Quantum Computers And Qubits Extend Moore's Law
One radical departure is to replace the conventional bit (that can be thought of as a transistor in the on or off state, i.e. with a value of 1 or 0) with a quantum bit, or qubit. This can simultaneously exist in both the 1 and 0 states, in essence representing a superposition of states. The power of using such an approach quickly becomes evident if one considers that a single qubit can be a superposition of two states, 0 or 1, while a pair of qubits can be a superposition of four states that could be represented as 00, 01, 10 and 11.
Using a pair of qubits, four calculations can be performed at once. With more coupled qubit states the number of simultaneous calculations becomes 2N, where N is the number of qubits. A 10 qubit system could perform more than one thousand simultaneous calculations, and a 20 qubit system could perform more than one million.
In 1998, B Kane of the University of New South Wales proposed a quantum computer based on a combination of Si process technologies and nuclear magnetic resonance (NMR) spectroscopy (Kane). Using NMR, each qubit is defined by the spin orientation of an individual atomic nucleus, which is the direction of its nuclear magnetic dipole (up or down). Each of these dipoles can either reinforce or oppose an externally applied magnetic field.
Earlier proposals along these lines attempted to use the dipole effects found in the small organic molecule chloroform suspended in solution, but Kane proposed instead to use a single 31P dopant atom buried in a Si matrix. While such a donor dopant atom would typically provide an electron to be transported throughout the Si lattice, at a temperature of around 100 mK, the donor electron is weakly bound to the 31P atom and its spin can impact the state of the nuclear-spin qubit.
A possible device structure, as shown in the figure, consists of interspersed A and J gates (DiVincenzo). By applying a voltage on the A gate, the bound electron can be polarized, changing its overlap with the donor nucleus. The interaction between the weakly bound electron and the P nucleus determines the relative energies of its two spin states, allowing a specific radio pulse (similar in approach to the way spins are flipped in NMR processes) to selectively change the state of only those nuclei that have a polarized electron. The J gates allow for the manipulation of adjoining weakly bound electrons, whose dimensions at 100 mK are around 20 nm. It is through the use of the J gates that quantum-based logic algorithms can be implemented.
An important aspect of qubits is that they need to be isolated from any degrees of freedom that may lead to their decoherence (change in spin state). When relying on the spin states of a donor atom in a semiconductor, nuclear spins in the host lattice can represent a large source with which donor spins can interact and quickly loose coherence. Because a P nucleus has a nuclear spin of I = ½, this interaction can be eliminated if the host lattice has a different spin state, i.e. I = 0. This is true in the case of isotopically pure 28Si. However, III-V semiconductor materials do not possess stable I = 0 isotopes, making this a Si-only approach.
Manipulating P atoms
While such a quantum computer based on the spin states of a P atom in a Si matrix is possible, it was well understood in 1998 that its implementation was not a problem based on theory or physics. Instead, the problem is the practical reality of how to fabricate the device shown in the figure, and specifically how to place single (or a few) P atoms beneath the surface of the A gates. Progress has actually been quite rapid.
The first step was reported in 2001 in a paper by researchers at the University of New South Wales (O Brien et al.), in which a precise array of P atoms (of 1-2 P atomic widths) were deposited on a Si surface through a process of hydrogen desorption induced by scanning tunneling microscopy (STM). While single atom placement has been demonstrated using STM, this has been done only for metal-on-metal applications, and is much more difficult for semiconductor-on-semiconductor applications due to the strong covalent bonds involved. Therefore, an alternate approach was developed in which an atomically clean and smooth Si surface prepared in a UHV system is exposed to an atomic hydrogen beam to form a primarily monohydride surface (in which one hydrogen atom bonds to each Si atom). A STM tip is then used to desorb hydrogen from the surface. It has been demonstrated that a resolution of