PHY202 (Spring 2024) Lecture Note No.

Kenji Nishiwaki *
Shiv Nadar Institution of Eminence, Department of Physics
Ver. 2 February 2024 (07h 48min in Indian ST)

Here, we will see the basic principles of quantum mechanics in finite-dimensional Hilbert space.

1 Probabilitistic Aspects of Quantum Mechanics

1.1 Probabilistic/Statistical Interpretation
As we discussed, in quantum mechanics, physical states are represented as (abstract) vectors in a Hilbert
space, where such vectors are sometimes called state vectors. This is one of the fundamental principles
of quantum mechanics. Another fundamental principle is about the probabilistic aspects of it:
babababababababababababababababababababababab

Born formula
For normalised state vectors with (as k|ψik = 1 and k|χik = 1), the probability to find the state
:::::::::::::::::::::::::::::
|χi in |ψi is given as
:::::::::

P (|ψi → |χi) = |hχ|ψi|2 , (1.1)

which is named as Born formula.a

a
If we do not put the normalisation on |ψi and |χi manifestly, the formula is modified as
2
|hχ|ψi|
P (|ψi → |χi) = . (1.2)
k|ψik2 · k|χik2

We remember that the inner product hχ|ψi between the two different state vectors is a complex number,
and we cannot interpret hχ|ψi itself as a probability. On the other hand, the squared absolute value of it
|hχ|ψi|2 can be interpreted as a probability. This is because |hχ|ψi|2 is a positive number or zero, and
owing to Cauchy-Schwarz inequality hψ|ψihχ|χi ≥ |hχ|ψi|2 , we can estimate the upper bound of it as

|hχ|ψi|2 ≤ k|ψik2 · k|χik2 = 1, (1.3)

where thus, we can conclude

0 ≤ |hχ|ψi|2 ≤ 1, (1.4)

and we can interpret |hχ|ψi|2 as a probability. Let us emphasise that state vectors should be normalised
for correct probabilistic interpretation. :::::
This :::::::::::
probability ::
is:::::::::::
recognised:::
as::::
the :::::::::
transition::::::::::::
probability :::::
from
the state |ψi to another state |χi instantaneously.
::::::::::::::::::::::::::::::::::::::::::::::

*
e-mail: kenji.nishiwaki@snu.edu.in.

1
 Next, we focus on a specific case that a physical state |ψi in H is represented as a linear combination
of a CONS {|χ1 i , |χ2 i , · · · } as
X X
|ψi = cr |χr i = hχr |ψi|χr i , (1.5)
r r

where |ψi should be normalised as k|ψik = 1. Note that the coefficients hχr |ψi are called wave functions
of the state |ψi under the CONS {|χ1 i , |χ2 i , · · · }.
Here, the probability to find the state |χs i in the state |ψi is calculated by the Born formula as

P (|ψi → |χs i) = |hχs |ψi|2 = |cs |2 , (1.6)

which is the probabilistic sense of the coefficients of the expansion. Due to the normalisation hψ|ψi = 1,
we get
X  X X
1 = hψ|ψi = hψ| hχr |ψi|χr i = |hχr |ψi|2 = P (|ψi → |χr i) , (1.7)
r r r

which expresses the property of probability that the sum of the probabilities of all possiblencases is unity. o
(r1 ) (r2 )
We know that we can construct a CONS from the eigenvectors of a Hermitian operator A, |vλ1 i , |vλ2 i , · · ·
b
and the probability of finding one of the degenerated eigenvectors belonging to λ is given as
kλ
! kλ D
X E X E2
(rλ ) (r )
P |ψi → vλ = vλ λ ψ , (1.8)
rλ =1 rλ =1

where this probability is interpreted as :::

the:::::::::::
probability:::
of::::::::::
observing::::
the::::::::::: λ::::::
eigenvalue:: when:::
we:::::::::
measure
the b in the state vector |ψi.
value of A
:::::::::::::::::::::::::::::::::::

• If |ψi is an eigenvector belonging to one of the eigenvalues λ, the possibility of finding one of the
degenerated eigenvectors belonging to λ is 1.
 E
b the probability P |ψi → Pkλ v (rλ ) represents the probability
• If |ψi is not an eigenvector of A, :::::::::::::: rλ =1 λ ::::::::::::::::::::::::

of observing the corresponding eigenvalue λ

:::::::::::::::::::::::::::::::::::::::::::::::::
in |ψi.

We focus on the following quantity for the normalised state vector |ψi (k|ψik = 1),
kλ D
XX E E
(r) (r)
hψ|A|ψi
b = hψ| A
b vλ ψ vλ
λ r=1

X Xkλ D E E
(r) (r)
= hψ| λ vλ ψ vλ
λ r=1
kλ
X X D E 2
(r)
= λ vλ ψ
λ r=1
" kλ
!#
X X E
(r )
= λ P |ψi → vλ λ , (1.9)
λ rλ =1
| {z }
Probability of observing λ in |ψi

2
where thus, the real number hψ|A|ψi
b is recognised as the expectation value of the Hermitian operator A
b
for the physical state |ψi. We introduce the notation,
h i
E|ψi A b = hAib |ψi := hψ|A|ψi
b , (1.10)

where the state |ψi is normalised, and we depicted the state for taking an average explicitly. When the
information of the state |ψi for taking an average is clear, frequently, this information is omitted.
Further, we will investigate the distribution of the operator,
∆A b − hAi
b := A b I,
b (1.11)
which measures the deviation from the expectation value of A.
b From
 2  2
∆Ab = A b − 2hAi bA b 2 I,
b + hAi b (1.12)
and thus
 2    
2
b − 2hAi b + hAi2
∆A
b = A bA b Ib
|ψi |ψi
  
2 D E2
= A
b − A
b
|ψi |ψi
    D
2 E2
= ψ A
b ψ − ψ A
b ψ , (1.13)

which is called the variance, and it is the simplest information on the deviation from the expectation
value. Variance is the amount squared and then averaged, so the actual size of the derivation is its square
root as,
s
   2 
σ|ψi A b := ∆Ab , (1.14)
|ψi

 is called the standard deviation. If no derivation from the expectation value of A exists for |ψi,
which b
σ|ψi A
b becomes zero.

• In Cn , the transition probability, the average, and the standard deviations are calculated based on
the standard techniques of linear algebra.
• In physics, physical quantities are defined such that their measurements are real numbers. In quantum
::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
theory, quantities that can actually be measured, i.e., physical quantities, are called observables.
::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
Observables are usually represented by Hermitian operators in H .
::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::

• If A
b and Bb are Hermitian operators, linear combination of A
b and Bb with real coefficients, k1 A+k
b 2Bb
(k1 , k2 ∈ R)nare also
o Hermitian operators, while
h i products AB and B A are not always Hermitian
the b b b b
operators. A, B (= AB + B A) and A, B (= A
b b b b b b b b bBb−B b A)
b are a Hermite operator and an
anti-Hermite operator, respectively.
Attension about normalisation of state vectors 

As we have discussed above, to derive probabilities and expectation values of Hermitian operators
appropriately, state vectors used for calculations should be normalised as unity. In the following, we
basically represent state vectors as normalised ones (as unity) unless otherwise noted.
 

