本文讨论氮化酶与 FeMo-co 的基态能量问题，以及这是否真需要量子计算机。背景上，19 世纪可用氮主要来自秘鲁海岸岛屿的鸟粪，直到 Fritz Haber 和 Carl Bosch 于 1909 年实现工业固氮后，实践意义才下降；但科学问题仍在。2011 年 Microsoft 会议后，2017 年又有论文把氮化酶视为量子计算基准，而 Chan 一直认为经典方法足以处理这一难题。
核心计算对象是 FeMo-co 的最低能电子构型。该簇含 7 个铁原子，每个原子有 4 或 5 个未成对电子；电子可能处于超过 78,000 种可行构型之中，因此其基态可视为这些构型的加权叠加。研究者认为经典计算需要逐步排除大量无关构型，而量子计算机理论上可直接表示初始猜测并演化到正确基态；但 Chan 指出，量子方法同样受制于高质量初始猜测这一瓶颈，且经典技术近年快速成熟。
Chan 团队自 2000 年他在 Cambridge 获博士以来持续发展压缩复杂量子态的方法，并将两种技术用于 FeMo-co。其一是从初始猜测出发，逐步调整少量电子行为，并证明更大范围的调整不会带来显著能量变化；其二是把初态分块，仅允许有限信息流通过，再证明只需考察到某个上限。两种方法得到相同的基态能量估计，并与实验观测一致。作者认为这为进一步模拟完整氮化酶反应提供了路线，但从基态到一串中间化学态仍然很难；Whitfield 也强调，量子计算更可能在“随时间演化”的问题上体现优势，而非单一基态数值。
The article examines whether nitrogenase truly requires quantum computers, against a historical backdrop in which usable nitrogen came from Peruvian guano until Haber and Bosch’s 1909 industrial nitrogen fixation reduced the practical pressure. The scientific puzzle remained: how a soil bacterium’s nitrogenase accomplishes what the Haber-Bosch process needs an industrial furnace to do. Microsoft later helped cast the enzyme as a quantum benchmark after a 2011 meeting and a 2017 PNAS paper, but Chan argued from the start that classical methods could still work.
The main target was FeMo-co’s ground-state energy. FeMo-co contains 7 iron atoms, each with 4 or 5 unpaired electrons, and the electrons can occupy more than 78,000 plausible configurations; the ground state is a weighted superposition of them. In principle the Schrödinger equation determines the answer, but in practice both classical and quantum approaches need a strong initial guess. Advocates of quantum computing expected an advantage from representing the state directly and evolving it forward, while Chan countered that quantum methods still face the same initial-guess bottleneck and that classical techniques have been rapidly improving.
Chan’s team, building on methods he had developed since earning his Cambridge PhD in 2000, used two compression strategies. One incrementally changed small numbers of electrons and showed that larger changes did not materially alter the energy; the other split the initial state into pieces and limited information flow between them, proving only a bounded amount needed to be tracked. Both methods produced the same FeMo-co ground-state energy and matched experimental data. The result does not solve the full enzyme-reaction problem, which requires a sequence of intermediate states, but it weakens the claim that quantum hardware is necessary for the hardest chemistry and shifts attention toward time-evolution problems, where quantum computers may still have the clearest edge.