Fact-checked by Grok 2 weeks ago

Spectral radius

The spectral radius of a square A \in \mathbb{C}^{n \times n} is defined as the largest among its eigenvalues, mathematically expressed as \rho(A) = \max \{ |\lambda| : \lambda is an eigenvalue of A \}. This quantity serves as a fundamental invariant in linear algebra, bounding the growth or decay of matrix powers and informing the properties of iterative algorithms. For any consistent \|\cdot\|, the spectral radius satisfies \rho(A) \leq \|A\|, highlighting its role as the tightest possible bound across all induced norms. Gelfand's formula provides a norm-independent characterization: \rho(A) = \lim_{k \to \infty} \|A^k\|^{1/k}, which holds for any and underscores the spectral radius's intrinsic nature. A pivotal theorem states that the sequence of powers A^k converges to the if and only if \rho(A) < 1, making the spectral radius a key criterion for asymptotic stability in linear dynamical systems and the convergence of fixed-point iterations. For Hermitian or normal matrices, the spectral radius equals the 2-norm, \rho(A) = \|A\|_2, while for unitary matrices it is exactly 1, and for nilpotent matrices it is 0. In the theory of nonnegative matrices, the Perron–Frobenius theorem asserts that \rho(A) is itself a real eigenvalue with a positive eigenvector, facilitating applications in , population dynamics, and optimization problems where matrix entries represent transitions or interactions. Beyond matrices, the concept extends to graph theory, where the spectral radius of a graph is the largest eigenvalue of its adjacency matrix; for a d-regular graph, this equals d, linking it to connectivity, expansion properties, and .

Definitions

Square matrices

The spectral radius of a square matrix A \in \mathbb{C}^{n \times n} is defined as \rho(A) = \max \{ |\lambda| : \lambda \text{ is an eigenvalue of } A \}. This measures the largest absolute value among the eigenvalues of A, capturing the scale of the matrix's spectral behavior in finite dimensions. Since A is finite-dimensional, the eigenvalues are finite in number and can be found as the roots of its characteristic polynomial. The eigenvalues of A are the complex numbers \lambda satisfying \det(A - \lambda I) = 0, where I is the n \times n identity matrix. The characteristic polynomial p_A(\lambda) = \det(A - \lambda I) is a monic polynomial of degree n, and its roots constitute the spectrum of A, denoted \sigma(A). Thus, the spectral radius can equivalently be expressed as \rho(A) = \sup \{ |\lambda| : \lambda \in \sigma(A) \}. In the complex case, the supremum is attained as a maximum due to the compactness of the spectrum. The concept of the spectral radius traces its origins to David Hilbert's foundational work on integral equations during 1904–1906, where spectral considerations arose in solving Fredholm-type problems, though the precise term and its formalization for finite square matrices developed later in the context of linear algebra. For example, consider the $2 \times 2 rotation matrix R(\theta) = \begin{pmatrix} \cos \theta & -\sin \theta \\ \sin \theta & \cos \theta \end{pmatrix}. The characteristic polynomial is p_{R(\theta)}(\lambda) = \lambda^2 - 2\cos\theta \, \lambda + 1 = 0, with roots \lambda = e^{i\theta} and \lambda = e^{-i\theta}./05%3A_Eigenvalues_and_Eigenvectors/5.04%3A_Complex_Eigenvalues) Both eigenvalues have modulus 1, so \rho(R(\theta)) = 1, reflecting the norm-preserving nature of rotations./05%3A_Eigenvalues_and_Eigenvectors/5.04%3A_Complex_Eigenvalues)

