1. Introduction

Mirrors and filters based on planar, multilayer thin film stacks of two (or more) dielectrics have long been central elements of optical technology. A ubiquitous example is the quarter wave stack (QWS), which exhibits high reflection bands at wavelengths satisfying a Bragg resonance condition. It was recently recognized [1–2] that omnidirectional reflection (near unity reflectance for all incidence angles and polarization states, over some ranges of wavelength) is possible from a QWS or similar multilayer structure. The key prerequisite is a sufficiently high refractive index contrast between the two dielectric materials that comprise the stack. As an example (assuming air incidence), if the lower refractive index material has n~1.5 then the higher refractive index material must have n~2.3 or greater [1–4]. The selection of viable materials is governed by the minimum index contrast requirement, and also by the need for excellent chemical, thermal, and mechanical compatibility between the materials. Planar omnidirectional reflectors with fundamental stop bands in the visible [5–6], near infrared [7–8], and mid infrared [2] have been reported. Further, novel waveguides [9–11] and cavities [12] based on these reflectors have been explored recently. In the majority of cases, the materials employed (such as Si/SiO₂, Si/Si₃N₄, or SiO₂/TiO₂) necessitate high temperature processing or annealing of the thin film layers. This is detrimental to the integration potential of the mirrors, and can lead to failure of the films (by crystallization, cracking or delamination) [6–7].

Chalcogenide glasses (sometimes classified as inorganic polymers, because of their bonding structure and thermo-mechanical properties) and organic polymers offer unique options for low temperature processing of high index contrast structures [13]. Hollow fibers with omnidirectional reflector claddings fabricated in chalcogenide glass and polymer have recently been the subject of much interest [14–15]. The fibers are drawn from a pre-form at elevated temperature, illustrating the compatibility (good mutual adhesion, well matched thermal and elastic properties, etc.) of these two classes of materials. We recently exploited this compatibility in the manufacture of low loss, high index contrast integrated waveguides [16]. Following that work and employing an approach similar to that used in [2], we have fabricated omnidirectional reflectors with fundamental stop bands near 1750 nm wavelength.

2. Fabrication

We chose the commercially available chalcogenide glass Ge₃₃As₁₂Se₅₅ (IG2 glass, Vitron AG, France) for this work. It is a well-known and widely characterized alloy (also known as AMTIR-1 and TI-20 glass [17–18]) with a reasonably high glass transition temperature (T_g~360 C) and interesting acousto-optic [17], magneto-optic [18], and 3^rd order nonlinear optic [19] properties. IG2 has high refractive index (n~2.6) and good transparency above its electronic absorption edge (for wavelengths longer than approximately 700 nm), making it a viable photonic crystal medium in the near infrared [20]. Thin films of IG2 glass were deposited by thermal evaporation onto room temperature substrates. The thermal evaporation was performed at a base pressure of ~2 × 10^-6 Torr. In order to improve film homogeneity and thickness uniformity, the substrate was fixed to a planetary rotation system. The typical deposition rate was 2-3 nm/second, and film thickness was controlled using a standard deposition rate and thickness monitor (MCM-160, McVac Inc.).

For the lower index layers, a commercial polymethylmethacrylate (PMMA) resist (950 PMMA A2, Microchem Corp.) was used in initial attempts. While the resultant multilayer structures exhibited promising optical properties, they did not behave well on cleaving (see Fig. 1(a)). Better results were obtained using a polyamide-imide (PAI) polymer (Torlon AI-10, Solvay Advanced Polymers), as shown in Fig. 1(b). PAI is a high performance, aromatic thermoplastic, used as a structural material in the automotive and aerospace industries [21]. We have recently investigated its use as an integrated optics material [22], including as a cladding material for chalcogenide glass based waveguides [16]. PAI has numerous attributes, including high thermal stability (T_g~280 C) and good adhesion to many other materials (it has been used as an adhesive layer for fluorinated polymers employed in nonstick cookware applications). It has good transparency in the wavelength range ~600 nm to ~1700 nm and a refractive index n~1.6. PAI powder was dissolved in a standard organic solvent and the solution was subsequently passed through a 1 micron filter paper to remove large particulates. Thin films were realized by spin-casting and curing at temperatures up to 160 C in a vacuum oven under a nitrogen atmosphere. Note that the underlying IG2 layers were simultaneously annealed with the curing of each PAI layer; this induces structural relaxation and enhances the stability of the glass layers.

