Open Access
Add to BibSonomy
Abstract
We demonstrate an ultra-high-bandwidth Mach-Zehnder electro-optic modulator (EOM), based on foundry-fabricated silicon (Si) photonics, made using conventional lithography and wafer-scale fabrication, oxide-bonded at 200C to a lithium niobate (LN) thin film. Our design integrates silicon photonics light input/output and optical components, such as directional couplers and low-radius bends. No etching or patterning of the thin film LN is required. This hybrid Si-LN MZM achieves beyond 106 GHz 3-dB electrical modulation bandwidth, the highest of any silicon photonic or lithium niobate (phase) modulator.
© 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Electro-optic modulation, the imprinting of a radio-frequency (RF) waveform on an optical carrier, is one of the most important photonics functions, being crucial for high-bandwidth signal generation, optical switching, waveform shaping, data communications, ultrafast measurements, sampling, timing and ranging, and RF photonics. Although silicon (Si) photonic electro-optic modulators (EOMs) can be fabricated using wafer-scale technology compatible with the semiconductor industry, such devices do not exceed a 3-dB electrical modulation bandwidth of about 50 GHz [1], whereas many applications require higher RF frequencies. Bulk Lithium Niobate (LN) [2] and etched LN modulators [3] can scale to higher bandwidths in the Mach-Zehnder Interferometer (MZI) configuration. But so far, LN modulators have not been integrated into the foundry Si photonics fabrication process popularized over the last decade.
It was shown nearly two decades ago that thin-film lithium niobate can be bonded to silicon after surface activation [4,5] and several reports of thin film LN-based electro-optic modulators have been published [3, 6–16]. However, as Fig. 1 shows, most bulk LN EOMs and thin film LN EOMs have demonstrated limited bandwidth, less than what can now be achieved simply by using Si [1], with the notable exception of the recent work published in [3]. Moreover, some limitations are evident in previous works, such as:
Fig. 1 Progress of lithium niobate (LN) electro-optic modulators (EOM) in terms of 3-dB electrical modulation bandwidths [29], including traditional bulk LN EOMs [2, 30–33] and thin film LN EOMs [3,6–15]. The horizontal dashed black line at 50 GHz represents the 3-dB electrical (3-dBe) modulation bandwidth achievable by an all-Si EOM made in a foundry process [1]. The 3-dBe modulation bandwidth of the Si-LN device in this work is well beyond 106 GHz, the frequency limit of our experimental capabilities. The marker reported by ‘This work’ is the 1.5 dBe modulation bandwidth. An asterik (*) indicates a reference where the 3-dBe modulation bandwidth was not provided and had to be estimated from sideband measurements.
The above-mentioned limitations are not present in our approach. Here, an un-patterned thin-film of LN was simply oxide-bonded at room temperature to the patterned and planarized Si waveguide circuits, with an anneal step at 200°C. In contrast, fabrication of bulk titanium-indiffused LN modulators and doped III–V or Si modulators require at least 600°C and typically 900–1000°C [34,35], limiting them to either standalone or front-end-of-line device fabrication. No patterning or etching of the LN film was performed in our approach. Oxide, rather than polymeric, bonding was used [36, 37]. The traveling wave electrodes were formed using thin deposited aluminum, rather than thick electroplated gold [2]. The demonstrated combination of beyond-100 GHz electrical (3-dB) modulation bandwidth and a design / fabrication approach and materials that are compatible with, and leverage, wafer-scale and foundry-based approaches popularized by Si photonics is a significant advancement from both a performance and a practical (e.g., fabrication, cost, scalability) perspective.
2. Hybrid LN/Si electro-optic Mach-Zehnder modulator
Mach-Zehnder modulators (MZM) consist of light being split equally into two pathways, each experiencing a voltage-driven optical phase shift, typically between 0 and π/2 radians in each arm with opposing signs, and recombining to result in controllable optical transmission. The MZM structure is one of the most widely-used EOMs in practical applications, and LN is a suitable and tested material for high-speed and high-bandwidth EOMs, with stand-alone MZM devices having reached 110 GHz optical (70 GHz electrical) 3-dB modulation bandwidth about two decades ago in an unpackaged device [2]. However, the traditional design and fabrication approach of LN MZMs, based on ion exchange or implantation into bulk LN, is a relatively slow, expensive, and labor-intensive process, which is not compatible with the complex, multi-functional integrated optics microchips currently being developed and deployed.
The use of thin-film LN introduces some new parameters into the design space. For example, the ability to match the optical and microwave indices by varying the dimensions of the LN layer and the rib-loading Si waveguide offer new design opportunities to achieve true optical-RF phase matching to very high frequencies without artificial velocity matching structures. Indeed, electro-optic modulation sidebands have been measured to several hundred gigahertz — though far beyond the 3-dB electrical modulation bandwidth point — for ridge waveguide LN EOMs [38].