Bounded linear operators

In the context of infinite-dimensional spaces, the spectral radius extends naturally to bounded linear operators on Banach spaces. For a bounded linear operator T on a complex Banach space X, the spectral radius \rho(T) is defined as \rho(T) = \sup \{ |\lambda| : \lambda \in \sigma(T) \}, where \sigma(T) denotes the spectrum of T, the set of all complex numbers \lambda such that T - \lambda I is not invertible in the algebra of bounded operators on X. This definition captures the growth rate of the operator's powers in a manner analogous to the finite-dimensional case, but relies on the resolvent set rather than direct eigenvalue computation. Unlike the point spectrum, which consists solely of eigenvalues (points \lambda for which T - \lambda I has a non-trivial kernel), the full spectrum \sigma(T) may include additional components such as the approximate point spectrum (where T - \lambda I is injective but not bounded below) and the residual spectrum. Thus, \rho(T) is determined by the entire spectrum, not just the eigenvalues, which can lead to \rho(T) exceeding the maximum modulus of any actual eigenvalues in infinite dimensions. This broader scope is essential for operators without point spectrum, such as shifts or multiplication operators on function spaces. A key property is that \rho(T) \leq \|T\|, where \|T\| is the operator norm induced by the norm on X, reflecting the fact that the spectrum is contained in the closed disk of radius \|T\| centered at the origin. Equality holds for normal operators on Hilbert spaces, where the spectral theorem ensures the norm equals the supremum of the moduli of spectral values. Further details on normal operators appear in the section on symmetric matrices. As an illustrative example, consider the multiplication operator T on the Hilbert space L^2[0,1] defined by (Tf)(x) = f(x) \cdot g(x), where g is a continuous complex-valued function on [0,1]. The spectrum \sigma(T) coincides with the image g([0,1]), a compact subset of \mathbb{C}, so \rho(T) = \max_{x \in [0,1]} |g(x)|. This case highlights how the spectral radius directly reflects the essential range of the multiplier function.

Graphs

In graph theory, the spectral radius of a finite undirected graph G is defined as \rho(G) = \rho(A), where A is the adjacency matrix of G, a real symmetric $0-$1 matrix with zeros on the diagonal (for simple graphs without loops) and A_{ij} = 1 if vertices i and j are adjacent. Since A is symmetric and nonnegative, its eigenvalues are real, and \rho(G) coincides with the largest eigenvalue of A, which is simple and positive by the . For directed graphs, the adjacency matrix A is defined similarly but need not be symmetric, so its eigenvalues may be complex; the spectral radius \rho(G) remains the maximum modulus among these eigenvalues. The spectral radius \rho(G) has significant combinatorial interpretations, particularly in bounding the growth of walks in the graph. The total number of walks of length n in G equals the sum of the entries of A^n, which is \sum_k \lambda_k^n over the eigenvalues \lambda_k of A; for large n, this quantity grows asymptotically as c \cdot \rho(G)^n for some constant c > 0 depending on the Perron eigenvector. A representative example is the cycle graph C_n on n vertices, whose adjacency matrix has eigenvalues $2 \cos(2\pi k / n) for integers k = 0, 1, \dots, n-1; thus, \rho(C_n) = 2. The spectral radius also connects to structural properties like diameter and connectivity: for a connected graph G with diameter d, \rho(G) \geq 2 \cos(\pi / (d+1)), with equality holding for the path graph on d+1 vertices.

Fundamental properties

Gelfand's formula

Gelfand's formula characterizes the spectral radius \rho(T) of a bounded linear operator T on a Banach space as
\rho(T) = \lim_{n \to \infty} \|T^n\|^{1/n},
where \| \cdot \| denotes the operator norm induced by the norm on the Banach space. This limit exists and equals \inf_{n \geq 1} \|T^n\|^{1/n}.
For square matrices, the formula takes the alternative form \rho(A) = \lim_{n \to \infty} \|A^n\|^{1/n}, where \| \cdot \| is any matrix norm. The value of the limit is independent of the choice of , a consequence of the equivalence of norms in finite-dimensional spaces for matrices or the submultiplicativity of the in the setting. The formula is named after , who established it in 1941 as part of his foundational work on normed rings (), building on earlier contributions by to in operator algebras and by Arne Beurling in 1938 for the matrix case. As a simple example, consider a A satisfying A^2 = 0. Then \rho(A) = 0, and \|A^n\|^{1/n} = 0 for all n \geq 2, so the limit is 0.

