ANALYSIS & PDE

Volume 9 No. 8 2016

M IKHAIL KOROBKOV AND TAI -P ENG T SAI

FORWARD SELF-SIMILAR SOLUTIONS

OF THE NAVIER–STOKES EQUATIONS IN THE HALF SPACE

msp
ANALYSIS AND PDE

msp
Vol. 9, No. 8, 2016
dx.doi.org/10.2140/apde.2016.9.1811

FORWARD SELF-SIMILAR SOLUTIONS

OF THE NAVIER–STOKES EQUATIONS IN THE HALF SPACE
M IKHAIL KOROBKOV AND TAI -P ENG T SAI

For the incompressible Navier–Stokes equations in the 3D half space, we show the existence of forward
self-similar solutions for arbitrarily large self-similar initial data.

1. Introduction

Let R+ 3 = {x = (x , x , x ) : x > 0} be a half space with boundary ∂ R3 = {x = (x , x , 0)}. Con-

1 2 3 3 + 1 2
3 × [0, ∞) → R3 and pressure
sider the 3D incompressible Navier–Stokes equations for velocity u : R+
3 × [0, ∞) → R,
p : R+
∂t u − 1u + (u · ∇)u + ∇p = 0, div u = 0, (1-1)
3 × [0, ∞), coupled with the boundary condition
in R+

u|∂ R+3 = 0, (1-2)

and the initial condition
u|t=0 = a, div a = 0, a|∂ R+3 = 0. (1-3)

The system (1-1) enjoys a scaling property: if u(x, t) is a solution, then so is

u (λ) (x, t) := λu(λx, λ2 t) (1-4)

for any λ > 0. We say that u(x, t) is self-similar (SS) if u = u (λ) for every λ > 0. In that case,
 
1 x
u(x, t) = √ U √ , (1-5)
2t 2t
where U (x) = u x, 12 . It is called discretely self-similar (DSS) if u = u (λ) for one particular λ > 1. To

get self-similar solutions u(x, t) we usually assume the initial data a(x) is also self-similar, i.e.,
a(x̂) x
a(x) = , x̂ = . (1-6)
|x| |x|
In view of the above, it is natural to look for solutions satisfying
C(C∗ )
|u(x, t)| ≤ or ku( · , t)k L 3,∞ ≤ C(C∗ ), (1-7)
|x|

MSC2010: 35Q30, 76D05.

Keywords: Forward self-similar solutions, Navier-Stokes equations, half space.

1811
1812 MIKHAIL KOROBKOV AND TAI-PENG TSAI

where C∗ is some norm of the initial data a. For 1 ≤ q, r ≤ ∞, we denote the Lorentz spaces by L q,r .
In such classes, with sufficiently small C∗ , the unique existence of mild solutions — solutions of the
integral equation version of (1-1)–(1-3) via a contraction mapping argument — has been obtained by
Giga and Miyakawa [1989] and refined by Kato [1992], Cannone, Meyer and Planchon [Cannone et al.
1994; Cannone and Planchon 1996], and Barraza [1996]. It is also obtained in the broader class BMO−1
in [Koch and Tataru 2001]. In the context of the half space (and smooth exterior domains), it follows
from [Yamazaki 2000]. As a consequence, if a(x) is SS or DSS with small norm C∗ and u(x, t) is a
corresponding solution satisfying (1-7) with small C(C∗ ), the uniqueness property ensures that u(x, t) is
also SS or DSS, because u (λ) is another solution with the same bound and same initial data a (λ) = a. For
large C∗ , mild solutions still make sense but there is no existence theory since perturbative methods like
the contraction mapping no longer work.
Alternatively, one may try to extend the concept of weak solutions (which requires u 0 ∈ L 2 (R3 )) to more
general initial data. One such theory is local-Leray solutions in L 2uloc , constructed by Lemarié-Rieusset
[2002]. However, there is no uniqueness theorem for them and hence the existence of large SS or DSS
solutions was unknown. Recently, Jia and Šverák [2014] constructed SS solutions for every SS u 0 which
is locally Hölder continuous. Their main tool is a local Hölder estimate for local-Leray solutions near
t = 0, assuming minimal control of the initial data in the large. This estimate enables them to prove
a priori estimates of SS solutions, and then to show their existence by the Leray–Schauder degree theorem.
This result is extended by Tsai [2014] to the existence of discretely self-similar solutions.
When the domain is the half space R+ 3 , however, there is so far no analogous theory of local-Leray

solutions. Hence the method of [Jia and Šverák 2014; Tsai 2014] is not applicable.
In this note, our goal is to construct SS solutions in the half space for arbitrary large data. By BCw we
denote bounded and weak-* continuous functions. Our main theorem is the following.
Theorem 1.1. Let  = R+ 3 and let A be the Stokes operator in  (see (5-5)–(5-7)). For any self-similar
1
vector field a ∈ Cloc (\{0}) satisfying div a = 0, a|∂ = 0, there is a smooth self-similar mild solution
u ∈ BCw ([0, ∞); L σ ()) of (1-1) with u(0) = a and
3,∞

ku(t) − e−t A ak L 2 () = Ct 1/4 , k∇(u(t) − e−t A a)k L 2 () = Ct −1/4 , ∀t > 0. (1-8)

Comments on Theorem 1.1:

(1) There is no restriction on the size of a.
(2) It is concerned only with existence. There is no assertion on uniqueness.
(3) Our approach also gives a second construction of large self-similar solutions in the whole space R3 ,
but for initial data more restrictive (C 1 ) than those of [Jia and Šverák 2014]. In fact, it would show
the existence of self-similar solutions in the cones

K α = {0 ≤ φ ≤ α} , for 0 < α ≤ π,

(in spherical coordinates), if one could verify Assumption 3.1 for e−A/2 a. We are able to verify it
only for α = π2 and α = π .
 FORWARD SELF-SIMILAR SOLUTIONS OF THE NAVIER–STOKES EQUATIONS IN THE HALF SPACE 1813

√
−1
(4) We have the uniform bound (1-7) for u 0 (t) = e−t A a and we show |u 0 (x, t)| . t + |x| in
Section 6. We expect u 0 (t) 6∈ L () for any q ≤ 3, and ku 0 (t)k L q → ∞ as t → 0+ for q > 3. The
q

difference v = u − u 0 is more localized: by interpolating (1-8), kv(t)k L q → 0 as t → 0+ for all

q ∈ [2, 3). Although kv(t)k L 3 () = C for t > 0, v(t) weakly converges to 0 in L 3 as t → 0+ , as
easily shown by approximating the test function by L 2 ∩ L 3/2 functions. Both u 0 (t) and v(t) belong
to L ∞ (R+ ; L 3,∞ (R+
3 )).

