A team of researchers from the UCLA Samueli School of Engineering has demonstrated a new approach to an old problem: measuring the spectra of light, also known as spectroscopy.
Taking advantage of scalable, cost-effective nano-fabrication techniques, as well as AI-driven algorithms, they built and tested a system that is more compact than traditional spectrometers, while also providing additional design advantages.
Spectroscopy is a central tool for many applications in life sciences, medicine, astrophysics and other fields. Traditional spectrometers divide light into its component colors so that the intensity of each can be measured.
This leads to many constraints and design tradeoffs: finer spectral resolution (with tight spacing between detectable colors or wavelengths) may require the use of more expensive hardware, increasing the physical footprint of the device and the potential Sacrificing signal strength.
This can be problematic for applications requiring high sensitivity, high spectral resolution, and compact system design.
It also presents challenges ahead for hyperspectral imaging, including capturing a spectrum for each pixel in an image, a technique commonly used to assess crop health or the spread of greenhouse gases among other uses Used for remote sensory functions such as environmental monitoring.
The approach of researchers at UCLA, driven by AI, re-applies the spectroscopy problem from the ground up. Instead of relying on splitting the light into a rainbow of component wavelengths, a nanostructured chip spectrally decomposes light using hundreds of unique spectral filters in parallel.
This chip uses plasmonic structures as spectral encoders, composed of 252 tiles, each with a distinctive nanoscale pattern that transmits a different spectrum of light.
In other words, the unknown spectrum of light measured is “encoded” in the transmission of each of these plasmonic tiles. This nanostructured encoder is fabricated through an imprint lithography process that can significantly reduce production costs and enable large production volumes to scale.
The light transmitted by the spectral encoder chip is captured using a standard, inexpensive image sensor, used regularly in our mobile phone cameras, to produce an image that is then re-encoded with the light from the image The unknown spectrum is fed into the neural network tasked with regrouping. Information.
This spectral reconstruction neural network was shown to give very fast accurate results compared to other computational spectroscopy approaches, which result in less than one-third of a millisecond. This new AI-powered spectrometer framework displays a path around the specific tradeoff between device cost, size, resolution, and signal strength.
“We are demonstrating a proof not only on the concept device here,” said Aydogan Ozcan, an associate professor of electrical and computer engineering at Nano Systems Systems (CNSI), California, whose group conducted the research.
“We are introducing an entirely new framework for chip-scale spectrometer design. Neural networks, training spectra, nano-encoder geometrics and materials; each of these components can be adapted for different applications or specific tasks Is, enabling compact, cost-effective. Effective spectrometers that produce high quality measurements for a given sample type or spectral regime. “
This AI-enabled on-chip spectrometer framework can explore various applications ranging from environmental monitoring of gases and toxins to medical diagnostics where spectral information is required to distinguish the presence of different biomarkers.
Researchers also noted that plasmonic tiles can be truncated downward (like a camera pixel grid) to perform hyperspectral imaging, which may be important, for example, autonomous remote sensing where compact, lightweight form factors Are required.
Other authors of work include electrical and computer engineering researchers Calvin Brown, Artem Goncharov, Zachary S. Ballard and Eunze Qiu, graduate students were Mason Fordham and Ashley Clemens, and Yair Revenson, assistant professor of electrical and computer engineering.