www.elsevier.com/locate/anihpc
Solitary waves for nonlinear Klein–Gordon equations coupled with Born–Infeld theory
Yong Yu
Courant Institute of Mathematical Sciences, 251 Mercer Street, New York, NY 10012, USA Received 21 February 2009; received in revised form 28 September 2009; accepted 5 October 2009
Available online 10 November 2009
Abstract
We consider the nonlinear Klein–Gordon equations coupled with the Born–Infeld theory under the electrostatic solitary wave ansatz. The existence of the least-action solitary waves is proved in both bounded smooth domain case andR3case. In particular, for bounded smooth domain case, we study the asymptotic behaviors and profiles of the positive least-action solitary waves with respect to the frequency parameterω. We show that whenκandωare suitably large, the least-action solitary waves admit only one local maximum point. Whenω→ ∞, the point-condensation phenomenon occurs if we consider the normalized least-action solitary waves.
©2009 Elsevier Masson SAS. All rights reserved.
1. Introduction
1.1. Background
The Born–Infeld geometric theory of electromagnetism is a nonlinear generalization of the classical Maxwell theory. It was introduced to overcome the infinite energy problem associated with a point-charge source in the original Maxwell theory. Based on the action principle of special relativity, Born proposed thefirstBorn–Infeld theory (cf. [2]
and [3]). It is defined by the action density LBI,1=b2
1−
1+FμνFμν 2b2
−det(gμν) (1.1)
wheregμν is the metric tensor of a(3+1)-dimensional Minkowskian space–time of the signature(+ − − −), and Fμν =∂μAν −∂νAμ is the field strength curvature induced from a gauge potential (connection 1-form)Aμ. Later, by considering the invariance principle, Born and Infeld reconsidered (1.1) and introduced thesecondBorn–Infeld theory in [8] and [9], which is defined by the action density
LBI,2=b2
−det(gμν)−
−det
gμν+Fμν
b . (1.2)
E-mail address:yuyong@cims.nyu.edu.
0294-1449/$ – see front matter ©2009 Elsevier Masson SAS. All rights reserved.
doi:10.1016/j.anihpc.2009.11.001
It is clear that when the Born–Infeld free parameter b→ ∞, both (1.1) and (1.2) reduce to the classical Maxwell theory.
In recent years, the Born–Infeld nonlinear electromagnetism has regained its importance due to its relevance in the theory of superstrings and membranes (cf. [29]). It has received much attention from both theoretic physicists and mathematicians (cf. [32,27,1,18,34,24]). Mathematically, motivated from Gibbons’ work (cf. [18]), in [34], Yang proposed an extended Born–Infeld equation, which can be used to unify the minimal surface equations and maximal hypersurface equations. Meanwhile, the author studied the existence of magnetostatic minimum-energy solutions in the Born–Infeld–Higgs model. Later, in Lin and Yang’s work (cf. [24]), the authors studied the gauged harmonic maps by extending the scalar Higgs fields in [34] to maps from a 2-surface into the standard 2-sphere. The coexistence of vortices and antivortices were obtained by studying the first-order system of self-dual and anti-self-dual equations.
The existence of cosmic strings induced by these vortices were also established.
1.2. Electrostatic solitary wave ansatz
As we know, the gauge potentialAμcan be coupled to a complex order parameterψthrough the minimal coupling rule. That is the formal substitution
∂
∂t → ∂
∂t −iA0,
∇ → ∇ −iA
where A=(A1, A2, A3) is a magnetic vector potential, and A0 is an electric potential. Therefore, in a flat Minkowskian space–time with metric(gμν)=diag[1,−1,−1,−1], we can define the Klein–Gordon–Maxwell La- grangian densityLKGMas
LKGM=1 2
∂ψ
∂t −iψ A0
2− |∇ψ−iAψ|2−m2|ψ|2 +1
p|ψ|p (1.3)
wherem is the mass of a particle, andp∈(2,6) is a constant describing the nonlinearity in (1.3). The nonlinear Born–Infeld–Klein–Gordon equations (NBIKG for short) are the Euler–Lagrange equations of the total action
S=
LBI,1+LKGM. (1.4)
Under the electrostatic solitary wave ansatz
ψ (x, t )=u(x)eiωt, A0= −φ(x), A=0
whereuandφare real-valued functions defined on a subset U ofR3, andωis a positive frequency parameter, the total action in (1.4) takes the form
FU(u, φ):=
U
1
2|∇u|2+1 2
m2−ω2 u2−1
p|u|pdx−Eu,U(φ) (1.5)
with
Eu,U(φ):=
U
b2
1−
1− 1 b2|∇φ|2
+ωu2φ+1
2φ2u2. (1.6)
The critical point (φ,u) ofFU satisfies the Euler–Lagrange equations associated with (1.5). By standard calculations, we get
⎧⎪
⎪⎨
⎪⎪
⎩
∇ · ∇φ
1−b12|∇φ|2=u2(ω+φ), u=
m2−(ω+φ)2
u− |u|p−2u.
