4.4.4 Discussion

The drive current improvements for the various optimizations performed are summarized in Table 4.4. The performance gains given refer to uniformly doped devices with a bulk doping concentration of 5.47$\cdot$10$^{17}$cm$^{-3}$ and 2.39$\cdot$10$^{18}$ cm$^{-3}$ for Device Generation A and Device Generation B, respectively. In this table the optimization results using a one-dimensional approach in the vertical direction are shown, too.

Table 4.4: Performance comparison of the various drive current optimization results.

	Device Generation A		Device Generation B
device	$I_{\mathrm{on}}$ ($\mu$A)	perf. gain	$I_{\mathrm{on}}$ ($\mu$A)	perf. gain
uniformly doped	258.5	-	130.8	-
one-dimensional vertical	290.3	12.3%	176.4	34.9%
two-dimensional	373.7	44.6%	224.2	71.4%
one Gaussian function	369.3	42.9%	214.0	63.6%
two Gaussian functions	373.2	44.4%	222.5	70.1%

Fig. 4.12 shows the vertical doping profiles after the one-dimensional optimization for both Device Generation A and Device Generation B. In this case only vertical doping variations were allowed in the inverted-T region already used for the two-dimensional approach. The resulting profiles consist of a doping layer close to the silicon surface which sets the threshold voltage and a buried layer under the source and drain wells preventing punchthrough.

Figure 4.12: The drive current optimization results of the purely vertical optimizations for Device Generation A (top) and Device Generation B (bottom).

It is clearly demonstrated that the complex results from two-dimensional optimizations can successfully be substituted by Gaussian functions without a substantial performance loss. This does not mean that the two-dimensional optimizations are useless. It rather means that the core information about how to improve the device performance is delivered by the two-dimensional approach. Based on this information, Gaussian functions can be introduced to make the doping profiles smooth.

The difference in performance gain between Method 1 and Method 2 is rather small, especially for Device Generation A, and might not justify the additional effort of a second Gaussian function.

For both device generations the doping peak in the channel region is almost identical when comparing Method 1 and Method 2. This indicates that they are equivalent techniques to prevent punchthrough.