Optical ridge type waveguides based on UV-curable polymer were fabricated by imprinting method. Positive tone resist patterned on a silicon wafer was used as a mould. The characterization of waveguides was carried out by coupling TE-polarized light from a tapered fiber into a waveguide with 30 mm length and mapping the intensity distribution with another tapered fiber at the output facet of a waveguide. Proper single- or multimode operation was observed depending on the waveguide width being either 2 µm or 6 µm. Experimental observations on the mode profiles were also supported by the simulation results. Average power transmissions of 32% at 1530 nm wavelength and 45% at 1310 nm wavelength were characterized. The results suggest that the simple mould fabrication process might be a useful technique for device prototyping and that the performance of replicated waveguides can meet the requirements for certain applications.
© 2009 OSA
Imprinting or nanoimprint lithography (NIL) has been considered as an emerging technology for patterning structures in micro- and nanoscale [1–4]. Important advantage of this technique includes a potential for high-throughput with low equipment costs. The operation principle of imprinting is fairly simple. Imprinting is essentially embossing method, where the pattern is replicated from a mould to another surface. The pattern transfer can be based on the surface deformation of thermoplastic media or alternatively UV-curing of polymer. In the first method, patternable material is heated above the glass transition temperature, when it accommodates to the profile of the mould. In the latter choice, UV-curable polymer in liquid phase is coated on a substrate and UV-light is exposed during the stamping step. This results in the formation of the replica from a mould. Restriction of UV-method is the requirement for the UV-transparency of the mould or the substrate. Nanoimprint technology cannot only create resist patterns for the subsequent etching step, but can also produce structures in polymer based functional devices.
Besides academic and industrial microelectronics community, imprinting has gained also notable attention in other research fields including photonics. Various photonics devices have been demonstrated, where imprinting technique has been utilized during fabrication. Passive devices include, for example, diffractive  and refractive [5,6] micro-lenses, planar waveguides [7–9] and gratings . Imprint technique has been also applied successfully to enhance the performance of active photonics devices, such as, distributed-feedback (DFB) single-frequency diode laser  and GaN-based light-emitting diode with photonic crystal structure for improved light extraction efficiency . Interestingly, the usage of imprinting in optical applications is not limited in the device fabrication, but this replication method has been applied to process alignment structures on micro-opto-electro-mechanical (MOEMS) packages .
The mould in nanoimprint fabrication possesses principally the same role as the photomask in lithography setting high quality requirements for the mould preparation. In imprinting, patternable polymer replicates very accurately the mould features down to nano-scale. This is one key advantage of imprinting technique, since with a high quality mould it is possible to fabricate almost ideal patterns. On the other hand, as a consequence of accurate replication, also non-idealities and defects in the mould are transferred into replicas. Generally in optical components, surface roughness is with high importance. This concerns especially planar waveguides where propagation loss due to scattering is extremely sensitive on the roughness. Moulds for the imprint fabrication of optical waveguides have been manufactured, for example, by utilizing diamond tooling  or reactive ion etching (RIE) . Heat treatment method can be further applied to improve the device operation by reducing the surface roughness of replicated waveguides .
The motivation of our work was to investigate the usage of patterned photoresist as a mould for UV-imprinting of polymer waveguides. The mould fabrication consists only in patterning of the resist layer. The subsequent etching of substrate is not required and one potential source for the increased surface roughness can be avoided. The mould preparation is also rapid providing thereby means for prototyping waveguide devices with short turn-around time. This article is divided in the following manner. After describing the waveguide manufacturing process by imprinting, results on the operation of waveguides are presented in order to evaluate their performance end to estimate their usability in applications. First, measured mode profiles of imprinted waveguides are shown and compared with simulation results. Second, transmittance of single-mode waveguides is characterized and the experimental results are compared with theoretically obtainable transmission values. Finally, the results are summarized with concluding remarks.
2. Fabrication of waveguides
The process flow used in this work to fabricate ridge waveguides is illustrated in Fig. 1 . The process can be divided into two parts: mould fabrication and actual waveguide replication. Standard positive resist patterning procedure was applied in the mould fabrication . Photoresist was first spin coated on a silicon wafer. This was followed by lithographic patterning of the resist layer, when UV-light was exposed through a shadow mask. Finally, the exposed areas were removed with a developer. This structure was later used as an imprinting stamp in replication step. The pattern height can be controlled by choosing a resist with an optimal viscosity and fine tuned by adjusting the spin coating speed. Cross section image from a patterned resist is shown in Fig. 2(a) . The mould surface is smooth illustrating the benefit of the used method.
