IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 24, NO.

4, JULY/AUGUST 2018 6101508

Characterization of Low Loss Waveguides

Using Bragg Gratings
Yi-Wen Hu , Yang Zhang, Pradip Gatkine , Joss Bland-Hawthorn,
Sylvain Veilleux, and Mario Dagenais , Fellow, IEEE

Abstract—A novel approach is developed for measuring small than 0.1 dB/m were fabricated by wafer bonding a thermal ox-
losses in highly transparent Si3 N4 /SiO2 waveguides on a silicon ide layer as the upper cladding and resulted in a waveguide
chip. The approach is particularly applicable to waveguides written with ultra-low confinement and reduced scattering losses at the
by high-resolution patterning techniques, such as e-beam lithog-
raphy, whose lengths cannot be easily increased beyond several core-cladding interfaces [3]. Ring resonators with high quality
centimeters. This method is based on measuring the transmission factor Q > 106 have also been fabricated in low loss waveg-
of an optical cavity formed by two highly reflective (R at least 0.999) uides and often have been used to extract the loss coefficient in
simple Bragg gratings and a uniform waveguide between the two these waveguides [4], [11]–[13]. Applications of ultra-low loss
gratings whose length can be varied to increase the loss fitting ac- waveguides are envisioned in optical gyroscopes [14]–[16], dis-
curacy. A theoretical model based on an ABCD matrix method is
developed and used for the final loss value fitting. Experimentally, persion compensation [17], in packet-switched networks [18],
a cavity with extinction ratio over –70 dB and quality factor Q optical filters [19]–[22], optomechanical sensing [4], [11], [23],
= 1.02×106 is realized. The fitting results show a waveguide loss [24] and in astrophotonics [25]–[30].
of 0.24 ± 0.01 dB/cm and a grating loss of 0.31 ± 0.01 dB/cm. The major sources of loss in waveguides are material ab-
These results are obtained with relatively high index contrast (Δn sorption, bending loss, and light scattering. An efficient and
> 0.001) gratings with 0.1-pm wavelength scanning resolution.
It is expected that with better design and wavelength scanning reliable method to measure these losses is critical for the design
technique, this approach is applicable more generally to measure of many photonic integration applications. In the past decades,
waveguide loss coefficients as low as 0.001 dB/cm. the most used loss measurement techniques are based on ei-
Index Terms—Optical planar waveguides, gratings, waveguide ther length variation method (fabricating waveguides of differ-
filters, optical losses. ent lengths) or ring resonator method (deriving waveguide loss
from its relationship with cavity Q). In this work, we have devel-
oped a new approach for characterizing small losses in highly
I. INTRODUCTION transparent Si3 N4 /SiO2 waveguides on silicon. It relies on mea-
OW loss dielectric waveguides are critical for many ap- suring the transmission of an on-chip Fabry-Perot (FP) cavity
L plications in integrated photonics. Several groups have
demonstrated low propagation loss waveguides [1]–[10]. In fact,
formed by two Bragg gratings and a straight waveguide between
them. Indeed, additional scattering loss will be introduced by
losses as small as 0.1 dB/m and 0.9 dB/m have now been re- these Bragg gratings. However, with a proper design, it does
ported in weakly and strongly confined waveguides, respectively not compromise loss measurement accuracy (see Section V-A).
[3], [5], [11]. Ultra-low losses in long waveguides (> 1 m) have Compared with the length variation method, this approach is
been measured using a coherent optical frequency domain re- particularly applicable to waveguides written by EBL whose
flectometry technique [3], [5]. The waveguides with losses less lengths cannot be easily extended beyond several centimeters.
Compared with the ring resonator method, this approach uses a
Manuscript received November 28, 2017; revised April 3, 2018; accepted FP cavity which is formed on a straight structure. Therefore it
April 3, 2018. Date of publication April 16, 2018; date of current version May
18, 2018. This work was supported in part by W. M. Keck Foundation, and in part does not involve bending loss, and can operate equally well for
by the National Science Foundation under Grant AST1711377. (Corresponding optical modes with any confinement factor. Moreover, different
author: Mario Dagenais.) with other loss measurement methods based on FP cavity [31],
Y.-W. Hu, Y. Zhang, and M. Dagenais are with the Department of Electrical
and Computer Engineering, University of Maryland, College Park, MD 20742 [33], we use highly reflective (R > 0.999) integrated gratings as
USA (e-mail:,[email protected]; [email protected]; [email protected]). the cavity mirror and a more precise numerical fitting procedure,
P. Gatkine is with the Department of Astronomy, University of Maryland, which gives much higher loss measurement accuracy (around
College Park, MD 20742 USA (e-mail:,[email protected]).
S. Veilleux is with the Department of Astronomy, University of Maryland, 0.01 dB/cm in our experiment).
College Park, MD 20742 USA, and also with the Joint Space-Science Institute, The paper is organized as follows: Section II discusses wave-
University of Maryland, College Park, MD 20742 USA (e-mail:, veilleux@ guide design and fabrication. In Section III, the physics model
astro.umd.edu).
J. Bland-Hawthorn is with the Sydney Institute for Astronomy, School of and theoretical fitting process are introduced. In Section IV, we
Physics, University of Sydney, Sydney, NSW 2006, Australia (e-mail:, jbh@ first measure waveguide propagation loss by analyzing several
physics.usyd.edu.au). Si3 N4 /SiO2 waveguides of different lengths. After recognizing
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. the limitation of this approach, we characterize the waveguide
Digital Object Identifier 10.1109/JSTQE.2018.2827663 loss again using our new approach. In Section V, we demonstrate

