Janee

Intermolecular forces
and their representation in

electronic structure calculations
János Ángyán
1
2
Contents
1 Introduction 7
1.1 What are intermolecular forces? . . . . . . . . . . . . . . . . . . 7
1.2 Intermolecular potential . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 Cluster expansion of the interaction energy . . . . . . . . . . 9
1.4 Ranges of interaction . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.5 Calculation of intermolecular interaction energy . . . . . . . . 10
2 Experimental sources of intermolecular potentials 13

2.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Trouton’s rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2.1 Well-depth from Trouton’s rule . . . . . . . . . . . . . . 14
2.3 Theory vs. experiment . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4 Lattice energies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4.1 Ionic crystal . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4.2 Rare gas crystal . . . . . . . . . . . . . . . . . . . . . . . . 17
2.4.3 Three-body forces and the structure of rare gas solids 18
2.5 Equation of state – virial coeffiicients . . . . . . . . . . . . . . 18
2.5.1 van der Waals equation of state . . . . . . . . . . . . . . 18
2.5.2 Equations of state - generalities . . . . . . . . . . . . . . 19
2.5.3 Equations of state - ideal gas . . . . . . . . . . . . . . . . 20
2.5.4 Equations of state - real gas . . . . . . . . . . . . . . . . 21
2.5.5 Temperature dependence of the virial coefficient . . . 22
2.5.6 Third virial coefficient . . . . . . . . . . . . . . . . . . . . 23
2.6 VRT spectroscopy - water dimer . . . . . . . . . . . . . . . . . . 23
2.6.1 VRT spectroscopy - water dimer . . . . . . . . . . . . . . 24
3 Perturbation theory 25
3.1 Rayleigh-Schroedinger perturbation theory . . . . . . . . . . . 25
3.1.1 Iterative solution . . . . . . . . . . . . . . . . . . . . . . . . 26
3.1.2 Example: H-atom in electric field . . . . . . . . . . . . . 26
3.1.3 Recursion formulae for RSPT corrections . . . . . . . . 27
3.1.4 Explicit formulae at low orders . . . . . . . . . . . . . . . 28
3.2 Energy with the first-order wave function . . . . . . . . . . . . 29
3
3.2.1 Deformation energy and perturbation energy . . . . . 30
3.3 Dalgarno’s (2n+1) theorem . . . . . . . . . . . . . . . . . . . . . 30
3.4 Pertubational correction of expectation values . . . . . . . . 31
3.5 Partitioning method . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4 Perturbation theory for intermolecular interactions 35

4.1 Intermolecular RSPT . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2 Partition of the Hamiltonian . . . . . . . . . . . . . . . . . . . . 35
4.2.1 Longuet-Higgins form of the interaction operator . . 36
4.3 The polarization approximation . . . . . . . . . . . . . . . . . . 37
4.3.1 Second order energy in the polarization approximation 37
4.3.2 Induction energy . . . . . . . . . . . . . . . . . . . . . . . . 38
4.3.3 Dispersion energy . . . . . . . . . . . . . . . . . . . . . . . 39
4.3.4 Summary of second order terms . . . . . . . . . . . . . 40
4.3.5 Third order induction energy . . . . . . . . . . . . . . . . 40
4.4 Non-additive interactions . . . . . . . . . . . . . . . . . . . . . . 41
5 Beyond the polarization approximation 43

5.1 Example of H atom + proton system . . . . . . . . . . . . . . . 43
5.1.1 Convergence of the polarization approximation for H+ 2 43
5.2 Example of two H atoms . . . . . . . . . . . . . . . . . . . . . . . 45
5.2.1 Behaviour of the RSPT interaction energy . . . . . . . . 46
5.3 Symmetry failure . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.4 Claverie’s analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
5.5 Symmetrized RS perturbation theory . . . . . . . . . . . . . . . 47
5.5.1 Separation of exchange contributions . . . . . . . . . . 49
5.6 Symmetry adapted perturbation theory (SAPT) . . . . . . . . 49
6 Molecules in electrostatic field 53

6.1 Electrostatic interaction Hamiltonian . . . . . . . . . . . . . . 53
6.2 Molecule in a uniform electric field . . . . . . . . . . . . . . . . 53
7 Multipole expansion – Cartesian formalism 55

7.1 Potential of a charge distribution . . . . . . . . . . . . . . . . . 55
7.2 Multipole moments . . . . . . . . . . . . . . . . . . . . . . . . . . 57
7.2.1 Traceless (Buckingham) multipole moments . . . . . . 58
7.2.2 Explicit form of higher multipole moments . . . . . . . 58
7.2.3 Translation of multipole moments . . . . . . . . . . . . . 59
7.3 Electric field, field gradient . . . . . . . . . . . . . . . . . . . . . 59
7.4 Energy of a charge distribution in non-uniform field . . . . . 59
7.5 Multipole interaction energy . . . . . . . . . . . . . . . . . . . . . 60
7.6 Calculation of cartesian interaction tensors . . . . . . . . . . 60
7.7 Example: Structure of the HX dimers . . . . . . . . . . . . . . 62
4
8 Multipole expansion – spherical harmonics formalism 65
8.1 Potential of a charge distribution . . . . . . . . . . . . . . . . . 65
8.1.1 Potential outside of the charge distribution . . . . . . . 67
8.1.2 Potential in the overlap region . . . . . . . . . . . . . . . 68
8.2 Exact multipolar part of the potential . . . . . . . . . . . . . . 69
8.3 Buehler-Hirschfelder bipolar expansion . . . . . . . . . . . . . 70
8.4 Spherical harmonics expansion of the interaction . . . . . . 70
8.5 Interaction tensors between local frame multipoles . . . . . 72
9 Long range perturbation theory 75

9.1 Induction energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
9.2 Multipole polarizabilities . . . . . . . . . . . . . . . . . . . . . . . 77
9.2.1 Translational invariance . . . . . . . . . . . . . . . . . . . 78
9.2.2 Multipole polarizabilities – spherical tensors . . . . . . 78
9.3 Distance dependence of the induction energy . . . . . . . . . 79
9.4 Convergence of the multipolar induction energy . . . . . . . . 79
9.5 Dispersion energy . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
9.5.1 Dipolar dispersion energy . . . . . . . . . . . . . . . . . . 81
9.5.2 Approximate dispersion formulae . . . . . . . . . . . . . 83
9.5.3 Solutions of the H + H dispersion problem . . . . . . . 84
9.6 Convergence of the 1/R-expansion . . . . . . . . . . . . . . . . 85
10 Supermolecule method 86
10.1 Overview of the supermolecule approach . . . . . . . . . . . . 86
10.2 Choice of the method . . . . . . . . . . . . . . . . . . . . . . . . . 86
10.3 Choice of the basis set . . . . . . . . . . . . . . . . . . . . . . . . 87
10.4 BSSE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
10.4.1 Nitrogen dimer . . . . . . . . . . . . . . . . . . . . . . . . . 88
10.4.2 Basis incompleteness correction . . . . . . . . . . . . . . 88
10.4.3 BSSE-free geometry optimization . . . . . . . . . . . . . 89
10.5 Size consistency problem . . . . . . . . . . . . . . . . . . . . . . 90
10.5.1 Size-consistency of MBPT methods . . . . . . . . . . . . 90
10.5.2 Other size-consistent methods . . . . . . . . . . . . . . . 91
11 DFT and intermolecular forces 92

11.1 DFT description of intermolecular complexes . . . . . . . . . 92
11.2 Gordon-Kim method . . . . . . . . . . . . . . . . . . . . . . . . . 92
11.3 Is dispersion missing from DFT? . . . . . . . . . . . . . . . . . 93
11.4 Comparison of some functionals . . . . . . . . . . . . . . . . . 94
11.5 Modern theories of intermolecular forces in DFT . . . . . . . 94
11.5.1 Adiabatic connection formula . . . . . . . . . . . . . . . . 94
11.5.2 Asymptotic dispersion energy functionals . . . . . . . 95
11.5.3 Dispersion functional from the electron gas . . . . . . 95
5
11.5.4 Seamless dispersion energy functionals . . . . . . . . . 96
11.5.5 Goerling-Levy perturbation theory . . . . . . . . . . . . 97
11.5.6 DFT + E(disp) method . . . . . . . . . . . . . . . . . . . . 98
11.5.7 DFT-SAPT method . . . . . . . . . . . . . . . . . . . . . . . 98
6
Chapter 1
Introduction
1.1 What are intermolecular forces?

N2 crystal
◦ why are the N2 molecules arranged
this way?
◦ what are the optimal intermolecular

distances?
◦ what is the physical origin of

the interactions that govern the
structure?
Intermolecular interactions
◦ fragments (subsystems) are well-defined and preserve their identity

◦ usually between closed shell molecules, ions or atoms
◦ much weaker than chemical bonding
◦ long range
7
1.2 Intermolecular potential
At infinitely large distance, there is no interaction:
the energy is the sum of isolated atoms
Etot (∞) = EA + EB (1.1)
At an arbitrary, finite R distance the total energy
Etot (R) = EA + EB + U (R) (1.2)

Ar2 dimer
has an interaction-dependent contribution,
U (R),the interaction energy or intermolecular pair-potential
U (R) = Etot (R) − Etot (∞) = Etot (R) − EA − EB (1.3)
The intermolecular pair-potential is the work to bring together the two systems from
infinity to the distance R, against the intermolecular force, F (R)
Z ∞
dEtot dU
U (R) = F (r)dr where F (R) = − =− (1.4)
R dR dR
◦ σ is the hard core

radius
U(R)
σ Rm
◦ ε is the well depth
−ε
◦ repulsive for R < Rm
◦ attractive for R > Rm

F(R)
◦ for σ < R < Rm nega-

Rm tive, but repulsive
8
For polyatomic subsystems the intermolecular interaction energy
U (R, ω; q A , q B )
depends on the relative position and orientation {R, Ω} and on the internal coor-
dinates {q A , q B }
Even if we neglect the intramolecular degrees of freedom, the PES is six-dimensional.
1.3 Cluster expansion of the interaction energy

The interaction energy can be decomposed to the sum of one-, two-, three-, etc. body
contributions
X 1X 1 X 1 X
UN = Ui + Uij + Uijk + Uijkl + . . . (1.5)
2 3! 4!
i ij ijk ijkl
◦ one-body: geometrical deformation of the monomers
◦ two-body: dominating term (additivity)
◦ three-body, etc.
Effective (system-dependent) pair-potential

X 1 X eff
UN = Ui + Uij (N ) (1.6)
2
i ij
may be quite different from the pure pair-potential.

Attention: a pair-potential, obtained from condensed-phase experimental data, is usually
not valid for binary complexes.
9
1.4 Ranges of interaction
Range short intermediate long

Subsystems difficult quite easy easy to identify
Overlap strong weak no
Description SM/PT
U(R)
SM PT
1.5 Calculation of intermolecular interaction en-

ergy
Two strategies
◦ supermolecule
U = ∆E = E(AB) − E(A) − E(B) (1.7)
difference of the total energies (large numbers)
– very general
– does not depend on the intensity of the interaction
– difficult to interpret in terms of the monomer properties
– energy components may have large errors with respect to the interaction
energy, e.g. (N2 )2
E(HF) = −109 a.u.
E(corr) = −1 a.u.
∆E = 0.0004 a.u.
10
◦ perturbation theory
∞
X
Ĥ = ĤA + ĤB + λV̂ U = ∆E = λn ∆E (n) (1.8)
n=1
expanded in power series of the strength of interaction
– leads directly to the interaction energy

– based on the monomer wavefunction/properties
– easy to interpret (physical insight, decomposition)
– provides basis of approximate analytical energy expressions
– is the expansion convergent?
11
12
Chapter 2
Experimental sources of
intermolecular potentials
2.1 Overview
◦ thermodynamical properties: heat of vaporization (Trouton’s rule)
◦ crystal structures
– ionic crystals
– rare gas solids
◦ physico-chemical properties: bulk modulus, phonon spectrum, etc.

◦ virial coefficients of real gases
◦ viscosity, thermal conductivity (collision integrals
◦ spectroscopy: VRT (vibration-rotation tunneling)
2.2 Trouton’s rule
Empirical relationship between enthalpy of vaporization, ∆Hvap and boiling point,

Tb at atmospheric pressure
∆Hvap ≈ 10RTb (2.1)
This observation can be explained by the fact that at T b the change in Gibbs free
energy is zero
∆Hvap = Tb ∆Svap (2.2)
13
Table 2.1: Pair potential well-depths from Trouton’s rule
Tb /K n (20Tb /n)/K (ε/kB )/K ε/kB /kJ mol−1 εlit.

He 4.2 12 7 11 0.09
Ar 87.0 12 145 142 1.18
Xe 166.0 12 277 281 2.34
CH4 111.5 12 186 180-300 1.5-2.50
H2 O 373.2 4 1866 ≈2400 ≈20
Approximate the entropy change by the liquid/gas volume change:
∆Svap = R ln (Vg /Vl ) ≈ R ln 1000 ≈ 7R (2.3)
The remaining contribution of 3R is due to liquid structure.
2.2.1 Well-depth from Trouton’s rule

The latent heat of evaporation can be approximated by the energy required to
separate the liquid to its constituents is ε (neglect zp energy). The total energy for
N molecules, each having n neighbours is
1
∆Hvap ≈ 10RTb ≈ NA nε
2
An estimation of of the well-depth, ε is given in terms of Tb
ε/kB ≈ 20Tb /n (2.4)
14
2.3 Theory vs. experiment
EXPERIMENT
{Dexp} {Dtheor}
model model
Uexp Utheor
QUANTUM CHEMISTRY
2.4 Lattice energies

2.4.1 Ionic crystal
◦ NaCl crystallizes in fcc lattice
◦ lattice parameter a = 5.64Å
◦ ions are in (0,0,0) and ( 12 , 12 , 12 )
◦ lattice energy Ulatt = −764.4 kJ/mol
Cohesion energy is the sum of the Coulomb (Madelung) and repulsion energy:
Ucoh = UC + UR (2.5)
where UC is a lattice sum of 1/r interactions, UR is the so-called Born-Mayer poten-
tial: X X
UC = Q2 (±)j rj−1 UR = B exp −ri /ρ = 6Be−a/2ρ (2.6)
j i
The lattice sum (Madelung energy) is conditionally convergent, so special techniques

(e.g. Ewald summation) are needed to evaluate them. For a NaCl lattice, the energy
15
can be expressed using the Madelung constant, α = 1.7476,
2α 2 · 1.7476
UC = −Q2 = −1389.9 × kJ/mol = −861.3kJ/mol (2.7)
a 5.64
The Born-Mayer potential has two unknown parameters, B and ρ.
◦ Repulsion energy
exp
UR = Ucoh − UC = −764.4 + 861.3 = 96.9kJ/mol (2.8)
◦ Condition of equilibrium in the minimum of the lattice energy

∂Ucoh 2α 6B −a/2ρ Uc UR
= −Q2 2 − e =− − =0 (2.9)
∂a a 2ρ a 2ρ
Effective ”ion radius” is obtained from the repulsion and Madelung energies
1 UR
ρ=− · a = 0.3164 (2.10)
2 UC
◦ The B parameter of the Born-Mayer potential is
UR
B= −a/2ρ
= 1.168 · 105 kJ/mol (2.11)
6e
The quality of this potential can be checked by calculating the bulk modulus
∂ 2 Ucoh
K=V (2.12)
∂V 2
where the volume of 1 mole NaCl is V = NA a3 /4. The volume derivative can be
calculated as lattice-parameter derivative:
−1
∂ ∂V ∂
= (2.13)
∂V ∂a ∂a
Bulk modulus for the NaCl structure
2 2
N A a3 4 ∂ 2 Ucoh

4 ∂ Ucoh
K= = (2.14)
4 3NA a2 ∂a2 9NA a ∂a2
Second derivative of the cohesion energy
∂ 2 Ucoh ∂ Q2 2α

1 ∂UR 2 1
= − − = UC + UR (2.15)
∂a2 ∂a a2 2ρ ∂a a2 4ρ2
Theoretical bulk modulus

4 2UC UR
K= 2
+ 2 = 14.8016 kJ/mol/Å3 (2.16)
9NA a a 4ρ
Conversion to gigapascal 1 kJ/mol/Å3 = 1.667 GPa, i.e.
K theor = 24.57 GPa K exp = 24 GPa
16
2.4.2 Rare gas crystal
◦ Ar crystallizes in fcc lattice
◦ lattice parameter a = 5.3109Å
◦ ions are in (0,0,0) and ( 12 , 12 , 12 )
◦ lattice energy Ulatt = −8.4732 kJ/mol

