5

On the way to show that the function $f(x)=\lim_{n\to{\infty}}\left(\frac{x}{n}+1\right)^n$ has this property: $f(a+b)=f(a)\cdot f(b)$, a math professor explained me that: $\lim_{n\to{\infty}}\left(\frac{ab}{n^2}+\frac{a+b}{n}+1\right)^n=\lim_{n\to{\infty}}\left(\frac{a+b}{n}+1\right)^n$,

because $\lim_{n\to{\infty}}(\frac{ab}{n^2})=0$, so we can delete the $\frac{ab}{n^2}$ part from the function.

But he didn't explain to me why we don't delete the $\frac{a+b}{n}$ part as $\lim_{n\to{\infty}}(\frac{a+b}{n})$ is equal to $0$ as well.

So there is still some mystery is this resolution for me. Please help to understand why it has to be like that.

-Edit: if possible I would like a solution not far from the way that my professor showed me. Or else please show me first why he induce me in a wrong way. I also don't want a solution that use the exp function because I'm only on the way to define it.

-Edit: the answer that was validated in this question (in the link below) do not satisfy me as the $a$ change to become $a\to\infty$ at the end of the demonstration that would break the sandwich order presented at the begening of it: Proving multiplicative property of the exponential function from its limit definition

-Edit: someone pointed out in the comment that by using the little o notation we can show that the term $\frac{ab}{n^2}$ could be neglected in $\lim_{n\to{\infty}}\left(\frac{ab}{n^2}+\frac{a+b}{n}+1\right)^n$ but from this rose a problem very similar to the problem that rose in the proof that the professor gave me. It is that with the little o notation we could also show that the $\frac{a+b}{n}$ term could be neglected in $\lim_{n\to{\infty}}\left(\frac{a+b}{n}+1\right)^n$

-Edit: the answer that was validated in the link below is not satisfying me because it use the property of $e^x$ in the proof as something aquired and as I said before: I'm only on the way to define $e^x$ so I don't want the use of it in the proof that I search by any mean. I also notice something that I think is a mistake in this proof and if you are interested you can watch the comment that I gave: Prove properties of exponential function using a limit definition

lazare
  • 387
  • 1
  • 9
  • Comments have been moved to chat; please do not continue the discussion here. Before posting a comment below this one, please review the purposes of comments. Comments that do not request clarification or suggest improvements usually belong as an answer, on [meta], or in [chat]. Comments continuing discussion may be removed. – Xander Henderson Nov 23 '24 at 13:00

3 Answers3

4

First a little lemma: Suppose $x_n \to 0$ then $\lim_n (1+{x_n \over n})^n =1$. Suppose $\epsilon \in (0,1)$ and $|x_n| < \epsilon$. Then \begin{eqnarray} |(1+{x_n \over n})^n -1| &=& |\sum_{k=1}^n \binom{n}{k} {1 \over n^k}x_n^k | \\ &\le& \sum_{k=1}^n \binom{n}{k} {1 \over n^k}|x_n|^k \\ &\le& \sum_{k=1}^n \binom{n}{k} {1 \over n^k}\epsilon^k \\ &=& \sum_{k=1}^n {n! \over (n-k)!k! n^k} \epsilon^k \\ &=& \sum_{k=1}^n {n\cdots(n-k+1) \over n\cdots n}{1 \over k!} \epsilon^k \\ &\le& \sum_{k=1}^n \epsilon^k = {\epsilon \over 1-\epsilon} \end{eqnarray} from which the result follows.

Then $1+{a+b \over n} + {ab \over n^2} = (1+{a+b \over n}) {1+{a+b \over n} + {ab \over n^2} \over 1+{a+b \over n}} = (1+{a+b \over n})(1+{1 \over n}{{ab \over n} \over {1+{a+b\over n}}})$. Setting $x_n = {{ab \over n} \over {1+{a+b\over n}}}$ and applying the lemma shows that $\lim_n (1+{a+b \over n} + {ab \over n^2})^n = \lim_n(1+{a+b \over n})^n $, which was the desired result.