We now outline our proof. Unlike previous approaches based on the evolution equations, we directly
prove the existence of the profile U in (1-5). It is based on the a priori estimates for U using the classical
Leray–Schauder fixed point theorem and the Leray reductio ad absurdum argument (which has been
fruitfully applied in recent papers of Korobkov, Pileckas and Russo [Korobkov et al. 2013; 2014a; 2014b;
2015a; 2015b] on the boundary value problem of stationary Navier–Stokes equations). Specifically, the
profile U (x) satisfies the Leray equations
−1U − U − x · ∇U + (U · ∇)U + ∇ P = 0, div U = 0 (1-9)
3 with zero boundary condition and, in a suitable sense,
in R+

U (x) → U0 (x) := (e−A/2 a)(x) as |x| → ∞. (1-10)

System (1-9) was proposed by Leray [1934], with the opposite sign for U + x · ∇U , for the study of
√
√
singular backward self-similar solutions of (1-1) in R3 of the form u(x, t) = U x/ −2t / −2t. Their
triviality was first established in [Nečas et al. 1996] if U ∈ L 3 (R3 ), in particular if U ∈ H 1 (R3 ) as assumed
in [Leray 1934], and then extended in [Tsai 1998] to U ∈ L q (R3 ), 3 ≤ q ≤ ∞. In the forward case and in
the whole space setting, we have

|U0 (x)| ∼ |x|−1 , V (x) := U (x) − U0 (x), |V (x)| . |x|−2 for |x| > 1; (1-11)

see [Jia and Šverák 2014; Tsai 2014]. In the half space setting, it is not clear if one can show a pointwise
decay bound for V . We show, however, that V (x) is a priori bounded in H01 (R+ 3 ), and use this a priori
1
bound to construct a solution. Due to lack of compactness of H0 at spatial infinity, we use the invading
method introduced by Leray [1933]: we approximate  = R+ 3 by  =  ∩ B , k = 1, 2, 3, . . . , where
k k
Bk is an increasing sequence of concentric balls, construct solutions Vk in k of the difference equation
(3-3) with zero boundary condition, and extract a subsequence converging to a desired solution V in R+ 3.

Our proof is structured as follows. We first recall some properties for Euler flows in Section 2, and
then use it to show that the Vk are uniformly bounded in H01 (k ) in Section 3. In Section 4, we construct
Vk using the a priori bound and a linear version of the Leray–Schauder theorem, and extract a weak
limit V using the uniform bound. The arguments in Sections 2–4 are valid as long as one can show that
U0 = e−A /2 a, A being the Stokes operator in , satisfies certain decay properties to be specified in
Assumption 3.1. In Section 5 we show that, for  = R+ 3 and those initial data a considered in Theorem 1.1,

U0 indeed satisfies Assumption 3.1. We finally verify that u(x, t) defined by (1-5) satisfies the assertions
of Theorem 1.1 in Section 6.
1814 MIKHAIL KOROBKOV AND TAI-PENG TSAI

Because our existence proof does not use the evolution equation, we do not need the nonlinear version
of the Leray–Schauder theorem as in [Jia and Šverák 2014; Tsai 2014]. As a side benefit, we do not need
to check the small-large uniqueness (cf. [Tsai 2014, Lemma 4.1]).

2. Some properties of solutions to the Euler system

1,q
For q ≥ 1, denote by D 1,q () the set of functions f ∈ Wloc () such that k f k D 1,q () = k∇ f k L q () < ∞.
Recall, that by the Sobolev embedding theorem, if q < n then for any f ∈ D 1,q (Rn ) there exists a constant
c ∈ R such that f − c ∈ L p (Rn ) with p = nq/(n − q). In particular,

f ∈ D 1,2 (R3 ) ⇒ f − c ∈ L 6 (R3 ), f ∈ D 1,3/2 (R3 ) ⇒ f − c ∈ L 3 (R3 ). (2-1)

Further, denote by D01,2 () the closure of the set of all smooth functions having compact supports in 
with respect to the norm k · k D 1,2 () , and H () = {v ∈ D01,2 () : div v = 0}. In particular,

H () ,→ L 6 (). (2-2)

(Recall that by the Sobolev inequality, k f k L 6 (R3 ) ≤ Ck∇ f k L 2 (R3 ) holds for every function f ∈ Cc∞ (R3 )
having compact support in R3 ; see [Adams and Fournier 2003, Theorem 4.31].)
Assume that the following conditions are fulfilled:

(E) Let  be a domain in R3 with (possibly unbounded) connected Lipschitz boundary 0 = ∂, and the
functions v ∈ H () and p ∈ D 1,3/2 () ∩ L 3 () satisfy the Euler system

(v · ∇)v + ∇ p = 0 in ,



div v = 0 in , (2-3)

v = 0 on ∂.


The next statement was proved in [Kapitanskiı̆ and Piletskas 1983, Lemma 4] and in [Amick 1984,
Theorem 2.2]; see also [Amirat et al. 1999, Lemma 4].

Theorem 2.1. Let the conditions (E) be fulfilled. Then

∃ p̂0 ∈ R : p(x) ≡ p̂0 for H2 -almost all x ∈ ∂. (2-4)

Here and henceforth we denote by Hm the m-dimensional Hausdorff measure Hm (F) = limt→0+ Hm
t (F),
P∞ S∞
where Hm
t (F) = inf i=1 (diam Fi )m : diam F ≤ t, F ⊂
i F
i=1 i .

3. A priori bound for Leray equations

Recall that the profile U (x) in (1-5) satisfies Leray equations (1-9) with zero boundary condition and
U (x) → U0 (x) at spatial infinity. Decompose

U = U0 + V, U0 = e−A/2 a. (3-1)
 FORWARD SELF-SIMILAR SOLUTIONS OF THE NAVIER–STOKES EQUATIONS IN THE HALF SPACE 1815

Because a is self-similar, u 0 ( · , t) = e−t A a is also self-similar, i.e., u 0 (x, t) = λu 0 (λx, λ2 t) for all λ > 0.
Differentiating in λ and evaluating at λ = 1 and t = 12 , we get

0 = U0 + x · ∇U0 + ∂t u 0 x, 12 = U0 + x · ∇U0 + 1U0 − ∇ P0

(3-2)

for some scalar P0 . Thus, the difference V (x) satisfies

−1V − V − x · ∇V + ∇ P = F0 + F1 (V ), div V = 0 (3-3)

for some scalar P, where

F0 = −U0 · ∇U0 , (3-4)
F1 (V ) = −(U0 + V ) · ∇V − V · ∇U0 , (3-5)

and V vanishes at the boundary and the spatial infinity.

For a Sobolev function f ∈ W 1,2 (), set
Z 1/2
k f k H 1 () := |∇ f |2 + 12 | f |2 . (3-6)


Denote by H01 () the closure of the set of all smooth functions having compact supports in  with respect
to the norm k · k H 1 () , and
1
H0,σ () = { f ∈ H01 () : div f = 0}.

Note that H01 () = { f ∈ W 1,2 () : f |∂ = 0, k f k H 1 () < ∞} for bounded Lipschitz domains.
We assume the following.