(after zpe correction)
Cohesion energy is the sum of pair potentials

NX
Ucoh = ULJ (rij ) (2.17)
2 ij
where N = 4, number of atoms in the unit cell, and ULJ is the Lennard-Jones
potential:
σ 12 σ 6
ULJ (r) = 4ε − (2.18)
r r
Lattice sums can be calculated from the
√ number of neighbours at the multiples of
the nearest-neighbour distance d = a/ 2:
shell 1 √2 √3 4 √5 i
r d 2d 3d 2d 5d fi · d
mi 12 6 24 12 24 mi
General form of the lattice sums
shells n shells
X σ X σ n σ n
mi = mi fi−n = pn (2.19)
i
fi d i
d d
Lattice sum for the 6- and 12-potentials
Pi 1 5 fcc(∞) hcp (∞)

p12 = P i mi fi−12 12 12.13114 12.13188 12.13229
p6 = i mi fi−12 12 14.01839 14.45392 14.45489
Cohesion energy of the rare-gas lattice in terms of lattice sums
σ 6
Nε σ 12
Ucoh = p12 − p6 (2.20)
2 d d
Stability condition
∂Ucoh σ 12 1 σ 6 1
= −12p12 + 6p6 =0 (2.21)
∂d d d d d
17
yields the σ parameter in the function of d, nearest-neighbour distance
r
2p12
d= 6 · σ = 1.09026σ (2.22)
p6
Lattice constant a = 1.54186σ.
Cohesion energy can be expressed in the function of ε.
" 2 #
Nε p6 p6
Ucoh = p12 − p6 (2.23)
2 2p12 2p12
can be expressed in the function of ε
N p26
Ucoh = − ε = −2.1525 N ε (2.24)
8 p12
Experimental lattice parameters from ”best” Lennard-Jones potentials:
ε σ a Ucoh a Ucoh
Ne
Ar
Kr
Xe
2.4.3 Three-body forces and the structure of rare gas solids
2.5 Equation of state – virial coeffiicients

2.5.1 van der Waals equation of state
Equation of state for an ideal gas (noninteracting point masses)
P V = RT (2.25)
where P is the pressure, V is molar volume, and R = 8.31J K −1 mol−1 is the gas
constant.
18
The standard molar volume V s (1
atm, T=273.15 Ko of real gases
are different from the ideal value
(22414 cm3 ). Considerable devi-
ations can be observed in the be-
haviour of the compression factor,
z = PRTV .
These deviations can be accounted for by the van der Waals equation of state, by
introducing two empirical constants, a and b

a
P+ (V − b) = RT (2.26)
V
◦ excluded volume in a binary collision 43 πσ 3
2
b = πNA σ 3 (2.27)
3
◦ reduction of pressure due to intermolecular attraction: the number of binary

2
interactions is proportional to the square of the density, (a/V )
Virial equation of state – power series of (1/V )
B2 B3 B4
z =1+ + 2 + 3 + ... (2.28)
V V V
2.5.2 Equations of state - generalities

The total energy is a function of xα external parameters
E = E(x1 , x2 , . . . xn ) (2.29)
Define generalized forces corresponding to external parameters
∂E
Xα = − (2.30)
∂xα
19
In statistical mechanics the equations of state describe the relationship of the exter-
nal parameters, the generalized forces and the temperature:
∂ ln Z
X α = kT (2.31)
∂xα
where Z is the partition function. For the special case of volume (xα = V ) and
pressure (X α = P )
∂ ln Z
P = kT (2.32)
∂V
The semi-classical partition function
ZZ Z
1
Z= ... e−(K+U )/kT d3 p1 d3 p2 . . . d3 pN d3 r1 d3 r2 . . . d3 rN (2.33)
N !h3N
is a product of two terms, depending on the kinetic and the potential energy, re-
spectively:
1 X 2 1X
K= pj U= ujk (2.34)
2m 2 jk
ZZ Z ZZ Z
1 −K/kT 3
Z= ... e 3 3
d p1 d p2 . . .d pN . . . e−U/kT d3 r1 d3 r2 . . .d3 rN (2.35)
N !h3N
| {z }| {z }
(2πmkT )3N/2 ZU
Partition function 3N/2

1 2πmkT
Z= · ZU (2.36)
N! h2
2.5.3 Equations of state - ideal gas

For an ideal gas uij → 0 (or if T is high kT → ∞)
e−U/kT → 1 ZU = V N (2.37)
and the corresponding partition function

1 3 3 2πm
ln Zid = ln + N ln V + ln kT + ln (2.38)
N! 2 2 h2
Equation of state for an ideal gas

∂ ln Zid N kT
P = kT = (2.39)
∂V V
or
P V = N kT (2.40)
20
2.5.4 Equations of state - real gas
For a real gas with low number density (n = N/V small) an approximate expression
can be obtained for the configurational partition function.
Take the average potential energy with β = 1/kT , which satisfies the following
relationship
U e−βU d3 r1 d3 r2 . . . d3 rN
RRR
∂
U = RRR −βU 3 3 = − ln ZU (2.41)
e d r1 d r2 . . . d3 rN ∂β
The logarithm of the partition function is
Z β
ln ZU (β) = N ln V − U (β 0 )dβ 0 (2.42)
0
In a low-density system the U , the average potential energy of 1/2N (N −1) molecule
pairs is equal N 2 /2-fold of the average potential energy of an arbitrary pair of
molecules:
1 1
U = N (N − 1) ≈ N 2 u (2.43)
2 2
The average potential energy of a pair of molecules
u(R)e−βu(R) d3 R
R Z
∂
u = R −βu(R) 3 =− ln e−βu(R) d3 R (2.44)
e dR ∂β
Since u(R) = 0 almost everywhere, excepted the small distances, it is worthwhile to

transform the integral as
Z Z
−βu(R) 3
1 + e−βu(R) − 1 d3 R = V + I(β)

e d R= (2.45)
The quantity in parentheses is the Mayer-function f (R) = e−βu(R) − 1, and its

integral over the intermolecular separation, R is
Z ∞
e−βu(R) − 1 R2 dR

I(β) = 4π (2.46)
0
Calculate the average potential energy for a pair of molecules

∂ ∂ I(β) 1 ∂I(β)
u= ln [V + I(β)] = ln V + ln 1 + ≈− · (2.47)
∂β ∂β V V ∂β
which is substituted in the expression of the configurational partition function
1 N2
ln ZU (β) = N ln V + [I(0) − I(β)] (2.48)
2 V
The equation of state is
P ∂ ln Z ∂ ln Z N 1 N2
= = = − I(β) (2.49)
kT ∂V ∂VU V 2V2
21
or Z ∞
PV PV 4π N
e−βu(R) − 1 R2 dR

= =1− (2.50)
RT N kT 2 V 0
This is the first term of the virial equation of state:
∞
PV X Bn+1
z= =1+ (2.51)
RT n=1
Vn
and the virial coefficient is

Z ∞
e−u(R)/kT − 1 R2 dR

B2 (T ) = −2πN (2.52)
0
2.5.5 Temperature dependence of the virial coefficient
T \R small (u > 0) large (u < 0)

low
+ –– B2 < 0
kT < u0
high
+ – B2 > 0
kT > u0
22
2.5.6 Third virial coefficient
B3 (T ) = B3add + ∆B3 (2.53)
For a strictly additive potential
ZZZ
add N 3
B3 = − 8π f12 f13 f23 r12 r13 r23 dr12 dr12 dr12 (2.54)
3
The non-additivity correction depends on the 3-body potential ∆U3
ZZZ
N 3
e−∆U3 /kT − 1 e−U3 /kT r12 r13 r23 dr12 dr12 dr12

∆B3 = − 8π ?? (2.55)
3
2.6 VRT spectroscopy - water dimer
Calculated splitting due to tunneling is very sensitive to the height and shape of the
barrier – stringent test of the PES.
23
2.6.1 VRT spectroscopy - water dimer
Experimental vs. calculated VRT splittings with different water-water potentials.

None of the 14 potentials was found to be fully satisfactory.
Fellers, Braly, Saykally, Leforestier, J. Chem. Phys. 110 (1999) 6306)
24
Chapter 3
Perturbation theory
3.1 Rayleigh-Schroedinger perturbation theory

We are looking for the ground state solution of the problem
Ĥψ = (Ĥ0 + V̂ )ψ = Eψ (3.1)
and we already know the (exact) solution of the zero-order problem
Ĥ0 ϕ0 = E0 ϕ0 (3.2)
Let us write ∆E = E − E0 , the energy correction
(Ĥ0 − E0 )ψ = (∆E − V̂ )ψ (3.3)
and in order to fix the phase of ψ, impose the intermediate normalization
hϕ0 |ψi = 1 (3.4)
Introduce the reduced resolvent operator as
R̂0 = (1 − |ϕ0 ihϕ0 |)(Ĥ0 − E0 )−1 (3.5)
which can be regarded as the inverse of the operator Ĥ0 −E0 in the space of functions
orthogonal to ϕ0
R̂0 (Ĥ0 − E0 ) = 1 − |ϕ0 ihϕ0 | (3.6)
Multiplying the Schrödinger equation by R̂0
R̂0 (Ĥ0 − E0 )ψ = R̂0 (∆E − V̂ )ψ (3.7)
using the definition of the resolvent and the intermediate normalization an equation
is obtained for the wave function
ψ = ϕ0 + R̂0 (∆E − V̂ )ψ (3.8)
25
After multiplication of the Schrödinger equation by hϕ0 |
hϕ0 |(Ĥ0 − E0 )|ψi = hϕ0 |(∆E − V̂ )|ψi (3.9)
we get the energy correction
∆E = hϕ0 |V̂ |ψi (3.10)
3.1.1 Iterative solution
These equations can be solved iteratively

∆En = hϕ0 |V̂ |ψn−1 i
(3.11)
ψn = ϕ0 + R̂0 (∆En − V̂ )ψn−1
To the lowest orders of iteration we find by using ψ0 = ϕ0 and R̂0 ϕ0 = 0

∆E1 = hϕ0 |V̂ |ψ0 i
ψ1 = ϕ0 − R̂0 V̂ ψ0
∆E2 = hϕ0 |V̂ |ψ1 i = hϕ0 |V̂ |ψ0 i − hϕ0 |V̂ R̂0 V̂ |ψ0 i = ∆E1 − hϕ0 |V̂ R̂0 V̂ |ψ0 i
(3.12)
ψ2 = ϕ0 − R̂0 (∆E2 − V̂ )ψ1
= ϕ0 − R̂0 (hϕ0 |V̂ |ψ0 i − hϕ0 |V̂ R̂0 V̂ |ψ0 i − V̂ )(ϕ0 − R̂0 V̂ ψ0 )
= ϕ(1) − R̂0 (V̂ − ∆E2 )R̂0 V̂ ϕ0
3.1.2 Example: H-atom in electric field

Hamiltonian of the H-atom in an electric field, Fz
1 1
Ĥ = Ĥ0 + ẑFz where Ĥ0 = − ∆ − (3.13)
2 r
The Hamiltonian of the isolated H-atom has
the lowest eigenvalue E0 = − 12 and eigenfunction ϕ0 = √1 e−r .
π
First iteration in the energy yields zero
∆E1 = Fz hϕ0 |ẑ|ψ0 i = µz · Fz = 0 (3.14)
First iteration in the wave function leads to the equation
ψ1 = ϕ0 − R̂0 V̂ ϕ0

(Ĥ0 − E0 )ψ1 = − V̂ − ∆E1 ϕ0

1 1 1 1
− ∆− + ψ1 = − √ Fz z e−r (3.15)
2 r 2 π
26
which has the solution
Fz r
ψ1 = − √ z − 1 e−r (3.16)
π 2
Second energy iteration
9
∆E2 = Fz hϕ0 |ẑ|ψ1 i = − · Fz2 (3.17)
4
This leads to a development of the energy in the powers of Fz
9
E = E0 + ∆E2 = E0 − · F2 (3.18)
4 z
the dipole polarizability is
∂2E

9
α=− = a.u. (3.19)
∂Fz2 Fz =0 2
3.1.3 Recursion formulae for RSPT corrections

Assuming that the iteration process converges, the exact wave function and energy
can be expanded in power series of a perturbation parameter λ
(Ĥ0 + λV̂ )ψ = ∆Eψ (3.20)
as
∞
X ∞
X
n (n)
∆E = λ ∆E and ψ= λn ψ (n) (3.21)
n=1 n=0
Substitute the series expansions in
∆E = hϕ0 |V̂ |ψi and ψ = ϕ0 + R̂0 (∆E − V̂ )ψ
leading to
X X
λn ∆E (n) = λm+1 hϕ0 |V̂ |ψ (m) i
n=1 m=0
X X
!
X (3.22)
n (n) m (m) k (k)
λ ψ = ϕ0 + R̂0 λ ∆E − λV̂ λ ψ
n=1 m=1 k=1
and collect terms of the same power to obtain the general recursion formulae
∆E (n) = hϕ0 |V̂ |ψ (n−1) i

n−1
(n) (n−1)
X (3.23)
ψ = −R̂0 V̂ ψ − ∆E (k) R̂0 ψ (n−k)
k=1
27
3.1.4 Explicit formulae at low orders
The reduced resolvent has the spectral resolution
X |ϕ0 ihϕ0 |
R̂0 = (3.24)
k6=0
Ek − E0
◦ First order
∆E (1) = hϕ0 |V̂ |ϕ0 i (3.25)
X hϕk |V̂ |ϕ0 i

ψ (1) = −R̂0 V̂ ϕ0 = − ϕk (3.26)
k6=0
Ek − E0
The wave function corrections are often expressed in terms of the expansion
(n)
coefficients ck = hϕk |ψ (n) i on the basis of the eigenfunctions of the zero order
Hamiltonian
(1)
X hϕk |V̂ |ϕ0 i
ck = − (3.27)
k6=0
Ek − E0
◦ Second order
Energy
E (2) =hϕ0 |V̂ |ψ (1) i = −hϕ0 |V̂ R̂0 V̂ |ϕ0 i

X hϕ0 |V̂ |ϕk ihϕk |V̂ |ϕ0 i (3.28)
=−
k6=0
Ek − E0
Wave function
ϕ(2) = R̂0 V̂ R̂0 V̂ ψ (1) + R̂0 E (1) R̂0 V̂ ψ (0) =

(3.29)
= R̂0 V̂ R̂0 V̂ ϕ0 − R̂0 hV̂ iR̂0 V̂ ϕ0 = R̂0 V R̂0 V̂ ϕ0
where we used the definitions
hV̂ i = hϕ0 |V̂ |ϕ0 i and V = V̂ − hV̂ i (3.30)
◦ Third order
∆E (3) = hϕ0 |V̂ |ψ (2) i = −hϕ0 |V̂ R̂0 V R̂0 V̂ |ϕ0 i

X X hϕ0 |V̂ |ϕk ihϕk |V |ϕl ihϕl |V̂ |ϕ0 i (3.31)
=
k6=0 l6=0
(Ek − E0 )(El − E0 )
28
Summary of energy corrections
∆E (1) = hV̂ i (3.32)

∆E (2) = hV̂ R̂0 V̂ i (3.33)
∆E (3) = hV̂ R̂0 V R̂0 V̂ i (3.34)

∆E (4) = hV̂ R̂0 V R̂0 V − hV R̂0 V i R̂0 V̂ i (3.35)
∆E (5) = hV̂ R̂0 (V R̂0 V R̂0 V − hV R̂0 V R̂0 V i
− V R̂0 hV R̂0 V i − hV R̂0 V iR̂0 V )R̂0 V̂ i (3.36)
3.2 Energy with the first-order wave function

The energy (Rayleigh quotient)
hψ|Ĥ + λV̂ |ψi

E= (3.37)
hψ|ψi
with the first order wave function, ψ = ϕ0 + λψ (1) = ϕ0 − λV̂ R̂0 ϕ0
hϕ0 − λϕ0 V̂ R̂0 |Ĥ + λV̂ |ϕ0 − λV̂ R̂0 ϕ0 i

E= (3.38)
1 + λ2 hϕ0 V̂ R̂0 |R̂0 V̂ ϕ0 i
After expanding the denominator and using that R̂0 Ĥ0 ϕ0 = 0
E = E0 + λE (1) − 2λ2 hϕ0 |V̂ R̂0 V̂ |ϕ0 i+

+ λ2 hϕ0 |V̂ R̂0 (Ĥ + λV̂ )R̂0 V̂ |ϕ0 i ×

(3.39)
× 1 − λ2 hϕ0 V̂ R̂0 |R̂0 V̂ ϕ0 i

and collecting terms of the same order
E = E0 + λE (1) −
− λ2 2hϕ0 |V̂ R̂0 V̂ |ϕ0 i−
(3.40)
− λ2 hϕ0 |V̂ R̂0 (Ĥ − E0 )R̂0 V̂ |ϕ0 i
+ λ3 hϕ0 |V̂ R̂0 (V̂ − E (1) )R̂0 V̂ |ϕ0 i
Use that E (1) = hV̂ i and R̂0 (Ĥ − E0 ) = 1 − |ϕ0 ihϕ0 | and obtain the previously
derived 3rd order energy expression
E = E0 + hϕ0 |V̂ |ϕ0 i − hϕ0 |V̂ R̂0 V̂ |ϕ0 i + hϕ0 |V̂ R̂0 (V̂ − hV̂ i)R̂0 V̂ |ϕ0 i (3.41)
29
This result can be generalized: the nth order wave function determines the (2n-1)th
order energy expression, provided the normalization is taken into account.
The perturbational energy is not a upper bound to the exact energy, but the
Rayleigh-quotient is an upper bound.
3.2.1 Deformation energy and perturbation energy

Up to second order the Rayleigh quotient can be written as a sum of two terms, the
expectation value of the zero order Hamiltonian
hϕ − λϕV̂ R̂0 |Ĥ|ϕ − λV̂ R̂0 ϕi (2)

= E0 + λ2 hϕ|V̂ R̂0 V̂ |ϕi = E0 + ∆Edef (3.42)
1+ λ2 hϕV̂ R̂0 |R̂0 V̂ ϕi
and the expectation value of the perturbation operator
hϕ − λϕV̂ R̂0 |λV̂ |ϕ − λV̂ R̂0 ϕi (2)

