Embedding Theorems in Lp Spaces
📂Lebesgue SpacesEmbedding Theorems in Lp Spaces
Theorem
If Ω⊂Rn is an open set and let’s assume vol(Ω)=∫Ω1dx<∞.
(a) For 1≤p≤q≤∞, if u∈Lq(Ω) then, u∈Lp(Ω) and
∥u∥p≤(vol(Ω))p1−q1∥u∥q
And Lq is embedded into Lp.
Lq(Ω)→Lp(Ω)
(b) For 1≤p≤q≤∞, if u∈L∞(Ω) then,
p→∞lim∥u∥p=∥u∥∞
(c) For all 1≤p<∞, if u∈Lp(Ω) and there exists K that is ∥u∥p≤K,
u∈L∞(Ω)and∥u∥∞≤K
Explanation
(a) When 1≤p≤q, generally, there is no inclusion relation between the space of Lp and Lq. However, if the domain’s volume is finite, then Lq⊂Lp holds.
(b) Since the above holds when q=∞, if u∈L∞, for all 1≤p<∞, u∈Lp and the limit of ∥⋅∥p converges to ∥⋅∥∞.
(c) The assumption about constant K does not mean there exists one for all p respectively, but means one K satisfies ∥u∥p≤K for all p. As (a) and (b) suggest, the larger p becomes, it turns into a smaller space, which can be considered as L∞(Ω)=1≤p<∞⋂Lp(Ω).
Proof
(a)
If either p=q or q=∞ holds, it is trivial that (1) and (2) are true. So, let’s assume 1≤p<q<∞ and u∈Lq(Ω).
Lemma: Generalized Hölder’s Inequality
If three constants α>0,β>0,γ>0 satisfy α1+β1=γ1 and f∈Lα(Ω),g∈Lβ(Ω), then fg∈Lγ(Ω) and the following inequality holds.
∥fg∥γ=(∫Ω∣f(x)g(x)∣γdx)1/γ≤∥f∥α∥g∥β
Substituting α=q,β=p1−q1,γ=p, f=u, and g=1 into the lemma above gives u∈Lq(Ω),1∈Lp1−q1(Ω), therefore,
⟹(∫Ω∣u(x)⋅1∣pdx)1/p≤∥u∥p≤=∥u∥q∥1∥p1−q1(∫Ω1dx)p1−q1∥u∥q (vol(Ω))p1−q1∥u∥q
Thus, u∈Lp(Ω) holds.
Embedding
- X is a subspace of Y.
- ∃M>0 such that ∥Ix∥Y≤M∥x∥X,x∈X
Also, if we let M=(vol(Ω))p1−q1
∥u∥p≤M∥u∥q
Therefore, by the definition of embedding, we can see that Lq(Ω)→Lp(Ω) is an embedding.
■
(b)
From the result of (a), the following holds.
p→∞limsup∥u∥p≤∥u∥∞
Meanwhile, for any ϵ>0, there exists a set A⊂Ω with a positive measure μ(A) satisfying the following condition.
∣u(x)∣≥∥u∥∞−ϵ,if x∈A
If such A does not exist, ∥u∥∞ would not satisfy the definition of ∥⋅∥∞, hence the existence of A is guaranteed. Therefore, the following formula holds.
∫Ω∣u(x)∣pdx≥∫A∣u(x)∣pdx≥∫A(∥u∥∞−ϵ)pdx≥μ(A)(∥u∥∞−ϵ)p
Thus, we obtain the following.
∥u∥p≥(μ(A))1/p(∥u∥∞−ϵ)
This must hold for all ϵ and any corresponding A, thus
p→∞liminf∥u∥p≥∥u∥∞
Therefore,
p→∞liminf∥u∥p≤p→∞limsup∥u∥p≤∥u∥∞≤p→∞liminf∥u∥p≤p→∞limsup∥u∥p
⟹p→∞liminf∥u∥p=∥u∥∞=p→∞limsup∥u∥p
⟹p→∞lim∥u∥p=∥u∥∞
■
(c)
Let’s assume for all 1≤p<∞, there exists K that is ∥u∥p≤K. And suppose either u∈L∞(Ω) or ∥u∥∞≤K does not hold. Then, according to the definition of ∥⋅∥∞, we can find a constant K1 and a set A⊂Ω that satisfies the following.
K1>Kandμ(A)>0and∣u(x)∣>K1 for x∈A
Then, as shown in the proof of (b), the following holds.
p→∞liminf∥u∥p≥K1>K
This contradicts the assumption that for all p, ∥u∥p≤K holds. Therefore,
u∈L∞(Ω)and∥u∥∞≤K
■