Lecture 8 A Priori Estimates for Poisson’s Equation. � Recall that N f = Ω Γ(x − y)f (y)dy is called the Newtonian Potential of f . Proposition 1 Suppose Ω is bounded domain, f ∈ L1 (Ω), and ω = N f is the Newto­ nian potential of f . Then ω ∈ C 1 (Rn ) and � Di ω(x) = Di Γ(x − y)f (y)dy, ∀x ∈ Ω. Ω Proof: Γ = C|x|2−n =⇒ |DΓ| ≤ C|x|1−n , therefore � v(x) = Di Γ(x − y)f (y)dy Ω � is well defined. (|v(x)| ≤ �f �L∞ Ω |Di Γ|dy ≤ C�f �L∞ .) Define η� (t) to be C ∞ function with properties: (1) η� (t) = 0 for t < �; (2) η� (t) = 1 for t > 2�; (3) 0 ≤ η� (t) ≤ 1; (4) |Dη� | ≤ 2� . Define ω� (x) to be � ω� (x) = Γ(x − y)η� (|x − y |)f (y)dy. Ω Then ω� (x) ∈ C1 and ω� (x) −→ ω uniformly. � (Di Γ(x − y) − Di (Γ(x − y)η� (|x − y |)))f (y)dy v(x) − Di ω� (x) = �Ω Di ((1 − η� (|x − y |))Γ(x − y))f (y)dy = Ω � Di ((1 − η� (|x − y |))Γ(x − y))f (y)dy = |x−y|≤2� So � 2 ( |Γ(x − y)| + |Di Γ(x − y)|)dy |x−y|≤2� � � � 2 |Γ(z)|dz + |Di Γ(z)|dz) ≤ sup |f |( � |z|≤2� |z|≤2� � � 2 C C ≤ sup |f |( dz + dz) n−2 n−1 � |z |≤2� |z| |z|≤2� |z| � � 2 ≤ C sup |f |( rdr + dr) � r≤� r≤� |v(x) − Di ω� (x)| ≤ sup |f | ≤ C · � sup |f |. Now we have ω� → ω and Di ω� → v uniformly on compact subsets as � → 0, thus ω ∈ C 1 (Rn ) and Di ω = v. � 1 Theorem 1 Let u ∈ C 2 (Ω), f ∈ L∞ (Ω) and Δu = f in Ω. Then for any compact subdomain Ω� ⊂⊂ Ω, �u�C 1 (Ω� ) ≤ C(�u�C 0 (Ω) + �f �L∞ (Ω) ). � Proof: Let ω be the Newtonian potential of f , i.e. ω(y) = Ω Γ(x − y)f (x)dx. Then from Green’s representation formula, � ∂ ∂u v(y) = u(y) − ω(y) = u(x) Γ(x − y) − Γ(x − y) dσx ∂νx ∂ν ∂Ω is a harmonic function. So � � �ω�C 0 (Ω� ) = sup | Γ(x − y)f (x)dx| ≤ �f �L∞ (Ω) sup y∈Ω� y∈Ω� Ω Ω C ds ≤ C�f �L∞ (Ω) , |x − y |n−2 and � Di Γ(y − x)f (x)ds � ≤ �f �L∞ (Ω) Di Γ(y − x)ds Di ω(y) = Ω Ω ≤ �f �L∞ (Ω) C ds |x − y | n−1 = C�f �L∞ (Ω) Thus �Di ω�C 0 (Ω) ≤ C�f �L∞ (Ω) , and so �ω�C 1 (Ω) ≤ C�f �L∞ (Ω) . Since v is harmonic, we have �Dv�C 0 (Ω� ) ≤ C�v�C 0 (Ω) ≤ C(�u�C 0 (Ω) + �ω�C 0 (Ω) ) ≤ C(�u�C 0 (Ω) + �f �L∞ (Ω) ). Thus �u�C 1 (Ω� ) ≤ �v�C 1 (Ω� ) + �ω�C 1 (Ω� ) ≤ C(�u�C 0 (Ω) + �f �L∞ (Ω) ). More over, one can show that for any 0 ≤ α < 1, �u�C 1,α (Ω� ) ≤ C(�u�C 0 + �f �L∞ ). This is not true for α = 1. 