= λhϕ|V̂ |ϕi − λ2 2hϕ|V̂ R̂0 V̂ |ϕi = E (1) + ∆Estab
1+ λ2 hϕV̂ R̂0 |R̂0 V̂ ϕi
(3.43)
The second order correction to the expectation value of the perturbation is twice the
second order energy correction and it is twice the energy raise of the wave function
due to the deformation of the wave function.
(2) (2)
∆Estab = −2∆Edef (3.44)
(2) (2) 1 (2)

∆E (2) = ∆Estab − ∆Edef = ∆Estab (3.45)
2
3.3 Dalgarno’s (2n+1) theorem

Although the recursion formulae would suggest that we need the (n − 1) order wave
function correction to calculate the n-th order energy correction, as the previous
example showed, there is a much stronger statement:
In order to obtain the (2n + 1) order energy correction, it is sufficient to
know the n-th order wave function correction.
Wave function
n
X
ψ= ψ (k) + O(λn+1 ) (3.46)
k
Energy
n X
X n
E= hψ (k) |Ĥ|ψ (l) i + O(λ2n+2 ) (3.47)
k l
30
Transformation formulae (Löwdin)
∆E (2n) = hϕ0 |V̂ |ψ (2n−1) i

n X
n
X (3.48)
= hψ (n) |V̂ |ψ (n−1) i − ∆E (2n−k−l) hψ (k) |ψ (l) i
k=1 l=1
∆E (2n+1) = hϕ0 |V̂ |ψ (2n) i

n X
n
(n) (n)
X (3.49)
= hψ |V̂ |ψ i− ∆E (2n+1−k−l) hψ (k) |ψ (l) i
k=1 l=1
Computational difficulties increase significantly at 2n orders.
3.4 Pertubational correction of expectation val-

ues
Expectation value of an arbitrary Hermitian operator, B̂ with a first order wave
function ψ ≈ (1 + R̂0 V̂ )ϕ
hϕ − λϕV̂ R̂0 |B̂|ϕ − λV̂ R̂0 ϕi

B= (3.50)
1 + λ2 hϕV̂ R̂0 |R̂0 V̂ ϕi
Expanding up to first order in the perturbation
hϕ|B̂ + B̂ R̂0 V̂ + V̂ R̂0 B̂ + V̂ R̂0 (B̂ − B0 )R̂0 V̂ |ϕi = B0 + ∆B (1) (3.51)
Suppose that the perturbation is of the form V̂ = aÂ = Âa† , then
∆B (1) = a† hÂR̂0 B̂ + B̂ R̂0 Âia = a† K(ÂB̂)a (3.52)
where K(ÂB̂) = hÂR̂0 B̂ + B̂ R̂0 Âi is the linear response function.
3.5 Partitioning method

Let us consider an orthonormal basis {ϕk } divided into two subsets, A and B,
containing nA and nB functions, respectively. The result, obtained by using only
nA functions can be improved by adding the nB extra functions, i.e. the secular
equations are partitioned as
AA
H AB
A A
H c c
BA BB B =E B (3.53)
H H c c
31
which is equivalent to the system of equations
H AA cA + H AB cB =EcA
(3.54)
H BA cA + H BB cB =EcB
A formal solution can be obtained by expressing cB from the second equation
cB = (E1BB − H BB )−1 H BA cA (3.55)
and inserting it into the first one leading to
H eff cA = EcA (3.56)
The effective Hamiltonian is an nA × nA matrix
H eff = H AA + H AB (E1BB − H BB )−1 H BA (3.57)
including implicitly the effect of the nB other functions. The effective Hamiltonian
depends on the yet unknown energy, E, therefore the solution should be obtained
by iteration.
Take the case nA = 1, cA = c1 = 1, then H eff is just a single element
E = H eff = f (E) (3.58)
Insert as a first approximation E = H11 and expand the inverse matrix, using
(I + ∆)−1 = I − ∆ + ∆2 leading to second order in the off-diagonal elements
X H1κ Hk1
E = H11 + (κ > 1) (3.59)
κ>1
H11 − Hκκ
The expansion coefficients in the cB vector (c.f. above)
cB = (E1BB − H BB )−1 H BA cA (3.60)
can be approximated similarly as

Hκ1
cκ = (3.61)
H11 − Hκκ
Analogous to RSPT, but
◦ basis is finite
◦ no a priori separation of the Hamiltonian is necessary
◦ complete set of eigenfunctions not needed
◦ analogies with RSPT to handled with caution
32
To make the connection with RSPT clear, choose the eigenfunctions of Ĥ0 as basis
(0)
Ĥ0 ϕk = Ek ϕk (3.62)
and the relevant matrix elements are

(0)
H11 = hϕ1 |Ĥ0 + V̂ |ϕ1 i = E1 + hϕ1 |V̂ |ϕ1 i (3.63)
H1k = hϕ1 |Ĥ0 + V̂ |ϕk i = hϕ1 |V̂ |ϕk i (3.64)
By application of the previously derived results we obtain for the perturbational

corrections to the i-th state
(0)
X hϕi |V̂ |ϕk ihϕk |V̂ |ϕi i
Ei = Ei + hϕi |V̂ |ϕi i + (0) (0)
(3.65)
k(6=i) Ei − Ek
X X hϕk |V̂ |ϕi i

ψi = ϕi + cik ϕk = ϕi − (0) (0)
ϕk (3.66)
k(6=i) k(6=i) Ei − Ek
(Quasi)-degenerate cases can be handled by taking(nA > 1).

In non-orthogonal basis sets the matrix equations are
AA
H AB
AA
M AB
A A
H c M c
=E (3.67)
H BA H BB cB M BA M BB cB
where M is the overlap (metric) matrix of the basis. The effective equation is
H eff cA = EM AA cA (3.68)
with the effective Hamiltonian
H AA + (H AB − EM AB )(E1BB − H BB )−1 (H BA − EM BA ) (3.69)
and the energy to second order is
H00 X [Hk0 − Mk0 (H00 /M00 )]2

E= + (3.70)
M00 k
[H00 /M00 − (Hkk /Mkk )]
33
34
Chapter 4
Perturbation theory for

intermolecular interactions
4.1 Intermolecular RSPT

The solution of the Schrödinger equation for the non-interacting complex

A B
Ĥ + Ĥ ϕk = E0 ϕk (4.1)
can be obtained from the solutions of the isolated subsystem Schrödinger equations
Ĥ A ψaA = EaA ψaA Ĥ B ψbB = EbB ψbB (4.2)

!
X X Zα X 1 X Z α Zα 0
Ĥ X = T̂i − + + (4.3)
i∈X α∈X
riα i,j∈X
rij a,b∈X Rαα0
i<j α<α0
as simple products (not antisymmetric for intermolecular electron exchange) of the

monomer wave functions
ϕk = ψaA ψbB (4.4)
and the eigenvalues are the sum of monomer eigenvalues
Ek = EaA + EbB (4.5)
4.2 Partition of the Hamiltonian

Since the subsystems are distinguishable (preserve their identities), each of the
N = NA + NB electrons and M = MA + MB nuclei can be assigned to one of
the subsystems. Subtracting from the total Hamiltonian
NA +NB MA +MB
! N +N MA +MB
X X Zα 1 AX B 1 X Zα Zα 0
Ĥ = T̂i − + + (4.6)
i=1 α=1
riα 2 i,j=1 rij α,β=1
Rαα0
35
the sum of monomer Hamiltonians one obtains the operator of the intermolecular
interaction
X X Zα Z β X X Z β X X Za X X 1
V̂ = − − + (4.7)
α∈A β∈B
Rαβ i∈A β∈B
riβ α∈A j∈B rαj i∈A j∈B rij
In the absence of any “natural” perturbation parameter, we shall consider a an

adiabatic switching of the interaction:
Ĥ(λ) = Ĥ A + Ĥ B + λV̂ (4.8)
4.2.1 Longuet-Higgins form of the interaction operator

The interaction operator can be written in a more compact and separable form,
based on the charge density operator, defined as
X X
%̂X (r) = Zα δ(r − Rα ) − δ(r − ri ) (4.9)
α∈X i∈X
with the matrix elements

X
hψa |%̂A (r)|ψa0 i = δaa0 Zα δ(r − Rα ) − P (aa0 |r, r 0 ) (4.10)
α∈X
where P (aa0 |r, r 0 ) is an element one the one particle density matrix.
The state charge density
X
%aa (r) = Zα δ(r − Rα ) − P (aa|r, r) (4.11)
α∈X
contains the contribution of both electrons and nuclei.

Transition charge density
%aa0 (r) = −P (aa0 |r, r) (4.12)
Using %̂(r) and the Coulomb-kernel function, T (r, r 0 ) = |r − r 0 |, the interaction

operator is Z Z
V̂ = dr dr0 %̂A (r)T (r, r0 )%̂B (r0 ) (4.13)
By a straightforward generalization of the Einstein summation convention
V̂ = %̂A B
r Tr,r 0 %̂r 0 (4.14)
36
4.3 The polarization approximation
Direct application of the RSPT to the the above problem with the eigenfunction of
Ĥ0 = HÂ + ĤB as zeroth order wave function is called the polarization approxima-
tion.
(n) (n−1)
Epol = hϕ0 |V̂ |ϕpol i (4.15)
n−1
X
(n) (n−1) (n−k)
ϕpol = −R̂0 V̂ ϕpol − ∆E (k) R̂0 ϕpol (4.16)
k=1
The first-order interaction energy in this approximation is

(1)
Epol = hψ0A ψ0B |V̂ |ψ0A ψ0B i (4.17)
Using the Longuet-Higgins operator

Z Z
(1)
Epol = dr dr0 hψ A |%̂A (r)|ψ A iT (r, r0 )hψ B |%̂B (r0 )|ψ B i (4.18)
it is easy to see that it is the Coulomb interaction of the charge densities of the two
subsystems. Z Z
(1)
Epol = dr dr0 %A 0 B 0
00 (r)T (r, r )%00 (r ) (4.19)
Electrostatic potential of the subsystems

Z
V (r) = dr 0 T (r, r 0 )%A
A 0
00 (r )
Z (4.20)
V (r) = dr 0 T (r, r 0 )%B
B 0
00 (r )
Alternative form of the first order interaction energy

Z Z
(1)
Epol = dr%00 (r)V (r) = dr%B
A B A
00 (r)V (r) (4.21)
4.3.1 Second order energy in the polarization approxima-

tion
The second-order interaction energy
(2)
X |hψ A ψ B |V̂ |ψ A ψ B i|2
0 0 a b
Epol = − A B
(4.22)
ab6=00
∆E0a + ∆E0b
37
can be written as a sum of three terms, each having a different physical meaning.
(2)
0 0 0 b
Epol = − B
induction A → B (4.23)
b6=0
∆E 0b

0 0 a 0
− A
induction A ← B (4.24)
a6=0
∆E 0a

0 0 a b
− A B
dispersion (4.25)
a6=0
∆E 0a + ∆E 0b
b6=0
These terms correspond to the following decomposition of the reduced resolvent

X |ψaA ψbB ihψaA ψbB |
R̂0 =
a,b
EaA − E0A + EaB − E0B
(ab)6=(00) (4.26)
= R0A OA
+ R0B OB
+ R0AB
= R0 (ind) + R0 (disp)
where ÔX = |ψ0X ihψ0X | is the projection operator to the ground state of subsystem
X.
4.3.2 Induction energy

∆E (2) (ind, A ← B) = −hψ0A ψ0B |V̂ R̂0A ÔB V̂ |ψ0A ψ0B i
1 B (4.27)
= −VrB h%A A B A A
r R̂0 %r 0 iVr = − Vr K(%r , %r 0 )Vr
B
2
0
The K(%A A
r , %r 0 ) = α(r, r ; ω = 0) charge density linear response function (suscepti-
bility) is a kind of generalized polarizability function, which gives the induced charge
density in a static external potential
Z
∆% (r) = dr 0 α(r, r 0 ; ω = 0)V B (r 0 )
A
(4.28)
The sum-over-states definition of the charge density susceptibility at the ω frequency

is
1 X [h0|%̂(r)|aih0|%̂(r 0 )|ai + h0|%̂(r 0 )|aih0|%̂(r)|ai] ω0a
α(r, r 0 ; ω) = 2
(4.29)
~ a6=0 ω0a − ω2
The induction energy can be regarded as the stabilization energy due to interaction
of the induced charge density with the electrostatic potential of the partner
Z
(2)
∆Estab = − dr∆%A (r)V B (r) (4.30)
38
partially compensated by the (positive) deformation energy, spent to polarize the
charge distribution
(2) 1 (2)
∆Edef = − ∆Estab (4.31)
2
Which means that the induction energy
(2) (2) 1 (2)

∆E (2) (ind, A ← B) = ∆Estab + ∆Edef = ∆Estab (4.32)
2
is the half of the stabilization energy (in the linear response approximation).
Note: do not mix up the terminology “polarization” and “induction”...
4.3.3 Dispersion energy

∆E (2) (disp) = −hψ0A ψ0B |V̂ R̂0AB V̂ |ψ0A ψ0B i
0 B 0
1 X X %A A B
ZZZZ
0a (r)%0a (r )%0b (s)%0b (s ) (4.33)
= T (r, s)T (r 0 , s0 ) A B
~ a6=0 b6=0 ω0a + ω0b
We can transform the double sum according to subsystems using the following iden-
tity
2 ∞
Z
1 x y
= dω (4.34)
x+y π 0 x + ω y + ω2
2 2 2
which leads to a separable form at the expense of an additional integral in the

frequency domain
ZZZZ
(2)
∆E (disp) = T (r, s)T (r 0 , s0 )×
2
Z
1 X ω0a A A
%0a (r)%A 0 X B B
ω0b %0b (s)%B 0 (4.35)
0a (r ) 0b (s )
dω
π A2 B2
~ a6=0 ω0a + ω2 b6=0 ω0b + ω2
We can recognize in the two separate sums the dynamic charge density susceptibil-
ities at imaginary frequencies of both subsystems, i.e.
1 X [h0|%̂(r)|aih0|%̂(r 0 )|ai + h0|%̂(r 0 )|aih0|%̂(r)|ai] ω0a

α(r, r 0 ; iω) = 2
(4.36)
~ a6=0 ω0a + ω2
Final expression of the dispersion energy (Casimir-Polder)

Z ∞ ZZZZ
~
(2)
∆E (disp) = dω drdr 0 dsds0 ×
2π 0 (4.37)
× αA (r, r 0 ; iω)T (r 0 , s0 )αB (s, s0 ; iω)T (r, s)
Corresponds to the Coulomb correlation of fluctuating charge densities of the two

systems. (Cf. fluctuation-dissipation theorem).
39
4.3.4 Summary of second order terms
%A B
r Trs %s %A B A
r Trs αss0 Ts0 r0 %r0
B
Trsαss A
0 Ts0 r 0 αr 0 r
electrostatic induction dispersion
4.3.5 Third order induction energy

Third order energy
∆E (3) = ϕ0 |V̂ R̂0 (V̂ − hV̂ i)R̂0 V̂ |ϕ0 i (4.38)
Decomposition of the reduced resolvent
R0 (ind) = R0A OB + OA R0B R0 (disp) = R0AB (4.39)
Pure induction part of the third order energy can be expanded

h i
∆E (3) (ind) = hϕ0 |V̂ R̂0 (ind) V̂ − hV̂ i R̂0 (ind)V̂ |ϕ0 i
= Trs Tr0 s0 Tr00 s00 × (4.40)
hϕA B A B
A B A B
A B A B
0 ϕ0 |%̂r %̂s R̂0 (ind) %̂r %̂s − h%̂r0 %̂s0 i R̂0 (ind) %̂r 00 %̂s00 |ϕ0 ϕ0 i
Two kinds of terms are obtained
◦ hyperpolarizability term (nonlinear response), e.g.
Trs Tr0 s0 Tr00 s00 h%̂A A A B B B B B B

r ih%̂r 0 ih%̂r 00 i h%̂s R̂0 (%̂s0 − h%̂s0 i)R̂0 %̂s00 i (4.41)
| {z }
1 B
6
β (s,s0 ,s00 )
40
◦ iterated linear response terms, e.g.
Trs Tr0 s0 Tr00 s00 h%̂A B B B A A A B

r i h%̂s R̂0 %̂s0 i h%̂r0 R̂0 %̂r 00 ih%̂s00 i (4.42)
| {z } | {z }
1 B 1 A 0 00
2
α (s,s0 ) 2
α (r ,r )
Graphical representation of the iterated linear response terms
Analogous terms appear in higher order terms, that can be iterated up to self-
consistency
4.4 Non-additive interactions

The Hamiltonian operator of a ternary complex
ĤA + ĤB + ĤC + V̂AB + V̂AC + V̂BC (4.43)
is ”additive”.
◦ First order energy is strictly additive

(1)
∆EABC = hψ A ψ B ψ C |V̂AB + V̂AC + V̂BC |ψ A ψ B ψ C i
= hψ A ψ B |V̂AB |ψ A ψ B ihψ C |ψ C i+
+ hψ A ψ C |V̂AC |ψ A ψ C ihψ B |ψ B i+ (4.44)
+ hψ B ψ C |V̂BC |ψ B ψ C ihψ A |ψ A i
(2) (2) (2)
= ∆EAB + ∆EAC + ∆EBC
◦ Second order energy

