Loading [MathJax]/extensions/TeX/boldsymbol.js

 

 

 

Final CN equations

We can rewrite the Crank-Nicolson scheme as follows \begin{equation*} V_{j}= \left(2\hat{I}+\alpha\hat{B}\right)^{-1}\left(2\hat{I}-\alpha\hat{B}\right)V_{j-1}. \end{equation*} We have already obtained the eigenvalues for the two matrices \left(2\hat{I}+\alpha\hat{B}\right) and \left(2\hat{I}-\alpha\hat{B}\right) . This means that the spectral function has to satisfy \begin{equation*} \rho(\left(2\hat{I}+\alpha\hat{B}\right)^{-1}\left(2\hat{I}-\alpha\hat{B}\right)) < 1, \end{equation*} meaning that \begin{equation*} \left|(\left(2+\alpha\mu_i\right)^{-1}\left(2-\alpha\mu_i\right)\right| < 1, \end{equation*} and since \mu_i = 2-2cos(\theta) we have 0 < \mu_i < 4 . A little algebra shows that the algorithm is stable for all possible values of \Delta t and \Delta x .