Large-Stroke Varifocal Mirror with Hydraulic Actuation for

Endoscopic Laser Surgery

André Geraldes1 , Andrea Jacassi2 , Paolo Fiorini3 and Leonardo S. Mattos4

Optical Lens Fixed

Abstract— Endoscopic laser tools allow using the precision Fiber Mirror
of surgical lasers in minimally invasive procedures . However
due to the difficulty of adjusting the focal length of the laser
Focal
during the procedure, ablation is typically performed with the Point
fiber laser in contact with the tissue, which often leads to
tissue carbonization. In order to overcome this problem, MEMS Varifocal Mirror

varifocal mirrors can be used. This paper presents a large-

stroke MEMS varifocal mirror for an endoscopic laser tool,
composed of a 4 mm circular silicon nitride membrane. Using Fig. 1. Schematic diagram of an endoscopic laser tool with adjustable
hydraulic actuation, large optical power can be achieved while focus using a MEMS varifocal mirror.
maintaining good optical quality. Experimental characterization
shows that the mirror can achieve an optical power range of
more than 64 m-1 with 9000 Pa of actuation.
results in large absorption of the laser at the surface of the
I. INTRODUCTION tissue leading to tissue carbonization [3].
In order to overcome this problem MEMS varifocal mir-
Lasers are used in many surgical interventions due to their rors can be used. Varifocal mirrors are deformable mirrors,
ability of performing precise incisions and ablations. One whose curvature can be dynamically changed, allowing to
example is the use of CO2 laser in phonomicrosurgery for control the focal length of the laser without mechanical
vocal cord’s tumor resection. In this procedure, the precision displacement. Being specifically designed for focus control
of the laser allows preserving more healthy tissue, which has they differentiate from other deformable mirrors for having
a significant impact on patient’s voice and recovery time [1]. few degrees of freedom, typically allowing to control only
Moreover the laser allows reducing bleeding by cauterizing the focal length and a few aberrations. This allows them to
small blood vessels while cutting, which improves the safety be very compact, fast and simple to control, while providing
of the procedure and prevents complications. large optical power range. Over the last decade, such kind
One of the main disadvantages of laser tools is that the of mirrors have already been incorporated in microscopy and
target site must be accessed by a straight line trajectory. imaging systems.
This poses ergonomic issues, specially when the area to be Several MEMS varifocal mirrors have been developed
operated cannot be properly exposed. With the development based on electrostatic actuation [4]–[8]. These mirrors con-
of fiber-coupled lasers, endoscopic laser tools have been sist on thin membranes deflected by electrostatic pressure,
proposed, allowing to deliver the laser closer to the target which is generated by applying an electric field between
site [2]. This provides improved visualization of the surgical the mirror and a counter electrode. This actuation method
site, while expanding the applicability of laser tools to other allows actuation frequencies in the order of kHz, but pro-
kinds of procedures. vides limited stroke and consequently limited optical power.
One of the main challenges in endoscopic laser surgery is Mirrors made from silicon membranes and nitride films [4]–
the precise focusing of the laser, since the distance between [6] typically provide a maximum stroke of 4 µm, while
the tool and the tissue can change during the procedure and polymeric membranes [8], [9] can reach up to 16 µm of
the tissue to be ablated is not always planar. Since adjustable stroke. For this reason, the diameter of electrostatic mirrors is
focusing systems based on moving lenses are usually too usually lower than 3 mm. Other disadvantages of electrostatic
large and bulky for an endoscopic system, the fiber tip is actuation include the fact that the mirror can only deflect
commonly placed in direct contact with the tissue. This in one direction and that the electrostatic pressure is non
∗ This
uniform, which creates spherical aberration on the mirror
work was not supported by any organization
1 André Geraldes is with the Department of Advanced Robotics, Istituto curvature.
Italiano di Tecnologia, Genoa, Italy [email protected] Using pneumatic actuation, a mirror with convex de-
2 Andrea Jacassi is with the Department of Plasmon flection has been developed [10]. In this case, the mirror
Nanotechnologies, Istituto Italiano di Tecnologia, Genoa, Italy
[email protected] membrane is connected to a sealed chamber, which is ac-
3 Paolo Fiorini is with the Department of Computer Science, University
tuated by an electrostatic membrane on its opposing side.
of Verona, 37134 Verona, Italy [email protected] The gas inside the chamber transfers the actuation from the
4 Leonardo S. Mattos is with the Department of Ad-
vanced Robotics, Istituto Italiano di Tecnologia, Genoa, Italy electrostatic membrane to the mirror, allowing to deflect the
[email protected] mirror in convex shape. Later this actuation mechanism has
been combined with an electrostatic mirror [11], allowing it r
R
to deflect in both directions. This 4 mm varifocal mirror was h
able to achieve a total stroke of 20 µm and an actuation speed w0 w(r)
of 2 kHz. Besides that, spherical aberration has been reduced
since the pneumatic pressure over the mirror is uniform. The p
same concept has been explored in [12], where a hydraulic
chamber has been used to transfer magnetic actuation to the
Fig. 2. Deflection model of a varifocal membrane mirror.
mirror membrane. In this case, the mirror could achieve 18
µm of stroke with 22.5 kPa of actuation, but the maximum
Gold coating
actuation frequency was 15 Hz. Cover
Other actuation mechanisms investigated so far include Silicon nitride membrane

