Abstract
A room-temperature multimodal sensor composed of PEDOT:PSS deposited on an AT-cut quartz crystal
microbalance (QCM) crystal has been fabricated. The nonlinear resistive and frequency sensor responses are
deconvolved using an arti cial neural network (ANN), which allows the single sensor to function simultane-
ously as a relative humidity (RH) sensor and a pressure sensor using only two electrodes. We demonstrate
that the predictive ability of the sensor is highly in
uenced by the data used to train the ANN. When training
sets are tailored to resemble the operating conditions of the sensor, the sensor achieves an average resolution
of < 3% RH from 0-100% RH, even after H2O saturation occurs on the surface. Our results indicate that
ANNs show strong promise for improving the resolution of low cost gas sensors and for expanding the range
of environmental conditions in which a given sensor can operate.
