Almost everywhere convergent subsequence in a Sobolev space
$begingroup$
Let $1 < p < n$ and let $p^ast$ denote the Sobolev conjugate of $p$, i.e. let
$$
p^ast := frac{np}{n-p}.
$$
Denote by $mathcal{D}^{1,p}(mathbb{R}^n)$ the space
$$
mathcal{D}^{1,p}left(mathbb{R}^nright) := left{ u in L^{p^ast}(mathbb{R}^n) : nabla u in L^p(mathbb{R}^n) right}
$$
equipped with the norm
$$
leftVert u rightVert := leftVert nabla u rightVert_{L^p(mathbb{R}^n)}.
$$
Assume now that $v_m rightharpoonup v$ weakly in $mathcal{D}^{1,p}(mathbb{R}^n)$ and that $(v_m)$ is uniformly bounded in $L^{p^ast}(mathbb{R}^n)$. After passing to a subsequence, can we assume that $v_m to v$ almost everywhere point wise?
Naturally my first idea was to try and use results about compact embeddings, but I am unaware of any such result involving this space $mathcal{D}^{1,p}(mathbb{R}^n)$.
Edit: Note that in this case we are working in the space $mathcal{D}^{1,p}(mathbb{R}^n)$ which is not the same as $W^{1,p}(mathbb{R}^n)$. In particular, the embedding theorems for $W^{1,p}(mathbb{R}^n)$ do not seem to apply here.
real-analysis functional-analysis pde sobolev-spaces
$endgroup$
add a comment |
$begingroup$
Let $1 < p < n$ and let $p^ast$ denote the Sobolev conjugate of $p$, i.e. let
$$
p^ast := frac{np}{n-p}.
$$
Denote by $mathcal{D}^{1,p}(mathbb{R}^n)$ the space
$$
mathcal{D}^{1,p}left(mathbb{R}^nright) := left{ u in L^{p^ast}(mathbb{R}^n) : nabla u in L^p(mathbb{R}^n) right}
$$
equipped with the norm
$$
leftVert u rightVert := leftVert nabla u rightVert_{L^p(mathbb{R}^n)}.
$$
Assume now that $v_m rightharpoonup v$ weakly in $mathcal{D}^{1,p}(mathbb{R}^n)$ and that $(v_m)$ is uniformly bounded in $L^{p^ast}(mathbb{R}^n)$. After passing to a subsequence, can we assume that $v_m to v$ almost everywhere point wise?
Naturally my first idea was to try and use results about compact embeddings, but I am unaware of any such result involving this space $mathcal{D}^{1,p}(mathbb{R}^n)$.
Edit: Note that in this case we are working in the space $mathcal{D}^{1,p}(mathbb{R}^n)$ which is not the same as $W^{1,p}(mathbb{R}^n)$. In particular, the embedding theorems for $W^{1,p}(mathbb{R}^n)$ do not seem to apply here.
real-analysis functional-analysis pde sobolev-spaces
$endgroup$
$begingroup$
In this post they seem to be using the compactness of the embedding $W^{1,2}(Omega) hookrightarrow L^p(Omega)$ for sufficiently nice $Omega$. However, I am not aware of any such embedding for the space $mathcal{D}^{1,p}(Omega)$. I do not see how this helps.
$endgroup$
– rolandcyp
Dec 31 '18 at 0:27
$begingroup$
Could you explain what you mean by $W^{1,p}$? I've seen it used to denote what you call $mathcal{D}^{1,p}$, and seen $mathcal{D}^{1,p}$ used to denote the closure of $C_c^{infty}$ in that space.
$endgroup$
– Sambo
Dec 31 '18 at 4:41
$begingroup$
In my case $W^{1,p}(Omega)$ denotes the space of functions in $L^p(Omega)$ having weak derivatives of the first order in $L^p(Omega)$.
$endgroup$
– rolandcyp
Dec 31 '18 at 6:08
$begingroup$
I see. I expect similar embeddings still hold. While I cannot point you directly to a reference, I do know of this paper which gives an embedding of a similar space (you'll see my confusion with the notation!)
$endgroup$
– Sambo
Dec 31 '18 at 7:00
add a comment |
$begingroup$
Let $1 < p < n$ and let $p^ast$ denote the Sobolev conjugate of $p$, i.e. let
$$
p^ast := frac{np}{n-p}.
$$
Denote by $mathcal{D}^{1,p}(mathbb{R}^n)$ the space
$$
mathcal{D}^{1,p}left(mathbb{R}^nright) := left{ u in L^{p^ast}(mathbb{R}^n) : nabla u in L^p(mathbb{R}^n) right}
$$
equipped with the norm
$$
leftVert u rightVert := leftVert nabla u rightVert_{L^p(mathbb{R}^n)}.
$$
Assume now that $v_m rightharpoonup v$ weakly in $mathcal{D}^{1,p}(mathbb{R}^n)$ and that $(v_m)$ is uniformly bounded in $L^{p^ast}(mathbb{R}^n)$. After passing to a subsequence, can we assume that $v_m to v$ almost everywhere point wise?
Naturally my first idea was to try and use results about compact embeddings, but I am unaware of any such result involving this space $mathcal{D}^{1,p}(mathbb{R}^n)$.
