Open Access
Add to BibSonomy
Abstract
Designing high-performance electro-optic polymer modulators requires insight into device-specific waveguide losses. Losses for rib waveguiding structures fabricated using SE0100C core and NOA73 cladding layers are reported. The waveguide loss was measured at 4.2 dB/cm, which is high compared to losses for different electro-optic rib waveguides reported at 1.2-2.9 dB/cm. The refractive index of unpoled SEO100C was measured at 1.655, expanding on the manufacturer’s reported optical constants of 1.65 (TE) and 1.7 (TM) for poled SEO100C. Utilizing the reported specific structure loss in device design for minimal loss will facilitate high performance electro-optic polymer modulator applications.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
In optical communication systems efficient use of optical power is critical, especially for long-haul communication. The introduction of integrated electro-optic (EO) modulators to a fiber optic network can significantly decrease the signal strength of the transmitter, creating the need for either higher power laser sources or the use of repeaters. Either option will add cost, complexity, and electrical power consumption to the system. Optimizing the EO modulator design and configuration to minimize the optical insertion loss is critical for widespread use in applications that demand high performance.
EO modulators are characterized by their insertion loss, EO coefficient (r33), and half-wave voltage (Vπ). Half-wave voltage in particular indicates the efficiency at which EO modulators modulate optical signals. Design choices such as device length, waveguide type, and waveguide material and will also directly impact a device’s half-wave voltage and insertion loss. Adjusting the active length in a modulator can fine-tune the half-wave voltage, as it is inversely proportional to the active length of EO material, but an increase in device length will also increase the insertion loss. In order to reduce the insertion loss related to fiber coupling, a waveguide mode size that matches the fiber mode is desirable. The single mode rib waveguide structure utilized in this paper is appealing in this context because it supports a larger mode than channel waveguides [1]. The rib waveguide mode closely matches input and output fiber mode sizes, resulting in greater coupling efficiency and a reduction in insertion loss.
EO material selection is critical to device performance. Maximizing the material’s EO coefficient is vital to minimizing the half-wave voltage but will also introduce material-specific characteristics contributing to device insertion loss. Lithium niobate is the dominant crystalline EO material used in commercially available EO modulators [2]. Limitations of lithium niobate are as follows: 1) The EO coefficient of lithium niobate is limited to 30.8 pm/V [3,4], 2) It has a relatively large dispersion between the indices of refraction for RF and optical signals; lithium niobate has a dispersion of 3.1 limiting operational bandwidth to 40 GHz [2,4,5], 3) Interfacing with silicon and other crystalline materials is very difficult based on its fragility and fixed crystal structure [4]. In particular, the lattice constant mismatch between lithium niobate (a = 5.15052 Å, c = 13.86496 Å) [6] and silicon (a = 5.43102 Å) [7] presents a challenge to directly interfacing the two materials.
EO polymers meet many of the limitations of lithium niobate: 1) The EO coefficients in EO polymers can exceed 100 pm/V [4,8–11], 2) EO polymers have a much smaller dispersion which allows them to modulate at higher frequencies than lithium niobate [4], 3) EO polymers are much easier to integrate with existing semiconductor technologies using techniques already present in semiconductor processing [4]. In reference to the EO polymer used in this study, SEO100C, it has a maximum poled EO coefficient of 140 pm/V [11] and a dispersion of less than 0.1 between the RF and optical frequency ranges [5].
There are numerous factors that can contribute to the insertion loss of EO polymer modulator devices including waveguide material, sidewall roughness, fiber coupling, and optical absorption by metal electrodes. All optical losses vary depending on the material set of a given device and, apart from fiber coupling losses, also depend on the waveguide length. Several benchmark figures reported at 1550 nm for rib waveguide structures made from different EO materials are noted in Table 1.
Table 1. Rib Waveguide Loss for various EO Materials
The 6.5 dB/cm loss measured in Table 1 for 3 µm SEO100C waveguides was not directly measured but instead was back-calculated from ring loss factor of a ring resonator modulator [14]. The associated ring loss factor was obtained from a fit of the empirical transmission spectrum of the ring resonator [14]. This measurement is closely linked to the coupling coefficient of the ring resonator so that any variation in the coupling coefficient will also impact the fit value of the ring loss factor. Furthermore, due to the structure of the ring resonator, bending losses could contribute to the reported value. The direct measurement of optical losses presented in this paper for straight SEO100C waveguides avoids bending loss and does not depend on the measurement of an additional device characteristic. Table 1 does not include the empirical loss measurements of 6 µm wide rib waveguides fabricated with the SEO100C polymer (core) and NOA73 (cladding) layers which are reported herein.
