PAPER www.rsc.

org/loc | Lab on a Chip

Polymer waveguide backplanes for optical sensor interfaces in

microfluidics{{
Kevin S. Lee,* Harry L. T. Lee and Rajeev J. Ram
Received 28th June 2007, Accepted 14th August 2007
First published as an Advance Article on the web 21st August 2007
DOI: 10.1039/b709885p

A polymer optical backplane capable of generic luminescence detection within microfluidic chips
is demonstrated using large core polymer waveguides and vertical couplers. The waveguides are
fabricated through a new process combining mechanical machining and vapor polishing with
elastomer microtransfer molding. A backplane approach enables general optical integration with
planar array microfluidics since optical backplanes can be independently designed but still
integrated with planar fluidic circuits. Fabricated large core waveguides exhibit a loss of 0.1 dB
cm21 at 626 nm, a measured numerical aperture of 0.50, and a collection efficiency of 2.86% in an
n = 1.459 medium, comparable to a 0.50 NA microscope objective. In addition to vertical couplers
for out-of-plane collection and excitation, polymer waveguides are doped with organic dyes to
provide wavelength selective filtering within waveguides, further improving optical device
integration. With large core low loss waveguides, luminescence collection is improved and
measurements can be performed with simple LEDs and photodetectors. Fluorescein detection via
fluorescence intensity with a limit of detection (3s) of 200 nM in a 1 mL volume is demonstrated.
Phosphorescence lifetime based oxygen detection in water in an oxygen controllable microbial cell
culture chip with a limit of detection (3s) of 0.08% or 35 ppb is also demonstrated utilizing the
waveguide backplane. Single waveguide luminescence collection performance is equivalent to a
back collection geometry fiber bundle consisting of nine 500 mm diameter collection fibers.

1. Introduction One common approach utilizing fiber optics places fibers

Luminescence based sensors are used for determining environ-
mental parameters such as dissolved oxygen or pH in
biological systems as well as for providing markers necessary
for cell identification and sorting. Such optical sensing
techniques can provide real time monitoring capabilities and
dynamic control without disturbing a biological system’s
equilibrium or introducing contaminants into the system. As
these new lab-on-a-chip systems become more integrated, the
number of sensors required increases, leading to an increase in
size and complexity of the off-chip detection systems. While
CCD camera imaging is well suited to applications with slow
timescale (,100 Hz) intensity measurements, spatially dis-
tributed systems requiring measurements at both high speed
and low intensity, such as fluorescence lifetime or flow
cytometry, need the speed and sensitivity of photodiodes and
non-imaging optics.1,2 Fiber optics, or more generally,
waveguides are typically used to decouple the location of
measurement points and photodetectors, increasing design
flexibility. In addition waveguides can reduce the number of
photodetectors in these systems by routing multiple signals to
a common photodetector.
MIT, EECS, 32 Vassar St. 26-459, Cambridge, MA, 02139, USA.
E-mail: kevbolee@mit.edu
{ The HTML version of this article has been enhanced with colour
images.
{ Electronic supplementary information (ESI) available: Maximum
waveguide width calculation, loss and roughness measurements and
system measurements. See DOI: 10.1039/b709885p
close to fluorescent sources to provide excitation light as well
as route collected signals to detectors with minimal loss.
However, in general, luminescence collected by a single fiber is
very low. In one instance, a 550 mm fiber performing excitation
and detection was only 0.0009% efficient in detecting
fluorescence from an oxygen sensitive ruthenium complex2
due to direct dye deposition on the small fiber. Fiber bundles
can overcome the low collection efficiency of single fibers when
collecting light from a distributed source, but typically require
many fibers, which can be expensive to assemble and not
useful when high spatial resolution is required. For example, a
fiber bundle system designed to collect fluorescence from
stained tissue samples consisted of thirty 100 mm fibers and still
required a photomultiplier tube to perform detection.3 In
addition, large fiber bundles are impractical for devices
requiring multiple sensors.
We have developed a platform of waveguide based passive
optical components utilizing the same polymer materials used
to fabricate microfluidic systems. Integration of excitation and
collection optics into microfluidic systems improves alignment
tolerances with external optical elements and also increases
system design flexibility. Many different designs have been
proposed for absorption or luminescence detection,4 including
many waveguide based geometries such as waveguide arrays,1
right angle collection,5 in-line absorption,6 evanescent cou-
pling,7 and hollow prisms.8 While each of the waveguide based
systems demonstrated optofluidic integration, all of the
designs either required waveguides to be in-plane or in direct
contact with fluid channels, or required fluidic channels to be

This journal is ß The Royal Society of Chemistry 2007 Lab Chip, 2007, 7, 1539–1545 | 1539
redesigned to accommodate waveguides. Such designs make
the layout of integrated optical systems dependent on fluidic
component locations and hinder device scalability. Previous
designs also did not consider the filters required to implement
on-chip fluorescence detection and often required lasers and
expensive detectors to detect a signal. In contrast to the strictly
planar designs demonstrated previously, waveguide back-
planes with vertical couplers allow for the integration of optics
into fluidic systems with minimal impact on the fluidic design.
