Experimental quantum information processing 43

computers be experimentally feasible? Or might there be some principle of physics which

fundamentally prohibits their eventual scaling? In the next two sections we address these
questions. We begin with a review of the famous ‘Stern–Gerlach’ experiment, which
provides evidence for the existence of qubits in Nature. We then widen our scope,
addressing the broader problem of how to build practical quantum information processing
systems.

1.5.1 The Stern–Gerlach experiment

The qubit is a fundamental element for quantum computation and quantum information.
How do we know that systems with the properties of qubits exist in Nature? At the time
of writing there is an enormous amount of evidence that this is so, but in the early days
of quantum mechanics the qubit structure was not at all obvious, and people struggled
with phenomena that we may now understand in terms of qubits, that is, in terms of two
level quantum systems.
A decisive (and very famous) early experiment indicating the qubit structure was
conceived by Stern in 1921 and performed with Gerlach in 1922 in Frankfurt. In the
original Stern–Gerlach experiment, hot atoms were ‘beamed’ from an oven through a
magnetic field which caused the atoms to be deflected, and then the position of each atom
was recorded, as illustrated in Figure 1.22. The original experiment was done with silver
atoms, which have a complicated structure that obscures the effects we are discussing.
What we describe below actually follows a 1927 experiment done using hydrogen atoms.
The same basic effect is observed, but with hydrogen atoms the discussion is easier
to follow. Keep in mind, though, that this privilege wasn’t available to people in the
early 1920s, and they had to be very ingenious to think up explanations for the more
complicated effects they observed.
Hydrogen atoms contain a proton and an orbiting electron. You can think of this elec-
tron as a little ‘electric current’ around the proton. This electric current causes the atom
to have a magnetic field; each atom has what physicists call a ‘magnetic dipole moment’.
As a result each atom behaves like a little bar magnet with an axis corresponding to the
axis the electron is spinning around. Throwing little bar magnets through a magnetic field
causes the magnets to be deflected by the field, and we expect to see a similar deflection
of atoms in the Stern–Gerlach experiment.
How the atom is deflected depends upon both the atom’s magnetic dipole moment –
the axis the electron is spinning around – and the magnetic field generated by the Stern–
Gerlach device. We won’t go through the details, but suffice to say that by constructing
the Stern–Gerlach device appropriately, we can cause the atom to be deflected by an
amount that depends upon the ẑ component of the atom’s magnetic dipole moment,
where ẑ is some fixed external axis.
Two major surprises emerge when this experiment is performed. First, since the
hot atoms exiting the oven would naturally be expected to have their dipoles oriented
randomly in every direction, it would follow that there would be a continuous distribution
of atoms seen at all angles exiting from the Stern–Gerlach device. Instead, what is seen
is atoms emerging from a discrete set of angles. Physicists were able to explain this by
assuming that the magnetic dipole moment of the atoms is quantized, that is, comes in
discrete multiples of some fundamental amount.
This observation of quantization in the Stern–Gerlach experiment was surprising to
physicists of the 1920s, but not completely astonishing because evidence for quantization
44 Introduction and overview

effects in other systems was becoming widespread at that time. What was truly surpris-
ing was the number of peaks seen in the experiment. The hydrogen atoms being used
were such that they should have had zero magnetic dipole moment. Classically, this is
surprising in itself, since it corresponds to no orbital motion of the electron, but based
on what was known of quantum mechanics at that time this was an acceptable notion.
Since the hydrogen atoms would therefore have zero magnetic moment, it was expected
that only one beam of atoms would be seen, and this beam would not be deflected by
the magnetic field. Instead, two beams were seen, one deflected up by the magnetic field,
and the other deflected down!
This puzzling doubling was explained after considerable effort by positing that the
electron in the hydrogen atom has associated with it a quantity called spin. This spin
is not in any way associated to the usual rotational motion of the electron around the
proton; it is an entirely new quantity to be associated with an electron. The great physicist
Heisenberg labeled the idea ‘brave’ at the time it was suggested, and it is a brave idea, since
it introduces an essentially new physical quantity into Nature. The spin of the electron
is posited to make an extra contribution to the magnetic dipole moment of a hydrogen
atom, in addition to the contribution due to the rotational motion of the electron.

Figure 1.22. Abstract schematic of the Stern–Gerlach experiment. Hot hydrogen atoms are beamed from an oven
through a magnetic field, causing a deflection either up (| + Z) or down (| − Z).

What is the proper description of the spin of the electron? As a first guess, we might
hypothesize that the spin is specified by a single bit, telling the hydrogen atom to go up or
down. Additional experimental results provide further useful information to determine if
this guess needs refinement or replacement. Let’s represent the original Stern–Gerlach
apparatus as shown in Figure 1.22. Its outputs are two beams of atoms, which we shall
call | + Z and | − Z. (We’re using suggestive notation which looks quantum mechanical,
but of course you’re free to use whatever notation you prefer.) Now suppose we cascade
two Stern–Gerlach apparatus together, as shown in Figure 1.23. We arrange it so that the
second apparatus is tipped sideways, so the magnetic field deflects atoms along the x̂ axis.
In our thought-experiment we’ll block off the | − Z output from the first Stern–Gerlach
apparatus, while the | + Z output is sent through a second apparatus oriented along the
x̂ axis. A detector is placed at the final output to measure the distribution of atoms along
the x̂ axis.
A classical magnetic dipole pointed in the +ẑ direction has no net magnetic moment
in the x̂ direction, so we might expect that the final output would have one central peak.
However, experimentally it is observed that there are two peaks of equal intensity! So
perhaps these atoms are peculiar, and have definite magnetic moments along each axis,
independently. That is, maybe each atom passing through the second apparatus can be
 Experimental quantum information processing 45

/ /

/ /

Figure 1.23. Cascaded Stern–Gerlach measurements.

described as being in a state we might write as | + Z| + X or | + Z| − X, to indicate

the two values for spin that might be observed.

