1

I am reading Prof. Ram Murty's Prime number and Irreducible polynomials. I am having problem in understanding a part of the following lemma:

Statement: Suppose that $\alpha$ is a complex root of a polynomial $$ f(x)=x^m+a_{m-1}x^{m-1}+\cdots+a_1x+a_0 $$ with coefficients $a_i$ equal to $0$ or $1$. If $|arg \ \alpha| \le \frac{\pi}{4}$ then $|\alpha| < \frac{3}{2}$. Otherwise $\mathfrak{R}(\alpha) < \frac{1+\sqrt{5}}{2\sqrt{2}}$.

Proof: The cases $m=1$ and $m=2$ can be verified directly. Assuming that $m \ge 3$, we compute for $z \ne 0$:

$$ \left| \frac{f(z)}{z^m}\right| \ge \left| 1+\frac{a_{m-1}}{z}+\frac{a_{m-2}}{z^2}\right| - \left(\frac{1}{|z|^3}+\cdots+\frac{1}{|z|^m}\right). $$

For $z$ satisfying $|arg \ z| \le \frac{\pi}{4}$ it is true that $\mathfrak{R}(\frac{1}{z^2}) \ge 0$, so for such $z$ we have

$$ \left|\frac{f(z)}{z^m}\right| > 1 - \frac{1}{|z|^2(|z|-1)} $$


  1. How to obtain the cases when $m=1,2$?

  2. How are the above two inequalities obtained?

  3. Where are we using the fact: For $z$ satisfying $|arg \ z| \le \frac{\pi}{4}$ it is true that $\mathfrak{R}(\frac{1}{z^2}) \ge 0$?

Saikat
  • 1,583
  • 2
    For $m = 1$, the possibilities for the polynomials are $x$ and $x + 1$. The possibilities for $\alpha$ are $0$ and $-1$. Similarly, for $m = 2$, there are only four polynomials and the possible $\alpha$s you can get $0$, $\pm \iota$, $-1$, $\frac{1}{2}(-1 \pm \sqrt{3} \iota)$. You can verify the claim individually for each. – Aryaman Maithani Jul 04 '21 at 10:54
  • What do you mean by "the above 2 inequality"? Do you mean both inequalities, or just the second one? – Paul Sinclair Jul 04 '21 at 20:21
  • @PaulSinclair: Sorry for the confusion. I mean both the inequalities. I am making it correct. – Saikat Jul 05 '21 at 00:53

1 Answers1

1

Here is a proof of the first inequality.

$$f(z) = z^m + a_{m-1}z^{m-1} + a_{m-2}z^{m-2} + a_{m-3}z^{m-3}+ \dots+a_0\\ \frac{f(z)}{z^m} = \left(1 + \frac{a_{m-1}}z + \frac{a_{m-2}}{z^2}\right) + \left(\frac{a_{m-3}}{z^3}+ \dots+\frac{a_0}{z^m}\right)\\ \frac{f(z)}{z^m} - \left(\frac{a_{m-3}}{z^3}+ \dots+\frac{a_0}{z^m}\right)= \left(1 + \frac{a_{m-1}}z + \frac{a_{m-2}}{z^2}\right)$$ $$\left|\frac{f(z)}{z^m}\right| + \left(\frac{|a_{m-3}|}{|z|^3}+ \dots+\frac{|a_0|}{|z|^m}\right)\ge \left|1 + \frac{a_{m-1}}z + \frac{a_{m-2}}{z^2}\right|\tag{1}$$ $$\left|\frac{f(z)}{z^m}\right| + \left(\frac{1}{|z|^3}+ \dots+\frac{1}{|z|^m}\right)\ge \left|1 + \frac{a_{m-1}}z + \frac{a_{m-2}}{z^2}\right|\tag{2}\\ \left|\frac{f(z)}{z^m}\right|\ge \left|1 + \frac{a_{m-1}}z + \frac{a_{m-2}}{z^2}\right| - \left(\frac{1}{|z|^3}+ \dots+\frac{1}{|z|^m}\right)$$ Where $(1)$ is the triangle inequality, and $(2)$ is because $\forall i, |a_i| \le 1$.

[this only works for $|z| > 1$]For the second inequality, note that if $|z| < 1$, then $$\begin{align}\frac{1}{|z|^3}+ \dots+\frac{1}{|z|^m} &= \frac{1}{|z|^3}\left(1 + |z|^{-1} + \dots+|z|^{-(m-3)}\right)\\ &=\frac{1}{|z|^3}\left(\dfrac{1-|z|^{-(m-2)}}{1-|z|^{-1}}\right)\\ &\le\frac{1}{|z|^2(|z|-1)}\end{align}$$ and if $|z| > 1$, then $$\begin{align}\frac{1}{|z|^3}+ \dots+\frac{1}{|z|^m} &\le \frac{1}{|z|^3}\left(1 + |z|^{-1} + \dots\right)\\ &=\frac{1}{|z|^3}\left(\dfrac1{1-|z|^{-1}}\right)\\ &=\frac{1}{|z|^2(|z|-1)}\end{align}$$ So if we can show $\left|1 + \frac{a_{m-1}}z + \frac{a_{m-2}}{z^2}\right| \ge 1$, the inequality you are after follows. This is where $\mathfrak R\left(\frac1{z^2}\right) \ge 0$ comes in. Though not mentioned, the argument restriction also implies $\mathfrak R\left(\frac1{z}\right) \ge 0$. Hence $$\mathfrak R\left(1 + \frac{a_{m-1}}z + \frac{a_{m-2}}{z^2}\right) \ge 1 + 0 + 0 = 1$$ But the modulus of any number with real part $\ge 1$ must be $\ge 1$ as well. Thus $$\left|1 + \frac{a_{m-1}}z + \frac{a_{m-2}}{z^2}\right| \ge 1$$ as desired.

Paul Sinclair
  • 43,643
  • I think that $|a_i| \ge 1$. Am I correct? – Saikat Jul 05 '21 at 01:45
  • No. The $a_i$ are all $0$ or $1$. Note that I also added an explanation for the second inequality. Egglog is correct in why the argument restriction implies $\mathfrak R\left(\frac 1{z^2}\right) \ge 0$. – Paul Sinclair Jul 05 '21 at 01:56
  • Yes yes, you are correct. I am sorry. – Saikat Jul 05 '21 at 01:58
  • I also just changed the argument for the second inequality a little, as the original argument implicitly assumed $|z| > 1$ to allow the infinite series to converge. – Paul Sinclair Jul 05 '21 at 02:11
  • 1
    @Saikat - my proof of the second inequality is flawed. It works fine for $|z| > 1$, but for $|z| < 1$, it does not work. I'll think about it some more. – Paul Sinclair Jul 05 '21 at 15:03
  • Can you please help me understand how do we get $(1)$ from triangle inequality? – Saikat Jul 05 '21 at 20:32
  • 1
    $$|a - b - c| = |a - (b + c)| \le |a| + |-(b+c)| = |a| + |b+c| \le |a| + |b| + |c|$$ and similarly for higher numbers of summands. – Paul Sinclair Jul 06 '21 at 01:17