What Are Matrices?

Matrices are rectangular arrays of numbers, symbols, or expressions arranged in rows and columns, used to represent linear transformations, systems of linear equations, and the relationships among variables in a wide range of physical and computational models. A matrix with m rows and n columns is called an m-by-n matrix, and its entries are identified by their row and column indices. Matrix algebra defines operations including addition, scalar multiplication, and matrix multiplication, each subject to rules that reflect the underlying geometry of linear transformations. The field of linear algebra, built largely on matrix theory, is foundational to virtually every quantitative discipline in science and engineering.

The systematic study of matrices developed through the nineteenth century, with Arthur Cayley formalizing matrix algebra in 1858 and Camille Jordan contributing the canonical form that bears his name. Properties such as eigenvalues, rank, determinant, trace, and condition number characterize a matrix's behavior and are central to stability analysis, data compression, and numerical computation. Engineering problems involving coupled variables, from circuit mesh analysis to mechanical vibration modes, are routinely cast as matrix equations and solved with software libraries such as LAPACK.

Matrix Operations and Decompositions

The core operations of matrix arithmetic, addition, transposition, and multiplication, obey a well-developed algebra in which multiplication is associative but not generally commutative. Matrix decompositions factor a matrix into simpler constituents that reveal its structure and enable efficient computation. The LU decomposition writes a square matrix as a product of lower- and upper-triangular factors, making Gaussian elimination applicable to successive right-hand sides at low cost. The eigenvalue decomposition expresses a diagonalizable matrix as a product of its eigenvector matrix, a diagonal eigenvalue matrix, and the inverse eigenvector matrix, capturing the matrix's action as a set of independent scalings along fixed directions. The singular value decomposition (SVD), which exists for every matrix regardless of shape or rank, provides the basis for least-squares solutions, dimensionality reduction, and principal component analysis. These decompositions are documented rigorously in the LAPACK Users' Guide maintained at Netlib.

Method of Moments

The method of moments is a numerical technique for solving linear operator equations, particularly integral equations that arise in electromagnetics and acoustics, by projecting them onto a finite-dimensional subspace and reducing them to a matrix equation. Unknown source distributions are expanded in a set of basis functions, the governing equation is tested with a set of weighting functions, and the result is a dense N-by-N impedance matrix whose solution yields the expansion coefficients. Roger Harrington formalized the method in the 1960s, and it now underpins commercial antenna and scattering solvers. As described in the National Academies survey of method-of-moments applications to large electromagnetic structures, efficient matrix compression and iterative solvers have extended the method to problems with millions of unknowns.

Mode Matching Methods

Mode matching is an analytical and semi-analytical technique used in guided-wave electromagnetics and acoustics to find scattering matrices and transmission coefficients at discontinuities in waveguides, cavities, and periodic structures. The fields on each side of a discontinuity are expanded in their respective sets of modal basis functions; continuity of tangential field components at the interface yields a system of linear equations that is cast as a matrix equation and solved by truncating the modal expansions. The accuracy improves as more modes are included in the truncation. Mode matching produces sparse, structured matrices that are cheaper to solve than the dense matrices of the method of moments, and it is the preferred approach for problems where the geometry consists of regular sections separated by abrupt junctions. An overview of the technique appears in the arxiv tutorial on method-of-moments and related matrix formulations in computational electromagnetics.

Applications

Matrices have applications in a wide range of disciplines, including:

• Circuit analysis, where nodal admittance matrices and mesh impedance matrices encode network topology
• Antenna and scattering simulation, where the method of moments discretizes integral equations into solvable linear systems
• Structural mechanics and finite-element analysis, where global stiffness and mass matrices govern static and dynamic response
• Machine learning, where matrix factorizations underlie recommendation systems, topic modeling, and neural network training
• Quantum mechanics, where matrix representations of operators describe state evolution and measurement outcomes