Critical line of exponents, scattering theories for a weighted

gradient system of semilinear wave equations

Xianfa Song∗
arXiv:2401.00706v1 [math-ph] 1 Jan 2024

Department of Mathematics, School of Mathematics, Tianjin University,

Tianjin, 300072, P. R. China

January 2, 2024

Abstract
In this paper, we consider the following Cauchy problem of a weighted gradient
system of semilinear wave equations
(
utt − ∆u = λ|u|α |v|β+2 u, vtt − ∆v = µ|u|α+2 |v|β v, x ∈ Rd , t ∈ R,
u(x, 0) = u10 (x), ut (x, 0) = u20 (x), v(x, 0) = v10 (x), vt (x, 0) = v20 (x), x ∈ Rd .

Here d ≥ 3, λ, µ ∈ R, α, β ≥ 0, (u10 , u20 ) and (v10 , v20 ) belong to H 1 (Rd ) ⊕ L2 (Rd ) or

Ḣ 1 (Rd ) ⊕ L2 (Rd ) or Ḣ γ (Rd ) ⊕ H γ−1 (Rd ) for some γ > 1. Under certain assumptions, we
establish the local wellposedness of the H 1 ⊕ H 1 -solution, Ḣ 1 ⊕ Ḣ 1 -solution and Ḣ γ ⊕ Ḣ γ -
solution of the system with different types of initial data.
Keywords: Weighted(or essential) gradient system; Wave equation; Critical line of
exponents; Well-posedness; Scattering.
2010 MSC: 35Q55.
Contents.
1 Introduction.
2 Preliminaries.
3 Global wellpossedness results on (1.1) when d = 3 and α + β ≤ 2.
4 Regular results on (1.1) in some special cases.
5 Ḣ 1 ⊕ Ḣ 1 scattering theory for (1.1) in some special cases.
6 Ḣ sc ⊕ Ḣ sc scattering theory for (1.1) in the supercritical case.
References.

1 Introduction
In this paper, we consider the following Cauchy problem

utt − ∆u = λ|u|α |v|β+2 u, vtt − ∆v = µ|u|α+2 |v|β v, x ∈ Rd , t ∈ R,


(1.1)
u(x, 0) = u10 (x), ut (x, 0) = u20 (x), v(x, 0) = v10 (x), vt (x, 0) = v20 (x), x ∈ Rd .

Here d ≥ 2, λ, µ ∈ R, α, β ≥ 0, (u10 , u20 ) and (v10 , v20 ) belong to H 1 (Rd ) ⊕ L2 (Rd ) or Ḣ 1 (Rd ) ⊕
L2 (Rd ) or Ḣ γ (Rd ) ⊕ H γ−1(Rd ) for some γ > 1. Besides the local wellposedness of the solution,
∗ E-mail: [email protected] or [email protected]

1
we are concerned with the global existence, regularity, asymptotic behavior and scattering
theory for (1.1) in this paper.
A large amount of work had been devoted to the scattering theory for the following Cauchy
problem
utt − ∆u = ±|u|p u, x ∈ Rd , t ∈ R,

(1.2)
u(x, 0) = u10 (x), ut (x, 0) = u20 (x), x ∈ Rd .
4
It is well known that p = n−2 is said to be a critical exponent of the H 1 solution of (1.2).
The global existence and scattering theory in different energy space in subcritical case can be
seen in [12, 14, 19, 21, 45, 46, 52, 55, 58]), while parallel results on (1.2) in the critical case
can be seen in [2, 15, 16, 29, 33] and those on (1.2) in the super-critical case can be seen in
[7, 17, 30, 32]. For the existence and scattering of solutions with small data at lower regularity,
we can refer to [26, 38, 44, 62]. For almost–sure scattering for the radial energy–critical nonlinear
wave equation, we can refer to [6, 12]. Some authors dealt with the regularity and asymptotic
behavior for the wave equation with a critical nonlinearity(see [23, 38, 50, 51]).
There are also many lectures on scattering theory for Klein-Gordon equation, see [5, 18, 19,
43, 46, 47, 59]. Meanwhile, the asymptotic behavior for Klein-Gordon equation in Schwarzschild
metric can be seen in [1, 11].
Some results on the wave equations with other type of nonlinearities were established.
Global solutions to the wave equations with nonlinearity of exponential growth in the critical
Sobolev space was established in [41]. Unique global existence and asymptotic behavior for
solutions to wave equations with non-coercive nonlinearity was established in [42]. Wave equa-
tions with time-dependent dissipation was discussed in [63]. Random data Cauchy theory for
nonlinear wave equations was studied in [39].
To analyze the behavior for the solutions to wave equations, different type of estimates were
established(see [3, 8, 22, 40, 48, 49, 56, 57, 61]). We can refer to these books [9, 25, 36, 54, 60]
and see more information on wave equations.
About the results on the global solution and scattering theory for a system of wave equa-
tions, we can refer to [10, 28, 34, 35] and the references therein. Differing to the equations have
the forms of utt − ∆u = Hv (u, v) and vtt − ∆v = Hu (u, v) in these references above, the equa-
tions in (1.1) have their special structure as follows: There exist two constants A = µ(α + 2),
B = λ(β + 2) and a function F (u, v) = λµ|u|α+2 |v|β+2 such that

∂F (u, v) ∂F (u, v)
= Aλ|u|α |v|β+2 u, = Bµ|u|α+2 |v|β v, (1.3)
∂u ∂v
which will be called as a weighted(essential) gradient system of wave equations below. We
will show an interesting phenomenon on (1.1) below. There exists a critical line of exponents
α+β = 2 when d = 3 in the following sense: The system always has a unique bounded H 1 ⊕H 1 -
solution for any initial data (u10 , u20 ), (v10 , v20 ) ∈ H 1 (R3 ) ⊕ L2 (R3 ) if α + β ≤ 2, while it fails to
be well-posed in H 1 (R3 )⊕H 1 (R3 ) for some (α, β) satisfying α+β > 2. While when d = 4, we call
(α, β) = (0, 0) is the critical point of exponents in the following sense: The system always has a
unique bounded H 1 ⊕ H 1 -solution for any initial data (u10 , u20 ), (v10 , v20 ) ∈ H 1 (R4 ) ⊕ L2 (R4 )
if (α, β) = (0, 0), while it fails to be well-posed in H 1 (R4 ) ⊕ H 1 (R4 ) if α + β > 0. We can define
the weighted energy of (1.1) as follows.
Z
Ew (u, v) = A(k∇uk22 + kut k22 ) + B(k∇vk22 + kvt k22 ) + λµ |u|α+2 |v|β+2 dx. (1.4)
Rd

2
Based on the conservation of weighted energy, we first get the global existence result, which is
one of main purposes in this paper. Our second goal is to obtain the regularity results on (1.1).
Last, we will establish scattering theory for (1.1) in supercritical case.
The rest of this paper is organized as follows. In Section 2, we give some notations and
useful lemmas. In Section 3, we will consider the global wellposedness of (1.1) when d = 2 and
d = 3. In Section 4, we will get the regularity results on (1.1). In Section 5, we will eatablish
Ḣ 1 ⊕ Ḣ 1 scattering theory for (1.1) in some special cases. In the last section, we will establish
scattering theory for (1.1) in supercritical case.

2 Preliminaries
In this section, we give some notations and lemmas.
Definition 2.1. If there exist two constants A, B and a function F (u, v) such that
∂F (u, v) ∂F (u, v)
= Af1 (u, v), = Bf2 (u, v), (2.1)
∂u ∂v
then we call the system
utt − ∆u = f1 (u, v), vtt − ∆v = f2 (u, v) (2.2)
as a weighted(essential) gradient system of wave equations.
Let F (u, v) = λµ|u|α+2 |v|β+2 , A = µ(α + 2) and B = λ(β + 2). Then
∂F (u, v) ∂F (u, v)
= Af1 (u, v) = Aλ|u|α |v|β+2 u, = Bf2 (u, v) = Bµ|u|α+2 |v|β v, (2.3)
∂u ∂u
which means that the system utt −∆u = λ|u|α |v|β+2 u, vtt −∆v = µ|u|α+2 |v|β v is a weighted(essential)
gradient system of wave equations.
We denote by F Fourier transform, and we write v̂ = F v for any v ∈ J ′ (Rd ). Define the
function spaces
s
Hps = Hps (Rd ) = {f ∈ J ′ (Rd ) : F −1 [(1 + |ξ|2 ) 2 fˆ] ∈ Lp (Rd )} (2.4)

equipped with the norm kf kHps = kF −1 [(1 + |ξ|2 ) fˆ]kLp and

s
2

Ḣps = Ḣps (Rd ) = {f ∈ J ′ (Rd ) : F −1 [|ξ|s fˆ] ∈ Lp (Rd )}, (2.5)

equipped with the norm kf kḢps = kF −1 [|ξ|s fˆ]kLp

1
Define the operators ω = (−∆) 2 , K(t) = ω −1 sin(ωt) and K̇ = cos(ωt). Denote
 R t2 α β+2
 G1 (t1 , t2 ; u, v)(t) = Rt1 K(t − τ )(λ|u(τ )| |v(τ )|
 u(τ ))dτ,
 t2 α+2 β
G (t , t ; u, v)(t) = K(t − τ )(µ|u(τ )| |v(τ )| v(τ ))dτ,

 2 1 2

t1
F1 (t0 ; u, v)(t) = G1 (t0 , t; u, v)(t), F2 (t0 ; u, v)(t) = G2 (t0 , t; u, v)(t), (2.6)

F1 (t2 ; u, v) − F1 (t1 ; u, v) = G1 (t1 , t2 ; u, v),




F2 (t2 ; u, v) − F2 (t1 ; u, v) = G2 (t1 , t2 ; u, v).


We will construct a local(in t) solution to the integral equation

 Rt 1 1

 u(t) = K̇(t)u10 + K(t)u20 + 0 (−∆)− 2 sin[(−∆) 2 (t − τ )](λ|u|α |v|β+2 u)dτ,

 := K̇(t)u10 + K(t)u20 + F1 (t; u, v) = A1 (u10 , u20 ),
(2.7)
 v(t) = K̇(t)v10 + K(t)v20 + 0t (−∆)− 2 sin[(−∆) 2 (t − τ )](µ|u|α+2 |v|β v)dτ,
R 1 1


:= K̇(t)v10 + K(t)v20 + F2 (t; u, v) = A2 (v10 , v20 ).


3
 Let
2η(r) γ(r) δ(r) 1 1
= = = − . (2.8)
(d + 1) (d − 1) d 2 r

We will recall some properties for the operators ω and K(t) below(see [40, 49, 56]).
Lemma 2.1. (i) For any ψ ∈ C0∞ (Rd ) and t > 0, 1 < p ≤ 2 ≤ q < ∞,
1 1
kK(t)ψkLq (Rd ) ≤ ct−(d−1)( 2 − q ) kψk d−1 − d+1 ,p . (2.9)
Ḣ 2 q (Rd )

(ii) For any ϕ ∈ Ls (Rd ),

kK(t)ϕkLr (Rd ) ≤ C|t|1−δ(r)+δ(s) kϕkLs (Rd ) (2.10)

with 
0 ≤ δ(r) − δ(s) ≤ min{1 + γ(r), d(1 − γ(r))},
(2.11)
1 < s, r < ∞ if d = 2,
A fact on the estimates for the operator K(t) acting in homogeneous Besov spaces below(see
Lemma 2.2 in [19]).
Lemma 2.2. (i) For 2 ≤ r ≤ ∞ and r1 + r̄1 = 1,
ρ+2η(r)−1
kK(t)ϕ; Ḃrρ k ≤ C|t|−γ(r) kϕ; Ḃr̄ k, (2.12)

ρ+2η(r)−1
where kϕ; Ḃr̄ k is the norm of ϕ in Besov space.
(ii) For d ≥ 2, 0 ≤ γ(r) ≤ 1,

kK(t)ϕ; Ḃrρ k ≤ C|t|−µ kϕ; Ḃrρ11 k (2.13)

for all ρ, ρ1 , r1 , µ satisfying

0 ≤ 1 + µ = ρ + δ(r) − ρ1 − δ(r1 )
1 1
≤ [γ(r) − γ(r1 )][1 + ] ≤ 1 + γ(r). (2.14)
2 γ(r)

We will recall some results on harmonic analysis below.