(2)
∆EABC = hψ A ψ B ψ C |(V̂AB + V̂AC + V̂BC )R̂0 (V̂AB + V̂AC + V̂BC )|ψ A ψ B ψ C i
(4.45)
41
The resolvent can be decomposed as
R̂0 = R̂0ABC
+R̂0AB ÔC + R̂0BC ÔA + R̂0CA ÔB 2-body dispersion
+ + R̂0A ÔB ÔC + R̂0B ÔC ÔA + R̂0C ÔA ÔB induction
Three types of contributions according to the interaction operators
– ”Diagonal” term
hψ A ψ B ψ C |V̂AB R̂0 V̂AB |ψ A ψ B ψ C i =

hψ A ψ B |V̂AB hψ C |R̂0 |ψ C iV̂AB |ψ A ψ B i = (4.46)
(2)
hψ A ψ B |V̂AB (R̂0AB + R̂0A ÔB + R̂0B ÔA )V̂AB |ψ A ψ B i = ∆EAB
– ”Off-diagonal” term
hψ A ψ B ψ C |V̂AB R̂0 V̂AC |ψ A ψ B ψ C i =

hψ A ψ B |V̂AB hψ C |R̂0 |ψ B iV̂AC |ψ A ψ C i =
(4.47)
hψ A ψ B |V̂AB |ψ B iR̂0A hψ C |V̂AC |ψ A ψ C i =
Trs Tr0 s0 %B A A A C
s h%̂r R̂0 %̂r 0 i%s0
non-additive 3-body induction interaction.
◦ Three-body dispersion (cf. Axilrod-Teller forces) appears in third order through

terms like
hV̂AB R̂0AB V̂BC R̂0BC V̂CA i (4.48)
42
Chapter 5
Beyond the polarization

approximation
The RSPT applied to intermolecular interaction is often called polarization approx-

imation, since it neglects effects that are related to the antisymmetry of the total
wave function. In the following we shall discuss the failure of the polarisation ap-
proximation and one of the possible methods to remedy this situation.
5.1 Example of H atom + proton system

Zeroth order Hamiltonian and wave function and the perturbation potential are
1 1 1 1
H0 = − ∆ − ϕ0 = 1sA V̂ = − + (5.1)
2 rA rB R
At R ≤ 0.05 a.u. (limit of united atoms) RSPT yields
8 16
Eel (R) = −2 + R2 + R3 + O(R4 ) (5.2)
3 3
For very large intermolecular distances, R > 12 a.u., the total energy can be ap-
proximated as
1 2.25 7.5
E(R) = Eel + = − 4 − 6 + O(R−7 ) (5.3)
R R R
5.1.1 Convergence of the polarization approximation for

H+
2
Pn
∆E (k)

k=1
∆(n) = 100 1 −
∆Eexact
43
R = 3.0 R = 12.5
n ∆E n ∆(n) ∆E n ∆(n)
1 3.3050(-03) 10.26 1.5000(-11) 100.000
2 -2.5686(-02) 71.14 -9.4250(-05) 27.809
3 -1.1074(-02) 56.87 -1.2599(-06) 26.844
4 -9.8501(-03) 44.17 -1.5711(-07) 26.723
5 -8.0099(-03) 33.84 -1.5409(-08) 26.711
6 -6.7955(-03) 25.08 -3.8934(-09) 26.708
8 -4.5279(-03) 12.03 -1.5287(-09) 26.706
12 -1.3808(-03) -0.41 -1.3304(-09) 26.702
16 +2.3496(-05) -2.40 -1.3244(-09) 26.698
20 +3.0523(-05) -1.07 -1.3241(-09) 26.694
30 -3.6059(-05) 0.30 -1.3241(-09) 26.683
38 -1.3906(-05) -0.07 -1.3241(-09) 26.675
Chalasinski et al. IJQC 11 (1977) 247
At large R the exact wave function (with intermediate normalization) is
ψ = ϕA + ϕB ≈ 1sA + 1sB (5.4)
i.e. the modification brought by the perturbation V̂ is not small at all.
ϕ0 ψ − ϕ0
−1/rA −1/rB
-10 -5 0 5 10
◦ near A R (a.u.)
– V̂ is small, therefore ϕ0 = 1sA is good approximation

◦ near B
– V̂ is big, therefore ϕ0 = 1sA is very bad approximation.
Expand the ψB = 1sB function by the eigenfunctions of A-centered basis
X
ψB = hψB |ak iak
k
At larger R bad convergence, since {ak } vanish exponentially at B.
∆E = hϕ0 |V̂ ϕA i + hϕ0 |V̂ ϕA i

(5.5)
rapid convergence obtained only at inf. order
44
5.2 Example of two H atoms
Interaction of two H atoms
1 1 1
V̂ = − − + (5.6)
r1B r2A r12
Unperturbed wave function
ϕ0 = a0 (1)b0 (2) (5.7)
The RSPT charge density is a simple superposition of the 1s densities,
%(r) = %1s (r − x)) + %1s (r + x)) (5.8)
In case of overlapping electron densities the antisymmetrized (triplet) wave

function is
1
ψ0 = p [a0 (1)b0 (2) − a0 (2)b0 (1)] (5.9)
2(1 − S 2 )
and the symmetry-adapted charge density is
1
%(r) = [%1s (r − x) + %1s (r + x) − 2S1s(r − x)1s(r + x)] (5.10)
1 − S2
Density
◦ depleted between nuclei

1 2
%(r = 0) = 2
2%1s (x) − 2S% 1s (x) = %1s (x) < 2%1s (x) (5.11)
1−S 1+S
◦ enhanced outside the nuclei, e.g. at r = 2x

1 1
%(2x) ≈ % 1s (x) + % 1s (3x) ≈ %1s (x) > %1s (x) (5.12)
1 − S2 1 − S2
The Hellmann-Feynmann forces will be repulsive.
◦ electron tunnels in both directions ⇒ exchange tunneling

1
◦ due to r12
the electronic motions become correlated ⇒ dispersion
In many-electron systems the RSPT ground state violates Pauli-principle ⇒

antisymmetrized product would be a better ψ0 , but it is not eigenfunction of Ĥ0 .
45
5.2.1 Behaviour of the RSPT interaction energy
∆E(R)
U(R)
∆EnRS (R)
◦ At short distances RSPT (polarization approximation) misses the repulsion:

no acceptable minimum
(RS)
◦ At large distance ∆E(R) − ∆En (R) vanishes exponentially, therefore the
polarization approximation is acceptable.
5.3 Symmetry failure

Why is the polarization approximation wrong at short and intermediate
distances?
◦ Antisymmetrizer commutes with total Hamiltonian

h i
Â, (Ĥ + λV̂ ) = 0 (5.13)
but neither with Ĥ0 , nor with V̂
[Â, Ĥ0 ] 6= 0 [Â, V̂ ] 6= 0 (5.14)
It follows from the commutation rules that
[Â, Ĥ0 ] = −λ[Â, V̂ ] (5.15)
i.e. non-zero zeroth-order and non-zero first order quantities are equal to
each other: no unambiguous definition of the perturbation order.
◦ Antisymmetrized product states of the subsystems
ÂϕA B
a ϕb (5.16)
46
form a non-orthogonal set. A hermitian operator has always a orthogonal
eigenfunctions. There can be no Hermitian Hamiltonian associated with
these zeroth-order wave functions.
◦ Symmetry dilemma: Zeroth order Hamiltonian Ĥ(λ = 0) = ĤA + ĤB has
lower symmetry than Ĥ(λ = 1)!
SN A SN B
N
ĤA + ĤB
Ĥ SNA +NB
The direct product the symmetric groups of rank NA and NB is a subgroup

of SNA +NB .
Consequence: the polarization approximation is only asymptotically convergent,
i.e. at a given intermolecular distance and orientation one cannot obtain the exact
energy as a power series of the λ perturbational parameter. In the practice, PA is
divergent at higher than 2nd order.
5.4 Claverie’s analysis

The lowest eigenfunction of Ĥ(λ = 0)
is a “bosonic” state, which is connected
with the physically forbidden mathemat-
ical ground state of Ĥ(λ = 1), while the
physical ground state of Ĥ(λ = 1) is con-
nected with an excited state of Ĥ(λ = 0).
Difference of p.g.s. and m.g.s. decreases
exponentially with R.
Since the m.g.s. becomes repulsive only
at chemical bond distances, polarization
approximation misses the repulsive ef-
fects.
Two possible strategies:
◦ abandon the usual partition and find a Ĥ0 for which Âϕ0 is eigenfunction
◦ maintain the partition, but reject RSPT
5.5 Symmetrized RS perturbation theory

Let the (NA + NB )-electron antisymmetrizer
1 X
Â = p (−1)P P (5.17)
(NA + NB )! P ∈SN
A +NB
47
and set the following zeroth order approximation to the wave function, satisfying
the intermediate normalization
ψ0 = N0 Âϕ0 N0 = hϕ0 |Âϕ0 i−1 (5.18)
First iteration in the energy
∆E1 = N0 hϕ0 |V̂ Âϕ0 i (5.19)
First iteration in the wave function
ψ1SRS = ϕ0 + N0 R̂0 (∆E1 − V̂ )Âϕ0 (5.20)
which is to be compared with the RSPT (polarization approximation, PA) result
ψ1P A = ϕ0 − R̂0 V̂ ϕ0 (5.21)
where the second term vanishes as R → ∞.
Separation of the V̂ -dependent component of the first wave function iteration
[Ĥ0 + V̂ , Â] = 0 (5.22)
−V̂ Â = Ĥ0 Â − Â(Ĥ0 + V̂ ) (5.23)
(∆E1 − V̂ )Â = Â∆E1 + Ĥ0 Â − ÂĤ0 − ÂV̂ (5.24)
(∆E1 − V̂ )Â = Â(∆E1 − V̂ ) + [Ĥ0 − E0 , Â] (5.25)
R̂0 [Ĥ0 − E0 , Â]ϕ0 = Âϕ0 − hϕ0 |Âϕ0 iϕ0 (5.26)
N0 R̂0 (∆E1 − V̂ )Âϕ0 = N0 Âϕ0 − N0 hϕ0 |Âϕ0 iϕ0 + N0 R̂0 Â(∆E1 − V̂ )ϕ0
= N0 Âϕ0 − ϕ0 + ψ (1) (5.27)
The first iteration of the wave function can be decomposed as

(exch)
ψ1 = ϕ0 + ψ0 + ψ (1) (5.28)
i.e. a large exchange correction
(exch)
ψ0 = N0 Âϕ0 − ϕ0 (5.29)
and a small V̂ -dependent correction,
ψ (1) = N0 R̂0 Â(∆E1 − V̂ )ϕ0 (5.30)
The second energy iteration gives
∆E2 = hϕ0 |V̂ ψ1 i = E (1) + E (2) (5.31)
with
E (2) = N0 hϕ0 |V̂ R̂0 Â(∆E1 − V̂ )ϕ0 i (5.32)
and due to the hermiticity of V̂ and Â
(1)
E (2) = N0 hϕpol |Â(V̂ − ∆E1 )ϕ0 i (5.33)
48
5.5.1 Separation of exchange contributions
At any order of the SRS perturbation theory, the energy correction can be
rigorously decomposed as a sum of exponentially decaying exchange component
and the polarization component
(n) (n)
E (n) = Epol + Eexch (5.34)
This can be done using the decomposition of the total antisymmetrizer as

NA !NB !
Â = ÂA ÂB (1 + P) (5.35)
(NA + NB )!
where P is the sum of inter-system permutations.
First order exchange energy
(1) hϕ0 |V̂ Pϕ0 i − hϕ0 |V̂ ϕ0 ihϕ0 |Pϕ0 i

Eexch = (5.36)
1 + hϕ0 |Pϕ0 i
Second order exchange energy
(1) (1)
(2) hϕpol |V̂ Pϕ0 i − hϕpol |V̂ ϕ0 ihϕ0 |Pϕ0 i
Eexch = (5.37)
1 + hϕ0 |Pϕ0 i
5.6 Symmetry adapted perturbation theory

(SAPT)
(Jeziorski, Moszynski, Szalewicz and Williams)
Zeroth-order wave Hamiltonian is the sum of isolated molecule Fock operators and
the solutions are expanded in a triple perturbation series
Ĥ = F̂ A + F̂ B + ζ V̂ + λA ŴA + λB ŴB (5.38)
Consecutive application of the RSPT (polarization approximation) and symmetry

adapted perturbation theory (exchange corrections) leads to the following
expression of the interaction energy:
∞
X ∞
X
(n) (n)
∆E = ∆Epol + ∆Eexch (5.39)
n=0 n=0
and both components are developed in the orders of the intramolecular correlation
operator
∞ X
X ∞ ∞ X
X ∞
(n) (nij) (n) (nij)
∆Epol = ∆Epol ∆Eexch = ∆Eexch (5.40)
i=0 j=0 i=0 j=0
49
Symbol Name Description
(10)
Epol Electrostatic Classical electrostatic interaction of un-
perturbed Hartree-Fock charge distribu-
(1n)
tions
Epol Electrostatic: correlation Correlation correction to the electrostatic
corrections energy from the nth order intramolecular
correlation effects on the charge distribu-
tions
(10)
Eexch Exchange (overlap) re- Repulsion of the closed shells: modifica-
pulsion tion of the interaction energy of Hartree-
Fock monomers due to the intermolecular
antisymmetrization
(1n)
Eexch Exchange (overlap) re- Intramolecular correlation correction to
pulsion: correlation cor- closed-shell repulsion
rections
Symbol Name Description
(20)
Eind Induction energy Energy arising from the distortion of each
molecule in the field of the unperturbed
Hartree-Fock charge distribution of the
other
(20)
Eind-exch Exchange induction Modification to the induction energy due
to antisymmetry effects
(20)
Edisp Dispersion energy Energy arising from the correlated fluctu-
ations of the of the unperturbed Hartree-
Fock charge distribution of each molecule
(20)
Edisp-exch Exchange dispersion Modification to the dispersion energy due
to antisymmetry effects
(2n)
Eind Induction energy: corre- Corrections to the induction energy due to
lation corrections intramolecular correlation effects
(2n)
Edisp Dispersion energy: corre- Corrections to the dispersion energy due
lation corrections to intramolecular correlation effects
50
SAPT decomposition of the He2 interaction energy
(After: Korona et al. J. Chem. Phys. 106 (1997), 5109)
51
52
Chapter 6
Molecules in electrostatic field
6.1 Electrostatic interaction Hamiltonian

Place a molecule in an external electrostatic field. The interaction Hamiltonian is
X
V̂ = qk V (r k ) with qk = e, Z (6.1)
k
Let us define the operator of the total charge density

X X
%̂(r) = Zα (r − Rα ) − δ(r − r i ) (6.2)
α i
permitting to express the interaction Hamiltonian in the compact form as

Z
V̂ = dr %̂(r) V (r) (6.3)
6.2 Molecule in a uniform electric field

Place molecules between two plates of a capacitor:
The electrostatic potential can be written in terms of the

3 components of the uniform electric field
X
V (r) = V0 + Fα r α = V 0 + Fα r α
α
(Einstein summation convention on repeated indices)
The interaction operator with the field of the capacitor simplifies as

Z
V̂ = dr %̂(r) (V0 + Fα rα ) = V0 q̂ + Fα µ̂α (6.4)
53
where the operators of the total charge and total dipole moment of the molecule are
Z Z
q̂ = dr %̂(r) µ̂α = dr %̂(r)rα (6.5)
In first order of the perturbation theory
∆E (1) = hψ0 |q̂|ψ0 i · V0 + hψ0 |µ̂α |ψ0 i · Fα = qV0 + Fα µα (6.6)
one retrieves the classical charge-potential and dipole-field interactions.

In second-order of the perturbation theory
X hψ0 |V̂ |ψk ihψk |V̂ |ψ0 i
∆E (2) = −hψ0 |V̂ R̂0 V̂ |ψ0 i = − (6.7)
Ek − E 0
k6=0
The matrix elements hψ0 |q̂|ψk i = 0, since the eigenstates are orthogonal. Only the
dipolar term survives
X
∆E (2) = − Fα hψ0 |µ̂α R̂0 µ̂β |ψ0 iFβ (6.8)
αβ
Using the definition of the dipolar linear response function (polarizability)

1 X
∆E (2) = − Fα ααβ Fβ (6.9)
2
αβ
In third order of the perturbation theory
∆E (3) = −hψ0 |V̂ R̂0 V R̂0 V̂ |ψ0 i (6.10)
Insert the interaction operator, and remember that V = V̂ − hV̂ i,

X
∆E (3) = − Fα Fβ Fγ ×
αβγ
X X
hψ0 |µ̂α |ψk ihψk |µ̂β |ψl ihψl |µ̂γ |ψ0 i
−
(Ek − E0 )(El − E0 )
k6=0 l6=0
X hψ0 |µ̂α |ψk ihψk |µ̂γ |ψ0 i (6.11)
hψ0 |µ̂β |ψ0 i =
(Ek − E0 )2
k6=0
1X
= Fα Fβ Fγ βαβγ
6
αβγ
where βαβγ is the first dipolar hyperpolarizability (non-linear response function).
54
Chapter 7
Multipole expansion – Cartesian

formalism
The complete determination of the complete charge density and various response
functions (and the associated deformation densities) is computationally heavy and
conceptually not very rewarding. As any distribution function, the essential features of
the charge distribution can be be characterized by its moments.
The mathematical analogy between the Longuet-Higgins interaction operator and the
electrostatic interaction energy allows us to generalize the results obtained for the latter
to the interaction operator too.
7.1 Potential of a charge distribution
Potential V (R) of the charge distribution %(r)

Z
V (R) = drT (R − r)%(r) (7.1)
can be expanded in Taylor series around r = 0 (|r| < |R|.
55
Taylor expansion of the Coulomb kernel

1 1 1 1
T (R − r) = − rα ∇α + rα rβ ∇ α ∇ β
R R 2! R
(7.2)
1 1
− rα rβ rγ ∇ α ∇ β ∇ γ + ...
3! R
Define the (Cartesian) Coulomb interaction tensors
1
T (R) =
R
1
Tα (R) = ∇α
R

1
Tαβ (R) = ∇α ∇β (7.3)
R

1
Tαβγ (R) = ∇α ∇β ∇γ
R

1
Tαβ...ν (R) = ∇α ∇β . . . ∇ν
R
and the electrostatic potential
Z Z
V (R) = T (R) dr%(r) − Tα (R) drrα %(r)
Z Z (7.4)
1 1
+ Tαβ (R) drrα rβ %(r) − Tαβγ (R) drrα rβ rγ %(r) + . . .
2! 3!
Define the multipole moments of the charge distribution
Z
q = dr%(r)
Z
mα = drrα %(r)
Z
Qαβ = drrα rβ %(r) (7.5)
Z
Oαβγ = drrα rβ rγ %(r)
Z
(n)
ξαβ...ν = drrαβ...ν %(r)
Cartesian multipole expansion of the potential
1 (−)n
V (R) = qT (R) − mα Tα (R) + Qαβ Tαβ (R) − . . . + ξαβ...ν Tαβ...ν (R) (7.6)
2! n!
56
7.2 Multipole moments
There are several conventions in the literature. The above-defined multipole moments
are symmetric tensors Qαβ = Qβα and their trace is nonzero
. X
Qαα = Qαα 6= 0 (7.7)
α
They can be called “unnormalized traced Cartesian” multipoles.

