You are on page 1of 11

Mechanical modeling of fluid-driven polymer lenses

Qingda Yang,1 Paul Kobrin,2,* Charles Seabury,2 Sridhar Narayanaswamy,2


and William Christian2
1
Department of Mechanical and Aerospace Engineering, University of Miami, Coral Gables, Florida 33124, USA
2
Teledyne Scientific & Imaging, 1049 Camino Dos Rios, Thousand Oaks, California 91360, USA
*Corresponding author: pkobrin@teledyne.com

Received 13 March 2008; revised 4 June 2008; accepted 10 June 2008;


posted 11 June 2008 (Doc. ID 93759); published 9 July 2008

A finite-element model (FEM) is employed to study the pressure response of deformable elastic mem-
branes used as tunable optical elements. The model is capable of determining in situ both the modulus
and the prestrain from a measurement of peak deflection versus pressure. Given accurate values for
modulus and prestrain, it is shown that the two parameters of a standard optical shape function (radius
of curvature and conic constant) can be accurately predicted. The effects of prestrain in polydimethylsi-
loxane (PDMS) membranes are investigated in detail. It was found that prestrain reduces the sensitivity
of the membrane shape to the details of the edge clamping. It also reduces the variation of the conic
constant with changes in curvature. Thus the ability to control the prestrain as well as thickness
and modulus is important to developing robust optical designs based on fluid-driven polymer lenses.
2008 Optical Society of America
OCIS codes: 310.6805, 160.5470.

1. Introduction tion and may find applications in fine adjustments


The concept of an optical lens with a variable such as autofocus, but it has a limited range of refrac-
tunable focal length (FL) has been of considerable tive power adjustment. The most straightforward
interest for a number of years [1,2]. The ability to tactic is to directly adjust the curvature of the pri-
construct auto focus, zoom, or self-steering optical mary refractive interface with air [1,2,710]. Ad-
systems that do not require a change in external phy- vances in the optical quality of flexible polymer
sical dimensions would enable small robust designs membrane materials, such as polydimethylsiloxane
for a number of important applications. As new ma- (PDMS), developed for ocular implants and contact
terials and alternative electromechanical systems lenses, have made this appear a viable technology.
have become available, the search for practical de- A number of groups have demonstrated a variation
signs has gained momentum. of FL by changing fluid pressure on a fluid-filled lens
For example, the ability of liquid crystals to change with flexible membranes, which will be referred to in
index in response to an applied field stimulated a this paper as a fluid-filled polymer lens [2,711]. The
number of designs for variable-focus flat lenses [3]. large selection of potential fluids, the high index dif-
However, the polarization requirement and the lim- ference with air, and the wider range of curvatures
ited dynamic range have restricted the use of this achievable give the optical designer considerably
more flexibility. The approach does require a pump-
technique. Electrowetting, the change in surface con-
ing system, but various ingenious mechanical de-
tact angle in response to an electric field, has also
signs have been proposed. [6,8,12].
been exploited to produce tunable optical elements
In most of the systems demonstrated, simple mod-
[46]. Electrowetting avoids the polarization restric-
els assuming spherical surfaces have been used. If
this were an adequate representation for fluid-filled
0003-6935/08/203658-11$15.00/0 polymer lenses, the entire optical design history
2008 Optical Society of America based on spherical surfaces would be available.

3658 APPLIED OPTICS / Vol. 47, No. 20 / 10 July 2008


Unfortunately, even a qualitative consideration of
the mechanics suggests that the spherical assump-
tion is not accurate. Edge bonding conditions and fi-
nite film stiffness result in inflection points that do
not occur on a sphere. While to first order this devia-
tion can be ignored in determining the basic FL,
the picture is not adequate to predict the ultimate
aberration-limited performance. In order to design
precise optical systems a higher-order shape function
is needed. In this work, a parametric mechanical
model is first developed for the deformation, based
on the geometric dimensions and film properties in-
cluding thickness, boundary conditions, modulus,
and prestrain. The material constants are then ex-
Fig. 1. Illustration of a thin-film polymer membrane (of radius a
tracted from a comparison with simple point mea- and thickness h) bonded to a rigid washer.
surements on real devices at various pressures.
Those values are then inserted into a finite-element
model (FEM model) to predict the precise shape over First, the influences of Youngs modulus, E, and
a wide range of operating conditions. This form is prestrain, 0 , on the relationship between peak de-
then fitted with a standard optical lens shape func- flection (or, in optical terminology, sag) of the mem-
tion containing a radius of curvature, R, and a conic brane, 0 , and the applied pressure, p, is examined.
constant, K and is used for predictive optical design. Since there is no analytical solution that can cover
The paper is organized as follows. In Section 2, a the regime of transition from small to large deflec-
parametric study based on a simple approximate ana- tion, an approximate solution is derived by using
lytical formulation for the deflection of a circular an analytical form based on the Galerkin method.
membrane under transverse pressure is discussed This method assumes that the deflection shape is
in order to highlight the significance of boundary of the form 1 r=a2 2 and is proportional to the
conditions and prestrain. In Section 3, specimen fab- peak deflection,
rication and membrane deformation measurement
procedures are described. In Section 4 a FEM-based Wr 0 1 r=a2 2 ; 1a
self-consistent parameter extraction process that can
accurately determine the in situ values of membrane where r is the radius of the point in cylindrical coor-
modulus and prestrain is established, and its effec- dinates. Equation (1a) automatically satisfies the
tiveness is demonstrated. Numerical predictions displacement boundary conditions and following
based on the determined in situ modulus and the standard Galerkin procedure; an approximate
prestrain are compared with experimental measure- analytical expression that relates pressure and peak
ments. Section 5 discusses the implication of pre- deflection is found to be
strain on polymer lens design and some practical
means to quickly assess prestrain from measured   2  2
p 1 h 0
data. Finally, in Section 6 some highlights of this 112 46 28 18 2

