Quasilinear parabolic problems with nonlinear boundary conditions
Quasilinear parabolic problems with nonlinear boundary conditions Quasilinear parabolic problems with nonlinear boundary conditions
Theorem 4.3.1 Let 1 < p < ∞, J = [0, T ], and Ω ⊂ R n+1 be a domain with compact C 2 -boundary Γ which decomposes according to Γ = ΓD ∪ ΓN and dist (ΓD, ΓN) > 0. Suppose the assumptions (H1)-(H4) are satisfied. Then (4.40) admits a unique solution in the space Z := H α p (J; Lp(Ω)) ∩ Lp(J; H 2 p(Ω)) if and only if the functions f, g, h are subject to the following conditions. (i) f ∈ H α p (J; Lp(Ω)); (ii) g ∈ B (iii) h ∈ B 1 α(1− 2p ) pp 1 1 α( − 2 2p ) pp (iv) f|t=0 ∈ B 2 2− pα pp 2− 1 p (J; Lp(ΓD)) ∩ Lp(J; Bpp (ΓD)); 1− 1 p (J; Lp(ΓN)) ∩ Lp(J; Bpp (ΓN)); (Ω), if α > 1 p ; 1 1 2(1− − α pα ) pp (v) ∂tf|t=0 ∈ B (Ω), if α > 1 + 1 p ; (vi) f|t=0 = g|t=0 on ΓD, if α > 2 2p−1 ; (vii) B(0, x, Dx)f|t=0 = h|t=0 on ΓN, if α > 2 p−1 ; (viii) ∂tf|t=0 = ∂tg|t=0 on ΓD, if α > 1 + 3 2p−1 . Before proving Theorem 4.3.1 we recall some general properties of variable transformations. Let Ω ⊂ R n+1 be a domain with compact C 2 -boundary Γ and x0 ∈ Γ. Without restriction of generality, we may assume that x0 = 0 and that n(x0) = (0, . . . , 0, −1); this can always be achieved by a composition of a translation and a rotation in R n+1 . It is easy to see that such affine mappings of R n+1 onto itself leave invariant all function spaces under consideration (i.e. Lp, Z, and the regularity classes of the data and of the coefficients), and they also preserve ellipticity (including the ellipticity constant in (H3)) as well as normality. Continuing, by definition of a C 2 -boundary, there exist an open neighbourhood U = U1 × U2 ⊂ R n+1 of x0 with U1 ⊂ R n and U2 ⊂ R as well as a function ζ ∈ C 2 (U1) such that Define ϑ : Ū → Rn+1 by Γ ∩ U = {x = (x ′ , y) ∈ U : y = ζ(x ′ )}, Ω ∩ U = {x = (x ′ , y) ∈ U : y > ζ(x ′ )}. ϑk(x) = x ′ k if k = 1, . . . , n and ϑn+1(x) = y − ζ(x ′ ). (4.42) Clearly, ϑ ∈ C 2 (U, R n+1 ) is one-to-one and satisfies Ω ∩ U = {x ∈ U : ϑn+1(x) > 0} as well as Γ ∩ U = {x ∈ U : ϑn+1(x) = 0}. By extending ζ to a function ˜ ζ ∈ C 2 (R n ) with compact support and defining ˜ ϑ by (4.42) with ζ being replaced by ˜ ζ, we get a C 2 -diffeomorphism ˜ ϑ of R n+1 onto itself extending ϑ and satisfying ˜ ϑ(x) = x for large values of |x|. Also, ˜ ϑ is a C 2 -diffeomorphic mapping from Ω0 := {x ∈ R n+1 : y > ˜ ζ(x ′ )} onto R n+1 + . For the Jacobian D ˜ ϑ(x), one obtains D ˜ � In 0 ϑ(x) = −∇x ′ ζ(x ˜ ′ ) 1 � , x ∈ R n+1 , which entails det D ˜ ϑ(x) = 1 for all x ∈ R n+1 . Notice also that D ˜ ϑ(0) = In+1. 74
