Post by Yifan Li

Our paper on fix pimd/langevin is out! I am very happy to contribute a comprehensive, efficient, and easy-to-use PIMD module to the LAMMPS ecosystem. In this work, we present an implementation of path integral molecular dynamics (PIMD) directly in LAMMPS, with support for commonly used features such as NVE, NVT, NpH, and NpT ensembles. A major goal of this project was to make PIMD fit naturally into the existing LAMMPS framework, so that users can simulate nuclear quantum effects (NQEs) within a highly efficient and widely used molecular dynamics package. Special thanks to Axel Gomez, a postdoctoral fellow here at Princeton, for his help with benchmarks, feedback, and manuscript preparation. Axel is an expert in simulating NQEs with PIMD, and this manuscript would not have been possible without his support. I also want to thank Yotam Feldman, Ofir Blumer, Jacob Higer, and Barak Hirshberg for contributing the fix pimd/langevin/bosonic module! Thank Weile Jia for providing valuable insights into profiling and performance optimization from the MPI side, which saved me a great deal of time during production simulations. I am grateful to our early users, including Chunyi Zhang, Kehan Cai, Li Fu, Shenzhen Xu, for running the code, helping with bug fixes, and providing valuable feedback. Thank Axel Kohlmeyer for his rigorous review and improving the code quality. I also want to thank my PhD advisor, Roberto Car, for his long-term support. If you want to run PIMD, give it a try! Our paper also serves as a practical tutorial: https://lnkd.in/eM2iq-9n See the official LAMMPS documentation for more information. Working on this project also left me with a broader question. I spent substantial effort learning the LAMMPS code base at a time when coding AI tools were far less developed. With today’s rapid progress in coding AI, do you feel it has become much easier to contribute to large software projects?