Using a polymer optical backplane, we will demonstrate a
potentially disposable high sensitivity lifetime based lumines-
cence detection system utilizing only light emitting diodes
(LED) and silicon photodetectors.
2. Design
In order to maximize the signal to noise ratio of the detection
system, stray excitation light must be minimized at the detector
and luminescence collection must be maximized. A back-
scattering collection geometry satisfies the first requirement
because only interface reflections contribute to stray light. For a
planar waveguide system, a backscattering collection geometry
can be achieved by forcing the light out of plane utilizing
waveguide vertical couplers. Due to the low index contrast of
the waveguides, the vertical couplers are metallized to maximize
efficiency. The backplane design is shown in Fig. 1a.
The second requirement, maximizing luminescence collec-
tion, depends on the location and behavior of the fluorescent
source. To demonstrate the backplane concept, we assume
little knowledge of the location of the emitting source and
Fig. 1 (a) A side and top view of the integrated oxygen sensor chip is
shown. The waveguides are designed to excite and collect luminescence
from the sensor located 1 mm above the silvered bends. (b,c) Ray
tracing simulations of the intensity distribution for the input
waveguide excitation and point source collection efficiency of the
output waveguide at the sensing plane for 45u bends and 30u bends,
respectively. System performance increases by 200% after replacing 45u
bends with 30u bends.
1540 | Lab Chip, 2007, 7, 1539–1545
optimize for collecting luminescence located 1 mm above the
backplane, or the thickness of a typical glass slide. Under this
constraint, the minimum core size of the integrated wave-
guides, which maximizes efficiency, is determined by the
distance h from the emitting source and the waveguide
numerical aperture NA = (ncore2 2 nclad2)1/2 which specifies
the sine of the maximum input angle to the waveguide. With
ncore = 1.56 and nclad = 1.459 for the materials used (section 3),
the NA is 0.55. Through ray geometry, eqn (1) for the
minimum width can be derived taking into account coupling
through next = nclad instead of air. This equation is derived in
the ESI.{ At h = 1 mm, the minimum core width is
approximately 800 mm.
2h
w~ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (1)
ðnext =NAÞ2 {1
Sensor excitation and collection efficiency is also influenced
by the waveguide vertical coupler bend angle. While optical
backplanes developed for other interconnect applications9
focus on coupling as much power out of the waveguide as
possible, a backplane for luminescence detection focuses on
maximizing both sensor excitation and collection of sensor
emission. To meet this condition, the sensor should be spatially
illuminated where the collection waveguide is most efficient.
This occurs when the excitation and collection waveguide bend
angles result in intensity profiles which overlap at the sensor
plane.
To determine the spatial distribution of excitation light
incident on the sensor, ray tracing simulations were performed
starting with a point source located at the waveguide input
and ending at the sensor surface. For the collection efficiency
of the output waveguide, the efficiency needs to be determined
for each position on the sensor plane. Rays representing
point source emissions were placed along the sensor plane, as
shown in Fig. 1a, and propagated through the collection
waveguide. The ratio of light exiting the output/filter
waveguide to light emitted by the point source determines
the collection efficiency at a particular point on the sensing
plane. By performing this simulation for points throughout
the surface plotted in Fig. 1b and 1c, a contour of the
collection area efficiency versus sensor position can be
determined. Fig. 1b and Fig. 1c show contours of the
excitation profiles (dashed lines) and point source collection
efficiencies (solid lines) at the sensing plane for 45u and 30u
vertical couplers.
For oxygen sensors used in this paper, fluorophor emission
is collected through a polystyrene encapsulating layer, 100 mm
glass, and then 1 mm of polydimethylsiloxane (PDMS). For a
well centered point source above this specific material stack, a
multimode fiber or microscope objective with a 0.55 NA
achieves a collection efficiency of 3.13% into the first lens
calculated from ray optics10 ignoring reflection effects. In ray
tracing simulations, the 30u coupler, 1 mm2 collection
waveguide design presented here achieves a maximum point
source collection of 2.86%, close to the multimode fiber
maximum, suggesting that this design would be suitable for
cytometry systems using similar NA single lens collection
geometries, which achieve collection efficiencies of 4%.10
This journal is ß The Royal Society of Chemistry 2007
 While the point source collection efficiency is large, the
overall system performance also includes the excitation
efficiency. To include excitation efficiency, the system effi-
ciency is recalculated as the ratio of emission leaving the
output waveguide to excitation entering the input waveguide.
