COLLEGE PARK, MD — December 11, 2018

IonQ announces the successful development of two state-of-the-art quantum computers with record-breaking performance. IonQ’s systems are the first in the market that store information on individual atoms. They are more accurate and can perform more complex calculations than any quantum computer built to date. In 2019 they will be made available to leading researchers for applications in medicine, chemistry, energy, logistics and other areas where conventional computers fall short.

The results were announced by Christopher Monroe, IonQ’s co-founder and chief executive, at the Quantum for Business conference in San Francisco on Tuesday.

Quantum computers harness the unusual physics of very small particles—quantum mechanics—to solve problems that are beyond the capability of conventional devices. To achieve this potential, however, they must isolate and manipulate quantum systems to create quantum versions of computer bits, called qubits. Engineers around the world have been racing to build computers that can control more qubits, for longer calculations, and with fewer errors.

The qubits in the new IonQ systems are individual atoms of the rare earth element ytterbium, suspended in a vacuum. Information is stored, processed and retrieved from the atoms using precisely aimed laser beams. This approach, called ion trapping, is theoretically powerful but technically challenging. The first quantum computers in the market—from companies like IBM, Google, and Rigetti—create qubits on silicon chips that are cooled to nearly absolute zero. While this approach is said to benefit from the mature manufacturing technologies of the semiconductor industries, solid-state quantum computers so far have proven to be unstable and error-prone, limiting them to short and simple calculations.

“In the lab, we saw the potential of ion trap computing,” Monroe said. “And so we started a company with an incredible team to commercialize it.” Monroe, Bice Seci-Zorn Professor of Physics at the University of Maryland, has been a leader in developing ion trap quantum computing since 1995. In 2015, he co-founded IonQ with collaborator Jungsang Kim, an engineering professor at Duke.

“After two years of work, our against-the-grain bet is paying off,” Monroe said. “The IonQ System is robust and industrial strength. Even at this early stage, the results show the ion trap design has all the advantages we expected and more.”

On measures of capacity, accuracy and other key benchmarks, IonQ’s system has surpassed all the other quantum computers in the market. It has stored 160 qubits and performed operations on 79 qubits, a record. Its gate fidelity—a measure of the accuracy of logical operations—is greater than 98% for both one-qubit and two-qubit operations on average in a 13-qubit configuration. This means it can handle longer calculations than other commercial quantum computers.

The potential of the IonQ system, and the flexibility of a trapped-ion architecture, can be seen in benchmark results using the Bernstein-Vazirani Algorithm. This application tests the ability of a computer to determine an encoded number—called an oracle—when allowed only a single yes/no question. For a 10-bit oracle (a number between 0 and 1023), a conventional computer would succeed 0.2% of the time. The IonQ system ran the algorithm on all possible 10-bit oracles and observed an average success rate of 73%. That’s a better result on a more complex version of the calculation than any result yet published for a quantum computer.

Preliminary benchmarks can be seen in the appendix. The company will publish further details in peer-reviewed journals in the coming months, including a new quantum chemistry simulation well beyond what has been done previously. IonQ’s engineers believe that the system’s performance will improve steadily.

“Atomic qubits are perfect, so we don’t need to spend any more time on the physics. Our work is now focused on systems engineering and integration,” said Kim. “Every quarter we make our lasers more precise and our ion trap package more capable through the increased sophistication of our control software. Each improvement dramatically expands our processing power.”

In 2019, the company will invite leading players in government and industry to a private beta test of its systems.

“If your company runs up against the limits of conventional computers in fields like medicine, logistics or finance, now is the time to act,” said Stewart Allen, IonQ’s Chief Product Officer. “If you start developing the algorithms and software for your business, you’ll be positioned to get a significant competitive advantage when the capacity of quantum computing hardware surpasses conventional computers. Based on the results we’ve seen, that day will be sooner than many people expect.”

**Media contact:**

Jessica Shapow

P: (301) 761-3948

E: press@ionq.co

See also: http://ionq.co

Preliminary benchmark test results on IonQ hardware as of December 10, 2018.

Qubits are the basic unit of information storage on a quantum computer. After they’re initialized, logical operations—called gates—are performed on them.

Maximum loaded 160 qubits

Single-qubit gates performed on up to 79 qubits

Two-qubit gates performed on all pairs of up to 11 qubits

Gate fidelity is a measure of the accuracy of a single gate. Gates that manipulate one qubit at a time are less complex and less error-prone than gates that operate on two qubits. The following benchmarks were captured on a fully-connected 11-qubit configuration.

Single-qubit gates >99%

Two-qubit gates >98%^{*}

Single-qubit gates 99.97%

Two-qubit gates 99.3%^{*}

Single-qubit gates >99%

Two-qubit gates >96%^{*}

^{*} not corrected for state preparation and measurement errors.

The Bernstein-Vazirani Algorithm is a basic test of the ability of a quantum computer to simultaneously evaluate possibilities that conventional computers must calculate one at a time. The complexity of the test is determined by the maximum length in bits of an oracle—an arbitrary number the computer must determine.

10-qubit oracle success rate 73.0%

Classical computer success rate ~0.2%