Introduction
Main results and remarks
In this paper, we are concerned with the following IBVP (initial boundary value problem) of 2-D quadratic quasilinear wave equations in exterior domains with homogeneous Dirichlet boundary conditions
{ □ u = Q ( ∂ u , ∂ 2 u ) , ( t , x ) ∈ [ 0 , + ∞ ) × K , u ( t , x ) = 0 , ( t , x ) ∈ [ 0 , + ∞ ) × ∂ K , ( u , ∂ t u ) ( 0 , x ) = ( ε u 0 , ε u 1 ) ( x ) , x ∈ K . (1) \begin{cases}
\Box u = Q(\partial u,\partial^2u), & (t,x)\in[0,+\infty)\times K,\\
u(t,x)=0, & (t,x)\in[0,+\infty)\times\partial K,\\
(u,\partial_tu)(0,x)=(\varepsilon u_0,\varepsilon u_1)(x), & x\in K. \tag*{(1)}
\end{cases} ⎩ ⎨ ⎧ □ u = Q ( ∂ u , ∂ 2 u ) , u ( t , x ) = 0 , ( u , ∂ t u ) ( 0 , x ) = ( ε u 0 , ε u 1 ) ( x ) , ( t , x ) ∈ [ 0 , + ∞ ) × K , ( t , x ) ∈ [ 0 , + ∞ ) × ∂ K , x ∈ K . (1)
where ∂ = ( ∂ 0 , ∂ 1 , ∂ 2 ) = ( ∂ t , ∂ x 1 , ∂ x 2 ) \partial=(\partial_0,\partial_1,\partial_2)=(\partial_t,\partial_{x_1},\partial_{x_2}) ∂ = ( ∂ 0 , ∂ 1 , ∂ 2 ) = ( ∂ t , ∂ x 1 , ∂ x 2 ) , □ = ∂ t 2 − Δ \Box=\partial_t^2-\Delta □ = ∂ t 2 − Δ with Δ = ∂ 1 2 + ∂ 2 2 \Delta=\partial_1^2+\partial_2^2 Δ = ∂ 1 2 + ∂ 2 2 , K = R 2 ∖ O K=\mathbb{R}^2\setminus\mathcal{O} K = R 2 ∖ O , the obstacle O ⊂ R 2 \mathcal{O}\subset\mathbb{R}^2 O ⊂ R 2 is compact and contains the origin, the boundary ∂ K = ∂ O \partial K=\partial\mathcal{O} ∂ K = ∂ O is smooth, ε > 0 \varepsilon>0 ε > 0 is small, ( u 0 , u 1 ) ∈ C ∞ ( K ) (u_0,u_1)\in C^\infty(K) ( u 0 , u 1 ) ∈ C ∞ ( K ) and supp ( u 0 , u 1 ) ⊂ { x ∈ K : ∣ x ∣ ≤ M 0 } \operatorname{supp}(u_0,u_1)\subset\{x\in K:|x|\le M_0\} supp ( u 0 , u 1 ) ⊂ { x ∈ K : ∣ x ∣ ≤ M 0 } with some fixed constant M 0 > 1 M_0>1 M 0 > 1 . In addition, the smooth nonlinearity is given by
Q ( ∂ u , ∂ 2 u ) = ∑ α , β = 0 2 Q α β ( ∂ u ) ∂ α β 2 u Q(\partial u,\partial^2u)=\sum_{\alpha,\beta=0}^{2}Q^{\alpha\beta}(\partial u)\partial_{\alpha\beta}^2u Q ( ∂ u , ∂ 2 u ) = α , β = 0 ∑ 2 Q α β ( ∂ u ) ∂ α β 2 u
with
Q α β ( ∂ u ) = ∑ γ = 0 2 Q α β γ ∂ γ u + ∑ γ , δ = 0 2 Q α β γ δ ∂ γ u ∂ δ u + O ( ∣ ∂ u ∣ 3 ) , (2) Q^{\alpha\beta}(\partial u)=\sum_{\gamma=0}^{2}Q^{\alpha\beta\gamma}\partial_\gamma u+\sum_{\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\partial_\gamma u\partial_\delta u+O(|\partial u|^3), \tag*{(2)} Q α β ( ∂ u ) = γ = 0 ∑ 2 Q α β γ ∂ γ u + γ , δ = 0 ∑ 2 Q α β γ δ ∂ γ u ∂ δ u + O ( ∣ ∂ u ∣ 3 ) , (2)
where the constants Q α β γ Q^{\alpha\beta\gamma} Q α β γ , Q α β γ δ Q^{\alpha\beta\gamma\delta} Q α β γ δ are naturally imposed to satisfy the symmetric conditions
Q α β γ = Q β α γ , Q α β γ δ = Q β α γ δ = Q α β δ γ . (3) Q^{\alpha\beta\gamma}=Q^{\beta\alpha\gamma},\qquad Q^{\alpha\beta\gamma\delta}=Q^{\beta\alpha\gamma\delta}=Q^{\alpha\beta\delta\gamma}. \tag*{(3)} Q α β γ = Q β α γ , Q α β γ δ = Q β α γ δ = Q α β δ γ . (3)
Meanwhile, for ξ = ( ξ 0 , ξ 1 , ξ 2 ) ∈ { ± 1 } × S 1 \xi=(\xi_0,\xi_1,\xi_2)\in\{\pm1\}\times S^1 ξ = ( ξ 0 , ξ 1 , ξ 2 ) ∈ { ± 1 } × S 1 , the first null condition is satisfied
Q 1 s t ( ξ ) = ∑ α , β , γ = 0 2 Q α β γ ξ α ξ β ξ γ ≡ 0 (4) Q_{1st}(\xi)=\sum_{\alpha,\beta,\gamma=0}^{2}Q^{\alpha\beta\gamma}\xi_\alpha\xi_\beta\xi_\gamma\equiv0 \tag*{(4)} Q 1 s t ( ξ ) = α , β , γ = 0 ∑ 2 Q α β γ ξ α ξ β ξ γ ≡ 0 (4)
and the second null condition is violated
Q 2 n d ( ξ ) = ∑ α , β , γ , δ = 0 2 Q α β γ δ ξ α ξ β ξ γ ξ δ ≠ 0. (5) Q_{2nd}(\xi)=\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\xi_{\alpha}\xi_{\beta}\xi_{\gamma}\xi_{\delta}\ne0. \tag*{(5)} Q 2 n d ( ξ ) = α , β , γ , δ = 0 ∑ 2 Q α β γ δ ξ α ξ β ξ γ ξ δ = 0. (5)
As usual, in order to find the smooth solutions of problem (1.1), the compatibility conditions of the initial data ( ε u 0 , ε u 1 ) (\varepsilon u_0,\varepsilon u_1) ( ε u 0 , ε u 1 ) with the boundary value u ∣ [ 0 , + ∞ ) × ∂ K = 0 u|_{[0,+\infty)\times\partial\mathcal{K}}=0 u ∣ [ 0 , + ∞ ) × ∂ K = 0 are necessary. That is, let J k u = { ∂ x α u : 0 ≤ ∣ α ∣ ≤ k } J_k u=\{\partial_x^\alpha u:0\le|\alpha|\le k\} J k u = { ∂ x α u : 0 ≤ ∣ α ∣ ≤ k } and ∂ t k u ( 0 , x ) = F k ( ε J k u 0 , ε J k − 1 u 1 ) \partial_t^k u(0,x)=F_k(\varepsilon J_k u_0,\varepsilon J_{k-1}u_1) ∂ t k u ( 0 , x ) = F k ( ε J k u 0 , ε J k − 1 u 1 ) ( 0 ≤ k ≤ 2 N ) (0\le k\le2N) ( 0 ≤ k ≤ 2 N ) , where F k F_k F k depends on Q ( ∂ u , ∂ 2 u ) Q(\partial u,\partial^2u) Q ( ∂ u , ∂ 2 u ) , ε J k u 0 \varepsilon J_k u_0 ε J k u 0 and ε J k − 1 u 1 \varepsilon J_{k-1}u_1 ε J k − 1 u 1 in terms of the equation and the initial data in problem (1.1), then the compatibility conditions up to ( 2 N + 1 ) (2N+1) ( 2 N + 1 ) -order mean that
F k vanish on ∂ K for all 0 ≤ k ≤ 2 N . (6) F_k\text{ vanish on }\partial\mathcal{K}\text{ for all }0\le k\le2N. \tag*{(6)} F k vanish on ∂ K for all 0 ≤ k ≤ 2 N . (6)
Let w h o m w_{\mathrm{hom}} w hom solve the homogeneous initial boundary value problem
{ □ w h o m = 0 , ( t , x ) ∈ [ 0 , + ∞ ) × K , w h o m ( t , x ) = 0 , ( t , x ) ∈ [ 0 , + ∞ ) × ∂ K , ( w h o m , ∂ t w h o m ) ( 0 , x ) = ( ε u 0 , ε u 1 ) ( x ) , x ∈ K . (7) \begin{cases}
\Box w_{\mathrm{hom}}=0, & (t,x)\in[0,+\infty)\times\mathcal{K},\\
w_{\mathrm{hom}}(t,x)=0, & (t,x)\in[0,+\infty)\times\partial\mathcal{K},\\
(w_{\mathrm{hom}},\partial_t w_{\mathrm{hom}})(0,x)=(\varepsilon u_0,\varepsilon u_1)(x), & x\in\mathcal{K}. \tag*{(7)}
\end{cases} ⎩ ⎨ ⎧ □ w hom = 0 , w hom ( t , x ) = 0 , ( w hom , ∂ t w hom ) ( 0 , x ) = ( ε u 0 , ε u 1 ) ( x ) , ( t , x ) ∈ [ 0 , + ∞ ) × K , ( t , x ) ∈ [ 0 , + ∞ ) × ∂ K , x ∈ K . (7)
Our main result can be stated as
Theorem 1.1. Suppose that the obstacle O \mathcal{O} O is star-shaped and (1.2)–(1.6) hold. Then there are two constants ε 0 , κ > 0 \varepsilon_0,\kappa>0 ε 0 , κ > 0 such that when ε ≤ ε 0 \varepsilon\le\varepsilon_0 ε ≤ ε 0 and
∥ u 0 ∥ H 2 N + 1 ( K ) + ∥ u 1 ∥ H 2 N ( K ) ≤ 1 with N ≥ 53 , (8) \lVert u_0\rVert_{H^{2N+1}(\mathcal{K})}+\lVert u_1\rVert_{H^{2N}(\mathcal{K})}\le1\quad\text{with }N\ge53, \tag*{(8)} ∥ u 0 ∥ H 2 N + 1 ( K ) + ∥ u 1 ∥ H 2 N ( K ) ≤ 1 with N ≥ 53 , (8) problem (1.1) has a classical solution u ∈ ⋂ j = 0 2 N + 1 C j ( [ 0 , e κ / ε 2 ) , H 2 N + 1 − j ( K ) ) u\in\bigcap_{j=0}^{2N+1}C^j([0,e^{\kappa/\varepsilon^2}),H^{2N+1-j}(\mathcal{K})) u ∈ ⋂ j = 0 2 N + 1 C j ([ 0 , e κ / ε 2 ) , H 2 N + 1 − j ( K )) . Moreover,
∑ ∣ α ∣ ≤ 2 N − 46 ∣ ∂ Z α u ∣ ≤ C ε ⟨ x ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ 0.001 − 1 , (9) \sum_{|\alpha|\le2N-46}|\partial Z^\alpha u|\le C\varepsilon\langle x\rangle^{-1/2}\langle t-|x|\rangle^{0.001-1}, \tag*{(9)} ∣ α ∣ ≤ 2 N − 46 ∑ ∣ ∂ Z α u ∣ ≤ C ε ⟨ x ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ 0.001 − 1 , (9) ∑ ∣ α ∣ ≤ 2 N − 46 ∑ i = 1 2 ∣ ∂ ‾ i Z α u ∣ ≤ C ε ⟨ x ⟩ − 1 / 2 ⟨ t + ∣ x ∣ ⟩ 0.001 − 1 , (10) \sum_{|\alpha|\le2N-46}\sum_{i=1}^{2}|\overline{\partial}_i Z^\alpha u|\le C\varepsilon\langle x\rangle^{-1/2}\langle t+|x|\rangle^{0.001-1}, \tag*{(10)} ∣ α ∣ ≤ 2 N − 46 ∑ i = 1 ∑ 2 ∣ ∂ i Z α u ∣ ≤ C ε ⟨ x ⟩ − 1/2 ⟨ t + ∣ x ∣ ⟩ 0.001 − 1 , (10) ∑ ∣ α ∣ ≤ 2 N − 52 ∣ Z α u ∣ ≤ C ε ⟨ t + ∣ x ∣ ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ 0.003 − 1 / 2 , (11) \sum_{|\alpha|\le2N-52}|Z^\alpha u|\le C\varepsilon\langle t+|x|\rangle^{-1/2}\langle t-|x|\rangle^{0.003-1/2}, \tag*{(11)} ∣ α ∣ ≤ 2 N − 52 ∑ ∣ Z α u ∣ ≤ C ε ⟨ t + ∣ x ∣ ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ 0.003 − 1/2 , (11) ∑ ∣ α ∣ ≤ 2 N − 46 ∣ ∂ Z α ( u − w h o m ) ∣ ≤ C ε 2 ⟨ t ⟩ − 1 , ∣ x ∣ ≤ t / 10 , (12) \sum_{|\alpha|\le2N-46}|\partial Z^\alpha(u-w_{\mathrm{hom}})|\le C\varepsilon^2\langle t\rangle^{-1},\quad|x|\le t/10, \tag*{(12)} ∣ α ∣ ≤ 2 N − 46 ∑ ∣ ∂ Z α ( u − w hom ) ∣ ≤ C ε 2 ⟨ t ⟩ − 1 , ∣ x ∣ ≤ t /10 , (12) where Z = { ∂ , Ω } Z=\{\partial,\Omega\} Z = { ∂ , Ω } with Ω = x 1 ∂ 2 − x 2 ∂ 1 \Omega=x_1\partial_2-x_2\partial_1 Ω = x 1 ∂ 2 − x 2 ∂ 1 , ⟨ x ⟩ = 1 + ∣ x ∣ 2 \langle x\rangle=\sqrt{1+|x|^2} ⟨ x ⟩ = 1 + ∣ x ∣ 2 , ∣ x ∣ = r = x 1 2 + x 2 2 |x|=r=\sqrt{x_1^2+x_2^2} ∣ x ∣ = r = x 1 2 + x 2 2 , ∂ ‾ i = ∂ i + x i ∣ x ∣ ∂ t \overline{\partial}_i=\partial_i+\frac{x_i}{|x|}\partial_t ∂ i = ∂ i + ∣ x ∣ x i ∂ t ( i = 1 , 2 ) (i=1,2) ( i = 1 , 2 ) are the good derivatives (tangent to the outgoing light cone ∣ x ∣ = t |x|=t ∣ x ∣ = t ). Furthermore, the following time decay estimates of local energy hold
∑ ∣ α ∣ ≤ 2 N − 25 ∥ ∂ α u ∥ L 2 ( K R ) ≤ C R ε ( 1 + t ) − 1 , (13) \sum_{|\alpha|\le2N-25}\|\partial^\alpha u\|_{L^2(\mathcal{K}_R)}\le C_R\varepsilon(1+t)^{-1}, \tag*{(13)} ∣ α ∣ ≤ 2 N − 25 ∑ ∥ ∂ α u ∥ L 2 ( K R ) ≤ C R ε ( 1 + t ) − 1 , (13) ∑ ∣ α ∣ ≤ 2 N − 25 ∥ ∂ α ( u − w h o m ) ∥ L 2 ( K R ) ≤ C R ε 2 ( 1 + t ) − 1 , (14) \sum_{|\alpha|\le2N-25}\|\partial^\alpha(u-w_{\mathrm{hom}})\|_{L^2(\mathcal{K}_R)}\le C_R\varepsilon^2(1+t)^{-1}, \tag*{(14)} ∣ α ∣ ≤ 2 N − 25 ∑ ∥ ∂ α ( u − w hom ) ∥ L 2 ( K R ) ≤ C R ε 2 ( 1 + t ) − 1 , (14) where R > 1 R>1 R > 1 is a fixed constant, K R = K ∩ { x : ∣ x ∣ ≤ R } \mathcal{K}_R=\mathcal{K}\cap\{x:|x|\le R\} K R = K ∩ { x : ∣ x ∣ ≤ R } and C R > 0 C_R>0 C R > 0 is a constant depending on R R R .
and for the lifespan T ˉ ε \bar{T}_{\varepsilon} T ˉ ε corresponding to the 3-D exterior domain problem without the null condition
lim ε → 0 + ε ln T ˉ ε = τ ˉ 0 = ( 1 2 max ( σ , ω ) ∑ α , β , γ = 0 3 Q α β γ ω ~ α ω ~ β ω ~ γ ∂ σ 2 F ˉ 0 ( σ , ω ) ) − 1 . \lim_{\varepsilon\to0+}\varepsilon\ln\bar{T}_{\varepsilon}=\bar{\tau}_{0}=\left(\frac{1}{2}\max_{(\sigma,\omega)}\sum_{\alpha,\beta,\gamma=0}^{3}Q^{\alpha\beta\gamma}\widetilde{\omega}_{\alpha}\widetilde{\omega}_{\beta}\widetilde{\omega}_{\gamma}\partial_{\sigma}^{2}\bar{F}_{0}(\sigma,\omega)\right)^{-1}. ε → 0 + lim ε ln T ˉ ε = τ ˉ 0 = 2 1 ( σ , ω ) max α , β , γ = 0 ∑ 3 Q α β γ ω α ω β ω γ ∂ σ 2 F ˉ 0 ( σ , ω ) − 1 .
where ω ~ = ( ω 0 , ω ) = ( ω 0 , ω 1 , ω 2 , ω 3 ) ∈ { ± 1 } × S 2 \widetilde{\omega}=(\omega_{0},\omega)=(\omega_{0},\omega_{1},\omega_{2},\omega_{3})\in\{\pm1\}\times\mathbb{S}^{2} ω = ( ω 0 , ω ) = ( ω 0 , ω 1 , ω 2 , ω 3 ) ∈ { ± 1 } × S 2 , x = r ω x=r\omega x = r ω with r = x 1 2 + x 2 2 + x 3 2 r=\sqrt{x_{1}^{2}+x_{2}^{2}+x_{3}^{2}} r = x 1 2 + x 2 2 + x 3 2 , σ = r − t \sigma=r-t σ = r − t and F ˉ 0 ( σ , ω ) = lim r → + ∞ r w h o m ( r − σ , r ω ) \bar{F}_{0}(\sigma,\omega)=\lim_{r\to+\infty}r w_{\mathrm{hom}}(r-\sigma,r\omega) F ˉ 0 ( σ , ω ) = lim r → + ∞ r w hom ( r − σ , r ω ) with w h o m w_{\mathrm{hom}} w hom being the solution corresponding to the 3-D case of (7) . Therefore, the precise position of shock formation for the 2-D or 3-D exterior domain problem (1) can be determined. It is worth noting that although the shock formation for the Cauchy problems of quasilinear wave equations or multidimensional hyperbolic conservation laws has been deeply studied (see [7 ] , [36 ] , [37 ] , [40 ] and [45 ] ), there are few results on the blowup mechanism of smooth solutions to the corresponding initial boundary value problems in exterior domains.
Remark 1.5. The quasilinear wave equation in (1) not only admits a lot of interesting physical backgrounds (for example, see [19 ] ) but also can be derived from some 2-D variational problem. Indeed, for z = ( z 0 , z 1 , z 2 ) ∈ R 3 z=(z^{0},z^{1},z^{2})\in\mathbb{R}^{3} z = ( z 0 , z 1 , z 2 ) ∈ R 3 , let L ( z ) \mathcal{L}(z) L ( z ) be a smooth scalar Lagrangian density and assume L z α z β ( 0 ) = m α β \mathcal{L}_{z^{\alpha}z^{\beta}}(0)=m^{\alpha\beta} L z α z β ( 0 ) = m α β with ( m α β ) α , β = 0 2 = diag ( 1 , − 1 , − 1 ) (m^{\alpha\beta})_{\alpha,\beta=0}^{2}=\operatorname{diag}(1,-1,-1) ( m α β ) α , β = 0 2 = diag ( 1 , − 1 , − 1 ) . It follows from the first variation ∑ α = 0 2 ∂ α ( L z α ( ∂ u ) ) = 0 \sum_{\alpha=0}^{2}\partial_{\alpha}(\mathcal{L}_{z^{\alpha}}(\partial u))=0 ∑ α = 0 2 ∂ α ( L z α ( ∂ u )) = 0 and Taylor expansion that
□ u = ∑ α , β , γ = 0 2 g α β γ ∂ γ u ∂ α β 2 u + ∑ α , β , γ , ν = 0 2 h α β γ ν ∂ γ u ∂ ν u ∂ α β 2 u + Q ( ∂ u ) , (17) \Box u=\sum_{\alpha,\beta,\gamma=0}^{2}g^{\alpha\beta\gamma}\partial_{\gamma}u\partial_{\alpha\beta}^{2}u+\sum_{\alpha,\beta,\gamma,\nu=0}^{2}h^{\alpha\beta\gamma\nu}\partial_{\gamma}u\partial_{\nu}u\partial_{\alpha\beta}^{2}u+Q(\partial u), \tag*{(17)} □ u = α , β , γ = 0 ∑ 2 g α β γ ∂ γ u ∂ α β 2 u + α , β , γ , ν = 0 ∑ 2 h α β γ ν ∂ γ u ∂ ν u ∂ α β 2 u + Q ( ∂ u ) , (17)
where g α β γ = − L z α z β z γ ( 0 ) g^{\alpha\beta\gamma}=-\mathcal{L}_{z^{\alpha}z^{\beta}z^{\gamma}}(0) g α β γ = − L z α z β z γ ( 0 ) , h α β γ ν h^{\alpha\beta\gamma\nu} h α β γ ν are constants with h α β γ ν = h β α γ ν = h α β ν γ h^{\alpha\beta\gamma\nu}=h^{\beta\alpha\gamma\nu}=h^{\alpha\beta\nu\gamma} h α β γ ν = h β α γ ν = h α β ν γ , and Q ( ∂ u ) = O ( ∣ ∂ u ∣ 3 ∣ ∂ 2 u ∣ ) Q(\partial u)=O(|\partial u|^{3}|\partial^{2}u|) Q ( ∂ u ) = O ( ∣ ∂ u ∣ 3 ∣ ∂ 2 u ∣ ) is a smooth function. If the first null condition holds but the second null condition is violated in (17) , then (17) fulfills the requirement on the quasilinear wave equation in (1) .
Remark 1.6. Consider the corresponding 2-D or 3-D Cauchy-Neumann problem of (1)
{ □ u = Q ( ∂ u , ∂ 2 u ) , ( t , x ) ∈ [ 0 , + ∞ ) × K , ∂ ∂ n u ( t , x ) = 0 , ( t , x ) ∈ [ 0 , + ∞ ) × ∂ K , ( u , ∂ t u ) ( 0 , x ) = ( ε u 0 , ε u 1 ) ( x ) , x ∈ K . (18) \left\{
\begin{aligned}
\Box u&=Q(\partial u,\partial^{2}u), && (t,x)\in[0,+\infty)\times K,\\
\frac{\partial}{\partial n}u(t,x)&=0, && (t,x)\in[0,+\infty)\times\partial K,\\
(u,\partial_{t}u)(0,x)&=(\varepsilon u_{0},\varepsilon u_{1})(x), && x\in K.
\end{aligned}
\right. \tag*{(18)} ⎩ ⎨ ⎧ □ u ∂ n ∂ u ( t , x ) ( u , ∂ t u ) ( 0 , x ) = Q ( ∂ u , ∂ 2 u ) , = 0 , = ( ε u 0 , ε u 1 ) ( x ) , ( t , x ) ∈ [ 0 , + ∞ ) × K , ( t , x ) ∈ [ 0 , + ∞ ) × ∂ K , x ∈ K . (18)
where n = ( 0 , n 1 , n 2 ) = ( 0 , n 1 ( x ) , n 2 ( x ) ) \mathbf{n}=(0,n_{1},n_{2})=(0,n_{1}(x),n_{2}(x)) n = ( 0 , n 1 , n 2 ) = ( 0 , n 1 ( x ) , n 2 ( x )) represents the unit outer normal direction of the boundary [ 0 , + ∞ ) × ∂ K [0,+\infty)\times\partial K [ 0 , + ∞ ) × ∂ K . In addition, the following admissible condition on Q ( ∂ u , ∂ 2 u ) Q(\partial u,\partial^{2}u) Q ( ∂ u , ∂ 2 u ) for Neumann boundary value problem (18) is imposed:
For any C 1 ( R 2 ) C^{1}(\mathbb{R}^{2}) C 1 ( R 2 ) smooth functions w w w and ψ \psi ψ verifying ∂ ∂ n w = ∂ ∂ n ψ = 0 \frac{\partial}{\partial n}w=\frac{\partial}{\partial n}\psi=0 ∂ n ∂ w = ∂ n ∂ ψ = 0 on [ 0 , + ∞ ) × ∂ K [0,+\infty)\times\partial K [ 0 , + ∞ ) × ∂ K ,
it holds that on [ 0 , + ∞ ) × ∂ K [0,+\infty)\times\partial K [ 0 , + ∞ ) × ∂ K ,
∑ α , β , μ = 0 2 Q α β μ n α ∂ β w ∂ μ ψ = 0 and ∑ α , β , μ , ν = 0 2 Q α β μ ν n α ∂ β w ∂ μ ψ ∂ ν ψ = 0. (19) \sum_{\alpha,\beta,\mu=0}^{2}Q^{\alpha\beta\mu}n_{\alpha}\partial_{\beta}w\partial_{\mu}\psi=0\quad\text{and}\quad\sum_{\alpha,\beta,\mu,\nu=0}^{2}Q^{\alpha\beta\mu\nu}n_{\alpha}\partial_{\beta}w\partial_{\mu}\psi\partial_{\nu}\psi=0. \tag*{(19)} α , β , μ = 0 ∑ 2 Q α β μ n α ∂ β w ∂ μ ψ = 0 and α , β , μ , ν = 0 ∑ 2 Q α β μν n α ∂ β w ∂ μ ψ ∂ ν ψ = 0. (19)
Under the compatibility conditions of the initial data ( ε u 0 , ε u 1 ) (\varepsilon u_{0},\varepsilon u_{1}) ( ε u 0 , ε u 1 ) with the boundary value ∂ ∂ n u ∣ [ 0 , + ∞ ) × ∂ K = 0 \left.\frac{\partial}{\partial n}u\right|_{[0,+\infty)\times\partial K}=0 ∂ n ∂ u [ 0 , + ∞ ) × ∂ K = 0 , there are some interesting results on the lifespan of problem (18) (see [5 ] , [9 ] , [19 ] , [23 ] , [34 ] , [42 ] –[44 ] and [47 ] ). However, compared with the Cauchy-Dirichlet problem (1) , the results on the precise lifespan of the Cauchy-Neumann problem (18) are still rather incomplete.
Previous results
The long time existence of small data smooth solutions to the quasilinear wave equations in the whole space R n \mathbb{R}^{n} R n or in the exterior domains have been extensively studied. We now give systematic reviews on the related results. Consider the following Cauchy problem of the quasilinear wave equation
{ □ w + ∑ α , β , μ = 0 n Q α β μ ∂ α β 2 w ∂ μ w + ∑ α , β , μ , ν = 0 n Q α β μ ν ∂ α β 2 w ∂ μ w ∂ ν w = 0 , t > 0 , x ∈ R n , ( w , ∂ t w ) ( 0 , x ) = ( ε w 0 C , ε w 1 C ) ( x ) , x ∈ R n (20) \left\{
\begin{aligned}
\square w+\sum_{\alpha,\beta,\mu=0}^{n}Q^{\alpha\beta\mu}\partial_{\alpha\beta}^{2}w\partial_{\mu}w+\sum_{\alpha,\beta,\mu,\nu=0}^{n}Q^{\alpha\beta\mu\nu}\partial_{\alpha\beta}^{2}w\partial_{\mu}w\partial_{\nu}w&=0,\quad t>0,\ x\in\mathbb{R}^{n},\\
(w,\partial_{t}w)(0,x)&=(\varepsilon w_{0}^{C},\varepsilon w_{1}^{C})(x),\quad x\in\mathbb{R}^{n}
\end{aligned}
\right. \tag*{(20)} ⎩ ⎨ ⎧ □ w + α , β , μ = 0 ∑ n Q α β μ ∂ α β 2 w ∂ μ w + α , β , μ , ν = 0 ∑ n Q α β μν ∂ α β 2 w ∂ μ w ∂ ν w ( w , ∂ t w ) ( 0 , x ) = 0 , t > 0 , x ∈ R n , = ( ε w 0 C , ε w 1 C ) ( x ) , x ∈ R n (20)
and the initial boundary value problem in exterior domains
{ □ w + ∑ α , β , μ = 0 n Q α β μ ∂ α β 2 w ∂ μ w + ∑ α , β , μ , ν = 0 n Q α β μ ν ∂ α β 2 w ∂ μ w ∂ ν w = 0 , t > 0 , x ∈ K , w = 0 , t > 0 , x ∈ ∂ K , ( w , ∂ t w ) ( 0 , x ) = ( ε w 0 B , ε w 1 B ) ( x ) , x ∈ K (21) \left\{
\begin{aligned}
\square w+\sum_{\alpha,\beta,\mu=0}^{n}Q^{\alpha\beta\mu}\partial_{\alpha\beta}^{2}w\partial_{\mu}w+\sum_{\alpha,\beta,\mu,\nu=0}^{n}Q^{\alpha\beta\mu\nu}\partial_{\alpha\beta}^{2}w\partial_{\mu}w\partial_{\nu}w&=0,\quad t>0,\ x\in K,\\
w&=0,\quad t>0,\ x\in\partial K,\\
(w,\partial_{t}w)(0,x)&=(\varepsilon w_{0}^{B},\varepsilon w_{1}^{B})(x),\quad x\in K
\end{aligned}
\right. \tag*{(21)} ⎩ ⎨ ⎧ □ w + α , β , μ = 0 ∑ n Q α β μ ∂ α β 2 w ∂ μ w + α , β , μ , ν = 0 ∑ n Q α β μν ∂ α β 2 w ∂ μ w ∂ ν w w ( w , ∂ t w ) ( 0 , x ) = 0 , t > 0 , x ∈ K , = 0 , t > 0 , x ∈ ∂ K , = ( ε w 0 B , ε w 1 B ) ( x ) , x ∈ K (21)
where n ≥ 2 n\geq2 n ≥ 2 , ( w 0 C , w 1 C ) ∈ C 0 ∞ ( R n ) (w_{0}^{C},w_{1}^{C})\in C_{0}^{\infty}(\mathbb{R}^{n}) ( w 0 C , w 1 C ) ∈ C 0 ∞ ( R n ) , ( w 0 B , w 1 B ) ∈ C 0 ∞ ( K ) (w_{0}^{B},w_{1}^{B})\in C_{0}^{\infty}(K) ( w 0 B , w 1 B ) ∈ C 0 ∞ ( K ) , ε > 0 \varepsilon>0 ε > 0 is small, K = R n ∖ O K=\mathbb{R}^{n}\setminus O K = R n ∖ O , the obstacle O ⊂ R n O\subset\mathbb{R}^{n} O ⊂ R n is compact and contains the origin, and the boundary ∂ K = ∂ O \partial K=\partial O ∂ K = ∂ O is smooth. Let T ε C T_{\varepsilon}^{C} T ε C and T ε D T_{\varepsilon}^{D} T ε D denote the lifespan of smooth solutions to the Cauchy problem (20) and the Dirichlet boundary problem (21) , respectively. For n = 3 n=3 n = 3 in (20) and (21) , set Q ( ξ ) = ∑ α , β , γ = 0 3 Q α β γ ξ α ξ β ξ γ Q(\xi)=\sum_{\alpha,\beta,\gamma=0}^{3}Q^{\alpha\beta\gamma}\xi_{\alpha}\xi_{\beta}\xi_{\gamma} Q ( ξ ) = ∑ α , β , γ = 0 3 Q α β γ ξ α ξ β ξ γ for ξ = ( ξ 0 , ξ 1 , ξ 2 , ξ 3 ) ∈ { ± 1 } × S 2 \xi=(\xi_{0},\xi_{1},\xi_{2},\xi_{3})\in\{\pm1\}\times S^{2} ξ = ( ξ 0 , ξ 1 , ξ 2 , ξ 3 ) ∈ { ± 1 } × S 2 . The previously known results on T ε C T_{\varepsilon}^{C} T ε C and T ε D T_{\varepsilon}^{D} T ε D can be briefly described in the following table:
Table 1.
