I am reading a paper regarding the pricing of certain financial derivatives making use of the finite element method. In this paper the following weighted Sobolev spaces are introduced:

$W_{0}$ = $ \{ \phi : \int_{\Omega} \phi^2 dx + \int_{\Omega} x^2 (\phi')^2 dx < \infty \} $

$W_{2,0}$ = $ \{ \phi : \int_{\Omega} \phi^2 dx + \int_{\Omega} x^2 (\phi')^2 \int_{\Omega} x^2 (\phi'')^2 dx < \infty \} $

With associated norms $\Vert . \Vert_{0} $ and $\Vert . \Vert_{2,0} $ , $\Omega = (0,a) $ and where the derivatives are take in the weak sense.

The author then proceeds to give the following result:

$\Vert \phi - \pi_{h}\phi \Vert_{0} \leq ch \Vert \phi \Vert_{2,0} $ $\forall \phi \in W_{2,0}$, where $\pi_{h}$ is the piecewise linear interpolant of $\phi$ over [0,a].

My problem is the definition of $\pi_{h}\phi$ as the functions in the wighted space need not even be defined at 0. So far I have managed to show that the functions in $W_{2,0}$ can be approximated by bounded $C^{\infty}$ functions, are bounded and have a classically differentiable version.

I have seen two possible options:

1) Show the embedding $W_{2,0} \hookrightarrow C(\bar{\Omega})$.

2) Show that $\pi_{h}$ is a bounded linear operator and can thus be extended from the bounded $C^{\infty}$ functions to $W_{2,0}$.

However I can't seem to get either of these two methods to work. Any ideas would be appreciated.

  • $\begingroup$ If you can prove that functions in $W_{2,0}$ are bounded and can be approximated by $C^\infty$ functions, then the same proof is very likely to show that these functions are continuous. $\endgroup$ – Michael Renardy Sep 7 '14 at 13:23
  • $\begingroup$ It is basically just the right continuity at 0 that is needed, but I couldn't see any way that this followed from boundedness or the density of $C^{\infty}$. $\endgroup$ – User76765 Sep 7 '14 at 14:08
  • $\begingroup$ If the embedding is untrue, it requires a bounded oscillating function, from which it follows that the derivative will be unbounded and oscillate between $+- \infty$ as x approaches 0, similarly for the second derivative. $\endgroup$ – User76765 Sep 7 '14 at 14:24
  • $\begingroup$ Okay, so I have shown that the interpolant is not bounded, so the embedding seems to be the only option. $\endgroup$ – User76765 Sep 9 '14 at 8:17
  • $\begingroup$ How did you prove functions in $W_{2,0}$ are bounded? Did you not do this by showing that the $L^\infty$ norm is bounded in terms of the $W_{2,0}$ norm? You also say you know $C^\infty$ functions are dense. So you have a uniformly convergent sequence of $C^\infty$ functions. This implies the limit is continuous. $\endgroup$ – Michael Renardy Sep 9 '14 at 10:31

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Browse other questions tagged or ask your own question.