Open Access
Add to BibSonomy
Abstract
We experimentally fabricate circular core 3-dimensional (3D) crossover single-mode polymer optical waveguides using a photomask free unique fabrication technique named the Mosquito method for realizing channel shuffling. The 3D crossover structure is accomplished by forming four cores (2 ch. × 2 ch.) with different heights: the last two channels cross over the first two channels with horizontal and vertical core bending. We compare the insertion losses between the fabricated 3D single-mode crossover waveguide and 3D S-bend core waveguides fabricated separately, which correspond to the lower and upper channels in the crossover waveguide. Then, we investigate the effect of the core crossover on the loss, and find that almost negligible additional loss is observed. The average insertion losses of this 6-cm long 3D crossover single-mode waveguide are 3.95 and 3.81 dB at 1310 nm, and 5.74 and 4.80 dB at 1550-nm wavelength, for the lower and upper channels, respectively. The interchannel crosstalk in this crossover waveguide is observed to be lower than −40 dB, while the 1 dB radial alignment tolerance is ± 1.7 and ± 2.1 µm at 1310 and 1550 nm, respectively. These results suggest that the fabricated circular core single-mode 3D crossover polymer waveguides could have a great impact for high-bandwidth-density on-board and inter-chip optical interconnect applications.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Optical interconnects are needed to meet the increasing bandwidth demands in modern data communication technologies because they are one of the potential solutions to improve wiring density and to reduce energy consumption [1] compared to copper-based electrical interconnects. In particular, there has been great anticipation for optical circuits on chips and on boards realized with optical waveguides. Optical waveguides composed of organic polymers have already drawn much attention for use in high-end servers/switches in data centers and high-performance computing (HPC) systems [2–4]. The current focus for optical interconnects is on the board-level and even on-chip interconnects, which are gradually developing with polymer waveguides. Polymer waveguides are versatile devices for short-reach on-board interconnects because they can easily be integrated on printed circuit boards (PCBs), and because of their possibilities in high-density wiring and low cost [5,6]. Multimode polymer waveguides are generally operated at an 850-nm wavelength with a combination of GaAs-based vertical cavity surface emitting laser (VCSEL) sources, which exhibit relaxed alignment tolerances.
If we can realize high-density optical circuits for on-board interconnects, channel-shuffling structures are needed for parallel processing operations [7] and switching networks [8]. We already demonstrated channel-shuffling multimode polymer waveguides [9] fabricated using the imprint method that we developed. This channel-shuffling polymer waveguide is the combination of 2D S-bend multimode rectangular cores, in which graded-index (GI) profiles are formed. Because of the tight optical confinement of the GI profiles [10], crossing losses at the core intersections are minimal, even if the crossing angle is as small as 10 degrees. Therefore, we reported that the GI core multimode waveguides could allow for a wide degree of freedom in channel-shuffling designs, namely crossing angles of the shuffled cores.
Meanwhile, we developed a unique fabrication method for polymer optical waveguides with GI cores named as the Mosquito method. The Mosquito method is used to easily fabricate multimode [11,12] polymer waveguides. The Mosquito method has an ability to form GI circular core. However, it should be noted that the core crossings on one plane are impossible to form by the Mosquito method because of its fabrication procedure, which is described in the later section. Alternatively, we have reported making the cores cross [13,14] using three-dimensional (3D) core alignment, because one of the strong features of the Mosquito method is the ability to form 3D wiring patterns using very compact equipment. Many researchers have demonstrated 3D waveguides for optical interconnects over the last couple of years. Therefore, various techniques, such as direct laser writing (DLW) [15,16] and femtosecond laser writing [17] have been used to fabricate 3D wiring patterns. However, the fabrication processes of these techniques are complicated and expensive. A key issue of this 3D crossover structure fabricated with the Mosquito method is that the cores are composed of horizontal and vertical bending with the optimum bending radius and crossing angle: this provides a height gap between the two (upper and lower) channel groups at the intersection. We reported earlier that we successfully fabricated 3D channel-shuffling multimode polymer waveguides to improve the wiring density for parallel signal transmission circuits as an application of 3D wiring pattern. Meanwhile, fiber flex circuits [18] have drawn much attention for interconnecting the systems with high-bandwidth-density wiring for cards in a shelf or shelf-to-shelf, however the fibers can break or exhibit high bending loss in short-reach applications, practically on PCBs. Therefore, an application of polymer waveguide to channel-shuffling is also reported in [19].
