Chapter 12

Quantum Measurement

The essence of quantum mechanics lies in the indeterminacy associated with the statistical
interpretation of the wave function. A quantum state Ψ does not uniquely determine the
outcome of a measurement. All Ψ provides is the statistical distribution of possible results.
The physical system does not have the attribute in question prior to the measurement. Rather,
the act of measurement itself creates the property, limited only by the statistical constraint
imposed by the wave function. An immediately repeated measurement collapses the wave
function and forces the system to take a stand, helping to create an attribute that was not there
previously.

12.1 EPR paradox

Let us consider the decay of π 0 into an electron and a positron:

π 0 → e− + e+ . (12.1)

Assuming π 0 was at rest, e− and e+ fly off in opposite directions. π 0 has spin 0, so the conser-
vation of angular momentum requires that e− and e+ are in the singlet spin configuration:

1
√ (↑− ↓+ − ↓− ↑+ ). (12.2)
2

If e− has spin up, e+ has spin down, and vice versa. We cannot tell which combination we will
get in any particular π 0 decay, but we can tell that we will get each combination half the time
on average.
Suppose that e− and e+ fly way off, say 100 light years, and you measure the spin of e−
and get spin up. Then, you can say immediately that someone 100 light years away will get
spin down, if he or she measures the spin of e+ . If you think that e− really had spin up (and
so e+ spin down) from the moment they were created, then you are WRONG. Neither particle

1
2 CHAPTER 12. QUANTUM MEASUREMENT

had either spin up or spin down until the act of measurement collapsed the e− wave function to
spin-up, and instantaneously produced the spin-down state of e+ 100 light years away.
This is what takes place in reality, though in 1935, Einstein, Podolsky and Rosen claimed
that this phenomenon was contradictory to the principle of locality — no influence can prop-
agate faster than the speed of light.

12.2 Bell’s inequality

To solve the EPR paradox, Einstein, Podolsky and Rosen assumed that, in addition to the wave
function Ψ, some hidden variable λ is needed to characterize the state of a system fully. A
number of theories concerning λ were proposed until 1964 when Bell proved that any hidden
variable theory is incompatible with quantum mechanics.

Figure 12.1: Bell’s version of the EPR experiment: detectors independently oriented in direc-
tions a and b.

Bell suggested a generalized EPR experiment — instead of orienting the e− and e+ detec-
tors along the same direction, he allowed them to be rotated independently. The first detector
measures the component of the e− spin in the direction of a unit vector a, and the second de-
tector measures the component of the e+ spin in the direction of a unit vector b. Let us record
the spins in units of ~/2, then each detector registers +1 for spin up or −1 for spin down along
the direction in question. Suppose now that we calculate the average value of the product of
the spins, for a given set of detector orientations. We call this average P (a, b). If b = a (i.e.
the detectors are parallel), then the detector configuration is the same as in the EPR paradox.
In that case, one particle is spin up while the other is spin down; the product is always −1, so

P (a, a) = −1. (12.3)

If b = −a (i.e. the detectors are anti-parallel), then

P (a, −a) = +1. (12.4)

12.2. BELL’S INEQUALITY 3

For arbitrary detector orientations, we get1

P (a, b) = −a · b. (12.5)

However, Bell proved that this result is incompatible with any local hidden variable theory,
based on the following argument.
Suppose that the state of the e+ e− system is characterized by the hidden variable(s) λ.
Then there exists some function A(a, λ) which gives the result of the e− measurement, and
B(b, λ) for the e+ measurement. These functions can only take on the values ±1:

A(a, λ) = ±1, B(b, λ) = ±1. (12.6)

If the two detectors are aligned, the results are perfectly anti-correlated:

A(a, λ) = −B(a, λ), for all λ. (12.7)

The average of the product of the spin measurements is

Z
P (a, b) = ρ(λ)A(a, λ)B(b, λ)dλ, (12.8)
Z
where ρ(λ) is the probability density for λ, and satisfies ρ(λ)dλ = 1. Using Eq.(12.7),
Eq.(12.8) can be written as
Z
P (a, b) = − ρ(λ)A(a, λ)A(b, λ)dλ. (12.9)

Letting c be any other unit vector,

Z
P (a, b) − P (a, c) = − ρ(λ)[A(a, λ)A(b, λ) − A(a, λ)A(c, λ)]dλ. (12.10)

Since [A(b, λ)]2 = 1,

Z
P (a, b) − P (a, c) = − ρ(λ)[1 − A(b, λ)A(c, λ)]A(a, λ)A(b, λ)dλ. (12.11)

Noting that

−1 ≤ A(a, λ)A(b, λ) ≤ +1, ρ(λ)[1 − A(b, λ)A(c, λ)] ≥ 0, (12.12)

1
√
When two spin-1/2 particles are in the singlet configuration |s ms i = |0 0i = (1/ 2)(↑↓ − ↓↑), then
(2) ~2 (1)
hSa(1) Sb i = − cos θ, where Sa is the component of the spin angular momentum of particle number 1 along
4
(2)
the unit vector â, Sb be the component of the spin angular momentum of particle number 2 along the unit vector
b̂, and θ is the angle between â and b̂.
4 CHAPTER 12. QUANTUM MEASUREMENT

we can derive from Eq.(12.11) that

Z
|P (a, b) − P (a, c)| ≤ ρ(λ)[1 − A(b, λ)A(c, λ)]dλ. (12.13)

Using A(c, λ) = −B(c, λ) and Eq.(12.8), we get

|P (a, b) − P (a, c)| ≤ 1 + P (b, c). (12.14)

This equation is the Bell inequality. Now suppose that a, b and c lie in a plane and c makes a
45◦ angle with a and b, quantum mechanics says1
1
P (a, b) = 0, P (a, c) = P (b, c) = − √ = −0.707. (12.15)
2
These values are incompatible with Eq.(12.14):

0.707 6≤ 1 − 0.707 = 0.293. (12.16)