piezoelectric [13] and electrothermal mirrors [14]. Elec- Silicon bulk

Fluidic chamber

trothermal mirrors are composed of a bimorph membrane, Sealing sheet

whose deflection is caused by the difference in thermal

Mirror holder
expansion of the two materials that compose the membrane.
These mirrors can achieve more than 21 µm of stroke and Actuation channel

actuation speed of 66 Hz, but the mirror diameter is typically

bellow 1 mm. Piezoelectric mirrors are composed of bimorph Fig. 3. Cross-sectional schematic diagram of the designed varifocal mirror
with hydraulic actuation mechanism.
membranes of piezoelectric materials, whose expansion and
contraction can be controlled by the applied voltage. These
mirrors can be significantly large and present good optical
quality, but the maximum stroke that can be achieved is also when the deflection is much larger than the membrane
very limited. thickness, the bending moments can be neglected [15]. In
Nevertheless none of the varifocal mirrors proposed so far this case, the membrane assumes the shape of a parabola,
is suitable for an endoscopic laser tool, due to the required described by
focal length range and compatibility with high-power laser.
r2
 
Mirrors with small diameters cannot provide tight focusing w(r) = w0 1 − 2 , (1)
at large focal lengths and result in high power density of the R
laser over the mirror, which can lead to damage to the mirror.
where R is the radius and w0 is the center deflection of
On the other hand, mirrors developed with diameter greater
the membrane. The relationship between w0 and the applied
than 2 mm, all present an optical power range lower than
pressure p is given [16] by
20 m-1 . In this work, we present a MEMS varifocal mirror
for an endoscopic laser tool, designed to achieve an optical
power range of at least 40 m-1 . In order to achieve large Eh2
   
4h Eh 7−ν
stroke without limitation on the mirror diameter, hydraulic p= σ0 + w0 + w03 , (2)
R2 2
3R (1 − ν) 3R4 1−ν
actuation has been used. Fig. 1 illustrates how such mirror
could be used in an endoscopic setup. where h is the membrane thickness, σ0 is the residual
The rest of the paper is organized as follows. In section II stress in the membrane and E and ν are respectively the
we present the mathematical model of a varifocal membrane Young’s modulus and the Poisson’s ratio of the membrane
mirror and the developed design of varifocal mirror with hy- material. Since the focal length of a parabola is given by
2
draulic actuation. Next the fabrication method of the mirror f = R /4w0 , the optical power of the mirror with curvature
is described in details in section III. Section IV describes the w(r) is
experimental characterization, while section V presents the
obtained results and a series of discussions. Finally section 1 4w0
P = = 2. (3)
VI presents our conclusions and expectations of future work. f R
II. MIRROR MODEL AND DESIGN Replacing (3) into (2) gives the following relationship
A. Mirror Deflection Model between the applied pressure and the optical power of the
mirror
When uniform pressure is applied to a membrane mirror,
the mirror tends to bend in the direction of the applied
pressure, as shown in Fig. 2. The deflection of the mirror Eh2 EhR2 (7 − ν) 3
 
