Open Access
Add to BibSonomy
Abstract
In this work, we experimentally demonstrate a four-mode polarization/mode insensitive 3-dB coupler based on an adiabatic coupler. The proposed design works for the first two transverse electric (TE) modes and the first two transverse magnetic (TM) modes. Over an optical bandwidth of 70 nm (1500 nm to 1570 nm), the coupler exhibits at most 0.7 dB insertion loss with a maximum crosstalk of -15.7 dB and a power imbalance not worse than 0.9 dB. A multimode photonic switch matrix using this optical coupler is proposed simultaneously exploiting wavelength division multiplexing (WDM), polarization division multiplexing (PDM), and mode division multiplexing (MDM). Based on the coupler experimental measurements, the switching system loss is estimated to be 10.6 dB with crosstalk limited by the MDM (de)multiplexing circuit.
© 2023 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Chip-scale optical interconnects have been considered as a promising solution to alleviate the limited transmission capacity in communication systems. The silicon-on-insulator (SOI) platform has been investigated in recent years as a platform for realizing optical interconnects due to its high refractive index contrast and compatibility with well-established complementary metal-oxide-semiconductor (CMOS) processes. To satisfy the increasing demand for high transmission capacity of optical networks, different multiplexing technologies have been exploited [1,2]. Wavelength division multiplexing (WDM) technology is widely used since it effectively improves the capacity of the optical interconnect. However, WDM requires multiple lasers, which translates into cost, packaging complexity, and power consumption. In recent years, mode division multiplexing (MDM) gained great attention for transmission such as few-mode fiber transmission [3], source-synchronous interconnects [4,5], and optical switches [6–8]. In this multiplexing technique, spatial and orthogonal modes are used as data channels carrying different signals to further address the challenge of Shannon’s limit [9]. Further, polarization division multiplexing (PDM) is an effective method for multiplexing signals carried on electromagnetic waves. This allows for two channels of information to be transmitted on the same carrier frequency by using two optical waves with orthogonal polarization states. Various MDM/PDM silicon photonics (SiPh) devices including mode (de)multiplexer, multimode crossing, polarization beam splitter and coupler, 3-dB couplers, and switches have been reported to realize MDM/PDM optical interconnects [10]. Among these, the multimode 3-dB coupler is of great importance for MDM optical interconnects as it provides the function for splitting as well as combining the multimode input signals. Recently, good performance achievements have been reached for single mode optical 3-dB couplers using multimode interference (MMI) [11], Y-junctions [12], directional couplers (DC) [13,14], and adiabatic couplers (AC) [15]. As for multimode design, mode insensitive 3-dB couplers have been demonstrated using tapered couplers for TE0-4 modes [16] and a directional coupler for the first two transverse magnetic (TM) modes [17]. PDM couplers that work for both TE and TM fundamental modes are also demonstrated using MMI based on subwavelength grating (SWG) structures [18], silicon bent directional couplers [19], and asymmetrical Mach-Zehnder Interferometer (MZI) [20]. However, an optical coupler that works for both MDM and PDM still remains to be proposed and experimentally validated. In [21], a 3-dB splitter that works for the first three TE modes and one TM mode was proposed and validated through simulation based on a 300 nm-thick SOI using a shallow etching process. Furthermore, an MMI-based dual-polarization and mode division multiplexed power splitter for four modes (TE0, TE1, TM0, TM1) was experimentally validated [22]. However, the device is implemented using a pixelated meta structure, which is difficult and expensive to fabricate. Table 1 compares the performance of the reported MDM optical couplers and our work. Our design achieves improvements in insertion loss (IL), crosstalk (XT) and power imbalance (PI) performance. Recently, multimode optical couplers with novel structure using densely packed waveguide array (DPWA) were also experimentally validated [23,24].
Table 1. Performance of reported multimode optical couplera
In this work, a broadband four-mode mode/polarization insensitive 3-dB optical power splitter for the fundamental and first order modes of both TE and TM is experimentally demonstrated. The proposed device is optimized based on an adiabatic coupler structure. The experimental results show that the measured insertion loss is less than 0.7 dB and the crosstalk is at most -15.7 dB within the wavelength range of 1500 nm to 1570 nm for all four modes. We exploit the coupler in a proposed photonic switch matrix supporting four modes and four wavelengths and estimate its loss based on the measured performance of the optical coupler. The rest of this paper is organized as follow. In section 2, the design topology and building blocks are presented and simulated. In section 3, the fabrication and measurement results are discussed. The coupler is used in a proposed multimode optical switch reported in section 4 after which we conclude in section 5.
