Optics Communications 311 (2013) 338–345

Contents lists available at ScienceDirect

Optics Communications
journal homepage: www.elsevier.com/locate/optcom

Analysis, design and fabrication of optical waveguides

for Mach–Zehnder Interferometry
Dibyo Sarkar a, Iqbal Jamal b, Sushanta K. Mitra a,n
a
Micro and Nano-scale Transport Laboratory, Department of Mechanical Engineering, University of Alberta, Edmonton, Canada AB T6G 2G8
b
AQL Management Consulting Inc., 4216 109A St NW, Edmonton, Canada AB T6J2R8

art ic l e i nf o a b s t r a c t

Article history: In this work, we developed an efficient monomodal waveguide with an integrated Mach–Zehnder
Received 14 June 2013 Interferometer (MZI) configuration. We considered two different types of MZI configurations, one with
Received in revised form an angular and another with an S-bend Y-junction. We were able to determine the critical cut-off angle
17 August 2013
for the angular Y-junction and the critical cut-off radius for the S-bend Y-junction. These critical
Accepted 24 August 2013
parameters, typically requiring nanoscale resolution, ensure minimal optical losses, providing an
Available online 11 September 2013
extremely sensitive waveguide system with applications in MEMS-based biosensing.
Keywords: & 2013 Elsevier B.V. All rights reserved.
Mach–Zehnder Interferometer
Y-junction
Monomode
Simulations
Waveguide
Biosensor

1. Introduction The evanescent field has a penetration depth of 100–200 nm [7]

for a 632.8 nm wavelength light emitted from a He–Ne laser. In
Optical waveguides are used in large number of science and number of MEMS applications, this evanescent field of the wave-
engineering areas with wide range of applications – from detec- guide is often manipulated to act as a sensing mechanism for
tion of biomolecules to homeland security. Typically, optical detection of biomolecules [8–11].
waveguides consist of a middle core layer and two outer cladding Fig. 1 shows a Mach–Zehnder Interferometer configuration
layers. The core layer has a higher refractive index than the waveguide, based on the principle of total internal reflection
cladding layers which allows light to propagate through the core (TIR), on a silicon substrate. Existing literature suggests that there
due to total internal reflections. There has been a sudden surge in are different interferometer configurations available for 3-dB
research towards design and fabrication of such optical wave- coupling such as Multimode Interferometers (MMI) [12–14];
guides [1–5]. However, still certain challenges exists in terms of Directional Couplers (DC) [15,16]. These interferometers have
obtaining desired “quality” of the waveguide. In this paper, we some inherent challenges and limitations which include lack of
provide a systematic approach towards the design and fabrication single mode light field in MMIs (lower sensitivity) and complexity
of waveguides, typically for MEMS (MicroElectroMechanicalSystem) of fabrication of DCs. In this present study, we have focussed on an
applications. MZI configuration waveguide with Y-junction 3-dB couplers which
Often, in MEMS applications, a ridge structure is fabricated in has shown to overcome these challenges [17–19]. The design
the core layer to confine the light waves and guide them through consists of two Y-junction couplers connecting a sensor and a
the waveguide [6]. Due to the difference in the refractive indices of reference arm. A portion of the sensor arm has its upper cladding
core and cladding layers, these layers will have different wave layer etched out exposing the core layer called the sensing area. All
equations. Hence, there would be a discontinuity in the electro- the biofunctionalization takes place at the sensing area allowing
magnetic field at the core-cladding interface. This discontinuity for the effect of the evanescent field to come into play at the core
is compensated by an exponentially decaying electromagnetic boundary. Incoming light is split into the two arms which later
field at the core-cladding interface called the evanescent field. recombine providing output signal containing information of the
changes that have occurred in the sensor arm. The sensor arm is
the site where the refractive index change takes place. This causes
n
Corresponding author. Tel.: þ 1 780 492 5017; fax: þ 1 780 492 2200. a phase shift in the sensor arm as compared to the reference arm.
E-mail address: sushanta.mitra@ualberta.ca (S.K. Mitra). As mentioned earlier, the distortion in the evanescent field for the

0030-4018/$ - see front matter & 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.optcom.2013.08.079
 D. Sarkar et al. / Optics Communications 311 (2013) 338–345 339

microfabrication process. Also, from the existing literature it is

observed that obtaining such precise sub-micron feature wave-
guide (as required for monomode waveguides), yet having all
necessary properties related to the core and cladding materials, is
of great challenge. Although there are quite a few novel fabrica-
tion technologies for fabricating sub-micron features, for example,
the work done by Chou et al. [1] and others [2–5], yet they are
accompanied by complexities while fabricating features with
dimensions as low as 4 nm. Our main motivation was to use
cheap and reproducible fabrication techniques which overcome
challenges typically associated with fabrication of such wave-
guides and obtain the desired feature sizes, as inferred from our
simulation study.

