VOL. 2, NO.

4, DECEMBER 2018 5001104

Sensor integration

Refractive Index Sensor Based on Gradient Waveguide Thickness

Guided-Mode Resonance Filter
Chan-Te Hsiung and Cheng-Sheng Huang
Department of Mechanical Engineering, National Chiao Tung University, Hsinchu 30010, Taiwan

Manuscript received October 30, 2018; revised November 21, 2018; accepted November 22, 2018. Date of publication November 26, 2018; date of current
version December 11, 2018.

Abstract—In this article, a novel refractive index sensor based on gradient waveguide thickness guided-mode resonance
(GWT-GMR) is proposed. The GWT can convert an spectral information into an spatial information, which can be measured
using a charged coupled device (CCD) instead of a bulky spectrometer. A GWT-GMR sensor of size only 1261 × 200 μm2
has a detection range of 0.109 refractive index units (RIU), sensitivity of 16 239 μm/RIU, and the limit of detection of 6.9 ×
10−4 RIU, which are sufficient for many practical applications. Multiplexing can be easily implemented because of sensor
compactness and the simple readout of the CCD. Moreover, the system is suitable for handheld devices or integration
with smartphones for potential biosensing applications.

Index Terms—Grating, guided-mode resonance (GMR), refractive index (RI) sensor, subwavelength structure.

I. INTRODUCTION resonance (GMR) [8] or a photonic crystal [9] and with high resolution
because of a high-quality factor of resonance, considerable flexibility
Increasing demand for public health and other applications in dif-
in material selection and structure design, and simple readout config-
ferent fields—such as disease diagnosis, food quality control, and
uration have been studied extensively and commercialized by several
environmental monitoring—has, in previous decades, spurred the de-
companies as desktop-sized biosensor systems. Currently, wavelength
velopment of various label-free (LF) biosensors on the macro- to mi-
modulation by using broadband light sources and high-resolution spec-
croscale [1]. In contrast to traditional biosensing based on commonly
trometers [10], [11] or tunable light sources and photodetectors [12],
used labels such as fluorescence dyes, LF biosensors have several
[13], is frequently used to track optical resonance and achieve a low-
advantages including real-time monitoring of the interaction between
detection limit. Moreover, other readout systems, based on angular
bio-entities, rapid and simple detection without a labeling procedure
modulation [14], intensity modulation [15], and phase modulation
being required, and less contamination than in dye labeling. The LF
[16], have been successfully demonstrated.
biosensors can be categorized based on their transduction mecha-
Triggs et al. [17] recently proposed a chirped GMR biosensor
nism, which is used to convert the interaction between surface-bound
with a graded duty cycle; this biosensor uses a laser as an illumi-
ligands and target analytes into a measurable signal that may be elec-
nation light source and a complementary metal-oxide-semiconductor
trical, mechanical, optical, electrochemical, or thermal. Optical-based
(CMOS) sensor for readout such that the sensor can be integrated with
LF biosensors have received considerable attention because of their
smartphones without the use of an additional diffraction grating to
numerous advantages, such as excellent sensitivity, freedom from elec-
disperse the spectral information in the camera.
tromagnetic interference, remote sensing ability, simple configuration,
Recently, we demonstrated a gradient grating period GMR to be
and multiplexing capability.
used as a compact spectrometer [18]. In this study, a gradient waveg-
The detection mechanism of most optical LF biosensors is based
uide thickness GMR (GWT-GMR) was proposed and applied in an
on the change in refractive index (RI) induced by the adsorption of
RI sensor with a limit of detection (LOD) comparable to that of the
interested analytes on the sensor surface, which changes the propaga-
sensor employing GMR with a graded duty cycle. The device can
tion characteristics. The variation in RI can be evaluated using several
be easily used for biosensing applications with appropriate surface
optical properties, such as intensity, resonant wavelength, coupling
functionalization.
angle, and polarization, depending on the design and detection setup
of the optical biosensor.
The optical fiber has been a popular platform for RI sensing ap-
II. METHODS
plications due to its lightweight and flexible structure characteristics
and potential remote sensing capability [2]–[7]. On the other hand, A. Design and Fabrication
for a planar platform, the most common optical biosensor is based on
surface plasmon resonance and has been widely developed and com- Resonance in planar dielectric waveguide grating structures has
mercialized. By contrast, optical biosensors based on a guided-mode been theoretically and experimentally studied [8], [19], [20]. With
an accurate device geometry and a material selection, the GMR can
be designed such that for a broadband light source at normal in-
Corresponding author: C.-S. Huang (e-mail: csh@nctu.edu.tw).
Associate Editor: D. Uttamchandani. cidence, a particular wavelength of light triggers device resonance
Digital Object Identifier 10.1109/LSENS.2018.2883471 and is reflected back, with the remaining light transmitted through.