Given a function v ∈ H 2 p(R n+1 + ) we consider the pull-back Θv defined on Ω0 by Θv(x) = v( ˜ ϑ(x)). Using now the notation x = (x1, . . . , xn+1), the function u = Θv satisfies n+1 � ∂xiu(x) = ∂¯xk v( ˜ ϑ(x))∂xi ˜ ϑk(x), k=1 n+1 � ∂xi∂xj u(x) = ∂¯xk v( ˜ ϑ(x))∂xi∂xj ˜ ϑk(x) + k=1 n+1 � k, l=1 ∂¯xk ∂¯xl v( ˜ ϑ(x))∂xi ˜ ϑk(x)∂xj ˜ ϑl(x) for x ∈ Ω0 and i, j = 1, . . . , n + 1. For a differential expression of the form we thus obtain (Eu)(x) = −c(x) : ∇ 2 xu(x) + c1(x) · ∇xu(x) + c0(x)u(x), x ∈ Ω0, (4.43) (Eu)(x) = − (D ˜ ϑ(x)c(x)D ˜ ϑ T (x)) : ∇ 2 ¯xv( ˜ ϑ(x)) + (D ˜ ϑ(x)c1(x) − D 2 ˜ ϑ(x) : c(x)) · ∇¯xv( ˜ ϑ(x)) + c0(x)v( ˜ ϑ(x)), x ∈ Ω0, with (D 2 ˜ ϑ(x) : c(x))k = � n+1 i, j=1 cij(x)∂xi ∂xj ˜ ϑk(x), k = 1, . . . , n + 1. So applying the push-forward operator Θ −1 to the function Eu on Ω0 gives Θ −1 Eu = (Θ −1 EΘ)Θ −1 u = E ˜ ϑ v, where E ˜ ϑ := Θ −1 EΘ is the second order differential operator (E ˜ ϑ ˜ w)(¯x) = −c ϑ(¯x) 2 : ∇¯xw(¯x) + c ˜ ϑ 1(¯x) · ∇¯xw(¯x) + c ˜ ϑ 0(¯x)w(¯x), ¯x ∈ R n+1 + , (4.44) with coefficients c ˜ ϑ (¯x) = (D ˜ ϑ c D ˜ ϑ T ) ◦ ˜ ϑ −1 (¯x), c ˜ ϑ1(¯x) = (D ˜ ϑc1 − D 2 ˜ ϑ : c) ◦ ˜ ϑ −1 (¯x), c ˜ ϑ 0(¯x) = c0( ˜ ϑ −1 (¯x)), ¯x ∈ R n+1 + , (4.45) see also [69, Section 5]. Observe that the preceeding formulas are also valid for functions on J × Ω0 resp. J × R n+1 + and differential operators (4.43) resp. (4.44) with time-dependent coefficients. In view of (4.45), it follows in particular that for an operator A(t, x, Dx) of the form (4.41) and satisfying the smoothness and ellipticity conditions in (H2) and (H3) with Ω replaced by Ω0, the transformed operator A ˜ ϑ (t, ¯x, D¯x) defined on J × R n+1 + enjoys the same properties, the ellipticity constant c0 appearing in (H3) remaining unchanged. Further, since D ˜ ϑ ≡ 1 and the derivatives of ˜ ϑ and ˜ ϑ−1 up to order 2 are bounded, the change of variable formula for the Lebesgue integral shows that Θ induces isomorphisms Θ (p) : Hm p (R n+1 + ) → Hm p (Ω0) for each p ∈ (1, ∞) and m = 0, 1, 2. Proof of Theorem 4.3.1. We begin this time with the sufficiency part. The overall plan can roughly described as follows. With the aid of localization w.r.t. space and the coordinate transformations discussed above, problem (4.40) is reduced to a finite number of related problems on Rn+1 and R n+1 + , respectively. For these problems, solution operators are available thanks to the Theorems 4.1.1, 4.2.3, and 4.2.4. So the local equations can be solved and, by summing over all ’local solutions’, we obtain a fixed point equation for v of the form v = v0 + R(v) =: G(v), where v0 is determined by 75
- Page 25 and 26: Further, K ∞ (α, θa) := {a ∈
- Page 27 and 28: Using (2.19) for aω and bω yields
