Exploring the promise of diffusion Monte Carlo for Materials Research

Event Sponsor: 
Materials Science Seminar
Start Date: 
Oct 27 2015 - 2:00pm
Building 223/Room S105
Argonne National Laboratory
Luke Shulenburger
Speaker(s) Title: 
Sandia National Laboratories
Olle Heinonen, Anouar Benali

Electronic structure calculations based on Density Functional Theory (DFT) play an increasingly important role in the development and characterization of materials. However, efforts to calculate properties of materials of interest commonly expose the accuracy limitations of current DFT approximations. This situation has led to a renewed interest in beyond-­‐DFT approximations. Since Ceperley and Alder’s pioneering calculations on the homogeneous electron gas provided the benchmarks that underpin most current DFT approximations, diffusion Monte Carlo (DMC) has been recognized as one of  the most accurate techniques for electronic structure calculations. Unfortunately, despite its modest computational complexity, the large expense of ab initio calculations rendered the method impractical for all but benchmark and model calculations. Within the last decade this situation has changed. The availability of capability class computers and well-­‐designed algorithms has  allowed DMC to become feasible for general application. Because the resources required are still quite large, it is essential to understand the size of the approximations in DMC calculations of real materials and how to minimize them. I will present a series of benchmarks showing the current state of the art in this promising method1 as well as paths towards improving its accuracy.2 I will also present several recent applications that take advantage of the unique strengths of DMC, including calculating the properties of two dimensional black phosphorus,3 determining the interaction of large closed shell molecules,4 and probing the magnetic state of strongly correlated transition metal oxides. I will close by comparing DMC to other modern beyond-­‐DFT methods in the context of materials research.
1Shulenburger and Mattsson, PRB 88, 245117 (2013)
2Shulenburger, Mattsson and Desjarlais, arXiv:1501.03850
3Shulenburger, Baczewski, Zhu, Guan and Tomanek, arXiv:1508.04788
4Benali, Shulenburger, Romero, Kim and von Lilienfeld, JCTC 10, 3417 (2014)
*Sandia National Laboratories is a multi-­‐program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE­‐AC04-­94AL85000.