1949-307X C 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.

See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

5001104 VOL. 2, NO. 4, DECEMBER 2018

Fig. 1. Schematic of three-layer GWT-GMR device and its working

principle.

The reflected wavelength (resonant wavelength, λ R ) can be calculated

using the second-order Bragg condition [21] as follows: Fig. 2. (a) Schematic of the sputtering setup. (b) Fabricated GWT-
GMR device. SEM top view (c) and cross-sectional views at a TiO2
thickness of (d) 124 nm and (e) 195 nm.
λ R = n eff . (1)

Resonance is substantially affected by n eff , the effective RI, and the

grating period (). n eff represents the weighted RIs of the structure
that supports the resonant mode. Therefore, n eff is related to the RIs
of a low-RI substrate, high RI of dielectric, and cover (sample) layers,
and the thickness of the high-RI dielectric layer. The formulation for
calculating n eff can be found in the relevant literature [22].
Based on the fundamental principle of GMR, different designs and
fabrication processes have been demonstrated to realize a linear vari-
able bandstop filter, including variable grating period [23], [24], vari-
able waveguide thickness [25], [26], and variable duty cycle [17]. The
GWT-GMR used in this work consists of a three-layered structure
as shown in Fig. 1, including a substrate of polyethylene terephtha-
late (PET), a replicated grating structure of ultraviolet (UV) curable
polymer (Norland 68), and the GWT of TiO2 layer. To fabricate the
GWT-GMR, a silicon master (10 × 14 mm2 ) with a grating period
of 360 nm and a grating depth of 85 nm was fabricated with electron
Fig. 3. (a) Cross-section of one period of the simulation model used
beam lithography and reactive ion etching. Then, a replica molding
in DiffractMOD. (b) Transmission spectra for different TiO2 thicknesses
process was used to transfer the grating pattern to a Norland 68 on and a sample RI of 1.333. (c) Transmission efficiency at different TiO2
a PET sheet. Finally, a layer of TiO2 with a gradient thickness was thicknesses for a fixed wavelength of 640 nm and a sample RI of 1.333.
(d) Transmission efficiency at different TiO2 thicknesses and with dif-
sputtered on the Norland 68 to complete a three-layer GWT-GMR.
ferent sample RIs for light of a fixed wavelength of 640 nm.
To deposit a TiO2 film, dc sputtering (Ion Tech Microvac 450CB)
was used. A TiO2 target with a diameter of 4 in. was placed at the
bottom of a slightly tilted holder. To deposit gradient TiO2 , a repli- device (CCD) with minima under this particular thickness as shown
cated grating structure of Norlan68/PET was attached to a fixture [see in Fig. 1.
Fig. 2(a)]. During deposition, the sample was kept stationary. A baffle Equation (1) indicates that λ R changes with n eff . The thicker the
was placed in front of the device [see Fig. 2(a)] to obtain a larger TiO2 , the greater the n eff , generating higher λ R . When the RI of the
gradient and more repeatable graded TiO2 deposition. The position sample is increased (with n eff ), the resonance shifts to thinner TiO2 to
relative to the target and the gap between the fixture and baffle could yield the same λ R . When the sample was applied on the GWT-GMR
be further optimized to yield a desirable gradient and deposition rate. sensor, the change in the surface RI resulted in a change of the resonant
The resulting thickness was approximately 100–200 nm with a gradi- position (thickness) for a fixed incident wavelength. The amount of
ent of 7.14 nm/mm. Fig. 2(b)–(d) presents a picture of the complete dip shifts in CCD could be correlated with the degree of change in the
