Student Seminar reports

Quantum Computation and Quantum Information Course

PH584
IIT Madras Chennai 600036,
Inst: Dr. Arvind
Jan-April 2006

May 7, 2006
Contents

1 Quantum Zeno Effect 6

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2 Main Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2 Quantum Bit Commitment and Quantum Coin Flipping 10

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 Main Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.1 Bit Commitment Problem . . . . . . . . . . . . . . . . 11
2.2.2 Coin Flipping . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.3 Realizing Coin Flipping using Bit Commitment . . . . 13
2.3 Quantum Bit Commitment and Quantum Coin Flipping . . . 13
2.3.1 An Attempt at Quantum Bit Commitment - BB84
scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3.2 General Model of Quantum Bit Commitment with
One-way communication . . . . . . . . . . . . . . . . . 14
2.4 Impossibility of Unconditional Quantum Bit Commitment . . 15
2.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

3 Positive Operator Value Measures 18

3.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2.1 UQSD–Unambiguous Quantum State Discrimincation 19
3.2.2 Optimal distinguishing strategies . . . . . . . . . . . . 20
3.3 Neumark’s Theorem . . . . . . . . . . . . . . . . . . . . . . . 21
3.4 The state after a POVM measurement . . . . . . . . . . . . . 22
3.5 Information gain from a measurement . . . . . . . . . . . . . 23
3.6 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

1
4 A Thermodynamic Approach To The Measure of Entangle-
ment 24
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.2 Entanglement of a Pure State . . . . . . . . . . . . . . . . . . 25
4.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

5 Bell’s Inequalities For Mixed States 28

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5.2 Bell’s Inequalities for Pure states . . . . . . . . . . . . . . . . 28
5.3 A More General Approach to Measurements . . . . . . . . . . 29
5.4 The Example . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.6 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

6 Approximate Quantum Cloning 32

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
6.1.1 No-Cloning Theorem . . . . . . . . . . . . . . . . . . . 33
6.2 Input state dependent QCM . . . . . . . . . . . . . . . . . . . 33
6.3 Input state independent QCM . . . . . . . . . . . . . . . . . . 35
6.4 QCM In the nighborhood of a given state . . . . . . . . . . . 36
6.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
6.6 REFENENCES . . . . . . . . . . . . . . . . . . . . . . . . . . 37

7 Single Qubit and CNOT Gates are Universal 38

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
7.2 Two-level unitaries are universal . . . . . . . . . . . . . . . . 38
7.3 Single qubit and CNOT gates are universal . . . . . . . . . . 40

8 Realization of Physical Quantum Computer 42

8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
8.2 Difference between Classical and Quantum Computer . . . . 43
8.3 Requirements and Difficulties . . . . . . . . . . . . . . . . . . 43
8.3.1 A set of qubit: . . . . . . . . . . . . . . . . . . . . . . 43
8.3.2 A set of Quantum Gates . . . . . . . . . . . . . . . . . 44
8.3.3 Detection . . . . . . . . . . . . . . . . . . . . . . . . . 44
8.3.4 Scalability . . . . . . . . . . . . . . . . . . . . . . . . . 44
8.3.5 Erase . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
8.4 Quantum Computers . . . . . . . . . . . . . . . . . . . . . . . 44
8.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
8.6 Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

2
9 Interferometry Using Large Matter Particles 47
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
9.1.1 Decoherence . . . . . . . . . . . . . . . . . . . . . . . . 48
9.2 Main Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
9.2.1 Test particle . . . . . . . . . . . . . . . . . . . . . . . 49
9.2.2 Experimental Set-Up . . . . . . . . . . . . . . . . . . . 49
9.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
9.4 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

10 Experimental test of quantum non-locality in three photon

GHZ entanglement 53
10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
10.2 Main Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
10.2.1 Method we are using . . . . . . . . . . . . . . . . . . . 54
10.2.2 How we do this? . . . . . . . . . . . . . . . . . . . . . 54
10.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
10.4 Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

11 Nonlocality for Two Particle Systems 59

11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
11.2 The Basis Sets and The Operators . . . . . . . . . . . . . . . 60
11.3 Proof of Nonlocality . . . . . . . . . . . . . . . . . . . . . . . 62
11.4 The Caveat . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
11.5 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

12 Reducing the complexity of reduction 64

12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
12.2 Complexity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
12.2.1 NP-completeness . . . . . . . . . . . . . . . . . . . . . 65
12.3 Reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
12.3.1 General defination of reduction . . . . . . . . . . . . . 66
12.4 Circuit family . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
12.5 Different kinds of reduction . . . . . . . . . . . . . . . . . . . 67
12.6 Main study . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
12.7 conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
12.8 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

13 Quantum Random Walk 69

13.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
13.2 Main Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

3
 13.2.1 The Classical Random Walk . . . . . . . . . . . . . . . 71
13.2.2 The Quantum Random walk . . . . . . . . . . . . . . 71
13.2.3 Random Walks in Computer Science . . . . . . . . . . 72
13.2.4 Decoherence . . . . . . . . . . . . . . . . . . . . . . . . 75
13.2.5 Open questions . . . . . . . . . . . . . . . . . . . . . . 75
13.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
13.4 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

14 Optimality of Grover’s algorithm 77

14.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
14.2 Main study . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
14.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
14.4 Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

15 RSA Cryptography 82
15.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
15.2 Main Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
15.2.1 Motivation and Procedure . . . . . . . . . . . . . . . . 83
15.2.2 Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . 83
15.2.3 Attacks on RSA . . . . . . . . . . . . . . . . . . . . . 84
15.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
15.4 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

16 Experimental Tests of Bell’s Inequality 86

16.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
16.2 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

17 Quantum computation by anyons 91

17.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
17.2 Toric codes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
17.3 Abelian anyons . . . . . . . . . . . . . . . . . . . . . . . . . . 93
17.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
17.5 Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

18 Privacy Amplification by Public Discussion 95

18.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
18.2 Main Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
18.2.1 Detection of transmission errors and tampering . . . . 96
18.2.2 Reconciliation . . . . . . . . . . . . . . . . . . . . . . . 97
18.2.3 Reduction of eavesdroppers information . . . . . . . . 97

4
 18.2.4 putting the concepts together . . . . . . . . . . . . . . 99
18.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

5
Chapter 1

Quantum Zeno Effect

R.Bharath, AE02B005

Abstract
In the field of quantum measurement theory, there is a seemingly paradoxi-
cal result obtained due to continuous measurements. This result is called as
Quantum Zeno effect. In this discussion we shall look at this in the light of
classical zeno’s paradox. Following this we shall go through an example in a
single qubit system to realise the notion in the light of this simple example.
Following this we shall look at the first discussion of the zeno’s paradox by the
authors B.Misra et al.

1.1 Introduction
We shall start the discussion on Quantum Zeno Effect by looking at the
classical zeno paradox which forms the logical basis for the Quantum Zeno
Paradox. Zeno’s set of paradoxes were put forth by Zeno of Elea to support
the doctrine of his teacher Paramenides, on the doctrine of plurality and
the very idea that ’All is one’ and our very intutional notion of change and
dynamics is mistaken and is nothing more than an illusion. There are many
paradoxes put forth by Zeno but the three most famous ones are - Achilles
and the Tortoise, the Dichotomy Argument and the Arrow in Flight. For
our discussion here the Paradox of an Arrow in Flight is the most relevant
one. We shall try to have a better understanding of the Arrow Paradox.

The Paradox of Arrow focusses on the instantaneous physical properties

of a moving arrow. Here in this case Zeno observes that if physical objects
exist discretely at a sequence of discrete instants of time and if no motion

6
occurs in an instant, then one may very well conclude that no motion occurs
at all in any given instant. This argument can very well be extended to the
context of continuous motion, wherein in any single slice of time there is no
physical difference between a moving and a non-moving arrow. Here Zeno
suggests that if all time is composed of instants (or ’nows’) may it be con-
tinuous or discrete, and if motion cannot exist in any instant, then motion
cannot exist at all. This argument against motion rests on the assumption
that time is composed of indivisible instants and nothing can move in this
instant. There are counterarguments to solve the Zeno’s Arrow paradox,
but we shall not get into it as it has very little relevance to Quantum Zeno
Effect. The bottomline is that our act of determining the exact position of
the arrow causes the arrow to freeze in time and the whole idea of velocity
makes no sense since an instant is of 0 seconds and according to the argu-
ment by zeno the arrow travels 0 distance,hence the very notion of velocity
and hence motion is compromised for the act of determing the position of
the arrow.
On similar lines is the quantum zeno paradox wherein we get a seemingly
paradoxical result due to continuous measurements on a temporally evolving
quantum system, wherein it is observed that the system does not evolve at
all in the light of the continuous measurements.

1.2 Main Study

Consider a single qubit system wherein a qubit |1i evolves temporally to an
orthogonal state by |0i by the time evolution operator exp(-iHT), where H
the hamiltonian is taken as the projection operator k(|0i+ |1i)(h0|+ h1|) and
T is the total time taken for the evolution to take place. If suppose we divide
the time interval T into N equal parts equal to dt, and then after letting
the system to evolve for dt we measure the state of the system, preferably
in the Z basis. we continue doing so for the whole of the n time intervals
of dt each, and after each measurement we determine the probability of the
system being in state |1i (or |0i ). To determine the probability after n
measurements for the system to be in the given state (|1i or |0i) we add the
probabilities we obtain after each measurement on a given eigenstate one
obtains in the previous measurement. Then we idealize the measurements
to being continuous by taking the time interval between every measurement
to tend to zero, and hence the number of measurements to tend to infinity
(n tends to infinity). The value of this probability should lie between [0,1].
For the sake of completeness a few details of calculations done for this sake

7
shall be discussed here.

exp(-iHdt) is reduced to the linear form ′ 1− iH ′

k sin(kdt) exp(−ikdt) and this
is operated on the prevalent states of the system. On operating the time evo-
lution operator on |1i we get x|0i+(1−x)|1i where x = i H k sin(kdt) exp(−ikdt).

We get the probabilities of obtaining state |0i as |x|2 = sin2 (kdt) and state
|1i as |1 − x|2 = cos2 (kdt). we now branch from each of the eigenstates and
in turn let the states evolve for dt and we measure the states in Z basis.
Continuing the branching in this fashion we evaluate the probabilities and
sum up all the probabilities. The series expression for the probability of
obtaining |1i after N measurements is as follows :
2N
sin2 (kdt) (1−sin (kdt))
cos2 (kdt) − cos2 (kdt) sin2 (kdt)(1 − sin2N −2 (kdt))
2N−2
+cos4 (kd ∗ t)( 1−(2sin(kd∗t))
cos(2kd∗t) )

On evaluating the limit of the above expression as dt → 0 ,we obtain the

value of the probability as 1. Hence from this it is evident that the quantum
system has not evolved from the state |1i at all in the light of the infinite
(and hence continuous) measurements made on the system. This gives a
glimpse of the Zeno’s paradox in a single qubit quantum system.

In the context of the first ever discussion on QZE by Misra et al, we find
that we must first seek to understand the operational meanings of continu-
ous measurements. For this sake the authors have defined the probabilities
P,Q,R for determining the outcomes of continuously ongoing measurements
in a given interval of time, and go on to say that we may consider the pro-
cess of continuous measurements to be the limiting case of successions of
practically instantaneous measurements. Here Q represents the probability
that the system prepared in an initial undecayed state at time 0 is found
to decay during a given interval. Through a series of rigorous mathematical
arguments the authors prove that Q = 0, and hence the Quantum Zeno
Paradox from the formalisms of Quantum mechanics. The authors finally
make the following conclusions of which they speculate that either of them
might hold true in the light of the Zeno’s Paradox in Quantum theory they
evoked through the formalisms of Quantum mechanics and mathematics.

1. Probabilities like Q do not have any operational meaning since there

is a speculation that Quantum theory denies the possibility of continuous

8
measurements
2. Measurements in real life are non-ideal, whereas zeno’s paradox is
based on the assumpt ion that measurements are ideal. And this very fact
might render the mathematical for malisms put forth to realize the effects
of the paradox might become invalid.
3. It may be wrong to assume the temporal evolution of a quantum
system can be described as a linear operator in time evolution. It might be
best described only in terms of persist ent interactions between the quantum
system and the classical measurement apparatus. when this is attempted
then the paradox might disappear.
4. Quantum mechanics is incomplete since it is not able to satisfactorily
explain the Paradox completely.

1.3 Conclusions
In the light of research done by quantum theorists in the recent past, the
conclusion is that there is no paradox in the phenomenon of QZE since
the quantum system is overwhelmed and in a way paralysed due to the
continual interactions of the measurement apparatus with the system, thus
not allowing the system to temporally evolve. Thus the quantum system
remains in its initial state in the light of continuous observations.

1.4 References
”Zeno’s Paradox in Quantum Theory” , Journal of Mathematical Physics,
Vol. 18, No. 4, April 1977

9
Chapter 2

Quantum Bit Commitment

and Quantum Coin Flipping

Aravindan, CS03B002

Abstract
The Bit Commitment and Coin Flipping problems are first introduced and
examined their relation is also discussed. Then the Quantum variants of the
problems are discussed with an example. Then, perfect Quantum Bit Com-
mitment are individually proved to be impossible. A sample Quantum Bit
Commitment protocol is introduced and finally,some difficulties and general
observations are made regarding the two problems.

2.1 Introduction
Quantum Cryptography is an applied field of theoretical Computer Sci-
ence, where Quantum Mechanics is used to arrive at solutions for various
problems in Cryptogrpahy, dealing with security and privacy. Most of Clas-
sical Cryptography that is practised today are based on assumptions re-
garding the computational difficulty of cerrtain problems like the Discrete
Log Problem, Prime Factorization and the Quadratic Residuosity problem.
Classical Cryptogrpahy has been surviving for the last forty years or so only
because there weren’t any good algorithms to solve these problems. But
with the advent of efficient Quantum Algorithms for Factorization [1] and
all Hidden Subgroup Problems, there is now serious cause for concern among
Cryptographists. However, without the advent of the Quantum Key Disri-
bution protocol [2] people have also started coming up with protocols and

10
methods to perform Cryptography.

The Quantum Coin Flipping and Bit Commitment problem are such prob-
lems, with the the classical Coin Flipping having been first introduced by
Blum[3]. In this problem, two parties try to agree on a particular value(bit),
without any prior information, and with no direct communication. These
problems find a number in of applications like Oblivious Transfer and Secure
Mutliparty Computation(eg. e-voting) , in which more than one party try
to communicate and compute some particular function securely without any
of their individual data getting revealed. In this paper, in the first section,
we will look at a detailed description of the Bit Commitment and the Coin
Flipping problems and also look at how they are related. The second section
contains a study of the Quantum variants of the problem and we will also
look at some sample protocols for these problems.In the next section we will
go on to show the Impossibility of Unconditional Bit Commitment. In the
last we will take one of the Weak Coin Flipping protocols as a case study
and we will finally look at a few concerns and difficulties in this field.

2.2 Main Study

2.2.1 Bit Commitment Problem
In the Bit Commitment problem a party Alice wants to communicate a bit
b (say the result of a coin toss) to Bob. Alice doesn not send the bit as
such but she sends something like a certificate or commitment that implies
b and commits to it. Both the parties wants the following conditions to
hold. From the commitment that Alice has sent, Bob must not be able to
infer what Alice has sent. Further, Alice must not be able to change her
commitment to some other bit, once it has already been sent. This problem
can be thought of as a scenario in which Alice writes her bit on a piece of
paper, locks it inside a safe and gives the safe to Bob and keeps the key with
herself. Now we set to formally define the problem and the general protocol
governing it.

• Alice decides on a bit b to commit to.

• Alice computes cb =Commit(b) and sends it to Bob.

• Later, Alice may choose to reveal the bit b by sending Reveal(b).

• Bob verifies that Alice had originally commited to Bob only.

11
Perfect Concealment:
Protocols having the perfect concealment property guarantee that even with
the iknowledge of cb , Bob gets no idea about b. Mathematically, if b’ is the
bit Bob guesses,
P rob(b′ = b|cb ) = P rob(b = b′ )
For Computational Concealment, we relax the requirement by stating that
any machine which has computationally (polynomially) bounded resources
should not be able to guess b any better.

Perfect Binding:
Protocols having perfect binding property guarantee that once Alice has
committed to a bit b, she can’t alter it. If b′A is the bit revealed by Alice
finally,
P rob(b′A 6= b) = 0
Protocols having Computational only dictate that computationally bounded
participants should satisfy the specified conditions.
It has been proved that protocols that are both Perfectly Binding and
Perfectly Concealing are impossible classically. We will show in the next
section that Perfectly Secure Bit Commitment is impossible Quantum me-
chanically too. Classically, perfectly binding and computationally hiding or
concealing have been realised using the RSA assumption and the Discrete
Log Assumption. Using Quantum mechanics it is easier to realize Conceal-
ment than Hiding.

2.2.2 Coin Flipping

In the Coin Flipping problem two parties Alice and Bob want to decide on
a coin toss without meeting with each other or involving any third party.
Consider a scenario when Alice and Bob have just been divorced and they
want to decide with their pet dog which both of them don’t want to part
with.At the same time they don’t want to meet or speak directly with each
other. They want to be able to ensure that no one cheated in the exchange.
Naive attempts like asking one of them to toss a coin and conveying the result
is not secure as the person tossing the coin can always claim he got the other
result and so on. Moreover, using entangled particles like Bell states, with
which one of them measure a bell qubit and the other’s is automaticallly
obtained is also not enough as this relies on a trusted Third party.

12
2.2.3 Realizing Coin Flipping using Bit Commitment
A secure Bit Commitment protocol can be used as a black box to realize a
secure Coin Flipping protocol as shown below. Let the Commitment phase
be represented by Commit and the revealing phase by Reveal.

• Alice decides on a bit bA and commits to it Commit(bA ).

• Alice sends Commit(bA ) to Bob.

• Bob chooses a bit bB and sends it to Alice.

• Alice reveals the bit bA by Reveal(bA .