2. Waveguide fabrication
Rib waveguides were fabricated on a silicon substrate, as shown in Fig. 1 and Fig. 2. First, a bottom gold electrode was deposited. A 4.3 µm layer of NOA73 was then spun over the gold and UV cured to form the lower cladding layer.
Fig. 1. Vertical Cross section of rib waveguide. Width (W) is 6 um, core height (H) is 2.7 um, and etched core height (h) is 2.033 um.
Fig. 2. Vertical cross-section of a rib waveguide structure fabricated with a polycarbonate core layer. Polycarbonate is the host polymer for the nonlinear chromophores of SEO100C and behaves similarly to SEO100C during fabrication. Polycarbonate cross-sections are much easier to image than SEO100C cross sections, due to their superior contrast with the cladding layers.
For the electro-optic core, the solid mixture of guest-host polymer and SEO100C chromophore powder provided by the manufacturer was mixed with dibromomethane so that a weight percentage of 6.3% was achieved. The SEO100C layer was then spin coated and thermally cured according to manufacturer’s specifications, producing a film 2.7 µm thick [11]. The waveguide pattern was transferred onto the core layer through a soft contact photolithographic process and RIE etched with an oxygen plasma to create rib waveguides with rib heights of 667 nm. A rib height of 667 nm and a film thickness (H) of 2.7 µm ensures that the waveguides are single mode [10,14].
A 4.3 µm layer of NOA73 was spun and cured to form the top cladding layer. The top layer of gold was deposited and a second photolithographic process was performed to pattern the top electrodes, followed by a chemical etch of the top electrode. A shadow mask was then aligned to the wafer and oxygen plasma was used in a RIE process to etch through the polymers and gain access to the bottom electrode. Finally, the devices were diced using a nickel dicing blade to produce clean end faces.
3. Loss in optical waveguides
The insertion loss of an optical waveguide device can be broken into the various loss contributions stemming from the material properties and the rib waveguide structure. The total insertion loss is described by (1).
3.1 Ensuring zero electrode loss
For the fabricated device, electrode loss must be considered along the entire waveguide length because the bottom electrode was un-patterned and covered the entire wafer area. If the cladding layers are sufficiently thick enough to isolate the optical waveguide mode from the metal electrodes the electrode loss can be eliminated. To verify this, the electrode loss was simulated using BeamPROP 8.1 for waveguides of the same dimensions as the fabricated waveguides, with top and bottom cladding layers of varying thicknesses (Fig. 3). At the cladding thickness fabricated, the electrodes contribute 0.018 dB/cm to the overall insertion loss. This value is less than 1% of the measured length dependent losses reported in the next section, justifying the assumption that absorptive electrode loss is negligible.
3.2 Application of waveguide loss
Knowledge of the waveguide loss is particularly useful when designing devices to fall below a maximum loss specification. Many common waveguide devices such as Mach-Zehnder interferometers, directional couplers, and ring resonators require bending waveguides, which introduce the potential for bending losses. Bending loss increases as the radius of curvature for a bend decreases, therefore a longer arc length is necessary to reduce bending loss [17]. With the increase in arc length there comes an increase in path length loss and a tension between the waveguide loss and bending loss. Since bending loss can be determined theoretically, a precise empirical value for waveguide loss is needed to calculate the optimal waveguide arc length required to minimize device loss. Furthermore, the waveguide arc length and radius of curvature will also impact the footprint of a device.
4. Experimental procedure and results
4.1 Cutback method
Coupling and waveguide losses can be experimentally determined by the cutback method [16]. It is important to be able to distinguish the waveguide loss from the coupling loss, as coupling efficiency can vary depending on the waveguide end face quality and the coupling method used. Unlike waveguide loss, coupling loss is independent of waveguide length and corresponds to the static loss in cutback method data.
To measure insertion loss in the cutback method study, single mode 6 µm fibers were used to couple 1550 nm light into and out of the waveguides (Fig. 4). The 6 µm fiber mode better matched the 6 µm waveguide mode than modes from the 9 µm fibers more commonly used at 1550 nm. By matching the fiber and waveguide modes more closely, lower coupling loss can be achieved.
Fig. 3. Simulated electrode loss for various cladding thicknesses between the core and the electrodes.
