N Seem Il Finished

Lecture Notes: Navier-Stokes Equations
Emil Wiedemann, Edited: Jack Skipper
Universität Ulm, Winter 2018/19

Preface
These are lecture notes for the advanced master's course on the 3D incompressible
Navier-Stokes equations at Universität Ulm in winter term 2018/19. Except for the rst
and the last chapter, the notes follow the excellent recent textbook [ ].4 Students are
1 3, 5].
encouraged to consult further literature, such as the classical books [ ,
Except for the very end of the course, I chose to work exclusively on the three-
dimensional torus such as to simplify the presentation. However all mentioned results
from the rst four chapters have a straightforward extension to the whole space R3 , or
to (suciently regular) bounded domains, which certainly represent the physically most
relevant case.
I would like to thank Dr. Jack Skipper for typing considerable parts of these notes.
Corrections and suggestions should be sent to emil.wiedemann@uni-ulm.de.
Contents
Preface 2
Chapter 1. Introduction 4
1.1. Physical Derivation (Sketch) 4
1.2. Elementary Mathematical Properties 5
1.3. Related Models 7
Chapter 2. Function Spaces and Weak Solutions 8

2.1. Fourier Characterisation of Sobolev Spaces 8
2.2. Helmholtz Decomposition 9
2.3. The Stokes Operator 11
2.4. Weak Solutions 12
Chapter 3. Existence of Weak Solutions 17

3.1. Galerkin Approximation 17
3.2. The Existence Proof 21
Chapter 4. Strong Solutions 25

4.1. Some More on Bochner Spaces 25
4.2. Properties of Strong Solutions 27
4.3. Local Existence of Strong Solutions 31
4.4. Regularity of Strong Solutions 33
4.5. Blowup and the Beale-Kato-Majda Criterion 36
Chapter 5. The Vanishing Viscosity Limit 38

5.1. The Periodic Case 38
5.2. Bounded Domains 41
Bibliography 44
3
CHAPTER 1
Introduction
The Navier-Stokes equations are
∂t ux, t u © ux, t ©px, t ν∆ux, t

div ux, t 0.
Here, x, t > Ω 0, T , where Ω ` Rd some domain, and we have the unknown velocity
eld
u Ω 0, T Rd ;
the unknown pressure eld
p Ω 0, T R;
and the given constant viscosity ν A 0. It can be written in components, for i 1, . . . , d:
d d
∂t ui Q uj ∂j ui ∂i p ν Q ∂x2 ui
j
j 1 j 1
d
Q ∂j uj 0.
j 1
The Navier-Stokes Equations (NSE) describe the time evolution of the velocity and
pressure of a viscous incompressible uid (e.g. water) without external forces.
1.1. Physical Derivation (Sketch)

Conservation of mass: At every time a volume element Ω̃ `` Ω should conserve the
mass of uid (incompressibility). This means that inow and outow of u into Ω̃ have to
balance:

ux, t nx dS x 0,
∂ Ω̃
where nx is the outer unit normal of the surface ∂ Ω̃ at the point x. For regular enough
boundary ∂ Ω̃ and u by the Gauss-Green theorem, the surface integral is equal to

div ux, t dx,
Ω̃
and since this should equal zero for every Ω̃, we conclude div u 0 everywhere in Ω.
Conservation of momentum/Newton's 2nd law: Consider a uid particle initially lo-
cated at x > Ω and denote its location at time t by X x, t (Lagrangian description).
Newton's 2nd law for this particle (point) reads F ma, and by assuming constant
density ( m 1) we obtain
Ẍ x, t F X x, t, t.
The particle trajectory map is determined by the ODE
Ẋ x, t uX x, t, t,

X x, 0 x,
4
1.2. ELEMENTARY MATHEMATICAL PROPERTIES 5
because the particle moves according to the ow of u. Therefore, by the chain rule,
Ẍ x, t ∂t uX x, t, t Ẋ x, t © uX x, t, t

∂t uX, t uX, t © uX, t
where the second term represents the phenomenon known as advection.

Even without external forces (like gravity), there are two kinds of internal forces:
The pressure: the uid pushes itself due to incompressibility, and a force results called
the pressure gradient © p. Example: rotating uid in a disk, where the pressure gradient
is precisely the centrifugal force so © p is the centripetal force, orthogonal to u.
The friction due to viscosity: In a discrete setting, the velocity dierences between
neighbouring uid particles would cause a friction force proportional to
ux hej , t ux, t.

Summing over all neighbours of x, we obtain
d
Q ux hej , t 2ux, t ux hej , t
h2
j 1
1
where is the appropriate scaling; indeed, then this expression is precisely the discrete
h2
Laplacian, which converges, as h 0, to ∆ux, t.
In total, we obtain
∂t uX, t u © uX, t ©pX, t ν∆uX, t,

i.e. the NSE.
1.2. Elementary Mathematical Properties

The incompressible NSE
∂t ux, t u © ux, t ©px, t ν∆ux, t

div ux, t 0,
have a good part of parabolic nature (i.e. the heat equation ∂t u ν∆u). The bad parts
non-linear advection term u © u and non-local terms © p and div u. Note there is no
evolution law for the pressure.
1.2.1. Energy balance. If u is smooth we can multiply the (NSE) by u and integrate
over (space) Ω to obtain

∂t u u dx u © u u dx © p u dx ν ∆u u dx. (1.1)
Ω Ω Ω Ω
The rst term of (1.1) becomes

1d 2 1d 2
SuS dx Yu , tYL2 Ω .
2 dt Ω 2 dt
Further, we note, by integrating by parts, that

u © u u dx Q uiuj ∂j ui dx Q ∂j uiuj ui dx Q ui∂j uj ui dx.
Ω Ω i,j Ω i,j Ω i,j
Thanks to incompressibility (div u 0 or Pi ∂iui 0) we see that the last term vanishes,
whereas the remaining one is precisely the negative of the left hand side, hence

u © u u dx 0.
Ω
6 1. INTRODUCTION
For the term involving the pressure in (1.1) we can also integrate by parts to obtain

© p u dx p div u dx 0
Ω Ω
again by incompressibility. For the last term on the RHS of (1.1) we integrate by parts
and see that

2
ν ∆u u dx ν S©uS dx.
Ω Ω
In total, after also integrating in time, we obtain

t
1 2 2 1 2
Sux, tS dx ν S©ux, sS dx ds Sux, 0S dx.
2 Ω 0 Ω 2 Ω
This suggests that u > L2t Hx1 Ω 9 Lt L2x Ω ª

is a suitable function space for NSE (the
so-called energy space ).
1.2.2. Elimination of pressure. Note that, by virtue of incompressibility (div u
0), the nonlinearity can be written in divergence form (using Einstein's summation con-
vention):
u © ui uj ∂j ui ∂j uj ui div u a ui ,
were we wrote u a uij ui uj and the divergence of a matrix eld is taken row-wise: Let
A Ω 0, T Rd d
then div A is a vector eld given by div Ai Pj ∂j Aij .
Hence the NSE can be written in divergence form as
∂t u divu a u ©p ν∆u,
div u 0.
Take the divergence of the NSE and we obtain
div ∂t u div divu a u div ©p µ∆ div u

and as div u 0 both the rst term and the last term vanish. Further, we note that
div ©p ∆p and so we obtain
∆p div divu a u,
which is a Poisson equation for the pressure. (In the case of a bounded domain this would
be supplemented by a Neumann boundary condition.)
If u > L2 then this can be solved by some distribution p, and we can write this (sym-
bolically) as
1
p ∆ div divu a u
and hence
1
∂t u divu a u ∆ div divu a u ν∆u.
1
The operator ∆
is given as a singular integral operator: E.g. in R3 we have

f y
∆ 1f
C dy.
R3 Sx y S
This is a non-local operator: Even if f is compactly supported, ∆ 1f

will, in general, not
be. For the NSE this means that uid particles may interact, through the pressure, even
when they are far away from each other.
1.3. RELATED MODELS 7
Existence of weak solutions Uniqueness Regularity

d 2 Yes Yes Yes
d 3 Yes (We will show) Unknown/no Unknown (Millennium Problem!)
Table 1. State of the Art for incompressible NSE
1.3. Related Models

1.3.1. Ideal uids, Euler. We can set ν 0 and thus model Ideal uids without
friction. This gives the Euler equations
∂t u divu a u ©p 0
div u 0.
Here without the parabolic term from the Laplacian, the mathematical theory is very
dierent.
1.3.2. Compressible uids. We can study compressible uids (like air) where we
have an extra non-negative scalar eld ρ modelling the density:
∂t ρu divρu a u ©pρ div S©u

∂t ρ divρu 0,
the (isentropic) compressible Navier-Stokes equations. Here, S denotes the Newtonian
stress tensor, and the pressure is now a constitutively given function of the density (e.g.
the polytropic pressure law pρ ργ , γ A 1 the adiabatic exponent).
1.3.3. Non-Newtonian uids. To study non-Newtonian uids (like blood), we re-

place the ∆u with div S©u, where S is non-linear, e.g. the p-Laplacian:
p 2
S©u S©uS u
©
and we recover the standard NSE for p 2. (To be precise, one usually uses only the
symmetric part of ©u.)
The NSE are widely used by physicists, engineers, geo-scientists etc. for atmospheric
and ocean dynamics, weather forecasting, turbulence theory, etc.
CHAPTER 2
Function Spaces and Weak Solutions

We choose as a domain the three dimensional torus T3 R3 ~2πZ3 ; it has the advantages
of being compact and having no physical boundaries at the same time. In other words, we
look for space periodic solutions:
ux 2πk, t ux, t ¦k > Z3 .

The analysis of functions on T3 is simplied by the Fourier series: For u > L1 T3 , meaning
2π 2π 2π
SuxS dx Sux1 , x2 , x3 S dx1 dx2 dx3 @ ª,
T3 0 0 0
we can dene the Fourier coecients

1
ûk e ik x
ux dx > Ck , k > Z3 .
2π 3 T3
If Pk>Z3 Sûk S @ ª, then the Fourier inversion formula says that
ux Q ûk eik x.

k>Z3
We only work with real-valued functions u, which implies that ûk û k , for k > Z3 .
By Plancherel's Theorem, for u > L T3 we have that
2

SuxS
2
dx 2π
3
Q Sûk S2,
T3 k>Z3
and in particular u > L2 if and only if û > l2 , i.e.
Q Sûk S 2
@ ª.
k>Z3
2.1. Fourier Characterisation of Sobolev Spaces

Let s > N, then one usually denes, with the multi-index α > Nd0 , the Sobolev norm
< =
3@ A
2
YuYH s T3
2
YuYL2 T3 Q Y∂ αuY2L 2 T3 2π @
@
@k>Z3
Q
Sûk S
2
Q Q S∂Å
α u S2 A ,
k A
A
SαSBs > SαSBs k >Z3 ?
where we used Plancherel's Theorem in the last equality. Note that the derivatives are
taken in the weak (distributional) sense.
Further, we can integrate by parts to see that

Å
1 ik x 1 ik x
∂ u
j k e ∂j ux dx ikj e ux dx ikj ûk
2π 3 T3 2π 3 T3
and hence
Å
S∂ αu S 2
Sk S
2SαS
Sûk S
2
k
and thus
2
YuYH s T3 2π
3
Q Q SûS2SkS2 α . S S
k>Z3 SαSBs
8
2.2. HELMHOLTZ DECOMPOSITION 9
It turns out (exercise!) that this is equivalent to the norm

2
YuYH s T3 2π
3
Q 1 Sk S
2s
Sûk S
2
.
k>Z3
Note that this is even well-dened when s >~ N!
Definition 2.1. Let s C 0, then H T3
s
contains all functions u > L1 T3 such that
YuYH s
2
2π
3
Q 1 Sk S
2s
Sûk S
2
@ ª.
k>Z3
When s > N, this denition coincides with the denition by weak derivatives. It will be
useful to consider homogeneous Sobolev spaces, where the zero-th Fourier mode is zero.
Definition 2.2. (1) The homogeneous space L̇2 T3 consists of all u > L2 T3
such that

uxdx 0 i.e. û0 0 .
T3
(2) The homogeneous Sobolev space Ḣ s T3 is dened as H s T3 9 L̇2 T3 , with the
norm
YuY
2
Ḣ s
2π
3
Q Sk S
2s
Sûk S
2
.
k>Z3 0
(3) For s @ 0, we dene H s

T3 as the dual space of Ḣ s T3 .
An element v>H s
T3 can be represented as
v x Q v̂k eikx

k>Z3 0
with
2
Yv YH s T3 Q SkS 2sSv̂k S2 @

ª .
kx0
Indeed, as the dual pairing is given by
v, u Q v̂k ûk Q v̂k û k ,
k x0 k x0
using Cauchy-Schwarz we obtain
1~2 1~2
Sv, uS B Q Sv̂k SSû k S Q SSkv̂kSsS Sû k SSkSs B Q SkS 2sSv̂k S2

