I’ve spoken before, in following blog posts, about the superb quantum computing R&D being done by the Martinis and Cleland groups at UCSB:
- To the Rescue, Musketeers! To the Rescue!
- The Greatest Obstacle to Building a Quantum Computer
- Quantum Computing’s American Riviera
- Lincoln Laboratory Joins the Race To Build a Quantum Computer
Since then, the UCSB group has been making steady progress in their QC efforts. Check out their latest paper:
Implementing the Quantum von Neumann Architecture with Superconducting Circuits,
by Matteo Mariantoni, H. Wang, T. Yamamoto, M. Neeley, Radoslaw C. Bialczak, Y. Chen, M. Lenander, Erik Lucero, A. D. O’Connell, D. Sank, M. Weides, J. Wenner, Y. Yin, J. Zhao, A. N. Korotkov, A. N. Cleland, John M. Martinis
IMHO, in light of this paper, I think anybody who still claims it will take 20 years before QCs arrive, is clueless.
Let me now give a short executive summary of the paper.
The paper demonstrates a “universal” (i.e., general purpose) quantum computing device with two qubits (However, the UCSB scientists see no impediments to adding substantially more qubits to their device (“scaling it”), and they expect to do this in the near future). Most of their device is built on a chip using standard microfabrication techniques and it utilizes standard radio and microwave frequency components. The chip does need to be cooled to milli-Kelvin temperatures. Luckily, refrigerators that achieve such temperatures have improved considerably in recent times, no longer relying on costly liquid helium.
The paper reports that two quantum algorithms were performed on the new rig: the quantum Fourier transform, with 66% fidelity, and the three-qubit Toffoli OR phase gate, with 98% fidelity. These two algorithms are typical for a gate-model QC, as opposed to an adiabatic-model QC, which is what D-Wave offers. These fidelity numbers are not stellar; actually, they are far below what is necessary for doing quantum error correction and Shor’s algorithm. (But they might be good enough — me and my algorithms, we are not too demanding — for performing my algorithm that improves upon Grover’s algorithm.)
The UCSB scientists call their QC design ideas the ReZQu architecture (ReZQu stands for Resonator/Zero/Qubit). Their architecture is von-Neumann-like. In the case of classical computers, a von Neumann architecture comprises a central processing unit (CPU) for performing operations and a memory (RAM) for holding instructions and data. Their QC still stores the instructions in a classical computer, but it has one section (called the quRAM) for storing quantum data and another section (called the quCPU) for doing quantum calculations.See Fig.2. The quCPU (blue box) includes two qubits Q1 and Q2 and the bus resonator B. The quRAM (magenta boxes) comprises two resonators M1 and M2 that serve as memory and two zeroing registers Z1 and Z2.
Resonators act like holding tanks for one or more photons (all of the photons held by a resonator are of the same frequency, so they occupy a degenerate energy level). On the contrary, the qubits have two discrete, separate energy levels (they actually have more than 2 levels, but only the lowest two are used. The higher levels have a different energy separation than the lowest two, so that transitions between the lowest two levels can be excited without exciting transitions to the higher levels). The zeroing registers are also two level systems. They are used for “dumping quantum information”, i.e., to bring a qubit they are put in contact with to its “zero” (i.e.. lowest) energy level.
The horizontal axis of Fig.2 measures distance in one spatial dimension and the vertical axis measures frequency. The frequencies on the vertical axis range from about 6 to about 8 MHz. M1, M2, B, Z1, Z2, i.e., all elements except the two qubits (Q1 and Q2) have a fixed, distinct (i.e, well separated from the others) frequency. Unlike the frequencies of the other elements, the frequencies of Q1 and Q2 are not fixed. They can be varied at will using something called z-pulses. The frequency of Q1 (respectively, Q2) can be adjusted so as to coincide with either M1, Z1 or B (respectively, M2, Z2, B).
That’s why I like to call them “elevator qubits”, because Fig.2 resembles a building with various floors (= frequencies or energies). An elevator (= a qubit) can move from one floor to another. Once it reaches a certain floor, it can absorb a person (= a quantum excitation) from that floor, or release a person into that floor. Each elevator (= qubit) can only hold one person at a time but the storage floors (= resonators) can hold multiple people.
If qubit Qj (j=1,2) were long-lived, there would be no need for giving it a memory Mj. The raison d’être for the memories is that memories are longer lived than qubits. In the present experiment, resonator excitations have a coherence time (i.e., a lifetime) of about 4 microseconds, whereas qubit excitations live about a tenth of that time. The UCSB scientists hope that in the future, they will identify and figure out how to avoid sources of noise which limit the life of resonator excitations. The UCSB device can perform a single elementary operation in about 10 nanoseconds, so its current 4 microsecond coherence time already allows it to perform a few hundred elementary operations.
If you want to delve more deeply into the details of the UCSB QC architecture, Martinis keeps a really excellent, highly pedagogical website. I particularly enjoyed reading the pdf files of the lectures he gave at a Summer School in Finland, on August 2007. These lectures can be found in the “Tutorials” section of his website. I also learned a lot by skimming through the various theses that are available in the “Theses” section of his website.