study are summarized. E 211 2 a a
  
0 h
840 1 v : 1b
2. Parametric Study with Consideration of Prestrain a a
To qualitatively understand the influence of possible
membrane prestrain on optical parameters (e.g., the It can be easily demonstrated that this solution ap-
conic constants and radius of curvature of a deformed proaches both small and large deflection limits cor-
membrane), a simple analytical expression relat- rectly. Also, note that in the limit of 0 0 and
ing applied pressure and peak deflection of a pres- with small 0 =a and h=a, the solution
sured membrane was derived by using the Galerkin  3
method (e.g., [13]) The necessary elastic and geo- p 16 h 0

metric parameters that have to be considered in a E 31 2 a a
membrane deformation analysis are illustrated in
Fig. 1. The deformation of such a circular membrane is the correct solution for small deformation of the
will approach two limiting cases. At low pressure, thin-plate-bending problem.
which is of interest to long FL applications, the de- Equation (1b) reveals that for a membrane with a
formation is small and can be described by small- given geometry h=a, the normalized peak deflection
deformation plate-bending theory. At sufficiently 0 =a is dependent on the normalized pressure p=E
high pressure the deformation approaches the so- and prestrain and is weakly dependent on Poissons
called membrane limit by biaxial stretching. The op- ratio v. In this study, the membrane material, PDMS,
erating regime of membranes in fluid-filled polymer is nearly incompressible, and there was no evidence
lenses is frequently between these two limits. that it is changed by the various manufacturing

10 July 2008 / Vol. 47, No. 20 / APPLIED OPTICS 3659


procedures, i.e., spin coating under varying tempera-
ture and electron beam irradiation. Further, as will
be seen later in the numerical analyses (Section 4),
the deformation of PDMS membranes can be
accurately characterized by the optimum choice of
Youngs modulus E and prestrain 0 , which indicates
that any uncertainty associated with a processing-
induced change of v will be compensated by E and
0 . Therefore a Poissons ratio of 0.48 was used in
all analyses hereafter.
It is immediately seen from Eq. (1b) that for any
thin-film, polymer-based membrane (h=a 1) the
prestrain will have significant influence in the small
deflection regime (small 0 =a). Note that, in this lim-
it, the normalized peak deflection is proportional to
the normalized pressure, and the proportionality
scales linearly with the level of prestrain,

 
p 40 h 0 0 h
1; 1 : 2a
E 1 a a a a

In the other limit, when 0 =a is sufficiently large,


the peak deflection is independent of the prestrain,

   
p 46 28 182 h 0 3 0
large : 2b
E 211 2 a a a

The effect of 2% prestrain on the normalized de-


flection versus pressure is shown in Fig. 2. Note that Fig. 2. (Color online) (a) Effects of prestrain (0 ) on the relation-
ship between normalized peak deflection (0 =a) and applied pres-
the bottom left corner corresponds to the small de-
sure (p=E) for various membrane geometries (h=a), and
flection limit and the upper right corner corresponds (b) magnified view of 0 =a p=E at typical operational regime.
to the membrane limit. It is obvious that the pre-
strain is of significant influence on the membrane de-
flection in the small-deformation limit. For typical 3. Specimen Fabrication and Deformation
fluid-driven polymer lenses, the operational regime Measurement
is indicated by the dashed box in Fig. 2(a), which To investigate the prestrain effects, specimens with
is magnified in Fig. 2(b). It is seen that the prestrain different levels of prestrain were fabricated, and
has significant influence on the sag versus pressure: their deformation profiles under various fluidic pres-
it affects not only the magnitudes, but also the char-
sures were measured. The membranes were fabri-
acteristics of the 0 =a versus p=E curves, especially
cated under different conditions and subjected to
the initial slopes. It will be shown below that the pre-
strain can also significantly influence the shape of different treatments; so their properties may vary
polymer membranes. significantly. Descriptions of the samples and the
It is interesting to note that in the small- measurement procedures are described below.
deformation limit (large FL), the prestrain can actu- A. Specimen Fabrication Procedure
ally increase the pressure sensitivity of polymer
based membranes, but it has less effect on the pres- The membranes in this study were made of Dow
sure sensitivity at large deformations (small FL). Corning Silgard 185 silicone. The precursor liquid
It is emphasized that Eqs. (1a) and (1b) are far was spun onto standard semiconductor-grade silicon
from a rigorous solution but are a convenient first- wafers. The wafers had previously been coated with a
order estimate of the influence of prestrain on mem- thin release layer. For the lower temperature curing
brane sag. While the 0 =a versus p=E relation given the release layer was photoresist. The silicone was
by Eq. (1b) is within 5% of our rigorous numerical spin coated and cured twice to achieve a total thick-
(FEM) solutions, the assumed shape function of ness of 120 m.
Eq. (1a) is generally erroneous. To accurately predict The support structures consisted of rigid circular
the deformed shape, detailed numerical models washers with a 10 mm inner diameter and a 12 mm
using FEM must be used, and this will be detailed or larger outer diameter. The baseline design used
in Section 4. machined aluminum polished to an optical-quality