• Alice and Bob decide on bA ⊕ bB

2.3 Quantum Bit Commitment and Quantum Coin

Flipping
The Quantum Bit Commitment and Quantum Coin Flipping are essentially
their Classical counterparts with the added power of Quantum Mechanics to
come to our aid. It is to be remembered that we are still trying to commit
and decided on abit and notonaqubit. At first glance, it looks like if secure
bit commitment is impossible, so is bit commitment.

2.3.1 An Attempt at Quantum Bit Commitment - BB84

scheme
This scheme is very similar to the Quantum Key Distribution (QKD) pro-
tocol proposed by Brenett and Bassard. We have a set of two mutually
unbiased bases Q0 and Q1 where Q0 is the standard perpendicular x-y like
basis. Q1 is a mutually unbiased basis to Q0 . The protocol is as given
below:

• Alice chooses a bit to commit to and picks the basis Q = Qb to encode

information. Commitment:

• Alice encodes a random w1 , w2 , ...wn in Qb basis and sends it to Bob.

• Bob picks up random set of basis Qi (as he is not aware of Alice’s

bases) and measure the encoded qubits in this basis as Wi . Revealing

13
 • Alice announces b and wi and Bob verifies if ∀i

Qi = Qwhenwi = Wi

Concealing Property: It is easy to see that this protocol works for an

honest Alice and Bob. To prove security against a dishonest Alice, we prove
the Concealement property by showing that ρB B
0 = ρ1 .i.e. we show that the
reduced density matrix in the Bob-system is same for both b=0 and b=1,
which means that bob can’t distinguish between b=0 and b=1. For Binding
property , Alice needs to guess wi for all Qi =QB which is negligible.
But the flaw in the proof was found only in ’94 when Alice got over the
Binding property by using Entangled photons. A similar construction will be
used to prove Impossibility of the general model, so we will cover this later.
We will now first define a General model for Quantum Bit Commitment,
and our impossibility will hold for all protocols in this model.

2.3.2 General Model of Quantum Bit Commitment with One-

way communication
• Alice and Bob, before the execution of the protocol, agree on two
| 0i and | 1i ∈ HA ⊗ HB where | 0i = ni αi | ei i⊗ | φi i and
P
states P
| 1i = ni βi | e′i i⊗ | φ′i i
i.e both Alice and Bob know | 0i and | 1i. Here | ei i are orthogonal
but | φi i are not necessarily so.

• An honest Alice makes a measurement on the first register (HA ) of | bi

where b = 0 or b = 1.

• Alice sends the second register to Bob.

• Alice opens commitment by sending b and i or j as the case may be.

• Bob performs measurements on the second register to verify Alice’s

claim.

This model proposed by Lo and Chau [5] covers almost all the protocols
that can be devised. Further the proof for impossibility presented below
can be easily extended to a very general model [8]. For example, in the
BB-84 based protocol,
1
| 0i = √ (| 0i⊗ |→i+ | 1i⊗ |↑i)
2

14
 1
| 1i = √ (| 0i⊗ |րi+ | 1i⊗ |տi)
2
. Further, the measurement is performed before the commitment is opened
which makes the adversary weaker and hence this does more than enough.
Hence this model suffices for BB-84 model.

2.4 Impossibility of Unconditional Quantum Bit

Commitment
Here, we split the proof into two parts. First we prove that realizing perfect
concealment and binding simultaneously is impossible. Then we show that
even if the Concealement is not perfect, perfect binding is not possible.

Perfect Concealment
Assume that perfect concealment exists. Hence Bob does not know whether
Alice sent qubits corresponding to b=0 or b=1 at commit stage. Therefore,
ρB B
0 ρ1 . i.e
T rA | 0ih0 |= T rA | 1ih1 |
By Schmidt decomposition,
p
| 0i = σk λk | sk iA ⊗ | θk iB
p
| 1i = σk λk | s′k iA ⊗ | θk′ iB
Since the system is different only in HB , a local Unitary operation on system
B, UB s.t UB | 0i =| 1i can be used to transform between the two qubits
and hence cheat by delaying her commitment from the Step 2 to Step 4.
Therefore Binding property is not satisfied.

Not Perfect concealment

Let F (ρb0 , ρB
1 ) = max|ψ0 i,|ψ1 i | < ψ0 |ψ1 > | and 0 ≤ F ≤ 1. We use two
results to get the required result:

• For every purification | ψ1 i of ρB

1 , we have a purification | ψi s.t
| < ψ|ψ1 > | = F

• any two purifications of ρB can be converted to another by unitary

transformations.

15
 Cheating Strategy:

• She prepares the state in | 0i.

• She commits only in Step 4.

• If she wants 0, she doesnt make any change.

• If she wants 1, she applies UA and gets | ψ0 i s.t | < ψ0 |1 > | = 1 − δ.

Since | 1i and | ψ0 i are very similar, she can easily cheat.

Hence the Binding property is violated. Thus by using an entangled sys-

tem, and delaying the measurement till Step 4 , we are able to change our
commitment. Hence Quantum Bit Commitment is impossible.

2.5 Conclusions
Here in this article, we have introduced the twin problems of Bit Commit-
ment and Coin Flipping and their Quantum counterparts as well. We then
gave a construction for Quantum Bit Commitment protocols and also dealt
with the Impossibility proofs. In Quantum Bit Commitment, Concealing
property is easier while Binding becomes tougher because of Entanglement.
This is just the opposite for Coin Flipping protocols. Further, even if the re-
quired Quantum protocols do exist, security is not fully guaranteed by their
use, since a public verifiable certificate or proof is not available to the out-
side world. Whereas in clasical cryptography, the classical communication
is available for all to seeand verify , this is not the case with Quantum Cryp-
tography since by No Cloning theorem copies can’t be created. In this case,
the defeated party can always claim that the interaction never took place.
Hence the protocol should guarantee that sufficient classical communication
should take place so that is avoided .

16
Bibliography

[1] The RSA Cryptosystem by Rivest,Shamir and Alderman.

[2] C. H. Bennett and G. Brassard, in Proceedings of IEEE International

Conference on Computers, Systems, and Signal Processing (IEEE, New
York, 1984), p. 175.

[3] Blum : The Bit Commitment Problem

[4] Lo, Hoi-Kwong, and H. F. Chau, ”Is quantum bit commitment really p
ossible?”, Phys. Rev. Lett. 78, (1997) 3410-3413.

[5] Lo, Hoi-Kwong, ”Insecurity of Quantum Secure Computations”, Phys.

Rev. A56, (1997) 1154-1162.

[6] D. Mayers, following Letter, Phys. Rev. Lett. 78, 3414 (1997).

[7] H.-K. Lo and H. F. Chau, in Proceedings of the Fourth Workshop on

Physics and Computation, PhysComp’ 96, Boston, 1996 (New England
Complex Systems Institute, Boston, 1996), p. 76.

[8] Hoi-Kwong Lo and H. F. Chau in arXiv:quant-ph/9711065 v1 26 Nov

1997: Why quantum bit commitment and ideal quantum coin tossing are
impossible.

17
Chapter 3

Positive Operator Value

Measures

Harshavardhan S, CS03B021

Abstract
No abstract has been provided!

3.1 Definitions
According to the third postulate ofP
Quantum Mechanics, a measure-
†
ment is a set of operators {Mm } (where m Mm Mm = I, m indicating the
output of the experiment), which when used to ”measure” a state |ψi would
†
yield an outcome m with a probability pm = hψ|Mm Mm |ψi, and produce a
Mm |ψi
resultant state pm . From the definition, it follows that pm = 1.
√

A more specific version of the above defined generalised notion of a

Measurement called the Projective Measurement or Von Neumann Mea-
surement, which is realisable in experiment, is widely used in practice. A
Projective
P Measurement is defined by an observable (Hermitiam operator)
M = m mPm (m being the eigenvalues and Pm being the projectors into
the eigenspaces corresponding the eigenvalues m), which upon ”measuring”
the state |ψi, yields the state P√mp|ψi
m
with probability pm = hψ|Pm |ψi, m
being the measurement outcome. Projective measurements have nice prop-
erties like:
Pm Pn = Pm δmn (3.1)

18
 E(|ψi, M ) = hψ|m|ψi (3.2)
†
Suppose we replace Mm Mm term in the generalized measurement no-
†
tion with Em = Mm Mm (Note P that by definition {Em } is a set of pos-
itive operators such that m Em = I). Now, if we are not concerned
much with the output states of the measurement, we can do away with
the set {Mm } and instead use the set {Em } for obtaining the outcome
probabilities of measurements, pm = hψ|Em |ψi. This set {Em } consti-
tutes a POVM. In general any set of positive operators {Em } such that
P
m Em = I is a POVM, whose measurement outcomes have probabilities
hψ|Em |ψi. Also, given a set of POVMs {Em }, one can always define an
equivalent set of measurement operators such that the outcome probabil-
ities of both the sets of√ measurements
pP are same. Infact
P √ the measurement
operators set {Mm = Em = ξ |ψ
k k k ihψk | = k ξk |ψk ihψk |} is one
such construction.
√ Infact, any general measurement {Mk } can be expressed
as {Mk = U Ek } where {Ek } is a POVM and U is some unitary operator.
Coming to the relation between projective measurements and POVMs, the
effect of a projective measurement in a higher dimensional space is a POVM
in the lower dimensional space as shall be illustrated in section 2.2.

3.2 Applications
We now describe 2 applications of POVMs:

3.2.1 UQSD–Unambiguous Quantum State Discrimincation

The problem of UQSD may be stated as follows: given a state and also the
information that the state belongs to the set of states {| ψ1 i, | ψ2 i, ..., | ψn i},
which are not necessarily orthogonal, how does one unambiguously report
back the identity of the given unknown state? Ofcourse, one cannot solve
this problem with 100% success rate. Also, projective measurements give us
no means of accomplishing this task even with a non-zero failure probability.
However, one can construct a POVM that will help establish the identity
of the given state without any ambiguity with a significant probability of
success, with the provision made that the technique has a non-zero failure
probability, in which case, one cannot reliably ascertain the identity of the
state. But in the success scenario of the POVM measurement, one can be
100% sure about the identity of the original state.

19
 Consider the set S = {| ψ1 i, | ψ2 i}. The POVM {E1 = I−|ψ 1 ihψ1 |
norm. , E2 =
I−|ψ2 ihψ2 |
norm. , E3 = 1 − E1 − E2 } accomplishes the UQSD task for the set S. If
one makes a measurement of a state that is a member of set S and gets the
output 1, one can be sure that the state was | ψ2 i. On the other hand, if
the output was 2, then the initial state was | ψ2 i without any doubt. But
on the other hand, if the output is 3, then one cannot ascertain the identity
of the given state.

In general, given a linearly independent set S = {| ψ1 i, | ψP 2 i, ..., | ψn i},

the set of POVMs E = {Ej =| ψj⊥ ihψj⊥ |}, where | ψj⊥ i = i cji | ei i, |
ψj i = i dji | ei i, C = [cji ]ji , D = [dji ]ji , C = D † achieves the UQSD task.
P
Also, USQD can be acheived only for a linearly independent set of states.

However, note that a POVM is in general not the optimal strategy of

obtaining information about a state if the amount of information gain is
the criteria (For a note on the quantification of the information gain from
measurements, refer the last section). For instance, consider the states
{ψ1 = cosθ | 0i + sinθ | 1i, ψ2 = sinθ | 0i + cosθ | 1i}. The observable
M = |ψ1 ihψ1 | − |ψ2 ihψ2 | gives us the maximal information about the state
through a projective measurement. The problem of completely optimizing
the information gain from a measurement, given apriori probability infor-
mation is difficult and remains only partially solved.

3.2.2 Optimal distinguishing strategies

Suppose we are given a set of state {(ρi , pi )} and we are to distinguish
between them with optimal information gain. To start with, we might ex-
tend the Hilbert space of the system by introducing an auxiliary ρaux and
considering projective measurements {Pk } in the extended space of states
{ρi ⊗ ρaux }.

Pki = P T r[Pk (ρi ⊗ ρaux )]

= mr,ns (Pk )mr,ns (ρi )nm (ρaux
 )sr
P P (3.3)
= mn rs (Pk )mr,ns (ρaux )sr (ρi )nm
= T r(Ak ρi ),

where X
(Ak )mn = (Pk )mr,ns (ρaux )sr . (3.4)
rs

20
 P P P
Ak =
k ( k Pk )mr,ns (ρaux )sr
rsP
= P rs Imr,ns (ρaux )sr (3.5)
= rs Imr,ns (ρaux )sr δsr
= I.
P
Therefore, {Pk } = { r s(Pk )mr,ns (ρaux )sr } forms a POVM.
Here, the projective measurements in the extended space translated into
POVMs of the space of interest – which in general – is the case. A projective
measurement in a higher dimensional space appears in a lower dimensional
space as a POVM.

Coming back to original problem, the problem of completely optimizing

the information gain from the measurement has not yet been solved fully.
However, several partial results have been obtained by E.B.Davies, which
are as follows. The optimal measurement strategy has POVMs of the form
Ak =| ψihψ |, where | ψi is not necessarily normalized. Also the cardinality
of the set of POVMs, n, satisfies the inequality d ≤ n ≤ d2 , where d is the
dimensionality of the space of ρi . Also, the maximum
P information
P gain that
can be obtained can be bounded by: Iav ≤ S( i pi ρi ) − i pi S(ρi ).

3.3 Neumark’s Theorem

Having seen the above relationships between different kinds of measurement
notions, one naturally arrives at the question of whether any POVM can be
realised as a projective measurement in some higher dimensionsal space, of
which the space of interest is a subspace. Neumark theorem states that the
above fact is indeed true. The following is a proof of the theorem for the
case where Ak are of the form Ak =| ψk ihψk |, k = 1 → N, dim(H) = n.
The space H is now extended to an N -dimensional K, by adding (N − n)
additional vectors. That is, we add (N − n) additional vectors vs satisfying
hvs |vt i = δst , hvs |ψk i = 0Pto the existing | ψr i so that the we give the set of
vectors, | wi i =| ψi i + N j=n+1 cij | vj i. These set of {| wi i} help us form
a projective measurement {|wi ihwi |} if and only if the set P of vectors {| wi i}
form an orthonormal set, that is if hwi |wj i = hψi |ψj i + N ∗ c
k=n+1 ik jk = δij ,
c

21
 Pn ∗
PN ∗
i.e., if l=1 ψil ψjl + k=n+1 cik cjk = δij .
 
ψ11 ψ12 ... ψ1n c1,n+1 c1,n+2 ... c1N
 ψ21 ψ22 ... ψ2n c2,n+1 c2,n+2 ... c2N 
⇐⇒ M = 
 
.. .. .. .. 
 . . . . 
ψN 1 ψN 2 . . . ψN n cN,n+1 cN,n+2 . . . cN N
(3.6)
is such that M M † = I.

P It is easy to construct PM such that M † M = I, because we already have

∗
k | ψk ihψk |= I, i.e., k ψki ψkj = δij . Having noticed this, we have in-
finitely many ways to choose | vl i, in order to make M † M = I (One may
use the Graeme-Schmidt orthomormalization process for this).

Given that M † M = I, we have for any | ψi,

| ψi = M (M −1 | ψi) = M (M † M )(M −1 | ψi)

= (M M † )(M M −1 ) | ψi) = (M M † ) | ψi. (3.7)

Therefore, M † M = I ⇒ M M † = I. Having constructed the set of projective

measurements {|wi ihwi |}, we notice that the POVM {Ai } is the projection
of this operator set in the lower space, Ai = P (|wi ihwi |)P † (P being the
projector into the subspace), and the effect of the constructed projective
measurement in the subspace of interest is equivalent to the effect of POVM
at hand. And hence the proof.

3.4 The state after a POVM measurement

Suppose we make a POVM {E √ m } on a state ρ. As we noticed earlier, the
measurement set {Mm = U Em } gives us this POVM for any choice of
unitary operator U . We have this choice of U , because we may choose to
implement a POVM by any confirming set of projective measurements. The
state of√ the √system after we perform the POVM on it and obtain outcome
†
m is U Empρm Em U . So the final state of the system would be
√ √
′
X U Em ρ Em U † Xp p
ρ = pm = U( Em ρ Em )U ′ (3.8)
m
pm m

22
3.5 Information gain from a measurement
Given a set of states {ρi } with apriori probabilities {pi } and a measurement
strategy which yields the output µ with probability Pµi if the initial state is
ρi , the probability Qiµ that the initial state was ρi , provided that the output
is µ is
Piµ pi
Qiµ = P (3.9)
j Pjµ pj

Having found the output µ, the uncertainity about the preparation is

X
Hµ = − Qiµ logQiµ (3.10)
i

The uncertainity after the measurement is

X
hHµ i = qµ Hµ , (3.11)
µ

where qµ is the probability of outcome µ. Finally, the information gain from

the experiment is X X
I =− pi logpi − qµ Hµ (3.12)
i q

3.6 References
[1] Asher Peres, Quantum Theory - Concepts and Methods.
[2] Nielsen, Chuang - Quantum Computation and Quantum Information.
[3] Preskill, Lecture notes for the Quantum Computation Course, Caltech.

23
Chapter 4

A Thermodynamic Approach
To The Measure of
Entanglement

Auditya Sharma, EP02B004

Abstract
Here we present arguments inspired from Thermodynamic principles to show
that the von Neumann entropy is a measure of entanglement for a bipartite
system in a pure state.

4.1 Introduction
Quantum Entanlement is a remarkable phenomenon which finds applications
in quantum information processing. A good understanding of entanglement
is vital for the growth of the field of quantum information theory. A signif-
icant problem without conclusive answers is the question of what measure
of entanglement to associate with a given state of a system. Many ad-hoc
measures have been suggested for different cases but no conclusive mea-
sure for the most general state has been agreed upon. Here we look at a
pure state and present arguments to show that the von-Neumann entropy
of the reduced density matrix is the measure of entanglement for a bipartite
system.

