METHODS FOR TREATING SOLVENT EFFECTS AND INTERMOLECULAR FORCES. Mark S. Gordon Iowa State University Ames Laboratory

Transcription

1 METHODS FOR TREATING SOLVENT EFFECTS AND INTERMOLECULAR FORCES Mark S. Gordon Iowa State University Ames Laboratory

2 OUTLINE Solvation Methods Explicit vs. implicit methods Explicit Methods TIP3P, TIP4P SPC, SPC/E EFP Method for Solvation Summary of EFP1 method for water Sample input files Monte Carlo method Example applications Generalized EFP Method (EFP2)

3 SOLVENT MODELS: CONTINUUM Mostly based on Onsager reaction field model Computationally efficient Sensitive to cavity size and shape Do not account for explicit solute-solvent interactions

4 SOLVATION MODELS: EXPLICIT Solvent molecules described using potentials Empirical potentials: lots of fitted parameters Ab initio potentials: computationally expensive Include explicit solute-solvent interactions Configurational sampling necessary: computationally demanding

5 Multi-Layered Approach to Solvation H H O H O H H O MM H H O H H O H Ab Initio Solute Continuum

6 DISCRETE/EXPLICIT MODELS TIP3P, TIP4P, TIP5P Jorgensen etal, JCP, 79, 926 (1983) Basically simple Lennard-Jones model m n ε mn = [q i q j / r ij ] + A / R 12 C / R 6 i j ε mn =interaction energy between waters m,n q O <0 = -2q H, A, C fitted to bulk properties Rigid molecules TIP4P adds a 4th center inside O toward Hs Very popular model; doesn t get OO radial distribution function right: misses 2nd peak

7 DISCRETE/EXPLICIT MODELS SPC (Simple Point Charge), SPC/E Berendsen etal, JPC, 91, 6269 (1987) SPC: m n ε mn = [q i q j / r ij ] + A / R 12 C / R 6 i j SPC same as TIP, but fitted to MD simulations of density and vaporization energy Rigid molecules Gets OO radial distribution function right Accounts for polarizability/induction

8 SPC/E RDF goo(r) Curve: SPC/E Points: exptl R, Angstroms

9 General Effective Fragment Potential Discrete solvation method Fragment potential is one electron contribution to the QM Hamiltonian if QM part is present Potentials are obtained by separate QM calculations depend on properties of isolated molecules can be systematically improved Internally frozen geometries

10 GOALS Develop general approach to accurately describe intermolecular interactions Solvent effects on Ions and molecules Chemical reaction mechanisms Spectral shifts Liquid behavior Liquid-gas, liquid-surface interface Van der Waals interactions Polymer interactions Enzyme interactions Incorporate all important physics No parameter fits Efficient

11

12 Effective Fragment Potential System is divided into an ab initio region for the solute and a fragment region for the solvent molecules. E = Eab initio + Einteraction

13 General EFP Method In the most general implementation EFP should include all relevant energy contributions: Einteraction = Ecoulomb + Epolarization + Eexch. rep. + Edispersion + Echarge transfer + Ehigher order terms

14 Interaction energy consists of : electrostatic, polarization and exchange repulsion/charge transfer term Einteraction= Ecoulomb + Epolarization + Eexchange repulsion/charge transfer K E interaction = V Elec k (µ,s) + V Pol Re l (µ,s) + V p m (µ, s) k =1 Hartree Fock based EFP L l =1 M m =1 Distributed Multipolar expansion Damping at small R LMO polarizability expansion Iterate to self-consistency Fit to Functional Form

15 Analytic gradients (first derivatives) have been derived and coded for all terms. So, one can perform geometry optimizations, transition state searches, dynamics

16 HIGHER LEVELS OF EFP1 DFT-based EFP (Ivana Adamovic) Same general approach Based on B3LYP Adds some level of correlation MP2-based EFP (Lyuda Slipchenko,Jie Song) Same general approach Separate fits for exchange repulsion+ct, dispersion More effective correlation, especially at long range EFP-EFP done, EFP-QM in progress

17 EFP1/DFT: APPLICATION TO S N 2 Microsolvation of Cl - + CH 3 Br (Ronnie Bierbaum) Solute=DFT/B3LYP, MP2, CCSD(T) Water=DFT, MP2, EFP/DFT Preliminary analysis CCSD(T) has minor effect relative to MP2 (< 1 kcal/ mol) Basis set effects are small (<1 kcal/mol) Basis set used: aug-cc-pvdz

