1. Introduction

The ferroelectric crystal lithium niobate (LiNbO₃) is a well-established material for a wide variety of photonic applications due to its excellent electro-optical, acousto-optical and nonlinear properties. Another benefit of LiNbO₃ is easy implementation of low-loss waveguides, the basic building blocks for integrated optics devices, such as electro-optic and acousto-optic modulators, nonlinear frequency converters and waveguide amplifiers as well as waveguide lasers [1]. In particular, there is a growing interest in the development of Er³⁺-doped waveguide lasers due to emission in the third telecommunication window around 1.55 µm [2-3]. Moreover, besides simple continuous wave fixed-wavelength free-running lasers, the above-mentioned properties of LiNbO₃ allow for the development of mode-locked pulsed and tunable waveguide lasers [4–7] as well as integration of a variety of nonlinear, electro- and acousto-optic devices on the same substrate.

In the last decade, ridge waveguides in LiNbO₃ have become of great importance for integrated optics due to their stronger light confinement. The smaller mode field sizes in ridge waveguides, along with the notably increased mode overlap, help to improve the performance of nonlinear, active and electro-optic photonic devices [8–11]. The first developed wet- or plasma-etched LiNbO₃ ridge waveguides showed rather high propagation losses of ≥ 1 dB/cm. In 2007 Hu et al. reported losses of 0.3 dB/cm in wet-etched titanium (Ti) in-diffused LiNbO₃ ridges [12] and, three years later, 0.05 dB/cm in similar waveguides with their sidewalls smoothened due to the inverse order of etch and Ti diffusion steps [13]. Meanwhile, over the last 14 years, precision machining employing a diamond blade (optical grade dicing) has been established as an alternative technique for low-loss ridge waveguide fabrication [14–16]. A main advantage of optical grade dicing is that it does not require photolithographic or etching steps. In 2014 Gerthoffer et al. reported an optimized process for fabrication of Ti:LiNbO₃ ridge waveguides with propagation losses of < 0.2 dB/cm due to waveguide sidewall smoothening during high temperature Ti diffusion [17].

In this paper, we report on the development of efficient erbium (Er)-doped Ti:LiNbO₃ ridge waveguide amplifiers and lasers. To improve both amplifier and laser performance and taking advantage of ridge geometry, an advanced fabrication technique of three-side Er/Ti deposition and diffusion was used allowing for higher Er concentrations and better overlap of Er diffusion profiles with the guided modes [18]. Internal gain of 14 dB at 1531 nm has been obtained in 4.6 cm-long waveguides for a coupled pump power of 200 mW at 1486 nm, which corresponds to 3 dB/cm signal amplification. For different reflectivity settings of the waveguide resonator ends, lasing at 1531 nm with a slope efficiency of 19% and at 1561 nm with 33% and an optimized output power of 50 mW was demonstrated in a 6.8 cm-long sample.

2. Fabrication and small-signal amplification in Er:Ti:LiNbO₃ ridge waveguides

2.1 Fabrication

The fabrication process of Er:Ti:LiNbO₃ ridge waveguides is based on a novel technique comprised of ridge definition using diamond blade dicing followed by three-side Er and Ti layers deposition and high-temperature in-diffusion. A detailed description of the fabrication steps can be found in [18]. The technical parameters of the best performing sample in the cited work were used. Ridges with widths ranging from 5 µm to 15 µm were cut into a congruent x-cut LiNbO₃ substrate along the y-axis of the crystal using a precision wafer saw. Two 11 nm-thick Er layers were evaporated at the angles ± 60 ° with respect to the substrate normal and in-diffused for 100 hours at 1120 °C. For waveguiding, 110 nm-thick Ti layers were deposited on top of the ridges and the sidewalls at ± 60 ° and in-diffusion for 3.7 hours at 1100 °C. Smooth end facets were obtained by dicing perpendicular to the y-axis.

2.2 Characterization of small-signal amplification

In the recent study of Er:Ti:LiNbO₃ ridge waveguide amplifiers, a single-pass small-signal internal gain up to 2.7 dB/cm at 1531 nm was measured in 2.5 cm-long samples for the TM modes (σ-polarization) in y-propagation waveguides under 200 mW of coupled pump power at 1486 nm [18]. The conducted simulations showed that there is a nearly linear dependence between the waveguide length and the optical gain for waveguide lengths up to 10 cm, making it possible to achieve more than 25 dB of single-pass small-signal amplification at 1531 nm.

