Researchers at Cornell University have produced the first direct images of atomic-scale defects inside working computer chips, mapping the exact positions of atoms in transistor structures just 15 to 18 atoms wide. The imaging method, built on advanced electron microscopy, exposes small imperfections the team nicknamed "mouse bites" — flaws that form during manufacturing and disrupt how electrons flow through a chip's channels.

The channel is the narrow path a current travels inside a transistor. At widths of 15 to 18 atoms, a single displaced atom changes how easily electrons pass through. Until now, no imaging technique could show those individual flaws in place inside a finished device structure. The Cornell team's microscopy approach resolves the atoms directly and records their positions.

What are "mouse bites"?

The defects take their name from their appearance in the images: small chunks of ordered atomic structure that appear to have been gnawed away. They arise during the complex manufacturing process, in which layers of material are deposited, etched and heated to build transistor structures atom by atom.

A mouse bite is not a design error. It is a manufacturing artifact — a place where the atoms did not settle exactly where the process intended. Inside a channel only 15 to 18 atoms across, such an artifact occupies a significant share of the current path. The researchers found these imperfections disturb electron flow through the channel, degrading the device's electrical behavior.

The practical point is simple: the industry now has a way to see, atom by atom, where manufacturing goes wrong.

Why atomic-scale imaging matters at advanced nodes

Chipmakers have pushed transistor dimensions toward the atomic limit. At a channel width of 15 to 18 atoms, there is very little margin between a good device and a faulty one. Process engineers have historically inferred defect types from electrical measurements and yield data, because no instrument could look inside the structure at atomic resolution.

The Cornell technique changes that. By mapping the exact positions of atoms inside transistor structures, it gives engineers a direct link between a physical flaw and its electrical effect. That link matters for yields: if a specific defect type correlates with current disruption, fabs can trace it to a specific step in the process and correct that step.

The method also turns defect inspection from statistics into observation. Instead of estimating how many imperfections a process produces, engineers can count the individual mouse bites and see precisely where they sit in the channel.

What comes next

The Cornell team has demonstrated the technique on transistor structures and identified the defect type and its electrical consequence. The work opens a path for chipmakers to inspect atomic-scale defects directly as geometries continue to shrink — and as channels approach the width of a dozen atoms, the ability to see each flaw in place is likely to become a standard tool rather than a laboratory achievement.