Measures uniquely determined by their restriction to $ C_0^\infty $

110 Views Asked by At

In Lars Hörmander's Linear Partial Differential Operators (1963 edition) the following is stated on the bottom of page 3.

... Further, which is important for the definition of distributions, a measure is uniquely determined by its restriction to $ C_0^\infty (\Omega) $. For if $ \mathrm{d} \mu $ is a measure in $ \Omega $ and $ \int u \: \mathrm{d} \mu = 0 $ for every $ u \in C_0^\infty (\Omega) $, then $ \int u \: \mathrm{d} \mu = \lim \int u_\epsilon \: \mathrm{d} \mu = 0 $ for every $ u \in C_0^0 (\Omega) $.

I would be very grateful if someone would be so kind to help me understand this comment. What does it mean for a measure to be "determined completely"? And why does the second sentence demonstrate the foregoing assertion? I am a little stuck...

N.B. $ u_\epsilon $ was previously defined by $ u_\epsilon(x) = \int u(x-\epsilon y) \: \varphi (y) \: \mathrm{d}y = \epsilon^{-n} \int u(y) \: \varphi(x-y/\epsilon) \: \mathrm{d}y $, where $ \varphi \in C_0^\infty (\mathbb{R}^n) $ with $ \mathrm{supp} (\varphi) = \{ x : \lvert x \rvert \leq 1 \} $, $ \varphi \geq 0 $ and $ \int \varphi \mathrm{d} x = 1 $ (i.e. $ \varphi $ is a mollifier).

2

There are 2 best solutions below

2
On BEST ANSWER

"Determined completely" means that if $\int f\,d\nu_1 = \int f\,d\nu_2$ for every $f \in C^\infty_0(U)$, then $\nu_1 = \nu_2$.

To see why the second sentence is true, let's suppose $\mu$ is a finite measure. We know that $u_\epsilon$ converges to $u$ pointwise. Moreover, if $C = \sup |u|$ (which is finite because $u$ is continuous and compactly supported), you can easily check that $|u_\epsilon| \le C$ as well. Hence $\int u_\epsilon\,d\mu \to \int u\,d\mu$ by the dominated convergence theorem (using $C$ as your dominating function).

Now if it holds for finite measures, then it also holds for signed measures $\mu$ (consider the Jordan decomposition $\mu = \mu^+ - \mu^-$).

Now suppose $\nu_1, \nu_2$ are finite measures and $\int f\,d\nu_1 = \int f\,d\nu_2$ for all $f \in C^\infty_0$. Let $\mu = \nu_1 - \nu_2$. Then $\int f\,d\mu = 0$ for all $f \in C^\infty_0$. As discussed above, it follows that $\int f\,d\mu = 0$ for all $f \in C^0_0$. This means $\mu$ must be the zero measure; this is the easy direction of the Riesz-Markov theorem, and can be proved using the monotone class theorem or some variant. If $\mu = 0$ then $\nu_1 = \nu_2$.

If $\nu_1, \nu_2$ are not necessarily finite, but are finite on compact sets, then you get the same conclusion by approximating $\Omega$ by a sequence of bounded sets.

1
On

This follows from the fact that $C^{\infty}_0 (\Omega )$ is dense in $C^0_0 (\Omega ) $ and monotone convergence theorem