Detection of hidden defects in materials using a single-pixel terahertz sensor

In the field of engineering and materials science, the detection of hidden structures or defects within materials is very important. Traditional terahertz imaging systems, which rely on the unique property of terahertz waves to penetrate visibly opaque materials, have been developed to reveal the internal structure of various materials of interest.

This capability provides unprecedented advantages in many applications for industrial quality control, safety inspection, biomedicine, and defense. However, most existing terahertz imaging systems have limited bandwidth and bulky setups, and they need raster scanning to obtain images of hidden features.

To change this paradigm, researchers at the UCLA Samueli School of Engineering and the California NanoSystems Institute developed a unique terahertz sensor that can rapidly detect hidden defects or objects in a target sample volume using a single-pixel spectroscopic terahertz detector.

Instead of traditional point-by-point scanning and digital imaging-based methods, this sensor inspects the volume of the test sample illuminated by terahertz radiation in an instant, without imaging or digitally processing the sample.

Led by Chancellor’s Professor of Electrical and Computer Engineering Dr. Aydoghan Ozcan and UCLA’s Northrop Grumman Endowed Chair Dr. Mona Jarahi, this sensor serves as an all-optical processor that can search for and classify unexpected sources of generated waves. by diffraction through hidden defects. The paper is published in the journal Communications of nature.

“It’s a change in how we view and use terahertz images and senses as we move away from traditional methods to more efficient, AI-based, all-optical sensing systems,” said Dr. Ozcan, who is also the Association’s deputy director. is California Institute for Nanosystems at UCLA.

This new sensor includes a series of diffraction layers that are automatically optimized using deep learning algorithms. After training, these layers are turned into a physical prototype using additive manufacturing approaches such as 3D printing. This allows the system to perform all-optical processing without the need for raster scanning or cumbersome digital image capture/processing.

“It’s like the sensor has its own built-in intelligence,” Dr. Ozcan said, drawing parallels with their previous AI-engineered optical neural networks. “Our design includes multiple diffraction layers that modify the input terahertz spectrum depending on the presence or absence of hidden structures or defects in the test materials. Think of it as giving our sensor the ability to “sense and respond” based on what it “sees”. “At the speed of light.”

To demonstrate their new concept, the UCLA team created a diffraction terahertz sensor using 3D printing and successfully detected hidden defects in silicon samples. These designs consisted of stacked wafers with one layer containing the defects and the other hiding them. The intelligent system accurately identified the presence of unknown hidden defects in different shapes and positions.

The team believes that their diffraction defect sensor framework can also work at other wavelengths, such as infrared and X-rays. This versatility heralds many applications, from manufacturing quality control to safety inspection and even cultural heritage preservation.

The simplicity, high throughput, and cost-effectiveness of this non-imaging approach promise transformative advances in applications where speed, efficiency, and accuracy are paramount.