Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I'm a beginner in PDE's and numerical methods, so please go slow :-D

I'm trying to show that the Lax-Wendroff scheme is stable for $|a\lambda| < 1$. The scheme is this:

$v_m^{n+1} = v_m^n - \frac{a \lambda }{2}\left(v_{m+1}^n - v_{m - 1}^n\right) + \frac{a^2 \lambda ^2 }{2}\left(v_{m+1}^n - 2v_m^n + v_{m - 1}^n\right)$

And I want to show that this scheme is stable without relying on Fourier methods... Instead I want to analyze the linear system that we can construct using the scheme.

So we can write out:

$v_1^{n+1} = v_1^n - \frac{a \lambda }{2}\left(v_{2}^n - v_{0}^n\right) + \frac{a^2 \lambda ^2 }{2}\left(v_{2}^n - 2v_1^n + v_{0}^n\right)$

$v_2^{n+1} = v_2^n - \frac{a \lambda }{2}\left(v_{3}^n - v_{1}^n\right) + \frac{a^2 \lambda ^2 }{2}\left(v_{3}^n - 2v_2^n + v_{1}^n\right)$

$v_3^{n+1} = v_3^n - \frac{a \lambda }{2}\left(v_{4}^n - v_{2}^n\right) + \frac{a^2 \lambda ^2 }{2}\left(v_{4}^n - 2v_3^n + v_{2}^n\right)$

...

So this gives a system of equations which we can write as a matrix equation:

$v^{n+1} = Av^{n} + f$ where $v^{n+1} = \begin{pmatrix}v_1^{n+1}\\v_2^{n+1}\\\vdots\\ v_{m-1}^{n+1}\end{pmatrix}$, $v^{n} = \begin{pmatrix}v_1^{n}\\v_2^{n}\\\vdots\\ v_{m-1}^{n}\end{pmatrix}$, f contains the boundary conditions and is $f = \begin{pmatrix}\frac{a \lambda + a^2 \lambda^2}{2} v_0^n\\0\\0\\\vdots\\0\\ \frac{-a \lambda + a^2 \lambda^2}{2} v_m^n\end{pmatrix}$ and A is the tridiagonal matrix with main diagonal having all entries being $1 - a^2 \lambda^2$, the super-diagonal having entries $\frac{a^2\lambda^2 - a \lambda}{2}$, and the sub-diagonal having entries $\frac{a^2\lambda^2 + a \lambda}{2}$.

Now, to make sure the scheme is stable, my understanding is that we want $||A|| \le 1$ where the norm can be any vector-induced norm.

So we will look at when $||A||_{\infty} \le 1$.

We see that $||A||_{\infty} = |\frac{a^2 \lambda^2 + a \lambda}{2}| + |1 - a^2\lambda^2| + |\frac{a^2\lambda^2 - a\lambda}{2}|$

Now we have several cases, depending on where $a\lambda$ lies on the number line.

If $a \lambda < -1$, then $||A||_{\infty} = \frac{a^2 \lambda^2 + a \lambda}{2} - \left(1 - a^2\lambda^2\right) + \frac{a^2\lambda^2 - a\lambda}{2}$.

If $-1 \le a \lambda < 0$, then $||A||_{\infty} = -\left(\frac{a^2 \lambda^2 + a \lambda}{2}\right) + \left(1 - a^2\lambda^2\right) + \frac{a^2\lambda^2 - a\lambda}{2}$.

If $0 \le a \lambda \le 1$, then $||A||_{\infty} = \left(\frac{a^2 \lambda^2 + a \lambda}{2}\right) + \left(1 - a^2\lambda^2\right) - \left(\frac{a^2\lambda^2 - a\lambda}{2}\right)$.

If $1 < a \lambda$, then $||A||_{\infty} = \left(\frac{a^2 \lambda^2 + a \lambda}{2}\right) - \left(1 - a^2\lambda^2\right) + \left(\frac{a^2\lambda^2 - a\lambda}{2}\right)$.

The first and last case, we boil down to $||A||_\infty = 2a^2\lambda^2 - 1$, and given $|a\lambda| > 1$, we see that $||A||_\infty > 1$ in these cases.

If $-1 \le a \lambda < 0$, then $||A||_{\infty} = 1 - a\lambda\left(a\lambda + 1\right)$ and we'd also have $||A||_\infty > 1$.

If $0 \le a \lambda \le 1$, then $||A||_{\infty} = 1 + a\lambda\left(-a\lambda + 1\right)$ and again we'd also have $||A||_\infty > 1$.

So it seems like no matter what, we'd have the scheme being unstable.

What is the mistake here?

share|improve this question
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.