Proof and corollaries

The proof of Gelfand's formula relies on the submultiplicativity of the operator norm in Banach algebras, which ensures that \|T^{n+m}\| \leq \|T^n\| \cdot \|T^m\| for any bounded linear operator T on a Banach space. This property implies that the sequence a_n = \|T^n\|^{1/n} satisfies a_{n+m} \leq (a_n a_m)^{n m / (n+m)} in a manner that bounds the lim sup and lim inf, leading to the existence of the limit \lim_{n \to \infty} \|T^n\|^{1/n}. Specifically, for any positive integer m, \|T^{m k}\|^{1/(m k)} \leq \|T^m\|^{1/m} for all k, so taking the infimum over m shows that the lim sup of a_n equals the infimum over m of a_m, establishing that the limit exists and equals this infimum. For the upper bound, consider any eigenvalue \lambda of T with |\lambda| = \rho(T). Then, for the corresponding eigenvector v \neq 0, \|T^n v\| = |\lambda|^n \|v\|, so |\lambda|^n \leq \|T^n\| \cdot \|v\|, implying \rho(T) \leq \liminf_{n \to \infty} \|T^n\|^{1/n}. More generally, without assuming eigenvalues exist, the spectral radius satisfies \rho(T^n) = \rho(T)^n \leq \|T^n\|, so \rho(T) \leq \liminf_{n \to \infty} \|T^n\|^{1/n}. For the lower bound, suppose \limsup_{n \to \infty} \|T^n\|^{1/n} < r. Then, for any \lambda with |\lambda| > r, the Neumann series \sum_{k=0}^\infty (T/\lambda)^k converges in the operator norm to (\lambda I - T)^{-1}, since the terms decay geometrically. Thus, \lambda \notin \sigma(T), implying \rho(T) \leq r, and taking the infimum over such r yields \rho(T) \geq \limsup_{n \to \infty} \|T^n\|^{1/n}. Combining bounds gives equality. A key is that the power series \sum_{n=0}^\infty T^n / r^n converges in the whenever r > \rho(T). This follows directly from the convergence when \|T / r\| < 1 in the effective radius, and it implies that the radius of convergence of the resolvent (r I - T)^{-1} = \frac{1}{r} \sum_{n=0}^\infty (T/r)^n is precisely $1/\rho(T), providing a holomorphic extension outside the spectrum. Another important corollary concerns power-bounded operators, where \sup_{n \geq 0} \|T^n\| < \infty. Gelfand's formula then yields \rho(T) = \lim_{n \to \infty} \|T^n\|^{1/n} \leq 1, since the norms are uniformly bounded. This has applications to stability in dynamical systems: if \rho(T) < 1, the iterates T^n x \to 0 for all x, ensuring asymptotic stability of the origin under the linear flow x_{n+1} = T x_n; conversely, if \rho(T) > 1, some directions grow unboundedly, indicating . As a numerical illustration, consider the 3×3 companion matrix A = \begin{pmatrix} 0 & 1 & 0 \\ 0 & 0 & 1 \\ 1 & 1 & 0 \end{pmatrix} associated to the characteristic polynomial x^3 - x - 1 = 0. The eigenvalues are one real root \lambda \approx 1.3247 (with |\lambda| = \rho(A)) and two complex conjugates with modulus less than 1. Approximating \rho(A) via \|A^n\|_2^{1/n} (using the spectral norm), the sequence converges to \approx 1.3247: for example, values for small n start above 1.5 and decrease toward the limit as n increases to 20 or more, demonstrating the practical utility of the formula for computation when direct eigenvalue solving is challenging.