To study the amplification performance in longer ridge waveguides, the single-pass small-signal gain and absorption in a 4.6 cm-long sample with an anti-reflection (AR) coated facet was measured. A schematic of the setup for the absorption/gain measurement is presented in Fig. 1(a). As a small-signal source, we used a fiber-coupled Er-doped fiber amplifier (EDFA) with a broad continuous spectrum covering the wavelength range from 1510 nm to 1630 nm. Using an attenuator, the EDFA signal was reduced to the −20 dBm level. As a pump source, a fiber-pigtailed 1486 nm laser diode with a maximum output power of 300 mW was used. To combine the pump and the signal we used a single-mode 1480 nm/1550 nm wavelength division multiplexer (WDM). The polarizations of the pump and the signal were adjusted by means of paddle polarization controllers. Due to higher Er³⁺ absorption at 1480 nm and emission around 1550 nm for σ-polarized light [19-20], the pump and signal waves were chosen to be σ-polarized, which for a x-cut y-propagation LiNbO₃ substrate corresponds to TM polarization. A combination of a fiber-pigtailed gradient index (GRIN) and an AR coated focusing lens was used to couple light into the waveguide, which avoids etalon effects that would otherwise arise during fiber-to-waveguide coupling between the cleaved fiber end and the sample facet. At the output facet, the amplified signal and the residual pump light were collected and separated by a similar combination of optical elements. For waveguide mode imaging, an additional beam splitter (BS) was used to reflect 30% of the output onto an infrared camera. The separated amplified signal light was coupled to an optical spectrum analyzer (OSA). For the optimization of the throughput signal, we utilized a 3-axis piezo positioning system with an auto-align option. A transmitted spectrum of the EDFA through the setup without the sample served as a reference. The internal small-signal absorption/gain was then determined from the ratio of the recorded transmitted spectrum to this reference spectrum with taking into account coupling and Fresnel reflection losses at the waveguide facets [21]. The evaluated coupling efficiencies into the ridge waveguides, based on the recorded mode intensity profiles, depended on the waveguide width and varied from 80% to 92%.

 

Fig. 1 (a) Schematic of the setup for the single-pass small-signal optical absorption/gain measurements. (b) Small-signal absorption/gain spectra at various coupled pump power levels measured for a 6 µm-wide ridge waveguide.

Download Full Size | PDF

Because of trade-off between Er concentration and scattering losses in the fabricated ridge waveguides, the best optical amplification, as well as lasing performance, is expected in the narrower waveguides with about 6-7 µm widths [18]. The results of the small-signal absorption/gain measurements in a 6 µm-wide ridge waveguide at various coupled pump power levels are presented in Fig. 1(b). Owing to the above-mentioned specific of the Er³⁺ absorption and emission spectra, the best amplifier performance was achieved for both signal and pump being σ-polarized. In this waveguide, as can be seen from the absorption measurement (black line), the high overlap between the guided mode and the Er diffusion profile results in a signal attenuation of −40 dB. By increasing the coupled pump power to 42 mW, we observed internal transparency (0 dB) at 1531 nm. At the maximum available coupled pump power of 200 mW, we measured the highest internal gain of 14.0, 6.0 and 5.9 dB at 1531, 1547 and 1562 nm wavelengths, respectively. The realized internal gain of 14.0 dB at 1531 nm corresponding to 3.0 dB/cm, exceeds the best literature values reported for Er:Ti:LiNbO₃ waveguides and is in good agreement with our simulated value of 14.8 dB for a 4.6 cm-long sample.

3. Laser characterization in Er:Ti:LiNbO₃ ridge waveguides

3.1 Laser setup

To make best use of the high internal gain in Er:Ti:LiNbO₃ ridge waveguides for efficient lasing around 1.55 µm wavelength, a 6.8 cm-long sample was chosen [2]. For the experimental investigation of lasing at 1531 nm and 1561 nm, we have used the two different laser setups shown schematically in Fig. 2(a) and 2(b).

 

Fig. 2 (a) Schematic diagram of the experimental setup of the Er:Ti:LiNbO₃ ridge waveguide laser operating at 1531 nm. (b) Setup of the ridge waveguide laser operating at 1561 nm.

Download Full Size | PDF

For lasing at 1531 nm, the ridge waveguides were pumped from each side by fiber-pigtailed 1486 nm laser diodes with maximum output power of 300 mW, see Fig. 2(a). To couple the pump light into the waveguide, we used the same combination of GRIN and AR-coated focusing lenses. The polarization of the pump light was again adjusted with paddle polarization controllers. At both waveguide end facets, the laser light was separated from the residual pump light by means of a WDM and the laser output power was measured with a power meter. To obtain the output power behind the focusing lenses, the measured power was corrected for the insertion losses of the WDMs. We measured the laser output for different output coupler reflectivities. In the first configuration, we achieved lasing with feedback provided solely by Fresnel reflection (14%) at the waveguide end facets. For the realization of variable output coupling, either a LiNbO₃ or a silicon (Si) plate was inserted between one lens and the sample, see Fig. 2(a). Between this plate and the end facet of the Er:Ti:LiNbO₃ waveguide a small, several micrometer wide air gap was left, forming a low-finesse Fabry-Perot resonator. By varying this air gap, the effective reflectivity R_eff at each particular wavelength can be adjusted within the range of 0% < R_eff < 44% and 5% < R_eff < 59% with the inserted LiNbO₃ and Si plate, respectively. To obtain the desired effective reflectivity for the laser wavelength and simultaneously the highest coupled pump power, the air gap width was fine-tuned with a piezo-drive within about 4 µm - 13 µm range.

