11.08.2013 Views

Untitled - Cdm.unimo.it

Untitled - Cdm.unimo.it

Untitled - Cdm.unimo.it

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

An Example in Two Dimensions 289<br />

The scheme (13.3.5)-(13.3.6) is implic<strong>it</strong>. However, we can invert the block diagonal<br />

matrix −An + ωIn in (13.3.5) by computing once and for all the inverse of the block<br />

− Ân + ω În. Similarly, by wr<strong>it</strong>ing (−KnAnKn + ωIn) −1 = Kn(−An + ωIn) −1 Kn, we<br />

can deal w<strong>it</strong>h (13.3.6). The theory shows that the fastest convergence of the method is<br />

obtained when ω = λn,1 λn,n−1, where λn,1 and λn,n−1 are the smallest and the<br />

largest eigenvalues of − Ân respectively.<br />

To conclude this section, we present the results of a numerical test. In (13.1.1) we<br />

take f(x,y) := 2π, ∀(x,y) ∈ Ω, i.e., a constant charge dens<strong>it</strong>y in Ω. The corresponding<br />

potential U is plotted in figure 13.3.1. Besides, in table 13.3.1, we report for different n<br />

the value of the approximating polynomial pn at the center of Ω. Chebyshev collocation<br />

nodes have been used. As n increases, these computed quant<strong>it</strong>ies converge to a lim<strong>it</strong>,<br />

which is expected to be the potential at the point (0,0).<br />

Figure 13.3.1 - Solution of problem (13.1.1) for f ≡ 2π.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!