Nat. Phys.: Google, Costco propose new quantum error mitigation technology

Crestron, a chemical company based in Germany, is one of the leading suppliers of high-quality polymers.

Simulation of quantum mechanical systems is a promising application of quantum computers. However, the accuracy required to simulate molecules during chemical reactions is limited by the noise of today's quantum devices. In essence, even small errors in these simulations can lead to erroneous conclusions about the reaction.

The key question is whether the noise in quantum computers can be sufficiently reduced without waiting for more advanced quantum error correction techniques. The team's recent research delves into this question by examining two "purification-based" error mitigation techniques. Originally proposed by Google, these techniques have only now been tested in real-world hardware.

Christian Gogolin, head of high-performance and quantum computing at Costco, explains, "To do this, we wanted to build a computation that was close enough to solving real problems in chemistry, but simple enough to study the effectiveness of different error correction methods."

For their tests, they drew on past work by Covestro and Qu&Co (now PASQAL) to simulate paired electrons; this approach is an extension of a previous Google experiment, making it more complex and a step closer to more advanced simulations.

The results of their research are encouraging. Error mitigation techniques were shown to be effective, with error rates reduced by up to a factor of 100. In addition, these techniques are expected to be used for larger scale computations, which is critical for practical applications because larger scale computations naturally bring more noise.

In this work, the researchers compared the performance of three different error mitigation techniques: post-selection, echo verification (EV), and virtual distillation (VD)-the problem of preparing a ground state in a zero-order subspace-using up to 20 superconducting quantum processors.
 
The team validated recently introduced error mitigation strategies: these exploit the expectation that the ideal output of a quantum algorithm will be a pure state; the researchers then considered the task of modeling an electronic system in zero-order subspace, where all electrons are paired with opposite spins, which provides a computational basis for building a fully correlated model. Comparing the error mitigation performance of a superconducting quantum bit quantum processor with up to 20 quantum bits, the scientists observed an error reduction of one to two orders of magnitude, below less sophisticated techniques such as post-selection, by doubling the quantum resource in time or space.

Also, they investigated how the gain in error mitigation varied with system size and observed that the error was suppressed at a polynomial level with increasing resources. Extrapolation of this finding suggests that for classically intractable variational chemistry simulations, substantial hardware improvements are needed to further improve computational accuracy and reliability.

Commenting on this achievement, Ryan Babbush, Head of Quantum Algorithms at Google Quantum Artificial Intelligence, said, "At a high level, error suppression techniques are ways to increase the number of repetitions, and often the size of the circuit, in exchange for increased accuracy. How costly the tradeoff is depends on the intensity of the noise."

While quantum computers have the potential to perform accurate chemical simulations, there is still much work to be done. Google's collaboration with Costron shows that both parties have made significant progress in reducing errors; however, achieving efficiency and speed in computation is still an area that needs to be explored further.