Global Uniform Boundedness of Solutions

arXiv:1710.04622v1 [math.AP] 12 Oct 2017

to viscous 3D Primitive Equations

with Physical Boundary Conditions
1
Ning Ju

Abstract
Global uniform boundedness of solutions to 3D viscous Primitive
Equations in a bounded cylindrical domain with physical boundary con-
dition is proved in space H m for any m > 2. A bounded absorbing
set for the solutions in H m is obtained. These results seem quite dif-
ficult to be proved using the methods recently developed in [8] and
[9]. A completely different approach based on hydrostatic helmholtz
decomposition is presented, which is also applicable to the cases with
other boundary conditions. Several important results about hydro-
static Leray projector are obtained and utilized. These results are
expected useful as well for solving some other problems for 3D viscous
Primitive Equations which may be hard due to non-periodic boundary
conditions.

Keywords: Primitive Equations, physical boundary condition, global

uniform estimates, hydrostatic helmholtz decomposition, hydrostatic
Leray projector.

MSC: 35B40, 35B30, 35Q35, 35Q86.

1 Introduction
Let D be a bounded open subset of R2 with smooth boundary D and

= D (h, 0) R3 ,
1
Department of Mathematics, Oklahoma State University, 401 Mathematical Sciences,
Stillwater OK 74078, USA. Email: ning.ju@okstate.edu,

1
 N. Ju 3D Viscous Primitive Equations 2

where h is a positive constant. Consider in the cylinder the system of

viscous Primitive Equations (PEs) of Geophysical Fluid Dynamics:
Conservation of horizontal momentum:

vt + (v )v + wvz + p + f v + L1 v = 0;

Hydrostatic balance:
pz + = 0;

Continuity equation:
v + wz = 0;

Heat conduction:
t + v + wz + L2 = Q.

The unknowns in the above system of 3D viscous PEs are the fluid
velocity field (v, w) = (v1 , v2 , w) R3 with v = (v1 , v2 ) and v = (v2 , v1 )
being horizontal, the temperature and the pressure p. The Coriolis rotation
frequency f = f0 ( + y) in the -plane approximation and the heat source
Q are given. The differential operators L1 and L2 are defined respectively
as:
Li := i i z2 ,

with positive constants i , i for i = 1, 2. In the above equations, and

denote horizontal gradient and Laplacian:
2
X
:= (x , y ) (1 , 2 ), := x2 + y2 i2 .
i=1

In the sequel, we also denote

3 := (, z ) = (x , y , z ).

The boundary of is partitioned into three parts: = t b l , where

t := {(x, y, z) : z = 0},
b := {(x, y, z) : z = h},
l := {(x, y, z) : (x, y) D}.
 N. Ju 3D Viscous Primitive Equations 3

The following set of physical boundary conditions will be used:

on t : vz + 1 v = 0, w = 0, z + 2 = 0,
on b : v = 0, w = 0, z = 0,
on l : v = 0, n = 0,

where 1 , 2 are non-negative constants and n is the normal vector of l .

Let ps (x, y, t) be the pressure on the top t . Then, the above system of
PEs can be re-written as
Z 0
vt + L1 v + (v )v + wvz + ps + (x, y, , t)d + f v = 0. (1.1)
z

t + L2 + v + wz = Q; (1.2)
Z 0
w(x, y, z, t) = v(x, y, , t)d, (1.3)
z
(vz + 1 v)|z=0 = v|z=h = v|(x,y)D = 0, (1.4)

w|z=h = 0, (1.5)

(z + 2 )|z=0 = z |z=h = n |(x,y)D = 0. (1.6)

The above system of PEs will be solved with suitable initial conditions:

v(x, y, z, 0) = v0 (x, y, z), (x, y, z, 0) = 0 (x, y, z). (1.7)

Assume Q is independent of time for simplicity of discussion, since results

to be presented for autonomous case can be extended to non-autonomous
case with proper modifications.
The notions of weak and strong solutions were introduced in [14], where
existence of weak solutions was proved, though uniqueness of weak solutions
is still not resolved yet. Local (in time) existence and uniqueness of strong
solutions were obtained in [4] and [17]. Global (in time) existence of strong
solutions was proved in [2] for the case when v satisfies a set of Neumann
type boundary conditions. See [10] for a different approach. For the case
when v satisfies the boundary conditions (1.4), see [11] for a proof of global
regularity of strong solutions. For the case when v satisfies a set of related
boundary conditions, see [5] for a different approach. Uniform boundedness
 N. Ju 3D Viscous Primitive Equations 4