can be modeled by considering the membrane as a clamped p = h σ0 + 2
P+ P . (4)
3R (1 − ν) 192 (1 − ν)
circular thin plate. This way, the shape of the deflected
membrane is determined by the bending moments acting at Therefore the optical power of the mirror P (p) can be
the rim of the membrane and the elastic stress generated obtained as a function of the applied pressure by inverting
by the straining of the membrane’s neutral fiber. For small (4). This can be done explicitly with Cardan’s formula or
deflections, the bending moments are dominant, however using a numerical approximation model.
 (a) (d) (g)
B. Mirror design with hydraulic actuation
The proposed design of the varifocal mirror with hydraulic
actuation is shown in Fig. 3. The membrane is constructed (b) (e) (h)
from a silicon nitride film deposited on a silicon wafer.
Due to the good adhesion between the silicon and the
nitride film, the rim of the membrane can be considered (c) (f) (i)
as a clamped circular plate. The membrane is released by
etching the silicon bulk from the bottom, creating a fluidic
chamber for actuation. When fluid is pumped in or out of Si Nitride Cr Au Resist Lor 7B
the chamber, uniform pressure is applied to the membrane,
Fig. 4. Schematic diagram of the fabrication process. (a) Si wafer
causing it to deflect as a paraboloid. This allows controlling (100) double-coated with LPCVD super low stress silicon nitride. (b) Cr
the focal length of the mirror without inducing other optical deposition and patterning. (c) RIE of the nitride. (d) DRIE of the Si. (e)
aberrations. The top surface of the membrane is coated with Cr etching and wafer dicing. (f) Si etching with TMAH/IPA solution. (g)
Photoresist patterning with underlying layer of LOR 7B. (h) Ti-Au layer
a gold film to provide high reflectivity to the mid and long- deposition. (i) Resist stripping.
infrared ranges.
The design parameters (i.e. R, h and σ0 ) were defined
based on the requirements of the clinical application. These chamber has been controlled by a microinjection system,
requirements include the focal length range, the diameter connected to the mirror through an actuation channel. This
of the laser spot, the overall dimensions of the varifocal allows designing the mirror almost independently of the
mirror, the maximum required actuation pressure and the actuation system, which facilitates testing different mirror
actuation frequency. In the case of endoscopic laser surgery, designs. The fluid media used was mineral oil. To attach
the focal length of the beam must be shifted from 15 to 45 the mirror to the microinjection system, an aluminum mirror
mm to compensate usual distances to different anatomical holder was used. The holder contains a 10.2 mm square hole
structures, while the diameter of the laser spot must typically for fitting the membrane mirror, aligned with a 2 mm center
be kept lower than 200 µm. The maximum diameter of the hole. When the mirror is in place, an aluminum cover is
varifocal mirror depends on the design of the endoscopic attached to the holder with screws, locking the mirror. A
