In the next few weeks, a research group at the Delft University of Technology in the Netherlands expects to receive an important package. Its contents promise to increase competition in the race to produce useful quantum computers.
Shipped from the research-and-development facilities of semiconductor giant Intel in Hillsboro, Oregon, the parcel holds the first quantum computer manufactured with the techniques used to fabricate silicon chips in conventional computers. Although the silicon method currently lags behind other approaches to building quantum computers, the company hopes that the technique could accelerate the development of devices that go beyond proof-of-concept curiosities, says James Clarke, who heads Intel’s quantum-hardware development. “I think you’ll hear a lot about silicon quantum computing this year,” Clarke says.
The relatively modest device represents the latest move in the push to give silicon a boost over other approaches. Some scientists also see promise in the silicon route. Physicists such as Michelle Simmons at the University of New South Wales (UNSW) in Sydney, Australia, are developing their own ways of building quantum computers using silicon. In May 2017, she founded an Aus$83 million (US$65 million) start-up called Silicon Quantum Computing, backed in part by the Australian government.
Quantum computers aim to exploit two small-scale phenomena to outperform their classical counterparts, which encode bits of information as 0s and 1s. In the quantum world, units of information are called qubits, and each qubit can exist simultaneously in a ‘superposition’ of both 0 and 1. Two bits can also be entangled, so that the state of one qubit determines the state of its partner. This enables quantum devices to conduct calculations in parallel.
Physicists in many labs have developed prototype quantum computers, which often operate at temperatures close to absolute zero. The frontrunners in the race use one of two methods to encode the qubits: single ions held in traps, or oscillating currents in superconducting loops. Both systems require exquisite control: the ion technique uses complex laser systems to read and write each qubit, and superconducting qubits must each have a device to control them using radio waves.
Proponents of the silicon technique see major advantages in using a semiconductor to code qubits. They can be manipulated much more simply using microscopic electric leads etched right onto the chip. And if the same large-scale manufacturing techniques for making chips could be transferred to the quantum realm, it could become easier to turn the technology into commercial products.
A long road
The idea of building quantum computers out of silicon is not new. Bruce Kane, an experimental physicist now at the University of Maryland in College Park, first suggested encoding qubits in the magnetic orientation, or ‘spin’, of phosphorus nuclei embedded in silicon 20 years ago1. At about the same time, David DiVincenzo, a theoretical physicist then at IBM in Yorktown Heights, New York, and his collaborator Daniel Loss at the University of Basel in Switzerland proposed a way of storing information in the spins of mobile electrons inside semiconductors2. Both proposals led to a number of experimental demonstrations but, for a long time, the quality of the materials limited progress.
Building a quantum computer using silicon took years of “not very flashy” developments in materials science and engineering, says physicist Jason Petta of Princeton University in New Jersey. Physicists at the UNSW Centre for Quantum Computation and Communication Technology, which Simmons directs, have done much of that groundwork. And Simmons developed a manufacturing technique that requires fewer control leads, preventing inevitable issues of crowding once quantum devices scale up, she says. “I want to engineer everything out that isn’t essential and make things as simple as possible.”
In 2017, two groups reached a milestone when they designed the first fully controllable two-qubit devices in silicon. Petta and his collaborators achieved that feat3, as did a separate team4 led by Lieven Vandersypen at Delft.
Intel, which is investing US$50 million over 10 years at Delft, is now manufacturing multiple-qubit electron-spin devices for Vandersypen, in the same type of factory where it develops microprocessor-fabrication techniques. Industrial partners can help by providing reliably identical devices, he says.
“We hope that we can accelerate spin qubits to compete” with the more mature approaches, Clarke says. Simmons’ start-up aims to build a ten-qubit machine within five years. Google, IBM and a number of other companies and academic labs are all using different techniques to build quantum computers with around 50 superconducting qubits — and so is Intel itself, which is hedging its bets by supporting more than one technical approach.
Nature 553, 136-137 (2018)
Kane, B. E. Nature 393, 133–137 (1998).
Loss, D. & DiVincenzo, D. P. Phys. Rev. A 57, 120–126 (1998).
Zajac, D. M. et al. Science 359, eaao5965 (2018).
Watson, T. F. et al. Preprint at https://arxiv.org/abs/1708.04214 (2017).