of strong solutions was proved in [6] for the case with Neumann type bound-
ary conditions and in [12] for the case with physical boundary conditions.
An initial motivation of this paper is to study uniform boundedness of
H2 solutions to the 3D primitive equations for the case with physical bound-
ary conditions (1.4). Global existence of the H 2 solutions for the case with
periodic boundary condition was proved in [15]. The method of [15] uses ex-
tensively integrations by parts. For non-periodic cases, the boundary terms
arising from those integrations by parts would cause trouble for a priori
estimates. The first success overcoming this difficulty was achieved in [9],
which proved global uniform boundedness of H 2 solutions for the 3D Primi-
tive Equations with v satisfying a set of Neumann type boundary conditions.
An interesting aspect of this approach is that not only it works for both the
periodic and the non-periodic case, it also requires less demanding condition
on Q than [15]. This approach was further improved in [8] which eliminated
all technical restrictions completely. Roughly speaking, the main idea of
[8] and [9] is obtaining uniform boundedness of k(vt , t )kL2 and kvz kH 1 first,
then using elliptic regularity to get uniform boundedness of k(v, )kH 2 , since
boundary conditions of vt , t and vz can be used in this approach. However,
it seems the approach of [8] and [9] can not help for the case with the phys-
ical boundary condition (1.4). The reason is that uniform boundedness of
kvz kH 1 is difficult for this case; while the nonlinear term wvz in (1.1) is
not controllable with just uniform boundedness of k(vt , t )kL2 alone. An-
other drawback of the method of [8] and [9] is that it seems not helpful
for uniform estimate of higher regularity, even for the case with Neumann
boundary conditions.
In this paper, a completely different approach is presented. The new ap-
proach can prove uniform boundedness of H m norm of (v, ) for any m > 2,
if D and Q are sufficiently smooth. Moreover, this approach works not only
for the case with boundary condition (1.4), but also for other typical bound-
ary conditions. The main new idea is the use of hydrostatic Leray projector
to be defined in Section 3. It allows treatment of pressure and boundary
conditions more convenient. To realize the goal, several fundamental results
for hydrostatic Leray projector and a new anisotrophic multi-linear estimate
 N. Ju 3D Viscous Primitive Equations 5

in Sobolev space are established. With the help of these results, the proofs
of the new approach for uniform boundedness appear direct and simplified.
The results on hydrostatic Leray projector are expected to be of general in-
terest and can be helpful for solving some other problems about the system
of 3D Primitive Equations which may be rather difficult due to non-periodic
boundary conditions.
The rest of this article is organized as follows:
In Section 2, we give the notations, briefly review the background results
and recall some important facts which are useful for later analysis. We will
also state and prove Lemma 2.2, which gives a useful new anisotrophic multi-
linear estimate in Sobolev space and will be used a few times in Sections 4
and 5.
In Section 3, we present and prove Theorems 3.1, 3.2, 3.3 and 3.4. These
fundamental results on hydrostatic Helmholtz decomposition and hydro-
static Leray projector will be used in several key steps of the analyses of
Section 4 and Section 5.
In Section 4, the result of global uniform boundedness of H 2 solutions
is stated as Theorem 4.1 and is proved. Theorem 4.2 on global uniform
boundedness of k(vt , t )k will also be proved.
In Section 5, global uniform boundedness for higher regularity will be
presented as Theorem 5.1 and be proved briefly.

2 Preliminaries
Notations to be used are basically standard. C and c denote generic positive
constants, the values of which may vary from one occurrence to another. The
relation operators 4 and are used such that, for real numbers A and B,

A4B if and only if A 6 C B,

AB if and only if c A 6 B 6 C A,

for some positive constants C and c independent of A and B.

Lp () and Lp (D) (1 6 p 6 ) are the classic Lebesgue Lp spaces with
 N. Ju 3D Viscous Primitive Equations 6

the norm denoted as

k kp := k kLp , k k := k k2 .

H m () and H m (D) (m > 1) (with norm k kH m ) denote the classic Sobolev

spaces for L2 functions with weak derivatives up to order m also in L2 .
Notations for vector and scalar function spaces may not be distinguished
when they are self-evident from the context.
Define as the vertical average of on :

1 0
Z
(x, y) = (x, y, z)dz.
h h

By the Holder inequality, it is easy to see that, for Lp (),

1
kkLp () = h p kkLp (D) 6 kkp , p [1, +]. (2.1)

Define the function spaces:

H := H1 H2 := {v (L2 ())2 | v = 0, v n|D = 0} L2 (),

V := V1 V2 := {v (H 1 ())2 | v = 0, v|b l = 0} H 1 ().

Notice that, when 2 = 0, V2 is chosen as

 Z 
1
V2 := H () : d = 0 .

We use d and dD to denote dxdydz and dxdy in integrals in and D

respectively, or we may simply omit them.
Define the bilinear forms: ai : Vi Vi R, i = 1, 2, such that
2
Z ! Z
X
a1 (v, u) = 1 vi ui + 1 vz uz d + 1 1 v u dD,
i=1 t
Z Z
a2 (, ) = (2 + 2 z z ) d + 2 2 dD,
t

and the linear operators Ai : Vi 7 Vi , i = 1, 2, such that

hA1 v, ui = a1 (v, u), v, u V1 ; hA2 , i = a2 (, ), , V2 ,

 N. Ju 3D Viscous Primitive Equations 7

where Vi (i = 1, 2) is the dual space of Vi and h, i is the corresponding

scalar product between Vi and Vi . We also use h, i to denote the inner
products in H1 and H2 .
Define:
D(Ai ) = { Vi , Ai Hi }, i = 1, 2.

Since A1
i is a self-adjoint compact operator in Hi , by the classic spectral
theory, the power Asi can be defined for any s R. Then

D(Ai ) = D(A1
i )
1
is the dual space of D(Ai ) and, with the denotation i = Ai2 ,
1
1
Vi = D(Ai2 ) = D(i ), Vi = D(Ai 2 ) = D(1
i ).

