Abstract
By leveraging mode-division multiplexing (MDM), capacity of on-chip photonic interconnects can be scaled up to an unprecedented level. The demand for dynamic control of mode carriers has led to the development of mode-division multiplexing switches (MDMS), yet the conventional MDMS is incapable of directly accessing an individual lower-order mode that propagates in a multi-mode bus waveguide, which hinders its scalability and flexibility. In this paper, we propose and demonstrate the first direct-access MDMS as a novel platform for scalable on-chip multi-mode networks. At first, the highly efficient mode exchangers are developed for TE0–TE2 and TE1–TE2 mode swap, which are then employed to realize the direct-access mode add-drop multiplexers with high performances. The direct-access MDMS is then achieved based on the proposed mode add-drop multiplexers, which can be used for dynamically adding and dropping any selected mode carrier in a three-channel MDM. Moreover, the novel direct-access scheme is also adopted to simultaneously harness wavelength and mode carriers, leading to a wavelength/mode-hybrid multiplexing system with an enhanced link capacity of twelve channels. To further verify the utility of the MDMS, a multi-mode hubbed-ring network is constructed, where one hub and three nodes are organized within a ring-like multi-mode bus waveguide. The reconfigurable network traffic of 6 × 10 Gbps data streams are obtained by using three eigen modes as signal carriers. The measurement results show low bit-error rates (<10−9) with low power penalties (<3.1 dB).
1 Introduction
The multi-core architecture has become the mainstream in improving performances of modern micro-processors by assembling numerous parallel computational cores on a single chip [1]. In the meantime, the continuous drive towards a highly parallel system has led to the ever-growing bandwidth demand in dense intra-chip data communications and complex on-chip network traffics. Nevertheless, the commonly used on-chip electric interconnects face critical challenges in fulfilling the stringent bandwidth requirement, due to severe power dissipation and significant latency penalties at high frequencies [2]. Recently, on-chip photonic interconnects have emerged as a potential approach to break the performance ceiling, which is promised by the intrinsically large bandwidth and ultra-low propagation losses [3]. Moreover, the parallelism offered by multi-dimensional multiplexing enables the simultaneous transmission of multiple data streams in a single waveguide, so that the link capacity can be efficiently scaled up. The wavelength-division multiplexing (WDM) is to utilize different wavelengths as signal carriers [4, 5]. The WDM technology has been extensively used, especially for long-haul optical communications. However, WDM systems usually require a great number of laser diodes working simultaneously, and it is essential to monitor each wavelength carrier and compensate the wavelength drift caused by thermal effect, which gives rise to power consumption and therefore hinders its applications in power-sensitive on-chip operations.
The power-efficient on-chip photonic interconnects can be realized by adopting mode-division multiplexing (MDM) to boost the capacity within each single-wavelength carrier [6]. The MDM is to use different eigen modes as signal carriers, and the major advantage of MDM is its excellent scalability, since the channel number of MDM can be increased by simply choosing a wider multi-mode waveguide that supports more eigen modes. Over the recent years, a variety of devices have been reported to facilitate the practical use of MDM, e.g. mode add-drop multiplexers (MADM) [7], [8], [9], multi-mode bends [10], [11], [12], multi-mode crossings [13], [14], [15], multi-mode splitters [16], [17], [18] and mode-division multiplexing switches (MDMS) [19], [20], [21], [22], [23], [24], [25], [26], [27]. Among them, the MDMS is a key element in MDM, which can be used to add/drop any desired eigen modes and re-allocate them at will, allowing a dynamic manipulation of mode carriers in multi-mode networks with tremendous flexibility. Various MDMSs have been demonstrated in the previous works by leveraging Y-junctions [19], [20], [21], asymmetric directional couplers (ADC) [22], [23], [24] and asymmetric micro-ring resonators (MRR) [25], [26], [27]. Usually, an MDMS is achieved by combining an MADM with a single-mode switch. However, the conventional MADM cannot individually add/drop a single lower-order mode in the multi-mode waveguide unless all the higher-order modes have been demultiplexed, as will be discussed in Section 2. Consequently, the conventional MDMS cannot select an individual mode carrier in a “direct-access” manner; in other words, some unwanted mode carriers also have to be accessed during switching operations. This will lead to a vastly increased device footprint and layout complexity especially when the channel number is large, as will be discussed in Section 2. Up to now, such “direct-access” conundrum has not been properly addressed for all the reported MDMSs.
In this paper, we propose and demonstrate the first direct-access MDMS for scalable on-chip multi-mode networks. We firstly develop two types of novel mode exchangers (MEs) for swapping lower-order modes (i.e. TE0 and TE1) with the highest-order mode (i.e. TE2). The proposed MEs are then employed to implement the direct-access MADM that is capable of directly coupling arbitrary eigen modes to/from a multi-mode bus waveguide. After that, the direct-access MDMS is realized by combining the proposed MADMs with Mach–Zehnder switches. Furthermore, the proposed MADMs are also combined with tunable MRRs to manipulate wavelength/mode-hybrid carriers to further enhance the capacity. The reconfigurable multi-mode hubbed-ring network (HRN) is then developed to further prove the utility of the MDMS. For the proposed HRN, multiple eigen modes (i.e. TE0–2) rather than multiple wavelengths are used as carriers to organize network traffics with a ring-like topology, based on which, reconfigurable on-chip photonic interconnects between one hub and three nodes can be realized.
2 Design, results and discussion
The direct-access MADM is the footstone for realizing the MDMS. Therefore, it is worthwhile to further explain the “direct-access” conundrum in MADM to clearly show the context of this work. For a multiplexing system, “direct-access” means that only the selected carrier will be added and dropped while the other carriers will pass through without (de)multiplexing. For WDM, this functionality can be quite easy to realize by leveraging optical notch filters, e.g. MRR. However, the scenario is completely different for MDM. A three-channel MDM is discussed here as a proof of concept. The first three TE modes (i.e. TE0–2) are employed as carriers, and the corresponding signal channels are denoted as CH0-2. Figure 1A illustrates the schematic of conventional MADM. It can be seen that, to add/drop CH0 (carried by TE0), all the other higher-order modes (i.e. TE1–2) must be sequentially demultiplexed by ADC2 and ADC1 and bridged to the other two mirrored ADCs. Consequently, the conventional MADM fails to work in a “direct-access” manner, since idle carriers (i.e. TE1–2) are also (de)multiplexed. The underlying cause of this issue is two-fold. Firstly, the mode selectivity of ADC solely relies on the phase-matching between adjacent waveguides [8]. We calculate effective indices of TE0–2 modes in a silicon-on-insulator (SOI) waveguide with varied widths and a fixed thickness of H wg = 0.22 μm by using the finite-element method (FEM) (see Figure 1C). For a multi-mode waveguide with W wg = 1.2 μm, one can set the access waveguide widths as W wg = 0.37 μm and W wg = 0.58 μm to achieve selective TE2-TE0 and TE1–TE0 coupling (see the dash lines in Figure 1C). However, the complete coupling of TE0 requires the same widths between multi-mode and access waveguides. As a result, the undesired TE1–2 modes will also be coupled. Secondly, even if the phase-matching condition can be satisfied, it is still challenging to extract lower-order modes (e.g. TE1) with ADCs due to weak coupling strength. Figure 1D shows the calculated field profiles for TE0–2 with H wg = 0.22 μm and W wg = 1.2 μm. One can see that, compared with TE2, the evanescent field is much weaker for TE0-1. The coupling strength can be enhanced by choosing a narrower multi-mode waveguide, but higher-order modes will be cut off at a smaller width.
