1. Introduction

Optical interconnects offer a promising solution to the communication bottleneck imposed by conventional electronic circuitry and constitute an intensive area of research both in industry and academia. Silicon photonics based waveguides function by having a large refractive index contrast between a silicon core and a buried oxide under-cladding and are now being introduced for chip-level optical interconnects. Such an approach is not practical for board-level (1-20 cm) dimensions because of the waveguide losses and the fabrication cost. Polymer based optical interconnects are cost effective and have the additional advantages of scalability, process flexibility and high coupling efficiency to the optical fibers. Single mode polymer waveguides provide an opportunity for communication between integrated photonic chips and single-mode optical fibers on the package and board level [1]. Several techniques for the fabrication of polymer waveguides have been reported such as photolithography [2], reactive-ion etching [3], UV imprinting [4–11], laser ablation [12], direct-write lithography [13–16] and micromolding in capillaries (MIMIC) [17,18]. The small refractive index contrast between the polymer materials and typical cladding layers limits the bending radii and the choice of the substrates. Cladding layers are typically used in polymer waveguides mainly to generate symmetric field profiles to ensure high-coupling efficiency with corresponding symmetric inputs (fibre inputs, VCSELs or other laser sources) and to protect the underlying waveguides. In order to use air as upper cladding to achieve sharper bends, a substrate with a refractive index value comparable to air is desirable for compact polymer waveguides with symmetric field profiles. Different techniques have previously been adopted to achieve sharp bends for polymer waveguides but those techniques result in asymmetric field profile [19,20]. In principle, a honeycomb like material having refractive index comparable to air can be used as a low index under-cladding for polymer waveguides to achieve sharp bends with symmetric field profiles.

With this latter point in mind it has been known for some time that 3D Colloidal photonic crystals (CPCs) can be fabricated using the “bottom-up” colloidal self-assembly approach and used for various applications including sensing [21,22], amplification [23] and manipulation of light [24,25]. While the “bottom-up” colloidal self-assembly approach is both simple and cost effective, the deliberate creation of defects in self-assembled 3D CPCs is not so straightforward [26]. The engineering of artificial extrinsic defects, including point, line and planar defects, is critical to the fabrication of functional photonic devices such as optical waveguides, splitters and couplers. Such artificial defects in CPCs may offer functionality to the CPC, e.g. a line defect can direct the propagation of light and act as an optical waveguide [27–29]. Laboratory controlled colloidal self-assembly tends to form high quality face centered cubic (FCC) colloidal crystals. Thus the self-assembly route requires to be combined with other micro-engineering techniques, such as lithography, in order to introduce the features required for waveguides, cavities, etc., that have the added functionality to make useful devices.

The combination of colloidal self-assembly and lithography has been used to build line defects within self-assembled CPCs. Ye et al. [30] fabricated micrometer-scale air-core line defects within inverted colloidal crystals. Photoresist line patterns were made on a silicon substrate using optical lithography followed by the self-assembly of a polystyrene (PS) CPCs on top of the line pattern and substrate. After inverting the structure, and removal of the PS and photoresist, an air-core line defect between the substrate and the so called ‘inverse opal’ was produced. Three similar methods have been used to fabricate line defects within CPCs using photolithography [31–33]. In these cases, the line defects were buried within the CPC using multiple stages. CPC is first deposited onto a silicon substrate, onto which is printed a temporary pattern of polymeric material, then a second CPC is deposited in order to completely bury the pattern. Finally the polymeric material is selectively removed leaving behind a colloidal crystal with an air-core line defect. Another approach used was to construct a Si₃N₄ rib waveguide surrounded by a silica CPC [34]. While these linear defects mentioned above have been proposed as optical waveguides, only the Si₃N₄ rib waveguide paper reports the output power transmission spectrum of their colloidal-clad waveguide.

Turning now to the MIMIC approach, this is a soft lithography method, which is a simple and easy to use, and has potential for applications in mass production. MIMIC can be used for the production of micrometer sized lines and other shapes onto substrates, and has been proposed for use in producing waveguides [17,18]. The flexible micro-mold plays an important role in micro-channel formation on a flat solid substrate and has been used in the fabrication of micro gas sensors [35], photonic biosensors [36] and CPC miniaturized spectrometers [37].

In this paper, we utilize for the first time the MIMIC approach combined with a novel co-assembly method developed in our laboratories for the formation of a low refractive index layer in the form of an inverse colloidal photonic crystal onto which polyimide waveguides are subsequently fabricated.

 

Fig. 1 Schematic of the fabrication process: Firstly a co-crystallized PMMA-SiO₂ film is deposited on a silicon substrate to act as an under-cladding for the polyimide waveguides. The waveguides are made by depositing Durimide 112A and printing a prepared PDMS mold under vacuum. Chloroform is used after soft-baking the Durimide to swell and release the PDMS mold and also remove the PMMA spheres from the underlying co-crystallized film. This results in a low refractive index SiO₂ inverted opal beneath the Durimide waveguides. An oxygen plasma is used to thin the residual layer. The plasma also aids the removal of any retained PMMA spheres.

