Sie sind auf Seite 1von 4

J. Micromech. Microeng. 8 (1998) 8487.

Printed in the UK

PII: S0960-1317(98)86813-7

Deep wet etching of borosilicate glass using an anodically bonded silicon substrate as mask
Thierry Corman, Peter Enoksson and Goran Stemme
Department of Signals, Sensors and Systems, Royal Institute of Technology, S-100 44 Stockholm, Sweden Received 29 August 1997, accepted for publication 10 December 1997 Abstract. Deep wet etching of borosilicate glass using an anodically bonded silicon substrate as mask is presented. Depths of 500 m or more can be achieved very easily. The structured glass wafer can be bonded anodically on the same side to another silicon wafer, after having removed the bonded silicon mask. A lateral underetching 1.5 times larger than the depth was measured. An application using this masking technique is also presented. It consists of using the anodically bonded frame of a resonant silicon structure as a mask for deep glass etching to increase the gap between the glass wall and the resonator, thus yielding a high Q -factor.

1. Introduction Besides silicon, various types of glass are widely used in microsensor and microactuator technologies, due to the advantages they can provide. For instance, they are electrically insulating, resistant to many chemicals, can be microstructurable, are transparent at certain wavelengths and can be joined to many different metals [1]. The main requirements for glass in the eld of microsystem technology are that it should be microstructurable, preferably using a standard lithography process (batch fabrication); contain easily movable ions to perform anodic bonding; have a thermal expansion coefcient matching the thermal expansion coefcient of silicon in order to avoid thermal stress. These requirements are generally very difcult to full with only one particular type of glass. Generally, easily bondable glass is difcult to etch because of its chemical composition and easily structurable glass is difcult to bond because its thermal expansion coefcient does not match the thermal expansion coefcient of silicon and/or because of the absence of easily movable positive sodium ions [1]. This paper presents a simple and reliable technique to etch deep in borosilicate glass using a anodically bonded silicon substrate as mask. The proposed solution meets the requirements needed in microsystem technology, i.e., the glass is both easy to structure using a standard lithography technique and easy to anodically bond afterwards.
Tel: +46 8 7907789; fax: +46 8 100858; e-mail: corman@s3.kth.se 0960-1317/98/020084+04$19.50 c 1998 IOP Publishing Ltd

2. Etching techniques Many techniques have been used to structure siliconbondable glass. (a) Deep etching in borosilicate glass using a polysilicon lm deposited by LPCVD has been reported [2]. The LPCVD polysilicon mask can be used to etch through a 500 m thick Pyrex glass wafer. This method is compatible with standard lithography techniques and no underetching (except the isotropically induced one, i.e. one times the etch depth) is generated. The problem with this method is that during the high temperature LPCVD deposition process, the Na+ ions diffuse out of the glass and contaminate the furnace. This contamination will seriously degrade the electrical characteristics of silicon. A furnace so contaminated could be dedicated for unclean MEMS applications, but the high cost of the equipment combined with rentability considerations make it very difcult to dedicate a furnace just for unclean applications. This masking method might be used just before a furnace cleaning. A similar method using PECVD silicon carbide as mask has also been reported [3]. (b) Etch masks of Cr+Au+resist are often used [46]. The chromium layer is used to improve the adhesion of gold or resist to glass. Although this method is simple and compatible with standard lithography techniques, the mask cannot be used for very large etch depths since pin-holes appear after a while. Moreover, a large lateral underetching of chromium is present. This can be reduced by heating the substrate before and after the mask deposition [7]. These drawbacks (poor adhesion to substrate and bad resistance

Etching of borosilicate glass

(a)

(b)

(c)

(d)

Figure 1. SEM photos showing the cross-section of etched cavities in Pyrex 7740 glass wafers at four different depths. As a mask layer, a previously structured and anodically bonded 0.5 thick silicon wafer was used. The depths reached are (a) 100 m, (b) 190 m, (c) 350 m and (d) 470 m, obtained by submerging the glass wafers in a solution of HF (50%):H2 O (1:5). The silicon bonded mask in (d) is still unaffected after 470 m and the etching can continue.

to HF) limit the utilization of this type of mask to shallow etchings of less than 50 m (Pyrex glass). (c) Other methods like laser structuring [1], electrochemical discharge drilling [4, 7] or sand blasting can also be used. These methods enable the obtaining of straight walls in glass, but they do not meet the desired standard batch process requirements and present problems of accuracy. 3. Experiments The glass wafers used for the tests are 500 m thick borosilicate Pyrex 7740 glass wafers, which are commonly used for anodic bonding in microsensor and microactuator technology. The masking layer used for the experiments is a silicon wafer previously microstructured in KOH and anodically bonded to the Pyrex glass wafer. The substrates are then submerged in a solution of HF (50%):H2 O (1:5) for the etching experiments. The concentration 1:5 was chosen to obtain smooth and regular surfaces. The corresponding

etching speed at this concentration was 12.9 m h1 . The etching was stopped at four different depths: 100 m, 190 m, 350 m and 470 m. The results are illustrated in gure 1, showing SEM photos of the same magnication taken at the four different depths. A back-side protection consisting of a dummy Si wafer joined to glass with black wax (Picein 105) was used for the 350 m and 470 m deep-etched wafers. The cross-section shown in gure 2 represents a global view of the silicon mask which is almost hanging over the etched glass substrate, due to the HF isotropic etching of glass. Figure 3 shows a SEM photo taken after having nearly etched through a 500 m thick glass wafer. From this photo, it can be seen that even a thin layer of silicon (formed in the V-groove during the KOH etching) can protect the glass substrate. To verify whether the glass wafer is still bondable on the same side after the etching, a new silicon wafer was anodically bonded to the glass after the removal of the silicon mask in KOH.
85

T Corman et al

Figure 2. SEM photo showing an etch depth of 350 m in Pyrex 7740 glass using a bonded silicon substrate as mask layer.

Figure 4. Utilization of a resonator frame as mask for deep glass etching to minimize damping and increase the Q -factor.

Figure 3. SEM photo illustrating the prole obtained after the etching through a Pyrex 7740 glass wafer of 0.5 mm.

4. Results The tests have shown that it is possible to etch a cavity of 500 m and even more, keeping a very good mask without any defects after the deep etching. This etch mask has a very good resistance to HF and the anodic bonding provides a very good adhesion to glass. As illustrated in gure 1, for all depths, the lateral underetching measured is 1.5 times larger than the vertical etching. A possible explanation for this larger lateral etching could be the presence of a thin layer of SiO2 (formed during the anodic bonding process [8]) at the glass silicon interface which is etched faster than glass. In HF, an LPCVD TEOS deposited oxide is etched approximately 1.5 times faster than Pyrex glass, and eight times faster than a thermal grown oxide. This underetching is not signicant compared to other masking methods like, for example, Cr+Au, which presents much larger lateral undercutting (about ve times larger [5], which was also veried by our measurements). Due to the application, a 0.5 mm thick Si wafer was
86

used, even though the results show that a much thinner mask could have been used. When making the mask design, the anisotropic KOH etching of silicon, which forms inclined walls, should be taken into consideration along with the silicon thickness to be etched, as it can be clearly seen from gures 2 and 3. To be able to dene accurate lithography patterns, it is favourable to have a thin silicon substrate as mask. Another way to avoid problems of layout design and pattern accuracy is to use deep reactive ion etching (DRIE) which generates straight walls in the silicon substrate. Very thin silicon wafers of a few micrometers are commercially available. The use of these wafers as mask enables very good accuracy to be obtained, especially if they are used in combination with deep reactive ion etching. We also performed some experiments by sputtering a silicon lm on Pyrex glass wafers which was subjected to an anodic bonding process. The result obtained was similar: the anodically bonded Si lm could better withstand HF than the non-bonded lm. However, the use of a sputtered bonded Si lm as mask has been limited by the appearance of pin-holes. Another important result is that after removing the silicon-bonded mask in KOH and washing the structured glass in H2 SO4 :H2 O2 (2.5:1), it is possible to anodically bond a new silicon wafer on the same glass side. This result clearly demonstrates the possibility of bonding a glass wafer anodically several times on the same side using a proper surface treatment. 5. Application: resonator structure frame [6] used as mask to minimize damping and increase the Q -factor The idea is to use the frame of a silicon vibrating structure as mask in order to increase the distance between the glass lid wall and the oscillating surface. This will minimize the damping due to the squeezed gas between the walls during the oscillations, thus yielding a high Q-factor. Figure 4 illustrates the processing steps for this application. First, a shallow depth (20 m) is etched using a conventionally Cr+Au+resist mask to make the opening

Etching of borosilicate glass

under the resonator. The structure is then anodically bonded to the glass lid. The bonded frame of the resonator can then be used as the mask to etch deeper in the glass. Finally, another glass lid is bonded on the other side to encapsulate the resonator inside a low pressure cavity. Using this method, with a solution of HF (50%):H2 O (1:5), a glass lid with a 175 m deep cavity was fabricated. Another glass lid having a 20 m deep cavity was anodically bonded in vacuum on the double-stack wafer to encapsulate the silicon resonator. The Q-value obtained for this structure was 5000 which was ve times larger than the Q-value measured for a similar structure having a 20 m deep cavity on both sides. 6. Conclusions A technique for deep wet etching of borosilicate glass using anodically bonded silicon as mask is presented. One advantage of this method is the possibility of etching deep in glass with a relatively small underetching. The microstructured glass can be anodically bonded again on the same side to another silicon wafer, after the silicon mask removal. This technique, which only requires anodic bonding before the etching, is very simple and reliable and can nd many interesting applications.

References
[1] H ulsenberg D et al 1994 New glasses for microsystem technologies Proc. Micro System Technol. pp 25968 [2] Gr etillat M-A, Paoletti F, Thi ebaud P, Roth S, Koudelka-Hep M and de Rooij N F 1996 A new fabrication method of borosilicate glass capillary tubes with lateral inlets and outlets Proc. Eurosensors X pp 25962 [3] Flannery A F, Mourlas N J, Storment C W, Tsai S, Tan S H and Kovacs G T A 1997 PECVD silicon carbide for micromachined transducers Proc. Transducers 97 pp 21720 [4] Diepold T and Obermeier E 1995 Smoothing of ultrasonically drilled holes in borosilicate glass by wet chemical etching Proc. MME95 (Copenhagen) pp 358 [5] Shoji S, Kikuchi H and Torigoe H 1997 Anodic bonding below 180 C for packaging and assembling of MEMS using lithium alminosilicate- -quartz glass-ceramic Proc. MEMS97 pp 4827 [6] Corman T, Enoksson P and Stemme G 1997 Gas damping of electrostatically excited resonators Sensors Actuators A 61 24955 [7] Shoji S and Esashi M 1990 Photoetching and electrochemical discharge drilling of Pyrex glass Tech. Dig. 9th Sensor Symp. pp 2730 [8] Ko W H, Suminto J T and Yeh G J 1985 Bonding techniques for microsensors Micromachining and Micropackaging of Transducers (Amsterdam: Elsevier) pp 4161

87

Das könnte Ihnen auch gefallen