Fig. 4. Experimental coupling setup. Input fiber is on the right, output fiber is on the left. Fibers were mounted on 3-axis micrometer stages.
The sample used had eight parallel straight rib waveguides fabricated according to the procedure described in the Waveguide Fabrication section above. The waveguide length was measured with calipers, the power from the input fiber was measured, each of the eight waveguides were coupled through, and power from the output fiber was measured. From the input and output fiber power measurements, the insertion loss was determined for each of the eight waveguides. The waveguides were cut back from the output end with a dicing saw so that the input end faces remained unchanged and the output end faces were of consistent quality across all waveguide lengths. After the sample was diced, the measurement process was repeated based on the new length. A linear regression of the insertion loss was performed to characterize the length-dependent and coupling losses. Figure 5 shows the collected data points with the linear regression. The y-intercept is the average coupling loss due to both the input and output fibers. The total coupling loss was found to be 7.2 ± 0.2 dB. The length-dependent losses were found to be 4.2 ± 0.2 dB/cm. Because electrode loss is negligible, the length-dependent losses represent the characteristic waveguide loss as defined in (2).
Fig. 5. Experimental results from the cutback method. Insertion loss of eight waveguides were measured at each waveguide length. Measurements were made at 1550 nm.
The 4.2 dB/cm measured waveguide loss is significantly less than the 6.5 dB/cm loss reported in [14]. The 2.3 dB/cm difference in loss can be attributed to several factors. The 3 µm waveguides utilized a taller rib height, thereby presenting a greater surface area for optical losses due to sidewall roughness [14]. The 6.5 dB/cm loss was calculated based on the loss factor in a ring resonator. Because of the bend in the ring resonator, a greater percentage of optical power would be present at the waveguide sidewall than in a straight waveguide due to the outward shift an optical mode experiences as it propagates in the ring waveguide [14,17,18]. Having more power present at the waveguide sidewall would also lead to an increase in roughness losses.
4.2 Ellipsometry
In addition to measuring the optical waveguide losses of the fabricated SEO100C structure the refractive index of the SEO100C mixed during fabrication was measured. Having an accurate measurement of the refractive index for a material will allow for more accurate modeling and design of future waveguide structures. The refractive index was found using transmission and reflection ellipsometry performed with a J.A. Woolam VASE Ellipsometer. Samples were prepared for ellipsometry by spinning a 2.7 µm film of SEO100C on an ITO glass slide. For ellipsometry to be effective, the thickness of the film needs to be on the same order as the desired wavelength [19]. The wavelengths at which the ellipsometry was performed are all within an order of magnitude of the 2.7 µm film thickness. The transparent substrate allowed for transmission studies to be conducted in addition to reflection ellipsometry. To improve the accuracy of the reflection ellipsometry results, an area of the back side of the substrate was roughened to eliminate the back reflections from the rear glass/air interface. The samples were measured with an optical profilometer to determine the film thickness and several sets of transmission and reflection ellipsometric data were collected. Knowing the sample thickness, the optical properties were determined by fitting the ellipsometric data.
A bare ITO glass slide was measured to accurately model the substrate the SEO100C sample was prepared on. Using the base model of the ITO slide, a model was created for the SEO100C sample. The fitting process was performed in the NIR range and SEO100C’s refractive index was fit from 1100 nm to 1700 nm, a range which includes the telecommunications wavelengths of 1310 nm and 1550 nm. Datasets from different locations on the sample were simultaneously fit. A Sellmeier equation was used to describe the dispersion curve and fit while matching the thickness of the modeled layer to the measured value. Using the manufacturer’s reported refractive index of 1.7 for poled SEO100C, an initial curve was chosen and then incrementally changed to minimize the error between the model and the collected data. The Sellmeier equation is shown in (3), with the coefficients listed in Table 2. For the listed coefficients the wavelength is given in microns. This fitting process found a refractive index value of 1.655 at 1550 nm and the measured dispersion curve of SEO100C is shown in Fig. 6.
Fig. 6. Dispersion curve for SEO100C as measured by ellipsometry. The refractive index at 1550 nm is highlighted with a value of 1.655.
Table 2. Sellmeier Coefficients for SEO100C
5. Conclusion
Electro-optic polymer rib waveguides were fabricated and their associated waveguide and coupling losses were characterized, providing information necessary to optimize device designs and structures. Complete knowledge of the SEO100C waveguide loss is useful in the design of integrated photonic devices using the SEO100C/NOA73 material set. This becomes particularly relevant when a design balance must be implemented between waveguide and bending losses, such as the S-bends in a Mach-Zehnder modulator or the ring of a ring resonator. Additionally, the loss data gathered here can be useful when designing a device’s size relative to its insertion loss.
