Hello, A yearly allocation for the LCRC cluster has been requested with the following updated information: Submitter/PI: Kristy Mardis Project Name: Solar_EPR Division: CSE Project title: Solar energy conversion in organic photovoltaics Associated funding: DOE Division of Chemical Sciences, Geosciences, and Biosciences Other Systems: Dr. Mardis has three small linux clusters (12/24/48 nodes) which are housed at her home institution. Her research group is the sole users of these systems. Orca, pqs, gabedit, and gamess all can be run on these systems. Science: This work is focused on the elucidating the electronic structures of donor-acceptor polymers which are the outstanding solar conversion material for organic photovoltaic (OPV) devices. Energy conversion process in OPV is the initial charge separation which yields a positive polaron on the polymer and a negative polaron on the electron acceptor, often a fullerene derivative. Stronger delocalization of the positive polaron on the polymer minimizes wasteful charge recombination. The delocalization correlates with solar energy conversion efficiency and is determined by the electronic structures of the organic molecules. We propose to use density functional theory (DFT) to investigate delocalization in polymers and small oligomers that are currently used in OPVs and have been suggested to be a future replacement of the polymers. This work will test hypotheses centered on the role of conformation and chemical structure on electron transfer. The first aim is to evaluate the applicability of current DFT methods for modeling EPR parameters of OPV systems. The second aim is to evaluate the extent of polaron delocalization along the oligomers. The third aim is to explore the effect of solvent and conformation, especially conformational changes between solution and thin film, on the calculated magnetic resonance parameters in order to assess the feasibility of this approach for modeling thin film OPV systems. These results will facilitate interpretation of our multifrequency Electron Paramagnetic Resonance (EPR) data of polymer and small molecule (oligomer) solar cells. Project description: Prior work suggested polaron delocalization in polymers and larger oligomers ranged over multiple subunits; up to 60 Angstroms. However, due to our limited existing computational power we were not able to assess structures in the size. Furthermore, test calculations indicated sensitivity of the delocalization to both conformation and presence of external charges in the surrounding of the molecule. The current work seeks to investigate a series of recently synthesized well-defined oligomers to test the limits of the delocalization. We intend to run a series of gas-phase model DFT optimizations to determine the low energy oligomer conformations using standard ab initio optimization codes (NWCHEM). PTB7 oligomers ranging in size from n=1, 2, 3 and 6 and P3HT oligomers ranging from n=5 to n=30 will be optimized using both traditional hybrid and long-range corrected functionals. Furthermore, Orca 4.0.0.2 (referred to as Orca 4 in the following) will be used to assess the suitability of the PBEh-3c composite method for optimizing larger oligomers at lower computational cost. The resultant conformers will be used to calculate EPR parameters (g-tensors and hyperfine coupling constants) using Orca 4. Oligomers of varying length will be investigated to evaluate delocalization ranges and sensitivity of electronic structure parameters to oligomer length. The initial structure building and analysis of spin densities will be carried out on our lab computers. Additionally, optimizations and EPR parameter calcu lations will be performed on a variety of conformations with differing point group assignments for the C60 fullerene. Experimental EPR data show a broad peak giving g values in a range of 1.994-1.996 for the thin film sample that is not present in the solution sample. Comparison with distorted calculated structures should allow the assignment of this peak. The primary usage of blues/bebop will be in the optimization, frequency calculation, and EPR parameter calculation steps of the work. The optimization subroutines of NWCHEM are highly parallelized and can be performed on bebop. Orca does not scale well in terms of the number of processors, with scaling deteriorating after 16 processors. Test calculations on bebop show that going from 12 to 18 processors increases the rate by 1.27 rather than the theoretical 1.5 