1. Introduction

An integrated-optic (IO) geometry for active waveguide devices possesses several desirable features. It allows active components to be combined with passive components on a single substrate, resulting in a reduction in device size and potentially cost [1, 2]. Furthermore, higher rare earth (RE) dopant concentrations may be possible in IO waveguides than in optical fibers. In some glasses the presence of RE ions may increase the tendency towards crystallization during reheating for the fiber draw process. In IO waveguides, reheating is not needed, and any RE dopant concentration is possible in principle as long as the composition lies within the glass-forming region. As a result, active IO devices have the potential for high gain per unit length.

The chalcogenide glasses (ChG’s) are good candidates for use in RE doped amplifiers and lasers operating at wavelengths from the telecommunications bands through the long-wave IR. They exhibit low phonon energies, resulting in wide IR transmission windows and permitting radiative transitions that are not possible in silica and phosphate glasses due to multiphonon quenching [3]. ChG’s have been used previously in the fabrication of fiber lasers [4] as well as in optically written IO lasers [5]. They have not, however, been previously used in erbium-doped waveguide amplifiers (EDWA’s).

Gallium lanthanum sulfide (GLS) is a ChG that may be particularly well-suited for use in EDWA’s [6, 7, 8, 9, 4, 10]. It has a wide IR transmission window, with transmission of 50% or higher through a 1 mm thickness over a wavelength range of 0.5 to 10μm [7]. Its high glass transition temperature, T _g = 580° C, makes GLS appropriate for use in high-temperature applications. GLS exhibits good environmental stability; it has been shown that the attenuation in uncoated GLS fiber increases only slightly after several years of storage [10]. Additionally, high RE dopant concentrations without clustering are possible owing to the fact that the RE ions substitute for the lanthanum ion in the glass matrix [8]. The potential of GLS as a laser material has been demonstrated in Nd³⁺-doped bulk glass lasers [9] and Nd³⁺-doped fiber lasers [4]. The spectral properties of Er³⁺-doped GLS have been well-characterized [11].

Previous efforts to fabricate IO waveguides in GLS glass have utilized the fact that the glass exhibits a photoinduced refractive index change when exposed to above bandgap light. Waveguides written in polished bulk samples with propagation losses of < 0.5 dB/cm have been demonstrated [5]. Patterning waveguides in a thin film of sputtered GLS, however, possesses several advantages over this method. A sputtered film can be made with high compositional uniformity and with fewer of the local defects that are sometimes present in bulk glass. Furthermore, deposition results in improved design flexibility; sputtered films can be prepared on a variety of substrates, so integration with other IO devices is more straightforward. This method has the additional benefit that glass is deposited only were needed, potentially reducing cost.

In this paper we describe the fabrication process for EDWA’s in sputtered films of GLS and discuss their gain characteristics. The waveguides are patterned by use of the lift-off technique and standard photolithography. Our initial results show an internal gain of 2.8 dB/cm for a 2.4cm long sample — a total internal gain of 6.7 dB. This research demonstrates the feasibility of fabricating active IO devices in sputtered films of GLS. The same fabrication techniques could be applied with other rare earth dopants to create mid-IR sources, taking advantage of radiative transitions that are quenched in other hosts.

 

Fig. 1. Schematic diagram of the lift-off fabrication process showing a cross section of two waveguides.

Download Full Size | PDF

2. Experiment

Both slab waveguides and strip waveguides were fabricated by sputtering. Several 2 in. (5.08 cm) diameter fused silica substrates were coated with a 60 Å layer of Cr via sputtering to act as an adhesion layer. On the strip waveguide samples, a thick (3.7μm) polyimide lift-off layer was spun onto the substrate followed by a 1.1μm thick layer of positive photoresist. The positive photoresist was patterned with straight lines 10μm in width by exposing it through a photomask. The photoresist and lift-off layer were developed, and the underlying lift-off layer acquired an undercut profile owing to its higher (relative to the positive photoresist’s) etch rate.

