I want to ask MSE to confirm the correctness of the alternate solution and its mistake.

My previous attempt was wrong,That's why I made a new initiative.

Using the definition of the derivative I wanted to find the derivative of the function $f(x)=a^x$.

Here $a≠0,a\in \mathbb{R^{+}}$ and $a^x=e^{x \ln a}$

I wrote these:

$(a^x)'=\lim_{\delta x\to 0}\frac{a^{x+\delta x}-a^x}{\delta x}=a^x×\lim_{\delta x\to 0}\frac {a^{\delta x}-1}{\delta x}$

Now,I must find $\lim_{\delta x\to 0}\frac {a^{\delta x}-1}{\delta x}$

I tried to do something:

$$a^{\frac{\delta x}{a^{\delta x}-1}}=\left(1+\left(a^{\delta x}-1\right)\right)^{\frac{1}{a^{\delta x}-1}}$$

$$\lim_{\delta x \to 0} {a^{{\frac {\delta x}{a^{\delta x}-1}}}}=\lim_{\delta x \to 0} {{\left(1+\left(a^{\delta x}-1\right)\right)^{\frac{1}{a^{\delta x}-1}}}}$$

$$a^{\lim_{\delta x\to 0}{{{\frac {\delta x}{a^{\delta x}-1}}}}}=\lim_{\delta x \to 0} {{\left(1+\left(a^{\delta x}-1\right)\right)^{\frac{1}{a^{\delta x}-1}}}}$$

$$a^{\lim_{\delta x\to 0}{{{\frac {\delta x}{a^{\delta x}-1}}}}}=e$$

$$\lim_{\delta x\to 0}{{{\frac {\delta x}{a^{\delta x}-1}}}}=\log_a{e}$$

$$\frac{1}{\lim_{\delta x\to 0}{{{\frac {\delta x}{a^{\delta x}-1}}}}}=\frac{1}{\log_a{e}},\log_a{e}≠0$$

$$\lim_{\delta x\to 0} \frac {1}{ \frac{\delta x}{a^{\delta x}-1}}=\ln a$$

$$\lim_{\delta x\to 0}\frac {a^{\delta x}-1}{\delta x}=\ln a$$

I used:

  • $\lim_{n\to 0}{n}=\lim_{\delta x\to 0}{(a^{\delta x}-1)}=a^0-1=0$

  • $\lim_{\delta x \to 0} {{\left(1+\left(a^{\delta x}-1\right)\right)^{\frac{1}{a^{\delta x}-1}}}}=\lim_{n\to 0}{(1+n)^{\frac 1n}}=e$


$$(a^x)'=\lim_{\delta x\to 0}\frac{a^{x+\delta x}-a^x}{\delta x}=a^x×\lim_{\delta x\to 0}\frac {a^{\delta x}-1}{\delta x}=a^x \ln a$$

Is this method/way/solution correct?

Thank you!

  • $\begingroup$ Is this how $e$ was discovered ? $\endgroup$ – Rene Schipperus Mar 6 '18 at 0:00
  • $\begingroup$ $\lim_{n\to\infty} {(1+1/n)^n}=e$ $\endgroup$ – MathLover Mar 6 '18 at 0:07
  • 5
    $\begingroup$ One can't answer this without knowing how you define $a^x$ in the first place. If you can use $a^x=e^{x\ln a}$ and $\lim_{x\to0}\frac{e^x-1}{x}=1$, then a simple substitution suffices. $\endgroup$ – egreg Mar 6 '18 at 11:38
  • $\begingroup$ @egreg I tried to make up some deficiencies.. $\endgroup$ – MathLover Mar 6 '18 at 11:53
  • $\begingroup$ You should first write the definition of $a^x$ (see Jack D'Aurizio answer). Without this you can't solve the problem. $\endgroup$ – Paramanand Singh Mar 8 '18 at 8:33

There are some subtleties here.

  1. How is $a^x$ defined, for a generic $a>0$ and some $x\in\mathbb{R}$? The most common ways are to define $a^x$ directly as $\exp\left(x\log a\right)$, or to consider a sequence of rational numbers $\left\{\frac{p_n}{q_n}\right\}_{n\geq 0}$ convergent to $x$ and let $a^x=\lim_{n\to +\infty}a^{\frac{p_n}{q_n}}$ ;

  2. If we have $\lim_{n\to +\infty}f(n)=L$ (limit of a sequence) it is not granted that $\lim_{x\to +\infty} f(x)=L$ (limit of a function) without further assumptions on $f$. For instance $\lim_{n\to +\infty}\sin(\pi n)=0$ but $\lim_{x\to +\infty}\sin(\pi x)$ does not exist, so we have to be careful in deriving $\lim_{x\to 0}\frac{a^x-1}{x}=\log a$ from $\lim_{n\to +\infty}\frac{a^{1/n}-1}{1/n}=\log a$;

  3. On the other hand, there is no need to over-complicate things: $\lim_{x\to 0}\frac{a^x-1}{x}$ is just the value of the derivative of $a^x$ at the origin. If we know/prove that $a^x=\exp\left(x\log a\right)$, then $\lim_{x\to 0}\frac{a^x-1}{x}=\log a$ is a straightforward consequence of the chain rule, given the differentiability of the exponential function.

