Cornell researchers have used advanced electron microscopy to identify “mouse bite” defects in 3D transistors for the first time.
By visualising how atomic-scale roughness slows down electrons, this new imaging tool helps companies like TSMC debug and optimise the next generation of high-performance computer chips
Researchers at Cornell University, in collaboration with industry giants TSMC and ASM, have developed a high-resolution 3D imaging technique that can detect atomic-scale defects in computer chips. Published in Nature Communications on February 23, 2026, the study introduces a method to visualise “mouse bite” defects—tiny structural irregularities that can sabotage the performance of modern electronics.
As transistors shrink to the size of individual molecules, this tool provides a critical way for engineers to debug and refine the fabrication of chips for everything from smartphones to quantum computers.
Semiconductors: The challenge of atomic-scale transistors
At the heart of every computer chip is the transistor, a tiny switch that controls the flow of electrical current. In modern high-performance chips, these switches have transitioned from flat, “suburb-like” layouts to vertical, 3D structures known as Gate-All-Around (GAA) transistors. These components are now so small—often only 15 to 18 atoms wide—that even a minor structural flaw can significantly impede the flow of electrons.
Professor David Muller, who led the research, compares a transistor to a water pipe. If the walls of the pipe are rough, the flow of water (or electrons) slows down. Identifying exactly where and why these “rough walls” occur is essential for maintaining the speed and efficiency of next-generation processors used in AI data centres and mobile devices.
Detecting “mouse bites” with electron ptychography
To see these defects, the team used a computational imaging method called electron ptychography. This technique utilises an Electron Microscope Pixel Array Detector (EMPAD), a Guinness World Record-holding sensor co-developed by Muller’s group. By capturing how electrons scatter as they pass through a transistor and using complex algorithms to reconstruct the patterns, the researchers achieved unprecedented 3D clarity at the atomic level.
The imaging revealed “mouse bites”—rough patches at the interfaces where the gate material meets the transistor channel. These defects arise during the hundreds of chemical etching and heating steps required to build a chip. Previously, engineers had to rely on indirect measurements or 2D projections to guess what was happening inside these 3D structures. This new method provides a direct, atomic-resolution map of the damage.
Improving yield in the age of 3D chips
The ability to “see” defects after every step of the manufacturing process allows for much finer engineering control. In prototype 5nm-thick channels, the researchers found that only about 60% of the silicon atoms maintained a perfect “bulk-like” structure, with the rest being affected by strain and roughness.
This discovery is particularly vital for the development of quantum computers and advanced AI chips, where structural precision is non-negotiable. By partnering with TSMC and ASM, the Cornell team has ensured that this “atomic debugger” can be integrated into industrial workflows, helping to ensure that the billions of transistors on a single chip all function as intended.