copper.hat
  • 178,207
  • 1
    I was going to mention your lemma in a slightly different form as a comment to question, but before that saw your answer. +1 – Paramanand Singh Jul 04 '24 at 14:08
  • 1
    In particular I have used it on many occasions on this site and a sample is available at https://math.stackexchange.com/a/3000717/72031 dealing with complex exponential. – Paramanand Singh Jul 04 '24 at 14:15
  • @ParamanandSingh That's funny, I would not have immediately recognised the lemma as equivalent had you not brought it to my attention! – copper.hat Jul 04 '24 at 18:49
  • $(0,1)$ is an interval? – lazare Jul 05 '24 at 10:35
  • @lazare yes. ${}$ – copper.hat Jul 05 '24 at 13:46
  • Can you explain a bit more why you find $|(1+{x_n \over n})^n -1| \le \sum_{k=1}^n \binom{n}{k} {1 \over n^k}\epsilon^k$ it's a bit teleported. – lazare Jul 05 '24 at 19:35
  • 1
    @lazare I elaborated a bit... – copper.hat Jul 05 '24 at 22:52
  • Thank you for your developement. Is there a reason why in this: $\lim_n (1+{x_n \over n})^n =1$ it is not precised if $n$ goes to infinity or something else? – lazare Jul 06 '24 at 13:51
  • @lazare I do not understand what you are asking. $\lim_n (1+{x_n \over n})^n =\lim_{n \to \infty} (1+{x_n \over n})^n $. – copper.hat Jul 06 '24 at 16:00
  • Somehow you understood that what I asked because you replied well to it : $\lim_n (1+{x_n \over n})^n =\lim_{n \to \infty} (1+{x_n \over n})^n$ – lazare Jul 06 '24 at 16:04
  • So inside $\lim_{n \to \infty} (1+{x_n \over n})^n$ there is $(\frac{x_n} {n})$ witch contain $x_n\to0$ and $n\to\infty$ something that goes to $0$ divided by something that goes to $\infty$ that's not an indeterminate case? – lazare Jul 06 '24 at 16:10
  • @lazare I did not know what you were asking, I just guessed. I am not sure what you are asking about indeterminate. For any $n$ the form ${x_n \over n}$ is well defined. The analysis above shows that the limit is $1$. There is no indeterminate case. – copper.hat Jul 06 '24 at 16:56
  • So $\frac{h\to0}{n\to\infty}$ is a total legal determinate case??! I'm sorry but it look to big for me. I need some explanation for that or a less various confirmations from other members. – lazare Jul 06 '24 at 17:21
  • It's ok I get it. I'm agree with your limit. Sometime my brain block on some things. – lazare Jul 06 '24 at 17:32
  • If $x_n\to0$ and $|x_n| < \epsilon$. So, $x_n$ go to $0$ as a negative. Am I right? – lazare Jul 06 '24 at 19:49
  • 1
    @lazare no. There are no sign restrictions on $x_n$. – copper.hat Jul 06 '24 at 20:16
3

Remark. This solution works only for real numbers $a, b$.

Update (Thank Hilbert's Minion for pointing out the issue).

First, we have, for all $n \ge 4|a| + 4|b| + 4$,
$$\left(\frac{ab}{n^2}+\frac{a+b}{n}+1\right)^n \le \left(\frac{(a + b)^2}{4n^2} + \frac{a+b}{n}+1\right)^n = \left(\frac{a+b}{2n}+1\right)^{2n} \tag{1}$$ where we use $ab \le (a + b)^2/4$ (easy). (Note: If $n \ge 4|a| + 4|b| + 4$, then $\frac{ab}{n^2}+\frac{a+b}{n}+1 > 0$ and $\frac{a+b}{2n}+1 > 0$.)

Second, we have, for all $n \ge 4|a| + 4|b| + 4$,
\begin{align*} \left(\frac{ab}{n^2} +\frac{a+b}{n}+1\right)^n &= \left( \frac{a+b}{n}+1\right)^n\left(1 + \frac{\frac{ab}{n^2}}{\frac{a+b}{n}+1}\right)^n\\ &\ge \left( \frac{a+b}{n}+1\right)^n \left(1 + \frac{\frac{ab}{n^2}}{\frac{a+b}{n}+1}\cdot n\right)\\ &= \left( \frac{a+b}{n}+1\right)^n \left(1 + \frac{\frac{ab}{n}}{\frac{a+b}{n}+1}\right) \tag{2} \end{align*} where we use the Bernoulli inequality $(1 + u)^r \ge 1 + ur$ for all $u > -1$ and $r \ge 1$.

