Resman
Resman
University of Ljubljana
SEMINAR
Quantum computing
1
Contents
1. Abstract 2
2. Introduction 3
3. Qubit 3
3.1. Big molecules and spins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.2. Ion traps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
5. Algorithms 10
5.1. Deutsch’s problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
6. Errors 12
6.1. Error correcting codes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
7. Conclusion 13
8. References 14
1. Abstract
The aim of this seminar is to inform the reader of the basics of quantum computation. How
information is implemented is described as well as actions on logical one and zero in quantum
systems. Efficient algorithms for quantum computers are mentioned. I wanted to focus the
work on how these computers are actually made. Unfortunately each implementation could be
a seminar on its own, so only the basics are covered.
2
2. Introduction
As the world evolves human being is eager to develop and use new, better, faster technologies.
Since R.P. Feynman 1982 showed the potential of quantum systems, the research in this direc-
tion has not been stopped. Boost for the research was the fact of quantum systems being in
linear combination of eigenstates. First just theories on how to use this parallelism were devel-
oped and now we already have simple but functioning quantum computers working with eight
qubits.
3. Qubit
Classical information is stored in bits, strings of ones and zeroes representing numbers, letters,
etc. In quantum informational technology we use quantum bits - qubits[8]. Logical one and
zero are represented by two orthogonal states of the system, which form a computational basis.
So any other state of the system can be written as a linear conbination of these two:
. We can create a superposition of qubits. With two qubits we can for example produce
a product(separable) state
In our discussion we will ignore the normalizing factor, as the equations are easier to compre-
hend without it. It can be, of course, restored easily. Expanding the written example we gain a
superposition of all possible states of the system
at once, if is
As written, a quantum register can store all the numbers between and
the number of qubits in the system. Be aware that if there are qubits in the system, the system
has states. If we follow a notation !"# $! , we get
%& ' (
what in decimal equals
3
Some strictly make a difference between entangled and separable states. Entangled states are
the states, that cannot be written as a product:
' %
There are however some differences i.e. entangled states are harder to produce. Their measure-
ment on the other hand is more accurate [9, p.139].
When seeking for a proper implementation of a computer, we must also consider interaction
between qubits, which is a must. A typical qubit is therefore polarization of a photon, an array
of quantum dots, a nuclear spin or any other state of an atom. As optical systems are big and
therefore not promising, they are not of our interest. We will discuss implementation of qubits
using the last two techniques.
Nuclear spin has been interesting because of its isolation from electronic and vibrational mech-
anisms[2]. A suggestion was to create an apparatus to address single nuclear spins one at a time.
This approach would avoid the thermal problem because by definition a single system is always
in a pure quantum state. Unfortunately the experimental approach for this is very difficult to
realize because of the high sensitivity required. The method would be an atomic resolution
magnetic scanning force probe. The signal induced by a single nucleus is namely extremely
weak.
A better way than addressing single nucleus is to use a bigger molecule, a multispin molecular
system such as (2,3) dibromothiophene.
cies to act on specific qubit. Spins of the same molecule interact through dipolar interactions.
That is the way to a two qubit computer.
As we need a stronger signal we use more molecules of the same kind . Now statistics
come in. In thermal equilibrium the spins of each molecule are arranged in some distribution
of energy eigenstates. The lower the energy the more populated the state is. We describe this
populations with a density matrix [2]. Let us have a look at a matrix for two state (one qubit)
4
6 @
@ N
N and denote the population probabilities for the two energy levels: "
where
Since
, the matrix can be linerised: -
.
-
"
3
"! -
The second part is only a small deviation ( ' 1 ), but this is the thing we measure.
" # '&
-%$
For example of our molecule (two qubit system) the deviation of the density matrix can be
+ , --
-
'& ".)(**
*+ .
Chirac and Zoller[3] have proposed this scheme, which is scalable to an arbitrary number of
qubits. Each qubit is comprised of a pair of the ions’ internal states. Ions are trappe in a
harmonic potential, that has to be narrow, so that the ions do not move a lot.
132
Figure 2: Ions are trapped in the paraboloidal potential[10]. They are a few
apart[4], so that we can adress a single one with a laser. Here two ions
are trapped.
4
Applying a potential , the ions align themselves in the plane of the ring electrode. The ring
electrode is eliptical, so that the ions are aligned on a line.
5
Ions, usually = , are confined in a coaxial resonator based radio frequency (rf) trap, named
Paul trap[5].
Figure 2[4]: Ions are trapped in the ellipsoidal part of the ring electrode.
This kind of trap is capable of strong confinement.
This means, that the ions
2
cannot move(oscillate) more than
exciting radiation( + ).
, where is the wavelength of the
6
4. Gates and networks
Qubits are useless if there is nothing to do with them. That is why we seek for operators which
will somehow calculate with them.
A quantum logic gate is a device that performs a fixed unitary operation on selected qubits in a
fixed period of time.
A quantum network is a device consisting of quantum logic gates whose computational steps
are synchronized in time.
Why does a gate have to be unitary? If we take an arbitrary gate U, which is in general time
dependent and use it on the eigenfunction, we get a propagated function, which has to satisfy
the Schroedinger equation
,".:
?
So we obtain an operator equation
" : and 6 the? solution
".= 1 DGHF 6 ?
If we express the exponential function into series, we see that
" = DGHF And so prove that
is unitary operator if is hermitian ( " 6 ):? 6 ?
" = 1 DGHF = DGHF "
Figure 3[6]: A laser beam is separated and optically lead to a specific ion.
There are two partial beams per ion. These are needed as will be seen futher
in the seminar.
7
4.3. Hadamard gate
For production of superposition of qubits from a pure state we use Hadamard gate.
"
—- H —- 3 29
-
—- H —-
This is how we construct previously described register of all qubits from a pure state
— H — — H —
Notice the effect of using Hadamard twice:
Hadamard gate can be easily implemented in NMR by rotation around axes tilted at
0/
between the z and x axes. Such a rotation can be achieved directly using standard NMR tech-
niques.
When trying to create gates that are not bijective, we must remember that they should be unitary
and so reversible. We must think of a gate that is a logical gate but can be reversed. A simple
gate that does the trick is C-not or XOR gate. It takes two qubits, which can be entangled, so
the computational basis is
( %( '(
This gate flips the second qubit if the first is .
"
/
Figure 5: C-not gate flips the second qubit (spin) if the first is in logical .
Here we see that we have to have an interaction between qubits. This is the only way to flip
the second qubit on basis of the first. In bulk spin-resonance computers this interaction is spin
coupling. In ion traps ions interact with phonons. This is the quantized centre-of-mass motion.
Phonons (interactions) can be controlled with two different angles of the lasers. The photon
gives its momentum to the data bus. When computer absorbs another photon coming from op-
posite direction, the center-of-mass stops.
9
5. Algorithms
Algorithm is a set of instructions which are applied to the system to obtain a solution. The
algorithms are that make quantum computers useful, since some things can be done faster with
quantum parallelism, while other not. The biggest difference is in NP problems. In informa-
tional sciences we have two kind of algorithms, P and NP. First, polynomial, are such that the
time is scaled as a power of input. NP problems (not polynomial) are problems, for which
a polynomial algorithm does not exist. With quantum computers some NP problems can be
solved in polynomial time. This is the only advantage of quantum computers. The most well
known problems of this kind are [1]:
This is the first problem, which was ”invented” to prove the potential of quantum computing[1].
Our aim is to determine whether an unknown one bit function is constant or balanced. This
means does the output of the function change with the input or not.
f(0) f(1)
1 1
We have these four possibilities for the function f: 0 0
0 1
1 0
10
Classically we have to evaluate the function twice, on logical 1 and 0 and from the output
determine the nature of unknown function f. This can be time consuming if the function does
not evaluate quickly. The quantum algorithm determines the nature of the function with its
single evaluation.
As quantum computers use reversible logic, it is not possible to implement directly. So we
design a propagator :
Calculating this operator on superposition yields:
: : , " %(
," ,"
" 2
This way the value of the function stays in the phase factor. With using as a superposition,
we can do better:
: 2 ,"
With knowing how to implement Hadamard gate it is now not difficult to build a network,
solving the Deutsch’s problem.
11
6. Errors
To keep a quantum computer functioning properly we need error correcting codes. There are
many sources of error:
Phase error
The qubit may gain an arbitrary phase. For example
Phase of the whole function is of no significance but a relative phase < I
is catastrophic.
Small errors
tween qubits.
precision
can occur because of limited
' of the
'
instrumentation as well as interactions be-
Cloning
We cannot copy qubit registers.
For correcting we need some extra information and must encode a single qubit using three
qubits $! . This way the probability for an error is bigger, but we know how to correct it.
Correcting bit flip error:
To do this we have to measure two qubits at once. This operation projects the state of the
whole qubit back on itself. Let us have a look at an example
"
". % '
We have to measure ! 2( . The two bit number we get tells us which qubit has
been flipped. In our example that is the third:
! 2(
;" ( 2 ".+)
12
Correcting small errors:
Suppose we have a modified qubit:
" & &
When measuring the same quantities as the previous time ! 2( , we get a projection
of the qubit on itself. With probability we obtain the measurement (0,0), which
-
projects the damaged qubit onto the undamaged qubit . The other possible measure-
ment is (1,1), with probability - which means that is projected onto , what
we can repair.
This algorithms were developed by Shor and do not disturb the qubit. From measurement we
gain information and project the qubit on a set of discrete states. Using the measured informa-
tion we know, onto which state the qubit has been projected on and can reconstruct it.
7. Conclusion
Since Feynman showed the possibilities of quantum computers the research has not stopped.
Theory is well developed. With the three described gates we can do any qubit transformation
we want. Some powerful algorithms have been developed. These are the only advantage of the
quantum computers. They are however useless when we want to calculate everyday arithmetic.
Just some special problems can be efficiently solved by a quantum system. That is the main
reason, why they probably will not be in the stores in the next decade. There are of course other
restrictions, which in contrary can and will be passed. These are implementation problems.
The biggest computer build was a 7 qubit spin system. Now it is known that there is no future
in the spin systems, so attention of the researchers is directed towards ion traps and quantum
dots.
13
8. References
3. Quantum Computation with cold trapped ions; Cirac, Zoller; Physical rev. let., vol 74, p.
4091;1995
4. Cooling the Collective motion of trapped ions to initialize a quantum register; King,
Wood,..; Physical rev. let., vol 81, p. 1525; 1998
9. Lecture Notes for Physics 229: Quantum Information and Computation; John Preskill;
California Institute of Techology; September 1998
10. Quantum Information Processing; Jaewan Kim; Dept of Physics, KAIST; 1998
14