Do anyone has a hint how to proof by estimate $ e^{-ax} \leq (1-x)^{a} + \frac{1}{2}ax^2 $ for $x,a \in \mathbb{R}$ and $0 \leq x \leq 1$ and $a > 1$.
Simple starting with the definition of $e$ leads me to (if I am not wrong):
$e^{-xa} = \sum_{n \geq 0} \frac{(-1)^n}{n!}(ax)^n = 1 - ax + \frac{(ax)^2}{2}-\frac{(ax)^3}{6}+...$.
But the estimation needs a tiny trick, my ideas with estimate are too rough. I just began with thinking that $-a \leq -ax \leq 0$ and then trying to extrude it to all the other terms, but I don't know how to go further.
Otherwise it is easy to show that $e^{-x} \geq 1-x$ but this is the opposite beginning of what I want so I stop with that, too.
Thank you in advance. If you just have an input this is enough, I may want to solve it on my own, if there is an idea how to start.
Let $g(x)$ be given by
$$g(x)=(1-x)^a+\frac12ax^2-e^{-ax}$$
for $x\in[0,1]$ and $a>1$.
Applyng Bernoulli's Inequality, we see that for $x\in[0,1]$ and $a\ge1$
$$g(x)\ge 1-ax+\frac12ax^2-e^{-ax}$$
Let $h(x)= 1-ax+\frac12ax^2-e^{-ax}$. Note that $h(0)=0$ and for $a\ge 1$
$$h'(x)=a(e^{-ax}-(1-x))\ge a(e^{-ax}-(1-ax))\ge 0$$
From the mean value theorem, $h(x)\ge 0$ and therefore for $x\in[0,1]$ and $a\ge 1$
$$h(x)\ge 0\implies g(x)\ge 0\implies e^{-ax}\le (1-x)^a+\frac12ax^2$$
as was to be shown!