18 GAS PHASE 0.00 separate reactants Cl Br IRC -6.2 DFT (4.4) -1.5 MP2 (10.2) transition state IRC -7.0 DFT -6.1 MP2 products DFT MP2 ion-molecule reactant DFT MP2 ion-molecule product

19 DFT

20 Forward central barrier and relative differences as a function of n (kcal/mol) Forward E a ΔE a n=0 DFT n=1 EFP/DFT DFT n=2 EFP/DFT DFT n=3 (isomer 2) EFP/DFT DFT n=0 MP n=1 MP2/EFP MP n=2 MP2/EFP MP n=3 (isomer 2) MP2/EFP MP n=4 (isomer 5) MP2/EFP

21 Forward central barrier and relative differences as a function of n (kcal/mol) Forward E a ΔE a n=0 DFT n=1 EFP/DFT DFT n=2 EFP/DFT DFT n=3 (isomer 2) EFP/DFT DFT n=0 MP n=1 MP2/EFP MP n=2 MP2/EFP MP n=3 (isomer 2) MP2/EFP MP n=4 (isomer 5) MP2/EFP

22 Forward central barrier and relative differences as a function of n (kcal/mol) Forward E a ΔE a n=0 DFT n=1 EFP/DFT DFT n=2 EFP/DFT DFT n=3 (isomer 2) EFP/DFT DFT n=0 MP n=1 MP2/EFP MP n=2 MP2/EFP MP n=3 (isomer 2) MP2/EFP MP n=4 (isomer 5) MP2/EFP

23 CONCLUSIONS EFP reproduces QM at small fraction of cost DFT greatly underestimates central barrier DFT reproduces solvent effect on central barrier MP2 for solute with EFP for solvent good compromise Reproduces both absolute and solvent-shifted MP2 central barriers Suggests EFP1/MP2 will be just as good

24 Water hexamer isomers Binding DFT MP2 CCSD HF Energy B3LYP (T) prism cage book cyclic boat basis set: DH(d,p) -units: kcal/mol

25 Water hexamer isomers Binding Energy DFT B3LYP EFP1/ DFT MP2 CCSD (T) HF EFP1/ HF prism cage book cyclic boat basis set: DH(d,p) -units: kcal/mol

26 Water hexamer isomers Binding Energy DFT B3LYP EFP1/ DFT MP2 EFP1/ MP2 CCSD (T) HF EFP1/ HF prism cage book cyclic boat basis set: DH(d,p) -units: kcal/mol

27 Initial structure: 62 waters, at least 26 ps equilibration Timestep size = 1 fs 2.5 goo(r) EFP1/HF EFP1/DFT EFP1/MP2 SPC/E Exp (THG) r (Angstroms) Exp (THG): X-ray; Sorenson et. al. J. Chem. Phys. 113, 9149 (2000).

28 Initial structure: 512 waters, at least 5 ps equilibration Timestep size = 1 fs goo(r) EFP1/HF EFP1/DFT EFP1/MP2 SPC/E Exp (THG) r (Angstroms) Exp (THG): X-ray; Sorenson et. al. J. Chem. Phys. 113, 9149 (2000).

29 GENERAL EFP METHOD: EFP2 Jan Jensen Mark Freitag Ivana Adamovic Hui Li Lyuda Slipchenko Dan Kemp Tony Smith Peng Xu Mol. Phys., 89, 1313 (1996) JCP, 108, 4772 (1999) JCP, 112, 7300 (2000) Mol. Phys., 103, 379 (2005) Mol. Phys., 107, 999 (2009)

30 Generalized EFP2 Method Interaction energy consists of : electrostatic, polarization, exchange repulsion, dispersion, and charge transfer terms Einteraction= ECoulomb + EPolarization + Eexrep + Edispersion + ECharge Transfer No Parameter Fits Distributed Multipolar expansion Screening: near field LMO polarizability expansion Iterate to self-consistency From first principles using LMO overlaps Distributed LMO dispersion from first principles

