Integrated Refractive Index Optical Ring
Resonator Detector for Capillary Electrophoresis
Hongying Zhu,†Ian M. White,†Jonathan D. Suter,†Mohammed Zourob,‡and Xudong Fan*,†
Department of Biological Engineering, 240D Life Sciences Center, University of Missouri-Columbia, Columbia, Missouri
65211, and Institute of Biotechnology, University of Cambridge, Tennis Court Road, Cambridge CB2 1QT, UK
We developed a novel miniaturized and multiplexed, on-
capillary, refractive index (RI) detector using liquid core
optical ring resonators (LCORRs) for future development
of capillary electrophoresis (CE) devices. The LCORR
employs a glass capillary with a diameter of ∼100 µm and
a wall thickness of a few micrometers. The circular cross
section of the capillary forms a ring resonator along which
the light circulates in the form of the whispering gallery
modes (WGMs). The WGM has an evanescent field
extending into the capillary core and responds to the RI
change due to the analyte conducted in the capillary, thus
permitting label-free measurement. The resonating nature
of the WGM enables repetitive light-analyte interaction,
significantly enhancing the LCORR sensitivity. This LCORR
architecture achieves dual use of the capillary as a sensor
head and a CE fluidic channel, allowing for integrated,
multiplexed, and noninvasive on-capillary detection at any
location along the capillary. In this work, we used electro-
osmotic flow and glycerol as a model system to demon-
strate the fluid transport capability of the LCORRs. In
addition, we performed flow speed measurement on the
LCORR to demonstrate its flow analysis capability. Finally,
using the LCORR’s label-free sensing mechanism, we
accurately deduced the analyte concentration in real time
at a given point on the capillary. A sensitivity of 20 nm/
RIU (refractive index units) was observed, leading to an
RI detection limit of 10-6RIU. The LCORR marries
photonic technology with microfluidics and enables rapid
on-capillary sample analysis and flow profile monitoring.
The investigation in this regard will open a door to novel
high-throughput CE devices and lab-on-a-chip sensors in
Capillary electrophoresis (CE) is a rapid, high-resolution
analytical tool and has been used extensively in many chemical
and biomedical applications,1,2ranging from genome sequencing,3
proteomics,4and clinical and environmental sample analysis5,6to
chemical cytometry.7Detection of CE is typically carried out at
the terminal end of the capillary.1,2However, in this detection
scheme, only the final separated analyte is analyzed, and the
detailed information regarding how samples move in the capillary
is completely lost.8,9To take full advantage of the high resolution
and low sample volume achieved by CE, on-capillary detection is
highly desirable.8,9As compared to end-capillary detection, on-
capillary detection can analyze sample separation by providing
vital information about the flow profile, such as flow rate and
dispersion, while also permitting detection with low sample
volume.10,11In particular, for CE modes that exhibit self-concentra-
tion and self-focusing effects, such as capillary isoelectric focusing,
on-capillary detection becomes critical due to its capability of direct
monitoring of self-focusing dynamics and fast sample analysis.12
On-capillary detection has been accomplished using UV
absorption and laser-induced fluorescence (LIF) spectroscopy;2,8,13,14
however, the sensitivity in UV absorption is generally low due to
the short light absorption path and small sample volume. In
addition, UV absorption suffers from nonlinear response to sample
concentration.9Although LIF is the most sensitive detection
technique, this method requires the presence of chromophores
that are either inherent to the analyte or introduced through
labeling. In the former case, intrinsic chromophores may not
readily fit the laser lines available in the lab, and their fluorescence
is normally weak, leading to the poor signal-to-noise ratio. In the
latter situation, labeling processes are laborious and may interfere
with the analyte’s biochemical functions.
Recently, refractive index (RI) has been used in CE for on-
capillary detection9,11-20and end-capillary detection,21as well as
* To whom correspondence should be addressed. Phone: 573-884-2543.
Fax: 573-884-9676. E-mail: email@example.com.
†University of Missouri-Columbia.
‡University of Cambridge.
(1) Kraly, J.; Fazal, M. A.; Schoenherr, R. M.; Bonn, R.; Harwood, M. M.; Turner,
E.; Jones, M.; Dovichi, N. J. Anal. Chem. 2006, 78, 4097-4110.
(2) Landers, J. P. Handbook of Capillary Electrophoresis, 2nd ed.; CRC Press:
Boca Raton, 1996.
(3) Dovichi, N. J.; Zhang, J. Angew. Chem., Int. Ed. 2000, 39, 4463-4468.
(4) Guzman, N. A.; Phillips, T. M. Anal. Chem. 2006, 77, 60A-67A.
(5) Hu, S.; Dovichi, N. J. Anal. Chem. 2002, 74, 2833-2850.
(6) Thormann, W.; Lurie, I. S.; McCord, B.; Marti, U.; Cenni, B.; Malik, N.
Electrophoresis 2001, 22, 4216-4243.
