Showing that the supremum of a given function is finite.

91 Views Asked by At

Consider arbitrary elements $n \in \mathbb N$ and $ \lambda \in \mathbb R$ such that $0 < \lambda < n$.

Problem. My goal is to prove that the supremum

$$ \sup_{r > 0} f(r) $$

is finite, where $f$ is defined as

$$ f(r) = \begin{cases} r^{n-\lambda}\log_2(r+2) \quad &\text{ if } 0 < r \leqslant 1 \\[.2cm] r^{-\lambda}\log_2(r+2) &\text{ if } r > 1.\end{cases} $$

Attempt. I thought about starting by differentiating $f$. Doing so, we obtain

$$ f'(r) = \begin{cases} r^{n-\lambda-1}\left( (n-\lambda)\log_2(r+2) + \frac{r}{r\ln(2) + \ln(4)} \right) \quad &\text{ if } 0 < r \leqslant 1, \\[.2cm] r^{-\lambda - 1}\left( \frac{r}{r\ln(2) + \ln(4)} - \lambda \log_2(r+2)\right) &\text{ if } r > 1.\end{cases} $$

After this, I attempted to determine the zeros of $f'$ but it turns out this isn't an easy task (at least for me). Therefore, I skipped to try and study the signal of the derivative $f'$. It's clear that for $0 < r \leqslant 1$ the derivative $f'$ is always positive, since every member involved in positive.

As for the $r > 1$, to study the signal of $f'$ we must study the expression

$$ \frac{r}{r\ln(2) + \ln(4)} - \lambda \log_2(r+2). $$

Again, just like studying the zeros of the derivative, this turned out to be not so easy. With this in mind, I am requesting help to proceed further.

Thanks for any help in advance.

1

There are 1 best solutions below

6
On BEST ANSWER

The question is not asking you to find the supremum, only prove that it is finite. So it is sufficient to prove that $f(r)$ is bounded.

First, while not completely necessary, it's nice to note that $f(r)$ is always positive. As you've noted, for $0 < r \leq 1$, $f(r)$ has positive derivative, so it's increasing.

Then for $r > 1$, we can use Lázaro's hint from the comments that $\lim_{r \rightarrow \infty} \frac{\log r}{r^\lambda} = 0$, and with a little tweaking that gives us that $\lim_{r \rightarrow \infty} f(r) = 0$. Then think back to the original definition of a limit at infinity - we know that for any $M \in \mathbb{R}^+$, there is a value $R$ such that $r \geq R \implies |f(r)| \leq M$.

So pick an $M$. We know that $\sup_{r > R} f(r) < M$. We also know that $f$ is continuous, which means that it is bounded on $(0, R)$, so set $M' = \sup_{r \in (0, R)} f(r)$. Then $\sup_{r > 0} f(r) \leq \max(M, M') < \infty$.