- Page 29 and 30: 2.7 Evolutionary integral equations
- Page 31 and 32: Example 2.8.1 For J = [0, T ] and a
- Page 33: We conclude this section by illustr
- Page 36 and 37: kernel a. The operator B is inverti
- Page 38 and 39: x := f(0) ∈ X exists and we are l
- Page 40 and 41: with two positive constants C1, C2
- Page 42 and 43: with two positive constants C1 and
- Page 44 and 45: derivative theorem to this pair of
- Page 46 and 47: 3.2 A general trace theorem Let X b
- Page 48 and 49: 3.3 More time regularity for Volter
- Page 50 and 51: Theorem 3.4.2 Suppose X is a Banach
- Page 52 and 53: Our next objective is to show neces
- Page 54 and 55: Let u1 be the restriction of v1 to
- Page 56 and 57: Proof. We begin with the necessity
- Page 59 and 60: Chapter 4 Linear Problems of Second
- Page 61 and 62: The strategy for solving (4.1) is n
- Page 63 and 64: Since ψj ≡ 1 on supp ϕj, we may
- Page 65 and 66: Turning to (c), let g ∈ Ξi+1 and
- Page 67 and 68: endowed with the norm | · | Y T 2
- Page 69 and 70: We remark that the constant C2 stem
- Page 71 and 72: One can then construct functions a
- Page 73 and 74: analogous to (4.17), shows that S i
- Page 75: Apply now V#, i+1 := I + k ∗ A#(
- Page 79 and 80: v is a solution of (4.40) on Ji+1 :
- Page 81 and 82: Chapter 5 Linear Viscoelasticity In
- Page 83 and 84: where δij denotes Kronecker’s sy
- Page 85 and 86: problem ⎧ ⎪⎨ ⎪⎩ ∂tv −
- Page 87 and 88: To see the converse direction, supp
- Page 89 and 90: and up solves � Aup − ∆xup
- Page 91 and 92: It can be written as where l(z, ξ)
- Page 93 and 94: elongs to H∞ (Σ π 2 +η × Ση
- Page 95 and 96: which allows us to write the first
- Page 97 and 98: Chapter 6 Nonlinear Problems 6.1 Qu
- Page 99 and 100: sufficiently small, say T ≤ T1
- Page 101 and 102: (d) bD ∈ C(J0 × ΓD × U0), ∃C
- Page 103 and 104: which entails (6.14). Corresponding
- Page 105 and 106: substitution operators to be studie
- Page 107 and 108: for all t, τ ∈ J, ξ, η ∈ K,
- Page 109 and 110: to write where h2(t, τ, x) = h21(t
- Page 111 and 112: Lemma 6.2.3 Let 0 < s < s0 < 1, ρ
- Page 113 and 114: Bibliography [1] Albrecht, D.: Func
- Page 115 and 116: [54] Lunardi, A.: On the heat equat
- Page 117 and 118: kleines T mit Hilfe des Kontraktion
- Page 119: Personal Details Curriculum Vitae N
Given a function v ∈ H 2 p(R n+1<br />
+ ) we consider the pull-back Θv defined on Ω0 by<br />
Θv(x) = v( ˜ ϑ(x)). Using now the notation x = (x1, . . . , xn+1), the function u = Θv<br />
satisfies<br />
n+1 �<br />
∂xiu(x) = ∂¯xk v( ˜ ϑ(x))∂xi ˜ ϑk(x),<br />
k=1<br />
n+1 �<br />
∂xi∂xj u(x) = ∂¯xk v( ˜ ϑ(x))∂xi∂xj ˜ ϑk(x) +<br />
k=1<br />
n+1 �<br />
k, l=1<br />
∂¯xk ∂¯xl v( ˜ ϑ(x))∂xi ˜ ϑk(x)∂xj ˜ ϑl(x)<br />