Assumption 3.1 (boundary data at infinity). Let  be a domain in R3 . The vector field U0 :  → R3
satisfies div U0 = 0 and
kU0 k L 6 () < ∞, k∇U0 k L 2 () < ∞. (3-7)

Note that from Assumption 3.1 and (3-4) it follows, in particular, that
Z Z
F0 · η ≤ C, (η · ∇)U0 · η ≤ C (3-8)
 

for any η ∈ H0,σ 1

() with kηk H 1 () ≤ 1 (by virtue of the evident imbedding H0,σ 1
() ,→ L p for all
0,σ
p ∈ [2, 6]).
If it is valid in , it is also valid in any subdomain of  with the same constant C. We show in
Section 5 that for  = R+ 3 and a satisfying (5-1), U = e−A/2 a satisfies (5-3) and hence Assumption 3.1.
0
This is also true if  = R3 and a is self-similar, divergence free, and locally Hölder continuous.

Theorem 3.2 (a priori estimate for bounded domain). Let  be a bounded domain in R3 with connected
Lipschitz boundary ∂, and assume Assumption 3.1 for U0 . Then for any function V ∈ H01 () satisfying

−1V + ∇ P = λ(V + x · ∇V + F0 + F1 (V )), div V = 0 (3-9)

1816 MIKHAIL KOROBKOV AND TAI-PENG TSAI

for some λ ∈ [0, 1], we have the a priori bound

Z
kV k2H 1 () = |∇V |2 + 12 |V |2 ≤ C(U0 , ).




Remark. Note that C(U0 , ) is independent of λ ∈ [0, 1].

Proof. Let the assumptions of the theorem be fulfilled. Suppose that its assertion is not true. Then there
exists a sequence of numbers λk ∈ [0, 1] and functions Vk ∈ H01 () such that

−1Vk − λk Vk − λk x · ∇Vk + ∇ Pk = λk (F0 + F1 (Vk )), div Vk = 0, (3-10)

and moreover, Z
Jk2 := |∇Vk |2 → ∞. (3-11)


Multiplying (3-10) by Vk and integrating by parts in , we obtain the identity

λk
Z Z
Jk2 + |Vk |2 = λk (F0 − Vk · ∇U0 )Vk . (3-12)
2  

Consider the normalized sequence of functions

bk = 1 Vk ,
V P
bk = 1
Pk . (3-13)
Jk λk Jk2
Since Z
bk |2 ≡ 1,
|∇ V

bk , which converges weakly in W 1,2 () to some function
we could extract a subsequence, still denoted by V
V ∈ H01 (), and strongly in L 3 (). Also we could assume without loss of generality that λk → λ0 ∈ [0, 1].
Multiplying the identity (3-12) by 1/Jk2 and taking a limit as k → ∞, we have
λ0
Z Z Z
1+ |V |2 = −λ0 (V · ∇U0 )V = λ0 (V · ∇V )U0 . (3-14)
2   

In particular, λk is separated from zero for large k.

Multiplying (3-10) by 1/(λk Jk2 ), we see that the pairs (V
bk , P
bk ) satisfy the equation

bk = 1 1 1V bk + 1 F0 − U0 · ∇ V
 
V
bk · ∇ V
bk + ∇ P bk + V
bk + x · ∇ V bk − Vbk · ∇U0 . (3-15)
Jk λk Jk
Take an arbitrary function η ∈ Cc,σ
∞ (). Multiplying (3-15) by η, integrating by parts and taking a

limit, we obtain finally Z

(V · ∇V ) · η = 0. (3-16)


Since η ∈ Cc,σ
∞ () is arbitrary, we see that V is a weak solution to the Euler equation

(V · ∇)V + ∇ P = 0 in ,


div V = 0 in , (3-17)
V = 0 on ∂,

 FORWARD SELF-SIMILAR SOLUTIONS OF THE NAVIER–STOKES EQUATIONS IN THE HALF SPACE 1817

for some P ∈ D 1,3/2 () ∩ L 3 (). By Theorem 2.1, there exists a constant p̂0 ∈ R such that P(x) ≡ p̂0
on ∂. Of course, we can assume without loss of generality that p̂0 = 0, i.e., P(x) ≡ 0 on ∂. Then by
(3-14) and the first line of (3-17), we get
λ0
Z Z Z
1+ 2
|V | = −λ0 U0 · ∇ P = −λ0 div(P · U0 ) = 0.
2   
The obtained contradiction finishes the proof of the theorem. 
Theorem 3.3 (a priori bound for invading method). Let  = R+ 3 , and assume Assumption 3.1 for U .
0
Take a sequence of balls Bk = B(0, Rk ) ⊂ R3 with Rk → ∞, and consider half-balls k =  ∩ Bk . Then
for functions Vk ∈ H01 (k ) satisfying

−1Vk − Vk − x · ∇Vk + ∇ Pk = F0 + F1 (Vk ), div Vk = 0, (3-18)

we have the a priori bound Z

|∇Vk |2 + 21 |Vk |2 ≤ C(U0 ),


k

where the constant C(U0 ) is independent of k.

Proof. Let the assumptions of the theorem be fulfilled. Suppose that its assertion is not true. Then there
exists a sequence of domains k and a sequence of solutions Vk ∈ H01 (k ) of (3-18) such that
Z
Jk2 := kVk k2H 1 (k ) = |∇Vk |2 + 12 |Vk |2 → ∞.


(3-19)
k

Multiplying (3-18) by Vk and integrating by parts in k , we obtain the identity

Z
2
Jk = (F0 − Vk · ∇U0 )Vk . (3-20)
k

Consider the normalized sequence of functions

bk = 1 Vk ,
V bk = 1 Pk .
P (3-21)
Jk Jk2
Multiplying (3-18) by 1/Jk2 , we see that the pairs (V
bk , P
bk ) satisfy the equation

V
bk · ∇ V bk = 1 (1V
bk + ∇ P bk + Vbk + x · ∇ Vbk + F0 − U0 · ∇ V bk · ∇U0 ).
bk − V (3-22)
Jk
Since Z
bk |2 + 1 |V
b 2


|∇ V 2 k | ≡ 1,
k

bk , which converges weakly in W 1,2 () to some function

we could extract a subsequence, still denoted by V
V ∈ H01 (), and strongly in L 2 (E) for any E b .
Multiplying the identity (3-20) by 1/Jk2 and taking a limit as k → ∞, we have
Z
1 = (−V · ∇U0 )V. (3-23)

1818 MIKHAIL KOROBKOV AND TAI-PENG TSAI

Take an arbitrary function η ∈ Cc,σ

∞ (). Multiplying (3-22) by η, integrating by parts and taking a limit,

we obtain finally Z
(V · ∇V ) · η = 0. (3-24)


Since η ∞ ()
∈ Cc,σ is arbitrary, we see that V is a weak solution to the Euler equation

(V · ∇)V + ∇ P = 0 in ,


div V = 0 in , (3-25)
V = 0 on ∂,


with some P ∈ D 1,3/2 ()∩ L 3 (). More precisely, since V, ∇V ∈ L 2 (), we have P ∈ D 1,q () for every
q ∈ 1, 23 . Consequently, P ∈ L s () for each s ∈ 32 , 3 . In particular, P ∈ L 3 () and ∇ P ∈ L 9/8 ().
   

