Spectral Theorem

Eigenvalues and eigenvectors of symmetric matrices
The symmetric eigenvalue decomposition theorem
Rayleigh quotients

Eigenvalues and eigenvectors of symmetric matrices

Let

be a square, n times n

symmetric matrix. A real scalar lambda

is said to be an eigenvalue of

if there exist a non-zero vector $u in mathbf{R}^n$ such that

The vector

is then referred to as an eigenvector associated with the eigenvalue lambda

. The eigenvector

is said to be normalized if |u|_2 = 1

. In this case, we have

The interpretation of

is that it defines a direction along

behaves just like scalar multiplication. The amount of scaling is given by lambda

. (In German, the root ‘‘eigen’’, means ‘‘self’’ or ‘‘proper’’). The eigenvalues of the matrix

are characterized by the characteristic equation

where the notation det

refers to the determinant of its matrix argument. The function with values t rightarrow p(t) :=det( t I - A)

is a polynomial of degree

called the characteristic polynomial.

From the fundamental theorem of algebra, any polynomial of degree

has

(possibly not distinct) complex roots. For symmetric matrices, the eigenvalues are real, since lambda = u^TAu

when

, and

is normalized.

Spectral theorem

An important result of linear algebra, called the spectral theorem, or symmetric eigenvalue decomposition (SED) theorem, states that for any symmetric matrix, there are exactly

(possibly not distinct) eigenvalues, and they are all real; further, that the associated eigenvectors can be chosen so as to form an orthonormal basis. The result offers a simple way to decompose the symmetric matrix as a product of simple transformations.

Theorem: Symmetric eigenvalue decomposition

We can decompose any symmetric matrix $A in mathbf{S}^n$ with the symmetric eigenvalue decomposition (SED)

$A = sum_{i=1}^n lambda_i u_iu_i^T = U Lambda U^T, ;; Lambda = mbox{bf diag}(lambda_1,ldots,lambda_n) .$

where the matrix of U := [u_1 , ldots, u_n] is orthogonal (that is, U^TU=UU^T = I_n ), and contains the eigenvectors of , while the diagonal matrix Lambda contains the eigenvalues of .

Here is a proof. The SED provides a decomposition of the matrix in simple terms, namely dyads.

We check that in the SED above, the scalars lambda_i

are the eigenvalues, and u_i

's are associated eigenvectors, since

The eigenvalue decomposition of a symmetric matrix can be efficiently computed with standard software, in time that grows proportionately to its dimension

. Here is the matlab syntax, where the first line ensure that matlab knows that the matrix

is exactly symmetric.

Matlab syntax

>> A = triu(A)+tril(A',-1);
>> [U,D] = eig(A);

Rayleigh quotients

Given a symmetric matrix

, we can express the smallest and largest eigenvalues of

, denoted $lambda_{rm min}$ and $lambda_{rm max}$ respectively, in the so-called variational form

$lambda_{rm min}(A) = min_{x} : left{ x^TAx ~:~ x^Tx = 1 right} , ;; lambda_{rm max}(A) = max_{x} : left{ x^TAx ~:~ x^Tx = 1 right} .$

The term ‘‘variational’’ refers to the fact that the eigenvalues are given as optimal values of optimization problems, which were referred to in the past as variational problems. Variational representations exist for all the eigenvalues, but are more complicated to state.

The interpretation of the above identities is that the largest and smallest eigenvalues is a measure of the range of the quadratic function x rightarrow x^TAx

over the unit Euclidean ball. The quantities above can be written as the minimum and maximum of the so-called Rayleigh quotient x^TAx/x^Tx

Historically, David Hilbert coined the term ‘‘spectrum’’ for the set of eigenvalues of a symmetric operator (roughly, a matrix of infinite dimensions). The fact that for symmetric matrices, every eigenvalue lies in the interval $[lambda_{rm min},lambda_{rm max}]$ somewhat justifies the terminology.