An introduction to Relativistic Quantum Chemistry

Lucas Visscher Vrije Universiteit Amsterdam

The extra dimension

Hamiltonian Dirac-Coulomb-Breit

Basisset

Complete

NR Minimal Hartree-Fock

Method Full CI

Course outline: lecture 1+2

Relativistic Quantum Chemistry

Special Relativity The Dirac equation

Free particles Second quantization and QED : a short detour Hydrogenic atom

Approximate Hamiltonians

Breit-Pauli perturbation theory The regular approximation (ZORA) The Douglas-Kroll-He method Four-component methods Direct perturbation theory

Course outline: lecture 3

Effective Core Potentials Basic assumptions Ab initio Model Potentials Energy-Consistent Pseudopotentials Shape-Consistent Pseudopotentials Computational aspects

All-electron or valence-only ? Wave Function Theory or Density Functional Theory ? Spin-orbit or scalar relativistic ?

Relativistic effects in chemistry

Dissociation energies Bond lengths and bond strengths Dipole moments NMR shieldings Electric Field Gradients
4

Coordinate transformation

Galilean transformation
x = x " + vt y = y" z = z"

dx d( x " + vt) dx " w= = = + v = w" + v dt dt dt

Simple addition of velocities, no speed limit

137
Maxwell

Thought experiment 1

Two rotating double stars A and B

Does their light reach earth at different times ? Do we observe one star at two positions ? NO -> The speed of light (c) does not depend on the motion of the emitting stars Is there some immobile substance (ether) that transmits the radiation? NO -> Need better theory of mechanics
7

Thought experiment 2

Take two observers inside and outside a moving train The train passes the stationary observer, waiting for the railroad sign, on its way to a nearby tunnel... They both know the speed of light and wonder when the light of the railroad sign will illuminate the tunnel The observer outside has an easy job : t = distance / c The observer inside needs to correct for the fact that the tunnel is moving towards him (and the light) and gets a slightly smaller t Their conclusion: with c constant, t needs to be relative

Special relativity
c constant and t variable gives
x 2 + y 2 + z 2 = c 2t 2 x "2 + y "2 + z "2 = c 2 t "2

Galileo
!
x = x " + vt y = y" z = z"

We need a new transformation

x = " ( x # + vt #) y = y# z = z# t = $ ( t # + %x #)
Scaling factor No dependence on y and z since clocks in the yz plane would disagree (reciprocal relation between the frames)

Lorentz transformation

Substitute this ansatz in the unprimed equations and solve v 2 $1/ 2 v " = # = (1$ 2 ) %= 2 c c Lorentz transformation
x = " ( x # + vt #) y = y# Generalize to 3d z = z# $ vx # ' t = "& t # + 2 ) % c ( Time and spatial coordinates transform into each other 4-dimensional space-time coordinate system ! Nonrelativistic limit (c ) gives Galileo transformation

& ( v # r ")($ %1) ) r = r " + v( + $t "+ 2 v ' * & ( v # r ") ) t = $ (t" + + 2 c * '

10

Relativistic Quantum Mechanics

1905 : STR

Einstein : E = mc2 Schrdinger equation Dirac equation Tomonaga, Schwinger & Feynman

1926 : QM

1928 : RQM

1949 : QED

11

Non-relativistic quantization
The nonrelativistic Hamiltonian function
"2 H =T +V = + q# (r ) 2m " = p $ qA
Quantization
!

# H " ih ; p " $ih% #t & (r,t) = ih # & (r,t) H #t h 2 iqh q2 2 H =$ % + %'A +A'% + A + q( (r) 2m 2m 2m

12

Spin and non-relativistic quantization 1

We can also write the the Hamiltonian function as

2m # i# j = &ij + i'ijk # k
Quantization

(# $ % ) E = q" +

Kronecker delta and Levi-Civita tensor

2 1 H = q" + # $ %ih& + qA 2m 2 2 h2 q2 % + + iqh # $ & , # $ A = q" #$& #$A 2m 2m 2m

{ (
(

)}

[(

)(

)]

13

Spin and non-relativistic quantization 2

(! " u)(! " v) = (u " v) + i! " (u # v )
h 2 q2 2 H =" # + q$ + A 2m 2m iqh qh + #%A +A%# " & % #'A +A '# 2m 2m

)
chain rule

" # A (r ) f (r) = " # ( f (r)A (r ))

!

A is a multiplicative operator

= "f (r) # A (r ) + f (r)" # A (r )

= $A # "f (r) + Bf (r)

Use definition of B

q2 2 q H = T + q" + iqA # $ + A % & # B 2 2

in atomic units

Spin in NR quantum mechanics

The Pauli Hamiltonian in two-component form
& 1 2 q2 2 q A " Bz (" # + q$ + iqA % # + 2m 2 ( 2 q ( " ( Bx + iBy ) ' 2 ) q Bx " iBy ) + ( 2 + 1 2 q2 2 q + " # + q$ + iqA % # + A + Bz 2 2m 2 * "

Second derivatives w.r.t. position, first derivative w.r.t. time Linear in scalar, quadratic in vector potential Can not be Lorentz-invariant Ad hoc introduction of spin.The anomalous g-factor (ratio magnetic moment to the intrinsic angular momentum) is not well explained No spin-orbit coupling
15

Relativistic quantization 1
Take the classical relativistic energy expression
E " q# = [ m c + c $
2 4 2 2 1/ 2

" E = mc 2 "

Without EM-fields

Quantization recipe gives "# 2 4 2 2! ih = m c + c $ # % q&# "t After series expansion of the square root this could provide relativistic corrections to the Schrdinger Equation

!
Disadvantage : Difficult to define the square root operator in terms of a series expansion (A and p do not commute). Not explored much.
16

Relativistic quantization 2
Eliminate the square root prior to quantization

( E " q# ) = m 2c 4 + c 2$ 2
Quantization

% " (2 2 4 2 2 'ih # q$ * + = ( m c + c , )+ & "t ) Klein-Gordon Equation

Lorentz invariant No spin # " * (r )" (r )dr = f (t)
!

Charge is conserved, particle number is not

The KG-equation is used for mesons (that have no spin)

17

Relativistic quantization 3
Define a new type of square root

E " q# = $mc 2 + c % & '

[% i,% j ]+ = 2(ij )
Quantization

[% i , $ ] + = 0

) $2 =1

"# ih = $mc 2 + c% & ' + q( # "t

The Dirac equation

Suitable for relativistic description of electrons

18

The Dirac equation

(' ( r,t ) ("mc + c # $ % + q& )' (r,t ) = ih (t
2

First derivatives with respect to time and position Linear in scalar and vector potentials
!

Can be shown to be Lorentz invariant Alpha and Beta are conventionally represented by the following set of 4-component matrices
$ 0 #y' $ 0 #z' $ 0 #x' $I 0 ' "x = & ) "z = & ) * =& ) "y = & ) #y 0 ( #z 0 ( #x 0 ( 0 +I ( % % % %

19

The Dirac Hamiltonian

H = "mc 2 + c # $ % + q& ( mc 2 + q& 0 c% z c(% x ' i% y )+ * 2 0 mc + q& c(% x + i% y ) 'c% z * = * c% z c(% x ' i% y ) 'mc 2 + q& 0 * 2 'c% z 0 'mc + q& , )c(% x + i% y )

Four component wave function : why ?

!

1) Spin doubles the components 2) Negative energy solutions : E < -mc2

20

Densities
Charge density
"(r,t ) = q# (r,t )# (r,t )

Current density
! j (r,t ) = q" (r,t ) c# " (r,t )

Conservation relation
!

"#(r,t ) + $ % j (r,t ) = 0 "t

21

Time-independent Dirac equation

The nuclei do not move with relativistic speeds with respect to each other Take a stationary frame of reference (BornOppenheimer approximation) Separate the time and position variables " (r,t) = ih #" (r,t) H Time dependent Dirac equation #t " (r,t) = $(r)%(t)

H $(r) = E $(r) %(t) = e Et / ih

Time independent Dirac equation

22

Free particle Dirac equation

Take simplest case : = 0 and A = 0 Use plane wave trial function

$ a1 ' & ) ik#r & a 2 ) "(r) = e &a3 ) & ) %a 4 (

Non-relativistic functional form with constants ai that are to be determined

( ) (E * mc ) a
2

E * mc 2 a1 * chk z a 3 * chk* a 4 = 0
2

After insertion into time-independent Dirac equation

* chk + a 3 + chk z a 4 = 0

*chk z a1 * chk* a 2 + E + mc 2 a 3 = 0 *chk + a1 + chk z a 2

2 4

( ) + ( E + mc ) a

k = k x iky

=0
!
23

Free particle Dirac equation

Two doubly degenerate solutions

(E

! m2 c 4 ! c 2 h 2 k 2 ) = 0

E+ = + m2 c 4 + c2 h 2 k 2 E! = ! m2 c 4 + c 2 h 2 k 2

Compare to classical energy expression

E = m2 c 4 + c2 p2

Quantization (for particles in a box) and prediction of negative energy solutions

24

Free particle Dirac equation

Wave function for E = E+

chk z chk + a 2 = 0 ; a 3 = a1 ; a 4 = a1 2 E + + mc E + + mc 2 h k " p << mc cpz

For particles moving with nonrelativistic velocities

pz a 3 = a1 # a1 2mc mc 2 + m 2 c 4 + c 2 p 2 p+ a 4 # a1 2mc

Upper components are the Large components Lower components are the Small components
25

Free particle Dirac equation

Wave function for E = E-

a4 = 0 chk z pz a1 = a 3 # a3 2 "2mc E " " mc chk + p+ a2 = a3 # a3 2 "2mc E " " mc

Role of large and small components is reversed Charge conjugation symmetry Can we apply the variational principle ? Variational Collapse

26

Dirac sea of electrons

Electronlike continuum solutions

Electronlike bound solutions

Positronlike continuum solutions

All negative energy solutions are filled The Pauli principle forbids double occupancy Holes in the filled sea show up as particles with positive charge : positrons (discovered in 1933) Infinite background charge

27

Second Quantization

Introduce a m-dimensional Fock space F(m)

States are defined by the occupation number vector n

n = n1, n2 ,K, nm ni = 0,1

The vacuum has all n=0

vac = 0, 0,K, 0

We use an orthonormal basis

n k = !nk vac vac = 1

28

Second Quantization

Second quantized operators

Creation operator
ai n1 ,K, ni ,K, nm = 0 (ni = 1) ai n1 ,K, ni ,K, nm = Ci n1 ,K,1,K, nm (ni = 0) ai vac = 0,K,1,K, 0

Annihilation operator
ai n1 ,K, ni ,K, nm = Ci n1 ,K, 0,K, nm (ni = 1) ai n1 ,K, ni ,K, nm = 0 (ni = 0) ai vac = 0

Define all operators in terms of these elementary operators

!=
m k ,l =1

"!

a al kl k

29

Fock space Hamiltonian

Positive and negative energy solutions define a Fock space Hamiltonian

H Total = H ++ + H +" + H "+ + H ""

E #E + E #E "

H ++ =

$
p,q

p H pq a a q
E #E + E #E "

H "" =

$
% ,&

H%& a% a &

H +": pair creation =

$ $
p

p H p% a a% H%p a% a p

E #E " E #E +

H "+: pair annihilation =

$ $
%
p

30

Renormalization
1.

Subtract energy from the occupied negative energy states

H QED = H Total " E 0 = H Total "

2.

H Total

Re-interpretation
p a = bp ap = bp a! = b!
Due to the anticommutation relation

!
3.

a! = b!

Normal ordered Hamiltonian

electrons el. pos.

positrons

H QED =

"
p,q

H pq b bq + p

"" (
p

H p# b b# + H#p b# b p $ p

) "
# ,%

H#% b# b%

31

Quantum Electro Dynamics

Positive energy for positrons

E(1p;0e) = K, 1,K;K H QED K,1,K;K
Neg. Pos.
positron states

Neg.

Pos.

= K,1,K;K "

%H
# ,$

#$ # $

b b K,1,K;K = "E & ' mc 2

Total charge is also redefined

QED Qvac = "e vac N QED vac electron states positron states

= "e vac

$
p

b b p " p

$b b
#

# #

vac = 0

32

Dressed particles

The QED Hamiltonian depends on the positive and negative energy solutions of the Dirac equation. The Dirac equation depends on the external potential Common choices

Free particle solutions (Feynman,1948) Fixed external potential (Furry,1951) External + some mean-field potential (fuzzy)

Particles in one representation are quasiparticles (dressed with ep-pairs) in another representation Different no-pair approximations possible
33

Electron-electron interaction

Add quantized EM-field and interaction term

= p ! states;e ! states; photons HQED , full = H e + p + H photons + He + p, photons

Electron-electron interaction is automatically retarded by the finite velocity of light Corrections to the Dirac equation and the instantaneous Coulomb interaction can be derived

Feynman (NP 1965) diagrams

Breit interaction (1929) (Order c-2) Vacuum Polarization + Self Energy = Lamb shift (NP 1955) (c-3)
34

Electron-electron interaction

Three terms up to order c-2

Coulomb"Breit

1 (1,2) = r12 1 " 2 c# 1 $ c# 2 c r12 1 " 2 c# 1 $ % 1 c# 2 $ % 2 r12 2c

)(

Coulomb, Gaunt and retardation terms

First correction describes the current-current interaction Second correction describes retardation

35

Dirac-Coulomb-Breit Hamiltonian

Second quantization is merely convenient for our purposes, but becomes essential when going beyond the No-Pair approximation Page 68 of Book 1 has everything we need:
= " h D a a + 1 " ( gC + g B ) a a a a H ij i j ijkl ijkl i k l j 2 i, j,k,l i, j

Matrix elements are complex and (therefore) have less permutational symmetry We want to compute these matrix elements, so we need to go back to first quantization and basis set expansion techniques.....
36

MO-integrals in quaternion form

L. Visscher, J. Comp. Chem. 23 (2002) 759.
X Y $ V (r1 )$!W (r1 )$" (r2 )$ # (r2 ) VW,X G! ,"# Y =

%%

r1 2
&1 =0 &1 = 0 X,& ' ' c p 3 3
1

dr1 dr2

X B!X,& 1 2 e &1 2 ,pq

c!X,& 2 e * 1 e& 2 q &

Dirac G pqrs ( Coulomb &1 2& 3 4 NL NL

XX,& XXY Y = ' ' ' ' B! , pq1 2 G!"#YB"#Y,& 3 4 (&1 2, &3 4 = 0,1, 2,3) ,rs X Y ,! " ,#

L,S L ,S N X NY

Lvy G pqrs ( Leblond

LL LLLL LL,0 ' ' B!,,0 G!"# B"# ,rs pq

,! " ,#

G spinfree pqrs

XX,0 XXY Y = ' ' ' ' B !, pq G!"#YB"#Y,0 ,rs X Y ,! " ,# NL NL

L ,S L ,S NX NY

G Two( spinor& 1 2&3 4 pqrs

LL ,& ' ' B! ,pq

12

,! " ,#

LLLL LL,& G!"# B"# ,rs 3 4 (&1 2 , & 3 4 = 0,1, 2, 3)

37

Discussed in Lecture 1

("mc
!

+ c # $ % + q& )' (r ) = E' (r )

Dirac equation

Lorentz invariance Renormalization (QED) Choosing the reference Dirac Hamiltonian in QED: we need orbitals No-Pair approximation and second quantized Hamiltonian Breit interaction
38

Electronlike continuum solutions

Electronlike bound solutions

Positronlike continuum solutions

One more exact solution of the Dirac equation

The hydrogenic atom

% 2 Z ' mc " r ' ' c# $ p & ( c # $ p *%+ L ( r)( %+ L ( r)( ' * = E' S * Z *&+ S ( r)) &+ ( r)) "mc 2 " * r)

This equation can be solved exactly by separating the radial and angular variables The derivation and energy is found in various textbooks.

39

The hydrogenic atom

The exact energy expression

# % % Z /c 2 E = mc / 1+ $ 1 Z2 % n " j " + ( j + 1/2) 2 " 2 % & 2 c '2 % % ( % % )

Scalar relativistic corrections : Spin-orbit couping :

j=ls

Z2 E =" 2
NR

!
40

41

Orbital stabilisation
Alkali metals H, Li, Na, K, Rb, Cs, Fr, 119 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0 50 100 Nuclear Charge 150 nonrelativistic relativistic

42

Spin-orbit splitting
B, Al, Ga, In, Tl,13 Group 113
0.4

0.3

0.2 nonrelativistic relativistic relativistic 0.0 0 50 100 150 Nuclear Charge

0.1

43

Orbital destabilisation
Zn, Cd, Hg, 112 Group 12
0.9 0.8 0.7 0.6 0.5 0.4 0.3 0 50 100 Nuclear Charge 150 nonrelativistic relativistic relativistic