2. Principles and simulations

2.1. Design considerations

Fig. 1. Schematic of a biosensor using Mach–Zehnder Interferometer configuration.

The design for the waveguide is based on the principle of

sensing arm with respect to the reference arm is captured by a monomodal propagation of light. To ensure that there is mono-
photodiode [9–11]. modal propagation of light as opposed to multimodal propagation
This work, for the first time, advances the quantitative basis for in order to nullify the effect of Intermodal Dispersion (or Differential
the design of Si-based Mach–Zehnder Interferometers meant for Mode Delay), the waveguide should have specific dimensions. The
optimal biomolecule detection. MZI based biosensors have seen core and cladding dimensions are already well established in the
huge changes in their design since their advent in 1993 by literature [31,32]. For example, for a waveguide to be monomodal,
Heideman et al. [20]. Although polymers such as PMMA (poly- one can define a characteristic quantity, V, given as [33]
methylmethacrylate), polyamide, polysterene, etc. had been used
to fabricate waveguides since the 1970's [21] and more recently 2π a 2
V¼ ðnCo n2Cl Þ1=2 ð1Þ
have been used in MZI-based biosensing [5,22–24], silicon based λ
waveguides have always been the preferred choice due to their where, a is the core half width, nCo and nCl are the refractive indices
much higher sensitivity [10,20]. As already discussed, in MZIs, of the core (silicon nitride) and cladding (silicon dioxide), respec-
Y-junction couplers are more advantageous than MMIs or DCs and tively and λ is the wavelength of the source. The V number is
this has ensured that most of the recent studies on MZI-based typically less than 2.405 for a monomodal waveguide. Eq. (1) shows
biosensors have frequently used Y-junctions with circular bends of that for the V number to be less than 2.405, the half width
radius 80 mm [10,11,25,26]. However, there is little discussion on (a) should be less than 177 nm, which means that the total width
the design of the Y-junctions used on these MZI configurations. of the core can be of maximum dimension of 354 nm. However, the
Therefore, a very relevant question pertains to quantifying the sensing area of the biosensor would have the upper cladding layer
physical dimensions of Y-junction MZIs based on silicon technol- etched out and replaced with a layer of biomolecules (bacteria),
ogy. But to the authors’ best knowledge, there is no literature on which have a lower index of refraction [34,35]. This constrains the
theoretical analysis and design of Si-based MZI. Our study bridges core width even further and hence a safe design would have a
this crucial gap by providing an explanation through theory and 250 nm wide core.
simulations as to why such dimensions have been used over the The ridge plays the most important role in deciding the modal
years and where the cut-off dimension of the Y-junction lies. behavior of a waveguide. Finding out the dimensions of the ridge is
Optical waveguides (or fibers) can be classified into two broad non-trivial and involves computing mode profiles and solving wave
categories: monomode and multimode. Monomode optical fiber equations. This has been done through simulations using RSOFT
allows only the lowest order bound mode (or the zeroth mode) to CADTM , a photonic design software, by RSoft Design Group, Inc.
propagate whereas a multimode fiber, as the name suggests, Another key design aspect is the Y-junction of MZI system. An MZI
allows higher modes to propagate through it. The use of mono- configuration essentially has two Y-junctions, one that diverges the
mode waveguides for optical biosensing, in order to achieve high input to the sensor and reference arms while the other converges
sensitivity, has been known in the literature for over two decades the sensor and reference arms to the output. The highest amount of
[10,11,27–30]. Therefore, in order to achieve a high sensitivity losses in such a configuration occurs due to bending losses in the
immunosensor that can detect biomolecule concentration levels Y-junctions. Therefore the Y-junctions need to be designed such that
lower than 100 CFU/ml, it is imperative to design an optical there is minimal bending loss without allowing cross coupling of
waveguide capable of rejecting any mode other than the funda- signal between the sensor and reference arms.
mental mode from propagating through it.
In this paper, we have performed simulations on various planar
waveguide dimensions to establish that monomode waveguides 2.2. Numerical analysis
indeed provide the highest sensitivity. Detailed theoretical analysis
of the waveguides have also been presented. Through our simula- Fig. 2 is a graphic representation, generated by RSOFT CADTM , of
tions we have obtained optimal dimensions for a Mach–Zehnder the boundary conditions used for the simulations. The 1:5 μm
Interferometer (MZI) configuration with the highest sensitivity in lower cladding layer and 1:75 μm upper cladding layer have a
the form of lowest optical losses during light propagation. Such refractive index of 1.46, representing the SiO2 cladding layers. The
simulation results present to us a range of waveguide dimensions 250 nm core layer, with a refractive index of 2.00, represents the
that can be used for monomode propagation. We selected an Si3N4 core layer and has a 4 μm by 4 nm ridge. The total length of
optimal dimension among those simulated waveguides for further the chip in the z-direction (not shown in figure) is 30 mm.
340 D. Sarkar et al. / Optics Communications 311 (2013) 338–345