Edit: Note that in this case we are working in the space $mathcal{D}^{1,p}(mathbb{R}^n)$ which is not the same as $W^{1,p}(mathbb{R}^n)$. In particular, the embedding theorems for $W^{1,p}(mathbb{R}^n)$ do not seem to apply here.
real-analysis functional-analysis pde sobolev-spaces
$endgroup$
Let $1 < p < n$ and let $p^ast$ denote the Sobolev conjugate of $p$, i.e. let
$$
p^ast := frac{np}{n-p}.
$$
Denote by $mathcal{D}^{1,p}(mathbb{R}^n)$ the space
$$
mathcal{D}^{1,p}left(mathbb{R}^nright) := left{ u in L^{p^ast}(mathbb{R}^n) : nabla u in L^p(mathbb{R}^n) right}
$$
equipped with the norm
$$
leftVert u rightVert := leftVert nabla u rightVert_{L^p(mathbb{R}^n)}.
$$
Assume now that $v_m rightharpoonup v$ weakly in $mathcal{D}^{1,p}(mathbb{R}^n)$ and that $(v_m)$ is uniformly bounded in $L^{p^ast}(mathbb{R}^n)$. After passing to a subsequence, can we assume that $v_m to v$ almost everywhere point wise?
Naturally my first idea was to try and use results about compact embeddings, but I am unaware of any such result involving this space $mathcal{D}^{1,p}(mathbb{R}^n)$.
Edit: Note that in this case we are working in the space $mathcal{D}^{1,p}(mathbb{R}^n)$ which is not the same as $W^{1,p}(mathbb{R}^n)$. In particular, the embedding theorems for $W^{1,p}(mathbb{R}^n)$ do not seem to apply here.
real-analysis functional-analysis pde sobolev-spaces
real-analysis functional-analysis pde sobolev-spaces
edited Dec 31 '18 at 0:24
rolandcyp
asked Dec 30 '18 at 21:30
rolandcyprolandcyp
2,149422
2,149422
$begingroup$
In this post they seem to be using the compactness of the embedding $W^{1,2}(Omega) hookrightarrow L^p(Omega)$ for sufficiently nice $Omega$. However, I am not aware of any such embedding for the space $mathcal{D}^{1,p}(Omega)$. I do not see how this helps.
$endgroup$
– rolandcyp
Dec 31 '18 at 0:27
$begingroup$
Could you explain what you mean by $W^{1,p}$? I've seen it used to denote what you call $mathcal{D}^{1,p}$, and seen $mathcal{D}^{1,p}$ used to denote the closure of $C_c^{infty}$ in that space.
$endgroup$
– Sambo
Dec 31 '18 at 4:41
$begingroup$
In my case $W^{1,p}(Omega)$ denotes the space of functions in $L^p(Omega)$ having weak derivatives of the first order in $L^p(Omega)$.
$endgroup$
– rolandcyp
Dec 31 '18 at 6:08
$begingroup$
I see. I expect similar embeddings still hold. While I cannot point you directly to a reference, I do know of this paper which gives an embedding of a similar space (you'll see my confusion with the notation!)
$endgroup$
– Sambo
Dec 31 '18 at 7:00
add a comment |
$begingroup$
In this post they seem to be using the compactness of the embedding $W^{1,2}(Omega) hookrightarrow L^p(Omega)$ for sufficiently nice $Omega$. However, I am not aware of any such embedding for the space $mathcal{D}^{1,p}(Omega)$. I do not see how this helps.
$endgroup$
– rolandcyp
Dec 31 '18 at 0:27
$begingroup$
Could you explain what you mean by $W^{1,p}$? I've seen it used to denote what you call $mathcal{D}^{1,p}$, and seen $mathcal{D}^{1,p}$ used to denote the closure of $C_c^{infty}$ in that space.
$endgroup$
– Sambo
Dec 31 '18 at 4:41
$begingroup$
In my case $W^{1,p}(Omega)$ denotes the space of functions in $L^p(Omega)$ having weak derivatives of the first order in $L^p(Omega)$.
$endgroup$
– rolandcyp
Dec 31 '18 at 6:08
$begingroup$
I see. I expect similar embeddings still hold. While I cannot point you directly to a reference, I do know of this paper which gives an embedding of a similar space (you'll see my confusion with the notation!)
$endgroup$
– Sambo
Dec 31 '18 at 7:00
$begingroup$
In this post they seem to be using the compactness of the embedding $W^{1,2}(Omega) hookrightarrow L^p(Omega)$ for sufficiently nice $Omega$. However, I am not aware of any such embedding for the space $mathcal{D}^{1,p}(Omega)$. I do not see how this helps.
$endgroup$
– rolandcyp
Dec 31 '18 at 0:27
$begingroup$
In this post they seem to be using the compactness of the embedding $W^{1,2}(Omega) hookrightarrow L^p(Omega)$ for sufficiently nice $Omega$. However, I am not aware of any such embedding for the space $mathcal{D}^{1,p}(Omega)$. I do not see how this helps.
$endgroup$
– rolandcyp
Dec 31 '18 at 0:27
$begingroup$
Could you explain what you mean by $W^{1,p}$? I've seen it used to denote what you call $mathcal{D}^{1,p}$, and seen $mathcal{D}^{1,p}$ used to denote the closure of $C_c^{infty}$ in that space.
$endgroup$
– Sambo
Dec 31 '18 at 4:41
$begingroup$
Could you explain what you mean by $W^{1,p}$? I've seen it used to denote what you call $mathcal{D}^{1,p}$, and seen $mathcal{D}^{1,p}$ used to denote the closure of $C_c^{infty}$ in that space.
$endgroup$
– Sambo
Dec 31 '18 at 4:41
$begingroup$
In my case $W^{1,p}(Omega)$ denotes the space of functions in $L^p(Omega)$ having weak derivatives of the first order in $L^p(Omega)$.
$endgroup$
– rolandcyp
Dec 31 '18 at 6:08
$begingroup$
In my case $W^{1,p}(Omega)$ denotes the space of functions in $L^p(Omega)$ having weak derivatives of the first order in $L^p(Omega)$.
$endgroup$
– rolandcyp
Dec 31 '18 at 6:08
$begingroup$
I see. I expect similar embeddings still hold. While I cannot point you directly to a reference, I do know of this paper which gives an embedding of a similar space (you'll see my confusion with the notation!)
