JID:YJMAA AID:22848 /FLA Doctopic: Partial Differential Equations [m3L; v1.250; Prn:8/01/2019; 12:54] P.

1 (1-20)
J. Math. Anal. Appl. ••• (••••) •••–•••

Contents lists available at ScienceDirect

Journal of Mathematical Analysis and Applications

www.elsevier.com/locate/jmaa

Weak solutions with improved regularity for the nonhomogeneous

asymmetric fluids equations with vacuum
P. Braz e Silva a,∗,1 , F.W. Cruz a , M.A. Rojas-Medar b , E.G. Santos c
a
Departamento de Matemática, Universidade Federal de Pernambuco, Recife, PE, 50740-560, Brazil
b
Instituto de Alta Investigación, Universidad de Tarapacá, Casilla 7D, Arica, Chile
c
Departamento de Matemática, Universidade Federal da Paraíba, João Pessoa, PB, 58059-900, Brazil

a r t i c l e i n f o a b s t r a c t

Article history: We prove the existence of local and global in time weak solutions for the equations
Received 23 August 2018 of nonhomogeneous incompressible asymmetric fluids in a bounded domain Ω  R3 ,
Available online xxxx when the initial vacuum is allowed, i.e., the initial density ρ0 is not necessarily
Submitted by D. Wang
strictly positive. Our solutions here are more regular than previous results and so
Keywords: satisfy the initial conditions in a stronger sense. The global in time solutions are
Asymmetric fluids obtained under smallness assumptions on the initial data and forcing terms.
Existence © 2019 Elsevier Inc. All rights reserved.
Weak solutions
Vacuum

1. Introduction

We consider the nonhomogeneous incompressible asymmetric fluids equations

⎧  
⎪
⎪ ρ ut + (u · ∇)u − (μ + μr )Δu + ∇p = 2μr curl w + ρf ,
⎪
⎨ div u = 0,
  (1)
⎪
⎪ ρ wt + (u · ∇)w − (ca + cd )Δw − (c0 + cd − ca )∇(div w) + 4μr w = 2μr curl u + ρg,
⎪
⎩ ρ + u · ∇ρ = 0,
t

in the cylinder Ω × (0, T ), where Ω  R3 is a bounded open set and (0, T ) is a time interval with 0 < T ≤ ∞.
We complement this system with the following initial and boundary conditions

u(x, t) = w(x, t) = 0, ∀ (x, t) ∈ ∂Ω × (0, T ), (2)

ρ(x, 0) = ρ0 (x), u(x, 0) = u0 (x), w(x, 0) = w 0 (x), ∀ x ∈ Ω. (3)

* Corresponding author.
E-mail addresses: pablo@dmat.ufpe.br (P. Braz e Silva), wergete@dmat.ufpe.br (F.W. Cruz), marko.medar@gmail.com
(M.A. Rojas-Medar), eduardo@mat.ufpb.br (E.G. Santos).
1
Partially supported by CNPq (Brazil), grant #309491/2015-0.

https://doi.org/10.1016/j.jmaa.2018.12.075
0022-247X/© 2019 Elsevier Inc. All rights reserved.
JID:YJMAA AID:22848 /FLA Doctopic: Partial Differential Equations [m3L; v1.250; Prn:8/01/2019; 12:54] P.2 (1-20)
2 P. Braz e Silva et al. / J. Math. Anal. Appl. ••• (••••) •••–•••

In system (1), the unknowns ρ(x, t) ∈ R+ , u(x, t) ∈ R3 , p(x, t) ∈ R and w(x, t) ∈ R3 are, respectively,
the mass density, the linear velocity, the pressure distribution and the angular velocity of rotation of the fluid
particles as functions of the position x and of the time t. The positive constants μ, μr , c0 , ca and cd represent
viscosity coefficients and satisfy the inequality c0 + cd > ca . The functions f and g are given external forces.
By ∇, Δ, div and curl we denote the gradient, Laplacian, divergence and rotational operators, respectively;
ut , wt and ρt stand for the time derivatives of u, w and ρ.
For the derivation of equations (1) and a discussion about their physical meaning, see [13,14,23]. Con-
cerning applications, the micropolar fluid model has been used, for example, in lubrication theory [15,25],
as well as in modeling blood flow in thin vessels [2].
Observe that if μr = 0 and w = 0 = g, then system (1) is reduced to the nonhomogeneous Navier–Stokes
equations (see [1,16,17,19–22,26,28,29]). Concerning the model considered in this paper, let us recall that,
under certain assumptions, using linearization and an almost fixed point Theorem, Lukaszewicz established
in [24] the existence of weak solutions for short time (see also [23]). The existence of weak solutions with
ρ ≥ 0 was studied by Braz e Silva and Santos in [5,6] in the N s,q context (Nikolskii spaces). In [11,12],
strong solutions were established by Conca et al. using the linearization and by the method of successive
approximations. By using the semi-Galerkin spectral approximations Boldrini et al. [4] proved the existence
and uniqueness of strong solutions (the rate of convergence of this method was studied in [3]). In [7,8] the
authors prove that there exists a small time interval where the fluid variables converge uniformly as the
viscosities tend to zero. In the limit, they find a nonhomogeneous, non-viscous, incompressible asymmetric
fluid governed by an Euler-like system in the case R3 . In [9] the authors established the existence of local in
time semi-strong solutions and global in time strong solutions for the system of equations describing flows
of viscous and incompressible asymmetric fluids with variable density in general three dimensional domains
with boundary uniformly of class C 3 . Under suitable assumptions, uniqueness of local semi-strong solutions
was also proved. Here, we prove existence of weak solutions with improved regularity, so improving the
results in [5,6]. Initial conditions here are satisfied in a stronger sense. We show the existence of local in
time solutions. Moreover, under smallness assumptions on the initial data and forcing terms, we show that
the solutions are actually global in time.
This paper is organized as follows. In Section 2, the basic notation is stated and the main results are
formulated, while that in Section 3 we prove our existence results of local and global in time weak solu-
tions.

2. Preliminaries and results

2.1. Function spaces

Throughout this work, we use bold face letters to denote vector fields in R3 , as well as to indicate spaces
whose elements are of this nature. We denote by D(Ω) the space of C ∞ functions with compact support
in Ω and by D (Ω) be the space of distributions over Ω. The usual Lebesgue spaces are denoted by Lp(Ω),
1 ≤ p ≤ ∞. We denote Lp norms by  · p . The usual L2 (Ω) inner product and norm are respectively
indicated by (· , ·) and  · . We use the usual notation for Sobolev spaces

W m,p (Ω) = {v ∈ Lp (Ω) ; Dα vp < +∞ , |α| ≤ m} ,

for a multi-index α, a nonnegative integer m and 1 ≤ p ≤ ∞. The usual norm in W m,p (Ω) is denoted by
·m,p . We denote by W0m,p (Ω) the closure of D(Ω) in W m,p (Ω). When p = 2, we will set H m (Ω) := W m,2 (Ω)
and H0m (Ω) := W0m,2 (Ω). For m ∈ N, 1 ≤ p < ∞ and 1 < q ≤ ∞, where q is conjugate to p, the
Sobolev spaces W −m,q (Ω) represent the topological dual of W0m,p (Ω). When m = 1 and q = 2, we will set
H −1 (Ω) := W −1,2 (Ω). We write
JID:YJMAA AID:22848 /FLA Doctopic: Partial Differential Equations [m3L; v1.250; Prn:8/01/2019; 12:54] P.3 (1-20)
P. Braz e Silva et al. / J. Math. Anal. Appl. ••• (••••) •••–••• 3

