Open Access
Add to BibSonomy
Abstract
Novel broadband operating 1×8 channel wavelength-division-multiplexed (WDM) optical filter is proposed and experimentally demonstrated. The proposed 1×8 channel WDM filter is composed of third-stage delayed interferometers with point-symmetric Mach-Zehnder type wavelength insensitive couplers (PS-MZ-WINCs). We verify that the PS-MZ-WINC operates as a key component for the broadband operating WDM optical filter in terms of good interchannel balance and low crosstalk without accompanying with an additional excess loss. The fabricated silicon-nanowire 1×8 channel WDM filter with the channel spacing of ∼9-nm exhibited a clear demultiplexing filter response over the wavelength range of >100-nm in the O-band regime, together with interchannel balance of <0.5 dB, and low excess loss of <0.9 dB. The proposed scheme could be very effective for increasing channel count and/or channel spacing for several kinds of WDM applications.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
System performances continue to require an abrupt increase in the aggregate bandwidth of chip-to-chip interconnects based on multi-package parallel systems [1,2]. To date, wavelength division multiplexing (WDM) has been investigated as one of key technologies to make bandwidth density of optical interconnects much higher [3–5].
In our previous works, we reported several kinds of 1×4 channel optical (De)MUXs based on multistage delayed interferometers (MDI) [6,7] for processing WDM signals in silicon photonics based integrated transceivers [8]. Usually, a broad operating wavelength range has a great impact for increasing available channel count (NCh) and/or channel spacing (Δν) in the MDI-type WDM filter schemes. A directional coupler (DC) used in the MDI-type WDM filter normally has a lower insertion loss compared with the case where a multimode interference (MMI) coupler [9] is used.
However, the DC fundamentally has a sinusoidally-varying coupling ratio against a wavelength. That is, the optical coupling ratio changes in a complementary manner between the two outputs of DC. As a result, without considering so-called phase errors by fabrication imperfections, the imbalanced coupling ratio of the DC can cause to deteriorate spectral response of MDI-type WDM filter in terms of insertion loss, interchannel balance and crosstalk, which inherently restricts the operating range of WDM filters. To overcome these drawbacks, there have been many reports for alleviating wavelength sensitive coupling ratio by modifying the coupling region of DC to the anti-symmetrical adiabatic scheme [10] or by bending the DC configuration [11]. In Ref. 10, wideband operation of 60-nm span was experimentally verified with a low crosstalk of ≤–20dB. However, the excess loss of the device was ≥2dB, and the interchannel balance did not keep constant within the 4-channel operating range. Also, the channel count was limited to 1×4. Meanwhile, in Ref. 11, although very low excess loss of 1dB was achieved with wideband operation of 80-nm span, the channel count was limited to be 1×4.
In our previous work, to solve the aforementioned problems, we adopted the Mach-Zehnder interferometric wavelength insensitive coupler (MZ-WINC) as an optical coupler, and experimentally demonstrated a wide operating range by locating the two identical Y-axis-symmetric Mach-Zehnder interferometric wavelength insensitive couplers (YS-MZ-WINCs) at the back and forth of each DI [12]. We experimentally demonstrated wideband operating range of ≥100-nm wavelength span in a 1×4 scheme. However, as the channel count increases to 1×8, the advantage reported in Ref. 12 was severely downgraded in terms of insertion loss, interchannel balance and spectral crosstalk.
On the other hand, as discussed in Ref. 13, K. Jinguji et al identified that a point-symmetric Mach-Zehnder interferometric wavelength insensitive coupler (PS-MZ-WINC) can provide much better flatband filter spectral response in silica based planar waveguides. This idea was applied to the 2×2 silicon based optical switching device to enlarge the operating wavelength range [14]. To date, however, there have been little report on WDM filter applications where the PS-MZ-WINC is adopted as an optical coupler in the MDI scheme.
In this work, in order to make available spectral range much wider for larger NCh and wider Δν, we propose a novel MDI-type WDM filter based on the PS-MZ-WINCs, clarify that the PS-MZ-WINC plays a major role in extending available operating wavelength range maintaining a low insertion loss, low interchannel imbalance and low spectral crosstalk of the device, and theoretically verify a superior spectral characteristics to the previously reported WDM filters based on the normal DC [6] and the YS-MZ-WINC [12] from the viewpoint of available spectral range and scalability of the channel count. Based on the proposed filter scheme, we experimentally demonstrate 1×8 channel silicon-nanowire-based device operating in >100-nm-wide O-band spectral range.