Definition 2.3. A pair (q, r) of positive real numbers is said to be wave admissible if
2 ≤ q ≤ +∞, 2 ≤ r < +∞ and  
1 d−1 1 1
≤ − .
q 2 2 r
Lemma 2.4.(Strichartz estimate, [27]) Take two admissible pairs (q1 , r1 ) and (q2 , r2 ),
1
let q2+ q̄12 = 1, r12 + r̄12 = 1,

1 d d 1 d
+ = −s= + − 2. (2.15)
q1 r1 2 q̄2 r̄2
Then for any I ⊂ R,

k(u, ∂t u)kCt (I;Ḣ s ⊕Ḣ s−1 ) + kukLqt 1 Lrx1 (I×Rd )

≤ k(u(0), ∂t u(0))kḢ s ⊕Ḣ s−1 + k|∇|(∂t2 u − ∆u)kLq̄2 Lr̄x2 (I×Rd ) . (2.16)
t

4
 The next lemmas are about product rule and fractional chain rules, we can see [4, 31, 37]
for more details.
Lemma 2.5. Assume that s ∈ (0, 1], 1 < r, r1 , r2 , q1 , q2 < +∞ and satisfy 1r = r11 + q11 =
1 1
r2 + q2 . Then

k|∇|s (f g)kLrx . kf kLrx1 k|∇|s gkLqx1 + k|∇|s f kLrx2 kgkLqx2 .

1 1 1
Lemma 2.6. Assume that s ∈ (0, 1], 1 < q, q1 , q2 < +∞ and satisfy q = q1 + q2 ,
G ∈ C 1 (C). Then

k|∇|s G(u)kLqx . kG′ (u)kLqx1 k|∇|s ukLqx2 .

Lemma 2.7. Assume that G is a Hölder continuous function of order 0 < p < 1. Then
s s
k|∇|s G(u)kLqx . k|u|p− σ kLqx1 k|∇|σ uk σ sq2
Lx σ

for every 0 < s < p, 1 < q < ∞ and ps < σ < 1 satisfying 1q = q11 + q12 and (1 − pσ
s
)q1 > 1.
Lemma 2.8. Assume that G is a Hölder continuous function of order 0 < p ≤ 1, 0 < s <
σp < p, 1 < q, q1 , q2 , r1 , r2 , r3 < ∞ and satisfy
s
(1 − p)r1 > 1, (p −
)r2 > 1,
σ
1 1 1 1 1 1
= + = + + .
q q1 q2 r1 r2 r3
Then

k|∇|s [w · (G(u + v) − G(u))]kLqx

  σs
s
p− σ
. k|∇| σ
wkLqx1 kvkpLpq2 + kwkLr1 kvk s k|∇|σ uk sr3 + k|∇| vk σ sr3 .
x L(p− σ )r2 Lx σ Lx σ

3 Local wellposedness results on (1.1)

In this section, we will consider the global wellposedness of H 1 ⊕ H 1 solution to (1.1).

3.1 H 1 ⊕ H 1 global wellposedness results when d = 2 and d = 3

Let d = 2 or d = 3,

2η(r) γ(r)


 1 ≤ l, r, q, q1 ≤ ∞, (d+1) = (d−1) = δ(r)
d =
1
2 − r1 ,
1 < r < ∞ if d = 2, |γ(r)| ≤ 1 if d = 3,




(α+β+2)d
l ≤ min{1 + γ(r), d(1 − γ(r))}, (3.1)
 (α+β+2) 1



 q + q1 ≤ 1,
 η = 2 − (α + β + 2)( d + 1 ) > 0.

1 l q

And for any interval I containing 0, denote

X0 (I) = Lq (I, Ll ), X1 (I) = Lq1 (I, Lr ). (3.2)

5
 The following lemma shows some form of equivalence between the system of partial dif-
ferential equations (1.1) and the system of integral equations (2.7) and gives the uniqueness of
the solution to (2.7).
Lemma 3.1. Let d ≥ 2, I be a bounded interval containing 0, u10 ∈ X0 (I) ∩ X1 (I) and
v10 ∈ X0 (I) ∩ X1 (I). Then
(1) For any t0 ∈ I, F1 (t0 ; u, v) and F2 (t0 ; u, v) ∈ X1 (I) are continuous functions of t0
with values F1 (t0 ; u, v) ∈ X1 (I) and F2 (t0 ; u, v) ∈ X1 (I). And for any (u1 , v1 ), (u2 , v2 ) ∈
[X1 (I) ∩ X0 (I)]2 , there holds
kF1 (t0 ; u1 , v1 ) − F1 (t0 ; u2 , v2 ); X1 (I)k + kF2 (t0 ; u1 , v1 ) − F2 (t0 ; u2 , v2 ); X1 (I)k
≤ C[ku1 − u2 ; X1 (I)k + kv1 − v2 ; X1 (I)k]
2
( !)
X
2 η1 α+β+2 α+β+2
× |I| + |I| [kui ; X0 (I)k + kvi ; X0 (I)k ] . (3.3)
i=1

(2) For any t1 , t2 ∈ I, G1 (t1 , t2 ; u, v) and G2 (t1 , t2 ; u, v) are continuous functions of t1 , t2

and G1 (t1 , t2 ; u, v) ∈ X1loc (R), G2 (t1 , t2 ; u, v) ∈ X1loc (R). For any bounded interval J ⊃ I and
(u1 , v1 ), (u2 , v2 ) ∈ [X1 (I) ∩ X0 (I)]2 , there holds
kF1 (t1 , t2 ; u1 , v1 ) − F1 (t1 , t2 ; u2 , v2 ); X1 (J)k + kF2 (t1 , t2 ; u1 , v1 ) − F2 (t1 , t2 ; u2 , v2 ); X1 (J)k
≤ C[ku1 − u2 ; X1 ([t1 , t2 ])k + kv1 − v2 ; X1 ([t1 , t2 ])k]
2
( !)
X
2 η1 α+β+2 α+β+2
× |J| + |J| [kui ; X0 ([t1 , t2 ])k + kvi ; X0 ([t1 , t2 ])k ] . (3.4)
i=1

(3) For any t0 ∈ I, F1 (t0 ; u, v) and F2 (t0 ; u, v) satisfy F1 (t0 ; u, v) = λ|u|α |v|β+2 u and
F2 (t0 ; u, v) = µ|u|α+2 |v|β v in D′ (I ×Rd ) respectively, and for any t1 , t2 ∈ I, G1 (t1 , t2 ; u, v) =
0 and G2 (t1 , t2 ; u, v) = 0 in D′ (Rd+1 ).
(4) Let u10 , v10 ∈ X1loc (I). Then the system (2.7) has at most one solution in X1loc (I) ∩
X2loc (I).
Proof: (1) and (2)
By (2.10) and (2.11), using Hölder inequality, we have
kK(t − τ )[f1 (u1 (τ ), v1 (τ )) − f1 (u2 (τ ), v2 (τ ))]kr + kK(t − τ )[f2 (u1 (τ ), v1 (τ )) − f2 (u2 (τ ), v2 (τ ))]kr
≤ C|t − s|1−δ(r)+δ(s) (k[f1 (u1 (τ ), v1 (τ )) − f1 (u2 (τ ), v2 (τ ))]ks + k[f2 (u1 (τ ), v1 (τ )) − f2 (u2 (τ ), v2 (τ ))]ks )
≤ C|t − s|1−δ(r)+δ(s) [ku1 (τ ) − u2 (τ )kr + kv1 (τ ) − v2 (τ )kr ]
× {ku1(τ )kα+β+2
l + ku2 (τ )kα+β+2
l + kv1 (τ )kα+β+2
l + kv2 (τ )kα+β+2
l } (3.5)
with
(α + β + 2)d
δ(r) − δ(s) = . (3.6)
l
By the expressions of F1 (u, v), F2 (u, v), G1 (u, v) and G2 (u, v), using (3.5) and Young’s inequal-
ity for the integral, we can obtain (3.3) and (3.4).
(3) can be obtained by the standard duality argument and some elementary computations,
we omit the details here.
(4) Assume that (u1 , v1 ) and (u2 , v2 ) be two solutions of (2.7) with the same initial data
(u0 , v0 ). Then
u1 − u2 = F1 (u1 , v1 ) − F1 (u2 , v2 ), v1 − v2 = F2 (u1 , v1 ) − F2 (u2 , v2 ).

6
Replacing I in (3.3) by a sufficient small interval J containing 0 such that
2
( !)
2 η1
X
α+β+2 α+β+2 1
C |J| + |J| [kui ; X0 (J)k + kvi ; X0 (J)k ] ≤ ,
i=1
2

we can get
1
[ku1 − u2 ; X1 (J)k + kv1 − v2 ; X1 (J)k] ≤ [ku1 − u2 ; X1 (J)k + kv1 − v2 ; X1 (J)k],
2
which means that (u1 , v1 ) = (u2 , v2 ) in J. Iterating the process, we know that (u1 , v1 ) = (u2 , v2 )
everywhere in I. 
For any interval I ⊂ R and for suitable values of ρ, r and q, we will introduce the notations
X2 (I) = Lq (I, Ḃrρ ) and let Bi (I, R) be the closed ball of radius R in Xi (I), i = 1, 2. We have
the following estimate
Lemma 3.2. Let d ≥ 2, ρ and r satisfy q ≥ (α + β + 3) and 0 ≤ ρ < 1,

(d − 1) d
0 ≤ γ(r) ≤ (α + β + 2)( − ρ) ≤ 1 + γ(r),
, (3.7)
(d + 1) r
d 1
η2 = 2 − (α + β + 2)( − ρ + ) > 0. (3.8)
r q

Let I be a bounded open interval containing 0 and u, v ∈ X2 (I). Then

(1) F1 (u, v) ∈ X2 (I), F2 (u, v) ∈ X2 (I), and there holds

kF1 (u, v); X2 (I)k + kF2 (u, v); X2 (I)k

≤ C{|I|2 (ku; X2 (I)k + kv; X2 (I)k) + |I|η2 (ku; X2 (I)kα+β+3 + kv; X2 (I)kα+β+3 )}. (3.9)

(2) For any bounded interval J ⊃ I and for any t1 , t2 ∈ I, G1 (t1 , t2 ; u, v) and G2 (t1 , t2 ; u, v)
are continuous functions of t1 , t2 with values in X2 (J), and there holds

kG1 (t1 , t2 ; u, v); X2 (J)k + kG2 (t1 , t2 ; u, v); X2 (J)k

≤ C{|J|2 (ku; X2 ([t1 , t2 ])k + kv; X2 ([t1 , t2 ])k)
+ |J|η2 (ku; X2 ([t1 , t2 ])kα+β+3 + kv; X2 ([t1 , t2 ])kα+β+3 )}. (3.10)

Proof: First we prove that

θ θ
kf1 (u, v); Ḃl,m k + kf2 (u, v); Ḃl,m k
θ θ
≤ C(ku; Ḃk,m k + kv; Ḃk,m k)(k|u|α+β+2 ks + k|v|α+β+2 ks ) (3.11)

for 0 < θ < 1, 1 ≤ l ≤ k ≤ ∞, 1 ≤ m ≤ ∞ and 1s = 1l − k1 . Here Ḃl,mθ

was defined as that in
the Appendix of [18]
1
X
θ
Ḃl,m = {w ∈ L′ : { 2jmθ kϕj ∗ wkm
l }
m ≡ kw; Ḃ θ
l,m k < ∞}, (3.12)
j
α
L = {ϕ : D û(0) = 0 for any multindex α}. (3.13)

Here ϕ is the completion of S(Rd ), S(Rd ) is the set of all complex-valued rapidly decreasing
infinitely differentiable functions defined on the d-dimension real Euclidean space Rd .

7
 Let τy be the space translation by y ∈ Rd . Then
1
∂f1
Z
|τy f1 (u, v) − f1 (u, v)| ≤ |τy u − u| [| ((ϑ1 τy u + (1 − ϑ1 )u), τy v)|]dϑ1
0 ∂u
Z 1
∂f1
+ |τy v − v| [| (u, (ϑ2 τy v + (1 − ϑ2 )v))|]dϑ2 , (3.14)
0 ∂v
Z 1
∂f2
|τy f2 (u, v) − f2 (u, v)| ≤ |τy u − u| [| ((ϑ3 τy u + (1 − ϑ3 )u), τy v)|]dϑ3
0 ∂u
Z 1
∂f2
+ |τy v − v| [| (u, (ϑ4 τy v + (1 − ϑ4 )v))|]dϑ4 . (3.15)
0 ∂v

Applying Hölder’s inequality to (3.14) and (3.15), we can get

kτy f1 (u, v) − f1 (u, v)kl + kτy f2 (u, v) − f2 (u, v)kl

≤ C0 [kτy u − ukk + kτy v − vkk ][k|u|α+β+2 ks + k|v|α+β+2 ks ], (3.16)

which can deduce (3.11) immediately.

