9
$\begingroup$

Iwaniec [1] proved that $$ \pi(x+x^\theta)-\pi(x) < \frac{(2+\varepsilon)x^\theta}{\eta(\theta)\log x},\ x>x_0(\varepsilon,\theta). $$ with $$ \eta(\theta)=\frac{15\theta-2}{9}. $$

(Actually, he proves that a function $\eta(\theta)>\theta$ exists, and that this is an admissible choice. This choice gives nontrivial information for $\theta>1/3.$ He gives others like $\eta_1(\theta)=(1+\theta)/2$ for $\theta>1/2.$)

Two other questions have asked about the primes in this interval

Fundamentally, I'd like information on $$ f(\theta) := \limsup_{x\to\infty} \frac{\pi(x+x^\theta)-\pi(x)}{x^\theta/\log x} $$

What modern results are available? The result above is $f(\theta) \le 18/(15\theta-2)$ for $1/3<\theta<1.$

[1] Henryk Iwaniec. On the Brun–Titchmarsh theorem. Journal of the Mathematical Society of Japan, 34:1, pp. 95–123, 1982.

$\endgroup$
1
  • 1
    $\begingroup$ Quibble: I would say that what Iwaniec's choices do is not to give non-trivial information for $\theta>1/3$ or $\theta>1/2$ - rather, it is the case that, for $\theta$ fulfilling those inequalities, Iwaniec gives something better than Brun-Titchmarsh. Of course, if you consider Brun-Titchmarsh (with $q=1$) to be trivial, your statement is correct. $\endgroup$ Jan 7 at 6:30

2 Answers 2

6
$\begingroup$

I found one paper that improves on the quoted result (in a narrow range): $$\eta(\theta )=\frac{100\theta -45}{11}\qquad\text{is admissible for}\qquad \frac{6}{11}<\theta \le \frac{11}{20}.$$ See Lou-Qi: Upper bounds for primes in intervals (Chinese), Chinese Ann. Math. Ser. A 10 (1989), 255-262.

$\endgroup$
1
  • $\begingroup$ Would you happen to read Chinese, GH? $\endgroup$ Jan 15, 2022 at 19:46
6
$\begingroup$

Montgomery (1) gives a list of 40 exponent pairs $(\kappa,\lambda)$ which can be plugged into Iwaniec's formula

$$ \eta(\theta)=\left(1+\frac{1-\lambda+2\kappa}{3-\lambda-\kappa/2}\right)\theta - \frac{\kappa}{3-\lambda-\kappa/2} $$

to yield bounds for $0<\theta\le1/2.$ Of these, 36 are optimal in some interval; adding the zeta function value for $\theta>1/2$ yields

$$ \eta(\theta)=\begin{cases} \theta,&\text{ if }0<\theta\le1/9\\ \frac{1047\theta-2}{1029},&\text{ if }1/9\le\theta\le514/3597\\ \frac{531\theta-2}{515},&\text{ if }514/3597\le\theta\le322/2061\\ \frac{369\theta-2}{354},&\text{ if }322/2061\le\theta\le194/1101\\ \frac{444\theta-4}{417},&\text{ if }194/1101\le\theta\le130/669\\ \frac{189\theta-2}{176},&\text{ if }130/669\le\theta\le46/219\\ \frac{498\theta-8}{451},&\text{ if }46/219\le\theta\le2/9\\ \frac{228\theta-4}{205},&\text{ if }2/9\le\theta\le362/1569\\ \frac{1161\theta-22}{1037},&\text{ if }362/1569\le\theta\le182/743\\ \frac{1167\theta-26}{1027},&\text{ if }182/743\le\theta\le52/201\\ \frac{921\theta-22}{805},&\text{ if }52/201\le\theta\le342/1291\\ \frac{254\theta-8}{215},&\text{ if }342/1291\le\theta\le226/833\\ \frac{669\theta-22}{563},&\text{ if }226/833\le\theta\le118/415\\ \frac{116\theta-4}{97},&\text{ if }118/415\le\theta\le186/641\\ \frac{589\theta-22}{487},&\text{ if }186/641\le\theta\le94/303\\ \frac{587\theta-26}{473},&\text{ if }94/303\le\theta\le310/959\\ \frac{1439\theta-66}{1153},&\text{ if }310/959\le\theta\le286/855\\ \frac{1327\theta-66}{1049},&\text{ if }286/855\le\theta\le127/376\\ \frac{351\theta-22}{265},&\text{ if }127/376\le\theta\le74/217\\ \frac{393\theta-26}{293},&\text{ if }74/217\le\theta\le45/128\\ \frac{325\theta-22}{241},&\text{ if }45/128\le\theta\le62/171\\ \frac{347\theta-26}{251},&\text{ if }62/171\le\theta\le37/98\\ \frac{281\theta-22}{201},&\text{ if }37/98\le\theta\le66/173\\ \frac{823\theta-66}{585},&\text{ if }66/173\le\theta\le426/1093\\ \frac{317\theta-26}{224},&\text{ if }426/1093\le\theta\le682/1733\\ \frac{1243\theta-114}{851},&\text{ if }682/1733\le\theta\le438/1103\\ \frac{1601\theta-150}{1089},&\text{ if }438/1103\le\theta\le36/89\\ \frac{1651\theta-162}{1107},&\text{ if }36/89\le\theta\le486/1181\\ \frac{659\theta-66}{439},&\text{ if }486/1181\le\theta\le474/1141\\ \frac{1465\theta-150}{969},&\text{ if }474/1141\le\theta\le1626/3865\\ \frac{595\theta-66}{383},&\text{ if }1626/3865\le\theta\le120/281\\ \frac{1427\theta-162}{911},&\text{ if }120/281\le\theta\le422/973\\ \frac{1249\theta-150}{781},&\text{ if }422/973\le\theta\le846/1921\\ \frac{923\theta-114}{571},&\text{ if }846/1921\le\theta\le1542/3469\\ \frac{470\theta-60}{287},&\text{ if }1542/3469\le\theta\le34/75\\ \frac{15\theta-2}{9},&\text{ if }34/75\le\theta\le1/2\\ \frac{\theta+1}{2},&\text{ if }1/2<\theta\le7/12\\ 2,&\text{ if }7/12<\theta<1. \end{cases} $$

This is hardly a satisfactory answer, giving only 20-year-old results and neglecting Vinogradov's method for small $\theta,$ but it may be useful so I'll leave it here.

I’ve added Huxley’s result $$ \pi(x)-\pi(x-y) \sim \frac{y}{\log x}\text{ for }x^\theta \le y \le x/2 $$ with $\theta>7/12$.

  1. Hugh L. Montgomery, Harmonic Analysis as found in Analytic Number Theory, in Twentieth Century Harmonic Analysis—A Celebration, Springer, 2001
  2. M. N. Huxley, On the difference between consecutive primes, Invent. Math. 15 (1972), pp. 164-170.
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.