3
1.2 Commutative physical quantities and Joint probability
As we introduced, for two operators in a Hilbert space H , A
b and B,
b the commutator is defined as
h i
A, B := A
b b bBb−B b A,
b (1.15)
h i
which measures the noncommutativity between A b and B.b If A, bB b =b 0, the two products A
bBb and BbAb
care called commutative and A bBb=B b A.b Note that all of the physical quantities in classical theory are
commutative.
We focus on two Hermitian operators, where eigenvectors of one of them construct a CONS individ-
ually (precisely speaking, after orthonormalisation of degenerated eigenvectors),
E E D E
b v (ra ) = a v (ra ) , (ra ) (ra0 )
A A=a A=a vA=a vA=a 0 = δaa0 δra ra0 (ra = 1, 2, · · · , ka ) , (1.16)
E E D E
b v (sb ) = b v (sb ) , (sb ) (sb0 )
B B=b B=b vB=b vA=b 0 = δbb0 δsb sb0 (sb = 1, 2, · · · , kb ) , (1.17)

where a and b represent one of the eigenvalues, ra and rb discriminate eigenvectors belonging to degener-
ated eigenvalues, and ka and kb denote the numbers of degenerated eigenvectors for a and b, respectively.
Sometimes, simultaneous eigenvectors are found between operators, where in the case of A b and B,
b the
concrete properties are as follows,
   
(ta,b ) ( ta,b ) (ta,b ) ( ta,b )
b v
A A=a,B=b = a vA=a,B=b , Bb v
A=a,B=b = b vA=a,B=b , (1.18)

where we can take the set of them as a CONS,

 
(ta,b ) (ta0 ,b0 )
vA=a,B=b vA=a0 ,B=b0 = δaa0 δbb0 δta,b t0a0 ,b0 (ta,b = 1, 2, · · · , ka,b ) . (1.19)

From the following simple calculation,

    
( a,b ) (ta,b ) (ta,b ) ( a,b )
  t  t
B A vA=a,B=b = (ba) vA=a,B=b = (ab) vA=a,B=b = AB vA=a,B=b ,
b b b b (1.20)

we recognise that simultaneous eigenstates between A

b and Bb are well defined if A
b and B
b are commutative.
According to the Born formula in Eq. (1.1), the probability
 
ka,b  ka,b  2
X (ta,b ) X ( ta,b )
P |ψi → vA=a,B=b  = vA=a,B=b ψ , (1.21)
ta,b =1 ta,b =1

describes that of observing one of the simultaneous eigenvectors of the two operators A
b and B
b with the
focused eigenvalues a and b, which is rephrased as the probability of observing the eigenvalues A = a
and B = b simultaneously in |ψi. Since orthonormalised simultaneous eigenvectors construct a CONS,
we get
 
ka,b 
X X (ta,b )
P |ψi → vA=a,B=b  = 1. (1.22)
all possible pairs:{a,b} ta,b =1

4
 We will see a concrete example of the two operators in C3 ,
   
5 1 0 0 8 0
b = 1 5 0 ,
E Fb = 8 0 0 , (1.23)
0 0 6 0 0 8

where it is easy to check that they are commutative,

 
8 40 0
Eb Fb = 40 8 0  = FbE.
b (1.24)
0 0 48

Also, it is straightforward to derive their eigenvalues as

E = 6, 6, 4, F = 8, 8, −8, (1.25)

and their corresponding orthonormal eigenvectors (through the Gram-Schmidt method, if necessary) as
     
1 0 1
(1)
E 1   (2)
E 1  
vE=6 = √ 1 , vE=6 = 0 ,  |vE=4 i = √ −1 , (1.26)
2 0 1 2 0
     
1 0 1
(1)
E 1   (2)
E 1  
vF =8 = √ 1 , vF =8 = 0 ,
  |vF =−8 i = √ −1 . (1.27)
2 0 1 2 0

Note that the choice of the corresponding orthonormal vectors belonging to a degenerated eigenvalues is
not unique, e.g., we can take the other set instead,
   
1 1
(1)
E 1 (1)
E
(2)
E 1 (2)
E
veE=6 = √ 1 = veF =8 , veE=6 = √  1  = veF =8 . (1.28)
2 1 2 −2

Here, apparently, a concrete representation of the simultaneous eigenvectors is

     
1 0 1
(1)
E 1 (2)
E 1
vE=6,F =8 = √ 1 , vE=6,F =8 = 0 , |vE=4,F =−8 i = √ −1 . (1.29)
2 0 1 2 0

Hereafter, we focus on the state


1
1
|ψi = √ −1 (1.30)
3 1

5
and transition probabilities to various eigenstates.
 E D E 2
(1) (1)
P |ψi → vE=6 = vE=6 ψ = 0,

(2)
E D
(2)
E 2 1
P |ψi → vE=6 = vE=6 ψ = ,
3
2
P (|ψi → |vE=4 i) = |hvE=4 | ψi|2 = , (1.31)
3
 E D E2 1
(1) (1)
P |ψi → veE=6 = veE=6 ψ = ,
9
 E D E2 2
(2) (2)
P |ψi → veE=6 = veE=6 ψ = , (1.32)
9
 E D E2
(1) (1)
P |ψi → vF =8 = vF =8 ψ = 0,
 E D E2 1
(2) (2)
P |ψi → vF =8 = vF =8 ψ = ,
3
2 2
P (|ψi → |vF =−8 i) = |hvF =−8 | ψi| = , (1.33)
3
 E D E2
(1) (1)
P |ψi → vE=6,F =8 = vE=6,F =8 ψ = 0,
 E D E2 1
(2) (2)
P |ψi → vE=6,F =8 = vE=6,F =8 ψ = ,
3
2 2
P (|ψi → |vE=4,F =−8 i) = |hvE=4,F =−8 | ψi| = , (1.34)
3
which leads to
X2  E 1
(r)
P (E = 6|ψ) = P |ψi → vE=6 = , (1.35)
r=1
3
2
X  E
(r)
= P |ψi → veE=6 , (1.36)
r=1
2
P (E = 4|ψ) = P (|ψi → |vE=4 i) = , (1.37)
3
X2  E 1
(r)
P (F = 8|ψ) = P |ψi → vF =8 = , (1.38)
r=1
3
2
P (F = −8|ψ) = P (|ψi → |vF =−8 i) = , (1.39)
3
and also
X2  E 1
(r)
P (E = 6 and F = 8|ψ) = P |ψi → vE=6,F =8 = , (1.40)
r=1
3
2
P (E = 4 and F = −8|ψ) = P (|ψi → |vE=4,F =−8 i) = , (1.41)
3
P (E = 6 and F = −8|ψ) = 0, (1.42)
P (E = 4 and F = 8|ψ) = 0. (1.43)
From the above calculations, a number of things can be read.

6
 • As seen in Eqs. (1.31) and (1.32), the probability of going to one of the degenerate orthonormalised
eigenvectors depends on the choice of eigenvector. However, the probability of going to one of
the degenerate eigenvectors is independent of the choice of eigenvector. This also means that the
probabilities of observing E = 6 from |ψi and observing F = 8 from |ψi are also independent of
the choice of degenerate (orthonormalised) eigenvectors (for degenerated eigenvalues).

• As seen in Eqs. (1.31), (1.32) and (1.34), as a matter of course, if individual eigenvectors can
be taken as the simultaneous eigenvectors, the probabilities of finding one of them in |ψi is the
same with that of finding both of them in |ψi. This point is also observed in, e.g., P (E = 6|ψ),
P (F = 8|ψ) and P (E = 6 and F = 8|ψ).

