See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228954364

Sliding-blade MEMS iris and variable optical attenuator

Article  in  Journal of Micromechanics and Microengineering · December 2004

DOI: 10.1088/0960-1317/14/12/015

CITATIONS READS

77 567

4 authors, including:

H. Zou Hadi Veladi

Civil Aviation University of China University of Tabriz
24 PUBLICATIONS   411 CITATIONS    35 PUBLICATIONS   161 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

A High Optical Transmittance and Low Cost Touch Screen without Patterning View project

Microfluidic View project

All content following this page was uploaded by Hadi Veladi on 28 November 2015.

The user has requested enhancement of the downloaded file.

INSTITUTE OF PHYSICS PUBLISHING JOURNAL OF MICROMECHANICS AND MICROENGINEERING
J. Micromech. Microeng. 14 (2004) 1700–1710 PII: S0960-1317(04)81596-1

Sliding-blade MEMS iris and variable

optical attenuator
R R A Syms, H Zou, J Stagg and H Veladi
Optical and Semiconductor Devices Group, Electrical and Electronic Engineering
Department, Imperial College London, Exhibition Road, London SW7 2BT, UK
E-mail: r.syms@ic.ac.uk

Received 24 May 2004, in final form 27 July 2004

Published 14 September 2004
Online at stacks.iop.org/JMM/14/1700
doi:10.1088/0960-1317/14/12/015

Abstract
An iris-type variable aperture fabricated using microelectromechanical
systems (MEMS) technology is described. The device contains a number of
shutter blades, which are each driven by a separate microactuator, and
translated synchronously to create a variable polygonal aperture. The optical
performance of devices with different numbers of blades is compared using
simple analytic models and diffraction theory. The mechanism is simulated
by finite element analysis. Four-blade devices driven by buckling mode
electrothermal actuators are formed by double-sided patterning and deep
reactive ion etching of bonded silicon-on-insulator and characterized
experimentally. Symmetric deflections are obtained, and used to create a
square pupil. Variable attenuation is demonstrated using optical fibres with
thermally expanded cores.

1. Introduction lying on-axis, and are then translated in quasi-tangential

Miniature optical attenuators based on controllable apertures
have a variety of applications, including communications and
imaging systems. Variable optical attenuators (VOAs) with
a single shutter blade as shown in figure 1(a) have been
constructed using several different microelectromechanical
systems (MEMS) technologies [1–3]. Generally, the blade
motion is in plane, but depending on the configuration the
optical beam may travel parallel to or perpendicular to the
substrate. When operated dynamically, such devices may act
as optical choppers [4–6]. Devices with two shutter blades
operating symmetrically, as shown in figure 1(b), have the
advantage of predictable wavelength-dependent loss (WDL),
since the main peak of the far-field diffraction pattern obtained
when an apertured beam exits the device must lie on-axis.
Electrostatically driven twin-blade devices have recently been
demonstrated [7], and a translating aperture has been used in
a spectrometer [8].
More complex aperture configurations have so far
attracted little attention. However, as we show here, a two-
blade device is merely a special case of a general N-blade
polygonal shutter, which can be created from a set of
symmetric triangular elements. These elements are initially
tiled together to close the pupil, with one of their vertices
directions to open the pupil, with each element sliding against
its neighbours. Figures 1(c) and (d ) show four-blade and
eight-blade devices based on this principle, respectively. In
the former case, a square aperture is created, and in the latter,
an octagonal one. These devices may all be constructed in
‘normally open’ variants.
As the number of blades rises, the pupil tends to a
circle. Pupils with increased symmetry may be useful in
VOAs with low WDL and polarization-dependent loss (PDL).
Alternatively, when combined with other variable components,
such as deformable mirrors [9] and movable or tunable lenses
[10, 11], they may find applications in miniature imaging
systems [12] or adaptive optics [13].
The use of a multi-blade shutter is clearly similar to the
conventional photographic iris, and hence we refer to it here
as an iris VOA. In a macroscopic iris, the blade motions are
linked by a sliding rotary drive. Using MEMS technology,
which allows complex mechanisms to be duplicated very
simply and accurately, it is possible to obtain a similar effect
using a number of entirely separate linear actuators operating
together. In a two-blade device, the shutter blades will clearly
strike each other if driven in the opposite direction to that
used to open the pupil. However, in a four-blade device, the
blades may be driven in either direction without colliding. This

0960-1317/04/121700+11$30.00 © 2004 IOP Publishing Ltd Printed in the UK 1700

 Sliding-blade MEMS iris and variable optical attenuator

Shutter TEC fibre Iris TEC fibre

blade z1 z2
Optical
beam 2w0 Blade
motion
(a) a
E f (x, y) 2a Ea E a† Ef
Input A(x, y) Output
(a)
2a
1.0

(b)
0.8

0.6 N=1
N=2

η
N=4
2a
0.4 N =∞

0.2

(c)
0.0
-1.0 -0.5 0.0 0.5 1.0 1.5 2.0
(b) Normalized aperture size a/w0

Figure 2. (a) Optical model of an iris-type VOA; (b) variation of

transmission with normalized aperture a/w0 , for irises with
2a different numbers of blades, assuming that z1 = z2 = 0.