this analysis is mainly the index distribution in the x and y

directions (see Fig. 2). The wavelength of the light at the input
(z¼ 0) is also provided and the FDBP numerical analysis computes
the wave field throughout the domain ð0 r z r 30 mmÞ.
The next part of the simulations is to calculate the number of
modes in the waveguide. As the waveguide needs to be a
monomode waveguide, modal analysis is a critical part of the
whole design. In order to do this, a propagating beam method is
used [40]. This technique generates correlation functions by
solving the wave equations to compute the modes in a structure.
The general solutions to Eqs. (6) and (8) can be expressed as
eigenfunction expansions:

ψ ðx; y; zÞ ¼ ∑An un ðx; yÞeiβn z ð9Þ

n
Fig. 2. Cross-sectional refractive index layout of the waveguide at z¼ 0. The z-axis
′
is into the plane of the paper. The color bar to the right indicates the change in
refractive index of the waveguide. (For interpretation of the references to color in
ψ ′ðx; y; zÞ ¼ ∑A′n u′n ðx; yÞeiβn z ð10Þ
n
this figure caption, the reader is referred to the web version of this article.)
where u is the eigenfunction and A is a constant. Here, β is the
propagation constant, which describes the behavior of a mode in
RSOFT CADTM uses a Finite Difference Beam Propagation (FDBP) the waveguide. The relation between β n and β′n is given as [40]
method to solve exact and paraxial wave equations derived from the
Helmholtz equation [36]. The computational effort is directly propor- βn ¼ nCl k½1ð1 þ 2βn′ =nCl kÞ1=2  ð11Þ
tional to the number of grid points used in the numerical simulation. Therefore, once the simpler Fresnel (parabolic) equation (Eq. (10))
Light propagation within the waveguide can be described in is solved and βn′ is calculated, it is easy to find the propagation
terms of the Helmholtz equation as [37] constant (βn ) for the exact wave equation or the Helmholtz
equation.
∇2 ϕðrÞ þ k n2 ϕðrÞ ¼ 0
2
ð2Þ
RSoft uses a correlation function that relates the contribution of
where, ϕ is the electric field potential, r ¼ rðx; y; zÞϵR , kð ¼ 2π =λÞ is
3 the field at z ¼0 over the whole range of the function (parabolic).
the free space wave number and nð ¼ nðx; y; zÞÞ is the refractive The correlation function (C) is represented as,
index of the device. Z Z
The electric field potential (ϕ) in Eq. (2) gives the spatial CðzÞ ¼ ψ ′ðx; y; 0Þψ ′ðx; y; zÞ dx dy ð12Þ
dependence of the electric field E. The time dependence of the
electric field can be re-written in terms of the field potential For different modes ðj ¼ 0; 1; 2…Þ, one can write
through the exponential term, as shown here [6], ψ ′ðx; y; zÞ ¼ ∑A′n;j u′n;j ðx; yÞeiβ′n z ð13Þ
n;j
iωt
Eðx; y; z; tÞ ¼ ϕðx; y; zÞe ð3Þ
Substituting Eq. (13) in Eq. (12), one obtains
In a typical waveguide, the electric field potential varies rapidly in
the z-direction (direction of propagation of the field). We factor CðzÞ ¼ ∑jA′n;j j2 eiβ′n z ð14Þ
n;j
out this rapid variance by substituting with a slowly varying field
ψ , which can be written as [38] A Fourier transform of the correlation function is performed with
respect to the axial z-direction. The resulting spectra display sharp
ϕðx; y; zÞ ¼ ψ ðx; y; zÞe inCl kz
ð4Þ resonances corresponding to mode groups, and the positions and
where, nCl is a reference refractive index or, in this case, the heights of these resonances determine the mode properties ðβ Þ.
cladding refractive index (refer Fig. 2). The Fourier transform of Eq. (14) is
Z 1
The Helmholtz equation, provided in Eq. (2) can be expanded as,
CðβÞ ¼ ∑jA′n;j j2 eiðββ′n Þz dz ð15Þ
∂2 ϕ ∂2 ϕ ∂2 ϕ n;j 1
þ k n2 ϕ ¼ 0
2
þ þ ð5Þ
∂x2 ∂y2 ∂z2 By introducing a Delta function [41]
Z 1
Now replacing Eq. (4) in Eq. (5), we get
δðββ′n Þ ¼ eiðββ′n Þz dz ð16Þ
1
∂2 ψ ∂2 ψ ∂2 ψ ∂ψ
þ 2inCl k þ ψ ðn2 n2Cl Þk ¼ 0
2
þ þ ð6Þ one can re-write Eq. (15) as [40],
∂x2 ∂y2 ∂z2 ∂z
′
which is often referred as the exact wave equation. CðβÞ ¼ ∑jA′n;j j2 δðβ βn Þ ð17Þ
n;j
Based on the paraxial approximation that ψ varies very slowly
with z, one can write [39] Since δð0Þ ¼ 1, Eq. (17) will give a spectra where the maxima is at
β ¼ β′n . This gives the β′n values which in turn lead to βn values
∂2 ψ ∂ ψ
⪡ ð7Þ from Eq. (11). These values are used to find the field profiles by
∂z2 ∂z
placing them in the eigenfunctions (Eqs. (9) and (10)).
which gives rise to a Parabolic or Fresnel form of the wave
equation: 2.3. Results of simulation
∂2 ψ ′ ∂2 ψ ′ ∂ψ ′
þ ψ ′ðn2 n2Cl Þk ¼ 0
2
þ 2 þ 2inCl k ð8Þ The propagation constant β provides information related to the
∂x2 ∂y ∂z
amount of losses in each mode. Since, the main focus is on
Since the waveguide is translationally invariant, i.e., the refractive developing low loss fundamental mode waveguides, different
index does not vary in the z-direction ðn ¼ nðx; yÞÞ, the input for combinations of width (w) and height (h) (refer to Fig. 1) of the
 D. Sarkar et al. / Optics Communications 311 (2013) 338–345 341

