The fabrication and measurement results are presented in this chapter. The batch assembly of 90° devices is demonstrated, followed by the discussion of the angular deviation of the assembled devices. 90° devices fixed by V-shaped hinges only without traditional hinges are also tested. A corner cube reflector based on this concept is fabricated and tested. Finally, the positioning and angular accuracy of 45°
devices are presented.
4-1 Batch assembly
Three microprobes were connected as an array by vacuum tape to push an array of 90° mirrors. The microprobe has a 10 µm diameter precision tip connected to a 4-cm -long shaft. The push pads are 250 × 250 µm2 and mirror plates of 760 × 760 µm2. The microprobe array was operated by a micropositioner, as shown in Figure 4-1. The distance of the microprobes was the same as the distance of the push pads, which was 2 mm.
The demonstration of batch assembly is shown in Figure 4-2. The different tip positions of the handmade microprobe array affected the respective assembly of each mirror. The problem can be easily solved by the make-to-order microprobes. After the microprobe array was pushed down, the middle mirror was first assembled (Figure 4-2 (b)). Then the other two mirrors were assembled successively as the pushing depth of the mircoprobes was increased (Figure 4-2 (c)(d)). After the microprobes were removed, the mirrors were assembled successfully (Figure (e)). The assembly time of three mirrors was 40s. Compared to 25s of a single mirror, the assembly time was reduced by 46%.
Figure 4-1 Setup for batch assembly.
Figure 4-2 Sequence of batch assembly of three mirrors, (a) probes are aligned to the push pads, (b) the middle mirror was pushed up first, (c) other mirrors were assembled subsequently, (d) mirrors were assembled at 90°.
(a) (b)
(c) (d)
Micropositioner
Array of probes
2 mm
Figure 4-2 Sequence of batch assembly of three mirrors (continued), (e) the probes were moved away (f) SEM of assembled mirrors.
4-1-1 Problems
In the assembly process, the three mirrors and the three tips form two lines. If the two lines are not aligned to each other, batch assembly may fail. This is important for batch assembly on the wafer level. Figure 4-3 is an example of unsuccessful assembly of three mirrors. The right-hand side mirror was not assembled due to this problem (Figure 4-3 (d)). For batch assembly of three mirrors, 1° misalignment of the two lines results in 70 µm deviation of the microprobes in x axis. Although the one push operation has large positioning tolerance due to the large push pad area (250 × 250 µm2), the displacement due to the misalignment is magnified, especially for more than 3 mirrors.
(e) (f)
Figure 4-3 Sequence of a failed batch assembly, only two samples were pushed up.
4-2 Reflective coating
A reflective coating might be applied to the mirror. The residual stress in the coated metal will result in a curved mirror that will induce aberrations. Increasing the rigidity of the mirror by using a thicker mirror plate can solve the problem. An aluminum-coating experiment was conducted to measure the residual stress of aluminum, as shown in Figure 4-4.
If an isotropic stress is assumed and the thickness of the film is much smaller than that of the substrate, the stress can be expressed as [30],
(a) (b)
(c) (d)
Probe tip line
Push-pad line Offset
2 mm x
y θ
2
3 (12 )
s s
s f
E d
r d
σ δ
ν
⋅ ⋅ ∆
= − (4-1)
where σ is the intrinsic compressive stress, ∆δ is difference in bowing of the substrate before and after, r is the radius of the substrate, Es is the Young’s Modulus of the substrate, vs is the Poisson’s ratio of the substrate, ds and df are the thickness of the substrate and the film, respectively. In Figure 4-4 (b), the aluminum is 1000 Å thick.
The radius of the substrate is 760 µm. The result shows that the residual stress of aluminum is compressive. The intrinsic compressive stress can be calculated as 0.68×108 Pa, which is comparable to the value 1.5×108 Pa in literature [31]. By substituting the stress into Equation 4-1, the thickness of the SOI device layer should be increased to 50 µm so that the bowing is small than λ 10 0.043= µm
(a)
(b)
Figure 4-4 (a) Mirror before coating (b) mirror after coating.
4-3 V-shaped hinge
Positioning accuracy of the conventional and the V-shaped hinges are compared.