• Since the simultaneous eigenvectors for the pairs {E = 6, F = −8} and {E = 4, F = 8} do not ex-
ist, and thus the corresponding probabilities P (E = 6 and F = −8|ψ) and P (E = 4 and F = 8|ψ)
are zero. Here, we should focus on the fact that the probabilities P (E = 6|ψ) and P (F = −8|ψ)
are nonzero, which suggests that the probability distributions of E and F are correlated.

1.3 Projective measurements and Wave function collapse

Let us now further our understanding of state vectors |ψi and their measurement from the state normal-
isation hψ|ψi = 1 and how to calculate the expectation value of a Hermitian operator A b in Eq. (1.9).
Please keep in mind that |ψi is considered as the object that abstractly represents the information on the
quantum state ψ.
At first, we focus on the normalisation of |ψi,
kλ D
XX ED E
(r ) (rλ )
1 = hψ|ψi = ψ vλ λ vλ ψ
λ rλ =1
XD E XD E
= ψ Π
bλ ψ = b2 ψ
ψ Π λ
λ λ
XD E X E 2
= Π
b λψ Π
b λψ = Π
b λψ , (1.44)
λ λ

with the projective operator toward the directions covered by the eigenvectors belonging to λ of a Hermitian
operator,
kλ
X ED
(r ) (r )
Π
b λ := vλ λ vλ λ , (1.45)
rλ =1

where we used the properties of Π

b λ,
 2    †  
Π
bλ b2 = Π
=Π b λ, Π
bλ b† = Π
=Π b λ. (1.46)
λ λ

7
 
E
The vector Πb λψ = Π b λ |ψi is the projected state of |ψi onto the subspace of the entire vector space
constructed by the basis vectors belonging to the eigenvalue λ,
kλ D
X E E
b λ |ψi = (r ) (r )
Π vλ λ ψ vλ λ . (1.47)
rλ =1

Also, we can see

kλ D
E 2 X E 2
(r )
Π
b λψ = vλ λ ψ
rλ =1
kλ
!
X E
(r )
= P |ψi → vλ λ . (1.48)
rλ =1
 
Remember that this projection operator onto the λ-space is independent of the choice of the CONS as
 
kλ ED kλ0 ED
(s ) (s )
X (r ) (r )
X X
Π
bλ = vλ λ vλ λ  vλ0 λ0 vλ0 λ0 
rλ =1 λ0 sλ0 =1
kλ0
kλ X X E D
(s )
X (r )
= vλ λ δλλ0 δrλ ,sλ0 vλ0 λ0
rλ =1 λ0 sλ0 =1
kλ0 ED
(s ) (s )
X
= vλ0 λ0 vλ0 λ0
sλ0 =1
kλ
X ED
(s ) (s )
= vλ λ vλ λ . (1.49)
sλ =1

Also, the following property is also very important for λ 6= λ,

e

kλ
!  ke 
λ  0
X (s )
ED
(s )
X (s )
0
(s )
Π
b λΠ
be =
λ vλ λ vλ λ  vλe λe vλe λe 
sλ =1 s0e =1
λ

kλ kλ E  0
(s0λe ) (s )
e
X X (s ) (s )
= vλ λ vλ λ vλe vλe λe
sλ =1 s0e =1
λ
| {z }
=0
= 0. (1.50)

We can make a similar representation for the expectation value,

ka0 D
ka X ED ED E
(ra0 ) (ra0 )
XX (ra ) (ra )
hAi
b |ψi = hψ|A|ψi
b = ψ vA=a vA=a A vA0 =a0 va0
b ψ
a,a0 ra =1 ra0 =1 | {z }
=a0 δaa0 δra ra0

8
 ka
XX D ED E
(ra ) (ra )
= a ψ vA=a vA=a ψ
a ra =1
X D E X D E X E 2
= a ψ Π
b A=a ψ = a Πb A=a ψ Π
b A=a ψ = a Π
b A=a ψ . (1.51)
a a a

After the preparation of the quantum state |ψi, if we obtain certain information on the eigenvalue
of a Hermitian operator A, b our knowledge of |ψi increases. As we discussed, in general, a state vector
abstractly represents our knowledge of the corresponding quantum state. Measurement in quantum theory
is an operation that yields some information about a state vector. Taking everything above together, it
can be concluded that the quantum state must change its form before and after the measurement
operation. In quantum theory, the ideal case of a measurement operation is formulated as follows.
Attension about discrimination between operators and their eigenvalues 

Since operators are discriminated against for numbers by the hat symbol, e.g., as O,
b we can simply
represent their eigenvalues with the same alphabet without the hat symbol, e.g., as O. On the other
hand, sometimes, we adopt the more clear distinction for an operator, e.g., A,
b as A = a, where a
specific value is represented by the corresponding small letter.
 
babababababababababababababababababababababab

Setting up the ideal measurement in quantum theory

Suppose that an ideal measurement of a physical quantity described by a Hermitian operator A b is
made on a state vector |ψi (this is the form immediately before the measurement), the measured
b a. In this case, the state vector |ψafter i immediately
value is one of the discrete eigenvalues of A,
after the measurement is given by

b A=a |ψi
Π
|ψafter i → |ψA=a i := , (1.52)
b A=a |ψi
Π

where the state vector is renormalised (as a unit vector) after the projection operation by Π
b A=a
on |ψi.

Note that the ideal measurement is frequently also called the projective measurement since it is constructed
by a projection operator.1

Here, we will revisit the example of the two operators in C3 discussed in the previous subsection,
   
5 1 0 0 8 0
b = 1 5 0 ,
E Fb = 8 0 0 ,
0 0 6 0 0 8
1
The situation discussed by projective operators is idealistic, where the physical quantity A
b is measured, the measured value
a is observed, and the physical state is fully formed into the eigenvector belonging to the eigenvalue a of A.
b In general, measured
values can be deviated from a, and the state after the measurement can be different from the corresponding eigenvector. Such
general situations can be treated by quantum measurement theory.

9
where projection operators are easily constructed by use of the eigenvectors in Eqs. (1.26) and (1.27) as,
    1 1 
1 0 0
(1) (2) 1   1

2
1
2
1
ΠE=6 = ΠE=6 + ΠE=6 = √
b b b 1 √ 1, 1, 0 + 0 0, 0, 1 = 2 2 0
  
2 0 2 1 0 0 1
(1) (2)
=Πb
F =8 + ΠF =8 = ΠF =8 ,
b b (1.53)
 1
+ 2 − 12 0
  
1
1 1
= √ −1 √ 1, −1, 0 = − 12 + 12 0 = Π


Π
b E=4 b F =−8 , (1.54)
2 0 2 0 0 0

where we can check the following consistency condition easily,

Π
b E=6 + Π
b E=4 = Ib = Π
b F =8 + Π
b F =−8 . (1.55)

For the quantum state,


1
1
|ψi = √ −1 ,
3 1

the normalised projected states after the corresponding projective measurements are calculated with ease,
 
0
ΠE=6 |ψi
b
|ψE=6 i := = 0 = |ψF =8 i ,
 (1.56)
ΠE=6 |ψi
b 1
 
1
ΠE=4 |ψi
b 1  
|ψE=4 i := =√ −1 = |ψF =−8 i , (1.57)
b E=4 |ψi
Π 2 0

The following (conditional) probabilities are calculated,

P(E = 6 → F = −8|ψ) = P(F = −8|ψE=6 ) = |hvF =−8 | ψE=6 i|2 = 0, (1.58)