Table 1 the two arms (d) not exceeding 100 μm, since the arms need to be
Propagation losses in the fundamental mode or zeroth order mode (the only mode close to each other to ensure that there is no phase change signal
of light propagating in a monomode waveguide) for different combinations of ridge
at the output due to the difference in variations of temperature,
dimensions.
humidity, etc. in the two arms [11].
w (μm) h Simulation results show that for a constant value of the arm
length L, the Sbend configuration shows less loss than an angular
1 nm 3 nm 4 nm configuration. For example, for L ¼ 22 mm, the loss incurred in
2 1.185  10 7
2.769  10 9
1.353  10  9
Sbend is 0.4 dB/cm whereas for angular it is 0.56 dB/cm. On
3 4.93  10  8 9.401  10  10 1.125  10  10 repeating such simulations for various dimensions, it is found that
4 2.496  10  8 3.87  10  10 1.373  10  10 the best results were seen for angular bends with θ o 2:51 and
5 1.315  10  8 8.977  10  11 3.12  10  10 Sbends with R 415 mm. Fig. 4 compares angular bend structure
6.518  10  9 9.585  10  11 2.487  10  10
6
with different θ values (11 and 3.371). The one with an angle
7 2.849  10  9 2.995  10  10 Bimode
higher than 2.51 is observed to have large loss (nearly 3 dB/cm)
and the presence of a higher order mode in the vertical arms. This
can be explained by a phenomenon called intermodal scattering.
Sharp bends in the MZI design cause intermodal scattering, where
a fraction of power from the fundamental mode is transferred to a
higher order mode [6]. The appearance of a higher order mode in
the waveguide just after the diverging (or input) Y-junction can
explain why the loss distribution is not symmetric (refer to Fig. 4
(ii)(b)). Higher order modes are much more susceptible to bend
losses as compared to the fundamental mode and therefore
the losses are much higher in the waveguide region just after
the output Y-junction than in the section just before the input
Y-junction.
Decreasing θ to a much lower value and increasing R to a much
higher value is constrained by the size of the chip. One possible
solution is to decrease d and bring the arms closer. However, the
two MZI arms cannot be brought closer than 70 μm as cross
coupling of signal between the two arms can occur.