By (3.11) and the Sobolev inequality, we can obtain

kK(t − τ )f1 (u, v); Ḃrρ k + kK(t − τ )f2 (u, v); Ḃrρ k
′ ′
≤ C|t − τ |−µ [ku; Ḃrρ kα+β+3 + kv; Ḃrρ kα+β+3 ] (3.17)

for ρ′ ≤ ρ and
d d
(α + β + 3)( − ρ) = ′ − ρ′ = ρ + δ(r) − ρ′ − δ(r′ ) = 1 + µ.
r r
For given r in range (3.7), if (3.8) holds, then r′ satisfies (2.14) and (2.13) holds for these
ρ, r, ρ′ , r′ and µ. Applying (2.13) to f1 (u, v) and f2 (u, v), then estimating the time integral by
Young’s inequality, we can obtain (3.9) and (3.10). 
Now we will prove the local existence and uniqueness of the solution to (2.7).
Proposition 3.1. Let d ≥ 2, ρ, r, q and q1 satisfy 0 ≤ ρ < 1, 1 ≤ q ≤ q1 ≤ ∞, (3.7)
and (3.8) hold. Then for any R > 0, there exists T (R) > 0 such that for any u10 , v10 ∈
B2 (I, R) ∩ X1 (I) with I = [−T (R), T (R)], (2.7) possesses a solution (u, v) satisfying u, v ∈
B2 (I, 2R) ∩ X1 (I) and

[ku; X1 (I)k + kv; X1 (I)k] ≤ 2[ku10 ; X1 (I)k + kv10 ; X1 (I)k].

And it is unique in X2 (I) ∩ X1 (I).

Proof: Let dl = dr − ρ in Lemma 3.1 and Lemma 3.2 such that (3.1), (3.7) and (3.8) hold
with η1 = η2 = η. Meanwhile, using Sobolev embedding, we have

Ḃrρ ⊂ Ll , kukl ≤ C̄ku; Ḃrρ k, kvkl ≤ C̄kv; Ḃrρ k,

X2 (·) ⊂ X0 (·), ku; X0 k ≤ C̄ku; X2 k, ku; X0 k ≤ C̄ku; X2 k.

If we choose T = T (R) small enough such that

1
C{4T 2 + (2T )η (2R)α+β+2 (1 + 2C̄ α+β+2 )} ≤ , (3.18)
2

8
then apply the estimates of Lemma 3.1 and Lemma 3.2 to the operators A1 (u10 , u20 ) and
A2 (v10 , v20 ), we find that they make the set S = B2 (I, 2R) ∩ X1 (I) invariant and they are
contracting on S in the norm of X1 (I). Since B1 (I, R1 ) ∩ B2 (I, 2R) is w∗ -compact in X1 (I) ∩
X2 (I) for any R1 > 0, especially, it is compact in the w∗ -topology of X1 (I), therefore, it is
w∗ -closed and strongly closed in X1 (I) and consequently S is strongly closed in X1 (I). By the
contraction mapping theorem, we obtain the existence and uniqueness of a solution to (2.7) in
S. While the uniqueness result is the directly following from Lemma 3.1 (3). 
To establish the existence of the solution to (2.7) corresponding to initial data with finite
energy, we will recall some facts below.
Fact 1: Let d ≥ 2,

Xe = {(ϕ, ψ) : ϕ ∈ H 1 , ψ ∈ L2 } = H 1 ⊕ L2 , (3.19)

and (ϕ, ψ) ∈ Xe is associated a solution of the free equation

ϕ(t) = K̇(t)ϕ + K(t)ψ (3.20)

in L(R, H 1 ). Then for any (ϕ, ψ) ∈ Xe , ϕ ∈ X2 (R) and satisfies

kϕ; X2 (R)k ≤ C(k∇ϕk2 + kψk2 ). (3.21)

Here X2 (R) = Lq (R, Ḃrρ ) and ρ, r, q satisfy

(
0 ≤ δ(r) ≤ d2 , −1 ≤ σ = ρ + δ(r) − 1 < 21 ,
(3.22)
σ ≤ γ(r)
2 ,
1
q = max(0, σ).

Fact 2: Let d ≥ 2. There exist ρ, r and q such that 0 ≤ ρ < 1, (3.7), (3.8) and (3.22) hold.
By Fact 1, Fact 2, using Lemma 3.1 with dl = dr − ρ and Lemma 3.2, we can obtain the
results on the existence of the solution to (2.7) corresponding to initial data with finite energy
as follows.
Proposition 3.2. Let d ≥ 2. X1 and X2 correspond to values of ρ, r, q provided by Fact 2
and q1 ≥ q. Then
(1) For any (u10 , u20 ) ∈ Xe and (v10 , v20 ) ∈ Xe , there exists T > 0 depending only on
k(u10 , u20 ); Xe k and k(v10 , v20 ); Xe k such that (2.7) has a unique solution in X1 (I) ∩ X2 (I)
with I = [−T, T ].
(2) For any (u10 , u20 ) ∈ Xe and (v10 , v20 ) ∈ Xe , and any interval I, (2.7) has at most one
solution (u, v) with u, v ∈ X1loc (I) ∩ X2loc (I).
For (u, v) ∈ X1 (I) ∩ X2 (I), we define the weighted energy
Z
Ew (u, v) = µ(k∇uk22 + kut k22 ) + λ(k∇vk22 + kvt k22 ) + λµ |u|α+2 |v|β+2 dx. (3.23)
Rd

Choosing an even nonnegative function h1 ∈ L∞ (Rd ) with compact support and such that
kh1 k1 = 1, we can define hj (x) = j d h1 (jx),

f1j (u, v) = hj ∗ f1 (hj ∗ u, hj ∗ v), f2j (u, v) = hj ∗ f2 (hj ∗ u, hj ∗ v),

and
Z
Ejw (u, v) = µ(k∇uk22 + kut k22 ) + λ(k∇vk22 + kvt k22 ) + λµ |hj ∗ u|α+2 |hj ∗ v|β+2 dx. (3.24)
Rd

9
Consider the regularized system

u = hj ∗ u(0) + F1j (u, v), v = hj ∗ v (0) + F2j (u, v), (3.25)

where F1j and F2j are defined by (2.7) with f1 , f2 replaced by f1j , f2j .
Noticing the fact that the convolution with hj is a contraction in the spaces Lr and Ḃrρ , all
the estimates above hold with f replaced by fj , we know that Proposition 3.2, part (1) holds
for the system (3.25) with the same T , independently of j, and the solution (uj , vj ) to (3.25)
obtained in part (1) converges to the solution of (2.7) in X1 (I) when j → ∞. Moreover, the
following properties hold
(1) For any nonnegative integer k, (uj , u̇jt ), (vj , v̇jt ) ∈ L1 (I, H k+1 ⊕H k and (uj , vj ) satisfies
the system

uj = f1j (uj , vj ), vj = f2j (uj , vj ).

(2)

Ejw (uj (t), vj (t)) = Ejw (hj ∗ u0 , hj ∗ v0 ) ≡ Ejw (3.26)

and the estimates

kuj (t)k2 + kvj (t)k2 ≤ Ce(Ejw , t) ≤ Ce(Ēw , t), (3.27)

k∇uj (t)k22 + kujt (t)k22 + k∇vj (t)k22 + kvjt (t)k22 2
≤ C[ėt (Ejw , t)] ≤ C[ėt (Ēw , t)] 2
(3.28)

with

Ēw = sup Ejw < ∞ (3.29)

and
1
e(Ew , τ ) = ku0 k2 cosh(a|τ |) + (Ew + a2 ku0 k22 ) 2 a−1 sinh(a|τ |)
1
+ kv0 k2 cosh(a|τ |) + (Ew + a2 kv0 k22 ) 2 a−1 sinh(a|τ |). (3.30)

Particularly, (uj , u̇jt ) and (vj , v̇jt ) are locally bounded in H 1 ⊕ L2 uniformly in j.
The global existence and uniqueness results for finite energy solutions of (2.7) are based
on the following conservation of energy for these solutions with finite energy initial data.
Proposition 3.3. Let d ≥ 2, (u10 , u20 ), (v10 , v20 ) ∈ Xe and I be an open interval con-
taining 0. Let ρ, r and q satisfy 0 ≤ ρ < 1, (3.7), (3.8), (3.22) and q1 = ∞. Assume
that (u(0) , v (0) ) satisfies (3.20) and (u, v) is a solution of (2.7) in X1 (I) ∩ X2 (I). Then
(u, ut ), (v, vt ) ∈ L(I, H 1 ⊕ L2 and (u, v) satisfies the conservation of energy

Ew (u(t), v(t)) = Ew (u0 , v0 ) ≡ Ew (3.31)

and

ku(t)k2 + kv(t)k2 ≤ Ce(Ew , t), (3.32)

k∇u(t)k22 + kut (t)k22 + k∇v(t)k22 + kvt (t)k22 ≤ C[ė(Ew , t)] 2
(3.33)

for all t ∈ I.
Proof: We just need to prove the result in any bounded subinterval I ′ ⊂⊂ I containing 0.

10
 Let

R = sup [ku(0) + G1 (t, u, v); X2 (I ′ )k + kv (0) + G2 (t, u, v); X2 (I ′ )k]. (3.34)

s∈I ′

By (3.4), R is finite. Let T = T (R) be defined as in the proof of Proposition 2.1. By the results
of Proposition 2.1, for any t ∈ I ′ , solving (2.7) with initial time t by contraction, the solution
(u, v) can be recovered in the interval I ′ ∩ [t − T, t + T ]. If we cover I ′ by a finite number of
intervals Ik of length 2T and centers at tk = (1 − ǫ)kT for some ǫ > 0, then we can deduce the
result in I ′ by the corresponding one in Ik for successive values of k = 0, ±1, ±2,.... Therefore,
we only need to prove this proposition in the special case where I is a small interval containing
0 where (2.7) can be solved by the contraction method of Proposition 2.1 below.
Approximating (u, v) in I by the regularized solutions (uj , vj ) of (3.25), using (3.28), (3.30)
and standard compactness arguments, it follows that (uj , vj ) converges to (u, v) in the w∗ -sense
in L∞ (I, H 1 ) and (ujt , vjt ) converges to (ut , vt ) in the w∗ -sense in L∞ (I, L2 ). Moreover, (u, ut )
and (v, vt ) satisfy (3.32) and (3.33) for almost all t in I. Here ut and vt are the derivatives of
u and v in D′ (I, L2 ). Hence (u, v) ∈ C(I, Ls ) because u, v ∈ C(I, Lr ) ∩ L∞ (I, H 1 ) and ut , vt ∈
L∞ (I, L2 ). By Corollary 2.1, (u, v) satisfies (1.1) in D′ (I × Rd ) and utt , vtt ∈ L∞ (I, H −1 ).
Consequently, ut , vt ∈ C(I, H −1 ) and ut , vt ∈ Cw (I, L2 ). (3.32) and (3.33) hold for all t ∈ I by
the continuity properties of (u, ut ) and (v, vt ).
By uniform boundedness in H 1 and convergence in C(I, Lr ), using interpolation, we know
2d
that uj and vj respectively converge to u and v in C(I, Ls ) for all s satisfying 2 < s < d−2 .
Similar to (3.63) and (3.64) in [18], it is easy to show that uj and vj respectively converge to
u and v in C(I, L2 ), ujt and vjt respectively converge to ut and vt weakly in L2 for each t ∈ I.
Consequently, uj (t) and vj (t) respectively converge to u(t) and v(t) weakly in H 1 for each t ∈ I.
Letting the limit j → ∞ in (3.26), by the convergence of uj , vj to u, v in C(I, Ls ) for
2 ≤ s ≤ α + β + 4, we can directly show that
Z Z
α+2 β+2
|hj ∗ u| |hj ∗ v| dx → |u|α+2 |v|β+2 dx, Ejw (u(t), v(t)) → Ew (u0 , v0 ).
R3 R3

