Why does $$\lim_{n \to \infty} \left(1+\frac{1}{n}\right)^n=e$$ but $$\lim_{n \to \infty} \left(1-\frac{1}{n}\right)^n=e^{-1}$$

Shouldn't the limits be the same since $\left(1+\frac{1}{n}\right) \to 1$?

  • 7
    $\begingroup$ No............. $\endgroup$ Dec 2 '17 at 20:11
  • 3
    $\begingroup$ $\lim_{n \to \infty} \left(1+\frac an\right)^n = e^a$, for $a \in \mathbb{R}$. $\endgroup$
    – TheSimpliFire
    Dec 2 '17 at 20:12
  • 1
    $\begingroup$ Generally speaking, $1^{+\infty}$ is an indeterminate form. The limit can be anything. $\endgroup$ Dec 2 '17 at 20:14
  • 2
    $\begingroup$ @amWhy I think there is a reason given for why the two limits should be the same. It is not a good reason and exposes a misunderstanding, and I think it is clear that OP is trying to understand why their intuition isn't working. $\endgroup$ Dec 2 '17 at 20:52
  • 1
    $\begingroup$ Oh, sure, @MarkBennet, now that you've edited the question to improve its quality. My comment was left prior to your edit, and was valid wrt the original answer. Don't chastise someone for observing a problem with the post, after you have edited it to improve it for the asker. $\endgroup$
    – amWhy
    Dec 2 '17 at 21:04

If the limits were the same, they would have to be $1$, since $$\left(1+\frac1n\right)^n>1$$ for all $n$, while $$\left(1-\frac1n\right)^n<1$$ for all $n$.

So if you expect the first limit to be $e$ and you know that $e>1$, you cannot expect the second one to be equal to it.

Edit: (to incorporate two of Paramanad Singh comments). One can obtain from Bernoulli's Inequality that $$\left(1+\frac1n\right)^n\geq2,\ \ n\geq1.$$ Which implies that $e\geq2$ as soon as one knows that the limits exists. This already shows that the limits cannot be the same. More interestingly, from $$ \left(1-\frac{1}{n}\right)^{n}=\dfrac{1}{\left(1+\dfrac{1}{n-1}\right)^{n-1}}\cdot\left(1-\frac{1} {n}\right),n>1 $$ one gets that $$\lim_{n\to\infty}\left(1-\frac{1}{n}\right)^{n}=e^{-1}.$$

  • 1
    $\begingroup$ (+1) very simple explanation of why the limits are different. $\endgroup$
    – zwim
    Dec 2 '17 at 20:51
  • $\begingroup$ (+1) More interestingly you can show that expression $(1+(1/n))^{n}$ is always greater than $2$ (via Bernoulli inequality ) and the expression $(1-(1/n))^{n}$ is trivially less than $1$ so that there is no way they can tend to the same limit. This deals with the part "you know that $e>1$" of your answer. $\endgroup$
    – Paramanand Singh
    Dec 3 '17 at 3:22

It's the same reason that you can't argue that $\lim_{n \to \infty}(1 + \frac{1}{n})^n = 1$, even though $1 + \frac{1}{n} \to 1$ and $1^n = 1$. The issue is that when we say that the limit of an expression is a certain value, we just mean the expression gets very close to that value - we make no promises about how fast.

When $n$ is very large, $1 - \frac{1}{n}$ is very nearly $1$ - but a number very nearly $1$, when raised to a very large power, can be very small. For example, $0.99^{10000}$ is zero to more than forty decimal places. So that exponent of $n$ can "pull" the value away from one - for $(1 - \frac{1}{n})^n$, it pulls it down, while it pulls $(1 + \frac{1}{n})^n$ up.


Just to give you some intuition, note that since $1-x < 1/(1+x)$ for $0<x<1,$

$$ (1-1/n)^n < \frac{1}{(1 + 1/n)^n}$$

for $n>1.$ Now the right side $\to 1/e.$ Thus if the limit of the left side exists, it has to be $\le 1/e.$


Expand both using the binomial theorem

$$\left(1+\frac 1n\right)^n=1^n+n\cdot\frac 1n\cdot1^{n-1}+\binom n2\left(\frac 1n\right)^21^{n-2}+\binom n3\left(\frac 1n\right)^31^{n-3}+\dots =$$$$=1+1+\frac {n(n-1)}{2n^2}+\frac {n(n-1)(n-2)}{6n^3}+\dots$$while $$\left(1-\frac 1n\right)^n=1-1+\frac {n(n-1)}{2n^2}-\frac {n(n-1)(n-2)}{6n^3}+\dots$$

Subtract the second from the first and you get $$2+\frac {n(n-1)(n-2)}{3n^3}+\dots$$ a sum of positive terms. So the difference between the two expressions is at least $2$. If the limits exist (they do), the difference between the limits must be at least $2$.

If you analyse the difference you will note that it is increasing with $n$ (but bounded) so even though the $(1\pm \frac 1n)$ part tends to zero, raising to the $n^{th}$ power in this case gives expressions which get further apart rather than closer together.


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.