for x ∈ Ω0 and i, j = 1, . . . , n + 1. For a differential expression of the form<br />
we thus obtain<br />
(Eu)(x) = −c(x) : ∇ 2 xu(x) + c1(x) · ∇xu(x) + c0(x)u(x), x ∈ Ω0, (4.43)<br />
(Eu)(x) = − (D ˜ ϑ(x)c(x)D ˜ ϑ T (x)) : ∇ 2 ¯xv( ˜ ϑ(x))<br />
+ (D ˜ ϑ(x)c1(x) − D 2 ˜ ϑ(x) : c(x)) · ∇¯xv( ˜ ϑ(x)) + c0(x)v( ˜ ϑ(x)), x ∈ Ω0,<br />
<strong>with</strong> (D 2 ˜ ϑ(x) : c(x))k = � n+1<br />
i, j=1 cij(x)∂xi ∂xj ˜ ϑk(x), k = 1, . . . , n + 1. So applying the<br />
push-forward operator Θ −1 to the function Eu on Ω0 gives<br />
Θ −1 Eu = (Θ −1 EΘ)Θ −1 u = E ˜ ϑ v,<br />
where E ˜ ϑ := Θ −1 EΘ is the second order differential operator<br />
(E ˜ ϑ ˜<br />
w)(¯x) = −c<br />
ϑ(¯x) 2<br />
: ∇¯xw(¯x) + c ˜ ϑ<br />
1(¯x) · ∇¯xw(¯x) + c ˜ ϑ<br />
0(¯x)w(¯x), ¯x ∈ R n+1<br />
+ , (4.44)<br />
<strong>with</strong> coefficients<br />
c ˜ ϑ (¯x) = (D ˜ ϑ c D ˜ ϑ T ) ◦ ˜ ϑ −1 (¯x), c ˜ ϑ1(¯x) = (D ˜ ϑc1 − D 2 ˜ ϑ : c) ◦ ˜ ϑ −1 (¯x),<br />
c ˜ ϑ<br />
0(¯x) = c0( ˜ ϑ −1 (¯x)), ¯x ∈ R n+1<br />
+ , (4.45)<br />
see also [69, Section 5].<br />
Observe that the preceeding formulas are also valid for functions on J × Ω0 resp.<br />
J × R n+1<br />
+ and differential operators (4.43) resp. (4.44) <strong>with</strong> time-dependent coefficients.<br />
In view of (4.45), it follows in particular that for an operator A(t, x, Dx) of the form<br />
(4.41) and satisfying the smoothness and ellipticity <strong>conditions</strong> in (H2) and (H3) <strong>with</strong><br />
Ω replaced by Ω0, the transformed operator A ˜ ϑ (t, ¯x, D¯x) defined on J × R n+1<br />
+ enjoys<br />
the same properties, the ellipticity constant c0 appearing in (H3) remaining unchanged.<br />
Further, since D ˜ ϑ ≡ 1 and the derivatives of ˜ ϑ and ˜ ϑ−1 up to order 2 are bounded, the<br />
change of variable formula for the Lebesgue integral shows that Θ induces isomorphisms<br />
Θ (p) : Hm p (R n+1<br />
+ ) → Hm p (Ω0) for each p ∈ (1, ∞) and m = 0, 1, 2.<br />
Proof of Theorem 4.3.1. We begin this time <strong>with</strong> the sufficiency part. The overall<br />
plan can roughly described as follows. With the aid of localization w.r.t. space and<br />
the coordinate transformations discussed above, problem (4.40) is reduced to a finite<br />
number of related <strong>problems</strong> on Rn+1 and R n+1<br />
+ , respectively. For these <strong>problems</strong>, solution<br />
operators are available thanks to the Theorems 4.1.1, 4.2.3, and 4.2.4. So the local<br />
equations can be solved and, by summing over all ’local solutions’, we obtain a fixed<br />
point equation for v of the form v = v0 + R(v) =: G(v), where v0 is determined by<br />
75