(1.7)
In this article,Uis assumed to be a bounded smooth domain inR3or the wholeR3space. WhenU=R3, we omit the subscriptionUfrom (1.5) and (1.6). That is, we useF andEuto denoteFR3 andEu,R3, respectively. As a convention in this paper, in an integral expression, if its integration domain isR3, we will omitR3from the integral expression.
1.3. Main results
The existence of solitary waves has been well studied in different systems (cf. [4–7,10–13,16,22,25]). Particularly in [13] and [25], the authors considered the system of nonlinear Klein–Gordon equation coupled with the Born–Infeld type equations. Here, the Born–Infeld type equations refer to the second order expansion of the original Born–Infeld equations when the Born–Infeld parameterb→ ∞. In this article, we study the original Born–Infeld equations.
Firstly, for any fixedu∈L2(U ), we study the variational problem
φmin∈MEu,U(φ). (1.8)
The configuration space M:=D(U )∩
φ∇φL∞(U )b
(1.9) naturally arises from the physical constraint (quantity under the square root in (1.6) should be nonnegative) and the finite energy condition (Eu,U(φ) <∞) of (1.6). In (1.9),D(U )denotes the completion ofC0∞(U;R)with respect to the norm
φD(U ):= ∇φL2(U )+ ∇φL4(U ).
It is embedded intoL∞(U )continuously (cf. Proposition 8 in [17]). Therefore,Mis a topological space by equipping with the uniform norm topology whenU is a bounded smooth domain or the locally uniform norm topology when U=R3. The existence of global minimizer for the variational problem (1.8) will be studied in Section 2.1 by an application of the direct method in the Calculus of Variations. The minimizer is unique because of the convexity of the functionalEu,U. Therefore, we can define a nonlinear operator
Φ:L2(U )→M, (1.10)
which sends one L2(U ) functionu to the unique minimizer of the variational problem (1.8). In Section 2.3, the operatorΦis proved to be a continuous map betweenL2(U )andMif we equipL2(U )with its strong topology.
Plug the minimizerΦ(u)into (1.5), the functional J[u] :=FU
u, Φ(u)
, (1.11)
which is defined onH1(U )or a subspace ofH1(U ), is strongly indefinite. In Proposition 3.1, the functionalJ will be proved to beC1differentiable in the sense of Fréchet. We emphasize here that, with the loss of theC1regularity onΦ, our method relies only on the continuity of the operatorΦ. In Section 3.2, we will assumeU=R3. By theZ2
Mountain Pass Theorem, we will prove the existence of infinite many critical points ofJ with radial symmetry when κ=m
ω (1.12)
is suitably large (see Theorem 3.3). Furthermore, we will show that among all nonzero critical points ofJwith radial symmetry, there exists one that attains the least-J-action (see Theorem 3.8). In Section 3.3,Uis a bounded smooth domain. We will study the positive critical points of the functionalJ with boundary values set to be 0. Therefore, we work on the functional
J+[u] :=
U
1
2|∇u|2+1 2
m2−ω2 u2− 1
pup+dx−Eu,U Φ(u)
. (1.13)
J+ is defined onH01(U ). Based on the well-known Mountain Pass Lemma due to Ambrosetti and Rabinowitz, we will show in Theorem 3.18 and Theorem 3.19 that ifκ is suitably large, then among all nonzero critical points ofJ+, there exists one that attains the least-J-action. We call it the positive least-J-action critical point ofJ. Compared to the magnetostatic minimum-energy solutions for the Born–Infeld–Higgs model or the gauged harmonic map model in [34] and [24], our results in Theorems 3.3, 3.8, 3.18 and 3.19 verify the existence of the least-action solutions in the context of electrostatic fields. Moreover, in Theorem 3.18, we generalize the existence of the Mountain Pass solution in [26] and [33] to a coupled system combining a semilinear elliptic equation with an electric potential governed by the Born–Infeld theory.
