Spectral study of propagation losses of GaAs/AlOx nonlinear waveguides (original) (raw)
2011, 2011 Conference on Lasers and Electro-Optics Europe and 12th European Quantum Electronics Conference (CLEO EUROPE/EQEC)
Continuously tunable sources with room-temperature operation are required in the mid-infrared region for applications such as spectroscopy or pollutants monitoring. In this spectral range, optical parametric oscillators (OPOs) are more versatile than laser diodes, the tunability of which is limited to a few tens of nanometers. Recently, relatively compact OPOs have started to be available, but they often remain incompatible with the necessary portability out of research laboratories. The perspective of guided-wave OPOs is promising, because of the high efficiency granted by mode confinement. In this context, GaAs is a good alternative to popular LiNbO 3 thanks to its higher χ (2) , broader transparency range, and optoelectronic integrability. We therefore focus on GaAs/AlOx form-birefringent waveguides [1] designed for type-I spontaneous parametric down conversion of a TM 00 pump at 1.06 µm into TE 00 signal-idler modes in the 1.2-4.2 µm range. Despite their encouraging 1500% W-1 cm-2 conversion efficiency (i.e. 4.5% parametric amplification in 3 mmlong samples for 30 mW of pump power) [2,3], OPO threshold has been hindered so far by the optical losses induced by AlAs oxidation process. The impact of the latter on the roughness of GaAs/AlOx interfaces can be revealed by TEM microscopy, showing an increase from 0.3 nm before oxidation to at least 0.53 nm after it. A simple, semi-analytical, Rayleigh-type model shows that the latter value of roughness accounts for scattering losses of 1.0±0.4 cm-1 for a TE 00 mode at 1.55 µm, in agreement with the measured values [4]. A deeper insight in the origin of these losses is provided by their measurement in a broad spectral range, which we have spanned with tunable lasers around 1.09 µm (Yb-YAG Koheras), 1.33 µm (Tunics-XS), 1.55 µm (Tunics-Plus) and 2.12 µm (NanoPlus DFB). Fig.1 shows the TE 00 experimental losses and the model prediction band, with roughness parameters given by TEM study of our samples. A good agreement is apparent, except at the pump wavelength, where the Fabry-Perot fringe method is less reliable, and tends to overestimate the losses, because of its sensitivity to multi-mode propagation and of an insufficient tunability of our source.