6
$\begingroup$

I have found that $\left|\frac{\zeta'(x)}{\zeta^2(x)}\right|\leq \frac{1}{x-\frac{1}{2}}$ for all real $x$ such that $x>1$ seems to be true. I have plotted the inequality and got this inequality holding. Can someone help me to prove it?

Edit: It seems to be too hard to prove it so if you know any similar inequalities can you share their proofs?

$\endgroup$
16
  • 1
    $\begingroup$ For $x$ close to $1$ this is true by the Taylor expansion, for $x$ close to $+\infty$ this is true by the Dirichlet series expansion, for the intermediate $x$ this is true by a computer. But this is not a research level question. $\endgroup$ Commented Nov 2 at 8:26
  • 6
    $\begingroup$ @AlekseiKulikov : If for some range this is true only "by a computer", then how can it not be a research question? $\endgroup$ Commented Nov 2 at 9:37
  • 2
    $\begingroup$ @StevenLandsburg in the same way as writing two thousand-digit numbers $A, B$ and a two-thousand-digit number $C$ and asking if $A*B>C$ is not a research level question, even though I will not be able to solve it without a computer. $\endgroup$ Commented Nov 2 at 10:31
  • 2
    $\begingroup$ @AlekseiKulikov in your case the computer can give us the proof since it is based on some algorithm that can do all the computation needed to be sure that the result is true or false but my question is about an infinite continuous range and this function is highly sensitive to precision errors that's why it should be proven before we can believe that it's true in general. $\endgroup$
    – Haidara
    Commented Nov 2 at 10:50
  • 4
    $\begingroup$ @AlekseiKulikov In mathematics, not every heuristic reasoning counts as a proof. Simply checking an inequality for finitely points with a computer with finite precision will always never be a proof! Also, there is more to mathematical results than simply being true or false -- we want to know why they are true (or false!). However, I make no claims that this question here is of particular interest, which I cannot judge. $\endgroup$ Commented Nov 2 at 11:37

1 Answer 1

6
$\begingroup$

In the range $1<x<1.5$, the known bounds $$-\frac{\zeta'(x)}{\zeta(x)}<\frac{1}{x-1}<\zeta(x)$$ imply that $$-\frac{\zeta'(x)}{\zeta^2(x)}<1<\frac{1}{x-1/2}.$$ In the range $1.5\leq x<1.7$, we have by monotonicity that $$-\frac{\zeta'(x)}{\zeta(x)}\leq-\frac{\zeta'(1.5)}{\zeta(1.5)}<1.6<\frac{\zeta(1.7)}{1.7-0.5}<\frac{\zeta(x)}{x-1/2}.$$ In the range $1.7\leq x<2$, we have by monotonicity that $$-\frac{\zeta'(x)}{\zeta(x)}\leq-\frac{\zeta'(1.7)}{\zeta(1.7)}<1<\frac{\zeta(2)}{2-0.5}<\frac{\zeta(x)}{x-1/2}.$$ In the range $2\leq x<2.6$, we have by monotonicity that $$-\frac{\zeta'(x)}{\zeta(x)}\leq-\frac{\zeta'(2)}{\zeta(2)}<0.6<\frac{\zeta(2.6)}{2.6-0.5}<\frac{\zeta(x)}{x-1/2}.$$ In the range $2.6\leq x<3$, we have by monotonicity that $$-\frac{\zeta'(x)}{\zeta(x)}\leq-\frac{\zeta'(2.6)}{\zeta(2.6)}<0.3<\frac{\zeta(3)}{3-0.5}<\frac{\zeta(x)}{x-1/2}.$$ In the range $3\leq x$, the known bound $\psi(r)<1.04 r$ for the second Chebyshev function implies that $$-\frac{\zeta'(x)}{\zeta^2(x)}<-\frac{\zeta'(x)}{\zeta(x)}=x\int_2^\infty\frac{\psi(r)}{r^{x+1}}\,dr<\frac{2.08x}{x-1}2^{-x}<\frac{1}{x-1/2}.$$ The quoted bounds can be found in Delange (1987), Corollary 1.14 of Montgomery-Vaughan: Multiplicative number theory I, and Theorem 12 of Rosser-Schoenfeld (1962).

$\endgroup$
13
  • 2
    $\begingroup$ Just to be sure, in the intermediate regime you did not prove bounds like $-\frac{\zeta'(1.5)}{\zeta(1.5)} < 1.6$ by hand, but rather made an electronic computing machine, colloquially known as a computer, to tell you this number to enough decimal places? $\endgroup$ Commented Nov 3 at 8:53
  • 1
    $\begingroup$ @AlekseiKulikov That's right. Probably there are more elegant ways to verify the original inequality. For example, it would be helpful to know that $-\zeta'(x)/\zeta^2(x)$ is decreasing. $\endgroup$
    – GH from MO
    Commented Nov 3 at 9:32
  • 1
    $\begingroup$ Ok nice proof. Also I think it can be generalized to other inequalities I will try to do it myself. Thank you! $\endgroup$
    – Haidara
    Commented Nov 3 at 9:51
  • 1
    $\begingroup$ @Haidara Please ask your new question in a new post. $\endgroup$
    – GH from MO
    Commented Nov 3 at 12:53
  • 2
    $\begingroup$ @AlekseiKulikov You're a funny guy. $\endgroup$ Commented Nov 3 at 13:52

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .