Lecture 1: The superposition principle

Read Baym Chapter 1, Chapter 3; Sakurai Chapter 1

1 Why do we need to learn quantum mechanics (QM)?

Without quantum mechanics, we cannot understand the world.

1. Why can we stand up without collapsing?

2. Why do there exist metal/insulator/semi-conductor, magnetism, superconductivity/superfluidity?
3. What is the origin, past, future, and the fate of the earth, solar system, and even the
universe?

2 Questions for further thinking

1. For a microscopic system, should it always be in a state described by a wavefunction?

2. What is the superposition principle?
3. How to normalize eigen-wavefunctions for an operator with continuous eigen-spectra?
4. How to use wavefunctions to represent probability of measurement?
5. What are the constraints for wavefunctions of identical particles?
6. Why to use linear operator to represent observable?
7. What is the precise meaning of quantization?

3 Fundamental principles and notions of QM

1. Correspondence principle
2. Space-time symmetry
3. Statistical interpretation of wavefunctions
4. Superposition principle
5. Fundamental quantization condition
6. Schödinger equation for the pure state and quantum Liouville equation for the mixed
state
7. Bose and Fermi statistics

1
4 How to describe classic systems?

4.1 Classic particles

The state of a classical systems of N-particles can be represented by their coordinates and mo-
menta as (~r1 , ...,~rN ;~p1 , ..~pN ) which form a complete set of observables. Any other observable
F can be expressed as a function of coordinates and momenta as F(~r1 , ..,~rN ;~p1 , ...,~pN ). This
is called the “purely mechanical description”. If at an initial time t0 , (~r1 , ...,~rN ;~p1 , ..~pN ,t0 )
are known, their values at any later time t are also determined by the Hamilton’s canonical
equation

d ∂H d ∂H
qi = , pi = − , (i = 1, ..., n). (1)
dt ∂pi dt ∂qi
On the other hand, in many situations we are not able to determine all the values of coordi-
nates and momenta precisely, and thus have to use the “statistical description”. If coordinates
and momenta are repeatedly measured many times, a positive-definite probability distribution
function is obtained ρ(~r1 , ...,~rN ;~p1 , ...,~pN ). The value of the observable F is expressed as
R
d~r1 ...d~rN d~p1 ..d~pN F(~r1 , ..,~rN ;~p1 , ...,~pN )ρ(~r1 , ...,~rN ;~p1 , ...,~pN )
F̄ = R . (2)
d~r1 ...d~rN d~p1 ..d~pN ρ(~r1 , ...,~rN ;~p1 , ...,~pN )
If the system is time-dependent, the distribution function is explicitly dependent on time as
ρ(~r1 , ...,~rN ;~p1 , ...,~pN ;t), and its time evolution follows the Liouville’s equation

∂ρ N n dqi ∂ρ d pi ∂ρ o
+∑ + = 0. (3)
∂t i=1 dt ∂qi dt ∂pi

4.2 Classic waves

Consider an EM wave propagating along the z-direction. Its electric fields can be written
as Ex (r,t) = Ex0 cos(kz − ωt + αx ), Ey (r,t) = Ey0 cos(kz − ωt + αy ), ~B = ẑ × ~E. In principle,
classic physics does not need complex number. It is more convenient to represent as

Ex (r,t) = Ex ei(kz−ωt) , Ey (r,t) = Ey ei(kz−ωt) , (4)

where Ex = Ex0 eiαx and Ey = Ey0 eiαy . The polarization of the EM wave is defined as i) If
Ey(x) = 0, then polarization of the EM wave is along x(y)-direction. ii) If Ey = ±iEx , then
the EM wave is right(left) circularly polarized. Any general state of this EM wave can be
decomposed as linear superposition of either the linear polarized bases i), or the circularly
polarized basis.

