Exciton Confinement in Strain-engineered InAs/InGaAs/GaAs Metamorphic Quantum Dots for Telecom Wavelength Emission

Lasers with long wavelength emission (1.3 µm to 1.55 µm) are useful for telecommunications due to low attenuation in optical fibers. Although lasers based on self-assembled quantum dots (QDs) emitting at 1.3 µm are commercially available, producing longer wavelength devices has proven challenging. One possible route for achieving this is the use of strain-engineered InAs/InGaAs/GaAs quantum dots grown on metamorphic InGaAs layers, where InGaAs is used as upper and lower confining layers (CLs) for exciton confinement. The emission wavelength is controlled by two independent parameters.1-2 (1) By changing the amount of In in the upper confining layer (UCL) and lower confining layer (LCL). This changes the band discontinuities between the QDs and confining layers and also affects the mismatch (strain). (2) Thickness, d, of the LCL which affects only the mismatch between the QDs and CLs. By this method it should be possible to optimize the confinement whilst also extending the wavelength to 1.55 µm: emission as long as 1.59 µm in similar metamorphic nanostructures has been reported.3 Measuring the emission wavelength is straightforward, but how about the confinement? Here we do this by studying exciton properties (radius, aB and reduced mass, ?) using low temperature magneto-photoluminescence.4 We have characterized twelve samples of strain-engineered InAs/InGaAs/GaAs QDs at 2 K and in magnetic fields 0 - 17 T. TEM measurements confirm that the dot morphology is independent of both d and CL In content. For samples where the dot-CL mismatch is low (5 to ~6 %), we are able to reach the high-field regime, allowing us to determine aB and ?. We find that ? increases with increasing mismatch, but see no clear trend in aB. A correlation between mismatch and ? is not unexpected, but note that aB will depend on mismatch (via ?) and band offset (CL In content). The diamagnetic coefficient (which is proportional to (a_B^2)/?) decreases with increasing mismatch for all samples, consistent with the trend in ? seen at low mismatch, but at high mismatch (>6%) this effect saturates. We also find that the high-field regime is not reached at high mismatch (>6.4%), so aB must, in general, be smaller at high mismatch than at low mismatch.