3. Materials and dimensions

For the fabrication of the optical waveguides, silicon nitride

(R.I.¼2.00) is chosen as the core material while silicon dioxide
(R.I.¼1.46) forms the cladding layers. Silicon nitride was selected
as the core of the waveguide because it is chemically stable
and does not allow liquids to diffuse into it from outside during
biofunctionalization steps, which are often used in biosensor
application [31]. With silicon nitride as the core, silicon dioxide
was selected as the upper and lower cladding layers since it is a
well established fact that a large difference in core and cladding
refractive indices has a direct effect on improving sensitivity of the
waveguide [29,42]. Moreover, since both silicon nitride and silicon
dioxide are transparent, light propagates through the waveguide
without significant attenuation [43].
The cross sectional view of the waveguide is shown in Fig. 5.
Ideally the core layer should have a height of 250 nm and the
Fig. 3. (a) Angular Y-junction and (b) Sbend Y-junction. upper and lower cladding layers should be more than 1:5 μm so
that the cladding layer is sufficiently thicker than the core
ridges are tried out to find an optimal design. It is found that thickness in order to have low attenuation losses [31,32]. Based
waveguides with ridge width below 8 μm and height between on the simulation results, angular Y-junctions with θ ranging from
1 and 4 nm showed monomodal behavior. Table 1 enlists different 0.91 to 2.51 and S-bend Y-junctions with R ranging from 20 mm to
combinations of the ridge width (w) and height (h) with the 180 mm, can be the ideal dimensions for MZI system. However, in
modulus of the propagation losses (calculated from β) occurring in all of the designs the ridge height and width was fixed at 4 nm
each of them. and 4 μm, respectively. The length of the waveguide is typically
Based on the calculated values of monomode losses within the about 30 mm.
waveguide, the dimensions chosen for fabricating the waveguides
are h ¼4 nm and w ¼ 4 μm, which provides one of the minimal
losses, as highlighted in Table 1. 4. Fabrication
The next step is to design the MZI structures. We perform
simulations on two different geometries. One of them has straight We fabricated waveguides with MZI structures of 4 nm height
arms with an opening angle θ called the angular Y-junction while and 4 μm width. Such low aspect ratio features possess inherent
the other one has an S-bend of radius R called the Sbend Y-junction complications in fabrication. The fabrication involved nine major
(see Fig. 3). All simulations were done with the distance between steps and all of the fabrication steps were performed in the
342 D. Sarkar et al. / Optics Communications 311 (2013) 338–345

Fig. 4. (i) a) Transverse electric field through the entire length of the MZI configuration (θ ¼ 11). (b) Variation of power with length; (ii) (a) Transverse electric field through
the entire length of the MZI configuration ðθ ¼ 3:371Þ. (b) Variation of power with length.

Fig. 6(b) is the microscopic image of a diverging Y-junction of

one of the MZI structures with an angular Y-junction of one
degree. Fig. 6(c) is the microscopic image of the area on one of
the MZI arms where the upper silicon dioxide cladding layer is
completely etched to reveal the silicon nitride core layer (also
called the sensor area).

Fig. 5. Cross sectional view of the waveguide with dimensions.