GWT-GMR device and scanning electron microscopy (SEM) top and RI of the sample.
cross-sectional views at different thicknesses.
C. Simulation
B. Work Principle
To validate the proposed GWT-GMR RI sensor, a simulation tool
Fig. 1 illustrates the working principle of GWT-GMR as an RI sen- (DiffractMOD, RSoft Design Group) based on rigorous coupled-wave
sor. When the light of a particular wavelength (640 nm) is incident on analysis was employed to estimate the sensor performance parame-
the GWT-GMR device at normal incidence, the light resonates in areas ters, including the sensor’s sensitivity and LOD. Fig. 3(a) presents an
with a particular thickness such that it is reflected back at this location example simulation model with rounded corners based on the SEM
and is transmitted through in areas with other thicknesses. Therefore, examination [see Fig. 2(d) and (e)] (the x- and y-directions are not to
a dip-like intensity distribution can be measured by a charged coupled scale). Fig. 3(b) presents transmission spectra for transverse magnetic
 VOL. 2, NO. 4, DECEMBER 2018
Fig. 4. Transmission spectra for (a) TE- and (b) TM-polarization at dif-
ferent locations (TiO2 thicknesses). Resonant wavelength as a function
of GWT-GMR position for (c) TE- and (d) TM-polarization.
(TM) polarization at different TiO2 thicknesses when the sample’s
RI was 1.333. If the wavelength is fixed at 640 nm (dashed line), the
transmission efficiency can be obtained as a function of TiO2 thickness
[see Fig. 3(c)]. The same processes were performed for different sam-
ple RIs from 1.333 to 1.442. The results [see Fig. 3(d)] indicate that
for different sample RIs, there is a corresponding TiO2 thickness for
which a light of 640-nm wavelength can excite the resonance resulting
in the lowest transmission efficiency.
D. Transmission Measurement
To confirm that the fabricated GWT-GMR device is a linear
variable bandstop filter, a transmission setup was used to obtain the
transmission characteristics of GWT-GMR. The GWT-GMR sensor
was mounted on a translational stage with a resolution of 10 μm. An
optical fiber (Ocean Optics Inc., Amersham, U.K., QP400-2-UV-VIS)
with a core of diameter 400 μm was connected to a broadband
light source (Ocean Optics Inc., LS-1-LL) and used for illumination
at different locations (which had different TiO2 thicknesses). The
transmitted light was transverse electric (TE)- or TM-polarized before
it was collected by another fiber, which was connected to a commer-
cial spectrometer (Ocean Optics Inc., USB2000+VIS-NIR-ES). A
translational stage was moved in 1 mm increments, and transmission
spectra were obtained for both TE- and TM-polarization at 15
positions from 0 (with thickest TiO2 ) to 14 mm [see Fig. 4(a) and (b)].
The full widths at half maximum (FWHM) for TE-polarization were
14.3–42.9 nm, broadening with increasing thickness. By contrast,
for TM-polarization, the FWHMs were 9.9–15.1 nm, and TM-
polarization was, thus, used for subsequent RI measurement because
of its narrower linewidth, which would result in potentially higher
resolution. The broader linewidth from TE resonance can be attributed
to better mode confinement [8]. For TE- and TM-polarization, an
approximately linear relation was observed between resonance and
position along the device [see Fig. 4(c) and (d)].
E. RI Measurement
A narrowband of light (640 nm) generated using a monochromator
(Spectral Products, Putnam, CT, USA, DK242) was coupled to a fiber
(Ocean Optics Inc., Largo, FL, USA, QP600-2-UV-VIS). The light
5001104
