DFT METHODS IN GAUSSIAN

Ivan Rostov,
Australian National University,
Canberra E-mail: Ivan.Rostov@anu.edu.au
Variety Of Methods In Computational Chemistry

Cost
CCSD(T)

MP2 energy correction

DFT methods
Ab initio
Hartree-Fock QM

Semiempirical QM

MM Force Fields

Accuracy 2
Variety of Methods in Computational Chemistry

Quality Size dependence

 Ab initio MO Methods
 CCSD(T) quantitative (1~2 kcal/mol) but expensive ~N6
 MP2 semi-quantitative and doable ~N4
 HF qualitative ~N2-3
 Density Functional Theory
 DFT semi-quantitative and cheap ~N2-3
 Semi-empirical MO Methods
 AM1, PM3, MNDO semi-qualitative ~N2-3
 Molecular Mechanics Force Field
 MM3, Amber, Charmm semi-qualitative (no bond-breaking) ~N1-2

3
 Quantum Chemistry Basics

ˆ   E, where r , r r ,  ,   

H 1 2 N 1 2 N

Variational principle:
ˆ  EE
H ˆ
exact  exact H exact

Born-Oppenheimer (clamped-nuclei) approximation

electrons are fast and moves in the field of fixed nuclei
N M N
1
ˆ   Zm 1 ZmZn
    
2 i 1 i m1 i 1 rim i  j rim mn Rmn

2
H

Hartree-Fock Approximation:
  
EHF  min ESD , where SD  det1 ( x1 )  2 ( x2 )   N ( xN )
1
 SD  N N!

4
 Density Functional Theory Basics
      
 r1     * r1 , r2  rN ,  1 ,  2  N    r1 , r2  rN ,  1 ,  2  N d 3r2 d 3r3  d 3rN d 1d 2  d N

Hohenberg-Kohn Theorems (1964)

 
1. ext
V r   Vext  r 
0  N , Z A , RA   Hˆ  0  E0 and other properties 
Therefore, instead of  dependent on 4N coordinates we would need just 0 dependent on just 3
coordinates
2. The variational principle for DFT
E   E 0 

E   Tk    E Ne    Eee   or
E   Tk    E Ne    J    Exc  
If we would know how to express each of those four terms

J     1
 
1  r  r2    M  r

dr1dr2 ; E Ne     Z m    dr
  
2 r12 m 1 r  Rm

What about Tk[] and Exc[]?

Thomas, Fermi (1927)
Tk    3 2    r  5 / 3dr poor accuracy, as was formulated for the uniform electron gas
3 2/3  
5
10
 Density Functional Theory Basics

Kohn-Sham formalism resolves the problem with the kinetic energy term
1 N  
Tk     i* r  2i r dr
2
 N
 2
 r    i r 

The big unknown left is

EXC    EX    EC  

The Hartree-Fock case:

   
 r  r  r  j 2  
r
EC    0
1
EXHF    
i 1 j 1 i 2
dr1dr2 ;
2 i, j r12
6
 Exchange functionals Ex[]
1/ 3
• Slater:ρ4/3 with theoretical coefficient a =2/3. E     9  3  a  r 4 / 3 d 3 r
Keyword: Used Alone: HFS, Comb. Form: S
X  
8  
• Xαρ4/3 with the empirical coefficient of 0.7, usually used when this exchange functional
is used without a correlation functional
Keyword: Used Alone: XAlpha, Comb. Form: XA.
• Becke 88: Becke's 1988 functional, which includes the Slater exchange along with
corrections involving the gradient of the density. E    f  r ,  r d 3r
Keyword: Used Alone: HFB, Comb.Form: B.
X 
• Perdew-Wang 91: The exchange component of Perdew and Wang's 1991 functional.
Keyword: Used Alone: N/A, Comb. Form: PW91.
• Modified PW91: as modified by Adamo and Barone.
Keyword: Used Alone: N/A, Comb. Form: MPW.
• Gill 96: The 1996 exchange functional of Gill.
Keyword: Used Alone: N/A, Comb. Form: G96.
• PBE: The 1996 functional of Perdew, Burke and Ernzerhof.
Keyword: Used Alone: N/A, Comb. Form: PBE.
• OPTX: Handy's OPTX modification of Becke's exchange functional.
Keyword: Used Alone: N/A, Comb. Form: O.
• TPSS: The exchange functional of Tao, Perdew, Staroverov, and Scuseria.
Keyword: Used Alone: N/A, Comb. Form: TPSS. E    f  r ,  r ,  2  r d 3r
X  7
 Correlation functionals Ec[]

