Let $M$ be a differentiable manifold, $TM$ and $T^*M$ a tangent and cotangent bundle of $M$ and let $\Gamma (TM),\ \Gamma (T^*M)$ be spaces of smooth sections of $TM$ and $T^*M$. Let $T_s^r (M)$ denote a space of smooth tensor fields on $M$.

Since vector field on $M$ is a smooth section of $TM$, clearly
$$T_0^1 (M)=\Gamma (TM)$$ and similarly $$T_1^0 (M)=\Gamma (T^*M).$$

My question is related to something I saw in some lecture notes on differential geometry. There was a definition of a tensor field as a smooth section of $TM \otimes \dots \otimes TM \otimes T^*M \otimes \dots \otimes T^*M$. But since these bundles are not vector spaces, it seems a little suspicious to me. I think the product of (co)tangent bundles may be DEFINED "fiber-wise" in the following sense

$$ TM \otimes TM \equiv \bigcup_{p \in M} \{T_pM \otimes T_pM\} \ \tag{*}$$ (disjoint union). This would also correspond with the definition of tensor field I know. Since tensor field of some type is a map which assigns to every point $p \in M$ a tensor from $T_pM \otimes \dots \otimes T_pM \otimes T_p^*M \otimes \dots \otimes T_p^*M$ smoothly, one has

$$ T_s^r (M) = \Gamma \left( \bigcup_{p \in M} \{T_pM \otimes \dots \otimes T_pM \otimes T_p^*M \otimes \dots \otimes T_p^*M\}\right).$$

So my question is whether the following equalities are true:

$$ \bigotimes^r TM \otimes \bigotimes^s T^*M = \bigcup_{p \in M} \{\bigotimes^r T_pM \otimes \bigotimes^s T_p^*M\}\ \tag{1}$$

$$\Gamma \left( \bigotimes^r TM \otimes \bigotimes^s T^*M \right) = \bigotimes^r \Gamma (TM) \otimes \bigotimes^s \Gamma (T^*M) \ \tag{2}$$

and if the tensor product of (co)tangent bundles is not defined the way I proposed in $(*)$, what is the correct definition or from which more general concept it follows?


I think you are looking for the concept of a smooth functor. Essentially, any functorial construct on vector spaces, such as taking duals, tensor products and direct sums, can be applied to a vector bundle, if for all vector spaces $V, W$ the map $$\mathcal F : \operatorname{Hom}(V,W) \to \operatorname{Hom}(\mathcal F(V), \mathcal F(W))$$ is smooth using the usual definition of what it means for a map between vector spaces to be smooth.

Thus if $E$ and $F$ both bundles over $M$ with fibers $V$ and $W$,, we can find a bundle denoted $E \otimes F$, also over $M$, the fiber of which is $V\otimes W$. So your first equality is true. In fact the proof is to do as you suggest and take the disjoint union of the fiber-wise tensor product. Then you only have to check that this gives a smooth bundle, but since the map $(A, B) \mapsto A\otimes B$ is smooth for linear transformations $A$ and $B$ between finite-dimensional vector spaces, this is the case.

On the level of sections, you also have $$\Gamma(E \otimes F) \cong \Gamma(E)\otimes\Gamma(F)$$ but the tensor product on the right is the tensor product of $C^\infty(M)$-modules. There is not strict equality here, but the spaces are naturally isomorphic. Here, naturally means that you can construct an isomorphism that doesn't depend on arbitrary choices. (To clarify: you probably know that if $V$ and $W$ are vector spaces of the same dimension, you can find an isomorphism $V \cong W$ by choosing bases for $V$ and $W$. But this isomorphism depends on which bases you choose, so they are not naturally isomorphic. However, $V \otimes W$ and $W \otimes V$ are naturally isomorphic. Can you think of why?)

  • $\begingroup$ To be precise, on the level of sections, don't we have an isomorphism $\Gamma(E\otimes F)\cong\Gamma(E)\otimes\Gamma(F)$ and not a strict equality? $\endgroup$ – Alex Nelson Mar 31 '14 at 20:54
  • $\begingroup$ I have no experience with functors but it seems very powerful. I'll check it out and try to absorb your answer. Thank you! $\endgroup$ – AlanHarper Mar 31 '14 at 22:00
  • $\begingroup$ @AlexNelson Yes, you are correct. I've made an edit. $\endgroup$ – Robin Ekman Apr 1 '14 at 12:19
  • $\begingroup$ Dear @Robin, There is a natural map $\Gamma(E)\otimes\Gamma(F)\to\Gamma(E\otimes F)$, but is it really true that this is an isomorphism all the time? This is not such an easy thing in the category of, e.g., schemes. I actually asked a question about this (here: math.stackexchange.com/questions/492166/…) but never received an answer. $\endgroup$ – Keenan Kidwell Apr 1 '14 at 19:22
  • $\begingroup$ I do not know if it is an isomorphism for schemes. But it is in the smooth category. You can find a proof in Manifolds and Differential Geometry by Jeffrey Lee, Proposition 7.32. Actually the first four sections of Chapter 7 are all relevant to this question. $\endgroup$ – Robin Ekman Apr 1 '14 at 19:24

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.