GLS deposition was performed as described in Ref. [12]. A bulk glass ingot with a molar composition of 70Ga₂S₃:23La₂S₃:6La₂O₃:1Er₂S₃, from ChG Southampton Ltd., was ground into a powder, and a sputter target was formed by hot pressing this powder into a 3 in. (7.62 cm) diameter disk. GLS was deposited by RF magnetron sputtering in a sputter-up geometry onto the photoresist-coated and uncoated substrates. This process was carried out in an Ar atmosphere with a base pressure of 5 mT and a flow rate of 20sccm. An energy density of approximately 1 W/cm² was used. The substrate temperature was maintained at 100° C. The resulting deposition rate was 11 Å/min. A schematic diagram of the fabrication process for strip waveguides is shown in Fig. 1.

A strip waveguide sample was cleaved perpendicular to the waveguides on either end, resulting in 2.4 cm long devices. The lift-off layer and photoresist were left in place while measuring the amplifier’s properties. A schematic diagram of the amplification measurement setup is shown in Fig. 2. The outputs of a fiber-coupled 1480nm wavelength diode laser and a fiber-coupled 1550nm wavelength diode laser were combined by use of a multiplexer. The 1480nm laser acted as the pump, and the 1550nm laser, modulated with a square wave, acted as the signal. Light from the output of the multiplexer was coupled into the input of the waveguide with a silica fiber, and light at the output of the waveguide was collected with a lens and focused onto an InGaAs photodiode. Filters were used to isolate the 1550nm signal. The gain was measured with an oscilloscope.

3. Results

Several measurements were made on slab waveguide samples. The refractive index was measured with an ellipsometer and found to be 2.5 at a wavelength of 633 nm, the same as that of bulk GLS. The crystallinity of the film was evaluated with XRD, and it was found to be glassy, exhibiting no sharp diffraction peaks. The composition of the film was evaluated with EDS (neglecting oxygen which did not have measurable EDS peaks) and found to be Ga_29.7La_10.6S_59.2Er_0.5 — only slightly different from the melt composition of Ga₂₉La₁₂S_58.5Er_0.5. As described in Ref. [12], the sputtered film composition does not vary with thickness. RMS surface roughness, measured with a stylus profilometer on a slab waveguide, was found to be approximately 8 Å, sufficiently small to provide low scattering losses.

 

Fig. 2. Schematic diagram of the amplification measurement setup.

Download Full Size | PDF

 

Fig. 3. SEM image of (a) a cross section and (b) a view with a 10° inclination of a GLS amplifier. This waveguide was patterned with a 10μm mask opening.

Download Full Size | PDF

SEM images of a cleaved strip waveguide sample are shown in Fig. 3. The undercut lift-off layer and the overhanging positive photoresist can be seen in the images. The positive photoresist is obscured slightly in the first image by debris from the lift-off layer that were created when the sample was cleaved. During deposition, GLS filled the channel creating a strip waveguide that is approximately 11μm wide at its base and 2.2μm in height. The top surface of the waveguide is curved slightly because its edges were partially shielded by the overhanging positive PR during deposition.

Spontaneous emission was measured with both a 980nm and a 1480nm wavelength pump. Light was coupled into the waveguide with a single mode silica fiber. Green luminescence resulting from upconversion is visible along the entire length of the waveguide. Light at the waveguide’s output was collected with another single mode silica fiber and analyzed with an OSA. The magnitude of the spontaneous emission as a function of wavelength is shown in Fig. 4. When pumped with a 1480nm source, the peak emission occurred at a wavelength of 1537nm with a smaller local maximum at a wavelength of 1547nm. When pumped with a 980 nm source, the peak emission occurred at a wavelength of 1551nm with a smaller local maximum at a wavelength of 1538 nm. The difference between the two spectra may be the result of the difference in absorption cross section for the two wavelengths. The absorption at 980 nm is greater than it is at 1480 nm, so the 980nm pump is absorbed, on average, closer to the input of the waveguide. While the emission cross section should be approximately the same for either pump, the emitted light is subject to a longer effective path length in the case of a 980 nm pump. An absorption peak is present in Er³⁺-doped GLS at 1538 nm, so attenuation caused by reabsorption near 1538nm is greater when a 980nm pump is used. The fluorescence lifetime was also measured with each pump source and found to be 1.1± 0.1ms in both cases.

 

Fig. 4. Spontaneous emission from a waveguide when pumped with 1480 nm and 980 nm wavelength light.

