Ferroelectric quantum dots enable phototransistors that adapt to low light and store visual memory, supporting motion recognition and in-sensor learning in neuromorphic systems.
(Nanowerk Spotlight) 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.
Biological and artificial vision systems with adaptive and dynamic sensing capabilities. a) The human visual system receives external stimuli through photoreceptors, including cones and rods, and continuously adapts by updating synaptic connections to form visual dynamic memory. Highly photosensitive rod cells are essential for weak light detection and scotopic adaptation. b) A floating gate configuration is used to implement the low-light in-sensor visual adaptation and dynamic memory (LADM) system. The synthesis of FE-QDs involves a ligand exchange process, where the long-chain PVDF-SH ligands replace the short-chain OA ligands. The FE-QD’s advantage is its ability to realize LADM functions within a single device, facilitated by the ferroelectric PVDF-SH ligand and the photoelectric response of QDs. c) Schematic diagram of car motion recognition under low-light conditions, using conventional sensing, adaptive sensing, and adaptive & dynamic sensing, respectively. (Image: Reprinted from DOI:10.1002/adma.202504117, CC BY) (click on image to enlarge)
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 team also evaluated the photoresponse of the device across multiple wavelengths, including ultraviolet, blue, and green light. They observed a strong response under ultraviolet illumination, with distinct changes in current and clear separation between different voltage-induced states. Compared to a control device using unmodified quantum dots, the ferroelectric version showed higher sensitivity, longer retention, and more stable behavior under repeated cycling.
To explore practical applications, the researchers used their device to simulate learning in a neural network. They configured a three-layer system with 784 input nodes, 100 hidden nodes, and 10 outputs to classify images from the MNIST database of handwritten digits. By using light pulses and polarization states to adjust the conductance of the device, they were able to encode synaptic weights directly into the material. The network reached a classification accuracy of 92.2 percent, close to the 98.2 percent achieved using ideal digital weights. The conductance changes were stable across many programming cycles, showing low variation and good reproducibility.
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.
To extend the system to motion detection, the team simulated four types of car movement: approaching, retreating, and lateral motion. They compared three types of sensors. A conventional one failed to detect motion under low light. A sensor with adaptive behavior could detect the car but not track its direction. Only the full adaptive and dynamic system using ferroelectric quantum dots was able to detect both presence and direction of motion with perfect accuracy. The data fed into a convolutional neural network achieved 100 percent classification accuracy, demonstrating the potential of the system for real-time recognition in poor lighting.
This work integrates light sensing, memory storage, and adaptation into a single material platform. It solves the charge confinement problem in quantum dots using ferroelectric polymers that also stabilize the film and enable electrical control. The resulting device can be tuned, programmed, and reconfigured using light and voltage, with long-term stability and low power requirements. It operates without a separate processor or memory unit, and its stability in air makes it suitable for practical use.
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.
If this article was useful, support our independent nanotechnology reporting with any amount.
Your contribution funds the next explainer and keeps Nanowerk open for everyone.
For authors and communications departmentsclick to open
Lay summary
Prefilled posts
Nanowerk Newsletter
Get our Nanotechnology Spotlight updates to your inbox!
Thank you!
You have successfully joined our subscriber list.
Become a Spotlight guest author! Join our large and growing group of guest contributors. Have you just published a scientific paper or have other exciting developments to share with the nanotechnology community? Here is how to publish on nanowerk.com.
