Fig. 5.9 – Theoretical influence of indium segregation rate on a lattice parameter. Correlation with PL peak position.
preparation of III-N materials can be subject to difficulties like gallium implantation [8][9]) since it allows the detection of composition variations over areas with a diameter of 20nm [10]. This is the next step for an optimal measure of the indium segregation and its quantification. It would strongly enhance the photoluminescence results analysis. Also, previously, we did some TEM measurements on sample 1.
The EDX analyses on the InAlN layer of sample 1 (cf. figure 5.10) do not appear to demonstrate any segregation of indium. This could be for a variety of reasons. Perhaps this sample had less segregation, perhaps the measurement lacks sensitivity due to the effects of projection on the segregation or that the contrast of the blue color is not large enough to perceive this segregation. Finally, perhaps further work with XRD RSM, using longer more detailed scans would enable the measurement of two different peaks, to clearly identify the different phases in the layers.
5.2 | Gallium pollution in MOCVD reactors growing Ga containing alloys
In almost all of the InGaAlN layers we investigated previously, the XPS measurements showed gallium incorporated in various quantities. We decided to understand how the gallium content evolved with growth parameter and to see if this was linked with the TMIn flow. Finally we found a technical solution to grow pure InAlN samples.
Fig. 5.10 – EDX analysis of TEM pictures of InAlN/GaN (sample 1)
5.2.1 | Calculation of a virtual TMGa flow from pollution sources in the reactor To understand how growth parameters have an impact on gallium incorporation, we calculated an effective TMGa flow coming from the pollution sources as if there were a real TMGa flow. The calcula-tion was performed as follows. First we divide the total InGaAlN quaternary layer with a thickness tInGaAln into three InN, GaN, AlN binary layers with respective thicknesses :tInN = %In×tInGaAlN, tGaN = %Ga×tInGaAlN, andtAlN = %Al×tInGaAlN. In the same way we can define the growth rate of each binary by :V gInN = %In×V gInGaAlN,V gGaN = %Ga×V gInGaAlN, andV gAlN = %Al×V gInGaAlN. Now, we assume that our effective TMGa flow is normalized compared to TMAl flow. It is simpler to normalize the layers relative to the TMAl, because indium is volatile and this can complicate the calculations. For this normalization, the ratio of TMGa to TMAl in the gas phase is considered to be the same as the ratio of Ga content to Al content in the solid phase. This gives the equation :
T M GaEf f ectivef low
T M AlRealf low = V gGaN
V gAlN (5.1)
If we apply equation 5.1 to our experimental points we can produce the graph shown in figure 5.11.
This is a plot of many different samples grown at various temperatures, growth rates and indium flows.
Despite this wide variety of parameters, the graph shows a linear evolution of TMGa effective flow coming from the pollution versus real TMIn flow ; with only the carrier gas and the chamber pressure significantly changing the behavior. We verified that there was no external source of gallium reaching the reactor, and so we have to assume that the gallium is desorbing from the GaN deposition on the walls which occurred during the growth of our buffer layers.
5.2. GALLIUM POLLUTION IN MOCVD REACTORS GROWING GA CONTAINING
ALLOYS 143
Fig. 5.11 – Effective TMGa flow coming from the pollution versus real TMIn flow. Except for two points indicated by arrows, all the other samples were grown at 100mbar.
Figure 5.11 implies that increasing the TMIn flow will always increase the pollution by gallium, and therefore suggests that the TMIn is somehow catalyzing the desorption of gallium from the chamber.
The only parameter which seem to affect the gallium desorption are the carrier gas and the pressure.
The use of hydrogen as a carrier gas increases the pollution, but this change may be more related to a decrease of the indium content therefore promoting an increase of the gallium content. In any case, as we have shown in the previous chapter, this is a difficult condition for InAlN growth. The reduction of chamber pressure appears to cause a slight improvement in gallium pollution, but the effect is very limited.
We did not see a clear effect of temperature on this relationship, which means that potentially a reduction in TMIn flow accompanied by a reduction in temperature (which has been previously seen to increase indium incorporation) could be a way to keep the indium constant in the layers while reducing the gallium pollution. However, this would only give a slight decrease, so a more radical solution is required.
5.2.2 | Cleaning test with Cl2 : growth of pure InAlN
As the gallium appears to be coming from the wall and showerhead deposition during the GaN layer growth preceding the InGaAlN layers, we performed a test using the Cl2 cleaning of the chamber to avoid this effect (cf. figure 5.12). After the GaN layer was grown, the sample was removed from the chamber and stored in a load-lock under low pressure N2 atmosphere. The chamber was cleaned, before reintroducing the wafer to the chamber and performing growth of the InAlN.
Fig. 5.12 – Cleaning process between GaN and InAlN layers.
The resulting InAlN layers are totally free of gallium, when using either carrier gas, confirming the hypothesis of the source of the gallium as the wall or the showerhead in the growth chamber. And even more interestingly, this significantly increased the indium content in the layers. Under nitrogen we got 18% indium and 11% under hydrogen, where without the cleaning, using the same condi-tions, we had previously incorporated only 7.5% and 4% respectively. This may suggest that indium was in competition with gallium for its incorporation into the layer, or that the interaction between TMIn and the chamber, which resulted in the gallium pollution, was also consuming some of the indium.
The surface morphology of these four samples, with and without clean, and under the two carrier gases, are shown in figure 5.13. There is not a huge difference with and without cleaning, but globally we see a smoother surface morphology after cleaning, even though we have a higher indium content. This is again a positive result, showing the potential of using the cleaning between wafers to get high quality layers without gallium incorporation.
Finally, the sheet resistance of the layer with 18% indium was measured, resulting in a value of 250 Ohm/sq, showing that the GaN channel layer was not strongly impacted by its removal from the chamber during the cleaning. 250 Ohm/sq was also the sheet resistance value measured on the wafer grown without a cleaning process. That suggests that such a low 2DEG resistivity is more related to the presence of an AlN spacer [11][12][13][14] we placed on both samples (before and after cleaning, under nitrogen). A similar layer with 7.5% indium, the wafer without a cleaning process between the GaN and the InGaAlN layers, showed us a sheet resistivity above 350 Ohm/sq.