and going to 36 processors gives a speedup of only 1.76 vs 3 and going to 2 nodes of 36 (72 processors) resulted in only a speedup of 2.23 rather than a factor of 6. Thus, running orca on blues rather than bebop is by far a better choice were two nodes are only 32 processors (and a reasonable speedup of 1.6 versus 2 theoretical of going from 1 node to 2 is still present). Furthermore, Orca 4 is a highly accurate program for calculating EPR parameters and there is no alternative program at this level available. Howe ver, ORCA needs significant memory for oligomers of greater than n=2 (for PTB7 oligomers) and n=10 (for P3HT oligomers) and the C60 fullerene. For all these systems, the bigger memory nodes on blues are necessary (biggpu). Thus we request most of our allocation to be on blues rather than fully transitioning to bebop (until and unless bigmem nodes are available on bebop). Test calculations performed on our small local workstation for the largest oligomer we calculated so far took about 200 (small basis set) to 20,000 (large basis set) CPU hours (using PQS as the optimization engine) for optimization and frequency calculations. The existing allocation is being used to set-up and run optimizations using Density Functional Theory using NWCHEM (B3LYP and CAM-B3LYP) and Orca (PBEh-3c). Calculations for the PTB7 n=2, 3, and 4 oligomers with all cis and all trans conformations are completed. This work will include the largest PTB7 oligomer (n=6) and mixed cis/trans conformations. These calculations indicate that each n=4 conformation requires between 800-1400 CPU hours (using timings on 16 nodes/16 processors per node on blues) depending on the functional chosen. The EPR parameter calculation (g-tensor and hyperfine calculations) took 1200 CPU hours (using Orca 3.0.3; no significant speedup is expected for version 4). As P3HT has a less substituted monomeric unit, the number of conformations scales more favorably with the number of oligomer units. However, it is expected that n>25 sized oligomers will be investigated before the maximum extent of delocalization is reached. Again, optimizations will b e performed primarily in NWCHEM, but EPR parameters and the PBEh-3c optimizations (not implemented in NWCHEM yet) will be performed in ORCA. Thus, in summary, for PTB7 we request time to for the n=6 oligomer which has 10 rotatable bonds leading to 55 possible conformations (we plan to sample 1/3 of them) for a total of 81,000 hours (CPU x # of conformation x tested functionals = 1500 x 18 x 3) for the NWCHEM optimizations using B3LYP and a medium basis set. Completing the smaller n=3 and n=4 PTB7 oligomers is expected to take 36,000 hours (1200 hours x 10 conformers x 3 functionals) The EPR parameter calculations will take 158,400 hours (1200 hours x 66 structures x 2 for biggpu node) CPU hours. For the P3HT work, the smaller conformations (up to n=10) will be performed on local systems and we also hope to primarily use the PBEh-3c approach which performs significantly faster in Orca. Using LCRC resources for n = 12, 15, 20, and 25 oligomers and assuming 6 representative low energy conformations per oligomer would require 6000 hours for optimization (4 oligomers x 6 conformations x 250 hours for PBEh-3C opt of n = 15) and 6000 hours for EPR parameter calculation (smaller conformers cost about 8 CPU hours; larger about 120 CPU hours/conformation x factor of 2 for biggpu node). Finally, the fullerene C60 work is expected to involve optimizations of distorted structures in six different point groups taking an estimated 30,000 hours. The EPR calculations only involve g-values and thus will be performed on our local systems (no LCRC resources needed). This totals to 317,160 CPU hours. We are requesting an allocation of 80,000 per quarter equally spaced as Dr. Mardis is on sabbatical and thus will be able to use the hours evenly throughout the allocation period. Furthermore, we are requesting that more than half of that time be on blues where Orca is more efficient and where the biggpu nodes are located (which are required for the EPR parameter calculations). We expect 3 project members (Mardis, Niklas, Poluektov). No non-parallel jobs will be run and all visualization and analysis will be done on local machines. Industry partnership: Project URL: Current FY Hours Used: undetermined amount New FY Requested allocation: 320000 Q1: 80000 Q2: 80000 Q3: 80000 Q4: 80000 Justification: Storage requirements: 1TB Thank You, The LCRC Accounts System