1. Introduction

Chalcogenide glasses (ChGs) have a number of unique properties that make them attractive for fabricating planar optical waveguides. Their high linear refractive index (2-3) and broad infrared transparency suggest that compact optical circuits made from sub-wavelength, single mode waveguides can be fabricated for telecommunications, optical sensing and mid-IR science. Their low photon energy makes them attractive hosts for rare earth doping which could lead to waveguide lasers operating on long wavelength infrared transitions [1]. Their large ultra-fast third order optical nonlinearity has already been successfully used for all-optical signal processing of telecommunications signals [2], and in the future should lead to the generation of broad-band super-continuum further into the mid-IR than achievable with other materials leading to bright sources for spectroscopy and sensing [3]. Finally there exists a wide range of glass-forming systems and compositions and this allows material properties to be fine-tuned for specific applications. Motivated by such possibilities over the last decade interest has increased in the fabrication and use of high performance planar waveguides made from chalcogenide glasses. In this paper we review the current state of chalcogenide waveguide devices and present recent results on the production of buried channel and rib waveguides as well as highly nonlinear nanowires from Ge_11.5As₂₄Se_64.5 chalcogenide glass.

2. Material properties

Chalcogenide glasses contain one or more of the chalcogen elements from group 16 of the periodic table (S, Se, Te but excluding O) covalently bonded to networks formers such as As, Sb, Ge, Si, etc. They are, therefore, composed of relatively weakly bonded heavy elements and this leads to many of their most important optical and physical properties. Relatively low bond energies results in optical gaps in the visible or near IR as well as low to moderate glass transition temperatures (T_g≈100-400C). The low vibrational energies of the bonds extends optical transparency to ≈8µm for sulphides; ≈14µm for selenides and beyond 20µm for tellurides although impurities often exist in bulk glasses or films which significantly limit long wavelength transmission. For example “high purity” chemical elements used for glass fabrication often contain significant quantities of C, O and H and these lead to absorption bands between 1.4 and 15µm [4]. As a consequence approaches for purifying the starting materials have been developed that can reduce impurity levels to ≈10⁻⁵wt% and this is adequate particularly for waveguide applications where path lengths can be short (10 s cm) [5].

One of the most widely studied and technologically important properties of ChGs is their photosensitivity [6,7]. This is associated with local changes in the chemical bonds induced by light via the formation of self-trapped excitons which change the valency of neighboring atoms and hence the bonding. When these photo-induced excitons annihilate either the original bonds are restored or, because of the inherent steric flexibility of chalcogenides, different bond configurations are created which can change the macroscopic properties of the glass, such as its band gap or refractive index. In this case the phenomenon is known as photo-darkening and the change the refractive index can be used for direct writing of waveguides and Bragg gratings into chalcogenide films. Photo-darkening is, however, just one example of the rich range of phenomena associated with optically-induced changes of the chemical bonds. Others include photo-diffusion, photo-fluidity, photo-crystallization and photo-induced birefringence some of which have been used for waveguide fabrication. It is worth noting that bond switching can also be induced by energy in the form of x-rays, electron or ion beams or heat and leading to applications as x-ray or electron beam resists [8].

Whilst these phenomena have been a fertile ground for research over several decades, photo-sensitivity is generally undesirable in optical waveguides. Its presence underlines the fact that the glass network is unstable and, therefore, may evolve changing the waveguide properties when exposed to high optical fluence, such as those necessary for all-optical signal processing, or when stored even at room temperature. Thus, it is important to ask whether glasses exist that are thermally stable and have negligible photosensitivity and how can we identify them from within the wide range of glass forming compositions that exist? In this context it is worth noting that a photo-stable glass in the Ge-As-Se system has been reported which indicates that this is indeed a valid question [9].

A picture of how the properties of ChGs depend on composition is emerging from the literature. The work of Phillips [10], Thorpe [11], Tanaka [12] and Boolchand [13] have substantially improved understanding of the network topology at least of ternary chalcogenides such as the Ge-As-Se system. Thorpe showed the existence of a phase transition from a “floppy” to “rigid” phase at MCN≈2.4 whilst Tanaka identified a second transition from “rigid” to “stressed rigid” phase at MCN≈2.67 (MCN = Mean Coordination Number which equals the sum of the products of valency times atomic abundance of the constituent atoms). Using temperature modulated differential scanning calorimetry, Boolchand [13] claimed a so-called “intermediate” phase existed in Ge-As-Se glasses for MCN in the range 2.28-2.44 and that this corresponded to a self-organizing network characterized by zero non-reversing heat flow. Recent studies comparing the properties of Ge-As-Se films with those of bulk glass suggest that for some compositions films formed by thermal evaporation also self organize so that they form with chemical bonds indistinguishable from those of the bulk glass [14]. This is in fact quite unusual for chalcogenides since creating a film by physical vapour deposition onto a cold substrate is a highly non-equilibrium process which results in the film containing large numbers of defect bonds or molecular clusters. As a result it is usual for evaporated ChG films to have different physical properties (band-gap, refractive index, density, etc) from the bulk glass. Thus the self-organizing nature of networks is important for film forming and creates films that have been shown to be thermally stable. This has motivated us to study compositions such as Ge_11.5As₂₄Se_64.5 which is an example of a thermally stable nonlinear optical glass suitable for waveguide fabrication.

3. Approaches to waveguide fabrication

The first reports of waveguides made from ChGs appeared in the 1970s from researchers in Russia [15] and Japan [16,17]. Ohmachi [16] demonstrated the acousto-optic effect in 1-D slab waveguides made from As₂S₃ deposited on a Lithium Niobate crystal. Towards the end of the 1970s photo-darkening was applied to make 3-D waveguides (2-D waveguides with bends) in As-S-Se-Ge glasses [17]. By irradiating the films with an Ar laser operating at 514nm, photo-induced index changes as high as 0.03 were achieved at 1064nm and the lowest losses were reported to be 0.4dB/cm at 1064nm.

Following these initial reports, there was little progress in the field until the late 1990s when new research on waveguide fabrication appeared. Meneghini et al. [18] reported the use of ion implantation both for waveguide writing and also to achieve doping of the chalcogenide waveguide with Nd ions. Photo-darkening using light at the band edge was again applied to write waveguides in films of As-S(-Se) [19], Ga-La-S [20] and AsSeTe/AsSe [21] whilst Efimov et al. [22] showed that femtosecond pulses at 850nm, well away from the band edge were equally effective. Photo-darkening was shown to be a useful route for fabricating Bragg gratings into planar ChG films [23]. Waveguide gratings were first produced reported with a reflectivity of 86% at 1550nm in As-S-Se/As-S channel waveguides by exposure to near bandgap light at 514nm [24]. More recently ultra-strong apodized gratings with in–band rejection of >27dB (measurement limited) with low chirp and low side-lobe levels were written into As₂S₃ rib waveguides [25] and used for experiments on all-optical regeneration [26]. Similarly high performance sampled gratings have also been demonstrated [27,28].

In-diffusion of silver into ChG films can be used for waveguide fabrication due to the increase in the refractive index associated with silver doping. Fick et al. [29] reported the use of a thermal process to in-diffuse a pre-patterned silver layer into a 500nm thick As₂S₃ film to produce a channel waveguides. Photo-diffusion has also been applied to produce waveguides in As₂Se₃ [30,31] and GeS₂ [32]. The high index contrast (0.3-0.4) available by this method makes the production of compact waveguides feasible.

Whilst photo-darkening or silver in-diffusion are relatively simple processes for locally changing the refractive index of a ChG thin film, they are not necessarily desirable methods for producing waveguides which must be stable when exposed to high intensity light or operate at elevated temperatures. As a result it is preferable to create waveguide by physically structuring chalcogenide films for technologically-important application such as sensing, all-optical processing and mid-IR science by applying standard lithographic procedures involving either lift-off processes or etching.

In 1999 Viens et al. [33] reported the use of wet chemical etching to create structured waveguides in As-S(-Se) glasses. This appears to be the first attempt to use a photolithographic process although in this case the use of an NH₄OH-based developer resulted in isotropic etching and relative poor waveguide profiles. Nevertheless, losses around 1dB/cm at 1300nm were reported for 5µm wide waveguides made from 1.25µm thick As₂₄Se₃₈S₃₈ films etched to create a 0.4µm high rib and over-coated with an As₂S₃ cladding.

Subsequently two methods have become popular: lift-off techniques and those based on dry (plasma) etching. A problem, illustrated by the wet etching approach described above, is that many chalcogenides are soluble in alkaline photoresist developers. As a consequence care must be taken to prevent chemical attack of the ChG films that results in roughening of the film surfaces and enhanced optical losses.

