Professional Documents
Culture Documents
Abstract. We demonstrate that the three dimensional incompressible magnetohydrodynamics (MHD) system is ill-posed due to the discontinuity of weak solutions in a wide range of spaces. Specifically, we construct initial data which
has finite energy and is small in certain spaces, such that any Leray-Hopf type
of weak solution to the MHD system starting from this initial data is discontinuous at time t = 0 in such spaces. An analogous result is also obtained for the
Navier-Stokes equation which extends the previous result of ill-posedness in
1
B ,
by Cheskidov and Shvydkoy to spaces that are not necessarily critical.
The region of the spaces where the norm inflation occurs almost touches L2 .
KEY WORDS: magneto-hydrodynamics system; ill-posedness; discontinuity of solutions.
CLASSIFICATION CODE: 76D03, 35Q35.
1. Introduction
The three dimensional incompressible magneto-hydrodynamics (MHD) system
is given by:
ut 4u + u u b b + p = 0,
(1.1)
bt 4b + u b b u = 0,
u = 0,
b = 0,
if (u(x, t), b(x, t)) solves (1.1) with the initial data (u0 (x), b0 (x)). A space that is
invariant under the above scaling is called critical space. Examples of critical spaces
associated with the above scaling in three dimension are
3
1+ p
1
1
, BM O1 , B ,
.
H 2 , L3 , B p|p<,
1
Notice that B ,
is the largest critical space for both the NSE and the MHD
system. In the periodic case there is no distinction between homogeneous and
1+ 3
data in B, , such that mild solutions with this data become arbitrarily large in
1
after an arbitrarily short time. This result was later extended to generalized
B ,
1
Besov spaces smaller than B,p
, p > 2 by Yoneda [17]. Moreover, in [7] Cheskidov
and Shvydkoy proved the existence of discontinuous Leray-Hopf solutions of the
1
with arbitrarily small initial data. Contrary to
Navier-Stokes equations in B ,
the Bourgain-Pavlovic construction where the energy transfers from high to low
modes to produce the norm inflation, the norm discontinuity in [7] is due to the
forward energy cascade generated by local interactions. In [6] Cheskidov and Dai
considered fractional Navier-Stokes equations and showed that the natural space
for norm inflation is critical only when the power of the Laplacian is one. When
the power od the laplacian is larger than one, the norm inflation occurs not only in
critical spaces, but also in subcritical and supercritical.
For the MHD system, Miao, Yuan and Zhang [15] proved the existence of a
global mild solution in (BM O1 )2 for small initial data and uniqueness of such
2
solution in C([0, ); BM O1 ) . Later, Dai, Qing and Schonbek [8] established
several different types of norm inflation phenomena for the three dimensional
1
MHD system in the largest critical space (B ,
)2 , by adopting the idea of [3].
Since the MHD system describes the coupling of velocity field and magnetic field,
the authors were able to construct different initial data to produce different types
of norm inflation. In particular, the magnetic field can develop norm inflation in
short time even when the velocity remains small and vice versa. In [5] Cheskidov
and Dai used their approach from [6] to extend the norm inflation results to even
wider range of spaces that included critical, subcritical, and supercritical.
In this paper we further investigate the method of [7] to study the ill-posedness
problem of the NSE and the MHD system in a large class of spaces which may
contain critical, supercritical and subcritical spaces. First, modifying the initial
data construction, we are able to obtain discontinuous weak solutions to the NSE
in certain Besov spaces. Namely, we prove that
Figure 1. The region where the discontinuity for the NSE occurs
on the plane of the smoothness index s vs 1/r.
Theorem 1.1. Let 1 < r and 23 < 2 which satisfy either
3
or
2+ <3 ;
r
3
3
11
3
3 + < 4,
2 +
r ,
.