Unabridged multipoles (Applequist)
1
µαβ...ν =
ξαβ...ν (7.8)
n!
Traceless cartesian multipoles (Buckingham)
Potential of a quadrupole
1
V (2) (R) = Qαβ Tαβ (R) (7.9)
2
1
Since the T (R) = R function satisfies the Laplace equation

2 1
∇ =0
R
(7.10)
X 1 X
∇α ∇α = Tαα (R) = 0
α
R α
we can add to the quadrupole potential an arbitrary quantity
λδαβ Tαβ (R) (7.11)
without changing its value

1
V (2) (R) = (Qαβ + λδαβ ) Tαβ (R) (7.12)
2
Let us choose λ as the trace of Qαβ
1 1X
λ = − Qαα = − Qαα (7.13)
3 3 α
The new expression of the quadrupole potential

Z
1 1 1
V (2) (R) = · dr 3rα rβ − r2 δαβ %(r)Tαβ (R) = Θαβ Tαβ (R)

(7.14)
2 3 3
where have introduced the traceless cartesian quadrupole moment
Z
1
dr 3rα rβ − r2 δαβ %(r)

Θαβ = (7.15)
2
57
7.2.1 Traceless (Buckingham) multipole moments
General definition
(−)n ∂n
Z
(n) 2n+1 1
Mαβ...ν = dr%(r)r (7.16)
n! ∂rα ∂ . . . rν r
Electrostatic potential
X (−)n (n) (n)
V (R) = Mαβ...ν Tαβ...ν (R) (7.17)
n
(2n − 1)!!
where the notation (2n − 1)!! means
(2n − 1)!! = 1 · 3 · 5 · . . . (2n − 1) (7.18)
The traceless multipoles do not contain the full information about the %(r) charge
distribution, since they are undetermined up to an arbitrary spherically symmetric
component.
On the contrary, one can reconstruct the full charge density from the traced multipoles.
Consider the Fourier transform of %(r)
Z
%(k) = dreikr %(r) (7.19)
and expand the exponential

Z Z Z
1
%(k) = dr%(r) +ikα dr rα %(r) − kα kβ dr rα rβ %(r) + . . . (7.20)
2!
| {z } | {z } | {z }
q mα Qαβ
7.2.2 Explicit form of higher multipole moments

◦ Quadrupole moment Θαβ
X 3 1
Θαβ = q a [ aα aβ − a2 δαβ ] (7.21)
a
2 2
Six distinct components, five independent components.
◦ Octopole moment Ωαβγ

X 5
Ωαβγ = q a [ aα aβ aγ − a2 (aα δβγ + aβ δγα + aγ δαβ )] (7.22)
a
3
10 distinct components, seven independent components.
◦ Hexadecapole moment Φαβγγ
◦ 2n moments...
58
7.2.3 Translation of multipole moments
◦ Dipole moment of a point charge distribution
X
µα = q i rαi (7.23)
i
In a new frame obtained after a translation by a vector a

X X X
µ0α = q i (rαi − aα ) = q i rαi − aα q i = µα − aα q tot (7.24)
i i i
Dipole of a neutral system (qtot = 0) is invariant to translation of origin.

◦ Quadrupole moment
X
q i 3rαi rβi − (ri )2 δαβ

Θαβ = (7.25)
i
After translation by a
1 3
Θ0αβ = Θαβ + q tot (3aα aβ − a2 δαβ ) − (µtot aβ + µtot tot
β aα ) − δαβ µα aα (7.26)
2 2 α
7.3 Electric field, field gradient

The electric field, field gradient
Fα (R) = −∇α V (R) Fαβ (R) = −∇α ∇β V (R) (7.27)
and higher derivatives of the potential of a multipolar charge distribution can be

expressed with the same interaction tensors
Fαl11 ...ν1 = −∇α . . . ∇ν1 V (R) =
∞
X (−)l2 (7.28)
− T l1 +l2 M l2
(2l2 − 1)!! α1 ...ν1 α2 ...ν2 α2 ...ν2
l2
7.4 Energy of a charge distribution in

non-uniform field
We choose the origin inside the charge distribution, %(r) and expand the external
potential in Taylor series around the origin
1 1
V (r) = V (0) + rα Vα (0) + rα rβ Vαβ (0) + rα rβ rγ Vαβγ (0) + . . . (7.29)
2! 3!
Using the expression of the interaction energy
Z Z Z
U = dr%(r)V (r) = dr%(r)V (0) + dr%(r)rα Vα (0)
Z Z (7.30)
1 1
+ dr%(r)rα rβ Vαβ (0) + dr%(r)rα rβ rγ Vαβγ (0) + . . .
2! 3!
59
where we recognize the definition of the unnormalized traced multipoles,
1 1
U = M V + Mα V α + Mαβ Vαβ + Mαβγ Vαβγ + . . . (7.31)
2! 3!
By the virtue of the Laplace equation, Vαα = 0, we can subtract the trace of Mαβ , i.e.
1
3 Mαα from the quadrupolar energy
1 1 1 1
Mαβ Vαβ = Vαβ (Mαβ − δαβ Mαα ) = Vαβ Θαβ (7.32)
2 2 3 3
By similar manipulations at higher orders
1 1
U = qV + µα Vα + Θαβ Vαβ + Ωαβγ Vαβγ
3 3·5
1 1 (7.33)
+ Φαβγδ Vαβγδ + ξ (n) Vα...ν
3·5·7 (2n − 1)!! α...ν
The energy is the scalar product of the field with dipole, filed gradient with quadrupole
and higher field gradients with the corresponding multipole tensors.
Attention: the result may depend on the choice of origin!
7.5 Multipole interaction energy

Take two sets of multipoles. Taylor series of the Coulomb interaction
1 1 1
0
= = =
|τ − τ | |B + s − A − r| |R − (r − s)|
(7.34)
1 1 1 1
= + (rα − sα )∇α + (rα − sα )(rβ − sβ )∇α ∇β + ...
R R 2! R
which leads directly to an expression of the interaction energy of two multipole charge
distributions in terms of their traced Cartesian moments
U (R) = q A q B T (R) + (mA B A
α q − q mα )Tα (R)
1 (7.35)
+ (QA q B + q A QB A B B A
αβ − mα mβ − mα mβ )Tαβ (R) + . . .
2! αβ
Using the traceless (Buckingham) moments
X (−)`1
U (R) = M (`1 )[A] T (`1 +`2 ) (R)Mα(`22...ν
)[B]
(7.36)
(2`1 − 1)!!(2`2 − 1)!! α1 ...ν1 α1 ...ν1 α2 ...ν2 2
` 1 `2
7.6 Calculation of cartesian interaction tensors

They are defined as
1 1
T = = R−1 = q (7.37)
R Rx2 + Ry2 + Rz2
Using the following derivation “rules”
∇α Rβ = δαβ ∇α Rn = nRα R(n−2) ∇α R−n = −nRα R−(n+2) (7.38)
60
◦ Charge-dipole
Tα (R) = ∇α T (R) = −Rα R−3 (7.39)
◦ Dipole-dipole/charge-quadrupole
Tαβ (R) = ∇α ∇β T (R)

= −∇α Rβ R−3 = −δαβ R−3 + 3Rα Rβ R−5 (7.40)
2 −5
= (3Rα Rβ − R δαβ )R
◦ Dipole-quadrupole/charge-octopole
Tαβγ (R) = ∇α ∇β ∇γ T (R) = δα (3Rβ Rγ − R2 δβγ )R−5

= 3δαβ Rγ R−5 + 3δαγ Rβ R−5 + 3δβγ Rα R−5 + 3 · 5 · Rα Rβ Rγ R−7 (7.41)
= 3 5Rα Rβ Rγ − R2 (Rα δβγ + Rβ δγα + Rγ δαβ ) R−7

Interaction tensors
◦ invariant to interchange of indices

◦ traceless, since ∇2 (1/R) = 0
◦ 2n+1 independent components
Cartesian tensor formulation
◦ simple in low orders

◦ easily extended algorithms
◦ global frame
◦ heavy formalism
◦ too much components (reducible and
redundant)
61
7.7 Example: Structure of the HX dimers
Competition of three multipolar interaction. Angular dependence for linear molecules
(symmetric tops) is the following
◦ dipole-dipole
µA µB
Uµµ = − (2 cos θA cos θB − sin θA sin θB cos ϕ) (7.42)
R3
◦ dipole-quadrupole
3µA ΘB
UµΘ = [cos θA (3 cos2 θB − 1) − sin θA sin 2θB cos ϕ] (7.43)
4R4
◦ quadrupole-quadrupole
ΘA ΘB 3
UΘΘ = [1 − 5 cos2 θA − 5 cos2 θB − 15 cos2 θA cos2 θB
R5 4 (7.44)
+ 2(4 cos θA cos θB − sin θA sin θB cos ϕ)2 ]
The structure is characterized by θA ≈ 0 and ϕ = 0, µA = µB = µ and

ΘA = ΘB = Θ. The sum of the 4 interaction terms (two for UµΘ )
µ2
U =− · 2 cos θ
R3
µΘ 3
+ 4 · [(3 cos2 θ − 1) − 2 cos θ] (7.45)
R 2
Θ2
+ 5 · 3(3 cos2 θ − 1)
R
Θ
Introduce the parameter λ = µR
µ2 3
−2 cos θ + λ [(3 cos2 θ − 1) − 2 cos θ] + λ2 3(3 cos2 θ − 1)

U= 3
R 2 (7.46)
µ2
3λ(1 + 2λ)(3 cos2 θ − 1) − 2(2 + 3λ) cos θ

= 3
2R
62
Electrostatic interaction energy
is plotted for the
◦ HF dimer
µ = 0.72ea0 ,
Θ = 1.875ea20
R = 5.1a0 ;
minimum at 68 degrees
(exp. 60)
◦ HCl dimer
µ = 0.433ea0 ,
Θ = 2.8ea20
R = 6.8a0 ;
minimum at 79 degrees
63
64
Chapter 8
Multipole expansion – spherical

harmonics formalism
8.1 Potential of a charge distribution
The distance |R − r| can be expressed with the help

of the cosine rule
(R − r) · (R − r) = R2 + r2 − 2Rr cos γ (8.1)
and the Coulomb-kernel can be written in two alternative forms

 h i−1/2
1 r 2 r


R

 1 + R − 2 R cos γ if r < R,
T (R − r) = (8.2)
 h i−1/2
 1 1 + R 2 − 2 R cos γ


r r r if r > R.
The expression in brackets is just the generator function of the Legendre polynomials
∞
X
(1 + t2 − 2t cos γ)−1/2 = Pl (cos γ) tl |t| < 1 (8.3)
l
65
The first few Legendre polynomials
P0 (x) = 1
P1 (x) = x
1
P2 (x) = (3x2 − 1)
2
1
P3 (x) = (5x3 − 3x)
2
1
P4 (x) = (35x4 − 30x2 + 3)
8
and at an arbitrary order they can be obtained from the recurrence relation
(2l + 1)xPl = (l + 1)Pl+1 + lPl−1 (8.4)
Depending on the space domains, either the first (r < R), or the second (R < r) form of
the generator function can be applied:
P∞ l
r
 l Rl+1 Pl (cos γ) if r < R

T (R − r) = (8.5)
P∞ Rl

l r l+1 Pl (cos γ) if r > R
The formulae for the two domains of space (r < R, r > R) are usually expressed in the
following condensed form
∞ l
X r<
T (R − r) = l+1
Pl (cos γ) (8.6)
l=0
r>
Using the addition theorem of spherical harmonics

l
X
4π ∗
Pl (cos γ) = Ylm (ω)Ylm (Ω) (8.7)
2l + 1
m=−l
where ω = (θ, φ) and Ω = (Θ, Φ) are the angular components of the polar coordinates
of r and R, respectively.
The (complex) spherical harmonics for m ≥ 0
1/2
m 2l + 1 (l − m)!
Ylm (ω) = (−) P`m (cos θ) exp(imφ) (8.8)
4π (l + m)!
and for m ≤ 0
∗
Ylm (ω) = (−)m Ylm (ω) (8.9)
Here we used the associated Legendre polynomials, that are defined for m > 0 through
the derivatives of Pl (x) as
m
2 m/2 d
P`m (x) = (1 − x ) P` (x) (8.10)
dx
66
P00 (cos θ) = 1
P10 (cos θ) = cos θ
P11 (cos θ) = sin θ
1
P20 (cos θ) = (3 cos2 θ − 1)
2
P21 (cos θ) = 3 sin θ cos θ
P22 (cos θ) = 3 sin2 θ
Finally, we obtain the spherical harmonics form of the interaction kernel
X∞ X l l
4π r< ∗
T (R − r) = Y (ω)Ylm
l+1 lm
(Ω) (8.11)
2l + 1 r
l=0 m=−l >
8.1.1 Potential outside of the charge distribution

If the charge distribution %(r) vanishes in the points R where we want to calculate the
potential, the condition r < R is always satisfied
∞ X ` Z
X 4π ∗
V (R) = Y`m (Ω)R−`−1 drY`m (ω)r` %(r) (8.12)
2l + 1
`=0 m=−`
Define the modified spherical harmonics

1/2
4π
C`m (ω) = Y`m (ω) (8.13)
2l + 1
as well as the regular, R`m (r), and irregular, I`m (r), spherical harmonics
R`m (r) = r` C`m (ω) I`m (r) = r−`−1 C`m (ω) (8.14)
Using these functions, in terms of complex spherical harmonics multipole moments, Q`m
Z
Q`m = drR`m (r)%(r) (8.15)
the potential becomes

∞ X
X ` Z ∞ X
X `
V (R) = (−)m I`m (R) drR`m (r)%(r) = (−)m I`m (R)Q`m (8.16)
`=0 m=−` `=0 m=−`
It is more practical to use real spherical harmonics instead the complex ones
r
1
(−)m Q`m + Q`m

Q`mc = (8.17)
2
r
1
(−)m Q`m − Q`m

Q`ms = (8.18)
2
Relationship of the real spherical harmonics and Cartesian multipole moments
67
(n) (n)
Q`mc/s ξα...ν Mα...ν
Q00 q q
Q10 mz µz
Q11c mx µx
Q11s my µy
1
Q20 2 (2Qzz − Qxx − Qyy ) qΘzz
√ 4
Q21c 3Qxz Θxz
√ q3
4
Q21s 3Qyz 3 Θyz
√
3 √1 (Θxx − Θyy )
Q22c 2 (Qxx − Qyy ) 3 q
√ 4
Q22s 3Qxy 3 Θxy
The absolute value of the multipole moment

`
2 X X
Q` = Q`m Q∗`m = Q2`κ (8.19)
m=−` κ
This quantity is invariant with respect to the orientation of the frame.
8.1.2 Potential in the overlap region

Take a charge distribution
%(r) = f (ω)σ(r) (8.20)
with a radial part, which vanishes only at r → ∞. The general expression of the
potential is
∞ Z ∞ l l Z
4π 2 r<
X X
∗
V (R) = drr l+1 σ(r) dωYlm (ω)Ylm (Ω) (8.21)
2l + 1 0 r >
l=0 m=−l
The radial integral should be divided in two parts

R ∞
rl Rl
Z Z
2
I(R) = drr l+1 σ(r) + drr2 σ(r) (8.22)
0 R R rl+1
For instance, take an exponential function (STO)
σ(r) = rn e−ar (8.23)
Use the integration rules