Next we provide a detailed account of the conclusions in the table above.
When n ≥ 4 n\geq4 n ≥ 4 , the global existence of small data smooth solutions to both the problems (20) and (21) has been well known, i.e., T ε C , T ε D = ∞ T_{\varepsilon}^{C},T_{\varepsilon}^{D}=\infty T ε C , T ε D = ∞ . The Cauchy problem (20) is referred to [14 , 26 , 28 , 35 ] and see [5 , 12 , 39 , 44 ] for the exterior domain problem (21) .
When n = 3 n=3 n = 3 and the null condition is satisfied (i.e., Q ( ξ ) ≡ 0 Q(\xi)\equiv0 Q ( ξ ) ≡ 0 ), the global existence of small data smooth solutions for both the Cauchy problem (20) and the initial boundary value problem (21) can be found in [6 , 27 ] and [11 , 22 , 24 , 38 ] , respectively. While the null condition is violated (that means, Q ( ξ ) ≢ 0 Q(\xi)\not\equiv0 Q ( ξ ) ≡ 0 ), the sharp lower bounds T ε C , T ε D ≥ e C / ε T_\varepsilon^C,T_\varepsilon^D\geq e^{C/\varepsilon} T ε C , T ε D ≥ e C / ε have been established for problem (20) (see [20 ] ) and problem (21) (one can be referred to [9 , 25 , 39 ] ), respectively.
When n = 2 n=2 n = 2 and all Q α β μ Q^{\alpha\beta\mu} Q α β μ vanish (i.e., Q α β μ ≡ 0 Q^{\alpha\beta\mu}\equiv0 Q α β μ ≡ 0 for all α , β , μ = 0 , 1 , 2 \alpha,\beta,\mu=0,1,2 α , β , μ = 0 , 1 , 2 ), the global smooth solutions with the second null condition (Q 2 n d ( ξ ) ≡ 0 Q_{2nd}(\xi)\equiv0 Q 2 n d ( ξ ) ≡ 0 ) and the lower bound of the lifespan without the second null condition (Q 2 n d ( ξ ) ≢ 0 Q_{2nd}(\xi)\not\equiv0 Q 2 n d ( ξ ) ≡ 0 ) for both the problems (20) and (21) have been shown, see [18 , 21 , 30 ] for the global existence, see [10 , 15 , 29 ] for T ε C ≥ e C / ε 2 T_\varepsilon^C\geq e^{C/\varepsilon^2} T ε C ≥ e C / ε 2 and [31 ] for T ε D ≥ e C / ε 2 T_\varepsilon^D\geq e^{C/\varepsilon^2} T ε D ≥ e C / ε 2 , respectively.
When n = 2 n=2 n = 2 and Q α β μ ≠ 0 Q^{\alpha\beta\mu}\neq0 Q α β μ = 0 for some α , β , μ = 0 , 1 , 2 \alpha,\beta,\mu=0,1,2 α , β , μ = 0 , 1 , 2 , as is well known from [1 , 2 , 3 , 14 ] that T ε C = ∞ T_\varepsilon^C=\infty T ε C = ∞ for Case 1 (i.e., Q 1 s t ( ξ ) ≡ 0 Q_{1st}(\xi)\equiv0 Q 1 s t ( ξ ) ≡ 0 and Q 2 n d ( ξ ) ≡ 0 Q_{2nd}(\xi)\equiv0 Q 2 n d ( ξ ) ≡ 0 ), the lifespan verifies lim ε → 0 + ε T ε C = τ 0 > 0 \lim_{\varepsilon\to0+}\varepsilon\sqrt{T_\varepsilon^C}=\tau_0>0 lim ε → 0 + ε T ε C = τ 0 > 0 for Case 2 (namely, Q 1 s t ( ξ ) ≢ 0 Q_{1st}(\xi)\not\equiv0 Q 1 s t ( ξ ) ≡ 0 in (4) ), lim ε → 0 + ε 2 ln T ε C = τ ~ 0 > 0 \lim_{\varepsilon\to0+}\varepsilon^2\ln T_\varepsilon^C=\widetilde{\tau}_0>0 lim ε → 0 + ε 2 ln T ε C = τ 0 > 0 for Case 3 (that is, Q 1 s t ( ξ ) ≡ 0 Q_{1st}(\xi)\equiv0 Q 1 s t ( ξ ) ≡ 0 but Q 2 n d ( ξ ) ≢ 0 Q_{2nd}(\xi)\not\equiv0 Q 2 n d ( ξ ) ≡ 0 in (5) ), respectively. For the exterior domain problem (21) , T ε D = ∞ T_\varepsilon^D=\infty T ε D = ∞ is shown in [17 ] for Case 1 . Recently, the authors in [33 ] proved T ε D ≥ C ε 2 ∣ ln ε ∣ 3 T_\varepsilon^D\geq\frac{C}{\varepsilon^2|\ln\varepsilon|^3} T ε D ≥ ε 2 ∣ l n ε ∣ 3 C for Case 2 , which contains ∣ ln ε ∣ − 3 |\ln\varepsilon|^{-3} ∣ ln ε ∣ − 3 loss than the corresponding Cauchy problem. In [8 ] , lim inf ε → 0 + ε T ε D ≥ τ 0 \liminf_{\varepsilon\to0+}\varepsilon\sqrt{T_\varepsilon^D}\geq\tau_0 lim inf ε → 0 + ε T ε D ≥ τ 0 is established for Case 2 and lim sup ε → 0 + ε T ε D ≤ τ 0 \limsup_{\varepsilon\to0+}\varepsilon\sqrt{T_\varepsilon^D}\leq\tau_0 lim sup ε → 0 + ε T ε D ≤ τ 0 is also shown for some radial problem.
This paper focuses on the remaining Case 3 of the 2-D quasilinear wave equations in exterior domains, namely, we will prove the sharp lifespan T ε D ≥ e C / ε 2 T_\varepsilon^D\geq e^{C/\varepsilon^2} T ε D ≥ e C / ε 2 for problem (1) under the first null condition and without the second null condition.
Sketches of the proof
Some essential difficulties for studying the 2-D exterior domain problem of quasilinear wave equations have been illustrated in [17 ] . Note that the Lorentz boost fields t ∂ i + x i ∂ t t\partial_i+x_i\partial_t t ∂ i + x i ∂ t (i = 1 , 2 i=1,2 i = 1 , 2 ), which play key roles in the related Cauchy problem, are no longer applicable for the exterior domain problems since these fields contain the large normal component factor t t t , see [25 page 110] . On the other hand, it is also difficult to apply the scaling vector field t ∂ t + r ∂ r t\partial_t+r\partial_r t ∂ t + r ∂ r for the 2-D exterior domain problem since the local energy of 2-D wave equation decays in the rate ( 1 + t ) − 1 (1+t)^{-1} ( 1 + t ) − 1 , while the local energy of the related 3-D problem decays exponentially in time, one can be referred to [41 , 46 ] . It is emphasized that the exponential decay rate of local energy for the 3-D problem is essential in proving the key Lemma 2.8 of [38 ] . Moreover, as discussed in [13 Page 485] , the crucial KSS estimates in 3-D problem [25 , 38 , 39 ] fail in the corresponding 2-D problem.
When both the first and second null conditions hold, the global existence of smooth solutions to the 2-D quadratic quasilinear wave equations in exterior domains has been proved in [17 ] and the following crucial pointwise estimates and local energy decay estimates have also been established
∑ ∣ a ∣ ≤ N + 1 ∣ ∂ Z a u ∣ ≤ C ε ⟨ x ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ − 1 , \sum_{|a|\leq N+1}|\partial Z^a u|\leq C\varepsilon\langle x\rangle^{-1/2}\langle t-|x|\rangle^{-1}, ∣ a ∣ ≤ N + 1 ∑ ∣ ∂ Z a u ∣ ≤ C ε ⟨ x ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ − 1 ,
∑ ∣ a ∣ ≤ N + 1 ∣ ∂ Z a u ∣ ≤ C ε ⟨ x ⟩ − 1 / 2 ⟨ t + ∣ x ∣ ⟩ 0.001 − 1 . \sum_{|a|\leq N+1}|\partial Z^a u|\leq C\varepsilon\langle x\rangle^{-1/2}\langle t+|x|\rangle^{0.001-1}. ∣ a ∣ ≤ N + 1 ∑ ∣ ∂ Z a u ∣ ≤ C ε ⟨ x ⟩ − 1/2 ⟨ t + ∣ x ∣ ⟩ 0.001 − 1 .
∑ ∣ a ∣ ≤ N + 1 ∣ Z a u ∣ ≤ C ε ⟨ t + ∣ x ∣ ⟩ 0.001 − 1 / 2 ⟨ t − ∣ x ∣ ⟩ 0.001 − 1 / 2 , \sum_{|a|\le N+1}|Z^a u|\le C\varepsilon\langle t+|x|\rangle^{0.001-1/2}\langle t-|x|\rangle^{0.001-1/2}, ∣ a ∣ ≤ N + 1 ∑ ∣ Z a u ∣ ≤ C ε ⟨ t + ∣ x ∣ ⟩ 0.001 − 1/2 ⟨ t − ∣ x ∣ ⟩ 0.001 − 1/2 ,
∑ ∣ a ∣ ≤ 2 N − 41 ∥ ∂ a u ∥ L 2 ( K R ) ≤ C R ε ( 1 + t ) − 1 . (22) \sum_{|a|\le2N-41}\|\partial^a u\|_{L^2(\mathcal{K}_R)}\le C_R\varepsilon(1+t)^{-1}. \tag*{(22)} ∣ a ∣ ≤ 2 N − 41 ∑ ∥ ∂ a u ∥ L 2 ( K R ) ≤ C R ε ( 1 + t ) − 1 . (22)
In the paper, in order to prove Theorem 1.1 for t ≤ e κ / ε 2 t\le e^{\kappa/\varepsilon^2} t ≤ e κ / ε 2 , we intend to prove some estimates analogous to (22) through overcoming the difficulties mentioned above. However, due to the absence of the second null condition, only a slightly weaker estimate (9) can be obtained than that in the first line of (22) , which will be enough to show Theorem 1.1 . In this process, we identify the following favorable structure for the cubic nonlinearity in (1)
∑ α , β , γ , δ = 0 2 Q α β γ δ ∂ α β 2 u ∂ γ u ∂ δ u = 1 3 ∂ t [ ∑ α , β , γ , δ = 0 2 Q α β γ δ ω α ω β ω γ ω δ ( ∂ t u ) 3 ] + ∑ α , β , γ , δ = 0 2 Q α β γ δ ( ∂ α β 2 u ∂ γ u ∂ δ u − ω α ω β ω γ ω δ ∂ t 2 u ( ∂ t u ) 2 ) , (23) \sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\partial_{\alpha\beta}^{2}u\partial_{\gamma}u\partial_{\delta}u
=\frac{1}{3}\partial_t\left[\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\omega_\alpha\omega_\beta\omega_\gamma\omega_\delta(\partial_tu)^3\right]
+\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\left(\partial_{\alpha\beta}^{2}u\partial_{\gamma}u\partial_{\delta}u-\omega_\alpha\omega_\beta\omega_\gamma\omega_\delta\partial_t^2u(\partial_tu)^2\right), \tag*{(23)} α , β , γ , δ = 0 ∑ 2 Q α β γ δ ∂ α β 2 u ∂ γ u ∂ δ u = 3 1 ∂ t α , β , γ , δ = 0 ∑ 2 Q α β γ δ ω α ω β ω γ ω δ ( ∂ t u ) 3 + α , β , γ , δ = 0 ∑ 2 Q α β γ δ ( ∂ α β 2 u ∂ γ u ∂ δ u − ω α ω β ω γ ω δ ∂ t 2 u ( ∂ t u ) 2 ) , (23)
which is composed of a divergence part and a second null condition part.
Based on (23) and the proof procedure of (22) for both null conditions, we may roughly focus on the following special cubic nonlinear problem which does not contain the quadratic nonlinearity with the first null condition and the cubic nonlinearity with the second null condition
□ w = ∂ t [ ( ∂ t w ) 3 ] . (24) \Box w=\partial_t\left[(\partial_t w)^3\right]. \tag*{(24)} □ w = ∂ t [ ( ∂ t w ) 3 ] . (24)
To solve (24) , at first, through investigating the solution of (7) by some delicate analyses, we can obtain the following new estimate (see § 4 below)
∣ ∂ Z a w h o m ( t , x ) ∣ ≤ C ε ⟨ x ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ − 1 , x ∈ K . (25) |\partial Z^a w_{\mathrm{hom}}(t,x)|\le C\varepsilon\langle x\rangle^{-1/2}\langle t-|x|\rangle^{-1},\qquad x\in\mathcal{K}. \tag*{(25)} ∣ ∂ Z a w hom ( t , x ) ∣ ≤ C ε ⟨ x ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ − 1 , x ∈ K . (25)
This, together with the weighted L ∞ − L ∞ L^\infty-L^\infty L ∞ − L ∞ estimate for the solution w − w h o m w-w_{\mathrm{hom}} w − w hom of the related nonlinear wave equation with divergence form in Lemma 2.5 , yields that for t ≤ e κ / ε 2 t\le e^{\kappa/\varepsilon^2} t ≤ e κ / ε 2 ,
∣ ∂ w ( t , x ) ∣ ≤ C ε 3 ln ( 2 + t + ∣ x ∣ ) ⟨ x ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ − ( 1 − η ) , η ∈ ( 0 , 1 / 2 ) . |\partial w(t,x)|\le C\varepsilon^3\ln(2+t+|x|)\langle x\rangle^{-1/2}\langle t-|x|\rangle^{-(1-\eta)},\qquad\eta\in(0,1/2). ∣ ∂ w ( t , x ) ∣ ≤ C ε 3 ln ( 2 + t + ∣ x ∣ ) ⟨ x ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ − ( 1 − η ) , η ∈ ( 0 , 1/2 ) .
Thus, (9) can be attained for t ∈ [ 0 , e κ / ε 2 ] t\in[0,e^{\kappa/\varepsilon^2}] t ∈ [ 0 , e κ / ε 2 ] , but which does not admit the same good spacetime decay rate as in the first line of (22) due to the appearance of the large factor ln ( 2 + t + ∣ x ∣ ) \ln(2+t+|x|) ln ( 2 + t + ∣ x ∣ ) .
Next, we explain how to control the energy bounds of problem (1) for t ∈ [ 0 , e κ / ε 2 ] t\in[0,e^{\kappa/\varepsilon^2}] t ∈ [ 0 , e κ / ε 2 ] , which is one of the key points in proving Theorem 1.1 . Due to the essential differences from [17 ] , the desired energy estimate can not be established only by (9) –(11) . Indeed, in the region ∣ x ∣ ≤ t / 10 |x|\le t/10 ∣ x ∣ ≤ t /10 , it follows from (9) and the standard energy estimate for the quadratic nonlinear wave equation that
d d t E ( t ) ≤ C ε E ( t ) ( 1 + t ) 1 − δ 0 , \frac{d}{dt}E(t)\le\frac{C\varepsilon E(t)}{(1+t)^{1-\delta_0}}, d t d E ( t ) ≤ ( 1 + t ) 1 − δ 0 C εE ( t ) ,
where δ 0 = 10 − 3 \delta_0=10^{-3} δ 0 = 1 0 − 3 and E ( t ) = ∑ ∣ a ∣ ≤ N 0 ∥ ∂ ∂ a u ( t ) ∥ L 2 E(t)=\sum_{|a|\le N_0}\|\partial\partial^a u(t)\|_{L^2} E ( t ) = ∑ ∣ a ∣ ≤ N 0 ∥ ∂ ∂ a u ( t ) ∥ L 2 . Together with Gronwall’s Lemma, this yields
E ( t ) ≤ E ( 0 ) e C ε ( 1 + t ) δ 0 , E(t)\le E(0)e^{C\varepsilon(1+t)^{\delta_0}}, E ( t ) ≤ E ( 0 ) e C ε ( 1 + t ) δ 0 ,
which means that E ( t ) E(t) E ( t ) grows exponentially in time and can not be uniformly bounded for t ∈ [ 0 , e κ / ε 2 ] t \in[0,e^{\kappa/\varepsilon^2}] t ∈ [ 0 , e κ / ε 2 ] . On the other hand, if one intends to use the estimate analogous to the first line of (22) , then it holds that in the region ∣ x ∣ ≤ t / 10 |x| \le t/10 ∣ x ∣ ≤ t /10 ,
d d t E ( t ) ≤ C ε E ( t ) 1 + t and thus E ( t ) ≤ E ( 0 ) e C ε ln ( 2 + t ) = E ( 0 ) ( 2 + t ) C ε , (26) \frac{d}{dt}E(t) \le\frac{C\varepsilon E(t)}{1+t}
\quad\text{and thus}\quad
E(t) \le E(0)e^{C\varepsilon\ln(2+t)}=E(0)(2+t)^{C\varepsilon}, \tag*{(26)} d t d E ( t ) ≤ 1 + t C εE ( t ) and thus E ( t ) ≤ E ( 0 ) e C ε l n ( 2 + t ) = E ( 0 ) ( 2 + t ) C ε , (26)
which also means that E ( t ) E(t) E ( t ) is not bounded for t ∈ [ 0 , e κ / ε 2 ] t \in[0,e^{\kappa/\varepsilon^2}] t ∈ [ 0 , e κ / ε 2 ] . If one still continues to apply the energy estimate (26) as in [17 ] , then it follows from the Sobolev embedding that
∣ ∂ Z a u ( t , x ) ∣ ≤ C ε ⟨ x ⟩ − 1 / 2 ( 2 + t ) C ε . (27) |\partial Z^a u(t,x)| \le C\varepsilon\langle x\rangle^{-1/2}(2+t)^{C\varepsilon}. \tag*{(27)} ∣ ∂ Z a u ( t , x ) ∣ ≤ C ε ⟨ x ⟩ − 1/2 ( 2 + t ) C ε . (27)
Therefore, the estimate (9) can not be derived by the weighted L ∞ − L ∞ L^\infty-L^\infty L ∞ − L ∞ estimate in Lemma 2.5 with (25) and (27) . To overcome this essential difficulty, a better decay estimate (including the small factor ln ( 2 + t ) − 2 \ln(2+t)^{-2} ln ( 2 + t ) − 2 for large t t t ) will be established for the solution w h o m w_{\mathrm{hom}} w hom of (7) (see Section 4 below)
∣ ∂ Z a w h o m ( t , x ) ∣ ≲ ε ( 1 + t ) − 1 ln ( 2 + t ) − 2 , ∣ x ∣ ≤ t / 4. (28) |\partial Z^a w_{\mathrm{hom}}(t,x)| \lesssim\varepsilon(1+t)^{-1}\ln(2+t)^{-2}, \qquad|x| \le t/4. \tag*{(28)} ∣ ∂ Z a w hom ( t , x ) ∣ ≲ ε ( 1 + t ) − 1 ln ( 2 + t ) − 2 , ∣ x ∣ ≤ t /4. (28)
If the better estimate (12) is shown, then it follows from some related methods in [17 ] and (28) that for t ∈ [ 0 , e κ / ε 2 ] t \in[0,e^{\kappa/\varepsilon^2}] t ∈ [ 0 , e κ / ε 2 ] ,
d d t E ( t ) ≤ C E ( t ) ( ε ( 1 + t ) − 1 ln ( 2 + t ) − 2 + ε 2 ( 1 + t ) − 1 ) \frac{d}{dt}E(t) \le CE(t)\left(\varepsilon(1+t)^{-1}\ln(2+t)^{-2}+\varepsilon^2(1+t)^{-1}\right) d t d E ( t ) ≤ C E ( t ) ( ε ( 1 + t ) − 1 ln ( 2 + t ) − 2 + ε 2 ( 1 + t ) − 1 )
and
E ( t ) ≤ E ( 0 ) e C ( ε + ε 2 ln ( 2 + t ) ) ≤ E ( 0 ) e C ( ε + κ ) , E(t) \le E(0)e^{C(\varepsilon+\varepsilon^2\ln(2+t))} \le E(0)e^{C(\varepsilon+\kappa)}, E ( t ) ≤ E ( 0 ) e C ( ε + ε 2 l n ( 2 + t )) ≤ E ( 0 ) e C ( ε + κ ) ,
which leads to the uniform bound of the energy E ( t ) E(t) E ( t ) .
To prove the key estimate (12) , we will refine the estimate in Lemma 2.5 in the region ∣ x ∣ ≤ t / 10 |x| \le t/10 ∣ x ∣ ≤ t /10 by the Fourier method and the Littlewood-Paley theory (see Lemma 4.2). Namely, for the solution w w w of the nonhomogeneous linear Cauchy problem
{ □ w = F ( t , x ) , ( t , x ) ∈ [ 0 , + ∞ ) × R 2 , ( w , ∂ t w ) ( 0 , x ) = ( 0 , 0 ) , x ∈ R 2 , \left\{
\begin{aligned}
\Box w &= F(t,x), && (t,x)\in[0,+\infty)\times\mathbb{R}^2,\\
(w,\partial_t w)(0,x) &= (0,0), && x\in\mathbb{R}^2,
\end{aligned}
\right. { □ w ( w , ∂ t w ) ( 0 , x ) = F ( t , x ) , = ( 0 , 0 ) , ( t , x ) ∈ [ 0 , + ∞ ) × R 2 , x ∈ R 2 ,
where supp y F ( s , y ) ⊂ { y : 9 s 10 ≤ ∣ y ∣ ≤ s + M 0 } \operatorname{supp}_y F(s,y)\subset\{y:\frac{9s}{10}\le|y|\le s+M_0\} supp y F ( s , y ) ⊂ { y : 10 9 s ≤ ∣ y ∣ ≤ s + M 0 } and M 0 > 0 M_0>0 M 0 > 0 , then it holds that for any constant μ > 0 \mu>0 μ > 0 and ∣ x ∣ ≤ t 10 |x|\le\frac{t}{10} ∣ x ∣ ≤ 10 t ,
∣ ∂ t w ( t , x ) ∣ ≲ 1 ⟨ t ⟩ sup ( s , y ) ∈ [ 0 , t ] × R 2 ⟨ y ⟩ 1 / 2 ⟨ s + ∣ y ∣ ⟩ ( min { ⟨ y ⟩ , ⟨ s − ∣ y ∣ ⟩ } ) 1 + μ ∣ ( 1 − Δ ) 2 F ( s , y ) ∣ . (29) |\partial_t w(t,x)|\lesssim\frac{1}{\langle t\rangle}\sup_{(s,y)\in[0,t]\times\mathbb{R}^2}\frac{\langle y\rangle^{1/2}\langle s+|y|\rangle}{\left(\min\{\langle y\rangle,\langle s-|y|\rangle\}\right)^{1+\mu}}|(1-\Delta)^2F(s,y)|. \tag*{(29)} ∣ ∂ t w ( t , x ) ∣ ≲ ⟨ t ⟩ 1 ( s , y ) ∈ [ 0 , t ] × R 2 sup ( min {⟨ y ⟩ , ⟨ s − ∣ y ∣ ⟩} ) 1 + μ ⟨ y ⟩ 1/2 ⟨ s + ∣ y ∣ ⟩ ∣ ( 1 − Δ ) 2 F ( s , y ) ∣. (29)
which eliminates the large factor ln ( 2 + t + ∣ x ∣ ) \ln(2+t+|x|) ln ( 2 + t + ∣ x ∣ ) on the right hand side of (29) appeared in Lemma 2.5. Combining (29) with the energy estimate of v = u − w h o m v=u-w_{\mathrm{hom}} v = u − w hom (see Section 6) and the Sobolev embedding, one has the following important pointwise estimate (see (214) )
∑ ∣ a ∣ ≤ 2 N − 15 ∣ ∂ Z a v ( t , x ) ∣ ≤ C ε ⟨ x ⟩ − 1 / 2 . (30) \sum_{|a|\le2N-15}|\partial Z^a v(t,x)|\le C\varepsilon\langle x\rangle^{-1/2}. \tag*{(30)} ∣ a ∣ ≤ 2 N − 15 ∑ ∣ ∂ Z a v ( t , x ) ∣ ≤ C ε ⟨ x ⟩ − 1/2 . (30)
Based on (30) , through applying some related methods for treating the exterior domain problem of the quadratic quasilineat wave equation with null forms in [17 ] (see Lemmas 2.11–2.12 below) and utilizing the basic structure expression (23) , the pointwise estimates (9) -(12) will be obtained by the weighted L ∞ − L ∞ L^\infty-L^\infty L ∞ − L ∞ estimates of the IBVP established in § 5 following the steps below
∑ ∣ a ∣ ≤ 2 N − 15 ∣ Z a v ∣ ≤ C ε ⟨ t + ∣ x ∣ ⟩ 0 + ⇒ ∑ ∣ a ∣ ≤ 2 N − 17 ∣ ∂ ˉ Z a v ∣ ≤ C ε ⟨ x ⟩ − 1 + , ∣ x ∣ ≥ 1 + t / 2 , ⇓ ∑ ∣ a ∣ ≤ 2 N − 21 ∣ Z a v ∣ ≤ C ε ⟨ t + ∣ x ∣ ⟩ − 1 / 2 , ∑ ∣ a ∣ ≤ 2 N − 21 ∣ ∂ Z a v ∣ ≤ C ε ⟨ x ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ − 1 / 2 ⟨ t ⟩ 0 + , ⇓ ∑ ∣ a ∣ ≤ 2 N − 27 ∣ Z a v ∣ ≤ C ε ⟨ t + ∣ x ∣ ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ − 0.4 , ∑ ∣ a ∣ ≤ 2 N − 23 ∣ ∂ ˉ Z a v ∣ ≤ C ε ⟨ x ⟩ − 3 / 2 + , ∣ x ∣ ≥ 1 + t / 2 , ∑ ∣ a ∣ ≤ 2 N − 25 ∥ ∂ a v ∥ L 2 ( K R ) ≤ C ε 2 ( 1 + t ) − 1 , ⇓ ∑ ∣ a ∣ ≤ 2 N − 36 ∣ ∂ Z a v ∣ ≤ C ε ⟨ x ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ − 1 + ⟨ t ⟩ 0 + , ⇓ ∑ ∣ a ∣ ≤ 2 N − 46 ∣ ∂ Z a v ∣ ≤ C ε 2 ⟨ t ⟩ − 1 , ∣ x ∣ ≤ t / 10 , ∑ ∣ a ∣ ≤ 2 N − 46 ∣ ∂ Z a v ∣ ≤ C ε ⟨ x ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ − 1 + , ⇓ ∑ ∣ a ∣ ≤ 2 N − 46 ∣ ∂ ˉ Z a v ∣ ≤ C ε ⟨ x ⟩ − 1 / 2 ⟨ t + ∣ x ∣ ⟩ − 1 + ⇒ ∑ ∣ a ∣ ≤ 2 N − 52 ∣ Z a v ∣ ≤ C ε ⟨ t + ∣ x ∣ ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ − 1 / 2 + . \begin{aligned}
\sum_{|a|\le2N-15}|Z^a v| &\le C\varepsilon\langle t+|x|\rangle^{0+}
\Rightarrow
\sum_{|a|\le2N-17}|\bar{\partial}Z^a v| \le C\varepsilon\langle x\rangle^{-1+},\ |x|\ge1+t/2,\\
\Downarrow\\
\sum_{|a|\le2N-21}|Z^a v| &\le C\varepsilon\langle t+|x|\rangle^{-1/2},
\qquad
\sum_{|a|\le2N-21}|\partial Z^a v|
\le C\varepsilon\langle x\rangle^{-1/2}\langle t-|x|\rangle^{-1/2}\langle t\rangle^{0+},\\
\Downarrow\\
\sum_{|a|\le2N-27}|Z^a v| &\le C\varepsilon\langle t+|x|\rangle^{-1/2}\langle t-|x|\rangle^{-0.4},
\qquad
\sum_{|a|\le2N-23}|\bar{\partial}Z^a v|
\le C\varepsilon\langle x\rangle^{-3/2+},\ |x|\ge1+t/2,\\
\sum_{|a|\le2N-25}\|\partial^a v\|_{L^2(K_R)} &\le C\varepsilon^2(1+t)^{-1},\\
\Downarrow\\
\sum_{|a|\le2N-36}|\partial Z^a v|
&\le C\varepsilon\langle x\rangle^{-1/2}\langle t-|x|\rangle^{-1+}\langle t\rangle^{0+},\\
\Downarrow\\
\sum_{|a|\le2N-46}|\partial Z^a v|
&\le C\varepsilon^2\langle t\rangle^{-1},\ |x|\le t/10,
\qquad
\sum_{|a|\le2N-46}|\partial Z^a v|
\le C\varepsilon\langle x\rangle^{-1/2}\langle t-|x|\rangle^{-1+},\\
\Downarrow\\
\sum_{|a|\le2N-46}|\bar{\partial}Z^a v|
&\le C\varepsilon\langle x\rangle^{-1/2}\langle t+|x|\rangle^{-1+}
\Rightarrow
\sum_{|a|\le2N-52}|Z^a v|
\le C\varepsilon\langle t+|x|\rangle^{-1/2}\langle t-|x|\rangle^{-1/2+}.