tool, but should not exceed the few millimeters range. The latex rubber sealing sheet is used to fill the gaps between
actuation pressure must be kept low in order to allow using the mirror and the holder and to prevent leakage of the oil.
a miniaturized actuation system and the actuation frequency The external length of the mirror holder is 40 mm.
would ideally be above 30 Hz. As it can be seen, simulta-
neous fulfillment of all these requirements is challenging to III. MIRROR FABRICATION
achieve.
The fabrication process of the varifocal mirror is shown in
The radius of the mirror determines the diameter of laser
Fig. 4. The process starts with a (100) silicon wafer coated
beam that can be used, therefore limits the numerical aperture
with 500 nm of super low stress low-pressure chemical
of the system. In order to be able to focus the laser beam
vapor deposition (LPCVD) silicon nitride on both sides. A
tightly, the numerical aperture must be
chromium layer of 250 nm is deposited on the bottom of the
R λM 2 wafer and patterned to expose a circular window of 4 mm
NA = > , (5) of diameter. This patterning also exposes the borders of the
f 2πω0
mirror, to allow precisely dicing it from the wafer. Using the
where λ is the wavelength of the beam, M 2 is the patterned chromium layer as a mask, the bottom nitride layer
beam quality factor and ω0 is the radius of the focal spot. is removed with reactive-ion etching (RIE), then the silicon
Assuming a CO2 laser (λ = 10.6 µm), with M 2 = 2.5 bulk is etched with a BOSCH deep reactive-ion etching
and a focal spot of 200 µm of diameter, N A > 0.0422 (DRIE) process, which allows drilling as circular hole in
is required. Considering the maximum focal length as 45 the wafer with vertical walls. Since the BOSCH process
mm, the radius of the mirror must be R > 1.9 mm. For also etches the nitride, the DRIE must be interrupted before
this reason, the mirror radius was designed as 2 mm. The etching the entire wafer. This is achieved by measuring the
membrane thickness affects the actuation pressure, as the etching rate of the DRIE initially and adjusting the etch
later is proportional to the former. However it also affects duration to leave around 30 µm of silicon. Once this process
the mechanical stability of the membrane, therefore it cannot is finished, the wafer is diced into individual mirrors, since
be too low. In this case, a membrane of 500 nm was used. at this point of the fabrication process the borders of the
The residual stress depends mainly on the fabrication process mirrors are already very thin.
and is usually measured after the mirror fabrication. The size After that, the chromium layer and the remaining silicon
of the silicon piece diced from the wafer after releasing the are etched to release the silicon nitride membrane. To main-
membrane has been chosen as 10 mm x 10 mm, in order to tain the hole circular, the remaining silicon is etched with
facilitate handling the mirror. a solution of TMAH with 30% of IPA at 95 ◦ C for about
At this stage of the project, the pressure in the fluidic 8 hours. The TMAH/IPA solution provides almost isotropic
 Microinjection system