2. Building blocks and design topology
The proposed four-mode dual polarization 3-dB coupler is based on the design first proposed in [21] with large operating bandwidth and promising simulation performance. In this study, an SOI wafer with a 300 nm-thick silicon core layer is considered as the fabrication platform limiting its validation in commercial 220 nm-thick fabrication processes. Besides, the minimum waveguide gap distance for this structure is 50 nm, which is difficult to achieve because the minimum fabrication limit is controlled above 60 nm. Figure 1(a) illustrates the working principle of the PDM-MDM circuit used with the main building blocks. The PDM-MDM circuit has two input ports (I1 and I2) and two output ports (O1 and O2) and consists of polarization beam combiners (PBC) and splitters (PBS), mode specific multiplexers (Mux) and demultiplexers (DeMux), and the designed 3-dB coupler. At the input ports, mode-specific vertical grating couplers enable either the TE or TM fundamental mode. After being combined through the PBC, both TE and TM fundamental modes propagate in the same waveguide to the multiplexers. The four modes (TE0, TE1, TM0, TM1) are thus generated to evaluate the 3-dB coupler. At the output part of the 3-dB coupler, the demultiplexers and PBS structures convert the higher order modes back to fundamental modes for off-chip photodetection. The polarization beam combiners and splitters (PBC/PBS) and adiabatic coupler–based mode multiplexers are shown in Fig. 1(c-e). In [8], we report the parameters of the TE mode multiplexer while the parameters of the TM mode multiplexer and PBC/PBS are optimized based on the work reported in [24]. The narrowest waveguide gap is specifically narrowed from 0.5 µm to 0.1 µm to match the TE multiplexer’s gap. As TE and TM multiplexers are connected to build an entire PDM-MDM structure. Two multiplexers with different gap structures have different output spacing. Thus, we need to add S-bend waveguides to satisfy the height difference between them. However, if the two gaps are the same, straight waveguides can be used instead of S-bend to reduce the transmission loss. Besides, the mode has better transmission with a smaller gap as it contributes the stronger coupling effect. Afterwards, the device length is optimized for transmission using ANSYS/Lumerical FDTD solutions.
Fig. 1. (a) Schematic of the PDM-MDM circuit; (b) Schematic of the PBC/PBS design; (c) Schematic of the mode multiplexer for TE mode and for (d) TM mode (e) Schematic of the 3-dB coupler design. Figure 1(a-e) are not to scale.
The 3-dB coupler is schematically shown in Fig. 1 with dimensions not to scale. The device is designed for fabrication on a 220 nm waveguide thickness SOI chip. The gap distance between the two input waveguide arms starts at 500 nm (Gapin) to then gradually decreases to 80 nm (GapAC) as the waveguides taper to a width of 1 µm (W3)$.$ The asymmetric adiabatic coupler has different widths for its input with one narrow input waveguide (W1 = 800 nm) and a larger input waveguide (W2 = 1200 nm) to realize the optical power splitting. At the output of the coupler, two S-bends efficiently separate the output ports by 2 mm to avoid further optical field overlap.
The adiabatic coupler part is first optimized by an eigenmode expansion (EME) method using ANSYS/Lumerical MODE solutions. The optimization is based on transmission with respect to propagation length. We optimized the width values (${W_1}$, $\; {W_2},\; {W_3})$ and the adiabatic coupler lengths (${L_1},\; {L_2}$) for the lowest insertion loss and lowest modal crosstalk performance. Figure 2(a) shows the simulated effective index as a function of the waveguide width. To support four modes, ${W_1}$ and ${W_2}$ should be larger than 800 $\textrm{nm}.$ We set the upper limit of ${W_1}$ and ${W_2}$ to 1.2 ${\mathrm{\mu} \mathrm{m}}$ to avoid higher order modes excitation. According to Fig. 2 (a), the effective index of $T{E_0}$ mode saturates and remains constant for lengths greater than 1.5 $\mathrm{\mu }$m even if the waveguide width increases. TM1 mode only emerges after a width greater than 800 $\textrm{n}$m (cut-off condition for TM1) before which its effective index is very similar to that of the cladding $\textrm{Si}{\textrm{O}_2}$ (1.44) since it is not guided. When the waveguide width changes, the effective index of one mode may correspond to another mode contributing to crosstalk. A larger width difference ensures that the effective index difference is sufficiently large to reduce modal crosstalk and coupling between the two waveguides. Thus, ${W_1}$ is set as 0.8 µm and ${W_2}$ is set as 1.2 µm. The width of $Ga{p_{in}}\; $ is selected to be 500 n$\textrm{m}$ providing enough distance avoiding possible coupling between the two input ports. $Ga{p_{AC}}$ should be as small as possible to enhance the coupling and reduce the length of the adiabatic coupler. As such, $Ga{p_{AC}}$ is selected as 80 $\textrm{nm}$ limited by fabrication. $Ga{p_{out}}$ is chosen to be 2 $\mathrm{\mu}\textrm{m}$ which is large enough to avoid coupling between the two output arms. It is worthwhile to note that this design could be used to process higher order modes by increasing the waveguide width (${W_1}$ and ${W_2}$) and the length of AC.