(7) Krylow, S. N.; Starke, D. A.; Arriaga, E. A.; Zhang, Z.; Chan, N. W. C.; Palcic,
M. M.; Dovichi, N. J. Anal. Chem. 2000, 72, 872-877.
(8) Bruno, A. E.; Gassmann, E.; Pericles, N.; Anton, K. Anal. Chem. 1989, 61,
(9) Bruno, A. E.; Krattiger, B.; Maystre, F.; Widmer, H. M. Anal. Chem. 1991,
(10) Staller, T. D.; Sepaniak, M. J. Electrophoresis 1997, 18, 2291-2296.
(11) Pawliszyn, J. Anal. Chem. 1988, 60, 2796-1801.
(12) Wu, X.-Z.; Pawliszyn, J. Electrophoresis 2002, 23, 542-549.
(13) Schrum, K. F.; Lancaster, J. M., III; Johnston, S. E.; Gilman, S. D. Anal.
Chem. 2000, 72, 4317 - 4321.
(14) Swinney, K.; Bornhop, D. J. Electrophoresis 2000, 21, 1239-1250.
(15) Bornhop, D. J.; Dovich, N. J. Anal. Chem. 1987, 59, 1632-1636.
(16) Markov, D.; Dotson, S.; Wood, S.; Bornhop, D. J. Electrophoresis 2004, 25,
(17) Wang, Z.; Bornhop, D. J. Anal. Chem. 2005, 77, 7872-7877.
Anal. Chem. 2007, 79, 930-937
Analytical Chemistry, Vol. 79, No. 3, February 1, 2007
10.1021/ac061279q CCC: $37.00© 2007 American Chemical Society
Published on Web 12/19/2006
in microchip-based fluidic systems.22,23RI detection features a
number of advantages in comparison to UV or fluorescence
detection. First, it is noninvasive to virtually all analytes; second,
it is relatively simple to implement and provides a sensing signal
regardless of the existence of absorbing or fluorescent chro-
mophores. As a result, RI detection is considered to be a
universally applicable method in microfluidics detection.14,17,18,23
Moreover, RI detection measures the analyte concentration instead
of mass; the detection signal does not scale down with the sample
volume, which makes RI detection particularly attractive when
ultrasmall (pico-/nanoliter) detection volume is involved.17,23RI
detection has been implemented in the form of forward scatter-
ing,9,18,23backscattering,17surface plasmon resonance,21waveguide,22
and Fabry-Perot interferometry.24Detection limits ranging from
10-5to 10-9RIU have been reported.2,9,11-24
Although RI detection has been shown to be a promising
technique in CE development, miniaturization and multiplexing
of RI detectors have been a challenge.21The equipment involved
is usually bulky and highly sophisticated, which limits the RI
detection to only one or two spots along the capillary. Further-
more, due to relatively large dimensions of the detection spot on
the capillary, the resolution in flow spatial profile may be
In this article, we have developed a novel miniaturized and
multiplexed on-capillary RI detection architecture based on liquid
core optical ring resonators (LCORRs). The concept of the LCORR
is illustrated in Figure 1. The LCORR employs a piece of glass
capillary with a diameter of a few tens to a few hundreds of
micrometers and a wall thickness of a few micrometers. The
circular cross section of the capillary, shown in Figure 1B, forms
an optical ring resonator along the capillary. The ring resonator
supports the circulating mode of the light, called whispering
gallery modes (WGMs),25which can be launched by bringing the
LCORR into contact with an optical waveguide perpendicular to
the LCORR. The capillary wall is sufficiently thin (<4 µm) so that
the WGM evanescent field extends beyond the capillary interior
surface and detects the analyte in the capillary. Despite the small
physical size of the ring resonator, the resonating nature of the
WGM enables repetitive light-analyte interaction, thus signifi-
cantly enhancing the LCORR detection length and sensitivity. The
effective detection length Leffcan be characterized by the Q-factor
of the ring resonator, which determines the number of light
circulation cycles supported by the resonator,25
where λ is wavelength and n is the RI of the LCORR (n ) 1.45).
For example, for an LCORR with a Q-factor of 106and λ )
980 nm, Leffcan be as long as 10 cm. Such a long detection length
is achieved without sacrificing the CE resolution, in contrast to
other capillary designs such as “Z-shaped” and “bubble-shaped”
capillaries.2Moreover, as shown in Figure 1, the LCORR is
scalable to a 2-dimensional array, which not only is important for
high-throughput CE development, but also allows us to set up a
reference channel to reduce thermal induced noise, the biggest
limitation in RI detection.14,17
The LCORR is unique in that it achieves dual use of the
capillary as a sensor head and as a CE fluidic channel. The LCORR
belongs to the field of optical ring resonator sensors that has
recently been under intensive investigation for bio/chemical
detection.26-35As discussed in the Theory Section, these sensors
utilize RI as the sensing signal; the WGM spectral position shifts
in response to the RI change induced by either the bulk solution
change near the resonator surface or binding of the analyte to
the surface. A detection limit on the order of 10-7RIU and 10-6
(18) Swinney, K.; Markov, D.; Bornhop, D. J. Analyst. 1999, 124, 221-225.
(19) Krattiger, B; Bruin, G. J. M.; Bruno, A. E. Anal. Chem. 1994, 66, 1-8.
(20) Deng, Y.; Li, B. Appl. Opt. 1998, 37, 998-1005.
