E.3 Numerical Results

A comparison of threshold voltages based on the definitions 1,4,5, and 6, i.e., , (using the EKV equations), , and is shown in Figs. E.1-E.3. The data were obtained by device simulations of a device with $\ensuremath{L}\xspace =1\mu\rm m$, $\ensuremath{t_{\mathit{ox}}}\xspace =5\rm n m$, and a constant channel doping of $\ensuremath{N_{\mathit{ch}}}\xspace =\rm 4\cdot10^{17} cm^-3$. What can be seen in these figures - apart from the expected qualitative dependencies - is that the threshold voltages differ by some 100mV over wide ranges of parameters. Figure E.4 shows the unique threshold current as a function of the gate length. The deviation of this curve from a straight line^E.3 indicates a difference between definitions 2 and 6. The reason for decrease of at very small channel lengths is the threshold voltage drop and the increased output conductance, which both tend to lower the value of determined according to definition 6. In other words: the peak of the $\ensuremath{I_{\mathit{T}}}\xspace (L)$ is an indicator for short-channel effects.

Figure E.1: Threshold voltage vs. channel doping

Figure E.2: Threshold voltage vs. oxide thickness

Figure E.3: Threshold voltage vs. channel length

Figure E.4: Unique threshold current vs. channel length

Footnotes

... line^E.3

or $1/L_\mathit{eff}\approx 1/(L-\Delta L)$, which would mean even a larger difference.