Fig. 2. (a) Simulated effective index of TE and TM modes in a 220 nm thick silicon waveguide as a function of waveguide width; (b) Normalize transmission from ${I_1}$ as a function of length. (c) Normalize transmission from ${I_2}$ as a function of length.
The idea of adiabatic transition in the field of integrated photonics was first defined in [25]. The AC does not require a precisely critical power-transfer length leading to smaller wavelength and polarization dependencies compared to other interference-based couplers. The AC works by converting the input mode in a single waveguide into an even or odd supermode over two waveguides separated by a small gap ($Ga{p_{AC}}$). The modes from narrow waveguide input (${W_1}$) produce odd supermodes, while the modes from wide waveguide ${W_2}$ input generate even supermodes. The phase condition, as seen in Fig. 3, is what distinguishes even from odd modes. Two in-phase modes that are respectively contained in the two waveguides make up even supermodes. While odd supermodes have two antiphase components. From the adiabatic tapering process, therefore, one input mode can be fully converted into its corresponding supermodes at the output ports. Figure 2(b,c) show power transmission from ${I_1}$ and ${I_2}$ individually as a function of the AC length. It can be observed that when the device length is close to 0, almost all the power from ${I_1}$ would propagate into ${O_1}$. As the length increases, the transmission of ${O_1}$ decreases while the power of ${O_2}$ increases with both finally converging to 0.5, evenly. A similar process for ${I_2}$ input occurs which is shown in Fig. 2(c). The energy excited from ${I_2}$ is transferred from ${O_2}$ to ${O_1}$, and equally distributed as the length increases. Note that the fundamental TE mode requires the longest coupling length, as it is confined with weaker coupling efficiency. The length of the AC is optimized to 1.6$\; \textrm{mm}$ with a tradeoff between the device length and conversion efficiency. Similarly, ${L_2}\; $is optimized to 60 µm for better transmission with S-bend radius R of $1\; \textrm{mm}$ to reduce bending loss and mode crosstalk.
Fig. 3. Mode propagations in the 3-dB coupler device for different modes launched at a wavelength of 1550 nm with the corresponding total field as a percentile. The cross-sectional fields at the output are shown on the right of each figure.
The simulated mode propagations for the overall structure of the four modes are shown in Fig. 3 with their transmission percentage. All four input modes are split into two output ports with a nearly 50% power splitting ratio realizing the 3-dB power splitting. The corresponding mode profiles at the output ends of the coupler are at the right of each simulation in this figure. Insertion loss (IL), modal crosstalk (XT), and power imbalance (PI) are three main parameters for MDM system evaluation and can be calculated as Eqs. (1)-(3) [21]:
3. Experimental results and discussion
The design was fabricated by Applied Nanotools Inc. (ANT). The silicon device layer is patterned using a 100 KeV electron-beam lithography (EBL) followed by an inductively coupled plasma-induced reactive ion etching (ICP-RIE) process. A 2.2 $\mathrm{\mu}\textrm{m}$ thick SiO2 cladding is deposited by plasma-enhanced chemical vapor deposition (PECVD). An optical microscopic image is shown in Fig. 4 labelled with all the corresponding structures including 16 grating couplers (GCs), two PBCs and two PBSs for polarization conversion, four (de)multiplexs for higher mode propagation, and a 3-dB coupler in the middle. For the continuous (CW) measurement, a tunable C-band laser is first regulated to a specific polarization status (TE or TM) by an off-chip polarization controller (PC) and injected into the chip via one GC. On the output side, light is first collected by another GC to a fiber. The optical output is measured by an optical power meter. Two different GCs were used to excite TE or TM modes since GCs are polarization sensitive. To experimentally test both TE and TM modes simultaneously, test structures corresponding to all the building blocks, i.e., the PBC/PBS and the (de)multiplexers for TE and TM modes, are first measured to obtain their inherent insertion loss and modal crosstalk performance. The