44

Orbital contraction

<r> of the outermost s-orbital

Alkali metals 8 7 6 5 4 3 2 1 0 50 100 150 Nuclear Charge nonrelativistic relativistic

45

Orbital contraction

<r> of the outermost p-orbital

Group 13 5.0 4.5 4.0 3.5 3.0 2.5 2.0 0 nonrelativistic relativistic relativistic 50 100 Nuclear charge 150

46

Orbital expansion

<r> of the outermost d-orbital

Group 12 2.0 1.8 1.6 1.4 1.2 1.0 0.8 20 nonrelativistic relativistic relativistic 70 Nuclear Charge 120
47

The hydrogenic atom

The exact energy expression

# % % Z /c 2 E = mc / 1+ $ 1 Z2 % n " j " + ( j + 1/2) 2 " 2 % & 2 c '2 % % ( % % )

Can be expanded to

(1+ x )

"

1 2

1 3 = 1" x + x 2 "K 2 8

# ' * Z6 Z2 Z4 % 3 n % 2 ! E = mc " 2 + 4 2 $ " ( + O, 4 / 2n 2n c % 4 j + 1 % +c . & 2)

48

Approximate Hamiltonians

Find operators that can describe these scalar relativistic and spinorbit coupling corrections in molecular systems Start by decoupling the large and small component equations

V" L + c # $ p" S = E" L c # $ p" L + (V % 2mc 2 )" S = E" S

Rewrite the lower equation as

$ E # V '#1 * + p L " S (r ) = &1# " (r ) 2 ) % 2mc ( 2mc *+p L = K ( E,r ) " (r ) 2mc
!

# E " V &"1 K ( E,r ) = %1" ( $ 2mc 2 '

49

Approximate Hamiltonians

Substitute in the upper equation

$ 1 ' L % (" # p)K ( E,r )(" # p) + V (* (r ) = E* L (r ) & 2m )

Unnormalized Elimination of the Small Component

!
# E " V &"1 K ( E,r ) = %1" ( $ 2mc 2 '

50

Pauli Hamiltonian

Crudest approximation :

K ( E,r ) = 1

$ 1 ' L % (" # p)(" # p) + V (* (r ) = E* L (r ) & 2m ) $ p2 ' L! Schrdinger equation + V (* (r ) = E* L (r ) % & 2m )

Take K=1 but keep the magnetic field

% 1 ( L & (" # $ )(" # $ ) + V )+ (r ) = E+ L (r ) Pauli equation ' 2m *

51

Breit-Pauli Hamiltonian

Find an operator that normalizes the wave function :

" = N" L
1 N = 1+ (# $ p)K 2 (# $ p) 4m 2c 2

Multiply the UESC equation by N-1

% 1 ( "1 N & (# $ p) K ( E,r)(# $ p) + V ) N N+ L ( r) = N "1E+ L ( r) ' 2m *

"1

% 1 ( "1 N & (# $ p)K ( E,r )(# $ p) + V ) N + (r ) = EN "2+ (r ) ' 2m *

"1

! !
52

Breit-Pauli Hamiltonian

Use series expansions and keep terms up to order c-2

% ("1/ 2 1 N "1 = '1+ (# $ p)K 2 (# $ p)* & 4m 2c 2 ) 1 = 1" (# $ p)K 2 (# $ p) + K 8m 2c 2 p2 = 1" + O(c "4 ) 8m 2c 2 % ("1 "2 = 1+ 1 (# $ p)K 2 (# $ p) N ' 4m 2c 2 !* & ) !

# E " V &"2 K = %1" = 1+ O(c "2 ) 2 ( $ 2mc '

2

p2 = 1" + O(c "4 ) 4m 2c 2

53

Breit-Pauli Hamiltonian

Expansion of K
# ( E " V )& ( E " V ) + O(c "4 ) K ( E,r ) = %1+ = 1" ( 2mc 2 ' 2mc 2 $
"1

Subsitute and keep only terms to order c-2

"1%V + 1 (# $ p)K (# $ p)(N "1+ = EN "2+ N ' * & ) 2m

% ( 1 " # p)V (" # p) $ Ep 2 $ Tp 2 $ [ p 2 ,V ] + * ( ' , Ep 2 / 2 +T+ + 'V *+ = . E $ 2 2 2 21 4m c 4m c 0 ' * & )

!
54

Breit-Pauli Hamiltonian

The energy dependent term on the lhs was cancelled by the rhs 1 (" # p)V (" # p) $ Tp2 $ [p2 ,V ] + 2 H BP = T + V + 2 2 4m c Further simplify the equation using

(" # p)V (" # p) = (pV ) # p + Vp2 + i" # (pV ) $ p

%

1 2 1 p ,V ] = % (p2V ) % (pV ) # p % Vp 2 [ + 2 2
2 4 BP = T + V " p V " p + i# $ (pV ) % p H 8m 2c 2 8m 3c 2 4m 2c 2

Result : The Breit-Pauli equation

Darwin

Mass-Velocity

Spin-Orbit
55

Expectation values for the hydrogen atom

<H
Darwin

Z4 >= 3 2 2n c

(l = 0)

< H Darwin >= 0 (l > 0)

!
!