Denote uκ,ω by one positive least-J-action critical point of the functional J+. With respect to the parameters κ and ω, in Section 4, we focus on the asymptotic behaviors and profiles of the functions in relation to uκ,ω andΦω(uκ,ω). We adopt a singular perturbation method. In fact, in Section 4.3, we will prove the point-condensation phenomenon for the normalized functionvκ,ω=ω2−2puκ,ω. More precisely, it will be proved that whenκis a suitably large constant, in the limit ofω→ ∞,vκ,ωadmits only one local maximum pointPκ,ω. In particular,
vκ,ω(Pκ,ω)
κ2−1 1
p−2 (1.14)
is uniformly bounded from below by a constant depending only onκ. Meanwhile, the normalized functionvκ,ω→0 inCloc1,α(U\Pκ,ω).
2. Minimizer ofEu,U
2.1. Existence of the minimizerΦ(u)
DenoteD1,2(U )by the completion ofC0∞(U;R)with respect to the norm φD1,2(U ):= ∇φL2(U ).
D(U )is continuously embedded into bothD1,2(U )andL∞(U ). Moreover,D1,2(U )is continuously embedded into L6(U )by Sobolev inequality.
In the following, we begin to study the variational problem (1.8) in this section. First, let us state a lemma.
Lemma 2.1.If{φn} ⊆M andφn φinD1,2(U ), thenφn→φuniformly inU¯ ifUis a bounded smooth domain;
φn→φlocally uniformly inR3, ifU=R3.
Proof. Assume that{φnk}is any subsequence of{φn}. Since{φnk} ⊆Mandφnk φ weakly inD1,2(U ),{∇φnk}is uniformly bounded inL2(U )andL4(U ). IfUis a bounded smooth domain, by Morrey’s inequality, we then have
φnk
C0,14(U )¯ CφnkD(U ),
which implies that {φnk}is equicontinuous on U. Apply Arzelá–Ascoli’s theorem, we can extract a subsequence,¯ which is denoted by{φnkl}, such thatφnkl ⇒φinU. Because the subsequence¯ {φnk}is arbitrary, we haveφn⇒φ inU.¯ R3case is similar. We omit the proof. 2
In the following, we use the direct method in the Calculus of Variations (cf. Theorem 1.2 in [31]) to show that Theorem 2.2.For everyu∈L2(U ), there exists a uniqueΦ(u)∈Msuch that
Eu,U Φ(u)
= inf
φ∈MEu,U(φ).
In fact, we need to verify that:
(1) Mis a weakly closed subset ofD1,2(U );
(2) Eu,U is coercive and weakly sequentially lower semicontinuous onMwith respect toD1,2(U ).
Lemma 2.3.Mis a weakly closed subset ofD1,2(U ).
Proof. Note that,D1,2(U )is reflexive andMis convex. We only need to prove the strong closedness ofMinD1,2(U ).
Choose{φn} ⊆M,φn→φinD1,2(U ). Up to a subsequence, we can assume∇φn→ ∇φalmost everywhere inU.
Then∇φL∞(U )b. We also have
U
|∇φn− ∇φ|44b2
U
|∇φn− ∇φ|2→0.
Hence,φ∈D(U ). Under the definition ofMin (1.9),φ∈M. 2 Pay attention to the fact that
F (x, φ, p)=b2
1−
1− 1 b2|p|2
+ωu2(x)φ+1
2u2(x)φ2−1
2ω2u2(x)
is convex inp. Then,F is a Caratheodory function. In the following, we apply Theorem 1.6 in [31] and show that Lemma 2.4.Eu,U(·)is coercive and weakly sequentially lower semicontinuous with respect toD1,2(U ).
Proof. The coercivity ofEu,U with respect toD1,2(U )is a result of the fact that Eu,U(φ)
U
1
2|∇φ|2dx−1 2ω2
U
u2dx.
Let us now prove the weakly sequentially lower semicontinuity ofEu,U. Assume that{φn} ⊆Mandφn φweakly inD1,2(U ). From Lemma 2.1, we know thatφn⇒φlocally inU. That is,∀UU, we haveφn⇒φinU. Hence, φn→φinL1(U). Since∇φn∇φweakly inL2(U ), we get∇φn∇φweakly inL1(U). Apply Theorem 1.6 in [31],
Eu,U(φ)lim inf
n→∞ Eu,U(φn). 2
Remark 2.5.Use the same method as the above, we can prove that if{φn} ⊆Mandφn φweakly inD1,2(U ), then
U
b2
1−
1− 1 b2|∇φ|2
lim inf
n→∞
U
b2
1−
1− 1 b2|∇φn|2
.
Now we complete this section by the proof of Theorem 2.2.
