Open Access
Add to BibSonomy
Abstract
We report a fabrication process of a polymer, and mirror-based out-of-plane optical coupler. In the process, a pre-formed mirror blank made of a buffer coat material is re-exposed by a laser direct writing tool with low numerical aperture of 0.1. The fabrication process is inherently fast because of the low numerical aperture (NA) process. The surface figure of the mirror is controlled under 0.04 waves in root-mean-square (RMS) at 1.55 μm wavelength, with mirror angle of 45 ± 1 degrees. Nominal insertion loss of 8.5dB of the mirror-based coupler was confirmed with polymer waveguides fabricated simultaneously.
© 2017 Optical Society of America
1. Introduction
The scalability in data transfer rate and wiring density of metallic wire is facing physical limitations imposed by increased clock frequency [1]. Planar and integrated optical interconnects with lasers, waveguides, modulators, detectors, and dispersive devices for wavelength division multiplexing are expected to pace with the rapidly increasing demand for high speed, high density data transmission [2]. While such complete integration is the most ideal embodiment, an out-of-plane optical coupler is a viable solution for semi-integrated optical interconnects for which external light sources are employed. It provides a low cost yet moderately dense optical interconnect by leveraging flip-chip mounting technology [3,4] which enables discrete integration of light sources and detectors to a planar photonic circuit, for example, coupling of light between Vertical Cavity Surface-Emitting Lasers (VCSELs), waveguides, and photodiodes via micro lenses [5,6] and optical interfaces such as OptoBump [5].
Such out-of-plane coupling at a 90 degree angle, has been demonstrated by using diffraction gratings [6,7], evanescent couplers [8], and embedded mirrors [9–11]. Diffraction gratings are designed for specific wavelengths, thus fabrication tolerance is tight in general, though they have high coupling efficiency. For evanescent couplers, a precise control of the fabrication process is needed due to exponentially decreasing coupling efficiency as the separation from the waveguide increases [12]. In contrast, embedded mirrors have less limitation in wavelength, provided the mirror surface is precisely fabricated along with controlling mirror angle. Fabrication methods of mirror-based couplers are reported for standalone mirrors by oblique exposure [13], and embedding a coupler directly into the waveguide by mechanically cutting the edge of the waveguide [14,15], as well as wet etching processes of Si substrates [16]. Although the processes provide small coupling loss ranging from 0.35 to 4 dB, a lithographic process which has an affinity to the existing process and a yield which keeps up with established processes is highly desired for a streamlined integration of mirror fabrication into the device.
The laser direct writing, or gray scale lithography is adopted to fabricate blazed gratings and tapered waveguides for mode matching [17] and has a fast exposure rate, typically 100mm2/sec. The high exposure rate is assured, when the characteristic length scale of the structure is large enough compared to the spot size, defined by λ/NA. In contrast, for a smaller structure whose length scale is comparable to the wavelength, write resolution has to be increased by decreasing the focal volume to assure optical quality of surface figure.
Thus, laser direct writing has a flexibility in terms of writing speed, area, aspect ratio and resolution. However, all the parameters are not simultaneously satisfied by a single tool, and a major trade-off exists among writing speed, area, and resolution dominated by the NA of the tool. Although it is known that the write speed in non-linear writing in theory is independent of the focal volume [18], practical latency factors, for example, time to adjust dose by optical modulator, settle time for the stage movement, accumulates and total write time inversely scales with focal size for linear writing thus it scales as NA. For the 2D writing, NA2 dependence is expected. Apparently, a low NA and 1D writing scheme is the most promising solution which minimizes the total writing time. Therefore, for large scale production of an optical device such as an optical coupler, the low NA, 1D writing approach with a larger FOV optics is a promising solution to assure its scalability of the fabrication speed and area.
In this paper, we report a low NA and 1D laser direct writing for a 1 to 1 aspect ratio structure: a 45 degree mirror-based optical coupler, while keeping a high write rate of 2600 mm2/min. We demonstrate that simultaneous control of the mirror angle and surface figure is achievable even by using the low NA method.
In section 2, we describe how a low NA gray scale and laser direct writing process enables precise fabrication of a 45 degree optical surface. The integration of the mirror with polymer waveguides is also discussed. In section 3, experimental evaluation of coupling loss is presented, followed by a discussion on particular advantages of the proposed approach, as well as limitations, in section 4.