Third, we have $$\lim_{n\to \infty} \left(\frac{a+b}{2n}+1\right)^{2n} = \lim_{n\to \infty} \left(\frac{a+b}{n}+1\right)^{n}, \quad \lim_{n\to \infty} \left(1 + \frac{\frac{ab}{n}}{\frac{a+b}{n}+1}\right) = 1. \tag{3}$$

From (1)-(3), by the squeeze theorem, we have $$\lim_{n\to{\infty}}\left(\frac{ab}{n^2} +\frac{a+b}{n}+1\right)^n= \lim_{n\to{\infty}}\left(\frac{a+b}{n}+1\right)^n.$$

River Li
  • 49,125
  • How we know that $( \frac{\frac{ab}{n^2}}{\frac{a+b}{n}+1})\gt-1$ ? – lazare Jul 05 '24 at 15:31
  • @lazare Can you prove it yourself using $n \ge 4|a| + 4|b| + 4$? – River Li Jul 05 '24 at 15:36
  • I just tried but I don't see. – lazare Jul 05 '24 at 15:43
  • @lazare We split into two cases. If $\frac{\frac{ab}{n^2}}{\frac{a+b}{n}+1} \ge 0$, then $\frac{\frac{ab}{n^2}}{\frac{a+b}{n}+1} > -1$. If $\frac{\frac{ab}{n^2}}{\frac{a+b}{n}+1} < 0$, first you can prove that $\frac{a+b}{n} + 1 \ge \frac34$, second you can prove that $\frac{ab}{n^2} \ge -\frac{1}{4}$. – River Li Jul 05 '24 at 22:15
  • Can you go more straight to the demonstration that you want to share. It's like I have guess 75% of your results. I will not try myself to find why when: $\frac{\frac{ab}{n^2}}{\frac{a+b}{n}+1} < 0$ then $\frac{a+b}{n} + 1 \ge \frac{3}{4}$ it's boring. May be that if the demonstration was less labyrinthic I would have more will to play but that is not the case, it's actually a tough one and I don't feel like there is a light at the end of the tunnel or at less not enough to spend so much energy on it. – lazare Jul 06 '24 at 14:21
  • @lazare OK. We do it step by step. First, If $n \ge 4|a| + 4|b| + 4$, we have $$\frac{a + b}{n} + \frac{1}{4} = \frac{a + b + n/4}{n} \ge \frac{a + b + |a| + |b| + 1}{n} = \frac{(a + |a|) + (b + |b|) + 1}{n} > 0$$ where we use $a + |a| \ge 0$ and $b + |b| \ge 0$. Any question? – River Li Jul 06 '24 at 14:25
  • To what $\frac{a + b}{n} + \frac{1}{4}$ is related in your previous message? – lazare Jul 06 '24 at 15:02
  • @lazare $\frac{a+b}{n} + 1 \ge \frac34 \iff \frac{a+b}{n} + \frac14 \ge 0$. In detail, $\frac{a+b}{n} + 1 - \frac34 \ge \frac34 - \frac34$ because $1 - \frac34 = \frac14$ and $\frac34 - \frac34 = 0$. – River Li Jul 06 '24 at 15:06
  • Why $\frac{a+b}{n} + 1 \ge \frac{3}{4}$ ? – lazare Jul 06 '24 at 15:21
  • @lazare We need to prove that $\frac{a+b}{n} + 1 \ge \frac{3}{4}$. Since $\frac{a+b}{n} + 1 \ge \frac34 \iff \frac{a+b}{n} + \frac14 \ge 0$, it suffices to prove that $\frac{a + b}{n} + \frac{1}{4} \ge 0$. – River Li Jul 06 '24 at 15:25
1

For all $p > 1$, as $h \to {0}_{+}$,

$$\begin{align} {\left( 1 + a h + C {h}^{p} \right)}^{1 / h} & = \exp \left( \frac {1}{h} \ln \left( 1 + a h + C {h}^{p} \right) \right) \\ & = \exp \left( \frac {a h + C {h}^{p}}{h} \cdot \frac {\ln \left( 1 + a h + C {h}^{p} \right)}{a h + C {h}^{p}} \right) \to \exp \left( a \right). \end{align}$$

Above, we have used the following result:

$$\frac {\ln \left( 1 + h \right)}{h} \to 1.$$


EDIT:

Here, as requested, we want to disuse the exponential function as it has yet to be defined. To this end, note that for all $n \in \mathbb {N}$,

$$\begin{align} {\left( 1 + \frac {a}{n} + \frac {C}{{n}^{p}} \right)}^{n} & = \sum_{k = 0}^{n} \binom {n}{k} {\left( 1 + \frac {a}{n} \right)}^{n - k} {\left( \frac {C}{{n}^{p}} \right)}^{k} \\ & = {\left( 1 + \frac {a}{n} \right)}^{n} \cdot \sum_{k = 0}^{n} \binom {n}{k} {\left( \frac {C / {n}^{p}}{1 + a / n} \right)}^{k}. \end{align}$$