As shown in Figure 4-5 (a), the hinge pin of the conventional hinge is at the center of the rotational axis. Figures 4-5 (b) and (c) show the right and left sides of the hinge pin after assembly. The two gaps should be both 10 µm if the hinge pin is at the center of the rotational axis. In Figures 4-5 (b) and (c), the gaps are about 13 µm and 7 µm, respectively. About 3 µm offset of the hinge pin position was measured.
Figure 4-5 (a) Layout of the conventional hinge, (b) right side, (c) left side.
Compared to the conventional hinge, the SEM photographs in Figure 4-6 show the locking accuracy of the V-shaped hinge. In Figure 4-6 (a), the distances from the rotational axis to the two sides of the hinge pin are 2.5 µm and 8.5 µm, respectively.
Rotational axis
Since the thickness of silicon is 5 µm, the distance between the hinge pin and the two sides of the SU-8 should be 0 µm (Gap 1) and 6 µm (Gap 2), respectively, after assembly. In Figure 4-6 (c), Gap 2 is broader than the 5-µm-thick hinge pin. Thus the axis deviation is less than 1 µm in the V-shaped hinge.
Figure 4-6 (a) Layout of the V-shaped hinge, (b) SEM photograph of the V-shaped hinge, (c) zoom in of the V-shaped hinge.
Mirror
V-shaped hinge Rotational axis
(a)
(b) (c)
5µm 8.5µm
2.5µm
Gap 1
Gap 2 Hinge pin
SU-8 step
4-3-1 Strength of the V-shaped hinge
The assembled 90° mirror without side latches but with only V-shaped hinge is shown in Figure 4-7 (a). Figures 4-7 (b) and (c) show the SEM of the same sample taken in two different measurements. The mirror was measured at different angles because it was affected by the air flow of the pumping procedure in SEM. Although the V-shaped hinge can position the axis more precisely than the conventional hinge, the robustness of the assembled devices still needs to be improved.
(a)
Figure 4-7 (a) SEM photograph of 90° mirror without side latch, (b) and (c) the same sample taken at two measurements.
(b) (c)
91.0o
92.5o Viewing direction in (b) and (c)
4-4 Angular accuracy of 90° mirrors
The causes for angular inaccuracy are discussed in the following. The possible causes are listed and their effects are discussed.
Step in the SU-8 photoresist
Because the oxide deposition process in Step G is not a planar process, a step is formed in the bottom of the SU-8 photoresist and structure that affects mirror angle after assembly, as shown in Figure 4-8. Figure 4-8 (a) shows the layout design. Figure 4-8 (b) shows the cross section of the SU-8 step after assembly. Since the step locked the mirror at the rotational axis, the angles of the assembled devices are all less than 90° [7].
The step can be removed by polishing after the oxide deposition process. Another easy solution of this problem is to change the location of the step in the layout design, as shown in Figure 4-9. Because the mirror plate is not anchored after releasing, the offset of the SU-8 layer is not related to the angle of the flip-up mirror. The offset of the SU-8 layer can change the location of the SU-8 step. The separation between the locking point of the V-shaped side latch and the SU-8 step can solve the problem.
Figure 4-8 Step in the SU-8 photoresist, (a) layout design (b) cross section, (c) schematic, (d) SEM photograph of the SU-8 step [7].
Figure 4-9 Changing the SU8 step position, (a) layout design, (b) cross section.
SU-8 photoresist
Side latchLocking point of V-shaped side latch
Step
(b) (d)
Flip-up direction
SU-8 photoresist SOI Substrate
Side latchLocking point of V-shaped side latchLock less than 90°
Step
Side latches
Another reason that results in angular deviation is the side latches. As Figure 4-10 shows, the mirror plate was not locked into the center of the V-shaped opening. The main reasons are the spring and the friction force between the silicon and the SU-8 photoresist. The lifted-up angle of the side latches affects the angular accuracy. Figure 4-11 shows the SEM of different angles of the flip-up side latches and Figure 4-12 shows the statistics of the comparison. The angles were calculated by measuring the offset of the top of the mirror compared to the bottom using the ruler in an optical microscope. The magnification of the object lens and the eyepiece are 50 and 10, respectively. The resolution of the ruler is 1 mm. Thus the measurement resolution is
1 mm 2 µm pump. In Figure 4-12, the structures are fabricated at average 89.8±0.3°, 89.7±0.4°, and 89.4±0.4°, which are 10°, 20°, and 40° structures, respectively. The angles of the mirrors with 10° side latches are more exact than that with 20° and 40° side latches.