1077-260X © 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.
6101508 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 24, NO. 4, JULY/AUGUST 2018

caused by unstable writing conditions which introduces scatter-

ing loss. The second category is to minimize horizontal and ver-
tical offset (especially gaps) between e-beam write-fields (WFs)
caused by alignment limitations, which can bring stitching loss.
Much work has been done to obtain ultra-low losses by focus-
ing on reducing loss of the first category [11], [35]. However, for
long waveguide gratings involving several ten to hundred WFs
[25], the second issue can also be extremely critical. The reason
is that it brings not only additional loss, but it also leads to accu-
mulation of phase errors, which could lead to failure of the entire
grating. In our experiment, to handle this issue appropriately, we
first measure the stitching error in SEM and then use a proper
WF zoom factor to minimize any possible gaps between the
WFs. Moreover, we use Raith e-Line’s Height Sensing function
to dynamically control the beam focus point when moving from
one WF to another during exposure. With the above two meth-
ods, we can achieve a decent loss level around 0.31 dB/cm for
the grating and 0.24 dB/cm for the uniform waveguide, which
Fig. 1. (a) Waveguide cross section illustration (dimensions are not to scale). will be discussed thoroughly in Section IV.
(b) TE mode profile at 1550 nm. (c) TM mode profile at 1550 nm.

III. BRAGG GRATING MODELING AND DESIGN

by simulation that millimeter-length structures are enough to
measure losses down to 0.01 dB/cm. A further factor of 10 im- On-chip gratings are important for various photonic applica-
provement (0.001 dB/cm) is achievable with a centimeter-scale tions, such as sensing, filters and lasers. Among different types
length. We summarize our main results in Section VI. of gratings, Bragg gratings is one of the most promising candi-
dates to be generally employed in photonic integrated circuits
(PICs), mostly for its simplicity and compatibility with uniform
II. WAVEGUIDE DESIGN AND FABRICATION
waveguides. On the other hand, when it comes to the ultra-low
A. Waveguide Core Design loss regime, their extra scattering loss poses a challenge for re-
As shown in Fig. 1(a), our waveguide core is formed by 2 μm searchers to make them competitive with their ring resonator
× 100 nm low pressure chemical vapor deposition (LPCVD) counterparts.
Si3 N4 . This core dimension is chosen for getting a reasonably In this work, we aim at using Bragg gratings to implement
confined mode with acceptable propagation loss. In general, high reflectivity, low loss cavities: phase-shifted Bragg gratings
thinner and wider core sizes are preferred for achieving lower (PSBG) and Bragg grating Fabry-Perot cavities (BGFP). Fur-
loss waveguides [8]. However, those modes tend to have a small thermore, we demonstrate how to use them as a characterization
confinement factor, especially in the vertical direction, which tool for low loss waveguides. In this section, we will first focus
makes them less attractive in applications requiring bends of on their modeling and design.
small radii of curvature. For example, in arrayed waveguide Our theoretical model is based on the ABCD matrix method
gratings (AWGs), large bending radii of arrayed waveguides which is convenient to use for studying the transmission of op-
will make the device footprint increase appreciably [27], [34]. tical waves in periodic structures. We will briefly discuss it, and
Therefore, we choose 100 nm as a trade-off value for the a detailed theory can be found in [36]. What we consider here
Si3 N4 thickness. Additionally, 2 μm width is selected to ensure are two Bragg gratings with a physical cavity length L0 in the
operation in single mode at a wavelength of 1.4 μm and above. center, as illustrated in Fig. 2(a). In the design, the grating on
each side consists of N pairs of alternating width (w1 and w2 )
waveguide segment with period Λ = λc / 2neff . λc is the stop-
B. Fabrication band center wavelength. Starting from basic electromagnetic
In this paper, we use 10 µm thick thermal oxide to effectively equations, we can derive a matrix equation relating the field
reduce substrate leakage to the silicon substrate. The waveguides amplitude at two neighboring grating periods,
are patterned by 20kV Raith e-Line e-beam system with PMMA      
resist. After development, we use a liftoff process to form a an −1 A B an
reverse Chromium (Cr) mask. The mask is used for etching down = ×
bn −1 C D bn
Si3 N4 by Inductively Coupled Plasma (ICP) Etching. Lastly,
we remove the Cr and deposit 4 μm plasma-enhanced chemical
vapor deposition (PECVD) oxide layer as the upper cladding. Here an and bn are the electric field amplitude in the nth
We use EBL to write the full structure, which can be 1-3 cm grating period for a right-traveling and a left-traveling plane
long, the challenge of which can be classified into two cate- wave. Note that the ABCD coefficients are determined by the
gories. The first category is to reduce the sidewall roughness electrical boundary conditions. For the uniform cavity part,
HU et al.: CHARACTERIZATION OF LOW LOSS WAVEGUIDES USING BRAGG GRATINGS 6101508