Efficiency at the sensor plane is determined by taking the
overlap of the excitation and collection areas given in Fig. 1c.
The total system efficiency resulting from a maximum angular
distribution excitation is only 0.40% with 45u couplers and a
factor of 2.8 times better at 1.14% for optimized 30u couplers.
To further improve the collection efficiency of the waveguides,
the collection waveguide can be expanded in width to
maximize collection given the large area sensor illumination.
A width expansion to 3 mm brings the system efficiency to
1.73%. While the collection efficiency of this waveguide system
is approximately half of the maximum achievable with a
microscope (3.13%), it provides a compact design, incorporat-
ing both excitation and collection.
3. Fabrication
As mentioned in section 2, the 1 mm standoff between the
sensors and waveguides requires that waveguides be fabricated
on a mm scale to perform efficient luminescence collection.
Standard lithography processes are not amenable to fabricat-
ing master molds with features of this size since mm
thicknesses require multiple spin coating and exposure steps.
In SU-8 processes, layers of 100 mm thickness already suffer
from substrate delamination due to stress.11 As a result, a
computer numerical control (CNC) machining approach is
used to fabricate the mm thick master molds necessary for
polymer molding.
The general waveguide fabrication process is shown in Fig. 2.
Polycarbonate positive master molds were machined using a
CNC mill (Sherline 2000) and a 1 mm diameter square end mill
(Microcut USA) to fabricate 800 mm 6 1 mm channels or ribs
(Fig. 2a). To make the waveguide mold as smooth as possible,
milling was performed around the waveguide structure,
defining the sidewalls only, with a first pass at 2000 rpm and
a 25 mm min21 feed speed, and a second pass removing an
extra 100 mm at 2000 rpm and a 50 mm min21 feed speed to
Fig. 2 The fabrication process for making large core polymer channel
waveguides with a CNC milling process. The core preparation process
for providing integrated filter materials is not shown.
This journal is ß The Royal Society of Chemistry 2007
improve finish quality. In this way, we took advantage of the
pre-polished polycarbonate surface as one of the waveguide
walls. 30u vertical couplers are then milled with 120u full angle
drill mills (Microcut USA) at each detection location with the
same machining parameters.
Roughness due to milling was then reduced through solvent
vapor polishing where a vapor of methylene chloride diffuses
into the plastic surface and causes reflow12 (Fig. 2b). The
degree of polishing could be controlled by varying the solvent
pressure or also by varying the exposure time. Under optimal
polishing conditions, the roughness average (Ra) from sanded
polycarbonate samples could be reduced from 1000 nm to
70 nm.13 By combining vapor polishing with CNC milling, this
method becomes viable for both optical and microfluidic
master mold fabrication.
PDMS (Dow Corning Sylgard 184) was used to replicate the
polycarbonate master molds, forming the waveguide cladding
(Fig. 2c). Since the refractive index difference between PDMS
(n = 1.459 at 600 nm via ellipsometry) and the core material,
polyurethane (NOA 71 Norland Products) (n = 1.56 given by
manufacturer), was too low to reflect light out of plane using
total internal reflection, the vertical couplers were coated with
300 nm of silver to improve reflectivity. By first applying air
plasma to modify PDMS into a glass-like layer (Fig. 2d), high
quality silver films could be deposited with e-beam evapora-
tion (Fig. 2e) on the molded PDMS cladding.14 A 100 mm sheet
of PDMS, made by spin coating onto a fluorinated silicon
wafer, was then air plasma bonded to the top cladding mold to
seal the channels and provide a flat surface for interfacing with
fluidic chips (Fig. 2f).
Once the hollow channels in PDMS were complete, the
waveguide core material was introduced. Before waveguide
core injection, all of the PDMS channels were surface treated
with air plasma to increase their surface energy. This reduces
the contact angle between liquids and cured PDMS, increasing
capillary forces as well as binding strength at the core cladding
interface.15 Without first applying a surface treatment, bubbles
formed along the sidewalls and delamination of the core from
the cladding occurred due to shrinkage after curing. In this
work, no bubbles were observed, and while 6% shrinkage
occurred, there was no sign of delamination. For all
waveguides, NOA71 was filled into the channels via direct
injection and capillary filling (Fig. 2g).