The lift-off techniques circumvents this problem and involves the deposition of the ChG film onto a pre-patterned resist [34]. However side-wall roughness remains an issue and both graded index coatings and reflow methods have been used to reduce the waveguide losses [35,36]. Typically it is possible to reduce losses by around a factor of two using either of these methods, for example, the losses in 800nm x 400nm As₂S₃ nanowires were reduced from ≈6db/cm to ≈3.2 dB/cm at 1550nm by the reflow approach. Waveguides and ring resonators produced using the lift-off method have been successfully used as platforms for optical sensing [37,38].

To use dry-etching thin protective coatings such as SU8 or PMMA can be applied to the ChG film beneath the photo-resist to protect the surface of the ChG film from chemical attack by the photo-resist developer [39,40]. Both CF₄/O₂ and CHF₃ provide good selectivity during dry etching resulting in near vertical waveguide sidewalls. However, most ChG films deposited by thermal evaporation or pulsed laser deposition are composed of molecular clusters of different phases which may etch at different rates resulting in the production of a roughened, grainy etched surface [2,41]. This can be reduced by thermal annealing of the films prior to etching – which reduces phase separation in the films [41,42] – and by careful choice of plasma chemistry and process conditions [43]. By combining these approaches dispersion-engineered rib waveguides 2µm x 850nm have been produced in As₂S₃ with losses as low as 0.3dB/cm in TM mode. These have been successfully used for a wide variety of experiments on nonlinear optics and all-optical signal processing [44–47]. Dry etching with CF₄/O₂ plasma has also been applied to the production of waveguides for the mid-IR around 10.6µm in Te₂As₃Se₅ glass [48].

The relatively low glass transition temperatures characteristic of chalcogenide glasses makes it attractive to employ hot embossing to produce optical waveguides. One of the first reports of experiments on embossing was applied to As₂S₃ films using a hard stamp made from silica [49]. The embossing was carried out at a temperature of 225C at a pressure of 5 10⁶ Pa. Although submicron structures were successfully demonstrated significant problems were reported due to the appearance of crystals on the embossed surfaces. Later Lian et al. [50] reported the use of a silicon stamp to pattern an As₂S₃ film at 245C. In this case single mode waveguides were successfully demonstrated albeit with a rather high loss around 2.9dB/cm at 1550nm. Most recently it has been shown possible to use a soft PDMS stamp to emboss As₂₄Se₃₈S₃₈ glass films which have a rather low glass transition temperature of 120C [51]. Embossing was carried out at 190C and the use of a soft stamp made it possible to produce waveguides between 1.3µm and 3.3µm wide and 1µm high over a full 100mm wafer with exceptionally smooth side-walls. Waveguide losses for 4µm wide ribs were determined by the cut-back method to be 0.24dB/cm and 0.29dB/cm at 1550nm for TM and TE modes respectively. These results indicate that it is indeed possible to use hot embossing to produce quite low loss structures with small mode areas.

Whilst the low glass transition temperature (T_g) of As₂₄Se₃₈S₃₈ was an advantage for demonstrating the use of a soft stamp to mold rib waveguides, such a low T_g is not ideal from the device point of view. As a result we have now extended the technique showing that it is possible to use PDMS stamps at temperatures that are high enough (up to 300C) to emboss more robust glasses such as As₂S₃. 300C is well above the temperatures where crystallization or surface evaporation has been reported in As₂S₃ [50,43], however, we have found that these can largely be eliminated by applying a thin coating to stabilize the surface of the As₂S₃ film prior to embossing.

In the experiments reported here, therefore, we spun a 40nm thick coating of SU8 on top of an 850nm thick As₂S₃ layer that had been deposited using thermal evaporation onto an oxidized silicon wafer. The double chamber embossing tool described in [51] was used with a PDMS stamp replicated from an etched master. Embossing was carried out at a pressure of 2 atmospheres for 5 minutes at a temperature around 260C before the sample was flash cooled. In these conditions As₂S₃ rib waveguides were successfully produced over a full 100mm wafer. An optical micrograph showing the cross section of the typical rib is shown in Fig. 1(a) . It is apparent that some distortion of the profile has occurred compared with the results in As₂₄Se₃₈S₃₈ [51], and in particular the rib height increased from 850nm to ≈1µm and a significant slope is apparent on the sides of the waveguide. We performed cut-back measurements at 1550nm by coupling light in and out of the waveguide using lensed fibers with mode-field diameter of 2.5µm (Fig. 2 ). These indicated the waveguide losses were around 0.55dB/cm for a waveguide with nominal cross section 2.6µm x 0.85µm and were comparable with the values measured for similar structures produced by dry etching [40]. It is also worth noting that the wavelength dependence of the losses obtained from white light measurements was consistent with a λ⁻² scaling rather than the λ⁻⁴ reported in [51] which suggests the issues with Rayleigh scattering identified in the bulk of As₂₄Se₃₈S₃₈ did not occur in As₂S₃.

