You are on page 1of 12

Simulation and interferometer results of MEMS deformable mirrors

Bautista Fernández and Joel Kubby


Department of Electrical Engineering
University of California, Santa Cruz
1156 High Street, MS:SOE2
Santa Cruz, CA 95064

ABSTRACT
Various types of large stroke actuators for Adaptive Optics (AO) were simulated individually and as part of a mirror
system consisting of actuators bonded to face plates with different boundary conditions. The actuators and faceplate
were fabricated using a high aspect ratio process that enables the fabrication of 3-dimensional Micro-Electro-Mechanical
System (MEMS) devices. This paper will review simulation results along with measurements of the displacement of the
actuators utilizing a white-light interferometer. Both simulations and interferometer scans have shown the ability of the
actuators to achieve displacement of 1/3 of the gap between the spring layer and the counter electrode.

Keywords: Adaptive Optics, Actuators, Large-Stroke, High Aspect Ratio MEMS

1. Introduction
Micro-Electro-Mechanical System (MEMS) deformable mirrors (DM) currently in development for adaptive optics
(AO) have not been able to meet the needs of large stroke, displacement of 10µm, of future telescopes like the Thirty-
Meter-Telescope (TMT). Such DM’s require a large amount of process development, costing time and money, and have
limited stroke (displacement). 1 Research efforts in MEMS “large-stroke” actuators for deformable mirrors (DM) have
used the Sandia ultra-planar, multi-level MEMS technology (SUMMiT) 2 to fabricate segmented AO mirrors and the
MEMSCAP polysilicon surface micromachining process3 to fabricate continuous-face-sheet AO mirrors.4 Boston
Micromachines Corporation (BMC) 5 fabricate their DM’s by modifying a 2-dimensional “surface micromachining”
fabrication process developed by Howe and Muller consisting of thin 1-2µm thick layers of structural polysilicon and
sacrificial oxide. 6

The individual actuators and mirror systems discussed in this paper use high aspect ratio technology, 7, 8 from HT-Micro9,
and do not require process development to achieve large strokes. Various mirror designs with different actuator and
assorted spring constants have been simulated with IntelliSuite’s Finite Element Analysis software.10

2. Designs
The individual actuators and mirror systems were created with L-Edit and simulated with Intellisuite’s Finite Element
Analysis (FEA) software. The purpose of modeling the actuators and mirror systems was to acquire an understanding of
how much displacement a design would obtain for a given voltage. The simulations will also allow us to pick the best
actuator and mirror boundaries combination to utilize in our next fabrication run.

The simulated mirror systems consisted of two mirror membrane designs. It consisted of:

• Mirror-membrane with a stiffening ring at its edges overlapping actuators


• Mirror-membrane supported at its edges by a post at the center of actuators

These mirror systems with different boundary conditions were simulated with actuators that were either:

• Square actuator supported by folded springs at the edge centers


• Square actuator supported by eight folded springs at the corners
• Circular actuator supported by folded springs at 90°, 180°, 270°, and 360° of its perimeter
The actuator 3-dimentional designs made with Intellisuite are show in Figure1. Figure 2, contains mirror system designs
with different boundary conditions created with L-edit. The thickness of the actuator spring layer is 5µm and mirror
layer is 2µm.

(A) (B) (C)


Figure 1: (A) Center edge support, (B) Corner eight-spring support, (C) Circular actuator 3-D models.

(A) (B) (C)


Figure 2: Mirror system designs with (A) Posts (B) Ring and (C) Folded springs as boundary conditions.
2.1 Actuator Simulation
The actuators were simulated individually to obtain a better comprehension of their response. A tensile stress of 50MPa
was utilized on all the simulations. Simulation of the center edge support actuator, showed it achieve a displaced
~9.81µm with an applied voltage of 152V. Table 1 shows the displacement simulation results obtained for the center,
corner, and supported edge of the actuator. Figure 3 contains the corresponding displacement vs. voltage plot.

