Review Article Volume 1 Issue 5
Departamento de Matematica, Universidade Estadual de Maring
Correspondence: Larkin NA, Departamento de Matemática,Universidade Estadual de Maringá, Av. Colombo 5790, Agência UEM, 87020-900, Maringá, PR, Brazil
Received: June 26, 2018 | Published: September 27, 2018
Citation: Larkin NA, Padilha MV. Decay of solutions for 2D navier-stokes equations posed on rectangles and on a half-strip. Open Acc J Math Theor Phy. 2018;1(5):203-208. DOI: 10.15406/oajmtp.2018.01.00035
Initial-boundary value problems for 2D Navier-Stokes equations posed on rectangles and on a half-strip were considered. The existence and uniqueness of regular global solutions on rectangles and their exponential decay as well as exponential decay of generalized solutions on a half-strip have been established.
Keywords: navier-stokes equations, lipschitz and smooth domains, decay in bounded and unbounded domains
The main goal of this work is establishing of sharp estimates for the exponential decay rates of solutions to initial-boundary value problems for the 2D Navier-Stokes equations:
ut+(u⋅∇)u=νΔu−∇p, in Ω×(0,t),ut+(u⋅∇)u=νΔu−∇p,inΩ×(0,t), (1)
∇u=0 in Ω, u|∂Ω=0,∇u=0inΩ,u|∂Ω=0, (2)
u(x,y,0)=u0(x,y),u(x,y,0)=u0(x,y),
(3)
where
ΩΩ
is either a bounded rectangle or a half-strip in
ℝ2R2
with the homogeneous Dirichlet condition on the boundary of
Ω.Ω.
The question of decay of the energy for generalized solutions had been stated by J.Leray1 and attracts till now attention of many pure and applied mathematicians2‒11. In all of these papers, the decay rate of ∥u∥(t)L2(Ω)∥∥u∥(t)L2(Ω) was controlled by the first eigenvalue of the operator A=−PΔ,A=−PΔ, where PP is the projection operator on solenoidal subspace of L2(Ω).L2(Ω). Associated with stability questions, problems on dimensions of attractors and nonlinear spectral manifolds also have been studied.2,6,7
It is well-known that solutions of the 2D Navier-Stokes equations posed on smooth bounded domains with the Dirichlet boundary conditions are globally regular.9,11‒14 On the other hand, the question of regularity is not obvious in the case of bounded Lipschitz domains and unbounded Lipschitz and smooth domains. It has been proved that for Lipschitz domains, bounded and unbounded, there exists a unique global generalized solution.9,11,14
u,ut ∈L∞(0,∞;L2(Ω))∩L2(0,∞;H1(Ω)),u,ut∈L∞(0,∞;L2(Ω))∩L2(0,∞;H1(Ω),)
but it was not clear whether
u ∈ L∞(0,∞;H2(Ω))u∈L∞(0,∞;H2(Ω))
at least for bounded Lipschitz domains.
In this work, we have established this fact for rectangles making use of ideas.15 The following inequality holds for rectangles
∥u∥2H2(Ω)(t)+∥ut∥2L2(Ω)(t)≤C∥u0∥2H2(Ω)exp(−ν(π2L2+π2B2)t)∥∥u∥2H2(Ω)(t)+∥ut∥2L2(Ω)(t)≤C∥u0∥2H2(Ω)exp(−ν(π2L2+π2B2)t)
and
∥u∥2H1(Ω)(t)+∥ut∥2L2(Ω)(t)≤C∥u0∥2H2(Ω)exp(−νπ2B2t)∥∥u∥2H1(Ω)(t)+∥ut∥2L2(Ω)(t)≤C∥u0∥2H2(Ω)exp(−νπ2B2t)
for a half-strip.
Our paper has the following structure: Chapter I is Introduction. Chapter 2 contains notations and auxiliary facts. In Chapter 3, existence and uniqueness of global generalized solutions on either bounded or unbounded Lipschitz domains have been established. In Chapter 4, regularity and decay of solutions on rectangles and on a half-strip have been studied.
Let ΩΩ be a domain in ℝ2.R2. Define as in:11
Dx=∂∂x, Dy=∂∂y, Dj=DxjxDyjy=∂|jx+jy|∂xjx∂yjy.Dx=∂∂x,Dy=∂∂y,Dj=DxjxDyjy=∂|jx+jy|∂xjx∂yjy.