24
4.2 Entanglement of a Pure State
There is a lot of similarity between the principles of Thermodynamics and
Entanglement. This is because of inherent similarities in the way the quan-
tities are defined in the two fields. Also there is an inherent similarity at
a more formal level. There is a principle that Entanglement obeys which
is the analogue of the second law of Thermodynamics. The second Law of
Thermodynamics says that it is impossible to construct a perpetual motion
machine. This directly implies that two reversible engines operating between
the same temperatures are equally efficient. The formal correspondence with
entanglement is as follows: The laws of nature are such that it is impossible
to increase the entanglement between remote systems by local operations
alone. Here local operations include unitary operations, measurement and
classical communication between the two remote systems. The analogue of
a reversible engine here is a reversible transformation. Suppose Alice and
Bob are two experimenters who are separated in space. Suppose they carry
between them k entangled pairs of qubits, with one member of each pair
with Alice and the other with Bob. Additionally, let us also suppose that
there are further pairs of qubits at the disposal of our experimenters. Also
each of the k pairs that are entangled are identical in the state of entan-
glement. Now suppose Alice and Bob perform local reversible operations to
transfer the entanglement to n pairs. Even if n ≥ k there is no violation of
the above law because the amount of entanglement in each of the n pairs
may now be less than what was there before. Now, we put forth the ques-
tion: Is it possible to that in the process entanglement has got destroyed? In
′
other words, is there is superior transformation, which would require only k
′
pairs, where k ≤ k to finally give the final state of the n pairs obtained by
the previous transformation? We can invoke our principle here to show that
′
this is not possible. For, if such a k did exist, then we could perform an
′
initial transformation from the k pairs to the n pairs and then, use the other
transformation (since it is reversible) to arrive at k pairs. So effectively, by
the end of this procedure we have succeded in increasing the entanglement
′
from k entangled pairs to k pairs finally, thus violating the principle stated
′
above. Therefore, k = k . So, from the discussion above, we have shown
that the entanglement content of the k pairs is equivalent to the entangl-
ment content of the n pairs obtained by local reversible transformations.
The reversible transformations we have considered are in fact, consistent
with quantum mechanics. Bennett, Bernstein, Popescu and Schumacher [2]
have shown that it is possible, with local operations only to transform k

25
 ′
E
systems in an entangled state ψAB into n systems in a different entangled
state |ψAB i. The transformation is reversible when the number of systems
becomes arbitrarily large. That is, the ratio, n/k is a constant in the limit
k → ∞. And without loss of generality, we may choose the final n pairs to
be the singlet states which are all maximally entangled. So, our problem
effectively reduces to finding the entanglent content of n pairs of singlets.
Next, we ask the question: what function of n would be a suitable measure
of the entanglement content of the n singlets. Again we appeal to thermo-
dynamic principles for the answer. Reversibility requires us to go to the
limit of infinite n, and for infinite n, there is no way we can define the en-
tanglement content. So we will have to define the entanglement per system.
We borrow the formal principle from thermodynamics: the thermodynamic
limit requires us to define intensive quantities. Therefore we must have en-
tanglement of n singlets proportional to n. Therefore, we can say that the
entanglement of the k systems in the state |ψAB i is equal to the entan-
glement content in n systems in the singlet state |SAB i. That is, we have
kE(|ψAB i) = nE(|SAB i), where E denotes the entanglement measure. Thus
n
E(|ψAB i) = lim E(|SAB i) (4.1)
n,k→∞ k
The proportionality constant E(|SAB i) simply defines a conventional limit
and we set it to 1. We have shown that the measure of entanglement for
pure states is got from the limit above. Now it remains to compute the
limit. Bennet, Bernstein, Popescu and Schumacher [2] have computed this
limit: E(|ψAB i) equals the entropy of entanglement of the state |ψAB i. The
entropy of entanglement is the von Neumann entropy of the reduced den-
sity matrix seem by either Alice or Bob, and equals the Shannon entropy
of the squares of the coefficients of the entangled state in the Schmidt de-
compostion[2]. The entropy of entanglement is zero for a pair of systems in
a product state and it is one for a pair of systems in the singlet state; it is
never negative. Also the measure is intensive as required from arguments
above. A qualitative understanding of this result may be obtained as fol-
lows. It is clear that both entanglement and entropy are very closely related
concepts because, each of these quantities, in a sense, is a measure of the
lack of information we have of the system. Now, from the point of view of
either Alice or Bob, the only information they have of their own subsystem
is contained completely in their respective reduced density matrix. This
reduced density matrix may be interpreted as a mixed state with different
probabilities for the different states in which the respective subsystem can
exist. The measure of entropy that may be obtained from this set of prob-

26
abities is the Shannon entropy and this can be reasonably taken to be a
measure of the entanglement of the system, because it is the measure of the
lack of information of the exact state of either sub-system. This is also the
von-Neumann entropy by definition.

4.3 Conclusions
We have first established, using arguments taken from the related field of
thermodynamics, a limit equation to obtain the entanglement of a pure
state. We have then provided qualitative arguments for the result that the
calculation of the limit leads to von-Neumann entropy as the measure of
entanglement for a bipartite entangled system.

References
1. S. Popescu, D. Rohrlich, arXiv: quant-phy/9610044

2. C.H Bennett, H.Bernstein, S. Popescu, and B. Schumacher, Phys. Rev.

A53, 2046 (1996)

3. Asher Peres, Quantum Theory: Concepts and Methods, Kuwer Aca-

demic Publishers.

27
Chapter 5

Bell’s Inequalities For Mixed

States

Supradeepa V. R., EP02B010

Abstract
A Seminar Report done as a part of the Course on Quantum Computation and
Quantum Information Theory

5.1 Introduction
Bell’s pioneering paper showed that, when measurements are performed on
two quantum systems separated in space then their results are correlated in
a manner, which is not explained by a local hidden variables model. The
main problem, that has not been answered even today is the classification
of Quantum states into local and non local ones. The case for pure states is
fully solved, but for mixed states only partial results are known. In the paper
that I reviewed, Sandy Popescu gives strong arguments for his contention
that we have to rethink the usual way we think about Bell’s Inequalities
when we talk of mixed states.

5.2 Bell’s Inequalities for Pure states

Before we go forward to address the main points, we would look at the Bell’s
Inequality for pure states again. Let us say Alice and Bob arbitrarily choose
to measure {Q,R} or {S,T} simultaneously, let us consider the expression

28
QS + RS + RT − QT = (Q + R)S + (R − Q)T = ±2.

If p(q, r, s, t) is the probability of (Q = q, R = r, ..), then

P
E(QS + RS + RT − QT ) = p(q, r, s, t)(qr + rs + st − qt) ≤ 2.

Hence, E(QS) + E(RS) + E(RT ) − E(QT ) ≤ 2.

|01i−|10i
Now let us consider |ψi = √
2

−Z√
2 −X2 Z2√
−X2
Let, Q = Z1 , R = X1 , S = 2
, T = 2
, then
√
E(QS) + E(RS) + E(RT ) − E(QT ) = 2 2

The above state violates the inequality. The basic question of which pure
states are classical and which are not can be answered without going be-
yond the usual scheme. Every entangled pure state violates some usual Bell
inequality and hence is non local. The only pure states which do not yield
non local correlations when a single ideal measurement is performed are the
Direct Product states for which no further analysis is required.

5.3 A More General Approach to Measurements

Bell’s proof and almost all subsequent proofs have a common aspect and that
is that they all consider the case, when a single ideal local measurement is
made on either of the systems. Ideal measurements as in measurements
which follow the postulates of Quantum mechanics, like measurements cor-
respond to Hermitian operators in a n dimensional Hilbert space, and the
only possible results of a measurement are the eigenvalues of the measured
operator. Such an approach however is very limited. We can start conceiving
more complex measurements, like for example a sequence of measurements.
The question that arises now is that will considering general local measure-
ments yield anymore information, the answer to this is Yes, and that will
be the subject matter for the rest of this report.

In case of pure states too, considering general measurements might yield

more information, but there is essentially no need for that. However this
changes dramatically when one starts talking about mixed states. Now we
will look into a situation where, when the two systems separated in space

29
are subjected to single ideal measurements, the corresponding results are
classical but when subjected to more general measurements the correspond-
ing correlations are non classical. In our case the more general measurement
is just two successive single ideal measurements.

5.4 The Example

The density matrices which we analyze are called the Werner Matrices. Con-
sider two systems separated in space, each system in a d dimensional Hilbert
space whose density matrix is given by
I (dxd)
+ 2 di≤j,i,j=1 |Sij ihSij |
P
d
W =
d2
where, |Sij i = |ii1 |ji2√−|ji
2
1 |ii2

An important property of these density matrices is that they cannot be

decomposed into mixtures of pure direct product states. As direct product
states are the only classical pure states, and as Werner matrices are not
mixtures of such states, one intuitively feels that these density matrices
are non classical. However Werner has proven that these matrices do not
violate any standard Bell inequality. Werner has given an explicit local
hidden variables model to simulate these correlations. But the question is,
is there a local hidden variable model which can simulate the results of any
measurements performed on the two systems prepared in such a mixed state.

Let us now look at a slightly general measurement on these Werner ma-

trices. First each particle is subjected to a measurement of a two di-
mensional projection operator, P = |1i11 h1| + |2i11 h2| on particle 1 and
Q = |1i22 h1| + |2i22 h2| on particle 2. After the direct measurement is per-
formed and the result is registered, an observer situated near particle 1
chooses at random to measure one of the operators A or A’, whose exact
form we will see. We note that the decision of which operator to measure
is taken only after the measurement of P. Similarly an observer situated
near particle 2 chooses randomly between a measurement of B or B’. This
scheme is identical to that used in the standard Bell’s inequalities except the
difference that, here A,A’,B,B’ are not measured directly on the particles in
their original state, but after the measurement of P and Q.

The operators A,A’,B and B’ each have three different eigenvalues 1, -

1, 0. The eigenvalues 1 and -1 are non degenerate and the corresponding

30
eigenstates belong to the subspaces {|1i1 , |2i1 } and {|1i2 , |2i2 }, respectively.
The eigenvalue 0 is highly degenerate and corresponds to the rest of the
Hilbert space. The nondegenerate part of these operators are choosen in
such a way that they yield maximum violation of the CHSH inquality for
the singlet state |S12 i. That is,
√
hS12 |AB + AB ′ + A′ B − A′ B ′ |S12 i = 2 2.
Now let us what happens if we subject an ensemble of pairs of particles in
the Werner state to the measurements described above. According to the
results obtained in the measurements of P and Q the original ensemble splits
into four sub ensembles given by {0,0}, {0,1}, {1,0}, {1,1}. The important
point we have to note here is, if the initial ensemble was classical, behaving
according to a hidden variable model, then each of the subensembles is
classical. But then we get a contradiction with the Quantum mechanical
predictions. The ensemble corresponding to P = 1 and Q = 1 is given by,
P QW QP 2d I (2x2)
W′ = = ( + |S12 ihS12 |)
N 2d + 4 2d
Where N is a normalization factor. In this state the CHSH inequality is
violated.
2d √
T rW ′ (AB + AB ′ + A′ B − A′ B ′ ) = 2 2 ≥ 2, f ord ≥ 5.
2d + 4

5.5 Conclusions
In conclusion, although a local hidden variable model can simulate all the
correlations which arise when only a single ideal measurement is performed
on each of the two particles, such a model cannot account for the correla-
tions which arise when two consecutive measurements are performed on each
particle. So when we have to talk of entanglement and locality for mixed
states then we also have to talk in terms of more general measurements.
This example clearly shows that a Bell inequality type of an approach is
inadequate to talk of entanglement properties of Mixed states.

5.6 References
1. Bell’s Inequalities and Density Matrices: Revealing ”Hidden” Nonlocality
- Sandy Popescu, PRL-2619-1995.
2. R.F.Werner - Phys Rev A 40, 4277(1989).
3. Nielson and Chaung - Quantum Computing and Quantum Information.

31
Chapter 6

Approximate Quantum
Cloning

K.V.S. Vinodh, EP03B006

Abstract
Vinoth has not supplied any abstract!

6.1 Introduction
Lets consider a quantum state |S >a in a two-level space with the basis
denoted by |0 >a and |1 >a , to be made copy of. The subscript ’a’ just
denotes that the state corresponds to original system which is to be copied.
|S >a is a linear combination of |0 >a and |1 >a which when fed into a
quantum-copying machine(QCM) in addition to |S >a , will give copy |S >b
identical to original state— Qualitatively a QCM can be understood as

|S >a |K >b |Q >x −→ |S >a |S >b |Q′ >x (6.1)

Here subscript ’b’ denotes the copy state, |K >b can be any state which acts
like a blank paper in a copying machine for practical purposes it may be
assumed as |0 >b . To simplify the notation we can eliminate this |K >b . So
QCM can be simply written as

|S >a |Q >x −→ |S >a |S >b |Q′ >x (6.2)

Now the question arises, will quantum mechanics allow such kind of QCM,
answer is ”NO” thus no-cloning theorem comes into picture.

32
6.1.1 No-Cloning Theorem
This tells us that quantum states cannot be cloned ideally for a arbitrary
input state, its later extended to even mixed states. It was shown that if
a arbitrary mixed state is sent through a QCM the two states which result
out whose reduced density operators are not identical. Now question arises
if we cannot make exact copy how close a copy can we come up with.
Approximate copying is important in fields like quantum cryptography and
quantum computing, where if we can have a measure of how similar the
copy is, we can actually copy the required information in that fraction of
the copy, in other words we will take in the part of the information which
was copied and eliminate the rest of it, for our given practical purpose.
We now try to look into three different QCMs.

• Input state dependent QCM

• Input state independent QCM

• QCM in the neighborhood of a given state

Then we will try to see about the entanglement of the copy state and the
original sate i.e. |S >b and |S >a . this is important because any measure-
ments made on the state |S >b would effect the original state |S >a

6.2 Input state dependent QCM

|0 >a Q >x −→ |0 >a |0 >b |Q0 >x (6.3)
|1 >a Q >x −→ |1 >a |1 >b |Q1 >x (6.4)

x< Q|Q >x = x < Q0 |Q0 >x = x < Q1 |Q1 >x = 1 (6.5)
x< Q0 |Q1 >x = x < Q1 |Q0 >x = 0 (6.6)
take a input state |S >a = α|0 >a + β|1 >a then,

|S >a |Q >x −→ α|0 >a |0 >b |Q0 >x + β|1 >a |1 >b |Q1 >x (6.7)

Distance between density operators ρ1 and ρ2 is defined by Hilbert-Schmidt

norm
D = ||ρ1 − ρ2 ||2 (6.8)

33
Where Hilbert-Schmidt norm of a operator A is defined as ||A|| = T r(A ∗ A)
This can be used as a measure of similarity between to density matrices.
Better measure of similarity is given by fidelity measure, which is defined as

F = T r(ρ1 1/2 ∗ ρ2 ∗ ρ1 1/2 ) (6.9)

(values ranging from 0 to 1) This is more promising definition to state the

similarity. Interpretation of Hilbert-Schmidt norm as probability distribu-
tions fail at infinite dimensional space and for finite dimensional space it
becomes less efficient as the
dimensionality increases. Where as fidelity measures are more reliable even
at higher dimensions. But these Hilbert-Schmidt norm is easier to calculate.
Coming to our area of interest in two dimensional space Hilbert-Schmidt
norm is a reasonable measure for the similarity.

Da = T r[ρ(id) a − ρ(out) a ] (6.10)

ρ(id) a = α2 |0 >a < 0| + αβ|0 >a < 1| + βα|1 >a < 0| + β 2 |1 >a < 1| (6.11)

ρ(out) a = T rb [(ρ(out) ab ] = α2 |0 >a < 0| + β 2 |1 >a < 1| (6.12)

2
using them we can find Da = 2(αβ) = 2α2 (1 − α2 ) As we can observe this
QCM is input state dependent as well as the output state Entangled. This
entanglement can be measured by the measure of distance between actual
two-bit density operator and the direct product of density operators ρ(out) a
and ρ(out) b , this is given by
2
D (1) ab = T r[ρ(out) ab − ρ(out) a ∗ ρ(out) b ] (6.13)

Then the efficiency of the QCM is given by the measure of distance

between the actual obtained density operator and ideal density operator.
Which is defined as
2
D(2) ab = T r[ρ(out) ab − ρ(id) ab ] (6.14)

To complete the the picture we define the third distance measure between
the ideal density operator and the product of independent density operators.
(3) 2
Dab = T r[ρ(id) ab − ρ(out) a ∗ ρ(out) b ] (6.15)

Where,
ρ(id) ab = ρ(id) a ∗ ρ(id) b (6.16)

34
 ρ(out) ab = T rx [(ρ(out) abx )] (6.17)
ρ(out) a = T r[b (ρ(out) ab )ρ(out) b ] = T r[a (ρ(out) ab )] (6.18)
by simple mathematics it can be shown that

D1 ab = Da ∗ Db (6.19)

D(2) ab = Da + Db (6.20)
(3)
Dab = Da + Db − D (1) ab (6.21)
Da maximum at |S >a = |+ >a or|− >a This copying would be worst for
those states.

6.3 Input state independent QCM

|0 >a |Q >x −→ |0 >a |0 >b |Q0 >x +[|0 >a |1 >b +|1 >a |0 >b |Y0 >x (6.22)
|0 >a |Q >x −→ |1 >a |1 >b |Q0 >x +[|0 >a |1 >b +|1 >a |0 >b |Y1 >x (6.23)
By unitarity or transformation we get

x< Qi |Qi >x + 2x < Yi |Yi >x = 1 (6.24)

x< Y0 |Y1 >x = x < Y1 |Y0 >x = 0; i = 0, 1 (6.25)

for further simplification we define |Qi >x and |Yi >x are orthogonal i.e.
x < Q i |Yi >x = 0; i = 0, 1 now we define

x< Yi |Yi >x = e (6.26)

and
x< Qi |Yj >x = n/2; i 6= jandi, j = 0, 1 (6.27)
0 ≤ e ≤ 1/2 & 0 ≤ n ≤ 2e1/2 [1 − 2e]1/2 this results from Schwarz inequality
let the input state be

|S >a = α|0 >a + β|1 >b (6.28)

As defined earlier we will now try finding the distance (measure or

similarity) between ideal density and obtained density matrices i.e.
2
D(2) ab = T r[ρ(out) ab − ρ(id) ab ] (6.29)

35
 we will solve for e and n such that this distance D (2) ab is independent of
input state i.e the value of D (2) ab does not contain the term α. by
mathematical simplification we can find e = 1/6 and n = 2/3 D (2) ab would
be then input state independent and is equal to 2/9
Using this information we can also fine the states as
√ √
|Y0 >x = [ (1/6), 0] |Y1 >x = [0, (1/6)]
√ √
|Q0 >x = [0, (2/3)] |Q1 >x = [ (2/3), 0]
We can find that |Y0 >x = (1/2)|Q1 >x & |Y1 >x = (1/2)|Q0 >x Then
the QCM simplifies to
√ √
|0 >a |Q >x −→ (2/3)|00 > | ↑> + (1/3)|+ > | ↓> (6.30)
√ √
|1 >a |Q >x −→ (2/3)|00 > | ↓> + (1/3)|+ > | ↑> (6.31)

6.4 QCM In the nighborhood of a given state

Consider a QCM define by

|1 >a |Q >x −→ |1 >a |1 >b |Q1 >x (6.32)

|0 >a |Q >x −→ [|0 >a |1 >b + |1 >a |0 >b )|Q1 >x (6.33)
|Q >x and |Q1 >x are initial and final states of the copying machine.
This QCM duplicate one basis completely and other is completely changed,
in fact |0 >a is transformed to a state which doesnot have overlap with its
perfect copy state,|0 >a |0 >b to the matter of fact it creates copies of the
states in the neighborhood of |1 >a Now lets try find the efficiency of copy
in the given neighborhood, consider the state

|S >a = α|0 >a + β|1 >a (6.34)

which is in the neighborhood of |1 >a i.e.β = 1 − δβ, 0 < δβ << 1 and
|u|2 + |v|2 = 1

|S >a |Q >x −→ (α|11 > +β|+ >)|Q1 >x (6.35)

√
D2ab = (3 − 2 2)|u|2 + |δβ|2 (6.36)
This actually has a order of |α|2
Now lets consider slightly better QCM

36
 √
|1 >a |Q >x −→ (1/2)[|11 > |Q1 >x + |00 > |Q0 >x ] (6.37)
√
|0 >a /Q > x −→ (1/2)[|01 > +|10 >]|Q1 >x (6.38)

D(2) ab = |δβ|2 + 2(|δβ||α|)2 (6.39)

This actually has the order of|α|4 thus the states in the neighborhood are
copied more efficiently.

6.5 Conclusions
No-cloning theorem prohibits the idea of a arbitrary quantum state to be
completely duplicated. But if one relaxes on the idea of complete replication,
we can define different Quantum Copying Machine which can perform the
work of copying the given states to the required accuracy. Thus we defined
few QCMs and tried to estimate the quality of the copy by using the distance
measure. Main problem with these QCMs is that the resulting copy and the
original are entangled hence they cant be treated independently i.e. in any
measurement is made on either element other gets effected automatically.

6.6 REFENENCES
V. Buzek & M.Hillery, Quantum copying: beyond no-cloning theorem; Phys-
ical Review Volume 54,Number 3: September 1996

37
Chapter 7

Single Qubit and CNOT

Gates are Universal

J. Karthik MT02B011

Abstract
The single qubit and CNOT gates are shown to be universal for quantum
computation Here we explicitly construct the solution and prove that CNOT
and single qubit gates are enough to approximate any unitary operator to
arbitrary accuracy.

7.1 Introduction
A set of gates is said to be universal for quantum computation if any unitary
operation can be approximated to arbitrary accuracy by using a circuit
involving these gates only Here, we prove by construction that single qubit
and CNOT gates are universal.
The first construction shows that any unitary operator can be expressed
as the product of unitary operators each of which acts on a subspace spanned
by two computational basis states. The second construction proves that
any 2-level unitary operator can be approximated using a circuit containing
single qubit and CNOT gates. This completes the proof.

7.2 Two-level unitaries are universal

Two-level unitaries are the unitaries that act only on a subspace spanned
by two of the computational basis states. In this section, we prove that any

38
unitary operator U can be decomposed into a product of two-level unitaries.
Considering a 3x3 unitary U
 
a b c
U= d e f 
g h j

We can find two-level unitaries U1 , U2 ,and U3 such that U1 U2 U3 = U It

follows that U1−1 U2−1 U3−1 = U where Ui−1 is also the hermitian conjugate
of Ui because U is unitary. If this is true, the decomposition into two-level
unitaries is proved for the 3x3 case. Now if we contruct matrices U1 and U2
such that  ∗
b∗

a
p p 0
U1 =  pb −a 0 
 
p
0 0 1
and ′ ′
a∗ c∗
 
q 0 q
U2 =  0 1 0 
 
′ ′
c −a
q 0 q

where, U1 U =  ′ ′ ′ 
a d g
′ ′
U1 U =  0 e h 
′ ′ ′
c f j
√ √
p = a2 + b2 and q = a′ 2 + c′ 2
Now we find that
 ′′ ′′ 
1 d g
′′ ′′
U2 U1 U =  0 e h 
′′ ′′
0 f j

Since U2 , U1 and U are unitary it follows that their product is also unitary.
′′ ′′
Therefore g and d are 0. Finally, if we set U3 to be equal to
 
1 0 0
′′ ′′
U2 U1 U =  0 e ∗ h ∗ 
′′ ′′
0 f ∗ j ∗

It is easy to verify that U3 U2 U1 = U and that the unitaries Ui are indeed

two-level unitaries. Thus the proof for the 3x3 case.

39
 If U acts on a d dimensional space we find two-level unitary matrices U1 ,
U2 . . . Ud−1 such that Ud−1 Ud−2 . . . U1 has a 1 in the top left hand corner
that zeroes everywhere else in the first row and column. We then repeat the
procedure for the d − 1 by d − 1 submatrix and so on. We finally obtain
matrices V1 , V2 etc such that U = V1 V2 . . . Vk where k is at the most equal
to (d − 1) + (d − 2) + . . . + 1.
Thus, by construction it is proved that any unitary matrix can be written
as a product of two-level unitaries.

7.3 Single qubit and CNOT gates are universal

In this section, we prove that every two-level unitary matrix can be approx-
imated by the single qubit and CNOT gates.
Let the two-level unitary act on a subspace spanned by the basis states s
and t (which we assume are represented in binary, for convenience). We now
write a sequence of binary numbers starting with s and ending with t such
that the adjacent members differ by a single bit. Let the members of the
sequence be gi such that g1 = s and gm = t. The following operations result
in the application of U which acts nontrivially only on s and t.(Only the
states occuring in the sequence are manipulated in the following method.
Therefore its enough if we follow them)
We first swap the following pairs. g1 - g2 , g2 - g3 . . . gm−1 - gm−2 . The
swapping of two states can be accomplished by doing a controlled bit flip on
the bit that these two states differ in, conditional on the values of the other
bits being identical to those in these two states. After the swapping, we do
a controlled U operation on gm−1 and gm with the target bit being the one
which these two differ in and all the others are control bits. We then undo
the first m − 1 swaps which will complete the application of U .
This method proves that two-level unitaries can be implemented by using
n−1 bit controlled operations. Further, these n−1 bit controlled operations
can be approximated by using only single qubit and CNOT gates.
Since any unitary matrix can be written as a product of two-level uni-
taries, it follows that any unitary matrix can be approximated to arbitrary
accuracy by single qubit and CNOT gates alone.

Bibliography
1. Michael A. Nielsen and Isaac Chuang, Quantum Computation and
Quantum Information, Cambrige University Press, Cambridge, 2000

40
2. K. R. Parthasarathy, Lectures on Quantum Computation and Quan-
tum Error Correcting Codes, Indian Statistical Institute, 2001

41
Chapter 8

Realization of Physical
Quantum Computer

Brajesh Tiwari, PH04C009

Abstract
In my present study I am going to give difference between classical computer
and Quantum computer,Requirements to make quantum computers, Different
ways to to get these requirements, difficulties to implement it and some details
of making it by CQED (one atom and one photon method) and Quantum Dot.

8.1 Introduction
Present is the age of computer.We can not imagine our daily life without
computer.The decrease in size as well as the increase in speed of classical
computer has its limitation due to quantum confinement effect,switching
speed (due to saturation current) and leakage current.To improve the speed
further and decrease the size we have to think new ways to implement
computer which are the Quantum Mechanical ideas(superposition princi-
ple).These Quantum Mechanical ideas give the effective parallelism called
Feynman Quantum Parallelism i.e. inputs are processed (transformed)
to outputs in one step via a unitary operation which increase the speed dras-
tically and size also.The challenge is to implement or design Hamiltonian
interaction which evolve the system such that the states are transformed as
our will.The statement ”design Hamiltonian interaction and the operation
(Unitary) that transform inputs into outputs”is the core problem to make
a Quantum Computer.

42
8.2 Difference between Classical and Quantum Com-
puter
Computation is the process which transforms inputs to output.
(1)In case of classical computers or presently existing computers the number
of input may be one or more but output is only one while in case of quantum
computer the number of inputs and outputs are same.
(2)In classical computer output is come from multiple steps while in quan-
tum computer output is come from only one step.All the results are there
in in the output with certain finite probability.
(3)In classical computer the fundamental unit of information are binary
number or Bit while in quantum computer fundamental unit is Qubit not a
binary but more Quaternary in nature.
(4)In classical computer information is encoded in a series of bits, and these
bits are manipulated via Boolean logic gates arranged in succession to pro-
duce an end result while a quantum computer manipulates qubits by exe-
cuting a series of quantum gates, each a unitary transformation acting on a
single qubit or pair of qubits and the succession of gates perform a unitary
transformation to a set of qubit with some initial states.The qubit can then
be measured which gives the output computational result.

8.3 Requirements and Difficulties

Quantum Computation can be done on a classical computer but the simu-
lation of the problem is exponentially hard and incredibly inefficient while
quantum computer do it ease.

8.3.1 A set of qubit:

A two level system perfectly identified and which form a quantum resister
that can be manipulated and measured in a controlled way.The qubits can
be either in superposition state or entangled state which gives extraordinary
power to the quantum computer.The state of qubit must kept almost pure
otherwise power of superposition will not effective i.e.the qubits must be well
isolated from the environment such a way that the process of decoherence
is sufficiently slow.
Decoherence gives a major problem to a physical realization of quantum
computer. Decoherence is caused by interactions with a second system
which may be thought of as either ”the environment” or as ”a measuring

43
device”.The interaction may be considered as quantum measurement.The
measuring device or environment can entangled with the system.Both sys-
tem and environment must evolve in same state but if environment has many
degree of freedom then it not happens so i.e.system behaves like classical sta-
tistical ens emble rather than superposition state.Decoherence is extremely
fast process in macroscopic objects.It is difficult to maintain superposition
state of many particle system in which each particle may separated from all
other.

8.3.2 A set of Quantum Gates

The controlled manipulation of qubits means we can perform an unitary
operating U on the qubit.Any arbitrary instructions between the qubits can
be decomposed as product of gates belonging to small set called universal
gate.
1)Single-qbit gate; phase gate, excitations
2)Two-qbit gate; Controlled- Not (CNOT),Controlled Phase

8.3.3 Detection
This process requires the interaction between measurement apparatus in an
irreversible way which gives whether each qubit is either in 0 or 1 state.

8.3.4 Scalability
The difficulty for performing gates, measurements, etc. should not grow
(exponentially) as the number of qubits increases otherwise, the gain in
quantum algorithms would be lost.

8.3.5 Erase
We must able to prepare the initial state of the system e.g 0 state in com-
putational basis.
The most important problem is the necessity of finding quantum system with
sufficient isolation and a controlled interaction.There are very few systems
which fulfills the above requirement of making Quantum Computer.

8.4 Quantum Computers

NMR quantum computer: Decoherence Time (DT) is 10−2 to 108 , Opera-
tion Time (OT) is 10−3 to 10−6

44
Ion Trap (In+ ): DT is 10−3 , OT is 10−7
Optical Cavity QED: DT is 10−5 and OT is 1014
Quantum Dot: DT is 10−6 and OT is 10−9
Electron Spin: DT is 10−3 and OT is 10−7
Electro GaAs: DT is 10−10 and OT is 10−13

I am going to deal with CQED only.To realize the quantum computer

first we have to make quantum resister a two level atom which ground state
is 0 state and excited state is the 1 state and then controlled unitary oper-
ation like C-NOT, Hadamard transformation,Controlled phase gate.
To realize C-NOT gate; If an atom in an optical lattice trap is brought close
to the second atom in an adjacent well of the lattice then the energy level
of valence electro in each of the atoms are shifted owing to the fluctuation
of dipole moment(oscillatory) of the other atom.The shift is greater for the
higher levels thus electron in excited state peaks up phase more slowly then
are in the ground state.This conditional phase shift (i.e. the amount of
phase depends on the state of control electron) forms this basis of this kind
of conditional state change is performed by a quantum C-Not gate.
Optical Lattice has important role so I describe it.Optical lattice is a peri-
odic light shift potentials that are formed by interference of two or more laser
beams.Atoms are cooled and localized in the potential minima.It has larger
coherence time in compression to solid state lattice.Optical lattice is able to
trap atoms because electromagnetic fields induces dipole moment.the shift-
ing in light depends on detuning.The interaction between dipole moment
(oscillatory) and electromagnetic field of laser modify the energy of atom.If
laser frequency is greater then atomic transition frequency within the atom
the atom is pulled towards the max intensity region and vice a versa i.e atom
is trapped in bright and dark region.
Jayen Cumming Model which is a two level atom interacting with quantized
field.In this model there are two approximations first one is dipole approxi-
mation and then Rotating Wave Approximation (RWA).
The Jayen Cumming Hamiltonian is

H = 1/2h̄ω0 + h̄ωa† a + h̄λ(σ+ a + σ− a† ) (8.1)

λ = (dg)/h̄ (8.2)
g = −(h̄ω/ǫ0 V )1/2 sin(kz) (8.3)
d = he | dˆ | gi W here | σ+ i =| eihg | and | σ− i =| gihe |
consider initialy the state‘ is | ii =| ei | ni the f inal state is either

45
| gi | n + 1i or | gi | n − 1i gives‘ the general solution as
√ √
| ψ(t)i = cos(λt n + 1) | ei | ni − i sin(λt n + 1) | gi | n + 1i (8.4)
√
Rabbi frequency is define defined as Ω = 2λ n + 1
If Ωt = π/2 then we can get entangled state for no photon in excited sate
and one photon in ground state
√
| ψπ /2i = 1/ 2(| e, 0i − i | g, 1i) (8.5)

and for all other Ω t it can be used for different type of operations on
the qubit.

8.5 Conclusions
Using two level atom interact with radiation field we are able to make qubits
which is first requirement and we can realize phase gate as well as Hadamard
gate.Two atom trapped in optical lattice can able to manipulate second atom
conditionally by the first atom i.e. we can treat it as Controlled Phase gate
(C-NOT gate).Science Quantum Dots can be assumed as artificial atom
embedded on a surface can also be used for the same purpose which can be
handled easily.
To make a physical quantum computer we need well isolated qubits which
can be manipulated in controled way and must maintain its superposition
state (less Decoherence) for a long time to extract informatin from that.

8.6 Reference
Books
Experimental Quantum Computation and Information Course CXLVIII
Pub:Societa Italiana De Ficica Bolona-Italy
Introductory Quantum Optics by C.Gerry and P.Knight Pub:Cambridge
University.
Paper
T.Radke and S Fritzsche Computer Physics Communications 173(2005Aug)91-
113
M.G Tanner D.G Haske D.A Williams Microelectronics Engineering xxx(2006)xxx
xxx

46
Chapter 9

Interferometry Using Large

Matter Particles

Chirag Patidar, Ph04C010

Abstract
In 1999 Anton Zeilinger and his colleague at the university of Vienna demon-
strated the wave nature of C60 molecules[1]. An obvious motivation for doing
experiements with such large particles is the everyday experience that particle
don‘t spread like wave,rather they have a well defined position whenever they
are observed.While microscopic objects like electron,nuetron etc. shows wave
like behaviour in some situations.How to understand this quantum to classical
transition;linking to incompatible description of reality? It turns out to that
macroscopic C60 molecules also show wave like behaviour if all the sources of
decoherence are eliminated from the experimental set up.This strenghen wave
particle daulity as one of the fundamental principle of nature.

9.1 Introduction
Albert Einstien one of the great thinker of 20th century proposed the idea
of duality of light in order to explain photo-electric effect in 1905. Carrying
further his idea Louis-Victor de broglie proposed the idea of wave particle
daulity in 1924. He gave a relationship between momentum of particle and
wavelength associated with it. The idea was this if light can behave in some
situations as wave and in some other situations as particle than matter
particles should also behave like that. This idea is one of the corner stones
of quantum mechanics.
It was 1927 when first experiemental proof of wave particle daulity was

47
given by Davision and Germon for microscopic particles.They observed the
diffraction of electron on Ni crystal. After that in 1930, Esterman andStem
made a beam of healium atom diffracted on the surface of a Nacl crystal.
Ater that many more experiements were performed using microscopic small
particles like to show wave particle daulity.
In 1999 Zeilinger and his group has shown that C60 molecule shows wave
like nature. It provides solid proof of the idea of wave particle daulity as a
fundamental princile of nature.It is now well stablished fact that decoher-
ence is the process which destroy this wave nature and make our everyday
experiencce classical.

9.1.1 Decoherence
Decoherence is the process which makes a quantum system behave like a
classical system. It arises because of interaction of system with environment
and measuring instrument.Because of this system gets entangled with the
wavefunction of environment and we loose phase infromation. Processes
which cause decoherence are following
1.Emission of radiation
2.Absorption of radiation
3.Collision
4.scattering
There are mathematical models by which decoherence can be explained[2].
Let us assume that system is in state

| Ψi = a | 0i + b | 1i (9.1)

with aa∗ +bb∗ =1 at time t=0. So at time t=0 we have density operator

aa∗ ab∗
 
ρ= (9.2)
a∗ b bb∗
Now suppose that system can evolve with probability p by the unitary trans-
formation U1 and by probability (1-p) by the uniary transformation U2 .
Then at time t=t we have the density operator

ρt = pU1 ρU2† + (1 − p)U1 ρU2† (9.3)

Now let me generalise the idea by taking rotation as unitary transformation

R(θ) with the probability given by guassian distributation having variance

48
λ and centred at zero.Then at time t=t the state of the system is gven by
Z∞
aa∗ ab∗ e−λ
 
−θ 2
†
ρ= R(θ)ρR (θ)e 2λ2 dθ = ∗ −λ (9.4)
a be bb∗
−∞

Here off diagonal elements of the density matrix depend uopn the variance of
guassian distributation of the random variable angle. Off diagonal elements
of the density dies exponentially as variance increases and so as the phase
coherence between two states dies exponentially. if the diagonal elements
are zero then the system will behave like a classical system.

