11.08.2013 Views

Untitled - Cdm.unimo.it

Untitled - Cdm.unimo.it

Untitled - Cdm.unimo.it

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Results in Approximation Theory 117<br />

We observe that the inequal<strong>it</strong>y (6.6.2) is satisfied for all the Jacobi weights. There-<br />

fore, by (6.1.7), the rate of convergence can be estimated in the general case. For<br />

instance, for any f ∈ Ck ( Ī), k ≥ 1, the following inequal<strong>it</strong>y holds:<br />

(6.6.7) f − Iw,nf L 2 w (I) ≤ C<br />

k 1<br />

<br />

d<br />

n<br />

kf dxk <br />

<br />

C 0 (Ī)<br />

, ∀n ≥ k,<br />

where w(x) = (1 − x) α (1 + x) β , α > −1, β > −1, x ∈ I. The same statement applies<br />

for the error f − Ĩw,nf. We expect that (6.6.7) can be further refined by weakening the<br />

assumptions on f and, w<strong>it</strong>h the same rate of convergence, replacing the norm on the<br />

right-hand side of (6.6.7) w<strong>it</strong>h a weaker one (see also section 6.8).<br />

By means of inverse inequal<strong>it</strong>ies, we also give a bound to the derivative of the error<br />

as follows:<br />

(6.6.8) (f − Iw,nf) ′ L 2 w (I) ≤ (f − Π 1 w,nf) ′ L 2 w (I) + (Π 1 w,nf − Iw,nf) ′ L 2 w (I)<br />

≤ f − Π 1 w,nf H 1 w (I) + Cn 2 Π 1 w,nf − Iw,nf L 2 w (I),<br />

where Π 1 w,n is defined in (6.4.3) and where we used the inequal<strong>it</strong>y (6.3.6) for the<br />

polynomial p := Π 1 w,nf − Iw,nf ∈ Pn. Assuming a sufficient regular<strong>it</strong>y for f, the last<br />

terms in (6.6.8) tend to zero for n → +∞.<br />

It is clear now that the aliasing errors Aw,nf and Ãw,nf, introduced in section<br />

4.2, decay to zero for n → +∞, w<strong>it</strong>h a rate depending on the smoothness of f. From<br />

(6.2.4) we know that<br />

(6.6.9) f − Πw,n−1fL2 w (I) ≤ f − Iw,nfL2 w (I), ∀f ∈ C 0 ( Ī), ∀n ≥ 1.<br />

We can be more precise about this estimate.<br />

Theorem 6.6.2 - For any n ≥ 1 and any f ∈ C0 ( Ī), we have<br />

(6.6.10) f − Iw,nf 2 L 2 w (I) = f − Πw,n−1f 2 L 2 w (I) + Aw,nf 2 L 2 w (I),<br />

(6.6.11) f − Ĩw,nf 2 L 2 w (I) = f − Πw,nf 2 L 2 w (I) + Ãw,nf 2 L 2 w (I).

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

Saved successfully!

Ooh no, something went wrong!