5. Results

NanoFab facility and the Micro and Nano-scale Transport Labora- 5.1. Thickness of layers and refractive indices
tory at the University of Alberta.
First, the Si wafers are cleaned in a piranha solution with Ellipsometry is used for determining the thickness and more
3 parts H2SO4 and 1 part H2O2, for 15 min. Then in order to obtain importantly the refractive indices of the SiO2 cladding layers and
a high quality oxide layer with high uniformity to form the lower the Si3N4 core layer. This system measures the thickness optically
cladding, thermal oxidation is preferred since it is very important and hence is a non-destructive technique. The thermally grown
to have maximum uniformity before deposition of the core Si3N4 SiO2 layer (refer to Section 4) is measured using an ellipsometer
layer. Low Pressure Chemical Vapor Deposition (LPCVD) of Si3N4 is and it revealed a thickness of 1:3525 7 0:0007 μm. The refractive
performed on the thermally oxidized Si wafer. LPCVD nitride can index was found to be 1.4668 70.0012 which is close to the target
easily be deposited in a very pure and uniform way which leads to refractive index of 1.46. The LPCVD nitride (refer to Section 4)
high thermal stability and low etch rates [44]. Low etch rate is an thickness was 252.6 70.3 nm with a refractive index of 2.0844 7
essential advantage of LPCVD because it can allow etching of 0.0051 which is again close to the target refractive index of 2. The
features with depths as small as 4 nm. Then, in the most important upper cladding layer was found to be 1:9964 70:0030 μm thick
step of the fabrication procedure, photolithography and Reactive with a refractive index of 1.4631 70.0012. All measurements were
Ion Etching (RIE) are used to print the MZI configuration on the taken in the spectral range of 0:3 μm–0:8 μm.
Si3N4 core. The mask used for the photolithographic step was
designed using a mask generation software from Tanner EDA, 5.2. AFM
L-Edit. The SiO2 upper cladding layer does not demand too much
of uniformity but it is very essential that the stress induced by the The ridge dimensions are measured using Atomic Force Micro-
deposition of a thick upper cladding layer is well controlled. Hence scopy (AFM). These AFM measurements are done after the first RIE
Plasma Enhanced Chemical Vapor Deposition (PECVD) process was step. Fig. 7 shows 2-D and 3-D images of a 50  50 μm area on the
used for forming the upper cladding layer. MZI chip where the diverging Y-junction is located. The RMS
Till this point, a complete MZI waveguide with two reference roughness was also calculated over various sections of the chip
arms has been fabricated. In order to create a sensing area on one before and after PECVD of the SiO2 upper cladding layer and it was
of the arms (to be called as the sensor arm), another photolitho- found to be in the range of 0.5–0.6 nm, which is acceptable.
graphic and RIE step is performed. In this case, the SiO2 upper
cladding layer is selectively etched up to the core layer over a
20 mm by 100 μm area, thereby exposing the core-cladding 6. Experimental setup
boundary where the intensity of the evanescent field is the
strongest (refer Fig. 1). Once the chips were fabricated, the end Fig. 8 is a picture of the experimental setup used for analyzing
faces were polished using diamond lapping and polishing films the fabricated MZI chips. Light from a He–Ne laser (633 nm) is
(15 μm–0:1 μm). This step was essential to prevent scattering of coupled onto the MZI chip using a 40  objective lens. Light
light at the input and output of the waveguide. emitted by the chip is directly coupled into a multimode fiber
 D. Sarkar et al. / Optics Communications 311 (2013) 338–345 343

Fig. 6. (a) Schematic of an MZI configuration. (b) Microscopic image of the diverging Y-junction (showing θ ¼ 11). (c) Microscopic image of a portion of the sensor area
revealing the silicon nitride core layer.

Fig. 7. (a) 2-D AFM image of a diverging Y-junction scanned over an area of 50  50 μm on the MZI chip. (b) 3-D image of the Y-junction.

In order to analyze the fabricated waveguides, loss measure-

ments are conducted for the MZI chips. The loss per unit length
can be written as [45]

10 log ðP in =P out Þ
dB=cm ¼ ð18Þ
L

where, Pin and Pout are the input and output powers of the
waveguide, respectively, and L is the length of the whole chip
(30 mm).
Initially, the power at the input end of the waveguide (or the
output from the laser), Pin, was compared to the power emanated
by the waveguide, Pout. It is to be noted that, during experiments,
the output from the laser is in the order of milliwatts (mW)
whereas the output from the waveguide is in microwatts ðμWÞ.
This indicates that a fair amount of light from the laser failed to
enter the waveguide thereby creating anomalies in the loss
analysis. Therefore, instead of measuring P in as the power from
the laser, the waveguide is cut into half and the corresponding
output power from the chip of length L/2 is taken to be Pin while
Fig. 8. Laboratory setup of our experiment used for power measurements of the
Pout remained the output power from an entire chip. Using this in
intensity of the light emanating from the waveguide.
Eq. (18) and making the length half of its original value
ðL ¼ 15 mmÞ, the actual loss per cm of the waveguide was
(62:6 μm core diameter). The multimode fiber transmits the light determined. Fig. 9(a) and (b) shows the image of monomodal
into a photodiode for power measurements. The chip is placed on light field at the output end of the entire chip and at the section
an XYZ stage with a piezo controlled resolution of 20 nm. In order cut across the mid-way between the input and the output
to align the laser light into the sub-wavelength core (250 nm), Y-junctions, respectively. The two spots in such mid-section
a CMOS camera, connected to a PC, is placed above the chip to (Fig. 9(b)) are due to light emitting from the two arms (reference
simplify the optical alignment of the device. and sensing) of the fabricated MZI.
344 D. Sarkar et al. / Optics Communications 311 (2013) 338–345