V Center Corners Edge Corner - Edge -


(µm) (µm) (µm) Center Center
Difference Difference
(µm) (µm)
152.00 -9.81 -9.44 -9.27 0.38 0.54
140.00 -7.26 -6.98 -6.86 0.28 0.40
120.00 -4.53 -4.36 -4.28 0.17 0.25
100.00 -2.81 -2.70 -2.65 0.11 0.16
80.00 -1.66 -1.60 -1.57 0.06 0.09
60.00 -0.88 -0.84 -0.83 0.04 0.05

Table 1: FEA simulation result of a square actuator Figure 3: Displacement of a square actuator supported
supported by folded springs at the edge centers. by folded springs at the edge centers.
Figure 4 shows the simulated displacement of top and side view for the center edge support actuator.

Figure 4: Center edge support actuator displacement simulation.

The simulation of the corner eight-spring support actuator is capable of obtaining a displacement of ~9.76µm with an
applied voltage of 156V. Table 2 shows the displacement simulation results obtained for the center, corner, and
unsupported edge of the actuator. The corresponding displacement vs. voltage is shown on Figure 5.

V Center Corners Edge Corner - Edge -


(µm) (µm) (µm) Center Center
Difference Difference
(µm) (µm)
156.00 -9.76 -8.84 -9.39 0.92 0.37
140.00 -6.65 -6.04 -6.40 0.61 0.25
120.00 -4.22 -3.84 -4.06 0.38 0.16
100.00 -2.64 -2.41 -2.55 0.23 0.09
80.00 -1.57 -1.43 -1.51 0.14 0.06
60.00 -0.83 -0.76 -0.80 0.07 0.03

Table 2: FEA simulation result of a square actuator Figure 5: Displacement of a square actuator supported
supported by springs at the corners. by springs at the corners.

Figure 6 shows the simulated displacement of top and side view for the corner eight-spring support actuator.

Figure 6: Corner eight-spring support actuator displacement simulation.

The circular actuator is capable of obtaining a displacement of ~9.87µm with an applied voltage of 167V. Table 3 show
the displacement simulation results obtained for the center, supported perimeter, and unsupported perimeter of the
actuator. The corresponding displacement vs. voltage is found on Figure 7.
V Center Perimeter Support Center - Center -
(µm) (µm) (µm) Support Perimeter
Difference Difference
(µm) (µm)
167.00 -9.87 -9.47 -9.17 -0.70 -0.40
150.00 -6.79 -6.51 -6.31 -0.48 -0.28
140.00 -5.48 -5.26 -5.09 -0.39 -0.22
120.00 -3.59 -3.45 -3.34 -0.25 -0.14
100.00 -2.30 -2.21 -2.14 -0.16 -0.09
80.00 -1.39 -1.34 -1.30 -0.09 -0.05
70.00 -1.01 -0.97 -0.94 -0.07 -0.04
60.00 -0.76 -0.73 -0.71 -0.05 -0.03
Table 3: FEA simulation result of a circular actuator. Figure 7: Displacement of a circular actuator.

Figure 8 shows the simulated displacement of top and side view for the circular actuator.

Figure 8: Circular actuator displacement simulation.

All of the actuators behaved as expected. The rectangular actuator supported at the corners did maintain the
displacement of the corners to be less than that of the center. The actuator supported at the center edge had a corner to
center difference of 0.38µm.

A circular actuator was crated to prevent premature pull-in attributed to the unsupported corners of the square actuators
displacing farther than the center of the actuators. It was observed that the perimeter of the circular actuator would be
displaced by 0.7µm less than the center. Simulation of this actuator in a mirror system, as discussed below, showed that
the perimeter did not surpass the displacement of the center of the actuator. One drawback of such an actuator is the
increase in the voltage required to displace 10µm. The voltage increase is due to the smaller electrode area.

2.2 Mirror system simulation


The individual actuators were simulated as part of a mirror systems composed of an array of 3x3 actuators. The center
actuator in the 3x3 mirror systems was displaced to approximately 10µm, thus allowing the surrounding actuators to act
as support springs.

