On Shusterman’s Goldbach-type problem for sign patterns of the Liouville function
Alexander P. Mangerel · 2024
Read On Shusterman’s Goldbach-type problem for sign patterns of the Liouville function by Alexander P. Mangerel (2024) 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-2412-17199.txt.
Introduction
Alexander Mangerel wrote a mathematics paper in two thousand twenty-four on a Goldbach-style question about the signs of the Liouville function. That function labels each whole number plus or minus one based on its prime factors. Shusterman asked whether every large even number splits into two parts that both carry the minus label.
Assuming a famous unproven hypothesis about prime patterns, Mangerel answers yes for all large enough even numbers. The result rests on a counting bound: for prime totals, every pair of sign patterns occurs often among the splits. The proof leans on a classical expansion of fractions, the Pierce expansion, which may serve other two-part problems. The statement mirrors the Goldbach idea that even numbers are sums of two primes.
Catalog record
| CREATOR | Alexander P. Mangerel |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 74 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-17199
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
On Shusterman’s Goldbach-type problem for sign patterns of the Liouville function Alexander P. Mangerel Abstract . Let λ \lambda be the Liouville function. Assuming the Generalised Riemann Hypothesis for Dirichlet L L -functions (GRH), we show that for every sufficiently large even integer N N there are a , b ≥ 1 a,b\geq 1 such that a + b = N and λ ( a ) = λ ( b ) = − 1 . a+b=N\text{ and }\lambda(a)=\lambda(b)=-1. This conditionally answers an analogue of the binary Goldbach problem for the Liouville function, posed by Shusterman. The latter is a consequence of a quantitative lower bound on the frequency of sign patterns attained by ( λ ( n ) , λ ( N − n ) ) (\lambda(n),\lambda(N-n)) , for sufficiently large primes N N . We show, assuming GRH, that there are constants C , N 0 > 0 C,N_{0}>0 such that for each pattern ( η 1 , η 2 ) ∈ { − 1 , + 1 } 2 (\eta_{1},\eta_{2})\in\{-1,+1\}^{2} and each prime N ≥ N 0 N\geq N_{0} , | { n < N : ( λ ( n ) , λ ( N − n ) ) = ( η 1 , η 2 ) } | ≫ N e − C ( log log N ) 6 . |\{n<N:(\lambda(n),\lambda(N-n))=(\eta_{1},\eta_{2})\}|\gg Ne^{-C(\log\log N)^{6}}. The proof makes essential use of the Pierce expansion of rational numbers n / N n/N , which may be of interest in other binary problems. 1. Introduction 1.1. Main results The binary Goldbach conjecture asserts that every even integer N>2 is representable as N=p+q , where p and q are prime. Equivalently, for every even N>2 we expect the non-trivial lower bound (1) \sum_{1\leq n<N}1_{\mathbb{P}}(n)1_{\mathbb{P}}(N-n)>0 to hold, writing \mathbb{P} to denote the set of all primes. This famous claim remains out of reach of modern methods, even assuming the Generalised Riemann Hypothesis (GRH). Inspired by the heuristic relationship between the distribution of primes and the Liouville function \lambda , it is natural to consider analogous problems about the behaviour of pairs (\lambda(n),\lambda(N-n)) , where 1\leq n<N . Here, in contrast to the case of primes, there is no parity obstruction to the value distribution of \lambda(N-n) , so we need not restrict ourselves to N even. Moreover, there are several different analogous problems that can be posed in this direction. The first concerns obtaining the simplest non-trivial bound on the correlation \mathcal{L}_{\lambda}(N):=\sum_{1\leq n<N}\lambda(n)\lambda(N-n), namely |\mathcal{L}_{\lambda}(N)|<N-1 . In our previous paper [ 7 ] we obtained this bound for all N\geq 11 by elementary arguments. The next analogous problem to ask is about the sign patterns i.e., pairs in \{-1,+1\}^{2} that are attained by (\lambda(n),\lambda(N-n)) , for 1\leq n<N . The result of [ 7 ] showed that the products \lambda(n)\lambda(N-n) must change sign along 1\leq n<N , which implies the existence of integers 1\leq a,b<N such that \lambda(a)=\lambda(N-a)\text{ and }\lambda(b)=-\lambda(N-b). Using the involution b\mapsto N-b , we find that both the patterns 1 1 1 Here and elsewhere we abbreviate +1 