 

Fig. 1 (a) Optical micrograph of the cross-section of an As₂S₃ waveguide with a 40nm thick SU8 coating; 1(b) a micrograph without the coating present showing more vertical waveguide sidewalls; 1(c) a dark-field micrograph of the embossed surface in the presence of the SU8 coating; 1(d) a similar dark-field micrograph of the embossed surface without SU8 coating showing crystallization of the surface.

Download Full Size | PDF

 

Fig. 2 Insertion loss vs length for TE mode of embossed As₂S₃ waveguide nominally 2.6µm wide x 850nm high.

Download Full Size | PDF

Several factors could be responsible for the observed distortion of the rib. These include too high a viscosity in the As₂S₃ combined with the thin side-cladding inhibiting lateral flow of the glass. Alternatively the SU8 coating could lack the necessary elasticity to conform the molded surface. Since changing the embossing temperature had little effect on the rib shape, we decided to emboss a sample without the SU8 coating. A micrograph of the resulting rib waveguide is shown in Fig. 1(b) and less distortion is apparent with more vertical edges suggesting that the coating is responsible for much of the distortion in Fig. 1(a). However, the uncoated embossed surface is covered with micro-crystals (Fig. 1(d)) whereas the coated embossed surface was very clean (Fig. 1(c)). Hence the challenge ahead appears to select a more elastic or thinner surface coating which should reduce distortion whilst simultaneously suppressing crystallization. Nevertheless, these results indicate that embossing with a soft stamp is a viable approach for producing low loss chalcogenide waveguides and could be a low cost route for fabricating chips for applications such as sensing.

4. Waveguides and nanowires in Ge_11.5As₂₄Se_64.5 glass

Our research has shown that glasses from the intermediate phase of the Ge-As-Se system corresponding to compositions with mean coordination numbers (MCN) around 2.45-2.5 should be the best for waveguide fabrication (MCN = sum of the products of the valency times the abundance of the constituent atoms) [14,52]. We have demonstrated that films in this range deposited by thermal evaporation have physical properties (refractive index, band gap, density) indistinguishable from those of the bulk glass [14]. In contrast, films with higher (or lower) MCN have quite different properties from the bulk. A consequence is that thermal annealing of films in this region of the IP does not change their properties whereas outside this region large changes can occur. We thus believe Ge_11.5As₂₄Se_64.5 as a good material for waveguide fabrication. Three test structures have been evaluated: (1) dispersion engineered air-clad rib waveguides; (2) dispersion engineered nanowires intended for all-optical signal processing [2]; and (3) single mode all-chalcogenide buried channel waveguides intended for applications in the mid-IR around 3µm.

Dispersion-engineered nonlinear optical waveguides are an ideal platform for all-optical processing of data at extreme bit rates. By achieving a high nonlinear parameter, γ = ωn₂/cA_eff (where A_eff is the mode area and n₂ the instantaneous Kerr nonlinearity), device lengths can shrink into the centimeter range and this helps achieve operation over very wide bandwidths at moderate powers. Most functionalities required from all-optical processors can be achieved using either four wave mixing (FWM) or cross phase modulation (XPM) with the nonlinear phase shift required from the device Δφ = γPL_eff < 1, where P is the pulse power, and L_eff the effective length of the device which may be less than the physical length of the waveguide in the presence of losses or pulse walk-off. When γ≳10W⁻¹m⁻¹ all-optical processing of data streams can be performed at average powers of a few hundred mW. In the past five years nonlinear waveguide devices made from either silicon or chalcogenide glasses have proven effective for wavelength conversion, mid-span spectral inversion, parametric amplification, de-multiplexing, RF spectral analysis, regeneration and impairment monitoring. Whilst silicon has offered the highest nonlinear parameters (γ≈150W⁻¹m⁻¹) [53], it suffers from the disadvantage of significant two-photon absorption at telecommunication frequencies. In comparison two-photon absorption and free carrier effects are negligible in ChGs. In what follows we describe the properties of waveguide devices fabricated from Ge_11.5As₂₄Se_64.5 ChG that has been identified as having favourable properties for all-optical devices.

