ebook img

The initial-boundary value problem for some quadratic nonlinear Schr\"odinger equations on the half-line PDF

0.31 MB·
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview The initial-boundary value problem for some quadratic nonlinear Schr\"odinger equations on the half-line

THE INITIAL-BOUNDARY VALUE PROBLEM FOR SOME QUADRATIC NONLINEAR SCHRO¨DINGER EQUATIONS ON THE HALF-LINE 6 1 0 MA´RCIO CAVALCANTE 2 c e Instituto de Matem´atica, Universidade Federal do Rio de Janeiro D Ilha do Fund˜ao, 21945-970, Rio de Janeiro RJ-Brazil 8 1 Abstract. Weprovelocalwell-posednessfortheinitial-boundaryvalueproblemassociated tosomequadraticnonlinearSchr¨odingerequationsonthehalf-line. Theresultsareobtained ] inthelowregularitysettingbyintroducingananalyticfamilyofboundaryforcingoperators, P following theideas developed in [14]. A . h t a m 1. Introduction [ 2 This paper is concerned with the initial-boundary value problem (IBVP) on the right v half-line for some quadratic nonlinear Schro¨dinger equations, namely 5 6 i∂ u ∂2u =N (u,u), (x,t) (0,+ ) (0,T), i = 1,2,3, 5 t − x i ∈ ∞ × 2 u(x,0) = u0(x), x (0,+ ), (1.1) ∈ ∞ 0 u(0,t) = f(t), t (0,T), 1. ∈ 0 where N (u,u)= u2, N (u,u) = u2 or N (u) = u2. 1 2 3 6 | | The appropriate spaces for the initial and boundary data is motivated by the behavior of 1 v: thesolutions forlinear Schro¨dingerequation inR. Lete−it∂x2 thelinear homogeneous solution i group in R for the Schro¨dinger equation. The following smoothing effect X ar ke−it∂x2φkL∞x H˙ 2s4+1(Rt) ≤ ckφkH˙s(R), can be found in [16], where this inequality is sharp in the sense that 2s+1 cannot be replaced 4 by any higher number. We are thus motivated to consider the IBVP (1.1) in the setting u0 Hs(R+) and f(t) H2s4+1(R+). (1.2) ∈ ∈ Ourgoal instudying(1.1)-(1.2)is toobtain low regularity results. Thenweconsiders 0, ≤ where compatibility conditions between the values u(0) and f(0) are not required. AMS Subject Classifications: 35Q55. E-mail address: [email protected], [email protected]. 1 2 IBVP FOR SOME QUADRATIC NONLINEAR SCHRO¨DINGER EQUATIONS ON THE HALF-LINE The motivation to study the IBVP (1.1)-(1.2) come from the associated initial-value prob- lem (IVP) in R, that is, i∂ u ∂2u= N(u,u), (x,t) R (0,T), t − x ∈ × (1.3) (u(x,0) = u0(x), x R. ∈ The IVP (1.3) has been studied by several authors of the last decades (see e.g. the mono- graph by Linares and Ponce [19]). Here we mainly discuss the principal results for quadratic nonlinearities. Cazenave and Weissler [6] and Tsutsumi [20] obtained local well-posedness (LWP), i.e., existence, uniqueness and continuity of the data-to-solution map, in Hs(R) for s 0, for any quadratic nonlinear term, i.e., N(u,u) cu2. The proof is based on the ≥ | | ≤ | | version of the