Summary of Session II: Quantum Computing and Technology
Rapporteur: Daniel Gottesman, Berkeley EECS
The first speaker was Bill Risk, from IBM. He gave us a summary of the overall IBM effort in quantum computation:
Historical: Pioneering work was done by Landauer and Bennett on thermodynamics of computation and reversible computation. Quantum key distribution was first invented and demonstrated at IBM.
Nuclear magnetic resonance: Chuang et al. of IBM Almaden performed Shor's algorithm with a seven-qubit molecule. The molecule itself was synthesized at IBM.
Ion traps: The IBM ion trap does not seem to experience as severe a heating problem as other ion traps.
Solid-state implementations: IBM is investigating various possibilities.
Quantum cryptography: An experimental quantum key distribution system is set up at IBM; detectors are a crucial problem.
No IBM/Berkeley day would be complete without a speaker from Caltech, so the second speaker was John Preskill, who gave us a summary of theoretical quantum information research at Caltech:
Algorithms: An algorithm for solving Pell's equation due to Sean Hallgren (formerly a Berkeley graduate student).
Cryptography: Impossibility proof of quantum coin tossing due to Alexei Kitaev.
Entanglement: Entanglement has non-locality properties impossible in classical physics. Dave Bacon (another former Berkeley student) and Ben Toner have been studying how much communication is needed to simulate this non-locality.
Error correction: Using topologically-inspired quantum codes, Dennis, Kitaev, Landahl, and Preskill have shown quantum information can be stored for arbitrarily long times if the error rate per gate is less than 10-4. The construction uses only local quantum gates, which is important for many implementations.
John Kubiatowicz of Berkeley EECS spoke next about architectural issues of building quantum computers:
Entropy exchange: Fault-tolerant quantum computation requires a constant source of initialized states; i.e., high-entropy random qubits must be replaced with cold standardized qubits. This ability must be dispersed throughout the computer and must be present everywhere.
Wires: The difficulty of making wires for qubits will be a big difference between classical and quantum architectures. Qubits cannot be copied or amplified, unlike classical bits. Quantum wires may consist of SWAP gates for short wires, and quantum teleportation for longer distances.
The next speaker was Birgitta Whaley, of the Berkeley Chemistry department. She spoke about theoretical issues relating to implementations of quantum computation:
Decoherence-free subspaces (DFS): A DFS is a way of encoding quantum states so that they are immune to the effects of errors, without the need for active correction processes. This only works for certain restricted types of errors, such as collective errors (which affect all qubits equally and coherently), but might be useful for eliminating the dominant error modes in a quantum computer.
Encoded universality: To create a universal quantum computer, we need the ability to perform any unitary transformation. However, it is acceptable if we only perform such transformations on states with a particular encoding. This strategy allows universal quantum computation based on, for instance, just the exchange interaction, which is easily available in many solid-state systems.
We actually had two speakers from Caltech; the second, Hideo Mabuchi, was the next speaker of the day. He talked about experimental quantum computation at Caltech:
Quantum networks: Photons are good for quantum communication, whereas atoms are good for quantum computation and storage. Caltech has a program to build quantum networks by trapping atoms in cavities, allowing an interface between photons and atoms.
Quantum feedback control: By studying quantum control using active real-time feedback, one can hope to get better control of a quantum system. Mabuchi's group has performed an adaptive homodyne measurement using feedback to get better accuracy than possible otherwise.
The last speaker was Mike Crommie, of the Berkeley Physics department. He spoke about issues relating to condensed matter qubits:
Measurement: The scanning tunneling microscope is a useful tool for measurement in various solid-state systems. It might help measuring spins in the Kane proposal (single phosphorus atoms implanted in silicon electrodes) or on a surface.
Decoherence times: The rate of decoherence is a critical parameter, since if it is too fast, quantum computation will be impossible in a system. STMs help to measure it.
Creation of entanglement: Entanglement can be created in many solid-state systems via the exchange interaction.
Various ideas came up multiple times during the day in different guises:
Error Control: Every speaker said something relating to error control in a quantum computer. On the one hand, there was the issue of experimental determination of error rates (Risk, Crommie). On the other, various types of error control techniques were described: error correction (Preskill), decoherence-free subspaces (Whaley), and active feedback control (Mabuchi). The need for an entropy exchange component to quantum computers (Kubiatowicz) comes from the need for error correction. Also, long-range quantum cryptography (Risk) requires some sort of quantum repeater (Mabuchi) based on quantum error correction.
Communication: Another recurring theme was quantum communication as a separate task from quantum computation. This relates to quantum cryptography and quantum networks (Risk, Mabuchi), and also the difficulty of creating wires for a quantum computer (Kubiatowicz).
Scalability: To create a large quantum computer, we will need to scale significantly beyond the current under-10-qubit range of current quantum computers. The speakers seemed generally in agreement that a solid-state system (Risk, Crommie) was the way to do this. Only in a large quantum computer do architectural issues (Kubiatowicz) become a problem. The exchange interaction (Whaley, Crommie) was mentioned as a good candidate for control of quantum information in these systems.