Lecture #9

Plan:
1. Laguerre polynomials and the radial function
2. Concepts in probability
3. The Heisenberg uncertainty principle and the commutator
4. Variational method
5. Rayleigh-Ritz trial functions
6. Derivation of the secular equation
7. H2+ ion – full problem and associated challenges
8. Born-Oppenheimer approximation the electronic Hamiltonian
9.

Questions you should be able to answer by the end of today’s lecture:

1. Understand the idea behind the variational method for approximating
eigenfunctions.
2. Why does bonding occur?
3. How to calculate simple approximations to the molecular orbital in an ionized
Hydrogen molecule.
4. How do the molecular orbitals change as the interatomic distance is decreased
5. What physical insight leads to the separation of the electronic and nuclear
coordinates?
6. How to choose trial wave functions and use the variation principle.

Math Tools:

1. Basic probability concepts

2. Variation methods
3. Minimization of functionals

References:

1. Merzbacher chapter 8.
2. Lowe, chapter 7.
 Hydrogen atom eigenfunctions
unlm = Rnl ( r ) Yl m (θ , ϕ )
EI
En = −
n2
 2  ( n − l − 1) !  r  2l +1  2r 
r 3 l
−
R (r ) = e na0
    Ln −l −1  
 n  2n ( n + l ) !  na0   na0 

Where L2nl−+l1−1 is the generalized Laguerre polynomial (using Mathematica’s definition)

Plotting the r 2 R102 ( r ) , r 2 R202 ( r ) , r 2 R302 ( r ) distributions:
123 123 123
(1s )2 ( 2s )2 ( 3s )2

I 2 M3 Hn - l - 1L !
R@r_, n_, l_D := ã n Hn + lL ! & J N
-r 2r l
2 n Hn + lL!3
n
n

LaguerreLAn - l - 1, 2 l + 1, E;
2r

PlotA9r2 R@r, 1, 0D2, r2 R@r, 2, 0D2, r2 R@r, 3, 0D2=,

n

8r, 0, 40<, PlotRange ® 880, 40<, 80, .6<<E

0.5

0.4

0.3

0.2

0.1

5 10 15 20 25 30 35 40

I 2 M Hn - l - 1L !
R@r_, n_, l_D := ã n Hn + lL ! & J N
3
-r 2r l
2 n Hn + lL!3
n
n

LaguerreLAn - l - 1, 2 l + 1, E;
2r

PlotA9r2 R@r, 3, 2D2, r2 R@r, 4, 3D2, r2 R@r, 4, 2D2=,

n

8r, 0, 40<, PlotRange ® 880, 40<, 80, .12<<E

 0.12 3d
0.1
4f
0.08 4d
0.06

0.04

0.02

5 10 15 20 25 30 35 40

LaguerreL[n, a, x] gives the generalized Laguerre polynomial Ln HxL

Mathematica note
a

SphericalHarmonicY[l, m, q, f] gives the spherical harmonic Yl Hq, fL

m

The Heisenberg Uncertainty Principle and it’s connection to the commutator

We mentioned in class that only operators that commute can be measured
“simultaneously” and are thus termed compatible this follows from the fact that they have
common eigenfunctions. Implying that for a set of non-commuting observables
measurement of one physical quantity and the subsequent collapse of the wavefunction to
the corresponding specific eigenfunction will invariably associate an uncertainty with
respect the measurement of the other.

Review of some basic concepts in probability

The set of possible results of a test, experiment or trial are called outcomes {Oi } we
assume that each outcome is assigned a probability of occurrence pi .The set of
probabilities { pi } is called a probability distribution. A variable X which takes on the
value X i if outcome Oi occurs is called a random variable. A function of a random
variable is itself a random variable.
The expectation value (or mean value) of the random variable X for the given probability
distribution is,
X = ∑ pi X i
i
The variance of a random variable is given by:
( ∆X ) (X − ) = ∑( X − X )
2 2
≡ pi = X 2 − X
2 2
X
i
The standard deviation is defined by:
∆X = X2 − X
2

Discuss: standard deviation as a measure of uncertainty in a measurement (plot the value

of a measurement vs. the experiment number).

The covariance of two random variables is,

∆X ∆Y = ( X − X )(Y − Y ) = XY − X Y
if X and Y are uncorrelated (independent) the covariance vanishes.
Let us now use these basic probability concepts to derive a rigorous form of the
Heisenberg uncertainty principle.

