Vous êtes sur la page 1sur 55

Paul Dirac (1902-1984 )

Dirac Notation
A powerful & concise formalism (also known as bra-ket / bracket notation) Two mathematical traditions: Heisenberg matrix mechanics vs. Schrdinger wave mechanics (matrices & vectors) (integrals). Computationally different but formally equivalent. Each has strengths & weaknesses. Dirac notation is a useful first step when a QM calculation is set up. (Then choose matrix or wave mechanics depending on which is computationally easier).

Dirac notation: Its advantages


Only a small number of elements ( ket, bra, bra-ket pair, ket-bra product, the completeness condition) We can build all of quantum theory from this Flexibility: straightforward to translate from one language to another (e.g. Fourier Transforms). Generality: setting up the problem doesnt require any computation Simple and elegant Widely used in QM textbooks so you have to get to know it!

Dirac notation: Kets, Bras, and Bra-Ket Pairs The Ket: |


is a vector

it contains what is known (e.g. p=2, or x=3) if we put a function into the Ket, e.g. | it represents a system in an eigenstate (also called a state vector)

represents the initial state of a system

The Bra: |
is also a vector represents the final state of the system. e.g. x=3| specifies final state has position x=3. can be used in conjunction with a Ket. e.g. x=3| is the probability amplitude of at position x=3. can be squared to give a probability. e.g. |x=3||2 is the probability of finding the particle at x=3.

alternatively, a Bra can contain the variable in which to express the ket. e.g.

is a function defining in coordinate space x.

example: for particle in a box,

x = (x ) = 2 sin ( x ) = 2 sin (x )
We might equally want in terms of momentum, in which case,

p = ( p ) which is a different function to (x )


But remember, contains all the information about the system so x and p are formally equivalent. i.e. If you have one you can transform into the other.

The Bra-Ket pair: |


Can be thought of as a dot product or vector projection - the projection of the content of the Ket onto the content of the Bra

e.g.

is the projection of the state onto the state . is equal to the probability amplitude that a system starting in state will end up in state .

A Ket-Bra product is sometimes referred to as an overlap integral. It is formally equivalent to the integral notation you have seen previously:

= *d

Similarly, the self-overlap Bra-Ket is equivalent to an integral as follows:

= *d
(we will see a proof of this in a few slides)

Further comments on Bra-Ket notation


can be complex, but projection of onto itself must be real (i.e. a real probability). Thus the bra must be the complex conjugate its corresponding Ket:

=
Hence,

will then be real

Also, the Ket is the transpose of its corresponding Bra. (See maths catch-up notes: matrices section)

Ket-Bra products, | |

(a.k.a. Projection operators)

The operator for projection onto some vector x is

x x

When this operator is applied to a state vector

we get

x x

This means the contribution of the state

to

In higher dimensions, we can think of a vector as being equal to the sum of its components in each orthogonal direction.

e.g. for simple vector in 2 dimensions,

a,

the projection onto each of its component parts in the orthogonal (and normalised) direction vectors i and j is given by

a= i ia + j ja

By comparing coefficients, we see that i.e. it is an identity operator.

i i + j j =1

This is known as the completeness condition.

Generally, when there are n discrete, orthogonal basis states, the sum of the identity operators over these states is given by

n
n

n =1

When the basis states are not discrete (e.g. position) , the sum becomes an integral:

n dx = 1

Putting in an a dummy wavefunction, we get:

= n n
n

= n n dx

This is a linear superposition in the basis set n.

Insertion of the identity operator into a bra-ket

= x x dx = * ( x ) ( x )dx
Notice that here we have replaced the vertical bar with the identity operator (and summed over all basis states) once the problem is written down (i.e. the basis set is chosen) we can evaluate this integral using traditional techniques.

Hence we have shown the equivalence of a Dirac Bra-Ket projection with the overlap integral of wave mechanics

We could show the same result by thinking of | and | as linear superpositions of the eigenstates of position, x:

= x x dx

and

= x' x' dx'


(assuming continuous basis states)

combining these as a bra-ket gives

= x' x' x x dx' dx = x x dx


The middle term disappears and we replace x by x, because position eigenstates are orthogonal. i.e.

x' x = 0

unless x = x in which case

x' x = x x = 0

Examples of Superposition

The following few slides show how superposition can be applied to QM problems. The examples themselves arent important, but the techniques are...

Particle Spin
The electron possesses an internal angular momentum called the spin. It can assume two values (up and down) in some arbitrarily chosen direction. Lets say that we specify the z direction as direction in which the spin is defined. Then the eigenstates can be given (in units of h/4) by the vectors below: spin up in the z direction:

1 S z ,up = 0 0 S z ,down = 1

spin down in the z direction:

These are kets because they denote the state of the system

These vectors form a complete, discrete basis set (i.e. they describe all the states that are possible).

