I have some function $F(\omega): \mathbb R\to\mathbb C$. The function $F(\omega)$ has both roots and singularities. Fortunately, I can calculate positions of singularities analytically.

So my problem is to find roots inside domain between singularities. Here I have difficulty, because I cannot estimate if there is one, many or none roots inside the domain. Now I scan $\omega$ through the domain for change of simultaneous change of real and imaginary part of $F(\omega)$, what is both inaccurate and ineffective.

If there some method to estimate number of roots inside a domain?

I don't know if this is a right place to post questions about computational math, so don't hesitate to show me the right place to post :)

  • $\begingroup$ If $C$ is a closed contour that does not enclose any singularity then the number of zeroes inside $C$ is given by $$\int_C\frac{F'}{F}d\omega.$$ If there are poles enclosed by $C$ you have to add them to the result to get the only the number of zeroes. $\endgroup$
    – WimC
    Jun 1 '12 at 16:29
  • $\begingroup$ Thank you for your answer. I'll try this technique, the problem here is computation of $F'$. Am I right that contour can be just a segment in real axis? $\endgroup$
    – galadog
    Jun 1 '12 at 16:46
  • $\begingroup$ I forgot the factor $(2 \pi i)^{-1}$. $\endgroup$
    – WimC
    Jun 1 '12 at 16:49
  • $\begingroup$ @galadog no, the contour must be a closed loop in the plane. That integral calculates the number of zeros inside the loop. $\endgroup$ Jun 1 '12 at 16:59

See Computing the Zeros of Analytic Functions for theory and code.

Despite the title, it handles computing zeros and poles of meromorphic functions.

  • $\begingroup$ Are we sure that the above reference really helps? From looking at the statement of the question, it appears we are talking about a function from the reals into the complex plane, rather than an analytic or meromorphic function. (Unless I have it all wrong). $\endgroup$
    – Old John
    Jun 1 '12 at 16:48
  • $\begingroup$ @OldJohn: good point. Could still work, but that would depend a bit on the actual form of $F$. $\endgroup$
    – WimC
    Jun 1 '12 at 16:52
  • $\begingroup$ @OldJohn, right, I just assumed the function could be extended to the complex plane. $\endgroup$
    – lhf
    Jun 1 '12 at 17:00

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.