1. Introduction

Waveguide-based molecular sensing can be used in a plethora of applications such as environmental gas sensing and biomedical sensing [1]. In order to analyse a liquid or gaseous sample, a purpose-designed optical waveguide can be utilised, along with a light source and a detector. Most molecules exhibit absorption bands in the mid-infrared (mid-IR) spectral region. More specifically, the functional group region (2.5–7 µm) and the fingerprint region (7–25 µm) [2] contain absorption bands that are unique to different molecules/specimens. Thus, for the fabrication of an on-chip waveguide-based sensing device, optical materials that are transparent in the mid-IR region are required.

Glasses have been extensively used in the field of photonics for centuries. Oxides and chalcogenides are the two most common optical glass families that are found in applications ranging from telecommunication fibres to optical sensors and lenses. However, oxide-based glasses suffer from strong absorption at wavelengths above 4 µm [3], limiting their use to visible and short-wave infrared applications. Chalcogenide glasses, on the other hand, have good mid-IR transparency, in some cases up to λ = 20 µm [4]. In addition to their wide infrared transparency, they are also characterised by their large refractive index (usually between 2 and 3), high optical non-linearity [5,6] and they can be deposited as thin films using standard techniques. For these reasons, chalcogenide glasses have been attracting significant interest as they meet all the requirements for the realisation of label-free integrated optical sensing devices and nonlinear optical devices for signal processing.

Despite their attractive optical properties, chalcogenides are very delicate materials [7] and require special care when used in the fabrication of integrated optical devices. It is well known that the propagation losses of integrated optical devices are much higher than those found in optical fibres. Such increased losses mainly stem from contamination and surface roughness. The predominant route for fabrication of optical waveguides usually consists of waveguide-patterning of a deposited film using photolithography and an etching process. As chalcogenides are very sensitive materials, they are very prone to attack by a wide range of species including alkaline solutions such as photolithographic developers and ionised gases [7,8]. This tends to produce defects on the surface of the layers which in turn can contribute to elevated propagation losses.

Germanium (Ge) based waveguide platforms suitable for long-wave IR applications have been recently reported in the literature [9,10]. These Ge based waveguides have shown propagation losses as low as 0.5 dB/cm between λ = 5.1 and 8 µm [9]. Despite the very low achieved propagation losses in this wavelength range, Si-O absorption as well as free carrier and multiphonon related absorptions can introduce higher losses at longer wavelengths (above 8 µm) [9]. Chalcogenide waveguides on the other hand, do not present such strong characteristic absorptions in the long-wave IR and most importantly can be easier and more cost effective to fabricate. Chalcogenide waveguide platforms fabricated using a) standard photolithographic patterning [4,11,12], b) hot-embossing [7,8], c) micro-transfer moulding [13] and d) photodarkening [14] are found in the literature. Waveguides with propagation losses lower than 0.5 dB/cm at λ = 1550 nm are reported [7,12] and have potential for the realisation of integrated optical devices. However, high temperature processes like hot-embossing are unsuitable for many chalcogenide materials [7] and this along with the complexity of chalcogenide waveguide patterning by etching [12] show that improved methods are required for fabricating low loss chalcogenide waveguides [8]. Waveguide propagation losses depend strongly on sidewall roughness, and in waveguides formed by etching the sidewalls, it is often necessary to use (partially-etched) rib waveguides [7,12] to mitigate the lossy interaction of the propagating mode with the roughness of the etched sidewalls. However, for devices relying on the interaction of the optical mode with the outer medium, such as sensing devices where high sensitivity is crucial, a fully-etched (ridge) waveguide geometry is preferable, thus minimising sidewall roughness becomes critical. Furthermore, ridge waveguides can help improve the compactness of integrated devices by providing reduced bend losses.

Excess propagation losses due to surface roughness can be reduced by avoiding etching the sidewalls of the waveguide. This can be done by patterning and structuring an easily processable substrate like silicon (Si) to create pedestal-like structures and then depositing the optical layers on top of these pedestals. For example, pedestal waveguides on dry etched Si substrates were reported by Carvalho et al. [15–17]. In this article, we demonstrate a waveguide platform which is based on the fabrication of Si pedestals using standard photolithography and a simple and cost-effective wet etch process (unlike the dry-etched approach used in [15–17]), followed by deposition of the optical cladding and core layers. The optical layers consist of GeAsSeTe (IG3) and GeAsSe (IG2) as the core and cladding, respectively. Example refractive indices of the bulk materials at λ = 7.7 µm are 2.7926 (IG3) and 2.5036 (IG2) [18]. With appropriate design, this waveguide platform has potential for use in several applications including biochemical sensing, security and nonlinear optical processing.