Fig. 9. CMOS camera images, through a 5  objective, of the output end of (a) the full chip and (b) the mid-section of the chip. The images show a monomode light field.
The two bright spots in (b) refers to the light emitting from the reference and the sensor arms of the MZI chip.

wave equations gave mode profiles of different waveguide config-

urations. A 4 nm high and 4 μm wide ridge structure was chosen
as the dimension for fabrication. Second, the MZI configuration is
designed. The choice of waveguide configurations have been done
through loss analysis. For the first time, a dimensional constraint
of the MZI configuration has been quantified for optical biosen-
sors. From our simulation results, we have inferred that wave-
guides with Y-junctions above a 2.51 angle or below a 15 mm
radius will have extremely high losses. The fabrication steps
involved in making sub-micron optical MZI waveguides have also
been discussed. Finally, we performed loss analysis experiments
on the fabricated MZI chips to confirm our simulation results and
found them to be in congruence with each other. The character-
ization of fabricated waveguide through measuring the loss
component further confirmed that the fabricated waveguide can
be used for biosensors and other MEMS applications.

Acknowledgments

The financial supports from NSERC-CRD program and AQL Inc.

are highly appreciated. We also gratefully acknowledge the useful
discussions with Stefania Dante, PhD student at Research Center
on Nanoscience and Nanotechnology (CIN2), and Spanish National
Fig. 10. (a) The variation of propagation loss with the radius of curvature of Research Council (CSIC), Barcelona, Spain.
Y-junction graph for Sbend MZI. (b) The variation of propagation loss with the
angle of Y-junction graph for angular MZI. References

The losses calculated range from 0.1 dB/cm to 0.7 dB/cm which [1] S. Chou, P. Krauss, Microelectronic Engineering 35 (1997) 237.
[2] R. Soref, J. Larenzo, IEEE Journal of Quantum Electronics 22 (1986) 873.
follow the simulation results very closely. A comparison of the loss
[3] A. Rickman, G. Reed, F. Namavar, Journal of Lightwave Technology 12 (1994) 1771.
values calculated through simulations and experiments for differ- [4] A. Crunteanu, M. Pollnau, G. Jänchen, C. Hibert, P. Hoffmann, R. Salathé,
ent MZI configurations is given in Fig. 10. Propagation loss (dB/cm) R. Eason, C. Grivas, D. Shepherd, Applied Physics B 75 (2002) 15.
values are plotted against increasing R values for Sbend (Fig. 10(a)) [5] R. Bruck, E. Melnik, P. Muellner, R. Hainberger, M. Lämmerhofer, Biosensors
and Bioelectronics 26 (2011) 3832.
and θ values for angular (Fig. 10(b)) MZI configurations. The chips [6] R.R. Syms, J.R. Cozens, Optical Guided Waves and Devices, McGraw-Hill,
used for loss measurements do not have a sensing area etched on London, 1992.
them. During biosensing experiments, the etched sensor area [7] A. Rohrbach, Biophysical Journal 78 (2000) 2641.
[8] F. Brosinger, H. Freimuth, M. Lacher, W. Ehrfeld, E. Gedig, A. Katerkamp,
would encounter losses due to the discontinuity in the upper F. Spener, K. Cammann, Sensors and Actuators B: Chemical 44 (1997) 350.
cladding. However, as biomolecules start immobilizing on the [9] O. Parriaux, G. Veldhuis, Journal of Lightwave Technology 16 (1998) 573.
sensor surface, it is expected that the losses would start to [10] F. Prieto, B. Sepulveda, A. Calle, A. Llobera, C. Domínguez, A. Abad, A. Montoya,
L. Lechuga, Nanotechnology 14 (2003) 907.
decrease with an increase in concentration of the in situ biomo- [11] L.M. Lechuga, K. Zinoviev, L. Fernandez, J. Elizalde, O. Hidalgo, C. Dominguez,
lecule solution. Proceedings of SPIE 7220 (2009) 72200L. (1-8).
[12] L. Spiekman, Y. Oei, E. Metaal, F. Green, I. Moerman, M. Smit, IEEE Photonics
Technology Letters 6 (1994) 1008.
[13] L.B. Soldano, E.C. Pennings, Journal of Lightwave Technology 13 (1995) 615.
7. Conclusions [14] C. Themistos, B. Rahman, Applied Optics 41 (2002) 7037.
[15] C.P. Wen, IEEE Transactions on Microwave Theory and Techniques 18 (1970) 318.
[16] R. Forber, E. Marom, IEEE Journal of Quantum Electronics 22 (1986) 911.
In this study, we have performed simulations for determining
[17] M. Izutsu, Y. Nakai, T. Sueta, Optics Letters 7 (1982) 136.
waveguide dimensions for monomodal propagation of light using [18] L.H. Frandsen, P.I. Borel, Y. Zhuang, A. Harpøth, M. Thorhauge, M. Kristensen,
RSOFT CADTM . Extensive theoretical analysis for mode calculation W. Bogaerts, P. Dumon, R. Baets, V. Wiaux, et al., Optics Letters 29 (2004) 1623.
in waveguides has been shown. Two aspects of the design of MZI [19] A. Liu, L. Liao, D. Rubin, J. Basak, Y. Chetrit, H. Nguyen, R. Cohen, N. Izhaky,
M. Paniccia, Semiconductor Science and Technology 23 (2008) 064001. (1–7).
waveguides are taken into consideration. First, a monomodal [20] R. Heideman, R. Kooyman, J. Greve, Sensors and Actuators B: Chemical 10
waveguide was designed. Numerical solutions of the theoretical (1993) 209.
 D. Sarkar et al. / Optics Communications 311 (2013) 338–345 345

