Diophantine approximation by primes.

*(English)*Zbl 1257.11035Let \(\lambda_1, \lambda_2, \lambda_3\) be non-zero real numbers, not all of the same sign and not all in rational ratio. The author uses the Davenport-Heilbronn method to show that for every real number \(\lambda_0\) and every \(\delta>0\) the Diophantine inequality
\[
|\lambda_0+\lambda_1p_1+\lambda_2p_2+\lambda_3p_3| < (\max p_i)^{-2/9+\delta}
\]
has infinitely many solutions in primes \(p_1, p_2, p_3\). The approach dates to work of R. C. Vaughan [Proc. Lond. Math. Soc., III. Ser. 28, 373–384 (1974; Zbl 0274.10045)], who obtained the exponent \(1/10\) in place of \(2/9\), and incorporates ideas of G. Harman [J. Lond. Math. Soc., II. Ser. 44, No. 2, 218–226 (1991; Zbl 0754.11010)], who had obtained the best previously known exponent of \(1/5\). As expected, one must look for solutions in \((X,2X]^3\), where \(X=q^{3/2}\) and \(q\) is a denominator of a convergent to some \(\lambda_i/\lambda_j\). In fact, it is well-known that there exist coefficients \(\lambda_i\) and arbitrarily large \(X\) for which the inequality has no solution in \((X,2X]^3\). The author’s set-up uses the Brüdern-Fouvry vector sieve to obtain a lower bound for the characteristic function of pairs of primes in terms of upper and lower bounds \(\rho^{\pm}(n)\) for the characteristic function of primes. The functions \(\rho^{\pm}(n)\) are constructed by repeated application of Buchstab’s identity and numerical integration, in such a way that the mean square error in approximating the respective counting functions is small enough to handle the major arc. The author further argues that the exponential sums arising from these \(\rho^{\pm}(n)\) can be decomposed into suitable type I and type II sums, which form the basis of the minor arc analysis. The strategy is to bound the measure of the set on which all three generating functions are nearly as large as \(X^{7/9+\delta}\), and this is the heart of the paper. Large exponential sums \(S(\lambda_i\alpha)\) produce good rational approximations \(a_i/q_i\) to \(\lambda_i\alpha\), and upper bounds for averages of \(S(a_i/q_i+\beta_i)\) over \(q_i\) then yield upper bounds on the number of such \(a_i/q_i\) that can occur. For large moduli the type II average estimates are obtained via the large sieve, and for small moduli they are obtained from classical mean value estimates for Dirichlet polynomials.

Reviewer: Scott T. Parsell (West Chester)

##### MSC:

11D75 | Diophantine inequalities |

11P32 | Goldbach-type theorems; other additive questions involving primes |

11N36 | Applications of sieve methods |

Full Text:
DOI

**OpenURL**

##### References:

[1] | Harman, Prime-detecting sieves, vol. 33: London Mathematical Society Monographs (New Series) (2007) · Zbl 1220.11118 |

[2] | Heath-Brown, Can. J. Math. 34 pp 1365– (1982) · Zbl 0478.10024 |

[3] | DOI: 10.1112/jlms/s2-44.2.218 · Zbl 0754.11010 |

[4] | DOI: 10.1112/jlms/s2-27.1.9 · Zbl 0504.10018 |

[5] | DOI: 10.1007/BF01403187 · Zbl 0219.10048 |

[6] | Brüdern, Compositio Math. 102 pp 337– (1996) |

[7] | Vaughan, The Hardy–Littlewood method, vol. 125: Cambridge Tracts in Mathematics (1997) |

[8] | Baker, Sieve methods, exponential sums and their applications in number theory pp 1– (1997) |

[9] | DOI: 10.1112/plms/s3-28.2.373 · Zbl 0274.10045 |

[10] | DOI: 10.1112/jlms/s2-25.2.201 · Zbl 0443.10015 |

[11] | Schwarz, J. Reine Angew. Math. 212 pp 150– (1963) |

[12] | Baker, J. Reine Angew. Math. 228 pp 166– (1967) |

[13] | Saffari, Ann. Inst. Fourier 27 pp 1– (1977) · Zbl 0379.10023 |

[14] | Iwaniec, Analytic number theory, vol. 53: American Mathematical Society Colloquium Publications (2004) |

[15] | Heath-Brown, J. Reine Angew. Math. 389 pp 22– (1988) |

[16] | DOI: 10.1112/plms/s3-72.2.241 · Zbl 0874.11052 |

This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. It attempts to reflect the references listed in the original paper as accurately as possible without claiming the completeness or perfect precision of the matching.