To optimize the bandwidth for all-optical processing small device dispersion is desirable. We have determined the group velocity dispersion (GVD) for Ge_11.5As₂₄Se_64.5 glass using an SCI Filmtek 4000 wafer mapper to be –650ps/nm/km at 1550nm. This needs to be reduced to near zero in the waveguide devices by employing dispersion engineering that offsets waveguide dispersion against the material dispersion. The group velocity dispersion for Ge_11.5As₂₄Se_64.5 rib waveguides 3 or 4 µm wide was calculated as a function of film thickness assuming a constant 50% edge depth for various cladding indices. Near zero or anomalous dispersion could only be obtained in the TM mode for air-clad ribs made from films ≈700-800nm thick whilst the TE mode dispersion remained large and normal. The larger material dispersion of Ge_11.5As₂₄Se_64.5 compared with As₂S₃ meant that zero dispersion designs with polymer or polysiloxane claddings could only be achieved using much thinner films (≈500nm thick). Thus, we first fabricated air-clad ribs from films 765nm thick with 50% etch depth for which the calculated GVD for waveguides 3µm and 4µm wide are shown in Fig. 3 . Zero dispersion occurs in these waveguides at 1512nm and 1525nm for 3µm and 4µm wide structures respectively for TM modes and is slightly anomalous at around 50-80 ps/nm/km at 1550nm. For TE modes the dispersion remains normal and is around –450 ps/nm/km.

 

Fig. 3 LHS: group velocity dispersion (GVD) for a 765nm x 4µm air-clad rib waveguide in Ge_11.5As₂₄Se_64.5 with 50% etch depth; RHS: GVD for 765nm x 3µm rib air clad rib waveguide with 50% etch depth.

Download Full Size | PDF

The Ge_11.5As₂₄Se_64.5 films were produced using thermal evaporation from a baffled box at a deposition rate of ≈1Å/s. The as-deposited films were annealed in a vacuum oven at 150C before being patterned using contact lithography and etched using CHF₃ plasma. Roughness of the etched surfaces was assessed using both and atomic force microscope and using a Veeco NT9100 optical profiler and was ≈1.5nm RMS. The resulting waveguides were coated with a ≈10-15nm of Al₂O₃ using atomic layer deposition. This layer was found to be effective in passivating the etched surfaces and, in particular, substantially increased the power handling capability of these waveguides. An SEM image showing the cross section of 765nm x 4µm rib waveguide is shown in Fig. 4 .

 

Fig. 4 SEM image of the cross section of 4µm wide Ge_11.5As₂₄Se_64.5 air clad rib waveguide.

Download Full Size | PDF

Optical losses were measured by the cut-back method and ranged were 0.75dB/cm and 0.65dB/cm for TM and TE modes respectively for 4µm wide ribs rising to 0.82 and 0.76dB/cm for the 3µm wide ribs. The nonlinear parameters for TM mode were calculated to be 21W⁻¹m⁻¹ and 30W⁻¹m⁻¹ for the 4µm and 3µm wide ribs respectively based on the material nonlinearity, n₂ = 8.6x10⁻¹⁴cm²/W obtained from Z-scan measurements [52].

The nonlinear properties of these rib waveguides were assessed through a series of measurement on self-phase modulation and power-dependent transmission of 2ps duration pulses coupled in to the rib waveguides from a mode-locked fiber laser. The output spectra were recorded on an optical spectrum analyzer and the power dependence of the transmission was compared with the result of simulations based on the split-step Fourier method. This allowed the nonlinear figure of merit (FOM = n₂/βλ) to be determined as ≈60 for this material. Based on the measured value of n₂ for bulk glass the value of the two-photon absorption coefficient was, therefore, β≈9x10⁻¹²cm/W.