Illustration of “direct-access” conundrum in mode add-drop multiplexers (MADM). (A) The schematic configuration of conventional MADM. (B) The schematic configuration of direct-access MADM. The inset illustrates the mode-swap process in a mode exchanger (ME). (C) The calculated effective indices (n eff) of TE0–2 modes as functions of waveguide width (W wg). The inset shows a multi-mode waveguide cross-section. (D) The calculated field profiles for TE0–2 modes in a multi-mode waveguide with H wg = 0.22 μm and W wg = 1.2 μm. The dashed boxes show the enlarged field profiles at waveguide edges. ADC, asymmetric directional coupler.
Here, an ME-based configuration is proposed to solve this “direct-access” conundrum, as schematically shown in Figure 1B. The add-drop of CH0 is considered as an example. The ME is capable of swapping mode carriers between two signal channels. The signal-carrier mapping relation is shown in the inset of Figure 1B. At the input port, the carrier for CH0 is transformed from TE0 to TE2, while the carrier for CH2 is transformed from TE2 to TE0. Thus, the incident CH0 can be directly dropped by ADC2 through a TE2–TE0 coupling. The carrier for CH2 is reverted from TE0 to TE2 by another identical ME at output port. On the other hand, the launched TE1 mode will pass through ME without any change, and thus CH1 (carried by TE1) will go straight into the output port. For the adding operation, the CH0 signal is firstly loaded onto TE2 mode by ADC2 through a TE0−TE2 coupling. The carrier for CH0 is then converted from TE2 to TE0 by the output ME. For such a design, CH0 can be directly added or dropped without demultiplexing the mode carriers for CH1–2. Furthermore, the mapping relation between CH i and TE i can be perfectly maintained at the output port. Here, we take the add-drop of CH0 as an example, but this method should work for all the involved signal channels. The prime advantage for this design is its superior scalability and flexibility. In a conventional MADM, the number of ADCs can be expressed as:
where N ADC is the required number of ADCs, C is the largest mode order used in the MDM system, and M is the mode order of carrier to be added or dropped. It is clear that the ADC number will rapidly increase with the channel number of MDM especially when a lower-order mode needs to be added or dropped from a large capacity. In contrast, for the present direct-access MADM, only two ADCs (N ADC = 2) and two MEs are required, no matter how large the channel number is, and thus the system can be easily scaled up. Moreover, this scheme will not introduce additional ADCs (i.e. N ADC = 2C) even if it is necessary to add/drop all the mode carriers. In the following subsections, we will present the realization of MDMS and relevant devices (i.e. ME and MADM) based on the proposed “direct-access” scheme. The fabrication and characterization methods for these devices can be found in Supporting Information, Sections S1 and S2.
2.1 Mode exchanger
The previously reported MEs are usually based on free-geometric or sub-wavelength structures with a complex design flow and a lack of universality [28], [29], [30], [31]. Here, a simple and general design methodology is present for high-performance MEs, as schematically illustrated in Figure 2A. The proposed structures are based on long-period waveguide gratings (LPWGs). Two different types of MEs are discussed: ME02 for TE0–TE2 mode swap and ME12 for TE1–TE2 mode swap, as illustrated in the insets of Figure 2A. The sinusoidal functions are utilized to define the shape of LPWGs. The grating structures are applied on both side-walls of ME02 since TE0 and TE2 are both symmetric modes. In contrast, for ME12, width variations are applied only on a single side-wall and the trajectory of other side-wall is straight and uniform since symmetries of TE1 and TE2 are reversed. Moreover, the waveguide width linearly increases over 0 ≤ L ≤ N meij Λmeij /2 region, while linearly decreases over N meij Λmeij /2 ≤ L ≤ N meij Λmeij region, in order to enhance the working bandwidth:
where W meij (L) is the width function for ME ij , W me,m and W me,M are minimum and maximum widths, N meij is the grating-period number of ME ij , and Λmeij is the grating pitch of ME ij . Thus, the corresponding trajectory functions can be expressed as:
where Y meij,± denotes trajectory function for ME ij , Δmeij is the grating strength of ME ij .
Design and results for mode exchangers (ME). (A) The schematic configuration of mode exchangers (ME02, ME12) for TE0–TE2 and TE1–TE2 mode swap. The insets show mode-swap processes. (B) The calculated effective-index (n eff) dispersion curves of TE0–2 modes. The index-mismatch (δn eff) compensations for (C) ME02 and (D) ME12. The calculated mode-swap efficiencies (MSEme02, MSEme12) for (E) ME02 and (F) ME12 with varied grating strengths (Δme02, Δme12) and grating pitches (Λme02, Λme12). (G) The calculated light propagation profiles for ME02 (left column) and ME12 (right column) as TE0–2 modes are launched. (H) The optical microscope image of fabricated devices with port identifiers labeled. The insets are scanning electron microscope (SEM) images of ME12 and ME02. (I) The calculated (upper row) and measured (lower row) transmittance matrices for ME02 (left column) and ME12 (right column) at the 1.55-μm wavelength. (J) The measured transmittance spectra for ME02 (upper row) and ME12 (lower row). CADC, cascaded asymmetric directional couplers.