Furthermore,
Z Z ∞Z
d
4/3 2
|P(r ω)|4/3 dω dr


|P| = −R
S+
R R S1+ dr
Z Z 1/9 Z 8/9
. 1/3
|P| |∇ P| ≤ |P|3
|∇ P| 9/8
,
|x|>R |x|>R |x|>R

where S R+ = {x ∈  : |x| = R} is the corresponding half-sphere. Hence, we conclude that

Z
|P|4/3 → 0 as R → ∞. (3-26)
S+
R

Analogously, from the assumption U0 ∈ L 6 (), ∇U ∈ L 2 (), it is very easy to deduce that
Z
|U0 |4 → 0 as R → ∞. (3-27)
S+
R

On the other hand, by (3-23) and the first line of (3-25) we obtain
Z Z Z Z
1 = (V · ∇)V · U0 = − ∇ P · U0 = − lim div(P · U0 ) = − lim P(U0 · n) = 0, (3-28)
  R→∞  R R→∞ S+
R

where  R =  ∩ B(0, R) and the last equality follows from (3-26)–(3-27). The obtained contradiction
finishes the proof of the theorem. 

4. Existence for Leray equations

The proof of the existence theorem for the system of equations (3-3)–(3-5) in bounded domains  is
based on the following fundamental fact.
Theorem 4.1 (Leray–Schauder theorem). Let S : X → X be a continuous and compact mapping of a
Banach space X into itself , such that the set

{x ∈ X : x = λSx for some λ ∈ [0, 1]}

is bounded. Then S has a fixed point x∗ = Sx∗ .

 FORWARD SELF-SIMILAR SOLUTIONS OF THE NAVIER–STOKES EQUATIONS IN THE HALF SPACE 1819

Let  be a domain in R3 with connected Lipschitz boundary 0 = ∂, and set X = H0,σ 1
().
For functions V1 , V2 ∈ H0,σ (), write hV1 , V2 i H =  ∇V1 · ∇V2 . Then the system (3-3)–(3-5) is
1
R

equivalent to the following identities:

Z
hV, ζ i H = G(V ) · ζ, ∀ζ ∈ Cc,σ
∞
(), (4-1)

where G(V ) = V + x · ∇V + F(V ), F(V ) = F0 + F1 (V ),

F0 (x) = −U0 · ∇U0 , (4-2)

F1 (V ) = −(U0 + V ) · ∇V − V · ∇U0 . (4-3)
1
Since H0,σ () ,→ L 6 (), by the Riesz representation theorem, for any f ∈ L 6/5 () there exists a unique
mapping T ( f ) ∈ H0,σ
1
() such that
Z
hT ( f ), ζ i H = f · ζ, ∀ζ ∈ Cc,σ
∞
(), (4-4)

and moreover,
kT ( f )k H ≤ k f k X 0 ,
where Z
kfk X0 = sup f · ζ.
ζ ∈Cc,σ
∞ (), kζ k ≤1 
H

Then the system (3-3)–(3-5)∼(4-1) is equivalent to the equality

V = T (G(V )). (4-5)

Theorem 4.2 (compactness). If  is a bounded domain in R3 with connected Lipschitz boundary 0 = ∂,
1
and Assumption 3.1 holds for U0 , then for X = H0,σ () the operator S : X 3 V 7→ T (G(V )) ∈ X is
continuous and compact.
Proof. (i) For V, Ṽ ∈ X , setting v = Ṽ − V ,

F(Ṽ ) − F(V ) = −(U0 + V + v) · ∇v − v · ∇(U0 + V ).

Thus we have
kS(Ṽ ) − S(V )k X
. kvk L 2 + k∇vk L 2 + kF(Ṽ ) − F(V )k L 6/5
. kvk L 2 + k∇vk L 2 + kU0 k L 3 k∇vk L 2 + kV + vk L 3 k∇vk L 2 + k∇U0 k L 2 kvk L 3 + kvk L 3 k∇V k L 2
. (1 + kV k X + kvk X )kvk X . (4-6)

(ii) By the Sobolev theorems, we have the compact embedding X ,→ L r () for all r ∈ [1, 6). Thus if a
sequence Vk ∈ X is bounded in X , i.e., kVk k L 2 () + k∇Vk k L 2 () ≤ C, then we can extract a subsequence
Vkl which converges to some V ∈ X in L 3 () norm: kVkl − V k L 3 () → 0 as l → ∞. Then using the
condition Vkl ≡ V ≡ 0 on ∂ and integration by parts, it is easy to see that kF(Vkl ) − F(V )k X 0 → 0 and,
consequently, kG(Vkl ) − G(V )k X 0 → 0 as l → ∞. 
1820 MIKHAIL KOROBKOV AND TAI-PENG TSAI

Corollary 4.3 (existence in bounded domains). Let  be a bounded domain in R3 with connected Lipschitz
boundary ∂, and assume Assumption 3.1 for U0 . Then the system (3-3)–(3-5) has a solution V ∈ H0,σ
1
().
Proof. This is a direct consequence of Theorems 4.1–4.2 and 3.2. 
Theorem 4.4 (existence in unbounded domains). Let  = R+
3 , and assume Assumption 3.1 for U . Then
0
the system (3-3)–(3-5) has a solution V ∈ H0,σ ().
1

Proof. Take balls Bk = B(0, k) and consider the increasing sequence of domains k =  ∩ Bk from
1
Theorem 3.3. By Corollary 4.3 there exists a sequence of solutions Vk ∈ H0,σ (k ) of the system (3-3)–(3-5)
in k . By Theorem 3.3, the norms kVk k H 1 () are uniformly bounded, thus we can extract a subsequence
0,σ
Vkl such that the weak convergence Vkl * V in W 1,2 (0 ) holds for any bounded subdomain 0 ⊂ . It
is easy to check that the limit function V is a solution of the system (3-3)–(3-5) in . 

5. Boundary data at infinity in the half space

In this section we restrict ourselves to the half space  = R+ 3 with boundary 6 = ∂ R3 and study the
+
decay property of U0 = e−A/2 a. Our goal is to prove the following lemma, which ensures Assumption 3.1
under the conditions of Theorem 1.1.
Write x ∗ = (x 0, −x3 ) given x = (x 0, x3 ) ∈ R3 , and hzi = (1 + |z|2 )1/2 for z ∈ Rm .
Lemma 5.1. Suppose a is a vector field in  = R+
3 satisfying

1
a ∈ Cloc (\{0}; R3 ), div a = 0, a|∂ = 0,
(5-1)
a(x) = λa(λx), ∀ x ∈ , ∀λ > 0.

Let U0 = e−A/2 a, where A is the Stokes operator in . Then

|∇ k U0 (x)| ≤ ck [a]1 (1 + x3 )− min(1,k) (1 + |x|)−1 , ∀k ∈ Z+ = {0, 1, 2, . . . }, (5-2)

and, for any 0 < δ  1,

|∇U0 (x)| ≤ cδ [a]1 x3−δ hxi2δ−2 , (5-3)

where [a]m = supk≤m, |x|=1 |∇ k a(x)|.

m
If we further assume a ∈ Cloc , m ≥ 2, and ∂3k a|6 = 0 for k < m, then |∇ k U0 (x)| ≤ ck [a]m hx3 i−k hxi−1
for k ≤ m.
Estimates (5-2) and (5-3) imply, in particular,

U0 ∈ L 4 () ∩ L ∞ (), ∇U0 ∈ L 2 (), (5-4)

and hence Assumption 3.1 for U0 is satisfied.