9.2 Main Study

9.2.1 Test particle
C60 is a molecule that consists of 60 carbon atoms, arranged as 12 pentagons
and 20 hexagons. The shape is the same as that of a soccer ball.There are 120
symmetry operations, like rotations around an axis or reflections in a plane,
which map the molecule onto itself.In their experiement they sublimated
C60 fullerenes in an oven at temperatures between 900 and 1000K.

Figure 9.1: Molecular structure of C60

9.2.2 Experimental Set-Up

The emerging molecular beam was passed through two collimation slits,
each about 10m wide, separated by a distance of 1.04m. Then it traversed
a free-standin nanofabricated SiNx grating consisting of nominally 50 nm
wide slits with a 100 nm period. At a further distance of 1.25m behind the

49
 Figure 9.2: Experimental Set-Up

diffraction grating,the interference pattern was observed using a spatially

resolving detector. It consisted of a beam from a visible argon-ion laser,
focused to a gaussian waist of 8m width (this is the size required for the
light intensity to drop to 1/e2 of that in the centre of the beam). The light
beam was directed vertically, parallel both to the lines of the diffraction
grating and to the collimation slits. By using a suitable mirror assembly, the
focus could be scanned with micrometre resolution across the interference
pattern. The absorbed light then ionized the C60 fullerenes via heating and
subsequent thermal emission of electron.The detection region was found to
be smaller than 1mm in height, consistent with a full Rayleigh length of
800m. A significant advantage of the thermionic mechanism is that it does
not detect any of the residual gases present in the vacuum chamber

9.3 Conclusions
The interference pattern produced by C60 molecules clearly shows a cen-
trel maxima and first order maxima and minima which can be fitted using
kirchoff diffracton theory. This is a clear manifestation of the wave nature
of C60 molecules like other microscopic particle. Position of the maxima
depends on the temperature of molecules, grating period, distance between
grating and detector.The most probable velocity of C60 molecules is 220 m/s
at 900K.Debroglie wavelength of molecules at this velocity is 2.5pm which
has been calculated using λ = h/M v,here M is mass of a C60 molecule.This

50
Figure 9.3: a,diffraction pattern of C60 molecules and b,pattern without
grating

shows that a single molecule as a whole travels in the path and don‘t get de-
composed into fragments in the path.Now to destroy interefernce we need to
have some means by which we can have path information of the molecule.But
to have path information we need λ ≪ d. So the wavelength 418 nm of the
laser is not sufficient to induce decoherence by single photon scattering but
multiple photon can induce decoherence that is why we have to use as low as
possible photons for detecting molecules. Other processes which can cause
decoherence is emission of photon due to vibrational energy and rotationalal
energy,scattering due to air molecules. At T=900 K, as in their experiment,
each C60 molecule has on average a total vibrational energy of Ev = 7 eV
[3] stored in 174 vibrational modes, four of which may emit infrared radia-
tion at λvib = 719 m [4] each with an Einstein coefficient of Ak 100 s−1 [3].
During its time of flight from the grating towards the detector (τ ≈ 6ms) a
C60 molecule may thus emit on average 23 such photons.
In addition, hot C60 has been observed[5] to emit continuous blackbody
radiation, in agreement with Planck’s law, with a measured integrated emis-
sivity ofǫ = 4.5(±2.0) × 10−5 [3]. For a typical value of T 900 K, the average
energy emitted during the time of flight can then be estimated as only Ebb
= 0.1 eV. This corresponds to the emission of (for example) a single photon
at lambda 10 m. Absorption of blackbody radiation has an even smaller in-
fluence as the environment is at a lower temperature than the molecule.And
in last mean free path of molecules at pressure 5 × 107 mbar and 900K is

51
100 m so colossion with background air molecules can be neglected.So all
the source of decoherence has been eliminated in the experiment and as a
result diffraction pattern formed.
Observation of quantum interference with fullerenes also shows that for
obserbing interference high symmetry of molecule is not necessary. fullerenes
are found as a 50-50%mixture of highly symmetric 12 C60 and less symmetric
12 C 13 C and 12 C 13 C isotopomeric variants.Interesting thing is that all
59 58 2
the fillerenes contribute to interefence because if this was not the case then
the background count would be very high.
So finally I can conclude that if all the sourece of decoherence can be
eliminated then we will be able to see wave nature of particles in our day to
day life.

9.4 References
1. Markus Arndt, Olaf Nairz, Julian Vos-Andreae, Claudia Keller, Gerbrand
van der Zouw, Anton Zeilinger. Nature 401, 680-682 (14 Oct 1999) Letters
to Editor
2. Quantum Computation and Quantum Information.Michael A.Nielsen &
Isaac L.Chuang. Edition 2000.
3. Kolodney,E., Budrevich,A. & Tsipinyuk B. Unimolecular rate constants
and cooling mechanisms of superhot C6 0 molecules. Phys. Rev. Lett. 74,
510-513 (1995).
4. Krtschmer,W., Lamb,L. D., Fostiropoulos,K. & Huffman,D. R. A new
form of carbon. Nature 347, 354-358 (1990).
5. Mitzner,R. & Campbell,E. E. B. Optical emission studies of laser des-
orbed C6 0. J. Chem. Phys. 103, 2445-2453 (1995).

52
Chapter 10

Experimental test of
quantum non-locality in
three photon GHZ
entanglement

Dileep.M, PH04C021

Abstract
Bells theorem finds a conflict between quantum physics and local realism. Sim-
ilarly when we observe the three photon GHZ states the same conflict remains
out aim is study these quantum physical and local realisms to get the proof of
the conflict.

10.1 Introduction
Bells theorem states that certain statistical correlation predicted by quan-
tum physics for measurements on two particle systems can not be understood
within a realistic picture based on local properties of each individual particle
even if they are separated by large distances.
It was discovered that there is a conflict between quantum mechanical
and local realistic predictions. But verifying this is very difficult as it requires
entanglement between at least three particles.
Here in this article, we are reporting an experimental confirmation of
this conflict using a method to observe three photon entanglement or GHZ
states.

53
 Figure 10.1: Experimental set-up for GHZ tests of quantum nonlocality

10.2 Main Study

10.2.1 Method we are using
Here w are conducting three specific experiments involving measurements
of polarization correlation between three photons which will lead to predic-
tions of a fourth experiment. By these experiments we are observing that,
quantum physical prediction are mutually contradictory to the prediction
based on local realism.
What we found from the experiment is the results of the fourth experi-
ment is in agreement with the quantum prediction and there is conflict with
local realism.

10.2.2 How we do this?

As shown in Figure 10.1 pairs of polarization-entangled photons(one photon
H polarized and the other V ) are generated by a short pulse of ultraviolet
light ( 200 fs, λ = 394 nm).Observation of the desired GHZ correlations
requires fourfold coincidence and therefore two pairs. The photon registered
at T is always H and thus its partner in ~b must be V. The photon reflected
at the polarizing beam-splitter (PBS) in arm ~a is always V, being turned
into equal superposition of V and H by the l/2 plate, and its partner in arm
~b must be H. Thus if all four detectors register at the same time, the two
photons in D1 and D2 must either both have been V V and reflected by the

54
last PBS or HH and transmitted. The photon at D3 was therefore H or V,
respectively. Both possibilities are made indistinguishable by having equal
path lengths via ~a and ~b to D1 (D2) and by using narrow bandwidth filters
(F ≈ 4 nm) to stretch the coherence time to about 500 fs, substantially
larger than the pulse length. This effectively erases the prior correlation
information and, owing to indistinguishability,the three photons registered
at D1, D2 and D3 exhibit the desired GHZ correlations predicted by the
state of equation [10.1], where for simplicity we assume the polarizations at
D3 to be defined at right angles relative to the others. Polarizers oriented
at 45deg and l/4 plates in front of the detectors allow measurement of linear
H ′ /V ′ (circular R/L) polarization from which we can compare the results
with the prediction made by quantum physical and local realism. Let us
take the entangled three particle state
1
|ψi = √ (|Hi1 |Hi2 |Hi3 + |V i1 |V i2 |V i3 ) (10.1)
2
Where H and V denote horizontal and vertical linear polarization resp. If
we do the measurements of linear polarization along direction H/V rotated
by 45 degrees with respect to H/V or of circular polarization L/R. these
new polarization can be expressed in terms of original one as
1
|H′i = √ (|Hi + |V i)
2
1
|V ′i = √ (|Hi − |V i)
2
1
|R′i = √ (|Hi + i|V i)
2
1
|L′i = √ (|Hi − i|V i)
2
For our convenience we use the notation one of H/V as x measurement and
one of R/L as y measurement.One obtains quantum prediction for measure-
ments of these new polarizations. For example, measurements of circular
polarization on say photon 1,2 and linear polarization on photon 3 denoted
by yyx experiment The state will become as
1
|ψi = (|Ri1 |Li2 |H ′ i3 + |Li1 |Ri2 |H ′ i3 + |Ri1 |Ri2 |V ′ i3 + |Li1 |Li2 |V ′ i3 )
2
(10.2)
From the expression we can say that
1. the result obtained is maximally random

55
 Figure 10.2: A typical experimental used in the GHZ argument

2. Given any two results of measurement of any two photons we can

predict with certainty the result of the corresponding measurement
performed on third photon.
Similarly we can predict for other measurement like yyx, yxy and xyy.
In the case of local realism Einsteins locality will tell you that no in-
formation can travel faster that light. So for simultaneous measurements
we can say that measurement of one will never depend on the other two
and their outcomes. This is the yyx experiment measuring circular polar-
ization on photons 1 and 2 and linear polarization on the third. ~a, Fourfold
coincidences between the trigger detector T, detectors D1 and D2 (both
set to measure a right-handed polarized photon), and detector D3 (set to
measure a linearly polarized H ′ (lower curve) and V ′ (upper curve) photon
as a function of the delay between photon 1 and 2 at the nal polarizing
beam-splitter). We could adjust the time delay between paths ~a and ~b in
Figure 10.1 by translating the final polarizing beam-splitter (PBS) and us-
ing additional mirrors (not shown in Figure 10.1) to ensure overlap of both
beams, independent of mirror displacement. At large delay, that is, outside
the region of coherent superposition, the two possibilities HHH and VVV
are distinguishable and no entanglement results. In agreement with this ex-
planation, it was observed within the experimental accuracy that for large
delay the eight possible outcomes in the yyx experiment (and also the other
experiments) have the same coincidence rate, whose mean value was chosen
as a normalization standard. b, At zero delay maximum GHZ entanglement
results; the experimentally determined fractions of RRV9 and RRH9 triples
(out of the eight possible outcomes in the yyx experiment) are deduced from
the measurements at zero delay. The fractions were obtained by dividing

56
 Figure 10.3: Predictions of quantum mechanics and of local realism

the normalized fourfold coincidences of a specific outcome by the sum of all

possible outcomes in each experimenthere, the yyx experiment.
Figure ?? tells that the fraction of occurrence of the yyo experiment is
more for RRV ′ state than RRH ′ state. This is the experimental result.
So, the correlation predicted by equation [10.2] is that each photon carries
elements of reality for both x and y measurements that determining the
specific individual measurement result.
The elements of reality for different experiments are
+1 for H ′ and -1 for V ′ +1 for R and -1 for L
so, Y1 Y2 X3 = −1, Y1 X2 Y3 = −1, X1 Y2 Y3 = −1.
to reproduce the quantum prediction we need to do a fourth experiment
of x on all three photons. This xxx experiment will tell you the outcome
predicted by local realisms.For Einsteins locality Yi Yi = H

X1 X2 X3 = (X1 Y2 Y3 )(Y1 X2 Y3 )(Y1 Y2 X3 )

= −1

From this the local realistic possible result for XXX is given by the wave
function.
1
|ψi = (|H ′ i1 |H ′ i2 |H ′ i3 +|H ′ i1 |V ′ i2 |V ′ i3 +|V ′ i1 |H ′ i2 |V ′ i3 +|V ′ i1 |V ′ i2 |V ′ i3 .
2
(10.3)
From these things we can conclude that none of the local realistic model
predicts none of the terms that was previously predicted by quantum ap-
proach. This is the GHZ contradiction between local realisms and quantum
physics.
From the above fig we can see that experimental results are more in
agreement with the quantum physics.
In Bells inequalities for two photons, the conflict between local realism
and quantum physics arises because of statistical prediction. But in case of
GHZ states the conflict arises even for the case of definite prediction.

57
 There are some doubts weather such a experimental setup can be used for
local realisms But this has been disproved recently. The procedure permits
valid GHZ test for local realism In essence both Bell and GHZ argument
exhibit a conflict between detection events and the ideas of local realism.
Here what we did was, we first conducted three spatially separated po-
larization measurements. If the results obtained are in agreement with the
prediction for a GHZ state.Then for an XXX experiment, our consequent
experiment using a local realistic theory is exactly the opposite of our ex-
pectation using quantum physics.

10.3 Conclusion
From the experiment we proved the conflict between quantum physics and
local realisms. We observed that the experimental results are in agreement
with the quantum physics and but in conflict with the local realisms.

10.4 Reference
Zeilinger et. al, NATURE, VOL 403, 3 FEBRUARY 2000.

58
Chapter 11

Nonlocality for Two Particle

Systems

Joshi Anand D, ph04C014

Abstract
No abstract was provided

11.1 Introduction
The Bell inequality is given by

E(QS) + E(RS) + E(RT ) − E(QT ) <= 2 (11.1)

We find that this inequality does not hold for the maximally entangled state,
| ψi = |01i−|10i
√
2
and for the operators, √ √
Q = Z1 , R = X1 , S = (−Z2 − X2 )/ 2 and T = (Z2 − X2 )/ 2
We find that the Bell inequality√is violated for the case presented, the
sum of average values is less than 2 2 and not 2. Hence, there should have
crept into the derivation some flaw. The flawed argument is that of ’locality’
and ’realism’ which we define as below.
Realism is the assumption that physical properties PQ , PR , PS , PT have
definite values Q, R, S, T independent of any observation. It means that
the Moon exists even when nobody is looking at it.
Locality is the assumption that Alice performing her measurement does
not influence result of Bob’s measurement and vice versa. Therefore, both
can carry out their measurements without any effect on each other.

59
11.2 The Basis Sets and The Operators
In this report I have given a proof that quantum mechanics is essentially
nonlocal if we assume realism. It is a general proof, it however comes with
a caveat; it fails for states which are maximally entangled.
Consider basis set for two particles comprising of orthogonal states |
+ii and | −ii We can write any two particle general entangled state as a
Schmidt decomposition as,

| Ψi = α | +i1 | −i2 − β | +i1 | −i2 (11.2)

given the constraints that,

α2 + β 2 = 1. (11.3)
We introduce another basis set related by the earlier one as,

| +ii = b | ui ii + ia∗ | vi i (11.4)

| −ii = ia | ui ii + b∗ | vi i (11.5)
with the inverse relations easily obtainable as,

| ui i = b∗ | +ii − ia∗ | −ii (11.6)

| vi i = −ia | +ii + b∗ | −ii . (11.7)

Using the new basis set, we will write state defined earlier as,

| Ψi = (αb2 + βa2 ) | u1 i | u2 i + i(αa∗ b − βab∗ )[| u1 i | v2 i+ | v1 i | u2 i]

− [α(a∗ )2 + β(b∗ )2 ] | v1 i | v2 i. (11.8)

For reasons that will be subsequently explained, we equate first term of the
above equation to zero i.e. a2 /α = −b2 /β = k2 , say. This implies,
√ p
a = k αandb = ik β (11.9)

k2 = 1/(|α| + |β|) (11.10)

p
| Ψi = − αβ[| u1 i | v2 i+ | v1 i | u2 i] + (|α| − |β|) | v1 i | v2 i (11.11)
With these relations, we therefore rewrite the state as,
√
αβ p
| Ψi = [ p | u1 i + |α| − |β| | v1 i]
|α| − |β|
√
αβ p αβ
[p | u2 i + |α| − |β| | v2 i] − | u1 i | u 2 i (11.12)
|α| − |β| |α| − |β|

60
 Having got so far writing the equations, we will introduce another basis
set which is
| ci i = A | ui i + B | vi i (11.13)
| di i = −B ∗ | ui i + A∗ | vi i (11.14)
whose inverse relations cen be obtained as,

| ui i = A∗ | ci i − B | di i (11.15)

| vi i = B ∗ | ci i + A | di i (11.16)
where we have defined coefficients A and B as,
√
αβ |α| − |β|
A= p and B = p (11.17)
1 − |αβ| 1 − |αβ|

and we can write the entangled state as,

| Ψi = N (| c1 i | c2 i − A2 | u1 i | u2 i) (11.18)

where
1 − |αβ|
N= (11.19)
|α| − |β|
Substituting equations (13) and (15) in equation (18), we have the two-
particle entangled states written in following four equivalent forms. In short,
what we are doing here is generating basis sets such that in the 4 forms given,
we have in 1st form both the particles in | ui i, | vi i basis, in 2nd form, particle
one in | c1 i, | d1 i basis and particle two in | u2 i, | v2 i basis and so on.

| Ψi = N (AB | u1 i | v2 i + AB | v1 i | u2 i + B 2 | v1 i | v2 i) (11.20)

| Ψi = N (| c1 i(A | u2 i + B | v2 i) − A2 (A∗ | c1 i − B | d1 i) | u2 i) (11.21)

| Ψi = N ((A | u1 i + B | v1 i) | c2 i − A2 | u1 i(A∗ | c2 i − B | d2 i)) (11.22)
| Ψi = N (| c1 i | c2 i − A2 (A∗ | c1 i − B | d1 i)(A∗ | c2 i − B | d2 i)) (11.23)
Now we define two observables Ui and Di with corresponding projection
operators which are non-commuting in general.
Ui =| ui ihui | Di =| di ihdi |
The operators can have values either 0 or 1 and the subscript i indicates
the particle on which they will operate. We now use these two operators to
measure values on states in different forms constructed in equations (20) to

61
(23). We measure U1 and U2 on (20), since there is no term of the form
| u1 i | u2 i , we have,
U1 U2 = 0. (11.24)
We also measure D1 and U2 on (21), U1 and D2 on (22) and D1 and D2 on
(23). In (21), only | d1 i | u2 i term contains | d1 i, therefore, we have,

if D1 = 1 then U2 = 1. (11.25)

Similarly, from (22), we have,

if D2 = 1 then U1 = 1 (11.26)

and from (23), we get the result,

D1 = 1 and D2 = 1 (11.27)

where, the probability of getting outcome (25) is |N A2 B 2 |2 and it will be

used to prove some things later.