2. Low NA laser direct writing process for 45 degree mirror
2.1 Device structure
The device of our interest is schematically depicted in Fig. 1. An external light source(s), e.g., a VCSEL, and potentially with an integrated micro lens array [5–7], is discretely integrated into a planar waveguide having dimensions of (w) x (h), for example by flip chip bonding [3]. The mirror-based coupler (height: T) directs the light from the source to the waveguides which are a distance d apart. A secondary coupler (not shown) re-directs the light to a photo detector (not shown).
The dimensions T, d, h, and w are determined as follows. Typical index contrast [19,20] of polymer waveguides is on the order of 0.01, which gives rise to the NA of the waveguide, [21]. The height of the mirror T and spacing d between the edge of the waveguide and the surface of the mirror are designed to avoid beam clip-off at the mirror surface. The condition is given by.
where is the Rayleigh range, and w0 is the diameter of the beam waist. The 2nd term of Eq. (1) is further simplified as , since is feasible in our process, and because Rayleigh range for an beam is about 25 µm, divergence of the beam at the surface of the mirror is negligible. The cross sectional dimensions of the waveguide, 9 (w) x 7 (h) µm are determined by a mode solver package so that the waveguide supports a single mode.Figure 2 shows an overall process flow. An 11 µm layer of cladding material, Epoclad (Microresist Technology, Berlin, Germany), is spin coated onto a 500 µm Silicon substrate preconditioned by an adhesion promoter HDMS. The cladding layer undergoes a flood exposure (350 mJ/cm2 at 365 nm) using a model MA6 mask aligner (Suss Microtec, Germany), development (90 sec bath, SU-8 developer) and hard bake (90°C for 5 min on a hot plate). Following the hard bake, an 8 µm layer of buffer coat material, WPR-5100 (JSR, Sunnyvale, CA), is spin coated (3000rpm for 55sec) onto the cladding layer. Although, the final coupler height was 5.5μm, an 8μm layer is used during processing to accommodate a loss of 2-3 µm of mirror height during the development process explained in the later discussion section. The buffer coat material undergoes a flood exposure (250 mJ/cm2 at 365 nm) to pattern a latent image of a mirror blank with width 13 µm which is locally exposed by a focused laser [22,23] to form a 45 degree surface, which is explained in detail in next section.
Figure 2 Fabrication process of a 45 degree optical coupler with an under cladding layer on a Si substrate. A binary mask patterns a latent image of a mirror blank inside a buffer coat material. Local exposure at a distance L from the left edge of the mirror blank by direct laser writing form a mirror surface after development. Waveguide core is coated and patterned followed by coating of an upper cladding layer.
2.3 Low NA and selective exposure of mirror blank
The surface angle and figure of the mirror are controlled by the location and dose at the selective exposure step by direct laser writing by Maskless Lithography Tool (MLT) [22,24,25]. Figure 3 depicts schematics of the MLT. The MLT uses a flying spot scanning laser system to apply exposure patterns using an 8-bit (0-255) grayscale. The grayscale level in the bitmap file corresponds to the modulation of laser power controlled by an Acoustic Optical Modulator (AOM). The laser is expanded (not shown), and relayed to AOM by relay optics. The scan optics employs one-dimensional scan along the x-direction while the sample stage synchronously moves along y-direction. All the components are controlled by personal computer (PC). The MLT operates at a wavelength of 365nm with a typical continuous laser power of 450mW.
Fig. 3 Block diagram of the MLT. A 365 nm laser enters an optical relay system which directs the beam to an AOM. The beam is modulated as referenced by the input bitmap pattern to produce a variable exposure as the sample is translated.
Figure 4 shows the shape of the mirror blank after development as a function of the beam location L, as measured from the left edge of a 13um wide mirror blank. As the exposure location (dose of 1140 mJ/cm2) is shifted from the center of the blank towards the right edge, the shape of the mirror surface continuously changes as depicted in Figs. 4(a), (b), (c), and (d). Surface topology of the mirror surface (right slope of the structure depicted in Figs. 4(a), (b), (c), and (d)) is plotted in Fig. 4(e) with a 45 degree reference line (black solid line). At a distance L = 10 µm a flat surface figure with a root-mean-square (RMS) figure of 0.04 waves for λ = 1.55 µm with 54 degrees slope angle is formed.
Fig. 4 Left: (a-d) Cross sectional SEM showing the effect of shifting exposure of laser direct writing L: location of the focused UV spot measured from the left edge of the mirror blank on surface topology of mirror surface. Also indicated is RMS error in waves (λ = 1.55 µm): (a) L = 8µm, (b) 9µm, (c) 10µm, and (d) 11µm, all with dose = 1140 mJ/cm2. Right: (e) Surface topology of the mirrors depicted in (a-d) is plotted.