For collection waveguides with integrated filters, NOA71
was first directly mixed with an organic dye (pinacyanol
iodide) which absorbs at 600 nm in order to provide integrated
filtering of the excitation light. The dye was allowed to saturate
the prepolymer. Integrating filters into the waveguides can
improve both coupling and collection efficiency by removing
1 to 5 mm thick external filters and allowing LEDs and
detectors to be placed closer to the waveguide facets. This can
eliminate reflection losses from external filters and reduce free
space divergence losses associated with an increased path
length. Similar dye doped polymer filters have also been
produced for bulk PDMS layers.16 After the channels were
filled with core material, the waveguide array was cured under
UV illumination. While diffusive mixing between dye doped
and clear NOA71 occurred before complete cure was reached,
the dye was still able to fully absorb excitation light. No
Lab Chip, 2007, 7, 1539–1545 | 1541
photobleaching of the dye absorption was observed after a 24 h
UV-exposure.
After core curing, the waveguide ends were cleaved to the
appropriate size (Fig. 2h). Since cleaving elastomeric wave-
guides results in poor quality end faces, additional NOA71 was
deposited on each end and cured to repair the cut ends (Fig. 2i).
This smoothed each of the end faces and improved input
coupling dramatically.
The fabricated device shown in Fig. 3 consists of a
waveguide backplane interfaced with a fluidic device capable
of controlling dissolved oxygen concentration. Design and
fabrication of this fluidic system has been previously
described.17 The fluidic device consists of 4 wells in parallel
with integrated peristaltic mixers separated by a thin gas
permeable membrane. Pressurized gas actuation of the
membrane translates into diffusion of the gas into the fluid
in the well. By placing an oxygen sensor in each well, the
oxygen concentration of the fluid can be measured and
controlled with changes in gas concentration. This microfluidic
system is ideal for testing lifetime based sensors since the
dissolved oxygen in the wells of the chip can be controlled. The
backplane contains one waveguide sensor for each well
coupled to the oxygen sensor inside the well.
The oxygen sensors used for this experiment are composed
of a mixture of platinum(II) octaethylporphine ketone (PtOEP-
K Frontier Scientific) and polystyrene deposited on glass disks.
Glass disks were etched with HF prior to deposition of the
polymer/dye solution to enhance binding of the fluorophor
film to the sensor substrate. This phosphorescent dye has a
maximum absorption at 592 nm and emits at 759 nm with a
12% quantum yield.18
4. Results and discussion
Before using the waveguide backplane for luminescence
measurements, the fabricated system must be characterized.
We first measure straight waveguide test structures to
determine the loss, numerical aperture, and surface roughness
effects of the fabricated waveguides. We then characterize the
waveguide filters and the silvered vertical couplers.
Total system performance is then characterized, including
the input and output coupling efficiency between an LED and
Fig. 3 The fabricated device is shown consisting of 4 waveguide
sensors butt coupled to photodetectors. The optical backplane rests
underneath a miniature bioreactor array containing 4 integrated
oxygen sensors. Darker sections of the waveguides result from mixing
the core polymer with organic dye and acts as a filter for the 590 nm
excitation source.
1542 | Lab Chip, 2007, 7, 1539–1545
the backplane. After demonstrating that luminescence collec-
tion can be performed above the noise limit of the detection
system, the backplane is integrated with a fluidic chip capable
of performing dissolved oxygen control in water and
measurements of oxygen concentration are performed.
A. Waveguide characterization
Fabricated large core waveguides (800 mm 6 1 mm) exhibited
a propagation loss of 0.10 dB cm21 at 632 nm from a HeNe
laser via the cut-back method. The measured waveguide
numerical aperture (NA) was 0.50 (50% intensity). The NA
reduction from the expected 0.55 to 0.50 in the large core
waveguides reduces the collection efficiency and most likely
results from sidewall roughness still present after the polishing
process. By assuming uniform scattering within the core and
ignoring interface effects at the waveguide ends, the scattered
light from the waveguide is also directly proportional to the
power inside the waveguide. Measurements of the scattered
light intensity versus the length of the waveguide can therefore
serve as an alternative method to measure loss. An exponential
fit to the scattered light intensity imaged using an 8-bit CCD
camera yielded a similar propagation loss as the cut-back
method of 0.13 dB cm21.
To compare with previously reported polymer waveguides,
90 mm 6 90 mm 6 70 mm waveguides were fabricated.
Perpendicular scattering power measurements yielded a
waveguide loss of 1.26 dB cm21 at 632 nm. If the majority
of loss results from sidewall scattering, then a 106 increase in
loss is expected for a 106 decrease in the width of the
waveguide core. This loss value compares well with the 1 dB
cm21 of other previously reported CNC machined wave-
guides19 but was achieved with more tolerant machining
parameters due to the additional vapor polishing step. The loss
is also comparable to 75 mm 6 50 mm 6 70 mm waveguides
fabricated by IBM with measured losses of 1.2 dB cm21 for an
NA of 0.29.20 Details of the loss measurements and numerical
aperture measurements are provided in the ESI.{
B. Dye doped filters
Filter waveguides were also fabricated to determine their
absorption and fluorescence characteristics. At maximum
solubility in NOA71, the filters containing pinacyanol iodide
(Sigma Aldrich) achieved an absorption per unit length of
71.8 dB cm21 at 590 nm and less than 0.30 dB cm21 at 760 nm
as measured with an Ocean Optics white light source and
spectrometer through a 1 mm path length of pinacyanol iodide
doped NOA71. Waveguide coupled dye autofluorescence was
measured by creating a 70 mm long dye doped waveguide and
injecting a 632 nm HeNe laser at the input of the waveguide.
Since the input is entirely absorbed, dye fluorescence can be
measured by measuring the waveguide output power. This
results in a 1% output power relative to the input laser power
as measured with the photodiode circuit used for actual
experiments. Pinacyanol iodide has a short lifetime (330 ps in
viscous fluids such as glycerol),21 and therefore the fluores-
cence emission adds constant intensity and phase to the
detected signal at kHz frequencies. For frequency lock-in
measurements, a filter dye with a slower lifetime would be
This journal is ß The Royal Society of Chemistry 2007
preferable to provide both absorption based attenuation and
frequency response attenuation.
C. Silvered vertical couplers
To test the effects of air plasma on silver surface deposition
quality, 300 nm silver films were deposited on both air plasma
treated and native flat PDMS surfaces. While silver evaporated
on air plasma treated surfaces achieved reflectivities of 98% as
expected, silver evaporated on untreated PDMS surfaces
resulted in microcracks, reducing reflectivity to 90%. In
addition to silver deposition quality, bend roughness from
milling resulted in variations in surface flatness. In comparison
to flat surfaces, measurements on vertical coupler surfaces
resulted in a reduction in reflectivity from 98% to 96%, in
agreement with Monte Carlo simulations of a rough bend
incorporating measured roughness.
D. System performance
By characterizing all of the coupling, roughness, and material
losses in the experimental system, simulated system perfor-
mance was found to agree with experimental system perfor-
mance. Overall efficiency from LED output to collected
luminescence was calculated to be 0.0109% while measured
results were 0.01%. A detailed description of the system losses
and characterization measurements are described in ESI.{
To compare the waveguide collection efficiency with
conventional fiber bundles, we also tested a fiber bundle with
a 1 mm core diameter PMMA excitation fiber surrounded by
0.5 mm core diameter PMMA collection fibers. Measurements
performed with the fiber bundle on the same oxygen sensor
required 9 collection fibers in order to yield a similar efficiency
of 0.0095%. Out of 9 collection fibers, a maximum efficiency of
0.0014% was measured for a single fiber.
E. Intensity detection (fluorescein fluorescence)
For intensity based fluorescence detection, a 500 mm thick 1 6
2 cm2 area microfluidic well was placed above the waveguide
detection area. Sodium fluorescein (Sigma Aldrich) in deio-
nized water was pumped through the well in increasing
concentrations starting with concentrations below the detec-
tion limit of the system. Due to the close excitation and
emission spectra, external filters were used. The input
excitation to the waveguide consisted of a filtered (Omega
Optical HQ480/406) 470 nm fiber coupled LED with an
output power of 8.6 mW. The fiber from the LED was coupled
to a second glass fiber with the excitation filter placed between
to eliminate PMMA autofluorescence, and the second fiber
was butt coupled to the waveguide. A fluorescence filter
(Omega Optical HQ510LP) was also placed between the
collection waveguide output and the silicon photodiode. The
amplifier circuit provided a gain of 3 6 106 V A21.
Fig. 4 shows the collected fluorescence intensity for different
concentrations of sodium fluorescein with the background
signal measured when DI water is in the well subtracted. From
the plot, we can determine the 3s noise level for an average of
ten 1 second acquisitions, where s is the standard deviation of
the measured intensity when the well contains no fluorescein.
This journal is ß The Royal Society of Chemistry 2007
Fig. 4 Collected fluorescence intensity versus concentration at an
input power of 8.6 mW. The measurement setup is also shown. From
the standard deviation of the data for undetectable concentrations, a
limit of detection of 200 nM can be estimated for sodium fluorescein.
From these measurements, the estimated limit of detection
(LOD) for sodium fluorescein is 200 nM with a linear
detection range up to 100 mM. With a 500 mm channel height
and an illumination area of 2 6 1 mm2 estimated in section 2,
the detection volume is 1 mL. The LOD achieved by this system
is considerably higher than other similar fluorescence detec-
tors, where an LOD of less than 0.04 nM was acheived,22,23
mainly resulting from autofluorescence of the waveguide core
material in combination with the filters used. Measurements of
the background signal show a larger detected signal than the
noise floor of the photodetector circuit by over 2 orders of
magnitude. Improvements to the LOD for fluorescein detec-
tion are possible if other core materials are used which
generate less autofluorescence.
F. Lifetime detection (oxygen sensing)
An example application replaces an array of optical fiber
bundles with the waveguide backplane for measurements of
dissolved oxygen in a microbioreactor chip.17 Sampling
oxygen quickly and accurately is necessary for feedback
control of the dissolved oxygen concentration. Due to the
high density of fluidics and control valves in this device, strictly
planar waveguides can not be implemented along side the
microfluidics. The experimental setup for measuring oxygen
concentration with the waveguide backplane consisted of four
590 nm fiber-coupled LEDs modulated at 5 kHz. To filter IR
emission from the LEDs, short pass filters (BG-39) were
placed at the end of the LED fibers and butt coupled to
secondary PMMA fibers. The second fibers were then butt
coupled to the input waveguides and four silicon photodiodes
were placed at each output guide with the option of placing an
off-chip filter (RG-9) between the output guide and the
photodiodes. Unlike the 470 nm excitation, no autofluores-
cence was detected at the photodetector with a 590 nm
excitation and RG-9 filter. The concentration of dissolved
oxygen in water was controlled by the peristaltic mixer which
mixes the liquid in the chamber with pressurized nitrogen or
pressurized air.17
The phosphorescence of the PtOEP-K dye, which composes
the oxygen sensors, is quenched by molecular oxygen, resulting
Lab Chip, 2007, 7, 1539–1545 | 1543
in a reduction in the phosphorescence lifetime. By measuring
the phase shift of the collected phosphorescence relative to the
modulated input, a signal related to the phosphorescence
lifetime of the sensor can be determined. Each sensor exhibited
a maximum phase change of 30u with a standard deviation of
0.5u from nitrogen to air when the off-chip filter was used as
shown in Fig. 5a. Sensor calibration was performed by mixing
ratios of nitrogen and air using an electrically controlled
switching valve (Lee Co.) operating at 20 Hz. Oxygen
concentration could be varied between 0% and 21% by
changing the duty cycle of the switch between nitrogen and
air. Fitting the phase response versus frequency to a first order
model resulted in phosphorescence lifetimes ranging from
19.8 ms to 58 ms, which is comparable with previously reported
calibrations.18 The linearity of the inset in Fig. 5a shows that
the extracted lifetimes are accurately described by Stern–
Volmer quenching kinetics. Using data from the frequency
response versus oxygen concentration, a calibration curve
relating phase to oxygen concentration can be determined at
specific frequencies as shown in Fig. 5b.
A step response of the oxygen concentration in water,
measured by switching the peristaltic actuation gas from
nitrogen to air, is shown in Fig. 5c. Since the lifetime based
sensor measures fluorescence quenching due to oxygen, the
noise increases when the well is air mixed versus nitrogen
mixed. The noise at maximum concentration (21% oxygen)
was obtained from the standard deviation (s) of ten 1 second
acquisitions during air mixing. By converting the deviation in
phase into a deviation in concentration, the worst case
resolution (3s) for detecting oxygen in water was estimated.
Fig. 5 (a) A measured phase vs. frequency plot (dotted) is shown for
the oxygen sensors as measured with the waveguides and off-chip
filters. Curve fits (solid) using phosphorescence lifetimes ranging from
19.8 ms to 58 ms for air and nitrogen, respectively, are also shown. (b)
The inset shows the calibration curve of normalized phase vs. %
saturated oxygen concentration at 5 kHz. (c) Measured dissolved
oxygen concentration step responses from the chip shown in Fig. 3
with 4 different waveguide sensors utilizing both off-chip filters
(dashed) and on-chip filters (solid). (d) Repeatability measurements
showing deviations in measured oxygen concentration for the four
waveguides with on-chip (square) and off-chip filters (circle).
