5.1.3 Factor Interpretation

Once principal components are obtained, these main components are naturally subject to interpretation. Initially, it is not determined at all, that the principal components have a direct physical meaning, they primarily simplify the data set. Only secondary exclusions and comparisons allow for their interpretation. Factor 1 can be considered to be strongly influenced by the sheet charge density ${\it n}_{\mathrm{sheet}}$, if not the sheet carrier density itself. The following reasoning explains this idea:

• ${\it n}_{\mathrm{sheet}}$ is strongly negatively correlated to ${\it R}_{\mathrm{S}}$ and ${\it R}_{\mathrm{D}}$, which is true for Factor 1.
• ${\it n}_{\mathrm{sheet}}$ is negatively correlated to ${\it BV}_{\mathrm{GD}}$, that is increasing ${\it n}_{\mathrm{sheet}}$ leads lower breakdown voltage, which is one reason to reduce the $\delta$-doping in power HEMTs.
• Both ${\it C}_{\mathrm{gs}}$ and ${\it C}_{\mathrm{gd}}$ are positively correlated to ${\it n}_{\mathrm{sheet}}$.
• An increased sheet charge density leads to an increase of ${\it C}_{\mathrm{ds}}$, which is true for Factor 1.

For Factor 2, no direct physical correlation could be found.

More generally, the variations in III-V HEMTs processing are different from those typical from silicon CMOS due to the wafer size. For 2-4 inch wafers the statistical sample is smaller and singular events and systematic changes mix. A list of typical variations is given in the following, some origins of these variations are given in the next section.

• the gate-to-channel separation ${\it d}_\mathrm{gc}$ [50].
• the channel sheet carrier concentration [153].
• the gate length ${\it l}_{\mathrm{g}}$.
• the alignment of the gate contact [153].
• the alignment of additional masks for double recess HEMTs.
• the etch depth for the cap etching of the second recess.

Stepping from 4 inch to 6 and 8 inch wafers the importance shifts from singular events to statistical variations due to the increase of the cell numbers in the sample.