Q SkS2sSûk S2 @ ª.
kx0 k x0 kx0 kx0
2.2. Helmholtz Decomposition

Consider now vector-valued maps u > L̇2 T3 ; R3 , i.e. u u1 , u2 , u3 with u1 , u2 , u3 >
2 3
L̇ T ; R. Recall the incompressibility (divergence-free) condition div u 0, which in
terms of Fourier coecients reads
3 3
0 Å
div uk Q ∂Ä
j uj i Q kj ûj k ik ûk .
j 1 j 1
(Note that now ûk > C3 .) This motivates the following denition of solenoidal (i.e.
divergence free or incompressible) vector elds.
Definition 2.3. We dene the space H H T3 as
u > L̇ T ; R2 3 3
k uk 0 ¦ k x 0 .
Note that H may contain vector elds that are not in H 1 and hence div is not well
2
dened simply by taking derivatives. H is equipped with the L norm.
10 2. FUNCTION SPACES AND WEAK SOLUTIONS
Lemma 2.4. Every u > H T3 is weakly divergence free in the sense that

ux © φx dx 0
T3
for all φ > Ḣ 1 T3 .
Proof. We can write

u as a Fourier series as ux Pjx0 ûj eij x, then using orthogo-
nality in the form eij x e ik x
dx δjk , we obtain

ux © e
ik x
dx Q ûj e ij x
ik e
ik x
dx ik ûk dx 0
T3 T3 j x0 T3
since u > H T3 . Further, since e

ik x
k>Z3 0 form an orthonormal basis (ONB) of
Ḣ T1 d
, the computation extends to any φ > Ḣ 1 T3 .
3
Definition 2.5. The space G G T is dened as
G g > L̇ T ; R
2 3 3
g © φ for some φ > Ḣ 1 T3 .
Hence Lemma 2.4 says that G and H are orthogonal subspaces of L̇2 .
Theorem 2.6 (Helmholtz decomposition). L̇2 G`H, i.e. for all u > L̇2 T3 ; R3 there
exist unique g > G and h > H such that

u gh and g h dx 0.
T3
Moreover, if u > Ḣ s T3 then g © φ for φ > Ḣ s 1
T3 and h > Ḣ s T3 .
Proof. We can write u as a Fourier series and see that ûk k>Z3 > l2 . We can then
write each ûk as a linear combination of k and a vector wk perpendicular to k . Thus for
all k x 0, let ûk αk k wk with k wk 0 and αk > C. Note that by orthogonality
2 2 2 2
Sûk S Sαk S Sk S Swk S (2.1)
and hence
ux Q ûk eik x Q αk k wk eik x Q iαk ©eik x Q wk eik x © φx hx,
kx0 kx0 k x0 k x0
where
φ x Q iαk eik x and hx Q wk eik x.

k x0 k x0
Note that
Yφ Y
2
Ḣ 1 Q Sαk S2SkS2 and YhYL̇2 Q Swk S2
k x0 k x0
and so, as
Q Sαk S2SkS2 Swk S

2
Q Sûk S2 @ ª ,
k x0 k x0
hence φ > Ḣ 1 and h > L̇2 and thus g > G and h > H .
Further, suppose that u > Ḣ s T3 and multiply (2.1) by Sk S2s to conclude that φ > Ḣ s 1
and h > Ḣ s .
Finally, we must show uniqueness. Suppose that u h1 ©φ1 h2 ©φ2 for h1 , h2 > H
and © φ1 , ©φ2 > G, then
h1 h2 ©φ1 φ2 0 implies that Yh1 h2 ©φ1 ©φ2 Y2 0

and then we can apply Lemma 2.4 to see that
Yh1 h2 Y2 Y©φ1 ©φ2 Y2 0

and so h1 h2 and © φ1 © φ2 .
2.3. THE STOKES OPERATOR 11
Definition 2.7. The orthogonal projection from L̇2 T3 onto H is called the Leray
projection : If u hg with h>H and g > G, then Pu h.
In Fourier series (exercise!)
Pux Q ûk
ûk k
Sk S2
k eik x .
k x0
Note that the following useful lemma only holds true when the spatial domain has no
physical boundaries.
Lemma 2.8. P commutes with derivatives, i.e. P∂xj ∂xj P.

Proof. In Fourier series we see that
Ä
∂
Å Ä j uk k ûk k ûk k
P∂j uk ∂ j uk k ikj ûk ikj k ikj ûk k ∂Å
j Puk
Sk S2 Sk S2 Sk S2
and so we are done.
2.3. The Stokes Operator

Definition 2.9. The space V V T3 is given by V H 9 H 1 T3 ; R3 , with the Ḣ 1 -
norm.
That is, V consists of weakly divergence-free vector elds with extra regularity H 1.
Definition 2.10 (Stokes operator). The Stokes operator is dened as P∆, in the
domain V 9 H 2 T3 ; R3 .
We notice that from Lemma 2.8 that if u>V 9 H 2, then
P∆u ∆Pu ∆u
since u > H. Hence the Stokes operator is simply ∆. However, on bounded domains this
is no longer true in general we cannot necessarily commute derivatives with the Leray
projector. (On a bounded domain one includes information on the boundary condition in
the denition of the space H; this amounts to a weak formulation of the slip condition
u n 0 on ∂Ω.)
Theorem 2.11. There exists a family wk k>N of smooth vector elds on T3 such that
(1) wk is an orthonormal basis of H ,
(2) wk are eigenfunctions of the Stokes operator with eigenvalues 0 @ λ1 B λ2 B B
λj , ª
(3) wk form an orthogonal basis of V .
Proof. For each k > Z3 0 choose vectors mk , m k > R3 such that
Y Ù
mk k , m k k , mk m k ,
Ù Ù
Y Ymk cosk xYL2 T3 Ymk sink xYL2 T3 1,

Y m k m k .

Then mk cosk x 8 mk sink x ` H : Indeed,

mk cosk x dx 0, mk sink x dx 0,
T3 T3
and k mk cosk x, k mk sink x 0 by choice of mk . Hence, we have an orthonormal
family in H whose members are also in the domain of the Stokes operator (because they
are smooth). Moreover,
∆mk cosk x mk Q ∂j2 cosk x mk Sk S2 cosk x Sk S

2
mk cosk x,
2
so we can dene Sk S λk , and similar for sink x, whence 2 is proved after re-labelling
3
indices from Z 0 to N. To show that these functions form a Hilbert basis of H , let
u>H and write
ux Q ûk eik x 1

2
Q
ûk eik x
1
2
û k e ik x Q

kx0 k x0 kx0
1
2 kx0
Qûk û k cosk x

1
2 kx0
iûk û Q k sink x,
and we see that both ûk û k and iûk û

k are in R3 and perpendicular to k. This
becomes, for some ak , bk , ck , dk , αk , βk > R,
Q a k m k bk m k cosk x Q ck mk dk m k sink x
k x0 k x0
Q a k m k b k mk cosk x
Q ck mk d k mk sink x
k x0 k x0
Q αk mk cosk x Q βk mk sink x.
kx0 k x0
Finally, show orthogonality in V: Indeed, if wk , wl are two of the given eigenfunctions of
the Stokes operator with eigenvectors λk , λl (k x l), then using integration by parts,
©wk , ©wl wk , ∆wl λ l wk , w l 0
by orthogonality in H.
º
Remark 2.12. The Ḣ 1 -norm of wk is λk , because

2 2
Ywk Y S©wk S dx © wk © wk dx wk ∆wk dx
Ḣ 1
T3 T3 T3

2
λk Swk S dx λk Ywk Y2H λk .
T3
We have now dened enough machinery so that we can consider weak solutions to the
NSE.
2.4. Weak Solutions

Suppose that u, p is a smooth solution of the NSE. Then
∂t u u © u ν∆u © p > G.
The requirement of v>G is equivalent, by the Helmholtz decomposition, to

v φ dx 0
T3
for all φ > H. It will be convenient to choose φ from the smooth class of functions
Dσ φ > Cc T3 0, ª

ª
div φt 0 ¦ t C 0.
Note that we do not restrict φ to be zero at t 0. So if φ > Dσ , then the NSE imply
ª
ª
ª

∂t u φ dx dt u © u φ dx dt ν ∆u φ dx dt 0
0 T3 0 T3 0 T3
where the term involving the pressure has been projected away by the choice of test
function. Integration by parts in the ∂t and the ∆ terms gives
ª
ª

u ∂t φ dx dt u © u φ dx dt
0 T3 0 T3
ª

ν © u ©φ dx dt u0 φ0 dx. (2.2)
0 T3 T3
2.4. WEAK SOLUTIONS 13
On the other hand, we have already derived the energy equality:

t
1 2 2 1 02
Sux, tS dx ν S©ux, sS dx ds Su S dx,
2 Ω 0 Ω 2 Ω
which suggests that u > L2 0, ª; V 9 L 0, ª; H is the appropriate function space. Note
ª
that for φ > Dσ and u > L H 9 L2 V , (2.2) is well-dened. Thus, we have the following:
ª
Definition 2.13 (Weak Leary-Hopf solution of the NSE). A vector eld u>L ª
0, ª; H 9
2
L 0, ª; V is called a weak (Leray-Hopf ) solution of the NSE if (2.2) holds for all φ > Dσ .
It will be convenient to check that this denition can actually be tested on a smaller
class of test functions than Dσ . Therefore, set
N
D̃σ φ Q dk twk x dk > Cc ª
0, ª¡ ,
k 1
where wk is the eigenbasis of the Stokes operator from Theorem 2.11. Clearly we have
that D̃σ ` Dσ .
Lemma 2.14. If u > L 0, ; H L2 0, ; V satises
ª
ª 9 ª (2.2) for all φ > D̃σ , then it
even satises (2.2) for all φ > Dσ , i.e. it is a weak solution.
Proof. Let φ > Dσ , then for every t C 0, φt > H , and we can write
Q dk twk x
ª
φx, t
k 1
since wk form a Hilbert basis of H. Set
N
φN Q dk twk x > D̃σ .
k 1
Then φN φ in C 0, ª; V . Indeed,
Q Q
ª ª
sup Yφt φN tY2V sup Y dk twk Y2V sup λk d2k t,

t t k N 1 t k N 1
º
as wk are orthogonal in V and Ywk YV λk (from remark after Theorem 2.11). This
then becomes, as λk increases to ª,
λ2k d2k t

Q Q
ª ª
1
sup B sup λ2k d2k t
t k N 1 λ k λN t k N 1
Q
ª
1
sup ∆wk xdk t, ∆wk xdk tL2
λN t k N 1
by orthogonality. We then see that we can bound this above by
1 1
sup Y ∆φYL2 T3 B sup YφtYH 2 T3 0
λN t λN t
as N ª, since supt YφtYH 2 T3 is independent of N .
Furthermore, ∂t φN ∂t φ in L2 0, T ; L2 T3 because

Q
ª ª ª
2
Y∂t φ ∂t φN YL2 T3 dt Y dk twk xY2L2 T3 dt

0 0 k N 1

Q
ª ª
2
dk t dt 0
0 k N 1
as N ª, where the latter convergence follows from
∂t φ > Cc T3 0, ª ` Cc

ª ª
0, ª; H
and hence
Q dk t2
ª
sup
sup Y∂t φY2L2 T3 @ ª.
t k 1 t
It follows that
ª
ª

∂t φN u dx dt ∂t φ u dx dt,
0 0
ª
ª

© φN © u dx dt © φ ©u dx dt,
0 ª
0
ª

0
φN 0 u x©u dx dt φ0 u0 x dx dt.
0 0
For the remaining term u ©u φN dx dt, we will use the Sobolev embedding H 1 T3 `
L6 T3 , so it follows that
sup YφN φYL6 T3 B C sup YφN φ YV 0

t t
as N ª. Thus, by Hölders's inequality,

V u © u φN φ dx dtV B SuSS©uSSφN φS dx dt
B YuYL2t L3x Y©uYL2t L2x YφN φYLª 6
t Lx
0
as N ª. (We note that using again the same embedding theorem we have YuYL2 L3
t x
B
C YuYL2t L6x B C YuYL2t Hx1 and so YuYL2 L3
t x
@ ª.) So if we consider the equation for a weak
solution to the NSE (2.2) with φN used as a test function, then we see that every term
will converge to the corresponding one with φ > Dσ and so (2.2) follows for φ > Dσ .
For later reference, we prove another lemma which allows us to test a weak solution
with functions of the form χ t1 ,t2 φ for φ > Dσ , for almost every t1 @ t2 , where χ denotes the
indicator function of a set. This is a consequence of the Lebesgue dierentiation theorem,
which we recall without proof:
Theorem 2.15 (Lebesgue dierentiation theorem). Let Ω ` Rn be measurable and

f> L1loc Ω , then for almost every x > Ω we have

1
lim Sf y f xS dy 0.
0 SB xS B x
A point x for which the statement of the dierentiation theorem is true is called a
Lebesgue point of f; the theorem thus says that, given a locally integrable function on a
domain, almost every point in that domain is a Lebesgue point.
Lemma 2.16. Let u be a weak (Leray-Hopf) solution of NSE. Then

t2 t2
t1 T3 t1 T3
t2
ν © u ©φ dx dt ut1 φt1 dx ut2 φt2 dx (2.3)
t1 T3 T3 T3
for every φ > Dσ and almost all 0 B t1 @ t2 , including t1 0.
Remark 2.17. Later we will see that this is even true for all (and not just almost all)
times.
Proof. We prove only the case t1 0 in detail. So let t2 A 0 and consider a smooth
(cut-o ) function ζR R such that
Y ζ C 0,
2.4. WEAK SOLUTIONS 15
ζ t 1 for t B 1 and ζ t 0 for t C 1,

Y
ζ is monotone decreasing.
Y
Then, for every A 0, set
t t2
ζ t ζ .

