In a new study published in Nature, researchers at ETH Zurich and Empa created a novel image sensor built of perovskite. This semiconductor material allows for greater color reproduction and fewer image artifacts with less light. Perovskite sensors are also ideal for machine vision.
Thin-film technology: One of the two perovskite-based sensor prototypes the researchers have used to demonstrate that the technology can be successfully miniaturized. Image Credit: Empa / ETH Zürich
All smartphones and digital cameras have image sensors. They detect colors similarly to the human eye. Individual cone cells in the retina identify red, green, and blue (RGB). Individual pixels in image sensors absorb and transform wavelengths into electrical signals.
The great majority of image sensors are composed of silicon. This semiconductor material often absorbs light over the whole visible spectrum. To make RGB image sensors, the entering light must be filtered. Pixels for red have filters that block (and waste) green, blue, and so forth. Each silicon image sensor pixel consequently gets about one-third of the available light.
Maksym Kovalenko and his team at ETH Zurich and Empa have suggested a revolutionary technique that allows them to use each photon of light for color detection. They have been investigating perovskite-based image sensors for over a decade.
Stacked Pixels
Their revolutionary image sensor is based on lead halide perovskite. This crystalline compound is also a semiconductor. In contrast to silicon, it is especially easy to produce, and its physical qualities vary according to its precise chemical makeup. This is precisely what the researchers are utilizing in developing perovskite image sensors.
Perovskite will absorb red light if it includes slightly more iodine ions. The researchers added additional bromine to the green solution and more chlorine to the blue solution without using filters. The perovskite pixel layers are transparent to other wavelengths, allowing them to flow through. This implies that the red, green, and blue pixels of the image sensor can be stacked on top of each other, as opposed to silicon image sensors, which have the pixels placed side by side.
This design allows perovskite-based image sensors to catch three times as much light as traditional image sensors with the same surface area while simultaneously offering three times greater spatial resolution. Researchers from Kovalenko’s team demonstrated this a few years ago, first with individual larger pixels formed of millimeter-sized single crystals.
They have created two completely working thin-film perovskite image sensors for the first time.
We are developing the technology further from a rough proof of principle to a dimension where it could actually be used.
Maksym Kovalenko, ETH Zurich
A typical path of development for electronic components: “The first transistor consisted of a large piece of germanium with a couple of connections. Today, 60 years later, transistors measure just a few nanometers,” stated Kovalenko.
Perovskite image sensors are currently in their early phases of development. However, using the two prototypes, the researchers demonstrated that the technology can be shrunk. The sensors, which were manufactured using typical industrial thin-film methods, have at least surpassed their intended vertical dimensions.
Of course, there is always potential for optimization.
Sergii Yakunin, Study Co-Author and Lecturer ETH Zurich
The researchers rigorously tested two prototypes, each featuring different readout technologies, proving perovskite's advantages. Their results show these new sensors are more light-sensitive, offer more precise color reproduction, and can achieve significantly higher resolution than traditional silicon technology. Because each pixel captures all available light, the sensors also eliminate common digital photography artifacts like demosaicing and the moiré effect.
Machine Vision for Medicine and the Environment
Perovskite image sensors, however, have applications beyond consumer digital cameras. The material's qualities make it particularly ideal for use in machine vision. The human eye determines the emphasis on red, green, and blue. These image sensors operate in RGB format since our eyes view in RGB mode.
However, while addressing certain jobs, it is best to define additional appropriate wavelength ranges that the computer image sensor should scan. There are often more than three, known as hyperspectral imaging.
Perovskite sensors provide a significant advantage in hyperspectral imaging. Researchers can carefully regulate the wavelength ranges absorbed by each layer.
“With perovskite, we can define a larger number of color channels that are clearly separated from each other,” added Yakunin.
Silicon’s wide absorption spectrum necessitates several filters and advanced computer algorithms.
“This is very impractical even with a relatively small number of colors,” summed up Kovalenko.
Perovskite-based hyperspectral image sensors might be employed in medical analysis as well as automated monitoring of agriculture and the environment.
In the following phase, the researchers plan to shrink the size and increase the number of pixels in their perovskite image sensors. Their two prototypes have pixel sizes ranging from 0.5 to 1 millimeter. Pixels in commercial image sensors are in the micrometer range.
“It should be possible to make even smaller pixels from perovskite than from silicon,” said Yakunin.
The electronic connections and processing procedures must be adapted to the new technology.
“Today's readout electronics are optimized for silicon. But perovskite is a different semiconductor, with different material properties,” added Kovalenko.
However, researchers are certain that these problems can be solved.
Journal Reference:
Wang, H., et al. (2025) A Review of Perovskite-Based Photodetectors and Their Applications. Nature. doi.org/10.3390/nano12244390