Deeptech startup H.I.Council performs world's first 108-qubit FeMoCo molecule calculation on IBM quantum computer, paving the way for chemical precision calculations previously thought solvable only by quantum computers.

H.I.Council, a deeptech company located in Shibuya, Tokyo, and represented by Futoshi Hamanoue, has achieved a world-first by performing a 108-qubit electron structure calculation of the FeMoCo molecule, the active center of nitrogen fixation enzymes, on a real IBM quantum computer (ibm_pittsburgh, Heron r2, 156 qubits). The results have been published in academic papers. This research demonstrated that the chemical precision calculation of FeMoCo, a long-standing "killer application" target for quantum computers, can be achieved for a 48-qubit system using a classical DMRG-AFQMC pipeline leveraging quantum chemistry knowledge. Simultaneously, it quantitatively clarified the limitations of current quantum computers, termed the "coherence wall," for the first time. The papers are available on the preprint server ChemRxiv (DOI: 10.26434/chemrxiv.15001770/v2) and the open-access repository Zenodo (DOI: 10.5281/zenodo.19463795). Related technology is patent pending (JP Patent Application 2025-182361). The background of this research is that FeMoCo (iron-molybdenum cofactor, MoFe7S9C) is the active center of the nitrogenase enzyme, responsible for biological nitrogen fixation. Accurate calculation of its electron structure is crucial for designing new catalysts in the agrochemical and fertilizer industries. In 2017, Reiher et al. estimated the computational scale required for FeMoCo electron structure calculation to be 108 qubits, positioning it as a "killer application" for quantum computers. Since then, FeMoCo has been recognized as a long-term goal in the quantum computing industry. However, current NISQ quantum computers are limited by noise, and no prior research had successfully completed FeMoCo-scale calculations on real machines and quantitatively validated the results. H.I.Council's research achieved four main results: First, they performed the world's first 108-qubit FeMoCo calculation on a real machine, reaching a statistical error of ±0.67 mHa using phase-less auxiliary-field quantum Monte Carlo (ph-AFQMC). This is unprecedented precision for this scale on quantum hardware, though further systematic error improvement is needed for chemical precision. Second, through multi-scale spatial correlation measurements, they quantified the systematic decay of correlation amplitude with increasing circuit depth. They showed that for 108 qubits, the extractable correlation signal is capped at r ≈ 0.03, naming this the "coherence wall," which is the first systematic measurement of the quantitative limits of current quantum computers. Third, they achieved chemical precision (+1.07 mHa) for a 48-qubit FeMoCo electron structure calculation using classical computational methods deeply rooted in quantum chemistry (DMRG multi-determinant trial wavefunction + ph-AFQMC). This demonstrates that chemical precision for FeMoCo, previously believed to be solvable only by quantum computation, can be achieved by leveraging quantum mechanics knowledge through alternative classical quantum chemistry methods. This signifies not a refutation of quantum computers, but the successful realization of one of multiple pathways to translate the understanding of quantum mechanics into practical applications, representing a "generalized quantum advantage." Fourth, based on these findings, they proposed three concrete directions for improving next-generation quantum hardware: moving away from isotropic thermal equilibration, prioritizing connectivity quality over qubit count, and designing hardware bias to maintain the information backbone. These are verifiable hypotheses presented as a call to the quantum hardware community.