V(Ω) = {u ∈ D(Ω) ; div u = 0 in Ω} ,

and denote by H and V the closure of V(Ω) in L2 (Ω) and H 1 (Ω), respectively.
If X is a Banach space and 0 < T ≤ ∞, we denote by Lp (0, T ; X) the Banach space of X-valued
(classes of) functions defined a.e. in [0, T ] that are Lp -integrable in Bochner’s sense and denote its norm
by  · Lp (0,T ;X) . Evidently, Lp (0, T ; Lp (Ω)) = Lp (Ω × (0, T )). Frequently, we will consider Banach spaces
Lp (0, T ; X) with X = W m,q (Ω). In such cases, for any v ∈ Lp (0, T ; W m,q (Ω)), v(t) stands for the function
v(·, t). We will also use the space of k-times differentiable functions on [0, T ) with values in X: C k ([0, T ); X).
We use the notation D(0, T ; X) for the space of X valued C ∞ functions defined [0, T ] and with compact
support in (0, T ). Accordingly, we indicate the space of distributions with values in X by D (0, T ; X).
Given a function f : (0, T ) → X and h > 0, let τh f : (−h, T − h) → X be the translated function of f ,
defined by (τh f ) (t) := f (t + h). For 1 ≤ q ≤ ∞, 0 < s < 1, the Nikolskii space N s,q is defined by

−s
N s,q
(0, T ; X) := f ∈ L (0, T ; X) ; sup h
q
τh f − f Lq (0,T −h;X) < ∞ .
h>0

The space N s,q (0, T ; X) is a Banach space with respect to the norm


f N s,q (0,T ;X) := f Lq (0,T ;X) + sup h−s τh f − f Lq (0,T −h;X) .
0<h<T

2.2. Auxiliary results

In order to achieve our ends, we need some preliminary results.

Lemma 1. (See [29, Lemma 1.4, p. 263]) Let X and Y be two Banach spaces, such that X → Y with a
continuous injection. If a function φ belongs to L∞ (0, T ; X) and is weakly continuous with values in Y , then
φ is weakly continuous with values in X.

Lemma 2. (See [27, Lemma 1.5, p. 22]) Let Ω ⊂ R3 be open and bounded, with Lipschitz boundary. If
h ∈ D (0, T ; H −1 (Ω)) satisfy h, v = 0 in D (0, T ), for all v ∈ V, then there exists g ∈ D (0, T ; L2 (Ω))
such that h = ∇g. Moreover, g may be chosen such that the map h → g is linear and continuous from
W s,r (0, T ; H −1 (Ω)) into W s,r (0, T ; L2 (Ω)), for all s ∈ R and 1 ≤ r ≤ ∞

Lemma 3. (See [28, Lemma 3, p. 1096]) Let Ω ⊂ R3 be open and bounded, with Lipschitz boundary.

(i) For 1 ≤ r ≤ ∞, W 1,r (Ω) → Lr∗ (Ω) with continuous imbedding, where

1 1 1
= − if r < 3, r∗ is any finite real if r = 3, r∗ = ∞ if r > 3.
r∗ r 3

The imbedding W 1,r (Ω) → Ls (Ω) is compact if s < r∗ .

(ii) For 1 ≤ s ≤ r ≤ ∞ the product is continuous, if t ≥ 1, in

1 1 1
W 1,r (Ω) × W 1,s (Ω) → W 1,t (Ω) , = + .
t r∗ s

(iii) For 1 ≤ r ≤ ∞, 1 < s ≤ ∞, the product is continuous, if 1/r + 1/s ≤ 1, in

1 1 1
W 1,r (Ω) × W −1,s (Ω) → W −1,t (Ω) , = + .
t r∗ s
JID:YJMAA AID:22848 /FLA Doctopic: Partial Differential Equations [m3L; v1.250; Prn:8/01/2019; 12:54] P.4 (1-20)
4 P. Braz e Silva et al. / J. Math. Anal. Appl. ••• (••••) •••–•••

Lemma 4. (See [27, Lemma 1.3, p. 17]) Let X → B → Y be Banach spaces, the imbedding X → B being
compact. Then, the following imbeddings are compact:

∂φ
(i) L (0, T ; X) ∩ φ ;
q
∈ L (0, T ; Y ) → Lq (0, T ; B), if 1 ≤ q ≤ ∞.
1
∂t

∞ ∂φ
(ii) L (0, T ; X) ∩ φ ; ∈ L (0, T ; Y ) → C([0, T ]; B), if 1 < r ≤ ∞.
r
∂t
(iii) Lq (0, T ; X) ∩ N s,q (0, T ; Y ) → Lq (0, T ; B), if 0 < s ≤ 1, 1 ≤ q ≤ ∞.

Moreover, if K ∈ L1 (0, T ) is a given function, then:

(iv.a) Every set

  
 ∂φ 
F ≡ (bounded in L∞ (0, T ; X)) ∩ φ ;   ≤ K, a.e. in [0, T ]
 ∂t 
Y

is relatively compact in C([0, T ]; B).

(iv.b) Every set
  
 ∂φ 
∞
F ≡ (bounded in L (0, T ; X)) ∩ φ ;   ≤ K + ψ, a.e. in [0, T ], ψLr (0,T ) ≤ C ,
 ∂t 
Y

where C is a constant, is relatively compact in C([0, T ]; B) if r > 1.

Lemma 5. (See [28, Lemma 6, p. 1098]) Let g ∈ W 1,1 (0, T ) and k ∈ L1 (0, T ) satisfy

dg
≤ F (g) + k in [0, T ] , g(0) ≤ g0
dt

where F is bounded on bounded sets from R into R, that is,

∀a > 0 ∃A > 0 such that |x| ≤ a ⇒ |F (x)| ≤ A.

Then for every ε > 0, there exists Tε > 0 independent of g such that

g(t) ≤ g0 + ε ∀ t ≤ Tε .

Theorem 6. (See Theorem 2.1 in [5,6]) Let Ω ⊂ R3 be an open bounded set with Lipschitz boundary. Given
T > 0, if u0 ∈ H, w0 ∈ L2 (Ω), ρ0 ∈ L∞ (Ω), ρ0 ≥ 0, and f , g ∈ L1 (0, T ; L2 (Ω)), then there exist

u ∈ L2 (0, T ; V ), w ∈ L2 (0, T ; H 10 (Ω)), p ∈ W −1,∞ (0, T ; L2 (Ω)), ρ ∈ L∞ (0, T ; L∞ (Ω)),

such that

ρu, ρw ∈ L∞ (0, T ; L2 (Ω)) ∩ N 4 ,2 (0, T, W −1,3 (Ω)),

1

inf ρ0 ≤ ρ(x, t) ≤ sup ρ0 ,

Ω Ω

satisfying the equations (1) in Ω × (0, T ), the boundary conditions (2), and the weak initial conditions
JID:YJMAA AID:22848 /FLA Doctopic: Partial Differential Equations [m3L; v1.250; Prn:8/01/2019; 12:54] P.5 (1-20)
P. Braz e Silva et al. / J. Math. Anal. Appl. ••• (••••) •••–••• 5

ρ t=0 = ρ0 ,
⎛ ⎞
 
⎝ ρu · v dx⎠ (0) = ρ0 u0 · v dx, ∀ v ∈ V ,
Ω Ω
⎛ ⎞
 
⎝ ρw · z dx⎠ (0) = ρ0 w0 · z dx, ∀ z ∈ H 10 (Ω).
Ω Ω

Our first result is the following

Proposition 7. Let (ρ, u, p, w) be a solution given by Theorem 6. We give improved regularity to this solution:

(i) If ρu, ρw ∈ C ([0, T ] ; X), where X is any Banach space such that L2 (Ω) → X, then

(ρu) (t) , (ρw) (t) ∈ L2 (Ω) for all t ∈ [0, T ] ,

 
[(ρu) (0) − ρ0 u0 ] · v dx = 0 ∀ v ∈ H, [(ρw) (0) − ρ0 w0 ] · z dx = 0 ∀ z ∈ L2 (Ω),
Ω Ω

(ρu) (0) = ρ0 u0 + ∇λ where λ ∈ H 1 (Ω) , and (ρw) (0) = ρ0 w0 .

