Open Access
Add to BibSonomy
Abstract
Optical isolators are necessary components in photonic integrated circuits to ensure system stability by blocking the back-reflected light waves. In this study, a transverse electric mode magneto-optical isolator based on an asymmetric microring resonator without additional polarization rotators in series is demonstrated. Cerium-substituted yttrium iron garnet was integrated with silicon-on-insulator substrates by wafer bonding to break the Lorentz reciprocity on-chip after applying a unidirectional magnetic field. The isolator shows a 22 dB isolation ratio and a 4.3 dB insertion loss at an operation wavelength of 1572.62 nm.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
With the surge in global data volume, low-loss and high-performance photonic integrated circuits (PICs) have attracted wide research interest and are used in communication systems [1,2]. Among on-chip optical components, integrated optical nonreciprocal devices, represented by isolators, have been considered a long-standing challenge [3,4]. Optical isolators, which block the back-reflected light wave from reaching the laser cavity and optical amplifier to reduce the intensity and phase noise in the systems, are still a missing link in current PICs [5,6].
So far, the mechanisms for breaking the Lorentz reciprocity on-chip have been optical nonlinear effects [7,8], dynamic modulation [9,10], and magneto-optical (MO) effects [11,12]. Optical nonlinear devices, however, cannot function as isolators owing to their dynamic reciprocity [13]. Dynamic modulation requires complex control circuits that increase integration difficulty and energy consumption [14]. Therefore, the MO effect is the most suitable technical approach for the design and preparation of high-performance integrated optical isolators [15,16].
In recent years, the performance of integrated MO isolators has significantly improved [17]; however, there are still several challenges in the realization of transverse electric (TE) mode isolators. On the one hand, the nonreciprocal phase shift (NRPS) for the TE mode requires material distribution asymmetry in the in-plane direction. Thus, MO materials need to be deposited on the sidewall of the waveguide. The poor crystallinity of the deposited MO materials remains an issue for improving device performance [18,19]. High-performance TE mode isolators can be realized by placing polarization rotators in series with a transverse magnetic (TM) mode isolator. However, this requires an additional device footprint and tends to increase the insertion loss [20,21]. On the other hand, Mach–Zehnder interferometer (MZI) isolators require opposite magnetic fields in the two arms when the push-pull operation is employed [22]. A radial-direction magnetic field is required in the microring resonator (MRR) isolators demonstrated in [23]. The difficulty of magnetic-field integration remains an issue in these devices. Monolithic integration enables the control of the distribution of MO materials [24,25] or the fabrication of vertical tapers using grayscale lithography [26], allowing the isolators to work under a unidirectional magnetic field. Another approach is to deposit a permanent magnet film to achieve the passive integration of magnetic fields [27]. However, they all require multiple alignment lithography, which complicates the fabrication process and increases the cost.
The device structure proposed in this study was an MRR MO isolator based on asymmetric microring. The MO material was integrated by wafer bonding which has excellent integration temperature compatibility with other on-chip integrated components. The device operates under a unidirectional magnetic field, which simplifies the magnetic-field integration. Additionally, it was fabricated using a single lithography process. We combined “built-in” polarization rotators with the microring, enabling the device to use the NRPS of the TM mode in the microring to achieve TE mode isolation. Because the coupler used for coupling the microring with the bus line functions as a polarization rotator, there is no substantial increase in the footprint compared with the TM mode MRR isolator. Owing to the asymmetric structure, the mode propagating in the microring is different on the two sides of the microring, which avoids the cancelation of the NRPS under a unidirectional magnetic field. In the fabricated device, a high isolation ratio (IR) and low insertion loss (IL) were obtained.