Bounds and estimates

For matrices

A fundamental upper bound on the spectral radius \rho(A) of an n \times n matrix A is given by any induced matrix norm \|A\|, satisfying \rho(A) \leq \|A\|. For example, the infinity norm \|A\|_\infty = \max_i \sum_{j=1}^n |a_{ij}|, which is the maximum absolute row sum, provides \rho(A) \leq \max_i \sum_{j=1}^n |a_{ij}|. The offers a more refined localization of eigenvalues, implying an upper bound on \rho(A). It states that every eigenvalue of A lies in at least one of the disks in the centered at a_{ii} with R_i = \sum_{j \neq i} |a_{ij}| for i = 1, \dots, n. Consequently, \rho(A) \leq \max_i (|a_{ii}| + R_i) = \max_i \left( |a_{ii}| + \sum_{j \neq i} |a_{ij}| \right). For lower bounds, the trace of A provides a simple estimate: since \operatorname{tr}(A) = \sum_{i=1}^n \lambda_i where \lambda_i are the eigenvalues, it follows that \rho(A) \geq |\operatorname{tr}(A)| / n. More advanced trace-based lower bounds exist, such as those comparing \sum_{j=1}^n |\lambda_j|^2 to powers of the trace, yielding sharper estimates for \rho(A) in certain cases. For non-negative matrices, the Collatz-Wielandt formula characterizes \rho(A) via a min-max over positive vectors x > 0: \rho(A) = \max_{x > 0} \min_i \frac{(Ax)_i}{x_i} = \min_{x > 0} \max_i \frac{(Ax)_i}{x_i}, assuming A is irreducible. As an illustrative special case, consider a row-stochastic P (non-negative entries with row sums equal to 1). Here, \rho(P) = 1, with the all-ones vector as the corresponding right eigenvector. Equality holds in the infinity norm bound, since \|P\|_\infty = 1.

For graphs

The spectral radius of the A_G of a d- G is exactly d, as d is an eigenvalue corresponding to the all-ones eigenvector. A fundamental upper bound states that the spectral radius \rho(G) satisfies \rho(G) \leq \Delta(G), where \Delta(G) is the maximum of G, with equality G is regular or biregular bipartite. This inequality follows from the spectral radius being at most the infinity norm of A_G, which equals \Delta(G). For bipartite graphs, the spectrum of A_G is symmetric about zero, so \rho(G) = \lambda_{\max}(A_G) = -\lambda_{\min}(A_G). The spectral radius is related to the matching number \nu(G), as the bipartite graphs maximizing \rho(G) for a fixed \nu(G) = \beta are the complete bipartite graphs K_{\beta, n-\beta}. The Alon–Boppana bound provides that for any family of d-regular graphs with diameter tending to infinity, the second-largest eigenvalue \lambda_2(G) \geq 2\sqrt{d-1} + o(1), implying fundamental limits on the construction of expander graphs where the is constrained relative to the fixed spectral radius d. As an illustrative example, the K_n has adjacency matrix eigenvalues n-1 (multiplicity 1) and -1 (multiplicity n-1), yielding \rho(K_n) = n-1.

Special cases

Symmetric matrices

For real symmetric matrices, the spectral radius \rho(A) coincides with the spectral norm \|A\|_2, defined as the largest singular value of A, and equals the maximum absolute value of its eigenvalues, \max_i |\lambda_i|, since all eigenvalues are real. This equality holds because the spectral norm of a symmetric matrix is precisely the radius of its spectrum. The spectral radius admits a variational characterization through the Rayleigh quotient: \rho(A) = \max_{\|x\|_2 = 1} |x^T A x|, where the maximum is attained at an eigenvector corresponding to the eigenvalue of largest absolute value. This formulation leverages the self-adjoint nature of symmetric matrices to provide bounds and approximations for the dominant eigenvalue. As a consequence, matrix inequalities simplify for symmetric matrices. Specifically, \rho(A) = \|A\|_2 \leq \|A\|_F \leq \sqrt{n} \|A\|_2, where \|A\|_F is the Frobenius and n is the matrix ; the left inequality follows from the general spectral radius bound by induced norms, while the right arises from the expression \|A\|_F^2 = \sum_i \lambda_i^2 \leq n \max_i \lambda_i^2 = n \|A\|_2^2. A concrete example is the n \times n symmetric tridiagonal with 0s on the and 1s on the sub- and super-diagonals, which arises in discretizations of the one-dimensional Laplacian . Its eigenvalues are explicitly \lambda_k = 2 \cos\left(\frac{k \pi}{n+1}\right) for k = 1, \dots, n, so the spectral radius is \rho(A) = 2 \cos\left(\frac{\pi}{n+1}\right), approaching 2 as n increases.

