In this section, the problems encountered during fabrication and solutions are discussed.
(1) Over etch
The first SiO2 (anchor) layer is etched by poly RIE. Multiple etches are used to control the etching depth. The residual oxide thickness is etched by observing the color of the oxide layer using an optical microscope. Figure 3-3 shows a case where the oxide layer is overetched.
Overetch
Figure 3-3: Overetch (2) Annealing
The anneal process is conducted after all poly layers are deposited. Figure 3-4 is the poly structure without annealing. The plate is curved. Figure 3-5 is another poly structure with annealing. The structure is flat.
Figure 3-4: Without annealing
Figure 3-5: With annealing (3) Gold adhesion
Another problem in the process is the adhesion of gold. At first, the wet etching using 10%KI + 5%I2 + 85% H2O is used to pattern gold. Figure 3-6 shows that the gold layer peels off after wet etching. On the other hand, Figure 3-7 shows a gold layer patterned by lift-off. The gold layer sticks with the poly structure layer well.
Figure 3-6: Wet etching patterned gold layer peels off.
Figure 3-7: Lift-off patterned gold layer sticks well.
(4) Thermal stress
Thermal stress is produced when pushing wafers into and pulling wafers out of the furnace. It can make the structure crack. Figure 3-8 shows the fissures that are produced by the thermal stress in the plate when the speed of pushing the wafer into the furnace is too fast. The problem can be avoided by pushing the wafer slowly.
Fissures
Figure 3-8: Plate with fissures (5) B.O.E. dipping
The dipping process between deposition processes is very important to remove
the SiO2 on the top layer. In Figure 3-9, the 5000Å chromium (Cr) layer peeled off from the poly structure layer after releasing. The dipping time in Figure 3-9 is 1 minute in HF : H2O = 1 : 100. However, it was observed that after dipping Poly2 in B.O.E. for 3 minutes, the metal layer – 3000Å gold + 100Å chromium (Cr) bends the beam and still sticks to the beam perfectly.
Figure 3-9: The metal layer peels off.
○6 Releasing time
In Figure 3-10, the SiO2 is not cleaned completely because of short releasing time (about 30 minutes). If the releasing time is long enough, the SiO2 is cleared after etching by B.O.E. for 2 hours, as shown in Figure 3-11.
SiO2
Figure 3-10: Releasing time is not enough.
Without SiO2
Figure 3-11: Releasing time is enough.
CHAPTER 4
Measurement
There are three main parts in this thesis. One is the plate self-assembled by residual stress beams, one is the Fresnel lens, and the other is the combination of those two parts to build a vertical Fresnel lens. In this chapter, the results of the three parts are discussed. In Section 4-1, two MUMPS samples are shown. In Section 4-2, a Fresnel lens experiment is shown. And finally, the result of the vertical Fresnel lenses is in Section 4-3.
4-1 MUMPs
Two Multi-User MEMS Processes (MUMPs) runs using gold-polysilicon stress beams were tested and the results are discussed.
4-1.1 First MUMPs Run
Figure 4-1 shows the scanning electron microscopy (SEM) pictures of a device from the first MUMPs run sample. This figure is cut the unnecessary part. The lens plate can not be raised by stress induced beams successfully.
Figure 4-1: First MUMPs run result
Figure 4-2: Magnified picture of a hinge
The cause of the failure to raise the lens plate is that the hinge bar is too wide to rotate in the staple, based on the MUMPs layout rules. As shown in Figure 4-2, the space under the staple for the hinge to rotate is too small.
Height
= 4.75μm Width = 5μm
Figure 4-3: Layout and profile of the hinges
Figure 4-3 is the layout and profile of the hinges. The width of the hinge bar is 5µm, but the height of the staple is 4.75µm [(First Oxide=2.0µm) + (Poly1=2.0µm) + (Second Oxide=0.75µm) = 4.75µm]. The width of the hinge bar is wider than the inner space of the staple so the space for hinge rotating is not enough. The solution is to reduce the width of the hinge bar.
Height (Tip to substrate)
Figure 4-4: Device1 measurement result
There are two devices in the first run samples. In device1, the plate is 400×450 µm2 and the beams are 500×300 µm2. Figure 4-4 is the device1 measurement result by a WYKO interferometer. From the Y profile, the deflection is not due to the curvature of the plate caused by the residual stress. The deflection is the result from that the plate is flipped up by stress induced beams. The tip displacement of the 450 µm long plate composed of two structure layers (Poly1 and Poly2) is 55.0µm and the tip deflection angle is 7ْ .
