next up previous contents
Next: 6.2 Design of Experiments Up: 6. Constrained Optimization of Previous: 6. Constrained Optimization of

6.1 Device Structure Definition and Simulation

Figure 6.1: MINIMOS planar MOSFET structure
\includegraphics[scale=0.5]{mm-mos.eps}

The device topography conforms to the planar idealized device structure (see Fig. 6.1). The source and drain contact lengths (LS, LD) and the gate poly thickness (GCTHE) are set to the nominal gate length \ensuremath{L_{\mathit{nom}}}. The spacer widths (SGAP, DGAP) are set to $\ensuremath{L_{\mathit{nom}}}\xspace /4$. The transistor width was assumed as $\ensuremath{W}\xspace =2\ensuremath{L}\xspace .$ The doping profile is given analytically in terms of joined half Gaussian functions which are computed according to the primary device structure parameters. The channel profile is assumed constant with a concentration of \ensuremath{N_{\mathit{ch}}}. The source/drain profiles are idealized as rectangular boxes with a supersteep Gaussian fall-off, which is achieved by $\sigma_\mathit{lat} \ll$ and $\sigma_\mathit{vert,low} \ll$ and a vertical offset of $-\ensuremath{X_{\mathit{j}}}\xspace $. The underdiffusion length is set to $\ensuremath{L_{\mathit{nom}}}\xspace /8$. The concentration of the source/drain profiles is assumed as $\rm 10^{20}cm^{-3}$ which is roughly the maximum achievable electrically active arsenic concentration. Note, that there is no second source/drain doping and, also, there is no LDD doping. Instead, the shallow dopings serve only to provide the best contact between source/drain and the channel.

The primary device structure parameters, which are also technology parameters, are \ensuremath{L_{\mathit{nom}}}, \ensuremath{t_{\mathit{ox}}}, and \ensuremath{N_{\mathit{ch}}}. The channel doping is determined automatically (cf. Section C) according to a nominal off-state current \ensuremath{I_{\mathit{off,nom}}} at worst-case conditions which are $\ensuremath{V_{\mathit{DD}}}\xspace = 1.1\cdot\ensuremath{V_{\mathit{DD,nom}}}\xspace $ and $\ensuremath{L}\xspace = 0.9\cdot\ensuremath{L_{\mathit{nom}}}\xspace $. Note that \ensuremath{I_{\mathit{off}}} is a very indirect measure for circuit functionality and performance, which are the actual targets. However, to take \ensuremath{I_{\mathit{off}}} as a design parameter in optimizations enables physically and numerically well-behaved algorithms.

The device simulations were carried out with MINIMOS 6.0, which means that all effects covered by a drift-diffusion simulation are accounted for. Quantization effects in the channel are indirectly considered by the choices of \ensuremath{t_{\mathit{ox}}} as explained in Section 6.2. This, as well as the precise definition of the device structure and the optimization constraints in Table 6.3 should be kept in mind when comparing absolute values, e.g., of \ensuremath{I_{\mathit{on}}} or \ensuremath{t_{\mathit{d}}} to other published data.


next up previous contents
Next: 6.2 Design of Experiments Up: 6. Constrained Optimization of Previous: 6. Constrained Optimization of

G. Schrom