2 � But if f ∈ C α (Ω), then �u�C 2 (Ω� ) ≤ C(�u�0C + �f �L∞ (Ω) ) and �u�C 2,α (Ω� ) ≤ C(�u�αC + �f �L∞ (Ω) ). C 1,α estimate for Newtonian Potential (Ω Bounded) � � ω(x) = Γ(x − y)f (y)dy =⇒ Di ω(x) = Di Γ(x − y)f (y)dy. Ω Ω Theorem 2 If f ∈ L∞ , then ω ∈ C 1,α (Ω). Proof: Take x, x ∈ Ω, let δ = |x − x|, and ξ = 12 (x + x). � Di ω(x) − Di Ω(x) = (Di Γ(x − y) − Di Γ(x − y))f (y)dy Ω � ≤ �f �L∞ (Ω) |Di Γ(x − y) − Di Γ(x − y)|dy Ω � ≤ �f �L∞ (Ω) ( |Di Γ(x − y) − Di Γ(x − y)|dy Bδ (ξ) � |Di Γ(x − y) − Di Γ(x − y)|dy) + Ω−Bδ (ξ) = �f �L∞ (Ω) (I + II), where � � |Di Γ(x − y)|dy + I≤ |Di Γ(x − y)|dy Bδ (ξ) Bδ (ξ) � � |Di Γ(x − y)|dy + ≤ B 3δ (x) |Di Γ(x − y)|dy B 3δ (x) �2 2 ≤C B 3δ (x) 1 dy + |x − y|n−1 2 � B 3δ (x) 1 dy |x − y)|n−1 2 3δ ≤C· 2 = C |x − x0 |, and � |Di Γ(x − y) − Di Γ(x − y)|dy II = Ω−Bδ (ξ) � ≤ |x − x| � ≤C ·δ |DDi Γ(x̂ − y)|dy Ω−Bδ (ξ) |y−ξ|≥δ 1 dy |ˆ x − y |n 3 Since we have |y − ξ | ≤ |y − x ˆ| + |ˆ x − ξ| δ ≤ |y − x| ˆ + 2 1 ≤ |y − x| ˆ + |y − ξ | 2 1 =⇒ |y − ξ | ≤ |y − x̂|. 2 Thus � 1 II ≤ C · δ dy | y − ξ |n |y−ξ|≥δ � R 1 ≤C ·δ dr δ r ≤ C · δ(log R − log δ) ≤ C · δ(log R + cδ α−1 ) (� − log δ ≤ Cδ α−1 f or 0 ≤ α < 1) = Cδ + Cδ α ≤ Cδ α = C |x − x|α . Combine the above results, we get ω ∈ C 1,α (Ω). Further more, we get the C 1,α estimate �ω�C 1,α (Ω) ≤ C�f �L∞ (Ω) , 0 ≤ α < 1. � Just as last time, this implies Corollary 1 Suppose Δu = f, f ∈ L∞ (Ω), Ω� ⊂⊂ Ω, then for 0 ≤ α < 1, �u�C 1,α (Ω� ) ≤ C(�u�C 0 (Ω) + �f �L∞ (Ω) ). Remark 1 Look at the above proof and assume f ∈ Lp (Ω). � Di ω(x) − Di ω(x) = (Di Γ(x − y) − Di Γ(x − y))f (y)dy Ω � � ≤ { |Di Γ(x − y) − Di Γ(x − y)|q dy}1/q { |f (y)|p dy}1/p Ω Ω 1 1 p + = 1 =⇒ q = . p q p−1 In the 2nd part of the above proof, we had � II ≤ { ||x − x||DD ˆ x − y)||q }1/q i Γ(ˆ Ω � 1 )1/q ≤ C · δ( nq |y−ξ|≥δ |y − ξ | ≤ C · δ · (δ −nq+n )1/q =C ·δ 1−n+ n q . 4 Let α = 1 − n + n p−1 p , then p(α − 1 + n) = pn − n, i.e. p = If Δu = f, f ∈ Lp (Ω), p = n 1−α , Ω� n 1−α , ⊂⊂ Ω, then �u�C 1,α (Ω� ) ≤ C(�f �Lp (Ω) + �u�C 0 (Ω) ). Later we will show �u�W 2,p (Ω� ) ≤ C(�f �Lp (Ω) + �u�C 0 (Ω) ). So C 1,α estimate follows by Sobolev Embedding theorem. 5 we have: