These projects have come up in a couple of emails lately which means it's time to write a blog post.
PM6 in GAMESS
+Jimmy Charnley Kromann is working on implementing PM6 in GAMESS. It is working for elements up to F. However, heavier elements require semi-empirical integrals involving d-orbitals which are not in GAMESS. We have gotten some integral code from James Stewart (who has been a great help in general!) that Jimmy is working on implementing in GAMESS.
The first step is to reproduce PM6 calculations for molecules with non-d-orbitals atoms using the new integral code. As I remember, the 1-electron Hamiltonian is correct, but the 2-electron part of the Fock matrix is still wrong. It looks like the integrals are computed correctly but indexed wrongly. Jimmy is busy with classes now but will be back on the case in a few weeks.
Once PM6 is working we also plan to implement DH2 and DH+.
Semiempirical PCM calculations
+Casper Steinmann is spending his vacation working on semiempirical PCM calculations, in particular the implementation suggested by Chudinov et al. Casper is making great progress so this post will likely be out of date soon and some of the results I refer to are only a few days old.
It looks like the energy calculations are working correctly in the sense that we get very similar solvation energies to Chudinov et al. and RHF/STO-3G calculations. Gradients have just been implemented but some geometry optimizations fail because the energy changes very little while the maximum gradient is still too large, indicating a numerical problem. Other geometry optimizations work fine.
The implementation is only for s- and p-functions so far.
Computational efficiency and future directions
Jimmy tested the computational efficiency of PM6 in GAMESS and MOPAC2009 on some large molecules. An energy + gradient calculation is slower in GAMESS but the geometry optimization converges faster resulting in a overall time savings, but much more testing needs to be done. We don't know yet how efficient the PCM implementation is yet.
One future goal is to be able to increase the computational efficiency for large molecules such as proteins through a combination of parallelization and fragmentation a la FMO.
This work is licensed under a Creative Commons Attribution 3.0 Unported License.
PM6 in GAMESS
+Jimmy Charnley Kromann is working on implementing PM6 in GAMESS. It is working for elements up to F. However, heavier elements require semi-empirical integrals involving d-orbitals which are not in GAMESS. We have gotten some integral code from James Stewart (who has been a great help in general!) that Jimmy is working on implementing in GAMESS.
The first step is to reproduce PM6 calculations for molecules with non-d-orbitals atoms using the new integral code. As I remember, the 1-electron Hamiltonian is correct, but the 2-electron part of the Fock matrix is still wrong. It looks like the integrals are computed correctly but indexed wrongly. Jimmy is busy with classes now but will be back on the case in a few weeks.
Once PM6 is working we also plan to implement DH2 and DH+.
Semiempirical PCM calculations
+Casper Steinmann is spending his vacation working on semiempirical PCM calculations, in particular the implementation suggested by Chudinov et al. Casper is making great progress so this post will likely be out of date soon and some of the results I refer to are only a few days old.
It looks like the energy calculations are working correctly in the sense that we get very similar solvation energies to Chudinov et al. and RHF/STO-3G calculations. Gradients have just been implemented but some geometry optimizations fail because the energy changes very little while the maximum gradient is still too large, indicating a numerical problem. Other geometry optimizations work fine.
The implementation is only for s- and p-functions so far.
Computational efficiency and future directions
Jimmy tested the computational efficiency of PM6 in GAMESS and MOPAC2009 on some large molecules. An energy + gradient calculation is slower in GAMESS but the geometry optimization converges faster resulting in a overall time savings, but much more testing needs to be done. We don't know yet how efficient the PCM implementation is yet.
One future goal is to be able to increase the computational efficiency for large molecules such as proteins through a combination of parallelization and fragmentation a la FMO.
This work is licensed under a Creative Commons Attribution 3.0 Unported License.
No comments:
Post a Comment