6 Interpretation

In both PL and time resolved PL an effect is observed which manifests itself as a blueshift of the spectrum as the excitation wavelength is shortened. A possible explanation is that there is a saturation effect at play. The band-to-band Si NC PL saturates, while the Si=O bonds at the edges of the NC do not21. To investigate whether or not the laser power and wavelength dependence is the same effect, I’ve tried to quantitatively estimate the energy each of these methods put into the NCs. A common method of displaying this information is plotting the PL amplitude as function of the average excited NC fraction. The average excited NC fraction can be calculated from the laser power and wavelength, and the silicon absorption curve.


PIC
Figure 19: The fraction of Si=O defect related PL and band-to-band NC PL at similar concentration of excited NCs versus the photonenergy.



PIC
Figure 20: A proposed alternative energy level for the Si=O defect.


In figure 19 the fraction of Si=O and bandgap-related NC PL is plotted. This was done because the absolute amplitudes weren’t comparable between all measurements. The ratio is independent of this, so with figure 19 is possible to tell something about the Si=O effect relative to the bandgap-related NC PL, namely that our suspicions based on the spectral measurements are true. The downside is that I can’t tell for sure whether this is anywhere near saturation.

6.1 What can be improved?

Calculating the average excited NC fraction assumes a uniform NC distribution inside the sample, and ignores any edge effects at the surface between the NCs and the substrate. It would be helpful to know more about the exact composition of the NC sample, but also the average excited NC fraction should ideally be measured, and not calculated.

Another improvement would be to extend and systematize the ad-hoc characterization method used in this paper. With few measurements, a lot of new physics is laid bare. The literature consists mostly of establishing the presence of effects, rarely of complete NC models that explain much of the observations. Creating a systematic in which all characteristics present in a NC sample are included, will make the creation of new complete models and theories easier and more clear.

Five concrete recommendations for future research are therefore: