Execution time is between one or two months using i7 processor with four cores for a Gaussian 09 job that contains more than 50 heavy atoms. Is there any way to reduce this time? And how much the reduction can be? What is the alternative without sacrificing accuracy (method and basis vectors)?
I need to know a bit more about what calculations you are running to be able to help you.
What do you mean with the term "heavy atoms'? Different people have different definitions of 'heavy'.
What calculation type, level of theory, and basis set are you currently using?
Dear Nathaniel Gunby
Heavy atoms usually mean atoms with atomic number around 50 or higher (relativistic effects might be included). Some times a long computational time may be even with 500 carbon atoms using basis such as 6-31(d). The kind of calculation can be geometrical optimization or frequency calculations. My question is how to reduce computational time and how much it will be reduced by another computational hardware or software without changing the implemented theory.
Dear Mudar
Thank you for clarifying. I thought that was what you meant by heavy atoms, but some people use the term to refer to boron and heavier.
For large molecules, a two-step optimisation where you use a lower level of theory/small basis set to get close to the minimum before optimising that geometry with the methods you want to use can reduce computational time, but that doesn't work so well with heavy atoms where relativistic effects make the level of theory and basis set more important. This doesn't help with computing frequencies, which (as far as I am aware) are hard to speed up.
For systems with a large number of electrons, increasing the available memory is important for speed. If you can't increase the memory enough to avoid swapping to disk, using an SSD instead of a traditional spinning-disk hard drive for the swap space can improve performance.
General speed optimisations are still useful here. One is to use direct algorithms (which re-compute values instead of storing them) when it reduces swapping, Another, which works if you are interested in different conformers of one molecule, is to optimise a single conformer the use its final orbitals as an initial guess for the others, which should reduce the number of iterations for the wavefunction to converge.
I hope this is useful.
Mr. Abdulsattar,
Deffinitively in all chases the application of DFT approachers is highly recommendable. They provided a very balanced strategy between the comoputational-cost and accuracy. For large (bio)macromolecules the molecular mechanics/dynamics (MM/MD) calculations are recommendably.
A. For pure organic molecules suitable methods are described in the paper 1:
Paper 1. International Journal of Biological Macromolecules, 64, 2014, 383–391, Macromolecular ensembles of cyclodextrin crystallohydrates and clathrates – experimental and theoretical gas – and condense phase study, Bojidarka Ivanova, Michael Spiteller
The utilization of Truhlar' functionals such as M05, M06, M06-2X provide highly accurate information about the structure an dproperties. Details of the theoretical background is given in papers 2 and 3, where are compared also the data with other ab initio and DFT methods:
Paper 2:Theoretical Chemistry Accounts, 2008, 120, 215-241,The M06 suite of density functionals for main group thermochemistry, thermochemical kinetics, noncovalent interactions, excited states, and transition elements: two new functionals and systematic testing of four M06-class functionals and 12 other functionals, Yan Zhao, Donald G. Truhlar
Paper 3: Theoretical Chemistry Accounts, 2011, 129, 219-227, Coupled cluster, B2PLYP and M06-2X investigation of the thermochemistry of five-membered nitrogen containing heterocycles, furan, and thiophene, Pablo A. Denis
Towards the besis sets very accurate information is achieved utilizing the Dunning's correlation consisted basis sets such as cc-pVDZ, cc-pVTZ, cc-pVQZ, cc-pV5Z, aug-cc-pVDZ, are related. But their utilization is relatively time consuming. So that some of the Pople's basis sets such as 6-31+G*, 6-31G(3df, 3pd), 6-311G or related with suitable polarization and diffusion function also provide very accurate data at a relatively balansing computational costs.
B. For inorganic or metal organic compounds suitable theoretical levels are shown in the paper 4.
Paper 4: Adsorption of uranium composites onto saltrock oxides – experimental and theoretical study, Journal of Environmental Radioactivity, 135, 2014, 75-83, Bojidarka Ivanova, Michael Spiteller
The utilization of quasirelativistic effective core pseudopotentials
from Stuttgart-Dresden (SDD) provide highly meaningful information at relatively significantly reduced computational time (resp. costs).
C. For large (bio)macromolecules, theoretical levels are described in paper 5:
Paper 5: International Journal of Biological Macromolecules, 65, 2014, 314–324, Evodiamine and rutaecarpine alkaloids as highly selective transient receptor potential vanilloid 1 agonists, Bojidarka Ivanova, Michael Spiteller
Dear Nathaniel Gunby
Thanks for your answer. I tried some of the methods that you described. I hope I can get useful results trying the others.
Dear Bojidarka B. Ivanova
Thanks for your detailed answer. Actually, I am interested in inorganic compounds and I will try to benefit from the literature mentioned in your answer.
- Badges
- Science topic