This is somewhat unrelated to what I normally do in mathematics, which is why it may be obvious to some of you, but I was puzzled by this:

If we look for classical solutions on $[0,1]$ to

$$-y''(x) = \lambda y(x)$$ with initial conditions $y(0)=1, y'(0)=0$ then the solution is $y_{\lambda}(x)=\cos(\sqrt{\lambda}x).$ Now consider $f(\lambda)=y_{\lambda}(1)$ Now, if we allow $\lambda$ to be complex (which is why I call it $z$ in the following) and always pick the correct branch of the square root, then a plot showed me that we have in fact $ \frac{1}{\left\lvert f(z) \right\rvert} =O\left({\left\lvert \textbf{Im}(z) \right\rvert^{-1}}\right)$ by which I mean that there is a constant $C_R>0$ such that

$\frac{{\left\lvert \textbf{Im}(z) \right\rvert}}{\left\lvert f(z) \right\rvert} \le C_R$ for $z \in \mathbb{C}\backslash \mathbb{R}\cap B(0,R) $ and $B(0,R)$ is an arbitrary ball in the complex numbers.

Numerically, I found some evidence that if $V$ is an even and let's say smooth function with respect to $\frac{1}{2}$ then the same holds for solutions to $-y''(x)+V(x)y(x)=\lambda y(x)$ with the same initial conditions. I am not really sure how to show this theoretically, if it is even true?



  • $\begingroup$ Are you sure that you wrote your initial condition correctly? $\endgroup$ – Alexandre Eremenko Feb 9 '17 at 7:55
  • $\begingroup$ @AlexandreEremenko sorry, yes you are right $\endgroup$ – Kinzlin Feb 9 '17 at 7:59
  • $\begingroup$ The solutions are entire functions of $z$ (for arbitrary $V$), there's no need to "pick a branch of the square root." $\endgroup$ – Christian Remling Feb 9 '17 at 19:00

For the equation $-y''+V(x)y=zy$, and the solution defined by $y(0,z)=1,\; y'(0,z)=1$, if we set $f(z)=y(1,z)$, then $$f(z)=\cos\sqrt{z}+O(|z|^{-1/2}\exp(|\Im z|),\quad z\to\infty.$$ The only assumption is that $V$ is continuous on $[0,1]$. See, for example, Levitan, Sargsjan, Introduction to spectral theory, AMS 1975, Chapter I, Lemma 2.2.

| cite | improve this answer | |
  • $\begingroup$ @Kinzin: The constant in $O$ is explicit. Just look at the proof in the book. Also z can approach the real axis and even to be on the real axis; asymptotics works. $\endgroup$ – Alexandre Eremenko Feb 9 '17 at 16:04
  • $\begingroup$ Continuity of $V$ is not needed for this, it works the same way for locally integrable $V$. $\endgroup$ – Christian Remling Feb 9 '17 at 18:57

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.