  • $\begingroup$ Thank you JackD'aurizio ..My mind is very confused. I'm trying to figure it out right now..( $\endgroup$ – MathLover Mar 6 '18 at 1:32
  • $\begingroup$ by keeping your answer's guidance, I tried to do something.. $\endgroup$ – MathLover Mar 6 '18 at 2:30
  • $\begingroup$ @MathLover: you still have to clarify what is your definition of $a^x$. If $a^x=\exp(x\log a)$, it is pretty obvious that $\frac{d}{dx} a^x = a^x\log a$. $\endgroup$ – Jack D'Aurizio Mar 6 '18 at 2:33

User Jack D'Aurizio has raised some good points in his answer and I want to add some remarks in that direction. So this is not exactly an answer but more of a comment which does not fit in comment box.

For most students the first encounter with exponential and logarithmic functions is not a very fascinating one. They are presented with lots of nice (and even strange) properties of these functions and usually the definitions and proofs are absent. Armed with this information the student simply can not evaluate limits as the one in question and needless to say the proof / solution is incorrect.

Textbook authors should be more forthcoming and intellectually honest to say that a proper theory of these functions can not be provided at this stage and the properties of these functions have to be assumed (translate: mugged up) without any proof whatsoever. And they should not give exercises to prove these properties rather the exercises should use these assumed properties to handle more challenging problems. Thus the current question is pointless unless the proper theory of these functions has been developed.

Instead of asking "Evaluate $\lim\limits_{x\to 0}\dfrac{a^x-1}{x}$" it is better to give the following exercise :

Evaluate the limit $\lim\limits_{x\to 0}\dfrac{2^x+3^x-2}{x}$. You may use the limit formula $\lim\limits_{x\to 0}\dfrac{a^x-1}{x}=\log a$ and properties of logarithms.

Update: I decided to answer your question in simple English. Your approach has the following issues:

  • no definition of $a^x$ is provided.
  • you need to show that $\lim\limits_{\delta x\to 0}a^{\delta x} =1$
  • you need to define $e$ and show that $\lim\limits_{\delta x\to 0}(1+\delta x) ^{1/\delta x} = e$.
  • you need to also define $\log a$.

Fixing all the above issues is difficult. Some of these problems have been fixed. See below.

Further Update: the question has been edited to include some definitions (like for $a^x$) and some assumptions about $e$ and logarithmic function have been used. With these assumptions one can say that your approach is correct.

Also OP has raised a concern in comments: does one need to prove each and every theorem being used in order to solve a problem? NO!! But only when the theorem is well known and supposed to be considered basic. If one is trying to prove the basic results then one has to be extra cautious and mention the definitions and prove the basic results using those definitions. For example, one can use L'Hospital's Rule to evaluate a limit, but when one has to prove L'Hospital's Rule, it is necessary to go deeper and start with mean value theorems.

  • $\begingroup$ @MathLover: your method is not rigorous / wrong. But it is not your fault. Your textbook does not tell you the right method to handle such problems. I hope you can understand this comment. $\endgroup$ – Paramanand Singh Mar 6 '18 at 9:26
  • 1
    $\begingroup$ This is the first time I have received such downvotes and also a delete vote (without any comment). I am flagging for moderator attention. $\endgroup$ – Paramanand Singh Mar 8 '18 at 18:05
  • 1
    $\begingroup$ @MathLover : well that comment is now outdated and I will delete it. I don't mean to discourage you. I hope you will learn such topics in very near future. $\endgroup$ – Paramanand Singh Mar 8 '18 at 19:30
  • 1
    $\begingroup$ @MathLover : don't worry. When I was learning these things as a student I faced similar problems. I suggest you read the book A Course of Pure Mathematics by G H Hardy. It is one of the best books for calculus which helped me a lot. But you will need to improve your English also to read it because it uses lot of English instead of math symbols. $\endgroup$ – Paramanand Singh Mar 8 '18 at 20:14
  • 1
    $\begingroup$ @MathLover : it's an old book from 1940s. I don't think there is any translation into any language. But you can try Google translate. Best would be to find a friend in your country who knows good English and ask him to translate. Bye for now. Enjoy learning here. $\endgroup$ – Paramanand Singh Mar 8 '18 at 21:01

I thought it might be instructive to present a way forward the relies on pre-calculus tools only. To that end we proeed.