2
X D E2
(r)
P(E = 4 → F = 8|ψ) = P(F = 8|ψE=4 ) = vF =8 ψE=4 = 0, (1.59)
r=1
2
X D E 2
(r)
P(F = −8 → E = 6|ψ) = P(E = 6|ψF =−8 ) = vE=6 ψF =−8 = 0, (1.60)
r=1

P(F = 8 → E = 4|ψ) = P(E = 4|ψF =8 ) = |hvE=4 | ψF =8 i|2 = 0, (1.61)

where owing to the formula for the joint probability,

P (Ai ∩ Bj ) = P (Bj ∩ Ai ) = P (Bj |Ai ) P (Ai ) , (1.62)

10
we derive the following joint probabilities,
 
 h i
P(E = 6 and F = −8|ψ) = P
 F = −8 E = 6|ψ  P(E = 6|ψ)

| {z }
=ψE=6
1
= 0 × = 0, (1.63)
3 
 h i
P(E = 4 and F = 8|ψ) = PF = 8 E = 4|ψ 

 P(E = 4|ψ)
| {z }
=ψE=4
2
=0× = 0, (1.64)
3
where these results are consistent with the previous calculation based on the simultaneous eigenvectors.
Also, we can see
2
X D E 2
(r)
P(E = 6 → F = 8|ψ) = P(F = 8|ψE=6 ) = vF =8 ψE=6 = 1, (1.65)
r=1

P(E = 4 → F = −8|ψ) = P(F = −8|ψE=4 ) = |hvF =−8 | ψE=4 i|2 = 1, (1.66)

2
X D E2
(r)
P(F = 8 → E = 6|ψ) = P(E = 6|ψF =8 ) = vE=6 ψF =8 = 1, (1.67)
r=1

P(F = −8 → E = 4|ψ) = P(E = 4|ψF =−8 ) = |hvE=4 | ψF =−8 i|2 = 1, (1.68)

which leads to the joint probabilities

 
 h i
P(E = 6 and F = 8) = P
 F = 8 E = 6|ψ  P(E = 6|ψ)

| {z }
=ψE=6
1 1
=1× = , (1.69)
3 3 
 h i
P(E = 4 and F = −8) = P
F = −8 E = 4|ψ  P(E = 4|ψ)

| {z }
=ψE=4
2 2
=1× = . (1.70)
3 3
The two joint probabilities are also consistent with the previous derivation.

1.4 Quantum effects of Noncommutative physical quantities

h i
For two operators A
b and Bb in H , if the commutator between them is not zero operator A, bB b 6=
0⇔A
b b 6= B
bB b A,
b they are called noncommutative. In general, no simultaneous eigenvector exists (or

11
no sufficient number of simultaneous eigenvectors exists enough for constructing a CONS) between the
Hermitian operators, which are noncommutative.

Here, we focus on the very important example in C2 . We introduce the following three Hermitian
operators,
     
• 0 1 • 0 −i • 1 0
X = σx =
b , Yb = σy = , Zb = σz = , (1.71)
1 0 i 0 0 −1

which are named qubit operators or spin operators (for spin-1/2 states, other than the magnitude ~/2) and
σx , σy and σz are called Pauli matrices. Using the concrete representations in Eq. (1.71), we can evaluate
the products of X,
b Yb and Zb as follows:

Xb Yb = −Yb X
b = iZ,
b (1.72)
Yb Zb = −ZbYb = iX,
b (1.73)
ZbXb = −X b Zb = iYb , (1.74)
X
bXb (= Xb 2 ) = Yb Yb (= Yb 2 ) = ZbZb (= Zb2 ) = I,
b (1.75)

where now Ib is represented by the two-by-two identity matrix. Thus, we get

h i h i h i
X,
b Yb = 2iZ, b Yb , Zb = 2iX,
b Z,
b X
b = 2iYb , (1.76)

which means that any two of X, b Yb and Zb are noncommutative.

It is easy to check that their eigenvalues are

X = ±1, Y = ±1, Z = ±1, (1.77)

and their normalised eigenvectors can be taken as

   
1 1 1 1
|vX=+1 i := |x+ i = √ , |vX=−1 i := |x− i = √ , (1.78)
2 1 2 −1
   
1 1 1 i
|vY =+1 i := |y+ i = √ , |vY =−1 i := |y− i = √ , (1.79)
2 i 2 1
   
1 0
|vZ=+1 i := |z+ i = , |vZ=−1 i := |z− i = . (1.80)
0 1

Remember that eigenvectors belonging to different eigenvalues of an operator are orthogonal, e.g.,
hx+ |x− i = 0. Also, keep in mind that the overall phase factor of each eigenvector cannot be determined
uniquely. For example, the general form of the eigenvectors of X b is

eiθx+ 1 eiθx− 1
   
general general
|vX=+1 i = √ , |vX=−1 i = √ , (1.81)
2 1 2 −1
D E D E D E
general general general general general general
where θx+ , θx− ∈ R, vX=+1 vX=+1 = 1 = vX=−1 vX=−1 , and vX=+1 vX=−1 = 0.
From the above forms, it is explicitly shown that there is no simultaneous eigenvector among X, b Yb
and Z.
b Thereby, if the value of X b is determined, values of Yb and Zb should not be uniquely fixed. More

12
concrete information is given by the transition probabilities, e.g.,
1
P (Z = +1 → X = +1) = |hx+ | z+ i|2 = , (1.82)
2
2 1
P (Z = +1 → X = −1) = |hx− | z+ i| = , (1.83)
2
and the joint probability,

P (Z = +1 and X = +1) = 0 (∴ the absence of the simultaneous eigenvector) . (1.84)

The projection operators are constructed as,

   
1 1 1 1 1 −1
Πx+ := |x+ ihx+ | =
b , Πx− := |x− ihx− | =
b , (1.85)
2 1 1 2 −1 1
   
b y+ := |y+ ihy+ | = 1 1 −i b y− := |y− ihy− | = 1 1 i
Π , Π , (1.86)
2 i 1 2 −i 1
   
b z+ := |z+ ihz+ | = 1 0 b z− := |z− ihz− | = 0 0
Π , Π , (1.87)
0 0 0 1

where we can get the relationship between the eigenvectors belonging to different Hermitian operators,
e.g.,

b x− |z+ i = √1 (|x+ i + |x− i) ,

 
|z+ i = Ib |z+ i = Πb x+ + Π
2
  1
b x− |z− i = √ (|x+ i − |x− i) ,
|z− i = Ib |z− i = Πb x+ + Π
2
b z− |x+ i = √1 (|z+ i + |z− i) ,
 
|x+ i = Ib |x+ i = Πb z+ + Π
2
  1
b z− |x+ i = √ (|z+ i + |z− i) .
|x− i = Ib |x− i = Πb z+ − Π (1.88)
2

Next, we focus on the general state in C2 ,

 
c
|ψi = 1 , (1.89)
c2

where the normalisation condition requires |c1 |2 + |c2 |2 = 1. The expectation values of the operators X,b
Yb and Zb for |ψi are calculated as
  
D E D E
0 1 c1
X
b = ψ X ∗
b ψ = c1 , c 2∗
= c∗1 c2 + c∗2 c1 = 2 Re (c∗1 c2 ) , (1.90)
|ψi 1 0 c2
  
