Extremal lattices and the construction of lattice rules (Q620983)

Obtaining good lattice rules for the approximation of integrals is of essential interest in the theory of numerical integration. The present paper has review character and its main purpose is to survey some of recent results about obtaining extremal lattice rules and the construction of good lattice rules. In the Introduction some notations and concepts of the theory of the trigonometric degree of accuracy are presented. Notions of trigonometric degree, enhanced trigonometric degree, an effectiveness factor of a lattice rules, are reminded. The notion of a dual lattice, the construction of the dual lattice, and its importance are considered. In Section 2, the idea of searching good lattice rules is presented. The matrix approach to construct \(K\)-optimal lattice rules is shown and the notion of extremal lattice is introduced. In Section 3, the problem of how the extremal lattice can be obtained and how such a lattice can be used to bootstrap the construction of lattice rules of a higher degree, is considered. In Subsections 3.1 and 3.2, respectively, the three- and the four-dimensional cases are considered. Details, as a matrix \(B_{3}\) which generates an admissible lattice, coordinate matrix are developed. It is shown how \(B_{3}\) can be used to construct lattice rules of degree \(6k-1.\) By analogy with the three-dimensional case, the four-dimensional case is considered. In Subsection 3.3 the details of some of the higher dimensional cases are discussed. In dimensions \(n=5\) and \(n=6\) the extremal lattices are obtained. In Section 4, the problem of how the extremal lattice can be used to obtain closed expressions for sequences of lattice rules is considered. In Theorem 4 for any \(k \geq 1\) and \(r \in \{-4, -3, -2, -1, 0, 1 \},\) the trigonometric degree \(6k + r\) of the \(k\)th lattice rule, is obtained. In the Conclusion of the article, both approaches for constructing lattice rules, the first one being a computer search, and the second one comprising methods from the geometry of numbers, are discussed.