3660 APPLIED OPTICS / Vol. 47, No. 20 / 10 July 2008


surface. Because of the concern that the edge round- B. Membrane Deformation Measurement
ing induced by polishing would affect the edge con- Shape versus pressure measurements were made
straints and thus the lens shape, silicon washers using a noncontact optical technique at the facilities
were sometimes used. The silicon washers were cut of Micro Photonics Inc. (Irvine, California, www
to shape with laser etching, which left very sharp cor- .microphotonics.com). The micromeasure profiler
ners. A thin film of SiO2 was sputtered onto the rings. uses a polychromatic beam focused on the surface
The rings were dry pressure bonded to the supported by a highly achromatic lens and a spectrometer to de-
membrane after both surface were exposed to an oxy- tect the wavelength of a reflection maximum, which
gen plasma. Then the membrane was cut around the indicates the vertical height (Fig. 3). The sample is
ring, and the release layer was dissolved away. Very scanned in the XY plane under the probe to obtain
good bonding quality was achieved by this process, a 3D profile. Two different configurations were used,
and no debonding between washer and membrane but in both cases the vertical resolution was less than
was observed in any of the specimens. 100 nm, and the noise in the data is limited by the
Variation of prestrain and possibly modulus re- surface roughness of 500 nm rms. The horizontal
sulted from differences in cure temperature. As de- step was 10 m, roughly the same as the optical spot
tailed in Table 1, the cure temperature was 90C for size. The XY translation stage reproducibility is
baseline samples and 150C for the high-temperature 2 m according to the manufacturer and is small en-
ough to have no noticeable effect on this work. The
samples. Two mechanisms may contribute to pre-
reflection comes primarily from the airPDMS inter-
strain in the final mounted membranes. One is
face. The smaller optical index difference between
that there is some shrinkage as the film cures. The
the second PDMS surface and the pressurized fluid
other is that there is a large mismatch between the produces an insignificant reflection. At the edges of
coefficient of thermal expansion of the PDMS (6 the aperture, however, where the film is bonded to
104 =C) and the silicon substrate (3 106 =C). If the washer, the second PDMS surface reflection is
the film is assumed to be at zero stress when the stronger than the first, and some noise and uncer-
cure is complete at the cure temperature, this coeffi- tainty result. Also, the collection optics cannot mea-
cient of thermal expansion difference will result in sure the membrane when the slope exceeds 30. This
a tensile stress of a few percent when the film is cooled occurs at the outer radius of the membrane at the
to room temperature. This prestrain should in gen- highest pressures. The system is nevertheless very
eral increase as the cure temperature is increased, effective for measuring the range of curvatures
although it is very difficult to control it quantitatively. and apertures useful for optical devices.
Further, different cure temperatures may also lead The apparatus did allow for measurements at both
to variations in the Youngs modulus because of the positive and negative pressures. No evidence of peel-
temperature dependence of polymerization. There- ing at the edges was seen under positive pressure in
fore, a method that is capable of accurately determin- this data. With the film on top of the washer, the entire
ing in situ modulus and prestrain from an already aperture could be scanned (within the slope limit).
mounted membrane was developed and is described
4. FEM Analyses of Prestrained Membrane
in Section 4.
Deformation
In addition to the two sets of membranes prepared
at different cure temperatures, several films pre- As mentioned in Section 2, the approximate solution
pared with the baseline process (90C cure) were of Eqs. (1a) and (1b) should not be used to evaluate
further subjected to a flux of high-energy electron ir- deformation profiles for designing optical compo-
radiation. This process is known to increase the cross nents. In this section we use a rigorous, 3D FEM with
large deformation theory to study the effects of mod-
linking of the polymer and will increase the resulting
ulus and prestrain on the deformed profile.
modulus of the membrane [14]. This radiation
process may also alter the initial prestrain, because
volume and temperature variations were observed
during the irradiation process.

Table 1. Process Parameters for Fabrication of the PDMS


Films Used in this Study

Parameter Baseline High Temp


Spin speed 1000 RPM 1000 RPM
Spin time 60 s 60 s
No. of spins 2 2
Cure temperature 90C 150C
Cure time 4h 1h
Fig. 3. Test configuration for the pressure-dependent shape mea-
Final thickness 120 m 120 m
surements.