(21) Whelan, R. J.; Zare, R. N. Anal. Chem. 2003, 75, 1542-1547.
(22) Lenney, J. P.; Goddard, N. J.; Morey, J. C.; Snook, R. D.; Fielden, P. R.
Sens. Actuators, B 1997, 38-39, 212-217.
(23) Burggraf, N.; Krattiger, B.; de Moller, A. J.; de Rooij, N. F.; Manz, A. Analyst.
1998, 123, 1443-1447.
(24) Woodruff, S. D.; Yeung, E. S. Anal. Chem. 1982, 54, 2124-2125.
(25) Chang, R.; Campillo, A. Optical Processes in Microcavities; World Scientific
Pub Co., Inc.: Singapore, 1996.
(26) White, I. M.; Zhu, H.; Suter, J. D.; Hanumegowda, N. M.; Oveys, H.; Zourob,
M.; Fan, X. IEEE Sens. J., in press.
(27) Vollmer, F.; Braun, D.; Libchaber, A.; Khoshsima, M.; Teraoka, I.; Arnold,
S. Appl. Phys. Lett. 2002, 80, 4057-4059.
(28) Vollmer, F.; Arnold, S.; Braun, D.; Teraoka, I.; Libchaber, A. Biophys. J. 2003,
(29) Krioukov, E.; Greve, J.; Otto, C. Sens. Actuators, B 2003, 90, 58-67.
(30) Hanumegowda, N. M.; White, I. M.; Oveys, H.; Fan, X. Sens. Lett. 2005, 3,
(31) White, I. M.; Hanumegowda, N. M.; Fan, X. Opt. Lett. 2005, 30, 3189-
(32) Zhu, H.; Suter, J. D.; White, I. M.; Fan, X. Sensors 2006, 6, 785-795.
(33) Hanumegowda, N. M.; White, I. M.; Fan, X. Sens. Actuators, B 2006, 120,
(34) Hanumegowda, N. M.; Stica, C. J.; Patel, B. C.; White, I. M.; Fan, X. Appl.
Phys. Lett. 2005, 87, 201107-1 - 201107-3.
(35) White, I. M.; Oveys, H.; Fan, X. Opt. Lett. 2006, 31, 1319-1321.
Figure 1. (A) A conceptual illustration of multiplexed on-capillary
detection using a 2-dimensional LCORR array. A waveguide array
arranged transversely is in contact with the LCORR and launches
the WGM. (B) The cross section of an LCORR. The WGM circulates
along the LCORR circumference. Its evanescent field extends beyond
the interior surface of the LCORR and interacts with the analytes in
Analytical Chemistry, Vol. 79, No. 3, February 1, 2007
RIU has been reported in microsphere-based ring resonator
sensors and the LCORR, respectively.26,34Moreover, as compared
to current capillary technology, the LCORR adds new function-
alities of multiplexed on-capillary detection, because the LCORR
has ring resonator sensors naturally integrated along the capillary.
As shown in Figure 1, the detection light, that is, the WGM, can
be launched externally through an array of optical waveguides at
any location along the capillary, providing a sensitive, noninvasive,
and quantitative tool to monitor the analyte flowing in the core in
real time. As shown later, the waveguide lateral size is only a few
micrometers; thus, the detection position can be defined along
the capillary with a precision down to micrometers, which
significantly increases the resolution in determination of the flow
profile in the capillary in comparison with state of the art.2,13,14
Assuming that the LCORR has an inner diameter of 100 µm and
a detection window of 10 µm (determined by the extension of
the ring resonator along the LCORR), the detection volume for
each ring resonator is estimated to be <100 pL.
In this article, we first present the underlying detection theory,
followed by a discussion on fabrication and characterization of
the LCORR. Then we report on the use of the LCORR with electro-
osmotic flow (EOF) as a model system to demonstrate its
capability for fluid transport, flow analysis, and subsequent
quantitative sample analysis. We show that the LCORR is highly
compatible with current CE technology and has great potential
for development of novel miniaturized high-throughput CE de-
The WGM of the LCORR can fully be described using Mie
theory by considering a three-layered radial structure (core, wall,
and surrounding medium).35-37The radial distribution of the
WGM electrical field of an LCORR is governed by
where Jm and Hm
Hankel function of the first kind, respectively. The RI of the core,
wall, and the surrounding medium is described by n1, n2, and n3.
The terms r1and r2represent the inner and outer radius of the
LCORR, respectively, and km,lis the amplitude of the wave vector
in a vacuum for the lth-order radial WGM. The resonant
wavelength λm,l) (2π)/(km,l) can be obtained numerically from
eq 2 by matching the boundary conditions at r1 and r2. Two
examples of the electrical field distribution of the WGM are shown
in Figure 2.
