Singe Quantum Algorithms and architectures will ultimately need hardware on which to run, we've explored how the principal experimental efforts are striving to produce it. Even 15 years ago, a quantum computer was generally viewed by computer scientists and physicists alike as an intriguing but probably unattainable theoretical curiosity. But interest exploded in 1994 after Peter Shor, then at Bell Laboratories (now at MIT), published his famous quantum factoring algorithm capable of undermining widely used cryptosystems that relied on the difficulty of factoring large numbers. Today, several thousand physics, computer science, and engineering researchers in more than 100 groups in universities, institutes, and companies around the world are exploring the frontiers of quantum information, encompassing quantum computing, as well as recently commercialized quantum cryptography and quantum teleportation communication techniques. Accelerating progress on virtually all fronts in this worldwide research community is yielding confidence that a practical quantum computer is indeed achievable.

Quantum computing's potential has always been tantalizing: Exponentially scalable computing power that could solve problems beyond the capabilities of conventional computers. The key is exploiting the superposition of quantum-entangled information units, or qubits. But the research challenges are daunting: How to create and reliably compute with the qubits, which require the seemingly mutually exclusive conditions of exquisite classical control while being isolated from any external influences that could destroy the entanglement.

The computing power of a quantum computer grows exponentially with the number of qubits it uses. Dozens to hundreds of qubits will be needed for a quantum computer to solve interesting problems using quantum algorithms (along with appropriate quantum error-correction techniques needed to be sure the answer is correct). The qubits must also be connected by quantum communication channels into logic gates that can be manipulated to implement the algorithms.

However, merely having and connecting qubits is not sufficient for a quantum computer. They must remain entangled long enough to complete the number of gate operations required by the algorithm and mandatory error correction. Faster gate operation, higher fidelity (percentage of gate operations completed correctly), and greater error-correcting efficiency can speed the calculation or reduce the number of qubits needed to solve the problem.

More than a dozen different ways of creating qubits—each with its own strengths and challenges—have been developed to date. The following is a rundown of the leading candidates:

*Ion traps* use electrical and/or magnetic fields and
laser-cooling to create a "pseudo-molecule" quantum register with
micron-scale inter-ion spacing. Typically, the qubit is a
two-level motion mode for a trapped ion. The modes are modulated
by laser pulses. The ion motion acts like a data bus, and gates
are implemented by modulating neighboring ions.

"Our decoherence times can be up to 10 minutes—very long compared with other quantum computing techniques," says Dave Wineland of the National Institute of Standards and Technology, Boulder, CO. "But our gates are rather slow, about five microseconds for our two-qubit gates. " Since factoring a 100-to-200-digit number would require a million operations, even error-free implementation would take far longer than the qubit could be maintained. Researchers, led by Rainer Blatt of the Institut fur Experimentalphysik Universtität Innsbruck in Austria, recently set the record for qubit fidelity: 99.3%.

Integrating CMOS chips with ion traps is a recent innovation that permits quantum communication but uses classical control and measurement. One design created at Lucent by Richard Slusher (now at the Georgia Tech Quantum Institute) and Jungsang Kim (now at Duke) is an ion-trap analogue to the electron-based charge-coupled device chips used in digital cameras. "This design gives addressability to an enormous number of ions," says MIT's Isaac Chuang.

*Quantum dots* are a popular solid-state host for qubits.
Lieven Vandersypen's research group in Delft, The Netherlands,
reported last November that it had used an alternating electrical
field to control single electrons contained in gallium arsenide
quantum dots. Electrical control is more selective than the
previously used magnetic fields. Their fidelity of flipping a
single-electron spin (a simple gate) is a little lower than the
73% attained with magnetic-field control but is expected to
increase as they gain experience. Future research will venture
into no-spin hosts, such as silicon and carbon (carbon-12,
nanotubes, graphene), that are expected to have much longer
decoherence times.

*Linear optic* qubits are created by simultaneously
producing forward and backward photons and encoding their logical
states into vertical and horizontal polarizations. This approach
has the advantage of long decoherence times and compatibility
with fiber optics but needs higher photon-creation and -detection
efficiencies. Last December, Andrew White's group from the
University of Queensland reported that it had used a linear optic
circuit involving four qubits to find the prime factors of 15 (5
and 3) thus demonstrating that system's ability to perform the
core processes required for implementing Shor's algorithm.

In April, Prem Kumar of Northwestern University announced a quantum gate created within an optical fiber. A few years ago, Kumar showed that photons can remain entangled within a fiber for a distance of 100 kilometers. The recent result will be useful in creating quantum repeaters for a distributed quantum information network.

*Superconducting* qubits can be made in three types:
charge, flux, and phase. Each uses excitation states of Josephson
junctions: two superconductor pieces separated by an insulator
thin enough for Cooper pairs of electrons to tunnel across. This
approach is scalable, since superconductivity enables fast
control and readout and large, controllable couplings between
widely separated qubits. However, it also requires extremely low
temperatures—milliKelvins—and tends to have short
decoherence times (to date only a few microseconds). John
Martinis's group at the University of California, Santa Barbara,
recently measured single-qubit fidelity of 98% in a phase-qubit
system. And Robert Schoelkopf's group at Yale developed a
"transmon" qubit 300 microns long and very stable against
noise.

More than a dozen different ways of creating qubits have been developed to date.

*Hybrid approaches* combine the best features of their
parents. Chuang's research group is integrating ion traps into
superconducting qubits. "Ion traps are very hard to connect to
anything," Chuang says. "It would be really nice to have an ion
trap with wires coming in and out." Vandersypen says that as ion
traps get smaller they might eventually look similar to quantum
dots.

*D-Wave Systems*, a venture-backed company based in
Vancouver, BC, made news twice last year when it announced the
operation of a 16-qubit (in February) and 28-qubit (in November)
"adiabatic" quantum computer. However, many scientists are
skeptical of both the claims and their importance. The adiabatic
method, which is a quantum version of simulated annealing,
involves slowly evolving a system toward the solution. Skeptics
say this approach will prove to be neither fault-tolerant nor
scalable. Although the company revealed scant scientific detail
about its approach, its president and CEO, Geordie Rose, says it
has no proof yet of entanglement—the hallmark of quantum
computation.

*Diamond-based systems* are an intriguing recent entrant.
The qubit is the spin state of a nitrogen impurity adjacent to a
vacancy in the carbon crystal. Stemming from Tom Kennedy's 2003
research at the Naval Research Laboratory, "nitrogen-vacancy
color centers" have two compelling advantages: their spin state
can be both initialized and read out optically at room
temperature and weak spin-orbit coupling in diamond makes this
qubit well-decoupled from its environment. Several groups have
jumped in to study this system, and its decoherence time has
increased from about 50 microseconds to nearly a millisecond.

Challenges include devising ways to control individual qubits and couple them together. But David Awschalom of the University of California, Santa Barbara, is optimistic. "If you had told me," he says, "a few years ago that you were going to try to control a single electron at gigahertz frequencies in a solid-state material at room temperature, I'd have said, 'Good luck!' Now we're doing just that."

This system is also useful as an accessible test bed for studying spin interaction in solid-state materials that may contribute to the success of other systems and quantum physics knowledge in general. Awschalom's group recently watched quantum information from a single "nitrogen-vacancy center" spin disappear into the "bath" of spins associated with the much more common nitrogen impurities not associated with vacancies...then reappear. "In quantum physics, this is a big deal," Awschalom says. "It's an age-old problem. There have been 1,000 theory papers on this, but no experiments."

The recent across-the-board progress, however, has stimulated optimism in the eventual success and impact of quantum computing inconceivable 15 years ago. For example, IBM Research's David DiVincenzo, who proved in 1995 that quantum algorithms could be executed using only two-qubit operations and later devised seven widely accepted criteria for a practical quantum computer, says, "I'm confident that the quantum computer will eventually change the world and will deeply influence how information processing will be done in the future."

**©2008
ACM 0001-0782/08/0700 $5.00**

Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.

The Digital Library is published by the Association for Computing Machinery. Copyright © 2008 ACM, Inc.