Better Images For Humans And Computers

Image sensors are built into every smartphone and every digital camera. They distinguish colors in a similar way to the human eye. In our retinas, individual cone cells recognize red, green and blue (RGB). In image sensors, individual pixels absorb the corresponding wavelengths and convert them into electrical signals.

The vast majority of image sensors are made of silicon. This semiconductor material normally absorbs light over the entire visible spectrum. In order to manufacture it into RGB image sensors, the incoming light must be filtered. Pixels for red contain filters that block (and waste) green and blue, and so on. Each pixel in a silicon image sensor thus only receives around a third of the available light.

Maksym Kovalenko and his team associated with both ETH Zurich and Empa have proposed a novel solution, which allows them to utilize every photon of light for color recognition. For nearly a decade, they have been researching perovskite-based image sensors. In a new study published in the renowned journal Nature, they show: The new technology works.

Stacked pixels

The basis for their innovative image sensor is lead halide perovskite. This crystalline material is also a semiconductor. In contrast to silicon, however, it is particularly easy to process - and its physical properties vary with its exact chemical composition. This is precisely what the researchers are taking advantage of in the manufacture of perovskite image sensors.

If the perovskite contains slightly more iodine ions, it absorbs red light. For green, the researchers add more bromine, for blue more chlorine - without any need for filters. The perovskite pixel layers remain transparent for the other wavelengths, allowing them to pass through. This means that the pixels for red, green and blue can be stacked on top of each other in the image sensor, unlike with silicon image sensors, where the pixels are arranged side-by-side.

Thanks to this arrangement, perovskite-based image sensors can, in theory, capture three times as much light as conventional image sensors of the same surface area while also providing three times higher spatial resolution. Researchers from Kovalenko's team were able to demonstrate this a few years ago, initially with individual oversized pixels made of millimeter-large single crystals.

Now, for the first time, they have built two fully functional thin-film perovskite image sensors. "We are developing the technology further from a rough proof of principle to a dimension where it could actually be used," says Kovalenko. A normal course 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."

Perovskite image sensors are still in the early stages of development. With the two prototypes, however, the researchers were able to show that the technology can be miniaturized. Manufactured using thin-film processes common in industry, the sensors have reached their target size in the vertical dimension at least. "Of course, there is always potential for optimization," notes co-author Sergii Yakunin from Kovalenko's team.

In numerous experiments, the researchers put the two prototypes, which differ in their readout technology, through their paces. Their results prove the advantages of perovskite: The sensors are more sensitive to light, more precise in color reproduction and can offer a significantly higher resolution than conventional silicon technology. The fact that each pixel captures all the light also eliminates some of the artifacts of digital photography, such as demosaicing and the moiré effect.

Machine vision for medicine and the environment

However, consumer digital cameras are not the only area of application for perovskite image sensors. Due to the material's properties, they are also particularly suitable for use in machine vision. The focus on red, green and blue is dictated by the human eye: These image sensors work in RGB format because our eyes see in RGB mode. However, when solving specific tasks, it is advisable to specify other optimal wavelength ranges that the computer image sensor should read. Often there are more than three - so-called hyperspectral imaging.

Perovskite sensors have a decisive advantage in hyperspectral imaging. Researchers can precisely control the wavelength range they absorb by each layer. "With perovskite, we can define a larger number of colour channels that are clearly separated from each other," says Yakunin. Silicon, with its broad absorption spectrum, requires numerous filters and complex computer algorithms. "This is very impractical even with a relatively small number of colours," Kovalenko sums up. Hyperspectral image sensors based on perovskite could be used in medical analysis or in automated monitoring of agriculture and the environment, for example.

In the next step, the researchers want to further reduce the size and increase the number of pixels in their perovskite image sensors. Their two prototypes have pixel sizes between 0.5 and 1 millimeters. Pixels in commercial image sensors fall in the micrometer range (1 micrometre is 0.001 millimetre). "It should be possible to make even smaller pixels from perovskite than from silicon," says Yakunin. The electronic connections and processing techniques need to be adapted for the new technology. "Today's readout electronics are optimized for silicon. But perovskite is a different semiconductor, with different material properties," says Kovalenko. However, the researchers are convinced that these challenges can be overcome.