Fig. 3. Illustration of the bending waveguide design for measuring linear

propagation loss. We made five waveguides on the chip with length difference
of 0, 7, 14, 21, and 28 mm. The corresponding loss results are shown in Fig. 4.
Fig. 2. (a) Illustration of the Bragg grating design. For PSBG, cavity length
L 0 = Λ ∼ 550 nm. For BGFP, we made the cavity length L 0 = 1 mm, 3 mm a result, BGFP can have higher Q (Q ∼ ωτ , τ is the photon
and 6 mm. (b) Tilted view of the 2 μm ×100 nm straight waveguide. (c) Top
view of the λ/4 PSBG; the π phase shifted part can be recognized in the figure lifetime) than PSBG. This is confirmed experimentally in the
center. next section.

A = eik 1 z L 0 /2 , B = C = 0, D = e−ik 1 z L 0 /2 . Here k1z is the C. Loss Terms in the Model

propagation constant in width w1 waveguide segment. Transmission matrix terms, A, B, C, D depends on 3 vari-
The coupling coefficient κ is an important concept in coupled ables, k1z , k2z (propagation constant in width w2 waveguide
mode theory, which is quite useful for grating design [37]. For segment), and the grating period Λ. To take the loss terms into
a 1-d grating discussed here, if we denote the larger index as n1 consideration, we only need to add an imaginary part to k1z and
and smaller index as n2 , κ can be roughly estimated as κ = π/λ k2z . For example,
(n1 − n2 ), and the stopband width is [34]
2π 2π
λ2 |κ| k1z = n1 × − inim g × (2)
Δλ ∼ (1) λ λ
πnef f
Note that α = 4πnimg /λ, and we have two loss terms, the
For the transmission calculation, we still need to know the grating loss αg and the cavity loss αc (for PSBG, αc = αg ).
effective indices for waveguide width w1 and w2 through- By fitting our experimental and theoretical curve with different
out the studied spectral range, which are obtained using the lengths cavities, we can extract both αg and αc .
FIMMWAVE full-vectorial mode solver. Note that in the final
fitting of the transmission curve, we still need to slightly adjust IV. EXPERIMENTAL RESULTS
the index profile to account for the rounded edges of the grating
To characterize the transmission spectrum, we use a setup
structure [see Fig. 2(c)].
composed of a Keysight 81600B tunable laser and a Keysight
N7744A power meter. The tunable laser triggers the power meter
A. π Phase Shifted Bragg Grating during the wavelength scanning for fast spectral measurement
First we design a π (i.e., λ/4) PSBG (L0 = Λ) to get a narrow with 0.1 pm resolution. Polarization maintaining fibers are used
peak at the stopband center. The peak’s linewidth and intensity for optical mode launch and collection. Fiber rotators are also
are very sensitive to propagation loss, so we can use it to measure employed for the TE/TM control. The fiber input and output
the grating loss αg . To get high enough measurement resolution coupling are accomplished by two Newport VP-25XL XYZR
for state-of-the-art low loss waveguides, we want to design a motorized stages with 10 nm minimum incremental motion.
PSBG with as narrow linewidth as possible, which requires
a large κLg for high reflection, where Lg = N Λ is the grat- A. Uniform Waveguide
ing length. After careful theoretical and empirical studies, we
We first measure the waveguide loss with uniform waveguides
choose w1 = 2 μm, w2 = 2.15 μm, corresponding to a moder-
of different lengths. Two broadband sources centered at 1310
ate κ ∼ 10 cm−1 , which is a compromise for low scattering loss
and 1550 nm are deployed to characterize the loss from 1200
and decent grating coupling efficiency. As R ∼ tanh2 (κLg ), in
to 1630 nm. The layout design is shown in Fig. 3, the length
the layout design, we choose Lg to be around 3.5 mm to make
difference among them is 0, 7, 14, 21, and 28 mm. The extracted
κLg large enough to get single side mirror reflection R > 0.999.
loss results are shown in Fig. 4.
The lowest loss occurs at the longer wavelength of the spec-
B. Bragg Grating Fabry-Perot cavity trum in Fig. 4. Specifically, at 1630 nm, the TE loss is 0.16
The Q of PSBG is dominated by grating loss αg . In order to ± 0.18 dB/cm, and the TM loss is 0.32 ± 0.13 dB/cm. How-
measure uniform waveguide loss, we keep the same reflecting ever, when we move to a shorter wavelength like 1558 nm, TE
gratings and make L0 much larger to form a Fabry-Perot cavity. loss is 0.68 ± 0.25 dB/cm, and TM loss is 0.79 ± 0.20 dB/cm.
Since the cavity loss is always smaller than the grating loss, Two strong absorption peaks can be observed around 1390 nm
photons will stay inside for longer time before leaking out. As and 1505 nm. Previous studies have noticed them in Si3 N4 /SiO2
6101508 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 24, NO. 4, JULY/AUGUST 2018

Fig. 4. Linear propagation loss of Si3 N4 /SiO2 waveguide before annealing,

extracted from the sample in Fig. 3. The error bars are shown as the shaded area.