2
r
r
2
3
+3
r
There exists an initial data u0 Br,
(that depends only on ), such that every
Leray-Hopf weak solution u Cw ([0, T ); L2 ) L2 ([0, T ); H) to the NSE satisfies
(1.3)
3 +3
r
Br,
t0+
+3
+3
r
r
There exists an initial data (u0 , b0 ) Br,
Br,
Hopf weak solution (u(t), b(t)) to (1.1)-(1.2) satisfies
(1.4)
lim sup ku(t) u0 k r3 +3 + kb(t) b0 k
t0+
Br,
3 +3
r
Br,
T3
q () =
(1
for q 0,
q )
() for q = 1.
For a tempered distribution vector field v on the torus Tn we consider the LittlewoodPaley projections
X
(2.6)
vq (x) =
v(k)q (k)eikx , q 1.
kZn
X
q=1
vq
holds in the distribution sense. Essentially the sequence of the smooth functions
q forms a dyadic partition of the unit. To simplify the notation, we denote
vq =
q
X
vj ,
vq = vq1 + vq + vq+1 .
j=1
1/l
X
s
kf kBr,l
=
(sq kfq kr )l .
q1
Then
n
o
s
Br,l
(Tn ) = f S 0 : kf kB s < ,
r,l
+3
r
In this section we construct initial data u0 Br,
3
r +3
3
r +3
Br,
Br,
for the MHD system with finite energy. The construction is
similar to the one in [7].
Let , > 3/2. We take any strictly decreasing sequence {qj } such that
(3.9)
4
23
qi
qi+1 ,
4
q23
,
qi
i+1
4
q+3
.
qi
i+1
kk
.
|k|2
We denote
~e1 (k) = p(k)~e1 ,
k Z3 \ {0} ,
qj F (1 () + 2 ())
(3.10)
j1
Due to the fact that () is flat around spheres || = q , one can check that for c
small enough we have
F(Uqj )() =
qj 1 (),
F(Uqj 1 )() =
qj 2 (),
F(Uqj +1 )() = 0.
+3
r
Lemma 3.1. Let > 3/2. For all 1 < r , we have u0 , u0b Br,
3
r +3
b0 Br,
. In particular, u0 , u0b H
32 s
and b0 H
32 s
and
Proof: We only prove the conclusions for u0 . On the block Lj , for 1 < r < ,
we have, by the boundedness of the Leray-Hopf projection and the Lp estimate of
Drichihlet kernel (see [10])
1
1
k
(~e2 ()Lj )kr .
(Lj )kr
qj F
qj kF
3(1 r1 )
.
qj qj
+3
qrj
kUqj kr . 1,
+3
qrj
kUqj 1 kr . 1.
+3
r
Therefore, u0 B r,
.
qj
Z
1d +
Aj A
1d +
Cj
Z
1d
C
j
3
qj .
3
Therefore, u0 B ,
. In particular, for r = 2, the embedding B 2,2 H 2 s
holds for all s > 0. Similar conclusion holds for b0 .
3
3 p
kUqj 1 kp + kUqj kp . qj
3
3 p
kBqj kp . qj
The following estimates are essential to produce the discontinuity of the weak
solutions.
Lemma 3.3. Let u0 , u0b , b0 be defined as in (3.10)-(3.11). Then the trilinear
terms satisfy
B(u0 , u0 , Uqj ) B(u0b , u0b , Uqj ) 73
,
qj
B(u0b , b0 , Bqj ) q72
,
j
obtained in a similar way. Note that supp Ui supp Uj = for any |ij| 2.
We decompose the term as
X
q , U
q , Uq ) + B(U
q , U
q , Uq )
B(u0 , u0 , Uqj ) =
B(U
j
j
j
j
k
k
kj+1
q , Uq ) + B(U
q , Uq , Uq )
+ B(Uqj1 , U
j
j
j
j1
j
X
q k22 . 4
kU
qj
k
kj+1
.
32
qk
kj+1
4
qj
q23
j+1
Similarly,
|III| . kUqj k32
2
X
kj1
qk kUqk k .
4
qj1
23
qj
.
.
(4.13)
u = 0,
(4.14)
for some positive constants c1 and c2 . One contradiction argument will lead the
conclusion of the theorem.
Suppose that for every > 0 there exists t0 = t0 () > 0 such that ku(t)
U k r3 3+ < for all 0 < t t0 . Denoting w = u U , it follows
Br,
3 r3
kwp kr p
for all p 1.