2. Device configuration
Figure 1 shows the schematic diagrams of 1×8 channel MDI-type WDM optical filters based on (a) conventional DCs [6], (b) YS-MZ-WINCs [12], and (c) PS-MZ-WINCs. In either case, the 1×8 channel WDM filters require third-stage DIs to discriminate eight kinds of wavelength. The path length differences at each DI [shown as A, B and C in Fig. 1] are determined according to the relation shown in Table 1. The phase relation required for achieving the 1×8 filtering function is not limited to those shown in Table 1.
Fig. 1. Schematic diagrams of 1×8 channel multistage delayed interferometric (MDI) WDM filter schemes based on (a) directional couplers (DCs) [6], (b) Y-axis-symmetric Mach-Zehnder wavelength insensitive couplers (YS-MZ-WINCs) [12] and (c) proposed point-symmetric Mach-Zehnder wavelength insensitive couplers (PS-MZ-WINCs).
Table 1. Relation of optical path difference and amount of phase shift at each DI for three kinds of 1×8 channel WDM optical filters
For the device model shown in Fig. 1(b), the relation between output channels and their assigned wavelengths is different, because the phase relation between the first and second-stage DIs becomes out-of-phase [12]. It should be noted that the relation between output channels and their assigned wavelengths can be adjusted by choosing other phase shifter options. In either case, the channel spacing (Δν) is given by the path difference at first-stage DI, while the second and third stage DIs work as a spectral splicer to demultiplex as many wavelength components by reducing the path lengths differing by an approximately integer multiple and adding proper phase shifters. In this way, the relation between output channels and wavelength is given by the descriptions in Fig. 1(a).
As discussed in a previous section, the demultiplexing spectral behavior inherently has a wavelength sensitivity due to imbalanced splitting ratio of optical couplers used in each scheme. When it comes to using the YS-MZ-WINC in a WDM filter scheme, it influences not only optical splitting behavior but also relative phase relation within MDIs. This is the reason why the required amounts of phase shift at each DI in Table 1 are different from other schemes. In case of the proposed device, the phase shift at each DI is the same as the case shown in Fig. 1(a) since the PS-MZ-WINC does not have an effect on the relative phase relation within MDIs.
3. Theoretical analysis
Figure 2 shows top views of the DC and the WINC used in the 1×8 channel MDI-type WDM filters, the cross-sectional view of the waveguide and coupling region, together with the corresponding calculated coupling ratios for the DC and the WINC. For operating in O-band spectral range, the waveguide width (W) and the core layer thickness (T) were optimized to 340-nm and 200-nm for a single mode condition.
Fig. 2. (a) Top views of (a) DC and (b) WINC used in 1×8 channel MDI-type WDM filters, cross-sectional view of (c) waveguide and (d) coupling region, together with (e) the corresponding calculated coupling ratios for DC and WINC.
Since the spectral characteristics can be degraded by the degree of splitting ratio imbalance of optical couplers, we need to specify their spectral response in O-band regime. To validate optical coupling behavior against a wavelength in terms of the degree of wavelength sensitivity for the DC and the optimization of the phase shifter (δw) in the WINC for extending equivalent 3-dB splitting ratio, we performed an analytical calculation based on coupled mode theory and transfer matrix method [6]. Considering the Si-nanowire-waveguide DC with a bending curvature radius of 6 µm, we theoretically identified that the coupling coefficient as a function of wavelength [κDC(λ)] is given by 0.5×Sin(λ/34π) + 0.5 at the center wavelength λCN=1280 nm, which is comparable to the result obtained in our previous work [12], although the waveguide dimension (W = 440 nm and T = 220 nm) and operating wavelength range (C-band range) were quite different.
As discussed in [12], we optimized δw in the WINC to + 0.47π [radian] so as to extend equivalent 3-dB splitting range. As shown in Fig. 2(e), when we permit the deviation of ± 0.65dB from the perfect 3-dB splitting ratio, the WINC exhibited the available spectral range of >130-nm, which is nearly 4-times wider than the case of the DC (∼32-nm). Then, we theoretically analyze spectral characteristics of the 1×8 channel MDI-type WDM filters shown in Fig. 1. For simplicity, we assumed a lossless medium. The path length at each DI region was properly set by considering the equivalent index and dispersion property of the Si-nanowire waveguides. In this case, we set the Δν to ∼9-nm at λCN=1280 nm. Figure 3 shows the calculated spectral characteristics for the WDM filters based on (a) normal DCs, (b) YS-MZ-WINCs, and (c) PS-MZ-WINCs. δw was commonly set to + 0.47π [radian].