10 July 2008 / Vol. 47, No. 20 / APPLIED OPTICS 3661


While the mechanics would suggest an intuitive fully modeled by using contact elements. Such local
parameterization based on the physical constraints, details were found important because they dictate
the need to assess the effect of the shape on optical the bonding edge boundary conditions that have con-
designs requires conformation to the conventions siderable influence on membrane deflection profile,
commonly used in optics. In this study, the standard as shown in Fig. 5(a).
surface used in a popular ray tracing program (ZE- The prestrain was modeled by using a pseudo-
MAX Development Corporation, 3001 112th Avenue temperature method, that is, by inducing a false tem-
NE, Suite 202, Bellevue, Washington 98004-8017, perature change so that the intended level of pre-
USA) was chosen. The normalized two-parameter strain due to thermal contraction was developed
equation that characterizes the deformed membrane before applying the mechanical pressure load. Extra
surface or conic profile is care was taken so that the membrane thickness after
the pseudo-temperature excursion was equal to the
r r=a2 a=R measured thickness of a physical sample.
p ; 3
a 1 1 1 Kr=a2 a=R2 The FEM-predicted effects of 0 on the deformed
membrane profiles are plotted in Fig. 5(a) for a mem-
where R and K are the radius of curvature (which is brane with h=a 0:025. The FEM predicted de-
inversely proportional to the FL) and the conic con- formed shapes at prestrain levels of 0% , 2%, and
stant, respectively. They are related to the lengths
of the semi-major axis (b1 ) and the semi-minor axis
(b2 ) by

R b1 2 =b2 ; K b1 2 b2 2 =b1 2 : 4

In geometrical terms, the conic profile is a hyperbola


when K < 1, a parabola when K 1, an ellipse
when 1 < K < 0, a sphere when K 0, and an ellip-
soid when K > 0. Thus, with only two parameters, a
wide range of optically important shapes can be mod-
eled. By fitting the experimentally measured shape
and the model derived shape to this form the effect
on image quality can be readily evaluated.
A. Finite Element Model (FEM)
The 3D axial-symmetric FEM model is shown in
Fig. 4. The numerical model contains 10 layers of
axial-symmetric elements for the thin-film mem-
brane and one layer of axial-symmetric rigid ele-
ments for the washer. The washer deformation is
ignored because its Youngs modulus is several or-
ders of magnitude larger than that of the PDMS poly-
mer. Local details around the bonding edge including
rounded corners and possible contact (touching) of
the deformed membrane to the washer were care-

Fig. 5. (Color online) (a) FEM predicted normalized deformation


shape (r=a) as a function of normalized position (r=a) at var-
ious deformation levels for 0 0% (red dashed curves with cir-
cles), 0 2% (black single-dotteddashed curves), and 0 4%
(pink solid curves). Also superimposed are spherical shapes at
the corresponding peak deformation levels (blue double-dotted
dashed curves) for comparison. (b) Microsoft Excel fitted conic con-
stant (K) and normalized radius of curvature (R=a) from the defor-
mation data of (a) plotted as functions of applied pressure (p=E).
Fig. 4. (Color online) 3D axial-symmetric FEM of the bonded The solid curves refer to the left Y axis and dashed curves to the
membrane system used for membrane deformation calculations right Y axis