After writing
B(u, u, uqj ) B(U, U, Uqj )
=B(w, U, Uqj ) + B(u, w, Uqj ) + B(u, u, wqj ) = A + B + C,
we estimate each term through the Bonys para-product (c.f. [2]) decomposition as
follows.
X
q , Uq )
A=
B(wp0 , Up00 , Uqj ) + B(wqj , U
j
j
p0 ,p00 qj
|p0 p00 |2
+ B(w
qj , Uqj , Uqj ) rA = A1 + A2 + A3 rA .
with rA being the overlap of A2 and A3 . Later rB , rC , rD , rE , rF have the same
meaning. Combining H
olders inequality, and Bernsteins inequalities, we obtain
X
X
r
. 4
32
. 73
,
|A1 | kUqj k
kwp0 kr kUp00 k r1
qj
qj
p00
for > 32 ;
q , wq )| kUq k kU
q k r kwq kr
|A2 | = |B(Uqj , U
j
j
j
j r1
j
3 X
3
32+ r
4 r
73
. qj
p
. qj
pqj
for +
3
r
< 4;
r kU
|A3 | qj kUqj k r1
qj kr
qj k kw
X
42
3
73
p . qj
. qj
pqj
(4.15)
for
3
3
< < 3, and + < 4.
2
r
We decompose B similarly,
X
B=
B(up0 , wp00 , Uqj ) + B(uqj , w
qj , Uqj )
p0 ,p00 qj
|p0 p00 |2
+ B(
uqj , wqj , Uqj ) rB = B1 + B2 + B3 rB ;
2r
kB1 k . qj kUqj k r2
+ r3
2 r3
kup k2
pqj
9
2 2
qj
3
X qj + r 2
9
2
kup k2 . q2j kup k2
p
pqj
for +
3
r
2;
9
2
2r kw
qj , uqj ) kUqj k r2
qj kr kuqj k2 q2j kuk2 ;
|B2 | = B(Uqj , w
+ r3
2
2r . qj
|B3 | k
uqj k2 kwqj kr kUqj k r2
k
uqj k2
X
pqj
9
2 2
qj kuk2 ,
3
3
2 + r
qj
3
3
2 + r
qj
if +
kuk2 ,
3
r
if +
qj kuk2 ,
< 3,
3
r
if +
> 3,
3
r
= 3.
We thus obtain
(4.16)
|B| .
9
2 2
qj kuk2 ,
3
3
2 + r
qj
3
3
2 + r
qj
kuk2 ,
qj kuk2 ,
if 2 +
if +
3
r
if +
3
r
< 3,
> 3,
3
r
= 3.
Similarly,
C=
3 r3
p0 ,p00 qj
|p0 p00 |2
+ B(
uqj , uqj , wqj ) rC = C1 + C2 + C3 rC ;
10
|C1 | kwqj kr
4 r3
2r . qj
k
up k2 k
up k r2
pqj 2
pr
kup k22
pqj 2
3
2
. 2
for r ;
qj kuk2 ,
2
2
2
2r kwq kr .
|C2 | kuk2 k
uqj k r2
qj kuk2 ;
j
3
2r kuk2
uqj k2 .
|C3 | . qj kwqj kr k
uqj k r2
qj k
Thus,
2
3
|C| . 2
uqj k2 .
qj kuk2 + qj k
(4.17)
3
r
or if 3 < + < 4,
Z t0
B(u, u, uq ) B(U, U, Uq ) ds
j
j
0
Z t0
3/2+ r3 1/2
2
3
.73
t
+
t
+
k
uqj (s)k2 ds
qj
0
qj
qj
qj
0
0
Z t0
2+ r3 11
25
24
2 1/2
.73
t
+
t
+
k
u
(s)k
ds
;
qj
0
qj
2
qj
qj
qj
0
0
otherwise, if + 3r = 3,
Z t0
B(u, u, uqj ) B(U, U, Uqj ) ds
0
Z
2+ r3 11
1/2
25
24
73
2
qj t0 + qj + qj
.qj
t0 + qj
t0
k
uqj (s)k2 ds .
we can chose small enough and large enough j0 such that for j j0
Z t0
B(u, u, uqj ) B(U, U, Uqj ) ds c1 73
t0 .