The widths are chosen to be W me,m = 1.23 μm and W me,M = 1.29 μm as an initial setting. The idea is to use LPWGs to compensate the index mismatch between eigen modes. Therefore, the foremost issue is to determine the grating pitch according to the following equation [32]:
where the left term represents the effective-index difference between TE i and TE j , and the right term represents the effective-index compensation offered by LPWGs. The effective-index dispersion curves are calculated for TE0–2 at an average width (W wg = 1.26 μm) by using FEM, as shown in Figure 2B. Thus, grating pitches can be approximately set as Λme02 = 2.65 μm and Λme12 = 4.1 μm according to Eq. (7). The index-mismatch compensations for ME02 and ME12 are shown in Figure 2C and D. One can see that index differences can be perfectly canceled out at the central wavelength of 1.55 μm (see intersections between solid lines and dashed lines). The remanent index mismatch at other wavelengths is also quite small. Moreover, the mode swap should occur merely between target eigen modes since required index compensations are distinct between ME02 and ME12. Here, grating-period numbers are set as fixed values (N me02 = 4, N me12 = 3), and mode-swap efficiencies are calculated with varied grating strengths (Δmeij ) and grating pitches (Λmeij ) by using the finite-difference time-domain (FDTD) method, as shown in Figure 2E and F. For ME ij , the mode-swap efficiency (MSEmeij ) is defined as the normalized transmittance of TEj as TEi is launched. From the results, near-unity mode-swap efficiencies can be reached for both ME02 and ME12 (MSEme02 > 0.99, MSEme12 > 0.99) with the following optimal parameters: Δme02 = 0.22 μm, Δme12 = 0.18 μm, Λme02 = 2.65 μm and Λme12 = 4.325 μm. The light propagation profiles are then calculated for the optimized ME02 and ME12 by using FDTD, as shown in Figure 2G, where efficient and complete mode-swap processes can be observed. The transmittance spectra are also calculated by using FDTD. The calculation results can be found in Supporting Information, Section S3. The fabrication tolerance analysis for MEs is also obtained (see Supporting Information, Figures S3 and S4). Low IL and XT can still be maintained even with a large width deviation of ±20 nm, indicating outstanding reliability in massive productions.
Two sets of devices (ME02 and ME12) were fabricated closely on the same chip, as shown in Figure 2H. The cascaded ADCs (CADC) were employed as terminals for (de)multiplexing mode carriers in a multi-mode waveguide. The detailed discussion about CADC can be found in Supporting Information, Section S4. To measure the mapping relation between input TE m and output TE n modes for ME ij , one can choose Imeij,m and Omeij,n as input and output ports, respectively. The transmittance matrices were measured for the fabricated ME02 and ME12 at the central wavelength, as shown in Figure 2I. The calculated transmittance matrices are also given for comparison. It can be found that the measured insertion losses and crosstalk are as low as IL < 0.88 dB, XT < −15.8 dB, respectively. The pure losses of MEs were measured to be IL < 0.4 dB by deducting losses caused by CADCs. We also measured the transmittance spectra over the wavelength range from 1.52 μm to 1.58 μm, as shown in Figure 2J. From the spectra, low crosstalk of XT < −10 dB can be attained over the whole bandwidth (BW > 60 nm). The demonstrated ME02 and ME12 are mainly applicable for a three-channel MDM, but the proposed design methodology should not be hampered by the required channel number or mode orders, since the index compensation between any eigen-mode pair can be easily achieved by tailoring the grating pitch, which makes it a universal solution to arbitrary mode swap. It should also be noted that the mapping relation for passive ME is fixed once it is designed and fabricated since it is basically a passive device. However, the mode selectivity will not be hindered by the fixed functionality of ME, because the active mode add-drop can be easily obtained by combining MEs with ADCs and single-mode switches, as will be discussed later.
2.2 Direct-access mode add-drop multiplexer
Here, the direct-access scheme presented in Figure 1B is realized by employing the aforementioned MEs. Figure 3A schematically illustrates the proposed structures (MADM i ) for direct add-drop of CHi (carried by TE i ). Among them, MADM0/MADM1 is formed by combining ME02/ME12 with two ADC2 (as discussed in Figure 1B), whereas MADM2 is just a pair of mirrored ADC2 for TE2 (de)multiplexing. The optimization processes and experimental results for ADC2 can be found in Supporting Information, Section S4. The mode add-drop processes for MADM0–2 are illustrated in the insets of Figure 3A. We calculate the light propagation profiles for MADMs by using FDTD, as shown in Figure 3B. The dropping processes are displayed in the 1st–3rd rows. It can be seen that, for MADM i , only TE i is routed into the dropping port with negligible residual power in the multi-mode bus waveguide, showcasing direct-access operations with high efficiencies and low crosstalk. The 4th row shows adding operations for TE0–2. The input TE0 mode is firstly coupled into the multi-mode waveguide and converted to TE2 through a TE0–TE2 coupling. For MADM2, the excited TE2 mode goes directly into the output port, whereas for MADM0–1, an additional mode-swap process is performed to recover the signal-carrier mapping relation. The transmittance spectra are also calculated by using FDTD. The calculation results can be found in Supporting Information, Section S5.
Design and simulation results for direct-access mode add-drop multiplexers (MADM). (A) The schematic configuration of mode add-drop multiplexers (MADM0–2) for TE0–2 modes add-drop. The insets show mode add-drop processes. (B) The calculated light propagation profiles for MADM0 (left column), MADM1 (middle column) and MADM2 (right column) as TE0–2 modes are launched. ME, mode exchanger; ADC, asymmetric directional coupler.
Three sets of devices (i.e. MADM0–2) were fabricated closely on the same chip, as shown in Figure 4A. The CADCs were utilized to (de)multiplex TE0–2 (see Supporting Information, Section S4 for more information). To characterize mode-dropping processes in MADM i , we chose Imadmi,m as the input port to excite TE m , while measured transmittances at Omadmi,n (for the output TE n mode) and Dmadmi (for the dropped mode carrier). Figure 4B shows the measured transmittance matrices for mode-dropping operations. The measurement results show low losses (IL ≈ 0.17–1.52 dB) and low crosstalk (XT ≈ −37.2 ∼ −15.4 dB) at the central wavelength, which agrees well with calculation results (see the upper row of Figure 4B). We also measured the transmittance spectra as TE0–2 are dropped, as shown in Figure 4C, where dashed lines highlight dropped mode carriers. The measured crosstalk levels are as low as XT < −12 dB over a 60-nm wavelength span. To characterize mode-adding processes in MADM i , we chose Amadmi as the input port and measured transmittances at Omadmi,j (for the output TE j mode). The measured transmittance spectra are shown in Figure 4D. From the spectra, one can find low losses (IL < 0.9 dB) and low crosstalk (XT < −15 dB) around the central wavelength.
Measurement results for direct-access mode add-drop multiplexers. (A) The optical microscope image of fabricated devices with port identifiers labeled. (B) The calculated (upper row) and measured (lower row) transmittance matrices for MADM0 (left column), MADM1 (middle column) and MADM2 (right column) at the 1.55-μm wavelength. The measured transmittance spectra for MADM0 (upper row), MADM1 (middle row) and MADM2 (lower row) as light was launched from (C) input ports and (D) adding ports. MADM, mode add-drop multiplexer; CADC, cascaded asymmetric directional couplers.