Note that they are:

orthogonal, i.e. dot product = 0:

1 0 0 = 1 0 + 0 1 = 0 1

normalised, i.e. modulus = 1:

1 2 2 0 = 1 + 0 =1 0 2 2 1 = 0 +1 = 1

What if we want to know the spin vector in a different basis, i.e. the orthogonal direction, x ? Answer: We need to translate the eigenstates into the language we want, i.e. from spin along the z axis, to spin along the x direction. To do so we can take linear combinations of the two z-axis eigenstates:

S x ,up =

1 S z ,up + S z ,down 2

) )

S x ,down =

1 S z ,up S z ,down 2

Note the factor 1/2, which is required to normalise the linear combinations, because:

1 0 1 S z ,down + S z ,down = 0 + = 1 1 S z ,down S z ,down = 2

Exercise: Calculate the state vectors

S x ,up and S x ,down

Answer:

S x ,up =

1 S z ,up + S z ,down 2

S x ,down =

1 S z ,up S z ,down 2

1 2
1 2 1 2

1 0 + 0 1

1 2
1 2 1 2

1 0 0 1

S z ,up S z ,down S x ,up S x ,down S x ,up S x ,down S z ,up S z ,down

Notice that these two basis states are orthogonal (dot product = 0).

Similarly, these two basis states are also orthogonal (dot product = 0).

Also, all four state vectors are normalised. (i.e. modulus = 1)

So, and

S z ,up S z ,up = S z ,down S z ,down = 1 S z ,up S z ,down = 0

Exercise:
Determine whether

S x ,up

is orthogonal to

S z ,up

Answer: We need to find the angle between the two vectors, so we work out the dot product:

S x ,up S z ,up =

1 2 1 2

1 = 0

1 2

1 +

1 2

0 =

1 2

S x ,up S z ,up = S x ,up


1 = cos 2

S z ,up cos = 11cos

= 45o

i.e. they are not orthogonal. this means that S x ,up and S z ,up are not discrete basis states

10

Lets look further at one of those linear combinations, i.e.

S x ,up =

1 2 1 2

Given this starting state, Dirac notation allows us to calculate the probabilities of getting either spin up or spin down as the result of an experiment to measure spin along the x axis:

P(spin up) = S x ,up S x ,up

= 1 =1
2 2

P(spin down) = S x ,down S x ,up

=0

i.e. every experiment will give a measurement of spin up, when measured along the x axis.

What is the probability of getting a result of spin up if we try to measure spin along the z-axis? To find the probability, we must evaluate the following bra-ket pair: probability amplitude of ending up in spin up along z axis, given that the starting state is spin up along the x axis To evaluate the Bra-Ket pair we take the dot product of the two state vectors

S z ,up S x ,up 1 = 0 = 1 2
1 2 1 2

= 1

1 2

+ 0

1 2

This is the probability amplitude. Hence, probability=|probability amplitude|2 = 0.5

11

Similarly, the probability of getting a result of spin down along the z axis, given that the starting state is spin up along the x axis is given by

S z ,down S x ,up

1 2

Conclusions When the system starts in the state |Sx,up, every measurement of the spin in the x direction will give the same result: spin up. But measurement of spin in the z direction yields spin up in 50% of the experiments and spin down in the other 50%.

i.e Spin is well determined in the x-direction but not determined in the z direction.

NOTE: Linear superposition is NOT a mixture. We cannot regard |Sx,up as a 50/50 mix of |Sz,up & |Sz,down, because |Sz,up and |Sz,down are themselves linear superpositions of |Sx,up & |Sx,down:

S z ,up =

1 S x ,up + S x ,down 2

) )

S z ,down =

1 S x ,up S x ,down 2

If |Sx,up were a mixture of |Sz,up and |Sz,down, it would give an indefinite measurement of spin in the x-direction, despite the fact that its an eigenfunction of the x-spin operator.

12

Another example of linear superposition


The eigenfunctions for particle in a box form a complete basis. So we can use the completeness condition to write the wavefunction as a linear superposition of those eigenfunctions:

= n n
n

Projection of onto the eigenstate n i.e. the contribution that n makes to , in other words, just a number. Equivalent to cn coefficients you met with TRW. Thus,

= n cn
n

Consider a trial wave function for the particle in a box of length 1. (For the time being, lets just say we have pulled it out of a hat).

(x ) = 105 x 2 x 3

Is this trial wavefunction any good to describe any of the true eigenstates of the system? Remember that previously, we solved the particle in a box problem exactly and got the following expression for the complete basis:

n = 2 sin (nx ) n = 1, 2, 3,...

13

Because a state is given by a linear combination over the entire basis (due to the completeness condition), i.e.

= n cn
n

it is possible to expand the trial eigenstate (using software) into a linear combination of the true, orthogonal, eigenstates. The first few coefficients come out as follows:

105 x 2 x 3 = 0.935

2 sin (x )

-0.351

2 sin (2x ) +0.0035

2 sin (3x ) +...

Graphically, it looks like this:

= 0.935

-0.351

+0.0035

+...

The first coefficient is 0.935. This tells us that the trial function has a good deal of the true n=1 wavefunction in it.

So we can conclude that it is an OK approximation for the n=1 eigenstate.

However it wasnt quite right! It is an infinite expansion that includes bits of the entire basis set.

14

Operators, eigenvectors, eigenvalues, expectation values


In matrix mechanics, - Operators are matrices - States are vectors

Thus, matrices operate on vectors to give information about the state of the system. Note that a matrix, operating on (i.e. multiplying by) a vector gives a vector as the answer.

All observables have an associated operator. A system is either in a well defined state, or it isnt.

Example:
For spin, the operators are given (in units of h/4) by:

0 1 = S x 1 0

1 0 = S z 0 1

If we apply the spin operator to a spin state vector, we get the following:

0 1 S = S x x ,up 1 0
i.e.

1 2 1 2

1 2 1 2

S S x x ,up = S x ,up

with eigenvalue 1. Thus, |Sx,up is an eigenvector (or eigenfunction) of S x

15

on the other hand,

1 0 S S = z x ,up 0 1

1 2 1 2

1 2 1 2

= S x ,down

so we conclude that |Sx,up is not an eigenfunction of S z


Thus, Sx doesnt have a definite value in the z-direction. But even though we cant make any predictions about what the value of Sz will be, we can still obtain an expectation value for a large number of measurements...

A quick detour: general comments on expectation values


The expectation value for any observable, a, which has a corresponding operator , can be calculated as follows:

=a A = a A =a A A

A QM eigenvector equation

pre-multiply both sides by

a is a constant, so we can extract it


from the Bra-Ket

a=

expression for a.
divide by

to leave an

16

When the observable is energy, the operator is known as the Hamiltonian, and the expectation values for the energy is given by

E=

If is normalized, this simplifies to give a simple expression for the expectation value for energy:

E= H
Once more, this is analogous to the integrals you have seen already. i.e.

d E = *H

We can demonstrate this equivalence by looking at the expectation value for the position of a system in a state |:

x
the position operator Lets see what happens when we expand | into the states of the position operator:

= x x x dx x = x x x dx = * ( x ) x ( x )dx

simplification occurs because:

x =xx = x x x

17

Back to our problem.


Given that the system is in the state |Sx,up, what is the expectation value of a large number of experiments to measure spin in the z-axis? Answer:

S S S x ,up z x ,up =

1 2

1 2

1 0 0 1

1 2 1 2

=0

final state

operator for observable, i.e. z-axis spin

initial state

Notice that the Bra is the conjugate transpose of the Ket

The result of 0 is consistent with the idea that 50% of the time we will get spin +1 and 50% of the time we will get spin = 1 (hence giving a mean of 0).

The momentum operator in coordinate space


Wave-particle duality turns up once more! A particle or photon with wavelength has an un-normalized wavefunction of

x x = exp i 2
But de Broglies relation tells us that

p=

hence

h p

thus the momentum wavefunction in coordinate space is given by

x i 2px ipx = exp x p = exp i 2 = exp h ( p) h h

18

Now, just as in coordinate space,

x =xx x p =p p p

we can say that in momentum space,

But what about the momentum operator in coordinate space? We can use Dirac notation to help us out (surprise!)

p x p = x p p =p x p ipx ipx h d = p exp exp = h h i dx h d = x p i dx


Comparing the first line with the last line gives

= x p

h d x i dx

Matrix Elements

This section is based on Atkins MQM. See the book for more details...
The fundamental commutation relation of mechanics, implies that x and px are operators.

[x, px ] = ih

x and px can thus be represented by matrices

19

A matrix, M, is an array of elements (which can be complex).

Mrc is shorthand for the element of matrix M that occurs in and column c.

row r

For rules of matrix algebra, see PWB maths catch up notes, or Just the Maths on the web. Briefly, for multiplication: the product AB is another matrix C

Elements are given by

C rc = A rn B nc
n

For most matrices this is non commutative, so

AB BA 0

Heisenbergs matrix mechanics:


We represent position and linear momentum by the matrices x and px We require that

xp x p x x = ihI

I is the unit matrix: square with all diagonals = 1, and all other elements = 0

So far we have been looking at things like

n mA

(i.e. expectation value for state n to end up in state m)

These are sometimes abbreviated to element of a matrix.