In our fabrication approach, depicted in Fig. 2, MZMs were built on a silicon photonics platform, using photolithography on silicon-on-insulator (SOI) wafers (220 nm Si thickness, 3 μm oxide thickness) and did not require sub-resolution features, unlike most plasmonic or polymeric slot modulators [39,40]. Silicon thinning (down to 150 nm) and feature patterning were followed by oxide deposition and subsequent chemical mechanical polishing (CMP) and oxide thinning by a timed wet etch (diluted hydrofluoric acid) process. 600 nm x-cut LN films were commercially-procured (NanoLN, Jinan Jingzheng Electronics Co., Ltd.) in the form of a 75 mm wafer separated from a Si handle by a 2 μm layer of SiO2 (known as LN-on-insulator, or LNOI). SOI and LNOI wafers were diced and indvidiual dies were cleaned using an RTA-1 wet etch. Plasma surface activation (PSA) of both the SOI and LNOI dies was achieved with a microwave frequency plasma asher at 150 Watts and 1.103 mbar for 150 seconds. Dies were immediately placed in de-ionized water for 10 minutes. Finally, pairs of SOI and LNOI dies were dried with nitrogen gas and bonded at room temperature. A vacuum pen was used to bond the LNOI to the SOI without contaminating either of the PSA surfaces. Bonded samples were then transferred to a hot plate at room temperature, where a weight was placed on the backside of the LNOI handle before increasing the hot plate temperature to 200°C for 1 hour followed by 250°C for 1 hour, 300°C for 2 hours, and natural cooling back to room temperature. No further processing of the LN film was performed during this fabrication process (e.g., no etching [3,12,14,15,41] or sawing [42]). The bonded stack has been shown to withstand repeated temperature-cycling to at least 300°C [43], sufficient for the post-processing required here. In fact, several fabricated chips were repeatedly processed, after bonding, through multiple cycles of electrode formation, removal and re-formation, in search of the optimal dimensions. No debonding or noticeable degradation to the stability or quality of the samples occurred during these additional process steps.
Following bonding, the LNOI Si handle was removed with a XeF2 isotropic dry etch. Before the XeF2 etch, all other areas of the bonded sample were coated with a protective layer of temporary bonding material (BrewerBond 220, Brewer Science, Inc.) to prevent undesired etching of the SOI substrate and waveguiding features. The 2 μm SiO2 insulator was removed with a hydrofluoric acid-based wet etch. Aluminum was deposited by sputtering after first sputtering a 10 nm chromium adhesion layer, and coplanar waveguide electrodes were patterned using photolithography. The total electrode thickness is 1.6 μm, the center electrode width is 30 μm, the outer electrodes have widths of 100 μm, and the electrode spacing is 12 μm. 90° bends are used to connect to the probe launch and termination pads, which have the same dimensions as the coplanar waveguide transmission line of the EOM. A fully fabricated chip is shown in Fig. 2(c); a microscope image of the EOM is provided in Fig. 2(d). The electrodes used here are more than 15 times thinner than those used in [2], thus improving fabrication practicality. The silicon photonic features were made on a high-resistivity Si handle wafer, with a measured resistivity (after HF etch to remove native oxide) of around 6 ×103 Ω−cm. Using a silicon substrate instead of the traditional LN substrate can mitigate piezoelectric resonances, such as those observed and discussed in [44]. As described below, the optical input and output from the MZM section were through (crystalline) silicon photonic waveguiding structures.
Fig. 2 (a) Thin film x-cut lithium-niobate (LN) on insulator dies were bonded at room temperature to segmented dies of a patterned and planarized silicon-on-insulator (SOI) wafer which contained fabricated silicon photonic waveguide circuits. No etching or patterning of the LN film was performed. (b) Exploded representation of the EOM, where an unpatterned, un-etched LN thin film was bonded to a Mach-Zehnder interferometer fabricated in Si. Aluminum electrodes were deposited on a 50 nm SiO2 layer over the LN film. ‘SiP Region’ denotes the SiO2-clad region outside the bonded LN film containing Si waveguide circuits, such as feeder waveguides, bends, directional couplers, and path-length difference segments. (c) Top view of a representative fabricated hybrid Si-LN EOM test chip, which contains 60 EOM waveguide structures in parallel (in the north-south direction); for this report, test electrodes for use in push-pull configuration were only fabricated on one EOM device. (d) Composite microscope image of the EOM (not to scale). DC: directional coupler, PLD: path-length difference, GSG: ground-signal-ground, SiP: Si photonics.
The die shown in Fig. 2(c) was taken from a 150 mm SOI wafer. Due to the hybrid nature of the waveguide geometry — shown in Fig. 3(a) — the thickness of the SiO2 layer, defined by CMP and timed-etch thinning, has a significant impact on the hybrid MZM’s 3-dB electrical modulation bandwidth and VπL. A measured wafer map of final SiO2 thickness over test Si features is shown in Fig. 3(b). Negative values correspond to regions where the oxide removal process was too aggressive, and polished into the Si layer. Using this data, 3-dB electrical modulation bandwidth (Fig. 3(c)) and VπL (Fig. 3(d)) values were simulated to show how they would vary for MZMs fabricated from dies across the wafer. Because of the sensitivity of the 3-dB electrical modulation bandwidth to electrical-optical index matching (discussed in more detail in Section 4), relatively minor changes to SiO2 thickness alter the bandwidth limitations of the device. This is clearly seen in Fig. 3(c), where the simulated 3-dB electrical modulation bandwidth varies from 70 GHz to greater than 250 GHz. VπL is also affected by the thickness of the oxide layer and its simulated values range from about 6.4 V.cm to 8.1 V.cm. Improved processing and planarization control will improve the yield of hybrid modulators from a full wafer.