 
(ii) If u ∈ C ([0, T ] ; H) and w ∈ C [0, T ] ; L2 (Ω) , then
 
ρu, ρw ∈ C [0, T ] ; H −1 (Ω) ,

(ρu) (0) = ρ (0) u (0) = ρ0 u0 in Ω,

(ρw) (0) = ρ (0) w (0) = ρ0 w0 in Ω.

 
If in addition u ∈ C ([0, T ] ; V ) and w ∈ C [0, T ] ; H 10 (Ω) , then
 
ρu, ρw ∈ C [0, T ] ; W −1,∞ (Ω) .
 
(iii) If u ∈ C ([0, T ] ; H), w ∈ C [0, T ] ; L2 (Ω) and ρ0 > 0 a.e. in Ω, then

u (0) = u0 , w (0) = w0 in Ω.

Proof. In what follows, we will work as in [28, Proposition 13, p. 1110]. We will only prove the results for
u, since the proof of results for the angular velocity w is similar.

(i) By hypothesis and by Theorem 6, we obtain

 
ρu ∈ C ([0, T ] ; X) ∩ L∞ 0, T ; L2 (Ω) .

By Lemma 1, we concluded that ρu is weakly continuous with values in L2 (Ω). Thus, (ρu) (t) ∈
L2 (Ω), ∀ t ∈ [0, T ]. In particular,

(ρu)(t) → (ρu)(0) in L2 (Ω) as t → 0+ .

Now, since (ρu) (0) ∈ L2 (Ω), we have

JID:YJMAA AID:22848 /FLA Doctopic: Partial Differential Equations [m3L; v1.250; Prn:8/01/2019; 12:54] P.6 (1-20)
6 P. Braz e Silva et al. / J. Math. Anal. Appl. ••• (••••) •••–•••

⎛ ⎞
  
(ρu) (0) · v dx = ⎝ ρu · v ⎠ (0) dx = ρ0 u0 · v dx, ∀v ∈ V .
Ω Ω Ω

Therefore

[(ρu) (0) − ρ0 u0 ] · v dx = 0, ∀v ∈ V .
Ω

By density, this is satisfied for all v ∈ H. Finally, by Lemma 2, there exists λ ∈ L2 (Ω) such that

(ρu) (0) − ρ0 u0 = ∇λ.

Since (ρu) (0) and ρ0 u0 ∈ L2 (Ω), we conclude that λ ∈ H 1 (Ω).

(ii) Since u ∈ C ([0, T ] ; H), we have u (0) ∈ H and

u (t) → u (0) in H, as t → 0+ . (4)

 
By Theorem 6, ρ ∈ C [0, T ] ; W −1,∞ (Ω) and ρ (0) = ρ0 . Therefore,

ρ (t) → ρ0 in W −1,∞ (Ω) strong.

Again by Theorem 6, we have ρ ∈ L∞ (Ω × (0, T )). Thus, ρ (t)L∞ is bounded a.e. in (0, T ). Since
L∞ (Ω) → W −1,∞ (Ω), it follows that ρ is weakly continuous on [0, T ] into L∞ (Ω), i.e.,


η, ρ (t) → η, ρ0 , ∀ η ∈ (L∞ (Ω)) .

Given g ∈ L1 (Ω), we define the functional φg : L∞ (Ω) → R by φg (h) := Ω
g · h dx. Thus
 
g · ρ (t) dx → g · ρ0 dx.
Ω Ω

Hence

∗
ρ (t)  ρ0 in L∞ (Ω) , as t → 0+ . (5)

By Minkowski’s inequality, for all g ∈ L2 (Ω), we have

ρ (t) u (t) g − ρ0 u (0) g1 ≤ (ρ (t) − ρ (0)) u (0) g1 + (u (t) − u (0)) ρ (t) g1 .

By convergence (5), we conclude that

(ρ (t) − ρ (0)) u (0) g1 → 0, as t → 0+ .

Moreover, note that

(u (t) − u (0)) ρ (t) g1 ≤ u (t) − u (0)2 ρ (t)∞ g2 .

Now, by convergence (4), we obtain

ρ (t) u (t)  ρ0 u (0) in L2 (Ω). (6)

JID:YJMAA AID:22848 /FLA Doctopic: Partial Differential Equations [m3L; v1.250; Prn:8/01/2019; 12:54] P.7 (1-20)
P. Braz e Silva et al. / J. Math. Anal. Appl. ••• (••••) •••–••• 7

Since the imbedding L2 (Ω) → H −1 (Ω) is compact, we have

ρ (t) u (t) → ρ0 u (0) in H −1 (Ω) strong. (7)

Thus
 
ρu ∈ C [0, T ] ; H −1 (Ω) .

Now, let u be continuous on [0, T ] into V . Similarly to obtaining (6), we get

ρ (t) u (t)  ρ0 u (0) in L6 (Ω).

Since the imbedding L6 (Ω) → W −1,∞ (Ω) is compact, we conclude that

 
ρu ∈ C [0, T ] ; W −1,∞ (Ω) .

Since

[(ρu) (0) − ρ0 u0 ] · v dx = 0, ∀v ∈ H
Ω

and ρ (0) u (0) = ρ0 u (0), we have


[ρ0 u (0) − ρ0 u0 ] · v dx = 0, ∀ v ∈ H.
Ω


Choosing v = u (0) − u0 ∈ H, we obtain ρ0 |u (0) − u0 |2 dx = 0. Since ρ0 ≥ 0, we get
Ω

ρ0 |u (0) − u0 |2 = 0 a.e.in Ω, (8)

i.e., ρ0 u (0) = ρ0 u0 a.e. in Ω, and thus

(ρu) (0) = ρ (0) u (0) = ρ0 u0 a.e. in Ω.

(iii) Since ρ0 > 0 a.e. in Ω, by (8) we obtain

u (0) = u0 a.e. in Ω.

This completes the proof of Proposition 7. 2

2.3. Main results

Our aim is to prove the following theorems:

Theorem 8 (Existence of local weak solutions). Let Ω ⊂ R3 be an open bounded set such that ∂Ω is W 2,∞
 
(or C 2 ). Assume that u0 ∈ V , w0 ∈ H 10 (Ω), ρ0 ∈ L∞ (Ω), ρ0 ≥ 0 and f , g ∈ L2 0, T ; L2 (Ω) .
JID:YJMAA AID:22848 /FLA Doctopic: Partial Differential Equations [m3L; v1.250; Prn:8/01/2019; 12:54] P.8 (1-20)
8 P. Braz e Silva et al. / J. Math. Anal. Appl. ••• (••••) •••–•••

(i) There exist (ρ, u, p, w) satisfying the properties stated in Theorem 6 and, for some T∗ > 0, these
functions are such that
 
u ∈ L2 0, T∗ ; H 2 (Ω) ∩ L∞ (0, T∗ ; V ) ,
   
w ∈ L2 0, T∗ ; H 2 (Ω) ∩ L∞ 0, T∗ ; H 10 (Ω) ,
 
p ∈ L2 0, T∗ ; L6 (Ω) ,
 
ρu, ρw ∈ C [0, T∗ ] ; W −1,∞ (Ω) ,

∂ρu ∂ρw  
, ∈ L2 0, T∗ ; W −1,6 (Ω) ,
∂t ∂t
∂ρ  
∈ L∞ 0, T∗ ; W −1,6 (Ω) ,
∂t
(ρu) (0) = ρ0 u0 + ∇λ, where λ ∈ W 1,6 (Ω) ,

(ρw) (0) = ρ0 w0 .

