batteries and Catalysts

The arXiv preprint (abs/2606.06617) introduces PSWF-LR, leveraging prolate spheroidal wave functions for exponent-aware mollification and atom-grid spreading to handle arbitrary 1/r^p long-range terms with fewer Fourier modes than Ewald summation. This preprint, not yet peer-reviewed, reports benchmarks across ionic, polar, and interfacial systems showing threefold MD speedups and extended simulation scales beyond prior memory limits. Standard Ewald approaches in earlier MLIPs such as those in the 2022 J. Chem. Phys. work on reciprocal-space networks often demand dense grids that scale poorly for production runs; PSWF-LR treats the decay exponent as a physical prior, yielding measurable gains in energy and force accuracy. Related work in the 2023 Nature Communications paper on equivariant long-range potentials similarly identified dispersion bottlenecks in semiconductors yet lacked the compact PSWF representation. Limitations include reliance on synthetic benchmarks without explicit sample sizes disclosed and absence of real-world materials validation, leaving open questions on transferability to disordered systems. The advance directly supports faster screening of battery electrolytes and catalyst surfaces, areas where conventional MLIPs have lagged behind DFT accuracy at scale.