This action might not be possible to undo. Are you sure you want to continue?

In many problems of mathematical physics and variational calculus it is not suﬃcient to deal with the classical solutions of diﬀerential equations. It is necessary to introduce the notion of weak derivatives and to work in the so called Sobolev spaces. Let us consider the simplest example — the Dirichlet problem for the Laplace equation in a bounded domain Ω ⊂ Rn : u = 0, x∈Ω u(x) = ϕ(x), x ∈ ∂Ω, (∗)

where ϕ(x) is a given function on the boundary ∂Ω. It is known that the Laplace equation is the Euler equation for the functional

n

l(u) =

Ω j=1

∂u ∂xj

2

dx.

We can consider (∗) as a variational problem: to ﬁnd the minimum of l(u) on the set of functions satisfying condition u|∂Ω = ϕ. It is much easier to minimize this functional not in C 1 (Ω), but in a larger class. 1 Namely, in the Sobolev class W2 (Ω). 1 W2 (Ω) consists of all functions u ∈ L2 (Ω), having the weak derivatives ∂j u ∈ L2 (Ω), j = 1, . . . , n. If the boundary ∂Ω is smooth, then the trace of u(x) on ∂Ω is well deﬁned and relation u|∂Ω = ϕ makes sense. (This follows from the so called boundary trace theorem“ for Sobolev spaces.) ” 1 If we consider l(u) on W2 (Ω), it is easy to prove the existence and uniqueness of solution of our variational problem. 1 The function u ∈ W2 (Ω), that gives minimum to l(u) under the condition u|∂Ω = ϕ, is called the weak solution of the Dirichlet problem (∗).

We’ll study the Sobolev spaces, the extension theorems, the boundary trace theorems and the embedding theorems. Next, we’ll apply this theory to elliptic boundary value problems.

1

§1: Preliminaries

Let us recall some deﬁnitions and notation. Deﬁnition An open connected set Ω ⊂ Rn is called a domain. By Ω we denote the closure of Ω; ∂Ω is the boundary. Deﬁnition We say that a domain Ω ⊂ Ω ⊂ Rn is a strictly interior subdomain of Ω and write Ω ⊂⊂ Ω, if Ω ⊂ Ω. If Ω is bounded and Ω ⊂⊂ Ω, then dist {Ω , ∂Ω} > 0. We use the following notation: ∂u x = (x1 , x2 , . . . , xn ) ∈ Rn , ∂j u = , ∂xj α = (α1 , α2 , . . . , αn ) ∈ Zn + |α| = α1 + α2 + . . . + αn , Next, Deﬁnition u = (∂1 u, . . . , ∂n u) , is a multi–index ∂ |α| u ∂xα1 ∂xα2 ...∂xαn n 1 2

n j=1

∂αu =

| u| =

|∂j u|2

1/2

**Lq (Ω), 1 ≤ q < ∞ , is the set of all measurable functions u(x) in Ω such that the norm
**

1/q

u is ﬁnite.

q,Ω

=

Ω

|u(x)|q dx

**Lq (Ω) is a Banach space. We’ll use the following property: Let u ∈ Lq (Ω), 1 ≤ q < ∞. We denote
**

|z|≤ρ

Here u(x) is extended by zero on Rn \Ω. Jρ (u; Lq ) is called the modulus of continuity of a function u in Lq (Ω). Then Jρ (u; Lq ) → 0 as ρ → 0. 2

Jρ (u; Lq ) = sup

n

|u(x + z) − u(x)|q dx

1/q

.

Deﬁnition Lq,loc (Ω), 1 ≤ q < ∞, is the set of all measurable functions u(x) in Ω such that Ω |u(x)|p dx < ∞ for any bounded strictly interior subdomain Ω ⊂⊂ Ω. Lq,loc (Ω) is a topological space (but not a Banach space). We say that uk −→ u in Lq,loc (Ω), if uk − u Ω ⊂⊂ Ω

k→∞ k→∞ q,Ω

−→ 0 for any bounded

**Deﬁnition L∞ (Ω) is the set of all bounded measurable functions in Ω; the norm is deﬁned by u ∞,Ω = ess sup |u(x)|
**

x∈Ω

Deﬁnition C l (Ω) is the Banach space of all functions in Ω such that u(x) and ∂ α u(x) with |α| ≤ l are uniformly continuous in Ω and the norm u

C l (Ω)

=

|α|≤l

x∈Ω

sup |∂ α u(x)|

is ﬁnite. If l = 0, we denote C 0 (Ω) = C(Ω). Remark If Ω is bounded, then u ty of u, ∂ α u, |α| ≤ l < ∞ follows from the uniform continui-

C l (Ω)

Deﬁnition C l (Ω) is the class of functions in Ω such that u(x) and ∂ α u, |α| ≤ l, are continuous in Ω. Remark Even if Ω is bounded, a function u ∈ C l (Ω) may be not bounded; it may grow near the boundary. Deﬁnition ∞ C0 (Ω) is the class of the functions u(x) in Ω such that a) u(x) is inﬁnitely smooth, which means that ∂ α u is uniformly continuous in Ω, ∀α; b) u(x) is compactly supported: supp u is a compact subset of Ω. 3

be a function such that ∞ ω ∈ C0 (Rn ). ωρ (x) = 0 if n Deﬁnition wρ is called a molliﬁer. (5) . Deﬁnition uρ (x) is called a molliﬁcation or regularization of u(x). Let ω(x). We extend u(x) by zero on Rn \Ω and consider the convolution ωρ ∗ u =: uρ uρ (x) = In fact. and ω(x)dx = 1. the integral is over Ω ∩ {y : |x − y| < ρ}. we may take 1 c exp − 1−|x|2 ω(x) = where constant c is chosen so that condition (1) is satisﬁed. Deﬁnition of molliﬁcation The procedure of molliﬁcation allows us to approximate function u ∈ Lq (Ω) by smooth functions. n ωρ (x − y)u(y)dy. For ρ > 0 we put ωρ (x) = ρ−n ω ∞ Then ωρ ∈ C0 (Rn ). 4 if if 0 |x| < 1 |x| ≥ 1 x ρ . ω(x) ≥ 0. (2) |x| ≥ ρ. and let u ∈ Lq (Ω) with some 1 ≤ q ≤ ∞. ωρ (x) ≥ 0. x ∈ Rn . (1) For example. (3) (4) ωρ (x)dx = 1. x ∈ Rn .§2: Molliﬁcation of functions 1. Let Ω ⊂ Rn be a domain. ω(x) = 0 if |x| ≥ 1.

3) Let u ∈ Lq (Ω) with some q ∈ [1. We have |uρ (x)| ≤ ≤ 5 = n |u(y)|q dy n ωρ (x − y)|u(y)|dy ∞ n u ωρ (x − y)dy =1 dy |u(y)|q ⇒ |uρ (x)|q ≤ n ωρ (x − y)|u(y)|q dy n ωρ (x − y) |u(y)|q dy n ωρ (x − y) uρ n ≤ u q. ∞]. since ωρ (x − y) = 0. and α ∂ α uρ (x) = n ∂x ωρ (x − y)u(y)dy. 2) uρ (x) = 0 if dist {x. Proof: Case 1 : 1 < q < ∞.2.Ω . the operator Yρ : u → uρ is a linear continuous operator from Lq (Ω) to Lq (Rn ) and Yρ Lq (Ω)→Lq ( n ) ≤ 1. y ∈ Ω. This follows from ωρ ∈ C ∞ . Ω} ≥ ρ. Then q. In other words. Properties of molliﬁcation 1) uρ ∈ C ∞ (Rn ). 1 o Let 1 + q = 1. (6) 1/q n ωρ (x − y)|u(y)|q dy ωρ (x − y)dx =1 . we have q |uρ (x)| = ≤ ωρ (x − y)1/q ωρ (x − y)1/q u(y)dy 1/q n n =1 By (4). we obtain n |uρ (x)|q dx ≤ = dx n n Case 2 : q = ∞. By the H¨lder inequality and (4).

ρ → 0. 1/q .Ω Proof The proof is based on the following property: if u ∈ Lq (Ω) (and u(x) is extended by 0). We integrate the inequality |uρ (x)| ≤ and obtain: n n n n ωρ (x − y)|u(y)|dy 4) Let u ∈ Lq (Ω). uρ − u q.) Case 1 : 1 < q < ∞. By (4) and (5) we have |uρ (x) − u(x)| ≤ = n ωρ (x − y) (u(y) − u(x)) dy ωρ (x − y)1/q ωρ (x − y)1/q (u(y) − u(x)) dy n Then. Lq ) → 0 n ωρ (x − y) |u(y) − u(x)|q dy |uρ (x)| dx ≤ dx ωρ (x − y)|u(y)|dy = n |u(y)|dy (7) as ρ → 0. then sup |z|≤ρ n |u(x + z) − u(x)|q dx (Jρ (u. 1 ≤ q < ∞. 1/q =: Jρ (u. Consequently. Then q. Lq ) is called the modulus of continuity of u in Lq . by the H¨lder inequality. it follows that o 1/q n =1 6 |uρ (x) − u(x)| ≤ ωρ (x − y)dy uρ − u n → 0 as →0 as ρ → 0.⇒ uρ ∞ ≤ u ∞ Case 3 : q = 1.

then uρ − u C(Ω ) →0 as ρ→0 7 = dz ωρ (z) ⇒ |uρ (x) − u(x)| dx ≤ ⇒ uρ − u n ≤ Jρ (u. then we may loose continuity. q. uρ − u C(Ω) However. If u ∈ C(Ω) and we extend u(x) by zero. In general. |u(y + z) − u(y)| dy . 0 as ρ → 0. Remark If q = ∞. there is NO such property. Lq ) → 0 as n ≤ sup |u(y + z) − u(y)|q dy dx n ωρ (x − y) |u(y) − u(x)| dy n = |uρ (x) − u(x)|q dx ≤ dx n n ωρ (x − y) |u(y) − u(x)|q dy n dz ωρ (z) |u(y + z) − u(y)|q dy dz |z|<ρ =1 ωρ (z) ρ → 0. We have |uρ (x) − u(x)| ≤ ωρ (x − y) |u(y) − u(x)| dy n n x−y=z |z|<ρ → ≤ Jρ (u.Hence. we have the following property: 5) If u ∈ C(Ω) . Ω ⊂⊂ Ω and Ω is bounded. since L∞ –limit of smooth functions uρ (x) must be a continuous function. L1 ) 0 as ρ → 0. Case 2 : q = 1. Lq ))q . n n x−y=z |z|<ρ |z|≤ρ = (Jρ (u.

**Proof Let ρ < dist {Ω ; ∂Ω}. Then
**

uρ (x) − u(x)

=

x−y=z

n

ωρ (x − y) (u(y) − u(x)) dy ωρ (z) (u(x − z) − u(x)) dz ρ→0

⇒ sup |uρ (x) − u(x)|

x∈Ω

≤ →

x∈Ω |z|≤ρ

sup sup |u(x − z) − u(x)|

0 as

(since u(x) is continuous in Ω).

8

=

n

∞ §3: Class C0 (Ω)

∞ By C0 (Ω) we denote the class of inﬁnitely smooth functions in Ω with compact support: ∞ u ∈ C0 (Ω)

⇔

u ∈ C ∞ (Ω) and supp u ⊂ Ω.

Theorem 1

∞ C0 (Ω) is dense in Lq (Ω), 1 ≤ q < ∞

**Proof Let u ∈ Lq (Ω) and ε > 0. Let Ω be a bounded domain, Ω ⊂⊂ Ω, and u We put u(ε) (x) = Then u − u(ε)
**

(ε) q,Ω q,Ω\Ω

ε ≤ . 2 if x ∈ Ω if x ∈ Ω\Ω ≤

ε 2

u(x) 0

(ε)

**property 4) of molliﬁcation, uρ − u(ε) Hence, uρ − u Note that uρ ∈
**

(ε) q,Ω ∞ C0 (Ω)

≤

ε 2.

**Let uρ (x) be the molliﬁcation of u(ε) (x). By
**

(ε) q,Ω

for suﬃciently small ρ.

≤ ε for suﬃciently small ρ. if ρ < dist {Ω , ∂Ω} .

**Theorem 2 Let u ∈ L1,loc (Ω), and suppose that u(x)η(x)dx = 0,
**

Ω ∞ ∀η ∈ C0 (Ω).

(8)

Then u(x) = 0,

a. e. x ∈ Ω.

**Theorem 2 is an analog of the Main Lemma of variational calculus. Proof 1) First, let us prove that u(x)η(x)dx = 0
**

Ω

9

for any η ∈ L∞ (Ω) with compact support supp η ⊂ Ω. Suppose that supp η ⊂ Ω , where Ω is a bounded domain and Ω ⊂⊂ Ω. Then Let Ωρ0 = {x : dist {x; Ω } < ρ0 }, and let χρ0 (x) = By (8), u(x)ηρ (x)dx = 0,

Ω ∞ ηρ ∈ C0 (Ω) if ρ < dist Ω ; ∂Ω =: 2ρ0 .

1 0

if x ∈ Ωρ0 otherwise

ρ < ρ0 .

(9)

**Since η ∈ L1 (Ω), by property 4) of molliﬁcation, ηρ − η 1,Ω → 0 as ρ → 0. Then there exists a sequence {ρk }k∈ , ρk → 0, ρk < ρ0 , such that ηρk (x) −→ η(x)
**

k→∞ k→∞

**for almost every x ∈ Ω. for a. e. x ∈ Ω. ∞ ), we have
**

∞,

Then also ηρk (x)u(x) −→ η(x)u(x) Using property 3) (that ηρ ∞ ≤ η

**and the right-hand side in (10) belongs to L1 (Ω). Then, by the Lebesgue Theorem,
**

Ω

|u(x)ηρk (x)| ≤ χρ0 (x)|u(x)| η

u(x)ηρk (x)dx −→

k→∞ Ω

By (9), the left-hand side is equal to zero. Hence, Ω u(x)η(x)dx = 0. 2) Now, let Ω be a bounded domain such that Ω ⊂⊂ Ω. We put η(x) = Then u(x)η(x) =

u(x) |u(x)|

0

, if u(x) = 0, x ∈ Ω , otherwise |u(x)| 0 ,x∈Ω , x ∈ Ω\Ω

**Since η(x) is L∞ –function with compact support supp η ⊂ Ω ⊂ Ω, then, by part 1), 0=
**

Ω

u(x)η(x)dx =

Ω

It follows that u(x) = 0 for a. e. x ∈ Ω . Since Ω is an arbitrary bounded domain such that Ω ⊂⊂ Ω, then u(x) = 0, a.e. x ∈ Ω

10

(10)

u(x)η(x)dx.

|u(x)|dx.

and u(x)∂ α η(x)dx = (−1)|α| Ω Ω ∞ v(x)η(x)dx. a. ∂ α u may exist in the weak sense wihout existing in the classical sense. we don’t need the existence of derivatives of the smaller order (like in the classical deﬁnition).loc (Ω) are both weak derivatives of u. Then. 2) The weak derivative is deﬁned as an element of L1.§4: Weak derivatives 1. Remark 1) To deﬁne the weak derivative ∂ α u. Of course. ∀η ∈ C0 (Ω). v(x) = w(x).loc (Ω). ∞ ∀η ∈ C0 (Ω).e. (v(x) − w(x)) η(x)dx = 0. and is denoted by ∂ α u. 11 . w ∈ L1. by (11). If u(x) is suﬃciently smooth to have continuous derivative ∂ α u. Hence. Deﬁnition and properties of weak derivatives Deﬁnition 1 Let α be a multi–index. (11) Then v is called the weak ( or distributional ) partial derivative of u in Ω. x ∈ Ω. Ω By Theorem 2. Suppose that u.loc (Ω).loc (Ω) and v. v ∈ L1. Suppose that u ∈ L1. Properties of ∂ α u 1) Uniqueness Proof Uniqueness of the weak derivative follows from Theorem 2. so we can change it on some set of measure zero. we can integrate by parts: u(x)∂ α η(x)dx = Ω Ω (−1)|α| ∂ α u(x)η(x)dx. the classical derivative ∂ α u is also the weak derivative.

Proof Obviously. ∂Ω}.loc (Ω) and there exist weak derivatives v1 = ∂ α u1 .loc (Ω) and v = ∂ α u. Then vρ (x) = ∂ α uρ (x) if ρ < dist {x. Obvious 4) Molliﬁcation of the weak derivative Derivative of molliﬁcation is equal to molliﬁcation of derivative“. Proof Let ρ < dist {x. then for η(y) := ωρ (x − y) we have η ∈ C0 (Ω). = (−1)|α| Ω =∂ α (c1 u1 +c2 u2 ) 3) If v = ∂ α u in Ω . Suppose that u. This ” is true in any bounded strictly interior domain Ω ⊂⊂ Ω. u2 ∈ L1. α α Note that ∂x ωρ (x − y) = (−1)|α| ∂y ωρ (x − y). c2 ∈ C. the derivative ∂ α uρ in (12) is understood in the classical sense. v2 = ∂ α u2 ∈ L1.loc (Ω). We have uρ (x) = Ω ωρ (x − y)u(y)dy α Then ∂ α uρ (x) = Ω ∂x ωρ (x − y)u(y)dy. v ∈ L1. 12 . (c1 u1 + c2 u2 ) ∂ α ηdx = c1 Ω Ω c1 . ∞ Since ρ < dist {x. Hence.2) Linearity If u1 . then there exists ∂ α (c1 u1 + c2 u2 ) and ∂ α (c1 u1 + c2 u2 ) = c1 ∂ α u1 + c2 ∂ α u2 . u1 ∂ α ηdx + c2 Ω u2 ∂ α ηdx v2 ηdx Ω = (−1)|α| c1 Ω v1 ηdx + (−1)|α| c2 (c1 v1 + c2 v2 ) ηdx. α u = v. then v = ∂ α u in Ω for any Ω ⊂ Ω. ∂Ω} . ∂ α uρ (x) = (−1)|α| Ω α ∂y ωρ (x − y)u(y)dy. we obtain By deﬁnition of the weak derivative ∂ ∂ α uρ (x) = Ω ωρ (x − y)v(y)dy = vρ (x). (12) The functions uρ and vρ are smooth. ∂Ω}.

∂Ω} < ρo . ∞ ∀η ∈ C0 (Ω). in general. 1 ≤ q < ∞. Proof This follows from property 4) of molliﬁcation and property 4) of weak derivatives: ∂ α u = v ∈ Lq (Ω). Deﬁnition 1 ⇔ Deﬁnition 2 Proof 1) Deﬁnition 1 ⇐ Deﬁnition 2.Ω → 0 as ρ → 0. the weak derivative ∂ α u in Rn does not exist. Here α is a multi–index and |α| = l. Then v is called the weak derivative of u in Ω : ∂ α u = v. Another deﬁnition of the weak derivative Deﬁnition 2 Suppose that u. ∂ α uρ = vρ in Ω vρ − v q. → 0 as Remark If we extend u(x) by zero on Rn \Ω. we have convergence ρ→0 ∂ α uρ −→ ∂ α u in Lq (Ω ) only for bounded strictly interior domain Ω. for any bounded strictly interior domain Ω ⊂⊂ Ω. Hence. Then ∂ α uρ − ∂ α u q. then um ∂ α ηdx = (−1)|α| Ω Ω ∂ α um ηdx.Ω ( for suﬃciently small ρ) .loc (Ω) and there exists the weak derivative ∂ α u such that ∂ α u ∈ Lq (Ω). Exclusion: if u(x) = 0. then.loc (Ω). if dist {x. m→∞ m→∞ m ∈ N. such that um −→ u and ∂ α um −→ v in L1. ρ→0 and ∂ α u ∈ Lq (Ω). then ∂ α uρ − ∂ α u q.loc (Ω) and there exists a sequence um ∈ C l (Ω). 2.5) Suppose that u ∈ L1. Since um ∈ C l (Ω). ρ → 0. v ∈ L1. (13) 13 .Ω −→ 0.

