The spectral theorem says when a tangled matrix can be simplified to pure stretching
Many matrices scramble space in complicated ways, but a special class can be rewritten, in the right coordinates, as nothing more than stretching along perpendicular directions. The spectral theorem identifies that class. Cauchy proved it for real symmetric matrices, and John von Neumann's generalisation is arguably the central result of operator theory.
Diagonalising a matrix means finding a basis in which it becomes diagonal, with numbers only along the main diagonal and zeros elsewhere. Those diagonal numbers are the eigenvalues, and each marks a direction, an eigenvector, that the matrix simply scales without turning. Once in that form, heavy computations collapse into easy arithmetic on the diagonal entries.
The theorem guarantees this for self-adjoint operators, called Hermitian when complex and symmetric when real, and more broadly for so-called normal operators. Self-adjoint means the operator can be moved from one side of an inner product to the other without changing the result. That single property forces every eigenvalue to be a real number, as a short calculation with an eigenvector paired against itself shows.
The proof sketch has a pleasing rhythm. The fundamental theorem of algebra supplies at least one eigenvalue and eigenvector. Self-adjointness then ensures that the space perpendicular to that eigenvector is mapped into itself, so the same argument can be repeated inside that smaller space to find a second, perpendicular eigenvector, and so on. Finite induction finishes the job, producing a full set of mutually perpendicular eigenvectors that can be scaled to unit length.
The payoff is the spectral decomposition: the operator splits into a weighted sum of perpendicular projections, one per eigenvalue, carving the whole space into independent pieces. It is a special case of both the Schur decomposition and the singular value decomposition. Cauchy, who was also the first to treat determinants systematically, handled the finite real symmetric case. Infinite-dimensional spaces need a modified notion of diagonalisation, and there the theorem describes operators that behave like multiplication, which in abstract terms is a statement about commutative C*-algebras.
Source: Spectral theorem
