7 The well-posedness of Cauchy problem 0 0 for dissipative modified Korteweg de 2 n Vries equations ∗ a J 2 Wengu Chen1, Junfeng Li1,2 and Changxing Miao1 2 1Institute of Applied Physics and Computational Mathematics ] P.O.Box 8009, Beijing 100088, China P 2College of Mathematics, Beijing Normal University, A Beijing 100875, China . h t a E-mail:[email protected], junfl[email protected], m miao [email protected] − [ 1 v Abstract. In this paper we consider some dissipative versions of 2 the modified Korteweg de Vries equation u +u +|D |αu+u2u = 0 0 t xxx x x 6 with0< α ≤ 3. Weprovesomewell-posednessresultsontheassociated 1 Cauchy problem in the Sobolev spaces Hs(R) for s> 1/4−α/4 on the 0 basis of the [k; Z]−multiplier norm estimate obtained by Tao in [9] for 7 0 KdV equation. / h 2000 Mathematics Subject Classification: 35Q53, 35A07. t a m 1 Introduction : v i TheXs,bspaces,asusedbyBeals,Bourgain,Kenig-Ponce-Vega, Klainerman- X Machedonandothers,arefundamentaltoolstostudythelow-regularity r a behavior of nonlinear dispersive equations. It is of particular interest to obtain bilinear or multilinear estimates involving these spaces. By Plancherel’s theorem and duality, these estimates reduce to estimat- ing a weighted convolution integral in terms of the L2 norms of the component functions. In [9], Tao systematically studied the weighted ∗The work was partly supported by NNSF of China (No.10371080) and China Postdoctoral Science Foundation (No.2005038327). 1 convolution estimates on L2. As a consequence, Tao obtained sharp bilinear estimates for the KdV, wave, and Schro¨dinger Xs,b spaces. These estimates are usable for many applications, and Tao in his paper presented some selected applications of these estimates to prove multi- linear estimates and local well-posedness results. In this paper, we will get some bilinear estimate on some new type of Xs,b spaces adapted to the dissipative version of the modified Korteweg de Vries equation and obtain new local well-posedness of associated Cauchy problem. This new type of Bourgain spaces was first introduced by Molinet and Rib- aud in [6]. We defer to introduce Tao’s estimates here. Recently, Molinet and Ribaud consider the Cauchy problem asso- ciated with dissipative Korteweg de Vries equations u +u +|D |αu+uu = 0, t ∈ R , x∈ R, t xxx x x + (1) (cid:26) u(0) = ϕ, where |D |α denotes the Fourier multiplier operator with symbol |ξ|α. x Whenα = 1/2, equation (1)modelstheevolution ofthefreesurfacefor shallow water waves damped by viscosity. When α = 2, equation (1) is the so-called Korteweg de Vries Burgers equation which described the propagation ofsmallamplitudelongwaves insomenonlineardispersive media when dissipative effects occur (see[8]). In [6], Molinet and Ribaud proved the global well-posedness of (1) for data in Hs(R), s > −3 for all α > 0. Especially when α = 2, 4 they proved the global well-posedness of Korteweg de Vries Burgers equation for data in Hs(R), s > −3 − 1 . The surprising part of this 4 24 result is that the indice s = −3 − 1 is lower than the best known 4 24 indice s = −3/4 obtained by Kenig, Ponce and Vega in [5] for the KdV equation and lower than the indice s = −1/2 of the critical