and -1 as simply + and - . (+,-) and (-,+) appear among (\lambda(n),\lambda(N-n)) ; however, it is unclear from this which of (+,+) and/or (-,-) is attained by (\lambda(n),\lambda(N-n)) . In this direction, Shusterman [ 16 ] had earlier asked the following problem. Problem 1.1 (Shusterman) . Is it true that for every even integer N>2 there exist positive integers a,b such that a+b=N and \lambda(a)=\lambda(b)=-1 ? Shusterman’s problem is a weaker version of the binary Goldbach problem, as if N is an even integer such that N=p+q with p,q prime then \lambda(p)=\lambda(q)=-1 . For that matter, however, we may ask whether the same property holds for odd integers N as well, and whether for any such N the sign pattern (+,+) , in addition to (-,-) , must occur. It is well-known that a resolution of the binary Goldbach problem via sieve methods alone seems to fail due to the parity problem , namely that sieves are unable to resolve the parity of the number of prime factors of sifted integers (without implementing additional parity-breaking techniques of the kind introduced by Friedlander and Iwaniec, Harman and others). Naturally, Shusterman’s problem offers the same challenge. In this note we address problems like Problem 1.1 , conditionally on the GRH for Dirichlet L -functions (see Section 1.2 below for a discussion concerning this hypothesis, and how it may be weakened). As mentioned, GRH is not known to imply the validity of the binary Goldbach conjecture. Theorem 1.2 . Assume GRH. Then there is an effectively computable N_{0} such that if N\geq N_{0} is even then we can find 1\leq a,b<N such that a+b=N and \lambda(a)=\lambda(b)=-1 . The latter result is a consequence of the more general Theorem 1.3 below. Before stating it we recall that given y\geq 2 , an integer n\in\mathbb{N} is called y -friable (or y -smooth ) if p|n and p\in\mathbb{P} then p\leq y . Theorem 1.3 . Assume GRH. Let N be a positive integer. Then there is an effectively computable p_{0}\geq 2 such that either: (1) N is p_{0} -friable, or (2) for every (\eta_{1},\eta_{2})\in\{-1,+1\}^{2} there is a positive integer 1\leq a<N such that \displaystyle(\lambda(a),\lambda(N-a))=(\eta_{1},\eta_{2}). In particular, case (2) occurs for all sufficiently large 2 2 2 The word “squarefree” here may of course be replaced by “ k th power free”, for any k\geq 2 . squarefree integers N . Theorem 1.3 asserts that each of the sign patterns in \{-1,+1\}^{2} occurs at least once in the sequence of pairs (\lambda(n),\lambda(N-n)) , 1\leq n<N . This theorem follows from a significantly stronger quantitative assertion on the number of such sign patterns, when N is a sufficiently large prime . As mentioned, the main result of [ 7 ] yields |\mathcal{L}_{\lambda}(N)|<N-1 . The proof of Theorem 1.3 is based on the following key proposition that strengthens this last condition. Proposition 1.4 . Assume GRH. Then there is a constant C>0 such that for any prime N , |\mathcal{L}_{\lambda}(N)|\leq N-Ne^{-C(\log\log N)^{6}}. Consequently, for each pair (\eta_{1},\eta_{2})\in\{-1,+1\}^{2} there are 3 3 3 Note that we have not attempted to find the optimal power of \log\log N here, and that the exponent 6 here could be improved somewhat; however, our argument currently cannot be used to bring it down below 1 . \gg Ne^{-C(\log\log N)^{6}} integers 1\leq n<N such that (\lambda(n),\lambda(N-n))=(\eta_{1},\eta_{2}) . 1.2. Remarks on the main theorems Remark 1 . In the proofs of both Theorem 1.2 and Theorem 1.3 we do not actually need the full force of GRH. In fact, a zero-free region of the shape \left\{s\in\mathbb{C}:\text{Re}(s)>1-\frac{1}{(\log N)^{c}},\,|\text{Im}(s)|\leq(\log N)^{3}\right\}, where c\in(0,3/50) , would suffice for our purposes (albeit giving a weaker lower bound for the frequency of each sign pattern in Proposition 1.4 ). See Remark 9 below for a proof of this. Remark 2 . There has been much recent progress in studying the frequency and distribution of sign patterns of the Liouville function (and indeed other real-valued, non-pretentious multiplicative functions) as a consequence of the work of Matomäki, Radziwiłł, Tao, Teräväinen and others on, in particular, correlations of non-pretentious multiplicative functions, see e.g.,