$endgroup$
– Sambo
Dec 31 '18 at 7:00
$begingroup$
I see. I expect similar embeddings still hold. While I cannot point you directly to a reference, I do know of this paper which gives an embedding of a similar space (you'll see my confusion with the notation!)
$endgroup$
– Sambo
Dec 31 '18 at 7:00
add a comment |
1 Answer
1
active
oldest
votes
$begingroup$
First of all we need to understand what weak convergence in this space means. Observe that we have an isometric embedding $mathcal{D}^{1,p}(mathbb R^n) hookrightarrow L^p(mathbb R^n,mathbb R^n)$ given by $u mapsto nabla u.$ By standard results about dual spaces of subspaces, we get that
$$ mathcal{D}^{1,p}(mathbb R^n)' cong L^{p'}(mathbb R^n,mathbb R^n) / Y, $$
where
$$ Y = left{g in L^{p'}(mathbb R^n,mathbb R^n) middle| int_{mathbb R^n} g.nabla u = 0 , forall u in D^{1,p}(mathbb R^n) right}.$$
(Note I do not claim that $mathcal{D}^{1,p}(mathbb R^n)$ is complete in this argument $(dagger)$.) From this it is easy to see that $v_m rightharpoonup v$ weakly in $D^{1,p}(mathbb R^n)$ if and only if for all $g in L^{p'}(mathbb R^n, mathbb R^n),$
$$ int_{mathbb R^n} g.nabla v_m rightarrow int_{mathbb R^n} g.nabla v. $$
Now the idea is that a.e. convergence is a local property, so it suffices to restrict to a bounded domain where we get compact embeddings into $L^p.$ For this fix $Omega subset mathbb R^n$ a bounded domain, then extending by zero we get for any $g in L^{p'}(Omega,mathbb R^n),$
$$ int_{Omega} g.nabla v_m rightarrow int_{Omega} g.nabla v. $$
Now the restriction $(v_n|_{Omega})$ defines a bounded sequence in $W^{1,p}(Omega),$ where we have used Hölder and the uniform bound of $(v_n)$ in $L^{p^*}(mathbb R^n).$ We know by the Rellich-Kondrachov compactness theorem that there exists $u in W^{1,p}(Omega)$ and a subsequence $(v_{n_k}|_{Omega})$ such that $v_{n_k} rightharpoonup u$ weakly in $W^{1,p}(Omega)$ and $v_{n_k} rightarrow u$ a.e. also. Then for any $g in L^{p'}(Omega,mathbb R^n)$ we have,
$$ 0 = lim_{mrightarrow 0} int_{mathbb R^n} g.nabla v_{n_k} - g.nabla v = int_{mathbb R^n} g.nabla(u-v). $$
Hence $nabla(u-v) = 0$ a.e. in $Omega,$ and so there is $lambda_{Omega} in mathbb R$ such that $v_{n_k} rightarrow (v + lambda_{Omega})$ a.e. in $Omega.$
Now take a compact exhaustion $mathbb R^n = bigcup_j Omega_j$ and iteratively choose subsequences $v_{k,j}$ such that $v_{k,j} rightarrow v + lambda_i$ a.e. as $k rightarrow infty$ in $Omega_i$ for all $i leq j.$ Then observe by Fatou's lemma that,
$$ k_j^{p^*} |Omega_j| leq liminf_{k rightarrow infty} int_{Omega_j} |v_{k,j} - v|^{p^*} leq sup_{m} int_{mathbb R^n} 2^{p^*}(|v_m|^{p^*} + |v|^{p^*}) < infty, $$
so $k_j rightarrow 0$ as $j rightarrow infty.$ Hence the diagonal sequence $v_{n_k} = v_{k,k}$ satisfies $v_{n_k} rightarrow v$ a.e. in $mathbb R^n.$
Now as the limit is unique, we see that every subsequence of $(v_m)$ has a further subsequence which converges a.e. to $v.$ Hence the entire sequence converges a.e. to $v.$
Added later $(dagger)$: Writing $X = mathcal{D}^{1,p}(mathbb R^n)$ and $lVert cdot rVert = lVert nabla cdot rVert_{L^p(mathbb R^n,mathbb R^n)},$ it is not clear whether $(X,lVertcdotrVert)$ is complete (which is equivalent to its image being closed in $L^p(mathbb R^n,mathbb R^n)$). However I claim this is not necessary for the argument to work.
Let $overline X$ be the completion of $X$ with respect to this norm, which can be identified with the closure of $X$ in $L^p(mathbb R^n,mathbb R^n)$ identified via this isometric embedding. Then the restriction map $overline X' ni f mapsto f|_X$ defines an isometric isomorphism $X' cong overline X'.$ Also we have,
begin{align*}
Y &= left{ g in L^{p'}(mathbb R^n,mathbb R^n) middle| int_{mathbb R^n} g.nabla u = 0, forall u in X right} \
&= left{ g in L^{p'}(mathbb R^n,mathbb R^n) middle| int_{mathbb R^n} g.nabla u = 0, forall u in overline X right}
end{align*}
Identifying $L^{p'}(mathbb R^n,mathbb R^n) cong L^p(mathbb R^n,mathbb R^n)',$ we have $Y = X^{perp} = overline X^{perp}.$ Now the result about duality of subspaces is generally is stated for closed subspaces (see e.g. proposition 3.67 in section 3.16 of Introduction to Banach Spaces and Algebras by Allan & Dales), but in light of above we have,
$$ X' cong overline X' cong L^p(mathbb R^n,mathbb R^n)' / overline{X}^{perp} cong L^{p'}(mathbb R^n,mathbb R^n) / Y, $$
so it also holds even if $X$ is not complete.