2
 The energy of the EM wave satisfies the following relation because

|Ex |2 + |Ey |2
Etotal = V = N h̄ω, (5)
8π
where N is the number of photon. We define the state vector |Ψi = (ψx , ψy )T with ψx(y) =
q
V 2 2
8πN h̄ω Ex(y) , which satisfies the normalization condition |ψx | + |ψy | = 1. We can define a
set of basis as
   
1 0
|xi = , |yi = , (6)
0 1

which are for polarizations along x and y-directions respectively. Or, we can define We can
define the set of basis for circularly polarizations
   
1 1 1 0
|Ri = √ , |Li = √ . (7)
2 i 2 −i

1. Definition of a row vector hΨ| = (ψ∗x , ψ∗y ).

2. The scalar product between |Ψi and |Φi = (φx , φy ) is defined as hΨ|Φi = ψ∗x φx +
ψ∗y φy = hΦ|Ψi∗ .

3. Normalization condition of |ψx |2 + |ψy |2 = 1 can be represented as hΨ|Ψi = 1.

4. Any state can be decomposed into linear superpositions of a set of complete bases
ψx − iψy ψx + iψy
|Ψi = ψx |xi + ψy |yi = √ |Ri + √ |Li. (8)
2 2
In other words, we have
n o
|Ψi = |xihx|Ψi + |yihy|Ψi = |xihx| + |yihy| |Ψi
n o
= |RihR|Ψi + |LihL|Ψi = |RihR| + |LihL| |Ψi. (9)

The function of a polaroid is a projection operator. For example, the x-polaroid and y-
polaroid behave as

Px = |xihx|; Py = |yihy|. (10)

And the polaroids of R and L-circularly polarization behave as PR = |RihR| and PL = |LihL|.
For a general projection, which projects out a state |Φi, its operator is represented as Pφ =
|ΦihΦ|. For a general state |Ψi, after it passes the polaroid Pφ , it becomes Pφ |Ψi = hΦ|Ψi|Φi,
where hΦ|Ψi is the amplitude passing the polaroid. The EM wave intensity after the projec-
tion relative to before is |Pφ |Ψi|2 = |hΦ|Ψi|2 , or, the probability is |hΦ|Ψi|2 .

3
 Let us perform the rotation around the z-axis at θ which is denoted as R(θ). Under this
rotation, the bases of |xi and |yi change to |x´ i = R(θ)|xi and |y´ i = R(θ)|yi. The rotation
operation R(θ) can be represented as an unitary matrix as

(|x´ i, |y´ i) = (|xi, |yi)R(θ) (11)

where the transformation matrix

Ri j (θ) = hi|R(θ)| ji = hi| j´ i, (12)

where i, j equals to x, y, and j´ = x´ , y´ . More explicitly

hx|x´ i hx|y´ i
   
cos θ − sin θ
R(θ) = = . (13)
hy|x´ i hy|y´ i sin θ cos θ

Here R is an orthogonal matrix, generally speaking, it can be a unitary matrix, which can be
exponentialized as
 
− h̄i Sθ 0 −i
R(θ) = e , with S = h̄ . (14)
i 0

For a general state |Ψi, its projections on the two sets of bases satisfy
   
hx|Ψi ´ ´ −1 hx|Ψi
|Ψi = (|xi, |yi) = (|x i, |y i)R (θ) , (15)
hy|Ψi hy|Ψi

and thus
hx´ |Ψi hx´ |xi hx´ |yi
      
−1 hx|Ψi hx|Ψi
= R (θ) = . (16)
hy´ |Ψi hy|Ψi hy´ |xi hy´ |yi hy|Ψi

Define a state |Ψ´ i whose coordinates |x´ i and |y´ i bases are the same as those of |Ψi in the
bases of |xi and |yi, and thus

|Ψ´ i = |x´ ihx|Ψi + |y´ ihy|Ψi, (17)

which shows that

hx|Ψ´ i
  ´
hx |xi hx´ |yi
    
hx|Ψi hx|Ψi
= = R(θ) . (18)
hy|Ψ´ i hy´ |xi hy´ |yi hy|Ψi hy|Ψi

In other words, |Ψ´ i is the consequence of applying the rotation around z-axis at θ on the
state of |Ψi.
The operator S defined in Eq. 19 is a Hermitian operator, which describe an observable:
the spin angular momentum of EM wave along the z-direction. Its eigenvalues are ±h̄ with

4
eigenvecotrs ΨR,L
 
1
SΨR(L) = ±h̄ΨR(L) , ΨR(L) = . (19)
±i

The right and left circularly polarized light carry spin (projection along the z-axis) ±h̄.

5 Description of quantum systems

The quantum mechanical systems are similar to the classic waves but with significant differ-
ence. For the moment, we only consider states represented by wavefunctions which obey the
super-position principle. Such a principle states:

1) For a given system at a given time, the set of all the possible wavefunctions form a
linear space on the complex-number field.

