Jülich, 23 October 2019 – After rumors had been spreading for weeks, now it’s official. Google and collaborators, including Forschungszentrum Jülich, have achieved a milestone in quantum computing called quantum supremacy as the researchers claim in the renowned journal “Nature”. The term marks the moment when a quantum computer, for the first time, outperforms state-of-the-art conventional computers for a specific task. While the Google quantum processor takes about 200 seconds, the world’s fastest supercomputers would need approximately 10.000 years to perform an equivalent task. Researchers of the Jülich Supercomputing Centre under the lead of Kristel Michielsen took part in verifying and benchmarking the quantum processor with simulations carried out on the Jülich supercomputer JUWELS.
Already in the eighties, Richard Feynman proposed that quantum computers can solve certain problems, like the simulation of quantum systems in physics and chemistry, much faster than conventional computers. “The demonstration of this challenging idea requires the development of a sufficiently large quantum processor with a sufficiently low error rate and the search for a problem that is difficult for a conventional computer but simple for a quantum computer”, says Kristel Michielsen.
For this purpose, Google fabricated the quantum processor “Sycamore”, having 53 working qubits, the counterpart of classical bits in a quantum computer. As the benchmarking problem the researchers used the sampling of the output of a random quantum circuit in which random operations on qubits are carried out. Sampling the output of such a quantum circuit produces a set of bitstrings, for example 0000101, 1011100, of which the probability distribution resembles a speckle pattern, a wave pattern produced by laser light scattered from a random surface. Sampling from this probability distribution with a conventional computer becomes exponentially more difficult as the number of qubits and the number of operational cycles in the circuit increases.
Cross-entropy benchmarking, a method to compare how often each bitstring is observed experimentally with its corresponding ideal probability computed via simulation on a conventional computer, was used for verifying and benchmarking “Sycamore”. In the regime of quantum supremacy, the computing cost for the cross-entropy is prohibitively large.
The intractability for calculating the cross entropy on conventional supercomputers does not only depend on the number of qubits but also on the complexity of the random circuit. Several so-called quantum supremacy circuits with 53 qubits have been identified, which could only be performed on “Sycamore”.
While Google’s quantum processor needed 200 seconds to sample the output of such a quantum circuit 1 million times, a state-of-the-art supercomputer would require 10,000 years to perform the equivalent task. Thus, quantum processors have reached the regime of quantum supremacy, a milestone for quantum computing.
For the simulation of the quantum random circuits various algorithms and codes were implemented on SUMMIT, the currently most powerful supercomputer in the world, Google Cloud servers, and JUWELS, the Jülich supercomputer. The researchers of the Jülich Supercomputing Centre used the Jülich Universal Quantum Computer Simulator (JUQCS), which set the world record in simulating arbitrary quantum circuits on a universal quantum computer with 48 qubits, for benchmarking “Sycamore” with random quantum circuits with up to 43 qubits.
Copyright: Forschungszentrum Jülich / Ralf-Uwe Limbach
Original publication: “Quantum supremacy using a programmable superconducting processor” by Google AI Quantum and collaborators, Nature23 October 2019, DOI: 10.1038/s41586-019-1666-5
Nature has kindly agreed to publish this paper as “free in perpetuity,” meaning “it will be completely free for everyone to read and download.”