Replication process started by spin coating of commercial Ormocer  hybrid polymer on BK7 glass substrate. Resin was diluted in a thinner (Ormothin) in order to obtain a desired film thickness for the ridge waveguide purposes. Coating step was followed by the prebake at 85 °C for 2 minutes to evaporate the solvents. Patterned mould was then pressed against the glass wafer with soft Ormocer layer on it. During the pressing step UV-light was exposed through the glass substrate resulting in curing of liquid polymer layer. Replicated waveguide wafer was detached and, finally, waveguide facets were cleaved. A cross section image from a replicated ridge waveguide on a slab with a thickness of 0.8 µm is shown in Fig. 2(b).
The used fabrication process resulted in a waveguide structure, where the glass substrate acted as an under cladding layer. Dimensions of the ridge were determined by the mould structure. Surface roughness of the mould and replicated waveguides were analyzed by atomic-force-microcopy (AFM, Veeco Dimension 3100). The surface analysis of the mould covered the determination of the RMS roughness value on the positive tone layer, and also, on the bottom of the groove with a width of 6 µm, which is essentially the surface of the silicon wafer. Correspondingly, RMS surface roughness of the replicated waveguides was measured on the top of the waveguide ridge with a width of 6 µm and on the slab area. The scan area in each analysis was 4 µm × 4 µm and the measurements were carried out five times at separate parts of the samples to improve the reliability of the analysis. In the case of mould, RMS roughness values of 0.17 ± 0.03 nm and 0.48 ± 0.02 nm were determined from the bottom of the groove and from the surface of the resist layer, respectively. The surface roughness was slightly increased in the Ormocer replica, where a surface roughness value of 0.8 ± 0.3 nm was observed at the top of the ridge. A similar surface roughness with a RMS value of 0.8 ± 0.2 nm was further determined from the slab region. These results indicate that the used replication method used in this work can be utilized to fabricate device structures with a smooth surface. Refractive indices at 1530 nm wavelength used later in Testing and analysis section for Ormocer and glass substrate acting as an under cladding layer were 1.519 and 1.456, respectively. Corresponding indices at 1310 nm wavelength were 1.521 for Ormocer and 1.459 for the glass. The refractive indices were obtained by using Metricon prism coupling equipment, when a separate blank glass wafer and non-patterned Ormocer film were used in measurements.
3. Testing and analysis
Fabricated waveguides were analyzed by a measurement setup illustrated in Fig. 3 . The setup is based on the autoalign station (Newport) equipped with computer controlled nano-movers (50 nm repeatability) and an optical power meter. Laser sourced light was coupled into the waveguides with a tapered fiber attached to a nano-mover and the output intensity of the waveguides was monitored with an identical tapered fiber attached to another nano-mover. The light intensity was read by an optical power meter connected with the output fiber. TE-polarization state was adjusted by fiber polarization controller and verified before the actual measurements with an external polarizator. Commercially available (Oz Optics) tapered fibers produced 1/e2 spot size of 2.3 µm at 1530 nm. Correspondingly, the spot size at 1310 nm wavelength was 2.0 µm. The working distance of the fibers was about 14 µm.
At first, the intensity profiles at the output facet of the waveguides were examined to verify the proper waveguide operation. Initially, the input and output fibers were aligned optimally with the waveguide to produce a maximum light coupling at input and output facets of the waveguide. Figure 4(a) shows the output intensity profile of a waveguide with a ridge width of 2 µm i.e. similar to one shown in Fig. 2(b). A localized intensity maximum was observed when the input fiber was maintained at an optimal coupling position and the output fiber was scanned in xy-plane defined in Fig. 3. The measured intensity distribution suggests the proper waveguide operation. Figure 4(b) and 4(c) show the output intensity distributions at the output facet of a waveguide with a width of 6 µm. Inserts in these figures illustrate the focus spot of the input waveguide. When the input fiber was at the center of the waveguide, as illustrated in Fig. 4(b) with a red circle, the output distribution had a single peak indicating the excitation and propagation of the fundamental mode. When the input fiber was moved 2 µm towards the edge of the waveguide (along x-axis in Fig. 3), two intensity maxima were observed at the output facet due to coupling and propagation of the higher order mode. The experimental observations were also supported by the computational results for the TE-polarization state. Film-mode-matching (FMM) method of commercial Fimmwave software was used to simulate the optical field distributions of the propagating modes and their corresponding effective indices . According to the simulations, 2 µm waveguides showed single-mode operation at 1530 nm and 1310 nm wavelengths, whereas in a waveguide with a width of 6 µm, also higher order mode could propagate.
