Neural Network Learns to Read Molecules by Their Carbon Signature express

Original: "Graph Neural Network Predictions of Carbon 1s Binding Energies with Near-Experimental Accuracy"

· Adam E. A. Fouda, Joshua Zhou, Rodrigo Ferreira, Patrick Phillips, Valay Agarawal, Bhavnesh Jangid, Jacob J. Wardzala, Rui Ding, Junhong Chen, Nicole Tebaldi, Phay J. Ho, Laura Gagliardi, Linda Young

arXiv:2604.27070 · 2026-04-29 · CC BY 4.0 · 2 min

Chemical Physics Computational Physics

A graph neural network predicts the energy of carbon electrons by viewing a molecule as a network of bonds.

Every carbon atom in a molecule carries a passport — the energy with which its deepest electrons sit in their orbitals. This energy is sensitive to neighbors: add an oxygen nearby or a double bond, and the number in the passport changes. Chemists have long measured such energies with X-rays, but calculating each new molecule takes hours of machine time.

The new model works like a seasoned archivist who, from a single photo of a building, instantly determines the strength of its foundation. It sees a molecule not as a jumble of balls but as a graph — a network where atoms are nodes and bonds are edges. The neural network was trained on 8637 carbon centers in 2116 molecules ranging from 4 to 16 atoms, using data from multiconfigurational density functional theory (a sophisticated quantum calculation method). When tested on 570 real measurements in 113 molecules, some reaching 45 atoms, the average error was 0.33 electronvolts. That's roughly one hundredth of the energy of visible light.

For comparison: previous computational methods at the same level of theory gave an error of 0.27 eV, but only for the training set. The new model maintains accuracy even on unfamiliar large molecules.

A special feature of the model is chemically meaningful node features that reduce dependence on the depth of message passing in the graph. In simpler terms, the neural network doesn't need to 'ask' distant neighbors many times to understand an atom's environment. It's like gauging a person's mood not from their entire chain of conversations but from a few key words.

The model is also robust to distorted geometries: those that arise from thermal vibrations or early stages of a chemical reaction. It was tested on tautomers of avobenzone (45 atoms)—molecules that shift a hydrogen atom and are linked to UV protection. The neural network instantly distinguished the forms and produced energies consistent with experiment. This paves the way for rapid screening of drugs and materials.

Graph neural networks don't just memorize examples: they capture the rules of chemical environment. So a model trained on small molecules confidently handles large ones, where bond combinations are many times more numerous.

And here's the curious thing: carbon's passport changes by several electronvolts depending on neighbors — enough to distinguish graphite from diamond by X-ray spectrum alone.

Whole graph →

The energy of carbon's core electrons changes by several electronvolts depending on neighbors — enough to distinguish graphite from diamond by X-ray spectrum alone.

Scientists

Paul Dirac Fred Hoyle Margaret Burbidge William Lawrence Bragg Michael Berry

Concepts

graph neural network Graphs neural network predictive model molecular property prediction training machine learning materials generative model parameter training data prediction accuracy weight graphene gpu generalization molecular dynamics acceleration computational reaction analysis Generative Adversarial Networks transfer learning neuron

I am the author

Neural Network Learns to Read Molecules by Their Carbon Signature express