11.3 Proof of Nonlocality

We set out now to prove the nonlocal nature of the quantum world. If
we assume the realism, then we have a ’hidden variable’ λ. The values of
operators Ui and Di depend on this hidden variable and not on anything else.
For the sake of measurements, we can assume that after formation of states,
particles 1 and 2 can separate and incident on two apparatuses. Locality
implies that measurement outcome of one particle in no way influences that
of the other.
We now consider measurements and their outcomes carried out. From
equation (27), we have D1 =1 and D2 =1 while from (25) we infer that since
D1 =1, U2 =1. Also, due to locality, this value of U2 can be attributed to the
hidden variable. Hence, we write U2 (λ) = 1. If we consider (26), we similarly
get U1 = 1 and due to the assumption of locality, U1 (λ) = 1. Therefore, we
get, U1 (λ)U2 (λ) = 1, which is in clear contradiction to equation (22) where
we have U1 U2 = 1
The assumption of realism allows us to set up a hidden variable which
determines values of observables. This assumption and that of locality leads
to a contradiction. Hence we conclude that the realistic interpretation of
quantum mechanics is nonlocal.

62
11.4 The Caveat
This proof, although good enough for any general two-particle entangled
state, fails for maximally entangled state. It is mentioned earlier that the
probability of getting outcome (25) is |N A2 B 2 |2 . Using values of A, B in
equation (17) we write denoting the probability by γ,
2
γ = (|α| − |β|)2 |αβ|
(11.28)
(1 − |αβ|)2

If either α=0 or β=0, we get the trivial case of non-entangled state Ψ in

equation (2). For a two-particle system, the maximally entangled state
would correspond to α = β. This will make γ=0 and we will not be able
to run our nonlocality argument. The reason for inapplicability of the proof
presented to maximally entangled state is certain lack of symmetry possessed
by an entangled state and which is absent in a maximally entangled state.

11.5 References
1. L. Hardy, Phys. Rev. Lett., vol. 71, no. 11, pg. 1665 (1993).
2. Quantum Computation and Quantum Information, M. A. Nielsen and I.
L. Chang, Cambridge Univ. Press, 1st edition, 2002.

63
Chapter 12

Reducing the complexity of

reduction

Krishnendu Maity, PH04C018

Abstract
In 1998 Agrawal et al worked on the isomorphisms of complete sets.And the
main theorem they used there is Isomorphism theorem[1]. This theorem states
that all sets that are complete under(non-uniform) AC0 reductions are isomor-
phic under isomorphisms computable and invertible via(non-uniform) depth-
three AC0 circuits. In that paper some questions were left open. Later in
2001[2] he proposed some theorem to give the answer of all those question.
And in my work I have described the questions left open at that paper and
try to realize the theorem proposed by him to understand the answer of those
question.

12.1 Introduction
The computational problems are devided according to the difficulty to solve
the problem. Now there exist some suitable complexity classes[3] under
which many of the computational problem can turn out to be complete,
even under very restrictive notions of reducibility. Now in our present case
we will first try to realize the complexity classes(Section 2), and different
kind of problems under the complexity classes. In this present work to un-
derstand the theorem proposed by Agrawal we should have some knowledge
about reduction(section 3) process.So for that I will give some descriptions
about the differnt kind of reductions(section 5), and to understan the com-
plexity classes well I will give some idea about circuit family(section 4) and

64
differnt properties of circuit family which will also be helpful to understand
the proposed theorems by agrawal.And at the end we have discussed the
questions and the proposed theorem to understand the answer(section 6).

12.2 Complexity
Computational complexity is the study of the space and time resource re-
quired to solve the computational problem. Complexity classes are the group
of problem that can be solved with same resource. Suppose we are having
n bit input and we want to know wheather the number corrosponding to
input is prime or not? Now the chief distinction made in computational
complexity is between problems which can be solved using resources which
are bounded by a polynomial in n or which requires sources which grows
faster than any polynomial in n.In this case the resources required expo-
nential in size. Thus the entire computational problem is devided into two
parts.
Many computational problems are most clearly formulated as decision
problem, problems with answer yes or no for example if a number m is prime
or not. This is primality decision problem.Although most decision problem
can easily be stated in simple,familiar language but discussion of the general
properties of decision problems is greatly helped by the terminology called
language.
There are different types of complexity classes.Which are the following
1)P-The problems belong to this class can be solved in polynomial time.
2)NP- This stands for ”nondeterministic polynomial time”. Where the
term ”nondeterministic” is just a fancy way of talking about guessing a
solution. A problem is in NP if we can quickly test wheather a solution is
correct.
3)PSPACE- This class consists of those problem which can be solved
using resources which are few in sptial size, but not necessarily in time.
4)BPP- This is the class of problems that can be solved using randomized
algorithms in polynomial time, if a bounded probability of error is allowed
in the solution to the problem.
Among these classes only P and NP are very importent and we will
consider only those complexity classes for our consideration.

12.2.1 NP-completeness
In complexity classes the NP- complete problems are the most difficult prob-
lems in NP in the sense that they are the ones most likely not to be in P.

65
The reason is that if we could find a way to solve any NP-complete problem
quickly,then we can use to solve all the NP problem quickly.
A decision problem C is NP-complete if it is complete for NP, meaning
that
1)It is in NP.
2)It is NP-hard, that is every other problem in NP is reducible to it.

12.3 Reduction
The equivalance between the factoring decision problem and the factoring
problem proper is a special instance of one of the most importent idea in
computer science, an idea known as reduction. We know that some problems
can be viewed as special instances of other problem. A less trivial example of
reduction is the reduction of Hamiltonian Circuit problem to the Travelling
salesman decision problem.

12.3.1 General defination of reduction

A language B is said to be reducible to another language A if there exists a
Turing machine operating in polynomial time such that given an input x it
outputs R(x), and if x belongs to B if and only if R(x) belongs to A.Thus
if we have an algorithm for deciding A,then with a little extra effort we can
decide the language B. In this sense, the language B is essentially no more
difficult to decicde than the language A. Here we will consider many-one
reduction.

12.4 Circuit family

Circuit families are a convenient formalism to use in defining the complexity
classes of languages or function.
Now here we will discuss some properties of the circuit families. 1)Size
2)depth 3)uniformity
A circuit family is a set {Cn : n ∈ N} where each {Cn } is an acyclic
circuit with n Boolean inputs x1 ........xn , and some output gates y1 ........yr .
1)Size-Now {Cn } has size S(n) if each circuit Cn has at most S(n) gates.
2)Depth-It has depth d(n) if the length of the longest path from input
to output in Cn is at most d(n).
3)Uniformity-A family {Cn } is uniform if the function n to Cn is easy to
compute in some sence, means given n gates and the name of a gate g, one

66
can determine all of the desired information about g.
Now there are different kinds of uniformity, among which the two uni-
formity we are using here are
1)DLOGTIME Uniformity- It is one kind of uniformity that when the
input of the uniformity machine has length O(log n), then this notion of
uniformity is usually called ”DLOGTIME Uniformity”.
2)P-Uniformity- For instance if one is trying to model what can be com-
puted efficiency by circuits that are feasible to construct, then polynomial
time would seem to be the right notion of uniformity.

12.5 Different kinds of reduction

AC0 reduction:- A function is said to be in AC0 if there is a circuit family
Cn } of size nO (1) and depth O(1) consisting of unbounded fan-in AND,OR
and NOT gates such that for each input x of length n,the output of Cn on
input x is f(x). And they will obey the condition if x=y then f(x)=f(y).
TC0 reduction:- This is the classes of functions computed by the circuit
families of majority gates of size nO (1) and depth O(1), but they will not
follow the restriction like AC0 .
NC1 ,NC0 :- It belongs to the circuit family of size nO (1) and depth O(log
n) and consisting two fan in AND,NOT and OR gates.

12.6 Main study

The questions that were left open in Isomorphisms theorem we will consider
here. And in the present work Agrawal has given answar of these question by
some theorems.Now for every question he proposed one theorem and proved
all those theorem.
Now the questions are
1) Are the all sets complete for a well known complexity class(e.g.,NP)
under polynomial-time reductions are already complete under AC0 reduc-
tions?
2)Does the uniform version of the isomorphism theorem hold?
3)Is depth-three optimal, or are the complete sets isomorphic under iso-
morphisms computable by depth-two circuits?
Now the theorem proposed by the Agrawal to give the answer of these
questions are
1)There exists a set that is complete for NP under Dlogtime-uniform
AC [mod 2] reductions but not under non-uniform AC0 reductions.
0

67
 2)The sets complete for C(complexity classes) under P-uniform AC0 re-
ductions are all isomorphic under isomorphisms computable and invertible
by P-uniform AC0 circuits of depth-three.
3)There are sets complete for C under Dlogtime-uniform AC0 reductions
that are not isomorphic under any isomorphisms computed by AC0 circuits
of depth-two.

12.7 conclusion
The conclusions we can draw by seeing the theorems. Now from the first
theorem we are getting negetive answer of the first question. That is the first
theorem telling us that sets will not complete under AC0 reduction even if
it is complete for any other polynomial time reduction. The second theorem
is giving answer of the second question. From the second theorem we can
conclude that the uniform version of Isomorphism theorem holds. And the
third theorem is also giving the negetive answer of the thired question,i.e
according to this theorem depth-circuit is optimal.

12.8 References
[1]M. Agrawal, E. Allender and S. Rudich Reducing the circuit complex-
ity: an isomorphism theorem and a gap theorem. J. Coumpt.System(1998)
Sci.57, 127-143.
[2]Manindra Agrawal,Eric Allender,Russel Impagliazzo,Toniann Pitassi,
and Steven Rudich.
Reducing the complexity of reductions
comput.complex.10(2001), 117-138
[3]Michel A. Nielsen and Issac L. Chuang
Quantum Computation and Quantum Information
Cambridge University Press (2000)

68
Chapter 13

Quantum Random Walk

Partha Sarathi Pal, PH04C020

Abstract
This article will give some physical intution about random walk which will pro-
vide a general flavor about the phenomena. It will followed by more rigorous
definition along with some necessary termiologies to give an introduction about
the two main models of quantum random walk. Afterthat some computer sci-
ence and probability background in this matter. Some important algorithimic
results obtained from quantum random walk will be discussed.Here we can see
some differences of classical random walk and quantum random walk on the
basis os results obtained from the walk on the circle. Finally some philosoph-
ical aspects will be discussed with the view point of decoherence. Also some
open questions and future possibilities will be mentioned.

13.1 Introduction
In 1993 Y. Aharanov, L.Davidovich and N.Zagury first time used the term
”Quantum Random Walk”. Let us assume that a particle is on a line and its
position is described by a wave packet | ψx0 i localized around a position x.
The function of the corresponding wavepacket centered around x0 in given
by hx | ψx0 i. The translation of a particle corresponding to the step length l
can be expressed as the unitary operator Ul = exp(−iP l/h̄) where P is the
momentum operator. So that Ul =| ψx0 i=| ψx0−l i.
Let The particle has a spin 12 dof and Sz represents the operator corre-
sponding to the Z-component of the spin. The eigenstates are given by |↑i,

69
|↓i where
h̄ 1
Sz |↑i = |↑i ⇒ |↑i (13.1)
2 2
h̄ 1
Sz |↓i = − |↓i ⇒ − |↓i (13.2)
2 2
Here the matrix representations of this two spin states are given be |↑i =
(1 0)† and for |↓i = (0 1)† . The z-component of the spin is given by
 
1 1 0 1
Sz = = (|↑ih↑| − |↑ih↑|) (13.3)
2 0 −1 2
Then the reduced translation operator of the particle depending on its in-
ternal spin degree of freedom can be given by U = exp(−2iSz ⊗ P l). If the
spin of the particle is |↑i at the initial state . Then the wavefunction is in
the form |↑i⊗ | ψx↑0 i. Then after the operation of the translation operator it
will become |↑i⊗ | ψx↑0 −l i. Similarly for |↓i it will become |↓i⊗ | ψx↓0 −l i But
in reality the spion states appear in superposition states which is given by

| Ψin i = (α↑ |↑i + α↓ |↓i)⊗ | ψx0 i (13.4)

If we operate unitary operator on it then it will become

U | Ψin i = α↑ |↑i⊗ | ψx0 −l i + α↓ |↓i⊗ | ψx0 +l i (13.5)

↑ −iP l ↓ iP l
U | Ψin i = (α |↑i ⊗ e + α |↓i ⊗ e ) | ψx0 i (13.6)

A rotation operator of spin can be given by

 
cosθ −sinθ
R(θ) = (13.7)
sinθ cosθ
If we operate R(θ) from the left then it will become

R(θ)U | Ψin i = [(α↑ cosθe−iP l −α↓ sinθeiP l ) |↑i+(α↑ sinθe−iP l +α↓ cosθeiP l ) |↓i]⊗ | ψx0 i
(13.8)
Now we are doing measurement Mz to establish the state of the particle.
|↑i ⊗ (I − iP lδ↑ ) | ψx0 i

Mz R(θ) U| Ψin i = (13.9)
|↓i ⊗ (I − iP lδ↓ ) | ψx0 i
The probabilities of |↑iand |↓i are given by
2
p↑ = | α↑ cosθ − α↓ sinθ | (13.10)
2
p↓ = | α↑ sinθ + α↓ cosθ | (13.11)

70
and the displacements are given by

α↑ cosθ + α↓ sinθ
lδ↑ = l (13.12)
α↑ cosθ − α↓ sinθ
α↑ sinθ − α↓ cosθ
lδ↓ = l ↑ (13.13)
α sinθ + α↓ cosθ
Here in these two cases the values of displacements can be much larger
than the value of l. If we choose for |↑i case tanθ =| α↑ /α↓ | (1 + ǫ)
with l/∆x ≪| ǫ |≪ 1 then the value of lδ↑ ≈ −2l/ǫ will be larger than l in
several orders of magnitude. This is significantly different from the Classical
phenomena. This exception may be useful in modern quantum information
processing.

13.2 Main Study

13.2.1 The Classical Random Walk
The classical random walk in 1-dimenssion can be defined by a linear 1-
dimensional lattice on which the walker take steps of length l. At every
step the walker has the finite probability(p) of moving to his left and q=(1-
p) moving to his right.this can be decided by unbiased coin toss, where
p=q=1/2 at every step. these probabilities at every step are independent
to its previous steps. P(x,N) is the probability of finding the walker at a
position x =nl where n is the integer after N steps.
The P(x,t) is a binomial distribution which becomes Gaussian if we allow
thwe walk in the to continue for a long time.
1 −(x−µ)2
P (x, t)dx = √ e 2σ2 dx (13.14)
2πσ
√
where µ =p(p − q)N l is the mean value of x and σ = 2 pqN √l is the standard
deviation h(∆x2 )i. When p = q = 1/2, µ = 0 and σ = t. From this√we
can conclude that in time t the walker travels a distance proportional to t.

13.2.2 The Quantum Random walk

The key idea behind the quantum random walk is is to iterate the walk ie
the unitary and rotaion operators are operated repeatedly without doing
measurement. Here once unitary operator and rotation operater operates
then it means one step in the walk. This can be illustrated by two models,
they are

71
 1.Discrete time quantum random walk.
2.contineous time quantum random walk.

13.2.3 Random Walks in Computer Science

Random walks are the corestones of theoretical computer science. It is
used as algorithimic tools to solve varity of problems. they used to provide
a general paradigm for sampling and exploring exponentially large sets of
combinatorial structures.

Examples of Random Walk

The behaviour of an algorithm that use random walk depende on quantities
like mixing time or its expected hitting times between two vertices of the
underlying graph. The later point can be illustrated by the examples of
1. S-T Connectivity
2. 2-SAT

Classical Random Walk

The simple random walk on the indirected graph can G(V,E) can is described
by repeated application in a stochatistic matrix M, where Mi,j = d1i is an
edge of G and di is the degree of i. For a connected and nonbipartite G
the distribution of random walk can be given by p~T =M T p~0 which converges
to a stationary distribution π~0 . For d-regular G the limiting probability
distribution is uniform over the nodes of the graph. we can approach to
this problem by setting up a random walk on a graph whose nodes are the
objects. For our convenience we take the walk in the sample form. Here the
initial point is choosen in randomly. This algorithm is only efficient only
when random walk approaches fast to the limiting distribution.
There are a few definitions of limiting distribution. The frequently used
quantity is mixing time is given by

Mǫ = minT | ∀t ≥ T, p~0 :k p~t − ~π k≤ ǫ, (13.15)

Total variation distance to measure the distance between two distributions

p~,~
q: X
k ~p − ~q k= | p~i − q~i | (13.16)
i

The mixing time is related to the gap between the largest eigenvalue λ1
= 1 of the stochastic matrix M, and the second largest eigenvalue λ2 can be

72
expressed in the following way
λ2 1
≤ Mǫ ≤ (maxi logπ~i −1 + logǫ− 1) (13.17)
(1 − λ2 )log2ǫ 1 − λ2

this relation provides very useful connectionbetween mixing time and the
seconed largest eigenvalue λ2 of the transition matrix M. It actually says
that λ2 is the only eigenvalue of M which matters in the mixing behaviour
of the walk. This is not same in case of quantum case.
Circle: For the case of random walk in the N-circle the mixing time is
quadratic, Mǫ ∼ N 2 l̇og(1/ǫ) The probability of hitting jth point close to 1
is T ∼ N 2 .
Hypercube: For the d-dimensional hypercube the mixing time scales
with d as Mǫ ∼ dlogdlog(1/ǫ). The probability of hitting from one corner
000. . .00 to opposite corner 111. . .11 depends on the dimentionality d as
T ∼ 2d .