Proof of Theorem 2.2. Lemma 2.3 and Lemma 2.4 together with Theorem 1.2 in [31] imply that there exists a minimizer ofEu,U onM. The convexity of the functionalEu,U ensures the uniqueness of the minimizer. 2 2.2. Some properties of the operatorΦ
From Theorem 2.2, we can construct the operatorΦ, which is defined as in (1.10). Since for a fixedu∈L2(U ), Φ(u)is the unique minimizer of the variational problem (1.8), then
Proposition 2.6.∀u∈L2(U )fixed,Φ=Φ(u), we have
U
|∇Φ|2
1−b12|∇Φ|2
U
−u2(ω+Φ)Φ. (2.1)
Proof. ∀λ∈(0,1),λΦ∈M. ConsiderEu,U(λΦ)as a function ofλ, then d
dλEu,U(λΦ)=
U
λ |∇Φ|2 1−b12|∇Φ|2λ2
+ωu2Φ+λu2Φ2. (2.2)
SinceΦattains the minimum ofEu,U inM, we have lim inf
λ→1−
d
dλEu,U(λΦ)0. (2.3)
Apply Fatou’s Lemma, (2.2) and (2.3) may imply (2.1). 2
SinceD(U )is embedded intoL∞(U )continuously, then(ω+Φ)Φ∈L∞(U ). By the assumption thatu∈L2(U ), we can conclude from Proposition 2.6 that the minimizerΦ=Φ(u)of the variational problem (1.8) satisfies
U
|∇Φ|2
1−b12|∇Φ|2<+∞. Therefore, we can defineK⊆Mas
K:=
ψ∈M
U
|∇ψ|2
1−b12|∇ψ|2
<+∞
. (2.4)
The minimizerΦ=Φ(u)of the variational problem (1.8) is also the minimizer of the problem to minimize the energy functionalEu,U onK. Therefore, we may get the following proposition.
Proposition 2.7(Variational Inequality).∀u∈L2(U )fixed,Φ=Φ(u)is the minimizer ofEu,UonM. Then,∀ψ∈K,
U
∇Φ
1−b12|∇Φ|2· ∇(Φ−ψ )+(ω+Φ)(Φ−ψ )u20. (2.5)
Proof. ∀ψ∈Kand 0< λ <1,λΦ+(1−λ)ψ∈K. We know thatΦ(u)attains the minimum ofEu,U onK, thus, lim inf
λ→1−
d dλEu,U
λΦ+(1−λ)ψ 0.
Apply Lebesgue’s Dominated Convergence theorem. We complete the proof. 2
As an application of the variational inequality (2.5), we prove the a priori upper and lower bounds ofΦ(u)for a fixedu∈L2(U ).
Proposition 2.8.∀u∈L2(U ), it results inΦ(u)0. Moreover,Φ(u)(x)−ωifu(x)=0.
Proof. Apply the variational inequality (2.5) and setψ= −Φ−, we have
U
∇Φ
1−b12|∇Φ|2· ∇
Φ+Φ− +ω
Φ+Φ− u2+
Φ+Φ−
Φu20.
That is,
Φ0
∇Φ+
1−b12|∇Φ+|2· ∇Φ++ωΦ+u2+Φ+Φ+u20.
So we get∇Φ+=0. Hence,Φ0. If we setψ= −ω+(ω+Φ)+, then we have
Φ−ω
|∇Φ|2
1−b12|∇Φ|2+u2(Φ+ω)20. (2.6)
Hence,Φ(u)(x)−ωwheneveru(x)=0. 2
In the rest of this section, we assume thatU=R3. Letf be a function defined onR3. With respect to fixedx0∈R3 andg∈O(3), we can define translation and rotation onf as
Tx0f (x):=f (x+x0), Tgf (x):=f (gx), ∀x∈R3. (2.7) From the definition ofK in (2.4), it is easy to prove thatK is translation–rotation invariant. Pay attention to the uniqueness result in Theorem 2.2. We prove the fact thatΦ is even and commutes with the group of rototranslations in Proposition 2.9 below.
Proposition 2.9.∀u∈L2(R3),∀g∈O(3),∀x0∈R3, we have
Φ(Tx0u)=Tx0Φ(u), (2.8)
Φ(Tgu)=TgΦ(u), (2.9)
Φ(u)=Φ(−u). (2.10)
Proof. We know that Eu
Φ(u)
=ETx
0u
Tx0Φ(u) .
SinceK is translation–rotation invariant, we haveTx0Φ(u)∈K. Note that,Φ(Tx0u)attains the minimum ofETx
0u
onM. Then, Eu
Φ(u) ETx
0u
Φ(Tx0u) . Sincex0is arbitrary, we conclude that
Eu Φ(u)
=ETx
0u
Φ(Tx0u)
=ETx
0u
Tx0Φ(u) .
