I've been reading a bit about tensors on Wikipedia (so correctness not guaranteed here) and I have a question. The order of a tensor $T$ is defined as $n+m$, where $n$ denotes the number of covariant indices and $m$ the number of contravariant indices. Wikipedia gives linear transformations as examples of second order tensors and scalars as examples of 0 order tensors. Here's what I don't understand:

For the sake of concreteness, let's take a particular example. Consider the vector space $\mathbb{R}^n$ (over $\mathbb{R}$) with the linear transformation $T: \mathbb{R}^n \rightarrow \mathbb{R}^n$, $T: v \mapsto \alpha v$ for some $\alpha \in \mathbb{R}$. It is my understanding that in any basis we may represent this transformation by the array $\alpha I$, where $I$ is the $n \times n$ identity matrix. Two indices are required to uniquely specify each element of this array, hence $T$ should be a second order tensor, as dictated by Wikipedia. However, what if we simply represented $T$ by the array $(\alpha)$? This is just a scalar, so by Wikipedia this is a 0 order tensor. Does the order of a tensor depend on our representation of it somehow?

Edit: $(\alpha)$ is to be interpreted as a $1 \times 1$ matrix.

  • $\begingroup$ If you write it as a scalar it is no longer a matrix. The matrix representation is always $n\times n$. $\endgroup$ Jan 16, 2016 at 15:56
  • $\begingroup$ I put the scalar in parentheses here to denote a $1 \times 1$ matrix. $\endgroup$
    – M10687
    Jan 16, 2016 at 15:57
  • 2
    $\begingroup$ A $1\times 1$ matrix doesn't act on $\mathbb R^n$ if $n>1$, just as a $6\times 7$ matrix can't be applied to $\mathbb R^{10}$. Writing it as a scalar is akin to giving it a name, not changing the matrix representation. $\endgroup$ Jan 16, 2016 at 16:00

1 Answer 1


The "type" of a tensor $T$—i.e., of a tensor-valued function, or "tensor field"—determines how $T$ transforms under change of coordinates.

  • If you view a scalar as a scalar-valued function it behaves like a $0$-tensor.

  • If you view a scalar as the operation of scalar multiplication, it behaves like a $2$-tensor.

Strictly speaking those are distinct interpretations, not mere matters of notation. (Matrix notation reflects this distinction, incidentally: You can't (matrix-)multiply a $1 \times 1$ matrix $(\alpha)$ by an $n \times n$ matrix.)


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.