THEOREMS OF QUANTUM MECHANICS

In order to develop methods to treat many-electron systems

(atoms & molecules), many of the theorems of quantum
mechanics are useful.

Useful Notation

The matrix element Amn is defined by

Amn = ∫ φm* A φn dτ

= <φmAφn> bracket notation

= (φmAφn)

= <mAn>

The overlap integral between two functions is

∫ φm*φn dτ = <φmφn> = <mn>.

The complex conjugate of the overlap integral is

[∫ φm*φn dτ]* = ∫ φn*φm dτ = <nm>

= [<mn>]*

Also <mm>* = <mm>.

Theorems

The average value of a physical quantity must be a real number.

 Let A be the linear operator for the property A. So if A is
real, then

<A> = <A>*

and A is said to be a Hermitian Operator. For a Hermitian

Operator:

<A> = ∫ ψ* Aψ dτ = <A>* = (∫ ψ* Aψ dτ)*

= ∫ ψ (Aψ)* dτ

Using the above relation, prove ∫ f* Ag dτ = ∫ g (Af) * dτ.

If ψ = f + cg & A is a Hermitian operator, then

∫ (f + cg) * A(f + cg) dτ = ∫ (f + cg)[ A(f + cg)] * dτ.

Left-hand side = ∫ (f + cg) * A(f + cg) dτ

= ∫ (f* + c*g*) A(f + cg) dτ

= ∫ (f* + c*g*) Af dτ + ∫ (f* + c*g*) Acg dτ

= ∫ f*Afdτ + c*∫ g*Afdτ + c ∫ f*Ag dτ + cc*∫ g*Ag dτ

By symmetry, the right-hand side

= ∫ f(Af)*dτ + c∫ g(Af)*dτ + c*∫ f(Ag) *dτ + cc*∫g(Ag) * dτ

Since A is Hermitian

& ∫ f*Afdτ = ∫ f(Af)*dτ & ∫ g*Ag dτ = ∫g(Ag) * dτ,

then, from the left- & right-hand sides:

 c*∫ g*Afdτ + c ∫ f*Ag dτ = c∫ g(Af)*dτ + c*∫ f(Ag) *dτ.

Since this must hold for all c’s, it must hold for c=i or 1. Set c=1
to give one eqn., set c=i to give a second eqn. Then add.

c=1: ∫ g*Afdτ + ∫ f*Ag dτ = ∫ g(Af)*dτ + ∫ f(Ag) *dτ

c=i: -∫ g*Afdτ + ∫ f*Ag dτ = ∫ g(Af)*dτ - ∫ f(Ag) dτ; i* = -i

Add eqns.⇒ ∫ f*Ag dτ = <φiAφj> = <φjAφi>* = ∫ g(Af)*dτ.

So for a Hermitian operator,

<φiAφj> = <φjAφi>* or <iAj> = <jAi>*

or Aij = Aji*

What Operators Are Hermitian?

Is V(x), the potential energy operator, Hermitian?

<φjV(x)φi>* = ∫-∞∞φj (Vφi) * dx

V* = V & V is just a multiplicative operator (no square roots,

etc). So

∫-∞∞φj V * φi * dx = ∫-∞∞φj Vφi * dx = ∫-∞∞φi *Vφj dx

<φiV(x)φj> Hermitian

Is px = -ih∂/∂x Hermitian? (px* = ih∂/∂x)

<φi px φj> = ∫-∞∞φi * (-ih∂/∂x)φj dx = -ih∫-∞∞φi *φj’ dx

Integration by parts: ∫ uv’ = uv - ∫vu’

So -ih∫-∞∞φi *φj’ dx = -ih[φi * φj-∞∞ - ∫-∞∞φj (φi*)’ dx]

Assume φi & φj are well-behaved (i.e. =0) at +∞, so that the first
term on the right equals 0. Then

<φi px φj> = ih ∫-∞∞φj(φi*)’ dx = ∫-∞∞φj( ih ∂/∂x)φi * dx

= ∫-∞∞φj(px) *φi * dx = <φj px φi>* Hermitian

(Prove: T, the kinetic energy operator, is Hermitian).

