Whereas the PL peak energy monotonically changes with the Bi fraction and P in, a different behavior is observed with the spectrum full-width at half maximum (FWHM). The observation of the spectral broadening in Figure 2 suggests an increase of the FWHM with adding Bi. However, this is true only at high excitation intensity, as it is shown in the inset of Figure 4, where there is a clear PL narrowing effect with Bi% at low P in.

This can be explained in terms of clustering effects and localized exciton states induced by Bi incorporation. At low excitation power, the PL signal is dominated by localized exciton recombination, whose energy distribution shrinks with increasing Bi, moving from a set of quasi-discrete energy levels to a quasi-band formation with a larger density of states (see illustration in the top of Figure 4 inset), and hence resulting in an enhanced contribution to the PL spectrum. Figure 4 PL FWHM PHA-848125 price vs. P in for the PLX3397 in vitro three samples. The inset shows the FWHM vs. Bi%, for the three excitation power densities and a scheme of Bi cluster state distribution. With increasing incident power, the localized levels saturate, giving rise to delocalized excitons and to an increase in the FWHM. This is probably due to inhomogeneous broadening caused by fluctuations in the local Bi composition, valence band potential, and strain distribution, and eventually

band filling. The change in the FWHM with P in is illustrated in Figure 4 for three samples, where the two different processes depending on the P in clearly appear. All five samples follow the same u-shaped trend, with a minimum FWHM in the P in region between 0.5 and 20 mW, Loperamide as already observed by Mazur et al. [16] in GaAsBi QW samples under CW excitation power. The excitation power corresponding to this minimum for each sample

will be referred as P MIN. At low intensity, excitons tend to be highly localized and cannot be separated, so they recombine radiatively. By increasing P in, filling of the localized states occurs, and delocalized excitons start recombining, with the PL emission energy approaching the theoretical Varshni curve. From previously reported Arrhenius plot in a similar sample, we observed that there is a continuous set of activation energies for these excitons (some of which can be cured by thermal annealing) [15]. Therefore, their contribution is expected to be always present, but predominant at the lowest P in Target Selective Inhibitor Library supplier values. In order to discriminate the contribution of delocalized and localized excitons, an efficient way consists in separating them in two families, in a similar way as reported by Mazur et al. [16], and fit all PL spectra by two Gaussians. Figure 5 shows, for example, the GaAsBi PL transition of sample 1, which is strongly asymmetric, together with the Gaussian fitting of the two exciton recombination-related peaks. Figure 5 Fitting (black line) of the normalized sample 5 PL spectrum (circles) with the sum of two Gaussian curves.