Quantum computers are inherently so error-prone that they can only be operated with error correction. Two such procedures are just reaching the limit for practical application.
Sycamore demonstrated its “superiority” back in 2019: Back then, the quantum computer from the Google Quantum AI group performed a certain task, which was devised for this purpose, faster than a classic computer. Now the team has also developed the error correction of such a system on the basis of this device.
If the former was still important from an academic point of view, but not practical, the group has now “taken a big step towards practical application”, says Tommaso Calarco when asked by the Science Media Center. The scientist from Forschungszentrum Jülich is involved in building quantum computers, but was not involved in the current publication.
In the journal “Nature”, Google Quantum AI demonstrates the use of two error correction methods. These procedures are essential for every quantum computer. Because its qubits – the quantum physical counterparts of the classic computer bits – are highly sensitive to environmental influences and often change their state during the calculation. As a result, they repeatedly produce incorrect results. Noticing and correcting this without destroying the fragile state is the task of the error correction process.
In the Sycamore computer, each of the 54 qubits consists of a tiny, superconducting circuit. A common method for error correction is to link these units in such a way that each qubit that is needed for the calculation technically consists of several hardware qubits. If one of them is faulty, this reveals itself when comparing it with the linked hardware qubits.
Google Quantum AI shows two different methods
This principle can be put into practice in a wide variety of ways. Google Quantum AI has now implemented two of them. In the first method, the so-called repetition code, the hardware qubits are linked like a chain. It is the easier of the two techniques; It can be used to capture either only one or the other of the two possible types of errors. The second method, the surface code, consists of a two-dimensional arrangement of the qubits and allows both types of defects to be detected simultaneously.
One of the most important results of the study is the experimental proof of a theoretically predicted phenomenon: as the number of hardware qubits per (logical) qubit increases, the error rate decreases exponentially. The researchers at Google Quantum AI were indeed able to measure this for the case of chain-shaped repetition codes by increasing the number of interconnected hardware qubits from 5 to 21.
Each newly added hardware qubit improves the accuracy disproportionately. “If this were not the case, the resources required for error correction would grow so rapidly that the advantages of the quantum algorithms themselves would be eaten up,” write Sven Ramelow and Helen Chrzanowski in their comment on the study at the Science Media Center. Both also research quantum systems at the Berlin Humboldt University.
Another important step forward in the current study is the fact that the Google team has maintained this error correction mechanism for over 50 calculation cycles. The error rate was partly at a level that is necessary for the actual use of the quantum computer.
In contrast, independent experts rate the implementation of the more complex, but much more practice-relevant method of the surface code as not yet sufficient. Here, the scientists at Google Quantum AI have so far not been able to reduce the error rate to a level that is needed for real calculations.
In general, the final step – interconnecting the qubits including error correction so that they can perform calculations – is still pending. Nevertheless, quantum computer expert Peter Zoller from Innsbruck speaks of a success through decisive technological improvements at all levels in the Sycamore processor. “The road to error correction is a marathon, progress in this direction is incremental but remarkable.”
Spectrum of science
We are partners of Spektrum der Wissenschaft and would like to make well-founded information more accessible to you. Follow Spectrum of Science if you like the articles.
Original article on Spektrum.de