Compact and silicon-on-insulator-compatible hybrid
plasmonic TE-pass polarizer
M. Z. Alam,* J. Stewart Aitchison and M. Mojahedi
Department of Electrical & Computer Engineering, University of Toronto, Toronto, Ontario M5S3G4, Canada
*Corresponding author: malam@waves.utoronto.ca
Received July 5, 2011; revised November 4, 2011; accepted November 16, 2011;
posted November 17, 2011 (Doc. ID 150478); published December 24, 2011
Hybrid plasmonic waveguides consisting of a metal plane separated from a high-index medium by a low-index
spacer have recently attracted much interest. Here we show that, by suitably choosing the dimensions and material
properties of the hybrid waveguide, a very compact and broadband TE-pass polarizer can be implemented. Finite-
difference time-domain simulation indicates that the proposed device can provide large extinction ratio with low
insertion loss for the TE mode. © 2011 Optical Society of America
OCIS Codes: 130.3120, 240.6680, 260.3910, 310.6628.
Silicon-on-insulator (SOI) technology has attracted a lot
of interest as a platform for integrated optics in recent
years. Because of the high dielectric contrast of silicon
with air and silica, SOI waveguides can guide light in a
very confined manner. The SOI platform is also compa-
tible with complementary metal-oxide semiconductor
technology and offers the potential of integration of elec-
tronics and photonics. Unfortunately, the high dielectric
contrast makes SOI waveguides highly polarization de-
pendent and this issue must be taken into account when
designing any optical system. For example, the polariza-
tion state in a standard optical fiber varies randomly and
connecting it directly to an SOI optical chip may disrupt
the proper function of the latter. One solution to this pro-
blem is to use a polarization diversity scheme where the
two polarization states are separated at the input of the
SOI chip and processed separately. This results in an
increased system size and complexity. Another solution
that is satisfactory for many practical situations is to use
a polarizer to extinguish the unwanted polarization state.
The metal clad optical polarizer consisting of a dielec-
tric waveguide separated from a metal surface by a buffer
layer is one of the most common types of integrated TE-
pass polarizer [
1]. Such a polarizer can offer a very high
extinction ratio and low insertion loss for the TE mode.
However, proper operation of the metal clad polarizer re-
quires exact phase matching between the surface plas-
mon and dielectric waveguide modes. Hence it is very
sensitive to dimensional variations. A gap plasmon wave-
guide resonantly coupled to a dielectric waveguide can
provide a very high extinction ratio over a broad wave-
length range, but the TE insertion loss for the device is
significant [
2]. A shallowly etched SOI waveguide polar-
izer has also been proposed [
3] that utilizes the polariza-
tion dependence of leakage loss to extinguish the TM
mode while the TE mode is well guided. The polarizer
is very simple to fabricate and provides excellent ex-
tinction ratio and low insertion loss for the TE mode.
However, the polarizer is 1 mm long, compromising com-
pactness, which is a key advantage of SOI photonics. A
compact SOI compatible broadband TE-pass polarizer
with low insertion loss for the TE mode is yet to be
reported.
Recently, we proposed a hybrid plasmonic waveguide
that consists of a metal plane separated from a high-
index medium by a low-index spacer layer [
4,5]. Since
then a number of hybrid plasmonic waveguides have
been proposed [
4–7] and various practical applications
have been suggested [
8–10]. A hybrid waveguide that
is especially suitable for SOI technology is shown in
Fig.
1(a) [7]. It consists of a metal slab of size w × t
separated from a silicon slab of dimension w × d (high-
index medium) by a low-index spacer of dimension
w × h (low-index medium). The structure is compatible
with SOI fabrication technology and can be easily inte-
grated with other SOI devices. The fundamental TM
mode in the hybrid guide is concentrated in close vicinity
of the metal. In contrast, the fundamental TE mode is
concentrated in the high-index region. As a result, the
TM mode is always more lossy for such a guide [
8].
For a proper choice of waveguide dimensions, the TM
mode propagating though the hybrid waveguide will suf-
fer very high attenuation, while the TE mode will be re-
latively unaffected. Inserting such a hybrid waveguide
w
t
h
d
T
(a)
D
H
T
(b)
Metal
Silica
Silicon
(c)
Output silicon
waveguide
Hybrid
polarizer
Input silicon
waveguide
Spacer
Fig. 1. Cross sections of (a) a hybrid waveguide and (b) input
and output silicon waveguides, and (c) a three-dimensional
schematic of the complete TE-pass polarizer.
January 1, 2012 / Vol. 37, No. 1 / OPTICS LETTERS 55
0146-9592/12/010055-03$15.00/0 © 2012 Optical Society of America
section between two input/output silicon waveguides
[shown in Fig.
1(b)] will result in a TE-pass polarizer
[shown in Fig.
1(c)].