Moreover,
D(Ai ) Vi Hi Vi D(Ai ) ,

where the embeddings above are all compact and each space above is dense
in the one following it.
Define the norm k kVi by:
1 1
k k2Vi = ai (, ) = hAi , i = hAi2 , Ai2 i = hi , i i , i = 1, 2.

Then, for any = (1 , 2 ) V1 and V2

kk 4 kkV1 , kk 4 kkV2 .

Therefore, for any = (1 , 2 ) V1 and V2 ,

kkV1 kkH 1 , kkV2 kkH 1 . (2.2)

Recall the following definitions of weak and strong solutions:

Definition 2.1 Suppose Q L2 (), (v0 , 0 ) H and T > 0.

The pair (v, ) is called a weak solution of the 3D viscous PEs (1.1)-(1.7)
on the [0, T ] if it satisfies (1.1)-(1.7) in weak sense (see e.g. [14] for details)
and that
(v, ) L ([0, T ]; H) L2 (0, T ; V ).
 N. Ju 3D Viscous Primitive Equations 8

Moreover, if (v0 , 0 ) V , a weak solution (v, ) is called a strong solution

of (1.1)-(1.7) on the time interval [0, T ] if, in addition, it satisfies

(v, ) L ([0, T ]; V ) L2 (0, T ; D(A1 ) D(A2 )).

Recall the following lemma which will be used in the a priori estimates
in Sections 4 and 5. See [1] and [6] for the proof.
Lemma 2.1 Suppose that v, H 1 () and L2 (). Then,
Z 0   1 1
 4 kvk 12 kvk 2 1 kk 12 kk 2 1 kk.
 

 v(x, y, )d ,  H H
z
Finally, we prove the following anisotrophic estimate in Sobolev spaces. It
will be used as well in Sections 4 and 5.
Lemma 2.2 Suppose , H 1 () and L2 (). Then,
Z
1 1 1
|| 4kk 4 (kk + kz k) 4 (kk + kk) 2

1 1 1
kk 4 (kk + kz k) 4 (kk + kk) 2 kk.
Proof:
Z Z 0
|| 6 kkL4 (D) kkL4 (D) kkL2 (D) dz
h
Z 0 1
1
4 kkL2 2 (D) kkL2 (D) + kkL2 (D) 2

h
1
1
kkL2 2 (D) kkL2 (D) + kkL2 (D) 2
kkL2 (D) dz
1 1
6kkL (L2 ) kkL (L2 )
2 2
z D z D
Z 0

1
kkL2 (D) + kkL2 (D) 2
h

1
kkL2 (D) + kkL2 (D) 2
kkL2 (D) dz
1 1 1 1
4kkL2 (L ) kkL2 (L ) (kk + kk) 2 (kk + kk) 2 kk
2 2
D z D z

Notice that
Z
kk2L2 (L ) = kk2L
z
dD
D z
ZD
4 kkL2z (kkL2z + kz kL2z ) dD 4 kk2 (kk2 + kz k2 ).
D
 N. Ju 3D Viscous Primitive Equations 9

This finishes the proof of Lemma 2.2.

3 Hydrostatic Decomposition
First, we prove the following Helmholtz type hydrostatic decomposition.
Recall that the space H1 was defined in Section 2.

Theorem 3.1
(L2 ())2 = H1 G,
where
G := {u (L2 ())2 | uz = 0, u = q, q H 1 (D)}.

Proof: Let (H1 ) be the orthogonal completement of H1 with respect to the

inner product in (L2 ())2 , i.e.

(L2 ())2 = H1 (H1 ) .

We need only to show that G = (H1 ) .

Claim 1: G (H1 ) .
Proof of Claim 1: Let u G. Thus, u = q for some q H 1 (D) and uz = 0.
Let v V1 , where

V1 := {v (C ())2 | v vanishes in a neighborhood of b l , v = 0}.

Then, Z Z
hu, vi = h (q) vdD = h q v = 0.
D D
Recall that V1 is dense in V1 (see [13]). Hence, V1 is dense in H1 , since V1
is dense in H1 . Therefore, u (H1 ) . This proves Claim 1.
Claim 2: (H1 ) G.
Proof of Claim 2: Let u = (u1 , u2 ) (H1 ) . Then, u (L2 ())2 and

hu, vi = 0, v H1 .

Step 1. Choose the special v = (z , 0) with C0 (). Then, v H1 , since

v = (0, 0). Thus,

hu1 , z i = hu, vi = 0, C0 ().

 N. Ju 3D Viscous Primitive Equations 10

Therefore, z u1 = 0 as the weak derivative of u1 . Similarly, z u2 = 0. Thus,

uz = 0, and u (L2 (D))2 .

Step 2. Choose v (L2 ())2 such that vz = 0 and

v = 0, n v = 0|D = 0.

Since v H1 , we have hu, vi = 0 for all such vs. Thus, there exists a
q H 1 (D), such that
u = q.

Therefore, u G. This proves Claim 2.