Varifocal mirror

Pressure sensor
Arduino
Fig. 5. Photograph of the fabricated 4 mm varifocal mirror in a 10 mm x
10 mm silicon piece (1-cent euro coin shown as reference) Pre-focusing lens

etching of the silicon, as long as the IPA concentration is

kept constant. This is achieved by adding IPA to the solution Wavefront sensor

during the etching process to compensate the evaporated Optical fiber

IPA. After the membrane is released, a gold coating is
applied using a lift-off process. In this process, the top of the
membrane is spun with two layers of photoresist (an S1813 Fig. 6. Experimental setup used for characterizing the fabricated varifocal
mirror.
layer on top of a LOR 7B layer) and patterned to expose a
circular window aligned with the released membrane. Since
the LOR 7B is developed faster than the S1813, it produces the focal length of the beam cannot be measured. Therefore
an undercut in the exposed window, which improves the the mirror deflection curve cannot be fully determined. In
performance of the lift-off. Then a titanium-gold layer (7 order to solve this problem, a pre-focusing lens can be used
nm of titanium and 200 nm of gold) is evaporated on top of for converging or diverging the beam before it reaches the
the wafer. Finally, the resist is removed with solvent, leaving mirror. In this configuration, the focal length of the system
the gold coating only on the circular membrane. Fig. 5 shows is given by
a photograph of the fabricated mirror.
IV. EXPERIMENTAL CHARACTERIZATION 1 1 1
= + , (8)
A. Focal length measurement f fM f0
The performance of the fabricated mirror has been char- where fM is the focal length of the mirror and f0 is the
acterized by measuring the focal length and analyzing the focal length of the system when the mirror is perfectly flat. In
optical quality of a laser beam reflected by the mirror. This this way, the range of fM that cannot be measured changes
has been achieved using a Shack-Hartmann wavefront sensor with f0 . Therefore, the complete fM (p) curve of the mirror
(SHWFS). The SHWFS is composed of a CCD camera and can be obtained by combining the results of characterization
an array of microlenses, which allows sampling the intensity experiments performed with different values of f0 .
and curvature of the incoming beam at multiple locations,
reconstructing the entire wavefront of the beam. The wave- B. Experimental Setup
front is reconstructed using a sequence of orthogonal zernike
Fig. 6 shows the complete experimental setup used for
polynomials [17]. This reconstruction method is convenient
characterizing the fabricated varifocal mirror. The laser
as each zernike polynomial is associated with one type of
source used was a 10 mW fiber-coupled LED with a 200
optical aberration, including defocus, coma, astigmatism and
µm multi-mode optical fiber (Thorlabs, USA). The laser has
spherical aberration. From the defocus coefficient, one can
been pointed at the varifocal mirror within the mirror holder
obtain the radius of curvature of the beam, which is related
with an angle of incidence of 45◦ . To allow adjusting f0 , a
to the distance to the focal spot by:
plano-convex lens has been placed between the mirror and
the fiber tip, inside a linear support. The distance between
  z 2 
r
RoC(z) = z 1 + , (6) the lens and the fiber was adjusted manually. In order to limit
z
the laser beam to the diameter of the mirror, a 2 mm circular
where z is the distance to the focal spot and zr is the
2 aperture was placed before the lens.
Rayleigh range of the beam, given by zr = πω0/λM 2 . When
The wavefront sensor WFS150-5C (Thorlabs, USA) has
z  zr , RoC(z) ≈ z, therefore the focal length of the
been placed in front of the varifocal mirror, also at an
system can be obtained as
angle of 45◦ . The hydraulic pressure applied to the mirror
has been changed using a microinjection system CellTram®
f = d − RoC(z), (7)
vario (Eppendorf, Germany), which has a resolution of 2
with d being the distance between the mirror and the nL. The microinjection system has been controlled by an
measurement plane of the SHWFS. The main limitation of Arduino Nano, which was also used to monitor the pressure
this measurement method is that it can only be used if z  in the actuation channel, measured by an HSC board mount
zr , which means there is an interval around f = d, for which differential pressure sensor (Honeywell, USA).
C. Static Characterization
40
The static characterization of the mirror consisted in Exp 1 (f 0 = 65.56 mm)
Exp 2 (f 0 = -81.95 mm)
obtaining the experimental relationship between the optical 30
Parametric fitting
power of the mirror PM = 1/fM and the applied pressure

Optical power of the mirror (m -1 )

20
in the fluidic chamber. It also included the analysis of the
optical quality of the mirror based on the zernike coefficients 10
of the reconstructed wavefronts. This was performed with
two sets of experiments. In the first one, the pre-focusing 0
lens has been adjusted to converge the laser beam, resulting
in f0 = 65.56 mm. In the second one, the position of the lens -10

