How do I differentiate these functions in $\mathbb{Q}_2$?




I've been struggling with this for a couple of days now.

Please bear in mind I'm working without a teacher here so no laughing...

$\displaystyle f(x)=x^2: f'(x)=\lim_{\lvert a\rvert_2\to0}\frac{x^2-(x-a)^2}{a}=\lim_{\lvert a\rvert_2\to0}\frac{a^2+2ax}{a}=2x$

$f(x)=1/x:\displaystyle f'(x)=\lim_{\lvert a\rvert_2\to0}\frac{\frac{1}{x}-\frac{1}{x-a}}{a}=\lim_{\lvert a\rvert_2\to0}\frac{-1}{x(x-a)}=-x^{-2}$

But I think we would need to define $\displaystyle f'(x)=0$ for $x=0$?

As for $f(x)=\log(x)$ I have little idea where to start. Other than knowing that $\log$ is defined in $\mathbb{Q}_2$

  • $\begingroup$ The 2-adic logarithm $\log(1+x)$ is "naturally" defined only when $|1-x|_2<1$. In other words,inside $\Bbb{Q}_2$ you need $x\in 2\Bbb{Z}_2$. See Wikipedia. $\endgroup$ – Jyrki Lahtonen Aug 12 '17 at 13:39
  • 1
    $\begingroup$ Note: $\mathbb{Q}_2$ are the $2$-adic numbers. Anyone who claims to compute $\log'$ here should first show us the definition of $\log$ they are using in $\mathbb Q_2$. $\endgroup$ – GEdgar Aug 12 '17 at 13:48
  • 1
    $\begingroup$ Your first two computations are right, and the same will work in any topological field. Note $1/x$ is not defined for $x=0$, so we should not define its derivative there either. $\endgroup$ – GEdgar Aug 12 '17 at 13:53
  • 1
    $\begingroup$ Yes, and you can differentiate that series term-by-term, and get $1/(1+x)$ as the derivative. Valid when $x\in p\Bbb{Z}_p$. $\endgroup$ – Jyrki Lahtonen Aug 12 '17 at 15:20
  • 1
    $\begingroup$ @RobertFrost. I am just sitting next door ! Just come for a drink. $\endgroup$ – Claude Leibovici Aug 12 '17 at 16:32

Let's follow Wikipedia and define the 2-adic (Iwasawa) logarithm on $\Bbb{Q}_2^*$ as follows. Recall that the group $\Bbb{Q}_2^*$ is a direct product $$ \Bbb{Q}_2^*=\langle 2\rangle\times(1+2\Bbb{Z}_2). $$ The usual Taylor series of $\log(1+x)$ familiar from calculus $$ x-\frac{x^2}2+\frac{x^3}3-\frac{x^4}4+\cdots=\sum_{n=1}^\infty(-1)^{n+1}\frac{x^n}n $$ converges w.r.t. the 2-adic topology whenever $x\in2\Bbb{Z}_2$. This is because for all $n$ we have $\nu_2(x^n)=n\nu_2(x)\ge n$ but we easily see that $\nu_2(n)\le n/2$. Therefore the general term $x^n/n\to0$ which is sufficient for convergence in a complete non-archimedean case. We can thus use this series and define $$ \log(1+x)=\sum_{n=1}^\infty(-1)^{n+1}\frac{x^n}n $$ whenever $x\in 2\Bbb{Z}_2$. The familiar power series identities then show that the rule $$\log(xy)=\log x+\log y\qquad(*)$$ holds whenever $x,y\in1+2\Bbb{Z}_2$. We can extend the definition of the logarithm to all of $\Bbb{Q}_2$ (actually even further, see WP) in such a way that $(*)$ holds by declaring that $\log(2)=0$, and, as a consequence of $(*)$ that $$ \log(2^n(1+x))=\log(1+x)=x-\frac{x^2}2+\cdots $$ for all integers $n$ and all $x\in 2\Bbb{Z}_2$.

The usual business of termwise differentiation of a power series carries over to the 2-adics. Therefore when $x,a\in1+2\Bbb{Z}_2$ we get $$ \lim_{x\to a}\frac{\log x-\log a}{x-a}=1-(a-1)+(a-1)^2-(a-1)^3+\cdots=\frac1a $$ because the geometric series converges as $a-1\in2\Bbb{Z}_2$.

As we took care to have $(*)$, the usual calculation of the real limit goes thru, and the result $\log'(x)=1/x$ holds everywhere.

  • $\begingroup$ Thanks. I will have to digest over the coming days. $\endgroup$ – samerivertwice Aug 13 '17 at 21:04
  • $\begingroup$ Does $\langle2\rangle$ mean $\{\ldots1/2,1,2,4,\ldots\}$? And what does $X^{*}$ mean in general? Is it $X\setminus X^{\times}$? I did search long and hard before asking. $\endgroup$ – samerivertwice Aug 17 '17 at 18:04
  • 1
    $\begingroup$ If $R$ is a ring then $R^*$ is the group of units of that ring. I'm not sure about you, but some authors use $R^\times$ to mean the same thing. $\endgroup$ – Jyrki Lahtonen Aug 19 '17 at 6:50
  • $\begingroup$ Thanks, that is all it was. $\endgroup$ – samerivertwice Aug 19 '17 at 6:52


$$f(x)=\log(x)\implies f'(x)=\lim_{ a\to 0}\frac{\log(a+x)-\log(x)}{a}=\lim_{ a\to 0}\frac{\log\left(1+\frac a x\right)}{a}$$ Now, using equivalents $$\log\left(1+\frac a x\right)\sim \frac a x$$

  • $\begingroup$ Thanks. Is this much different in $\mathbb{Q}_2$ compared with $\mathbb{R}$? $\endgroup$ – samerivertwice Aug 12 '17 at 13:26
  • $\begingroup$ @RobertFrost. At this point, I have the feeling to be a Martian here. I am not a mathematician. So, .... $\endgroup$ – Claude Leibovici Aug 12 '17 at 13:47
  • $\begingroup$ @ClaudeLeibovici what part of Mars are you from? I know it well. $\endgroup$ – samerivertwice Aug 12 '17 at 13:50

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.