A mn

implying that it is the

n mA

is often referred to as a matrix element of

20

a diagonal matrix element is a Bra-Ket of the form (i.e. Bra and Ket refer to the same state.)

m mA

Sums over the products of bra-kets of the form

s sA c rA

are actually matrix multiplications:

s sB B c = r A c r A
invoke the completeness condition in reverse

i.e. the sum is a single matrix element (bra-ket) of the product of the operators.

Exercise
Use the completeness condition to prove that the eigenvalues of the square of a Hermitian operator are non-negative. i.e. prove that for

2 x = x x A

then x 0 if is Hermitian

Remember, Hermicity means that Lets multiply both sides by

n = nA m mA

2 x = x x x = x x x = x 1 = x xA

We must therefore show that this expectation value 0

21

2 x xA A x = xA n nA x = x A
n

write the operator 2 as

employ the completeness condition


*

=
n

n xA n xA
2

write the Bra-Ket on the right the other way round (it becomes the conjugate: remember the definition of a Hermitian operator) for a complex number z, z(z*) is always positive, so the sum over n will be positive.

n = x A
n

QED: the square of the modulus will never be negative

Exercise Show that if then =0 A

f ) = A f (A
*

for any real function f,

22

The Variational Method


When we cant solve the Schrdinger Equation exactly, we must use an approximate method.

Real chemical problems cannot be solved analytically, so this is important! Variational method lies at the root of (amongst many other things) MO theory, i.e. by building wavefunctions from linear combinations of hydrogenic wavefunctions.

The Variation Principle: If an arbitrary wave function is used to calculate the energy, then the value calculated is never less than the true energy

If we adjust the coefficient(s) of the arbitrary wavefunction until we get the lowest energy, then those coefficients are best i.e. closest to the true wavefunction.

We may be able to lower the energy by using a more complicated trial wavefunction (e.g. a by taking a linear combination of known eigenstates of a similar problem, such as hydrogenic atomic orbitals)

23

Detour: Is it valid to split the expression for energy into a linear combination of eigenstates?
Lets expand in terms of the eigenstates, |n of the system, using the completeness condition:

E= H n n = H
n

Note: is normalised here

If |n is an eigenstate of the system, the Schrdinger equation allows us to replace

n H

by

En En = n E

Whats more, multiplication commutes, so

So

E= n E n
n

We have to figure out how to evaluate these integrals. if we are solving in coordinate space (x) then we can expand them, (once more, using the completeness condition)

E = x x n dxE n x x dx
n

E=
n

( x
*

x n dx En

) ( n x

x dx

E=
n

( x

n x dx En

) ( n x

x dx

these are just numbers, leading us to...

24

E = cn E n
2 n

This is the result that you saw in TRWs lectures i.e. when you have a linear superposition of wavefunctions,

= c11 + c2 2 + ...
then the expectation value for the energy is equal to the square of the coefficients, multiplied by the energy of each state. i.e.

E = c1 E1 + c2 E2 + ...
2 2

Final Comments

Comparison of

with

tells us that

H = n E n
n

i.e. the Hamiltonian can be written as a sum of projection operators involving the eigenstates (i.e. energy levels)

25

Back to the Variational principle Why is the energy always greater than the true energy?
Lets revisit particle in a box Weve already seen how this system can be solved exactly, so its a good (i.e. simple!) system to demonstrate the results of the variational method. Well not actually solve the problem here rather we look at the results and then revisit the variational method later. We pick an trial wavefunction:

(x ) = ax(1 x )

Recall that the eigenfunctions of the Hamiltonian (i.e. the complete solution) for L=1 is given by

n = 2 sin (nx )

with

En =

n 2 2 2

Our trial function, , is not an eigenfunction of the Hamiltonian

So this system does not have a well defined energy.

All we can do is calculate the average for many experimental measurements of E.

Each measurement will give a particular energy, En, (i.e. one of the eigenvalues of the Hamiltonian).

The coefficient |cn|2 gives the probability of En being measured.

26

Minimising the energy in the above problem gives an optimised trial wavefunction of:

( x ) = 30 x(1 x )
Which you may recall looks very much like the true E1 eigenstate of the particle in a box.

Using software to find the expansion coefficients of the true eigenfunctions gives:

c12 = 0.9987 c32=0.0014 c52 = 0.00006 c72 = 0.00001 All others are 0 (or really small)

Thus there is a 97.87% chance of getting

E1 =

12 2 = 4.93 2

Note that in the real world, we dont have the exact eigenfunctions at our disposal thats precisely why were employing the variational method! Thus, this example is for illustration only

In this example, the first term in the linear expansion is by far the most important, Thus we can see that is a pretty good approximation to the first pure eigenstate, E1.

