Low Frequency Interdigital Capacitive Sensor using Deep Reactive Ion Etching and Bulk Micromachining of Silicon Mahdiyeh Mehran

Shamsodin Mohajerzade

Nano-Electronic and Thin Film Lab University of Tehran Tehran, IRAN [email protected]

Nano-Electronic and Thin Film Lab University of Tehran Tehran, IRAN [email protected]

structures at micro-scale level. Before the invention of deep reactive ion etching (DRIE for short) by Bosch in 1995, most of the fabrication was based on surface machining, a process capable of creating layered structures with excellent selectivity [11, 12]. For such processes, polysilicon films with thicknesses from tenth of micrometer to a few micrometers are used and the employment of a sacrificial layer and certain heat treatment steps complete the fabrication. By the evolution of high precision vertical etching techniques either by Bosch or by a cryogenic plasma method, the realization of deep and vertical interdigital structures has become possible where the problems and limitations of surface machining tool has been surmounted. The evolution of complex finger-type structures needs plasma-based reactive ion etching with either a sequential etching/passivation technique or a cryogenic assisted high aspect ratio etching. Aspect ratios as high as 100 and feature sizes as small of 0.2 micrometer can be realized with adjusting the etching parameters. In this paper a bulk micromachined accelerometer is reported using our recently developed hydrogen assisted deep reactive ion etching method. A combination of anisotropic etching of (100) silicon substrates from back-side of the substrates is used along with deep etching of interditigal features from the front-side. By proper controlling of the etching parameters one can achieve features below one micrometer and aspect ratios around 50. Also the evolution of grass or black silicon which is an adverse side-effect of reactive ion etching can be controlled using the hydrogenation step. Formation of the cavity underneath the interdigital structure besides grass insertion on the back side of it allows inclusion of liquid to add to the effective mass of the device and to realize a low frequency sensing device suitable for earthquake prediction.

Abstract— A novel fabrication procedure for implementation of suspended, interdigital capacitive sensor on silicon-based membranes is reported. The process employs a combination of anisotropic back-side micromachining with front-side vertical deep reactive ion etching of silicon. The formation of cavities underneath the interdigital structure beside the evolution of nano-grass on the back side of the sample, allows inclusion of liquid to add to the effective mass of the device and to realize a low frequency sensing device suitable for earthquake prediction. The etching of vertical pieces is feasible using a hydrogen-assisted reactive ion etching. Also the evolution of grass on the back and front sides of the sample can be controlled by etching parameters. Keywords-component; accelerometer, Mechanical resonant frequency, DRIE, black silicon.

I.

INTRODUCTION

There are increasing demand toward small size, light weight and low power sensing systems in all application domains such as micro accelerometers, actuators and pressure sensors. Accelerometers have a widespread range of applications such as inertial navigation and guidance, seismology, and space microgravity measurements [1-3]. Also the fabrication of nock sensors, air-bag activation sensing elements and even microphones is another serious line of research for such devices. Due to its high demand, the research to fabricate accelerometers has reached a maturity in terms of the process and the materials used. Usually accelerometers are classified based on the form of transduction mechanism which are capacitive piezoelectric, piezoresistive, thermal, electron tunnelling and Bragg grating methods. The capacitive form is one of most widely techniques which is fully compatible with MEMS-based technologies [4-8]. It is because of their high sensitivity, good dc response and noise performance, low drift, low temperature sensitivity, low power dissipation and simple structure [9, 10]. One of the main methods for the fabricating capacitive accelerometers is based on bulk-micromachining technology of silicon substrates with (100) orientation [8, 9]. The micromachining is a key element in realizing sophisticated

II.

FABRICATION PROCEDURE

The fabrication process of these sensors depends on both back-side anisotropic and front-side reactive ion etching steps. Since the evolution of deep vertical features is not possible with wet etching processes, the reactive ion etching technique

Proceedings of ICEE 2010, May 11-13, 2010 978-1-4244-6761-7/10/$26.00 ©2010 IEEE

Pr o c e e di ngo fI CEE2 0 1 0 , I s f a ha nUni v e r s i t yo fTe c hno l o g y , 1 1 1 3Ma y2 0 1 0 450