$endgroup$
1
$begingroup$
In the beginning, when you discuss the dual space of the subspace $D^{1,p}(mathbb{R}^n)$, don't we first need to show that $D^{1,p}(mathbb{R}^n)$ is closed as a subspace of $L^p(mathbb{R}^n;mathbb{R}^n)$?
$endgroup$
– Quoka
Dec 31 '18 at 18:48
$begingroup$
@Quoka I don't think closedness is needed for the argument to work. See my edit, where I've added more details showing what to do if it isn't.
$endgroup$
– ktoi
Jan 1 at 11:07
add a comment |
Your Answer
StackExchange.ready(function() {
var channelOptions = {
tags: "".split(" "),
id: "69"
};
initTagRenderer("".split(" "), "".split(" "), channelOptions);
StackExchange.using("externalEditor", function() {
// Have to fire editor after snippets, if snippets enabled
if (StackExchange.settings.snippets.snippetsEnabled) {
StackExchange.using("snippets", function() {
createEditor();
});
}
else {
createEditor();
}
});
function createEditor() {
StackExchange.prepareEditor({
heartbeatType: 'answer',
autoActivateHeartbeat: false,
convertImagesToLinks: true,
noModals: true,
showLowRepImageUploadWarning: true,
reputationToPostImages: 10,
bindNavPrevention: true,
postfix: "",
imageUploader: {
brandingHtml: "Powered by u003ca class="icon-imgur-white" href="https://imgur.com/"u003eu003c/au003e",
contentPolicyHtml: "User contributions licensed under u003ca href="https://creativecommons.org/licenses/by-sa/3.0/"u003ecc by-sa 3.0 with attribution requiredu003c/au003e u003ca href="https://stackoverflow.com/legal/content-policy"u003e(content policy)u003c/au003e",
allowUrls: true
},
noCode: true, onDemand: true,
discardSelector: ".discard-answer"
,immediatelyShowMarkdownHelp:true
});
}
});
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
StackExchange.ready(
function () {
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fmath.stackexchange.com%2fquestions%2f3057220%2falmost-everywhere-convergent-subsequence-in-a-sobolev-space%23new-answer', 'question_page');
}
);
Post as a guest
Required, but never shown
1 Answer
1
active
oldest
votes
1 Answer
1
active
oldest
votes
active
oldest
votes
active
oldest
votes
$begingroup$
First of all we need to understand what weak convergence in this space means. Observe that we have an isometric embedding $mathcal{D}^{1,p}(mathbb R^n) hookrightarrow L^p(mathbb R^n,mathbb R^n)$ given by $u mapsto nabla u.$ By standard results about dual spaces of subspaces, we get that
$$ mathcal{D}^{1,p}(mathbb R^n)' cong L^{p'}(mathbb R^n,mathbb R^n) / Y, $$
where
$$ Y = left{g in L^{p'}(mathbb R^n,mathbb R^n) middle| int_{mathbb R^n} g.nabla u = 0 , forall u in D^{1,p}(mathbb R^n) right}.$$
(Note I do not claim that $mathcal{D}^{1,p}(mathbb R^n)$ is complete in this argument $(dagger)$.) From this it is easy to see that $v_m rightharpoonup v$ weakly in $D^{1,p}(mathbb R^n)$ if and only if for all $g in L^{p'}(mathbb R^n, mathbb R^n),$
$$ int_{mathbb R^n} g.nabla v_m rightarrow int_{mathbb R^n} g.nabla v. $$
Now the idea is that a.e. convergence is a local property, so it suffices to restrict to a bounded domain where we get compact embeddings into $L^p.$ For this fix $Omega subset mathbb R^n$ a bounded domain, then extending by zero we get for any $g in L^{p'}(Omega,mathbb R^n),$
$$ int_{Omega} g.nabla v_m rightarrow int_{Omega} g.nabla v. $$
Now the restriction $(v_n|_{Omega})$ defines a bounded sequence in $W^{1,p}(Omega),$ where we have used Hölder and the uniform bound of $(v_n)$ in $L^{p^*}(mathbb R^n).$ We know by the Rellich-Kondrachov compactness theorem that there exists $u in W^{1,p}(Omega)$ and a subsequence $(v_{n_k}|_{Omega})$ such that $v_{n_k} rightharpoonup u$ weakly in $W^{1,p}(Omega)$ and $v_{n_k} rightarrow u$ a.e. also. Then for any $g in L^{p'}(Omega,mathbb R^n)$ we have,
$$ 0 = lim_{mrightarrow 0} int_{mathbb R^n} g.nabla v_{n_k} - g.nabla v = int_{mathbb R^n} g.nabla(u-v). $$
Hence $nabla(u-v) = 0$ a.e. in $Omega,$ and so there is $lambda_{Omega} in mathbb R$ such that $v_{n_k} rightarrow (v + lambda_{Omega})$ a.e. in $Omega.$
Now take a compact exhaustion $mathbb R^n = bigcup_j Omega_j$ and iteratively choose subsequences $v_{k,j}$ such that $v_{k,j} rightarrow v + lambda_i$ a.e. as $k rightarrow infty$ in $Omega_i$ for all $i leq j.$ Then observe by Fatou's lemma that,
$$ k_j^{p^*} |Omega_j| leq liminf_{k rightarrow infty} int_{Omega_j} |v_{k,j} - v|^{p^*} leq sup_{m} int_{mathbb R^n} 2^{p^*}(|v_m|^{p^*} + |v|^{p^*}) < infty, $$
so $k_j rightarrow 0$ as $j rightarrow infty.$ Hence the diagonal sequence $v_{n_k} = v_{k,k}$ satisfies $v_{n_k} rightarrow v$ a.e. in $mathbb R^n.$
Now as the limit is unique, we see that every subsequence of $(v_m)$ has a further subsequence which converges a.e. to $v.$ Hence the entire sequence converges a.e. to $v.$
Added later $(dagger)$: Writing $X = mathcal{D}^{1,p}(mathbb R^n)$ and $lVert cdot rVert = lVert nabla cdot rVert_{L^p(mathbb R^n,mathbb R^n)},$ it is not clear whether $(X,lVertcdotrVert)$ is complete (which is equivalent to its image being closed in $L^p(mathbb R^n,mathbb R^n)$). However I claim this is not necessary for the argument to work.
Let $overline X$ be the completion of $X$ with respect to this norm, which can be identified with the closure of $X$ in $L^p(mathbb R^n,mathbb R^n)$ identified via this isometric embedding. Then the restriction map $overline X' ni f mapsto f|_X$ defines an isometric isomorphism $X' cong overline X'.$ Also we have,
begin{align*}
Y &= left{ g in L^{p'}(mathbb R^n,mathbb R^n) middle| int_{mathbb R^n} g.nabla u = 0, forall u in X right} \
&= left{ g in L^{p'}(mathbb R^n,mathbb R^n) middle| int_{mathbb R^n} g.nabla u = 0, forall u in overline X right}
end{align*}
Identifying $L^{p'}(mathbb R^n,mathbb R^n) cong L^p(mathbb R^n,mathbb R^n)',$ we have $Y = X^{perp} = overline X^{perp}.$ Now the result about duality of subspaces is generally is stated for closed subspaces (see e.g. proposition 3.67 in section 3.16 of Introduction to Banach Spaces and Algebras by Allan & Dales), but in light of above we have,
$$ X' cong overline X' cong L^p(mathbb R^n,mathbb R^n)' / overline{X}^{perp} cong L^{p'}(mathbb R^n,mathbb R^n) / Y, $$
so it also holds even if $X$ is not complete.
$endgroup$
1
$begingroup$
In the beginning, when you discuss the dual space of the subspace $D^{1,p}(mathbb{R}^n)$, don't we first need to show that $D^{1,p}(mathbb{R}^n)$ is closed as a subspace of $L^p(mathbb{R}^n;mathbb{R}^n)$?
$endgroup$
– Quoka
Dec 31 '18 at 18:48
$begingroup$
@Quoka I don't think closedness is needed for the argument to work. See my edit, where I've added more details showing what to do if it isn't.
$endgroup$
– ktoi
Jan 1 at 11:07
add a comment |
$begingroup$
First of all we need to understand what weak convergence in this space means. Observe that we have an isometric embedding $mathcal{D}^{1,p}(mathbb R^n) hookrightarrow L^p(mathbb R^n,mathbb R^n)$ given by $u mapsto nabla u.$ By standard results about dual spaces of subspaces, we get that
$$ mathcal{D}^{1,p}(mathbb R^n)' cong L^{p'}(mathbb R^n,mathbb R^n) / Y, $$
where
$$ Y = left{g in L^{p'}(mathbb R^n,mathbb R^n) middle| int_{mathbb R^n} g.nabla u = 0 , forall u in D^{1,p}(mathbb R^n) right}.$$
(Note I do not claim that $mathcal{D}^{1,p}(mathbb R^n)$ is complete in this argument $(dagger)$.) From this it is easy to see that $v_m rightharpoonup v$ weakly in $D^{1,p}(mathbb R^n)$ if and only if for all $g in L^{p'}(mathbb R^n, mathbb R^n),$
$$ int_{mathbb R^n} g.nabla v_m rightarrow int_{mathbb R^n} g.nabla v. $$
Now the idea is that a.e. convergence is a local property, so it suffices to restrict to a bounded domain where we get compact embeddings into $L^p.$ For this fix $Omega subset mathbb R^n$ a bounded domain, then extending by zero we get for any $g in L^{p'}(Omega,mathbb R^n),$
$$ int_{Omega} g.nabla v_m rightarrow int_{Omega} g.nabla v. $$
Now the restriction $(v_n|_{Omega})$ defines a bounded sequence in $W^{1,p}(Omega),$ where we have used Hölder and the uniform bound of $(v_n)$ in $L^{p^*}(mathbb R^n).$ We know by the Rellich-Kondrachov compactness theorem that there exists $u in W^{1,p}(Omega)$ and a subsequence $(v_{n_k}|_{Omega})$ such that $v_{n_k} rightharpoonup u$ weakly in $W^{1,p}(Omega)$ and $v_{n_k} rightarrow u$ a.e. also. Then for any $g in L^{p'}(Omega,mathbb R^n)$ we have,
$$ 0 = lim_{mrightarrow 0} int_{mathbb R^n} g.nabla v_{n_k} - g.nabla v = int_{mathbb R^n} g.nabla(u-v). $$
Hence $nabla(u-v) = 0$ a.e. in $Omega,$ and so there is $lambda_{Omega} in mathbb R$ such that $v_{n_k} rightarrow (v + lambda_{Omega})$ a.e. in $Omega.$
Now take a compact exhaustion $mathbb R^n = bigcup_j Omega_j$ and iteratively choose subsequences $v_{k,j}$ such that $v_{k,j} rightarrow v + lambda_i$ a.e. as $k rightarrow infty$ in $Omega_i$ for all $i leq j.$ Then observe by Fatou's lemma that,
$$ k_j^{p^*} |Omega_j| leq liminf_{k rightarrow infty} int_{Omega_j} |v_{k,j} - v|^{p^*} leq sup_{m} int_{mathbb R^n} 2^{p^*}(|v_m|^{p^*} + |v|^{p^*}) < infty, $$
so $k_j rightarrow 0$ as $j rightarrow infty.$ Hence the diagonal sequence $v_{n_k} = v_{k,k}$ satisfies $v_{n_k} rightarrow v$ a.e. in $mathbb R^n.$
Now as the limit is unique, we see that every subsequence of $(v_m)$ has a further subsequence which converges a.e. to $v.$ Hence the entire sequence converges a.e. to $v.$
Added later $(dagger)$: Writing $X = mathcal{D}^{1,p}(mathbb R^n)$ and $lVert cdot rVert = lVert nabla cdot rVert_{L^p(mathbb R^n,mathbb R^n)},$ it is not clear whether $(X,lVertcdotrVert)$ is complete (which is equivalent to its image being closed in $L^p(mathbb R^n,mathbb R^n)$). However I claim this is not necessary for the argument to work.
