Piezoelectric cantilever resonator is one of the most promising platforms for real-time sensing of volatile organic compounds (VOCs). However, it has been a great challenge to eliminate the cross-sensitivity of various VOCs for these cantilever-based VOC sensors. Herein, a virtual sensor array (VSA) is proposed on the basis of a sensing layer of GO film deposited onto an AlN piezoelectric cantilever with five groups of top electrodes for identification of various VOCs. Different groups of top electrodes are applied to obtain high amplitudes of multiple resonance peaks for the cantilever, thus achieving low limits of detection (LODs) to VOCs. Frequency shifts of multiple resonant modes and changes of impedance values are taken as the responses of the proposed VSA to VOCs, and these multidimensional responses generate a unique fingerprint for each VOC. On the basis of machine learning algorithms, the proposed VSA can accurately identify different types of VOCs and mixtures with accuracies of 95.8 and 87.5%, respectively. Furthermore, the VSA has successfully been applied to identify the emissions from healthy plants and "plants with late blight" with an accuracy of 89%. The high levels of identifications show great potentials of the VSA for diagnosis of infectious plant diseases by detecting VOC biomarkers.

Multifunctional environmental sensing is crucial for various applications in agriculture, pollution monitoring, and disease diagnosis. However, most of these sensing systems consist of multiple sensors, leading to significantly increased dimensions, energy consumption, and structural complexity. They also often suffer from signal interferences among multiple sensing elements. Herein, we report a multifunctional environmental sensor based on one single sensing element. A MoS 2 film was deposited on the surface of a piezoelectric microcantilever (300 × 1000 μm 2 ) and used as both a sensing layer and top electrode to make full use of the changes in multiple properties of MoS 2 after its exposure to various environments. The proposed sensor has been demonstrated for humidity detection and achieved high resolution (0.3% RH), low hysteresis (5.6%), and fast response (1 s) and recovery (2.8 s). Based on the analysis of the magnitude spectra for transmission using machine learning algorithms, the sensor accurately quantifies temperatures and CO 2 concentrations in the interference of humidity with accuracies of 91.9 and 92.1%, respectively. Furthermore, the sensor has been successfully demonstrated for real-time detection of humidity and temperature or CO 2 concentrations for various applications, revealing its great potential in human−machine interactions and health monitoring of plants and human beings.

No abstract