A low-cost, versatile optical coupling structure featuring a monolithic integration of a polymeric waveguide, beam ducts, and end-reflectors has been designed, prototyped, and demonstrated to be capable of sustaining high misalignment errors whilst maintaining a reasonable coupling efficiency. The soft lithography fabrication process of this interconnection design allows for significant advantages over traditional designs in terms of misalignment tolerance, manufacture cost and speed, as well as 3-D integration capability.
© 2005 Optical Society of America
In the microelectronics industry, the exponential growth of computing hardware today requires a dramatic increase in the throughput of electronic components for purposes such as data transmission between the central processing unit (CPU) and its peripheral chips. Although the microelectronics industry has steadily improved the performance of conventional electronic circuits to solve such problems, traditional printed circuit boards (PCBs) inevitably suffer inherent problems such as crosstalk, limited line length, and poor flexibility in layout design, as bitrates between chips with large scale integration (LSIs) increase to the magnitudes of a few gigabits per second.
Optical chip-to-chip interconnection is a promising method of providing interconnections between LSIs because of the capability for high bandwidth and high density of optical links, as well as the fact that it is unaffected by electromagnetic interference (EMI) of electronic connects. Recent efforts in this field have concentrated on integrating optical interconnections into conventional PCBs [1–6]. In particular, polymer waveguides have been integrated into PCBs as an optical layer [3, 5, 7] to produce hybrid electro-optical PCBs. This, however, is hindered by difficulties. Firstly, there is a significant difficulty during assembly of providing adequate alignment precision between the vertical-cavity surface-emitting laser (VCSEL) (or the photodiode, PD) and the waveguide layer as to obtain a link with acceptable optical losses. Secondly, the divergence-induced loss of the beam from the VCSEL into the optical layer embedded in thick PCB layers, and from the optical layer to the PD window, is also significant. Several efforts have been made to employ various optical elements to connect and alter the light paths, including the use of a 45°-ended waveguide, a micro-lens , optical coupler , and micro-prisms . Yet the high misalignment sensitivity and complex packaging requirements still remain with such designs.
As an effort to solve the aforementioned problems, a novel structure for chip-to-chip optical interconnection applications is proposed in this paper. Soft lithography techniques [9, 10] are utilized to fabricate proposed structures at a high efficiency and low cost.
Most prior designs of optical interconnections usually call for a high-precision VCSEL/PD alignment, often with an alignment-error requirement of less than a few micrometers; this exhibits an immediate difficulty in the assembly of electro-optical PCBs. In order to increase the tolerance of the interconnection waveguide to alignment errors, we present a waveguide architecture of two beam ducts integrated into both ends of a straight rectangular waveguide. Fig. 1 displays the ducts, as well as other elements involved in providing high-speed optical communications within a board; this includes a VCSEL, waveguide couplers, and a polymer-based channel waveguide functioning as the optical layer. The driving electrical pulses modulate the VCSEL, and the light received at the photodiode through the waveguide demodulates back as electric signals on the surface of the PCB. The specialized integrated beam ducts in the design allows for greater misalignment tolerances in the VCSEL/waveguide coupling and in the waveguide/PD coupling; the acceptable error can be in the magnitude of several hundred micrometers. Traditionally, the micro-fabrication of such a three-dimensional optical polymer structure is costly, but thanks to the recent availability of soft lithography techniques, the transfer of 3-D features onto polymers can be simplified to a mold-printing process while maintaining a surface roughness of as low as several tens of nanometers .
3. Architectures and ray tracing simulations
As shown in Fig. 1, the optical components are structured into an optical connection through the PCB, and coupling between the VCSEL (or PD) and polymeric waveguide occurs through the substrate. The light is first emitted from the VCSEL; then, its path is altered by 90° at a 45° mirror plane and thus reflected into a beam duct. Assisted by the duct, the beam subsequently travels into the straight waveguide with minimal loss, and again expands at the opposite beam duct. From there it reflects down perpendicularly to reach the PD as an optical signal.