Ω Similarly. Thus. Prove this yourself. be a sequence of bounded domains such that (m) Ωm ⊂⊂ Ω. m→∞ We want to ﬁnd a sequence um ∈ C ∞ (Ω) such that um −→ u and m→∞ ∂ α um −→ v in L1.loc (Ω). We put (m) um (x) = uρm (x).loc (Ω). Then um ∈ C ∞ (Ω) and um −→ u in L1. x ∈ Ω. ∞ ∀η ∈ C0 (Ω). prove that ∂ α um −→ v in L1. m→∞ 14 . u∂ α ηdx = (−1)|α| Ω Ω vηdx. using property 4) of molliﬁcation.loc (Ω). 2) Deﬁnition 1 ⇒ Deﬁnition 2.For η ﬁxed. by property 5) of ∂ α u. Let {Ωm } . v = ∂ α u in the sense of Deﬁnition 2. Let u. v ∈ L1. m→∞ Next. Consider the molliﬁcation of u(m) : uρ ∈ C ∞ (Ω). and let v = ∂ α u in the sense of Deﬁnition 1. Let {ρm }m∈ be a sequence of positive numbers such that ρm → 0 as m → ∞. It means that v = ∂ α u in the sense of Deﬁnition 1.loc (Ω). m ∈ N. We put Ωm ⊂ Ωm+1 and Ωm = Ω. the left–hand side of (13) tends to (um − u) ∂ α ηdx ≤ max |∂ α η| Ω u∂ α ηdx as m → ∞: m→∞ Ω supp η |um − u| dx −→ 0. m∈ u(m) (x) = u(x) 0 if x ∈ Ωm otherwise Then u(m) ∈ L1 (Ω). Conse- vηdx. the right–hand side of (13) tends to (−1)|α| quently.

the operator ∂ α is closed.loc (Ω) and v. Remark The conclusion of Theorem 3 remains true under weaker assumptions that um ηdx −→ m→∞ Ω uηdx and Ω Ω ∂ α um ηdx −→ m→∞ Ω vηdx. Weak derivatives of the product of functions Proposition If u.loc (Ω) and um −→ u in L1. In other words. then q ∂j (uv) = (∂j u) v + u (∂j v) .Theorem 3 Let um ∈ L1. ∂j u ∈ L1. x ∈ Ω 0 otherwise 15 v (x) = ˜ v(x) . ↓m→∞ vηdx. We put u(x) = ˜ u(x) .loc (Ω).) 3. (It means that um → u and ∂ α um → v in D (Ω). Let Ω be a bounded domain such that supp η ⊂ Ω ⊂⊂ Ω. x ∈ Ω 0 otherwise . Proof By Deﬁnition 1.loc (Ω) with some 1 1 < q < ∞. Proof 1) Case 1: 1 < q < ∞ ∞ Let us ﬁx η ∈ C0 (Ω). ∂j v ∈ C(Ω).loc (Ω) and ∂ α um −→ v in L1. ∞ ∀η ∈ C0 (Ω).loc (Ω). Then v = ∂ α u. ∂j v ∈ Lq . Ω ∞ ∀η ∈ C0 (Ω) ↓m→∞ Ω u∂ α ηdx = (−1)|α| ∞ ∀η ∈ C0 (Ω) ⇒ v = ∂ α u in the sense of Deﬁnition 1. Suppose that there m→∞ exist weak derivatives ∂ α um ∈ L1. ∂j u ∈ Lq. 1 + q = 1 or if u. for ∂ α um we have um ∂ α ηdx = (−1)|α| Ω Ω m→∞ ∂ α um ηdx.loc (Ω). and v.

supp η → 0. as ρ → 0. ∂j v = ∂j v in Ω (it is clear from Deﬁnition 1). By property 4) of molliﬁcations. (14) ˜ ˜ Ω uρ vρ ∂j ηdx −→ ˜ ˜ ≤ ≤ + ρ→0 uv∂j ηdx Ω (15) Ω We have Ω (˜ρ vρ − uv ) ∂j ηdx u ˜ ˜˜ Ω (˜ρ − u) vρ ∂j ηdx + u ˜ ˜ q.Ω → 0 as ρ → 0 Next.(16) it follows that Ω uv∂j ηdx = − ((∂j u) v + u (∂j v)) ηdx Ω ∞ This identity is proved for any η ∈ C0 (Ω). ∂j vρ − ∂j v ˜ ˜ uρ vρ ∂j ηdx = − ˜ ˜ = − Let us show that Ω ∂j uρ − ∂j u ˜ ˜ q. Since uρ . ˜ ˜ uρ − u ˜ ˜ q. ∂j u ∈ Lq (Ω ). Prove yourself 16 . vρ − v ˜ ˜ q . ˜ ˜ So. ∂j v ∈ Lq (Ω ). ∂j u = ∂j u in Ω . ˜ ˜ By property 5) of weak derivatives.Ω → 0. we have ˜ ˜ ∂j (˜ρ vρ ) ηdx u ˜ Ω Ω Ω (∂j uρ ) vρ ηdx − ˜ ˜ uv ∂j ηdx = ˜˜ uρ (∂j vρ ) ηdx.Then u ∈ Lq (Ω ).supp η q .Ω Ω u (˜ρ − v ) ∂j ηdx ˜ v ˜ uρ − u ˜ ˜ →0 vρ ˜ q . v ∈ Lq (Ω ).Ω max |∂j η| → 0 Similarly.Ω vρ − v ˜ ˜ →0 q . as ρ → 0. we can show that ((∂j uρ ) vρ + uρ (∂j vρ )) ηdx ˜ ˜ ˜ ˜ Ω as ρ→0 ((∂j u) v + u (∂j v )) ηdx ˜ ˜ ˜ ˜ Ω ρ→0 → = ((∂j u) v + u (∂j v)) ηdx (16) Ω From (14) .Ω max |∂j η| + bounded u ˜ q. vρ are smooth functions. It means (by Deﬁnition 1) that there exists the weak derivative ∂j (uv) and ∂j (uv) = (∂j u) v + u (∂j v) 2) Case q=1. → 0.

j = 1. n. Then u ∈ L ˜ We put u(y) = u(f ˜ ˜ 1. . . ˜ We denote um (y) = um (f −1 (y)). using that ∂um −→ ∂xj in L1. m→∞ Ω is a bounded domain such that Ω ⊂⊂ Ω. and ∂u ˜ = ∂yk Proof Since there exist weak derivatives C 1 (Ω) n j=1 ∂u ∂xj ∂xj ∂yκ ∂u ∂xj ∂u exists a sequence um ∈ such that um −→ u and ∂um −→ ∂xj in ∂xj L1. . j = 1. −1 (y)). Then um ∈ C 1 (Ω). Indeed.loc (Ω). . .loc (Ω). .loc (Ω). . . by Deﬁnition 2. . one can show that ∂xj ∂ um ˜ ∂yk n = j=1 ∂um ∂xj ∂xj ∂yκ m→∞ n j=1 −→ ∂u ∂xj ∂xj ∂yκ ˜ in L1. since |J(x)| is bounded in Ω . there exist weak derivatives ∂u ˜ ∂yk and ∂u ˜ ∂yk 17 n = j=1 ∂u ∂xj ∂xj ∂yk .loc (Ω). for every bounded domain ˜ ˜ ˜ Ω ⊂⊂ Ω we have ˜ Ω ∈ L1. n ∂um ∂xj ∂ um ˜ = ∂yk ∂xj ∂yκ j=1 m→∞ ˜ ˜ Let us check that um −→ u in L1. and um −→ u in L1. ˜ Let y = f (x) be a diﬀeomorphism of class C 1 and f (Ω) = Ω. m→∞ ∂u Similarly. Change of variables Suppose that u ∈ L1.loc (Ω). by usual rule (for ˜ ˜ classical derivatives). . and. .loc (Ω) and there exist weak derivatives ∂j u ∈ L1. n. then there m→∞ m→∞ |˜m (y) − u(y)| dy u ˜ = = ˜ Ω um (f −1 (y)) − u(f −1 (y)) dy |um (x) − u(x)| |J(x)| dx as m→∞ Ω → 0 ˜ Here Ω = f −1 (Ω ) and J(x) is the Jacobian of the transformation f (x) ∂y J(x) = det ∂x . .loc (Ω) Then.4. Here the right-hand side tends to zero. Let us show that there exist ∂u ˜ weak derivatives ∂yκ . κ = 1. . . . n.loc (Ω) for all j = 1.loc (Ω). n. that Deﬁnition 1 and Deﬁnition 2 are equivalent ). . (We can construct this sequence like in the proof.

|α| = l. Then u(x) is a polynomial of order ≤ l − 1 in Ω. . The same is true for weak derivatives. and ∂ α u = ∂ α u = 0. by property 4) of ˜ ∂ α u. x ∈ Ω . so. x ∈ Ω . ˜ By property 4) of molliﬁcation. ∂Ω }. Thus. Therefore. . Consider ˜ ˜ the molliﬁcation uρ (x). 2) Now it is easy to complete the proof by the standard procedure. uρ (x) is a smooth function in Ω and all ˜ ˜ its derivatives of order l are equal to zero. where Pl−1 is a polynomial of order ≤ l − 1. Theorem 4 Suppose that u ∈ L1. uρ − u ˜ 1.Thus. j = 1. then. |α| = l. Proof 1) Let Ω be a bounded domain such that Ω ⊂⊂ Ω. . 18 and k∈ Ωk = Ω. e. i. 5. n. closed!) subspace in L1 (Ω ). If ρ < dist {Ω . It follows that (ρ) (ρ) uρ (x) = Pl−1 (x). Pl−1 −→ u (ρ) ρ→0 in L1 (Ω ). ∂ α uρ (x) = (∂ α u)ρ (x).Ω →0 as ρ → 0. x∈Ω.loc (Ω) and there exist weak derivatives ∂ α u for any multi-index α such that |α| = l (l ∈ N) and ∂ α u = 0 in Ω. . For ordinary derivatives we have the following property: ∂u if ∂xj = 0 in Ω. |α| = l. in Ω . for weak derivatives we have the usual rule of change of variables. The set of all polynomials in Ω of order ≤ l − 1 is a ﬁnite-dimensional (and. The same is true for derivatives of higher order. ∂ α uρ = 0 in Ω . . x ∈ Ω 0 otherwise Then u ∈ L1 (Ω ) . . ˜ ˜ Hence. Let Ω be another bounded domain such that Ω ⊂⊂ Ω ⊂⊂ Ω. Let {Ωk }k∈ be a sequence of bounded domains such that Ωk ⊂⊂ Ω. Ωk ⊂ Ωk+1 . We put u(x) = ˜ u(x) . the limit u(x) must be also a polynomial of order ≤ l − 1: u(x) = Pl−1 (x). then u = const.

The weak dx derivative coincides with the classical derivative almost everywhere. 6. xm one has u(xj ) − u(xj ) < ε. b] → R is absolutely continuous if and only if there exists a function v ∈ L1 (a. b] → R is called absolutely continuous. b ]) xj − xj < δ. b). (⊂ [ a. if for any ε > 0 there exists δ > 0 such that for any ﬁnite set of disjoint intervals x1 .We have poved that for each domain Ωk u(x) = Pl−1 (x). . with m j=1 x2 . then there exists derivative du du dx for almost every x ∈ (a. Absolute continuity property The existence of the weak derivative is related to the absolute continuity property. x ∈ [a. m j=1 We’ll use the following facts: 1) u : [a. xm . b] . b) and dx = v (∈ L1 (a. x ∈ Ω. (k+1) (k) (k) x ∈ Ωk Then Pl−1 (x) is continuation of Pl−1 (x). 19 . A measurable function u(x) is absolutely continuous on [a. . Theorem 5 Let n = 1. 2) If u : [a. ⇒ There exists a polynomial Pl−1 (x) such that u(x) = Pl−1 (x). b)). . x1 . Recall the deﬁnition of absolute continuity for function of one variable. b] → R is absolutely continuous. but continuation of a polynomial is unique. x2 . Deﬁnition Function u : [a. . b) such that x u(x) = u(a) + a v(t)dt. b] if and only if there exists the weak derivative du ∈ L1 (a.

= 0. By statement 1) (already proved). b). The obtained identity b u a dη =− dx b vηdx. a. dx dx Integrate this identity over (a. then u = c + w is also absolutely continuous. du dw d (u − w) = . Proof 1) u(x) is absolutely continuous. b) ⇒ u is a. Then the product η u is also absolutely continuous. it means that in the class of functions equivalent to u.) By Theorem 4. dx c. there exists the weak derivative dw dx which coincides with the classical one and with v. dx x Consider w(x) = a v(t)dt. b). Then (Since η(x) = 0 near a and b). Hence. Next. So. e. ⇒ ∃ weak derivative du ∈ L1 . There exists classical derivative dw dx = v. 20 . Then w(x) is absolutely continuous. b). i. a ∞ ∀η ∈ C0 (a. b). If x = a. There exists the classical derivative d(η u) for dx almost every x ∈ (a. b). dx dx dx (the weak derivative is equal to zero. vη + u a dη dx dx = 0. We have the usual rule: dη d (η u) = vη + u .Remark When we speak about measurable functions. when we say that a measurable function u(x) is absolutely continuous. =0 x u(x) = u(a) + a v(t)dt.. means that v is the weak derivative 2) ∃ weak derivative v = du ∈ L1 (a. Thus. we have u(a) = c + w(a) = c. b b d(η u) a dx dx = 0. let η ∈ C0 (a. there exists an absolutely continuous representative.e. du dx by Deﬁnition 1. c. u − w = const. that are equal to each other almost everywhere. If u(x) is a. x ∈ (a. Since w(x) is absolutely consinuous. b). then there exists the classical derivative du = v almost everywhere dx ∞ and v ∈ L1 (a. Thus. we mean not just one function but a class of functions.

∂x2 in Ω. if they exist. . by Theorem 6. ψ ∈ L1 (0. it means that there exists weak ∂2u ∂2u derivative ∂x1 ∂x2 and ∂x1 ∂x2 =0. . not having derivatives of lower order. b(x )] (as a function of one variable xj ). 1]. Exercise: Prove Theorem 6. 1). By Deﬁnition 1 of weak derivative. 1)2 and u(x1 . for ∀ η ∈ C0 (Ω). xj−1 . and in x2 for x1 ﬁxed. ∂u ∂u Then. 2) Suppose that the domain Ω ⊂ Rn is devided by a smooth (n-1)–dimensional surface Γ into two parts Ω1 and Ω2 . there exists the weak derivative ∂x1 ∂x2 = 0.u(x) is absolutely continuous. b(x )] ⊂ Ω. xn ) and write x = {x . e. We denote x = (x1 . . u Ω ∂2 η dx = ∂x1 ∂x2 = 1 ∂2 η dx1 dx2 + ∂x1 ∂x2 0 0 0 1 1 ∂2 η dx2 + dx1 ϕ(x1 ) 0 0 ∂x1 ∂x2 1 1 ϕ(x1 ) ∂2 η dx1 dx2 ∂x1 ∂x2 0 1 1 ∂2 η dx1 dx2 ψ(x2 ) 0 0 ∂x1 ∂x2 ψ(x2 ) =0 1 =0 = 0.loc (Ω) and there exists the weak derivative ∂xj ∈ L1. but ϕ. . Theorem 6 Let Ω ⊂ Rn . xj }. n > 1. classical derivative = weak derivative = v ∈ L1 (a. then u(x) must be also absolutely continuous in x1 for x2 ﬁxed. x ∈ (a. . Then for almost every x the function u(x . b). 7. This example shows that functions may have derivative of higher order. So. . b(x )] are some intervals such that {x } × [a(x ).) ∂2u However. b). it has classical derivative for a. . Suppose that [a(x ). Ω = Ω1 ∪ 21 . xj ) is absolutely continuous on interval [a(x ).loc (Ω). u(x1 . xj+1 . x2 ) = ϕ(x1 ) + ψ(x2 ). (Since. x2 ) does not have weak derivatives ∂x1 . . ∂u Let u ∈ L1. ∞ Indeed. where ϕ and ψ are not absolutely continuous on [0. Examples 1) Let Ω = (0.

k = 1. Let u1 ∈ C 1 (Ω1 ). . then cos (∠(n. ∂xj ∂xj For this consider continuous functions |x|α . |x| > δ u(δ) (x) = δα . it follows that ∃ ∂xj = vj ∂u(δ) L1 (Ω) ∂xj −→ 22 . α > −n + 1. |x| > δ 0 . then the integral over Γ is ∂u equal to zero. In this case. Prove that there exist the weak derivatives ∂u ∂u and = αxj |x|α−2 . if cos (∠(n. ∂u If Γ (u1 (x) − u2 (x)) η cos (∠(n. |x| ≤ δ ∂u(δ) = ∂xj αxj |x|α−2 . we can integrate by ∞ parts in Ωk : for η ∈ C0 (Ω) we have: u Ω u1 (x) . if Γ is parallel to the axis 0xj . by Theorem 3. Let n(x) be the unit normal vector to Γ exterior with respect to Ω1 . then there exists ∂xj . ⇒ Even if u1 |Γ = u2 |Γ . . j = 1. Exercise Let Ω = {x ∈ Rn : |x| < 1}. then ∂xj does not exist. u2 ∈ C 1 (Ω2 ). x ∈ Ω1 u2 (x) . weak derivatives do not exist. .Ω2 ∪ Γ. 2. 0xj )) = 0. there exists the weak derivative ∂xj and ∂u = ∂xj ∂u1 ∂xj ∂u2 ∂xj in Ω1 in Ω2 ∂u Also. |x| ≤ δ vj := αxj |x|α−2 . u(x) = If u1 |Γ = u2 |Γ . Since uk (x). n. is a C 1 –function in Ωk . From the previous example we know that ∃ L1 (Ω) Check that u(δ) −→ u and ∂u Then. 0xj )) dS(x) If u1 = u2 on Γ (we have NO jump on Γ). and let u(x) = |x|α . then. in general. For example. 0xj )) = 0. 0xj )) dS(x) = 0. . the tangential derivative exists. x ∈ Ω2 ∂η dx = ∂xj u1 Ω1 = − + Γ Ω1 ∂η ∂η dx + dx u2 ∂xj ∂xj Ω2 ∂u2 ∂u1 η dx − η dx + ∂xj Ω2 ∂xj (u1 (x) − u2 (x)) η cos (∠(n.

Ω |α|≤l 23 . ∂ α um −→ vα as m → ∞. l If p = 2. l We introduce the (standard) norm in Wp (Ω): u l Wp (Ω) = Ω |α|≤l |∂ α u|p dx 1/p . Proof l Let {um } be a fundamental sequence in Wp (Ω). l Then we say that u ∈ Wp (Ω). there exist functions u. Then also um → u. l In other words. Wp (Ω) is a Banach space. |α| ≤ l. ∂ α um → vα in L1. Hence. Deﬁnition of Wp (Ω) ◦ (1 ≤ p < ∞. um −→ u l Wp (Ω) as m → ∞. Proposition l Wp (Ω) is complete. vα = ∂ α u. v)W l (Ω) = 2 ∂ α u(x)∂ α v(x)dx. such that ∂ α u ∈ Lp (Ω). vα ∈ Lp (Ω) such that Lp (Ω) Lp (Ω) um −→ u. 0 2) Wp (Ω) = Lp (Ω).Ω is equivalent to the standard norm. l ∈ Z+ ) Deﬁnition Suppose that u ∈ Lp (Ω) and there exist weak derivatives ∂ α u for any α with |α| ≤ l (all derivatives up to order l). Remark 1) The norm |α|≤l ∂αu p.loc (Ω). that all sequences {∂ m p Since the space Lp (Ω) is complete .l l §5: The Sobolev spaces Wp (Ω) and Wp (Ω) l 1. the space W2 (Ω) is a Hilbert space with the inner product (u. By Theorem 3. It is equivalent to the fact α u } for |α| ≤ l are fundamental sequences in L (Ω).

f −1 ∈ C l (Ω). if u ∈ Wp ˜ p c1 u l Wp (Ω) ≤ u ˜ l ˜ Wp (Ω) ≤ c2 u l Wp (Ω) . (17) C l (Ω) The constants c1 . ˜ ≤ c2 u with the constant c2 depending only on 24 . ˜ Proof: For simplicity. ∂xj ∂yk Ω Here J(x) = det f (x) and the constant c := maxj. Using the properties of weak derivatives (see section 4 Change of varia” l bles“ in § 4). u(y) = u(f −1 (y)). ∂xj ∂yk ˜ Ω dy ˜ ∈ Lp (Ω): = ≤ n n p Ω j=1 ∂u ∂xj ∂xj ∂yk max j=1 n j=1 x∈Ω ∂xj |J(x)|1/p ∂yk ∂u ∂xj p 1/p |J(x)| dx 1/p ∂u ∂xj p 1/p dx Ω ≤ c dx . c2 do not depend on u. there exist the weak derivatives ∂u ˜ = ∂yk Let us check that ∂u ˜ ∂yk p ∂u ˜ ∂yk 1/p n j=1 1 u ∈ Wp (Ω). ˜ ∂u ∂xj . so that l (Ω). Theorem 7 ˜ Let f : Ω → Ω be a diﬀeomorphism of class C l . f C 1 and . they depend only on f and f −1 C l (Ω) . let us prove Theorem 7 in the case l = 1. Thus. we can show that the class Wp (Ω) is invariant with respect to smooth (C l -class) change of variables. We have By section 4 in §4. and ˜ Then.l l For W2 (Ω) another notation H l (Ω) is often used: W2 (Ω) = H l (Ω). ˜ f ∈C l (Ω) . then u = u ◦ f −1 ∈ W l (Ω). |J(x)|1/p |˜(y)|p dy = u u ˜ 1 ˜ Wp (Ω) f −1 Ω |u(x)|p |J(x)| dx ≤ (max |J(x)|) 1 Wp (Ω) Ω |u(x)|p dx.k maxx∈Ω depends only on the norms f Also we have ˜ Ω C 1 (Ω) and f −1 ˜ C 1 (Ω) . Prove the lower estimate in (17) yourself (for this C1 change the roles of u and u in the argument).

