For the diffusion equation $\dfrac{\partial \rho}{\partial t} = D\,\Delta \rho$, I'm trying to show that Crank-Nicolson is second order. I've isolated the truncation error $\tau_i^{n+1}$ to be $ D\left(\frac{\rho(x_{i+1}, t_{n+1}) + \rho(x_{i-1}, t_{n+1}) - 2\rho(x_i, t_{n+1})}{2(\Delta x)^2} + \frac{\rho(x_{i+1}, t_n) + \rho(x_{i-1}, t_n) - 2\rho(x_i, t_n)}{2(\Delta x)^2}\right)$, but I'm having trouble actually showing that it's $O((\Delta t)^2 + O((\Delta x)^2)$.

How should I proceed? A simple hint (ie how to to the Taylor Series) would be just fine!

Thank you!

  • $\begingroup$ is $D$ the derivative or just a scalar? $\endgroup$ – Ellya May 6 '14 at 20:18
  • $\begingroup$ Hint: try to develop each of the first long fraction near the point $(x_i,t_{n+1})$ and the second near $(x_i,t_n)$ up to fourth degree. $\endgroup$ – TZakrevskiy May 6 '14 at 20:20

Taylor's in two variables:

$f(x,y)=f(a,b)+\sum_{1\le r+s\le n}\frac{\partial^{r+s}f(a,b)}{\partial x^r\partial y^s}\frac{(x-a)^r}{r!}\frac{(y-b)^s}{s!}+E_n(x,y)$

Where the error term $E_n(x,y)=\sum_{r+s=n+1}\frac{\partial^{r+s}f(\xi,\eta)}{\partial x^r\partial y^s}\frac{(x-a)^r}{r!}\frac{(y-b)^s}{s!}$

Where $(\xi,\eta)$ is some value on the line between the points $(a,b)$ and $(x,y)$.

And as the commenter pointed out you will probably need a fourth order approximation to cancel with the $(\Delta x)^2$ on the bottom of the fraction.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.