Then H = T + V is Hermitian.

PROVE: The eigenvalues of a Hermitian operator are real. (This

means they represent a physical quantity.)

For A φi = b φi, show that b = b* (b is real).

If A is Hermitian, then ∫ φi *Aφi dτ = ∫ φi (Aφi)* dτ.

Or, ∫ φi *bφi dτ = ∫ φi (bφi)* dτ = ∫ φi b*φi* dτ

Then b∫ φi *φi dτ = b*∫ φi φi* dτ = b*∫ φi *φi dτ

So b = b*
PROVE: The eigenfunctions of a Hermitian operator can be
chosen to be orthogonal.

Show that, if B F = s F & B G = t G & t is not equal to s,

then <FG> = 0.

Since B is Hermitian
 <F B G> = <G B F>*

Or <F tG> = <G s F>*

So t <F G> = s*<G F>* = s*<FG> = s<FG>

(t-s) <FG> = 0

t not equal to s ⇒ <FG> = 0

The requirement that t is not equal to s means that F & G

are independent eigenfunctions that have different
eigenvalues (i.e. they are non-degenerate)

PROVE: That in the case of degenerate eigenfunctions, we can

construct from these eigenfunctions a new eigenfunction that will
be orthogonal.

Remember: We have shown that any linear combination of

degenerate eigenfunctions corresponding to the same eigenvalue
is also an eigenfunction with the same eigenvalue.

Let B F = s F & B G = s G

Let φ1 & φ2 be the new eigenfunctions that will be

orthogonal.

Set φ1 = F, φ2 = G + c F. Find c such that <φ1φ2> = 0.

Procedure: Schmidt Orthogonalization

<φ1φ2> = <F G + cF> = <F G > + c <F F >

If <φ1φ2> = 0, then c = - <F G >/<F F >

(Unless otherwise noted, assume all eigenfunctions are
orthogonal & normalized: <φiφj> = δij = 0 unless i = j; δii = 1)

Expansion in terms of eigenfunctions:

We can use the eigenfunctions of a Hermitian operator to
describe an arbitrary well-behaved function. We can expand the
arbitrary function in terms of all (or a complete set) of
eigenfunctions of the operator.

Let f be an arbitrary well-behaved function that obeys the

same boundary conditions as the complete set φi

f = Σ a i φi a i is an expansion coefficient
i

To find a i (formal solution):

<φj f> = <φj Σ a i φi> = Σ a i<φj  φi> = Σ a iδji = a j

i i i

Or f = Σ <φi f> φi
i
So, the eigenfunctions of a Hermitian operator form a complete
orthonormal set with real eigenvalues

Eigenfunctions of Commuting Operators:

In Chapter 5 we stated that a wavefunction can be

simultaneously an eigenfunction of two different operators if
those operators commute. Or, more exactly, a necessary
condition for the existence of a complete set of simultaneous
eigenfunctions of two operators is that the operators commute
with each other. This means that the physical properties
associated with the operators can be measured simultaneously.
PROVE: If there exists a common complete set of
eigenfunctions for two linear operators, then the operators
commute.

Let φi be the complete set of eigenfunctions of the

operators A & B.

A φ i = s iφ i & B φ i = t iφ i

Show that [A,B] = 0 or (AB - BA)f = 0 where f is an

arbitrary function.

We can expand f in terms of the complete set of

eigenfunctions of A & B:

f = Σ c i φi
i

So (AB - BA) f = (AB -BA) Σ c i φi = Σ c i(AB -BA) φi

i i
= Σ c i(ABφi -BAφi) = Σ c i(A t i φi -B s i φi)
i I

= Σ c i(t i A φi - s i B φi) = Σ c i(t i s i φi - s i t i φi)

i i

= Σ c it i s i (φi - φi) = 0
i

The important point here is that both operators must have a

common complete set of eigenfunctions. The existence of just
one eigenfunction in common is not enough to guarantee that
[A,B] = 0.
Look over the proofs for:

If A & B commute, we can select a common complete set of

eigenfunctions for them.

If A is a Hermitian operator with eigenfunction φi such that Aφi =

s iφi & [A,B] = 0, then Bij = <φiB φj> = 0 (si not = sj).