As $n \to \infty$, for all $p > 1$,

$$\begin{align} \left| \sum_{k = 0}^{n} \binom {n}{k} {\left( \frac {C / {n}^{p}}{1 + a / n} \right)}^{k} \right| & \le \sum_{k = 0}^{n} \binom {n}{k} {\left| \frac {C / {n}^{p}}{1 + a / n} \right|}^{k} \\ & \le \sum_{k = 0}^{n} {n}^{k} \, {\left| \frac {C / {n}^{p}}{1 + a / n} \right|}^{k} \\ & \le \sum_{k = 0}^{\infty} {\left( \frac {\left| C \right| / \, {n}^{p - 1}}{1 - \left| a \right| / \, n} \right)}^{k} = {\bigg( 1 - \frac {\left| C \right| / \, {n}^{p - 1}}{1 - \left| a \right| / \, n} \bigg)}^{- 1} \to 1. \end{align}$$

Meanwhile,

$$\displaystyle \left| \sum_{k = 0}^{n} \binom {n}{k} {\left( \frac {C / {n}^{p}}{1 + a / n} \right)}^{k} \right| \ge 1 - \binom {n}{1} \cdot \frac {\left| C \right| / {n}^{p}}{1 + \left| a \right| / n} \to 1.$$

So

$$\left| \sum_{k = 0}^{n} \binom {n}{k} {\left( \frac {C / {n}^{p}}{1 + a / n} \right)}^{k} \right| \to 1.$$

Simon
  • 1,486
  • Wow I didn't exept that we need to go this far to get a solution. Is my professor so bad that he indicate me a dead end proof? – lazare Jul 03 '24 at 23:15
  • I edited my post. I don't want an exp function if possible in the solution. – lazare Jul 03 '24 at 23:21
  • @lazare, I have edited my post. Not very rigorous, though. – Simon Jul 04 '24 at 00:35
  • How do you find that: $\begin{align} {\left( 1 + \frac {a}{n} + \frac {C}{{n}^{p}} \right)}^{n} & = \sum_{k = 0}^{n} \binom {n}{k} {\left( 1 + \frac {a}{n} \right)}^{n - k} {\left( \frac {C}{{n}^{p}} \right)}^{k} \ \end{align}$ ? – lazare Jul 05 '24 at 19:47
  • 1
    @lazare, by the binomial theorem: https://www.quora.com/How-do-you-prove-the-binomial-theorem/answers/4158038 – Simon Jul 05 '24 at 23:02
  • But actualy $( 1 + \frac {a}{n} + \frac {C}{{n}^{p}})^{n}$ is a trinomial. – lazare Jul 06 '24 at 15:38
  • 1
    @lazare, it is apparently a trinomial, but you can see it as a binomial where $1 + a / n$ is a term and $C / {n}^{p}$ is the other. (Likewise, you can make it a $100$-nomial if you feel like.) – Simon Jul 06 '24 at 20:52
  • Ok for the binomial formula. Now, how do you get this? $\sum_{k=0}^{n} \binom{n}{k} {(1+\frac{a}{n})}^{n - k} {(\frac{C}{{n}^{p}} )}^{k} = {( 1 + \frac {a}{n} )}^{n} * \sum_{k = 0}^{n} \binom {n}{k} {( \frac{C / {n}^{p}}{1 + a / n})}^{k}$

    I mean, how did you cancel the $-k$ in the $(1+\frac{a}{n})^{n-k}$ and how it is related to the change in this right side: $(\frac{C/n^p}{1+a/ n})^{k}$ ?

    – lazare Jul 09 '24 at 03:11
  • 1
    @lazare, see$$\begin{align} {\Big( 1 + \frac {a}{n} \Big)}^{n - k} & = {\Big( 1 + \frac {a}{n} \Big)}^{n} {\Big( 1 + \frac {a}{n} \Big)}^{- k}, \ {\Big( 1 + \frac {a}{n} \Big)}^{- 1} ;; & = \frac {1}{1 + a / n}. \end{align}$$ Can you take it from here? – Simon Jul 09 '24 at 05:25
  • Ok... Why $(1+\frac{a}{n})^{-k}(\frac{C}{n^p})^k$ would not be equal to $((1+\frac{a}{n})(\frac{C}{n^p}))^0=1$? – lazare Jul 09 '24 at 06:24
  • 1
    @lazare, ${a}^{x} {b}^{- x} = \exp \left[ x \left( \ln \left( a \right) - \ln \left( b \right) \right) \right] = 1$ for all $x \in \mathbb {R}$ if and only if $\ln \left( a \right) = \ln \left( b \right)$ (i.e., $a = b$). – Simon Jul 09 '24 at 06:32
  • Oh yea, you are right it's an illegal move. – lazare Jul 09 '24 at 06:33