| Abstract: |
Resources are requested to benchmark calculations in support of three collaborations of the Cooksy group at San Diego State University (SDSU) with SDSU experimentalists studying structures and dynamics of organometallic complexes. We propose to carry out electronic structure calculations (primarily geometry optimizations and harmonic frequency analysis) with a variety of methods drawn from density functional theory (DFT) and ab initio methods employing larger basis sets than we are able to implement using our current SDSU-campus facilities. Two of these projects were presented to us in the last two weeks, and it would be impossible for us to make significant progress on all three within our available resources. The projects are briefly outlined below.
1. Optimizing structural and configurational selectivity of Ru-based zipper catalysts; Douglas B. Grotjahn, PI. The Grotjahn group developed a catalyst that effectively converts terminal alkenes to central alkenes (e.g., 1-hexene to 3-(E)-hexene). More recently, they have developed a variation of this catalyst that exhibits relatively quick conversion of the 1-alkene to the 2-alkene, and a much slower second step, so that the 2-alkene may be effectively separated from the other products. They are also hoping to design catalysts that will enhance production of less stable (Z) configurations. We have been collaborating with the Grotjahn group to model the mechanism for the reaction, using B3LYP-D2 and MN15 functionals in conjunction with cc-pVNZ basis sets and a COSMO solvation model. A challenge in this project is that the kinetically favored reaction pathway is predicted to have three viable candidates for the rate-determining step: a hydrogen transfer step, a hindered torsion, and dissociation of the product from the catalyst. Initial calculations found that single-point energies were fairly insensitive to DFT method and basis set, but we have evidence now that re-optimizing the geometries with a larger basis set may have a significant impact on the calculated barriers. With a cc-pVDZ basis set (aug-cc-pVDZ for the Ru), the reaction system encompasses 622 contracted functions (1315 primitives). The average cpu time per B3LYP optimization step on our present cluster is 165 minutes. Expanding the basis set to cc-pVTZ or pc-2 more than doubles the number of functions, corresponding to a roughly tenfold increase in cpu demands. We can narrow these more resource-intensive calculations to the chief candidates for the rate-determining step, but we will need to benchmark a suitable balance of DFT method and basis set for extending this work to other catalysts and substrates.
2. Perovskites as catalysts in asymmetric synthesis; Yong Yan, PI. Yan, an assistant professor in our department, asked us to model the approach of an amine with hindered rotation to the surface of a PbBr perovskite metal-organic framework, in hopes of explaining the origin of observed stereoselectivity in the ensuing reaction. We have a crystal structure for the perovskite surface, which we plan to model as a large cluster with suitable bond terminations at the edges, frozen to its x-ray geometry. The cluster we hope to use consists of 583 atoms, which we will treat using a 3-layer ONIOM model, including molecular mechanics for organic ligands greater than 5 Å from the reactive site, and DFT with a small LANL2DZ basis set for the PbBr framework. Even so, we anticipate that a meaningful calculation will entail some 3000 basis functions. The closest system in size to this that we have analyzed previously is a 200-atom Pt-based nanocage, which we were able to characterize using a similar combination of basis sets and methodologies, and we have scaled the demands of that project for our estimate below of the resources needed for the perovskite calculations.
3. Aerobactin photochemistry; Carl Carrano, PI. Carrano is a AAAS Fellow, recognized for his contributions to the field of bioinorganic chemistry of marine microorganisms. He has asked us to calculate the relative stabilities of several forms of selected siderophores that play a crucial role in iron photoreduction and decarboxylation reactions. We will calculate the relative energies of the ligands as well as the complexes, before and after photoactivation, and at different levels of protonation. Part of the challenge of this work is to obtain accurate reaction enthalpies and free energies, taking into account the loss of CO¬2 and proton exchange with the environment. These calculations will require counterpoise corrections for the basis set superposition error on modeling the loss of CO2 and H nuclei from the complex, as well as COSMO-RS calculations to improve the modeling of solvated protons. The Cooksy group uses COSMOtherm for the COSMO-RS calculations, but these require input from a previous DFT/COSMO calculation.
The bulk of the calculations will be carried out using Gaussian, with some jobs (such as energy decomposition analysis) likely to be submitted to GAMESS.
We estimate the total cpu-hour requirements at 42,600 core-hours, based on (44 distinct structures of interest) times (50 optimization steps per structure) times (19.4 core-hours per step). .Core-hours here assume a factor of 2 speed-up for AVX2-enabled chips compared to the SSE 4.0 chips currently used. Additional time has been factored in for frequency analysis of stationary point geometries. We will generally try to accelerate optimizations with large basis sets by optimizing with smaller basis sets first, and in many cases we may find that single-point energies are sufficient. Any place that we are able to save time from our allotment will provide added opportunity to benchmark different methods and basis sets against experimental measurements. |
0000-0002-2794-4269