Fig. 3 (a) Cross-section of the MZM. hCMP is the thickness of the CMP SiO2 layer between the Si features and the LN film. (b) Measured hCMP across the 150 mm wafer. Discrete measurements made on test sites are interpolated in this smoothly-varying color representation. An outline of the chip measured in this work is shown with dashed lines. (c) Calculated 3-dB electrical modulation bandwidth, and (d) Calculated VπL for an electrical frequency of 10 GHz, both based on the measured SiO2 thickness shown in panel (b), and without changing any other physical parameters or structural definitions.
3. Integration of LN EOM section with silicon photonics structures
As we have reported elsewhere, complex waveguiding circuits can be built up with a single bonded layer and multiple vertical transitions, using only silicon layer lithography to control the light path both outside the bonded region and also under the bonded LN area [45]. For the MZM reported here, the silicon photonics region outside the bonded LN area included four types of optical waveguide structures: fully-etched tapers for light input and output from the chip (edge couplers), single-mode broadband directional couplers (>15 dB extinction ratio (ER) throughout 1525 nm–1575 nm, maximum ER of approximately 30 dB), path-length difference (PLD) segment (including spline curve bends), and adiabatic waveguide tapers for inter-layer transitions (Si-to-LN and vice-versa). Precise foundry processing of the Si photonic features results in accurate and repeatable formation of the directional coupler splitting ratio. Since the LN layer is neither patterned nor etched in our design, there was no alignment issue at the bonding step; the Si features alone determine the optical propagation path.
Adiabatic waveguide tapers were designed to achieve a vertical inter-layer transition (from Si to LN, and the reverse). As shown in Fig. 4, the design uses the TE-polarized fundamental guided mode, which is also used in conventional silicon photonics at 1.5 μm wavelengths [46]. Since the refractive index of Si at these wavelengths (approximately 3.5) is significantly higher than that of LN (approximately 2.2), the large index difference enables control of the mode size and location (i.e., mainly in the Si rib or the LN slab) through lithography of the Si layer alone. Thus, only the width of the Si waveguide (w) was varied in our design; when w > 600 nm, light at 1.55 μm is mostly confined within the Si rib with confinement factor ΓSi = 64% (Mode A) and ΓSi = 58% (Mode B). For w = 320 nm, light is guided in Mode C and “sees” the LN slab layer, with confinement fraction in the LN layer calculated as ΓLN = 81% and ΓSi = 5%. Longitudinal Poynting vector simulations of these modes are shown in Fig. 4(c). We do not reduce the width of the Si rib (w) in Mode C further, in order to have a laterally confined optical mode despite the high value of ΓLN, which does not experience high optical loss from the metal electrodes. Also, by not letting the mode expand further in width, the Mode A to Mode B to Mode C transition loss is kept low; this transition is estimated from simulations to be 0.1 dB, and is described in more detail in previous work [45]. A benefit of these high-bandwidth modulators is that ΓLN has less variation with small errors in fabricated waveguide dimensions than plasmonic or polymeric slot waveguide MZMs, and the minimum feature sizes are easily achieved by today’s silicon photonics foundry processing technology.
Fig. 4 (a) Schematic of the EOM (not to scale, not showing electrodes), including two 3-dB directional couplers (DC) and a waveguide segment for path-length difference (PLD). Three optical waveguide modes are used, labeled as A, B, and C. Modes A (Si under SiO2) and B (Si under LN) have Si rib width w = 650 nm whereas mode C has w = 320 nm. (b) Dispersion curves (effective index versus w) in the hybrid region; w values for modes B and C are chosen to stay within the single-mode region of operation. An adiabatic waveguide transition (variation in w) is designed to evolve from mode B to C and vice versa. (c) Calculated Poynting vector components along the direction of propagation. Modes A and B are Si-guided and have a similar confinement fraction in Si. Mode C, with LN confinement factor (ΓLN) greater than 80%, is used in the phase-shifter segments.
Vertical, inter-layer transitions to and from the hybrid Si-LN region occur only where needed, inside the perimeter of the bonded region. Optical losses between Modes A and B are minimized by keeping the Si waveguide wide (w = 650 nm) when crossing into the hybrid region. Thus, the edges of the bonded thin film, even if rough on the scale of the optical wavelength, do not significantly affect optical propagation. This makes the back-end integration of thin-film LN simple and feasible, without requiring precision etching or patterning of either LN or silicon after bonding. Integration may also avoid the challenges of traditional packaging of LN EOMs: the 70-GHz unpackaged LN modulator of [2] was reported to achieve a 3-dB bandwidth, when packaged into a stand-alone module, of only about 30 GHz [47].