1544 | Lab Chip, 2007, 7, 1539–1545
For off-chip glass filtered devices, the minimal resolution was
measured at 0.4% oxygen. At 100% nitrogen, a similar
measurement results in an LOD at 0.08% oxygen. For a
dissolved oxygen concentration of 9.2 ppm in air saturated
water, this translates to an LOD of 35 ppb oxygen. This LOD
is comparable to the original work characterizing PtOEP-K18
as well as other fiber based optodes2,24 and is lower than other
waveguide oxygen sensors where LODs of 600 and 300 ppb
oxygen were achieved7,25 It should be noted that while
dynamic range is not affected by intensity variations due to
the ratiometric nature of the phase measurement, both the
LOD and resolution improve with increasing intensity since
the signal to noise ratio is only between 30 and 56 for air and
nitrogen, respectively. On-chip filtered waveguides measured a
worst case resolution of 5% oxygen for air saturated water, and
an LOD of 2% oxygen for nitrogen saturated water. The
decrease in resolution can be attributed to autofluorescence in
the dye doped collection waveguide which further reduces the
signal to noise ratio.
To test the waveguide sensor repeatability, each waveguide
sensor was exposed to the same set of nitrogen and oxygen
mixtures and the responses of different sensors to the same
concentrations were compared. Gas mixtures were created
with the same switching valve used for calibrations. Sensors
were exposed to six different oxygen concentrations and the
measured concentration was taken as the average of 20 acquisi-
tions. Fig. 5d shows the standard deviation of the measured
concentration between the four sensors. From the data, we can
see that the off-chip filtered waveguides measure concentration
very consistently, with intersensor deviations of only 0.025%
and 0.14% oxygen concentration for nitrogen purged and air
saturated conditions. On-chip filtered waveguides performed
worse, due mainly to signal contributions from stray excitation
and dye autofluorescence. This reduced the full scale phase
range to only 6u, inflating phase deviations. Even with a
reduced phase range, on-chip filtered waveguides, using the
same calibration curve measured in Fig. 5b, measure
intersensor deviations of 0.22% and 0.53% oxygen under
nitrogen purged and air saturated conditions, respectively, for
averaged measurements.
5. Conclusions
The polymer based optical backplane fabricated in PDMS
consisted of large area waveguides, silvered vertical couplers,
and filters. A fabrication method combining CNC milling and
vapor polishing was shown to generate sufficiently smooth
surfaces for waveguiding. Measured propagation loss for
800 mm 6 1 mm waveguides was 0.10 dB cm21 at 632 nm
and 30u silvered vertical coupler reflectivity was 96%. By
fabricating waveguides in a polymer backplane platform, low
cost optical detection solutions can be designed easily for
applications where CCDs are either too expensive or incapable
of detection, such as at IR wavelengths or for high frequency
measurements. Organic dye doping of waveguides provided
integrated filters, further reducing the complexity and cost of
off-chip optics. The fabricated backplane demonstrated
accurate oxygen concentration measurements via the phos-
phorescence lifetime of sensors located within a complex
This journal is ß The Royal Society of Chemistry 2007
microfluidic device with required input powers as low as 17 mW
for an SNR of 56. Simulations of the system indicate a point
source collection efficiency of 2.86%, suggesting that the
backplane could be used for fast particle fluorescence detection
systems such as flow cytometers in addition to the demon-
strated distributed area luminescence detection system.
References
1 C. H. Chen, High-Sensitivity Scattering-based Detection under
Symmetrical Arrayed-Waveguide Platform, presented at the IEEE
Summer Topicals – Optofluidics, Quebec City, QC, Canada, July
17–19, 2006, paper TuA4.2.
2 P. A. S. Jorge, P. Caldas, C. C. Rosa, A. G. Oliva and J. L. Santos,
Optical fiber probes for fluorescence based oxygen sensing, Sens.
Actuators, B, 2004, 103, 290–299.
3 B. W. Pogue and G. Burke, Fiber-optic bundle design for
quantitative fluorescence measurement from tissue, Appl. Opt.,
1998, 37(31), 7429–7436.
4 M. Dandin, P. Abshire and E. Smela, Optical filtering technologies
for integrated fluorescence sensors, Lab Chip, 2007, 7, 955–977.
5 D. A. Chang-Yen and B. K. Gale, An Integrated Optical Glucose
Sensor Fabricated Using PDMS Waveguides on a PDMS
Substrate, Proc. SPIE–Int. Soc. Opt. Eng., 2004, 5345, 98–107.
6 K. Windhorn, L. Meixner, S. Drost, H. Schroder, E. Kilgus,
W. Scheel, H. Reichl, F. Ebling, M. Kuchler, J. Nester, T. Otto and
T. Gebner, Polymer BioMEMs with integrated optical and fluidic
functionality, Proc. Micro Syst. Technol. 2003, 2003, 416–423.
7 D. A. Chang-Yen and B. K. Gale, An integrated optical oxygen
sensor fabricated using rapid-prototyping techniques, Lab Chip,
2003, 3, 297–301.