3662 APPLIED OPTICS / Vol. 47, No. 20 / 10 July 2008


4% (see legend) are plotted for four different defor- transition regime as well. In this study, an optimiza-
mation levels (0 =a 0:021; 0:055; 0:10; 0:22). For tion procedure has been used to adjust E and 0 to
comparison, the corresponding spherical shapes hav- match the predicted 0 versus p curve with the mea-
ing the same peak deflection are also superimposed. sured one. The linear elastic nature of the system
It is seen that the prestrain can greatly improve the guarantees that the extracted E and 0 approach
optical profiles in that they can greatly reduce the the true solution. The self-consistency of the para-
boundary effects and make the deformed profiles clo- meter determination procedure can then be vali-
ser to a spherical shape. This is particularly impor- dated by comparing predictions of the membrane
tant at relatively small deformations. Without the deformation profiles against experimental measure-
prestrain, the deformed shapes are significantly dif- ments over the entire range (r=a from 0 to 1).
ferent from spherical shapes and are closer to (an un- Figure 6 gives such an example. In this particular
wanted) hyperbolic shape. However, some boundary sample, the PDMS membrane had been subjected to
effects persist (especially at large deformations) even electron irradiation to enhance its elastic modulus.
for a large prestrain. Therefore, for practical pur- The experimentally measured 0 versus p data are
poses, only the central 80% of the membrane (80% depicted in Fig. 6(a) as filled dots. The FEM calcu-
aperture) should be used in an optical design, and lated 0 versus p curves of three different E; 0 com-
this convention will be followed in the remainder binations are also plotted. The numerical 0 versus p
of this paper. curve is sensitive to both E and 0 . However, it is
The numerically computed deformation data was found that the best fit to the experimental data can-
fitted to the standard conic surface in normalized co- not be achieved by adjusting E or 0 individually.
ordinates using the optimizing solver function in a They must be adjusted simultaneously to achieve
Microsoft Excel spread sheet by setting the target the best fit. The optimum combination was found
function to be in the form of Eq. (3). The resulting to be 0 1:3% and E 12:9 MPa. Additional simu-
normalized radius of curvature, R=a, and conic con- lations were also carried out to examine the possible
stant, K, are plotted in Fig. 5(b) as functions of nor- influence of Poissons ratio. It was concluded that
malized pressure, p=E, with the solid curves such effects were secondary and they can be fully ac-
referring to the left vertical axis and dashed curves counted for by a minor change of E. Therefore, to
to the right vertical axis. The effect of prestrain is keep the optimization process simple, we will use
clearly evident in the variation at low pressure (long E and 0 only and keep Poissons ratio as a constant
FL), but the curves merge together at high pressure of 0.48.
(short FL). That is, 0 can dramatically reduce the The determined E and 0 are then used to predict
membrane deformation (=a) at small p=E, and hence the deformed profiles of the same membrane under a
greatly increase R=a. However, the influence di- wide range of deformation levels, as shown by the so-
minishes as pressure increases. Both of these limits lid curves in Figs. 6(b) and 6(c). It can be seen that
are consistent with the results of the approximate the numerically predicted shapes follow the experi-
analytical solution (Fig. 2). mental curves closely, at both large [Fig. 6(b)] and
The existence of prestrain also has important in- small deformations [Fig. 6(c)]. The excellent agree-
fluences on K (solid curves in Fig. 5(b), referring to ment between numerically predicted and experimen-
the left vertical axis). With 0 0, the deformed tally measured deformation profiles clearly
membrane is highly hyperbolic (negative K values) demonstrates that the determined membrane modu-
at small pressure (i.e., small deflection) and quickly lus and prestrain are true properties of the mem-
becomes ellipsoidal as pressure increases. When brane and suggests that they can be used for
pressure is sufficiently large (i.e., large deformation), design purposes.
K remains slightly ellipsoidal (K > 0), but asympto-
tically approaches 0 (spherical shape) at high pres- C. FEM Analyses of Experiments
sure. As the prestrain increases, the change in K For each of the three types of membrane listed in
becomes more gradual, as evidenced by the 0 Table 2, the deformed profiles at different pressure
2% and 0 4% curves. More interestingly, when levels were carefully measured. The peak deflection
0 is sufficiently large (for example 0 4%) the conic versus pressure data was then used to obtain the
constant becomes almost independent of the radius Youngs modulus and prestrain in the membrane
of curvature within a fairly large range of pressures through FEM and the parameter determining proce-
(or equivalently, deformation). This is desired by op- dure described in Subsection 4.B. The obtained E and
tical designers because a constant K could signifi- 0 for each membrane are listed in Table 2, together
cantly simplify the designing procedure. with the membrane geometries.
Although there is some scatter in the data, the
B. Determining in situ E and 0 mean values for each type of sample do show the ex-
Figure 5 suggests that the prestrain significantly pected differences. The films cured at 90C have
influences membrane shape, and hence the optical 1:9% prestrain, whereas the films cured at 150C
performance, particularly at low pressure. Unfortu- have 2:5%. There is also a small difference in the
nately, this is a difficult region to accurately measure modulus, which, though not surprising, is not ex-
the shape, and therefore one has to make use of the plained definitively. The irradiated films did show

10 July 2008 / Vol. 47, No. 20 / APPLIED OPTICS 3663


bonded. It is clear from these variations that the abil-
ity to extract the material parameters from
fabricated membranes is essential to being able to
eventually engineer optimized devices.
The repeatability of experiments is reasonably
good. It can be seen from Table 2 that FEM deter-
mined E and 0 are fairly consistent among speci-
mens, as long as the local details of the washers are
reasonably reproduced. Therefore in the following
we choose to use only one typical datum from each spe-
cimen group for detailed analysis and discussion.
Three typical 0 versus p curves are shown in
Fig. 7(a), where the symbols are measured data
points and the solid lines are the numerical best fits
using the E and 0 shown in the corresponding boxes.
Figure 7(b) is the same data but is plotted in normal-
ized scales (0 =a versus p=E). Note that the apparent
convergence of the non-normalized data at low pres-
sures is the result of the compensating effects of
the higher modulus and the lower prestrains in

Fig. 6. (Color online) (a) Best fit of FEM result of peak deflection
(0 ) as a function of applied pressure (p) (red solid curve) as com-
pared with the experimentally measured data (squares). Two
other FEM curves of 0 p using different moduli and prestrains
are also plotted to show the influence of these two parameters on
the 0 p data. (b) FEM predicted deformation profiles (black solid
curves) under various high pressure levels, compared with the ex-
perimentally measured deformation (curves with high-frequency
undulations). (c) Comparison of FEM predicted deformations
(black solid curves) with experimental measurements (curves with
undulations) at lower pressure levels.

Fig. 7. (Color online) (a) FEM calculated 0 p using best-fit E


the expected increase in modulus. However they also and 0 as indicated for three membranes (solid curves) compared
show a reduction in prestrain, which was not ex- with experimentally measured data (symbols). (b) Normalized
pected, but may be explained as a virtual annealing 0 =a p=E curves showing the different initial slopes that are inver-
effect as the polymer chains are broken and re- sely proportional to the level of prestrains in the three membranes.

