Commercial Quantum Computing Pushes On

　　Intel Labs has announced 17-qubit CMOS superconducting demonstration platform that it says brings quantum computing closer to commercial development. Intel delivered the prototype this week to research partner QuTech (Delft, Netherlands), which will test it on a suite of quantum algorithms to prove the design’s commercial relevance.

　　Intel entered the quantum computer race in 2015, when it invested $50 million to advance quantum computing in a collaborative development effort with QuTech. The researchers aim to accelerate the development of commercially useful quantum computers by pairing Intel’s CMOS design and manufacturing expertise with QuTech’s expertise in connecting, controlling, and measuring multiple, entangled qubits.

　　At this year’s International Solid-State Circuits Conference (ISSCC 2017), the collaborators demonstrated key circuit blocks for an integrated cryogenic-CMOS control system that cools to 20 milli-Kelvin (250 times colder than deep space), presenting their work in a paper titled “15.5 Cryo-CMOS circuits and systems for scalable quantum computing.” They have also demonstrated a scalable “surface code” error-correction scheme that enables spatial multiplexing, describing that work in an American Physical Society (APS) paper co-authored by multiple QuTech engineers and David Michalak, a quantum computing researcher-in-residence at Intel.

　　The research collaborators are working on two parallel efforts to perfect quantum values: a spin-qubit fabrication flow on Intel’s 300-millimeter CMOS process and the packaging advances in the superconducting prototype announced this week. With the new packaging system (see photos), the prototype realizes 17 qubits for quantum computing via an architecture that supports full error correction, improves yield, and boosts performance, according to Intel.

　　Managing the effort at Intel is a duo dubbed “a superposition of two Jims” by Intel colleagues: Jim Clarke, director of quantum hardware, and Jim Held, director of emerging technology research. (Superposition, the founding principle of quantum computing, is the ability to harness two values — 1 and 0 — simultaneously in a single qubit.)

　　“Intel’s quantum computer hardware is relatively young, but it is moving fast,” said Clarke. “I liken it to the Apollo mission, which against all odds reached the moon in just a few years. Likewise, the Intel-QuTech collaboration is a quick dash to commercial quantum computers.”

　　Held said the collaborative program is assembling “an entire software stack for quantum algorithms, from qubit operations to the hardware and software architectures required and the quantum applications themselves. In 2016, we built a large-scale qubit simulator with 42 qubits — since extended to 45 qubits — running on an Intel supercomputer, so that we have a platform to develop quantum software that is ready for use at the same time our quantum hardware is ready for commercialization."

　　Intel claims its quantum computer architecture solves many of the problems encountered over the years by other teams, such as those at D-Wave Systems Inc., Google Labs, IBM, Microsoft Labs, Quantum Circuits Inc., Rigetti Computing, and the U.S. National Institute of Standards and Technology (NIST).

　　For instance, instead of cooling its hardware to a consistent temperature, such as the easily obtained temperature of liquid nitrogen (77 K, or –320 F) or even that of liquid helium (4 K, or –452 ), Intel used various helium isotopes to cool its qubits to extremely low temperatures (20 mK, or –459 F). The goal of the extreme cooling is a more error-free, and thus more commercially relevant, design, Clarke said.

　　“However, the key circuit blocks for our integrated cryogenic-CMOS control system only need the more easily obtainable cooling to 4 K,” he said.

　　Intel has also moved away from the standard wire bonding techniques that other labs have used for quick proof-of-concept demonstrations, instead opting for a scalable method that enables 10 to 100 times more signals into and out of the qubits.

　　“We have codesigned the chip and its package for the long term, in order to realize a quantum computer that will be more commercially relevant and more general-purpose and that will ultimately add significantly to Intel’s bottom line at the end of the quantum computer race,” said Clarke. Whereas D-Wave, for example, is pursuing quantum annealing to solve optimization problems, Intel aims to solve many more problems that have proved intractable for conventional digital computers. Nonetheless, Intel has stopped short of promising a “universal” quantum computer that would solve all such problems.

　　Intel’s Components Research group in Oregon and its Assembly Test and Technology Development team in Arizona collaborated on the codesign of the chip and packaging technologies, which minimize the radio-frequency (RF) interference between qubits.

　　Of course, Intel’s 17-qubit superconducting chip is just a proof of concept; D-Wave, in contrast, has leapt from 1,000 to 2,000 superconducting qubits this year to solve commercial optimization problems using quantum annealing. Intel says it chose 17 qubits as the minimum number necessary to prove its surface-code error-correction scheme, which it says would be scalable to commercially relevant quantum computers with spatial multiplexing.