This result means that there exists no hidden variable that accommodates quantum mechanics
with the principle of locality.
Actually there are many things that travel faster than light — if a bug flies across the beam
of a movie projector, the speed of its shadow is proportional to the distance to the screen.
The distance can be as large as you like, so the shadow can move at arbitrarily high velocity.
However, the shadow does not carry any energy, nor can it transmit a message from one point to
another. We call this type of influence ethereal. On the other hand, a causal influence cannot
propagate faster than light. Special Relativity says that there exist inertial frames in which a
signal travelling faster than light propagates backward in time. In that case the effect precedes
the cause, which leads to inescapable logical anomalies.
In the EPR paradox, the measurement of e− does influence the outcome of the e+ measure-
ment (i.e. anti-correlation between the two spins), but the measurement of e− does not cause a
particular outcome for e+ . The person at the e− detector cannot control the outcome of his/her
own measurement, that is, the person cannot make a given e− come out spin-up. All he/she can
do is to decide whether to make a measurement at all, and the person at the e+ detector cannot
tell whether e− was measured or not, because the lists of data at the two detectors, consid-
ered separately, are completely random. Only when we compare the two lists of data later do
we discover the anti-correlations. In another reference frame the e+ measurements take place
before the e− measurements, but this does not result in any logical anomaly. Therefore, the
anti-correlation of the two spins is a delicate influence of ethereal type.
Thus, we distinguish two types of influence:

• “Causal” kind: an influence that produces actual changes in some physical property of
the receiver, detectable by measurements on that subsystem alone.
12.3. NO-CLONE THEOREM 5

• “Ethereal” kind: an influence that does not transmit energy or information, and for which
the only evidence is a correlation in the data taken on the two separate subsystems — a
correlation which cannot be detected by examining either list alone.

Causal influences cannot propagate faster than light, but ethereal ones can. The influences
associated with the collapse of the wave function is an ethereal kind.

12.3 No-clone Theorem

Measurements are destructive in quantum mechanics — they collapse wave functions and alter
the state of the system. Why don’t we just create clones (i.e. identical copies) of the original
state and measure those clones, leaving the original system itself unaffected by the measure-
ments? Suppose we were able to create such a “quantum Xerox machine”. Schematically,
the machine takes as input a particle in state |ψi (the original one to be copied) plus a second
particle in state |Xi (a “blank sheet of paper”), and spits out two particles in the state |ψi:

|ψi|Xi → |ψi|ψi. (12.17)

This machine could clone |ψ1 i and |ψ2 i separately2 :

|ψ1 i|Xi → |ψ1 i|ψ1 i, |ψ2 i|Xi → |ψ2 i|ψ2 i. (12.18)

But this machine cannot clone a linear combination of |ψ1 i and |ψ2 i, because

|ψi(α|ψ1 i + β|ψ2 i) → α|ψ1 i|ψ1 i + β|ψ2 i|ψ2 i (12.19)

6= (α|ψ1 i + β|ψ2 i)(α|ψ1 i + β|ψ2 i). (12.20)

Unless the machine can clone α|ψ1 i + β|ψ2 i, it cannot clone any state, because any state can
be expressed as a linear combination of states by changing the bases3 .

12.4 Schrödinger’s cat

What exactly is a measurement? How can we tell when a measurement has occurred? To pose
this essential question, Schrödinger posed a cat paradox.
A cat is placed in a chamber, together with a Geiger counter that contains a tiny amount of
radioactive substance. If one of the atoms decays, the counter triggers and activates a hammer
which breaks a container of cyanide and kills the cat. Suppose we left the entire system, say,
for an hour. If no atom has decayed, the cat is living. If the first decay has occurred, the cat is
2
|ψ1 i and |ψ2 i could be spin up and spin down, if the particle is an electron.
3 (z) 1 (x) (x)
For instance, recall χ+ = √ (χ+ + χ− ).
2
6 CHAPTER 12. QUANTUM MEASUREMENT

dead. We do not know whether the cat is alive or dead. So, at the end of the hour, the wave
function of the cat can be written as
1
ψ = √ (ψalive + ψdead ), (12.21)
2
which means that the cat is neither alive nor dead until you open the door of the chamber and
observe the cat inside. Your observation forces the cat to “take a stand”: dead or alive, and if
you find the cat to be dead, then it’s really you who killed him by looking in the window.
Obviously this is nonsense. There is something absurd about the idea of a macroscopic
object being in a linear combination of two different states. The answer is that the triggering
of the Geiger counter constitutes the “measurement”, not the intervention of a human observer.
The measurement occurs at the moment when the microscopic system, described by quantum
mechanics, interacts with the macroscopic system, described by classical mechanics, in such a
way as to leave a permanent record. The macroscopic system itself does not occupy a linear
combination of distinct states. Keep in mind that the measurement does not necessarily entail
human participation.

12.5 Quantum Zeno effect

A measurement collapses the wave function and makes it “take a stand”. An immediately
repeated measurement reproduces the same value, and this fact carries directly observable con-
sequences. Suppose we take an unstable system (e.g. an atom in an excited state), and subject
it to repeated measurements. Each observation collapses the wave function, and it is possible
to delay the transition to the lower state. This phenomenon is called the quantum Zeno effect.
Let us imagine a system starts out in the excited state ψ2 , which has a natural lifetime τ
for transition to the ground state ψ1 . For times substantially less than τ , the probability of
transition is proportional to t. Since the transition rate is 1/τ ,
t
P2→1 = . (12.22)
τ
If we make a measurement after a time t, the probability that the system is still in the upper
(ψ2 ) state is
t
P2 (t) = 1 − . (12.23)
τ
Suppose we do find the atom to be in the upper state. In that case, the wave function collapses
back to ψ2 , and the process starts all over again. If we make a second measurement at time 2t,
 2
t 2t
P2 (2t) = 1 − ≈1− , (12.24)
τ τ
12.6. NEUTRINO OSCILLATION 7

which is the same as it would have been if we had never made the first measurement at time t.
However, for extremely short times, the probability of a transition is not proportional to t.
If t is extremely small, Eq.(??) becomes
Z ∞  2 
2 2 sin [(ω0 − ω)t/2]
Pb→a (t) = |P| ρ(ω) dω (12.25)
0 ~2 0 (ω0 − ω)2
Z ∞
[(ω0 − ω)t/2]2
 
