Abstract
We consider the mixed boundary value problem for the Stokes and steady Navier–Stokes equations in two-dimensional domain
with corner points on boundary, where the boundary conditions change their type. Our goal is to prove some regularity results near
corner points, where the boundary conditions change.
© 2007 IMACS. Published by Elsevier B.V. All rights reserved.
Using polar coordinates (r, ω) and the substitution r = eτ and applying the complex Fourier transform with respect to
τ, we get
∂2 (p̂ τ
1e )
−
− (iλ − 1)2 (p̂ τ
1 e ) = H̃1 (λ, ω) in G = (0, ω0 ), (6)
∂ω 2
where (p̂ τ τ
1 e ) = F (p̂1 e ). We can write the general solution of (6) in the form
(p̂ τ
1 e )(λ, ω) = 4c1 cos(iλ − 1)ω + 4c2 sin(iλ − 1)ω + Ip (λ, ω), (7)
where Ip (λ, ω) is particular solution with respect to the right-hand side H̃1 (λ, ω). We reduce our problem – using
known localization principle – to a boundary value problem in a special domain, namely in the infinite angle K. In the
same way as for Eq. (5), we rewrite the Stokes system (1) and (2)
∂2 (û
1,1 )
∂(p̂ τ
1e )
− − (iλ)2 (û
1,1 ) + (iλ − 1) cos ω(p̂
τ
1e )− sinω = F̃1 , (8)
∂ω 2 ∂ω
∂2 (û
2,1 )
∂(p̂ τ
1e )
− − (iλ)2 (û
2,1 ) + (iλ − 1) sinω(p̂
τ
1e )+ cosω = F̃2 , (9)
∂ω 2 ∂ω
∂(û
1,1 ) ∂(û
2,1 )
(iλ)(û
1,1 ) cos ω − sin ω + (iλ)(û
2,1 ) sinω + cosω = 0, (10)
∂ω ∂ω
where F̃1 (λ, ω) = F (F̂1 ), F̃2 (λ, ω) = F (F̂2 ), F̂1 = f̂1,1 e2τ , F̂2 = f̂2,1 e2τ . The general solution of (8) and (9) has the
form
u
1,1 (λ, ω) = c3 cos(iλω) + c4 sin(iλω) + c1 cos(iλ − 2)ω + c2 sin(iλ − 2)ω + I1 (λ, ω), (11)
u
2,1 (λ, ω) = c5 cos(iλω) + c6 sin(iλω) − c1 sin(iλ − 2)ω + c2 cos(iλ − 2)ω + I2 (λ, ω). (12)
Now we transform the “Dirichlet–Dirichlet” boundary conditions to the forms
u
1,1 (λ, 0) = u
2,1 (λ, 0) = u
1,1 (λ, ω0 ) = u2,1 (λ, ω0 ) = 0 (13)
and denote by LλDD the operator of problem (8)–(10) with boundary conditions (13), respectively, the
“Neumann–Neumann” boundary conditions to the forms
∂u
1,1 ∂u
2,1 τ
= − p̂1 e = 0, ω = 0, (14)
∂ω ∂ω
∂u
1,1 τ
∂u
2,1 τ
+ sinω0 (p̂ 1e )= − cosω0 (p̂ 1 e ) = 0, ω = ω0 , (15)
∂ω ∂ω
and denote by LλNN the operator of problem (8)–(10) with boundary conditions (14) and (15), respectively, the mixed
“Dirichlet–Neumann” boundary conditions to the forms
u
1,1 = u2,1 = 0, ω = 0, (16)
∂u
1,1 τ
∂u
2,1 τ
+ sinω0 (p̂ 1e )= − cosω0 (p̂ 1 e ) = 0, ω = ω0 , (17)
∂ω ∂ω
and denote by LλDN the operator of problem (8)–(10) with boundary conditions (16) and (17).
If λ is no eigenvalue of Lλ[··] , then for any right-hand side from L2 (G) there exists a uniquely determined solution
2
(ũ1 , p̃1 ) ∈ [W 2,2 (G)] × W 1,2 (G). The eigenvalues of Lλ[··] are the zeros of determinants D[··] (λ) of corresponding
matrixes of coefficients c1 , c2 , c3 , c4 , c5 and c6 (substituting the general solution to the corresponding boundary
conditions and the continuity equation we get three linear systems of six equations) given by
DDD (λ) = DNN (λ) = (iλ)2 sin2 [ω0 ] − sin2 [(iλ)ω0 ] = 0, λ = 0, (18)
DDN (λ) = (iλ)2 sin2 [ω0 ] − cos2 [(iλ)ω0 ] = 0. (19)
10 M. Beneš / Mathematics and Computers in Simulation 76 (2007) 8–12
Let (u,p) ∈ [W1,2 (Ω)]2 × L2 (Ω) be the corresponding uniquely determined weak solution. From theory of V.A.
