In mathematics, the spectral radius of a square matrix is the maximum of the absolute values of its eigenvalues.[1] More generally, the spectral radius of a bounded linear operator is the supremum of the absolute values of the elements of its spectrum. The spectral radius is often denoted by ρ(·).

Definition

edit

Matrices

edit

Let λ1, ..., λn be the eigenvalues of a matrix ACn×n. The spectral radius of A is defined as

 

The spectral radius can be thought of as an infimum of all norms of a matrix. Indeed, on the one hand,   for every natural matrix norm  ; and on the other hand, Gelfand's formula states that  . Both of these results are shown below.

However, the spectral radius does not necessarily satisfy   for arbitrary vectors  . To see why, let   be arbitrary and consider the matrix

 .

The characteristic polynomial of   is  , so its eigenvalues are   and thus  . However,  . As a result,

 

As an illustration of Gelfand's formula, note that   as  , since   if   is even and   if   is odd.

A special case in which   for all   is when   is a Hermitian matrix and   is the Euclidean norm. This is because any Hermitian Matrix is diagonalizable by a unitary matrix, and unitary matrices preserve vector length. As a result,

 

Bounded linear operators

edit

In the context of a bounded linear operator A on a Banach space, the eigenvalues need to be replaced with the elements of the spectrum of the operator, i.e. the values   for which   is not bijective. We denote the spectrum by

 

The spectral radius is then defined as the supremum of the magnitudes of the elements of the spectrum:

 

Gelfand's formula, also known as the spectral radius formula, also holds for bounded linear operators: letting   denote the operator norm, we have

 

A bounded operator (on a complex Hilbert space) is called a spectraloid operator if its spectral radius coincides with its numerical radius. An example of such an operator is a normal operator.

Graphs

edit

The spectral radius of a finite graph is defined to be the spectral radius of its adjacency matrix.

This definition extends to the case of infinite graphs with bounded degrees of vertices (i.e. there exists some real number C such that the degree of every vertex of the graph is smaller than C). In this case, for the graph G define:

 

Let γ be the adjacency operator of G:

 

The spectral radius of G is defined to be the spectral radius of the bounded linear operator γ.

Upper bounds

edit

Upper bounds on the spectral radius of a matrix

edit

The following proposition gives simple yet useful upper bounds on the spectral radius of a matrix.

Proposition. Let ACn×n with spectral radius ρ(A) and a consistent matrix norm ||⋅||. Then for each integer  :

 

Proof

Let (v, λ) be an eigenvector-eigenvalue pair for a matrix A. By the sub-multiplicativity of the matrix norm, we get:

 

Since v ≠ 0, we have

 

and therefore

 

concluding the proof.

Upper bounds for spectral radius of a graph

edit

There are many upper bounds for the spectral radius of a graph in terms of its number n of vertices and its number m of edges. For instance, if

 

where   is an integer, then[2]

 

Symmetric matrices

edit

For real-valued matrices   the inequality   holds in particular, where   denotes the spectral norm. In the case where   is symmetric, this inequality is tight:

Theorem. Let   be symmetric, i.e.,   Then it holds that  

Proof

Let   be the eigenpairs of A. Due to the symmetry of A, all   and   are real-valued and the eigenvectors   are orthonormal. By the definition the spectral norm, there exists an   with   such that   Since the eigenvectors   form a basis of   there exists factors   such that   which implies that

 

From the orthonormality of the eigenvectors   it follows that

 

and

 

Since   is chosen such that it maximizes   while satisfying   the values of   must be such that they maximize   while satisfying   This is achieved by setting   for   and   otherwise, yielding a value of  

Power sequence

edit

The spectral radius is closely related to the behavior of the convergence of the power sequence of a matrix; namely as shown by the following theorem.

Theorem. Let ACn×n with spectral radius ρ(A). Then ρ(A) < 1 if and only if

 

On the other hand, if ρ(A) > 1,  . The statement holds for any choice of matrix norm on Cn×n.

Proof

Assume that   goes to zero as   goes to infinity. We will show that ρ(A) < 1. Let (v, λ) be an eigenvector-eigenvalue pair for A. Since Akv = λkv, we have

 

Since v ≠ 0 by hypothesis, we must have

 

which implies  . Since this must be true for any eigenvalue  , we can conclude that ρ(A) < 1.

Now, assume the radius of A is less than 1. From the Jordan normal form theorem, we know that for all ACn×n, there exist V, JCn×n with V non-singular and J block diagonal such that:

 

with

 

where

 

It is easy to see that

 

and, since J is block-diagonal,

 

Now, a standard result on the k-power of an   Jordan block states that, for  :

 

Thus, if   then for all i  . Hence for all i we have:

 

which implies

 

Therefore,

 

On the other side, if  , there is at least one element in J that does not remain bounded as k increases, thereby proving the second part of the statement.

Gelfand's formula

edit

Gelfand's formula, named after Israel Gelfand, gives the spectral radius as a limit of matrix norms.

Theorem

edit

For any matrix norm ||⋅||, we have[3]

 .

Moreover, in the case of a consistent matrix norm   approaches   from above (indeed, in that case   for all  ).

Proof

edit

For any ε > 0, let us define the two following matrices:

 

Thus,

 

We start by applying the previous theorem on limits of power sequences to A+:

 

This shows the existence of N+N such that, for all kN+,

 

Therefore,

 

Similarly, the theorem on power sequences implies that   is not bounded and that there exists NN such that, for all kN,

 

Therefore,

 

Let N = max{N+, N}. Then,

 

that is,

 

This concludes the proof.

Corollary

edit

Gelfand's formula yields a bound on the spectral radius of a product of commuting matrices: if   are matrices that all commute, then

 

Numerical example

edit

Consider the matrix

 

whose eigenvalues are 5, 10, 10; by definition, ρ(A) = 10. In the following table, the values of   for the four most used norms are listed versus several increasing values of k (note that, due to the particular form of this matrix, ):

k      
1 14 15.362291496 10.681145748
2 12.649110641 12.328294348 10.595665162
3 11.934831919 11.532450664 10.500980846
4 11.501633169 11.151002986 10.418165779
5 11.216043151 10.921242235 10.351918183
       
10 10.604944422 10.455910430 10.183690042
11 10.548677680 10.413702213 10.166990229
12 10.501921835 10.378620930 10.153031596
       
20 10.298254399 10.225504447 10.091577411
30 10.197860892 10.149776921 10.060958900
40 10.148031640 10.112123681 10.045684426
50 10.118251035 10.089598820 10.036530875
       
100 10.058951752 10.044699508 10.018248786
200 10.029432562 10.022324834 10.009120234
300 10.019612095 10.014877690 10.006079232
400 10.014705469 10.011156194 10.004559078
       
1000 10.005879594 10.004460985 10.001823382
2000 10.002939365 10.002230244 10.000911649
3000 10.001959481 10.001486774 10.000607757
       
10000 10.000587804 10.000446009 10.000182323
20000 10.000293898 10.000223002 10.000091161
30000 10.000195931 10.000148667 10.000060774
       
100000 10.000058779 10.000044600 10.000018232

Notes and references

edit
  1. ^ Gradshteĭn, I. S. (1980). Table of integrals, series, and products. I. M. Ryzhik, Alan Jeffrey (Corr. and enl. ed.). New York: Academic Press. ISBN 0-12-294760-6. OCLC 5892996.
  2. ^ Guo, Ji-Ming; Wang, Zhi-Wen; Li, Xin (2019). "Sharp upper bounds of the spectral radius of a graph". Discrete Mathematics. 342 (9): 2559–2563. doi:10.1016/j.disc.2019.05.017. S2CID 198169497.
  3. ^ The formula holds for any Banach algebra; see Lemma IX.1.8 in Dunford & Schwartz 1963 and Lax 2002, pp. 195–197

Bibliography

edit
  • Dunford, Nelson; Schwartz, Jacob (1963), Linear operators II. Spectral Theory: Self Adjoint Operators in Hilbert Space, Interscience Publishers, Inc.
  • Lax, Peter D. (2002), Functional Analysis, Wiley-Interscience, ISBN 0-471-55604-1

See also

edit