The time T∗ depends only on the data Ω, f , g, ρ0 , u0 and w0 .

(ii) If in addition

1
∈ L6/5 (Ω),
ρ0

then

t → u(t) is continuous at t = 0 into H,

t → w(t) is continuous at t = 0 into L2 (Ω),

u(0) = u0 , w(0) = w0 ,

(ρu)(0) = ρ0 u0 , (ρw)(0) = ρ0 w0 .

(iii) If moreover

inf ρ0 > 0,
Ω

then
 
u ∈ C ([0, T∗ ] ; V ) , w ∈ C [0, T∗ ] ; H 10 (Ω) ,

∂u ∂w  
∈ L2 (0, T∗ ; H) , ∈ L2 0, T∗ ; L2 (Ω) .
∂t ∂t

With respect to the global existence of weak solutions, we prove

Theorem 9 (Existence of global weak solutions). Under the hypotheses in Theorem 8 and assuming that
 
f , g ∈ L∞ 0, ∞; H 1 (Ω) and that the norms

u0 H 1 , w0 H 1 , f L∞ (0,∞;L2 (Ω)) , gL∞ (0,∞;L2 (Ω))

0 0
JID:YJMAA AID:22848 /FLA Doctopic: Partial Differential Equations [m3L; v1.250; Prn:8/01/2019; 12:54] P.9 (1-20)
P. Braz e Silva et al. / J. Math. Anal. Appl. ••• (••••) •••–••• 9

are sufficiently small, the weak solution (ρ, u, p, w) of problem (1)–(3) exists globally in time. Moreover,
there exists a positive constant M such that
 
2 2
sup ∇u(·, t) + ∇w(·, t) ≤ M.
t≥0

3. Proof of the theorems

3.1. Proof of Theorem 8

In what follows, we will work as in [28, Theorem 14, p. 1110].

Proof of part (i)

3.1.1. Suitable basis for V and L2 (Ω)

We denote by P the orthogonal projection from L2 (Ω) onto H, induced by the usual Helmholtz decom-
position of L2 (Ω), i.e., L2 (Ω) = H ⊕ H ⊥ . The operator P Δ : V ∩ H 2 (Ω) ⊂ H → H is a bijection of
V ∩ H 2 (Ω) on H that is self-adjoint and whose inverse is compact. It is characterized by the relation

(P Δw, v) = −(∇w, ∇v), for all w ∈ V ∩ H 2 (Ω), v ∈ V .

Its eigenfunctions and eigenvalues are denoted by vm and λm , respectively, where

v m ∈ V ∩ H 2 (Ω), P Δv m = −λm v m with 0 < λ1 ≤ λ2 ≤ · · · , λj → ∞.

Moreover, it is well known that {vm }∞

m=1 form an orthogonal complete system in H and a basis in V . From
Sobolev’s theorem

P Δv m = −λm v m ∈ C(Ω).

By [18, Theorem 2, p. 67] v m ∈ W 2,q (Ω), for all q < ∞, which implies

v m ∈ C 1 (Ω).

We also consider the Laplace operator −Δ with domain H 10 (Ω) ∩ H 2 (Ω). We denote by z m and γm ,
respectively, the eigenfunctions and eigenvalues of −Δ (with Dirichlet boundary condition), where

z m ∈ H 10 (Ω) ∩ C ∞ (Ω), −Δz m = γm z m with 0 < γm → +∞.

It will be assumed that {z m }∞ 2

m=1 form an orthogonal complete system in L (Ω) (see [10, Theorem 9.31,
p. 311]).
Let V m be the subspace of V spanned by {v 1 , · · · , v m } and W m the subspace of L2 (Ω) generated by
{z , · · · , z m }.
1

3.1.2. Approximate problem

The solutions of problem (1)–(3) can be obtained by using the semi-Galerkin method determined by the
spaces V m and W m . More precisely, for each fixed m ∈ N, we consider the following finite dimensional
problem: find Tm ∈ (0, T ], ρm ∈ C 1 (Ω × [0, Tm ]), um ∈ C 1 ([0, Tm ]; V m ) and wm ∈ C 1 ([0, Tm ]; W m )
satisfying the equations
JID:YJMAA AID:22848 /FLA Doctopic: Partial Differential Equations [m3L; v1.250; Prn:8/01/2019; 12:54] P.10 (1-20)
10 P. Braz e Silva et al. / J. Math. Anal. Appl. ••• (••••) •••–•••


t + (u · ∇) u − f ) · v + (μ + μr ) ∇u · ∇v − 2μr w · curl v) dx = 0,
(ρm (um m m m m
(9)
Ω

t + (u · ∇) w − g) · z + (ca + cd ) ∇w · ∇z
(ρm (wm m m m

+ (c0 + cd − ca ) div wm · div z + 4μr wm · z − 2μr curl um · z) dx = 0, (10)

t + u · ∇ρ = 0,
ρm m m
(11)

for t ∈ (0, Tm ) and for all v ∈ V m , z ∈ W m , and the initial conditions

ρm |t=0 = ρm
0 , u |t=0 = u0 and w |t=0 = w 0 in Ω,
m m m m
(12)

where ρm m m
0 , u0 and w 0 are any functions satisfying

0 ∈V , 0 → u0 in L (Ω),
m 2
um um (13)

0 ∈W , 0 → w 0 in L (Ω),
m 2
wm wm (14)
∗ ∞
0 ∈ C (Ω),
ρm 1
ρm
0  ρ0 in L (Ω), (15)

1 1
+ inf ρ0 ≤ ρm
0 ≤ + sup ρ0 . (16)
m Ω m Ω

We say that the triplet (ρm , um , wm ), solution of problem (9)–(12), is an m-th approximate solution.
Through the characteristics method, it is not difficult to prove that the Cauchy problem (9)–(12) admits
exactly one solution (ρm , um , wm ) defined in a time interval [0, Tm ]. Actually it is possible to prove that
Tm = T for all m ∈ N.

Remark 10. It is well-known that, according to the classical method of characteristics, the solution of initial
value problem

t + u · ∇ρ = 0,
ρm m m
ρm (0) = ρm
0

is given by

0 (y (s; x, t)) |s=0 ,

ρm (x, t) = ρm m

where the particle trajectory y m (s; x, t) is defined by the solution of the Cauchy problem

dy m
= um (y m , s), y m |s=t = x .
ds

From this and by (16), we get the bound

1 1
+ inf ρ0 ≤ ρm ≤ + sup ρ0 ≤ 1 + sup ρ0 =: β. (17)
m Ω m Ω Ω
JID:YJMAA AID:22848 /FLA Doctopic: Partial Differential Equations [m3L; v1.250; Prn:8/01/2019; 12:54] P.11 (1-20)
P. Braz e Silva et al. / J. Math. Anal. Appl. ••• (••••) •••–••• 11

