Assuming Goldbach's conjecture, let's define for a sufficiently large integer $ n $ the quantity $ r_{0}(n) : =\inf\{r\geq 0,(n-r,n+r)\in\mathbb{P}^{2}\} $.

Under GRH, what is the best upper bound for $ R(x) : =\sum_{n\leq x}r_{0}(n) $?

  • $\begingroup$ It's certainly $O(x^2)$ :) $\endgroup$
    – Wojowu
    Feb 26, 2018 at 16:24
  • $\begingroup$ And less than $ \frac{x(x-1)}{2} $! $\endgroup$ Feb 26, 2018 at 17:09
  • 6
    $\begingroup$ I think this question suffers from a common barrier: "assuming Goldbach's conjecture" doesn't postulate or imply any sort of quantitative information on the sizes of the primes involved. (Of course we have heuristics, but that doesn't sound like what you're looking for here.) And GRH isn't known to be of any help here. So I'd be surprised if one could do any better with GRH than without GRH. One could try to modify existing almost-all results for Goldbach's conjecture to get $o(x^2)$ for the sum you propose.... $\endgroup$ Feb 26, 2018 at 18:21

1 Answer 1


By a result of Jia (Three primes theorem in a short interval. VII., Acta Math. Sinica (N.S.) 10 (1994), 369-387), we have for any $\epsilon>0$, $$\#\{n\leq x:\ r_0(n)>x^{7/12+\epsilon}\}\ll\frac{x}{\log^2 x}. $$ As a consequence, we have the uniform bound (under the Goldbach conjecture) $$R(x)\ll\frac{x^2}{\log^2 x}.$$ I believe that with current technology one can also establish, for some small but effective $\delta>0$, $$\#\{n\leq x:\ r_0(n)>x^{1-\delta}\}\ll x^{1-\delta}.$$ This would yield (under the Goldbach conjecture) the stronger bound $R(x)\ll x^{2-\delta}$.

Added. Baker and Harman (The three primes theorem with almost equal summands, R. Soc. Lond. Philos. Trans. Ser. A Math. Phys. Eng. Sci. 356 (1998), no. 1738, 763–780) improved the exponent $7/12+\epsilon$ above to $4/7$.

  • 1
    $\begingroup$ GH, I didn't notice it at first look, but I think you forgot a hash sign, cause a set can't be upper bounded per se. $\endgroup$ Feb 26, 2018 at 21:30
  • $\begingroup$ @SylvainJULIEN: Thank you! I fixed this. $\endgroup$
    – GH from MO
    Feb 26, 2018 at 21:39
  • $\begingroup$ Maybe a further interesting question would be to investigate the behavior of $R_{t}(x) : =\sum_{n\leq x}r_{0}(n)^{t} $ as well as the map $ t\mapsto\frac{\partial}{\partial t}R_{t}(x) $ for $ x $ fixed. $\endgroup$ Feb 26, 2018 at 22:04
  • $\begingroup$ @SylvainJULIEN: That's equivalent to investigating the distribution of $r_0(n)$ as a whole. $\endgroup$
    – GH from MO
    Feb 26, 2018 at 22:23

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.