2) Consider a state represented by the wavefunction aΨA + bΨB (a, b are complex con-
stants) which is a superposition of wavefunctions ΨA and ΨB . The measurements of an
observable F in the state aΨA + bΨB only yield values from those from measurements in the
states ΨA or ΨB .

3) If in the state Ψ, the measurement of F can yield the value of f , then Ψ can be decom-
posed as

Ψ = ΨA + ΨB . (20)

In the state represented by ΨA , the measurement of F only yields the value of f ; while such
measurements in the state represented by ΨB never yield the value of f . ΨA is called the
eigenstate of F with the eigenvalue f .

4) Under time evolution, the linear relation between wavefunctions does not change. For
example, if at t0 consider a state Ψ(t0 ) = aΨA (t0 ) + bΨB (t0 ), then at time t, we have Ψ(t) =
aΨA (t) + bΨB (t).
Examples
1) For a spinless particle, we use ψ~r´ (~r) to represent the state of particle localized at ~r´ .
In such a state, you can only find the Rparticle location at ~r´ , and later on we can show that
ψ~r´ (~r) = δ(~r −~r´ ). Then in the state r´ ≤R ψ~r´ (~r)ψ(~r´ )d 3~r´ , you can never find the particle
outside the ball BR (0). Inside this ball, you can find the particle at the neighborhood d~r1
around the location of~r1 . The corresponding component is ψ~r1 (~r)ψ(~r1 )d~r1 .
2) Suppose a spinless particle is located at~r0 at time t0 . This wavefunction is denoted as
ψ~r0 (~r) = δ(~r −~r0 ). At a later time t, the amplitude of this particle located at ~r is denoted as

5
K(t~r;t0~r0 ), which is called the propagator. Then for an arbitrary wavefunction ψ(~r,t0 ), how
does it evolve with time?
Solution:
Z
Ψ(~r,t0 ) = d~r0 ψ~r0 (~r)ψ(~r0 ,t0 ), (21)

in which ψr0 (~r) is the coordinate eigen-function, and ψ(~r0 ,t0 ) is the amplitude. By the su-
perposition law, at time t1 , we have
Z Z Z
Ψ(~r,t1 ) = d~r0 K(t1~r;t0~r0 )ψ(~r0 ,t0 ) = d~r1 ψ~r1 (~r) d~r0 K(t1~r1 ;t0~r0 )ψ(~r0 ,t0 ). (22)

At an even later time t2 , we have

Z Z
Ψ(~r,t2 ) = d~r1 K(t2~r;t1~r1 ) d~r0 K(t1~r1 ;t0~r0 )ψ(~r0 ,t0 ). (23)

Thus the propagator K satisfies the following property

Z
K(t2~r;t0~r0 ) = d~r1 K(t2~r;t1~r1 )K(t1~r1 ;t0~r0 ), (24)

for any t2 > t1 > t0 .

6 The statistical interpretation of wavefunctions

Different from classical waves, the QM wavefuncitons has the statistical interpretation. For
simplicity, we use the case of single particle for example. It can be straightforwardly gener-
alized to many-particle case.
1) A wavefunction Ψ(~r) is an amplitude. |Ψ(~r)|2 d~r represent the probability to find the
system configuration in the neighborhood d~r around~r.
2) In the state of Ψ, the probability of the results of measuring an observable F can be
represented through bilinear function of Ψ and Ψ∗ .
For two eigenstates Ψ1 and Ψ2 of F with two different eigenvalues f1 and f2 , we su-
perpose them for a new state Ψ12 = a1 ψ1 + a2 ψ2 in which we can only find either f1 or f2
for measuring F. First of all, Ψ12 cannot be an eigenstate of F. Otherwise, say, it has the
eigenvalue of f1 , then we can make Ψ12 − a1 ψ1 = a2 ψ2 . Since both Ψ and ψ1 possess the
eigenvalue f1 , ψ2 should be also an eigenstate of F with the same eigenvalue. This contra-
dicts to our initial assumption, thus the measurement in Ψ cannot give a definitive result of F.
The question is what is the relative probability of finding f1 and f2 denoted by W f1 (F, Ψ12 )
and W f2 (F, Ψ12 ), respectively. As assumed above, they should be expressed by bilinear of