Z ∞
n!
drrn e−ar =
0 an+1
∞ n (8.24)
n! X (aR)k −ar
Z
n −ar
drr e = n+1 e
R a k!
k=0
68
The radial integral
n+2+l
X (aR)k n+1−l
X (aR)k
n! 1 −ar 2l+1 −ar
I(R) = · l+1 1 − e +R e (8.25)
an+1 R k! k!
k=0 k=0
For a 1s function (l = 0, n = 0)
α3 −2αr
σ(r) = e (8.26)
π3
the potential

1 −2αR
V(1s)2 (R) = 1 − (1 + αR)e (8.27)
R
In general, the potential of a charge distribution can be separated to a multipolar and
penetration component.
8.2 Exact multipolar part of the potential

In quantum chemistry, molecular charge densities are usually developed on Gaussian
basis functions X X
%(r) = Zα δ(r − R) − Pµν χ∗µ (r)χν (r) (8.28)
µν
For this case, the multipolar potential can be exactly evaluated.
Let us consider the χν (r) functions as primitive Gaussian functions of the general form
χ(r) = R(r)Ylm (θ, φ) (8.29)
The electronic contribution to the charge density is the sum of two types of terms:
◦ One-center densities, both basis functions are on the same atomic center
%µν (r) = Rµ (r)Rν (r)Ylµ mµ (ω)Ylν mν (ω) (8.30)
Apply the sum rule (Clebsch-Gordan series)
X m̄µ mν m
Ylµ mµ (ω)Ylν mν (ω) = Klµ lν l Ylm (ω) (8.31)
lm
with
|lµ − lν | < l < lµ + lν and − mµ + mν + m = 0 (8.32)
The one-center densities can be written as a finite series of multipolar potentials
(maximal order is lµ + lν ) around the natural expansion centre.
◦ Two-center densities can be handled in an analogous way, by using the Gaussian
addition theorem, which leads to the natural expansion centre, the “barycenter”
αµ Rµ + αν Rν
Rµν = (8.33)
αµ + αν
The multipole expansion wrt to the barycenter is finite, with the highest rank of
multipole of lµ + lν .
69
8.3 Buehler-Hirschfelder bipolar expansion
The interaction kernel T (r AB ) is expanded as

∞ l<
1 X X |m|
= JlA lB (r A r B ; R)YlA m (ωA )YlB m (ωB ) (8.34)
r AB
lA lB =0 m=−l<
|m|
The space is divided on 4 regions and the JlA lB function takes different forms in these
regions. Region I. corresponds to the “standard” multipole expansion.
8.4 Spherical harmonics expansion of the

interaction
One can obtain the bipolar expansion by applying the previously derived result for
|r B − r A | < |R|
∞ X
l
1 X
= (−)m Rlm (r B − r A )Ilm (R) (8.35)
|R + r B − r A |
l=0 m=−l
According to the addition theorem of regular spherical harmonics

1/2
X X
l−m (2l + 1)!
Rlm (r B + r A ) = δlA +lB ,l (−) ×
(2lA )!(2lB )!
lA lB mA mB (8.36)

lA lB l
× R (r )R (r )
mA mB m lA mA A lB mB B

lA lB l
where a Wigner 3j coefficient.
mA mB m
70
Use that Rlm (−r) = (−)l Rlm (r), the multipole expansion of the interaction energy
1/2
lB (2lA + 2lB + 1)!
X X
U (R) = (−) ×
m m m
(2lA )!(2lB )!
l A lB A B (8.37)

l lB lA + lB
× A QA B
lA mA QlB mB IlA +lB ,m (R)
mA mB m
The multipole moments are expressed in a global coordinate system.
Transform the multipoles to a local coordinate system
X
l
Q̃lk = Qlm Dmk (Ω) (8.38)
m
where Ω = (α, β, γ) is the rotation that takes the global axes to the local ones, and
l (Ω) are the elements of the (hermitian) Wigner rotation matrices.Inversely, the
Dmk
global multipole components can be written in terms of the local ones
X X
Qlm = l
Q̃lk Dkm (Ω−1 ) = l
Q̃lk [Dkm (Ω)]∗ (8.39)
k k
Insert the local-frame multipole moments and define a new orientation-dependent

function
−1
kA kB lA −lB −j lA lB j
S̄lA lB j = i ×
0 0 0
(8.40)
X l lB lA + lB
× [DklAA mA (ΩA )]∗ [DklBB mB (ΩB )]∗ Cjm (θ, φ) A
mA mB m
mA mB m
where θ, φ are the polar angles of the intermolecular vector. In terms of these functions
and expanding the Wigner 3j symbols one obtains
X X l A + l B mA mB −lA −lB −1
U (R) = Q̃A B
lA mA Q̃lB mB S̄lA lB lA +lB R (8.41)
lA
lA lB mA mB m
or after transformation to real components

X X l A + l B κA κB −lA −lB −1
U (R) = Q̃A B
lA κA Q̃lB κB S̄lA lB lA +lB R (8.42)
lA
lA l B κ A κ B m
One can define the spherical analogue of the interaction tensors

lA + lB
Tl1 κ1 ,l2 κ2 = S̄lκ11l2κl21 +l2 R−l1 −l2 −1 (8.43)
lA
and the electrostatic interaction energy (operator) takes the simple form
X X X
U= Q̃A
lA κA TlA κA ,lB κB Q̃lB κB = Q̃A AB B
t Ttu Q̃u (8.44)
l A lB κ A κ B m tu
with t, u = {lκ}.
The T tensors can be expressed in terms of the unit vectors of the respective local
coordinate systems, and the vector R.
71
8.5 Interaction tensors between local frame
multipoles
Following the derivation of Hättig and Heß (Mol.Phys. 81 (1994) 813), the interaction
energy in space-fixed frame
%A (r)%B
ZZ
SF (s)
U= drds SF (8.45)
|s − r|
Let us write the rotation matrices from the space-fixed (global) to the molecule-fixed
(local) frame as R̂(ΩA ) and R̂(ΩB ), where Ω = (α, β, γ) are the Euler angles, and the
coordinates
r = A + R̂(ΩA )r A and s = B + R̂(ΩB )r B (8.46)
Use the notation R = B − A and that the length of a vector does not change by
rotation
|s − r| = |R + R̂(ΩB )r B − R̂(ΩA )r A | = |R̂−1 (ΩA )R + R̂−1 (ΩA )R̂(ΩB )r B − r A | (8.47)
Electrostatic energy in molecule-fixed charge distributions
%A B
ZZ
MF (r A )%MF (r B )
U= dr A dr B =
|R̂−1 (ΩA )R + R̂−1 (ΩA )R̂(ΩB )r B − r A |
X∞ X Z (8.48)
= Ql1 κ1 · dr B Il1 κ1 (R̂−1 (ΩA )R + R̂−1 (ΩA )R̂(ΩB )r B ) · %B
A
MF (r B )
l1 κ1
The irregular spherical harmonics can be expanded in Taylor series
−1 −1
Il1 κ1 (R̂A R + R̂A R̂B r B ) =
∞
X 1 −1 −1
= (R̂A R̂B r B · ∇R̂−1 R )l2 · Il1 κ1 (R̂A R) =
l2 ! A
l2 =0
∞
X 1 −1
= (r B · ∇R̂−1 R )l2 · Il1 κ1 (R̂A R) (8.49)
l2 ! B
l2 =0
and in terms of spherical tensors as
−1 −1
Il1 κ1 (R̂A R + R̂A R̂B r B ) =
∞ X
X 1 −1
= Rl κ (r B ) · Rl2 κ2 (∇R̂−1 R ) · Il1 κ1 (R̂A R) (8.50)
(2l2 − 1)!! 2 2 B
l2 =0 κ2
Using the identity Ilκ (R) = (−)l Rlκ (∇)(1/R), the interaction energy becomes
∞ XX
X ∞ X
U= QA B AB
l1 κ1 · Ql2 κ2 · Tl1 κ1 l2 κ2 (8.51)
l1 κ1 l2 κ2
72
with TlAB
1 κ 1 l2 κ 2
interaction tensor
(−)l1 1 −1 −1 1
TlAB
1 κ 1 l2 κ 2
(R, ΩA , ΩB ) = · Rl2 κ2 (R̂B ∇R )Rl1 κ1 (R̂A ∇R ) (8.52)
(2l1 − 1)!! (2l2 − 1)!! R
The rotated internuclear vector and the rotated differential operators can be expressed in
terms of direction cosines and the internuclear distance R. The rotation matrices can be
written in terms of the unit vectors of the molecule fixed frame in the space-fixed system
R̂A = (e(A) (A) (A)

x , ey , ez ) and R̂B = (e(B) (B) (B)
x , ey , ex )
(A)
the intermolecular unit vector, eAB = R/R and the direction cosines, rαA = eα · eAB
and
(B)
cαβ = eβ · e(A)
α .
For instance
1 2
T20,00 = (3rαA − 1)R−3
2 (8.53)
T1α,1β = (3rαA rβB + cαβ )R−3
Computation of the general expressions in both cartesian and spherical tensors are
quite costly and scale as L6 , where L is the order of 1/R expansion. Using an
intermediate coordinate transformation, Hättig derived general recurrence relations,
which scale as L4 (CPL, 260(1996)341).
73
74
Chapter 9
Long range perturbation theory
9.1 Induction energy

Substitution of the expanded interaction operator leads at the first order to the
multipole expression of the electrostatic interaction energy. At the second order, we have
in the Cartesian tensor formulation
X 1
E (2) (ind,B) = − B
×
b6=0
∆E0b
(9.1)
h00|T q̂ A q̂ B + Tα (q̂ A µ̂B A B A B A B A B
α − µ̂α q̂ ) + Tαβ (q̂ Θ̂αβ + Θ̂αβ q̂ − µ̂α µ̂β ) + . . . |0bi×
h0b|T q̂ A q̂ B + Tα0 (q̂ A µ̂B A B A B A B A B

α0 − µ̂α0 q̂ ) + Tα0 β 0 (q̂ Θ̂α0 β 0 + Θ̂α0 β 0 q̂ − µ̂α0 µ̂β 0 ) + . . . |00i
Performing the integrations on the A subsystem, we obtain the corresponding ground

state multipole moments. Grouping the terms according to the multipole rank of the B
subsystem, and taking into account that hψ0B |q̂ B |ψbB i = 0,
E (2) (ind,B) =
X h0|µ̂B B
α |bihb|µ̂α0 |0i
− (Tα q A − Tαβ µA
α + . . .) B
(Tα0 q A − Tα0 β 0 µA
α0 + . . .)−
b6=0
∆E 0b
X h0|Θ̂B B
αβ |bihb|µ̂α0 |0i
A
− (Tαβ q + Tαβγ µA
γ + . . .) B
(Tα0 q A − Tα0 β 0 µA
β 0 + . . .)−
b6=0
∆E0b (9.2)
X h0|µ̂B B
α |bihb|Θ̂α0 β 0 |0i
− (Tα q A − Tαβ µA
α + . . .) B
(Tα0 β 0 q A + Tα0 β 0 γ 0 µA
γ + . . .)−
b6=0
∆E 0b
X h0|Θ̂B B
αβ |bihb|Θ̂α0 β 0 |0i
− (Tαβ q A + Tαβγ µA
γ 0 + . . .) B
(Tα0 β 0 q A − Tα0 β 0 γ 0 µA
γ 0 + . . .) . . .
b6=0
∆E 0b
75
Introduce the multipole polarizabilities
X h0|µ̂α |nihn|µ̂β |0i + h0|µ̂β |nihn|µ̂α |0i
ααβ =
∆E0n
n6=0
X h0|µ̂α |nihn|Θ̂βγ |0i + h0|Θ̂βγ |nihn|µ̂α |0i
Aα,βγ = (9.3)
∆E0n
n6=0
1 X h0|Θ̂αβ |nihn|Θ̂γδ |0i + h0|Θ̂γδ |nihn|Θ̂αβ |0i

Cαβ,γδ =
3 ∆E0n
n6=0
and identify the parentheses as the electric field, field gradient, etc.
FαA = −(Tα q A − Tαβ µA

α + . . .)
A
(9.4)
Fαβ = −(Tαβ q A + Tαβγ µA
γ + . . .)
The induction energy in terms of the field (and its gradients) and the multipolar
polarizabilities
1 1 1 A B
E (2) (ind,B) = − FαA αα,α
B A
0 Fα 0 − FαA AB A A
α,α0 β 0 Fα0 β 0 − Fαβ Cαβ,α0 β 0 Fα0 β 0 + . . . (9.5)
2 3 6
In the spherical tensor formalism, the general polarizability component
X h0|Q̂`κ |nihn|Q̂`0 κ0 |0i + h0|Q̂`0 κ0 |nihn|Q̂`κ |0i
α`κ,`0 κ0 = (9.6)
∆E0n
n6=0
and the induction energy takes the form

1 XX B 1X
E (2) (ind,B) = − A A
α`κ`0 κ0 V`κ V `0 κ 0 = − ∆QB A
`κ` V`κ (9.7)
2 0 0
2
`κ ` κ `κ
where ∆QB
`κ` are the induced moments of B
X
∆QB
`κ =
B
α`κ` A
0 κ 0 V `0 κ 0 (9.8)
`0 κ 0
The multipolar induction energy can be truncated
◦ according to the powers of (1/R)N

does not ensure that the induction energy is negative
◦ according to the maximum rank of the multipole operators (L)
always negative
76
9.2 Multipole polarizabilities
Dipole polarizability is a symmetric tensor
 
αxx αxy αxz
α= αyy αyz  (9.9)
αzz
Units are volume (Å3 or bohr3 ); the SI units are Fm2 . 1 a.u. = 0.14818 Å3 =
0.16488×10−40 Fm2 .
Proportional to the volume, as it can be seen by introducing an average excitation
energy and using the resolution of identity
X 2h0|r̂α |nihn|r̂α |0i 2
h0|r̂α2 |0i − h0|r̂α |0ih0|r̂α |0i

α= ≈ (9.10)
∆E0n U
n6=0
and inversely proportional to the excitation energy.

As any second-rank tensor, it can be decomposed into irreducible parts, according to
α = α(0) + α(1) + α(2) (9.11)
where
(0) 1
ααβ = Tr(α)δαβ
3
(1) 1
ααβ = (ααβ − αβα ) = 0 (9.12)
2
(2) 1 1
ααβ = (ααβ + αβα ) − Tr(α)δαβ
2 3
α(1) vanishes, since α is symmetric. In a principal-axis system we have the
decomposition
   
α 0 0 αxx − α 0 0
α = 0 α 0 +  0 αyy − α 0  (9.13)
0 0 α 0 0 αzz − α
The first trace-invariant quantity that can be formed is the mean polarizability
1 1
α = Trα = (αxx + αyy + αzz ) (9.14)
3 3
and the second trace-invariant is the polarizability anisotropy, γ, defined as the positive
root of
1
γ 2 = [3Tr(α2 ) − Tr(α)2 ] (9.15)
2
or in components
1
γ 2 = [3ααβ αβα − ααα αββ ] (9.16)
2
77
which can be written in a principal axis system
1
γ 2 = [(αxx − αyy )2 + (αyy − αzz )2 + (αzz − αxx )2 ] (9.17)
2
and makes clear that this quantity vanishes for spherically symmetric systems.
For linear molecules αzz = αk and αxx = αyy = α⊥ and the polarizability anisotropy,
γ
γ = αk − α⊥ (9.18)
For linear molecules the dipole-quadrupole polarizability can be written in terms of the
unit vector along the symmetry axis, e = (ex , ey , ez ) as
1
Aαβγ = Ak (3eα eβ eγ − eα δβγ )
2 (9.19)
+ A⊥ (eβ δαγ + eγ δαβ − 2eα eβ eγ )
where Ak = Azzz and A⊥ = Axxz = Ayyz .
For tetrahedral molecules the dipole-dipole polarizability is isotropic, the
dipole-quadrupole polarizability contains only one independent principal axis component,
A = Axyz . This is the first anisotropic polarizability for tetrahedral symmetry.
9.2.1 Translational invariance

The dipole-dipole polarizability is always origin-independent.
Higher-rank multipole polarizabilities are (always) origin-dependent. For instance, the
Az,zz component involves the Θ̂zz operator. After a shift of the origin by c = (0, 0, c)
Θ̂C O O
zz = Θ̂zz − 2cµ̂z (9.20)
and the dipole-quadrupole polarizability at the new origin is

X h0|µ̂O |nihn|Θ̂O − 2cµ̂O |0i
z zz z
AC
z,zz = = AO
z,zz − 2cαzz (9.21)
∆E0n
n6=0
This means that the origin (and the dipole polarizability) should always be specified
when Aα,βγ is given.
9.2.2 Multipole polarizabilities – spherical tensors

