Tunable Large Free Spectral Range Microring Resonators in Lithium Niobate On Insulator
Tunable Large Free Spectral Range Microring Resonators in Lithium Niobate On Insulator
Tunable Large Free Spectral Range Microring Resonators in Lithium Niobate On Insulator
70 µm-radius ring. This work will enable efficient on-chip filtering in LNOI and precede future, more
complex, microring resonator networks and nonlinear field enhancement applications.
fabrication process, etching down to 300 nm trenches in waveguide cross-section, obtained via focused ion beam
LNOI, critical for advanced photonic components. We (FIB) slicing and scanning electron microscopy (SEM),
further report 3 pm/V electro-optic tuning of a 70 µm- shows a sidewall angle of 75◦ and an etch depth of 350 nm
radius microring resonator—to the authors’ knowledge, Fig. 1(c). Finally, the waveguide facets were diced using
this is the largest to date in Z-cut LNOI. We expect optical grade dicing to facilitate butt-coupling. The length
the microring resonators in this work to pave the way of the chip, after all processing steps were completed, is 6
towards on-chip filtering in LNOI with ring networks, as mm.
well as field enhancement applications such as switching
and nonlinear photon generation.
III. EXPERIMENTAL RESULTS
II. DESIGN AND FABRICATION In order to confirm that the photonic components are
not limited by the propagation loss, loss measurements
Microring resonators with radii 30–90 µm were designed were performed prior to the characterization of the micror-
to obtain an FSR from 1.5 to 5.7 nm and were simu- ings, using the Fabry-Perot loss measurement technique
lated using the commercially available software, Lumer- [34]. Laser light at 1550 nm wavelength is coupled into
ical. Rings of varying radii were fabricated to analyze and out of the polished facets of the waveguide using
the FSR and performance for the TE and TM modes. polarization maintaining (PM) lensed fibers with a mode
The small bending loss needed for good operation of a field diameter of 2 µm . A typical optical transmission
30 µm microring resonator required high-index contrast spectrum for TM (the TE and TM modes have a similar
single mode waveguides at 1550 nm. A mode solver was response) is shown on the Fig. 2 (a). Linear inverse tapers
used to determine the dimensions required to ensure a 200 µm long down to 200 nm width, at the waveguide
sufficiently small TM polarization bend radius. The de- ends, are used to improve the mode matching between
sign of the waveguide includes the following parameters: the lensed fibre and the waveguide [35], and improve the
rib height, top width, sidewall angle, refractive indices signal to noise ratio of the Fabry-Perot measurements. As
of the waveguide and claddings, and film thickness. The the waveguide narrows, the mode field diameter at the
cross-section of a Z-cut rib waveguide cladded with SiO2 input and output of the waveguide significantly increases,
is shown in Fig. 1(b). The small gap of 300 nm was allowing improved mode matching with the mode of the
chosen to heavily overcouple the rings and obtain wide lensed fiber. The total input and output coupling and
bandwidth resonances, rather than extremely narrow res- propagation loss is 8 dB for a 6 mm long chip, compared
onances that require very precise wavelength tuning to to the 15 dB loss achieved with the straight waveguide
access. A simulation of the Q-factor as a function of without tapering section. The estimated propagation loss
coupling region gap for the TM mode at 1550 nm of a 30 is less than 0.5 dB/cm for both the TE and TM modes,
µm microring resonator is shown in Fig. 1(a), and indi- which is in agreement with the results obtained in our
cates that the 3 dB bandwidth of the microring resonances previous work [24]
(FWHM) is FWHM = λres /Q =∼155 pm, where λres is The fabricated microring resonators were characterized
the wavelength of the resonance and Q is the Q-factor. by sweeping the wavelength of the laser between 1530
The simulation was performed using Lumerical Mode; the to 1610 nm and recording their spectral responses with
measured losses, as per the Fabry-Perot measurements a commercially available high-speed InGaAs photodiode.
presented in Fig. 2(a), were taken into account in the The laser light was injected into and out of a 6 mm bus-
simulation model. waveguide via PM lensed fibers. To decrease the chance
The photonic components were fabricated by the pro- of interference between multiple oscillations inside of the
cess developed and described in our previous work [24]. photonic component, the inverse tapering section was not
The process starts with 500 nm thick LN film, which is implemented for the microrings—this led to a drop in the
fabricated using the smart-cut technique on 2 µm of SiO2 mode matching efficiency. We observe that both TE and
layer and supported by a 500 µm LN substrate. The next TM (Fig. 2(b)) modes reaches the largest FSR for the ring
fabrication steps rely on electron beam lithography and with the smallest radius 30 µm ; however, the TE and TM
lift-off of the e-beam evaporated metal layer to obtain a modes show different results in terms of the achievable
hard mask defining the photonic components. The scan- Q-factor for this geometry. A Q-factor of ∼ 9000 was
ning electron microscopy image (SEM) of a waveguide achieved for the TM mode whilst for the TE mode the Q-
to a ring coupling region just after the metal lift-off pro- factor is significantly smaller ∼ 1200. As the radius of the
cess, is shown in Fig. 1(e). The components were then ring increases, the Q-factor for TM mode remains almost
dry etched in a reactive ion etcher Fig. 1(d). Following unchanged (Fig. 2(b)); meanwhile, for the TE mode, it
etching, the waveguides were cladded with 3 µm thick significantly increases (Fig. 2(b)) and the highest Q-factor
plasma-enhanced chemical vapor deposition (PECVD) has been achieved for the ring with radius of 90 µm Fig.
SiO2 . The presented structures were etched deeper than 2(d). This dissimilarity can be attributed to the difference
in our previous work to achieve the necessary index con- in the bending loss between both modes. It was deduced
trast, reducing the waveguide bending radius. The rib by using our numerical model (Fig. 1(a)) and the value
3
FIG. 1. (a) Simulation of Q-factor as a function of the coupling region gap for a 30 µm radius microring—the light red dashed
lines demarcate the simulated Q-factor (∼10000) for the microring shown in (d). (b) Design of a single mode LNOI rib waveguide
at 1550 nm wavelength; the top width is 650 nm, the bottom width is 840 nm and the waveguide height is 350 nm. Scanning
electron microscope pictures: (c) cross-section taken by FIB slicing and SEM imaging; (d) an etched 30 µm radius ring with a
300 nm gap between the bus waveguide and the ring prior to PECVD SiO2 cladding; (e) false-color image of the coupling region
after the lift-off process and prior to etching; the false-red highlights the metal etch mask.
of intrinsic quality factor that TM mode bending loss ∼ 0.272, whilst the TE mode is lower ∼ 0.247. As the
for microring resonator of 30 µm of radius is around 1.5 TM mode has a higher index contrast, a smaller bend
dB/cm, while for TE mode it estimated to be around 12 radius is achieved, enabling smaller microring resonators
dB/cm for the ring with the same radius. By comparing to be realized. The TE mode bending loss decreased
theoretical and experimental results, the effective index with increasing microring resonator radius, leading to an
for the TE mode is 1.85 and for the TM mode is 1.72 and improvement in the Q-factor.
the TM mode was confined to have an index contrast of
The group indices for the TE and TM modes respec-
4
FIG. 2. (a) Transmission spectrum of the inverse taper with 200 nm width LNOI used for calculating the propagation loss of
the TM mode.(b) Measured Q-factors of the ring resonators versus their radius for the TE and TM modes. The blue curve
corresponds to the TE mode and the red curve corresponds to the TM mode. (c) Spectral response of the ring with radius 30
µm for TM mode. (d) Spectral response of the ring with radius 90 µm for TE mode.
tively, nTE
g and nTM
g , are deduced from the fully-vectorial and Fig. 3(f), show the measured power distribution at a
mode solver using the Sellmeier equations for lithium wavelength of 1550 nm in a 3 × 3 µm window; each cell
niobate: nTMg = 2.33 and nTEg = 2.38. The FSR can be defined by the white grid lines represents a single pixel (a
calculated using FSR = λ2 /(ng L), where L is the circum- single power measurement). The measured power distri-
ference of the ring (L = 2πR), R is the radius of the ring. bution is performed by sweeping the fiber over the output
The simulation curve is plotted with the measured FSR facet of the waveguide, resulting in a convolution between
for different microring resonator dimensions in Fig. 3(a) the fiber mode and the waveguide mode, smearing and
and Fig. 3(d). The simulated E-field distributions of the enlarging the appearance of the waveguide mode.
fundamental waveguide modes at a wavelength of 1550
nm (found using an in-house mode solver) are included to
the figures as insets: Fig. 3(b) for the TE mode, and Fig.
3(e) for the TM mode. Also included as insets, Fig. 3(c)
5
FIG. 3. (a) Measured FSR as the function of a microring resonator radius for the TE mode and (d) for TM mode; the blue
circles are measured values, whilst red line is theoretically predicted dependence of FSR on microring resonator radius; (b) the
simulated electrical field distribution for the TE mode and (e) for the TM mode; (c) measured optical power distribution at
the output of the chip for the TE mode and (f) for the TM mode, where black lines schematically show the actual waveguide
dimensions.
IV. ELECTRO-OPTIC RESONANT ring resonators and were fabricated on the same chip, the
WAVELENGTH TUNING propagation loss in the rings are concluded to be equally
low loss.
An electrode consisting of Cr (20 nm) and Al (500 nm) is The demonstrated ring resonators are designed to be
deposited directly on the upper cladding of the waveguide. strongly overcoupled, increasing their 3 dB resonance
The separation between the electrode and the waveguide bandwidth (and, conversely, reducing their Q-factor). A
is designed to be 3 µm, which is estimated to be close 300 nm gap in the bus waveguide to microring coupling
enough that the electric field extending from the electrode region provides strong overcoupling. The potential of
can effectively influence the LNOI waveguide, but far the nanofabrication process used in this work [24] could
enough that the optical loss is not increased. Figure 4(a) be further extended to photonic components including
shows the simulation result of the static electric potential grating couplers and compact directional couplers.
performed using a finite element solver, with the voltage The Q-factor measurements show that it is possible to
applied across the top and bottom electrodes. The bottom achieve small and high performance microring resonators
electrode, serving as a ground plane, is made from Cr (10 for the TM mode—critical for electro-optic and nonlinear
nm), Au (100nm) and Cr (10 nm). applications. Meanwhile, the TE mode bending losses
To demonstrate the electro-optic tuning of the device significantly limit the Q-factor of the smaller radius mi-
we apply a DC voltage from 0 V down to -55 V to the top croring resonators; however, increasing the ring radius
electrode of the ring resonator with a radius of 70 µm. leads to a substantial increase in Q-factor. It demon-
The resonance shifts with the applied voltage as shown strated that the TM mode can achieve a smaller bend
in Fig 4(b) corresponding to a EO tunability of 3 pm/V. losses than the TE mode, as the index contrast of the TE
fundamental mode is less than that of the TM fundamen-
tal mode, as verified by both our in-house mode solver,
V. DISCUSSION and by the Q-factor simulations conducted in Lumerical
Mode for the 30 µm ring.
The Fabry-Perot transmission measurements were con- It was found that the theoretically predicted results
ducted on straight waveguides with inverse tapers at both for the microring resonators demonstrated in this paper
ends and indicate low propagation loss for this platform. are in a good agreement with the experimental results
The overall insertion loss of the waveguides is dominated (Fig. 3). The deviation for ng is less than 2% leading to
by mode-mismatch between waveguide and optical fiber, precise agreement between the designed and measured
despite the significant improvement of provided the in- FSRs for different ring geometries. Using 350 nm deep
verse tapers. Given that the straight waveguides measured ribs, a small TM bend radius was achieved to enable 30
have identical dimensions to the waveguides used in the µm TM microring resonators with an FSR of 5.7 nm.
6
FIG. 4. (a) Simulation results of electrical field. b) Spectrum of the optical resonances when voltage from -55 to 0 V is applied.
This result is competitive with other high-index contrast advanced fabrication enables minimal separation (300 nm)
leading platforms, such as SiN and AlN, between monolithically defined adjacent features, whilst
For comparison, we report in Table I a summary of maintaining smooth waveguide sidewalls. We have
experimental results on resonant wavelength tuning. It verified that the optical characteristics of the fabricated
can be seen the tunable ring resonators have been realized microring resonators correspond well with the design
in a multitude of photonic platforms. Silicon has reported and simulation. We have further demonstrated 3pm/V
very high EO tuning with large FSR [27]. More recent EO tuning of a 70 µm radius microring. These results
work has shown good performance in hybrid Si on LN, will precede more complex photonic devices in LNOI,
although it requires extra fabrication steps [28, 29]. While ranging from precise filtering with multistage microring
using AlN has so far resulted in limited tunability [30], resonators to electro-optically tunable devices.
LNOI photonics presents a promising approach to tun-
able ring resonators [31–33]. The results presented in this
Funding
work combine good EO tunability with simple fabrication
Australian Research Council Centre for Quantum Com-
process of Z-cut LNOI single mode waveguides, which are
putation and Communication Technology CE170100012;
readily compatible with other single mode photonic com-
Australian Research Council Discovery Early Career Re-
ponents and and will enable future low-loss and tunable
searcher Award, Project No. DE140101700; RMIT Uni-
filtering in LNOI.
versity Vice-Chancellors Senior Research Fellowship.
