Metallization introduced corrosion and parylene protection of surface micromachined polysilicon film with submicron capacitive gap

Microelectronic Engineering 97 (2012) 20–25

Contents lists available at SciVerse ScienceDirect

Microelectronic Engineering journal homepage: www.elsevier.com/locate/mee

Short Note

Metallization introduced corrosion and parylene protection of surface micromachined polysilicon film with submicron capacitive gap Yiming Zhang, Yunda Wang, Ming Cai, Ying Wang, Yilong Hao, Jing Chen ⇑ National Key Laboratory of Science and Technology on Micro/Nano Fabrication, Institute of Microelectronics, Peking University, Beijing, China

a r t i c l e

i n f o

Article history: Received 4 September 2011 Received in revised form 1 February 2012 Accepted 15 March 2012 Available online 23 March 2012 Keywords: MEMS Polysilicon corrosion HF release Parylene

a b s t r a c t Corrosion of polysilicon occurs in the hydrofluoric acid (HF), which may change the surface morphology, film thickness and even its mechanical strength; furthermore, long time exposure to HF and the presence of Au metallization appear to promote this corrosion. This unusual phenomenon was observed during the releasing of the clamped–clamped beams with a capacitive gap of 100 nm, where obvious color change of the surface polysilicon was revealed. Then the influence on electrical performance of the structure was evaluated as well as the relationship between the maximum releasing time and the concentration of HF acid. After that, a novel protective coating was developed to prevent the corrosion. Parylene was demonstrated to be an effective protection mask in HF wet-etching process, where thickness dominated the yield because parylene film swelled in HF etchant. With a parylene protection layer of 0.5–2 lm, Al pads can withstand 40 min exposure to 40% HF. Finally, a parylene layer of 2 lm was deposited and patterned on top of the metal electrode, which proved to be very effective for isolating the polysilicon from galvanic corrosion during releasing. Ó 2012 Elsevier B.V. All rights reserved.

1. Introduction Polycrystalline silicon (polysilicon) is widely employed in Microelectromechanical System (MEMS) [1–3], while Au (with an adhesion layer such as Cr/Ti/TiW) and Al are the most popular metallization layers for making electrical contacts. In surface micromachining, Au was preferred because that HF would attack Al during the releasing, where Au could still remain its integrity. In this process, the final step is usually dissolving the sacrificial oxide in concentrated hydrofluoric acid (HF). However, it is reported that corrosion of polysilicon occurs during the HF release, which may change the surface morphology, film thickness and even its mechanical strength [4–7]. Furthermore, long time exposure to HF and the presence of Au metallization appeared to promote this corrosion. Polysilicon that was electrically connected to Au was etched in the HF release step such that those features became slightly thinner and rougher than the features that were not in electrical contact with Au [8–11]. On the contrary, no such changes were reported when the metallization layer was Al. However, to improve the performance of sensors and actuators, the capacitive gap of the surface micromachined device is usually reduced to submicron even nanometer scale. A relatively long time is required to fully release the device, in which case Au metallization is employed. Although it is recognized that the influence introduced

⇑ Corresponding author. E-mail address: [email protected] (J. Chen). 0167-9317/$ - see front matter Ó 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.mee.2012.03.010

by polysilicon corrosion may jeopardize its performance, this phenomena has not yet been well investigated. In this paper, surface micromachined clamped–clamped beams (CC-beam) made of P-doped polysilicon with metallization layers of Cr/Au and Al were fabricated to characterize this unusual phenomenon. Changes of the surface morphology were revealed, and the influence on electrical performance of the structure was evaluated. The release criteria and the origin of this effect were discussed. Finally, a parylene protection method was developed to prevent the polysilicon corrosion.

2. Experiment Clamped–clamped beams (CC-beam) with a capacitive gap of 100 nm were manufactured using 2-layer polysilicon (P-doped) surface micromachining process in the National Key Laboratory on Micro/Nano Fabrication Technology, Peking University. Cr/Au and Al were sputtered and patterned respectively for making the electrical contacts. Fig. 1 shows the schematic cross section of the CC-beam. The thickness of Poly1 was 0.3 lm and Poly2 was 2 lm. Both polysilicon layers were P-doped by implantation (Poly1: dosage 1e16 cm2; energy 80 KeV; Poly2: dosage 2e16 cm2; energy 80 K eV), followed by annealing at 1100 °C in Nitrogen atmosphere for 1 h. The sacrificial layer was LPCVD SiO2 of 100 nm. The diced sample chips were released in HF solution of different concentrations. The surface features were observed with optical microscope and SEM.