The angle of 54 degrees is further reduced while keeping the surface figure by controlling dose D given by where is beam speed, is beam spot size, and is power density of the beam, and N is a number of repeated exposures. In our experiment, , , and were used [26]. Figure 5 shows mirror angle as a function of dose controlled by number of passes. While dose is less than or greater than 646 mJ/cm2, the angle is larger than 45 degrees, Figs. 5(a), (c), and (d). As the dose is increased beyond 1100 mJ/cm2 the surface takes on a concave shape, Fig. 5(d). We found that a slope of 45 degrees is achieved by D ~646 mJ/cm2 with N = 17. For this optimum dose the surface figure is 0.06 µm P-V (0.04 waves RMS for λ = 1.55 µm) as summarized in Fig. 5(e) with a 45 degree reference line (black solid line).
Fig. 5 Left: (a-d) Cross sectional SEM of mirror, (a) 55 degrees for 266 mJ/cm2, 45 degrees for (b) 646 mJ/cm2, (c) 48 degrees for 798 mJ/cm2, and (d) 52 degrees for 1140 mJ/cm2. Right: (e) Surface topology after development for dosages of (a)-(d).
The underlying mechanism of formation of a straight slope with a 45 degree angle can be understood as follows. First of all, the shape of the mirror is formed by single exposure of a focused UV laser having a Gaussian intensity profile. Specifically, we utilized the side slope of the Gaussian dose profile. Since the side slope of the Gaussian is not a linear one, we take advantage of the nonlinearity of the material in removal of the volume of the film as a dose. In our dose level, the nonlinearity of the material is such that the removal rate decreases as dose increases. Figure 6 shows topology of the developed WPR-5100 photo resist after exposing it with a linear gradient exposure in space. Within the region of the linear gradient exposure that was applied (around 300 μm in x), a part of the region (starting around 150 μm to 250 μm, in x), we see aforementioned nonlinear response. This tendency can be seen in Figs. 4(b) and 4(c) that residual photo resist was observed around the location where the center of the writing beam was located. At the same time one may notice that over the range of L, we always observe that at least a part of the developed structure has straight region. The parameter L is optimized such that the straight region is formed at the edge of the mirror blank while not exposing the mirror blank with the higher intensity region of the focused spot, Fig. 4(e). The nonlinearity is again used to control the angle of the mirror. While repeating the writing by passing beam multiple times over the mirror blank, more material removed at low dose region (close to the top of the mirror), as compared to the high intensity region (bottom edge of the mirror), thus the angle of mirror decreases as number of passes increase. The tendency continues up to the dose 646 mJ/cm2 (Fig. 4(b)). Beyond this dose level, one may notice that the angle of slope increases again while the height of the mirror decreases and the boundary of the mirror starts shifting towards the left. As Fig. 5 shows, the appropriate amount of shift of the writing beam is identified as L = 10 µm. The L = 10µm condition is needed to match the nonlinearity of the material (depicted in Fig. 6) to the dose profile (side profile of Gaussian) which leads to the formation of a straight slope. As dose increases, the optimum condition for L is no longer satisfied since more material is removed from the right edge of blank as we see decrease in height and shift of slope towards the left. Thus, the effective value for L increases to >10 µm for which the slope angle increases as observed in Fig. 4(d) for L = 11um. In Fig. 5(d), the topology of mirror surface looks more concave because of the residual material observed at right edge. The existence of the residual material is due to mismatch of the nonlinearity of material to side profile of the Gaussian dose. For this case, less material around the right edge of the mirror was removed since a stronger part of the writing beam is no longer matched to the right edge of the blank. In contrast, the left part of the mirror is mostly intact, but has a steeper angle because this portion is exposed by the tail of the Gaussian intensity where there is less variation of intensity profile as compared to the center portion of Gaussian intensity profile.
Fig. 6 Topology of WPR5100 after applying linear exposure dose. Nonlinear response to the dose can be seen around 150-250 μm.
2.4 Au coating of mirror surface
Next, the sample is developed using NMD-3 2.38% (TOK, Milpitas, CA) for 5 min. Finally, the sample is hard baked using a hot plate for 60 min at 150°C. The mirror surface is selectively coated with a 100 nm layer of Au by using a lithographic wet etching technique [26,27].