In device2, the plate is 400×550 µm2 and the beams are 800×400 µm2. Figure 4-5 is the device2 measurement result. From the X profile, the deflection is also caused of by the flip-up by stress induced beams. The tip displacement of the 550µm long plate composed of single structure layer (Poly2) is 55.6µm and the tip deflection angle is 5.8ْ .
Height (Tip to substrate)
Figure 4-5: Device2 measurement result
Figure 4-6 is the tip displacement of the test cantilever beams in this MUMPS die. From the run data in the MEMSCAP website, the thickness of Poly1 is about 2.0µm, the compressive stress in poly1 is 13MPa. The thickness of Poly2 is about 1.5 µm, and the compressive stress in poly2 is 18MPa. The thickness of gold is about 0.55µm, and the tensile stress in gold is 33MPa. The calculated curve of the cantilever beams based on the run data is also shown in Figure 4-6. The calculation result based on the run data is quite different from the experiment result. The other variable
parameter, for example, Young’s modulus is needed to be measured and considered.
Figure 4-6: Tip displacement of test beams in the first MUMPs run sample In these two devices, the angle of device1 (length = 450µm) is larger than the angle of device2 (length = 550µm). From Chaper 2, a longer beam should have a larger bending height and tip deflection angle. From the measurement results, the longer stress beams do not raise the plate higher. The cause is that the longer beam is softer and has less actuation force.
4-1.2 Second MUMPs Run
The second MUMPs sample is a modified version of the first one. The width of the hinge bar is reduced to 2µm. It is successful to assemble the plate to approach the vertical position by the residual stress beams. Two devices are tested in this sample.
z Device1
Figure 4-7: Vertical device1
Figure 4-8: Side view of the vertical device1
Figure 4-7 is the SEM picture of the device whose residual stress beam length and width are 500µm and 300µm. The residual stress beam is composed of 1.5µm thickness Poly2 and 0.5µm thickness Gold. The length and width of the lens plate are 450µm and 400µm. The plate is composed of Poly1 and Poly2. This device is similar
to the first MUMPS version but the width of the hinge bar is reduced to 2µm and the shape of the beam tip is changed to a V-shape.
Figure 4-8 is the side view of the SEM picture in Figure 4-10. The measured angle from Figure 4-8 by a protractor is about 90ْ .
Figure 4-9: Curvature of the residual stress beam with the vertical plate Plate
Figure 4-10: Curvature of the residual stress beam without vertical plate
The curvature of the stress induced beam of this device is measured by WYKO, as shown in Figure 4-9. The end deflection perpendicular to the unreleased position is 43.0µm. The black area in front of the beam tip is the vertical plate.
Figure 4-10 is the beam deflection perpendicular to the unreleased position without the plate. The end deflection height without the plate is 51.9µm. The height of the end deflection is larger than the beam with the plate. This result is caused by the fact that the plate touches and presses the residual stress beam, as shown in Figure 4-11.
Figure 4-11: Magnified picture of beam and plate
Because there is no fixing mechanism in the design, the plate sways easily. The final angle of the plate is decided by the residual stress beams and the location of the hinge bar. Figures 4-12 shows the hinges in this vertical device. The 2µm wide hinge bar is located in the position that makes the plate vertical and fixed in the staple. If the hinge bar is not fixed, a swaying plate appears. Figure 4-13 is a swaying device and the rotation angle of the plate is more than 90ْ .
Figure 4-12: An overview of the hinge group
Figure 4-13: More than 90ْ device1
Next, the minimum rotation space for the hinge bar is discussed. Figure 4-14(a)
and (b) are the layout and side view of the hinge bar. The block A in Figure 4-14(b) is the rotation space of the hinge bar. The diagonal line (j) that is from the Poly2 layer to the Poly0 layer is the minimum.