2. Device design
2.1 Device structure and operation principle
An illustration of the proposed isolator prepared on a silicon-on-insulator (SOI) wafer is shown in Fig. 1(a). The isolator is composed of an asymmetrical microring and an input/output tapered waveguide with widths WR and WS, respectively. The cross-sectional structure of the MO waveguides is shown in Fig. 1(b). The 220-nm-thick Si core layer with a cerium-substituted yttrium iron garnet (Ce:YIG) top cladding integrated by wafer bonding provides structural asymmetry in the out-of-plane direction. After the application of a magnetic field in the x-axis direction, the propagation constant of the fundamental TM mode in the MO waveguide degenerates when propagating along the + z- and -z-directions. The difference is defined as the NRPS, which can be calculated as
Fig. 1. (a) Top view sketch of the device, the widths of microring and input/output waveguide are WR and WS, respectively. The lower microring waveguide is divided into six parts (Part1-6) for designed mode conversion. (b) Cross-section of the magneto-optical waveguide. Mode field distribution at specific locations (c) A, (d) B, and (e) C, as shown in (a).
The lower waveguide of the microring was divided into six parts (Part1-6) to realize the designed mode conversion. The input TE0 mode (Fig. 1(c)) first reached the TE0 to TE1 mode coupler in Part4, and partial energy was converted to the TE1 mode (Fig. 1(d)) and coupled to the microring. Owing to the relatively small fabrication tolerances of traditional directional couplers [28], this design uses complementary side-by-side tapered waveguides. By designing a nonadiabatic mode evolution, mode crosstalk was used to achieve energy coupling with better robustness. Part3 and Part5 were the buffer areas to ensure that the propagation length of the TM mode in the upper waveguide was fixed when the length of the coupler was changed. The TE1 mode in the microring then propagated to the connection waveguide in Part2 and reached the TE1 to TM0 mode converter in Part1. Using the adiabatic transmission of the hybrid mode to convert the mode efficiently, the converted TM0 mode (as shown in Fig. 1(e)) propagated and accumulated an NRPS of $\pi $ in the upper MO waveguide to achieve isolator operation. Then, by the reverse mode conversion process in Part6 and Part4, partial energy was converted into the TE0 mode and output. In backward transmission, owing to the NRPS, the resonant wavelength of the microring changed such that the forward and backward transmission characteristics were different, leading this device to work as a TE-polarized isolator.
2.2 Structural parameter design
We assumed that the refractive indices of the materials were ${n_{Si}}$ = 3.48, ${n_{Si{O_2}}}$ = 1.444, and ${n_{Ce:YIG}}$ = 2.2. The Faraday rotation of Ce:YIG was 4500 °/cm based on a previous experiment [29]. The effective refractive index (neff) of the transmission modes in a single MO waveguide varied with the waveguide width at a wavelength of 1550 nm, as shown in Fig. 2(a). The mode converter was designed using the hybrid mode of TE1 and TM0 within a width range where the neff of the first- and second-order modes (mode1 and mode2) were approaching (purple area in Fig. 2(a)). The TE ratio and NRPS of mode1 with different WR values near the hybridization point are shown in Fig. 2(b). Hybrid modes with lower TE ratios have larger NRPSs. The TE ratios were 95% and 5% when the WR was 975 and 775 nm, respectively. These can be considered the two ends of the mode converter.
Fig. 2. (a) Effective refractive index of the modes varying with the waveguide width of a single magneto-optical waveguide, where the purple (pink) areas indicate the width ranges for the mode converter (coupler). (b) The simulated TE ratio defined by $E_x^2/(E_x^2 + E_y^2)$ and NRPS of mode1 as a function of the microring width WR. (c) The TE1 ratio of supermodes in the coupler varies with waveguide width; the black dotted lines indicate the designed widths. (d) Mode conversion efficiency as a function of tapered waveguide length, dotted lines indicate the designed length. (e) The confinement factor of mode1 in Ce:YIG and Si as a function of the microring width WR.
For the mode coupler, we designed the width of one end as the WR of 975 nm, connected to the mode converter. Additionally, we chose WS = 500 nm, same as the input/output single-mode waveguide, as there was no coupling between the two waveguides. While keeping (WR + WS) constant and the gap width of 200 nm, WR (WS) becomes wider (narrower) around the coupling point where the neff of the TE0 mode coincides with that of the TE1 mode (pink area in Fig. 2(a)). There are two supermodes near the coupling point, i.e., supermode1 and supermode2. The power coupling efficiency between the supermodes in coupler and the TE1 mode in a single MO waveguide was defined as the TE1 ratio of the supermodes. By simulating the TE1 ratio of the supermodes at different WR (WS), as shown in Fig. 2(c), the other end of the coupler can be designed as WR = 1020 nm and WS = 455 nm, where the TE1 mode ratio of supermode1 and supermode2 are 99% and 1%, respectively. Finally, the output waveguide width was increased from 455 to 500 nm using a 130-µm-long linear taper after bending the waveguide.