2c2 qj
0
In the second and third cases, we assume
3 11
< 0.
r
2
Then for small enough and large enough j0 , the same estimate holds. Going back
to (4.14) it implies
2,
2 +
k
uqj (t0 )k22 kUqj k22 E(t0 ) + c1 73
t0 /2,
qj
11
for all j > j0 , which shows that u(t0 ) has infinite energy, a contradiction. In the
end, we collect the conditions on the parameters and obtain that, either
3
< 2,
2
3
< 2,
2
2+
r
3
,
2
3
< 3;
r
or
3+
3
< 4,
r
2 +
3
11
<
.
r
2
Denote Eb (t) =
(5.18)
Rt
0
1
1
kbq (t)k22 kBqj k22 Eb (t) + c1 q72
t
j
2 j
2
Z t
c2
|B(u, b, bqj ) B(u0b , b0 , Bqj )| + |B(b, u, bqj ) B(b0 , u0b , Bqj )|ds
0
1
t c2 R(t)
kBqj k22 Eb (t) + c1 q72
j
2
where R represents the remainder term, and c1 , c2 are positive constants. Again,
we use a contradiction argument to show the conclusion of the theorem.
Assume for every > 0 there exists t0 = t0 () > 0 such that
(5.19)
ku(t) u0b k
+ kb(t) b0 k
3+
B r, r
< ,
3+
B r, r
t0 , for all j j0 .
2c2 qj
In the following we compute the second term in R to obtain the desired estimate.
The first term can be estimated similarly.
Let w = u u0b and y = b b0 . Note that by the assumption (5.19), we have
(5.21)
3 r3
kwp kr p
3 r3
kyp kr p
for all p 1.
12
We decompose D as
X
D=
+ B(
yqj , (u0b )qj , Bqj ) rD
D1 + D2 + D3 rD
with rD being the overlap of D2 and D3 .
Applying the H
olders inequality, (5.21) and (3.12) yields,
X
r
|D1 | kBqj k
kyp0 kr k(u0b )p00 k r1
p0 ,p00 qj ,|p0 p00 |2
3 r3
. 4
qj
p0
pr00
72
,
qj
for + > 3;
)q , yq )| kBq k kU
q k r kyq kr
|D2 | = |B(Bqj , (U
j
j
j
j
j r1
X
3+ r3
4 r3
. q72
. qj
p
j
pqj
for +
3
r
< 4;
42
r kU
|D3 | k
yqj kr kBqj k r1
qj k . qj
3
. q72
p
j
pqj
(5.22)
We decompose E as,
E=
2r
|E1 | kBqj k r2
kbp0 k2 kwp00 kr
+ r3
. q2j
2 r3
p0
kbp0 k2
p0 qj
9
. q2j
p0 qj
9
. q2j
for +
3
r
qj
p0
+ r3 2
kbp0 k2
kbk2
2;
9
2r kw
|E2 | = |B(Bqj , w
qj , bqj )| kBqj k r2
qj kr kbqj k2 . q2j
kbk2 ;
13
2r . k
2r
|E3 | kbqj k2 kwqj kr kBqj k r2
bqj k2 kBqj k r2
3 r3
pqj
3
3
2 + r
qj
kbk2
3 r3
p
X
pqj
9
2
kbk2 ,
qj
3
3
2 + r
qj
3
3
2 + r
qj
kbk2 ,
3
r
+ 3r
if +
< 3;
if
> 3,
qj kbk2 ,
if +
3
r
= 3.
Hence
|E| .
(5.23)
9
2
kbk2 ,
qj
3
3
2 + r
qj
3
3
2 + r
qj
kbk2 ,
qj kbk2 ,
if 2 +
if +
3
r
if +
3
r
< 3;
> 3,
3
r
= 3.