A design library of hybrid Si-LN components was created to aid in simulations and design within the Lumerical Interconnect simulation environment [48]. Light input and output was achieved using tapered single-mode, polarization-maintaining fibers. From test structures, an optical propagation loss of −0.6 dB/cm in the hybrid Si-LN region (Mode ’C’ in Fig. 4(c)) was measured. The propagation losses in the Si-only regions were about −1.3 dB/cm and are kept short in this design. The edges of the silicon photonic chip were lightly polished, but not fully prepared or packaged; hence, the edge coupling loss was about −3 dB per edge and the total fiber-to-fiber insertion loss was −13.6 dB. The calculated intrinsic loss of the full MZM (not including edge couplers), based on the measured propagation loss (−0.6 dB/cm in the EOM, −1.3 dB/cm outside the EOM) and the device length (0.5 cm for the EOM, 1.67 cm outside the EOM), as well as the simulated inter-layer transition loss estimates (−0.1 dB each), should be about −2.9 dB. However, the actual insertion loss was around −7.6 dB. The additional loss (−4.7 dB) is likely due to higher-than-expected attenuation in the broadband directional couplers and tapers, as well as an estimated (from simulations) additional −0.4 dB loss due to the electrical lines, which pass directly over the optical mode at the electrode bends (see Fig. 2(d)). It is also possible that the propagation losses in this particular device were higher than expected due to the multiple re-fabrication steps performed on this chip in search of the optimum electrode structure, which left organic and metallic residue on the lithium niobate film’s surface near the waveguides.
4. EOM measurements
At low speeds, the MZM demonstrated a high extinction ratio (> 20 dB) as shown in Fig. 5(a), with VπL = 6.7 V.cm at dc for an L = 0.5 cm device. This value of VπL is in good agreement with the expected value of approximately 6.4 V.cm from Fig. 3(d). A tabulation of VπL values for various thin-film LN EOMs is presented in [49] with values between 1 and 25 V.cm having been reported. As there is a wide spread in the reported values, a couple of comments are in order: the devices reported here were intended to maximize operating bandwidth, and were not specifically designed to minimize VπL.
RF measurements were performed on a bare-die chip using 50-Ω probes rated to 110 GHz and using laboratory equipment and RF waveguide components also rated and calibrated to about 110 GHz. The RF driving waveform was either from an RF oscillator (up to 67 GHz) or three different frequency multipliers (sequentially covering the range of frequencies up to 106 GHz). GSG probes were used for both launch and termination. Calibration of the signal pathway was performed using a high-frequency RF power sensor. To inform a computational model of the expected behavior, electrical S-parameters were measured using a Vector Network Analyzer up to 110 GHz, as shown in Fig. 5(b), and analyzed using standard algebraic transformations and lossy transmission line circuit analysis [50]. The artifact near 90 GHz is attributed to a defect in the probing pads caused by repeated contact probing, as it was not repeatable in other measured devices. Otherwise, the |S21|2 trend is smoothly-varying and follows expected behavior for an EOM using the Pockels effect in lithium niobate. The microwave refractive index, nm, was fitted to a power-law equation and the characteristic impedance, Zc, was fitted to a first-order polynomial of the RF frequency. As shown in Fig. 5(c), nm = 2.25 and Zc varied between 53.4 and 55.1 Ω from dc to 110 GHz. The microwave loss, αm, when fitted to a power-law equation, followed approximately an f1/4 dependence, in contrast with the typical f1/2 dependence in traditional EOMs [2]. Because this device has relatively thin electrodes and a silicon substrate, αm is not due solely to conductor loss, and instead, includes a combination of conductor, substrate, and radiation losses.
The method described in [51] was used to detect signals and modulation sidebands at an optical wavelength of 1560 nm. With the modulator biased at quadrature, the difference (log scale) between the optical intensity of the first sideband and carrier signal was used to extract the modulation index, and thus the frequency response, from 106 GHz down to 2 GHz (providing a safe margin for the 0.18 GHz resolution of the OSA). The peak-to-peak RF drive amplitude was about 0.7 Volts to 1 Volt up to 67 GHz, with variations in the RF drive amplitude due to a slight drop of power from the signal generator. Frequency multipliers were used for the frequency range above 67 GHz up to 106 GHz, in which case the peak-to-peak RF drive amplitude varied from about 0.5 Volts to 4 Volts, depending on the response of the multipliers. The electro-optic response is shown in Fig. 6, and an (electrical) 3-dB bandwidth was observed to lie well beyond 106 GHz, the limit of our measurement capabilities. The measurement matches well with the calculated response (solid black line in Fig. 6), which was obtained by using the measured values of Fig. 5(c) in Eq. (1). Electro-optic measurements at frequencies beyond 67 GHz require RF multipliers and time-consuming calibrations using appropriate waveguides, cables, probes and detectors for each frequency band, as observed elsewhere [52]. Some of the scatter in the measurements at the highest frequencies arises from the calibration of the frequency extenders, which have nonlinear and discontinuous dispersion curves. It may also be possible that, despite best efforts, high-speed probe contact varied between the calibration measurements and the EO response measurement, causing some of the ripple seen in Fig. 6. The measured flat-spectrum modulation response is consistent with our simulation based on electrical S-parameter measurements, which predicts flat frequency response to even higher frequencies.
Fig. 5 (a) Normalized optical transmission of the Mach-Zehnder interferometric electro-optic Modulator (MZM), versus dc voltage at optical wavelength λ = 1560 nm. Fitted VπL = 6.7 V.cm for device length L = 0.5 cm. (b) Measured electrical S-parameters of the MZM’s coplanar-waveguide transmission line. (c) Left y-axis: extracted microwave phase index nm and microwave loss αm (dB/cm) over the dc-110 GHz frequency range. Right y-axis: characteristic electrode impedance Zc (Ω). (d) Modest-speed eye diagram (20 Gbit/s) using on-off keying (OOK) modulation, generated using off-the-shelf optical communications test equipment. The measured signal-to-noise ratio is approximately 9.0.
