A research group has developed a new near-infrared sensor that is easy to make, comparable in size to sensors in smartphones, and ready for immediate use in industrial process monitoring and agriculture.

FREMONT, CA: A research group at TU/e has developed a novel near-infrared sensor that is simple to produce, similar in size to smartphone sensors, and suitable for application in industrial process monitoring and agriculture. This breakthrough was recently published in Nature Communications.

The human eye is a specialized sense organ capable of receiving visual images which are relayed in the brain.  The eye provides crucial information about the world around us by converting visible light into signals for different colours using three photoreceptor cells. When our brain combines the information, it develops a prediction about what they imply based on our previous experience.

Although the human eye is impressive, it is far from being the most advanced natural light sensor out there. For instance, the eyes of the Mantis shrimp have 16 different cells that are sensitive to ultraviolet, visible and near-infrared light. Measuring the spectrum in the infrared is particularly attractive for applications in industry and agriculture, but the fundamental problem that exists is: current near-infrared spectrometers are simply too large and expensive. However, experts have solved this issue by developing a near-infrared sensor that fits on a compact chip. And similar to the eye of the Mantis shrimp, it has 16 different sensors, but they are all sensitive in the near-infrared. Besides, miniaturization of the sensors while keeping costs low was a major challenge.

The experts have been looking into this technology for a while and have now successfully integrated the spectral sensors on a chip while also addressing another important issue: data efficiency. When a sensor monitors light, the resultant signal is usually used to rebuild the material's optical spectrum or optical fingerprint. The data is then analysed using sensing algorithms.

The researchers demonstrate that spectrum reconstruction is not required in this new method. To put it another way, the sensors' signals can be delivered directly to the analysis algorithms which significantly simplifies the design requirements for the device.