College of Science breakthrough reveals atomic defects key to next-generation memory

Physicist Matt Graham and his team have pulled off something no other lab in the world can do: detect all the different types of atomic-scale defects that make transistors made from amorphous semiconductors tick.

Using a laser-based technique they developed, one so sensitive it can spot a single missing oxygen atom, the group has opened a new window into how next-generation memory materials behave. The work, funded by Samsung Electronics, centers on transistors used in DRAM, the memory chips found in most computers.

The breakthrough comes at a critical moment for the semiconductor industry. As artificial intelligence and other data-intensive technologies place increasing demands on computer memory, researchers are searching for alternatives to silicon that are cheaper, more energy efficient and capable of powering future chip designs. But until now, engineers have no reliable way to measure the relative concentrations of the different atomic defects that determine how transistors made from these materials perform.

Graham and his team published their results in Advanced Functional Materials, a top-tier, high-impact scientific journal.

Their approach gives the semiconductor industry a powerful new tool for improving transistors, the tiny switches that control electrical current in a chip, by showing engineers why some perform better than others. The main appeal of their method is that it is conducted on live transistors used for computer memory and display panels.

What is an amorphous semiconductor?

Amorphous semiconductors are a newer class of semiconductors that don’t have any crystalline order, a completely random arrangement of atoms, rather than a repeating structure. Silicon, by contrast, has a diamond cubic crystal structure, meaning each atom is covalently bonded to its four nearest neighbors.

IGZO, the specific semiconductor material Graham works on, is made of indium gallium zinc oxide. The material was originally prized for a different reason entirely: it's transparent, allowing light to pass through it. That’s why IGZO has been commonplace in smartphone and TV displays for roughly two decades. Only recently has it been considered as a potential material for computing.

Two men work in a laboratory with glasses and lazers.

Matt Graham's research group is called the Micro-Femto Energetics Lab. One of their primary goals is to better understand electron flow and relaxation in emerging materials, transistors, solar cells and other devices.

“There is this problem right now of the memory bottleneck effect,” said Graham. “Where industry is able to scale the CPU computing power, but the memory effect is not scaling. As a result, memory has become the limiting factor, and with the expansion of artificial intelligence, we need new materials beyond silicon to tackle the problem.”

Compared to silicon, IGZO is cheaper to produce and can be more energy efficient in certain circumstances, said Måns Mattsson, a graduate researcher and lead author of the study. It does have a tradeoff though. Because it isn’t a pure crystal, IGZO transistors tend to be less reliable than silicon ones, without the same consistency in switching on and off.

One of the biggest benefits is the ability to create layers. Most computer chips are currently two-dimensional because silicon requires extremely high temperatures to process, which would melt any layers underneath. IGZO, on the other hand, uses a much lower processing temperature, allowing for the creation of three-dimensional layers. This stacking allows developers to increase computing speeds.

Silicon also requires doping, or the intentional addition of tiny amounts of impure atoms to a pure material to control its ability to conduct electricity. IGZO doesn’t require doping because it naturally contains defects that serve the same purpose.

Defects aren’t all bad, as the term might imply. Some are responsible for providing the free electrical charges needed for current to flow, while others can trap charges and reduce performance.

Mattsson and Graham’s technique works by shining laser light on the material to free electrons trapped inside these defects, allowing them to move and produce a measurable current, effectively making the invisible defect visible. They can now relay that insight back to chip engineers, helping identify which defects cause performance losses in advanced transistors and in the future, predict transistor performance from measured defect states, accelerating development of new chip technologies.

The next step for the team is to develop a machine-learning approach that would let other labs extract similar defect information from just a transistor’s electrical performance, without needing the specialized laser setup Graham’s lab currently relies on.

“There is this problem right now of the memory bottleneck effect."

The measurement technique is patent pending. The project has been funded by Samsung Electronics for four years through the company’s Global Research Outreach Award and a College of Science Research and Innovation Seed Program Industry Partnership Award. Oregon State University co-authors on the paper include graduate student Jared Parker, Professor Paul Cheong, and lab alumni Chris Malmberg and Kyle Vogt.

Beyond IGZO, the team hopes to show the method works on everything from crystalline transistors to solar cells. Researchers who have read the group’s paper are already sending samples. The aim, Graham said, is to give engineers what they’ve been missing: not just whether a transistor works, but why.

What is an amorphous semiconductor?

How can AI accelerate discovery while keeping scientists at the center? Scott Clark explores the question Oct. 5

Yunker Lecture 2026: 'Three Frontiers in the Search for Signs of Life Beyond Earth’

College of Science AI Showcase: From classrooms to cosmos

A physics student's journey from the lab to the Great Pyramids