Sobolev space for the dissipative Burgers equation u −u +uu = 0 (see [1], t xx x [4]). In [7], Molinet and Ribaud improved this result by introducing a new Bourgain type space and working in the space. They showed that Korteweg de Vries Burgers equation is globally well-posed in Hs(R) for s > −1 and in some sense ill-posed in Hs(R) for s < −1. The main purpose of this paper is to consider the Cauchy problem for the following dissipative versions of the mKdV equation on the real line u +u +|D |αu+u2u = 0, t ∈ R t xxx x x + (2) (cid:26) u(0) = ϕ 2 withα ∈(0, 3]. Andweprovelocalwell-posednessresultsontheassoci- ated Cauchy problemintheSobolev spaces Hs(R)for s > 1/4−α/4 on the basis of the [k; Z]−multiplier norm estimate obtained by Tao in [9] for KdV equation. We also get the global well-posedness for 1 < α≤ 3 when s > 1 − α. It is worth pointing out that the method used to 4 4 obtain the multilinear estimates by Molinet and Ribaud in [7] does not work when α is small enough. If we consider the case 0 < α ≤ 1, we can only get that problem (2) is local well-posed for s > 1 − α by run- 2 2 ning the approach of [7]. The basic ideas of processing the multilinear estimates on Xs,b spaces are using Cauchy-Schwarz inequality and the algebraic smoothing relation in [7]( resonance identity in [9]). In [9], Taoutilizesdyadicdecompositionandorthogonality beforeresortingto Cauchy-Schwarz. Theadvantages of dyadic decomposition and orthog- onality lead to a better estimate when the algebraic smooth relation brings little benefit. It is clear we get a better local well-posedness for the dissipative version of the mKdV equation than the local well-posedness of mKdV. This explicates that the dissipative term in (2) plays a key role for the low regularity of the equation. We are interested in finding out the smooth effect of the dissipative term. From the scaling analysis, the critical index of the mKdV equation is −1, while the critical index of 2 the equation with only dissipative term is 1− α. Thus we conjecture 2 thatthemeanvalue 1−α ofthetwocritical indicesisthecritical index 4 4 for our dissipative version of the mKdV. For instance, we consider the Korteweg de Vries Burgers equation, −1 is exactly the mean value of −3 the critical scaling index of KdV and −1 the critical scaling index 2 2 of Burgers equation. Unfortunately, we can not show the ill-posedness below the index 1 − α. Recently, the first two authors of this paper 4 4 [3] showed that when α = 2, (2) is ill-posedness for s < −1 in some 2 sense. A similar argument can also get the same ill-posedenss for our problem. 1.1 Notations For a Banach space X, we denote by k·k the norm in X. We will X use the Sobolev spaces Hs(R) and their homogeneous versions H˙s(R) equipped with the norms kuk = k(1 − ∆)s/2uk and kuk = Hs L2 H˙s 3 k|D|s/2uk . Recall that for λ > 0, L2 kf(λt)k ≤ (λ−1/2+λs−1/2)kf(t)k , kf(λt)k ∼ λs−1/2kf(t)k . (3) Hs Hs H˙s H˙s s,b We also consider the corresponding space-time Sobolev spaces H x,t endowed with the norm kuk2 = < ξ >2s< τ >2b |uˆ(ξ, τ)|2dξdτ, (4) Hs,b ZR2 where < · >=(1+|·|2)1/2. Let U(·) be the unitary group in Hs(R), s ∈ R, which defined the free evolution of the KdV