2.3 Direct-access mode-division multiplexing switch
The direct-access MDMS can be realized by exploiting the aforementioned MADMs. Here, we propose and demonstrate a cascaded mode-division multiplexing switches (CMDMS) to obtain dynamic switching operations in a three-channel MDM, where three independent MDMSs (i.e. MDMS0–2) are successively cascaded to add/drop TE0–2, as schematically illustrated in Figure 5A. Each MDMS is built by assembling the aforementioned MADM with a Mach–Zehnder switch. Here, a normally-bar switch (NBS) is proposed and demonstrated to enhance the working bandwidth and reduce the power consumption by using bent directional couplers (BDCs) and a multi-sectional phase shifter (PS), as schematically illustrated in Figure 5B. The BDCs can provide a flattened coupling-ratio spectrum to ensure high extinction ratios over a broad bandwidth [33]. The multi-sectional PS is introduced to induce a phase bias at off state [34]:
where φ nbs,bias is the phase bias, n nbs,1 and n nbs,2 are TE0 effective indices with W wg = W nbs,1 and W wg = W nbs,2, L nbs,s and L nbs,r are waveguide lengths of shifting and reference sections. Thus, the bar-coupling can be achieved at off state by properly choosing waveguide widths and lengths of PS to induce a phase bias of φ nbs,bias = π. More simulation details about BDCs and PS can be found in Supporting Information, Section S6. The phase bias of π can be compensated in the switching operation by heating arm #1 (see the yellow region in Figure 5B), leading to the cross-coupling at on state. Figure 5C shows the measured transmittance spectra for the fabricated NBS (see also Figure 11 in Supporting Information). A high extinction ratio (ER > 16.7 dB) was measured at the central wavelength with a switching power of 30.9 mW. The flattened transmittance spectra can be observed over the whole bandwidth at both on and off states. More measurement results and fabrication-tolerance analysis for NBS can be found in Supporting Information, Section S6. Thus, if all the NBSs are at off state, the mode carrier (i.e. TE i ) dropped by MADM i will be coupled back into the multi-mode waveguide, and thus all the mode carriers will go straight into the output port with the initial signal-carrier mapping relation preserved. On the other hand, if NBS i is switched on, the corresponding signal channel CH i (carried by TE i ) can be added and dropped at Amdms,i and Dmdms,i , as illustrated in Figure 5D. One should note that the cascading configuration is proposed just to verify the direct-access operations for arbitrary mode carriers, and a stand-alone MDMS also works especially when only few mode carriers need to be handled within a large capacity. For example, a separate MDMS0 is fully functional if TE0 is the only mode carrier to be added and dropped.
Design and all-off measurement results for direct-access mode-division multiplexing switches (MDMS). (A) The schematic configuration of cascaded MDMSs (CMDMS). (B) The schematic configuration of normally-bar switch (NBS). (C) The measured transmittance spectra for NBS at off state (upper panel) and on state (lower panel). (D) The illustration of mode add-drop operations. (E) The optical microscope image of fabricated devices with port identifiers labeled. (F) The measured transmittance matrix for CMDMS at all-off state (λ = 1.55 μm). (G) The measured transmittance spectra for CMDMS at all-off state as light was launched from input ports. BDC, bent directional coupler; PS, phase shifter; CADC, cascaded asymmetric directional coupler.
Figure 5E shows the microscope image of fabricated devices, which consists of of two mirrored CADCs for mode (de)multiplexing and a CMDMS in between. One can refer to Supporting Information, Section S4 for more details about CADC. We firstly characterized the transmission responses as all the NBSs were switched off. The TE i mode was excited in the multi-mode waveguide by choosing Imdms,i as the input port. The output transmittances of TE i were measured at Omdms,i , while the transmittances of dropped mode carrier were measured at Dmdms,i . The transmittance matrix was measured at the 1.55-μm wavelength, as shown in Figure 5F. From the results, the light transmission from Imdms,i to Omdms,i is dominant, whereas the transmittance at Dmdms,i is negligible, indicating that no mode carriers were dropped. Low losses and crosstalk were measured (IL ≈ 1.35–1.56 dB, XT ≈ −28.8 ∼ −14.8 dB) at all-off state. We also measured the transmittance spectra over the wavelength range from 1.52 μm to 1.58 μm, as shown in Figure 5G. Over the whole bandwidth, crosstalk levels were measured to be XT < −10 dB. One might find some dense ripples in the measured spectra, which is mainly caused by the interference effect. Specifically, the extinction ratio of NBS is not high enough to fully eliminate cross-coupling, and a weak interference will be accumulated as multiple NBSs are cascaded. The ripples can be depressed by further improving the extinction ratio of NBS (see Supporting Information, Section S6).
Next, dropping operations were conducted by switching on NBS0–2, as shown in Figure 6A. The left panel (i.e. 1st–3rd columns) and right panel (i.e. 4th–6th columns) show the transmittance spectra at output and dropping ports, respectively. For clarity, we only display the measurement results for Imdms,i -to-Omdms,i and Imdms,i -to-Dmdms,i transmittances. It can be seen that, when NBS i was switched on, transmittances at the corresponding Omdms,i port were efficiently reduced to a relatively low level (<−15 dB), and the dropped mode carriers were transferred to Dmdms,i with low losses (IL < 1 dB), showcasing dropping processes of selected mode carriers. After that, we measured the transmittance spectra when light is launched from Amdms,i as NBS i was switched on to verify the adding functionality of CMDMS, as shown in Figure 6B. From the spectra, low losses (IL < 1 dB) and low crosstalk (XT ≈ −18 ∼ −12 dB) can be observed for Amdms,i -to-Omdms,i transmissions. We also measured eye-diagrams and bit-error rates (BER) with a 10 Gbps data stream to verify data links built in CMDMS under different states (see Supporting Information, Section S7). Based on the single-port method, wide-opened eye-diagrams with ER > 10 dB have been observed. The power penalties were measured to be as low as ≈1.2–2.6 dB at BER = 10−9.
Add-drop operation results for direct-access mode-division multiplexing switches (MDMS). (A) The measured transmittance spectra at output ports (left penal) and dropping ports (right panel) as light is launched from input ports and NBS0-2 were switched on. The dashed boxes highlight dropped mode carriers. (B) The measured transmittance spectra at output ports as light was launched from adding ports while NBS0-2 were switched on. CMDMS, cascaded mode-division multiplexing switch; NBS, normally-bar switch.
