From this page: https://en.wikipedia.org/wiki/Chebyshev_function#Asymptotics_and_bounds

A theorem due to Erhard Schmidt states that, for some explicit positive constant $K$, there are infinitely many natural numbers $x$ such that $$ψ(x)>x+K√x$$

Since the original paper is in German. I read this paper: https://projecteuclid.org/download/pdf_1/euclid.acta/1485887467

My question is: Is the result of Schmidt conditional to Riemann Hypothesis.

  • 1
    $\begingroup$ Even if Schmidt's proof uses RH, it is relatively easy to prove much stronger lower bounds assuming RH is false. $\endgroup$
    – Wojowu
    Dec 28 '19 at 11:36
  • $\begingroup$ @Wojowu: I know that. I am asking on that result if it is related to RH or its negation. $\endgroup$
    – Safwane
    Dec 28 '19 at 11:38

1. It is known unconditionally that, as $x$ tends to infinity, $$\psi(x)-x=\Omega_{\pm}(x^{1/2}).\tag{$1$}$$ This is Corollary 15.4 in Montgomery-Vaughan: Multiplicative number theory I.

2. In fact Hardy and Littlewood proved the stronger result $$\psi(x)-x=\Omega_{\pm}(x^{1/2}\log\log\log x).\tag{$2$}$$ This is Theorem 15.11 in Montgomery-Vaughan: Multiplicative number theory I.

3. Schmidt (1903) proved the analogue of $(1)$ for the function $$f(x):=\sum_{k=1}^\infty\frac{1}{k}\pi(x^\frac{1}{k}).$$ His proof is essentially the same as of the above quoted Corollary: if the Riemann Hypothesis is false, then one has a better result, while if the Riemann Hypothesis is true, then one has a precise form of the stated result with an implied constant given in terms of the lowest lying zero of $\zeta(s)$. So Schmidt's result is unconditional as well, but it differs slightly from the statement attributed to him in the Wikipedia article.

4. Hardy and Littlewood (1916) attribute $(1)$ to Schmidt, and they quote it as Theorem 2.241. Precisely, they say that "This is substantially the well-known result of Schmidt". The stronger statement $(2)$ is Theorem 5.8 in their paper.

P.S. As Greg Martin kindly pointed out, $(2)$ is really due to Littlewood (1914).

  • $\begingroup$ But still the result appeared in Wikipedia is true and unconditional as I read in the linked paper. Page 139, Theorem 2.241. $\endgroup$
    – Safwane
    Dec 28 '19 at 12:39
  • 1
    $\begingroup$ @Helena: I agree, and this is what item 1 in my post is about. I also added items 2 and 4 in my post. $\endgroup$
    – GH from MO
    Dec 28 '19 at 14:00
  • 4
    $\begingroup$ Remark: the result (2) is actually due to Littlewood alone (even though its full proof first appeared in a paper of Hardy and Littlewood). $\endgroup$ Dec 29 '19 at 5:30
  • 1
    $\begingroup$ @GregMartin: Thank you! See the "P.S." section in my post. $\endgroup$
    – GH from MO
    Dec 29 '19 at 6:02

It is elementary, take $K=1/30 < 1/|\rho_0|$ where $\rho_0\approx 1/2+i14.13$ is the first zero, if $\psi(x)-x\le Kx^{1/2}$ for $x$ large enough, then $x-\psi(x)+K x^{1/2}+C\ge 0$ for all $x$, where $C$ is a suitable real constant. Let $$F(s):=\int_1^\infty (x-\psi(x)+Kx^{1/2}+C) x^{-s-1}dx= \frac1{s-1}+\frac{\zeta'(s)}{s\zeta(s)}+\frac{K}{s-1/2}+\frac{C}{s}.$$ By the non-negativity of the integrand, it has a singularity at its abscissa of convergence $\sigma$. But the RHS is analytic on $(1/2,\infty)$, thus $\sigma=1/2$. And by the non-negativity again we have, as $\Re(s) \to 1/2$, $$|F(s)|\le F(\Re(s))\sim \frac{K}{\Re(s)-1/2}.$$ This contradicts that $$F(s)\sim\frac{\zeta'(s)}{s\zeta(s)}\sim\frac{1/\rho_0}{s-\rho_0}\qquad\text{as $s\to \rho_0$}.$$

  • 3
    $\begingroup$ Showing that $K$ arbitrary large works should be much more difficult $\endgroup$
    – reuns
    Dec 29 '19 at 2:20

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.