2. Methods

2.1 Si pedestal fabrication

For the fabrication of the Si pedestals a simple photolithography and etch process was utilised on Si wafers as shown in Fig. 1. Si (110) wafers can be used for anisotropic etching when exposed to appropriate etch solutions [19,20]. The most common etchants are KOH, EDP and TMAH. In this work, KOH was used as it provides good Si:etch mask selectivity and is not as toxic as the other two options. Precise alignment of the channels with the appropriate crystal direction, in this case <112>, allows fabrication of high aspect ratio features with vertical sidewalls [20]. We fabricated samples with Si pedestals using two alignment methods (i) the hexagon alignment method [21] and (ii) simple alignment to the primary flat of the wafer (<111>). The first method is accurate but requires a two-stage process and we confirmed that our simple second method, avoiding the hexagon patterning, provided similar vertical-walled pedestals after careful optimisation/alignment to the <111 > flat. The process used to create the Si pedestals was as follows: S1813 photoresist was spun on oxidised wafers with SiO₂ thickness of ∼ 380 nm and subsequently soft baked in an oven for 30 minutes at 90 °C. The wafers were then exposed to UV light through a light field chrome mask and developed in MF-319 developer. The SiO₂ mask was patterned using a 7:1 buffered hydrofluoric acid solution. The KOH solution concentration used for Si etching was 23% by weight and was produced by mixing high purity KOH pellets and DI water at elevated temperatures using a magnetic stirrer, until homogenous. KOH etching then took place in a beaker in a hot water bath which maintained the temperature to approximately 80 °C. The wafers were rinsed with DI water and dried using nitrogen after each developing and etching step.

 

Fig. 1. Si pedestal waveguide fabrication process

Download Full Size | PDF

2.2 Optical layer deposition

The chalcogenide waveguide layer deposition was performed using a thermal evaporator (BOC Edwards 306a) at room temperature and base pressure < 6×10⁻⁴ Pa. The evaporation source materials used were small pieces of unpolished glass plates supplied by Vitron. The etched silicon pedestal substrates were mounted in the evaporator on a rotating carousel. The deposition rate was controlled by adjusting the source current to achieve a deposition rate of ∼1.2 nm/s. To limit the material spitting observed during preliminary deposition of IG2 and IG3, baffled box sources were used. The resulting average layer thicknesses were 7.7 ± 0.7 µm and 4.1 ± 0.2 µm for IG2 (cladding) and IG3 (core), respectively. A ∼250 nm IG2 capping layer was also deposited on the waveguides for protection against environmental contamination once removed from the chamber.

2.3 Optical characterization

The refractive indices of the evaporated IG2 and IG3 films were measured at a wavelength of 1553 nm using the prism coupling method (Metricon) and ellipsometry (λ = 192–1697 nm), for comparison with the refractive indices of the bulk glasses [18] and to confirm the index contrast between core and cladding materials. While channel waveguide performance is measured in the Mid-IR, measurements at 1553 nm on thin films allow direct comparison between two experimental techniques and the bulk data at a spot wavelength.

Light of wavelengths between 7 and 11 µm from a tunable quantum cascade laser (QCL, Pranalytica) was end-fire coupled into the pedestal ridge waveguides via a ZnSe objective lens. The input focal spot diameter was chosen to closely match the expected modal width of the waveguide and was aligned for maximum excitation of the fundamental mode. The intensity of the scattered light emanating from the top surface of the waveguide was captured using a thermal imaging camera (FLIR) and fitted in MATLAB. The mid-IR propagation loss of the waveguides was estimated by fitting an exponential decay to the out-of-plane scattered light from the waveguide [15,22]. To image the propagating mode of the waveguides, another ZnSe objective lens was used which collected the output light spot and focused it onto the sensor of a thermal imaging camera.

3. Results and discussion

Figure 2 shows scanning electron microscope (SEM) images of a)–b) cross section and c) top view of the 18.5 µm wide pedestals (∼26.5 µm waveguide width). After close inspection, the pedestal sidewalls were found to be nearly vertical. These pedestals were patterned using simple primary flat alignment during photolithography. It is evident that the chalcogenide optical layers are of high quality as they show no signs of large grain structure and no interfacial cracks between the optical layers. Deposition can also be observed on the sidewalls of the silicon pedestal and primarily on the right side, due to the position of the evaporation source.

 

Fig. 2. a) SEM image of the cross section of the 26.5 µm wide pedestal waveguide, b) zoomed-in version of a), c) SEM top-view of waveguides.

Download Full Size | PDF

The rough surface seen between the pedestal waveguides in Fig. 2(c) is a result of {110} plane etching in the KOH solution. This is not expected to interfere with the operation of the waveguides deposited on the unetched top surfaces of the silicon pedestals.

