Hello, A yearly allocation for the LCRC cluster has been requested with the following updated information: Submitter/PI: John J. Low Project Name: Electric_Interface Division: MSD Project title: Electrified Interfaces from First-Principle Studies Associated funding: DOE BES-CEES EFRC, DOE BES-JCESR Other Systems: Science: Electrified Interface of solid and solid/liquid electrolytes in electrochemical system is an important component that of great interest to electric energy storage research at Argonne (i.e. BES-JCESR, BES-CEES-EFRC, DOE-EERE), and also among the scientific community. Within this wide area of research problems, the effect of external bias voltage on the interface at the atomistic scale is the basis of our fundamental understanding, design and application in this field. To include the external bias voltage in atomistic simulation based on first-principle study (DFT), the theoretical treatment with the application of a finite homogeneous electric field proposed by Nunes et. al. [1] is a reasonable and practical method based on DFT calculations for the fundamental studies. Within this renew proposal, the key parts of the proposed study will consists of two independent yet related topics: (I) the electrified interface of organic electrolyte/oxide, (II) electrified interface of organic electrolyte/CrSi2. For the case study of electrified interface of organic electrolyte/oxide, our interest will continue to focus on oxide solid-phase instability at external electric potential that exemplified by metal dissolution from manganese (Mn)-based oxide cathode surface that found in next generation Li-ion battery system [2]. To go in parallel with our experimental colleagues, the modeled system will conitue to focus on the spinel LiMn2O4 (LMO)/ GenII electrolyte (i.e. 1.0 LiPF6 in ethylene carbonate: ethyl methyl carbonate electrolyte with 3:7 ratio by weight) developed at ANL. In the coming FY2016, the primary study will continue be Mn desolvation, dissolution and diffusion at the cathode interface as we planned on FY2015, but will only focus on [100] LMO. For the case study of electrified interface of organic electrolyte/CrSi2, our interest is impired by the experimental work by Fister et. el. [3]. A layer of Li cations have been observed by Fister et al. at the interface between CrSi2 and a electrolyte contain LiPF6, ethylene carbonate and propylene carbonate at well define voltages. [3] We will use to a combination of quantum mechanical and force-field calculation to model Li cations at this electrified interface and compare them to experimental results. Project description: Experiments Planned: The project can be divided into two related problems: 1) Electrified Interfaces of spinel [100] LMO/electrolyte According to our knowledge, it is found that the surfaces of [100] LMO is susceptible to dissolution than [111] LMO surface. Therefore this surface appears to be a logical modeled system to focus on. To address the problems that related to our experimental colleagues findings, the simulation of Mn dissolvation, dissolution and diffusion into electrolyte interfaces will be performed with and without the water molecules (i.e. moisture that will cause side-effect) in the simulation. 2) Electrified Interfaces of CrSi2/electrolyte We will used the OPLS-AA forcefield and LAMMPS to generate a starting configuration for the bilayer at the CrSi2/electrolyte. Molecular dynamics is the most efficient way to preequilibrate, allowing each ion to attain proper coordination and a low energy distribution in the bilayer. The bilayer in carbonates is about 30Å wide. We plan to carry out calculations to span the voltages where the layer of Li cations were observed. The first pass will involve a molecular dynamics followed by geometry optimization with quantum mechanics. This should yield a qualitative correct description of the electrified CrSi2/electrolyte interface. If sufficient CPU time is available we will carry out ab initio molecular dynamics trajectories on the most interesting optimized geometries. Computational Methods and Software The density functional theory simulation code OPENMX, VASP and Quantum Espresso will be employed in most of the work. Visualization The public domain graphics packages VMD, VESTA, XCRYSDEN and PWGUI are convenient for depicting three dimensional atomic configurations, making movies of dynamical processes and setting up calculations. Scalability Issues VASP is well known to have scalability issues. Although it is very efficient code for models contains 100s of atoms, the parallel scalability for systems of this size is only 100s of cores. Although better parallel scalability can be obtained for larger models, the size scalablity becomes a major issue as O(N3) terms become a bottleneck in the simulation. The scalability of Quantum Espresso is similar to that of VASP. OpenMX is an O(N) code with good parallel scalability. A recent publication by Duy and Osaki [4] has demonstrated the excellent parallel scalability of OpenMX to over a 100K cores on the K computer. Calculation size The smallest cells for which the effects of interest can realistically be simulated for [100] LMO/electrolyte AIMD simulation will contain about four - five hundred atoms, with about 200 atom slab of LMO, and 200 - 300 atoms of electrolyte that represent ESM screening media in OPENMX. All numerical shortcuts (gamma-point only, soft pseudopotentials) will be exploited. We expect to run on 256 processors. Of the calculations proposed here, the AIMD simulations (item 1) will have the largest CPU-time requirements per unit MD-time step of 1 fs. The AIMD simulations performed in NVT-ensemble simulations may require about 1 processor-hr per time step, for a cell size of about 500 atoms (using perhaps 256 processors). Obtaining a reasonably good MD statistics calculation may require of the order of 200 k processor hrs. Thus for two separate sample simulation (i.e. with and without H2O molecule), the total 400k core-hour is needed in this problem. Our model includes a three-layer slab of CrSi2 about 15Å wide containing 108 atoms (Cr36O72), a 30Å layer of electrolyte containing 6 LiPF6 molecules and 60 ethylene carbonate molecules. We estimate a ten quantum mechanical geometry optimizations with different charges on the CrSi2 slab. Each geometry optimization will require 3,000 core-hours. Ten geometry optimizations will require 30,000 core hours. The voltages from these preliminary calculations will help us select three geometries for ab initio molecular dynamics (AIMD). A picosecond of AIMD will require 60,000 core-hours. Three trajectories at different levels of charging will require 180,000 core-hours. The voltages from these AIMD simulations will suggest an additional three AIMD simulations with different levels of charging to attain the desire voltage in the simulation. The total amount of time for the geometry optimization and AIMD simulations of CrSi2/Li6(PF6)6EC6 will require 400,000 core-hours. Thus, requirements for the entire project are thought to be of order 800,000 core-hours. References: 1. R. W. Nunes and X. Gonze, Phys. Rev. B 63, 155107 (2001). 2. J. Lu et. al. Nature Communications 5, 5693 (2014). 3. T. Fister et. al. J. Phys. Chem. C 116(2012)22341. 4. T. Ohwaki, M. Otani, T. Ikeshoji, T. Ozaki, J. Chem. Phys. 136, 134101 (2012); T.V.T. Duy, T. Osaki, Comp. Phys. Comm. 185, 777 (2014). Industry partnership: Project URL: Current FY Hours Used: undetermined amount New FY Requested allocation: 800000 Q1: 200000 Q2: 200000 Q3: 200000 Q4: 200000 Justification: We have been testing the beta version of VASP on GPUs. We expect that the GPU version of VASP will be three times faster than the current version. This should increase the efficiency of our VASP calculations. See http://www.mcs.anl.gov/~jlow/VASP/Scaling%20data%20for%20VASP%20on%20Blues.p... for details on the parallel scaling of VASP. See http://www.sciencedirect.com/science/article/pii/S0010465513004013 for details for the parallel scaling of OPENMX. See http://training.uhem.itu.edu.tr/docs/18hazirannano/PW-III-para.pdf for details on the parallel scaling of Quantum Espresso. Storage requirements: Thank You, The LCRC Accounts System