worst measured values for these test structures for their IL and modal crosstalk are 0.9 dB and -15.4 dB, respectively, over the wavelength range from 1500 nm to 1570 nm. As shown in Fig. 4, when light is launched from any one of the input ports (GCs labeled 1-8 in blue colour), $T{M_1}$ , $T{E_1}$ , $T{M_0}$ , $T{E_0}$ , $T{M_0}$ , $T{E_0}$ , $T{M_1}$ , $T{E_1}$ are excited, respectively. Similarly, for the output ports (GCs labeled 1-8 in red colour). $T{M_1}$ , $T{E_1}$ , $T{M_0}$ , $T{E_0}$ , $T{M_0}$ , $T{E_0}$ , $T{M_1}$ , $T{E_1}$ are detected by the optical power meter, respectively. We can only excite one input port and one output port at a time in the measurement. When we choose the same mode order ($Input_{GC\; 1} \to $ $Output_{GC\; 1}, {{I_1} - {O_1}\; :T{M_1}{\; } - {\; }T{M_1}}$ for example), that is how we evaluate the IL performance. After normalizing to the$\; {I_1} - {O_1}\; :T{M_1}{\; } - {\; }T{M_{1\; \; }}$transmission curve of the test structure, the IL values are obtained (Fig. 5). When the mode orders are different between the input and output ports ($Inpu{t_{GC\; 1}} \to \; $ $Outpu{t_{GC\; 2\; to\; 4\; }}, {{I_1} - {O_1}\; :T{M_1}{\; } - {\; }T{E_{1}}, \; {\; }T{M_1}{\; } - {\; }T{M_{0}}, \; T{M_1}{\; } - {\; }T{E_0}} )$, the measurement curves indicate the XT performance as well, i.e., the transmission from one optical mode to another optical mode. These transmission curves are normalized to the input mode straight-through case (${I_1} - {O_1}\; :T{M_1}{\; } - {\; }T{M_1})\; $as a reference to show the XT value. The XT is labeled with a vertical arrow in the first spectrum of Fig. 6 for clarification.
Figure 5 shows the transmission curves of the 3 dB splitting. Measurement results below 1570 nm are presented as the measurement spectra showed prominent Fabry-Perot (FP) fringes for wavelength larger than 1570 nm. These FP fringes come from the cavity generated by in-waveguide reflection between the two grating couplers which was also observed by our previous measurement [26]. It is worth noting that the grating coupler used in this work is provided by the ANT PDK [27]. According to the test results, the 3-dB bandwidths of the GCs are 30.6 nm (TE) and 47.5 nm (TM) around the central wavelength of 1550 nm. Thus, we have used a moving average smooth filter algorithm (Polyfit in MATLAB) to obtain an outline of the test results. The IL and PI performance are calculated based on these Polyfit curves.
Figure 6 shows the corresponding measured and normalized spectra for the four-mode inputs. The MDM system is a two input ports $\textrm{by}$ two output ports ($2 \times 2$). Thus, for each mode, there are four diagrams showing the transmission from 1) the upper input to the upper output (I1→O1), 2) upper input to the lower output (I1→O2), 3) lower input to the upper output (I2→O1), and 4) the lower input to the lower output (I2→O2).
The experimental results of the optical coupler show that the measured IL for the four modes varies from 0.06 dB to 0.68 dB. The largest IL comes from the $T{M_1}$ mode (I1→ O2). This could be explained by the fact that the width of the AC input is relatively small and close to the cut off condition for $T{M_1}$ mode. The XTs for all four modes are within the range of -31.7 dB to -15.7 dB over the 70 nm optical bandwidth range. The largest XT is between $T{E_0}$ and $T{E_1}$. That is mainly because these two modes have similar effective indices.
According to the measurement results shown in Fig. 5, the power imbalance caused by the asymmetry of the AC is also significant and should be noticed. Table 2 summarizes the experimental results for the 3-dB coupler within 70 nm of bandwidth range from 1500 nm to 1570 nm. To implement an on-chip MDM system interfacing with off-chip single mode optical fiber, the (de)multiplexer is an essential part to excite higher order modes. As such, we cannot measure the XT of the 3-dB coupler without it. The experimental structure shows that the whole MDM circuit has a similar XT level (-15.7 dB) as the test structure (-15.4 dB). Therefore, it can be concluded that the XT performance of the MDM is mainly affected by the (de)multiplexers. By further optimizing the multiplexing building blocks, the crosstalk performance of MDM circuits can largely be improved.
