A three-dimensional map locates 6,648 carbon atoms in graphene

A carbon atom map now has a third dimension. Researchers working with twisted bilayer graphene located 6,648 individual carbon atoms, reporting a positional precision of 12 picometers, or twelve trillionths of a meter. The collaboration includes ShanghaiTech University, UCLA and MIT. Its preprint appeared in April 2025; the journal paper followed on November 3, 2026.

Graphene consists of a single layer of carbon atoms. Rotating one sheet slightly before stacking it over another creates a structure that need not remain perfectly flat. Atoms can shift sideways while the sheets bend. A face-on image superimposes the two layers, obscuring some of those height differences.

Imagine assigning seats from photographs of two crowded rows. One picture may show many people without revealing exactly who sits behind whom. Views from other directions help resolve the arrangement. Here, an electron beam scans a small region and records scattering patterns. Computation turns the measurements into images and combines the different views into a three-dimensional reconstruction.

Taking more pictures is not a complete solution. Carbon scatters electrons weakly, while a stronger exposure can damage the specimen. The authors describe collecting thirteen views of the same region and improving reconstruction to separate signals from the closely spaced sheets.

Three-dimensional atomic imaging already had a history. A 2023 study used related methods to resolve zirconium and tellurium structures inside carbon nanotubes, averaging repeating units. The advance is therefore not the first three-dimensional picture of any atom. It extends reliable localization to individual carbon atoms in a particular specimen, rather than relying mainly on stronger signals from heavier elements.

The resulting map shows that the graphene sheets do not bend identically everywhere. ShanghaiTech reports that calculations based on the measured structure predict changes in the energies electrons can occupy. This connects a structural measurement to a predicted electronic effect; it is not a demonstration of a faster commercial chip.

For nonspecialists, the broader idea is straightforward. Knowing a material's ingredients does not fully explain its behavior. Their actual arrangement matters too. Replacing an idealized diagram with coordinates measured in a real sample could help researchers understand why apparently similar materials behave differently, then decide which features to change.

First preprint: April 11, 2025, 11:20 Beijing time / April 10, 2025, 11:20 PM EDT. Journal publication: November 3, 2026.