D E D E
0 −i c1
Yb = ψ Yb ψ = c1 , c2∗ ∗
= −ic∗1 c2 + ic∗2 c1 = 2 Im (c∗1 c2 ) , (1.91)
|ψi i 0 c 2
  
D E D E
1 0 c1
Z
b = ψ Z ψ = c1 , c 2
b ∗ ∗
= c∗1 c1 − c∗2 c2 = |c1 |2 − |c2 |2 , (1.92)
|ψi 0 −1 c2

13
where they are real variables (as generally proved for Hermitian operators before). We focus on the
following combination,
D E 2 D E 2 D E 2
X
b + Yb + Zb = (c∗1 c2 + c∗2 c1 )2 + (−ic∗1 c2 + ic∗2 c1 )2 + (c∗1 c1 − c∗2 c2 )2
|ψi |ψi |ψi

= 2c∗1 c∗2 c1 c2 + (c∗1 c1 )2 + (c∗2 c2 )2

= (c∗1 c1 + c∗2 c2 )2 = 12 = 1. (1.93)
D E D E D E
• If any one of X
b , Yb and Zb is +1 or −1, the two others should be zero. A concrete
|ψi |ψi E
D |ψiD E D E
example of this situation is X b = 0, Yb = 0, and Zb = +1. Here, the zero can be
|z+ i |z+ i |z+ i
recognised as a 50%-50% mixture of +1 and −1 eigenvalues. Therefore, if one of the eigenvalues
of X,
b Yb and Zb is completely fixed, the other two become totally indeterminate. This represents a
nontrivial correlation between noncommutative physical quantities.

• If one expectation value is close to +1 or −1 (eigenvalues), then this physical state can be said
to be close to the corresponding eigenstate. In this case, this state can be understood to be nearly
indeterminate with respect to the other two eigenvalues.

1.5 Considering the sequential Stern-Gerlach (SG) experiments

We focus on the state describing the outmost electron’s spin of the Ag atom emitted from the furnace in
the sequential SG experiments,

e = √1 (|z+ i + |z− i) ,
|ψi (1.94)
2
where Z is randomised. In the following calculations, the relations summarised in Eq. (1.88) are useful.

• SG-b
z Case:
  D E 2 1
P Z = +1|ψe = z+ ψe = , (1.95)
2
  D E 2 1
P Z = −1|ψe = z− ψe = . (1.96)
2

• SG-b
z zb Case: If Z = +1 is observed in the first SG apparatus, the corresponding state after the
measurement should be the following normalised projected state,
E b Z=1 |ψi
Π e
ψeZ=1 := = |z+ i . (1.97)
b Z=1 |ψi
Π e

So, we get
  D E 2
P Z = +1|ψeZ=1 = z+ ψeZ=1 = 1, (1.98)
  D E 2
P Z = −1|ψeZ=1 = z− ψeZ=1 = 0. (1.99)

14
 • SG-b
zx
b Case: Using the projected state in Eq. (1.97), we can get
  D E 2 1
P X = +1|ψZ=1 = x+
e ψeZ=1 = , (1.100)
2
  D E 2 1
P X = −1|ψZ=1 = x−
e ψeZ=1 = . (1.101)
2

• SG-bzx
bzb Case: If X = +1 is observed in the second SG apparatus, the corresponding state after the
second measurement should be
E
E ΠX=1 ψZ=1
b e
ψeZ=1→X=1 := E = |x+ i . (1.102)
Π
b X=1 ψeZ=1

So, we get
  D E 2 1
P Z = +1|ψeZ=1→X=1 = x+ ψeZ=1→X=1 = , (1.103)
2
  D E 2 1
P Z = −1|ψeZ=1→X=1 = x− ψeZ=1→X=1 = . (1.104)
2

The above results clearly declare that quantum mechanics provides the correct probabilities for all of the
(sequential) SG experiments!

1.6 Interference effect in C2

We focus on the following linear combination of |z+ i and |z− i,
1
|ψ(α) i := √ eiα/2 |z+ i + e−iα/2 |z− i


2
iα/2
 iα/2 
e −iα

1 e
= √ |z+ i + e |z− i = √ −iα/2 , (1.105)
2 2 e

where α is a real parameter and the state is normalised as hψ(α) |ψ(α) i = 1. This is a generalisation of the
first state for the SG experiments, defined in Eq. (1.94). Owing to the straightforward calculation,

2 1
P Z = +1|ψ(α) = z+ ψ(α) = , (1.106)
2

2 1
P Z = −1|ψ(α) = z− ψ(α) = , (1.107)
2
the state in Eq. (1.105) is also considered as a 50%-50% mixture of |z+ i and |z− i. Also, the above result
mentions that no difference emerges in the SG-b z apparatus if we replace the first stage |ψi
e in Eq. (1.94)
into |ψ(α) i in Eq. (1.105). We continue similar calculations in the SG-b z zb and SG-bzxb apparatuses based
on the normalised projected state,

ψ(α;Z=1) |ψ(α) i
ψ(α;Z=1) := = eiα/2 |z+ i , (1.108)
ψ(α;Z=1) |ψ(α) i

15
which leads to

2
P Z = +1|ψ(α;Z=1) = z+ ψ(α;Z=1) = 1, (1.109)

2
P Z = −1|ψ(α;Z=1) = z− ψ(α;Z=1) = 0, (1.110)

2 1
P X = +1|ψ(α;Z=1) = z+ ψ(α;Z=1) = , (1.111)
2

2 1
P X = −1|ψ(α;Z=1) = z− ψ(α;Z=1) = , (1.112)
2

where all of the results are equivalent
to those based on |ψi.
e From above, we can recognise the important
iα/2
point: the overall phase factor e does not contribute to the probabilities (and also to expectation
:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::

values)
:::::::
since this part is factorised and eiα/2 = 1. Also, we can get the corresponding results for picking
up Z = −1 at the first step,

2
P Z = +1|ψ(α;Z=−1) = z+ ψ(α;Z=−1) = 0, (1.113)

2
P Z = −1|ψ(α;Z=−1) = z− ψ(α;Z=−1) = 1, (1.114)

2 1
P X = +1|ψ(α;Z=−1) = z+ ψ(α;Z=−1) = , (1.115)
2

2 1
P X = −1|ψ(α;Z=−1) = z− ψ(α;Z=−1) = . (1.116)
2
 
Here, we evaluate the combined probability for the situation that first we observe Z = +1 or Z = −1
in ψ(α) at the first apparatus, and we observe X = +1 after that:


P [Z = +1 or Z = −1] → X = +1 ψ(α)





= P X = +1 ψ(α);Z=+1 · P Z = +1 ψ(α) + P X = +1 ψ(α);Z=−1 · P Z = −1 ψ(α)
1 1 1 1 1
= · + · = . (1.117)
2 2 2 2 2
 
Also, we focus on the probability of finding the X = +1 result in the generalised first state ψ(α;Z=1) .
The calculation itself is straightforward:

2
P X = +1|ψ(α) = x+ ψ(α)
1 iα/2
2
= e + e−iα/2
2
1 iα/2
∗ iα/2
+ e−iα/2 + e−iα/2


= e e
4
1
1
= 1 + 1 + eiα + e−iα = (1 + cos α) , (1.118)
4 2



where P [Z = +1 or Z = −1] → X = +1 ψ(α) is not matched with P X = +1|ψ(α) in general (except
for α = ±π/2). The difference between the two processes is with or without identifying the intermediate
state before the X measurement.

