I have a very sparse non-linear system $N(u) = 0$ that can be solved as a time-dependent ODE, $\frac{du}{dt} = N(u)$, and explicitly integrated until $\frac{du}{dt} = N(u) = 0$, e.g. by forward euler, $u^{n+1} = u^n+\Delta t N(u^n)$.

In the process of this numerical integration, is there any way to recover the eigenvalues of the system by examining something about the growth/response in $u$ and it's dependence on the timestep? Gershgorin's theorem provides a bound on the spectrum, but in practice this bound's range is too large to be useful. I don't need exact eigenvalues, but relatively tight approximations would be great. I suspect there's something about this in the literature of linear time-invariant (LTI) theory, but I don't know where to start looking.


Your Answer

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

Browse other questions tagged or ask your own question.