2.4 Direct-access wavelength/mode-division multiplexing switch
The wavelength/mode-hybrid multiplexing has drawn tremendous interests for its ability to push the capacity limit while ensure relatively low cost and power consumption by transmitting signals with both wavelength and mode carriers [35], [36], [37]. Thus, it is essential to develop the wavelength/mode-division multiplexing switch (WMDMS) for handling wavelength/mode-hybrid carriers. Here, the cascaded WMDMSs (CWMDMS) are proposed and demonstrated as a proof of concept. In essence, the proposed scheme is just the aforementioned CMDMS with NBSs replaced by tunable MRRs, as schematically illustrated in Figure 7A. Three mode carriers (i.e. TE0–2) and four wavelength carriers (i.e. λ 0 ≈ 1.5445 μm, λ 1 ≈ 1.5461 μm, λ 2 ≈ 1.5477 μm and λ 3 ≈ 1.5491 μm) are considered in this scheme, and the signal channel carried by TE i /λ j is denoted as CH ij . Thus, totally twelve wavelength/mode-hybrid carriers can be supported. The tunable MRR is realized based on a race-track resonator with a pair of BDCs [38], as shown in Figure 7B. Figure 7C shows the measured transmittance spectra as different electric power is applied onto the tunable MRR (see also Supporting Information, Figure S14). One can see that the utilized wavelength carriers can be totally covered by the wavelength-tuning rang with a large wavelength-tuning slope of ≈0.095 nm/mW (see Figure 7D). More details about tunable MRR can be found in Supporting Information, Section S8. The static resonant wavelength of MRR is deviated from the utilized wavelength carriers (i.e. λ 0–3). Thus, at all-off state, the incident CH ij will propagate directly into output ports. To add/drop CH ij (carried by TE i /λ j ), one can choose WMDMSi to select TE i , while apply the specific electric power (P j ) onto MRR i to select λ j , as illustrated in Figure 7E. One should note that the proposed scheme is just a proof of concept, and it is unnecessary to cascade all the three WMDMSs if only few wavelength/mode-hybrid carriers need to be added and dropped from a large capacity. For example, one can use a stand-alone WMDMS0 to perform the carrier manipulation if CH00 is the only signal channel to be handled.
Design and all-off measurement results for direct-access wavelength/mode-division multiplexing switches (WMDMSs). (A) The schematic configuration of cascaded WMDMSs (CWMDMS). (B) The schematic configuration of tunable micro-ring resonator (MRR). (C) The measured transmittance spectra at the CRO port of MRR as applied electrical power is varied. (D) The measured tuning efficiency of MRR. (E) The illustration of wavelength/mode add-drop operations. (F) The optical microscope image of fabricated devices with port identifiers labeled. The inset shows the enlarged view of a single WMDMS. (G) The measured transmittance matrix for CWMDMS at all-off state. (H) The measured transmittance spectra for CWMDMS at all-off state as light is launched from input ports. BDC, bent directional coupler; CADC, cascaded asymmetric directional coupler; CMRR, cascaded micro-ring resonator.
Figure 7F shows the microscope images of fabricated devices. A pair of mirrored CADCs were fabricated at input and output ends of CWMDMS for mode (de)multiplexing (see Supporting Information, Section S4 for more information). Three sets of cascaded MRRs (CMRR) were placed at the output end as four-channel wavelength demultiplexers (see Supporting Information, Section S9 for more information). Thus, to launch CH ij into the multi-mode waveguide, one can choose Iwmdms,i as the input port, while align the wavelength of tunable laser to λ j . The output transmittances of CH ij can be measured at Owmdms,ij . The adding and dropping ports for CH ij are Awmdms,i and Dwmdms,i , respectively. We firstly measured the transmittance matrix as zero electric power was applied (see Figure 7G). It can be seen that the incident CH ij was completely routed to the corresponding output port (i.e. Owmdms,ij ). The insertion losses and crosstalk were measured to be IL < 2.25 dB and XT < −15.2 dB for all the signal channels. Figure 7H shows the measured transmittance spectra over the wavelength range from 1.542 μm to 1.552 μm, where low crosstalk (XT < −15 dB) can be observed over the whole bandwidth.
We then characterized the transmission responses for dropping operations, as shown in Figure 8A. The left panel (i.e. 1st–3rd columns) shows Iwmdms,i -to-Owmdms,i transmittances, while the right panel (i.e. 4th–6th columns) shows Iwmdms,i -to-Dwmdms,i transmittances. The applied electric power was set as P 0 = 24.2 mW, P 1 = 41.1 mW, P 2 = 57.9 mW and P 3 = 72.6 mW for dropping λ 0–3, respectively. From the spectra, when MRR i was tuned with electric power of P j , the Owmdms,ij transmittance at λ j was dramatically reduced with ER > 15 dB, and the dropped CH ij was efficiently transferred to the corresponding dropping port (i.e. Dwmdms,i ) with low losses (IL < 2 dB). The 3 dB bandwidths are slightly different between tunable MRR and CMRR filter due to fabrication errors, which leads to the notch-like response shown in Figure 8A [27]. The adding operations for CH ij were then conducted by injecting light into Awmdms,i while applying electric power of P j onto MRR i . The measured transmittance spectra are shown in Figure 8B. The crosstalk levels were measured to be XT < −15 dB, while insertion losses were measured to be IL < 2 dB. The single-port method was used to measure eye-diagrams and bit-error rates (BER) as a 10 Gbps data stream was coupled into the device (see Supporting Information, Section S10). Figure S16A shows the measured eye-diagrams, where clear patterns and high extinction ratios (ER > 10 dB) can be observed. From Figure S16B, power penalties were measured to be ≈ 0.9–2.2 dB to ensure low bit-error rates of BER = 10−9.
Add-drop operation results for direct-access wavelength/mode-division multiplexing switches (MDMSs). (A) The measured transmittance spectra at output ports (left panel) and dropping ports (right panel) as light was launched from input ports while different electric power (P 0–3) was applied on MRR0–2. The dashed boxes and arrows highlight the dropped wavelength/mode-hybrid carriers. (B) The measured transmittance spectra at output ports as light was launched from adding ports and different electric power (P 0–3) was applied on MRR0–2. CWMDMS, cascaded wavelength/mode-division multiplexing switch; MRR, micro-ring resonator.
2.5 Reconfigurable multi-mode hubbed-ring network
The proliferation of on-chip photonic interconnects has created a rising demand for high-performance optical networks [39]. The hubbed-ring network (HRN) has been widely used in optical communication systems (e.g. metropolitan area networks), taking advantage of its outstanding scalability and robustness [40], [41], [42]. Typically, an HRN can be built by connecting one central hub and numerous access nodes with a single feeding ring. The network traffic of the HRN is organized on a multi-point-to-point (MP2P) basis. In other words, the optical paths are built merely between one central hub and each access node in a duplex manner, whereas inter-node links are forbidden. Here, an HRN with three access nodes (denoted as nodes #1–3) is investigated as an example, as schematically illustrated in Figure 9A, where there are totally six optical paths contained in a single ring-like bus waveguide. The hub-to-node (H2N) link is supported by paths #1–3 (see solid arrows), while the node-to-hub (N2H) link is supported by paths #1′–3′ (see dashed arrows). One can find that these optical paths overlap with each other, which makes it inevitable to use the multiplexing technology to achieve parallel data transmissions. The conventional HRNs are usually constructed based on WDM [39], as shown in the inset of Figure 9A, which requires a large number of laser diodes with different wavelengths working simultaneously at the central hub. As an improvement, we propose to use MDM to substitute WDM, leading to a novel concept of multi-mode HRN, as shown in the inset of Figure 9A. For this approach, the optical paths in HRN are carried by multiple eigen modes (i.e. TE0–2) rather than multiple wavelengths (i.e. λ0–2), in order to improve the power efficiency. The reconfigurability is another desired attribute for HRN, which refers to the ability to deploy the mode carrier for each optical path. It should be noted that there are usually a large number of nodes in HRN and each node only requires a limited number of carriers, which makes it inefficient to add-drop all the mode carriers. Therefore, the aforementioned MDMS should just fit in the application scenario due to its ability to directly access any eigen mode without demultiplexing other idle carriers.