PARITY OPERATOR - a quantum mechanical operator that

has no classical mechanical equivalent

Π f(x,y,z) = f(-x,-y,-z)

The parity operator, Π, replaces the Cartesian coordinates with

their negative values.

Example: Π (x2 - z eAy) = (x2 + z e-Ay)

[In Cartesian coordinates, Π (x,y,z) = (-x,-y,-z). What about

spherical polar coordinates? The allowed ranges for the
variables are:

0 < r < ∞, 0 < θ < π, 0 < φ < 2π

To move into the quadrant of (-x,-y,-z),

r → r, φ → π + φ, θ → π - θ]

Find the eigenvalues of the parity operator:

Π gi = ci g i

First, find Π2:

 Π2 f(x,y,z) = Π [Π f(x,y,z)] = Π f(-x,-y,-z) = f(x,y,z)

So Π2 = 1 (unit operator)

Then Π2 gi = Π [Πgi] = Π ci gi = ci Π gi = ci ci gi = ci 2gi

So ci 2 = 1 & ci = +1

Find the eigenfunctions of the parity operator:

Π gi = ci g i

Or Π gi (x,y,z) = + gi (x,y,z)

And Π gi (x,y,z) = gi (-x,-y,-z)

If ci = +1, gi (x,y,z) = gi (-x,-y,-z) & g is an even function

If ci = -1, gi (x,y,z) = -gi (-x,-y,-z) & g is an odd function

So the eigenfunctions of Π are all the possible well-behaved

even & odd functions.

The parity relationships are useful in constructing variational

wavefunctions & molecular wavefunctions (later chapters).

If Π & H commute, we can select a common set of

eigenfunctions.

H = -h2/(2m) (∂2/∂x2 + ∂2/∂y2 + ∂2/∂y2) + V

[H, Π] = -h2/(2m)[∂2/∂x2, Π] - h2/(2m)[∂2/∂y2, Π]

- h2/(2m)[∂2/∂z2, Π] + [V, Π]

Consider [∂2/∂x2, Π] = ∂2/∂x2 Π - Π ∂2/∂x2:

 Π ∂2/∂x2φ(x,y,z) = ∂/∂(-x) ∂/∂(-x) φ(-x,-y,-z)

= [-∂/∂x][-∂/∂x] φ(-x,-y,-z)

= ∂2/∂x2 φ(-x,-y,-z) = ∂2/∂x2 Π φ(x,y,z)

So [∂2/∂x2, Π] = 0 = [∂2/∂y2, Π] = [∂2/∂z2, Π]

Consider [V, Π] = VΠ - ΠV

ΠV(x,y,z) φ(x,y,z) = V(-x,-y,-z) φ(-x,-y,-z)

If V is an even function, V(x,y,z) = V(-x,-y,-z)

ΠV(x,y,z) φ(x,y,z) = V(x,y,z) φ(-x,-y,-z)

= V(x,y,z) Πφ(x,y,z); V & Π commute.

Otherwise they don’t commute.

So [H, Π] = 0 if V is an even function. When V is even, we can

choose the ψ so that they are even or odd, i.e. have definite
parity. This is used in the Variation Method to construct the
appropriate wavefunction.

MEASUREMENT & SUPERPOSITION OF STATES:

The basic method is a scheme for calculating the probabilities of

various possible outcomes of a measurement.

Example: If the state function, Ψ(x,t) is known, the probability

of finding the particle between x & x+dx is Ψ(x,t) 2dx.
In general, consider the property, G: How can we calculate the
probability for each possible result of the measurement of G?
(Assume there are N particles & three coordinates; Let q
represent the position coordinates.)

