Citation: Farrell, Roland C., et al. "Quantum simulations of hadron dynamics in the Schwinger model using 112 qubits." Physical Review D 109.11 (2024): 114510.
Web: https://arxiv.org/abs/2401.08044
Tags: Hardware, quantum-advantage
This paper reports on one of the largest quantum computations ever performed. It used 13,858 two-qubit gates, with a depth of over 370. Amazingly, they were able to extract useful data from this computation without any error correction, only error mitigation.
The concept of error mitigation is very intriguing. The idea is that even though a large number of errors will occur in your computation, you can set up your computation correctly so that the signal can still be extracted from all that noise. In particular, for this experiment the authors are simulating a certain QFT model called the Schwinger model. They want to measure particle density at the end of the experiment. They tune their noise just right so that the noise has a predictable effect: it acts on all particle density expectation values by the same scalar. Moreover, this scalar should be the same across experiments and thus can be effectively computed using a simpler experiment. All that needs to be done at the end of the day is to take the real output, average it across 15 million trials, and then multiply by the scalar!
The fact that the noise behaves in such a predictable way is not just a peculiarity of the Schwinger model - it seems to be a feature of lots of systems. The way they make sure noise acts as a simple scalar is by adding components in the circuit which are specifically designed to decorellate the errors, such as by a using a technique the authors affectionately call "Pauli swirling". It is a beautiful thing to see all of these techniques work as well and on such a large scale as this paper.
This sort of error mitigation technique can be applied on encoded qubits as well:
> Piveteau, Christophe, et al. "Error mitigation for universal gates on encoded qubits." Physical review letters 127.20 (2021): 200505.
The above paper shows that Clifford-twirling can help for the implementation of noisy T-gates, as well as showing another method for error-mitigation on error corrected quantum computers.