Direct measurement of the losses for SEO100C have not been previously reported for fabricated rib waveguide structures. For the rib waveguide structure fabricated with respective core and cladding materials of SEO100C and NOA73, the device loss was characterized using the cutback method. The overall device loss was subdivided into three component losses: waveguide, coupling, and electrode losses. Simulations showed electrode loss to be negligible due to the cladding layer thickness providing adequate isolation between the optical mode and electrode metal. Waveguide loss at 1550 nm was determined using the cut back method, and the coupling loss was extrapolated from the determined waveguide loss. The measured waveguide loss of 4.2 ± 0.2 dB/cm includes the scattering and absorption losses inherent to the waveguide materials as well as the scattering loss due to waveguide sidewall roughness. An ellipsometer was used to measure a refractive index of 1.655 at 1550 nm for unpoled SEO100C. The extinction coefficient, however, was below the resolution limit for the ellipsometer. Further study is needed to differentiate the sidewall roughness component of the waveguide loss from the material components.
The 4.2 dB/cm measured waveguide loss was higher than generally expected, as other organic EO polymer waveguide losses around 1-2 dB/cm have been reported elsewhere [21]. The difference between the 4.2 dB/cm loss reported in this work and the 6.5 dB/cm loss reported in [14] can be attributed to different device geometries and differing dimensions of the waveguides measured, particularly the deeper etch depth which can increase roughness losses.
Changes in the waveguide geometry and fabrication process could reduce the waveguide losses. Designing a waveguide with a shorter rib height would reduce roughness losses by reducing the surface area of the sidewalls. However the reduced rib height will result in lower confinement of the optical mode [10,14]. Additionally the photolithography and RIE processes used to define the waveguides produce slanted sidewalls which can increase losses from the sidewall roughness [13]. The losses from chromophore scattering can be reduced by using a lower concentration of chromophore to host polymer, at the expense of a lower EO coefficient.
Despite the high 4.2 dB/cm loss measured for the waveguides in this work, the high EO coefficient of materials like SEO100C allow modulator devices with low driving voltages to be made smaller than those made from materials with lower EO coefficients. Since the EO coefficient of SEO100C is approximately four times larger than that of lithium niobate, a modulator made from SEO100C can be four times shorter than an equivalent modulator made from lithium niobate. The decrease in device size from the high EO coefficient serves to counteract the increased loss per unit length. Furthermore, the active region of most EO waveguide devices is usually smaller than the overall waveguide length. If devices can be designed such that EO polymers can be integrated with lower loss passive materials, there is no need to propagate light through EO polymer in the non-active regions of a device. Adiabatic taper designs have been developed to integrate EO polymer waveguides with passive waveguide materials [22]. Such designs show promise to retain the benefits of a high EO coefficient while mitigating the higher waveguide loss native to these materials by introducing lower loss passive materials for the non-active device lengths.
Funding
Naval Surface Warfare Center Crane and Alion Science and Technology Corporation, Proposal Number 1704010.
Acknowledgments
The authors would like to acknowledge the researchers from Soluxra LLC, particularly Dr. Jingdong Luo for the support provided regarding the processing of the SEO100C EO polymer.
Disclosures
The authors have no conflicts of interest to disclose.
References
1. S. P. Pogossian, L. Vescan, and A. Vonsovici, “The single-mode condition for semiconductor rib waveguides with large cross section,” J. Lightwave Technol. 16(10), 1851–1853 (1998). [CrossRef]
2. 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(1), 69–82 (2000). [CrossRef]
3. J.-M. Liu, Photonic Devices (Cambridge University Press, 2005).
4. Lumera Corporation, “Applications of Electro-optic Polymers and Devices: Breaking the High Frequency, Broad Bandwidth Barrier,” (February 2008).
5. Y. Enami, A. Seki, S. Masuda, T. Joichi, J. Luo, and A. K. Y. Jen, “Bandwidth Optimization for Mach–Zehnder Polymer/Sol–Gel Modulators,” J. Lightwave Technol. 36(18), 4181–4189 (2018). [CrossRef]
6. S. B. Abraham and P. Marsh, “Defect structure dependence on composition in lithium niobate,” Acta Crystallogr., Sect. B: Struct. Sci. 42(1), 61–68 (1986). [CrossRef]
7. “Lattice Parameter of Silicon” (NIST, 2018), retrieved 2019, https://physics.nist.gov/cgi-bin/cuu/Value?asil.