Meanwhile, single-mode polymer waveguides are a promising component to improve the connectivity to photonic integrated circuits (PICs) operating under the single-mode condition [20]. Actually, single-mode polymer waveguides are currently expected to be used as an interface between Si photonics chips and single-mode fibers (SMFs) [21] with adiabatic coupling. So, the research activity on single-mode polymer waveguides for on-board and inter-chip connections has resumed over the last couple of years, although intensive research on them had already been done 20-30 years ago. In order to couple to the Si-photonics chips and to SMFs, single-mode polymer waveguides are required to transmit the signal at the 1310 and 1550-nm wavelengths. Furthermore, single-mode polymer waveguides need to have almost the same mode-field diameters (MFDs) as those of SMFs for low-loss connections [22]. Therefore, single-mode 3D crossover polymer waveguides can be the potential components in on-board and inter-chip wiring circuits to enhance the bandwidth distance products even in short-reach applications.
In this paper, we design a low-loss crossover structure with single-mode cores as an example of 3D optical circuit and fabricate the designed crossover waveguide applying the Mosquito method to enhance the bandwidth and wiring-density for parallel signal transmission on PCBs and PICs. The single-mode 3D crossover structure is composed of horizontal and vertical core bending, including the optimal bending radius and crossing angle to add a height gap between two cores at the intersection.
2. The Mosquito method
A microdispenser (model no.: ML-808GX, Musashi Engineering Corp.) and desktop robot (model no.: SM300 DSS-3A, Musashi Engineering Corp.) are the main tools of the Mosquito method. The fabrication scheme of this technique is illustrated in Fig. 1. First, a liquid monomer for the cladding is spread on a glass substrate by filling the inside of a rectangular silicone-rubber frame fixed on the substrate. Then, another viscous liquid monomer for the core is dispensed into the cladding monomer from the needle-tip of a syringe by applying pressure with a microdispenser. Finally, both the liquid monomers are cured under UV light exposure and then post-baked. After curing, the obtained waveguide is carefully peeled off from the glass substrate. It should be mentioned that in order to fabricate single-mode polymer waveguides, we should apply UV light to cure the monomers immediately after completing the core monomer dispensing. Otherwise, the dispensed core monomer diffuses deep into the cladding monomer and expand the core diameter, which is not suitable for satisfying the single-mode condition. Hence, it is a vital issue that the temperature of both monomers, particularly the cladding monomer, be kept constant during the core dispensing so that the monomer diffusion between the core and cladding can be controlled. The controllability of the monomer diffusion leads to the variability of the index profile, which is another essential advantage of the Mosquito method. The other advantages of the Mosquito method are in the photomask-free procedure, cost-effective, the ability to form circular cores even with 3D wiring patterns, and the eco-friendliness because almost no materials are discarded. Meanwhile, we have already found some drawbacks of the Mosquito method, which are the difficulty in forming core crossing structures on one plane (two-dimensional), the difficulty in fabricating thin-film waveguide, and the low accuracy of core position. We are now investigating how to address these weak spots.
3. Design and fabrication
3.1 The single-mode conditions
We employ organic-inorganic hybrid resins named SUNCONNECT supplied by Nissan Chemical Corporation to fabricate the single-mode circular core polymer waveguides with 3D crossover structures. The core monomer is NP-005 (n: 1.575 at 1550-nm wavelength and viscosity is 76,000 cP), and the cladding monomer is NP-211 (n: 1.567 at 1550-nm wavelength and viscosity is 3,900 cP). An important aspect of these resins is the low absorption loss at 1550 nm inherent to common organic polymers, as we already reported in [23]. The low absorption loss is obtained by their chemical structures: the concentration of carbon-hydrogen (C-H) bonding per unit volume in these resins (particularly the resin for the core) is lower than those of the other organic polymers applied to fabricate the conventional optical waveguides. The absorption loss due to C-H stretching vibration inherent to the material could decrease. The key parameters to satisfy the single-mode condition are the core diameter, the relative refractive index difference (Δ), and the operating wavelength. The Δ of the materials used is calculated to be 0.507%. The single-mode conditions at 1550 and 1310-nm wavelengths are simulated using a commercially available mode solver, FIMMWAVE. Here, as we already reported, in the cores of the waveguide fabricated with the Mosquito method, parabolic GI profiles could be formed. So, a parabolic index profile is assumed for the simulation by approximating the profile with the well-known power-law form having an index exponent g of 2.0. It is well known that the requirement for the core diameter to satisfy the single-mode condition can be relaxed by forming such a parabolic profile. Further, the cut-off core diameter for the step-index (SI) circular core single-mode waveguides is also calculated following the above parameters. From Fig. 2, it is revealed that the core diameter for GI circular core should be smaller than 12 and 14 µm under 1310 and 1550-nm wavelength operations, respectively, to satisfy the single-mode condition. However, for SI circular core, the core diameter should be smaller than 6 and 7 µm at 1310 and 1550-nm wavelengths, respectively. Hence, it is obvious that the core can be relaxed almost 6 or 7 µm for the GI circular core compared to the SI circular core. It was already experimentally confirmed that we could control the core diameter by adjusting the following parameters in the Mosquito method: the needle-scan velocity, the dispensing pressure, and the inner diameter of the needle.