/ / /

/ / /

Figure 1.24. Three stage cascaded Stern–Gerlach measurements.

Another experiment, shown in Figure 1.24, can test this hypothesis by sending one
beam of the previous output through a second ẑ oriented Stern–Gerlach apparatus. If
the atoms had retained their | + Z orientation, then the output would be expected to
have only one peak, at the | + Z output. However, again two beams are observed at
the final output, of equal intensity. Thus, the conclusion would seem to be that contrary
to classical expectations, a | + Z state consists of equal portions of | + X and | − X
states, and a | + X state consists of equal portions of | + Z and | − Z states. Similar
conclusions can be reached if the Stern–Gerlach apparatus is aligned along some other
axis, like the ŷ axis.
The qubit model provides a simple explanation of this experimentally observed be-
havior. Let |0 and |1 be the states of a qubit, and make the assignments
| + Z ← |0 (1.56)
| − Z ← |1 (1.57)
√
| + X ← (|0 + |1)/ 2. (1.58)
√
| − X ← (|0 − |1)/ 2 (1.59)
Then the results of the cascaded Stern–Gerlach experiment can be explained by assuming
that the ẑ Stern–Gerlach apparatus measures the spin (that is, the qubit) in the computa-
tional basis |0, |1, and
√ the x̂ Stern–Gerlach
√ apparatus measures the spin with respect to
the basis (|0 + |1)/ 2, (|0 − |1)/ 2. For example, in the cascaded ẑ-x̂-ẑ experiment,
√
if we assume that the spins are in the state | + Z = |0 = (| + X + | − X)/ 2 after
exiting the first Stern–Gerlach experiment, then the probability for obtaining | + X
out of the second apparatus is 1/2, and the probability for | − X is 1/2. Similarly, the
probability for obtaining | + Z out of the third apparatus is 1/2. A qubit model thus
properly predicts results from this type of cascaded Stern–Gerlach experiment.
46 Introduction and overview

This example demonstrates how qubits could be a believable way of modeling systems
in Nature. Of course it doesn’t establish beyond all doubt that the qubit model is the
correct way of understanding electron spin – far more experimental corroboration is
required. Nevertheless, because of many experiments like these, we now believe that
electron spin is best described by the qubit model. What is more, we believe that the
qubit model (and generalizations of it to higher dimensions; quantum mechanics, in other
words) is capable of describing every physical system. We now turn to the question of
what systems are especially well adapted to quantum information processing.

1.5.2 Prospects for practical quantum information processing

Building quantum information processing devices is a great challenge for scientists and
engineers of the third millennium. Will we rise to meet this challenge? Is it possible at
all? Is it worth attempting? If so, how might the feat be accomplished? These are difficult
and important questions, to which we essay brief answers in this section, to be expanded
upon throughout the book.
The most fundamental question is whether there is any point of principle that prohibits
us from doing one or more forms of quantum information processing? Two possible
obstructions suggest themselves: that noise may place a fundamental barrier to useful
quantum information processing; or that quantum mechanics may fail to be correct.
Noise is without a doubt a significant obstruction to the development of practical
quantum information processing devices. Is it a fundamentally irremovable obstruction
that will forever prevent the development of large-scale quantum information process-
ing devices? The theory of quantum error-correcting codes strongly suggests that while
quantum noise is a practical problem that needs to be addressed, it does not present a
fundamental problem of principle. In particular, there is a threshold theorem for quan-
tum computation, which states, roughly speaking, that provided the level of noise in a
quantum computer can be reduced below a certain constant ‘threshold’ value, quantum
error-correcting codes can be used to push it down even further, essentially ad infini-
tum, for a small overhead in the complexity of the computation. The threshold theorem
makes some broad assumptions about the nature and magnitude of the noise occurring in
a quantum computer, and the architecture available for performing quantum computa-
tion; however, provided those assumptions are satisfied, the effects of noise can be made
essentially negligible for quantum information processing. Chapters 8, 10 and 12 discuss
quantum noise, quantum error-correction and the threshold theorem in detail.
A second possibility that may preclude quantum information processing is if quan-
tum mechanics is incorrect. Indeed, probing the validity of quantum mechanics (both
relativistic and non-relativistic) is one reason for being interested in building quantum
information processing devices. Never before have we explored a regime of Nature in
which complete control has been obtained over large-scale quantum systems, and perhaps
Nature may reveal some new surprises in this regime which are not adequately explained
by quantum mechanics. If this occurs, it will be a momentous discovery in the history of
science, and can be expected to have considerable consequences in other areas of science
and technology, as did the discovery of quantum mechanics. Such a discovery might also
impact quantum computation and quantum information; however, whether the impact
would enhance, detract or not affect the power of quantum information processing can-
not be predicted in advance. Until and unless such effects are found we have no way of
knowing how they might affect information processing, so for the remainder of this book