Fig. 3. Calculated spectral characteristics for 1×8 channel MDI-type WDM filters based on (a) normal DCs [Fig. 1(a)], (b) YS-MZ-WINCs [Fig. 1(b)], and (c) PS-MZ-WINCs [Fig. 1(c)].
Here, we define the filter operating condition when the device satisfy both <1dB of interchannel imbalance for the eight output ports and <–15dB of the spectral crosstalk at each grid. If we consider point-to-point optical links (i.e. Ethernet, datacenter application), total crosstalk within the integrated device is dominated by an incoherent crosstalk with negligible beat noises rather than a coherent crosstalk with considerable beat noises arising from the identical wavelength components. Thus, the crosstalk of <–15dB can be a figure of merit to assure a power penalty of <0.5dB to maintain error-free communication at a receiver side [15]. As shown in Fig. 3(a), the spectral response for the normal DC based WDM filter tends to be severely deteriorated as the wavelength is deviated by ± 27 nm (spectral range at around from λ1 to λ7) from the λCN, which is attributed to the significant asymmetric κDC, as shown in Fig. 2(e). This can also be physically interpreted that the device architecture provides an insufficient functionality for the cases of Δν >6.7 nm.
As seen in Fig. 3(b), when the YS-MZ-WINC is adopted, the available spectral range was estimated to be as wide as ± 35 nm. Although the WINC can markedly relax the problem on κDC(λ), the enhancement of the available spectral range was limited by + 16 nm. Moreover, despite of the same path difference relation (see Table 1) at each DI as the case shown in Fig. 3(a), each Δν for 8 channels tends to be slightly wider as the wavelength shifts to the shorter or longer waveguide side from the λCN. We understood as follows. When the WINC is positioned Y-axis-symmetrically at the front and back of each DI, it also influences over the optical path difference relation of WDM filter scheme as well as optical coupling characteristics. In other words, since the spectral responses including the DI and the WINCs with δW are correlated, the response cannot be considered as the simple combination of both contributions. Overall, the phase balance between MDIs are degraded, unless we exclude the aforementioned correlation contributions. Consequently, as shown in Fig. 3(b), spectral crosstalk and deviation of Δν tend to be prominent as the assigned wavelength is deviated from the λCN. Needless to say, the degree of spectral degradation becomes more significant as the NCh increases.
These drawbacks can be overcome by adopting the PS-MZ-WINC. As shown in Fig. 3(c), the WDM filter scheme exhibits the available spectral range of >130-nm. By locating the WINCs point-symmetrically at the front and back of each DI, the WDM filter scheme can completely remove the correlation between coupling characteristics and optical path differences. Therefore, the performance of the proposed WDM filter scheme is not limited by NCh. Needless to say, the advantage of the proposed scheme is not influenced by Δν as long as the equivalent 3dB split ratio is secured by the WINC structure.
4. Experimental demonstrations
4.1 Device fabrication
193-nm ArF-immersion lithography process was used to fabricate the proposed 1×8 channel MDI-type WDM filter on a 300-mm silicon-on-insulator (SOI) wafer with a 200-nm-thick Si layer and a 2-µm-thick buried oxide layer. The Si-nanowire waveguide width was adjusted to 340 nm for satisfying the single-mode condition in O-band regime. All the waveguide stripes were passivated by SiO2 after a dry etching step. In our previous work, by adopting high-precision process based on the ArF-immersion lithography, we were able to achieve very low propagation loss to be ≤0.5dB/cm in C-band spectral regime and ≤1.5 dB/cm in O-band spectral regime [16].