Table 2. Experimental results of the optical coupler with worst results in bold characters
Considering the possible performance degradation caused by fabrication variations, the fabrication tolerance of the designed optical coupler is investigated. In simulation, only $Ga{p_{AC}}$ is changed and the other parameters remain constant. In reality, all parameters are affected by fabrication tolerance. While not exhaustive of an investigation, four optical couplers with widths ${\pm} {\; }$20 nm from the originally designed widths (${\textrm{W}_1}$= 0.8 µm, ${\textrm{W}_2}$= 1.2 µm, ${\textrm{W}_3}$= 1.0 µm, first row of Table 3) are measured to assess the coupler’s tolerance to fabrication variations. The experimental results of these structures are compared in Table 3 over 70 nm of optical bandwidth range from 1500 nm to 1570 nm. The worst results occur for a narrower width (last row in bold). A fabrication error leading to a narrower width by up to 20 nm results in a 0.39 dB increase in IL (from 0.68 dB to 1.07 dB), 1.4 dB increase in XT (from -15.66 dB to -13.92 dB), and 0.41 dB increase in PI (from 0.91 dB to 1.32 dB). This worst degradation is primarily due to the waveguide width being insufficiently large to support a$\; $stable transmission for $\textrm{T}{\textrm{M}_1}$, particularly when ${\textrm{W}_1}$ is lower than 0.8 µm.
Table 3. Experimental results of the optical coupler with varying waveguide widths
4. Proposed multimode switch
In this section, a multimode photonic switch matrix is developed that simultaneously exploits WDM, PDM, and MDM. The 3-dB coupler presented in section 3 serves as one of the MDM building blocks for a four-mode by four coarse wavelength division multiplexed (CWDM) optical switching system, as shown in Fig. 7. The switching system accommodates four modes with different mode sequences and polarization states ($T{E_0},T{E_1},T{M_0},T{M_1}$). Four CWDM wavelengths spaced by 18 nm in the wavelength range from 1500 nm to 1570 nm. The WDM-PDM-MDM switching circuit is based on an 8$\; \times \; $8 banyan topology validated by our group [28]. The 8 × 8 optical switch matrix consists of 12 Mach-Zehnder interferometer (MZI) switches (${S_1}$ to ${S_{12}}$), each utilizing two 3-dB couplers along with a mode/polarization insensitive phase shifter as designed. The design of PDM-MDM phase shifter is still under investigation. It is worth noting that an MDM switch has been proposed and experimentally demonstrated by our group using a TE mode insensitive phase shifter [8]. As for the realization of a polarization/mode insensitive design, two possible methods are worth investigating. First, a SWG based structure in the phase shifter has been validated by our group [29]. For the same polarization state (either TE or TM), the thermos-optic coefficient (dneff/dT) of different mode orders converge to a specific value as the waveguide width increases [8]. However, the converging value differs between TE and TM modes. Specifically, TM thermo-optic coefficient is smaller than TE modes because TM modes have more overlap with the oxide which has a thermo-optic coefficient one order of magnitude smaller than Silicon. By implementing the SWG structure and adjusting its parameters, it is possible to decrease dneff/dT for TE modes and increase the value of TM modes to converge to a specific point realizing a polarization-mode insensitive phase shifter. Other designs were proposed such as adding a square SOI nanophotonic waveguide pad to achieve polarization-insensitive phase delay when heated [30]. The second method consists of separating the TE and TM modes as suggested in a novel polarization insensitive phase shifter reported in [31]. The phase shifter generates a phase difference between the two interfering signals, enabling thermally changing states. When the phase shifter is disabled, the optical switch is in the bar state (Fig. 7(a)). In the bar state, the two inputs are directly connected to two corresponding outputs. When the phase shifter is on (i.e., applying a π phase shift), the optical switch is in the cross state (Fig. 7(b)), each input is routed to the opposite output. By controlling the state of each MZI optical switch, the space switching function for the four modes can be realized.
Fig. 7. (a) Bar state of the optical switch; (b) Cross state of the optical switch; (c) Schematic structure of the four-mode optical switch in bar state; (d) Schematic structure of four-mode optical switch in cross state; (e) Schematic structure of the proposed four modes by four CWDM optical switching system.