• VWN: Vosko, Wilk, and Nusair 1980 correlation functional fitting the RPA solution to
the uniform electron gas, often referred to as Local Spin Density (LSD) correlation.
• VWN V(VWN5): Functional which fits the Ceperly-Alder solution to the uniform
electron gas.
• LYP: The correlation functional of Lee, Yang, and Parr which includes both local and
non-local terms.
• PL (Perdew Local): The local (non-gradient corrected) functional of Perdew (1981).
• P86 (Perdew 86): The gradient corrections of Perdew, along with his 1981 local
correlation functional.
• PW91 (Perdew/Wang 91): Perdew and Wang's 1991 gradient-corrected correlation
functional.
• B95 (Becke 95): Becke's τ-dependent gradient-corrected correlation functional
(defined as part of his one parameter hybrid functional.
• PBE: The 1996 gradient-corrected correlation functional of Perdew, Burke and
Ernzerhof.
• TPSS: The τ-dependent gradient-corrected functional of Tao, Perdew, Staroverov, and
Scuseria.

8
 Popular combinations of Ex[] and Ec[]
• SVWN=LSDA
• SVWN5
• BLYP
Hybrid functionals
• B3LYP
Exchybr  a0 ExHF  1  a0 ExLDA  ax ExB88  EcLDA  ac EcGGA
a0  0.2; ax  0.72; ac  0.81

• B3P86, B3PW91, B1B95 (1 parameter), B1LYP, MPW1PW91, B98, B971, B972,

PBE1PBE etc.

• You can even construct your own. Gaussian provides such a functionality:
Exc = P2EXHF + P1(P4EXSlater + P3ΔExnon-local) + P6EClocal + P5ΔECnon-local

IOP(3/76),IOP(3/77) and IOP(3/78) setup P1 - P6

B3LYP =
BLYP IOp(3/76=1000002000) IOp(3/77=0720008000) IOp(3/78=0810010000)
9
 New functionals in revision E01 of G03
Hybrid M05, MO5-2X with the same parameterization scheme but different set of
parameters (25!)

As reported by Donald Truhlar and Yan Zhao, M05 and MO5-2X outperform other
parameterized hybrid functionals in nonmetallic thermochemical kinetics,
thermochemistry and noncovalent interactions. MO5-2X is especially good for
calculation of the bond dissociation energies, stacking and hydrogen-bonding
interactions in nucleobase pairs

10
Time-Dependent DFT
Runge-Gross theorem
 

 r, t 

Vext r , t  ˆ r , t 
H r 1,,r n , t 

Runge-Gross equations:
   1 M
  
', t  3    
i  j r , t             d r '   xc r , t  j r , t 
2 Z m r
t  2 m 1 r  R r  r' 
  2
 r , t     j r , t  ;  j t  0   0j
j

 Exc r , t 
and  xc r , t   
r , t 
Linear response of the KS approximation
steff     K st ,uv  Puv  
uv

nst
Pst    steff  
 s   t    11
Time-Dependent DFT
A  B 1/ 2 A  B A  B 1/ 2 FI  I2FI where F  A  B 
1 / 2
XY
X ai  Pai  ; Yai  Pia  
Aia , jb   ij ab   a   i   K ia , jb
Bia , jb  K ia ,bj
 2 E xc
 ia jb     r1  a r1 
   *  3 3
  j r2 b r2 d r1d r2  K ia , jb
* HF
K ia , jb 
 r1  r1 
i

ia jb    i* r1  a r1   j r2 b* r2 d 3r1d 3r2
  1  
r12
K iaHF , jb  cx   ja ib  cx  0 for pure functional s

 2  
f I  EI  E0  0 r 0   d ia  a   i  FiaI 
2 2 1/ 2

3 3  ia 

where d ia   i r r  a* r d 3r
  

12
 Time Dependent DFT (TD-DFT) is widely used to calculate
molecular electronic excitation energies.

 Sufficiently accurate to be useful

 Sufficiently economical to apply to large molecules

 Not as accurate as highly correlated methods such as

CASPT2 or CC3

 Problems with Rydberg and Charge Transfer States, double

excitations, intensities
Pyrrole Rydberg States
CASPT2 B3LYP PBE0
A2 5.22 4.64 4.93
B1 5.87 5.32 5.61
A2 5.97 5.31 5.61
B1 5.97 5.54 5.85
B2 6.09 6.06 6.28
B1 6.40 5.87 6.16
A2 6.42 5.78 6.04
A2 6.51 6.00 6.14
B2 6.53 6.32 6.53
A1 6.54 6.11 6.41
MAE .46 .23