The length of the upper waveguide of the microring was designed by the accumulated NRPS of $\pi $. Considering that the NRPS of the TM0 mode in a 775-nm-wide MO waveguide was 6.11 rad/mm as shown in Fig. 2(b), the length was determined to be 510 µm. The radius of the bending waveguide of the microring was designed to be 50 µm to ensure that the bending loss could be ignored. The mode conversion efficiency varied with the converter length, as indicated by the blue line in Fig. 2(d). The mode conversion efficiency was higher with longer L1,6. When L1,6 was 150 µm, an almost complete mode conversion was achieved. The tapered waveguide in Part2 involved no mode conversion and was, therefore, designed to be 30 µm to ensure negligible loss.
The coupler length must satisfy the critical coupling condition of the MRR. We first calculated the propagation loss of the light wave in the microring for one propagation cycle. Because there was no optical absorption of air and SiO2 at a wavelength of 1550 nm, the propagation loss was mainly caused by the optical absorption of Ce:YIG and the scattering loss of the Si waveguide. The confinement factors of mode1 in the Ce:YIG and Si layers, that is, ${\Gamma _{Ce:YIG}}$ and ${\Gamma _{Si}}$, respectively, at different WR are shown in Fig. 2(e). Assuming losses of ${\alpha _{Ce:YIG}}$ = 60 dB/cm and ${\alpha _{\textrm{Si}}}$ = 3 dB/cm, the propagation loss of the microring was calculated to be ${\alpha _R}$ = 4.3 dB for mode1 by [30]
Table 1. Structural parameters
3. Device preparation
Figure 3 shows the processing flow for the isolator fabrication. First, 200-nm-thick SiO2 was deposited by P-CVD as a mask layer on an SOI substrate with a 3-µm-thick SiO2 under cladding and a 220-nm-thick Si core layer. After coating with the ZEP520A electron-beam (EB) resist, it was patterned using EB lithography. The pattern was then transferred to the SiO2 and Si layers using a two-step reactive ion etching (RIE) process for CF4 and SF6. After removing the SiO2 mask by HF solution wet etching, Ce:YIG grown on the SGGG substrate was integrated on the SOI chip by surface-active wafer bonding. Device performance can be characterized after dicing.
4. Device characterization
4.1 Performance
Figure 4(a) shows a micrograph of the fabricated device, where the green square represents the bonded Ce:YIG/SGGG chip. Considering the fabrication error, multiple devices with different coupler lengths were fabricated to ensure that some devices could satisfy the critical coupling condition. Focusing lenses containing polarizers were employed to couple input and output light to improve the TE mode purity of the light. After testing with TE-polarized light wave input, we found that the device with L4 of 90 µm had the largest extinction ratio. The forward and backward transmission spectra of the device after application of an external magnetic field to magnetize Ce:YIG in the film plane are shown in Fig. 4 (b) which includes coupling loss due to the focusing lenses. The green line is the transmission spectrum of a straight waveguide that has the same length as the isolator with an air upper cladding. At a wavelength of 1572.62 nm, the device exhibits a 22 dB isolation ratio and a 4.3 dB insertion loss.
Fig. 4. (a) Micrograph of fabricated devices. The blue arrow indicates the direction of the applied magnetic field. (b) Transmission spectra of reference waveguide and isolator in forward and backward transmission.
