Semiconductor sensors, which are inexpensive and consume little energy, are used whenever frequent measurements are needed: for instance, the ripeness of fruits and vegetables can be determined by the concentration of gases in greenhouses, and the condition of perishable products can be assessed by CO2 content in refrigerators. They are also used in healthcare: patients with type 1 diabetes exhale air that contains acetone, the amount of which can be used to determine whether an insulin injection is needed.
However, semiconductor sensors have a drawback: due to their sensitivity, they react to a number of substances, which makes it difficult to obtain information about the content of a specific compound. The selectivity of sensors can be increased via imitating the human sense of smell.
“The electrical resistance of the sensory material, which is carbon nanotube fabric in our case, changes in response to the presence of many gases. Its reaction varies depending on different gases in different concentrations, which opens a technological opportunity. After all, the human sense of smell is also designed in such a way that epithelium receptors react to a variety of substances. But different groups of receptors react to each smell differently. And the strength of activation of the neurons associated with them helps the brain recognize each smell,” graduate student Konstantin Zamansky is quoted as saying by Skoltech.
By the same principle, a few semiconductor sensors can be used simultaneously, each reacting to a type of gas. This forms an artificial nose that can react to the presence of a gas mixture in the air. The Skoltech scientists have taken a step forward in this direction by obtaining a multidimensional signal from the same sensor at different temperatures.
The authors created a sensor in which the sensory material, a piece of fabric made of single-wall carbon nanotubes, was suspended between two electrodes. This solution made it possible to heat or cool the sensor almost instantly: within 40 seconds, the device measures resistance at 400 different temperatures in the range from 25°C to 125°C. These 400 values represent an odor pattern that can be recognized with 90% accuracy using a machine learning model.
Thanks to the solution developed by Skoltech, one sensor can be used to detect several different gases at once. As a result, this study will make it possible to reduce the cost of detecting harmful impurities.