Fig. 5. TE propagation loss after annealing. The measurement is done with Fig. 6. PSBG grating experimental (black solid line) and theoretical fit-
the same sample in Fig. 4, after 2 hour annealing in 1150C. ting (dashed lines) results. In (a) and (c), the wavelength range is 4 nm.
In (b) and (d), the wavelength range is 20 pm. Fitting parameters: (1558
nm) n 1 = 1.47584, n 2 = 1.47691; Δn = 0.00107, κ = 10.81 cm−1 , stop-
platforms [38], [39], and they are generally attributed to O-H, band width Δλ = 0.563 nm. (1629 nm) n 1 = 1.47171, n 2 = 1.47273; Δn =
0.00102, κ = 9.85 cm−1 , Δλ = 0.561 nm.
N-H, Si-H bonds in the two materials. These peaks can be largely
removed by annealing at high temperature, as shown in Fig. 5. TABLE I
It can also be seen that the measurement error increased after PSBG (CENTERED AT 1629 NM) FINE FITTING RESULTS FOR FIG. 6
annealing. A possible reason is that the annealing process in-
troduced larger coupling efficiency fluctuation among different
waveguides. This is also a main disadvantage of this loss mea-
surement method, whose measurement error depends not only
on the waveguide fabrication stability, but also on the coupling
efficiency stability.
a
B. π Phase Shifted Bragg Grating The center peak intensity is the maximum transmission the center peak can reach. The
center peak intensity will be exactly 0 dB if both the grating loss α g and cavity loss α c
As we can see, the previous method shows a measurement are 0.
b
The linewidth measured in experiment is 2.2 pm.
error of at least 0.1 dB/cm, which is undesirable for current low
loss waveguide. Now we used two PSBG to demonstrate the plotted another two fitting curves with ± 0.1 dB/cm in Fig. 6,
loss fitting at two different wavelengths, 1558 and 1629 nm, and the fitting accuracy can actually reach about ± 0.01 dB/cm with
compare with the loss obtain in IV. A. 0.1 pm wavelength resolution, as shown in Table I.
The fitting process is as follows: first, we scale and shift the in- For the 1629 nm PSBG, the narrowest linewidth we can get
dex profile to fit the stopband position and width; next we adjust is 2.2 pm, corresponding to a Q ∼ 7.4 × 105 . The extinction
grating loss αg to find the best fitting curve, as shown in Fig. 6. ratio (ER) of the 1558 nm grating is 50 dB, and that of the
The fitting parameters are also shown in the caption. An impor- 1629 nm grating is 60 dB (limited to 50 dB in measurement by
tant criteria is that we must use the same wavelength dependent the uncoupled fiber input light). It implies a single side mirror
effective index profile for the fitting curves. The 1558 nm PSBG reflection R ∼ 0.999 around 1629 nm.
has a larger index contrast than the 1629 nm PSBG because the
shorter wavelength mode is more confined and more sensitive
C. Bragg Grating Fabry-Perot Cavity
with width variation.
From the theoretical fit, we get grating loss αg = 0.64 dB/cm We can extract the grating loss from the PSBG sample.
at 1558 nm and αg = 0.31 dB/cm at 1629 nm. Although we only However, we still cannot tell the exact straight waveguide
HU et al.: CHARACTERIZATION OF LOW LOSS WAVEGUIDES USING BRAGG GRATINGS 6101508

Fig. 8. The enlarged figure for BGFP transmissions. Red, blue and green lines
Fig. 7. Experimental and simulation curves for Bragg grating cavity with denotes cavity loss α c = 0.14 dB/cm, 0.24 dB/cm and 0.34 dB/cm. Grating loss
length (a) 1 mm, (b) 3 mm and (c) 6 mm. For the blue simulation curve, we α g is set as 0.4 dB/cm. (a) 1 mm cavity with linewidth 2.0 pm. (b) 3 mm cavity
use grating loss 0.41 dB/cm and cavity (straight waveguide) loss 0.24 dB/cm. with linewidth 1.8 pm. (c) 6 mm cavity with linewidth 1.6 pm. Note that the
The wavelength range is 4 nm in all four panels. The y axes are set to be legend in (a) also holds for (b) and (c). The x range are all 8 pm, y axes are set
the same. Fitting parameters: n 1 = 1.47860, n 2 = 1.47979; Δn = 0.00119, to be the same.
κ = 11.5 cm−1 , Δλ = 0.649 nm.

