Week 6: Poincaré inequality, difference quotients and the trace operator
Poincaré inequality
The following inequality is an important consequence of the Rellich-Kondrachov inequality. Given f∈L1(Ω), write (f)Ω:=μ(Ω)1∫Ωf.
Theorem 6.1 (Poincaré inequality). Let Ω be a connected bounded domain of class C1. For any p∈[1,∞] there exists a constant C>0 depending only on n, p and Ω such that whenever u∈W1,p(Ω),
∣∣u−(u)Ω∣∣p≤C⋅∣∣Du∣∣p.
Proof. Suppose this is not the case so that for all k∈N, there exists a uk∈W1,p(Ω) such that
∣∣uk−(uk)Ω∣∣p>k⋅∣∣Duk∣∣p.
Define vk:=∣∣uk−(uk)Ω∣∣puk−(uk)Ω for each k∈N. Clearly ∣∣vk∣∣p=1, (vk)Ω=0, and
∣∣Dvk∣∣Ω=∣∣uk−(uk)Ω∣∣p1⋅∣∣Duk∣∣p<k1.
In particular, {vk}k=1∞ is a bounded sequence in W1,p(Ω) so that by the Rellich-Kondrachov theorem, we may pass to a subsequence convergent in Lp(Ω), which we again denote by {vk}. Set v=Lplimk→∞vk. We immediately obtain that ∣∣v∣∣p=1 and (v)Ω=0. On the other hand, if φ∈C0∞(Ω),
i.e. ∫Ω∂iφ⋅v=0 so that v∈W1,p(Ω) with Dv≡0. Since Ω is connected (exercise 3 of tutorial sheet 1), we deduce that there exists a constant C0 such that v≡C0. Since (v)Ω=0, we must have C0=0 so that v≡0, but then this contradicts the statement that ∣∣v∣∣p=1!
Remark 6.2. In the case of a ball B(x,r), we can get a clearer picture of how the constant in Poincaré's inequality depends on r using scaling techniques as follows: If Ω=B(0,1), then we know that there is a constant C(n,p)>0 such that for all u∈W1,p(B(0,1)),
∣∣u−(u)B(0,1)∣∣p,B(0,1)≤C⋅∣∣Du∣∣p,B(0,1).(*)
Now suppose v∈W1,p(B(x,r)) and define u:B(0,1)→K such that
z↦v(x+rz).
It is not hard to see that u∈W1,p(B(0,1)), and its weak derivative is given by Du(z)=r⋅Dv(x+rz) for almost all z∈B(0,1). Now, the Poincaré inequality (*) holds for u, and it is easily seen by changing variables in the various integrals that (u)B(0,1)=(v)B(x,r), ∣∣u−(u)B(0,1)∣∣p,B(0,1)=r−n/p∣∣v−(v)B(x,r)∣∣p,B(x,r) and ∣∣Du∣∣p,B(0,1)=r1−pn∣∣Dv∣∣p,B(x,r) so that (*) reduces to
∣∣v−(v)B(x,r)∣∣p,B(x,r)≤Cr∣∣Dv∣∣p,B(x,r),
i.e. the constant in Poincaré's inequality linearly depends on r.
Difference quotients
Suppose u∈C1(Ω) and recall that for each x∈Ω and i∈{1,…,n},
∂iu(x)=h→0limhu(x+hei)−u(x),
and for any U⋐Ω, the right-hand limit exists uniformly in x∈U. More generally, given u∈Lloc1(Ω), i∈{1,…,n} and h∈R, we define the difference quotientδihu:Ω→K by
δihu(x):=hu(x+hei)−u(x).
It turns out that if u∈H1,p(Ω), then δihuh→0∂iu in Llocp(Ω). This is a consequence of the following lemma.
Lemma 6.3. Suppose u∈Hk,p(Ω) and U⋐Ω. The inequality
∣∣δihu∣∣k−1,p,U≤∣∣u∣∣k,p,Ω
holds whenever 0<∣h∣<dist(U,∂Ω), and limh→0∣∣δihu−∂iu∣∣k−1,p,U=0, i.e. δihuh→0∂iu in Wlock−1,p(Ω).
Proof. First suppose u∈Ck,p(Ω) and note that for any α, Dαδihu≡δihDαu on U. Moreover, whenever v∈C1,p(Ω), we compute that
Now suppose u∈Hk,p(Ω) and {um}⊂Ck,p(Ω) is a sequence having u as its Wk,p-limit. Applying (#) to {um} and taking the limit m→∞ implies the first claim. To establish the second claim, we use the triangle inequality to deduce that
Hence, h→0limsup∣∣δihu−∂iu∣∣k−1,p,U≤ε for all ε>0, i.e. h→0limsup∣∣δihu−∂iu∣∣k−1,p,U=0, establishing the claim.
Note that a direct consequence of this lemma is the following: If u∈Hk,p(Ω), then for any U⋐Ω, ∣∣δihu∣∣k−1,p,U is bounded from above, and this bound is independent of h and U. We shall show that the converse is true in the case where p∈]1,∞[. First, we recall an important theorem from functional analysis and one of its consequences.