4.2 Discussion
We examined the fabrication error by SEM measurement, as shown in Fig. 5(a), and obtained the actual width parameters. By simulating with the actual parameters and L4 of 90 µm at the operating wavelength of 1572.62 nm, the $|t |$ of the coupler was 0.63, which was in line with the design expectation. The confinement factor for different microring widths WR was re-simulated, as shown in Fig. 5(b). The propagation loss of the microring can be calculated using Eq. (2) as ${\alpha _R}$ = 4.1 dB, and the corresponding attenuation coefficient $\alpha $ was 0.63. Simulations proved that the device operated under the critical coupling condition, and the assumed material loss in the simulations was reasonable. The transmission efficiency of the MRR can be calculated as [31]
Fig. 5. (a) The design width of the device at specific locations compared with the actual width by SEM measurement. (b) The confinement factor of mode1 (TE0 mode) in Ce:YIG and Si varying with the WR (WS) based on actual width parameters and operating wavelength.
Substituting $\alpha $ = $|t |$ = 0.63 into Eq. (3), the IL caused by the MRR at a non-resonant wavelength ($\varphi $ = (2n+1)$\pi $) was 0.9 dB. This value is much lower than ${\alpha _\textrm{R}}$ because only partial energy is coupled into the microring, and the rest does not experience the propagation loss of the microring. The remaining IL of 3.4 dB was from three origins: 0.7 dB caused by the difference in the operation wavelength in forward and backward transmission due to insufficient NRPS, which originated from the hybrid mode containing the TM component in the mode converter. Because the TM component experienced an NRPS (as shown by the red line in Fig. 2(b)) with the opposite sign with respect to the upper waveguide, it contributed to reducing the overall NRPS. A relatively large MO loss was observed in the input/output waveguides additionally covered by the MO material. The length of the input/output MO waveguide was 1525 µm, which was slightly longer than the Ce:YIG chip length of 1500 µm, owing to the bending waveguides. By simulating the confinement factor of the linear tapered structure at the input/output waveguide, as shown in Fig. 5(b), the MO loss in the input/output MO waveguides was estimated to be 2.3 dB. There was a mode mismatch loss of 0.2 dB for the junction structure between the air cladding and MO cladding waveguides at each edge of the Ce:YIG chip. Thus, the contribution of the mode-mismatch loss accounted for 0.4 dB in total. The overall loss breakdown of the isolator is listed in Table 2.
Table 2. Loss breakdown
If we could further control the wafer bonding accuracy, such as using transfer printing, the device loss could be greatly reduced by covering only the upper waveguide of the microring with the MO material. The MO loss in the input/output waveguides would be eliminated. There would be no NRPS of the hybrid modes in the mode converter, so the IL due to insufficient NRPS would be also eliminated. Simultaneously, IR would be slightly improved. The junction loss would be included in the propagation loss of the microring. In this case, the junction loss is 1.5 dB for the TM mode at each edge, which is much higher compared with the TE mode. But the MO loss, which is caused only by the 510-µm-long TM-mode MO waveguide, is calculated to be 2 dB. The attenuation coefficient $\alpha $ corresponding to the total propagation loss of 5 dB becomes 0.56. By assuming $\alpha $ = $|t |$ to ensure the critical coupling condition, the IL at the non-resonant wavelength is calculated to be 1.4 dB by Eq. (3). Additionally, higher IR could be achieved by further optimizing the extinction ratio of the MRR. Therefore, the performance of the isolator still has a potential for improvement.
5. Conclusions
By designing the asymmetric microring structure, we realized a TE mode isolator without serial polarization rotators. The device operates under a unidirectional magnetic field, which simplifies the magnetic-field integration. The fabricated isolator showed an IR of 22 dB and an IL of 4.3 dB at a wavelength of 1572.62 nm. The device performance can be further improved by controlling the distribution of MO materials and is expected to achieve a lower IL of 1.4 dB. This device demonstrates the possibility of further simplifying the optical component integration process and reducing the device footprint, which enables higher-density PICs.
Funding
Core Research for Evolutional Science and Technology (JPMJCR18T4); Japan Society for the Promotion of Science (19H02190); New Energy and Industrial Technology Development Organization (JPNP13004, JPNP16007); China Scholarship Council.
Disclosures
The authors declare no conflicts of interest.
Data availability
The data that support the findings of this study are available from the corresponding authors upon reasonable request.