2 2
≈ |P| ρ(ω) dω (12.26)
0 ~2 0 (ω0 − ω)2
2 Z ∞
2 2t
= 2
|P| ρ(ω)dω ∝ t2 . (12.27)
0 ~ 4 0
So, the probability of a transition can be written as P2→1 = αt2 , and the probability that the
system is still in the upper state after the two successive measurements is

P2 (2t) = (1 − αt2 )2 ≈ 1 − 2αt2 , (12.28)

whereas it would have been 1 − α(2t)2 ≈ 1 − 4αt2 if we had never made the first measurement.
If we repeatedly examine the system at time T /n, 2T /n, 3T /n, . . ., T , then
(  2 )n
T α
P2 (T ) = 1 − α ≈ 1 − T 2, (12.29)
n n

which goes to 1 in the limit n → ∞ — a continuously observed unstable system never decays at
all. In fact, the experiment is impractical for spontaneous transitions, but in can be done using
induced transitions, and the quantum Zeno effect has been confirmed by actual experiments.

12.6 Neutrino oscillation

There exit three kinds of neutrinos: electron neutrino (νe ), muon neutrino (νµ ) and tau neutrino
(ντ ). These are the states of neutrinos produced by particle interactions, and are called the
flavor eigenstates. On the other hand, the eigenstates of the Hamiltonian are called the mass
eigenstates:

H|ν1 i = E1 |ν1 i, H|ν2 i = E2 |ν2 i, H|ν3 i = E3 |ν3 i. (12.30)

The point is that the flavor eigenstates are not the mass eigenstates, but a mixture of them. For
simplicity, let us consider νe and νµ alone, and express them as:




|νe i|νµ i = cos θ sin θ − sin θ cos θ |ν1 i|ν2 i . (12.31)

Imagine a neutrino is produced at t = 0 as a νµ :

|ν(0)i = |νµ i = − sin θ|ν1 i + cos θ|ν2 i. (12.32)

8 CHAPTER 12. QUANTUM MEASUREMENT

At time t, the mass eigenstates pick up phase factors:

|ν(t)i = − sin θ|ν1 ie−iE1 t/~ + cos θ|ν2 ie−iE2 t/~ . (12.33)

The probability that the neutrino is observed as a νe at time t is:

Pνe →νµ = |hνe |ν(t)i|2 (12.34)

= |(cos θhν1 | + sin θhν2 |)(− sin θ|ν1 ie−iE1 t/~ + cos θ|ν2 ie−iE2 t/~ )|2 . (12.35)

Using hν1 |ν1 i = hν2 |ν2 i = 1 and hν1 |ν2 i = hν2 |ν1 i = 0,

Pνe →νµ = | − sin θ cos θe−iE1 t/~ + sin θ cos θe−iE2 t/~ |2 (12.36)
2 2 −i(E1 −E2 )t/~ i(E1 −E2 )t/~
= sin θ cos θ(1 − e −e + 1) (12.37)
  
2 2 (E1 − E2 )t
= sin θ cos θ 2 − 2 cos (12.38)
~
 
(E1 − E2 )t
= sin2 (2θ) sin2 (12.39)
2~
Meanwhile, special relativity says
s
p m 2 c2
E = p2 c2 + m2 c4 = pc 1+ . (12.40)
p2

The neutrino’s mass is very small (m2 c4  p2 c2 ), so

1 m 2 c2 m2 c3
 
E ≈ pc 1 + = pc + . (12.41)
2 p2 2p
Also, E ≈ pc holds although the neutrino may change its flavors, so

∆m2
E1 − E2 = , where ∆m2 ≡ (m1 c2 )2 − (m2 c2 )2 . (12.42)
2E
Moreover, v ≈ c, so t ≈ L/c, where L is the flight distance of the neutrino. Eq.(12.39)
becomes
∆m2 L
 
2 2
Pνe →νµ = sin (2θ) sin . (12.43)
4E~c
∆m2 [eV2 ]
 
2 2
= sin (2θ) sin 1.27 L[m] (12.44)
E[MeV]
The probability that the flavor of the neutrino changes from νµ to νe oscillates according to
the flight distance of the neutrino. This phenomenon is called the neutrino oscillation. The
12.7. BLACKBODY RADIATION 9

discovery of neutrino oscillation is the evidence for the non-zero mass of the neutrino. When
all the three kinds of neutrinos are involved, the mixing matrix is called the MNS matrix4 and
written as
−iδ13
   

1 0 0 c 13 0 s 13 e c 12 s 12 0

|νe i|νµ i|ντ i = 0 c23 s23   0 1 0  −s12 c12 0 |ν1 i|ν2 i|ν3 i
0 −s23 c23 −s13 eiδ13 0 c13 0 0 1
(12.45)
where sij and cij represent sin θij and cos θij , respectively, and δ13 is the CP (Charge-Parity)
violating phase.
Also in the quark sector there is a similar mixing of the d (down), s (strange), and b (bottom)
quarks, and the mixing matrix is named the CKM5 matrix.

12.7 Blackbody radiation

In §?? we learned that the energy density of the black body is described by Planck distribu-
tion:
8πhν 3 1
I(ν) dν [J m−3 ] = 3 hν/k T −1
dν. (12.46)
c e B

Let us derive this equation in two ways — One is Planck’s original approach, and the other is
a modern approach using statistical mechanics as well as quantum mechanics.

