Let $V : [a,b] \to \mathbb{R}$ be smooth, strictly increasing and $V(a) = 0$. Suppose that $f : [a,b] \to \mathbb{R}$ is smooth and satisfies $f^{\prime \prime} (x) + V(x) f(x) = 0$ on $[a,b]$. Can we then bound $\sup_{x \in [a,b]} |f(x)|$ in terms of $f(a) , f(b) , f^{\prime} (b), V(b)$? I intentionally don't put $f^{\prime} (a)$ into the list.
-
1$\begingroup$ If $V(x)$ were ro be negative the answer is yes, by the maximum principle. If $b-a$ is small enough there is a maximum principle on thin domains so again, yes, but I presume that otherwise the answer is probably no. $\endgroup$– leo monsaingeonCommented Jun 8, 2021 at 6:42
-
1$\begingroup$ by the way: any estimate based on such maximum principle approach would actually not involve any derivatie $f'(a),f'(b)$ $\endgroup$– leo monsaingeonCommented Jun 8, 2021 at 9:11
-
$\begingroup$ @AlexandreEremenko In your answer there, you estimate in terms of the data at the end were the potential is smallest. So here it will correspond to the end $a$, I believe. $\endgroup$– SashaCommented Jun 8, 2021 at 12:47
-
$\begingroup$ @leomonsaingeon Thank you. Unfortunately, I don't know what is the maximum principle in this context. I will try to look it up. $\endgroup$– SashaCommented Jun 8, 2021 at 12:48
-
1$\begingroup$ What you added is not enough. Consider $y(x)=A\sin(x/A)$ on the interval $[0,\pi A]$. We have $y(\pi A)=0,y'(\pi A)=1,y(0)=0$ and $V=1/A$. By small perturbation you can make the potential strictly increasing, and $V(\pi A)$ whatever you want. $\endgroup$– Alexandre EremenkoCommented Jun 8, 2021 at 14:13
1 Answer
It will be more convenient (at least to me) to turn the problem right to left, restating it as follows:
Let $V\colon[a,b]\to\mathbb{R}$ be smooth, strictly decreasing and $V(b) = 0$. Suppose that $f\colon[a,b]\to\mathbb{R}$ is smooth and satisfies $f''(x)+V(x) f(x)=0$ on $[a,b]$. Can we then bound $\sup_{x \in [a,b]} |f(x)|$ in terms of $f(a),f'(a),f(b),V(a)$?
Such a bound does not exist.
Indeed, take any natural $n$. Let $x_0:=0$. For all $k=0,\dots,n-1$, let $x_{k+1}:=x_k+k$. Let $a:=x_0=0$ and $b:=x_n$.
For all $k=0,\dots,n-1$ and all $x\in[x_k,x_{k+1})$, let
$$f(x):=(-1)^k\,k\,\sin\frac{\pi(x-x_k)}k,$$
$$V(x):=\frac{\pi^2}{k^2},$$
with $f(x_n):=0$ and $V(x_n):=0$.
Then $f\in C^1[a,b]$, $V$ is nonincreasing, $V(b)=0$, $f''+Vf=0$ on $[a,b]\setminus\{x_1,\dots,x_n\}$, $f(a)=0$, $f'(a)=\pi$, $f(b)=0$, $V(a)=\pi^2$. However, $|f((x_{n-1}+x_n)/2)|=n-1$, which is unbounded as $n\to\infty$, even though the values of $f(a),f'(a),f(b),V(a)$ are fixed.
It remains to smooth $V$ and $f$ near the points $x_k$ appropriately, also modifying $V$ to make it strictly decreasing. (It should be clear that such qualitative conditions as smoothness and strictness cannot affect the presence or absence of a quantitative bound.)
-
-
$\begingroup$ @username : That is a different story. $\endgroup$ Commented Jun 8, 2021 at 23:45
-
$\begingroup$ @username : Your question has now been answered at mathoverflow.net/a/394933/36721 $\endgroup$ Commented Jun 9, 2021 at 15:22
-
-
$\begingroup$ @username : I am glad you liked it. $\endgroup$ Commented Jun 10, 2021 at 16:57