Block Matrices

Matrices partitioned into submatrices, letting large structured problems be manipulated in coarse pieces.

Partitioning

A block matrix is a matrix whose entries are grouped into rectangular submatrices, or blocks, treated as single units. When the partitions are compatible, blocks obey the same arithmetic rules as scalars: block addition adds corresponding blocks, and block multiplication follows the row-by-column rule with matrix products in place of scalar products, provided the inner block dimensions conform.

Why partition

Block structure exposes and exploits problem structure. Coupled systems, such as coordinates split into position and momentum, or a discretization split by physical region, naturally form blocks. Working at the block level clarifies the algebra and enables algorithms that solve each block or coupling separately, which is essential for parallel and hierarchical solvers.

Block triangular and diagonal

A block-diagonal matrix has nonzero blocks only on the diagonal; its determinant is the product of the block determinants, and its eigenvalues are the union of the blocks' eigenvalues. A block-triangular matrix shares the determinant property, which is why decoupling a system into block-triangular form simplifies both analysis and solution.

The Schur complement

For a two-by-two block matrix with blocks A, B, C, D, the Schur complement of A is D - C A^{-1} B. It appears when eliminating one block of variables and governs the invertibility and conditioning of the whole system. Schur complements underlie block elimination, domain decomposition, and many preconditioners for large sparse systems.

import numpy as np
A = np.eye(2); B = np.ones((2, 2)); C = np.zeros((2, 2)); D = 2*np.eye(2)
M = np.block([[A, B], [C, D]])
print(M.shape)   # (4, 4) assembled from blocks

Multiphysics simulations that couple, for example, a magnetic-field region to a plasma region assemble their operators as block matrices, and block elimination via Schur complements is how the coupled system is solved efficiently.