\end{aligned} ∣ a ∣ ≤ 2 N − 15 ∑ ∣ Z a v ∣ ⇓ ∣ a ∣ ≤ 2 N − 21 ∑ ∣ Z a v ∣ ⇓ ∣ a ∣ ≤ 2 N − 27 ∑ ∣ Z a v ∣ ∣ a ∣ ≤ 2 N − 25 ∑ ∥ ∂ a v ∥ L 2 ( K R ) ⇓ ∣ a ∣ ≤ 2 N − 36 ∑ ∣ ∂ Z a v ∣ ⇓ ∣ a ∣ ≤ 2 N − 46 ∑ ∣ ∂ Z a v ∣ ⇓ ∣ a ∣ ≤ 2 N − 46 ∑ ∣ ∂ ˉ Z a v ∣ ≤ C ε ⟨ t + ∣ x ∣ ⟩ 0 + ⇒ ∣ a ∣ ≤ 2 N − 17 ∑ ∣ ∂ ˉ Z a v ∣ ≤ C ε ⟨ x ⟩ − 1 + , ∣ x ∣ ≥ 1 + t /2 , ≤ C ε ⟨ t + ∣ x ∣ ⟩ − 1/2 , ∣ a ∣ ≤ 2 N − 21 ∑ ∣ ∂ Z a v ∣ ≤ C ε ⟨ x ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ − 1/2 ⟨ t ⟩ 0 + , ≤ C ε ⟨ t + ∣ x ∣ ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ − 0.4 , ∣ a ∣ ≤ 2 N − 23 ∑ ∣ ∂ ˉ Z a v ∣ ≤ C ε ⟨ x ⟩ − 3/2 + , ∣ x ∣ ≥ 1 + t /2 , ≤ C ε 2 ( 1 + t ) − 1 , ≤ C ε ⟨ x ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ − 1 + ⟨ t ⟩ 0 + , ≤ C ε 2 ⟨ t ⟩ − 1 , ∣ x ∣ ≤ t /10 , ∣ a ∣ ≤ 2 N − 46 ∑ ∣ ∂ Z a v ∣ ≤ C ε ⟨ x ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ − 1 + , ≤ C ε ⟨ x ⟩ − 1/2 ⟨ t + ∣ x ∣ ⟩ − 1 + ⇒ ∣ a ∣ ≤ 2 N − 52 ∑ ∣ Z a v ∣ ≤ C ε ⟨ t + ∣ x ∣ ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ − 1/2 + .
whose details are given in § 7 . Due to these obtained conclusions, Theorem 1.1 is eventually proved.
The paper is organized as follows. In § 2 , several basic lemmas, some important results and good properties of null structures are stated. In § 3 , the key Lemma 2.5 is proved in details. Two crucial estimates are derived in § 4 . In § 5 , some new pointwise estimates for the exterior domain problem are established. In § 6 , the energy estimates are shown to remain uniformly bounded for t ∈ [ 0 , e κ / ε 2 ] t\in[0,e^{\kappa/\varepsilon^2}] t ∈ [ 0 , e κ / ε 2 ] . In § 7 , the key pointwise estimates are established step by step and the proof of Theorem 1.1 is finished.
Notations:
K = R 2 ∖ O \mathcal{K}=\mathbb{R}^2\setminus\mathcal{O} K = R 2 ∖ O , the obstacle O ⊂ R 2 \mathcal{O}\subset\mathbb{R}^2 O ⊂ R 2 is compact and contains the origin, K R = K ∩ { x : ∣ x ∣ ≤ R } \mathcal{K}_R=\mathcal{K}\cap\{x:|x|\le R\} K R = K ∩ { x : ∣ x ∣ ≤ R } , R > 1 R>1 R > 1 is a fixed constant which may be changed in different places.
The cutoff function χ [ a , b ] ( x ) = χ [ a , b ] ( ∣ x ∣ ) \chi_{[a,b]}(x)=\chi_{[a,b]}(|x|) χ [ a , b ] ( x ) = χ [ a , b ] ( ∣ x ∣ ) , χ [ a , b ] ( s ) ∈ C ∞ ( R ) \chi_{[a,b]}(s)\in C^\infty(\mathbb{R}) χ [ a , b ] ( s ) ∈ C ∞ ( R ) with a , b ∈ R a,b\in\mathbb{R} a , b ∈ R , 0 < a < b 0<a<b 0 < a < b , 0 ≤ χ [ a , b ] ( s ) ≤ 1 0\le\chi_{[a,b]}(s)\le1 0 ≤ χ [ a , b ] ( s ) ≤ 1 and
χ [ a , b ] ( s ) = { 0 , s ≤ a , 1 , s ≥ b . \chi_{[a,b]}(s)=
\begin{cases}
0, & s\le a,\\
1, & s\ge b.
\end{cases} χ [ a , b ] ( s ) = { 0 , 1 , s ≤ a , s ≥ b .
N 0 = { 0 , 1 , 2 , ⋯ } \mathbb{N}_0=\{0,1,2,\cdots\} N 0 = { 0 , 1 , 2 , ⋯ } and N = { 1 , 2 , ⋯ } \mathbb{N}=\{1,2,\cdots\} N = { 1 , 2 , ⋯ } .
∂ 0 = ∂ t \partial_0=\partial_t ∂ 0 = ∂ t , ∂ 1 = ∂ x 1 \partial_1=\partial_{x_1} ∂ 1 = ∂ x 1 , ∂ 2 = ∂ x 2 \partial_2=\partial_{x_2} ∂ 2 = ∂ x 2 , ∂ x = ∇ x = ( ∂ 1 , ∂ 2 ) \partial_x=\nabla_x=(\partial_1,\partial_2) ∂ x = ∇ x = ( ∂ 1 , ∂ 2 ) , ∂ = ( ∂ t , ∂ 1 , ∂ 2 ) \partial=(\partial_t,\partial_1,\partial_2) ∂ = ( ∂ t , ∂ 1 , ∂ 2 ) .
For ∣ x ∣ > 0 |x|>0 ∣ x ∣ > 0 , define ∂ ˉ i = ∂ i + x i ∣ x ∣ ∂ t \bar{\partial}_i=\partial_i+\frac{x_i}{|x|}\partial_t ∂ ˉ i = ∂ i + ∣ x ∣ x i ∂ t (i = 1 , 2 i=1,2 i = 1 , 2 ) and ∂ ˉ = ( ∂ ˉ 1 , ∂ ˉ 2 ) \bar{\partial}=(\bar{\partial}_1,\bar{\partial}_2) ∂ ˉ = ( ∂ ˉ 1 , ∂ ˉ 2 ) .
r = ∣ x ∣ r=|x| r = ∣ x ∣ , x = r ( cos θ , sin θ ) x=r(\cos\theta,\sin\theta) x = r ( cos θ , sin θ ) with θ ∈ [ 0 , 2 π ] \theta\in[0,2\pi] θ ∈ [ 0 , 2 π ] , Ω i j = x i ∂ j − x j ∂ i \Omega_{ij}=x_i\partial_j-x_j\partial_i Ω ij = x i ∂ j − x j ∂ i with i , j = 1 , 2 i,j=1,2 i , j = 1 , 2 , Ω = ∂ θ = Ω 12 \Omega=\partial_\theta=\Omega_{12} Ω = ∂ θ = Ω 12 , Z = { Z 1 , Z 2 , Z 3 , Z 4 } = { ∂ t , ∂ 1 , ∂ 2 , Ω } Z=\{Z_1,Z_2,Z_3,Z_4\}=\{\partial_t,\partial_1,\partial_2,\Omega\} Z = { Z 1 , Z 2 , Z 3 , Z 4 } = { ∂ t , ∂ 1 , ∂ 2 , Ω } , Z ~ = χ [ 1 / 2 , 1 ] ( x ) Z \widetilde{Z}=\chi_{[1/2,1]}(x)Z Z = χ [ 1/2 , 1 ] ( x ) Z .
∂ x a = ∂ 1 a 1 ∂ 2 a 2 \partial_x^a=\partial_1^{a_1}\partial_2^{a_2} ∂ x a = ∂ 1 a 1 ∂ 2 a 2 for a ∈ N 0 2 a\in\mathbb{N}_0^2 a ∈ N 0 2 , ∂ a = ∂ t a 0 ∂ 1 a 1 ∂ 2 a 2 \partial^a=\partial_t^{a_0}\partial_1^{a_1}\partial_2^{a_2} ∂ a = ∂ t a 0 ∂ 1 a 1 ∂ 2 a 2 for a ∈ N 0 3 a\in\mathbb{N}_0^3 a ∈ N 0 3 and Z a = Z 1 a 1 Z 2 a 2 Z 3 a 3 Z 4 a 4 Z^a=Z_1^{a_1}Z_2^{a_2}Z_3^{a_3}Z_4^{a_4} Z a = Z 1 a 1 Z 2 a 2 Z 3 a 3 Z 4 a 4 for a ∈ N 0 4 a\in\mathbb{N}_0^4 a ∈ N 0 4 .
For f ≥ 0 f\geq0 f ≥ 0 and g ≥ 0 g\geq0 g ≥ 0 , f ≲ g f\lesssim g f ≲ g or g ≳ f g\gtrsim f g ≳ f denotes f ≤ C g f\leq Cg f ≤ C g for a generic constant C > 0 C>0 C > 0 independent of ε > 0 \varepsilon>0 ε > 0 , and f ≈ g f\approx g f ≈ g means f ≲ g f\lesssim g f ≲ g and g ≲ f g\lesssim f g ≲ f .
Denote ∥ u ( t ) ∥ = ∥ u ( t , ⋅ ) ∥ \|u(t)\|=\|u(t,\cdot)\| ∥ u ( t ) ∥ = ∥ u ( t , ⋅ ) ∥ with norm ∥ ⋅ ∥ = ∥ ⋅ ∥ L 2 ( K ) \|\cdot\|=\|\cdot\|_{L^2(K)} ∥ ⋅ ∥ = ∥ ⋅ ∥ L 2 ( K ) , ∥ ⋅ ∥ L 2 ( K R ) \|\cdot\|_{L^2(K_R)} ∥ ⋅ ∥ L 2 ( K R ) , ∥ ⋅ ∥ H k ( K ) \|\cdot\|_{H^k(K)} ∥ ⋅ ∥ H k ( K ) .
W μ , ν ( t , x ) = ⟨ t + ∣ x ∣ ⟩ μ ( min { ⟨ x ⟩ , ⟨ t − ∣ x ∣ ⟩ } ) ν W_{\mu,\nu}(t,x)=\langle t+|x|\rangle^\mu(\min\{\langle x\rangle,\langle t-|x|\rangle\})^\nu W μ , ν ( t , x ) = ⟨ t + ∣ x ∣ ⟩ μ ( min {⟨ x ⟩ , ⟨ t − ∣ x ∣ ⟩} ) ν for μ , ν ∈ R \mu,\nu\in\mathbb{R} μ , ν ∈ R .
∣ Z ≤ j f ∣ = ( ∑ 0 ≤ ∣ a ∣ ≤ j ∣ Z a f ∣ 2 ) 1 / 2 |Z^{\leq j}f|=\left(\sum_{0\leq|a|\leq j}|Z^af|^2\right)^{1/2} ∣ Z ≤ j f ∣ = ( ∑ 0 ≤ ∣ a ∣ ≤ j ∣ Z a f ∣ 2 ) 1/2 and Z ≤ 1 f = ( f , Z f ) Z^{\leq1}f=(f,Zf) Z ≤ 1 f = ( f , Z f ) .
Preliminaries
Several useful lemmas
In this subsection, some lemmas on the local energy decay estimate, the elliptic estimate, Gronwall’s lemma and the Sobolev embedding are listed.
Lemma 2.1. Suppose that the obstacle O \mathcal{O} O is star-shaped and K = R 2 ∖ O K=\mathbb{R}^2\setminus\mathcal{O} K = R 2 ∖ O . Let w w w be the solution of the IBVP
{ □ w = G ( t , x ) , ( t , x ) ∈ [ 0 , ∞ ) × K , w = 0 , ( t , x ) ∈ [ 0 , ∞ ) × ∂ K , ( w , ∂ t w ) ( 0 , x ) = ( w 0 , w 1 ) ( x ) , x ∈ K , (31) \begin{cases}
\Box w=G(t,x), & (t,x)\in[0,\infty)\times K,\\
w=0, & (t,x)\in[0,\infty)\times\partial K,\\
(w,\partial_tw)(0,x)=(w_0,w_1)(x), & x\in K, \tag*{(31)}
\end{cases} ⎩ ⎨ ⎧ □ w = G ( t , x ) , w = 0 , ( w , ∂ t w ) ( 0 , x ) = ( w 0 , w 1 ) ( x ) , ( t , x ) ∈ [ 0 , ∞ ) × K , ( t , x ) ∈ [ 0 , ∞ ) × ∂ K , x ∈ K , (31)
where supp x ( w 0 ( x ) , w 1 ( x ) , G ( t , x ) ) ⊂ { x : ∣ x ∣ ≤ R 1 } \operatorname{supp}_x(w_0(x),w_1(x),G(t,x))\subset\{x:|x|\leq R_1\} supp x ( w 0 ( x ) , w 1 ( x ) , G ( t , x )) ⊂ { x : ∣ x ∣ ≤ R 1 } for R 1 > 1 R_1>1 R 1 > 1 . Then for fixed constant R > 1 R>1 R > 1 , m ∈ N m\in\mathbb{N} m ∈ N and η ∈ ( 0 , 1 ] \eta\in(0,1] η ∈ ( 0 , 1 ] , it holds
∑ ∣ a ∣ ≤ m ∥ ∂ a w ∥ L 2 ( K R ) ≲ ⟨ t ⟩ − 1 ln − 2 ( 2 + t ) ( ∥ w 0 ∥ H m ( K ) + ∥ w 1 ∥ H m − 1 ( K ) ) + ⟨ t ⟩ − η ∑ ∣ a ∣ ≤ m − 1 sup 0 ≤ s ≤ t ⟨ s ⟩ η ∥ ∂ a G ( s ) ∥ L 2 ( K ) . (32) \sum_{|a|\leq m}\|\partial^aw\|_{L^2(K_R)}
\lesssim\langle t\rangle^{-1}\ln^{-2}(2+t)(\|w_0\|_{H^m(K)}+\|w_1\|_{H^{m-1}(K)})
+\langle t\rangle^{-\eta}\sum_{|a|\leq m-1}\sup_{0\leq s\leq t}\langle s\rangle^\eta\|\partial^aG(s)\|_{L^2(K)}. \tag*{(32)} ∣ a ∣ ≤ m ∑ ∥ ∂ a w ∥ L 2 ( K R ) ≲ ⟨ t ⟩ − 1 ln − 2 ( 2 + t ) ( ∥ w 0 ∥ H m ( K ) + ∥ w 1 ∥ H m − 1 ( K ) ) + ⟨ t ⟩ − η ∣ a ∣ ≤ m − 1 ∑ 0 ≤ s ≤ t sup ⟨ s ⟩ η ∥ ∂ a G ( s ) ∥ L 2 ( K ) . (32)
Proof. One can see the end of Section 3 in [46 ] or Lemma 3.2 of [31 ] .
Lemma 2.2 (Lemma 3.1 of [31 ] ). Let w ∈ H j ( K ) w \in H^j(\mathcal{K}) w ∈ H j ( K ) and w ∣ ∂ K = 0 w|_{\partial\mathcal{K}}=0 w ∣ ∂ K = 0 with integer j ≥ 2 j \ge2 j ≥ 2 . Then for any fixed constant R > 1 R>1 R > 1 and multi-index a ∈ N 0 2 a\in\mathbb{N}_0^2 a ∈ N 0 2 with 2 ≤ ∣ a ∣ ≤ j 2\le|a|\le j 2 ≤ ∣ a ∣ ≤ j , one has
∥ ∂ x a w ∥ L 2 ( K ) ≲ ∥ Δ w ∥ H ∣ a ∣ − 2 ( K ) + ∥ w ∥ H ∣ a ∣ − 1 ( K R + 1 ) . (33) \|\partial^a_x w\|_{L^2(\mathcal{K})} \lesssim\|\Delta w\|_{H^{|a|-2}(\mathcal{K})}+\|w\|_{H^{|a|-1}(\mathcal{K}_{R+1})}. \tag*{(33)} ∥ ∂ x a w ∥ L 2 ( K ) ≲ ∥Δ w ∥ H ∣ a ∣ − 2 ( K ) + ∥ w ∥ H ∣ a ∣ − 1 ( K R + 1 ) . (33)
Lemma 2.3. For any positive function f ( t ) , g ( t ) , h ( t ) ≥ 0 f(t),g(t),h(t)\ge0 f ( t ) , g ( t ) , h ( t ) ≥ 0 satisfying
f ( t ) ≤ g ( t ) + ∫ 0 t f ( s ) h ( s ) d s , f(t)\le g(t)+\int_0^t f(s)h(s)\,ds, f ( t ) ≤ g ( t ) + ∫ 0 t f ( s ) h ( s ) d s , then it holds
f ( t ) ≤ e ∫ 0 t h ( s ) d s sup 0 ≤ s ≤ t g ( s ) . f(t)\le e^{\int_0^t h(s)\,ds}\sup_{0\le s\le t}g(s). f ( t ) ≤ e ∫ 0 t h ( s ) d s 0 ≤ s ≤ t sup g ( s ) .
Proof. The proof follows from Gronwall’s inequality and direct computations.
Lemma 2.4 (Lemma 3.6 of [31 ] ). Given any function f ( x ) ∈ C 0 2 ( K ‾ ) f(x)\in C^2_0(\overline{\mathcal{K}}) f ( x ) ∈ C 0 2 ( K ) , one has that for all x ∈ K x\in\mathcal{K} x ∈ K ,
⟨ x ⟩ 1 / 2 ∣ f ( x ) ∣ ≲ ∑ ∣ a ∣ ≤ 2 ∥ Z a f ∥ L 2 ( K ) . (34) \langle x\rangle^{1/2}|f(x)|\lesssim\sum_{|a|\le2}\|Z^a f\|_{L^2(\mathcal{K})}. \tag*{(34)} ⟨ x ⟩ 1/2 ∣ f ( x ) ∣ ≲ ∣ a ∣ ≤ 2 ∑ ∥ Z a f ∥ L 2 ( K ) . (34)
Some pointwise estimates of Cauchy problem
In order to prove Theorem 1.1 , the following spacetime weighted estimates for the solution of 2-D linear wave equation will be crucial.
Lemma 2.5. Let w w w be the solution of the Cauchy problem
{ □ w = ∑ α = 0 2 ∂ α H α ( t , x ) , ( t , x ) ∈ [ 0 , ∞ ) × R 2 , ( w , ∂ t w ) ( 0 , x ) = ( 0 , 0 ) , x ∈ R 2 . \begin{cases}
\Box w=\sum_{\alpha=0}^{2}\partial_\alpha H^\alpha(t,x), & (t,x)\in[0,\infty)\times\mathbb{R}^2,\\
(w,\partial_t w)(0,x)=(0,0), & x\in\mathbb{R}^2.
\end{cases} { □ w = ∑ α = 0 2 ∂ α H α ( t , x ) , ( w , ∂ t w ) ( 0 , x ) = ( 0 , 0 ) , ( t , x ) ∈ [ 0 , ∞ ) × R 2 , x ∈ R 2 . Then for H 0 ( t , x ) ≡ 0 H^0(t,x)\equiv0 H 0 ( t , x ) ≡ 0 and η ∈ ( 0 , 1 / 2 ) \eta\in(0,1/2) η ∈ ( 0 , 1/2 ) , one has
⟨ x ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ 1 − η ∣ w ∣ ≲ ln ( 2 + t + ∣ x ∣ ) ∑ i = 1 2 ∑ ∣ a ∣ ≤ 1 sup ( s , y ) ∈ [ 0 , t ] × R 2 ⟨ y ⟩ 1 / 2 W 1 , 1 ( s , y ) ∣ Z a H i ( s , y ) ∣ . (35) \langle x\rangle^{1/2}\langle t-|x|\rangle^{1-\eta}|w|
\lesssim\ln(2+t+|x|)\sum_{i=1}^{2}\sum_{|a|\le1}\sup_{(s,y)\in[0,t]\times\mathbb{R}^2}\langle y\rangle^{1/2}W_{1,1}(s,y)|Z^aH^i(s,y)|. \tag*{(35)} ⟨ x ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ 1 − η ∣ w ∣ ≲ ln ( 2 + t + ∣ x ∣ ) i = 1 ∑ 2 ∣ a ∣ ≤ 1 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 sup ⟨ y ⟩ 1/2 W 1 , 1 ( s , y ) ∣ Z a H i ( s , y ) ∣. (35) where W μ , ν ( t , x ) = ⟨ t + ∣ x ∣ ⟩ μ ( min { ⟨ x ⟩ , ⟨ t − ∣ x ∣ ⟩ } ) ν W_{\mu,\nu}(t,x)=\langle t+|x|\rangle^\mu(\min\{\langle x\rangle,\langle t-|x|\rangle\})^\nu W μ , ν ( t , x ) = ⟨ t + ∣ x ∣ ⟩ μ ( min {⟨ x ⟩ , ⟨ t − ∣ x ∣ ⟩} ) ν ; for general H 0 ( t , x ) ≢ 0 H^0(t,x)\not\equiv0 H 0 ( t , x ) ≡ 0 and η ∈ ( 0 , 1 / 2 ) \eta\in(0,1/2) η ∈ ( 0 , 1/2 ) , it holds that
⟨ t + ∣ x ∣ ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ 1 / 2 − η ∣ w ∣ ≲ ln ( 2 + t + ∣ x ∣ ) ∑ α = 0 2 ∑ ∣ a ∣ ≤ 1 sup ( s , y ) ∈ [ 0 , t ] × R 2 ⟨ y ⟩ 1 / 2 W 1 , 1 ( s , y ) ∣ Z a H α ( s , y ) ∣ . (36) \langle t+|x|\rangle^{1/2}\langle t-|x|\rangle^{1/2-\eta}|w|
\lesssim\ln(2+t+|x|)\sum_{\alpha=0}^{2}\sum_{|a|\le1}\sup_{(s,y)\in[0,t]\times\mathbb{R}^2}\langle y\rangle^{1/2}W_{1,1}(s,y)|Z^aH^\alpha(s,y)|. \tag*{(36)} ⟨ t + ∣ x ∣ ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ 1/2 − η ∣ w ∣ ≲ ln ( 2 + t + ∣ x ∣ ) α = 0 ∑ 2 ∣ a ∣ ≤ 1 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 sup ⟨ y ⟩ 1/2 W 1 , 1 ( s , y ) ∣ Z a H α ( s , y ) ∣. (36)
Proof. Its proof will be postponed to § 3 .
As a corollary of Lemma 2.5 , we have
Corollary 2.6. Let w w w be the solution of the Cauchy problem
{ □ w = F ( t , x ) , ( t , x ) ∈ [ 0 , ∞ ) × R 2 , ( w , ∂ t w ) ( 0 , x ) = ( 0 , 0 ) , x ∈ R 2 . \begin{cases}
\Box w = F(t,x), & (t,x) \in[0,\infty) \times\mathbb{R}^{2},\\
(w,\partial_{t}w)(0,x) = (0,0), & x \in\mathbb{R}^{2}.
\end{cases} { □ w = F ( t , x ) , ( w , ∂ t w ) ( 0 , x ) = ( 0 , 0 ) , ( t , x ) ∈ [ 0 , ∞ ) × R 2 , x ∈ R 2 . Then for η ∈ ( 0 , 1 / 2 ) \eta\in(0,1/2) η ∈ ( 0 , 1/2 ) , it holds that
⟨ x ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ 1 − η ∣ ∂ t 2 w ∣ ≲ ln ( 2 + t + ∣ x ∣ ) ∑ ∣ a ∣ ≤ 2 sup ( s , y ) ∈ [ 0 , t ] × R 2 ⟨ y ⟩ 1 / 2 W 1 , 1 ( s , y ) ∣ Z a F ( s , y ) ∣ . (37) \langle x\rangle^{1/2}\langle t-|x|\rangle^{1-\eta}|\partial_{t}^{2}w| \lesssim\ln(2+t+|x|)\sum_{|a|\leq2}\sup_{(s,y)\in[0,t]\times\mathbb{R}^{2}}\langle y\rangle^{1/2}W_{1,1}(s,y)|Z^{a}F(s,y)|. \tag*{(37)} ⟨ x ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ 1 − η ∣ ∂ t 2 w ∣ ≲ ln ( 2 + t + ∣ x ∣ ) ∣ a ∣ ≤ 2 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 sup ⟨ y ⟩ 1/2 W 1 , 1 ( s , y ) ∣ Z a F ( s , y ) ∣. (37)
Proof. Note that
w ( t , x ) = ∫ 0 t sin ( t − s ) ∣ ∇ ∣ ∣ ∇ ∣ F ( s ) d s , w(t,x)=\int_{0}^{t}\frac{\sin(t-s)|\nabla|}{|\nabla|}F(s)\,ds, w ( t , x ) = ∫ 0 t ∣∇∣ sin ( t − s ) ∣∇∣ F ( s ) d s ,
∂ t 2 w ( t , x ) = F ( t , x ) + Δ w ( t , x ) = F ( t , x ) + ∫ 0 t sin ( t − s ) ∣ ∇ ∣ ∣ ∇ ∣ Δ F ( s ) d s . (38) \partial_{t}^{2}w(t,x)=F(t,x)+\Delta w(t,x)=F(t,x)+\int_{0}^{t}\frac{\sin(t-s)|\nabla|}{|\nabla|}\Delta F(s)\,ds. \tag*{(38)} ∂ t 2 w ( t , x ) = F ( t , x ) + Δ w ( t , x ) = F ( t , x ) + ∫ 0 t ∣∇∣ sin ( t − s ) ∣∇∣ Δ F ( s ) d s . (38)
In addition, (35) can be written as
⟨ x ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ 1 − η ∣ ∫ 0 t sin ( t − s ) ∣ ∇ ∣ ∣ ∇ ∣ ∇ G ( s ) d s ∣ ≲ ln ( 2 + t + ∣ x ∣ ) ∑ ∣ a ∣ ≤ 1 sup ( s , y ) ∈ [ 0 , t ] × R 2 ⟨ y ⟩ 1 / 2 W 1 , 1 ( s , y ) ∣ Z a G ( s , y ) ∣ . (39) \langle x\rangle^{1/2}\langle t-|x|\rangle^{1-\eta}\left|\int_{0}^{t}\frac{\sin(t-s)|\nabla|}{|\nabla|}\nabla G(s)\,ds\right|
\lesssim\ln(2+t+|x|)\sum_{|a|\leq1}\sup_{(s,y)\in[0,t]\times\mathbb{R}^{2}}\langle y\rangle^{1/2}W_{1,1}(s,y)|Z^{a}G(s,y)|. \tag*{(39)} ⟨ x ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ 1 − η ∫ 0 t ∣∇∣ sin ( t − s ) ∣∇∣ ∇ G ( s ) d s ≲ ln ( 2 + t + ∣ x ∣ ) ∣ a ∣ ≤ 1 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 sup ⟨ y ⟩ 1/2 W 1 , 1 ( s , y ) ∣ Z a G ( s , y ) ∣. (39)
Combining (39) with (38) yields (37) . □
Lemma 2.7. Let w w w be the solution of the Cauchy problem
{ □ w = H ( t , x ) , ( t , x ) ∈ [ 0 , ∞ ) × R 2 , ( w , ∂ t w ) ( 0 , x ) = ( w 0 , w 1 ) , x ∈ R 2 . \begin{cases}
\Box w = H(t,x), & (t,x) \in[0,\infty) \times\mathbb{R}^{2},\\
(w,\partial_{t}w)(0,x) = (w_{0},w_{1}), & x \in\mathbb{R}^{2}.