< H MV

# ' Z4 % 3 n % >= 4 2 $ " ( 1% 2n c % 4 l + & 2)

4 l SO >= Z <H 2n 3c 2 l(2l + 1)( l + 1)

( j = l + 1/2) ( j = l "1/2)

<H

SO

Z4 "l "1 >= 3 2 2n c l(2l + 1)( l + 1)

56

Approximate relativistic Hamiltonians

Can we improve upon the Breit-Pauli Hamiltonian ? A short wish list : 1. The Hamiltonian should resemble the Schrdinger Hamiltonian as much as possible 2. It should describe the scalar relativistic effects 3. It should describe the spin-orbit coupling effect 4. It should be variationally stable 5. It should be easy to implement 6. Errors relative to the Dirac solutions should be small and systematically improvable 7. It should be well-named....
57

Regular approximations

What did we do wrong ? Check the expansion parameter

# ( E " V )& ( E " V ) + O(c "4 ) K ( E,r ) = %1+ = 1" ( 2mc 2 ' 2mc 2 $
"1

E should be small relative to 2mc2

Orbital energies vary over a range of -0.1 to 5,000 au Twice the rest mass energy is 37,558 au This difference should be large enough

V should be small relative to 2mc2

The potential is dominated by the nuclear attraction close to the nuclei Z V "# r Take r = 10-4 au and Z=6 (carbon) : V = 60,000 au Is this inside the nucleus ? No : the RMS radius is 4.7 10-5 au for C.
58

Regular approximations

Can we find a better expansion parameter ? Yes !

"1 # ( E " V )& ) ,"1 V , ) E K ( E,r ) = %1+ = +1" 1+ . 2 ( 2. + * 2mc - * 2mc 2 " V 2mc ' $ "1

E should be small relative to 2mc2 - V

V is negative which improves the expansion close to the nuclei

Zeroth order in this expansion

$1 + 1 . % V ( " # p) + V /1 ZORA (r ) = E1 ZORA (r ) , (" # p)'1$ 2* ( & 2mc ) - 2m 0

Zeroth order equation does describe SO-coupling and scalar relativistic corrections EC ! Gauge dependence of the energy V "V + C E " E + C # 2

Affects ionization energies, structures Gauge independence can be achieved various ways

2mc

59

Approximations to K(E,r) for the 1s orbital of Fm99+

60

Uranium atom
8

1s orbital
6

DIRAC ZORA NR

r!! (a.u.)

0 10-5 10-4 10-3

r (a.u.)

10-2

10-1

100

61

Uranium atom
0.5

7s orbital DIRAC ZORA NR

0.4

r!! (a.u.)

0.3

0.2

0.1

0.0 10-5 10-4 10-3 10-2

r (a.u.)

10-1

100

101

62

Foldy-Wouthuysen transformations

Use an energy-independent unitary transformation to decouple the large and small component equations
UH DU "1U# iD = EU# iD $H + 0 ' FW D "1 H = UH U = & ") % 0 H ( $# iFW ' FW 4,(+) D(+) #i = U# i = & ) Picture change % 0 ( 1 1 $ ' " " 2 2 X 1 (1+ X X ) ) = & (1+ X X ) X= K (".p) U & 1 1 ) 2mc "2 ) & X 1+ XX " 2 ) (1+ XX ) ( % (

Exact expressions are only known!for the free particle problem

63

Douglas-Kroll-Hess method

Idea

Transform bare-nucleus Hamiltonian with the known free-particle tranformation matrix, followed by additional transformations to reduce size of remaining off-diagonal elements to some order in the potential

Assumptions

The transformation is based on the Furry picture : potential does not include mean-field of electrons The conventional implementations neglect the transformation of the twoelectron interaction and often also the SO-coupling terms

Advantages-Disadvantages
Method is variationally stable Slight modification of existing code required (replacement of one-electron nuclear attraction integrals), fast implementation Good results in practice, significant improvement over Breit-Pauli Complicated operators, matrix elements can not be calculated analytically Two-electron terms are hard to evaluate Interactions with external field need to be represented by transformed operators (picture change)
64

Douglas-Kroll-Hess method

The second-order Hamiltonian

" H DKH 2 H (2) = $ (2) # H 21
(2) H12 % (2) ' H 22 &

The Douglas-Kroll-Hess Hamiltonian

1 1 H DKH 2 = E p + A p V + (".Pp )V (".Pp ) A p + W1 E pW1 + W12 E p + E pW12 2 2

E p = m 2c 4 + c 2 p 2
!

Free particle energy operator Kinematical factor

Ap =

E p + 2mc 2 2E p

Rp =
!

c" # p = " # Pp 2 E p + mc

A p R pV (p, p") A p " # A pV (p, p") R p " A p " W1 = E p + E p"

65

Higher order approaches

General operator transformations

Barysz-Sadlej-Snijders (1997) Reiher (200X) van Wllen (200X)

Higher order two electron effects

Samzow, Hess, Jansen (1992) Park and Almlf (1994) Hirao (2003-present)

Infinite order via matrix representations

Ilias and Jensen (2005)

66

Four-component methods

Idea

Expand Dirac equation in separate basis sets for the large and small components Use kinetic balance condition to prevent variational collapse

Advantages-Disadvantages
No approximations made Matrix elements over the operators are easily evaluated Many more two-electron integrals need to be handled The Fock matrix is twice as large No picture change problems

67

Hartree-Fock Self Consistent Field

occupied orbitals
1.

2 Construct Fock operator F = # "c + c$ % p + V +

'J
j

&Kj

2.

Find eigensolutions
! Check convergence
!

F" ( r1) = #" ( r1)

3.

{" new } = {" old } ?

E HF = E Kinetic + E Potential + E Elec. Re p.

4.