12.7.1 Density of states

First, let us derive the density of states6 for electromagnetic radiation (i.e. light waves) confined
in a cube of side length L (cavity radiation). The derivation is similar to the free electron gas
model that we learned in §??. Due to the boundary condition at the walls of the cube, the light
waves have the amplitudes proportional to:
n π  n π  n π 
x y z
sin x sin y sin z , (12.47)
L L L
where nx = 1, 2, 3, . . . , ny = 1, 2, 3, . . . , nz = 1, 2, 3, . . . . (12.48)

The wave vector is

 
nx π ny π nx π
k ≡ (kx , ky , kz ) = , , . (12.49)
lx ly lx
4
Maki-Nakagawa-Sakata matrix
5
Cabibbo-Kobayashi-Maskawa matrix
6
g(ν)dν is the number of states whose frequencies are within ν and ν + dν.
10 CHAPTER 12. QUANTUM MEASUREMENT

In the 3D k space, each block of volume π 3 /L3 contains two states, since light waves are
transverse and have two degrees of freedom with respect to their polarization. Noting that the
states in the space with nx > 0, ny > 0, nz > 0 are occupied, the number of states contained in
one octant of a shell of thickness dk is
1
4πk 2 dk × 2 k 2 L3
8 = 2 dk. (12.50)
π3 π
L 3

Using k = 2πν/c, Eq.(12.50) becomes

4π 2 ν 2 2π L3 8πν 2 3
π dν = L dν. (12.51)
c2 c π3 c3
Thus, the density of states per unit volume is

8πν 2
g(ν)dν = dν. (12.52)
c3

12.7.2 Planck’s quantum hypothesis

In 1900, Planck postulated that light energy can only be emitted and absorbed in discrete bun-
dles called quanta7 — the energy of light of frequency ν is determined by8

E = n hν (n = 0, 1, 2, 3, . . .). (12.53)

According to Maxwell-Boltzmann distribution law, the probability that a particle in a thermal

equilibrium at temperature T has energy  is proportional to exp(−/(kB T )). So, the average
energy of light of frequency ν is
∞  
X nhν
nhν exp −
kB T
Ē = n=0∞   . (12.54)
X nhν
exp −
n=0
kB T

The denominator becomes

∞  
X nhν 1
exp − = (12.55)
n=0
kB T 1− e−hν/(kB T )
7
Today we call them photons.
8
h is called Planck constant today and known to have the value of 6.626 × 10−34 [m2 kg s−1 ].
12.7. BLACKBODY RADIATION 11

Letting ξ = 1/(kB T ), the numerator becomes

∞ ∞
X d X −nhνξ d
−1 hνe−hνξ
nhνe−nhνξ = − e =− 1 − e−hνξ = (12.56)
n=0
dξ n=0 dξ (1 − e−hνξ )2
Putting Eq.(12.55) and Eq.(12.56) into Eq.(12.54),
hνe−hν/(kB T ) hν
Ē = = . (12.57)
1 − e−hν/(kB T ) ehν/(kB T ) − 1
Multiplying Eq.(12.57) by Eq.(12.52), we get Planck distribution Eq.(12.46).

12.7.3 Modern approach

The energy levels of a 1D harmonic oscillator is given by
   
1 1
n = n + ~ω = n + hν (n = 0, 1, 2, . . .). (12.58)
2 2
In statistical mechanics, when a closed system is in thermal contact with a heat bath and only
the thermal energy exchange occurs between them, the probability distribution of each con-
stituent particle being in a state with energy n is
  ∞  
1 n X n
Pn = exp − , where Z ≡ exp − . (12.59)
Z kB T n=0
kB T

Z is called the canonical partition function. For Eq.(12.58), Z becomes

∞   
X hν 1 exp(−hν/(2kB T ))
Z= exp − n+ = . (12.60)
n=0
kB T 2 1 − exp(−hν/(k B T ))

The Helmholtz free energy is

  
1 hν
φ = −kB T ln Z = hν + kB T ln 1 − exp − . (12.61)
2 kB T
The total free energy of the whole system is obtained by the superposition of individual parti-
cles represented by harmonic oscillators:
X 1  
hνi

F = hνi + kB T ln 1 − exp − (12.62)
i
2 kB T

The internal energy of the system is

  X 
2 d F 1 hνi
E = −T = hνi + (12.63)
dT T i
2 exp(hνi /(kB T )) − 1)
12 CHAPTER 12. QUANTUM MEASUREMENT

The first term in Eq.(12.63) represents the zero-point energy that exists even at absolute zero
temperature. Using Eq.(12.52), the energy of the system relative to the zero-point energy can
be written as
Z ∞
hν
E = g(ν)dν (12.64)
0 exp(hν/(kB T )) − 1
Z ∞
8πh ν3
= dν (12.65)
0 c3 exp(hν/(kB T )) − 1

The integrand is exactly Planck distribution Eq.(12.46).

Chapter 13

Quantum Computation and Information

13.1 Classical Computation

Before we can discuss quantum computation, it is useful to introduce the fundamental concepts
behind classical computation: (1) the universal Turing machine, (2) algorithmic efficiency, and
(3) computational complexity.
In 1936 Alan Turing introduced a universal model for computation, in the form of his
description of a hypothetical computing machine, which today bears his name.

Definition 13.1.1 (Turing Machine). A Turing machine consists of “. . .an unlimited memory
capacity obtained in the form of an infinite tape marked out into squares, on each of which
a symbol could be printed. At any moment there is one symbol in the machine; it is called
the scanned symbol. The machine can alter the scanned symbol, and its behavior is in part
determined by that symbol, but the symbols on the tape elsewhere do not affect the behavior
of the machine. However, the tape can be moved back and forth through the machine, this
being one of the elementary operations of the machine. Any symbol on the tape may therefore
eventually have an innings.” 1

A schematic example of a particular Turing machine is shown in Figure 13.1.

Figure 13.1: Schematic of one type of Turing machine.

1
A.M. Turing (1948). “Intelligent Machinery”. The Turing Archive, pp. 3.

13
14 CHAPTER 13. QUANTUM COMPUTATION AND INFORMATION

Definition 13.1.2 (Universal Turing Machine). A universal Turing machine is a Turing ma-
chine that can simulate an arbitrary Turing machine on arbitrary input.
We can further generalize this concept of a universal Turing machine to a programming
language which can simulate a Turing machine.
Definition 13.1.3 (Turing Completeness). A programming language is Turing complete if it
has the ability to simulate a Turing machine.
If a Turing complete programming language can simulate any Turing machine, it is theoreti-
cally capable of expressing all task accomplishable by all Turing machines. The Church-Turing
thesis generalizes this finding to all computational devices, algorithms, and functions.
Thesis 13.1.1 (Church-Turing). A function on the natural numbers is computable if and only
if it is computable by a Turing machine.

a b NOT a a AND b
0 0 1 0
0 1 1 0
1 0 0 0
1 1 0 1
a b a OR b a XOR b
0 0 0 0
0 1 1 1
1 0 1 1
1 1 1 0
a b a NAND b a NOR b
0 0 1 1
0 1 1 0
1 0 1 0
1 1 0 0
Figure 13.2: Circuit representations and truth tables for six basis logic gates: (a) NOT a, (b) a
AND b, (c) a OR b, a XOR b, (e) a NAND b, and (f) a NOR b.