in figure 2(a). Here, a fibre-coupled device is constructed by

(d)
Optical beam
Blades
(e) Substrate
Figure 1. Shutter-type VOAs with (a) one, (b) two, (c) four and
(d ) eight blades; (e) cross-section of a through-substrate device.
characteristic allows some protection against damage caused
by under-damped oscillations. If the blades and their motions
lie in plane, a complete iris may be formed, by removing the
substrate beneath the pupil as shown in figure 1(e).
In this paper, we consider such multiple blade
arrangements. We begin in section 2 by presenting simple
scalar models of their optical performance that both ignore and
include the effects of diffraction. Configurations for MEMS
devices based on different actuation methods are considered
in section 3, and a simple buckling mode electrothermal
mechanism is simulated by finite element analysis. Prototype
four-blade devices with electrothermal drives are demonstrated
by deep reactive ion etching of bonded silicon-on-insulator
(BSOI) in section 4. They are characterized electrically,
mechanically and optically, and a complete VOA based on
placing the pupil between two cleaved fibre ends. To reduce
diffraction, fibres with thermally tapered ends that support an
expanded optical mode may be used as the input and output
coupling optics [14, 15], following arrangements developed
for other in-line systems [16].
2.1. Diffractionless model
The overall transfer efficiency η may be estimated when the
axial separations z1 and z2 are so close to zero that diffraction
may be neglected, or when they are sufficiently large that it
must be taken into account. In each case, the overall efficiency
can be written as η = ηA ηC . Here ηA is a term describing the
effect of the aperture on the field emitted by the input fibre and
ηC describes the subsequent coupling efficiency into the output
fibre. If the guided mode supported by the two fibres is defined
by the transverse electric field distribution Ef (x, y), and the
field just after the aperture is Ea (x, y) = A(x, y) Ef (x, y),
where A(x, y) is a pupil function, the efficiency can be found
from standard scalar theory as [17]
ηA =
Ea , Ea /
Ef , Ef 
ηC = |
Ea , Ef |2 /{
Ea , Ea 
Ef , Ef }. (1)
η = |
Ea , Ef | /{
Ef , Ef }
2 2

fibres with thermally expanded cores (TECs) is demonstrated. Here,
E1 , E2  is the inner product of two fields E1 (x, y) and
Conclusions are presented in section 5. E2 (x, y), given by



2. Optical model of an iris VOA
E1 , E2  = E1 E2∗ dx dy. (2)
A

We begin by constructing theoretical optical models of a Here ∗ denotes complex conjugation, and the integral is taken
multiple-blade iris VOA, based on the arrangement shown over the whole (x, y) plane.

1701
R R A Syms et al

Assuming that a single Gaussian mode  is supported

 by  2s
both the fibres, we may write Ef = E0 exp −(x 2 + y 2 ) w02 ,
where w0 is the beam radius. The effect of the aperture
may then be described simply by limiting the field Ea at its
extremity. In the case of a square pupil, for example, we may
write A(x, y) = 1 for |x|
a and |y|
a, and A(x, y) = 0
for |x| > a and |y| > a and where 2a is the width of the pupil. 2(a+s)
Integrals of the type shown in equation (1) may often be
evaluated analytically. Defining ηN as the transfer efficiency
of an iris with N blades, we obtain for example:
√
η1 = {1 + erf( 2a/w0 )}2 /4 (a)
√
η2 = erf 2 ( 2a/w0 ) 0
√ (3)
η4 = erf 4 ( 2a/w0 ) -5
   2

Transmission (dB)
s/w0 = 0.1000
η∞ = 1 − exp −2a 2 w02 -10 s/w0 = 0.0500
where erf(x) is the error function. Equations (3) suggest that s/w0 = 0.0250
-15 s/w0 = 0.0125
the efficiencies depend only on the normalized aperture a/w0 , s/w0 = 0.0000
as shown in figure 2(b). Apart from the case of a single -20 (i) w = 25 µm, z = 50 µm
blade (η1 ), there is little to choose between the functions;
the efficiency rises smoothly from zero at a/w0 = 0, when -25 (ii) w = 4 µm, z = 50 µm
the pupil is entirely closed, and tends to unity at around
a/w0 = 2. For irises with more than four blades, the variations -30
lie between the curves for N = 4 and N = ∞ (circular 0.0 0.5 1.0 1.5 2.0
aperture). For the single-blade iris, the transfer efficiency is (b) Normalized aperture size a/w0
0.25 when a/w0 = 0, because only half the aperture is then
obscured; negative values of a/w0 are needed to reduce the 35
efficiency to zero.
Maximum attenuation (dB)