The general multipole polarizabilities, defined in terms of the (irreducible) multipole
moment operators, Q̂`m
X h0|Q̂` |nihn|Q̂`2 m2 |0i + h0|Q̂`2 m2 |nihn|Q̂`1 m1 |0i
1 m1
α`1 m1 ,`2 m2 = (9.22)
∆E0n
n6=0
is a reducible spherical tensor quantity, which can be decomposed into irreducible parts
as X
α`1 `2 :kq = C(`1 `2 k; m1 m2 q)α`1 m1 ,`2 m2 (9.23)
m1 m2
78
where C(`1 `2 k; m1 m2 q) are the Clebsch-Gordan coefficients.
Since C = 0, unless k = `1 + `2 , `1 + `2 − 1, . . . , |`1 − `2 |, the nonzero irreducible parts of
the dipole-dipole polarizability can be only with k = 0, 2, for the dipole-quadrupole
polarizability, k = 1, 3, etc. For instance,
r
1
α11:00 = (αxx + αyy + αzz )
3
r (9.24)
2
α11:20 = γ
3
9.3 Distance dependence of the induction

energy
Interaction of a spherically symmetric polarizability ααβ = αδαβ with multipoles:
◦ charge
1 1 1 X αRα2 1 q2α
E(ind, q) = − qTα ααβ Tβ q = − q 2 Tα Tα α = − q 2 = − (9.25)
2 2 2 α
R3 R3 2 R4
◦ dipole
1 1 µ2 α(3 cos2 θ + 1)
E(ind, µ) = − µα Tαβ αββ 0 Tβ 0 α0 µα0 = − (9.26)
2 2 R6
◦ quadrupole
E(ind, Θ) ∼ R−8 (9.27)
9.4 Convergence of the multipolar induction

energy
The multipole-expanded interaction Hamiltonian of a H-atom + proton system is
∞ n
1 X r
Vmult = 1− Pn (cos θ) (9.28)
R R
n=0
The exact second order multipole energy (Dalgarno & Lynn, 1957)
∞
X (2n + 2)!(n + 2)
E (2) = − (9.29)
n=1
n(n + 1)R(2n+2)
The ratio of successive terms
2n(n + 3)(2n + 3)
lim =∞ (9.30)
n→∞ (n + 2)R2
indicates that for any value of R the series is divergent.
79
9.5 Dispersion energy
Dispersion interaction energy in spherical tensor formalism
XX X X
E (2) (disp) = −
a6=0 b6=0 À mA `B mB
(9.31)
h00|Q̂A B A B
À mA TÀ mA ,`B mB Q̂`B mB |abihab|Q̂`0 m0 T`0A m0A ,`0B m0B Q̂`0 0 |00i
A A B mB
× A + ∆E B
∆E0a 0b
Applying the Casimir-Polder method

2
E (2) (disp) = − T` m ,` m T 0 0 0 0 ×
~π A A B B À mA ,`B mB
Z ∞ h0|Q̂A A ∞ B B
À mA |aiha|Q̂`0 m0 |0iω0a X h0|Q̂`B mB |bihb|Q̂`0 0 |0iω0b (9.32)
B mB
X
A A
× dω 2 + ω2 2 + ω2
ω0a ω0b
a6=0 b6=0
Dynamic (frequency-dependent) multipole polarizabilities in molecule-fixed frame

X 2ω0n h0|Q̂`m |aiha|Q̂`0 m0 |0i
α`m,`0 m0 (ω) = 2 − ω2 (9.33)
ω0a
n6=0
related to the charge density susceptibility

ZZ
α`m,`0 m0 (ω) = drdr 0 R`m (r)α(r, r 0 |ω)R`0 m0 (r 0 ) (9.34)
Dispersion energy
X
E (2) (disp) = − TÀ mA ,`B mB T`0A m0A ,`0B m0B XÀB
A mA `B mB `
0 0 0 0 (9.35)
A mA ,`B mB
The Casimir-Polder or dispersion integrals are defined as

Z
AB ~
XÀ mA `B mB `0 m0 ,`0 = dωαÀA mA ,`0 m0 (iω)α`BB mB ,`0 m0 (iω) (9.36)
A A B 2π A A B B
The dispersion integrals can be calculated by numerical quadrature as

M
~ X
X AB = w(ωj )αA (iωj )αB (iωj ) (9.37)
2π
j=1
With a Gauss-Chebyshev quadrature scheme the grid points are chosen as

2j − 1
ωj = cot π (9.38)
4M
and the weights
2π
w(ωj ) = (9.39)
4M sin2 (π(2j − 1)/4M )
80
Typically, with 5 and 7 grid points one has an accuracy of 0.1 %.
Numerically more efficient procedure (for high-rank calculations) is to contract first the
polarizabilities with the interaction tensors (in local frame) and perform the numerical
integration afterwards (Hättig, 1996).
In this form the interaction tensors and X AB are both reducible: they transform
according to the double and quadruple product group of SO(3). In terms of irreducible
tensors, the dispersion coefficients are of the form (Wormer)
X X
C`LAA`L0 B`BL`0 = XÀB 0 0 0
A mA `B mB ` m ,` m
0 · factor (9.40)
A B A A B B
mA m0A mB m0B
The algebraic factor depends on the Wigner 3j and 9j coefficients. These coefficients
are coupled to the following form
CnLA LB L (n = À + `B + `0A + `0B + 2)
For atoms in S-state, for example,

∞
(2)
X
Edisp =− C2n R−2n
n=3
n−2
X
C2n = C(`; n − k − 1) (9.41)
`=1
(2` + 2`0 )! 1
Z
0
C(`; ` ) = dωαÀ (iω)α`B0 (iω)
(2`)!(2`0 )! 2π
The dispersion integral depends on the properties of the individual molecules and
independent of the intermolecular geometry. The complete expression can be expressed
in terms of irreducible spherical tensor components.
◦ dipole-dipole leading term R−3 R−3 ∼ R−6
◦ dipole-quadrupole would be R−3 R−4 ∼ R−7 ,

after averaging over all orientations, it gives zero
◦ quadrupole-quadrupole term R−4 R−4 ∼ R−8
◦ dipole-octopole and quadrupole-qudrupole R−10
9.5.1 Dipolar dispersion energy

The leading multipolar term in the dispersion interaction energy
X X h00|µ̂A B A B
α Tαβ µ̂β |abihab|µ̂γ Tγδ µ̂δ |00i
(2)
E (disp) = − A + ∆E B
(9.42)
a6=0 b6=0
∆E0a 0b
81
◦ Casimir-Polder formula
∞ ∞
h0|µ̂A A
α |aiha|µ̂γ |0iω0a h0|µ̂B B
β |bihb|µ̂δ |0iω0b
Z
(2) 2 X X
E (disp) = −Tαβ Tγδ dω 2 2
~π a6=0
ω0a + ω2 b6=0
ω0b + ω2
(9.43)
◦ Unsöld approximation
∞ X ∞ A B h0|µ̂A |aiha|µ̂A |0i h0|µ̂B |bihb|µ̂B |0i
X ∆E0a ∆E0b α γ β δ
E (2) (disp) = −Tαβ Tγδ A B A B
a6=0 b6=0
∆E0a + ∆E0b ω0a ω0b
(9.44)
In order to factor this latter expression, we use the identity
A ∆E B
∆E0a 0b UA UB
A B
= (1 + ∆ab ) (9.45)
∆E0a + ∆E0b UA + UB
with
A + 1/U − 1/E B
1/UA − 1/E0a B 0b
∆ab = A + 1/E B
(9.46)
1/E0a 0b
that can be made negligibly small by choosing appropriate average excitation

energies UA and UB
UA UB
E (2) (disp) ≈ − A B
Tαβ Tγδ ααγ αβδ (9.47)
4(UA + UB )
In both expressions we can separate the orientation-independent spherically averaged

and various orientation-dependent components, by using the decomposition of the
polarizabilities to irreducible parts. The spherically averaged component becomes in
both cases
6
Tαβ Tγδ αA δαγ αB δβδ = αA αB Tαβ Tαβ = αA αB 6 (9.48)
R
leading to the general expression
C6
E (2) (disp) ≈ − (9.49)
R6
with the C6 coefficient
Z
C6 = 3~ dωαA (iω)αB (iω) (Casimir-Polder)
(9.50)
3UA UB
C6 ≈ αA αB (London)
2(UA + UB )
Experimental oscillator strength distributions can be used to determine “experimental”

C6 coefficients.
Roughly proportional to the square of the polarizability/volume.
Some typical values
82
system C6 C8 C10
H· · · H 6.5 124.4 1135
He· · · He 1.46 13.9 182
Ne· · · Ne 6.6 57 700
Ar· · · Ar 64.3 1130 25000
Kr· · · Kr 133 2500 60000
Xe· · · Xe 286
Combining rules can be deduced from the London-formula

q
C6AB ≈ C6AA C6BB (9.51)
9.5.2 Approximate dispersion formulae

The average excitation energy of the London formula is an empirical parameter, either
the first ionization potential (quite bad) or the first excitation energy (somewhat better)
is used.
Similar expressions can be derived from the Casimir-Polder expression, by considering
some general properties of the average dynamic dipole-dipole polarizabilities
1 X 2ω0k |h0|x̂|ki|2
α(iω) = 2 + ω2 (9.52)
~ ω0k
k6=0
We are looking for the simplest, one-term approximation in the form

ω2
α(iω) ≈ ·a (9.53)
ω2 + ω2
Parameters a and ω will be found from the asymptotic behaviour of α(iω).
For ω = 0 one gets the static polarizability, a = α(0).
For ω → ∞ one gets the Thomas-Reiche-Kuhn sum rule, the number of electrons
1 X n
α(iω) → 2 2 2∆E0n x20n = 2 2 (9.54)
~ ω ~ ω
n6=0
In this limit we have

a n
=
ω 2 /ω 2 ~2 ω 2
n
ω2 = (9.55)
~2 α(0)
r
1 n
ω=
~ α(0)
which leads to the Mavroyannis-Stephen (Slater-Kirkwood) approximation
3αA αB
C6 ≈ (9.56)
(αA /nA )1/2 + (αB /nB )1/2
83
Take another, equivalent form of the dynamic polarizability
1 X |r on |2 1 X |r on |2
α(ω) = α+ (ω) + α+ (−ω) = + (9.57)
3~ ω0n + ω 3~ ω0n − ω
n6=0 n6=0
For ω = 0
1
α+ (ω) = α(0) (9.58)
2
For ω → ∞
1X 1 1
~ωα+ (ω) → |r 0n |2 = h0|r 2 |0i − h0|r|0i2 = (∆r)2 (9.59)
3 3 3
n6=0
Considering a one-term approximation

a0
α+ (ω) = (9.60)
~(ω 0 + ω)
Taking the limiting cases one obtains
1 2 (∆r)2
a0 = α(0) ω= (9.61)
2 3~ α
and we get the Salem-Tang-Karplus approximation
αA αB
C6 = (9.62)
αA /(∆r A )2 + αB (∆r B )2
For dimers it is identical to the Alexander upper bound
3 1
C6 ≤ S(−2)S(−1) = α(∆r B )2 (9.63)
4 2
9.5.3 Solutions of the H + H dispersion problem

The multipolar interaction Hamiltonian for two H atoms lying on the z axis, separated
by a distance R0 is
V̂ = − 2[R0−3 ξ1 ξ2 cos θ1 cos θ2

+ βξ1 ξ22 cos θ1 (3 cos2 θ2 − 1) (9.64)
+ γξ12 ξ22 (3 cos2 θ1 2
− 1)(3 cos θ2 − 1) + . . .
√
= r1,2 , α = ( 6/2)R0−3 , β = ( (30)/4)R0−4 and
p
with polar
√ coordinates ξ1,2
γ = ( 70/8)R−5 .
By direct summation over the excited states, Eisenchitz and London obtained for the
dipolar dispersion energy
12 X ∆E z 2 ∆E z 2
E (2) = − 0n 0n 0m 0m (9.65)
R06 n,m

1 − n12 1 − m12 2 − n12 − 1
m2
84
It is difficult to make converge this expression, because of the discrete-continuum
matrix elements. The best value is C6 = 6.47.
The difficulties of the sum-over-states solution can be avoided by solving directly for the
first-order wave function using the Ansatz proposed by Slater and Kirkwood,
ψ(r 1 , r 2 ) = ψ0 (r1 , r2 )[1 + φ(r 1 , r 2 )] (9.66)
with the ground state unperturbed wave function of the dimer, ψ0 (r1 , r2 ). The
two-particle correlation function, φ, satisfies the differential equation
1 2
∇ φ + (∇ ln ψ0 ) · ∇φ) − v = 0 (9.67)
2
where v can be one of the multipolar terms of the interaction Hamiltonian. Taking the
correlation function in the form
vR(ξ, ξ 0 )
φ= (9.68)
E0
leads to a differential equation for R(ξ, ξ 0 ).
Slater and Kirkwood (1931) obtained an approximate solution, leading to C6 = 6.23.
Pauling and Beach (1935) used special orbitals to construct the Hamiltonian matrix and
obtained C6 = 6.49903, C8 = 124.399 and C10 = 1135.21. Recent exact solutions
obtained by Choy (P.R.A. 62 (2000) 012506) using orthogonal polynomials, confirm
these values.
9.6 Convergence of the 1/R-expansion

◦ The 1/R expansion of the total interaction energy converges asymptotically to the
ground state interaction energy.
◦ For each finite R, however, the 1/R expansion may diverge. It had been proven for
H2 molecule and H+ 2 ion.
◦ The 1/R expansion of the second-order polarization energy for H+
2 ion (induction
energy) diverges for all finite R.
◦ The 1/R expansion of the second-order polarization energy for H2 molecule, i.e. its
dispersion energy, diverges for all finite R.
◦ Given a finite one-particle GTF or STF basis set, the multipole expansion of each
order of the polarization series converges, as long as the smallest spheres around
the respective nuclei and basis function centers do not touch. However, the limit of
this convergence may be quite different from the exact interaction energy.
85
86
Chapter 10
Supermolecule method
10.1 Overview of the supermolecule approach

◦ interaction energy is given by a very small difference of large numbers
∆E(R) = E AB (R) − E A − E B (10.1)
◦ advantages
– straightforward
– the whole PES can be treated on an equal footing
◦ drawbacks
– small difference of large numbers

– systematic errors are much larger than ∆E itself
– limited to size-consistent methods
– difficult to interpret
10.2 Choice of the method

◦ semiempirical methods: inappropriate
◦ DFT: problems with dispersion (see later)
◦ ab initio
– Hartree-Fock: misses dispersion, but reasonable for polar systems, H-bonds
– Configuration Interaction: inappropriate due to size inconsistency (excepted

full CI)
– MBPT: at low orders (MP2) polarizabilities are of only UCHF quality
– couple cluster: CCSD(T) is considered as the best compromise
87
10.3 Choice of the basis set
Requirements:
◦ First-order energy: correct multipoles and tails of monomers → diffuse functions

◦ Second-order energy: good (static and dynamic) polarizabilities → polarization
functions
– dipole polarizabilities: d functions

– quadrupole polarizabilities: f functions
– ...
◦ balanced basis set: valence and polarization part of similar quality
– good valence - poor polarization : underestimates induction

– poor valence - good polarization : overestimates induction
Precise calculations require very large basis sets, (multiple zeta quality + f,g,h orbitals).
High angular momentum functions can be replaced by bond functions.
10.4 BSSE
Since the basis functions stick on the atoms, one has a different basis set (of unequal
quality) in the function of the intermolecular separation/orientation.
∆E = EA∪B (R, Ω) − EA∪B (R = ∞, Ω) = EA∪B (R, Ω) − EA − EB (10.2)
At small separations the partner basis functions improve the monomer description
(nothing to do with the physical interaction) and lead to an extra stabilization of the
complex.
Boys-Bernardi counterpoise correction:
At each geometry evaluate monomer energies in the dimer basis
∆E = EA∪B (R, Ω) − EA (A ∪ B|R, Ω) − EB (A ∪ B|R, Ω) (10.3)
Problems: e.g. monomer properties in the dimer basis have lower symmetry
88
10.4.1 Nitrogen dimer
without BSSE correction with BSSE correction
0.4 0.4
0.2 0.2
RHF
0 RHF 0
MP2
-0.2 -0.2
-0.4 MP2 -0.4

6-311+G(d)
3 3.5 4 4.5 5 5.5 6 3 3.5 4 4.5 5 5.5 6
0.4 0.4
0.2 0.2
RHF
0 RHF 0
MP2
-0.2 -0.2
AUG-cc-pVTZ -0.4 MP2 -0.4
3 3.5 4 4.5 5 5.5 6 3 3.5 4 4.5 5 5.5 6
10.4.2 Basis incompleteness correction

It is more judicious to consider the CP correction as a basis set incompleteness
correction. If the monomer basis were complete, the difference
EA (A ∪ B|R, Ω) − EA (10.4)
would vanish. For incomplete basis set, this error (per monomer) represents an extra
stabilisation, that should be subtracted from the total energy:
corr
EA∪B (R, Ω) = EA∪B (R, Ω) − (EA (A ∪ B|R, Ω) − EA ) − (EB (A ∪ B|R, Ω) − EB ) (10.5)
New total energy functional, that can be used to calculate BSSE-corrected dimer
properties, and one can even define BSSE-free effective Hamiltonian. (Mayer, 2003)
The interaction energy is defined with respect to the “true” monomer energies:
corr
∆E = EA∪B (R, Ω) − EA − EB (10.6)
89
10.4.3 BSSE-free geometry optimization
0.2
0.1
-0.1
-0.2
-0.3
-0.4
-0.5
3 3.5 4 4.5 5 5.5 6
◦ BSSE correction at the minimum of the uncorrected surface is larger than in the
true minimum.
◦ Find the minimum of the corrected dimer energy!
Gradient optimizations using ”composite” energy expression (and gradients).
Fig. 1 Structure of the water dimer.
Basis set PES R r1 α β ∆E