16
 The origin of this difference can be clarified if we rewrite the two probabilities in the following matter,

2 2
P [Z = +1 or Z = −1] → X = +1 ψ(α) = hx+ | z+ i z+ ψ(α) + hx+ | z− i z− ψ(α) ,
(1.119)

2
P X = +1|ψ(α) = x+ ψ(α)
2
= hx+ | z+ i z+ ψ(α) + hx+ | z− i z− ψ(α)
2 2
= hx+ | z+ i z+ ψ(α) + hx+ | z− i z− ψ(α)


∗ 
+ 2 Re hx+ | z+ i z+ ψ(α) hx+ | z− i z− ψ(α) ,
(1.120)

where we used the completeness

relation |z+ ihz+ | + |z− ihz− | = I.
b The emergence of the interference
:::::::::::::::::::::::::::::::::
term in P X = +1|ψ (α) is the origin of the difference.
:::::::::::::::::::::::::::::::::::::::::::::::::::::

 
1.7 A and B 6= A + B in eigenvalue
b b b b

Aswe discussed, the addition of the operators A b in a Hilbert space H is defined and represented
b and B
  
as Ab+B b . We will zoom in on a ‘strange’ aspect of the eigenvalues of A, b Bb and Ab+B b .
When we take A b and Bb as Xb and Yb in C2 , the addition of them is defined as
 
  0 1−i
L
b := A
b+B
b = , (1.121)
1+i 0

which is a Hermitian operator. Thinking naively, candidates for the eigenvalues of L

b seem to be
   
b + an eigenvalue of Yb → −2, 0, or 2.
an eigenvalue of X (1.122)
√
Nevertheless, the actual eigenvalues of L b are ± 2, which is none of them. This seemingly strange
phenomenon occurs when no simultaneous eigenvector exists (or no sufficient number of simultaneous
eigenvectors exists enough for constructing a CONS) between the Hermitian operators.
 

Wehcan rephrase
i this statement as follows: iftwo Hermitian
 operators A
b and B
b are non-commutative
as A,
bB b 6= b0, in general, an eigenvalue of A b+B b cannot be represented as a simple addition of
eigenvalues of A
b and B.b
 

1.8 Uncertainty relation of Robertson

For two Hermitian operators A,
b Bb and an arbitrary normalised state in a Hilbert space H , the following
inequality holds,
    1 Dh iE
b · σ|ψi B
σ|ψi A b ≥ A,
bB b , (1.123)
2 |ψi

which is named the uncertainty relation of Robertson (or Robertson’s inequality).

17
 We will prove the relation in Eq. (1.123). First, by use of the Hermitian operators measuring the
deviations for A
b and B
b as defined in Eq. (1.11),

∆A b − hAi
b=A b I,
b ∆B b − hBi
b=B b I,
b (1.124)

we construct two states |φi and |χi from |ψi as follows:

|φi := ∆A
b |ψi , |χi := ∆B
b |ψi . (1.125)

When we substitute the above |φi and |χi for the Cauchy-Schwarz inequality,

hφ|φihχ|χi ≥ |hφ|χi|2 , (1.126)

we get
  2    2  D E2
ψ ∆A b ψ · ψ ∆B b ψ ≥ ψ ∆A ∆B ψ
b b . (1.127)
| {z } | {z }
b))2
=(σ|ψi(A b ))2
=(σ|ψi(B

Here, we can focus on the rephrasing in terms of the anti-commutator and the commutator between ∆A b
and ∆B,
b

b = 1 ∆A, b + 1 ∆A,
n o h i
∆A b ∆B b ∆B b ∆B b . (1.128)
2 2
Since, as we discussed and it is easy to check, the two terms are anti-Hermite and Hermite, where their
eigenvalues are purely imaginary and real, respectively. Thus, we get
D E2 1 D n o E2 1 D h i E2
ψ ∆A b ∆Bb ψ = ψ ∆A, b ∆B b ψ + ψ ∆A, b ∆Bb ψ , (1.129)
4 4
where we combine this result and Eq. (1.128) and we reach
  2   2 1 D n o E2 1 D h i E 2
σ|ψi A
b · σ|ψi B b ≥ ψ ∆A, b ∆B b ψ + ψ ∆A,
b ∆B
b ψ
4 4
1 D h b bi E 2
≥ ψ ∆A, ∆B ψ . (1.130)
4
Finally, the identity (it is easy to show it),
h i h i
∆A,
b ∆B
b = A,bB
b , (1.131)

brings us to the form,

  2   2 1 D h i E 2
σ|ψi A
b · σ|ψi B
b ≥ ψ A,
bBb ψ (≥ 0) . (1.132)
4
If we take the square root of the above inequality, we reach Eq. (1.123) and Robertson’s inequality is
proved. We can read the following properties:
Dh iE      
• If A,
bBb 6= 0, there are three possibilities: (i) σ|ψi A
b 6= 0 and σ|ψi B b 6= 0, (ii)
|ψi
           
σ|ψi A
b = 0 and σ|ψi B b = ∞, (iii) σ|ψi A b = ∞ and σ|ψi B b = 0.

18
 h i
• If the condition is realised (remember that A,bB
b is anti-Hermitian),
h i
A,
bB b = ik Ib (k ∈ R) , (1.133)
the Robertson’s inequality in Eq. (1.123) takes the simplified form,
      k 2
σ|ψi Ab · σ|ψi B b ≥ , (1.134)
4
where the lower bound does not depend on |ψi.

For completeness, we derive the condition for the situation that Robertson’s inequality is saturated,
where the equality in Cauchy-Schwarz inequality and the condition
D n o E
ψ ∆A, b ∆B b ψ =0 (1.135)
should be simultaneously realised. We remember the equality condition for the Cauchy-Schwarz inequal-
ity, where one of the following conditions holds
|φi = |nulli , |χi = |nulli , |χi = t |φi (t ∈ C) . (1.136)
  2 D h i E
• If |φi = |nulli, hφ|φi = σ|ψi A b = hnull|nulli = 0. Here, ψ A, bB b ψ becomes zero
and the second condition in Eq. (1.135) is realised, and Robertson’s inequality is saturated.
  2 D h i E
• If |χi = |nulli, hχ|χi = σ|ψi B b = hnull|nulli = 0. Here, ψ A, bB b ψ becomes zero
and the second condition in Eq. (1.135) is realised, and Robertson’s inequality is saturated.
• If |χi = t |φi (t ∈ C), this is rephrased as
∆B b |ψi = t · ∆Ab |ψi
   
⇔ b − tA
B b |ψi = hBi b − thAib |ψi . (1.137)
The second condition is represented as
D n o E
0 = ψ ∆A, b ∆B b ψ
D E D E
= ψ ∆A b ∆B b ψ + ψ ∆B b ∆A
b ψ
D E D E
∗
= t · ψ ∆A ∆A ψ + t · ψ ∆A ∆A ψ
b b b b
  2 
∗
= (t + t ) ψ ∆A b ψ , (1.138)
| {z }
b))2
=(σ|ψi(A
 
where we used the condition ∆B b |ψi = t·∆A b |ψi. If σ|ψi Ab = 0, the second condition is realised.
 