Y. Zhang et al. / Microelectronic Engineering 97 (2012) 20–25

21

Fig. 1. Cross section of the CC-beam.

Fig. 3. Surface change of polysilicon after release.

Fig. 4. SEM image of the released cc-beam. Fig. 2. Schematic of the electrical test.

The electrical resistances of Poly1 and Poly2 were tested on a probe station after releasing, as shown in Fig. 2. Rp1 represented the resistance of Poly1, and Rp2 represented the resistance of Poly2. The IV curve was obtained by a semiconductor network analyzer HP 4145B as shown in Fig. 2, and pull-in voltage could be determined.

3. Results and discussion Samples with Au/Cr metallization were released in concentrated HF (40%) with time varying from 2 to 18 min. An obvious color change of Poly1 could be observed in the microscope, as shown in Fig. 3. The color of Poly1 changed to green after 6 min

22

Y. Zhang et al. / Microelectronic Engineering 97 (2012) 20–25

Fig. 5. Relationship between Poly1 thickness and releasing time.

Table 1 Electrical characterization (40% HF). Release time (min)

2

4

6

8

10

12

14

16

Rp1 (X) Rp2 (X) Pull-in

236 214 F

256 218 F

268 219 F

270 222 P

299 230 P

309 236 P

9k 253 F

1 263 F

Fig. 6. Surface of Al samples after HF release.

Table 2 Electrical characterization (40% HF:H2O = 1:1). Release time (min)

8

12

14

16

18

20

22

Rp1 (X) Rp2 (X) Pull-in

225 217 F

229 226 F

240 240 F

298 246 F

343 298 F

453 308 P

1 316 F

Table 3 Electrical characterization (40% HF:H2O = 1:5). Release time (min)

10

20

30

40

60

Rp1(X) Rp2(X) Pull-in

221 220 F

339 238 F

1 254 F

1 306 F

1 321 F

Fig. 7. The section view of testing chip.

releasing and grew darker when the releasing time further increased. Poly1 was almost invisible after 18 min. However, Poly2 did not change much under the microscope. Fig. 4(a) is the SEM photograph of a CC-beam with Cr/Au after 4 min releasing, whose Poly1 electrode became green in Fig. 3. The dark parts in SEM verified the color change. Fig. 4(b) is the SEM photograph of the Poly1 electrode released in HF for 18 min. The surface roughness greatly increased. Thickness of Poly1 was measured by the surface profiler after releasing. The polysilicon is considered to dissolute for the following reactions [12–14].

Si þ 2Fab þ 4HFaq þ kþ ! H2 SiF6 þ H2;aq þ ð2 - kÞe

ðk 6 2Þ ð1Þ

Si þ 4HOab þ kþ ! SiO2 þ 2H2 O þ ð4  kÞe SiO2 þ 6HFaq ! H2 SiF6 þ 2H2 O

ðk 6 4Þ

ð2aÞ ð2bÞ

In Eqs. (1), (2a), and (2b), the subscript ab denotes the surfaceadsorbed ion species, the subscript aq denotes the aqueous species, while k+ and e represent the holes and electrons, respectively. Reactions (1) and (2) may occur simultaneously on the surface. When the release time was less than 10 min, the surface polysilicon was dissolved directly into the electrolytic solution, and porous silicon was formed as Eq. (1). During this period, the thickness of Poly1 did not change too much as shown in Fig. 5. With the release time further increased from 10 to 16 min, the polysilicon had developed porosity during the HF exposure, and the oxidants could easily diffuse through the pores, which greatly promoted reaction in Eq. (2a), a thick surface oxide formed at a faster rate than the dissolution in Eq. (2b). The thickness of Poly1

Y. Zhang et al. / Microelectronic Engineering 97 (2012) 20–25

23

Fig. 8. Samples covered with 2 lm thick parylene with 200 lm edge distance in 40% HF for different etching times, Al not etched.