We denote for scalar functions f(x,y,t)f(x,y,t) by Lp(Ω), 1<p<+∞ the Banach space with the norm
∥f∥pLp(Ω)=∫Ω|f|p dxdy, p∈(1,+∞), ∥f∥L∞(D)=ess supD|f(x,y)|.
For p=2, L2(Ω is a Hilbert space with the scalar product
(u,v)=∫Ωu(x,y)v(x,y)dxdy and the norm ∥u∥2=∫Ω|u(x,y)|2dxdy.
The Sobolev space
Wm,p(Ω)
is a Banach space with the norm
∥u∥Wk,p(Ω)=∑0≤|α|≤k∥Dαu∥Lp(Ω).
When
p=2, Wm,2(Ω)=Hm(Ω)
is a Hilbert space with the following scalar product and the norm:
((u,v))Hm(Ω)=∑|j|≤m(Dju,Djv), ∥u∥2Hm(Ω)=∑|j|≤m∥Dju∥2.
Let D(Ω) or D(ˉΩ) be the space of C∞ functions with compact support in Ω or ˉΩ . The closure of C∞ functions in Wm,p(Ω) is denoted by Wm,p0(Ω) and (Hm0(Ω) when p=2).
Define the auxiliary spaces which are projections for the solenoidal vector functions,
V={u∈D(Ω),∇u=0}, V=the closure of V in H10(Ω),
H=the closure of V in L2(Ω),
The space
H
is eqquiped with the natural
L2
inner product. The space
V
will be equipped with the scalar produt
((u,v))=(Dxu,Dxv)+(Dyu,Dyv)
when Ω is bounded. If Ω is unbounded, we define the inner product as the sum of the inner products as following:
[[u,v]]=(u,v)+((u,v)).
We use the usual notations of Sobolev spaces
Wk,p,
Lp
and
Hk
for vector functions and the following notations for the norms:
i) For vector functions
u(x,y,t)=(u1(x,y,t),u2(x,y,t)),
∥u∥pLp(Ω)=∫Ω(|u1|p+|u2|p )dxdy,
∥u∥Wk,p(Ω)=∑0≤|α|≤k∥Dαu1∥Lp(Ω)+∥Dαu2∥Lp(Ω), p∈(1,+∞).
The closures of V in L2(Ω) and in H10(Ω) are the basic spaces in our study. We denote them by H and V respectively.. Obviously V is a subspace of H10(Ω).
Define the operator
(u⋅∇)u=(u1u1x+u1u2x+u2u1y+u2u2y).
Lemma 4.1 (The Steklov Inequality)16 Let v∈H10(0,L). Then
π2L2∥v∥2(t)≤∥vx∥2(t). (4)
Proof. Let v(t)∈H10(0,π) , then by the Fourier series,
∫π0v2t(t) dt≥∫π0v2(t) dt.
Inequality (4) follows by a simple scaling.
Lemma 4.2 (Differential form of the Grownwall Inequality) Let I=[t0,t1]. Suppose that functions a,b:I→ℝ are integrable and a function a(t) may be of any sign. Let u:I→ℝ be a differentiable function satisfying
u′(t)≤a(t)u(t)+b(t),fort∈Iand u(t0)=u0, (5)
then
u(t)≤u0e∫tt0a(t) dt+∫tt0e∫st0a(r) drb(s) ds.
(6)
Proof. Multiply (5) by the integrating factor e∫st0a(r) dr and integrate from t0 to t.
The next Lemmas will be used in estimates:
Lemma 4.3 (See: 11,14 ) Let v∈H10(Ω) , then
∥v∥L4(Ω)≤21/4∥v∥1/2L2(Ω)∥∇v∥1/2L2(Ω). (7)
Lemma 4.4 (See: 14 ) Let b(u,v,w)=((u⋅∇)v,w), then
|b(u,v,w)|≤21/2∥u∥1/2∥u∥1/2V∥v∥V∥w∥1/2∥w∥1/2V
∀u,v,w∈H10(Ω)
. If
u∈L2(0,∞;V)∩L∞(0,∞;H)
, then we can define the operator
Bu
such that
Bu
belongs to
L2(0,∞;V′)
and
(Bu,v)=b(u,u,v),
∥Bu∥L2(0,∞;V′)≤21/2∥u∥L∞(0,∞;H)∥u∥L2(0,∞;V).
Let Ω be a bounded Lipschitz domain. Given u0∈H, consider the following problem:
(ut−νΔu+∇p+(u⋅∇)u=0inΩ×(0,t),∇u=0inΩ×(0,t),u=0 on ∂Ω×(0,t),t>0,u(x,y,0)=u0(x,y),inΩ (8)
equivalent to the variational problem given by,11
(u′+Au+Bu=0in(0,t),t>0u(0)=u0, (9)
where Au∈V′ such that (Au,v)=−ν((u,v)) for all v∈V and Bu∈V′ such that
(Bu,v)=b(u,u,v). (10)
Theorem 5.1 Given u0∈H2(Ω)∩V, there exists a unique generalized solution u to (8) such that for all Φ∈V, Φ|∂Ω=0 it satisfies the following integral identity:
∫Ω{utΦ+ν(uxΦx+uyΦy)−u(u⋅∇)Φ}dxdy=0, (11)
where
u∈L∞(0,∞;V), ut∈L∞(0,∞;H)∩L2(0;∞;V).
Proof. The estimates that follow may be established on Gallerkin approximations.14,9 We estimate:
Estimate I - u∈L∞(0,∞;H)∩L2(0,∞;V) .
Multiply (9) by u to obtain
(ut,u)(t)+(Au,u)(t)=0. (12)
It follows from here that
ddt∥u∥2(t)+2ν∥u∥2(t)V=0. (13)
Integrating (13) over (0,t), , we get
∥u∥2(t)+2ν∫t0∥u∥2V(s) ds=∥u0∥2, t>0.
(14)
Hence
u∈L∞(0,∞;H)∩L2(0,∞;V)
.
Estimate II - ut∈L∞(0,∞;H)∩L2(0,∞;V) .
Derivating (9) and multiplying by ut, we get
ddt∥ut∥2(t)+2ν∥ut∥2V(t)+2b(ut,u,ut)(t)=0. (15)
By Lemma 4.4,
2|b(ut,u,ut)(t)|≤23/2∥ut∥(t)∥ut∥V∥u∥V
≤ν∥ut∥2V+23ν∥u∥2V(t)∥ut∥2(t)
and (15) becomes
ddt∥ut∥2(t)+ν∥ut∥2V(t)≤ϕ(t)∥ut∥2(t), (16)
where ϕ(t)=23ν∥u∥2V(t). Making use of Lemma 4.2, we obtain
∥ut∥2(t)≤∥ut∥2(0)e∫t0 ϕ(s) ds. (17)
To prove that ∥ut∥(0) is in H, multiply equation (9) by ut(t) to get
∥ut∥2(t)+ν((u,ut))(t)+b(u,u,ut)(t)=0.
In particular, for t=0 we have
∥ut(0)∥2=ν(Δu0,ut(0))−b(u0,u0,ut(0)), (18)
where ut(0)=limt→0ut(t), [19]. From this
∥ut(0)∥≤ν∥Δu0∥+∥Bu0∥. (19)
By the Hölder inequality,
|b(u,u,v)|≤∥u∥L4(Ω)(2∥∇u∥L4(Ω)+∥u2x∥L4(Ω)+∥u1y∥L4(Ω))∥v∥
≤C∥u∥V∥u∥H2(Ω)∥v∥, ∀u∈H2(Ω),∀v∈L2(Ω).
(20)
Hence
|Bu0|≤C∥u0∥V∥u0∥H2(Ω)≤C∥u0∥2H2(Ω) (21)
and by (18), ut(0)∈H. This and (16) imply that
ut∈L∞(0,∞;H)∩L2(0,∞;V).
Returning to (12), we calculate
ν∥u∥2V(t)=(u,ut)(t)≤∥u∥(t)∥ut∥(t), (22)
hence u∈L∞(0,∞;V). This and (17) prove validity of (11) and consequently the existence part of Theorem 3.1. Uniqueness of the generalized solution, u, ut∈L∞(0,∞;H)∩L2(0,∞;V) has been established.9,14
Remark 5.1 Estimates u,ut∈L∞(0,∞;H∩L2(0,∞;V) were established first for Lipschitz domains9,14 and were valid also for unbounded domains with a natural condition lim|x|→+∞u(x,y,t)=0. We repeat them because we will need these estimates while establishing decay of solutions in bounded and unbounded Lipschitz domains.
Consider the Poisson problem in a bounded rectangle Ω
(Δu=f(x,y), (x,y)∈Ω,u|∂Ω=0, (23)
Remark 6.1 It has been proved10 that for
Ωπ={x=(x1,...,xn), 0<xi<π; i=1,...,n}
the following inequality holds
∥u∥W2,p(Ωπ)≤C(Ω)∥f∥Lp(Ωπ).
It is easy to generalize this result for any rectangle in ℝ2.
Theorem 6.1 The problem (23) posed in rectangle Ω={(x,y)∈ℝ2,0<x<L; 0<y<B}, where f∈Lp(Ω), 1<p≤2, has a solution u∈W2,p(Ω). Moreover,
∥u∥W2,p(Ω)≤cΩ∥f∥Lp(Ω). (24)
Returning to the original problem for the Navier-Stokes equations,
(ut−νΔu+∇p+(u⋅∇)u=0inΩ×(0,t)∇u=0inΩ×(0,t),u=0in∂Ω×(0,t),t>0,u(x,y,0)=u0(x,y),inΩ, (25)
where u is a vector function from ℝ2 into ℝ2 and p is a real function from ℝ2 into ℝ, and making use of Galerkin approximations, we establish the following result.
Theorem 6.2 Given u0∈H2(Ω)∩V, the problem (25) has a unique solution (u,p) such that
u∈L∞(0,∞;V∩H2(Ω)),ut∈L∞(0,∞;H)∩L2(0,∞;V),
∇p∈L∞(0,∞;H).
(26)
Moreover,
∥ut∥(t)+∥u∥(t)H2(Ω)+∥∇p∥(t)≤Ce−12χt,
(27)
where
χ=ν(π2L2+π2B2)
and
C
depends on
∥u0∥H2(Ω).
Proof. Decay of L2 Norm
By definition,
∥u∥2V(t)=∥ux∥2(t)+∥uy∥2(t).
Since u|∂Ω=0, making use of Lemma 4.1, we get
∥ux∥2(t)≥π2L2∥u∥2(t), ∥uy∥2(t)≥π2B2∥u∥2(t).
This implies
∥u∥2V(t)≥(π2L2+π2B2)∥u∥2(t).
(28)
Returning to (12), we obtain
ddt∥u∥2(t)+2ν(π2L2+π2B2)∥u∥2(t)≤0. (29)
Define
χ=ν(π2L2+π2B2).
Then (29) implies
∥u∥2(t)≤∥u0∥2 e−2χt.
(30)
Decay of
H1
Norm
Rewrite (15) in the form
ddt∥ut∥2(t)+ν∥ut∥2V(t)−ϕ(t)∥ut∥2(t)≤0, (31)
where ϕ=23ν∥u∥2V(t). Acting similarly to the proof of (29), we obtain
∥ut∥2V(t)≥(π2L2+π2B2)∥ut∥2(t). (32)
Hence (31) reduces to the form
ddt∥ut∥2(t)+(χ−ϕ(t))∥ut∥2(t)≤0. (33)
By Lemma 4.2,
∥ut∥2(t)≤∥ut(0)∥2 e∫t0 ϕ(s) ds e−χt. (34)
Since u∈L2(0,∞,V), then by (14),
∫t0ϕ(s) ds ≤ 2ν ∥u0∥2, t>0,
and it follows from (13) that
ν∥u∥2V(t)≤(u,ut)(t)≤∥u∥(t)∥ut∥(t)
≤∥u0∥∥u0′∥ e2ν∥u0∥ e−χt e−12χt.
(35)
Therefore
∥u∥2V(t)≤1ν∥u0∥∥u0′∥e2ν∥u0∥ e−32χt
(36)
and
∥u∥2H10(Ω)(t)≤(1ν∥u0∥∥u0′∥e2ν∥u0∥+∥u0∥2)e−32χt.
(37)
Decay of
H2
-Norm
In order to estimate
∥u∥H2(Ω)(t),
we will use Theorem 6.1. First write (8) as
Δu=f=1ν(ut+∇p−(u⋅∇)u).
We estimate
|b(u,u,v)|(t)=|((u⋅∇),v)(t)|≤c2∥u∥(t)L4(Ω)∥u∥(t)H10(Ω)∥v∥(t)L4(Ω)
≤C∥u∥2(t)H10(Ω)∥v∥(t)L4(Ω)
(38)
and by (30),
∥(u⋅∇)u∥(t)L4/3(Ω)≤Ce− 32χt.
Returning to (9), we obtain
∥Au∥L4/3(Ω)(t)≤∥Bu∥L4/3(Ω)(t)+∥ut∥L2(Ω)(t). (39)
It follows by (38) and (34) that ∥Au∥(t)L4/3≤Ce− 12χt. By Theorem of de Rham,17 one can check that there exists ∇p such that11
−∇p=ut+Au+Bu (40)
and
∥∇p∥L4/3(Ω)(t)≤∥ut∥L2(Ω)(t)+∥Au∥L4/3(Ω)(t)
+∥Bu∥L4/3(Ω)(t)≤Ce− 12χt.
(41)
Since f∈L4/3(Ω), due to Theorem 6.1,
∥u∥(t)W2,43(Ω)≤∥ut∥L2(Ω)(t)+∥∇p∥L4/3(Ω)(t)
+∥(u⋅∇)u∥L4/3(Ω)(t)
(42)
and by (42), we get ∥u∥(t)W2,4/3(Ω)≤Ce− 12χt. By the Sobolev theorems,
∥u∥L∞(Ω)(t)≤C∥u∥W2,4/3(Ω)(t)≤Ce−12χt.
(43)
This implies
∥Bu∥(t)≤C∥u∥(t)L∞(Ω)∥u∥(t)H10(Ω)∈L2(Ω).
To prove that the norms ∥ut∥L2, ∥∇p∥L2(Ω) and ∥(u⋅∇)u∥L2(Ω) have exponential decay, we use the equality (10)
∥(u⋅∇)u∥(t)=∥Bu∥(t),
where Bu∈L2(Ω)′ such that
〈Bu,v〉=∫Ω(u1u1xv1+u1u1yv2+u2u2xv1+u2u2yv2) dΩ
for every v∈L2(Ω). We calculate
|b(u,u,v)|(t)≤C∥u∥L∞(Ω)(t)∥u∥H10(Ω)(t)∥v∥(t). (44)
Since the right-hand side of (44) has exponential decay for every v∈L2(Ω), it follows
∥(u⋅∇)u∥(t)≤Ce−54χt. (45)
Returning to (9), we obtain the decay rate for the operator Au
∥Au∥(t)≤∥Bu∥(t)+∥ut∥(t).
It follows from (34) and (45) that ∥Au∥(t)≤Ce−12χt. By (40),
∥∇p∥L2(t)≤∥ut∥(t)+∥Au∥(t)+∥Bu∥(t)≤Ce−12χt. (46)
Since now
f∈L2(Ω),
substituting (34), (45), (46) into (24) and making use of Theorem 6,1, we prove
∥u∥H2(Ω)(t)≤Ce−12χt.
It means that a unique generalized soliution is regular.
The proof of Theorem 6.2 is complete.
Theorem 7.1 Consider the half-strip Ω={(x,y)∈ℝ2;0<x, 0<y<B}. Given u0∈H2(Ω)∩V, the following problem:
(ut−νΔu+∇p+(u⋅∇)u=0inΩ×(0,t),∇u=0inΩ×(0,t),u=0on∂Ω×(0,t),t>0,limx→∞|u(x,y,t)|=0, t>0,u(x,y,0)=u0(x),inΩ (47)
has a unique solution (u,p) such that
u∈L∞(0,∞;H10(Ω)),ut∈L∞(0,∞;L2(Ω)),
∇p∈L∞(0,∞;L2(Ω)).
(48)
Moreover,
∥ut∥(t)+∥u∥H10(Ω)(t)+∥∇p∥L4/3(Ω)(t)≤C2e−12θt, (49)
where θ=νπ2B2 and C2 depends on ν, ∥u0∥H2(Ω).