Strichartz estimate for the free Schro¨dinger group e−it∂x2 found in [11]. Kenig, Ponce and Vega [17] showed LWP in Hs(R) for s ( 3,0] for the nonlinear terms u2 or u2, ∈ −4 and s ( 1,0] for nonlinearity u 2, by using Fourier restriction norm method introduced by ∈ −4 | | Bourgain in [3]. Bejenaru and Tao [1] showed LWP in Hs(R) for s 1 when N(u,u) = u2, ≥ − by an iteration using a modification of the standard Bourgain spaces. These authors also proved local ill-posedness (LIP) for s < 1, in the sense that the data-to-solution map fails to be continuous. In [18] Kishimoto obta−ined LWP in Hs(R) for s 1 and LIP for s < 1 ≥ − − when N(u,u) = u2. The solutions we construct to the IBVP (1.1)-(1.2) shall have the following properties. Definition 1.1. A function u(x,t) will be called a distributional solution of (1.1) on [0,T] if (a) Well-defined nonlinearity: ubelongs tosome space X withthe property that u X ∈ implies N (u,u), i= 1,2,3 is a well-defined distribution; i (b) u(x,t) satisfies the equation (1.1) in the sense of distributions on the set (x,t) ∈ (0,+ ) (0,T); (c) Spac∞e tr×aces: u C [0,T]; Hs(R+) and in this sense u(,0) = u ; ∈ x · 0 (d) Time traces: u∈ C(cid:0)R+x; H2s4+1(0,T(cid:1)) and in this sense u(0,·) = f. In our case, X shall be th(cid:0)e restriction on R(cid:1)+ (0,T) of the functions belonging to the × Space Xs,b C Rt; Hs(Rx) C Rx; H2s4+1(Rt) , where Xs,b is the Bourgain space, where ∩ ∩ (cid:0) (cid:1) (cid:0) (cid:1) 1 2 u = ξ 2s τ ξ2 2b uˆ(ξ,τ)2dξdτ . k kXs,b h i h − i | | (cid:18)Z Z (cid:19) The space Xs,b, with b > 1, is typically employed in the study of the IVP (1.3). As in the 2 work of Colliander and Kenig [7], to handle with the Duhamel boundary forcing operator in our study of the IBVP, we need to take b < 1, force us to intercept Xs,b with the space 2 C R ; Hs(R ) , because for b < 1, Xs,b is not embedded into C R ; Hs(R ) . An important t x 2 t x point here is that for b > 1 the bilinear estimates in Bourgain Spaces Xs,b were already (cid:0) (cid:1) 2 (cid:0) (cid:1) obtained in [17], while here we need to prove these estimates in the case b < 1. 2 Our main results are summarized in the following theorem. Theorem 1.1. Let s = 3/4 for i = 1,3 and s = 1/4. For any initial-boundary data i 2 − − (u0,f) Hs(R+) H2s4+1(R+), with s (si,0], there exists a positive time Ti, depending only ∈ × ∈ IBVP FOR SOME QUADRATIC NONLINEAR SCHRO¨DINGER EQUATIONS ON THE HALF-LINE 3 of ku0kHs(R+) and kfkH2s4+1(R+), and a distributional solution ui(x,t) of the IBVP (1.1)-(1.2) with nonlinearity Ni(u,u), which is defined in C [0,T]; Hs(R+) C R+; H2s4+1(0,T) . ∩ Moreover, the map (u0,f) 7−→ ui from Hs(R+)(cid:0)× H2s4+1(R+) i(cid:1)nto C(cid:0)[0,Ti]; Hs(R+x) (cid:1)∩ C R+x; H2s4+1(0,T) is analytic. (cid:0) (cid:1) (cid:0)As