6-31G** CP-corr 2.991 0.967 2.7 129.5 -21.21
Standard 2.913 0.967 9.8 99.1 -19.41
cc-pVDZ CP-corr 3.048 0.969 -0.1 138.2 -17.61
Standard 2.912 0.970 5.7 104.5 -15.02
cc-pVTZ CP-corr 2.973 0.965 3.6 129.6 -18.70
Standard 2.908 0.966 5.1 115.6 -18.28
cc-pVQZ CP-corr 2.935 0.964 4.6 127.2 -19.71
Standard 2.901 0.965 4.8 121.0 -19.62
aug-cc-pVDZ CP-corr 2.977 0.972 5.7 122.3 -18.62
Standard 2.921 0.973 6.4 119.6 -18.45
aug-cc-pVTZ CP-corr 2.933 0.968 5.5 124.7 -20.21
Standard 2.905 0.969 4.9 124.3 -19.75
aug-cc-pVQZ CP-corr 2.917 0.966 5.6 124.8 -20.59
Standard 2.903 0.966 5.1 124.7 -20.54
Hobza et al. Phys. Chem. Chem. Phys., 1999, 1, 3073-3078
90
10.5 Size consistency problem
H2 dimer
minimal basis
0.005
0.0045
0.004
0.0035
0.003
0.0025
0.002
0.0015
0.001 CI
0.0005 RHF
0
2 2.5 3 3.5 4 4.5 5
◦ CISD interaction energy does not tend to zero!
◦ CISD monomer wf:

2 2
σgA − λσuA
◦ product wf of the dimer:

2 2 2 2
(σgA − λσuA )(σgB − λσuB )
◦ contains up to quadruple excitations:

2 2
λ2 σuA σuB
◦ unbalanced description of monomer and dimer
10.5.1 Size-consistency of MBPT methods

N-mer of non-interacting H2 molecules
Second order energy
(2)
N
X |hΨ|Ŵ |Ψ21i 2̄1̄i i|2
i i
2
N K12
E = = (10.7)
i=1
2(ε1 − ε2 ) 2(ε1 − ε2 )
is N times the monomer MP2 energy.
91
Third order energy correction
2
N K12 (J11 + J22 − 4J12 + 2K12 )
E (3) = (10.8)
4(ε1 − ε2 )2
The general result is true: the MPn energy correction is size-consistent.
10.5.2 Other size-consistent methods

The size consistency can be qualitatively discussed by invoking the one- and
two-particle excitation operators
X X
T̂1 = Tar a+
r a a T̂2 = rs + +
Tab ar as ab aa (10.9)
ar abrs
◦ SDCI wave function of a complex
(1+T̂1A +T̂2A )|ΨA i(1+T̂1B +T̂2B )|ΨB i = |ΨAB (SDCI)i+(T̂1A T̂2B +. . .)|ΨA ΨB i
which is not of SCDI form
◦ coupled cluster wave functions (CCSD) of a complex
exp (T̂1A + T̂2A )|ΨA i exp (T̂1B + T̂2B )|ΨB i = exp (T̂1A + T̂1B + T̂2A + T̂2B )|ΨA ΨB i
is still of the CCSD form.
92
Chapter 11
DFT and intermolecular forces
11.1 DFT description of intermolecular

complexes
◦ History: Gordon-Kim
◦ Diagnosis: present-day functionals
◦ Remedies
– Parameterized (hybrid) functionals

– Adiabatic connection formula – response functions
∗ asymptotic theories
∗ seamless functionals
– DFT+Edisp approaches
11.2 Gordon-Kim method

Simple DFT method to calculate short-range interaction energy of complexes
(typically rare gas dimers) in the overlap region.
The density is a simple superposition
ρAB = ρA + ρB (11.1)
of Hartree-Fock densities of the monomers.

The interaction energy is calculated from the Thomas-Fermi kinetic energy
functional
3
Ek [ρ] = CF ρ5/3 = (3π 2 )2/3 ρ5/3 (11.2)
10
93
and Dirac exchange functional
3 3 1/3 4/3
Ex [ρ] = −Cx ρ4/3 = − ρ (11.3)
4 π
and an interpolated uniform electron gas formula for the correlation energy.
Z
5/3 5/3 5/3
∆E =CF ρAB (r) − ρA (r) − ρB (r) dr
Z Z Z
ZA Z B ZA ZB
+ − + ρAB (r)dr + ρA (r)dr + ρB (r)dr
rA rB rA rB
(11.4)
Z A ZB
+ + J[ρAB ] − J[ρA ] − J[ρB ]
R
Z
4/3 4/3 4/3
− Cx ρAB (r) − ρA (r) − ρB (r) dr + Ec [ρAB ] − Ec [ρA ] − Ec [ρB ]
Improvements were necessary for the self-interaction correction (SIC), but the
overall agreement with experiment is quite satisfactory.
System r0 (calc.) r0 (exp.) ∆E ∆E (exp.)
He-He 2.49 2.96 62.5 16.5
Ne-Ne 2.99 3.09 56 63
Ar-Ar 3.62 3.70 175 195
Kr-Kr 3.89 3.87-4.08 248 270-340
Xe-Xe 4.15 4.5 417 342-360
11.3 Is dispersion missing from DFT?

◦ First studies by Pulay (1994) and Becke (1995)
– LDA severely overbinds
– GGA is repulsive
◦ Role of exchange functional; Zhang (1997)
– LDA overbinds
– B88-PW91 repulsive
– B86-PW91, PW91-PW91, PBE-PW91 yields quite good results
– similar results with LYP correlation functional
– choice of exchange is crucial
◦ Role of correlation functional; Pérez-Jordá (1999)
– HF+LYP and HF+Becke overbinds
– HF+Wilson-Levy too short equilibrium, but reasonable
– still wrong asymptotical behaviour
94
11.4 Comparison of some functionals
Wesolowski reported some interaction energies of the CO2 · · · He complex
Method 0 deg 30 deg 60 deg 90 deg
SAPT 90.88 12.39 -26.14 -26.38
CCSD(T) 93.62 15.83 -24.72 -25.70
BPW91 409.24 244.66 87.84 47.01
BLYP 231.29 125.37 48.76 33.24
B3LYP 164.79 79.05 25.69 17.83
PW91PW91 22.2 -62.3 -86.1 -70.4
LG86 58.6 5.3 -4.3 1.9
Hybrid functional and the PW91PW91 and the Lacks-Gordon functionals are
much better than the BPW91 and BLYP functionals. This can be explained by
the incorrect long-range behaviour of the B88 exchange functional.
11.5 Modern theories of intermolecular forces

in DFT
Recent attempts to remedy the dispersion energy problem:
One has to go ”beyond” conventional Kohn-Sham DFT theory
11.5.1 Adiabatic connection formula

Exact expression of the correlation energy by the ACF
e2
ZZ
1
Ec [ρ] = d3 rd3 r 0
2 |r − r 0 |
Z 1 (11.5)
dλ h ρ̂(r) − ρ(r) ρ̂(r 0 ) − ρ(r 0 ) iλ − δ(r − r 0 )hρ̂(r)i

×
0
where h. . .iλ means that the integration is performed with a potential Vλ that
keeps the density equal to ρ(r). The spontaneous fluctuations of the system can be
related to the ground state linear response function via the fluctuation-dissipation
theorem
~ ∞
Z 1
e2
ZZ Z
1 3 3 0 0 0

Ec [ρ] = d rd r × dλ dω χ λ (r, r ; iω) − χ 0 (r, r ; iω)
2 |r − r 0 | 0 π 0
(11.6)
The density-density response function (charge density susceptibility) for a KS

system (Vλ=0 = V KS (r)) is
X (fj − fk )
χ0 (r, r 0 ; ω) = ϕj (r)ϕk (r)ϕk (r 0 )ϕj (r 0 ) (11.7)
ε j − εk − ~ω
j,k
95
The interacting response is given by the screening equation
Z
χλ (r, r ; ω) = χ0 (r, r ; ω) + d3 r 00 Qλ (r, r 00 ; ω)χλ (r 00 , r 0 ; ω)
0 0
(11.8)
with the kernel

λe2
Z
00
drχ0 (r, r 0 ; ω) + fxcλ (r, r 00 ; ω)

Qλ (r, r ; ω) = 00
(11.9)
|r − r |
The RPA response function is obtained by setting the exchange-correlation kernel,
fxc = 0. It is important that the KS solutions satisfy the self-interaction condition.
11.5.2 Asymptotic dispersion energy functionals

In the case of non-overlapping (distinguishable) subsystems the screening equation
can be solved by second order PT with respect to the Coulomb potential that
couple the subsystems
λe2 λe2
ZZ ZZ
~ 0
(2)
E =− dr 1 dr 1 dr 2 dr 02
2π a b |r 1 − r 2 | |r 01 − r 02 |
Z ∞ (11.10)
0 0
× dωχa,λ=1 (r 1 , r 1 ; iω)χb,λ=1 (r 2 , r 2 ; iω)
0
This is essentially the Casimir-Polder formula. All the approximate and multipolar
forms, discussed previously, apply.
11.5.3 Dispersion functional from the electron gas

The electronic displacement, p(r, t) in an oscillating external field is given by the
dynamic polarizability tensor
Z
pα (r, t) = exp(−iωt) dr 0 ααβ (r, r 0 ; iω)Eβext (r 0 ) (11.11)
and it is related to the charge density susceptibility by

1 ∂2
χ(r, r 0 ; ω) = − ααβ (r, r 0 ; ω) (11.12)
e2 ∂rα ∂rβ
Dobson and Dinte considered the approximate polarizability of the homogeneous
electron gas
n0
ααβ (q; ω) ≈ δαβ (11.13)
m[ω − ωP2 (n0 )]
2
with the uniform number density, n0 and the plasma frequency,

ωP = (4πn0 e2 /m)1/2 . Taking the space Fourier transform
n0
ααβ (r − r 0 ; ω) ≈ δαβ δ 3 (r − r 0 ) (11.14)
m[ω 2 − ωP2 (n0 )]
96
the simplest generalization to an inhomogeneous system
ρ(r)
ααβ (r, r 0 ; ω) ≈ δαβ δ 3 (r − r 0 ) (11.15)
m[ω 2 − ωP2 (ρ(r))]
Applying the relationship between the charge density response function and the
polarizability

0 1 ρ(r) 3 0
χ(r, r ; ω) ≈ − 2 ∇r ∇r0 δ (r − r ) (11.16)
e m[ω 2 − ωP2 (ρ(r))]
Substitution into the second order formula, integrating by parts and performing
the frequency integral analytically, we get for two non-overlapping systems
p
3~e1/2
Z Z
(2) 1 ρa (r 1 )ρb (r 2 )
E ≈− 3/2 1/2
dr 1 dr 2 6 p p (11.17)
4(4π) m a b r12 ρa (r 1 ) + ρb (r 2 )
This expression gives a severely overestimated dispersion energy for atoms, unless
one introduces an integration cutoff (Rapcewitz & Ashcroft, 1991; Andersson,
1998).
11.5.4 Seamless dispersion energy functionals

The asymptotic formula does not apply to the overlap region: no clear separation
of the dispersion and other correlation effects (already taken into account by the
local and/or gradient corrected functional). Kohn, Meir and Makarov (1998)
proposed to split to Coulomb potential to a short- and a long-range part
1
= Usr (|r − r 0 |) + Ulr (|r − r 0 |) (11.18)
|r − r 0 |
Possible choices
◦ Yukawa potential
exp(−κr)
Usr =
r
◦ Error function (Ewald)
erfc(κr)
Usr =
r
The adiabatic connection formula was applied with λUlr and the exact
exchange-correlation energy formula was obtained as
(sr) 1
Exc = Exc − Ulr (0)N + Epol (11.19)
2
97
(sr)
The quantity Exc is the exchange-correlation energy with only the short-range
part of the Coulomb potential and the “polarization energy” is given by
ZZ Z ∞ Z 1
~ 3 0 0
Epol = − 3
d rd r Ulr (|r − r |) dω dλImχ̃λ (r, r 0 , ω) (11.20)
2π 0 0
The response function, χ̃ is calculated from the screening equation with Ulr instead
of the full Coulomb interaction.
The asymptotic behaviour is obtained after transformation to the time domain,
leading to the C6 coefficient
αa (t1 )αzz
b
Z Z
3 (t2 )
C6 = dt1 dt2 zz (11.21)
π t1 + t2
R
with αzz = dr1 dr2 χ(r1 , r2 )z1 z2 . The time-dependent polarizabilities were
obtained from a stepwise integration of the time-dependent Schrödinger equation
for H and He with the
◦ exact xc-functional: very good agreement with the best values (He–He 1.45
(1.458); H–He 2.81 (2.817))
◦ LDA xc-functional: 28% too high C6 for He dimer.
11.5.5 Goerling-Levy perturbation theory

The following leading correlation energy term has been obtained from the
adiabatic connection formula by Görling and Levy
Ec(2) =Ec∆HF + EcMP2 =

occ X
virt
X |hj|v̂ N L − v̂x |vi|2
x
= +
j v
εj − εv (11.22)
occ virt
1 X X |hjk|vwi − hjk|wvi|2
+
4 jk vw εj + εk − εv − εw
Provided that the subsystems are distinguishable the leading intermolecular

perturbation correction leads to an approximate dispersion energy
X X X X |(φa φa0 |φb φb0 )|2
Edisp ≈ (11.23)
a a0 b b0
∆εaa0 + ∆εbb0
Asymptotic dispersion energies (C6 coefficients) were determined by Görling et al.

(Theochem, 501 (2000) 271).
98
11.5.6 DFT + E(disp) method
In a more pragmatic spirit, several authors implemented damped multipolar
expansions of the dispersion energy
X C6ab
E(disp) = − f6 (Rab ) · 6
(11.24)
ab
Rab
where the atom-atom dispersion coefficients are determined from atomic

polarizabilities (via London-type formulae) and combining rules. The damping
function may take, e.g. the Tang-Toennis form
n
X (bR)k
fn (R) = 1 − exp(−bR) (11.25)
k=0
k!
Some recent examples
◦ Elstner et al. JCP 114(2001) 5149

◦ Wu at al. JCP 115 (2001) 8748
◦ Wu and Yang JCP 116 (2002) 515
11.5.7 DFT-SAPT method

Williams and Chabalowski (JPC A 105 (2001) 646) proposed to use Kohn-Sham
orbitals obtained from standard DFT calculations in the SAPT correction
formulae (without intramolecular correlation terms)
(10) (20) (20) (20) (20)
∆E(DFT-SAPT) = Epol + Eind + Edisp + Eexch-ind + Eexch-disp (11.26)
◦ double counting of correlation effects is avoided
◦ response properties are calculated in the uncoupled version (KS response

function)
◦ appropriateness of KS eigenvalues is questionable
99

Janee

Uploaded by

Copyright:

Available Formats

Janee

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Janee

Uploaded by

Copyright:

Available Formats

Intermolecular forces

and their representation in

2 Experimental sources of intermolecular potentials 13

4 Perturbation theory for intermolecular interactions 35

5 Beyond the polarization approximation 43

6 Molecules in electrostatic field 53

7 Multipole expansion – Cartesian formalism 55

9 Long range perturbation theory 75

11 DFT and intermolecular forces 92

1.1 What are intermolecular forces?

◦ what are the optimal intermolecular

◦ what is the physical origin of

◦ fragments (subsystems) are well-defined and preserve their identity

Etot (∞) = EA + EB (1.1)

At an arbitrary, finite R distance the total energy

Etot (R) = EA + EB + U (R) (1.2)

U (R) = Etot (R) − Etot (∞) = Etot (R) − EA − EB (1.3)

◦ σ is the hard core

◦ attractive for R > Rm

◦ for σ < R < Rm nega-

1.3 Cluster expansion of the interaction energy

◦ one-body: geometrical deformation of the monomers

◦ two-body: dominating term (additivity)

Effective (system-dependent) pair-potential

may be quite different from the pure pair-potential.

Range short intermediate long

1.5 Calculation of intermolecular interaction en-

expanded in power series of the strength of interaction

– leads directly to the interaction energy

◦ physico-chemical properties: bulk modulus, phonon spectrum, etc.

2.2 Trouton’s rule

Empirical relationship between enthalpy of vaporization, ∆Hvap and boiling point,

Tb /K n (20Tb /n)/K (ε/kB )/K ε/kB /kJ mol−1 εlit.

Approximate the entropy change by the liquid/gas volume change:

∆Svap = R ln (Vg /Vl ) ≈ R ln 1000 ≈ 7R (2.3)

The remaining contribution of 3R is due to liquid structure.

2.2.1 Well-depth from Trouton’s rule

ε/kB ≈ 20Tb /n (2.4)

2.4 Lattice energies

◦ NaCl crystallizes in fcc lattice

◦ lattice parameter a = 5.64Å

◦ ions are in (0,0,0) and ( 12 , 12 , 12 )

◦ lattice energy Ulatt = −764.4 kJ/mol

The lattice sum (Madelung energy) is conditionally convergent, so special techniques

◦ Condition of equilibrium in the minimum of the lattice energy

◦ Ar crystallizes in fcc lattice

◦ lattice parameter a = 5.3109Å

◦ ions are in (0,0,0) and ( 12 , 12 , 12 )

◦ lattice energy Ulatt = −8.4732 kJ/mol

Cohesion energy is the sum of pair potentials

Lattice sum for the 6- and 12-potentials

Pi 1 5 fcc(∞) hcp (∞)

can be expressed in the function of ε

2.4.3 Three-body forces and the structure of rare gas solids

2.5 Equation of state – virial coeffiicients

◦ excluded volume in a binary collision 43 πσ 3

◦ reduction of pressure due to intermolecular attraction: the number of binary

Virial equation of state – power series of (1/V )

2.5.2 Equations of state - generalities

Define generalized forces corresponding to external parameters

Partition function 3N/2