Qunova gets an 80-qubit chemistry workflow to agree with a leading classical answer

Qunova Computing used a quantum computer and a conventional supercomputer together to calculate the energy of an iron-sulfur cluster—a small molecular structure found in proteins that move electrons and help chemical reactions happen.

Calculating this energy helps scientists understand how a molecule behaves. Better calculations could eventually help researchers develop catalysts, materials and medicines.

The two computers shared the work. IBM’s quantum computer used 80 physical qubits—the individual computing units on its chip—to generate possible arrangements of the molecule’s electrons. Fugaku, RIKEN’s conventional supercomputer, then evaluated selected arrangements and calculated an energy estimate.

Qunova supplied HI-VQE, the algorithm and software that organized this process. It guided the quantum computer, selected promising arrangements and used the supercomputer’s calculations to improve the next round. The aim was to reach an accurate answer while keeping the amount of computing manageable.

The final estimate closely matched a result from DMRG, one of the strongest conventional methods for this problem. The difference was within “chemical accuracy,” a commonly used tolerance for chemistry calculations. This is a meaningful benchmark: the combined workflow reached a demanding accuracy target on a difficult molecular problem. The reported final value included an extrapolation—an estimate of where the result would converge as the calculation was expanded.

This approach can work with today’s imperfect quantum computers without using error-corrected logical qubits. Logical qubits combine multiple physical qubits to protect quantum information while a calculation runs. Here, conventional software instead used chemical rules to filter or adjust some noisy measurement results. It processed the output; it did not repair the qubits themselves. Enough useful information still had to survive the quantum calculation.

Why this matters: Related methods have also been explored by IBM, RIKEN and Lockheed Martin. For businesses, they offer a possible route to useful quantum chemistry before large, fully error-corrected machines arrive. Qunova’s contribution is its particular way of selecting and refining the electron arrangements, together with the performance reported in this experiment.

Qunova also reported that one intermediate stage used about 300 times fewer supercomputer resources than an earlier hybrid method. That is encouraging progress, although the comparison covers one stage rather than the complete calculation.

Our call is MONITOR. Qunova has shown that its approach can calculate a difficult molecule’s energy accurately. A valuable next result would show how it helps researchers choose a catalyst or design a material, with a better answer, less computing time or lower cost than conventional methods.