On the level of distribution of Goldbach primes and its applications

Mizuki Akeno · 2026

Read On the level of distribution of Goldbach primes and its applications by Mizuki Akeno (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-29559.txt.

Introduction

Akeno proves results about Goldbach primes, the primes that appear in representations of even numbers as a sum of two primes. The Goldbach setting is classic number theory. His object is the set of primes p where both p and the complementary value are prime for a given even target. The key quantity is its level of distribution, a measure of how evenly that set spreads across arithmetic progressions.

He proves that for almost all even integers, this set of Goldbach primes has a stated level of distribution. Two applications follow. First, almost all even integers can be written as a sum of two primes whose difference plus one is an almost-prime with few prime factors. Second, a similar statement holds for almost all even integers divisible by six.

Catalog record

CREATORMizuki Akeno
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE86 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.29559

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

On the level of distribution of Goldbach primes and its applications Mizuki Akeno Abstract . We prove that, for almost all even integers N > 0 N>0 , the set of Goldbach primes ℙ ∩ ( N − ℙ ) \mathbb{P}\cap(N-\mathbb{P}) has a level of distribution 1 / 6 1/6 . As applications, we show that almost all even integers N > 0 N>0 can be written as the sum of two primes p 1 , p 2 p_{1},p_{2} such that p 1 − p 2 + 1 ∈ ℙ 4 p_{1}-p_{2}+1\in\mathbb{P}_{4} . We also prove an analogous result with 2 ​ p 1 ​ p 2 + 1 ∈ ℙ 13 2p_{1}p_{2}+1\in\mathbb{P}_{13} for almost all integers N > 0 N>0 with 6 | N 6\mid N . 2020 Mathematics Subject Classification 11N36, 11P32 1. Introduction Let \mathbb{P} denote the set of primes, and let \mathbb{P}_{k} denote the set of products of at most k primes. It is well known that, for any A>0 and X>2 , one has \sum_{m+n=N}\Lambda(m)\Lambda(n)=\mathfrak{S}(N)N+O\left(\frac{X}{(\ln{X})^{A}}\right) for all but O(X(\ln{X})^{-A}) integers N\in(X/2,X] . Here, \mathfrak{S}(N) is the singular series, defined by \mathfrak{S}(N)=\prod_{p}\left(1+\frac{c_{p}(N)}{\varphi(p)^{2}}\right)=\sum_{q}\frac{\mu(q)^{2}}{\varphi(q)^{2}}c_{q}(N). A Bombieri–Vinogradov type theorem for \mathbb{P}\cap(N-\mathbb{P}) has been studied in the following form: (1) \sum_{d_{1}\leq D_{1}}\sum_{d_{2}\leq D_{2}}\lambda^{(1)}_{d_{1}}\lambda^{(2)}_{d_{2}}\left(\sum_{\begin{subarray}{c}m+n=N\\ m\equiv l_{1}\bmod d_{1}\\ n\equiv l_{2}\bmod d_{2}\end{subarray}}\Lambda(m)\Lambda(n)-M_{0}\right)\ll\frac{X}{(\ln{X})^{A}} for almost all N , where l_{1},l_{2} are non-zero integers, (\lambda^{(1)}_{d}),(\lambda^{(2)}_{d}) are divisor bounded sequences and M_{0}=M_{0}(d_{1},l_{1},d_{2},l_{2},N) is a suitable approximation of the sum over m,n . The case D_{1}=X^{1/2} , D_{2}=1 follows from the Bombieri–Vinogradov theorem and a standard circle method argument. See also [ 8 ] , [ 15 ] . With a minor adaptation to almost-prime weights, this implies that, for almost all positive integers N\equiv 4\bmod 6 , the equation p_{1}+p_{2}=N , p_{1}+2\in\mathbb{P}_{2} is solvable. Tolev’s argument [ 17 ] yields an estimate of the form ( 1 ) with D_{1}=X^{1/2},D_{2}=X^{1/3} . He applied this to show that, for almost all positive N\equiv 4\bmod 6 , the equation p_{1}+p_{2}=N,p_{1}+2\in\mathbb{P}_{5},p_{2}+2\in\mathbb{P}_{7} is solvable. Matomäki’s argument [ 13 ] yields an estimate of the form ( 1 ) with D_{1}=X^{1/2},D_{2}=X^{1/2} , under the assumption that one of the sequences \lambda^{(i)}_{d} is well-factorable. As an application, she proved the solvability of p_{1}+p_{2}=N,p_{1}+2\in\mathbb{P}_{2},p_{2}+2\in\mathbb{P}_{7} for almost all N\equiv 4\bmod 6 . A related result for p_{1}+p_{2}=N,p_{1}+2\in\mathbb{P}_{2},p_{2}+2\in\mathbb{P}_{2} can be obtained, by different methods, from the work of Matomäki and Shao [ 11 ] . See also [ 4 ] and [ 5 ] . One can extend ( 1 ) to the case where l_{i} is allowed to depend on d_{i} for each i=1,2 and D_{1}=X^{1/2},D_{2}=X^{1/3} , with little additional effort. See, for example, [ 1 , Theorem 2.1] . Thus, one may obtain similar results for the equation p_{1}+p_{2}=N,f_{1}(p_{1})\in\mathbb{P}_{r_{1}},f_{2}(p_{2})\in\mathbb{P}_{r_{2}} for suitable low-complexity functions f_{i} such as polynomials and suitable r_{1},r_{2}\geq 2 . The basic strategy in the proofs of the above-mentioned results is to rewrite the contribution of \Lambda in ( 1 ) as \int_{\mathbb{T}}\left(\sum_{d_{1}\leq D_{1}}\lambda_{d_{1}}\sum_{\begin{subarray}{c}m\leq X\\ m\equiv l_{1}\bmod d_{1}\end{subarray}}\Lambda(m)e(m\alpha)\right)\left(\sum_{d_{2}\leq D_{2}}\lambda_{d_{2}}\sum_{\begin{subarray}{c}n\leq X\\ n\equiv l_{2}\bmod d_{2}\end{subarray}}\Lambda(n)e(n\alpha)\right)e(-N\alpha)d\alpha, where e(\cdot)=e^{2\pi i\cdot} and \mathbb{T}=\mathbb{R}/\mathbb{Z} , and then to use major- and minor-arc estimates for \sum_{d_{i}\leq D_{i}}\lambda_{d_{i}}\sum_{\begin{subarray}{c}n\leq X\\ n\equiv l_{i}\bmod d_{i}\end{subarray}}\Lambda(n)e(n\alpha). Thus, in this approach the residue classes l_{i} may depend on d_{i} , but not on N ; this independence is used, in particular, when applying Bessel’s inequality over the minor arc. In this paper, we consider ( 1 ) in the case where l_{i} may also depend on N . We prove the following result. Theorem 1.1 . Let \theta\in(0,1/6) , let A>0 and let k\geq 0 . Let w_{1},w_{2} be smooth functions supported on [\eta,1] for some \eta>0 . For all integers N\in(X/2,X] with O(X(\ln{X})^{-A}) exceptions, we have \sum_{d\leq X^{\theta}}\tau(d)^{k}\sup_{l:(d,l(N-l))=1}|R_{d,l}(N)|\ll\frac{X}{(\ln{X})^{A}} where R_{d,l}(N)=\sum_{\begin{subarray}{c}m+n=N\\ m\equiv l\bmod d\end{subarray}}\Lambda(m)\Lambda(n)w_{1}(m/X)w_{2}(n/X)-\frac{\mathfrak{S}(N)}{\psi_{N}(d)}\sum_{m+n=N}w_{1}(m/X)w_{2}(n/X) and \psi_{N} is defined by \psi_{N}(d)=d\prod_{\begin{subarray}{c}p|d\\ p|N\end{subarray}}\left(1-\frac{1}{p}\right)\prod_{\begin{subarray}{c}p|d\\ p\nmid N\end{subarray}}\left(1-\frac{2}{p}\right). Here and throughout, we use the convention that the supremum over an empty set is 0 . Indeed, we prove a more general version in which one of the \Lambda factors is replaced by a more general sequence such as the indicator function of rough numbers. See Theorem 3.1 . By a slight modification of the result of Maier-Pomerance [ 9 , Theorem 3.1] concerning the distribution of generalized twin primes in arithmetic progressions, one can