3664 APPLIED OPTICS / Vol. 47, No. 20 / 10 July 2008


Table 2. Membrane Geometries and Calculated Modulus and Prestrain

Specimen E (MPa) 0 (%) a (mm) h (mm) v


PDMS membrane cured at 150C
PDMS_HT1 2.30 (6.45) 2.4 (0.88) 5.015 0.121 0.48
PDMS_HT2 2.25 (6.54) 2.5 (0.86) 5.015 0.121 0.48
PDMS_HT3 2.60 (7.89) 2.7 (0.90) 5.015 0.121 0.48
PDMS membrane subjected to electron-beam irradiation after curing at 90C
PDMS_EB1 10.3 (17.6) 0.65 (0.45) 5:00 0:250 a 0.123 0.48
PDMS_EB2 11.7 (16.7) 0.35 (0.34) 5.015 0.108 0.48
PDMS_EB3 11.5 (16.4) 0.4 (0.34) 5.015 0.109 0.48
PDMS membrane cured at 90C (baseline)
PDMS_1 2.75 (6.06) 2.3 (1.12) 5:000 0:300 a 0.124 0.48
PDMS_2 2.85 (6.40) 1.65 (0.81) 5.015 0.124 0.48
PDMS_3 2.90 (5.99) 1.1 (0.57) 5:015 0:235 a 0.124 0.48
a
The numbers after indicate the local radius of curvature of the rounded hole edge of the rigid washer, which has to be explicitly
modeled as illustrated in Fig. 4. The rest of the specimens have sharp edge.

the PDMS_EB samples. In the normalized plot the of p=E, before they become unstable. Note that, even
differences in initial slope reflect the differences in at large p=E values, the numerical predictions are
0 as described by Eq. (2a). A quick calculation using not completely identical to experimental results.
the normalized initial slopes gives prestrains of 0.4%
for PDMS_EB2, 1.7% for PDMS_2, and 2.3% for
PDMS_HT1, which are in excellent agreement with
the numerically obtained values.
Using the E and 0 values derived from the 0 ver-
sus p data, FEM simulations of the membranes at
different pressure levels were performed. These were
then fitted with the standard conic surface, and
values of radii of curvature and conic constants were
obtained and compared with the experimental
values. Figure 8(a) shows the comparisons of normal-
ized radius, R=a, as a function of normalized pres-
sure p=E. The predicted R=a versus p=E relation is
remarkably successful. The numerical results (solid
curves) are almost identical to the experimental
curves (dotted curve with symbols). Note the rela-
tively large difference between samples PDMS_2
and PDMS_HT1, despite the fact that their 0 versus
p curves are almost identical (Fig. 7(a)).
The predicted K versus p=E curves are shown in
Fig. 8(b). The predictions follow experimental curves
over a large range of p=E, but the agreement in the
small p=E range is plagued by the large oscillation of
experimental data. There are two reasons for the
lack of consistency of experimental fitting. The first
is noise in the data. At small pressures and deforma-
tions, the fractional errors in pressure and height in-
crease (p error 0:14 KPa), as can be seen from
Fig. 6(c). Also, since K represents the deviation from
a spherical shape, it is a second-order effect, and the
fitting becomes increasingly unstable for a flat
surface.
From Fig. 8(b), the effect of 0 on the K versus R=a
relation is clearly seen. Of the three samples,
PDMS_HT1 has the largest prestrain (0 2:4%),
and K remains nearly constant with a value of about
Fig. 8. (Color online) (a) Comparison of FEM predicted (solid
2.5 over a fairly large range of p=E. This is a desirable curves) and experimentally measured (symbols) normalized ra-
condition, since, in developing a multielement vari- dius of curvature (R=a) as functions of normalized pressure
able design, a changing K value can be problematic (p=E) for the three membranes with different mechanical proper-
for optimization. As the level of prestrain decreases, ties. (b) Conic constant K predicted (solid) and extracted from ex-
the K values, in general, increase with the decrease perimental measurements (symbols).

