Tensor confusion - Are coordinates a type of tensor?

280 Views Asked by At

(All of this question uses $V$ as the real vector space where tensors are being built from).

I am currently trying to learn about tensor notation, and I am running into a road block with understanding what exactly constitutes as a vector.

In the formal definition, a tensor seems to be defined bluntly as “an element of the tensor space over $V$”. For example:

  • $(0,1)$-tensors are elements of $V$.
  • $(1,0)$-tensors are elements of $V^*$.
  • $(1,1)$-tensors are elements of $V \otimes V^*$

However now comes my confusion. When we think of a vector $v \in V$ with respect to a (say three dimensional) basis $\{e_1, e_2, e_3\}$, then in tensor notation we write $v = a^i e_i$, where $a^1, a^2, a^3$ are the coordinates in this basis.

Then $a^i$ gets called a “contravariant tensor” (in other words a $(1,0)$-tensor), but how is it a tensor with respect to the formal definition? Surely each $a^i \in \mathbb{R}$, and $a^i \notin V^*$.

The same goes for something like the metric tensor, which I see denoted as $g_{i j}$, but this is just the components and again just a bunch of elements from $\mathbb{R}$?

2

There are 2 best solutions below

4
On BEST ANSWER

A contravariant tensor here means an $(1,0)$ tensor, that is, an element of $V^*$, i.e. a linear map $V\to\Bbb R$.
Specifically here, we get the elements of the dual basis of $e_1,\dots, e_n$, which are just the coordinate mappings: $\sum_ia^ie_i\mapsto a^j$.

The metric tensor is of type $(2,0)$, as it is a bilinear map $V\times V\to\Bbb R $, which are in one-to-one correspondence with elements of $V^*\otimes V^*$.

0
On

Tensors are invariant objects so their coordinates aren’t themselves tensors The confusion comes from authors identifying tensors with their components because the math is easier, especially when dealing with higher rank tensors.