Chemical Computing Group Excellence Award for Student Research in Developing a Novel Computational Method

Posted on October 30, 2024

Graduate student Robert Lavroff (Alexandrova group) has been awarded the 2024 Chemical Computing Group Excellence Award for outstanding research in computational chemistry.

Lavroff is one of the five awardees who will receive a travel stipend to attend the 2025 ACS National Meeting in San Diego, along with a one-year license for the Molecular Operating Environment software.

The prestigious award is open twice a year to graduate students across North, South, and Central America who sign up to present their research at an upcoming ACS meeting either in Spring or Fall. Winners are selected based on the quality and significance of the research they will present.

As a fifth-year theoretical and computational chemistry Ph.D. student in Professor Anastassia Alexandrova’s group, Lavroff works on methods and applications for materials with strong electron-electron and electron-phonon interactions. At the ACS meeting, he will present his work titled “Aperiodic fragments in periodic solids: Achieving the thermodynamic limit of solid-state defect calculations with hundreds to thousands times speedup,” which is based on a novel computational method he and his collaborators have developed in the “Cryscor” quantum chemistry software. These collaborators are Dr. Denis Usvyat at Humboldt University of Berlin, Professor Lorenzo Maschio at the University of Torino, and Dr. Daniel Kats, Dr. Nikolay Bogdanov, and Professor Ali Alavi at the Max Planck Institute for Solid State Research.

Defects such as dopants and vacancies are crucial in catalysis and materials science, yet understanding their electronic structure can require very costly computations. Traditional computational methods construct periodic supercells with hundreds to thousands of atoms to meet the realistic condition of non-interacting defects. This approach is both computationally expensive and prone to unphysical errors if careful benchmarking is not done. Alternatively, to speed up and maintain the accuracy of computation, Rob and the collaborators begin with a pristine, small cell in the absence of defects and later embed a defective site in the periodic environment, referred to as an “aperiodic fragment” (Fig. 1. An example can be found in Science 101 below). The innovative method successfully approaches the expected carbon-fluorine bond breaking energy in a test fluorographane system using only tens of atoms in the calculation, making it hundreds of times faster than the conventional supercell method. The approach allows a variety of quantum chemistry methods and can compute excited states of defects as well.

Figure 1. Difference between conventional supercell method (left) vs. aperiodic fragment (right). (“HF” stands for Hartree-Fock, which is a standard, inexpensive quantum chemistry method often used as a starting point for more accurate, expensive calculations referred to as “post-HF” methods.)

“We’re excited to add more features and apply our method to other interesting defects in solids.” Lavroff said. “I’m grateful to my advisors, Anastassia and Denis, and to all my collaborators.”

A version of Lavroff’s work can be found online, and he recently presented it in a 15-minute talk (39:40) at the Virtual International Seminar on Theoretical Advancements.

Lavroff joins a list of previous members in Alexandrova group who have also won the CCG Excellence Award, including Zisheng Zhang (PhD 2024), PJ Robinson (UCLA undergraduate, Ph.D. from Columbia), Zhihao Cui (UCLA undergraduate, Ph.D. from Caltech), as well as Professor Alexandrova herself, who also won this Award at their times.

Science 101

Could you provide us an example to help us better understand your method?

Lavroff: In a conventional calculation of a graphitic single-atom catalyst (SAC) illustrated below, you’d need a graphene supercell that can house the 6 carbon sites that are replaced with the 4 nitrogens and the metal, plus any additional carbons needed to prevent these SAC sites from interacting from each other.

Figure 2. Graphene supercell with a 6-carbons-site defect: 4 carbons substituted with nitrogen and then the remaining vacancy being filled with some metal atom. (This supercell is tens of atoms, but is still likely much too small to prevent the defect from interacting with its repeated image.)

This ends up being at least tens, probably hundreds, of atoms for this 2D system, and for a 3D crystal, it would be even worse. In our approach instead, you can do a periodic calculation on graphene using the primitive graphene unit cell (which has only 2 atoms). And then once you have the periodic solution, you cut out the aperiodic fragment (which needs to be at least big enough to house the 6 carbon sites mentioned above), which is embedded in the solution of the periodic environment from your 2-atom graphene calculation, and then go from there.

Article by Zhuoying Lin, zylin@g.ucla.edu.