Silicon chips set to go atomic

By Eric Smalley, Technology Research News

Prototype quantum computers, which use atoms or subatomic particles to carry out computations, are laboratory oddities that run gamut from lasers fired into special crystals to test tubes of liquid placed in the same type of machine that produces medical MRIs.

Quantum computers could be fantastically fast at solving large problems like cracking secret codes, but they remain largely theoretical, which is why researchers are trying different approaches to see if it's possible to make large, practical computers that use the states of subatomic particles to store and process information.

Researchers from Stanford University and Keio University in Japan are exploring familiar ground in an attempt to make quantum devices as similar as possible to ordinary, classical computers. The team has come up with a way to build quantum computers entirely out of the traditional material of computers -- silicon.

"Silicon-based architectures are exciting because of all the engineering that has already gone into silicon technology over the last 40-plus years," said Thaddeus Ladd, a researcher at Stanford University.

Quantum computers use atoms or subatomic particles -- qubits -- to represent bits of information. The nucleus of an atom can act like a tiny magnet, and depending on which way its magnetic field is oriented, it represents a 1 or a 0. Conventional computers use the presence or absence of electric current running through transistors to indicate the 1s and 0s of digital information.

When an atom is isolated from its environment the nucleus is in the weird quantum state of superposition, meaning it is in some mixture of all possible states. A qubit in superposition is a mix of 1 and 0, and a string of qubits in superposition can represent every combination of 1s and 0s at the same time.

The power of a quantum computer comes from this ability to represent and check every combination at once to find the number that represents the solution to a problem. Ordinary computers have to check each possible answer one at a time. "One [quantum] computer does the work of many computers," said Ladd.

In the largest prototype quantum computer to date, researchers used liquid nuclear magnetic resonance (NMR) technology to manipulate seven qubits. NMR, which is the technology behind magnetic resonance imaging (MRI) machines, uses magnetic fields and radio waves to shift and measure the atomic nuclei in the molecules that make up a liquid.

Researchers generally agree that liquid NMR quantum computers, however, cannot be made larger than 10 qubits because the strength of the radio signals from the qubits decreases exponentially compared to the noise with each additional qubit. To live up to its vast potential, a quantum computer must contain thousands or millions of qubits. Today's classical computers contain millions of transistors.

The Stanford and Keio University researchers' proposal also uses NMR technology, but in a solid form. Their idea is to embed atoms of the silicon isotope silicon29, whose nuclei act like magnets, in ordinary silicon, whose nuclei do not. This isolates the silicon29 nuclei, reducing the noise problem.

Another challenge to making quantum computers is preserving the fragile states of qubits. Energy from the environment can disturb the qubits, causing them to decohere. Decoherence is like when a spinning top falls over or is touched so that it changes direction, said Ladd. The longer a qubit remains coherent, the more operations a quantum computer can carry out.

The embedded silicon scheme addresses the decoherence problem as well, protecting the magnetic nuclei so that their quantum states last longer.

The researchers' scheme calls for using a string of silicon29 nuclei to form a qubit. If ordinary silicon wafers were cut so the crystal structure of the edges forms steps, silicon29 atoms deposited on the edges would move to the corners of the steps, making chains, said Ladd.

Measuring the magnetic orientation of a chain of a few hundred atomic nuclei is difficult compared to measuring the magnetization of a group of billions of nuclei in the molecules of a liquid. "Such a small nuclear magnetization has never before been measured," said Ladd.

The researchers' architecture calls for measuring and controlling the qubit chains by putting them in magnetic fields of different strengths so that each chain responds to a different radio frequency. "Essentially each qubit has its own radio channel," said Ladd. "By adjusting the frequency of the radio waves we generate with an antenna, we can tune in to each nucleus. These radio frequency signals can rotate the nuclei any way we want," he said.

The quantum computer would carry out logic operations through intricate sequences of rotations.

The scheme is not the first proposed for making a quantum computer out of silicon. University of Maryland researcher Bruce Kane devised a plan to make qubits out of phosphorous atoms placed at regular intervals in a silicon chip. The qubits in that scheme would be more prone to decoherence and would likely be even more challenging to read than those in the researchers' all-silicon approach, said Ladd.

Kane's proposal may have more promise in a long run, however, because it is more likely to lead to machines with the thousands or millions of qubits necessary to make practical quantum computers, said Ladd. "The big difference is that our proposal is more short-term and realistic, while Kane's is more far-reaching and therefore more uncertain," he said.

The Stanford and Keio University researchers' proposal is interesting and would probably be easier to implement than other types of silicon quantum computers, but it will not scale up to large numbers of qubits, said Kane.

The researchers' work may solve some of the problems in current NMR computers, said Eli Yablonvitch, a professor of electrical engineering at UCLA. Using strings of silicon29 nuclei would have many advantages over using liquids, he said.

However, "it would remain painfully slow, with a clock speed of less than 1 kHz," Yablonvitch said. Magnetic fields cause atomic nuclei to resonate at relatively low frequencies. The radio signals used to control the qubits must match their low resonant frequencies, which limits the speed at which the computer can carry out instructions.

For this reason, other research teams are working with electrons rather than atomic nuclei. This type of silicon quantum computer "would have many of the same advantages but could have a clock speed of 1 GHz," said Yablonvitch. Like nuclei, electrons also act like tiny magnets, but they can be manipulated with ultrafast pulses of laser light. The downside to using electrons is that their quantum states are short-lived compared to the states of nuclei.

It's difficult to estimate how long it will take to develop full-scale quantum computers, said Ladd. It is likely to take at least 20 years, but this is "wild speculation," he said.

Ladd's research colleagues were Jonathan Goldman, Fumiko Yamaguchi and Yoshihisa Yamamoto of Stanford University and Eisuke Abe and Kohei Itoh of Keio University in Japan. They published the research in the July 1, 2002 issue of the journal Physical Review Letters. The research was funded by the Japan Science and Technology Agency, NTT and the Defense Advanced Research Projects Agency (DARPA).

Timeline:   > 20 years
Funding:   Government, Corporate
TRN Categories:   Physics; Quantum Computing and Communications
Story Type:   News
Related Elements:  Technical paper, "All-Silicon Quantum Computer," Physical Review Letters, July 1, 2002


July 24/31, 2002

Page One

Disks set to go ballistic

Two-step queries bridge search and speech

Implant links nerve cells to electronics

Silicon chips set to go atomic

Light switch promises powerful computers


Research News Roundup
Research Watch blog

View from the High Ground Q&A
How It Works

RSS Feeds:
News  | Blog  | Books 

Ad links:
Buy an ad link


Ad links: Clear History

Buy an ad link

Home     Archive     Resources    Feeds     Offline Publications     Glossary
TRN Finder     Research Dir.    Events Dir.      Researchers     Bookshelf
   Contribute      Under Development     T-shirts etc.     Classifieds
Forum    Comments    Feedback     About TRN

© Copyright Technology Research News, LLC 2000-2006. All rights reserved.