An immediate application of Theorem 3.1 is the definition of the hydro-
static Leray projector P as the orthogonal projection of (L2 ())2 onto H1
with repect to the inner product of (L2 ())2 .
Next, we give the following refined decomposition result for (L2 ())2 :

Theorem 3.2 Let D be an open bounded set of class C 2 . Then,

(L2 ())2 = H1 G1 G2 , (3.1)

where H1 , G1 and G2 are mutually orthogonal spaces and

G1 :={u (L2 ())2 | uz = 0, u = q, q H 1 (D), q = 0},

G2 :={u (L2 ())2 | uz = 0, u = q, q H01 (D)}.

Moreover, the following decomposition is valid:

u = Pu + (q1 + q2 ), u (L2 ()), (3.2)

where

q1 = 0, n q1 |D = n (u q2 ), (3.3)
q2 = u H 1 (D), q2 H01 (D). (3.4)

Proof: In the following, we prove (3.2), from which (3.1) follows as well. For
any given u (L2 ())2 , the Dirichlet problem (3.4) has a unique solution
 N. Ju 3D Viscous Primitive Equations 11

q2 . Therefore, the Neumann problem (3.3) is well defined. Moreoever, by

the Stokes formula and (3.4), we have
Z Z Z
n (u q2 ) ds = (u q2 ) dD = (u q2 ) dD = 0.
D D D

Then, the Neumann problem (3.3) has a solution q1 H 1 (D), which is

unique up to an additive constant. Obviously,

qi Qi , i = 1, 2.

It is easy to see that q1 and q2 are orthogonal in (L2 ())2 , since

Z Z
q1 q2 d =h q1 q2 dD
Z D Z
=h (n q1 )q2 ds h (q1 )q2 dD = 0,
D D

where the Stokes formula and the definitions of q1 and q2 given by (3.3) and
(3.4) are used. Finally, we show that

u := u (q1 + q2 ) H1 .

Notice that, by (3.3) and (3.4),

u = u q1 q2 = u q2 = 0,

and
u n|D = (n u + n q1 + n q2 ) |D = 0.

Thus, u H1 . This proves (3.2) and hence (3.1) as well.

The importance of (3.2) is that it gives more concrete information about
the decomposition (3.1). Especially, it provides detailed relation between Pu
and u, via (3.3) and (3.4). An immediate and very important consequence
of (3.2) is the following result.

Theorem 3.3 For any u (L2 ())2 , the following statements are valid:

(a) Suppose |uz | L1loc (). Then, in the sense of distribution,

(Pu)z = uz .
 N. Ju 3D Viscous Primitive Equations 12

(b) Suppose |u|, |u| Lr (), r (1, ) and D C 2 . Then, there exists
a constant c = c(r, D) > 0 such that:

k|(Pu)|kLr () 6 c(r, D)(kuk(Lr ())2 + k|u|kLr () ).

(c) Suppose u (W 1,r ())2 , r (1, ) and D C 2 . Then, there exists

a constant c = c(r, D) > 0 such that:

k|(, z )Pu|kLr () 6 c(r, D)kuk(W 1,r ())2 .

(d) Suppose u (W m,r ())2 , r (1, ) and D C m+1 . Then, there

exists a constant c = c(r, D) > 0 such that:

k|m
3 Pu|kLr () 6 c(r, D)kuk(W m,r ())2 .

Proof: Theorem 3.3 (a) follows from (3.2) immediately. By (3.2), we also
have
1
k|(Pu)|kLr () 6 kuk(Lr ())2 + h r k|2 q1 |kLr (D) + k|2 q2 |kLr (D) .

From (3.3), it follows that

k|2 q1 |kLr (D) 4k|u q2 |kW 11/r,r (D)
4kuk(W 1,r (D))2 + kq2 k(W 1,r (D))2
1
4h r (kuk(Lr ())2 + k|u|kLr () ) + kq2 k(W 1,r (D))2
From (3.4), it follows that
1
kq2 k(W 1,r (D))2 4 k ukLr (D) 4 h r k|u|kLr () .

In the above estimates, (2.1) has been used. Therefore,

k|(Pu)|k(Lr ())2 4 kukLr () + k|u|kLr () .

This proves Theorem 3.3 (b). Theorem 3.3 (c) is an immediate consequence
of (a) and (b). Proof of Theorem 3.3 (d) is similar to that of (c).
The last important result of the section is about the relation between
the anisotrophic Laplace operator L1 and the hydrostatic Stokes operator A1
through the hydrostatic Leray projector P. Recall that the operator A1 was
already defined in Section 2.
 N. Ju 3D Viscous Primitive Equations 13

Theorem 3.4 As an isomorphism from V1 onto (V1 ) , A1 satisfies:

A1 u = PL1 u, u V1 .

Proof: It suffices to prove A1 = PL1 only. Suppose u D(A1 ) and

V1 . Then, L1 u (L2 ())2 . By Theorem 3.1, there exists a q G such
that
Z
hPL1 u, i = hL1 u q, i = hL1 u, i = [1 u + 1 uzz ] d.

Since u D(A1 ) and V1 , we have

2
!
Z 0 Z  Z 0 Z X
hu, i = u dD dz = ui i dD dz
h D h D i=1
2 2
!
Z 0 Z X Z X
= n ui i ds ui i dD dz
h D i=1 D i=1
Z 2
X
= ui i d,
i=1
Z Z 0  Z  0 Z 0 
huzz , i = uzz dz dD = uz  uz z dz dD

D D z=h
Zh Z h

= 1 u dD uz z d.
t

Thus,
hPL1 u, i = hA1 u, i , V1 ,

that is
PL1 u = A1 u, u D(A1 ).

