Google is trying out error correction on its quantum processor.
The current generation of quantum hardware is called “NISQ”: noisy, intermediate-scale quantum processors.
“Intermediate-scale” refers to a qubit count that usually runs into the dozens, while “noisy” refers to the truth that current qubits often encounter errors, caused by problems setting or reading qubits, or by the qubit losing its state during calculations.
Google’s quantum computing group mentioned that the layout of its processor was chosen because it makes it easier to perform error correction.
The team now runs two different error correction schemes on the processor, and the results show that error correction clearly works, although more qubits and a lower inherent error rate are required before a correction makes sense.
Google chose a geometry in which all internal qubits are connected to four neighbors while those at the edge have only a pair of connections.
In what is likely the clearest demonstration yet, the researchers started the linear error correction system with a chain of five qubits and gradually added more until the chain reached 21 qubits.
As the chain gained more and more qubits, it became more robust, with the error rate between the five-chain and the 21-chain decreasing by a factor of 100. Errors still occurred, however, so the error correction was not error-free, and performance remained stable for up to 50 rounds of error checking.
Errors also occurred in the second error correction configuration, but most were caught, and the exact nature of the errors was generally derivable. Since the setup requires a more precise geometry to work, the team did not extend them beyond a limited number of qubits.
Ultimately, the entire system was in part extremely underperforming: the researchers attribute the underperformance to the effects of cosmic rays or local radiation sources hitting the chip.
Although the problems are not widespread, they are a problem and will increase as the number of qubits continues to grow, simply because the processors will be an ever-growing target.
Even with a 21 qubit chain, the error rate at the end was about one in 100,000 operations, which is enough to expect that a calculation can be continued to catch and correct errors.
All 21 qubits were used to encode a single logical qubit. Even the largest of the current processors could only hold two qubits with these systems.
For more information, read the original story in Arstechnica.