Fig. 5. (a)–(c) Intensity distribution at the CCD for three sucrose con-
centrations. Black and red curves represent original data and fitted Fano
curve, respectively. (d) Average shift in pixels with respect to 0% as a
function of sucrose concentration with 5% increments. (e) Average shift
in pixels with respect to 24% as a function of sucrose concentration in
0.1% increments.
was TM-polarized before it was incident on the GWT-GMR sensor
attached to a linear CCD (ToupTek Photonics Co., Hangzhou, Zhe-
jiang, China, LHCCD01304). During the experiment, 0% solution
(deionized water) was initially used, and sucrose was then added up to
40% in increments of 5%. Each sample was aspirated out before ap-
plying the subsequent sample. The experiment was performed thrice,
and between each run, the GWT-GMR sensor was rinsed with water.
Fig. 5(a)–(c) presents the intensity distribution for three sucrose solu-
tions. To better define the minima, the Breit–Wigner–Fano resonance
model was used to fit the original curve by using Origin Pro. Fig. 5(d)
plots the average shift in pixels with respect to 0% as a function of con-
centration. To further investigate the detection limit, experiments were
performed for sucrose solutions between 24% and 26% with 0.1% in-
crements and using the same procedure; Fig. 5(e) presents the results.
III. RESULTS AND DISCUSSIONS
A. Sensitivity and Resolution
The simulation results shown in Fig. 3(d) indicate that when the
RI was increased from 1.333 to 1.442, the TiO2 thickness decreased
from 154.5 to 145.5 nm, inducing a total lateral shift of 1261 μm
with a thickness gradient of 7.14 nm/mm. Device sensitivity can be
defined as the amount of lateral shift corresponding to the amount of
RI variation, which was 11 568.8 μm/RIU (RIU refers to refractive
index units). Note that the definition of sensitivity employed in this
study differs from that commonly used in the relevant literature of
spectrometer measurements, which is the amount of peak wavelength
shift corresponding to the amount of RI variation. The sensitivity used
in this study is related to the size of GWT-GMR. A flatter gradient can
be used to improve sensitivity at the cost of sensor footprint, which
is discussed in the subsequent section. The theoretical LOD (LODT )
was determined as the CCD resolution divided by the sensitivity and
indicated the minimum RIU that can be resolved for a single pixel.
The CCD used in this study had pixels of size 8 μm, leading to an
LODT of 6.9 × 10−4 RIU, which was demonstrated to be sufficient for
several applications [17].
Sensitivity was experimentally calculated as the slope of the linear
fitted line in Fig. 5(d), which was 16 239 μm/RIU. This sensitivity is
5001104
slightly higher than that predicted by the model. This difference can be
attributed to the different RIs of the GMR filter used in the simulation
or a difference between the sample RIs used in the simulation and
experiment. The experimental LOD (LODE ) was three times the stan-
dard deviation at 0%, which was 0.5 pixels, divided by sensitivity. The
resulting LODE of 7.39 × 10−4 RIU is comparable to that obtained by
Trigg et al. [17] (LOD of 2.37 × 10−4 RIU). Although this LODE is
not optimal compared with those of some RI biosensors in the relevant
literature, Trigg et al. [17] demonstrated that this LOD is appropriate
for several practical applications.
B. Potential Improvement
Fig. 3(d) indicates that when the RI changes from 1.333 to 1.442,
the corresponding resonant thickness changes from 154.5 to 145.5 nm.
For the current fabrication, the thickness gradient was 7.14 nm/mm,
and the resulting device is 1.261 mm in length such that the LODT that