Download Full Size | PDF

2. Waveguide fabrication and characterization

Co-crystallized colloidal poly methyl methacrylate (PMMA) and silica opaline thin films were deposited onto silicon wafer substrates using 368 nm diameter PMMA spheres and hydrolyzed silica from a tetraethyl orthosilicate (TEOS) precursor solution by a co-assembly method used in our laboratory [38,39], which is similar to that described by Hatton [40] and Wang [41]. In a typical process, an aqueous suspension of PMMA particles of 0.05-0.1% by volume was stirred for 1 hour. After this a solution of TEOS: Ethanol: 0.1M Hydrochloric acid (HCl), 1:8:1 by volume was added to the PMMA suspension and stirred for a further 30 minutes. Hydrophilic silicon wafers were held vertically in a vial in which the suspension and precursor had been previously added; the vial was then placed in an oven at 55°C for 24 hours. Upon evaporation of the solvent, a co-crystallized PMMA-silica film was produced, typically with 6-12 layers of PMMA particles assembled into an FCC structure, bonded by a silica backbone as shown in Fig. 2(a, b).

The reflection response of the deposited co-crystallized structure was measured at an angle of 10° to the normal in order to observe the Bragg reflections. Light from a white light source (Ocean Optics HL2000) was collimated and incident on the crystal. The reflected light was collected and coupled to a spectrometer (Ocean Optics USB4000) in order to measure the reflecting wavelengths. A weak Bragg reflectance peak was measured at a wavelength of 809 nm as shown in Fig. 2(c). The effective refractive index of the co-crystallized structure was measured to be 1.44 ± 0.01 by using angle resolved spectroscopy of the Bragg peak [42]. The low intensity of this reflection was due to the small refractive index contrast between PMMA spheres and the silica backbone. The resonance wavelength of the silica inverted opal crystal after the removal of the PMMA spheres was also investigated. A strong Bragg reflection peak was measured at 612 nm with a Full Width Half Maximum (FWHM) of 41 nm as shown in Fig. 2(c). The refractive index of the inverted structure was calculated to be 1.01 ± 0.01. The presence of a Bragg peak around 600 nm shows that inverted crystal will act as a low refractive index substrate and will not affect the spectral propagation through the waveguide at telecommunications wavelengths. A PMMA colloidal crystal without the silica backbone was also prepared using the 368 nm PMMA spheres in order to calculate the refractive index of the PMMA spheres. Reflection measurements from this PMMA CPC displayed a Bragg peak at 811 nm as shown in Fig. 2(c). The effective refractive index of the PMMA CPC was measured to be 1.34 ± 0.01 from which the refractive index of the PMMA spheres was calculated to be around 1.44 ± 0.01, which is very similar to that of silica. The small refractive index contrast between the PMMA spheres and the silica backbone explains the weak Bragg peak at 809 nm observed for the co-crystallized sample.

 

Fig. 2 Scanning Electron Microscope (SEM) images of (a) co-crystallised PMMA-SiO₂ film deposited on silicon substrate on which the polyimide was deposited for the waveguide (It should be noted that the PMMA spheres are readily shrunk in the SEM beam), (b) the edge of the co-crystallised film. (c) Measured reflection at 10° to the surface normal for PMMA CPC, PMMA-SiO₂ co-crystallised and SiO₂ inverted opal structures.

Download Full Size | PDF

A flexible micro-mold of poly-dimethylsiloxane (PDMS) was produced by replication of a master mold on a silicon wafer. The master mold had waveguides of different widths (1 µm-10 µm) at a spacing of 50 µm. The master mold was fabricated onto the silicon wafers by standard photolithography methods. The PDMS micro-molds were made from a PDMS elastomer kit (Dow Corning Sylgard 184). The base-to-catalyst mixing ratio from the PDMS kit was 10:1. The mixture was poured over the silicon master mold and degassed in a vacuum chamber for 20 minutes, and then cured at 60°C overnight. The PDMS micro-mold was then peeled from the master mold. Prior to infiltration of the capillary lines in the micro-mold, the PDMS mold and the co-crystallised PMMA-silica film were cleaned in air plasma for 5 minutes. A small droplet of Fuji Film polyimide adhesive having a refractive index of 1.8 around telecom wavelengths, Durimide 112A [43], was deposited onto the PMMA-silica film and pressure applied. The pressure was maintained while the substrate and mold was placed into a vacuum chamber at 0.1 MPa for 1 hour. Upon removal from the vacuum the Durimide 112A was soft-baked at 135°C for approximately 90 seconds, according with the manufacturers’ guidelines. After curing, chloroform was used for 30-60 seconds, both to swell the PDMS mold to release it from the Durimide MIMIC lines, and secondly to remove the PMMA spheres, thus leaving behind Durimide capillary lines on top of an inverted silica opal structure, Fig. 1. Durimide 112A was chosen because of its high refractive index value and for its stability against solvent attack (e.g. chloroform and acetone) which was necessary for the processing of the final waveguides. This fabrication method can also be applied to different polymer waveguide materials.