regards the q(cid:1)uestion of regularity this result is similar of the classical results obtained by Kenig, Ponce and Vega in [17] for the pure IVP (1.3) on R. The proof of Theorem 1.1 involves the approach of Colliander and Kenig [7], in their treatment of the generalized Korteweg-de Vries on the half-line. In this work they introduced anewmethodtosolveIBVPfornonlineardispersivepartialdifferentialequationsbyrecasting these problems as IVP with an appropriate forcing term. We also use the ideas contained in Holmer’s works [13] and [14], in his study of nonlinear Schro¨dinger and Korteweg-de Vries (KdV) equations on the half-line, by adapting the method of [7]. The method consists in solving a forced IVP in R, analogous to the (1.1), i∂ u˜ ∂2u˜ = N(u˜,u˜)+D(x)h(t); (x,t) R (0,T), u˜(tx,0−) =xu˜ (x), x R. ∈ × (1.4) 0 (cid:26) ∈ where u˜ is a nice extension of u and D is a distribution supported in R . The boundary 0 0 − forcing function h(t) is selected to ensure that u˜(0,t) = f(t), t (0,T). (1.5) ∈ Upon constructing the solution u˜ of (1.4), we obtain a distributional solution of (1.1), by restriction, as u = u˜ R+ (0,T) . |{ × } The solution of (1.4) satisfying (1.5) is constructed using the classical restricted norm method of Bourgain (see [3] and [17]) and the inversion of a Riemann-Liouville fractional integration operator. The crucial point in this work is the study of the Duhamel boundary forcing operator of [13], t f(x,t) = 2e−iπ4 ei(t−t′)∂x2δ0(x) 1f(t′)dt′, L Z0 I−2 where is the the Riemann-Liouville fractional integral of order 1/2 given by 1 I−2 − 1 t f(t) = (t s) 1/2f (s)ds, for f C (R+), I−21 ′ Γ(1/2) Z0 − − ′ ∈ 0∞ in the context of Bourgain Spaces. In [13] the operator was studied in the context of L Strichartz estimates, to solve the IBVP associated to the classical nonlinear Schro¨dinger equation, with nonlinearity λ uα 1u. We will obtain the Bourgain spaces estimates, − | | ψ(t) f(x,t) c f , k L kXs,b ≤ k kH2s4+1(R+) 0 for 1 < s < 1, where ψ(t) is a smooth cutoff function. To extend this estimate for others −2 2 values of s we define an analytic family of boundary forcing operators, analogous to the one in [14] in the study of the IBVP for the KdV equation on the half-line, in the low regularity setting. This family of operator generalizes the single operator of [13]. L 4 IBVP FOR SOME QUADRATIC NONLINEAR SCHRO¨DINGER EQUATIONS ON THE HALF-LINE We have applied the same technical to study the IBVP associated to the Schro¨dinger- Korteweg-de Vries system on the half-line [8]. We do not explore here the uniqueness of solutions obtained in Theorem 1.1. Actually we have uniqueness in the sense of Kato (see [15]) only for a reformulation of the IBVP (1.1) as an integral equation posed