Quantum computer with separate CPU and memory represents significant breakthrough

By Brian Dodson

The UCSB von Neumann quantum computer. The small black squares are the superconducting qubits, and the meandering lines are the memory resonators. (Photo: Erik Lucero)

To date, quantum computers have been implemented so that programming their operation was, in essence, hardwired into their essential structure. Although many useful demonstrations of quantum computing have resulted from such special-purpose devices, they are basically one-problem computers which cannot easily be reprogrammed or scaled to attack larger problems. As early models of practical quantum computers, they don't make the grade.

• 
• 
• 

The basis of essentially all practical classical computers is the Von Neumann architecture, which comprises a central processing unit (CPU) to do calculations, a memory which holds both data and CPU instructions, and an interface which allows the input and output of the CPU to change the information in memory. This architecture is easily scalable to nearly any size and capacity desired.
Recently, John Martinis' research group at the University of California at Santa Barbara has created the first general-purpose programmable quantum computer.
Their quantum computer uses superconducting circuits to form quantum computer equivalent of a Von Neumann architecture.
The result is the first universal (general purpose) quantum computer. To illustrate the importance of this design, the UCSB circuit, which has two qubit registers and two entangled memories, has been used to simulate a three-qubit logic gate. Such ability to solve problems having more active information than the capacity of the CPU is enabled by implementation of the Von Neumann architecture.

How quantum computers work

The irreducible carrier of quantum information is the qubit, named in analogy to the classical bit. But whereas the bit is a simple on-off signal, a qubit is in essence a unit vector whose direction is described by a pair of angles, θ and Φ. These angles describe the superposition of pure quantum states which makes up the quantum information in the qubit. While a bit defines a single binary parameter (+ 1), a qubit defines a continuous complex variable.
When a quantum operation is carried out on a qubit, these angles change, thereby changing the quantum information in that qubit. All quantum computation in the end reduces to combined rotations of quantum states.

Superconducting circuitry

There are a number of reasons that superconducting circuitry was chosen for implementation of the UCSB Von Neumann quantum computer. Superconducting structures which can store a qubit of information are easily constructed using standard microfabrication techniques. Additionally, such structures couple easily to MHz and GHz radio waves, which provides effective control of the computer operations using well understood electronics.
A larger physical dimension, however, implies there are likely to be more ways in which superconducting qubits can lose coherence through unintentional environmental interactions. This does lead to shorter coherence times than are achieved in other physical implementations of qubits, about 4 microseconds for the UCSB circuitry.
However, the key parameter is how many quantum operations can be made within the coherence time. In the case of the UCSB computer, several hundred operation cycles can be carried out without losing quantum coherence. While encouraging, the number of coherent quantum operations must be significantly increased to support a large-scale superconducting quantum computer.
The low-level organization of the UCSB quantum computer is called Resonator/zero-Qubit architecture (RezQu). This consists of a set of superconducting qubits (in the current example, two qubits). Each of the superconducting qubits is capacitatively coupled to a dedicated memory resonator, as well as to a common resonant quantum information bus. The bus is used to couple qubits during computational operations, while the memory resonators are used for storing the current state of the qubits. When a qubit is passed into its memory resonator, the qubit is placed in the ground state.
Using their new architecture, the UCSB group was able to implement the three-qubit Toffoli OR phase gate with 98% fidelity. Universal quantum computation can be carried out using combinations of this Toffoli gate and simple qubit rotations. However, it does not currently appear that 98% fidelity represents a sufficiently small error to permit conventional error-correcting codes to function properly. Thus, the UCSB Von Neumann quantum computer is potentially capable of universal computation, limited only by memory resources and quantum coherence time, but requires increased fidelity to fulfill this potential.
The Paper entitled Implementing the Quantum von Neumann Architecture with Superconducting Circuits is published online in the journal Science.

Sources: UCSB, physicsworld.com