We used these rib waveguides to assess continuous wave (CW) power handling of this glass. For these experiments, a narrowband CW source was amplified using an erbium-doped fiber amplifier (EDFA) to a maximum power of around 1W. This was coupled into the waveguide using a tapered lensed fiber with a mode-field diameter of ≈2.5µm. After taking coupling and reflection losses into account around 250mW of average power could be coupled into the ribs. Initially waveguides without Al₂O₃ coatings were tested and in this instance optical damage was found to occur erratically at average powers around 30-50mW in the waveguide. The morphology of the damage had the nature of a “fiber fuse” that is observed to occur in silica optical fibers at high power loadings when a damage track originates at a defect deep within in the fiber and causes a thermal wave to travel from this site back towards the fiber input destroying the glass in its track. In the case of Ge_11.5As₂₄Se_64.5 the same phenomenon was observed with damage starting up to a few centimeters into the waveguide and melting the waveguide core along a “snaking” path back to the input facet. A typical damage track is shown in Fig. 5 . The application of the thin Al₂O₃ coating dramatically changed the threshold for damage. After coating the waveguides could withstand the full coupled power without incurring damage and this corresponded to an average power density in around 20MW/cm².

 

Fig. 5 “fiber fuse” damage in a 3µm wide Ge_11.5As₂₄Se_64.5 rib waveguide without Al₂O₃ coating exposed to CW power around 40mW. The track extends over a distance of about 1cm in this case and appears to originate at a defect in the waveguide. The molten damage snakes along the waveguide back to the entrance surface.

Download Full Size | PDF

Whilst these Ge_11.5As₂₄Se_54.5 ribs provide improved nonlinearity over similar structures fabricated from As₂S₃, air cladding leaves them vulnerable to mechanical damage during handling or to deterioration due to contamination of the waveguide surfaces. As a result buried channel waveguide are preferable. To achieve zero or anomalous dispersion using a low index cladding, such as the Inorganic Polymer Glass (IPG) material with refractive index ≈1.51 that we commonly to clad As₂S₃ waveguides, requires the film thickness to be reduced to around 500nm. We recently reported experiments on Ge_11.5As₂₄Se_64.5 nanowires 630nm x 500nm fabricated by electron beam lithography using a PMMA resist and etched using CHF₃ plasma [54]. A record nonlinear parameter for a glass waveguide of 136W⁻¹m⁻¹ was predicted for these structures and this was confirmed via measurements of the conversion efficiency to the idler in experiments on continuous wave (CW) FWM. In these first devices the optical losses were moderate at between 2.6db/cm and 3.2dB/cm for TM and TE modes respectively at 1550nm. Through process improvements this has now been reduced to ≈1.5dB/cm for a 550nm x 500nm nanowire with γ≈150W⁻¹m⁻¹. A cross section of these most recent nanowires is shown in Fig. 6 .

 

Fig. 6 An SEM image showing the cross section of an IPG coated of 500nm x550nm Ge_11.5As₂₄Se_64.5 nanowire with loss of 1.5dB/cm (TM mode)

Download Full Size | PDF

Whilst the devices described above have all been designed to operate in the telecommunications bands around 1550nm, operation in the mid-IR requires the cladding as well as the core materials to be transparent at long wavelengths. This rules out the use of oxides or most polymers as bottom and top claddings. To overcome this issue we have fabricated our first buried channel waveguides in an all-chalcogenide system based on a 1.7µm thick Ge_11.5As₂₄Se_64.5 core evaporated on top of a 2µm thick As₃₆S₆₄ lower cladding whose refractive index is 2.37 at 1550nm thus providing an index contrast of ≈0.3. After patterning the core layer was fully-etched using CHF₃ plasma and the core over-coated with an additional layer of As₃₆S₆₄ either 2-3µm thick. A challenge when cladding a structured waveguide using thermal evaporation is to eliminate voids at the corners of the etched core which can appear due to masking effects. To achieve good corner filling, the upper cladding was, therefore, evaporated onto the patterned substrate as it was rotated around an axis tilted at an angle of 57° relative to the direct line from the wafer axis to the thermal source. This, in principle, leads to the coating thickness being the same on both the horizontal and vertical surfaces. As is evident from the optical micrographs in Fig. 7 , good corner filling was achieved with no evidence of voids visible in these optical micrographs although as the cladding gets thicker masking effects eventually appear which creates “notches” at the upper cladding boundary. However calculations of the mode field show that for an index contrast of 0.3, the field does not extend to the “notches” even for a wavelength of 3.5µm.