Acknowledgments
We thank Jochen Schröder for discussions. This work was
performed in part at the Melbourne Centre for Nanofab-
VI. CONCLUSION rication in the Victorian Node of the Australian National
Fabrication Facility (ANFF) and the Nanolab at Swin-
We have analyzed in detail the performance of large burne University of Technology. The authors acknowledge
FSR microring resonators in Z-cut LNOI, fabricating the facilities, and the scientific and technical assistance,
rings of varying radii and reporting their characterization of the Australian Microscopy & Microanalysis Research
for TE and TM polarizations. The demonstrated Facility at RMIT University.
20, 12261 (2012). [24] I. Krasnokutska, J.-L. J. Tambasco, X. Li, and A. Peruzzo,
[11] W. H. P. Pernice, C. Xiong, C. Schuck, and H. X. Opt. Express 26, 897 (2018).
Tang, Applied Physics Letters 100, 223501 (2012), [25] C. Wang, M. Zhang, B. Stern, M. Lipson, and M. Lončar,
https://doi.org/10.1063/1.4722941. (2017), 1701.06470.
[12] P. P. Absil, J. V. Hryniewicz, B. E. Little, P. S. Cho, R. A. [26] H. Liang, R. Luo, Y. He, H. Jiang, and Q. Lin, Optica
Wilson, L. G. Joneckis, and P.-T. Ho, Opt. Lett. 25, 554 4, 1251 (2017).
(2000). [27] G. Li, X. Zheng, J. Yao, H. Thacker, I. Shubin, Y. Luo,
[13] T. A. Ibrahim, V. Van, and P.-T. Ho, Opt. Lett. 27, 803 K. Raj, J. E. Cunningham, and A. V. Krishnamoorthy,
(2002). Opt. Express 19, 20435 (2011).
[14] C. Ciminelli, F. Dell’Olio, M. N. Armenise, F. M. Soares, [28] L. Chen, M. G. Wood, and R. M. Reano, Opt. Express
and W. Passenberg, Opt. Express 21, 556 (2013). 21, 27003 (2013).
[15] C.-Y. Chao, W. Fung, and L. J. Guo, IEEE Journal of [29] L. Chen, Q. Xu, M. G. Wood, and R. M. Reano, Optica
Selected Topics in Quantum Electronics 12, 134 (2006). 1, 112 (2014).
[16] H. Jung, C. Xiong, K. Y. Fong, X. Zhang, and H. X. [30] H. Jung, K. Y. Fong, C. Xiong, and H. X. Tang, Opt.
Tang, Opt. Lett. 38, 2810 (2013). Lett. 39, 84 (2014).
[17] H.-T. Kim and M. Yu, Opt. Express 24, 9501 (2016). [31] C. Wang, M. Zhang, B. Stern, M. Lipson, and M. Lončar,
[18] D. Janner, D. Tulli, M. Garcı́a-Granda, M. Belmonte, and Opt. Express 26, 1547 (2018).
V. Pruneri, Laser & Photonics Reviews 3, 301 (2009). [32] A. Guarino, G. Poberaj, D. Rezzonico, R. Degl’Innocenti,
[19] J. L. Jackel, C. E. Rice, and J. J. Veselka, Applied Physics and P. Günter, Nature Photonics 1, 407 (2007).
Letters 41, 607 (1982), https://doi.org/10.1063/1.93615. [33] S. Y. Siew, S. S. Saha, M. Tsang, and A. J. Danner, Ieee
[20] O. Stepanenko, E. Quillier, H. Tronche, P. Baldi, and Photonics Technology Letters 28, 573 (2015).
M. D. Micheli, J. Lightwave Technol. 34, 2206 (2016). [34] R. Regener and W. Sohler, Applied Physics B 36, 143
[21] G. Poberaj, H. Hu, W. Sohler, and P. Günter, Laser & (1985).
Photonics Reviews 6, 488 (2012). [35] J. Cardenas, C. B. Poitras, K. Luke, L. Luo, P. A. Morton,
[22] S. Y. Siew, S. S. Saha, M. Tsang, and A. J. Danner, and M. Lipson, IEEE Photonics Technology Letters 26,
IEEE Photonics Technology Letters 28, 573 (2016). 2380 (2014).
[23] M. Zhang, C. Wang, R. Cheng, A. Shams-Ansari, and
M. Lončar, Optica 4, 1536 (2017).