MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $f$ be a function analytic on an open subset $D\subset \mathbb{C}$, and let $\gamma:[0,1] \to D$ be a line segment. $g = f\circ\gamma$ is another curve in the complex plane; is it possible to for $g$ to cross a straight line infinitely often, where the crossing points accumulate towards a point? That is, does there exist a point $\alpha$ and a ray $R$ emanating from the point $\alpha$ such that for all $\epsilon > 0$, $g$ crosses the ray $R$ infinitely many times in the $\epsilon$-ball around $\alpha$?

We're trying to show something about analytic continuation, but we cannot rule out pathological beasts like these.


share|cite|improve this question
Oh dear, this question looks as though it has the potential to obsess. I thought I had an answer but it collapsed as I tried to write it down. – gowers Apr 22 '10 at 7:02
Please define "cross the ray". Maybe I am missing something, but if I take $f(z) = z$ and $R = \gamma$ then $g$ is valued in the ray. Does that count? – BCnrd Apr 22 '10 at 7:19
Are you thinking of the case when the inverse image of these crossings converges to a point outside D? – Zsbán Ambrus Apr 22 '10 at 8:57
up vote 6 down vote accepted

The image of $[0,1]$ is compact and so must contain the purported accumulation point. It makes no loss to assume that $\gamma(t)=t$, $f(0)=0$ is the accumulation point, and the line in question is the real axis. Then $f(z)=a_n z^n+a_{n+1}z^{n+1}+\cdots$ where $a_n$ is nonzero and $n$ is a positive integer.

At this stage I'll assume there is a sequence of reals $t_1>t_2>\cdots$ tending to zero with each $f(t_j)$ real. we want to show that all the $a_k$ are real. Then considering $f(t_j)/t_j^n$ we get $a_n$ real. Now consider $f(z)-a_n z^n$ in place of $f(z)$. We get $a_{n+1}$ real etc. So $f$ takes reals to reals so there's no "crossing" of the real axis.

In general, there must be a sequence of distinct points $(t_j)$ in $[0,1]$ where the curve crosses the line and which tends (by Bolzano-Weierstrass) to a point $t\in[0,1]$. Replace $[0,1]$ by $[0,t]$ or $[t,1]$ (one of these has infinitely many $t_j$) and rescale the interval to $[0,1]$.

share|cite|improve this answer
Very nice. My mistake was to try to use the no-isolated-zeros theorem rather than imitating its proof. – gowers Apr 22 '10 at 12:12
I follow everything, except for the base step to establish $a_n\in\mathbb{R}$. I presume that it's because $f(z)/z^n$ has an accumulation of real values on values from the $[0,1]$ interval, and since the image of $[0,1]$ contains the accumulation point, $f(0)/0^n = a_n$ must be real valued as well. – Henry Yuen Apr 22 '10 at 17:18
Each $f(t_j)/t_j^n$ is real. As $j\to\infty$, $t_j\to0$ and so $f(t_j)/t_j^n\to a_n$. As the limit of a sequence of real numbers, $a_n$ is real. – Robin Chapman Apr 22 '10 at 17:39

I am little worried by Gowers' comment as this seems fairly straightforward to me.

Let $L:{\mathbb C} \to {\mathbb R}$ be a nonconstant affine map vanishing on your ray. I assume that by "line segment" you really mean that $\gamma$ is linear (or affine). In any case, as long as $\gamma$ is real analytic (by which I mean that it extends to a real analytic function on some open neighborhood of $[0,1]$), so is $h := L \circ f \circ \gamma$. A real analytic function on a compact interval is either always zero or has only finitely many zeros. If $h$ is identically zero, then your curve will stay on the line containing your ray and can enter or leave the ray only finitely many times. If $h$ is not identically zero, then your curve meets the ray only finitely many times.

share|cite|improve this answer
My comment does not imply that the question isn't straightforward. I am struggling to understand your argument, however. Are your "real analytic" functions complex-valued? I should add that the fact that I am struggling to understand your argument is compatible with its being clear and correct. – gowers Apr 22 '10 at 12:09
No, my functions are real valued. The function L has the form L(z) = a Re(z) + b Im(z) + c for some real numbers a, b and c. I am using the fact that if F(z) is complex analytic, then via the usual identification of C with R^2, the induced function G:R^2 -> R^2 given by (a,b) -> (Re(F(a+ib)),Im(F(a+ib))) is real analytic, a result one can see by expanding the Taylor series for F(a+ib). – Thomas Scanlon Apr 22 '10 at 16:09

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.