 

Fig. 7 Optical micrographs of buried channel waveguides using a Ge_11.5As₂₄Se_64.5 core surrounded by a As₃₆S₆₄ cladding. In the image on the LHS the core dimensions are ≈1.7µm x 1.7µm whilst on the RHS the core is 3.7µm x 1.7µm. No evidence the existence of voids near the corners of the etched core are visible. The notches on the top interface between the As₃₆S₆₄ and air is caused by masking effects during the deposition of the cladding layer.

Download Full Size | PDF

The nonlinear parameter at 1550nm for these buried channel waveguides was calculated to be ≈16W⁻¹m⁻¹ at 1550nm for the 1.7µm x 1.7µm square core rib (Fig. 7: LHS) although the dispersion is large and normal at around –500ps/nm/km. This dispersion in this structure should be polarization independent. The optical losses were measured to be 0.2dB/cm for TM mode but 0.45dB/cm in TE mode at 1550nm. This significant polarization dependent loss suggests the roughness at the vertical etched interfaces was significant since the TE mode is more sensitive to roughness on the vertical interfaces. The 1.7µm x 1.7µm waveguides become single mode beyond about 3.5µm. At present a suitable source at 3.5µm has not been available to determine performance in this long wavelength region.

5. Conclusions

Over the past decade there has been considerable progress to develop low loss optical waveguides in chalcogenide glasses for applications in optical sensing and all-optical processing. Both of these benefit from the creation of very small waveguides. For sensing small waveguides push the evanescent field outside the waveguide so that it interacts effectively with the environment, whereas for all-optical processing small waveguides enhance the intensity and thereby the nonlinear response of the device. To date the smallest nanowires produced in ChGs have mode areas around 0.20µm² at 1550nm and have losses around 1.5dB/cm. Fabricated from highly nonlinear Ge_11.5As₂₄Se_64.5 glass, leads to a nonlinear parameter of 150W⁻¹m⁻¹ – comparable with that of a silicon nanowire but with the advantage of negligible nonlinear absorption and no free carrier effects. This is expected to be an excellent platform for all-optical processing.

By designing the structures to take advantage of dispersion-engineering these nanowires display anomalous dispersion in the telecommunications bands and can produce broadband supercontinuum at record low powers in very short devices (<2cm) [54]. So far the supercontinuum spectra that have been reported span an octave around the pump wavelength at 1550nm, but clearly the high nonlinearity and good transmission in the mid-IR suggest that supercontinuum can be also generated at much longer wavelengths and this is now the subject of active research [3].

For larger waveguides the optical losses can be much lower: for example we have reported As₂S₃ rib waveguides with mode areas ≈5.7µm² and optical losses around 0.05dB/cm at 1550nm [55]. These are comparable with the lowest losses achieved in silica planar waveguides and indicate that complex circuits requiring long path lengths are also feasible.

Whilst there has been a lot of work on waveguides for the telecommunications band, there are far fewer reports on devices operating at longer wavelengths that take advantage of the intrinsic Mid-IR transparency of ChGs. However, there is a strong motivation to fabricate waveguide operating in the molecular fingerprint region of the optical spectrum between 2 and 20µm both for sensing illicit or dangerous materials via their spectroscopic signatures and for applications such as mid-IR astronomy. The commercialization of quantum cascade lasers, has provided the sources which can be used with ChG waveguide chips making many sensing applications feasible. Some reports of the production of waveguides for the mid-IR have appeared [48,56], and losses as low as 0.5dB/cm at 8.4µm have been reported [56].

Producing low loss waveguides operating over wide ranges in the mid-IR raises a number of technical challenges. Firstly, impurity absorption in the glass films must be carefully controlled to preserve optical transparency. This will require careful attention to the purity of the starting materials used to make films and to the processes used to make the devices. To obtain transmission at the longest wavelengths, telluride glasses will be needed but these are generally far less chemically stable than the selenides and sulphides and are often sensitive, for example, to water vapour. A possible solution is to move to the chalco-halide systems that have been shown to have greater stability whilst maintaining long wavelength transparency [57].

To conclude, there are exciting opportunities to exploit the novel properties of chalcogenides in optical waveguide devices. Recent results are very encouraging showing it is possible to obtain low loss or highly nonlinear waveguide devices tailored for important applications. Many exciting developments can be anticipated as these structured are applied to create new opportunities in infrared science.