4

I'm in high school I always wonder why does the below limit

$$ \lim_{x\to0}(1+x)^{\frac{1}{x}} \tag{1}$$

necessarily converges to a finite number the places where I searched in the internet I found that it's a definition

but why does it really needs to converge to a specific number I searched for it in many places some of the explanations which I found on quora ,YouTube and etc

some said that we can just verify it by plugging in some small number

some were kind of circular proof only like they took $\ln$ in expression (1) then they used L hospitals rule

but as we know while finding the derivative of the log function we ourself assume that the limit(1) is a Constant and then we proceed.

Also another way I thought was to consider $(1+\frac{1}{x})^{x}$ as a function I differentiated it and we clearly see as $x \to \infty$ it's slope tends to 0 thus it must be a constant

but again in between I had to differentiate the log function so I want to ask that is there any way to really show the (1) approaches a finite number.

Or do I need some more mathematical knowledge to do it?

Gary
  • 36,640
Anonymous
  • 537
  • 1
    @peterwhy log's differentiation are again being used I've mentioned this point in my question – Anonymous Apr 22 '24 at 16:52
  • 2
    It would help if you used punctuation. You would need to start with a definition of what $a^b$ means, before you can talk about circularity in arguments. – copper.hat Apr 22 '24 at 17:01
  • @copper.hat could you please elaborate your point a bit more. – Anonymous Apr 22 '24 at 17:16
  • I don't understand your question. While finding the limits of a logarithmic function, what do we take constant? – Gwen Apr 22 '24 at 17:49
  • 3
    @AnshuGupta How do you define $(1+x)^\frac{1}{x}$? – copper.hat Apr 22 '24 at 18:08
  • 1
    I don't think this is a duplicate of what peterwhy linked, because it is one of the standard ways to define exponential function by first proving $\lim_n(1+1/n)^n$ exists, so even though there are ways to define exponential function without the existence of the limit, it really misses the point that highschool student is asking a very reasonable question that can be answered without heavy machinery like Taylor series or differential equations, which otherwise needs to be used if you want the existence of exponential function. – Ennar Apr 22 '24 at 18:47
  • @copper.hat sorry I'm not able to get on what point are you trying refer me. (Also isn't my question really different from the question you mention because answers in that questions have used L hospitals rule which I have mentioned in my question) – Anonymous Apr 22 '24 at 19:32
  • 1
    Slope tending to 0 doesn't imply convergence to a constant - consider $\sqrt x$. – Karl Apr 22 '24 at 20:07
  • 2
    @AnshuGupta In order to prove something, one starts from a definition of what the quantity $(1+x)^\frac{1}{x}$ means. – copper.hat Apr 22 '24 at 20:17
  • How do you feel about defining $e$ as the solution $X=e$ in $\int_1^X \frac{dx}x=1$ ? – PM 2Ring Apr 23 '24 at 12:34
  • related https://math.stackexchange.com/q/59313 – user Apr 30 '24 at 15:33
  • related https://math.stackexchange.com/q/3825609 – user Apr 30 '24 at 15:34

4 Answers4

7

I'll start off with a disclaimer: this will be a long and technical answer. I'll take a risk that this might be much more technical than you asked for, but it can't be avoided if you are really looking for understanding this limit beyond just techniques to calculate it. If you simply want to know how it's calculated, I refer you to the link that peterwhy provided in the comments.

In fact, that's how I recommend anyone to calculate it. In this answer I do something else.

What I'll cover is usually done at university level at a first course in mathematical analysis, but I'll just stick to the stuff we need for this particular limit and will provide links that you can use if you are not familiar with something I use.

The whole point of this answer is to convince you that $\lim_{x\to 0}(1+x)^{\frac 1x}$ exists essentially because $\lim_n(1+\frac 1n)^n$ exists and those two are equal. If you are happy with just that as an explanation, you can read section 2. and ignore the rest of the answer.