Green tensor for the nonstationary Stokes system in the half space. Consider the nonstationary Stokes
3,
system in the half space R+

∂t v − 1v + ∇ p = 0, div v = 0, 3
for x ∈ R+ , t > 0, (5-5)
v|x3 =0 = 0, v|t=0 = a. (5-6)
 FORWARD SELF-SIMILAR SOLUTIONS OF THE NAVIER–STOKES EQUATIONS IN THE HALF SPACE 1821

In our notation,
v(t) = e−t A a. (5-7)

It is shown by Solonnikov [2003, §2] that, if a = ă satisfies

div ă = 0, ă3 |x3 =0 = 0, (5-8)

then Z
vi (x, t) = Ğ i j (x, y, t)ă j (y) dy (5-9)
3
R+
with

Ğ i j (x, y, t) = δi j 0(x − y, t) + G i∗j (x, y, t), (5-10)

∂ ∂
Z
G i∗j (x, y, t) = −δi j 0(x − y ∗, t) − 4(1 − δ j3 ) E(x − z)0(z − y ∗ , t) dz,
∂x j R2 ×[0,x 3 ] ∂ xi

where E(x) = 1/(4π|x|) and 0(x, t) = (4πt)−3/2 e−|x| /(4t) are the fundamental solutions of the Laplace
2

and heat equations in R3 . (A sign difference occurs since E(x) = −1/(4π|x|) in [Solonnikov 2003].)
Moreover, G i∗j satisfies the pointwise bound
√
−k √
−3−|k 0 |−|`0 | −cy 2 /t
|∂ts Dxk D `y G i∗j (x, y, t)| . t −s−`3 /2 t + x3 3 t + |x − y ∗ | e 3 (5-11)

for all s ∈ N = {0, 1, 2, . . .} and k, ` ∈ N3 [Solonnikov 2003, (2.38)].

Note that Ğ i j is not the Green tensor in the strict sense since it requires (5-8). There is no known
pointwise estimate for the Green tensor; cf. [Solonnikov 1964; Kang 2004].
We now estimate U0 = e−A/2 a for a satisfying (5-1). By (5-9) and (5-10),
Z Z
U0,i (x) = 0 x − y, 2 ai (y) dy +
1
G i∗j x, y, 12 a j (y) dy =: U1,i (x) + U2,i (x).



(5-12)
3 3
R+ R+

By (5-11), for k ∈ Z+ and using only |a(y)| . 1/|y 0 |,

Z
2 1
|∇ U2 (x)| .
k
(1 + x3 )−k (1 + x3 + |x 0 − y 0 |)−3 e−cy3 0 dy
R+3 |y |
Z
1
. (1 + x3 )−k (1 + x3 + |x 0 − y 0 |)−3 0 dy 0
R2 |y |
Z
1
= (1 + x3 )−k−2 (1 + |x̄ − z 0 |)−3 0 dz 0
R 2 |z |
. (1 + x3 )−k−2 (1 + |x̄|)−1
= (1 + x3 )−k−1 (1 + x3 + |x 0 |)−1 , (5-13)

where x̄ = x 0 /(1 + x3 ). To estimate U1 , fix a cut-off function ζ (x) ∈ Cc∞ (R3 ) with ζ (x) = 1 for |x| < 1.
We have
Z Z
∇ U1,i (x) =
k
0 x − y, 2 ∇ y (1 − ζ (y))ai (y) dy + ∇xk 0 x − y, 12 (ζ (y)ai (y)) dy, (5-14)
1

k


3 3
R+ R+
1822 MIKHAIL KOROBKOV AND TAI-PENG TSAI

using a|6 = 0. Hence, for k ≤ 1,

Z
2 /2 2 /4
|∇ U1 (x)| .
k
e−|x−y| hyi−1−k dy + e−x . hxi−1−k . (5-15)
R3

We can get the same estimate for k ≥ 2 if we assume ∇ k a is defined and has the same decay. On the
other hand, we can show |∇xk U1 (x)| . hxi−2 for k ≥ 2 if we place the extra derivatives on 0 in the first
integral of (5-14).
Combining (5-13) and (5-15), we get (5-2) and the last statement of Lemma 5.1.
Write
− = {x ∈  : 1 + x3 > |x 0 |}, + = {x ∈  : 1 + x3 ≤ |x 0 |}. (5-16)

By (5-13) and (5-15), we have shown (5-3) in − (with δ = 0).

It remains to show (5-3) in + .

Estimates using boundary layer integrals. Set ε j = 1 for j < 3 and ε3 = −1. Thus x ∗j = ε j x j . Let ā(x)
be an extension of a(x) to x ∈ R3 with

ā j (x) = ε j a j (x ∗ ), if x3 < 0.
3 and a| = 0, it follows that div ā = 0 in R3 . Let u(x, t) be the solution of the
Since div a = 0 in R+ 6
nonstationary Stokes system in R3 with initial data ā, given simply by
Z
u i (x, t) = 0(y, t) āi (x − y) dy.
R3

It follows that u i (x, t) = εi u i (x ∗, t). Thus

∂3 u i (x, t)|6 = 0, for i < 3; u 3 (x, t)|6 = 0. (5-17)

We have |∇ k a(y)| . |y|−1−k for k ≤ 1. By the same estimates leading to (5-15) for U1 , we have

∇xk u i x, 12 . hxi−1−min(1,k) , for k ≤ 2.

(5-18)

Thus u x, 21 satisfies (5-3).

Using the self-similarity condition

u(x, t) = λu(λx, λ2 t), ∀λ > 0, (5-19)

from (5-18) we get √
−1−m
,

|x| + t m ≤ 1,
|∇xm u i (x, t)| . −1/2 √
−2 (5-20)
t |x| + t , m = 2.
Now decompose
v = u − w.

Then w satisfies the nonstationary Stokes system in R+ 3 with zero force, zero initial data, and has boundary

value
w j (x, t)|x3 =0 = u j (x 0 , 0, t), if j < 3; w3 (x, t)|x3 =0 = 0. (5-21)
 FORWARD SELF-SIMILAR SOLUTIONS OF THE NAVIER–STOKES EQUATIONS IN THE HALF SPACE 1823

Using (5-21), it is given by the boundary layer integral

X Z tZ
wi (x, t) = K i j (x − z 0 , s)u j (z 0 , 0, t − s) dz 0 ds, (5-22)
j=1,2 0 6

where, for j < 3,

1
K i j (x, t) = −2δi j ∂3 0 − ∂ j Ci , (5-23)
Z π
yi − xi
Ci (x, t) = ∂3 0(y, t) dy (5-24)
6×[0,x3 ] |y − x|3
[Solonnikov 1964, pp. 40, 48]. (Note that the K i3 ( j = 3) have extra terms.) They satisfy for j < 3
√
−k √
−2−`
|∂tm Dx`0 ∂xk3 Ci (x, t)| ≤ ct −m−(1/2) x3 + t |x| + t (5-25)

[Solonnikov 1964, pp. 41, 48].

