Professional Documents
Culture Documents
Navier-Stokes Equations∗†
Dongyi Wei
School of Math Sciences and BICMR, Peking University,
arXiv:1508.03318v1 [math.AP] 13 Aug 2015
Abstract
Smooth solutions to the axially symmetric Navier-Stokes equations obey the fol-
lowing maximum principle:kruθ (r, z, t)kL∞ ≤ kruθ (r, z, 0)kL∞ . We first prove the
global regularity of solutions if kruθ (r, z, 0)kL∞ or kruθ (r, z, t)kL∞ (r≤r0 ) is small
compared with certain dimensionless quantity of the initial data. This result im-
proves the one in Zhen Lei and Qi S. Zhang [10]. As a corollary, we also prove the
global regularity under the assumption that |ruθ (r, z, t)| ≤ | ln r|−3/2 , ∀ 0 < r ≤
δ0 ∈ (0, 1/2).
Key words axially symmetric, Navier-Stokes Equations, Regularity
Criterion.
MSC2010 35Q30, 76D05, 76D07, 76N10.
1 Introduction
In the cylindrical coordinate system with (x1 , x2 , x3 ) = (r cos θ, r sin θ, z), an axially
symmetric solution of the Navier-Stokes equations is a solution of the following form
u(x, t) = ur (r, z, t)er + uθ (r, z, t)eθ + uz (r, z, t)ez , p(x, t) = p(r, z, t),
where x
1x2 x x
2 1
er = ,
, 0 , eθ = − , , 0 , ez = (0, 0, 1).
r r r r
In terms of (ur , uθ , uz , p), the axially symmetric Navier-Stokes equations are as follows
u2
∂t ur + u · ∇ur − △ur + ur2r − rθ + ∂r p = 0,
∂t uθ + u · ∇uθ − △uθ + u2θ + ur uθ = 0,
r r
(1.1)
∂ u
t z + u · ∇u z − △u z + ∂ z p = 0,
∂r (rur ) + ∂z (ruz ) = 0.
∗
E-mail address: jnwdyi@163.com (D.Y. Wei).
†
1
It is well-known that finite energy smooth solutions of the Navier-Stokes equations satisfy
the following energy identity
Z t
2
ku(t)kL2 + 2 k∇u(s)k2L2 ds = ku0 k2L2 < +∞. (1.2)
0
ω = ∇ × u = ωr er + ωθ eθ + ωz ez ,
where
1
ωr = −∂z (uθ ), ωθ = ∂z (ur ) − ∂r (uz ), ωz = ∂r (ruθ ).
r
Denote
ωθ ωr ∂z uθ
Ω= , J= =− ,
r r r
then
2 uθ
∂t Ω + u · ∇Ω − △ + ∂r Ω + 2 J = 0,
r r
(1.5)
2 ur
∂t J + u · ∇J − △ + ∂r J − (ωr ∂r + ωz ∂z ) = 0.
r r
M1 = (1 + kΓ0 kL∞ )ku0 kL2 and M0 = (kJ0 kL2 + kΩ0 kL2 )M13 .
2
1/4 −1/2
Remark 1.1. Choose r0 > 0 such that M0 ≥ r0 M1 and use (1.4), then we can
obtain the following global regularity condition
1/4
(b) kΓ0 kL∞ ≤ (1 + ln(C0 M0 + 1))−3/2 ,
this condition depends only on the initial value and is very useful especially when kΓ0 kL∞
is very small, in this sense it improves the result in [10]. On the other hand, if we take
r0 → 0+ in condition (a) we can obtain an important corollary.