The key difference between previous designs and our current design lies on the structure of the two ends of the polymeric waveguide. Figure 2 displays a comparison among the designs. Structure type-I was proposed to utilize a rectangular waveguide and two 45° end-mirrors with a uniform cross-sectional dimension of 50×50μm . However, the manufacture of this design required complex micromachining to create its optical surfaces while mounted on the PCB: a difficult task when embedded in circuits. Type-II architecture introduced new elements into the design - two optical rods, integrated with TIR reflectors, mountable on holes on the PCB . The design further proposed for an increase in the dimensions of both the waveguide and optical rods in comparison to type-I; the waveguide core cross-section has been extended to 100×65μm, and the rod core/clad diameters are respectively 62.5/125μm VCSEL-side and 100/125μm PD-side. Although type-II architecture allows for the components to be pre-machined off-board, it did not solve the problem of very high sensitivity to misalignment errors in both architectures.
To overcome this shortcoming, a new optical connection structure for chip-to-chip interconnection applications, type-III, is proposed in this paper, which utilizes the addition of special features appended to the waveguide to significantly increase its tolerances to misalignment. For the designs of the type-III prototype, the dimensions of the 45° mirror is increased to 500×500μm while keeping the cross-section of the rectangular waveguide at 300×300μm. To achieve a smooth coupling between the end-mirrors and the waveguide, we inserted two beam ducts in-between. This allows for the cross-section to gradually decrease from the coupling-optimized 500×500μm to the mode-reduction-optimized 300×300μm, as demonstrated in Fig. 2.
As PCB mounting requires that the bottom surface of the interconnection be one planar facet, the beam duct may only have three slant surfaces with which to guide the beam. Conventional equations and calculations for minimizing refractive loss in beam ducts are mostly for symmetrical structures; with such an asymmetrical design for the beam duct, data on the optimization of its physical dimensions is not readily available.
In an effort to search for an optimized structure, the entrance window, mirror plane, beam duct, and waveguide were simulated by ray tracing with a commercial software ZEMAX®, in which the desired number of rays, emitted from randomly selected points on the window of the VCSEL, pass through free space toward the TIR reflector. The resulting output power is determined by monitoring the number of rays received in the waveguide, as well as by observing any potential rays escaping the entire interconnection structure.
The software ZEMAX® defines this type of ray tracing with non-sequential components, or NSC, which is different from non-sequential surfaces, or NSS. NSC ray tracing in ZEMAX® supports the definition and placement of multiple light sources, multiple objects, and multiple detectors in space. Automatic processing of reflection, refraction, and total internal reflection (TIR) are supported for an expansive range of 3D objects and materials, including diffractive optics. We utilized ZEMAX® to simulate the efficiency of the beam duct by varying the length of the duct section while the output light in the waveguide was monitored. Using the software, we recorded the resulting output power as a function of lengths in the range of 3.5mm to 7.5mm, but preserving the TIR reflector’s triangular prism structure at the proposed 0.5×0.5mm. With this technique, we observed that the coupling efficiency peaked at around 6mm, as shown in Fig. 3. The phenomenon was explained by Fig. 4, in which it is shown that for beam duct lengths over or under 6mm, a portion of the rays leak through the structure due to incident angles being greater than critical angles. Three typical situations with beam duct lengths of 3.5mm, 6mm, and 7.5mm are shown in Fig. 4.
3.2 Architecture comparison
A high coupling efficiency between the waveguide-VCSEL and waveguide-PD is of critical concern, for it enables the low-power operation of the VCSEL. Furthermore, when a small-aperture VCSEL is used to operate at a high bandwidth, for example a 3μm aperture for a 10 GHz operation, the coupling efficiency is of concern because of the large spatial divergence of the VCSEL’s light.
The coupling efficiency as dependent on misalignment error between the waveguide and the TIR reflector structure is again simulated by ray tracing with ZEMAX®. This simulation requires a few assumptions. First, the spatial distribution of the VCSEL is assumed to be a Gaussian profile. Though the true spatial distribution is not completely Gaussian, it is approximately Gaussian, and thus the profile is used to simplify calculations. Secondly, the software assumes that light within the acceptance angle of the waveguide are completely accepted; this however is clearly a necessary assumption for a theoretical simulation.
We utilized ZEMAX® to simulate the coupling efficiency for three structures of past and present design: type-I, type-II, and type-III from Fig. 2. The simulated misalignment sensitivities are reported in Fig. 5. The specifications of the optical structure used in the simulation is listed in Table 1. Furthermore, experimental performance measurements from the fabricated type-III prototype is shown alongside the simulated measurements in Fig. 5.
