RII Track-4: Big Data and Massive Computation Approaches to Non-Equilibrium Dynamics of Strongly Correlated Materials

点击翻译按钮获取中文摘要

Non-technical DescriptionIn some materials, the motion of elections through the material structure can be highly correlated, such that the electrons behave as cars move in heavy traffic; they cannot maneuver freely and their motions are strongly influenced by others. Materials that exhibit electron correlations also exhibit intriguing properties, such as metal-insulator transitions and unconventional superconductivity. However, whether electron correlation is the sole cause of these observed effects has perplexed scientists for decades. Overcoming this gap in knowledge could open up revolutionary opportunities for novel transistor and ultrafast device applications. In this project, the PI will use the supercomputing capabilities at Oak Ridge National Laboratory to tackle this challenging problem. Advanced simulations will be performed using tens of thousands of computing processors to model the behavior of electron correlation systems at the atomic scale. The research will generate more than 100 terabytes of data, requiring that Big Data techniques be employed to effectively process and analyze the huge data volume. The project includes efforts to broaden participation in the next-generation scientific computing workforce. The research topics address several of the 10 Big Ideas for Future NSF Investments and the Grand Challenges in Basic Energy Sciences, thereby also having potential impacts on U.S. science leadership and an energy-sustainable future.Technical DescriptionThe project's focus is on modeling quantum many-body phenomena driven away from equilibrium, with the goal of understanding correlated electrons' non-equilibrium behaviors revealed by time-domain spectroscopies at ultra-short time scales. Matrix diagonalization over 100 billion basis states will be tackled by matrix-free and dataflow computing. Equilibrium and non-equilibrium wavefunction-based quantum impurity solvers also will be developed. Non-linear time series regression will be further implemented to alleviate the computational cost for time-evolution calculations. The resulting codes will be employed to simulate ultrafast photon-based spectroscopies on vanadium dioxide (VO2), using effective single-band Hubbard model and multi-orbital Hamiltonian from Wannier projection. The role of structure transition will be addressed by restricted phonon calculations. These simulations could significantly advance the understanding of non-equilibrium phenomena and photo-induced phase transitions in VO2 and other strongly correlated transition-metal oxides. Open-source softwares also will be made freely available to the public for parallel cloud computing to further benefit the scientific community for numerical studies of non-equilibrium many-body problems.