 

Fig. 1. (a) SEM image of the cleaved facet of a multilayer stack of IG2 glass and PMMA polymer. The polymer layers deformed badly on cleaving, and there was clear loss of adhesion between layers. Inset: transmission scan of the PMMA-based multilayer, providing evidence of a stop band in the 1500 to 2200 nm wavelength range. (b) SEM image of the cleaved facet of a 5.5 period multilayer stack of IG2 glass and PAI polymer (starting and ending with polymer). The silicon substrate is visible at the bottom of the image. The wafer was cleaved after a brief immersion in liquid nitrogen.

Download Full Size | PDF

Along with providing a sufficient index contrast for omnidirectional reflection (a normalized omnidirectional bandwidth of ~6% is predicted [3]), IG2 and PAI have sufficient thermo-mechanical compatibility for processing of multilayer stacks. Aside from similar glass transition temperatures, they have reasonably well matched thermal expansion coefficients (approximately 12×10^-6 C^-1 and 30×10^-6 C^-1 for IG2 and PAI, respectively) and Young’s modulii (approximately 15-20 GPa for IG2 films and approximately 5 GPa for PAI films).

Multilayer reflectors were deposited on 4 inch, single side polished silicon wafers. The target layer thicknesses (~170 nm and ~270 nm for IG2 and PAI, respectively) were based on the goal of achieving an omnidirectional reflection band centered near 1550 nm, according to the design rules described elsewhere [3]. The targeted layer structure is essentially a QWS, with its normal incidence fundamental Bragg resonance at approximately 1725 nm wavelength.

 

Fig. 2. (a) SEM image of a 6.5 period multilayer stack, starting and ending with a PAI polymer layer. (b) Surface relief map obtained by AFM in tapping mode over a 1 μm × 1 μm area.

Download Full Size | PDF

SEM images of typical multilayer stacks are shown in Figs. 1 and 2. We fabricated stacks with up to 13 layers, but stacks with higher numbers of layers appear feasible. Numerous cleaved edges were examined, and there was no sign of film delamination or cracking when PAI was used as the low index material. As we have noted previously [16], PAI is a tough polymer with a tendency to stretch and deform on cleaving. The cleaved edges shown were obtained by dipping the samples in liquid nitrogen just prior to cleaving. This resulted in improved cleaves, but some mechanical deformation within the PAI layers was still apparent (especially for the layers nearest the silicon substrate). Under no circumstances did we observe significant loss of adhesion between IG2 and PAI layers. The SEM images indicated that the as-grown layers were slightly thicker than targeted. However, there was good consistency for both the glass and polymer layers. As-grown layer thickness was estimated as ~190 nm and ~290 nm for the IG2 and PAI layers respectively. These estimates were used in the modeling described in Section 3. Since the first submission of this manuscript, we have grown mirrors with improved control over the layer thickness and having omnidirectional bands near 1550 nm wavelength. We estimate that the error in layer thickness is within +/- 5 nm for both the evaporated glass and spun-cast polymer layers.

3. Characterization

In the following, we focus on the results for a 6.5 period stack (see Fig. 2(a)). The surface morphology was characterized using an SEM, with a scanning profilometer, and by atomic force microscopy (AFM). Long-range scans revealed excellent planarity. From short-range scans (in the 0.3 to 13 μm range) by the AFM (Digital Instruments Nanoscope IV) in tapping mode (see Fig. 2(b)), the RMS roughness was consistently estimated to be on the order of 0.2 nm. Such smooth interfaces are critical to the application of such mirrors to waveguides and microcavities, and indicate that thermal expansion related stresses (which can cause wrinkling, etc.) were not a significant problem for our process.