ρ→0 l Then for molliﬁcations uρ (x) we have uρ −→ u in Wp (Ω). ∞ Then um ∈ C0 (Ω1 ) and um −→ u as m → ∞ ˜ ˜ ˜ (since um − u Wp (Ω1 ) = um − u Wp (Ω) ). and let ◦ u(x) = ˜ u(x) 0 x∈Ω x ∈ Rn \Ω. u ∈ Wp (Ω1 ). Proposition l Let u ∈ Wp (Ω). ˜ Proof l Wp (Ω) l ∞ By deﬁnition of Wp (Ω).l 2. ˜ ◦ l Wp (Ω1 ) Theorem 8 l Let u ∈ Wp (Ω) and let ◦ u(x) = ˜ u(x) 0 x∈Ω x ∈ Rn \Ω. ˜ ˜ l l l Hence. Wp (Ω) is a subspace in the space Wp (Ω). l Then u ∈ Wp (Ω1 ) for any Ω1 such that Ω ⊂ Ω1 . In particular. 25 . ˜ l u ∈ Wp (Rn ). there exists a sequence um ∈ C0 (Ω) such that ◦ um −→ u as m → ∞. ◦ ◦ l l So. Deﬁnition of Wp (Ω) ◦ Deﬁnition ∞ l l The closure of C0 (Ω) in the norm of Wp (Ω) is denoted by Wp (Ω). We put um (x) = ˜ um (x) 0 x∈Ω otherwise .

by deﬁnition of u and deﬁnition ˜ ˜ ˜ Remark l If u(x) is an arbitrary function in Wp (Ω). Integration by parts Proposition l l Let u ∈ Wp (Ω) and v ∈ Wp (Ω). ˜ ˜ l uρ −→ u in Wp (Ω). |α| ≤ l. So. we have weak derivative ∂ ◦ ∂ α uvm dx = (−1)|α| Ω Ω u ∂ α vm dx.Proof l We have already proved that u ∈ Wp (Rn ). Then ∂ α u ∈ Lp (Rn ). Ω u ∂ α vm dx Ω m→∞ −→ u ∂ α v dx. (18) ∂ α uvdx = (−1)|α| Ω Ω u ∂ α vdx. ˜ then. uρ −→ u in Wp (Ω). Proof ∞ l Let vm ∈ C0 (Ω) and vm → v as m → ∞ in Wp (Ω). u(x) does not have weak derivatives in Rn . and uρ = uρ . 26 . ρ→0 l of molliﬁcation. in general. property 4) and 5) of ∂ ∂ α uρ −→ ∂ α u in Lp (Ω). But. u = u in Ω. So. and u(x) is deﬁned as above. where ◦ 1 p + 1 p = 1. Ω We have ∂ α u (vm − v) dx ≤ ≤ u |∂ α u|p dx 1/p Ω l Wp Ω Ω |vm − v|p dx 1/p l Wp (Ω) vm − v (Ω) → 0 as m → ∞. ˜ ˜ It means that ρ→0 ρ→0 |α| ≤ l. (19) Let us show that ∂ α u vm dx Ω m→∞ −→ ∂ α u v dx. ◦ 3. (See example 2 ˜ l l in Section 7 of §4). Then |α| ≤ l. By Deﬁnition 1 of the α u. in general. Wp (Ω) = Wp (Ω). By ˜ ˜ α u (molliﬁcation of the weak derivative).

Hence. Separability By Vpl (Ω) we denote the linear space of all vector–valued functions v = {vα }|α|≤l such that vα ∈ Lp (Ω). l and J is injective.Ω . We know that Lp (Ω) is a separable Banach space if 1 ≤ p < ∞. ˜ From Theorem 3 it follows that Vpl (Ω) is a closed subspace of Vpl (Ω). We introduce the norm in Vpl (Ω): v l Vp (Ω) = |α|≤l vα p. It follows that Wp (Ω) is separable if 1 ≤ p < ∞. (Since any subspace of some sepa˜ Vp p rable space is also separable. The space Wp (Rn ) Proposition l l ∞ l Wp (Rn ) = Wp (Rn ) . Then Vpl (Ω) is the direct product of a ﬁnite number (equal to the number of multi–indices α with |α| ≤ l) of Lp (Ω). Tending to the limit in (19) as m → ∞. In other words. which is equivalent to the standard norm) to l l (Ω): Vp l J : Wp (Ω) → Vpl (Ω). u ∈ Wp (Ω). we can identify Wp (Ω) with Vpl (Ω). 4. Such an operator is called isometric. it preserves the norm: Ju Vpl (Ω) = ||u||Wp (Ω) . l Now. |α| ≤ l. l 5.Ω . ˜ The range Ran J = Vpl (Ω) is a linear set in Vpl (Ω) consisting of l vector–valued functions v of the form v = {∂ α u}|α|≤l .Ω u (∂ α vm − ∂ α v) dx ≤ ≤ u Ω |u|p dx 1/p Ω |∂ α vm − ∂ α v|p dx 1/p l Wp (Ω) vm − v l Wp (Ω) → 0 as m → ∞. l (Ω) is separable together with V l (Ω).) l l ˜ Since J is isometric. C0 (Rn ) is dense in Wp (Rn ). Then so is Vpl (Ω). Ju = {∂ α u}|α|≤l . consider the transformation J from Wp (Ω) (equipped with the norm ||u||Wp (Ω) = |α|≤l ∂ α u p. Then J is a linear operator. we obtain (18). ◦ 27 .

Next. (R) (R) ∞ l Then uρ ∈ C0 (Rn ) and uρ → u(R) as ρ → 0 in Wp (Rn ). Then uρ (R) n = n p ∂ α u(R) (x) − ∂ α u(x) dx =0 for |x|≤R 1/p 1/p ∂ u α (R) (x) − ∂ u(x) dx ∂ u(x) dx β p 1/p α p −u n) < ε. let u ∈ Wp (Rn ) and let ε ε > 0. . We ﬁnd R so large that u(R) − u W l ( n ) < 2 . β Note that derivatives ∂x ζ |x| R u(R) (x) = 0 if |x| ≥ 2R. we consider molliﬁcation uρ of u(R) . ∞ l l It follows that C0 (Rn ) is dense in Wp (Rn ). β u p ∈ L . ∂ u 1 p (R) Now. |x| R ζ(t) = 0 if t ≥ 2. we ﬁnd ρ so p l Wp ( n) l Wp ( 28 small that uρ (R) − u(R) ε < 2 . Indeed. Thus. We put u(R) (x) = u(x)ζ u(R) (x) = u(x) if |x| ≤ R. ζ(t) = 1 if 0 ≤ t ≤ 1. Then l Let u ∈ Wp (Rn ). = |x|>R ≤ c |β|≤|α| |x|>R → 0 as R → ∞ l This expression tends to zero as R → ∞. x ∈ Rn . p. |β|≤|α| a.Proof Let ζ ∈ C ∞ (R+ ) be such that 0 ≤ ζ(t) ≤ 1. (R) → u as R → ∞ in W l (Rn ). since u ∈ Wp (Rn ). we obtain the inequality ∂ α u(R) (x) ≤ c Then for |α| ≤ l we have ∂ α u(R) − ∂ α u ∂ β u(x) . and so. .e. are uniformly bounded with respect to R ≥ 1. Calculating the derivatives of u(R) (x).

let u ∈ C0 (Ω).6. The Friedrichs inequality Theorem 9 l If Ω is a bounded domain in Rn . 29 .Ω .l. ∂xn 0 Here x = (x1 . xn ). ∞ 1) So.l. Let Q be a cube with the edge d = diamΩ. o |u(x)|p ≤ xn 0 p ∂u (x . We can choose the coordinate system so that Q = {x : 0 < xj < d. by the H¨lder inequality. xn−1 .Ω ≤ (diamΩ) Ω ∂u ∂xn p 1/p dx ≤ (diamΩ)—u—p. such that Ω ⊂ Q. . (22) This is inequality (20) for l = 1.Ω ≤ (diamΩ)l —u—p. xn ) = (x . Obviously. y)dy. . We have proved that u p. (21) ∞ l ∞ Since C0 (Ω) is dense in Wp (Ω). p We integrate both sides of this inequality: Ω |u|p dx = Q |u|p dx p p ≤ +1=p dp/p 0 d dxn Q p ∂u ∂xn p dx = d p Ω ∂u ∂xn dx. . . . . x ∈ Q. n}.Ω . . 1/p (20) —u—p. xn ) ∂xn p dxn 1 Here 1 + p = 1. xn ∂u u(x) = (x . j = 1. it suﬃces to prove (20) for u ∈ C0 (Ω). We extend u(x) by zero to Q\Ω.1.Ω . Then. then for any function u ∈ Wp (Ω) we have ◦ u Here p.Ω = Proof ◦ ∂αu |α|=l p p. y) dy ∂xn xn p/p x 1dy 0 ≤ dp/p ≤ dp/p d 0 ∂u(x .

p.Ω Remark l Inequality (20) is not valid for all u ∈ Wp (Ω). ⇒ Ω |u| dx ≤ d ∂lu ∂xl n dx ≤ dlp —u—p . but u p. we iterate (22): ∂u ∂xn p p Ω dx ≤ d lp Ω p Ω ∂2u ∂x2 n p p dx.2) In order to prove (20) with l > 1.l.l. etc.Ω = 0.Ω = 0. then —u—p. 30 . Example If Ω is a bounded domain and u(x) = Pl−1 (x)(= 0) is a polynomial of order ≤ l − 1.

Let us show that uk − u Wp (Ω) → 0 as k → ∞. Consider a sequence of domains Ωk = x : k−1 x ∈ Ω . We’ll consider the class of domains for which the answer is YES. Domains of star type A natural question: l Can we approximate functions in Wp (Ω) by smooth functions? The answer depends on domain Ω. Then ∂ uk − ∂ u α α p. k l Clearly.§6. We put uk (x) = u( k−1 x). l Let u ∈ Wp (Ω). 31 . This follows from the property of Lp –functions: if u ∈ Lp (Ω).) k Let α be a multi–index with |α| ≤ l.Ω = Ω k−1 k k−1 k |α| ∂ u |α| α k−1 x − ∂ α u(x) dx k ∂ u Ω α 1/p ≤ 1− k−1 x k ≤c u l Wp (Ω) p 1/p dx + →0 as k→∞ + Ω ∂αu p k−1 x − ∂ α u(x) dx k →0 as k→∞ 1/p . Proof Let us use the coordinate system with origin 0. then c |z(x)|≤ k sup Ω |u(x + z(x)) − u(x)|p dx −→ 0. uk ∈ Wp (Ωk ). if any half–line starting at point 0 intersects ∂Ω only in one point. Theorem 10 Let Ω be a bounded domain of star type with respect to a point 0. l We have uk − u p. we can“. l Then C ∞ (Ω) is dense in Wp (Ω).Ω = Ω u p k−1 x − u(x) dx k 1/p −→ 0 as k → ∞. ” Deﬁnition We say that a bounded domain Ω is of star type with respect to a point 0. k ∈ N. k→∞ d (In our case z(x) = − x and |x| ≤ diamΩ = d ⇒ |z(x)| ≤ k . k Then Ωk+1 ⊂ Ωk and Ω ⊂ Ωk .

xn > 0} be a half–ball. and a sequence l uk (x) := uk.l Hence. 32 .ρ ∈ C ∞ (Ω) and uk. so that ρk → 0 as k → ∞. uk −→ u in Wp (Ω). Ω is of star type with l respect to any interior point 0 . Remark Let Ω = {x : |x| < 1. We can choose a sequence {ρk }.ρk (x) tends to u(x) in Wp (Ω): ˜ ∞ (Ω) and uk ∈ C ˜ uk − u ˜ k→∞ l Wp (Ω) −→ 0.ρ −→ uk in Wp (Ω) (since Ω is bounded and Ω ⊂⊂ Ωk ). Suppose that u ∈ Wp (Ω) and u(x) = 0 ε if |x| > 1 − ε.ρ (x). Consider molliﬁcations uk. Then uk ∈ C ∞ (Ω) and uk (x) = 0 if |x| > 1 − 2 for ˜ ˜ suﬃciently large k. Then ρ→0 k→∞ l uk.

and um − u Wp (K+ ) → 0 as m → ∞. We start with the case l = 1. Using construction of Theorem 10 (and Remark after Theorem 10). functions of class Wp ◦ Theorem 11 Suppose that Ω ⊂ Rn is a bounded domain such that Ω is a compact ˜ ˜ manifold of class C 1 . ⇒ vm 1 Wp (K) (24) . 1 follows that vm ∈ Wp (K) (see §4. Example 2). We put 1 vm (x) = um (x) x ∈ K+ um (x . Then there exists a linear bounded extension operator 1 1 ˜ Π : Wp (Ω) → Wp (Ω) such that (Πu) (x) = u(x). we can ﬁnd a sequence um (x) such that um ∈ C ∞ (K+ ). vm ∈ C ∞ (K− ). um (x) = 0 near Σ+ . −xn ) x ∈ K− . 1 Let us show that v ∈ Wp (K) and v 1 Wp (K) = 21/p u 1 Wp (K+ ) . For the norm of vm we have: Then vm ∈ C(K). It is a natural question if we can extend ˜ function ∈ Wp ˜ l (Ω). xn > 0} be a half–ball.§7: Extension theorems l We can always extend a function u ∈ Wp (Ω) by zero and the extended l (Ω) in Ω (⊃ Ω). We extend u to the left half–ball K− = {x : |x| < 1. vm ∈ C ∞ (K+ ). ◦ x ∈ Ω. Proof We proceed in three steps: Step 1 1 Let Ω = K+ = {x : |x| < 1. vm (x) = 0 near ∂K. xn < 0} as follows: v(x) = u(x) x ∈ K+ u(x . −xn ) x ∈ K− . and let u ∈ Wp (K+ ) and u(x) = 0 near Σ+ = {x ∈ ∂K+ : |x| = 1}. Subsection 7. It ◦ vm p 1 Wp (K) = K (|vm (x)|p + | vm (x)|p ) dx K+ = 2 (|um (x)|p + | um (x)|p ) dx = 21/p um 33 1 Wp (K+ ) . (23) Here K = {x : |x| < 1}. Let Ω be a domain in Rn such that Ω ⊂ Ω.

−xn ) x ∈ K− . x ∈ U . ∂vm (x) = ∂xj m→∞ ∂ ∂xj x ∈ K+ (um (x . −yn ) y ∈ K− . n − 1. f (U ∩ Ω) = K+ . . y ∈ K+ . m→∞ 1 Wp (K). f : U → K. m→∞ ∂um (x) ∂xj 1 Since um −→ u in Wp (K+ ). x ∈ K+ wn (x) = − . ˜ We extend u(y) on K− like in step 1: ˜ v (y) = ˜ u(y) ˜ y ∈ K+ u(y . v|Ω = u. . f (U ∩ ∂Ω) = ∂K+ \Σ+ . ∂v ∂xj By Theorem 3.Next. and 1 Consider the function v(x) = v (f (x)). 1 ˜ ˜ v(x) by zero on Ω\U . −xn ) x ∈ K− ∂u(x) ∂xn ∂u ∂xn (x j = 1. v ∈ Wp (K). f ∈ C f −1 ∈ C 1 (K). Step 3 (general case) Let Ω be a bounded domain such that Ω is a compact manifold of class 34 . . where wj (x) = ∂u ∂xj (x ∂u(x) ∂xj x ∈ K+ . Then u ∈ Wp (K+ ) ˜ ˜ and u(y) = 0 near Σ+ . Then v ∈ Wp (Ω). where U is a neighbourhood of 0 ∈ ∂Ω. We extend ˜ ◦ ◦ v 1 ˜ Wp (Ω) = v 1 Wp (U ) ≤ c1 v ˜ 1 Wp (K) ≤ c1 21/p u ˜ 1 Wp (K+ ) ≤ c2 c1 21/p u =c 1 Wp (Ω) . . . f (U ) = K. Relation (23) follows from (24) by the limit procedure Step 2 1 Suppose that u ∈ Wp (Ω) and supp u ⊂ U . −xn )) x ∈ K− . ˜ ◦ 1 As it was proved in step 1. and ˜ v ˜ 1 Wp (K) = 21/p u ˜ 1 Wp (K+ ) . Then v ∈ Wp (U ). it follows that vm −→ v in Lp (K) and m→∞ ∂vm ∂xj −→ wj in Lp (K). 1 We consider the function u(y) = u(f −1 (y)). such that U ⊂ Ω and ∃ diﬀeomorphism ˜ x 1 (U ). f −1 C 1 and on p. The constant c depends on f C 1 . in K and ∂v ∂xj = wj . Thus. there exist weak derivatives vm −→ v in (as m → ∞).

1 ˜ of step 2. Finally. . and vj W 1 (Ω) ≤ cj uj W 1 (Ω) . 35 . v 1 ˜ Wp (Ω) ≤ 1 ˜ Wp (Ω) ≤ cj uj j=1 1 Wp (Ω) ≤c uj j=1 where c = max1≤j≤N {cj }. . ˆ Thus.. UN such that either Uj ⊂ Ω or Uj is a neighbourhood of some point x(j) ∈ ∂Ω. We put Πu = v = j=1 1 ˜ Then v ∈ Wp (Ω). we constructed the linear continuous extension operator 1 1 ˜ Π : Wp (Ω) → Wp (Ω). then uj ∈ Wp (Ω) (since supp ζj ⊂ Uj ) and we can extend uj (x) ˜ by zero to Ω\Ω : uj (x) x ∈ Ω vj (x) = ˜ 0 x ∈ Ω\Ω.. 1 ˜ Wp (Ω) ≤ cˆ u c c= ˆ j=1 cj . fj (Uj ) = K. ◦ C 1 with boundary ∂Ω. we can extend uj (x) to some function vj ∈ Wp (Ω) such that vj (x) = uj (x). . x ∈ Ω. . j=1 ζj (x) = 1. −1 fj N C1 and on p. We represent u(x) as u(x) = Let u ∈ Wp j=1 uj (x). Finally. . ˜ p p If U ∩ ∂Ω = ∅. U2 .N such that 1 If Uj ⊂ Ω.N ∞ ζj ∈ C0 (Rn ). j=1 There exists a partition of unity {ζj (x)}j=1.. U2 . N 1 Wp (Ω) .. j=1 ˜ We can choose the sets U1 . and 1 Wp (Ω) . fj (Uj ∩ ∂Ω) = ∂K+ \Σ+ . By the result ◦ The constant cj depends on fj C1 . N x ∈ Ω. supp ζj ⊂ Uj . N ◦ N j=1 vj (x) N vj . 1 Wp (Ω) ≤ cj u ˆ 1 Wp (Ω) . then uj (x) satisﬁes assumptions of step 2. The constant cj depends on ζj ˆ v Hence. N 1 (Ω). UN so that N Uj ⊂ Ω. . . uj 1 Wp (Ω) = ζj u C1 . fj (Uj ∩ Ω) = K+ . Ω ⊂ N Uj . and ∃ a diﬀeomorphism −1 fj ∈ C 1 (Uj ). where uj (x) = ζj (x)u(x). fj ∈ C 1 (K). . x ∈ Ω. v(x) = vj j=1 = N j=1 uj (x) = u(x). Then (by deﬁnition of such manifolds) there exists a ﬁnite number of open sets U1 .

l Πu Lp (Ω) ≤ c2 u Lp (Ω) . Similar extension theorem is true for unbounded domain Ω ⊂ Rn satisfying the following condition. 3. v = Πu. Ω and Ω. fj (∂Ω ∩ Uj ) = ∂K+ \Σ+ . Now we consider the case l > 1 Theorem 13 Suppose that Ω ⊂ Rn is a bounded domain such that Ω is a compact ˜ ˜ manifold of class C l . c2 depend on l. ∀x ∈ Ω. Then there exists a linear bounded extension operator 1 1 ˜ Π : Wp (Ω) → Wp (Ω) ◦ such that (Πu) (x) = u(x). for x ∈ Ω ◦ and Besides. Ω and Ω. fj (Uj ) = K. i. 36 ≤ c1 u Wp (Ω) . Suppose that there exist bounded open sets {Uj } . ˜ l ˜ Wp (Ω) .Remark It is clear from the proof that for the constructed extension operator Π we have v Lp (Ω) ≤ c u Lp (Ω) . bourhood of a point x j j −1 fj ∈ C 1 (K). ˜ ˜ The constant c depends on p. j ∈ N. Here either Uj ⊂ Ω or Uj is a neighj=1 (j) ∈ ∂Ω and ∃ a diﬀeomorphism f ∈ C 1 (U ). and that N (x) ≤ N < ∞. x ∈ Ω. Moreover. We omit the proof.) Theorem 12 ˜ Under the above conditions on Ω ⊂ Rn . let Ω be a domain in Rn ∞ ˜ such that j=1 Uj ⊂ Ω.. 2. Suppose also that each point x ∈ Ω belongs only to a ﬁnite number N (x) of sets Uj . such that Ω ⊂ ∞ Uj . suppose that the norms fj C 1 (Uj ) −1 and fj C 1 (K) are uniformly bounded for all j ∈ N. (Πu) (x) = u(x). fj (Ω ∩ Uj ) = K+ . (This means that the multiplicity of covering is ﬁnite. Let Ω be a domain in Rn such that Ω ⊂ Ω. Πu ˜ The constants c1 . e. Then there exists a linear bounded extension operator l l ˜ Π : Wp (Ω) → Wp (Ω). p.

v ∈ Wp (K). . The only diﬀerence is that we consider diﬀeomorphisms of class C l .K ≤ cα ∂ α u p. We put + v(x) = l−1 j=0 cj u(x) x ∈ K+ jx ) x ∈ K . In particular. . . ◦ ◦ 37 . . let u ∈ C ∞ (K+ ) and u(x) = 0 near Σ+ . . It suﬃces to assume that Ω is domain of class C 1 (for arbitrary l!) or. 1. fj ∈ Lip1 ). v Wp (K) ≤ c1 u Wp (K+ ) . . j = 0. m = 0. . ◦ Remark 1) The conclusion of Theorem 13 remains true under weaker assumptions on the domain Ω. l Then v ∈ Wp (K). We extend u(x) by zero to Rn \K+ . . are chosen so that ∂mv ∂mv (x . So. l 2) Extension theorems allow us to reduce the study of functions in Wp (Ω) l ˜ to the study of functions in Wp (Ω). even that Ω is Lipschitz domain (it means that −1 diﬀeomorphisms fj . . . cl−1 exist. Next. l − 1. it suﬃces to consider smooth functions u ∈ C ∞ (K+ ).) Hence . . These conditions are equivalent to the following system of linear equations for c0 . u(x . where Ω = K+ and u(x) = 0 near Σ+ .Proof Like in the proof of Theorem 11. c1 . l − 1. . So. . the question reduces to the case. The determinant of this system is not zero. −0). . Moreover. . cl−1 : l−1 j=0 −2j m cj = 1. (We use that v ∈ C ∞ (K+ ). . . v ∈ C ∞ (K− ) and v ∈ C l−1 (K). from the fact that l ∞ ˜ l ˜ C0 (Ω) is dense in Wp (Ω) it follows that C ∞ (Ω) is dense in Wp (Ω). It is easy to check that v ∈ Wp (K). The argument is the same as in proof of Theorem 11. for arbitrary domain Ω. . l Hence. +0) = (x . v(x) = 0 near ∂K. −2 n − The constants cj . |α| ≤ l. we use the covering Ω ⊂ N Uj and the j=1 partition of unity. such constants c0 . . if domain Ω satisﬁes conditions of Theorem 13. l l l Obviously. and ∂ α v p. ∂xm ∂xm n n m = 0. l − 1. .K+ .