Basis set: aug-ccpvtz+R 14

14
 CASPT2 B3LYP

W1 A’’ 5.62 5.49

W2 A’’ 5.79 5.73

NV1 A’ 6.39 7.19

NV2 A’ 6.49 7.23

CT1 A’ 7.18 6.06

CT2 A’’ 8.07 6.24

15
•Problems with Rydberg and Charge Transfer States are due to
incorrect long range potentials (also problems with the
response kernel, self-interaction)
•Standard DFT functionals are too short range
•Modifications to the long range part of the exchange potential
are needed
•One approach is to use range-separated functionals constructed
from different short range (high density) and long range (low
density) forms. Short and long range components evaluated
using different techniques
•Other approaches are possible e.g. orbital dependent potentials

16
 Asymptotic Behavior of VXC(R).

lim Vxc r     I   max
1
Must be: r 
D.J. Tozer, N.C. Handy (1998)
r
This is not observed for all model potentials listed earlier (exponential asymptotic)
Solution (T. Yanai and K. Hirao group, 2004)

1 1  erf ( r12 ) erf ( r12 )

First term goes to Ex, while second calculated together
 
r12 r12 r12 with J
 
x
2
erf ( x)    2
exp t dt
(T. Yanai, D.P.Tew, N.C.Handy, 2004)  0

1 1  a  erf ( r12 ) a  erf ( r12 )

  CAM-B3LYP: a = 0.19; =0.46,  = 0.33
r12 r12 r12

17
 Pyrrole Rydberg States

CASPT2 B3LYP PBE0 CAM-

B3LYP
A2 5.22 4.64 4.93 5.10
B1 5.87 5.32 5.61 5.82
A2 5.97 5.31 5.61 5.83
B1 5.97 5.54 5.85 6.08
B2 6.09 6.06 6.28 6.38
B1 6.40 5.87 6.16 6.42
A2 6.42 5.78 6.04 6.37
A2 6.51 6.00 6.14 6.45
B2 6.53 6.32 6.53 6.72
A1 6.54 6.11 6.41 6.75
MAE .46 .23 .13
18
Basis set: aug-ccpvtz+R
 CASPT2 B3LYP HCTH-AC CAM-
B3LYP
W1 A’’ 5.62 5.49 5.43 5.61

W2 A’’ 5.79 5.73 5.70 5.84

NV1 A’ 6.39 7.19 6.87 7.06

NV2 A’ 6.49 7.23 6.98 7.36

CT1 A’ 7.18 6.06 5.16 6.74

CT2 A’’ 8.07 6.24 4.67 7.85

19
 Retinal Proteins Chromophores

Table 1. Excitation energies (eV) and oscillator strengths for 6-cis-11-cis PSB11.1
•CASPT2 calculations performed on geometry optimized with the state averaged CAS(12,12)/6-31G(d). All TD-DFT calculations employed geometry
optimized with B3LYP/6-31G(d).

S1 S2
Method Basis
E f E f
Experiment (in methanol) 2.79, 2.82
Experiment – solvent blue shift 2.41,2.43
CASPT2(12,12)1 6-31G(d) 2.41 3.52
TD-BP86 6-31G+(d) 2.14 0.79 2.70 0.73
TD-B3LYP 6-31G(d) 2.35 1.14 3.14 0.57
TD-B3LYP 6-31G+(d) 2.31 1.23 3.09 0.43
TD-B3LYP SV(P) 2.33 1.13 3.13 0.58
TD-CAMB3LYP 6-31G(d) 2.50 1.51 3.69 0.33
TD-CAMB3LYP 6-31G+(d) 2.46 1.51 3.65 0.33
TD-CAMB3LYP SV(P) 2.48 1.50 3.67 0.33
TD-CAMB3LYP TZV(P) 2.45 1.50 3.63 0.33 20
Retinal Proteins Chromophores

21
Retinal Proteins Chromophores

22
Retinal Proteins Chromophores

23
DFT references
1. W. Koch, M.C. Holthausen, A Chemist’s Guide to Density Functional Theory (Wiley-
VCH Verlag GmbH, 2001)
2. M.E. Casida in Recent Advances in Density Functional Methods, Part 1 (World
Scientific, Singapore, 1995)
3. M.E. Casida in Recent Developments and Applications of Modern Density Functional
Theory, Theoretical and Computational Chemistry, vol 4., ed. by J.M. Seminario
(Elsevier, Amsterdam, 1996).
4. Marques M.A.L. and Gross E.K.U. Annu. Rev. Phys. Chem 55, 427 (2004).

24
 Acknowledgements

 Fujitsu Company for financial support and giving us opportunity

to visit the beautiful country of Japan
 Hiro Hotta for constant help
 Professor Shinkoh Nanbu for invitation to the Kyushu University
 You, audience, for your attention

25