Uniqueness of the matrix representation of tensors

81 Views Asked by At

Note that both maps below satisfy the universal property of the tensor product. \begin{align*} \mathbb{R}^2 \times \mathbb{R}^2 &\rightarrow \mathbb{R}^{2 \times 2} &\mathbb{R}^2 \times \mathbb{R}^2 &\rightarrow \mathbb{R}^{2 \times 2} \\ \begin{bmatrix} a_1 \\ a_2 \\ \end{bmatrix} \times \begin{bmatrix} b_ 1 \\ b_2 \\ \end{bmatrix} &\mapsto \begin{bmatrix} a_1 b_ 1 & a_1 b_2 \\ a_2 b_1 & a_2 b_2 \\ \end{bmatrix} &\begin{bmatrix} a_1 \\ a_2 \\ \end{bmatrix} \times \begin{bmatrix} b_ 1 \\ b_2 \\ \end{bmatrix} &\mapsto \begin{bmatrix} a_1 b_ 2 & a_1 b_1 \\ a_2 b_2 & a_2 b_1 \\ \end{bmatrix} \\ (a_i) \times (b_i) &\mapsto (c_{ij} = a_i b_j) &(a_i) \times (b_i) &\mapsto (c_{ij} = a_i b_{\tau(j)}) \text{ where $\tau = (12) \in S_2$} \end{align*} Similarly, for $(\sigma_1, \sigma_2, \dots, \sigma_d) \in S_{n_1} \times S_{n_2} \times \dots S_{n_d}$, the map
\begin{gather*} K^{n_1} \times K^{n_2} \times \dots \times K^{n_d} \ \xrightarrow{\Gamma_{(\sigma_1, \sigma_2, \dots, \sigma_d)}} \ K^{n_1 \times n_2 \times \dots \times n_d} \\ (x_i^1) \times (x_i^2) \times \dots \times (x_i^d) \ \mapsto \ (c_{{i_1}{i_2}\dots{i_d}} = x^1_{\sigma_1(i_1)} x^2_{\sigma_2(i_2)} \dots x^d_{\sigma_d(i_d)}) \end{gather*} satisfies the universal property.

Are these the only maps that satisfy the universal property? Formally, is it true that:

if $K^{n_1} \times K^{n_2} \times \dots \times K^{n_d} \ \xrightarrow{\Gamma} \ K^{n_1 \times n_2 \times \dots \times n_d}$ satisfies the universal property of the tensor product, then $\Gamma = \Gamma_{(\sigma_1, \sigma_2, \dots, \sigma_d)}$ for some $(\sigma_1, \sigma_2, \dots, \sigma_d) \in S_{n_1} \times S_{n_2} \times \dots S_{n_d}$.

I asked a similar question here Proof of unique coordinatization of tensor space once bases are chosen but I think I confused people in how the question was stated, so I am writing it in a different way here. Thank you for any help you can offer.

1

There are 1 best solutions below

1
On

No, there are many more such "tensor maps". For instance, (I think, as you didn't specify exactly which tensor product property you wanted) you can compose with any automorphism of $\mathbb{R}^{2 \times 2}$ to create a different tensor map.

For instance, for any invertible matrix $P$, $(a,b) \longmapsto P^{-1}\Gamma_{\operatorname{id},\operatorname{id}}(a,b)P$ should satisfy the property.

For a concrete example, take $P=\begin{bmatrix}0 & 1\\-1 & 0\end{bmatrix}$ and consider the map $$\begin{bmatrix}a_1\\ a_2\end{bmatrix} \times \begin{bmatrix}b_1\\ b_2\end{bmatrix} \longmapsto \begin{bmatrix} a_2b_2 & -a_2b_1 \\-a_1b_2 & a_1b_1\end{bmatrix}.$$