Thus, ζ (restricted to t C 0) is a smooth approximation of the indicator function χ 0;t2 .
If φ > Dσ , then the product ζφ is still in Dσ , so using it as a test function in the weak
formulation of NSE gives
ª
ª

u ∂t ζ φ dx dt u © u ζ φ dx dt
0 T3 0 T3
ª

ν © u ©ζ φ dx dt u0 φ0 dx
0 T3 T3
(for the term involving u0 , ζ 0
). The two integrals
note that 1 for suciently small
including space derivatives are easily seen to converge as ζ converges almost 0: Indeed,
everywhere to χ 0,t2 , and the integrand u ©u ζ φ is bounded pointwise by Su ©uSSφS,
uniformly in , which is of course integrable. Hence, by the dominated convergence theorem,
ª
ª
t2
u ©u ζ φ dx dt u ©u χ 0,t2 φ dx dt u ©u φ dx dt
0 T3 0 T3 0 T3
as 0, and likewise
ª
t2
© u ©ζ φ dx dt © u ©φ dx dt
0 T3 0 T3
(of course the space derivative does not hit ζ , which depends only on time).
The rst integral, which contains the time derivative, is a bit more delicate. We
compute
ª
ª
ª

u ∂t ζ φ dx dt u ζ φ dx dt u ζ ∂t φ dx dt,
0 T3 0 T3 0 T3
and the latter integral is seen, as before, to converge to
t2
u ∂t φ dx dt.
0 T3

For the integral involving ζ , observe that by denition,

1
t t2
ζ t ζ ,

ª
which also implies, by the fundamental theorem of calculus, that
0 ζ t dt

1 for every
A 0. Note also that ζ
is supported in B t2 . Therefore,
ª

V u ζ φ dx dt ut2 φt2 dxV
0 T3 T3
t2
B
Sζ tS V ut φt ut2 φt2 dxV dt
t2 T3
t2
1
B Yζ Y C 1 V ut φt ut2 φt2 dxV dt 0
t2 T3
as 0, provided t2 is a Lebesgue point of the map

t ( ut φt dx.
T3
Since, by Lebesgue's theorem, this is the case for almost every t2 A 0, by collecting all
terms we nally arrive at (2.3) in the case t1 0.
In the general case, we would use the test function ζ tξ tφx, t, where ζ is as
before and
t1 t
ξ t ζ .

The passage to the limit 0 can then be achieved in exactly the same way as above.
CHAPTER 3
Existence of Weak Solutions

3.1. Galerkin Approximation
3.1.1. A toy example: the heat equation. To illustrate the Galerkin method in
the simplest possible setting, let us consider the Cauchy problem for the heat equation on
the torus:
∂t u ∆u on T3 ,
(3.1)
ut 0 u0 .
Let wk k>N be an eigenbasis of ∆ and project the problem to the nite dimensional
subspace PN H spanw1 , . . . , wN : If ut is in this space for every t, then so is ∂t ut,
and thanks to the eigenfunction property also ∆ut > PN H . Therefore, the projected
version of the heat equation simply reads
∂t uN ∆uN on T3 ,
(3.2)
uN t 0 PN u0 .
This equation is known as the Galerkin heat equation of order N, and we want to solve
it in PN H . To this end, take the ansatz uN x, t P N N
l 1 dl twl x, insert it into (3.2),
multiply by wk (k 1, . . . N ), and integrate in space:
N N
dk t λk dk t 0,
dN
k 0 u
0
, w k L 2 ,
where we used orthonormality and the eigenfunction property of the wl . This is a system
(actually a decoupled one in this simple case) of linear ordinary dierential equations,
which has a global smooth solution by standard ODE theory.
We wish to let N ª and hope to obtain a solution to the original problem in the
limit. To this end, observe that multiplication of (3.2) with its solution uN and integration
in space yields (in analogy to NSE) the energy equality
t
1 2 2
SuN x, tS dx S©uN x, sS dx ds
2 T3 T3
0

1 1
0
SPN u xS dx B 0
Su xS dx,
2 T3 2 T3
and thus a uniform (in N) bound of the Galerkin sequence in L L2

ª
9 L2 H 1 . By the
Banach-Alaoglu Theorem, we may therefore take a weakly*-convergent subsequence (not
relabelled), so that uN @u>L

ª
L2 . Hence for every φ > Cc T3 0, ª,
ª
we have
ª
ª

uN ∂t φ dx dt u∂t φ dx dt,
0 T3 0 T3
ª
ª

uN ∆φ dx dt u∆φ dx dt,
0 T3 0 T3
17
18 3. EXISTENCE OF WEAK SOLUTIONS
and nally

0
PN u φ t 0dx u0 φt 0dx,
T3 T3
because PN u0 u0 in L2 . It thus follows that u is a weak solution of (3.1). Note that we
2 1
did not make any use of the L H -bound.
3.1.2. Galerkin for NSE. Recall the basis of eigenfunctions of ∆ (now viewed as
the Stokes Operator) from Theorem 2.11. Let PN H spanw1 , . . . , wN and consider the
projection operator PN L̇2 PN H given by
N
PN u Q u, wj wj
j 1
using the L2 inner product. Clearly, for all u > H we have PN u u in H , as an orthonormal
basis of H (i.e. in the L2 -norm). Indeed,
Q Q
ª ª
2 2 2
YPN u uYL2 Y u, wj wj Y Su, wj S 0
j N 1 j N 1
as N ª.
Definition 3.1. The N -th order Galerkin approximation of the NSE with initial data
u0 > H is the solution of the equation
∂t uN PN uN © uN ν∆uN , (3.3)
uN 0 PN u0 .
Note we have not yet proved that such uN exists! Again we have projected away the
pressure. Like for the heat equation, we want to take N ª.
From the energy estimate, we expect a uniform bound of the form

t
sup SuN x, tS
2
dx 2ν S©uN x, sS
2
dx ds B C @ ª, (3.4)
t Ω 0 Ω
so by the Banach-Alaoglu Theorem we will be able to pass to weak limits:
uN @ u, © uN @ © u
2
weakly in L .
So far, the general strategy seems similar as for the heat equation. However, for NSE,
there are two main issues to solve:
First, the Galerkin approximation (3.3) and thus the resulting system of ODEs now
feature a quadratic term, so that the ODE solution is prima facie only obtained up to
a possibly nite blow-up time; indeed, the simplest quadratic ODE, ẋ x2 , does exhibit
nite-time blow-up. Even worse, the existence interval 0, TN might depend on N and
could therefore, in the worst case, converge to zero as N ª, so that in the limit, we
would be left with nothing. It turns out, luckily, that such blow-up scenarios can rather
easily be ruled out by virtue of the nite dimensional energy equality (3.4).
Once we have globally existing Galerkin approximants uN N >N which satisfy the uni-
form bound (3.4), we need to establish that the weak limit u is a weak solution. For the
heat equation (and more generally for linear equations), this was trivial. However, for the
NSE, in the weak formulation we have the nonlinear term
ª

uN © uN φ dx dt
0 T3
and it is not clear whether
uN © uN @ u © u!

3.1. GALERKIN APPROXIMATION 19
For example in 1D we see that
uN x sinN x @0 but u2N x sin2 N x @~ 0,

meaning that weak limits and nonlinearities do, in general, not converge. However, if
we knew uN u strongly and © uN @ © u weakly, then it would follow uN © uN @
u © u weakly. The strong convergence of uN will be obtained by means of the following
compactness result, which can be seen as a time-dependent version of the classical Rellich
compactness theorem:
Lemma 3.2 (Aubin-Lions). Let 0 @ T @ ª and assume for some 1 @ p, q B ª that

YuN YLq 0,T ;V Y∂t uN YLp 0,T ;V B C,
for a constant C x C N . Here, V is the dual space of V . Then there exists a subsequence

uNj j >N such that uNj u strongly in Lq 0, T ; H , for some u > Lq 0, T ; H .
Proof. Consider for k > N the map t ( uN t, wk L 2. It is not dicult to see
(exercise!) that this map is weakly dierentiable with weak derivative ∂t uN t, wk , which
is a well-dened Lp function since ∂t uN > Lpt V
and wk > V , and so (cf. exercise) s (
uN s, wk is (absolutely) continuous and
s

uN s, wk uN s , wk ∂t uN , wk dt
s
for all s, s > 0, T .

Again by continuity we may invoke the mean value theorem for
integrals to conclude there exists an s > 0, T

such that
T

1
uN s , wk uN , wk dt.
T 0
Hence,
s
sup SuN s, wk S B SuN s , wk S V
∂t uN , wk dtV
0BsBT s
T T
1
B YuN YL2 T3 Ywk YL2 T3 dt Y∂t uN YV Ywk YV dt.
T 0 0
We can now use Hölder's inequality combined with the facts that Ywk YL2 T3 1 and
º
Y wk Y V λk to obtain the bound
1 1 1 » »
1 p1
B T

qY uN YLq 0,T ;H T Y∂t uN YLp 0,T ;V λ k B C1 λ k C2
T
where we have used that Y∂t uN YLp 0,T ;V is uniformly bounded and by Poincaré's inequality
we have YuYH B C YuYV , which gives a uniform bound on YuN YLq 0,T ;H . It follows that
Pk uN P k
j 1 uN , wj wj is in C 0, T ; H and
k » »
sup YPk uN sYH B Q C1 C2 λj B kC1 C2 λk
s>0,T j 1
as λj is increasing.
Claim 1: For every k , Pk uN N >N has a subsequence converging in C 0, T ; H . For
this, we will use the Arzelà-Ascoli Theorem. As Pk H is nite-dimensional, it suces to
check that the sequence is uniformly bounded and equicontinuous. From the previous
estimate we already have uniform boundedness, so all we need to check is equicontinuity.
Since Pk H is nite-dimensional, Y YH and Y YV are equivalent in Pk H , and so

t2
YPk uN t2 Pk uN t1 YH ] ∂t Pk uN s ds]
t1 H
t2
B Y∂t Pk uN sYH ds
t1
t2
B Ck Y∂t Pk uN sYV ds
t1
t2
1
p
t2 1 p1
B Ck p
Y∂t Pk uN sYV ds ds
t1 t1
1
B C̃k St2 t1 S1
p
where we used Hölder's inequality and that Y∂t Pk uN YLp t1 ,t2 ;V is uniformly bounded in
N. Thus we may apply the Arzelà-Ascoli theorem, which proves the claim.
Claim 2: uN has a subsequence that is Cauchy in Lq 0, T ; H . Clearly Claim 2
implies the theorem. By a diagonal argument (exercise), we can select a subsequence (still
denoted uN ) so that Pk uN is convergent in Lq 0, T ; H for all k>N (see Claim 1).
q
We will show this sequence is Cauchy in L 0, T ; H .
Claim 2a: For every δ A 0 there exists k > N such that
T
YPk uN s uN sYH
q
ds @ δ
0
for all N C k . Indeed, we know that C C YuN YLq 0,T ;V and as ©wj , ©wk ∆wj , wk
λj wj , wk λj δjk , then
T T ª
q
2

CC Y©uN sY
q
L2
ds Q
j 1
λj SuN s, wj L2 S2

ds
0 0
T
q
2

Q
ª
C λj SuN s, wj L2 S2 ds,

0 j k1
so as λj are increasing we obtain
T
q
2 T
2
Q
q ª q
C λk 2
1 SuN s, wj L2 S ds 2
λk 1 YPk uN s uN sY
q
L2
ds,
j k1

0 0
and now Claim 2a follows from the fact that λ ª.