For the LCORR with the radius much larger than wavelength,
the WGM resonance condition can be approximated by a simple
(1)are the mth Bessel function and the mth
where r is the ring outer radius, and m is an integer number given
in eq 2, which represents the angular momentum term. neffis the
effective RI experienced by the WGM and is determined by the
RI of the core (sample), capillary wall, and the surrounding
medium (e.g., air).37Equation 3 shows that resonance occurs for
wavelengths when an integer multiple of that wavelength matches
The WGM has an evanescent field that extends beyond the
dielectric surface and into the core. The bulk solution change near
the interior surface or binding of molecules to the interior surface
leads to a change in neff, which in turn changes the WGM
resonance condition, as indicated by eq 3. Utilization of the shifts
in the WGM spectral position as the sensor signal enables a label-
free detection that conveys quantitative and kinetic information
about the flow in the core. We have developed an in-house
simulation tool based on eq 2 that allows us to simulate the WGM
behavior under various experimental conditions, such as wall
thickness, LCORR size, and the RI of the core.
To achieve adequate sensitivity, the WGM needs to have
sufficient evanescent exposure in the core. Figure 2 compares
the WGM radial distribution for a thick-walled (5 µm) and thin-
walled (2.5 µm) LCORR of 100-µm o.d. (outer diameter) using
Mie theory.35-37For the thick-wall case, the amount of the WGM
in the core is negligible and the WGM is, thus, insensitive to the
RI change in the core. In contrast, when the wall becomes thin,
a significant fraction of light is present in the core, resulting in a
much higher sensitivity (inset in Figure 2B).
Materials. Ethanol, glycerol (99%), sodium phosphate dibasic
(Na2HPO4, 99%), and hydrofluoric acid (HF, 48%) were purchased
from Sigma-Aldrich (St. Louis, MO). Hydrofluoric acid is a
particularly dangerous inorganic acid that should be handled with
(36) Bohren, C. F.; Huffman, D. R. Absorption and Scattering of Light by Small
Particles; John Wiley & Sons, Inc.: New York; 1998.
(37) Suter, J. D.; White, I. M.; Zhu, H.; Fan, X. Appl. Opt., in press.
(38) Knight, J. C.; Dubreuil, N.; Sandoghdar, V.; Hara, J.; Lefevre-Seguin, V.;
Raimond, J. M.; Haroche, S. Opt. Lett. 1996, 21, 698-700.
Figure 2. WGM radial distribution for (A) a thick-walled LCORR
(o.d./i.d. ) 100/90 µm) and (B) a thin-walled LCORR (o.d./i.d. ) 100/
95 µm). The dashed lines show the interior and exterior surfaces.
The structure of the LCORR is water core, n ) 1.333; glass wall,
n ) 1.45; and surrounding medium, n ) 1.0. Wavelength ) 980 nm.
The inset is the theoretically calculated WGM response to the RI
change in the core for the thick-walled LCORR (curve A) with a
sensitivity of 6 × 10-4nm/RIU and the thin-walled LCORR (curve B)
with a sensitivity of 11 nm/RIU.
BJm(km,ln2r) + CHm
(r e r1)
(1)(km,ln2r) (r1e r e r2)
(r g r2) }
Analytical Chemistry, Vol. 79, No. 3, February 1, 2007
consideration for safety. All chemical agents were used without
purification and were prepared in the 18-MΩ water generated by
the Easypure-UV system from Barnstead (Dubuque, IA). The
glycerol solutions were prepared by initially dissolving the glycerol
in 0.001 M Na2HPO4buffer and then further diluting in Na2HPO4
to the desired concentrations. Fused-silica and aluminosilicate
glass tubes with 1.2-mm o.d. and ∼0.9-mm i.d. (inner diameter)
were purchased from Sutter Instruments (Novato, CA).
Fabrication and Characterization of the LCORR. Since the
thin-walled LCORRs are not readily available commercially, we
assembled an in-house pulling station that allowed us to fabricate
the LCORRs by rapidly stretching a fused-silica or aluminosilicate
glass tube while heating the center section, as illustrated in Figure
3. For a given initial tube size, the final LCORR o.d. is determined
by the pulling and feed-in speed, both of which are controlled by
a computer. For our experiments, the LCORR o.d. is designed to
be in the proximity of 100 µm. The pulling process was terminated
immediately when a sufficiently long LCORR was achieved. After
systematic investigation in pulling parameters and quality check
of the final size and the wall thickness using an optical microscope,
we found that the original o.d./i.d. ratio and circularity can well
be maintained after pulling. Therefore, the wall thickness of the
LCORR thus prepared was approximately 10 µm. With this
method, LCORRs of tens of centimeters in length can be made,
limited only by the pulling stage’s travel distance. For our
experiment, by cutting off both ends, only the center portion of
the original LCORR was used.
To further reduce the wall thickness to below 4 µm, various
concentrations of HF were pumped through the LCORR via a
peristaltic pump to slightly etch the LCORR interior wall. This
etching process, similar to the one used for fused-silica etching
reported in refs 35 and 39, was well-controlled and took 20-60
min. When the desired wall thickness was reached, pure water
was pumped through the LCORR to terminate the etching process.
Despite the thin wall, the final LCORRs still retain relatively strong
mechanical strength and can easily be handled without damage.
Experimental Setup and Detection Schemes. The details
of the experimental setup are schematically shown in Figure 4.
We used two LCORRs of 115-µm (LCORR no. 1) and 130-µm
(LCORR no. 2) o.d. with a length of ∼2 cm in the experiment.