Although powerful, the definition of a Turing machine is somewhat unintuitive. For this
reason, it is often useful to recast a Turing machine in terms of logic gates, such as NOT, AND,
OR , XOR , NAND , and NOR , as described in Figure 13.2. Using logic gates, one may simulate a
universal Turing machine using just the NOT, AND, and OR gates. This may be simplified even
further to the universal NAND gate.
Theorem 13.1.1 (Universal NAND Gate). Any function on bits can be computed from the
composition of NAND gates alone, which is thus known as a universal gate.
13.2. QUANTUM COMPUTATION 15

When combined together with Church-Turing thesis, the Universal NAND Gate theorem
implies that any Turing machine can be simulated by a composition of the universal NAND
gate.
Since most programming languages are Turing complete, that is, they contain a NAND gate,
the Church-Turing thesis implies most programming languages and computational devices are
both equivalent and optimal. This idea of a universal Turing machine, or a Turing complete
programming language, being somehow optimal is espoused in the so-called strong version of
the Church-Turing thesis.
Thesis 13.1.2 (Strong Church-Turing). Any algorithmic process can be simulated efficiently
using a Turing machine.
However, to understand the strong Church-Turing thesis we first need to define what is
meant by an efficient and inefficient simulation. From the area of computational complexity we
obtain the following definitions.
Definition 13.1.4 (Efficient Algorithm). An algorithm is said to be efficient if it runs in time t
which grows polynomially with the size of the problem to be solved N , that is, t ≈ O(N α ) for
α ∈ R.
Definition 13.1.5 (Inefficient Algorithm). An algorithm is said to be inefficient if it requires
superpolynomial (typically exponential) time in the size of the problem to be solved N , that is,
t ≈ O(αN ) for α ∈ R.
So any algorithm whose execution time grows polynomially, O(N α ), with the problem size
is said to be efficient, while any algorithm whose execution time grows exponentially, O(αN ),
with the problem size is said to be inefficient.
However, in the 1970s, Solovay and Strassen showed that by using a randomized algorithm,
one could determine with high probability whether a number was prime or composite, whereas
no efficient deterministic test for primality was known. This lead to the first modification of
the strong Church-Turing thesis:
Thesis 13.1.3 (Probabilistic Strong Church-Turing). Any algorithmic process can be simulated
efficiently using a probabilistic Turing machine.
This alteration to the original strong Church-Turing thesis begs the question Is there an-
other model of computation that allows one to efficiently solve problems that are not efficiently
solvable within Turing’s model of computation? It is this question which led David Deutsch in
1985 to propose the quantum model of computation.

13.2 Quantum Computation

Since the laws of physics are ultimately quantum mechanical, Deutsch suggested that for a
computational device to efficiently simulate an arbitrary physical system, the computational
16 CHAPTER 13. QUANTUM COMPUTATION AND INFORMATION

device itself must be based upon the principles of quantum mechanics. This led Deutsch to
propose a quantum Turing machine, or universal quantum computer.
To do so, Deutsch employed the quantum analogue of the universal NAND gate for quan-
tum systems, that is, the controlled-NOT (CNOT) gate, or matrix operator UCN , shown in Fig-
ure 13.3.
 
|Ai ——•—— |Ai 1 0 0 0 |00i
|  0 1 0 0  |01i
| UCN =  
 0 0 0 1  |10i
|
|Bi ——- ⊕—— |B ⊕ Ai 0 0 1 0 |11i
|00i → |00i; |01i → |01i; |10i → |11i; |11i → |10i.
|A, Bi → |A, B ⊕ Ai

Figure 13.3: Schematic, matrix, and functional representations of the CNOT or controlled-NOT
gate, or matrix operator UCN , where ⊕ is addition modulo two.

From Figure 13.3 we see two important features of quantum logic gates:

1. Quantum logic gates are quantum operators, and for this reason, are linear, unitary, and
hence reversible.

2. Quantum logic gates act on one or more quantum states which may be represented by a
two state basis, {|0i, |1i}, or qubit.

Essentially, the CNOT gate flips the second qubit if the first qubit is in the |1i state.

Definition 13.2.1 (Qubit). A qubit is a two level quantum system, |ψi, which may be repre-
sented as
|ψi = α|0i + β|1i
where α, β ∈ C and |α|2 + |β|2 = 1, so that |0i and |1i form an orthonormal basis set over the
two-dimensional complex vector space of the qubit.

Since a qubit |ψi is a superposition over a two-level orthonormal vector space, it may be
represented in the form
θ θ
|ψ = cos |0i + eiϕ sin |1i.
2 2
It is this representation, where we may consider θ to be the azimuthal angle and ϕ to be the
polar angle in spherical coordinates, which gives rise to the Bloch sphere representation of a
qubit shown in Figure 13.4.
Four of the most important single qubit gates are shown in Figure 13.5.
13.2. QUANTUM COMPUTATION 17

Figure 13.4: Bloch sphere representation of the qubit |ψi.

 
1 1 1
(a) Hadamard — H — √2
1 −1
 
1 0
(b) phase —S —
0 i
 
1 0
(c) π/8 —T — iπ/4
0 e
 
0 1
(d) Pauli-X —X —
1 0
 
0 −i
(d) Pauli-Y —Y —
i 0
 
1 0
(d) Pauli-Z —Z —
0 −1

Figure 13.5: Single qubit (a) Hadamard, (b) phase, (c) π/8, (d) Pauli-X, (c) Pauli-Y , and (f)
Pauli-Z gates.

The Hadamard gate puts a single qubit into a superposition of |0i and |1i, so that
|0i + |1i
|0i — H — √
2
|0i − |1i
|1i — H — √
2
(α + β)|0i + (α − β)|1i
α|0i + β|1i — H — √
2
18 CHAPTER 13. QUANTUM COMPUTATION AND INFORMATION

This can be interpreted as a rotation on the Bloch sphere, as shown schematically in Figure 13.6

Figure 13.6: Visualization of the Hadamard gate on the Bloch sphere, acting on the input state
|0i+|1i
√
2
.