\end{cases} { □ w = H ( t , x ) , ( w , ∂ t w ) ( 0 , x ) = ( w 0 , w 1 ) , ( t , x ) ∈ [ 0 , ∞ ) × R 2 , x ∈ R 2 . Then for μ , ν ∈ ( 0 , 1 / 2 ) \mu,\nu\in(0,1/2) μ , ν ∈ ( 0 , 1/2 ) , we have
⟨ t + ∣ x ∣ ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ μ ∣ w ∣ ≲ A 3 , 0 [ w 0 , w 1 ] + sup ( s , y ) ∈ [ 0 , t ] × R 2 ⟨ y ⟩ 1 / 2 W 1 + μ , 1 + ν ( s , y ) ∣ H ( s , y ) ∣ . (40) \langle t+|x|\rangle^{1/2}\langle t-|x|\rangle^{\mu}|w| \lesssim A_{3,0}[w_{0},w_{1}]+\sup_{(s,y)\in[0,t]\times\mathbb{R}^{2}}\langle y\rangle^{1/2}W_{1+\mu,1+\nu}(s,y)|H(s,y)|. \tag*{(40)} ⟨ t + ∣ x ∣ ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ μ ∣ w ∣ ≲ A 3 , 0 [ w 0 , w 1 ] + ( s , y ) ∈ [ 0 , t ] × R 2 sup ⟨ y ⟩ 1/2 W 1 + μ , 1 + ν ( s , y ) ∣ H ( s , y ) ∣. (40) ⟨ t + ∣ x ∣ ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ 1 / 2 ∣ w ∣ ≲ A 3 , 0 [ w 0 , w 1 ] + sup ( s , y ) ∈ [ 0 , t ] × R 2 ⟨ y ⟩ 1 / 2 W 3 / 2 + μ , 1 + ν ( s , y ) ∣ H ( s , y ) ∣ . (41) \langle t+|x|\rangle^{1/2}\langle t-|x|\rangle^{1/2}|w| \lesssim A_{3,0}[w_{0},w_{1}]+\sup_{(s,y)\in[0,t]\times\mathbb{R}^{2}}\langle y\rangle^{1/2}W_{3/2+\mu,1+\nu}(s,y)|H(s,y)|. \tag*{(41)} ⟨ t + ∣ x ∣ ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ 1/2 ∣ w ∣ ≲ A 3 , 0 [ w 0 , w 1 ] + ( s , y ) ∈ [ 0 , t ] × R 2 sup ⟨ y ⟩ 1/2 W 3/2 + μ , 1 + ν ( s , y ) ∣ H ( s , y ) ∣. (41) ⟨ t + ∣ x ∣ ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ 1 / 2 ∣ w ∣ ≲ A 3 , 0 [ w 0 , w 1 ] + ln ( 2 + ∣ t − ∣ x ∣ ∣ ) sup ( s , y ) ∈ [ 0 , t ] × R 2 ⟨ y ⟩ 1 / 2 W 3 / 2 , 1 + ν ( s , y ) ∣ H ( s , y ) ∣ (42) \langle t+|x|\rangle^{1/2}\langle t-|x|\rangle^{1/2}|w| \lesssim A_{3,0}[w_{0},w_{1}]+\ln(2+|t-|x||)\sup_{(s,y)\in[0,t]\times\mathbb{R}^{2}}\langle y\rangle^{1/2}W_{3/2,1+\nu}(s,y)|H(s,y)| \tag*{(42)} ⟨ t + ∣ x ∣ ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ 1/2 ∣ w ∣ ≲ A 3 , 0 [ w 0 , w 1 ] + ln ( 2 + ∣ t − ∣ x ∣∣ ) ( s , y ) ∈ [ 0 , t ] × R 2 sup ⟨ y ⟩ 1/2 W 3/2 , 1 + ν ( s , y ) ∣ H ( s , y ) ∣ (42) and
⟨ t − ∣ x ∣ ⟩ 1 + μ ∣ ∂ w ∣ ≲ A 4 , 1 [ w 0 , w 1 ] + ∑ ∣ a ∣ ≤ 1 sup ( s , y ) ∈ [ 0 , t ] × R 2 ⟨ y ⟩ 1 / 2 W 1 + μ + ν , 1 ( s , y ) ∣ Z a H ( s , y ) ∣ . (43) \langle t-|x|\rangle^{1+\mu}|\partial w| \lesssim A_{4,1}[w_{0},w_{1}]+\sum_{|a|\leq1}\sup_{(s,y)\in[0,t]\times\mathbb{R}^{2}}\langle y\rangle^{1/2}W_{1+\mu+\nu,1}(s,y)|Z^{a}H(s,y)|. \tag*{(43)} ⟨ t − ∣ x ∣ ⟩ 1 + μ ∣ ∂ w ∣ ≲ A 4 , 1 [ w 0 , w 1 ] + ∣ a ∣ ≤ 1 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 sup ⟨ y ⟩ 1/2 W 1 + μ + ν , 1 ( s , y ) ∣ Z a H ( s , y ) ∣. (43) where A κ , s [ f , g ] = ∑ Γ ~ ∈ { ∂ 1 , ∂ 2 , Ω } ( ∑ ∣ a ∣ ≤ s + 1 ∥ ⟨ z ⟩ κ Γ ~ a f ( z ) ∥ L ∞ + ∑ ∣ a ∣ ≤ s ∥ ⟨ z ⟩ κ Γ ~ a g ( z ) ∥ L ∞ ) A_{\kappa,s}[f,g]=\sum_{\widetilde{\Gamma}\in\{\partial_{1},\partial_{2},\Omega\}}\left(\sum_{|a|\leq s+1}\|\langle z\rangle^{\kappa}\widetilde{\Gamma}^{a}f(z)\|_{L^{\infty}}+\sum_{|a|\leq s}\|\langle z\rangle^{\kappa}\widetilde{\Gamma}^{a}g(z)\|_{L^{\infty}}\right) A κ , s [ f , g ] = ∑ Γ ∈ { ∂ 1 , ∂ 2 , Ω } ( ∑ ∣ a ∣ ≤ s + 1 ∥ ⟨ z ⟩ κ Γ a f ( z ) ∥ L ∞ + ∑ ∣ a ∣ ≤ s ∥ ⟨ z ⟩ κ Γ a g ( z ) ∥ L ∞ ) .
Proof. See (59) , (60) , (62) of [17 ] (or (92) , (93) , (94) and (A.13) of [32 ] ) for (40) , (41) , (43) , respectively. The estimate (42) can be obtained from (92) and (A.13) of [32 ] .
Lemma 2.8 (Lemma 2.8 of [19 ] ). Let w w w be the solution of the Cauchy problem
{ □ w = H ( t , x ) , ( t , x ) ∈ [ 0 , ∞ ) × R 2 , ( w , ∂ t w ) ( 0 , x ) = ( 0 , 0 ) , x ∈ R 2 . \begin{cases}
\Box w = H(t,x), & (t,x) \in[0,\infty) \times\mathbb{R}^{2},\\
(w,\partial_t w)(0,x) = (0,0), & x \in\mathbb{R}^{2}.
\end{cases} { □ w = H ( t , x ) , ( w , ∂ t w ) ( 0 , x ) = ( 0 , 0 ) , ( t , x ) ∈ [ 0 , ∞ ) × R 2 , x ∈ R 2 . where supp x H ( t , x ) ⊂ { x : ∣ x ∣ ≤ R } \operatorname{supp}_{x} H(t,x) \subset\{x: |x| \le R\} supp x H ( t , x ) ⊂ { x : ∣ x ∣ ≤ R } . Then we have that for μ , ν > 0 \mu,\nu> 0 μ , ν > 0 and μ + ν < 1 / 2 \mu+\nu< 1/2 μ + ν < 1/2 ,
⟨ t + ∣ x ∣ ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ μ ∣ w ∣ ≲ ∑ ∣ a ∣ ≤ 2 sup ( s , y ) ∈ [ 0 , t ] × R 2 ⟨ s ⟩ 1 / 2 + μ + ν ∣ ∂ a H ( s , y ) ∣ (44) \langle t+|x|\rangle^{1/2}\langle t-|x|\rangle^{\mu}|w| \lesssim\sum_{|a|\le2}\sup_{(s,y)\in[0,t]\times\mathbb{R}^{2}}\langle s\rangle^{1/2+\mu+\nu}|\partial^{a}H(s,y)| \tag*{(44)} ⟨ t + ∣ x ∣ ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ μ ∣ w ∣ ≲ ∣ a ∣ ≤ 2 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 sup ⟨ s ⟩ 1/2 + μ + ν ∣ ∂ a H ( s , y ) ∣ (44) and
⟨ t ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ ∣ ∂ w ∣ ≲ sup ( s , y ) ∈ [ 0 , t ] × R 2 ⟨ s ⟩ ∣ ( 1 − Δ ) 3 H ( s , y ) ∣ . (45) \langle t\rangle^{1/2}\langle t-|x|\rangle|\partial w| \lesssim\sup_{(s,y)\in[0,t]\times\mathbb{R}^{2}}\langle s\rangle|(1-\Delta)^{3}H(s,y)|. \tag*{(45)} ⟨ t ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ ∣ ∂ w ∣ ≲ ( s , y ) ∈ [ 0 , t ] × R 2 sup ⟨ s ⟩ ∣ ( 1 − Δ ) 3 H ( s , y ) ∣. (45)
Null conditions
Set
Q 0 ( f , g ) = ∂ t f ∂ t g − ∑ i = 1 2 ∂ i f ∂ i g , Q_{0}(f,g)=\partial_{t}f\partial_{t}g-\sum_{i=1}^{2}\partial_{i}f\partial_{i}g, Q 0 ( f , g ) = ∂ t f ∂ t g − i = 1 ∑ 2 ∂ i f ∂ i g ,
Q μ ν ( f , g ) = ∂ μ f ∂ ν g − ∂ ν f ∂ μ g , μ , ν = 0 , 1 , 2. (46) Q_{\mu\nu}(f,g)=\partial_{\mu}f\partial_{\nu}g-\partial_{\nu}f\partial_{\mu}g,\quad\mu,\nu=0,1,2. \tag*{(46)} Q μν ( f , g ) = ∂ μ f ∂ ν g − ∂ ν f ∂ μ g , μ , ν = 0 , 1 , 2. (46)
Lemma 2.9. Suppose that the constants N α β γ N^{\alpha\beta\gamma} N α β γ satisfy
∑ α , β , γ = 0 2 N α β γ ξ α ξ β ξ γ = 0 for ξ = ( ξ 0 , ξ 1 , ξ 2 ) ∈ { ± 1 } × S . \sum_{\alpha,\beta,\gamma=0}^{2}N^{\alpha\beta\gamma}\xi_{\alpha}\xi_{\beta}\xi_{\gamma}=0\quad\text{for }\xi=(\xi_{0},\xi_{1},\xi_{2})\in\{\pm1\}\times\mathbb{S}. α , β , γ = 0 ∑ 2 N α β γ ξ α ξ β ξ γ = 0 for ξ = ( ξ 0 , ξ 1 , ξ 2 ) ∈ { ± 1 } × S . Then for smooth functions f f f , g g g and h h h , it holds that
∣ ∑ α , β , γ = 0 2 N α β γ ∂ α β 2 f ∂ γ g ∣ ≲ ∣ ∂ ˉ ∂ f ∣ ∣ ∂ g ∣ + ∣ ∂ 2 f ∣ ∣ ∂ ˉ g ∣ \left|\sum_{\alpha,\beta,\gamma=0}^{2}N^{\alpha\beta\gamma}\partial_{\alpha\beta}^{2}f\partial_{\gamma}g\right|\lesssim|\bar{\partial}\partial f||\partial g|+|\partial^{2}f||\bar{\partial}g| α , β , γ = 0 ∑ 2 N α β γ ∂ α β 2 f ∂ γ g ≲ ∣ ∂ ˉ ∂ f ∣∣ ∂ g ∣ + ∣ ∂ 2 f ∣∣ ∂ ˉ g ∣ ∣ ∑ α , β , γ = 0 2 N α β γ ∂ α f ∂ β g ∂ γ h ∣ ≲ ∣ ∂ ˉ f ∣ ∣ ∂ g ∣ ∣ ∂ h ∣ + ∣ ∂ f ∣ ∣ ∂ ˉ g ∣ ∣ ∂ h ∣ + ∣ ∂ f ∣ ∣ ∂ g ∣ ∣ ∂ ˉ h ∣ . (47) \left|\sum_{\alpha,\beta,\gamma=0}^{2}N^{\alpha\beta\gamma}\partial_{\alpha}f\partial_{\beta}g\partial_{\gamma}h\right|\lesssim|\bar{\partial}f||\partial g||\partial h|+|\partial f||\bar{\partial}g||\partial h|+|\partial f||\partial g||\bar{\partial}h|. \tag*{(47)} α , β , γ = 0 ∑ 2 N α β γ ∂ α f ∂ β g ∂ γ h ≲ ∣ ∂ ˉ f ∣∣ ∂ g ∣∣ ∂ h ∣ + ∣ ∂ f ∣∣ ∂ ˉ g ∣∣ ∂ h ∣ + ∣ ∂ f ∣∣ ∂ g ∣∣ ∂ ˉ h ∣. (47) and
Q 12 ( f , g ) = 1 r ( ∂ r f ∂ θ g − ∂ θ f ∂ r g ) = ∑ i = 1 , 2 ∂ i ( x i f Ω g ∣ x ∣ 2 ) − ∂ θ ( f ∂ r g ∣ x ∣ ) , x ≠ 0 , Q_{12}(f,g)=\frac{1}{r}(\partial_{r}f\partial_{\theta}g-\partial_{\theta}f\partial_{r}g)=\sum_{i=1,2}\partial_{i}\left(\frac{x_{i}f\Omega g}{|x|^{2}}\right)-\partial_{\theta}\left(\frac{f\partial_{r}g}{|x|}\right),\quad x\ne0, Q 12 ( f , g ) = r 1 ( ∂ r f ∂ θ g − ∂ θ f ∂ r g ) = i = 1 , 2 ∑ ∂ i ( ∣ x ∣ 2 x i f Ω g ) − ∂ θ ( ∣ x ∣ f ∂ r g ) , x = 0 , ∣ Q 12 ( f , g ) ∣ ≲ ( 1 + ∣ x ∣ ) − 1 ( ∣ Z f ∣ ∣ ∂ g ∣ + ∣ ∂ f ∣ ∣ Z g ∣ ) . (48) |Q_{12}(f,g)|\lesssim(1+|x|)^{-1}(|Zf||\partial g|+|\partial f||Zg|). \tag*{(48)} ∣ Q 12 ( f , g ) ∣ ≲ ( 1 + ∣ x ∣ ) − 1 ( ∣ Z f ∣∣ ∂ g ∣ + ∣ ∂ f ∣∣ Z g ∣ ) . (48)
Proof. Since the proof of (47) is rather analogous to that in Section 9.1 of [4 ] and [16 Lemma 2.2] , we omit it here. In addition, (48) comes from (35) of [17 ] directly.
Lemma 2.10. Let u u u be the solution of (1) and Q ( ∂ u , ∂ 2 u ) Q(\partial u,\partial^{2}u) Q ( ∂ u , ∂ 2 u ) be defined by (2) . Suppose that the symmetric conditions (3) and the first null condition (4) hold. Then for any multi-index α \alpha α , one has
□ Z α u = Z α Q ( ∂ u , ∂ 2 u ) \square Z^{\alpha}u=Z^{\alpha}Q(\partial u,\partial^{2}u) □ Z α u = Z α Q ( ∂ u , ∂ 2 u ) = ∑ α , β , γ = 0 b + c ≤ α 2 Q a b c α β γ ∂ α 2 Z b u ∂ γ Z c u + ∑ α , β , γ , δ = 0 b + c + d ≤ α 2 Q a b c d α β γ δ ∂ α β 2 Z b u ∂ γ Z c u ∂ δ Z d u , (49) =\sum_{\substack{\alpha,\beta,\gamma=0\\ b+c\leq\alpha}}^{2}Q_{abc}^{\alpha\beta\gamma}\partial_{\alpha}^{2}Z^{b}u\partial_{\gamma}Z^{c}u+\sum_{\substack{\alpha,\beta,\gamma,\delta=0\\ b+c+d\leq\alpha}}^{2}Q_{abcd}^{\alpha\beta\gamma\delta}\partial_{\alpha\beta}^{2}Z^{b}u\partial_{\gamma}Z^{c}u\partial_{\delta}Z^{d}u, \tag*{(49)} = α , β , γ = 0 b + c ≤ α ∑ 2 Q ab c α β γ ∂ α 2 Z b u ∂ γ Z c u + α , β , γ , δ = 0 b + c + d ≤ α ∑ 2 Q ab c d α β γ δ ∂ α β 2 Z b u ∂ γ Z c u ∂ δ Z d u , (49) where Q a b c α β γ Q_{abc}^{\alpha\beta\gamma} Q ab c α β γ and Q a b c d α β γ δ Q_{abcd}^{\alpha\beta\gamma\delta} Q ab c d α β γ δ are constants, in particular Q a a 0 α β γ = Q α β γ Q_{aa0}^{\alpha\beta\gamma}=Q^{\alpha\beta\gamma} Q aa 0 α β γ = Q α β γ and Q a a 00 α β γ δ = Q α β γ δ Q_{aa00}^{\alpha\beta\gamma\delta}=Q^{\alpha\beta\gamma\delta} Q aa 00 α β γ δ = Q α β γ δ hold. Furthermore, for any ( ξ 0 , ξ 1 , ξ 2 ) ∈ { ± 1 } × S (\xi_{0},\xi_{1},\xi_{2})\in\{\pm1\}\times\mathbb{S} ( ξ 0 , ξ 1 , ξ 2 ) ∈ { ± 1 } × S , one has
∑ α , β , γ = 0 2 Q a b c α β γ ξ α ξ β ξ γ ≡ 0. (50) \sum_{\alpha,\beta,\gamma=0}^{2}Q_{abc}^{\alpha\beta\gamma}\xi_{\alpha}\xi_{\beta}\xi_{\gamma}\equiv0. \tag*{(50)} α , β , γ = 0 ∑ 2 Q ab c α β γ ξ α ξ β ξ γ ≡ 0. (50) Proof. See Lemma 6.6.5 of [14 ] .
Lemma 2.11 (Structures of Q 0 i Q_{0i} Q 0 i null forms, Lemma 2.3 of [17 ] ). For any functions f f f , g g g , and α = 0 , 1 , 2 \alpha=0,1,2 α = 0 , 1 , 2 , we have that for x ≠ 0 x\ne0 x = 0 , i = 1 , 2 i=1,2 i = 1 , 2 ,
Q 0 i ( ∂ α f , g ) = ∑ μ = 0 2 { C i α 12 , μ Q 12 ( ∂ μ f , g ) + C i α 0 , μ ( ω ) Q 0 ( ∂ μ f , g ) + C i α μ ( ω ) □ f ∂ μ g } + ∑ μ , ν = 0 2 C i α μ ν , 1 ( ω ) Ω ∂ μ f ∂ ν g + C i α μ ν , 2 ( ω ) Ω g ∂ 2 f ∣ x ∣ , (51) Q_{0i}(\partial_{\alpha}f,g)=\sum_{\mu=0}^{2}\left\{C_{i\alpha}^{12,\mu}Q_{12}(\partial_{\mu}f,g)+C_{i\alpha}^{0,\mu}(\omega)Q_{0}(\partial_{\mu}f,g)+C_{i\alpha}^{\mu}(\omega)\square f\partial_{\mu}g\right\}
+\sum_{\mu,\nu=0}^{2}\frac{C_{i\alpha}^{\mu\nu,1}(\omega)\Omega\partial_{\mu}f\partial_{\nu}g+C_{i\alpha}^{\mu\nu,2}(\omega)\Omega g\partial^{2}f}{|x|}, \tag*{(51)} Q 0 i ( ∂ α f , g ) = μ = 0 ∑ 2 { C i α 12 , μ Q 12 ( ∂ μ f , g ) + C i α 0 , μ ( ω ) Q 0 ( ∂ μ f , g ) + C i α μ ( ω ) □ f ∂ μ g } + μ , ν = 0 ∑ 2 ∣ x ∣ C i α μν , 1 ( ω ) Ω ∂ μ f ∂ ν g + C i α μν , 2 ( ω ) Ω g ∂ 2 f , (51) where C i α 12 , μ C_{i\alpha}^{12,\mu} C i α 12 , μ are constants, C i α 0 , μ ( ω ) C_{i\alpha}^{0,\mu}(\omega) C i α 0 , μ ( ω ) , C i α μ ( ω ) C_{i\alpha}^{\mu}(\omega) C i α μ ( ω ) , C i α μ ν , 1 ( ω ) C_{i\alpha}^{\mu\nu,1}(\omega) C i α μν , 1 ( ω ) and C i α μ ν , 2 ( ω ) C_{i\alpha}^{\mu\nu,2}(\omega) C i α μν , 2 ( ω ) are polynomials of ω \omega ω . Furthermore, it holds that
Q 0 i ( ∂ α f , g ) = ∑ μ = 0 2 { C i α 12 , μ Q 12 ( ∂ μ f , g ) + C i α 0 , μ ( ω ) Q 0 ( ∂ μ f , g ) + C i α μ ( ω ) □ f ∂ μ g } + ∑ μ = 0 2 ∂ μ ( C ( ω ) g Ω ∂ f ∣ x ∣ ) + ∂ θ ( C ( ω ) g ∂ 2 f ∣ x ∣ ) + C ( ω ) g Ω ∂ f ∣ x ∣ 2 , (52) Q_{0i}(\partial_{\alpha}f,g)=\sum_{\mu=0}^{2}\left\{C_{i\alpha}^{12,\mu}Q_{12}(\partial_{\mu}f,g)+C_{i\alpha}^{0,\mu}(\omega)Q_{0}(\partial_{\mu}f,g)+C_{i\alpha}^{\mu}(\omega)\square f\partial_{\mu}g\right\}
+\sum_{\mu=0}^{2}\partial_{\mu}\left(\frac{C(\omega)g\Omega\partial f}{|x|}\right)+\partial_{\theta}\left(\frac{C(\omega)g\partial^{2}f}{|x|}\right)+\frac{C(\omega)g\Omega\partial f}{|x|^{2}}, \tag*{(52)} Q 0 i ( ∂ α f , g ) = μ = 0 ∑ 2 { C i α 12 , μ Q 12 ( ∂ μ f , g ) + C i α 0 , μ ( ω ) Q 0 ( ∂ μ f , g ) + C i α μ ( ω ) □ f ∂ μ g } + μ = 0 ∑ 2 ∂ μ ( ∣ x ∣ C ( ω ) g Ω ∂ f ) + ∂ θ ( ∣ x ∣ C ( ω ) g ∂ 2 f ) + ∣ x ∣ 2 C ( ω ) g Ω ∂ f , (52) where C ( ω ) C(\omega) C ( ω ) stands for the generic polynomial of ω \omega ω .
Lemma 2.12. Let u u u be the solution of (1) and Q ( ∂ u , ∂ 2 u ) Q(\partial u,\partial^{2}u) Q ( ∂ u , ∂ 2 u ) be defined by (2) . Suppose that the symmetric conditions (3) and the first null condition (4) hold. Then for any multi-index α \alpha α , one has
□ Z α u = ∑ b + c ≤ α { ∑ μ , ν = 0 2 ∣ x ∣ − 1 ( C ( ω ) ∂ μ ν 2 Z b u Ω Z c u + C ( ω ) ∂ μ Z b u ∂ ν Z c u ) + ∑ μ = 0 2 [ C a b c 12 , μ Q 12 ( ∂ μ Z b u , Z c u ) + C a b c 0 , μ ( ω ) Q 0 ( ∂ μ Z b u , Z c u ) + C ( ω ) □ Z b u ∂ μ Z c u ] } + Z α [ ∑ α , β , γ , δ = 0 2 Q α β γ δ ∂ α β 2 u ∂ γ u ∂ δ u ] , (53) \square Z^{\alpha}u=\sum_{b+c\leq\alpha}\left\{\sum_{\mu,\nu=0}^{2}|x|^{-1}\left(C(\omega)\partial_{\mu\nu}^{2}Z^{b}u\Omega Z^{c}u+C(\omega)\partial_{\mu}Z^{b}u\partial_{\nu}Z^{c}u\right)
+\sum_{\mu=0}^{2}\left[C_{abc}^{12,\mu}Q_{12}(\partial_{\mu}Z^{b}u,Z^{c}u)+C_{abc}^{0,\mu}(\omega)Q_{0}(\partial_{\mu}Z^{b}u,Z^{c}u)+C(\omega)\square Z^{b}u\partial_{\mu}Z^{c}u\right]\right\}
+Z^{\alpha}\left[\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\partial_{\alpha\beta}^{2}u\partial_{\gamma}u\partial_{\delta}u\right], \tag*{(53)} □ Z α u = b + c ≤ α ∑ { μ , ν = 0 ∑ 2 ∣ x ∣ − 1 ( C ( ω ) ∂ μν 2 Z b u Ω Z c u + C ( ω ) ∂ μ Z b u ∂ ν Z c u ) + μ = 0 ∑ 2 [ C ab c 12 , μ Q 12 ( ∂ μ Z b u , Z c u ) + C ab c 0 , μ ( ω ) Q 0 ( ∂ μ Z b u , Z c u ) + C ( ω ) □ Z b u ∂ μ Z c u ] } + Z α α , β , γ , δ = 0 ∑ 2 Q α β γ δ ∂ α β 2 u ∂ γ u ∂ δ u , (53)
where C a b c 12 , μ C_{abc}^{12,\mu} C ab c 12 , μ are constants and C a b c 0 , μ ( ω ) C_{abc}^{0,\mu}(\omega) C ab c 0 , μ ( ω ) are polynomials of ω \omega ω . Moreover, one has
□ Z a u = ∑ b + c ≤ a { ∂ θ ( C ( ω ) Z c u ∂ 2 Z b u ∣ x ∣ ) + ∑ μ = 0 2 ∂ μ ( C ( ω ) Z c u Z ∂ Z b u ∣ x ∣ ) + C ( ω ) Z c u Z ∂ Z b u ∣ x ∣ 2 + ∑ μ = 0 2 [ C a b c 0 , μ ( ω ) Q 0 ( ∂ μ Z b u , Z c u ) + C ( ω ) ∂ μ Z c u □ Z b u ] } + Z a [ ∑ α , β , γ , δ = 0 2 Q a b c α β γ δ ∂ α β 2 u ∂ γ u ∂ δ u ] . (54) \Box Z^a u=\sum_{b+c\leq a}\left\{\partial_\theta\left(\frac{C(\omega)Z^cu\partial^2Z^bu}{\lvert x\rvert}\right)+\sum_{\mu=0}^{2}\partial_\mu\left(\frac{C(\omega)Z^cuZ\partial Z^bu}{\lvert x\rvert}\right)+\frac{C(\omega)Z^cuZ\partial Z^bu}{\lvert x\rvert^2}+\sum_{\mu=0}^{2}\left[C_{abc}^{0,\mu}(\omega)Q_0(\partial_\mu Z^bu,Z^cu)+C(\omega)\partial_\mu Z^cu\Box Z^bu\right]\right\}+Z^a\left[\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q_{abc}^{\alpha\beta\gamma\delta}\partial_{\alpha\beta}^2u\partial_\gamma u\partial_\delta u\right]. \tag*{(54)} □ Z a u = b + c ≤ a ∑ { ∂ θ ( ∣ x ∣ C ( ω ) Z c u ∂ 2 Z b u ) + μ = 0 ∑ 2 ∂ μ ( ∣ x ∣ C ( ω ) Z c u Z ∂ Z b u ) + ∣ x ∣ 2 C ( ω ) Z c u Z ∂ Z b u + μ = 0 ∑ 2 [ C ab c 0 , μ ( ω ) Q 0 ( ∂ μ Z b u , Z c u ) + C ( ω ) ∂ μ Z c u □ Z b u ] } + Z a α , β , γ , δ = 0 ∑ 2 Q ab c α β γ δ ∂ α β 2 u ∂ γ u ∂ δ u . (54)
Proof. Note that the null condition (50) ensures (see [27 ] or [38 Page 79] )
∑ α , β , γ = 0 2 Q a b c α β γ ∂ α β 2 Z b u ∂ γ Z c u = ∑ μ = 0 2 [ C a b c 0 , μ Q 0 ( ∂ μ Z b u , Z c u ) + ∑ α , β = 0 2 C a b c α β , μ Q α β ( ∂ μ Z b u , Z c u ) ] . (55) \sum_{\alpha,\beta,\gamma=0}^{2}Q_{abc}^{\alpha\beta\gamma}\partial_{\alpha\beta}^2Z^bu\partial_\gamma Z^cu=\sum_{\mu=0}^{2}\left[C_{abc}^{0,\mu}Q_0(\partial_\mu Z^bu,Z^cu)+\sum_{\alpha,\beta=0}^{2}C_{abc}^{\alpha\beta,\mu}Q_{\alpha\beta}(\partial_\mu Z^bu,Z^cu)\right]. \tag*{(55)} α , β , γ = 0 ∑ 2 Q ab c α β γ ∂ α β 2 Z b u ∂ γ Z c u = μ = 0 ∑ 2 C ab c 0 , μ Q 0 ( ∂ μ Z b u , Z c u ) + α , β = 0 ∑ 2 C ab c α β , μ Q α β ( ∂ μ Z b u , Z c u ) . (55)
where C a b c 0 , μ C_{abc}^{0,\mu} C ab c 0 , μ and C a b c α β , μ C_{abc}^{\alpha\beta,\mu} C ab c α β , μ are constants. The equations (53) and (54) can be obtained by (49) , (55) with (51) and by (48) , (52) , respectively.
Proof of Lemma 2.5
In order to prove Lemma 2.5, we first establish the following basic estimate.
Lemma 3.1. For μ ∈ ( 0 , 1 / 2 ) \mu\in(0,1/2) μ ∈ ( 0 , 1/2 ) , it holds that
⟨ t + ∣ x ∣ ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ μ ∣ w ( t , x ) ∣ ≲ ln ( 2 + t + ∣ x ∣ ) ∑ ∣ a ∣ ≤ 1 sup ( s , y ) ∈ [ 0 , t ] × R 2 ⟨ y ⟩ 1 / 2 W 1 , 1 ( s , y ) ∣ Z a H ( s , y ) ∣ . (56) \langle t+\lvert x\rvert\rangle^{1/2}\langle t-\lvert x\rvert\rangle^\mu\lvert w(t,x)\rvert\lesssim\ln(2+t+\lvert x\rvert)\sum_{\lvert a\rvert\leq1}\sup_{(s,y)\in[0,t]\times\mathbb{R}^2}\langle y\rangle^{1/2}W_{1,1}(s,y)\lvert Z^aH(s,y)\rvert. \tag*{(56)} ⟨ t + ∣ x ∣ ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ μ ∣ w ( t , x )∣ ≲ ln ( 2 + t + ∣ x ∣) ∣ a ∣ ≤ 1 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 sup ⟨ y ⟩ 1/2 W 1 , 1 ( s , y ) ∣ Z a H ( s , y )∣ . (56) provided that w ( t , x ) w(t,x) w ( t , x ) is the solution of
{ □ w = ∂ t H ( t , x ) , ( t , x ) ∈ [ 0 , ∞ ) × R 2 , ( w , ∂ t w ) ( 0 , x ) = ( 0 , 0 ) , x ∈ R 2 . \begin{cases}
\Box w=\partial_tH(t,x), & (t,x)\in[0,\infty)\times\mathbb{R}^2,\\
(w,\partial_tw)(0,x)=(0,0), & x\in\mathbb{R}^2.
\end{cases} { □ w = ∂ t H ( t , x ) , ( w , ∂ t w ) ( 0 , x ) = ( 0 , 0 ) , ( t , x ) ∈ [ 0 , ∞ ) × R 2 , x ∈ R 2 .
Proof. By Poisson’s formula, one has
w ( t , x ) = 1 2 π ∫ ∣ x − y ∣ < t ∫ 0 t − ∣ x − y ∣ ∂ s H ( s , y ) ( t − s ) 2 − ∣ x − y ∣ 2 d s d y . (57) w(t,x)=\frac{1}{2\pi}\int_{\lvert x-y\rvert<t}\int_{0}^{t-\lvert x-y\rvert}\frac{\partial_sH(s,y)}{\sqrt{(t-s)^2-\lvert x-y\rvert^2}}\,ds\,dy. \tag*{(57)} w ( t , x ) = 2 π 1 ∫ ∣ x − y ∣ < t ∫ 0 t − ∣ x − y ∣ ( t − s ) 2 − ∣ x − y ∣ 2 ∂ s H ( s , y ) d s d y . (57)
Set ρ = ∣ x − y ∣ \rho=\lvert x-y\rvert ρ = ∣ x − y ∣ , τ = t − s − ρ \tau=t-s-\rho τ = t − s − ρ and K ( ρ , τ ) = 1 τ ( 2 ρ + τ ) K(\rho,\tau)=\frac{1}{\sqrt{\tau(2\rho+\tau)}} K ( ρ , τ ) = τ ( 2 ρ + τ ) 1 . We decompose (57) into the following parts
2 π w ( t , x ) = ∫ ρ < t ∫ 0 t − ρ ( 1 − χ [ 1 / 2 , 1 ] ( τ ) ) K ( ρ , τ ) ∂ s H ( s , y ) d s d y + ∫ ρ < t ∫ 0 t − ρ χ [ 1 / 2 , 1 ] ( τ ) K ( ρ , τ ) ∂ s H ( s , y ) d s d y = ∫ ρ < t ∫ 0 t − ρ ( 1 − χ [ 1 / 2 , 1 ] ( τ ) ) K ( ρ , τ ) ∂ s H ( s , y ) d s d y − ∫ ρ < t χ [ 1 / 2 , 1 ] ( t − ρ ) K ( ρ , t − ρ ) H ( 0 , y ) d y + ∫ ρ < t ∫ 0 t − ρ ∂ τ ( χ [ 1 / 2 , 1 ] ( τ ) K ( ρ , τ ) ) H ( s , y ) d s d y . (58) \begin{aligned}
2\pi w(t,x)={}&\int_{\rho<t}\int_0^{t-\rho}(1-\chi_{[1/2,1]}(\tau))K(\rho,\tau)\partial_sH(s,y)\,ds\,dy\\
&+\int_{\rho<t}\int_0^{t-\rho}\chi_{[1/2,1]}(\tau)K(\rho,\tau)\partial_sH(s,y)\,ds\,dy\\
={}&\int_{\rho<t}\int_0^{t-\rho}(1-\chi_{[1/2,1]}(\tau))K(\rho,\tau)\partial_sH(s,y)\,ds\,dy\\
&-\int_{\rho<t}\chi_{[1/2,1]}(t-\rho)K(\rho,t-\rho)H(0,y)\,dy\\
&+\int_{\rho<t}\int_0^{t-\rho}\partial_\tau\bigl(\chi_{[1/2,1]}(\tau)K(\rho,\tau)\bigr)H(s,y)\,ds\,dy. \tag*{(58)}
\end{aligned} 2 π w ( t , x ) = = ∫ ρ < t ∫ 0 t − ρ ( 1 − χ [ 1/2 , 1 ] ( τ )) K ( ρ , τ ) ∂ s H ( s , y ) d s d y + ∫ ρ < t ∫ 0 t − ρ χ [ 1/2 , 1 ] ( τ ) K ( ρ , τ ) ∂ s H ( s , y ) d s d y ∫ ρ < t ∫ 0 t − ρ ( 1 − χ [ 1/2 , 1 ] ( τ )) K ( ρ , τ ) ∂ s H ( s , y ) d s d y − ∫ ρ < t χ [ 1/2 , 1 ] ( t − ρ ) K ( ρ , t − ρ ) H ( 0 , y ) d y + ∫ ρ < t ∫ 0 t − ρ ∂ τ ( χ [ 1/2 , 1 ] ( τ ) K ( ρ , τ ) ) H ( s , y ) d s d y . (58)
On the other hand, it is easy to know ∣ ∂ τ K ( ρ , τ ) ∣ = ρ + τ τ ( 2 ρ + τ ) K ( ρ , τ ) ≤ K ( ρ , τ ) τ |\partial_\tau K(\rho,\tau)|=\frac{\rho+\tau}{\tau(2\rho+\tau)}K(\rho,\tau)\le\frac{K(\rho,\tau)}{\tau} ∣ ∂ τ K ( ρ , τ ) ∣ = τ ( 2 ρ + τ ) ρ + τ K ( ρ , τ ) ≤ τ K ( ρ , τ ) . Then we have
∣ ( 1 − χ [ 1 / 2 , 1 ] ( τ ) ) K ( ρ , τ ) ∣ + ∣ ∂ τ ( χ [ 1 / 2 , 1 ] ( τ ) K ( ρ , τ ) ) ∣ ≲ K ( ρ , τ ) 1 + τ . (59) |(1-\chi_{[1/2,1]}(\tau))K(\rho,\tau)|+|\partial_\tau(\chi_{[1/2,1]}(\tau)K(\rho,\tau))|\lesssim\frac{K(\rho,\tau)}{1+\tau}. \tag*{(59)} ∣ ( 1 − χ [ 1/2 , 1 ] ( τ )) K ( ρ , τ ) ∣ + ∣ ∂ τ ( χ [ 1/2 , 1 ] ( τ ) K ( ρ , τ )) ∣ ≲ 1 + τ K ( ρ , τ ) . (59)
Substituting (59) into (58) yields
∣ w ( t , x ) ∣ ≲ M t ( X ( t , x ) + ∫ 0 t k ( τ ) Y ( t − τ , x ) d τ ) , (60) |w(t,x)|\lesssim M_t(X(t,x)+\int_0^t k(\tau)Y(t-\tau,x)d\tau), \tag*{(60)} ∣ w ( t , x ) ∣ ≲ M t ( X ( t , x ) + ∫ 0 t k ( τ ) Y ( t − τ , x ) d τ ) , (60)
where
M t = ∑ ∣ a ∣ ≤ 1 sup ( s , y ) ∈ [ 0 , t ] × R 2 W ( s , y ) ∣ Z a H ( s , y ) ∣ , X ( t , x ) = ∫ ∣ x − y ∣ < t ( 1 + ∣ y ∣ ) − 5 / 2 t 2 − ∣ x − y ∣ 2 d y , M_t=\sum_{|a|\le1}\sup_{(s,y)\in[0,t]\times\mathbb{R}^2}W(s,y)|Z^aH(s,y)|,\qquad X(t,x)=\int_{|x-y|<t}\frac{(1+|y|)^{-5/2}}{\sqrt{t^2-|x-y|^2}}dy, M t = ∣ a ∣ ≤ 1 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 sup W ( s , y ) ∣ Z a H ( s , y ) ∣ , X ( t , x ) = ∫ ∣ x − y ∣ < t t 2 − ∣ x − y ∣ 2 ( 1 + ∣ y ∣ ) − 5/2 d y ,
k ( τ ) = 1 τ ( 1 + τ ) , Y ( t , x ) = ∫ ∣ x − y ∣ < t ∣ x − y ∣ − 1 / 2 W ( t − ∣ x − y ∣ , y ) d y k(\tau)=\frac{1}{\sqrt{\tau}(1+\tau)},\qquad Y(t,x)=\int_{|x-y|<t}\frac{|x-y|^{-1/2}}{W(t-|x-y|,y)}dy k ( τ ) = τ ( 1 + τ ) 1 , Y ( t , x ) = ∫ ∣ x − y ∣ < t W ( t − ∣ x − y ∣ , y ) ∣ x − y ∣ − 1/2 d y
and W ( s , y ) = ⟨ y ⟩ 1 / 2 W 1 , 1 ( s , y ) W(s,y)=\langle y\rangle^{1/2}\mathcal{W}_{1,1}(s,y) W ( s , y ) = ⟨ y ⟩ 1/2 W 1 , 1 ( s , y ) .
Next we estimate X ( t , x ) X(t,x) X ( t , x ) and Y ( t , x ) Y(t,x) Y ( t , x ) . For notational convenience, set r = ∣ x ∣ r=|x| r = ∣ x ∣ , p = t + r p=t+r p = t + r , q = ∣ t − r ∣ q=|t-r| q = ∣ t − r ∣ , λ = ∣ y ∣ \lambda=|y| λ = ∣ y ∣ and ζ = t − r \zeta=t-r ζ = t − r .
We assert that
X ( t , x ) ≲ ⟨ t + r ⟩ − 1 / 2 ⟨ t − r ⟩ − 1 / 2 . (61) X(t,x)\lesssim\langle t+r\rangle^{-1/2}\langle t-r\rangle^{-1/2}. \tag*{(61)} X ( t , x ) ≲ ⟨ t + r ⟩ − 1/2 ⟨ t − r ⟩ − 1/2 . (61)
If p ≤ 8 p\le8 p ≤ 8 , then (61) can be obtained by
∫ ∣ x − y ∣ < t d y t 2 − ∣ x − y ∣ 2 = 2 π t . (62) \int_{|x-y|<t}\frac{dy}{\sqrt{t^2-|x-y|^2}}=2\pi t. \tag*{(62)} ∫ ∣ x − y ∣ < t t 2 − ∣ x − y ∣ 2 d y = 2 π t . (62)
In what follows, p ≥ 8 p\ge8 p ≥ 8 is assumed. Denote
X R ( t , x ) = ∫ ∣ x − y ∣ < t , R / 2 ≤ ∣ y ∣ ≤ R d y t 2 − ∣ x − y ∣ 2 , X_R(t,x)=\int_{|x-y|<t,\ R/2\le|y|\le R}\frac{dy}{\sqrt{t^2-|x-y|^2}}, X R ( t , x ) = ∫ ∣ x − y ∣ < t , R /2 ≤ ∣ y ∣ ≤ R t 2 − ∣ x − y ∣ 2 d y ,
X ~ R ( t , x ) = ∫ ∣ x − y ∣ < t , ∣ y ∣ ≤ R d y t 2 − ∣ x − y ∣ 2 . (63) \widetilde{X}_R(t,x)=\int_{|x-y|<t,\ |y|\le R}\frac{dy}{\sqrt{t^2-|x-y|^2}}. \tag*{(63)} X R ( t , x ) = ∫ ∣ x − y ∣ < t , ∣ y ∣ ≤ R t 2 − ∣ x − y ∣ 2 d y . (63)
If R = 2 n ≥ p / 4 R=2^n\ge p/4 R = 2 n ≥ p /4 , then it follows from (62) that
∑ R = 2 n ≥ p / 4 R − 5 / 2 X R ( t , x ) ≲ p − 5 / 2 t ≲ p − 3 / 2 . (64) \sum_{R=2^n\ge p/4}R^{-5/2}X_R(t,x)\lesssim p^{-5/2}t\lesssim p^{-3/2}. \tag*{(64)} R = 2 n ≥ p /4 ∑ R − 5/2 X R ( t , x ) ≲ p − 5/2 t ≲ p − 3/2 . (64)
Note that r = ∣ x ∣ ≤ ∣ x − y ∣ + ∣ y ∣ < t + R r=|x|\le|x-y|+|y|<t+R r = ∣ x ∣ ≤ ∣ x − y ∣ + ∣ y ∣ < t + R , otherwise, X ~ R ( t , x ) \widetilde{X}_R(t,x) X R ( t , x ) vanishes.
If R = 2 n ∈ [ 1 , p / 4 ] R=2^n\in[1,p/4] R = 2 n ∈ [ 1 , p /4 ] , then t > p − R 2 ≥ 3 p 8 ≥ 3 R 2 t>\frac{p-R}{2}\ge\frac{3p}{8}\ge\frac{3R}{2} t > 2 p − R ≥ 8 3 p ≥ 2 3 R holds. By the rotation of coordinates, one can assume x = ( r , 0 ) x=(r,0) x = ( r , 0 ) , y = ( z , h ) y=(z,h) y = ( z , h ) and z ± ( h ) = r ± t 2 − h 2 z_\pm(h)=r\pm\sqrt{t^2-h^2} z ± ( h ) = r ± t 2 − h 2 . Note that ∣ y ∣ ≤ R |y|\le R ∣ y ∣ ≤ R holds in the integral domain of X ~ R ( t , x ) \widetilde{X}_R(t,x) X R ( t , x ) . This implies that ∣ h ∣ ≤ R ≤ 2 t / 3 |h|\le R\le2t/3 ∣ h ∣ ≤ R ≤ 2 t /3 , ∣ z ∣ ≤ R ≤ p / 4 |z|\le R\le p/4 ∣ z ∣ ≤ R ≤ p /4 , z + ( h ) − z ≥ 5 − 2 2 p z_+(h)-z\ge\frac{\sqrt{5}-2}{2}p z + ( h ) − z ≥ 2 5 − 2 p and t 2 − ∣ x − y ∣ 2 = ( z − z − ( h ) ) ( z + ( h ) − z ) t^2-|x-y|^2=(z-z_-(h))(z_+(h)-z) t 2 − ∣ x − y ∣ 2 = ( z − z − ( h )) ( z + ( h ) − z ) . Therefore, one can achieve that for R = 2 n ∈ [ 1 , p / 4 ] R=2^n\in[1,p/4] R = 2 n ∈ [ 1 , p /4 ] ,
X R ( t , x ) ≲ X ~ R ( t , x ) ≲ p − 1 / 2 ∫ − R R ∫ − R ≤ z ≤ R , z ≥ z − ( h ) ( z − z − ( h ) ) − 1 / 2 d z d h ≲ R 3 / 2 p − 1 / 2 . (65) X_R(t,x)\lesssim\widetilde{X}_R(t,x)\lesssim p^{-1/2}\int_{-R}^{R}\int_{-R\le z\le R,\ z\ge z_-(h)}(z-z_-(h))^{-1/2}dzd h\lesssim R^{3/2}p^{-1/2}. \tag*{(65)} X R ( t , x ) ≲ X R ( t , x ) ≲ p − 1/2 ∫ − R R ∫ − R ≤ z ≤ R , z ≥ z − ( h ) ( z − z − ( h ) ) − 1/2 d z d h ≲ R 3/2 p − 1/2 . (65)
For q = ∣ t − r ∣ ≤ 2 q = |t-r| \le2 q = ∣ t − r ∣ ≤ 2 , (65) leads to
X ~ 1 ( t , x ) + ∑ n ≥ 0 ∑ R = 2 n ≤ p / 4 R − 5 / 2 X R ( t , x ) ≲ p − 1 / 2 . (66) \widetilde{X}_1(t,x)+\sum_{n\ge0}\sum_{R=2^n\le p/4}R^{-5/2}X_R(t,x)\lesssim p^{-1/2}. \tag*{(66)} X 1 ( t , x ) + n ≥ 0 ∑ R = 2 n ≤ p /4 ∑ R − 5/2 X R ( t , x ) ≲ p − 1/2 . (66)
For q ≥ 2 q\ge2 q ≥ 2 and t ≥ r = ∣ x ∣ t\ge r=|x| t ≥ r = ∣ x ∣ , when R = 2 n ∈ [ q / 4 , p / 4 ] R=2^n\in[q/4,p/4] R = 2 n ∈ [ q /4 , p /4 ] , (65) gives
∑ n ≥ 0 ∑ R = 2 n ∈ [ q / 4 , p / 4 ] R − 5 / 2 X R ( t , x ) ≲ ∑ n ≥ 0 ∑ R = 2 n ≥ q / 4 R − 1 p − 1 / 2 ≲ p − 1 / 2 ( 1 + q ) − 1 ; (67) \sum_{n\ge0}\sum_{R=2^n\in[q/4,p/4]}R^{-5/2}X_R(t,x)\lesssim\sum_{n\ge0}\sum_{R=2^n\ge q/4}R^{-1}p^{-1/2}\lesssim p^{-1/2}(1+q)^{-1}; \tag*{(67)} n ≥ 0 ∑ R = 2 n ∈ [ q /4 , p /4 ] ∑ R − 5/2 X R ( t , x ) ≲ n ≥ 0 ∑ R = 2 n ≥ q /4 ∑ R − 1 p − 1/2 ≲ p − 1/2 ( 1 + q ) − 1 ; (67)
when R = 2 n ≤ q / 4 = ∣ t − r ∣ / 4 R=2^n\le q/4=|t-r|/4 R = 2 n ≤ q /4 = ∣ t − r ∣/4 , it is easy to get t 2 − ∣ x − y ∣ 2 ≥ t 2 − ( r + R ) 2 ≳ p q t^2-|x-y|^2\ge t^2-(r+R)^2\gtrsim pq t 2 − ∣ x − y ∣ 2 ≥ t 2 − ( r + R ) 2 ≳ pq and
X ~ R ( t , x ) ≲ R 2 ( p q ) − 1 / 2 , \widetilde{X}_R(t,x)\lesssim R^2(pq)^{-1/2}, X R ( t , x ) ≲ R 2 ( pq ) − 1/2 ,
X ~ 1 ( t , x ) + ∑ n ≥ 0 ∑ R = 2 n ≤ q / 4 R − 5 / 2 X R ( t , x ) ≲ ( p q ) − 1 / 2 . (68) \widetilde{X}_1(t,x)+\sum_{n\ge0}\sum_{R=2^n\le q/4}R^{-5/2}X_R(t,x)\lesssim(pq)^{-1/2}. \tag*{(68)} X 1 ( t , x ) + n ≥ 0 ∑ R = 2 n ≤ q /4 ∑ R − 5/2 X R ( t , x ) ≲ ( pq ) − 1/2 . (68)
For q ≥ 2 q\ge2 q ≥ 2 and t ≤ r t\le r t ≤ r , the analogous estimate for (67) can be obtained. In this case, one has R ≥ q = r − t R\ge q=r-t R ≥ q = r − t in the integral domain of X ~ R ( t , x ) \widetilde{X}_R(t,x) X R ( t , x ) in (63) , otherwise, ∣ y ∣ ≤ R ≤ ∣ x ∣ − t ≤ ∣ x − y ∣ + ∣ y ∣ − t |y|\le R\le|x|-t\le|x-y|+|y|-t ∣ y ∣ ≤ R ≤ ∣ x ∣ − t ≤ ∣ x − y ∣ + ∣ y ∣ − t holds and X ~ R ( t , x ) \widetilde{X}_R(t,x) X R ( t , x ) vanishes.
Therefore, the estimate (61) can be obtained by combining (63) –(68) with the dyadic decomposition in space R 2 \mathbb{R}^2 R 2 .
• Estimate of Y ( t , x ) Y(t,x) Y ( t , x )
We assert that
Y ( t , x ) ≲ ln ( 2 + t + ∣ x ∣ ) ln ( 2 + ∣ t − ∣ x ∣ ∣ ) ⟨ t + ∣ x ∣ ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ 1 / 2 . (69) Y(t,x)\lesssim\frac{\ln(2+t+|x|)\ln(2+|t-|x||)}{\langle t+|x|\rangle^{1/2}\langle t-|x|\rangle^{1/2}}. \tag*{(69)} Y ( t , x ) ≲ ⟨ t + ∣ x ∣ ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ 1/2 ln ( 2 + t + ∣ x ∣ ) ln ( 2 + ∣ t − ∣ x ∣∣ ) . (69)
For fixed x ≠ 0 x\ne0 x = 0 , the Jacobian determinant of the coordinate transformation y = ( y 1 , y 2 ) → ( λ , ρ ) y=(y_1,y_2)\to(\lambda,\rho) y = ( y 1 , y 2 ) → ( λ , ρ ) is
∣ ∂ ( λ , ρ ) ∂ ( y 1 , y 2 ) ∣ = ( λ + ρ + r ) ( λ + ρ − r ) ( r + λ − ρ ) ( r − λ + ρ ) 2 λ ρ , (70) \left|\frac{\partial(\lambda,\rho)}{\partial(y_1,y_2)}\right|=\frac{\sqrt{(\lambda+\rho+r)(\lambda+\rho-r)(r+\lambda-\rho)(r-\lambda+\rho)}}{2\lambda\rho}, \tag*{(70)} ∂ ( y 1 , y 2 ) ∂ ( λ , ρ ) = 2 λ ρ ( λ + ρ + r ) ( λ + ρ − r ) ( r + λ − ρ ) ( r − λ + ρ ) , (70)
where λ = ∣ y ∣ \lambda=|y| λ = ∣ y ∣ , ρ = ∣ x − y ∣ \rho=|x-y| ρ = ∣ x − y ∣ and r = ∣ x ∣ r=|x| r = ∣ x ∣ .
Set α = λ + t − ρ \alpha=\lambda+t-\rho α = λ + t − ρ and β = λ + ρ − t \beta=\lambda+\rho-t β = λ + ρ − t .
From ρ < t \rho<t ρ < t , one has
α ∈ ( q , p ) , β ∈ ( − ζ , α ) , λ = 1 2 ( α + β ) , ρ = 1 2 ( 2 t − α + β ) , ∣ ∂ ( λ , ρ ) ∂ ( α , β ) ∣ = 1 / 2. (71) \alpha\in(q,p),\qquad\beta\in(-\zeta,\alpha),\qquad\lambda=\frac{1}{2}(\alpha+\beta),\qquad\rho=\frac{1}{2}(2t-\alpha+\beta),\qquad\left|\frac{\partial(\lambda,\rho)}{\partial(\alpha,\beta)}\right|=1/2. \tag*{(71)} α ∈ ( q , p ) , β ∈ ( − ζ , α ) , λ = 2 1 ( α + β ) , ρ = 2 1 ( 2 t − α + β ) , ∂ ( α , β ) ∂ ( λ , ρ ) = 1/2. (71)
Substituting (70) and (71) into Y ( t , x ) Y(t,x) Y ( t , x ) together with the facts of ρ ≤ ( p + β ) / 2 \rho\le(p+\beta)/2 ρ ≤ ( p + β ) /2 and λ ( 1 + λ ) − 1 / 2 ≲ λ 1 / 2 \lambda(1+\lambda)^{-1/2}\lesssim\lambda^{1/2} λ ( 1 + λ ) − 1/2 ≲ λ 1/2 yields
Y ( t , x ) ≲ ∫ q p Z ( α , ζ ) d α ( 1 + α ) ( α − ζ ) ( p − α ) with Z ( α , ζ ) = ∫ − ζ α λ 1 / 2 d β β + ζ min { 1 + λ , 1 + ∣ β ∣ } , (72) Y(t,x)\lesssim\int_q^p\frac{Z(\alpha,\zeta)\,d\alpha}{(1+\alpha)\sqrt{(\alpha-\zeta)(p-\alpha)}}\quad\text{with}\quad Z(\alpha,\zeta)=\int_{-\zeta}^{\alpha}\frac{\lambda^{1/2}\,d\beta}{\sqrt{\beta+\zeta}\min\{1+\lambda,1+|\beta|\}}, \tag*{(72)} Y ( t , x ) ≲ ∫ q p ( 1 + α ) ( α − ζ ) ( p − α ) Z ( α , ζ ) d α with Z ( α , ζ ) = ∫ − ζ α β + ζ min { 1 + λ , 1 + ∣ β ∣ } λ 1/2 d β , (72)
where ζ = t − r \zeta=t-r ζ = t − r .
We now deal with Z ( α , ζ ) Z(\alpha,\zeta) Z ( α , ζ ) . Note that
Z ( α , ζ ) ≲ Z 1 ( α , ζ ) + α 1 / 2 Z 2 ( α , ζ ) . (73) Z(\alpha,\zeta)\lesssim Z_1(\alpha,\zeta)+\alpha^{1/2}Z_2(\alpha,\zeta). \tag*{(73)} Z ( α , ζ ) ≲ Z 1 ( α , ζ ) + α 1/2 Z 2 ( α , ζ ) . (73)
where Z 1 ( α , ζ ) = ∫ − ζ α λ 1 / 2 d β β + ζ ( 1 + λ ) Z_1(\alpha,\zeta)=\int_{-\zeta}^{\alpha}\frac{\lambda^{1/2}\,d\beta}{\sqrt{\beta+\zeta(1+\lambda)}} Z 1 ( α , ζ ) = ∫ − ζ α β + ζ ( 1 + λ ) λ 1/2 d β and Z 2 ( α , ζ ) = ∫ − ζ α d β β + ζ ( 1 + ∣ β ∣ ) Z_2(\alpha,\zeta)=\int_{-\zeta}^{\alpha}\frac{d\beta}{\sqrt{\beta+\zeta(1+|\beta|)}} Z 2 ( α , ζ ) = ∫ − ζ α β + ζ ( 1 + ∣ β ∣ ) d β .
For Z 1 ( α , ζ ) Z_1(\alpha,\zeta) Z 1 ( α , ζ ) , set z = β + ζ z=\beta+\zeta z = β + ζ , one can arrive at
Z 1 ( α , ζ ) ≲ ∫ 0 α + ζ d z z ( 1 + α − ζ + z ) ≲ ln ( 2 + α ) , (74) Z_1(\alpha,\zeta)\lesssim\int_{0}^{\alpha+\zeta}\frac{dz}{\sqrt{z(1+\alpha-\zeta+z)}}\lesssim\ln(2+\alpha), \tag*{(74)} Z 1 ( α , ζ ) ≲ ∫ 0 α + ζ z ( 1 + α − ζ + z ) d z ≲ ln ( 2 + α ) , (74)
where the last inequality comes from the transformation of z = ( 1 + α − ζ ) sinh 2 ψ z=(1+\alpha-\zeta)\sinh^2\psi z = ( 1 + α − ζ ) sinh 2 ψ .
For Z 2 ( α , ζ ) Z_2(\alpha,\zeta) Z 2 ( α , ζ ) , we now show
Z 2 ( α , ζ ) ≲ ln ( 2 + q ) ( 1 + q ) − 1 / 2 with q = ∣ ζ ∣ . (75) Z_2(\alpha,\zeta)\lesssim\ln(2+q)(1+q)^{-1/2}\quad\text{with }q=|\zeta|. \tag*{(75)} Z 2 ( α , ζ ) ≲ ln ( 2 + q ) ( 1 + q ) − 1/2 with q = ∣ ζ ∣. (75)
Indeed, for ζ = − q ≤ 0 \zeta=-q\leq0 ζ = − q ≤ 0 , one has
Z 2 ( α , ζ ) ≲ ( 1 + q ) − 1 ∫ q 1 + 2 q d β β − q + ∫ 1 + 2 q ∞ d β ( 1 + β ) 3 / 2 ≲ ( 1 + q ) − 1 / 2 ; (76) Z_2(\alpha,\zeta)\lesssim(1+q)^{-1}\int_q^{1+2q}\frac{d\beta}{\sqrt{\beta-q}}+\int_{1+2q}^{\infty}\frac{d\beta}{(1+\beta)^{3/2}}\lesssim(1+q)^{-1/2}; \tag*{(76)} Z 2 ( α , ζ ) ≲ ( 1 + q ) − 1 ∫ q 1 + 2 q β − q d β + ∫ 1 + 2 q ∞ ( 1 + β ) 3/2 d β ≲ ( 1 + q ) − 1/2 ; (76)
for ζ = q > 0 \zeta=q>0 ζ = q > 0 and q ≥ 1 q\geq1 q ≥ 1 , analogously to (76) , we have
Z 2 ( α , ζ ) ≲ q − 1 ∫ − q − q / 2 d β β + q + q − 1 / 2 ∫ − q / 2 q d β 1 + ∣ β ∣ + ∫ q ∞ d β ( 1 + β ) 3 / 2 ≲ ( 1 + q ) − 1 / 2 ln ( 2 + q ) ; (77) \begin{aligned}
Z_2(\alpha,\zeta)&\lesssim q^{-1}\int_{-q}^{-q/2}\frac{d\beta}{\sqrt{\beta+q}}+q^{-1/2}\int_{-q/2}^{q}\frac{d\beta}{1+|\beta|}+\int_q^\infty\frac{d\beta}{(1+\beta)^{3/2}}\\
&\lesssim(1+q)^{-1/2}\ln(2+q); \tag*{(77)}
\end{aligned} Z 2 ( α , ζ ) ≲ q − 1 ∫ − q − q /2 β + q d β + q − 1/2 ∫ − q /2 q 1 + ∣ β ∣ d β + ∫ q ∞ ( 1 + β ) 3/2 d β ≲ ( 1 + q ) − 1/2 ln ( 2 + q ) ; (77)
when q ≤ 1 q\leq1 q ≤ 1 , it holds that
Z 2 ( α , ζ ) ≲ ∫ − q 2 − q d β β + q + ∫ 2 − q ∞ d β ( 1 + β ) 3 / 2 ≲ 1. (78) Z_2(\alpha,\zeta)\lesssim\int_{-q}^{2-q}\frac{d\beta}{\sqrt{\beta+q}}+\int_{2-q}^{\infty}\frac{d\beta}{(1+\beta)^{3/2}}\lesssim1. \tag*{(78)} Z 2 ( α , ζ ) ≲ ∫ − q 2 − q β + q d β + ∫ 2 − q ∞ ( 1 + β ) 3/2 d β ≲ 1. (78)
Collecting (76) -(78) yields (75) .