27

The Variational Theorem says that no matter how hard you try in constructing trial wavefunctions, you cant get lower than the true ground state energy.

Why?
The only way can give the correct result for the energy of the ground state of the particle in the box, is if our trial wavefunction was (by chance!) equal to the ground state eigenfunction itself. In which case,

c1 = 1,

and all others,

cn >1 = 0

For any other trial wavefunction, c1

< 1, and other values, cn >1, are non zero. Thus the energy has to be greater than E1.

We will shortly be applying the Variational method to the He atom,

...but first a little diversion into Perturbation Theory, which is a different approximate method.

28

Perturbation theory for obtaining the energy of He


The Schrdinger equation has not been solved exactly for He

The Schrdinger equation:

= E H +V = E T

h2 2 ze 2 ze 2 e2 2 2m 1 + 2 + 4 r + 4 r + 4 r = E 0 1 0 2 0 3

(See questions from workshop in first PWB session.)

For simplicity, from now on, we will omit the constant factor 1/40 from the Coulomb potential. Now lets rewrite the Hamiltonian in a slightly different way:

h 2 2 ze 2 h 2 2 ze 2 e 2 H = + r + 2m 2 r 2m 1 r 1 2 3

= H

H0

+ H'

Note that H0 is the sum of the H-like Hamiltonians for each of the two electrons:

H 0 = h0 (r1 ) + h0 (r2 )

with

2e 2 h2 2 h0 (r ) = r 2m

29

Weve already solved the Schrdinger equation for the H atom and obtained an s atomic orbital as the ground state. The 1s wavefunction of a H atom,

1s (r ) , satisfies

h0 (r )1s (r ) = E1s 1s (r )
Energy of hydrogenic 1s orbital What if we ignore the electron-electron interaction in He? (i.e. what if we ignore the Hamiltonian H ) Then the 2 electrons will then be independent of one another and we need only consider the 2 electron system with H0. We shall call this a reference Hamiltonian.

Each electron will follow its own H-like Hamiltonian h0 and occupy the 1s orbital independently. Note: here we neglect here the of the electron and implicitly assume that they are antiparallel. The Schrdinger equation for He is then

H 0 (0 ) (r1 , r2 ) = E (0 ) (0 ) (r1 , r2 )
with solution:

(0 ) (r1 , r2 ) = 1s (r1 )1s (r2 )

This is consistent with the Aufbau configuration, He: (1s)2

30

Question:
What assumption allows us to say that

(0 ) (r1 , r2 ) = 1s (r1 )1s (r2 )

Answer: This is the orbital approximation

Exercises
1. Using

h0 (r )1s (r ) = E1s 1s (r )

, show that
(0 )

(0 ) (r1 , r2 ) = 1s (r1 )1s (r2 )

is a solution of H 0

(r1 , r2 ) = E (0 ) (0 ) (r1 , r2 )

2. Express E(0) in terms of E1s. How can you explain this in terms of the assumed relation between the two electrons?

3. If we define the energy of the 1s orbital of the H atoms as

EH = hcRH = 13.6 eV
Express E1s in terms of EH and z.

31

We use this approximate wavefunction to evaluate total energy. i.e.

E (0 ) H (0 ) = (0 ) H 0 + H ' (0 ) = (0 ) H 0 (0 ) + (0 ) H ' (0 ) = E0 + ( 0 ) H ' ( 0 )


z = 2 for He

We use the 1s H-like wavefunction,

zr 1 z 2 exp 1s (r ) = a a 0 0
Bohr radius

...and we can explicitly write the electron-electron interaction energy as

(0 )

H'

(0 )

1 z = 2 a 0

2 z (r1 + r2 ) e2 r12 exp dr1dr2 a 0

the integration is complicated but tractable. The answer is:

(0 ) H ' (0 ) =

5 zE H 4

32

Recall from TRW sessions that for Hydrogenic atoms,


2 1 ze EH (n ) = 2 2 n 4 0 h 2

So for He, we have a ground state energy of

5 11 E1 E (0 ) + (0 ) H ' (0 ) = 2 z 2 z E H = E H = 74.80 eV 4 2
(Remember, we are ignoring the 1/40 here )

c.f. experimental value of 78.98 eV

Now back to the Variational Method, applied to He.


We start with a trial wavefunction that is the product of two 1s atomic hydrogenic wavefunctions (as before), but we swap the charge z for a variable parameter :

(r1 + r2 ) 1 (0 ) (r1 , r2 ) = 1s (r1 )1s (r2 ) = exp a a 0 0


3

we can think of as the effective nuclear charge which takes into account the shielding effect of the second electron.

