Assume a function $M(t)$ strictly increasing, essentially growing exponentially, and asymptoptically growing at a known rate $\bar{g}$, i.e. $\lim_{t\to\infty}M'(t)/M(t) = \bar{g}$

In a set of awful integro-differential equations, I need to use the inverse of $M(t)$, denoted by $q(\cdot)$ for simplicity, i.e. $t = M^{-1}(M) \equiv q(M)$.

The function $M(t)$ will be solved numerically with some sort of spectral collocation method with a polynomial basis (see below). My general problem is: with a guess on $M(t)$, how can I find $q(M)$?

Since this equation is growing exponentially, I don't think I can use a polynomial basis for $M(t)$ directly. Instead, I suspect that I should guess a Chebyshev basis for $g(t)$ constraining for some large $T$ that $g(T) = \bar{g}$. If so, then $$M(t) \equiv \exp\left(\int_0^t g(\tau) d \tau \right)$$

Which works fine for my computations. In particular, it is easy with a Chebyshev basis to find the partial integrals $\int_0^t g(\tau) d \tau$ for $t$ nodes of the Chebyshev basis (through Calculating integrals for a function approximated by Chebyshev polynomials)

But if so, how can I find the $q(M)$ function, or a basis of it in $M$ space? Obviously, I could just evaluate $M(t)$ at the Chebyshev nodes of $t$ to find the appropriate $M$, and then use some sort of interpolation to find the approximating $q(M)$, but I would prefer an approach that has a chance of using auto-differentiation by throwing it into my collocation method.

For example, is there a way to translate this into a nonlinear ODE using the inverse-function theorem, and if so does this simplify substantially given that this is a Chebyshev basis?



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Browse other questions tagged or ask your own question.