Design and results for the reconfigurable multi-mode hubbed-ring network (HRN). (A) The illustration of HRNs based on wavelength-division multiplexing (WDM) and mode-division multiplexing (MDM). (B) The optical microscope image of fabricated HRN. The dashed boxes show enlarged views of cascaded mode add-drop multiplexers (CMADM, upper panel) and cascaded mode-division multiplexing switches (CMDMS, lower panel). The HRN consists of one CMADM as a central hub and three CMDMSs as access nodes #1–3. The port and switch identifiers are also labeled. The illustrations of hub-to-nodes (H2N) and nodes-to-hub (N2H) links for HRN at (C) state A and (D) state B. The shaded boxes illustrate which normally-bar switches (NBS) are turned on. The measured transmittance matrices for the HRN at (E) state A and (F) state B (λ = 1.55 μm). The measured transmittance spectra for the HRN at (G) state A and (H) state B. The measured (I) eye-diagrams and (J) bit-error rates (BER) with 10 Gbps data links at λ = 1.55 μm using multi-port method. CGMB, curvature-gradient multi-mode bend; B2B, back-to-back.
Here, we propose and demonstrate a reconfigurable multi-mode HRN by utilizing the aforementioned MADM and MDMS. Figure 9B shows the microscope images of fabricated devices. One set of cascaded MADMs (CMADM) was employed as a central hub, while three sets of cascaded MDMSs (denoted as CMDMS1-3) were used as three access nodes (i.e. nodes #1–3), as shown in the right panel of Figure 9B. The central hub and access nodes were connected by a ring-like multi-mode bus waveguide with four curvature-gradient multi-mode bends (CGMB) at corners. More details about CGMB can be found in Supporting Information, Section S11. The input/output port at central hub is denoted as H1–6, while the input/output port at node #i is denoted as N i1–i6 (see the right panel of Figure 9B). One can choose to switch on NBS ij (see the left panel of Figure 9B) to select TE j as the mode carrier for path #i (′) between the central hub and access node #i. Two connection states were implemented to verify the reconfigurability of HRN, as illustrated in Figure 9C and D. At state A, we chose to switch on NBS10, NBS21 and NBS32. In this case, mode carriers for paths #1(′), #2(′) and #3(′) were TE0, TE1 and TE2, respectively. The H2N/N2H links were characterized by choosing H2/N11, H4/N23 and H6/N35 as input ports, while choosing N12/H1, N24/H3 and N36/H5 as output ports. At state B, we chose to switch on NBS12, NBS20 and NBS31. In this case, carriers for paths #1(′), #2(′) and #3(′) were TE2, TE0 and TE1, respectively. The H2N/N2H links were characterized by choosing H2/N15, H4/N21 and H6/N33 as input ports, while choosing N16/H1, N22/H3 and N34/H5 as output ports. The single-wavelength measurement was firstly performed to obtain transmittance matrices at λ = 1.55 μm for both states A and B, as shown in Figure 9E and F. From the results, it can be seen that optical paths were successfully established for both H2N and N2H links. The measured insertion losses are IL ≈ 1.76–4.17 dB at state A and IL ≈ 1.25–4.05 dB at state B. The crosstalk was also measured to be as low as XT < −17.6 dB. The transmittance spectra were then measured at states A and B, as shown in Figure 9G and H. From the spectra, low crosstalk levels of XT < −10 dB can be achieved over a 45 nm bandwidth. After that, 10 Gbps data streams were injected into each optical path built in the reconfigurable multi-mode HRN to further evaluate its overall performance. Here, the rigorous multi-port method was used to characterize data transmissions in the HRN. Figure 9I shows the measured eye-diagrams at states A and B, where low noises and high extinction ratios (ER > 10 dB) can be observed. The bit-error rates were also measured at the central wavelength, as shown in Figure 9J. Low power penalties of ≈1.9–3.1 dB were experimentally observed under low bit-error rates of BER <10−9. Thus, the aggregate transmission bandwidth should be 6 × 10 Gbps for parallel transmissions of totally six optical paths in HRN.
The above results for states A and B are present just to show the feasibility, and more connection states can be supported by the reconfigurable multi-mode HRN. Actually, totally M!/(M − N)! connection states can be achieved for an HRN with M mode carriers and N access nodes. We believe these experimental results can be a solid demonstration of viability of using the MDMS to build scalable on-chip networks. Moreover, the results shown in Figure 9G and H have demonstrated the broadband characteristic of the HRN, which makes it potential to further expand the capacity and add more access nodes to the network by introducing wavelength/mode-hybrid multiplexing, as discussed in Section 2.4.
3 Summary and conclusion
In conclusion, we have proposed and demonstrated, to the best of our knowledge, the first direct-access MDMS as a novel platform for dynamically controlling mode carriers. In the first place, the ME and MADM are developed to address the “direct-access” conundrum. The direct-access MDMS is then realized by employing MADM as a building block. The proposed scheme is a combination of direct-access MADMs and normally-bar switches (NBSs), which is capable of adding and dropping any selected mode carrier in a three-channel mode-division multiplexing system. We have also proposed and demonstrated a WMDMS by introducing tunable micro-ring resonators (MRRs) to achieve an expanded link capacity with totally twelve wavelength/mode-hybrid carriers. In principle, these proposed schemes are viable for any MDM systems regardless of the channel number. We have summarized and compared the performance of several reported MDMSs in Supporting Information, Section S12. It can be found that our proposed MDMS and WMDMS can provide low crosstalk, low losses and small footprint. The device footprint can be further reduced by introducing thermal-isolation trenches [43] and direct thermal contacts [44] to further improve the thermo-optical efficiency of NBSs. A reconfigurable multi-mode hubbed-ring network (HRN) is then constructed based on the developed MADM and MDMS. The mode carriers are utilized to support six parallel optical paths between one central hub and three access nodes with a large bandwidth. The different connection states can be built in the multi-mode HRN, taking advantages of its great reconfigurability. In this work, we have provided a general and systematic solution to how to dynamically control mode carriers in a direct-access manner, which we believe is potential for large-scale on-chip photonic interconnects.
Funding source: National Key Research and Development Program of China
Award Identifier / Grant number: 2019YFB2203603
Funding source: National Science Fund for Excellent Young Scholars
Award Identifier / Grant number: 61922070
-
Author contribution: H. X. designed and characterized devices. H. X. and C. L. measured data transmission responses. Y. S. and D. D. supervised the research. The manuscript was written through contributions of all authors.
-
Research funding: This work was supported by National Key Research and Development Program of China (2019YFB2203603) and the National Science Fund for Excellent Young Scholars (61922070).
-
Conflict of interest statement: The authors declare no conflicts of interest regarding this article.
References
[1] A. Shacham, K. Bergman, and L. P. Carloni, “Photonic networks-on-chip for future generations of chip multiprocessors,” IEEE Trans. Comput., vol. 57, pp. 1246–1260, 2008, https://doi.org/10.1109/tc.2008.78.Search in Google Scholar
[2] H. J. Caulfield and S. Dolev, “Why future supercomputing requires optics,” Nat. Photonics, vol. 4, pp. 261–263, 2010, https://doi.org/10.1038/nphoton.2010.94.Search in Google Scholar
[3] M. Smit, J. van der Tol, and M. Hill, “Moore’s law in photonics,” Laser Photon. Rev., vol. 6, pp. 1–13, 2012, https://doi.org/10.1002/lpor.201100001.Search in Google Scholar
[4] P. Dong, “Silicon photonic integrated circuits for wavelength-division multiplexing applications,” IEEE J. Sel. Top. Quant. Electron., vol. 22, pp. 370–378, 2016, https://doi.org/10.1109/jstqe.2016.2575358.Search in Google Scholar
[5] D. Liu, H. Xu, Y. Tan, Y. Shi, and D. Dai, “Silicon photonic filters,” Microw. Opt. Technol. Lett., vol. 63, pp. 2252–2268, 2021.10.1002/mop.32509Search in Google Scholar
[6] H. Xu, D. Dai, and Y. Shi, “Silicon integrated nanophotonic devices for on-chip multi-mode interconnects,” Appl. Sci., vol. 10, p. 6365, 2020, https://doi.org/10.3390/app10186365.Search in Google Scholar
[7] Y. He, Y. Zhang, Q. Zhu, et al.., “Silicon high-order mode (de)multiplexer on single polarization,” J. Lightwave Technol., vol. 36, pp. 5746–5753, 2018, https://doi.org/10.1109/jlt.2018.2878529.Search in Google Scholar
[8] D. Dai, J. Wang, and Y. Shi, “Silicon mode (de)multiplexer enabling high capacity photonic networks-on-chip with a single-wavelength-carrier light,” Opt. Lett., vol. 38, pp. 1422–1424, 2013, https://doi.org/10.1364/ol.38.001422.Search in Google Scholar PubMed
[9] D. Dai, C. Li, S. Wang, et al.., “10-channel mode (de)multiplexer with dual polarizations,” Laser Photon. Rev., vol. 12, p. 1700109, 2018, https://doi.org/10.1002/lpor.201700109.Search in Google Scholar
[10] L. H. Gabrielli, D. Liu, S. G. Johnson, and M. Lipson, “On-chip transformation optics for multimode waveguide bends,” Nat. Commun., vol. 3, p. 1217, 2012, https://doi.org/10.1038/ncomms2232.Search in Google Scholar PubMed
[11] C. Sun, Y. Yu, G. Chen, and X. Zhang, “Ultra-compact bent multimode silicon waveguide with ultralow inter-mode crosstalk,” Opt. Lett., vol. 42, pp. 3004–3007, 2017, https://doi.org/10.1364/ol.42.003004.Search in Google Scholar
[12] H. Xu and Y. Shi, “Ultra-sharp multi-mode waveguide bending assisted with metamaterial-based mode converters,” Laser Photon. Rev., vol. 12, p. 1700240, 2018, https://doi.org/10.1002/lpor.201700240.Search in Google Scholar
[13] H. Xu and Y. Shi, “Dual-mode waveguide crossing utilizing taper-assisted multimode-interference couplers,” Opt. Lett., vol. 41, pp. 5381–5384, 2016, https://doi.org/10.1364/ol.41.005381.Search in Google Scholar
[14] H. Xu and Y. Shi, “Metamaterial-based Maxwell’s fisheye lens for multimode waveguide crossing,” Laser Photon. Rev., vol. 12, p. 1800094, 2018, https://doi.org/10.1002/lpor.201800094.Search in Google Scholar
[15] S. H. Badri, H. R. Saghai, and H. Soofi, “Multimode waveguide crossing based on a square Maxwell’s fisheye lens,” Appl. Opt., vol. 58, pp. 4647–4653, 2019, https://doi.org/10.1364/ao.58.004647.Search in Google Scholar PubMed
[16] H. Xu and Y. Shi, “Ultra-broadband dual-mode 3 dB power splitter based on a Y-junction assisted with mode converters,” Opt. Lett., vol. 41, pp. 5047–5050, 2016, https://doi.org/10.1364/ol.41.005047.Search in Google Scholar PubMed
[17] W. Chang, X. Ren, Y. Ao, et al.., “Inverse design and demonstration of an ultracompact broadband dual-mode 3 dB power splitter,” Opt. Express, vol. 26, pp. 24135–24144, 2018, https://doi.org/10.1364/oe.26.024135.Search in Google Scholar PubMed
[18] H. Xie, Y. Liu, Y. Wang, et al.., “An ultra-compact 3-dB power splitter for three modes based on pixelated meta-structure,” IEEE Photon. Technol. Lett., vol. 32, pp. 341–344, 2020, https://doi.org/10.1109/lpt.2020.2975128.Search in Google Scholar
[19] C. Sun, Y. Yu, G. Chen, and X. Zhang, “Integrated switchable mode exchange for reconfigurable mode-multiplexing optical networks,” Opt. Lett., vol. 41, pp. 3257–3260, 2016, https://doi.org/10.1364/ol.41.003257.Search in Google Scholar PubMed
[20] Y. Xiong, R. B. Priti, and O. Liboiron-Ladouceur, “High-speed two-mode switch for mode-division multiplexing optical networks,” Optica, vol. 4, pp. 1098–1102, 2017, https://doi.org/10.1364/optica.4.001098.Search in Google Scholar
[21] C. Sun, W. Wu, Y. Yu, et al.., “De-multiplexing free on-chip low-loss multimode switch enabling reconfigurable inter-mode and inter-path routing,” Nanophotonics, vol. 7, pp. 1571–1580, 2018, https://doi.org/10.1515/nanoph-2018-0053.Search in Google Scholar
[22] S. Wang, H. Wu, H. K. Tsang, and D. Dai, “Monolithically integrated reconfigurable add-drop multiplexer for mode-division-multiplexing systems,” Opt. Lett., vol. 41, pp. 5298–5301, 2016, https://doi.org/10.1364/ol.41.005298.Search in Google Scholar