Notice that D(A1 ) is dense is V1 . Therefore, considered as an element of

(V1 ) ,
PL1 u = A1 u, u V1 .

 N. Ju 3D Viscous Primitive Equations 14

4 Uniform Boundedness of k(A1v, A2)k

In this section, we prove uniform boundedness of k(A1 v, A2 )k and existence
of a bounded absorbing set of it in R+ . More precisely, we prove the following
theorem.

Theorem 4.1 Suppose Q L2 () and (v0 , 0 ) D(A1 )D(A2 ). Let (v, )

be the unique strong sloution of problem (1.1)-(1.7). Then,

(A1 v, A2 ) L (0, ; H), (A1 v, t ) L2 (0, T ; V ), T (0, ).

Moreover, there exists a bounded absorbing set of k(A1 v, A2 )k in R+ .

Proof: Step 1. Estimate of kA1 vk.

First, apply the hydrostatic Leray projector P to (1.1) and use Theorem 3.4
to get  Z 0 
vt + A1 v = P (v )v + wvz + d + f v . (4.1)
z

Next, apply A1 to (4.1) and then take inner product with A1 v to arrive at

1d 3
kA1 vk2 + kA12 vk2
2 dt
  Z 0  
= A1 P (v )v + wvz + d + f v , A1 v
z
 Z 0 
1
3

A1 P (v )v + wvz + d + f v kA1 vk.

2 2
6
z

Notice that P does not commute with A1 due to no-slip boundary condition.
Therefore, Theorem 3.3 is crucial here in dealing with the right-hand side
of the above inequality, from which it then follows that
1 d 3
kA1 vk2 + kA12 vk2
2 dt  Z 0  3


P (v )v + wvz + d + f v 1 kA1 vk
2
4
z H
Z 0
3
d + f v

(v )v + wvz + kA1 vk,
2
4
z H1
 N. Ju 3D Viscous Primitive Equations 15

where we have also used norm equivalence (2.2). Noticing that v|b l =
w|t b = 0, we have
0
Z


(v )v + wvz + d + f v

z 1
Z 0 H

4k3 [(v )v + wvz ]k +
d
+ kvkV1
z H1
4k3 [(v )v + wvz ]k + kvkV1 + kA2 k.

Therefore,
d 3
kA1 vk2 + kA12 vk2 4 k3 [(v )v + wvz ]k2 + kvk2V1 + kA2 k2 . (4.2)
dt
Now, we estimate the first term on the right side of (4.2). By Lemma 2.2,
we have

k[(v )v]k 6k|v||2 v|k + k|v|2 k

1 3 1 3 1 3
4kvk 4 kvkV41 k2 vk 4 k2 vkV41 + kvk 2 kvkV21
1 3 1 3 3 1 3
4kvk 4 kvkV41 kA1 vk 4 kA12 vk 4 + kvkV21 kA1 vk 2 ,
kz [(v )v]k 4k(v )vz k + k(vz )vk
1 3 1 3 1 3 1 3
4kvk 4 kvkV41 kvz k 4 kvz kV41 + kvz k 4 kvz kV41 kvk 4 kvkV41
1 3 1 3 3 1 3
4kvk 4 kvkV41 kA1 vk 4 kA12 vk 4 + kvkV21 kA1 vk 2 .

Therefore,
1 3 1 3 3
k3 [(v )v]k2 4kvk 2 kvkV21 kA1 vk 2 kA12 vk 2 + kvkV1 kA1 vk3
3
(4.3)
4C kvk2 kvk6V1 kA1 vk2 + kvkV1 kA1 vk3 + kA12 vk2 .

Next, we use (1.3) and Lemma 2.1 and to get

k(wvz )k 6kw(vz )k + k(w) vz k

1 1 1 1
4kvk 2 kvkV21 kvz k 2 kvz kV21
1 1 1 1
+ k2 vk 2 k2 vkV21 kvz k 2 kvz kV21
1 3 1
4kvkV21 kA1 vkkA12 vk 2 .
 N. Ju 3D Viscous Primitive Equations 16

Similar to the above estimates, we use Lemma 2.1 and Lemma 2.2 to get
kz (wvz )k 6kwvzz k + k( v)vz k
1 1 1 1
4kvk 2 kvkV21 kvzz k 2 kvzz kV21
1 3 1 3
+ kvk 4 kvkV41 kvz k 4 kvz kV41
1 3 1 1 3
4kvkV21 kA1 vkkA12 k 2 + kvkV21 kA1 vk 2 .
Therefore,
3
k3 (wvz )k2 4kvkV1 kA1 vk2 kA12 vk + kvkV1 kA1 vk3
3
(4.4)
4C kvk2V1 kA1 vk4 + kvkV1 kA1 vk3 + kA12 vk2 .
Combining (4.2), (4.3) and (4.4) with > 0 sufficiently small, we have
d 3
kA1 vk2 + kA12 vk2
dt
4kvk2 kvk6V1 kA1 vk2 + kvkV1 kA1 vk3
+ kvk4V1 kA1 vk4 + kvk2V1 + kA2 k2 (4.5)

4(kvk2 kvk6V1 + kvkV1 kA1 vk + kvk4V1 kA1 vk2 )kA1 vk2

+ kvk2V1 + kA2 k2 .
Now, an application of Gronwall lemma to (4.5) immediately yields the
following boundedness:

A1 v L (0, T ; H1 ) L2 (0, T ; V1 ), T (0, ).