Next, recall that (for the k from Claim 2a) Pk uN is Cauchy in Lq 0, T ; H , so there is
an N0 > N such that
T
YPk uN s Pk uM sY
q
ds @ δ
0
for all N, M A N0 . By the triangle inequality,
YuN uM YLq 0,T ;H B YuN Pk uN YLq 0,T ;H YPk uN Pk uM YLq 0,T ;H

1
YuM Pk uM YLq 0,T ;H B 3δ q
and since δA0 was arbitrary, Claim 2 follows and we are done.
Only one ingredient is missing before we can embark on the existence proof for weak
solutions:
3.2. THE EXISTENCE PROOF 21
Lemma 3.3 (L
p
Let Ω be a measure space and u > Lp Ω Lq Ω for
interpolation). 9
some 1 B p B q B ª . Then u > L Ω for all p B r B q, and YuYLr B YuYαLp YuY1Lqα , where
r
q .
1 α 1 α

r p
p q
Proof. Using Hölder's inequality with
αr , 1αr we see that
αr 1αr
p p
SuS
r
dx SuS
αr 1αr
SuS dx B SuS
p
dx
q
SuS dx
Ω Ω Ω Ω
α 1α r
YuYLp YuYLq
and so we are done.
3.2. The Existence Proof

Theorem 3.4 (Existence of weak solutions). For every u > H there exists a weak so-
0
lution of the NSE with initial data u0 . Moreover, this solution satises ∂t u > L4loc~3 0, T ; V .
Proof. Step 1: Existence of Galerkin approximations, locally in time. Recall the

N -th order Galerkin equation:
∂t uN PN uN © uN ν∆uN ,

uN 0 PN u0 .
We take the ansatz
N
uN x, t Q dNj twj x
j 1
and multiply the Galerkin equation by wk , and integrate:
N N
∂t Q dN
j twj xwk x dx PN Q dNj tdNl twj x © wl x wk x dx
T3 j 1 T3 j,l 1
N
ν Q dN
j t∆wj x wk x dx.
j 1 T3
By orthogonality of wk in L2 and the eigenfunction property, this gives
N
N
dk t νλk dj t
N
Q dNj tdNl tBkjl 0,
j,l 1
with k 1, . . . , N and

Bkjl wj x © wl x wk x dx.
T3
Indeed, PN v, wk L2 v, wk L2 for all v > L2 by a simple linear algebra argument.
N
This is a system of N ODEs for the N unknown functions dk (k 1, . . . , N ), with
initial condition

dN
k 0 u0 xwk x dx.
T3
(The latter is obtained by multiplying
N
uN x, 0 Q dNj 0wj x
j 1
by wk and employing the initial condition uN x, 0 PN u0 x. 0

Note again PN u x, wk x
0
u , wk x).
By classical ODE theory (Picard-Lindelöf/Cauchy-Lipschitz) there exists a time TN A 0

and a solution dk k 1,...,N > C 0, TN of this system.
N 1
Step 2: Show TN ª for all N , via energy estimates. Let s > 0, TN . Multiply the
Galerkin equation at time s with uN s and integrate in x to get

∂t uN s uN s dx PN uN s © uN s uN s dx
T3 T3

ν ∆uN s uN s dx.
T3
Observe each integral in order: for the rst we see that

1 d 2
∂t uN s uN s dx SuN sS dx,
T3 2 dt T3
for the second,

PN uN s © uN s uN s dx uN s © uN s uN s dx
T3 T3
3
Q ujN s∂xj ulN sulN s dx
j,l 1 T3
3
Q ∂xj ujN sulN s2 dx
j,l 1 T3
3
Q ∂xj ujN sulN s∂xj ulN s dx 0
j,l 1 T3
using incompressibility, and for the last term we see that

2
ν ∆uN s uN s dx ν
S©uN sS dx.
T3 T3
Hence for all s > 0, TN we obtain the (nite-dimensional) energy equality
1 d 2 2
YuN sYL2 T3 ν Y©uN sYL2 T3 0.
2 dt
We note that this equality implies (after integration in s) that
ª
1 1 1 0 2
YuN tYL2 B Yu0 YL2 ds B
2 2 2
sup and Y©uN sYL2 Yu YL2 ,
t 2 2 0 2ν
thus uN are uniformly bounded in Lª
0, ª; H 9 L2 0, ª; V .
In particular, since uN x, s P N N
k 1 dk swk x and since wk is an ONB in L2 ,
d
YuN sYL2
2
Q SdNk sS2
k 1
N
and this is bounded in s. It follows that dk sk 1,...,N is uniformly bounded in s and
hence TN ª.
Step 3: Bound ∂t uN in an appropriate norm (in order to apply Aubin-Lions).

Let φ>V and take the L2 -inner product with the Galerkin equation:
∂t uN , φ ν ∆uN , φ PN uN © uN , φ

ν ∆uN , φ uN © uN , PN φ,
where for the last equality we used the self-adjointness of the projection PN . For the rst
term
Sν ∆uN , φS ν S©uN , ©φS B ν Y©uN YL2 YφYV ,

3.2. THE EXISTENCE PROOF 23
and for the second

1 1
SuN © uN , PN φS B YuN YL3 Y©uN YL2 YPN φYL6 B YuN YL2 2 YuN YL2 6 Y©uN YL2 YPN φYL6
using interpolation (Lemma 3.3). Now we can use the Sobolev embedding Yv YL6 B C Yv YḢ 1
C Y v YV C Y©v YL2 and the projection property YPN φYV B Yφ YV to obtain
1 3
SuN © uN , PN φS B C YuN YL2 Y©uN YL2 YφYV .
2 2
It follows from the denition of the dual/operator norm

1 3
Y∂t uN YV B ν Y©uN YL2 C YuN YL2 2 Y©uN YL2 2
and thus for any 0@T @ª
T 4
T T 2
4
Y∂t uN YV 3
ds B Cν Y©uN s Y 3 ds C 2
YuN Y 3 2 Y©uN sYL2
L
ds.
0 0 0
Using Hölder's inequality in time on both terms (rst L3 (on 1), L3~2 second L ª
, L1 ) we
obtain
4
< T =
1 3
1 @ 2A 2
B CνT 3 @
@ Y©uN sY
2
ds A
A C YuN YL3 ª 0,T ;H YuN sY2L2 0,T ;V
@ 0 A
> ?
which becomes
1 4 2
B CνT 3 YuN sYL3 2 0,T ;V C YuN YL3 ª 0,T ;H YuN sY2L2 0,T ;V
which is bounded uniformly in N (for xed ν and T !) owing to the energy estimates from
Step 2.
Step 4: Extract a convergent subsequence.
By Banach-Alaoglu, there is a subsequence uNj j such that
uNj @u

weak- in Lª
0, ª; H , and another subsequence uNj,l l such that
© uNj,l @w in L2 0, ª; L2 T3 .

It is easy to show that (exercise!) w © u. For any xed 0 @ T @ ª, extracting yet another
subsequence if necessary (not relabelled), the uniform bound on ∂t uN gives
∂t uN @ ∂tu

in L4~3 0, T ; V
.
Even better, by a diagonal argument we obtain a subsequence such that
∂t uN @ ∂tu

in
4~3
Lloc 0, ª; V
.
Choosing yet another subsequence and applying again a diagonal argument, we obtain
by Lemma 3.2 (Aubin-Lions)
uN u
strongly in L2loc 0, ª; H .
Step 5: Show the limit is a solution of NSE.
By Lemma 2.14, it suces to choose a test function of the form
m
φx, t Q dk twk x > D̃σ .
k 1
Let T be so large that suppdk ` 0, T for all k 1, . . . , m. By the Galerkin equation,

for any N C m we have
ª
ª

uN ∂t φ dx dt ν ©uN © φ dx dt
0 T3 0 T3
ª

uN © uN φ dx dt u0 φ0 dx.
0 T3 T3
Weak convergence uNj @u

in Lª
0, T ; L2 T3 gives
ª
ª

uN ∂t φ dx dt u ∂t φ dx dt
0 T3 0 T3
and © uN

@ © u

in L2 0, T ; L2 T3 gives

ª ª
©uN © φ dx dt © u ©φ dx dt.
0 T3 0 T3
Only the non-linear term needs some more attention, we see that
uN © uN u © u uN u © uN u © uN u.

Firstly,
ª
T
V uN u © uN φ dx dtV B Cφ YuN uYL2 Y©uN YL2 dt
0 T3 0
B Cφ YuN uYL2 0,T ;L2 Y©uN YL2 0,T ;L2
and we see that this converges to zero as Y©uN YL2 0,T ;L2 is bounded and YuN uYL2 0,T ;L2
converges to zero as N ª . Secondly,
ª

u © uN u φ dx dt 0
0 T3
as N ª as u > L2t,x and © uN u converges to zero weakly in L2t,x . It follows that
ª
ª

uN © uN φ dx dt u © u φ dx dt
0 T3 0 T3
and so
ª

u ∂t φ ν ©u ©φ u © u φ dx dt u0 φ0 dx.
0 T3 T3
Hence u is a weak solution.
The bound on the time derivative allows us to obtain a useful continuity property in
time, typical for balance equations in continuum mechanics:
Proposition 3.5. The solution constructed in Theorem 3.4 is (after alteration on a

set of times of measure zero, if necessary) contained in the space C 0, ; V , and it ª

satises the statement of Lemma 2.16 even for all (and not just almost all) 0 B t1 @ t2 .
4~3
Proof. This will be proved in the exercises, based on the property ∂t u > Lloc 0, ª; V
.

In fact, one can show that this proposition is true for every Leray-Hopf solution, not
just the (possibly particular) one constructed in Theorem 3.4. Of course, if Leray-Hopf
solutions are unique, this distinction is unnecessary, but uniqueness is still unknown in
three dimensions.
CHAPTER 4
Strong Solutions
4.1. Some More on Bochner Spaces
For this entire chapter, let 0 @ T @ ª be arbitrary but xed. We collect a few technical
results to be used later.
Proposition 4.1. Let X be a Banach space and suppose u, w > L1 0, T ; X . Then the
following are equivalent:
(1) ∂t u w in the weak sense;
(2) There exists ξ > X such that, for a.e. t > 0, T ,
t
ut ξ ws ds;
0
(3) For every v > X , in the weak sense it holds that

d
u, v w, v .
dt
Moreover, if one (and thus all) of these conditions holds, then u can be altered on a nullset
of times so that it belongs to C 0, T ; X .
Proof. Exercise.
Proposition 4.2. Let u > L2 0, T ; Ḣ 1 T3 and ∂t u > L2 0, T ; H 1
T3 . Then,
(1) u > C 0, T ; L̇2 T3 ;
(2) the map t ( 21 YutY2L̇2 is weakly dierentiable with
d1 2
YutY 2
L̇
ut, ∂t ut for a.e. t > 0, T ;
dt 2
(3)
max YutYL̇2 B C YuYL2 Ḣ 1 Y∂t uYL2 H 1 .

t> 0,T
Proof. Let η > Cc R be a standard mollier (in the time variable), that is, η C 0,
ª
R η θ dθ 1, supp η ` B1 0, and η η SθS. Set η t 1 t and, for any f > L1loc R,
f f η , i.e.

f t f t τ η τ dτ.