Figure 4 shows the top-view of the fabricated 1×8 channel third-stage WDM filter based on the PS-MZ-WINCs. The device size is 450-µm-long and 220-µm-wide, which means that the waveguide propagation loss is less than 0.09 dB in the WDM filter. That is, the influence of the waveguide propagation loss over the insertion loss of the device is negligible. We also fabricated a reference S-shaped bent waveguide (not shown) adjacent to the fabricated device for accurately estimating the excess loss within the filter architecture including several PS-MZ-WINCs. The input of the 1×8 channel WDM filter was set to the Input-3 [see Fig. 4]. In the following section, in order to accurately analyze the intensity and phase behaviors of each DI, we also characterize the filter spectral responses when we input the signal from the channel of Input-1, Input-2, Input-4 and Input-5.
Fig. 4. Top view of fabricated Si-nanowire 1×8 channel MDI-type WDM optical filter based on PS-MZ-WINCs.
At each PS-MZ-WINC, the phase shifters (δW1 and δW2) that were defined as the delayline (δL = 132 nm) were commonly set to be + 0.47π radian at λCN=1280 nm. Also, to define the Δν of 9-nm at λCN=1280 nm, ΔL was set to be 5 µm at the C1 region shown in Fig. 4. Other path difference and additional phase shift conditions at other DIs were adjusted according to the relation shown in Table 1. Once again, the Δν is a typical value and is not limited by the spectral window as long as the PS-MZ-WINC provides a required coupling response.
4.2 Measured spectral characteristics
The transmission characteristics were measured for the fabricated devices. As a light source, amplified spontaneous emission lights of three kinds of super luminescent emitting diodes were used to cover spectral range from λ=1240 to 1380 nm. Spectral response shorter than 1240 nm was not measured due to the limited operating range of optical sources. The polarization of input light was adjusted to be a linearly polarized TE-mode, and the output light was measured by a spectrum analyzer within a 140-nm-wide spectral range.
Figure 5 shows the measured spectra for the proposed 1×8 channel WDM filter based on the PS-MZ-WINCs. The transmittance depicted in Fig. 5 was estimated by normalizing the transmittance of the S-shaped reference waveguide. That is, the peak transmittance at each channel grid from 0-dB indicates the excess loss (excluding pure waveguide propagation loss) within the fabricated 1×8 WDM filter architecture. As clearly seen in Fig. 5, the fabricated device exhibited ∼9-nm-spaced 8 channel filtering response with good interchannel balance of <0.5dB for >100-nm-wide spectral range. The excess loss was estimated to be ∼0.4dB at the output channel 7 (best case) and ∼0.9 dB at the output channel 4 (worst case).
As mentioned in a previous section, considering the waveguide propagation loss of ≤1.5 dB/cm in the O-band spectral regime [16], the fabricated device with the ∼600-µm-long interaction lengths is expected to have the pure waveguide propagation loss of ≤0.09dB. This physically means that the total insertion loss of fabricated device is dominated by the excess loss within the filter architecture including several PS-MZ-WINCs rather than a waveguide propagation loss. The measured excess loss and interchannel balance were better than the recently reported results in MDI-type devices [10,11] or arrayed waveguide grating (AWG)-type devices [17,18].
As theoretically indicated in Fig. 3(c), we believe the device would still be stably working at the shorter wavelength range (λ<1240 nm), although spectral response was not measured due to the limited operating range of experimental setups. Nearly the same function could also be achieved within the spectral range shown in Fig. 5, if we adjust the λCN to 1310 nm. Since the PS-MZ-WINC operates as a 3dB coupler working in a wide wavelength range and gives little impact on the function of MDI, we can flexibly design the path length of the MDI and the number of DIs with different Δν (i.e. coarse WDM) and NCh (i.e. 4 or 16).
4.3 Analysis of the measured spectral response
The proposed device exhibited a low insertion loss and good interchannel balance. However, out-of-band spectral responses for each output channel were somewhat different from the theoretically expected ones shown in Fig. 3(c), which causes to increase spectral crosstalk. Since fabrication process is usually accompanied with some imperfections, spectral response of device might be deteriorated. To make these causes clear, we theoretically analyzed the experimental data.
Figure 6 shows the measured spectra for the fabricated device when the light was incident on (a) Input-1 or Input-5 (see Fig. 4), (b) Input-2, (c) Input-4, and their corresponding calculated spectra (d) fitted to Fig. 6(a), (e) fitted to Fig. 6(b), and (f) fitted to Fig. 6(c). It is noted that the dotted lines in Figs. 6(d)–6(f) indicate the ideal cases with no excess phase error (δW1= δW2= +0.47π [rad.]). The relation of output colors is the same as that shown in Fig. 1(c).