The reason was that the friction force of the 20° and 40° devices was greater than the 10° devices, as describe in Chapter 2. The restoring force of the side latches parallel to the x direction can not drive the mirror to the desired locking position effectively.
Figure 4-10 Offset of the mirror plate.
Side latch
0 1 2 3 4 5
Number of devices
89.0~89.2 89.4~89.6 89.8~90.0 90.2~90.4 Mirror angle
10°: 89.8±0.3° 20°: 89.7±0.4° 40°: 89.4±0.4°
Figure 4-11 Side latches with different flip-up angles, (a) 10°, (b) 20°, (c) 40°.
Figure 4-12 Measured angle of mirrors with different side latches angles.
4-5 Corner cube reflector
A corner cube reflector was assembled by the one-push method. The optical micrographs are shown in Figure 4-13 and the SEM photographs are shown in Figure 4-14. As Figure 4-14 shows, the corner cube reflector is constructed by two 90°
(a) (b)
(c)
mirrors. The reflecting zone of the corner cube reflector is (760µm)3. The three reflecting surfaces are the front side of Mirror 1, backside of Mirror 2, and the surface of substrate underneath Mirror 2. The angles of the corner cube reflector surfaces measured by an optical microscope were 89.9° between Mirror 1 and substrate, 89°
between Mirror 2 and substrate, and 90° between Mirror 1 and Mirror 2, with a measurement resolution of 0.15°. The deviation from the Mirror 2 and substrate needs to be improved. The deviation can be solved by adding locking mechanism between Mirror 1 and Mirror 2.
Figure 4-13 (a) Assembled corner cube reflector, (b) surfaces of the corner cube reflector.
Figure 4-14 SEM photograph of the fabricated corner cube reflector.
Lifted-up direction
Push pad and through-wafer hole
(a) (b)
Mirror1
Mirror 2 Substrate
underneath Mirror 2
Through-wafer hole Reflecting zone
4-5-1 Optical measurement
Light reflected from the corner cube reflector was measured optically as shown in Figure 4-15. The laser beam was first weakly focused by a lens with a focal length of 10 cm. The focused light was then divided by the beam splitter. The dotted line in Figure 4-15 is the reflected light by the beam splitter whereas the solid line is the transmitted light. The reflected light was reflected again by the inner surface of the beam splitter and projected onto the screen. The transmitted light was reflected by the corner cube and the beam splitter and then projected onto the screen. The distance between two projected spots was measured to derive the angle of the corner cube reflector. The experimental setup is shown as Figure 4-16.
Figure 4-15 Illustration of the optical measurement.
Figure 4-16 (a) Experimental setup, (b) highlight of the corner cube reflector.
Laser
Before the measurement, the optical path is aligned two steps. First the laser beam is aligned to be parallel to the optical table (Figure 4-17 (a)). Then the beam splitter and the lens are placed orthogonal to the laser beam (Figure 4-17 (b)), which can be obtained from the overlap between the reflected optical spot from the beam splitter and the emission aperture of the laser.
Figure 4-17 Optical path alignment, (a) laser beam alignment, (b) lens and beam splitter alignment.
Figure 4-18 shows the projected spots on the screen placed at d =10 mm and 60 mm away from the beam splitter. The scale of the grid paper is 0.1 mm/div. The calculation of the deviation angle θ is shown in Figure 4-19. The distance between the screens in Figures 4-19 (a) and (b) is 50mm. The distance between the two spots is increased by 1 mm. The angle θ can be calculated as 1 3 mm 2 mm o
= tan 1.15
50 mm
θ − − = . The corner cube is rotated by various angles and the result is shown in Table 4-1.
Overlap
(a) (b)
Figure 4-18 Reflected spots on the screen placed at (a) 10 mm, (b) 60 mm away from the beam splitter.
Figure 4-19 Angular deviation of the corner cube reflector.
Table 4-1 Angular deviation with various corner cube reflector orientations.
Measurement No. Deviation
1 1.15°
2 0.92°
3 1.03°
4 1.60°
5 1.26°
6 1.15°
Average deviation 1.19° ± 0.4°
2mm 2mm
50mm
θ 1mm Corner cube reflector
θ
Beam splitter
10mm
Reflected spot from the corner cube reflector Reflected spot from the beam splitter
(a) (b)
4-6 45° structures
Two types of 45° structures were designed in Chapter 2. The fabricated devices are presented in the following sections. The problems encountered in the assembly process will also be discussed.
4-6-1 Device 1
The SEM of the fabricated Device 1 are shown in Figure 4-20. Figure 4-20 (a) shows a 45° device before assembly. Figure 4-20 (b) shows the highlight of the support structure. Figure 4-20 (c) shows a 135° device. Figure 4-20 (b) shows the close-up view of the support structure.
Figure 4-20 Fabricated Device 1, (a) 45° structure, (b) support of 45° structure (b) 135° structure, (d) support of 135° structure.
(a) (b)
(c) (d)
Anchor
Anchor
However, the assembly process failed due to a step in the bottom surface of the SU-8 structures, as shown in Figures 4-21 and 4-22. Because the oxide deposition process is not a planar process, the subsequently coated SU-8 will have a mechanical step in the bottom surface, which will result in the locking phenomenon during assembly. The associated photographs of the optical microscope are shown in Figure 4-21. Figure 4-22 is the cross section of Figure 4-21. The assembly process failed eventually. Figure 4-23 shows the SEM of a 135° device. In Figure 4-23 (a), a 135°
device is assembled without the restoring force beams. However, the angle of the device without restoring force beams is unknown because the restoring force beams take no effect. In Figure 4-23 (b), the hinge pin was locked by the mechanical step and not on the axis because the imbalance of the force on the hinge pin since the restoring force beams on other side has been broken.
Figure 4-21 Mechanical steps in SU-8, (a) 45° device, (b) 135° device.
Mechanical steps
Mechanical steps
(a) (b)
Cross section line in Figure 4-22
Figure 4-22 Cross sectional view, (a) After SU-8 process. (b) problem of the assembly process.
Figure 4-23 SEM of 135° devices, (a) restoring force beam was destroyed due to the step, (b) the step lock the structure out of rotational axis.
(a) (b)
Out of rotational axis
Locked
Anchor Restoring force beam
SOI Substrate
Buried oxide PECVD Oxide SU-8 photoresist
Step
Silicon
Anchor
SOI Substrate
SU-8 photoresist
Movement of silicon Locked
Anchor
(b) (a)
4-6-2 Device 2
Mirrors without SU-8 hinges and supports were also assembled to verify the feasibility of the proposed concept. Figure 4-24 shows the assembly process of the fabricated structure. Probe 1 was aligned with the push pad of the mirror (Figure 4-24 (a)). The mirror was pushed over 65° and Probe 1 stayed on the push pad (Figure 4-24 (b)). Then Probe 2 pushed the push pad of the support to 45~55° (Figure 4-24 (c)). As Probe 1 was removed from the mirror, the mirror lay firmly on the support due to the torsional beams of the mirror (Figure 4-24 (d)). Probe 2 was removed subsequently and then the support also lay on the mirror due to the torsional beams of the support (Figure 4-24 (e)). Finally the mirror and the support were interlocked at the desired angle.
Figure 4-24 Assembly process of a 45° mirror without SU-8, (a) before assembly, (b) the mirror was first aligned and pushed up by Probe 1.
(a) (b)
Probe 1
800µm
Probe 1
Mirror
Torsional beam
Support
Figure 4-24 Assembly process of a 45° mirror without SU-8 (continued), (c) the support was pushed up by Probe 2, (d) Probe 1 was removed. (e) probe 2 was removed.
The SEM photographs of the assemble mirror are shown in Figure 4-25. Figure 4-25 (a) shows an assembled 45° mirror, whereas Figure 4-25 (b) shows the side view.