Corollary 1.1. Let δ0 ∈ (0, 1/2), u be the strong solution of the axially symmetric
Navier-Stokes equations with initial value u0 ∈ H 2 and kΓ0 kL∞ < ∞. If
Denote q
2
− 43
K(ε) = exp 2ε − 1 − 1 , K0 (ε) = exp(ε− 3 − 1)
3
the presence of swirl, the global regularity problem is still open. Recently, tremendous
efforts and interesting progress have been made on the regularity problem of the axi-
ally symmetric Navier-Stokes equations [1][2][3][4][5][9][10]. There are many significant
results under the sufficient conditions for regularity of axially symmetric solution of type
ur 2 3 3
ωθ ∈ Lp (0, T ; Lq (R3 )) and ∈ Lp (0, T ; Lq (R3 )), + ≤ 2, < q < +∞
r p q 2
in [1]. It has been shown in [4] that the axially symmetric solution is smooth in R3 ×(0, T ]
when r d uθ ∈ Lp (0, T ; Lq (R3 )) with
2 3 2 3
d ∈ [0, 1), (p, q) ∈ ,∞ × ,∞ , + ≤ 1 − d ,
1−d 1−d p q
in particular, global regularity is obtained if |Γ| ≤ Cr α for some α > 0, C > 0. In [10]
global regularity is obtained if |Γ| ≤ C| ln r|−2 for some C > 0. Clearly, our Corollary
1.1 improves the one in [10].
Here global regularity means T ∗ = +∞, and we only need to prove Ω ∈ L∞ (0, T ∗; L2 (R3 )),
hence we can use the results in [4] or [14] to obtain global regularity. We can also use
ur ur
Lemma 2.1 in section 2 to obtain ∇ ∈ L∞ (0, T ∗ ; L2 (R3 )), and ∈ L∞ (0, T ∗; L6 (R3 )),
r r
then the results in [1] or [6] imply the global regularity.
This paper is organized as follows, in section 2 we will give some notations and
3 important Lemmas, in section 3 we will first follow the proof in [10] then use the
Lemmas in section 2 to conclude the proof.
4
Proof. By virtue of ∂r (rur ) + ∂z (ruz ) = 0, we can find the stream function ψθ such
that
1
ur = −∂z ψθ , uz = ∂r (rψθ ),
r
then we can compute
2 ψθ 2 ur
− △ + ∂r = Ω, △ + ∂r = ∂z Ω.
r r r r
therefore,
u
2 Z
ur 2 ur
r
∇
2 ≤ − △ + ∂r dx
r L r r r
ur ur
u
Z Z
r
=− ∂z Ωdx = ∂z Ωdx ≤
∇
kΩkL2 ,
r r r L2
hence,
u
r
∇
2 ≤ kΩkL2 .
r L
Since
2
2 ur
k∂z Ωk2L2
=
△ + ∂r
r r
L2
1 ur
2
u
2
ur 1 ur
Z
r
=
△
+ 4
∂r
+ 4 △ ∂r rdrdz,
r L2 r r L2 r r r
and
u
2
u
2
r
2 r
ur
△
2 =
∇
, ∂r = 0,
r L r L2 r r=0
1 ur
2
Z
ur 1 ur 1 ur 2 ur 2
Z
△ ∂r rdrdz =
∂r
+
∂z − ∂r (0, z, t)dz ≥ 0,
r r r r r L2 2 r r
u
2
r
we can obtain
∇2
≤ k∂z Ωk2L2 , this completes the proof Lemma 2.1.
r L2
Denote
Z r
v
′ ′
v(r, z, t) = |uθ (r , z, t)|dr , for r > 0, a(t) =
(t)
,
0 r L∞
5
Z r′
′ ′
Proof. For r > 0, z ∈ R, t > 0 as v(r , z , t) = ′ ′
|uθ (r, z ′ , t)|dr, by Hölder inequality
0
we have
r′ r′
|uθ (r, z ′ , t)|2
Z Z
′ ′ 2
|v(r , z , t)| ≤ rdr dr
0 0 r
′
r ′2 r +∞
|uθ (r, z, t)|2
Z Z
= dr −∂z dz
2 0 z r
Z ′ Z +∞
r ′2 r
= dr 2Juθ (r, z, t)dz
2 Z0
u z
θ
≤ r ′2 |J| (r, z, t)rdrdz
r
uθ
′2
≤ r kJ(t)kL
(t)
.