Proof. Obviously, the variational formulation of (47) is also (9). Repeating the proof of Theorem 5.1 (see Remark 3.1), we can proof the existence and uniqueness of the generalized solution18 to problem (47). Note that (14) holds for the problem (47). Using the Steklov inequality with respect to variable y, we obtain
∥uy∥2≥π2B2∥u∥2,
hence, similarly to (13),
ddt∥u∥2(t)+2νπ2B2∥u∥2(t)+2ν∥ux∥2(t)≤0. (50)
By Lemma 5, ∥u∥2(t)≤∥u0∥2e−2νπ2B2t. (51)
Since (31) holds for the problem (47), making use of Lemma 4.4, we estimate
ddt∥ut∥2(t)+2ν∥ut∥2(t)V≤2∥ut∥(t)∥u∥(t)V∥ut∥(t)V (52)
which we rewrite as
ddt∥ut∥2(t)+ν∥ut∥2V(t)−2ν∥u∥2V(t)∥ut∥2(t)≤0. (53)
By Lemma 4.1,
∥uty∥2(t)≥π2B2∥ut∥2(t)
and (53) becomes
ddt∥ut∥2(t)+[νπ2B2 − 2ν∥u∥2V(t)]∥ut∥2(t)≤0. (54)
By Lemma 5, (54) provides
∥ut∥2(t)≤∥ut∥2(0)e23ν∫t0∥u∥2V(s) dse−νπ2B2t,
hence
∥ut∥2(t)≤∥ut∥2(0)e2ν∥u0∥e−νπ2B2t. (55)
Returning to (35), we estimate
∥u∥2V(t)≤1ν∥ut∥(t)∥u∥(t)≤1ν∥u0∥∥ut∥(0)e2ν∥u0∥e−ν3π22B2t. (56)
Decay for Pressure
In order to obtain decay for
∥∇p∥L4/3(Ω)(t),
we start with
∥(u⋅∇)u∥L4/3(Ω)(t)=∥Bu∥L4(Ω)′(t),
where L4(Ω)′ is the dual of the space L4(Ω). Since
Au=−ut−Bu,
repeating calculations of (38) and making use of (34), we get
∥Au∥(t)L4/3(Ω)≤c1e−12θt. Observing that (40) holds for the problem (47), we obtain
∥∇p∥L4/3(Ω)(t)≤∥ut∥L2(Ω)(t)+∥Au∥L4/3(Ω)(t)
+∥Bu∥L4/3(Ω)(t)≤c2e−12θt.
(57)
Jointly (55), (56) and (57) prove (48), (49).
In our work, we tried to respond some questions posed by J. Leray,1 namely, regularity of global solutions of the Navier-Stokes equations and their decay. Therefore, our results can be divided in two parts: the first one concerns decay of global regular solutions of the 2D Navier-Stokes equations posed on rectangles.19 It is known that there exist global regular solutions for the 2D Navier-Stokes equations posed on smooth bounded domains,4,10,11,14 but regularity in nonsmooth (Lipschitz) domains, such as rectangles, is not obvious. For bounded rectangles, we have established the existence of an unique global regular solution which decays exponentially as We demonstrated that the decay rate is different for different norms, see (26), (30), (36), whereis defined by the geometrical characteristics of a domain
The second part of our work concerns decay of solutions for the 2D Navier-Stokes equations posed on a half-strip. In existing publications,3–11 the decay rate of is controlled by the first eigenvalue of the operatorwhereis the projection operator on solenoidal subspace of It is clear that this approach does not work in unbounded domains On the other hand, our approach based on the Steklov inequality with respect toallowed us to estimate the decay rate of a generalized solution for the 2D Navier-Stokes equations posed on a half-strip.
We must emphasize that this estimate is the first in the history which gives an explicit value of the decay rate for unbounded domains. Results established in our work can be used in constructing of numerical schemes for solving initial-boundary value problems for the Navier-Stokes equations appearing in Mechanics of viscous liquid. From the physical point of view, decay estimates show that the decay rate of perturbations of solutions caused by the initial data is bigger for bigger values of viscosityand smaller values of the width and length of the rectangles and the width of a half-strip.
None.
The authors declare that there are no conflict of interest regarding the publication of this paper.
©2018 Larkin, et al. This is an open access article distributed under the terms of the, which permits unrestricted use, distribution, and build upon your work non-commercially.