-4
$\begingroup$

Limit of $(n^2+n)^{1/2} - (n^3+n^2)^{1/3}$ as $n\to\infty$

Help me please solving this.

  • 3
    Limit $n \to ?? $2017-01-07

2 Answers 2

1

The approach is standard, using the fact that $a-b = \dfrac {a^k - b^k} {a^{k-1} + a^{k-2}b + \dots + a b^{k-2} + b^{k-1}}$. In our case, $a = \sqrt{n^2 + n} = \sqrt[6]{(n^2 + n)^3}$ and $b = \sqrt[3]{n^3 + n^2} = \sqrt[6]{(n^3 + n^2)^2}$, so we shall take $k=6$. It follows that

$$\sqrt{n^2 + n} - \sqrt[3]{n^3 + n^2} = \sqrt[6]{(n^2 + n)^3} - \sqrt[6]{(n^3 + n^2)^2} = \frac {(n^2 + n)^3 - (n^3 + n^2)^2} {\sum \limits _{i=1} ^6 \sqrt{n^2 + n}^{6-i} \sqrt[3]{n^3 + n^2}^{i-1}} = \\ \frac {(n^6 + 3n^5 + 3n^4 + n^3) - (n^6 + 2n^5 + n^4)} {n^5 \sum \limits _{i=1} ^6 \sqrt{1 + \frac 1 n}^{6-i} \sqrt[3]{1 + \frac 1 n}^{i-1}} = \frac {1 + 2 \frac 1 n + \frac 1 {n^2}} {\sum \limits _{i=1} ^6 \sqrt{1 + \frac 1 n}^{6-i} \sqrt[3]{1 + \frac 1 n}^{i-1}} \to \frac {1 + 0 + 0} {\sum \limits _{i=1} ^6 1 \cdot 1} = \color{red} {\frac 1 6} .$$

  • 0
    I always give +1 for the hard way :D2017-01-07
  • 0
    @SimpleArt: I don't know whether it's hard or not, but it surely is elementary (high-school level, to be precise). Your approach requires familiarity with asymptotics, which means more conceptual layers, which are not really needed to solve the problem.2017-01-07
  • 0
    Difficulty to me lies not in content for this, but the tedious-ness and prone-ness to error, which is why I try to avoid :-/ (and whether or not something is not needed does not matter, if you can use a tool, use it)2017-01-07
  • 0
    Thanks for the answer tho your highschool and mine were obviously very different.2017-01-07
2

We may apply binomial expansion to see that

$$(n^2+n)^{1/2}=n+\frac12+\mathcal O(n^{-1})$$

$$(n^3+n^2)^{1/3}=n+\frac13+\mathcal O(n^{-1})$$

So as $n\to\infty$,

$$(n^2+n)^{1/2}-(n^3+n^2)^{1/3}\sim\color{red}{\frac16}+\mathcal O(n^{-1})$$


For explanation, we have the binomial expansion:

$$(a+b)^k=a^k+ka^{k-1}b+\mathcal O(a^{k-2}b^2)$$

If you look at your regular old binomial expansion, the first number is obviously $a^k$ and the second is always the exponent times $a^{k-1}b$, and it goes on with different coefficients on and on. However, its a little harder figuring out the coefficients beyond this point, so we just sum it all up with $\mathcal O(a^{k-2}b^2)$, which basically says it is equal to that multiplied by some finite constant. Terms beyond that are even smaller than $a^{k-2}b^2$, so they get combined into it all.

For example,

$$(n^2+n)^{1/2}=(n^2)^{1/2}+\frac12(n^2)^{-1/2}n+\mathcal O((n^2)^{-3/2}n^2)\\=n+\frac12+\mathcal O(n^{-1})$$

Notice that regardless of constant, as $n\to\infty$, $\mathcal O(n^{-1})\to0$, so this method is very simple and effective for these problems.