Consider now u as given in the statement and extend it to t > R by zero, so that its
mollication is well-dened on all of 0, T . For , δ A 0, uδ and u are smooth in time, and
we may thus use the standard Leibniz rule to compute

d 2
Yu t uδ tYL2 2 u t uδ t ∂t u t ∂t uδ t dx. (4.1)
dt T3
Observe that for a.e. s > 0, T , u s us in H 1 T3 and ∂t u s @ ∂tus

in H 1
T3
(exercise). Pick such an s and integrate (4.1) from s to t to obtain
t
Yu t uδ tYL2
2
B Yu s uδ sY2L2 2 Su τ uδ τ , ∂t u τ ∂t uδ τ S dτ
0
25
26 4. STRONG SOLUTIONS
By choice of s, the rst expression on the right hand side converges, as , δ 0, to zero,
and so does the dual pairing under the integral, for almost every τ. But then the integral
itself converges by dominated convergence (exercise).
It follows that u A0 is Cauchy in C 0, T ; L2 T3 , and since this space is Banach,
it follows u u > C 0, T ; L T 2 3
, whence (1) is established.
For (2), again by the classical Leibniz rule,

d 2
Yu tYL2 2 u t ∂t u t dx
dt T3
and hence, after integration from s to t,
t
2 2
Yu tYL2 Yu sYL2 2 u τ ∂t u τ dτ dx, (4.2)
s T3
and the same equality follows for u instead of u by similar convergence arguments as
before. Application of Proposition 4.1 (2) then gives the desired characterisation of the
weak time derivative.
Finally, for (3), integrate (4.2) in s over 0, T to arrive at
T
T Yu tY2L2 B Yu sYL2
2
ds T YuY2L2 Ḣ 1 Y∂t uY2L2 H 1 ,
0
and thus (3) follows because t is arbitrary and the right hand side is independent of t.
Corollary 4.3. Let u, v > L2 0, T ; Ḣ 1 T3 and ∂t u, ∂t v > L2 0, T ; H 1
T3 . Then
the map t ( u, v is absolutely continuous with weak derivative
d
u, v ∂t u, v u, ∂t v .
dt
Proof. This follows from the polarisation identity
1 2 2 2
u, v Yu v Y YuY Yv Y
2
and the preceding proposition.
The following can be seen as an extension of Proposition 4.2 to higher order Sobolev
spaces:
Proposition 4.4. Let n > N0 and suppose u > L2 0, T ; Ḣ n 2

T3 and ∂t u > L2 0, T ; Ḣ n T3 .
Then, u > C 0, T ; Ḣ n 1 T3 , and

max YutYḢ n1 B C YuYL2 Ḣ n2 Y∂t uYL2 Ḣ n .

t> 0,T
Proof. We indicate only the formal argument and remark that the rigorous proof
proceeds exactly as in Proposition 4.2 by time mollication.
So assume u is smooth and take for simplicity n 0, then we compute
d 2 d 2
YutY 1 Y©utYL2
dt Ḣ dt

2 © u ∂t ©u dx
T3

2 ∆u ∂t u dx B YutY2H 2 Y∂t utY2L2 ,
T3
and the estimate
sup YutYḢ n1 B C YuYL2 Ḣ n2 Y∂t uYL2 Ḣ n

t> 0,T
follows, as before, by integrating rst from s to t and then by integrating in s from 0 to

T.
4.2. PROPERTIES OF STRONG SOLUTIONS 27
4.2. Properties of Strong Solutions

Definition 4.5 (Strong solutions). A strong solution of NSE is a weak solution with
additional regularity
u>L ª 1 3 2
0, T ; H T 9 L 0, T ; H T .
2 3
Strong solutions have very nice properties, like energy conservation, uniqueness, and
smoothness; however, given initial data u0 > V , a strong solution is known to exist only on
a possibly nite time interval (whether or not the existence time can actually be nite is
precisely the Navier-Stokes Millennium Problem).
Lemma 4.6. Let u be a strong solution, then ∂t u, u © u, ∆u > L2 0, T ; L2 T3 .
Proof. From the assumption u > L2 0, T ; H 2 it follows immediately that ∆u >
2 2 2
L 0, T ; L . For the nonlinear term, note that H embeds continuously into L , so
ª
that
T T
2
SuS S©uS
2
dx dt B 2
YutYLª S©ux, tS
2
dx dt B C YuY2Lª H 1 YuY2L2 H 2 .
0 T3 0 T3
It remains to estimate the time derivative. In view of Lemma 2.16 and the remark after
Proposition 3.5, for every smooth divergence-free vector eld φ>C ª
T3 (note, no time
dependence here) and every t > 0, T we have
t
0
ut φ dx u φ dx ν ©u ©φ u © u φ dx ds
T3 T3 0 T3
t
0
u φ dx ν∆u φ u © u φ dx ds,
T3 0 T3
as u has weak second space derivatives. Proposition 4.1 allows us to take the time derivative
of this equality to obtain, for every t > 0, T ,

∂t u φ dx ν∆u φ u © u φ dx.
T3 T3
Let ψ > C T3 be any

ª
smooth vector eld and ψ φ ©π its Helmholtz decomposition,
so that div φ 0. Then, as ut > H ,

∂t u ψ dx ∂t u φt dx
T3 3
T
ν∆u Pψ u © u Pψ dx
T3

Pν∆u u © u ψ dx,
T3
and since ψ was arbitrary, it follows that ∂t u P∆u u © u. By the previous estimates,
this is indeed in L2 0, T, L2 .

Lemma 4.7. Let u be a strong solution. Then, for any w > L2 0, T ; H ,
T
∂ t u u © u ν∆u w dx dt 0.
0 T3
Note that the integral in fact is well-dened by the previous lemma.
Proof. By similar arguments as in the proof of Lemma 2.14, the space D̃σ is dense in
2
L 0, T ; H , and therefore it suces to consider w > D̃σ . Using such w as a test function
in the weak formulation gives

uT wT dx u0 w0 dx
T3 T3
T
u ∂t w ν ©u ©w u © u w dx dt.
0 T3
But clearly, since u is a strong solution,
T T
© u ©w dx dt ∆u w dx dt,
0 T3 0 T3
and by Proposition 4.1 and Corollary 4.3 also
T T
0
uT wT dx u w0 dx u ∂t w dx dt ∂t u w dx dt.
T3 T3 0 T3 0 T3
Putting everything together, we arrive at the conclusion.
Lemma 4.8. For any u > V , we have

u © u u dx 0.
T3

Proof. Write bu, u, u
T3 u © u u dx 0, then this is a trilinear form. For
smooth vector elds, we have already established bu, u, u 0 (in the formal derivation of
the energy equality). Therefore, if u u η denotes a standard mollication, we have
Sbu, u, uS B Sbu, u, u bu , u, uS

Sbu , u, u bu , u , uS Sbu , u , u bu , u , u S.
But the rst term is estimated as

Su u SS©uSSuS dx 0
T3
as 0: This follows from the embedding H 1 ` L6 ` L3 and Hölder's inequality with
1 1 1
6 , 2 , 3 . The other two terms are estimated in the same way.
Theorem 4.9 (Energy equality). A strong solution satises the energy equality, i.e.
for every s @ t we have
t
1 2 2 1 2
SutS dx ν S©ux, τ S dτ SusS ds.
2 T3 s T3 2 T3
Proof. In Lemma 4.7, we may take w χ s,t u and thus obtain
T t
0 ∂t uu ©uν∆u uχ s,t dx dτ ∂t uu ©uν∆u u dx dτ.
0 T3 s T3
1 d 2
By Proposition 4.2, u ∂t u 2 dt YuYL2 , and
t
1 2 1 2
u ∂t u dx dτ SutS dx SusS ds,
s T3 2 T3 2 T3
whereas (by a very simple approximation argument)

2
∆u u dx S©uS dx,
T3 T3
and nally

u © u u dx 0
T3
thanks to Lemma 4.8. This completes the proof.
4.2. PROPERTIES OF STRONG SOLUTIONS 29
Lemma 4.10. Let u be a weak solution of NSE and U > L2 0, T ; H 2 V a vector eld 9
with ∂t U > L 0, T ; L2 . Then U is a valid test function in the denition of weak solution
2
for u, that is,

ut U t dx u0 U 0 dx
T3 T3
t
u ∂t U ν ©u ©U u ©u U dx ds (4.3)
0 T3
for every t > 0, T .

Proof. We only give a sketch. As before (e.g. in the proof of Lemma 2.14), we
approximate U by elds in D̃σ by means of the projection operator PN onto the span of
the rst N Stokes eigenfunctions. For every thus obtained UN , (4.3) is valid. One then
takes the limit N ª. The only term requiring some care is
T3 ut U t dx, as we need
to converge pointwise in t. But by Proposition 4.4 (with n 0), we have
sup YUN UM tYḢ 1 B C YUN UM YL2 Ḣ 2 Y∂t UN UM YL2 L̇2 ,

t> 0,T
meaning that UN is Cauchy, and thus convergent, in C 0, T ; H 1 . The proof now
proceeds as in Proposition 4.4.
Theorem 4.11 (Weak-strong uniqueness). Let U be a strong solution and u a weak

solution that satises the energy inequality, and assume both solutions share the same initial
data u0 > H . Then u U almost everywhere.
Proof. The proof relies on an estimate of the relative energy between u and U , dened
as

1 2
Erel t Sux, t U x, tS dx.
2 T3
In the course of the computation, we use three ingredients:
(1) The weak formulation for u, tested with U, as justied by Lemma 4.10:

ut U t dx u0 U 0 dx
T3 T3
t
u ∂t U ν ©u ©U u ©u U dx ds;
0 T3
(2) The pointwise solution property of U, integrated against u:

T
∂t U U © U ν∆U u dx dt 0,
0 T3
as justied by Lemma 4.7;

(3) The energy (in)equalities for u and U, as justied by assumption and by Theo-
rem 4.9, respectively:
t
1 1
SutS
2
dx ν S©ux, τ S
2
dτ B SusS
2
ds
2 T3 s T3 2 T3
and similarly for U.

Using (1), we obtain

1 2
Erel t Sux, t U x, tS dx
2 T3

1 2 1 2
Sux, tS dx SU x, tS dx ut U t dx
2 T3 2 T3 T3

1 2 1 2
Sux, tS dx SU x, tS dx
2 T3 2 T3
t
u ∂t U ν ©u ©U u © u U dx ds u0 U 0 dx.
0 T3 T3
By, the energy inequalities, (3) and the assumption u0 U 0, the sum of the inital terms is
non-positive and can thus be neglected in the estimate, so that
t
Erel t B u ∂t U ν ©u ©U u © u U dx ds
0 T3
t t
2 2
ν S©uS dx ds ν S©U S dx ds.
0 T3 0 T3
Integrating by parts and using (2), we can write this as
t
Erel t B u ∂t U νu ∆U U © U u dx ds
0 T3
t
2
ν S©u ©U S dx ds R (4.4)
0 T3
t
2
ν S©u ©U S dx ds R,
0 T3
where
t
R U © U u u © u U dx ds.
0 T3
Similar arguments as in Lemma 4.8 yield

U © u u dx 0, u © U U dx 0,
T3 T3
whence
t
R U © U u u u © U u U dx ds,
0 T3

and nally an application of the formula
T3 U © U u U u dx 0 gives
t
R U u © U u U dx ds.
0 T3
We thus obtain the estimate
t
SRS B SU uSS©U © uSSU S dx ds
0 T3
t t (4.5)
Bν S©U ©
2
uS dx ds C ν 2
YU sYLª SU
2
uS dx ds,
0 T3 0 T3
a2 2
where we used the weighted Young's inequality ab B 2δ 2
δ 2 b2 for suitable δ, depending on
ν.
In total, using (4.5) in (4.4), we obtain
t
Erel t B 2
YU sYLª Erel s ds,
0
4.3. LOCAL EXISTENCE OF STRONG SOLUTIONS 31
and since U is a strong solution, YU YLª

2
> L1 0, T , and we may then use Grönwall's
inequality to conclude Erel 0, which implies the theorem.
Theorem 4.12 (Weak-strong stability). Let U be a strong solution and u a weak solu-
tion that satises the energy inequality, with initial data U 0 > H and u0 > H , respectively.
Then there exists a constant, depending only on the norm of U in L H 1 L2 H 2 and on ª
9
the viscosity ν , such that for all t > 0, T

Yut U tYL2 T3 B eCt Yu0 U 0 YL2 T3 .
Proof. The proof is almost identical to the one of the preceding theorem, only that
the initial terms do not cancel. It is left as an exercise. The reader might also want to give
an explicit formula for the constant C in terms of U and ν.
4.3. Local Existence of Strong Solutions

Lemma 4.13 (Agmon's inequality). Let u > Ḣ T , then, for some constant C ,
2 3
YuYLª B C YuY1H~12 YuY1H~22 .

Proof. We split the Fourier series of u into low and high frequencies, with M A0 to
be chosen later:
ux Q ûk eik x

kx0
Q ûk eik x Q ûk eik x

S k S BM Sk SAM
Q ûk eik x
Sk SSk S
1
Q ûk eik x Sk S2 Sk S 2
S k S BM S k S AM
1~2 1~2
1 1
B YuYḢ 1
SkSBM
Q Sk S2
YuYḢ 2 Q
SkSAM Sk S4
.
It is easy to see that
1~2 1~2
1
Q
SkSBM Sk S2
BC
1
SxS2
dx
BM 0B1 0
M 1~2
r2
C dr B CM 1~2 ,
1 r2
and similarly
1~2 1~2
1
Q
SkSAM Sk S4
BC
1
SxS4
dx
R3 BM 0
1~2
ª
r2
C dr B CM 1~2
,
M r4
so that in total
YuYLª B C M 1~2 YuYḢ 1 M 1~2

YuYḢ 2 .
YuYḢ 2
The choice M YuYḢ 1
now yields the result.
Theorem 4.14 (Existence of strong solutions). Let u0 > V . There exists a constant
C A 0, depending only on the viscosity ν , such that there exists a strong solution at least
on the interval 0, T , where T C Y u0 YL42 . ©

Proof. Recall the Galerkin equation (3.3),
∂t uN PN uN © uN ν∆uN ,

uN 0 PN u0 ,
for which we have seen to have global smooth solutions. Multiply this equation by ∆uN
and integrate in space to obtain

2
∂t uN ∆uN dx ν S∆uN S dx uN © uN ∆uN dx 0.
T3 T3 T3
This means

1 d 2 2
Y©uN YL2 ν Y∆uN YL2 dx uN © uN ∆uN dx, (4.6)
2 dt T3
and we wish to estimate the right-hand side. For this, we use Agmon's inequality and the
estimate YuN YH 2 B C Y∆uN YL2 (which is trivial to show in Fourier):

V uN © uN ∆uN dxV B YuN YLª Y©uN YL2 Y∆uN YL2
T3
1~2 1~2
B C YuYH 1 YuYH 2 Y©uN YL2 Y∆uN YL2
B C Y©uN Y3L~22 Y∆uN Y3L~22 .

ap bq
Young's inequality ab B p
1
q for p
1
q 1, applied here with p 4 and q 4
3 , gives

1
V uN © uN ∆uN dxV B C ν Y©uN Y6L2 ν Y∆uN Y2L2 ,
T3 2
so that (4.6) becomes
d
Y©uN YL2 ν Y∆uN YL2 B C ν Y©uN YL2 .
2 2 6
(4.7)
dt
Setting aside the Laplacian term for the moment, we see that Y Y©uN Y2L2 satises the
ordinary dierential inequality
Y t B CY t3 ,

Y 0 YPN ©u YL2
0 2
and hence Y t B X t for the solution of the corresponding equation X
CX 3 , X 0
Y©u0 Y2L2 C Y 0. It is not dicult to compute X t explicitly as
Y©u0 Y2L2
X t ¼ .
1 2CtY©u0 Y4L2
3
If we set T 8C Y©u0 Y4 2
, we obtain X T 2Y©u0 Y2L2 and therefore Y t is uniformly
L
bounded in 0, T , which in turn means that the uN are bounded in L ª
0, T ; V , uniformly
in N.
2
Coming back to (4.7) and integrating from 0 to T, we observe (recalling that Y©uN Y 2
L
0 2
is bounded, on 0, T , by 2Y©u Y 2 )
L
T T
ν 2
Y∆uN YL2 dt B 0 2
Y©PN u YL2
2
Y©uN T Y 2
L C Y©uN YL2
6
dt
0 0
T
B 0 2
Y©u YL2 C Y©uN YL2
6
dt
0
B Y©u0 Y2L2 2CT Y©u0 Y6L2
7 0 2
Y©u YL2
4
4.4. REGULARITY OF STRONG SOLUTIONS 33
by choice of T . It follows that the uN are also bounded, uniformly in N, in the space
L2 0, T ; H 2 T3 .
We know already that a subsequence of uN N >N converges to a weak solution of
NSE. Selecting from this sequence another subsequence that converges additionally in
L ª
0, T ; V 9L2 0, T ; H 2 T3 (this exists by the Banach-Alaoglu theorem, and the bounds
just derived), we see that the weak solution is in fact in L ª
0, T ; V 9 L2 0, T ; H 2 T3 ,
and is thus a strong solution up to time T.
4.4. Regularity of Strong Solutions

It turns out that even when only u0 > V , the strong solution of the NSE will automati-
smooth on the (open!) time interval 0, T , for any
ª
cally be C T up to which the solution
exists in the strong sense. This is an eect of parabolic regularisation.
First we need a crucial Banach algebra property:
Let s A 23 , then H s T3 embeds continuously into L

Theorem 4.15.
ª
T3 and forms
a Banach algebra, that is,
Yuv YH s B C YuYH s Yv YH s ,
where the constant depends only on s.
Proof. We give the proof only for u, v > Ḣ s , as this is the situation we shall need.
s
However the general statement follows similarly (by replacing Sk S by 1 Sk Ss ). So let
x>T 3
and using Hölder's inequality, we obtain
1~2 1~2
SuxS W Q ûk e ik x
W B Q Sûk S Q Sûk SSkS SkS s s
B Q Sûk S SkS 2 2s
Q SkS 2s
.
kx0 kx0 k x0 kx0 k x0
But the rst factor is precisely the homogeneous H s -norm of u, and for the second factor
we compute

Q SkS
ª
1
2s
BC dx C r2 2s
dr,
kx0 R3 B0 1 SxS2s 1
which is nite if and only if 2 2s @ 1, i.e. sA 3

2 . Thus we obtain the embedding assertion
as required.
Let now u, v > Ḣ s T3 , then by the above computation the Fourier series of u and v
are absolutely convergent, and thus we may form the Cauchy product to calculate

uxv x Q ûk eik x Q v̂j eij x Q Q ûk l v̂l eik x, (4.8)
kx0 j x0 k > Z3 l>Z3
Ã
so that u vk Pl>Z 3 ûk l v̂l
(the Fourier transform of the product becomes a convolution).
As another ingredient we recall the inequality Sk S
s
B C Sk lSs SlSs , where C depends
only on s A 0.
We can now estimate
Yuv YH s
2
Q 1
2s
Sk S SÃ
uv k S 2
kx0
RR RR2
2s RRR R
Q 1 Sk S RR Q û v̂ RRR
RR 3 kl l RRR
k>Z3 RRl>Z RR
RR RR2
RR RR
BC Q Q RR
RR 1 s s
Sk l S 1 Sl S ûk l v̂l R
R
RR
k>Z3 RRRl>Z3 RR
R
RR RR2 RR RR2
RR R R RR
B C Q Q R
RRR
s RR
1 Sk lS ûkl v̂l RR C
RR
RR
RR
R
Q Q
1 SlS ûkl v̂l RRR .
s
RR
k>Z3 RRRl>Z3 RR k>Z3 RRRl>Z3 RR
Let us dene the function 1 S©S ux

s
Pkx01 Sk Ss ûk eik x , then obviously
Y1 S©S uYL2

s
B C YuYH s ,
and similarly for v. By (4.8), then, we see that
Q 1 s
Sk lS ûkl v̂l

s
1 S©S uv k
,
and Q 1
s
SlS ûkl v̂l u1 S©Ss v k ,
,
l > Z3 l > Z3
so that (using Plancherel's Theorem and Hölders)
Yuv YH s
2
BC QS s
1 S©S uv k S
, 2
C Q S u1 ©
s
S S v k S
, 2
kx0 kx0
s 2 s 2
C Y1 S©S uv Y 2 Yu1 S©S v Y 2
L̇ L̇
BC 2 2
Ḣ
2 2
Yv YLª YuY s YuYLª Yv Y s
Ḣ
BC 2 2
Yv Y s YuY s ,
Ḣ Ḣ
where in the nal step we used the embedding
s ª
H `L .
Theorem 4.16 (Higher regularity). Let m C 2 and u0 > V H m T3 . Then the strong 9
solution of NSE with existence interval 0, T even satises

u>L ª
0, T ; H
m 2
9 L 0, T ; H
m 1
.
Proof. Let uN denote the smooth Galerkin approximations, as usual. The existence
proof for strong solutions yielded uniform (in N) bounds for
YuN YLª H 1 YuN YL2 H 2 .
By induction, we will deduce such an estimate for m instead of 1. So let us assume the
induction hypothesis
sup YuN YLª H m1 YuN YL2 H m .

N
Taking the Hm inner product of the Galerkin equation with uN and then using Theo-
rem 4.15, we obtain
1 d 2 2
YuN YH m ν Y©uN YH m uN © uN , uN H m
2 dt
B YuN ©uN YH m YuN YH m
B YuN YH m Y©uN YH m YuN YH m
1
B ν Y©uN Y2H m C ν YuN Y4H m .
2
Note that in the last step, we made use of the Cauchy-Schwarz inequality with (ab B
a2 C b2 ).
This results in
d
YuN YH m ν Y©uN YH m B C ν YuN YH m YuN YH m
2 2 2 2
g tYuN Y2H m , (4.9)
dt
where g C ν YuN Y2H m > L1 0, T uniformly in N, by virtue of the induction hypothesis.
Grönwall's inequality therefore yields
t
2
YuN tYH m B 0 2
Yu YH m exp g s ds .
0
Since the right hand side is nite and independent of N, we obtain the bound
sup YuN YLª H m @ ª.

N
4.4. REGULARITY OF STRONG SOLUTIONS 35
Going back to estimate (4.9) and integrating in time, we get (using the bound just derived)
T
2
YuN T YH m
0 2
Yu Y
Hm ν Y©uN Y
2
Hm B g t dt T YuN YLª H m ,
0
which entails
T
1
Y©uN YH m
2
B 0 2
Yu YH m g t dt T YuN YLª H m .
ν 0
As the right hand side is nite and independent of N, we obtain the desired bound
sup YuN YL2 H m1 @ ª.

N
The sequence uN converges to the strong solution u of NSE, but by the bounds just
obtained and the Banach-Alaoglu Theorem, it also converges weakly* in L H m and weakly
ª
2 m1
in L H . It follows that u is contained in these spaces, as claimed.
Theorem 4.17 (Space regularity). Let u be a strong solution of NSE on 0, T . Then,
for every 0 @ @ T and every m > N, u > C 0, T ; H m T3 .
Proof. By denition, u > L2 0, T ; H 2 , and therefore for almost every s1 > 0, ,
us1 > H 2 . Choosing such s1 and using us1 as initial data, and keeping in mind the
uniqueness of strong solutions, we obtain from Theorem 4.16 u > L s1 , T ; H , and so
2 3
we may choose an s2 > s1 , such that us2 > H . In this way we obtain a sequence
3
0 @ s1 @ s2 @ s3 @ . . . @ such that u > L sm , T ; H m 1 9 L2 sm , T ; H m 2 .

ª
Observe that this implies ∂t u > L , T ; H

2 m 1
, because
∂t u Pu © u ν∆u.
Clearly, the last term is in L2 , T ; H m 1

, but also the nonlinear one: Since u >L ª
, T ; H m1
and © u > L2 , T ; H m 1
, by the Banach Algebra property also u © u > L2 , T ; H m 1

(so this is actually better than required).
Thus, by virtue of Proposition 4.4, this implies
u > C , T ; H m
and so we are done.
Theorem 4.18 (Time regularity). Let u be a strong solution, A 0, and j, k > N. Then
∂tj u > L ª
, T ; H k T3 .
Proof. We proceed by induction over j.

We use once more
∂t u Pu © u ν∆u. (4.10)
Similarly as in the previous proof, we bound YPu © uYH k at each time by YuYH k YuYH k1 ,
and Y∆uYH k B YuYH k2 . But both these are bounded, uniformly in t, in the respective
Sobolev norms by virtue of Theorem 4.17.
For the induction step, dierentiate (4.10) j1 times with respect to t to obtain
j 1
∂tj u Q j i 1P∂tiu

© ∂ t
j 1i
u ν∆∂tj 1 u.
i 0
But by induction hypothesis, the rst j 1 time derivatives are in L ª
0, T ; H m for every
m, and so we can apply similar arguments as in the induction base to conclude.
Recalling that a function which is contained in Sobolev spaces of arbitrary order is in
fact smooth, we obtain:
Corollary 4.19. Let u be a strong solution of NSE on 0, T . Then, for every A 0,

u>C ª
T 3
, T .
4.5. Blowup and the Beale-Kato-Majda Criterion

It can not be excluded that a strong solution ceases to be strong after nite time. There
is a rich theory of possible blow-up scenarios, although it might turn out that blowup can
actually not happen (this is the Millennium Problem).
4.5.1. Vorticity. Let u be a strong solution on 0, T , then it is smooth (in space) for
any t > 0, T . Dene the curl operator, which acts on smooth vectorelds u > C T3 ; R3 ª
and yields another such vectoreld, by
curl ui ijk ∂j uk ,

where ijk is the Levi-Civita symbol, i.e. ijk 1 if i, j, k is an even permutation of
1, 2, 3, ijk 1 if i, j, k is an odd permutation, and ijk 0 otherwise. Note we applied
the summation convention.
It is easy to see that div curl u 0 for any choice of u, and also curl ©p 0 for any scalar
eld p. Taking the curl of NSE and denoting ω curl u (which is known as the vorticity of
the ow), we obtain
∂t ω curlu © u ν∆ω.

