Perron–Frobenius theorem
A nonnegative square matrix has its Collatz–Wielandt value as an eigenvalue with a nonnegative eigenvector, and for irreducible matrices this eigenvector is positive and the eigenvalue is the spectral radius.
Results
Let $A$ be a nonnegative real $n\times n$ matrix, $n\ge1$. The Perron root is $$ r(A) = \sup_{v \geq 0,\ v \neq 0}\ \min_{i\,:\,v_i > 0} \frac{(Av)_i}{v_i}. $$
Theorem (Perron–Frobenius). [Per07]
- $r(A)$ is an eigenvalue of $A$ with a nonnegative eigenvector.
- If $A$ is irreducible, then $A$ has an eigenvector with positive entries and a positive eigenvalue. Moreover $r(A)\in\operatorname{spec}(A)$ and $|\mu|\le r(A)$ for every real eigenvalue $\mu$.
- If $A$ is primitive (some power is entrywise positive), then there is exactly one nonnegative eigenvector of sum $1$ with positive eigenvalue.
- If $Av=rv$ with $v$ entrywise positive and $r>0$, then $r=r(A)$.
Proof
Existence comes from upper semicontinuity of the Collatz–Wielandt function on the standard simplex. The irreducible case reduces to the primitive case through $1+A$.
References
[Per07]; G. Frobenius, “Über Matrizen aus nicht negativen Elementen,” Sitzungsberichte der Königlich Preussischen Akademie der Wissenschaften (1912), 456–477.
- [Per07] Zur Theorie der Matrices. Mathematische Annalen, 64(2), 248-263, 1907.
BibTeX
@article{Perron1907, author = {Perron, Oskar}, title = {Zur {T}heorie der {M}atrices}, year = {1907}, journal = {Mathematische Annalen}, volume = {64}, number = {2}, pages = {248--263}, doi = {10.1007/BF01449896}, url = {https://doi.org/10.1007/BF01449896} }