Some additive problems of Goldbach's type (Q5933531)
From MaRDI portal
scientific article; zbMATH DE number 1599278
Language | Label | Description | Also known as |
---|---|---|---|
English | Some additive problems of Goldbach's type |
scientific article; zbMATH DE number 1599278 |
Statements
Some additive problems of Goldbach's type (English)
0 references
8 March 2002
0 references
Let \(\lambda\) be a positive algebraic irrational number, and write \(E_{\lambda}(N)\) for the number of natural numbers \(n\leq N\) such that the equation \(p_1+[\lambda p_2]=n\) has no solution in primes \(p_1\), \(p_2\), where \([x]\) means the integer part of \(x\). In [Dokl. Math. 59, 193-195 (1999; Zbl 0961.11033)], \textit{G. I. Arkhipov, K. Buriev} and \textit{V. N. Chubarikov} proved the bound \(E_{\lambda}(N)\ll N^{7/9+\varepsilon}\) (throughout, the letter \(\varepsilon\) denotes any fixed positive number, as usual). The paper under review first points out that a sharper estimate for \(E_{\lambda}(N)\) follows immediately from the work of \textit{J. Brüdern, R. J. Cook} and \textit{A. Perelli} [``The values of binary linear forms at prime arguments.'' In: Sieve methods, exponential sums and their applications in number theory, Cambridge Univ. Press (1997; Zbl 0924.11085)]. By the latter work, in fact, one finds that for all but \(O(N^{2/3+\varepsilon})\) natural numbers \(n\leq N\), there exist primes \(p_1\), \(p_2\) such that \[ |p_1+\lambda p_2-n-1/2|<1/4. \] The latter inequality implies \(p_1+[\lambda p_2]=n\), and the superior bound \(E_{\lambda}(N)\ll N^{2/3+\varepsilon}\) thus follows. This fine observation clearly presents the most important idea of this paper. Namely, when an integer part occurs in a diophantine equation, one may consider an associated diophantine inequality instead. By this strategy, one may occasionally obtain a better conclusion than what is established by applying the circle method directly to the original diophantine equation, and this is indeed the case in several interesting problems. When two or more integer parts occur in an equation, or when one looks into asymptotic formulae for the number of solutions of such equations, the situation is not as straightforward as the first example above. But adding some supplementary technique to the main idea, the author manages to deal with these kinds of problems as well, and establishes the following three theorems in this paper: (i) Let \(\lambda_1\), \(\lambda_2\) be positive algebraic numbers such that 1, \(\lambda_1\), \(\lambda_1/\lambda_2\) are linearly independent over the rationals. Then for all but \(O(N^{5/6+\varepsilon})\) natural numbers \(n\leq N\), the equation \([\lambda_1 p_1]+[\lambda_2 p_2]=n\) has a solution in primes \(p_1\), \(p_2\). (ii) Let \(\lambda_1\), \(\lambda_2\), \(\lambda_3\) be positive algebraic numbers such that both triples 1, \(\lambda_1\), \(\lambda_1/\lambda_2\) and 1, \(\lambda_1\), \(\lambda_1/\lambda_3\) are linearly independent over the rationals, and put \(r(n)=\sum (\log p_1)(\log p_2)(\log p_3)\), where the sum is over the primes \(p_1\), \(p_2\), \(p_3\) satisfying \([\lambda_1 p_1]+[\lambda_2 p_2]+[\lambda_3 p_3]=n\). Then one has \[ r(n)=n^2/(2\lambda_1\lambda_2\lambda_3)+ O(n^2e^{-\sqrt{\log n}}). \] Moreover, under the Riemann hypothesis (for the Riemann zeta function alone), the last error term may be replaced by \(O(n^{75/38+\varepsilon})\). (iii) Let \(1<\gamma<3/2\), and put \(\theta=\min\{3/\gamma-2,1/3\}\). Then for all but \(O(N\exp(-(\log N)^{\theta-\varepsilon}))\) natural numbers \(n\leq N\), there exist a prime \(p\) and a natural number \(k\) such that \[ p+[\exp((\log k)^{\gamma})]=n. \] Of course, the method of this paper seems effective for various other problems, too. As another example, the author briefly mentions the investigation of the supremum of the real numbers \(c\) such that the equation \([p_1^c]+[p_2^c]+[p_3^c]=n\) is soluble in primes for every large \(n\).
0 references
Goldbach-type theorems
0 references
circle method
0 references