33

If we optimise to minimise the energy, will be less then z. This lower effective z will mean that the orbitals are bigger. We can justify this as being due to e-e repulsion. So our job is to find the minimum of the function

E ( ) = H
The integral is similar to the one we saw for perturbation theory, and the result is

5 E ( ) = 2 2 + 4 z EH 4
This quadratic equation has 1 minimum which determines the best Question: find . Is it bigger or smaller than z? How do you interpret it?

Comparison of Perturbation Theory and Variational Method for He.

1st order perturbation E1 / eV -74.80

Variation

Experiment

-77.45

-78.98

IP / eV

20.40

23.05

24.58

34

Perturbation theory some generalisations


This section doesnt include the derivations! See text books if interested...

Method:
1. Divide the Hamiltonian into a known, reference Hamiltonian, H0, and a small remaining part, V:

H = H0 +V
(0 ) (0 )

2. Excluding V, solve to get a wavefunction n(0) and En:

H 0 n

(0 )

= En n

(n = 0,1,2,...)

3. Use the solved expressions for n(0) and En to find approximate solutions to the full Schrdinger equation:

H n = En n (n = 0,1,2,...)

The energy can be worked out as follows:


(0) V k( 0 ) n 2

En = En

(0 )

(0) (0) V n + n + k n

En

(0 )

Ek

(0 )

+ ...

In the He example, we only used the first term i.e. it was first order.

Weaknesses: perturbation theory breaks down when there are degeneracies (i.e. En(0) = Ek(0)).

n =

( 0) n

+
k n

(0) V k( 0) n

En

(0 )

Ek

(0 )

k( 0 ) + ...
(this is a mixing coefficient)

35

The modification to n(0) to give n is given by other functions k(0) The mixing coefficients are proportional to the integral i.e. the bigger the overlap between k(0) n(0) the higher the mixing coefficient a.k.a. the transition matrix element
( 0) n V k( 0 )

The mixing coefficients are also proportional to

En

(0 )

1 (0 ) Ek

This means that the coefficient is bigger when the energies are similar. i.e. nearly degenerate states mix better

The Variational Method as applied to the LCAO-MO Method


(i.e. Linear combination of atomic orbitals-molecular orbitals)

Diatomic Molecules

AB

The Variational Method is versatile: any sensible trial function will do. Often a linear superposition of basis states is easiest:

( x ) = cn f n ( x )
n

coefficients taken as variational parameters Remember, the aim is to optimise the variational parameters to minimise the energy

36

For a diatomic molecule, there are two atoms, each supplying one electron to the covalent bond. Thus our linear combination of hydrogenic basis functions will look like this:

= c A A + cB B
The basis wavefunctions are normalised but at this stage, not necessarily orthogonal. i.e. the overlap integral, S is given by

The coefficients are real numbers

S AB = B A 0

Recall what we saw last week on matrix elements:

We can write the expectation value,

A H B
as a matrix element:

H AB

37

E AA = A H A E BB = B H B
E AB = B H A E AB = A H B

Expectation values for the energy of the atoms.

Expectation value for resonance energy of electrons delocalised across both atoms.

These matrix elements may be... calculated (as for He atom earlier) or estimated from spectroscopy or simply taken as parameters (see Hckel

theory later)

E=

Expectation value for energy

c A A + cB B H c A A + cB B c A A + cB B c A A + cB B

Substitute linear combination for

c A A H c A A + c A A H cB B + cB B H c A A + cB B H cB B c A A c A A + c A A cB B + cB B c A A + cB B cB B c A c A A H A + c AcB A H B + cB c A B H A + cB cB B H B c A c A A A + c AcB A B + cB c A B A + cB cB B B c Ac A H AA + c AcB H AB + cB c A H BA + cB cB H BB c Ac A S AA + c AcB S AB + cB c A S BA + cB cB S BB c A H AA + 2c AcB H AB + cB H BB 2 2 c A + 2c AcB S AB + cB


2 2

expand Bra-Kets

extract constants

write integrals as matrix elements

gather like terms, and notice that SAA = SBB = 1 provided that A and A are normalised

38

So, if the expectation value for energy is given by,

c H + 2c AcB H AB + cB H BB E = A AA2 2 c A + 2c AcB S AB + cB


2 2

we must then vary the parameters cA and cB in order to minimise this energy. The minimum will occur when

E =0 c A

and

E =0 cB

The differentiation is easier if we rearrange as follows:

E=

E H =0
in full,

E c A + 2c AcB S AB + cB c A H AA + c AcB (H AB + H BA ) + cB H BB = 0
2 2 2 2

) (

So now we differentiate this

39

E c A + 2c AcB S AB + cB c A H AA + c AcB (H AB + H BA ) + cB H BB = 0
2 2 2 2

) (

Differentiation with respect to cA (product rule)