The expectation value of an operator gives us the mean value of the physical quantity for
a particular state. One would approach this value if a measurement of A would be
performed on a large collection of identical states ψ .
∞

ψ Aψ
ˆ ∫ ψ ∗ Aˆψ dv
Aˆ = = - ∞∞
ψ ψ
∫ ψ ψ dv
∗

-∞
The definition of an Hermitian operator is:
∗ ∗
∫ ( Fψ ) ψ dv = ∫ψ Fψ dv
This implies that:

( ∆Aˆ ) ( Aˆ − Aˆ ) ( Aˆ − Aˆ )ψ , ( Aˆ − Aˆ )ψ
2 2
≡ =

and,

( ) ( ) ( Bˆ − Bˆ )ψ , ( Bˆ − Bˆ )ψ
2 2
∆Bˆ ≡ Bˆ − Bˆ =
define,
(
ψ a = Aˆ − Aˆ ψ )
ψb = ( Bˆ − Bˆ )ψ
Using the Schwarz inequality
2
ψ a ,ψ a ψ b ,ψ b ≥ ψ a ,ψ b
which can be interpreted geometrically as the cosine of an angle has to be smaller or
equal to 1.

and substituting the expressions defined above,

( )( ) ( ) ( ) ( )( )
2 2 2 2
∆Aˆ ∆Bˆ ≥ Aˆ − Aˆ ψ , Bˆ − Bˆ ψ = ψ , Aˆ − Aˆ Bˆ − Bˆ ψ
Lets us define,
 Aˆ , Bˆ  = iCˆ
 
where the i has been added to ensure that C is hermitian.
(
Aˆ − Aˆ Bˆ − Bˆ + Bˆ − Bˆ )( ) ( ) ( Aˆ − Aˆ ) + i Cˆ = Fˆ + i Cˆ
( )(
Aˆ − Aˆ Bˆ − Bˆ = ) 2 2 2
 2

( ∆Aˆ ) ( ∆Bˆ )  i   i  ˆ i ˆ  1 ˆ
2 2 2 2
≥ ψ ,  Fˆ + Cˆ ψ =  Fˆ + Cˆ  F − C  = F
ˆ + C
 2   2  2  4
1 1
( )( ) 1
2 2
≥ Cˆ =  Aˆ , Bˆ  → ∆Aˆ ∆Bˆ ≥  Aˆ , Bˆ 
4 4 2

The variation method

The problem: in most cases of physical interest the eigenfunctions of the Hamiltonian
cannot be readily found. In such cases one needs to resort to a systematic procedure for
searching for functions which approximate the eigenfunctions. The calculus of variations
provides us with a powerful tool for solving such problems where H is known but not
un and En .
The variation method is based on the idea that if we try to find a trial function φ that
minimizes average E it will be greater or equal than the true ground state eigenfunctions
un .
∞

∫φ
∗
Hˆ φ dv
E = Hˆ = -∞
∞
≥ Eground state
∫ φ φ dv
∗

-∞
Proof:
Because H is Hermitian its true eigenfunctions are a complete set therefore any function
and in particular the trial function can be expanded in terms of eigenfunctions:
n
χ ( x ) = ∑ cn un ( x )
i =1
n
χ Hˆ χ = ∑ cn En
2

i =1
and
n
χ χ = ∑ cn
2

i =1

For E = Hˆ = Eground it is necessary and sufficient for all cn = 0 except for n=1.
Note: it is possible to find eigenvalues other than the ground energy using the variation
method:

The Rayleigh –Ritz trial function

In this method a trial function is chosen as a linear combination of independent and - in
most cases orthogonal and normalized - basis functions χ . Thus the trial function is of
the form:
N
χ = ∑ cnφn
n =1
The trial function can be viewed as a vector in a basis of φ where the c’s are the
components of the vector.
The expectation value is expressed as,

χ Hˆ χ
∑c
i , j =1
∗
i φi Hˆ φ j c j
H = =
χ χ N

∑c c
i , j =1
∗
i j

note: the expression φi Hˆ φ j is a complex number obtained by performing the integral

∫ φ Hˆ φ dv and is called the i,j matrix element of H or H

i j ij .
The condition for achieving a stationary (or minimum) value of H with respect to
variations in the trial wavefunction is:
∂ ˆ ∂ ˆ
∗
H = 0 and H =0
∂ci ∂c j
N
 N  N

∂
∑ φi Hˆ φ j c j  ∑ ci∗c j  − ci ∑ ci∗ φi Hˆ φ j c j
H =
j =1  i , j =1  i , j =1
=0
∂ci∗  N ∗ 
2

 ∑ ci c j 
 i , j =1 
N
 N  N
→ ∑ φi Hˆ φ j c j  ∑ ci∗c j  = ci ∑ ci∗ φi Hˆ φ j c j
j =1  i , j =1  i , j =1
N

N ∑c ∗
i φi Hˆ φ j c j
∑φ i Hˆ φ j c j = ci
 N ∗ 
i , j =1
= Hˆ ci = Eci
 ∑ ci c j 
j =1

 i , j =1 
Yielding n equations with n unknowns since for each ci there is a separate equation, E
being the expectation value corresponding to the optimal choice of the trial function. This
equation is known as the secular equation and we are looking for it’s roots (i.e the E
values).