[23] Y. Zhang, Y. He, Q. Zhu, C. Qiu, and Y. Su, “On-chip silicon photonic 2 × 2 mode- and polarization-selective switch with low inter-modal crosstalk,” Photon. Res., vol. 5, pp. 521–526, 2017, https://doi.org/10.1364/prj.5.000521.Search in Google Scholar
[24] L. Yang, T. Zhou, H. Jia, et al.., “General architectures for on-chip optical space and mode switching,” Optica, vol. 5, pp. 180–187, 2018, https://doi.org/10.1364/optica.5.000180.Search in Google Scholar
[25] B. Stern, X. Zhu, C. P. Chen, et al.., “On-chip mode-division multiplexing switch,” Optica, vol. 2, pp. 530–535, 2015, https://doi.org/10.1364/optica.2.000530.Search in Google Scholar
[26] S. Wang, X. Feng, S. Gao, et al.., “On-chip reconfigurable optical add-drop multiplexer for hybrid wavelength/mode-division-multiplexing systems,” Opt. Lett., vol. 42, pp. 2802–2805, 2017, https://doi.org/10.1364/ol.42.002802.Search in Google Scholar
[27] L. W. Luo, N. Ophir, C. P. Chen, et al.., “WDM-compatible mode-division multiplexing on a silicon chip,” Nat. Commun., vol. 5, p. 3069, 2014, https://doi.org/10.1038/ncomms4069.Search in Google Scholar PubMed
[28] D. Chen, X. Xiao, L. Wang, Y. Yu, W. Liu, and Q. Yang, “Low-loss and fabrication tolerant silicon mode-order converters based on novel compact tapers,” Opt. Express, vol. 23, pp. 11152–11159, 2015, https://doi.org/10.1364/oe.23.011152.Search in Google Scholar PubMed
[29] Z. Cheng, J. Wang, Z. Yang, et al.., “Sub-wavelength grating assisted mode order converter on the SOI substrate,” Opt. Express, vol. 27, pp. 34434–34441, 2019, https://doi.org/10.1364/oe.27.034434.Search in Google Scholar
[30] J. Guo, C. Ye, C. Liu, et al.., “Ultra-compact and ultra-broadband guided-mode exchangers on silicon,” Laser Photon. Rev., vol. 14, p. 2000058, 2020, https://doi.org/10.1002/lpor.202000058.Search in Google Scholar
[31] G. Zhang and O. Liboiron-Ladouceur, “Scalable and low crosstalk silicon mode exchanger for mode division multiplexing system enabled by inverse design,” IEEE Photon J., vol. 13, pp. 1–13, 2021, https://doi.org/10.1109/jphot.2021.3061112.Search in Google Scholar
[32] A. Yariv, “Coupled-mode theory for guided-wave optics,” IEEE J. Quant. Electron., vol. 9, pp. 919–933, 1973, https://doi.org/10.1109/jqe.1973.1077767.Search in Google Scholar
[33] S. Chen, Y. Shi, S. He, and D. Dai, “Low-loss and broadband 2 × 2 silicon thermo-optic Mach–Zehnder switch with bent directional couplers,” Opt. Lett., vol. 41, pp. 836–839, 2016, https://doi.org/10.1364/ol.41.000836.Search in Google Scholar
[34] H. Xu, D. Dai, and Y. Shi, “Low-crosstalk and fabrication-tolerant four-channel CWDM filter based on dispersion-engineered Mach–Zehnder interferometers,” Opt. Express, vol. 29, pp. 20617–20631, 2021, https://doi.org/10.1364/oe.428352.Search in Google Scholar
[35] J. B. Driscoll, C. P. Chen, R. R. Grote, et al.., “A 60 Gb/s MDM-WDM Si photonic link with <0.7 dB power penalty per channel,” Opt. Express, vol. 22, pp. 18543–18555, 2014, https://doi.org/10.1364/oe.22.018543.Search in Google Scholar
[36] D. Dai, J. Wang, S. Chen, S. Wang, and S. He, “Monolithically integrated 64-channel silicon hybrid demultiplexer enabling simultaneous wavelength- and mode-division-multiplexing,” Laser Photon. Rev., vol. 9, pp. 339–344, 2015, https://doi.org/10.1002/lpor.201400446.Search in Google Scholar
[37] S. Wang, H. Wu, M. Zhang, and D. Dai, “A 32-channel hybrid wavelength-/mode-division (de)multiplexer on silicon,” IEEE Photon. Technol. Lett., vol. 30, pp. 1194–1197, 2018, https://doi.org/10.1109/lpt.2018.2839533.Search in Google Scholar
[38] P. Chen, S. Chen, X. Guan, Y. Shi, and D. Dai, “High-order microring resonators with bent couplers for a box-like filter response,” Opt. Lett., vol. 39, pp. 6304–6307, 2014, https://doi.org/10.1364/ol.39.006304.Search in Google Scholar PubMed
[39] V. C. López and L. Velasco, “Elastic optical networks: architectures, technologies, and control,” in Optical Networks, From Physical Layer to Service Offerings, 1st ed. Cham, Switzerland, Springer International Publishing, 2016.10.1007/978-3-319-30174-7Search in Google Scholar
[40] N. Bouabdallah and H. Perros, “Cost-effective single-hub WDM ring networks: a proposal and analysis,” Comput. Network., vol. 51, pp. 3878–3901, 2007, https://doi.org/10.1016/j.comnet.2007.04.014.Search in Google Scholar
[41] J. Manchester, P. Bonenfant, and C. Newton, “The evolution of transport network survivability,” IEEE Commun. Mag., vol. 37, pp. 44–51, 1999, https://doi.org/10.1109/35.783124.Search in Google Scholar
[42] D. Chiaroni, G. B. Santamaria, C. Simonneau, et al.., “Packet OADMs for the next generation of ring networks,” Bell Labs Tech. J., vol. 14, pp. 265–283, 2010, https://doi.org/10.1002/bltj.20415.Search in Google Scholar
[43] P. Sun and R. M. Reano, “Submilliwatt thermo-optic switches using free-standing silicon-on-insulator strip waveguides,” Opt. Express, vol. 18, pp. 8406–8411, 2010, https://doi.org/10.1364/oe.18.008406.Search in Google Scholar PubMed
[44] J. Zhang and S. He, “Cladding-free efficiently tunable nanobeam cavity with nanotentacles,” Opt. Express, vol. 25, pp. 12541–12551, 2017, https://doi.org/10.1364/oe.25.012541.Search in Google Scholar
Supplementary Material
The online version of this article offers supplementary material (https://doi.org/10.1515/nanoph-2021-0441).
© 2021 Hongnan Xu et al., published by De Gruyter, Berlin/Boston
This work is licensed under the Creative Commons Attribution 4.0 International License.