Non-negative matrices

For an irreducible non-negative matrix A, the Perron-Frobenius theorem asserts that the spectral radius \rho(A) is a simple positive real eigenvalue, known as the Perron root, which admits a positive eigenvector, and all other eigenvalues \lambda satisfy |\lambda| \leq \rho(A). This dominant eigenvalue governs the long-term asymptotic behavior of powers of A, with the corresponding eigenvector providing a positive direction of growth. A non-negative irreducible matrix is primitive if some power A^k is positive, in which case the strict inequality |\lambda| < \rho(A) holds for all non-Perron eigenvalues, ensuring faster convergence to the dominant behavior. Simple bounds on the spectral radius follow from the row sums of A: if r_{\min} and r_{\max} denote the minimum and maximum row sums, respectively, then r_{\min} \leq \rho(A) \leq r_{\max}, with equality if and only if all row sums are equal. For primitive matrices, the inequalities are strict unless A is regular (i.e., a positive scalar multiple of a doubly stochastic matrix). The Collatz-Wielandt characterization provides a variational formula for the spectral radius of a non-negative matrix A: \rho(A) = \inf_{x > 0} \max_i \frac{(Ax)_i}{x_i} = \sup_{x > 0} \min_i \frac{(Ax)_i}{x_i}, where the infimum and supremum are taken over positive vectors x, and equality holds for the Perron eigenvector. This min-max principle highlights the connection between the spectral radius and Rayleigh-like quotients adapted to the non-negative cone. In applications, the spectral radius of a non-negative matrix models the growth rate in discrete-time systems; for example, in Markov chains with P (row-stochastic, hence non-negative with row sums equal to 1), \rho(P) = 1 is the dominant eigenvalue, corresponding to the , and the second-largest eigenvalue modulus determines the mixing time to . Similarly, in age- or stage-structured population models, the A (non-negative Leslie or similar) has \rho(A) as the asymptotic growth rate r, where r > 1 implies population expansion, r < 1 implies decline, and scaling fertilities adjusts r to a target value via \rho of a modified fertility-survival product. A classic example is the Fibonacci matrix A = \begin{pmatrix} 1 & 1 \\ 1 & 0 \end{pmatrix}, which is irreducible and non-negative; its \det(A - \lambda I) = \lambda^2 - \lambda - 1 = 0 yields eigenvalues \frac{1 \pm \sqrt{5}}{2}, so \rho(A) = \frac{1 + \sqrt{5}}{2} \approx 1.618, the golden ratio, with positive eigenvector \begin{pmatrix} \phi \\ 1 \end{pmatrix} where \phi = \rho(A). This illustrates the theorem, as the other eigenvalue has modulus less than \rho(A), and powers of A generate numbers scaled by \rho(A)^n.

References

  1. [1]
    What Is the Spectral Radius of a Matrix? - Nick Higham
    Jan 12, 2024 · The spectral radius \rho(A) of a square matrix A\in\mathbb{C}^{n\times n} is the largest absolute value of any eigenvalue of A.
  2. [2]
    [PDF] BACKGROUND ON GRAPHS
    May 4, 2022 · ä Definition : a graph is d regular if each vertex has the same degree d. Proposition: The spectral radius of a d regular graph is equal to d.
  3. [3]
    The Characteristic Polynomial
    The characteristic polynomial of A is the function f ( λ ) given by f ( λ )= det ( A − λ I n ) . We will see below that the characteristic polynomial is in ...
  4. [4]
    Matrix Spectrum -- from Wolfram MathWorld
    The eigenvalues of a matrix A are called its spectrum, and are denoted lambda(A). If lambda(A)={lambda_1,...,lambda_n}, then the determinant of A is given ...
  5. [5]
    [PDF] On the origin and early history of functional analysis - DiVA portal
    During the years 1904–1906, Hilbert published six papers on integral equations which ... For the parameter λ, Riesz proved the spectral radius theorem saying that ...
  6. [6]
    [PDF] FUNCTIONAL ANALYSIS | Second Edition Walter Rudin
    Rudin, Walter, (date). Functional analysis/Walter Rudin.-2nd ed. p. em. -(international series in pure and applied mathematics).
  7. [7]
    [PDF] Chapter 9: The Spectrum of Bounded Linear Operators
    Although the spectral radius of a self-adjoint operator is equal to its norm, the spectral radius does not provide a norm on the space of all bounded operators.
  8. [8]
    Spectral Radius of Graphs
    The basic one of these is the adjacency matrix, which encodes existence of edges joining vertices of a graph. Knowledge of spectral properties of the adjacency ...
  9. [9]
    [PDF] On the spectral radius of a directed graph
    Such matrices correspond to (possibly in nite) directed graphs with nite in- and out-degrees at every vertex. Theorem 1 trivially extends to this class of.
  10. [10]
    Walks and the spectral radius of graphs - ScienceDirect.com
    Given a graph G, write μ(G) for the largest eigenvalue of its adjacency matrix, ω(G) for its clique number, and wk(G) for the number of its k-walks.
  11. [11]
    [PDF] Algebraic Graph Theory - CMU School of Computer Science
    Algebraic graph theory / Chris Godsil, Gordon Royle. p. cm. - (Graduate ... Now, we will find the eigenvalues of the cycle Cn. Take the vertex set of.
  12. [12]
    A lower bound for the spectral radius of graphs with fixed diameter
    We determine a lower bound for the spectral radius of a graph in terms of the number of vertices and the diameter of the graph.
  13. [13]
    None
    ### Gelfand's Formula for Bounded Operators on Banach Space
  14. [14]
    [PDF] an elementary proof of the spectral radius formula for matrices
    Theorem 2.1. All norms on a finite-dimensional vector space are equivalent ... The spectral radius formula holds for any matrix and any norm: ‖An‖1/n ...
  15. [15]
    [PDF] On Beruling-Gelfand's spectral radius theorem - Auburn University
    Jan 24, 2007 · Theorem 3.1. (Beruling, 1938, Gelfand, 1941) Let A ∈ Cn×n. where r(X) is the spectral radius of X and kXk is the spectral norm of A. We now ...
  16. [16]
    [PDF] Math 261y: von Neumann Algebras (Lecture 2)
    Sep 2, 2011 · Our first nontrivial result is the following: Theorem 11 (Gelfand's Spectral Radius Formula). Let A be a Banach algebra and let x ∈ A. Then.
  17. [17]
    [PDF] POWER-BOUNDED ELEMENTS AND RADICAL BANACH ...
    The well-known spectral radius formula implies that, if x is power-bounded, then r(x) ≤ 1; also, if x is doubly power-bounded, then Sp x ⊆ Γ. A further simple ...<|control11|><|separator|>
  18. [18]
    [PDF] Matrix norms and singular value decomposition
    the induced norm measures the amount of \ampli cation" the matrix A provides ... spectral radius of A, defined as: (A) maxi j ij and the i's are the ...
  19. [19]
    [1605.07239] Optimizing Gershgorin for Symmetric Matrices - arXiv
    May 23, 2016 · The Gershgorin Circle Theorem is a well-known and efficient method for bounding the eigenvalues of a matrix in terms of its entries.
  20. [20]