The estimated IL varies for different optical paths and depends on the amount of waveguide crossings and state of the switch. Each routing path contains three stages (MZI elements) to realize the optical switching process. Stage 1 consists of four MZI structures labeled ${S_1}$ to ${S_4}$. Without any design change in the 3-dB coupler, the estimated footprint of the switching system is 16 ${\times} $ 7400 $\mathrm{\mu }{\textrm{m}^2}$, making its implementation challenging. By shortening the PDM-MDM 3-dB coupler from 1.6 ${\mathrm{\mu} \mathrm{m}}$ to 1 ${\mathrm{\mu} \mathrm{m}}$, the overall size of the switch can be reduced to 16 ${\times} $ 6300 $\mathrm{\mu }{\textrm{m}^2}$ which consists of $4 \times 3$ MZI structures (3.88 ${\times} $ 2100 $\mathrm{\mu }{\textrm{m}^2}$). The experimental measured IL of 1 ${\mathrm{\mu} \mathrm{m}}$ length 3-dB coupler is 1 dB. The total IL of the MZI structure is thus estimated to be 2.2 dB. This accounts for a phase shifter loss of 0.2 dB but mainly comes from the two 3-dB couplers (2 ${\times} $ 1 dB). Stage 2 also consists of four MZI structures (${S_5}$ to ${S_8}$). The propagation loss of multimode crossings accommodating four modes have at most 0.7 dB IL and a crosstalk of less than -35 dB [32]. The MDM-WDM-PDM optical signal coming from Stage 2 propagates through the final stage with four MZIs labeled ${S_9}$ to ${S_{12}}$. IL is the smallest for the shortest path (${I_0} - {O_0},{I_8} - {O_8}$), with no MZI in crossbar state and no crossing. The worst case for IL is the optical paths which include three MZIs in cross state (${I_0} - {O_7},{I_7} - {O_0}$). As such, the largest IL of the 8 × 8 optical switching system is estimated to be 10.6 dB with four waveguide crossings employed. It is estimated that the switch XT would be limited by the (de)multiplexers. Table 4 shows the approximate value of the optical loss estimated from the building blocks. A WDM-PDM-MDM system has been experimentally validated and reported in [33]. The system has a maximum IL of 12.8 dB and XT of -11.3 dB over 40 nm bandwidth. Besides, another WDM-PDM-MDM system with the help of microring (MRR) is also measured to exhibit 9 dB IL and -15.6 dB XT [34]. Our system is promising to achieve comparable IL and XT performance.
Table 4. Optical loss estimation of the 8${\; } \times {\; }$8 switching system
To improve the overall performance of the system, the next step is to reduce the crosstalk of the multiplexer in the system. There are two main optimization methods. The first one is to keep the existing and conventional multiplexer structure and perform gradient descent deep learning optimization on the main parameters in the result. The gradient descent algorithm can be iterated by the computer to find the optimized parameter value that corresponds to the best transmission in the specific range [35]. The second method is to design the device by using pixelated waveguides as in [36] for a mode multiplexer with -20 dB XT performance. There are still limitations for implementing the WDM-PDM-MDM system, such as the large size which remains a concern. Inverse design is a method that has been developed in recent years to achieve high density and high performance integrated silicon-based devices through the use of multiple algorithms. For multiplexers, inverse design can effectively reduce the size of integrated silicon-based devices while maintaining a high level of performance [37].
5. Conclusion
In summary, we propose and experimentally demonstrate a mode and polarization insensitive four-mode 3-dB coupler for the MDM system. The four-mode 3-dB coupler consists of an adiabatic coupler. The footprint of the designed structure is 3.88 ${\mathrm{\mu} \mathrm{m}}\; \times $ 1660 ${\mathrm{\mu} \mathrm{m}}$. The measurement shows low insertion loss which is less than 0.7 dB and low modal crosstalk which is less than -15.7 dB over 70 nm bandwidth ranging from 1500 nm to 1570 nm. Power imbalance of the device is less than 0.9 dB. The device is optimized for $T{E_0},T{E_1},T{M_0},T{M_1}$ and has a significant measurement performance improvement over past published works. By further optimizing the mode (de)multiplexers, the modal crosstalk can be improved. We believe such a device could find its applications in high-performance WDM-PDM-MDM compatible multimode optical switch interconnects with a maximum 10.6 dB insertion loss.
Funding
Huawei Technologies Canada, Co., Ltd.
Acknowledgments
The authors thank Ajaypal Singh Dhillon, Sunami Morrison, and José Garcia Echeverria for their support through technical discussions while conducting this research.
Disclosures
The authors declare that there are no conflicts of interest related to this article.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. D. Zhou, C. Sun, Y. Lai, Y. Yu, and X. Zhang, “Integrated silicon multifunctional mode-division multiplexing system,” Opt. Express 27(8), 10798–10805 (2019). [CrossRef]
2. Y. Shi, Y. Zhang, Y. Wan, Y. Yu, Y. Zhang, X. Hu, X. Xiao, H. Xu, and B. Pan, “Silicon photonics for high-capacity data communications,,” Photon. Res. 10(9), A106–A134 (2022). [CrossRef]
3. S. Jiang, L. Ma, Z. Zhang, X. Xu, S. Wang, J. Du, C. Yang, W. Tong, and Z. He, “Design and characterization of ring-assisted few-mode fibers for weakly coupled mode-division multiplexing transmission,” J. Lightwave Technol. 36(23), 5547–5555 (2018). [CrossRef]
4. C. Williams, B. Banan, G. Cowan, and O. Liboiron-Ladouceur, “A source-synchronous architecture using mode-division multiplexing for on-chip silicon photonic interconnects,” IEEE J. Sel. Top. Quantum Electron. 22(6), 473–481 (2016). [CrossRef]
5. C. Williams, D. Abdelrahman, X. Jia, A. I. Abbas, O. Liboiron-Ladouceur, and G. E. R. Cowan, “Reconfiguration in source-synchronous receivers for short-reach parallel optical links,” IEEE Trans. VLSI Syst. 27(7), 1548–1560 (2019). [CrossRef]
6. P. Guo, W. Hou, L. Guo, Z. Ning, M. S. Obaidat, and W. Liu, “WDM-MDM silicon-based optical switching for data center networks,” ICC IEEE International Conference on Communications (IEEE, 2019), pp. 1–6.
7. Y. Zhang, R. Zhang, Q Zhu, Y. Yuan, and Y. Su, “Architecture and devices for silicon photonic switching in wavelength, polarization and mode,” J. Lightwave Technol. 38(2), 215–225 (2020). [CrossRef]
8. G. Zhang, H. R. Mojaver, A. Das, and O. Liboiron-Ladouceur, “Mode insensitive switch for on-chip interconnect mode division multiplexing systems,” Opt. Lett. 45(4), 811–814 (2020). [CrossRef]
9. G. Li and X. Liu, “Focus issue: space multiplexed optical transmission,” Opt. Express 19(17), 16574–16575 (2011). [CrossRef]
10. C. Li, D. Liu, and D. Dai, “Multimode silicon photonics,” Nanophotonics 8(2), 227–247 (2018). [CrossRef]
11. I. Kim, J. Y. Kim, J. Yoon, H. Yoon, H. H. Park, and H. Kurt, “Experimental demonstration of inverse-designed silicon integrated photonic power splitters,” Nanophotonics 11(20), 4581–4590 (2022). [CrossRef]
12. Y. Zhang, S. Yang, A. E. Lim, G. Q. Lo, C. Galland, T. Baehr-Jones, and M. Hochberg, “A compact and low loss Y-junction for submicron silicon waveguide,” Opt. Express 21(1), 1310–1316 (2013). [CrossRef]
13. S. Zhao, W. Liu, J. Chen, Z. Ding, and Y. Shi, “Broadband arbitrary ratio power splitters based on directional couplers with subwavelength structure,” IEEE Photonics Technol. Lett. 33(10), 479–482 (2021). [CrossRef]
14. C. Ye and D. Dai, “Ultra-compact broadband 2× 2 3 dB power splitter using a subwavelength-grating-assisted asymmetric directional coupler,” J. Lightwave Technol. 38(8), 2370–2375 (2020). [CrossRef]
15. D. Mao, Y. Wang, E. EI-Fiky, L. Xu, A. Kumar, M. Jaques, A. Samani, O. Carpentier, S. Bernal, M. S. Alam, J. Zhang, M. Zhu, P. C. Koh, and D. V. Plant, “Adiabatic coupler with design-intended splitting ratio,” J. Lightwave Technol. 37(24), 6147–6155 (2019). [CrossRef]
16. W. Chen, J. Lin, H. Li, P. Wang, S. Dai, Y. Liu, R. Yao, J. Li, Q. Fu, T. Dai, and J. Yang, “Broadband multimode 3 dB optical power splitter using tapered couplers,” Opt. Express 30(26), 46236–46247 (2022). [CrossRef]
17. Y. Luo, Y. Yu, M. Ye, C. Sun, and X. Zhang, “Integrated dual-mode 3 dB power coupler based on tapered directional coupler,” Sci. Rep. 6(1), 23516 (2016). [CrossRef]
18. S. Han, W. Liu, and Y. Shi, “Ultra-broadband dual-polarization power splitter based on silicon subwavelength gratings,” IEEE Photonics Technol. Lett. 33(15), 765–768 (2021). [CrossRef]
19. X. Chen, W. Liu, Y. Zhang, and Y. Shi, “Polarization-insensitive broadband 2 × 2 3 dB power splitter based on silicon-bent directional couplers,” Opt. Letters 42(19), 3738–3740 (2017). [CrossRef]
20. D. Dai, Z. Wang, J. Peters, and J. E. Bowers, “Compact Polarization Beam Splitter Using an Asymmetrical Mach–Zehnder Interferometer Based on Silicon-on-Insulator Waveguides,” IEEE Photonics Technol. Lett. 24(8), 673–675 (2012). [CrossRef]
21. L. Han, B. P.-P. Kuo, N. Alic, and S. Radic, “Ultra-broadband multimode 3 dB optical power splitter using an adiabatic coupler and a Y-branch,” Opt. Express 26(11), 14800–14809 (2018). [CrossRef]
22. Y. Liu, Z. Wang, Y. Liu, X. Wang, X. Guo, D. Li, Y. Yao, Q. Song, J. Du, Z. He, and K. Xu, “Ultra-compact mode-division multiplexed photonic integrated circuit for dual polarizations,” J. Lightwave Technol. 39(18), 5925–5932 (2021). [CrossRef]
