0

I was reading about duals, double duals and "natural identifications", and read that T** is identified with T. However, the statement did not come with a proof, and I cannot find one online.

$T : V → W$ is a map between finite-dimensional vector spaces. T** : V** → W** is the associated map between double duals.

The definition of natural identification is if $E^{(V)}: V \cong V^{**}$ and $E(W): W \cong W^{**}$ are the natural isomorphisms, then we have $T^{**} ◦ E^{(V)} = E^{(W)} ◦ T.$

I am very new to dual spaces, so would appreciate a full proof and explanation over hints so that I can slowly study in my own time, thanks.

yw_2003
  • 319
  • What exactly do you need proof of? "Identified with" is not really a result in and of itself, just a way of looking at things. You could prove that there exists a correspondence between $T \in \mathcal{L}(V, W)$ and a $T^{}$. To prove the uniqueness of this correspondence you can use the result in your third last line. Do you get how you would define $T^{}$ given $T$? – Vercingetorix Nov 14 '22 at 23:58
  • 2
    Before getting to maps, first are you comfortable with the identification of the spaces $V$ and $V^{**}$? – blargoner Nov 15 '22 at 01:02
  • @struggling You might find this post to be helpful – Ben Grossmann Nov 15 '22 at 01:50

1 Answers1

2

I recommend looking at my earlier post; the map I call $\alpha$ in that context is what you refer to as $E^{(V)}$. As I explain in this post, the "natural" identification between $V$ and $V^{**}$ is the map $E^{(V)}:V \to V^{**}$ given by $E^{(V)}(v) = E_v$ for $v \in V$, where $E_v \in V^{**}$ is the map from $V^*$ to the underlying field $F$ given by $E_v(f) = f(v)$, for any $f \in V^*$.

Informally, for any $v \in V$, $E_v$ describes the "natural" way in which a vector $v$ can be used to make a linear map $f:V \to F$ into an element of $F$.

Now, we also need to understand what $T^{**}$ means. First, we define $T^*$: any map $T:V \to W$ induces a dual map $T^*:W^* \to V^*$. For any $f \in W^*$, this map is defined such that $T^*(f) = f \circ T$. We can take the dual map of this dual map to get the map $T^{**}:V^{**} \to W^{**}$, which we call the "double dual" of the original map $T$. For any $\phi \in V^{**}$, this map is defined so that $T^{**}(\phi) = \phi \circ T^{*}$

Your question, then, is why it holds that $T^{**} \circ E^{(V)} = E^{(W)} \circ T$. Note that both sides of this equation describe a linear transformation, which is to say that they describe a function (which the special property of "linearity"). The two functions $T_1 = T^{**} \circ E^{(V)}$ and $T_2 = E^{(W)} \circ T$ from $V$ to $W^{**}$ are the same if for every input $v \in V$ (the common domain of these functions), we have $T_1(v) = T_2(v)$ (i.e. the two functions produce the same output). That is, we need to show that for every $v \in V$, we have $$ (T^{**} \circ E^{(V)})(v) = (E^{(W)} \circ T)(v). $$ Now, we're actually in the same situation as before: both sides of the equation describe an element of $W^{**}$, which is to say that both sides describe a linear transformation from $W^*$ to $F$. Again, the two functions are the same if they take every input $f \in W^*$ to the same output. That is, we need to show that for every $f \in W^{**}$, $$ [(T^{**} \circ E^{(V)})(v)](f) = [(E^{(W)} \circ T)(v)](f). $$ Finally, we have an equation between two numbers whose truth we need to confirm. To simplify the left-hand side, note that \begin{align} [(T^{**} \circ E^{(V)})(v)](f) &= [(T^{**}(E^{(V)}(v))](f) = [T^{**}(E_v)](f) \\ &= [E_v \circ T^*](f) = E_v(T^*(f)) = E_v(f \circ T) \\ &= (f \circ T)(v) = f(T(v)). \end{align} For the right-hand side, \begin{align} [(E^{(W)} \circ T)(v)](f) &= [E^{(W)}(T(v))](f) = E_{T(v)}(f) = f(T(v)). \end{align} So, the two sides are indeed the same, which means that $[(T^{**} \circ E^{(V)})(v)]$ and $[(E^{(W)} \circ T)(v)]$ are the same element of $W^{**}$, which means that $T^{**} \circ E^{(V)} = E^{(V)} \circ T$, which was what we wanted.

azif00
  • 20,792
Ben Grossmann
  • 225,327