For lasing at 1561 nm, we chose the setup schematically shown in Fig. 2(b). As before, a fiber-pigtailed 1486 nm laser diode with a maximum output power of 300 mW was used as a pump source. Using a high reflective (HR) mirror for both pump and laser wavelengths at the rear end facet of the sample, the coupled pump light was reflected and therefore passed the laser resonator twice. Again, we studied the laser performance for different effective reflectivities at the sample’s front facet. In the first configuration, lasing was obtained with feedback provided solely by Fresnel reflection. For the realization of variable reflectivity at pump and laser wavelengths, a LiNbO₃ plate was inserted between the lens and the front waveguide facet.

3.2 Results and discussion

For the reason, mentioned in Subsection 2.2, the best lasing performance was obtained in a 6 µm-wide ridge. The coupling efficiency into the waveguide for the pump wave, evaluated based on the recorded mode intensity profiles, was 85%, independent of the laser configuration.

In Fig. 3(a), the laser output power at 1531 nm is plotted versus the coupled pump power for the laser configuration with an inserted Si plate. The air gap was tuned to 12.6 µm to obtain the highest effective reflectivity of 59% at 1531 nm and the lowest one of 5% for the pump light. Here and in the following, the output power is the sum of the outputs behind the two focusing lenses. The lasing threshold was reached at a pump power of 117 mW. For the maximum coupled pump of 316 mW we measured an output power of 37.5 mW with a slope efficiency of 19%, which is comparable to the previously reported values, see Table 1. For the configurations with the feedback provided solely by Fresnel reflection at the waveguide end facets and with an inserted LiNbO₃ plate, which are not shown for the sake of clarity in Fig. 3(a), we measured efficiencies of 17% and 18%, respectively. Since the slope efficiency kept increasing with the reflectivity of the input coupler, a better performance can be expected for even higher input coupler reflectivity. Furthermore, our calculations show that reducing scattering losses from the current 0.4 dB/cm to 0.25 dB/cm would increase the achievable slope efficiency up to 30%.

 

Fig. 3 Laser performance of the Er:Ti:LiNbO₃ ridge waveguide of 6 µm width: (a) Laser output power at 1531 nm versus coupled pump power for the laser configuration with an inserted Si plate, see setup in Fig. 2(a). Inset: The laser output power, measured over a 30 min time span. (b) Laser output power of the ridge waveguide laser at 1561 nm versus coupled pump power for front facet reflectivities R_OC of 14% (red) and 30% (blue).

Download Full Size | PDF

Table 1. Performance of previously reported Er:LiNbO₃ channel waveguide laser. The abbreviations (l) and (i) stand for launched and incident pump power, respectively.

View Table

The measured output power at 1561 nm versus the coupled pump power for the reflectivities 14% and 30% at the front sample facet is shown in Fig. 3(b). For the Fresnel reflection feedback of 14% for both pump and laser wavelengths, laser oscillation sets in at 45 mW. A slope efficiency of 27% and a maximum laser power of 40 mW were measured. For the laser configuration with an inserted LiNbO₃ plate, the best laser performance was obtained for an air gap width of about 4.5 µm resulting in effective reflectivities of 30% and 0% at 1561 nm and 1486 nm, respectively. With a lasing threshold of 41 mW, a maximum laser output power of 50 mW with a slope efficiency of 33% was achieved for 200 mW of coupled pump power. This is the highest slope efficiency ever reported, to the best of our knowledge, for an Er:Ti:LiNbO₃ waveguide laser. The laser emission has exhibited clear multi-longitudinal-mode behavior with the spectrum envelope having a FWHM width of about 0.3 nm and 0.5 nm at 1531 nm and 1561 nm, respectively. With OSA resolution comparable to the laser modes spacing, the individual longitudinal modes were only marginally resolved with an estimated free spectral range of about 8 pm which agrees well with the theoretically expected value. The laser output power, measured over a 30 minutes time span, was found to be stable, see inset of Fig. 3(a). The absence of photorefractive damage effects, despite high infrared light intensities and significant amount of green up-converted emission, can be explained by the improved photorefractive damage resistance of ridge waveguides compared to their channel counterparts [23].

4. Summary

In conclusion, we have demonstrated Er:Ti:LiNbO₃ ridge waveguide amplifiers and lasers fabricated in congruent x-cut substrates. We used a novel fabrication technique of three-side evaporation and in-diffusion for Er and Ti incorporation into pre-defined ridges, which improves the amplifier performance by allowing for higher Er concentrations and better overlap of Er diffusion profiles with the guided modes. A total single-pass small-signal internal gain up to 14 dB at 1531 nm in a 4.6 cm-long sample for the TM modes in y-propagation waveguides under 200 mW of coupled pump power at 1486 nm was measured. Furthermore, we demonstrated efficient Er:Ti:LiNbO₃ ridge waveguide lasers operating at 1531 nm and 1561 nm with slope efficiencies of 19% and 33%, respectively. Our future studies will be focused on the optimization of the amplifier / laser performance by means of erbium/ytterbium co-doping [24].