We now show (5-3) for w x, 12 in the region + : 1 + x3 ≤ |x 0 |.

For t = 21 and i, k ∈ {1, 2, 3},

X Z 1/2Z 1
∂xk wi x, 2 = −
1
∂k Ci (x − z 0 , s)∂z j u j z 0, 0, 12 − s dz 0 ds

j=1,2 0 6 π
Z 1/2Z
2∂k ∂3 0(x − z 0, s)u i z 0 , 0, 12 − s dz 0 ds


− 1i<3
0 6
= I1 + I2 . (5-26)

Above, we have integrated by parts in tangential directions x j in I1 .

By (5-20) and (5-25),
Z 1/2Z
√
−1 √
−2 0 q 1
−2 0
|I1 | . s −1/2 x3 + s |x − z 0 | + s |z | + 2 − s dz ds.
0 6

Fix 0 < ε ≤ Splitting 0, 12 as 0, 14 ∪ 41 , 21 , and making the change of variable s → 12 − s in

1
4, 2
1 1




2. ,
we get
Z 1/4Z
√
−2 0
|I1 | . x3−2ε s −1+ε |x 0 − z 0 | + x3 + s (|z | + 1)−2 dz 0 ds
0 6
Z 1/4Z √
−2 0
+ (x3 + 1)−1 (|x 0 − z 0 | + x3 + 1)−2 |z 0 | + s dz ds.
0 6

Integrating first in time and using, for 0 < b < ∞, 0 ≤ a < 1 < a + b, and 0 < N < ∞, that
Z 1
ds C
≤ a+b−1 , (5-27)
0 s (N + s) (N + 1)1−a
a b N
Z 1  
ds 1 2N + 2
≤ C min , log , (5-28)
0 s (N + s)
a 1−a N 1−a N
1824 MIKHAIL KOROBKOV AND TAI-PENG TSAI

where the constant C is independent of N , we get

Z
|I1 | . x3−2ε (|x 0 − z 0 | + x3 )−2+2ε (|x 0 − z 0 | + x3 + 1)−2ε (|z 0 | + 1)−2 dz 0
6
2|z 0 |2 + 2
Z  
1
+ (x3 + 1) (|x − z | + x3 + 1) min
−1 0 0 −2
, log dz 0 .
6 |z 0 |2 |z 0 |2

Dividing the integration domain into |z 0 | < 21 |x 0 |, 12 |x 0 | < |z 0 | < 2|x 0 |, and |z 0 | > 2|x 0 |, we get

|I1 | . x3−2ε hxi−2+δ , for x ∈ + (5-29)

for any 0 < δ  1. Taking ε = 21 δ and ε = 12 , we get

(1 + x3 )|I1 | . x3−δ hxi−2+2δ , for x ∈ + . (5-30)

To estimate I2 for i < 3 (note I2 = 0 if i = 3), we separate two cases. If k < 3, integration by parts
gives Z 1/2Z
2∂3 0(x − z 0 , s)∂zk u i z 0 , 0, 21 − s dz 0 ds.


I2 = −
0 6
2
Using ue−u ≤ C` (1 + u)−` for u > 0 and any ` > 0,

|x| −3
 
−2 x 3 −x 2 /4s √
−3
∂3 0(x, s) = cs √ e ≤ cs −2
1+ √ = cs −1/2 |x| + s . (5-31)
s s
Hence I2 can be estimated in the same way as I1 , and (5-30) is valid if I1 is replaced by I2 and k < 3.
When k = 3, by ∂t 0 = 10 and integration by parts,
Z 1/2Z  X 
∂ 2j − ∂t 0(x − z 0 , s)u i z 0 , 0, 21 − s dz 0 ds


I2 = 2
0 6 j<3
XZ 1/2Z
2∂ j 0(x − z 0 , s)∂z j u i z 0 , 0, 12 − s dz 0 ds


=
j<3 0 6
Z 1/2Z
20(x − z 0 , s)∂t u i z 0 , 0, 12 − s dz 0 ds


+
0 6
Z Z 
− z 0 , 21 − µ u i (z , 0, µ) dz −
0
20(x − z , µ)u i z , 0, − µ dz
0 0 1



− lim 20 x 2
µ→0+ 6 6
= I3 + I4 + lim (I5,µ + I6,µ ).
µ→0+

Here I3 can be estimated in the same way as I1 , and (5-30) is valid if I1 is replaced by I3 . For I4 , since
∂t u i = 1u i , by estimate (5-20) for ∇ 2 u,

|x − z 0 |2 −3/2 1
Z 1/2Z    −1/2  r −2
−3/2 0 1
|I4 | . s 1+ −s |z | + −s dz 0 ds. (5-32)
0 6 4s 2 2
 FORWARD SELF-SIMILAR SOLUTIONS OF THE NAVIER–STOKES EQUATIONS IN THE HALF SPACE 1825

We have a similar estimate as I1 with the following difference: we have to use the estimate (5-27) during
the integration over each subinterval s ∈ 0, 14 and s ∈ 14 , 21 ; for the second subinterval we apply (5-27)
   

with a = 12 , b = 1, N = |z 0 |2 .
For the boundary terms, the integrand of I5,µ is bounded by e−|x−z | /2 |z 0 |−1 and converges to 0 as
0 2

µ → 0+ for each z 0 ∈ 6. Thus lim I5,µ = 0 by the Lebesgue dominated convergence theorem. For I6,µ ,
Z
1 1
|I6,µ | . µ−1/2 e−x3 /(4µ) 0R2 (x 0 − z 0 , µ) 0 dz 0 . µ−1/2 e−x3 /(4µ) 0 ,
2 2
(5-33)
6 hz i hx i
which converges to 0 as µ → 0+ for any x ∈ .
We conclude that, for either k < 3 or k = 3, (5-30) is valid if I1 is replaced by I2 and hence, for any
0 < δ  1,
(1 + x3 ) ∂k wi x, 12 . x3−δ hxi−2+2δ , ∀ x ∈ + , ∀i, k ≤ 3.


(5-34)

Combining (5-18) and (5-34), we have shown (5-3) in + , concluding the proof of Lemma 5.1. 

6. Self-similar solutions in the half space

In this section we first complete the proof of Theorem 1.1, and then give a few comments.
Proof of Theorem 1.1. By Lemma 5.1, for those a satisfying the assumptions of Theorem 1.1, U0 = e−A/2 a
satisfies (5-2) and (5-3), and hence Assumption 3.1 is satisfied. By Theorem 4.4, there is a solution
V ∈ H0,σ1
(R+3 ) of the system (3-3)–(3-5).

Noting U0 ∈ C ∞ (R+ 3 ) by (5-2), the system (3-3)–(3-5) is a perturbation of the stationary Navier–

Stokes system with smooth coefficients. The regularity theory for the Navier–Stokes system implies that
∞ 3 3.


