Decomposition of forms in the proof of Poincaré's lemma

351 Views Asked by At

On page 34 of Bott and Tu, Differential Forms in Algebraic Topology, where the author proves that the maps $\pi: \Bbb R ^n \times \Bbb R^1 \rightarrow \Bbb R^n$, $s: \Bbb R^n \rightarrow \Bbb R^n \times \Bbb R^1$ ($x \mapsto (x,0))$ induce an isomorphism $$H^*(\Bbb R^n \times \Bbb R^1) \rightarrow H^*(\Bbb R^n)$$ in the de Rham cohomology, he stated that

Every form on $\Bbb R^n \times \Bbb R$ is uniquely a linear combination of the following two types of forms:

  1. $(\pi^* \phi ) f(x,t)$,
  2. $(\pi^* \phi) f(x,t) dt$,

where $\phi$ is a form on the base $\Bbb R^n$.

The author states the same argument works when $\Bbb R^n$ is in general replaced by a manifold $M$. But the argument here requires that we work locally on charts $\{U_\alpha \times \Bbb R^1 \}$?

I could not really make sense of how to write this "coordinate independent" argument out rigorously:

I know we can take a partition of unity such that all forms are of linear combination of given form. But I do not make sense of "uniqueness".

1

There are 1 best solutions below

0
On

Not entirely sure if this is what you're stuck on, but:

More explicitly, uniqueness in the general case means that given any form $\omega \in \Omega^q(M \times \mathbb R) \subset \Omega^*(M \times \mathbb R)$, there exists a unique decomposition $\omega = \omega_1 + \omega_2$ where $\omega_1 = (\pi_1^* \phi_1)f_1$ and $\omega_2 = (\pi_1^* \phi_2)(\pi_2^* dt)f_2$ for some $\phi_1 \in \Omega^q(M \times \mathbb R)$, $\phi_2 \in \Omega^{q-1}(M \times \mathbb R)$, and $f_1, f_2 \in C^\infty(M \times \mathbb R)$. Here $\pi_1$ and $\pi_2$ are the projections onto the respective factors. Note that the $\phi_i$ and the $f_i$ themselves are not necessarily unique.

An important point for the question is that the form $\pi_2^* dt$ (or just $dt$) exists uniquely, independent of the coordinate (up to some coefficient which could be absorbed into the factor $(\pi_1^* \phi_2) f_2$ instead) due to the product structure. It's characterized by the property that its null space at each point is the embedded corresponding tangent space of $M$; add any form of the type $(\pi_1^* \phi) f \neq 0$ to it and we lose that property. And it's natural to give special treatment to $dt$ in the proof: The extra factor of $\mathbb R$ is the difference between the two base manifolds, after all.