Week 12 Notes, Math 8610, Tanveer

1. Navier-Stokes with Boundaries

We now turn to analysis of Navier-Stokes equations with solid bound-
aries on which no-slip condition boundary conditions will be imposed.
The notes in this part follow Temam’s book on Navier-Stokes equa-
tions.
We will assume that the domain Ω is an open set on one side of a
boundary Γ that is locally Lipschitz, i.e. locally in a neighborhood of a
point x ∈ Γ, we have a representation xj = θ (x1 , x2 , · · · xj−1 , xj+1 , · · · xn )
for some j, where θ is a Lipschitz function. It is known that such a
a set Ω is locally star-shaped, i.e. for each point xj ∈ Γ, there exists
an open neighborhood Oj such that there Oj′ = Ω ∩ Oj is star-shaped
with respect to one of its points. It is clear that if Γ is finite, then it
has a finite open cover {Oj }j∈J for a finite set J. We may arrange the
same to be true even when Γ is not finite.
Definition 1.1. Let D (Ω) and be the space of C ∞ functions with com-
pact support contained Ω.
Definition 1.2. Let V be the space of functions
(1.1) V = {u ∈ D (Ω) , ∇ · u = 0}
We define the closure of V in L2 (Ω) and H01 (Ω) to be H and V respec-
tively.

2. Steady Navier-Stokes equation with stationary

boundaries
First, we assume Ω to be a Lipschitz, bounded and open set in Rn .
We assume f ∈ L2 (Ω). We look for solution u : Rn → Rn , p : Rn → R
satisfying
(2.2) − ν∆u + (u · ∇)u + ∇p = f in Ω

(2.3) ∇ · u = 0 in Ω

(2.4) u = 0 on Γ
If u, p and f were smooth functions satisfying (2.2)-(2.4), then clearly
u ∈ V , and for any v ∈ V, inner product of (2.2) with v and integration
leads to
(2.5) ν (u, v)1 + b(u, u, v) = (f, v)0 ,
1
2

where recall that (, )m is the inner-product in the H m Hilbert space

and b is a trilinear functional of its arguments defined by
Z
(2.6) b(u, v, w) = ui vj,i wj dx
Ω

It is useful to identify the given f ∈ L2 with f ∈ V ′ the dual space of

V , such that the right side of (2.7) may be interpreted as hf, vi. We
then rewrite (2.7) as
(2.7) ν (u, v)1 + b(u, u, v) = hf, vi
Equation (2.7) provides the the weak formulation of steady Navier-
Stokes equation–we require that the solution u ∈ V satisfies (2.7) for
any v ∈ V (note V is the closure of V in H01 norm), where f ∈ V ′
is a given function. It may be shown that any such solution to (2.7)
satisfies (2.2) in a distributional sense.
3. Weak solutions to steady Navier Stokes
Our focus right now will be to prove solutions to (2.7) exist(1).
Lemma 3.1. The trilinear form b is defined and trilinear continuous
on H01 (Ω) × H01 (Ω) × H01 (Ω) for 2 ≤ n ≤ 4 and for bounded Ω.
Proof. If u, v, w ∈ H01 (Ω), and 2 < n ≤ 4,
(3.8) u ∈ L2n/(n−2) (Ω) , Dv ∈ L2 (Ω) , w ∈ Ln (Ω),
where we used Sobolev embedding theorems, definition of H01, the ob-
2n
servation that n ≤ n−2 for 2 ≤ n ≤ 4 and the fact that L2n/(n−2) (Ω) ⊂
n
L (Ω) for finite bounded domain Ω. By the Holder inequality,
Z 
(3.9)  ui vj,i wj dx ≤ c′ kui kL2n/(n−2) (Ω) kvj,i kL2 (Ω) kwj kLn (Ω)
 
Ω
It follows that
 
(3.10) b(u, v, w) ≤ c(n, Ω)kukH01 (Ω) kvkH01 (Ω) kwkH01 (Ω)
 

When n = 2, the same results (3.10) hold since

Z 
(3.11) u v w dx ≤ kui kL4 (Ω) kvj,k kL2 (Ω) kwkL4 (Ω) ,
 
 i j,i j
Ω
and for n = 2 and Sobolev inequality gives kgkL4 (Ω) ≤ c′ kgkH 1(Ω) .
Continuity of the trilinear form b in each argument follows immediately
from (3.10).

