0
$\begingroup$

I am looking for an example of a linear space $V$ over $\mathbb{R}$ and an endomorphism $\phi : V \to V$ such that each $a \in \mathbb{R}$ is an eigenvalue of $\phi$. One relatively simple example would be an endomorphism given by the matrix

$$\begin{bmatrix} a_1 & 0 & 0 & \cdots \\ 0 &a_2& 0 & \cdots \\ 0 & 0 & a_3&\cdots \\ \vdots & \vdots & \vdots & \ddots \end{bmatrix}$$ where $a_1, a_2, \dots$ are all eigenvalues associated with unit vectors $e_1, e_2, \dots$. We could possibly have infinitely many such eigenvalues. My problem here is that this still seems to be slightly complicated as the material I am going over has not introduced infinite dimensional vector spaces. Is there a more straightforward finite dimension example I am somehow missing?

$\endgroup$
  • 3
    $\begingroup$ In the finite dimensional case, you will have only finitely many eigenvalues $\endgroup$ – D_S Mar 6 '17 at 18:45
  • 1
    $\begingroup$ Eigenvectors associated to distinct eigenvalues are linearly independent, so this sort of construction is the best you can do; the dimension of the underlying vector space is an upper bound on the number of distinct eigenvalues, even in the infinite-dimensional case. $\endgroup$ – Qiaochu Yuan Mar 6 '17 at 18:55
  • $\begingroup$ @D_S One question. It bothers me that in my example, the set of eigenvectors $E$ (and the corresponding set of eigenvalues) satisfies $\lvert E \rvert = \lvert \mathbb{N} \rvert < \lvert \mathbb{R} \rvert$. Doesn't this mean that we can't claim that every $a \in \mathbb{R}$ is an eigenvalue? $\endgroup$ – Zelazny Mar 7 '17 at 7:34
  • 1
    $\begingroup$ Right, you only have countably many eigenvalues. In general, the cardinality of eigenvalues cannot exceed the dimension of your vector space. $\endgroup$ – D_S Mar 7 '17 at 9:39
  • 1
    $\begingroup$ Your example can be modified. Let $$V = \bigoplus\limits_{x \in \mathbb{R}} \mathbb{R}$$ and for $x \in \mathbb{R}$, let $e_x$ be the element with $1$ in the $x$ place, $0$ elsewhere. This is a basis for $V$. You can define a linear operator $T$ on $V$ by sending $e_x$ to $xe_x$. $\endgroup$ – D_S Mar 7 '17 at 9:41
2
$\begingroup$

As others have said, we need to consider something over an infinite dimensional space, however you slice it. As you've said, however, infinite matrices are a bit bizarre.

Here's a more "familiar" example: let $V$ denote the vector space of all smooth (infinitely differentiable) functions. Consider $\phi:V \to V$ to be the "derivative map", defined by $\phi(f) = f'$. We note that for any $\lambda \in \Bbb R$, the equation $$ \phi(f) = \lambda f $$ has the solution $f(x) = e^{\lambda x}$. So, every real number is an eigenvalue of $\phi$.

$\endgroup$
  • $\begingroup$ There are a lot of very strange things about my choice of vector space and linear map from a functional analysis perspective. However, it certainly does the job. $\endgroup$ – Omnomnomnom Mar 6 '17 at 20:00
  • $\begingroup$ This is a great example and it doesn't seem to suffer from the problem I mentioned above. Thanks! $\endgroup$ – Zelazny Mar 7 '17 at 8:14

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.