Chinese Physics Letters, 2016, Vol. 33, No. 2, Article code 024203 Two-Mode Converters at 1.3 μm Based on Multimode Interference Couplers on InP Substrates * Fei Guo(郭菲), Dan Lu(陆丹)**, Rui-Kang Zhang(张瑞康), Hui-Tao Wang(王会涛), Wei Wang(王圩), Chen Ji(吉晨) Affiliations Key Laboratory of Semiconductor Materials, and Beijing Key Laboratory of Low Dimensional Semiconductor Materials and Devices, Institute of Semiconductors, Chinese Academy of Sciences, Beijing 100083 Received 25 October 2015 *Supported by the National Basic Research Program of China under Grant No 2014CB340102, and the National Natural Science Foundation of China under Grant Nos 61474111 and 61274046.
**Corresponding author. Email: ludan@semi.ac.cn
Citation Text: Guo F, Lu D, Zhang R K, Wang H T and Wang W et al 2016 Chin. Phys. Lett. 33 024203 Abstract Two-mode converters at 1.3 μm, aiming at applications in mode-division multiplexing in Ethernet systems, are proposed and experimentally demonstrated. Based on multimode interference couplers, the two-mode converters with 50% and 66% mode conversion efficiencies are designed and fabricated on InP substrates. Mode conversion from the fundamental mode (TE$_{0})$ to the first order mode (TE$_{1})$ is successfully demonstrated within the wavelength range of 1280–1320 nm. The 1.3-μm mode converters should be important devices in mode-division multiplexing systems in Ethernet systems. DOI:10.1088/0256-307X/33/2/024203 PACS:42.79.Sz, 42.79.Gn, 42.25.Bs © 2016 Chinese Physics Society Article Text In recent years, the mode-division multiplexing (MDM) technology has been proposed and demonstrated in combination with the wavelength division multiplexing (WDM) technology to improve the transmission capacity of fibers in 1.55-μm long haul transmission systems.[1,2] As the key components in MDM systems, mode converters can be realized by several techniques, including free-space optics-based methods such as phase plate,[3] spatial light modulator (SLM);[4] fiber-based methods such as long-period fiber Bragg grating (LPFBG),[5] photonic lanterns;[6] and planar waveguide-based methods such as directional couplers[7] and multimode interference (MMI) couplers.[8,9] Among them, the planar waveguide-based technique provides a favorable solution for its low insertion loss, compact structure, and integration possibility with other devices. Many device structures have been realized on silicon-on-insulator (SOI) platforms. However, due to the lack of integrable active components in SOI platforms, it is only possible to realize integrated few-mode optical transmitters by hybrid integration of III–V materials on the SOI platform,[10] which requires complex knowledge and complicated processing technologies for both platforms. Meanwhile the photonic integration technology based on the InP material system can provide a possibility that both the active components such as lasers, modulators, detectors and the passive components such as the few-mode converters, multiplexers, and wavelength division multiplexing components can be integrated in the same material system. The MMI couplers have the advantages of small size, low loss, large fabrication tolerance and large optical bandwidth in comparison with other planar waveguide-based methods, and mode converter-multiplexers based on MMI couplers have been investigated and fabricated on InP substrates in the 1.5-μm wavelength by Leuthold et al. in 1998.[11] This structure provides a core component for integrated few-mode optical transmitters in long-haul optical systems working in the 1.5-μm wavelength range. With the standardization of 802.3ba 40 G/100 G Ethernet standard,[12] optical transmitters working at a wavelength of 1.3 μm are becoming increasingly important. It can be envisioned that the MDM technology will also be introduced in the Ethernet system in the near future. However, there is still no research on the 1.3-μm band few-mode devices. In this Letter, we report the design and fabrication of two mode converters working around 1.3 μm based on MMI couplers on InP substrates. Mode converters with mode-conversion efficiencies of 50% and 66% are demonstrated. Mode conversion from the TE$_{0}$ mode to the TE$_{1}$ mode was successfully realized in the wavelength range of 1280–1320 nm. The 1.3-μm mode converters are possible candidates for monolithically integrated few-mode transmitters in future Ethernet systems.
cpl-33-2-024203-fig1.png
Fig. 1. The basic structure of the MMI-based two-mode converters with (a) 50% and (b) 66% mode conversion efficiencies.