in R, such that solves (1.1). As there are many ways to transform the IBVP (1.1) into an integral equation, we do not have uniqueness in the strong sense. We believe that the method given by Bona, Sun and Zhang in [4] can be applied here. These authors introduced the concept of mild solutions, to treat the question of uniqueness for the Korteweg-de-Vries equation on the half-line. In [5] the same authors applied this method to get uniqueness of the solutions for the classical nonlinear Schro¨dinger equation, with nonlinearity uλ 1u on the half-line and on the bounded interval. − | | Thispaperis organized as follows: in thenextsection, we discusssomenotation, introduce function spaces and recall some needed properties of these function spaces, and review the definition and basic properties of the Riemann-Liouville fractional integral. Sections 3, 4 and 5, respectively, are devoted to the needed estimates for the linear group, the Duhamel boundary forcing operators and the Duhamel inhomogeneous solution operator. In Section 6 we define the Duhamel boundary forcing operators class, adapted from [14], and prove the needed estimates for it. In Section 7, we prove the bilinear estimates. Finally, Section 8 is devoted to prove Theorem 1.1. 2. Notations, Function Spaces, Riemman-Liouville Fractional Integral 2.1. Notations. For φ = φ(x) S(R), φˆ(ξ) = e iξxφ(x)dx will denote the Fourier trans- − ∈ form of φ. For u = u(x,t) S(R2), uˆ = uˆ(ξ,τ) = e i(ξx+τt)u(x,t)dxdt will denote its ∈ R − space-time Fourier transform, u(ξ,t) denotes its space Fourier transform and u(x,τ) x t F R F its time Fourier transform. Let ξ = 1 + ξ . Define (τ i0)α as the limit, in the sense h i | | − of distributions, of (τ γi)α, when γ 0 . χ denotes the characteristic function of an − A − → arbitrary set A. 2.2. Function Spaces. For s 0 define φ Hs(R+) if exists φ˜ such that φ(x) = φ˜(x) for ≥ ∈ x > 0, in this case we set kφkHs(R+) = infφ˜kφ˜kHs(R). For s ≥ 0 define Hs(R+) = φ Hs(R+); supp φ [0, ) . 0 { ∈ ⊂ ∞ } For s < 0, define Hs(R+) as the dual space to H s(R+), and define Hs(R+) as the dual 0− 0 space to H s(R+). − Also definethe C (R+) = φ C (R); suppφ [0, ) and C (R+) as those members 0∞ { ∈ ∞ ⊂ ∞ } 0∞,c ofC (R+)withcompactsupport. WeremarkthatC (R+)isdenseinHs(R+)foralls R. 0∞ 0∞,c 0 ∈ Throughoutthe paper, we fix a cutoff function ψ C (R) such that ψ(t) =1 if t [ 1,1] ∈ 0∞ ∈ − and supp ψ [ 2,2]. ⊂ − The following lemmas state elementary properties of the Sobolev spaces. For the proofs we refer the reader [7]. Lemma 2.1. For −12 < s < 21 and f ∈ Hs(R), we have kχ(0,+∞)fkHs(R) ≤ ckfkHs(R). Lemma 2.2. If 0 ≤ s < 12, then kψfkHs(R) ≤ ckfkH˙s(R) and kψfkH˙−s(R) ≤ ckfkH−s(R) , where c only depends of s and ψ. IBVP FOR SOME QUADRATIC NONLINEAR