If we define $a^x$ by the expression $e^{\log(a)\,x}$, then we can write

$$\frac{a^x-1}{x}=\frac{e^{\log(a)\,x}-1}{x}\tag 1$$

In THIS ANSWER, I used only the limit definition of the exponential function along with Bernoulli's Inequality to show the exponential function satisfies the inequalities

$$1+x\le e^x\le \frac{1}{1-x}\tag2$$

for $x<1$.

Applying $(2)$ to $(1)$ reveals for $0<x<1$

$$ \log(a) \le \frac{a^x-1}{x}\le \frac{\log(a)}{1-\log(a) x}\tag3$$

and for $x<0$

$$\frac{\log(a)}{1-\log(a)x}\le \frac{a^x-1}{x}\le \log(a)\tag 4$$

Finally, applying the squeeze theorem to $(3)$ and $(4)$ yields the coveted limit

$$\bbox[5px,border:2px solid #C0A000]{\lim_{x\to 0}\frac{a^x-1}{x}=\log(a)}$$

And we are done!

  • $\begingroup$ Thank You for the answer..I read completely and Understand.. $\endgroup$ – MathLover Mar 6 '18 at 23:11
  • $\begingroup$ You're welcome! My pleasure. $\endgroup$ – Mark Viola Mar 6 '18 at 23:46
  • $\begingroup$ Very nice derivation as usual from you! $\endgroup$ – gimusi Mar 8 '18 at 19:37
  • $\begingroup$ @gimusi Thank you ... as usual. You just made my day. $\endgroup$ – Mark Viola Mar 8 '18 at 20:19

I guess one way to proceed is as follows. I use $\Delta$ to simplify notation. You need $$\lim_{\Delta \to 0} \frac{a^{\Delta} - 1}{\Delta}$$. Let $a = e^b$ or $b = \ln(a)$. multiplying numerator and denominator by $b$, you get the above equals $$b \cdot \lim_{\Delta \to 0} \frac{e^{b\Delta} -1}{b\Delta}$$ which is the derivative of $e^x$ at $x=0$ which is 1. So the expression you want is $b = \ln a$.


So here is a quick idea: let $\delta x = \log_a(1 + \frac{1}{n}) \to 0$ $$ \lim_{\delta x \to 0} \frac{a^{\delta x} - 1}{\delta x} = \lim_{n \to \infty} \frac{1}{n\cdot \log_a(1 + \frac{1}{n})} = \lim_{n \to \infty} \left(\log_a(1 + \frac{1}{n})^n \right)^{-1} \to \log_a(e)^{-1} = \ln(a)$$

  • $\begingroup$ This is true but the first equality needs to be properly justified: see the second point of my answer. $\endgroup$ – Jack D'Aurizio Mar 6 '18 at 1:20

If we accept

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

Then we try to prove:

\begin{equation} \lim_{x \rightarrow 0} \frac{e^{x} - 1}{x} = 1 \end{equation}

This will solve our problem, because then:

$$\lim_{h \rightarrow 0}\frac{a^{h} - a}{h} = \lim_{h \rightarrow 0}\frac{e^{h\ln(a)} - 1}{h} = \ln(a)\lim_{t \rightarrow 0}\frac{e^{t} - 1}{t} = \ln(a)$$

Where we use the substitution $t = h\ln(a)$. Now use the binomial theorem:

$$\lim_{x \rightarrow 0}\frac{e^{x} - 1}{x} = \lim_{x \rightarrow 0}\frac{\lim_{n \rightarrow \infty}\sum_{k = 0}^{n}\binom{n}{k}\frac{x^{k}}{n^{k}} - 1}{x} = 1 + \lim_{x \rightarrow 0}\lim_{n \rightarrow \infty}\sum_{k = 2}^{n}\binom{n}{k}\frac{x^{k-1}}{n^{k}}$$

Since $\binom{n}{k} \leq n^{k}$ for all $k \geq 0$, we have:

$$\left|\sum_{k = 2}^{n}\binom{n}{k}\frac{x^{k-1}}{n^{k}}\right| \leq \sum_{k = 1}^{n-1} |x|^{k} \leq \sum_{k=1}^{\infty}|x|^{k} = \frac{|x|}{1 - |x|}$$

And so:

$$\lim_{x \rightarrow 0}\lim_{n \rightarrow \infty} \sum_{k = 2}^{n}\binom{n}{k}\frac{x^{k-1}}{n^{k}} = 0$$

  • $\begingroup$ While this is rigorous, it does not discuss how $\log$ and equation $a^h=e^{h\log a} $ come into picture. Also one needs to show that $e^{x+y} =e^xe^y$ if one tries to find the derivative of $a^x$. $\endgroup$ – Paramanand Singh Mar 6 '18 at 7:51

enter image description here $$\phantom{extra characters. Please convert the image above to Latex.}$$


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.