used in this paper is described in more detail prior to the fabrication procedure of the sensor structure. In the following section, first the reactive ion etching, assisted by a hydrogenation step is described. The formation of vertical features and the creation of black silicon are described and a method to manipulate the etching process is explained. In the following part the application of such technique to realize interdigital sensor is presented. A. Hydrogen-assisted Deep Reactive Ion Etching (HDRIE) Deep reactive ion etching (RIE) of silicon has become a key technology in the fabrication of deep vertical structures for bulk micromachining. The most frequently used technique is a process developed by Robert Bosch GmbH [13]. An alternative process, which relies on cooling the silicon substrates to cryogenic temperatures using liquid nitrogen, is also being exploited [14, 15]. In an attempt to turn around the use of polymeric coating (e.g. C4F4) during the sequential etching steps in Bosch process, we have used hydrogen as recently reported in Ref [16]. Due to the importance of this part we explain it more in detail. Essentially this technique employs an RF-plasma (13.56 MHz) at two (or three) successive steps of passivation and etching sub cycles to perform the etching process. A mixture of H2/O2 gases with typical flows of 100 and 85sccm and a trace value of SF6 is used during the passivation step while the etching step is achieved using SF6 as the inlet gas with a typical flow of 30 sccm. The plasma power is usually set at 150 and 130W with the time of 50 and 10 seconds for the two subcycles of passivation/etching, respectively. The incorporation of O2/H2 and SF6 during the passivation step leads to the formation of a protecting layer over the side-walls of the vertical craters while H+-ion bombardment helps to remove this protective layer from the very bottom of the crater. However grasses are one of the side effects of DRIE, our vertical etching process can be adjusted to obtain grass-free structures both at micro and nano-scale with aspect ratios of the order 50.

Figure 1. The SEM images of some vertically etched samples. (a) Rods with 0.6 μm width and 10μm height are realized. (b) An array of silicon walls with the same height and spacing of 5μm. (c) , (d) arrayal formation of high aspect ratio features on silicon substrates.

By increasing the flow of SF6 from 30 to 50 sccm gradually, one can remove the silicon underneath such defect sites before a serious unwanted structure is formed which eventually turns out as grass. The creation of deep trenches is then feasible using this adjustable etching technique which in part (d) an SEM image of deep trenches can be seen. “Fig. 2” shows the SEM images of a grass-full structure on silicon substrate. Grass formation is known to be an unwanted effect in all RIE-based processes however, in this technique and by using proper flow of gases during the passivation and etching subcycles one can manipulate such Si grass formation. A heavy coverage of the etched surface by nano-grass throughout the sample as a result of high hydrogen energy is observed in the “Fig. 2”.

Using this method various features are obtained on the Si substrate with sizes down to 100nm in width. The etch rate can be varied between 0.1 to 0.5 μm/min. “Fig.1” and “Fig. 2” collect several SEM images of the samples prepared for this investigation. As seen from these images, it is possible to obtain grass-free features with desired shapes and heights. Part (a) in “Fig. 1” shows vertically etched silicon columns with 0.6μm width and a height of 10μm without a serious under etching. Also in part (b) one can see several lines of silicon with a height of 10μm and spacing of 5 μm. The surface of the etched silicon is free of grass or black-silicon. Part (c) shows the SEM image of arrays of silicon micro-scale rods with over 10μm height and below one μm width and aspect ratios more than 40. The spacing could be 1 μm. To obtain a grass-free surface without losing the verticality one has to increase the SF6 flow during the etching sub-cycle. We believe the grass formation in our technique is mainly due to the residual passivation coatings at the very bottom of the etched craters.

Figure 2. SEM images of improper vertical etching. A heavy coverage of the etched surface by nano-grass throughout the sample as a result of high hydrogen energy.

Pr o c e e di ngo fI CEE2 0 1 0 , I s f a ha nUni v e r s i t yo fTe c hno l o g y , 1 1 1 3Ma y2 0 1 0 451

B. Sensor Fabrication Process

It is worth mentioning that the originally deposited Chromium layer which acted as Final removal of such a layer is possible by a lift-off process where the underneath SiO2 oxide is removed in an HF-solution without adversely affecting any other part of the sensor. At last the sample is metal-coated and wire-bonded to allow measurements. Table I shows the geometrical specifications of the sensor. In the structure of sensor, S-shape springs are used which are more flexible. As the spring constant (K) is related to its geometry, shape and consisting material, we use S-shape spring to have smaller values of K. This smaller K tends to the reduction of the mechanical resonant frequency which is a goal in the evolution of this sensor. Capacitors in this structure are in differential form. This form omits the nonlinearity in the capacitor changing with respect to the displacement. “Fig. 4” presents the SEM images of the sensor. In the insets in this image the structures of the spring as well as blades have been highlighted to observe the verticality and well-definition of such components.