Theorem 6.4 (Banach-Alaoǧlu). Suppose X is a reflexive Banach space, i.e. the mapping
ι:Xx↦→(X∗)∗(f↦f(x))
is an isomorphism. If {xn}⊂X is a bounded sequence with ∣∣xn∣∣≤C, then it has a weakly convergent subsequence {xnk}k=1∞. Moreover, the weak limit x of {xnk} satisfies the inequality ∣∣x∣∣≤C.
Examples of reflexive Banach spaces include all finite-dimensional spaces and Lp(Ω) for p∈]1,∞[. We now apply this theorem to deduce some information on bounded sequences in Sobolev spaces.
Lemma 6.5. Suppose p∈]1,∞[ and {um}⊂Wk,p(Ω) is a bounded sequence (i.e. ∣∣um∣∣k,p≤C for some C>0 independent of m) such that umm→∞u in Lp(Ω). Then u∈Wk,p(Ω) and for all ∣α∣≤k, Dαumm→∞Dαu in Lp(Ω).
Proof. Fix α with ∣α∣≤k. The boundedness of {um} in Wk,p(Ω) implies that for each multiïndex α with ∣α∣≤k, Dαum is bounded in Lp(Ω). By applying Banach-Alaoǧlu, we may pass to a subsequence {umk} such that Dαumkk→∞vα in Lp(Ω). Now, since for any φ∈C0∞(Ω)
∫Ωumk⋅Dαφ=(−1)∣α∣∫ΩDαumk⋅φ,
and the left-hand side (resp. right-hand side) is a bounded linear function of umk (resp. Dαumk), we may take limits to deduce that
∫Ωu⋅Dαφ=(−1)∣α∣∫Ωvα⋅φ,
i.e. Dαu exists and is equal to vα∈Lp(Ω). Since α was arbitrary, we see that u∈Hk,p(Ω). To see that Dαumm→∞Dαu in Lp(Ω), we first note that for φ∈C0∞(Ω),
For fixed ε>0, we may choose φ such that ∣∣v−φ∣∣≤2Cε. Taking the limsup of both sides as m→∞ then ε↘0 implies Dαumm→∞Dαu in Lp(Ω).
Corollary 6.6. Let p∈]1,∞[, k∈N and u∈Wk−1,p(Ω). If there is a constant C>0 such that for all U⋐Ω, i∈{1,…,n}, h with 0<∣h∣<dist(U,∂Ω)
∣∣δihu∣∣k−1,p,U≤C,
then u∈Wk,p(Ω).
Proof. Let U⋐Ω. By the Banach-Alaoǧlu theorem, for each i∈{1,…,n} there exists a sequence hm→0 such that δihmum→∞vi in Lp(U), and by the preceding lemma vi∈Wk−1,p(U) with Dαδihm⇁Dαvi in Lp(U) so that, in particular, for φ∈C0∞(U) and ∣α∣≤k−1,
where in the last step follows from a change of variables. Now, since φ is smooth and u locally summable, we may interchange limit and integral to deduce that
∫RnDαvi⋅φ=(−1)∣α∣+1∫Rnu⋅∂iDαφ,
i.e. Dα+eiu exists weakly on U and is equal to Dαvi∈Lp(U). This implies in particular that u∈Wk,p(U). Finally, let Ωδ={x∈Ω:dist(x,∂Ω)>δ}∩B(0,δ−1) and ∣α∣=k−1. The bound
∫Ωδ∣δihDαu∣p≤Cp
implies that ∫Ωδ∣∂iDαu∣p≤Cp, i.e. for all α with ∣α∣=k, ∫Ωδ∣Dαu∣p≤Cp. Taking the limit δ↘0 implies that ∣∣Dαu∣∣p,Ω≤C so that u∈Wk,p(Ω).
Trace operator
We now turn our attention to the question of `boundary values' of a function u∈W1,p(Ω) in the case where Ω is a bounded domain of class C1. Since the boundary ∂Ω is a set of measure zero, the quantity u∣∂Ω is a priori not well defined. We can however make sense of this function using the approximation techniques we have been using.
The following lemma is our starting point and tells us how the W1,p norm of a compactly supported C1 function on R+n controls its Lp norm on ∂R+n.
Lemma 6.7. Let u∈C01(R+n). The following inequality holds:
∣∣u∣∣p,∂R+n≤p⋅∣∣u∣∣1,p,R+n.
Proof. We compute, applying the divergence theorem and Young's inequality:
Remark 6.8. Note that this lemma says that the mapping C01(R+n)⊂H1,p(R+n)→Lp(∂R+n) given by u↦u∣∂Ω is continuous. Using the same techniques as Theorem 2.11, it may be shown that H1,p(R+n)=C01(R+n) so that by Lemma 4.5, it extends uniquely to a continuous mapping H1,p(R+n)→Lp(∂R+n), which we call the trace operator. In particular, if u∈H01,p(Rn), then u=limm→∞um in W1,p(Rn) for a sequence {um}m=1∞⊂C0∞(R+n), and by Lemma 4.5,