A recent technical examination has challenged the validity of flagship quantum chemistry results published by IBM in 2025, according to Hacker News Front Page. The audit focuses on the computational accuracy of iron–sulfur clusters, specifically questioning whether the quantum calculations achieved the intended electronic spin states required for reliable chemical modeling.

Researchers previously claimed that quantum computers are now useful for chemistry applications, specifically citing calculations for iron–sulfur clusters. However, the new analysis indicates that the electronic states converged during these simulations demonstrate spin-squared values (⟨S²⟩) between 4.7 and 7.0, contradicting the target singlet state of ⟨S²⟩ = 0. When utilizing IBM’s provided software driver, the subspace size expanded 4.00x, moving from 48,600 to 194,481 determinants, yet the ground energy shift remained under a nanohartree.

Further analysis of IBM’s own archived data, which included 2,457,600 shots on [2Fe-2S] from December 2023 and 3,163,742 outcomes on [4Fe-4S] from April 2024, showed significant energy deviations. The [2Fe-2S] samples sat 248 mHa away from the reference, while the [4Fe-4S] cluster converged to a triplet state with a variance of 3e-6, missing the reference energy by 1,438 mHa.

Comparison of IBM Hardware Data Analysis

Why It Matters

This critique exposes a widening gap between publicized quantum utility and empirical reproducibility. By demonstrating that current quantum algorithms for complex molecular systems may consistently collapse into incorrect spin states, the report highlights that "quantum advantage" in chemistry remains sensitive to hardware-software integration. This audit suggests that until software pipelines can enforce correct physical constraints, quantum chemical simulations may struggle to outperform established classical methods like selected-CI, potentially stalling the adoption of quantum workflows in industrial pharmaceutical and materials research.