The uniqueness result in Theorem 2.2 immediately implies (2.8). The proofs for (2.9) and (2.10) are similar. We omit the discussions here. 2
As an application of (2.8) in Proposition 2.9 and the variational inequality (2.5) in Proposition 2.7, let us consider aW2,2estimate on the minimizerΦ(u).
Lemma 2.10. If ∇Φ(u)∈ Lp1(R3), ∇u∈L2(R3)∩Lp2(R3) and u∈L2(R3)∩Lp3(R3) where 1pi ∞ (i=1,2,3)satisfying p1
1 +p12 +p13 =1, then
i,j
∂ijΦ(u)26ω∇Φ(u)
Lp1(R3)∇uLp2(R3)uLp3(R3). (2.11) Proof. Assume that Φ =Φ(u), Φ1, :=Φ(x+e1) and u1, :=u(x+e1) where e1=(1,0,0). Since K is translation–rotation invariant,Φ1,∈K. By Proposition 2.7, we have
∇Φ
1−b12|∇Φ|2· ∇(Φ−Φ1,)+ω(Φ−Φ1,)u2+(Φ−Φ1,)Φu20. (2.12) From (2.8) in Proposition 2.9, we know thatΦ1,=Φ(u(· +e1)). Apply Proposition 2.7 again, we have
∇Φ1,
1−b12|∇Φ1,|2· ∇(Φ1,−Φ)+ω(Φ1,−Φ)u21,+(Φ1,−Φ)Φ1,u21,0. (2.13) By adding the inequalities (2.12) and (2.13), we have
∇Φ1,−Φ
2+
ω+Φ(u)Φ1,−Φ
u2(x+e1)−u2(x)
0.
Therefore,
∇Φ1,−Φ
2+2u
ω+Φ(u)Φ1,−Φ
u(x+e1)−u(x)
+ ω+Φ(u)Φ1,−Φ
u(x+e1)−u(x)
u(x+e1)−u(x)
0. (2.14)
Since we know that if 1p∞, Φ1,−Φ
Lp(R3)
∇ΦLp(R3). In particular, ifp= ∞,
Φ1,−Φ
L∞(R3)
b. (2.15)
Then, we know that
ω+Φ(u)Φ1,−Φ
u(x+e1)−u(x)
2
bω+Φ(u)
L∞(R3)∇u2L2(R3). Therefore, from (2.14) and Proposition 2.8, we have
∇Φ1,−Φ
22
ω+Φ(u)Φ1,−Φ
u(x+e1)−u(x)
u
+O() 2ω
Φ1,−Φ
u(x+e1)−u(x)
u
+O(). (2.16)
By applying Hölder’s inequality on (2.16) and letting→0, we get ∇∂1Φ(u)22ω∇ΦLp1(R3)∇uLp2(R3)uLp3(R3)
where 1pi∞(i=1,2,3)satisfying p1
1 +p12 +p13 =1. 2
Apply Lemma 2.10 by settingp1= ∞,p2=p3=2. The following proposition holds:
Proposition 2.11.Ifu∈H1(R3), then∇Φ(u)∈H1(R3;R3), which satisfies
i,j
∂ijΦ(u)26ωb∇uL2(R3)uL2(R3). (2.17)
2.3. Continuity of the operatorΦ
In this section, we study the continuity of the operator Φ in two cases. In Proposition 2.12, we prove that Φ:L2(U )→M is a continuous map if we equipL2(U )with the strongL2(U )topology. If we restrict the opera- torΦon
L:=L2(U )∩L125(U )∩L3(U ) (2.18)
and equip L with the strong L125(U )∩L3(U ) topology, then in Proposition 2.13, we show that Φ :L→M is a continuous map. Here,un→uunder the strongL125(U )∩L3(U )topology means thatun→uin bothL125(U )and L3(U ) strongly. As before,M is a topological space equipped with the uniform norm topology ifU is a bounded smooth domain or with the locally uniform norm topology ifU=R3.
Proposition 2.12.Φ:L2(U )→Mis a continuous map.
Proof. Assume thatun→ustrongly inL2(U ). In the following, we show that∀{unk} ⊆ {un}, there exists a subse- quence{unkl}, such that whenUis a bounded smooth domain,Φ(unkl)⇒Φ(u)inU¯; whenU=R3,Φ(unkl)⇒Φ(u) locally inR3.
Note that,{unk}is uniformly bounded inL2(U ). By Propositions 2.6 and 2.8,{Φ(unk)}is uniformly bounded in D(U ). Hence, it is also uniformly bounded inL∞(U ). Then, there exist a subsequence{un
kl}andf ∈M, such that Φ(unkl) f inD1,2(U ). By Lemma 2.1,Φ(unkl)⇒f inU¯ whenUis a bounded smooth domain orΦ(unkl)⇒f locally inR3whenU=R3. For the rest of the proof, we only need to show thatf =Φ(u).
BecauseΦ(un
kl)is the minimizer ofEunk
l,U onMandun
kl →uinL2(U ), we have Eunk
l,U
Φ(un
kl) Eunk
l,U
Φ(u)
→Eu,U Φ(u)
. Note that,{Φ(un
kl)}is uniformly bounded inL∞(U )and converges tof pointwisely. Whenun
kl→uinL2(U ), we have
U
Φ(unkl)u2n
kl→
U
f u2,
U
Φ(unkl)2u2n
kl→
U
f2u2. (2.19)
From Remark 2.5 and (2.19), we get Eu,U(f )lim inf
l→∞ Eunk
l,U
Φ(unkl) lim
l→∞Eunk
l,U
Φ(u)
=Eu,U
Φ(u) . By the uniqueness result in Theorem 2.2,f=Φ(u). 2
Proposition 2.13.Φ:L→Mis a continuous map.Lis defined in(2.18).
Proof. Assume thatun→uinL. WhenU is a bounded smooth domain, the result has already been included in Proposition 2.12. Now, we assumeU=R3. Notice Proposition 2.8 and apply Hölder’s inequality on the right-hand side of (2.1). We have
∇Φ(un)2+ 1
2b2∇Φ(un)4ωun2
L125(R3)
Φ(un)
L6(R3). (2.20)
Sinceun→uinL, we have{un}uniformly bounded in L125(R3). From (2.20),{Φ(un)}is uniformly bounded in D(R3). Hence, it is also uniformly bounded inL∞(R3)andL6(R3). Then, there exist a subsequence, still denoted by {Φ(un)}, andf ∈M, such thatΦ(un) f inD1,2(R3). Apply Lemma 2.1,Φ(un)⇒f locally inR3. For the rest of the proof, we only need to show thatf =Φ(u).
Note that,
u2nΦ(u)−u2Φ(u)
u2n−u2
L65
Φ(u)
L6C∇Φ(u)
L2un−u
L125 →0, (2.21)
u2nΦ(u)2−u2Φ(u)2
u2n−u2
L32
Φ(u)2
L6C∇Φ(u)2
L2un−uL3→0. (2.22) Hence, we have
nlim→∞Eun Φ(u)
=Eu Φ(u)
. SinceEun(Φ(un))Eun(Φ(u)), then
lim sup
n
Eun Φ(un)
Eu Φ(u)
. (2.23)
In another way, note that,{Φ(un)}is uniformly bounded inL6(R3),L∞(R3)and converges tof pointwisely. When un→uinL, we can apply the Lebesgue’s Dominated Convergence theorem and similar arguments as in (2.21) and (2.22) to show that
u2nΦ(un)→
u2f,
u2nΦ(un)2→
u2f2. (2.24)
Because of Remark 2.5 and (2.24), we have lim inf
n Eun
Φ(un)
Eu(f ). (2.25)
Obviously, (2.23) and (2.25) imply thatEu(f )=Eu(Φ(u)). Therefore, we havef =Φ(u)by the uniqueness result in Theorem 2.2. 2
Remark 2.14.Ifun→uinL2(U )orL, then
nlim→∞Eun,U Φ(un)
=Eu,U Φ(u)
. (2.26)
3. Existence of solitary wave solutions
In this section, we prove the existence of critical points ofJ andJ+. Firstly, let us study theC1differentiability of J andJ+in the sense of Fréchet.
3.1. Differentiability ofJ andJ+
In this section, the functionalJ is defined onH, which is a subspace ofH1(U ). Naturally,His endowed with the metric fromH1(U ). The main purpose of this section is to prove
Proposition 3.1.J∈C1(H;R)in the sense of Fréchet.
Proof. It is sufficient to prove
vlimH1→0
J[u+v] −J[u] −DJ[u]v
/vH1=0 where
DJ[u]v:=
U
∇u· ∇v+ m2−
ω+Φ(u)2
uv− |u|p−2uv. (3.1)
We splitJ[u+v] −J[u] −DJ[u]vinto three parts. That is, J[u+v] −J[u] −DJ[u]v=A+B+C
where A=
U
1
2∇(u+v)2−1
2|∇u|2− ∇u· ∇v, B= −1
p
U
|u+v|p− |u|p−p|u|p−2uv,
C=Eu, U Φ(u)
−Eu+v,U
Φ(u+v) +
U
1 2
m2−ω2 v2+
2ω+Φ(u)
Φ(u)uv.
