FMO-QMC strategy using CASINO CASINO workshop@Towler Institute, Tuscany, Italy. FMO-QMC strategy using CASINO (Fragment Molecular Orbital method) Ryo Maezono MAEZONO.Ryo@nims.go.jp National Institute for Materials Science, Tsukuba, Japan.
Huge molecules… Very interesting target for precise QMC… FMOイントロ(1) Huge molecules… ‘Fr1’ ‘Fr2’ ‘Fr3’ Very interesting target for precise QMC… - The whole molecule is too huge… Generating the whole trial nodes is difficult… - Quantum dynamics is believed to be important only within each fragment. For distant fragments, the interaction is well described by classic electrostatic fields, as believed.
What is FMO? Dividing the whole molecule calculation ‘Fr1’ ‘Fr2’ ‘Fr3’ ‘Glycine_trimer’ (Fragment Molecular Orbital method) Dividing the whole molecule calculation into calculations on each fragment. Approximation; Now implemented in ‘GAMESS’ and ‘ABINIT_MP’, ‘NWChem’
FMO近似 FMO approximation Interactions only between adjacent fragments is taken into account. -Two fragments example; ; … III IV I II -Multi fragments extension; ; ( )
Procedure… -FMO-SCF; -Similar procedure on QMC… More correlations! ; 計算手順(1) Procedure… -FMO-SCF; Iterations to reach convergence… ; -Similar procedure on QMC… by ‘GAMESS’ or ‘ABINIT_MP’ Cusp_corr., Jastrow, DMC, … More correlations!
Procedure… -Similar procedure on QMC… - More correlations! 計算手順(2) Procedure… -Similar procedure on QMC… by ‘GAMESS’ or ‘ABINIT_MP’ by ‘GAMESS’ or ‘ABINIT_MP’ ; Cusp_corr., Jastrow, DMC, … - More correlations! - No huge gwfn.data is required! based on partial gwfn.data for each fragment…
Future perspective… Accurate treatment of ‘reaction centre’ 将来構想 Future perspective… Accurate treatment of ‘reaction centre’ without annoying huge whole size… (e.g., DNA stacking…) vdW force ‘Engine’ ‘Framework’ - FMO - MP2, MP4 etc… - QMC…
As a kick-off work… FMO approx. is reasonable at the level of QMC?? 最初の構想 As a kick-off work… FMO approx. is reasonable at the level of QMC?? ?? (@ MP2 they confirmed.) ‘Fr1’ ‘Fr2’ ‘Fr3’ - FMO-SCF (ABINIT_MP or GAMESS). FMO; -687.387353735997 Full; -687.3875159324 - VMC with Jastrow. - DMC - HFVMC (VMC w/o Jastrow) Full Whole QMC is feasible.
Trial implementation… 實装の試行 Trial implementation… - by FMO-SCF as ‘gwfn.fmo’ Code extension to include - by FMO-SCF as ‘mol_density.fmo’ (by ‘ABINIT_MP’) input, gwfn.data mol_density.fmo, gwfn.fmo - Ion positions of other fragments. - Charge densities of other fragments accumulated on each cells.
Result at HFVMC… Full-SCF ; -687.3875159324 HFVMC ABINIT_MP fr1 -122.86(7) -122.7992645203 0 fr2 -307.8(1) -307.6073508713 - fr3 -683.5(2) -683.4622078107 0 fp12 -518.4(4) -518.311823403056 0 fp13 -885.1(1) -885.152669732591 0 fp23 -1219.5(2) -1218.91371271375 - FMO-SCF ; -687.387353735997 Full-SCF ; -687.3875159324 FMO-HFVMC ; -687.8(6) Full-HFVMC ; -686.9(9) N.B., No cusp corrections. - Large amount of accumulation, - careful choice of ‘NVMCAVE’ - ‘Core stacking’ --> lower estimation (*Andrea’s paper)
Technical issues… - Cusp corrections: - Filters: - Code extensions: 技術的事項 Technical issues… - Cusp corrections: ‘Floating gaussian centre’ is used… code extension by Mike. - Filters: {gwfn.data_fr_j} ‘ABINIT_MP’ etc. ‘CASINO’ {density.data_fr_j} - Code extensions: - Calculation including electro-static field from other fragments. (trial by RM)
Summary… ○ FMO-SCF --> FMO-QMC enables us… まとめ Summary… ○ FMO-SCF --> FMO-QMC enables us… to concentrate only on the active site without treating huge trial wavefunctions. where precise handling of …for increasing system size MP2 QMC electronic correlation is required. ○ On-going work… - VMC --> DMC ○ Future… - DNA stacking, Biological reactions etc. where the correlation expected to be important…