Let $overline X$ be the completion of $X$ with respect to this norm, which can be identified with the closure of $X$ in $L^p(mathbb R^n,mathbb R^n)$ identified via this isometric embedding. Then the restriction map $overline X' ni f mapsto f|_X$ defines an isometric isomorphism $X' cong overline X'.$ Also we have,
begin{align*}
Y &= left{ g in L^{p'}(mathbb R^n,mathbb R^n) middle| int_{mathbb R^n} g.nabla u = 0, forall u in X right} \
&= left{ g in L^{p'}(mathbb R^n,mathbb R^n) middle| int_{mathbb R^n} g.nabla u = 0, forall u in overline X right}
end{align*}
Identifying $L^{p'}(mathbb R^n,mathbb R^n) cong L^p(mathbb R^n,mathbb R^n)',$ we have $Y = X^{perp} = overline X^{perp}.$ Now the result about duality of subspaces is generally is stated for closed subspaces (see e.g. proposition 3.67 in section 3.16 of Introduction to Banach Spaces and Algebras by Allan & Dales), but in light of above we have,
$$ X' cong overline X' cong L^p(mathbb R^n,mathbb R^n)' / overline{X}^{perp} cong L^{p'}(mathbb R^n,mathbb R^n) / Y, $$
so it also holds even if $X$ is not complete.
$endgroup$
1
$begingroup$
In the beginning, when you discuss the dual space of the subspace $D^{1,p}(mathbb{R}^n)$, don't we first need to show that $D^{1,p}(mathbb{R}^n)$ is closed as a subspace of $L^p(mathbb{R}^n;mathbb{R}^n)$?
$endgroup$
– Quoka
Dec 31 '18 at 18:48
$begingroup$
@Quoka I don't think closedness is needed for the argument to work. See my edit, where I've added more details showing what to do if it isn't.
$endgroup$
– ktoi
Jan 1 at 11:07
add a comment |
$begingroup$
First of all we need to understand what weak convergence in this space means. Observe that we have an isometric embedding $mathcal{D}^{1,p}(mathbb R^n) hookrightarrow L^p(mathbb R^n,mathbb R^n)$ given by $u mapsto nabla u.$ By standard results about dual spaces of subspaces, we get that
$$ mathcal{D}^{1,p}(mathbb R^n)' cong L^{p'}(mathbb R^n,mathbb R^n) / Y, $$
where
$$ Y = left{g in L^{p'}(mathbb R^n,mathbb R^n) middle| int_{mathbb R^n} g.nabla u = 0 , forall u in D^{1,p}(mathbb R^n) right}.$$
(Note I do not claim that $mathcal{D}^{1,p}(mathbb R^n)$ is complete in this argument $(dagger)$.) From this it is easy to see that $v_m rightharpoonup v$ weakly in $D^{1,p}(mathbb R^n)$ if and only if for all $g in L^{p'}(mathbb R^n, mathbb R^n),$
$$ int_{mathbb R^n} g.nabla v_m rightarrow int_{mathbb R^n} g.nabla v. $$
Now the idea is that a.e. convergence is a local property, so it suffices to restrict to a bounded domain where we get compact embeddings into $L^p.$ For this fix $Omega subset mathbb R^n$ a bounded domain, then extending by zero we get for any $g in L^{p'}(Omega,mathbb R^n),$
$$ int_{Omega} g.nabla v_m rightarrow int_{Omega} g.nabla v. $$
Now the restriction $(v_n|_{Omega})$ defines a bounded sequence in $W^{1,p}(Omega),$ where we have used Hölder and the uniform bound of $(v_n)$ in $L^{p^*}(mathbb R^n).$ We know by the Rellich-Kondrachov compactness theorem that there exists $u in W^{1,p}(Omega)$ and a subsequence $(v_{n_k}|_{Omega})$ such that $v_{n_k} rightharpoonup u$ weakly in $W^{1,p}(Omega)$ and $v_{n_k} rightarrow u$ a.e. also. Then for any $g in L^{p'}(Omega,mathbb R^n)$ we have,
$$ 0 = lim_{mrightarrow 0} int_{mathbb R^n} g.nabla v_{n_k} - g.nabla v = int_{mathbb R^n} g.nabla(u-v). $$
Hence $nabla(u-v) = 0$ a.e. in $Omega,$ and so there is $lambda_{Omega} in mathbb R$ such that $v_{n_k} rightarrow (v + lambda_{Omega})$ a.e. in $Omega.$
Now take a compact exhaustion $mathbb R^n = bigcup_j Omega_j$ and iteratively choose subsequences $v_{k,j}$ such that $v_{k,j} rightarrow v + lambda_i$ a.e. as $k rightarrow infty$ in $Omega_i$ for all $i leq j.$ Then observe by Fatou's lemma that,
$$ k_j^{p^*} |Omega_j| leq liminf_{k rightarrow infty} int_{Omega_j} |v_{k,j} - v|^{p^*} leq sup_{m} int_{mathbb R^n} 2^{p^*}(|v_m|^{p^*} + |v|^{p^*}) < infty, $$
so $k_j rightarrow 0$ as $j rightarrow infty.$ Hence the diagonal sequence $v_{n_k} = v_{k,k}$ satisfies $v_{n_k} rightarrow v$ a.e. in $mathbb R^n.$
Now as the limit is unique, we see that every subsequence of $(v_m)$ has a further subsequence which converges a.e. to $v.$ Hence the entire sequence converges a.e. to $v.$
Added later $(dagger)$: Writing $X = mathcal{D}^{1,p}(mathbb R^n)$ and $lVert cdot rVert = lVert nabla cdot rVert_{L^p(mathbb R^n,mathbb R^n)},$ it is not clear whether $(X,lVertcdotrVert)$ is complete (which is equivalent to its image being closed in $L^p(mathbb R^n,mathbb R^n)$). However I claim this is not necessary for the argument to work.