A short computation shows curlu © u u © ω ω © u, so we arrive at the vorticity
equation
∂t ω u © ω ν∆ω ω © u.
The right hand side is called the vortex stretching term ; in 2D it is not present, and the
1
vorticity equation is simply a linear transport-diusion equation that satises a maximum
p ª
principle in any L norm (including L ). This is another, very important way, to see why
the 2D NSE are so much better behaved than the 3D NSE.
Lemma 4.20. Let u > C ª

T3 ; T3 be divergence-free, then
Yω YL2 Y©uYL2 .
Proof. The computation goes like this:

2
Sω S ijk ∂j uk imn ∂m un

δjm δkn δjn δkm ∂j uk ∂m un

2
∂j uk ∂j uk ∂j uk ∂k uj ∂j uk ∂j uk S©uS ,

where we used ∂j uk ∂k uj 0 owing to an integration by parts and the divergence-free
property.
4.5.2. Blowup.
Definition 4.21. Let u0 > V
u be a corresponding Leray-Hopf solution of NSE. A
and
time T A 0 is called the blowup time for the solution if u is a strong solution on 0, T

for any A 0, but it is not a strong solution on 0, T .

A few remarks are in order. First, it is possible that no blowup occurs and hence no
(nite) blowup time exists this is trivially so e.g. for the zero solution. Secondly, if there
is a blowup, then the blowup time is uniquely determined by u0 : Indeed, as long as the
strong solution exists, it is unique in the class of Leray-Hopf solutions (Theorem 4.11).
1
Of course to solve this equation, the nonlocal coupling between u and ω needs to be taken into
account. The coupling law is known by the name of Biot-Savart and is a Fourier multiplier operator of
order 1.
4.5. BLOWUP AND THE BEALE-KATO-MAJDA CRITERION 37
is the smallest time at which Y©uYL2 T

A more substantial remark is that T ª.
Indeed, the proof of Theorem 4.14 shows that u > L 0, T ; H

2 2
as long as u>L
ª
0, T ; H 1 ,
so that a solution that exits in the former space will thereby also exit in the latter.
Theorem 4.22 (Beale-Kato-Majda). Let u0 > V and u be a corresponding Leray-Hopf

solution of the NSE. If T A 0 is such that
T
Yω YLª dt @ ª,
0
then u is a strong solution on 0, T .
Proof. We multiply the vorticity equation by ω and integrate to obtain

∂t ω ω u © ω ω ν Sω S2 dx ω © u ω dx.
d

The rst term equals
dt Sω S2 dx, the second one is zero by the usual computation involving
the divergence-free property, and the third is non-negative and can be dropped. Hence,
using Lemma 4.20,

d
Yω Y B dx B Yω YLª Y©uYL2 Yω YL2 B Yω YLª Yω Y2L2 ,
2
Sω SS©uSSω S
dt
and so by Grönwall's inequality,
T
2
Y©uT YL2 B 0 2
Yu YV exp Yω tYLª dt .
0
Since, by assumption, the right hand side is nite, then so is Y©uT YL2 , and following the
remark after Denition 4.21, we conclude.
CHAPTER 5
The Vanishing Viscosity Limit

So far we kept νA0 constant. It has become clear that virtually all estimates during
this course have crucially relied on ν A 0, and blow up as ν 0 . In fact, except for the
Beale-Kato-Majda criterion, all the results presented so far are false or, in case of existence
of weak solutions, completely unknown for ν 0, in which case the resulting system is
known as the Euler equations.
An obvious question that arises in the study of turbulent ows (for which a dimension-
less number proportional to ν 1
, the Reynolds number, is typically very large) is whether
the (Leray-Hopf ) solutions of the NSE converge, as ν 0 , to a solution of the Euler equa-
tions, if the latter exists. This is known as the viscosity limit problem. It turns out there
is a crucial dierence between the cases with and without physical boundaries.
5.1. The Periodic Case

We give here a particularly elegant way to handle the viscosity limit, due to P.-L. Li-
3
ons [ , Chapter 4.4].
5.1.1. Dissipative Solutions of the Euler Equations. We consider now the Euler
equations,
∂ t u u © u © p 0
div u 0,
1
2
whose energy
2 T3 SuS dx is formally conserved by a similar computation as for NSE.
Therefore the function space L ª
0, T ; H appears suitable for the study of solutions.
For the following formal computation, suppose u is a smooth solution with data u0 ,
and let U >C ª
T 3
0, T ; R 3
be any divergence-free eld. Denote
E U ∂t U PU © U ,
which in a sense measures how far U is from being a solution of Euler.

Then, subtracting the equations for u and U, we get
∂ t u © u U u U © U © π E U
for some scalar eld π. Next, multiply this by uU and integrate to obtain

1 d 2
Su U S dx u © u U u U dx
2 dt T3 3
T
u U © sym U u U dx E U u U dx,
T3 T3
1 t
where © sym 2 © © denotes the symmetric gradient. Note that the second integral
vanishes by the usual integration by parts argument, so we can estimate

1 d
Su U S
2
dx B Y©sym U YLª Su U S
2
dx E U u U dx.
2 dt T3 T3 T3
38
5.1. THE PERIODIC CASE 39
Grönwall's inequality then leads to

t
1
Su U St
2
dx B exp Y©sym U YLª ds Su
0
U 0S2 dx
2 0
t t
(5.1)
exp Y©sym U YLª dτ E U u U dx ds.

0 s
Recall from the exercises that C 0, T ; L2w is the space of functions in L 0, T ; L2
ª
that are weakly continuous in the sense that, for every t > 0, T , us @ ut weakly in
L2 , as s t.
Our formal computation motivates the following denition:
Definition 5.1 (dissipative solutions). Let C 0, T ; L2w , with u > L ª

0, T ; H 9
u0 u . Then u is a dissipative solution of Euler if (5.1) holds for every U > C 0, T ; H
0
such that E U > L 0, T ; L and ©sym U > L 0, T ; L .

1 2 1 ª
Remark 5.2. (1) The function spaces in this denition are chosen precisely such
that each term in (5.1) is well-dened.
(2) Choosing U 0, we obtain simply

1 1
SutS
2
dx B Su S
02
dx ¦t C 0, (5.2)
2 2
meaning that energy is not produced (but preserved or dissipated). This explains
the terminology.
(3) It can be shown that every solution in the sense of distributions that satises the
weak energy inequality (5.2) is also a dissipative solution in the sense of the given
denition. Conversely, there exist dissipative solutions that are not solutions in
the sense of distributions. As we shall see, however, dissipative solutions are
useful regardless of any ontological debates as to whether dissipative solutions are
really solutions of the Euler equations.
Proposition 5.3 (Weak-strong uniqueness). Suppose U > C 0, T ; H is such that

© sym U
ª
,
and is a solution of Euler in the sense that E U 0 almost every-
> L1 0, T ; L
where. Then any dissipative solution with u0 U 0 is equal to U almost everywhere.
Proof. This is a direct consequence of (5.1).
Note that dissipative solutions coincide, in particular, with the smooth solution as long
as the latter exists.
Lemma 5.4.Let u > L 0, T ; H C 0, T ; L2w , with u0 u0 . Then u is a dissipative

ª
9
solution of Euler if (5.1) holds for every smooth divergence-free U > C T3 0, T . ª

Proof. Let U be as in Denition 5.1. First we observe that then ∂t U > L1 0, T ; L2 .
Indeed,
∂t U PU © U E U .
By assumption, E U has the required regularity.

Moreover, a simple calculation yields
1 2
U © U 2©sym U U ©SU S ,
2
so that
PU © U 2P©sym U U .
But since the latter is the product of a matrix eld in L1 L ª
and a vector eld in L L2 ,
ª
we see that PU © U > L1 L2 .

40 5. THE VANISHING VISCOSITY LIMIT
1
Next, let η a standard mollier in x and set, as usual, η x 3
η x , so that U
U η is a smooth function in the space variables. Moreover, by the integrability of

∂t U , just shown, ∂t Uε > L1 0, T ; C k for every k > N, and by Proposition 4.1 this entails
Uε > C 0, T ; C k for every k > N.
Thus, we may compute
E U ∂t U PU © U
∂t U PU © U PU © U PU © U
(5.3)
E U 2P©sym U U 2P©sym U U
E U 2P ©sym U U ©sym U U .
On the one hand, E U E U in L1 L2 as 0. On the other hand, since ©sym U >

L1 L and U > L L2 , both ©sym U U and ©sym U U converge to ©sym U U in L1 L2 ,
ª ª
1 2
so the second expression in the last line of (5.3) converges to zero in L L . Hence, E U
1 2
converges to E U in L L .
Therefore, all the integrals in (5.1) converge appropriately as 0, so that (5.1) is
valid for U if it was valid for each U .
Time regularity can be guaranteed by regularising also in t, which poses little problem
since there is no nonlinearity of ∂t U .
5.1.2. The Viscosity Limit.

Theorem 5.5. Suppose U > C 0, T ; H is such that sym U > L1 0, T ; L , and is a ©
ª
solution of Euler in the sense that E U 0 almost everywhere. Let uν ν A0 be a family
of Leray-Hopf solutions, satisfying the weak energy inequality, with uν 0 U 0 for every
ν A 0. Then,
lim uν U strongly in L ª
0, T ; H .
ν 0
Proof. First we will show that a subsequence of uν converges weakly to a dissipative
solution of Euler. It suces to use smooth test elds, as shown in Lemma 5.4. So let
v > C c T
ª 3
0, T ; R 3
be divergence-free. Then, using v as a test eld in the denition
of weak solution of NSE, for every t > 0, T we have
t t
uν ∂t v dx ds uν © uν v dx ds
0 T3 0 T3
t
ν uν
© © v dx ds U 0 v 0 dx uν t v t dx.
0 T3 T3 T3
By denition of E v and the divergence-free property of both elds (see Lemma 4.8), we
thence get
t t
uν E v dx ds uν v © v uν v dx ds
0 T3 0 T3
t
ν uν
© © v dx ds U 0 v 0 dx uν t v t dx.
0 T3 T3 T3
We recall the energy inequality for uν ,

t
1 1
Suν tS
2
dx ν S©uν x, τ S
2
dτ B SU 0S
2
ds,
2 T3 0 T3 2 T3
5.2. BOUNDED DOMAINS 41
and further observe that

t
1 2 1 2
Sv tS dx Sv 0S dx v ∂t v dx ds
2 T3 2 T3 0 T3
t
1 2
Sv 0S dx v E v v ©v dx ds
2 T3 0 T3
t
1 2
Sv 0S dx v E v dx ds.
2 T3 0 T3
Using these ingredients (and simply dropping the H 1 -term in the energy inequality for uν ),
we can estimate

1 1
Suν t v tS
2
dx B SU 0 v 0S
2
dx
2 T3 2 T3
t
uν v © v uν v dx ds
0 T3
t t
ν ©uν © v dx ds E v uν v dx ds
0 T3 0 T3
and further
t
1 1
Suν t v tS
2
dx B SU 0 v 0S
2
dx Y©sym v Y Suν v S2 dx ds
2 T3 2 T3 0 T3
t t
Cν Y©uν YL2 ds E v uν v dx ds.
0 0 T3
But note that, by virtue of the energy inequality,
Y©uν YL1 L2 B C T Y©uν YL2 L2 B C T ν 1~2

YU 0YL2 ,
and hence Grönwall's inequality yields

t
1
Suν v S t dx B exp
2
Y©sym v YLª ds SU 0 v 0S
2
dx
2 0
t t
exp Y©sym v YLª dτ E v uν v C T ν 1~2 YU 0YL2 dx ds.
0 s
ª
As ν 0, the last term vanishes, and the uniform bounds on uν in L H give weak*-
convergence in that space, so that on the right hand side, uν can be replaced by the
weak* limit u. For the left hand side, one can show that uν even converges in the space
C 0, T ; L2w , and since the functional u ( Su v tS2 dx is weakly lower semicontinuous,

the left hand side can only decrease in the limit.
5.2. Bounded Domains

As soon as physical boundaries are involved, the viscosity limit gets much more dicult.
While the theory of the NSE as presented in these notes carries over to smooth bounded
domains in a rather straightforward way, the limit ν 0 behaves very dierently. The
reason is as follows: the NSE are usually equipped with Dirichlet boundary conditions (u
0 on ∂Ω), which in the context of uid mechanics are called no-slip boundary conditions.
Since the passage ν 0 formally turns a second-order system into one of rst order,
we cannot impose the same conditions on Euler. The most common choice are the slip
boundary conditions u n 0 on ∂Ω, where n denotes the outer unit normal. This change
of boundary conditions causes the formation of a boundary layer.
As an analogy, consider the 1D heat equation ν∆u on 0, 1 0, T with Dirichlet
∂t u
boundary condition and initial data u0 1. The boundary condition will instantaneously
lead the (smooth) solution to attain u0 u1 0, and for small times the solution will
42 5. THE VANISHING VISCOSITY LIMIT
be approximately 1 in the interior and decay to zero steeply in a neighbourhood of the

º
boundary points of size νt. The same is expected for the NSE.
Definition 5.6. Let Ω ` R3 be a smooth bounded domain, then u > C 0, T ; L2w Ω 9
2 1
L 0, T ; H0 Ω is said to be a weak solution of the NSE if it is weakly divergence-free and
satises
ª
ª

0 Ω
ª
0 Ω

0
ν © u ©φ dx dt u φ0 dx uT φT dx
0 Ω Ω Ω
for every divergence-free φ > C 1 Ω 0, T such that φ 0 on ∂Ω.