8. R. Himmelhuber, O. D. Herrera, R. Voorakaranam, L. Li, A. M. Jones, R. A. Norwood, J. Luo, A. K. Y. Jen, and N. Peyghambarian, “A Silicon-Polymer Hybrid Modulator—Design, Simulation and Proof of Principle,” J. Lightwave Technol. 31(24), 4067–4072 (2013). [CrossRef]
9. Y. Jouane, Y. C. Chang, D. Zhang, J. Luo, A. K. Jen, and Y. Enami, “Unprecedented highest electro-optic coefficient of 226 pm/V for electro-optic polymer/TiO(2) multilayer slot waveguide modulators,” Opt. Express 22(22), 27725–27732 (2014). [CrossRef]
10. O. Min-Cheol, Z. Hua, Z. Cheng, H. Erlig, C. Yian, B. Tsap, D. Chang, A. Szep, W. H. Steier, H. R. Fetterman, and L. R. Dalton, “Recent advances in electrooptic polymer modulators incorporating highly nonlinear chromophore,” IEEE J. Sel. Top. Quantum Electron. 7(5), 826–835 (2001). [CrossRef]
11. Soluxra LLC, “Processing Description of SEO100C,” (unpublished.).
12. M. F. Volk, S. Suntsov, C. E. Ruter, and D. Kip, “Low loss ridge waveguides in lithium niobate thin films by optical grade diamond blade dicing,” Opt. Express 24(2), 1386–1391 (2016). [CrossRef]
13. X.-L. Zhao, Y.-B. Yue, T. Liu, J. Sun, X.-B. Wang, X.-Q. Sun, C.-M. Chen, and D.-M. Zhang, “Optimized design and fabrication of nanosecond response electro–optic switch based on ultraviolet-curable polymers,” Chin. Phys. B 24(4), 044101 (2015). [CrossRef]
14. A. Hosseinzadeh, “Linear Ring Resonator Modulator for Microwave Photonic Links,” PhD (Michigan Technological University, Ann Arbor, 2018).
15. S. M. Lindecrantz and O. G. Hellesø, “Estimation of Propagation Losses for Narrow Strip and Rib Waveguides,” IEEE Photonics Technol. Lett. 26(18), 1836–1839 (2014). [CrossRef]
16. A. R. Tynes, A. D. Pearson, and D. L. Bisbee, “Loss Mechanisms and Measurements in Clad Glass Fibers and Bulk Glass,” J. Opt. Soc. Am. 61(2), 143–153 (1971). [CrossRef]
17. E. A. J. Marcatili and S. E. Miller, “Improved relations describing directional control in electromagnetic wave guidance,” The Bell Syst. Tech. J. 48(7), 2161–2188 (1969). [CrossRef]
18. K. L. Kruse, “Multi-mode and single mode polymer waveguides and structures for short-haul optical interconnects,” (Michigan Technological University, 2015).
19. J. A. Woolam Co., Guide to Using WVASE32 Spectroscopic Ellipsometry Data Acquisition and Analysis Software (J. A. Woolam Co., Inc., Lincoln, NE, 2008), p. 8.
20. M. J. Briseno, “Electro-Optic Contact Poling of Polymer Waveguide Devices and Thin Films,” M.S. (Michigan Technological University, Ann Arbor, 2017).
21. M. Eich, L. Dalton, B. Robinson, A. Jen, P. Ried, B. Eichinger, P. Sullivan, A. Akelaitis, D. Bale, M. Haller, J. Luo, S. Liu, Y. Liao, K. Firestone, N. Bhatambrekar, S. Bhattacharjee, J. Sinness, S. Hammond, N. Buker, R. Snoeberger, M. Lingwood, H. Rommel, J. Amend, S.-H. Jang, A. Chen, and W. Steier, “Electro-optic coefficients of 500 pm/V and beyond for organic materials,” in Linear and Nonlinear Optics of Organic Materials V, (2005).
22. Y. Enami, G. Meredith, N. Peyghambarian, and A. K. Y. Jen, “Hybrid electro-optic polymer/sol–gel waveguide modulator fabricated by all-wet etching process,” Appl. Phys. Lett. 83(23), 4692–4694 (2003). [CrossRef]