From figure 2(b), we may see that an iris with one blade 30
requires a single actuator with a travel of around 3w0 to cover 25
the useful range of operation, while an iris with two blades
requires two separate actuators, each with a travel of 2w0 . 20
Similarly, a device with four blades requires four actuators;
15
because each moves in a tangential direction (figure 1), the
travel needed to achieve an aperture a = 2w0 is increased 10
√
to 2 2w0 . Devices with even larger numbers of blades
5
require an even longer travel, tending to infinity for perfectly
circular pupils. Hence, a four-blade device may be a useful 0
compromise between increased symmetry and complexity. As 0.0 0.1 0.2 0.3 0.4 0.5
w0 increases, the travel length must also clearly increase. (c) Normalized blade offset s/w0

Figure 3. (a) Four-blade iris with offset blades; (b) variation of

2.2. Blade clearances
In practice, it may be convenient to fabricate the whole
mechanism using single-layer patterning. Since there must
be a finite clearance between the blades, the aperture may
never be fully closed in this case. Figure 3(a) shows the layout
of a four-blade iris, with each blade offset by s from the ideal
position shown in figure 1(c), in both the x- and y-directions.
With this modification, the expression previously found for the
transfer efficiency becomes
√ √
η4 = {erf 2 ( 2(a + s)/w0 ) + erfc( 2(a + s)/w0 )
√ √
× [erf( 2(a + s)/w0 ) − erf( 2(a − s)/w0 )]}2 (4)
where erfc(x) is the complementary error function. Figure 3(b)
shows the variation in dB of η4 with a/w0 , for different values
of the normalized offset s/w0 . As can be seen, the iris now
works less effectively as an attenuator, and the transmission
transmission with a/w0 , for different values of normalized blade
offset s/w0 ; (c) variation of maximum attenuation with s/w0 .
steadily increases as s/w0 rises. The minimum transmission
η4 min is obtained when a/w0 = 0, when the blades are aligned
as shown in the left-hand diagram in figure 3(a). η4 min may
be obtained from equation (4) as
√ √
η4 min = erf 2 ( 2s/w0 ){1 + erfc( 2s/w0 )}2 . (5)
Figure 3(c) shows the variation of η4 min with s/w. To achieve
an attenuation of more than 30 dB, we require s/w0 < 0.01.
Similarly, for 20 dB maximum attenuation, s/w0 < 0.03 is
needed. Although both these values are small, s itself may
lie in the micron range if w0 is sufficiently large. Single-
layer fabrication may therefore be used to construct devices
designed for use with expanded optical beams.

1702

2.3. Effect of diffraction
When the separations z1 and z2 are non-zero, diffraction must
be taken into account. Unfortunately, the output fibre may
effectively lie in the near or far field, depending on the pupil
size. Diffractive analyses of one- and two-blade MEMS VOAs
have already been carried out [7, 18]. According to scalar
theory, propagation from the input fibre to the aperture may be
described by the standard rules for Gaussian beams (see e.g.
[19]). Ignoring some distance-dependent phase terms, such a
beam can be described by the field:
E(r, z) = A0 (w0 /w) exp{−(x 2 + y 2 )/w2 }
× exp{−jk0 (x 2 + y 2 )/2R}. (6)
Here, A0 is a constant, the term (w0 /w) is required for power
conservation in an expanding beam, exp{−(x 2 + y 2 )/w2 }
describes the radial amplitude variation, while exp{−jk0 (x 2 +
y 2 )/2R} is the radial variation in phase. The parameter w
is the beam radius away from the waist, R is the radius of
curvature of the phase-front and k0 = 2π/λ, where λ is the
wavelength. The values of w and R vary with distance z from
the waist according to
w 2 = w02 {1 + (z/z0 )2 } R = z{1 + (z0 /z)2 }. (7)
 increasingly complex diffraction patterns are obtained in the
Here, z0 = k0 w0 2 is a characteristic distance measured in
2
the direction of propagation, which represents the transition
between near-field and far-field behaviour.
To find the field falling on the output fibre from the field
emerging from the aperture, more general diffraction theory
must be used. Again according to scalar theory (see e.g. [20]),
the diffraction of an optical field E specified on the (x, y)
plane to a field E † on the (x † , y † ) plane a distance z away is
described by the integral:



E † (x † , y † ) = {f (θ)/jλ0 r}E(x, y) exp(−jk0 r) dx dy.
A
(8)
√ † †
Here r = {(x − x) + (y − y) + z } is the distance between
2 2 2
the points (x, y) and (x † , y † ), and f (θ) is a slowly varying
angular function designed to exclude backward travelling
waves. For normal incidence on the object plane, we may
take f (θ) = {1 + cos(θ)}/2, where θ is the angle between
the z-axis and the vector r = (x † − x)i + (y † − y)j + zk.
Because the most severe diffraction occurs after the aperture,
we assume that z1 = 0 and concentrate on the problem of
determining the field Ea† at the output fibre using equation (8).
The new field Ea† may then be used in equation (1) to find the
effect on efficiency.
To illustrate the effect of diffraction, figure 3(b) shows
two further sets of data found by numerical integration. In
case (i) we assume that w0 = 25 µm and z2 = 50 µm,
and in case (ii) that w0 = 4 µm and z2 = 50 µm. For
case (i), the numerical data are virtually super-imposed on the
analytic result, suggesting that diffraction is unimportant for
expanded beams and small distances. For (ii), they differ from
the analytic result, and transmission is reduced even when the
aperture is fully open. Since the beam is now Gaussian at
the output fibre, this limiting efficiency may be found from
equations (6), (7) and (1) as
 2 2   2 
ηmax = 4w2 w02 w + w02 + k0 w 2 w02 2R . (9)
It is simple to verify that equation (9) reproduces the results
Sliding-blade MEMS iris and variable optical attenuator
of the diffraction integral under these conditions. When
the aperture is partly open, the transmission is also reduced
compared with the analytic result.
These results can be understood from typical diffraction
patterns at the output fibre. The patterns in figures 4(a) and
(b) show contours of equal intensity for four-blade devices in
cases (i) and (ii), respectively, for different values of a/w0 and
zero blade clearance (s/w0 = 0). The results are normalized
to a peak intensity of unity, and the iris is fully open in the left-
hand figure and increasingly closed moving to the right-hand
figure. Although the output beam gradually reduces in size, it
lies in the near field in all the figures shown. Only when the
pupil is closed further still does significant diffraction occur.
Considerably more diffraction occurs in case (ii). Here the
output lies in the far field throughout, and the beam spreads
even when the aperture is fully open.
The effect of clearance slots is to introduce additional
fringes into the diffraction pattern. For example, figure 4(c)
shows diffraction patterns obtained for case (i), for different
values of the normalized blade clearance s/w0 , assuming now
that the blades are fully aligned (a/w0 = 0). The clearance
is gradually reduced moving to the right-hand figure, and
process.
3. MEMS sliding blade iris mechanism
In this section, we consider configurations for a MEMS
sliding blade iris. A variety of suitable actuators exist,
including electrostatic, electromagnetic, electrothermal and
piezoelectric types [21]. Electrostatic and electrothermal types
may be fabricated directly in silicon, without the need for
other functional materials. Electrothermal actuators might be
preferred for their simplicity, and electrostatic actuators for
small power consumption and large bandwidth. In each case,
relatively long travel is required for use with expanded mode
fibres, which are needed to reduce the effect of diffraction.
Long travel in turn requires a large actuator span, so that the
mechanism will inevitably be much larger than the pupil.
3.1. Actuation and static response
Figure 5(a) shows how a four-blade iris might be driven by
a set of four buckling-mode electrothermal actuators [22–24].
Here, paired beams of length L1, width W1 and separation S1
have been used to obtain straight-line motion. These beams
support a shutter blade of breadth B on two rigid beams of
larger width W2 and separation S2 . The actuators are connected
electrically, so they are all driven simultaneously by passing
a current between diagonal corners. In this configuration, the
circuit consists of two actuators in series, in parallel with two
other similarly arranged actuators.
When the current flows, the temperature of the actuator
beams rises, so that they expand due to linear thermal
expansion. However, being constrained at their roots, they
buckle laterally. If the beams have a high aspect ratio, and are
given an initial central offset , in-plane motion in the outward
direction may be forced. Furthermore, this motion may be
continuous (unlike a straight beam, which buckles only at a
1703
R R A Syms et al
a/w0 = 2 a/w0 = 1
(a)
a/w0 = 2 a/w0 = 1
(b)
s/w0 = 0.2 s/w0 = 0.1
(c)
Figure 4. Diffraction patterns from a four-blade iris for different values of a/w0 , assuming that z1 = 0, s/w0 = 0, and (a) w0 = 25 µm and
z2 = 50 µm and (b) w0 = 4 µm and z2 = 50 µm; (c) patterns obtained for case (a), assuming now that a/w0 = 0 and s/w0 varies.
critical axial load). A separation D is provided between each
shutter and its neighbour’s support beams, to allow overshoot.
Alternatively, figure 5(b) shows how a four-blade device
might be driven by a set of electrostatic comb drive actuators,
which are now equipped with folded portal frame flexure
mechanisms to ensure straight-line motion over the distance
involved [25]. The fixed electrodes are connected by
the surrounding land and the moving electrodes by the flexure
suspension system, so that all four drives may be operated
from a single voltage source.
The operation of buckling mode actuators is well
understood (see e.g. [23]), and here we give only brief details.
Figure 6(a) shows a simulation of an electrothermal device in
its thermally excited state, carried out using the commercial
finite element software ANSYS 6.0, which illustrates the
operational principle. The assumed dimensional parameter
values were L1 = 2 mm, W1 = 8 µm, S1 = 200 µm, W2 =
20 µm, S2 = 100 µm, B = 300 µm and  = 8 µm.
A Young’s modulus of 168 GPa, a Poisson’s ratio of 0.22
and a density of 2330 kg m−3 were used, to model a device
1704
a/w0 = 0.6
a/w0 = 0.6
s/w0 = 0.05
formed in single crystal silicon. The other material parameters
were a resistivity of 2300  m, a linear thermal expansion
coefficient of 2.06 × 10−6 ◦ C−1 and a thermal conductivity of
150 W m−1 ◦ C−1. The elements were of type SOLID98. Here,
translation of the iris blades in the x and y directions results in
the creation of a square pupil, whose axes are rotated by 45◦
from the directions of motion.
3.2. Dynamic behaviour
The bandwidth of an electrothermal device is defined by
the first-order time constant associated with a thermal drive,
and the mechanical time constant of the mass–spring–damper
system formed by the suspension and suspended parts.
Because an electrothermal mechanism normally has a small
mass, the thermal time constant is likely to be longer in a
practical geometry. Furthermore, the power consumption is
continuous, and determined by the amount of conduction and
convection cooling. In contrast, in an electrostatic device,
the bandwidth is mainly determined by the mechanical time
 Sliding-blade MEMS iris and variable optical attenuator