E c A + 2c AcB S AB + cB c A H AA + 2c AcB H AB + cB H BB 2 2 E + c A + 2c AcB S AB + cB =0 c A c A c A
2 2 2 2

) (

E (2c A + 2cB S AB ) + c A + 2c AcB S AB + cB


2

E ) (2c H c
A A

AA

+ 2cB H AB ) = 0

When

E = 0 this becomes: c A

E (2c A + 2cB S AB ) (2c A H AA + 2cB H AB ) = 0

gathering up all the terms in cA and cB gives:

(H AA E )cA + (H AB ES AB )cB = 0
Similiarly, from

E = 0 we get: cB

(H BA ESBA )cA + (H BB E )cB = 0

These are simultaneous equations which we can write as a matrix equation

H AA E H ES BA BA

H AB ES AB cA 0 = H BB E cB 0

Now is a good time to review the catch-up maths notes if you havent done so already. Make sure that you can find the eigenvalues and eigenvectors of a 2x2 matrix

40

we could write the 2x2 matrix in this eigenvector equation as:

H AA E H ES BA BA

H AB ES AB H AA = H BB E H BA

H AB 1 S AB E H BB S BA 1

Thus, the full equation will be:

H AA H BA

H AB 1 S AB c A 0 E 0 S c = H BB 1 BA B

In an abstract and more general form (i.e. which is applicable to any number of electrons and not just the 2 we are looking at here), it is:

(H ES ) c = 0

This allows us to get a value for the energy, E, of the MOs, and the coefficients of the AO linear combinations.

Solving the secular equation to find E


In order to have non-zero solutions (i.e. c0) the matrix must have a determinant of 0. i.e.

H ES

H ES = 0

This is called the secular equation and it can be solved to give the energy, E, of the MOs.

When we have the eigenvalues, E , we can then find the atomic orbital coefficients, c, from

(H ES ) c = 0

(Actually, in our 2 electron case, it will only give a ratio for cA / cB i.e. you have to guess one number and the other can be calculated. Ultimately, normalisation will give the absolute values of cA and cB)

41

We can simplify things at this stage by choosing an orthonormal set of atomic orbitals as the basis set. Then the overlap integrals become either 1 or 0:

S AB = B A = 0 S AA = A A = 1

This means that the overlap matrix turns into the identity matrix:

S=I

Hence,

(H ES )c = 0 (H EI )c = 0
Hc Ec = 0 Hc = Ec

This is a matrix eigenvector problem which is easily solved. (Check the maths notes for how to do it by hand for a 2x2 matrix.) Reminder: this isnt the Schrdinger equation! H is a matrix of energy integrals, and not an operator matrix. c just gives the coefficients of the linear combination of atomic orbitals. For bigger matrices, this is done by diagonalisation using computers.

42

For a 2x2, we can do it long hand...

H AA H BA

H AB c A cA = E c H BB cB B

eigenvector equation

H AB c A H AA E H =0 H BB E BA cB
H AA E H BA H AB H BB E =0

take the eigenvalue into the matrix

set the determinant to 0 expand the determinant

(H AA E )(H BB E ) H AB H BB = 0
E 2 (H AA + H BB )E (H AA H BB H AB H BB ) = 0

This is a quadratic equation

E 2 (H AA + H BB )E (H AA H BB H AB H BB ) = 0

Substitute the coefficients of E into the quadratic formula to solve for E:

E=

1 (H AA + H BB ) 2

( H AA H BB )2 + 4a(H AA H BB H AB H BB )

We wont do it here, but you would then substitute these values for E into the equation

(H ES ) c = 0

to find c1/c2

43

Hckel MO Theory
A simple treatment of the MOs of conjugated planar hydrocarbons A series of approximations to help us solve the secular equation It is a semi-empirical method as it relies on experimental data. We consider only one pz atomic orbital per C atom involved in the delocalised system., which comes from a 2pz AO We describe the molecular orbitals as a linear combination of these 2pz atomic orbitals.

To simplify, we make some drastic assumptions:

That the basis set, the 2pz atomic orbitals, are orthonormal, i.e.

S=I

(Note that if this were strictly true there would be no bonding! Neighbouring atoms will have p overlap, but much smaller than the 2p overlap) All the overlap integrals (Sij) are set to 0. All the diagonal elements of the Hamiltonian matrix are represented by a single parameter, . i.e. H11 = H22 = ... = The off diagonal elements are set to different common parameter, Hij = , but only when atoms i and j are connected. Otherwise, Hij = 0

44

Implications of setting diagonal energy elements to , and all non diagonal energy elements are 0 unless atoms are touching:

This implies that the energy is purely atomic in nature. It is consistent with a picture of bonding that is fully delocalised across the whole system and the environment of the atom can be neglected. A result is that conformational isomers such as cis- / trans- are not distinguished from one another.