1. Intro

First of all, as pointed out in the comments, to even consider the limit $\lim_{x\to 0}(1+x)^{\frac 1x}$, you'd need to define what $(1+x)^{\frac 1x}$ is for all real numbers $x$. What we use here is definition of $a^b = e^{b\ln a}$ (I discuss why we use such a definition in this answer if you are interested), so $(1+x)^{\frac 1x} = e^{\frac 1x \ln(1+x)}.$ From there, one can simply find out the limit just like I mentioned at the start, but I don't think this is a satisfying answer if one is not familiar with how exponential function is defined in terms of limits and the whole process of showing that the definition is good.

However, I will not use $(1+x)^{\frac 1x} = e^{\frac 1x \ln(1+x)},$ I'll just take for granted that such a function exists, but I'll use the same ideas one uses to establish definition of exponential function. Concretely, one can define $e^x=\lim_n(1+\frac xn)^n$ for any real $x$ (this is not the only way one can define it), but I'll only consider the special case when $x=1$ (and later $x = -1$), which you are probably familiar with: $\lim_n(1+\frac 1n)^n = e\approx2.718,$ and then relate this to your limit.


2. Existence of $\lim_n(1+\frac 1n)^n$

So, let's start with a proof that $\lim_n(1+\frac 1n)^n$ exists. We do it by using monotone convergence theorem, i.e. we want to show that the sequence $(1+\frac 1n)^n$ is increasing and bounded above, therefore it is convergent.

To show that it is increasing we can use AM-GM inequality:

\begin{align}\sqrt[n+1]{(1+\frac 1n)^n} = \sqrt[n+1]{1\cdot(1+\frac 1n)\ldots(1+\frac 1n)}&\leq\frac{1+(1+\frac 1n)+\ldots+(1+\frac 1n)}{n+1}\\ &= \frac{1+n(1+\frac 1n)}{n+1} = 1+\frac 1{n+1}.\end{align} Applying $(n+1)$-st power to both sides gives us $(1+\frac 1n)^n \leq (1+\frac 1{n+1})^{n+1}$ for all positive integers $n$.

To show that it is bounded above we use binomial theorem:

\begin{align}(1+\frac 1n)^n = \sum_{k=0}^n\binom nk\frac 1{n^k} &= \sum_{k=0}^n\frac 1{k!}\frac{n(n-1)\ldots(n-k+1)}{n^k}\leq \sum_{k=0}^n\frac 1{k!} \\&= 1 + (1+\frac 1{2!}+\frac 1{3!}+\ldots+\frac 1{n!}) \\&\leq 1 + (1+\frac 1{2}+\frac 1{4}+\ldots+\frac 1{2^{n-1}}) = 1+\frac{1-\left(\frac 12\right)^n}{1-\frac 12}\\ &\leq 1+\frac{1}{1-\frac 12} = 3 \end{align}

where we also used formula for sum of finite geometric sequence and this finishes the proof that $(1+\frac 1n)^n$ converges to some number less than or equal to $3$. This is how one can define the number $e$: it is equal to that limit.


3. Towards $\lim_{x\to 0}(1+x)^{1/x}$

We are now ready to come back to the limit you asked about. You had a good idea by considering $\lim_{x\to\infty}(1+\frac 1x)^x$ instead, but it's not the full picture. You see, $\frac 1x$ approaches $0$ when $x$ goes to $\infty$, but only from above, so $\lim_{x\to\infty}(1+\frac 1x)^x = \lim_{x\to 0^+}(1+x)^{\frac 1x}$, where notation $x\to 0^+$ means that we are only considering positive $x$ approaching $0$ (see one-sided limit).

To calculate the limit you want we also need to see what happens when $x$ approaches $0$ from bellow. We can do it by calculating $\lim_{x\to -\infty}(1+x)^{\frac 1x} = \lim_{x\to 0^-}(1+\frac 1x)^x$. If this limit turns out to be the same as the last one, then we can conclude that $\lim_{x\to 0}(1+x)^{\frac 1x}$ exists and is equal to those two limits.