Fig. 6. Measured spectra for proposed 1×8 channel WDM filter when the light was launched from (a) Input-1 or Input-5, (b) Input-2, (c) Input-4, and their corresponding calculated spectra with fitting conditions: (d) fitted to (a), (e) fitted to (b), and (f) fitted to (c). Dotted lines in (d)-(f) indicate spectra with no excess phase deviation (δW1=δW2=0.47π rad.).
To begin with, in order to identify the main factor for influencing filter spectral shape and crosstalk, we characterized spectral response for the two single DIs (C1 and C4 at the third-stage DI). Based on the coupled mode theory, spectral response of the waveguide-type optical filter can be given by focusing on only the phase relations in each optical path, if the interaction loss is sufficiently short enough for us to neglect the contribution of the optical loss in the waveguide. Generally, the aforementioned phase relation becomes easy to be degraded as the interaction length becomes longer, simply because the probability of the excess phase errors in the waveguide by the waveguide width or the core thickness fluctuations becomes higher [19].
Compared with the calculated spectra for ideal case (dotted lines) shown in Fig. 6(d), the measured spectra [see Fig. 6(a)] shows a smaller spectral extinction ratio at each grid, which means that the split intensity balance for the PS-MZ-WINC was worse than the ideal case (nearly 50:50). Meanwhile, the relative wavelength differences between multiple sinusoidally-varying spectra tend to be matched with theoretical ones, which indicates that the relative phase relations between the two third-stage DIs is close to those shown in Table 1. That is, although the interaction length is much longer at the delayline (C1 and C4 at the third-stage DI), there was no significant phase errors by fabrication imperfections, which justifies our high precision process based on 300-mm waferscale ArF-immersion lithography technology [16]. Through these findings, we can understand that the main reason for the degraded extinction ratio is the split intensity imbalance at the PS-MZ-WINCs.
Based on these quantitative justifications, through iterative analytic calculations as a function of δW1 and δW2, we obtained moderate spectral agreement when we set δW1 and δW2 to be + 0.64π [rad.], +0.45π [rad.] at the PS-MZ-WINC with C1, and + 0.65π [rad.], +0.46π [rad.] at the PS-MZ-WINC with C4, as can be seen as the solid lines in Fig. 6(d). As a result, in spite of identical set value, only the δW1 was markedly deviated by >+0.17π [rad.], while the δW2 was calibrated to be comparable to the ideal value (<+0.02π [rad.]).
These trends were also confirmed at the second-stage DIs. As shown in Figs. 6(b) and 6(c), we observed spectral responses without including the first-stage DI by launching the light into Input-2 and Input-4. When we compared the measured results with calculated ones as shown in Figs. 6(e) and 6(f), the trends on the spectral extinction ratio at each grid and the spectral difference were similar to those shown in Fig. 6(d). Through these results, we clarified the following aspects. The excessive phase errors at each DI region (A∼C) were negligible enough not to influence over the spectral shape or spectral crosstalk. Also, spectral degradations by the unintentionally increased δW1 were reproducibly observed at all PS-MZ-WINCs, while the estimated values of δW1 and δW2 kept nearly constant. Therefore, we expect that the spectral degradation was mainly influenced by the deviation of δW1.
Through our analytic calculations, we verified that the excess phase error of +0.17π [rad.] causes the WINC split intensity ratio to be nearly 30:70. Other factors (i.e. DI regions (A∼C) and δW2) had little influence on spectral response. As described in our previous work in Ref. 16, the fabrication process technology we used gave highly constant controllability in waveguide width (σW = 1.3 nm) and core thickness (σH = 0.24 nm) across the whole 300-mm SOI wafer. Since the fabricated device size was sufficiently compact (450-µm-long and 250-µm-wide), we believe that the width and height variations of the waveguide core dimension did not prominentaly contribute to the spectral degradation.