TABLE II
BGFP (CENTERED AT 1625.8 NM) FINE FITTING RESULTS FOR FIG. 8
loss. We fabricated three additional Bragg gratings with cen-
ter length 1 mm, 3 mm and 6 mm. Thus, the transmission
contains the information from both αg and αc . The fitting pro-
cess is similar to that of the PSBG, and the results are shown
in Fig. 7.
In this BGFP sample we kept the grating design parame-
ters the same as for the PSBG sample. However, ER can reach
−70 dB in experiment and −80 dB by fitting. A one-side mirror
reflectivity R ∼ 0.9999 is achieved. There are two reasons that
we get higher ER than in Fig. 6. First, we increase the offset
between the input and the output waveguides to reduce the back-
ground light intensity; secondly, we have higher index contrast
Δn due to a slightly thicker LPCVD Si3 N4 layer, which can be
confirmed in the Δn fitting parameters (0.00119 vs 0.00102). V. FURTHER DISCUSSIONS
In the enlarged plots of Fig. 8, it can be seen clearly that the
longer cavity leads to a narrower linewidth, consistent with our A. Loss Reduction
expectations. The best Q we get is 1.02 × 106 ∼ δλ = 1.6 pm. As shown in Fig. 2(b), there exists some obvious roughness
To our knowledge, this is the best Q obtained with on-chip Bragg along waveguide boundaries, which is our device’s dominant
gratings. From Table II, we can see that the resolution is still loss source. The roughness comes from e-beam lithography and
around ± 0.01 dB/cm. lift-off process, but the latter probably plays a more important
We also studied the effective cavity length Leff [illustrated role. In future work, we plan to use lift-off resist (LOR) or a
in Fig. 2(a)] of BGFP. From Fig. 7, we measured the FSR negative resist [11] to reduce the edge roughness and thereafter
for each cavity. They are 0.481 nm, 0.234 nm and 0.129 nm, the loss.
which corresponds to Leff = 1.86 mm, 3.82 mm and 6.93 mm, Another loss term in our measurement is the scattering loss
respectively. The difference between Leff and L0 is quite stable, due to mode mismatch at the interface between the straight
about 0.8-0.9 mm, which is favorable to design specific FSR waveguide and the grating. This loss will become more influ-
cavities. ential if higher width contrast grating is used. However, it can
6101508 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 24, NO. 4, JULY/AUGUST 2018

TABLE III
12 MM BGFP MODELING RESULTS RELATED WITH FIG. 9

C. Lower κ Grating
Another direction to optimize our approach is to use a lower κ
grating for getting lower αg . Now our grating κ is about 10 cm−1 .
If we decrease it and increase Lg to keep κLg a constant, we
will also see an improvement in the loss measurement limit.
The underlying physics is that for the center cavity, its peak
linewidth δλ depends only on the loss terms αg , αc , Leff and
grating reflection κLg . As long as κ drops and we extend Lg
to keep the mirror reflectivity the same, the cavity Q will be
Fig. 9. BGFP simulation to demonstrate its potential for measuring very low increased. In experiment, a 1/10 factor decrease (κ = 1 cm−1 )
loss coefficients. We keep the grating loss α g = 0.1 dB/cm, and set the cavity is definitely achievable [20]. Further reduced grating κ might be
loss α c = 0, 0.001 and 0.01 dB/cm.
limited by fabrication stability and accuracy.

D. Loss Measurement Bandwidth

be virtually eliminated by reducing the sidewall modulation to
zero at these interfaces. Since we use a simple Bragg grating as a reflective mirror,
there is only one stopband in the wavelength range of interest
and the bandwidth is also quite narrow. To measure losses at
B. Lower Limit of Measurable Waveguide Loss different position, we have to use multiple waveguides or mul-
Considering the fact that people have already achieved losses tiple cavities in a line. However, combined with the complex
down to 0.001 dB/cm similar platforms [11], it is quite important grating technologies we developed for on-chip waveguides [25]
to assess our approach’s lower limit of measurable loss. or fibers [30], the proposed approach is actually quite viable
Since our αg can be 0.31 dB/cm, only 0.07 dB/cm higher than to be used for loss measurement with over 100 nm wavelength
the straight waveguide loss, it is reasonable to assume that our range. This is very attractive for ultra-fast, ultra-broadband loss
current Bragg grating design introduces an extra scattering loss characterization.
< 0.1 dB/cm. Now we can simulate whether our approach is ap-
plicable for the current best waveguide loss level ∼ 0.001 dB/cm
VI. CONCLUSION
[3], [11].
In Fig. 9, we keep αg = 0.1 dB/cm, and set αc as 0, 0.001 We have proposed an alternative method for measuring ultra-
and 0.01 dB/cm. If we look at the green curve, which stands low loss coefficients in a waveguide, which is both accurate
for 0.01 dB/cm loss level, it is obvious that our approach can and efficient. We have also demonstrated on-chip Bragg grating
easily handle it even with cavity length 6 mm. For 0.001 dB/cm cavities with Q = 1.02 × 106 , which as far as we are concerned
loss, however, we may have to use longer cavities to increase the is the highest experimentally observed Bragg grating cavity Q.
sensitivity. In Table III, we exhibited some numerical values ex- The straight waveguide loss measured with this new approach
tracted from Fig. 9. For a 12 mm BGFP with αc = 0.001 dB/cm, is 0.24 ± 0.01 dB/cm. The loss measurement limit and accuracy
although the linewidth only changes by 0.006 pm compared with generally scales with the grating and waveguide loss, making
αc = 0 dB/cm, the center peak intensity can drop more than this approach promising to measure waveguide loss down to
1 dB, which can easily be measured in the experiment without 0.001 dB/cm.
much difficulty. This result implies that we can still characterize
the waveguide loss, even if it is much lower than the grating
loss. The reason is that we can always try to increase the num- ACKNOWLEDGMENT
ber of grating periods to achieve high enough reflection so that The authors would like to thank Dr. T. Zhu and UMD Ph.D.
the cavity transmission becomes adequately sensitive within the student S. Xie for constructive discussions, and also acknowl-
expected waveguide loss range. edge the support of the Maryland NanoCenter and its AIMLab.
HU et al.: CHARACTERIZATION OF LOW LOSS WAVEGUIDES USING BRAGG GRATINGS 6101508

