This paper studies spin-on glass (SOG) etching in T-shaped microchannels by hydrofluoric acid (HF). Since oxide etching by HF in microchannels is both reaction and diffusion limited, an etching model based on non-first order chemical reaction/steady-state diffusion sacrificial layer etching mechanism is presented to compensate for the etching effect at channel junction. Microchannels are formed on silicon substrate by deep reactive ion etching (DRIE). Samples with channel depth varying from 1μm to 6 μm are prepared by varying exposure time to reactant gas in DRIE chamber. Channel widths prior to the junction are varied from 2 μm to 10 μm while channel width beyond the junction is fixed at 5 μm. The channels are then filled with SOG by multiple spin, bake and cure processes. After etchback planarization using 5% HF solution, the samples are coated with 1.5 μm thick positive photoresist. An etch window is opened at channel fronts to expose underlying SOG. The samples are then time-etched in 5% HF solution and etch front propagation is observed under optical microscope through the transparent photoresist layer. It is observed that SOG etch rate in the microchannels is independent of channel width or channel depth. SOG etch rate at channel's T-junction is 0.67 times lower than etch rate in the straight channels preceding it due to HF concentration variation and etch product transfer rate variation effects. The proposed model fits experimental data well. Offset crosses vent pattern is determined as a good candidate for removing sacrificial oxide under an enclosed cap structure.
This paper presents a method to form thick spin-on glass (SOG) sacrificial layer for accelerometer encapsulation fabrication. SOG is chosen as the sacrificial material because it is easy to apply, has good thickness uniformity, and can be easily etched back before densification. Siloxane type SOG is applied on blank wafers and accelerometer patterns by multiple spin, bake, and cure processes. A series of gradual hot plate baking up to 250°C are experimented for each spun layer. After multiple spin and bake, the SOG layers are etched back in hydrofluoric acid (HF) solution of various concentrations to form rectangular encapsulation bases. 25 samples are prepared for SOG thickness uniformity characterization. Center thickness and four perimeter thickness measurements are taken for each sample using thin-film mapper. These five measurements are averaged to get sample thickness. Two surface profiler measurements are taken for each sample perpendicularly to each other using Tencor surface profiler. The minimum reading is subtracted from the maximum reading to get sample variation. Upon SEM inspection, mildly sloped etched walls from HF etching are observed. No surface cracking was visible. Shallow trench patterns are apparent on SOG deposited on accelerometer pattern. The average sample thickness is 5 μm with 3.7% thickness variation across samples. The average variation within each sample is 0.14 μm with an average of 2.6% thickness variation within sample. These thickness variations are acceptable for encapsulation structure deposition.