Honeywell and Quantinuum: Catalyst Design via Quantum Chemistry

Industrial nitrogen fixation via the Haber-Bosch process consumes roughly 1-2% of global energy supply annually, running at 400-500 degrees Celsius and 150-300 atmospheres of pressure to force nitrogen and hydrogen together over an iron catalyst. Nature solves the same problem at room temperature and atmospheric pressure using the nitrogenase enzyme, whose catalytic core is an iron-sulfur (Fe-S) cluster called the FeMo-cofactor. Understanding the electronic structure of this cluster well enough to engineer synthetic mimics has been a goal of computational chemistry for decades, but classical quantum chemistry methods scale exponentially with the number of correlated electrons and cannot treat the full active space of the FeMo-cofactor exactly. Honeywell and its quantum computing subsidiary Quantinuum identified this problem as a near-term target for quantum hardware.

The core computational approach is the Variational Quantum Eigensolver (VQE), which uses a quantum processor to prepare a parameterized trial wave function (the ansatz) and measure its energy, while a classical optimizer adjusts the parameters to minimize that energy toward the true ground state. Crucially, Quantinuum’s work employs active space methods borrowed from classical quantum chemistry: the full molecular orbital space is partitioned into a small active space containing the most strongly correlated electrons, which is treated on the quantum processor, while the remaining electrons are handled classically via methods like CASSCF or NEVPT2. For the Fe-S cluster models studied here, the relevant active space contains on the order of 12 electrons in 12 orbitals, a system that strains classical exact diagonalization but sits within reach of current trapped-ion hardware when compiled with high-fidelity gates.

Gate fidelity is the key differentiator that makes trapped-ion hardware compelling for chemistry over superconducting alternatives at current qubit counts. Quantinuum’s H2 processor achieves two-qubit gate fidelities above 99.9%, compared to roughly 99.0-99.5% for leading superconducting devices. For a VQE circuit involving hundreds of two-qubit gates, that fidelity difference compounds dramatically: a 300-gate circuit with 99.9% per-gate fidelity has roughly 74% probability of executing without error, while the same circuit at 99.0% drops to about 5%. Chemistry simulations require reaching so-called chemical accuracy, an energy precision of 1 kcal/mol (about 1.6 millihartree), which is the threshold at which computed reaction energies are trustworthy for predicting whether a catalytic cycle will proceed. Achieving this accuracy demands low noise, not simply more qubits.

The 2024 results demonstrated chemical accuracy for 12-electron active space Fe-S cluster models on the H2 system, a meaningful step because it establishes that current trapped-ion hardware, without error correction, can already treat the electronic structure of realistic transition metal clusters at useful precision. The pathway to industrial relevance requires extending this to the full 54-electron active space of the FeMo-cofactor, which will require either larger fault-tolerant hardware or more aggressive active space partitioning. Nevertheless, the demonstration closes the loop between quantum hardware performance and a concrete industrial chemistry problem, providing Honeywell’s process engineering teams with a credible timeline for when quantum-designed catalysts might enter the development pipeline.