| The elucidation of the structure of chemical compounds, is more and more effectuated, by a combination of experiment and theory (computing). Only through comparison of the experimentally measured spectrum with computationally determined spectra for either model or the actual compound, an assignment of the peaks in the spectrum can be made and the structure of the system is determined[1]. We intend to extend the parallel capabilities in GAMESS-UK[2]. There is already a closed and open shell serial hessian implementation, including Infrared intensities, but only RAMAN intensities for the closed shell case [5,6]. The situation is even grimmer for the parallel implementation. There only the closed shell infrared case is parallelised [8], including DFT. This is a parallel semi-direct implementation, which uses Global Arrays[3]. Unfortunately this version is rather memory hungry in the CHF equations, as all the right-hand-sides are simultaneously required in memory as well as the transformed 2-electron integrals. It is however a start that can be built upon. I therefore propose to clean up the program once and for all. This involves the following [7,8,9,10] - Implement a multipass CHF solver with perturbation data distributed. - Implementing integral direct capabilities in the CHF solver [4]. - Adding the polarisability derivatives, to calculate Raman intensity [5,6] - Extending the Closed Shell engine to handle open and unrestricted cases (DFT). A lot of the basis machinery is already present, but quite a bit of interfacing must be done. After the project we'll have a program that can yield Infrared and Raman spectra for real life molecules with basis sets that will match experimental accuracy. We expect the code to parallelize efficiently over 512 processors. |