3.1.3. Estimates for the approximate solutions

Let (ρm , um , wm ) be an approximate solution of problem (9)–(12). Taking v = 2um
t in equation (9),
z = 2wm t in equation (10) and adding the resulting equations, one obtains
   
d d
2 ρm |um
t | dx + 2
2
ρm |wm
t | dx + μ
2
|∇um |2 dx + (ca + cd ) |∇wm |2 dx
dt dt
Ω Ω Ω Ω
 
d d
+ (c0 + cd − ca ) |div w m |2 dx + μr |curl um − 2wm |2 dx
dt dt
Ω Ω
 
= −2 ρm (um · ∇) um · um
t dx − 2 ρm (um · ∇) wm · wm
t dx
Ω Ω
 
+2 ρm f · um
t dx + 2 ρm g · wm
t dx. (18)
Ω Ω

By Cauchy–Schwarz and Young inequalities, we bound each term on the right-hand side of identity (18) as
follows:
  
−2 ρm (um · ∇)um · um t dx = 2 −(ρm 1/2
) (um
· ∇)u m
, (ρm 1/2 m
) ut
Ω
 2  2
 1/2   
≤ 2 (ρm ) (um · ∇)um  + 1/2 (ρm )1/2 um
t  ,
  
−2 ρm (um · ∇) wm · wm
t dx = 2 −(ρm 1/2
) (um
· ∇)w m
, (ρm 1/2 m
) w t
Ω
 2  2
 1/2   1/2 
≤ 2 (ρm ) (um · ∇)wm  + 1/2 (ρm ) wm t  ,
  
m 1/2 1/2
2 ρm f · um
t dx = 2 (ρ ) f , (ρm ) um t
Ω
 2
2  1/2 
≤ 2β f  + 1/2 (ρm ) um t  ,
  
m 1/2 1/2
2 ρm g · wm
t dx = 2 (ρ ) g, (ρm ) wm t
Ω
 2
2  1/2 
≤ 2β g + 1/2 (ρm ) wm t  .

Therefore, from identity (18) one gets

   
d d
ρ m
|um
t |
2
dx + ρ m
|wm
t |
2
dx + μ |∇u | dx + (ca + cd )
m 2
|∇wm |2 dx
dt dt
Ω Ω Ω Ω
 
d d
+ (c0 + cd − ca ) |div wm |2 dx + μr |curl um − 2wm |2 dx
dt dt
Ω Ω
 2  2
 1/2   1/2  2 2
≤ 2 (ρm ) (um · ∇)um  + 2 (ρm ) (um · ∇)wm  + 2β f  + 2β g . (19)

Now, taking v = P Δum in equation (9), with  > 0 to be chosen later on, we find
JID:YJMAA AID:22848 /FLA Doctopic: Partial Differential Equations [m3L; v1.250; Prn:8/01/2019; 12:54] P.12 (1-20)
12 P. Braz e Silva et al. / J. Math. Anal. Appl. ••• (••••) •••–•••

  
− (μ + μr ) ∇u · ∇P Δu dx = 
m m
ρ m
um
t · P Δu dx + 
m
ρm (um · ∇) um · P Δum dx
Ω Ω Ω
 
− 2μr curl wm · P Δum dx −  ρm f · P Δum dx. (20)
Ω Ω

Using again Cauchy–Schwarz and Young inequalities, we bound the term in the right-hand side of identity
(20) as follows:

t · P Δu dx = (ρ ut , P Δu )
ρm um m m m m

Ω
 
−1  m 1/2 m 2 1 2
≤ β (μ + μr ) (ρ ) ut  + (μ + μr ) P Δum  ,
 4
ρm (um · ∇) um · P Δum dx = (ρm (um · ∇) um , P Δum )
Ω
 2 1
−1  m 1/2 m  2
≤ β (μ + μr ) (ρ ) (u · ∇)um  + (μ + μr ) P Δum  ,
 4
−2μr curl wm · P Δum dx = 2μr (curl wm , −P Δum )
Ω
−1 2 1 2
≤ 8μ2r (μ + μr ) curl wm  + (μ + μr ) P Δum  ,
 8
− ρm f · P Δum dx = (ρm f , −P Δum )
Ω
−1 2 1 2
≤ β 2 (μ + μr ) f  + (μ + μr ) P Δum  .
4

Applying the estimates above to identity (20) we obtain

 2  2
 2 −1  1/2  −1  m 1/2 m
(μ + μr ) P Δum  ≤ β (μ + μr ) (ρm ) um t  + β (μ + μr ) (ρ ) (um
· ∇)u 
8
−1 2 −1 2
+ β 2 (μ + μr ) f  + 8μ2r (μ + μr ) curl wm  . (21)

We can find a similar differential inequality for wm . Indeed, if we take z = δΔwm in Eq. (10), with δ > 0
to be chosen later on, we see that
 
−δ (c0 + cd ) ∇wm · ∇Δwm dx − δ (c0 + cd − ca ) div wm · div Δwm dx
Ω Ω
  
−4δμr wm · Δw m dx = δ t · Δw dx + δ
ρm wm m
ρm (um · ∇) wm · Δwm dx
Ω Ω Ω
 
−2δμr curl um · Δwm dx − δ ρm g · Δwm dx. (22)
Ω Ω

Now, bound each term on the right-hand side of (22) as follows:


t · Δw dx = (ρ w t , Δw )
ρm wm m m m m

Ω
 
−1  m 1/2 m 2 1 2
≤ β (ca + cd ) (ρ ) wt  + (ca + cd ) Δwm  ,
4
JID:YJMAA AID:22848 /FLA Doctopic: Partial Differential Equations [m3L; v1.250; Prn:8/01/2019; 12:54] P.13 (1-20)
P. Braz e Silva et al. / J. Math. Anal. Appl. ••• (••••) •••–••• 13


ρm (um · ∇) wm · Δwm dx = (ρm (um · ∇) wm , Δwm )
Ω
 2 1
−1  m 1/2 m  2
≤ β (ca + cd ) (ρ ) (u · ∇)wm  + (ca + cd ) Δwm  ,
 4
−2μr curl um · Δwm dx = 2μr (curl um , −Δwm )
Ω
−1 2 1 2
≤ 8μ2r (ca + cd ) curl um  + (ca + cd ) Δwm  ,
 8
− ρm g · Δwm dx = (ρm g, −Δwm )
Ω
−1 2 1 2
≤ β 2 (ca + cd ) g + (ca + cd ) Δwm  .
4

Therefore, using the estimates above into identity (22), one obtains

δ 2 2
(ca + cd ) Δwm  + δ(c0 + cd − ca )∇div wm 2 + 4δμr ∇wm 
8
 2  2
−1  1/2  −1  m 1/2 m
≤ δβ (ca + cd ) (ρm ) wm t  + δβ (c a + cd ) (ρ ) (um
· ∇)w 
−1 2 −1 2
+ δβ 2 (ca + cd ) g + 8δμ2r (ca + cd ) curl um  . (23)

Adding inequalities (21) and (23), we conclude that

 2 δ 2
(μ + μr ) P Δum  + (ca + cd ) Δwm 
8 8
 2  2
−1  1/2  −1  m 1/2 
≤ β (μ + μr ) (ρm ) um t  + δβ (ca + cd ) (ρ ) wmt 

 2  2
−1  m 1/2 m  −1  1/2 
+ β (μ + μr ) (ρ ) (u · ∇)um  + δβ (ca + cd ) (ρm ) (um · ∇)wm 
−1 2 −1 2
+ β 2 (μ + μr ) f  + δβ 2 (ca + cd ) g
−1 2 −1 2
+ 8μ2r (μ + μr ) curl wm  + 8δμ2r (ca + cd ) curl um  . (24)

Now, adding estimates (19) and (24) and choosing

μ + μr ca + cd
= , δ= ,
2β 2β

we obtain the following:

1   
 m 1/2 m 2 1  m 1/2 m 2 (μ + μr )2 2 (ca + cd )2 2
(ρ ) ut  + (ρ ) wt  + P Δum  + Δwm  (25)
2 2 16β 16β
d
2 2 2 2