The phase gate adds a phase of ϕ = π/2, that is, i = e−π/2 , so that

α|0i + β|1i — S — α|0i + iβ|1i

Similarly, the π/8 gate adds a phase of ϕ = π/4, so that

α|0i + β|1i — S — α|0i + βeiπ/4 |1i,

while the Pauli-Z gate adds a phase of ϕ = π, so that

α|0i + β|1i — S — α|0i − β|1i,

Note that applying the π/8 gate twice is equivalent to the phase gate, and applying the π/8 gate
four times is equivalent to the Pauli-Z gate, that is,

—S —S —≡—T —
and
— S — S — S — S — ≡ — T — T — ≡ — Z —.

Theorem 13.2.1 (Universal CNOT Gate). Any multiple qubit logic gate may be composed from
CNOT and the single qubit Hadamard and π/8 gates.
13.2. QUANTUM COMPUTATION 19

It is important to note that although the CNOT gate UCN (see Figure 13.3) and the single
qubit Hadamard and π/8 gates (see Figure 13.5) are the universal gates in quantum computing,
the CNOT gate is not equivalent to a classical NAND gate. Rather, the CNOT gate is the quantum
generalization of the classical XOR gate shown in Figure 13.2(d), with an auxiliary bit for A.
This brings up an important question: Can a quantum computer simulate any classical
computer? One may initially assume the obvious answer is yes, as all physical systems are
essentially classical. However, the system of quantum logic gates described by Deutsch is
inherently reversible, and coherent, which is not the case for all classical gates. For this reason,
the universal classical NAND gate does not have a direct quantum generalization, as the NAND
gate is inherently irreversible. To address this issue, we may employ ancillary bits (or qubits),
and in so doing, formulate reversible versions of irreversible gates.
A useful reversible classical logic gate is the Toffoli gate, shown in Figure 13.7.

Inputs Outputs
a0 b 0 c 0
 
a b c 1 0 0 0 0 0 0 0 |000i
a ——•—— a 0 0 0 0 0 0  0 1 0 0 0 0 0 0  |001i
|  
| 0 0 1 0 0 1 
 0 0 1 0 0 0 0 0 
 |010i
| 0 1 0 0 1 0  0 0 0 1 0 0 0 0  |011i
b ——•—— b  
| 0 1 1 0 1 1  0 0 0 0 1 0 0 0  |100i
| 1 0 0 1 0 0

0 0 0 0 0 1 0 0

|101i
|
 
 
c ⊕
——-—— c ⊕ ab 1 0 1 1 0 1  0 0 0 0 0 0 0 1  |110i
1 1 0 1 1 1 0 0 0 0 0 0 1 0 |111i
1 1 1 1 1 0

Figure 13.7: Circuit representation, truth table, and matrix representation of the reversible
Toffoli gate.

Since the Toffoli gate is reversible and has a unitary matrix representation, it may be sim-
ulated using a composition of quantum logic gates. Moreover, by applying a Toffoli gate to a
and b with a 1 in the third ancillary bit, one may obtain the NAND in the third ancillary bit, as
shown in Figure 13.8.
Since the Toffoli gate can simulate the universal NAND gate using an ancillary bit, and there
exists a direct quantum generalization of the reversible Toffoli gate, it follows that a universal
quantum computer can efficiently simulate any classical computer.
However, single qubits are rather limited. To exploit the power a quantum system to per-
form computations, we must employ entanglement. A prototypical example are the Bell states
|βab i shown in Figure 13.9.
So, by applying a Hadamard gate to one qubit to put rotate a pure |0i or |1i state into
a superposition of |0i+|1i
√
2
or |0i+|1i
√
2
, and then applying the CNOT gate, we obtain two entan-
gled qubits. It can be shown that the four Bell states have the highest degree of entanglement
20 CHAPTER 13. QUANTUM COMPUTATION AND INFORMATION

a ——•—— a
|
|
|
b ——•—— b
|
|
|
1 ⊕
——-—— 1 ⊕ ab = ¬(ab)

Figure 13.8: Circuit representation of a Toffoli gate with an ancillary 1 in the third bit providing
the result of the classical NAND gate on bits one and two in the third bit.

In Out
|00i+|11i
|00i √ ≡ |β00 i
2 a — H —•——
|01i+|10i |
|01i √
2
≡ |β01 i | |βab i
|10i−|11i
|
|10i √ ≡ |β10 i b ———- ⊕——
2
|01i−|10i
|11i √
2
≡ |β11 i

Figure 13.9: Input-output quantum “truth table” and quantum circuit to create Bell states.

possible between two qubits.

It is the |β00 i Bell state which we can use to perform quantum teleportation, as shown in
Figure 13.10.

|ψi ———•—— H —— M1 ======

===========•
| k
Alice | k
———⊕—————- M2 ======• k

Bob
|β00 i { —————————————– X
k
M2
k
—— Z M1 —— |ψi

Figure 13.10: Quantum circuit for teleporting a qubit. The qubit |ψi to be teleported and the
first of the two entangled qubits in the Bell state |β00 i belong to Alice, while the second of the
two entangled qubits in the Bell state |β00 i belongs to Bob. M1 and M2 are measurements,
producing classical bits denoted by double lines.

Suppose the state to be sent by Alice is |ψi = α|0i + β|1i, then the algorithm acts as
13.3. COMPUTATIONAL COMPLEXITY 21

follows:
|00i + |11i
(α|0i + β|1i) √
2
α|0i(|00i + |11i) + β|1i(|10i + |01i)
CNOT √
2
α(|0i + |1i)(|00i + |11i) + β(|0i − |1i)(|10i + |01i)
H
2
|00i(α|0i + β|1i) + |01i(β|0i + α|1i) + |10i(α|0i − β|1i) + |11i(−β|0i + α|1i)
=
2
So we have four equally probably outcomes:
1. Alice’s qubits are in the |00i state, so Bob’s qubit is in the |ψi state.
2. Alice’s qubits are in the |01i state, so Bob applies a Pauli-X gate to get his qubit in the
|ψi state.
3. Alice’s qubits are in the |10i state, so Bob applies a Pauli-Z gate to get his qubit in the
|ψi state.
4. Alice’s qubits are in the |11i state, so Bob applies a Pauli-X and PauliZ gate to get his
qubit in the |ψi state.
Since Alice’s qubits have equal probability of being in either |00i, |01i, |10i, or |11i, Bob
is only able to obtain |ψi by first knowing which state Alice’s qubits are in. This ensures that
(1) Alice must first measure the state of her two qubits, destroying the state |ψi so that the non-
cloning theorem is not violated, and (2) no information is transferred faster than the classical
communication of which state Alice’s qubits were in.