Shutter blade W1
L1
D
Shutter B ∆ I
support
beam
Via hole S2 W2

S1

Electrothermal drive

(a)
Motion Land
(a)

Electrostatic
comb drive

Straight-line
mechanism

(b)

(b)

Figure 5. Four-blade MEMS iris VOAs with (a) electrothermal and

(b) electrostatic drives.

constant, which is typically lower due to the increased mass

associated with comb electrodes, and power is consumed only
when the device is changing state.
The dynamical performance of an electrothermal device
may again be simulated using ANSYS. For example,
figures 6(b) and (c) show the two lowest order in-plane
eigenmodes of a single shutter blade driven by a buckling mode
electrothermal actuator, assuming the same parameters as
before. The fundamental mode occurs at ≈2.4 kHz frequency, (c)
and predominantly involves flexing of the drive beams. The
second mode mainly involves flexing of the shutter support Figure 6. (a) ANSYS simulations of a four-blade electrothermal
iris; (b) and (c) the lowest order in-plane mode shapes.
beams; because these are stiffer, this resonance occurs at a
higher frequency (≈5.3 kHz). Higher-order in-plane modes
of each type exist, together with additional out-of-plane modes phase, the pupil will open and close sinusoidally. However,
that involve twisting of the drive beams. slight differences in the resonant frequency or quality factor
Although the actuators are connected electrically, (Q-factor) of each oscillator will cause the blades to strike each
the different parts of the mechanism are not connected other. If all four actuators operate on the second mode, they
mechanically. Consequently, the four mass–spring oscillators will inevitably collide.
operate independently and do not display the collective motion All the modes above may be excited by mechanical
of a coupled oscillator system. If all four oscillators operate on noise or by shock inertial loads; however, the resonant
the fundamental in-plane mode, with the same amplitude and frequencies are high enough that some immunity is likely.

1705
R R A Syms et al
1. Substrate etch Bonded silicon
Oxide Substrate
2. Bonded layer etch
3. Sacrificial layer etch
4. Metallization Iris mechanism
Pupil
Figure 7. Process flow for device fabrication.
More importantly, electrical transients may provide suitable
excitation. To avoid damage, it may be important to develop a
mechanical linkage between neighbouring oscillators to force
phase- and frequency-locked operation, and this aspect is
currently being considered.
4. Fabrication and testing
In this section, we describe the fabrication and characterization
of prototype four-blade iris VOAs, driven by electrothermal
buckling mode actuators. As described before, small
clearances are required between the blades, which prevent
complete optical extinction. To close the openings,
overlapping features could be fabricated using multi-layer
processing. Here, we have ignored this aspect, and used a
single layer mechanism.
4.1. Device fabrication
Devices with the layout of figure 5(a) were fabricated by deep
reactive ion etching of BSOI [26]. The material was obtained
commercially from Analog Devices Belfast, and consisted of
4 in diameter (100) oriented wafers with a 2 µm thick SiO2
interlayer and a 25 µm thick bonded Si layer. The main device
parameters were as given in the previous section;√however, to
allow clearance, the blades were offset by s = 5/ 2 µm, and
to allow overshoot, a separation D = 10 µm was provided
between each blade mechanism and its neighbour.
Complete iris VOAs were formed by double-sided
processing as shown in figure 7. First, the substrate side
was patterned with the via hole mask, and etched down to the
oxide interlayer (step 1). Deep etching was performed using
a Surface Technology Systems inductively coupled plasma
1706
(a)
(b)
Figure 8. SEM photographs of a four-blade electrothermal MEMS
iris VOA as fabricated in BSOI: (a) overall layout, and (b) close-up
of shutter blades.
etcher, using a photoresist hard mask and cyclic etching with
SF6 and passivation with C4F8 [27]. The hard mask was
then removed by wet stripping followed by plasma etching
in O2. The device layer side was then patterned using a mask
aligner with through-wafer infrared illumination, taking care
not to fracture the thin layer over the pupil during spin coating
and exposure. The device layer was then patterned using the
mechanical layer mask, and etched again to the oxide, which
typically cracked slightly when the exposed silicon had been
completely removed (step 2). A small increase in pressure
caused by a leakage of the helium back-side coolant was then
a good indication that etching was complete. The hard mask
was then stripped.
The sacrificial oxide was then etched from beneath the
movable parts using buffered hydrofluoric acid (step 3). The
wafer was then freeze-dried in a water/methanol mixture and
finally metallized with 100 Å Cr and 300 Å Au by conformal
sputtering to allow electrical contacts (step 4). Electrical
isolation between different suspended parts is provided by a
large lateral undercut, which prevents metal tracking between
the pads. Figure 8 shows scanning electron microscope
photographs of a complete device. Figure 8(a) shows the
 Sliding-blade MEMS iris and variable optical attenuator