∑ φ ( Hˆ φ )
n

i j − δ ij E c j = 0
j =1

in matrix notation,
 H11 − E H12 .... H1n   c1 
  
 H 21 H 22 − E   c2  = 0
=
 M  M 
  
 H n1 H nn − E   cn 

which can only have a non-trivial solution if

  n 
( )
Det  ∑ φi Hˆ φ j − δ ij E  = 0
 j =1 
H11 − E H12 .... H 1n
H 21 H 22 − E
=0
M
H n1 H nn − E
note: if the basis set is not orthogonal this matrix has some additional terms see Lowe
page 195 for more details

Thus the problem of finding the approximate eigenfunctions and corresponding energy
spectrum has been transformed from one that involves functions and minimization to an
algebraic problem.

The Ritz Theorem: The expectation value of the Hamiltonian is stationary in the vicinity
of its discrete eigenvalues.

All of the intelligence goes into choosing a “good” trial function, which already contains
in it a large amount of the true eigenfunction. The trial function is chosen with the aid of
physical insight. Hartree Fock method for calculating eigenfunctions of many electron
Hamiltonians is also an example of the variation method.

Understanding the chemical bond - Approximating the wave functions of molecules

LCAO linear combination of atomic orbitals

The problem:

How do we calculate the eigenfunctions of molecules?

The system – simple diatomic molecule

I] Draw the potential

r1

r2 R

II] Write the full Hamiltonian

  
 
h 2 r 2 h 2 r 2 h 2 r 2  e2 e 2 e2 
H ( RA , RB , r , PA , PB , p ) = −
ˆ ∇r − ∇A − ∇B +  − − + 
3 nuclear{
2me 2m p 2m p r1 r2 R
1424 3  1424 
KE of electron  electron-nucleus repulsion

 attraction energy


where,
r r r
r1 = r − RA
r r r
r2 = r − RB

In general the wave functions should contain all of the coordinates of the system
Because the protons are 1836 times heavier than the electrons they have much smaller
velocity this means that at the time scales associated with the electronic movement
the protons are stationary.

The Born-Oppenheimer Approximation

(1) The electronic wavefunctions depend only on the positions of the nuclei and not
on their momentum
(2) The positions of the protons are a parameter not a QM variable
(3) Protons have a fixed separation R just add a constant to the energy.
(4) Reduce the three body problem to an electron in a potential
(5) The Hamiltonian is separable

Using the procedure we developed:

Write the Hamiltonian for the electron energy and nuclear repulsion energy.

h2 r 2 e2 e2 e2
H el − nr = −
ˆ ∇r − − +
2me r1 r2 R
14442444 3
electronic energy H el

Hˆ elec uelec ( r ) = Eelec uelec ( r )

Eelec + Enuc nuc rep = Eelec + 1/ R
What we have is a functional dependence of the electronic energy on the nuclear
separation Eelec ( R) .
  
 
h r 22
h r 2 
2
e 2
Hˆ ( RA , RB ) = − ∇A + − ∇ B + Eelec ( R ) + 
2m p 2m p  {R 
 nuclear repulsion 
 energy 
H nuc ( RA , RB ) unuc ( RA , RB ) = Enuc unuc ( RA , RB )
ˆ
These eigenfunction reperesent the translation, rotation and vibration states of the
crystal.

uTot ( RA , RB , r ) = uelec ( r ) unuc ( RA , RB )

We would like the find the exact energy eigenfunctions because this is difficult we will
find approximate solutions
Hˆ el u = Eu
Assume that our solution can be written as a linear combination of atomic orbitals
r r r
χ ( r ) = c1 u100 ( r1 ) + c2 u100 ( r2 )
123 1424 3
φ1 φ2

where
r r r
r1 = r − RA
r r r
r2 = r − RB
Since this wave function is not an energy eigenfunction we would like to minimize
the average energy or the expectation value with respect to the coefficients,

∫ χ Hˆ χ dv
Eav = Hˆ = -∞
∞

-∞
∫ χχ dv
Take a closer look at the integrals,

 
 c u ( rr ) + c u ( rr ) Hˆ c u ( rr ) + c u ( rr ) dv
( )
∫  1 123
100 1
1424
2 100
3
2
 1 100 1 2 100 2