Let $overline X$ be the completion of $X$ with respect to this norm, which can be identified with the closure of $X$ in $L^p(mathbb R^n,mathbb R^n)$ identified via this isometric embedding. Then the restriction map $overline X' ni f mapsto f|_X$ defines an isometric isomorphism $X' cong overline X'.$ Also we have,
begin{align*}
Y &= left{ g in L^{p'}(mathbb R^n,mathbb R^n) middle| int_{mathbb R^n} g.nabla u = 0, forall u in X right} \
&= left{ g in L^{p'}(mathbb R^n,mathbb R^n) middle| int_{mathbb R^n} g.nabla u = 0, forall u in overline X right}
end{align*}
Identifying $L^{p'}(mathbb R^n,mathbb R^n) cong L^p(mathbb R^n,mathbb R^n)',$ we have $Y = X^{perp} = overline X^{perp}.$ Now the result about duality of subspaces is generally is stated for closed subspaces (see e.g. proposition 3.67 in section 3.16 of Introduction to Banach Spaces and Algebras by Allan & Dales), but in light of above we have,
$$ X' cong overline X' cong L^p(mathbb R^n,mathbb R^n)' / overline{X}^{perp} cong L^{p'}(mathbb R^n,mathbb R^n) / Y, $$
so it also holds even if $X$ is not complete.
$endgroup$
First of all we need to understand what weak convergence in this space means. Observe that we have an isometric embedding $mathcal{D}^{1,p}(mathbb R^n) hookrightarrow L^p(mathbb R^n,mathbb R^n)$ given by $u mapsto nabla u.$ By standard results about dual spaces of subspaces, we get that
$$ mathcal{D}^{1,p}(mathbb R^n)' cong L^{p'}(mathbb R^n,mathbb R^n) / Y, $$
where
$$ Y = left{g in L^{p'}(mathbb R^n,mathbb R^n) middle| int_{mathbb R^n} g.nabla u = 0 , forall u in D^{1,p}(mathbb R^n) right}.$$
(Note I do not claim that $mathcal{D}^{1,p}(mathbb R^n)$ is complete in this argument $(dagger)$.) From this it is easy to see that $v_m rightharpoonup v$ weakly in $D^{1,p}(mathbb R^n)$ if and only if for all $g in L^{p'}(mathbb R^n, mathbb R^n),$
$$ int_{mathbb R^n} g.nabla v_m rightarrow int_{mathbb R^n} g.nabla v. $$
Now the idea is that a.e. convergence is a local property, so it suffices to restrict to a bounded domain where we get compact embeddings into $L^p.$ For this fix $Omega subset mathbb R^n$ a bounded domain, then extending by zero we get for any $g in L^{p'}(Omega,mathbb R^n),$
$$ int_{Omega} g.nabla v_m rightarrow int_{Omega} g.nabla v. $$
Now the restriction $(v_n|_{Omega})$ defines a bounded sequence in $W^{1,p}(Omega),$ where we have used Hölder and the uniform bound of $(v_n)$ in $L^{p^*}(mathbb R^n).$ We know by the Rellich-Kondrachov compactness theorem that there exists $u in W^{1,p}(Omega)$ and a subsequence $(v_{n_k}|_{Omega})$ such that $v_{n_k} rightharpoonup u$ weakly in $W^{1,p}(Omega)$ and $v_{n_k} rightarrow u$ a.e. also. Then for any $g in L^{p'}(Omega,mathbb R^n)$ we have,
$$ 0 = lim_{mrightarrow 0} int_{mathbb R^n} g.nabla v_{n_k} - g.nabla v = int_{mathbb R^n} g.nabla(u-v). $$
Hence $nabla(u-v) = 0$ a.e. in $Omega,$ and so there is $lambda_{Omega} in mathbb R$ such that $v_{n_k} rightarrow (v + lambda_{Omega})$ a.e. in $Omega.$
Now take a compact exhaustion $mathbb R^n = bigcup_j Omega_j$ and iteratively choose subsequences $v_{k,j}$ such that $v_{k,j} rightarrow v + lambda_i$ a.e. as $k rightarrow infty$ in $Omega_i$ for all $i leq j.$ Then observe by Fatou's lemma that,
$$ k_j^{p^*} |Omega_j| leq liminf_{k rightarrow infty} int_{Omega_j} |v_{k,j} - v|^{p^*} leq sup_{m} int_{mathbb R^n} 2^{p^*}(|v_m|^{p^*} + |v|^{p^*}) < infty, $$
so $k_j rightarrow 0$ as $j rightarrow infty.$ Hence the diagonal sequence $v_{n_k} = v_{k,k}$ satisfies $v_{n_k} rightarrow v$ a.e. in $mathbb R^n.$
Now as the limit is unique, we see that every subsequence of $(v_m)$ has a further subsequence which converges a.e. to $v.$ Hence the entire sequence converges a.e. to $v.$
Added later $(dagger)$: Writing $X = mathcal{D}^{1,p}(mathbb R^n)$ and $lVert cdot rVert = lVert nabla cdot rVert_{L^p(mathbb R^n,mathbb R^n)},$ it is not clear whether $(X,lVertcdotrVert)$ is complete (which is equivalent to its image being closed in $L^p(mathbb R^n,mathbb R^n)$). However I claim this is not necessary for the argument to work.