Chapter 2: Embedding Theorems Introduction Embedding theorems give relations between diﬀerent functional spaces. l > 0. then a function u ∈ Wp (Ω) is continuous (precisely. We deﬁne the embedding operator J : B1 → B2 . But the fact that for bounded domain Ω. 38 . Wp p embeddings are compact. l Another embedding theorem is that. where the constant c does not depend on u ∈ B1 . Deﬁnition If B1 → B2 and the embedding operator J : B1 → B2 is a compact operator. (This is the Rellich embedding theorem. Some embeddings are obvious. e. if l1 > l2 .) l r More general is the Sobolev embedding theorem : Wp (Ω) → Wq (Ω) under some conditions on p. which takes u ∈ B1 into the same element u considered as an element of B2 . is non–trivial. r (with q > p and r < l). then J B1 →B2 ≤ c. In particul (Ω) → L (Ω). The embedding theorems are very important for the modern analysis and boundary value problems. then we say that B1 is compactly embedded into B2 . if for any u ∈ B1 we have u ∈ B2 and u B2 ≤ c u B1 . x ∈ Ω). Deﬁnition Let B1 and B2 be two Banach spaces. u(x) coincides with a continuous function for a. The fact that B1 → B2 is equivalent to the fact that the embedding operator J : B1 → B2 is continuous linear operator. l l For example. l The trace embedding theorems show that functions in Wp (Ω) have traces on some surfaces of lower dimension. it is obvious that Wp1 (Ω) → Wp2 (Ω). if pl > n. The compactness of operator J is equivalent to the fact that any bounded set in B1 is a compact set in B2 . these lar. If u B2 ≤ c u B1 . ∀u ∈ B1 . We say that B1 is embedded into B2 and write B1 → B2 . l. q.

We always assume that K(x. y)| ≤ ε(ρ) → 0 as ρ → 0 y∈Ω |x−y|≤ρ (K is continuous in x). (1) . e. y)|t dy ≤ M |K(x. x ∈ D. y) is a measurable function on D × Ω. we need some auxiliary material about integral operators. (K is bounded). y)| ≤ L < ∞ sup sup |K(x. y) satisﬁes conditions c) and d) . 1. c) ess d) x. Then. Here 1 ≤ p < ∞. y ∈ Ω. u ∈ Lp (Ω) (1 ≤ p < ∞). b) D for a. We consider the integral operator (Ku) (x) = v(x) = Ω K(x.z∈D |K(x. the operator K is continuous or. y)u(y)dy. e. compact from Lp (Ω) to Lq (D). Lemma 1 If K(x. and K satisﬁes one or several of the following conditions: a) Ω |K(x. even. then K : Lp (Ω) → C(D) is compact operator.§1: Integral operators in Lp (Ω) In order to prove embedding theorems. y) − K(z. x ∈ D. We’ll show that under some conditions on the kernel K(x. where s > 0. by condition c). Proof Let u ∈ Lp (Ω) and v(x) = (Ku) (x). |v(x)| ≤ L |u(y)| dy ≤ L |u(y)| dy 39 p 1/p 1/p 1 dy Ω p = L|Ω|1/p u Ω Ω p.Ω . Let Ω ⊂ Rn and D ⊂ Rm be some bounded domains. where t ≥ 1.y∈Ω for a. y). or from Lp (Ω) to C(D). y)|s dx ≤ N sup x∈D.

Ω . y) satisﬁes conditions a) and b) with p s t some t < p and p + p ≥ 1. Lemma 2 1 1) If p > 1. (5) 4) If p = 1. y)) u(y)dy Ω 1/p ≤ ε(ρ) |u(y)| dy u p. (6) 40 .Ω . then v = Ku ∈ L∞ (D) and v ∞. and K(x.Ω . (4) 3) If p > 1. then. 1 + p = 1. (If p = 1 then (1) is also true with p = ∞. u ∈ L1 (Ω). y) satisﬁes condition a) with t = p . then v = Ku ∈ Lq (D) (for u ∈ Lp (Ω)). y) satisﬁes condition b) with s = q ≥ 1. if |x − z| ≤ ρ (x. u ∈ Lp (Ω). and K(x. and K(x. z ∈ D). Then v = Ku ∈ Lq (D) and v q.) Next. y) − K(z. s t where q ≥ p is deﬁned from the relation q + p = 1. this set is compact in C(D). if u belongs to some bounded set in Lp (Ω): u p.1 1 where p + p = 1. by condition d).D ≤ N 1/q u 1. ≤ ε(ρ)|Ω| (2) From (1) and (2) it follows that. y) satisﬁes condition c). By the Arzela Theorem. then the set of functions {v} is uniformly bounded ( v C(D) ≤ L|Ω|1/p c) and equicontinuous (|v(x) − v(z)| ≤ ε(ρ)|Ω|1/p c. u ∈ Lp (Ω).Ω .D ≤ M 1/p u p. (3) 2) If p = 1. then v = Ku ∈ L∞ (D) and v ∞. It means that the operator K : Lp (Ω) → C(D) is compact.Ω ≤ c. We have v q. and K(x. |v(x) − v(z)| = Ω (K(x. if |x − z| ≤ ρ).D ≤L u 1. |Ω|1/p = 1.Ω .D ≤ M 1/p N 1/q u p. u ∈ L1 (Ω).

**Proof 1) Let p > 1. Using that
**

s q

+

t p

= 1, we obtain:

p

**|K(x, y)u(y)| = |K(x, y)|s/q |u(y)|p/q |u(y)|1− q |K(x, y)|t/p We apply the H¨lder inequality for the product of three functions: o |f1 (y)f2 (y)f3 (y)| dy ≤
**

1 p1

1 p1 1 p2 1 p3

Ω

Ω

|f1 | dy

p1

Ω

|f2 | dy

p2

Ω

|f3 | dy

p3

with

+

1 p2

+

1 p3

= 1.

pq q−p ,

We take p1 = q, p2 = |v(x)| ≤ ≤

p3 = p . Then

Ω

|K(x, y)u(y)| dy

Ω

|K(x, y)| |u(y)| dy

1− p q p,Ω

s

p

1 q

Ω

|u(y)| dy

1 q

p

1 1 −q p

Ω

|K(x, y)| dy

t

1 p

≤M 1/p ( by cond. a) )

≤ M

1 p

u

Ω

|K(x, y)| |u(y)| dy

s

p

.

Note that in the case q = p, we simply apply the ordinary H¨lder o inequality and obtain the same result. We have |v(x)|q ≤ M q/p u ⇒

D q−p p,Ω q−p p,Ω q−p p,Ω Ω

|K(x, y)|s |u(y)|p dy dx

Ω

|v(x)|q dx ≤ M q/p u = M q/p u

D

|K(x, y)|s |u(y)|p dy

D

Ω

|u(y)|p dy

|K(x, y)|s dx

≤N ( by cond b ))

≤ NM This gives estimate (3).

q/p

u

q p,Ω .

**2) Let p = 1 and s = q ≥ 1. If q > 1, we have |K(x, y)u(y)| = (|K(x, y)|q |u(y)|)1/q |u(y)|1/q , Then, by the H¨lder inequality, o |v(x)| ≤ ≤
**

Ω

1 1 + = 1. q q

|K(x, y)u(y)| dy

Ω

|K(x, y)|q |u(y)| dy 41

1/q Ω

1/q

|u(y)| dy

⇒

D

|v(x)|q dx ≤ =

Ω

dx

D Ω

|K(x, y)q | |u(y)| dy

D

u

q/q 1,Ω

|u(y)| dy

q 1,Ω

|K(x, y)|q dx

u

q/q 1,Ω

≤N ( by cond b) with s=q)

≤ N u

.

**This implies (4) (in the case q > 1). If q = 1, then |v(x)| ≤ ⇒
**

D

Ω

|K(x, y)| |u(y)| dy dx

Ω

|v(x)| dx ≤ =

D

|K(x, y)| |u(y)| dy

D

Ω

|u(y)| dy

1,Ω .

|K(x, y)| dx

≤N

≤ N u This implies (4) (in the case q = 1).

**3) Let p > 1, and condition a) is satisﬁed with t = p . Then, by the H¨lder inequality, o |v(x)| ≤ ≤ |K(x, y)u(y)| dy
**

Ω

Ω

|K(x, y)|p dy

≤M 1/p p,Ω .

1/p Ω

|u(y)|p

1/p

≤ M 1/p u This yields (5).

**4) Let p = 1 and K satisﬁes condition c). Then |v(x)| ≤ which gives (6). |K(x, y)| |u(y)| dy ≤ L |u(y)| dy = L u
**

1,Ω ,

Ω

Ω

42

**Remark Statements 1) and 2) mean that the operator K : Lp (Ω) → Lq (D) is continuous, and K K
**

Lp (Ω)→Lq (D) Lp (Ω)→Lq (D)

≤ N 1/q ,

≤ M 1/p N 1/q ,

p > 1;

(7) (8)

p = 1.

**Under conditions of 3) and 4) the operator K : Lp (Ω) → L∞ (D) is continuous and K
**

Lp (Ω)→L∞ (D) Lp (Ω)→L∞ (D)

K

≤ L,

≤ M 1/p ,

p > 1;

(9) (10)

p = 1.

2.

Now, we’ll show that under some additional assumptions on K(x, y), the operator K is compact. We’ll assume that K(x, y) can be approximated by Kh (x, y) (as h → 0) and Kh (x, y) are bounded and continuous in x. Suppose that Kh (x, y), 0 < h < h0 , satisﬁes conditions c) and d) (where L = L(h) and ε(ρ) = ε(ρ; h) depend on h). 1) Suppose that K(x, y) and Kh (x, y), 0 < h < h0 , satisfy conditions of Lemma 2(1) with common t, s, M, N , and |Kh (x, y) − K(x, y)|t dy ≤ mh −→ 0, for a. e. x ∈ D; (11) |Kh (x, y) − K(x, y)|s dx ≤ nh −→ 0, for a. e. y ∈ Ω.

h→0 h→0

Lemma 3

Ω

(12)

D

**Then the operator K : Lp (Ω) → Lq (D) is compact. 2) Suppose that K(x, y), Kh (x, y), 0 < h < h0 , satisfy conditions of Lemma 2(2) with common s = q, N , and |Kh (x, y) − K(x, y)|q dx ≤ nh −→ 0, for a. e. y ∈ Ω. (13)
**

h→0

D

Then the operator K : L1 (Ω) → Lq (D) is compact. 3) Suppose that K(x, y), Kh (x, y) satisfy conditions of Lemma 2(3) with common t = p , M , and condition (11) is satisﬁed with t = p . Then the operator K : Lp (Ω) → C(D) is compact. (Here p > 1.) 43

D → 0 as h → 0. 2) Similarly. 44 . Thus. K : Lp (Ω) → Lq (D) is compact. we agree that Ωn ≡ Ω.D 1/p ≤ mh u p. Hence.Ω . the operator Kh : Lp (Ω) → Lq (D) is also compact. it follows that vh − v ∞. . and Kh − K Lp (Ω)→C(D) ≤ mh 1/p → 0 as h → 0. 1) From conditions (11). vh − v ∞. u ∈ Lp (Ω). n. Hence. 3) From condition (11) with t = p and estimate (5). Here x ∈ Ω or x ∈ Ωm . the operator K maps Lp (Ω) into C(D). So. if p = 1. Now we apply Lemmas 1–3 to the study of the operator (Kj u) (x) = Ω xj − y j u(y)dy. It follows that K : L1 (Ω) → Lq (D) is compact. either D = Ω or D = Ωm .Proof We denote (Kh u) (x) = vh (x) = Ω Kh (x. then K : Lp (Ω) → C(D) is also compact operator. |x − y|n j = 1. Since (Kh u) (x) = vh (x) is uniformly continuous. 3. from condition (13) and estimate (8) it follows that Kh − K L1 (Ω)→Lq (D) 1/q ≤ nh → 0 as h → 0. h→0 Since Kh : Lp (Ω) → C(D) is compact and Kh −→ K in the operator norm. y)u(y)dy. By Lemma 1. (12) and the estimate (7) it follows that Kh − K Lp (Ω)→Lq (D) ≤ mh 1/p nh 1/q → 0 as h → 0. Obviously. . . then v(x) is also uniformly continuous: v ∈ C(D). If m = n. Thus. the embedding C(D) → Lq (D) (for a bounded domain D) is continuous. Hence. the operator Kh : Lp (Ω) → C(D) is compact. . K is the limit in the operator norm of compact operators Kh . where Ωm is some section of Ω by m–dimensional hyper–plane (m < n).

Proof The proof is based on Lemmas 2 and 3. 1) Case 1 < p ≤ n Suppose that conditions 1) are satisﬁed and. We have Ω |Kj (x. then θ ≤ 1.) 2) If p > n. y)|t dy = Ω |xj − yj |t dy ≤ |x − y|tn dy Ω |x − y|t(n−1) . Obviously. Then q p + q s t Since q ≥ p. We put n m n n θ =1− + . if m < n. (If m = n. Since m ≤ n. t ≥ 1. q ≥ p. m .Lemma 4 1) Suppose that 1 ≤ p ≤ n. Then Ω |Kj (x. Thus. n − t(n − 1) 0 dn−t(n−1) r n−1 dr r t(n−1) where κn is the square of the unit sphere Sn−1 in Rn . then Ωm is arbitrary section of Ω by m–dimensional hyper–plane. Let d = diamΩ. n − t(n − 1) 45 . Note that t(n − 1) < n (since t(n − 1) < t(n − 1 + θ) = n) and s(n − 1) < m (since s(n − 1) < s(n − 1 + θ) = m). In particular. n − p < m ≤ n. Thus. then p + p ≥ 1. the numbers t and s satisfy conditions of Lemma xj −y 2(1). y)|t dy ≤ y=x+rξ ξ∈ n−1 B(x) dy |x − y|t(n−1) d = κn = κn . condition a) is satisﬁed with M= κn dn−t(n−1) < ∞. Clearly. then Ωm = Ω. This integral converges since t(n − 1) < n. q ≥ 1 and 1 − n + m > 0. that q ≥ p > 1. Let us check that Kj (x. Ω ⊂ B(x). B(x) = {y ∈ Rn : |x − y| ≤ d}. y) = |x−y|j satisfy conditions a) and b) n with these t and s. then the operator Kj : Lp (Ω) → C(Ωm ) is compact. moreover. t= = n m. p q Then the operator Kj : Lp (Ω) → Lq (Ωm ) is compact. Kj : Lp (Ω) → C(Ω) is compact. Hence. p q n−1+θ p + q s= m = n−1+θ n p m t s + = 1. t < p .

y)|s dx ≤ Bm (y ) dx κm dm−s(n−1) . y)| ≤ |Kj (x. conditions of Lemma 2(1) are satisﬁed and. Clearly. Kjh satisfy condition c). r We put Ψh (r) = Ψ( h ). y. = m − s(n − 1) |x − y |s(n−1) Hence. and 0 ≤ Ψ(r) ≤ 1. 1 Ψ(r) = 0 if 0 ≤ r ≤ 2 . y)|s dx ≤ dx ≤ |x − y|s(n−1) dx . Kj : Lp (Ω) → Lq (Ωm ) is continuous. Ψ(r) = 1 if r ≥ 1. r ∈ [0. condition b) is satisﬁed with N= κm dm−s(n−1) < ∞. Consider the m–dimensional ball Bm (y ) = {x ∈ Πm : |x − y | ≤ d}. Then Ψh (r) = 0 if 0 ≤ r ≤ h . ﬁnally. Let us check 46 . but it does not depend on Ωm (it is one and the same for all sections Ωm of dimension m). Clearly. 2 Ψh (r) = 1 if r ≥ h. be a smooth function such that Ψ ∈ C ∞ ([0. The integral is ﬁnite. condition d) for Kjh is also satisﬁed. |x − y|n Obviously. So. Consider the kernels Kjh (x. since s(n − 1) < m. |x − y |s(n−1) Ωm Ωm Ωm where y is the projection of point y onto the hyper–plane Πm (which contains Ωm : Ωm ⊂ Πm ). we want to ﬁnd the operators Kjh satisfying conditions of Lemma 3. Let Ψ(r). Then Ωm |Kj (x. s. ∀r. m − s(n − 1) The constant N depends on m and on d = diamΩ. Kjh satisfy conditions a) and b) together with K with the same constants t. y) are bounded: |Kjh (x. M and N . Thus. ∞). y) = xj − y j Ψh (|x − y|). therefore. |Kjh (x. For this. ∀x. Ωm ⊂ Bm (y ). Kjh (x. Hence. We want to prove that this operator is compact.Let us check condition b): |Kj (x. Kjh is uniformly continuous in both variables. y)| ≤ Ψh (|x − y|) ≤ |x − y|n−1 1 h n−1 2 = L(h). ∞)). And. y)| . So.

y)|t dy = |x−y|<h Ω |xj − yj |t (1 − Ψh (|x − y|))t dy tn |x − y| ≤1 ≤ = |x−y|<h κn hn−t(n−1) dy |x − y|t(n−1) n − t(n − 1) → 0 as h → 0. now Ωm = Ω. Thus. Hence. and any compact set in Lp (Ωm ) is also compact in Lq (Ωm ). since m > n − p). the operator Kj : Lp (Ω) → Lq (Ωm ) is compact. e. (12) is satisﬁed. Next. n − q(n − 1) . Let us check that condition b) with q s = q is true: |Kj (x. Since Ωm is a bounded domain. . y)|q dx ≤ 47 κn dn−q(n−1) . Then m = n. Thus. so. y) − Kj (x. 3) Case p=1 Condition n − p < m ≤ n means that n − 1 < m ≤ n. condition 1 − n + m > 0 means that p q n 1 − n + n > 0 ⇒ 1 ≤ q < n−1 . Next. y)|s dx = ≤ = |xj − yj |s (1 − Ψh (|x − y|))s dx sn |x−y|<h |x − y| dx |x − y |s(n−1) {x∈Ωm :|x−y |<h} Ωm Ωm κm hm−s(n−1) m − s(n − 1) → 0 as h → 0. condition 1 − n + m > 0 is p p true. y)|q dx = |xj − yj |q dx ≤ |x − y|nq dx . |x − y|(n−1)q Ω Ω Ω The integral is ﬁnite since (n − 1)q < n ⇒ Ω |Kj (x. It follows that the operator Kj : Lp (Ω) → Lq (Ωm ) is compact. |Kjh (x.condition (11): |Kjh (x. then we apply the result that Kj : Lp (Ω) → Lp (Ωm ) is compact (i. y) − Kj (x. we apply 1) with p = q. then Lp (Ωm ) → Lq (Ωm ) (if q < p). 2) If 1 ≤ q < p. (Recall that we assumed q ≥ p > 1). Then all conditions of Lemma 3(1) are satisﬁed. (11) is true.

y) − Kj (x. y)|p dy ≤ κn dn−p (n−1) dy ≤ < ∞. Indeed. then p < n . conditions of Lemma 3(3) are satisﬁed.Also. n − p (n − 1) |x − y|p (n−1) 1 p Ω Ω 1 1 1 Since p > n. then the operator Kj : Lp (Ω) → Lq∗ (Ωm ) is continuous (but not compact). y)|q dx ≤ κn hn−q(n−1) → 0 as h → 0. 48 . y) with it) satisfy condition a) with t = p : |Kj (x. y)|p dy |x−y|≤h κn hn−p (n−1) n − p (n − 1) → 0 as h → 0. dy |x − y|p (n−1) Thus. n m =0 + p q∗ ⇔ q∗ = mp n−p . Next. y) − Kj (x. It follows that the operator Kj : Lp (Ω) → C(Ω) is compact. the kernels Kj (x. where 1− Without proof. Ω |Kjh (x. y) (and Kjh (x. p = 1 − Hence n > p (n − 1). 4) Case p ¿ n Let us check that conditions of Lemma 3(3) are satisﬁed. >1− ≤ ≤ 1 n = n−1 n . the operator Kj : L1 (Ω) → Lq (Ω) is compact. Lemma 5 If 1 < p < n. n − q(n − 1) Ω By Lemma 3(2). condition (13) is satisﬁed: |Kjh (x. n − p < m ≤ n.