By Sobolev embedding theorem, it is easy to prove A, B=o(vH1). We thus focus on the proof of C=o(vH1).
In the following, we will prove Co(vH1)and Co(vH1)successively. Then we complete the proof of Propo- sition 3.1.
We first prove Co(vH1(R3)). SinceΦ(u+v)minimizesEu+v,UonM, we have CD:=Eu,U
Φ(u)
−Eu+v,U Φ(u)
+
U
1 2
m2−ω2 v2+
2ω+Φ(u)
Φ(u)uv.
Therefore, Co(vH1)by the fact that D=
U
1 2
m2−ω2
v2−ωΦ(u)v2−1
2Φ(u)2v2=o vH1
.
Let us now prove Co(vH1). SinceΦ(u)minimizesEu,U onM, CEu,U
Φ(u+v)
−Eu+v,U
Φ(u+v) +
U
2ω+Φ(u)
Φ(u)uv+1 2
m2−ω2 v2.
Set
Θ:=uv
Φ(u+v)−Φ(u)
2ω+Φ(u)+Φ(u+v) , we get
Co vH1
−
U
Θ. (3.2)
In the following, we prove
vlimH1→0
1 vH1
U
Θ
=0 (3.3)
in two cases, then from (3.2), we get Co(vH1).
Case 1(Bounded Smooth Domain).Under definition ofΘ, we have by applying Hölder’s inequality that 1
vH1
U
Θ
2ω+Φ(u)+Φ(u+v)
L∞(U )Φ(u+v)−Φ(u)
L∞(U )uL2(U ).
From Propositions 2.6, 2.8 and the fact thatD(U )is embedded intoL∞(U )continuously, we know that, asv→0 inH1(U ),2ω+Φ(u)+Φ(u+v)L∞(U )is uniformly bounded by a constant depending onb,ωanduL2(U ). By Proposition 2.12, we get (3.3) in the bounded smooth domain case.
Case 2(The whole Euclidean space R3). Similarly as in Case 1, whenv→0 in H1(R3),Φ(u+v)L∞(R3) is uniformly bounded by a constantC=C(b, ω,uL2(R3)). Sinceu∈L2(R3),∀ >0, there existsR >0 large enough, such thatuL2(BcR)< . Therefore, by Hölder’s inequality,
1 vH1(R3)
BcR
Θ
CuL2(BcR)< C. (3.4)
FixRabove. Since we know from Proposition 2.12 that ifv→0 inL2(R3), thenΦ(u+v)⇒Φ(u)inB¯R. We thus get
1 vH1(R3)
BR
Θ
CΦ(u+v)−Φ(u)
L∞(B¯R)uL2(R3)→0, asvH1(R3)→0. (3.5) By (3.4) and (3.5), we get (3.3) whenU=R3. 2
Remark 3.2.If we consider the functionalJ+defined as in (1.13), then by using the same method as in the proof of Proposition 3.1, we getJ+∈C1(H;R)in the sense of Fréchet.
3.2. Solitary wave solutions –U=R3
The functionalJ in (1.11) is defined onH1(R3). According to Proposition 3.1,J isC1differentiable in the sense of Fréchet. In this section, we study the existence of critical points ofJ by theZ2 Mountain Pass Theorem. More precisely, we prove
Theorem 3.3. If (p2 −1)m2> p2ω2, then in H1(R3), there exist infinitely many critical points of J with radial symmetry. Moreover, the critical points ofJ satisfy the equation
−u+ m2−
ω+Φ(u)2
u− |u|p−2u=0. (3.6)
Before we prove Theorem 3.3, let us state two lemmas, which are simple applications of (2.9) and (2.10).
Lemma 3.4.∀u∈H1(R3),J[u] =J[−u].
Lemma 3.5.∀u∈H1(R3),∀g∈O(3), we haveJ[Tgu] =J[u].
Lemma 3.4 and Lemma 3.5 imply that the functionalJ is even andTg invariant. Therefore, by the principle of symmetric criticality (cf. [28]), we may restrictJon the radially symmetric subspaceHr1(R3)⊂H1(R3). That is, Lemma 3.6.Ifu∈Hr1(R3)is a critical point ofJ|Hr1(R3), thenuis a critical point ofJ.
BecauseJ is invariant under translations, there is a lack of compactness onH1(R3). For this reason, we restrict J to the subspaceHr1(R3), which is a natural constraint forJ in the sense of Lemma 3.6. Then, there is no lack of compactness. Indeed, we have
Lemma 3.7.If(p2−1)m2>p2ω2, the functionalJ|Hr1(R3)satisfies the Palais–Smale condition.
