I know that the derivative of $b^x$ is just $b^x \log{(b)}$, and I've seen it being derived using chain rule and such (not that I understand how it's done, I just learned about $e$ today so using the chain rule to derive $b^x$ is outside my scope as of now). In any case I was trying to find out how you could derive $b^x$ using the definition of the derivative, $[f(x+h) - f(x)]/h$. I couldn't find it on the internet so I was wondering if someone here could show me. Thanks in advance!

  • $\begingroup$ Actually I think it's $b^x\log(b)$ not $b^x\log(x)$. Does that help? $\endgroup$ – Gregory Grant Apr 21 '15 at 20:33
  • $\begingroup$ Aha, sorry, brain fart lol. And not really I'm afraid. $\endgroup$ – inglavatin Apr 21 '15 at 20:34
  • $\begingroup$ I guess you need to show $\lim_{h\rightarrow 0}\frac{b^h-1}{h}=\log(b)$. I'm not sure how difficult that is, but I don't think it's trivial. $\endgroup$ – Gregory Grant Apr 21 '15 at 20:37
  • $\begingroup$ How do you define $\log b$ without knowing $e$? $\endgroup$ – celtschk Apr 21 '15 at 20:40
  • 1
    $\begingroup$ @celtschk Area below the hyperbola, as usual. $\endgroup$ – Alamos Apr 21 '15 at 20:43

$$ \frac{d}{dx} b^x = \lim_{h \to 0} \frac{b^{x+h}-b^x}{h} = b^x \lim_{h \to 0} \frac{b^h-1}{h}. $$ At this point, you have to evaluate this limit. You have a couple of options here. Obviously L'Hôpital's not going to do. One way is, if your definition of $e^{y}$ is $$ \exp{y} = \sum_{k=0}^{\infty} \frac{y^k}{k!}, $$ then write $b^h = \exp{(h\log{b})} $, and you can evaluate the limit using the first few terms of the series.

On the other hand, if your definition is $$ \exp{y} = \lim_{n \to \infty} \left( 1 + \frac{y}{n} \right)^n, $$ you have a bit more of a problem. Essentially what this comes down to is the rearrangement $$ y = \left( 1+\frac{x}{n} \right)^n \to x = n(y^{1/n}-1) $$ Now, set $h=1/n$ and this looks like your limit, so you conclude that the limit is $\log{b}$. (I would like to see a cleaner version of this, should anyone have one to offer: the rearrangement to get the inverse doesn't look completely justified the way I normally do it.)

And if your definition of $\log{x}$ is an integral, and $\exp{x}$ as its inverse, then you're probably best proving it equivalent to one of the above first...

  • $\begingroup$ Nevermind ... I was just nitpicking. Or even pettifogging ;) $\endgroup$ – String Apr 21 '15 at 20:58
  • $\begingroup$ @String I for one don't mind those being discreetly edited away by others... still, a good knock to one's ego :-) $\endgroup$ – Chappers Apr 21 '15 at 21:00

Here's how one might do it with the definition $\log b = \int_1^b \frac{\mathrm dx}{x}$. However I'm not sure if the exchange of derivative and limit in the first step is allowed.

On one hand we have: $$\frac{\mathrm d}{\mathrm db}\lim_{h\to 0}\frac{b^h-1}{h} = \lim_{h\to 0}\frac{\mathrm d}{\mathrm db}\frac{b^h-1}{h} = \lim_{h\to 0}\frac{hb^{h-1}}{h} = \frac{1}{b}$$ On the other hand we have: $$\frac{\mathrm d}{\mathrm db}\log b = \frac{\mathrm d}{\mathrm db}\int_1^b\frac{\mathrm dx}{x} = \frac1b$$ Therefore we have $$\lim_{h\to 0}\frac{b^h-1}{h} = \log b+C$$ To determine $C$ we note that with $b=1$, we have on one hand $$\lim_{h\to 0}\frac{1^h-1}{h} = 0$$ and on the other hand $$\int_1^1\frac{\mathrm dx}{x} = 0$$ and therefore $C=0$.

  • $\begingroup$ Hang on, how are you interchanging the limits in $d/db$ and $\lim_{h \to 0}$? $\endgroup$ – Chappers Apr 21 '15 at 21:35

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.