(i) (ii)

(iii) (iv)
(a)

50
Horizontal
Actuator displacement (µm)

40 actuator pair

30 Vertical
actuator pair
20

10

0
Blades aligned
-10
0.0 0.2 0.4 0.6 0.8 1.0 1.2
(b) Electrical power (W)

Figure 9. (a) Optical microscope views of iris at drive powers of (i) 0 mW, (ii) 80 mW, (iii) 460 mW, (iv) 1.1 W; (b) experimental variation
of actuator displacement with electrical drive power.

overall layout, while figure 8(b) shows a close-up of the
shutter blades. There is good alignment of the suspended
structure.
4.2. Electromechanical characterization
Electrical characterization was performed by passing a current
between two diagonally opposed contacts. Measurements of
deflection were obtained using an optical microscope equipped
with a ×20 objective, a video camera and a calibrated on-
screen cursor measurement system. The power delivered into
the device was determined as a product of the voltage across
the contacts and the current flowing. The device resistance
was 52  at zero power, rising slowly to around 58  as the
power increased to 1.1 W.
The blades deflected smoothly as the drive power
increased. However, some discrepancies from ideal behaviour
were observed. For example, figure 9(a) shows optical
microscope views of the blade positions at different powers.
There is a small deflection of the blades at zero power
(photograph i), in the opposite direction to that obtained at
moderate power (80 mW in photograph (ii), rising to 1.1 W
in (iv)). The static deflections are slightly different in the
horizontal and vertical directions. These results suggest that
the actuators each have a small static offset, since the blades
should be aligned at zero power. The offsets are attributed to
tensile stress in the Cr metallization, which acts to straighten
the pre-buckled suspension.
From these photographs, the actuator displacements may
be found. Figure 9(b) shows the variations of the horizontal
and vertical deflections with drive power, found as the average
of the deflection provided by the two actuators concerned in
each case. The two variations are very similar. The static
deflection is negative, with an average value of ≈−3 µm and
a difference of ≈2 µm between the horizontal and vertical
directions. The deflection increases quasi-linearly with power.

1707
R R A Syms et al
(i)
(a)
Peak-to-peak actuator deflection (µm)
100
10
1
.1
(b)
Figure 10. (a) Dark field optical microscope photographs of iris (i) stationary and (ii) oscillating; (b) frequency response characteristic.
Zero deflection, or ideal blade alignment, is achieved at
≈0.08 W. The maximum deflection achieved is ≈42 µm, at
1.1 W (or ≈8 V drive voltage). This deflection is equivalent
√
to an aperture size of a = 42/ 2 = 30 µm, and is limited
by thermal damage to the metal coating of the actuator beams,
which occurred at around 1.2 W power.
The frequency response was measured from dark field
optical microscope photographs of the type shown in
figure 10(a). Here, photograph (i) shows the blade positions
obtained with the device stationary, while photograph (ii)
shows the blades oscillating in synchronism. The blurred lines
represent the motion of small dust particles. Figure 10(b)
shows the variation of peak-to-peak actuator displacement
with drive frequency, obtained using a 15 V p-p sinusoidal
voltage. Since the device is power driven, the mechanical
response is at twice the frequency shown. The low frequency
roll-off (i.e., the 3 dB point) of the drive is at ≈100 Hz. With
this bandwidth, only a single mechanical resonance could be
excited, at ≈1.7 kHz, corresponding to a mechanical frequency
of 3.4 kHz. The motion was of the type shown in photograph
(ii) of figure 10(a), and hence corresponded to the fundamental
in-plane mode.
The resonance had a high (>10) Q-factor, but it was found
that the exact frequency and Q-factor both depended on several
aspects. Firstly, the resonant frequencies for the horizontal
and vertical actuators were slightly different, suggesting a
slight lack of symmetry in the device shape. Secondly, the
values of the resonant frequency and the Q-factor were strongly
dependent on the drive power. The frequency increased very
1708
(ii)
Drive Mechanical
roll-off resonance
1 10 100 1000 10000
Electrical drive frequency (Hz)
significantly as the power (and hence the oscillation amplitude)
rose, varying by a factor of approximately 2 (from 2.4 kHz
to 4.2 kHz) over the range investigated. The lower frequency
corresponds broadly with the prediction of the ANSYS model.
As the frequency rose, the Q-factor decreased, and there was
evidence of nonlinear behaviour. These effects are attributed
partly to alterations in the spring constant of the suspension
with temperature and partly to spring hardening.
4.3. Optical characterization and VOA operation
Complete variable optical attenuators were constructed using
TEC fibre from Kadence Photonics [28] for input and output
coupling. The mode field radius w0 at the thermally treated
end was first estimated from a butt-coupling experiment.
Two TEC fibres were coupled together and the transmission
measured as one fibre was translated laterally using a calibrated
micropositioner as shown in figure 11(a). For two similar
Gaussian modes, the transmission across the joint varies with
lateral offset δ as [17]
  
