Quantinuum has always had a strategy of providing users with qubits that demonstrated the highest possible measure of qubit quality, as measured by both gate fidelity metrics as well as the Quantum Volume measure originally proposed by IBM. In fact, in 2020, they declared a goal of continuing to improve the Quantum Volume measure of their machines by a factor of 10X each year. And to their credit, they have continued to meet or beat this goal every year.
Now, they have reached a new level of qubit quality performance on their 20 qubit H1-1 processor and achieved a two-qubit gate fidelity measure of 99.914% and a Quantum Volume measurement of 1,048,576 (or 220). The Quantum Volume benchmark was achieved with a circuit that used 20 qubits successfully running a randomized circuit that is 20 layers deep. Although the circuit used in a Quantum Volume test would not be useful for a commercial application, the advantage of this benchmark is that it is hard to “game” to unfairly show better results. We should also point out that this same machine is also in daily use serving end users. So it is not a special machine developed by engineering just to be used for these tests.
We should note that the company has maxed out the capability of the H1-1 processor, since that machine only contains 20 qubits. Another impressive result of Quantinuum’s test is that they achieved very high consistency in performance amongst all the qubits. In other quantum hardware implementations there can be a wide range of qubit quality performance with some qubits being good and other being not-so-good. In this latest test, every qubit pair on this H machine achieved a 99.9%+ two qubit gate fidelity.
The engineers at Quantinuum achieved this result by implementing various improvements in the laser and optics systems inside the machine. The original H1 processor was introduced in October 2020 with 10 qubits and a Quantum Volume level of 128 (or 27). And the company has been implementing many incremental upgrades over the years to both improve qubit quality and double the number of qubits.
The next step in their development will be to take all the improvements and learnings from the H1 and apply them to the next generation H2 processor. Currently the H2 supports 32 qubits but the company intends on expanding this in the future as well as improving the qubit fidelity metrics to levels similar to what they have achieved with the H1.
This demonstration is one more step towards reaching a Quantum Advantage level that can provide commercial benefit. Although many people are investigating error correction techniques, we should point out that error correction becomes increasingly more effective and efficient as the raw physical qubit fidelities improve. Although there is still disagreement amongst quantum researchers on whether commercial useful applications will be able to run on uncorrected NISQ level machine, the chances of this occurring also become higher as the raw qubit fidelities improve. So this development will be helpful for use in both non error corrected NISQ implementation as well as more efficient error correction implementations in fault tolerant machines.
We can’t describe this development as reaching a Sputnik moment, but it is another forward step in the long path of providing high quality quantum computation and it is synergistic with Quantinuum and Microsoft’s research results last month of demonstrating error correction circuits on their processors. For more information about this result, you can view a news release available on the Quantinuum website here. Also, for those who want to dig into the technical details, Quantinuum has posted the technical data of their Quantum Volume test on GitHub here and also posted on GitHub the hardware specifications here.
April 16, 2024