8 A. Llobera, R. Wilke and S. Buttgenbach, Poly(dimethylsiloxane)
hollow Abbe prism with microlenses for detection based on
absorption and refractive index shift, Lab Chip, 2004, 4, 24–27.
9 J. Moisel, J. Guttmann, H. P. Huber, O. Krumpholz, M. Rode,
R. Bogenberger and K. P. Kuhn, Optical backplanes with
integrated polymer waveguides, Opt. Eng., 2000, 39, 673–679.
10 P. M. Goodwin, W. P. Ambrose, J. C. Martin and R. A. Keller,
Spatial Dependence of the Optical Collection Efficiency in Flow
Cytometry, Cytometry, 1995, 21, 133–144.
11 S. J. Hong, S. Choi, Y. Choi, M. Allen and G. May,
Characterization of Low-Temperature SU-8 Photoresist
Processing for MEMS Applications, IEEE Conf. Workshop
ASMC 2004, 2004, 404–408.
12 V. D. McGinniss, Vaporous Solvent Treatment of Thermoplastic
Substrates, U. S. Pat. 4529563, July 16, 1985.
This journal is ß The Royal Society of Chemistry 2007
13 K. Lee, H. L. T. Lee and R. J. Ram, Large Core Polymer
Waveguides for Optical Backplanes in Microfluidic Systems,
LEOS Summer Topical Meetings, 2006, 2006, 38–39.
14 H. Schmid, H. Wolf, R. Allenspach, H. Riel, S. Karg, B. Michel
and E. Delamarche, Preparation of metallic films on elastomeric
stamps and their application for contact processing and contact
printing, Adv. Funct. Mater., 2003, 13, 145–153.
15 S. Bhattacharya, A. Datta, J. M. Berg and S. Gangopadhyay,
Studies on Surface Wettability of Poly(Dimethyl) Siloxane
(PDMS) and Glass Under Oxygen-Plasma Treatment and
Correlation With Bond Strength, J. Microelectromech. Syst.,
2005, 14(3), 590–597.
16 O. Hofmann, X. Wang, A. Cornwell, S. Beecher, A. Raja,
D. D. C. Bradley, A. J. deMello and J. C. deMello,
Monolithically integrated dye-doped PDMS long-pass filters for
disposable on-chip fluorescence detection, Lab Chip, 2006, 6,
981–987.
17 H. L. T. Lee, P. Boccazzi, R. J. Ram and A. Sinskey,
Microbioreactor arrays with integrated mixers and fluid injectors
for high-throughput experimentation with pH and dissolved
oxygen control, Lab Chip, 2006, 6, 1229–1235.
18 D. B. Papkovsky, G. V. Ponomarev, W. Trettnak and P. O’Leary,
Phosphorescent Complexes of Porphyrin Ketones: Optical
Properties and Application to Oxygen Sensing, Anal. Chem.,
1995, 67, 4112–4117.
19 D. Snakenborg, G. Perozziello, H. Klank, O. Geschke and
J. P. Kutter, Direct milling and casting of polymer-based optical
waveguides for improved transparency in the visible range,
J. Micromech. Microeng., 2006, 16, 375–381.
20 C. M. Olsen, J. M. Trewheila, B. Fan and M. M. Oprysko,
Propagation Properties in Short Lengths of Rectangular Epoxy
Waveguides, IEEE Photon. Technol. Lett., 1992, 4(2), 145–148.
21 S. Schneider, Flashlamp-pumped mode-locked dye lasers, Philos.
Trans. R. Soc. London, Ser. A, 1980, 298, 233–245.
22 R. Irawan, C. M. Tay, S. C. Tjin and C. Y. Fu, Compact
fluorescence detection using in-fiber microchannels – its potential
for lab-on-a-chip applications, Lab Chip, 2006, 6, 1095–1098.
23 R. Irawan, S. C. Tjin, X. Fang and C. Y. Fu, Integration of
optical fiber light guide, fluorescence detection system, and
multichannel disposable microfluidic chip, Biomed. Microdev.,
2007, 9, 413–419.
24 E. J. Park, K. R. Reid, W. Tang, R. T. Kennedy and R. Kopelman,
Ratiometric fiber optic sensors for the detection of inter- and intra-
cellular dissolved oxygen, J. Mater. Chem., 2005, 15, 2913–2919.
25 C. S. Burke, O. McGaughey, J. M. Sabattie, H. Barry, A.
K. McEvoy, C. McDonagh and B. D. MacCraith, Development of
an integrated optic oxygen sensor using a novel, generic platform,
Analyst, 2005, 130, 41–45.
Lab Chip, 2007, 7, 1539–1545 | 1545