I saw the following question in my linear algbra book, and found it rather strange:

"Let $V$ be a real vector space of dimension $n$. Let $L,K \colon V \rightarrow \mathbb{R}$ be linear transformations, so that $\ker(L) \subset \ker(K)$. Prove that $K=\lambda L$ for a $\lambda \in \mathbb{R}$, using the following steps:

  1. Prove this for $K=0$.
  2. Assume $K \neq 0$. Prove that $\dim(\ker(L)) = \dim(\ker(K))$.
  3. Now prove that $K = \lambda L$ for $K \neq 0$."

What is meant by "$K=0$"? How is a linear transformation ever equal to a number? I have a feeling that they might mean something else than just "$K$", but I'm not sure.

I tried something for "K maps every vector to $0$", but that doesn't really seem to be what is meant, because then I obviously get that $K(v) = \lambda L(v)$ when $\lambda = 0$, which would mean that this question allows me to just choose whatever $\lambda$ to make the statement true and I'm not sure that is the case here. Also, how would I continue for $K \neq 0$?

Can anyone give me a hint?



How is a linear transformation ever equal to a number?

Here, you're supposed to understand (from context) that $0$ doesn't mean the number, it means the function $0: V\to \mathbb{R}$ which sends every element to (the number) $0$. It's confusing notation.

then I obviously get that $K(v)=\lambda L(v)$ when $\lambda = 0$

That's correct.

Also, how would I continue for $K\neq 0$?

Have you tried using the hint in part 2 of the question? You're going to need to use a special fact about the target vector space $\mathbb{R}$. Think 'rank-nullity theorem'.

  • $\begingroup$ So I am supposed to prove that it is always possible to find a $\lambda$ so that $K = \lambda L$? Or should I find a specific $\lambda$ that will make this statement true? I already did some rather "dodgy" maths with said theorem to find that $\ker(L) = \ker(K)$, and thus as $K(v)$ and $L(v)$ are both elements of $\mathbb{R}$, let's say not equal to $0$, then I can find a $\lambda$ for every $v$. But that seems rather trivial, no? Shouldn't I do something else? $\endgroup$ – Linde Jan 21 '18 at 23:58
  • $\begingroup$ Given some fixed $K$ and $L$, there should be an element $\lambda$ such that, for every $v\in V$, we have $K(v) = \lambda L(v)$. That is, $\lambda$ is allowed to depend on $K$ and $L$, but not on $v$. $\endgroup$ – Billy Jan 22 '18 at 0:05
  • $\begingroup$ But the rest of what you've said looks fine to me. You should indeed find that $\ker(L) = \ker(K)$ (and you should be able to write down exactly what the dimension of this kernel is too). After that, well, perhaps pick a nice basis for $V$? $\endgroup$ – Billy Jan 22 '18 at 0:07
  • $\begingroup$ Wow, thanks, I hadn't thought of choosing a basis! $\endgroup$ – Linde Jan 22 '18 at 0:37
  • $\begingroup$ If I'm not mistaken, it just gives me that $\dim(\ker(K))=n-1$, which means I can choose a basis for V that consists of $n-1$ vectors that are elements of $\ker(L)$, and then one extra vector, say $v$. Since every $\lambda$ would do for the first $n-1$ vectors, I can say “let $\lambda$ be $K(v)/L(v)$, in which $L(v)\neq 0$ since $ v \notin ker(L)$“, and that should work, since a linear transf. is determined entirely by the images of the basis of V? If what I just said is correct, then thank you a huge lot for helping me out! $\endgroup$ – Linde Jan 22 '18 at 0:47

Yes, that is what is meant. Asserting that $K=0$ means that $(\forall v\in V):K(v)=0$.


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.