    Some lower bounds for the spectral radius of matrices using traces
    Feb 1, 2010 · If A is any complex matrix, two lower bounds for ∑ j = 1 n | λ j | 2 are compared, and furthermore a new lower bound for the spectral radius is ...
  21. [21]
    [PDF] New proofs for a bound on the spectral radius of the Hadamard ...
    For convenience, we recall the Collatz-Wielandt formula, which gives a (double- sided) bound on the spectral radius of an irreducible non-negative matrix A, ...
  22. [22]
    What Is a Stochastic Matrix? - Nick Higham
    Dec 13, 2022 · A stochastic matrix is an n\times n matrix with nonnegative entries and unit row sums. If A\in\mathbb{R}^{n\times n} is stochastic then Ae = e.
  23. [23]
    Eigenvalues of Graphs - American Mathematical Society
    The adjacency matrix of the complete graph Kn of order n is the matrix A = Jn − In where Jn is the n × n matrix of all 1s. This graph has eigenvalues −1,...,−1, ...
  24. [24]
    Spectral radius of graphs with given matching number - ResearchGate
    Aug 6, 2025 · In this paper, we show that of all graphs of order n with matching number β, the graphs with maximal spectral radius are Kn if n=2β or 2β+1; ...
  25. [25]
    Hypergraphs and hypermatrices with symmetric spectrum
    Apr 15, 2017 · A graph is bipartite if and only if its adjacency matrix has a symmetric spectrum. Recall that the spectrum of a complex square matrix A is ...
  26. [26]
    [PDF] A generalized Alon-Boppana bound and weak Ramanujan graphs
    A basic eigenvalue bound due to Alon and Boppana holds only for regular graphs. In this paper we give a generalized Alon-Boppana.
  27. [27]
    [PDF] arXiv:1710.07809v1 [math.HO] 21 Oct 2017
    Oct 21, 2017 · The maximum value of the Rayleigh quotient RA is the largest eigenvalue λ of A, and the maximum is achieved only at eigenvectors for A and λ. i ...
  28. [28]
    An Extension of the Rayleigh Quotient to the Spectral Radius ... - arXiv
    Dec 30, 2019 · The Rayleigh quotient, which provides the classical variational characterization of the spectral radius of Hermitian matrices, can be extended to nonsymmetric ...
  29. [29]
    [PDF] Chapter 4 Vector Norms and Matrix Norms - UPenn CIS
    Observe that property (3) of proposition 4.5 says that. A 2 ≤ A F ≤. √ n A 2. , which shows that the Frobenius norm is an upper bound on the spectral norm.
  30. [30]
    [PDF] Tridiagonal Toeplitz Matrices: Properties and Novel Applications
    The eigenvalues are allocated symmetrically with respect to δ. The spectral radius of the matrix (2) is given by ρ(T) = max ½¯¯¯. ¯δ. + 2p|στ|ei(α+β)/2 cos π.
  31. [31]
    [PDF] A new proof of the Perron–Frobenius theorem: a variational approach
    Jul 19, 2024 · Furthermore,. (1.20) means that the Perron root of irreducible nonnegative matrices is stable under arbitrary perturbations, which may be ...<|separator|>
  32. [32]
    [PDF] Generalized Sharp Bounds on the Spectral Radius of Digraphs - arXiv
    Mar 4, 2013 · The following classical result gives a bound on the spectral radius of the nonnegative matrix A in terms of its row sums. Theorem 1.1 (Frobenius) ...
  33. [33]
    A Collatz-Wielandt characterization of the spectral radius of order ...
    Dec 27, 2011 · We give a Collatz-Wielandt type formula, which characterizes the growth rate of the orbits in terms of eigenvectors in the closed cone or super-eigenvectors in ...
  34. [34]
    [PDF] Lecture 17 Perron-Frobenius Theory
    Perron-Frobenius Theory. • Positive and nonnegative matrices and vectors. • Perron-Frobenius theorems. • Markov chains. • Economic growth. • Population dynamics.
  35. [35]
    [PDF] Applications of Perron-Frobenius Theory to Population Dynamics
    More generally, we show how to achieve a given growth rate for the model by scaling the fertility matrix. Demographic interpretations of the results are given.
  36. [36]
    Matrices, eigenvalues, Fibonacci, and the golden ratio - The DO Loop
    Oct 5, 2010 · It was noted that the ratio between adjacent terms in the Fibonacci sequence approaches the "Golden Ratio," 1.61803399.... This post explains ...Missing: spectral | Show results with:spectral