# Tensor algebra question [closed]

1)Why embedding of ( not necessarily finite-dimensional) vector spaces $V\rightarrow W$ produces embedding of tensor algebras $T(V)\rightarrow T(W)$. I can prove it using Hamel basis in $W$ but is there a nicer ( more functorial ) argument? 2) How to prove the same statement for modules over an algebra instead of vector spaces?

-

## closed as too localized by Mariano Suárez-Alvarez♦, Pietro Majer, Martin Brandenburg, Dmitri Pavlov, Dan PetersenJun 29 '11 at 7:58

This question is unlikely to help any future visitors; it is only relevant to a small geographic area, a specific moment in time, or an extraordinarily narrow situation that is not generally applicable to the worldwide audience of the internet. For help making this question more broadly applicable, visit the help center.If this question can be reworded to fit the rules in the help center, please edit the question.

A better place to ask this would have been math.stackexchange.com, by the way; see the FAQ for details on the reason. – Mariano Suárez-Alvarez Jun 28 '11 at 16:58
Even for modules, the ultimate reason is the universal property of T(V), as a limit object. From this it immediately springs out the structure of functoriality and adjunction, according to general categorical facts. A nice reading is, of course, Mac Lane's Categories for the Working Mathematician, in particular, the chapter about adjoint functors. – Pietro Majer Jun 28 '11 at 17:06
Do you really mean "embedding" or just a morphism? In the case of modules, only split monomorphisms are mapped to split monomorphisms (in fact, via any functor), but $T(-)$ does not preserve monomorphisms in general. – Martin Brandenburg Jun 28 '11 at 17:25
Martin, by embedding I mean injective map, not just a morphism. Thank you for the answer. – MathAndMe Jun 28 '11 at 17:40
Martin, by the way, can you give some easy example of injective map of modules over rings such that corresponding map of tensor algebras is not injective? – MathAndMe Jun 28 '11 at 17:55

If $V$ is a subspace of $W$, consider the inclusion $f:V\to W$ and any map $g:W\to V$ such that $g\circ f=1_V$; to construct $g$, you need to use bases or something equivalent, for it does not exist over, say, a general ring...

Now $T(-)$ is a functor, so $T(g)\circ T(f)=T(1_V)=1_{T(V)}$. It follows that the map $T(f):T(V)\to T(W)$ is injective.

-
Yes, that's what I had in mind, I was asking to see if it is inavoidable. Thank you! – MathAndMe Jun 28 '11 at 17:42

Let me give another answer to 1).

In general, given a linear mapping $$\phi \colon E \to F$$ it extends uniquely to a homomorphism $$T(\phi) \colon T(E) \to T(F).$$ The proof can be made coordinate-free, in fact it follows from the universal property of $T(E)$ applied to the map $$\eta \colon E \to T(F),$$ where $\eta=i \circ \phi$ and $i \colon F \to T(F)$ is the natural embedding.

By construction it follows

$$T(\phi)(x_1 \otimes \ldots \otimes x_p)=\phi x_1 \otimes \ldots \otimes \phi x_p.$$

If $\psi \colon F \to G$ is another linear map one obtains $$T(\psi \circ \phi)=T(\psi) \circ T(\phi),$$ hence $T(\phi)$ is injective [risp. surjective] whenever $\phi$ is injective [resp. surjective].

For more details, see for instance [Greub, Multilinear Algebra, Chapter III].

-
"By construction it follows" - no it doesn't. – darij grinberg Jun 28 '11 at 20:03
Why not? Given an arbitrary associative algebra $A$ with unit element $e$, and a homomorphism $\eta \colon E \to A$, by the universal property of $T(E)$ there exists a unique homomorphism $h \colon T(E) \to A$ such that $h(1)=e$ and which extends $\eta$; this is given precisely by $h(x_1 \otimes \ldots \otimes x_p)=\eta x_1 \ldots \eta x_p$ (see Greub's book). Now apply this result with $A=T(F)$ and $\eta=i \circ \phi$. Am I missing something? – Francesco Polizzi Jun 28 '11 at 21:48
@darij: Certainly, if you believe that $\phi$ extends to a unique (ring) homomorphism, then it must be the one Francesco gives on pure tensors, as that certainly is a homomorphism extending $\phi$, isn't it? The part that I don't see how it follows is the injectivity/surjectivity. Over any ring, the unique homomorphism extending $\phi$ is the Francesco's, but Martin says that $T$ does not preserve monomorphisms in general. – Theo Johnson-Freyd Jun 29 '11 at 1:02
@Theo: Over a field, the injectivity/surjectivity follows from the same functorial argument as in Mariano's answer. I have edited the post to make this clearer – Francesco Polizzi Jun 29 '11 at 7:45