Let
y(t) := kA1 v(t)k2 , h(t) := kvk2V1 + kA2 k2 .
g1 (t) := kvk2 kvk6V1 + kvkV1 kA1 vk + kvk4V1 kA1 vk2 .
The following integrals with t > 0
Z t+1 Z t+1 Z t+1
y(s)ds, h(s)ds, g1 (s)ds
t t t
are all uniformly bounded for strong solutions. Then, an application of the
uniform Gronwall lemma (see [3] and [16]) to (4.5) yields the folloing uniform
boundedness:

A1 v L (0, ; H1 ) L2 (0, T ; V1 ), T (0, ).

 N. Ju 3D Viscous Primitive Equations 17

Moreover, there is a bounded absorbing set for A1 v in H1 . This finishes

proof of uniform boundedness of kA1 vk.
Step 2. Estimate of kA2 k.
We will estimate kA2 k in a way different from that of Step 1. This is
because the method of Step 1 will not work for Q L2 ().
Take inner product of (1.2) with A2 t , we get
 1 
1 1 d 1
kA22 t k2 + kA2 k2 = hQ, A2 t i + A22 [v + wz ], A22 t
2 dt
 1  (4.6)
d 1
= hQ, A2 i + A2 [v + wz ], A2 t .
2 2
dt
Therefore,
1 d d 1
kA22 t k2 + kA2 k2 6 2 hQ, A2 i + kA22 [v + wz ]k2 .
dt dt
Noticing that v|b l = w|t b = 0, we have
1 d
kA22 t k2 + kA2 k2 2 hQ, A2 i 4 k3 [v + wz ]k2 .

(4.7)
dt
Now, we estimate the right-hand side of (4.7). We have

k3 (v )k 4k(v )3 k + k|3 v|||k 4 kA1 vkkA2 k + k3 vk3 kk6 ,

k(wz )k 4kwz k + kz wk 4 kwk kA2 k + k2 vk3 kz k6 ,
kz (wz )k 4kwzz k + kz |v|k 4 kwk kA2 k + kvk3 kz k6 .

By interpolation inequality and Sobolev embedding, we have

1 1 1 3 1
k3 vk3 4 kvkV21 kA1 vk 2 , k2 vk3 4 kA1 vk 2 kA12 vk 2 , kk6 , kz k6 4 kA2 k.

Using (1.3), we have

z 0
Z  Z
 
|w(x, y, z)| =  v(x, y, ) d  6 |v(x, y, )| d.
h h

By Agmons inequality,
Z 0 1 1
|w(x, y, z)| 6 kv(, , )k(L
2
2 (D))2 kv(, , )k(H 2 (D))2 d
2

h
1 1 3 1
4h kvkV21 kA12 vk 2 .
2
 N. Ju 3D Viscous Primitive Equations 18

Thus,
1 1 3 1
kwk 6 h 2 kvkV21 kA12 vk 2 .

Collecting all the above estimates after (4.7), we obtain

k3 (v + wz )k2 4 g2 (t)kA2 k2 ,

where
3 3
g2 (t) := kvkV1 kA1 vk + kA1 vk2 + hkvkV1 kA12 vk + kA1 vkkA12 vk.

Then, (4.7) implies

1 d
kA22 t k2 + kA2 Qk2 4 g2 (t)(kA2 Qk2 + kQk2 ). (4.8)
dt
By (4.5) and uniform boundedness of kA1 vk, we have

g2 L2 (0, T ), T (0, ).

Thus, we conclude immediately from (4.8) by Gronwall lemma that

A2 L (0, T ; H2 ), t L2 (0, T ; V2 ), T (0, ).

R t+1
Moreover, by (4.5), we see that t g2 (s) ds is uniformly bounded for all
t > 0 and it is also obvious that
Z t+1 Z t+1
2 2
kA2 (s) Qk ds 6 kQk + kA2 (s)k2 ds
t t

is uniformly bounded for all t > 0, since (v, ) is a strong solution. Thus,
an application of uniform Gronwall lemma to (4.8) yields an absorbing set
for kA2 k in R+ and the uniform boundedness:

A2 L (0, ; H2 ).

Remark: If Q H 1 (), then, it is easy to see that one can use the
method in Step 1 of the proof of Theorem 4.1 to estimate kA2 k and obtain
the following additional regularity for :

A2 L2 (0, T ; V2 ), T (0, ).
 N. Ju 3D Viscous Primitive Equations 19

With Theorem 4.1, we obtain the following Theorem 4.2. Notice that
only (u0 , 0 ) D(A) is assumed here. Especially, (ut (0), t (0)) H is not
assumed, nor is it assumed that the system of 3D primitive equations is
valid at t = 0, as the later implies (ut (0), t (0)) H when (u0 , 0 ) D(A).

