2

How to find the $\lim_{n \to \infty} \left(\dfrac{(n+1)(n+2)\cdots(n+2n)}{n^{2n}}\right)^{1/n}$? I know how to find it for the indeterminate form of $1^{\infty}$ by converting it into $0/0$ form, but this cannot be converted into any known indeterminate form: $(0/0)^0$ . Can we convert it into an integral and then try to solve at as infinity is involved?. May someone help? Also please don't use any theorems or formula for limits except perhaps L'Hospital rule which I know. If you do please provide its proof as well.

Sloan
  • 2,303
Matt
  • 1,150

3 Answers3

7

First "trick:" convert to the exponential form.

$$ \left(\frac{\prod_{k=1}^{2n}(n+k)}{n^{2n}}\right)^{\frac{1}{n}} =\exp\left({\frac{1}{n}\ln\left(\frac{\prod_{i=1}^{2n}(n+k)}{n^{2n}}\right)}\right) $$ Now, let us focus on the exponent: $$ \frac{1}{n}\ln\left(\frac{\prod_{k=1}^{2n}(n+k)}{n^{2n}}\right) = \frac{1}{n}\ln\left(\prod_{k=1}^{2n}\frac{n+k}{n}\right) = \frac{1}{n}\ln\left(\prod_{k=1}^{2n}\left(1+\frac{k}{n}\right)\right) = \frac{1}{n}\sum_{k=1}^{2n}\ln\left(1+\frac{k}{n}\right) $$ At that point, it starts to really look like a Riemann sum, so let us massage it a little bit more: $$ \frac{1}{n}\ln\left(\frac{\prod_{k=1}^{2n}(n+k)}{n^{2n}}\right) = \frac{1}{n}\sum_{k=1}^{2n}\ln\left(1+\frac{k}{n}\right) = \frac{2}{2n}\sum_{k=1}^{2n}\ln\left(1+2\frac{k}{2n}\right) = 2\cdot\frac{1}{2n}\sum_{k=1}^{2n}f\!\left(\frac{k}{2n}\right) $$ for $f\colon [0,1]\to \mathbb{R}$ defined by $f(x)=\ln(1+2x)$.

We have$^{(\dagger)}$ $$\frac{1}{2n}\sum_{k=1}^{2n}f\!\left(\frac{k}{2n}\right)\xrightarrow[n\to\infty]{} \int_0^1 f = \frac{1}{2}(3\ln 3 -2)$$ and by continuity of the exponential your limit will be $$\left(\frac{\prod_{k=1}^{2n}(n+k)}{n^{2n}}\right)^{\frac{1}{n}} \xrightarrow[n\to\infty]{} e^{2\int_0^1 f} = e^{3\ln 3 -2)} = \frac{27}{e^2}.$$


$(\dagger)$ Here, we use the following theorem, which essentially follows from the definition of Riemann integration:

Theorem. (Riemann sums converge to the integral.) Let $f\colon [a,b]\to\mathbb{R}$ be a continuous function. Then $$ \frac{1}{n}\sum_{k=1}^n f\left(a+k\frac{b-a}{n}\right) \xrightarrow[n\to\infty]{} \int_a^b f $$ and $$ \frac{1}{n}\sum_{k=0}^{n-1} f\left(a+k\frac{b-a}{n}\right) \xrightarrow[n\to\infty]{} \int_a^b f. $$

This is a particular case of a slightly more general theorem ($f$ only needs to be Riemann integrable, and here we took a regular subdivision of $[a,b]$ in intervals of the same length $\frac{b-a}{n}$ instead of an arbitrary subdivision.)

In our case, $a=0$ and $b=1$, and we take a subdivision with $m\stackrel{\rm def}{=} 2n$ points: $$ \frac{1}{m}\sum_{k=1}^m f\left(\frac{k}{m}\right) \xrightarrow[m\to\infty]{} \int_0^1 f. $$

