Accelerating the discovery of materials

The government-sponsored Materials Genome Initiative demonstrates the importance of accelerating the discovery of materials, the impact of which would be felt in all technology disciplines. The high throughput approach to materials discovery takes advantage of high-volume computing and advances in machine learning to increase the accuracy and efficiency of material prediction. The current bottleneck to the high-throughput approach is the computation of large, theoretical databases of materials, the limiting factor being the slow convergence of a difficult numerical integral. We explain the hidden complexities of this unusual numerical problem and our approach to increasing its rate of convergence. Our tests on realistic 2D models show a ten-fold increase in computational efficiency over the traditional approach.