For nearly a decade, one molecule has been the "holy grail" of quantum computing: FeMoCo, the iron-molybdenum cofactor at the heart of how certain bacteria turn air into fertilizer. A Tokyo-based deep-tech startup called H.I.Council has just performed the world's first 108-qubit-scale calculation of this molecule on a real quantum computer. The results reveal both the promise and the hard limits of today's quantum hardware.

What Is FeMoCo and Why Does It Matter?

Some microorganisms on Earth can convert atmospheric nitrogen gas into ammonia at room temperature and normal pressure, a feat humans can only replicate industrially using the Haber-Bosch process, which requires extreme heat (over 400°C) and pressure (200+ atmospheres). This industrial process accounts for roughly 1.8% of global CO₂ emissions and consumes 3–5% of the world's natural gas.

The secret behind nature's trick lies in an enzyme called nitrogenase, whose active center contains a molecule known as FeMoCo (MoFe₇S₉C). If scientists could fully understand how FeMoCo works, it could lead to revolutionary new catalysts that produce fertilizer far more efficiently, a breakthrough with massive implications for agriculture, energy, and climate.

But FeMoCo's electronic structure is so complex that classical supercomputers cannot accurately simulate it. In 2017, a team led by Markus Reiher at ETH Zurich estimated that simulating FeMoCo would require around 108 qubits, and designated it as quantum computing's "killer application." Since then, cracking FeMoCo has been a north-star goal for the entire quantum computing industry.

H.I.Council's Breakthrough: Four Scales on Real Hardware

H.I.Council, a deep-tech company based in Shibuya, Tokyo (founded by Futoshi Hamanoue), ran FeMoCo electronic structure calculations at four scales, 48, 56, 80, and 108 qubits, on IBM's latest quantum processor, ibm_pittsburgh (Heron r2, 156 qubits).

The team achieved four key results.

First, they completed the world's first 108-qubit FeMoCo calculation on actual quantum hardware. Using a method called phaseless auxiliary-field quantum Monte Carlo (ph-AFQMC), they reached a statistical error of just ±0.67 milliHartree, an unprecedented level of precision at this scale. However, full chemical accuracy (matching experimental results) remains a challenge requiring further systematic error reduction.

Second, by measuring spatial correlations across multiple scales, they systematically quantified how correlation signals decay as circuit depth increases. At the 108-qubit scale, the extractable correlation signal was capped at roughly r ≈ 0.03, even after applying about 20 different Bayesian estimation methods. They named this limit the "coherence wall," marking the first systematic measurement of current quantum hardware's quantitative boundaries.

Third, and perhaps most surprisingly, using classical computational methods rooted in quantum chemistry (DMRG multi-determinant trial wavefunctions combined with ph-AFQMC), they achieved chemical accuracy (+1.07 milliHartree) for FeMoCo at the 48-qubit scale. This is a landmark result because the quantum computing industry has long believed FeMoCo's chemical accuracy could only be achieved via quantum computation. H.I.Council showed that quantum mechanics knowledge can be leveraged through a different route, classical quantum chemistry, to reach the same goal.

Fourth, the team proposed three specific directions for next-generation quantum hardware improvement: moving away from isotropic thermalization, prioritizing connectivity quality over raw qubit count, and designing hardware bias to preserve information backbones. These are testable hypotheses offered as constructive suggestions to the quantum hardware community.

"Broad Quantum Advantage", A New Perspective

A crucial takeaway from this research is that it does not deny quantum computers. Rather, it demonstrates that there are multiple pathways to translating quantum mechanics knowledge into practical results. Direct quantum computer execution and classical quantum chemistry methods both harness the same underlying understanding of quantum physics, just through different routes.

The research team frames this as "broad quantum advantage", a perspective gaining traction across the field. Quantum-inspired methods, quantum-classical hybrid approaches, and pure quantum computation are all different forms of putting quantum mechanics to work. By evaluating both direct quantum execution and classical quantum chemistry within the same experimental framework, H.I.Council has clarified where each approach currently stands and what challenges remain.

Open Science and a Call for Verification

H.I.Council openly acknowledges that the "coherence wall" was observed specifically on IBM's Heron r2 hardware. They are calling on the broader community, including teams working with Google, IonQ, Quantinuum, Pasqal, Fujitsu, and RIKEN, to independently verify whether similar limitations exist on other quantum platforms.

The paper makes all IBM Quantum job IDs and related data publicly available, enabling any external researcher to retrieve and analyze the original bitstrings directly. In a field where trillions of yen (tens of billions of dollars) in public funding are being invested globally, this kind of transparent, evidence-based verification is essential for investors, policymakers, and taxpayers alike.

Looking ahead, H.I.Council plans to apply its pipeline to calculating the reaction intermediates of the full nitrogen fixation catalytic cycle (the eight intermediate states known as the Lowe-Thorneley cycle), aiming to unravel the mechanism of biological nitrogen fixation at the molecular level. The paper has been published as open access on ChemRxiv and Zenodo, with related technology under patent application in Japan.

In Japan, this research is sparking debates about the real potential of quantum computing versus the hype. Some are impressed by a small startup tackling a problem that major tech giants have circled for years. Others question whether the "coherence wall" finding suggests we're further from useful quantum computing than industry marketing implies. What's the quantum computing landscape like in your country? We'd love to hear your perspective in the comments.

References