10 July 2008 / Vol. 47, No. 20 / APPLIED OPTICS 3665


This is because the conic constant is more sensitive to ulus of the PDMS_EB2 membrane (EEB2 11:7 MPa)
the boundary conditions than is the radius of curva- is about 5.1 times the modulus of the PDMS_HT1
ture. Despite considerable effort, the details of the membrane (EHT1 2:3 MPa). If both lenses are de-
bonding contain subtleties that can not be completely formed to have identical radii (for example, R=a
captured in the model (see Fig. 4). Fortunately, for 20), the required pressure for the PDMS_HT1 mem-
optical design, the effect of K is greatest for small brane is p 0:00012EHT1 0:28 KPa, which is actu-
FLs (small R), whereas the predictions are most ac- ally larger than the pressure required for the
curate and less important at long FLs (large R near PDMS_EB2 membrane (p 0:00002EEB2 0:23
flat conditions). KPa), owing to the much larger 0 in the PDMS_HT1
membrane. However, at the large deformation of
R=a 3, the required pressure for PDMS_HT1 mem-
5. Discussion
brane is p 0:0116EHT1 26:7 KPa, which is less
It has been clearly demonstrated that the variations than half of that required by the PDMS_EB2 mem-
in process parameters can influence the material brane, p 0:005EEB2 58:5 KPa. This explains the
modulus and prestrain in spin-cast films. Our nu- apparent contradiction regarding the p curves
merical procedure to obtain the in situ prestrain shown in Fig. 7(a). Similarly, increasing polymer
and Youngs modulus of a polymer lens has been pro- membrane thickness (increasing h=a in Fig. 2) can
ven to be very effective. More important, numerical also enhance the pressure sensitivity at small de-
predictions of membrane optical performance para- formations rather dramatically because the pressure
meters based on the determined membrane proper- is proportional to the third power of membrane thick-
ties are in excellent agreement with experimental ness in this limit in the absence of 0. However,
measurements. Especially, the numerical predictions this method suffers the same drawback as the high-
on the radius of curvature of a deformed membrane modulus membrane at large deformation; i.e., the
at various pressure levels are of very high fidelity pressure required is proportional to the thickness.
(Fig. 8). Using such a high-fidelity numerical model,
a number of issues important for polymer-based lens B. Quick Estimate of Prestrain in Thin-Film
design can be investigated. Polymer Membranes

A. Effects of Prestrain
If the Youngs modulus is known, it is possible to
quickly extract the prestrain level in a polymer mem-
The existence of prestrain is actually beneficial to brane of a given geometry (i.e., known h=a) from the
the optical performance of thin-film polymer mem- measured 0 p data [e.g., Fig. 7(a)]. This can be
branes because it helps to reduce the negative in- done through measuring the initial slope in the
fluence of boundary conditions on the deformed small-deformation range. If the membrane thickness
membrane profile (Fig. 5(a)). The deformed lens is very small (h=a 1), the initial slope is directly
profile with prestrain is closer to a spherical shape, related to the prestrain by Eq. (2a). With E known,
which is typically assumed in optical lens design. the prestrain can be quickly assessed by
Further, the prestrain helps to maintain a near-
constant value of K within a large pressure range
(Figs. 5(b) and 8(b)), which will greatly facilitate 1 a a
0 ; 5
optical design of polymer membranes. 4 h kE
Another important benefit of prestrain is that it will
lead to much improved pressure sensitivity (p=0 ) of where k 0 =p is the initial slope of the 0 p curve.
polymer membranes in the small-deformation re- Unfortunately, for most thin polymer membranes the
gime. Pressure sensitivity is important for the accu- elastic modulus itself is, strictly speaking, unknown.
rate control for variable-field-of-view applications. Nevertheless, the bulk modulus of the polymer mem-
According to Eqs. (2a) and (2b) and as clearly shown brane may be used for a quick, back-of-the-envelope
in Fig 7(b), at small deformation p=E is proportional to estimate of the prestrain in the membrane.
0 . However, it is noted that the prestrain effect fades Also, note that at large deformations (0 =a > 1) the
away gradually as membrane deformation increases 0 p relation is independent of the prestrain
(0 =a increases), which means there is no extra bur- [Eq. (2b)]. This feature can be used, in theory, to cal-
den for pressure increase in the small-field-of-view culate the Youngs modulus (E) of a polymer mem-
operating regime. This is preferable to other pressure brane. However, few polymer membranes can
sensitivity enhancement methods, such as increasing survive such a large deformation without bursting.
membrane modulus or thickness. At the deformation level involved in this study,
Increasing the modulus of a polymer membrane 0 =a 0:2, it is readily seen that the membrane mod-
will improve the pressure sensitivity in the small- uli obtained (as listed in the parentheses in Table 2)
deformation regime. However, the increased modulus are typically overestimated by a factor of about 23
requires a proportionally increased pressure to (assuming a fixed Poissons ratio of 0.48).
achieve the same deformation for large deformations,
which could be an extra burden for a miniature lens C. Optical Design with Considerations of Prestrain
system. This can be illustrated using the PDMS_HT1 In almost all of the previously reported fluid-driven
and PDMS_EB2 curves shown in Fig. 7(a). The mod- polymer lens designs, it was implicitly assumed that