A mirror system consisting of a membrane with a 200µm wide stiffening ring overlapping the square actuators supported
at the edge center was simulated. Simulation results are shown on Table 4 and the corresponding displacement vs.
voltage is shown on Figure 9.
V Center Corners Corner -
(µm) (µm) Center
Difference
(µm)
193.00 -9.61 -9.98 -0.37
173.00 -7.39 -7.55 -0.16
153.00 -5.61 -5.66 -0.05
133.00 -4.15 -4.15 0.00
113.00 -2.95 -2.95 0.00
93.00 -1.99 -1.99 0.00
73.00 -1.25 -1.23 0.02
50.00 -0.67 -0.65 0.02
Table 4: Simulation result of a square actuator Figure 9: Displacement of a square actuator supported
supported by folded springs at the edge centers. by folded springs at the edge centers.

This mirror system required 193.00 volts to displace the center of the mirror membrane by 9.61µm. A drawback on this
model was that the unsupported actuator corners would tent to displace more than the center of the actuator. The edge
center support actuator model should be avoided in future designs to prevent premature pull-in caused by actuator
corners coming into contact with the counter electrode.

For the same type of mirror boundary conditions but with corner eight-spring support actuators, a pull-in voltage of
199.00V was needed for the center to be displaced to ~9.79µm and the corners to 9.59µm. Compare to the previous
design, this mirror system is able to control the corner displacement and will be consider for future fabrication runs.
Table 5 contains simulation results for this mirror system, where center corresponds to the displacement of the center
part of the mirror membrane/ actuator, and corners and edge correspond to the corners and unsupported edge of the
actuator. Displacement vs. voltage is found on Figure 10.

V Center Corners Edge Corner - Edge -


(µm) (µm) (µm) Center Center
Difference Difference
(µm) (µm)
199.00 -9.79 -9.59 -9.88 0.20 -0.09
180.00 -7.70 -7.44 -7.69 0.26 0.01
160.00 -5.90 -5.64 -5.85 0.26 0.05
140.00 -4.41 -4.18 -4.35 0.23 0.06
120.00 -3.19 -2.99 -3.13 0.20 0.06
100.00 -2.19 -2.05 -2.14 0.14 0.05
80.00 -1.43 -1.32 -1.39 0.11 0.04
60.00 -0.85 -0.78 -0.83 0.07 0.02
Table 5: Simulation result of a mirror system with ring Figure 10: Displacement vs. voltage of a mirror system
boundary condition and corner eight-spring actuators. with ring boundary condition and corner eight-spring
actuators.

The last of the actuators simulated with the ring boundary condition, was the circular actuator. Unlike the previous
designs, the area of the circular actuator is smaller thus a higher voltage would be expected to displace the mirror
membrane close to 10µm. Table 6 show that the required voltage needed to displace the ring boundary mirror by
~9.83µm is 217V. In Table 6, perimeter corresponds to the unsupported outer-area of the circular actuator. Although the
device requires more voltage, it is more stable since there are no corners to worry about premature pull-in. Figure 11
contains the corresponding displacement vs. voltage for this mirror system.
V Center Perimeter Perimeter -
(µm) (µm) Center
Difference
(µm)
217.00 -9.83 -9.91 -0.08
200.00 -8.11 -8.14 -0.03
180.00 -6.41 -6.42 -0.01
160.00 -4.98 -4.98 -0.00
140.00 -3.76 -3.76 0.00
120.00 -2.73 -2.74 -0.01
100.00 -1.90 -1.91 -0.01
80.00 -1.16 -1.17 -0.01
60.00 -0.73 -0.72 0.01
Table 6: Simulation result of mirror system with Figure 11: Displacement vs. voltage of mirror
circular actuators. system with circular actuators.

Comparing the center displacement to the actuators perimeter is was observed that there exists a displacement difference
between the center and the perimeter of approximately 0.08µm. This type of mirror system will be taken into account
when deciding of the best actuator and mirror system to use in the next fabrication run.

The last boundary condition used was of a mirror membrane supported by posts that are on top of the actuators. This
type of boundary condition allows the edge of the mirror to move with the actuators supporting it. Table 7 contains the
corresponding simulation results, and Figure 12 contains the displacement vs. voltage plot.