V ∈ Cloc R+ . The vector field U = U0 + V is thus a smooth solution of the Leray equations (1-9) in R+
√
√
The vector field u(x, t) defined by (1-5), u(x, t) = U x/ 2t / 2t, is thus smooth and self-similar.
Moreover, 1

x

v(x, t) = u(x, t) − e a = √ V √
−t A
2t 2t
satisfies

kv(t)k L q (R+3 ) = kV k L q (R+3 ) (2t)(3/2q)−(1/2) and k∇v(t)k L 2 (R+3 ) = k∇V k L 2 (R+3 ) (2t)−1/4 .

This finishes the proof of Theorem 1.1. 

√ √
Remark. Let u 0 (x, t) = (e−t A a)(x) = U0 x/ 2t / 2t. We have u 0 ( · , t) → a as t → 0+ in L 3,∞ (R+ 3 ).

Indeed, by (5-2), |U0 (x)| . hxi−1 ∈ L 3,∞ ∩L q , q > 3. We have ku 0 (t)k L q (R+3 ) = kU0 k L q (R+3 ) (2t)(3/2q)−(1/2) ,
which remains finite as t → 0+ only if q = (3, ∞), and
1 1 1
|u 0 (x, t)| . √ · √ =√ . (6-1)
t 1 + |x|/ t t + |x|
This is consistent with the whole space case  = R3 .
For the difference V (x), we only have its L q (R+
3 ) bounds, and not pointwise bounds as (1-11) in [Jia

and Šverák 2014; Tsai 2014].

1826 MIKHAIL KOROBKOV AND TAI-PENG TSAI

Acknowledgments

The authors are deeply indebted to K. Pileckas and R. Russo for valuable discussions. In particular, they
pointed out the way to simplify the proofs for a stronger result.
This work was initiated when Korobkov visited the Center for Advanced Study in Theoretical Sciences
(CASTS) of the National Taiwan University. The research of Korobkov was partially supported by the
Russian Foundation for Basic Research (projects 14-01-00768-a and 15-01-08275-a) and by the Dynasty
Foundation. The research of Tsai was partially supported by the Natural Sciences and Engineering
Research Council of Canada grant 261356-13.

References
[Adams and Fournier 2003] R. A. Adams and J. J. F. Fournier, Sobolev spaces, 2nd ed., Pure and Applied Mathematics 140,
Elsevier, Amsterdam, 2003. MR 2424078 Zbl 1098.46001
[Amick 1984] C. J. Amick, “Existence of solutions to the nonhomogeneous steady Navier–Stokes equations”, Indiana Univ.
Math. J. 33:6 (1984), 817–830. MR 763943 Zbl 0563.35059
[Amirat et al. 1999] Y. Amirat, D. Bresch, J. Lemoine, and J. Simon, “Existence of semi-periodic solutions of steady Navier–
Stokes equations in a half space with an exponential decay at infinity”, Rend. Sem. Mat. Univ. Padova 102 (1999), 341–365.
MR 1739546 Zbl 0940.35158
[Barraza 1996] O. A. Barraza, “Self-similar solutions in weak L p -spaces of the Navier–Stokes equations”, Rev. Mat. Iberoamer-
icana 12:2 (1996), 411–439. MR 1402672 Zbl 0860.35092
[Cannone and Planchon 1996] M. Cannone and F. Planchon, “Self-similar solutions for Navier–Stokes equations in R3 ”, Comm.
Partial Differential Equations 21:1-2 (1996), 179–193. MR 1373769 Zbl 0842.35075
[Cannone et al. 1994] M. Cannone, Y. Meyer, and F. Planchon, “Solutions auto-similaires des équations de Navier–Stokes”,
pp. VIII-1–10 in Séminaire: Équations aux dérivées partielles, 1993–1994, École Polytech., Palaiseau, 1994. MR 1300903
Zbl 0882.35090
[Giga and Miyakawa 1989] Y. Giga and T. Miyakawa, “Navier–Stokes flow in R3 with measures as initial vorticity and Morrey
spaces”, Comm. Partial Differential Equations 14:5 (1989), 577–618. MR 993821 Zbl 0681.35072
[Jia and Šverák 2014] H. Jia and V. Šverák, “Local-in-space estimates near initial time for weak solutions of the Navier–Stokes
equations and forward self-similar solutions”, Invent. Math. 196:1 (2014), 233–265. MR 3179576 Zbl 1301.35089
[Kang 2004] K. Kang, “On boundary regularity of the Navier–Stokes equations”, Comm. Partial Differential Equations 29:7-8
(2004), 955–987. MR 2097573 Zbl 1091.76012
[Kapitanskiı̆ and Piletskas 1983] L. V. Kapitanskiı̆ and K. I. Piletskas, “Spaces of solenoidal vector fields and boundary value
problems for the Navier–Stokes equations in domains with noncompact boundaries”, Trudy Mat. Inst. Steklov. 159 (1983), 5–36.
In Russian; translated in Proc. Math. Inst. Steklov 159 (1984), 3–34. MR 720205 Zbl 0528.76029
[Kato 1992] T. Kato, “Strong solutions of the Navier–Stokes equation in Morrey spaces”, Bol. Soc. Brasil. Mat. (N.S.) 22:2
(1992), 127–155. MR 1179482 Zbl 0781.35052
[Koch and Tataru 2001] H. Koch and D. Tataru, “Well-posedness for the Navier–Stokes equations”, Adv. Math. 157:1 (2001),
22–35. MR 1808843 Zbl 0972.35084
[Korobkov et al. 2013] M. V. Korobkov, K. Pileckas, and R. Russo, “On the flux problem in the theory of steady Navier–Stokes
equations with nonhomogeneous boundary conditions”, Arch. Ration. Mech. Anal. 207:1 (2013), 185–213. MR 3004771
Zbl 1260.35115
[Korobkov et al. 2014a] M. Korobkov, K. Pileckas, and R. Russo, “The existence of a solution with finite Dirichlet integral for
the steady Navier–Stokes equations in a plane exterior symmetric domain”, J. Math. Pures Appl. (9) 101:3 (2014), 257–274.
MR 3168911 Zbl 1331.35263
[Korobkov et al. 2014b] M. Korobkov, K. Pileckas, and R. Russo, “The existence theorem for the steady Navier—Stokes
problem in exterior axially symmetric 3D domains”, preprint, 2014. Submitted to Math. Ann. arXiv 1403.6921
 FORWARD SELF-SIMILAR SOLUTIONS OF THE NAVIER–STOKES EQUATIONS IN THE HALF SPACE 1827