SCHRO¨DINGER EQUATIONS ON THE HALF-LINE 5 Lemma 2.3. Let f ∈ H0s(R+), with −∞ < s < +∞. Then kψfkH0s(R+) ≤ ckfkH0s(R+). For s,b R, we introduce the classical Bourgain spaces Xs,b and the modified Bourgain spaces Ws,∈b related to the Schro¨dinger equation as the completion of S (R2) under the norms ′ 1 2 u = ξ 2s τ ξ2 2b uˆ(ξ,τ)2dξdτ k kXs,b h i h − i | | (cid:18)Z Z (cid:19) 1 2 u = τ s τ ξ2 2b uˆ(ξ,τ)2dξdτ . k kWs,b h i h − i | | (cid:18)Z Z (cid:19) tα−1 2.3. Riemman-Liouville fractional integral. The tempered distribution + is defined Γ(α) as a locally integrable function for Re α > 0 by tα+−1, f = 1 +∞tα 1f(t)dt. − Γ(α) Γ(α) * + Z0 For Re α >0, we have that tα 1 tα+k 1 +− = ∂k + − , Γ(α) t Γ(α+k) ! for all k N. This expression can be used to extend the definition, in the sense of distribu- tions) of∈t+α−1 to all α C. Γ(α) ∈ A change of contour calculation shows that b tα 1 +− (τ) = e−12πα(τ i0)−α, (2.1) Γ(α) − ! where (τ i0) α is the distributional limit. If f C (R+), we define − − ∈ 0∞ tα 1 f = +− f. α I Γ(α) ∗ Thus, when Re α> 0, 1 t f(t)= (t s)α 1f(s)ds, α − I Γ(α) − Z0 t and f = f, f(t)= f(s)ds, and f = f . Also = . I0 I1 0 I−1 ′ IαIβ Iα+β The following lemmas, whose proofs can be found in [14], state some useful properties of R the Riemman-Liouville fractional integral operator. Lemma 2.4. If f C (R+), then f C (R+), for all α C. ∈ 0∞ Iα ∈ 0∞ ∈ Lemma 2.5. If 0 ≤ α < ∞ and s ∈R, then kI−αhkH0s(R+) ≤ ckhkH0s+α(R+). Lemma 2.6. If 0 ≤ α < ∞, s ∈ R and µ ∈ C0∞(R), then kµIαhkH0s(R+) ≤ ckhkH0s−α(R+), where c= c(µ). tα−1 For more details on the distribution + see [10]. Γ(α) 6 IBVP FOR SOME QUADRATIC NONLINEAR SCHRO¨DINGER EQUATIONS ON THE HALF-LINE 3. Linear Version We define the unitary group associated to the linear Schro¨dinger equation as 1 e−it∂x2φ(x) = eixξeitξ2φˆ(ξ)dξ, 2π Z it follows that (i∂t −∂x2)e−it∂x2φ(x) = 0, (x,t) ∈ R×R, (3.1) (e−it∂x2φ(x) t=0 = φ(x), x ∈ R. Lemma 3.1. Let s R, 0< b <1. (cid:12)If φ Hs(R), then ∈ (cid:12) ∈ (a) (Space traces) ke−it∂x2φ(x)kC Rt;Hs(Rx) ≤ ckφkHs(R); (b) (Time traces) kψ(t)e−it∂x2φ(x(cid:0))kC(Rx;H2(cid:1)s4+1(Rt)) ≤ ckφkHs(R); (c) (Bourgain spaces) kψ(t)e−it∂x2φ(x)kXs,b ≤ ckψkH1(R)kφkHs(R). Proof. The assertion in (a) follows from properties of group e−it∂x2, (b) was obtained in [16] and the proof of (c) can be found in [12]. (cid:3) 4. The Duhamel Boundary Forcing Operator We now introduce the Duhamel boundary forcing operator of Holmer [13]. For f C (R+), define the boundary forcing operator ∈ 0∞ t f(x,t) = 2e−iπ4 e−i(t−t′)∂x2δ0(x) 1f(t′)dt′ L Z0 I−2 1 t 1 ix2 = (t t′)−2e−4(t−t′) 1f(t′)dt′. √π Z0 − I−2 The equivalence of the two definitions is clear from the formula x e−iπ4sgn a ei4xa2 (ξ)= e−iaξ2, a R. F 2a 1/2√π ∀ ∈ ! | | From this definition, we see that (i∂t −∂x2)Lf(x,t) = 2eiπ4δ0(x)I−12f(t), x,t ∈ R, (4.1) ( f(x,0) =0, x R. L ∈ The following lemma, whose proof can be found in [13], states some continuity properties of f(x,t). L Lemma 4.1. Let f C (R+). ∈ 0∞,c (a) For fixed t, f(x,t) is continuous in x for all x R and ∂ f(x,t) is continuous in x L ∈ L x for x = 0 with 6 lim ∂x f(x,t)= eiπ4 1/2f(t), lim ∂x f(x,t)= eiπ4 1/2f(t). x 0− L − I− x 0+ L I− → → IBVP FOR SOME QUADRATIC NONLINEAR SCHRO¨DINGER EQUATIONS ON THE HALF-LINE 7 (b) For N,k nonnegative integers and fixed x, ∂k f(x,t) is continuous in t for all t R+. tL ∈ We also have the pointwise estimates, on [0,T], ∂k f(x,t) + ∂ f(x,t) c x N, | tL | | xL | ≤ h i− where c= c(f,N,k,T). Let f ∈ C0∞(R+), set u(x,t) = e−it∂x2φ(x)+L(f −e−it∂x2φ(x) x=0). Then, by Lemma 4.1 (a) u(x,t) is continuous in x. Thus u(0,t) = f(t) and u(x,t) solves the problem (cid:12) u(i(∂xt,−0)∂=x2)φu((xx),,t) = 2eiπ4δ(x)I−1/2(f −e−it∂x2φ(x) x=(cid:12)0), x(x,t)R∈, R, (4.2)  (cid:12) ∈ u(0,t) = f(t), (cid:12) t R. ∈ This would suffice to solve the linear analogue of the half-line problem.  5. Nonlinear versions We define the Duhamel inhomogeneous solution operator as D t w(x,t) = i e−i(t−t′)∂x2w(x,t′)dt′, D − Z0 it follows that (i∂ ∂2) w(x,t) = w(x,t), (x,t) R R, wt(−x,0x) =D0, x R. ∈ × (5.1) (cid:26) D ∈ Lemma 5.1. Let s R. Then: ∈ (a) (Space traces) If 1 < c< 0 , then −2 kψ(t)Dw(x,t)kC Rt;Hs(Rx) ≤ ckwkXs,c; (b) (Time traces) If 1 < c < 0, then (cid:0) (cid:1) −2 ψ(t) w(x,t) ckwkXs,c, if − 21 < s ≤ 21, k D kC(Rx;H2s4+1(Rt)) ≤ (c( w Ws,c + w Xs,c), for all s R; k k k k ∈ (c) (Bourgain spaces estimates) If 1 < c 0 b c+1, then −2 ≤ ≤ ≤ kψ(t)Dw(x,t)kXs,b ≤ kwkXs,c. Remark 5.1. We note that the Ws,b (time-adapted) Bourgain spaces, used in Lemma 5.1 (b), are introduced in order to cover the full values of s. Proof. To prove (a), we use 2χ (t)= sgn(t)+sgn(t t), then (0,t) ′ ′ ′ − t eitτ eitξ2 ψ(t) ei(t−t′)∂x2w(,t′)dt′ = ψ(t) eixξ − wˆ(ξ,τ)dτdξ · τ ξ2 Z0 Zξ Zτ − eitτ eitξ2 eitτ eitξ2 = ψ(t) eixξ − wˆ(ξ,τ)dτdξ+ψ(t) eixξ − wˆ(ξ,τ)dτdξ τ ξ2 τ ξ2 Zξ Z|τ−ξ2|≤1 − Zξ Z|τ−ξ2|>1 − := w +w . 1 2 8 IBVP FOR SOME QUADRATIC NONLINEAR SCHRO¨DINGER EQUATIONS ON THE HALF-LINE To estimate w , we use ψ(t)eitτ eitξ2 c, when τ ξ2 1, more the Cauchy-Schwarz 1 τ−ξ2 ≤ | − | ≤ − inequality. (cid:12) (cid:12) To estimate w , we use(cid:12)(cid:12) ψ(t)eitτ eitξ(cid:12)(cid:12)2 c τ ξ2 1, when τ ξ2 > 1, and the Cauchy- 2 τ−ξ2 ≤ h − i− | − | − Schwarz inequality. To pro(cid:12)ve (b), we tak(cid:12)e θ(τ) C (R) such that θ(τ) = 1 for τ < 1 and (cid:12) (cid:12) ∈ ∞ | | 2 supp θ [ 2, 2], then (cid:12) (cid:12) ⊂ −3 3 t eitτ eitξ2 Fx ψ(t) ei(t−t′)∂x2w(x,t′) (ξ) = ψ(t) τ −ξ2 wˆ(ξ,τ)dτ (cid:18) Z0 (cid:19) Zτ − eit(τ ξ2) 1 1 θ(τ ξ2) = ψ(t)eitξ2 − − θ(τ ξ2)wˆ(ξ,τ)dτ +ψ(t) eitτ − − wˆ(ξ,τ)dτ τ ξ2 − τ ξ2 Zτ − Zτ − 1 θ(τ ξ2) ψ(t)eitξ2 − − wˆ(ξ,τ)dτ := w + w w . − τ ξ2 Fx 1 Fx 2−Fx 3 Zτ − By the power series expansion for eit(τ−ξ2), w1(x,t) = ∞ ψk(t)e−it∂x2φk(x), where ψk(t) = k! k=1 X iktkθ(t)andφˆ (ξ) = (τ ξ2)k 1θ(τ ξ2)wˆ(ξ,τ)dτ. UsingLemma3.1 (b), it sufficestoshow k τ − − − that φk Hs(R) c u Xs,c. Using the definition of φk and the Cauchy-Schwarz inequality, k k ≤ k Rk 2 ∞ φ 2 = c ξ 2s (τ ξ2)k 1θ(τ ξ2)uˆ(ξ,τ) dξ k kkHs(Rx) Zξh i Z{τ:|τ−ξ2|≤23}Xk=1 − − − ! c ξ 2s τ ξ2 2c uˆ(ξ,τ)2dτdξ. ≤ h i h − i | | Zξ Zτ This completes the estimate for w . To treat w , we change variables η = ξ2, then 1 2 + 2 kw2kC Rx;H2s4+1(Rt) ≤ cZτhτi2s2+1 (cid:18)Zη=0∞hτ −ηi−1η−12 wˆ(±η12,τ) dη(cid:19) dτ. (5.2) (cid:12) (cid:12) By Cauchy-S(cid:0)chwarz (5.2)(cid:1)is bounded by (cid:12) (cid:12) c τ 2s2+1G(τ) ∞ τ η −2b η −12 wˆ( η12,τ) 2dηdτ, h i h − i | | ± Zτ Zη=0 (cid:12) (cid:12) where G(τ) = ηhτ −ηi−2+2b|η|−12hηi−sdη. Separating(cid:12)G(τ) in th(cid:12)e regions |η| ≤ 1, 2|η| ≤ 1 τ , τ 2η , wRe can obtain G(τ) c τ −2. | | | | ≤ | | ≤ h i Now suppose 1 < s 1. Applying the Cauchy-Schwarz inequality we obtain −2 ≤ 2 kw2kC Rx;H2s4+1(Rt) ≤ Zτhτi2s2+1G(τ)Zξhτ +ξ2i2chξi2s|wˆ2(ξ,τ)|2dξdτ, where G(τ) = c(cid:0) ηhτ −ηi−2−(cid:1)2c|η|−21hηi−sdη. Separating in the regions η 1, 2η τ , τ 2η and using that 1 < s 1, R | | ≤ | | ≤ | | | | ≤ | | −2 ≤ 2 2s+1 we can obtain G(τ) c τ − 2 . This completes the estimate for w2. To obtain w3, we ≤ h i IBVP FOR SOME QUADRATIC NONLINEAR SCHRO¨DINGER EQUATIONS ON THE HALF-LINE 9 write w3 = ψ(t)e−it∂x2φ(x), where φˆ(ξ)= 1−θτ(τξ−2ξ2)wˆ(ξ,τ)dτ. Using Lemma 3.1 (b) and the Cauchy-Schwarz inequality, we obtain − R kw3kC(Rx;H2s4+1(Rt)) = ckψ(t)e−it∂x2φ(x)kC(Rx;H2s4+1(Rt)) ≤ ckφkHs(R) dτ c ξ 2s wˆ(ξ,τ)2 τ ξ2 2cdτ dξ. ≤ h i | | h − i τ ξ2 2 2c Zξ (cid:18)Zτ Z h − i − (cid:19) Since c > 1, we have 1 dτ c. This completes the estimate for w . −2 τ ξ2 2−2c ≤ 3 The statement (c) is ahst−andiard estimate and can be found in [12]. (cid:3) R Let Λ(u)(t) = e−it∂x2u˜0+ (N(u,u))(t)+ h(t), D L twehnesrioen, ho(ft)u=inχ(R0,.+∞) f(t)−e−it∂x2u˜0 x=0 −D(N(u,u))(t) x=0 (0,+∞) and u˜0 is a good ex- 0 (cid:0) (cid:12) (cid:12) (cid:1)(cid:12) Using (4.2) and (5.1), we see that (cid:12) (cid:12) (cid:12) (i∂t −∂x2)Λ(u)(t) = N(u,u)+2eiπ4δ0(x)I−21h(t) and Λ(u)(0,t) = f(t). Thus,awaytosolvetheIBVP(1.1)istoprovethatΛ(ormorepreciselyitstimetruncated versions) defines a contraction in a suitable Banach space. However, as pointed out before, since we are interested in low regularity results, we will use the auxiliary Bourgain spaces. Because of the estimate for the Duhamel forcing operator in our study, we need to take b < 1, in order to Lemma 6.2 (c) to be valid. However, the space Xs,b, with b < 1, fails to be 2 2 embedded into C R ; Hs(R ) . For this reason, we choose to work with the Banach space Z t x given by (cid:0) (cid:1) Z = C Rt; Hs(Rx) C Rx; H2s4+1(Rt) Xs,b. ∩ ∩ In Section 7 we will obtain the bilinear estimates associated to nonlinearities N (u,u) = u2 (cid:0) (cid:1) (cid:0) (cid:1) 1 and N (u,u) = u2, in Bourgain spaces Xs,b, for 3 < s 0. 3 −4 ≤ Unfortunately, we will able to prove the estimate for the Duhamel boundary forcing oper- ator in Bourgain spaces ψ(t) f(x,t) f , k L kXs,b ≤ k kH2s4+1(R+) 0 if 1 < s < 1. In the next section, inspired by [14], to obtain this estimate on the full −2 2 interval ( 3,0], we will define an analytic family of boundary operators λ, for λ C, such −4 L ∈ that 0 = and L L (iλ∂ft(−0,∂tx2))=Lλefi(3λx4π,ft)(t=).2eiπ4 xΓλ−(−λ)1I−12−λ2f(t), t(x,tR).∈ R2, L ∈ Due to the support properties of xλ−−1, (i∂ ∂2) λf(x,t) = 0 for x > 0, in the sense of Γ(λ) t − x L distributions. For any 3 < s 0, we will be able to address the right half-line problem −4 ≤ (1.1) by replacing to λ for suitable λ = λ(s). L L 10 IBVP FOR SOME QUADRATIC NONLINEAR SCHRO¨DINGER EQUATIONS ON THE HALF-LINE 6. The Duhamel Boundary Forcing Operator Class Define, for λ C, such that Re λ > 2, and f C (R+), ∈ − ∈ 0∞ xλ 1 λf(x,t)= −− λf (,t) (x). L "Γ(λ) ∗L I−2 · # (cid:0) (cid:1) It follows that 1 + λf(x,t) = ∞(y x)λ 1 f (y,t)dy, for Reλ > 0. (6.1) − λ L Γ(λ)Zx − L I−2 (cid:0) (cid:1) For Re λ > 2, by using (4.1), we see that − 1 + λf(x,t)= ∞(y x)λ+1∂2 f (y,t)dy L Γ(λ+2) Zx − yL I−λ2 (6.2) = 1 +∞(y x)λ+1 i∂t(cid:0) λf ((cid:1)y,t)dy 2eiπ4xλ−+1 1/2 λ/2f(t). Γ(λ+2) Zx − I−2 − Γ(λ+2)I− − (cid:0) (cid:1) From (4.1) we see that xλ 1 (i∂t−∂x2)Lλf(x,t) = 2ei4π Γ−(−λ)I−21−λ2f(t), in the sense of distributions. Using Lemma 4.1, we see that λf(x,t) is well defined for Re λ > 2 and t [0,1]. L − ∈ As xλ−−1 = δ ,then 0f(x,t) = f(x,t)and(6.2)implies 1f(x,t)= ∂ ( )f(x,t). Γ(λ) 0 L L L− xL I1/2 (cid:12)λ=0 The dom(cid:12) inated convergence Theorem and Lemma 4.1 imply that, for fixed t [0,1] and Re λ > 1(cid:12)(cid:12), λf(x,t) is continuous in x for x R. ∈ − L ∈ The next Lemma states the values of λf(x,t) at x = 0. L Lemma 6.1. Let f C (R+). If Re λ > 1, then ∈ 0∞ − λf(0,t) = ei3λ4πf(t). (6.3) L Proof. From (6.2), we have + yλ+1 λf(0,t) = ∞ ∂2 f (y,t)dy. (6.4) L Z0 Γ(λ+2) yL I−λ2 (cid:0) (cid:1) By complex differentiation under the integral sign, (6.4) proves that λf(0,t) is analytic L in λ, for Re λ > 1. By analyticity, we shall only compute (6.3) for 0 < λ < 2. − For the computation in the range 0 < λ < 2, we use the representation (6.1), to obtain 1 + λf(0,t) = ∞(y)λ 1 f (y,t)dy − λ L Γ(λ)Z0 L I−2 = 1 t(t t′)−21 (cid:0) 1 λf(cid:1)(t′) +∞yλ−1 1 e4−(ti−yt2′)dydt′. Γ(λ)Z0 − I−2−2 Z0 √π

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.