can be achieved is 6.9 × 10−4 RIU. Because the LODT was calculated
by dividing the CCD resolution by the sensitivity, the LODT can be
improved by using a CCD with higher resolution or a TiO2 thickness
with a smaller gradient and sacrificing device footprint. For example,
if the gradient is reduced by half, and a CCD with 2 μm/pixel is used,
the LODT will be 8.35 × 10−5 RIU. Furthermore, the GWT-GMR
sensor can be further optimized in terms of grating period and material
selection. Rather than directly attaching the GWT-GMR sensor to the
CCD, an objective lens can be incorporated between them to further
improve the LOD [17].
Multiplexing can be easily implemented using a CCD and a GWT-
GMR sensor with a compact design. For example, the proposed GWT-
GMR sensor with a detection range of 0.109 RIU has a size of 0.2 ×
1.261 mm2 . A CCD of approximately 5 × 4 mm2 was sufficient for
the simultaneous detection of a panel of biomolecules with multiple
replicates.
IV. CONCLUSION
In this article, we demonstrated a unique RI sensor that employs
GWT-GMR. The sensor converts an spectral information into an spa-
tial information, allowing a simple readout device, such as a CCD or
CMOS, to replace relatively expensive and bulky spectrometers. A
sensor with a small size of 0.2 × 1.261 mm2 can yield a sensitivity of
16 239 μm/RIU, an LODE of 6.9 × 10–4 RIU, and a detection range of
0.109 RIU, which are sufficient for numerous practical applications.
The device can be further improved by several means, such as opti-
mization of the GMR geometry and material selection, using a smaller
gradient, and using a higher resolution CCD for improved LOD. The
compact size of the GWT-GMR sensor is beneficial for handheld
devices and integration with smartphones for numerous biosensing
applications.
ACKNOWLEDGMENT
This work was supported by the Ministry of Science and Technology (MOST), Taiwan
under Grants 106-3114-E-009-004 and 103-2221-E-009-120. The authors would like
to thank the Nano Facility Center, National Chiao Tung University, and the National
Nano Device Laboratories, Taiwan, for their support in fabricating and characterizing the
gradient waveguide thickness guided-mode resonance device. This manuscript was edited
by the Wallace Academic Editing.
VOL. 2, NO. 4, DECEMBER 2018
REFERENCES
[1] C. Ciminelli, C. M. Campanella, F. Dell’Olio, C. E. Campanella, and M. N. Ar-
menise, “Label-free optical resonant sensors for biochemical applications,” Prog.
Quantum Electron., vol. 37, pp. 51–107, 2013.
[2] E. Reyes-Vera, N. Gomez-Cardona, and P. Torres, “Label-free biosensor based on a
dual-core transversally chirped microstructured optical fiber,” Proc. SPIE, vol. 9157,
2014, Art. no. 91578J.
[3] P. Torres et al., “Two-core transversally chirped microstructured optical fiber refrac-
tive index sensor,” Opt. Lett., vol. 39, no. 6, pp. 1593–1596, 2014.
[4] M. R. Quan, J. J. Tian, and Y. Yao, “Ultra-high sensitivity Fabry–Perot interferometer
gas refractive index fiber sensor based on photonic crystal fiber and Vernier effect,”
Opt. Lett., vol. 40, no. 21, pp. 4891–4894, 2015.
[5] N. D. Gomez-Cardona, E. Reyes-Vera, and P. Torres, “Multi-plasmon resonances in
microstructured optical fibers: Extending the detection range of SPR sensors and a
multi-analyte sensing technique,” IEEE Sensors J., vol. 18, no. 18, pp. 7492–7498,
Jul. 2018.
[6] K. W. Li et al., “Ultrasensitive measurement of gas refractive index using an optical
nanofiber coupler,” Opt. Lett., vol. 43, no. 4, pp. 679–682, 2018.
[7] K. W. Li et al., “Spectral characteristics and ultrahigh sensitivities near the dispersion
