3

This is problem 35 of section 5.A of Axler's book Linear Algebra Done Right (3rd Edition). The problem goes as follows:

Suppose $V$ is finite-dimensional, $T$ is a operator on $V$, and $U$ is invariant under $T$. Prove that each eigenvalue of $T/U$ is an eigenvalue of $T$.

The author then adds the remark:

"The next exercise [36] asks you to verify that the hypothesis that $V$ is finite-dimensional is needed for the exercise above [35]."

Clarifying the notation for those not familiar with, $V/U$ is the quotient space defined as $$ V/U = \{v + U : v \in V \}. $$

where $$ v + U = \{ v + u : u \in U \}. $$

and the quotient operator $T/U : V/U \to V/U$ is the linear map defined as:

$$ (T/U)(v + U) = Tv + U. $$

My attempt at proving the result goes as follows:

Suppose that $\lambda$ is an eigenvalue of $T/U$, it can easily be shown that this implies that there must exist a $ v \notin U $ such that $Tv - \lambda v \in U$. Let $u_1, \ldots, u_n $ be a basis of the subspace $U$. Because $U$ is a subspace and it is invariant under $T$, it follows that $Tu_i - \lambda u_i \in U$ for $i = 1, \ldots, n$. Thus, we have a list of $n+1$ vectors in a space of dimension $n$ (because $\dim U = n$), therefore, this list has to be linearly dependent. This implies that there exists a non-trivial (not all scalars are equal to $0$) linear combination of $Tv - \lambda v, Tu_1 - \lambda u_1, \ldots, Tu_n - \lambda u_n $ that equals $0$. This implies that $$ (T-\lambda I)(\text{linear combination of } v, u_1, \ldots, u_n) = 0. $$ The linear combination of $v, u_1, \ldots, u_n $ cannot be zero because $v$ is not in $U$ and the $u_i$'s are a basis of $U$. This shows that $\lambda$ is also a eigenvalue of T. As we wanted.

My problem is that the author says that the hypothesis that $V$ is finite-dimensional is needed for the proof. But here we have only used the hypothesis that $U$ is finite-dimensional. So i suppose this proof would still be valid in a infinite-dimensional vector space $V$ as long as the invariant subspace $U$ is finite-dimensional. Is that correct?

Thanks in advance. Please feel free to ask for clarification in any passage in the proof.

  • The proof looks good to me. I agree that we only require $U$ to be finite-dimensional. –  Aug 23 '23 at 19:23
  • 1
    Yes, this looks OK. It might be clearer about "the linear combination" is (e.g. written more explicitly, as in the rest of the proof). Note that this question was previously asked e.g. at https://math.stackexchange.com/questions/2179872/prove-each-eigenvalue-of-t-u-is-an-eigenvalue-of-t although the only answer has a number of comments in which users seek clarification. Note also that when $V$ is not finite dimensional, counterexamples (with $U$ infinite dimensional) are given at e.g. https://math.stackexchange.com/questions/1147732/eigenvalues-of-an-operator-induced-in-a-quotient-space – leslie townes Aug 24 '23 at 00:54

0 Answers0