TY - JOUR
T1 - Augmented coupling interface method for solving eigenvalue problems with sign-changed coefficients
AU - Shu, Yu Chen
AU - Kao, Chiu Yen
AU - Chern, I. Liang
AU - Chang, Chien C.
N1 - Funding Information:
The work was supported in part by the National Science Council of the Republic of China under Contract Nos. NSC96-2221-E-002-201 , NSC97-2628-M-002-020 and NSC97-2115-M-002-004 . C.-Y. Kao was partially supported by the U.S. National Science Foundation grant DMS-0811003 and Sloan fellowship. She is grateful to the MBI f for the hospitality and support.
PY - 2010/12/10
Y1 - 2010/12/10
N2 - In this paper, we propose an augmented coupling interface method on a Cartesian grid for solving eigenvalue problems with sign-changed coefficients. The underlying idea of the method is the correct local construction near the interface which incorporates the jump conditions. The method, which is very easy to implement, is based on finite difference discretization. The main ingredients of the proposed method comprise (i) an adaptive-order strategy of using interpolating polynomials of different orders on different sides of interfaces, which avoids the singularity of the local linear system and enables us to handle complex interfaces; (ii) when the interface condition involves the eigenvalue, the original problem is reduced to a quadratic eigenvalue problem by introducing an auxiliary variable and an interfacial operator on the interface; (iii) the auxiliary variable is discretized uniformly on the interface, the rest of variables are discretized on an underlying rectangular grid, and a proper interpolation between these two grids are designed to reduce the number of stencil points. Several examples are tested to show the robustness and accuracy of the schemes.
AB - In this paper, we propose an augmented coupling interface method on a Cartesian grid for solving eigenvalue problems with sign-changed coefficients. The underlying idea of the method is the correct local construction near the interface which incorporates the jump conditions. The method, which is very easy to implement, is based on finite difference discretization. The main ingredients of the proposed method comprise (i) an adaptive-order strategy of using interpolating polynomials of different orders on different sides of interfaces, which avoids the singularity of the local linear system and enables us to handle complex interfaces; (ii) when the interface condition involves the eigenvalue, the original problem is reduced to a quadratic eigenvalue problem by introducing an auxiliary variable and an interfacial operator on the interface; (iii) the auxiliary variable is discretized uniformly on the interface, the rest of variables are discretized on an underlying rectangular grid, and a proper interpolation between these two grids are designed to reduce the number of stencil points. Several examples are tested to show the robustness and accuracy of the schemes.
UR - http://www.scopus.com/inward/record.url?scp=77957754784&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=77957754784&partnerID=8YFLogxK
U2 - 10.1016/j.jcp.2010.09.001
DO - 10.1016/j.jcp.2010.09.001
M3 - Article
AN - SCOPUS:77957754784
SN - 0021-9991
VL - 229
SP - 9246
EP - 9268
JO - Journal of Computational Physics
JF - Journal of Computational Physics
IS - 24
ER -