turning point of optical microfiber couplers,” J. Lightw. Technol., vol. 36, no. 12,
pp. 2409–2415, Jun. 2018.
[8] S. S. Wang and R. Magnusson, “Theory and applications of guided-mode resonance
filters,” Appl. Opt., vol. 32, no. 14, pp. 2606–2613, 1993.
[9] B. T. Cunningham, M. Zhang, Y. Zhuo, L. Kwon, and C. Race, “Recent advances
in biosensing with photonic crystal surfaces: A review,” IEEE Sensors J., vol. 16,
no. 10, pp. 3349–3366, May 2016.
[10] D. Wawro, S. Tibuleac, R. Magnusson, and H. Liu, “Optical fiber endface biosensor
based on resonances in dielectric waveguide gratings,” Proc. SPIE, vol. 3911, pp. 86–
94, 2000.
[11] B. Cunningham, B. Lin, J. Qiu, P. Li, J. Pepper, and B. Hugh, “A plastic col-
orimetric resonant optical biosensor for multiparallel detection of label-free bio-
chemical interactions,” Sens. Actuators B, Chem., vol. 85, no. 3, pp. 219–226,
2002.
[12] C. F. R. Mateus et al., “Ultra-sensitive immunoassay using VCSEL detection sys-
tem,” Electron. Lett., vol. 40, no. 11, pp. 649–651, May 2004.
[13] C. F. R. Mateus et al., “Ultra-compact, high sensitivity label-free biosensor using
VCSEL,” Proc. SPIE, vol. 5328, pp. 140–146, 2003.
[14] X. Wei and S. M. Weiss, “Guided mode biosensor based on grating coupled porous
silicon waveguide,” Opt. Exp., vol. 19, pp. 11330–11339, 2011.
[15] Y. C. Lin, W. H. Hsieh, L. K. Chau, and G. E. Chang, “Intensity-detection-based
guided-mode-resonance optofluidic biosensing system for rapid, low-cost, label-free
detection,” Sens. Actuators B, Chem., vol. 250, pp. 659–666, 2017.
[16] P. K. Sahoo, S. Sarkar, and J. Joseph, “High sensitivity guided-mode-resonance
optical sensor employing phase detection,” Sci. Rep., vol. 7, 2017, Art. no. 7606.
[17] G. J. Triggs, Y. Wang, C. P. Reardon, M. Fischer, G. J. O. Evans, and T. F. Krauss,
“Chirped guided-mode resonance biosensor,” Optica, vol. 4, no. 2, pp. 229–234,
2017.
[18] H.-Y. Hsu, Y.-H. Lan, and C.-S. Huang, “A gradient grating period guided-mode res-
onance spectrometer,” IEEE Photon. J., vol. 10, no. 1, Feb. 2018, Art. no. 4500109.
[19] S. S. Wang and R. Magnusson, “Design of wave-guide-grating filters with sym-
metrical line-shapes and low side-band,” Opt. Lett., vol. 19, no. 12, pp. 919–921,
1994.
[20] A. Sharon, D. Rosenblatt, and A. A. Friesem, “Resonant grating waveguide struc-
tures for visible and near-infrared radiation,” J. Opt. Soc. Amer. A, Opt. Image Sci.
Vis., vol. 14, no. 11, pp. 2985–2993, 1997.
[21] R. Magnusson et al., “Photonic devices enabled by waveguide-mode resonance
effects in periodically modulated films,” Proc. SPIE, vol. 5225, pp. 20–34,
2003.
[22] I. D. Block, N. Ganesh, M. Lu, and B. T. Cunningham, “Sensitivity model for
predicting photonic crystal biosensor performance,” IEEE Sensors J., vol. 8, no. 3/4,
pp. 274–280, Apr. 2008.
[23] H. A. Lin and C. S. Huang, “Linear variable filter based on a gradient grating
period guided-mode resonance filter,” IEEE Photon. Technol. Lett., vol. 28, no. 9,
pp. 1042–1045, May 2016.
[24] C. L. Fang et al., “Tunable guided-mode resonance filter with a gradient grating
period fabricated by casting a stretched PDMS grating wedge,” Opt. Lett., vol. 41,
no. 22, pp. 5302–5305, 2016.
[25] D. W. Dobbs, I. Gershkovich, and B. T. Cunningham, “Fabrication of a graded-
wavelength guided-mode resonance filter photonic crystal,” Appl. Phys. Lett., vol. 89,
no. 12, 2006, Art. no. 123113.
[26] L. Y. Qian, D. W. Zhang, C. X. Tao, R. J. Hong, and S. L. Zhuang, “Tunable guided-
mode resonant filter with wedged waveguide layer fabricated by masked ion beam
etching,” Opt. Lett., vol. 41, no. 5, pp. 982–985, 2016.