You could alternatively note that

$$(n^2+n)^{1/2}-(n^3+n^2)^{1/3}=\frac{(1+x)^{1/2}-1}x-\frac{(1+x)^{1/3}-1}x$$

where $x=1/u$. As $u\to\infty$, $x\to0$, so

$$\lim_{x\to0}\frac{(1+x)^{1/2}-1}x-\frac{(1+x)^{1/3}-1}x=\left.\frac d{dx}(1+x)^{1/2}-(1+x)^{1/3}\right|_{x=0}$$

  • 0
    I am not familiar with the $O(n^{-1})$ notation. What is it callled ?2017-01-07
  • 2
    [Big O-notation](https://en.m.wikipedia.org/wiki/Big_O_notation).2017-01-07
  • 1
    @WiCK3DPOiSON It is called "O-notation". Basically, everything after the first two terms is smaller $n^{-1}$ times a constant.2017-01-07
  • 0
    Thanks but im pretty sure we didnt study this way. Its completely new for me . Any other way like multiplying to get fraction ?2017-01-07
  • 0
    @hogogog You could, and that would result in dealing with some pretty hefty fractions. Since this method is so easy, would you like me to explain a bit more?2017-01-07
  • 0
    Sure , why not . Thanks btw.2017-01-07
  • 0
    @hogogog No problem, and your welcome! I added more explanation to my answer for you :-)2017-01-07
  • 1
    Wow ., this is brilliant and I really dont know why we havent still done this . Really appreciated and didnt know i would get an answer so quick even tho i didnt mention n-> inf at the question .2017-01-07
  • 0
    @hogogog Well, the question is pretty trivial if $n$ isn't going to infinity. Its the only limit here you'd have problems with. And yes! It is brilliant! But sadly, I don't think you are going to learn it, at least not soon. If you do learn it, you'll probably learn it during "Taylor's theorem" section, which may also interest you :)2017-01-07
  • 0
    This should help whenever i have roots , right ?2017-01-07
  • 0
    @hogogog especially for roots, which are a super pain :-(2017-01-07
  • 0
    And could u also tell me with what would u multiply this or would you do something different ( if u had to do some beginners way and clearly longer ) coz Im not really sure my proffesors would tolerate this way .Just a tip i need , i dont want to trouble u.2017-01-07
  • 0
    @hogogog it's no biggie. I'd recommend multiplying by the following:$$\frac{\sqrt a+\sqrt[3]{b}}{\sqrt a+\sqrt[3]{b}}\frac{a+\sqrt a\sqrt[3]{b^2}+\sqrt[3]{b^4}}{a+\sqrt a\sqrt[3]{b^2}+\sqrt[3]{b^4}}$$where a and b are the things under the square root and cube roots.2017-01-07
  • 0
    Pffff , didnt expect this2017-01-07
  • 0
    Anyway , thanks for help . It troubled me for a day and this was my last hope . :)2017-01-07
  • 0
    @hogogog ...if you can use L'Hospital's rule:$$(n^2+n)^{1/2} - (n^3+n^2)^{1/3}=\frac{(1+\frac1n)^{1/2}-(1+\frac1n)^{1/3}}n$$2017-01-07
  • 0
    Havent heard about that either. Really wierd there is no simpler solution ( 0 notation is simple but positive we didnt do it ) nor some special trick .2017-01-07
  • 0
    @hogogog There is. Please see the end of my newly edited answer.2017-01-07
  • 0
    Huh this is whole new level for me2017-01-07
  • 0
    @hogogog :D Calculus is an interesting subject. You have limits, derivatives, and integrals. And then about 500 different ways to do each problem, many you will not see at first, but shall become obvious after you learn certain methods.2017-01-07
  • 0
    Probably . We usually did recursive series , sandwich theorem , cesaro - stolz theroem and beginner ways of finding limits usually with some tricksters to notice what to do. Nothing with binomial , derivations and Lhospital .2017-01-07
  • 0
    @hogogog :D Well, glad to help and glad to know you learned some interesting things!2017-01-07
  • 0
    Really appreciated.2017-01-08