Theorem 4.2 Suppose Q L2 () and (v0 , 0 ) D(A1 )D(A2 ). Let (v, )

be the unique strong solution of the problem (1.1)-(1.7). Then,

(vt , t ) L (0, ; H) L2 (0, T ; V ), T (0, ).

Moreover, there exists a bounded absorbing set of k(vt , t )k in R+ .

Proof:
By (4.1) and Lemma 2.1, we have

kvt k 4kA1 vk + k(v )vk + kwvz k + kk + kvk

1 1 1 1
4kA1 vk + kvk6 kvk3 + kvk 2 kvkV21 kvz k 2 kvz kV21 + kk + kvk
3 1
4(1 + kvkV1 )kA1 vk + kvkV21 kA1 vk 2 + kk + kvk.
(4.9)

By (1.2), we have

kt k 6kA2 k + k(v )k + kwz k

(4.10)
4kA2 k + kA1 vkkk + kA1 vkkA2 k.

Thus, uniform boundedness of k(vt , t )k and existence of an absorbing ball

of it in R+ follow from (4.9), (4.10) and Theorem 4.1.
Denote
u := vt = t v, := t = t .

By (1.1) and (1.2), we have

ut + L1 u + (u )v + (v )u + wt vz + wuz
Z 0 (4.11)
+ (ps )t + (x, y, , t)d + f u = 0,
z

t + L2 + u + v + wt z + wz = 0. (4.12)
 N. Ju 3D Viscous Primitive Equations 20

Taking the inner product of (4.9) with u and using the boundary conditions
(1.4) and (1.6), we obtain

1 d
kuk22 +kuk2V1
2 dt
= h(u )v, ui h(v )u, ui hwt vz , ui hwuz , ui
Z 0  D E (4.13)
h(ps )t , ui , u f u , u
z
Z 0 
= h(u )v, ui hwt vz , ui + , u
z

where we have used the fact that

D E
h(ps )t , ui = f u , u = h(v )u, ui + hwuz , ui = 0.

Taking the inner product of (4.10) with and using (1.6), we obtain

1 d
kk22 +kk2V2 + 2 k(z = 0)k22
2 dt
= hu , i hv , i hwt z , i hwz , i (4.14)

= hu , i hwt z , i

where we have used the fact that

hv , i + hwz , i = 0.

Following the a priori estimates of [7] using (4.13) and (4.14), we obtain

d
kuk2 + kk2 + kuk2V1 + kk2V2 4 g(t) kuk2 + kk2 ,



(4.15)
dt
where

g(t) := 1 + kvk42 + kk42 + kvz k22 kvz k22 + kz k22 kz k22 .

Thus, from (4.15), we obtain

(vt , t ) L2 (0, T ; V ), T > 0.

This finishes proof of Theorem 4.2.

 N. Ju 3D Viscous Primitive Equations 21

5 Uniform Boundedness for Higher Regularity

Theorem 5.1 Suppose the integer m > 3, Q H m1 () and (v0 , 0 )
V (H m ())3 . Let (v, ) be the unique strong sloution of problem (1.1)-
(1.7). Then,

(v, ) L (0, ; (H m ())3 ), (v, ) L2 (0, T ; (H m+1 ()3 ), T (0, ).

Moreover, there exists a bounded absorbing set of k(v, )k(H m ())3 in R+ .

Proof: The idea of the proof is to use induction and the method of the
proof of Theorem 4.1. We give an outline of the proof here. To simplify
the presentation without loss of generality, let us consider the following
simplified equation:

vt + A1 v = P [(v )v + wvz ] . (5.1)

Applying Am
1 to (5.1) and taking inner product with v, we get

1d m 2
k1 vk + km+1 vk2 = m1 P [(v )v + wvz ] , m+1


1 1 1 v
2 dt
1 1
6 km1 P [(v )v + wvz ] k2 + km+1 vk2 ,
2 1 2 1
that is
d m 2
k vk + km+1 vk2 6 km1 P [(v )v + wvz ] k2 (5.2)
dt 1 1 1

Now, we can use induction to prove the expected uniform boundedness of

kvk(H m ())2 for m > 3. We already have the result valid for m = 0, 1, 2.
Assume it is already valid for kvk(H k ())2 , with 0 6 k 6 m 1. Next, we
show uniform boundedness of kvk(H m ())2 . Using Theorem 3.3 as in the
proof of Theorem 4.1, we have

km1
1 P [(v )v + wvz ] k2 4 k [(v )v + wvz ] k2(H m1 ())2 .

