Aspects of silicon MEMS cantilever beams micromachining
ASPECTS OF SILICON MEMS CANTILEVER BEAMS MICROMACHINING
Prof.Dr.Eng. Georgeta Ionascu1, Assist.Prof.Dr.Eng. Lucian Bogatu1,
Dr.Physicist Elena Manea2, Dr.Eng. Ileana Cernica2, Eng. Elena Stamata1
1POLITEHNICAUniversity of Bucharest, Precision Engineering Department
2 National Institute for Research & Development in Microtechnology of Bucharest
1
The Romanian Review Precision Mechanics, Optics & Mecatronics, 2008 (18), No. 34
Aspects of Silicon MEMS Cantilever Beams Micromachining
Abstract – A simple method to etch cantilevers beams oriented in the <100> direction on (100) silicon wafers is presented. Back-etching of the wafer, heavily doped boron etch stop, or anodic oxidation etch stop are not necessary. The method requires only two levels of masking and uses silicon dioxide as passivation material. It can be adopted to produce micromachined solid-state pressure sensors, accelerometers, resonators and electrodes for recording of biopotentials.
Key words - bulk micromachining, etching, silicon cantilever
1. Introduction
Micromachined silicon cantilever beams are the most usual structure in the field of microelectromecha-nical systems (MEMS) [1].
MEMS cantilevers are commonly fabricated from silicon (Si), silicon nitride (SiN) or polymers. Without cantilever transducers, atomic force microscopy would not be possible. A large number of research groups are attempting to develop cantilever arrays as biosensor for medical diagnostic applications. MEMS cantilevers are, also, finding application as radio frequency filters and resonators. The principal advantage of MEMS cantilevers is their cheap and easy fabrication in large arrays [2].
The fabrication process typically involves undercutting the cantilever structure to release it, often with an anisotropic wet etching technique. The most direct one is to etch from both the front side and the back side of the wafer until cantilevers are released. This method requires special tools and procedures for the front-to-back masks alignment. Etching through the entire thickness of the wafer will reduce its mechanical strength.
The second method requires doping the beam with a boron concentration greater then 5·1019 cm-3, which is close to the solid solubility of boron in silicon. The heavily doped region will resist at etching in EDP (ethylene diamine pyrocatechol) and KOH (potassium hydroxide), but the resulted silicon cantilever will have a large lattice strain owing to the high concentration of boron. Microelectronics cannot be integrated onto the same substrate with the cantilever.
The third method is an electromechanically controlled etching, known as anodic oxidation etch stop method. It requires first doping the desired beam areas to form a p-n junction with the starting substrate. The p-n junction is reversed-biased during etching. The exposed p-type material will be etched away, but the n-type material will be left intact. The thickness of the beam is determined by the p-n junction depth. Long diffusion time is needed to obtain a deep junction, or a thick layer of silicon has to be grown by epitaxy.
This paper presents a simple method to etch undoped silicon cantilever beams, oriented in the <100> direction (at 45° compared to the normal orientation) on (100) silicon wafers.
The process requires two levels of masking (a two-step etching process) to release the cantilevers. The anisotropic etching is performed in KOH solution.
Silicon cantilevers with the same background doping level as that of the starting wafer and vertical edges are produced.
2. Technological process of micromachining
A layer of silicon dioxide 0.5 μm thick is grown on a (100) wafer. The wafer is spin-coated with a layer of positive photoresist.
The necessary masks to fabricate the cantilevers are shown in Figures 1 (a) and 1 (b). Crosshatched areas represent opaque regions of the masks. The two masks have the same pattern excepting the beam width on the mask 2, which is smaller than the beam width from the mask 1. The relationship between feature dimensions on mask 1 and mask 2 is derived below.
First, the wafer is patterned with mask 1. The wafer is oriented in such a way that the length of the cantilever beam is in the <100> or <010> direction of the wafer, as indicated in Figure 1(c). The wafer is then immersed in a bath of buffered oxide etch (BOE: HF-NH4F, 1:6, 32°C, 0.1 μm/min) to remove the silicon dioxide in the areas that are not covered by photoresist, which is followed by dissolving the photoresist in an acetone bath. Fig. 2(a) shows a cross section of the beam region after the photoresist has been removed. The wafer is now prepared to be etched in a KOH 20% solution at 80°C (1.4 μm/min).
Downward etching will take place in the unprotected areas, where the {100} planes of silicon are exposed. Lateral etching of silicon directly underneath the silicon dioxide masking layer will also occur. The lateral planes being etched are the {100} equivalent planes. These planes are normal to the substrate. The rate of downward etching is expected to be the same as the rate of the lateral etching. This behaviour will result in walls that are almost or completely vertical [3].
Fig. 2(b) shows the structure at the end of the first KOH etching. {111} planes are formed at the clamping region of the cantilever beam.
Fig. 1 Geometry and orientation of the masks: (a) – mask 1; (b) - mask 2; (c) – the length direction of the beam is aligned with either the <100> or the <010> direction of the (100) silicon wafer.