Compute energy
occupied orbitals

!
2

occupied orbitals

E HF =

" < i | $ #c
i

+ c% & p | i > +

"< i | V | i > + 2 "< i |

i i, j

occupied orbitals

Jj 'K j |i >

68

Fock operator
%V + # J " K j j ' j F =' 'c ($.p) " # K j & j

( * j * V " 2c 2 + # J j " K j * j ) c ($.p) " # K j

J j ( r1) =

!
!
=

" jL ( r2 )" jL ( r2 ) + " jS ( r2 )" jS ( r2 ) $ j ( r2 ) dr2 = # dr2 r12 r12

K j " iL ( r1) = K jLL" iL ( r1) + K jSL" iL ( r1)

" jL ( r2 )" iL ( r2 ) " jL ( r2 )" iL ( r2 ) L dr2" j ( r1) + # dr2" jS ( r1) r12 r12

K j " iS ( r1) = K jLS " iS ( r1) + K jSS " iS ( r1) ! =

"

S j

( r2 )" ( r2 ) dr " L
r12
2 j

S i

( r1) + #

"

S j

( r2)" iS ( r2) dr " S

r12
2 j

( r1)
69

Basis set expansion

Use different expansion sets for the large and small component parts of the wave function
L $ # (r )' L N S $ 0 ' S " i (r ) = *& )c i + *& S )c+i # r 0 ( =1 % + =1 % + ( )( NL

Kinetic balance condition

#$p L " (r ) 2mc Recovers the non-relativistic limit " S (r ) =

1N L L* S " # (r )T" % (r ) dr = ( $ " # (r )(& ' p)" (r ) dr ) $ 2 =1
L*
S

!
!

$ " S (r )(& ' p)" %L (r )dr

70

TLL =

1 LS SL (" # p) (" # p) 2

Resolution of identity

Choice of expansion functions

Large component

Atoms : Sturmians, Slaters or Gaussians Molecules : Spherical or Cartesian Gaussians Same type as large component Should fulfill kinetic balance relation

Small component

S L {" P } = {(# $ p)" P }

{ }
S "P

$#" L #" L #" L ' =% P , P , P( & #x #y # z )

Restricted KB
!

Unrestricted KB
71

Kinetic Balance
Dirac equation c ! Schrdinger equation

Condition : (! . p)(! . p) = p2 Cartesian Gaussian basis Large Component Small Component s s p p d d f

72

The small component density

The large component wave function resembles the non-relativistic wave function Exact relation between large and small component wave functions
#1 ( S #1 $ E # V ' ( ( L " = &1+ *" 2 ) 2c % 2c (

Small component wave function is related to the first derivative of large component wave function The small component density is an embarrassingly local quantity !
73

Electron Density of Uranyl

Large component Small component

Spinfree Dirac equation

Define an auxilliary function

1 " = (# $ p)% L 2mc
S
Relation holds by definition

Transform the Dirac equation accordingly

%V ' 'T & (% L ( %1 T 0 (%+ L ( + (" # p)V (" # p) $ T *' L * = E '0 T *' L * ' * *& , ) 2 &, ) 2 & ) 2mc ) 4mc

Separate scalar and spin-dependent part and neglect the spin-dependent terms if desired
75

Direct perturbation theory

Consider the modified Dirac equation

%V ' 'T & (% L ( %1 T 0 (%+ L ( + (" # p)V (" # p) $ T *' L * = E '0 T *' L * ' * *& , ) 2 &, ) 2 & ) 2mc ) 4mc

Non-relativistic limit is related to the Lvy-Leblond equation

#V % $T # 1 0&#) L & T &#) L & NR (% L ( = E % (% ( "T '$ * ' 0 0'$ * L ' $ % V ' &(" # p)

(" # p)(%+ u (

%1 0(%+ u ( *' l * = E ' *' l * $2m )&+ ) &0 0)&+ )

LL

Define a perturbation theory with as first (or second) order perturbations

!
#0 0 & H =% "2 ( $0 "Vc '
(1)

#0 0 & S =% ( 0 c "2 ' $

(1)

76

Discussed in Lecture 2
$ 1 ' L % (" # p)K ( E,r )(" # p) + V (* (r ) = E* L (r ) & 2m )
Two-component methods

Breit-Pauli perturbation theory Regular approach (ZORA) Douglas-Kroll-Hess method 4-component methods Direct Perturbation Theory

77

Valence-Only approaches

All-electron calculations are not always feasible or necessary Hierarchy of approximations for core electrons 1. Correlate the core electrons at a lower level of theory (e.g. MP2) 2. Include core electrons only at HF level of theory 3. Use atomic orbitals for core electrons (Frozen Core) 4. Model frozen core by a Model Potential 5. Model frozen core by a Relativistic Effective Core Potential Error correction and additional features 1. Estimate higher order correlation effects in another basis set 2. Use a core correlation potential 3. Use a core polarization potential 4. 5. Include valence relativistic effects in RECP

78

Frozen Core approximation

Consider the Fock operator

Nuclei

F = h kinetic "

#
A

ZA + rA

occupied orbitals

# (J
j

"K j)
* Z A = Z A " ZCore

Identify localized (atomic) core orbitals and partition

Nuclei kinetic

valence Z A Nuclei core A A F=h " # + # # Jc " K c + # Jv " K v rA A A c v ! Nuclei * valence Nuclei $ core core ' Nuclei core A ZA ZA kinetic A F=h " # + # J v " K v + # &" + # Jc ) " # # K c rA rA ( A c A v A % c

VCoulomb

VExchange

Coulomb potential goes to zero at large distance, contains correction due to imperfect screening of nuclei at short distance Exchange contribution depends on the overlap : short range Approximation made : atomic core orbitals are not allowed to change upon molecule formation
79

Core polarization and overlap

Polarizability of the core can modeled by a classical core polarization potential (see also book II, formula 45.9)
A VCPP = "

1 $ fAT # A fA 2 A

Field from the electrons and the other nuclei at the position of core A

Polarizability of core A

Need a cut-off factor in the field since the multipole expansion is only valid outside the core Can be extended to model core-correlation and core-valence correlation as well The overlap between cores is assumed to be zero and thus neglects the exchange repulsion and nuclear attraction between neighbour cores For large core calculations this requires a correction
80

