(295m) Inoses: Intelligent Nature-Inspired Olfactory Sensors Engineered to Sniff

The need to acquire real-time information about the air we breathe has been brought into the spotlight through recent events, such as wildfires, the COVID pandemic, hazardous spills, and increasingly stringent emissions regulations. Standard methods to determine the composition of gaseous samples typically rely on bulky and expensive spectroscopic equipment, necessitating the transportation of potentially hazardous materials to the monitoring site. This increases risk and contributes to a high environmental footprint. Despite significant advances, current portable sensors are still limited in their ability to accurately analyze a diverse array of volatile compounds and mixtures. Dogs are still employed in many scenarios (such as detection of explosives and contraband, as well as, more recently, early disease detection), highlighting the impressive capabilities of the natural olfactory system. Meanwhile, sensors for volatile organic compounds (VOCs) and other hazardous air pollutants (HAPs) are often reported to be unreliable, difficult to calibrate, suffer from drift, and are highly susceptible to humidity and temperature changes.

In order to address the need across various industries for low-cost, accurate and real-time solutions for identifying volatile mixtures, we have developed intelligent Nature-inspired Olfactory Sensors Engineered to Sniff (iNOSES). Distinctive to our approach, the device implements biologically inspired sniffing: it generates and self-optimizes adaptive patterns of inhale-exhale sequences, which capture non-equilibrium mass transport phenomena experienced by each compound or mixture due to differences in their physicochemical properties, mimicking critical natural odor discrimination mechanisms. The presence of analytes is recorded in the form of time-resolved sensor output that is featurized to enable machine learning (ML) techniques for analyte detection and classification. Thus, discrimination of analytes is achieved via the interactions between the “olfactory system” (our sniffing device) and the “brain” (represented by ML models). We demonstrated that a sensor of this type, which incorporates non-equilibrium mass-transport dynamics, temporal data collection, and ML modeling, substantially enhances the detection power of any artificial nose, including – as is in our case – a single, mesoporous one-dimensional photonic crystal, which we demonstrated could distinguish several volatile polar and non-polar compounds as well as their binary mixtures, and could predict physical properties of known and unknown compounds. The approach is hardware-agnostic and can be applied to practically any volatile sensor, allowing us to broaden the repertoire of detectable gasses. The vision of iNOSES is to enable individual consumers, government agencies such as the US EPA, residential and non-residential building entities, and more to have access to accurate, affordable, real-time volatile sensors to make reliable measurements and predictions of indoor and outdoor air composition. This is a critical step to improving human and ecosystem health as well as defining and standardizing emissions regulations.