The basic structures of the MMI-based two-mode converters are shown in Figs. 1(a) and 1(b). The design principle is similar to that reported in Ref. [11], where the 50% and 66% mode converters are based on the $4\times4$ MMI and $3\times3$ MMI couplers, respectively. The widths ($W_{0}$) of the input 1, outputs 1 and 3 are designed to support the TE$_{0}$ mode only, while the widths (2$W_{0}$) of input 2 and output 2 are designed to support both the TE$_{0}$ mode and the TE$_{1}$ mode. The lengths ($L_{\rm MMI}$) of the multimode section are 3$L_{\pi}$/4 and $L_{\pi}$ for the 50% and 66% mode converters, where $L_{\pi}$ is the beat length of the two lowest-order modes.[13] The phase relations of the general $4\times4$ MMI and $3\times3$ MMI are illustrated in Tables 1 and 2, respectively, where $i=1$, 2, 3 and 4 denoting (bottom-up) the number of the input waveguides and $j=1$, 2, 3 and 4 denoting (top-down) the number of the output waveguides.[13] In the 50% mode converter, the phase relations of the output modes of the outputs 1 ($j=1$), 2 ($j=2$ and 3), and 3 ($j=4$) are $-\pi$, $-7\pi$/4 and $-3\pi$/4, $-\pi$, respectively, when a TE$_{0}$ mode is injected from the input waveguide input 1 ($i=1$). In output 2, the two output TE$_{0}$ modes with the same amplitude and phase shift of $\pi$ are combined as the TE$_{1}$ mode, while the outputs in outputs 1 and 3 still maintain the TE$_{0}$ mode. In the 66% mode converter, the phase relations of the output waveguides outputs 1 ($j=1$), and 2 ($j=2$ and 3) are $-\pi$, $-2\pi$/3 and $-5\pi$/3 when a TE$_{0}$ mode is launched into the input waveguide input 1 ($i=1$). Similarly, in output 2, the two output TE$_{0}$ modes with the same amplitude and phase difference of $\pi$ are combined as the TE$_{1}$ mode, while the output modes in output 1 remain as the TE$_{0}$ mode. On the other hand, the TE$_{0}$ mode launched into input 2 will directly transmit to output 2 without mode conversion. When output 2 is designated as the output for the converter, a mode-multiplexed signal (TE$_{1}$ from input 1 and TE$_{0}$ from input 2) will be obtained if two signals are launched from inputs 1 and 2 simultaneously. Therefore, the two-mode converters can also be used as the mode multiplexers.
Table 1. The phase relations of the general $4\times4$ MMI (50% mode converter).
$N$=4 $j$=1 $j$=2 $j$=3 $j$=4
$i$=1 $-4\pi/4$ $-3\pi/4$ $-7\pi/4$ $-4\pi/4$
$i$=2 $-3\pi/4$ $-4\pi/4$ $-4\pi/4$ $-7\pi/4$
$i$=3 $-7\pi/4$ $-4\pi/4$ $-4\pi/4$ $-3\pi/4$
$i$=4 $-4\pi/4$ $-7\pi/4$ $-3\pi/4$ $-4\pi/4$
Table 2. The phase relations of the general $3\times3$ MMI (66% mode converter).
$N$=3 $j$=1 $j$=2 $j$=3
$i$=1 $-3\pi/3$ $-\pi 2/3$ $-5\pi/3$
$i$=2 $-2\pi/3$ $-3\pi/3$ $-2\pi/3$
$i$=3 $-\pi /3$ $-2\pi/3$ $-3\pi/3$
The cross section of the waveguide is shown in Fig. 2. The material structure includes a 500-nm InP buffer, a 300-nm InGaAsP core layer with a 1.15-μm bandgap and a 1.7-μm InP cladding layer. The widths ($W_{0}$) of input 1 and outputs 1 and 3 are set to 2 μm, and the widths (2$W_{0}$) of input 2 and output 2 are 4 μm. The multimode section width ($W_{\rm MMI}$) is set to 12 μm for both of the 50% and 66% mode converters. The three-dimensional beam propagation method (3D-BPM) was used to simulate the mode conversion properties, where the TE mode at the operating wavelength of 1.31 μm was considered. The optimized multimode section lengths ($L_{\rm MMI}$) of the 50% and 66% mode converters were 371 μm and 495 μm, respectively. Figures 3(a) and 3(b) show the simulated intensity distribution when a TE$_{0}$ mode is launched into input 1. As can be seen, the intensity from output 2 shows a clear TE$_{1}$ mode distribution.
cpl-33-2-024203-fig2.png
Fig. 2. The cross section of the waveguide.
cpl-33-2-024203-fig3.png
Fig. 3. The simulated intensity distribution when a TE$_{0}$ mode is launched into input 1 of (a) 50% mode converters and (b) 66% mode converters.
cpl-33-2-024203-fig4.png
Fig. 4. The simulated transmission spectra of the 50% mode converter.