4 r3
qj
r
kbp0 k2 kup00 k2 1
p0 p00
for
3
r
2;
2r kbq k2
|F2 | = |B(yqj , u
qj , bqj )| kyqj kr k
uqj k r2
j
2
2
. 2
uqj k2 kbqj k2 . 2
qj k
qj (kuk2 + kbk2 ),
4
3
2r kuq k2 kyq kr .
|F3 | kbqj k r2
qj kbqj k2 kuk2 . qj kbqj k2 .
j
j
3
2
2
2
3
|F | . 2
qj (kuk2 + kbk2 ) + qj kbqj k2 .
(5.24)
t0
t
+
k
b
(s)k
ds
. 72
t
+
qj
qj
2
0
qj
qj
qj
0
0
Z t0
1/2
. 72
t0 + t0 +
kbqj (s)k2 ds
qj
0
14
t0
kbqj (s)k2 ds
3
r
11
2
for + + <
and + 4.
Therefore we choose j0 large enough and small enough such that, for all j j0
Z t0
c1 72
|B(b, u, bqj ) B(b0 , u0b , Uqj )|ds
qj
t0 .
4c
2
0
The first term in the integral R can be estimated analogously and satisfies
Z t0
c1 72
|B(u, b, bqj ) B(u0b , b0 , Uqj )|ds
qj
t0 ,
4c
2
0
for all j j0 . Therefore, we have shown that the claim (5.20) holds under the
assumption (5.19). It follows from (5.18) and (5.20) that
1
c1
1
kbqj (t0 )k22 kBqj k22 Eb (t0 ) + 72
t0
2
2
2 qj
which implies kb(t0 )k2 is infinity. It is a contradiction which is obtained under the
conditions
3
5
3
3
3
< < 3,
< , + 4, 2 + < 3,
+ < 4;
2
2
2
r
r
or
3 3
3
3
3
3
11
r , < < 3, > , + 4, + 3, + < 4, + + <
.
2 2
2
r
r
r
2
Anagolous analysis will give a contradiction that ku(t0 )k2 is infinity at a certain
time t0 under an alternate assumption on the parameter triplet (r, , ).
References
[1] H. Bahouri, J. Chemin, and R. Danchin. Fourier analysis and nonlinear partial differential equations. Grundlehrender Mathematischen Wissenschaften, 343. Springer, Heidelberg,
2011.
[2] J. Bony. Calcul symbolique et propagation des singularit
e pour les
equations aux d
eriv
ees
partielles non lin
eaires. Ann. Ecole Norm. Sup., 14: 209246, 1981.
[3] J. Bourgain and N. Pavlovi
c. Ill-posedness of the Navier-Stokes equations in a critical space
in 3D. Journal of Functional Analysis, 255: 22332247, 2008.
[4] A. Cheskidov, P. Constantin, S. Friedlander and R. Shvydkoy. Energy conservation and
Onsagers conjecture for the Euler equations. Nonlinearity, 21(6): 12331252, 2008.
[5] A. Cheskidov and M. Dai. Norm inflation for generalized Magneto-hydrodynamic system.
Nonlinearity, 28: 129142, 2015.
[6] A. Cheskidov and M. Dai. Norm inflation for generalized Navier-Stokes equations. Indiana
University Mathematics Journal, 63(3): 869884, 2014.
[7] A. Cheskidov and R. Shvydkoy. Ill-posedness of basic equations of fluid dynamics in Besov
spaces. Proc. Amer. Math. Soc., 138(3):10591067, 2010.
[8] M. Dai, J. Qing, and M. E. Schonbek. Norm Inflation for Incompressible Magneto1,
Hydrodynamics System in B
. Advances in Differential Equations, 16(78): 725746,
2011.
[9] G. Duvaut, J. L. Lions. In
equations en thermo
elasticit
e et magn
etohydrodynamique. Arch.
Rational Mech. Anal., 46(4): 241279, 1972.
[10] L. Grafakos. Classical Fourier analysis. Second edition. Graduate Texts in Mathematics,
249. Springer, New York, 2008.
[11] L. Grafakos. Modern Fourier analysis. Second edition. Graduate Texts in Mathematics, 250.
Springer, New York, 2009.
15