On p. 11 of Warner's Foundations of Differntiable Manifolds and Lie Groups, he discusses partitions of unity. The theorem says

Let $M$ be a differentiable manifold and $\{U_\alpha: \alpha \in A \}$ an open cover of $M$. Then there exists a countable partition of unity $\{ \varphi_i : i = 1, 2, 3, \dotsc \}$ subordinate to the cover $\{U_\alpha\}$ with $\operatorname{supp}\varphi_i$ compact for each $i$. If one does not require compact supports, then there is a partition of unity $\{\varphi_\alpha \}$ subordinate to the cover $\{U_\alpha \}$ (that is, $\operatorname{supp} \varphi_\alpha \subset U_\alpha$) with at most countably many of the $\varphi_\alpha$ not identically zero.

I am fine with the first part of the theorem: he constructs a countable partition of unity $\varphi_i$ with compact supports. Then, for the second part, he says

If we let $\varphi_\alpha$ be identically zero if no $\varphi_i$ has support in $U_\alpha$, and otherwise let $\varphi_\alpha$ be the sum of the $\varphi_i$ with support in $U_\alpha$, then $\{\varphi_\alpha\}$ is a partition of unity subordinate to the cover $\{U_\alpha\}$ with at most countably many of the $\varphi_\alpha$ not identically zero. To see that the support of $\varphi_\alpha$ lies in $U_\alpha$, observe that if $\mathcal{A}$ is a locally finite family of closed sets, then $\overline{\bigcup_{A \in \mathcal{A}}A} = \bigcup_{A \in \mathcal{A}} A$. Observe, however, that the support of $\varphi_\alpha$ is not necessarily compact.

If $\sum_\alpha \varphi_\alpha$ is to be identically $1$, then we have to make sure that we're not counting any $\varphi_i$ twice, right? But we will count it twice if $\operatorname{supp}\varphi_i$ lies in $U_\alpha$ for more than one $\alpha$, right? So perhaps we have to renormalize afterward...that would be OK, but how do we know that $\operatorname{supp}\varphi_i$ doesn't lie in infinitely many $U_\alpha$?

Does it have to do perhaps with the particular construction of the $\varphi_i$ he's chosen? (I don't see how.)

For instance, a stupid example which seems to break this construction is the open cover of a manifold $M$ by infinitely many copies of the single set $M$. Then take the function $\varphi$ which is identically 1 on $M$...this works for the first part. But in defining $\varphi_\alpha$ in the second part, we seem to be trying to sum up infinitely many copies of $\varphi$.


I think you're right. A better way to define $\varphi_\alpha$ would be to choose a map $a\colon \mathbb Z^+\to A$ such that $\operatorname{supp}\varphi_i\subseteq U_{a(i)}$ for each $i$, and then define $\varphi_\alpha$ to be $$ \varphi_\alpha = \sum_{i: a(i)=\alpha} \varphi_i. $$ This is in fact the definition I used in my Introduction to Smooth Manifolds (2nd ed., p. 44).


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.