3666 APPLIED OPTICS / Vol. 47, No. 20 / 10 July 2008


the membranes were deformed into a spherical 6. Conclusions
shape, i.e., K 0 [1,2,47,9,15,16]. However, both In this paper, a numerical model is developed for the
analyses and experimental measurements in this pressure response of a deformable elastic membrane
study showed that K is nonzero: K is mostly around used in a tunable optical element. The model is cap-
2.5 for a wide range of FLs, with rather significant able of determining both the modulus and the pre-
deviations (from K 0) in the small-deformation re- strain from a measurement of peak deflection
gime (small p=E), depending on the prestrain level versus pressure. A simple analytical model can be
(Fig. 8). It is noted that the conic constant does have used to accurately determine the prestrain if the
a profound effect on such issues as image quality. modulus is known. Given accurate values for E
Figure 9 shows a calculation of the spot size for a sim- and 0 , we have shown that the two parameters of
ple biconvex lens with two different FLs but varying the standard optical shape function (curvature and
conic constant. For negative conic values the focus is conic constant) can be predicted. The accuracy of
considerably improved over a simple spherical the prediction of the conic constant is best at large
shape, and for positive values it is degraded. Clearly values of curvature and poorest near flat conditions.
ignoring the conic term would result in a disappoint- Fortunately, the sensitivity of optical designs to the
ing correlation between the optical model and real value of the conic constant is greatest at curvature.
performance. For the long FL case R 25 mm (R=a In general it was found that prestrain reduces the
6:25), the effect is much less than for the short FL sensitivity of the membrane shape to the details of
case, R 8 mm (R=a 2:0). the edge clamping (which are often difficult to model
The clear implication from Fig. 9 is that to achieve accurately). It also reduces the variation of the conic
better image quality from a single element, a nega- constant with changes in curvature.
tive K value is preferred. Unfortunately, both the Thus the ability to control the residual stress as
model analyses and experiments in this study show well as the thickness and modulus is important to
that this can rarely be achieved with uniform mem- developing robust optical designs based on fluid-
brane geometry and modulus, except at very small filled polymer lenses.
deformation levels (long FL, large R=a or small The authors acknowledge Michael Rubner and his
p=E). However, increasing prestrain can bring the de- students for the irradiation of the PDMS mem-
formation profile closer to K 0. The analysis model branes. This work was supported by the Defense
in this paper shows that optimum lens shapes with Advanced Research Projects Agency (DARPA) under
K < 0 may be achieved by systematic change of the contract HR0011-04-C-0042.
modulus or membrane thickness as a function of r.
Such attempts are currently ongoing.
References
1. N. Sugiura and S. Morita, Variable-focus liquid-filled optical
lens, Appl. Opt. 32, 41814186 (1993).
2. M. Agarwal, R. A. Gunasekaran, P. Coane, and K. Varahram-
yan, Polymer-based variable focal length microlens system,
J. Micromech. Microeng. 14, 16651673 (2004).
3. G. Commander, S. E. Day, and D. R. Selviah, Variable focal
length microlenses, Opt. Commun. 177, 157170 (2000).
4. B. Berge and J. Perseux, Variable focal length controlled by
an external voltage: an application of electrowetting, Eur.
Phys. J. E 3, 159163 (2000).
5. T. S. Krupenkin, T. S. Yang, and P. Mach, Tunable liquid
micro-lens, Appl. Phys. Lett. 82, 316318 (2003).
6. A. Werber and H. Zappe, Tunable microfluidic microlenses,
Appl. Opt. 44, 32383244 (2005).
7. D. Y. Zhang, V. Lien, Y. Berdichevstky, J. Choi, and Y. H. Lo,
Fluidic adaptive lens with high focal length tunability, Appl.
Phys. Lett. 82, 31713172 (2003).
8. D. Y. Zhang, N. Justis, V. Lien, Y. Berdichevstky, and Y. H. Lo,
High-performance fluidic adaptive lenses, Appl. Opt. 43,
783787 (2004).
9. R. A. Gunasekaran, M. Agarwal, A. Singh, P. Dubasi, P. Coane,
and K. Varahramyan, Design and fabrication of fluid con-
trolled dynamic optical lens system, Opt. Lasers Eng. 43,
686703 (2005).
Fig. 9. (Color online) ZEMAX calculation for the focal spot size of 10. P. S. Moran, S. Dharmatilleke, A. H. Khaw, K. W. Tan,
a single double convex fluidic lens with diameter 8 mm, aperture M. L. Chan, and I. Rodriguez, Fluidic lenses with variable
6 mm, material n 1:67, and 550 nm. Triangles are for a lens focal length, Appl. Phys. Lett. 88, 041120 (2006).
with radius of curvature R 25 mm, and the squares are for 11. D. Y. Zhang, N. Justis, and Y. H. Lo, Fluidic adaptive lens of
R 8 mm. Above the plot are schematic representations of lenses transformable lens type, Appl. Phys. Lett. 84, 41944196
with negative (left) and positive (right) conic constants. (2004).

10 July 2008 / Vol. 47, No. 20 / APPLIED OPTICS 3667


12. H. Oku, K. Hashimoto, and M. Ishikawa, Variable-focus lens 15. J. Chen, W. Wang, J. Fang, and K. Varahramyan, Variable-
with 1 kHz bandwidth, Opt. Express 12, 21382149 (2004). focusing microlens with microfluidic chip, J. Micromech. Mi-
13. M. Sheplak and J. Dugundji, Large deflections of clamped cir- croeng. 14, 675680 (2004).
cular plates under initial tension and transitions to mem- 16. H. Ren, D. Fox, P. A. Anderson, B. Wu, and S. T. Wu, Tunale-
brane behavior, ASME J. Appl. Mech. 65, 107115 (1998). focus liquid lens controlled using a servo motor, Opt. Express
14. J. Walish, personal communication (MIT, 2007). 14, 80318036 (2006).

3668 APPLIED OPTICS / Vol. 47, No. 20 / 10 July 2008

You might also like