“Fig. 3” shows schematic of different fabrication steps for the final sensor. The process begins with growing a thermal oxide (0.3μm) on a low resistivity ( = 0.01  · cm) 500 μm p-type wafer, then a 250 nm LPCVD Si3N4 layer deposited on it which is used as KOH mask in the latter step. A 50nm Ebeam chromium layer is deposited on the backside of the wafer which is used as a mask for patterning Si3N4. These layers patterned using lithography with subsequent etching steps which tends to the formation of openings for generating membranes. Membranes are formed using 8 molar KOH solution in 60oC (with 18μm per hour, etch rate) with 30μm thickness. Then after removal of the remaining Si3N4 by the use of under etching SiO2 sacrificial layer, phosphorous doping is performed. Next a thick thermal oxide (0.3 μm) is grown which is followed with deposition of a 50nm E-beam chromium layer on the front side of wafer. The front-side mask has features of the order of 3μm and it is aligned with the previously created cavities from the back side using standard photolithography. Once the Cr layer on top of the sample is patterned, the sensor realization and evolution of vertical features is achieved using our recently developed H2-assisted sequential passivation/etching process in a RIE system. Once deep and vertical features are obtained, the sample is placed in a phosphorous diffusion furnace to obtain adequate conductive capacitor-electrodes. To avoid shortcircuiting the opposite electrodes the vertical etching is continued for removal of nearly 1 to 2 μm of silicon from the bottom surface of unsuspended parts.

TABLE I GEOMETRICAL SPECIFICATIONS OF THE FABRICATED INTERDIGITAL SENSOR.

Spring

Height : below 30 μm Length : 1mm Proof Mass

Length : 350

(μm)

(c)

(e)

(g)

Width : 10 Height : 30 Capacitor plates overlap : 300 Plates space in differential structure : 5 & 10

(b)

Phosphorous doping

Width : 20 μm Height :30 μm

Comb-drive Capacitors

(a)

Length : 1.3mm Width : 20 μm

(d)

(f)

(h)

Figure 3. The process flow for the fabrication of the interdigited structure. (a) oxide growth with LPCVD Si3N4 deposition and chromium layer at back side, (b) opening windows through thermal oxide/ Si3n4 /Cr layer, (c) Membrane formation using 8 molar KOH solution in 60c, (d) removing oxide/Si3n4 /Cr layers, (e) Phosphorous doping, (f) opening windows through thermal oxide/Cr layer, (g) HDRIE process for generating the structures of sensor, (h) removal of oxide and Cr layers.

Figure 4. The SEM image of the fabricated sensor. Arrows highlights the spring and he finger-type capacitor electrodes. The inner solid square is the effective “mass” of the device and the region where the interaction with liquid is more dominant.

Pr o c e e di ngo fI CEE2 0 1 0 , I s f a ha nUni v e r s i t yo fTe c hno l o g y , 1 1 1 3Ma y2 0 1 0 452

This means that we could introduce them whenever are wanted or omit them from the DRIE process. After grass insertion on the back side of the sensor as is showed in “Fig. 6”, the mechanical resonant frequency reduced to 2.25KHZ; however the water inclusion in the backside cavity of the sensor tends to the reduction of this frequency to 1.3KHZ. At last we fill the back side cavity with oil, this let us to achieve a low frequency of 100HZ which means that such a structure is capable of being used in earthquake prediction. V. Figure 5. Optical microscope image of sensor when it is deformed due to being wetted with water droplets.

We have demonstrated an effective and inexpensive method for the fabrication of a low frequency capacitive accelerometer on (100) silicon membrane. After creating membrane with the use of anisotropic etching of silicon in KOH solution, the featured structures are obtained by hydrogenation-assisted vertical etching of silicon membranes and substrates. Presence of cavity allows exploiting of liquid which helps reaching to smaller resonant frequencies with higher sensitivity. Further results of testing this structure are ongoing.

To better observe the full suspension of the structure, we have wetted the sample with droplets of water then placed it under an optical microscope which is capable of image recording. “Fig. 5” reveals the attachment of neighboring blades with spring deformation, moreover in the lower image of this figure one can observe highly deformed springs on either sides of the suspended inner part. This highly deformation of the springs shows the flexibility beside the rigidity of the sensor. III.

SUMMARY AND CONCLUSION

REFERENCES [1]

RESULTS AND DISCUSSIONS

An important parameter of such suspended devices is their mechanical resonance frequency. To measure the mechanical resonant frequency of the sensor we use Mickelson interferometer. We obtained 3KHZ as the mechanical resonant frequency of this sensor according to this method. Although this value is not low enough for many low frequency applications, it can be lowered by further thinning the spring parts or by coating the top inner-part. In this paper instead of using conventional methods to lower the resonance frequency, we prefer to keep this parameter at this value and incorporate liquid in the cavity underneath the sensing part. Using this method the frequency of measurement is well below the natural resonance frequency of the suspended part. For the improvement of the silicon adhesion to the liquids such as water and oil, we apply nano structures to the backside of the proof mass. These nano structures which are so-called grass or black silicon, give the hydrophilic besides oleophilic property to the silicon. As explained before, grass or black silicon is one of the side effects of DRIE which we have controlled them well.