Since (uj (t), ujt (t)) and (vj (t), vjt (t)) converge to (u(t), ut (t)) and (v(t), vt (t)) in H 1 ⊕ L2 , we
get

Ew (u(t), v(t)) ≤ Ew (u0 , v0 ) for all t ∈ I. (3.35)

The time reversal invariance of (1.1) and the results of Proposition 2.1 imply (3.31) for all t ∈ I.
Since u, v ∈ C(I, Ls ) for 2 ≤ s ≤ α + β + 4, we know that ku(t)k22 + k∇u(t)k22 + kut (t)k22 and
kv(t)k22 + k∇v(t)k22 + kvt (t)k22 are continuous functions of time, and the weak continuity implies
strong continuity of (u, ut ) as well as (v, vt ) in H 1 ⊕ L2 as a function of time.
Similar to the arguments above, it is easy to show that (uj (t), ujt (t)) and (vj (t), vjt (t))
respectively converge to (u(t), ut (t)) and (v(t), vt (t)) in H 1 ⊕ L2 for each t ∈ I. 
Now we can establish the global existence and uniqueness result for finite energy solutions.
Theorem 3.1. Let d = 2 or d = 3, (u10 , u20 ) ∈ Xe and (v10 , v20 ) ∈ Xe . Then (2.7)
with (3.20) has a unique solution (u, v) such that (u, ut ) ∈ C(R, Xe ), (v, vt ) ∈ C(R, Xe ) and
(3.31)–(3.33) hold.
Suppose that ρ, r, q and q1 satisfy 0 ≤ ρ < 1, q1 ≥ q, (3.7), (3.8) and (3.22). Then the
solution (u, v) is unique and u, v ∈ X1loc (R) ∩ X2loc (R).
Proof: Under the assumptions on ρ, r, q and q1 , by Proposition 3.2, (2.7) with initial
data (u10 , u20 ), (v10 , v20 ) ∈ Xe can be solved locally in time with u(t), v(t) ∈ X1 (I) ∩ X2 (I).

11
Here the length of the interval I depends on the norms of (u10 , u20 ) and (v10 , v20 ) in Xe .
By Proposition 3.3, the local solutions thereby obtained satisfy u(t), v(t) ∈ C(I, Xe ), (3.31),
(3.32) and (3.33), which imply the global existence of the solution (u(t), v(t)) with u(t), v(t) ∈
X1loc (R) ∩ X2loc (R) ∩ C(R, Xe ).
Note that a function in L∞loc (R, Xe ) belongs to X1loc (R) ∩ X2loc (R). We obtain the unique-
ness of the solution to (2.7) with u, v in C(R, Xe ), actually in L∞loc (R, Xe ). 

4 Regular results on (1.1) when d = 3, (α, β) = (0, 2) and

when d = 4, (α, β) = (0, 0)
In this section, we discuss the regular results on (1.1) when d = 3, (α, β) = (0, 2) and when
d = 4, (α, β) = (0, 0). First, we give some notations which were used in [50, 58]. Let z = (x, t)
denote the point in space-time Rd × R.

K(z0 ) = {z = (x, t)||x − x0 | ≤ t0 − t}

denote the backward light cone with vertex at z0 = (x0 , t0 ) ∈ Rd × R,

M (z0 ) = {z = (x, t)||x − x0 | = t0 − t}

its mantle, and

D(t; z0 ) = {z = (x, t) ∈ K(z0 )}

its spacelike sections for t fixed. For S < T and Q ⊂ Rd × R, let

QTS = {z = (x, t) ∈ Q|S ≤ t ≤ T }

be the truncated region, while

∂KST = D(S) ∪ D(T ) ∪ MST .

Given a vector-value function (u, v) on a cone K(z0 ), let

1
ew (u, v) = [µ(α + 2)(|ut |2 + |∇u|2 ) + λ(β + 2)(|vt |2 + |∇v|2 )] + λµ|u|α+2 |v|β+2
2
be its weighted energy density and
Z
Ew (u, v, D(t, z0 )) = ew (u, v)dx
D(t;z0 )

be its local weighted energy in a ball,

1 y y
dz0 (u, v) = [µ(α + 2)| ut − ∇u|2 + λ(β + 2)| vt − ∇v|2 ] + λµ|u|α+2 |v|β+2
2 |y| |y|

be the flux density with respect to M (z0 ), where y = x − x0 , and

Z
F lux(u, v, MST (z0 )) = dz0 (u, v)do.
T (z )
MS 0

12
 Lemma 4.1. Let (u, v) be a regular solution of (1.1) on K(z0 ). Then
1
Ew (u, v, D(T ; z0 )) + √ F lux(u, v, MST (z0 )) = Ew (u, v, D(S; z0 )) (4.1)
2
for 0 < S ≤ T ≤ t0 .
Proof: Multiplying the equation of u by µ(α + 2)ut and that of v by λ(β + 2)vt in (1.1),
we get
µ(α + 2)(utt − ∆u + λ|u|α |v|β+2 u)ut + λ(β + 2)(vtt − ∆v + µ|u|α+2 |v|β v)vt
d
= ew (u, v) − div[µ(α + 2)∇uut + λ(β + 2)∇vvt ] = 0. (4.2)
dt
Let y = x − x0 and
1 y
η = √ ( , 1)
2 |y|
be the outward unit normal on M (z0 ). Then the weighted energy flux through M (z0 ) is given
by
η · [(−µ(α + 2)∇uut − λ(β + 2)∇vvt ), ew (u, v)]
1 µ(α + 2) 2y λ(β + 2) 2y
= √ { [|∂t u|2 − · ∇uut + |∇u|2 ] + [|∂t v|2 − · ∇vvt + |∇v|2 ]
2 2 |y| 2 |y|
+ λµ|u|α+2 |v|β+2 }
1
= √ dz0 (u, v). (4.3)
2
Integrating (4.2) over Kts (z0 ) and using (4.3), we obtain (4.1). 
Lemma 4.2. Let (u, v) be a regular solution of (1.1) on K(z0 ) \ {z0 }. Then
Z
|u|α+2 |v|β+2 dx → 0 as S → t0 . (4.4)
D(S;z0 )

Proof: We can shift z0 to the origin and use Morawetz identity to prove the lemma.
Multiplying the equation of u by µ(α + 2)[tut + x · ∇u + (d−1)u
2 ] and that of v by λ(β + 2)[tvt +
(d−1)v
x · ∇v + 2 ], summing them up, we get
 
d−1 λµ α+2 β+2
∂t tQ0 + [µ(α + 2)uut + λ(β + 2)vvt ] − div(tP0 ) + |u| |v| = 0, (4.5)
2 d
where
x x
Q0 = ew (u, v) + [µ(α + 2)( · ∇u)ut + λ(β + 2)( · ∇v)vt ],
t t
x µ(α + 2)(|ut |2 − |∇u|2 ) + λ(β + 2)(|vt |2 − |∇v|2 )
 
α+2 β+2
P0 = − λµ|u| |v|
t 2
   
x (d − 1)u x (d − 1)v
+ µ(α + 2)∇u ut + · ∇u + + λ(β + 2)∇v vt + · ∇v + .
t 2t t 2t
Integrating (4.5) over the truncated cone KST and letting T → 0, we get
 
(d − 1)
Z Z
0=− SQ0 + [µ(α + 2)uut + λ(β + 2)vvt ] dx + λµ|u|α+2 |v|β+2 dxdt
D(S) 2 KS
Z  
1 (d − 1)
+√ tQ0 + x · P0 + [µ(α + 2)uut + λ(β + 2)vvt ] do
2 MS 2
:= (I) + (II) + (III). (4.6)

13
Obviously,

(II) ≥ 0. (4.7)

On MS , we have |x| = −t and

 
1 x · ∇u 2 x · ∇v 2
Z
(III) = √ −|x| µ(α + 2)|ut − | + λ(β + 2)|vt − | do
2 MS |x| |x|
 
1 (d − 1) x · ∇u x · ∇v
Z
+√ µ(α + 2)u(ut − + λ(β + 2)v(vt − do.
2 MS 2 |x| |x|
√
Now parameterizing MS by y → (y, −|y|) and letting ṽ(y) = v(y, −|y|), we have do = 2dy,
y · ∇ũ x·∇u ∇ṽ x·∇v
|y| = |x| − ut , and y · |y| = |x| − vt . Then

µ(α + 2)|y · ∇ũ|2 + λ(β + 2)|y · ∇ṽ|2

Z
(III) = − dy
DS |y|
(d − 1) y · ∇ũ y · ∇ṽ
Z
− [µ(α + 2)ũ + λ(β + 2)ṽ ]dy
2 DS |y| |y|
d − 1 2 (d − 1)2 2 d − 1 2 (d − 1)2 2
Z  
=− |y|−1 |y · ∇ũ + ũ| − ũ + |y · ∇ṽ + ṽ| − ṽ dy
DS 2 4 2 4
2 2
d−1 y · ∇(ũ) + y · ∇(ṽ)
Z
+ dy.
4 DS |y|

Integrating by parts and using the original coordinates again, we have

 
1 1 d−1 2 d−1 2
Z
(III) = √ |tut + x · ∇u + u| + |tvt + x · ∇v + v| do
2 MS t 2 2
d−1
Z
+ (u2 + v 2 )do
2 ∂D(S)
d−1
Z
≥ So(1) + (u2 + v 2 )do.
4 ∂D(S)

That is,
d−1
Z
−(III) ≤ −So(1) − (u2 + v 2 )do. (4.8)
4 ∂D(S)

14
Meanwhile,
" #
2 2
1 1 (d − 1) (d − 1) · ∇u 1 − 1
Z
xu x d u
(I) = − S |ut |2 + ∇u + − u− ( )2 2 dx
D(S) 2 2 2 |x|2 2 |x|2 2 2 |x|
" #
2
1 1 (d − 1) xv (d − 1) x · ∇v 1 d − 1 2 v2
Z
2
− S |vt | + ∇v + − v− ( ) dx
D(S) 2 2 2 |x|2 2 |x|2 2 2 |x|2
(d − 1) (d − 1)
Z Z
α+2 β+2
− λµ|u| |v| dx − [ut (x · ∇u + u) + vt (x · ∇v + v)]dx
D(S) D(S) 2 2
" #
2
1 1 (d − 1) xu (d − 1)(d − 3) u2
Z
=− S |ut |2 + ∇u + + dx
D(S) 2 2 2 |x|2 8 |x|2
" #
2
1 1 (d − 1) xv (d − 1)(d − 3) v 2
Z
2
− S |vt | + ∇v + + dx
D(S) 2 2 2 |x|2 8 |x|2
(d − 1) (d − 1)
Z Z
− [ut (x · ∇u + u) + vt (x · ∇v + v)]dx − λµ|u|α+2 |v|β+2 dx
D(S) 2 2 D(S)
d−1
Z
− (u2 + v 2 )do
4 ∂D(S)
d−1
Z Z
≥− Sλµ|u|α+2 |v|β+2 dx − (u2 + v 2 )do. (4.9)
D(S) 4 ∂D(S)

Using (4.6)–(4.9), we can get that

Z
−S λµ|u|α+2 |v|β+2 dx + So(1) ≤ 0,
D(S)

which implies that

Z
|u|α+2 |v|β+2 dx → 0 as S → t0 . 
D(S)

1 1
Let k · kq,s,τ denote the norm in the space Lq ([s, τ ]; Ḃq2 (D(t; Z0 ))), Ḃq2 (Rd ) be Besov space
1
and Ḃq2 (D(t)) be the local Besov space, q < 2d.
Proposition 4.3 Suppose that (u, v) is a classical solution to equations (1.1) on K(z0 ) \
1
{z0 }. Then (u, v) is bounded in Lq ([O, t0 ]; Ḃq2 (D(t; z0 ))) and

kukq,0,t0 < C(z0 , E(u, D(t0 ; z0 ))). (4.10)

Proof: Similar to (1.8) in [50], for any p ∈ [1, ∞], we have

kuk 1 + kvk 1
Lq ([s,τ ];Ḃq2 (D(t,z0 ))) Lq ([s,τ ];Ḃq2 (D(t,z0 )))
1
≤ C[Ew (u, v; D(s, z0 ))] 2 + k|u|α |v|β+2 uk 1
Lp ([s,τ ];Ḃp2 (D(t,z0 )))

