Say you're given a set of eigenvalues and eigenvectors, is it always possible to determine the original matrix A?

I recognize that if a matrix A is diagonalizable, the geometric multiplicity equals the algebraic multiplicity and you can form you matrix of Eigenvectors, call it Q, and then your diagonal matrix D (which consists of the eigenvalues that correspond to the correct eigenvector column in Q) gives you the result

$A = QDQ^-1$

but assuming we aren't given the original matrix A, is it possible to "reverse" this process and find A? Thanks


Assuming $A$ is $n\times n$, if you are given $n$ linear independent eigenvectors and the corresponding eigenvalues, then the $QDQ^{-1}$ formulation works.

If the matrix does not have this spanning set of eigenvectors, then you need the generalized eigenvectors, c.f. https://en.wikipedia.org/wiki/Generalized_eigenvector, to construct a matrix $P$ and the almost diagonal Jordan normal form $J$ such that $A=PJP^{-1}$.

Jordan normal forms, c.f. https://en.wikipedia.org/wiki/Jordan_normal_form, exist for all $n\times n$ matrices over the field $\mathbb{C}$, so the answer is "yes" if you are also given generalized eigenvectors and "no" if you are only given the eigenvectors of a non-diagonalizable matrix.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.