REFERENCES [25] T. Zhu, Y. Hu, P. Gatkine, S. Veilleux, J. Bland-Hawthorn, and M. Da-

genais, “Arbitrary on-chip optical filter using complex waveguide Bragg
[1] R. Adar, M. R. Serbin, and V. Mizrahi, “Less than 1 dB per meter propaga- gratings,” Appl. Phys. Lett., vol. 108, no. 10, Mar. 2016, Art. no. 101104.
tion loss of silica waveguides measured using a ring resonator,” J. Lightw. [26] T. Zhu, Y. Hu, P. Gatkine, S. Veilleux, J. Bland-Hawthorn, and M. Dage-
Technol., vol. 12, no. 8, pp. 1369–1372, Aug. 1994. nais, “Ultrabroadband high coupling efficiency fiber-to-waveguide coupler
[2] B. Little, “A VLSI Photonics Platform,” in Proc. Opt. Fiber Commun. using Si 3N4 /SiO2 waveguides on silicon,” IEEE Photon. J., vol. 8, no. 5,
Conf., 2003, Paper ThD1. pp. 1–12, Oct. 2016.
[3] J. F. Bauters et al., “Planar waveguides with less than 0.1 dB/m propa- [27] P. Gatkine, S. Veilleux, Y. Hu, J. Bland-Hawthorn, and M. Dagenais,
gation loss fabricated with wafer bonding,” Opt. Express, vol. 19, no. 24, “Arrayed waveguide grating spectrometers for astronomical applications:
pp. 24090–24101, Nov. 2011. new results,” Opt. Express, vol. 25, no. 15, pp. 17918–17935, Jul. 2017.
[4] H. Lee et al., “Chemically etched ultrahigh-Q wedge-resonator on a silicon [28] J. Bland-Hawthorn and P. Kern, “Astrophotonics: A new era for astro-
chip,” Nat. Photon., vol. 6, no. 6, pp. 369–373, Jun. 2012. nomical instruments,” Opt. Express, vol. 17, no. 3, pp. 1880–1884, Feb.
[5] H. Lee, T. Chen, J. Li, O. Painter, and K. J. Vahala, “Ultra-low-loss 2009.
optical delay line on a silicon chip,” Nature Commun., vol. 3, May 2012, [29] S. G. Leon-Saval, A. Argyros, and J. Bland-Hawthorn, “Photonic lanterns:
Art. no. 867. A study of light propagation in multimode to single-mode converters,” Opt.
[6] D. J. Moss, R. Morandotti, A. L. Gaeta, and M. Lipson, “New CMOS- Express, vol. 18, no. 8, pp. 8430–8439, Apr. 2010.
compatible platforms based on silicon nitride and Hydex for nonlinear [30] J. Bland-Hawthorn et al., “A complex multi-notch astronomical filter to
optics,” Nature Photon., vol. 7, no. 8, pp. 597–607, Aug. 2013. suppress the bright infrared sky,” Nature Commun., vol. 2, pp. 581–587,
[7] M. J. R. Heck, J. F. Bauters, M. L. Davenport, D. T. Spencer, and J. 2011.
E. Bowers, “Ultra-low loss waveguide platform and its integration with [31] R. G. Walker, “Simple and accurate loss measurement technique for semi-
silicon photonics,” Laser Photon. Rev., vol. 8, no. 5, pp. 667–686, Sep. conductor optical waveguides,” Electron. Lett., vol. 21, no. 13, pp. 581–
2014. 583, Jun. 1985.
[8] D. T. Spencer, J. F. Bauters, M. J. R. Heck, and J. E. Bowers, “Integrated [32] T. Feuchter and C. Thirstrup, “High precision planar waveguide prop-
waveguide coupled Si3N4 resonators in the ultrahigh-Q regime,” Optica, agation loss measurement technique using a Fabry-Perot cavity,” IEEE
vol. 1, no. 3, pp. 153–157, Sep. 2014. Photon. Technol. Lett. vol. 6, no. 10, pp. 1244–1247, Oct. 1994.
[9] K. Wörhoff, R. G. Heideman, A. Leinse, and M. Hoekman, “TriPleX: A [33] A. Grieco, B. Slutsky, and Y. Fainman, “Characterization of waveguide
versatile dielectric photonic platform,” Adv. Opt. Technol., vol. 4, no. 2, loss using distributed Bragg reflectors,” Appl. Phys. B, vol. 114, no. 4,
pp. 189–207, 2015. pp. 467–474, Mar. 2014.
[10] M. Belt, M. L. Davenport, J. E. Bowers, and D. J. Blumenthal, “Ultra-low- [34] W. Bogaerts et al., “Compact wavelength-selective functions in silicon-on-
loss Ta2O5-core/SiO2-clad planar waveguides on Si substrates,” Optica, insulator photonic wires,” IEEE J. Sel. Topics Quantum Electron., vol. 12,
vol. 4, no. 5, pp. 532–536, May 2017. no. 6, pp. 1394–1401, Nov./Dec. 2006.
[11] X. Ji et al., “Ultra-low-loss on-chip resonators with sub-milliwatt para- [35] R. J. Bojko, J. Li, L. He, T. B.-J., M. Hochberg, and Y. Aida, “Electron
metric oscillation threshold,” Optica, vol. 4, no. 6, pp. 619–624, Jun. beam lithography writing strategies for low loss, high confinement silicon
2017. optical waveguides,” J. Vac. Sci. Technol. B, vol. 29, no. 6, Oct. 2011, Art.
[12] Q. Li, A. A. Eftekhar, M. Sodagar, Z. Xia, A. H. Atabaki, and A. Adibi, no. 06F309.
“Vertical integration of high-Q silicon nitride microresonators into silicon- [36] A. Yariv and P. Yeh, “Wave propagation in periodic media,” in Photonics:
on-insulator platform,” Opt. Express, vol. 21, no. 15, pp. 18236–18248, Optical Electronics in Modern Communications, 6th ed. New York, NY,
Jul. 2013. USA: Oxford Univ. Press, 2007, ch. 12.
[13] K. Luke, A. Dutt, C. B. Poitras, and M. Lipson, “Overcoming Si3N4 film [37] J. Bland-Hawthorn, A. Buryak, and K. Kolossovski, “Optimization al-
stress limitations for high quality factor ring resonators,” Opt. Express, gorithm for ultrabroadband multichannel aperiodic fiber Bragg grating
vol. 21, no. 19, pp. 22829–22833, Sep. 2013. filters,” J. Opt. Soc. Amer., vol. 25, no. 1, pp. 153–158, Jan. 2008.
[14] C. Ciminelli, F. Dell’Olio, C. E. Campanella, and M. N. Armenise, “Pho- [38] C. H. Henry, R. F. Kazarinov, H. J. Lee, K. J. Orlowsky, and L. E. Katz,
tonic technologies for angular velocity sensing,” Adv. Opt. Photon., vol. 2, “Low loss Si3N4-SiO2 optical waveguides on Si,” Appl. Opt., vol. 26,
no. 3, pp. 370–404, Sep. 2010. no. 13, pp. 2621–2624, Jul. 1987.
[15] S. Srinivasan, R. Moreira, D. Blumenthal, and J. E. Bowers, “Design [39] M. J. Shaw, J. Guo, G. A. Vawter, S. Habermehl, and C. T. Sullivan,
of integrated hybrid silicon waveguide optical gyroscope,” Opt. Express, “Fabrication techniques for low-loss silicon nitride waveguides,” Proc.
vol. 22, no. 21, pp. 24988–24993, Oct. 2014. SPIE, vol. 5720, pp. 109–118, Jan. 2005.
[16] T. Huffman, M. Davenport, M. Belt, J. E. Bowers, and D. J. Blumenthal,
“Ultra-low loss large area waveguide coils for integrated optical gyro-
scopes,” IEEE Photon. Technol. Lett., vol. 29, no. 2, pp. 185–188, Jan. Yi-Wen Hu received two B.S. degrees in physics/
2017. computer software in 2011, and the M.S. degree in op-
[17] R. Moreira, S. Gundavarapu, and D. J. Blumenthal, “Programmable eye- tics in 2014 from Peking University, Beijing, China.
opener lattice filter for multi-channel dispersion compensation using an in- He is currently working toward the Ph.D. degree in
tegrated compact low-loss silicon nitride platform,” Opt. Express, vol. 24, electrical engineering at the University of Maryland,
no. 15, pp. 16732–16742, Jul. 2016. College Park, MD, USA.
[18] E. F. Burmeister et al., “Photonic integrated circuit optical buffer for From 2011 to 2014, he was a Research Assis-
packet-switched networks,” Opt. Express, vol. 17, no. 8, pp. 6629–6635, tant with Peking University focused on optical mi-
Apr. 2009. crocavities and optomechanical sensing. Since 2014,
[19] T. Huffman, D. Baney, and D. J. Blumenthal, “High extinction ratio widely he has been a Research Assistant with the University
tunable low-loss integrated Si3N4 third-order filter,” arXiv:1708.06344., of Maryland with interests in on-chip optical waveg-
Aug. 2017. uides/gratings/cavities for applications in communications, quantum informa-
[20] D. T. Spencer, M. Davenport, S. Srinivasan, J. Khurgin, P. A. Morton, and tion, and astrophotonics. He was the recipient of Clark School of Engineering
J. E. Bowers, “Low kappa, narrow bandwidth Si3N4 Bragg gratings,” Opt. Distinguished Graduate Fellowship in 2014.
Express, vol. 23, no. 23, pp. 30329–30336, Nov. 2015.
[21] T. Komljenovic and J. E. Bowers, “Monolithically integrated high-Q Rings
for narrow linewidth widely tunable lasers,” IEEE J. Quantum Electron., Yang Zhang received the B.S. degree in electrical
vol. 51, no. 11, pp. 1–10, Nov. 2015. engineering from Fudan University, Shanghai, China,
[22] J. F. Bauters, J. R. Adleman, M. J. R. Heck, and J. E. Bowers, “Design and in 2012. He is currently working toward the Ph.D.
characterization of arrayed waveguide gratings using ultra-low loss Si3N4 degree in electrical engineering at the University of
waveguides,” Appl. Phys. A, vol. 116, no. 2, pp. 427–432, Aug. 2014. Maryland, College Park, MD, USA.
[23] Y.-W. Hu, Y.-F. Xiao, Y.-C. Liu, and Q. Gong, “Optomechanical sensing His research interests include CIGS solar cell,
with on-chip microcavities,” Frontiers Phys., vol. 8, no. 5, pp. 475–490, optical phased array, and silicon nitride photonics.
Oct. 2013.
[24] Y.-W. Hu, B.-B. Li, Y.-X. Liu, Y.-F. Xiao, and Q. Gong, “Hybrid photonic–
plasmonic mode for refractometer and nanoparticle trapping,” Opt. Com-
mun., vol. 291, pp. 380–385, Mar. 2013.
6101508 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 24, NO. 4, JULY/AUGUST 2018

