1
$\begingroup$

Suppose that $\lim_{n\rightarrow\infty}\sqrt[n]{a_n}=1$ and $\lim_{n\rightarrow\infty}\frac{b_n}{a_n}=g$ where $g\in (0,\infty)$. We must prove that $\lim_{n\rightarrow\infty} \sqrt[n]{b_n}=1$.

Intuition tells me that this should be done by contradiction. But I can't get it to work. Here is what I came up with so far:

We know that for every $\epsilon>0$, for sufficiently large $n$ we have $\sqrt[n]{a_n}<\epsilon+1$. Suppose that $\lim_{n\rightarrow\infty} \sqrt[n]{b_n}\neq 1$. That means that there exists $\epsilon'$ such that, for sufficiently large $n'$ we have $\sqrt[n]{b_n}>\epsilon' + 1$.

Let's take $\epsilon'$ and $n=\max(n, n')$. Then we have:

$\sqrt[n]{b_n}>\epsilon' + 1$

$\sqrt[n]{a_n}<\epsilon' + 1$

And here, I think, something smart should be done to conclude that, in that case $\lim_{n\rightarrow\infty}\frac{b_n}{a_n}\neq g$. But I just can't see what to do. Any hints?

$\endgroup$
5
  • 1
    $\begingroup$ I'd argue directly. For large enough $n$, we have $$\frac{g}{2} < \frac{b_n}{a_n} < 2g.$$ $\endgroup$ – Daniel Fischer Dec 7 '14 at 13:57
  • $\begingroup$ What if $g=+\infty$? $\endgroup$ – qiubit Dec 7 '14 at 14:05
  • $\begingroup$ Doesn't $g \in (0, \infty)$ mean that $0<g<\infty$? $\endgroup$ – desos Dec 7 '14 at 14:06
  • 1
    $\begingroup$ That is excluded by $g\in (0,\infty)$. If $\frac{b_n}{a_n}\to +\infty$, then it need not hold that $\sqrt[n]{b_n}\to 1$. $\endgroup$ – Daniel Fischer Dec 7 '14 at 14:07
  • $\begingroup$ Is there a contrarguement to that? $\endgroup$ – qiubit Dec 7 '14 at 14:10
1
$\begingroup$

Let $\varepsilon > 0$. From $\lim_{n \to \infty} b_n/a_n = g \in (0,\infty)$ follows that there exists $n_0 \in \mathbb{N}$ s.t. \begin{align*} n \geq n_0 &\implies \left| \frac{b_n}{a_n} - g \right| < \varepsilon \\ &\iff g - \varepsilon < \frac{b_n}{a_n} < g + \varepsilon \end{align*} We can assume that $g - \varepsilon > 0$. Now \begin{align*} \implies \sqrt[n]{g-\varepsilon} &< \frac{\sqrt[n]{b_n}}{\sqrt[n]{a_n}} < \sqrt[n]{g + \varepsilon} \\ \iff \sqrt[n]{g-\varepsilon}\sqrt[n]{a_n} &< \sqrt[n]{b_n} < \sqrt[n]{g + \varepsilon}\sqrt[n]{a_n} \end{align*}

The left and right sides of the last inequality tend to $1$ as $n \to \infty$, so the result follows.

$\endgroup$
2
  • $\begingroup$ How does $(g - \epsilon)^{1/n}$ tend to $1$ and $n \to \infty$ $\endgroup$ – Amad27 Dec 7 '14 at 14:25
  • $\begingroup$ Because $g > 0$ we can consider epsilons that are less than $g$ so $g - \varepsilon$ is positive and I assumed it's a known result that then $\sqrt[n]{g-\varepsilon}$ tends to 1. $\endgroup$ – desos Dec 7 '14 at 14:29
2
$\begingroup$

Assume $\lim_{n \to \infty} (b_n)^{1/n} = M \ne 1$ and assume that the limit of $b_n$ exists:

$|(b_n)^{1/n} - M| < \epsilon$ if $n > N$

Also, $\lim_{n\to 1} (b_n)^{1/n} = b_1 \implies | (b_n)^{1/n} - 1| < \epsilon$

Let $\epsilon = |b_1 - M|/2$

$|b_1 - M|/2 = |b_1 - (b_n)^{1/n} + (b_n)^{1/n} - M|/2 = |-((b_n)^{1/n} - b_1) + ((b_n)^{1/n} - M)|/2 < |(b_n)^{1/n} - b_1|/2 + |(b_n)^{1/n} - M|/2$

$\therefore |(b_n)^{1/n} - M| < |b_1 - M|/2 < |(b_n)^{1/n} - M|/2 < |(b_n)^{1/n} - b_1|/2 + |(b_n)^{1/n} - M|/2 < |b_1 - M|$

A contradiction.

$\endgroup$
1
$\begingroup$

Is this solution correct?

For sufficiently large $n$: $\frac{g}{2} < \frac{b_n}{a_n} < 2g$. If we take the n-th degree root of all these terms we get:

$\sqrt[n]{\frac{g}{2}} < \frac{\sqrt[n]{b_n}}{\sqrt[n]{a_n}}<\sqrt[n]{2g}$. $g$ is finite so $\lim$ of left and right side evaluates to $1$. That means that $\lim \frac{\sqrt[n]{b_n}}{\sqrt[n]{a_n}}=1$. But $\lim \sqrt[n]{a_n}=1$ so in order for that to be true $\lim \sqrt[n]{b_n}$ must be equal to $1$ as well.

$\endgroup$
1
  • 1
    $\begingroup$ Looks ok to me. $\endgroup$ – desos Dec 7 '14 at 14:25

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.