V Center Corners Edge Corner - Edge -


(µm) (µm) (µm) Center Center
Difference Difference
(µm) (µm)
190.00 -9.91 -9.48 -9.86 0.43 0.05
180.00 -8.45 -8.06 -8.38 0.39 0.07
160.00 -6.17 -5.84 -6.09 0.33 0.08
140.00 -4.46 -4.20 -4.38 0.26 0.08
120.00 -3.13 -2.94 -3.07 0.19 0.06
100.00 -2.11 -1.97 -2.07 0.14 0.04
80.00 -1.35 -1.25 -1.31 0.10 0.04
60.00 0.79 -0.73 -0.77 0.06 0.02

Table 7: Simulation result of mirror system with Figure 12: Displacement vs. voltage of mirror
circular actuators. system with circular actuators.

Figure 13 show a front view of the simulated mirror systems consisting of a mirror membrane supported by post at the
center of the actuators.

Figure 13: Simulation front view of mirror system with corner eight spring actuators.
Simulation of this type of boundary mirror in conjunction with either the actuators supported at the corners with eight
springs or the circle actuators, showed promising result. For the mirror system with corner eight-spring actuators, the
difference between the center and the corners was of ~0.43µm. When modeling with the circle actuators, the difference
from the perimeter of the actuator to its center dropped to 0.09µm. The results of the simulation are shown Table 8, and
Figure 14 contains a displacement vs. voltage plot for such mirror system.

V Center Perimeter Perimeter -


(µm) (µm) Center
Difference
(µm)
202.00 -9.91 -9.82 0.09
180.00 -7.30 -7.24 0.06
160.00 -5.42 -5.35 0.07
140.00 -3.96 -3.89 0.07
120.00 -2.81 -2.75 0.06
100.00 -1.91 -1.87 0.04
80.00 -1.23 -1.20 0.03
60.00 0.73 -0.70 0.03
Table 8: Simulation result of mirror system with Figure 14: Displacement vs. voltage of mirror system with
circular actuators. circular actuators.

Figure 15 show a front view of the simulated mirror systems consisting of a mirror membrane supported by post at the
center of the actuators.

Figure 15: Simulation front view of mirror system with corner eight spring actuators.

The most promising mirror system that will be consider for future fabrication runs, consist of the a mirror-membrane
supported at its edges by a post at the center of the circular actuators as show in Figure 15. Another mirror boundary
condition that is currently being evaluated consists of a mirror membrane support around its edges by folded springs.
The L-edit design of such system is shown of Figure 2C.

3. Measurement results
Some of the actuator models have been fabricated with HT-Micro’s precision fabrication technology (HTPF™) and pull-
in measurements have been obtained with the use of a white-light interferometer. The values measured differ to the
values obtain in the simulation due to parts being manufactured with a smaller height than what was expected. The
smaller the gap on the actuator the smaller the voltage required for pull-in.

Actuator arrays were manufactured on two types of substrates.

1. Alumina substrate
2. Glass substrate
Figure 16: L-edit die layout.

The substrates were sectioned into six dies that contained various actuator designs. Figure 16 contains a picture of one of
the dies created with L-edit and send to manufacturing. The actuators included in this fabrication process contained
support at the edges with and with out etch holes, support at the corners, and octagonal actuators supported at four edges.
Bow-tie test structures11, 12 were also fabricated to determine if tensile of compressive stress are present.

The actuators fabricated on the glass substrate showed resolution and alignment improvement over the first wafer run.
For this wafer run, we were able to obtain line features in the order of 10µm. The structure on Figure 17 was used to
measure misalignment between layers. It was measured that a misalignment of ~10µm exists between layers. Figure 18
contains measurements obtained with a white-light interferometer.

Figure 17: 3-D Interferometer image of Figure 18: Interferometer misalignment results.
an alignment structure.

