2
$\begingroup$

Let $C$ be a simple closed curve in the complex plane, and let $f$ be holomorphic on an open set containing $C$. Is there a condition on the signs of Im $f$ and Re $f$ on $C$ that guarantees the existence of a zero of $f$ in the interior of $C$? (Shih's theorem is not quite what I want; there the condition is $\text{Re }\overline{z} \cdot f(z) > 0$.) (See M.-H. Shih, "Bolzano's Theorem for Functions of a complex variable, Am. Math. Monthly v. 89 (1982), 210-211.)

If the number of zeros of $f$ inside $C =$ the number of zeros of the identity in $f(C)$ then it should be sufficient that $f(C)$ passes through each quadrant in cyclic order (which can be rewritten as conditions on the imaginary and real parts as originally requested.) When is this equality the case?

$\endgroup$

2 Answers 2

2
$\begingroup$

"The number of zeros of identity" is the strange expression: this number is always $0$ or $1$. Assuming that $f$ is analytic in an open set containing $C$ and its interior region, and has no zeros on $C$, the number of zeros inside $C$ is always equal to the index of the curve $f(C)$ about $0$, $$\frac{1}{2\pi i}\int_{f(C)}\frac{dw}{w}=\frac{1}{2\pi i}\int_C\frac{f'(z)}{f(z)}dz.$$ All other counting formulas for zeros follow from this, as Robert Israel wrote.

This can be counted in terms of the numbers of crossing of the coordinate axes by $u=\Re f$ and $v=\Im f$. For example, consider those points on $C$ where $u(z)>0$ and $v$ changes sign at $z$ (as $z$ moves on $C$ in positive direction). Then the index equals to the number of changes from $-$ to $+$ minus the number of changes from $+$ to $-$ at these points.

$\endgroup$
4
  • $\begingroup$ Of course, this only works if $f$ is defined on all the 'interior' of the curve. $\endgroup$ Sep 7, 2015 at 13:14
  • $\begingroup$ I used the expression "number of zeros of the identity in $f(C)$" to denote the winding number of $f(C)$ around zero. Then (it seems to me) the equality I mentioned holds for holomorphic functions, and then one version of the condition I asked for would be (as I said) that $f(C)$ passes through all four quadrants in cyclic order--this is just a condition on the real and imaginary parts of $f(z), z \in C$. Is this indeed a sufficient condition to conclude that the open set bounded by $C$ contains a zero of $f$? $\endgroup$
    – student
    Sep 7, 2015 at 15:00
  • $\begingroup$ Your sufficient condition is not stated clearly in this remark, but perhaps if you state it clearly it will become correct. $\endgroup$ Sep 7, 2015 at 18:12
  • $\begingroup$ @Simon Henry: you are right. I thought that this was stated in the question but now I see it was not:-) $\endgroup$ Sep 7, 2015 at 18:14
1
$\begingroup$

I don't know what kind of condition you're looking for, but whatever condition it is must essentially work through the Argument Principle: as $z$ goes around the (positively oriented) curve $C$, it must force $f(z)$ to have winding number $\ge 1$ around $0$.

$\endgroup$
1
  • $\begingroup$ My edit (adding the second paragraph) and your reply just overlapped. Are you saying that the answer to my question "when is this equality the case" is "always"? $\endgroup$
    – student
    Sep 7, 2015 at 5:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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