Measuring from the point of a 3dB insertion loss for all three configurations in Fig. 5, it follows that the theoretical misalignment tolerance was expanded to nearly 400μm in the proposed type-III design, more than a factor of four in comparison to the type-I design and a factor of eight compared to the type-II.
It is also notable from Fig. 5 that the type-I design performed significantly better with a notably gentler curve in comparison to the type-II design, despite the type-I having considerably smaller dimensions for both its 45° reflectors and its waveguide. It follows that reflector geometry and coupling structure, when implemented in an optimized overall design, play a decisive role in the improvement of misalignment tolerance, and that numeric cross-sectional dimensions alone is not notably correlated with coupling efficiency.
The architecture of the interconnection as described in this paper has been prototyped using soft lithography, a known efficient optoelectronic-microelectronic fabrication method that can produce satisfactorily the required features in both the horizontal and vertical directions for the proposed polymeric three-dimensional (3-D) tapered waveguide. Soft lithographic techniques adopted to fabricate the optical coupling structure are low in capital cost, and straightforward and accessible in implementation. They can circumvent the diffraction limitations of projection photolithography, provide access to quasi-three-dimensional structures and generate patterns, and also can be used with a wide variety of materials and surface chemistries, marking it as a promising solution for the creation of optoelectronic devices varying from a few micrometers to several hundreds of micrometers. We report here the fabrication of the integrated type-III interconnection.
In the fabrication of the prototypes for this design, a silicone elastomer (Sylgard® 184, Dow Corning) was applied onto a micro-engraved and surface-polished master. The elastomer was then cured at 80°C for two hours to form a minimal-shrinkage negative mold of the master. The elastomer negative mold is partially UV transparent, allowing it to be capable of pattern printing with UV polymers. In our fabrication process, a UV-curable polyurethane acrylate prepolymer (DeSolite® optical fiber secondary coating, DSM Desotech) with a refractive index of 1.50 was poured into the negative, and subsequent UV exposure cured the DeSolite®. The finished structure may then be detached from the elastomer negative mold to form the desired interconnection, a precise, smooth replica of the original micro-machined positive master. The hard-polymer final product may be subsequently cladded with a softer prepolymer, DeSolite® optical fiber primary coating with a refractive index of 1.48.
This technique can be used to print 3-D structures into any polymer-based thin film material that often cannot be fabricated by conventional lithographic approaches. The fabricated interconnection before cladding with primary coating is shown in Fig. 6, and its experimental coupling efficiency performance was shown in Fig. 5.
5. Analysis of pulse shape due to time delay
Following the above coupling efficiency analysis, we conducted a study on pulse broadening as a result of increasing the waveguide cross-section. The result shows that pulse broadening caused by the size change is negligible.
The most common form of digital transmission involves logic 1 being represented by a light pulse, and logic 0 represented by darkness. In the case of an optical interconnection, the light pulses are radiated by a light source such as a VCSEL, and subsequently enter the optical structure, within which each pulse is broken down into a set of small pulses carried by an individual mode. A fundamental problem thus with increasing the physical dimensions of the waveguide is that at the output, individual pulses may recombine, overlap, and form one deformed pulse with the rising edge from the fundamental mode and the falling edge from the critical mode, which may be difficult to resolve from the subsequent signal by the PD due to coinciding of edges. Thus, inter-modal dispersion caused by the existence of multiple modes within a large optical structure may cause limitations in allowable bitrates. We analyzed how the dispersion restricts the bandwidth of the structure.
Due to the geometry of the proposed architecture, traditional calculations used for simulating bandwidth of fibers is not applicable to this coupling structure. We therefore modeled the design utilizing a MATHCAD® script sampling discrete modes of a simulated 2D structure. Within the process, the output beam of the VCSEL is approximated by a beam of Gaussian distribution with a 1/e2 beam width of 8°–12°, and is represented digitally by fan of rays each with an incident angle of θi, forming the discretization of the input beam. The resulting pulses may hence be depicted by the following equation:
where p[t-d(θi)] is a function describing individual pulse shape; a trapezoid-shaped input pulse is utilized as shown in Fig. 7. d(θi) is a time delay function equating arrival delay according to input incidence angle θi. w(θi) serves as a Gaussian weight function of θi which modifies the pulse amplitude in accordance with the intensity distribution of each discrete beam, with α being the 1/e2 angular width, and A the normalization constant.