Each LCORR was connected to two sample reservoirs (9 mm in
diameter and 5 mm in height) through UV-curable adhesives. Two
hundred volts from a high-voltage source from Spellman (New
York, NY) was placed across the LCORR, resulting in an electric
field of ∼100 V/cm. A digital ammeter from Omega (Stamford,
CT) was used to monitor the current passing through the LCORR.
An optical fiber taper with a diameter of ∼3 µm, fabricated by
stretching a single-mode optical fiber under flame,40was brought
in contact with the LCORR to couple the light from a 980-nm,
tunable diode laser from New Focus (San Jose, CA; spectral line
width <0.001 pm; repeatability <0.003 pm) into the WGM.
We have developed two approaches to detect the WGM
spectral position. In the first approach, a photodetector (no. 1 in
Figure 5) was used to monitor the light at the terminal end of the
optical fiber, whereas in the second approach, a photodetector
(no. 2 in Figure 5) was placed above the LCORR. In both
approaches, the laser periodically scanned in wavelength at a
constant power. As shown in Figure 5B, when the laser wave-
length matches the WGM resonance condition, the light couples
into the ring resonator and causes the measured transmission
power to drop, leaving a spectral dip at detector no. 1. In the
meantime, the light coupled into the LCORR is scattered off the
LCORR surface and can be detected as a spectral peak by detector
no. 2. Both the measured signals can be used to indicate the WGM
spectral position, which shifts in response to the RI change in
the LCORR core. The first approach is easy to implement, and
the second scheme is more suitable when multiple LCORRs are
used for high-throughput LCORR CE development.
In our experiment, since we used only one LCORR as a model
system, only detector no. 1 was employed. The laser scanning
rate was 5 Hz with a scanning range of 100 pm; the output power
was 1-2 mW. The entire measurement system was controlled
by a computer through a data acquisition card from National
Instruments (Austin, TX). The output power at detector no. 1 for
each scan was recorded for post-analysis using in-house spectral
(39) White, I. M.; Hanumegowda, N. M.; Oveys, H.; Fan, X. Opt. Express 2005,
(40) Cai, M.; Vahala, K. Opt. Lett. 2000, 25, 260-262.
Figure 3. The LCORR fabrication process.
Figure 4. (A) Schematic of EOF experimental setup with a single
capillary and a single optical fiber taper. (B) A picture of the
experimental setup. A thermal shield is used to reduce the thermal
fluctuation caused by air flow. White lines are drawn along the LCORR
and the taper to guide the eye. (C) A zoomed-in picture of the LCORR
in contact with the taper in the absence of the thermal shield.
Analytical Chemistry, Vol. 79, No. 3, February 1, 2007
dip detection software. Figure 4B shows a picture of the actual
system. Note that a thermal shield was used to reduce the
temperature fluctuation induced by air convection. A zoomed-in
picture in Figure 4C shows the details of the taper-LCORR system.
RESULTS AND DISCUSSION
Characterization of the LCORR. The LCORR sensor utilizes
RI changes in the sample, which are detected by shifts in the
WGM resonant wavelength. To perform quantitative detection,
the magnitude of the WGM spectral shift must be calibrated to
changes in magnitude of the RI, which is the sensitivity curve for
the LCORR. We measured this by passing an ethanol-water
mixture with well-characterized RI into the LCORR34,41while
monitoring the resulting WGM spectral shift. To avoid any
potential problem in thermally induced WGM shift caused by Joule
heat generated by EOF, we connected the LCORR to a peristaltic
pump for sample delivery for this characterization. Nevertheless,
the detection principle is the same as described in the previous
Figure 6 shows the WGM obtained from LCORR no. 2 as an
example to demonstrate how to track the WGM response. The
line width ∆λ was 0.8 pm, corresponding to a Q-factor of 1.2 ×
106(Q ) λ/δλ). The WGM shifted to a longer wavelength when
the ethanol-water mixture with a higher RI was pumped into the
LCORR initially filled with pure water (n ) 1.333). The WGM of
LCORR no. 1 could be tracked in the same manner, and its
Q-factor was 2 × 105. As shown in Figure 7A, the sensorgrams
for both LCORRs were built by monitoring the evolution of the
WGM spectral position when the ethanol was pumped into the
LCORR, followed by water rinsing after each ethanol plug. Plotting
the RI change versus the respective WGM spectral shift, Figure
7B shows a good linear fit with a sensitivity of 6.6 and 20 nm/
RIU for LCORR nos. 1 and 2, respectively. These sensitivities
correspond to a wall thickness of 2.8 µm for LCORR no. 1 and
2.3 µm for LCORR no. 2, according to the in-house simulation
tool based on Mie theory.35,37
Glycerol Calibration Curve. Upon establishment of the
LCORR sensitivity, we are able to measure the RI calibration curve
for glycerol, the analyte used in our experiment. To avoid any
potential problem introduced by EOF, we used a peristaltic pump
to drive the glycerol of various concentrations into LCORR no. 2.
The sensorgram for each concentration was recorded, and the
WGM spectral shift was then converted to the RI change using
20 nm/RIU, as plotted in Figure 8.