The simulated wavelength dependences of the transmission spectra are shown in Figs. 4 and 5, when a TE$_{0}$ mode is injected into input 1. It can be seen that when the TE$_{0}$ mode is converted to the TE$_{1}$ mode, the insertion losses are 3.04 dB and 1.8 dB, respectively, for the 50% and 66% mode converters at the operating wavelength of 1.31 μm. Considering the intrinsic losses of 3 dB and 1.8 dB in the 50% and 66% mode converters, the simulated excess losses for the two structures at 1.31 μm are negligible, while in the whole scanning range starting from 1.2 μm to 1.4 μm, the excess losses are below 6 dB and 8.3 dB, respectively. In our later experiment, the wavelength was scanned from 1280 nm to 1320 nm, in the range the simulated excess losses are below 0.5 dB and 0.7 dB, respectively. These simulated results indicate that the mode converters have relatively large optical bandwidth over 40 nm, where the excess losses are below 1 dB. The transmission curves of TE$_{0}$ modes transmitted from outputs 1 and 3 are also plotted in Figs. 4 and 5.
cpl-33-2-024203-fig5.png
Fig. 5. The simulated transmission spectra of the 66% mode converter.
The epitaxial materials of the devices were grown on InP substrates by the metal organic chemical vapor deposition (MOCVD). Then, the ridge waveguide was formed by the dry etching process using inductively coupled plasma (ICP). The scanning electron microscope (SEM) pictures of the fabricated devices are shown in Figs. 6(a) and 6(b). The input waveguides of the converters are designed to be cosine S-Bend shape to facilitate fiber coupling. The width $W_{0}$ of input 1 is 2 μm, and the widths 2$W_{0}$ of input 2 and output 2 are 4 μm. The value of $W_{\rm MMI}$ of the 50% and 66% mode converters is 12 μm. The values of $L_{\rm MMI}$ of the 50% and 66% mode converters are 371 μm and 495 μm, respectively. After fabrication, the devices were measured by using a dedicated fiber coupled passive device test setup, as schematically shown in Fig. 7. A tunable laser source working around 1.3 μm was used as the injection light. The output light from the device was divided into two parts by a beam splitter: one was detected by a photo detector, while the other was collected by a charge coupled device (CCD) camera. The images of the output modes distribution of the 50% mode converters are shown in Figs. 8(a) and 8(b). Figures 8(a) and 8(b) correspond to the case when the light is launched into inputs 1 and 2, respectively, while those of the 66% mode converters are shown in Figs. 8(c) and 8(d). It can be seen from Figs. 8(a) and 8(c) that the TE$_{0}$ mode launched into input 1 has been converted to the TE$_{1}$ mode at output 2. On the other hand, the TE$_{0}$ mode launched into input 2 still maintains the TE$_{0}$ mode, as shown in Figs. 8(b) and 8(d). Due to the influence of fabrication error, there is small crosstalk when the light is injected into input 2.
cpl-33-2-024203-fig6.png
Fig. 6. The SEM pictures of the fabricated devices: (a) 50% mode converter and (b) 66% mode converter. Inset: zoom-in view of the MMIs.
cpl-33-2-024203-fig7.png
Fig. 7. The device testing setup for monitoring the field pattern of the output from the mode converter.
cpl-33-2-024203-fig8.png
Fig. 8. The images of the output modes' distribution when the 1.3 μm light was injected into inputs 1 and 2, respectively: (a) and (b) 50% mode converter; (c) and (d) 66% mode converter.
The optical current of the devices under test (DUT) collected by the photo detector are shown in Figs. 9 and 10, respectively, and they are compared with a 3-μm straight waveguide to measure the relative insertion loss (RIL). The purpose of such a comparison is to minimize the influences of the coupling loss, leaving only the differences in the transmission loss and excessive loss between the DUTs and the straight waveguide. The blue and red lines represent the output light from DUT when the light is injected into inputs 1 and 2, respectively, in the wavelength range of 1280–1320 nm, while the dashed pink lines are the output light from the 3-μm straight waveguide. The ripples in the tested curves are due to the interface reflection between the device facets.[14,15] When calculating the RIL, we use the upper envelopes of the transmission spectrum of the DUT in contrast to the 3-μm straight waveguide. The RILs of the 50% and 66% mode converters are shown in Figs. 11 and 12, respectively. Due to the length difference between the straight waveguide and the converters, the resonance peaks and valleys caused by the interface reflection for the converters and the 3-μm straight waveguide are different,[14] causing the output power of the converters sometimes larger than the 3-μm straight waveguide, i.e. RIL$>$0 dB in some cases.
