Michael Berger’s August 27, 2025 Nanowerk Spotlight article takes us into the field of ‘brainlike’ vision (neuromorphic vision), Note: Links have been removed,
Human eyes adjust effortlessly to darkness. A few minutes after stepping into a dim room or driving through a tunnel at night, vision sharpens, objects come into focus, and movement is easier to track. This ability comes from a combination of immediate sensitivity and short-term memory built into the retina itself. Rod cells detect weak light. Neurons store patterns. Together, they let the brain build a coherent picture even in near darkness.
Artificial vision systems do none of this. Cameras capture light, but they rely on separate memory and processing units to interpret it. In dim conditions, this pipeline breaks down. Signals get noisy. Processing lags. And without a way to remember what was just seen, tracking motion or recognizing shapes becomes unreliable. The result is a major weakness in technologies that need to see in the dark, from autonomous vehicles to low-power robotics and surveillance.
Solving this problem requires more than just better sensors. It requires hardware that behaves more like a retina, adapting to weak light while storing and processing visual information locally. The field of retinomorphic vision aims to build such systems by mimicking biological principles in electronic devices. But one of the biggest technical barriers has remained unresolved.
Even the most light-sensitive materials, such as quantum dots, struggle to generate usable signals in low-light environments because the charges they produce remain locked together and don’t travel. Without charge separation, there’s no current to store, no memory to form, and no adaptation to achieve.
A study published in Advanced Materials (“Ferroelectric Quantum Dots for Retinomorphic In‐Sensor Computing”) presents a solution to this problem. The researchers developed ferroelectric quantum dots that combine strong light absorption with built-in electric fields. These fields help separate photo-generated charges, enabling a new kind of device that can detect, adapt to, and remember visual information in real time and low-light conditions.
My hat’s off to whomever produced this illustration as it is unusually accessible to someone like me, i.e., not particularly skilled at decrypting scientific illustrations,

Back to Berger’s August 27, 2025 article,
The team started with cadmium selenide quantum dots, a well-studied material known for its efficient light absorption. They wrapped the dots in a shell of zinc cadmium sulfide and replaced the usual surface ligands with specially designed polymer chains. These polymers, made from polyvinylidene fluoride, are ferroelectric. That means they contain internal dipoles that reorient under an applied voltage, generating a small electric field. This field counteracts the force that normally holds electrons and holes together inside a quantum dot, making it easier for the charges to separate.
The researchers synthesized the ferroelectric polymer using a controlled polymerization method and modified it with sulfur-based groups that strongly attach to the surface of the dots. This chemical structure not only introduces the desired electrical behavior but also helps prevent the dots from clumping together in a film, ensuring even distribution for device fabrication. Tests confirmed that the modified quantum dots preserved their optical properties and showed stable emission under illumination. Measurements also confirmed ferroelectric switching, with clear polarization loops and reversible shifts in surface potential.
To turn the material into a functional device, the team built a synaptic phototransistor. In this structure, the quantum dot film acts as a floating gate layer between insulators and a semiconducting channel. When light strikes the device, it creates charges in the dots. The direction of the ferroelectric polarization determines how easily these charges tunnel through to the channel. Once there, the charges modulate the current in the device, which continues to flow even after the light is turned off. This persistent current functions as a kind of memory, storing information about the visual input.
The researchers showed that applying different voltages changed the behavior of the device. A positive voltage aligned the internal dipoles to help charges flow, boosting memory retention. A negative voltage did the opposite, reducing the stored current. This switchable behavior mimics synaptic plasticity in biological systems, where the strength of a signal pathway changes based on input. The device demonstrated both short-term and long-term memory effects, depending on the strength and duration of the light pulses. In tests, the photocurrent persisted for over seven hours under ambient conditions without encapsulation, a result attributed to the chemical stability of the fluorinated polymer ligands.
…
The researchers also tested a form of sensory adaptation similar to how the human eye adjusts to low light over time. They projected a target pattern onto a 3 by 3 sensor array under dim conditions, along with a distracting background signal. Without polarization, the pattern was hard to detect. But when the ferroelectric function was activated, the pattern gradually became clearer with each pulse of light. This behavior mimicked the gradual increase in contrast that occurs in the retina during scotopic adaptation. Even after the light was removed, the stored signal remained visible, demonstrating that the sensor could both adapt and remember.
…
By combining material innovation with device design, the researchers have developed a sensor that moves beyond passive detection. It responds to light like a photoreceptor, stores information like a memory cell, and adapts its behavior like a synapse. The system could provide a foundation for vision hardware that processes information where it is captured, enabling fast, energy-efficient operation in conditions that currently challenge conventional cameras and sensors.
I wonder if this work will find its way into self-driving cars. Having just read a book about some of the problems with Tesla’s self-driving cars (2025 book, “The Tesla Files: The Definitive Exposé of the World’s Most Powerful Businessman and the Rise and Fall of his Empire” by Sönke Iwersen, Michael Verfürden), this would (if feasible) seem like an improvement over what they have currently.
Getting back to the research, here’s a link to and a citation for the paper,
Ferroelectric Quantum Dots for Retinomorphic In-Sensor Computing by Tingyu Long, Huanyu Zhou, Jaewan Ko, Hongwei Tan, Jaemin Lim, Yanfei Zhao, Daehan Kang, Eojin Yoon, Gyeong-Tak Go, Somin Kim, Seung-Woo Lee, Chan-Yul Park, Hyojun Choi, Hyeran Kim, Hyung Joong Yun, Sung Hyuk Park, Kwan Sik Park, Jeong Woo Park, Mungeun Kim, Yong Soo Cho, Ho Won Jang, Wenqiang Yang, Min Hyuk Park, Wan Ki Bae, Sebastiaan van Dijken, Joona Bang, Tae-Woo Lee. Advanced Materials DOI: https://doi.org/10.1002/adma.202504117 First published: 23 August 2025
This paper is open access.