To estimate the right-hand side of the above inequality, we just need to

estimate L2 norm of
3m1 [(v )v + wvz ] , (5.3)
 N. Ju 3D Viscous Primitive Equations 22

since if these terms which contain the highest order derivatives can be well
controlled, then those terms involving lower order derivatives can be treated
exactly in the same way. To simplify our presentation, we just need to
estimate
kzm1 [(v )v]k, and km1 (wvz )k,

since all other terms appearing in (5.3) can be treated similarly and behave
no worse. By Leibniz rule, we have
m1
X
kzm1 [(v )v] k 6 kzk v zm1k vk.
k=0

By Lemma 2.2,
1 3 1 3
kzk v zm1k vk 4 kvkH
4
k kvkH k+1 kvkH mk kvkH mk+1 .
4 4 4

Hence, by induction hypothesis for uniform boundedness of kvk(H m1 )2 , we

have
m1
X 1 3 1 3
kzm1 [(v )v] k 4 kvkH
4
k kvkH k+1 kvkH mk kvkH mk+1
4 4 4

k=0
 1
 m1
X 3 1 3
4 max kvkH k 4
kvkH
4
k+1 kvkH mk kvkH mk+1
4 4
06k6m1
k=0
m1
X 3 1 3
4 kvkH k+1 kvkH mk kvkH
4 4 4
mk+1
k=0
1 3 3
m kvkH m+1 + kvkH m .
4 4 4
4kvkH

Therefore,
1 3 3
kzm1 [(v )v] k2 4kvkH
2
m kvk m+1 + kvkH m
2
H
2
(5.4)
41 + C kvk2H m + kvk2H m+1 .

Similarly, we have
m1
X
km1 (wvz )k 4
k
( w) (m1k vz ) ,

k=0
 N. Ju 3D Viscous Primitive Equations 23

where we use to include all the products occurring here. By (1.3) and
Lemma 2.1, we get
m1
X
km1 (wvz )k2 4 kvkH k+1 kvkH k+2 kvkH mk kvkH m+1k
k=0
(5.5)
4kvkH m kvkH m+1 6 C kvk2H m + kvk2H m+1 .

Therefore, from (5.2) with consideration of(5.4) and (5.5), we have for m >
3,
d m 2
k vk + km+1 vk2 4 1 + kvk2H m . (5.6)
dt 1 1

Applying Glonwall lemma first and then uniform Gronwall lemma to (5.6)
proves uniform boundedness of kvkH m for (5.1) and existence of a bounded
absorbing set for kvkH m in R+ . The proof of the theorem for problem(1.1)-
(1.7) is similar and is thus omitted.

References
[1] C. Cao and E.S. Titi, Global Well-posedness and Finite Dimensional Global
Attractor for a 3-D Planetary Geostrophic Viscous Model, Comm. Pure
Appl. Math. 56, (2003), 198-233.
[2] C. Cao and E.S. Titi, Global well-posedness of the three-dimensional primitive
equations of large scale ocean and atmosphere dynamics, Ann. of Math.
(2) 166 (2007), no. 1, 245-267.
[3] C. Foias and G. Prodi, Sur le comportement global des solutions non station-
naires des equations de Navier-Stokes en dimension 2, Rend. Sem. Mat.
Univ. Padova, 39 (1967), 1-34.
[4] F. Guillen-Gonzaez, N. Masmoudi and M.A. Rodrguez-Bellido, Anisotropic
estimates and strong solutions of the Primitive Equations, Diff. Integral
Eq. 14(2001), no. 1, 1381-1408.
[5] M. Hieber and T. Kashiwabara, Global well-posedness of the three-dimensional
primitive equations in Lp-space. Arch. Ration. Mech. Anal. 221, 1077-1115
(2016).
[6] N. Ju, The global attractor for the solutions to the 3D viscous primitive equa-
tions, Discrete and Continuous Dynamical Systems, 17 (2007), no. 1, 159-
179.
 N. Ju 3D Viscous Primitive Equations 24

[7] N. Ju, The finite dimensional global attractor for the 3D viscous primitive
equations, Discrete and Continuous Dynamical Systems, 36 (2016), no.
12, 7001-7020.
[8] N. Ju, On H 2 solutions and z-weak solutions of the 3D primitive equations.
Indiana University Mathematics Journal, 66, (2017), no. 3.
[9] N. Ju and R. Temam, Finite Dimensions of the Global Attractor for 3D Prim-
itive Equations with Viscosity. J. Nonlinear Sci. 25 (2015), no. 1, 131-155.
[10] G. Kobelkov, Existence of a solution in the large for ocean dynamics equa-
tions, J. Math. Fluid Mech. 9 (2007), no. 4 588-610.
[11] I. Kukavica and M. Ziane, On the regularity of the primitive equations of the
ocean, Nonlinearity, 20, (2007), no. 12, 2739-2753.
[12] I. Kukavica and M. Ziane, Uniform gradient bounds for the primitive equations
of the ocean. Differential Integral Equations 21, (2008), no. 9-10, 837-849.
[13] J. Lions, R. Temam and S. Wang, New formulations of the primitive equations
of atmosphere and applications, Nonlinearity 5, (1992), 237-288.
[14] J. Lions, R. Temam and S. Wang, On the equations of the large scale Ocean,
Nonlinearity 5, (1992), 1007-1053.
[15] M. Petcu, On the three dimensional primitive equations, Adv. Dif. Eq., 11
(2006), 1201-1226.
[16] R. Temam, Infinite Dimensional Dynamical Systems in Mechanics and
Physics Spring-Verlag, New York, Applied Mathematical Sciences Series,
vol. 68, 1988. Second augmented edition, 1997.
[17] R. Temam and M. Ziane, Some mathematical problems in geophysical fluid dy-
namics, Handbook of Mathematical Fluid Dynamics, vol 3, S. Friedlander
and D. Serre Editors, Elsevier, 2004, 535-658.