Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

In S.S.Chern's Lectures on Differential Geometry, I don't understand the following text in Chapter 2, which introduces the tensor product:

The tensor product $V^*\otimes W^*$ of the vector spaces $V^*$ and $W^*$ refers to the vector space generated by all elements of the form $v^*\otimes w^*$, $v^*\in V^*$, $w^*\in W^*$. It is a subspace of ${\mathcal L}(V,W;{\mathbb F})$. We need to point out that any element in $V^*\otimes W^*$ is a finite linear combination of elements of the form $v^*\otimes w^*$, but generally cannot be written as a single term $v^*\otimes w^*$ (the reader should construct examples).

Here are my questions:

  • What does the first sentence mean? Does it mean $$V^*\otimes W^*:=\operatorname{span}\{v^*\otimes w^*|v^*\in V^*, w^*\in W^*\} $$ or $$V^*\otimes W^*:=\{v^*\otimes w^*|v^*\in V^*, w^*\in W^*\} ?$$
  • In the context, it is only defined that $$v^*\otimes w^*(v,w)=v^*(v)\cdot w^*(w).$$ What's the "finite linear combination of elements of the form $v^* \otimes w^* $" supposed to be defined? And what's the example "the reader needs to construct"?
share|cite|improve this question
By the way: when $V$ and $W$ are finite-dimensional, the span of these elements is all of $L(V, W; F)$. – Dylan Moreland Jul 12 '11 at 2:36
up vote 11 down vote accepted

The exposition you're quoting may be somewhat confusing if you're used to how tensor products are usually defined in a general setting (see e.g. the Wikipedia article). Usually, the tensor product is defined as a new vector space, either through a universal property or by explicit construction using equivalence classes. In your case however, it's being regarded as a subspace of an existing vector space, the space of all bilinear functions from $V\times W$ to $\mathbb F$. That allows the book to speak of linear combinations without introducing these as formal expressions, since it's already known how to form linear combinations of bilinear functions from $V\times W$ to $\mathbb F$.

To answer your questions specifically: Yes, the vector space generated by a set of elements is the span of those elements. There's a subtle difference in that the formulation "the vector space generated by $S$" can also be used to refer to the free vector space over $S$, whereas the formulation using the span can't be thus used and only serves to identify a subspace of a vector space already otherwise defined.

Concerning the examples to be constructed, consider linearly independent functions $v^*_1,v^*_2\in V^*$ and $w^*_1,w^*_2\in W^*$, and form $v^*_1\otimes w^*_1+v^*_2\otimes w^*_2$; I think you'll find that you can't express this in the form $v^*\otimes w^*$ for any $v^*\in V^*$ and $w^*\in W^*$.

share|cite|improve this answer
It's unfortunate that we (collectively)have inherited some bit of a culture in which "formal expression" can mean "marks on a page" or "meaningless symbols", without protest. It is true that the set-theoretic "construction" of many things is unsatisfactory, but one could, and I'd strongly recommend it, refer to the property one requires/needs, so that the object-in-question is unique-up-to-unique-isomorphism, at least. Existence/construction truly is a different matter. E.g., "polynomials" are not "expressions", but elements of a universal object in a category of commutative rings... – paul garrett Jul 11 '11 at 23:06
To be honest, I'm not sure which part of the answer, if any, you're criticizing. I wrote first that the tensor product can be defined "either through a universal property or by explicit construction using equivalence classes". Then I specifically addressed the OP's concern how the finite linear combinations are defined. Does your complaint concern the fact that I wrote only "introducing these as formal expressions" and not "either introducing these as formal expressions or postulating their existence in a universal object"? – joriki Jul 11 '11 at 23:16
I'm sorry, I should have been more careful not to imply (negativist) criticism of your answer. I mean to criticize something widespread in the mathematical environment, that amounts to laziness or ignorance, and/but should surely lead to confusion on the part of any serious, thinking reader. So, in fact, it is necessary, and, therefore, a good thing, to address a "questioner"'s confusion derived from the failings of the environment. I did not mean to be negative to you about addressing this! Sorry to have left such an impression! And/but we (collectively) might change the feeble/lazy texts... – paul garrett Jul 12 '11 at 0:39
@paul: can't we define tensor algebra as the largest model of associative algebra axioms that contains the given vector space? – Alexei Averchenko Aug 2 '11 at 6:05
@Alexei A... There are indeed various possibilities. Without in any way faulting joriki's response to the question as posed, for a definition/characterization I was thinking of a "tensor" algebra of a v.s. V as being as assoc alg A with a v.s. map V->A such that for any v.s. map V->B to an assoc alg B, there is a unique assoc alg map A->B making the obvious triangle commute. No coordinates, no interpretation of symbols, etc. But, as in the question as posed, usually one must work through a transitional viewpoint, as well. – paul garrett Aug 2 '11 at 12:38

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.