Figure 7(a) shows the final sample with Au coated mirrors and polymer waveguide array. The mirror-based optical coupler and the region above is coated by gold whereas the region below the optical coupler remains uncoated. The reader may notice a series of 6 mirror stripes above the optical coupler. These stripes are not used for the purpose of coupling but for alignment purposes only.
Fig. 7 (a) Optical micrograph showing top down view of optical waveguides and coupler before applying the upper cladding layer. The gold coated region covers the optical coupler only but not the waveguides. (b) Cross sectional SEM of optical coupler, (c) Cross sectional SEM of waveguide. Note: (b) and (c) are cleaved from opposite side of sample used in (a). The output mode is fit to a Gaussian curve with 1/e2 diameter = 13.1 µm (d).
2.5 Integration of waveguides
The selectively Au coated mirror is further processed by spin coating an 11 µm thick layer of Epocore (Microresist Technology, Berlin, Germany). The Epocore layer is patterned into waveguides (exposure dose = 350 mJ/cm2). The sample is developed and hard baked (90 sec bath, SU-8 developer with IPA rinse, 90°C for 5 min on a hot plate) followed by top coating with an upper cladding layer of Epoclad. In Fig. 7, a cross section of the rectangular waveguide, 7.78 µm x 9.92 µm, is shown along with the top micrograph of the device. The output mode from the cleaved edge of the waveguide is measured by illuminating the other side of the waveguide via the mirror and is fit to a Gaussian curve with 1/e2 diameter = 13.1 µm.
3. Optical characterization of mirror insertion loss
The insertion loss of the mirror-based optical coupler includes beam clipping, decrease in coupling efficiency due to imperfection of surface figure, and reflectivity of gold coating, and is given by,
where α is total device insertion loss, β is waveguide loss, γ is loss due to alignment, δ is Fresnel loss, e is an intrinsic mode mismatch loss between input beam and waveguide modes.Table 1 summarizes loss measurement results along with loss terms simulated for alignment error, Fresnel loss and mode mismatch.
Table 1. Loss measurement results
The total device insertion loss is measured by using a free-space coupling method in the configuration depicted in Fig. 1. A 1550nm laser (8155A, HP) is collimated by an f = 50 mm achromatic doublet and focused onto the mirror surface by an NA 0.1 microscope objective (80.7325, Rolyn Optics). Light reflected by the mirror is coupled into the front facet of a 1.25 cm waveguide. The output facet of the waveguide is imaged by an NA = 0.25 microscope objective (80.7325, Rolyn Optics) and f = 100 mm tube lens onto an IR detector (S132C, Thorlabs). Total device loss is measured as a ratio of the output power to the input power (typically, [W] and [W] respectively), and compensating it by loss of detection optics of 2 dB. The waveguide loss, 3 dB/cm, is separately measured by using a cutback method [28], and a total loss of 3.75 dB for the 1.25 cm waveguide is measured.
Loss due to misalignment, mode mismatch, and mirror clipping for T = 5 μm, is numerically obtained by mode solver package and FDTD simulations [21]. The loss due to alignment errors of the input spot to the waveguide in the range of 1 to 2μm can range from 0.25 to 1dB. We estimate that typical alignment error is about ± 1 μm, as a result of the 1μm sensitivity inherent in the Vernier micrometers used for translating the sample [29], which corresponds to 0.25 dB of loss. Fresnel loss is calculated and summed over losses 0.212 dB at air to dielectric (n_cladding = 1.5631), dB for dielectric (n_cladding = 1.5631) to dielectric (n_core = 1.5752) at the input edge of the waveguide, and 0.222 dB for dielectric (n_core = 1.5752) to air at the exit side of the waveguide.
The output mode is fit to a Gaussian curve with 1/e2 diameter = 13.1 µm. With an input NA 0.1 and λ = 1.55 um, the 1/e2 mode size at the input side of the waveguide is 12.7 µm. Due to the mode mismatch we estimate that 0.96 dB of coupling loss exists for the rectangular waveguide we fabricated.
4. Discussion
We estimated 8.51 dB of mirror insertion loss exists for the sample we fabricated. Ideally, the extent of the mirror is about the size of 1/e2 diameter of the input beam or T = 13.1 μm. Our sample has T = 5.5 um, thus a part of the beam is clipped by mirror itself, and loss due to the clipping is expected. Simulations by mode solver package and FDTD analysis for the structure depicted in Fig. 1 shows that for T = 5.5 and 13.1 μm, clipping loss is 2.5 dB and 0.4 dB, respectively.
