We now similarly construct trace operators for bounded domains of class C1.
Theorem 7.1. Suppose Ω is a bounded domain of class C1. There is a unique bounded linear operator T∂Ω:H1,p(Ω)→Lp(Ω) such that for all u∈C1(Ω), T∂Ωu=u∣∂Ω. Moreover, if φ∈C1(Ω), then for any v∈H1,p(Ω), T(v⋅φ)=Tv⋅Tφ.
Proof. Let u∈C1(Ω). As before, we note that ∂Ω⊂⋃x∈∂ΩΦx−1(B(0,21)) for a suitable collection of C1 diffeomorphisms {Φx:Nx→B(0,1)} taking x to 0 and ∂Ω∩Nx to B(0,1)∩Rn−1. By compactness, ∂Ω is contained in the finite union ⋃i=1NΦxi−1(B(0,21)) for some {xi}i=1N⊂∂Ω.
Let {ψi}i=1N be a partition of unity subordinate to this cover. Note that
Altogether, we see that ∣∣u∣∣p,∂Ω≤C(Ω,n,p)⋅∣∣u∣∣1,p,Ω so that the `restriction to the boundary mapping' C1(Ω)⊂H1,p(Ω)→Lp(∂Ω) is continuous. By Lemma 4.5, it extends uniquely to a continuous linear mapping T:H1,p(Ω)→Lp(∂Ω), which is what we sought to prove. The last statement is an immediate consequence of the definition of T.
The trace operator allows us to nicely characterise H01,p(Ω).
Theorem 7.2. Suppose Ω is a bounded domain of class C1. Then u∈H01,p(Ω) iff u∈H1,p(Ω) and T∂Ωu≡0.
Proof. (⇒) Clearly u∈H1,p(Ω). Since u=limm→∞um in W1,p(Ω) for {um}⊂C01(Ω) so that Tum≡0, we see that T∂Ωu=limm→∞T∂Ωum=0, the limit being taken in Lp(∂Ω).
(⇐) Assume the same setup as the proof of Theorem. The assertion that T∂Ωu≡0 is equivalent to the existence of {um}m=1∞⊂C1(Ω) such that umm→∞u in W1,p(Ω) and um∣∂Ωm→∞0 in Lp(∂Ω), which implies in particular that (ψi⋅um)∘Φxi−1m→∞(ψi⋅u)∘Φxi−1 in W1,p(R+n) and (ψi⋅um)∘Φxi−1∣∣∂R+nm→∞0 in Lp(∂R+n). It therefore suffices to show that if v∈H1,p(R+n) is compactly supported in R+n is such that Tv≡0, then v∈H01,p(R+n).
Let {vm}⊂C01(R+n) be such that vm→v in W1,p(R+n). We may assume that suppvm⊂K⋐R+n for all m∈N and a suitable compact set K. For x∈∂R+n and xn≥0 we compute that
Now fix a function χ∈C0∞(R) with 0≤χ≤1, χ∣[0,1]≡1 and suppχ⊂[0,2]. We claim that the function
vm′(x,xn):=v(x,xn)⋅(1−χ(mxn)),
whose support is contained in (Rn×[m2,∞[)∩suppv⋐R+n is an approximating sequence for v; mollifying this sequence would then establish the claim. Now, it is clear that
Since Tv=liml→∞Tvl in L2(∂R+n), the former parenthetical term tends to 0 as l→∞. On the other hand, we may estimate the latter term from above by 2p/p⋅∫Rn−1×[0,2/m]∣Dvl∣p, which tends to p2p⋅∫K∩Rn−1×[0,2/m]∣Dv∣p as l→∞. Altogether,
It turns out that linear functionals on H0k,p(Ω) of the form (#) are actually the most general ones.
Theorem 7.4. (H0k,p(Ω))∗=H−k,p(Ω), i.e. for any F∈(H0k,p(Ω))∗, there exist {fα}∣α∣≤k⊂Lq(Ω) such that for all u,
F(u)=∣α∣≤k∑∫Ωfα⋅Dαu
and ∣∣F∣∣=(∑∣α∣≤k∫Ω∣fα∣q)1/q.
Before proving this theorem, we recall the following fundamental result from functional analysis, which says that we can always extend continuous linear functionals in such a way as to preserve their norm.
Theorem 7.5 (Hahn-Banach). If X is a Banach space, U⊂X a linear subspace and φ∈U∗, then there exists a φ′∈X∗ such that φ′∣U≡φ and ∣∣φ′∣∣X∗=∣∣φ∣∣U∗.
We shall also make use of the normed space ((Lp(Ω))N,∣∣⋅∣∣p) with norm given by
∣∣(f1,…,fN)∣∣:=(i=1∑N∫Ω∣fi∣p)1/p.
