Lecture 19: How to Build Your Own Quantum Computer

Guest Lecturer: Isaac Chuang

Scribed by: Fen Zhao

Department of Mathematics, MIT

November 13, 2003

1 The DiVicenzo Criteria

The DiVicenzo Criteria list four things required for quantum computing: robust qubits, a universal
gate set, a fiducial input state, and projective measurements.

1.1 Robust Qubits

A quantum qubit is based on 2 level quantum systems. One must also remember that a tensor
product Hilbert space is needed; for example a harmonic oscillator is not a tensor product space
and therefore makes a bad quantum computer. In general, a two level atom would make a good
quantum computer.
One must also have a long coherence time. This is characterizes how the environment interacts
with your ideal qubit system. The imperfect qubit system will have many states besides the two
you are interested in. One can think of decoherence as the effects of all the interactions outside
your idea set of interactions.

|Ψ(t)� = e−iHt |Ψ(0)�

⎡ ⎤
a b g ...
⎢c d h ⎥
H = ⎢e f i
⎢ ⎥
⎥
⎣ ⎦
.. ..
. .

In the Hamiltonian above, the elements in bold represent the your ideal set of interactions, and
everything else is the non­ideal part that causes decoherence.
There are many sources of decoherence. Gravity causes decoherence if one states weights more
thant the other. There may be stray long range fields, typically associated with charge. There can
be leakage into larger Hilbert spaces; a two state atom may have higher energy levels that the state
can move to. However, there does exists true finite spaces in nature, such as spin.
There are two measurements of decoherence, T1 and T2 . T1 is called the “longitudinal coherence
time,” or the “spin lattice time,” or the “spontaneous emission time,” or the“amplitude damping.”
It measures the loss of energy from the system. One can do an experiment to determine T1 . First

1
P. Shor – 18.435/2.111 Quantum Computation – Lecture 19 2

T1 T2

1 1
probability
of being in
probability |0> state
of being in
|1> state 1/2

time time

Figure 1: expected results for experiments for T1 and T2

initialize the qubit to the ground state |0�. Then apply X = |0� �1| + |1� �0|, and wait for time t
and measure the probability of being in the |1� state. We expect an exponential decay e−t/Ti .
T2 is called the “transverse coherence time,” or the “spin­spin relaxation time,” or the “phase
coherence time,” or the “elastic scattering time,” or the “phase damping.” One can do an experi­
ment to determine T2 . First initialize the qubit to the ground state |0�. Then apply the Hadamard
transform H to get the state to |0�+
√
|1�
2
, wait for time t, apply H again, and measure the probability
of being in the |0� state. We expect that the measurement goes to 1/2 after a long time because
after a long time, most likely something popped the state into either |0� or |1� state, which after
the H transform sends the state to |0�±|1�
√
2
.
In general T1 > T2 .

1.2 Universal Gate Sets

There are many different universal gate sets. We have seen that CNOT and single qubit gates form
a universal gate set. CNOT, the Hadamard gate, and π/8 gate is another universal gate set.
In practice experimentalists use controlled Hamiltonians that they turn on and off for certain
time intervals.

H = Hsys + P1 (t)X + P2 (t)Y...

For example, for an NMR setup, we may have

π�
H= Jσz ⊗ σz + P1 (t)I ⊗ σx + P2 (t)I ⊗ σy ...
2
In the real world, there is no such thing as time dependent Hamiltonians. So how do we
perform a sequence of operations? It is a just an approximation; fundemantally we will always have
decoherence and will need fault tolerance.
What happens is that our classical controls are actually quantum systems, and we must take
into action the back action of the control system on our system. P1 (t)X is just an approximation;
in reality, we have a Jaynes­Cumming type interaction Hamiltonian:
H = �ωN + δZ + g(a† σ− + aσ+ )
where σ� = X±Y2 . In less abstract terms, there is decoherence that results after a photon interacts
with a qubit because the photon will carry away information about the state of the qubit.
P. Shor – 18.435/2.111 Quantum Computation – Lecture 19 3

2 Implementations
In general, the challenge of quantum computing lies in the fact that quantum systems have short
lifetimes, and that we need to control it externally.

System T (sec)
NMR 102 to 108
Ion Trap 10−3
Dots 10−6
Microwave Cavity 100
Optical Cavity 10−5

Table 1: Some relaxation times for different implementations

2.1 Cavity QED

The Hamiltonian is described by atom, photon, and atom­photon interactions. The qubit is the sin­
gle photon. Gates for π/8 and Hadamard are implemented by beam splitter and the like. Turchette
managed to achieve a control Z gate.

2.2 Ion Trap

The Hamiltonian is described by spin, photon­phonon (vibrational mode ) atom­photon­phonon
interactions. The qubit is the atom (spin) and the phonon. The Deutsch­Joza algorithm has been
implemented with ion traps. The system has a lifetime of O(1 − 100) (order of) milliseconds. Pulsed
lasers are the universal gate set. The challenge is to get a stable O(10) Hz laser and cooling the
ions to absolute zero.

2.3 NMR
The Hamiltonian is described by spins, spin­spin, and external control photon interactions. The
spins interact with chemical bonds. The gates are implemented by radio frequency magnetic field
pulses. Factoring with 6 qubits has been accomplished. In terms of robustness, 1 H, 13 C, 9 F , 14 N
has coherence times of O(1) seconds.

2.4 All Silicon Quantum Computer

This implementation implants individual atoms into silicon surrounding and controls the atom
electronically. It has a coherence of 100 ms at T = 9.2K. This implementation is scalable and uses
current techniques of classical computer engineering.

3 Quantum Cryptography
Currently two companies make quantum cryptography systems. The purpose of quantum cryptog­
raphy is to make it more likely to detect an eavesdropper. They are based on the fact that an
eavesdropper measuring a quantum system transmitted will collapse the system.
P. Shor – 18.435/2.111 Quantum Computation – Lecture 19 4

We can relate various techniques of quantum computing to features of it’s implementation. Data
compression is related to cooling. Error correction is related to control and T1 and T2 . Noisy coding
is related to precision of measurements. Cryptography is related to entanglement and non­locality.