To scale up the proposed method to fabricate a mirror having over 10 μm of height while keeping the 45 degrees flat surface, we envision that the key findings discussed before can be still applicable. Specifically, we can separate out two figures of merit: flatness and angle by initial intensity of the dose profile (a side of Gaussian intensity profile of a focused laser) and accumulated dose (number of passes). Technically, a thicker initial film height is needed. As a matter of fact, the WPR-5200 film [30] (available from the same manufacturer) can be deposited as thick as 20 μm Second, to apply the dose condition for the initial film thickness of 7 μm for even thicker film (for example 14 μm), our model for formation of the mirror indicates that the Gaussian exposure profile used for a 7 μm height needs to be scaled by factor of 2 ( = 14 μm/7 μm) assuming a weak absorption of the exposing light over the film thickness. Currently in our laser writing tool, the 1/e2 beam width is set to be 2.1 μm [22]. For a thicker film the beam width needs to be doubled to take advantage of the same nonlinearity of the material over a doubled mirror region. To do so, we have at least two options: using lower NA 0.05 focusing optics, or reduce beam size by factor of 2 by using an inverse telescope type optics while still using NA 0.1 focusing lens. Finally, the angle of the mirror is controlled by number of passes, which in first order is doubled since the volume of photo resist to be removed is doubled too. The mirror height problem can also be solved by adopting a cladding-core-air type waveguide. The device at this point is designed in cladding-core-cladding. Alternatively, cladding-core-air type can be used, and is beneficial in terms of reducing required mirror height by pushing the mode closer to the substrate and number of required processes. The effect of adopting a cladding-core-air design was simulated using a mode solver as illustrated in Fig. 8. Figure 8(a) shows that when Epoclad is used for under cladding and upper cladding, the location of the supported mode is centered with respect to the waveguide dimensions. When air is used for the upper cladding layer, the index contrast between the core and the upper cladding increases resulting in a lower mode position as seen in Fig. 8(b). In Fig. 8(c) we can observe an expected shift of 0.8µm. By pushing the mode toward the substrate, coupling efficiency of the mirror can be improved. FDTD simulations indicated a reduction in mode mismatch loss of 0.8dB.
Fig. 8 Modal analysis of waveguide (outlined in black). (a) Supported waveguide mode when Epoclad is used for upper and lower cladding using Epocore for the core material. (b) Supported waveguide mode when no upper cladding layer is used. (c) Cross section along the z axis of the supported waveguide modes with and without upper cladding layer.
We observed that the initial film thickness is reduced by 2-3 um. The reason is that, along with the removal of solvent of the photoresist, a background exposure present during laser direct writing due to an imperfect extinction ratio by the acousto-optic modulator (AOM) causes a small unintentional global exposure. During development, this small exposure is evident in the removal of areas which were not exposed directly. The loss of height is adjusted again by adding it to initial film thickness.
Theoretical reflectivity of Au at 1550 nm wavelength is 99% while we measured 95% with our device which shows 0.17 dB of loss exists. With the ideal mirror height, and mode matched input along with ideal alignment condition and reflectivity of mirror, the mirror insertion loss can be improved to 5.03 dB.
The material we used for waveguide Epocore/Epoclad is an SU-8 derivative and is a very well-known polymer used for waveguides at 1550 nm [31–33]. In addition, loss in the range of 2-3 dB/cm is typical for polymer waveguides operating at 1550 nm. Despite this high loss, the material is well understood and documented, therefore it is easy to process and replicate. The material is also extremely robust. For the purpose of demonstrating fabrication of a mirror coupler the loss of the waveguide is not essential, however, smaller loss alternatives, i.e., inorganic materials are also compatible with our process.
Once the waveguide and coupler dimensions are optimized, the estimated coupling loss of the mirror can be reduced to comparable numbers to other vertical couplers such as grating couplers (1.5 dB [6] to 4.5 dB [7]), 45 degree mirrors by oblique exposure (2 dB to 4 dB [9,34–36]) and by laser ablation or dicing (0.35 dB [11] and up to 2 dB [37]).
From the viewpoint of fabrication, our approach has a uniqueness which frees limitations of other fabrication methods. For example, direct implementation of a 45 degree mirror at the edge of waveguide is only effective for optical architecture placing a light source at the opposite side of the transparent substrate [22, 23]. Wet chemical etching of substrate with KOH or mechanical dicing of the edge of the waveguide have been demonstrated as well, though eliminating wet etching, and eliminating mechanical process, is essential towards fully CMOS compatible process. One may notice that we still relied on a wet process, for example, lift off of the gold mirror. The process can be replaced with nanostencil lithography evaporation techniques [38].
