7
$\begingroup$

Is there any hint to prove that for every $1 \le p < \infty $ the Schwartz Class is dense in $L^p$?

Thanks so much.

$\endgroup$
  • 5
    $\begingroup$ It might be easier to show that the compactly supported smooth functions are dense. These are contained in the Schwartz space of course. $\endgroup$ – Cameron Williams Dec 25 '15 at 20:08
  • $\begingroup$ @CameronWilliams how to do that? $\endgroup$ – user115608 Dec 25 '15 at 20:16
6
$\begingroup$

Hint:

Step 1. For every $f\in L^p$, and every $\varepsilon>0$, there exists an $M>0$, such that $$ \|\,f-\chi_{[-M,M]\,}f\|_{L^p}<\varepsilon, $$ where $\chi_A$ is the characteristic function of $A$. Set $g=\chi_{[-M,M]\,}f$.

Set 2. Let now $K_t(x)=(\pi t)^{1/2}\mathrm{e}^{-x^2/t}$, $t>0$, and define $g_t=g*K_t$. Then $$ \lim_{t\searrow 0}\|g_t-g\|_{L^p}=0, $$ and $g_t\in\mathscr S$, for all $t>0$.

| cite | improve this answer | |
$\endgroup$
  • 1
    $\begingroup$ thanks,but I think there are mistakes in the sign of the exponents of $K_t $,they both are negative. And then how do you say $g_t $ is a Schwartz function? $\endgroup$ – user115608 Jan 7 '16 at 18:26
2
$\begingroup$

Let me consider the case $1<p<\infty$.

Hint: $C_c^\infty(\mathbb{R}^n)\subset\mathcal{S}(\mathbb{R}^n)$ + Hahn-Banach Theorem + du Bois-Reymond Lemma.

Details:

Let $1<p<\infty$ and $C_c^\infty(\Omega)$ be the space of all compactly supported smooth functions $u:\Omega\subset\mathbb{R}^n\to\mathbb{R}$. Let $\phi:L^p(\Omega)\to\mathbb{R}$ be a continuous linear functional such that \begin{align}\phi(u)=0,\qquad\forall\ u\in C_c^\infty(\Omega).\tag{1}\end{align} Then, by the Riesz representation theorem, there exists $v\in L^{p'}(\Omega)$ such that \begin{align}\phi(f)=\int_\Omega fv\;dx,\quad\forall\ f\in L^p(\Omega).\tag{2}\end{align} Then, by $(1)$, \begin{align}\int_\Omega fv\;dx=\phi(f)=0,\quad\forall\ f\in C_c^\infty(\Omega).\tag{3}\end{align} By the du Bois-Reymond Lemma, it follows from $(3)$ that $v=0$ in $L^{p'}(\Omega)$. Thus, by $(2)$, \begin{align}\phi(f)=\int_\Omega f\cdot 0\;dx=0,\quad\forall\ f\in L^p(\Omega).\end{align} So, by (a corollary of) the Hahn-Banach Theorem, $C_0^\infty(\Omega)$ is dense in $L^p(\Omega)$.

The case $p=1$ (and other approach for the case $1<p<\infty$) can be found in the Brezis book.

| cite | improve this answer | |
$\endgroup$
  • 3
    $\begingroup$ I don't think that this is a good answer, since proving the Du Bois Reymond Lemma is essentially as hard as proving density itself. $\endgroup$ – PhoemueX Dec 25 '15 at 22:18
1
$\begingroup$

Let $f$ be a continuous function with compact support. Because such functions are dense in $L^p, 1\le p <\infty,$ it's enough to show $f$ can be approximated by Schwartz functions in all of these $L^p$ spaces.

Suppose $f$ is supported in $[-a,a].$ Let $\epsilon>0.$ By Weierstrass, there is a polynomial $q$ such that $|q-f|<\epsilon$ in $[-a,a].$

Define the functions

$$\begin {cases}\varphi_n(t) = \exp (1/[n(t^2-a^2)]),& t\in (-a,a)\\ 0, & |t| \ge a\\ \end {cases}$$

Each $\varphi_n$ is positive on $(-a,a)$ with support $[-a,a].$ We also have $\varphi_n \in C^\infty(\mathbb R),$ bounded by $1$ everywhere, and $\varphi_n \to\chi_{(-a,a)}$ pointwise everywhere. Of course each $\varphi_n$ is in the Schwartz space.

Now $q\varphi_n$ extends to be in $C_c^\infty(\mathbb R)$ in the obvious way. Thus

$$\int_{\mathbb R}|q\varphi_n - f|^p = \int_{-a}^a |q\varphi_n - f|^p \le 2^{p-1}\left (\int_{-a}^a |q\varphi_n - q|^p + \int_{-a}^a |q-f|^p\right ).$$

If $n$ is large enough, the last integral is less than $2a\epsilon^p,$ and this is enough for what we want.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ how do you say continuous compact support functions are dense in $L^p $? $\endgroup$ – user115608 Jan 7 '16 at 18:30
  • $\begingroup$ That's a standard result. $\endgroup$ – zhw. Jan 7 '16 at 18:36
  • $\begingroup$ Should it be $2^p$ in the last inequality? $\endgroup$ – Sam Wong Feb 3 at 12:33
  • $\begingroup$ @SamWong $2^9$ will work, but $2^{p-1}$ comes from convexity: $((1/2)x+(1/2)y)^p\le (1/2)x^p+(1/2)y^p.$ $\endgroup$ – zhw. Feb 3 at 17:51
  • $\begingroup$ @zhw. makes sense. thanks. $\endgroup$ – Sam Wong Feb 3 at 18:08

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.