We now turn our attention to the notion of a weak solution to a (divergence-form) elliptic PDE. We will assume for simplicity that all functions considered are real-valued.
Weak solutions to divergence-form elliptic equations via Lax-Milgram
Our goal will be to establish the existence of solutions to the following Dirichlet problem for u:Ω→R:
We will first consider the so-called weak formulation of this problem; to this end, assume the following setup throughout this lecture:
Ω⊂Rn is an open bounded set.
For each i,j∈{1,…,n}, aij∈L∞(Ω) and for all v∈Rn\{0} and almost all x∈Ω, the bounds
λ0∣v∣2≤i,j=1∑naij(x)vivj
and ∣(aij(x))i,j=1n∣≤Λ hold.
For each i∈{1,…,n}, bi∈L∞(Ω) with ∣b∣≤β.
c∈L∞(Ω) and the bound ∣c∣≤γ holds.
f∈L2(Ω).
Note that if u is twice differentiable and satisfies the equation Lu=f, then multiplying by a function φ∈C0∞(Ω) and integrating the leading order term by parts, we end up with the equation
where Ω is a bounded domain of class C1 and g∈T∂Ω(H1,2(Ω)), i.e. the equation (**) holds for all φ∈C0∞(Ω) and g=T∂Ωg′ for some g′∈H1,2(Ω). By considering u′:=u−g′ so that u′∈H01,2(Ω), we then see that
If g′ may be taken to be in C2(Ω), then the right-hand side may be written in the form of the right-hand side of (**) and we have reduced the case of general boundary values to that of zero boundary values. More generally, the right-hand side is actually an element of H0−1,2(Ω), considered as a function of φ. We will discuss this version of (*) in our investigations.
Note that if (**) holds for all φ∈C0∞(Ω), then by approximation it also holds for φ∈H01,2(Ω). Moreover, if we let
and F(φ):=∫Ωf⋅φ, then we may abbreviate (**) as finding a u∈H01,2(Ω) such that for all φ∈H01,2(Ω),
BL(u,φ)=F(φ).(**)’
The mapping B:H01,2(Ω)×H01,2(Ω)→R is called the bilinear form associated to L. As it turns out, there is a general functional-analytic principle that allows us to treat equations of the form (**)' under suitable assumptions on B.
Theorem 8.3 (Lax-Milgram). Suppose X is a Hilbert space, F∈X∗ and B:X×X→K is a bilinear form with the following properties:
B is bounded, i.e. there exists a C0>0 such that for all x,y∈X, ∣B(x,y)∣≤C0⋅∣∣x∣∣⋅∣∣y∣∣; and
B is coërcive, i.e. there exists a C1>0 such that for all x∈X, ℜB(x,x)≥C1⋅∣∣x∣∣2.
Then there exists a unique x0∈X satisfying the equation
B(x0,y)=F(y)(**)
for all y∈X.
Proof. Note that uniqueness immediately follows from coërcivity and bilinearity: If x0 and x0′ solve (**), then for all y∈X, B(x0−x0′,y)=0 so that
If {xi}i=1∞⊂X is a Cauchy sequence, then it is clearly a Cauchy sequence in X and therefore converges to some x∈X. On the other hand, the associates sequence of {χi}i=1∞⊂X satisfying the equation
so that χii→∞χ in X, which in turn implies by the boundedness of B that for all y∈X,
⟨x,y⟩=i→∞lim⟨xi,y⟩=i→∞limB(χi,y)=B(χ,y),
i.e. x∈X so that X is complete.
Now suppose x∈X, which implies that (X∋z↦⟨x,z⟩)∈(X)∗, whence we may find an x∈X such that for all z∈X,
⟨x,z⟩=⟨x,z⟩⇔⟨x−x,z⟩=0.
Since y↦B(x−x,y) is in X∗, we may find an x0∈X such that ⟨x0,y⟩=B(x−x,y) for all y∈X, i.e. x0∈X, which implies that
0=ℜ⟨x0,x−x⟩=ℜB(x−x,x−x)≥C1∣∣x−x∣∣2,
i.e. x=x so that x∈X⇒X=X. Since X is a Hilbert space, the Riesz representation theorem yields that every element F∈X∗ is of the form ⟨x,⋅⟩ for some x∈X, which establishes the claim.