Fig. 9. Samples covered with 2 lm thick parylene with 100 lm edge distance in 40% HF for different etching times, Al not etched.

increased because the inherent volume expansion involved in silicon oxidation (the molecular volume of SiO2 is more than twice the atomic volume of Si) [8]. After 16 min HF releasing, the dissolution resulted in a coarse consumption of Si, as the entire surface of the anode became covered in oxide, which was vulnerable to HF, as a result, the thickness of the polysilicon decreased drastically. Electrical tests were also conducted, where the samples were released in HF solution of 3 different concentrations, with the results listed in Table 1–3. Changes of Rp1 and Rp2 were presented in the tables, together with the pull-in test results: ‘‘1’’ indicated the electrical failure of the polysilicon; ‘‘F’’ indicated that the device failed the test either for the releasing incomplete or the polysilicon damaged; ‘‘P’’ indicated that pull-in could be detected, which revealed the CC-beam was operational. Samples with Al metallization were also released for comparison. There were no detectable changes regarding the surface morphology and the electrical performance of the polysilicon. Fig. 6 shows a sample with Al releasing for 6 min (a) and 8 min (b) in 40% HF solution. Although the Al layer was almost completely eroded, the surface of Poly1 remained unchanged.

Fig. 10. Samples covered with 2 lm thick parylene of various edge distances in 40% HF solution for 40 min, Al not etched except that with 20 lm edge distance (top right corner).

24

Y. Zhang et al. / Microelectronic Engineering 97 (2012) 20–25

up, Poly1 could be completely removed as Rp1 became ‘‘1’’ in Tables 1–3. Moreover, this corrosion also had a relation with the locations. As shown in Figs. 3 and 4, the corrosion of Poly1 was severe near the CC-beam. The HF concentration also played an important role. A more diluted solution could alleviate the erosion. Unfortunately, the releasing time had to be much longer to fully release the structure, which also induced polysilicon corrosion therefore lead to failure of the whole device. It was important to determine an appropriate process time for the device releasing. As listed in Tables 1–3, the higher the HF concentration, the wider the process window. Concentrate HF was favored in this regard as a general rule. One reason for this phenomenon was galvanic corrosion, where metallization played an important role [15]. Au metallization could promote anodic oxidation, particularly in HF. Once polysilicon and Au were electrically connected in HF, galvanic corrosion started. Then the electrochemical oxidation and subsequent dissolution of the oxides caused the ultimate ‘‘dissolve’’ of Poly1 [16,17]. 4. Parylene protection

Fig. 11. Samples covered with 4 lm (a) and 8 lm (b) thickness parylene in 40% HF solution for 7 min, most Al etched.

Table 4 Summary of the protective testing results. Thickness (lm)

Edge distances (lm)

Etching time (min)

Result

0.5/1/1.5 2

20–200 20 30–200 20–200

1–40 1–40

Al not etched Al etched Al not etched Most Al etched after 7 min

4 6 8

1–40

It is clear that Poly1 was attacked in HF for samples with Au metallization, for the changes of surface color, surface topography as well as the electrical properties. When the releasing time went

For samples with Al metallization, exposure to HF did not change the properties of polysilicon. However, Al could not stand in HF for a long time. Recently, several polymers protective for HF solution were illustrated, which could be applied by spin casting and be easily removed by solvent stripping or plasma ashing. Compared to these proprietary polymers, parylene, which was widely used and deposited by CVD at room-temperature, could be conformally coated on large scale topography and substrates of diverse shapes with various thicknesses [18]. It is very attractive to employ parylene as protection mask for HF wet-etching. Testing chips were first prepared to verify and evaluate the resistance of parylene to concentrated HF. As shown in Fig. 7, 2 lm-thick Al film was patterned on bare silicon, which was covered by parylene film of various edge distances (20–200 lm) and thicknesses (2–8 lm). Parylene type C film was deposited in Parylene Deposition System 2010 and patterned by O2 plasma, the substrates were silanized with A174 to enhance the adhesion before deposition. Then, the chips were put into 40% HF solution at 24 °C for up to 40 min. No film decomposing or lifting were observed during the process. After that, parylene on the samples was ashed by oxygen plasma to verify the integrity of Al. The results were shown in Figs. 8–11 and summarized in Table 4. It was observed that edge distance was not significant when it exceeded 20 lm. However, the film thickness was critical: Al squares were well protected by films of no more than 2 lm thick, while further increasing of the thickness results in severe Al damage even with a short etching time, as shown in Fig. 11. Parylene cannot be resolved in HF, and parylene-silicon interface was well sealed after the release as shown in Fig. 12. However,

Fig. 12. SEM of Al with 2 lm parylene after HF etching.