Fig. 2. Condition of single-mode circular core diameters: (a) GI circular core and (b) SI circular core.
3.2 Design of the 3D crossover structures
First, we introduce a basic sample design of the 3D crossover structure to set the structural parameters. A general structure with 3D crossover cores is shown in Fig. 3, where all the channels are bent both horizontally and vertically to provide a suitable height gap at the intersection, however, they are not simultaneous. This structure is developed by combining S-bend structures. The details of the S-bend structure are shown in Fig. 4(a). In order to realize the channel-shuffling structure shown in Fig. 3, lower channel-group (green) is dispensed first, followed by the upper one (yellow). The interchannel pitch in the horizontal direction is set to 250 µm to preserve a high connectivity to widely deployed single-mode fiber ribbons. In order to realize a low-loss waveguide, the structural parameters such as the bending radius (r shown in Fig. 3(c)) and the height gap (h1+h2 in Fig. 3(b)) are investigated to determine the optimum 3D crossover structures. We already confirmed in [24] that if the height gap in the intersection was too small, the optical loss increased. Meanwhile, the bending radius is an important parameter to keep the total loss as low as possible. In addition, how we can minimize the height gap is another key issue to make the waveguide compact.
Fig. 3. Sample design of the 3D crossover structures: (a) top view, (b) side view, and (c) magnified image of S-bend curves.
Fig. 4. Measurement setup for bending loss: (a) S-bend waveguide with bending angle 5°, and (b) straight waveguide.
First, we investigate the smallest bending radius by fabricating and simulating 3D S-bend waveguides in which the bending radius is varied. The S-bend structure shown in Fig. 4(a) is explained here. The coordinates of points A to F in the S-bend structures (needle-tip path) are shown in Table 1, where the origin is set at point A. From A to B, the cores are straight on the same height and parallel to the z-axis. Then, from B to C, the core ascends to the upper level on a straight slope maintaining the x coordinate same as those of A and B, and it connects to an arc of CC′ with variable radii in a range of 5-40 mm, and 5-degree central angle. Since this arc is placed on the same height as C, the core is bent only in the horizontal direction. From C′ to D, the core remains on the same height with a straight line and changes the x and z coordinates, and it also connects to an arc of DD′ on the same height. Finally, it descends to the original height from D′ to E on a straight line, followed by another straight line of EF parallel to the z-axis. The bending loss simulation is carried out using the beam propagation method (BPM). In addition, the optimum bending radius is also experimentally evaluated. The measurement conditions for the insertion loss are shown in Figs. 4(a) and 4(b), with a 1550-nm laser diode (LD) as a light source. Waveguide samples of S-bend core with different bending radii are fabricated, and the bending loss is evaluated at 1550 nm. Here, an SMF ((PA-A2): from Sumitomo Electric Industries Ltd., with 7.9 and 8.0-µm mode-field diameters (MFDs) along with the horizontal and vertical directions, respectively is used as a launching probe, while a 105-µm core step-index multimode fiber (105-SI MMF) is for detection. From the bending loss dependence on the bending radius in Fig. 5, we find that the simulated bending loss gradually decreases with the increase of bending radius. The same tendency is also observed in the experimental result as shown in Fig. 5. It should be noted that the experimental bending loss is much higher than that of the simulated when the bending radius is smaller than 15 mm because the needle tip cannot accurately follow the smooth curves under the small bending portion due to the high scan speed (80 mm/sec), and the deformation of the cores from the circular shape is observed due to such an abrupt motion of the needle scan. However, when the bending radius is larger than 15 mm, a good agreement is observed between the experimental and simulated results. Therefore, we set the suitable bending radius as 20 mm (including 5 mm margin) for the 3D crossover single-mode polymer waveguide to reduce the excess bending loss.