was changed making the beam divergent, with f0 = −81.95

-20
mm. For each experiment, the precise value of f0 was
obtained before the experiment by replacing the varifocal -30
mirror with a flat chromium mirror and measuring the radius
of curvature of the reflected beam with the SHWFS. The -40
-6000 -4000 -2000 0 2000 4000 6000
distance between the mirror and the SHWFS was kept Pressure (Pa)
constant, with d = 47.8 mm in both experiments.
For each experiment, the pressure in the fluidic chamber Fig. 7. Experimental results of the static characterization experiments.
was changed between -5000 and 4000 Pa. Sign conventions
are used as defined in Fig. 2, i.e. positive pressure causes
the membrane to deflect in concave shape (positive optical
length of a collimated beam from 32.15 mm to +∞ and
power), while negative pressure results in convex deflection
from −∞ to -30.26 mm. If instead, a convergent beam is
(negative optical power). For each experiment, pressure was
used with f0 = 25 mm, this optical power range could shift
changed in both increasing and decreasing directions, in
the focal length of the system from 13.81 to 143.6 mm. The
order to check for hysteresis.
obtained results also show the absence of hysteresis in the
D. Step Response hydraulic actuation system.
In order to understand the dynamic behavior of the mirror, The model used for the parametric fitting was p = A +
3
the step response of the system was measured. For this, BPM + CPM , and the obtained values for the coefficients
a step input was applied to the mirror by actuating the were A = −244.6, B = 78.71 and C = 0.05965.
microinjection system during an interval of 50 ms, while This parametric fit matches the membrane deflection model
the corresponding change in radius of curvature of the beam described in (4), except for the offset term, A. The presence
was measured continuously with the SHWFS. This step input of this offset in the p(P ) relationship is most likely due to the
corresponded to changing the pressure in the fluidic chamber thermal stress between the membrane and the gold coating,
from 0 to -65 Pa, which was the maximum variation the which was not taken into account on the membrane deflection
microinjection system could perform in this time interval. model and causes the membrane to bend even in the absence
The sampling time of the SHWFS for acquiring the beam of actuation. It can be seen from the plot that when the
and reconstructing the wavefront was 133 ms. The settling applied pressure is 0, the optical power of the mirror is
time of the system has been calculated as the time taken for 3.09 m-1 . The linear term B can be used to estimate the
the optical power of the mirror to stabilize around its steady- residual stress σ0 of the membrane. Assuming the nominal
state value. Since the magnitude of the step input was small values for h and R and using the E and ν coefficients of the
in comparison with the barometric fluctuations present in the nitride only (ignoring the effect of the gold coating), gives
experimental environment (typically around 2 Pa), an interval us σ0 = 112.43 MPa. This result is consistent with typical
of 5% was used instead of 2% to define the stabilization of values for silicon nitride films reported in the literature. The
the system. RMSE between the experimental data and the parametric fit
is 70.12 Pa, which corresponds to an average deflection error
V. RESULTS AND DISCUSSION of 0.86 m-1 .
A. Static Characterization Analyzing the zernike coefficients, the most significant
The results obtained for the static characterization are aberration observed in the beam was astigmatism. The mag-
presented in Fig. 7. This plot shows the measurements from nitude of coefficient Z6 grows with the applied pressure,
both experiments as well as the obtained parametric fitting. reaching -2 µm for p = −5000 Pa. This result is expected
The points with invalid measurement have been omitted from since the laser has been pointed at the mirror with an angle
the plot and excluded from the parametric fit. As it can of 45◦ , which results in off-axis aberrations [18]. In fact, it
be seen, the fabricated varifocal mirror can achieve a large has been shown that when the angle of incidence is large, the
optical power range, changing its curvature from 31.10 m-1 main off-axis aberration is astigmatism. All the remaining
to -33.05 m-1 for a pressure range of 9000 Pa (from 4000 zernike coefficients were bellow 180 nm, including coma and
to -5000 Pa). This optical power range could shift the focal spherical aberration. The coefficient for spherical aberration
 predicted paraboloid. This confirms the result obtained in
3.1
[11] that hydraulic and pneumatic actuation generate uniform
3 pressure over the membrane, resulting in lower spherical
2.9 aberration than electrostatic actuation. However the observed
off-axis astigmatism is an issue as it makes the focal spot
Optical power of the mirror (m-1)