Some of the devices that were tested had the spring layer removed intentionally to inspect between such a layer and the
counter electrode. Inside some of the devices there seemed to be leftover residue that was scattered on-top of the counter
electrode. Other devices had similar scattered material that was dark in color. Inspecting the alumina wafer also
revealed that we were able to obtain 10µm line features. On the alumina wafer, like in glass wafer, some of the devices
had the spring layer removed. Exactly like glass wafer, some of these devices had either chemical residue or dark chunks
of material. Overall, when comparing the two wafers it seem that the alumina wafer had less debris, scratches than the
glass wafer.

With the use of a white light interferometer bowtie structures were analyzed, thus allowing us to determine that the
actuators manufactured on both wafers had tensile stress. Initially before fabrication took place, the manufacture had
express that a 20MPa tensile stress was achievable. After the wafers had been fabricated there tensile measurements
increased between 6 and 7 times the initial quote. Our own measurements indicated an increase of 3 to 7 times the
original quote.
This unexpected increase in tensile stress will affect the manufacture parts. It will cause the device to have an increase in
voltage and to bow (deform). Interferometer scans showed that the actuator had a bow of approximately 1.82µm.
Figure 19 contains measurements take with the interferometer of an actuators surface bow.

Figure 19: X and Y axes bowing of the spring layer of a square center edge support actuator.

The bowing will prevent taking an accurate response measurement of the actuators. Since most of the actuators
manufactured are rectangular, the tensile stress causes the actuator corners to lift higher than the center, thus the height
of the corners are not the same. In turn, this bowing will prevent the actuator from displacing evenly. The corners
change in height by 100’s of nanometers to +1µm.

Given the high tensile stress achieved, new boundary conditions need to be created such as that show in Figure 2C. The
minimal tensile stress that the manufacture believes can be achieve is ~50MPa. The manufacture believes that the new
tensile stress can be obtained by utilizing a substrate that has a closer match coefficient of thermal expansion (CTE) to
gold.

The interferometer also revealed that the height of the manufactured devices was less than the 36.5µm required height.
That is 5µm for the actuator spring layer, 30µm gap, and 1.5µm for the counter electrode. The alumina wafer had an
average height of ~30.2µm with a standard deviation of 1.1µm, whereas the glass wafer had an average height of
29.5µm with a standard deviation of 2.7µm. Figure 20 shows a picture of one a 2 x 3 actuator array fabricated with
HTPF™ along with an interferometer scan of the top right actuator.

(A) (B)

Figure 20: (A) Top view of a 2x3 array of fabricated actuators (B) 3-D imaged obtained with a white-light interferometer.
Given the unexpected height obtained for both alumina and glass wafers, our devices had a gap of ~23.7µm and 23µm,
respectively. Due to the lower gap, the pull-in voltage of the fabricated actuators would be lower than expected from the
simulations.

Although the gap was smaller than expected, the actuators were able to achieve a displacement of 1/3 gap.
Measurements take with an interferometer are show in Figure 21. The actuator that was measure in Figure 21B pull-in at
a corner, thus the center of its spring layer did not go into contact with the counter electrode. The voltages of the
measured actuators were smaller than the simulation results due to the actuators containing a smaller gap. If the gap had
been 36.5µm as required it is expected that the actuators would be able to achieve ~10µm of stroke (displacement). A
closer inspections revealed that the actuators contained stringers and residue as shown if Figure 22. The stringers could
contribute to premature pull-in and to the actuator corners being pull-in, thus preventing the center of the actuator from
displacing evenly as show on Figure 21B. Internal residue could also attribute to the actuator pull-in at a corner.

(A) (B)
Figure 21: Interferometer results (A) Pull-in= 133.5V, Displacement= 8.19µm (B) Pull-in= 133.00V, Displacement= 9.07µm.

(A) (B)
Figure 22: (A) Stringer (B) Material residue.

Mirror-Membrane
To make sure that the new mirror boundary conditions, Figure 2C, are able to handle higher tensile stress (~50MPa), the
faceplate was simulated with stress only. From the simulation is was found out that such a faceplate would stay
relatively flat when facing 50MPa of tensile stress. One concern is that the corner actuators would tent to tilt diagonally
but not enough for the corners to come into contact with the membrane. Using circular actuator can help resolve the
diagonal tilting that the square actuator experience due to their lack of corners. On a 5x5 array simulation, it was shown
that the diagonal tilting was minimal on the inner diagonal actuators from the 5x5 corner actuators.