+ k|u|α+2 |v|β vk 1 (4.11)

Lp ([s,τ ];Ḃp2 (D(t,z0 )))

15
and
1
kukq,s,τ + kvkq,s,τ . [Ew (u, v; D(s, z0 ))] 2 + k|u|α+β+3 kp,s,τ + k|v|α+β+3 kp,s,τ , (4.12)
α+β+3 α+β+3 α+β+2
k|u| kp,s,τ . k|ũ| kp,s,τ . k|ũ| kLp1 kũkq,s,τ
α+β+2
. k|u| kLp1 (Ksτ (z0 )) kukq,s,τ (4.13)
α+β+3 α+β+3 α+β+2
k|v| kp,s,τ . k|ṽ| kp,s,τ . k|ṽ| kLp1 kṽkq,s,τ
α+β+2
. k|v| kLp1 (Ksτ (z0 )) kvkq,s,τ (4.14)
k|u| α+β+2
k L p1 = kukα+β+2
L p2 , k|v|α+β+2 kLp1 = kvkα+β+2
L p2 . (4.15)
Here
(d + 1) 2(d + 1)
p1 = , p2 = . (4.16)
2 d−2
By Sobolev embedding,
d−2
kukLp2 (D(t;z0 )) . kukθ 1 kuk1−θ
Lα+β+4 (D(t;z0 ))
, θ=
Ḃq2 (D(t;z0 )) d−1
kvk Lp2 (D(t;z 0 ))
. kvkθ 1 kvk1−θ
Lα+β+4 (D(t;z0 ))
,
Ḃq2 (D(t;z0 ))

then (4.12) implies that

1
kukq,s,τ + kvkq,s,τ . [Ew (u, v; D(s, z0 ))] 2 + sup kukϑLα+β+4(D(t;z0 )) kukγq,s,τ
s≤t≤τ

+ sup kvkϑLα+β+4(D(t;z0 )) kvkγq,s,τ . (4.17)

s≤t≤τ

By Lemma 4.2, we have

sup kukϑLα+β+4(D(t;z0 )) + sup kukϑLα+β+4(D(t;z0 )) ≤ ǫ. (4.18)
s≤t≤τ s≤t≤τ

Note that Ew (u, v; D(s; z0 )) is bounded by the initial energy Ew (u0 , v0 ). This implies that for
any ǫ > 0, and s close to t0 and s < r < t0 , we have
kukq,s,τ + kvkq,s,τ ≤ C(Ew (u0 , v0 )) + ǫkukγq,s,τ + kvkγq,s,τ . (4.19)
Choosing ǫ < ǫ0 2−γ [C(Ew (u0 , v0 ))]1−γ , we can obtain (4.10). 
Theorem 4.4. Suppose that (u, v) is a solution to equations (1.1) with smooth initial data.
If d = 3, (α, β) = (0, 2) or d = 4, (α, β) = (0, 0), then (u, v) is regular.
Proof: Without loss of generality, we suppose that (u10 , u20 ) and (v10 , v20 ) have compact
supports by finite propagation speed. Let (u, v) ∈ C ∞ (Rd × [0, t0 ]) be the unique maximal
solution of (1.1) and consider x0 ∈ Rd . Then (u, v) may be extended smoothly to a neighborhood
of z0 = (x0 , t0 ). By differentiating equations (1.1) and using Strichartz estimate
kukLq (Rd+1 ) . ku10 k 1 + ku20 k 1 + k|u|α |v|β+2 ukLp(Rd+1 ) ,
H 2 ,2 (Rd ) H − 2 ,2 (Rd )

kvkLq (Rd+1 ) . kv10 k 1 + kv20 k 1 + k|u|α+2 |v|β vkLp (Rd+1 ) (4.20)

H 2 ,2 (Rd ) H − 2 ,2 (Rd )

we obtain the following estimate for any first-order derivative:

kDukLq (Ksτ (z0 )) + kDvkLq (Ksτ (z0 ))
. C(Ew (Du, Dv, D(S; z0 ))) + k|u|α |v|β+2 |Du|kLp (Ksτ (z0 )) + k|u|α+1 |v|β+1 |Dv|kLp (Ksτ (z0 ))
+ k|u|α+2 |v|β |Dv|kLp (Ksτ (z0 )) + k|u|α+1 |v|β+1 |Du|kLp (Ksτ (z0 ))
:= C(Ew (Du, Dv, D(S; z0 ))) + (I) + (II) + (III) + (IV ). (4.21)

16
Similar to the computations in the proof of Proposition 4.3, we can get
(I) + (II) + (III) + (IV ) . [kDukLq + kDvkLq ][kukα+β+2 α+β+2
Lp2 (K τ (z0 )) + kvkLp2 (K τ (z0 )) ]
s s

. sup [kukβLα+β+4(D(t;z0 )) + kvkβLα+β+4 (D(t;z0 )) ][kukγ−1

q,s,τ + kvkγ−1
q,s,τ ]
s≤t≤τ

× [kDukLq (Ksτ (z0 )) + kDvkLq (Ksτ (z0 )) ]. (4.22)

By Lemma 4.2, the coefficients in front of [kDukLq (Ksτ (z0 )) + kDvkLq (Ksτ (z0 )) ] on the right hand
side of (4.21) can be small enough, which implies that Du ∈ Lq (K(z0 )) and Dv ∈ Lq (K(z0 )).
Differentiating the equation again, we can get the equations of D2 u and D2 v. If d = 3,
(α, β = (0, 2), there will appear the additional terms |v|3 DuDv and |v|2 u|Dv|2 in the equation
of D2 u, |u|3 DuDv and |u|2 v|Du|2 in the equation of D2 v; If d = 4, (α, β = (0, 0), there will
appear the additional terms vDuDv and u|Dv|2 in the equation of D2 u, uDuDv and v|Du|2 in
the equation of D2 v. Using Hölder’s inequality and the Sobolev embedding theorem, we obtain
k|v|3 DuDvkLp + k|v|2 u|Dv|2 kLp + k|u|3 DuDvkLp + k|u|2 v|Du|2 kLp
. [kDukLq + kukLq + kDvkLq + kvkLq ]ρ [kD2 ukLq + kD2 vkLq ]σ (4.23)
if d = 3, (α, β) = (0, 2) and
kvDuDvkLp + ku|Dv|2 kLp + kuDuDvkLp + kv|Du|2 kLp
. [kDukLq + kukLq + kDvkLq + kvkLq ]ρ [kD2 ukLq + kD2 vkLq ]σ (4.24)
if d = 4, (α, β) = (0, 0), where σ ≤ 1. If d = 3, then u, v ∈ W 2,q (K(z0 )) implies that
|v|4 u, |u|4 v ∈ H 2,2 (K(z0 )), which in turn implies that D3 u, D3 v ∈ L∞ ([0, t0 ]; L2 (D(t; z0 ))).
Using energy estimate and Grownwall inequality, we can get
Dk u, Dk v ∈ L2 ([0, t0 ]; L2 (D(t; z0 ))) for all k (4.25)
and
Ew (u, v, D(t; z0 )) → 0 as t → t0 . (4.26)
Consequently, for given ǫ > 0, Ew (u, v, D(s; z0 )) < ǫ for some s < t0 . Since (u, v) is smooth
in the neighbourhood of D(s; z0 ), hence Ew (u, v, D(s; z̄)) < ǫ,
sup[kukL6(D(s;z̄)) + kvkL6 (D(s;z̄)) ] < ǫ when d = 3 (4.27)
s≤t

and
sup[kukL4(D(s;z̄)) + kvkL4 (D(s;z̄)) ] < ǫ when d = 4 (4.28)
s≤t

keep true for z̄ in a neighborhood of z0 . By the standard argument similar to Remark 3.4 in
[58], (u, v) can be smoothly extended to this neighborhood.

5 Ḣ 1 × L2 scattering results when (α, β) = (0, 2) in dimen-

sion 3 and (α, β) = (0, 0) in dimension 4
In this section, we focus on the wellposedness and scattering for the solution (u, v) of (2.7)
with (u, ut ), (v, vt ) ∈ Ḣ 1 × L2 when α + β = 2 in dimension 3 and (α, β) = (0, 0) in dimension
4.

17
 First, we prove the following local wellposedness result.
Theorem 5.1. Let (α, β) = (0, 2), ρ = 5 when d = 3 and (α, β) = (0, 0), ρ = 3 when
d = 4, (u10 , u20 ), (v10 , v20 ) ∈ Ḣ 1 (Rd ) × L2 (Rd ) satisfying

ku10 k2 + k∇u10 k2 + ku20 k2 + kv10 k2 + k∇v10 k2 + kv20 k2 < δ, (5.1)

where δ > 0 is sufficiently small. Then there exists a unique solution (u, v) of (2.7) and satisfies
(u, ut ), (v, vt ) ∈ V ⊕ L∞ (R, L2 (Rd )), where V := Lρ (R, Ḣ s,r (Rd )) ∩ L∞ (R, Ḣ 1 (Rd )) with

(d − 1)ρ − (d + 1) 2(d − 1)ρ

s := , r := . (5.2)
(d − 1)ρ (d − 1)ρ − 4
2(d−1)ρ
Proof: By the results of Theorem 2 in [46] and its Corollary, we can choose q = (d−1)ρ−4 (because
2 < ρ < ∞, d = 3 or d = 4) such that

K̇(t)u10 + K(t)u20 ∈ V, K̇(t)v10 + K(t)v20 ∈ V,

kK̇(t)u10 + K(t)u20 ; V k ≤ cδ, kK̇(t)v10 + K(t)v20 ; V k ≤ cδ.

For (u1 , v1 ), (u2 , v2 ) ∈ V ⊕ V , using the decay estimate as that of Theorem 0(a) in [?], we have
t
sin[ω(t − τ )]
Z
k [λ|u1 |α |v1 |β+2 u1 (τ ) − λ|u2 |α |v2 |β+2 u2 (τ )]dτ kḢ s,r
−∞ ω
t
sin[ω(t − τ )]
Z
+k [µ|u1 |α+2 |v1 |β v1 (τ ) − µ|u2 |α+2 |v2 |β v2 (τ )]dτ kḢ s,r
−∞ ω
Z t
2
≤c (t − τ )− ρ k[λ|u1 |α |v1 |β+2 u1 (τ ) − λ|u2 |α |v2 |β+2 u2 (τ )]kḢ s̄,r′ dτ
−∞
Z t
2
+c (t − τ )− ρ k[µ|u1 |α+2 |v1 |β v1 (τ ) − µ|u2 |α+2 |v2 |β v2 (τ )]kḢ s̄,r′ dτ, (5.3)
−∞

. Letting r̃1′ = (d+2)ρ+2

′ ′
d+1
where s̄ = (d−1)ρ 2dρ , we know that the imbedding Ḣ 1,r̃ ⊂ Ḣ s̄,r holds. We
will use Theorem 1 on Page 119 in [54] to establish the following estitimates.
If d = 3, then

λk[|u1 |α |v1 |β+2 u1 (τ ) − |u2 |α |v2 |β+2 u2 (τ )]kḢ s̄,r′ + µk[|u1 |α+2 |v1 |β v1 (τ ) − |u2 |α+2 |v2 |β v2 (τ )]kḢ s̄,r′
X
≤c k|v1 |4 Dk u1 + u1 |v1 |3 Dk v1 − |v2 |4 Dk u2 − u2 |v2 |3 Dk v2 kLr̃′
|k|=1
X
+c k|u1 |4 Dk v1 + v1 |u1 |3 Dk u1 − |u2 |4 Dk v2 − v2 |u2 |3 Dk u2 kLr̃′
|k|=1
X
≤c k|v1 − v2 |(|v1 |3 + |v2 |3 )Dk u1 kLr̃′ + k|u1 − u2 ||v1 |3 Dk v1 kLr̃′
|k|=1

+ k|v1 − v2 ||u2 |(|v1 |2 + |v2 |2 )Dk v2 kLr̃′ + k|v2 |4 (Dk u1 − Dk u2 )kLr̃′ + k|u2 ||v2 |3 (Dk v1 − Dk v2 )kLr̃′
X
+c k|u1 − u2 |(|u1 |3 + |u2 |3 )Dk v1 kLr̃′ + k|v1 − v2 ||u1 |3 Dk u1 kLr̃′
|k|=1