Figure 4-25 (c) shows the angular interlock components. Figure 4-25 (d) shows the torsional beam and the mechanical stop which is used to prevent offset of the torsional beam.
(c) (d)
(e) Probe 2
Figure 4-25 (a) An assembled 45° device, (b) side view of 45° device, (c) Highlight of the locking mechanism, (d) torsional beam and mechanical stop (continued).
4-6-2-1 Angle measurement
The angles of the assembled devices are defined by the layout design and should be 45°. The angles of the fabricated and assembled devices were measured by using SEM photographs. The resolution of the measured angle is 0.2°. The resolution is limited by the pixel of the SEM figures. Figure 4-26 shows the SEM photograph of an assembled sample and the geometric scale. The angle was calculated as 45.8°. Table 4-2 lists the angle measurement results of 7 samples. The average angle of them was 45.89°± 0.2°.
(a) (b)
(c) (d)
Torsional beam Mirror
Support
Through-wafer hole
Interlock
Mechanical stop Torsional beam
Mirror
Figure 4-26 An assembled mirror at 45.8°.
Table 4-2 Angle measurement (resolution: 0.2°)
Sample number Measured angle
1 45.9°
2 45.8°
3 46.0°
4 45.9°
5 45.8°
6 46.1°
7 45.7°
Average angle 45.89°± 0.2°
The reason for the deviation from 45° is the bending of the supports. As the Figure 4-27 shows, the bent supports influence the assembled angle. Additionally, the thickness of the device layer was neglected in the angular calculation and it results in 0.4° deviation.
The deformation of the bent supports was measured by a WYKO NT1100 interferometer. Figure 4-28 (a) shows the 3-D profile of the bent supports. Figure 4-28 (b) shows the 2-D profile of the AB cut in Figure 4-28 (a). A finite element simulation is shown in Figure 4-29. Figure 4-29 (b) shows the simulated side view of the support.
Mirror
(484,76)
(212,360) (615,357)
From 4-28 (b), the bending occurs where the support structure has a smaller width, thus a weaker mechanical strength. In Figure 4-29 (b), the bending is observed in the same region. Therefore the torsional beam design needs to be improved. The bending can be eliminated by increasing the width of the support. A simulation with 50 µm wider support is shown in Figure 4-30. The support in Figure 4-30 is straight and the deviation due to the support can be solved.
Figure 4-27 The bent supports, (a) side view, (b) perspective view.
Figure 4-28 Bending of the supports after assembly, (a) 3-D profile.
Mirror
Bending
B
(a)
A
(a) (b)
Figure 4-28 Bending of the supports after assembly, (b) 2-D analysis.
(a) (b) Figure 4-29 (a) Simulation of the support, (b) side view.
Figure 4-30 (a) Simulation of a thicker support, (b) side view.
A
B A
Bending
(b) Straight
Straight
Bending
Straight
Straight
4-6-2-2 Discussion
In assembly experiments, short torsional beams (e.q. 112 µm long) were easily broken during assembly procedure, as shown in Figure 4-31. Long beams (e.g. longer than 400 µm) were soft enough to avoid breaking. The maximum shear stress τmax in a 400-µm-long torsional beam at θ = 80° can be calculated by Equation 2-2 to be 1.2 GPa. Compared to the yield strength of 7 GPa, the safety factor is 5.8. Therefore the original design for a safety factor of 1.4 (τmax= 5 GPa in a 112-µm-long beam) is too aggressive. However, long beams with less restoring force may not drive the support and mirror to interlock firmly due to the friction force and result in more angular deviation. The trade-off between the two factors needs to be considered more carefully in the future.
Figure 4-31 Broken torsional beam.
Broken beam
4-7 Summary
Batch assembly of more than one 90° devices was demonstrated and verified. It can reduce the assembling time on the wafer level. The V-shaped hinge was compared with the conventional hinge and the function of the V-shaped hinge was verified. The angular accuracy of the corner cube reflector was measured optically. The deviation is 1.19° ± 0.4° and needs to be improved. The reasons of angular inaccuracy were discussed. The 90° device was improved to average 89.8±0.3°, compared to our previous study, 89.2±0.3° [7]. The 45° devices were assembled at average 45.9±0.2°.
The 45° devices need to be improved.