5.2 Development of components via DRIE
5.2.2 Frontside photolithography and masking
After completion of the KOH etching the wafers were prepared for etching from the opposite side as seen in Figure 5.1. The DRIE would be used to form structures etched all the way through the remaining silicon layer. The silicon nitride layer used before for KOH etching is not resistant to the DRIE process therefore other materials were used for form an etch mask. Because the etch depth would be on the order of 150 um a relatively robust etch
mask was required. Two different types of etch masks were used, the first a chromium etch mask and the second a KMPR photoresist etch mask.
5.2.2.1 Chromium Etch Mask Process
For the fabrication of the silicon gate mesh as well as the silicon chip
window/collimator a chromium etch mask was formed. The chromium etch mask was formed by DC magnetron sputter deposition using a Lesker PVD 75. Approximately 2000A of Cr was deposited from a 3 in target onto each substrate at a power of 300W at a rate of just under 2 A/s in ~20 min.
After removal from the sputter system the wafers were spin coated with ~1.5um Shipley 1813 positive photoresist, exposed via a dark field mask, and developed for 60 s in MF 319 developer. The photolithography step created a pattern by removing the photoresist in the shape of the holes to be etched in the DRIE. For each of these structures most of the wafer was covered in photoresist with only the cumulative area of the small holes left behind. Next the wafer was hard baked at 125 C for 60 s prior to chromium etching. Room-
temperature CR-7 chromium etchant solution was used to etch the chromium layer. The wafers were etched for approximately 120 s total. The chrome mostly cleared after about 90 s, but additional overetch time was included to fully etch the chrome and possibly eliminate or undercut smaller defects in the features.
After etching the chrome through the photoresist layer the underlying silicon nitride layer was exposed. This was etched in the SAMCO RIE-1C according to the process described earlier; this allowed the silicon layer to be exposed directly to the DRIE process. Additionally the 1.5 um photoresist layer was left in place to provide additional protection during the DRIE process.
Figure 5.1: Schematic, cross sectional view of DRIE process using a chromium etch mask. a) Silicon Nitride wafer following KOH etching of backside is coated with a 2000A Cr layer. b) Photoresist coating and pattern formation. c) Profile after etching of Cr layer via chrome etch and of silicon nitride via RIE. d) DRIE of exposed silicon layer e) Wafer after completion of DRIE; holes etch completely through the silicon layer, photoresist may be completely consumed after extensive etching. f) The silicon gate mesh is produced following the same DRIE process only using a different photolithography pattern.
The chromium etch mask proved to be somewhat reliable although in some cases chromium layer adhesion loss during the DRIE process affected the device outcome. These failures may have been affected by the lack of wafer cleanliness. Standard cleaning and drying methods in our facilities are more rudimentary and leave open more possibility for contamination.
5.2.2.2 KMPR Photoresist Etch Mask Process
For the fabrication of the multi-slit structure an initial attempt was made to form a chromium etch mask, but after etching in DRIE this mask showed the above mentioned adhesion loss resulting in extreme broadening of the slit width. The slit was designed at approximately 10um, but after Cr etch mask adhesion loss the slit width etched into the substrate was closer to 40um.
As a result of this failure an alternate DRIE masking material was used. Microchem company’s KMPR 1010 photoresist was used to form the etch mask. KMPR is a negative, resin-based photoresist which has particularly high performance for DRIE applications. Typical postive photoresists have lower selectivities of 15:1 or 20:1 when undergoing RIE or DRIE. KMPR can exhibit etch selectivities as high as 100:1, allowing for deep etching into the silicon using only 1-10 um of photoresist. Also other available photoresists are limited to coating thicknesses of 2-3 um. Additionally KMPR lithography produces highly vertical lithography sidewalls, because of its highly transparent nature, allowing for greater fidelity in pattern transfer during the etch.
In this case KMPR 1010 photoresist was diluted with Microchem’s SU-8 2000 thinner to reduce the solids concentration to a level similar to that of KMPR 1005. KMPR 1010 typically spin coats a nominal 10 um layer at 3000 rpm spin speeds, while KMPR 1005 form an approximately 5 um layer. The amount of additional solvent needed was calculated to be around 20% by volume; therefore approximately 8 ml of thinner was added to 40 ml of KMPR 1010.
The KMPR 1005 photoresist mixture was then spin coated onto the silicon substrate at 3000 rpm for 30 s and soft baked for 5 min at 100C. Subsequently the sample and mask features were aligned and exposed with a 20 s dose of about 240 mJ and given a post exposure bake of 100 C for 2 min. The sample was then developed in Microposit’s CD26 developer, an aqueous solution with 2.4% Tetramethylammonium hydroxide, for 3 min with moderate agitation. Inspection of the film after development showed a relatively defect free surface and an estimated film thickness of 7 um, a more than adequate thickness of KMPR assuming moderate etch selectivity.
The KMPR process was similar to the chromium etch mask process shown in Figure 5.1. The Cr deposition, lithography, and etching of Cr and silicon nitride steps in a), b), and c) were replaced with the spin coating, exposure, and development of KMPR, along with an etching of the silicon nitride layer. The substrate then was placed in the DRIE with an outcome similar to f) except free of the Cr layer.