Table 1. Coordinates in mm of the points assigned in Fig. 4(a) when bending radius and angle are 10 mm and 5°, respectively
Next, the height gap to provide for the two crossing cores is discussed. The optimum height gap is investigated by fabricating several waveguide samples, as preliminary experiments. Here, it should be easily understood that the loss of the lower cores decreases with increasing the height gap, because the influence of the upper channel deceases, in particular the liquid monomer flow due to the needle scan to dispense the upper cores is weaker with the increase of the height gap. Hence, sufficiently large height gap contributes to decrease the loss to the minimum value. However, we also have another limitation that is the waveguide thickness: with increasing the thickness, the effect of volume shrinkage during UV curing increases resulting in warpage. In the case of the single-mode waveguide, even a couple of µm warpage leads to high coupling loss at the waveguide edges. Therefore, the height gap is required to be as small as possible, which is a trade-off relation with the loss.
Finally, we reach a solution for the crossover structures illustrated in Fig. 6: Fig. 6(a) indicates the top-view pattern of the over-crossing waveguide. This structure is the actual needle-tip scanning pattern in the Mosquito method, while the numbers in Fig. 6(b) are the parameters (length) in mm. In this design, the last two channels (Ch. 3 and 4 illustrated in orange) cross over the first two (Ch. 1 and 2 in blue) by proving a height gap at the intersection. The crossing angle is set to 12° because larger crossing angle is not suitable to form a crossover structure in a compact footprint. In Fig. 6(b), how the height gap is formed at the intersection between the lower and upper channel groups is illustrated. The important points A to F in order to express the core structure are assigned in Figs. 6(a) and 6(b), and their coordinates on Ch. 2 and 3 are indicated in Table 2. Here, the origin is set the point on line A, in the middle of Ch. 2 and 3 on a height of 0.375 mm. The core bending in the vertical direction (both upward and downward) is required to provide a suitable height gap at the intersection for the upper and lower cores. Here, the needle-tip starts scanning at a 0.375-mm height from the substrate, at point A in Fig. 6(b), which is defined as the standard height. For the 1st core group (Ch. 1 and 2), from B to C the needle-tip path descends to a position of 0.115 mm lower than the standard height along a 15-mm lateral distance tracing on a straight path, then, from C to C′, form a horizontal bending following to an arc. Then, the needle-tip path keeps the same height along a straight line at the intersection, and finally it gradually ascends to the standard height on a straight slope over a lateral distance of 15 mm.
Fig. 6. Schematic design of the 3-dimensional channel-shuffling needle-tip paths: (a) top view and (b) side view.
Table 2. Coordinates in mm of the points assigned on Ch. 2 and Ch. 3 in Fig. 6(a)
Similarly, for the last channel group (Ch. 3 and 4), the needle-tip path ascends to a position of 0.125 mm higher than the standard height along a 15-mm distance on a straight line, and then form a horizontal bending following to an arc. The needle-tip path also remains the same height on a straightaway at the intersection and finally moves back to the standard height with a straight slope over a lateral distance of 15 mm.
It is investigated that in the 3D crossover waveguides, if the height gap increases from the designed value of 0.240 mm as shown in Fig. 6(b), it requires more severe vertical bending, which could lead higher bending loss. Furthermore, the larger height gap requires thicker waveguide, which increases the influence of the volume shrinkage as aforementioned. Hence, in order to fabricate a thin crossover waveguide, the optimum height gap is determined to be 0.240 mm. Here, it should be noted that the upper channels slightly influence the lower channels in the case of symmetric height gap crossover structure. In order to further decrease the upper core effect, the upper cores in this crossover waveguide are dispensed slightly higher position (10 µm) compared to the symmetric one. Besides, the bending radius is set to be larger than 15 mm to reduce the additional bending loss, as shown in Fig. 5. Therefore, this 3D crossover structure has a crossing angle and bending radius of 12° and 20 mm, respectively, while the height gap at the intersection is 0.240 mm (240 µm) and the inter-core pitch is 250 µm.