Similarly, if σ|ψi B
b = 0, the second condition is also realised. For these cases, it is obvious that
 
the second condition for saturating Robertson’s inequality holds. We focus on σ|ψi A b 6= 0 and
 
σ|ψi Bb 6= 0, where the necessary condition is

t + t∗ = 0 ⇒ [t is pure imaginary as t = iκ (κ ∈ R)] . (1.139)

19
    
To summarize, (i) if σ|ψi A
b = 0 or σ|ψi B b = 0, Robertson’s inequality is saturated; (ii) if
   
σ|ψi Ab 6= 0 and σ|ψi Bb 6= 0, saturating Robertson’s inequality requires

   
b − iκA
B b |ψi = hBi
b − iκhAi
b |ψi , (1.140)
 
which means that the state |ψi should be an eigenstate of the non-Hermitian operator B b .2 Such
b − iκA
a state is called a minimal uncertainty state or a generalised coherent state.3

2 How to describe Time evolutions

2.1 Time evolution operator and Born formula
As mentioned, the Born formula shown in Eq. (1.1) can be understood as the transition probability from
the state |ψi to |χi, where the transition is completed instantaneously. This point can be rephrased as |ψi
and |χi applied for the Born formula should be (part of) the information on a focused physical system at a
certain time. However, on the other hand, physics often discusses the time evolution of physical systems.
Since even quantum systems are considered to have time evolution, we shall discuss below how the time
evolution of state vectors can be handled in quantum systems.
 
2 b − iκA b |ψi = hBi
b |ψi) and the imaginary part (A b |ψi = hAib |ψi) is
If the decomposition of B b into the real part (B
possible, an appropriate |ψi is a simultaneous eigenvector between A b and B.
b Nevertheless, if A b and Bb are noncommutative,
such decomposition is not justified.
3
The condition shown in Eq. (1.140) is the necessary condition for a consistent realisation of a generalised coherent state
for Ab and B,b but NOT a sufficient condition. If we cannot realise Eq. (1.140) no matter how we choose κ, no generalised
coherent state exists for the pair of A
b and B.
b

20
 
In quantum theory, we consider time evolution is also described by an operator. We introduce the
operator U
b (t) that advances the time of the operated state vector by t, which means that if the ψ-state
at t = 0 is introduced as |ψ(0)i (= |ψ(t = 0)i), the corresponding state at the time t should be
U
b (t)|ψ(0)i;

U(t)
b
|ψ(0)i −−→ |ψ(t)i = U
b (t)|ψ(0)i . (2.1)

Based on this operation, we get new expressions for the Born formula.

• A state is prepared at t = 0 as |ψ(0)i. After a time t has elapsed, the probability of finding the
state |χi in |ψ(t)i is given as
   
t U(t)
b momentary
P |ψ(0)i → − |χi := P |ψ(0)i −−→ |ψ(t)i −−−−−→ |χi

= |hχ|ψ(t)i|2
D E 2 D E2
= χ U (t) ψ(0)
b = χ U (t) ψ(0)
b . (2.2)

• The expectation value of a Hermitian operator Ab for the state |ψ(t)i at time t is calculated as
follows:
D E D E
A
b = ψ(t) A b ψ(t)
|ψ(t)i
D E
= Ub (t)ψ(0) A b Ub (t)ψ(0)
D E
† bb
= ψ(0) U (t) A U (t) ψ(0) .
b (2.3)
 

2.2 Schrödinger equation

In quantum mechanics/theory, the following principle dictates the time evolution of a state vector.
babababababababababababababababababababababab

Schrödinger equation
The following equation, named Schrödinger equation, dictates the time evolution of a physical
state |ψ(t)i:

d
i~ |ψ(t)i = H
b |ψ(t)i , (2.4)
dt

where the operator Hb called Hamiltonian describes the details of time evolutions of a focused
system. In principle, H
b is a Hermitian operator.

The following points are important:

• Schrödinger equation is a linear differential equation. Therefore, under the given initial condition

21
 at time t = 0 |ψ(0)i, the configuration at time t is uniquely determined.

• Taking the initial time at t = 0 is merely a widely adopted convention. An initial time can be set
anywhere, e.g., at t = τ , where |ψ(t + τ )i = U
b (t)|ψ(τ )i.

• The physical dimension of a Hamiltonian operator is equal to that of energy (remember that
[~] = [Energy × Time]]).

• There is no absolute principle that provides a Hamiltonian describing a given system. The form of
a Hamiltonian is determined after careful consideration of the properties of the physical system of
interest.
Hereafter, we adopt the convention,
d d hχ(t)| d |ψ(t)i
hχ(t) | ψ(t)i = |ψ(t)i + hχ(t)|
dt  dt   dt 
dχ(t) dψ(t)
= ψ(t) + χ(t) , (2.5)
dt dt

where the first step is a simple application of the Leibniz rule for derivatives. If the two states |ψ(t)i and
|χ(t)i follows the Schrödinger equation as,
d d
i~ |ψ(t)i = H
b |ψ(t)i , i~ |χ(t)i = H
b |χ(t)i , (2.6)
dt dt
we can follow
   
d dχ(t) dψ(t)
hχ(t) | ψ(t)i = ψ(t) + χ(t)
dt dt dt
   
i b i b
= − Hχ(t) ψ(t) + χ(t) − Hψ(t)
~ ~
i b
D E i D E
= Hχ(t) ψ(t) − χ(t) Hψ(t)
b
~ ~
i D
†
E i D E
= χ(t) H ψ(t) −
b χ(t) Hψ(t)
b
~ ~
i D E i D E
= χ(t) Hψ(t) −
b χ(t) Hψ(t) = 0,
b (2.7)
~ ~
where we used the Hermitian condition for the Hamiltonian (H b † = H)
b at the last step. This result is
understood that the inner product hχ(t) | ψ(t)i does not depend on time (as long as H
b is Hermitian), where
this is called the conservation of the inner product or norm.
Next, we formally solve the Schrödiner equation for the state

|ψ(t)i = U
b (t)|ψ(0)i , (2.8)
d b   
i~ U (t)|ψ(0)i = H b U b (t)|ψ(0)i , (2.9)
dt
where |ψ(0)i is an arbitrary state, and it is time-independent. Therefore, the above form is equivalent to
d b
i~ U (t) = H
bUb (t) , (2.10)
dt

22
where Ub (0) should equal Ib for consistency. It is easy to check that the solution is formally described by
the exponential operator,
∞  n
−iHt/~
X 1 i b
U (t) = e := − Ht , (2.11)
b b

n=0
n! ~
owing to
   
d −iHt/~ i b −iHt/~ −iHt/~ i b
e = − Ht e =e − Ht . (2.12)
b b b
dt ~ ~
Also, via the relation,
b (t)† U
hψ(t) | ψ(t)i = hψ(0)| U b (t) |ψ(0)i = hψ(0) | ψ(0)i , (2.13)
we recognise that the relationship should hold,
b (t)† U
U b (t) = I,
b (2.14)
b (t) is a unitary operator.4 We can easily check the
which means that the time evolution operator U
following properties of Ub (t),
   −1
Ub (t) U
b (s) = U b (t + s) , U
b (t) = U b (t) =Ub (−t) . (2.16)