Y. Zhang et al. / Microelectronic Engineering 97 (2012) 20–25

25

Fig. 13. Optical micrograph of parylene protection after 24 h relaxation. Fig. 15. SEM of the CC-beam resonator after HF wet etching release.

parylene as the protection layer was developed to prevent the polysilicon corrosion. Acknowledgements This work is supported by the National Natural Science Foundation of China (NSFC) Grant No. 60706029 and the National Science and Technology Major Project of China (Project No. 2009ZX02038-02). References Fig. 14. Schematic of parylene protection approach.

many bubbles were observed during the process, as shown in Figs. 8–10. After a relaxation of 24 h, most bubbles had disappeared, as shown in Fig. 13. As shown in Fig. 14, a parylene layer was coated on top of the metal electrode and then the devices were released in HF solution with a chemical resistant protection. Based on the previous experiments, parylene film thickness was critical. When the film thickness was less than 2 lm, Al squares were well protected, however, Al was etched if thickness further increased. The parylene layer was ashed in O2 plasma after release. A parylene layer of 2 lm successfully prevented Al electrodes of the CC-beam resonators from HF acid attacking during releasing (Fig. 15). 5. Conclusion Electrochemical (anodic) corrosion of polysilicon plays a significant role in the surface micromachining. In this study, clamped–clamped beams with a capacitive gap of 100 nm were manufactured to characterize this unusual phenomenon. Changes of the surface morphology were revealed, and the influence on electrical performance of the structure was evaluated, which indicate long time exposure to HF and the presence of Au metallization promote the galvanic corrosion. The release criteria and the origin of this effect were discussed. Finally, a practical method using

[1] Richard S. Muller, Microelectronic Engineering 53 (1–4) (2000) 47–54. [2] E. Marigó, J.L. Muñoz-Gamarra, G. Vidal, J. Giner, F. Torres, A. Uranga, N. Barniol, Microelectronic Engineering 88 (8) (August 2011) 2325–2329. [3] E. Herth, E. Algré, B. Legrand, L. Buchaillot, Microelectronic Engineering 88 (5) (2011) 724–728. [4] T.A. Lober, R.T. Howe, Surface-micromachining processes for electrostatic microactuator fabrication, in: IEEE Solid-State Sensor and Actuator Workshop, Hilton Head, SC, June 6–9, 1988, pp. 59–62. [5] J.A. Walker, K.J. Gabriel, M. Mehregany, Mechanical integrity of polysilicon films exposed to hydrofluoric acid solutions, in: Proceedings from IEEE MEMS, Napa Valley, CA, February 11–14, 1990, pp. 56–60. [6] D. Miller, K. Gall, C. Stoldt, Electrochemical and Solid-State Letters 8 (2005) G223–G226. [7] O.N. Pierron, D.D. Macdonald, C.L. Muhlstein, Applied Physics Letters 86 (2005) 211919. [8] Harold Kahn, Chris Deeb, IEEE Journal of MEMS 14 (5) (2005). [9] I. Chasiotis and W. G. Knauss, The influence of fabrication governed surface conditions on the mechanical strength of thin film materials, in: MRS Symp. Proc., vol. 657, 2001, pp. EE2.2.1–EE2.2.6. [10] E.K. Chan, K. Garikipati, R.W. Dutton, IEEE Design & Test of Computers 16 (1999) 58–65. [11] I. Chasiotis, W. Knauss, Journal of the Mechanics and Physics of Solids 51 (2003) 1533–1550. [12] David C. Miller, William L. Hughes, Zhong-Lin Wang, Ken Gall, Conrad R. Stoldt, Journal of Microelectromechanical Systems 16 (1) (2007) 87–101. [13] X.G. Zhang, S.D. Collins, R.L. Smith, Journal of the Electrochemical Society 136 (5) (1989) 1561–1565. [14] X.G. Zhang, Journal of the Electrochemical Society 138 (12) (1991) 3750–3756. [15] I. Chasiotis, IEEE Transactions on Device and Materials Reliability 4 (2004) 176–188. [16] F. Ronkel, J.W. Schultze, Journal of Porous Materials 7 (2000) 11–16. [17] V. Parkhutik, Solid-State Electronics 43 (1999) 1121–1141. [18] Sung-Won Youn, Hiroshi Goto, Masaharu Takahashi, Shoji Oyama, Yasuhiko Oshinomi, Kinya Matsutani, Ryutaro Maeda, Microelectronic Engineering 85 (1) (2008) 161–167.