Strain-engineered InAs/InxGa1-xAs/GaAs metamorphic quantum dots as a model system for studying exciton confinement

The confinement of carriers in self-assembled quantum dots (QDs) is fundamental to their properties, underpinning the observation of novel physics and motivating their use in a range of applications. Here we report a unique investigation of exciton confinement in strain-engineered InAs/InxGa1 xAs/GaAs metamorphic structures in which the band-offset and QD strain are systematically varied. This is done by controlling two independent parameters of the relaxed InxGa1-xAs lower confining layer, CL [1]. Changing the fractional In content, x, results in modifications of the band discontinuities between the QDs and CLs and also affects the QD-CL mismatch f that determines the QD strain. Changing the lower CL thickness, d, only affects f. Other QD properties such as size and composition are kept invariant [1]. Fig. 1: Diamagnetic shift, ?, as a function of x and d. Samples which reach the high-field regime are grey, whilst those with a PL shift that is parabolic to the highest B are black. Lines show the QD-CL lattice mismatch, f. We have studied 26 such InAs/InxGa1-xAs/GaAs QD samples (with another 9 in progress) at 2 K using photoluminescence (PL) in magnetic fields, B, up to 17 T. This allows us to probe properties of the exciton wave-function [2], and hence study the confinement of carriers in the dots. The 'bubble plot' in Fig. 1 summarizes our results by plotting the diamagnetic coefficient, (???a?_B^2)/?, (size of the bubbles) as a function of x and d. Samples which reach the high-field limit where the PL energy shifts linearly with B (allowing us to determine Bohr radius, aB and reduced mass, µ) are indicated in grey, whilst those whose energy shift remains parabolic, (implying that aB < 9 nm) are in black. The lines are drawn for constant f, so following these changes only the QD-CL band offset, whilst going vertically through the figure only changes (reduces) the strain. Immediately obvious and striking in the figure is the ten-fold increase in ? going from f ~7% to f ~5%. We attribute this to wave-function spill-over resulting from small barriers (large x) and to low ?, which, in turn, is a consequence of low f. More intriguing is the non-monotonic variation in ? with increasing f at x of about 0.3 (going vertically). Analysis of the grey data reveals that this may be due the dependence of ? on mismatch: it increases linearly as f goes from 5% to 6%, but then drops dramatically. The physics behind this is presently not understood and is the subject of further investigation. In contrast, there seems to be no correlation whatsoever between aB and f.