4. Existence of $\lim_{x\to\infty}(1+\frac 1x)^x$

To calculate a limit of a function at $\infty$, we consider all sequences $(x_n)$ of real numbers that tend to $\infty$ and need to show that $\lim_n(1+\frac 1{x_n})^{x_n}$ exists for all of them and all of those limits are equal (see limit of a function).

The idea here is that we already know that this limit exists in the case of $x_n = n$ and is equal to $e$. This is also true for any such sequence $(x_n)$ of integers, because in that case the sequence $(1+\frac 1{x_n})^{x_n}$ is a subsequence of the convergent sequence $(1+\frac 1n)^n$, and any subsequence of a convergent sequence converges to the same limit. We can actually use that to conclude that it's true for all sequences $(x_n)$ that tend to $\infty$, not just sequences of integers.

So let $(x_n)$ be such a sequence and for every $x_n$, and let $a_n$ be the unique integer such that $a_n \leq x_n < a_n+1$. We then have $$1+\frac 1{a_n+1} \leq 1+ \frac 1{x_n} \leq 1 + \frac 1{a_n}$$ and $$\left(1+\frac 1{a_n+1}\right)^{a_n} \leq \left(1+ \frac 1{x_n}\right)^{x_n} \leq \left(1 + \frac 1{a_n}\right)^{a_n+1}.\tag{1}$$

The limit of the RHS sequence is $\lim_n(1 + \frac 1{a_n})^{a_n+1} = \lim_n (1 + \frac 1{a_n})^{a_n}\cdot\lim_n(1+\frac 1{a_n}) = e\cdot 1 = e$ and the limit of the LHS sequence is $\lim_n(1 + \frac 1{a_n+1})^{a_n} = \lim_n (1 + \frac 1{a_n+1})^{a_n+1}\cdot\lim_n(1+\frac 1{a_n})^{-1} = e\cdot 1^{-1} = e.$

By the squeeze theorem from $(1)$ it follows that $\lim_n(1+ \frac 1{x_n})^{x_n} = e$ for any sequence $(x_n)$ that tends to $\infty$, and therefore $\lim_{x\to\infty}(1+\frac 1x)^x = e$.


5. Existence of $\lim_{x\to -\infty}(1+\frac 1x)^x$.

I will make this as short as possible and just refer to what we previously proved. First note that $$\lim_{x\to -\infty}\left(1+\frac 1x\right)^x =\lim_{x\to \infty}\left(1-\frac 1x\right)^{-x} = \frac 1{\lim_{x\to \infty}\left(1-\frac 1x\right)^{x}}.$$

Now, by the same type of arguments as previously, $\lim_{x\to \infty}(1-\frac 1x)^{x}$ will be equal to $\lim_n(1-\frac 1n)^n$. To see that $\lim_n(1-\frac 1n)^n$ exists, we do the same as for $\lim_n(1+\frac 1n)^n$, we prove that $(1-\frac 1n)^n$ is increasing sequence by AM-GM inequality just like in section 2. and is bounded above since $(1-\frac 1n)^n < 1$ for all positive integers $n$. Therefore, $\lim_n(1-\frac 1n)^n$ exists and let's denote that limit as $L$.

To calculate $L$ let's do the following:

$$e\cdot L=\lim_n\left(1+\frac 1n\right)^n\cdot \lim_n\left(1-\frac 1n\right)^n = \lim_n\left(1-\frac 1{n^2}\right)^n = \lim_n\left(\left(1-\frac 1{n^2}\right)^{n^2}\right)^{\frac 1n} \stackrel{(*)}{=} 1$$ and therefore $L = e^{-1}$. This proves that $\lim_{x\to -\infty}\left(1+\frac 1x\right)^x =\lim_{x\to \infty}\left(1-\frac 1x\right)^{-x} = \frac 1{\lim_{x\to \infty}\left(1-\frac 1x\right)^{x}} = e.$


