Yesterday I learned about dual spaces when reading about spaces of linear maps. The concept of a linear map and why linear maps form a vector space is clear to me. But there are some details about the dual space and its basis that I could not fully understand.

The text I am reading states the following:

Furthermore, for fixed vector spaces $U$ and $V$ over $K$, the operations of addition and scalar multiplication on the set $\operatorname{Hom}_K(U,V)$ of all linear maps from $U$ to $V$ makes $\operatorname{Hom}_K(U,V)$ into a vector space over $K$.

Given a vector space $U$ over a field $K$, the vector space $U^{*} = \operatorname{Hom}_K(U,K)$ plays a special role. It is often called the dual space or the space of covectors of $U$. One can think of coordinates as elements of $U^{*}$. Indeed, suppose that $U$ is finite-dimensional and let $e_{1},...,e_{n}$ be a basis of $U$. Every $x \in U$ can be uniquely written as $$x=\alpha_{1}e_{1}+...+\alpha_{n}e_{n}, \alpha_{i} \in K.$$ The scalars $\alpha_{1},...\alpha_{n}$ depend on $x$ as well as on the choice of basis, so for each $i$ one can write the coordinate function $$e^{i}: U \to K, e^{i}(x)=\alpha_{i}.$$ It is routine to check that each $e^{i}$ is a linear map, and indeed the functions $e^{1},...,e^{n}$ form a basis of the dual space $U^{*}$.

Now I have two questions:

1) The text states that $\operatorname{Hom}_K(U,V)$ is a vector space for fixed $U$ and $V$. This is perfectly clear to me, but is it correct that the dual space is $U^{*}=\operatorname{Hom}_K(U,K)$, i.e. it consists of all linear maps from $U$ to the field $K$? At first I thought this was a typo, but from what I've read on wikipedia and from other sources the notation seems to be correct. It also seems to make no sense to say the coordinate functions form a basis for $\operatorname{Hom}_K(U,V)$.

2) It is easy to see that the coordinate functions $e^{1},...,e^{n}$ are linear maps and I have also tried to check that the claim that they form a basis for $U^{*}$. However, I am unsure if my proof is correct and I think this is mainly because of my confusion stated in the first question.

My proof goes as follows:

We need to prove that $e^{1},...,e^{n}$ are linearly independent and span $U^{*}$.

First note that linear maps are uniquely determined by their action on a basis. Now let $0_{UK}:U \to K, 0_{UK}(u)=0$ $\forall u$ be the zero map. To prove linear independence we need to show that

$(*)$ $b_{1}e^{1}+...+b_{n}e^{n}=0_{UK} \implies b_{i}=0$ $\forall i$

or in other words the only linear combination of $e^{1},...,e^{n}$ that gives $0_{UK}$ is the trivial linear combination. Now assume $b_{i}=0$ for some $i$, then(*) clearly fails if $x$ has a non-zero ith coordinate and so $b_{1}e^{1}+...+b_{n}e^{n}$ is not the zero map.

To prove that $e^{1},...,e^{n}$ span $U^{*}$ we need to prove that any vector in $U^{*}$ (every linear map) can be written as a linear combination of $e^{1},...,e^{n}$, i.e. $$T(u)=k_{1}e^{1}+...k_{n}e^{n}$$ for any vector $u \in U$. To see this note that $$\begin{align*} T(u)&=T(\alpha_{1}e_{1}+...\alpha_{n}e_{n}) \\ &=\alpha_{1}T(e_{1})+...\alpha_{n}T(e_{n}) \\ &=e^{1}T(e_{1})+...e^{n}T(e_{n}) \end{align*}$$ where $T(e_{i})$ are scalars by definition of $T$.

Is my proof correct or have I missed anything?

Thanks very much for any hints and comments.

  • $\begingroup$ What do you denote (1) which has to hold for any $x$? $\endgroup$ – Bernard Jan 2 '20 at 9:42
  • $\begingroup$ Sorry, this is a typo. I changed (1) to $(*)$ later. Actually I wanted to remove this part since it became clear to me when I wrote it down. Thanks. $\endgroup$ – DerivativesGuy Jan 2 '20 at 9:56
  • The linear independence can be shown more directly, suppose that $$b_1e^1 + b_2e^2 + \cdots + b_ne^n = \textit0 \quad \textrm{for some } b_1,\dots,b_n \in K$$ where $\textit0 : U \to K$ is the zero map. Now, for all $i$, $1\leq i\leq n$, $$\begin{align} b_i = b_ie^i(e_i) &= \sum_{j=1}^n b_j e^j(e_i) \\ &= \Big( \sum_{j=1}^n b_j e^j \Big)(e_i) = \textit0(e_i) = 0 \end{align}$$ so, $b_1 = b_2 = \cdots = b_n = 0$.

    • Also, to check that $e^1,\dots,e^n$ spans $U^*$, let $f$ be arbitrary in $U^*$, and observe that for any $x\in U$ written as $x = \alpha_1e_1 + \cdots + \alpha_n e_n$, we have $$\begin{align} f(x) &= \sum_{j=1}^n \alpha_j f(e_j) \\ &= \sum_{j=1}^n f(e_j) \alpha_j \\ &= \sum_{j=1}^n f(e_j) e^j(x) = \Big( \sum_{j=1}^n f(e_j)e^j \Big) (x) \end{align}$$ so, $$f = f(e_1)e^1 + f(e_2)e^2 + \cdots + f(e_n)e^n$$

Your proof seems a bit confusing. You would make it clearer with the initial observation that any linear map from a vector space $U$ to a vector space $V$ is entirely determined by its values at the vectors $e_i$ of a basis of $U$. The rest should follow almost immediately.

  • $\begingroup$ Thanks for your comment. You are right, I have implicitly taken this for granted since it was stated earlier in the text I am reading. I will add it to the proof. $\endgroup$ – DerivativesGuy Jan 2 '20 at 10:35
  • $\begingroup$ And then all verifications have to be made only for the vectors of the basis. $\endgroup$ – Bernard Jan 2 '20 at 10:37
  • $\begingroup$ Yes, this is clear to me. I think the main confusion for me was what I have stated in question 1). If my understanding is correct, then my proof should be fine. $\endgroup$ – DerivativesGuy Jan 2 '20 at 10:42

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.