2
$\begingroup$

Background

Let $V,W$ be finite dimensional vector spaces, $V^*$ the dual space of $V$ and $\mathrm{Hom} \ (V,W)$ the vector space of all linear transformations from $V$ to $W$. The universal mapping property states the following:

Let $\phi$ be the bilinear map $(v,w) \rightarrow v\otimes w$ of $V \times W$ into $V \otimes W$. Then whenever $U$ is a vector space and $l: V \times W \rightarrow U$ is a bilinear map, there exists a unique linear map $\tilde{l}: V \otimes W \rightarrow U$ so that $\tilde{l}\circ\phi=l$

Attempt

Let $F: V^* \times W \rightarrow \mathrm{Hom} \ (V,W)$ be the bilinear map defined by $(f,w)(v)=f(v)\cdot w$. Then by the above property there is a unique linear map $\alpha: V^* \otimes W \rightarrow \mathrm{Hom} \ (V,W)$ such that $\alpha\circ \phi=F$.

Question

How can I show that $\alpha$ is an isomorphism?

$\endgroup$
  • 1
    $\begingroup$ If you want to show those spaces are isomorphic it suffices verifying the universal property for $(Hom(V, W), F)$ since tensor product is unique up to isomorphism. $\endgroup$ – PtF Apr 2 '16 at 17:32
1
$\begingroup$

It is reather easy to see that $F$ is surjective. Choose a basis $w_1,\dots,w_m$ for $w$ and consider a linear map $\Phi:V\to W$. Expanding $\Phi(v)=\sum_i\Phi_i(v)w_i$ defines $\Phi_1,\dots,\Phi_m\in V^*$ and one immediately verifies that $\Phi=F(\sum_i\Phi_i\otimes w_i)$.

For injectivity, I think a bit of additional work has to be done. The main point is to show that the elements $f\otimes w$ span $V^*\otimes W$. To do this using the universal property, you can take the subspace $A\subset V^*\otimes W$ spanned by all such elements and the qoutient projection $\pi:V^*\otimes W\to (V^*\otimes W)/A$. Then $\pi$ vanishes on all elements of the form $f\otimes w$ by construction, so by uniqueness in the universal property, it must be the zero map. But since $\pi$ is surjective, this implies that $(V^*\otimes W)/A=\{0\}$ and hence $A=V^*\otimes W$. Having that at hand, it follows readily that for the Basis above, also the elements of the form $f\otimes w_i$ for $f\in V^*$ and $i=1,\dots,m$ span $V^*\otimes W$. This implies that any element of the tensor product can be written as $\sum_i f_i\otimes w_i$ for some $f_i\in V^*$ and the basis elements $w_i$. But then $F(\sum_i f_i\otimes w_i)=0$ easily implies $f_i=0$ for all $i$ and thus injectivity of $F$.

Alternatively, you can use the same line of argument to show directly that choosing bases of $V$ and $W$, the tensor products of the basis elements form a basis for the tensor product.

$\endgroup$
1
$\begingroup$

Consider bases of $V$ and $W$, $\{v_1,\dots,v_m\}$ and $\{w_1,\dots,w_n\}$. For each $i$ and $j$, consider the linear map $$ f_{ij}\colon V\to W, \qquad f_{ij}(v_k)=\begin{cases} w_j & \text{if $k=i$}\\[4px] 0 & \text{if $k\ne i$} \end{cases} $$ It's easy to prove that $\{f_{ij}:1\le i\le m,1\le j\le n\}$ is a basis for $\operatorname{Hom}(V,W)$: this is basically the method for defining the associated matrix to a linear map between finite dimensional vector spaces, given bases.

Let $\{v_1^*,\dots,v_m^*\}$ be the dual basis on $V^*$ and note that $$ \alpha(v_i^*\otimes w_j)=\phi(v_i^*,w_j) $$ and that, for $k=1,2,\dots,m$, $$ \phi(v_i^*,w_j)(v_k)=v_i^*(v_k)w_j=f_{ij}(v_k) $$ so $\alpha(v_i^*\otimes w_j)=f_{ij}$. In particular $\alpha$ is surjective and the set $\{v_i^*\otimes w_j:1\le i\le m,1\le j\le n\}$ is linearly independent. Since it is also clearly a spanning set for $V^*\otimes W$, it is a basis.


Note. The universal property has been used for defining $\alpha$. In case $V$ is not finite dimensional, the map $\alpha$ is not surjective any more; the set $\{f_{ij}\}$ can be defined from bases, but it's not a basis for $\operatorname{Hom}(V,W)$; it's just linearly independent. So the map $\alpha$ is injective in general, but it is surjective if and only if $V$ is finite dimensional.

$\endgroup$

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.