Let's elaborate on $(*)$. There is a lemma that states that if $a>0$ is constant, then $\lim_n\sqrt[n]a = 1$. This can be proven directly from binomial theorem and squeeze theorem without use of exponential function. Furthermore, since $(1-\frac 1{n^2})^{n^2}$ is increasing and bounded above by $1$, we have $$\left(1-\frac 1{2^2}\right)^{2^2} \leq \left(1-\frac 1{n^2}\right)^{n^2} \leq 1,\ n\geq 2$$ which implies $$\sqrt[n]{\left(1-\frac 1{2^2}\right)^{2^2}} \leq \sqrt[n]{\left(1-\frac 1{n^2}\right)^{n^2}} \leq \sqrt[n]{1},\ n\geq 2,$$ so taking the limit proves $(*)$ by squeeze theorem.


With this we are done, $\lim_{x\to 0^+}(1+x)^{\frac 1x} = \lim_{x\to 0^-}(1+x)^{\frac 1x} = e$, so $\lim_{x\to 0}(1+x)^{\frac 1x} = e$.

Ennar
  • 24,364
  • When you were computing $eL$ is it always correct to substitute the that $e=\lim_{n\to\infty}(1+\frac{1}{n})^{n}$ because I've encountered a problem $$\lim_{n \to \infty}\frac{e^n}{(1+(\frac{1}{n}))^{n^2}}$$and directly substituting the def of e gives 1 as the answer but the answer is acutually$√e$ – Anonymous Apr 23 '24 at 08:30
  • @AnshuGupta, you have a good eye, I used something more special here, but it was getting late so I skipped over it. You can't just substitute however you want. General rule is that if $a_n$ and $b_n$ converge to $a$ and $b$ respectively, then $a_n^{b_n}$ converges to $a^b$. It's important that $a$ and $b$ are finite here, you know that it doesn't work in case $a = 1$ and $b = \infty$ :) I'll do it in other way, purposely avoiding exponential function (which is important since all of these ideas can be used to establish definition of exponential function, as I mentioned). – Ennar Apr 23 '24 at 10:57
  • I am grateful for your support and time ,thanks a lot ! – Anonymous Apr 23 '24 at 14:55
  • @AnshuGupta, you are welcome, I'm glad a high school student is this interested in understanding mathematics, rather than just calculations. If you need more help with anything I've written here, I'm happy to explain. – Ennar Apr 23 '24 at 15:00
5

This illustration is traditional in French mathematics education, might go back to Cauchy..

I'm taking $x >0$ giving $1+x > 1$

One thing is automatic, $$ \frac{1}{x} < 1 + \frac{1}{x} $$ so that $$ \left(1 +x \right)^{ \frac{1}{x} } < \left(1 +x \right)^{1 + \frac{1}{x} } $$

Here is the illustration part: the left hand side "starts" low. Say we are talking about $x=1,$ our starting point on the left is $2.$ The starting point on the right is $4.$

So far so good: now let $x$ decrease from $1$ towards $0.$ The left hand side increases and the right hand side decreases. They crash into each other, like two trains going in opposite directions on the same track.

$$ \begin{array}{rrr} x & LHS & RHS \\ 1 & 2 & 4 \\ 0.5 & 2.25 & 3.375 \\ 0.333 & 2.370 & 3.16 \\ 0.25 & 2.441 & 3.052 \\ 0.2 & 2.488 & 2.986 \\ 0.166 & 2.521 & 2.942 \\ 0.143 & 2.546 & 2.910 \\ 0.125 & 2.566 & 2.887 \\ 0.111 & 2.581 & 2.868 \\ 0.1 & 2.594 & 2.853 \\ 0.091 & 2.604 & 2.841 \\ 0.083 & 2.613 & 2.831 \\ 0.077 & 2.621 & 2.822 \\ 0.071 & 2.627 & 2.815 \\ 0.067 & 2.633 & 2.808 \\ 0.062 & 2.638 & 2.803 \\ 0.059 & 2.642 & 2.798 \\ 0.056 & 2.646 & 2.793 \\ ... &... & ...\\ ... &... &... \\ 0.014 &2.699 & 2.737 \\ 0.0141 &2.6993 & 2.7374 \\ 0.0138 &2.6996 &2.7371 \\ 0.1369 & 2.6998 & 2.7368 \\ 0.0135 & 2.7001 & 2.7366 \\ 0.0133 & 2.7003 & 2.7363 \\ & & \\ & & \\ \end{array} $$

