New technique advances carbon-fiber composites.
Seventeen years after physicist Richard Feynman speculated that a quantum computer might be better at simulating quantum mechanics than a traditional computer, MIT researchers have succeeded in programming a prototype quantum computer to do just that.
MIT's quantum computer can only count to four, but it has demonstrated quantum simulation for the first time. Research such as this may one day result in commercial hardware capable of computations beyond the reach of any conceivable conventional computer.
Associate Professor of Nuclear Engineering David G. Cory, Raymond Laflamme of Los Alamos National Laboratory and colleagues have come up with a general scheme for quantum simulation that would work on any quantum computer. In a paper in the June 28 issue of Physical Review Letters, the researchers illustrate the scheme on a liquid state nuclear magnetic resonance (NMR) quantum computer developed at MIT.
"What we've had this quantum computer demonstrate -- the dynamics of harmonic and anharmonic oscillators -- is very simple. A first-year quantum mechanics student could do it on paper. But this is probably the first reachable application of information processing on a quantum system," said Ching-Hua Tseng, an MIT postdoctoral associate on the nuclear engineering research team and co-author of the paper.
The other authors are Shyamal Somaroo and Timothy F. Havel of Harvard Medical School.
YEARS IN THE WORKS
In the 1980s, Benioff, Bennett, Deutsch, Feynman and Landauer studied the possibility of quantum simulation on a quantum computer.
It wasn't until 1994 -- when AT&T's Peter Shor (PhD 1985) discovered a quantum computer algorithm that would efficiently find the prime factors of composite integers -- that interest in the field was renewed. Integer factorization, because it takes a conventional computer a long time, is the basis for cryptosystems used for security. If quantum computers could do these calculations in seconds instead of years, they would be a major threat to cryptosystems. A quantum computer, operating at the speed of a conventional computer, could break a program's code as easily as a master safecracker could open a cheap padlock.
Around the same time as the 1994 discovery, Seth Lloyd, associate professor of mechanical engineering at MIT, proposed that a quantum computer could be built from an array of coupled two-state quantum systems, each of which can store one qubit (a measure of information).
Dr. Cory's research group, along with Neil A. Gershenfeld and colleagues in MIT's Artificial Intelligence Laboratory and Isaac Chuang at IBM, independently helped develop the quantum computer. The simulation scheme Dr. Cory's group used borrows heavily from average Hamiltonian theory, an NMR formalism introduced by John S. Waugh, Institute Professor emeritus of chemistry at MIT.
At the moment, quantum computers don't possess even the calculating power of a pocket calculator.
But the potential is there for a computer that would far surpass a regular computer in power and efficiency. Because quantum mechanics allows a quantum computer's components to represent many states simultaneously, it should be able to perform many computations simultaneously, making it an enormously powerful tool.
A quantum computer may be able to quickly solve problems involving weather prediction and fluid flow -- problems so big they couldn't be stored in a conventional computer's memory.
"The power of quantum computers is astronomical. If you took every molecule of water on the Earth and used it to build a classical computer, all you need are 86 qubits and you have a quantum computer that is more powerful than this theoretical classical computer," Dr. Cory said. A 40-qubit quantum computer cannot be simulated by even the largest existing computers in the world.
"Given that a quantum computer is based on something that physics allows and that there are such compelling reasons why we want it, I am certain that we will find the technology to make it. Don't ask me to predict how it will happen, because there are very important challenges to face. This has not been handed to us on a platter."
ENCODING ON SINGLE ATOMS
A quantum computer is based on the physical theory called quantum mechanics, which permits only probable or statistical calculation of the observed features of subatomic particles. One of these features is spin, or angular momentum.
In proposing how a quantum computer would work, Feynman suggested that ones and zeroes (representing the "on" or "off" states of the electronics of a conventional computer) could be represented by independent quantum states. According to quantum mechanics, the protons in a molecule act as small bar magnets with only two possible quantum states.
Similar to a compass, the spins align in a magnetic field to be either up or down. This is the equivalent of the conventional computer's "on" and "off" states. After data are stored in this manner, the computer also must incorporate logical gate operations, or commands, to perform calculations.
A single bit of information in the quantum realm is known as a qubit. A qubit can represent one, zero or the two states at once. It is possible to create a command that causes a controlled interaction among two or more qubits, producing a coherent change in the state of one qubit that is contingent on the state of another. This is the basis of quantum computation.
CONTROLLING WITH SPIN
Theoretically, a quantum computer could be fashioned out of anything. Any physical system is by definition a quantum mechanical system. The challenge is to isolate and manipulate the system to function as a computer. Cold trapped ions, electrons in semiconductor structures, superconducting devices and photons have been put forth as possible bases for a quantum computer.
Drs. Cory, Tseng and their colleagues are among a small handful of researchers using NMR to experiment with qubits.
NMR techniques are widely used to study the structure of molecules and to image internal structures within the human body. NMR allows scientists to manipulate the atomic spins of nuclei by applying an electromagnetic pulse to molecules diluted in a liquid. Sending a pulse for a specific amount of time generates a known signal. The signal is amplified by the molecules acting in parallel.
Controlling the spin of the liquid molecules performs the gate operation: it tells the computer to perform the mathematical function needed to solve the problem. It then arrives at an answer. In this case, the computer simulates a quantum mechanical system by carefully forcing its qubits to evolve like another quantum system.
MAKING ATOMS COOPERATE
The main obstacle to constructing a practical quantum computer is control -- the difficulty of engineering the quantum states required, the phenomenon of decoherence (the tendency for these quantum states to lose their coherence properties through interactions with the environment) and the quantum measurements required to read out the result of a quantum computation.
A quantum computer's very flexibility and potential is based on its building blocks' ability to interact on many levels. For this reason, it is also difficult to control. The atoms "talk" to each other and to materials in their environment, thereby losing their coherence or ability to concentrate on the task at hand. Recently discovered quantum error correction schemes should go a long way toward solving this problem, Dr. Tseng said.
In the face of all these challenges, researchers are unsure whether a quantum computer can be efficiently scaled up beyond a prototype. Dr. Tseng pointed out that the MIT researchers took the realistic route of creating a quantum information processor, a machine tailored toward a specific purpose, rather than attempting a full-fledged general computer capable of a variety of uses.
However, even primitive quantum computer research helps scientists better understand underlying physical principles of quantum mechanics and, in turn, use what is physically possible to inspire new avenues in computer science. Simulating the behavior of quantum mechanics is a useful research tool because "if you can calculate the behavior of a system you can't normally calculate, you can find out more about its behavior and structure and do research on that," Dr. Tseng said.
This work is supported by the Defense Advanced Research Projects Agency (DARPA) and the Department of Energy (DOE).
A version of this article appeared in the July 14, 1999 issue of MIT Tech Talk (Volume 44, Number 1).