2
$\begingroup$

I would like to know if there are any rare cases where performing a basis change for describing a transformation on a different basis also changes its eigenvalues.
I know that in general this does not happen, but I would like to know if there are any extreme cases where it does happen.

  • 4
    An eigenvalue of a linear transformation $T:V\to W$ is defined as $\lambda$ such that there is nonzero $v\in V$ which satisfies $Tv=\lambda v$; This description is basis-free, so changing basis cannot change eigenvalues2017-02-07
  • 1
    @user160738: Presumably you meant $T:V\to V$? Unless the range is a subspace of the domain, there cannot be an eigenvector.2017-02-07
  • 1
    @hardmath Oh, yes; thanks for pointing that out2017-02-07

1 Answers 1

2

No. If $T: V \rightarrow V$ is a linear operator where $A$ is the matrix of $T$ with respect to a basis $\mathcal{B}$, and $C$ is the matrix of $T$ with respect to another basis $\mathcal{B}^{'}$, then we necessarily have that $A=PCP^{-1}$ for some invertible matrix P. In other words, $A$ is conjugate (or similar) to $C$. Matrices that are similar always have the exact same eigenvalues.

  • 0
    Do you know if something like this could happen in the field of functional analysis(i.e. in infinite dimensional spaces)?2017-02-07