Integration of light sources, VCSEL and/or fiber need coupling optics due to the divergence of light [5–7]. With an aid of emerging wafer-level and stacked planer optics [39,40], coupling optics can be integrated with active or passive self-alignment techniques including flip chip (controlled collapse chip connection) bonding [3,4]. For example, an external substrate with a pre-fabricated light source and micro optics can be brought in contact with the solder bumps which are then reflowed in order to make the bond. In our design, a flat region just outside of the mirror coupler can be used as a spot for these solder bumps. Alternatively, it is possible to utilize the buffer coat material used in coupler fabrication to create an additional structure such as a trench for alignment. This could provide a key structure to align planer optics stacked on top of the mirror coupler. Simultaneous fabrication of both the trench and the optical coupler is possible therefore no added fabrication time is necessary.
Also the proposed method has an intrinsic advantage over the high NA approach in fabrication speed. Two photon laser direct writing can be used for fabrication of a 45 degree coupler, however, the write rate by using high NA optics, thus small focal volume, is substantially slower (10 mm/sec) [41]. In contrast, our fabrication process utilizes a low NA maskless lithography approach which enables 3 orders of magnitude faster exposure rates (~20,000mm/s).
5. Conclusions
We have demonstrated a complete process for a mirror-based out-of-plane optical coupler. By a novel usage of a buffer coat material as a pre-formed mirror blank, and re-patterning it by a low NA and 1D laser direct writing lithography tool, fabrication of a one-to-one aspect ratio structure with a fast write rate is feasible. Integration to a polymer waveguide shows that mirror insertion loss is measured as 8.51 dB when coupling with a 1550 nm laser to the device. Detailed loss analysis indicates further reduction of loss down to 5 dB is possible by adjustments to the fabrication process parameters resulting in a coupler height matched to the mode size. The mirror coupler has an insertion loss comparable to other methods while providing additional benefits in fabrication speed and affinity to the existing processes.
Acknowledgments
We would like to thank JSR Micro, Inc. for providing WPR-5100 buffer coat material and Intel Corp. for their support in the early stages of this research.
References and links
1. J. W. Goodman, F. J. Leonberger, S.-Y. Kung, and R. A. Athale, “Optical interconnections for VLSI systems,” Proc. IEEE 72(7), 850–866 (1984). [CrossRef]
2. E. Timurdogan, Z. Su, K. Settaluri, S. Lin, S. Moazeni, C. Sun, G. Leake, D. D. Coolbaugh, B. R. Moss, M. Moresco, V. Stojanovic, and M. R. Watts, “An ultra low power 3D integrated intra-chip silicon electronic-photonic link,” 2015 Opt. Fiber Commun. Conf. Exhib. OFC Proc. 3 pp. (2015). [CrossRef]
3. K. S. Kaur, A. Z. Subramanian, P. Cardile, R. Verplancke, J. Van Kerrebrouck, S. Spiga, R. Meyer, J. Bauwelinck, R. Baets, and G. Van Steenberge, “Flip-chip assembly of VCSELs to silicon grating couplers via laser fabricated SU8 prisms,” Opt. Express 23(22), 28264–28270 (2015). [CrossRef] [PubMed]
4. H. Lu, J. S. Lee, Y. Zhao, C. Scarcella, P. Cardile, A. Daly, M. Ortsiefer, L. Carroll, and P. O’Brien, “Flip-chip integration of tilted VCSELs onto a silicon photonic integrated circuit,” Opt. Express 24(15), 16258–16266 (2016). [CrossRef] [PubMed]
5. Y. Ishii, S. Koike, Y. Arai, and Y. Ando, “SMT-compatible large-tolerance “OptoBump” interface for interchip optical interconnections,” IEEE Trans. Adv. Packag. 26(2), 122–127 (2003). [CrossRef]
6. M. Dai, L. Ma, Y. Xu, M. Lu, X. Liu, and Y. Chen, “Highly efficient and perfectly vertical chip-to-fiber dual-layer grating coupler,” Opt. Express 23(2), 1691–1698 (2015). [CrossRef] [PubMed]
