This project will advance layered materials genome (LMG). Functional layered materials (LM) will dominate materials science in this century. The attractiveness of LMs lies not only in their outstanding electronic, optical, magnetic and chemical properties, but also in the possibility of tuning these properties in desired ways by building van der Waal (vdW) heterostructures composed of unlimited combinations of atomically thin layers. We will perform 10 5 - atom nonadiabatic quantum molecular dynamics (NAQMD) and 10 10 - atom reactive molecular dynamics (RMD) simulations on Aurora for computational synthesis and characterization of LMs. The unprecedented simulations will (1) guide the synthesis of stacked LMs by chemical vapor deposition (CVD), exfoliation and intercalation, and (2) discover function - property - structure relationships in LMs with a special focus on far - from - equilibrium electron ic processes .
Our metascalable (or “design once, scale on new architectures”) simulation approach achieves portable performance on current and future computing platforms based on a novel divide - conquer - “recombine” (DCR) algorithmic framework for (1) l ean divide - and - conquer density functional theory (LDC - DFT) for NAQMD simulations with minimal O (N) prefactor, and (2) extended Lagrangian RMD (XRMD) to eliminate the speed - limiting charge iterations in RMD simulations . Key to metascalability is global - loca l separation achieved by our globally scalable /reproducible and locally fast (G SL F) solvers based on (1) a new scalable and reproducible global summation method , and (2) fast shift - collapse (SC) computation of local n - tuples .
Our codes are scalable beyond petaflop/s . We have performed the largest ever (1 billion atoms) RMD and (10 4 atoms) quantum molecular dynamics (QMD) simulations on 786,432 IBM Blue Gene/Q cores at Argonne Leadership Computing Facility (ALCF). Our 39.8 trillion electronic degrees - of - freedom QMD and 68 billion - atom RMD benchmarks have achieved parallel efficiency exceeding 0.98 and 51% of the theoretical floating - point performance on 786,432 Blue Gene/Q cores. P erformance - portability of our simulation algorithms ha s been verified on general - purpose graphics processing units (GPGPUs) and Intel Xeon Phi.