This faceplate has been fabricated with HTPF™ and is still going under constant modification. The current faceplate was
fabricated on top of fixed post to determine print-through. The fabricated faceplate has a thickness of ~2.5µm. Although
there have been some problems with parts of the membrane acting as shrink-wrap on the post, we have been able to
obtain partial release of a face place the size of a 16x16 array of actuators. The faceplate has been fabricated on a
substrate that is both closely CTE match to gold and has an RMS smoothness in the order of nanometers.
Figure 23A contains a 3D image of the folded spring boundary condition faceplate obtained with the white
interferometer. Figure 23B contains the interferometer measurement of the substrate surface been used to fabricate the
mirror-membrane.

(A) (B)
Figure 23: (A) 3D interferometer image, (B) substrate surface measurement.

Conclusion
The simulation of the actuators showed that they all behaved as expected. The rectangular actuator supported at the
corners did maintain the displacement of the corners to be less than that of the center. The actuator supported at the
center edge had a corner to center difference of 0.38µm. The perimeter of the circular actuator is displaced by 0.7µm
less than the center at a voltage of 167.00V. Simulation of this actuator in a mirror system showed that the perimeter did
not surpass the displacement of the center of the actuator. One drawback of such an actuator is the increase in the voltage
required to displace 10µm. From the simulations it was observed that the center edge support actuators would have their
corners pull-in prematurely, thus this design will be avoided in future fabrication runs.

The main candidates for the next fabrication run are the corner eight-spring actuators and the circular actuators along
with the post and folded spring mirror boundary conditions.

The heights of the fabricated parts in the alumina and glass wafers were smaller than expected, but a 1/3 gap
displacement was achieved. It was noticed that some of the actuators pull-in at the corners, this could be attributed to the
corners not being supported, stringers, or material residue.

The next step would be to fabricate an entire mirror system composed of an array of 16x16 actuators with post and
folded spring boundary conditions.

Acknowledgement
This work has been supported in part by the National Science Foundation Science and Technology Center for Adaptive
Optics, managed by the University of California at Santa Cruz under Cooperative Agreement No. AST-9876783. This
research was also supported in part by a Special Research Grant from the University of California, Santa Cruz.
References
1. Bautista Fernandez, Joel Kubby, Design processing and materials for large-stroke actuators, Proc. SPIE Vol. 6467
(2007)
2. http://www.sandia.gov/mstc/technologies/micromachines/tech-info/technologies/trilevel.html, last accessed on
1/03/2007
3. http://www.memscap.com/, last accessed on 1/03/2007
4. P. Krulevitch, Moems spatial light modulator development at the center for adaptive optics moems and miniaturized
systems III, Proc. SPIE, Vol: 4983, pp. 227-233 (2003).
5. http://www.bostonmicromachines.com, last accessed on 1/01/2007
6. R. T. Howe, R. S. Muller, Polycrystalline silicon micromechanical beams, Proc. Electrochemical Society Spring
Meeting, pp. 184-185 (1982)
7. Todd Christenson, X-Ray-Based-Fabrication, Chp. 18, The MEMS Handbook, CRC Press, pp. 18-1 – 18-46 (2002)
8. B. Fernández, J. A. Kubby, Large stroke actuators for adaptive optics, Proc. SPIE Vol. 6113, pp. 238- 248 (2006)
9. http://htmicro.com, last accessed on 1/01/2007
10. http://www.intellisensesoftware.com/products.html, last accessed on 01/03/2007
11. Paul M. Zavracky, George G. Adams, Paul D. Aquilino, Strain Analysis of Silicon-on-Insulator Films Produced by
Zone Melting Recrystallization, Journal of Microelectromechanical Systems, Vol. 4, NO.1, pp. 42 - 48 (1995)
12. Yogesh B. Gianchandani, Khalil Najafi, Bent-Beam Strain Sensors, Journal of Microelectromechanical Systems,
Vol. 5, NO.1, pp. 52 - 58 (1996)

You might also like