Quantum computers and Circuits

The quantum computer works according to quantum mechanics, it is basi-
cally unitary transformation on its state space.
Qubit: The classical computers work on the srings of bits. The quantum
computers acts on qubits. Each qubit has two basis states | 0ior | 1i. A
state can be superposition of the basis states. Physically it can be realized
by spin-1/2 particles.
Gates: In classical computation it consists of several circuits which is
basically made by gates (AND, OR, NOT) which acts on one or more bits.
Here we used to implement the unitary operators physically. They are called
CNOT gates  
1 0 0 0
 0 1 0 0 
CN OT =   0 0 0 1 
 (13.18)
0 0 1 0
The main advabntage is the quantum computers are also able to do the
work of their classical counterpart. The main advantage is that the reverse
unitary computation does require any extra circuitry. A quantum computer
can efficiently implement any discrete time quantum random walkwhenever
the classical computer can implement the classical random walk only.

73
Results from quantum Random Walk
The calssical random walk is independent of initial states - i.e. it looses its
memory. that is not true for quantum random walk. There all transfor-
mations are unitary thus reversible, there it willnot loose its memory and
it is able to recollect its previous state. Thus it does not converge to a
stationary distribution. To give the proper explanation of mixing we have
to introduce some parameter for forgetting. It is called Cesaro limit. The
probability distribution c−t is an average distribution over the measurement
result between 1 and t:
t
−t 1X s
c = p~ (13.19)
t
s=1

With this definition we can see that c~t converge to a stationary distribution.
For the analysis of quantum random walk behaviour we have to follow the
wave nature of the unitary evolution Ut . To observe the classical wave vector
at time t, | Ψt i = U t | Ψ0 i, to a probability vector p~ti . The probability of
measurement of the particle in the position i at time t, p~ti = | h↑, i | Ψt i |2 +
| h↑, i | Ψt i |2 where we write |↑, ii short for |↑i⊗ | ii. Let {(λk , | vk i) :
1 . . . 2N } be the eigenvalues andPeigenvectors of U. The initial state can be
expanded as | Ψt i = Ut | Ψ0 i = 2N t
k=1 ak λk | vk i. Putting all values together
we can get for the ith component
t 2N
1X X X
c~ti = ak al ∗ (λk λl ∗ )s hα, i | vk ihvl | α, ii. (13.20)
t s=1
α=↑,↓ k,l=1

When t → ∞ we get
t 
1
X
∗ s 1 λk = λl
(λk λl ) → (13.21)
t limt→∞ t(1−(λ1k λl ∗ )) = 0 λk =
6 λl
s=1

hence
2N
c~ti →
X X
ak al ∗ hα, i | vk ihvl | α, ii = π~i (13.22)
α=↑,↓ k,l=1
λk =λl

Here π~i = α=↑,↓ 2N 2 2

P P
k=1 | ak | α, i | vk i . from these above two equations we
cam see the significant difference between between classical and quantum
walk. For classical case the stationary distribution π~i is indepenmdent of
the initial state. this is not true for the case for quantun walk.

74
 Another thing is that in classical case the mixing time is dependent on
the seconed largest eigenvalue of the transition matrix M. But in quantum
case the rate of convergence to ~π is dependent on the following terms.
t
1X 1 − (λk λl ∗ )s 1
(λk λl ∗ )s = ∗ ≤ (13.23)
t t(1 − (λk λl )) t | λ k − λl |
s=1

The mixing time is dependent on | λk − λl | and the expression for mixing

time for the walk on the circle is given by Mǫ ≤ N logN ǫ3
. This gives a
quadratic speedup over the classical walk on the circle which mixes in the
time proportional to N 2 .
The difference between these two walks can also be shown with the help
of the results obtained from the walks on hypercube and on oracle separation.

13.2.4 Decoherence
The crucial difference between the quantum and the classical walk are the
quantum coherence which is only available in quantum case. If any one want
to go to quantum to classical one then he has to remove the decoherence
part from quantum one.

13.2.5 Open questions

The quantum random walk on the circle of even degree or on the other
graphs does not converg to the uniform distribution as its eigenvalued are
degenerate. Rather, its stationary distribution depends on the starting state.
The exponential hitting time on both hypercube and glued trees are
dependent on the symmetry of the graph. Rapid hitting may have high
symmetry.
There are lots of graphs that has to be studied in the context of random
walk.
The connection between two quantum walk models are not clear.

13.3 Conclusions
A beautiful framework for quantum random walk is done here. Several
differences of the quantum random walk with its classical counterpart is
given. The quantum random walk can be utilised in constructing new and
faster algorithm.

75
13.4 References
J. Kempe Quantum Random Walk - an introductory overview; arxiv:quant-
ph/0303081, 09 Dec 2005.
M.D.Prabha, Dr. Arul Lakshminarayanan Quantum Random Walk in one
dimension in one dimension (A Project Report), Physics Department, IIT
Madras, April 2005.

76
Chapter 14

Optimality of Grover’s
algorithm

Posimsetti Anandarao PH04C021

Abstract
In this my
√ study report, we discuss about a new method for proving lower boun-
nds (O( N ) quantum queries) on quantum unordered search of N-elelments list
which was most famous Grover’s algorihtm. Thus we show Grover’s algorithm
is optimal.

14.1 Introduction
Before we discuss about ”quantum querymodel”, i will give a breif explaina-
tion about ”query model”. In this model the algorithm calls the input
element (or) elements, each call is known as query. Here the number of
queries that algorithm makes is crucial and it tells about the complexity of
algorithm. Examples for query model are
1. Grover’s Algorithm for quantum unordered search.
2. Period finding problem. etc
Classical query model runs the algorithm with one input each time and
after that it modifies input. Using classical query model desired element
is found with O(N ) queries on average in unorered search of N-element
list. On the other hand quantum query model runs the algorithm with a
superposition of inputs. Constructing quantum query algorithm with out
proving lower bounds on the number of queries that algorithm needs does not
make any sense. Here we discuss about proving lower bounds on quantum

77
unordered search using the method suggested by Andris Ambainis. This
implies that Grover’s algorithm is optimal.

14.2 Main study

Consider a bipartite system H = HA ⊗HI where HI is an ”input space”, HA
is the workspace of algorithm. Let S be a subset of possible inputs {0, 1}N .
HI is spanned by basis vectors |xi corresponding to inputs x ∈ S.
At the begining , the algorithm part is in its starting state |0i, the oracle
part is in a uniform superposition of some set of inputs and the two parts
are not entangled. The starting state is
X
|ψstart i = |0i ⊗ αxi |xi i.
xi ∈S

Algorithm performs a seqence of unitary transformations (T queries)

U0 → O → U1 → O → . . . → UT −1 → O → UT .

on the starting state. Where Uj ’s are unitary transformations that does not
depend on the input bits x1 , x2 , . . . , xN . O’s are query transformations that
are also unitary. In each query oracle O operation on superposition of inputs
changes the phase on the |ji component of |ψi.Therefore after a query |ψi
becomes X
|ψ′i = |0i ⊗ { αxi |xi i − αxj |xj i}.
xi 6=xj

Suppose xk be the marked element such that f (xk ) = z, After T queris the
final state will be X
|ψend i = αxi ,z |φxi i ⊗ |xk i.
xi ∈S

Which is an entangled state, |φxi i are algorithm work bits and total number
of qubits in ensillar space is remains the same. Consider reduced density
matrix ( after tracing out by HA ) before a query

ρ = |ψihψ|

becomes
ρ′ = |ψ′ihψ′|
after the query. The diagonal elements of the reduced density matrix doesn’t
change after any query. Only off diagonal elements in j th row and j th column

78
get modified. So summation over off-diagonal elements of ρ changes after
each query. Let’s define sum of absolute values of ρt (reduced density matrix
after t queries) X
St = |(ρt )mn |.
m6=n
√
Let’s choose |αx | = 1/ N
Number of off-diagonal elements in N ×N matrix is N ×(N −1).Therefore
for starting state
N
X
S0 = |(ρ0 )mn |
m,n=1,m6=n
N
X
= 1/N
m,n=1,m6=n
= N (N − 1) × 1/N
= (N − 1).

Let’s find out ST assuming algorithm gives correct answer with probability
at least 1 − ǫ. Let |φxi i and |φxk i be the final algorithm work bits for the
inputs xi and xk . Let {|νi} be the basis for algorithm work bits.Then
X
|φxi i = aν,z |νi,
ν
X
|φxk i = bν,z |νi.
ν

mnth element of reduced density matrix will be

X
(ρend )mn = αm αn a∗ν,z bν,z .
ν,z

Define error probability

X X
ǫ= |aν,z |2 = |bν,z |2 .
ν,z:z6=f (xi ) ν,z:z=f (xi )

79
Consider
X X
| a∗ν,z bν,z | ≤ |aν,z ||bν,z |
ν,z ν,z
X X
≤ |aν,z ||bν,z | + |aν,z ||bν,z |
ν,z:z=f (xi ) ν,z:z6=f (xi )
s X s X s X s X
≤ |aν,z |2 |bν,z |2 + |aν,z |2 |bν,z |2
ν,z:z=f (xi ) ν,z:z=f (xi ) ν,z:z6=f (xi ) ν,z:z6=f (xi )
p p
≤ ǫ × (1 − ǫ) + ǫ × (1 − ǫ)
p
≤ 2 × ǫ × (1 − ǫ).

Therefore ,
p
|(ρend )mn | ≤ |αm ||αn | × 2 ǫ(1 − ǫ)
p
≤ 1/N × 2 ǫ(1 − ǫ)
p
ST ≤ 1/N × 2 ǫ(1 − ǫ) × N (N − 1)
p
≤ 2 ǫ(1 − ǫ)(N − 1).

Till now we determined S0 and ST the deference is

p
(S0 − ST )min = (1 − 2 ǫ(1 − ǫ)) × (N − 1).

This is the minimum change in sum of off-diagonal elements of ρ after first

T queries. If we devide this value with the change in St for a single query we
get the minimum value of T that is nothing but lower bound on our query
model. Now our task is to find St−1 − St for any t ∈ {1, 2, . . . , T }.
X X
St−1 − St = |(ρt−1 )mn | − |(ρt )mn |
m6=n m6=n
X
≤ |(ρt−1 )mn − (ρt )mn |.
m6=n

As we discussed earlier only elements in nth row and nth column are differ

80
in phase, those are α∗m αn in ρ and −α∗m αn in ρ′.There fore,
X X
|(ρt−1 )mn − (ρt )mn | = 4|α∗m αn |
m6=n m6=n
X
≤ 4|αn | |αm |
m6=n
√ sX
≤ 4 N − 1|αn | |αm |2
m6=n
√ p
≤ 4 N − 1|αn | 1 − |αn |2
√
≤ 2 N −1
√
St−1 − St ≤ 2 N −1

Minimum value of T is
(S0 − ST )min
Tmin =
(St−1 − St )max
p
(1 − 2 ǫ(1 − ǫ)) × (N − 1)
= √
2 N −1
p √
= (1 − 2 ǫ(1 − ǫ)) N − 1/2.

Hence optmality of Grover’s algorithm is proved.

14.3 Conclusions
√
Using quantum query model optimality of Grover’s algorithm ( O( N )
quantum queries for searching unordered list of N-elements) is proved.

14.4 Reference
Andris Ambainis, arXiv:quant-ph/0002066 v1 24 Feb 2000.

81
Chapter 15

RSA Cryptography

Raj Kumar, PH04C022

Abstract
RSA is a trusted cryptographic metyhod to send encrypted data through a
public channel. It assures that Eve the eavesdropper can never be able to get
the message provided used prime numbers have been selected wisely.

15.1 Introduction
RSA way of cryptography was introduced by three great minds Rivest,
Shamir and Aldeman in the year 1977. It was so great that even now
three and half decade later people are using it offcourse with trust. Method
envolves the selection of two prime numbers with some restictions depend-
ing on factorization methods known. Idea lies on the fact that it is quiet
impossible to find the factors of the product of two using classical computer
in limited time. For instance a challenge put in 1978 to factorize

RSA-129 =
Decimal Digits= 129
Decimal Digit Sum=105443

1143816257578888676692357799761466120102182967212423625625
6184293570693524573389783059712356395870505898907514759929
0026879543541

with cash prize amounting100$ till1988butwasbrokeninyear1994

82
butnowstillsomechallengese.gonewith:
DecimalDigits : 309

13506641086599522334960321627 88059699388814756056670275244
85143851526510604859533833940 28715057190944179820728216447
15513736804197039641917430464 96589274256239341020864383202
11037295872576235850964311056 40735015081875106765946292055
63685529475213500852879416377 32853390610975054433499981115
0056977236890927563

DecimalDigitSum : 1369existsf or10,000$.

15.2 Main Study

15.2.1 Motivation and Procedure
Motivation is to send secure milittary data, banking account number, pass-
word which can be used for reverse engineering, millitary and financial gains.
RSA cryptographic technique can be depicted as:
Alice wants to send a encrypted message to Bob. She posts a message
through public channel that she is willing to send secure message to Bob. In
response to this Bob announces e,n publicly. Alice uses e,n to encrypt her
message ’m’ to ’c’ and sends to Bob which he decrypts and gets message.
Eve the eavesdroper contantly vigils the communication channel and gets
e,n,c but despite her best effort she can not retrive the message unless she
has exponential high computing facility.

15.2.2 Algorithm
Example:

1. Bob takes p= 885320963 ; q= 238855417

2. Calculates n=p*q= 211463707796206571 and V=(p-1)*(q-1) =

83
 3. Evaluates e= 9007 such that e and V are coprime.
Bob sends e, n to Alice through public channel.

4. Alice encrypts message m= 30120 (m < n). to c= me [mod n] =301209007 [mod

n] =113535859035722866[mod n]
Bob receives c.

5. He calculated ’d’ from: d*e=1 mod[V] =¿ d=116402471153538991

6. Bob decrypts the message as k= cd [mod n]

= 113535859035722866116402471153538991 [mod 211463707796206571]
= 30120 =m.

15.2.3 Attacks on RSA

If ’n’ and ’V’are known:
If somehow Eve manages to know n and V then she can decrypt message c
sent to Bob.
She follows the following procedure:

1. She formulate equation: X2 -(n-V+1)X+n=0.

2. Calculates roots of equation:

X= (n-V+1) + [(n-v+1)2 -4*n]1/2
Since n-V+1 = p*q-(p-1)*(q-1) +1 = p+q and n=p*q

So roots are p and q. Now Eve can follow same procedure as Bob to find d
and thus can decrypt the message.

If r can be calculated from me∗r = 1 mod n

In that case we concludes that me and n are co-prime for a ’r’ and ’r’ divides
V according to CRT (Chinese Remainder Theorem). By evaluating d a
multiplicative inverse of mod[r] such that e*d = k*r+1. (for any integer k
)
Eve can decrypt the message ’c’ as: ce∗d (mod n) = c1+k∗r (mod n) = c
k∗r
* c (mod n) = c (mod n) So, Eve can retrive the message.

84
Time Attack by Kocher 1995
Kocher says if Eve knows the hardware of Bob’s computer by that he means
time required for computing a calculation. She assumes message has to
converted to binary form and she formulates a algorithm necessarily used
by Bob for decryption. According to her :

1. Let bk be the Alice’s message in binary form to be decoded by Bob.

2. If bk = 1 then bit will be decoded and extra time (decoding time)will

be consumed and if bk =0 usual time spends.

3. Let t1i = time(extra) taken to decode a bit ie. for bk =1. t2i =usual
processing time taken by computer i.e for bk =0 and bk =1. Ti = t1i
+t2i = total time used to process a bit.

So, if Eve knows every Ti then she can estimate the status of bk and so
the message.

15.3 Conclusion
RSA is a promising method being used for cryptography. It’s beauty lies
in selection of of prime numbers which offcorse should be choosen with pre-
cautions as p,q don’t have smaller prime factors, p-q shouldn’t be small etc.
These conditions comes from the known factrozation methods.
Although it is not perfect and security can be breached by any of the men-
tioned methods. But still it is the most reliable cryptographic method.

15.4 References
1. www.rsa.com.
2. www.wikipedia.org.

85
Chapter 16

Experimental Tests of Bell’s

Inequality

M. S. Ramkarthik, Roll number PH06D006

Abstract
A review is presented regarding the experimental verification of the Bell’s in-
equality and thereby the implication of it to locally realistic theories. We have The
essentially described the Aspect-Grangier-Dalibard-Roger Experiments using
single photons.
famous Bell’s inequality was derived by Clauser - Shimony in the popular
form in terms of the ensemble averages of the properties which are intrinsic
to a quantum system.It’s mathematical form is as follows.

E(QS) + E(RS) + E(RT ) − E(QT ) ≤ 2 (16.1)

Here Q,R,S and T denote the objective properties of the system which can
assume any form depending on the physical system under consideration and
E denotes the ensemble averages of those obective properties.
When these ensemble averages are calculated for a specific physical sys-
tem we find that the Bell’s inequality is violated.
The point to be noted here is that,the main inputs which went for de-
riving the Bell’s inequality are
1. The assumption that the physical properties of the objective param-
eters exist independent of observation which is nothing but the assumption
of realism

2. Changing any parameter by measurement does not affect any other

86
parameter, Meaurement is local. This is the assumption of locality

When experiments were performed to measure the LHS of the Bell’s

inequality it boiled down to a number which is greater than 2. This means
that Locally Realistic theories like hidden variables are not correct.
Since the Bell’s inequality challenged Quantum Mechanics, Aspect and
his colleagues took the task of performing an experiment using single photons
using specific polarizations to verify the truth of Bell’s inequality.
Aspect Experiments consists basically of counting the coincidences of
photons emmited by a cascade process from an excited calcium atom when
made to pass through suitable polarizers.
Once we measure the coincidence rates between the arrival and the de-
tection of photons for the various orientation of the polarizers we can recast
the Bell’s inequality in the form

R++ (a, b) − R+− (a, b) − R−+ (ab) + R−− (a, b) = E(a, b)expt (16.2)

Where the R’s Refers to the coincidence rates of the photons in the coun-
ters and + corresponds to the vertically polarized photon and - corresponds
to horizontally polarized photon.
So the Bell’s inequality in the case of one photon experiments as de-
scribed above will have the form in terms of E as

E(ab) + E(cb) + E(cd) − E(ad) ≤ 2 (16.3)

Where a,b,c,d refers to the four different orientation of the polarizers.