We now establish some estimates on the bilinear form associated to L with the goal of applying the Lax-Milgram theorem.
Lemma 8.4 (energy estimates). Suppose BL is the bilinear form associated to L. There exist constants C0,C1>0 and C2>0 such that for all u,v∈H01,2(Ω),
∣BL(u,v)∣≤C0∣∣u∣∣1,2⋅∣∣v∣∣1,2
and
BL(u,u)+C2∣∣u∣∣22≥C1∣∣u∣∣1,22.
Proof. First note that by the triangle inequality and Hölder's inequality,
where CPoinc is the constant appearing in the Poincaré inequality for u∈H01,2(Ω). This establishes the claim.
As may be gleaned from this lemma, BL is almost coërcive, save for the constant C2 getting in the way. However, we can still use Lax-Milgram to solve a slightly modified version of our problem.
Theorem 8.5 (first existence theorem). There exists a constant C≥0 such that whenever μ≥C, there exists a unique weak solution to the Dirichlet problem
Lu+μuu∣∂Ω=fon Ω=0.
Proof. Let BL+μI be the bilinear form corresponding to the elliptic operator L+μI and BL that corresponding to L. For all u,v∈H01,2(Ω), we have that
BL+μI(u,v)=BL(u,v)+μ∫Ωu⋅v.
This bilinear form is bounded, since by Hölder's inequality and Lemma 8.4,
provided μ≥C:=C2. Since the linear mapping H01,2(Ω)∋v↦∫Ωf⋅v is continuous, Lax-Milgram implies that there exists a unique u∈H01,2(Ω) such that for all v∈H01,2(Ω),
BL+μI(u,v)=∫Ωf⋅v,
i.e. there exists a unique weak solution to the aforementioned Dirichlet problem.
Remark 8.6. The proof of the first existence theorem actually gives us a bit more: Given any F∈H0−1,2(Ω), there exists a unique solution u to the equation BL+μI(u,v)=F(v) for all v∈H01,2(Ω). We formally say that such a u solves the equation Lu+μu=F weakly. Using coërcivity, we also note that such a solution satisfies an estimate of the form
∣∣u∣∣1,2≤const(λ0,CPoinc)⋅∣∣F∣∣.
Remark 8.7. In light of the first existence theorem and the preceding remark, we see that the mapping H01,2(Ω)∋u↦B(u,⋅)∈H0−1,2(Ω) is an isomorphism. We formally express this by saying that L+μI:H01,2(Ω)→H0−1,2(Ω) is an isomorphism.
Remark 8.8. From the above proofs, we immediately see that if bi≡0 and c≥0, then the Dirichlet problem corresponding to L:=−∑i,j=1naij∂i∂j+c admits a unique weak solution, since in this case C2=0. This includes in particular the operator L=−Δ.
Existence via Fredholm Alternative
To get a clearer picture of what gets in the way of the existence and uniqueness of solutions to Lu=f, we employ the so-called Fredholm alternative from functional analysis.
Recall that if X is a Hilbert space, D⊂X is a dense subspace, and L:D→X is a linear mapping, the adjoint of L is the unique mapping L∗:D′→X such that
for all u∈D and v∈D′, ⟨Lu,v⟩=⟨u,L∗v⟩; and
if L0∗:D0′→X is any other linear mapping satisfying the relation ⟨Lu,v⟩=⟨u,L0∗v⟩ for all u∈D and v∈D0′, then D0′⊂D′ and L0∗=L∣D0′.
Note that L is not necessarily continuous (if it were, then L could be uniquely extended to all of X).
Theorem 8.9 (Fredholm alternative). Suppose X is a Hilbert space and K:X→X is compact, i.e. if {xi}i=1∞ is a bounded sequence, then {Kxi}i=1∞ has a convergent subsequence. Either one of the following alternatives holds:
For all g∈X and G∈X∗, there exist uniquef∈X and F∈X∗ such that the equations
(I−K)f=g(#)
and
(I−K∗)F=G(#)’
hold.
The homogeneous equations
(I−K)φ=0(##)
and
(I−K∗)Φ=0(##)’
each have (the same finite number of) non-trivial solutions φ∈X and Φ∈X∗. In this case, a necessary and sufficient condition that (#) and (#)' possess solutions is that g and G satisfy the relations ⟨Φ,g⟩=0 and ⟨G,φ⟩=0 whenever φ and Φ solve (##) and (##)' respectively.
We now reformulate the Dirichlet problem so as to make use of the Fredholm alternative. Suppose u∈H01,2(Ω) solves Lu=f on Ω (in the weak sense), where f∈L2(Ω), assume μ≥0 is as in the first existence theorem and set Lμ:=L+μI, which by the first existence theorem and subsequent remarks is an isomorphism H01,2(Ω)→H0−1,2(Ω). Noting that L=Lμ−μI, we see that
Lu=f⇔(Lμ−μI)u=f⇔(I−μLμ)u=Lμ−1f.
Setting K:=μLμ−1, we also note that if u∈L2(Ω) satisfies (I−K)u=μ−1Kf, then u=Ku+μ−1Kf∈H01,2(Ω) so that by reversing the steps above, Lu=f in the weak sense. Moreover, the mapping K:L2(Ω)→H01,2(Ω)⊂L2(Ω) is compact, since by Remark 8.6, if {fi}⊂L2(Ω) is a bounded sequence, then {Kfi} is bounded by the continuity of K (as a mapping into H01,2(Ω)), but then Rellich-Kondrakov implies that {Kfi} has a subsequence converging in L2(Ω), which establishes the claim. Finally, to fully leverage the Fredholm alternative, we will need a more explicit description of K∗. We first define some useful notions.
Definition 8.10. If bi∈C1(Ω), the elliptic operator L∗ defined by
L∗φ=−i,j=1∑n∂i(aij∂jφ)−i=1∑nbi∂iφ+(c−divb)φ
is called the formal adjoint of L. More generally, we define the adjoint bilinear formBL∗:H01,2(Ω)×H01,2(Ω)→R such that
(v,u)↦BL∗(v,u)=BL(u,v).
Finally, if F∈H0−1,2(Ω), we say that v∈H01,2(Ω) is a weak solution of the adjoint problem
L∗vv∣∂Ω=Fon Ω=0
if for all u∈H01,2(Ω), BL∗(v,u)=F(u).
Remark 8.11. If aij,b∈C1(Ω) and u,v∈C0∞(Ω), we see that BL(u,v)=⟨Lu,v⟩=⟨u,L∗v⟩=BL∗(v,u).
As before, we formally write L∗ for the mapping H01,2(Ω)→H0−1,2(Ω) such that v↦(u↦BL∗(v,u)). Since BL∗ satisfies the same estimates as BL, we see that Lμ∗=(L∗+μI):H01,2(Ω)→H0−1,2(Ω) is also an isomorphism, and a quick calculation shows that the adjoint of K defined before is given by
K∗=μ(Lμ∗)−1.
In particular, Φ∈L2(Ω) solves (##)' iff
Φ−μ(Lμ∗)−1Φ=0⇔Lμ∗Φ−μΦ=0⇔L∗Φ=0,
i.e. iff Φ solves the adjoint problem.
Finally, if Φ solves (##)', then for f∈L2(Ω),
⟨Φ,Lμ−1f⟩=0⇔μ−1⟨Φ,Kf⟩=0⇔⟨K∗Φ,f⟩=0⇔⟨Φ,f⟩=0.
Altogether, the Fredholm alternative yields the following existence theorem:
Theorem 8.12 (second existence theorem). One of the following statements is true:
For any f∈L2(Ω), the Dirichlet problem (*) possesses a unique solution u∈H01,2(Ω).
There exist a finite number of nontrivial solutions to the homogeneous problem
Luu∣∂Ω=0on Ω≡0
and the same number of nontrivial solutions Φ∈H01,2(Ω) to the adjoint homogeneous problem
L∗ΦΦ∣∂Ω=0on Ω≡0.
A solution to (*) exists iff for any solution Φ to the adjoint homogeneous problem, ⟨Φ,f⟩=0.