∞ Then for any u ∈ C0 (Ω) we have u(x) = Ω E(x − y)( u)(y)dy. and let uk ∈ C0 (Ω). . The integral representation for functions in Wp (Ω) Lemma 6 1 Let Ω ⊂ Rn be a bounded domain. e. let u ∈ Wp (Ω). Consider the fundamental solution of the Poisson equation E(z) = δ(z): E(z) = 1 − κn (n−2)|z|n−2 1 κ2 ln |z| n>2 n = 2. ∂zj κn |z|n j j ∂ Then. Then ◦ u(x) = 1 κn n j=1 Ω xj − yj ∂u dy. assume that u ∈ C0 (Ω). Let u ∈ Wp (Ω). |x − y|n ∂yj for a. . n κn Ω |x − y| ∂yj j=1 k→∞ 49 . uk −→ u in Wp (Ω). x ∈ Ω. n By Deﬁnition 1 of weak derivatives. For uk we have n 1 xj − yj ∂uk uk (x) = dy. 1 ∞ 1 2) Now. The function E(z) has weak derivatives 1 zj ∂E(z) = . (14) Proof ∞ 1) First. n.1 §2: Embedding theorems for Wp (Ω) ◦ 1 1. j = 1. ∂yj ∂yj Then. . |x − y|n ∂yj ∞ ∀u ∈ C0 (Ω). we have x −y n Ω E(x − y)( u)(y)dy = − n j=1 Ω j=1 Ω ∂E(x − y) ∂u dy. . ∂yj E(x − y) = − κ1 |x−y|n . 1 u(x) = κn ◦ xj − yj ∂u dy.

|x − y|n ∂yj 1 ∀u ∈ Wp (Ω). q<q 1 2) What does it mean that Wp (Ω) → Lq (Ωm ) in the case m < n ? 1 A function u ∈ Wp (Ω) is a measurable function in Ω. ◦ 1 2. j=1 ∂y By Lemma 4.Ωm ≤c u 1 Wp (Ω) .k Thus. by the limit procedure as k → ∞ we obtain: know that uk −→ u in Lp (Ω) and k→∞ ∂uk k→∞ ∂u ∂yj −→ ∂yj in Lp (Ω). p q 1 2) If p > n. ∞ First we consider u ∈ C0 (Ω). This linear operator can be extended by continuity to a continuous operator 1 T : Wp (Ω) → Lq (Ωm ). then Wp (Ω) is compactly embedded into C(Ω).) Hence. then this embedding is compact. If ∗ . We k→∞ Kj ( ∂uk ) −→ ∂yj ∂u Kj ( ∂yj ) in Lp (Ω). q < ∞ and q ≤ n−p = q ∗ . each operator Kj is compact from Lp (Ω) to Lp (Ω). In the case 1 − n + m > 0. Then u(x) = 1 κn n j=1 Ω xj − yj ∂u dy. Ωm is a set of measure zero. it can be changed on any set of measure zero. ◦ 1 1) If 1 ≤ p ≤ n. ∞ ∀u ∈ C0 (Ω). We have the estimate ◦ ◦ ◦ ◦ Tu q. m > n − p. (since Kj : Lp (Ω) → Lp (Ω) is continuous operator. then Wp (Ω) p q is embedded into Lq (Ωm ). ◦ Comments 1) Let us distinguish the case m = n (Ωm = Ω): np 1 If 1 ≤ p ≤ n. then Wp (Ω) → Lq (Ω). this embedding is compact. where Ωm = Ω (if m = n ) and Ωm is any section of Ω by m–dimensional plane (if m < n). 50 . Embedding theorems for Wp (Ω) ◦ Theorem 1 Let Ω ⊂ Rn be a bounded domain. uk = κ1n n Kj ∂uj . q < ∞ and 1 − n + m ≥ 0. ∞ ∞ Then T : C0 (Ω) → C0 (Ωm ) is a linear operator. and put T u = u|Ωm .

k→∞ 1 Wp (Ω) k. m > n − p. Here B1 . Then Dj : Wp (Ω) → Lp (Ω) is continuous operator. Proof of Theorem 1 ∂u 1 By Dj we denote operators Dj u = ∂xj . 2) Let p > n. . e. the p q operator Kj : Lp (Ω) → Lq (Ωm ) is continuous.1 ∞ Let u ∈ Wp (Ω).Ωm ≤ c uk − uj W 1 (Ω) −→ 0. then. j = 1. For p = 1 – without proof. the embedding operator n J = κ−1 n j=1 1 Kj Dj : Wp (Ω) → Lq (Ωm ) ◦ is continuous. w = T u. κ−1 n j=1 1 Kj Dj : Wp (Ω) → C(Ω) ◦ 51 . There exists limit T uk −→ w in Lq (Ωm ). . Then ∃ uk ∈ C0 (Ω). ◦ uk − u k→∞ Then T uk − T uj q. So.) If p > 1 and 1 − n + m = 0 (i. the embedding operator n J= κ−1 n j=1 1 Kj Dj : Wp (Ω) → Lq (Ωm ) ◦ is compact. Then representation (14) can be written as ◦ u= 1 κn n j=1 Kj Dj u. Then p q conditions of Lemma 4(1) are satisﬁed. B2 . then A2 A1 : B1 → B3 is compact. by Lemma 4(2). p Hence {T uk } is a Cauchy sequence in Lq (Ωm ). the embedding operator n J= is compact. Hence.j→∞ −→ 0. . (We use the fact that if A1 : B1 → B2 is continuous operator and A2 : B2 → B3 is compact operator. n. By deﬁnition. operators Kj : Lp (Ω) → C(Ω) are compact. Then. Hence. operator Kj : Lp (Ω) → Lq (Ωm ) is compact. (15) 1) Suppose that 1 ≤ p ≤ n. B3 are Banach spaces. q = q ∗ ). q < ∞ and 1 − n + m > 0. . Hence. by Lemma 5..

Proof ˜ ˜ ˜ Let Ω ⊂ Rn be a bounded domain such that Ω ⊂ Ω. (For example. 1) If 1 ≤ p ≤ n. but they do not depend on Ωm (they are one and the same for any section Ωm ). Embedding theorems for Wp (Ω) Theorem 2 Let Ω ⊂ Rn be a bounded domain of class C 1 . q. m > n − p. 52 . Ω is a ball of suﬃciently large diameter. (17) depend only on diamΩ. then Wp (Ω) is compactly embedded into C(Ω). we have n u C(Ω) ≤ κ−1 n n j=1 Kj Lp (Ω)→C(Ω) ∂j u p. 1 u ∈ Wp (Ω). In the case 1 − n + m > 0. 1 3. If u ∈ Wp (Ω). c= Π .Remark 1) Under conditions of Theorem 1(1).Ωm ≤ κ−1 n j=1 n Kj Lp (Ω)→Lq (Ωm ) Dj u Lp (Ω) ≤ c ∂j u j=1 p. (17) Using the estimates from Lemma 4 it is easy to see that the constants in estimates (16). we have the estimate n u q. this p q embedding is compact. (16) 2) Under conditions of Theorem 1(2). Then both statements 1 of Theorem 1 are true for Wp (Ω). 1 u ∈ Wp (Ω). then p q 1 Wp (Ω) is embedded into Lq (Ωm ).) By Theorem 11 (Chapter 1). there exists a linear continuous extension ope1 1 ˜ 1 1 ˜ rator Π : Wp (Ω) → Wp (Ω). then v = Πu ∈ Wp (Ω).Ω ◦ ≤ c u 1 Wp (Ω) .Ω ≤ c ∂j u j=1 p. m. and ◦ ◦ v 1 ˜ Wp (Ω) ≤c u 1 Wp (Ω) . q < ∞ and 1 − n + m ≥ 0. 1 2) If p > n.Ω ◦ ≤ c u 1 Wp (Ω) . p. n.

and Ωm is the section of Ω by the same Πm . this embedding is compact. then Π is bounded 1 ˜ set in Wp (Ω). ˜ R : Lq (Ωm ) → Lq (Ωm ) is the restriction operator. ◦ u 1 Wp (Ω) (18) a) Let 1 − n p + m q > 0.Ω .Ωm = ≤ ≤ RJΩ Πu ˜ JΩ Πu ˜ JΩ ˜ ◦ q. let JΩ : Wp (Ω) → Lq (Ωm ) be 1 ˜ ˜ the embedding operator and let JΩ : Wp (Ω) → Lq (Ωm ) be the em˜ ◦ ◦ 1 1 ˜ bedding operator. Hence. by ˜ Theorem 1.1 ˜ ˜ 1) Under conditions of part 1). u q. Let 1 be some bounded set in Wp (Ω). m n if 1 − p + q > 0. Π : Wp (Ω) → Wp (Ω) is the extension operator. where ˜ ˜ Πm is m–dimensional plane. ◦ ⇒ u q.Ωm = 0.Ωm ≤ c u 1 Wp (Ω) . Π ◦ ◦ ˜ is compact in C(Ω).Ωm 1 ˜ ˜ Wp (Ω)→Lq (Ωm ) Π 1 1 ˜ Wp (Ω)→Wp (Ω) =c1 =c2 1 ∀u ∈ Wp (Ω). By similar arguments. If is a bounded set in W 1 (Ω). by Theorem 1(2). ◦ . 1 Compare (18) with estimate (16): in the case u ∈ Wp (Ω) we can n estimate u q. Now it is impossible.) ◦ 1 ˜ Then Π = {v = Πu : u ∈ } is a bounded set in Wp (Ω). 1 ˜ 2) Under condition p > n. Hence.Ωm by the norms of derivatives j=1 ∂j u p. 1 ˜ ˜ b) Let 1 − n + m = 0. by Theorem 1. (It is clear for u = const = 0 : u q. Comments 1 1) Under conditions of Theorem2(1). ˜ 1 We have the estimate for all u ∈ Wp (Ω). (Here Ωm ⊂ Πm . Then.) 53 ◦ p is compact in C(Ω).Ω ˜ q. Wp (Ω) is compactly embed˜ ded into C(Ω). we show that 1 embedding Wp (Ω) → Lq (Ωm ) is also continuous. Wp (Ω) compactly embedded into Lq (Ωm ). this set is compact in Lq (Ωm ). but ∂j u ≡ 0. Then is compact in Lq (Ωm ) (because functions in are restrictions of functions in Π 1 back to Ω).Wp (Ω) → Lq (Ωm ). In this case embedding Wp (Ω) → Lq (Ωm ) is p q continuous (but not compact). Then JΩ = RJΩ Π.

m. Examples. p q Example. 3) If Ω is bounded and p > n. p < ∞. Let u(x) = |x|λ with 1 − 1 Then u ∈ Wp (Ω). n p + m q ≥ 0. n. 1 ∀u ∈ Wp (Ω). (17) depend only on diamΩ and on p. then Wp (Ω) → C(Ω). since n − 1 + p(λ − 1) > −1 ⇔ λ > 1 − 54 . n p < λ < −m.2) Similarly. then Wp (Ω) → Lq (Ωm ). p ≤ q < ∞ and 1 − 1 Wp (Ω) → Lq (Ωm ). All conditions in Theorems 2. (19) The constants in estimates (18). q Ω | u|p dx ≤ c Ω |x|p(λ−1) dx 1 = cκn 0 r n−1+p(λ−1) dr n . but u ∈ Lq (Ω). 4. q. Now 1) does not follow from 2). Comments.) Let us formulate the analog of Theorem 2 for unbounded domain. Then 1) If p ≥ 1. Let Ω = {x ∈ Rn : |x| < 1}.). on the properties of ∂Ω. 1 1) If 1 − n + m < 0. Theorem 3 Suppose that Ω ⊂ Rn is unbounded domain satisfying conditions of Theorem 12 (Chapter 1). 1 2) If p > n. we have the estimate u C(Ω) ≤ C u 1 Wp (Ω) . under conditions of Theorem 2(2). but not compact. then 1) follows from 2). then Remark 1) In Theorem 3 embeddings are continuous. / λ−1 Indeed. m > n − p. (While constants in estimates (16). so. 2) In part 1) we have condition q ≥ p (we don’t need this condition in Theorem 2. 3 are precise. | u| ≤ c |x| . (19) depend on Π and.

then Wp (Ω) → Lq (Ω). Example ∞ Let u ∈ C0 (Rn ) and let x(k) be a sequence of points x(k) ∈ Rn such 55 . Consider u(x) = ln | ln |x||. Let − n < λ < − n . | u(x)| ≤ |x|| ln |x|| . 5) For unbounded domains embeddings from Theorem 3 are not compact. We have q ∗ > p. Then / Ω | u(x)|n dx ≤ dx Ω |x|n | ln |x||n 1/e 0 1/e = κn = κn < ∞. since m > n − p. Ω |u|p dx < ∞. Wn ∞ n (Here q ∗ = ∞. Example Let Ω = {x ∈ Rn : |x| > 1} . since m − 1 + qλ < −1(⇔ λ < − Here Ωm is a section of Ω by some m–dimensional plane Πm such that point 0 ∈ Πm . But for p = n > 1. 1 (Ω) → C(Ω) and even W 1 (Ω) → L (Ω) . but u ∈ Lq (Ω). / m 3) The critical exponent“ q ∗ is deﬁned by the relation 1 − n + q∗ = 0. e 1 1 Then u ∈ Wn (Ω). u(x) = |x|λ . 1 4) If p = n = 1.Also. However. if p < q. b). q = ∞. Then u ∈ Wn (Ω). Then q p 1 u ∈ Wp (Ω). but u ∈ L∞ (Ω). Indeed. then 1 Wp (Ω) → Lq (Ωm ) for all q < ∞ (if Ω is bounded) and all p ≤ q < ∞ 1 (if Ω is unbounded). Check yourself. Also. then Wp (Ω) → C(Ω). Ω = (a. but not for q > q ∗ . 1 2) For unbounded domains. Here p < n. then any function u ∈ W1 (Ω) is absolutely continuous. 1 Wp (Ω) → Lq (Ωm ) for q ≤ q ∗ . p ” mp (q ∗ = n−p ). This follows from Theorem 5 (Chapter 1). Ω 0 r n−1 dr r n | ln r|n dr r| ln r|n 1 |u(x)|n dx < ∞. If p ≥ n.) Example Let Ω = x ∈ Rn : |x| < 1 . If p > n. |u(y)|q dy = Ωm Ωm |y|qλ dy 1 = κm 0 r m−1+qλ dr m ).

p < n. But {v } is not compact in L (Rn ). 56 . m > n − p. dimΓ = m. 1 6) For bounded domains Ω and q = q ∗ embedding Wp (Ω) → Lq∗ (Ω) is not compact. diﬀemorphisms Important case Γ = ∂Ω (then also ΩΓ = ∂Ω). For q < q ∗ this embedding is compact. Conditions: m = n − 1 > n − p ⇒ p > 1. But ukj C( n ) = u C( n ) = 0. Theorem 4 Let Ω ⊂ Rn be a bounded domain of class C 1 . n suppose that there exists a subsequence ukj such that ukj −→ u0 in j→∞ j→∞ −n 5. Then the set 1 {uk } is bounded in Wp (Rn ). 1 Wn (Ω) → Lq (∂Ω). dimΓ = n − 1. If n m 1 − p + q > 0. but {wk } is not compact in Lq∗ (Ω). Let 1 ≤ p ≤ n. and vk (x) = k p u x . Example ∞ 1 Let u ∈ C0 (Rn ). Contradication. u ∈ C0 (Rn ). Indeed. ∀q < ∞. wk (x) = k p −1 u(kx). ∀q ≤ q ∗ . then u0 (x) ≡ 0. Let Γ ⊂ Rn be a manifold p q 1 of class C 1 . Example ∞ Ω = {x : |x| < 1}. If n = p > 1. Then Wp (Ω) → Lq (ΩΓ ). (Obviously. 1 ≤ q < ∞ and 1 − n + m ≥ 0. then q ∗ = ∞. Check this yourself. Here p > n. then Wp (Ω) → Lq (∂Ω). Then 1 {wk } is bounded in Wp (Ω).that x(k) → ∞ as k → ∞. p q n−p 1 If q ∗ < ∞ (1 < p < n).) Uj . Without proof (The proof is based on Theorem 2 and using of covering fj and patition of unity. Embeddings on submanifolds Instead of the section of Ω by m–dimensional planes we can consider sections of Ω by some m–dimensional manifolds.) But the set {uk } is not compact in C(Rn ). Then vk ∈ Wp (Rn ) and {vk } k 1 (Rn ). then this embedding is compact. Let ΩΓ = Γ ∩ Ω. uk Wp ( n ) = 1 u Wp ( n ) = const. We put uk (x) = u(x − x(k) ). Since ukj −→ 0 in C(Ω) for any bounded domain Ω (simply ukj ≡ 0 in Ω for suﬃciently large j). 1 − n + n−1 ≥ 0 ⇔ q ≤ (n−1)p = q ∗ . 1 C(Rn ). Thus. the is bounded in Wp p k 1 embedding Wp (Rn ) → Lp (Rn ) is not compact.

l−j l−j−1 Jj : Wqj (Ω) → Wqj+1 (Ω). |α| ≤ l − j − 1. 1) If p ≥ 1. . then at least one of Jj is compact (at least for n n one index j we have 1 − qj + qj+1 > 0).Ω ≤ c ∂αu 1 Wqj (Ω) ≤c u ˜ l−j Wqj (Ω) for all α with |α| ≤ l − j − 1. l − r − n + n ≥ 0. 1. .l §3: Embedding theorems for Wp (Ω) Theorem 5 Let Ω ⊂ Rn be a bounded domain of class C 1 . . Each operator Jj is continuous . . . l−j l−j−1 l−j It follows that Wqj (Ω) → Wqj+1 (Ω). . . then Wp (Ω) → C r (Ω) and this embedding is compact. If θ > 0. j = 0. qs = q and 1 − qj + qj+1 ≥ 0. l 2) If p(l − r) > n. . . J1 J0 . . q1 . then this embedding is p q compact. We have: 3 1 0 r l−s l−2 l−1 l l Wp (Ω) = Wq0 (Ω) → Wq1 (Ω) → Wq2 (Ω) → . s − 1. 1 By Theorem 2(1). then J is also continuous. . . s − 1. Indeed. 57 . qs are deﬁned p q p q n n uniquely from the equations 1 − qj + qj+1 = 0. . then p q l r Wp (Ω) → Wq (Ω). αu ∈ L then ∂ qj+1 (Ω) . J J J Js−1 l r ⇒ The embedding operator J : Wp (Ω) → Wq (Ω) is represented as J = Js−1 . such numbers exist (but they are not unique). . Wqj (Ω) → Lqj+1 (Ω). n n q0 = p. 0 ≤ r < l. . . 1 ≤ q < ∞. . 1 1 Then ∂ α u ∈ Wqj (Ω) for |α| ≤ l − j − 1. Jj is a continuous operator. If n n θ = s − p + q > 0. and ∂αu qj+1 . . In this case J is also compact. j = 0. . qs such that qj ≥ 1. Proof 1) We put s = l − r and ﬁx the numbers q0 . If l − r − n + n > 0. l−j−1 ⇒ u ∈ Wqj+1 (Ω) and u l−j−1 Wqj+1 (Ω) ≤c u l−j Wqj (Ω) . Such numbers exist due to condition s − n + n ≥ 0. let u ∈ Wqj (Ω). We denote the embedding operator by Jj . → Wqs (Ω) = Wq (Ω). Since Wqj (Ω) → Lqj+1 (Ω). If l − r − n + n = 0. . then q0 .

the embedding p 1 l Wp (Ω) → C(Ω) is compact. It follows that Wp (Ω) → C l−1 (Ω) and this l (Ω). where J1 : Wp (Ω) → Wq (Ω) (J1 is compact by part r+1 → C r (Ω) (J is compact by case a). then q ∗ = ∞. In l (Ω) → L (Ω) (for l ≥ 1) is compact. r < l. In this case Wp (Ω) → Lq (Ω) ∀q < ∞ (and this embedding is compact). However.2) Let p(l − r) > n ⇔ l − r − n > 0 p Case a) r = l − 1 l − l + 1 − n > 0 ⇔ p > n.) Case b) r < l − 1 Then there exists a number q such that q > n and l−(r+1)− n + n > 0. (If u ∈ Wp p |α| ≤ l − 1. for the proof we need another l integral representation for u ∈ Wp (Ω) (including derivatives of higher order). Embedding p l Wp (Ω) → Lq (Ω) is compact for q < q ∗ . embedding Wp p Remarks 1 1) The embedding theorem for Ωm with m < n (Wp (Ω) → Lq (Ωm )) can l be also generalized for Wp (Ω). e. l − r − n =: ε > 0. Particular cases 1) Let r = 0. ◦ 58 . l Hence. then Wp (Ω) → C(Ω) and this embedding is compact. 1) of Theorem 5) and J2 : Wq 2 since q > n). and continuous for q = q ∗ . l 2) If pl = n. particular. n < q < 1−ε . l 2) The embedding theorems for Wp (Ω) can be also generalized for the case of unbounded domains. J : Wp (Ω) → C r (Ω) is compact. pl < n.) l Then we can represent the embedding operator J : Wp (Ω) → C r (Ω) l r+1 as J = J2 J1 . l r 4) Let q = p. p q (Indeed. then q ∗ < ∞. We can ﬁnd q > n such that 1 − n < ε. The critical exponent q ∗ is deﬁned from the conditinp n on l− n + q∗ = 0 ⇔ q ∗ = n−lp . p q n i. Then embedding Wp (Ω) → Wp (Ω) is compact. l But Wp (Ω) → L∞ (Ω). then ∂ α u ∈ W 1 (Ω) → C(Ω) for embedding is compact. l 3) If pl > n. By Theorem 2(2). Since pl < n..

k = 1. . There is a one–to–one correspondence of elements x ∈ X and coordinates N Proof From (1) it follows that N N x ≤ ξ k=1 k xk ≤ ξ k=1 k 2 1/2 N 1/2 xk k=1 N 2 2 . It means that there exists a system of linear independent elements x1 . e. ∀ y ∈ X. 2 1/2 (1) N k ξ = ξ k k=1 . . that this k=1 ξ functional has all properties of the norm. xN : N x= k=1 ξ k xk . . x ≤ γ x . Then x is a continuous function of ξ on the closed bounded set ξ ∈ CN : |ξ| = 1 in CN . Since x > 0. . . Therefore. M.. ξ k ∈ C . N. We denote x = . e. the space X with this norm is complete). . Proposition Any other norm x on X is equivalent to x . Check yourself. X is a Banach space with respect to this norm (i. ∀ x ∈ X. then by the Weierstrass Theorem. .l Equivalent norms in Sobolev spaces Wp (Ω) (lecture by prof. γ= k=1 xk > 0. all norms on X are equivalent to each other. Now we restrict the continuous function x to the unit sphere x = 1. Thus. Then y = y y y ≥ β y . dim X = N < ∞. 1/2 i. such that any x ∈ X can be represented as a linear combination of x1 . . . (3) 59 . . . From (2) and from the triangle inequality it follows that x − x ≤ x−x ≤γ x−x . (2) Now. xN ∈ X. let us prove the opposite inequality. The mapping x → ξ is an isometric isomorphism of X and CN (with the standard norm). Let us check that the function x is continuous on X with respect to x . Finitedimensional linear spaces and norms in these spaces Let X be a linear space. x ≥ β > 0 for x = 1. . . x x . Birman) 1.

