Abstract:In the absence of coating, the only way to improve the sensitivity of silicon microcantilever-based density sensors is to optimize the device geometry. Based on this idea, several microcantilevers with different shapes (rectangular-, U-and T-shaped microstructures) and dimensions have been fabricated and tested in the presence of hydrogen/ nitrogen mixtures (H2/N2) of various concentrations ranging from 0.2% to 2%. In fact, it is demonstrated that wide and short rectangular cantilevers are more sensitive to gas density changes than U-and T-shaped devices of the same overall dimensions, and that the thickness doesn't affect the sensitivity despite the fact that it affects the resonant frequency. Moreover, because of the phase linearization method used for the natural frequency estimation, detection of a gas mass density change of 2 mg/l has been achieved with all three microstructures. In addition, noise measurements have been used to estimate a limit of detection of 0.11 mg/l for the gas mass density variation (corresponding to a concentration of 100 ppm of H2 in N2), which is much smaller than the current state of the art for uncoated mechanical resonators.

No abstract

The uncoated silicon microcantilever (USMC) operated in the dynamic mode is a new concept in the field of microcantilever-based chemical sensors. Due to the absence of a sensitive layer, this kind of microsensor can only be used for specific applications where it is known that only one chemical species may be varying in concentration, such as monitoring hydrogen release in radioactive waste disposal facilities. Usually, the relative variation of the USMC resonant frequency expected for low concentrations (≤2%) of hydrogen in nitrogen is below 50ppm. As a result, the measurement of both the resonant frequency, f r , and the quality factor, Q, by classical methods, based on the gain spectrum (resonant peak and-3dB bandwidth), is not sufficiently accurate. In this paper, new measurement methods for monitoring f r and Q variations are proposed: (1) variation of gain and phase at fixed frequencies and (2) polynomial approximations of gain and phase spectra. The performance study of these characterization methods shows that monitoring f r by using phase linearization yields the best signal-to-noise ratio (e.g., 100 at 0.6% of H 2 in N 2), with 0.02% as a limit of detection for hydrogen.