6
Ψ12 and Ψ∗12 as

W f 1 (F, Ψ12 ) = A11 a∗1 a1 + A12 a∗1 a2 + A∗12 a∗2 a1 + A22 a∗2 a2 ,
W f 2 (F, Ψ12 ) = B11 a∗1 a1 + B12 a∗1 a2 + B∗12 a∗2 a1 + B22 a∗2 a2 . (25)

All the coefficients A’s and B’s are independent of a1,2 , but are determined by Ψ1,2 . If a1 =
0, we should have W f1 = 0 and thus A22 = 0. By a similar reasoning, B11 = 0. Because
probability is non-negative, the cross-terms a∗1 a2 and a∗2 a1 vanish which proves A12 = B12 = 0.
Also A11 and B11 should be positive numbers.
The sum of W f 1 and W f 2 is the total probability 1, which should be equal to the normalized
integral of |Ψ12 |2 .
Z Z
A11 a∗1 a1 + B22 a∗2 a2 = a∗1 a1 ψ∗1 (~r)ψ1 (~r)d~r + a∗2 a2 ψ∗2 (~r)ψ2 (~r)d~r
Z Z
+ a∗1 a2 ψ∗1 (~r)ψ2 (~r)d~r + a∗2 a1 ψ∗2 (~r)ψ1 (~r)d~r. (26)

Since a1,2 are arbitrary, we have

Z
W f 1 (F, Ψ12 ) = d~r [a1 Ψ1 (~r)]∗ [a1 Ψ1 (~r)],
Z
W f 2 (F, Ψ12 ) = d~r [a2 Ψ1 (~r)]∗ [a2 Ψ1 (~r)],
Z
0 = d~rΨ∗1 (~r)Ψ2 (~r). (27)

If we choose the normalization of Ψ1,2 as |Ψ1,2 |2 dq = 1, we have W f 1 = |a1 |2 and W f 2 =

R

|a2 |2 . We have also proved that two eigenstates of an observable F with different eigenvalues
are orthogonal.
More generally, according to the measurement of an arbitrary observable F on an arbitrary
wavefunction Ψ, Ψ can be expressed as the linear superposition of eigenstates F as
Z
Ψ = ∑ Ψ f0 + d f ´Ψ f ´ , (28)
f0

where f0 represents the discrete eigenvalues of F, and f ´ represents the continuous eigen-
values. This equation means the completeness of eigenstates of a physical observable. We
define the inner product
Z
(ΨA , ΨB ) = d~rΨ∗A (~r)ΨB (~r), (29)

then we have (Ψ f0 , Ψ) represents the probability of finding f0 in the measurement of F. If

f0 6= f0´ , we have (Ψ f0 , Ψ f ´ ) = 0. For the continuous spectra, ( ∆ f d f ´ Ψ´f , Ψ) represents the
R
0
probability of finding f ´ to f ´ + ∆ f in the measurement of F.

7
7 Expressing observables in terms of operators
Starting from Eq. 28, define an operator P( f0 , F)Ψ = Ψ f0 for discrete eigenvalues, and
P( f ´ , F)Ψd f ´ = ψ f ´ d f ´ for continuous eigenvalues. P( f0 , F) satisfies [P( f0 , F)]2 = P( f0 , F)
and so does P( f ´ , F), and thus they are projection operators. Define operator F̂ as
Z
F̂ = ∑ f0 P( f0 , F) + d f ´ f ´ P( f ´ , F), (30)
f0

then
Z
F̂Ψ = ∑ f0 Ψ f0 + f ´Ψ f ´ d f ´, (31)
f0

The expectation value of F can be expressed as

∑ f0 f0 (ψ f0 , ψ f0 ) + ∑∆ f f ´ ( ∆ f d f ´ Ψ f ´ , d f ´Ψ f ´ )
R R
∆f (Ψ, F̂Ψ)
F̄ = = . (32)
(Ψ, Ψ) (Ψ, Ψ)

8 Complete set of observables

A complete set of observables {F1 , F2 , ....} is defined as follows:
1) All the observables are compatible, i.e., they have common and independent eigenvalues.
2) A set of eigenvalues of this set of observables determine a pure state.
Again we will only use the single particle case for its simplicity. We denote x, y and z and
~r = (x, y, z) for the coordinate operators. The compatibility among x, y and z is assumed. Let
us consider x and following Eq. 31 to make the decomposition of the wavefunction Ψ(~r) as
Z
Ψ(~r) = dx´ Ψx´ (~r),
Z
x̂Ψ = dx´ x´ Ψx´ (~r). (33)

where Ψx´ (~r) is the eigenstate of r̂x . This shows that

Ψx´ (~r) = Ψ(~r)δ(x´ − x), (34)

and the expression of r̂ j,x in the coordinate representation as

x̂Ψ(~r) = xΨ(~r). (35)