+ μ ∇um  + (ca + cd ) ∇wm  + (c0 + cd − ca ) div wm  + μr curl um − 2wm 
dt
5  1/2
2 5 
  1/2
2
 2 2
≤ (ρm ) (um · ∇)um  + (ρm ) (um · ∇)wm  + 4μ2r curl um  + 4μ2r curl wm 
2 2
5β 2 5β 2
+ f  + g .
2 2
JID:YJMAA AID:22848 /FLA Doctopic: Partial Differential Equations [m3L; v1.250; Prn:8/01/2019; 12:54] P.14 (1-20)
14 P. Braz e Silva et al. / J. Math. Anal. Appl. ••• (••••) •••–•••

Note that, by standard interpolation, Hölder and Young inequalities, one has

5 m 1/2 m
2
 5 2 2
(ρ ) (u · ∇)um  ≤ β um 6 ∇um 3
2 2
5 3
≤ Cβ ∇um  P Δum 
2
50β 3 C 2 m 6 (μ + μr )2 2
≤ ∇u  + P Δum  , (26)
(μ + μr )2 32β
5 m 1/2 m
2
 5 2 2
(ρ ) (u · ∇)wm  ≤ β um 6 ∇wm 3
2 2
5 2
≤ Cβ ∇um  ∇wm  Δwm 
2
50β 3 C 2 m 4 m 2 (ca + cd )2 2
≤ ∇u  ∇w  + Δwm  . (27)
(ca + cd )2 32β

Using bounds (26) and (27) in inequality (25), one gets

1   
 m 1/2 m 2 1  m 1/2 m 2 (μ + μr )2 2 (ca + cd )2 2
(ρ ) ut  + (ρ ) wt  + P Δum  + Δwm 
2 2 32β 32β
d
2 2 2 2

+ μ ∇um  + (ca + cd ) ∇wm  + (c0 + cd − ca ) div wm  + μr curl um − 2wm 
dt



2 2 2 2 6 6
≤ C f  + g + C ∇um  + ∇wm  + ∇um  + ∇wm  . (28)

 
Now, let us bound ∇um (0)2 and ∇wm (0)2 . Let u0,i be the coordinates of u0 in the basis v i , i.e.,

u0 = u0,i v i ∈ V ,
i

and choose


m
um
0 = u0,i v i ∈ V m .
i=1

0 → u0 in H. Thus, hypothesis (13) holds. Furthermore

As this basis is orthogonal in H, um
 
∇v i , ∇v j = 0, if i = j.

Hence
 
m  
∇um (0)2 = |∇um
0 | dx =
2
|u0,i |2 |∇v i |2 dx ≤ |∇u0 |2 dx = ∇u0 2 . (29)
Ω i=1 Ω Ω

Analogously, we can write


w0 = w0,i z i ∈ H 10 (Ω),
i

 
where w0,i are the coordinates of w0 in the basis z i . Choose
JID:YJMAA AID:22848 /FLA Doctopic: Partial Differential Equations [m3L; v1.250; Prn:8/01/2019; 12:54] P.15 (1-20)
P. Braz e Silva et al. / J. Math. Anal. Appl. ••• (••••) •••–••• 15


m
wm
0 = w0,i z i ∈ W m .
i=1

 
0 → w 0 in L (Ω), and then hypothesis (14) is satisfied.
Since z i is a basis orthogonal in L2 (Ω), wm 2

Moreover, since
 i 
∇z , ∇z j = 0, if i = j,

we have
 
m  
∇wm (0)2 = |∇wm
0 | dx =
2
|w0,i |2 |∇z i |2 dx ≤ |∇w0 |2 dx = ∇w0 2 . (30)
Ω i=1 Ω Ω

Due to the estimates (29), (30) and Lemma 5, the inequality (28) implies that there exists T∗ > 0 (inde-
pendent of m) such that



2 2 2 2 2 2
∇um  + ∇wm  + div wm  + curl um − 2wm  (t) ≤ C ∇u0  + ∇w0  + 1, ∀ t ∈ [0, T∗ ] .

Therefore

um is bounded in L∞ (0, T∗ ; V ) (31)

and
 
wm is bounded in L∞ 0, T∗ ; H 10 (Ω) . (32)

Integrating inequality (28) with respect to time from 0 to T∗ , we conclude that

1/2  
(ρm ) um
t is bounded in L2 0, T∗ ; L2 (Ω) , (33)
1/2  
(ρm ) wm t is bounded in L2 0, T∗ ; L2 (Ω) , (34)
 
P Δum is bounded in L2 0, T∗ ; L2 (Ω) , (35)
 
Δwm is bounded in L2 0, T∗ ; L2 (Ω) . (36)

Thus
 
um is bounded in L2 0, T∗ ; H 2 (Ω) ∩ L∞ (0, T∗ ; V ) , (37)
   
wm is bounded in L2 0, T∗ ; H 2 (Ω) ∩ L∞ 0, T∗ ; H 10 (Ω) . (38)

3.1.4. Passing to the limit and additional properties of the solutions

By bounds (37), (38), Alaoglu’s theorem and reflexivity of Lp (Ω) spaces, we have
 
u ∈ L∞ (0, T∗ ; V ) ∩ L2 0, T∗ ; H 2 (Ω)

and
   
w ∈ L∞ 0, T∗ ; H 10 (Ω) ∩ L2 0, T∗ ; H 2 (Ω) ,
JID:YJMAA AID:22848 /FLA Doctopic: Partial Differential Equations [m3L; v1.250; Prn:8/01/2019; 12:54] P.16 (1-20)
16 P. Braz e Silva et al. / J. Math. Anal. Appl. ••• (••••) •••–•••

where u and w are the limits of um and wm , respectively. Since V → H 10 (Ω) → L6 (Ω), by (31) we
 
conclude that um is bounded in L∞ 0, T∗ ; L6 (Ω) . Similarly, we conclude that wm also is bounded in
 
L∞ 0, T∗ ; L6 (Ω) . Therefore, by the transport equation (11), we obtain

∂ρm  
is bounded in L∞ 0, T∗ ; W −1,6 (Ω) .
∂t
We also have that ρ, which is the limit of ρm , satisfies

∂ρ  
∈ L∞ 0, T∗ ; W −1,6 (Ω) .
∂t
 
Since um is bounded in L2 0, T∗ ; H 2 (Ω) and H 2 (Ω) → W 1,6 (Ω), it follows that
 
um is bounded in L2 0, T∗ ; W 1,6 (Ω) .

Analogously, we have
 
wm is bounded in L2 0, T∗ ; W 1,6 (Ω) .

By part (iii) of Lemma 3, the product W 1,6 (Ω) × W −1,6 (Ω) → W −1,6 (Ω) is continuous. Thus

∂ρm  
um is bounded in L2 0, T∗ ; W −1,6 (Ω) ,
∂t
∂ρ m  
wm is bounded in L2 0, T∗ ; W −1,6 (Ω) .
∂t
Since ρm is bounded in L∞ (0, T∗ ; L∞ (Ω)), by bounds (33) and (34) it follows that
 
ρm um
t is bounded in L
2
0, T∗ ; L2 (Ω) ,
 
ρm w m
t is bounded in L
2
0, T∗ ; L2 (Ω) .

Moreover, by the embedding L2 (Ω) → W −1,6 (Ω), we get

 
ρm um
t is bounded in L
2
0, T∗ ; W −1,6 (Ω) ,
 
ρm wmt is bounded in L
2
0, T∗ ; W −1,6 (Ω) .