13.3 Computational Complexity

At this point, we may wish to know whether quantum computers efficiently solve problems
classical computers cannot. To discuss this it is useful to consider the set of computational
complexity classes P, NP, PSPACE, and BQP.
Definition 13.3.1 (P). P is the set of all problems having solutions which require polynomial
time in the size of the problem, t ≈ O(N α ) for α ∈ R, on a classical computer.
Definition 13.3.2 (NP). NP is the set of all problems having solutions which require non-
polynomial time in the size of the problem, t ≈ O(αN ) for α ∈ R, on a classical computer.
Theorem 13.3.1 (P ⊆ NP). Any problem having a solution which require polynomial time
in the size of the problem, t ≈ O(N α ) for α ∈ R, also has solutions which require non-
polynomial time in the size of the problem, t ≈ O(β N ) for β ∈ R, so that P ⊆ N.
22 CHAPTER 13. QUANTUM COMPUTATION AND INFORMATION

Figure 13.11: The relationship between classical and quantum complexity classes.

Definition 13.3.3 (PSPACE). PSPACE is the set of all problems having solutions which re-
quire space which grows polynomially with the problem size, on a classical computer.

Definition 13.3.4 (BQP). BQP is the set of all problems having solutions which require poly-
nomial time in the size of the problem, t ≈ O(N α ) for α ∈ R, on a quantum computer.

As shown in Figure 13.11 and discussed previously, any algorithm which can be performed
efficiently on a classical computer can be efficiently simulated on a quantum computer. This
means P ⊆ BQP. However, although it is generally accepted that quantum computers can
perform efficiently certain tasks which take exponential time on a quantum computer, such as
the quantum Fourier transform, this is very difficult to prove. For this reason, we cannot be
sure that P 6= BQP. Similarly, although BQP ⊆ PSPACE, we cannot be sure that BQP 6=
PSPACE, as it is still not known whether P 6= PSPACE.
It is also useful to consider the set of NP-complete problems. These are problems in NP
whose most efficient known solution takes time t which grows exponentially in the problem
13.4. QUANTUM INFORMATION 23

size N , that is, t ≈ O(αN ) for some α ∈ R, and for which any efficient solution to one
problem could be used to efficiently solve all other NP-complete problems. For this reason,
it is anticipated that no efficient solution exists for any NP-complete problem. It should be
noted that the most efficient solution for NP-complete problems, based on the quantum search
algorithm, requires a quantum computer. Although this solution still takes exponential time
√ problem size, it runs for the square root of the classical run time, that is, tquantum ≈
in the
O( tclassical ).

13.4 Quantum Information

The main idea of quantum information is that even when using a noisy quantum channel with
zero capacity, so long as one can send qubits in both directions and one has a separate classical
channel with non-zero capacity, one can send quantum information. This is accomplished as
follows:

1. Alice prepares a large number m of Bells states.

2. Alice sends half of each Bell pair through the noisy channel to Bob.

3. Alice and Bob now perform entanglement distillation and obtain pure Bell pairs.

4. Alice prepares her qubit state and uses her Bell pairs to teleport the state to Bob.

Note that this procedure for transmitting qubits requires a classical channel to (1) perform the
entanglement distillation and (2) teleport the prepared qubit state from Alice to Bob.

13.5 Quantum Algorithms

Before describing the two main types of quantum algorithms, those based on the quantum
Fourier transform,√which provides an exponential speed-up, and the quantum search algorithm,
which provide a N speed-up, we should first put quantum algorithms on a firmer basis.
Specifically, we must first address the problem of decoherence time in quantum systems, which
limits the stability of a quantum system. In fact, this problem may be solved through the use of
quantum error correction codes.

13.5.1 Quantum Error Correction

The most straight-forward efforts to address errors in a quantum computer are hampered by the
following three constraints of quantum information:

1. No cloning,
24 CHAPTER 13. QUANTUM COMPUTATION AND INFORMATION

(a) (b)
|ψi ——•——–•—- H ———— • ——
• —— ——— • ——–
• ——
⊕—— H ——•———
• ———
⊕—— |ψi
| | | | | | | | | |
| | | | | | | | | |
| | |0i ———
⊕——| —— ——— ⊕——–
| ——
•— | | |
| | | | | | | |
| | | | | | | |
| | |0i ——— ⊕—— ———
—— ⊕——
——– •— | | |
| | | | |
| | | | |
|0i ——⊕
——–|—- H ———— • ——
• —— ——— • ——–
• —— | ———
⊕—— H ——⊕——— •—
| | | | | | | |
| | | | | | | |
| |0i ———
⊕——| —— ——— ⊕——–
| ——
•— | |
| | | | | |
| | | | | |
| |0i ——— ⊕—— ———
—— ⊕——
——– •— | |
| | |
| | |
|0i ————–
⊕ —- H ———— • ——
• —— ——— • ——–
• —— ⊕———
⊕—— H ——–——— •—
| | | | |
| | | | |
|0i ———
⊕——| —— ——— ⊕——–
| ——
•—
| | |
| | |
|0i ——— ⊕—— ———
—— ⊕——
——– •—
Figure 13.12: (a) Encoding and (b) decoding circuits for the Shor nine qubit code.

2. Errors are continuous, and

3. Measurement destroys quantum information.

However, even within these constraints, quantum error correction is possible.

