next up previous contents
Next: 5.7 Example Up: 5. Contacts and Boundaries Previous: 5.5 Contact Model

5.6 Contact Voltage Variable

Having a separate solution variable for the contact voltage avoids numerical problems with large arguments of the Bernoulli function B. Using a Scharfetter-Gummel discretization scheme the expression for the current between two grid points i and j reads

Iij = C1 . (B($\displaystyle \Delta$) . nj - B(- $\displaystyle \Delta$) . ni) (5.25)
$\displaystyle \Delta$ = C2 . ($\displaystyle \psi_{j}^{}$ - $\displaystyle \psi_{i}^{}$) + C3 (5.26)

with Ci being material parameters. Applying the contact voltage directly to the boundary grid point could cause large arguments of B and hence numerical problems. This is avoided by having a separate variable for the contact voltage. At the beginning of the iteration procedure the constitutive relation for $ \psi_{C}^{}$ is violated and will only successively be adapted which guarantees numerical stability (see Fig. 5.2).

Figure 5.2: Effect of a separate potential variable on the initial-guess of the potential: a) with a separate potential variable the potential stays smooth inside the semiconductor region. b) directly applying the contact potential gives a large discontinuity of the potential.
\begin{figure}
\begin{center}
\resizebox{16cm}{!}{
\psfrag{a} {$\scriptstyle a)$...
...ncludegraphics[width=16cm,angle=0]{figures/sep-pot.eps}}\end{center}\end{figure}


next up previous contents
Next: 5.7 Example Up: 5. Contacts and Boundaries Previous: 5.5 Contact Model
Tibor Grasser
1999-05-31