equation, i.e. U(t) = exp itP(D ) , (5) x (cid:16) (cid:17) where P(D ) is the Fourier multiplier with symbol P(ξ) = ξ3. Since x the linear symbolof equation (2) is i(τ−ξ3)+|ξ|α, by analogy with the spacesintroducedbyBourgainin[2]forpurelydispersiveequationsand Molinet and Ribaud for KdV-Burgers equation, we define the function s,b space X endowed with the norm α kuk = k < i(τ −ξ3)+|ξ|α >b< ξ >s uˆ(ξ, τ)k . (6) Xαs,b L2(R2) So that kuk ∼ k <|τ −ξ3|+|ξ|α >b< ξ >s uˆ(ξ, τ)k . (7) Xαs,b L2(R2) Note that since F(U(−t)u)(ξ, τ) = F(u)(ξ, τ + ξ3), one can re- s,b express the norm of X as α kuk = k< iτ +|ξ|α >b< ξ >s uˆ(ξ, τ +ξ3)k Xαs,b L2(R2) = k< iτ +|ξ|α >b< ξ >s F(U(−t)u)(ξ, τ)k . (8) L2(R2) s,b For T ≥ 0, we consider the localized spaces X endowed with the α,T norm kuk = inf {kwk , w(t) = u(t)on[0, T]}. Xαs,,bT w∈Xαs,b Xαs,b Finally we denote by W(·) the semigroup associated with the free evo- lution of the equation (2), i.e. ∀t ≥ 0, F (W(t)ϕ)(ξ) = exp[−t|ξ|α+itξ3]ϕˆ(ξ), x 4 and we extend W(·) to a linear operator defined on the whole real axis by setting ∀t∈ R, F (W(t)ϕ)(ξ) = exp[−|t||ξ|α+itξ3]ϕˆ(ξ). (9) x 1.2 Main results We will mainly work on the integral formulation of (2), i.e. 1 t u(t) = W(t)ϕ− W(t−t′)∂ (u3(t′))dt′, t ≥ 0. (10) x 3Z 0 Actually, toprovethelocalexistenceresult,weshallapplyafixedpoint argument to the following truncated version of (10) u(t) = ψ(t) W(t)ϕ− χR+(t) tW(t−t′)∂ (ψ3(t′)u3(t′))dt′ , (11) h 3 Z0 x T i wheret ∈ Rand,inthesequelofthispaper,ψ isatimecut-offfunction satisfying ψ ∈ C∞(R), suppψ ⊂ [−2, 2], ψ ≡ 1on[−1, 1], 0 and ψ (·) = ψ(·/T). Indeed, if u solves (11) then uis a solution of (10) T on [0, T], T < 1. Let us first state our global well-posedness result on the real line. In this paper we restrict us with 0 < α ≤ 3. Theorem 1 Let ϕ ∈ Hs(R), s > 1/4 − α/4. Then there exist some T > 0 and a unique solution u of (10) in Z =C([0, T], Hs)∩Xs,1/2. (12) T α,T Moreover the map ϕ 7→ u is smooth from Hs(R) to Z . T If the dissipative term is strong enough, we can also get the global well-posedness. Theorem 2 Let ϕ ∈ Hs(R), s > 1/4 − α/4 and 1 < α ≤ 3. The existence time of the solution of (10) can be extended to infinity. And u belongs to C((0, +∞), H∞(R)). 5 Remark: We can not get the global well-posedness for the case 0 < α ≤ 1, since that in this case one need a higher order a priori estimate than L2. While in the case 1 < α ≤ 3, we need only the a priori estimate in L2. It will be a interesting question to find out a suitable a priori estimate for the dissipative vision of the mKdV and then extend the local well-posedness to the global well-posedness when the dissipative term is very weak. This paper is organized as follows. In Section 2 we prove linear s,1/2 estimates in the function space X and in Section 3 we introduce α Tao’s [k; Z]−multiplier norm estimate and derive some trilinear esti- mate for the nonlinear term ∂ (u3) from Tao’s estimate. For Section x 4, we consider the local well-posedness while the global well-posedness will be in Section 5. 2 Linear Estimates In this section we study the linear operator ψ(·)W(·) as well as the linear operator L defined by t L : f 7→ χR+(t)ψ(t)Z W(t−t′)f(t′)dt′. (13) 0 2.1 Linear estimate for the free term Lemma 1 Let s ∈R. There exists C > 0 such that kψ(t)W(t)ϕk ≤ Ckϕk , ∀ϕ∈ Hs(R). (14) Xαs,1/2 Hs 6 Proof. By definition of k·k , Xαs,1/2 kψ(t)W(t)ϕk Xαs,1/2 = < ξ >s< i(τ −ξ3)+|ξ|α >1/2 F ψ(t)e−|t||ξ|αeitξ3ϕˆ(ξ) (τ) t (cid:13) (cid:16) (cid:17) (cid:13)L2 (cid:13) (cid:13) ξ,τ (cid:13) (cid:13) = < ξ >s ϕˆ(ξ) < iτ +|ξ|α >1/2 F ψ(t)e−|t||ξ|α (τ) t (cid:13)(cid:13) (cid:13) (cid:16) (cid:17) (cid:13)L2τ(cid:13)(cid:13)L2 (cid:13) (cid:13) (cid:13) (cid:13) ξ (cid:13) (cid:13) (cid:13) (cid:13) ≤ C < ξ >s ϕˆ(ξ) < τ >1/2 F ψ(t)e−|t||ξ|α (τ) t (cid:13)(cid:13) (cid:13) (cid:16) (cid:17) (cid:13)L2τ(cid:13)(cid:13)L2 (cid:13) (cid:13) (cid:13) (cid:13) ξ (cid:13) (cid:13) (cid:13) (cid:13) +C < ξ >s+α/2 ϕˆ(ξ) F ψ(t)e−|t||ξ|α (τ) t (cid:13)(cid:13) (cid:13) (cid:16) (cid:17) (cid:13)L2τ(cid:13)(cid:13)L2 (cid:13) (cid:13) (cid:13) (cid:13) ξ (cid:13) (cid:13) ≤ Ckϕ(cid:13)k , (cid:13) (15) Hs where we used the fact kψ(t)e−|t||ξ|αkHb ≤ C < ξ >α2(2b−1), ∀0 ≤ b ≤ 1, t which can be obtained from (19) in [7]. 2.2 Linear estimates for the forcing term Lemma 2 For w ∈ S(R2) consider k defined on R by ξ eitτ −e−|t||ξ|α k (t)= ψ(t) wˆ(ξ, τ)dτ. (16) ξ ZR iτ +|ξ|α Then, it holds for all ξ ∈R 2 < iτ +|ξ|α >1/2 F (k ) t ξ (cid:13) (cid:13)L2(R) (cid:13) (cid:13) (cid:13) |wˆ(ξ, τ)| (cid:13) 2 |wˆ(ξ, τ)|2 ≤ C dτ + dτ (17) h(cid:16)ZR < iτ +|ξ|α > (cid:17) (cid:16)ZR < iτ +|ξ|α > (cid:17)i By a little modification of Proposition 2 in [7], we can obtain the proof of Lemma 2. Lemma 3 Let s ∈R. 7 a) There exists C > 0 such that for all υ ∈S(R2), t (cid:13)χR+(t)ψ(t)Z0 W(t−t′)υ(t′)dt′(cid:13)Xαs,1/2 (cid:13) (cid:13) (cid:13) (cid:13) |υˆ(ξ, τ +ξ3)| 1/2 ≤ C kυk + < ξ >2s ( dτ)2dξ (18) h Xαs,−1/2 (cid:16)Z Z < iτ +|ξ|α > (cid:17) i b) For any 0 < δ < 1/2 there exists C > 0 such that for all δ s,−1/2+δ υ ∈ X α t (cid:13)χR+(t)ψ(t)Z0 W(t−t′)υ(t′)dt′(cid:13)Xαs,1/2 ≤ CδkυkXαs,−1/2+δ. (19) (cid:13) (cid:13) (cid:13) (cid:13) Proof. Assume that υ ∈ S(R2). Taking the x−Fourier transform we get t χR+(t)ψ(t)Z W(t−t′)υ(t′)dt′ 0 t = U(t)hχR+(t)ψ(t)ZReixξZ0 e−|t−t′||ξ|αFx(U(−t′)υ(t′))(ξ)dt′dξi. Set ω(t′) = U(−t′)υ(t′). Writing F (ω)(ξ, t) with the help of its time x Fourier transform and using Fubini’s theorem one infers that t χR+(t)ψ(t)Z W(t−t′)υ(t′)dt′ 0 t = U(t)hχR+(t)ψ(t)ZR2eixξe−t|ξ|αωˆ(ξ, τ)Z0 eit′τet′|ξ|αdt′dξdτi eitτ −e−t|ξ|α = U(t)hχR+(t)ψ(t)ZR2eixξ iτ +|ξ|α ωˆ(ξ, τ)dξdτi. (20) We set k (t) = ψ(t) eitτ−e−|t||ξ|αωˆ(ξ, τ)dτ. Since ω(t) = U(−t)υ(t) ∈ ξ R iτ+|ξ|α S(R2), it is clear thRat for any fixed ξ ∈ R, k is continuous on R and ξ kξ(0) = 0. Then it is not too hard to derive that kχR+(t)kξ(t)kHb ≤ t kk (t)k , 0 ≤ b ≤ 1, and b 6= 1/2. The case b = 1/2 follows by ξ Hb t 8 Lebesgue dominated convergence theorem. Thus in view of (20) t (cid:13)χR+(t)ψ(t)Z0 W(t−t′)υ(t′)dt′(cid:13)Xαs,1/2 (cid:13) (cid:13) (cid:13) (cid:13) t = (cid:13)(cid:13) < iτ +|ξ|α >1/2< ξ >s Fx,t(cid:16)U(−t)χR+(t)ψ(t)Z0 W(t−t′)υ(t′)dt′(cid:17)(cid:13)(cid:13)L2ξ,τ (cid:13) t (cid:13) ≤ (cid:13)(cid:13)< ξ >s (cid:13)Fx(cid:16)U(−t)χR+(t)ψ(t)Z0 W(t−t′)υ(t′)dt′(cid:17)(cid:13)Hτ1/2(cid:13)(cid:13)L2 (cid:13) (cid:13) (cid:13) (cid:13) ξ (cid:13) (cid:13) t (cid:13) (cid:13) +(cid:13)(cid:13)< ξ >s+α/2 (cid:13)Fx(cid:16)U(−t)χR+(t)ψ(t)Z0 W(t−t′)υ(t′)dt′(cid:17)(cid:13)L2τ(cid:13)(cid:13)L2 (cid:13) (cid:13) (cid:13) (cid:13) ξ (cid:13) (cid:13) ≤ (cid:13) <(cid:13) ξ >s kχR+(t)kξ(t)kHt1/2(cid:13)L2 +(cid:13) < ξ >s+α/2 kχR+(t)kξ(t)kL2t(cid:13)(cid:13)L2 (cid:13) (cid:13) ξ (cid:13) (cid:13) ξ ≤ (cid:13) < ξ >s kk (t)k + (cid:13)< ξ >(cid:13)s+α/2 kk (t)k (cid:13) (cid:13) ξ Ht1/2(cid:13)L2 (cid:13) ξ L2t(cid:13)L2 (cid:13) (cid:13) ξ (cid:13) (cid:13) ξ (cid:13) (cid:13) (cid:13) (cid:13) ≤ C < ξ >s < iτ +|ξ|α >1/2 F (k (t) . t ξ (cid:13)(cid:13) (cid:13) (cid:13)L2τ(cid:13)(cid:13)L2 (cid:13) (cid:13) (cid:13) (cid:13) ξ (cid:13) (cid:13) (cid:13) (cid:13) Then (18) follows directly from Lemma 2 together with the last esti- mate. To prove (19) we first assume that υ ∈ S(R2). Applying Cauchy- Schwartz inequality in τ on the second term of the right-hand side of (18), one obtains (19) for υ ∈ S(R2). The result for υ ∈ Xs,−1/2+δ α follows by density. 3 Tao’s [k; Z]−multiplier norm estimate and its application In this section we introduceTao’s [k; Z]−multiplier norm estimate and derive the trilinear estimate needed to obtain the local existence result from Tao’s multiplier norm estimate for KdV equation. Let Z be any abelian additive group with an invariant measure dξ. For any integer k ≥ 2, we let Γ (Z) denote the hyperplane k Γ (Z) := {(ξ ,··· , ξ )∈ Zk : ξ +···+ξ =0} k 1 k 1 k 9 which is endowed with the measure f := f(ξ ,··· , ξ , −ξ −···−ξ )dξ ···dξ . 1 k−1 1 k−1 1 k−1 ZΓk(Z) ZZk−1 A [k; Z]−multiplier is defined to be any function m : Γ (Z) → C k which was introduced by Tao in [9]. And the multiplier norm kmk [k;Z] is defined to be the best constant such that the inequality k k m(ξ) f (ξ ) ≤ kmk kf k , (21) j j [k;Z] j L2(Z) (cid:12)(cid:12)ZΓk(Z) jY=1 (cid:12)(cid:12) jY=1 (cid:12) (cid:12) holds for all test functions f on Z. Tao systematically studied this j kind of weighted convolution estimates on L2 in [9]. To state Tao’s results, we use some notation he used in his paper. We use A . B to denote the statement that A ≤ CB for some large constant C which may vary from line to line and depend on various parameters, and similarly use A ≪ B to denote the statement A ≤ C−1B. We use A ∼ B to denote the statement that A . B . A. Any summations over capitalized variables such as N , L , H are j j presumed to be dyadic, i.e., these variables range over numbers of the form 2k for k ∈ Z. In this paper, we will only consider the [3; Z]- multiplier. Let N , N , N > 0. It will be convenient to define the 1 2 3 quantities N ≥ N ≥ N to be the maximum, median, and max med min minimum of N , N , N respectively. Similarly define L ≥ L ≥ 1 2 3 max med L whenever L , L , L > 0. And we also adopt the following sum- min 1 2 3 mation conventions. Any summation of the form L ∼ ··· is a sum max over the three dyadic variables L , L , L & 1, thus for instance 1 2 3 := . LmXax∼H L1,L2,L3X&1:Lmax∼H Similarly, any summation of the form N ∼ ··· sum over the three max dyadic variables N , N , N > 0, thus for instance 1 2 3 := . Nmax∼XNmed∼N N1,N2,N3>0:XNmax∼Nmed∼N If τ, ξ and h(·) are given, we also adopt the convention that λ is short- hand for λ := τ −h(ξ). 10