The last condition that the test function has to vanish on the boundary is decisive.
2 Let Ω ` R3 be a smooth bounded domain, and uν ν A0 a

Theorem 5.7 (Kato 1984 [ ]).
family of weak solutions of the NSE satisfying the energy inequality, with initial u0 . Suppose
there exists a smooth solution u of Euler with initial datum u0 . Then, uν u strongly in
L 0, T ; L2 Ω if and only if
ª
T
2
lim ν S©uν S dx dt 0, (5.4)
ν 0 0 Ων
where Ων x > Ω distx, ∂Ω @ ν .

Proof. We only give a proof sketch. The dicult direction is to show that (5.4)
implies the desired convergence. We have

2 2 2
Suν uS dx Suν S dx SuS dx 2 uν u dx
Ω Ω
Ω
Ω
(5.5)
B2 Su S
02
dx 2 uν u dx.
Ω Ω
For the last integral, we would like to use u as a test function in the weak formulation of
the NSE. However, this is not allowed since only u n 0 at the boundary, and it may well
be that the tangential component of u is non-zero.
Kato's idea is now to cuto u near the boundary. To this end, let η > C ª
0, ª
such that η 0 0, η x 1 for every x C 1, and η is monotone non-decreasing. Set
dx distx, ∂Ω and let
dx
ην x η .
ν
It is now tempting to use ην u as a test function, as this now satises the correct boundary
condition. However this function will, in general, not be divergence-free!
By Poincaré's Lemma, however, since div u 0, there exists a smooth potential φ Ω
3
R such that curl φ u and φ 0 at ∂Ω. Set
vν curl ην φ ,
then vν is divergence-free, is zero on the boundary, and agrees with u except on Ων .
Moreover one can show various estimates such as
Y©vν YL2 B Cν 1~2

. (5.6)
In particular, Yu vν YL2 o1 as ν 0, and therefore, (5.5) turns into

Suν uS2 dx B 2 Su S
02
dx 2 uν vν dx 01.
Ω Ω Ω
5.2. BOUNDED DOMAINS 43
Using vν as a test function and following a computation similar to the proof of Theo-
rem 4.11, we arrive at
t
2
Yuν t utYL2 B o 1 C Yuν uY2 Rν s ds,
0
where

Rν t uν © uν vν uν ν ©uν © vν dx.
Ω
t
It remains to show
0 Rν ds 0.
We consider only the second term, using (5.6):

V ν ©uν vν dxV B ν Y©uν YL2 Y©uYL2 ν Y©uν YL2 Ων Y©u ©vν YL2 Ων
©
Ω
B Cν Y©uν YL2 Cνν 1~2
Y©uν YL2 Ων .
The time integral of the second term goes to zero by assumption, and so does the integral
of the rst term by virtue of the energy inequality:
t t 1~2
ν Y©uν YL2 ds B CT ν 1~2
ν
2
Y©uν YL2 ds B CT v 1~2 0.
0 0

Bibliography
[1] P. Constantin and C. Foias. Navier-Stokes equations. University of Chicago Press, 1988.
[2] T. Kato. Remarks on zero viscosity limit for nonstationary Navier-Stokes ows with boundary. Seminar
on nonlinear partial dierential equations. Springer, New York, NY, 1984.
[3] P.-L. Lions. Mathematical topics in uid mechanics. Vol. 1. Incompressible models. Oxford Lecture
Series in Mathematics and its Applications, 3. Oxford Science Publications. The Clarendon Press,
Oxford University Press, New York, 1996.
[4] J. C. Robinson, J. L. Rodrigo, and W. Sadowski. The Three-Dimensional NavierStokes equations:
Classical theory. Cambridge Studies in Advanced Mathematics, Vol. 157. Cambridge University Press,
2016.
[5] R. Temam. Navier-Stokes equations. Theory and numerical analysis. Revised edition. Studies in Math-
ematics and its Applications, 2. North-Holland Publishing Co., Amsterdam-New York, 1979.
44

N Seem Il Finished

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

N Seem Il Finished

Uploaded by

Copyright:

Available Formats

Lecture Notes: Navier-Stokes Equations

Emil Wiedemann, Edited: Jack Skipper

Universität Ulm, Winter 2018/19

Chapter 2. Function Spaces and Weak Solutions 8

Chapter 3. Existence of Weak Solutions 17

Chapter 4. Strong Solutions 25

Chapter 5. The Vanishing Viscosity Limit 38

∂t ux, t  u © ux, t  ©px, t ν∆ux, t

1.1. Physical Derivation (Sketch)

Ẋ x, t uX x, t, t,

Ẍ x, t ∂t uX x, t, t  Ẋ x, t © uX x, t, t

where the second term represents the phenomenon known as advection.

ux  hej , t  ux, t.

∂t uX, t  u © uX, t  ©pX, t ν∆uX, t,

1.2. Elementary Mathematical Properties

∂t ux, t  u © ux, t  ©px, t ν∆ux, t

In total, after also integrating in time, we obtain

This suggests that u > L2t Hx1 Ω 9 Lt L2x Ω ª

u © ui uj ∂j ui ∂j uj ui  div u a ui ,

div ∂t u  div divu a u  div ©p µ∆ div u

This is a non-local operator: Even if f is compactly supported, ∆ 1f

Existence of weak solutions Uniqueness Regularity

Table 1. State of the Art for incompressible NSE

1.3. Related Models

∂t ρu  divρu a u  ©pρ div S©u

1.3.3. Non-Newtonian uids. To study non-Newtonian uids (like blood), we re-

Function Spaces and Weak Solutions

ux  2πk, t ux, t ¦k > Z3 .

ux Q ûk eik x.

and in particular u > L2 if and only if û > l2 , i.e.

2.1. Fourier Characterisation of Sobolev Spaces

It turns out (exercise!) that this is equivalent to the norm

(3) For s @ 0, we dene H  s

v x Q v̂k eikx

v, u Q v̂k ûk Q v̂k û k , 

2.2. Helmholtz Decomposition

Definition 2.3. We dene the space H H T3  as

since u > H T3 . Further, since e

φ x  Q  iαk eik x and hx Q wk eik x.

Q Sαk S2SkS2  Swk S

h1  h2   ©φ1  φ2  0 implies that Yh1  h2  ©φ1  ©φ2 Y2 0

Yh1  h2 Y2  Y©φ1  ©φ2 Y2 0

Lemma 2.8. P commutes with derivatives, i.e. P∂xj ∂xj P.

and so we are done.

2.3. The Stokes Operator

(3) wk  form an orthogonal basis of V .

Y Ymk cosk xYL2 T3  Ymk sink xYL2 T3  1,

Then mk cosk x 8 mk sink x ` H : Indeed,

 ∆mk cosk x mk Q ∂j2 cosk x  mk Sk S2 cosk x Sk S

ux Q ûk eik x 1

and we see that both ûk û k  and iûk  û

2.4. Weak Solutions

Dσ φ > Cc T3  0, ª

On the other hand, we have already derived the energy equality:

sup Yφt  φN tY2V sup Y dk twk  Y2V sup λk d2k t,

λ2k d2k t

∂t φ > Cc T3  0, ª ` Cc

sup YφN  φYL6 T3  B C sup YφN  φ YV 0

Theorem 2.15 (Lebesgue dierentiation theorem). Let Ω ` Rn be measurable and

Lemma 2.16. Let u be a weak (Leray-Hopf) solution of NSE. Then

ζ t 1 for t B 1 and ζ t 0 for t C 1,

Then, for every  A 0, set

∂t ux, t u © ux, t ©px, t ν∆ux, t

Ẋ x, t uX x, t, t,

Ẍ x, t ∂t uX x, t, t Ẋ x, t © uX x, t, t

ux hej , t ux, t.

∂t uX, t u © uX, t ©pX, t ν∆uX, t,

∂t ux, t u © ux, t ©px, t ν∆ux, t

This suggests that u > L2t Hx1 Ω 9 Lt L2x Ω ª

u © ui uj ∂j ui ∂j uj ui div u a ui ,

div ∂t u div divu a u div ©p µ∆ div u

∂t ρu divρu a u ©pρ div S©u

1.3.3. Non-Newtonian uids. To study non-Newtonian uids (like blood), we re-

ux 2πk, t ux, t ¦k > Z3 .

ux Q ûk eik x.

(3) For s @ 0, we dene H s

v x Q v̂k eikx

v, u Q v̂k ûk Q v̂k û k ,

Definition 2.3. We dene the space H H T3 as

since u > H T3 . Further, since e

φ x Q iαk eik x and hx Q wk eik x.

Q Sαk S2SkS2 Swk S

h1 h2 ©φ1 φ2 0 implies that Yh1 h2 ©φ1 ©φ2 Y2 0

Yh1 h2 Y2 Y©φ1 ©φ2 Y2 0

(3) wk form an orthogonal basis of V .

Y Ymk cosk xYL2 T3 Ymk sink xYL2 T3 1,

Then mk cosk x 8 mk sink x ` H : Indeed,

∆mk cosk x mk Q ∂j2 cosk x mk Sk S2 cosk x Sk S

ux Q ûk eik x 1

and we see that both ûk û k and iûk û

Dσ φ > Cc T3 0, ª

sup Yφt φN tY2V sup Y dk twk Y2V sup λk d2k t,

λ2k d2k t

∂t φ > Cc T3 0, ª ` Cc

sup YφN φYL6 T3 B C sup YφN φ YV 0

Theorem 2.15 (Lebesgue dierentiation theorem). Let Ω ` Rn be measurable and

ζ t 1 for t B 1 and ζ t 0 for t C 1,

Then, for every A 0, set

∂t uN PN uN © uN ν∆uN , (3.3)

uN © uN @ u © u!

uN x sinN x @0 but u2N x sin2 N x @~ 0,

for all s, s > 0, T .

Since Pk H is nite-dimensional, Y YH and Y YV are equivalent in Pk H , and so

C λj SuN s, wj L2 S2 ds,

YuN uM YLq 0,T ;H B YuN Pk uN YLq 0,T ;H YPk uN Pk uM YLq 0,T ;H

∂t uN PN uN © uN ν∆uN ,

By orthogonality of wk in L2 and the eigenfunction property, this gives

by wk and employing the initial condition uN x, 0 PN u0 x. 0

∂t uN , φ ν ∆uN , φ PN uN © uN , φ

Sν ∆uN , φS ν S©uN , ©φS B ν Y©uN YL2 YφYV ,

It follows from the denition of the dual/operator norm

© uNj,l @w in L2 0, ª; L2 T3 .

Let T be so large that suppdk ` 0, T for all k 1, . . . , m. By the Galerkin equation,

uN © uN u © u uN u © uN u © uN u.

max YutYL̇2 B C YuYL2 Ḣ 1 Y∂t uYL2 H 1 .

Proposition 4.4. Let n > N0 and suppose u > L2 0, T ; Ḣ n 2

max YutYḢ n1 B C YuYL2 Ḣ n2 Y∂t uYL2 Ḣ n .

sup YutYḢ n1 B C YuYL2 Ḣ n2 Y∂t uYL2 Ḣ n

follows, as before, by integrating rst from s to t and then by integrating in s from 0 to

Let ψ > C T3 be any

Note that the integral in fact is well-dened by the previous lemma.

Sbu, u, uS B Sbu, u, u bu , u, uS

But the rst term is estimated as

for every t > 0, T .

sup YUN UM tYḢ 1 B C YUN UM YL2 Ḣ 2 Y∂t UN UM YL2 L̇2 ,

as justied by Lemma 4.7;

the viscosity ν , such that for all t > 0, T

ux Q ûk eik x

Q ûk eik x Q ûk eik x

YuYLª B C M 1~2 YuYḢ 1 M 1~2

∂t uN PN uN © uN ν∆uN ,

Let s A 23 , then H s T3 embeds continuously into L

which is nite if and only if 2 2s @ 1, i.e. sA 3

Let us dene the function 1 S©S ux