2

We consider $P(z)=a_{0}+a_{1}z+\cdot+a_{n-1}z^{n-1}+a_{n}z^n$, with $a_{0},\ldots,a_{n-1},a_{n} \in \mathbb{C}$ and $a_{n}\neq0$. Let $R=\max_{0\leq k\leq n-1}\left | \frac{a_k}{a_n} \right |$ and $S=\sum_{k=0}^{n-1}\left | \frac{a_k}{a_n} \right |$.

Can you help me establish the two following ?

a) Any complex root of $P$ has modulus less than or equal to $\max(1,S)$.

b) Any complex root of $P$ has modulus less than or equal to $1+R$.

It is worth noting that the approximation in b) is often better than that in a). Thank you for any hint or answer.

Davide Giraudo
  • 172,925
  • 3
    Did you ask 7 questions in the last 12 hours, including 5 in the last 2 hours? You might wish to slowdown a little... By the way, mentioning where you are stuck, what you tried and where you failed is well considered on this site. – Did Jan 11 '12 at 10:54
  • 1
    @Didier: The OP asked another one: http://math.stackexchange.com/questions/98138/the-graph-of-cot-is-the-image-of-the-graph-of-tan-by-a-simple-transformati – Paul Jan 11 '12 at 10:58
  • 1
    and one more: http://math.stackexchange.com/questions/98139/inequality-for-with-cot – Paul Jan 11 '12 at 11:00
  • 1
    @Paul: Indeed. Sigh. – Did Jan 11 '12 at 11:06
  • I understand your amazement for so many questions but I'm doing this because I found some interesting problems about complex numbers on the internet (more precisely here : http://www.eleves.ens.fr/home/kortchem/cg/complexes_expbs.pdf). Maybe this is stupid, but I wanted to post the questions before trying to solve them so that I would have the solutions ready when I would be stuck. I hope this does not disturb much. Or does it ? –  Jan 11 '12 at 11:09
  • And for this problem, the case $\left | z \right |\leq1$ is trivial. Thus we can suppose $\left | z \right |>1$.When $z$ is big, thepreponderant term of $P(z)$ is $a_{n}z^n$ and this can help us get a majorization. For b) I guess we should use geometric series (with ratio $\left | z \right |$ or $\left | \frac{1}{z} \right |$? –  Jan 11 '12 at 11:13
  • 1
    DidierPiau : "By the way, mentioning where you are stuck, what you tried and where you failed is well considered on this site." I'll try to act accordingly in the future. –  Jan 11 '12 at 11:14
  • 4
    I wanted to post the questions before trying to solve them... This is outrageous, if you ask me. – Did Jan 11 '12 at 11:16
  • 5
    @Didier: the beauty of it is that he's trying to sell us the hints in his exercise sheet as his guesses. – t.b. Jan 11 '12 at 11:31
  • And is there any issue with that ? Maybe this could have helped someone find the answer. Anyway, next time I'll try to think some time before asking questions. Sorry for that and thank you for the advice. –  Jan 11 '12 at 11:38
  • And I did not sell them as guesses. Just mentioned them here for you to know. –  Jan 11 '12 at 11:52
  • 4
    And I did not sell them as guesses... Obviously, you have a lot to learn about the generally accepted practices of attribution in mathematics. You copy verbatim several sentences from a document without mentioning the source, as if these were your own, and you do not even see the problem?! Since you ask, probably rhetorically: Is there an issue with that? let me answer: yes there is! – Did Jan 11 '12 at 12:12
  • @t.b. Yes. And now, a propitious change of username... Maybe momo1729 is supposed to cover the tracks of IsmailLemhadri. – Did Jan 11 '12 at 12:33
  • Just pardon me to say that I did mention the source in a previous comment (http://www.eleves.ens.fr/home/kortchem/cg/complexes_expbs.pdf). What I want to insist on is that you should not view my comment as an attempt to show you that I have made guesses about the problem ; this would simply be useless because I openly mentioned that I had not tackled it yet. –  Jan 11 '12 at 12:36
  • And by the way, the username change certainly does not come with the intention of hiding anything or whatsoever. I guess I should just not publish my real name on the Internet -that's something I should have done from the day I registered, even though M.SE encourages users to do so. –  Jan 11 '12 at 12:39

1 Answers1

4

Fix $\delta>0$. To solve the first question, we apply Rouché's theorem to the circle of center $0$ and radius $\max\{1,S\}+\delta$. Indeed, if $|z|=\max\{1,S\}+\delta$, $$|P(z)-a_nz^n|=\left|\sum_{j=0}^{n-1}a_jz^j\right|\leqslant \sum_{j=0}^{n-1}|a_j||z|^j<\sum_{j=0}^{n-1}|a_j|(\max\{1,S\}+\delta)^n\leqslant |a_nz^n|.$$

Milten
  • 7,031
Davide Giraudo
  • 172,925