3.3 Fabrication for the 3D crossover single-mode polymer waveguide
After carrying out several preliminary experiments for parameter optimization, we decide to employ a constant needle scan velocity of 80 mm/s and a dispensing pressure of 500 kPa, using a needle with 125/230-µm inner/outer diameters, respectively. The cladding monomer on the substrate is maintained at approximately 18 °C. The top view of the fabricated single-mode 3D crossover polymer waveguide is shown in Fig. 7, where the last two cores cross over the first two cores maintaining a 0.240-mm height gap at the intersection. A digital microscope (KEYENCE CORPORATION, VHX-5000) is used to observe the cross-sections of this waveguide. The cross-sectional images of the start and end facets (of dispensing directions) are shown in Figs. 8(a) and 8(b), respectively. From the cross-sectional images, it is evident that the cores are aligned almost linearly on the same height from the substrate at both end facets of the waveguide. The cores look almost perfectly circular with a diameter of approximately 10.8 µm, and the interchannel pitch is controlled to be 250 µm.
Fig. 8. (a) Start and (b) end facets of the cross-sectional images, and (c) magnified image of one core.
3.4 Fabrication for the 3-dimensional S-bend lower and upper core waveguides
We also fabricate 3D S-bend lower and upper channel single-mode polymer waveguides separately to evaluate the impact of crossover structures on the optical properties. These S-bend lower and upper channel waveguides are fabricated according to the same fabrication conditions as those for the 3D crossover single-mode waveguides mentioned above. The fabricated cores of the 3D S-bend lower and upper channel waveguides are also circular, and the core diameter is approximately 10.8 µm. The cross-sectional images of the lower and upper channel waveguides and their magnified cores are shown in Figs. 9(a)–9(d).
Fig. 9. Images of cross-sections and magnified cores: (a) and (b) are for the lower channel waveguide, and (c) and (d) are for the upper channel waveguide, respectively.
4. Characterization
4.1 Near-field patterns
Near-field patterns (NFPs) are a key for analyzing the MFDs and optical intensity distribution inside the core. We measure the NFPs using a mode profiler (SYNERGY OPTOSYSTEMS CO., LTD., M-Scope Type S) at 1550 and 1310-nm wavelengths on the output side of the waveguides, where the same SMF, which is used for the insertion loss measurement in Fig. 4, is butt-coupled to the core by minimizing the air-gap between the waveguide and the probe. The normalized output intensity profiles of this single-mode 3D crossover waveguides are shown in Fig. 10. The MFDs determined from the NFPs at 1/e2 of peak intensity are obtained along with the X and Y-directions, which are 7.7 µm and 7.6 µm, respectively at 1550 nm, while 6.8 µm and 6.6 µm, respectively at 1310 nm. We also understand from Fig. 10 that the normalized output intensity curves approximate to a Gaussian profile. Besides, the NFPs of the 3D S-bend lower and upper channel waveguides are also measured using the same setup as that for Fig. 10. The measured MFDs for 3D crossover and 3D S-bend single-mode waveguides along with the X and Y-directions are summarized in Table 3. We confirm that the MFDs are precisely controlled to the desired values, not only in the S-bend core waveguides, but also in the 3D crossover waveguides. Here, in Table 3, we find that the measured MFDs in the X-direction are slightly larger than those in the Y-direction in many channels. This is because the cores have a small dip on the top, as shown in the magnified image of the core cross-section in Fig. 8(c). We already found that the combination of core and cladding monomers as well as the dispensing condition could make the dip on the top. Meanwhile, under the different dispensing conditions, even elliptic cores could be formed. However, we already found a way to address these core deformation issues, which will be published elsewhere.
Fig. 10. NFPs and Gaussian curve fitting to the normalized intensity profiles at (A) 1550 nm and (B) 1310 nm.
Table 3. The measured MFDs for waveguides and fiber
The MFDs of the SMF used are measured in the horizontal and vertical directions and also listed in Table 3. The MFDs of the single-mode 3D crossover waveguide shown in Fig. 10 are almost the same as those of the SMF, which leads to low coupling loss between them.
The MFD with respect to the core diameter is calculated using FIMMWAVE at 1310 and 1550 nm by approximating a parabolic index profile with power-law form in a circular core, as shown in Fig. 11. Meanwhile, the experimentally measured MFDs of the crossover single-mode waveguide are also shown with respect to the measured core diameters (approximately) in Fig. 11. Here, the MFDs measured in the horizontal directions are employed for the analysis. In Fig. 11, the calculated MFD is in a range between 7.3 and 7.4 µm at 1550 nm, while 6.5 to 6.6 µm at 1310 nm, when the core diameter is controlled to a range between 10.5 to 11.0 µm. Meanwhile, the experimental results show that the MFD at 1550 nm is between 7.6 to 7.8 µm and between 6.7 to 6.9 µm at 1310 nm. It is found that the experimental MFDs show just a 0.4 µm difference from the calculated values at maximum. Since it is very difficult to measure the core diameter accurately from the cross-sectional images of the digital microscope, such a small difference shown in Fig. 11 could be within a measurement error.