It is straightforward to check that (Lp(Ω))N is a Banach space. We also note the following lemma.
Lemma 7.6. (Lp(Ω)N)∗≃(Lq(Ω))N, i.e. if G∈((Lp(Ω))N)∗, then there exist unique g1,…,gN∈Lq(Ω) such that for all u=(u1,…,uN)∈(Lp(Ω))N,
G(u)=i=1∑N∫Ωgiui
and ∣∣G∣∣=∣∣(g1,…,gN)∣∣q.
Proof. Let G∈(Lp(Ω)N)∗ and note that for any u∈Lp(Ω)N,
G(u)=G(u1,0,…,0)+G(0,u2,0,…,0)+⋯+G(0,…,0,uN).
Note that for each i∈{1,…,n}, Lp(Ω)∋ui↦G(…,ui,…)∈K is continuous so that by the canonical isomorphism Lp(Ω)∗≃Lq(Ω), there exists a unique gi∈Lq(Ω) such that for all ui∈Lp(Ω),
G(…,ui,…)=∫Ωgiui,
i.e. G(u)=∑i=1N∫Ωgiui. By the same computation employed in the discussion above, we see that ∣∣G∣∣≤∣∣(g1,…,gN)∣∣q. To see that equality holds, choose u such that
By the Hahn-Banach theorem, F′ may be extended to all of Lp(Ω)N in such a way that ∣∣F′∣∣(Lp(Ω)N)∗=∣∣F∣∣H0k,p(Ω)∗, and by Lemma , there exists an f=(fα)∣α∣≤k∈Lq(Ω)N such that ∣∣F′∣∣(Lp(Ω)N)∗=∣∣f∣∣qand for all v=(vα)∣α∣≤k∈Lp(Ω)N,
F′(v)=∣α∣≤k∑∫Ωfαvα;
in particular,
F(u)=F′(ι(u))=∣α∣≤k∑∫Ωfα⋅Dαu.
Remark 7.7. An inspection of the above considerations shows that we have not made essential use of the fact that we are dealing with H0k,p(Ω) rather than Wk,p(Ω); in particular, we have the same characterisation of Wk,p(Ω)∗ in terms of Lq(Ω)N.
Elliptic equations
We now turn our attention to elliptic partial differential equations.
Definition 7.8. A linear second order elliptic operator is an expression of the form
L:=−i,j=1∑naij∂i∂j+i=1∑nbi∂i+c,
where (aij)i,j=1n is a symmetric positive-definite matrix, and aij,bi,c:Ω→K. We say that L is uniformly elliptic if there exists a constant θ>0 such that for all v=(v1,…,vn)∈Rn\{0} and x∈Ω,
i,j=1∑naij(x)vivj≥θ∣v∣2.
Remark 7.9. We say that L is in divergence form if for all u∈C2(Ω), Lu may be written in the form
Lu=−i,j=1∑n∂i(aij∂ju)+i=1∑nbi∂iu+cu,
where (aij)i,j=1n has the same properties as before.
Given a second-order elliptic operator L as above and functions f:Ω→K and g:∂Ω→K, we will be interested in the question of whether there is a solution u:Ω→K to the elliptic boundary value problem
Luu=fon Ω=gon ∂Ω
and how smooth u is expected to be given appropriate smoothness conditions on f and g. We call such a boundary-value problem the Dirichlet problem associated with L.
Solutions to the Dirichlet problem associated with an elliptic operator in divergence form typically arise as minimisiers or critical points of energy functionals defined on suitable spaces of functions. The following lemma illustrates this. We assume that all functions are real-valued.
Lemma 7.10. Suppose (aij)i,j=1n is a positive-definite, symmetric matrix, aij∈C1(Ω), c,f∈C0(Ω) and g∈C0(∂Ω). Define the energy functional
E(u)=21∫Ωi,j=1∑naij∂iu∂ju+cu2−∫Ωfu.
If u∈C2(Ω) is a minimiser of E amongst all functions v∈C2(Ω) with v∣∂Ω≡g, i.e. u∣∂Ω≡g and E(u)≤E(v) for all such v, then u solves the following Dirichlet problem:
−i,j=1∑n∂i(aij∂ju)+c⋅uu∣∂Ω=fon Ω=g(**)
Proof. Let φ∈C0∞(Ω). The minimisation condition implies that for allt∈R, E(u+tφ)≤E(u), i.e. t↦E(u+tφ) attains its minimum at t=0. In particular,
which holds for all φ∈C0∞(Ω), immediately implying
Remark 7.11. Note that if aij,c∈L∞(Ω), then E(u) is also defined for u∈W1,2(Ω), as is the intermediate equation (***). We say that (***) is the weak formulation of (**) and call the minimisation posed in Lemma the variational formulation of (**).