AVS 50th International Symposium
    Applied Surface Science Wednesday Sessions
       Session AS-WeP

Paper AS-WeP16
Using MEMS Microarrays and Neural Networks to Identify Preferred Surface Chemistry in Application-Specific Gas Sensing

Wednesday, November 5, 2003, 11:00 am, Room Hall A-C

Session: Poster Session
Presenter: Z. Boger, National Institute of Standards and Technology
Authors: Z. Boger, National Institute of Standards and Technology
R.E. Cavicchi, National Institute of Standards and Technology
D.C. Meier, National Institute of Standards and Technology
C.B. Montgomery, National Institute of Standards and Technology
S. Semancik, National Institute of Standards and Technology
Correspondent: Click to Email
Surface chemical interactions between target analytes and sensing films are of primary importance in determining gas sensor performance. However, selecting the most appropriate sensing materials and conditions to detect certain analytes in different gas monitoring applications is very difficult. We present a novel approach for identifying well-suited thin film and surface compositions, preferred microstructures, and operating temperatures for conductometric gas sensing. It employs response databases measured from multi-element MEMS microarrays as well as artificial neural networks (ANN) signal processing. The array elements are microsensor platforms that contain varied sensing film types (TiO@sub 2@ and pure and surface-modified SnO@sub 2@), and can be programmed to operate individually at a variety of temperatures (20 °C to 500 °C), while the composition of test gases is changed. Recursive ANN pruning and re-training techniques are used to identify the more relevant inputs (materials, temperatures) for recognizing and quantifying a specific analyte in a given background. We describe several example cases involving pure and metal-dosed oxide films for: 1) recognition of 6 gases (H@sub 2@, ethanol and varied VOCs); 2) estimation of concentrations for binary mixtures of H@sub 2@, methanol and acetone in the 5-200 ppm (micromole/mole) range; 3) concentration estimation of chemical warfare agents (sarin, tabun, sulfur mustard) and simulants in the range of several ppb (nanomole/mole) to several ppm. Greatly reduced subsets of materials and temperatures, in the range of 3-15 out of a possible 80-1260 combinations, have been shown to carry the most effective analytical information. These selected interfaces and operating conditions provide a basis from which to understand, and then predict, the adsorption, desorption and reaction behavior that are critical to solid state transduction processes for varied classes of gaseous compounds.