In other words, the phase error by etched sidewalls with a typical standard deviation value (i.e. line edge roughness) has a randomly distributed statistical nature [19]. Considering this, the resultant phase error should be larger in a relatively longer path length region (i.e. including ΔL, 2ΔL, 4ΔL) rather than in a relatively shorter path length region (i.e. including phase shifter in WINCs). But, the estimated phase errors in the fabricated device did not show such kind of trend. Meanwhile, the phase change in waveguides can be generated by some other extrinsic reasons such as unexpected stitching errors of the waveguide patterns in photomask formation, lithography control condition or hard mask formation on top of the waveguide stripes for top-down etching process. If the equivalent waveguide widths between two or multiple optical paths within the interferometer regions are slightly different by virtue of some extrinsic reasons, it causes some portion of phase deviation from the ideal state in spite of nearly identical line edge roughness condition. However, it is extremely difficult to specify the reasons by usual measurement techniques such as SEM (scanning electron microscope) or TEM (transmission electron microscope), because these approaches marginally have data accuracy of line edge roughness. Increase in the number of samples will help us make some statistical data and let us bring to more accurate conclusion. More investigation for the deviation of δW1 will be needed in our future work.
Subsequently, in an attempt to simulate entire spectral response for the 1×8 WDM filter scheme, we mainly considered the deviation of δW1 in each PS-MZ-WINC. Figure 7 shows the calculated spectra for the proposed 1×8 channel WDM filter (a) when the phase shifter δW1 and δW2 was set to + 0.64π [rad.] and + 0.45π [rad.] at the PS-MZ-WINC with A, and other parameters were the same as those in Figs. 6(d)-(f), and (b) when all phase shifters δW1 and δW2 were set to + 0.50π [rad.] and + 0.45π [rad.], respectively.
Fig. 7. Calculated spectra for proposed 1×8 channel WDM filter with the light input from Input-3: (a) same fitting condition in Fig. 6, and (b) nearly ideal state (δW1=+0.50π [rad.], δW1=+0.45π [rad.])
As can be seen in Fig. 7, although the excess phase errors were not considered at each DI, the calculated spectra exhibited similar spectral shapes shown in the measured one (see Fig. 5). Low insertion loss with nearly constant parabolic spectral shapes over 100-nm-wide spectral range justifies that the excess phase errors at each DI region were sufficiently small enough not to influence on spectral response.
On the other hand, we made clear that the degraded spectral crosstalk at each grid was caused by an imbalanced splitting ratio for the first WINCs at each DI. Considering reproducible nature of the problem on δW1, it is expected that we could exclude this cause by optimizing the entire fabrication process. As shown in Fig. 7(b), nearly the same δW1 and δW2 (δW1=+0.50π [rad.], δW2=+0.45π [rad.]) enables us to fully suppress the spectral crosstalk of <–20dB at each grid, maintaining low excess loss and good interchannel balance over broader spectral range.
4.4 Discussion
Table 2 shows a brief performance comparison of several kinds of all-passive-type silicon-wire-based WDM filters from the viewpoint of channel count, available wavelength range, excess loss and spectral crosstalk. As seen in Table 2, compared with AWG-type scheme, it is evident that MDI-type scheme is more suitable for obtaining low excess loss, together with wideband operating window.
Table 2. Performance comparison for all-passive-type silicon-wire-based WDM filters from the viewpoint of channel count, operating wavelength range, excess loss and spectral crosstalk
Among the all-passive MDI-type WDM filters in Table 2, the proposed device exhibited potentially the widest operating wavelength range of ≥100-nm span and the lowest excess loss, along with the highest channel count. As a matter of course, the measured spectral crosstalk was not good enough to surpass the previously reported results. Since it was verified that this drawback originated from the specific local areas, we understand this drawback could be overcome by optimizing the fabrication process. Alternative efficient way to further lower the spectral crosstalk is to use the double filtering scheme by optically connecting the same MDI architecture one more time [6,20].
Meanwhile, recently, silicon-nanowire-waveguide MDI-type WDM filter based on external phase control was reported [21]. Accurate adjustment of the phase by monitoring the transmission state at each DI made it possible to realize 1×16 channel filtering operation with low crosstalk of <–20 dB, although the channel spacing was set to 50 GHz in a frequency domain. The proposed device scheme could also be applied to the externally phase controllable WDM filter applications. The proposed PS-MZ-WINC works as a highly efficient lower loss and broadband 3-dB optical coupler. Therefore, by combining the externally phase control scheme in the MDI, we could achieve more advanced WDM filter functionalities such as very narrowed channel spacing (<100-GHz) for dense WDM applications.