A goo d TE-pass polarizer should have (a) low insertion
loss for the TE mode, (b) high insertion loss for the TM
mode, and (c) compact size. For ease of discussion in this
work, we designate the ratio of propagation loss of the
two polarizations as
η
Propagation loss of TM mode dB∕μm
Propagation loss of TE mode dB∕μm
. (1)
Fulfillment of Condition (a) requires good effective
mode index matching for the TE mode between the in-
put/output silicon waveguide and the hybrid waveguide
section. To simultaneously satisfy Conditions (a) and (b),
η has to be large. If the propagation loss for the TM mode
per unit length is large, a high TM insertion loss will be
achieved over a short polarizer length, which satisfies
Condition (c). The material properties and polarizer
dimensions must be carefully chosen to simultaneously
satisfy all three conditions.
Noble metals (gold and silver) can provide low propa-
gation loss for plasmonic modes and are the metals of
choice for most plasmonic devices. However, in the cur-
rent case, we prefer large losses for the hybrid mode and
a metal having a large imaginary part of permittivity in
the near-IR; for example, chromium (ε
r
−6.7 41 × i
at 1.55 μm wavelen gth) is a better choice. Chromium
has very good adhesion to dielectric surfaces and, hence,
will make the fabrication of the device less challenging.
Two possible choices of spacer material are silica
(ε
r
2) and silicon nitride (ε
r
4). Since the latter
choice gives a larger propagation loss for the TM mode
[
8], we choose silicon nitride as the spacer material.
Since the choice of waveguide dimensions affects the
two modes differently, the waveguide dimensions need
to be properly chosen to achieve a large value of η.
We investigated the effects of waveguide dimensions
on the modal properties by using the commercial finite
element code COMSOL Multiphysics. Figure
2 shows
the effects of changing h and d for fixed w aveguide width
(w 550 nm). As shown in Fig.
2(a), propagation loss
for the TM mode drops for large silicon thickness (d)
and spacer thickness (h). The propagation loss of the TE
mode drops at a faster rate with increasing d and h and,
hence, as shown in Fig.
2(b), η increases. Figures 2(c)
and
(d) show the variations of the real part of effective
mode index N
eff
for the TE and TM modes, respectively.
Here, effective mode index N
eff
is defined as N
eff
k∕k
0
,
where k is the real part of the propagation constant of the
guided mode and k
0
ω∕c is the free-space wavenum-
ber. A large silicon thickness (d) results in a large N
eff
for the TE mode, which, in turn, results in a better match
between the hybrid guide and the input/output silicon
waveguides. The final device dimensions chosen are:
w 550 nm, d 120 nm, h 500 nm, and t 200 nm
to simultaneously satisfy Conditions (a), (b), and (c).
For these dimensions and a chromium layer on top of
the spacer, propagation loss for the TM mode is
3.7 dB∕μm and the value of η is 152, whereas for gold
these values become 0.2 dB∕μm and 143, respectively.
Therefore, chromium is a better choice for implementing
a compact hybrid TE-pass polarizer.
The finite-difference time-domain method (FDTD) is
well known for its ability to accurately predict the perfor-
mance of plasmonic devices. Therefore, the commercial
FDTD code Lumerical was used to investigate the perfor-
mance of the final design. Because of the small skin
depth of the metal, a small mesh size inside the metal
is required for simulation of plasmonic devices. To
achieve high accuracy without excessive demand of com-
putational resources, we used a nonuniform mesh with a
finer mesh in the metal and relatively coarse mesh in the
dielectric. We carried out simulations with different mesh
sizes and found that decreasing the minimum mesh size
below 10 nm does not significantly affect the simulation
results. Therefore, we used a nonuniform mesh with a
minimum mesh size of 10 nm in our simulations. Material
properties for silicon, silica, and chromium for various
wavelengths are taken from [
11]. The material properties
of silicon nitride at telecommunication bands are not
available from [
11]. Different permittivity values of sili-
con nitride are reported in literature by different groups,
but the values are all close to 4. Therefore, for silicon ni-
tride, we used a constant permittivity value (ε
r
4). The
permittivity of silicon nitride changes by less than 2%
over the wavelength range of our investigation (1.4–
1.6 μm ) and neglecting the dispersion of silicon nitride
should not significantly affect the results. The input/out-
put silicon waveguide dimensions are chosen to be D
350 nm and H 310 nm. To minimize device length, no
taper is used between the hybrid waveguide and silicon
waveguides. The computational volume is 30 μm long,
4 μm wide in the lateral direction, and 3.5 μm wide in
the vertical direction, and is terminated with perfectly
matched layers. The simulations were carried out on
multiple processors in parallel on the high-performance
computational facility West-grid.
Figure
3 shows the insertion loss spectrum for the TE
and TM modes of a 17-μm-long polarizer. The hybrid
Fig. 2. (Color online) Variations of the modal characteristics
with spacer and silicon thickness (h and d). (a) Propagation
loss of the TM mode. (b) Ratio of TM and TE loss (η).