2.8
to become elliptical. This is undesirable as it degrades the
2.7 focusing power of the beam and changes the shape of the
2.6
ablation crater. This problem can be compensated
√ by using
an elliptical mirror with aspect ratio of 2, which is able
2.5
to focus an off-axis beam with incidence angle of 45◦ with
2.4 almost no astigmatism [19].
2.3
The main drawback observed in the proposed mirror is the
actuation speed of the hydraulic actuation system. However,
2.2
it must be noted that the tubes connecting the microinjection
2.1 system to the varifocal mirror are significantly long. This
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Time (ms)
causes the volume of the actuation channel to be much
larger than the volume of the fluidic chamber of the mirror,
Fig. 8. Step response of the system. increasing the inertia of the actuation fluid. In order to
increase the actuation speed, the hydraulic system must be
redesigned for reducing the volume of the actuation channel.
was bellow 45 nm for all pressure values. This would also be important for the miniaturization of the
actuation system.
B. Step Response
Fig. 8 shows the obtained step response of the system. As VI. CONCLUSION
it can be seen the step input shifts the optical power of the
mirror from 3.09 to 2.27 m-1 . The dashed lines indicate the In this work a 4 mm silicon nitride varifocal mirror
5% interval around 2.27 m-1 . The obtained 5% settling time with hydraulic actuation has been presented. The mirror has
was 398 ms, which corresponded to the third measurement of been fabricated using surface and bulk micromachining. The
the SHWFS after the step input. This suggests the maximum hydraulic actuation system used is composed of a microin-
actuation frequency for this mirror to be 2.5 Hz. However, jection system connected to the mirror’s fluidic chamber
since the temporal resolution of the measurement was limited through an actuation channel. This system allowed deflecting
by the low sampling rate of the SHWFS, it is possible for the the mirror in both concave and convex shape.
true settling time of the system to be lower than that. The step The mirror has been tested by shifting the focal length of
response also showed an overshoot, with the mirror achieving a fiber-coupled laser beam. The experimental results showed
a minimum curvature of 2.12 m-1 . This corresponds to an that the mirror can achieve an optical power range of more
overshoot of 18.51%, however since the overshooting part than 64 m-1 , changing its curvature from 31.1 m-1 to -
of the curve is characterized by a single measurement of the 33.05 m-1 , with 9000 Pa of actuation. This curvature change
SHWFS, it is not possible to determine if this corresponds is enough to shift the focal length of a convergent beam
to the maximum overshoot of the system. In order to obtain from 13.81 to 143.6 mm, which would allow controlling
more precise results, it is necessary to replace the SHWFS the focal length of the laser over the entire surgical site.
with a sensing device with higher sampling rate. The mirror presented low optical aberrations, except for the
astigmatism consequent of the off-axis configuration. This
C. Discussion must be compensated by making the mirror elliptical. The
The obtained results show that the proposed varifocal actuation speed of the system was not very fast, due to
mirror can achieve very large optical power range, resulting the large volume of the hydraulic actuation channel, but the
in a focal length range larger than the one required for settling time of the system has been shown to be lower than
endoscopic laser surgery. The static characterization also 400 ms.
showed high precision and good agreement with the mathe- In future works, we expect to improve the mirror design
matical model. This suggests that this kind of mirror could to eliminate the off-axis astigmatism and to improve the
be controlled in open loop, as long as the pre-bending and actuation speed of the hydraulic system. Additionally, we
the residual stress of the membrane are determined. Both expect to test the operation of this mirror with a CO2 fiber
these parameters depend on the fabrication process and can laser to verify its ability to focus the high-power laser.