[Korobkov et al. 2015a] M. Korobkov, K. Pileckas, and R. Russo, “An existence theorem for steady Navier–Stokes equations in
the axially symmetric case”, Ann. Sc. Norm. Super. Pisa Cl. Sci. (5) 14:1 (2015), 233–262. MR 3379492 Zbl 1321.35137
[Korobkov et al. 2015b] M. V. Korobkov, K. Pileckas, and R. Russo, “Solution of Leray’s problem for stationary Navier–
Stokes equations in plane and axially symmetric spatial domains”, Ann. of Math. (2) 181:2 (2015), 769–807. MR 3275850
Zbl 1318.35065
[Lemarié-Rieusset 2002] P. G. Lemarié-Rieusset, Recent developments in the Navier–Stokes problem, Chapman & Hall/CRC
Research Notes in Mathematics 431, Chapman & Hall/CRC, Boca Raton, FL, 2002. MR 1938147 Zbl 1034.35093
[Leray 1933] J. Leray, “Étude de diverses équations intégrales non linéaires et de quelques problèmes que pose l’hydrodynamique”,
J. Math. Pures Appl. 12 (1933), 1–82. MR 3533002 Zbl 0006.16702
[Leray 1934] J. Leray, “Sur le mouvement d’un liquide visqueux emplissant l’espace”, Acta Math. 63:1 (1934), 193–248.
MR 1555394 Zbl 59.0763.02
[Nečas et al. 1996] J. Nečas, M. Ružička, and V. Šverák, “On Leray’s self-similar solutions of the Navier–Stokes equations”,
Acta Math. 176:2 (1996), 283–294. MR 1397564 Zbl 0884.35115
[Solonnikov 1964] V. A. Solonnikov, “Estimates for solutions of a non-stationary linearized system of Navier–Stokes equations”,
Trudy Mat. Inst. Steklov. 70 (1964), 213–317. In Russian; translated in Amer. Math. Transl. (2) 75 (1968), 6–121. MR 0171094
[Solonnikov 2003] V. A. Solonnikov, “Estimates for solutions of the nonstationary Stokes problem in anisotropic Sobolev spaces
and estimates for the resolvent of the Stokes operator”, Uspekhi Mat. Nauk 58:2(350) (2003), 123–156. In Russian; translated
in Russian Math. Surveys 58 (2003), 331–365. MR 1992567 Zbl 1059.35101
[Tsai 1998] T.-P. Tsai, “On Leray’s self-similar solutions of the Navier–Stokes equations satisfying local energy estimates”,
Arch. Rational Mech. Anal. 143:1 (1998), 29–51. MR 1643650 Zbl 0916.35084
[Tsai 2014] T.-P. Tsai, “Forward discretely self-similar solutions of the Navier–Stokes equations”, Comm. Math. Phys. 328:1
(2014), 29–44. MR 3196979 Zbl 1293.35218
[Yamazaki 2000] M. Yamazaki, “The Navier–Stokes equations in the weak-L n space with time-dependent external force”, Math.
Ann. 317:4 (2000), 635–675. MR 1777114 Zbl 0965.35118

Received 16 Aug 2015. Revised 20 Jun 2016. Accepted 13 Sep 2016.

M IKHAIL KOROBKOV: [email protected]
Sobolev Institute of Mathematics, 4 Acad. Koptyug Avenue, 630090 Novosibirsk, Russia
and
Novosibirsk State University, 2 Pirogova Str., 630090 Novosibirsk, Russia
TAI -P ENG T SAI : [email protected]
Department of Mathematics, University of British Columbia, Vancouver, BC V6T 1Z2, Canada
and
Center for Advanced Study in Theoretical Sciences, National Taiwan University, Taipei, Taiwan

mathematical sciences publishers msp

 Analysis & PDE
msp.org/apde

EDITORS

E DITOR - IN -C HIEF
Patrick Gérard
[email protected]
Université Paris Sud XI
Orsay, France

B OARD OF E DITORS

Nicolas Burq Université Paris-Sud 11, France Werner Müller Universität Bonn, Germany
[email protected] [email protected]
Massimiliano Berti Scuola Intern. Sup. di Studi Avanzati, Italy Gilles Pisier Texas A&M University, and Paris 6
[email protected] [email protected]
Sun-Yung Alice Chang Princeton University, USA Tristan Rivière ETH, Switzerland
[email protected] [email protected]
Michael Christ University of California, Berkeley, USA Igor Rodnianski Princeton University, USA
[email protected] [email protected]
Charles Fefferman Princeton University, USA Wilhelm Schlag University of Chicago, USA
[email protected] [email protected]
Ursula Hamenstaedt Universität Bonn, Germany Sylvia Serfaty New York University, USA
[email protected] [email protected]
Vaughan Jones U.C. Berkeley & Vanderbilt University Yum-Tong Siu Harvard University, USA
[email protected] [email protected]
Vadim Kaloshin University of Maryland, USA Terence Tao University of California, Los Angeles, USA
[email protected] [email protected]
Herbert Koch Universität Bonn, Germany Michael E. Taylor Univ. of North Carolina, Chapel Hill, USA
[email protected] [email protected]
Izabella Laba University of British Columbia, Canada Gunther Uhlmann University of Washington, USA
[email protected] [email protected]
Gilles Lebeau Université de Nice Sophia Antipolis, France András Vasy Stanford University, USA
[email protected] [email protected]
Richard B. Melrose Massachussets Inst. of Tech., USA Dan Virgil Voiculescu University of California, Berkeley, USA
[email protected] [email protected]
Frank Merle Université de Cergy-Pontoise, France Steven Zelditch Northwestern University, USA
[email protected] [email protected]
William Minicozzi II Johns Hopkins University, USA Maciej Zworski University of California, Berkeley, USA
[email protected] [email protected]
Clément Mouhot Cambridge University, UK
[email protected]

PRODUCTION
[email protected]
Silvio Levy, Scientific Editor

See inside back cover or msp.org/apde for submission instructions.

The subscription price for 2016 is US $235/year for the electronic version, and $430/year (+$55, if shipping outside the US) for print and
electronic. Subscriptions, requests for back issues from the last three years and changes of subscribers address should be sent to MSP.
Analysis & PDE (ISSN 1948-206X electronic, 2157-5045 printed) at Mathematical Sciences Publishers, 798 Evans Hall #3840, c/o Uni-
versity of California, Berkeley, CA 94720-3840, is published continuously online. Periodical rate postage paid at Berkeley, CA 94704, and
additional mailing offices.
APDE peer review and production are managed by EditFlow® from MSP.
PUBLISHED BY
mathematical sciences publishers
nonprofit scientific publishing
http://msp.org/
© 2016 Mathematical Sciences Publishers
 ANALYSIS & PDE
Volume 9 No. 8 2016

Estimates for radial solutions of the homogeneous Landau equation with Coulomb potential 1773
M ARIA P IA G UALDANI and N ESTOR G UILLEN
Forward self-similar solutions of the Navier–Stokes equations in the half space 1811
M IKHAIL KOROBKOV and TAI -P ENG T SAI
Decay of solutions of Maxwell–Klein–Gordon equations with arbitrary Maxwell field 1829
S HIWU YANG
Invariant distributions and the geodesic ray transform 1903
G ABRIEL P. PATERNAIN and H ANMING Z HOU
Multiple vector-valued inequalities via the helicoidal method 1931
C RISTINA B ENEA and C AMIL M USCALU
Structure of modular invariant subalgebras in free Araki–Woods factors 1989
R ÉMI B OUTONNET and C YRIL H OUDAYER
Finite-time blowup for a supercritical defocusing nonlinear wave system 1999
T ERENCE TAO
A long C2 without holomorphic functions 2031
L UKA B OC T HALER and F RANC F ORSTNERI Č