# Prove that $\lim\limits_{n\rightarrow\infty} \left(1+\frac{1}{a_{n}} \right)^{a_{n}}=e$ if $\lim\limits_{n\rightarrow\infty} a_{n}=\infty$

What would be the nicest proof of the following theorem:

If $\lim\limits_{n \rightarrow \infty} a_{n} = \infty$, then $\lim\limits_{n \rightarrow \infty} \left(1 + \frac{1}{a_{n}} \right) ^ {a_{n} } = e$.

If $\lim\limits_{n \rightarrow \infty}b_{n} = 0$, then $\lim\limits_{n \rightarrow \infty} \left(1 + b_{n} \right) ^ {\frac {1} {b_{n}} } = e$.

I somehow failed to find a proof here on the website and in the literature.

• Use the fact $(1+1/n)^n$ is monotone. – Pedro Tamaroff Dec 6 '17 at 17:06
• Are the $a_n$ and $b_n$ integers or real numbers? – Arthur Dec 6 '17 at 17:10
• @Arthur If they were integers, the theorem would follow from the fact that the limit of any subsequence is in fact the limit of the whole sequence, I think. But they are not necessarily integers. – Theta Dec 6 '17 at 17:18
• "What would be the nicest proof", asked 35 minutes ago ... answer already accepted. Hmmm ... I am sure there will be more people coming with various interesting proofs, but ... leave the question open for at least a week before choosing the "nicest" ;) – rtybase Dec 6 '17 at 17:38
• @rtybase Haha, I should have added I meant nicest in terms of being elementary, then RRL's answer is hard to beat. :D – Theta Dec 6 '17 at 17:43

Hint:

$$\left(1 + \frac{1}{\lfloor a_n \rfloor+1} \right)^{\lfloor a_n \rfloor} \leqslant \left(1 + \frac{1}{a_n} \right)^{a_n} \leqslant \left(1 + \frac{1}{\lfloor a_n \rfloor} \right)^{\lfloor a_n \rfloor+1},$$

and

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

• That's infact the same way you follow to demonstrate the basic limits, I don't really see any big insight in this. – gimusi Dec 6 '17 at 17:32
• Just a way to apply the squeeze theorem when $a_n$ is not an integer, given the limit result for $(1 + 1/n)^n$ where $n$ is an integer. – RRL Dec 6 '17 at 17:36
• @RRL Thank you, a brilliant answer; somehow making $a_{n}$'s integers was what I was looking for (but didn't come up with a way to do this). – Theta Dec 6 '17 at 17:37
• @Theta: You're welcome. – RRL Dec 6 '17 at 17:52

Given that $\lim_{n\to\infty }b_n = \lim_{n\to\infty }\frac{1}{a_{n}}=0$ we have,

$$\lim_{n \to\infty} \left(1 + \frac{1}{a_{n}} \right) ^ {a_{n} } = \lim_{n \to\infty} \exp\left(\frac{\ln\left(1 + \frac{1}{a_{n}} \right)}{\frac1{a_{n}}} \right)= \lim_{h \to0} \exp\left(\frac{\ln\left(1 +h \right)}{h} \right)=e$$

similarly

$$\lim_{n \to\infty} \left(1 +b_n \right) ^ { \frac{1}{b_{n}}} = \lim_{n \to\infty} \exp\left(\frac{\ln\left(1 + b_n\right)}{b_n} \right)= \lim_{h \to0} \exp\left(\frac{\ln\left(1 +h \right)}{h} \right)=e$$

One does L'Hospital to show that $\lim_{x\rightarrow\infty}\left(1+\dfrac{1}{x}\right)^{x}=e$, then one does sequential characterisation of limit. Similar reasoning applied to $b_{n}\rightarrow 0$ case.

• As sequences are introduced earlier in a calculus course, supposse I don't have suffient knowledge on L'Hospital and functions in general, could you suggest a solution? – Theta Dec 6 '17 at 17:17
• @RRL has provided a fantastic answer. – user284331 Dec 6 '17 at 17:23

For the function-sequence criteria, they are direct consequence of the two basic limits:

$$\lim_{n \rightarrow \infty} \left(1 + \frac{1}{{n}} \right) ^ {{n} } = e$$

$$\lim_{n \rightarrow \infty} \left(1 + {n} \right) ^ {\frac {1} {{n}} } = e$$

by a simple change of variables.

NOTE

$$\lim_{m \rightarrow \infty} \left(1 + {m} \right) ^ {\frac {1} {{m}} } = e$$

is obtained from the first by the substitution $$m=\frac1n$$

MSE REFERENCE

Proving the limit of a function of a sequence is equal to the function of the limit of that sequence

• No, it's not a direct consequence by a simple change of variables. No one said anything about the $a_n$ being integers. – Arthur Dec 6 '17 at 17:10
• @Arthur: If you read this answer carefully, he never required $n$ to be an integer. – Bumblebee Dec 6 '17 at 17:14
• I'm referring to the function-sequence criteria. – gimusi Dec 6 '17 at 17:15
• When we read it carefully, the limits are not even right. – Raskolnikov Dec 6 '17 at 17:16
• @Arthur You can also take a look here: math.stackexchange.com/questions/554910/… – gimusi Dec 6 '17 at 17:18

Let's do the 1st one ...

One of the logarithmic inequalities says: $$\frac{x}{1+x} < \ln (1 + x) < x, \forall x > -1$$

Because $\lim\limits_{n\rightarrow \infty} a_n \rightarrow \infty$, then $a_n > 0$ and $\frac{1}{a_n} > 0$ from some $n$ onwards. So $$0<\frac{1}{a_n+1}=\frac{\frac{1}{a_n}}{1+\frac{1}{a_n}} < \ln\left(1 + \frac{1}{a_n}\right) < \frac{1}{a_n}$$ Given $e^x$ is ascending: $$1<e^{\frac{1}{a_n+1}} < 1 + \frac{1}{a_n} < e^{\frac{1}{a_n}}$$ and from some $n$ onwards $$e^{\frac{a_n}{a_n+1}} < \left(1 + \frac{1}{a_n}\right)^{a_n} < e$$ Squeeze theorem finishes the proof.

For $|t|<1$, note that $t-{1 \over 2} t^2 \le \log(1+t) \le t$ and so $|\log(1+t)-t| \le {1 \over 2} t^2$.

Hence for $|x|>1$ we have $|\log(1+{1 \over x})-{1 \over x}| \le {1 \over 2} ({1\over x})^2$ and so $|x\log(1+{1 \over x})-1| \le {1 \over 2} {1\over |x|}$.

Hence $\lim_{x \to \infty} x\log(1+{1 \over x}) =\lim_{x \to \infty} \log(1+{1 \over x})^x = 1$ from which it follows that $\lim_{x \to \infty} (1+{1 \over x})^x = e$.