cpl-33-2-024203-fig9.png
Fig. 9. The optical currents of the 50% mode converter and the 3-μm-wide straight waveguide.
cpl-33-2-024203-fig10.png
Fig. 10. The optical currents of the 66% mode converter and the 3-μm-wide straight waveguide.
cpl-33-2-024203-fig11.png
Fig. 11. The relative insertion losses of the 50% mode converters in comparison with the 3-μm-wide straight waveguide.
The calculated RIL when the light is injected into inputs 1 and 2 are below 3.9 dB and 3.7 dB, respectively, for the 50% mode converter, and lower than 1 dB and 3.8 dB, respectively, for the 66% mode converter in the wavelength range of 1280–1320 nm. The optical bandwidth of the converters can be further improved by introducing tapered structure to the input and output waveguides to improve the optical mode overlapping.[15,16]
cpl-33-2-024203-fig12.png
Fig. 12. The relative insertion losses of the 66% mode converter in comparison with the 3-μm-wide straight waveguide.
In conclusion, two-mode converters working around 1.3 μm have been designed through a BPM method and fabricated on InP substrates. Mode converters with 50% and 66% efficiencies are simulated and manufactured. The simulation shows that the excessive losses of the 50% and 66% mode converters are below 0.5 dB and 0.7 dB, respectively, within the wavelength range of 1280–1320 nm. Mode conversion is successfully demonstrated with the fabricated device. The relative insertion losses of TE$_{0}$ modes to TE$_{1}$ mode and TE$_{0}$ mode to TE$_{0}$ mode are below 3.9 dB and 3.7 dB for the 50% mode converter and below 1 dB and 3.8 dB for the 66% mode converter, in the wavelength range of 1280–1320 nm. The 1.3-μm mode converters may find possible applications in mode-division multiplexing systems in the Ethernet systems.
References A design method of a fiber-based mode multi/demultiplexer for mode-division multiplexingUltra-Broadband Tapered Mode-Selective Couplers for Few-Mode Optical Fiber NetworksSPIE ProceedingsOn-chip two-mode division multiplexing using tapered directional coupler-based mode multiplexer and demultiplexerMode multiplexing and demultiplexing devices using multimode Interference couplersTwo-mode de/multiplexer based on multimode interference couplers with a tilted joint as phase shifterHybrid Integrated Platforms for Silicon PhotonicsMultimode interference couplers for the conversion and combining of zero- and first-order modesOptical transceivers for 100 gigabit Ethernet and its transport [100 gigabit ethernet transportOptical multi-mode interference devices based on self-imaging: principles and applicationsHigh precision planar waveguide propagation loss measurement technique using a Fabry-Perot cavity1.3-μm 1 × 4 MMI coupler based on shallow-etched InP ridge waveguidesOptical bandwidth and fabrication tolerances of multimode interference couplers
[1] Saitoh F et al 2010 Opt. Express 18 4709
[2] Riesen N and Love J D 2013 IEEE Photon. Technol. Lett. 25 2501
[3]Koebele C et al 2011 European Conference on Optical Communication (Geneva 18–22 September)
[4]Neo P L et al 2007 OFC/NFOEC (Anaheim 25–29 March)
[5]Hanzawa N et al 2011 OSA/OFC/NFOEC (Los Angeles 6–10 March) OWA4
[6] Li R et al 2011 Proc. SPIE (Shanghai 13–16 November) 8309 83091A
[7] Ding Y et al 2013 Opt. Express 21 10376
[8] Kawaguchi Y and Tsutsumi K 2002 Electron. Lett. 38 1701
[9] Han L et al 2015 Opt. Lett. 40 518
[10] Liang D et al 2010 Materials 3 1782
[11] Leuthold J et al 1998 J. Lightwave Technol. 16 1228
[12] Anderson J and Traverso M 2010 IEEE Commun. Mag. 48 S35
[13] Soldano L B and Pennings E C M 1995 J. Lightwave Technol. 13 615
[14] Feuchter T and Thirstrup C 1994 IEEE Photon. Technol. Lett. 6 1244
[15] Guo F et al 2014 J. Semicond. 35 024012
[16] Besse P A et al 1994 J. Lightwave Technol. 12 1004