Is the assumption that supports do not increase really needed in the Peetre theorem?

68 Views Asked by At

Consider the Peetre theorem, as stated here on wikipedia. Isn't the assumption that $D$ not increase supports reduntant? I would think that the fact that $D$ is a morphism of sheaves already implies this.

Let us consider the simplest example, $D$ is an endmorphism of the (vector space valued) sheaf of locally-defined smooth functions $\mathbb{R} \to \mathbb{R}$. Now, there are two ways to interpret the assumption that $D$ does not increase supports, corresponding to the two common defintions of the support of a smooth map $f : U \to \mathbb{R}$:

  1. Sometimes, the support of $f$ is just $\{x \in U : f(x) \neq 0\}$.
  2. Sometimes, the support is the closure of the above set.

I think it is safe to assume that we are not using the first definition, since even the ordinary derivative increases supports if they are defined this way. Consider $f(x) = x$ on $\mathbb{R}$. Then, the support of $f$ is $\mathbb{R}\setminus \{0\}$, but the derivative $f'(x)=1$ is supported on the whole line.... so in the sense the support has increased.

OK, so we may assume that when we say $D$ does not increase supports that we are talking about the closed supports. But now, let $f : U \to \mathbb{R}$ be a locally defined smooth function. I claim that $D$ does not increase the support of $f$ simply by virtue of being a sheaf morphism. Indeed, let $x$ belong to $U \setminus \overline{\mathrm{supp}}(f)$. Then, there is a small open set $V \subset U$ with $x \in V$ and $V \cap \overline{\mathrm{supp}}(f) = \varnothing$. Since $D$ is a sheaf morphisms, we have that $$(Df)|_V = D(f|_V) = D( \text{the zero function on }V) = \text{the zero function on }V,$$ by linearity. Thus, $x$ is also not in the support of $Df$, and the support is not increased.