2) ϕ(cx) = |c|ϕ(x). 1 ≤ p < ∞. ” l The standard norm in Wp (Ω). η ∈ CN . Ω ⊆ Rd . l ∈ N.l 2. Thus. l which is equivalent to the standard one. Example l X = Wp (Ω). 60 . ∀ x ∈ X. Such new norms in Wp (Ω) are trivial. (5) Then we can rewrite (4) as u p l Wp (Ω) = η p n . ∀ x ∈ X. ϕ(x) = Ω u(x)dx . then (6) will automatically deﬁne some norm in Wp (Ω). (4) η p N = s=1 |η s |p . is u l Wp (Ω) Let N be the number of all multi–indices α with |α| ≤ l. Example The norm u Lp (Ω) + max1≤|α|≤l ∂ α u Lp (Ω) is equivalent to the standard l norm in Wp (Ω). a seminorm ϕ has all properties of the norm besides one: from ϕ(x) = 0 it does not follow x = 0. if 1) 0 ≤ ϕ(x) < ∞. Trivial“ equivalent norms in Wp (Ω). |α| ≤ l. ∀ c ∈ C. In CN we introduce the norm of lp –type by the formula N = ∂αu |α|≤l p Lp (Ω) 1/p . (6) where η = ∂ α u Lp (Ω) . Give yourself several examples of new trivial“ norms in ” l Wp (Ω). The notion of seminorm. 3. Deﬁnition A functional ϕ on a linear space X is called a seminorm on X. 3) ϕ(x1 + x2 ) ≤ ϕ(x1 ) + ϕ(x2 ). This functional is equal to zero for any l u ∈ Wp (Ω) with zero mean value. If we replace the norm (5) in relation (6) by any other (equivalent!) norm l of vector η in CN .

by property 5).l 4. Assume that Ω ⊂ Rn is bounded and ∂Ω ∈ C 1 . therefore. Then it follows l that u ∈ Pl . Besides. and. Then there exists l a sequence {um }. functional (7) is homogeneous and satisﬁes the triangle inequality. which is equivalent to the standard norm. m l l−1 Since the embedding Wp (Ω) → Wp (Ω) is compact it follows from (10) l−1 that there exists a subsequence vmj . (8) is not true. by (9). Theorem l Let ϕ be a functional on Wp (Ω) satisfying conditions 1) – 5). it suﬃces to check that u l Wp (Ω) ≤ C—u—Wp (Ω) . if —u—Wp (Ω) = 0. —u—Wp (Ω) = l ∂αu p Lp (Ω) |α|=l + ϕ(u)p (7) Proof Obviously. Let ϕ be a seminorm on Wp (Ω) which satisﬁes properties: 4) ϕ(u) ≤ c u nuous in l Wp (Ω) l (Ω). u=0. then u = 0 (ϕ is non–degenerate on the l subspace Pl ⊂ W2 (Ω)). conti- 5) If u ∈ Pl and ϕ(u) = 0. (8) Suppose the opposite. Taking account of property 4). Then for any C > 0. and. p l Wp (Ω) . General theorem about equivalent norms in Wp (Ω). which converges in Wp (Ω) to some l−1 v0 ∈ Wp (Ω): vmj − v0 l−1 Wp (Ω) → 0 as j → ∞ (12) 61 . um ∈ Wp (Ω) such that m—um —Wp (Ω) ≤ um l We put vm = um um W l (Ω) . (10) (11) —vm —Wp (Ω) ≤ l 1 → 0 as m → ∞. ϕ(u) = 0. vm l Wp (Ω) = 1. functional l (7) is a norm on Wp (Ω). Next. (9) Then. then ∂ α u = 0 for ∀ α with |α| = l. Then the functional 1/p l deﬁnes the norm in Wp (Ω). Thus.) Wp (It means that ϕ is bounded. By Pl we denote the class l of all polynomials in Rn of order ≤ l − 1. l l u ∈ Wp (Ω).

v0 = 0. we have vmj −→ v0 as j → ∞. Mention that. (13) Since the operator ∂ α is closed in Lp (Ω). By (14) and property 4). Together with (14) this contradicts to (10). Control question Where did we use that Ω is bounded and ∂Ω ∈ C 1 ? 5. By property 5). Conditions 1) . Let ϕ(u) = ω u(x)dx . For ω = Ω and p = 2 this is the classical Poincare inequality. Exercise In the case p = 2. Let l ≥ 2 and ϕ(u) = u Theorem. Thus. (14) From (11) it follows that ϕ(vmj ) → 0 as j → ∞. we did not use any explicit construction and we did not obtain any upper bound for the constant C. Then by (12) and (13). ω ⊆ Ω. 2. Then the Theorem implies that u p Lp (Ω) —u—Wp (Ω) = l ∂αu p Lp (Ω) + u |α|=l p Lp (Ω) (15) ≤C Ω | u|p dx + p u(x)dx ω . However. ϕ(v0 ) = 0. By 1/p is equivalent to the standard one. conditions 1) – 5) are satisﬁed. ∂ α v0 = 0 for ∀ α with |α| = l. Let l = 1. 1. Examples. l = 2. Clearly. Now Pl consists of constants. 0 < |α| < l.2) are obviously satisﬁes. we have proved rather general theorem. ω is a measurable set such that mesn ω > 0. 62 . is estimated by the norm (15). ϕ(vmj ) → ϕ(v0 ) as j → ∞.From (11) it follows that ∂ α vmj j→∞ Lp (Ω) −→ 0 for ∀ α with |α| = l. Additions. It follows that ∂ α u Lp (Ω) . which in particular cases implies a number of concrete inequalities (proved before by special tricks). in the proof of inequality (8). v0 ∈ Pl . l Wp (Ω) v0 ∈ Pl . prove this estimate using Fourier transform. the norm Lp (Ω) .

This generalizes and strengthens the Friedrichs inequality |u|2 dx ≤ C | u|2 dx + |u|2 dS . . Properties 1) – 5) are satisﬁed. 63 . if u ∈ P2 . it is always so. (17) Condition 4) follows from the trace embedding theorem. if Γ = ∂Ω. (16) We have to check condition 5). Thus. Let l = 2. Then. Let us check 5): ϕ(u) = 0 ⇔ u|Γ = 0. xn . the norm (7) with l = 2 and such ϕ(u) is equivalent to the 2 standard norm in Wp (Ω). By Theorem (on equivalent norms) we obtain p u p Lp (Ω) ≤C Ω | u|p dx + udS Γ . it follows that u = 0. . l Wp (Ω) ∂Ω i. Condition 4) follows from the estimate |u|p dS ≤ C u p . x1 . In the case where Γ does not lie in some plane. i. then u is orthogonal in L2 (ω) to the basis in P2 . If ϕ(u) = 0. from the trace embedding theorem. Γ ⊂ ∂Ω. We put ϕ(u) = Γ udS . Ω Ω ∂Ω 4. then condition u|Γ = 0 and u = 0 is equivalent to the fact that Γ is a plane part of the boundary. 5. mesd−1 Γ > 0. . We put n ϕ(u) = ω u(x)dx + k=1 ω xk u(x)dx . Let ω ⊆ Ω be a measurable set.3. u|Γ = 0. it follows that u = 0. e. . mesd ω > 0. Let l = 2. . If u ∈ P2 (u(x) is a linear function). of linear combinations of the basis funcrtions 1. and u(x) = 0 is equation of this plane. We put ϕ(u) = Γ |u|dS. In particular. from u ∈ P2 . Consider P2 as a ﬁnite–dimensional subspace in L2 (ω). e. mesd−1 Γ > 0. . P2 consists of linear functions. Let l = 1. Then the norm (7) with l = 2 2 and ϕ(u) given by (17) is equivalent to the standard norm in Wp (Ω). Γ ⊂ ∂Ω.

this fact follows from the inequality j ξ k | ≤ |ξ j |2 + |ξ k |2 . The norm n —u—Wp (Ω) = l k=1 ∂lu ∂(xk )l p 1/p + u Lp (Ω) p Lp (Ω) is equivalent to the standard one.6. which does not follow from Theorem. we discuss one example. 2 For example. in W2 (Rd ) = H 2 (Rd ). 2|ξ 64 . In conclusion.

n S(Rn ) is called the Schwartz class. if f ∈ S(Rn ). but not Banach space. n |f (x)|2 dx = f (ξ) dξ. ∀k. We can introduce topology in S(Rn ). The inverse transformation F −1 is given by the formula n f (x) = (2π) −n 2 n f (ξ)eixξ dξ. Fourier transform. Deﬁnition Let f ∈ S(Rn ). It is known that the Fourier transform F can be extended by continuity to L2 (Rn ). 65 2 . x∈ sup (1 + |x|)k |∂ α ϕ(x)| < ∞. ∀α. F −1 : S(Rn ) → S(Rn ). It is known that f ∈ S(Rn ). Deﬁnition We say that ϕm −→ ϕ in S(Rn ). We deﬁne the transformation F : f → f . For ϕ ∈ S(Rn ) all derivatives ∂ α ϕ(x) are rapidly decreasing as |x| → ∞. f ∈ L2 (Rn ). n S(Rn ) is a topological space. and F is unitary operator in L2 (Rn ): F : L2 (Rn ) → L2 (Rn ).Chapter 3: Sobolev spaces H s (Rn ) §1: Classes S(Rn ) and S (Rn ). f (ξ) = (2π) −n 2 n f (x)e−ixξ dx. Deﬁnition S(Rn ) is a class of functions ϕ ∈ C ∞ (Rn ) such that for any multi–index α and any k ∈ N. So. F is called the Fourier transformation. if x∈ m→∞ m→∞ sup (1 + |x|)k |∂ α ϕm (x) − ∂ α ϕ(x)| −→ 0. F : S(Rn ) → S(Rn ) is a linear operator.

the space of linear continuous functionals on S(Rn )). A functional f ∈ S (Rn ) is called the Fourier image of f . ϕ := f. S (Rn ) is called the space of slowly increasing distributions. onto 66 . Deﬁnition Let f ∈ S (Rn ). ∀ϕ ∈ S(Rn ). It is known that F : S (Rn ) −→ S (Rn ).Deﬁnition By S (Rn ) we denote the dual space to S(Rn ) (i. Sometimes. The Fourier transformation is extended to the class S (Rn ). ϕ ∈ S(Rn ). If v ∈ S (Rn ). if f . F −1 : S (Rn ) → S (Rn ).. ϕ we denote the meaning of functional v on function ϕ. e. by v. ϕ .

Let u ∈ H l (Rn ). Consider the Fourier image Then u(x) = (2π) −n 2 n For the derivatives ∂ α u(x). the norm l norm in W2 (Rn ). The inner product in H s (Rn ) is deﬁned by Theorem 1 ∞ H s (Rn ) is the closure of C0 (Rn ) with respect to the norm u Hs . u(ξ)eixξ dξ. n |α|≤l |ξ α |2 |u(ξ)|2 dξ. u 2 Hl( n) = |∂ α u|2 dx = u(ξ) = (2π) −n 2 n u(x)e−ixξ dx. we have ∂ α u(ξ) = (iξ)α u(ξ) = i|α| ξ α u(ξ). ≤ c2 1/2 n (1 + |ξ|2 )l |u(ξ)|2 dξ. We introduce the space with this norm. .§2: Spaces H s (Rn ) 1. is equivalent to the standard s ∈ Rn . Deﬁnition (1 + |ξ|2 )l |u(ξ)|2 dξ H s (Rn ) = u ∈ S (Rn ) : n (1 + |ξ|2 )s |u(ξ)|2 dξ < ∞ . then 2 Hl( n) c1 n (1 + |ξ|2 )l |u(ξ)|2 dξ ≤ u n Thus. We can use the Fourier transform and express the norm in Let Ω = R l W2 (Rn ) = H l (Rn ) in terms of the Fourier image. n . v)H s ( n) = n (1 + |ξ|2 )s u(ξ)v(ξ)dξ. Deﬁnition of H s (Rn ) l l We know that the spaces W2 (Ω) (l ∈ N) are Hilbert spaces : W2 (Ω) = H l (Ω). Then |α|≤l n Since c1 (1 + |ξ|2 )l ≤ |α|≤l |ξ α |2 ≤ c2 (1 + |ξ|2 )l (prove this!). 67 (u. now we consider arbitrary l (not only l ∈ N).

Obviously. Then wk ∈ C0 (Rn ) and wk (ξ)(1 + |ξ|2 )s/2 → u∗ (ξ) in L2 (Rn ). Then ukj ∈ s ukj − uk 2 Hs = ≤ n n |ukj (ξ) − uk (ξ)|2 1 + |ξ|2 dξ l≥s. We have: j→∞ w(ξ) = u(ξ). ∞ 2) Let us show that each element of the closure of C0 (Rn ) in H s – ∞ norm belongs to H s (Rn ). Then also uk ∈ S(Rn ) L2 (Rn ). we ﬁx h ∈ C0 (Rn ) such that h(x) = 1 for |x| ≤ 1. u∗ (ξ) = u(ξ)(1 + |ξ|2 )s/2 ∈ L2 (Rn ). Then m w ∈ S (Rn ). If u ∈ H s (Rn ). It means that um (ξ)(1 + |ξ| m tal sequence in L2 (Rn ). For this. We put uk = F −1 wk . We put w(ξ) = u∗ (ξ)(1 + |ξ|2 )−s/2 . F −1 w = u ∈ S (Rn ). 68 ≤ c ∂ α uk (x) 1 − h |∂ α uk (x)|2 dx . then uk −→ u in H s (Rn ). then u∗ (ξ) = u(ξ)(1 + |ξ|2 )s/2 ∈ L2 (Rn ). there exists a sequence vk ∈ C0 (Rn ) such that vk (ξ) −→ u∗ (ξ) in L2 (Rn ). i. u − u is the Cauchy sequence in H m l H s ( n ) → 0 as 2 )s/2 =: u∗ (ξ) is a fundamenm. there exists a limit m→∞ u∗ (ξ) −→ u∗ (ξ) in L2 (Rn ). It follows that ukj −→ uk in H s (Rn ). It remains to approximate functions uk ∈ S(Rn ) by functions ∞ ∞ ukj ∈ C0 (Rn ) (in the H s –norm). We put vk (ξ)(1 + |ξ|2 )−s/2 = ∞ wk (ξ).l∈ n |ukj (ξ) − uk (ξ)|2 1 + |ξ|2 dξ 2 Hl l = ukj − uk For l ∈ N we can use another norm (which is equivalent to the standard one): ukj − uk 2 Hl |α|≤l ≤ c ˜ |β|≤l |x|>j → 0 as j → ∞. therefore. Suppose that um ∈ C0 (Rn ) and {um } s (Rn ). wk ∈ S(Rn ). since h . x j 2 dx x j = 1 for |x| ≤ j. Since L2 (Rn ) is complete. . l → ∞. and. We put ukj (x) = uk (x)h ∞ C0 (Rn ) and x j k→∞ and wk = uk .Proof 1) Let us show that any u ∈ H s (Rn ) can be approximated by functions ∞ in C0 (Rn ). ∞ ∞ Since C0 (Rn ) is dense in L2 (Rn ). Since uk (ξ)(1 + |ξ|2 )s/2 −→ u∗ (ξ) = u(ξ)(1 + |ξ|2 )s/2 in k→∞ k→∞ . e. m→∞ um (ξ)(1 + |ξ|2 )s/2 −→ u(ξ)(1 + |ξ|2 )s/2 in L2 (Rn ).

and let uj . m→∞ 2. We have: n L2 ( n) L2 ( = u Hs v H −s . uj −→ u in j→∞ H s (Rn ). We have Hs uj (ξ)vj (ξ)dξ uj (ξ)(1 + |ξ|2 )s/2 · vj (ξ)(1 + |ξ|2 )−s/2 dξ. Then there exists the limit j→∞ j→∞ n uvdx ≤ u Proof We have n Since uj −→ u in H s (Rn ). Duality of H s and H −s . vj ∈ C0 (Rn ). each element of the closure ∞ of C0 (Rn ) in · H s belongs to H s (Rn ). vj −→ v in H −s (Rn ). Theorem 2 ∞ Let u ∈ H s (Rn ). u(x)v(x)dx. v H −s . n . 69 ≤ As u uvdx = n (As u) (ξ)(A−s v) (ξ)dξ A−s v n) uj (x)vj (x)dx −→ j→∞ = uj (x)vj (x)dx = n n (As u) (ξ)(A−s v) (ξ)dξ =: We denote this limit by n u(x)v(x)dx. Then. it follows that uj (ξ)(1 + |ξ|2 )s/2 −→ u(ξ)(1 + |ξ|2 )s/2 =: (As u) (ξ) in L2 (Rn ). The fact that vj −→ v in H −s (Rn ) means that vj (ξ)(1 + |ξ|2 )−s/2 −→ v(ξ)(1 + |ξ|2 )−s/2 =: (A−s v) (ξ) in L2 (Rn ). Thus. v ∈ H −s (Rn ). (1) lim n uj (x)vj (x)dx. by (1). we have n n j→∞ j→∞ j→∞ j→∞ It is clear that the limit limj→∞ n uj vj dx does not depend on the choice of the sequences {uj } and {vj }.It means that um −→ u in H s (Rn ).