Transmissions of 2 µm waveguides with single-mode operation were analyzed in more detail at two wavelengths of 1530 nm and 1310 nm to estimate their performance and evaluate their usability in applications. Both the input and output fibers were scanned to optimize the fiber positions for the maximum power transmission through the waveguide. Table 1 shows the transmittance values and corresponding average loss of 5 measured waveguides referenced with a power that was obtained by measuring the power transmission from input fiber to output fiber at optimal coupling positions without a waveguide between them. The highest observed transmission values are underlined in Table 1.
Average transmittance of 32% was obtained at 1530 nm wavelength and 45% at 1310 nm wavelength. These results are now compared to the ideal total transmittance, which can be approximated asTable 2 on page 9. Intensity attenuation during the light propagation in a waveguide is expressed by the formula 15]. Reflections at the facets can be expressed by 18].Figs. 5(a) and 5(b) presenting the model fields of E fiber and E wg at 1530 nm wavelength, respectively. Gaussian profile in Fig. 5(a) is symmetric, while the intensity of the waveguide mode in Fig. 5(b) penetrates deeper into the substrate than into the air due to lowered index contrast between Ormocer and BK7 glass. Figure 5(c) presents the fiber-waveguide Γ2 coupling distribution based on Eq. (4), when the field profiles in Figs. 5(a) and 5(b) are applied. The computational coupling distribution resembles the experimentally measured coupling efficiency shown in Fig. 4(a). One should note, however, the difference in peak values. While in Fig. 4(a) the intensity is normalized to unity, in Fig. 5. (c) the maximum value of Γ2 is 0.9. This is theoretically the maximum intensity coupling efficiency between the waveguide and the used fiber, when they are aligned optimally and there is no scattering at the waveguide facet. Theoretical maximum coupling value at 1310 nm wavelength is 0.93.
The intensity profile and the spot size of tapered fiber were verified in order to validate the Gaussian filed distribution model. The analysis was carried out by coupling light directly from the input fiber into the output fiber without waveguides, as was made during the measurement of reference power described above. The results were then compared by substituting E wg=E fiber in Eq. (4), when it describes theoretical fiber-to-fiber coupling scheme. Figure 6(a) shows the measured and calculated coupling distribution at 1530 nm. Good correlation was observed suggesting that Gaussian field profile is reasonable approximation to model the field profiles of the tapered fibers. Same procedure was utilized at 1310 nm wavelength with similar results and it is illustrated in Fig. 6(b).
Computationally and experimentally obtained waveguide-fiber coupling distributions are illustrated in Figs. 7 and 8 for 1530 nm and 1310 wavelengths, respectively. Plots in Fig. 7(a) present the lateral sweep and in Fig. 7(b) the vertical sweep over the maximum coupling spot. They are actually normalized lines at zero-axis from Fig. 4(a) (measured) and Fig. 5(c) (calculated). Experimental and computational results showed fairly good agreement. In Fig. 7(b), a slight deviation can be observed in the vicinity of −2 µm, where the experiments showed higher coupling values than the theoretical model suggested. This indicates that the actual optical field of the guided mode penetrates deeper into the glass substrate than it was expected.Figures 8(a) and 8(b) show the same sweeps than in Fig. 7, but at a wavelength of 1310 nm. Results were consistent with the measurements carried out at 1530 nm wavelength. Simulated and measured plots were congruent with a slight deviation in vertical sweep at −2 µm region.
Theoretical optical field profiles were experimentally verified to depict well the intensity distributions of fabricated waveguides and tapered fibers used in characterization. This results in the presumption that field overlap calculation according to Eq. (4) represents, with reasonable accuracy, the maximum obtainable coupling between fiber and waveguide. The terms in Eq. (1) representing the total theoretical transmission of waveguide are listed in Table 2. These include the overlap term Γ2 from Eq. (4), reflection R according to Eq. (2) and intensity attenuation during light propagation in a waveguide P, see Eq. (3).