Fig. 6 Electro-optic response of the EOM for both sidebands ((a) and (b)) from the optical spectrum analyzer. Solid black line: calculated response from electrical S-parameters of Fig. 5(c); black circles: electro-optic response from sideband OSA measurements. The 1.5 dB electrical modulation bandwidth is reached at the measurement limit of 106 GHz. The extrapolated 3 dB electrical modulation bandwidth indicated by the continuation of the solid black line is well over 200 GHz.
It is important to provide a justification for the measured ultra-high bandwidth EOM response, based on the detailed theory that has been developed for LN EOMs over the last few decades. The optical modulation response of an electro-optic Mach-Zehnder modulator based on phase modulation (a good approximation of the Pockels effect in LN) is [29]:
where m(ω) is the (linear) optical modulation response, no is the optical group index, nm and αm are the microwave index and loss coefficient, L is the phase-shifter arm length, ω is the RF frequency, ZL (RL) and ZG (RG) are the load and generator impedances (and, respectively, resistances), Z0 is the electrode transmission line characteristic impedance, Zin is the RF line input impedance, and using the definitions F(u) = [1−exp(u)]/u; u± = ±αmL + j(ω/c)[±nm − no]L. Complex RF propagation constant γm = αm + jωnmc, and c is the speed of light. Eq. (1) is a low-pass filter type response, whose 3-dB electrical roll-off frequency (f3dB,el) is maximized by matching of the optical and RF indices, matching the load and generator impedances, and minimizing the RF loss, but is not affected by the optical propagation loss. Lower optical propagation loss (αo) improves overall transmission, but for values of αo < 1 dB/cm and device lengths < 1 cm, the actual measured losses are dominated by non-idealities, such as imperfect chip coupling, or non-unitary power splitting at directional couplers, and further optical loss reduction plays only a minor role.The implications of Eq. (1) can be understood by assessing its behavior with regard to each significant parameter in turn, each of which can play an important role in limiting the 3-dB electrical modulation bandwidth. Assuming velocity and impedance matching, the RF-loss limited bandwidth results in a 3-dB point of αm(f3dB,el).L = 6.4 dB. Using L = 0.5 cm for the device under test, and requiring the 3-dB electrical frequency f3dB,el ≥ 100 GHz, we need αm(100 GHz) ≤ 12.8 dB.cm−1. As shown in Fig. 5, measurements showed αm(100 GHz) = 7.7 dB.cm−1, well under the limit, and thus, RF losses are not a limitation. Assuming impedance matching and no RF loss, the bandwidth limitation f3dB,el.L = (0.13/Δn) GHz.m, where Δn = nm−no. Thus, achieving f3dB,el ≥ 100 GHz for a 0.5 cm long device requires Δn = nm−no ≤ 0.26. Calculations for our device indicate that Δn ≈ 0.07, which is well under the threshold for achieving 100 GHz electrical modulation bandwidth. Moreover, the electrode length times bandwidth product ([53]) “LB” = c/Δn = 428 GHz.cm (i.e., zero modulation response at f = 428 GHz for a 1 cm long device), whereas the fabricated device which achieves beyond 106 GHz modulation has L = 0.5 cm, a factor of two shorter. Thus, neither RF losses nor index matching are fundamental limitations, as they have been in the past.
Using Vπ = 13.4 V, a data rate of 150 Gbps, and a load resistance of 50 Ω, the potential power consumption of this terminated and matched device is calculated to be Wbit = 1.5 pJ/bit [54]. Future embodiments of this technology can be designed to meet the low power, high bandwidth requirements of data centers and other short haul applications without significantly sacrificing modulation bandwidth. For example, a similar device with Vπ = 4 V and a data rate of 100 Gbps would have a power consumption of only Wbit = 0.2 pJ/bit.
5. Conclusion
This paper reports the first electro-optic Mach Zehnder modulator (MZM) based on single-mode silicon (Si) photonic circuits bonded at standard fabrication temperatures to an unpatterned, un-etched thin-film of lithium niobate (LN) which achieves greater-than-100-GHz electrical modulation bandwidth. Such a device can bring ultra-wide electro-optic bandwidths to integrated silicon photonics, and benefit applications in analog and digital communications, millimeter-wave instrumentation, analog-to-digital conversion, sensing, antenna remoting and phased arrays. The design of the EOM utilizes the well-known Pockels electro-optic effect in only the desired section of the light pathway. The input and output are in silicon photonics and, through the use of inter-layer vertical waveguide transitions, the device is not sensitive to the rough edges, if any, of the LN thin film. The fabrication process, which does not require etching or sawing of LN, is based on a standard silicon photonics foundry fabrication flow. The fabrication process thus brings lithium niobate, the traditional electro-optic material-of-choice in the first few decades of integrated optics, into compatibility with silicon photonics, the more recent platform for complex integrated optics.