Ab Initio Model Potentials

Replace the exact, non-local, frozen core potential by a model potential plus a projection operator
core core A c core A

A Frozen core

= # (J " Z
c

) " #K
c

A c

A A A $ VCoul + VExch + PCore

A VCoul =

A 2 1 $ c iA e"# i rA rA i

Density fit of spherical density, can be done to arbitrary precision

primitive core basis on A

A VExch = "# c

#r S
r,s,t,u

"1
rs

s K cA t S "1 u tu

Resolution of identity with non-orthogonal functions

!
P
A Core

core

= " c BcA c
c

Level shift that enforces orthogonality to the core

81

Ab Initio Model Potentials

No freely adjustable parameters Keeps nodal structure of the valence orbitals Core orbitals in the virtual space Relativistic effects can be included in the reference Fock operator

Cowan-Griffin Hamiltonian Wood-Boring Hamiltonian Douglas-Kroll-Hess Hamiltonian

Can also be used to generate no-valence MPs

Improves description of ions in crystals May require iterative generation scheme See example from the work of Seijo in the green hand-out
82

Effective Core Potentials

Reduce the basis set used to describe the valence orbitals

Fv " Fv + % (#v $ #c ) c c
c

Phillips and Kleinman : shift core orbitals to make them degenerate with the valence orbitals

{" v } # {" v }
!

Make nodeless pseudo-orbitals

L#1

A Frozen core

(rA ) " M ( rA ) + $ $ lml

A L l= 0 m l =#l
L#1 l +1/ 2 l A L

f lA ( rA ) lml

Scalar

A Frozen core

(rA ) " M ( rA ) + $ $ $ ljm j

l= 0 j= l#1/ 2 m l =#l

f ljA ( rA ) ljm j

Spin-Orbit

These pseudopotentials are determined via a fitting procedure. They take care of Coulomb and Exchange and core-valence orthogonality.
83

Shape consistent ECPs

American school : Christiansen, Ermler, Pitzer French school : Barthelat, Durand, Heully, Teichteil Make nodeless pseudo-orbitals that resemble the true valence orbital in the bonding region
%" (r ) v (r ) = & v " v (r ) # " ' f (r )

( r $ rC ) (r < rC )

Original orbital in the outer region Smooth polynomial expansion in the inner region

Absolute correlation energy may be overestimated relative to correlation calculations done with the real orbitals Fit is sometimes done to the large component of Dirac wave function (picture change error) Reasonable accuracy for bond lengths and frequencies Available in many program packages

84

Energy consistent ECPs

German school : Stoll, Preuss, Dolg Semi-empirical or ab initio approach that tries to reproduce the low-energy atomic spectrum (using correlated calculations)
$ Lowlying ' Levels 2) & min& # w I ( E IPP " E IReference ) ) & I ) % (

Provides good accuracy for many elements and bonding situations

Difference in correlation due to nodeless valence orbitals is automatically included in the fit Small cores are often necessary to obtain stable results Available in many program packages

85

Methods to treat relativity

Best method depends on system studied See exercise (and answer) 10 Closed shells and simple open shells

Use a size-extensive and economical method SOC-inclusive method may be required

Complicated open shells, bond breaking

MCSCF, Multi-Reference CI or MR-CC SOC-inclusive methods are usually required Mean-field description of SO (AMFI) is usually sufficient

Use best practice and experience from calculations on light elements

86

Visible Relativistic Effects

Non-relativistic gold is silver

The 5d-6s transition is shifted from the UV to the visible part of the spectrum by scalar relativistic effects

6s
Non-Relativistic

6s
Relativistic

5d

5d

Phosphorescence

Singlet-triplet transitions are allowed because the non-relativistic quantum numbers are not exact

87

Mercury: Dipole polarizability

Calculation via 4component time-dependent Hartree-Fock (4c TD-DFT is nowadays also possible) Relativistic Nonrelativistic Breit-Pauli Pert. Theory

Nonrelativistic Relativistic

A. B. C.

Experiment

BP PT

6s6p transitions Forbidden 1S0 3P1 Allowed 1S0 1P1

88

Closed shell molecules

Some studies with all-electron single reference methods Analyze relativistic effects for diatomic molecules

89

Atomization energies
Example: Halogen molecules Molecular energy is hardly affected by SOcoupling (SO quenching) First order perturbation theory
Nonrelativistic u g

Relativistic u,1/2 g,1/2 u,1/2 g,3/2 u,3/2 g,1/2

90

u g

Atomization energies
Atomic asymptotes are lowered by SO-coupling First order perturbation theory
Nonrelativistic
2P
SO-splitting

px

py

pz

2P

3/2

Relativistic

p1/2 p3/2 p3/2

91

Relativistic effect on atomization energies (kcal/mol)

Mainly SO-coupling : relativistic effect on atomization energies can be estimated by correction to the asymptote

Relativistic effect on harmonic frequencies (cm-1)

Bond weakening due to admixture of the antibonding sigma orbital. This is a second order spin-orbit effect

Relativistic effect on equilibrium distances ()

Note the underestimation by Hartree-Fock in At2

Atomization energies
Example: Hydrogen halides SO-coupling is again mostly quenched First order perturbation theory Strong sigma-pi mixing in ultra-relativistic H117
Nonrelativistic 1/2 1/2
95

Relativistic 1/2 3/2

Dissociation Energies
aug-pVTZ
(kcal/mol)

HF 0 -5 -10 -15 -20 -25

HCl

HBr

HI

HAt

HF HF+G MP2 CCSD CISD CISD+Q CCSD(T)