The refractive index dispersion graphs for IG3 and IG2 films deposited on SiO₂ glass substrates measured using ellipsometry as well as for the bulk glasses from [18] are given in Fig. 3. The black vertical line intersects the dispersion graphs at λ = 1553 nm where the refractive indices were measured using the prism coupling method; the respective refractive index values and index contrast for the films and bulk glasses as well as the average film thicknesses are presented on Table 1. The index contrast between the IG2 and IG3 films is found to be approximately 0.34. It is evident that the refractive indices of the films are larger than those of the bulk starting materials. Nonetheless, we found a similar difference in the refractive index of the films to that of the starting bulk materials, as also observed by Bulla et al. [23]. Depending on the deposition rate and the composition of the starting material, differences in optical properties between the starting material and deposited film are usually present and can be attributed to a change in stoichiometry [23]. This phenomenon is very common in evaporated films of multi-component materials, due to the differences in the vapour pressure of the elements present in the compound glass.

 

Fig. 3. Refractive index dispersion of optical films and bulk glasses. The vertical line intersects the refractive index graphs @ λ = 1553 nm.

Download Full Size | PDF

Table 1. Refractive index data for the evaporated films and bulk materials at λ = 1553 nm [18]

View Table

The simulated and measured mode intensity plots and profiles for a waveguide with width of 26.5 µm and core thickness of 4.1 µm are presented in Fig. 4. For the selected average waveguide core thickness of 4.1 µm and at λ = 11 µm, the waveguide supports only one mode for widths ≤ 10.25 µm. The simulated fundamental mode profile shows the expected ellipsoidal shape and according to the estimated effective index (2.879), it is highly confined within the core which is a result of the large waveguide dimensions. In contrast, the measured mode profile captured using a thermal imaging camera, shows a more circular shape and has a much larger mode field width (1/e²); x-direction: ∼ 38 µm and y-direction: ∼ 40 µm, compared to the simulated mode; x-direction: ∼19.25 µm and y-direction: ∼ 4.46 µm. This is because the objective lenses at this wavelength have a minimum resolvable feature size of about 30 µm [24], so the mode image cannot be well resolved. Nonetheless, the image clearly confirms confined waveguiding.

 

Fig. 4. TM mode profile of the 4.1 µm (thick) × 26.5 µm (wide) waveguide at λ = 7.7 µm a) simulated x-direction, b) simulated y-direction. Fitting of light intensity of thermal picture of measured waveguide mode c) x-direction, d) y-direction, e) Simulated mode image, f) Measured mode image

Download Full Size | PDF

The propagation loss estimated using the scattered light intensity fit method (Fig. 5(a)) is presented against λ between 7 and 11 µm for both TM and TE polarisation on Fig. 5(b). Figure 5(a) shows the intensity of the scattered light emanating from the top surface of the waveguide along its length for λ = 7 µm, and this clearly shows a decaying trend related to waveguide propagation loss. Figure 5(b) summarises losses for all wavelengths and both polarisations and it can be seen that losses as low as 0.5 dB/cm are estimated for TM polarisation. The data were extracted from two dimensionally identical waveguides and show very similar losses for both polarisations. The small variations observed are attributed to measurement errors. The overall trend of the loss is decreasing with increasing wavelength and it averages 0.9 dB/cm. Considering that the dominant loss mechanism in these waveguides is scattering, this is expected as light scattering tends to reduce at longer wavelengths. The peak observed at λ=8 µm for TE polarisation was attributed to absorption related to Ge-O-Ge vibrations [25,26]. We speculate that the oxygen introduced on the deposited films after they are removed from the evaporation chamber is adsorbed into the core and cladding layers giving rise to oxygen related absorption. Reduction of oxygen contamination could be achieved by utilising a different capping layer material. This issue is currently being studied and improvement of the fabrication process is underway.

 

Fig. 5. a) Scattering data fit at λ = 7 µm b) Propagation loss vs Wavelength estimated for two waveguides (1 and 2) of the same width (∼26.5 µm)

Download Full Size | PDF

4. Conclusion

Low loss mid-IR chalcogenide pedestal waveguides were fabricated using a simple photolithographic and wet etching process followed by evaporation of a novel combination of three optical layers. Using the scatter intensity fit method, propagation losses averaging 0.9 dB/cm were estimated for wavelengths between 7 and 11 µm. We believe that the high structural and optical quality of the deposited waveguide layers as well as the adequate index contrast between the core and cladding make this material combination a very attractive choice for the fabrication of integrated mid-IR photonic sensors. Furthermore, the findings of this work demonstrate, for the first time, that the difficulties accompanying the processing of chalcogenides can be overcome by utilising standard Si processing.