Thus
∂ρu ∂ρw  
, ∈ L2 0, T∗ ; W −1,6 (Ω) .
∂t ∂t
     
Moreover, we have ρu ∈ L∞ 0, T∗ ; L6 (Ω) , since u ∈ L∞ 0, T∗ ; H 10 (Ω) → L∞ 0, T∗ ; L6 (Ω) and ρ ∈
 
L∞ (0, T∗ ; L∞ (Ω)). Similarly ρw ∈ L∞ 0, T∗ ; L6 (Ω) . By part (ii) of Lemma 4, one gets
 
ρu, ρw ∈ C [0, T∗ ] ; W −1,6 (Ω) .

Now, observe that:

     
• div (ρuu) ∈ L2 0, T∗ ; W −1,6 (Ω) , since ρu ∈ L∞ 0, T∗ ; L6 (Ω) and u ∈ L2 0, T∗ ; W 1,6 (Ω) ;
   
• Δu ∈ L2 0, T∗ ; W −1,6 (Ω) , since u ∈ L∞ (0, T∗ ; V ) → L2 (0, T∗ ; V ) → L2 0, T∗ ; H 10 (Ω) →
 
L2 0, T∗ ; L6 (Ω) ;
JID:YJMAA AID:22848 /FLA Doctopic: Partial Differential Equations [m3L; v1.250; Prn:8/01/2019; 12:54] P.17 (1-20)
P. Braz e Silva et al. / J. Math. Anal. Appl. ••• (••••) •••–••• 17

     
• ρf ∈ L2 0, T∗ ; W −1,6 (Ω) , since ρf ∈ L2 0, T∗ ; L2 (Ω) → L2 0, T∗ ; W −1,6 (Ω) ;
       
• curl w ∈ L2 0, T∗ ; W −1,6 (Ω) , since w ∈ L∞ 0, T∗ ; H 10 (Ω) → L2 0, T∗ ; H 10 (Ω) → L2 0, T∗ ; L6 (Ω) .
   
Hence, we conclude that ∇p ∈ L2 0, T∗ ; W −1,6 (Ω) , and this implies that p ∈ L2 0, T∗ ; L6 (Ω) .

Remark 11. We denote by u ⊗ v the tensor product of u and v, i.e., (u ⊗ v)ij := vi uj . We set ρuv :=
ρu ⊗ v = (ρvi uj )i,j . Moreover, the i-th component of vector div (ρuv) in cartesian coordinates is given by

  3
∂(ρvi uj )
div (ρuv) i = .
j=1
∂xj

Thus, by Proposition 7, we have

(ρu) (0) = ρ0 u0 + ∇λ, (ρw) (0) = ρ0 w0 with λ ∈ H 1 (Ω) .

In particular, λ ∈ W 1,6 (Ω). This finishes the proof of part (i) of Theorem 8.

Proof of part (ii) By part (i) of Theorem 8,

∂ρ  
∈ L∞ 0, T ; W −1,6 (Ω)
∂t
and by Theorem 6, ρ (0) = ρ0 . Then,

ρ0 − ρ (t)−1,6 ≤ Ct,

for 0 ≤ t ≤ T∗ . Since w (t) is bounded in H 1 (Ω) for 0 ≤ t ≤ T∗ it follows, by part (iii) of Lemma 3, that

ρ0 w (t) − ρ (t) w (t)−1,3 ≤ ρ0 − ρ (t)−1,6 w (t)1,2 ≤ Ct.

Thus, the application

t → ρ0 w (t) − ρ (t) w (t)

is continuous in t = 0, with values in W −1,3 (Ω). Since W −1,6 (Ω) → W −1,3 (Ω) and ρw ∈ C([0, T∗ ];
W −1,3 (Ω)), it follows that

t → ρ0 w (t)
   
is continuous at zero into W −1,3 (Ω). Since w ∈ L∞ 0, T∗ ; H 10 (Ω) , we have that ρ0 w ∈ L∞ 0, T∗ ; L6 (Ω) .
By Lemma 1, since L6 (Ω) → W −1,3 (Ω), we conclude that

t → ρ0 w (t)

1  
is continuous at zero into L6 (Ω) weak. As, by hypothesis, ∈ L6/5 (Ω) = L6 (Ω) , we get that
ρ0

t → w (t)
 
is weakly continuous at zero into L1 (Ω). On the other hand, since w ∈ L∞ 0, T∗ ; H 10 (Ω) , by Lemma 1, it
follows that
JID:YJMAA AID:22848 /FLA Doctopic: Partial Differential Equations [m3L; v1.250; Prn:8/01/2019; 12:54] P.18 (1-20)
18 P. Braz e Silva et al. / J. Math. Anal. Appl. ••• (••••) •••–•••

t → w (t)

is weakly continuous at zero into L2 (Ω). Thus

(ρw) (0) = ρ (0) w (0) = ρ0 w0 ,

that is,

ρ0 (w (0) − w0 ) = 0 =⇒ w (0) = w0 a.e. in Ω.

For the linear velocity u, we have a similar result. This completes the proof of part (ii) of Theorem 8.

Proof of part (iii) We already know that (see (33) and (34))

1/2 1/2  
(ρm ) um
t , (ρm ) wm
t are bounded in L2 0, T∗ ; L2 (Ω) .

Furthermore, the triplet (ρm , um , wm ) satisfy all the estimates obtained in the proof of Theorem 6. In
particular, ρm ≥ inf Ω ρ0 . By assumption inf Ω ρ0 > 0, so we conclude that
 
um
t , wm
t are bounded in L2 0, T∗ ; L2 (Ω) .
 
t ∈ C ([0, T∗ ] ; V ) and w t ∈ C [0, T∗ ] ; H 0 (Ω) . Thus,
1
We also have that um m

 
um 2
t is bounded in L (0, T∗ ; H) , wm
t is bounded in L
2
0, T∗ ; L2 (Ω) ,

and hence
 
ut ∈ L2 (0, T∗ ; H) and wt ∈ L2 0, T∗ ; L2 (Ω) .
 
Since u, w ∈ L2 0, T∗ ; H 2 (Ω) , it follows that
 
u, w ∈ C [0, T∗ ] ; H 1 (Ω) .
 
Moreover, since u ∈ L∞ (0, T∗ ; V ) and w ∈ L∞ 0, T∗ ; H 10 (Ω) , we have that
 
u ∈ C ([0, T∗ ] ; V ) and w ∈ C [0, T∗ ] ; H 10 (Ω) .

This completes the proof of Theorem 8. 2

3.2. Proof of Theorem 9

Inequality (28) reads

1   
 m 1/2 m 2 1  m 1/2 m 2 (μ + μr )2 2 (ca + cd )2 2
(ρ ) ut  + (ρ ) wt  + P Δum  + Δwm 
2 2 32β 32β
d
2 2 2 2

+ μ ∇um  + (ca + cd ) ∇wm  + (c0 + cd − ca ) div wm  + μr curl um − 2wm 
dt



2 2 2 2 6 6
≤ C f  + g + C ∇um  + ∇wm  + ∇um  + ∇wm  .