5. Summary
Novel 1×8 channel broadband operating multiple delayed interferometric (MDI) WDM filter based on point-symmetric Mach-Zehnder wavelength insensitive couplers (PS-MZ-WINCs) was proposed, theoretically analyzed and experimentally demonstrated. To verify broad operating range of the proposed device in O-band wavelength regime, we formulated analytic calculation models on a Si-nanowire-waveguide-type directional coupler (DC) and a WINC, and theoretically discussed from the viewpoint of a wavelength sensitivity of optical coupling ratio. Through the spectral comparison of 1×8 channel MDI-type WDM filter schemes, we showed that the PS-MZ-WINC works as a prerequisite identifier to have a wideband operating range and a spectral superiority without accompanying with an excess loss, interchannel imbalance and spectral crosstalk.
On a basis of the theoretical analysis, the proposed 1×8 channel WDM filter was fabricated by using a 300-mm SOI waferscale ArF-immersion lithography process. We were able to observe a 9-nm-spaced 1×8 channel demultiplexing spectra with low excess losses of 0.4∼0.9 dB and spectral crosstalk of <–10∼–15dB over >100-nm-wide wavelength range in O-band regime. Although the measured spectral response matched well with the calculated result in terms of an excess loss and in-band spectral shape against a wavelength, the out-of-band spectral shapes were quite different, and the measured spectral crosstalk was higher than the theoretically estimated one (<–15∼–25dB) at each grid. By making iterative analytic calculations fit in with the experimental data, we were able to verify that the principally contributing factor for the spectral degradations was the deviation of the phase change (δW1) located at the front portion for each PS-MZ-WINC rather than random phase errors at each DI. Moreover, we confirmed reproducible deviation amount of δW1 from the optimum value at each PS-MZ-WINC. Thus, the spectral response of the proposed 1×8 channel device could be improved by relaxing the problem on δW1. The advantages of the proposed WDM filter scheme are not restricted to the channel count and channel spacing as long as the effective 3-dB splitting ratio is secured.
Acknowledgement
This paper is based on results obtained from a project commissioned by the New Energy and Industrial Technology Development Organization (NEDO).
References
1. Y. Arakawa, T. Nakamura, Y. Urino, and T. Fujita, “Silicon photonics for next generation system integration platform,” IEEE Commun. Mag. 51(3), 72–77 (2013). [CrossRef]
2. M. Bohn, P. Magill, M. Hochberg, D. Scordo, A. Novack, and M. Streshinsky, “Next-generation silicon photonic interconnect solutions,” in Proc OFC 2019, (Optical Society of America, Washington, D.C., 1900), M3J.3
3. A. K. Krishnamoorthy, R. Ho, X. Zheng, H. Schwetman, J. Lexau, P. Koka, G. Li, I. Shubin, and J. E. Cunningham, “Computer systems based on silicon photonic interconnects,” Proc. IEEE 97(7), 1337–1361 (2009). [CrossRef]
4. A. Liu, L. Liao, Y. Chetrit, J. Basak, H. Nguyen, D. Rubin, and M. Paniccia, “Wavelength division multiplexing circuit on silicon-on-insulator platform,” IEEE J. Sel. Top. Quantum Electron. 16(1), 23–32 (2010). [CrossRef]
5. Y. A. Vlasov, “Silicon CMOS-integrated nano-photonics for computer and data communications beyond 100G,” IEEE Commun. Mag. 50(2), s67–s72 (2012). [CrossRef]
6. S.-H. Jeong, D. Shimura, T. Simoyama, M. Seki, N. Yokoyama, M. Ohtsuka, K. Koshino, T. Horikawa, Y. Tanaka, and K. Morito, “Si-nanowire-based multistage delayed Mach–Zehnder interferometer optical MUX/DeMUX fabricated by an ArF-immersion lithography process on a 300 mm SOI wafer,” Opt. Lett. 39(13), 3702–3705 (2014). [CrossRef]