4.2 Insertion Loss
4.2.1 Insertion loss of single-mode 3D crossover polymer waveguide
Since the single-mode 3D crossover waveguide consists of horizontal and vertical core bending, it is very difficult to measure the propagation loss applying the cut-back method. So, we measure the insertion loss instead, which is the combination of the propagation loss, the coupling loss, and the bending loss. The insertion loss is measured under the same condition as those explained in the above sections. Here, two different LDs at 1550 and 1310-nm wavelengths are used as the light source. On the detection side, a 105-SI MMF is used for the probe, by which the coupling loss from the waveguide to the detection probe is negligible because the 105-SI MMF receives all the output from the waveguide core. Meanwhile, an SMF (the same one as the launching side) is used as the detection probe in order to assess the real connectivity of the fabricated single-mode waveguides with SMF links for off-chip interconnects. The results of insertion losses for a 6-cm long 3D crossover single-mode polymer waveguide are shown in Fig. 12.
Fig. 12. Insertion losses for 6-cm long 3D crossover single-mode waveguide at (a) 1310 nm and (b) 1550 nm.
The average losses for the lower channels (Ch.1 and 2) and upper channels (Ch. 3 and 4) of this single-mode crossover waveguide at both wavelengths are summarized in Table 4. At both wavelengths, the substitution of the detection probe from the 105-SI MMF for the SMF causes 0.70 to 0.90 dB extra loss, due to the coupling loss (including the Fresnel reflection). We also compare these results to those of the separately fabricated 3D S-bend lower and upper channel waveguides to evaluate the additional loss due to the crossover, which is discussed in the next section.
Table 4. Insertion losses of the 3D crossover single-mode waveguide
4.2.2 Insertion loss of the 3D S-bend lower and upper channel waveguides
We evaluate the insertion losses of the 3D S-bend waveguides under the same technique, as was employed in the crossover waveguide. The losses of the S-bend lower and upper channels are measured under the SMF/SMF condition only. The results are presented in Table 5. The average losses for the lower and upper channels at 1310 nm are 3.90 dB and 3.76 dB, respectively, whereas the losses at 1550 nm are slightly higher: the average losses for the lower and upper channels are 5.59 dB and 4.61 dB, respectively.
Table 5. Insertion losses of the 3D S-bend waveguides
The average results of the 3D S-bend waveguides are shown in Fig. 13, which can be compared to the loss of the 3D crossover single-mode waveguide, in Fig. 12. The losses for both upper and lower channels in the 3D S-bend waveguide are almost the same as those in the 3D crossover waveguide. From the results, it is noted that there is no significant difference in the insertion losses between the waveguides with and without the over-crossing cores. So, the crossover structure causes negligible excess loss if the crossover structure is optimally designed. Furthermore, the insertion loss for a sample straight core polymer waveguide with the same length (6-cm long) as that of the 3D crossover one is measured to be 3.35 and 4.63 dB at 1310 and 1550-nm wavelengths, respectively, very comparable to the loss of the 3D crossover waveguides. Thus, the 3D crossover single-mode polymer waveguide would be a promising component to increase the wiring density realizing the optical circuits for PCBs and PICs.
4.3 Alignment tolerance analysis
The alignment tolerance (AT) is a very important characteristic to connect the single-mode waveguides to light sources, PICs, fiber ribbons, and detectors. The measurement setup for the AT of the waveguides is illustrated in Fig. 14. To measure the AT on the output end, the launch probe (the same SMF as that for the insertion loss measurement) is precisely aligned to couple the light into the waveguide core, while another SMF probe is also aligned precisely to the launched core at the output end by monitoring the output power. Then, the probe at the output end scans horizontally with a step of 0.1 µm as shown by the arrow in Fig. 14. The ATs are measured at 1310 and 1550 nm, and the obtained loss curve is shown in Fig. 15. It is found that the 1 dB alignment tolerance is ± 2.1 µm at 1550 nm, whereas ± 1.7 µm at 1310 nm. The AT difference between the two wavelengths stems from the MFD difference between them, as shown in Fig. 11. The observed AT is comparable to the commonly employed SMFs, and thus the single-mode waveguides fabricated could be coupled to SMF with low coupling loss.