Substituting (74) , (75) into (73) , we have
Z ( α , ζ ) ≲ ln ( 2 + α ) + α 1 / 2 ln ( 2 + q ) ( 1 + q ) − 1 / 2 . (79) Z(\alpha,\zeta)\lesssim\ln(2+\alpha)+\alpha^{1/2}\ln(2+q)(1+q)^{-1/2}. \tag*{(79)} Z ( α , ζ ) ≲ ln ( 2 + α ) + α 1/2 ln ( 2 + q ) ( 1 + q ) − 1/2 . (79)
Based on (72) and (79) , we next estimate Y ( t , x ) Y(t,x) Y ( t , x ) . Note that
Y ( t , x ) ≲ ln ( 2 + q ) ( 1 + q ) − 1 / 2 ∫ q p d α ( 1 + α ) ( α − q ) ( p − α ) + ln ( 2 + p ) ∫ q p d α ( 1 + α ) ( α − q ) ( p − α ) . (80) Y(t,x)\lesssim\ln(2+q)(1+q)^{-1/2}\int_q^p\frac{d\alpha}{\sqrt{(1+\alpha)(\alpha-q)(p-\alpha)}}+\ln(2+p)\int_q^p\frac{d\alpha}{(1+\alpha)\sqrt{(\alpha-q)(p-\alpha)}}. \tag*{(80)} Y ( t , x ) ≲ ln ( 2 + q ) ( 1 + q ) − 1/2 ∫ q p ( 1 + α ) ( α − q ) ( p − α ) d α + ln ( 2 + p ) ∫ q p ( 1 + α ) ( α − q ) ( p − α ) d α . (80)
Let α = q + ( p − q ) sin 2 ψ \alpha=q+(p-q)\sin^2\psi α = q + ( p − q ) sin 2 ψ . Then the integral in the first line of (80) can be reduced to
∫ q p d α ( 1 + α ) ( α − q ) ( p − α ) = 2 ∫ 0 π / 2 d ψ 1 + q + ( p − q ) sin 2 ψ . (81) \int_q^p\frac{d\alpha}{\sqrt{(1+\alpha)(\alpha-q)(p-\alpha)}}=2\int_0^{\pi/2}\frac{d\psi}{\sqrt{1+q+(p-q)\sin^2\psi}}. \tag*{(81)} ∫ q p ( 1 + α ) ( α − q ) ( p − α ) d α = 2 ∫ 0 π /2 1 + q + ( p − q ) sin 2 ψ d ψ . (81)
If p − q ≤ 1 + q p-q\leq1+q p − q ≤ 1 + q , then one has
∫ 0 π / 2 d ψ 1 + q + ( p − q ) sin 2 ψ ≲ ( 1 + q ) − 1 / 2 ≲ ( 1 + p ) − 1 / 2 . (82) \int_0^{\pi/2}\frac{d\psi}{\sqrt{1+q+(p-q)\sin^2\psi}}\lesssim(1+q)^{-1/2}\lesssim(1+p)^{-1/2}. \tag*{(82)} ∫ 0 π /2 1 + q + ( p − q ) sin 2 ψ d ψ ≲ ( 1 + q ) − 1/2 ≲ ( 1 + p ) − 1/2 . (82)
If p − q ≥ 1 + q p-q \ge1+q p − q ≥ 1 + q , then it follows from sin ψ ≥ 2 ψ / π \sin\psi\ge2\psi/\pi sin ψ ≥ 2 ψ / π that
∫ 0 π / 2 d ψ 1 + q + ( p − q ) sin 2 ψ ≲ ∫ 0 π / 2 d ψ 1 + q + 4 ( p − q ) ψ 2 / π 2 \int_{0}^{\pi/2}\frac{d\psi}{\sqrt{1+q+(p-q)\sin^{2}\psi}}\lesssim\int_{0}^{\pi/2}\frac{d\psi}{\sqrt{1+q+4(p-q)\psi^{2}/\pi^{2}}} ∫ 0 π /2 1 + q + ( p − q ) sin 2 ψ d ψ ≲ ∫ 0 π /2 1 + q + 4 ( p − q ) ψ 2 / π 2 d ψ
≲ ( p − q ) − 1 / 2 arsinh ( p − q 1 + q ) 1 / 2 ≲ ln ( 2 + p ) ( 1 + p ) − 1 / 2 . (83) \lesssim(p-q)^{-1/2}\operatorname{arsinh}\left(\frac{p-q}{1+q}\right)^{1/2}\lesssim\ln(2+p)(1+p)^{-1/2}. \tag*{(83)} ≲ ( p − q ) − 1/2 arsinh ( 1 + q p − q ) 1/2 ≲ ln ( 2 + p ) ( 1 + p ) − 1/2 . (83)
Similarly, the integral in the second line of (80) can be treated as follows
∫ q p d α ( 1 + α ) ( α − q ) ( p − α ) = 2 ∫ 0 π / 2 d ψ 1 + q + ( p − q ) sin 2 ψ = 2 ∫ 0 ∞ d z 1 + q + ( 1 + p ) z 2 ≲ ( 1 + p ) − 1 / 2 ( 1 + q ) − 1 / 2 , (84) \begin{aligned}
\int_{q}^{p}\frac{d\alpha}{(1+\alpha)\sqrt{(\alpha-q)(p-\alpha)}}=2\int_{0}^{\pi/2}\frac{d\psi}{1+q+(p-q)\sin^{2}\psi} \\
&=2\int_{0}^{\infty}\frac{dz}{1+q+(1+p)z^{2}}\lesssim(1+p)^{-1/2}(1+q)^{-1/2}, \tag*{(84)}
\end{aligned} ∫ q p ( 1 + α ) ( α − q ) ( p − α ) d α = 2 ∫ 0 π /2 1 + q + ( p − q ) sin 2 ψ d ψ = 2 ∫ 0 ∞ 1 + q + ( 1 + p ) z 2 d z ≲ ( 1 + p ) − 1/2 ( 1 + q ) − 1/2 , (84)
where the transformation z = tan ψ z=\tan\psi z = tan ψ is used in the last line of (84) .
Substituting (3.26)-(84) into (80) yields (69) for x ≠ 0 x\ne0 x = 0 . In addition, in the case of x = 0 x=0 x = 0 , (69) can be analogously obtained.
We start to prove (56) . It follows from (60) , (61) and (69) that for μ ∈ ( 0 , 1 / 2 ) \mu\in(0,1/2) μ ∈ ( 0 , 1/2 ) ,
∣ w ( t , x ) ∣ ≲ M t ( ⟨ t + r ⟩ − 1 / 2 ⟨ t − r ⟩ − 1 / 2 + ln ( 2 + t + r ) ∫ 0 t k ( τ ) d τ ⟨ t − τ + r ⟩ 1 / 2 ⟨ t − τ − r ⟩ μ ) . (85) |w(t,x)|\lesssim M_t\left(\langle t+r\rangle^{-1/2}\langle t-r\rangle^{-1/2}+\ln(2+t+r)\int_{0}^{t}\frac{k(\tau)d\tau}{\langle t-\tau+r\rangle^{1/2}\langle t-\tau-r\rangle^{\mu}}\right). \tag*{(85)} ∣ w ( t , x ) ∣ ≲ M t ( ⟨ t + r ⟩ − 1/2 ⟨ t − r ⟩ − 1/2 + ln ( 2 + t + r ) ∫ 0 t ⟨ t − τ + r ⟩ 1/2 ⟨ t − τ − r ⟩ μ k ( τ ) d τ ) . (85)
For r = ∣ x ∣ > t r=|x|>t r = ∣ x ∣ > t , it is easy to know
t − τ + r ≳ t + r , ∣ t − τ − r ∣ = r + τ − t ≥ r − t > 0. t-\tau+r\gtrsim t+r,\qquad|t-\tau-r|=r+\tau-t\ge r-t>0. t − τ + r ≳ t + r , ∣ t − τ − r ∣ = r + τ − t ≥ r − t > 0.
This, together with the fact of ∫ 0 ∞ k ( τ ) d τ ≲ 1 \int_{0}^{\infty}k(\tau)\,d\tau\lesssim1 ∫ 0 ∞ k ( τ ) d τ ≲ 1 , implies
∫ 0 t k ( τ ) d τ ⟨ t − τ + r ⟩ 1 / 2 ⟨ t − τ − r ⟩ μ ≲ ⟨ t + r ⟩ − 1 / 2 ⟨ t − r ⟩ − μ . (86) \int_{0}^{t}\frac{k(\tau)d\tau}{\langle t-\tau+r\rangle^{1/2}\langle t-\tau-r\rangle^{\mu}}\lesssim\langle t+r\rangle^{-1/2}\langle t-r\rangle^{-\mu}. \tag*{(86)} ∫ 0 t ⟨ t − τ + r ⟩ 1/2 ⟨ t − τ − r ⟩ μ k ( τ ) d τ ≲ ⟨ t + r ⟩ − 1/2 ⟨ t − r ⟩ − μ . (86)
For r = ∣ x ∣ ≤ t r=|x|\le t r = ∣ x ∣ ≤ t , analogously to (77) , we have that for t − r ≥ 1 t-r\ge1 t − r ≥ 1 ,
∫ 0 ∞ k ( τ ) ( 1 + ∣ t − τ − r ∣ ) − μ d τ \int_{0}^{\infty}k(\tau)(1+|t-\tau-r|)^{-\mu}d\tau ∫ 0 ∞ k ( τ ) ( 1 + ∣ t − τ − r ∣ ) − μ d τ
≲ ⟨ t − r ⟩ − μ ∫ 0 ( t − r ) / 2 k ( τ ) d τ + ⟨ t − r ⟩ − 3 / 2 ∫ ( t − r ) / 2 2 ( t − r ) ( 1 + ∣ t − τ − r ∣ ) − μ d τ \lesssim\langle t-r\rangle^{-\mu}\int_{0}^{(t-r)/2}k(\tau)d\tau+\langle t-r\rangle^{-3/2}\int_{(t-r)/2}^{2(t-r)}(1+|t-\tau-r|)^{-\mu}d\tau ≲ ⟨ t − r ⟩ − μ ∫ 0 ( t − r ) /2 k ( τ ) d τ + ⟨ t − r ⟩ − 3/2 ∫ ( t − r ) /2 2 ( t − r ) ( 1 + ∣ t − τ − r ∣ ) − μ d τ
+ ∫ 2 ( t − r ) ∞ τ − 3 / 2 − μ d τ ≲ ⟨ t − r ⟩ − μ ; (87) \qquad+\int_{2(t-r)}^{\infty}\tau^{-3/2-\mu}d\tau
\lesssim\langle t-r\rangle^{-\mu}; \tag*{(87)} + ∫ 2 ( t − r ) ∞ τ − 3/2 − μ d τ ≲ ⟨ t − r ⟩ − μ ; (87)
when t − r ≤ 1 t-r\le1 t − r ≤ 1 , (87) follows from ∫ 0 ∞ k ( τ ) d τ ≲ 1 \int_{0}^{\infty}k(\tau)\,d\tau\lesssim1 ∫ 0 ∞ k ( τ ) d τ ≲ 1 . Therefore, we can arrive at
∫ 0 t k ( τ ) d τ ⟨ t − τ + r ⟩ 1 / 2 ⟨ t − τ − r ⟩ μ \int_{0}^{t}\frac{k(\tau)d\tau}{\langle t-\tau+r\rangle^{1/2}\langle t-\tau-r\rangle^{\mu}} ∫ 0 t ⟨ t − τ + r ⟩ 1/2 ⟨ t − τ − r ⟩ μ k ( τ ) d τ
≲ ⟨ t + r ⟩ − 1 / 2 ∫ 0 ( 1 + t + r ) / 2 k ( τ ) ( 1 + ∣ t − τ − r ∣ ) − μ d τ \lesssim\langle t+r\rangle^{-1/2}\int_{0}^{(1+t+r)/2}k(\tau)(1+|t-\tau-r|)^{-\mu}d\tau ≲ ⟨ t + r ⟩ − 1/2 ∫ 0 ( 1 + t + r ) /2 k ( τ ) ( 1 + ∣ t − τ − r ∣ ) − μ d τ
+ ⟨ t + r ⟩ − 3 / 2 ∫ ( 1 + t + r ) / 2 t ( 1 + t − τ + r ) − 1 / 2 d τ ≲ ⟨ t + r ⟩ − 1 / 2 ⟨ t − r ⟩ − μ , (88) \qquad+\langle t+r\rangle^{-3/2}\int_{(1+t+r)/2}^{t}(1+t-\tau+r)^{-1/2}d\tau
\lesssim\langle t+r\rangle^{-1/2}\langle t-r\rangle^{-\mu}, \tag*{(88)} + ⟨ t + r ⟩ − 3/2 ∫ ( 1 + t + r ) /2 t ( 1 + t − τ + r ) − 1/2 d τ ≲ ⟨ t + r ⟩ − 1/2 ⟨ t − r ⟩ − μ , (88)
where (87) is used. Therefore, (56) is proved by (85) , (86) and (88) .
Based on Lemma 3.1 , we now prove Lemma 2.5 .
Proof of Lemma 2.5. At first, we prove (35) for H 0 ( t , x ) ≡ 0 H^{0}(t,x)\equiv0 H 0 ( t , x ) ≡ 0 . Choosing μ = η > 0 \mu=\eta>0 μ = η > 0 , ν = 1 − η \nu=1-\eta ν = 1 − η and κ = 1 \kappa=1 κ = 1 in (B.6) of [31 ] , one then has
⟨ x ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ 1 − η ∣ v ∣ ≲ ln ( 2 + t + ∣ x ∣ ) ∑ i = 1 2 ∑ ∣ a ∣ ≤ 1 sup ( s , y ) ∈ [ 0 , t ] × R 2 ⟨ y ⟩ 3 / 2 ⟨ s + ∣ y ∣ ⟩ ∣ Z a H i ( s , y ) ∣ . (89) \langle x\rangle^{1/2}\langle t-|x|\rangle^{1-\eta}|v|\lesssim\ln(2+t+|x|)\sum_{i=1}^{2}\sum_{|a|\leq1}\sup_{(s,y)\in[0,t]\times\mathbb{R}^{2}}\langle y\rangle^{3/2}\langle s+|y|\rangle|Z^{a}H^{i}(s,y)|. \tag*{(89)} ⟨ x ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ 1 − η ∣ v ∣ ≲ ln ( 2 + t + ∣ x ∣ ) i = 1 ∑ 2 ∣ a ∣ ≤ 1 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 sup ⟨ y ⟩ 3/2 ⟨ s + ∣ y ∣ ⟩ ∣ Z a H i ( s , y ) ∣. (89)
Recall (B.7) of [31 ]
⟨ x ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ 1 − η ∣ v ∣ ≲ ln ( 2 + t + ∣ x ∣ ) ∑ i = 1 2 ∑ ∣ a ∣ ≤ 1 sup ( s , y ) ∈ [ 0 , t ] × R 2 ⟨ y ⟩ 1 / 2 ⟨ s + ∣ y ∣ ⟩ ⟨ s − ∣ y ∣ ⟩ ∣ Z a H i ( s , y ) ∣ . (90) \langle x\rangle^{1/2}\langle t-|x|\rangle^{1-\eta}|v|\lesssim\ln(2+t+|x|)\sum_{i=1}^{2}\sum_{|a|\leq1}\sup_{(s,y)\in[0,t]\times\mathbb{R}^{2}}\langle y\rangle^{1/2}\langle s+|y|\rangle\langle s-|y|\rangle|Z^{a}H^{i}(s,y)|. \tag*{(90)} ⟨ x ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ 1 − η ∣ v ∣ ≲ ln ( 2 + t + ∣ x ∣ ) i = 1 ∑ 2 ∣ a ∣ ≤ 1 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 sup ⟨ y ⟩ 1/2 ⟨ s + ∣ y ∣ ⟩ ⟨ s − ∣ y ∣ ⟩ ∣ Z a H i ( s , y ) ∣. (90)
Combining (89) with (90) yields (35) . It is pointed out that the estimates (B.6) and (B.7) of [31 ] are only proved for H 0 ( t , x ) ≡ 0 H^{0}(t,x)\equiv0 H 0 ( t , x ) ≡ 0 .
When H 0 ( t , x ) ≢ 0 H^{0}(t,x)\not\equiv0 H 0 ( t , x ) ≡ 0 , the estimate (36) can be achieved by (35) and (56) directly.
Some key estimates
We now recall the Littlewood-Paley decomposition. Choosing a smooth cutoff function ψ : R → [ 0 , 1 ] \psi:\mathbb{R}\to[0,1] ψ : R → [ 0 , 1 ] , which equals 1 on [ − 5 / 4 , 5 / 4 ] [-5/4,5/4] [ − 5/4 , 5/4 ] and vanishes outside [ − 8 / 5 , 8 / 5 ] [-8/5,8/5] [ − 8/5 , 8/5 ] , one sets
ψ k ( ξ ) = ψ ( ∣ ξ ∣ / 2 k ) − ψ ( ∣ ξ ∣ / 2 k − 1 ) , k ∈ Z , ξ ∈ R 2 . \psi_{k}(\xi)=\psi(|\xi|/2^{k})-\psi(|\xi|/2^{k-1}),\qquad k\in\mathbb{Z},\quad\xi\in\mathbb{R}^{2}. ψ k ( ξ ) = ψ ( ∣ ξ ∣/ 2 k ) − ψ ( ∣ ξ ∣/ 2 k − 1 ) , k ∈ Z , ξ ∈ R 2 .
Let P k P_{k} P k be the Littlewood-Paley projection onto frequency 2 k 2^{k} 2 k
F ( P k f ) ( ξ ) = ψ k ( ξ ) F f ( ξ ) , k ∈ Z . \mathcal{F}(P_{k}f)(\xi)=\psi_{k}(\xi)\mathcal{F}f(\xi),\qquad k\in\mathbb{Z}. F ( P k f ) ( ξ ) = ψ k ( ξ ) F f ( ξ ) , k ∈ Z .
Estimates of the homogeneous initial boundary value problem
Lemma 4.1. Let w h o m w_{\mathrm{hom}} w hom be the solution to the homogeneous initial boundary value problem (7) , then for any multi-index a a a , one has that
∣ ∂ Z a w h o m ( t , x ) ∣ ≲ ε ⟨ x ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ − 1 , |\partial Z^{a}w_{\mathrm{hom}}(t,x)|\lesssim\varepsilon\langle x\rangle^{-1/2}\langle t-|x|\rangle^{-1}, ∣ ∂ Z a w hom ( t , x ) ∣ ≲ ε ⟨ x ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ − 1 , ∣ Z a w h o m ( t , x ) ∣ ≲ ε ⟨ t + ∣ x ∣ ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ δ 0 − 1 / 2 , |Z^{a}w_{\mathrm{hom}}(t,x)|\lesssim\varepsilon\langle t+|x|\rangle^{-1/2}\langle t-|x|\rangle^{\delta_{0}-1/2}, ∣ Z a w hom ( t , x ) ∣ ≲ ε ⟨ t + ∣ x ∣ ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ δ 0 − 1/2 , ∣ ∂ Z ‾ a w h o m ( t , x ) ∣ ≲ ε ⟨ x ⟩ − 1 / 2 ⟨ t + ∣ x ∣ ⟩ − 1 , (91) |\partial\overline{Z}^{a}w_{\mathrm{hom}}(t,x)|\lesssim\varepsilon\langle x\rangle^{-1/2}\langle t+|x|\rangle^{-1}, \tag*{(91)} ∣ ∂ Z a w hom ( t , x ) ∣ ≲ ε ⟨ x ⟩ − 1/2 ⟨ t + ∣ x ∣ ⟩ − 1 , (91) where δ 0 = 10 − 3 \delta_{0}=10^{-3} δ 0 = 1 0 − 3 . Moreover, for ∣ x ∣ ≤ t / 4 |x|\leq t/4 ∣ x ∣ ≤ t /4 , it holds that
∣ ∂ Z a w h o m ( t , x ) ∣ ≲ ε ⟨ t ⟩ − 1 ln ( 2 + t ) − 2 . (92) |\partial Z^{a}w_{\mathrm{hom}}(t,x)|\lesssim\varepsilon\langle t\rangle^{-1}\ln(2+t)^{-2}. \tag*{(92)} ∣ ∂ Z a w hom ( t , x ) ∣ ≲ ε ⟨ t ⟩ − 1 ln ( 2 + t ) − 2 . (92)
Proof. It follows from the local energy decay estimate (32) that for any multi-index b b b ,
∥ ∂ b w h o m ( t , x ) ∥ L 2 ( K R ) ≲ ε ⟨ t ⟩ − 1 ln ( 2 + t ) − 2 . (93) \|\partial^{b}w_{\mathrm{hom}}(t,x)\|_{L^{2}(K_{R})}\lesssim\varepsilon\langle t\rangle^{-1}\ln(2+t)^{-2}. \tag*{(93)} ∥ ∂ b w hom ( t , x ) ∥ L 2 ( K R ) ≲ ε ⟨ t ⟩ − 1 ln ( 2 + t ) − 2 . (93)
where R > 0 R>0 R > 0 is a large constant. Thus, the estimates (91) and (92) in the region ∣ x ∣ ≤ 2 |x|\leq2 ∣ x ∣ ≤ 2 can be achieved.
In what follows, we focus on the case of ∣ x ∣ ≥ 2 |x| \ge2 ∣ x ∣ ≥ 2 and t ≥ 1 t \ge1 t ≥ 1 . Note that w c ( t , x ) = χ [ 1 , 2 ] ( x ) w h o m ( t , x ) w^{c}(t,x)=\chi_{[1,2]}(x)w_{\mathrm{hom}}(t,x) w c ( t , x ) = χ [ 1 , 2 ] ( x ) w hom ( t , x ) is the solution of the Cauchy problem
□ w c = − [ Δ , χ [ 1 , 2 ] ( x ) ] w h o m , ( w c , ∂ t w c ) ( 0 , x ) = ( ε χ [ 1 , 2 ] ( x ) u 0 ( x ) , ε χ [ 1 , 2 ] ( x ) u 1 ( x ) ) . (94) \square w^{c}=-[\Delta,\chi_{[1,2]}(x)]w_{\mathrm{hom}},\qquad(w^{c},\partial_{t}w^{c})(0,x)=(\varepsilon\chi_{[1,2]}(x)u_{0}(x),\varepsilon\chi_{[1,2]}(x)u_{1}(x)). \tag*{(94)} □ w c = − [ Δ , χ [ 1 , 2 ] ( x )] w hom , ( w c , ∂ t w c ) ( 0 , x ) = ( ε χ [ 1 , 2 ] ( x ) u 0 ( x ) , ε χ [ 1 , 2 ] ( x ) u 1 ( x )) . (94)
The first two inequalities in (91) follow from Lemmas 2.7-2.8 with μ = 1 / 2 − δ 0 \mu=1/2-\delta_{0} μ = 1/2 − δ 0 , ν = δ 0 / 2 \nu=\delta_{0}/2 ν = δ 0 /2 and (93) . The third inequality of (91) in the region ∣ x ∣ ≤ t / 2 |x|\le t/2 ∣ x ∣ ≤ t /2 can be obtained by the first inequality. In the region ∣ x ∣ ≥ t / 2 |x|\ge t/2 ∣ x ∣ ≥ t /2 , it is easy to get that (one can also see Section 4.6 of [30 ] or Section 7 of [17 ] )
∂ + ( r 1 / 2 w ) ( t , r x ∣ x ∣ ) − ∂ + ( r 1 / 2 w ) ( 0 , ( t + r ) x ∣ x ∣ ) = ∫ 0 t { ( r + t − s ) 1 / 2 □ w + ( r + t − s ) − 3 / 2 ( w / 4 + Ω 2 w ) } ( s , ( t + r − s ) x ∣ x ∣ ) d s . (95) \partial_{+}(r^{1/2}w)(t,r\tfrac{x}{|x|})-\partial_{+}(r^{1/2}w)(0,(t+r)\tfrac{x}{|x|})
=\int_{0}^{t}\{(r+t-s)^{1/2}\square w+(r+t-s)^{-3/2}(w/4+\Omega^{2}w)\}(s,(t+r-s)\tfrac{x}{|x|})\,ds. \tag*{(95)} ∂ + ( r 1/2 w ) ( t , r ∣ x ∣ x ) − ∂ + ( r 1/2 w ) ( 0 , ( t + r ) ∣ x ∣ x ) = ∫ 0 t {( r + t − s ) 1/2 □ w + ( r + t − s ) − 3/2 ( w /4 + Ω 2 w )} ( s , ( t + r − s ) ∣ x ∣ x ) d s . (95)
where ∂ ‾ 1 = x 1 r ∂ + − x 2 r Ω \overline{\partial}_{1}=\frac{x_{1}}{r}\partial_{+}-\frac{x_{2}}{r}\Omega ∂ 1 = r x 1 ∂ + − r x 2 Ω , ∂ ‾ 2 = x 2 r ∂ + + x 1 r Ω \overline{\partial}_{2}=\frac{x_{2}}{r}\partial_{+}+\frac{x_{1}}{r}\Omega ∂ 2 = r x 2 ∂ + + r x 1 Ω and ∂ ± = ∂ t ± ∂ r \partial_{\pm}=\partial_{t}\pm\partial_{r} ∂ ± = ∂ t ± ∂ r . Set w = Z a w h o m w=Z^{a}w_{\mathrm{hom}} w = Z a w hom in (95) . Then one obtains
∣ ∂ + ( r 1 / 2 Z a w h o m ) ( t , x ) ∣ ≲ ε ⟨ x ⟩ − 1 + ε ⟨ r + t ⟩ − 2 ∫ 0 t ( 1 + ∣ r + t − 2 s ∣ ) δ 0 − 1 / 2 d s |\partial_{+}(r^{1/2}Z^{a}w_{\mathrm{hom}})(t,x)|\lesssim\varepsilon\langle x\rangle^{-1}+\varepsilon\langle r+t\rangle^{-2}\int_{0}^{t}(1+|r+t-2s|)^{\delta_{0}-1/2}\,ds ∣ ∂ + ( r 1/2 Z a w hom ) ( t , x ) ∣ ≲ ε ⟨ x ⟩ − 1 + ε ⟨ r + t ⟩ − 2 ∫ 0 t ( 1 + ∣ r + t − 2 s ∣ ) δ 0 − 1/2 d s
≲ ε ⟨ t + ∣ x ∣ ⟩ − 1 , \lesssim\varepsilon\langle t+|x|\rangle^{-1}, ≲ ε ⟨ t + ∣ x ∣ ⟩ − 1 ,
where the second inequality in (91) is used. Thus, the third inequality of (91) is achieved.
