Researchers at Cornell University have used high-resolution 3D imaging to identify atomic-scale defects in computer chips for the first time. These small defects can interfere with chip performance and are a major concern for modern electronics.
This new imaging technology was developed through collaboration with Taiwan Semiconductor Manufacturing Company (TSMC) and Advanced Semiconductor Materials (ASM). The discovery could have implications for many areas of technology, as computer chips power devices ranging from smartphones and cars to AI data centers and quantum computers.
The results of this survey were announced on February 23rd. nature communications. PhD student Sheikh Karapetian served as lead author of the study.
“There is no other way to see the atomic structure of these defects, so this will be a very important characterization tool for debugging and fault-finding computer chips, especially during development,” said David Mueller, the Samuel B. Eckert Professor of Engineering at Cornell Duffield Polytechnic Institute, who led the project.
Why are small defects important in semiconductor chips?
Very small structural defects have long been a challenge for the semiconductor industry. As chips become more complex and their components shrink to the scale of individual atoms, even small irregularities can affect device operation.
At the heart of every computer chip is a transistor. This small component acts as a switch that controls the movement of electrical current. Each transistor contains a channel that opens and closes to regulate the flow of electrons.
“Transistors are like little pipes that pass electrons instead of water,” Mueller said. “As you can imagine, if the pipe walls are very rough, it slows down the process. So it’s becoming increasingly important to measure how rough the walls are and which walls are good and which are bad.”
From early transistors to 3D chip structures
Mr. Muller has been studying the physical limits of semiconductor technology for many years. From 1997 to 2003, he worked in the research and development department at Bell Laboratories, where the transistor was invented, studying how small these devices would eventually become.
When transistors first appeared in the mid-20th century, they were placed across a chip in a flat layout that spread outwards, like a suburb stretching across a landmass. Over time, engineers ran out of surface area and began stacking transistors vertically, creating complex three-dimensional structures resembling high-rise apartment buildings.
“The problem is that these 3D structures are smaller than the size of a virus. And these days, it’s much smaller. Rather, on the scale of molecules inside a cell,” Muller said.
Today, a single advanced chip can contain billions of transistors. As sizes continue to shrink, diagnosing performance issues becomes much more difficult.
“Nowadays, transistor channels are only about 15 to 18 atoms wide, which is very, very small and very complex,” Karapetyan says. “Right now, it’s all about where all the atoms are, and it’s very difficult to characterize.”
Advances in electron microscopy
Early in his career at Bell Labs, Mueller collaborated with fellow scientist Glenn Wilk ’90, who is now vice president of technology at ASM. The two researched ways to replace silicon dioxide, the dominant gate material at the time. Silicon dioxide used to leak too much current when devices got very small. Their research helped promote the use of hafnium oxide, which subsequently became a standard material used in computer processors and mobile devices starting in the mid-20s.
“I can tell you that the paper we published on how to characterize these materials using electron microscopy was read very carefully by many in the semiconductor community,” said Mueller, co-director of Cornell’s Kavli Institute for Nanoscale Science and the Cornell Center for Materials Research (CCMR). “When we came back to this project, it was very clear, and microscopy has come a long way. Back then it was like biplanes flying, and now we have jet planes.”
The “jet” Muller refers to is electronic ptychography. This computational imaging technique relies on the Electron Microscopy Pixel Array Detector (EMPAD), a technology co-developed by Muller’s research group. The detector records the detailed pattern created as the electrons pass through the transistor structure.
By comparing how these scattering patterns change from one scan point to another, researchers can reconstruct highly detailed images. The system is so accurate that it produces the highest resolution images ever taken, allowing scientists to see individual atoms with such clarity that the Guinness Book of World Records recognizes this ability.
Discovery of “mouse bite” defect
More than 25 years after their previous collaboration, Müller and Wilk have teamed up again with support from TSMC and its Corporate Analysis Laboratory group. Their goal was to apply EMPAD technology to modern semiconductor devices.
“You can think of this imaging technique as solving a giant puzzle, both in terms of experimental data acquisition and computational reconstruction,” Karapetyan says.
After collecting and reconstructing the image data, the researchers tracked the positions of the atoms within the transistor’s channel. This analysis revealed subtle roughness at the interface of these channels. Karapetyan described these irregular patterns as “rat bites.”
Defects formed during the optimized growth process used to fabricate the structure. The sample device created at the nanoelectronics research center Imec provided an ideal platform to test the imaging technique.
“Manufacturing modern devices requires hundreds or hundreds of chemical etching, deposition, and heating steps, all of which affect the structure in some way,” Karapetian said. “Before, we would look at projection images to understand what was actually going on. Now we can actually see the probe directly after every step and have a better idea of, oh, if we raise the temperature this high, this is what happened.”
Implications for future chips and quantum computing
Being able to directly observe atomic-level defects could impact almost any device that relies on advanced computer chips, including smartphones, laptops, and large data centers. It could also help develop emerging technologies such as quantum computers, where researchers need very precise control over the structure of materials.
“I think this tool allows us to do a lot more scientifically, and we have a lot more engineering control,” Karapetyan said.
Co-authors of the study include Steven Zeltmann, staff scientist at the Accelerated Realization, Analysis, and Discovery Platform for Interfacial Materials (PARADIM), and Ta-Kun Chen and Vincent Hou of TSMC.
This research was funded by TSMC. Microscopy facility support was provided by CCMR and PARADIM, which are funded by the National Science Foundation.
