Ultrasmall Silicon LED Enables High-Resolution Integrated Imaging

Facebook X LinkedIn Email
SINGAPORE, May 22, 2023 — Researchers at the Singapore-MIT Alliance for Research and Technology (SMART) developed a CMOS-based silicon LED that, though smaller than a wavelength, emits light at an intensity comparable to much larger silicon LEDs. In a separate study, the researchers used their LED as the illumination source in an in-line, lensless, centimeter-scale, all-silicon, holographic microscope.

They also developed an untrained deep neural network to perform holographic image reconstruction. They then used the deep neural networking algorithm to reconstruct objects measured by the microscope, including plant seeds and tissue samples.

By embedding a physics model within the neural network algorithm, the researchers eliminated the need for training data, and the untrained deep neural network incorporated total variation regularization for increased contrast and accounted for the wide spectral bandwidth of the image source. Further, it allows the use of light sources — like the LED — without prior knowledge of the source spectrum or beam profile.

The researchers envision the CMOS micro-LEDs being used with the machine learning framework for computational imaging applications such as live-cell tracking or spectroscopic imaging of biological materials. Objects could be examined by a compact microscope, without the need for bulky, additional optics, they said.

Though it has been known that a small light source with high intensity like the developed silicon LED is desirable for integrated photonics applications, developing an emitter that is compatible with integrated circuit technology has so far been a challenge.
(a): Photograph of a fully fabricated 300-mm wafer. (b): Close-up of a chip die. (c): Infrared micrograph with the LED turned on. (d): Holographic microscope setup. (e): Close-up of a reconstructed holographic image compared with the (f): ground truth. Courtesy of the Singapore-MIT Alliance for Research and Technology (SMART).
(Left): photograph of a fully fabricated 300-mm wafer. (Top center): close-up of a chip die. (Top right): infrared micrograph with the LED turned on. (Bottom left): holographic microscope setup. (Bottom center): Close-up of a reconstructed holographic image compared with the ground truth (bottom right). Courtesy of the Singapore-MIT Alliance for Research and Technology (SMART).
The emission spectrum of the newly developed LED is centered around 1100 nm. Its emission area is smaller than 0.14 μm2. At room temperature, the LED has a high spatial intensity of greater than 50 mW/cm2. Due to subwavelength confinement, emissions exhibit a high degree of spatial coherence. The researchers demonstrated this property when they integrated the LED into the holographic microscope. The microscope uses only one emitter to simultaneously illuminate about 9.5 million pixels of a CMOS imager.

The small emission area of the silicon LED ensures that the microscope has high spatial coherence without the need for a pinhole and that it results in a large numerical aperture setup, circumventing the limits in source-to-sample distance found in conventional lensless holography devices.

Next, the scene is reconstructed by the untrained deep neural network, which simultaneously performs spectral recovery by learning from the given single experimental diffraction intensity. Blind source spectrum recovery from a single diffracted intensity pattern is a departure from previous supervised learning techniques, the team said.

The research could also pave the way for the development of powerful on-chip emitters smaller than 1 μm — a long-standing challenge. The light in most photonic chips originates from off-chip sources, which lowers energy efficiency and limits scalability. The work of the SMART team demonstrates the feasibility of next-generation, on-chip imaging systems that could, for example, enable cameras in mobile devices to be converted into high-resolution microscopes by modifying the silicon chip and the software of the device.

In-line holography microscopes have already been used for particle tracking, environmental monitoring, biological sample imaging, and metrology. Further applications could include creating an array of silicon micro-LEDs in CMOS to generate programmable coherent illumination. The technology could also be used to advance the miniaturization of diagnostics for indoor farming and sustainable agriculture.
Illustration of the process of image reconstruction using the LED holographic microscope and neural network. Courtesy of Singapore-MIT Alliance for Research and Technology (SMART).
Illustration of the process of image reconstruction using the LED holographic microscope and neural network. Courtesy of Singapore-MIT Alliance for Research and Technology (SMART).
Iksung Kang, a research assistant at MIT at the time of the research, said that the developed LED could be combined into an array for higher levels of illumination needed for larger-scale applications. “In addition, due to the low cost and scalability of microelectronics CMOS processes, this can be done without increasing the system’s complexity, cost, or form factor,” Kang said.

Kang also said that the CMOS-based LED could be used to convert a mobile phone camera into a holographic microscope. “Furthermore, control electronics and even the imager could be integrated into the same chip by exploiting the available electronics in the process, thus creating an ‘all-in-one’ micro-LED that could be transformative for the field,” Kang said.

Professor Rajeev Ram said that the LED has a range of possible applications beyond lensless holography. “Because its wavelength is within the minimum absorption window of biological tissues, together with its high intensity and nanoscale emission area, our LED could be ideal for bio-imaging and biosensing applications, including near-field microscopy and implantable CMOS devices,” he said. “Also, it is possible to integrate this LED with on-chip photodetectors, and it could then find further applications in on-chip communication, NIR proximity sensing, and on-wafer testing of photonics.”

The research on the subwavelength Si LED integrated in a CMOS platform was published in Nature Communications ( The research on CMOS micro-LED holography with an untrained deep neural network was published in Optica (

Published: May 2023
integrated photonics
Integrated photonics is a field of study and technology that involves the integration of optical components, such as lasers, modulators, detectors, and waveguides, on a single chip or substrate. The goal of integrated photonics is to miniaturize and consolidate optical elements in a manner similar to the integration of electronic components on a microchip in traditional integrated circuits. Key aspects of integrated photonics include: Miniaturization: Integrated photonics aims to...
Micro-LED (micro-light-emitting diode) refers to a technology that involves the use of very small light-emitting diodes to create displays and lighting systems. These LEDs are miniature versions of traditional LEDs, typically on the order of micrometers in size. Micro-LED displays offer several advantages over other display technologies, including improved brightness, energy efficiency, and the potential for high resolution. Here are key characteristics and features of micro-LED technology: ...
holographic microscopy
The holographic recording of a microscopic specimen whereby magnification is achieved by alteration in a wavelength or radius of curvature between recording and reconstruction of the wavefront. With this technique the quality and field of the image can surpass those produced by microscopic methods. A microscopy technique that utilizes optical diffraction tomography (ODT), which enables users to quantitatively and noninvasively investigate the intrinsic properties of cells. ODT reconstructs...
integrated photonic devicesintegrated photonicsdevicesBiophotonicsmicro-LEDmicro-LEDsLEDsLight SourcesMaterialsCMOSholographic microscopyMicroscopyAsia PacificResearch & TechnologyeducationImagingSingapore-MIT Alliance for Research and Technology AIneural networkscomputational imaging

We use cookies to improve user experience and analyze our website traffic as stated in our Privacy Policy. By using this website, you agree to the use of cookies unless you have disabled them.