Breaking the Speed–Accuracy Barrier in Quantum Chemistry

Compute reaction barriers with unprecedented accuracy at a fraction of today’s computational cost, enabling reliable simulations at scale.

Problem 03OpenUpdated Nov 2026

Quantum chemistryDensity functional theoryComputational chemistry

Written with inputs from

Prof. Stefan GrimmeProfessor, University of Bonn

Predicting Reactions Before the Lab Requires Fast and Accurate Methods

Imagine being able to predict how molecules interact and react before ever stepping into a laboratory. Such a capability could transform how we discover new materials, design molecules, and develop advanced manufacturing processes. Rather than relying on costly and time-consuming trial and error, scientists could explore thousands of chemical reactions computationally, rapidly identifying promising molecules and processes for experimental validation.

Computational methods offer a path toward this vision, but a fundamental challenge remains: balancing accuracy and speed. Widely used methods like density functional theory (DFT) are computationally efficient but often struggle to predict chemical reactions accurately. More accurate approaches, such as coupled-cluster theory, come at a much greater computational cost and remain impractical at scale. This trade-off limits our ability to reliably predict and explore new chemistry at scale.

Quantum chemistry lets us predict how the electrons in molecules move, and that movement decides how a reaction happens: which bonds break, which form, and how fast. Here is a schematic of a trimethylaluminum molecule meeting water, the textbook reaction of atomic layer deposition. Their electron clouds merge, a hydrogen jumps across, and then methane drifts away.

The problem

Develop a more accurate exchange-correlation functional that improves DFT predictions without sacrificing computational efficiency.

Submissions

We evaluate submissions on a hidden benchmark of reference-quality molecular reaction energies, activation barriers, binding energies, and bond dissociation energies representative of atomic layer deposition (ALD) and chemical vapor deposition (CVD) precursor chemistry.

Submissions should include:

Method: A description of the proposed computational approach and its underlying scientific principles.
Implementation: Working code and instructions sufficient to independently run and evaluate the method.
Results: Evidence of predictive accuracy and computational efficiency on the public benchmark (GMTKN55).

Evaluation considers:

Predictive accuracy: Error against the hidden benchmark.
Speed and computational effort: Compute time per molecule.
Reproducibility: Sufficient documentation to verify the reported results.

LeaderboardProvisional

Error in kcal/mol (WTMAD-2, the weighted total mean absolute deviation of Goerigk et al., Phys. Chem. Chem. Phys. 19, 32184 (2017)). All functionals are evaluated using the def2-QZVP basis set on a single CPU core, and separately on a single A100 GPU. Each calculation is repeated three times, and the median runtime is used. Other common functionals, such as r²SCAN and Skala, are being evaluated and will be added to the leaderboard.

Error (kcal/mol)

WTMAD-2 · lower is better

Next: 04 · Tuning Materials with Light

Next problem All problems