0

The bumpfunction $\psi$ has the properties that given two values $a,b\in\mathbb{R}$ such that $a<b$.

$\psi$ defines a smooth function with the properties

$$f(n) = \begin{cases} e^{-\frac{1}{x-a}}\cdot e^{-\frac{1}{b-x}} &\mbox{if } a<x<b \\ 0 & \mbox{otherwise } \end{cases} $$

and $\lim_{x\downarrow a}f^{(n)}(x)=0$ and $\lim_{x\uparrow b}f^{(n)}(x)=0$ ,$f^{(n)}$ describes the $n$-th derivative. $n\in\mathbb{N}_0$

I know that the $n-$th derivative for $e^{-\frac{1}{x-a}}=a(x)$ can be given by the Formula $q_n(-\frac{1}{x-a})a(x)$. The $n$-th derivative for $e^{-\frac{1}{b-x}} = b(x)$ can be given by the formula $p_n(-\frac{1}{b-x})$ Where $p_n$ and $q_n$ is a polynomial.

I know that the $n$-th derivative for $f(x)= (a\cdot b)(x)$ exists. Because since $a$ and $b$ are smooth the product is also smooth and the $n$-th derivative can given by the recursive fomula:

$f^{(1)}=(a\cdot b)^{(1)}=a'b+b'a$ and $f^{(n)}=(f^{(n-1)})'$

Now I want to prove that $\lim_{x\downarrow a}f^{(n)}(x)=0$ and $\lim_{x\uparrow b}f^{(n)}(x)=0$ , $n\in\mathbb{N}_0$

I have made the assumption that the $n$th derivative is of the form $$q_n(-\frac{1}{x-a})+p_n(-\frac{1}{b-x})((a\cdot b)(x))$$

But when I made the inductionstep I was runing into a Problem

$(a\cdot b)(x)((q_n(-\frac{1}{x-a})+p_n(-\frac{1}{b-x}))'+(((-\frac{1}{(x-a)^2})+(-\frac{1}{(b-x)^2}))(q_n(-\frac{1}{x-a})+p_n(-\frac{1}{b-x}))))$

The Problem is that if I multiply $-\frac{1}{(x-a)^2}$ with $p_n(-\frac{1}{b-x})$ then I get a polynomial with in a 'mixed' form. That's not what I want because this would negate the inductionstep.

Therefore I am asking you for help on the proof that:

$\lim_{x\downarrow a}f^{(n)}(x)=0$ and $\lim_{x\uparrow b}f^{(n)}(x)=0$ , $n\in\mathbb{N}_0$

New2Math
  • 1,265
  • The formula for the derivateive can be given by $f^{(n)}(x)=\sum_{k=0}^{n}\binom{n}{k}a^{(k)}b^{(n-k)}(x)$ which was proved in this question : https://math.stackexchange.com/questions/3113783/proving-the-connection-between-the-binomial-theorem-and-the-product-rule-for-der Now for $\lim_{x\downarrow a} p_n(-\frac{1}{b-x})$ converges to a constant since $a\neq b$, the same goes for $\lim_{x\uparrow b}q_n(-\frac{1}{x-a})a(x)$ then for every summand of $f^{(n)}$ we have that for $\lim_{x\uparrow b}$ or $\lim_{x\downarrow a} $ one part converges to a constant while the other goes to $0$ – New2Math Feb 15 '19 at 13:27
  • Then the whole sum for each Limit goes also to $0$ – New2Math Feb 15 '19 at 13:28

0 Answers0