Assembly and Testing of
MEMS Mirror for Endoscopic
OCT
Dolly Creger
Utah State University
Mentor: William Tang, BME
Graduate Student Mentor: Jessica Ayers, BME
1
Abstract
The new arising technologies of microelectromechanical systems (MEMS) and
Endoscopic Optical Coherence Tomography (OCT) can be combined to facilitate internal
imaging without invasive procedures. This paper will illustrate the design of a 360º rotating
MEMS mirror capable of enabling the imaging of tubular structures within the body such as
esophagus, trachea, and artery walls. This device, only 1 mm in diameter, will be nearly half
the size of those currently on the market today. The MEMS device proposed will consist of a
rotary scratch drive motor and a 500-1000 µm polysilicon mirror. It has been hypothesized
that magnetism could be used to assemble the micromirrors. It was found that on such a small
scale the force of stiction was greater than the magnetic pull on the Nickel coated mirrors.
The force of stiction needs to be eradicated before any other force can lift the mirrors.
Possible means to relieve surface tension prior to assembly are vaporization of water or
shear-mode ultrasonic vibrations on the mirrors.
Key Terms
Chemical Vapor Deposition (CVD) - A process in which a controlled chemical reaction
produces a thin surface film.
Microelectromechanical Systems (MEMS) - micron-scale devices linking mechanical parts
to electrical components.
Optical Coherence Tomography (OCT) - Imaging technology which uses infrared light to
image cross-sections of biological tissue with micron scale resolution.
Photoresist- A light sensitive coating that is used to transfer the image of a mask onto the
2
surface of a wafer, followed by a reactive ion etch (RIE).
Stiction- The surface tension of liquid on a MEMS component.
Shear Mode Quartz- A piezoelectric effect in a lateral plane.
Introduction
Current medical technology requires a biopsy to be taken when examining the trachea,
GI tract, or any other tubular structure in the body. Biopsies not only cause tissue damage at
the site of excision, but the sample collected may not include any diseased areas which are
present in the tissue. Endoscopic Optical Coherence Tomography (OCT) has been employed
to image the body internally without the use of invasive technology. There are a number of
linear micro scanners available for internal scans, however there are few rotary scanners
available for tubular imaging.
The factor which limits the size for rotary scanners is the rotating mirror necessary for
360º scans. Using current microelectromechanical systems (MEMS) technology, a 1 mm 360º
rotational micro mirror is proposed in this paper. This MEMS mirror will enable the
fabrication of an endoscopic probe that is half the size of those currently available to the
market.
The MEMS device presented will consist of two major components. The first is a rotary
scratch drive array used for actuation. The scratch drive uses differences in electric potentials
to “scratch” along the substrate; a circular array of these devices will initiate rotation. To test
this, variances in voltages and frequencies were applied.
The second part of the MEMS device is a polysilicon mirror that will be used for both
transmission and receipt of signals. Two mirror designs will be tested. One will be sputtered
3
with gold for increased reflectivity, while the other will be fabricated upside down and will
flip up over itself for increased optical flatness.
The entire MEMS device will be fabricated in plane using a currently available
lithographic process. The mirrors will then be assembled a top the rotary scratch drive motor.
This process will enable the mirrors to rotate an entire 360º. It will be mounted at the end of
an endoscope and used to image cylindrical pathways within the body. Stiction must be
overcome in order to assemble the mirrors, several treatments were employed to accomplish
both these tasks.
Design and Fabrication
The device was designed to be fabricated using a commercially available polyMUMPs
process, to make it more reproducible and thus more
commercially viable. The device was designed using the
program L-Edit. The fabrication process began with a
blank, 100 mm, n-type silicon wafer, having 1-2 olhms
resistively. The Surface was then doped with
Figure 1: 592 µm x 527 µm L-Edit mirror/motor design
phosphorous using POCl3 as its source. Doping the surface helps reduce or prevent charge
feed through to the substrate from electrostatic devices on the surface.
Low-stress LPCVD (low pressure chemical vapor deposition) was then used to deposit
silicon nitride, as an electrical isolation layer, at a thickness of 600 n. Next a 500 nm layer of
LPCVD polysilicon film known as Poly 0 was then deposited. At this point the wafer was
coated in photoresist and the appropriate mask is placed above the wafer and is exposed to
UV light. This allows the photoresist covered by the mask to withstand decomposition in
4
acetone. This process allows the transfer of the mask to the wafer. After the photoresist is
patterned, the Poly 0 layer can then be etched in an RIE (Reactive Ion Etch).
A 2.0 µm phosphosilicate glass (PSG) sacrificial layer is then deposited by LPCVD and
annealed for 1 hr. in argon. The PSG, known as First Oxide, is removed to free the first
mechanical layer of polysilicon. The sacrificial layer is lithographically patterned with the
DIMPLES mask and the dimples are transferred into
the sacrificial PSG layer by RIE. The dimple depth for
this device was 750 nm.
The wafer was then patterned with the 3rd mask
layer, ANCHOR1 and reactive ion etched. This step
provided the anchor holes that were filled by the Poly 1
Fig 2: SEM Image of Rotary Scratch Drive
layer. Two micrometers of un-doped polysilicon was then deposited by LPCVD. Next 200 nm
of PSG was deposited and anneal. This step both doped the polysilicon and reduced its
residual.