+ k|u1 − u2 ||v2 |(|u1 |2 + |u2 |2 )Dk u2 kLr̃′ + k|u2 |4 (Dk v1 − Dk v2 )kLr̃′ + k|v2 ||u2 |3 (Dk u1 − Dk u2 )kLr̃′
:= (1) + (2) + (3) + (4) + (5) + (6) + (7) + (8) + (9) + (10). (5.4)

18
 If d = 4, then

λk[|u1 |α |v1 |β+2 u1 (τ ) − |u2 |α |v2 |β+2 u2 (τ )]kḢ s̄,r′ + µk[|u1 |α+2 |v1 |β v1 (τ ) − |u2 |α+2 |v2 |β v2 (τ )]kḢ s̄,r′
X
≤c k|v1 |2 Dk u1 + u1 v1 Dk v1 − |v2 |2 Dk u2 − u2 v2 Dk v2 kLr̃′
|k|=1
X
+c k|u1 |2 Dk v1 + v1 u1 Dk u1 − |u2 |2 Dk v2 − v2 u2 Dk u2 kLr̃′
|k|=1
X
≤c k|v1 − v2 |(|v1 | + |v2 |)Dk u1 kLr̃′ + k|u1 − u2 ||v1 |Dk v1 kLr̃′
|k|=1

+ k|v1 − v2 ||u2 |Dk v2 kLr̃′ + k|v2 |2 (Dk u1 − Dk u2 )kLr̃′ + k|u2 ||v2 |(Dk v1 − Dk v2 )kLr̃′
X
+c k|u1 − u2 |(|u1 | + |u2 |)Dk v1 kLr̃′ + k|v1 − v2 ||u1 |Dk u1 kLr̃′
|k|=1

+ k|u1 − u2 ||v2 |Dk u2 kLr̃′ + k|u2 |2 (Dk v1 − Dk v2 )kLr̃′ + k|v2 ||u2 |(Dk u1 − Dk u2 )kLr̃′
:= (1)′ + (2)′ + (3)′ + (4)′ + (5)′ + (6)′ + (7)′ + (8)′ + (9)′ + (10)′ . (5.5)

Using Hölder’s inequality, (1), (2), (3), (6),(7) and (8) in (5.4), as well as (1)′ , (2)′ , (3)′ , (6)′ , (7)′
and (8)′ in (5.5) can be estimated by

[ku1 − u2 kLr̃′ p̂ + kv1 − v2 kLr̃′ p̂ ][ku1 kρ−2

Lr̃′ (ρ−2)q̂
+ ku2 kρ−2
Lr̃′ (ρ−2)q̂
+ kv1 kρ−2
Lr̃′ (ρ−2)q̂
+ kv2 kρ−2
Lr̃′ (ρ−2)q̂
]
× [kDk u1 kLr̃′ r̂ + kDk v1 kLr̃′ r̂ + kDk u2 kLr̃′ r̂ + kDk v2 kLr̃′ r̂ ]. (5.6)

2 dρ(ρ−1)
Since the choice r̂ = r̃ ′ and p̂ = (ρ − 2)q̂ can lead to r̃′ p̂ = r̃′ (ρ − 2)q̂ = ρ+1 , and the
dρ(ρ−1)
d+2
embedding Ḣ s,r ⊂ L ρ+1 holds if and only if ρ = d−2 , we get

λk[|u1 |α |v1 |β+2 u1 (τ ) − |u2 |α |v2 |β+2 u2 (τ )]kḢ s̄,r′ + µk[|u1 |α+2 |v1 |β v1 (τ ) − |u2 |α+2 |v2 |β v2 (τ )]kḢ s̄,r′
ρ−2 ρ−2
≤ c[ku1 − u2 kḢ s,r + kv1 − v2 kḢ s,r ][ku1 kḢ s,r
+ ku2 kḢ s,r
+ kv1 kρ−2
Ḣ s,r
ρ−2
+ kv2 kḢ s,r
]
× [ku1 kḢ 1,2 + ku2 kḢ 1,2 + kv1 kḢ 1,2 + kv2 kḢ 1,2 ]
ρ−1 ρ−1 ρ−1 ρ−1
+ c[ku1 kḢ s,r
+ ku2 kḢ s,r
+ kv1 kḢ s,r
+ kv2 kḢ s,r
][ku1 − u2 kḢ 1,2 + kv1 − v2 kḢ 1,2 ]. (5.7)

Then we can obtain the estimate for the left hand side of (5.3) as follows:

LHS(5.3)
Z t
2 ρ−2
≤c (t − τ )− ρ [ku1 − u2 kḢ s,r + kv1 − v2 kḢ s,r ][ku1 kḢ s,r
+ ku2 kρ−2
Ḣ s,r
ρ−2
+ kv1 kḢ s,r
ρ−2
+ kv2 kḢ s,r
]
−∞

× [ku1 kL∞ (R,Ḣ 1,2 (Rd )) + ku2 kL∞ (R,Ḣ 1,2 (Rd )) + kv1 kL∞ (R,Ḣ 1,2 (Rd )) + kv2 kL∞ (R,Ḣ 1,2 (Rd )) ]dτ
Z t
2
+c (t − τ )− ρ [ku1 kḢ ρ−1
s,r
ρ−1
+ ku2 kḢ s,r
+ kv1 kḢρ−1
s,r
ρ−1
+ kv2 kḢ s,r
]
−∞

× [ku1 − u2 kL∞ (R,Ḣ 1,2 (Rd )) + kv1 − v2 kL∞ (R,Ḣ 1,2 (Rd )) ]dτ. (5.8)

1 1 1 1
Note that for 0 < θ < 1, ρ̃ = θ and p = q + ρ̃′ , the following general Young’s inequality holds

1
k ∗ gkLq (R) ≤ ckgkLp(R) .
|t|θ

19
Taking ρ̃ = ρ2 , q = ρ and p = ρ
ρ−1 , we have

t
sin[ω(t − τ )]
Z
k [λ|u1 |α |v1 |β+2 u1 (τ ) − λ|u2 |α |v2 |β+2 u2 (τ )]dτ kLρ (R,Ḣ s,r (Rd ))
−∞ ω
t
sin[ω(t − τ )]
Z
+k [µ|u1 |α+2 |v1 |β v1 (τ ) − µ|u2 |α+2 |v2 |β v2 (τ )]dτ kLρ (R,Ḣ s,r (Rd ))
−∞ ω
≤ c[ku1 kL∞ (R,Ḣ 1,2 (Rd )) + ku2 kL∞ (R,Ḣ 1,2 (Rd )) + kv1 kL∞ (R,Ḣ 1,2 (Rd )) + kv2 kL∞ (R,Ḣ 1,2 (Rd )) ]
Z +∞ ρ ρ ρ(ρ−2) ρ(ρ−2) ρ(ρ−2) ρ(ρ−2)
 ρ−1
ρ
ρ−1 ρ−1 ρ−1 ρ−1 ρ−1 ρ−1
[ku1 − u2 kḢ s,r + kv1 − v2 kḢ s,r ][ku1 kḢ s,r + ku2 kḢ s,r + kv1 kḢ s,r + kv2 kḢ s,r ]dτ
−∞
+ c[ku1 − u2 kL∞ (R,Ḣ 1,2 (Rd )) + kv1 − v2 kL∞ (R,Ḣ 1,2 (Rd )) ]
Z +∞  ρ−1
ρ
ρ ρ ρ ρ
× [ku1 kḢ s,r + ku2 kḢ s,r + kv1 kḢ s,r + kv2 kḢ s,r ]dτ
−∞

≤ c[ku1 − u2 kV + kv1 − v2 kV ][ku1 kρ−1

V + ku2 kρ−1
V + kv1 kρ−1
V + kv2 kρ−1
V ]. (5.9)

Since the imbedding Ḣ s,r ⊂ L2ρ holds if and only if ρ = d+2

d−2

t
sin[ω(t − τ )]
Z
k [λ|u1 |α |v1 |β+2 u1 (τ ) − λ|u2 |α |v2 |β+2 u2 (τ )]dτ kL∞ (R,Ḣ 1,2 (Rd ))
−∞ ω
t
sin[ω(t − τ )]
Z
+k [µ|u1 |α+2 |v1 |β v1 (τ ) − µ|u2 |α+2 |v2 |β v2 (τ )]dτ kL∞ (R,Ḣ 1,2 (Rd ))
−∞ ω
Z +∞
≤c k[|u1 |α |v1 |β+2 u1 (τ ) − |u2 |α |v2 |β+2 u2 (τ )]kL2 (Rd ) dτ
−∞
Z +∞
+c k[|u1 |α+2 |v1 |β v1 (τ ) − |u2 |α+2 |v2 |β v2 (τ )]kL2 (Rd ) dτ
−∞
Z +∞
≤c [ku1 kρ−1 ρ−1 ρ−1 ρ−1
L2ρ + ku2 kL2ρ + kv1 kL2ρ + kv2 kL2ρ ][ku1 − u2 kL + kv1 − v2 kL ]dτ
2ρ 2ρ
−∞

≤ c[ku1 kρ−1
V + ku2 kρ−1
V + kv1 kρ−1
V + kv2 kρ−1
V ][ku1 − u2 kV + kv1 − v2 kV ]. (5.10)

Denoting the transformations which map (u, v) into the right hand sides of the equations in
(2.7) respectively by T1 and T2 , we have shown

kT1 (u1 , v1 ) − T1 (u2 , v2 )kV + kT2 (u1 , v1 ) − T2 (u2 , v2 )kV

≤ c[ku1 kρ−1
V + ku2 kρ−1
V + kv1 kρ−1
V + kv2 kρ−1
V ][ku1 − u2 kV + kv1 − v2 kV ],

and
kT1 (u, v)kV + kT2 (u, v)kV ≤ cδ + c[kukρV + kvkρV ].
If cδ ≤ 41 δ1 and 4cδ1ρ−1 ≤ 21 , then ku1 kV , ku2 kV , kv1 kV , kv2 kV ≤ δ1 , and

1
kT1 (u1 , v1 ) − T1 (u2 , v2 )kV + kT2 (u1 , v1 ) − T2 (u2 , v2 )kV ≤ [ku1 − u2 kV + kv1 − v2 kV ]
2
kT1 (u, v)kV + kT2 (u, v)kV ≤ δ1 .

By the contraction mapping principle, there exists a unique solution (u, v) within the ball
kukV + kvkV ≤ δ1 . To show the uniqueness within the whole of V , we give the following

20
estimates

ku1 − u2 kLρ (I,Ḣ s,r (Rd )) + kv1 − v2 kLρ (I,Ḣ s,r (Rd ))
≤ c[ku1 kL∞ (I,Ḣ 1,2 (Rd )) + ku2 kL∞ (I,Ḣ 1,2 (Rd )) + kv1 kL∞ (I,Ḣ 1,2 (Rd )) + kv2 kL∞ (I,Ḣ 1,2 (Rd )) ]
× [ku1 kρ−2
Lρ (I,Ḣ s,r (Rd ))
+ ku2 kρ−2
Lρ (I,Ḣ s,r (Rd ))
+ kv1 kρ−2
Lρ (I,Ḣ s,r (Rd ))
+ kv2 kρ−2
Lρ (I,Ḣ s,r (Rd ))
]
× [ku1 − u2 kLρ (I,Ḣ s,r (Rd )) + kv1 − v2 kLρ (I,Ḣ s,r (Rd )) ]
+ [ku1 kρ−1
Lρ (I,Ḣ s,r (Rd ))
+ ku2 kρ−1
Lρ (I,Ḣ s,r (Rd ))
+ kv1 kρ−1
Lρ (I,Ḣ s,r (Rd ))
+ kv2 kρ−1
Lρ (I,Ḣ s,r (Rd ))
]
× [ku1 − u2 kL∞ (I,Ḣ 1,2 (Rd )) + kv1 − v2 kL∞ (I,Ḣ 1,2 (Rd )) ]

and

ku1 − u2 kL∞ (I,Ḣ s,r (Rd )) + kv1 − v2 kL∞ (I,Ḣ s,r (Rd ))
≤ [ku1 kρ−1
Lρ (I,Ḣ s,r (Rd ))
+ ku2 kρ−1
Lρ (I,Ḣ s,r (Rd ))
+ kv1 kρ−1
Lρ (I,Ḣ s,r (Rd ))
+ kv2 kρ−1
Lρ (I,Ḣ s,r (Rd ))
]
× [ku1 − u2 kLρ (I,Ḣ s,r (Rd )) + kv1 − v2 kLρ (I,Ḣ s,r (Rd )) ],

where I = (−∞, T̄ ).
Since u1 , u2 , v1 , v2 ∈ Lρ (R, Ḣ s,r (Rd )), we can choose |T̄ | large enough such that
1
ku1 − u2 kVT̄ + kv1 − v2 kVT̄ ≤ [ku1 − u2 kVT̄ + kv1 − v2 kVT̄ ]
2
with
VT̄ = Lρ (I, Ḣ s,r (Rd )) ∩ L∞ (I, L∞ (I, Ḣ 1,2 (Rd )),
hence u1 = u2 , v1 = v2 in VT̄ . Step by step, it is possible to replace T̄ by T̄ + ǫ with ǫ possibly
depending on u1 , u2 , v1 , v2 , just like the arguments above, we can obtain u1 = u2 , v1 = v2 in
V. 
As a direct result of this theorem, we have the following corollary
Corollary 5.2

ku(t) − u− −
0 (t)ke + kv(t) − v0 (t)ke → 0 as t → −∞. (5.11)

Here kw(t)ke denote the energy norm

1 1
kw(·, t)k2e = [k(−∆) 2 w(·, t)k2L2 + kw(·, t)k2L2 ].
2

Proof: Note that Lρ (R, Ḣ s,r (Rd )) ⊂ Lρ (R, L2ρ (Rd )), we have
Z t
ku(t) − u− −
0 (t)ke + kv(t) − v0 (t)ke ≤ c [k|u|α |v|β+2 u(τ )kL2 + k|u|α+2 |v|β v(τ )kL2 ]dτ
−∞
Z t
≤c [ku(τ )kρL2ρ + kv(τ )kρL2ρ ]dτ → 0
−∞

as t → −∞. 

