1
$\begingroup$

It is often stated that a choice of a basis for a vector space over $R$ gives an isomorphism between $V$ and $R^n$. However, I am having a hard time showing this and confused over some issues. For simplicity, let's stick to a two-dimensional space $V$. Let's choose a basis $\{e_1, e_2\}$ for the vector space. Then we have the linear map $\phi: R^2 \to V; \phi(v_1,v_2)=v_1 e_1+v_2 e_2$. To prove the isomorphism between $R^2$ and $V$, I need to construct a map $\phi^{-1}: V\to R^2: \phi^{-1} \phi=Identity$. I don't see any obvious way to construct such a map. However, if there is an inner product (denoted here by ".") on $V$, we can construct such a map in the following way. Suppose $e_1,e_2$ is an orthonormal basis. We then set $\phi^{-1}: V\to R^2; \phi^{-1}(v)=(e_1.v, e_2.v)$. It is obvious that $\phi^{-1}\phi(v_1,v_2)=(v_1,v_2)$. So my question is, is it true that we need an inner product and the existence of an orthonormal basis to prove the isomorphism between $V$ and $R^2$? Or I have misunderstood something here?

Edit: I also would like to mention that this question originated from noticing the fact that although a given vector can be written as $v=v_1e_1+v_2e_2$, there is no natural way to find the coefficients $v_1,v_2$. The inner product seems to determine the values of $v_1,v_2$.

$\endgroup$
2
  • $\begingroup$ Send your basis to a basis. That‘s induces an isomorphism. $\endgroup$ – Qi Zhu Sep 27 '20 at 12:44
  • $\begingroup$ +1 for the edit. That's true the inner product gives you a simple method to "pick" the suitable coefficients $\endgroup$ – Peter Melech Sep 27 '20 at 17:52
2
$\begingroup$

The inner product is not relevant to the question.

In order that $\phi\colon\mathbb{R}^2\to V$ is an isomorphism it is necessary and sufficient that it is injective and surjective (besides being linear, of course).

Since $\{e_1,e_2\}$ is a spanning set, the map is surjective.

Since $\{e_1,e_2\}$ is linearly independent, the map is injective.

$\endgroup$
1
$\begingroup$

You don't need an inner product. For $\{e_1,e_2,...,e_n\}$ to be a basis of the finite-dimensional vectorspace $V$ (by definition of a basis or as a consequence of it being a maximal linearly independent set) there exists for every vector $v\in V$ a $\textbf{unique}$ $n$-tuple $$\begin{pmatrix}v_1\\v_2\\ \cdot\\\cdot\\\cdot\\v_n\end{pmatrix}\in\mathbb{R}^n$$ such that $v=\sum_{j=1}^nv_je_j$ and you define $\phi^{-1}(v)=\begin{pmatrix}v_1\\v_2\\ \cdot\\\cdot\\\cdot\\v_n\end{pmatrix}$. Now it is easy to check that $\phi\circ\phi^{-1}=id_{V}$ and $\phi^{-1}\circ\phi=id_{\mathbb{R}^n}$.

$\endgroup$

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.