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.

252 Polynomial Approximation of Differential Equations<br />

(11.2.5) pn,k(sk) = pn,k+1(sk) 1 ≤ k ≤ m − 1,<br />

(11.2.6) p ′ n,k(sk) = p ′ n,k+1(sk) 1 ≤ k ≤ m − 1,<br />

(11.2.7) pn,1(s0) = σ1, pn,m(sm) = σ2.<br />

Basically, in (11.2.4) we collocated the differential equation at the nodes inside Sk,<br />

1 ≤ k ≤ m, and imposed the boundary cond<strong>it</strong>ions (11.2.7). Relations (11.2.5) and<br />

(11.2.6) require the continu<strong>it</strong>y of πn and <strong>it</strong>s derivative in the interval I, i.e., πn ∈ C 1 (I).<br />

It is clear that (11.2.4)-(11.2.7) is equivalent to a m(n+1)×m(n+1) linear system. To<br />

determine the corresponding matrix, we argue as in sections 7.2 and 7.4. For instance,<br />

the entries of the derivative matrix at the collocation nodes θ (n,k)<br />

j , 0 ≤ j ≤ n, are given<br />

by 2 ˜ d (1)<br />

ij /(sk − sk−1) 0≤i≤n<br />

0≤j≤n<br />

, 1 ≤ k ≤ m.<br />

In the ultraspherical case, we can replace (11.2.6) by<br />

(11.2.8)<br />

−1<br />

sk+1 − sk−1<br />

+ γk<br />

<br />

(sk − sk−1)p ′′ n,k + (sk+1 − sk)p ′′ <br />

n,k+1 (sk)<br />

<br />

p ′ n,k − p ′ <br />

n,k+1 (sk) = f(sk) 1 ≤ k ≤ m − 1,<br />

where γk, 1 ≤ k ≤ m, are su<strong>it</strong>able constants. In this case we only have πn ∈ C 0 (I).<br />

The use of formula (11.2.8) was proposed in funaro(1986) for Legendre nodes and<br />

funaro(1988) for Chebyshev nodes. This new cond<strong>it</strong>ion, which is closely related to<br />

(9.4.23), results from a variational formulation of problem (11.2.1). When the constants<br />

γk in (11.2.8) are chosen appropriately, numerical results are in general more accurate<br />

than those obtained w<strong>it</strong>h cond<strong>it</strong>ion (11.2.6). In both cases, we expect convergence of<br />

the approximating function πn to the exact solution U when n → +∞. A convergence<br />

analysis can be developed following the variational approach described in sections 9.3<br />

and 9.4. To illustrate the basic principles, we show the theory in a simple case. We<br />

recall that there is no loss of general<strong>it</strong>y if we assume σ1 = σ2 = 0 in (11.2.1).

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

Saved successfully!

Ooh no, something went wrong!