Demonstration of Electro-osmotic Flow. To demonstrate
the flow transport capability of the LCORR, we employed EOF
with glycerol as the analyte. EOF plays an important role in
capillary electrophoresis, because it is responsible for the bulk
fluid transport. Glycerol was chosen, because it is an electrically
(41) Ghoreyshi, A. A.; Farhadpour, F. A.; Soltanieh, M.; Bansal, A. J. Membr.
Sci. 2003, 211, 193-214.
Figure 5. (A) Schemes used for WGM spectral position detection.
(B) Transmission signal from detector no. 1 and scattering signal from
detector no. 2 indicate the WGM spectral position.
Figure 6. WGM spectral position (from LCORR no. 2) shifted in
response to the refractive index change in the LCORR core. Curves
for t ) 2 and 6 s are shifted downward by 1 and 2 V, respectively, for
clarity. The Q-factor was 1.2 × 106and was determined by λ/∆λ,
where ∆λ ) 0.8 pm is the full-width-at-half-maximum of the WGM
resonance and λ ) 980 nm.
Figure 7. (A) Sensorgrams of the WGM response to various
concentrations of ethanol for LCORR nos. 1 (bottom curve) and 2
(top curve). Ethanol concentrations (v/v) are labeled in the figure. Top
curve is vertically shifted by 40 pm for clarity. (B) LCORR sensitivity
curve obtained with a linear fit of the data in A. Sensitivity is 6.6 and
20.1 nm/RIU for LCORR nos. 1 and 2, respectively.
Analytical Chemistry, Vol. 79, No. 3, February 1, 2007
neutral marker for EOF flow measurement.22Due to HF etching,
the LCORR surface is negatively charged; therefore, EOF and
glycerol move toward the cathode, which allows us to easily
control the analyte flow direction by simply switching the voltage
In this experiment, LCORR no. 1 was first connected to two
reservoirs, as shown in Figure 4. The LCORR and reservoirs were
initially filled with 0.001 M Na2HPO4 buffer. Then glycerol,
premixed with 0.001 M Na2HPO4 buffer, was added to the
reservoir at the anode. The resulting glycerol concentration in
the reservoir was 7.1% (w/w). The liquid in both reservoirs was
maintained at the same level to avoid any pressure-driven flow.
Figure 9 shows the WGM response to glycerol. The measurement
baseline was established in the first 20 s in Figure 9 when the
core was filled with buffer. A 200-V power supply was subsequently
turned on at t ) 20 s to drive glycerol toward the cathode. At t )
100 s, an abrupt positive shift was observed in the WGM spectral
position, indicating that the glycerol reached the detection spot.
The glycerol and buffer reached equilibrium at the ring resonator
location 40 s after the onset, as evidenced by the saturation
behavior in the WGM shift. Then the polarity of the applied voltage
was changed, and 25 s later, the WGM had a negative shift,
suggesting that EOF drove glycerol out of the LCORR. The
sensorgram moved back to the original level, indicative of
complete rinsing. This process was repeated several times.
Flow Analysis with the LCORR. One of the advantages of
the LCORR is its capability of performing detection at any location
along the capillary. This feature will be very useful in flow analysis,
as exemplified in Figure 10, where we carried out flow speed
measurement using a two-channel system on LCORR no. 1. Each
channel consisted of a ring resonator whose location was defined
by the tapers in contact with the LCORR. The channel separation
was 4.6 mm. Glycerol was first added to the reservoir, and the
electric field of 90 V/cm was subsequently applied across the
LCORR at t ) 0 s. As shown in Figure 10B, at t ) 20 s, channel
1 started to shift while channel 2 remained virtually unchanged,
suggesting that the edge of glycerol flow had reached the ring
resonator in channel 1, but had not yet reached channel 2. At
t ) 70 s, channel 1 saturated, indicative of equilibrium between
glycerol and buffer solution. The same saturation was reached
for the second channel 70 s later. Assuming that the onset of
saturation corresponds to the time when the main body of glycerol
moves to the ring resonator, we obtain a flow speed of 0.066 mm/s
in the LCORR.
In the LCORR, the WGM monitors the flow passing by in real
time, and multiple detection positions can be defined at any
location; therefore, the LCORR exhibits a large dynamic range in
flow rate measurement, as compared to other technologies.13,16,42,43
Theoretically, there is no upper or lower limit in flow speed
measurement. In practice, the highest detectable flow speed is
determined by the channel separation and by the laser scanning
and data acquisition rate. Under the condition of 1-mm channel
separation and 100-Hz scanning rate, which can easily be met,
the LCORR can handle a flow rate up to 100 mm/s. Furthermore,
the flow rate can be precisely measured due to the high accuracy
in detection position determination and the high-speed sampling
rate. With higher number of channels involved, the flow rate
measurement accuracy can further be improved through
Thermal Expansion and Thermo-optic Effects. The ob-
served WGM shift could also be caused by thermal expansion
and thermo-optic effects resulting from Joule heat in EOF.
Thermal expansion changes the LCORR radius, whereas thermo-
optic and electro-optic effects change the RI of water (or buffer)
and the LCORR wall. All these effects lead to variations in the
resonance condition in eq 3.37During the experiment, the current
through the LCORR was ∼2 µA at 200 V for the glycerol
concentrations used, resulting in a power of 0.4 mW dissipated
(42) Weimer, W. A.; Dovichi, N. Appl. Opt. 1985, 24, 2981-2986.
