Measurement of far-infrared waveguide loss using a multisection single-pass technique Michel Rochat, a) Mattias Beck, and Je ´ ro ˆ me Faist b) University of Neucha ˆtel, CH-2000 Neucha ˆtel, Switzerland Ursula Oesterle Swiss Federal Institute of Technology, CH-1015 Lausanne, Switzerland Received 7 November 2000; accepted for publication 29 January 2001 Waveguide loss measurements based on a multisection single-pass technique have been performed for both mid-infrared and far-infrared quantum cascade structures. The far-infrared quantum cascade structures are based on a vertical transition active region emitting at 76  m, embedded in a double-plasmon waveguide. The measured waveguide loss of 4220 cm -1 agrees well with the calculated one based on free carrier absorption. © 2001 American Institute of Physics. DOI: 10.1063/1.1357444 In the mid-infrared, lasers using the quantum cascade QC technology have been demonstrated 1 and are showing high operating performances and wide spectral emission pos- sibilities 3.4–19 m. 2,3 Although various research groups have shown far-infrared FIR electroluminescence in QC structures, 4–7 laser action could not be demonstrated so far using this technology. In order to reach laser threshold, an active region with a population inversion together with an efficient resonator designed for far-infrared wavelengths is needed. In the mid-infrared, good resonators are obtained by growing low refractive index claddings below and above the high refractive index active region. This can be done by mo- lecular beam epitaxy as the needed cladding thickness is at least in the order of /2 in the material. In the far-infrared, however, waveguides based on pure dielectric confinement would require cladding thickness on the order of 10 m. This technological difficulty can be avoided by using a metal instead of dielectrics for the mode confinement. This tech- nique was first used with QC structures by Sirtori et al. in a waveguide based on surface plasmons. 8 Compared to waveguides based on dielectric confinement, this technique allows a drastic reduction of the cladding thickness, but re- sults in a higher waveguide loss as the mode penetrates partly into the metal. However, these waveguides achieve larger overlap factor, ( 70%) than a regular slab wave- guide ( 40%) with the same thickness of the waveguide core. Our FIR waveguide is based on a metal–dielectric– metal, or double plasmon confinement commonly used in microwave engineering. To make the device easier to manu- facture, a heavily doped GaAs layer replaced the metal layer below the active region. This waveguide design leads to similar guiding properties as a metal–semiconductor–metal confinement, at the expense of slightly higher calculated waveguide losses of 51 cm -1 instead of 38 cm -1 . Two struc- tures with similar active regions have been grown. They con- sist of a 120 period GaAs/AlGaAs active region embedded between two highly doped n 5 10 18 cm -3 GaAs layers. A parabolic graded AlGaAs transition region has been intro- duced between the doped and undoped regions to enable an abrupt carrier distribution profile. The elementary cells of both active regions are very similar to the far-infrared verti- cal transition QC structure published previously. 7 However, small modifications have been made, in order to increase the current injection into the exited state. The injection barriers have respectively been reduced to 5.5 and 4.7 nm instead of 6.0 nm for samples S1683 and S1806. In order to compen- sate space charge accumulation in the structures, the first well of the injector has been doped with Si. Capacitance–voltage 9 measurements performed on the struc- tures yield an average doping of n =3.410 15 cm -3 and n =7.410 15 cm -3 respectively. Compared to our first struc- ture, the maximum injection current before appearance of the negative differential resistance region was increased from 40 to 100 A cm -2 , respectively, towards more than 700 A cm -2 . The extraction barriers have also been shortened from 4 nm down to 3 and 2.3 nm to enhance the extraction of the elec- trons from the lower level. Conventional techniques usually applied to measure waveguide losses in semiconductor lasers 10–12 are not appli- cable in our case, as the observed luminescence intensity is too weak. The technique using a single, multisection device 13 is, however, feasible as it uses the luminescence generated by the QC structure inside the waveguide as the light source. In this configuration, separated electrical contacts of constant area are provided on top of the waveguide. The light inten- sity I 0 produced in a pumped section will travel down the waveguide to the edge of the sample with the intensity de- creasing accordingly to the following relation: I t =I 0 e - w L , 1 where I t is the transmitted light intensity measured at the edge of the waveguide, L is the optical path length measured from the pumped section to the sample edge, and  w is the waveguide loss. The initial light intensity I 0 , identical for all sections provided that they are pumped with same current, can be obtained by measuring the light intensity produced by a Electronic mail: michel.rochat@unine.ch b Electronic mail: jerome.faist@unine.ch APPLIED PHYSICS LETTERS VOLUME 78, NUMBER 14 2 APRIL 2001 1967 0003-6951/2001/78(14)/1967/3/$18.00 © 2001 American Institute of Physics Downloaded 29 Mar 2001 to 130.125.4.87. Redistribution subject to AIP copyright, see http://ojps.aip.org/aplo/aplcr.jsp