2
r L2
Hence, we get
′ ′ 2
2
v(r , z , t)
u
θ
a(t) = sup ≤ kJ(t)kL2
(t)
2.
r ′ r
L
r >0,z ∈R
′ ′
2
kΓk2L∞ + ε 3
Z Z Z
2
2 2 2
|uθ (t)| |f | dx ≤ ε 3 |∂r f | dx + C |f |2 dx, (2.2)
r12 r
r≥ 21
|uθ (t)| 2
Z Z
− 31
|f | rdrdz ≤ ε |∂r f |2 rdrdz. (2.3)
r
Z r1
′
In this case, f (r , z) = − ∂r f (r, z)dr for 0 < r ′ < r1 , by Hölder inequality we have
r′
r1 r1
dr
Z Z
′ 2 2
|f (r , z)| ≤ r|∂r f (r, z)| dr .
r′ r′ r
Consequently,
r1 r1
dr ′
Z Z Z Z
′ ′ 2 ′ 2 ′
|uθ (r , z, t)||f (r , z)| dr ≤ r|∂r f (r, z)| dr |uθ (r , z, t)| dr , (2.4)
0 r′ r
6
by the definition of v we have
Z r1 r1 r1
dr ′ dr
Z Z
′
|uθ (r , z, t)| dr = v(r, z, t) , (2.5)
0 r′ r 0 r
|Γ| ε
by the definition of a(t) we have v(r, z, t) ≤ ra(t). On the other hand, |uθ | = ≤
r r
ε
for 0 < r < r1 . Hence, if ≤ r ≤ r1 , then
a(t)
Z r
ε
v(r, z, t) = v , z, t + |uθ (r ′ , z, t)|dr ′
a(t) ε
Z r a(t)
ε ε ′ ra(t)
≤ a(t) + dr = ε 1 + ln .
a(t) ε
a(t)
r′ ε
εK(ε)
here we used 0 < r1 ≤ . By (2.4), (2.5), (2.6), we have
a(t)
Z Z
2 − 31
′ ′ ′
|uθ (r , z, t)||f (r , z)| dr ≤ ε r|∂r f |2 dr,
integrate in z, we obtain (2.3). Now we discuss general f . Take a smooth cut-off function
1
of r such that (i) φ′ ≤ 0, (ii) φ ≡ 1 if 0 ≤ r ≤ , (iii) φ ≡ 0 if r ≥ 1. Then we have
2
2 2 !
|uθ (t)| 2 r r
Z Z Z
|f |2 drdz
|f | rdrdz = |uθ (t)| φ f drdz + |uθ (t)| 1 − φ
r r1 r1
Z 2
r |Γ| 2
Z
− 31
≤ε ∂r φ f rdrdz + |f | drdz
r1 r≥ 21 r !
r
C 4kΓkL∞
Z Z Z
− 31 2 2
≤ε |∂r f | rdrdz + 2 |f | rdrdz + |f |2 rdrdz
r1 r≥ r21 r12 r1
r≥ 2
1 Z
−
kΓkL∞ + ε 3
Z
1
≤ ε− 3 |∂r f |2 rdrdz + C |f |2rdrdz,
r12 r≥ 2
r1
7
here we used (2.3) and the fact that
Z 2 Z " 2 2
∂r φ r r 2 2 ∂r φ r
f rdrdz = φ |∂r f | + |f |
r1 r1 r1
#
r r
+∂r |f |2φ ∂r φ rdrdz
r1 r1
" 2 2 #
r r r r
Z Z
2 2 2
= φ |∂r f | + |f | ∂r φ rdrdz − |f | ∂r φ ∂r φ r drdz
r1 r1 r1 r1
C
Z Z
≤ |∂r f |2 + 2 |f |2rdrdz.
r1 r≥ r21
Similarly we have
2 2 !
r r
Z Z Z
|uθ (t)|2 |f |2 rdrdz = |uθ (t)|2 φ f rdrdz + |uθ (t)|2 1 − φ |f |2 rdrdz
r1 r1
2
r |Γ|2 2
Z Z
≤ kΓkL∞ (r≤r1 ) |uθ (t)| φ f drdz + 2
|f | rdrdz
r1 r≥
r1 r
!2
C 4kΓk2L∞
Z Z Z
1 2 2
≤ εε −3
|∂r f | rdrdz + 2 |f | rdrdz + |f |2rdrdz
r1 r≥ r21 r12 r1
r≥ 2
2 Z
2
kΓkL∞ + ε
Z
2 3
≤ ε 3 |∂r f |2 rdrdz + C 2
|f |2 rdrdz.
