7

Considering that asymptotically, $2^x$ grows faster than $x^5$ (in the beginning, $x^5$ grows faster than $2^x$, but there will be a point where $2^x$ outgrows $x^5$) then $\dfrac{x^5}{2^x} \rightarrow 0$ as $x \rightarrow \infty$. Therefore,
$$\lim _{x\to \infty}\dfrac{x^5}{2^x} = 0$$

But in order to solve the limit, I applied L'Hospital's Rule five times

\begin{align} \lim _{x\to \infty}\dfrac{x^5}{2^x} & =\lim _{x\to \infty}\dfrac{5x^4}{2^x\ln 2}\\ & = \lim _{x\to \infty}\frac{20x^3}{\ln^2(2)\cdot 2^x} \\ & = \lim _{x\to \infty}\frac{60x^2}{\ln^3(2)\cdot 2^x} \\ & = \lim _{x\to \infty}\frac{120x}{\ln^4(2)\cdot 2^x} \\ & = \lim _{x\to \infty}\frac{120}{\ln^5(2)\cdot 2^x} \\ & = \frac{120}{\ln^5(2)}\cdot\lim _{x\to \infty}\frac{1}{2^x} \\ & = 0 \end{align}

What would be a more elegant way solve it without using L'Hospital's Rule?

Edit

Even though, the Limit: $\lim_{n\to \infty} \frac{n^5}{3^n}$ is similar, I found the link provided by Axion004, How to prove that exponential grows faster than polynomial? more interesting. Also, the answer provided by user trancelocation was very interesting and is what I was expecting.

7 Answers7

12

You can use the series expansion $e^t = \sum_{n=0}^{\infty}\frac{t^n}{n!}$ as follows:

For $x>0$ you have $$\frac{x^5}{2^x}= \frac{x^5}{e^{x\ln 2}}< \frac{x^5}{\frac{(x\ln 2)^6}{6!}}= \frac{6!}{\ln^6 2}\cdot \frac 1x$$

5

Hint :

\begin{align*} \frac{x^5}{2^x} &= \exp \left(5 \ln(x)-x \ln(2)\right) \\ &= \exp \left[x\left(5 \frac{\ln(x)}{x}-\ln(2) \right)\right] \end{align*}

Now, you have the very classical limit (which can be proved with elementary method) $$\lim_{x \rightarrow +\infty} \frac{\ln(x)}{x} = 0$$

so $$\lim_{x \rightarrow +\infty} \left(5 \frac{\ln(x)}{x}-\ln(2) \right) = -\ln(2)$$

so $$\lim_{x \rightarrow +\infty} \left[x\left(5 \frac{\ln(x)}{x}-\ln(2) \right)\right] = -\infty$$

and you are done.

TheSilverDoe
  • 29,720
  • 3
    But then how do you show $5 \ln(x) - x \ln(2) \to - \infty$ without converting back to $x^5/2^x$? – user56202 Oct 01 '21 at 15:03
  • @TheSilverDoe Thank you for the answer, but could you explain a bit further? I found intuitive that $\dfrac{x^5}{2^x} \rightarrow 0$ as $x \rightarrow \infty$, but I also would like to show, write down the process to get the solution. I am not sure how the hint would help. – 欲しい未来 Oct 01 '21 at 15:13
  • @user56202 I edited my answer. – TheSilverDoe Oct 01 '21 at 15:18
  • The inner limit is $\lim_{x\to\infty}\left(x\left(5\frac{\log(x)}{x}-\log(2)\right)\right)$. While $\log(x)/x\to 0$, $x\log(x)/x\to\infty$. Your argument is therefore flawed. – Mark Viola Oct 01 '21 at 16:35
  • 1
    @MarkViola Don't you think it's true that if $f(x) \to \infty$, $g(x) \to L \in \mathbb{R} - {0 }$, then $f(x) \cdot g(x) \to \text{sgn}(L) \cdot \infty$? – user56202 Oct 01 '21 at 16:40
  • 1
    @TheSilverDoe Thank you for the update. – user56202 Oct 01 '21 at 16:41
  • @MarkViola No, it is not, see the explanation by user56202. – TheSilverDoe Oct 01 '21 at 17:10
  • @TheSilverDoe That explains nothing about your lack of rigor herein. – Mark Viola Oct 01 '21 at 18:40
  • @user56202 Yes. And what is your point? – Mark Viola Oct 01 '21 at 18:40
  • @MarkViola Please don't call a "lack of rigor" anything you don't understand. Just apply user56202's comment with $f(x)=x$ and $g(x)=5\frac{\ln(x)}{x}-\ln(2)$ and you will see that the argument is not flawed. – TheSilverDoe Oct 02 '21 at 21:49
  • 1
    @MarkViola In this answer, $f(x) = x$ and $\displaystyle g(x) = 5 \frac {\ln(x)}{x} - \ln(2)$. $f(x) \to +\infty$ and $g(x) \to - \ln(2) < 0$, so $f(x)g(x) \to - \infty$, so $e^{f(x)g(x)} \to 0$. Is there anything wrong with this? – user56202 Oct 02 '21 at 21:49
  • @TheSilverDoe Stand down! It appears you've made some edits. And now your solution is fine. – Mark Viola Oct 02 '21 at 23:10
  • @user56202. No. – Mark Viola Oct 02 '21 at 23:12
  • @MarkViola All the edits had been made before your comments... But nevermind, if the solution is fine for everyone, then there is no need to continue the discussion. – TheSilverDoe Oct 03 '21 at 07:15
5