31 S ij 2 E exch 2 2 2ln S ij A B 2 S ij F ik S kj + F jl S li 2T ij π i A j B Summary If all approximations presented above are combined, E exch can be approximated as (provided LMOs are used) R ij i A j B S ij Z J R ij + 2 R il + Z I R Ij + 2 R kj R ij J B l B I B k A i A j B k A l B This equation requires only the computation of intermolecular overlap and electronic kinetic energy integrals, i.e. no two-electron integrals other than those calculated once for the isolated molecules. It contains no adjustable parameters, only fixed parameters computed for the isolated molecules, such as the LMOs in some AO basis, Fock matrices in the LMO bases, and the LMO centroids of charge Numerical Tests E exch E exch (kcal/mol) C s C 2v E exch (Morokuma) RHF/6-31++(d,p) Oxygen-Oxygen Distance (Å)

32 EFP2 Dispersion: Ivana Adamovic E disp = C 6 R 6 + C 7 R 7 + C 8 R Simplest approximation: terminate at C 6 term Calculate E disp using distributed LMO expansion, E disp = x,y,z k A j B αβγδ T αβ kj T γδ kj 0 dνα k αγ (iν) j α βδ (iν) = k A j B C 6 kj / R 6 kj In terms of dynamic polarizability over imaginary frequency range and field gradient components, T Accomplished using dynamic TDHF equations Amos et al., JPC, 89, 2186 (1985); CPL, 278, 278 (1997)

33 Overlap Damping Based on intermolecular overlap Already available for exchange repulsion No fitting required, no additional cost m 2ln S ) ( (i, j ) = 1 S f dmp 2 ij m n=0 n/ 2 ij n! Mol. Phys., 107, 999 (2009); J. Phys. Chem. 119, 2161 (2015)

34 Charge transfer (CT) HOMO E= LUMO E=0.011 Charge Transfer Virtual Virtual Occupied A B Occupied Ammonium nitrate RHF/6-31++G(d,p) - Excitation of electrons in A to the unoccupied orbitals of B - Significant in ionic systems

35 Water Hexamers bag boat book cage cyclic prism

36 Water Hexamer Errors in Total Interaction Energy (relative to CCSD(T)//MP2/aug-cc-pVTZ)

37 Water Hexamer Errors in Total Interaction Energy (relative to CCSD(T)//MP2/aug-cc-pVTZ)

38 Water Hexamer Errors in Total Interaction Energy (relative to CCSD(T)//MP2/aug-cc-pVTZ)

39 Water Hexamer (Prism) Contributions to the Total Interaction Energy

40 Water Hexamer (Prism) Contributions to the Total Interaction Energy

41 Water Hexamer (Prism) Contributions to the Total Interaction Energy

42 Relative CCSD(T) Interaction Energies of Water Hexamers

43 Ar Hexamer c2v Total Interaction Energy CCSD(T)/aug-cc-pVTZ = FMO2-CCSD(T)/aug-cc-pVQZ = d4h Total Interaction Energy CCSD(T)/aug-cc-pVTZ = FMO2-CCSD(T)/aug-cc-pVQZ = d3h Total Interaction Energy CCSD(T)/aug-cc-pVTZ = FMO2-CCSD(T)/aug-cc-pVQZ = -3.07

44 Total Interaction Energies Argon Hexamer Prism Bag Trigonal Planar Method D4h C2v D3h EFP MP2/ACCD MP2/ACCT MP2/ACCQ CC/ACCD CC/ACCT

45 2-Body Interaction Energy Argon Hexamer Prism Bag Trigonal Planar Method D4h C2v D3h EFP MP2/ACCD MP2/ACCT MP2/ACCQ CC/ACCD CC/ACCT

46 3-Body Interaction Energy Argon Hexamer Prism Bag Trigonal Planar Method D4h C2v D3h EFP MP2/ACCD MP2/ACCT MP2/ACCQ CC/ACCD CC/ACCT

47 Ar Hexamer: Error (Total Interaction Energy) Relative to CC/ACCT

48 Error (Total Interaction Energy) Relative to CC/ ACCT - Ar Hexamer

49 Error (Total Interaction Energy) Relative to CC/ ACCT - Ar Hexamer

50 Benzene Dimer: Lyuda Slipchenko Some structures dominated by dispersion Especially π π interactions HF cannot get this right DFT only with special effort Requires correlated methods: MP2, CCSD(T) Good test of EFP

51 Low-energy configurations of benzene dimer R 2 R R R 1 sandwich T-shaped parallel-displaced High-level ab initio calculations: M.O. Sinnokrot, E.F. Valeev, and C.D. Sherrill, JACS, 124, (2002); M.O. Sinnokrot and C.D. Sherrill, JPC, 108, (2004)