2.3 Energy eigenstate

Eigenvalues of a Hamiltonian are interpreted as possible energies (of a focused state). The eigenvalue
problem for H
b is formulated as
b |φn i = En |φn i ,
H (2.17)
where |φn i should be nonzero vectors. En and |φn i are called energy eigenvalues and energy eigenstates.
The eigenequation for H b is sometimes called the time-independent Schrödinger equation. Note that the
number of energy eigenvalues for a state is finite in a finite-dimensional Hilbert space, while it can be
infinite in an infinite-dimensional Hilbert space (as we will see later).
The following form from En and |φn i,
|ψ(t)i = e−iEn t/~ |φn i , (2.18)
is a solution of the Schrödinger equation as
d
i~ |ψ(t)i = En |ψ(t)i = H
b |ψ(t)i . (2.19)
dt
For this |ψ(t)i, the absolute value of the inner product between e−iEn t/~ |φn i and an arbitrary state |χi is
time-independent since
χ e−iEn t/~ φn = e−iEn t/~ |hχ | φn i| = |hχ | φn i| , (2.20)
which suggests that in energy eigenstates, transition probabilities are not time-varying.
4
We check the unitary relation by a straightforward calculation,
 †
b†
e−iHt/~ e−iHt/~ = e+iH t/~ e−iHt/~ = e+iHt/~ e−iHt/~ = e+iHt/~−iHt/~ = eO = I. (2.15)
b b b b b b b b b

23
2.4 Time evolution of Two-body state
We consider a simple system in C2 , where we take |χ1 i and |χ2 i as a CONS. Arbitrary state vector in C2
is described as a linear combination as

|ψi = c1 |χ1 i + c2 |χ2 i , (2.21)

where the normalisation condition requires |c1 |2 + |c2 |2 = 1.

Here, we take an initial state as

|ψ(0)i = c1 (0) |χ1 i + c2 (0) |χ2 i , (2.22)

with

c1 (0) = hχ1 | ψ(0)i , c2 (0) = hχ2 | ψ(0)i . (2.23)

Thus, if we measure the state at the time t = 0, we observe the states |χ1 i or |χ2 i with the probabilities
of |c1 |2 or |c2 |2 . Since H b |χ2 i are transformed vectors in C2 , we can parametrise them as
b |χ1 i and H

b |χ1 i := h11 |χ1 i + h12 |χ2 i ,

H b |χ2 i := h21 |χ1 i + h22 |χ2 i ,
H (2.24)

and thus, we get

b |ψi = c1 (h11 |χ1 i + h12 |χ2 i) + c2 (h21 |χ1 i + h22 |χ2 i)
H
= (c1 h11 + c2 h12 ) |χ1 i + (c2 h22 + c1 h12 ) |χ2 i . (2.25)

If we introduce the column-vector representation,

   
• 1 0
•
|χ1 i = , |χ2 i = , (2.26)
0 1
we can rewrite the above form as
    
b |ψi = c
• 1 h11 + c 2 h12 h11 h12 c1
H = , (2.27)
c2 h22 + c1 h12 h21 h22 c2
| {z }
•
=H

as we generally discussed before, the linear operator H

b is represented as a matrix H. We should be careful
†
that the Hermitian condition (H
b = H) b put the conditions on the coefficients,

h∗11 = h11 , h∗22 = h22 , h∗12 = h21 , (2.28)

where they may describe the magnitudes of the transitions of |χ1 i → |χ1 i, |χ2 i → |χ2 i, and |χ1 i ↔ |χ2 i,
respectively.
Hereafter, we focus on the simplified case parametrised by two parameters,

ε := h11 = h22 , α := h12 = h∗21 , (2.29)

where h21 = α. The Schrödinger equation reads

    
d c1 (t) ε α c1 (t)
i~ = , (2.30)
dt c2 (t) α ε c2 (t)

24
where we used the parametrisation,

|ψ(t)i = c1 (t) |χ1 i + c2 (t) |χ2 i . (2.31)

The above matrix equation is equivalent to the simultaneous equations,

dc1 dc2
i~ = εc1 + αc2 , i~ = αc1 + εc2 , (2.32)
dt dt
where we get the equivalent form,
d d
i~ (c1 + c2 ) = (ε + α) (c1 + c2 ) , i~ (c1 − c2 ) = (ε − α) (c1 − c2 ) . (2.33)
dt dt
The above forms are solved easily as

c1 (t) + c2 (t) = e−i(ε+α)t/~ (c1 (0) + c2 (0)) , c1 (t) − c2 (t) = e−i(ε−α)t/~ (c1 (0) − c2 (0)) , (2.34)

and thus, we reach

1 h −i(ε+α)t/~ i
c1 (t) = e (c1 (0) + c2 (0)) + e−i(ε−α)t/~ (c1 (0) − c2 (0))
2    
−iεt/~
h αt αt i
=e cos c1 (0) − i sin c2 (0) ,
~ ~
1 h i
c2 (t) = e−i(ε+α)t/~ (c1 (0) + c2 (0)) − e−i(ε−α)t/~ (c1 (0) − c2 (0))
2    
−iεt/~
h αt αt i
=e − i sin c1 (0) + cos c2 (0) . (2.35)
~ ~
We summarize the solutions in the matrix form,
cos αt αt
    


c1 (t) c1 (0) −iεt/~ −i sin
= U (t)
b , U (t) = e
b ~
.
~ (2.36)
−i sin αt cos αt


c2 (t) c2 (0) ~ ~

By using the information, we can calculate the transition probabilities from the initial state at t = 0 to any
state at an arbitrary time t.

2.5 Heisenberg Equation of motion

We will revisit the formula in Eq. (2.3) for evaluating the expectation value of a Hermitian operator A
b for
the time-evolved state |ψ(t)i at time t from the initial time t = 0,
D E D E
A
b = ψ(0) U b (t)† A
bUb (t) ψ(0) . (2.37)
|ψ(t)i

Here, we introduce the notation,

A b (t)† A
bH (t) := U bUb (t) = e+iHt/~
b b e−iHt/~
A
b
, (2.38)

and we can rewrite the above form as

D E D E D E
Ab = ψ(0) AH (t) ψ(0) =: AH (t)
b b , (2.39)
|ψ(t)i |ψ(0)i

25
where the time dependence originally introduced to the state |ψ(0)i is pressed into the opeator A.
b Thus,
the effects of time evolution can be imposed entirely on the state or entirely on the operator, where the
former scheme is named the Schrödinger picture, while the latter scheme is named the Heisenberg picture,
respectively. The expression defined in Eq. (2.38) is the operator Ab in the Heisenberg picture, which is
sometimes referred to as the Heisenberg operator (at time t) AH (t) or the Heisenberg operator of A
b b (at
time t) in short.
The time derivative of the Heisenberg operator AbH reads

dA
bH (t) d  +iHt/~ b e−iHt/~

= e A
b b
dt dt
i b +iHt/~ b e−iHt/~ b e−iHt/~ i b
= He A − e+iHt/~ A H
b b b b
~ ~
i bb bH (t) i H,
= H AH (t) − A b (2.40)
~ ~
which leads to

dA
bH (t) h i
i~ = AH (t) , H .
b b (2.41)
dt
The equation given in Eq. (2.41) is called the Heisenberg equation (for time-dependent operators in the
Heisenberg pictures).
h i
• If AH (t) , H = O
b b b (for any t > 0), W cH (t) is time independent and thus we get A bH (t) = A.
b We
h i h i
can also derive that, if A,bH b =O b (at t = 0), A bH (t) , H
b =O b owing to the Leibniz rules of
h i
operators. These result in the significant conclusion: if A, bH b = O,b expectation values of the
operator Ab does not depend on time (for an arbitrary state |ψ(0)i). Physical quantities that do not
change with time are called conserved quantities.

• Since Hb is commutative with H

b itself, H
b does not depend on time. Therefore, energy eigenvalues
are conserved quantities.

26