Pradip Gatkine received the B.Tech. degree from the Sylvain Veilleux received the B.S. degree in physics
Indian Institute of Technology Bombay, Mumbai, In- from the Université de Montréal, Montréal, QC,
dia, in 2014, and the M.S. degree in astronomy from Canada, in 1984, and the Ph.D. degree in astronomy
the University of Maryland, College Park, MD, USA, from the University of California, Santa Cruz, CA,
in 2016. He is currently working toward the Ph.D. de- USA, in 1989.
gree in astronomy at the University of Maryland. In 1995, he became an Assistant Professor with
His photonic research focuses on the development the Department of Astronomy, University of Mary-
of next generation of astrophotonic instrumentation land, College Park, MD, USA. He was promoted to
for upcoming large telescopes. His current work in- Associate Professor in 2000 and to Full Professor in
volves the development of moderate resolution as- 2005. Since 2003, he has been the Optical Director at
trophotonic arrayed waveguide grating spectrographs Maryland. He is the author of more than 200 refereed
and studying OH-suppression for astronomical telescopes. On the astrophysics articles. His research interests include the physics of active galactic nuclei and
side, he is interested in studying the early universe, especially using gamma- starburst galaxies, particularly the connection between the supermassive black
ray bursts as probes. He is currently working on studying the circumgalactic hole, the starburst, and the host galaxy, and the importance of feedback on
medium in the early universe. He is a member of the International Society of galaxy evolution and the intracluster medium. He is also the lead on a num-
Optics and Photonics and American Astronomical Society. ber of instrumentation projects for large ground-based optical and near-infrared
telescopes with applications for space missions and beyond.
Prof. Veilleux was an NSERC Fellow with the Institute for Astronomy, Hon-
olulu, HI, USA, from 1989 to 1992, and a Hubble Fellow with the National
Optical Astronomy Observatory, Tucson, AZ, USA, from 1992 to 1995. He is
also a Fellow of the Joint Space-Science Institute. He was the recipient of the
NASA Hubble Fellowship, the National Science Foundation CAREER Award,
the Humboldt Prize, the John Simon Guggenheim Memorial Foundation Fel-
lowship, and the Aspen Institute Italia Award.

Joss Bland-Hawthorn was born and educated in

the U.K. From 1985 to 1988, he was a Postdoctoral
Fellow with the Institute for Astronomy, Honolulu,
HI, USA. He then served as a tenured Professor in Mario Dagenais (F’10) received the Ph.D. degree in
physics with Rice University, Houston, TX, USA, physics from the University of Rochester, Rochester,
before moving to Sydney, NSW, Australia, in 1993. NY, USA, in 1978, working in quantum optics and
From 2000 to 2007, he was the Head of Instrument photon correlations under the direction of Prof. Man-
Science Group, Anglo-Australian Observatory dur- del. Together with J. Kimble, he made the first obser-
ing where he established the fields of astrophotonics vation of photon antibunching. From 1978 to 1980,
and galactic archaeology. With Birks and Leon-Saval, he was a Research Fellow with Harvard University,
he developed the photonic lantern. He has authored where he worked in nonlinear optics with Prof. N.
or coauthored 550 research papers in optics, photonics, physics, astrophysics, Bloembergen. From 1980 to 1987, he worked with
and astronomical instrumentation. During the past decade, he was the recipient GTE Laboratories on photonic switching and semi-
of the prestigious Federation and Laureate Fellowships. He was the recipient conductor lasers. In 1987, he joined the University of
of many awards including the 2011 Jackson Gwilt Medal, the 2015 WH Steel Maryland, College Park, MD, USA, where he has been a Professor in electri-
Medal, the 2017 Thomas Ranken Lyle Medal, and the 2018 Miller Professorship cal and computer engineering since 1991. He has more than 300 archival and
to Berkeley. He is a Fellow of the Australian Academy of Science and the Opti- conference publications. His research interests include Si3 N4 /SiO2 integrated
cal Society of America. He is currently the Director of the Sydney Institute for nanophotonic devices on Si, GaN light sources, and quantum dot GaAs/AlGaAs,
Astronomy and an Assistant Director of the Institute of Photonics and Optical CIGS, and perovskite photovoltaics. He is a Fellow of the Optical Society of
Science, University of Sydney, Sydney, NSW, Australia. America and a Fellow of the Electromagnetic Society.