Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated


Source. Theorem 2.3 (p. 5) and Corollaries 2.5, 2.7 and 2.8 (p. 8), Section 2, of Nathan McNew, The convex hull of the prime number graph, in: Irregularities in the Distribution of Prime Numbers, Springer, Cham (2018), 125--141, doi:10.1007/978-3-319-92777-0_7, cited at the page numbers 1--15 of the author's preprint named on the source card.

Statement

Setting (pp. 1--3). The prime number graph is the set of points (n,pn)(n,p_n), pnp_n the nnth prime. A convex prime is a prime pnp_n for which (n,pn)(n,p_n) is a vertex of the convex hull of this graph; c1<c2<⋯c_1<c_2<\cdots are the indices of the convex primes, so the convex primes are pc1<pc2<⋯p_{c_1}<p_{c_2}<\cdots.

Theorem 2.3 (p. 5). There is a constant B>0B>0 such that for all sufficiently large ii,

pci+1−pci≤pciexp⁡{−Blog⁡3/5pci(log⁡log⁡pci)1/5}.p_{c_{i+1}}-p_{c_i}\le p_{c_i}\exp\Bigl\{\frac{-B\log^{3/5}p_{c_i}}{(\log\log p_{c_i})^{1/5}}\Bigr\}.

Corollary 2.5 (p. 8). There is a constant B>0B>0 such that the number of convex primes up to xx is at least

exp⁡{Blog⁡3/5x(log⁡log⁡x)1/5}.\exp\Bigl\{\frac{B\log^{3/5}x}{(\log\log x)^{1/5}}\Bigr\}.

Corollary 2.7 (p. 8). The sum ∑i≥11/log⁡pci\sum_{i\ge1}1/\log p_{c_i} diverges. The paper derives it from the lower bound of Corollary 2.5 and presents it as the proof of Tutaj's Conjecture 1.3 (p. 3).

Corollary 2.8 (p. 8). lim⁡i→∞pci+1/pci=1\lim_{i\to\infty}p_{c_{i+1}}/p_{c_i}=1. This is Tutaj's Theorem 1.1 (p. 3), proved by Tutaj under the Riemann Hypothesis; here it holds unconditionally, since Theorem 2.3 gives pci+1−pci=o(pci)p_{c_{i+1}}-p_{c_i}=o(p_{c_i}).

The paper does not say that the BB of Corollary 2.5 is the BB of Theorem 2.3. Page 7 says that Corollary 2.5 settles a claim Pomerance made without proof, of a count at least eclog⁡3/5−ϵxe^{c\log^{3/5-\epsilon}x} for some c>0c>0.

Read depth. Claims checked: Theorem 2.3 and Corollaries 2.5, 2.7 and 2.8 were read clause by clause on the page images of the preprint; the proof of Theorem 2.3 (pp. 5--7) was read but not checked, and nothing here is independently reviewed.

Proof pointer

pp. 5--7. The prime number theorem with the best known error term puts every point of the graph between the two convex curves x=li(y)∓Dyexp⁡{−Alog⁡3/5y/(log⁡log⁡y)1/5}x=\mathrm{li}(y)\mp Dy\exp\{-A\log^{3/5}y/(\log\log y)^{1/5}\}. A boundary segment of the hull lies between them. Comparing its midpoint with the inner curve through a Taylor expansion of li\mathrm{li} bounds its vertical extent by a constant times ylog⁡yexp⁡{−Alog⁡3/5y/(2(log⁡log⁡y)1/5)}y\log y\exp\{-A\log^{3/5}y/(2(\log\log y)^{1/5})\}, which gives the gap bound. The corollaries follow from the gap bound (pp. 7--8).

Dependencies

The prime number theorem with error term O(xexp⁡{−Alog⁡3/5x/(log⁡log⁡x)1/5})O(x\exp\{-A\log^{3/5}x/(\log\log x)^{1/5}\}), equation (5), p. 5.

Bears on

No Erdős problem directly. Since every convex prime is a midpoint convex prime (p. 2), Corollary 2.5 bounds from below the number of nn with Mn>0M_n>0 in the notation of equation (23); it says nothing about how large MnM_n can be, which is what Problem 454 asks.