Sweating the small stuff Home

Chip design aims for quantum leap

By Eric Smalley, Technology Research News

The first step toward making phenomenally powerful quantum computers is capturing and manipulating individual subatomic particles, which is a bit like getting a fly to venture onto your desk, then perform tricks like "sit up" and "roll over" on command.

The second step is harnessing, controlling and coordinating thousands or millions of particles at once. Making a practical quantum computer also means doing this using ordinary electronics rather than exotic laboratory equipment.

University of Wisconsin researchers are tackling these issues with a quantum computer design that would incorporate thousands of individually-controlled electrons into a silicon chip that could be made much the same way as today's computer chips.

Practical quantum computers would be many orders of magnitude faster than today's computers for problems that involve massive amounts of data, like cracking secret codes and searching large databases.

The researchers' idea is to "trap single electrons in tiny silicon sandwiches about a millionth of an inch across," said Robert Joynt, a physics professor at the University of Wisconsin at Madison. The silicon sandwiches are quantum dots, microscopic specks of semiconductor material that can hold one or a few electrons.

These dots can represent bits of information because an electron acts like a tiny spinning top, and depending on which way it is spinning it can represent 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.

Proposals for making quantum computers out of quantum dots have been around for several years. The Wisconsin researchers' design plots out some of the difficult details -- it allows individual electrons to be loaded into the quantum dots and allows interactions between electrons held in neighboring dots is to be closely controlled.

Each dot would consist of a bottom layer of silicon germanium that has been chemically altered to allow electrons to flow more easily. This layer would serve as a reservoir of electrons.

The middle layers would consist of an extremely thin layer of silicon sandwiched between layers of unaltered silicon germanium. The silicon layer would hold the individual electron used by the quantum computer, and the silicon germanium layers would act as barriers to keep additional electrons out. The researchers could coax individual electrons to tunnel through the barriers to the silicon layer by changing the electrical current running through the chip.

Metal electrodes that move the electrons laterally would form the chip's top layer. The electrodes would be used to bring pairs of electrons in adjacent dots together to perform the basic logic operations of computing. The quantum interactions of a pair of electrons can be represented mathematically, and that math can be used to generate the binary logic that is the foundation of computing. This allows logic operations like adding binary numbers to be carried out by controlling the electron interactions.

"The biggest hurdle is fabrication," said Joynt. "This needs to be done with exquisite control of the quality of the material and to very high measurement specs," he said.

Because the quantum dots are made from layers of metal and semiconductors, like computer chips, the researchers' proposed device could be built using standard chipmaking processes, according to Joynt. "The dots are only slightly smaller than the features on commercial chips, which have millions of transistors," he said.

Unless the optical lithography used in the commercial chip industry improves, however, this minor decrease in size means that the researchers will have to use electron beam lithography, said Joynt. "This is slower and more expensive, but perhaps not prohibitively so," he said.

Today's optical lithography uses ultraviolet light with wavelengths ranging from 200 to 300 nanometers and can etch features as small as 130 nanometers. Electron beams can be focused with magnetic fields to around 10 nanometers and so can etch much smaller features.

The potential benefits of a practical quantum computer are enormous.

"Quantum computing is massively parallel," said Joynt. This means that many inputs can be processed at the same time, which makes for a computer that can solve problems that would take a regular computer "essentially forever" to work out, he said.

When an electron is isolated from its environment it is in the weird quantum state of superposition, meaning it is spinning in both directions at once. An electron in superposition can represent a mix of 1 and 0, and a string of electrons in superposition can represent every combination of 1s and 0s at the same time.

The power of a quantum computer comes from the ability to check every possible combination of numbers at once to find the answer to a problem that can have more possibilities than there are atoms in the universe. Ordinary computers have to check each possible answer one at a time.

Researchers have already come up with software that would allow quantum computers to crack secret codes and search massive databases.

The Wisconsin work is a good effort that adds "many realistic details" to quantum dot research, said IBM Research physicist David DiVincenzo. DiVincenzo and Daniel Loss, a physics professor at University of Basel in Switzerland, developed an earlier quantum dot quantum computer proposal.

"I am very encouraged generally by the efforts of the University of Wisconsin group," said DiVincenzo. "They have started a big, integrated effort involving both theory and experiment," he said.

Practical quantum computers are likely to take 25 years to develop, said Joynt. "And I'm an optimist," he said. "We are working on fabricating a prototype, step by step," he added. "The next step is to make sure that our [silicon layer] is properly trapping the electrons."

Joynt's research colleagues were Mark Friesen, Paul Rugheimer, Donald Savage, Max Lagally, Daniel van der Weide and Mark Eriksson. They published the research in the July 15, 2002 issue of the journal Physical Review B. The research was funded by the U.S. Army Research Office (ARO) and the National Science Foundation (NSF).