I have a question about tensors. Okay, you could say that this question would fits more properly on Math.StackExchange but, here it's more cozy. Anyway:

A vector, as we all learned in first year, could be written in this form: \begin{equation} \textbf{V} = \sum A_{i} \textbf{e}_{i} \end{equation}

Okay, the thing begins more "rich" with linear algebra when, due to the formalism, one could now tell the difference between a contravariant vector and a covariant vector. Respectively:

\begin{equation} \textbf{V} = \sum A^{i} \textbf{e}_{i} , \textbf{V} = \sum A_{i} \textbf{e}^{i} \end{equation}

If you are quite curious you may reach the study of tensors. Following the same "idea" one could now write a tensor (of rank 3) with the same "shape" of a vector. A contravariant tensor and a covariant tensor. Respectively:

\begin{equation} \textbf{T} = \sum \sum \sum T^{ijk} \textbf{e}_{i}\textbf{e}_{j}\textbf{e}_{k},\\ \textbf{T} = \sum \sum \sum T_{ijk} \textbf{e}^{i}\textbf{e}^{j}\textbf{e}^{k} \end{equation}


The first thing that we learn when begin to talk about tensors, in a linear algebra perspective, is the tensor product. Because of that I know that what I've wrote above is, in fact:

\begin{equation} \textbf{T} = \sum \sum \sum T^{ijk} \textbf{e}_{i}\otimes \textbf{e}_{j}\otimes \textbf{e}_{k} \end{equation}

The point is, I still did not grasp the notion that $\textbf{e}_{i}\otimes \textbf{e}_{j}$ could be a fundamental way to think about inner products for example, or even what $\textbf{e}_{i}\otimes \textbf{e}_{j}\otimes \textbf{e}_{k}$ really really means in this concept; can someone help with this?. I mean "producing tensors" in the form as above.

  • $\begingroup$ Hi Countto10. So, I'm Brazillian, and most of my sources are in portuguese. But, books like: Marion (classical mechanics), Hartle (general relativity); Greub (linear algebra) and Hoffman-Kunze (Linear algebra) are my main sources. $\endgroup$
    – M.N.Raia
    Jul 2 '17 at 12:50
  • $\begingroup$ I think one way to think is that they "eat" eachother. Ex: $[e_1\otimes e_2](e^1\otimes e^2)=e^1(e_1)\cdot e^2(e_2)$. Ex: $[e^1\otimes e^2](e_1\otimes e_2)=e^1(e_1)\cdot e^2(e_2)$ $\endgroup$
    – Emil
    Jul 2 '17 at 16:22
  • $\begingroup$ I also believe that contraction is when you just feed one of them. Ex: $C(2,1)(a\otimes b\otimes c,d)= b^i(d_i)a\otimes c$ (note: haven't verified this. also used a makeshift notation since I haven't found the real notation). $\endgroup$
    – Emil
    Jul 2 '17 at 16:34
  • $\begingroup$ Also, the components are $T^{ijk}=\mathbf{T}(e^i\otimes e^j \otimes e^k)$. The $e_i\otimes e_j \otimes e_k$ are there to handle all other cases (since T is multilinear you can let them handle the other possible arguments for T). Unless I am mistaken. $\endgroup$
    – Emil
    Jul 2 '17 at 16:49
  • $\begingroup$ One more thought, maybe the contraction example above should have been written like $C(2,4)(a\otimes b\otimes c\otimes d)$, to stay in tensor-land, so to speak.. $\endgroup$
    – Emil
    Jul 2 '17 at 20:09

due to the formalism, one could now tell the difference between a contravariant vector and a covariant vector

This is incorrect: covariant (respectively contravariant) does not refer to the vector, rather to the transformation properties of the components of such vector along some basis.

A vector $\mathbf{v}$ is an element of a vector space $V$. A co-vector is a map from a vector space into a field $\alpha\colon V\to\mathbb{F}$ and as such an element of the dual vector space $V^*$. A tensor of type $(r,s)$ is a map $\tau\colon V^r\times {V^*}^s\to\mathbb{F}$.

Given the above, a tensor product of two tensors $T\colon A\to X$, $S\colon B\to Y$ is another linear map defined as $$ T\otimes S\colon A\otimes B\to X\otimes Y $$ such that $(T\otimes S)(a\otimes b)=T(a)\otimes S(b)$ however you choose $a\in A, b \in B$.

By linearity you can construct tensor product of any types.


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.