References
1. D. Marpaung, C. Roeloffzen, R. Heideman, A. Leinse, S. Sales, and J. Capmany, “Integrated microwave photonics,” Laser Photonics Rev. 7(4), 506–538 (2013). [CrossRef]
2. W. Bogaerts, D. Pérez, J. Capmany, D. A. B. Miller, J. Poon, D. Englund, F. Morichetti, and A. Melloni, “Programmable photonic circuits,” Nature 586(7828), 207–216 (2020). [CrossRef]
3. Y. Yang, T. Liu, L. Bi, and L. Deng, “Recent advances in development of magnetic garnet thin films for applications in spintronics and photonics,” J. Alloys Compd. 860, 158235 (2021). [CrossRef]
4. Y. Shoji, K. Miura, and T. Mizumoto, “Optical nonreciprocal devices based on magneto-optical phase shift in silicon photonics,” J. Opt. 18(1), 013001 (2016). [CrossRef]
5. C. Zhang, P. Dulal, B. J. H. Stadler, and D. C. Hutchings, “Monolithically-integrated TE-mode 1D silicon-on-insulator isolators using seedlayer-free garnet,” Sci. Rep. 7(1), 5820 (2017). [CrossRef]
6. Y. Shoji and T. Mizumoto, “Magneto-optical non-reciprocal devices in silicon photonics,” Sci. Technol. Adv. Mater. 15(1), 014602 (2014). [CrossRef]
7. L. Fan, J. Wang, L. T. Varghese, H. Shen, B. Niu, Y. Xuan, A. M. Weiner, and M. Qi, “An all-silicon passive optical diode,” Science 335(6067), 447–450 (2012). [CrossRef]
8. Y. Shi, Z. Yu, and S. Fan, “Limitations of nonlinear optical isolators due to dynamic reciprocity,” Nat. Photonics 9(6), 388–392 (2015). [CrossRef]
9. C. R. Doerr, L. Chen, and D. Vermeulen, “Silicon photonics broadband modulation-based isolator,” Opt. Express 22(4), 4493–4498 (2014). [CrossRef]
10. D. B. Sohn, S. Kim, and G. Bahl, “Time-reversal symmetry breaking with acoustic pumping of nanophotonic circuits,” Nat. Photonics 12(2), 91–97 (2018). [CrossRef]
11. R. Yamaguchi, Y. Shoji, and T. Mizumoto, “Low-loss waveguide optical isolator with tapered mode converter and magneto-optical phase shifter for TE mode input,” Opt. Express 26(16), 21271–21278 (2018). [CrossRef]
12. S. Liu, Y. Shoji, and T. Mizumoto, “Mode-evolution-based TE mode magneto-optical isolator using asymmetric adiabatic tapered waveguides,” Opt. Express 29(15), 22838–22846 (2021). [CrossRef]
13. J. B. Khurgin, “Non-reciprocal propagation versus non-reciprocal control,” Nat. Photonics 14(12), 711 (2020). [CrossRef]
14. D. L. Sounas and A. Alù, “Non-reciprocal photonics based on time modulation,” Nat. Photonics 11(12), 774–783 (2017). [CrossRef]
15. M. C. Tien, T. Mizumoto, P. Pintus, H. Kromer, and J. E. Bowers, “Silicon ring isolators with bonded nonreciprocal magneto-optic garnets,” Opt. Express 19(12), 11740–11745 (2011). [CrossRef]
16. J. Fujita, M. Levy, R. M. Osgood Jr, L. Wilkens, and H. Dötsch, “Waveguide optical isolator based on Mach–Zehnder interferometer,” Appl. Phys. Lett. 76(16), 2158–2160 (2000). [CrossRef]
17. W. Yan, Y. Yang, W. Yang, J. Qin, L. Deng, and L. Bi, “On-chip nonreciprocal photonic devices based on hybrid integration of magneto-optical garnet thin films on silicon,” IEEE J. Sel. Top. Quantum Electron. 28(3), 1–15 (2022). [CrossRef]