(c) and (d) N
eff
for the TE and TM modes. Device dimensions
are w 550 nm, t 200 nm, and T 1.5 μm, and the wave-
length of operation is 1.55 μm.
56 OPTICS LETTERS / Vol. 37, No. 1 / January 1, 2012
waveguide placed between two silicon waveguides act as
a Fabry–Perot cavity, as evident from the oscillations in
the TE transmission spectrum [Fig.
3(a)]. The large
propagation loss diminishe s the Fabry–Perot effect in
case of TM mode and the TM transmission spectrum
[Fig.
3(b)] exhibits no oscillation. For a given extinction
ratio, the device is more compact than previously re-
ported metal clad pola rizers [
1, 2, 12].To the best of
our knowledge, the most compact broadband TE-pass
polarizer for SOI reported so far has a length of
120 μm[
2]. Simulation results for the polarizer reported
in [
2] predicted an insertion loss of more than 30 dB for
the TM mode and 2 to 4 dB insertion loss for the TE mode
over a bandwidth of 350 nm. In contrast, as shown in
Figs.
3(a) and (b), our proposed TE-pass polarizer suffers
much lower insertion loss for the TE mode (approxi-
mately 1 dB) and can achieve greater than 30 dB inser-
tion loss for the TM mode for more than 160 nm
bandwidth. Although the bandwidth is lower than that
reported in [
2], it covers the entire S and C bands and
the device length is only 14.2% of that reported in [
2].
Figure
4 shows the insertion losses for the TE and TM
modes for various polarizer lengths at 1.55 μm wave-
length. A large extinction ratio (difference between the
insertion losses for the TE and TM modes) can be
achieved for a very short polarizer length. For example,
an 8-μm-long polarizer provides an extinction ratio of ap-
proximately 20 dB. The TE insertion loss undergoes sinu-
soidal variation as a function of polarizer length due to
the Fabry–Perot effect. The insertion loss is slightly high-
er for a polarizer length of around 14 μm. For the opti-
mized design, most power for the TE mode propagates
through the hybrid waveguide from the input to output
silicon waveguide. However, part of the power leaks into
the buried oxide layer, reflects back at the buried oxide/
silicon substrate interface, and then couples to the output
guide. The amount of this coupling and, hence, insertion
loss, is maximum for a certain polarizer length. This
phenomenon is well known for waveguide polarizers
and is explained in more detail in [
13].
In conclusion, we have proposed a compact, SOI com-
patible hybrid TE-pass polarizer that can be directly in-
tegrated with silicon nanowires on SOI wafer without
using any taper. The extinction ratio is very high and
the insertion loss for the TE mode is comparable to or
less than other recently reported SOI compatible TE-pass
polarizers having much larger footprints.
References
1. V. K. Sharma, A. Kumar, and A. Kapoor, Opt. Commun.,
284, 1815 (2011).
2. I. Avrutsky, IEEE J. Sel. Top. Quantum Electron. 14, 1509
(2008).
3. D. Dai, Z. Wang, N. Julian, and J. E. Bowers, Opt. Express
18, 27404 (2010).
4. M. Z. Alam, J. Meier, J. S. Aitchison, and M. Mojahedi, in
Conference on Lasers and Electro-Optics/Quantum
Electronics and Laser Science Conference and Photonic
Applications Systems Technologies, OSA Technical Digest
Series (Optical Society of America, 2007), paper JThD112.
5. M. Z. Alam, J. Meier, J. S. Aitchison, and M. Mojahedi, Opt.
Express 18, 12971 (2010).
6. R. F. Oulton, V. J. Sorger, D. A. Genov, D. F. P. Pile, and X.
Zhang, Nat. Photon. 2, 496 (2008).
7. D. Dai and S. He, Opt. Express 17, 16646 (2009).
8. M. Z. Alam, J. S. Aitchison, and M. Mojahedi, Appl. Opt. 50,
2294 (2011).
9. R. F. Oulton, V. J. Sorger, T. Zentgraf, R-M. Ma, Christopher
Gladden, Lun Dai, G. Bartal, and X. Zhang, Nature 461, 629
(2009).
10. M. Z. Alam, F. Bahrami, J. S. Aitchison, and M. Mojahedi, in
CLEO 2011—Laser Applications to Photonic Ap-
plications, OSA Technical Digest (CD) (Optical Society
of America, 2011), paper JWA111.
11. E. D. Palik, Handbook of Optical Constants of Solids
(Academic, 1985).
12. G. Li and A. Xu, J. Lightwave Technol. 26, 1234 (2008).
13. Q. Wang and S.-T. Ho, IEEE Photon. J. 2, 49 (2010).
Fig. 3. Variations of insertion loss with wavelength for a 17-
μm-long polarizer for the final design. (a) TE mode. (b) TM
mode.
Fig. 4. (Color online) Variations of TE and TM insertion losses
with polarizer length for the final design.
January 1, 2012 / Vol. 37, No. 1 / OPTICS LETTERS 57
