Study Of SnO₂ Formaldehyde Gas Sensor Based On IC Technology

Formaldehyde was widely used in industrial, pharmaceutical and other areas of daily life. However, formaldehyde gas is very harmful to human health. Thin film Si-based micro-sensor possesses many advantages such as small volume, low-cost, easy to form cosmically and excessive function array, and so on. The main contents of this thesis includes, the electrode design of the thin film Si-based formaldehyde gas sensor, the process conditions research of the SnO₂ film preparation on silicon by sol-gel method, the fabrication of Pd-doped SnCO₂/Pt/SiO₂/Si formaldehyde gas sensor, the analysis of the higher response of the Pd-doped SnO₂ sensor than the one of the undoped SnO₂ sensor, the solution of the selectivity of the sensor to mixed gas by BP neural network.ANSYS software was used to optimize the narrow middle and wide exterior electrode structure for the gas sensor. The Pt electrode was fabricated using IC technology with oxidation, lithography, sputtering, peeling off. The SnO₂ film was fabricated on the silicon substrate by sol-gel method. The best process conditions have been found by the experiments such as sol configuration, coating, pre-baking, annealing, etc. The thin film has been characterized by TG-DTA, step profiler, SEM, AFM, XRD.The Pd-doped SnO₂/Pt/SiO₂/Si formaldehyde gas sensor was fabricated. The best heating temperature was 230℃in which the response of the sensor to formaldehyde was the best. The experiments showed that the response of the 1 at.% Pd-doped SnO₂ sensor was higher than that of undoped SnO₂ sensor. The lowest concentration to formaldehyde tested by Pd-doped SnO₂ sensor is 0.05 ppm. The response time and the recovery time were both about 50 s. XPS peak O 1s spectra of undoped and 1 at.% Pd-doped SnO₂ sensitive films were tested. The result showed that the ratio of adsorbed oxygen to lattice oxygen greatly increased for doped film, and then the reaction active sites increased. It is beneficial to the reaction of adsorbed oxygen and the reducing gas such as formaldehyde, the gas sensitive responsiveness to formaldehyde has been improved. The disturbing gas of ethanol was tested, the reason why the selectivity was not ideal had been analyzed.For solving the disturbing gas of ethanol to formaldehyde, a three-level BP neural network was use to identify the gases and measure the concentrations. There were 4 nodes in input layer, 18 nodes in hide layer and 2 nodes in output layer. Some reserved samples were predict using the trained network, the results show that, the network was able to distinguish quantificational formaldehyde, ethanol from mixed air using this method.