4
$\begingroup$

Mike McNulty, who is a postdoc working with me, showed me the following trick for looking at asymptotic behavior of ODEs near singular points that he found; my question: does it have a well-known name in the literature?

Preliminary

Consider an ODE $u'' = Vu$. The method of reduction of order / quadrature tells us that, given one solution $u_0$ that does not vanish on an interval $(a,b)$, one can generate another linearly independent solution $v$ by solving $$ \left( \frac{v}{u_0} \right)' = \frac{1}{u_0^2} \tag{*}$$

Method

Suppose now that $V$ is possibly singular, so that on the interval $(0,1)$ the solution $u_0$ looks like $u_0(x) = x^\lambda \omega(x)$, where $\omega$ is bounded away from zero, and $\lambda > 0$. Then we can write-down $v$ using the following Laurent-series like expansion (basically by fiddling around with (*) and differentiating by parts). Here $M$ is an arbitrary positive integer denoting the order of expansion:

$$ \left(\frac{v}{u_0} \right)' = \left[\sum_{k = 1}^{M} (-1)^{k-1} \frac{x^{k - 2\lambda}}{(k-2\lambda)(k-1-2\lambda) \cdots (1-2\lambda)} \left( \frac{1}{\omega^2(x)}\right)^{(k-1)}\right]' + (-1)^M \frac{x^{M-2\lambda}}{(M-2\lambda)\cdots(1-2\lambda)} \left(\frac{1}{\omega^2(x)}\right)^{(M)} $$

Note

This feels like something that should be classically known, but I can't find anything like it in the three ODE texts I own (Hartman, Birkhoff-Rota, Hale; though possibly I was looking at the wrong parts). Any references would be appreciated. I feel that it is a lot like the method of Frobenius, except we do not assume analyticity.

$\endgroup$
4
  • $\begingroup$ Don't know if this old question is still of interest, but I don't really see an innovative advantage to writing an asymptotic expansion for $v/u_0^2$ in terms of the derivatives of $1/\omega^2(x)$ rather than just replacing that function by its Taylor series and integrating $x^{-2\lambda}/\omega^2(x)$ term by term. Either way, you at most obtain an asymptotic series for $v(x)$ at $x=0$. Knowing the full (or even a sufficiently high order) expansion for $v(x)$, it is possible to construct $v(x)$ by Picard iteration (without analyticity assumptions), cf proof of Thm.17.1 in Wasow. $\endgroup$ Commented Jul 11, 2023 at 14:10
  • $\begingroup$ @IgorKhavkine: no, generally you don't expect this to give you too much of an advantage; but in the specific case, we know the explicit function $\omega$ so it is easier not to take the Taylor series expansion. The intent of this question is pretty much just to give credit if appropriate $\endgroup$ Commented Jul 11, 2023 at 14:31
  • $\begingroup$ Since a Poincaré asymptotic expansion (power series asympototics) is unique when it exists, the difference between your way of writing it (using $\omega(x)$) and the output of the Frobenius method consists only of lower order terms. In the matter of the expansion itself, I think then credit belongs to Frobenius (analyticity is irrelevant, as all you need are formal power series expansions of everything at the singular point). If you need something like existence by the singular case of Picard's theorem, some classical references are around Sec.3.3 of Erdélyi's Asymptotic Expansions (1956). $\endgroup$ Commented Jul 11, 2023 at 21:59
  • $\begingroup$ @IgorKhavkine: thank you for the reference! I will check it out later. $\endgroup$ Commented Jul 12, 2023 at 13:15

1 Answer 1

1
$\begingroup$

Have you tried W. Wasow, Asymptotic Expansions for Ordinary Differential Equations (Wiley, 1965, Dover reprint 2018)?

$\endgroup$
2
  • $\begingroup$ Nope, but I will (just requested the book from the library). Thanks for the pointer, will report back if I find something. $\endgroup$ Commented Feb 10, 2023 at 14:28
  • $\begingroup$ Forgot to report back. I got the book from interlibrary loan, and wasn't able to find that precise method. $\endgroup$ Commented Mar 12, 2023 at 19:37

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .