While farhi and his colleagues are determining what can be done with a quantum computer, others are hard at work engineering the hardware itself.
IBM’s DiVincenzo says a practical quantum computer must have five fundamental capabilities: It must provide for qubits-particles or groups of particles that can be isolated and placed in superposition, the indeterminate state in which they represent both ones and zeroes. It must be possible for operators to control the initial states of the qubits, analogous to setting them all to zero at the outset of a computation. The qubits must remain stable-in superposition long enough to perform an operation-anywhere from milliseconds to several seconds. It must be possible to implement quantum logic circuits that correspond to such Boolean operators as and, or, and not, which form the basis of traditional computer architecture. In classical computers these expressions are embodied in electrical circuits. The simplest logic gate, the not gate, converts an incoming digital one into a zero, and vice versa. To manipulate qubits, quantum circuits will have to employ techniques such as extremely precise control of magnetic fields or laser pulses.
The final requirement of a quantum computer is that it make the results of a calculation accessible to the user, through, for example, a visual readout.
Most quantum-computing experiments boil down to efforts that address one or more of DiVincenzo’s requirements. “There are probably a half-dozen serious proposals and 10 times that number that are not serious,” says Bruce Kane, who specializes in the science of single-electron devices at the University of Maryland.
Chuang and Gershenfeld, for example, used nuclear magnetic resonance to measure the spin of qubits in bulk materials-a vial containing a billion billion molecules custom-made from fluorine, carbon, iron, hydrogen, and oxygen. The spins of the nuclei of the five fluorine and two carbon atoms in each molecule functioned as interacting qubits to execute Shor’s algorithm. Although Chuang and Gershenfeld’s achievement in controlling and measuring the spins of seven qubits has been widely hailed, many in the field believe that scaleup of this approach will be extremely difficult. “The limitation is that every time you add qubitsthe signal-to-noise ratio decreases,” says Kane, referring to the amount of useful information-such as the excess of particles with one spin over particles with a different spin-that can be distinguished from random disturbances in the bulk fluorocarbon material.
Chuang himself acknowledges that his seven-bit quantum computer falls far short of the scale needed for meaningful computations. “To make it practical, we’ll have to get to thousands, if not hundreds of thousands, of qubits,” he says. A rival approach that uses nanoscale engineering techniques to build qubit containers, he adds, may be easier to scale up.
David J. Wineland and his team are studying this alternative at the Time and Frequency Division of the U.S. National Institute of Standards and Technology in Boulder, CO. They’re building miniature devices using electrodes that isolate ions in “traps” fashioned from electric fields. The virtue of this approach, Wineland says, is that ion traps are relatively easy to fabricate, may be linked together, and can hold more than one ion per trap. Wineland suggests that a string of ions confined in a single trap might function as a sort of quantum memory, and each additional qubit would expand storage capacity exponentially. Already, Wineland’s group has coaxed such qubits to stay in a state of superposition for up to 10 minutes. But one current weakness in this scheme is that it’s difficult to transfer quantum information between ions held in separate traps, a necessity for large-scale computations.