η = exp −δ 2 w02 . (10)
Figure 11(b) shows the experimental variation of transmission
at 1.55 µm wavelength, which shows good agreement with
equation (10) for w0 = 25 µm. The displacement in figure 9
then corresponds to a/w0 ≈ 1.2, enough to cover most of the
range of VOA operation.
Figure 11(c) shows the experimental arrangement of an
iris VOA. The two TEC fibres and the iris are each held
on a separate micropositioner, which was used to optimize

TEC fibre δ TEC fibre
(a) Input
1.0
0.8
Coupling efficiency
0.6
0.4
0.2 Experiment
Gaussian model
0.0
-50 -40 -30 -20 -10 0 10 20
(b) Lateral offset δ (µm)
(c)
Figure 11. (a) Arrangement for measurement of TEC fibre;
(b) experimental and model transverse field profile of TEC fibre;
(c) arrangement for evaluation of a four-blade iris VOA.
transverse alignment and set axial separations. The iris
was tilted to reduce Fabry–Perot interference, and back-
reflected light was coupled into the cladding of the input fibre.
Transmission was measured using a broadband source centred
on 1.55 µm wavelength (an Agilent 83438A erbium ASE
source) and a Ge photodiode. Figure 12(a) shows the variation
of transmission with electrical power, for z1 = 20 µm and
z2 = 80 µm. These results omit the static insertion loss (2 dB
at these separations). The transmission initially decreases,
as the static offset described earlier is removed, reaching
a minimum at ≈0.1 W. The transmission then increases
smoothly to a maximum as the power is raised, in broad
agreement with figure 2(b). A maximum relative attenuation
of ≈17 dB was obtained, when the blades are aligned.
Superimposed on these data is a theoretical simulation,
based √on equation (4). A normalized offset of s/w0 =
5/(25 2) = 0.1414 was assumed from previous results, and
the variation of the normalized aperture size a/w0 with power
was obtained by fitting a quadratic to the data of figure 9(b).
The theoretical prediction is similar to the experimental data,
Sliding-blade MEMS iris and variable optical attenuator
1.0
0.8
Transmission
0.6 Theory
Output
Experiment
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
(a) Electrical power (W)
30 40 50
(b)
Figure 12. (a) Variation of transmission with electrical drive power
for a four-blade MEMS iris VOA; (b) small-signal step response.
but larger peak attenuation is obtained experimentally. This
discrepancy can be ascribed to diffraction, which spreads light
passing through the blade clearances away from the output
fibre.
Figure 12(b) shows the small-signal step response of the
VOA. To obtain these data, a fixed voltage was first applied,
so that the iris was partially opened, and an additional square
wave of 40 Hz frequency was then applied. Here, the entire
voltage variation is shown, while the time-varying part of the
detected signal has been enhanced. Following an upward step
in voltage, the optical signal rises slowly to a maximum, due
to the first-order time constant of the thermal drive (≈1.6 ms).
Superimposed on the slow rise are rapid oscillations,
arising from the fundamental in-plane mechanical resonance
described earlier. The exact frequency again varied with the
drive power; here it is ≈4 kHz. The damping is light, and the
oscillations take time to die away. Here, the 1/e decay time is
t ≈ 5 ms. Amplitude modulation of the oscillations was also
observed; this can be ascribed to ‘beating’ of the resonances of
the individual actuators. However, the low bandwidth of the
drive limits its ability to excite large oscillation amplitudes, so
that catastrophic overshoot does not occur.
5. Conclusions
In conclusion, we have demonstrated a new type of variable
attenuator, which operates by synchronous motion of a set of
shutters, which are each driven by a separate linear actuator to
1709
R R A Syms et al
form a sliding blade iris. The optical performance of devices
with different numbers of blades has been compared, and
models have been constructed that both ignore and include the
effects of diffraction. It has been shown that useful multiple-
blade arrangements can be constructed for use with expanded
mode fibres by single-layer patterning, without the clearances
between blades excessively limiting the maximum achievable
attenuation.
The device is suitable for realization using
microelectromechanical systems technology and four-blade
devices with electrothermal drives have been fabricated by
double-sided patterning and deep reactive ion etching of
bonded silicon-on-insulator material. Electromechanical