[21] J. Harris, R. Shubert, J. Polky, Journal of the Optical Society of America 60 [33] D. Marcuse, Journal of the Optical Society of America 66 (1976) 216.
(1970) 1007. [34] M. Jonasz, G. Fournier, D. Stramski, Applied Optics 36 (1997) 4214.
[22] D. Esinenco, S. Psoma, M. Kusko, A. Schneider, R. Muller, Reviews on Advanced [35] A.E. Balaev, K.N. Dvoretski, V.A. Doubrovski, Proceedings of SPIE 5068 (2003) 375.
Materials Science 10 (2005) 295. [36] R. Scarmozzino, A. Gopinath, R. Pregla, S. Helfert, IEEE Journal on Selected
[23] R. Műller, P. Obreja, M. Kusko, D. Esinenco, C. Tibeica, G. Conache, L. Buia, Topics in Quantum Electronics 6 (2000) 150.
D. Apostol, V. Damian, M. Mateescu, et al., Proceedings of SPIE 5972 (2005) [37] P. Moar, S. Huntington, J. Katsifolis, L. Cahill, A. Roberts, K. Nugent, Y. Iino,
59720Z. (1-6). P. Davis, G. Margaritondo, G. Tromba, et al., Journal of Applied Physics 85
[24] R. Bruck, R. Hainberger, Proceedings of SPIE 7138 (2008) 71380N. (1–7). (1999) 20.
[25] K. Zinoviev, L.G. Carrascosa, J. Sánchez del Río, B. Sepúlveda, C. Domínguez, [38] A.W. Snyder, J. Love, Optical Waveguide Theory, Springer, 1983.
L.M. Lechuga, Advances in Optical Technologies (2008). [39] M. Feit, J. Fleck Jr, et al., Applied Optics 18 (1979) 2843.
[26] B. Sepulveda, J.S. Del Rio, M. Moreno, F. Blanco, K. Mayora, C. Domínguez, [40] M. Feit, J. Fleck Jr, et al., Applied Optics 19 (1980) 1154.
L.M. Lechuga, Journal of Optics A 8 (2006) S561. [41] F.W. Olver, D.W. Lozier, R.F. Boisvert, C.W. Clark, NIST Handbook of Mathematical
[27] P. Nellen, K. Tiefenthaler, W. Lukosz, Sensors and Actuators A 15 (1988) 285. Functions, Cambridge University Press, 2010.
[28] A. Sloper, J. Deacon, M. Flanagan, Sensors and Actuators B: Chemical 1 (1990) 589. [42] K. Tiefenthaler, W. Lukosz, Journal of the Optical Society of America B 6 (1989) 209.
[29] W. Lukosz, Biosensors and Bioelectronics 6 (1991) 215. [43] D. Jones, Biomedical Sensors, Momentum Pr, 2010.
[30] P. Nellen, W. Lukosz, Biosensors and Bioelectronics 8 (1993) 129. [44] K.R. Williams, K. Gupta, M. Wasilik, Journal of Microelectromechanical
[31] R. Heideman, P. Lambeck, Sensors and Actuators B: Chemical 61 (1999) 100. Systems 12 (2003) 761.
[32] B. Pal, Fundamentals of Fibre Optics in Telecommunication and Sensor [45] N. Grote, H. Venghaus, Fibre Optic Communication Devices, Springer, 2001.
Systems, Bohem Press, 1992.