2
Kondra’tev (see [1]), it follows that (u, p) ∈ [V 2,2 (Ω, 1 + δ)] × V 1,2 (Ω, 1 + δ) for a small positive real number δ.
2
In order to get more qualitative regularity results, we will investigate the weak solution (u, p) ∈ [V 2,2 (Ω, 1 + δ)] ×
V 1,2 (Ω, 1 + δ) using the theory developed in [3]. Since fi,1 ∈ V 0,2 (K, β) for β ≥ 0, we get
|x|2β |fi,1 (x)|2 dx = e2(τ+τβ) |f̂i,1 (τ, ω)|2 dτ dω < ∞, for all β ≥ 0.
K S
Furthermore, we have
∞ ∞ ∞
2
e2(τ+τβ) |f̂i,1 (τ, ω)|2 dτ = (e2τ |f̂i,1 (τ, ω)|) e2τh dτ = |F̂i (τ, ω)|2 e2τh dτ < ∞,
−∞ −∞ −∞
i = 1,2, h = β − 1, β ≥ 0. Hence, it follows that the Fourier transform of F̂i (τ, ω) is meaningful in the half-plane
h = Im λ ≥ −1 for almost ω ∈ (0, ω0 ). Following regularity result holds for our case: if no eigenvalues of Lλ lie
on the line Im λ = h = β − 1, β ≥ 0, then the inverse Fourier transform
∞+ih
1 iλτ
ûi,1,h (τ, ω) = √ ui,1 (λ, ω) e dλ = ui,1,h (x) ∈ V 2,2 (K, β), (20)
2π −∞+ih
i = 1,2, exists, where ûi,1,h (x) is uniquely determined solution of the corresponding problem in the infinite angle from
V 2,2 (K, β). Analogously, we get for the pressure p̂1 ∈ V 1,2 (K, β), where β = h + 1.
Corollary 1.2.1. For the “Dirichlet–Dirichlet” and “Neumann–Neumann” boundary conditions is for our main
problem the case ω0 = π. It follows from (18), that for ω0 = π no eigenvalue of LλDD and LλNN lie on the
2
line h = Im λ = s for s > −1, therefore (û1 , p̂1 ) ∈ [V 2,2 (K, β)] × V 1,2 (K, β), β is small positive number. It follows,
2
that (u, p) ∈ [V 2,2 (Ω, β)] × V 1,2 (Ω, β). For a bounded domain Ω we have the continuous imbedding (see [3])
Finally, we get the weak solution
2
(u, p) ∈ [W 2−β,2 (Ω)] × W 1−β,2 (Ω), β is small positive number. (21)
It follows from (19), that for arbitrary ω0 ∈ (0, 2π), no eigenvalues of LλDN lie on the line Im λ = h = δ − 1, where
h ≥ −1/4, so we get the regularity result
2
(u, p) ∈ [W 5/4,2 (Ω)] × W 1/4,2 (Ω).
SKETCH OF THE PROOF: Let g ∈ H. Then there exists uniquely determined w ∈ D, such that
((w, v)) = (g, v), for all v ∈ V.
Let u be a fixed point in D. Denote the operator Bu : D → H, given by the equation
(Bu (w), v) = b(u, w, v) + b(w, u, v) = ((u∇)w, v) + ((w∇)u, v), for all v ∈ V.
Lemma 2.3. Let u0 ≡ 0. Then (using the Local Diffeomorphism Theorem, see [2]) we get: Gu0 = S is a one-to-one
operator from D to H.
Acknowledgement
The research was supported by the grant CTU0600111 of the Czech Technical University.
References
[1] V.-A. Kondra’tev, Boundary value problems for elliptic equations on domains with conical or angular points, Trudy Moskov. Mat. Obshch. 16
(1967) (in Russian).
[2] P. Kučera, Some properties of solutions to the nonsteady Navier–Stokes equations with mixed boundary conditions on an infinite time interval,
In: M. Feistauer, R. Rannacher, K. Kozel, (eds.), Proceedings of the 3rd Summer Conference on Numerical Modelling in Continuum Mechanics.
Part 2. Matfyzpress, Publishing House of the Faculty of Mathematics and Physics, Charles University, Prague, 1997, pp. 375–383.
[3] A. Kufner, A.-M. Sändig, Some Applications of Weighted Sobolev Spaces, Teubner-Texte zur Mathematik, Band 100, Leipzig, 1987.