Let $overline X$ be the completion of $X$ with respect to this norm, which can be identified with the closure of $X$ in $L^p(mathbb R^n,mathbb R^n)$ identified via this isometric embedding. Then the restriction map $overline X' ni f mapsto f|_X$ defines an isometric isomorphism $X' cong overline X'.$ Also we have,
begin{align*}
Y &= left{ g in L^{p'}(mathbb R^n,mathbb R^n) middle| int_{mathbb R^n} g.nabla u = 0, forall u in X right} \
&= left{ g in L^{p'}(mathbb R^n,mathbb R^n) middle| int_{mathbb R^n} g.nabla u = 0, forall u in overline X right}
end{align*}
Identifying $L^{p'}(mathbb R^n,mathbb R^n) cong L^p(mathbb R^n,mathbb R^n)',$ we have $Y = X^{perp} = overline X^{perp}.$ Now the result about duality of subspaces is generally is stated for closed subspaces (see e.g. proposition 3.67 in section 3.16 of Introduction to Banach Spaces and Algebras by Allan & Dales), but in light of above we have,
$$ X' cong overline X' cong L^p(mathbb R^n,mathbb R^n)' / overline{X}^{perp} cong L^{p'}(mathbb R^n,mathbb R^n) / Y, $$
so it also holds even if $X$ is not complete.
edited Jan 1 at 11:04
answered Dec 31 '18 at 13:11
ktoiktoi
2,5611618
2,5611618
1
$begingroup$
In the beginning, when you discuss the dual space of the subspace $D^{1,p}(mathbb{R}^n)$, don't we first need to show that $D^{1,p}(mathbb{R}^n)$ is closed as a subspace of $L^p(mathbb{R}^n;mathbb{R}^n)$?
$endgroup$
– Quoka
Dec 31 '18 at 18:48
$begingroup$
@Quoka I don't think closedness is needed for the argument to work. See my edit, where I've added more details showing what to do if it isn't.
$endgroup$
– ktoi
Jan 1 at 11:07
add a comment |
1
$begingroup$
In the beginning, when you discuss the dual space of the subspace $D^{1,p}(mathbb{R}^n)$, don't we first need to show that $D^{1,p}(mathbb{R}^n)$ is closed as a subspace of $L^p(mathbb{R}^n;mathbb{R}^n)$?
$endgroup$
– Quoka
Dec 31 '18 at 18:48
$begingroup$
@Quoka I don't think closedness is needed for the argument to work. See my edit, where I've added more details showing what to do if it isn't.
$endgroup$
– ktoi
Jan 1 at 11:07
1
1
$begingroup$
In the beginning, when you discuss the dual space of the subspace $D^{1,p}(mathbb{R}^n)$, don't we first need to show that $D^{1,p}(mathbb{R}^n)$ is closed as a subspace of $L^p(mathbb{R}^n;mathbb{R}^n)$?
$endgroup$
– Quoka
Dec 31 '18 at 18:48
$begingroup$
In the beginning, when you discuss the dual space of the subspace $D^{1,p}(mathbb{R}^n)$, don't we first need to show that $D^{1,p}(mathbb{R}^n)$ is closed as a subspace of $L^p(mathbb{R}^n;mathbb{R}^n)$?
$endgroup$
– Quoka
Dec 31 '18 at 18:48
$begingroup$
@Quoka I don't think closedness is needed for the argument to work. See my edit, where I've added more details showing what to do if it isn't.
$endgroup$
– ktoi
Jan 1 at 11:07
$begingroup$
@Quoka I don't think closedness is needed for the argument to work. See my edit, where I've added more details showing what to do if it isn't.
$endgroup$
– ktoi
Jan 1 at 11:07
add a comment |
Thanks for contributing an answer to Mathematics Stack Exchange!
- Please be sure to answer the question. Provide details and share your research!
But avoid …
- Asking for help, clarification, or responding to other answers.
- Making statements based on opinion; back them up with references or personal experience.
Use MathJax to format equations. MathJax reference.
To learn more, see our tips on writing great answers.
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
StackExchange.ready(
function () {
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fmath.stackexchange.com%2fquestions%2f3057220%2falmost-everywhere-convergent-subsequence-in-a-sobolev-space%23new-answer', 'question_page');
}
);
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
$begingroup$
In this post they seem to be using the compactness of the embedding $W^{1,2}(Omega) hookrightarrow L^p(Omega)$ for sufficiently nice $Omega$. However, I am not aware of any such embedding for the space $mathcal{D}^{1,p}(Omega)$. I do not see how this helps.
$endgroup$
– rolandcyp
Dec 31 '18 at 0:27
$begingroup$
Could you explain what you mean by $W^{1,p}$? I've seen it used to denote what you call $mathcal{D}^{1,p}$, and seen $mathcal{D}^{1,p}$ used to denote the closure of $C_c^{infty}$ in that space.
$endgroup$
– Sambo
Dec 31 '18 at 4:41
$begingroup$
In my case $W^{1,p}(Omega)$ denotes the space of functions in $L^p(Omega)$ having weak derivatives of the first order in $L^p(Omega)$.
$endgroup$
– rolandcyp
Dec 31 '18 at 6:08
$begingroup$
I see. I expect similar embeddings still hold. While I cannot point you directly to a reference, I do know of this paper which gives an embedding of a similar space (you'll see my confusion with the notation!)
$endgroup$
– Sambo
Dec 31 '18 at 7:00