Scientists at IBM Research Tuesday (February 28, 2012) said they have achieved a major advance in quantum computing that will allow engineers to begin work on creating a full-scale quantum computer.
The breakthrough allowed scientists to reduce data error rates in elementary computations while maintaining the integrity of quantum mechanical properties in quantum bits of data, known as qubits.
The creation of a quantum computer would mean data processing power would be exponentially increased over what is possible with today's conventional CPUs, according to Mark Ketchen, the manager of physics of information at the IBM's TJ Watson Research Center in Yorktown Heights, NY. .
A qubit, like today's conventional bit, can have two possible values: a 0 or a 1. The difference is that a bit must be a 0 or 1, and a qubit can be a 0, 1, or a superposition of both.
"Suppose you take 2 qubits. You can be in 00, 01, 10, and 11 at the same time. For 3 qubits you can be in 8 states at the same time (000, 001, 111, etc.). For each qubit you double the number of states you can be in at the same time. This is part of the reason why a quantum computer could be much more powerful," Ketchen said.
While a quantum computer is still a long way from being a reality - probably 10 to 15 years --- advances in reducing error rates and retaining the integrity of quantum mechanical properties in qubits opens the door to experimentation with new microfabrication techniques, IBM said.
"We're finally to the point where devices are getting good enough where data checking and error correcting is possible. As you cross this threshold, there's a lot of excitement growing," Ketchen said.
IBM's team presented their quantum computing advances at the annual American Physical Society meeting today.
Unlike today's silicon-based semiconductors, IBM is employing superconducting qubits that use established microfabrication techniques developed for silicon technology but produced on a sapphire chip. That offers the potential to one day scale up to the manufacture of thousands or millions of qubits.
A picture of the Silicon chip housing a total of three qubits. The chip is back-mounted on a PC board and connects to I/O coaxial lines via wire bonds (scale: 8mm x 4mm). A larger assembly of such qubits and resonators is envisioned for a scalable architecture.
"Things are getting to the point where, while we may not be ready to build a quantum computer, it's time to start thinking what a computer like this would look like and what it would it be able to do," Ketchen said.
At the atomic level, atoms and their components such as electrons behave differently whereby they adopt the rules of quantum mechanical physics forming quantum systems. In these states, quantum systems can be manipulated in such a way that certain mathematical problems and logical operations can be solved using exponentially fewer time steps than what is possible with conventional machine computation. For example, a quantum computer could factor a very large integer number into its prime factors (e.g., 3 and 5 are factors of 15) on a practical time scale, whereas it would take the age of the universe to solve the same problem using conventional electronics.
For example, today's best multi-core processors can encrypt or decrypt a 150-digit number, but if you were interested in decrypting a 1,000-digit number, it would take roughly all the computational resources in the world to do it, Ketchen said.
"On a quantum computer [in theory], it might take a few hours," he said.
IBM is not alone in its research on quantum computing; both the University of California at Santa Barbara and Yale University are doing similar research. Ketchen, however, argued that only IBM has the resources to actually fabricate quantum-computing chips.
Currently, IBM can maintain the state of a gate electrode long enough to perform an operation. That operation is accurate 95% of the time, Ketchen said. The scientists want to get the accuracy above 99%, so as to reduce data errors to the point where it could be used to perform computations with an acceptable degree of accuracy.
"Once the data error rate gets small enough, you can combine a number of gates together and get a prefect qubit," Ketchen said. "Now things are getting into the ballpark where you can build something and get the right answer. It also says we have to start thinking more seriously about building more complicated things and putting them together."
Sign up for Computerworld eNewsletters.