With these rudimentary warm up let us discuss the experimental situa-

tion in a greater detail.
The experiment consists of of a source emitting 2 correlated photons of
opposite circular polarizations,in Aspect’s experiment Doubly excited cal-
cium was used to get the photons in the opposite circular polarizations.
These 2 photons propogate along the opposite directions and encounter
2 polarizing analysers P A1 and P A2
as shown in the figure.

Now these analysers will make the circularly polarized photon into ver-
tical polarization state and horizontal polarization state. the transmitted

87
being the vertical polarization state and the reflected being the horizontal
polarization state for P A1 and the same for P A2
The vertically polarized photon state and the horizontally polarized pho-
ton state as described above are made to fall on 2 photomultipliers and the
coincidence singles from four photo multipliers, two for each analyser and
the coincidence rates are counted.
These analysers are mounting on a platform which can be rotated about
an optical axis, so that the orientation of these analysers can be changed as
per will. The coincidencence circuit is to look for coincidences in arrival and
detection of photons A and B within 20 nanosecond time window.
If a and b are the two orientations of polarising analyser, then the angle
between them is b − a and the ensemble average

E(a, b) = cos2 [b − a] − sin2 [b − a] = cos[2(b − a)] (16.4)

The experiment was done for four orientations of the analysers, two
orientations for each. The values a and c correspond to P A1 and b and d
for polariser P A2 . The angle between the P A1 and P A2 were calculated
for each experiment involving two orientation. When the LHS of Bell’s
Inequalities are calculated for these parameters we get the value as 2.828.
This clearly disproves the Bell’s Inequality which implies Quantum The-
ory is inconsistent with Local Reality.
Aspect measures the coincidence rates for the specific arrangement when
P A1 is in orientation a, and P A2 is in orientation b. Now the coincidence
rates were calculated for the combination of horizontally and vertically po-
larised states of the photons and E(a,b) was calculated from the coincidence
rates using Eq(2).
In this case the LHS of the Bell’s inequality was found to be 2.697±0.015.

This clearly proves that the results are in favour of Quantum Theory
aginst the Locally Realistic theories like hidden variable theories.
It is a possibility that the photons were influenced before the experiment
was setup to avoid the photon from knowing before which path it will take.
Two acousto-optical Raman Nath cells were kept before the photons were
incident on the analysers. Even in this case the LHS of Bell’s inequality was
calculated to be equal to 2.404 ± 0.080 which is again a very clear violation
of Bell’s inequality.

88
 Figure 16.1: Diagram

16.1 Conclusion
1. Quantum Theory is saved.
2. Either Reality or Spooky Action at a distance with super luminal com-
munication.
3. Hidden variable theories not possible because they are locally realistic.
4. However the truth in the Aspect’s experiment cannot be taken as 100
percetn certain as the photons involved in the experiment can be emitted
with right physical characteristics as per hidden variable theories so as to
reproduce the quantum theory predictions.

16.2 References
• Quantum Theory and measurement : J.A.Wheeler , Zurek (princeton
Univ press)

• Aspect. et.al PRL (47),460,1981

89
Chapter 17

Quantum computation by
anyons

Pramod Dominic, Roll No PH05D015

Abstract
A two dimensional system with anyonic excitations can be considered as a
quantum computer. Unitary transformations can be performed by moving
excitations around each other.

17.1 Introduction
Start with a class of stabilizer codes associated with lattice on a torus.Qubits
live on the edges of the lattice whereas stabilizer operators correspond to the
vertices and faces. These operators can put together to make up a hamil-
tonian with local interaction.The ground state of this hamiltonian coinsides
with the protected space of the code.
Excitations in this model are anyons, means that the global wavefunction
acqires some globalphase factor when one excitation moves around the other.
One can operate on the ground state space by creating an excitation pair,
moving one of the excitations around the torus and anihilating with the
other one.

17.2 Toric codes

Consider a k × k square lattice on a torus. Let us attach a spin or a qubit
to each of the edges of the lattice. Total number of qubits will be 2k2 . For

90
each vertex S and face P consider operators of the following form.
Y Y
As = σjx , Bp = σjz (17.1)
j∈S j∈P

This operators commute each other since S and P have either 0 or 2

common edges .The operators As and Bp are hermitian and have eigen
values 1 or -1.
Let N be a hilbert space with n = 2n2 qubits. Define a protected
subspace L ⊂ N as

h = {|ǫi ∈ N : As |ǫi = |ǫi, Bp |ǫi = |ǫi ∀ s, p} (17.2)

This construction gives the information of a quantum code TOR(K)
called toric code.The operators As , Bp in this code are called stabililizer
operators. By making use of the relations between stabilizer operators, one
can find the dimensionality of the space L
Y Y
As = 1, Bp = 1 (17.3)
s p

So there are m = 2k2 − 2 independent stabilizer operators which gives

the dimensionality of the space L = 2n−m = 4. However there is a more
instructive way of computing the dimensionality. Let us find the algebra
L(L) of all operators in the space L. Which will give complete information
about the space L. Let F ⊆ L(N ) be the algebra of operators generated
by As and Bp .ClearlyL(L) ∼ = G/I, where G ⊃ F is the algebra of all the
operators which commute with As and Bp and I ⊂ G is ideal generated by
As − 1 and Bp − 1. The algebra G is generated by operators of the form
Y Y
Z= σjz , X = σjx (17.4)
j∈c j∈c′

where c is a loop on the lattice where as c’ is a cut, i.e. aloop on the

dual lattice. If the loop is contarctible it should be the product of Bp . It
follows that the algebara L(L) is generated by 4 operatorsZ1 , Z2 , X1 , X2
have the same commutation relations that of σ1z .σ2z , σ1x , σ2x . Which shows
that |ǫi corresponds to a state of two qubits.So the protected space is four
dimensional.//
Let us choose basis vectors in the space N by assigning a label zj = 0, 1 to
each edge j.The constraints Bp |ǫi = |ǫi say that the sum of the labels at the
boundary of the space should be zero. More exactly, only such basis vectors

91
contribute to a vector from the protected space. Such a basis is charectarised
by two topological nummbers:sums of zj along the loops cz1 and cz2 .Thus
for each of the4 possible combinations of the 4 possible combinations of the
topological numbers v1 , v2 there is one vector from the protected subspace,

2 −1)/2
X X X
|ζv1 ,v2 i = 2−(k |z1 , ....., zn i : zj = v1 , zj = v2 ,
z1 ,....,zn j∈cz 1 j∈cz 2
(17.5)
One can always create linear combination of these vectors.

17.3 Abelian anyons

Consider a hamiltonian of the following form,
X X
H0 = − As − Bp (17.6)
s p

One can always classify low energy excitations of the Hamiltonian. Eigen
vector of H0 is is also eigen vectors of As andBp . An elementary particle is
Q As |ζi = |ζi Bp |ǫi = |ǫi is violated because
created if one ofQthe constraints
of the relations s As = 1, p Bp = 1
So it is imposiible to create a single particle. However, it is possible to
create two particle state of the form |ψ x (t′ )i = S x (t′ )|ǫi |ψ z (t)i = S z (t)|ǫi
where|ǫi is an arbitrary ground state.
In the first case two particle are created at the end points of the string
t such particles live on the vertices of the lattice, called Z-type particles.
Correspondingly X-type particle live on the phases .S z (t), S x (t′ ) are called
string operators. They commute with everyAs and Bp except a few ones.
Let us see what happens if the particles move around the torus. Mov-
ing Z-type partilces alone a path czi or cz2 is equivalent to applying the
operatorZ1 orZ2 . Thus we can operate on the ground state space by creat-
ing a particle pair, moving one of the particle along the torus and anihiliting
it with the other one. Thus we can realize quantum gates. In the presence
of perturbation two particle state will not be an eigen-state anymore. The
propagation process is described by Schrodinger equation with some effec-
tive mass m − z. In the non-perturbed modelmz = mx = inf there are no
particle in the ground state but they can be created and anhilated particle
virtually. A virtual particle can tunnel through a torus before anihilating
with each other. Such processes contribute termsbz1 Z1 , bz2 Z2 , bx3 X1 , bx2 X2 |
to the ground state of effective Hamiltonian.

92
 If we move partilce around each other(for this we don’nt need a torus.
We can work on the plane). Moving X-particle around Z-particle gives a

|ψinitial i = S z (t′ )|ψ x (q)i (17.7)

and
|ψf inal i = S x (c)S z (t)|ψ x (q)i = −|ψinitial i (17.8)
Global wavefunctios acquires a phase factor -1. It is not like usual particle
boson and fermions which do not change sign in such a processes. Particles
with this unusual particles called abelion anyons. Abelion anyons are one-
dimensional representaion of braid groups. Note that anyons exist in real
solid state systems(eg. fractional quantum hall effect).
The operators Z1 , Z2 , X1 , X2 can be realized by moving particles along
the loops cz1 , cz2 Z2 , cx1 , cx2 |.These loops only exist on the torus, not on the
plane. Consider, however the process in which an x type and z type particle
go around the torus and trace their path backward. This corresponds to an
operator on the plane. Indeed, we can deform particles trajectories so that
one particle stays at rest and other going around it. Due to the anyonic
nature of the particles,We see that the X1 and Z1 anti commute.

17.4 Conclusion
It has shown that anyons can arise from a hamiltonian with local interac-
tions. these anyons can be used to perform universal quantum computation.

17.5 Reference
• Fault tolerent quantum computation by anyons,A.Yu. Kitaev.

93
Chapter 18

Privacy Amplification by
Public Discussion

Raviteja Upadrashta, EE05S024

Abstract
We are going to be discussing how a public (void of privacy) channel with per-
fect authenticity can be used to remove the defects of a channel with imperfect
privacy and no authenticity. We discuss about the protocols Bob and Alice
can carry out to decide with a high probability 1) the communication between
them has been corrupted by Eve’s tampering and channel noise; and 2) if these
errors introduced are not too severe, to correct the errors introduced in the
transmitted sequence giving out minimum information to Eve. We show that
this information leaked to Eve is less than 1 bit.

18.1 Introduction
Alice intends on sending a secret sequence of bits (called the key) to Bob over
an imperfect channel in the presence of a third party, Eve, trying to gain as
much information about the secret key as possible. The channel is imperfect
in the sense that 1) transmission errors can occur, 2) Eve gains partial infor-
mation by eavesdropping over the channel and also 3) Eve has the capacity
to tamper with the transmission by modifying, injecting new information,
jumbling up the order of the transmission, replacing the transmitted bits
with bits of her choice and/or completely suppress the communication be-
tween Alice and Bob. The only limitation that we impose on Eve is that
Alice and Bob know exactly about the maximum amount of information
that Eve can gain by tampering and eavesdropping.

94
 We intend to look at a series of protocols followed by Bob and Alice over
this given imperfect private channel without authentication and a public
channel with authentication to agree upon a bit string (key) about which
Eve has nearly (less than 1 bit) or exactly no information except for the
length of the final key. All this is done at the cost of reducing the length of
the sequence, sometimes, dramatically.
Authentication is the property of the channel which enables one to iden-
tify the person from whom he/she has received a message over the channel.
It is some kind of a unique address which identifies the sender and it cannot
be tampered with in anyway by a third party.
Tampering and eavesdropping are two different operations. By tamper-
ing with the transmitted bits Eve doesn’t gain any information directly.
Visualize this situation like Eve has an inverter in her possession and she
blindly applies this inverter to the incoming bit sequence which she has no
information about. She has tampered with the bit sequence and since she
had no clue to what the inputs were she now has no clue to what the outputs
are.
The amount of computing power granted to Eve is unlimited, even
though no channel performs quite this badly, is because if it works for this
case then it must work for practically all cases. However Eve can disrupt
communication between the two parties but she can never fool(except for
a very small probability) both of them into making them believe that they
have a 100% correlated sequence.

18.2 Main Study

18.2.1 Detection of transmission errors and tampering
Let Alice send a bit string x of length N over the private channel and let
y be the string of length N received by Bob. The string y received by Bob
might be different from Alice as there could have been transmission errors
and also Eve could have tampered with the bit sequence x. In such a case
how do Alice and Bob detect whether the strings x and y are equal or not?
A simple and practical way is for Alice to choose a random function
f : {0, 1}N → {0, 1}K . Where K is a security parameter. Note that the
amount of partial information leaking to Eve depends on only K and not on
the length of the string N. Alice then computes f(x) and sends the value and
the description of the function to Bob. Following which Bob computes the
value of f(y). f(y)=f(x) is strong evidence that y=x. The probability that a
transmission error or tampering by Eve goes undetected is 2−K .

95
 One way of implementing this function is for Alice to perform a random
permutation on the string x i.e jumble up the order of the bits in x and
then reveal the last K bits of the permuted N bit string. Bob does the same
operations on his string and checks if the last K bits of his permuted string
and Alice’s K bits are identical.
Note that Alice must send the description of the permutation or in gen-
eral the function f and the K bits i.e the value f(x) only after Bob confirms
the reception of the N bit string y. This is because Eve can modify (N-K)
bits in x based on the permutation and can get away undetected. Moreover
after the transmission of the string x over the private channel all subsequent
talks between Alice and Bob must take place over the public channel as the
scheme presented here and the schemes that are presented in sections ahead
rely heavily on authentication.

18.2.2 Reconciliation
Alice and Bob have detected that there are errors and x 6= y. So now their
task is to correct those errors. How do they achieve this?
If the number of errors are less i.e. around two then Bob can compute
all strings Z that differs in 2 places from y and then calculates f(z). A match
f(z)=f(y) indicates that Z is the actual string x sent by Alice to Bob. This
process is called bit twiddling.
When the number of errors are larger Alice can randomly choose an
error correcting code C and then she computes the parity bits C(x) and
sends C(x) and the description of C over the public channel. This implies
eve has full knowledge of the parity bits and effectively has at most C(x)
amount of information. Thus effectively Alice sends xC(x) to Bob receives
yC(x). Bob can now find x successfully. Another variation of this method
is for Alice and Bob to agree on a permutation which transforms x and y to
x0 and y0 .

18.2.3 Reduction of eavesdroppers information

From the above discussions it is clear that Eve has 2 sets of information
about x. One is from partial information leak by eavesdropping when x was
being sent over the private channel and complete eavesdropping when the
error detection protocol was being carried out over the public channel.

96
Elimination of eavesdroppers information over the public channel
Let us assume for the time being that Eve can only eavesdrop over the public
channel. Now in the error detection protocol Alice announced f(x) of length
K bits over the public channel. Thus, we can say that Eve gains K bits of
information (Shannon sense. Don’t confuse be K bits of information to be
K physical bits of x.). So if Alice can select a function g : {0, 1}N → {0, 1}R
where R ≤ N − K and send its description to Bob over the public channel
then they can agree on the final string to be g(x) of length R bits. Then
from the below 2 theorems we can say that Eve has nearly or exactly no
information about the final string. Thus from the second idea we can see
that by chopping of any K+S physical bits the information about the final
−S
string know to Eve is reduced to 2ln2 .

Theorem 1: Let N be the length of the originally transmitted bit string

and let K < N be the safety parameter used for error detection. Let π :
{0, 1}N → {0, 1}N be a randomly chosen permutation . Let f : {0, 1}N →
{0, 1}K and g : {0, 1}N → {0, 1}R be equitable functions defined by f (x) =
π(x) mod 2K and g(x) = π(x) div 2K . Then knowledge of π ( hence of f
and g ) and f (x) yields no information on g(x), except that it is of length
N − K.

A function is said to be equitable if 1{x|f (x) = a} = 2( j −i) for every binary

string a of length i where i < j and f : 0, 1j → 0, 1i

Theorem 2: Let N, K, R, S, F, G, Y and Z where N and K are as in the-

orem 1. Let S be any non-negative integer smaller than N − K. Let
R = N − K − S. Let X, F and G be three independent uniformly dis-
−S
tributed random variables ranging over {0, 1}N . Then I(Z; Y F G) < 2ln2

Elimination of eavesdroppers information over the private channel

In this section we assume that Eve eavesdrops only on the private channel.
Let Eve eavesdropping on the private channel be represented by a function
e : {0, 1}N → {0, 1}K Alice and bob have to come up with a function g
with parameters N, K, R, S as given in the previous section. If you observe
carefully we can see that e performs the role of function f in the previous
section but the difference being that Alice and Bob have no clue as to what
the function e is. Thus, it is not possible in general, for Alice and Bob to
eliminate Eve’s information about g(x) completely.

97
Theorem 3: Let e be any function , let S < N − K be a safety parameter,
and let R = N − K − S. If g is chosen randomly, the expected amount of
−S
information on g(x) given by knowledge of e, g, and e(x) is less than 2ln2
bits.

18.2.4 putting the concepts together

If eve obtained K bits of information on the string x over the private channel
and some additional L bits of information over the public channel then Eve
can have at most K + L bits of information about x.

18.3 Conclusions
If no eavesdropping occurred over the private channel, it is possible for Alice
and Bob to publicly verify that no transmission errors or tampering occurred
either, with a 2−K error probability, and end up with an entirely secret final
string that is only K bits shorter than the original private transmission.
This is optimal.
If a partial eavesdropping occurred over the private channel, leaking
upto K bits of information to Eve, in Shannon’s sense, it is still possible for
Alice and Bob to publicly verify that no transmission errors or tampering
occurred, with a 2−L error probability, and end up with a final string that
is K + L + S bits shorter original private transmission, on which eve has less
−S
than 2ln2 bits information on the average.

98
Bibliography

[1] Charles H. Bennett. Privacy Amplification By Public Discussion. SIAM

Journal Of Computing , Vol. 17, No.2, April 1988.

99