Funding
National Science Foundation (NSF EFMA-1640968); San Diego Nanotechnology Infrastructure (NSF ECCS-1542148, UCSD Nano3 cleanroom); United States Department of Energy’s National Nuclear Security Administration DE-NA0003525.
Acknowledgments
P.O.W. acknowledges NDSEG fellowship support. Sandia is a multimission laboratory operated by National Technology and Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International Inc., for the United States Department of Energy’s National Nuclear Security Administration under contract DE-NA0003525. The views expressed in this article do not necessarily represent the views of the U.S. Department of Energy or the United States Government. The authors thank C. Levy and B. Rabet for useful discussions, as well as P. Asbeck for shared equipment.
Disclosures
P.O.W. is founder and CEO of Lightning Photonics Co., a start-up company commercializing thin-film lithium-niobate electro-optic modulation technology.
References
1. J. Sun, M. Sakib, J. Driscoll, R. Kumar, H. Jayatilleka, Y. Chetrit, and H. Rong, “A 128 Gb/s PAM4 silicon microring modulator,” in Optical Fiber Communication Conference Postdeadline Papers, OSA Technical Digest (Optical Society of America, 2018), paper Th4A.7. [CrossRef]
2. K. Noguchi, O. Mitomi, and H. Miyazawa, “Millimeter-wave Ti:LiNbO3 optical modulators,” J. Light. Technol. 16, 615–619 (1998). [CrossRef]
3. M. Zhang, C. Wang, X. Chen, M. Bertrand, A. Shams-Ansari, S. Chandrasekhar, P. Winzer, and M. Lončar, “Ultra-high bandwidth integrated lithium niobate modulators with record-low Vπ,” in Optical Fiber Communication Conference Postdeadline Papers, OSA Technical Digest (Optical Society of America, 2018), paper Th4A.5.
4. H. Takagi, R. Maeda, N. Hosoda, and T. Suga, “Room-temperature bonding of lithium niobate and silicon wafers by argon-beam surface activation,” Appl. Phys. Lett. 74, 2387–2389 (1999). [CrossRef]
5. M. M. R. Howlader, T. Suga, and M. J. Kim, “Room temperature bonding of silicon and lithium niobate,” Appl. Phys. Lett. 89, 031914 (2006). [CrossRef]
6. V. Stenger, J. Toney, A. Pollick, J. Busch, J. Scholl, P. Pontius, and S. Sriram, “Engineered thin film lithium niobate substrate for high gain-bandwidth electro-optic modulators,” in CLEO: 2013, OSA Technical Digest (Optical Society of America, 2013), paper CW3O.3.
7. L. Cao, A. Aboketaf, Z. Wang, and S. Preble, “Hybrid amorphous silicon (a-Si:H)–LiNbO3 electro-optic modulator,” Opt. Comm. 330, 40–44 (2014). [CrossRef]
8. L. Chen, Q. Xu, M. G. Wood, and R. M. Reano, “Hybrid silicon and lithium niobate electro-optical ring modulator,” Optica 1, 112–118 (2014). [CrossRef]
9. L. Chen, J. Chen, J. Nagy, and R. M. Reano, “Highly linear ring modulator from hybrid silicon and lithium niobate,” Opt. Express 23, 13255–13264 (2015). [CrossRef] [PubMed]
10. A. Rao, A. Patil, J. Chiles, M. Malinowski, S. Novak, K. Richardson, P. Rabiei, and S. Fathpour, “Heterogeneous microring and mach-zehnder modulators based on lithium niobate and chalcogenide glasses on silicon,” Opt. Express 23, 22746–22752 (2015). [CrossRef] [PubMed]
11. S. Jin, L. Xu, H. Zhang, and Y. Li, “LiNbO3 thin-film modulators using silicon nitride surface ridge waveguides,” IEEE Photon. Technol. Lett. 28, 736–739 (2016). [CrossRef]
12. A. J. Mercante, P. Yao, S. Shi, G. Schneider, J. Murakowski, and D. W. Prather, “110 GHz CMOS compatible thin film LiNbO3 modulator on silicon,” Opt. Express 24, 15590–15595 (2016). [CrossRef] [PubMed]
13. A. Rao, A. Patil, P. Rabiei, A. Honardoost, R. DeSalvo, A. Paolella, and S. Fathpour, “High-performance and linear thin-film lithium niobate mach–zehnder modulators on silicon up to 50 GHz,” Opt. Lett. 41, 5700–5703 (2016). [CrossRef] [PubMed]
14. V. E. Stenger, J. Toney, A. Pollick, D. Brown, B. Griffin, R. Nelson, and S. Sriram, “Low loss and low Vπ thin film lithium niobate on quartz electro-optic modulators,” in 2017 European Conference on Optical Communication, 1–3 (2017).