After the anisotropic etching in KOH, the wafer is etched in BOE to remove all the silicon dioxide and then cleaned and oxidized, again, to obtain a new layer of silicon dioxide 1 μm thick, as shown in Fig. 2(c). The wafer is spin-coated with a layer of thick positive photoresist (thick enough and planar photoresist, which must embed the previous structures), patterned with mask 2.
After the unprotected silicon dioxide is etched away in BOE, the photoresist is removed in acetone. The wafer is then etched in KOH at 80°C, until bulk silicon is completely underetched in the areas directly underneath the beam. Fig. 2(d) shows the wafer at the end of the second etching in KOH and the obtained cantilever beam.
Fig. 2Obtaining the silicon cantilever beams with two level of masking (a two-step etching process): (a) – wafer prior to the first etching in KOH (having the silicon dioxide masking layer patterned with mask 1); (b) – wafer at the end of the first etching in KOH; (c) – wafer coated with a fresh layer of silicon dioxide, which will be patterned with mask 2; (d) – wafer at the end of the second etching in KOH and releasing the cantilever beams.
3. Dimensions of mask features
In order to design a silicon cantilever beam of width wb, thickness tb and length lb, we need for mask 1:
(1)
(2)
where w1 and l1 are, respectively, the width and the length of the beam pattern on mask 1.
During the first KOH etching, the wafer will be etched downwards in the unprotected areas for a depth of tb+(wb/2) μm. The same amount will be etched away laterally underneath the beam area, as illustrated in Fig. 2(b).
To release the silicon cantilever in the second KOH etching, we need for mask 2:
(3)
(4)
where w2 and l2 are, respectively, the width and the length of the beam pattern on mask 2; d, a quantity smaller than wb, is the tolerance for mask misregistration and misalignment.
Eqs. (1) and (3) represent the lowest limits of the feature dimensions. They involve the shortest etching time. From (3), the interspace between the back side of the beam and the bottom surface of the wafer is calculated to be d + wb. Larger interspaces can be obtained with higher values of d.
4. Experimental results and conclusions
Following the procedures from above, more experiments, with normal orientation and inclining the wafer at 45° relative to the mask, were performed, as results from Figures 3a and 3b.
(a)
(b)
Fig. 3SEM images of the cantilevers after removing the passivation layer of silicon dioxide: (a) with normal orientation; (b) – inclined at 45°.
The duration of the first KOH etching is crucial, as extent of the etching will determine the width and the thickness of the beam. Overetching in the first etch stage will result in a narrower and thicker beam. The duration of the second etching will affect the thickness of the beam as well as the interspace between the beam and the substrate. A longer second etching can trim down the extra beam thickness caused by overetching in the first etching step; however, it will not compensate for the shrinkage in the width.
Cantilever beams micromachined at 40 μm, respectively 30 μm depth are presented in Figures 4 and 5. By reducing the etching depth, a beam of larger width and smaller thickness results.
Fig. 4SEM photo of a cantilever beam micromachined at 40 μm depth.
Fig. 5SEM photo of a cantilever beam micromachined at 30 μm depth.
A technology for a planar photoresist layer at the second masking level must be available.
The beams are shorter than expected and have sharp-pointed free ends. The needle-shaped free ends are formed by the much faster etching of planes of higher Miller indices originating from the protruding corners.
In the SEM image of Fig. 4 we can observe these planes of higher Miller indices at the free end of the vertical wall. Their etch rates are found to be almost two times those of the {100} planes. Consequently, a compensation for the length contraction is needed when designing the masks set. Compensation can be achieved with the following relationship:
(5)
where l1 and l2 are the length dimensions of the beams on mask 1 and 2 respectively, lb is the desired final silicon beam length, and (wb/2) +tb is the amount of etching in the first KOH etching.
The above method can be used to fabricate silicon cantilevers as long as possible; the beam width, wb, beam thickness, tb, and mask misalignment tolerance, d, satisfy eqs. (1), (3) and also:
(6)
where tSi is the thickness of the silicon wafer.
The silicon cantilever beams fabricated in this way have the same background doping level as that of the starting wafer, which allows microelectronics to be integrated on the cantilevers. A nearly rectangular in-plane cross section is obtained. Etching is performed on one side of the wafer only and does not require front to back mask alignments. By modifying the mask patterns slightly, clamped-clamped beams or bridges can, also, be fabricated.
References
[1]Ionaşcu G.,Technologies of Microtechnics for MEMS (in Romanian), “Cartea Universitară” Publishing House, Bucharest, 2004.
[2]Ionascu G., Bogatu L., Sandu A., Manea E., Cernica I.,Modeling, Simulation and Technology of Microbeam Structures for Microsensor Applications, in U.P.B. Scientific Bulletin, Series D, Vol. 70, No. 3, p. 19-30, 2008.
[3]Choi W., Smits J., A Method to Etch Undoped Silicon Cantilever Beams, in Journal of Microelectromechanical Systems, vol 2, no. 2, p. 82-85, 1993.
1
The Romanian Review Precision Mechanics, Optics & Mecatronics, 2008 (18), No. 34