21
 Define
Z ∞
1 1
u+
0 (t) := u(t) + (−∆)− 2 sin[(−∆) 2 (t − τ )](λ|u|α |v|β+2 u)dτ, u+ + +
0 (0) = u , u0t (0) = ũ
+
t
Z ∞
1 1
v0+ (t) := v(t) + (−∆)− 2 sin[(−∆) 2 (t − τ )](µ|u|α+2 |v|β v)dτ, v0+ (0) = v + , v0t
+
(0) = ṽ + .
t

Exactly similar to the discussions above, we can get ku(t) − u+ +

0 (t)ke + kv(t) − v0 (t)ke → 0 as
t → +∞. Now we have proven
Theorem 5.3 The scattering operators S1 : (u10 , u20 ) → (u+ , ũ+ ) and S2 : (v10 , v20 ) →
(v , ṽ ) exist in the sense of energy norms in a whole neighbourhood of the origin in Ḣ 1,2 (Rd )×
+ +

L2 (Rd ).

6 Ḣ sc × Ḣ sc−1 scattering results in the energy supercritical

case
In this section, we discuss the local wellposedness and scattering results on (1.1) in the
2
energy supercritical case. Denote sc = d2 − α+β+2 . The space-time norm k · kL1t L2x means that
k · kL1t L2x ((−T,T )×Rd ) , while we will omit (−T, T ) × Rd) below, 0 < T ≤ +∞. We state the results
as follows.
Theorem 6.1 Let d = 3 or d = 4, (u10 , u20 ) ∈ Ḣ sc (Rd ) × Ḣ sc −1 (Rd ) and (v10 , v20 ) ∈
Ḣ sc (Rd ) × Ḣ sc −1 (Rd ). Then the Cauchy problem (1.1) possesses a unique solution (u, v) with
the maximal lifespan u : I × Rd → R and v : I × Rd → R.
Moreover, there exists δ0 > 0 and if

ku10 kḢ sc (Rd ) + ku20 kḢ sc −1 (Rd ) + kv10 kḢ sc (Rd ) + kv20 kḢ sc −1 (Rd ) ≤ δ0 ,

there exist functions pairs (u1± , u2± ) ∈ Ḣ sc (Rd ) × Ḣ sc −1 (Rd ) and (v1± , v2± ) ∈ Ḣ sc (Rd ) ×
Ḣ sc −1 (Rd ) such that

|∇|−1 sin(t − t0 )|∇|

    
u(t) cos((t − t0 )|∇| u1±
− → 0. (6.1)
ut (t) −|∇| sin(t − t0 )|∇| cos(t − t0 )|∇| u2± Ḣ sc ×Ḣ sc −1

and
|∇|−1 sin(t − t0 )|∇|
    
v(t) cos((t − t0 )|∇| v1±
− → 0. (6.2)
vt (t) −|∇| sin(t − t0 )|∇| cos(t − t0 )|∇| v2± Ḣ sc ×Ḣ sc −1

as t → ±∞.
Proof: First, applying Strichartz’s estimate to (2.7), we have

kukL∞Ḣxsc + kvkL∞ Ḣxsc . ku10 kḢxsc + ku20 kḢxsc −1 + k|∇|sc −1 (|u|α |v|β+2 u)kL1t L2x
t t

+ kv10 kḢxsc + kv20 kḢxsc −1 + k|∇|sc −1 (|u|α+2 |v|β v)kL1t L2x . (6.3)

Note that sc < 2. We discuss it in two cases.

Case 1: d = 3. We define the working space as follows.

kukXT + kvkXT = kukL∞ Ḣxsc + k|∇|sc −1 ukL2+ L∞−

x
+ kukL2(α+β+2)− L2(α+β+2)−
t t x

+ kvkL∞ Ḣxsc + k|∇|sc −1 vkL2+ L∞−

x
+ kvkL2(α+β+2)− L2(α+β+2)− . (6.4)
t t x

22
 By Lemma 2.5 and Lemma 2.6, we obtain

k|∇|sc −1 (|u|α |v|β+2 u)kL1t L2x + k|∇|sc −1 (|u|α+2 |v|β v)kL1t L2x
. [kukα+β+2 α+β+2
2(α+β+2)+ + kvk 2(α+β+2)+ ][k|∇|
sc −1
ukL∞− + k|∇|sc −1 vkL∞− ]
Lx Lx x x
L1t

. [kukα+β+2
2(α+β+2)− 2(α+β+2)+ + kvkα+β+2
2(α+β+2)− 2(α+β+2)+ ][k|∇|
sc −1
ukL2+L∞− + k|∇|sc −1 vkL2+ L∞− ]
Lt Lx Lt Lx t x t x

. kukα+β+3
XT + kvkα+β+3
XT . (6.5)

Letting (a′ , b′ ) = (1, 2) in Lemma 2.4, using Strichartz’s estimate, we get

k|∇|sc −1 ukL2+ L∞−

x
+ k|∇|sc −1 vkL2+ L∞−
x
t t
sc −1
. ku10 kḢxsc + ku20 kḢxsc −1 + k|∇| (|u|α |v|β+2 u)kL1t L2x
+ ku10 kḢxsc + ku20 kḢxsc −1 + k|∇|sc −1 (|u|α+2 |v|β v)kL1t L2x (6.6)

and

kukL2(α+β+2)− L2(α+β+2)+ + k|vkL2(α+β+2)− L2(α+β+2)+

t x t x
sc −1
. ku10 kḢxsc + ku20 kḢxsc −1 + k|∇| (|u| |v|β+2 u)kL1t L2x
α

+ ku10 kḢxsc + ku20 kḢxsc −1 + k|∇|sc −1 (|u|α+2 |v|β v)kL1t L2x . (6.7)

Putting the results of (6.4)–(6.7) together, we have

kukXT + kvkXT . ku10 kḢxsc + ku20 kḢxsc −1 + kv10 kḢxsc + kv20 kḢxsc −1 + kukα+β+3
XT + kvkα+β+3
XT .

If T small enough, we have

kukXT + kvkXT ≤ C[ku10 kḢxsc + ku20 kḢxsc −1 + kv10 kḢxsc + kv20 kḢxsc −1 ] := M (6.8)

for some M large enough.

Case 2: d = 4. We define the working space as

kukXT + kvkXT = kukL∞Ḣxsc + k|∇|sc −1 ukL2t L8x + kuk 2(α+β+2)

8(α+β+2)
t 3
Lt Lx

+ kukL∞ Ḣxsc + k|∇|sc −1 ukL2t L8x + kuk 2(α+β+2)

8(α+β+2) . (6.9)
t 3
Lt Lx

By Lemma 2.5 and Lemma 2.6, we obtain

k|∇|sc −1 (|u|α |v|β+2 u)kL1t L2x + k|∇|sc −1 (|u|α+2 |v|β v)kL1t L2x

. [kukα+β+2 α+β+2
8(α+β+2) + kvk 8(α+β+2) ][k|∇|
sc −1
ukL8x + k|∇|sc −1 vkL8x ]
Lx 3 Lx 3
L1t
α+β+2 α+β+2
. [kuk 8(α+β+2) + kvk 8(α+β+2) ][k|∇|sc −1 ukL2t L8x + k|∇|sc −1 vkL2t L8x ]
2(α+β+2) 3 2(α+β+2) 3
Lt Lx Lt Lx

. kukα+β+3
XT + kvkα+β+3
XT . (6.10)

Letting (a′ , b′ ) = (1, 2) in Lemma 2.4, using Strichartz’s estimate, we get

k|∇|sc −1 ukL2t L8x + k|∇|sc −1 vkL2t L8x

. ku10 kḢxsc + ku20 kḢxsc −1 + k|∇|sc −1 (|u|α |v|β+2 u)kL1t L2x
+ ku10 kḢxsc + ku20 kḢxsc −1 + k|∇|sc −1 (|u|α+2 |v|β v)kL1t L2x . (6.11)

23
and

kuk 2(α+β+2)
8(α+β+2) + k|vk 2(α+β+2)
8(α+β+2)
Lt Lx 3 Lt Lx 3

. ku10 kḢxsc + ku20 kḢxsc −1 + k|∇|sc −1 (|u|α |v|β+2 u)kL1t L2x

+ ku10 kḢxsc + ku20 kḢxsc −1 + k|∇|sc −1 (|u|α+2 |v|β v)kL1t L2x . (6.12)

Putting the results of (6.9)–(6.12) together, we obtain

kukXT + kvkXT . ku10 kḢxsc + ku20 kḢxsc −1 + kv10 kḢxsc + kv20 kḢxsc −1 + kukα+β+3
XT + kvkα+β+3
XT .

If T small enough, we have

kukXT + kvkXT ≤ C[ku10 kḢxsc + ku20 kḢxsc −1 + kv10 kḢxsc + kv20 kḢxsc −1 ] := M (6.13)

for some M large enough.

Let A(u, v) = (A1 (u0 , v0 ), A2 (u0 , v0 )) and B0 (M ) is the ball in XT × XT with the center
is (0, 0) and the radium is M . Then (6.13) means that A reflects the ball B0 (M ) to itself.
Moreover, A is contract and Banach’s fixed point theorem gives a unique solution of (1.1).
Using the bootstrap argument, if ku10 kḢxsc + ku20 kḢxsc −1 + kv10 kḢxsc + kv20 kḢxsc −1 < δ0
small enough, we have

kukXT + kvkXT . δ0 , (6.14)

which implies that

kukX∞ + kvkX∞ . δ0 .

Now we choose scattering states as

 Z +∞ 
|∇|−1 sin(−s|∇|)
   
u1± u10
= − (|u(s)|α |v(s)|β+2 u(s)ds (6.15)
u2± u20 0 cos(−s|∇|)

and
+∞
|∇|−1 sin(−s|∇|)
    Z  
v1± v10
= − (|u(s)|α+2 |v(s)|β v(s)ds (6.16)
v2± v20 0 cos(−s|∇|)

Therefore,

|∇|−1 sin(t − t0 )|∇|

    
u(t) cos(t − t0 )|∇| u1±
−
ut (t) −|∇| sin(t − t0 )|∇| cos(t − t0 )|∇| u2± Ḣ sc ×Ḣ sc −1
Z ∞
|∇|−1 sin(−s|∇|)

= kA(t) (|u(s)|α+2 |v(s)|β v(s))dskḢ sc ×Ḣ sc −1 (6.17)
t cos(−s|∇|)

Here

|∇|−1 sin(t − t0 )|∇|

 
cos(t − t0 )|∇|
A(t) = . (6.18)
−|∇| sin(t − t0 )|∇| cos(t − t0 )|∇|

24
Using Strichartz estimate, we can obtain

|∇|−1 sin(t − t0 )|∇|

    
u(t) cos(t − t0 )|∇| u1±
−
ut (t) −|∇| sin(t − t0 )|∇| cos(t − t0 )|∇| u2± Ḣ sc ×Ḣ sc −1
sc −1 α+2 β
. k|∇| (|u(s)| |v(s)| v(s))kLa′ Lb′ ((t,+∞)×RN )
t x

. kukα+β+3
X(t,+∞) + α+β+3
kvkX(t,+∞)
→ 0 as t → +∞, (6.19)

which means (6.1). Here

1 1 1 1
+ = 1, + =1 (6.20)
a a′ b b′
and
1 N N +2
′
+ ′ = . (6.21)
a b 2
(6.2) can be obtained similarly. 