(1)It is to be noted that if Ω is not bounded then b as defined in (2.6) need not make sense.
In that case, we have to introduce auxiliary space for v other than V
 3

Corollary 3.2. For an open bounded set Ω and u, v, w ∈ V and 2 ≤

n ≤ 4, b is trilinear continuous form on V × V × V .
Definition 3.3. For u, v ∈ H01 (Ω), we define B(u, v) the linear con-
tinuous form on V defined by
(3.12) hB(u, v), wi = b(u, v, w) , u, v ∈ H01 (Ω) for any w ∈ V
Lemma 3.4. For any open bounded set Ω, for 2 ≤ n ≤ 4,
(3.13) b(u, v, v) = 0 , for any u ∈ V, v ∈ H01 (Ω)

(3.14) b(u, w, v) = −b(u, v, w) , for any u ∈ V, v, w ∈ H01 (Ω)

Proof. It is enough to prove these equalities for u ∈ V and v ∈ D(Ω)
since they are dense in the given spaces. Integration by parts gives
 
1
Z Z
(3.15) b(u, v, v) = uj vi,j vi dx = ∂xj uj vi vi dx = 0
Ω Ω 2
Now, if we replace v in (3.13) by v + w, we obtain
(3.16)
0 = b(u, v+w, v+w) = b(u, v, w)+b(u, v, v)+b(u, w, v)+b(u, w, w) = b(u, v, w)+b(u, w, v)
and (3.14) follows.

Before we prove a theorem on existence of steady solutions, we will

need the following preliminary lemma:
Lemma 3.5. Let X be a finite dimensional Hilbert sapce with inner-
product ( , ) and norm .. Let P be a continuous mapping from X to

itself such that
 
(3.17) (P (ξ), ξ) > 0 , for ξ  = k > 0
 

Then, there exists ξ ∈ X with |ξ| ≤ k such that P (ξ) = 0.

Proof. Suppose that P has no zero in the closed Ball B ⊂ X centered
at 0 with radius k. Then,
kP (ξ)
(3.18) S(ξ) = −  
P (ξ)

 
maps the B back to itself and is continuous in ξ. Brower fixed point
theorem implies that S has a fixed point in B, i.e. there exists ξ0 ∈ B
so that
kP (ξ0 )
(3.19) ξ0 = −  
P (ξ0 )

4
 
Clearly from above ξ0  = k. Inner-product with ξ0 and use of (3.17)
 
leads to
k (P (ξ0), ξ0 )
(3.20) (ξ0 , ξ0 ) = −   < 0,
P (ξ0 )


which is a contradiction. Hence we must have a zero of P (ξ) for |ξ| ≤ k.

Lemma 3.6. If u(m) → u in V weakly and in L2 (Ω) strongly, then

(3.21) b(u(m) , u(m) , v) → b(u, u, v) , for any v ∈ V
Proof. We know that
Z


(3.22) b(u(m) , u(m) , v) = −b u(m) , v, u(m) = − u(m),i u(m),j ∂xi vj dx
Ω

Therefore, we have
(3.23) Z Z



b(u(m) , u(m) , v)−b(u, u, v) = − u(m),i u(m),j − uj vj,i dx− u(m),i − ui uj vj,i dx
Ω Ω

Since kDvkL∞ (Ω) < ∞, and u(m) m is a bounded sequence in L2 (Ω),


application of Cauchy-Schwartz inequality completes the proof.