Optical transmission scans were obtained using a spectrophotometer (Perkin-Elmer Lambda 900 UV/Vis/NIR dual beam spectrophotometer). A typical scan is shown in Fig. 3(a). Since the mirrors were grown on single-side polished silicon wafers, transmission spectra (specific to the multilayer stack) could be assessed only in the wavelength region above 1100 nm. Even at longer wavelengths, the rough back surface of the silicon wafers precluded quantitative analysis of the mirror’s transmittance. However, the shape of the transmission spectrum provides clear evidence for the expected normal incidence stop band in the 1700 to 2200 nm wavelength range (see the simulated plots shown in Fig. 4 for comparison).

 

Fig. 3. (a) Transmission scan of the 6.5 period stack, obtained at nearly normal incidence with a spectrophotometer. (b) Refractive indices used in the simulations for IG2 glass (dashed) and PAI polymer (solid).

Download Full Size | PDF

 

Fig. 4. Simulated and experimentally measured reflectance versus wavelength, for TM (left panel) and TE (right panel) polarized light. For each polarization, plots are shown for incidence angles of 0, 20, 34, 48, 62, and 76 degrees from normal (in order from top to bottom). The simulated curves are shown as solid blue lines and the measured points are indicated by the red symbols. The gap in the measured data near 1400 nm is inherent to the instrument used. The simulated spectra are for IG2 and PAI layer thicknesses of 190 nm and 290 nm, respectively. The ellipsometry instrument does not allow normal incidence reflectance curves to be obtained.

Download Full Size | PDF

The optical response was tested in reflection at various angles of incidence and for both TE and TM polarization, using a variable angle spectroscopic ellipsometry (VASE) instrument (J. A. Woollam Co VB-250 Ellipsometer Control Module coupled with HS-190 High Speed Monochromator System). The theoretical reflectance spectra were simulated using a standard transfer matrix approach and the refractive indices plotted in Fig. 3(b). The dispersion relation for PAI polymer is from our previous work [22]. The Sellmeier dispersion relation for IG2 glass was taken from reference [23]; we should note that the stated range of validity for that expression is 1-14 μm. However, we have previously extracted the refractive index dispersion for our evaporated IG2 films and find that the expression from [23] gives reasonable agreement down to 500 nm. The substrate index was fixed at n=3.5. For simplicity we have neglected the effects of absorption in our simulations, limiting their accuracy at shorter wavelengths. IG2 glass becomes absorptive below ~700 nm and PAI polymer below ~600 nm. In spite of this simplification, the main features of the experimental and theoretical spectra are in good agreement over the entire wavelength range considered.

The experimental and theoretical reflection spectra are plotted for comparison in Fig. 4. The only fitting parameters in the simulation were the layer thicknesses, which as mentioned were estimated from SEM images. There is excellent agreement between the spectra, especially for angles of incidence less than 48 degrees. A first-order stop band in the region above 1500 nm is verified, as is a third order stop band near 600 nm. The multilayer structures appeared increasingly red in reflected light as layers were added, in keeping with the existence of this latter stop band.

As is well known [1–4], an omnidirectional stop band (if present) is bounded by the short wavelength edge of the normal incidence stop band and the long wavelength edge of the glancing incidence stop band for TM polarized light. The as-fabricated mirrors are predicted to have an omnidirectional stop band with an approximate bandwidth of 150 nm, with the short wavelength band edge near 1700 nm. The presence of this stop band is partially verified by the reflectance spectra obtained, which were limited by the instrument used to wavelengths below 1700 nm.

4. Summary and conclusions

Using a chalcogenide glass and a high temperature organic polymer, we have demonstrated planar reflectors with an omnidirectional stop band centered near 1750 nm wavelength. Reflectors with up to 13 layers were fabricated. In addition to exhibiting the expected optical response, the multilayer structures showed good mechanical integrity, adhesion, and planarity. This was attributed to the similar thermo-mechanical properties of the materials used. Given the low temperature processes employed, these reflectors might enable new types of integrated waveguides and cavities on electronics platforms. Chalcogenide glasses are promising materials for integrated acousto-optics, magneto-optics, and nonlinear optics. Photonic bandgap media based on chalcogenide glasses are expected to enable novel all-optical switching and signal processing functions [24].