Finally, we focus on the proof of (92) . The solution to problem (94) can be decomposed into w c = w 1 c + w 2 c w^{c}=w_{1}^{c}+w_{2}^{c} w c = w 1 c + w 2 c , where w 1 c w_{1}^{c} w 1 c and w 2 c w_{2}^{c} w 2 c are the solutions of the Cauchy problems
□ w 1 c = H = − [ Δ , χ [ 1 , 2 ] ( x ) ] w h o m , ( w 1 c , ∂ t w 1 c ) ( 0 , x ) = ( 0 , 0 ) , \square w_{1}^{c}=H=-[\Delta,\chi_{[1,2]}(x)]w_{\mathrm{hom}},\qquad(w_{1}^{c},\partial_{t}w_{1}^{c})(0,x)=(0,0), □ w 1 c = H = − [ Δ , χ [ 1 , 2 ] ( x )] w hom , ( w 1 c , ∂ t w 1 c ) ( 0 , x ) = ( 0 , 0 ) ,
□ w 2 c = 0 , ( w 2 c , ∂ t w 2 c ) ( 0 , x ) = ( ε χ [ 1 , 2 ] ( x ) u 0 ( x ) , ε χ [ 1 , 2 ] ( x ) u 1 ( x ) ) , \square w_{2}^{c}=0,\qquad(w_{2}^{c},\partial_{t}w_{2}^{c})(0,x)=(\varepsilon\chi_{[1,2]}(x)u_{0}(x),\varepsilon\chi_{[1,2]}(x)u_{1}(x)), □ w 2 c = 0 , ( w 2 c , ∂ t w 2 c ) ( 0 , x ) = ( ε χ [ 1 , 2 ] ( x ) u 0 ( x ) , ε χ [ 1 , 2 ] ( x ) u 1 ( x )) ,
respectively. The estimate of ∂ α w 2 c \partial^{\alpha}w_{2}^{c} ∂ α w 2 c (α = 0 , 1 , 2 \alpha=0,1,2 α = 0 , 1 , 2 ) can be obtained by (43) . Note that ∂ t w 1 c \partial_{t}w_{1}^{c} ∂ t w 1 c and ∂ j w 1 c \partial_{j}w_{1}^{c} ∂ j w 1 c (j = 1 , 2 j=1,2 j = 1 , 2 ) are expressed as
∂ t w 1 c ( t , x ) = ∫ 0 t ( cos ( t − s ) ∣ ∇ ∣ ) H ( s ) d s = ∫ 0 t e i ( t − s ) ∣ ∇ ∣ + e − i ( t − s ) ∣ ∇ ∣ 2 H ( s ) d s = 1 2 ∫ 0 t ∫ R 2 ∑ k ∈ Z ( K k 0 + + K k 0 − ) ( t − s , x − y ) ( 1 − Δ ) 3 H ( s , y ) d y d s \begin{aligned}
\partial_{t}w_{1}^{c}(t,x)=\int_{0}^{t}(\cos(t-s)|\nabla|)H(s)\,ds=\int_{0}^{t}\frac{e^{i(t-s)|\nabla|}+e^{-i(t-s)|\nabla|}}{2}H(s)\,ds \\
&=\frac{1}{2}\int_{0}^{t}\int_{\mathbb{R}^{2}}\sum_{k\in\mathbb{Z}}(K_{k}^{0+}+K_{k}^{0-})(t-s,x-y)(1-\Delta)^{3}H(s,y)\,dy\,ds
\end{aligned} ∂ t w 1 c ( t , x ) = ∫ 0 t ( cos ( t − s ) ∣∇∣ ) H ( s ) d s = ∫ 0 t 2 e i ( t − s ) ∣∇∣ + e − i ( t − s ) ∣∇∣ H ( s ) d s = 2 1 ∫ 0 t ∫ R 2 k ∈ Z ∑ ( K k 0 + + K k 0 − ) ( t − s , x − y ) ( 1 − Δ ) 3 H ( s , y ) d y d s
with K k 0 ± ( t , x ) = 1 ( 2 π ) 2 ∫ R 2 e i ( ± t ∣ ξ ∣ + x ⋅ ξ ) ψ k ( ξ ) ( 1 + ∣ ξ ∣ 2 ) 3 d ξ K_{k}^{0\pm}(t,x)=\frac{1}{(2\pi)^{2}}\int_{\mathbb{R}^{2}}e^{i(\pm t|\xi|+x\cdot\xi)}\frac{\psi_{k}(\xi)}{(1+|\xi|^{2})^{3}}\,d\xi K k 0 ± ( t , x ) = ( 2 π ) 2 1 ∫ R 2 e i ( ± t ∣ ξ ∣ + x ⋅ ξ ) ( 1 + ∣ ξ ∣ 2 ) 3 ψ k ( ξ ) d ξ
and
∂ j w 1 c ( t , x ) = ∫ 0 t ( sin ( t − s ) ∣ ∇ ∣ ) R j H ( s ) d s = 1 2 i ∫ 0 t ∫ R 2 ∑ k ∈ Z ( K k j + − K k j − ) ( t − s , x − y ) P k R j ( 1 − Δ ) 3 H ( s , y ) d y d s . (96) \begin{aligned}
\partial_{j}w_{1}^{c}(t,x)=\int_{0}^{t}(\sin(t-s)|\nabla|)R_{j}H(s)\,ds \\
&=\frac{1}{2i}\int_{0}^{t}\int_{\mathbb{R}^{2}}\sum_{k\in\mathbb{Z}}(K_{k}^{j+}-K_{k}^{j-})(t-s,x-y)P_{k}R_{j}(1-\Delta)^{3}H(s,y)\,dy\,ds. \tag*{(96)}
\end{aligned} ∂ j w 1 c ( t , x ) = ∫ 0 t ( sin ( t − s ) ∣∇∣ ) R j H ( s ) d s = 2 i 1 ∫ 0 t ∫ R 2 k ∈ Z ∑ ( K k j + − K k j − ) ( t − s , x − y ) P k R j ( 1 − Δ ) 3 H ( s , y ) d y d s . (96)
with K k j ± ( t , x ) = 1 ( 2 π ) 2 ∫ R 2 e i ( ± t ∣ ξ ∣ + x ⋅ ξ ) ψ [ k − 1 , k + 1 ] ( ξ ) ( 1 + ∣ ξ ∣ 2 ) 3 d ξ K_{k}^{j\pm}(t,x)=\frac{1}{(2\pi)^{2}}\int_{\mathbb{R}^{2}}e^{i(\pm t|\xi|+x\cdot\xi)}\frac{\psi_{[k-1,k+1]}(\xi)}{(1+|\xi|^{2})^{3}}\,d\xi K k j ± ( t , x ) = ( 2 π ) 2 1 ∫ R 2 e i ( ± t ∣ ξ ∣ + x ⋅ ξ ) ( 1 + ∣ ξ ∣ 2 ) 3 ψ [ k − 1 , k + 1 ] ( ξ ) d ξ
where R j = ∂ j ∣ ∇ ∣ − 1 R_j=\partial_j|\nabla|^{-1} R j = ∂ j ∣∇ ∣ − 1 (j = 1 , 2 j=1,2 j = 1 , 2 ) is the Riesz transformation and the fact of ψ k ( ξ ) = ψ [ k − 1 , k + 1 ] ( ξ ) ψ k ( ξ ) \psi_k(\xi)=\psi_{[k-1,k+1]}(\xi)\psi_k(\xi) ψ k ( ξ ) = ψ [ k − 1 , k + 1 ] ( ξ ) ψ k ( ξ ) has been used. In the following, it suffices to deal with the estimates of ∂ j w 1 c \partial_j w_1^c ∂ j w 1 c since the estimate of ∂ t w 1 c \partial_t w_1^c ∂ t w 1 c can be treated similarly. Recall (4.28) and (4.30) of [17 ]
∣ P k R j ( 1 − Δ ) 3 H ( s , x ) ∣ ≲ 2 2 k ( 1 + 2 k − k + ∣ x ∣ ) − 5 ∥ ( 1 − Δ ) 3 H ( s , ⋅ ) ∥ L ∞ ( R 2 ) , ∣ K k j ± ( t , x ) ∣ ≲ 2 2 k − 6 k + ( 1 + 2 k ∣ x ∣ ) − 1 / 2 ( 1 + 2 k ∣ t − ∣ x ∣ ∣ ) − 2 . (97) \begin{aligned}
|P_kR_j(1-\Delta)^3H(s,x)| &\lesssim2^{2k}\left(1+2^{k-k_+}|x|\right)^{-5}\|(1-\Delta)^3H(s,\cdot)\|_{L^\infty(\mathbb{R}^2)},\\
|K_k^{j\pm}(t,x)| &\lesssim2^{2k-6k_+}(1+2^{k}|x|)^{-1/2}(1+2^{k}|t-|x||)^{-2}. \tag*{(97)}
\end{aligned} ∣ P k R j ( 1 − Δ ) 3 H ( s , x ) ∣ ∣ K k j ± ( t , x ) ∣ ≲ 2 2 k ( 1 + 2 k − k + ∣ x ∣ ) − 5 ∥ ( 1 − Δ ) 3 H ( s , ⋅ ) ∥ L ∞ ( R 2 ) , ≲ 2 2 k − 6 k + ( 1 + 2 k ∣ x ∣ ) − 1/2 ( 1 + 2 k ∣ t − ∣ x ∣∣ ) − 2 . (97)
Substituting (93) and (97) into (96) yields
∣ ∂ j w 1 c ∣ ≲ ε ∑ k ∈ Z 2 4 k − 6 k + I k , (98) |\partial_jw_1^c|\lesssim\varepsilon\sum_{k\in\mathbb{Z}}2^{4k-6k_+}I^k, \tag*{(98)} ∣ ∂ j w 1 c ∣ ≲ ε k ∈ Z ∑ 2 4 k − 6 k + I k , (98)
I k = ∫ 0 t ∫ R 2 d y d s ( 1 + s ) ln 2 ( 2 + s ) ( 1 + 2 k ∣ t − s − ∣ x − y ∣ ∣ ) 2 ( 1 + 2 k − k + ∣ y ∣ ) 5 . I^k=\int_0^t\int_{\mathbb{R}^2}\frac{dy\,ds}{(1+s)\ln^2(2+s)(1+2^k|t-s-|x-y||)^2(1+2^{k-k_+}|y|)^5}. I k = ∫ 0 t ∫ R 2 ( 1 + s ) ln 2 ( 2 + s ) ( 1 + 2 k ∣ t − s − ∣ x − y ∣∣ ) 2 ( 1 + 2 k − k + ∣ y ∣ ) 5 d y d s .
The integral I k I^k I k can be separated into the following three parts
I k = I 1 k + I 2 k + I 3 k , I^k=I_1^k+I_2^k+I_3^k, I k = I 1 k + I 2 k + I 3 k ,
I 1 k = ∫ ∣ x − y ∣ ≥ t / 2 ∫ 0 t d s d y ( 1 + s ) ln 2 ( 2 + s ) ( 1 + 2 k ∣ t − s − ∣ x − y ∣ ∣ ) 2 ( 1 + 2 k − k + ∣ y ∣ ) 5 , I_1^k=\int_{|x-y|\geq t/2}\int_0^t\frac{ds\,dy}{(1+s)\ln^2(2+s)(1+2^k|t-s-|x-y||)^2(1+2^{k-k_+}|y|)^5}, I 1 k = ∫ ∣ x − y ∣ ≥ t /2 ∫ 0 t ( 1 + s ) ln 2 ( 2 + s ) ( 1 + 2 k ∣ t − s − ∣ x − y ∣∣ ) 2 ( 1 + 2 k − k + ∣ y ∣ ) 5 d s d y ,
I 2 k = ∫ ∣ x − y ∣ ≤ t / 2 ∫ 0 t − ∣ x − y ∣ 2 d s d y ( 1 + s ) ln 2 ( 2 + s ) ( 1 + 2 k ( t − s − ∣ x − y ∣ ) ) 2 ( 1 + 2 k − k + ∣ y ∣ ) 5 , I_2^k=\int_{|x-y|\leq t/2}\int_0^{\frac{t-|x-y|}{2}}\frac{ds\,dy}{(1+s)\ln^2(2+s)(1+2^k(t-s-|x-y|))^2(1+2^{k-k_+}|y|)^5}, I 2 k = ∫ ∣ x − y ∣ ≤ t /2 ∫ 0 2 t − ∣ x − y ∣ ( 1 + s ) ln 2 ( 2 + s ) ( 1 + 2 k ( t − s − ∣ x − y ∣ ) ) 2 ( 1 + 2 k − k + ∣ y ∣ ) 5 d s d y ,
I 3 k = ∫ ∣ x − y ∣ ≤ t / 2 ∫ t − ∣ x − y ∣ 2 t d s d y ( 1 + s ) ln 2 ( 2 + s ) ( 1 + 2 k ∣ t − s − ∣ x − y ∣ ∣ ) 2 ( 1 + 2 k − k + ∣ y ∣ ) 5 . (99) I_3^k=\int_{|x-y|\leq t/2}\int_{\frac{t-|x-y|}{2}}^t\frac{ds\,dy}{(1+s)\ln^2(2+s)(1+2^k|t-s-|x-y||)^2(1+2^{k-k_+}|y|)^5}. \tag*{(99)} I 3 k = ∫ ∣ x − y ∣ ≤ t /2 ∫ 2 t − ∣ x − y ∣ t ( 1 + s ) ln 2 ( 2 + s ) ( 1 + 2 k ∣ t − s − ∣ x − y ∣∣ ) 2 ( 1 + 2 k − k + ∣ y ∣ ) 5 d s d y . (99)
For the integral I 1 k I_1^k I 1 k , one can see that ∣ y ∣ ≥ ∣ x − y ∣ − ∣ x ∣ ≥ t / 2 − t / 4 = t / 4 |y|\geq|x-y|-|x|\geq t/2-t/4=t/4 ∣ y ∣ ≥ ∣ x − y ∣ − ∣ x ∣ ≥ t /2 − t /4 = t /4 and
I 1 k ≲ ⟨ t ⟩ − 3 / 2 2 3 ( k + − k ) / 2 ∫ 0 t ∫ R 2 d s d y ( 1 + s ) ln 2 ( 2 + s ) ( 1 + 2 k − k + ∣ y ∣ ) 7 / 2 ≲ ⟨ t ⟩ − 3 / 2 2 7 ( k + − k ) / 2 . (100) I_1^k\lesssim\langle t\rangle^{-3/2}2^{3(k_+-k)/2}\int_0^t\int_{\mathbb{R}^2}\frac{ds\,dy}{(1+s)\ln^2(2+s)(1+2^{k-k_+}|y|)^{7/2}}
\lesssim\langle t\rangle^{-3/2}2^{7(k_+-k)/2}. \tag*{(100)} I 1 k ≲ ⟨ t ⟩ − 3/2 2 3 ( k + − k ) /2 ∫ 0 t ∫ R 2 ( 1 + s ) ln 2 ( 2 + s ) ( 1 + 2 k − k + ∣ y ∣ ) 7/2 d s d y ≲ ⟨ t ⟩ − 3/2 2 7 ( k + − k ) /2 . (100)
For the integral I 2 k I_2^k I 2 k , one has that t − s − ∣ x − y ∣ ≥ t − ∣ x − y ∣ 2 ≥ t / 4 t-s-|x-y|\geq\frac{t-|x-y|}{2}\geq t/4 t − s − ∣ x − y ∣ ≥ 2 t − ∣ x − y ∣ ≥ t /4 and
I 2 k ≲ ⟨ t ⟩ − 3 / 2 2 3 ( k + − k ) / 2 ∫ 0 t ∫ R 2 d s d y ( 1 + s ) ln 2 ( 2 + s ) ( 1 + 2 k − k + ∣ y ∣ ) 5 ≲ ⟨ t ⟩ − 3 / 2 2 7 ( k + − k ) / 2 . (101) I_2^k\lesssim\langle t\rangle^{-3/2}2^{3(k_+-k)/2}\int_0^t\int_{\mathbb{R}^2}\frac{ds\,dy}{(1+s)\ln^2(2+s)(1+2^{k-k_+}|y|)^5}
\lesssim\langle t\rangle^{-3/2}2^{7(k_+-k)/2}. \tag*{(101)} I 2 k ≲ ⟨ t ⟩ − 3/2 2 3 ( k + − k ) /2 ∫ 0 t ∫ R 2 ( 1 + s ) ln 2 ( 2 + s ) ( 1 + 2 k − k + ∣ y ∣ ) 5 d s d y ≲ ⟨ t ⟩ − 3/2 2 7 ( k + − k ) /2 . (101)
For the integral I 3 k I_3^k I 3 k , due to s ≥ t − ∣ x − y ∣ 2 ≥ t / 4 s\geq\frac{t-|x-y|}{2}\geq t/4 s ≥ 2 t − ∣ x − y ∣ ≥ t /4 , it holds that
I 3 k ≲ ⟨ t ⟩ − 1 ln − 2 ( 2 + t ) ∫ 0 t ∫ R 2 d s d y ( 1 + 2 k ∣ t − s − ∣ x − y ∣ ∣ ) 2 ( 1 + 2 k − k + ∣ y ∣ ) 5 ≲ ⟨ t ⟩ − 1 ln − 2 ( 2 + t ) 2 3 ( k + − k ) . (102) I_3^k\lesssim\langle t\rangle^{-1}\ln^{-2}(2+t)\int_0^t\int_{\mathbb{R}^2}\frac{ds\,dy}{(1+2^k|t-s-|x-y||)^2(1+2^{k-k_+}|y|)^5}
\lesssim\langle t\rangle^{-1}\ln^{-2}(2+t)2^{3(k_+-k)}. \tag*{(102)} I 3 k ≲ ⟨ t ⟩ − 1 ln − 2 ( 2 + t ) ∫ 0 t ∫ R 2 ( 1 + 2 k ∣ t − s − ∣ x − y ∣∣ ) 2 ( 1 + 2 k − k + ∣ y ∣ ) 5 d s d y ≲ ⟨ t ⟩ − 1 ln − 2 ( 2 + t ) 2 3 ( k + − k ) . (102)
Collecting (98) –(102) leads to
∣ ∂ j w 1 c ∣ ≲ ε ⟨ t ⟩ − 1 ln − 2 ( 2 + t ) ∑ k ∈ Z 2 k / 2 − 5 k + / 2 ≲ ε ⟨ t ⟩ − 1 ln − 2 ( 2 + t ) . |\partial_jw_1^c|\lesssim\varepsilon\langle t\rangle^{-1}\ln^{-2}(2+t)\sum_{k\in\mathbb{Z}}2^{k/2-5k_+/2}\lesssim\varepsilon\langle t\rangle^{-1}\ln^{-2}(2+t). ∣ ∂ j w 1 c ∣ ≲ ε ⟨ t ⟩ − 1 ln − 2 ( 2 + t ) k ∈ Z ∑ 2 k /2 − 5 k + /2 ≲ ε ⟨ t ⟩ − 1 ln − 2 ( 2 + t ) .
This completes the proof of Lemma 4.1.
Some estimates of Cauchy problem
Lemma 4.2. Let w w w be the solution of the linear Cauchy problem
{ □ w = F ( t , x ) , ( t , x ) ∈ [ 0 , + ∞ ) × R 2 , ( w , ∂ t w ) ( 0 , x ) = ( 0 , 0 ) , x ∈ R 2 , \begin{cases}
\square w = F(t,x), & (t,x) \in[0,+\infty) \times\mathbb{R}^{2},\\
(w,\partial_t w)(0,x) = (0,0), & x \in\mathbb{R}^{2},
\end{cases} { □ w = F ( t , x ) , ( w , ∂ t w ) ( 0 , x ) = ( 0 , 0 ) , ( t , x ) ∈ [ 0 , + ∞ ) × R 2 , x ∈ R 2 , provided that supp y F ( s , y ) ⊂ { y : 9 s 10 ≤ ∣ y ∣ ≤ s + M 0 } \operatorname{supp}_{y} F(s,y) \subset\{y : \frac{9s}{10} \le|y| \le s+M_{0}\} supp y F ( s , y ) ⊂ { y : 10 9 s ≤ ∣ y ∣ ≤ s + M 0 } and M 0 > 0 M_{0}>0 M 0 > 0 . Then for any constant R > 0 R>0 R > 0 and μ > 0 \mu>0 μ > 0 , it holds that
⟨ t ⟩ ∣ ∂ t w ( t , x ) ∣ ≲ sup ( s , y ) ∈ [ 0 , t ] × R 2 ⟨ y ⟩ 1 / 2 W 1 , 1 + μ ∣ ( 1 − Δ ) 2 F ( s , y ) ∣ . (103) \langle t\rangle|\partial_t w(t,x)| \lesssim\sup_{(s,y)\in[0,t]\times\mathbb{R}^{2}} \langle y\rangle^{1/2}\mathcal{W}_{1,1+\mu}|(1-\Delta)^{2}F(s,y)|. \tag*{(103)} ⟨ t ⟩ ∣ ∂ t w ( t , x ) ∣ ≲ ( s , y ) ∈ [ 0 , t ] × R 2 sup ⟨ y ⟩ 1/2 W 1 , 1 + μ ∣ ( 1 − Δ ) 2 F ( s , y ) ∣. (103) where x ∈ { x ∈ R 2 : ∣ x ∣ ≤ t 10 or ∣ x ∣ ≤ R } x\in\{x\in\mathbb{R}^{2}: |x|\le\frac{t}{10}\ \text{or}\ |x|\le R\} x ∈ { x ∈ R 2 : ∣ x ∣ ≤ 10 t or ∣ x ∣ ≤ R } .
Proof. Without loss of generality, we can assume t ≥ 20 M 0 + 10 R t\ge20M_{0}+10R t ≥ 20 M 0 + 10 R since (103) obviously holds for the case of t ≤ 20 M 0 + 10 R t\le20M_{0}+10R t ≤ 20 M 0 + 10 R . Due to { ∣ x ∣ ≤ R } ⊂ { ∣ x ∣ ≤ t / 10 } \{|x|\le R\}\subset\{|x|\le t/10\} { ∣ x ∣ ≤ R } ⊂ { ∣ x ∣ ≤ t /10 } , we only focus on the case of ∣ x ∣ ≤ t / 10 |x|\le t/10 ∣ x ∣ ≤ t /10 . Note that
∂ t w ( t , x ) = 1 2 ∫ 0 t ∫ R 2 ∑ k ∈ Z ( K ~ k 0 + + K ~ k 0 − ) ( t − s , x − y ) ( 1 − Δ ) 2 F ( s , y ) d y d s \partial_t w(t,x)=\frac{1}{2}\int_{0}^{t}\int_{\mathbb{R}^{2}}\sum_{k\in\mathbb{Z}}\left(\widetilde{K}_{k}^{0+}+\widetilde{K}_{k}^{0-}\right)(t-s,x-y)(1-\Delta)^{2}F(s,y)\,dy\,ds ∂ t w ( t , x ) = 2 1 ∫ 0 t ∫ R 2 k ∈ Z ∑ ( K k 0 + + K k 0 − ) ( t − s , x − y ) ( 1 − Δ ) 2 F ( s , y ) d y d s
with
K ~ k 0 ± ( t , x ) = 1 ( 2 π ) 2 ∫ R 2 e i ( ± t ∣ ξ ∣ + x ⋅ ξ ) ψ k ( ξ ) ( 1 + ∣ ξ ∣ ) 2 d ξ \widetilde{K}_{k}^{0\pm}(t,x)=\frac{1}{(2\pi)^{2}}\int_{\mathbb{R}^{2}}e^{i(\pm t|\xi|+x\cdot\xi)}\frac{\psi_{k}(\xi)}{(1+|\xi|)^{2}}\,d\xi K k 0 ± ( t , x ) = ( 2 π ) 2 1 ∫ R 2 e i ( ± t ∣ ξ ∣ + x ⋅ ξ ) ( 1 + ∣ ξ ∣ ) 2 ψ k ( ξ ) d ξ
and
∣ K ~ k 0 ± ( t , x ) ∣ ≲ 2 2 k − 4 k + ( 1 + 2 k ∣ x ∣ ) − 1 / 2 ( 1 + 2 k ∣ t − ∣ x ∣ ∣ ) − 2 , |\widetilde{K}_{k}^{0\pm}(t,x)|\lesssim2^{2k-4k_{+}}(1+2^{k}|x|)^{-1/2}(1+2^{k}|t-|x||)^{-2}, ∣ K k 0 ± ( t , x ) ∣ ≲ 2 2 k − 4 k + ( 1 + 2 k ∣ x ∣ ) − 1/2 ( 1 + 2 k ∣ t − ∣ x ∣∣ ) − 2 ,
∣ ∂ t w ∣ ≲ ∑ k ∈ Z 2 2 k − 4 k + J k sup ( s , y ) ∈ [ 0 , t ] × R 2 ⟨ y ⟩ 1 / 2 W 1 , 1 + μ ∣ ( 1 − Δ ) 2 F ( s , y ) ∣ . (104) |\partial_t w|\lesssim\sum_{k\in\mathbb{Z}}2^{2k-4k_{+}}J^{k}\sup_{(s,y)\in[0,t]\times\mathbb{R}^{2}}\langle y\rangle^{1/2}\mathcal{W}_{1,1+\mu}|(1-\Delta)^{2}F(s,y)|. \tag*{(104)} ∣ ∂ t w ∣ ≲ k ∈ Z ∑ 2 2 k − 4 k + J k ( s , y ) ∈ [ 0 , t ] × R 2 sup ⟨ y ⟩ 1/2 W 1 , 1 + μ ∣ ( 1 − Δ ) 2 F ( s , y ) ∣. (104)
where
J k = ∫ 0 t ∫ 9 s 10 ≤ ∣ y ∣ ≤ s + M 0 ( 1 + 2 k ∣ x − y ∣ ) − 1 / 2 d y d s ( 1 + s + ∣ y ∣ ) 3 / 2 ( 2 M 0 + s − ∣ y ∣ ) 1 + μ ( 1 + 2 k ∣ t − s − ∣ x − y ∣ ∣ ) 2 . J^{k}=\int_{0}^{t}\int_{\frac{9s}{10}\le|y|\le s+M_{0}}\frac{(1+2^{k}|x-y|)^{-1/2}\,dy\,ds}{(1+s+|y|)^{3/2}(2M_{0}+s-|y|)^{1+\mu}(1+2^{k}|t-s-|x-y||)^{2}}. J k = ∫ 0 t ∫ 10 9 s ≤ ∣ y ∣ ≤ s + M 0 ( 1 + s + ∣ y ∣ ) 3/2 ( 2 M 0 + s − ∣ y ∣ ) 1 + μ ( 1 + 2 k ∣ t − s − ∣ x − y ∣∣ ) 2 ( 1 + 2 k ∣ x − y ∣ ) − 1/2 d y d s .
The integral J k J^{k} J k can be separated into three parts
J k = J 1 k + J 2 k + J 3 k , J^{k}=J_{1}^{k}+J_{2}^{k}+J_{3}^{k}, J k = J 1 k + J 2 k + J 3 k ,
J 1 k = ∫ 0 2 t / 5 ∫ 9 s 10 ≤ ∣ y ∣ ≤ s + M 0 ( 1 + 2 k ∣ x − y ∣ ) − 1 / 2 d y d s ( 1 + s + ∣ y ∣ ) 3 / 2 ( 2 M 0 + s − ∣ y ∣ ) 1 + μ ( 1 + 2 k ∣ t − s − ∣ x − y ∣ ∣ ) 2 , J_{1}^{k}=\int_{0}^{2t/5}\int_{\frac{9s}{10}\le|y|\le s+M_{0}}\frac{(1+2^{k}|x-y|)^{-1/2}\,dy\,ds}{(1+s+|y|)^{3/2}(2M_{0}+s-|y|)^{1+\mu}(1+2^{k}|t-s-|x-y||)^{2}}, J 1 k = ∫ 0 2 t /5 ∫ 10 9 s ≤ ∣ y ∣ ≤ s + M 0 ( 1 + s + ∣ y ∣ ) 3/2 ( 2 M 0 + s − ∣ y ∣ ) 1 + μ ( 1 + 2 k ∣ t − s − ∣ x − y ∣∣ ) 2 ( 1 + 2 k ∣ x − y ∣ ) − 1/2 d y d s ,
J 2 k = ∫ 3 t / 5 t ∫ 9 s 10 ≤ ∣ y ∣ ≤ s + M 0 ( 1 + 2 k ∣ x − y ∣ ) − 1 / 2 d y d s ( 1 + s + ∣ y ∣ ) 3 / 2 ( 2 M 0 + s − ∣ y ∣ ) 1 + μ ( 1 + 2 k ∣ t − s − ∣ x − y ∣ ∣ ) 2 , J_{2}^{k}=\int_{3t/5}^{t}\int_{\frac{9s}{10}\le|y|\le s+M_{0}}\frac{(1+2^{k}|x-y|)^{-1/2}\,dy\,ds}{(1+s+|y|)^{3/2}(2M_{0}+s-|y|)^{1+\mu}(1+2^{k}|t-s-|x-y||)^{2}}, J 2 k = ∫ 3 t /5 t ∫ 10 9 s ≤ ∣ y ∣ ≤ s + M 0 ( 1 + s + ∣ y ∣ ) 3/2 ( 2 M 0 + s − ∣ y ∣ ) 1 + μ ( 1 + 2 k ∣ t − s − ∣ x − y ∣∣ ) 2 ( 1 + 2 k ∣ x − y ∣ ) − 1/2 d y d s ,
J 3 k = ∫ 2 t / 5 3 t / 5 ∫ 9 s 10 ≤ ∣ y ∣ ≤ s + M 0 ( 1 + 2 k ∣ x − y ∣ ) − 1 / 2 d y d s ( 1 + s + ∣ y ∣ ) 3 / 2 ( 2 M 0 + s − ∣ y ∣ ) 1 + μ ( 1 + 2 k ∣ t − s − ∣ x − y ∣ ∣ ) 2 . (105) J_{3}^{k}=\int_{2t/5}^{3t/5}\int_{\frac{9s}{10}\le|y|\le s+M_{0}}\frac{(1+2^{k}|x-y|)^{-1/2}\,dy\,ds}{(1+s+|y|)^{3/2}(2M_{0}+s-|y|)^{1+\mu}(1+2^{k}|t-s-|x-y||)^{2}}. \tag*{(105)} J 3 k = ∫ 2 t /5 3 t /5 ∫ 10 9 s ≤ ∣ y ∣ ≤ s + M 0 ( 1 + s + ∣ y ∣ ) 3/2 ( 2 M 0 + s − ∣ y ∣ ) 1 + μ ( 1 + 2 k ∣ t − s − ∣ x − y ∣∣ ) 2 ( 1 + 2 k ∣ x − y ∣ ) − 1/2 d y d s . (105)
For the integral J 1 k J_{1}^{k} J 1 k , one sees that t − s − ∣ x − y ∣ ≥ t − s − ∣ x ∣ − ∣ y ∣ ≥ t − 2 s − ∣ x ∣ − M 0 ≥ t 10 − M 0 ≥ t 20 t-s-|x-y|\ge t-s-|x|-|y|\ge t-2s-|x|-M_{0}\ge\frac{t}{10}-M_{0}\ge\frac{t}{20} t − s − ∣ x − y ∣ ≥ t − s − ∣ x ∣ − ∣ y ∣ ≥ t − 2 s − ∣ x ∣ − M 0 ≥ 10 t − M 0 ≥ 20 t holds, while for the integral J 2 k J_{2}^{k} J 2 k , one has ∣ t − s − ∣ x − y ∣ ∣ = s + ∣ x − y ∣ − t ≥ s + ∣ y ∣ − ∣ x ∣ − t ≥ t ( 7 50 − 1 10 ) = t 25 |t-s-|x-y||=s+|x-y|-t\ge s+|y|-|x|-t\ge t\left(\frac{7}{50}-\frac{1}{10}\right)=\frac{t}{25} ∣ t − s − ∣ x − y ∣∣ = s + ∣ x − y ∣ − t ≥ s + ∣ y ∣ − ∣ x ∣ − t ≥ t ( 50 7 − 10 1 ) = 25 t .
Therefore,
J 1 k + J 2 k ≲ 2 3 ( k + − k ) / 2 ( 1 + t ) − 3 / 2 ∫ 0 t ∫ ∣ y ∣ ≤ s + M 0 d y d s ( 1 + s + ∣ y ∣ ) 3 / 2 ( 2 M 0 + s − ∣ y ∣ ) 1 + μ ≲ 2 3 ( k + − k ) / 2 ( 1 + t ) − 1 . (106) \begin{aligned}
J_1^k+J_2^k &\lesssim2^{3(k_+-k)/2}(1+t)^{-3/2}\int_0^t\int_{|y|\le s+M_0}\frac{dy\,ds}{(1+s+|y|)^{3/2}(2M_0+s-|y|)^{1+\mu}}\\
&\lesssim2^{3(k_+-k)/2}(1+t)^{-1}. \tag*{(106)}
\end{aligned} J 1 k + J 2 k ≲ 2 3 ( k + − k ) /2 ( 1 + t ) − 3/2 ∫ 0 t ∫ ∣ y ∣ ≤ s + M 0 ( 1 + s + ∣ y ∣ ) 3/2 ( 2 M 0 + s − ∣ y ∣ ) 1 + μ d y d s ≲ 2 3 ( k + − k ) /2 ( 1 + t ) − 1 . (106)
For the integral J 3 k J_3^k J 3 k , ∣ y ∣ ≥ 9 t 25 ≥ 3 ∣ x ∣ |y|\ge\frac{9t}{25}\ge3|x| ∣ y ∣ ≥ 25 9 t ≥ 3∣ x ∣ and ∣ x − y ∣ ≥ 2 ∣ y ∣ 3 ≥ t / 5 |x-y|\ge\frac{2|y|}{3}\ge t/5 ∣ x − y ∣ ≥ 3 2∣ y ∣ ≥ t /5 hold. Thus, we have
J 3 k ≲ ( 1 + t ) − 2 2 ( k + − k ) / 2 J 31 k , (107) J_3^k\lesssim(1+t)^{-2}2^{(k_+-k)/2}J_{31}^k, \tag*{(107)} J 3 k ≲ ( 1 + t ) − 2 2 ( k + − k ) /2 J 31 k , (107)
where
J 31 k = ∫ 2 t / 5 3 t / 5 ∫ 9 s 10 ≤ ∣ y ∣ ≤ s + M 0 d y d s ( 2 M 0 + s − ∣ y ∣ ) 1 + μ ( 1 + 2 k ∣ t − s − ∣ x − y ∣ ∣ ) 2 . J_{31}^k=\int_{2t/5}^{3t/5}\int_{\frac{9s}{10}\le|y|\le s+M_0}\frac{dy\,ds}{(2M_0+s-|y|)^{1+\mu}(1+2^k|t-s-|x-y||)^2}. J 31 k = ∫ 2 t /5 3 t /5 ∫ 10 9 s ≤ ∣ y ∣ ≤ s + M 0 ( 2 M 0 + s − ∣ y ∣ ) 1 + μ ( 1 + 2 k ∣ t − s − ∣ x − y ∣∣ ) 2 d y d s .
