A quantum computer has run the largest quantum program yet. The computer made errors, but a mathematical method was able to extract a meaningful result anyway. This error-mitigation technique could help quantum computers become more useful than their imperfect hardware would otherwise allow them to be.
Quantum computers have the potential to solve problems that even the best conventional supercomputers can’t. Yet, except for in a few specific cases, most existing quantum computers are either too small or make too many errors to achieve this. at IBM and his colleagues have made progress on both fronts.
Physicists write quantum computer programs by working out so-called quantum circuits. A circuit’s size is defined by how many operations, like turning a 0 into a 1, it contains and how many qubits, or quantum bits, it uses. The larger it is the more complex a problem the associated program can solve. But there are also more chances for errors to creep in.
Advertisement
Previously, the team’s largest meaningful quantum circuit used 26 qubits and contained 1080 individual operations. Now, the researchers have run a quantum circuit with 127 qubits and over 1700 individual operations.
The researchers developed this circuit to calculate what happens to a chain of electron-like particles when they are abruptly forced to interact with each other. They chose this problem because conventional computers can solve only some versions of it but must approximate solutions for most.
The IBM team’s quantum computer had qubits made from tiny wires that conduct electricity perfectly. Such superconducting qubits are typically programmed by being hit with pulses of microwaves, so running the new program was equivalent to programming a microwave pulse sequence.
Having many qubits doesn’t guarantee useful computation results because quantum computers often make errors, says IBM team member . The team could not eliminate them all because some come from the computer’s hardware. Instead, the researchers worked out how to convert the error-laden answer to a clear one.
The trick was to repeat the calculation with pulses adjusted in such a way that the errors increase every time. By doing this, they learned how the result of the computation changes and how to mathematically reverse the errors. This reversal then produced a solution equivalent to what a computer that makes no errors would have calculated. It matched approximations made by conventional computers. Kim presented the work at the in New York City on 9 November.
Kandala says that this experiment shows that for some problems where previously the researcher’s best bet was to use approximations, error mitigation opens the door for large but imperfect quantum computers to do better.
Looking at the computation as a physics experiment, programming computational steps into so many qubits is impressive, says at the University of Illinois Urban-Champaign. To fully realise the promise of quantum computing, researchers will have to develop computers that correct their own errors as they go, but for now error mitigation is a necessity if existing machines are to be useful, says at the University of Waterloo in Canada.