The wafer was then coated with photoresist and the 4th layer was lithographically
patterned. The PSG was first etched to create a hard mask and then poly1 was etched by RIE.
After this step the photoresist I and PSG hard mask were removed.
The 2nd oxide layer .75 µm was then deposited and then patterned twice to allow contact
to both poly 1 and substrate layers.
The final step was to coat the flip up mirrors with two micrometers of gold to increase
optical reflectivity (Koester et al. 2-5).
5
Testing
In order to assemble the mirrors on top of the rotary scratch drive, the liquid underneath
the mirrors needed to be eradicated. Methanol, which has less surface tension than water, was
the liquid used to replace hydrofluoric acid. The chip was placed into a drying oven set to
400º C for 2 hours. Due to the tendency of water to adhere to itself it, an instantaneous
vaporization of the liquid surrounding the chip was tested. A class R laser was pulsed at a
wavelength of 2940 nm with the maximum output of 500 mJ for 400 microseconds.
The mirrors were coated with Nickel using a chemical
vapor deposition (CVD). With the new ferromagnetic coating,
the mirrors would be attracted to powerful rare earth length
of the mirrors for assembly. The magnets were hovered over
the chip while the chip remained submerged in Methanol.
Figure 3: Ni Coated Polysilicon Mirror
With liquid surrounding the device, there would be equal surface tension above and below the
mirror.
Results
Baking the chip in the oven served to evaporate the
water only near the edges of the mirrors. With water still
trapped in the inner portion of the mirrors, the mirrors could
not be assembled. The laser could not run continually enough to
vaporize the water instantaneously. Many iterations
were necessary to evaporate all the liquid on the
chip.
6
Figure 4, 5: Affects of stiction on polysilicon mirror
When the magnet was passed over the Nickel coated mirrors no noticeable attraction
occurred. On such a small scale the force of viscosity of the methanol may have been
stronger than the magnetic pull on such a thin film of Nickel.
By the second day the layer of Nickel applied was worn away perhaps by the methanol
or any residual layer of photoresist left underneath the mirror. On the first day, the gold
coated mirrors had shiny nickel on them, however by the end of the day it could be seen that
the nickel was being degraded.
Discussion
It can be concluded that stiction must be eradicated before assembly of the mirrors can
occur. New ideas that may be implemented is a continuous laser to vaporize all the liquid
instantaneously. It would be beneficial to find a laser whose wavelength is entirely absorbed
by water, but will be transparent to silicon.
Graph 1: Absorption of silicon and water, Jessica Ayers UCI
7
A backside etching approach may also be taken. If the silicon is etched away from
beneath the mirror, there will be nothing for stiction to adhere to. However, the mirror is
fragile enough that it would need to be supported by a few support bars that are thin enough
not to trap substantial amounts of liquid between them and the polysilicon mirror.
Another technique that may be used is Shear Mode Quartz. Previous authors have been
able to release components of MEMS devices under the presence of stiction, by using
ultrasonic vibrations. If a shear mode vibration were used, a lower frequency of vibration
would be necessary to release the mirrors (Kaajakari and Lal, 1999). Centrifugal force has also
been implemented as an application for active-assembly (Iwase and Shimoyama, 1997).
For assembly of the mirror, it would be necessary to increase the thickness of the nickel
coating. The thicker the nickel coating, the stronger pull the rare earth magnets will have on
the mirrors surface. According to the equation
Sensitivity) where S
(Equation 1: Magnetic
is the sensitivity factor which indicates the ease of lifting according to a
magnetic field. According to the equation, Vmag is the volume of the ferromagnetic material
which would coat the mirrors. Lh, Wh, and Th indicate the length, width, and thickness of the
hinge part, respectively. It can be found the thickness of Nickel coating required for sufficient
magnetic pull.
Photoresist hinges may also be implemented. It has been shown that the surface tension
of photoresist upon melting can assemble small mirrors (Syms, 1999). A laser would need to
be used that is transparent to photoresist as well as silicon in order to vaporize all of the liquid
beneath the mirrors.
8
Acknowledgements
This Project was supported by the National Science foundation, University of California,
Irvine, and the UROP-IMSURE program.
Works Cited
1. David Koester, Allen Cowen, Ramaswamy Mahadevan, Mark Stonefield, and Busbee
Hardy. “a MUMPs Process.” PolyMUMPs Design Handbook. Berkley: MEMSCAP, 2003.
chapter 1. pp. 2-5.
2. Ville Kaajakari and Amit Lal. “Pulsed Ultrasonic Release and Assembly of
Micromachines.” The 10th International Conference on Solid-State Sensors and Actuators.
University of Wisconsin-Madison. 1999.
3. Eiji Iwase and Isao Shimoyama. “Multi-Step Sequential Batch Self-Assembly of
Three-Dimensional Micro-Structures using Magnetic Field.” MEMS 2005 Technical Digest. 6.1
(1997): 10-17.
4. Richard R. A. Syms. “Surface Tension Powered Self-Assembly of 3-D
Micro-Optomechanical Structures.” Journal of Microelectromechanical Systems 8.4 (1999):
448-455.
9