Explicit formulas for wavelet-homogenized coefficients of elliptic operators (Q861338)

From MaRDI portal
Revision as of 21:26, 29 February 2024 by SwMATHimport240215 (talk | contribs) (‎Changed an Item)
scientific article
Language Label Description Also known as
English
Explicit formulas for wavelet-homogenized coefficients of elliptic operators
scientific article

    Statements

    Explicit formulas for wavelet-homogenized coefficients of elliptic operators (English)
    0 references
    0 references
    9 January 2007
    0 references
    The author employs the method of \textit{M. E. Brewster} and \textit{G. Beylkin} [Appl. Comput. Harmon. Anal. 2, No. 4, 327--349 (1995; Zbl 0840.65047)] to obtain wavelet homogenized equations for discrete approximations to equations involving elliptic differential operators (i) \(d/dx\mu(x)d/dx\), (ii) \(\mu(x)\Delta\). Given a vector \(u(k)\), \(k= 0,\dots, N-1\) expressed on an orthogonal wavelet basis, filters are employed to divide \(u\) into coarse and fine components \(u= u_s+ u_d\), \(u_s(k)\), \(u_d(k)\), \(k= 0,\dots, N/2-1\). An \(N\times N\) matrix \(P\) acting on \(u\) goes into block form \(Pu= \left[\begin{smallmatrix} A_p & B_p\\ C_p & T_p\end{smallmatrix}\right] u\) and (iii) \(P=f\) becomes \(A_p u_d+ B_p u_s= f_d\), \(C_p u_d+ T_p u_s= f_s\). Eliminating \(u_d\) leads to an equation in the \(u_s\) vectors (iv) \(R_p u_s= f_s- C_p A^{-1}_p f_d\), \(R_p= T_p- C_p A^{-1}_p B_p\). The solution of (iv) is the coarse scale average of the solution of (iii) and is a homogenization of (iii). Let (v) \(S=\Delta_-\,V\,\vee\Delta_+\), (vi) \(M= VL\) be discrete approximations to (i), (ii), where \(V\) is a diagonal matrix with entries \(\mu(k+{1\over 2}/N)\). If the values of \(\mu\) are periodic then the Fourier symbols of \(R_S\), \(R_M\) are detemined. The \(R_S\) symbol is found explicitly when \(\mu\) has alternating values \(\alpha\), \(\beta\). When \(\mu\) is not periodic \(R_S\) and \(R_M\) are approximated two ways using a mass-lumping procedure. Results of numerical tests of the accuracy of the approximations are described and show that the methods are effective.
    0 references
    numerical homogenization
    0 references
    elliptic operators
    0 references
    wavelets
    0 references
    numerical examples
    0 references

    Identifiers