r1 r
r≥ 21
8
Similarly, by applying the energy estimate to the equation of Ω, one obtains that
1d uθ
Z Z
2 2 2
kΩkL2 + k∇ΩkL2 + |Ω(0, z, t)| dz = −2 JΩdx. (3.2)
2 dt r
Notice that
ur uθ
Z Z
J(ωr ∂r + ωz ∂z ) dx = [∇ × (uθ eθ )] · J∇ dx
r r
uθ 1 1
ur
2
Z
= uθ eθ · ∇J × ∇ dx ≤ k∇Jk2L2 +
uθ ∇
,
r 2 2 r L2
d ur
Z
2
kJk2L2 + k∇Jk2L2 + 2 2
|J(0, z, t)| dz ≤
uθ ∇
. (3.3)
dt r L2
Now we estimate the right hand side of (3.2) and (3.3). under the assumption of condition
(b), let
n o −3/2
1/4 −1/2
ε = 1 + ln C0 max M0 , r0 M1 + 1 , (3.4)
uθ |uθ | 2 |uθ | 2
Z Z Z
1 1
−
−2 JΩdx ≤ ε 3 |Ω| dx + ε 3 |J| dx
r r 1
r
C(1 + ε 3 kΓkL∞ )
Z Z
2
≤ |∂r Ω| dx + |Ω|2 dx (3.5)
r(t)2 r(t)
r≥ 2
− 32 1
Cε (1 + ε 3 kΓk ∞ )
Z Z
2 2 L
+ε −3
|∂r J| dx + |J|2 dx.
r(t)2 r≥ 2r(t)
ur ur
Choosing f = ∂r , ∂z in (2.2), we have
r r
2 2
ur
ur 2 Cε 3 (1 + ε− 3 kΓk2L∞ ) u 2
Z Z
2
2 r
uθ ∇
2 ≤ ε 3 ∂r ∇ dx + ∇ dx. (3.6)
r L r r(t)2 r(t)
r≥ 2 r
Denote
1 2
M2 = 1 + ε 3 kΓkL∞ + ε− 3 kΓk2L∞ ,
9
Inserting (3.5), (3.6) into (3.2), (3.3), we have
2
!
d ε CM2 ur 2
Z
3 2
2 2 2 2
kJkL2 + kΩkL2 ≤ ε 3 ∇ + |Ω| + |J| dx. (3.7)
dt 2 r(t)2 r≥ r(t)
2
r
Denote 2
ε3
A(t) = kJ(t)k2L2 + kΩ(t)k2L2 ,
2
then
A(t) ∈ C[0, T ∗ ) ∩ C 1 (0, T ∗ ).
By Lemma2.1 and the fact that
ur ∇ur ur
∇ = − 2 er ,
r r r
u 2 u 2
2 2 2 r θ 2
|∇u| = |∇ur | + |∇uθ | + |∇uz | + + ,
r r
we obtain
ur 2
Z
u
2
r
∇ + |Ω| dx ≤
∇
2 + kΩk2L2 ≤ 2kΩk2L2 ,
2
r≥ r(t)
2
r r L
ur 2 C
Z Z
2 2
∇ + |Ω| + |J| dx ≤ 2
|∇u|2 dx,
r≥ 2
r(t) r r(t)
hence we have
d CM2 ur 2
Z
2
2 2
A(t) ≤ ε 3 ∇ + |Ω| + |J| dx
dt r(t)2 r≥ r(t)
2
r
CM2 C
Z
2
2 2 2
≤ min 2ε 3 kΩkL2 + kJkL2 , |∇u| dx (3.8)
r(t)2 r(t)2
k∇u(t)k2L2
CM2
≤ min A(t), .
r(t)2 r(t)2
εK(ε)
Fix t ∈ (0, T ∗), if r(t) = ≤ r0 , by Lemma 2.2, we have
a(t)
1
a(t)2 ≤ A(t) 2 k∇u(t)kL2 ,
1
1 a(t)2 A(t) 2 k∇u(t)kL2
2
= 2
≤ .
r(t) (εK(ε)) (εK(ε))2
And (3.8) implies
1
( 1
)
d CM2 A(t) 2 k∇u(t)kL2 A(t) 2 k∇u(t)k3L2
A(t) ≤ min A(t),
dt (εK(ε))2 (εK(ε))2
k∇u(t)k3L2
CM2 A(t)k∇u(t)kL2 1
= min A(t) 2 ,
(εK(ε))2 (εK(ε))2
4
CM2 A(t) 3 k∇u(t)k2L2
≤ 8 ,
(εK(ε)) 3
10
otherwise, we have r(t) = r0 and
d CM2 k∇u(t)k2L2
A(t) ≤ .
dt r04
d
F (A(t)) ≥ CM2 k∇u(t)k2L2 ,
dt
and we can use the energy identity to obtain
Z t
F (A(0)) − F (A(t)) ≤ CM2 k∇u(s)k2L2 ds ≤ C1 M2 ku0 k2L2 .