Figure 4-14(a): The hinge layout in this version
38.2 ْP
Block A
Figure 4-14(b): Hinge side view
In Figure 4-14(b), the angle of the diagonal from the Poly2 layer to the Poly0 layer in the staple is 38.2ْ . The i is the height of the staple space, j is the 38.2ْ
diagonal of the staple space, k is the 38.2ْ diagonal of Oxide1, l is the 38.2ْ diagonal of Oxide2, m is the 38.2ْ diagonal of Poly0, and n is the 38.2ْ diagonal of the hinge bar.
The height of staple space in this design is 2.75µm (2µm thickness Oxide1 + 0.75µm thickness Oxide2 = 2.75µm), as shown in Figure 4-17(b), and the length of the 38.2ْ diagonal of the space in the staple is 3.6µm {[The 38.2ْ diagonal of Oxide1
= 3.2µm (2µm csc38.2ْ = 3.2µm)] + [The 38.2ْ diagonal of Oxide2 = 1.2µm
= 2.8µm). The 38.2ْ diagonal of the staple space is larger than the 45ْ diagonal of the hinge bar. The redundant length is 0.8µm [(The 38.2ْ diagonal of the staple space = 3.6µm) – (The 45ْ diagonal of the hinge bar = 2.8µm) = 0.8µm]. Because the 38.2ْ
diagonal of the staple space is larger than the 45ْ diagonal of the hinge bar, the plate can be rotated.
If the layout of the block of Poly0 is changed, this space can be larger. If the location of the left Poly0 block is moved away from underneath the hinge bar, it will provide more space to rotate. The length of the 38.2ْ diagonal of the space will be 4.4µm {[The 38.2ْ diagonal of Oxide1 = 3.2µm (2µm× csc38.2ْ = 3.2µm)] + [The 38.2ْ diagonal of Oxide2 = 1.2µm (0.75µm csc38.2ْ = 1.2µm)] = 4.4µm}. The redundant space is 1.6µm.
×
z Device2
Figure 4-15 is the SEM picture of a device with length and width of 800µm and 400µm, respectively. This device is readily assembled after releasing without CO2
drying. The stress beams and all the other components in this die are stuck on the substrate. This situation will be discussed in Section 4-1.3.
Figure 4-16 is the side view SEM picture of Figure 4-20. The measured angle by a protractor is about 90.5ْ .
Figure 4-15: Vertical device2
Figure 4-16: Side view of vertical device2
Figure 4-17 is the SEM picture of one hinge in this device. The hinge bar is about 1µm thick and slips into the space between staple and Poly0 and then be lodged in the space. The lodging in the space between staple and Poly0 fixes the angle and position
of the plate. The end deflection of this 800µm×400µm beam is 140µm, as shown in Figure 4-18.
Figure 4-17: Magnified picture of one hinge in this hinge group
Figure 4-18: The end deflection of the beam in the device2 z Test cantilever beams
There are test patterns in these dies to measure the residual stress. The type of the
stress in all gold stress beams is tensile. The radius of curvature of test cantilever beams is about 0.84mm. The curves of the test beams are plotted in Figure 4-19. The run data of MUMPs is that the thickness of Poly2 is about 1.58 µm, the thickness of gold is about 0.54µm, Poly2 has an 11MPa compressive stress and gold has a 24MPa tensile stress. From the curves in Figure 4-19, the measured data of all beams are quite different from the curve of MUMPs run data. The radius of curvature of the run data is about 7.5mm. The other variable parameter, for example, Young’s modulus is needed to be measured and considered.
Figure 4-19: The curves of every beam in the second MUMPs run 4-1.3 Discussions
(1) Fixing mechanism
Figure 4-20 is the magnified picture between plate and beam. From Figure 4-20, the distance between plate and beam is too wide to fix the plate in a fixed position. To
close plate and beam can improve the unstable problem. The funnel shape can also be used to improve this problem.
Figure 4-20: Distance between plate and beam
(2) Seized by staple
Some devices are not flipped up to the vertical position because the hinge bar is seized in the staple too tightly to rotate the plate by residual stress beams, as shown in Figure 4-21.
This problem can be avoided by reducing the friction between the hinge bar and the staple or making the position of the hinge bar lower than the staple. Reducing the contact area between the hinge bar and the staple is a way that can reduce the friction.
Adding needles on upperside of the hinge bar or the underside of the staple will reduce the contact area. Or adding liquid that can reduce the friction of the poly surface between the hinge bar and the staple is also a method.
Figure 4-21: Hinge bar group in the not to be vertical device2