be determined experimentally, however it would be necessary Furthermore, the overall miniaturization of the actuation
to measure them over a large number of samples to determine system will be required to allow integrating this mirror
the consistency of the fabrication process. in an endoscopic setup. This could involve replacing the
The observed optical aberrations were low, which means microinjection system with a micropump or even switching
the mirror curvature did not deviate significantly from the for an embedded actuation approach.
 R EFERENCES
[1] L. S. Mattos, N. Deshpande, G. Barresi, L. Guastini, and G. Peretti, “A
novel computerized surgeon–machine interface for robot-assisted laser
phonomicrosurgery,” The Laryngoscope, vol. 124, no. 8, pp. 1887–
1894, 2014.
[2] A. S. Jacobson, P. Woo, and S. M. Shapshay, “Emerging technol-
ogy: flexible co2 laser waveguide,” Otolaryngology-Head and Neck
Surgery, vol. 135, no. 3, pp. 469–470, 2006.
[3] P. Janda, R. Sroka, B. Mundweil, C. S. Betz, R. Baumgartner, and
A. Leunig, “Comparison of thermal tissue effects induced by contact
application of fiber guided laser systems,” Lasers in surgery and
medicine, vol. 33, no. 2, pp. 93–101, 2003.
[4] Y. Shao, D. L. Dickensheets, and P. A. Himmer, “3-D MOEMS mirror
for laser beam pointing and focus control,” Selected Topics in Quantum
Electronics, IEEE Journal of, vol. 10, no. 3, pp. 528–535, may 2004.
[5] R. Hokari and K. Hane, “A Varifocal Convex Micromirror Driven by
a Bending Moment,” Selected Topics in Quantum Electronics, IEEE
Journal of, vol. 15, no. 5, pp. 1310–1316, 2009.
[6] T. Sasaki and K. Hane, “Varifocal Micromirror Integrated With Comb-
Drive Scanner on Silicon-on-Insulator Wafer,” Journal of Microelec-
tromechanical Systems, vol. 21, no. 4, pp. 971–980, aug 2012.
[7] A. Lukes, Sarah Jane andPhillip Himmer, E. J. Moog, S. R. Shaw,
and D. L. Dickensheets, “Feedback-stabilized deformable membrane
mirrors for focus control,” Journal of Micro/Nanolithography, MEMS,
and MOEMS, vol. 8, pp. 7–8, 2009.
[8] S. J. Lukes and D. L. Dickensheets, “SU-8 focus control mirrors
released by XeF2 dry etch,” pp. 7930 – 7930 – 6, 2011.
[9] S. J. Lukes, B. J. Lutzenberger, E. Dunbar, S. R. Shaw, and D. L.
Dickensheets, “Variable-focus SU-8 membrane mirror with enhanced
stroke using feedback control,” in 2009 IEEE/LEOS International
Conference on Optical MEMS and Nanophotonics, aug 2009, pp. 141–
142.
[10] M. J. Moghimi, C. R. Wilson, and D. L. Dickensheets, “Electrostatic-
pneumatic MEMS deformable mirror for focus control,” in 2012
International Conference on Optical MEMS and Nanophotonics, aug
2012, pp. 132–133.
[11] M. J. Moghimi and D. L. Dickensheets, “Electrostatic-Pneumatic
Membrane Mirror With Positive or Negative Variable Optical Power,”
Journal of Microelectromechanical Systems, vol. 24, no. 3, pp. 716–
729, jun 2015.
[12] A. A. Alzaydi, J. T. W. Yeow, and S. L. Lee, “Hydraulic controlled
polyester-based micro adaptive mirror with adjustable focal length,”
Mechatronics, vol. 18, no. 2, pp. 61–70, 2008.
[13] J. Brunne, R. Grunwald, and U. Wallrabe, “Adaptive piezoelectric
axicon mirrors for high power femtosecond laser applications,” in 2011
16th International Solid-State Sensors, Actuators and Microsystems
Conference, jun 2011, pp. 1558–1561.
[14] J. Morrison, M. Imboden, T. D. C. Little, and D. J. Bishop, “Elec-
trothermally actuated tip-tilt-piston micromirror with integrated vari-
focal capability,” Optics Express, vol. 23, no. 7, p. 9555, apr 2015.
[15] W. K. Schomburg, “Membranes,” in Introduction to Microsystem
Design. Springer, 2011, pp. 29–52.
[16] F. Zhao, “Nonlinear solutions for circular membranes and thin plates,”
no. March, p. 69260W, 2008.
[17] J. C. Wyant and K. Creath, “Basic wavefront aberration theory for
optical metrology,” Applied optics and optical engineering, vol. 11,
no. part 2, pp. 28–39, 1992.
[18] A. Gomez-Vieyra and D. Malacara-Hernández, “Geometric theory of
wavefront aberrations in an off-axis spherical mirror,” Applied Optics,
vol. 50, no. 1, pp. 66–73, jan 2011.
[19] P. A. Himmer and D. L. Dickensheets, “Off-axis variable focus and
aberration control mirrors,” in MOEMS Display and Imaging Systems,
vol. 4985. International Society for Optics and Photonics, 2003, pp.
296–304.