After some slight modifications, the proof of Lemma 3.7 is almost the same as the proof of Lemma 6 in [13]. We omit the discussion here. Now we begin to prove Theorem 3.3.
Proof of Theorem 3.3. Jis even. According to Lemma 3.7, we know thatJ|Hr1satisfies the Palais–Smale condition.
Therefore, by Theorem 9.12 in [30], we only need to show thatJ|Hr1 satisfies the following two geometric hypothesis:
(G1) ∃ρ >0 andα >0 such thatJ[u]α,∀uwithuHr1=ρ;
(G2) for every finite dimensional subspaceV ofHr1,∃R=R(V ) >0 such thatJ[u]0,∀u∈V withuHr1R. Since we know that∀u∈H1(R3),
C(m, ω)u2H1−1
pupLpJ[u]C(m)u2H1−1 pupLp.
Sobolev inequality implies that uLpCuH1. This helps us to prove (G1). In finite dimensional subspace V of Hr1(R3), theLp norm andH1 norm are equivalent. Thus inV,uH1 C(V )uLp,∀u∈V. This helps us to complete the proof ofG2. 2
Theorem 3.3 implies the existence of infinitely many critical points of the functionalJ with the radial symmetry.
In Theorem 3.8 below, we prove the existence of least-J-action critical point among all nonzero critical points ofJ with the radial symmetry. Note thatJ is strongly indefinite. It is not bounded from above or below, we need to restrict our functionalJon the manifold
Σ:=
u∈Hr1
R3 u≡0, uis a critical point ofJ .
Theorem 3.8.If(p2−1)m2>p2ω2, then F := inf
u∈ΣJ[u] =min
u∈ΣJ[u]>0. (3.7)
Proof. First, let us proveF >0. Ifu∈Σ, then
|∇u|2+ m2−
ω+Φ(u)2
u2− |u|p=0. (3.8)
Apply Sobolev embedding theorem,
|u|p=
|∇u|2+ m2−
ω+Φ(u)2 u2C
|u|p 2
p
(3.9) whereC >0 is a constant depending onmandω. Sinceu=0, we have
|u|pCpp−2 >0. (3.10)
That is,Σkeeps strictly away from 0. By (3.8) and the fact thatEu(Φ(u))0, we have J[u] 1
2 −1 p
|∇u|2+ 1
2−1 p
m2−1
2ω2 u2Cu2LpC(p, ω, m) >0. (3.11) Hence,F >0.
If {un} ⊆Σ such that J[un] →F, then {J[un]} is bounded and from (3.11), {un} is uniformly bounded in Hr1(R3). Therefore, we can extract a subsequence, denoted also by {un}, such that un u0 in Hr1(R3). Since Hr1(R3) →Ls(R3)compactly withs∈(2,6), we can further assume thatun→u0inLs(R3)withs=125, 3, p.
By Proposition 2.13, we know thatΦ(un)⇒Φ(u0)locally inR3. Since
∇un· ∇φ+ m2−
ω+Φ(un)2 unφ=
|un|p−2unφ, ∀φ∈C0∞ R3
, letn→ ∞,
∇u0· ∇φ+ m2−
ω+Φ(u0)2 u0φ=
|u0|p−2u0φ.
That is,u0is a weak radial solution of (3.6). Regarding (3.10) andLp convergence of{un}, we imply thatu0≡0.
Hence,u0∈Σ. In addition, notice Remark 2.14. We have F J[u0] =
1
2|∇u0|2+1 2
m2−ω2 u20−1
p|u0|p−Eu0
Φ(u0)
lim inf
n
1
2|∇un|2+1 2
m2−ω2
u2n−lim
n
1
p|un|p−lim
n Eun Φ(un)
=F.
i.e.F can be attained by the functionu0∈Σ. 2 3.3. Solitary wave solutions – bounded smooth domain
In this section,U is a bounded smooth domain. We are interested in the existence of the positive least-J-action critical point ofJ. Therefore, we study the functionalJ+, which is defined onH01(U ). Due to Remark 3.2,J+isC1 differentiable in the sense of Fréchet.
Follow the notations as in [26]. Givene0,e≡0,e∈H01(U )withJ+[e] =0, we define C:= inf
h∈Γ max
0t1J+ h(t )
(3.12) whereΓ is the set of all continuous paths joining the origin and this givene. In addition, we define
M[v] :=sup
t0
gv(t ), ∀v∈H01(U ), v0, v≡0 (3.13)