0
Therefore, if the condition (a)′ : F (A(0)) > C1 M2 ku0 k2L2 is satisfied, then
inf F (A(t)) > 0, sup A(t) < +∞, sup kΩ(t)k2L2 < +∞,
0<t<T ∗ 0<t<T ∗ 0<t<T ∗
(εK(ε))2
F (A(0)) ≥ F max A(0),
r03
− 31
(εK(ε))2
8
= 3 max A(0), 3
(εK(ε)) 3 ,
r0
2
! 21
1 ε 3
A(0) =2 kJ(0)k2L2 + kΩ(0)k2L2 ≤ kJ0 kL2 + kΩ0 kL2 ,
2
from the definition of M2 , M1 , M0 and (1.4) we have
2 2 1
M2 < ε− 3 (1 + kΓkL∞ )2 , M2 ku0k2L2 ≤ ε− 3 M12 , A(0) 2 M13 ≤ M0 .
11
hence we obtain
n 1
o
3/2
M2 ku0 k2L2
32 ε−1 M13 max A(0) 2 , εK(ε)/r0
≤
F (A(0)) 33/2 (εK(ε))4
n o
3/2
max M0 , εK(ε)M13 /r0
≤
33/2
(ε(εK(ε))
4
)
4 3 3
3 M C
0 ∗ C ∗ M 1
≤ 3− 2 max ,
K0 (ε)4 K0 (ε)3 r03/2
4 3 3
− 23 C∗ C∗ − 32 C∗
< 3 max 4
, 3 =3
C C C0
r 03 0
3 3 − 3
≤ 3− 2 = C1 2 ,
C1
Therefore the claim is true, this completes the proof of Theorem 1.1.
by (1.6), we have | ln r0 |−3/2 ≥ kΓkL∞ (r≤r0 ) , using the property of this r0 we have
n o −3/2
1/4 −1/2
1 + ln C0 max M0 , r0 M1 + 1
−3/2 −3/2
−1/2
= 1 + ln C0 r0 M1 + 1 > 1 + ln e−1 r0−1 = | ln r0 |−3/2 ,
Therefore condition (a) in Theorem 1.1 is satisfied, and we can use Theorem 1.1 to get
the global regularity, this completes the proof of Corollary 1.1.
Acknowledgments The author would like to thank the professors Gang Tian and
Zhifei Zhang for some valuable suggestions.
References
[1] D. Chae, J. Lee, On the regularity of the axisymmetric solutions of the Navier-Stokes
equations, Math. Z., 239 (2002), 645-671.
[2] C. C. Chen, R. M. Strain, T. P. Tsai, H. T. Yau, Lower bound on the blow-up rate
of the axisymmetric Navier-Stokes equations, Int. Math Res. Notices (2008), vol. 8,
artical ID rnn016, 31 pp.
12
[3] C. C. Chen, R. M. Strian, T. P. Tsai, H. T. Yau, Lower Bounds on the Blow-Up
Rate of the Axi-Symmetric Navier-Stokes Equations II, Electron. Comm. PDE. 3
(2009) 203-232.
[9] Z. Lei, E. A Navas, Q. S. Zhang, A priori bound on the velocity in Axially symmetric
Navier-Stokes Equations, arXiv:1309.6625v2.
[12] A. Majda, A. Bertozzi, Vorticity and incompressible flow. Cambridge Texts in Ap-
plied Mathematics, 27. Cambridge University Press, Cambridge, 2002.
[13] C. Miao, X. Zheng, On the global well-posedness for the Boussinesq system with
horizontal dissipation. Comm. Math. Phys. 321 (2013), no. 1, 33-67.
13