G φi(q) = g i φi(q)

The eigenfunctions of any Hermitian operator form a complete

set (i.e. they are all the linearly independent eigenfunctions). The
φi form a complete set, so we can expand any arbitrary
wavefunction in terms of them:

Ψ(q,t) = Σ c i(t)φi(q)
i

Require

∫ Ψ*Ψ dτ = 1,

where dτ is the volume element for the spatial coordinates (not

time). Then

1 = ∫ Σc i*(t)φi*(q) Σ c j(t)φj(q) dτ
i j

= Σc i*Σ c j ∫φi*(q)φj(q) dτ

= Σc i*Σ c jδij
i j

= Σc i2
i

Choose the c i’s so that 1 = Σc i2

i
Since Ψ(q,t) is a normalized state function, we can write the
average value of G as

<G> = ∫ Ψ(q,t)* G Ψ(q,t) dτ

= ∫ Σc i*(t)φi*(q) G Σ c j(t)φj(q) dτ
i j

= Σc i*Σ c j ∫φi*(q) G φj(q) dτ

i j

= Σc i*Σ c j ∫φi*(q) gj φj(q) dτ

i j
 = Σc i*Σ c j gj ∫φi*(q) φj(q) dτ
i j

= Σc i*Σ c j gj δij
i j

= Σc i2 gi
i

We have previously defined the average value in terms of the

probability of getting one of the eigenvalues, gi, when G is
measured (i.e. When a property is measured, we can only get
one of the eigenvalues as a result. No other numbers are
possible.)

Then <G> = Σ Pgi gi,

i

where Pgi is the probability of finding the eigenvalue gi. So

Pgi = c i2

We can predict the result of the measurement of G with certainty

only if all the c i’s except one are 0:

c i = 0, i not equal to k; c k not equal to zero.

Then c k2 = 1 & the result of the measurement will be gk &

ψk = φ k since all the other c’s are 0.

So Ψ(q,t) = Σ c i(t)φi(q)
i

gives the state function as a superposition of eigenstates, φi, of G.

The coefficient, c i, of φi in the expansion is related to the
probability of finding the eigenvalue gi when G is measured (i.e.
The larger the contribution of φi in the expansion - indicated by
the magnitude of ci - the larger the probability of measuring that
eigenvalue).

Calculate ci:

Ψ(q,t) = Σ c i(t)φi(q)
i

∫ φj*(q) Ψ dτ = ∫ φj*(q) [Σ c i(t)φi(q)] dτ

i

= Σ c i(t) ∫ φj*(q) φi(q) dτ

i

= Σ c i(t) δji = c j = <φj Ψ> = probability amplitude

i

So the probability of measuring gj as a value of G is:

 c j 2 = <φj Ψ>2 = ∫ φj*(q) Ψ (q,t) dτ2

So, if we know Ψ, we can predict the outcome of the

measurement of G.

Example: G = px linear momentum

φi = eik x/h;
i
g i = ki

We previously found the general form for the

wavefunction for a free particle in one dimension:

Ψ = a1 e-iEt/h ei√(2mE) x/ h + a2 e-iEt/h e-i√(2mE) x/ h

= c1 eik x/ h + c2 eik x/ h,
1 2

where c1 = a1 e-iEt/h, c2 = a2 e-iEt/h, k1 = √(2mE), k2 = -√(2mE).

So  c 12 =  a1 e-iEt/h2 =  a12 eiEt/h e-iEt/h =  a12

= the probability of getting k1 when measuring G.

And  c 22 =  a22

= the probability of getting k2 when measuring G.

The probability of getting any other number when measuring G

is zero.

So Ψn(q,t) = e-iE t/h ψ n(q) is a stationary state

n

& Hψ n(q) = E n ψ n(q)

For a stationary state, the probability density does not depend on

time:

Ψn(q,t) 2 = ψn(q) 2 is independent of t.

If we take a superposition of stationary states, Ψn,

Ψ = Σ cn Ψn = Σ cn e-iE t/h ψ n
n

n n

is not an eigenfunction of H, but is an eigenfunction of the time-

dependent Schrödinger Eq.

[-ih ∂/∂t + H] Ψ = 0

Ψ doesn’t have a definite energy because it is a combination of

the ψ n, each with energy E n. The probability of getting E n
when the energy is measured is
  cn e-iE t/h2 =  cn2.
n

Ψ is a nonstationary state because the probability density

depends on t due to cross terms of the form

e-iE t/h eiE

n n’
t/h

Physical example: If a system that is in a stationary state is

exposed to radiation (i.e. a laser field whose potential varies with
time) the state changes to a nonstationary state.