[2]

[3]

[4]

[5]

[6]

[7] [8]

[9] [10]

[11]

[12] [13] Figure 6. SEM image of the sensor from back side. Evolution of grass on the proof mass can be observed in the inset.

Z. Syed, P. Aggarwal, X. Niu, N. El-Sheimy, Economical and robust inertial sensor configuration for a portable navigation system, International Technical Meeting of the Satellite Division, vol. 2, 2007, pp. 2129–2135. D. Mougenot, N. Thorburn, “MEMS-based 3C accelerometers for land seismic acquisition: Is it time?,” Leading Edge (Tulsa, OK), vol. 23, n 3, Mar. 2004, pp. 246–250. P. Monajemi, F. Ayazi, “Design optimization and implementation of a microgravity capacitive HARPSS accelerometer,” IEEE Sensors, vol. 6, n 1, Feb. 2006, pp. 39–46. S. Rajgopal, D. Zula, S. Garverick, M. Mehregany, “A silicon carbide accelerometer for extreme environment applications,” Materials Science Forum, vol. 600, 2009, pp 859–862. C. Xue, S. Chen, H. Qiao, W. Zhang, J. Xiong, B. Zhang, G. Zhang, “Development of a novel two axis piezoresistive micro accelerometer based on silicon,” Sensor Letters, vol. 6, n 1, Feb. 2008, pp. 149–158. M. Miao,Q. Hu, Y. Hao, H. Dong, L. Wang, Y. Shi, S. Shen, “A bulk micromachined Si-on-glass tunneling accelerometer with out-of-plane sensing capability,” Proc. IEEE NEMS 2007, pp. 235–240. H.Y. Au, S.k. Khijwania, H.Y. Tam,” Fiber Bragg grating based accelerometer,” Proc. Optical Fibre Sensors, vol. 7004, 2008. L. Dong, J. Chen, H. Yan, W. Huo, Y. Li, L. Sun, “A novel capacitive micro-accelerometer with grid strip capacitances and sensing gap alterable capacitances,” Semiconductors, vol. 30, n 3, Mar. 2009, pp. 0340081–0340086. L.K. Baxter, “Capacitive Sensors,” IEEE Press, Piscataway, N.J, 1997. B. Bais and B.Y. Majlis, “Analysis of the Mechanical Behavior of a MEMS Areachanged Lateral Capacitive Accelerometer,” Proc. of National Symposium on Microelectronics, Malaysia, 2003. W. Yun, R.T. Howe and P.R. Gray, “Surface micromachined, digitally force-balanced accelerometer with integrated CMOS detection circuitry,” Tech. Digest IEEE Solid-State Sensors and Actuator Workshop, June 1992, pp. 21–25. Airbags boom when IC accelerometer sees 50G, Electronic Design, 8 Aug. 1991. F. Laermer and A. Schilp, “Method of anisotropically etching silicon,” U.S. Patent 5 501 893, Mar. 26, 1996.

Pr o c e e di ngo fI CEE2 0 1 0 , I s f a ha nUni v e r s i t yo fTe c hno l o g y , 1 1 1 3Ma y2 0 1 0 453

[15] K. Murakami, Y. Wakabayashi, K. Minami, and M. Esashi, “Cryogenic dry etching for high aspect ratio microstructures,” in Proc. IEEE MEMS Conf., Feb. 1993. [16] S. Mohajerzadeh et. al., “Deep vertical etching of Si wafers using a hydrogenation-assisted reactive ion etching,” IEEE JMEMS, Vol. 16, no. 4, 2007, pp. 912–918.

[14] M. J. de Boer, J. G. E. Gardeniers, H. V. Jansen, E. Smulders, M. J. Gilde, G. Roelofs, J. N. Sasserath, and M. Elwenspoek, “Guidelines for etching siliconMEMS structures using fluorine high-density plasmas at cryogenic temperatures,” J. Microelectromech. Syst., vol. 11, no. 4, Aug. 2002, pp. 385–401.

Pr o c e e di ngo fI CEE2 0 1 0 , I s f a ha nUni v e r s i t yo fTe c hno l o g y , 1 1 1 3Ma y2 0 1 0 454