4.4 Interchannel crosstalk
We already confirmed from the aspect of the insertion loss in the above section that the 3D crossover waveguides fabricated in this paper would be applicable to the high-density optical circuits. Here, the interchannel crosstalk is a concern in the 3D crossover single-mode polymer waveguides. Hence, the crosstalk is evaluated using the same experimental setup as shown in Fig. 14. To evaluate the crosstalk, the launch probe is precisely aligned and fixed, and the output SMF scans along with the horizontal direction to the neighbor cores with a step of 1 µm to record the output power. The results of the crosstalk for the single-mode crossover waveguide are shown in Fig. 16. It is obvious that the crosstalk (Ch. 3 is launched) of this 6-cm long crossover single-mode polymer waveguide is less than −40 dB at both 1310 and 1550 nm, even to the nearest channel. On the other hand, although this crossover waveguide has the horizontal and vertical core bending, no leaky light is observed from Ch. 3, and the light re-coupling from the nearest core does not occur. This low crosstalk indicates that this 3D crossover single-mode waveguide is acceptable for high-density optical interconnect applications.
Fig. 16. Interchannel crosstalk for 6-cm long single-mode 3D crossover waveguide at (a) 1310 nm and (b) 1550 nm.
5. Conclusion
We successfully fabricated 3D crossover circular core single-mode polymer waveguides applying the Mosquito method. This crossover structure is fabricated by dispensing the 3D S-bend lower and upper channel groups with a height-gap at the intersection. The insertion loss of the 3D crossover waveguide is very close to that of the separately fabricated 3D S-bend lower and upper channel waveguides. The average insertion losses for this single-mode crossover waveguide at 1310 nm are 3.95 dB and 3.81 dB, and at 1550 nm are 5.74 dB and 4.80 dB for lower and upper channels, respectively. The interchannel crosstalk is less than −40 dB, which is much more acceptable for parallel signal transmission. Therefore, this 3D crossover single-mode waveguide could be a new and original device to promote high-density wiring with the parallel signal transmission for on-board and inter-chip optical interconnects.
References
1. D. A. B. Miller, “Energy consumption in optical modulators for interconnects,” Opt. Express 20(S2), A293–A308 (2012). [CrossRef]
2. N. Bamiedakis, F. Shi, D. Chu, R. V. Penty, and I. H. White, “High-speed data transmission over flexible multimode polymer waveguides under flexure,” IEEE Photonics Technol. Lett. 30(14), 1329–1332 (2018). [CrossRef]
3. T. Bechtold and D. Hohlfeld, “Multi-physical simulation of high performance computing platform integrating polymer waveguides,” in Proceeding of IEEE conference on Thermal, Mechanical and Multi-Physics Simulation and Experiments in Microelectronics and Microsystems, (IEEE, 2016), pp. 1/6–6/6.
4. N. Bamiedakis, J. Chen, P. Westbergh, J. S. Gustavsson, A. Larsson, R. V. Penty, and I. H. White, “40 Gb/s data transmission over a 1-m long multimode polymer spiral waveguide for board-level optical interconnects,” J. Lightwave Technol. 33(4), 882–888 (2015). [CrossRef]
5. F. E. Doany, C. L. Schow, B. G. Lee, R. A. Budd, C. W. Baks, C. K. Tsang, J. U. Knickerbocker, R. Dangel, B. Chan, H. Lin, C. Carver, J. Haung, J. Berry, D. Bajkowski, F. Libsch, and J. A. Kash, “Terabit/s-class optical PCB links incorporating 360-Gb/s bidirectional 850 nm parallel optical transceivers,” J. Lightwave Technol. 30(4), 560–571 (2012). [CrossRef]
6. N. Bamiedakis, J. D. Ingham, R. V. Penty, I. H. White, J. V. DeGroot Jr., and T. V. Clapp, “Multimode polymer waveguides for high-speed optical interconnects,” Proc. SPIE 10567, 146 (2006). [CrossRef]
7. H. S. Stone, “Parallel processing with the perfect shuffle,” IEEE Trans. Comput. C-20(2), 153–161 (1971). [CrossRef]
8. P. Y. Chen, D. H. Lawrie, P. C. Yew, and D. A. Padua, “Interconnection networks using shuffles,” Computer 14(12), 55–64 (1981). [CrossRef]
9. K. Abe, Y. Oizumi, Y Taira, and T. Ishigure, “Low loss channel-shuffling polymer waveguides: design and fabrication,” in Proceeding of IEEE Electronic Components and Technology Conference (IEEE, 2017), pp. 526–531.