18. Y. Zhang, Q. Du, C. Wang, T. Fakhrul, S. Liu, L. Deng, D. Huang, P. Pintus, J. Bowers, C. A. Ross, J. Hu, and L. Bi, “Monolithic integration of broadband optical isolators for polarization-diverse silicon photonics,” Optica 6(4), 473–478 (2019). [CrossRef]
19. E. Ishida, K. Miura, Y. Shoji, H. Yokoi, T. Mizumoto, N. Nishiyama, and S. Arai, “Amorphous-Si waveguide on a garnet magneto-optical isolator with a TE mode nonreciprocal phase shift,” Opt. Express 25(1), 452–462 (2017). [CrossRef]
20. R. Ma, S. Reniers, Y. Shoji, T. Mizumoto, K. Williams, Y. Jiao, and J. van der Tol, “Integrated polarization-independent optical isolators and circulators on an InP membrane on silicon platform,” Optica 8(12), 1654–1661 (2021). [CrossRef]
21. P. Pintus, D. Huang, P. A. Morton, Y. Shoji, T. Mizumoto, and J. E. Bowers, “Broadband TE optical isolators and circulators in silicon photonics through Ce: YIG bonding,” J. Lightwave Technol. 37(5), 1463–1473 (2019). [CrossRef]
22. D. Huang, P. Pintus, Y. Shoji, P. Morton, T. Mizumoto, and J. E. Bowers, “Integrated broadband Ce: YIG/Si Mach–Zehnder optical isolators with over 100-nm tuning range,” Opt. Lett. 42(23), 4901–4904 (2017). [CrossRef]
23. D. Huang, P. Pintus, C. Zhang, Y. Shoji, T. Mizumoto, and J. E. Bowers, “Electrically driven and thermally tunable integrated optical isolators for silicon photonics,” IEEE J. Sel. Top. Quantum Electron. 22(6), 271–278 (2016). [CrossRef]
24. W. Yan, Y. Yang, S. Liu, Y. Zhang, S. Xia, T. Kang, W. Yang, J. Qin, L. Deng, and L. Bi, “Waveguide-integrated high-performance magneto-optical isolators and circulators on silicon nitride platforms,” Optica 7(11), 1555–1562 (2020). [CrossRef]
25. L. Bi, J. Hu, P. Jiang, D. H. Kim, G. F. Dionne, L. C. Kimerling, and C. A. Ross, “On-chip optical isolation in monolithically integrated non-reciprocal optical resonators,” Nat. Photonics 5(12), 758–762 (2011). [CrossRef]
26. Q. Du, C. Wang, Y. Zhang, Y. Zhang, T. Fakhrul, W. Zhang, C. Gonçalves, C. Blanco, K. Richardson, L. Deng, C. A. Ross, L. Bi, and J. Hu, “Monolithic on-chip magneto-optical isolator with 3-dB insertion loss and 40 dB isolation ratio,” ACS Photonics 5(12), 5010–5016 (2018). [CrossRef]
27. T. Murai, Y. Shoji, N. Nishiyama, and T. Mizumoto, “Nonvolatile magneto-optical switches integrated with a magnet stripe array,” Opt. Express 28(21), 31675–31685 (2020). [CrossRef]
28. Y. Shoji, A. Fujie, and T. Mizumoto, “Silicon waveguide optical isolator operating for TE mode input light,” IEEE J. Sel. Top. Quantum Electron. 22(6), 264–270 (2016). [CrossRef]
29. Y. Shoji, T. Mizumoto, H. Yokoi, I. W. Hsieh, and R. M. Osgood Jr, “Magneto-optical isolator with silicon waveguides fabricated by direct bonding,” Appl. Phys. Lett. 92(7), 071117 (2008). [CrossRef]
30. J. T. Robinson, K. Preston, O. Painter, and M. Lipson, “First-principle derivation of gain in high-index-contrast waveguides,” Opt. Express 16(21), 16659–16669 (2008). [CrossRef]
31. W. D. Sacher and J. K. Poon, “Dynamics of microring resonator modulators,” Opt. Express 16(20), 15741–15753 (2008). [CrossRef]