Clement C.
  • 67,323
  • How $\frac{n+k}{n^{2n}}$ becomes $\frac{n+k}{n}$? – Enrico M. Sep 09 '16 at 18:23
  • @FourierTransform It did not. $\frac{1}{n^{2n}} \prod_{k=1}^{2n} (n+k)$ became $\prod_{k=1}^{2n} \frac{n+k}{n}$. – Clement C. Sep 09 '16 at 18:24
  • Oh! Sorry. I read wrong. Great! – Enrico M. Sep 09 '16 at 18:25
  • I know this comment doesn't really add much, but I'm just wondering how you came up with this whole process. It seems so genius to me that I can only conclude that it took hours. – Polygon Sep 09 '16 at 18:28
  • Whenever I see something of the form $a_n^{b_n}$, my first instinct is that I don't understand powers, so I rewrite it $e^{b_n \ln a_n}$ (or $2^{b_n \log_2 a_n}$ if it's computer science). Then, after that, I don't understand products: if there are products inside the logarithm, I "convert" them into a sum of logarithms instead. (The rationale being, I know of more results on sum and series than on products). And a sum of the form $\frac{1}{n}\sum_k g(\frac{k}{n})$ begs for being recognized as a Riemann sum -- it does not always work, but it's worth trying for the 90% cases where it does. – Clement C. Sep 09 '16 at 18:32
  • very clever the link to the Riemann sum ! – G Cab Sep 09 '16 at 18:47
  • Can we do it without using integration ? – Matt Sep 12 '16 at 17:57
  • Also even if we use integration may you please elaborate how did you covert the summation to integration and why did you use that particular function only in a little detail. – Matt Sep 12 '16 at 17:59
  • @RaghavSingal I'll add details on the theorem (Riemann sums converging to the integral). As for getting rid of integration altogether: this is the simplest and most natural approach I can think of, and the most "elementary." Another approach (Fourier Transform's) relies on Stirling's approximation, which is a less elementary result (I'd say), and needs to be handled carefully (as noted in a comment to the answer there, in particular their equalities are wrong as stated). All in one, I'd deem the Riemann sum approach the simplest... – Clement C. Sep 12 '16 at 18:01
  • Okay, thank you very much, may you just explain the integration conversion part. – Matt Sep 12 '16 at 18:03
  • @RaghavSingal Added some details. – Clement C. Sep 12 '16 at 18:12
1

Maybe you can use the exponential form.

Forgive me for what I'm going to write, I know it's bad formalism but it may be effective:

$$\left(\frac{0}{0}\right)^0 = e^{0\cdot \ln\left(\frac{0}{0}\right)}$$

Then you may use hospital for the limit inside the logarithm and proceed...


Answer:-

$$ \text{let } y = \left(\dfrac{(n+1)(n+2)\cdots(n+2n)}{n^{2n}}\right)^{1/n}$$

$$\implies\log_e y = {1\over n}\log_e \left(\dfrac{(3n)!}{n! \times n^{2n}}\right)$$

$$\implies e^{{1\over n}\log_e \left(\dfrac{(3n)!}{n! \times n^{2n}}\right)} = y$$


$$\text{let } z = \log_e \left(\dfrac{(3n)!}{n! \times n^{2n}}\right)$$

$$\implies z = \log_e (3n)!- \log_e n! - 2n\log_e n$$

$$\color{red}{\implies z = 3n\log_e (3n) - 3n - (n\log_e n - n) - 2n\log_e n}$$

$$\implies z = 3n\log_e (n) + 3n\log_e 3 - 2n - n\log_e n - 2n\log_e n$$

$$\implies z = n(3\log_e 3 - 2) $$


$$\text{substituting the value of z in y}$$

$$y= e^{{1\over n}z}$$

$$\implies y= e^{3\log_e 3 - 2}$$

$$\text{Finally finding the limit}$$ $$\implies \lim_{n \to \infty} y = \lim_{n \to \infty} e^{3\log_e 3 - 2} = e^{3\log_e 3 - 2} = {e^{3\log_e 3}\over e^2} = {27\over e^2}$$

$$\color{red}{RED} \leftarrow \text{Stirling approximation}$$

Enrico M.
  • 26,114
0

HINT:

Let $A=\left(\dfrac{(n+1)(n+2)\cdots(n+2n)}{n^{2n}}\right)^{1/n}$

$$\implies\ln A=\dfrac1n\sum_{r=1}^{2n}\ln\left(1+\dfrac rn\right)$$

Now like Find $\lim\limits_{n \to \infty} \frac{1}{n}\sum\limits^{2n}_{r =1} \frac{r}{\sqrt{n^2+r^2}}$, let $2n=m$

$$\implies\ln A=2\cdot\dfrac1m\sum_{r=1}^m\ln\left(1+2\cdot\dfrac rm\right)$$

$$\text{As }\lim_{n \to \infty} \frac1n\sum_{r=1}^n f\left(\frac rn\right)=\int_0^1f(x)dx$$

$$\implies\ln A=2\int_0^1\ln(1+2x)\ dx$$

Now integrate by parts,

$$\int\ln(1+2x)\ dx=\ln(1+2x)\int dx-\int\left(\dfrac{d\{\ln(1+2x)\}}{dx}\cdot\int dx\right)dx$$

$$=x\ln(1+2x)-\int\dfrac x{1+2x}dx$$

Now for $\int\dfrac x{1+2x}dx,$ set $1+2x=y$ to ultimately find that

$$\ln A=2\left(\dfrac32\ln(1+2)-1\right)=3\ln 3-2=\ln\dfrac{3^3}{e^2}\text{ as }\ln(e)=1$$