52 Compare EFP2 with Symmetry Adapted Perturbation Theory Sandwich & T-shape: SAPT: aug-cc-pvdz M.O. Sinnokrot and C.D. Sherrill, JPC, 108, (2004) EFP2: G(3df,2p) electrostatic with charge-charge damping term exchange repulsion polarization dispersion

53 Electrostatic Term

54 Exchange Repulsion

55 Polarization

56 Dispersion

57 Total Binding Energy

58 Parallel-Displaced Configuration

59 Benzene Dimer: Summary method basis sandwich T-shape parallel-displaced R energy R energy R1 R2 energy MP2 aug-cc-pvdz MP2 aug-cc-pvtz MP2 aug-cc-pvqz CCSD(T ) CCSD(T ) EFP2 aug-cc-pvdz aug-cc-pvqz G (3df,2p) distances: angstroms, energies: kcal/mol EFP: ~.4 sec CCSD(T): ~10 days

60 Benzene Dimer: Summary method basis sandwich T-shape parallel-displaced R energy R energy R1 R2 energy MP2 aug-cc-pvdz MP2 aug-cc-pvtz MP2 aug-cc-pvqz CCSD(T ) CCSD(T ) EFP2 aug-cc-pvdz aug-cc-pvqz G (3df,2p) distances: angstroms, energies: kcal/mol EFP: ~.4 sec MP2/QZ: ~10 hours

61 Contributions to the Interaction Energy dimer Exchange Induction Dispersion Total sandwich T-shaped Electrostatic paralleldisplaced

62 DNA BASES: Q.T. SMITH Compare with Jurecka, P.; Hobza, P. J. Am. Chem. Soc. 2003, 125, 15608) RI-MP2 / ext. cc-pvtz or TZVPP geometries RI-MP2/aug-cc-pVXZ (X = D, T, Q) to CBS limit energies CCSD(T) correction added Counterpoise corrected for BSSE EFP generated with G(3df,2p) basis set

63 Guanine-Cytosine H-Bonded RI-MP2/TVZPP optimized Adenine-Thymine H-Bonded 1.76 Å 1.93 Å 1.91 Å 1.82 Å 1.92 Å CCSD(T)/CBS energy: kcal/mol CCSD(T)/CBS energy: kcal/mol EFP COUL REP 36.4 POL DISP TOTAL EFP COUL REP 26.7 POL -7.2 DISP -8.2 TOTAL -15.8

64 Guanine-Cytosine stacked RI-MP2/TVZPP optimized Adenine-Thymine Stacked RI-MP2/TVZPP optimized 3.40 Å 3.20 Å CCSD(T)/CBS energy: kcal/mol CCSD(T)/CBS energy: kcal/mol EFP COUL REP 20.6 POL -3.1 DISP TOTAL EFP COUL -7.8 REP 16.1 POL -0.7 DISP TOTAL -10.0

65 CURRENT LIMITATIONS EFP1 is for water only QM part can be any level of theory Solvent must be water EFP2 can be generated for any species QM-EFP energies are now done QM-EFP gradients are in progress

66 CONTINUUM SOLVATION MODELS Simplest is Onsager self-consistent reaction field (SCRF) model ($SCRF) Solute (QM) dipole moment µ polarizes medium (solvent) through dielectric ε Newly polarized solvent alters solute dipole moment Iterated to self-consistency Accomplished by adding new term to QM Hamiltonian

67 ONSAGER SCRF MODEL V σ = -r.r r = position vector R is proportional to molecular dipole moment R = gµ µ = dipole vector g = 2 (ε 1)/[(2ε+1)(a 3 )] a = radius of cavity Usually assume spherical cavity Radius can be pretty arbitrary Adds simple <χ i x χ j > integrals - very cheap

68 PCM MODEL Polarizable Continuum Model (PCM) Tomasi etal., Chem. Phys. Lett., 255, 327 (1996) Van der Waals type surface cavity Uses detailed knowledge of electrostatic potential Cavity dispersion potential determined from surface area Can include dispersion effects Interfaced with EFP in GAMESS: $PCM See both input and reference sections in manual

69 OTHER METHODS COSMO Initially developed by Klamt Further developed by Baldridge (U. Zurich) No gradients, not interfaced with EFP SVP Chipman (Notre Dame Radiation Laboratory) Accounts for charge leakage outside cavity SM5 Truhlar, Cramer (Minnesota) Highly parametrized, interfaced with GAMESS