E ( c1 , c2 ) =  
φ φ
r r r r
( )(
1 2

)
∫ c1 u100 ( r1 ) + c2u100 ( r2 ) c1 u100 ( r1 ) + c2 u100 ( r2 ) dv
We are now going to apply the variation principle to allow us to find the function
which minimizes E (Cauchy Reiman Theorem).
∂E
=0
∂ci
Get a number of integrals which we will label in the following way:
r r
∫u 100 ( r1 ) Hu
ˆ
100 ( r1 ) dr ≡ H11
r r
∫u 100 ( r2 ) Hu
ˆ
100 ( r2 ) dr ≡ H 22
r r
∫u 100 ( r1 ) Hu
ˆ
100 ( r2 ) dr ≡ H12
r r
∫u 100 ( r2 ) Hu
ˆ
100 ( r1 ) dr ≡ H 21

r r
∫u ( r1 ) u100 ( r1 ) dr ≡ S11 = 1 r
 since u100 ( ri ) is orthonormal
100
r r
∫u 100 ( r2 ) u100 ( r2 ) ≡ S 22 = 1  
r r
100 ( r1 ) u100 ( r2 ) dr ≡ S12 

∫u r r
r r  ≠ 0 overlap integral u100 ( ri ) u100 ( rj ) are non-orthogonal
∫u 100 ( r2 ) u100 ( r1 ) dr ≡ S 21 


The coulomb energy integral:

r ˆ r
∫ u100 ( r1 ) Hu100 ( r1 ) dr ≡ H11 = H 22

r  h2 2 e2  r r  e2  r
= ∫ u100 ( r1 )  − ∇ −  u100 ( r1 ) dr − ∫ u100 ( r1 )   u100 ( r1 ) dr
144444  2m42444444
r1 
3 1444424444
r2 
3
E1 coulomb integral

The second term in the coulomb integral describes the attractive electrostatic
interaction between the proton B and the electron around proton A tends to lower the
energy.
The resonance energy contribution:
r ˆ r
∫ u100 ( r1 ) Hu100 ( r2 ) dr ≡ H12 = H 21

r  h2 2 e2  r r  e2  r
= ∫ u100 ( r1 )  − ∇ −  u100 ( r2 ) dr − ∫ u100 ( r1 )   u100 ( r2 ) dr
14444442444444 2m r2 
3 1444424444
r1 
3
E1S resonant integral

The interaction leads to the splitting and the creation of two energy levels

The overlap integral tends to destabilize the molecule.

Minimizing the energy expectation value with respect to the coefficients leads us to
the determinant problem,
( H11 − ES11 ) c1 + ( H12 − ES12 ) c2 = 0
( H 21 − ES 21 ) c1 + ( H 22 − ES22 ) c2 = 0
  H11 H12   c1   S11 S12   c1 
H   =E  
 21 H22   c2   S21 S22   c2 
H11 − ES11 H 12 − ES12
=0
H 21 − ES21 H 22 − ES 22
 1 H11 + H12
  → E+ =
 1 1 + S12
1 H11 − H12
  → E- =
 −1 1 − S12

e2
E− +
R
H11 − H12

H atom energy
1 e2
− E− +
2 R
H11 − H12
= E−
1 − S12
Energy (a.u= 27.21eV)

H11
−1 H11 + H12
= E+
1 + S12

−1.7 H11 + H12

 We get two solutions for the approximate eigenfunctions,

r
ψσ (r ) =
1
( u ( rr ) + u ( rr ) )
2 (1 + S )
100 1 100 2

r
ψσ∗ (r ) =
1
( u ( rr ) − u ( rr ) )
2 (1 − S )
100 1 100 2

Eσ < Eσ ∗

Consequences:

1. Odd and even wave functions the odd wavefunction has a node which is located
between the two nucluei.
2. Draw the wavefunction evaluated along the internuclear separation line.

3. Splitting of the energy levels as the internuclear spacing decreases.

Eel

R
4. The total energy of the molecule (including the proton-proton) repulsion.

bonding anti
bonding
Et

Et
Ro R Ro R

1. The lower electron eigenfunction or state has a corresponding minimum in the

total energy while the higher energy electon stae rise monotonically.
2. The internuclear distance at which the system acheives a minimum is the bond
length which is 2 R0
3. The energy minimum in the lower curve indicates a stable molecule and the
corresponding MO is call bonding MO. The other MO is called antibonding.
4. The electron in the bonding state will have a wavefunction that leads to a high
probability of finding the electron in between the nuclei –this allows for the
attractive e-p forces to offset the repulsive p-p interactions.

5. Near the minimum energy,

K
E; ( R − R0 )
2

2
K – is the bond stiffness.
6. In diatomic MO we distinguished between π and σ type bonds. What is the
significance of this labeling scheme? Recall that the AO were labeled with s, p, d
labels which were associated with the orbital angular momentum and the
spherically symmetric potential. Since L^2 commuted with the Hamiltonian we
could find a basis set which was common to Lˆ2 , Lˆz and the Hamiltonian. In a
linear molecule the rotational symmetry around the z axis is conserved hence
there is a conservation of the Lˆz component of the angular momentum that is why
the MO are labeled according to their symmetries.