Theoretical total transmission at 1530 nm wavelength is 54%, while experimentally observed maximum transmission was 34% with average value of 32%. Correspondingly, at 1310 nm wavelength theoretical transmission is 67% vs. measured maximum value of 50% and average of 45%. Higher transmission at shorter wavelength is expected to be mainly due to the lower material absorption. The deviation between experimental and theoretical values is presumably due to scattering at waveguide facets and during the propagation in a waveguide. The facets were cleaved, and might cause variation in coupling efficiency between different waveguides. This might be the source for the deviation in different measured transmittance values.
The influence of the waveguide surface roughness on scattering loss depends on the waveguide geometry and refractive index contrast between the guiding layer and surrounding media [19–21]. Scattering loss increases with decreasing waveguide dimensions due to a stronger interaction of the guided modes with the waveguide surfaces. Furthermore, the scattering is stronger in high-index contrast waveguides compared to low-index contrast structures. For example, in sub-micron silicon-on-insulator (SOI) waveguides with high-index contrast the desired sidewall roughness can be less than 1 nm . On the other hand, in single-mode polymer waveguides coated with a cladding layer to produce low-index contrast, the acceptable surface roughness can be up to tens of nanometers . In this work, the waveguides did not have the over cladding layer leading to the index contrast of about 1.52 at waveguide-air interface. This kind of structure results in the high sensitivity of the propagation loss to the top surface and sidewall roughness. The cut-back technique, explained e.g. in Ref. 16., is a characterization method to distinguish the coupling and propagation losses. This would require well reproducible facet quality by using, for example, additional polishing step. Therefore, the analysis is limited in discussion of total power budget of the measurement setup. One can however consider, that since the total loss in measurements is about 2 dB higher than the theoretical loss, one can conclude that additional loss per unit length due to scattering must be below ~0.7 dB/cm. This outcome presumes that there is no scattering at input and output couplings.
Regardless of the actual loss mechanism, measured transmittance values up to 51% suggest that by using a simple mould preparation technique, it is possible to fabricate waveguides with reasonable good performance. The usage of the imprinting method results in a ridge type structure, when the dimensions of the ridge are solely determined by the mould. Instead, the thickness of the slab underneath the ridge depends on the resin properties and processing parameters. The thickness of slab layer might be further decreased or completely removed by subsequent process steps, such as, reactive-ion-etching . As it was demonstrated in this work, imprinted waveguides without an over cladding layer can have high transmittivity. These kinds of bare waveguide structures are useful, for example, in sensors devices, where the evanescence field probes the adsorbed material on a waveguide surface. It is also possible to coat the waveguides with a cladding layer to isolate optically the waveguides from the ambient and to provide mechanical shielding e.g. in short range interconnection applications. According to the AFM analysis, the surface roughness level of the positive tone resist based mould is less than 0.5 nm and replicated structures also possess very smooth surfaces with sub-nanometer RMS surface roughness value. This benefit can be potentially applied also in the fabrication of other polymer based devices besides the waveguides, when smooth surface is required.
In conclusion, single- and multimode optical waveguides based on UV-curable polymer were fabricated by imprinting method. Positive tone resist patterned on a silicon wafer was used as a mould allowing a simple and rapid mould preparation. RMS surface roughness of the replicated waveguides was 0.8 nm according to the AFM studies. The length of the waveguides was 30 mm, while their width was either 2 µm (single-mode) or 6 µm (multimode). The experimental observations on the mode profiles were also supported by the simulation results. Average power transmission of 32% at 1530 nm and 45% at 1310 nm were determined. These values include both the coupling loss and attenuation during light propagation in waveguides. The results suggest that the simple mould fabrication process might be useful technique for the fast prototyping of devices and the performance of waveguides can meet requirements for applications.
