On the Quantum Computing Bus: A Chat with Steven M. Girvin
Computing with qubits beyond the simplest examples requires the ability to store, retrieve and communicate those qubits without losing their quantum state.
Steven Girvin is the Eugene Higgins Professor of Physics and Professor of Applied Physics at Yale University. His primary focus at present is on the quantum electrodynamics of microwave electrical circuits and quantum information processing with superconducting qubits and cavities. A 2007 paper on which Prof. Girvin collaborated, Coupling Superconducting Qubits via a Cavity Bus, points in the direction of the sort of data bus necessary for meaningful quantum computing.Prof. Girvin spoke with me by phone from his office:
JW: Your team is going beyond the theoretical and experimental problems of quantum computation to the engineering problems of what we do with the intermediate results of quantum computation.
SG: It's a new kind of engineering. You're trying to build little machines that are quantum mechanical and do things ... perhaps quantum computation, but who knows what we'll be able to do with them ... circuits in which microwaves, which despite their name are also particles, play an important role.
JW: The net result for quantum computing is that you can transmit an intermediate result without losing its quantum state.
SG: You have a bus on a classical computer because you can't have wires going from every bit to every other one. You have a common highway.
A quantum bus has to be able to carry signals in a special way. For instance, there might or might not be a photon of microwave energy traveling on the bus, it has be coherent, it has to be dissipationless.
Photons are very nice for doing this because they can travel long distances and carry information from one quantum bit to another quantum bit.
You can use that information to entangle them, to perform logic operations like "if the first qubit is in a certain state, flip the second one" -- the CNOT operation. In order to have quantum computation, you need to have interaction between qubits.
JW: And you need the ability to store an intermediate result, do you not?
SG: That's a problem, how do you store it? You can store it in a qubit, it could be a zero, a one, or a superposition of both. It could be a really good qubit where superpositions live a long time.
Right now, our qubits are a little embarrassing. They don't stay coherent for very long. People have had the idea of using microwave resonators as a quantum memory to store results while you're waiting to use them.
JW: I spoke recently to Prof. Ian Walmsley of Oxford and he was using the absorption and re-emission of photons by electrons as a quantum communication channel and potentially a refreshable quantum DRAM.
SG: We're interested in exploring whether it's easier to make resonators that hold microwave photons in superpositions of zero and one live a long time compared to how long excited states of qubits live. We and other groups have experimented with transferring the quantum information in a qubit into photons in a resonator where you can park it for a long time, and when you want to use it later in the calculation, bringing it back.
The actual computation has to be done in the qubit, for technical reasons, but the storage can be in this resonator.
JW: How long is "a long time"?
SG: Not very long. The best coherence times for qubits ... I think Robert Schoelkopf's and Michel Devoret's group here at Yale have among the longest coherence times that have been found in superconducting qubits and it's only a few microseconds.
That sounds very short, but given the extremely rapid control and pulsing that you can do on nanosecond timescales, that's actually enough time to do dozens of operations.
To build a real, operating quantum computer, you're going to need to do tens of or hundreds of thousands of operations during coherence time. We have a ways to go yet.
JW: Thinking as a computer programmer, am I correct in factoring practical quantum computing into qubit processors, a bus, and semi-persistent or at least refreshable storage?
SG: Yes. You have to have all of the requirements of a classical computer, but everything you do has to be able to work on superpositions of the two possible states of the qubit, the two possible states of the memory, the two possible states of the bus, and so forth.
JW: Without the fabled "observation" which destroys quantum state.
SG: Exactly. If you look at what is going on during the computation it is ruined. That's the central engineering challenge.
In order to be quantum-coherent, in order to be in the superposition wherein you're uncertain whether the qubit is one or zero, you can't observe it, neither you, nor the environment, nor the electrical environment, nor air molocules in the room. Nobody can observe. Nobody can measure the state, otherwise it's ruined.
The way to fix that is to make a qubit that doesn't interact with anything, so that nobody can find out what state it is in.
At the end of the calculation, I have to measure the state of the qubit, so I have to make it interact with my measurement apparatus. Or, if I want to do logic, I have to make two qubits interact with each other.
JW: And they have to interact at a quantum level.
SG: Exactly. So you have to be able to have no interactions in order have no decoherence, and you also have to be able to produce strong interactions, because you need to do logic operations, or because you need to measure the answer.
That's the engineering challenge: to be able to have no interactions and no measurements when you don't want them, and to be able to have good, strong ones when you do want them.
That's not so easy.
SG: Yes. Everyone who is making superconducting qubits uses Josephson junctions in some form or other. But there is an amazing variety of different circuit designs, all of which use at least one Josephson junction, which have different properties, different features and different strengths and weaknesses.
We started with a circuit design that we thought was good, but it had a reputation based on reality of being very susceptible to noise from stray electric fields. We slowly figured out ways to get rid of that weakness while keeping some of the other advantages at the same time.
Martinis has a different kind of qubit using Josephson junctions much larger in surface area. This has certain strengths and weaknesses.
We're all trying to end up in the middle with qubits that aren't too susceptible to perturbations but which still can be controlled, can be measured, and can talk to each other.
JW: Does it look hopeful?
SG: Well ... the first experimental paper in our group was six years ago. The experimental paper that started the field from the NEC Quantum Group was 1999. They did the first crude experiment with a circuit that was kind of acting like a qubit.
If you had asked me then whether we would be able to do all the things we're doing now, running small quantum algorithms and entangling three qubits at a time, measuring single microwave photons in interesting ways, I would have said that it is too hard and that we would not have made so much progress.
On the other hand, if you look at what the actual engineering requirements are for a non-trivial quantum computer, being able to do hundreds of thousands of operations without error, having many qubits, and so forth, it's a very, very distant and difficult goal.
I jokingly tell lay audiences when I give talks that we have made so much progress that building a quantum computer has gone from "impossible" to "impossible divided by ten".
It's typical of engineering frontiers that, if you are optimistic, you overestimate what you can do in the short term and underestimate the progress in the long term.
At this point, there are people trying many different kinds of qubits: atoms, ions, photons, superconducting circuits. It's not really obvious what the optimal choice is yet.
If you set aside the big-picture goal of building a quantum computer, some of the ideas developed are already helping us make, for example, better atomic clocks. That alone makes it worthwhile, even if we never build a quantum computer.
JW: It seems to me that one of the most interesting things about quantum computing is that it's got thousands of people playing with these states of quantum mechanical entities that, in the past, people collected statistics on but never before so intensively worked with these states day by day.
It's unusual to have people in my field, condensed matter physics, write a paper and have people in atomic and optical physics interested in it. We're usually very narrowly specialized. It's not that our experiments are similar. It's that having the overarching goal of building a quantum computer has given us a common language. "How are you manipulating your quantum states? Do you have any tricks that I can use?"
It's been tremendously invigorating and fun!