7. H. Yamada, M. Nozawa, M. Kinoshita, and K. Ohashi, “Vertical-coupling optical interface for on-chip optical interconnection,” Opt. Express 19(2), 698–703 (2011). [CrossRef] [PubMed]
8. A. Flores, S. Song, J. J. Yang, Z. Liu, and M. R. Wang, “High-Speed Optical Interconnect Coupler Based on Soft Lithography Ribbons,” J. Lightwave Technol. 26(13), 1956–1963 (2008). [CrossRef]
9. F. Wang, F. Liu, and A. Adibi, “45 Degree Polymer Micromirror Integration for Board-Level Three-Dimensional Optical Interconnects,” Opt. Express 17(13), 10514–10521 (2009). [CrossRef] [PubMed]
10. M. Immonen, M. Karppinen, and J. K. Kivilahti, “Fabrication and characterization of polymer optical waveguides with integrated micromirrors for three-dimensional board-level optical interconnects,” IEEE Trans. Electron. Packag. Manuf. 28(4), 304–311 (2005). [CrossRef]
11. N. Hendrickx, J. Van Erps, E. Bosman, C. Debaes, H. Thienpont, and P. Van Daele, “Embedded Micromirror Inserts for Optical Printed Circuit Boards,” IEEE Photonics Technol. Lett. 20(20), 1727–1729 (2008). [CrossRef]
12. N. M. J. E. N. Glytsis, N. M. Jokerst, R. A. Villalaz, Sang-Yeon Cho, Shun-Der Wu, M. A. Zhaoran Huang, Brooke, and T. K. Gaylord, “Substrate-Embedded and Flip-Chip-Bonded Photodetector Polymer-Based Optical Interconnects: Analysis, Design, and Performance,” J. Lightwave Technol. 21(10), 2382–2394 (2003). [CrossRef]
13. J. Schrauwen, S. Scheerlinck, D. Van Thourhout, and R. Baets, “Polymer wedge for perfectly vertical light coupling to silicon,” in (2009), Vol. 7218, p. 72180B–72180B–8.
14. G. Jiang, S. Baig, and M. R. Wang, “Flexible polymer waveguides with integrated mirrors fabricated by soft lithography for optical interconnection,” J. Lightwave Technol. 31, 1835–1841 (2013). [CrossRef]
15. X. Lin, X. Dou, A. X. Wang, and R. T. Chen, “Polymeric Waveguide Array with 45 Degree Slopes Fabricated by Bottom Side Tilted Exposure,” Proc. SPIE 7944, 794411 (2011). [CrossRef]
16. C. Strandman, L. Rosengren, H. G. A. Elderstig, and Y. Backlund, “Fabrication of 45 deg; mirrors together with well-defined v-grooves using wet anisotropic etching of silicon,” J. Microelectromech. Syst. 4(4), 213–219 (1995). [CrossRef]
17. A. Sure, T. Dillon, J. Murakowski, C. Lin, D. Pustai, and D. Prather, “Fabrication and characterization of three-dimensional silicon tapers,” Opt. Express 11(26), 3555–3561 (2003). [CrossRef] [PubMed]
18. M. Häfner, C. Pruss, and W. Osten, “Laser direct writing,” Opt. Photonik 6(4), 40–43 (2011). [CrossRef]
19. “EpoCore & EpoClad Serien | micro resist technology GmbH,” http://www.microresist.de/en/products/negative-photoresists/uv-lithography-broadband-and-i-line-exposure/epocore-epoclad-serien.
20. B. E. A. Saleh and M. C. Teich, Fundamentals of Photonics, 2nd ed. (Wiley-Interscience, 2007).
21. “Lumerical Solutions, Inc. | Innovative Photonic Design Tools,” https://www.lumerical.com/.
22. J. M. Tamkin, B. Bagwell, B. T. Kimbrough, G. Jabbour, and M. R. Descour, “High-speed gray-scale laser direct-write technology for micro-optic fabrication,” in Micromachining Technology for Micro-Optics and Nano-Optics (2003), Vol. 4984, pp. 210–218.
23. M. A. Zaverton, J. Sierchio, Y.-S. Kim, D. Hansen, W. Bletcher, J. Tamkin, and T. Milster, “Maskless Lithography Tool for Fabrication and Inspection of Diffractive Optical Elements and Computer Generated Holograms,” in International Optical Design Conference and Optical Fabrication and Testing(2010), Paper OTuC5 (Optical Society of America, 2010), p. OTuC5. [CrossRef]
24. C. Summitt, S. Wang, L. Johnson, M. Zaverton, T. Milster, and Y. Takashima, “Flexible micro-optics fabrication by direct laser writing toward CMOS compatible 3D optical circuit,” in G. von Freymann, W. V. Schoenfeld, and R. C. Rumpf, eds. (2013), p. 86130K.
25. C. Summitt, S. Wang, L. Johnson, M. Zaverton, T. Ge, T. Milster, and Y. Takashima, “Micro-optics fabrication by mask-based and mask-less mixed lithography process towards 3D optical circuits,” in G. von Freymann, W. V. Schoenfeld, and R. C. Rumpf, eds. (2014), p. 89740C.
26. W. Eidelloth and R. L. Sandstrom, “Wet etching of gold films compatible with high Tc superconducting thin films,” Appl. Phys. Lett. 59(13), 1632–1634 (1991). [CrossRef]
27. R. J. Jackman, J. L. Wilbur, and G. M. Whitesides, “Fabrication of submicrometer features on curved substrates by microcontact printing,” Science 269(5224), 664–666 (1995). [CrossRef] [PubMed]
28. D. Derickson, ed., Fiber Optic Test and Measurement, 1st ed. (Prentice Hall, 1997).
29. “SM-13,” https://www.newport.com/p/SM-13.
30. “JSR Micro / Materials Innovation,” http://jsr-nv.savvybyte.com/index.php/ElectronicMaterials/PackagingMaterials/WPR/.
31. V. Prajzler, P. Nekvindova, P. Hybs, and V. Jerabek, “Properties of the Optical Planar Polymer Waveguides Deposited on Printed Circuit Boards,” Radioengineering 24(2), 442–448 (2015). [CrossRef]
32. Y. Yang, K. Chen, W. Jin, and K. S. Chiang, “Mode Converter with Polymer Long-Period Waveguide Grating,” in Asia Communications and Photonics Conference 2014 (2014), Paper AW4C.4 (Optical Society of America, 2014), p. AW4C.4. [CrossRef]
33. S. K. Mishra, B. Zou, and K. S. Chiang, “Surface Plasmon Resonance Sensor Based on a Polymer Waveguide with a Copper Thin-Film Overlay,” in Asia Communications and Photonics Conference 2015 (2015), Paper AM3H.6 (Optical Society of America, 2015), p. AM3H.6. [CrossRef]
34. G. Jiang, S. Baig, and M. R. Wang, “Flexible Polymer Waveguides With Integrated Mirrors Fabricated by Soft Lithography for Optical Interconnection,” J. Lightwave Technol. 31(11), 1835–1841 (2013). [CrossRef]
35. T. Yoshimura, M. Miyazaki, Y. Miyamoto, N. Shimoda, A. Hori, and K. Asama, “Three-dimensional optical circuits consisting of waveguide films and optical Z-connections,” J. Lightwave Technol. 24(11), 4345–4352 (2006). [CrossRef]
36. J. Inoue, T. Ogura, K. Kintaka, K. Nishio, Y. Awatsuji, and S. Ura, “Fabrication of Embedded 45-Degree Micromirror Using Liquid-Immersion Exposure for Single-Mode Optical Waveguides,” J. Lightwave Technol. 30(11), 1563–1568 (2012). [CrossRef]
37. Z. Zhang, M. Kleinert, A. Maese-Novo, G. Irmscher, E. Schwartz, C. Zawadzki, and N. Keil, “Multicore Polymer Waveguides and Multistep 45deg Mirrors for 3D Photonic Integration,” IEEE Photonics Technol. Lett. 26(19), 1986–1989 (2014). [CrossRef]
38. S. Aksu, M. Huang, A. Artar, A. A. Yanik, S. Selvarasah, M. R. Dokmeci, and H. Altug, “Flexible Plasmonics on Unconventional and Nonplanar Substrates,” Adv. Mater. 23(38), 4422–4430 (2011). [CrossRef] [PubMed]
39. Y. Aoki, J. R. Mizuno, Y. Shimada, and K. Iga, “Parallel and bi-directional optical interconnect module using vertical cavity surface emitting lasers (VCSELs) and 3-D micro optical bench (MOB),” in 1999 Digest of the LEOS Summer Topical Meetings: Nanostructures and Quantum Dots/WDM Components/VCSELs and Microcavaties/RF Photonics for CATV and HFC Systems (Cat. No.99TH8455) (1999), pp. III9–II10.
40. K. Iga, M. Oikawa, S. Misawa, J. Banno, and Y. Kokubun, “Stacked planar optics: an application of the planar microlens,” Appl. Opt. 21(19), 3456–3460 (1982). [CrossRef] [PubMed]
41. “3D printing on the micrometer scale - Nanoscribe GmbH,” https://www.nanoscribe.de/files/4514/8179/1302/DataSheet_PPGT_V04_2016_web.pdf.