(43) Chen, Z.; Milner, T. E.; Dave, D.; Nelson, J. S. Opt. Lett. 1997, 22, 64-66.
Figure 8. Calibration curve for glycerol. The slope is 0.138 per %
Figure 9. The spectral shift of the WGM of LCORR no. 1 versus
time for 7.1% (w/w) glycerol/buffer solution added in the reservoir
connected to the positive electrode. The voltage polarity was
alternated to drive the glycerol into and out of the LCORR.
Figure 10. (A) Schematic of flow speed measurement with two
channels. (B) Sensorgram from channels 1 and 2. The WGM shift of
each channel is normalized to its respective saturation level. The time
delay between the two channels was 70 s, as indicated by the dashed
lines. Glycerol concentration, 8.3% (w/w); electric field, 90 V/cm;
channel spacing, 4.6 mm; flow speed, 0.066 mm/s.
Analytical Chemistry, Vol. 79, No. 3, February 1, 2007
along the entire 2-cm LCORR. It is estimated that the temperature
increase was a few tenths of a degree,9which, on the basis of a
5-10 pm/K shift rate in an earlier study,37corresponds to a WGM
shift on the order of 1 pm. Therefore, we conclude that in our
experiment, the WGM shifts observed during the glycerol experi-
ment were caused by the RI change due to glycerol.
Quantitative Analysis with the LCORR. In addition to
providing a transport mechanism for the sample, the LCORR is
capable of on-capillary quantitative sample analysis. To demon-
strate this, different concentrations of glycerol were passed
through LCORR no. 2 while the WGM spectral position was
monitored. First, 200 µL of buffer was added to the cathode
reservoir and 150 µL of 0.001 M Na2HPO4buffer was added to
the anode. After establishing the baseline, 50 µL of glycerol was
injected into the reservoir at the anode. The glycerol mixture was
subsequently driven through the LCORR after 200 V was applied,
causing the WGM to shift. After each run, the LCORR was cleaned
by driving the glycerol completely out of the LCORR, as exempli-
fied in the inset in Figure 11. Then, the reservoirs were cleaned
and filled with fresh buffer, and an increased concentration of
glycerol was added to repeat the above experiment. During the
experiment, the liquid level on the anode side was kept the same
as or slightly lower than that on the cathode side to avoid any
Given the observed value of the WGM spectral shift, we can
deduce the RI change and, thus, the glycerol concentration in
the capillary, as shown in Figure 11. The RI changes were first
calculated using the sensitivity of 20 nm/RIU in Figure 7B, and
the corresponding glycerol concentrations were then obtained
using the glycerol calibration curve in Figure 8. Comparison
between the deduced (triangles) and the actual (squares) glycerol
concentrations shows a good agreement, attesting to the quantita-
tive analysis capability of the LCORR.
Detection Limit Estimation. Detection limit (DL) of the
LCORR can be estimated by
where σ is the standard deviation of the system noise in units of
pm, which determines the system spectral resolution; S is the
LCORR sensitivity in units of nm/RIU; and σ is mainly determined
by two factors, that is, the WGM spectral line width and the
LCORR thermal noise. For an LCORR with a Q-factor of 106(line
width is ∼1 pm), 0.01-0.02 pm (1/50 to 1/100 of the line width)
can be resolved relatively easily.44Thermal noise results in
fluctuations in the WGM spectral position through variations in
LCORR radius (thermal expansion effect) and in the RI of the
wall and the core (thermo-optic effect), as discussed previously.
Since water in the core has a negative thermo-optic coefficient, it
counteracts the WGM shift due to the thermal expansion and RI
change of the wall,37in contrast to other RI detection in CE, in
which a large thermo-optic effect of water is the dominant noise
source.14,17,23It has been shown that at a certain wall thickness,
the water thermal effect will completely cancel those from thermal
expansion and RI changes in the wall, resulting in zero thermal
noise.37In the LCORR design, we can utilize this phenomenon to
In a separate experiment, we characterized the noise level of
the LCORR filled with water by placing it on a copper plate on
the top of a thermoelectrical cooler from Melcor (Trenton, NJ)
controlled by a temperature controller (LDT-5910B, ILX Light-
wave, Bozeman, MT). Results plotted in Figure 12 show that σ )
0.0065 pm. Note that this noise includes the noise resulting from
data digitization, which is 0.004 pm. Therefore, we believe that
the thermal noise may be even smaller. Given S ) 20 nm/RIU
that has been achieved in LCORR no. 2, the RI detection limit is
estimated to be 1 × 10-6RIU, corresponding to a detection limit
of 80 µM for glycerol and other similar chemicals. Considering
that protein and DNA molecules typically have a differential RI of
0.2 mL/g,27,45,46the detection limit for protein and DNA is
estimated to be 5 µg/mL (approximately 100 nM for protein with
molecular weight of 50 kiloDaltons and 1 µM for 15-base single-
CONCLUSIONS AND FUTURE RESEARCH
We have developed a novel on-capillary RI detector based on
the LCORR, which integrates the sensitive and noninvasive ring
resonator sensors with the capillary. The fluid transport and
analysis capability, along with multiple-channel detection, have
been demonstrated. The detection limit of 1 × 10-6RIU was
achieved with an LCORR of 2.3-µm wall thickness. The LCORR
(44) Arnold, S.; Khoshsima, M.; Teraoka, I.; Holler, S.; Vollmer, F. Opt. Lett.
