Researcher seen from behind adjusting cables on an unbranded quantum cryostat, no logos or readable text
AI-generated image

Research

Qunova says HI-VQE reached chemical accuracy on an 80-qubit iron-sulfur cluster run

Published · Last updated

Short answer: South Korean start-up Qunova Computing says its HI-VQE algorithm reached chemical accuracy in an 80-qubit simulation of the 4Fe-4S iron-sulfur cubane cluster, using an IBM Quantum System Two at RIKEN together with RIKEN's Fugaku supercomputer. The claim comes from a company press release published on 12 November.

According to the release carried by The Quantum Insider, Qunova's extrapolated ground-state energy for the cluster was −327.2482 ± 0.0007 Hartree, which it says agrees within chemical accuracy with a leading classical reference from the density matrix renormalisation group method (−327.2471 Hartree). Qunova describes HI-VQE as the first quantum algorithm to reach chemical accuracy at 80 qubits and says the result is more accurate than recent extrapolated trimmed configuration interaction results.

The company also compared cost with an earlier quantum approach. Qunova says that against a previously published sample-based quantum diagonalisation result for the same molecule, HI-VQE reached a lower energy of −326.9821 Hartree by its second iteration using 2,518 Fugaku node-hours, which it puts at roughly 300 times less classical compute.

HI-VQE, short for handover iterative variational quantum eigensolver, is offered as a Qiskit Function on IBM's platform. The IBM documentation describes it as a hybrid method in which the quantum processor samples candidate electron configurations and classical computers calculate the resulting wave function, so that only a compact subset of configurations is kept.

Iron-sulfur clusters sit at the core of enzymes involved in nitrogen fixation, hydrogen production and electron transport, which is why the molecule is a common stress test for new chemistry methods. Qunova says it plans to add orbital rotation and optimisation to HI-VQE and is looking for systems where it consistently beats classical methods; the release does not claim that yet.

← All news