Theorem 3 If v ∈ H −s (Rn ). then 0=u∈H s ( ∞ 0=u∈C0 ( Proof 1) The mapping As : H s (Rn ) → L2 (Rn ). Thus. and put u = F −1 w. for v∗ = A−s v we have v H −s = = v∗ L2 n 0=u∈H s sup (As u) (ξ)(A−s v) (ξ)dξ u Hs n u(x)v(x)dx u Hs = 0=u∈H s sup From Theorems 2 and 3 it follows that l(u) = n uvdx is a linear continuous functional on u ∈ H s (Rn ) (if v ∈ H −s (Rn )) and the norm of this functional is equal to v H −s : l = sup 0=u∈H s |l(u)| = sup u Hs 0=u∈H s n u 70 = sup n) g(ξ)v∗ (ξ)dξ g L2 uvdx Hs v H −s = sup n uvdx Hs n) u n uvdx Hs u . The mapping A−s : H −s → L2 is deﬁned s similarly. s/2 Then u(ξ) = w(ξ) and u∗ (ξ) = u(ξ)(1 + |ξ|2 )s/2 = (As u)(ξ). It is known that in L2 we have n v∗ L2 = sup 0=g∈L2 We put u = A−1 g. (As u)(ξ) = u(ξ) 1 + |ξ|2 is a one–to–one isometric mapping. The inverse mapping A−1 : L2 (Rn ) → H s (Rn ) is deﬁned as follows: for s u∗ (ξ) u∗ ∈ L2 consider w(ξ) = (1+|ξ|2 )s/2 . g L2 = u H s . Then v∗ ∈ L2 . Thus. As u L2 = u H s . (2) . then u runs over H s . A−1 u∗ = F −1 w = u. . Then g(ξ) = (As u)(ξ). Indeed. 2) Let v ∈ H −s and v∗ (ξ) = (A−s v)(ξ). = v H −s . If g runs s over L2 .

e. c2 ∈ C. 2) If N = H. Since l(zj ) = 0. |l(u)| . l(v0 ) l(u) v0 . u ∈ H. Indeed. a) l(c1 u1 + c2 u2 ) = c1 l(u1 ) + c2 l(u2 ). 3) For ∀u ∈ H consider u − Since v0 ∈ N ⊥ . Then there exists such element v ∈ H that l(u) = (u. 4) Uniqueness If (u. we have u− Denote v= Indeed. v0 l(v0 ) = 0 ⇒ (u.Riesz Theorem Let H be a Hilbert space and l(u). ˜ ˜ ˜ 5) The norm of l. it follows that l(z) = 0. This element v is unique and l = v H . So.v)| ≤ v u H H for ∀ 0 = u ∈ H. v0 2 0 Then l(u) = (u. then l(u) = 0. ∀u ∈ H. if zj ∈ N and zj −→ z in H. ∀u ∈ H. ∀ u1 . l(v0 ) = 0. v) = (u. v ). then v − v ⊥H ⇒ v − v = 0. u2 ∈ H s . |(u. b) |l(u)| ≤ c u Hs . l u − l(u) l(v0 ) v0 l(u) l(v0 ) v0 ∈ N. Then N is a closed subspace in j→∞ j→∞ H. v)| |l(u)| = sup = v u H u H 0=u∈H H . v). It means that l : H s → C. Then. Indeed. i. v)H . u Hs The norm l of a functional l is deﬁned by the formula l = 0=u∈H s sup 71 . v0 ) = l(u) l(v0 ) v . be a continuous linear functional on H. If N = H. and for u = v we have |(u. there exists v0 ∈ N ⊥ . l = sup 0=u∈H |(u. In this case v = 0. ∀ c1 . then l(zj ) −→ l(z). then N ⊥ = {0} ( where N ⊥ is the orthogonal complement of N ). Let l(u) be a continuous linear functional on H s (Rn ). = l(u) − l(u) l(v0 ) l(v0 ) = 0. ∀u ∈ H s (Rn ). z ∈ N . . v0 2 . v0 = 0. Proof 1) Let N = Ker l = {z ∈ H : l(z) = 0}.v)| u H = v H.

v ∈ H −s . Then u = A−1 u∗ . (3) l = v Proof Consider the mapping H −s . and ˜ = w l L2 . n ∀ u ∈ H s (Rn ). We denote v(x) = F Then n So. .Theorem 4 Let l(u) be a linear continuous functional on H s (Rn ). Then there exists unique element v ∈ H −s (Rn ). (As u) (ξ) = u∗ (ξ) = u(ξ) 1 + |ξ|2 ˜ ∗ ) = l(A−1 u∗ ) = l(u). s/2 Then l(u) = ˜ ∗ ) = n u(ξ) 1 + |ξ|2 l(u w(ξ)dξ. We deﬁne the functional ˜ ∗ ) on L2 (Rn ) by the formula l(u s Then ˜ is a linear continuous functional on L2 (Rn ). l By the Riesz theorem for the functional ˜ there exists unique function l n ) such that w ∈ L2 (R ˜ ∗) = l(u n u∗ (ξ)w(ξ)dξ. 72 v(ξ) = w(ξ)(1 + |ξ|2 )s/2 . (4) As : H s (Rn ) → L2 (Rn ). l(u) = ˜ ∗ ) = n (As u) (ξ)(A−s v) (ξ)dξ = n uvdx. such that l(u) = and uvdx. −1 w(ξ)(1 + |ξ|2 )s/2 . l(u For the norm of the functional l we have: l = sup ˜ ∗) l(u |l(u)| l = ˜ = w = sup u Hs u∗ L2 0=u∗ ∈L2 0=u∈H s L2 = v Remark Theorem 4 means that H −s (Rn ) is dual to H s (Rn ) with respect to L2 –duality. and v H −s = w L2 . H −s . l(u s s/2 . |v(ξ)|2 (1 + |ξ|2 )−s dξ = n |w(ξ)|2 dξ. We have w(ξ) = (1 + |ξ|2 )−s/2 v(ξ) = (A−s v) (ξ).

n =1 We have 2 . ρ > 0. a) |ω(ρξ)| ≤ c. ∞ Since ω ∈ C0 (Rn ). then ω(ξ) belongs to the Schwartz class S(Rn ).3. 73 = (2π)−n/2 (2π)−n/2 ρ−n ω n x ρ e−ixξ dx ω(y)e−iyξρ dy n ω(y)dy = (2π)−n/2 . Theorem 5 If u ∈ H s (Rn ). Proof For the Fourier transform of the convolution uρ = ωρ ∗ u we have uρ (ξ) = (2π)n/2 ωρ (ξ)u(ξ). Hence. which is summable since u ∈ H s . For u ∈ H s (Rn ) consider molliﬁcations: uρ (x) = (ωρ ∗ u)(x). →0 as ρ→0∀ ξ The function under the integral is estimated by C(1 + |ξ|2 )s |u(ξ)|2 . By the Lebesgue Theorem. ωρ (ξ) = x =y ρ = ω(ρξ). b) limρ→0 ω(ρξ) = ω(0) = (2π)−n/2 Let us estimate the norm uρ − u Hs . Next. n ω(x)dx = 1. Molliﬁcations in H s (Rn ) Let ωρ (x) = ρ−n ω x ρ be a molliﬁer. Recall that ω ∈ ω(x) ≥ 0. uρ (ξ) − u(ξ) = (2π)n/2 ω(ρξ) − 1 u(ξ). uρ − u H s → 0 as ρ → 0. then uρ − u Hs ∞ C0 (Rn ). uρ − u 2 Hs = n (1 + |ξ|2 )s |u(ξ)|2 (2π)n/2 ω(ρξ) − 1 dξ. → 0 as ρ → 0. ∀ξ ∈ Rn .

. ∞ u ∈ C0 (Rn ). ˜ ˜ In fact.4. {uj } is the Cauchy sequence in C r (Rn ). such that u ∈ C r (Rn ): uj − u C r −→ 0. We have n Then. So. There exists a limit j→∞ j→∞ ≤C u Remark Theorem 6 is generalization of the embedding theorem: l W2 → C r if 2(l − r) > n. by the H¨lder inequality.e. Then H s (Rn ) → C r (Rn ). ∂ α u(x) = (2π)−n/2 n (iξ)α u(ξ)eixξ dξ. 74 If |α| ≤ r. ∞ u ∈ C0 (Rn ). o |∂ α u(x)| ≤ ≤ n |ξ α | |u(ξ)| dξ n |α|≤r x∈ max max |∂ α u(x)| ≤ C u n i. then 2 n |ξ α |2 dξ (1+|ξ|2 )s Hs Hs < ∞. e. u(x) = u(x). |ξ α |2 (1 + |ξ|2 )−s dξ 1/2 n |u(ξ)|2 (1 + |ξ|2 )s dξ j. ∀ u ∈ H s. We identify u = u. By (5). Embedding H s → C r Theorem 6 Let s > r + n . Hs . Then there exists a sequence uj ∈ C0 (Rn ). for a. uj − ul C r ≤ C uj − ul H s −→ 0. ˜ ˜ We have proved that H s → C r and u Cr uj −→ u in H s . Thus. . and s − r > n . x ∈ Rn (check this!). 2 Proof ∞ 1) Let u ∈ C0 (Rn ).l→∞ u(x) = (2π)−n/2 u(ξ)eixξ dξ. ∀α. 1/2 . u Cr ≤ C u ∞ 2) Let u ∈ H s (Rn ). . (5) .

|z|n+2s = g(ξ)|u(ξ)|2 dξ. ∀ t > 0. Then u Hs = n |u(ξ)|2 (1 + A|ξ|2s )dξ Obviously. ξ ∈ Rn . |z| The function g(ξ) is homogeneous in ξ of order 2s: izξ 2 n The function g(ξ) depends only on |ξ|: g(ξ) = n where the axis 0z1 has direction of vector ξ. |tz|n+2s |1 − eiz1 |ξ| |2 dz . It follows that g(ξ) = A|ξ|2s . [s] = k . n 1/2 |∂ α u(x) − ∂ α u(y)|2 dxdy |x − y|n+2{s} . {s} > 0. Corollary If s > 0 . Then u H s u Hs .5. We have n n n P arseval n | where g(ξ) = n |1−en+2sdz . n n n |u(x) − u(x + z)|2 dxdz |z|n+2s |u(ξ)|2 |1 − eizξ |2 dξdz |z|n+2s 1/2 . 75 u Hs = |∂ α u|2 dx + g(tξ) = |1 − eitzξ |2 dz = t2s |z|n+2s n |1 − eitzξ |2 d(tz) = t2s g(ξ). n = |u(x) − u(y)|2 dxdy |x − y|n+2s y=x+z = = |u| dx + 2 . then the norm |α|≤k n |α|=k n is equivalent to u Hs . c1 (1 + |ξ|2 )s ≤ 1 + A|ξ|2s ≤ c2 (1 + |ξ|2 )s . A > 0. Equivalent norm in H s with fractional s > 0 Theorem 7 If 0 < s < 1. the norm u u Hs Hs is equivalent to the norm |u(x) − u(y)|2 dxdy |x − y|n+2s 1/2 n n Proof Note that F : u(x) − u(x + z) → u(ξ)(1 − eizξ ).

Proposition Let s1 < s < s2 . We denote λ = ε s2 −s > 0. ε–inequalities“ ” Obviously.Proof u ∈ H s ⇔ u ∈ H k and ∂ α u ∈ H {s} with |α| = k. −s s−s1 2 −s . . |x − y|n+2{s} (6) s1 ρ ≥ 1. and put C(ε) = λ−(s−s1 ) = ε 1 ερs2 −s + C(ε)ρ−(s−s1 ) = (λρ)s2 −s + (λρ)−(s−s1 ) 76 |∂ α u|2 dx + n |∂ α u(x) − ∂ α u(y)|2 dxdy . Then for ∀ ε > 0 ∃ C(ε) > 0 such that u 2 Hs ≤ε u 2 H s2 + C(ε) u 2 H s1 Proof (6) is equivalent to the inequality 1 + |ξ|2 s ⇔ ρs ≤ ερs2 + C(ε)ρs1 . H s1 (Rn ) → H s2 (Rn ) for s1 > s2 . Then 1 obviously ≥ . It is easy to check that u 2 Hs check this! u 2 Hk + |α|=k ∂αu 2 H {s} by Theorem 7 |α|≤k n |α|=k n 6. ⇔ 1 ≤ ερ s2 −s ≤ ε 1 + |ξ|2 s2 + C(ε) 1 + |ξ|2 −(s−s1 ) + C(ε)ρ ρ ≥ 1.

x = (x1 . 0) = (2π)− n−1 2 n−1 u(x) = u(x . Consider the traces of functions on the hyper–plane xn = 0. −∞ ≤ −∞ 1 Here the second integral is ﬁnite. (7) u(ξ . 1 (γ0 u) (x ) = u(x . ξn )dξn −∞ (9) . since s > 2 . xn ). We deﬁne the trace operator ∞ ∞ γ0 : C0 (Rn ) → C0 (Rn−1 ). Then the trace operator γ0 : C0 (Rn ) → C0 (Rn−1 ) can be 2 extended by continuity to the linear continuous operator 1 γ0 : H s (Rn ) → H s− 2 (Rn−1 ).§3: Trace embedding theorems We write x ∈ Rn as x = (x . . Theorem 8 ∞ ∞ Let s > 1 . xn ) = (2π)−n/2 1 n−1 ) Hs( n u(ξ . xn−1 ). (γ0 u) (x ) = u(x . . dξ eix ξ 1 √ 2π ∞ γ0 u ≤C u n) . ξn )eixn ξn eix ξ dξ dξn . (8) =a2 dξn 2 + ξ 2 )s (a n = a a2s ∞ −∞ ∞ −∞ d 1+ ξn a ξn a 2 s = a1−2s 1−2s dt (1 + t2 )s =cs ∞ = cs a dξn 2 (1 + |ξ |2 + ξn )s 77 ⇒ −∞ = cs (1 + |ξ |2 ) 2 −s . ξn )dξn . . Then =γ0 u(ξ ) ⇒ Then γ0 u(ξ ) 2 ∞ 1 γ0 u(ξ ) = √ 2π ∞ u(ξ . . We have H s− 2 ( Proof ∞ 1) Let u ∈ C0 (Rn ). 0). We have ∞ −∞ |u(ξ)|2 (1 + |ξ|2 )s dξn ∞ −∞ 2 (1 + |ξ |2 +ξn )−s dξn .

e. k. We have k H ( n−1 ) j=0 78 u 2 Hs( n) ≤c ϕ 2 1 s− 2 1 n−1 ) Hs( 3 =c ϕj γj u ≤c u n) γ0 uj − γ0 ul 2 1 H s− 2 ( ≤ cs uj − ul 2 Hs( j. 1. By the limit procedure. {γ0 uj } is a Cauchy sequence in H s− 2 (Rn−1 ). 1 1 (Rn−1 ) → H s (Rn ). . the estimate (10) is extended to all u ∈ H s (Rn ). . ϕk ) ∈ H s− 2 (Rn−1 ). . s > k + 2 . uj ∈ C0 (Rn ). . Hs( n−1 ) By deﬁnition . if ϕ = (ϕ0 . Then the trace operators 1 j γj = γ0 ◦ ∂xn : H s (Rn ) → H s−j− 2 (Rn−1 ) are continuous for j = 0. Then ∃ {uj }. . Integrate over Rn−1 : n−1 n−1 ) ∞ 2) C0 (Rn ) is dense in H s (Rn ). . 2 1 H s−j− 2 ( ∞ Let u ∈ H s (Rn ). 1 such that. u = P ϕ (∈ H s (Rn )). . i. v = γ0 u. . u ∈ C0 . .Thus. . Corollary 1 Let k ∈ N and s > k + 2 . (10) j→∞ n) −→ 0. . . uj − u By (10). 1 Theorem 9 (extension theorem) Denote H s− 2 (Rn−1 ) = H s− 2 (Rn−1 )×H s− 2 (Rn−1 )×.×H s−k− 2 (Rn−1 ). then ϕj = γj u. j = 0. from (8) and (9) it follows that 1 + |ξ |2 s− 1 2 γ0 u(ξ ) 2 ∞ ≤ cs −∞ |u(ξ)|2 (1 + |ξ|2 )s dξn . n) ∞ . . 1 n) . k. . . . 1. γ0 u 2 1 H s− 2 ( ≤ cs u 2 Hs( 1 + |ξ |2 1 s− 2 γ0 u(ξ ) dξ 2 ≤ cs n |u(ξ)|2 (1 + |ξ|2 )s dξ. Then there exists a limit: 1 1 j→∞ γ0 uj −→ v ∈ H s− 2 (Rn−1 ) in H s− 2 (Rn−1 ). There exists a linear continuous operator P :H 1 s− 2 1 Let k ∈ Z+ . We have H s−j− 2 ( So.l→∞ n) −→ 0. ϕ1 .

V (ξ . j ∂xn V (ξ . xn ) = j=0 1 j x ϕj (ξ )h xn j! n 1 + |ξ |2 . . h(t) = 1 for |t| ≤ 1. Let us show that V (ξ . xj h(ρxn ) → ij n Then u(ξ) = j=0 k ρj+1 h(j) j+1 ij ϕj (ξ )(1 + |ξ |2 )− 2 h(j) j! ξn 1 + |ξ |2 . 0) = ϕ0 (ξ ). xn ). n−1 ∞ Since h ∈ C0 (R). . ξn ) is the Fourier image (in one variable xn → ξn ) of V (ξ . Clearly. k. s) < ∞. Here ϕj (ξ ) is the Fourier image of ϕj (x ). . then h ∈ S(R) and. xn ) such that u ∈ H s (Rn ). where V (ξ . xn ) is the Fourier image of the function u(x . We have: k j=0 j=0 n−1 1+|ξ | 2 1 + |ξ|2 = 1 + |ξ |2 + ξn = (1 + |ξ |2 )(1 + τ 2 ). We put k V (ξ . .. 2 79 u 2 Hs( n) ≤C dξ ϕj (ξ ) (1+|ξ |2 )s−j− 2 2 We write the integral as change variable: τ = √ ξn u 2 Hs( n) = n |u(ξ)|2 (1 + |ξ|2 )s dξ ϕj (ξ ) (1 + |ξ | ) dξ . h(j) (τ ) (1 + τ 2 )s dτ = C(j. xn ∈ R. ξ ∈ Rn−1 . and in the internal integral 1 h(j) (τ ) (1+τ 2 )s dτ. ξn ρ . j = 1. 0) = ϕj (ξ ). so. ξn ) = V (ξ . . 0 ≤ h(t) ≤ 1. Then 2 2 2 −j−1 k ≤ C h (j) ξn 1 + |ξ |2 2 n (1 + |ξ|2 )s dξ. 1 g ρ 1 ξn ρ . dξn .Proof ∞ Let h ∈ C0 (R). We put u(ξ . . Note that xj g(xn ) → ij g(j) (ξn ) = ij n g(ρxn ) → F F F dj j dξn g(ξn ). ξn ). 2 .

j = 0.k j=0 k j=0 So. ϕk ) → u is a linear continuous operator 1 from H s− 2 (Rn−1 ) to H s (Rn ) . . . . . 80 = c ϕj ⇒ u 2 Hs( n) ≤ C n−1 ϕj (ξ ) (1 + |ξ |2 )s−j− 2 dξ . 2 1 2 1 H s−j− 2 ( n−1 ) . k. . . the operator P : ϕ = (ϕ0 . ϕ1 . and γj u = ϕj . . . .

There are diﬀerent ways of deﬁnition of the Sobolev spaces H s (Ω).§4: Spaces H s (Ω) (survey) 1. Next. s 2) The spaces Wp (Ω) with fractional s ≥ 0 and p = 2 can be deﬁned by analogy with Def 2 (with “2“ replaced by “p“). 2 def |∂ α u(x)−∂ α u(y)|2 dxdy 2 α u Hs = . Deﬁnition 1 H s (Ω) is the class of restrictions to Ω of functions in H s (Rn ): u ∈ H s (Ω) ⇔ ∃ v ∈ H s (Rn ). If H s (Ω). the space H s (Ω) is deﬁned. then both deﬁnitions give the same space: Def 1 ⇔ Def 2. there exists a linear continuous extension operator Π : H s (Ω) → H s (Rn ). Deﬁnition of H s (Ω) Let Ω ⊂ Rn be a domain. α 2 if s = [s] |α|≤s Ω |∂ u| dx. (11) |α|≤k Ω |∂ u| dx + |α|=k Ω Ω |x−y|n+2{s} if s = [s] = k. and the following norm is ﬁnite: u H s < ∞. Deﬁnition 3 ∞ H s (Ω) is the closure of C0 (Ω) with respect to the norm (11). 3) The embedding theorems can be generalized for spaces of fractional order. s ≥ 0. {s} = s − k. Approach I. such that their weak derivatives up to order k = [s] also belong to L2 (Ω). Case s ≥ 0. is the Sobolev space in the sense of Def 2. ◦ ◦ 81 . v|Ω = u. Approach II. Comments 1) If Ω ⊂ Rn is a bounded domain of Lipschitz class. Deﬁnition 2 H s (Ω) is the set of functions in L2 (Ω).

Then there exists a covering {Uj }j=1. Trace embedding theorems Theorems 8 and 9 can be extended to the case of bounded domain Ω with smooth boundary.. x ∈ Ω .Deﬁnition 4 Let s > 0.N such that Ω ⊂ N Uj . supp ζj ⊂ Uj . . ϕ ∈ H s (Ω).. UM are of second kind. we denote u. ϕ | . P0 u(x) = ◦ ◦ 1 Then P0 : H s (Ω) → H s (Rn ) is continuous. . For x ∈ ∂Ω we have M ζj (x) = 1. ◦ 2) Let u ∈ H s (Ω). ∞ There exists a partition of unity {ζj }. fj . l ≥ |s|. ϕ H s (Ω) ◦ By analogy with H s (Rn ) and H −s (Rn ). H −s (Ω) is the space of linear continuous functionals on H s (Ω) with the norm u H −s = sup 0=ϕ∈H s (Ω) ◦ | u. if s = m + 2 . x ∈ Ω. U2 . i. by deﬁnition. then ∃ diﬀeomorphism fj : Uj → K. m ∈ Z+ . (This is true in some neighbourhood j=1 −1 of ∂Ω. . Ω Comments 1 1) H s (Ω) = H s (Ω) for s < 2 . UN are strictly interior. 2. . m ∈ Z+ . if s = m + 1 . fj (Uj ∩ ∂Ω) = ∂K+ \Σ+ = Γ...) Let u ∈ C l (∂Ω). or Uj ∩ ∂Ω = ∅. such that ζj ∈ C0 (Rn ). Let Ω ⊂ Rn be a bounded domain of class C l . vj = uj ◦ fj . N j=1 ζj (x) = 1. 2 4) H s (Ω) is invariant with respect to diﬀeomorphisms of class C l . and UM +1 . y ∈ Γ . x ∈ Rn \Ω 3) Let Ω ⊂ Rn be a bounded domain with Lipschitz boundary. . 0. . H −s (Ω) = H s (Ω) ◦ ◦ ∗ . fj (Uj ∩ Ω) = K+ . fj ∈ C l . Then vj ∈ C l (Γ). . uj (x) = u(x)ζj (x). We extend vj by zero to Rn−1 \Γ: vj (y) = ˜ vj (y). Then. ϕ = u(x)ϕ(x)dx. y ∈ Rn−1 \Γ 82 . e. Then H −s (Ω) coincides with the space of restrictions to Ω of distributions ∈ H −s (Rn ). 0. either Uj ⊂ j=1 −1 Ω. . u(x). supp vj ⊂⊂ Γ. l ∈ N. . for u ∈ H −s (Ω). Suppose that domains U1 .

s ≤ l. = γl−1 u = 0. . 1. ϕ1 . . l ∈ N. s > k + 2 . j = 0. k (where The proof is based on Theorem 8. . s ≤ l. ◦ vj ˜ 2 Hs( n−1 ) def = u H s (∂Ω) . . Let k ∈ Z+ . (12) ∂j ∂ν j 1 3 1 ϕj j=0 2 1 H s−j− 2 (∂Ω) =c ϕ 2 H 1 s− 2 (∂Ω) . .Then vj ∈ C l (Rn−1 ). Then u ∈ H l (Ω) = l W2 (Ω) if and only if u ∈ H l (Ω) and γ0 u = γ1 u = . . {ζj }) are equivalent to each other. . and partition of unity {ζj }. Let γj : C l (Ω) → C l−j (∂Ω) be the trace operator: are normal“ derivatives of u). . ∂j u . 1. k. diﬀeomorphisms {fj }. . and k u Theorem 12 2 H s (Ω) ≤c Let Ω ⊂ Rn be a bounded domain of class C l . × H s−k− 2 (∂Ω). . Theorem 10 (trace embedding theorem) Let Ω ⊂ Rn be a bounded domain of class C l . Theorem 11 (extension theorem) 1 Let Ω ⊂ Rn be a bounded domain of class C l . supp vj ⊂ Γ. where k ∈ Z+ . j = 0. . We denote H 1 s− 2 (∂Ω) = H s− 2 (∂Ω) × H s− 2 (∂Ω) × . . 1 s > k + 2 . ϕj ∈ H s−j− 2 (∂Ω). It can be proved that all such norms (for diﬀerent {Uj } . So. s− 1 2 There exists a linear continuous operator PΩ : H (∂Ω) → H s (Ω) 1 such that. . . ◦ 83 . . This norm depends on the choice of covering {Uj }. . k. l ∈ N. if ϕ = (ϕ0 . ” Then the operator γj can be extended (uniquely) to linear continuous 1 operator γj : H s (Ω) → H s−j− 2 (∂Ω). then γj u = ϕj . {fj } . Consider ˜ ˜ 1/2 M j=1 Deﬁnition H s (∂Ω) is the closure of C l (∂Ω) with respect to the norm (12). . . and using covering {Uj }. diﬀeomorphisms {fj } and partition of unity {ζj }. ϕk ). s ≤ l. . ∂ν j ∂Ω γj u = j = 0. . and u = PΩ ϕ. the class H s (∂Ω) is well– deﬁned.

