Hello, A yearly allocation for the LCRC cluster has been requested with the following updated information: Submitter/PI: Rees Rankin Project Name: DFT_TM_Catal_Screen Division: CNM Project title: DFT Screening Investigation on Catalytic and Electrocatalytic Phenomena of Transition Metals Associated funding: LDRD-$ and Basic Sciences Other Systems: Carbon compute cluster @ CNM here at ANL Science: In this project, we will expand on successful previous efforts (FY2009,FY2010,FY2011) to utilize the Density Functional Theory (DFT) framework to determine the geometry and associated energy(ies) of molecular reactants, intermediate moieties (e.g. reaction intermediates, and products) on the exposed surfaces of transition metals (TM ). By obtaining such properties, we can derive understanding that serves dual purposes. First, we can directly calculate the potential energy, and hence the free energy of reaction mechanisms of interest on these TM surfaces- whether they be ideal defect free surfaces (flat 111 type surfaces) or step type surfaces which represent the more reactive sites on realistic nanoparticles. Second, we can further populate our database of computational catalytic properties which will be used for predictive screening of additional reactions, catalysts, and products via interpolative schemes and Sabatier Volcano type analysis through previously de termined (and calculated) descriptors. This is particularly relevant when considering catalyst design involving multiple atomic species, such as in the case of binary or ternary alloys. These materials are of great interest for future catalytic needs for reasons of improved catalytic efficiency, product (de)selectivity, and economic viability. Project description: In the previous work in FY2010 and 2011, we performed DFT calculations to begin to study extensions of our transition metal surface catalytic beyond reduction of oxygen by hydrogen to selectively form peroxide. We have extended similar studies to examine CO2 electroreduction (CDR), and reduction of oxygen by lithium towards lithium-oxide products which may be found nucleating on catalysts in LI-Air batteries. We aim to further extend our calculations on TM catalyst surfaces to now include effects of support and high-complexity alloying on overall activity and selectivity for all previously studied reactions, as well as new ones including deNOx electrocatalysis—a critical industrial reaction. Additionally, we seek to perform AIMD finite temperature simulations to verify the structure and energy of interfacial junctions between catalyst and electrolyte in electrochemical cells relevant to the Oyxgen Reduction Reaction (ORR) and other reactions. Comp ared to our previously completed work, the new work proposed requires significantly more calculations of long time duration on multiple nodes, but with less expected reliance on time-exhausting NEB transition state calculations; if our new initial results using studies of reaction free energies alone proves insufficient, future proposal requests for additional time may be submitted to allow for completion of the aforementioned transition state barrier calculations. Our work will continue to supplement the computational materials-by-design and Materials Design Workbench database being developed. The results of populating the MDW database will allow far greater amounts of materials and catalyst reactions to be quickly predictively screened in the future. Calculations will be expected to be performed using the Vienna ab initio Simulation Program (VASP). This program , currently deployed on Fusion, uses a periodic plane-wave representation of the electronic wavefunctions of materials with a self-consistent minimization/solution under the Density Functional Theory (DFT) formalism. Recent improvements in the parallelization of VASP combined with the excellent parallel hardware capabilities in the LCRC Fusion system allow for reasonably efficient near-linear scaling for jobs running on up to ~ 80 cores (10 nodes). Superlinear scaling can occasionally now even be seemingly witnessed for small jobs in a single compute node with appropriate cache sizes. Additional calculations in the future may utilize the Dacapo DFT code, which performs similarly to VASP in the above regards Ongoing and previous calculations for this work have already been performed on the HPC Carbon cluster available at the CNM (ANL), and the Fusion system at LCRC (ANL) in FY2010/FY20101, and are continuing into FY2012. However, the resources available at CNM-ANL Carbon cluster are currently significantly overloaded even after the recent hardware expansion; small scale jobs can actually sit in the submission queue for time intervals longer than their execution related wall time. An allocation of 400k core hours on LCRC-Fusion for FY2012 would allow us to perform an estimated 40-50 AIMD simulations on the 10ps time scale (~200k core hours) to characterize the interface of catalyst/electrolyte structures and energies relevant to various electrocatalytic processes. The remaining portion of the allocation requested would be used to perform ~500 calculation sets (catalyst/adsorbate/site preference scan) for electrocatalysts relevant to the deNOx reactions proposed for study. Based on our previous results with attempts in developing Volcano Plots for prediction of novel alloy catalysts, we feel this allocation sufficient to perform all thermodynamic data calculations needed to parameterize an initio deNOx reactivity and selectivity volcano for pure-component transition metal electrocatalysts. The availability of high performance computing resources such as FUSION will greatly augment and accelerate our work in understanding these critical catalytic problems. Project URL: Current FY Hours Used: undetermined amount New FY Requested allocation: 400000 Q1: 125000 Q2: 125000 Q3: 75000 Q4: 75000 Justification: As a brief summary of typical results, Vasp 4.6 or Vasp 5.x+ with appropriate compilation, Infiniband, and modern processor cores scales well (near linear) out to approximately 80 cores or 10 (Fusion) nodes for most typical calculations on these similar HPC systems. VASP scaling does fall significantly in the regime of calculations with severely large file I/O overhead--however, this is not expected to apply to the jobs performed in the proposed work. We can provide sample benchmarks representative of the type of scaling efficiency our typical calculations have achieved to date on Fusion nodes if necessary. Thank You, The LCRC Accounts System