As an alternative way, by ratio test

$$\frac{\dfrac{(n+1)^5}{2^{n+1}}}{\dfrac{n^5}{2^n}}=\frac12\left(1+\frac1n\right)^5 \to \frac12 \implies \dfrac{n^5}{2^n} \to 0$$

and since $\forall x>0\quad \exists n$ such that $n\le x\le n+1$ we have

$$\dfrac{x^5}{2^x}\le \dfrac{(n+1)^5}{2^{n}}=2 \dfrac{(n+1)^5}{2^{n+1}} \to 0$$

user
  • 154,566
3

Let $x=5u$. Then

$$\lim_{x\to\infty}{x^5\over2^x}=5^5\left(\lim_{u\to\infty}{u\over2^u}\right)^5$$

so it suffices to compute $\lim_{u\to\infty}u/2^u$. Let's do this using an inequality starting with the binomial theorem:

$$2^n=(1+1)^n=1+{n\choose1}+{n\choose2}+\cdots+1\gt{n\choose2}={n(n-1)\over2}\ge{n^2\over4}$$

for integers $n\ge2$. It follows that

$$0\le{u\over2^u}\le{\lceil u\rceil\over2^{\lfloor u\rfloor}}\le{\lfloor u\rfloor+1\over\lfloor u\rfloor^2/4}=4\left({1\over\lfloor u\rfloor}+{1\over\lfloor u\rfloor^2}\right)\to0$$

so by the Squeeze Theorem, $\lim_{u\to\infty}u/2^u=0$.

Barry Cipra
  • 79,832
2

All we need is to know that $$ \lim_{t\to\infty}\frac{e^t}{t}=\infty \tag{1} $$

Let's prove that, for every $a>1$ and $b>0$ (not necessarily an integer), we have $$ \lim_{x\to\infty}\frac{x^b}{a^x}=0 \tag{2} $$ First of all, perform the substitution $x=by$, so $a^x=(a^y)^b$ and our limit becomes $$ \lim_{x\to\infty}b^b\Bigl(\frac{y}{a^y}\Bigr)^{b} \tag{3} $$ OK, if we can prove that the limit of the part in parentheses is $0$, we're done. It's quite similar to $(1)$, isn't it? Since $a^y=e^{y\log a}$, we can perform a further substitution $y\log a=z$ and we get $$ \lim_{y\to\infty}\frac{y}{a^y}=\lim_{z\to\infty}\frac{1}{\log a}\frac{z}{e^z} \tag{4} $$ which is indeed $0$ because of $(1)$. The assumption that $a>1$ has been used here, because in this case $\log a>0$.

Should we prove $(1)$? You find several proofs that don’t use l’Hôpital. Perhaps the simplest is to use the mean value theorem to prove that, for $t>0$, it holds that $$ e^t>1+t+\frac{t^2}{2} $$

egreg
  • 238,574
  • Hi very kind. Just for my recent question https://math.stackexchange.com/questions/4259505/easier-way-to-display-of-commutation-lemma-for-limit-operator have you a possibility to give me a simplest answer? Thank you very much. – Sebastiano Oct 03 '21 at 20:08
1

$x^5$ and $2^x$ are both increasing monotonically, this is important because it guarantees there's no "weird" behavior at any subset of the real line. At the same time, as you've noticed, $(2^x)'>(x^5)'$. This analysis guarantees the limit. Of course, to be rigorous about it, you'd have to prove both claims made here, so the L'hopital's rule solution might be the easiest method.

Sigma
  • 1,836
  • 15
  • 35
1

In this answer it is shown without using L'Hopital's rule that for every $n>0$,

$$\lim_{x\to\infty}\frac{x^n}{e^x}=0. \tag{1}$$

We can use $(1)$ to show that for any $n>0$ and $a > 1$,

$$\lim_{x\to\infty}\frac{x^n}{a^x}=0. \tag{2}$$

To do this, write $a^x = e^{(\log a)x}$ where $\log a$ is positive since $a>1$. Then, if we set $y=(\log a)x$,

$$\frac{x^n}{a^x}=\frac{x^n}{e^{(\log a)x}}=\frac{1}{(\log a)^n}\frac{y^n}{e^y}.$$

When $x\to\infty$, we know that $y\to\infty$ because $\log a >0$. Therefore the behavior of $x^n/a^x$ follows from that of $y^n/e^y$ which is zero by $(1)$. Hence your limit is zero as it is a special case of $(2)$ where $n=5$ and $a=2$.

Axion004
  • 10,056