Mainly SO-coupling : a good estimate for atomization energies can be obtained by correcting only the asymptote

Relativistic

shift

in

De

Vibrational Frequencies
aug-pVTZ
0 HF -50 HF -100 HF+G MP2 CCSD -150 CISD CISD+Q CCSD(T) -200 HCl HBr HI HAt

Relativistic shift in

!e (cm-1)

-250

Bond weakening due to admixture of the antibonding sigma orbital

Bond Lengths
aug-pVTZ
HF 0.012 0.010 0.008 0.006 0.004 0.002 0.000 -0.002 -0.004 -0.006 -0.008 HF HF+G MP2 CCSD CISD CISD+Q CCSD(T) HCl HBr HI HAt

Competition between scalar and spin-orbit effects Total effect is small (< 0.01 ) in this case

Relativistic

shift

in

re

()

dissociation energy (eV)

SR ZORA ZORA Exp

PbO

TlF

spin-orbit effect in: fragments (atoms) molecule

PbTe

AuH

HI Bi2 IF Au2 I2

TlI

TlH

99

Groundstate of thalliumhydride
K. Faegri Jr.. and L. Visscher, Theor. Chem Acc. 105 (2001) 265.

Goal : Provide benchmark values for this standard testcase Hamiltonian : Dirac-Coulomb-(Gaunt) Correlation space : up to 36 electrons (6s, 6p; 4f, 5s, 5p, 5d)
Method and # electrons corr. MP2* 14 DC-CCSD(T)* 14 DC-CCSD(T) 14 DC-CCSD(T) 20 DC-CCSD(T) 36 DCG-CCSD(T) 36 experiment Re (pm) 186.2 188.5 187.6 187.4 187.4 187.7 186.8 Ke (N/m) 121 111 113.3 112.1 111.1 111.9 114.4 ! (cm-1) 1437 1376 1385 1378 1371 1376 1391 De (eV) 1.83 2.07 2.00 1.98 1.98 2.06 2.06

*Seth, Schwerdtfeger and Faegri (1999) calculations with contracted basis sets. 100

Open shell molecules

Two studies with all-electron single and multireference methods Analyze relativistic effects

101

Fine structure splitting in radicals

Valence iso-electronic systems O2, FO, ClO Breit interaction and correlation should be included for accurate results
Fine structure splittings XO molecules
350

300

ClO

Experiment DCG-HartreeFock DCG-CCSD

250

200

FO
150 150

DCG-CCSD-T

O2
200 250 300 350 400

Experimental

FSS

(cm

-1

Platinumhydride

PtH molecule : jj-coupling instead of LS-coupling scheme Pt (5d96s1) + H (1s1) PtH (5d92)

103

Molecular properties

Relativistic effects on some molecular properties

104

Dipole moment of HI
Relativistic effects
0.25 0.24 0.23

dipole moment (au)

0.22 0.21 0.20 0.19 0.18 0.17 0.16 -1 -0.5 0 (R-Re) 0.5 1

NR CCSD(T) SF CCSD(T) DC CCSD(T) exp.

105

NMR: 1H shielding trends

50 45 40 35 30 25 20 HF HCl HBr HI NR SO SR Rel

106

NMR: 13C shielding trends

Data from Malkina et al., Chem. Phys. Lett. 1998 Mean-field SO method employed.
107

Extracting nuclear structure information from Spectroscopy & Quantum Chemistry

Nuclear Quadrupole Moments

The coupling between the nuclear quadrupole moment Q and the electric eld gradient (EFG) at the nucleus q gives an energy splitting that depends on the orientation of the nuclear spin. This can be observed with high precision in microwave (rotational) spectroscopy on diatomic molecules. Quantum chemistry gives q and can thus be used to obtain accurate values of Q or to predict and rationalize NQR or NMR observations.

Molecular rotation Nuclear spin

EQ =

e 2 qzz Q[3m2 ! I(I + 1)] I 4I(2I ! 1)

108

Iodine
Spread in Nuclear Quadrupole Moments (mbarn)
TlI AuI AgI CuI I2 IBr ICl IF HI
ZORA4-DFT(BP) DC-MP2 DC-CCSD DC-CCSD(T)

DC-HF

SF-HF

NR-HF

-500

-400

-300

-200

-100

100

200

109

Further reading

Relativistic Quantum Mechanics

R. E. Moss, Advanced molecular quantum mechanics. (Chapman & Hall, London, 1973). P. Strange, Relativistic Quantum Mechanics. (Cambridge University Press, Cambridge, 1998).

Relativistic Quantum Chemical methods

Relativistic Electronic Structure Theory - Part 1 : Fundamentals, ed. P. Schwerdtfeger (Elsevier, Adam, 2002). Theoretical chemistry and physics of heavy and superheavy elements, ed. U. Kaldor and S. Wilson (Kluwer, Dordrecht, 2003. Relativistic Effects in Heavy-Element Chemistry and Physics, edited by B. A. Hess (Wiley, Chichester, 2003). Relativistic Electronic Structure Theory - Part 2 : Applications, ed. P. Schwerdtfeger (Elsevier, Amsterdam, 2004).
110

Applications

Typographical errors

Page 601 & Formula 4 (answer to question 3) : c missing

j (r,t ) = q" (r,t )c#" (r,t )

Page 604 : division symbol missing

# % % Z /c 2 E = mc / 1+ $ 1 Z2 % 2 % n " j " 2 + ( j + 1/2) " c 2 & ' % % ( % % )
2

Answer to question 2

(" # X)(" # Y) = " i" j X iY j

1 1 [" i," j ]+ X iY j + 2 [" i," j ]$ X iY j 2 = %ij X iY j + i&ijk" k X iY j = = X # Y + i" # ( X ' Y)
111