15. C. Wang, M. Zhang, B. Stern, M. Lipson, and M. Lončar, “Nanophotonic lithium niobate electro-optic modulators,” Opt. Express 26, 1547–1555 (2018). [CrossRef] [PubMed]
16. A. J. Mercante, S. Shi, P. Yao, L. Xie, R. M. Weikle, and D. W. Prather, “Thin film lithium niobate electro-optic modulator with terahertz operating bandwidth,” Opt. Express 26, 14810–14816 (2018). [CrossRef] [PubMed]
17. X. Lansiaux, E. Dogheche, D. Remiens, M. Guilloux-Viry, A. Perrin, and P. Ruterana, “LiNbO3 thick films grown on sapphire by using a multistep sputtering process,” J. Appl. Phys. 90, 5274–5277 (2001). [CrossRef]
18. Y. Sakashita and H. Segawa, “Preparation and characterization of linbo3 thin films produced by chemical-vapor deposition,” J. Appl. Phys. 77, 5995–5999 (1995). [CrossRef]
19. J.-G. Yoon and K. Kim, “Growth of highly textured LiNbO3 thin film on Si with MgO buffer layer through the sol-gel process,” Appl. Phys. Lett. 68, 2523–2525 (1996). [CrossRef]
20. A. Guarino, G. Poberaj, D. Rezzonico, R. Degl’Innocenti, and P. Günter, “Electro–optically tunable microring resonators in lithium niobate,” Nat. Photonics 1, 407 (2007). [CrossRef]
21. M. J. Heck, J. F. Bauters, M. L. Davenport, J. K. Doylend, S. Jain, G. Kurczveil, S. Srinivasan, Y. Tang, and J. E. Bowers, “Hybrid silicon photonic integrated circuit technology,” IEEE J. Sel. Top. Quantum Electron. 19, 6100117 (2013). [CrossRef]
22. C. Wang, M. J. Burek, Z. Lin, H. A. Atikian, V. Venkataraman, I.-C. Huang, P. Stark, and M. Lončar, “Integrated high quality factor lithium niobate microdisk resonators,” Opt. Express 22, 30924–30933 (2014). [CrossRef]
23. J. Wang, F. Bo, S. Wan, W. Li, F. Gao, J. Li, G. Zhang, and J. Xu, “High-Q lithium niobate microdisk resonators on a chip for efficient electro-optic modulation,” Opt. Express 23, 23072–23078 (2015). [CrossRef] [PubMed]
24. J. D. Witmer, J. A. Valery, P. Arrangoiz-Arriola, C. J. Sarabalis, J. T. Hill, and A. H. Safavi-Naeini, “High-Q photonic resonators and electro-optic coupling using silicon-on-lithium-niobate,” Sci. Rep. 7, 46313 (2017). [CrossRef] [PubMed]
25. J. Chiles, M. Malinowski, A. Rao, S. Novak, K. Richardson, and S. Fathpour, “Low-loss, submicron chalcogenide integrated photonics with chlorine plasma etching,” Appl. Phys. Lett. 106, 111110 (2015). [CrossRef]
26. P. Rabiei, J. Ma, S. Khan, J. Chiles, and S. Fathpour, “Submicron optical waveguides and microring resonators fabricated by selective oxidation of tantalum,” Opt. Express 21, 6967–6972 (2013). [CrossRef] [PubMed]
27. Z. Ren, P. J. Heard, J. M. Marshall, P. A. Thomas, and S. Yu, “Etching characteristics of LiNbO3 in reactive ion etching and inductively coupled plasma,” J. Appl. Phys. 103, 034109 (2008). [CrossRef]
28. C. J. G. Kirkby, “Low-energy ion-beam processing damage in lithium niobate surface-acoustic-wave optical waveguide devices and its post-manufacture removal,” J. Mater. Sci. 27, 3637–3641 (1992). [CrossRef]
29. G. Ghione, Semiconductor Devices for High-Speed Optoelectronics (Cambridge University, 2009). [CrossRef]
30. D. W. Dolfi and T. R. Ranganath, “50 GHz velocity-matched broad wavelength LiNbO3 modulator with multimode active section,” Electron Lett. 28, 1197–1198 (1992). [CrossRef]
31. K. Noguchi, O. Mitomi, H. Miyazawa, and S. Seki, “A broadband Ti:LiNbO3 optical modulator with a ridge structure,” J. Light. Technol. 13, 1164–1168 (1995). [CrossRef]
32. M. M. Howerton, R. P. Moeller, A. S. Greenblatt, and R. Krahenbuhl, “Fully packaged, broad-band LiNbO3 modulator with low drive voltage,” IEEE Photon. Technol. Lett. 12, 792–794 (2000). [CrossRef]
33. J. Kondo, A. Kondo, K. Aoki, M. Imaeda, T. Mori, Y. Mizuno, S. Takatsuji, Y. Kozuka, O. Mitomi, and M. Minakata, “40-Gb/s x-cut LiNbO3 optical modulator with two-step back-slot structure,” J. Light. Technol. 20, 2110–2114 (2002). [CrossRef]
34. E. L. Wooten, K. M. Kissa, A. Yi-Yan, E. J. Murphy, D. A. Lafaw, P. F. Hallemeier, D. Maack, D. V. Attanasio, D. J. Fritz, G. J. McBrien, and D. E. Bossi, “A review of lithium niobate modulators for fiber-optic communications systems,” IEEE J. Sel. Top. Quantum Electron. 6, 69–82 (2000). [CrossRef]