7. S.-H. Jeong and Y. Tanaka, “Silicon-wire optical demultiplexers based on multistage delayed Mach–Zehnder interferometers for higher production yield,” Appl. Opt. 57(22), 6474–6480 (2018). [CrossRef]
8. T. Akiyama, T. Aoki, T. Simoyama, A. Sugama, S. Sekiguchi, Y. Sobu, S. Tanaka, Y. Tanaka, S.-H. Jeong, M. Nishizawa, N. Hatori, A. Hayakawa, and T. Mori, “Error-free loopback of a compact 25Gb/s × 4ch WDM transceiver assembly incorporating silicon (De)multiplexer with automated phase-error correction,” in Proc OFC 2018, (Optical Society of America, Washington, D.C., 1900), W1I.2
9. Z. Xiao, X. Luo, P. H. Lim, P. Prabhathan, S. T. H. Silalahi, T. Y. Liow, J. Zhang, and F. Luan, “Ultra-compact low loss polarization insensitive silicon waveguide splitter,” Opt. Express 21(14), 16331–16336 (2013). [CrossRef]
10. L. Chang, Y. Gong, L. Liu, Z. Li, and Y. Yu, “Low-loss broadband silicon-on-insulator demultiplexers in the O-band,” IEEE Photonics Technol. Lett. 29(15), 1237–1240 (2017). [CrossRef]
11. H. Xu and Y. Shi, “Flat-top CWDM (De)multiplexer based on MZI with bent directional couplers,” IEEE Photonics Technol. Lett. 30(2), 169–172 (2018). [CrossRef]
12. S.-H. Jeong and Y. Tanaka, “Silicon wire delayed interferometric wavelength division multiplexing filters operating in the broad spectral range,” J. Opt. Soc. Am. B 35(8), 1871–1877 (2018). [CrossRef]
13. K. Jinguji, N. Takato, Y. Hida, T. Kitoh, and M. Kawachi, “Two-port optical wavelength circuits composed of cascaded Mach-Zehnder interferometers with point-symmetrical configurations,” J. Lightwave Technol. 14(10), 2301–2310 (1996). [CrossRef]
14. J. Van Campenhout, W. M. J. Green, S. Assefa, and Y. A. Vlasov, “Low-power, 2 × 2 silicon electro-optic switch with 110-nm bandwidth for broadband reconfigurable optical networks,” Opt. Express 17(26), 24020–24029 (2009). [CrossRef]
15. G. P. Agrawal, Lightwave Technology: Telecommunication Systems (Wiley, 2005), chap. 9, pp. 351–353.
16. T. Horikawa, D. Shimura, H. Okayama, S.-H. Jeong, H. Takahashi, J. Ushida, Y. Sobu, A. Shiina, M. Tokushima, K. Kinoshita, and T. Mogami, “A 300-mm silicon photonics platform for large-scale device integration,” IEEE J. Sel. Top. Quantum Electron. 24(4), 1–15 (2018). [CrossRef]
17. S. Pitris, G. Dabos, C. Mitsolidou, T. Alexoudi, P. De Heyn, J. van Campenhout, R. Broeke, G. T. Kanellos, and N. Pleros, “Silicon photonic 8×8 cyclic Arrayed Waveguide Grating Router for O-band on-chip communication,” Opt. Express 26(5), 6276–6284 (2018). [CrossRef]
18. Y. Oonawa, H. Okayama, D. Shimura, H. Takahashi, H. Yaegashi, and H. Sasaki, “Polarisation insensitive wavelength de-multiplexer using arrayed waveguide grating and polarisation rotator/splitter,” Electron. Lett. 55(8), 475–476 (2019). [CrossRef]
19. T. Goh, S. Suzuki, and A. Sugita, “Estimation of waveguide phase error in silica-based waveguides,” J. Lightwave Technol. 15(11), 2107–2113 (1997). [CrossRef]
20. S. Assefa, S. Shank, W. Green, M. Khater, E. Kiewra, C. Reinholm, S. Kamlapurkar, A. Rylyakov, C. Schow, F. Horst, H. Pan, T. Topuria, P. Rice, D. M. Gill, J. Rosenberg, T. Barwicz, M. Yang, J. Proesel, J. Hofrichter, B. Offrein, X. Gu, W. Haensch, J. Ellis-Monaghan, and Y. Vlasov, “Cascaded Mach-Zehnder wavelength filters in silicon photonics for low loss and flat pass-band WDM (de-)multiplexing,” in Proceedings of 2012 IEEE International Electron Devices Meeting (IEEE, 2012), paper 33.8.
21. L. Han, B. P. Kuo, A. Pejkic, N. Alic, and N. Radic, “50 GHz silicon cascaded Mach-Zehnder wavelength filter and automatic phase error correction,” in Proc OFC 2019, (Optical Society of America, Washington, D.C., 1900), W3B.3