References
[1] A. Bachelot, Asymptotic completeness for the Klein-Gordon equation on the Schwarzschild
metric, Ann. Inst. H. Poincaré Phys. Théor., 61(1994), 411–441.

[2] H. Bahouri and J. Shatah, Decay estimates for the critical semilinear wave equations, Ann.
Inst. H. Poincaré Anal. Non Linéaire, 15(1998), 783–789.

[3] M. Beals and W. Strauss, Lp estimates for the wave equation with a potential, Comm.
Partial Differential Equations, 18(1993), 1365–1397.

[4] J. Bourgain and D. Li, On an endpoint Kato-Ponce inequality, Diff Integral Eq., 27(2014),
1037–1072.

[5] P. Brenner, On space-time means and everywhere defined scattering operators for nonlinear
Klein-Gordon equations, Math. Z., 186(1984), 383–391.

[6] B. Bringmann, Almost–sure scattering for the radial energy–critical nonlinear wave equa-
tion in three dimensions, Anal. PDE, 13(2020), 1011–1050.

[7] A. Bulut, The defocusing energy–supercritical cubic nonlinear wave equation in dimension
five, Trans. Amer. Math. Soc., 367(2015), 6017–6061.

[8] R. Côte, C. E. Kenig, A. Lawrie and W. Schlag, Characterization of large energy solutions
of the equivariant wave map problem: I, Amer. J. Math., 137(2015), 139–207.

[9] P. A. Deift, A. R. It·s and X. Zhou, Long–time asymptotics for integrable nonlinear wave
equations. Important developments in soliton theory, 181–204, Springer Ser. Nonlinear
Dynam., Springer, Berlin, 1993.

[10] D. Del Santo, V. Georgiev and E. Mitidieri, Global existence of the solutions and formation
of singularities for a class of hyperbolic systems, in: Geometrical Optics and Related Topics,
Cortona, 1996, in: Progr. Nonlinear Differential Equations Appl., vol. 32, Birkhäuser
Boston, Boston, MA, 1997, pp. 117–140.

25
[11] J. Dimock and B. S. Kay, Classical and quantum scattering theory for linear scalar fields
on the Schwarzschild metric. I, Ann. Physics, 175(1987), 366–426.

[12] B. Dodson, J. Lührmann and D. Mendelson, Almost sure scattering for the 4D energy-
critical defocusing nonlinear wave equation with radial data, Amer. J. Math., 142(2020),
475–504.

[13] B. Dodson, A. Lawrie, D. Mendelson and J. Murphy, Scattering for defocusing energy
subcritical nonlinear wave equations, Anal. PDE, 13(2020), 1995–2090.

[14] B. Dodson, and A. Lawrie, Scattering for the radial 3D cubic wave equation, Anal. PDE,
8(2015), 467–497.

[15] R. Donninger and J. Krieger, Nonscattering solutions and blowup at infinity for the critical
wave equation, Math. Ann., 357(2013), 89–163.

[16] T. Duyckaerts, H. Jia, C. Kenig, and F. Merle, Dynamics of the focusing critical wave
equation. Séminaire Laurent Schwartz–Équations aux dérivées partielles et applications.
Année 2015–2016, Exp. No. VIII, 9 pp., Ed. Éc. Polytech., Palaiseau, 2017.

[17] Y. L. Gao and J. Xue, Local well-posedness and small data scattering for energy super-
critical nonlinear wave equations, Appl. Anal., 100(2021), 663–674.

[18] J. Ginibre and G. Velo, The global Caucby problem for the nonLinear Klein-Gordon equa-
tion, Math. Z., 189(1985), 487–505.

[19] J. Ginibre and G. Velo, The global Cauchy problem for the nonlinear Klein-Gordon
equation-II, Ann. Inst. Henri Poincaré, 6(1989), 15–35.

[20] J. Ginibre and G. Velo, Conformal invariance and time decay for non linear wave equations.
I, Ann. Inst. Henri Poincaré, Physique théorique, 47(1987), 221–261.

[21] J. Ginibre and G. Velo, Scattering theory in the energy space for a class of nonlinear wave
equations, Commun. Math. Phys., 123(1989), 535–573.

[22] J. Ginibre and G. Velo, Generalized Strichartz inequalities for the wave equation, J. Funct.
Anal., 133(1995), 50–68.

[23] M. G. Grillakis, Regularity and asymptotic behavior of the wave equation with a critical
nonlinearity, Ann. Math., 132(1990), 485–509.

[24] K. Hidano, Scattering problem for the nonlinear wave equation in the finite energy and
conformal charge space, J. Funct. Anal., 187(2001), 274–307.

[25] A. Jeffrey and T. Kawahara, Asymptotic methods in nonlinear wave theory. Applica-
ble Mathematics Series. Pitman (Advanced Publishing Program), Boston, Mass.-London,
1982. x+256 pp. ISBN: 0-273-08509-3.

[26] L.V. Kapitanskii, Weak and yet weak solutions of semilinear wave equations, Comm. Par-
tial Differential Equations, 19(1994), 1629–1676.

[27] M. Keel and T. Tao T. Endpoint Strichartz estimates, Am J Math., 120(1998), 955–980.

26
[28] S. Katayama, Global existence for coupled systems of nonlinear wave and Klein–Gordon
equations in three space dimensions, Math. Z., 270(2012), 487–513.

[29] C. E. Kenig, Lectures on the energy critical nonlinear wave equation. CBMS Regional
Conference Series in Mathematics, 122. Published for the Conference Board of the Math-
ematical Sciences, Washington, DC; by the American Mathematical Society, Providence,
RI, 2015. xiv+161 pp. ISBN: 978-1-4704-2014-7.

[30] C. E. Kenig and F. Merle, Radial solutions to energy supercritical wave equations in odd
dimensions, Discrete Contin. Dyn. Syst., 31(2011), 1365–1381.

[31] C. Kenig, G. Ponce and L. Vega, Well-posedness and scattering results for the generalized
Korteweg-de Vries equation via contraction principle, Comm Pure Appl Math., 46(1993),
527–620.

[32] R. Killip and M. Visan, The defocusing energy–supercritical nonlinear wave equation in
three space dimensions, Trans. Amer. Math. Soc., 36(2011), 3893–3934.

[33] J. Krieger, K. Nakanishi and W. Schlag, Global dynamics of the nonradial energy-critical
wave equation above the ground state energy, Discrete Contin. Dyn. Syst., 33(2013), 2423–
2450.

[34] H. Kubo, On the small data global existence and scattering for systems of semilinear wave
equations. Hyperbolic problems and related topics, 219–234, Grad. Ser. Anal., Int. Press,
Somerville, MA, 2003.

[35] H. Kubo and K. Kubota, Scattering for systems of semilinear wave equations with different
speeds of propagation, Adv. Differential Equations, 7(2002), 441–468.

[36] P. D. Lax, Hyperbolic partial differential equations. With an appendix by Cathleen S.

Morawetz. Courant Lecture Notes in Mathematics, 14. New York University, Courant In-
stitute of Mathematical Sciences, New York; American Mathematical Society, Providence,
RI, 2006. viii+217 pp. ISBN: 978-0-8218-3576-0; 0-8218-3576-9.

[37] D. Li, On Kato-Ponce and fractional Leibniz, Rev Mat Iberoam., 35(2019), 23–100.

[38] H. Lindblad and C. Sogge, On existence and scattering with minimal regularity for semi
linear wave equations, J. Funct. Anal., 130(1995), 357–426.

[39] J. Lührmann and D. Mendelson, Random data Cauchy theory for nonlinear wave equations
of power–type on R3 , Comm. Partial Differential Equations, 39(2014), 2262–2283.

[40] B. Marshall, W. Strauss and S. Wainger, Lp -Lq estimates for the Klein-Gordon Equation,
J. Math. Pur. Appl., 59(1980), 417–440.

[41] M. Nakamura and T. Ozawa, Global solutions in the critical Sobolev space for the wave
equations with nonlinearity of exponential growth, Math. Z., 231(1999), 479–487.

[42] K. Nakanishi, Unique global existence and asymptotic behavior of solutions for wave equa-
tions with non-coercive nonlinearity, Comm. Partial Differential Equations, 24(1999), 185–
221.

27
[43] K. Nakanishi, Scattering theory for nonlinear Klein–Gordon equation with Sobolev critical
power, Internat. Math. Res. Notices, (1999), 31–60.

[44] H. Pecher, Local solutions of semilinear wave equations in H s+1 , Math. Meth. Appl. Sci.,
19(1996), 145–170.

[45] H. Pecher, Scattering for semilinear wave equations with small data in three space dimen-
sions, Math. Z., 198(1988), 277–289.

[46] H. Pecher, Nonlinear small data scattering for the wave and Klein–Gordon equation, Math.
Z., 185(1984), 261–270.

[47] H. Pecher, Low energy scattering for nonlinear Klein–Gordon equations, J. Funct. Anal.,
63(1985), 101–122.

[48] H. Pecher, Lp -Abschätzungen und klassische Lösungen fiir nichtlineare Wellengleichungen

I, Math. Z., 150(1976), 159–183.

[49] J. C. Peral, Lp estimates for the wave equation, J. Funct. Anal., 36(1980), 114–145.

[50] J. Shatah and M. Struwe, Regularity results for nonlinear wave equations, Ann. Math.,
138(1993), 503–518.

[51] J. Shatah and M. Struwe, Well–posedness in the energy space for semilinear wave equations
with critical growth, IMRN Inter. Math. Research Notices, 7(1994), 303–309.

[52] R. P. Shen, Energy distribution of radial solutions to energy subcritical wave equation with
an application on scattering theory, Trans. Amer. Math. Soc., 374(2021), 3827–3857.

[53] C. Sogge, Monographs in Analysis, II, Lectures on Nonlinear Wave Equations, International
Press, Boston, MA, 1995.

[54] E. M. Stein, Singular integrals and differentiabitity properties of functions, Princeton, New
Jersey: Princeton University Press. 1970

[55] W. Strauss, Nonlinear scattering theory at low energy, J. Funct. Anal., 41(1981), 110–133.
Ibid., 43(1981), 281–293.

[56] R. Strichartz, Convolutions with kernels having singularities on a sphere, Trans. Amer.
Math. Soc., 148(1970), 461–470.

[57] R. Strichartz, Restrictions of Fourier transforms to quadratic surfaces and decay of solu-
tions of wave equations, Duke Math. J., 44(1977), 705–714.

[58] M. Struwe, Semilinear wave equations, Bull. Amer. Math. Soc. (N.S.), 26(1992), 53–85.

[59] M. Struwe, Globally regular solutions to the u5 Klein-Gordon equation, Ann. Scuola Norm.
Sup. Pisa Cl. Sci., 15 (1988), 495–513.

[60] T. Tao, Nonlinear dispersive equations. Local and global analysis. CBMS Regional Confer-
ence Series in Mathematics, 106. Published for the Conference Board of the Mathematical
Sciences, Washington, DC; by the American Mathematical Society, Providence, RI, 2006.
xvi+373 pp. ISBN: 0-8218-4143-2.

28
[61] D. Tataru, Strichartz estimates in the hyperbolic space and global existence for the semi-
linear wave equation, Trans. Amer. Math. Soc., 353(2000), 795–807.

[62] B. X. Wang, Nonlinear scattering theory for a class of wave equations in H s , J. Math.
Anal. Appl., 296(2004), 74–96.

[63] J. Wirth, Wave equations with time–dependent dissipation. I. Non–effective dissipation, J.

Differential Equations, 222(2006), 487–514.

29