characterization has been performed and near-synchronous
blade motions have been obtained. Departures from ideal
performance have been identified. A demonstration VOA
has been constructed using expanded core fibres, and optical
attenuation over a range of 17 dB has been demonstrated.
Several potential performance improvements are now being
investigated. These include a reduction of drive power, an
increase in dynamic range and improvement of high-frequency
mechanical performance.
Acknowledgment
The authors are grateful to the Engineering and Physical
Sciences Research Council for financial support under grant
GR/R07844/01.7.
References
[1] Barber B, Giles C R, Askyuk V, Ruel R, Stulz L and Bishop D
1998 A fiber connectorized MEMS variable optical
attenuator IEEE Photon. Technol. Lett. 10 1262–4
[2] Marxer C, Griss P and de Rooij N F 1999 A variable optical
attenuator based on silicon micromechanics IEEE Photon.
Technol. Lett. 11 233–5
[3] Wood R, Dhuler V and Hill E 2000 A MEMS variable
attenuator IEEE/LEOS Int. Conf. on Optical MEMS
(Kauai, Hawaii, 21–24 Aug. 2004) pp 121–2
[4] Ching M T, Brenner R A and White R M 1994
Microfabricated optical chopper Opt. Eng. 33 3634–42
[5] Marxer C, Grétillat M A, Jaecklin V P, Baettig R,
Anthamatten O, Vogel P and de Rooij N F 1996 Megahertz
opto-mechanical modulator Sensors Actuators A 52 46–50
[6] Toshiyoshi H, Fujita H and Ueda T 1995 A piezoelectrically
operated optical chopper by quartz micromachining
IEEE/ASME J. Microelectromech. Syst. 4 3–9
[7] Li L and Uttamchandani D 2004 Design and evaluation of a
MEMS optical chopper for fibre optic applications
IEE Proc. Sci. Meas. Technol. 151 77–84
1710
View publication stats
[8] Vuilleumer R and Kraiczek K 1995 Variable-entrance-slit
system for precision spectrophotometers Sensors Actuators
A 50 87–91
[9] Bifano T G, Krishnamoorthy-Mali R, Dorton J, Perreault J,
Vandelli N, Horenstein M and Castanon D 1997
Continuous-membrane surface micromachined silicon
deformable mirror Opt. Eng. 36 1354–60
[10] Kwon S, Milanovic V and Lee L P 2002 Large-displacement
vertical microlens scanner with low driving voltage
IEEE Photon. Technol. Lett. 14 1572–4
[11] Krupenkin T, Yang S and Mach P 2003 Tunable liquid
microlens Appl. Phys. Lett. 82 316–8
[12] Volkel R, Eisner M and Weible K J 2003 Miniaturized imaging
systems Microelectr. Eng. 67–8 461–72
[13] Paterson C, Munro I and Dainty J C 2000 A low cost adaptive
optics system using a membrane mirror Opt. Exp. 6 175–85
[14] Hanafusa H, Horiguchi M and Noda J 1991
Thermally-diffused expanded core fibers for low-loss and
inexpensive photonic components Electron. Lett. 27 1968–9
[15] Kihara M, Matsumoto M, Haibara T and Tomita S 1996
Characteristics of thermally expanded core fiber IEEE J.
Lightwave Technol. 14 2209–14
[16] Sato T, Sun J, Kasahara R and Kawakami S 1999 Lens-free
in-line optical isolators Opt. Lett. 24 1337–9
[17] Syms R R A and Cozens J R 1992 Optical Guided Waves and
Devices (London: McGraw Hill)
[18] Dai X H, Zhao X L, Cai B C and Cao Y 2002 Modelling and
characterization of diffractive optical propagation inside
MEMS variable optical attenuator SPIE Proc. 4920 148–51
[19] Marcuse D 1972 Light Transmission Optics (New York: Van
Nostrand Reinhold)
[20] Born M and Wolf E 1980 Principles of Optics 6th edn
(Oxford: Pergamon)
[21] Fujita H 1998 Microactuators and micromachines Proc. IEEE
86 1721–32
[22] Sinclair M J 2000 A high force low area MEMS thermal
actuator Proc. 7th Intersoc. Conf. on Thermal and
Thermomechanical Phenomena ITHERM 2000 (Las Vegas,
NV 23–26 May 2000) pp 127–32
[23] Que L, Park J and Gianchandani Y B 2001 Bent-beam
electrothermal actuators—Part I: Single beam and cascaded
devices IEEE/ASME J. Microelectromech. Syst. 10 247–54
[24] Dhuler V R and Walters M D 2001 Microactuators including a
metal layer on distal portions of an arched beam US Patent
no 6,255,757, July 3
[25] Tang W C, Nguyen T-C H and Howe R T 1989 Laterally
driven polysilicon resonant microstructures Sensors
Actuators 20 25–32
[26] Klaassen E H, Petersen K, Noworolski J M, Logan J,
Maluf N I, Brown J, Storment C, McCulley W and
Kovacs T A 1996 Silicon fusion bonding and deep reactive
ion etching: a new technology for microstructures Sensors
Actuators A 52 132–9
[27] Hynes A M, Ashraf H, Bhardwaj J K, Hopkins J, Johnston I
and Shepherd J N 1999 Recent advances in silicon etching
for MEMS using the ASETM process Sensors Actuators
74 13–7
[28] www.kadencephotonics.biz