The Shor code can protect against the effects of an arbitrary error on a single qubit, by
representing a single qubit using multiple qubits. This redundancy, although costing some
overhead, is simply a prefactor for the purposes of analyzing algorithms. Moreover, it opens
the way for performing fault tolerant quantum computation.
The basic nine qubit algorithm involves the encoding of the qubits |0i and |1i as the logical
qubits |0L i and |1L i, that is,

(|000i + |111i)(|000i + |111i)(|000i + |111i)

|0i → |0L i ≡ √
2 2
(|000i − |111i)(|000i − |111i)(|000i − |111i)
|1i → |1L i ≡ √
2 2

The algorithm itself is shown in Figure 13.12. By encoding the qubit in one of the two states
| + ++i and | − −−i, the logical qubit is robust against both sign and phase flips, which are
corrected without requiring a measurement during the decoding step.
 13.6. QUANTUM CRYPTOGRAPHY 25

13.5.2 Quantum Fourier Transform

The most well-known application of the quantum Fourier transform is the factoring of a com-
posite number into its two primes using Shor’s factoring algorithm. This takes a problem which
is thought to have no efficient solution on classical computers, i.e., be part of NP, and provide
a solution of order O(N 3 ), in the complexity class BQP. The quantum Fourier transform is
defined as follows.
Definition 13.5.1 (Quantum Fourier Transform). The quantum Fourier transform on an or-
thonormal basis set |0i, . . . , |N − 1i is defined as
N −1
1 X 2πijk/N
|ji −→ √ e |ki. (13.1)
N k=0

The action of the quantum Fourier transform on an arbitrary state is then

N
X −1 N
X −1
xj |ji −→ yk |ki. (13.2)
j=0 k=0

where the amplitudes yk are the discrete Fourier transforms of the amplitudes xj . Since the
Fourier transform is a linear unitary transformation, it may be implemented on a quantum
computer as shown in Figure 13.13.

|j1 i – H –R1 – · · · –Rn−1 — Rn —————— ——-———— - ————————— ——- |0i + e2πi0.j1 ···jN |1i
| | |
|j2 i ——- —•—– · · · –—|———|—— H – · · · – Rn−2 — Rn−1 — · · · —————— ——- |0i + e2πi0.j2 ···jN |1i
| |
.. | | | |
. | | | |
| | | |
|jn−1 i ——-——– · · · –—•———|———— · · · –—•———|—— · · · — H — Rn ——- |0i + e2πi0.jN −1 jN |1i
|| || |
| | |
|jn i ——–——– · · · –——–——•———— · · · –——–——•—— · · · ————•—– H – |0i + e2πi0.jN |1i
Figure 13.13: Efficient circuit for the quantum Fourier transform.

Here, the Rk gate denotes a rotation by 2π/2k , that is,

 
1 0
Rk ≡ k . (13.3)
0 e2πi/2

13.6 Quantum Cryptography

Although quantum algorithms allow the efficient factoring of large composite numbers into two
primes, thus breaking RSA public key encryption, quantum algorithms also provide a means
26 CHAPTER 13. QUANTUM COMPUTATION AND INFORMATION

for provably secure communication, via quantum cryptography or quantum key distribution.
Quantum cryptography is based on the well-established private key encryption schemes, some-
times called a “one time pad”. A one time pad is a set of random numbers which are both
private and shared by the sender, Alice, and the receiver, Bob. Once Alice and Bob have the
same private key, Alice can “add” the private key to the message, send it publicly, and Bob can
subtract the key from the message to decode it. So long as the key is (1) private, (2) only used
once, and (3) truly random, then an eavesdropper, Eve, cannot decrypt the message.
Private key encryption has been used for centuries, and is still in use today. The problem
with such schemes is that both generating truly random keys is non-trivial, and keeping them
secure while passing them from sender to receiver is hard to ensure. This is where quantum
key distribution comes in.
The strength of quantum key distribution lies in the fact that Eve cannot gain any infor-
mation from the qubits transmitted from Alice to Bob without disturbing their state. Firstly,
by the no-cloning theorem, Eve cannot clone Alice’s qubit. Secondly, any attempt to distin-
guish between two non-orthogonal quantum states can only gain information at the expense of
introducing disturbance to the signal.
By transmitting non-orthogonal qubit states from Alice to Bob and checking for disturbance
in the transmitted states, they can obtain an upper bound on any noise or eavesdropping by
Eve. This is accomplished using a random selection of “check” qubits. Alice and Bob can
then perform information reconciliation and privacy amplification to distill a shared secret key
string. It is this private key which can then be used as a one time pad for provably secure private
communication.
13.6. QUANTUM CRYPTOGRAPHY 27

Class Project
Write an essay on a topic related to Quantum Measurement, Quantum Reality, Quantum Com-
putation, Quantum Information, Quantum Cryptography, Quantum Algorithms, or a related
field. Your essay should include:

1. A thesis (due Friday December 1st, 2017)

2. A graphic for the table of contents.

3. A short abstract of less than 150 words.

4. One to two paragraphs of introduction.

5. At least three different arguments supporting your thesis, and refutations of any well-
known counter-arguments.

6. At most 2000 words or four pages, single spaced, double column, in English.

7. A summary and conclusions.

8. Captions for any figures or schemes included, which must be referenced in the text.
Such figures and schemes are highly encouraged, but must be original and not copied
from other sources.

The essay should be written in English at the level of a “lay” audience, that is, with the scientific
background of a fourth semester Yachay Tech student.

1. A rough draft outline of the essay should be given to the instructor for comments and
suggestions on Monday December 4th, 2017. The instructor will provide comments
and further suggestions to any drafts sent until the final due date.

2. A final version of each essay is due Friday December 8th, 2017.

Each essay will be evaluated based on individual creativity, originality, scientific rigour,
clarity of expression, and soundness of argument by both a non-technical “lay” referee and a
technical referee, i.e., the instructor.
If a sufficient number of the essays are deemed to be of significant quality, originality, and
interest, then for those authors who wish to contribute, their essays may be compiled together in
a book form which will be “published” on the Yachay Tech website at yachaytech.edu.ec and on
the Cornell University archive at arXiv.org in the Quantum Physics section. This process may
entail further revision of the essays by each author to address the comments of the evaluators
prior to inclusion in the final book. All essays must be written in or converted to LATEX prior to
inclusion in the final book form.