References and links
1. S. Y. Chou, P. R. Krauss, and P. J. Renstrom, “Imprint of sub-25 nm vias and trenches in polymers,” Appl. Phys. Lett. 67(21), 3114–3116 ( 1995). [CrossRef]
2. L. J. Guo, “Recent progress in nanoimprint technology and its applications,” J. Phys. D Appl. Phys. 37(11), R123–R141 ( 2004). [CrossRef]
3. H. Lan, Y. Ding, H. Liu, and B. Lu, “Review of the wafer stage for nanoimprint lithography,” Microelectron. Eng. 84(4), 684–688 ( 2007). [CrossRef]
4. K. Pfeiffer, M. Fink, G. Ahrens, G. Gruetzner, F. Reuther, J. Seekamp, S. Zankovych, C. M. Sotomayor Torres, I. Maximov, M. Beck, M. Graczyk, L. Montelius, H. Schulz, H.-C. Scheer, and F. Steingrueber, “Polymer stamps for nanoimprinting,” Microelectron. Eng. 61–62(1-4), 393–398 ( 2002). [CrossRef]
5. S. Obi, M. T. Gale, C. Gimkiewicz, and S. Westenhöfer, “Replicated Optical MEMS in Sol-Gel Materials,” IEEE J. Sel. Top. Quantum Electron. 10(3), 440–444 ( 2004). [CrossRef]
6. J. Albero, L. Nieradko, C. Gorecki, H. Ottevaere, V. Gomez, H. Thienpont, J. Pietarinen, B. Päivänranta, and N. Passilly, “Fabrication of spherical microlenses by a combination of isotropic wet etching of silicon and molding techniques,” Opt. Express 17(8), 6283–6292 ( 2009). [CrossRef] [PubMed]
7. T. Han, S. Madden, M. Zhang, R. Charters, and B. Luther-Davies, “Low Cost Nanoimprinted Polymer Waveguides,” in IEEE Conference Proceedings on Optoelectronic and Microelectronic Materials and Devices (2008), pp. 185–188.
8. C.-Y. Chao, W. Fung, and L. J. Guo, “Polymer Microring Resonators for Biochemical Sensing Applications,” IEEE J. Sel. Top. Quantum Electron. 12(1), 134–142 ( 2006). [CrossRef]
9. W.-J. Lee, J. W. Lim,S. H. Hwang, and B. S. Rho, “Imprint Master Fabricated by Ultra Precision Machining for Optical Waveguide,” in IEEE Proceedings of Opto-Electronics and Communications Conference and the Australian Conference on Optical (2008), pp. 1–2
10. S.-Q. Xie, J. Wan, B.-R. Lu, Y. Sun, Y. Chen, X.-P. Qu, and R. Liu, “A nanoimprint lithography for fabricating SU-8 gratings for near-infrared to deep-UV application,” Microelectron. Eng. 85(5-6), 914–917 ( 2008). [CrossRef]
11. J. Viheriälä, J. Tommila, T. Leinonen, M. Dumitrescu, L. Toikkanen, T. Niemi, and M. Pessa, “Applications of UV-nanoimprint soft stamps in fabrication of single-frequency diode lasers,” Microelectron. Eng. 86(3), 321–324 ( 2009). [CrossRef]
12. K.-J. Byeon, S.-Y. Hwang, and H. Lee, “Fabrication of two-dimensional photonic crystal patterns on GaN-based light-emitting diodes using thermally curable monomer-based nanoimprint lithography,” Appl. Phys. Lett. 91(9), 091106 ( 2007). [CrossRef]
13. J. T. Kim, J. J. Ju, S. Park, and M.-H. Lee, “O/E Integration of Polymer Waveguide Devices by Using Replication Technology,” IEEE J. Sel. Top. Quantum Electron. 13(2), 177–184 ( 2007). [CrossRef]
14. H. J. Levinson, Principles of Lithography (SPIE Press, 2005), Chap. 2.
15. Datasheet for Ormoclad, Microresist Technology, http://www.microresist.de/products/ormocers/pdf/pi_ormocore_clad_en_07062201_ls_neu.pdf
16. FIMMWAVE software by Photon Design.
17. G. Keiser, Optical fiber communications (McGraw Hill Companies, 2000)
18. G. T. Reed, and A. P. Knights, Silicon photonics (Wiley, 2004)
19. D. K. Sparacin, S. J. Spector, and L. C. Kimerling, “Silicon Waveguide Sidewall Smoothing by Wet Chemical Oxidation,” J. Lightwave Technol. 23(8), 2455–2461 ( 2005). [CrossRef]
20. K. P. Yap, A. É. Delage, J. Lapointe, B. Lamontagne, J. H. Schmid, P. Waldron, B. A. Syrett, and S. Janz, “Correlation of Scattering Loss, Sidewall Roughness and Waveguide Width in Silicon-on-Insulator (SOI) Ridge waveguides,” J. Lightwave Technol. 27(18), 3999–4008 ( 2009). [CrossRef]
21. H. Ma, A. K.-Y. Jen, and L. R. Dalton, “Polymer-Based Optical Waveguides: Materials, Processing and Devices,” Adv. Mater. 14(19), 1339–1365 ( 2002). [CrossRef]