Some remarks on the compressed matrix representation of symmetric second-order and fourth-order tensors

The mathematical description of physical phenomena requires the use of scalars and tensors of various order. Since many of the second-order and fourth-order tensors used in continuum mechanics possess certain symmetries, a compressed vector or matrix representation, respectively, is frequently used in computational applications such as the FEM and BEM. Use of different storage schemes for different tensors lead to an hypothetical nonuniqueness of such matrix representations. The present paper offers a clarification by investigation of the structure of the underlying six-dimensional vector space. It identifies various types of matrix representations as covariant, contravariant or mixed-variant coordinates in that vector space and thus proves consistency of the matrix representation with classical tensor analysis in R-3. Furthermore, it is shown that an ortho-normal basis for the underlying tensor representation in R-3 does not automatically lead to a normalized space for the compressed matrix representation in R-6. Thus, distinction of covariant and contravariant coordinates is necessary even in that case. Theoretical findings are worked out in detail for symmetric second-order and fourth-order tensors in R3. Example applications on commonly used fourth-order tensors as well as a comparison of possible computational implementations close the paper. (C) 2001 Elsevier Science B.V. All rights reserved.