July 7, 2020 | Written by: Jay Gambetta and Doug McClure
Categorized: Quantum Computing
Share this post:
Six months after the first Quantum Volume 32 demonstration at the beginning of 2020, IBM now hosts eight quantum computing systems that cross the QV32 performance threshold available to IBM Q Network organizations. Six of these are completely new systems — three 27-qubit Falcon processors, and four 5-qubit Canary processors.
Improvements to hardware design and a new “target rotary” pulsing technique enabled the rapid roll-out of our QV32 systems. Target rotary increases the fidelity of two-qubit entangling operations while mitigating spectator errors. Our entire fleet now includes 22 systems that deliver generational learning from all aspects of quantum research and development to our clients and users.
Continuing to build bigger and better quantum systems. Eight quantum systems from our Falcon r4 (Paris, Montreal, and Toronto), Canary r3 (Athens, Bogota, Rome and Santiago), and Penguin r3 (Johannesburg) design now achieve QV32.
As quantum systems with increasing numbers of qubits are developed, performance benchmarking in complete generality becomes an intractable problem. We know that overall device performance requires high fidelity one- and two-qubit gates; however, the computational power of the system is also determined by other parameters such as the number of qubits, their connectivity, and unintended interactions between neighboring qubits.
To account for this, we need to consider a holistic metric such as the Quantum Volume: a hardware-agnostic metric defined to take into account the number of qubits, connectivity, as well as gate and measurement errors. Material improvements to underlying physical hardware, such as increases in coherence times and reduction of device crosstalk, as well as software improvements to circuit compiler efficiency, can both result in measurable progress in Quantum Volume.
*Note: IBM Quantum now include eight systems that now achieve QV32.
Reducing spectator errors
These new, more powerful systems reflect our recent efforts to reduce spectator errors, both by lowering the upper error-bound set by hardware design and with new control sequences.
“Spectator” errors are another important limitation to quantum volume: nearby qubits that aren’t participating in a gate can corrupt system performance in ways not apparent when the gate is performed in isolation. This corruption can be in the form of unwanted entanglement or classical crosstalk with the spectator qubits during the gate.
We are currently investigating tools to detect and mitigate spectator errors. In our recent arXiv release, “Reducing unitary and spectator errors in cross resonance with optimized rotary echoes,” we show that “target rotary” pulsing — a resonant drive of the target qubit during a cross-resonance (CR) gate — corrects gate errors that result from the standard CR echo pulse while simultaneously reducing undesired interactions between the target qubit and its spectators.
Not only does this work improve the performance of today’s IBM Quantum systems, the new models presented in this paper reflect a deeper understanding that will inspire the next generation of devices and further research. While Quantum Volume still needs to improve substantially to provide quantum advantage in practical commercial applications, techniques such as target rotary pulsing, which addresses often-neglected errors, bring us closer to this goal.
Also contributing to this article: IBM Research scientists Isaac Lauer, Easwar Magesan, Emily Pritchett, and Neereja Sundaresan.
Quantum starts here