We were thus able to obtain the optical path of each propagation mode. The delayed light pulses from all modes are finally overlapped and recombined to the resulting pulse as shown in Fig. 7. The specifications of the optical structure used in the time delay simulation is listed in Table 1. It follows from Fig. 7 that the pulse shape is negligibly affected by modal dispersion for data rates of up to 10GHz.
Our analysis, as well as experimental results gathered from fabricated prototypes, demonstrate that this proposed integrated architecture of a polymeric optical interconnection for conventional PCB implementation is advantageous in the aspects of misalignment tolerance, ease and low cost of fabrication, as well as relative simplicity in deployment. The structure of a large-dimension entrance window and a comparatively small waveguide with a beam duct “bridge” has been found to retain both the advantages of low refractive losses and negligible time delays of pulses in high-bandwidth transmissions. The soft-lithography-based fabrication process signify a cost-effective and simple production of polymeric waveguides with necessary 3-D optical structures.
This research is supported by the China National Science Foundation under the grant No. 6047719.
References and Links
1. N. McArdle, M. Naruse, and M. Ishikawa, “Optoelectronic parallel computing using optically interconnected pipelined processing arrays,” IEEE J. Sel. Top. Quantum Electron 5, 250–260 (1999) [CrossRef]
2. Y. Ishii, S. Koike, Y. Arai, and Y. Ando, “SMT-compatible optical-I/O chip packaging for chip-level optical interconnects,” in Proceedings of IEEE Electronic Components Technology Conference, 2001, 870–875.
3. D. Krabe and W. Scheel, “Optical interconnection by hot embossing for module and PCB technology-The EOCB approach,” in Proc. IEEE Electronic Components Technology Conf, 1999, 1164–1166.
4. R. T. Chen, L. Lin, C. Choi, Y. J. Liu, B. Bihari, L Wu, S. Tang, R. Wichman, B. Picor, M. K. Hibbs-Brenner, J. Bristow, and Y. S. Liu, “Fully embedded board-level guided-wave optoelectronic interconnects,” in Proc. IEEE , 88, 780–793 (2000) [CrossRef]
5. S. Lehmacher, A. Neyer, and F. Mederer, “Polymer optical waveguides integrated in printed circuit boards,” in Proc. 27th Eur. Conf. Optical Communication (ECOC’01), 3, 2001, 302–303.
6. H. Schroder, J. Bauer, F. Ebling, and W. Scheel, “Polymer optical interconnects for PCB polymers and adhesives in microelectronics and photonics,” in Proc. 1st Int. IEEE Conf, Oct. 2001, 337–343.
7. M. Kicherer, F. Mederer, R. Jager, H. Unold, K. J. Ebeling, S. Lehmacher, A. Neyer, and E. Griese, “Data transmission at 3-Gbit/s over intraboard polymer waveguides with GaAs VCSELs,” in Proc. 26th Eur. Conf. Optical Communication (ECOC’01), 3, 2000, 289–290.
8. E. Griese, “A high-performance hybrid electrical-optical interconnection technology for high-speed electronic systems,” IEEE Trans. Adv. Package. 24, 373–383 (2001)
9. Younan Xia and George M. Whitesides, “SOFT LITHOGRAPHY,” Annu. Rev. Mater. Sci. 28, 153–184 (1998) [CrossRef]
10. Y. Huang, G. T. Paloczi, J. Scheuer, and A. Yariv, “Soft lithography replication of polymeric microring optical resonators,” Opt. Express 11, 2452–2458 (2003), http://www.opticsexpress.org/abstract. cfm?URI=OPEX-11-20-2452 [CrossRef] [PubMed]
11. B.S. Rho, M.H. Cho, H.S. Cho, S. Kang, H.-H. Park, S.-W. Ha, and B.-H. Rhee, “Low-crosstalk and high-efficiency optical interconnection using 45°-ended connection rods,” Electron. Lett. 40, 730–732 (2004) [CrossRef]