Theorem 3.1. (Existence of steady solution): Let Ω be a bounded set

in Rn for 2 ≤ n ≤ 4 and f ∈ H −1 (Ω). Then there exists at least one
weak solution u ∈ V to steady Navier Stokes equation (2.7).
Proof. We will take as given the fact that Stokes operator −P∆, where
P is the
∞ Hodge projection, has an orthornormal set of eigen functions
w (i) i=1 that form a complete set in V and that each w (i) is smooth in

Ω. Then, for each fixed integer m ≥ 1, we use a Galerkin approximation
m
X
(3.24) u(m) = ξi,m w (i)
i=1

We require choice of coefficients ξi,m , if there exists one, so that we

satisfy the Galerkin approximation to (2.7):
(3.25)
ν u(m) , w (k) + b(u(m) , u(m) , w (k)) = hf, w (k) i , for k = 1, 2, · · · , m

Equation (3.25) constitute a system of nonlinear equations for ξ1,m , ξ2,m , ...ξm,m .
We will now prove that this nonlinear system has a solution. Define X
to be the space spanned by w (1) , w (2) , .. w (m) and the inner product
 5

in X will be the inner product (, )1 induced by V , while P = Pm is

defined by
(3.26)
(Pm (u), v) = (Pm (u), v)1 = ν(u, v)1 +b(u, u, v)−hf, vi , for any u, v ∈ X
From properties of b in Lemma 3.4, it is clear that Pm is a continuous
mapping in X, and we have
(3.27) (Pm (u), u) = νkuk21 − kf kV ′ kuk1 = kuk1 (νkuk1 − kf kV ′ )
It follows that (Pm (u), u) > 0 for kuk = k > ν1 kf kV ′ . Using lemma
3.5, we know there exists a solution u(m) to (3.25). Now, we seek to
determine limit of m → ∞. If we multiply (3.25) by ξk,m and sum over
k = 1, 2, ·, m, it follows that
νku(m) k21 + b u(m) , u(m) , u(m) = hf, u(m) i


(3.28)
Since b(u, u, u) = 0, the above gives rise to the uniform estimate in m:
1
(3.29) ku(m) k1 ≤
kf kV ′
ν
Since the sequence u(m) remains bounded in V , by Banach-Alouglu
theorem, there exists some u ∈ V and a subsequence m′ → ∞ so that
(3.30) u(m′ ) → u , weakly in V,
and therefore strongly in L2 (Ω) since H01 (Ω) is compactly embedded in
L2 (Ω). Now, we take v = wj for any fixed j, From (3.25) for m ≥ j, it
follows that


(3.31) ν u(m) , v + b(u(m) , u(m) , v) = hf, vi
Using Lemma 3.6, on the subsequence u(m′ ) , it follows that
(3.32) ν (u, v) + b(u, u, v) = hf, vi
Since this is true for any v = wj , it is also true for a linear combination
of wj and by density for v ∈ V , and we have always have a weak
solution u ∈ V to the Navier-Stokes equation.

Theorem 3.2. Uniqueness for large ν (small Reynolds number) As-

sume 2 ≤ n ≤ 4 and domain Ω is Lipschitz and bounded. If ν is
sufficiently large or equivalently kf kV ′ is sufficiently small, then there
exists unique weak solution u ∈ V to (2.7).
Proof. In (2.7), we substitute v = u to obtain
1
(3.33) νkuk21 = hf ui ≤ kf kV ′ kuk1 , implying kuk1 ≤ kf kV ′
ν
6

Also, if u∗ , u∗∗ ∈ V are two different solutions to (2.7), then it follows

from subtraction that w = u∗ − u∗∗ satisfies
(3.34) ν (w, v)1 + b(w, u∗, v) + b(u∗ , w, v) = 0
Now choose v = w and use Lemma 3.1 and (3.33) to obtain
c(n, Ω)
(3.35) νkwk21 ≤ c(n, Ω)kwk21 ku∗ k1 ≤ kf kV ′ kwk21
ν
which gives rise to the only possibility kwk1 = 0 when
c(n, Ω)
(3.36) kf kV ′ < 1
ν2