4

It seems that when the logarithmic derivative of a function exists, the function itself cannot be zero. But I failed to construct a convincent proof. Is this true?

$\dfrac{f'(x)}{f(x)}$ is bounded implies $f(x)$ cannot have zero as a limiting value.

Is this true?

will
  • 51

2 Answers2

3

Let $f:x\to \frac{1}{x}$ on $[1,+\infty)$, then we have $$\frac{f'(x)}{f(x)}=-\frac{1}{x},$$ so $\displaystyle\frac{f'}{f}$ is bounded on $[1,+\infty)$ Nevertheless $\displaystyle \lim_{+\infty}f=0.$

3

True when $x$ approaches a finite $x_0$, not when $x$ approaches $\pm \infty$, as explained below.

Assume $f$ is differentiable on $\mathbb{R}$.

Then the function $$ g(x):=\ln|f(x)| $$ is defined and differentiable on the open set $$ U=\{x\in\mathbb{R}\;;\; f(x)\neq 0\}. $$ Its derivative is $$ g'(x)=\frac{f'(x)}{f(x)}\qquad \forall\;x\in U. $$

Write $U=\bigcup_{n\geq 1}(a_n,b_n)$ as a countable disjoint union of open intervals. See this thread to see why this is possible.

Consider for instance a finite point $x_0=a_n$.

Note that $\lim_{x\rightarrow x_0^+}\ln |f(x)|=\lim_{x\rightarrow x_0^+}g(x)=-\infty$.

Now if $g'=(\ln |f|)'$ was bounded by, say, $M$ on $(x_0,x_0+\delta]$, the mean value theorem, and then the triangular inequality, would yield $$ |g(x)-g(x_0+\delta)|\leq M \delta\quad \Rightarrow\quad |g(x)|\leq |g(x_0+\delta)|+M\delta $$ for all $x\in(x_0,x_0+\delta]$.

Contradiction.

So the logarithmic derivative is unbounded when it approaches every finite boundary point of $U$.

Now regarding the behavior at $\pm\infty$, you can easily find examples where $f'/f$ is bounded while $f$ tends to $0$.

Julien
  • 44,791
  • I'm not sure I understood the $\Rightarrow$ properly. I get how the left-hand side is true but I'm not sure I understand the implication. Could you please add some more details? – xavierm02 Mar 04 '13 at 21:12
  • @xavierm02 Sorry but what lhs are you talking about? And what implication? – Julien Mar 04 '13 at 21:21
  • I'm talking about the only place you wrote $a\Rightarrow b$ and the thing I call the left-hand side is the $a$. – xavierm02 Mar 04 '13 at 21:30
  • 1
    @xavierm02 Ah, ok. This is the triangular inequality: $|g(x)|=|g(x)-g(y)+g(y)|\leq |g(x)-g(y)|+|g(y)|$. – Julien Mar 04 '13 at 21:32
  • Oh right >< With your formulation I thought the mean value theorem gave the implication too whereas it only gives the left-hand side. Idk how I missed something that obvious ><. Thanks :) – xavierm02 Mar 04 '13 at 21:36
  • You write $U=\bigcup_{n\geq 1}(a_n,b_n)$ which means you assume there are countably many zeroes at most. It must come from the fact $f$ is differentiable but I can't see how... Does it? – xavierm02 Mar 04 '13 at 21:40
  • 1
    @xavierm02 No, I use the fact that every open set in $\mathbb{R}$ is the disjoint union of countably many open intervals. – Julien Mar 04 '13 at 21:42
  • 1
    @xavierm02 See this thread: http://math.stackexchange.com/questions/318299/any-open-subset-of-bbb-r-is-a-countable-union-of-disjoint-open-intervals-co – Julien Mar 04 '13 at 21:46
  • Coming to this topic after a long delay! Thanks you for the many answers and coments. – will Jun 26 '14 at 14:16