Abstract: For solving PDE eigen-problem, one of the most difficulties in approximation is the big gap between polynomial behavior of eigen-values and wave behavior of eigen-functions. There are three main discrete approaches in computing: variational, difference and generalized matrix eigenvalues approach.The classical FEM had been successful for solving PDE eigen-problems. However, this approach faces some bottlenecks in high performance computing. Lack of accuracy and over mesh-refining are two of them. Both are close related to approximation theory. A question is arisen: Can we improve the fractal approximation by perturbing discrete Rayleigh-quotient if giving-up the monotone approximation? The answer is “Yes”. Moreover, an analysis on some higher-order schemes will be constructed in the talk. Numerical tests in PDE eigen-problem in 2-D and 3-D are given to match our arguments.

 

About the Speaker: Prof. Jia-Chang Sun graduated from CUST in 1964 and visited abroad for 4 years, including Department of Mathematics of UCSB and NA Group of Yale in 1980-1982, Minnesota Supercomputer Institute and Department of Mathematics of UCLA in 1988-1989, Mathematical Department of CUHK in 1993, Computer Science Department of Corolaro in 1997 and 2004, etc. He has joined in many major projects in China, such as 863 and 973 for 20 years. Now he is a member of academic committee of 973 "New Computational Model of Peta Scale Computing ".

  

Date&Time: July 16, 2013 (Tuesday), 9:30 - 10:30 a.m. 
Location: 606 Conference Room