23. K. Chen, Z. Nong, J. Zhang, X. Cai, S. He, and L. Liu, “Multimode 3 dB coupler based on symmetrically coupled waveguides for on-chipbrk mode division multiplexing,” J. Lightwave Technol. 35(19), 4260–4267 (2017). [CrossRef]
24. D. Dai, C. Li, S. Wang, H. Wu, Y. Shi, Z. Wu, S. Gao, T. Dai, H. Yu, and H. Tsang, “10-Channel Mode (de)multiplexer with Dual Polarizations,” Laser Photonics Rev. 12(1), 1700109 (2018). [CrossRef]
25. T. A. Ramadan and R. M. Osgood, “Adiabatic couplers: design rules and optimization,” J. Lightwave Technol. 16(2), 277–283 (1998). [CrossRef]
26. T. V. Vaerenbergh, P. Sun, S. Hooten, M. Jain, Q. Wilmart, A. Seyedi, and R. Beausoleil, “Wafer-level testing of inverse-designed and adjoint-inspired vertical grating coupler designs compatible with DUV lithography,” Opt. Express 29(23), 37021–37036 (2021). [CrossRef]
27. Y. Wang, X. Wang, J. Flueckiger, H. Yun, W. Shi, R. Bojko, N. A. F. Jaeger, and L. Chrostowski, “Focusing sub-wavelength grating couplers with low back reflections for rapid prototyping of silicon photonic circuits,” Opt. Express 22(17), 20652–20662 (2014). [CrossRef]
28. A. Das, H. R. Mojaver, G. Zhang, and O. Liboiron-Ladouceur, “Scalable SiPh-InP Hybrid Switch Based on Low-Loss Building Blocks for Lossless Operation,” IEEE Photonics Technol. Lett. 32(21), 1401–1404 (2020). [CrossRef]
29. K. H. R. Mojaver, G. Zhang, and O. Liboiron-Ladouceur, “Design of Silicon Photonic Mode-Sensitive Thermo-Optic Phase Shifter based on Subwavelength Grating Structures,” IEEE Photonics Conference (IPC), 1–2 (2022).
30. S. Wang and D. Dai, “Polarization-insensitive 2× 2 thermso-osptic Mach–Zehnder switch on silicon,” Opt. Lett. 43(11), 2531–2534 (2018). [CrossRef]
31. K. Suzuki, T. Yamada, O. Moriwaki, H. Takahashi, and M. Okuno, “Polarization-insensitive MZI switch composed of an LN phase shifter array and silica-based PLC-integrated polarization beam splitter,” Optical Fiber Communication Conference, p. OWI7 (2008).
32. S. Wu, X. Mu, L. Cheng, S. Mao, and H. Fu, “State-of-the-art and perspectives on silicon waveguide crossings: A review,” Micromachines 11(3), 326 (2020). [CrossRef]
33. Y. Zhang, R. Zhang, Q. Zhu, Y. Yuan, and Y. Su, “Architecture and. Devices for Silicon Photonic Switching in Wavelength, Polarization and Mode,” J. Lightwave Technol. 38(2), 215–225 (2020). [CrossRef]
34. X. Cao, S. Zheng, N. Zhou, J. Zhang, and J. Wang, “On-Chip Multi-Dimensional 1 × 4 Selective Switch With Simultaneous Mode-/Polarization-/Wavelength-Division Multiplexing,” IEEE J. Quantum Electron. 56(5), 1–8 (2020). [CrossRef]
35. Z. Lin, Y. Wen, Y. Chen, Y. Zhang, and S. Yu, “Inverse design of orbital angular momentum mode demultiplexer by combining wavefront matching method and gradient descent algorithm,” 2019 Conference on Lasers and Electro-Optics Europe & European Quantum Electronics Conference (CLEO/Europe-EQEC) (2019).
36. H. Xie, Y. Liu, S. Wang, Y. Yao, Q. Song, J. Du, Z. He, and K. Xu, “Highly Compact and Efficient Four-Mode Multiplexer Based on Pixelated Waveguides,” IEEE Photonics Technol. Lett. 32(3), 166–169 (2020). [CrossRef]
37. S. Molesky, Z. Lin, A. Y. Piggott, W. Jin, J. Vuckovíc, and A. W. Rodriguez, “Inverse design in nanophotonics,” Nat. Photonics 12(11), 659–670 (2018). [CrossRef]