Download Full Size | PDF

 

Fig. 5. A grayscale image (a) and a contour map (b) of the output at a 1.5μm wavelength of a waveguide pumped with 1480 nm light. Contours are shown from 0.2 to 0.9 of the maximum irradiance at intervals of 0.1.

Download Full Size | PDF

A spatial profile of the spontaneous emission of a waveguide was obtained by pumping it at a 1480nm wavelength and imaging the output face onto a vidicon camera. A filter was placed in front of the camera to isolate the 1.5μm light. The resulting image and a corresponding contour map are shown in Fig. 5. The dimensions of the profile — approximately 7 μm wide by 3μm high — are comparable to those of the waveguide. As can be seen in the figure, some light is present in the substrate, possibly due to the presence of leaky modes.

Finite difference modeling using semi-vector solutions on a uniform grid, as derived in Ref. [13], was applied to estimate the number of modes supported by this waveguide. The simulation predicts that the waveguide supports approximately 100 modes for both TE and TM polarized light. The image of the amplifier’s output shown in Fig. 5 is most likely the combination of the light from a large number of waveguide modes.

 

Fig. 6. Internal gain as a function of pump power.

Download Full Size | PDF

A plot of the internal gain as a function of pump power is shown in Fig. 6. Internal gain is defined here as 10log(P _g/P ₀) where P _g is the peak power of the square wave at the detector in the presence of the pump, and P ₀ is the peak power of the square wave with no pump present. The maximum internal gain of 6.7dB was observed with a pump power of 180mW. This value corresponds to an internal gain per unit length of 2.8 dB/cm. The amplification was measured with the lift-off layer and photoresist layer still in place. It was found that when these layers were removed propagation losses increase considerably. This effect may be due to increased surface roughness in the area in which the lift-off layer was in contact with the waveguide, resulting in increased scattering.

Propagation loss in the waveguide was estimated by use of the scattered-light method. Light from a 1650nm wavelength source was coupled into the waveguide. The source was chosen to be near the design wavelength but away from the absorption peaks of Er ³⁺. A lens collected the scattered light from a small area of the waveguide and focused it onto a detector. The lens and detector were scanned together along an axis parallel to the waveguide. It is assumed that the scattered light at any point along the waveguide is proportional to the power of the guided light that is present at that point. It is thus possible to determine the power as a function of z and obtain an estimate of the propagation loss. The measured loss was 2.4± 0.4dB/cm.

The internal gain approximately balances the scattering loss, but in order to realize net gain, the internal gain must be greater than the sum of the scattering loss and the absorption due to the presence of Er³⁺ ions. Measurement on a bulk sample yielded an absorbance of 0.27 at λ = 1550nm corresponding to an additional loss of ~5.4 dB/cm. Therefore, an increase of this magnitude in the internal gain is required before net gain is realized.

4. Conclusion

Sputtered films of Er³⁺ doped GLS show promise for active IO applications such as lasers and optical amplifiers. The material’s transparency in the mid-IR, high T _g, and the potential for high RE doping concentrations make it an attractive candidate for such applications. RF magnetron sputtering in combination with the lift-off technique has been shown to be a viable method of fabricating EDWA’s in this material. In future work, we plan to fabricate lasers operating at a wavelength of 1540nm in Er³⁺-doped GLS. We also plan to fabricate mid-IR IO sources, taking advantage of active transitions in this host that are quenched in silica.

Several improvements may result in lower-loss waveguides and thus lead to better performance amplifiers and lasers. While in this case the lift-off technique yielded reasonably good-quality waveguides,it may be possible to achieve lower loss waveguides by sputtering a uniform film and performing a dry etch. To date this approach has not been successful because the sputtered films are resistant to dry etching, but we are attempting to develop an effective process. The deposition of a cladding layer should significantly decrease the waveguide losses and may also enable the fabrication of single-mode devices with mode profiles that are well-matched to that of a single-mode optical fiber. The amplifiers described here do not produce enough gain to overcome scattering loss and the loss due to Er³⁺ absorption in the waveguide. Process refinements that increase the internal gain should result in amplifiers with net gain.

We have recently achieved a deposition rate for GLS of over 100 Å/min., significantly higher than the rate used in the fabrication of the amplifiers reported here. The higher rate will result in much shorter deposition times and make sputtering GLS more viable for use in commercial processes.