2003, 28, 272-274.
(45) Arakawa, T.; Kita, Y. Anal. Biochem. 1999, 271, 119-120.
(46) Harrington, R. E. J. Am. Chem. Soc. 1970, 92, 6957-6964.
Figure 11. Deduced glycerol concentrations (triangles) in LCORR
no. 2 vs actual concentrations in the reservoir (squares). Inset shows
the sensorgram when 0.475% (w/w) glycerol was driven into and out
of the LCORR by EOF.
Wavelength scanning step size ) 0.004 pm.
LCORR noise characterization. σ ) 0.0065 pm.
DL ) 3σ/S
Analytical Chemistry, Vol. 79, No. 3, February 1, 2007
system is highly compatible with current CE technology and can
further be implemented with integrated optical designs for
instrument miniaturization.47Moreover, the LCORR is capable of
detecting flow at multiple locations, which provides high accuracy
in tracking the flow profile in the capillary. In a futuristic design,
detection can be performed along the entire capillary so that the
sample separation can be monitored en route to the terminal end
of a capillary, which will enhance the CE effectiveness in resolution
and detection speed.
In addition to conventional CE, the LCORR may also be an
excellent platform for immunoaffinity CE (IACE)4,48and capillary
isoelectric focusing (CIEF).12IACE can potentially utilize the
LCORR’s capability of detecting the binding of the molecules to
the interior surface. We recently showed that the detection limit
for protein binding is on the order of 10 pg/mm2.47,49For IACE,
the inlet of the LCORR will be patterned with different capture
molecules for multianalyte detection.50,51Thus, the binding and
elution processes of analytes, as well as subsequent sample
separation, can be monitored using the same LCORR technology.
CIEF takes advantage of the high-density, multiplexed, on-capillary
detection capability of the LCORR that permits real-time analyte
concentration gradient monitoring and fast analyte detection.12
Much work has to be done to develop the LCORR into a full-
fledged technology. To increase the detection channel density and
to integrate and miniaturize the detection system, optical waveguide
arrays fabricated with photolithography will be used in replace-
ment of fiber tapers, as has been demonstrated in our recent
work.47These optical waveguides can be mass-produced with well-
defined spacing. High detection channel density in combination
with high detection resolution will produce tremendous informa-
tion regarding the flow in a capillary. Furthermore, a few strategies
will be implemented to improve the detection limit. For example,
we will introduce a reference channel in the proximity of the
detection channel on the same LCORR, which will significantly
reduce the common-mode noise.17,26Sensitivity will be increased
by using an even thinner-walled LCORR, as indicated in Figure
2. Switching the operating wavelength from 980 nm to a longer
wavelength, such as 1550 nm, will also enhance the sensitivity.
For example, the same LCORR with 2.3-µm-thick wall will yield a
sensitivity of 150 nm/RIU when working at 1550 nm, according
to our simulation tool based on eq 2. In a longer term, photonic
crystal structures using multiple concentric layers coated onto the
LCORR will be employed to further expose more light into the
core,52,53thus increasing the light-analyte interaction. With all
these implementations, the detection limit of 1 × 10-7RIU or lower
can be reasonably expected, which makes the LCORR a competi-
tive technology for high-throughput CE development.
The authors thank the support from a 3M Non-Tenured Faculty
Award, the Wallace H. Coulter Foundation, a MU Research
Council Award, and a Life Sciences Postdoctoral Fellowship.
Received for review July 14, 2006. Accepted November 9,
(47) White, I. M.; Oveys, H.; Fan, X.; Smith, T. L.; Zhang, J. Appl. Phys. Lett.
2006, 89, 191106-1-191106-3.
(48) Guzman, N. A. Anal. Bioanal. Chem. 2004, 378, 37-39.
(49) White, I. M.; Zhu, H.; Suter, J. D.; Oveys, H.; Fan, X. Proc. SPIE 2006,
6380, 63800F-1 - 63800F-7.
(50) Balakirev, M. Y.; Porte, S.; Vernaz-Gris, M.; Berger, M.; Arie, J.-P.; Fouque,
B.; Chatelain, F. Anal. Chem. 2005, 77, 5474-5479.
(51) Ligler, F. S.; Breimer, M.; Golden, J. P.; Nivens, D. A.; Dodson, J. P.; Green,
T. M.; Haders, D. P.; Sadik, O. A. Anal. Chem. 2002, 74, 713-719.
(52) Xu, Y.; Liang, W.; Yariv, A. Opt. Lett. 2003, 28, 2144-2146.
(53) Hart, S. D.; Maskaly, G. R.; Temelkuran, B.; Prideaux, P. H.; Joannopoulos,
J. D.; Fink, Y. Science 2002, 296, 510-513.
Analytical Chemistry, Vol. 79, No. 3, February 1, 2007