Laplace特征值问题的局部和并行有限元算法-计算数学专业毕业论文.pdfVIP

1 Introduction The local and parallel finite element discretization method was established by Xu and Zhou [6]. In recent years, many researchers have been developing the local and parallel finite element method, which has become an important and efficient numerical method for solving partial differential equations, especially for elliptic boundary value problems. The method has been successfully used for solving eigenvalue problems [2, 5, 6]. This technique is seen as one of the most efficient method so far, and it has a broad space for development in scientific computing. In Xu鈥檚 paper [6], the two authors show us some local and parallel finite element methods, which provide us theoretical foundation for solving different kinds of partial differential equations. And these techniques are continual updating and developing. The main idea of new local and parallel finite element method is to use a relatively coarse grid to approximate the low frequencies of the solution and then to use a fine grid and some linear algebraic systems to correct the resulted residual by some local and parallel procedure. These methods are motivated by the local error estimates. The concrete treat- ment is dividing an initial domain into a number of disjoint subdomains 戆 , . . . , 戆 , then 1 肀 enlarging each 戆 (肀 = 1, . . . , 肀? to obtain 惟 , applying the local algorithms in parallel in 肀 肀 all 惟 . 肀 In Bi鈥檚 recent research [2], some new local and parallel finite element three-scale schemes for a class of eigenvalue problems are established by combing the local defet- correction technique and the shifted-inverse