A residual layer (1.66 µm thick) of Duramide remained adjacent to the waveguides after the fabrication. This residual layer acts as a cover on the co-crystallized PMMA-silica film and can impede the complete removal of the PMMA spheres. We simulated the light propagation in the waveguides in order to study the effect of this residual layer on the waveguide mode. The calculated fundamental mode at 1550 nm wavelength for a 3.0 µm x 3.7 µm waveguide is shown in Fig. 3(a). Calculations show that the waveguide can accommodate a 2nd order transverse mode as shown in Fig. 3(b). The height of the waveguide increased due to the residual layer leading to the excitation of the 2nd order mode. Calculations show that a global etch-back of the residual layer can result in elimination of the leaky mode shown in Fig. 3(b). The calculated fundamental mode of the waveguide after etching of the residual layer is shown in Fig. 3(c). The fabricated waveguide is highly multimoded and while higher order modes can be excited into the waveguide by off-center excitation, those modes will have larger propagation losses. An oxygen plasma was used to remove the residual layer as shown in Fig. 3(d). This global etch-back also removes any remaining PMMA spheres. The waveguide sample is shown in Fig. 3 (e, f) where the residual layer has been etched leaving a waveguide with a height of almost 3.7 µm. The shape of the waveguide is not rectangular but trapezoidal. The shape of the waveguide can be controlled with increasing experience using this original fabrication method. It should be mentioned here that the sample was diced to get smaller chips. The dice cuts were at right angles to waveguides to get sharp facets.

 

Fig. 3 (a) Calculated fundamental mode for the waveguide (3.0 µm x 3.7 µm) with a residual layer. (b) The higher order mode that can be excited in the waveguide with a residual layer. (c) Calculations show that the fundamental mode is excited upon removal of the residual layer. (d) Sketch showing the elimination of the residual layer by plasma etching. (e) SEM image of top view of the waveguide after oxygen plasma treatment. (f) SEM image of edge of the waveguide of ~3.7 µm in height.

Download Full Size | PDF

The transmission response of the plasma treated waveguides was measured using a fiber-coupled super-continuum light source from Fianium covering the E, S, C and L bands of the communication wavelengths (1300 - 1600nm). Single mode lens ended fibers were used to couple to the input and output waveguides of the fabricated waveguides and the power maximized in order to measure the transmitted powers. The output fiber was coupled to an Optical Spectrum Analyzer (OSA) in order to measure the spectral response of the waveguide. Polarization of the input light was controlled using a polarizer. The measured transmission spectra for both TE and TM polarizations of the incident light for the 3.0 µm x 3.7 µm waveguide is shown in Fig. 4(a). The Fabry-Perot resonances from the multiple passes through the waveguide due to the facet reflections are shown in Fig. 4(b). The calculated cavity length associated with these resonances is in agreement with the ~1 mm waveguide length. The mode profile of the waveguide was measured by scanning the output fiber in horizontal and vertical directions, Fig. 4(c). This shows that waveguide was guiding the fundamental spatial mode as expected from the simulations.

 

Fig. 4 (a) Measured transmission for both TE and TM polarizations of light. Measurements show broadband transmission from the 3.0 µm x 3.7 µm waveguide. This loss also includes the facet coupling losses. This is the measured spectra for a single waveguide not average of multiple waveguides. The increased noise around 1300 nm wavelengths is due to the laser source. (b) Measured Fabry-Perot resonances for the same waveguide. The calculated length of 1mm corresponding to the resonances around 1540 nm wavelengths is in agreement with the actual waveguide lengths. (c) Measured output profile at 1550 nm wavelength showing the fundamental spatial mode.

Download Full Size | PDF

3. Conclusion

In conclusion, we have experimentally demonstrated a functioning air-clad polymer based waveguide fabricated on an inverted photonic crystal. Our new approach will enable the realization of very compact polymer waveguide circuits. Wavelength dependent waveguides for sensing applications can be obtained by designing the underlying photonic crystal to be resonant which will be reflected in the transmission response (resonance dip/peak) of the waveguide. The resonance will shift with the penetration of analytes into the pores of the underlying crystal resulting in a label-free sensor. In addition, we note that in principle, light emission and lasing could be obtained if the polymer was doped with suitably luminescent dyes or nanoparticles.