MechSE faculty Narayana Aluru and SungWoo Nam recently collaborated on an interdisciplinary research project that developed a curved image sensor array for image processing devices. Their results were published in the journal Nature Communications in November.
Image-processing technology uses light-sensitive material to transform optical input into electronic output. “Physically, what’s happening when you put light on the material is that it excites electrons,” graduate research assistant Amir Taqieddin said. “And the reaction taking place will produce an electrical output.”
Current image-processing technology, such as facial-recognition devices, identifies images by employing multiple lenses and flat-image sensor arrays. These systems go through a several-step process to achieve an identified image. The device developed in the study significantly reduces the number of steps in the process by using a single lens and curved-image sensor array to mimic the visual recognition system occurring in a human eye.
An international team of experimental researchers that included Nam tested different material types and found that a heterostructure containing a polymer (pV3D3) and MoS2 allowed for more efficient image processing. In the image sensor, charge carriers are trapped at the interface between the polymer and MoS2, and the interfacial effect realized short-term plasticity and long-term plasticity in a nanoscale device, mimicking synaptic plasticity of human brains. Aluru and graduate students, including Taqieddin, then ran theoretical simulations of the materials the experimentalists had tested to better explain the physics behind the materials’ behavior.
“Our neuromorphic image sensor array is inspired by the human visual recognition system. Our novel 2D material system realized short-term plasticity and long-term plasticity, just like human memory, so that we could avoid massive data processing and calculation, which you would often see in conventional processors (von Neumann architecture), in our recognition system,” said Juyoung Leem, a recent graduate research assistant in Nam’s group (PhD ME ’20; now a postdoctoral fellow at Stanford University). “We saw interesting phenomena in the experiment and designed the neuromorphic device, and theoretical simulations helped us understand the underlying physics and support our experimental data.”
Taqieddin and the team more deeply investigated materials properties causing the behaviors observed in the experiments by using the National Center for Supercomputing Applications’ (NCSA) Blue Waters as well as the San Diego Supercomputing Center’s (SDCS) Comet to simulate light hitting the MoS2-organic heterostructures.
“This is the first fundamental study to make the device and explain the physics,” Taqieddin said.
“Future integration of this device with real-life technology that uses artificial intelligence will be very interesting to see.”
The project makes for a nice showcase of theoretical and experimental research working together to produce a novel technology. “The interdisciplinary collaboration is critical to modern scientific research,” Leem said, “because we want to understand physics better by seeing them from different angles and eventually to manipulate the system to achieve engineering goals.”
“The theoretical work supports the experimental observation to make a complete picture of the phenomena we’re capturing,” Taqieddin said. “[The collaboration of these] can lead to more optimized devices with better control and efficiency.”
This work also had international collaborators from Seoul National University, South Korea, Professor Dae-Hyeong Kim (MatSE PhD ’09) and Dr. Changsoon Choi, who was a visiting student in Nam’s laboratory when the collaboration was initiated.