I should add: it is immediate that the ratio of the LHS and the RHS goes to $1$ while the difference goes to $0.$

Added: I've not put in any specific proof that the LHS increases and the RHS decreases as $x$ decreases. Once you believe those, it follows that every value in the LHS is smaller than every value on the RHS

Will Jagy
  • 146,052
  • 1
    +1, never saw this argument before :) – Ennar Apr 22 '24 at 18:52
  • 1
    @Ennar thank you; I learned it in high school in the U.S. long ago. There was a question on terminology by a French guy https://math.stackexchange.com/questions/4391704/what-are-suites-adjacentes-called-in-english . I asked Jean Marie about it in comment below his answer to a different question: https://math.stackexchange.com/questions/4393211/how-to-show-that-euler-constant-lies-in-0-1 – Will Jagy Apr 22 '24 at 19:33
  • 1
    How do you show that RHS goes down and LHS goes up with $x$ going down ? For $x=1/n$ it can be done by induction. This is a common way of defining the number $e$. – Ryszard Szwarc Apr 22 '24 at 20:30
  • I am French and never saw that (which is interesting). – Claude Leibovici Apr 23 '24 at 06:06
  • +1 Nice. ${}{}$ – copper.hat Apr 23 '24 at 20:45
4

The function $\lim_{x\to \infty}(1+\frac1x)^x$ will converge if $\lim_{n\to\infty}(1+\frac1n)^n$ does for natural $n$. By the binomial theorem $$(1 + \frac1n)^n = \sum_{k=0}^n\tbinom nk\frac1{n^k}$$ and so the $k$-th term is $$\frac1{k!}\frac{n(n-1)(n-2)\cdots(n-k+1)}{n^k}=\frac1{k!}(1-\frac1n)(1-\frac2n)(1-\frac3n)\cdots (1-\frac{k+1}n)$$

Then letting $n \to \infty$ we have $$\lim_{n\to \infty}\sum_{k=0}^n\tbinom nk\frac1{n^k}=\sum_{k=0}^\infty \frac1{k!}$$

and since the partial sums are all positive and increasing we need simply to show it's bounded to prove convergence by the monotone convergence theorem. For this we note that for $k\geq 0$ $$\frac1{k!} \leq \frac1{2^{k-1}}$$ and so we have $$\sum_{k=0}^\infty \frac1{k!} \leq \sum_{k=0}^\infty \frac1{2^{k-1}} = 2+\frac1{1-\frac12}=4$$ which shows that it is bounded and we are done.

CyclotomicField
  • 11,925
  • 1
  • 14
  • 32
1

Let $f(x)\equiv (1+x)^{1/x}$, we have $f(x)=exp\{log[(1+x)^{1/x}]\}$. Since $exp\{.\}$ is continuous, we have

$$\lim_{x\rightarrow 0} f(x) = exp\{\lim_{x\rightarrow 0} log[(1+x)^{1/x}]\} = exp\{\lim_{x\rightarrow 0} \frac{log(1+x)}{x}\}$$

Note that $log(1+x) = x - \frac{x^2}{2}+[\textit{higher order terms}]$, and hence the limit inside the large bracket is 1. This concludes $\lim_{x\rightarrow 0} f(x) = e \approx 2.718$.

WHLin
  • 234