Sound-Based Depth Estimation of Glass Microchannel in Laser-Induced Backside Wet Etching Using Wavelet Transform

Abstract

Laser-induced backside wet etching (LIBWE) has been proposed to fabricate high-quality micromachined components on transparent materials. However, the process is limited by poor repeatability when fabricating high-aspect-ratio structures, even under the same conditions due to uncertainties arising from the thermal process and the complex mechanisms associated with the indirect irradiation of the etching process. Such errors could lead to redundant trials and wastages when trying to achieve the desired dimension. To identify the factors causing these variations, we targeted the process sounds generated during the etching. This study uses a microphone to measure factors that result in variations in material removal quantity during the etching process under the same conditions. The sound was filtered at frequencies between 3 and 6 kHz, which were selected as characteristic frequencies for the process under various laser conditions. By integrating the root mean squared value of the detail coefficient of the wavelet transform, the depth estimation closely matched the measured depth of the fabricated part. This finding suggests that determining the etching rate from sound at a certain characteristic frequency during the LIBWE process is feasible; this approach can improve the accuracy and repeatability of the process. Based on this estimation mechanism, we designed a closed-loop feedback control system capable of fabricating highly accurate microchannels in the range of 80–120 μm with a maximum error of 5.6%.