Evaluate $$\lim\limits_{n\to\infty} \frac{3+\sqrt{3}+\sqrt[3]{3}+\dots+\sqrt[n]{3}-n}{\ln n}.$$

We haven't been taught how to do asymptotes, I think this problem should be solved using Cesaro-Stolz; $$\lim\limits_{n\to\infty} \frac{a_n}{b_n}=\lim\limits_{n\to\infty} \frac{a_{n+1}-a_n}{b_{n+1}-b_n}$$ Note that this is my second question on this site.It's the first year i learned about limits so don't expect that this problem solves in a hard way.


Yes, you are correct, we can use Stolz-Cesaro theorem: $$\lim_{n\to\infty} \frac{\sum_{k=1}^n(\sqrt[k]{3}-1)}{\ln(n)}\stackrel{SC}{=}\lim_{n\to\infty} \frac{\sqrt[n+1]{3}-1}{\ln (n+1)-\ln(n)}= \lim_{n\to\infty} \frac{e^{\left(\frac{\ln(3)}{n+1}\right)}-1}{\ln\left(1+\frac{1}{n}\right)}.$$ Now note that (see Arnaud Mortier's comment below) $$\lim_{t\to 0}\frac{e^{t}-1}{t}=\lim_{t\to 0}\frac{e^{t}-e^0}{t-0}=(e^t)'_{t=0}=e^0=1$$ and $$ \lim_{t\to 0}\frac{\ln(1+t)}{t}=1=\lim_{t\to 0}\frac{\ln(1+t)-\ln(1+0)}{t-0}=(\ln(1+t))'_{t=0}=\frac{1}{1+0}=1.$$ Can you take it from here?

  • $\begingroup$ what do you mean by $\exp(\ln(3)/n)-1$ , btw i understand the denumerator $\endgroup$ – Robert Jan 15 '18 at 15:34
  • $\begingroup$ i never used that relation before, i can't do it, and btw $\lim_n \ln(1+\frac{1}{n})=0 ?$ $\endgroup$ – Robert Jan 15 '18 at 15:39
  • $\begingroup$ $$\lim_{t\to 0}\frac{e^{t}-1}{t}=1\quad\mbox{and}\quad \lim_{t\to 0}\frac{\ln(1+t)}{t}=1.$$ $\endgroup$ – Robert Jan 15 '18 at 15:44
  • 1
    $\begingroup$ You can find the final result by using $\lim_{t\to 0}\frac{e^{t}-1}{t}=1$ and $\lim_{t\to 0}\frac{\ln(1+t)}{t}=1$. $\endgroup$ – Robert Z Jan 15 '18 at 15:53
  • 1
    $\begingroup$ The two relations you mention can be seen as the definition of the derivative applied to the maps $e^t$ and ln$(1+t)$ at $0$. $\endgroup$ – Arnaud Mortier Jan 15 '18 at 15:53

I thought it might be of interest to present an approach that does not rely on the Stolz-Cesaro Theorem, but rather uses only (i) a pair of inequalities of the exponential function, (ii) a result from the integral test, and (iii) the squeeze theorem. To that end, we present two primers with references to their proofs.

PRIMER $(1)$:

In THIS ANSWER, I showed using on the limit definition of the integral test and Bernoulli's Inequality that the exponential function satisfies the inequalities

$$\bbox[5px,border:2px solid #C0A000]{1+x\le e^x\le\frac{1}{1-x}} \tag1$$

for $x<1$.

PRIMER $(2)$:

As a result from the integral test (SEE THIS), we have for a function $f(x)$ that is positive valued and monotonically decreasing for $x\ge N$

$$\bbox[5px,border:2px solid #C0A000]{\int_N^{n+1}f(x)\,dx\le \sum_{k=N}^n f(k)\le f(N)+\int_N^n f(x)\,dx}\tag2$$


Equipped with Primers $(1)$ and $(2)$, we are ready to proceed.

We begin with the function $f(x)=\sqrt[x]{3}-1=e^{\frac1x\log(3)}-1$ and write the sum of interest as $\sum_{k=1}^n f(k)=f(1)+\sum_{k=2}^n f(k)=2+\sum_{k=2}^n f(k)$.

Inasmuch as $\lim_{n\to \infty}\frac{2}{\log(n)}=0$, we analyze the limit $\lim_{n\to \infty}\frac{\sum_{k=2}^nf(k)}{\log(n)}$.

Applying the left-hand sides of $(1)$ and $(2)$, we see that

$$\begin{align} \sum_{k=2}^nf(k)&\ge \int_2^{n+1}f(x)\,dx\\\\ &=\int_2^{n+1}(e^{\frac1x \log(3)}-1)\,dx\\\\ &\ge \int_2^{n+1}\frac{\log(3)}{x}\,dx\\\\ &=\log(3)(\log(n+1)-\log(2))\tag3 \end{align}$$

Applying the right-hand sides of $(1)$ and $(2)$, we see that

$$\begin{align} \sum_{k=2}^nf(k)&\le (\sqrt3-1)+\int_2^{n}f(x)\,dx\\\\ &=(\sqrt 3-1)+\int_2^{n}(e^{\frac1x \log(3)}-1)\,dx\\\\ &\le (\sqrt 3-1)+\int_2^{n}\frac{\log(3)}{x-\log(3)}\,dx\\\\ &=(\sqrt 3-1)+\log(3)(\log(n-\log(3))-\log(2-\log(3)))\tag4 \end{align}$$

Dividing the expressions in $(3)$ and $(4)$ by $\log(n)$, and applying the squeeze theorem, we obtain the coveted result

$$\bbox[5px,border:2px solid #C0A000]{\lim_{n\to\infty}\frac{\sum_{k=1}^n(\sqrt[k]{3}-1)}{\log(n)}=\log(3)}$$

And we are done!

  • 1
    $\begingroup$ Very nice. The quality of contents and presentation of your answer is impressive. (+1) $\endgroup$ – Markus Scheuer Jan 15 '18 at 21:55
  • 1
    $\begingroup$ @MarkusScheuer Thank you for the nice words Markus! Very appreciative. $\endgroup$ – Mark Viola Jan 15 '18 at 22:11
  • $\begingroup$ I second the comment from @MarkusScheuer +1. IMHO this seems to be novel use of the idea behind integral test. $\endgroup$ – Paramanand Singh Jan 16 '18 at 16:22
  • $\begingroup$ @paramanandsingh Thank you my friend. And Happy New Year to you. $\endgroup$ – Mark Viola Jan 16 '18 at 16:32

As noted already you are correct and by Stolz-Cesaro theorem: $$\lim_{n\to\infty} \frac{\sum_{k=1}^n(\sqrt[k]{3}-1)}{\ln(n)}\stackrel{SC}{=}\lim_{n\to\infty} \frac{\sqrt[n+1]{3}-1}{\ln\left(1+\frac{1}{n}\right)}=\ln 3$$


$$\frac{\sqrt[n+1]{3}-1}{\ln\left(1+\frac{1}{n}\right)}= \frac{n\left(\sqrt[n+1]{3}-1\right)}{\ln\left(1+\frac{1}{n}\right)^n} \to \ln 3$$

since by definition of $e$

$$\ln\left(1+\frac{1}{n}\right)^n\to \ln e=1$$

and by Taylor's espansion

$$\sqrt[n+1]{3}=e^{\frac{\ln 3}{n+1}}=1+\frac{\ln 3}{n+1}+o\left(\frac1n\right)$$

$$n\left(\sqrt[n+1]{3}-1\right)=\frac{n}{n+1}\ln 3+o(1)\to\ln 3$$


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.