Consider um (x) = u(x . and 0 ≤ h(t) ≤ 1. We have: u(x . . we reduce the question to the following. xn ) = 0 ∂u (x . j = 1. xn ). = K+ K+ ∩{ 1 0<xn < m } |u(x)|2 dx ∂u ∂xj 2 2 K+ |1 − hm (xn )|2 dx ∂u 1 K+ ∩{0<xn < m } ∂xj → 0 as m → ∞. γ0 u = u|Γ = 0. . 0 ≤ t ≤ 2 . xn ) ∈ K+ . n − 1. for a. “⇐“ Using covering. xn )hm (xn ). then hm (t) = 1 1 for t ≥ m . diﬀeomorphisms and partition of unity. ∂xj ∂xj ∂ ∂u(x) ∂hm (u(x) − um (x)) = (1 − hm (xn )) − u(x). h(t) = 1 for t ≥ 1. ≤ dx 84 . xn > 0} be the half–ball. t)dt. We have the following representation for u(x): xn ◦ u(x . um ∈ H + m→∞ Let us check that um − u H 1 (K+ ) −→ 0. “⇒“ Obvious. Let K+ = {x ∈ Rn : |x| < 1. h(t) = 0. xn ) − um (x . Suppose that u ∈ H 1 (K+ ). u ∈ H 1 (Ω) ◦ ⇔ u ∈ H 1 (Ω) and γ0 u = 0. We have to prove that u ∈ H 1 (K+ ). u(x) = 0 near Σ+ . 1 (K ).Proof For simplicity we prove Theorem 12 in the case l = 1. ∂xn (13) 1 We ﬁx a cut–oﬀ function h(t) such that h ∈ C ∞ (R+ ). We put hm (t) = h(mt). ∂u(x) ∂ (u(x) − um (x)) = (1 − hm (xn )) . . xn ) = (1 − hm (xn ))u(x . (x . Then um (x) = 0 near ∂K+ . . ∂xn ∂xn ∂xn Then |u(x) − um (x)|2 dx = K+ K+ |1 − hm (xn )|2 |u(x)|2 dx ≤1 ≤ ∂ (u − um ) dx ∂xj 2 → 0 as m → ∞. e.

t)dt dx ∂xn 2 ∂u (x . t) dt dxn dx ∂xn dx Next step: consider molliﬁcations of um (x): (um )ρ .K+ 2 ∂ (u − um ) dx ∂xn 1/2 = K+ |1 − hm (xn )| 2 ∂u ∂xn 2 1/2 dx + + K+ →0 as m→∞ 1/2 ∂hm 2 2 ∂xn |u| dx . 85 . we can approximate function u(x) by functions (um )ρ ∈ C0 (K+ ) in ρ→0 ◦ H 1 (K+ )–norm ⇒ u ∈ H 1 (K+ ). t) dt ∂xn 2 xn 0 =xn xn m2 xn 0 12 dt dx ≤ ≤ cm2 cm2 ∂u 1 K+ ∩{0<xn < m } ∂xn → 0 as m → ∞. ∞ Thus. 1 m2 K+ ∩{ 1 0<xn < m } xn 0 2 ∂u (x . ρ > 0. we obtain: Jm [u] = ≤ m2 K+ h (mxn ) h (mxn ) K+ ≤C 2 0 2 xn 2 ∂u (x . Then (um )ρ ∈ ∞ C0 (K+ ) for suﬃciently small ρ and (um )ρ − um H 1 (K+ ) −→ 0. =Jm [u] It remains to show that Jm [u] → 0 as m → ∞. ∂xn ∂xn Using (13). We have: ∂hm (x) ∂ = (h(mxn )) = mh (mxn ).

Then v(x) satisﬁes the integral identity v ϕdx = Ω Ω f ϕdx. Now we want to deﬁne weak“ solution of problem (2) under wide conditi” ons on ∂Ω and f . there exists unique (weak) solution v ∈ H 1 (Ω) of the Dirichlet problem (2). ◦ ◦ if v satisﬁes the identity (3) for any ϕ ∈ H 1 (Ω). (1) If Φ(x) is arbitrary function in Ω such that Φ|∂Ω = g. x ∈ Ω u|∂Ω = g. Dirichlet problem for the Poisson equation Let Ω ⊂ Rn be a bounded domain. f ∈ C(Ω) and solution v ∈ C 2 (Ω). the boundary is suﬃciently smooth. Let us formally multiply equation − v = f by the test ∞ function ϕ ∈ C0 (Ω) and integrate over Ω. Then. First. Then we can consider arbitrary domains. Consider the classical Dirichlet problem: − u = F. In the classical setting of problem (2). A function v ∈ H 1 (Ω) is called a weak solution of the Dirichlet problem (2) with f ∈ H −1 (Ω). Deﬁnition Let Ω ⊂ Rn be arbitrary bounded domain. The boun- dary condition v|∂Ω = 0 we understand in the sense that v ∈ H 1 (Ω). for any f ∈ H −1 (Ω). x ∈ Ω v|∂Ω = 0 (2) where f (x) = F (x) + Φ(x). then the function v(x) = u(x) − Φ(x) is solution of the problem − v = f.§5: Application to elliptic boundary value problems 1. ∞ ∀ϕ ∈ C0 (Ω). We have v H 1 (Ω) ≤ C f H −1 (Ω) . ϕ ∈ H 1 (Ω) (since ◦ H −1 (Ω) is the dual space to H 1 (Ω) with respect to L2 –duality). ϕ ∈ H 1 (Ω). Theorem 1 Let Ω ⊂ Rn be a bounded domain. ◦ ◦ ◦ (3) 1 The left–hand side is well–deﬁned for any v ∈ H 1 (Ω) = W2 (Ω). ◦ ◦ and the right–hand side is well–deﬁned for f ∈ H −1 (Ω). 86 ◦ . we’ll study problem (2) with homogeneous boundary condition.

ϕ] := Ω ◦ ◦ v ϕdx. it follows from (5) that v H 1 (Ω) ≤C f H −1 (Ω) . v]1/2 . 87 . (5) Thus. 1/2 deﬁnes an inner product in the space H 1 (Ω). ϕ H 1 (Ω) ◦ 0=ϕ∈H 1 (Ω) and ϕ H 1 (Ω) [ϕ. lf = sup ◦ 0=ϕ∈H 1 (Ω) exists unique function v ∈ H 1 (Ω) such that lf (ϕ) = [v. by the Riesz Theorem. The corresponding norm [v. for antilinear continuous functional lf on H 1 (Ω) there this functional is equal to the norm of v. ◦ lf (ϕ) is antilinear continuous functional on ϕ ∈ H 1 (Ω): |lf (ϕ)| ≤ f H −1 (Ω) ϕ H 1 (Ω) . by deﬁnition of the class H −1 (Ω). ϕ ∈ H 1 (Ω). v. Since. We rewrite (3) in the following form: [v.Proof 1) The form [v. ·]. |lf (ϕ)| f H −1 (Ω) = sup . ϕ]. ϕ]1/2 . (Now we consider H 1 (Ω) as the Hilbert space with the inner product [·. 2) The right–hand side of (3) is lf (ϕ) = Ω f ϕdx. ϕ] = lf (ϕ). ◦ (4) By the Riesz Theorem. ϕ]1/2 |lf (ϕ)| = [v.) Then. v]1/2 is equivalent to the standard norm v H 1 (Ω) = Ω |v|2 + | v|2 dx This follows from the Friedrichs inequality: |v|2 dx ≤ CΩ | v|2 dx. v is the unique solution of (4) (⇔ (3)). and the norm of ◦ ◦ [ϕ. Ω Ω (here it is important that Ω is bounded). ∀v ∈ H 1 (Ω) ◦ .

Equation − u = F in Ω is understood in the sense of distributions: u(x) is a weak solution of (1). g ∈ H 1/2 (∂Ω). ◦ (8) where f = F + G. From G ∈ H 1 (Ω) it follows that G ∈ H −1 (Ω) and G H −1 (Ω) ≤ C2 G H 1 (Ω) . Then v = u − G ∈ H 1 (Ω) and γ0 v = 0. (7) If u ∈ H 1 (Ω) and γ0 u = g. Let F ∈ H −1 (Ω). if u ∈ H 1 (Ω). (6) ≤ C1 g H 1/2 (∂Ω) . u(x) satisﬁes the identity u ϕdx = Ω Ω F ϕdx. x ∈ Ω. This is equivalent to the fact that v ∈ H 1 (Ω). Then there exists unique weak solution u ∈ H 1 (Ω) of problem (1). Now we return to the problem (1) with non–homogeneous boundary condition u|∂Ω = g. for given F ∈ H −1 (Ω) and g ∈ H 1/2 (∂Ω). We have u Proof 1) By Theorem 11 (extension theorem). Suppose that Ω ⊂ Rn is a bounded domain of class C 1 .2. Consider the problem − u = F. 88 . Function v is solution of the problem − v=f v|∂Ω = 0. and γ0 u = g. by Theorem 10 (trace embedding theorem) the trace operator γ0 (γ0 u = u|∂Ω ) is continuous from H 1 (Ω) onto H 1/2 (∂Ω): γ0 : H 1 (Ω) → H 1/2 (∂Ω). ∞ ∀ϕ ∈ C0 (Ω). there exists extension G = PΩ g ∈ H 1 (Ω) such that γ0 G = g and G H 1 (Ω) H 1 (Ω) ≤C F H −1 (Ω) + g H 1/2 (∂Ω) . for g ∈ H 1/2 (∂Ω). Then. Then f ∈ H −1 (Ω) and f H −1 ≤ F H −1 + C2 G H 1 (Ω) ≤ F H −1 + C1 C2 g H 1/2 . We look for solution u ∈ H 1 (Ω). γ0 u = u|∂Ω = g(x). Theorem 2 Let Ω ⊂ Rn be a bounded domain of class C 1 .

∀u. u] = Ω |u|2 dx > 0 if u = 0. Lemma The operator A is compact operator in H 1 (Ω). with spectral parameter λ. ϕ] = ◦ Ω u ϕdx. x ∈ Ω u|∂Ω = 0. there exists unique solution v ∈ H 1 (Ω) of the problem (8). By the Riesz theorem for such forms it can be represented as [Au. Dirichlet problem with spectral parameter Now we consider the problem − u = λu + f (x). ≤ C3 f + C1 g ≤ C3 F + (C1 C2 C3 + C1 ) g 3. ϕ] = [Aϕ. Aϕ]. u] = [u. ϕ]. Let f ∈ H −1 (Ω). Obviously. As before. It follows that A = A∗ . [Au. This is inner product in H 1 (Ω). ϕ ∈ H 1 (Ω). Proof ◦ ◦ ◦ ◦ H 1 (Ω) This follows from the embedding theorem: H 1 (Ω) is compactly embedded in L2 (Ω). so [Au. Ω uϕdx = Ω ϕudx . u. and u H1 ◦ ≤ v H1 + G H −1 H −1 H1 H 1/2 (∂Ω) H 1/2 (∂Ω) . Here Ω is bounded. ◦ ◦ The form Ω uϕdx. (10) is called a weak solution of problem (9). ∀ϕ ∈ H 1 (Ω). where A is a linear continuous operator in H 1 (Ω). we denote [u. ϕ ∈ is continuous sesquilinear form in H 1 (Ω). 89 . A function u ∈ H 1 (Ω) satisfying identity u ϕdx = λ Ω Ω ◦ ◦ (9) uϕdx + Ω f (x)ϕdx. Deﬁnition Let Ω ⊂ Rn be arbitrary bounded domain.By Theorem 1. and v H 1 (Ω) ≤ C3 f H −1 . A > 0. So. Then u = v + G is unique solution of the problem (1). Next.

{Auk } converges strongly in H 1 (Ω). ◦ ◦ antilinear functional on ϕ ∈ H 1 (Ω). we reduced the problem (9) to the abstract equation (12) with compact operator A in the Hilbert space H 1 (Ω). We analyse equation (12). l → ∞. ∀ϕ ∈ H 1 (Ω). then also Auk H 1 (Ω) is uniformly bounded. ◦ operator J : H 1 (Ω) → L2 (Ω) is compact. if and only if for any sequence {uk } which converges weakly in H. As before. which is equivalent to the equation u − λAu = v. The case v = 0 (which corresponds to f = 0): − u = λu x ∈ Ω u|∂Ω = 0 90 ◦ (13) . Thus. and f H −1 (Ω) v H 1 (Ω) . A is a continuous operator. ◦ (11) (12) where v ∈ H 1 (Ω) is given. the sequence {T uk } converges strongly in H.We’ll use the following property of compact operators: T is a compact operator in the Hilbert space H. the functional lf (ϕ) = ◦ unique element v ∈ H 1 (Ω) such that Ω f ϕdx = [v. A(uk − ul )] = Ω (uk − ul )(Auk − Aul )dx L2 (Ω) ≤ → 0 ◦ u k − ul →0 Auk − Aul ≤C L2 (Ω) as k. we can rewrite identity (10) in the form [u. ϕ] + [v. ϕ] = λ [Au. We have [A(uk − ul ). ϕ] . It follows that A is compact operator. Since the embedding ◦ ly converges in H 1 (Ω). there exists ◦ ◦ Ω f ϕdx (where f ∈ H −1 (Ω)) is continuous ∀ϕ ∈ H 1 (Ω). {uk } converges strongly in L2 (Ω). ϕ]. By the Riesz Theorem. and we are looking for solution u ∈ H 1 (Ω). it follows that uk H 1 (Ω) is uniformly bounded. using the properties of compact operators. Now. We want to check that {Auk } converges strongly in L2 (Ω). Since {uk } weak◦ Let {uk } be a weakly convergent sequence in H 1 (Ω).

There exists non–trivial solution only if λ = λj . Theorem 3 ⇔ u − λAu = 0 ⇔ Au = µu (where µ = The spectrum of the Dirichlet problem (13) is discrete. each eigenvalue is of ﬁnite multiplicity (i. 1 Then for the eigenvalues λj = µj of the Dirichlet problem (13) we have the following properties: 0 < λ1 ≤ λ2 ≤ . Then each eigenvalue corresponds to one eigenfunction uj : Auj = µj uj . . λj → ∞ as j → ∞. if λ = λj (= µj ). and u H 1 (Ω) ◦ ≤ Cλ f H −1 (Ω) .1 ) λ It is known that the spectrum of a compact operator is discrete: it consists of eigenvalues µj . We can ﬁnd unique solution u − λAu = v u = (I − λA)−1 v. In our case A = A∗ > 0. then the operator (I − λA)−1 is bounded. j ∈ N. that may accumulate only to point µ = 0. then for any f ∈ H −1 (Ω) there / exists unique (weak) solution u ∈ H 1 (Ω) of the problem (14). 91 . We enumerate eigenvalues in non–increasing order counting multiplicities µ1 ≥ µ2 ≥ . The case v = 0 (f = 0) − u = λu + f u|∂Ω = 0 ⇔ x∈Ω (14) 1 For compact operator A it is known that. ∀j ∈ N. All eigenvalues are positive and have ﬁnite multiplicities. . j ∈ N. then all eigenvalues µj are positive: µj > 0. e. Thus. The only accumulation point is inﬁnity: λj → ∞ as j → ∞. If λ ∈ {λj }j∈ (λ is not eigenvalue). . . we have the following theorem. H −1 (Ω) . We have : µj → 0 as j → ∞. and u Since v H 1 (Ω) Theorem 4 f H 1 (Ω) ≤ (I − λA)−1 =Cλ v H 1 (Ω) . j ∈ N. we arrive at the following theorem. Eigenfunctions {uj } are linearly independent.. . dim ker(A − µj I) < ∞).

. if v satisﬁes the solvability condition: v⊥ker(I − λj A). this solvability condition is equivalent to: f (x)ϕj (x)dx = 0. (k) where u0 is a ﬁxed solution. suppose that λ = λj . (15) (k) p j=1 cj ϕj Ω The solution u(x) is not unique. Let λ1 ≤ λ2 ≤ . j ∈ N.Now. 4. then problem (14) has solution for any f ∈ H −1 (Ω). There exists an orthogonal system of eigenfunctions {ϕj }j∈ : ϕj − λj Aϕj = 0. Indeed. but is deﬁned up to a summand with arbitrary constants cj . Since [v. . Solution is not unique and is represented as p u = u0 + j=1 cj ϕj . ϕl ] = 0. ∞ F = j=1 [F. Next. [Aϕj . . ϕl ] = Ω ϕj ϕl dx = µj [ϕj . ϕ] = Ω f (x)ϕ(x)dx. k = 1. . and ϕj . solution of the equation u − λj Au = v exists. p. which satisﬁes the solvability conditions (15). are corresponding (linearly independent) eigenfunctions. and cj are arbitrary constants. Here we repeat each λj according to its multiplicity. ϕl ] = 0. Hilbert–Schmidt Theorem Finally. {ϕj }j∈ is orthogonal basis in H 1 (Ω). ◦ . i. . . ·]) in ◦ (k) H 1 (Ω) to all eigenfunctions ϕj . [Au. . . Theorem 5 (k) If λ = λj is eigenvalue of the Dirichlet problem. . Thus. (k) k = 1. ϕ] = Ω uϕdx (by 1 Deﬁnition of operator A). . It means that v is orthogonal (with respect to the inner product [·. e. j = l. . p. Then. for any F ∈ H 1 (Ω). k = 1. 92 [ϕj . p. Aϕj = µj ϕj (where µj = λj ). corresponding to the eigenvalue λj (here p is the multiplicity of λj ). ϕj ] It is important that ϕj ⊥ϕl also in L2 (Ω). . and v = 0 (f = 0). be eigenvalues of the Dirichlet problem. . ◦ By the Hilbert–Schmidt Theorem. j = l. . ϕj ] ϕj [ϕj . . we can apply the Hilbert–Schmidt Theorem for compact operators and obtain the following result.

This system forms an orthogonal basis in L2 (Ω) and in H 1 (Ω) (with respect to the inner product [·. ◦ 93 . The last fomula can be extended to all F ∈ L2 (Ω). ϕj ] = [ϕj . and F = j=1 (F. ϕj ] = λj [F. ϕj ] Ω F ϕj dx 2 Ω |ϕj | dx ∞ = (F. ϕj ] = λj Ω Then [F. ϕj ] = λj [Aϕj . |ϕj |2 dx. ·]). ϕj )L2 (Ω) ϕj 2 L2 (Ω) 2 L2 (Ω) . [ϕj . Aϕj ] = λj Ω F ϕj dx. Then there exists an ortogonal system of eigenfunctions {ϕj }j∈ of the Dirichlet problem.We have [F. ϕj )L2 (Ω) ϕj ϕj . Theorem 6 Let Ω ⊂ Rn be a bounded domain.

Sign up to vote on this title

UsefulNot useful- ans2
- Theory
- Test Matematika Mid Semester 2 Class 7 2011
- Infinity 1
- Saharon Shelah- Black Boxes
- IJCMS29-32-2009 New Results of Fuzzy Alpha - Open Sets
- latha-al
- Uncertainty With Ordinal Likelihood Information
- measch5
- Basic Set Theory - Anon
- 1-s2.0-S1658365513000794-main
- Group Theory
- TARK15.a.ozgun
- Novikov (2004)
- Mirna Dzamonja and Saharon Shelah- On the existence of universal models
- Gsm 70 GSM 70Errata
- Analysis Solutions
- sc_jalc
- Pt Ramdom
- SBEIII HW2 Key
- The Understanding of Despair through the Game of Chess
- Derivatives
- Chapter0_MathBasics
- AV02-0783EN
- Simplification Techniques and Tricks
- Jonathan Lear- Aritstotelian Infinity
- Toc
- pigozzinotes
- Long Line
- Saharon Shelah- PCF Without Choice
- Suslina_Sobolevraeume

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue reading from where you left off, or restart the preview.