35. P. A. Stolk, H.-J. Gossmann, D. J. Eaglesham, D. C. Jacobson, C. S. Rafferty, G. H. Gilmer, M. Jaraíz, J. M. Poate, H. S. Luftman, and T. E. Haynes, “Physical mechanisms of transient enhanced dopant diffusion in ion-implanted silicon,” J. Appl. Phys. 81, 6031–6050 (1997). [CrossRef]
36. H. Han, L. Cai, and H. Hu, “Optical and structural properties of single-crystal lithium niobate thin film,” Opt. Mater. 42, 47–51 (2015). [CrossRef]
37. F. Niklaus, G. Stemme, J. Q. Lu, and R. J. Gutmann, “Adhesive wafer bonding,” J. Appl. Phys. 99, 031101 (2006). [CrossRef]
38. J. Macario, P. Yao, S. Shi, A. Zablocki, C. Harrity, R. D. Martin, C. A. Schuetz, and D. W. Prather, “Full spectrum millimeter-wave modulation,” Opt. Express 20, 23623 (2012). [CrossRef] [PubMed]
39. S. Zhu, G. Q. Lo, and D. L. Kwong, “Phase modulation in horizontal metal-insulator-silicon-insulator-metal plasmonic waveguides,” Opt. Express 21, 8320 (2013). [CrossRef] [PubMed]
40. M. Ayata, Y. Fedoryshyn, W. Heni, B. Baeuerle, A. Josten, M. Zahner, U. Koch, Y. Salamin, C. Hoessbacher, C. Haffner, D. L. Elder, L. R. Dalton, and J. Leuthold, “High-speed plasmonic modulator in a single metal layer,” Science 358, 630–632 (2017). [CrossRef] [PubMed]
41. G. Ulliac, V. Calero, A. Ndao, F. I. Baida, and M.-P. Bernal, “Argon plasma inductively coupled plasma reactive ion etching study for smooth sidewall thin film lithium niobate waveguide application,” Opt. Mater. 53, 1–5 (2016). [CrossRef]
42. N. Courjal, F. Devaux, A. Gerthoffer, C. Guyot, F. Henrot, A. Ndao, and M.-P. Bernal, “Low-loss LiNbO3 tapered-ridge waveguides made by optical-grade dicing,” Opt. Express 23, 13983 (2015). [CrossRef] [PubMed]
43. P. O. Weigel and S. Mookherjea, “Reducing the thermal stress in a heterogeneous material stack for large-area hybrid optical silicon-lithium niobate waveguide micro-chips,” Opt. Mater. 66, 605–610 (2017). [CrossRef]
44. J. L. Nightingale, R. A. Becker, P. C. Willis, and J. S. Vrhel, “Characterization of frequency dispersion in Ti indiffused lithium niobate optical devices,” Appl. Phys. Lett. 51, 716–718 (1987). [CrossRef]
45. P. O. Weigel, M. Savanier, C. T. DeRose, A. T. Pomerene, A. L. Starbuck, A. L. Lentine, V. Stenger, and S. Mookherjea, “Lightwave circuits in lithium niobate through hybrid waveguides with silicon photonics,” Sci. Rep. 6, 22301 (2016). [CrossRef] [PubMed]
46. Y. Vlasov and S. McNab, “Losses in single-mode silicon-on-insulator strip waveguides and bends,” Opt. Express 12, 1622–1631 (2004). [CrossRef] [PubMed]
47. M. M. Howerton and W. K. Burns, Broadband traveling wave modulators in LiNbO3 (Cambridge University, 2002).
48. P. O. Weigel and S. Mookherjea, “Process design kit and modulator simulation for hybrid silicon-lithium niobate integrated optics,” in Conference on Lasers and Electro-optics, OSA Technical Digest (Optical Society of America, 2017), paper JTu5A.126. [CrossRef]
49. A. Rao and S. Fathpour, “Compact lithium niobate electrooptic modulators,” IEEE J. Sel. Top. Quantum Electron. 24, 1–14 (2018).
50. D. M. Pozar, Microwave Engineering, 4th ed (Wiley, 2012).
51. Y. Shi, L. Yan, and A. E. Willner, “High-speed electrooptic modulator characterization using optical spectrum analysis,” J. Light. Technol. 21, 2358–2367 (2003). [CrossRef]
52. C. Hoessbacher, A. Josten, B. Baeuerle, Y. Fedoryshyn, H. Hettrich, Y. Salamin, W. Heni, C. Haffner, R. Schmid, D. L. Elder, D. Hillerkuss, M. Möller, L. R. Dalton, and J. Leuthold, “Broadband plasmonic modulator enabling single carrier operation beyond 100 Gbit/s,” in Optical Fiber Communication Conference, OSA Technical Digest (Optical Society of America, 2017), paper W4I.6. [CrossRef]
53. G. E. Betts, “Microwave bandpass modulators in lithium niobate,” Integr. Guid. Wave Opt. 1989 Tech. Dig. Ser. 4, 14–17 (1989).
54. S. Koeber, R. Palmer, M. Lauermann, W. Heni, D. L. Elder, D. Korn, M. Woessner, L. Alloatti, S. Koenig, P. C. Schindler, H. Yu, W. Bogaerts, L. R. Dalton, W. Freude, J. Leuthold, and C. Koos, “Femtojoule electro-optic modulation using a silicon-organic hybrid device,” Light. Sci. Appl. 4, e255 (2015). [CrossRef]