8
Coordinates as a complete set of observables We next prove x, y, z are complete set of
observable. We define their common eigenstate as ψx´ ,y´ ,z´ (~r), which satisfies

r̂a ψx´ ,y´ ,z´ (~r) = ra´ ψx´ ,y´ ,z´ (~r) = ra ψx´ ,y´ ,z´ (~r) (36)

where ra = x, y and z, ra´ = x´ , y´ and z´ . Then we conclude that ψx´ ,y´ ,z´ (~r) ∝ δ(x − x´ )δ(y −
y´ )δ(z − z´ ). The set of eigenvalues are {x, y, z}. It is clear that ψx´ ,y´ ,z´ (~r) form a complete set
for wavefunctions Ψ(~r).
Thus we reach the remarkable conclusion that unlike the case in classical mechanics,
coordinates themselves form a complete set of observable without momenta.

The common eigenvalues of a complete set of observable span a complete basis for wave-
functions
1) Completeness: {Fα } with {α} = {α1 , α2 , .....} is a complete set of observable, and
{Ψα1 ,α2 ,.... } represent all the independent common eigenstates.
2) Orthogonal conditions: (Ψα , Ψα´ ) = 0 if {α} 6= {α´ }.
3) Normalization. For a set of discretized eigen-value of {α} = {n}, for an arbitrary wave-
function, the probability to find {n} is a finite value which is proportional to (Ψ{n} , Ψ{n} ),
thus it should be normalizable as (Ψ{n} , Ψ{n} ) = δ{n},{n´ } .
For a continuously R distributed eigenvalues, for simplicity, we consider a single param-
eter for eigenvalue ∆α dαΨα . The the probabilityR to observeR Fα ’s eigenvalue lying in the
interval α → α + ∆α should be finite such that ( ∆α dαΨα , ∆α dαΨα ) is finite. Because
R α1 , ψα2 ) 6=
(ψ R 0 only at α1 = α2 , (Ψα1 ,α2 ) has to be proportional to δ(α1 − α2 ) in order for
( ∆α dαΨα , ∆α dαΨα ) to have a finite value. Thus for eigenstates with continuous spectra,
we need to normalize them using δ-function as

(Ψα , Ψ´α ) = δ(α − α´ ). (37)

After all the eigenstates are normalized, we can write down the decomposition as
Z
Ψ = ∑ a{n} Ψ{n} + {dα}a{α} Ψ{α} , (38)
{n}

where {m} and {n} are the the discrete eigenvalues and {α} represents the continuous eigen-
values; a{n} = (Ψ{n} , Ψ), and a{m,α} = (Ψ{m,α} , Ψ).
4) Completeness condition: For simplicity, we consider the case of one particle in one
dimension. ψx´ (x) is the eigenstate of x with the eigenvalue x´ . Let us decompose it in terms
of the common eigenstates of a complete set of observable {F} as

Ψx´ (x) = ∑ an (x´ )Ψn (x) (39)

n

9
with
Z
an (x´ ) = (Ψn , Ψx´ (x)) = Ψ∗n (x)δ(x − x´ )dx = Ψ∗n (x´ ). (40)

For simplicity, we neglect the continous spectra, which can be added back easily. Thus we
have the completeness condition as

∑ Ψn(x)Ψ∗n(x´) = δ(x − x´). (41)

n

9 Generalization of the concept of wavefunction

Consider the decomposition of wavefunction in terms of the complete set of F{α} , and for
simplicity, we only consider the case that all eigenvalues are discrete.

Ψ= ∑ a{α}Ψ{α}, (42)
{α}

where an = (Ψn , Ψ). an is the wavefunction of state Ψ in the representation of α, an |an |2 is

the probability of finding the eigenstate Ψn in the state of Ψ.
Check the expression in the coordinate representation
Z
Ψ(~r) = d~rΨ(~r)Ψx´ ,y´ ,z´ (~r), (43)

in which Ψ(~r) is the wavefunction in the coordinate representation.

10