Worked example: Butadiene

This molecule has 4 carbon atoms contributing to the system, so we use four 2pz atomic orbitals. Thus we are looking for 4 MOs that are linear combinations of the 4 atomic pz orbitals:

1 = c1 A + c2B + c3C + c4D

2 = c5 A + c6B + c7C + c8D 3 = c9 A + c10B + c11C + c12D 4 = c13 A + c14B + c15C + c16D

45

We assume no overlap integrals, so the equation is:

Hc = Ec
4pz atomic orbitals means that the Hckel Hamiltonian is a 4x4 matrix. Thus we have...

c1 0 0 c1 c2 0 c2 E = c 0 c 3 3 0 0 c c 4 4

The secular equation that we must solve is:

E 0 0 E 0 =0 E 0 E 0 0

This is solvable by hand (e.g. using Laplace expansion by co-factors).

46

If we really had to work it out by hand, the determinant of the secular equation would expand as follows:

( E )4 3 2 ( E )2 + 4 = 0
( E )
= 3 2

Notice that this is a quadratic in ( E)2

( 3 )
2
1

2 2

4 4

3 5 = 2 2

3 5 2 E = 2

E = + 1.618 E = 1.618 E = + 0.618 E = 0.618

Alternatively, its much easier using a computer. In Maple:

Notice that the answers are the same as before!

47

We neednt stop here: we can find the eigenvectors too, i.e. the coefficients of the four AOs in each MO...

If we go to the trouble of normalising these eigenvectors, we get the following coefficients for the linear superposition of atomic pz orbitals:

c1 0.372 0.602 0.602 0.372 c2 0.602 0.372 0.372 0.602 c = 0.602 , 0.372 , 0.372 , 0.602 3 c 0.372 0.602 0.602 0.372 4

48

MO Energy levels in Butadiene calculated using Hckel approximation (as viewed down the axis of the pz-orbitals)

Now that we have the coefficients for the linear superposition, we can sketch the wavefunctions

49

Homework...
1. Calculate the Hckel MO energies and coefficients for ethylene (C2H4) 2. For the allyl cation (C3H5+), work out the Hckel MO energies and if youre up to the challenge, calculate the coefficients as well.

3. Write down the Hckel Hamiltonian for benzene and for hexatriene (C6H8). How do they differ?

1. Ethylene

Hc = Ec

c1 c1 = E c c 2 2

c1 0 E = E c2 0

50

The secular equation is given by

E =0 E

Expanding the determinant:

( E )( E ) 2 = 0
2 2E + E 2 2 = 0
E = ( + ) E = ( )

solving, we get:

To find the coefficients for each MO, we must substitute eachvalue for E in turn into the eigenvector equation...

For E = + :

c1 c1 = ( + ) c c2 2

c1 + c2 = ( + )c1 c1 ( + )c1 = c2 c1 = c2 c1 = c2

51

For E = -

c1 c1 ( ) = c c 2 2

c1 + c2 = ( )c1 c1 ( )c1 = c2 c1 = c2 c1 = c2

Thus, the MOs are given by

1 = c1 A c1 B 2 = c1 A + c1 B
Finally, we must normalise:

where A and B are the wavefunctions for the pz atomic orbitals that form the basis.

c1 + c1 = 1
2 2 2

Hence, final answer is:

c1 = c1 =

1 2 1 2

1 = 2 =

1 2

A A +

1 2

B B

1 2

1 2

52

2. C3H5+

Hc = Ec 0 0 c1 c1 c2 = E c2 c 3 c3
only 2 electrons, but we must consider 3 pz atomic orbitals

0 c1 0 E E c2 = 0 0 E c3 0

E 0 E =0 0 E

( E )

E E +0 =0 E 0 E 0

( E )(( E )( E ) 2 ) ( ( E ) 0 ) = 0
...then expand brackets and solve for E

53

Alternatively, use Maple (or similar) Maple syntax... at the prompt, type the following:

set linear algebra mode enter matrix, called H solve the secular determinant for e

with(linalg): H:=Matrix([[a,b,0],[b,a,b],[0,b,a]]); eigenvalues(H);

Answers:

E = a, a + 2b, a 2b

Now we need to find the coefficients i.e. the eigenvectors of

Lets use Maple to do the whole thing in one go...

with(linalg): H:=Matrix([[a-E,b,0],[b,a-E,b],[0,b,a-E]]); eigenvectors(H);

54

Output looks like this:

eigenvalue multiplicity of this eigenvalue ditto for 2nd and 3rd eigenvalues

eigenvector corresponding to this eigenvalue Finally, remember to normalise the coefficients...

3. Benzene vs. hexatriene

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

55

Vous aimerez peut-être aussi