10. Y. Koike, Y. Takezawa, and Y. Ohtsuka, “New interfacial-gel copolymerization technique for steric GRIN polymer optical waveguides and lens arrays,” Appl. Opt. 27(3), 486–491 (1988). [CrossRef]
11. H. Toda and T. Ishigure, “Index profile design of graded-index core tapered polymer waveguide for low loss light coupling,” in Proceeding of IEEE CPMT Symposium Japan, (IEEE, 2016), pp. 149–150.
12. R. Kinoshita, D. Suganuma, and T. Ishigure, “Accurate interchannel pitch control in graded-index circular-core polymer parallel optical waveguide using the Mosquito method,” Opt. Express 22(7), 8426–8437 (2014). [CrossRef]
13. O. F. Rasel and T. Ishigure, “Fabrication for 3-dimensionally shuffled polymer waveguide with GI circular core using the Mosquito method,” in Proceeding of IEEE Photonics Conference (IEEE, 2017), pp. 657–658.
14. O. F. Rasel, A. Yamauchi, and T. Ishigure, “Low-loss 3-dimensional shuffling graded-index (GI) polymer optical waveguides for optical printed circuit boards,” IEICE Trans. Electron. E101.C(7), 509–517 (2018). [CrossRef]
15. A. Landowski, D. Zepp, S. Wingerter, G. Freymann, and A. Widera, “Direct laser written polymer waveguides with out of plane couplers for optical chips,” APL Photonics 2(10), 106102 (2017). [CrossRef]
16. M. Schumann, T. Buckmann, N. Gruhler, M. Wegener, and W. Pernice, “Hybrid 2D–3D optical devices for integrated optics by direct laser writing,” Light: Sci. Appl. 3(6), e175 (2014). [CrossRef]
17. J. Lv, X. Hao, and F. Chen, “Three-dimensional waveguide coupler/beam splitter in lithium niobate crystals by femtosecond laser writing,” in Proceeding of Conference on Lasers and Electro-Optics (CLEO), (OSA, 2017), paper JTu5A.72.
18. Y. Abe, M. Kobayashi, M. Hirayama, and R. Nagase, “scalable optical fiber wiring system for over 10,000-fiber shuffler,” IEICE Trans. Electron. E88-C(8), 1755–1763 (2005). [CrossRef]
19. A. R. Dangel, F. Horst, D. Jubin, N. Meier, J. Weiss, B. J. Offrein, B. W. Swatowski, C. M. Amb, D. J. DeShazer, and W. Ken Weidner, “Development of versatile polymer waveguide flex technology for use in optical interconnects,” J. Lightwave Technol. 31(24), 3915–3926 (2013). [CrossRef]
20. S. H. Tao, J. Song, Q. Fang, M. B. Yu, G. Q. Lo, and D. L. Kwong, “Improving coupling efficiency of fiber-waveguide coupling with a double-tip coupler,” Opt. Express 16(25), 20803–20808 (2008). [CrossRef]
21. R. Dangel, J. Hofrichter, F. Horst, D. Jubin, A. La Porta, N. Meier, I. M. Soganci, J. Weiss, and B. J. Offrein, “Polymer waveguides for electro-optical integration in data centers and high-performance computers,” Opt. Express 23(4), 4736–4750 (2015). [CrossRef]
22. R. Dangel, A. L. Porta, D. Jubin, F. Horst, N. Meier, M. Seifried, and B. J. Offrein, “Polymer waveguides enabling scalable low-loss adiabatic optical coupling for silicon photonics,” IEEE J. Sel. Top. Quantum Electron. 24(4), 1–11 (2018). [CrossRef]
23. K. Yasuhara, F. Yu, and T. Ishigure, “Circular core single-mode polymer optical waveguide fabricated using the Mosquito method with low loss at 1310/1550 nm,” Opt. Express 25(8), 8524–8533 (2017). [CrossRef]
24. O. F. Rasel and T. Ishigure, “3-dimensional channel-shuffling single-mode polymer waveguides: design and fabrication,” in Proceeding of IEEE CPMT Symposium Japan, (IEEE, 2018), pp.143–144.