establish such a result with \theta<c for some absolute constant c>0 . Their argument is based on the work of Montgomery and Vaughan [ 16 ] on the bounds for the exceptional set in the Goldbach conjecture. A computation by the author using standard zero-density estimates for Dirichlet L -functions shows that one may take c=\frac{10}{314}=0.0318\ldots , although this value is probably not optimal. We use methods that are different from those of [ 9 ] in both major and minor arcs. We establish a new exponential sum estimate. See Theorem 2.1 . The exponent 1/6 in Theorem 1.1 comes from this estimate. On the major arc, we use a mean value theorem of Choi and Kumchev [ 3 ] for Dirichlet polynomials, together with an argument of Maynard [ 14 ] . As an application of Theorems 1.1 and 3.1 , we consider the solvability of the equations of the form p_{1}+p_{2}=N,f(p_{1},p_{2})\in\mathbb{P}_{r} for some function f:\mathbb{Z}\times\mathbb{Z}\to\mathbb{Z} and r\geq 2 . Matomäki and Zúñiga-Alterman [ 12 ] showed that one can find \mathbb{P}_{3} if the original and the switched problem have level of distribution at least 0.267 . We extend their result and show that one can find \mathbb{P}_{4} if the original and the switched problem have level of distribution at least 0.1635 . See also the Appendix for similar results for \mathbb{P}_{k},2\leq k\leq 7 . Since 1/6=0.166\ldots>0.1635 , we obtain the following result. Theorem 1.2 . Let A>0 and X\geq 2 . For all but O(X(\ln{X})^{-A}) even integers N\in(X/2,X] , there exist primes p_{1},p_{2} such that p_{1}+p_{2}=N and p_{1}-p_{2}+1 has at most 4 prime factors. The numerical comparison in the Appendix indicates that the weights appearing in previous applications of the weighted sieve do not suffice for this purpose, since the resulting criteria remain above the level 1/6 . See the Appendix. We also use Theorem 1.1 and Richert’s weighted sieve to obtain Theorem 1.3 . Let A>0 and X\geq 2 . For all but O(X(\ln{X})^{-A}) integers N\in(X/2,X] with 6|N , there exist primes p_{1},p_{2} such that p_{1}+p_{2}=N and 2p_{1}p_{2}+1 has at most 13 prime factors. Note that the number 13 arises as 13=2\times 6+1 . 1.1. AI tool disclosure OpenAI Codex (GPT-5.5) was used in the preparation of this manuscript in the following ways: • assistance with the numerical computations reported in the Appendix and with the proof of the asymptotic statement there in the regime k\to\infty ; • assistance with searching for existing results relevant to the manuscript; • proofreading for minor typographical, stylistic, and mathematical errors. Apart from these uses, the arguments and proofs in the paper were written by the author. All AI-assisted material used in the proofs was checked by the author, who takes full responsibility for the content of the paper. 2. Exponential sum estimate In this section, we prove the following estimate for exponential sums. Theorem 2.1 . Let D\leq X and \varepsilon>0 . Let w:\mathbb{R}\to\mathbb{R} be a smooth function supported on [0,1] . Let \alpha\in[0,1],q\geq 1,(q,a)=1,|\alpha-a/q|\leq 1/q^{2} . Then, (2) \sum_{d\sim D}\sum_{b\in\mathbb{Z}/{d\mathbb{Z}}}\left|\sum_{n}\Lambda(n)e((\alpha+b/d)n)w(n/X)\right|^{2}\ll X^{2+\varepsilon}\left(\frac{D}{q}+\frac{qD}{X}+D^{2}X^{-1/3}\right) For the proof

Read the complete text