MolSSI Community Highlight: Dr. Danny Perez, Los Alamos National Laboratory
MolSSI Community Highlights spotlight exceptional examples of research and education enabled by MolSSI software and educational resources.
Dr. Danny Perez (Scientist IV in the Theory Division at Los Alamos National Laboratory) and his group focus on the development, implementation, and applications of long-timescale methods for atomistic simulation of materials through their work with the Exascale Atomistic Capability for Accuracy, Length, and Time (EXAALT) project. In this post, Dr. Perez tells us how he is using the MolSSI Driver Interface Project (MDI) to unlock access to many different DFT software packages to orchestrate the execution of very large numbers of calculations. Dr. Perez’s work has applications in the study of materials in extreme conditions and the generation of transferable datasets to parameterize machine learning potentials for materials.
MolSSI:
Hi Dr. Perez, we appreciate you taking the time to talk to us today! Your research field seems intriguing. Could you provide a brief introduction to your research area and share what your team is currently focusing on?
Dr. Perez:
My research focuses on the development, implementation, and applications of long-timescale methods for atomistic simulation of materials. I am especially interested in materials in extreme conditions, such as materials for fission and fusion energy, or particle accelerators. Recently, I became interested in the generation of transferable datasets to parameterize machine learning potentials for materials. I am also interested in applications of petascale and exascale computing in materials science.
MolSSI:
Can you share how MolSSI software tools or resources have been utilized to accomplish specific research objectives in your projects?
Dr. Perez:
I have collaborated with the MolSSI team as part of my role as PI in the Exascale Atomistic Capability for Accuracy, Length, and Time (EXAALT) project, which is one of the application projects supported by the DOE’s Exascale Computing Project. One of our objectives is to significantly decrease the time required to obtain reliable interatomic potentials for materials by integrating all steps, generation of training configurations, characterization with density functional theory, training of the potential, and validation. Our objective is to design scalable workflows to integrate and automate all of these steps.
MolSSI:
For this particular project, which MolSSI software tools or resources did your team employ, and what aspects or features did you find most beneficial?
Dr. Perez:
We are designing our workflows using MolSSI Driver Interface (MDI). We are especially interested in the plugin mode of MDI which enables us to control large numbers of instances of DFT and/or MD codes at scale using MPI without having to resort to file or socket-based mechanisms.
MolSSI:
How did MolSSI resources influence your software development efforts, and what led you to choose MolSSI resources for this purpose?
Dr. Perez:
In our case, we wanted to be able to orchestrate the execution of very large numbers of DFT calculations as part of an integrated workflow tool that can run at scale. The solution we converged to was to support engine codes that can be called through an API that can receive existing MPI communicators. Since such APIs are still not common or standardized, we faced the prospect of having to write specific code to support multiple engines. We quickly learned that the MolSSI MDI team aimed at standardizing such APIs, which would greatly simplify our work. After engaging with the MDI team, they endeavored to implement a new plugin mode which perfectly met our requirements in terms of integration with MPI codes at scale. This has led to a tremendous simplification of our code, as we only have to support MDI to unlock a whole ecosystem, instead of having to write engine-specific code.
MolSSI:
It’s fantastic to see the role MolSSI’s software has played in your project’s development! Could you share any scientific findings or results that your workflows have facilitated?
Dr. Perez:
We are still developing our workflow, but we hope to demonstrate the end-to-end generation of reliable interatomic potentials in under a day.
MolSSI:
Those are impressive goals, and it’s wonderful to see how MolSSI resources have contributed to your group’s success. Thank you for sharing your experiences and insights with us!
Although there are no publications yet that incorporate MolSSI software tools or resources in Dr. Perez’s work, we look forward to seeing the advancements his group will make in the near future.