Hence, using bounds (29) and (30), one obtains that

JID:YJMAA AID:22848 /FLA Doctopic: Partial Differential Equations [m3L; v1.250; Prn:8/01/2019; 12:54] P.19 (1-20)
P. Braz e Silva et al. / J. Math. Anal. Appl. ••• (••••) •••–••• 19

⎛ ⎞
t
 
∇um (t) + ∇wm (t) ≤ C ⎝∇u0 2 + ∇w0 2 + f (s)2 + g(s)2 ds⎠
2 2

t
 
+C ∇um (s)2 + ∇wm (s)2 + ∇um (s)6 + ∇wm (s)6 ds.
0

This inequality indicates that the approximations (ρm , um , wm ) are defined in the whole interval [0, +∞).
Therefore, if u0 H 1 , w0 H 1 , f L∞ (0,∞;L2 (Ω)) and gL∞ (0,∞;L2 (Ω)) are sufficiently small, there exists
0 0
M > 0 such that
 
2 2
sup ∇um (t) + ∇wm (t) ≤ M < + ∞. (39)
t≥0

Accordingly, (ρ, u, w), which is the limit of (ρm , um , wm ), is globally defined. Now, carrying out the esti-
mates for (ρm , um , wm ) to (ρ, u, w) according to the bound (39), one finishes the proof of Theorem 9. 2

Acknowledgments

P. Braz e Silva was partially supported by CNPq, Brazil, grant #309491/2015-0.

References

[1] S.N. Antontsev, A.V. Kazhikhov, V.N. Monakhov, Boundary Value Problems in Mechanics of Nonhomogeneous Fluids,
Stud. Appl. Math., vol. 22, North-Holland Publishing Co., Amsterdam, 1990.
[2] T. Ariman, M.A. Turk, N.D. Sylvester, On steady and pulsatile flow of blood, J. Appl. Mech. 41 (1974) 1–7.
[3] J.L. Boldrini, M.A. Rojas-Medar, On the convergence rate of spectral approximation for the equations for nonhomogeneous
asymmetric fluids, Math. Model. Numer. Anal. 30 (1996) 123–155.
[4] J.L. Boldrini, M.A. Rojas-Medar, E. Fernández-Cara, Semi-Galerkin approximation and strong solutions to the equations
of the nonhomogeneous asymmetric fluids, J. Math. Pures Appl. 82 (2003) 1499–1525.
[5] P. Braz e Silva, E.G. Santos, Global weak solutions for asymmetric incompressible fluids with variable density, C. R. Acad.
Sci. Paris Ser. I 346 (2008) 575–578.
[6] P. Braz e Silva, E.G. Santos, Global weak solutions for variable density asymmetric incompressible fluids, J. Math. Anal.
Appl. 387 (2012) 953–969.
[7] P. Braz e Silva, E. Fernández-Cara, M.A. Rojas-Medar, Vanishing viscosity for non-homogeneous asymmetric fluids in R3 ,
J. Math. Anal. Appl. 332 (2007) 833–845.
[8] P. Braz e Silva, F.W. Cruz, M.A. Rojas-Medar, Vanishing viscosity for non-homogeneous asymmetric fluids in R3 : the L2
case, J. Math. Anal. Appl. 420 (2014) 207–221.
[9] P. Braz e Silva, F.W. Cruz, M.A. Rojas-Medar, Semi-strong and strong solutions for variable density asymmetric fluids in
unbounded domains, Math. Methods Appl. Sci. 40 (2017) 757–774.
[10] H. Brezis, Functional Analysis, Sobolev Spaces and Partial Differential Equations, Universitext, Springer, New York, 2011.
[11] C. Conca, R. Gormaz, E. Ortega-Torres, M.A. Rojas-Medar, Existence and uniqueness of a strong solution for nonhomo-
geneous micropolar fluids, nonlinear partial differential equations and their applications, in: Collège de France Seminar,
vol. XIV, Paris, 1997/1998, in: Stud. Appl. Math., vol. 31, North-Holland, Amsterdam, 2002, pp. 213–241.
[12] C. Conca, R. Gormaz, E.E. Ortega-Torres, M.A. Rojas-Medar, The equations of non-homogeneous asymmetric fluids: an
iterative approach, Math. Methods Appl. Sci. 25 (2002) 1251–1280.
[13] D.W. Condiff, J.S. Dahler, Fluid mechanical aspects of antisymmetric stress, Phys. Fluids 7 (842) (1964) 842–854.
[14] A.C. Eringen, Theory of micropolar fluids, J. Math. Mech. 16 (1966) 1–18.
[15] C. Ferrari, On lubrication with structured fluids, Appl. Anal. 15 (1983) 127–146.
[16] A.V. Kazhikhov, Solvability of the initial–boundary value problem for the equations of the motion of an inhomogeneous
viscous incompressible fluid, Dokl. Akad. Nauk SSSR 216 (1974) 1008–1010.
[17] J.U. Kim, Weak solutions of an initial boundary value problem for an incompressible viscous fluid with nonnegative density,
SIAM J. Math. Anal. 18 (1987) 89–96.
[18] O.A. Ladyzhenskaya, The Mathematical Theory of Viscous Incompressible Flow, Gordon and Breach Science Publishers,
New York, London, Paris, 1969.
[19] O.A. Ladyzhenskaya, V.A. Solonnikov, The unique solvability of an initial–boundary value problem for viscous incom-
pressible inhomogeneous fluids, in: Boundary Value Problems of Mathematical Physics, and Related Questions of the
Theory of Functions 8, Zap. Nauchn. Sem. Leningrad. Otdel. Mat. Inst. Steklov. (LOMI) 52 (1975) 52–109, 218–219.
[20] J.-L. Lions, On some problems connected with Navier–Stokes equations, in: Nonlinear Evolution Equations, Proc. Sympos.,
Univ. Wisconsin, Madison, WI, 1977, in: Publ. Math. Res. Cent. Univ. Wisconsin, vol. 40, Academic Press, New York,
London, 1978, pp. 59–84.
JID:YJMAA AID:22848 /FLA Doctopic: Partial Differential Equations [m3L; v1.250; Prn:8/01/2019; 12:54] P.20 (1-20)
20 P. Braz e Silva et al. / J. Math. Anal. Appl. ••• (••••) •••–•••

[21] J.-L. Lions, On some questions in boundary value problems of mathematical physics, in: Contemporary Developments in
Continuum Mechanics and Partial Differential Equations, Proc. Internat. Sympos., Inst. Mat., Univ. Fed. Rio de Janeiro,
Rio de Janeiro, 1977, in: North-Holland Math. Stud., vol. 30, North-Holland, Amsterdam, New York, 1978, pp. 284–346.
[22] P.-L. Lions, Mathematical Topics in Fluid Mechanics. Vol. I: Incompressible Models, Oxford Lecture Ser. Math. Appl.,
vol. 3, The Clarendon Press, Oxford University Press, New York, 1996.
[23] G. Lukaszewicz, Micropolar Fluids. Theory and Applications, Model. Simul. Sci. Eng. Technol., Birkhäuser Boston, Inc.,
Boston, MA, 1999.
[24] G. Lukaszewicz, On non-stationary flows of incompressible asymmetric fluids, Math. Methods Appl. Sci. 13 (1990) 219–232.
[25] J. Prakash, P. Sinha, Lubrication theory for micropolar fluids and its application to a journal bearing, Internat. J. Engrg.
Sci. 13 (1975) 217–232.
[26] R. Salvi, The equations of viscous incompressible non-homogeneous fluids: on the existence and regularity, J. Aust. Math.
Soc. B 33 (1991) 94–110.
[27] J. Simon, Existencia de solución del problema de Navier–Stokes con densidad variable (Existence of solution for the variable
density Navier–Stokes problem), Lectures at the University of Sevilla, Spain, 1989 (in Spanish).
[28] J. Simon, Nonhomogeneous viscous incompressible fluids: existence of velocity, density, and pressure, SIAM J. Math. Anal.
21 (1990) 1093–1117.
[29] R. Temam, Navier–Stokes Equations: Theory and Numerical Analysis, Studies in Mathematics and Its Applications, vol. 2,
North-Holland Publishing Company, Amsterdam, New York, Oxford, 1977.