For x = 0 x=0 x = 0 , one easily gets
J 31 k ≲ 2 5 ( k + − k ) / 4 ( 1 + t ) . (108) J_{31}^k\lesssim2^{5(k_+-k)/4}(1+t). \tag*{(108)} J 31 k ≲ 2 5 ( k + − k ) /4 ( 1 + t ) . (108)
For x ≠ 0 x\ne0 x = 0 , let ϕ ∈ [ 0 , π ] \phi\in[0,\pi] ϕ ∈ [ 0 , π ] be the angle between x x x and y y y . Taking the coordinate transformation y = ( y 1 , y 2 ) → ( r ~ , λ ~ ) y=(y_1,y_2)\to(\tilde r,\tilde\lambda) y = ( y 1 , y 2 ) → ( r ~ , λ ~ ) :
r ~ = ∣ y ∣ , λ ~ = ∣ y ∣ − ∣ x − y ∣ ∣ x ∣ ∈ [ − 1 , 1 ] , \tilde r=|y|,\qquad\tilde\lambda=\frac{|y|-|x-y|}{|x|}\in[-1,1], r ~ = ∣ y ∣ , λ ~ = ∣ x ∣ ∣ y ∣ − ∣ x − y ∣ ∈ [ − 1 , 1 ] ,
whose Jacobian is
∣ ∂ ( r ~ , λ ~ ) ∂ ( y 1 , y 2 ) ∣ = ∣ sin ϕ ∣ ∣ x − y ∣ . \left|\frac{\partial(\tilde r,\tilde\lambda)}{\partial(y_1,y_2)}\right|=\frac{|\sin\phi|}{|x-y|}. ∂ ( y 1 , y 2 ) ∂ ( r ~ , λ ~ ) = ∣ x − y ∣ ∣ sin ϕ ∣ .
If ϕ ≥ π / 6 \phi\ge\pi/6 ϕ ≥ π /6 , then it holds that
1 − λ ~ 2 = ∣ x ∣ 2 − ( ∣ y ∣ − ∣ x − y ∣ ) 2 ∣ x ∣ 2 = ( ∣ x ∣ + ∣ x − y ∣ − ∣ y ∣ ) ( ∣ x ∣ + ∣ y ∣ − ∣ x − y ∣ ) ∣ x ∣ 2 = ( ∣ x ∣ + ∣ x − y ∣ − ∣ y ∣ ) ( ∣ x ∣ 2 + ∣ y ∣ 2 + 2 ∣ x ∣ ∣ y ∣ − ∣ x − y ∣ 2 ) ∣ x ∣ 2 ( ∣ x ∣ + ∣ x − y ∣ + ∣ y ∣ ) = 2 ∣ x ∣ ∣ y ∣ ( ∣ x ∣ + ∣ x − y ∣ − ∣ y ∣ ) ( 1 + cos ϕ ) ∣ x ∣ 2 ( ∣ x ∣ + ∣ x − y ∣ + ∣ y ∣ ) ≲ cos 2 ϕ 2 = sin 2 ϕ 4 sin 2 ϕ 2 ≲ ∣ x − y ∣ 2 ∣ ∂ ( r ~ , λ ~ ) ∂ ( y 1 , y 2 ) ∣ 2 . (109) \begin{aligned}
1-\tilde\lambda^2
&=\frac{|x|^2-(|y|-|x-y|)^2}{|x|^2}\\
&=\frac{(|x|+|x-y|-|y|)(|x|+|y|-|x-y|)}{|x|^2}\\
&=\frac{(|x|+|x-y|-|y|)(|x|^2+|y|^2+2|x||y|-|x-y|^2)}{|x|^2(|x|+|x-y|+|y|)}\\
&=\frac{2|x||y|(|x|+|x-y|-|y|)(1+\cos\phi)}{|x|^2(|x|+|x-y|+|y|)}\\
&\lesssim\cos^2\frac{\phi}{2}=\frac{\sin^2\phi}{4\sin^2\frac{\phi}{2}}\lesssim|x-y|^2\left|\frac{\partial(\tilde r,\tilde\lambda)}{\partial(y_1,y_2)}\right|^2. \tag*{(109)}
\end{aligned} 1 − λ ~ 2 = ∣ x ∣ 2 ∣ x ∣ 2 − ( ∣ y ∣ − ∣ x − y ∣ ) 2 = ∣ x ∣ 2 ( ∣ x ∣ + ∣ x − y ∣ − ∣ y ∣ ) ( ∣ x ∣ + ∣ y ∣ − ∣ x − y ∣ ) = ∣ x ∣ 2 ( ∣ x ∣ + ∣ x − y ∣ + ∣ y ∣ ) ( ∣ x ∣ + ∣ x − y ∣ − ∣ y ∣ ) ( ∣ x ∣ 2 + ∣ y ∣ 2 + 2∣ x ∣∣ y ∣ − ∣ x − y ∣ 2 ) = ∣ x ∣ 2 ( ∣ x ∣ + ∣ x − y ∣ + ∣ y ∣ ) 2∣ x ∣∣ y ∣ ( ∣ x ∣ + ∣ x − y ∣ − ∣ y ∣ ) ( 1 + cos ϕ ) ≲ cos 2 2 ϕ = 4 sin 2 2 ϕ sin 2 ϕ ≲ ∣ x − y ∣ 2 ∂ ( y 1 , y 2 ) ∂ ( r ~ , λ ~ ) 2 . (109)
If ϕ ≤ π / 6 \phi\le\pi/6 ϕ ≤ π /6 , then ∣ y ∣ ≥ 3 ∣ x ∣ |y|\ge3|x| ∣ y ∣ ≥ 3∣ x ∣ implies ψ > π / 2 \psi>\pi/2 ψ > π /2 , where ψ \psi ψ is the angle between x x x and x − y x-y x − y (or between − x -x − x and y − x y-x y − x ). On the other hand, it follows from 2 ∣ y ∣ / 3 ≤ ∣ x − y ∣ ≤ 4 ∣ y ∣ / 3 2|y|/3\le|x-y|\le4|y|/3 2∣ y ∣/3 ≤ ∣ x − y ∣ ≤ 4∣ y ∣/3 that sin ϕ sin ψ = ∣ y − x ∣ ∣ y ∣ ∈ [ 2 / 3 , 4 / 3 ] \frac{\sin\phi}{\sin\psi}=\frac{|y-x|}{|y|}\in[2/3,4/3] s i n ψ s i n ϕ = ∣ y ∣ ∣ y − x ∣ ∈ [ 2/3 , 4/3 ] . Analogously to (109) , we can achieve
1 − λ ~ 2 = ( ∣ x ∣ + ∣ x − y ∣ − ∣ x − ( x − y ) ∣ ) ( ∣ x ∣ + ∣ y ∣ − ∣ x − y ∣ ) ∣ x ∣ 2 = 2 ∣ x ∣ ∣ x − y ∣ ( 1 + cos ψ ) ( ∣ x ∣ + ∣ y ∣ − ∣ x − y ∣ ) ∣ x ∣ 2 ( ∣ x ∣ + ∣ x − y ∣ + ∣ y ∣ ) ≲ cos 2 ψ 2 = sin 2 ψ 4 sin 2 ψ 2 ≲ ∣ x − y ∣ 2 ∣ ∂ ( r ~ , λ ~ ) ∂ ( y 1 , y 2 ) ∣ 2 . \begin{aligned}
1-\tilde\lambda^2
&=\frac{(|x|+|x-y|-|x-(x-y)|)(|x|+|y|-|x-y|)}{|x|^2}\\
&=\frac{2|x||x-y|(1+\cos\psi)(|x|+|y|-|x-y|)}{|x|^2(|x|+|x-y|+|y|)}\\
&\lesssim\cos^2\frac{\psi}{2}=\frac{\sin^2\psi}{4\sin^2\frac{\psi}{2}}\lesssim|x-y|^2\left|\frac{\partial(\tilde r,\tilde\lambda)}{\partial(y_1,y_2)}\right|^2.
\end{aligned} 1 − λ ~ 2 = ∣ x ∣ 2 ( ∣ x ∣ + ∣ x − y ∣ − ∣ x − ( x − y ) ∣ ) ( ∣ x ∣ + ∣ y ∣ − ∣ x − y ∣ ) = ∣ x ∣ 2 ( ∣ x ∣ + ∣ x − y ∣ + ∣ y ∣ ) 2∣ x ∣∣ x − y ∣ ( 1 + cos ψ ) ( ∣ x ∣ + ∣ y ∣ − ∣ x − y ∣ ) ≲ cos 2 2 ψ = 4 sin 2 2 ψ sin 2 ψ ≲ ∣ x − y ∣ 2 ∂ ( y 1 , y 2 ) ∂ ( r ~ , λ ~ ) 2 .
Therefore,
∣ ∂ ( r ~ , λ ~ ) ∂ ( y 1 , y 2 ) ∣ − 1 ≲ ∣ x − y ∣ 1 − λ ~ 2 ≲ 1 + t 1 − λ ~ 2 . \left\lvert\frac{\partial(\tilde{r},\tilde{\lambda})}{\partial(y_1,y_2)} \right\rvert^{-1} \lesssim\frac{\lvert x-y\rvert}{\sqrt{1-\tilde{\lambda}^2}} \lesssim\frac{1+t}{\sqrt{1-\tilde{\lambda}^2}}. ∂ ( y 1 , y 2 ) ∂ ( r ~ , λ ~ ) − 1 ≲ 1 − λ ~ 2 ∣ x − y ∣ ≲ 1 − λ ~ 2 1 + t .
In the new coordinate ( r ~ , λ ~ ) (\tilde{r},\tilde{\lambda}) ( r ~ , λ ~ ) , the integral J 31 k J_{31}^{k} J 31 k can be treated as
J 31 k ≲ ( 1 + t ) ∫ 0 t ∫ 0 s + M 0 ∫ − 1 1 d λ ~ d r ~ d s 1 − λ ~ 2 ( 2 M 0 + s − r ~ ) 1 + μ ( 1 + 2 k ∣ t − s − r ~ + ∣ x ∣ ∣ λ ~ ∣ ) 2 . J_{31}^{k} \lesssim(1+t)\int_0^t\int_0^{s+M_0}\int_{-1}^{1}\frac{d\tilde{\lambda}d\tilde{r}ds}{\sqrt{1-\tilde{\lambda}^2}(2M_0+s-\tilde{r})^{1+\mu}(1+2^k\lvert t-s-\tilde{r}+\lvert x\rvert\lvert\tilde{\lambda}\rvert)^2}. J 31 k ≲ ( 1 + t ) ∫ 0 t ∫ 0 s + M 0 ∫ − 1 1 1 − λ ~ 2 ( 2 M 0 + s − r ~ ) 1 + μ ( 1 + 2 k ∣ t − s − r ~ + ∣ x ∣ ∣ λ ~ ∣ ) 2 d λ ~ d r ~ d s .
Set ζ ± = s ± r ~ \zeta_{\pm}=s\pm\tilde{r} ζ ± = s ± r ~ . Then has
J 31 k ≲ 2 5 ( k + − k ) / 4 ( 1 + t ) ∫ − 1 1 ∫ 0 t ∫ 0 2 t + M 0 d ζ + d ζ − d λ ~ 1 − λ ~ 2 ( 2 M 0 + ζ − ) 1 + μ ( 1 + ∣ t − ζ + + ∣ x ∣ ∣ λ ~ ∣ ∣ ) 5 / 4 J_{31}^{k}\lesssim2^{5(k_+-k)/4}(1+t)\int_{-1}^{1}\int_0^t\int_0^{2t+M_0}\frac{d\zeta_+d\zeta_-d\tilde{\lambda}}{\sqrt{1-\tilde{\lambda}^2}(2M_0+\zeta_-)^{1+\mu}(1+\lvert t-\zeta_++\lvert x\rvert\lvert\tilde{\lambda}\rvert\rvert)^{5/4}} J 31 k ≲ 2 5 ( k + − k ) /4 ( 1 + t ) ∫ − 1 1 ∫ 0 t ∫ 0 2 t + M 0 1 − λ ~ 2 ( 2 M 0 + ζ − ) 1 + μ ( 1 + ∣ t − ζ + + ∣ x ∣ ∣ λ ~ ∣∣ ) 5/4 d ζ + d ζ − d λ ~
≲ 2 5 ( k + − k ) / 4 ( 1 + t ) ∫ − 1 1 ∫ 0 t d ζ − d λ ~ 1 − λ ~ 2 ( 2 M 0 + ζ − ) 1 + μ \lesssim2^{5(k_+-k)/4}(1+t)\int_{-1}^{1}\int_0^t\frac{d\zeta_-d\tilde{\lambda}}{\sqrt{1-\tilde{\lambda}^2}(2M_0+\zeta_-)^{1+\mu}} ≲ 2 5 ( k + − k ) /4 ( 1 + t ) ∫ − 1 1 ∫ 0 t 1 − λ ~ 2 ( 2 M 0 + ζ − ) 1 + μ d ζ − d λ ~
≲ 2 5 ( k + − k ) / 4 ( 1 + t ) . (110) \lesssim2^{5(k_+-k)/4}(1+t). \tag*{(110)} ≲ 2 5 ( k + − k ) /4 ( 1 + t ) . (110)
Collecting (104) -(108) and (110) yields (103) .
Based on Lemma 4.2 and Lemma 2.5, we have
Corollary 4.3. Let w w w be the solution to the linear initial boundary value problem
{ □ w = ∂ t F ( t , x ) , ( t , x ) ∈ [ 0 , + ∞ ) × K , w ( t , x ) = 0 , ( t , x ) ∈ [ 0 , + ∞ ) × ∂ K , ( w , ∂ t w ) ( 0 , x ) = ( 0 , 0 ) , x ∈ K , \begin{cases}
\square w=\partial_tF(t,x), & (t,x)\in[0,+\infty)\times\mathcal{K},\\
w(t,x)=0, & (t,x)\in[0,+\infty)\times\partial\mathcal{K},\\
(w,\partial_tw)(0,x)=(0,0), & x\in\mathcal{K},
\end{cases} ⎩ ⎨ ⎧ □ w = ∂ t F ( t , x ) , w ( t , x ) = 0 , ( w , ∂ t w ) ( 0 , x ) = ( 0 , 0 ) , ( t , x ) ∈ [ 0 , + ∞ ) × K , ( t , x ) ∈ [ 0 , + ∞ ) × ∂ K , x ∈ K , provided that supp y F ( s , y ) ⊂ { y : 9 s 10 ≤ ∣ y ∣ ≤ s + M 0 } \operatorname{supp}_yF(s,y)\subset\{y:\frac{9s}{10}\leq\lvert y\rvert\leq s+M_0\} supp y F ( s , y ) ⊂ { y : 10 9 s ≤ ∣ y ∣ ≤ s + M 0 } and M 0 > 0 M_0>0 M 0 > 0 . Then for any constant R > 0 R>0 R > 0 and μ > 0 \mu>0 μ > 0 , it holds that
⟨ t ⟩ ∣ w ( t , x ) ∣ ≲ sup y ∣ F ( 0 , y ) ∣ + sup ( s , y ) ∈ [ 0 , t ] × R 2 ⟨ y ⟩ 1 / 2 W 1 , 1 + μ ∣ ( 1 − Δ ) 2 F ( s , y ) ∣ , (111) \langle t\rangle\lvert w(t,x)\rvert\lesssim\sup_y\lvert F(0,y)\rvert+\sup_{(s,y)\in[0,t]\times\mathbb{R}^2}\langle y\rangle^{1/2}\mathcal{W}_{1,1+\mu}\lvert(1-\Delta)^2F(s,y)\rvert, \tag*{(111)} ⟨ t ⟩ ∣ w ( t , x )∣ ≲ y sup ∣ F ( 0 , y )∣ + ( s , y ) ∈ [ 0 , t ] × R 2 sup ⟨ y ⟩ 1/2 W 1 , 1 + μ ∣( 1 − Δ ) 2 F ( s , y )∣ , (111) where x ∈ { x ∈ R 2 : ∣ x ∣ ≤ t 10 or ∣ x ∣ ≤ R } x\in\{x\in\mathbb{R}^2:\lvert x\rvert\leq\frac{t}{10}\text{ or }\lvert x\rvert\leq R\} x ∈ { x ∈ R 2 : ∣ x ∣ ≤ 10 t or ∣ x ∣ ≤ R } .
Proof. The solution can be decomposed into w = ∂ t w 0 + w r w=\partial_tw_0+w_r w = ∂ t w 0 + w r , where
□ w 0 = F ( t , x ) , ( w 0 , ∂ t w 0 ) ( 0 , x ) = ( 0 , 0 ) ; □ w r = 0 , ( w r , ∂ t w r ) ( 0 , x ) = ( 0 , − F ( 0 , x ) ) . \square w_0=F(t,x),\qquad(w_0,\partial_tw_0)(0,x)=(0,0);\qquad\square w_r=0,\qquad(w_r,\partial_tw_r)(0,x)=(0,-F(0,x)). □ w 0 = F ( t , x ) , ( w 0 , ∂ t w 0 ) ( 0 , x ) = ( 0 , 0 ) ; □ w r = 0 , ( w r , ∂ t w r ) ( 0 , x ) = ( 0 , − F ( 0 , x )) .
This, together with (41) and (103) , yields the proof of (111) .
Pointwise estimates of the linear initial boundary value problem
This section aims to establish some important pointwise estimates by use of Lemma 4.2.
Lemma 5.1. Suppose that the obstacle O ⊂ R 2 \mathcal{O} \subset\mathbb{R}^{2} O ⊂ R 2 is star-shaped, K = R 2 ∖ O \mathcal{K}=\mathbb{R}^{2}\setminus\mathcal{O} K = R 2 ∖ O and let w w w be the solution of the IBVP
{ □ w = F 0 ( t , x ) + F 1 ( t , x ) , ( t , x ) ∈ [ 0 , ∞ ) × K , w = 0 , ( t , x ) ∈ [ 0 , ∞ ) × ∂ K , ( w , ∂ t w ) ( 0 , x ) = ( w 0 , w 1 ) ( x ) , x ∈ K , (112) \begin{cases}
\Box w=F_{0}(t,x)+F_{1}(t,x), & (t,x)\in[0,\infty)\times\mathcal{K},\\
w=0, & (t,x)\in[0,\infty)\times\partial\mathcal{K},\\
(w,\partial_{t}w)(0,x)=(w_{0},w_{1})(x), & x\in\mathcal{K}, \tag*{(112)}
\end{cases} ⎩ ⎨ ⎧ □ w = F 0 ( t , x ) + F 1 ( t , x ) , w = 0 , ( w , ∂ t w ) ( 0 , x ) = ( w 0 , w 1 ) ( x ) , ( t , x ) ∈ [ 0 , ∞ ) × K , ( t , x ) ∈ [ 0 , ∞ ) × ∂ K , x ∈ K , (112) where ( w 0 , w 1 ) (w_{0},w_{1}) ( w 0 , w 1 ) has compact support, supp x F 0 ( t , x ) ⊂ { x : 9 t / 10 ≤ ∣ x ∣ ≤ t + M 0 } \operatorname{supp}_{x}F_{0}(t,x)\subset\{x:9t/10\leq|x|\leq t+M_{0}\} supp x F 0 ( t , x ) ⊂ { x : 9 t /10 ≤ ∣ x ∣ ≤ t + M 0 } and supp x F 1 ( t , x ) ⊂ { x : ∣ x ∣ ≤ t + M 0 } \operatorname{supp}_{x}F_{1}(t,x)\subset\{x:|x|\leq t+M_{0}\} supp x F 1 ( t , x ) ⊂ { x : ∣ x ∣ ≤ t + M 0 } . Then for μ ∈ ( 0 , 1 / 2 ) \mu\in(0,1/2) μ ∈ ( 0 , 1/2 ) , one has
⟨ t ⟩ ∥ ∂ ≤ 1 ∂ t w ( t , x ) ∥ L 2 ( K R ) ≲ ∥ ( w 0 , w 1 ) ∥ H 2 ( K ) + ∑ l = 0 , 1 ∑ ∣ a ∣ ≤ 1 sup s ∈ [ 0 , t ] ⟨ s ⟩ ∥ ∂ a F l ( s ) ∥ L 2 ( K 3 ) + ∑ ∣ a ∣ ≤ 5 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 , 1 + μ ( s , y ) ∣ ∂ a F 0 ( s , y ) ∣ + ∑ ∣ a ∣ ≤ 2 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 + μ , 1 ( s , y ) ∣ Z a F 1 ( s , y ) ∣ . (113) \begin{aligned}
\langle t\rangle\|\partial^{\leq1}\partial_{t}w(t,x)\|_{L^{2}(\mathcal{K}_{R})}
\lesssim{}&\|(w_{0},w_{1})\|_{H^{2}(\mathcal{K})}
+\sum_{l=0,1}\sum_{|a|\leq1}\sup_{s\in[0,t]}\langle s\rangle\|\partial^{a}F_{l}(s)\|_{L^{2}(\mathcal{K}_{3})}\\
&+\sum_{|a|\leq5}\sup_{(s,y)\in[0,t]\times\mathbb{R}^{2}\setminus\mathcal{K}_{2}}
\langle y\rangle^{1/2}W_{1,1+\mu}(s,y)|\partial^{a}F_{0}(s,y)|\\
&+\sum_{|a|\leq2}\sup_{(s,y)\in[0,t]\times\mathbb{R}^{2}\setminus\mathcal{K}_{2}}
\langle y\rangle^{1/2}W_{1+\mu,1}(s,y)|Z^{a}F_{1}(s,y)|. \tag*{(113)}
\end{aligned} ⟨ t ⟩ ∥ ∂ ≤ 1 ∂ t w ( t , x ) ∥ L 2 ( K R ) ≲ ∥ ( w 0 , w 1 ) ∥ H 2 ( K ) + l = 0 , 1 ∑ ∣ a ∣ ≤ 1 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ ∥ ∂ a F l ( s ) ∥ L 2 ( K 3 ) + ∣ a ∣ ≤ 5 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 , 1 + μ ( s , y ) ∣ ∂ a F 0 ( s , y ) ∣ + ∣ a ∣ ≤ 2 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 + μ , 1 ( s , y ) ∣ Z a F 1 ( s , y ) ∣. (113)
Remark 5.1. The support property of F 0 ( t , x ) F_{0}(t,x) F 0 ( t , x ) in (112) is essential in the proof of Theorem 1.1 without the second null condition. The estimate (113) actually improves the estimate (5.25) of [17 ]
⟨ t ⟩ ∥ ∂ t w ( t , x ) ∥ L ∞ ( K R ) ≲ ∥ ( w 0 , w 1 ) ∥ H 3 ( K ) + ∑ ∣ a ∣ ≤ 2 sup s ∈ [ 0 , t ] ⟨ s ⟩ ∥ ∂ a F ( s ) ∥ L 2 ( K 3 ) + ∑ ∣ a ∣ ≤ 2 sup y ∈ K ∣ ∂ a F ( 0 , y ) ∣ + ∑ ∣ a ∣ ≤ 3 sup ( s , y ) ∈ [ 0 , t ] × K ⟨ y ⟩ 1 / 2 W 1 + μ + ν , 1 ( s , y ) ∣ Z a F ( s , y ) ∣ , \begin{aligned}
\langle t\rangle\|\partial_{t}w(t,x)\|_{L^{\infty}(\mathcal{K}_{R})}
\lesssim{}&\|(w_{0},w_{1})\|_{H^{3}(\mathcal{K})}
+\sum_{|a|\leq2}\sup_{s\in[0,t]}\langle s\rangle\|\partial^{a}F(s)\|_{L^{2}(\mathcal{K}_{3})}\\
&+\sum_{|a|\leq2}\sup_{y\in\mathcal{K}}|\partial^{a}F(0,y)|
+\sum_{|a|\leq3}\sup_{(s,y)\in[0,t]\times\mathcal{K}}
\langle y\rangle^{1/2}W_{1+\mu+\nu,1}(s,y)|Z^{a}F(s,y)|,
\end{aligned} ⟨ t ⟩ ∥ ∂ t w ( t , x ) ∥ L ∞ ( K R ) ≲ ∥ ( w 0 , w 1 ) ∥ H 3 ( K ) + ∣ a ∣ ≤ 2 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ ∥ ∂ a F ( s ) ∥ L 2 ( K 3 ) + ∣ a ∣ ≤ 2 ∑ y ∈ K sup ∣ ∂ a F ( 0 , y ) ∣ + ∣ a ∣ ≤ 3 ∑ ( s , y ) ∈ [ 0 , t ] × K sup ⟨ y ⟩ 1/2 W 1 + μ + ν , 1 ( s , y ) ∣ Z a F ( s , y ) ∣ ,
where w w w is the solution of □ w = F \Box w=F □ w = F , and μ , ν > 0 \mu,\nu>0 μ , ν > 0 are constants.
Proof. Although the proof is somewhat similar to that of (5.25) in [17 ] through utilizing Lemma 4.2 , we still give the details for readers’ convenience. Let w 1 b , w 2 b w_{1}^{b},w_{2}^{b} w 1 b , w 2 b be the solutions to
{ □ w 1 b = ( 1 − χ [ 2 , 3 ] ( x ) ) ( F 0 ( t , x ) + F 1 ( t , x ) ) , □ w 2 b = χ [ 2 , 3 ] ( x ) ( F 0 ( t , x ) + F 1 ( t , x ) ) , w 1 b ∣ ∂ K = w 2 b ∣ ∂ K = 0 , ( w 1 b , ∂ t w 1 b ) ( 0 , x ) = ( w 0 , w 1 ) ( x ) , ( w 2 b , ∂ t w 2 b ) ( 0 , x ) = ( 0 , 0 ) , (114) \begin{cases}
\Box w_{1}^{b}=(1-\chi_{[2,3]}(x))(F_{0}(t,x)+F_{1}(t,x)),\\
\Box w_{2}^{b}=\chi_{[2,3]}(x)(F_{0}(t,x)+F_{1}(t,x)),\\
w_{1}^{b}|_{\partial\mathcal{K}}=w_{2}^{b}|_{\partial\mathcal{K}}=0,\\
(w_{1}^{b},\partial_{t}w_{1}^{b})(0,x)=(w_{0},w_{1})(x),\qquad
(w_{2}^{b},\partial_{t}w_{2}^{b})(0,x)=(0,0), \tag*{(114)}
\end{cases} ⎩ ⎨ ⎧ □ w 1 b = ( 1 − χ [ 2 , 3 ] ( x )) ( F 0 ( t , x ) + F 1 ( t , x )) , □ w 2 b = χ [ 2 , 3 ] ( x ) ( F 0 ( t , x ) + F 1 ( t , x )) , w 1 b ∣ ∂ K = w 2 b ∣ ∂ K = 0 , ( w 1 b , ∂ t w 1 b ) ( 0 , x ) = ( w 0 , w 1 ) ( x ) , ( w 2 b , ∂ t w 2 b ) ( 0 , x ) = ( 0 , 0 ) , (114)
respectively. It follows from the uniqueness of smooth solution to the IBVP with (112) and (114) that w = w 1 b + w 2 b w=w_{1}^{b}+w_{2}^{b} w = w 1 b + w 2 b . By (32) with m = 2 m=2 m = 2 and η = 1 \eta=1 η = 1 , one has
⟨ t ⟩ ∥ ∂ ≤ 1 ∂ t w 1 b ( t , x ) ∥ L 2 ( K R ) ≲ ∥ ( w 0 , w 1 ) ∥ H 2 ( K ) + ∑ l = 0 , 1 ∑ ∣ a ∣ ≤ 1 sup s ∈ [ 0 , t ] ⟨ s ⟩ ∥ ∂ a F l ( s ) ∥ L 2 ( K 3 ) . (115) \langle t\rangle\|\partial^{\leq1}\partial_{t}w_{1}^{b}(t,x)\|_{L^{2}(\mathcal{K}_{R})}
\lesssim\|(w_{0},w_{1})\|_{H^{2}(\mathcal{K})}
+\sum_{l=0,1}\sum_{|a|\leq1}\sup_{s\in[0,t]}\langle s\rangle\|\partial^{a}F_{l}(s)\|_{L^{2}(\mathcal{K}_{3})}. \tag*{(115)} ⟨ t ⟩ ∥ ∂ ≤ 1 ∂ t w 1 b ( t , x ) ∥ L 2 ( K R ) ≲ ∥ ( w 0 , w 1 ) ∥ H 2 ( K ) + l = 0 , 1 ∑ ∣ a ∣ ≤ 1 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ ∥ ∂ a F l ( s ) ∥ L 2 ( K 3 ) . (115)
Next, we turn to the estimate of w 2 b w_{2}^{b} w 2 b . Let w 2 c w_{2}^{c} w 2 c be the solution of the Cauchy problem
□ w 2 c = { χ [ 2 , 3 ] ( x ) ( F 0 ( t , x ) + F 1 ( t , x ) ) , x ∈ R 2 ∖ O , 0 , x ∈ O ‾ , (116) \Box w_{2}^{c}=
\begin{cases}
\chi_{[2,3]}(x)(F_{0}(t,x)+F_{1}(t,x)), & x\in\mathbb{R}^{2}\setminus\mathcal{O},\\
0, & x\in\overline{\mathcal{O}},
\end{cases} \tag*{(116)} □ w 2 c = { χ [ 2 , 3 ] ( x ) ( F 0 ( t , x ) + F 1 ( t , x )) , 0 , x ∈ R 2 ∖ O , x ∈ O , (116)
with ( w 2 c , ∂ t w 2 c ) ( 0 , x ) = ( 0 , 0 ) (w_{2}^{c},\partial_{t}w_{2}^{c})(0,x)=(0,0) ( w 2 c , ∂ t w 2 c ) ( 0 , x ) = ( 0 , 0 ) . It is easy to find that v 2 b = ∂ t w 2 b − χ [ 1 , 2 ] ( x ) ∂ t w 2 c v_{2}^{b}=\partial_{t}w_{2}^{b}-\chi_{[1,2]}(x)\partial_{t}w_{2}^{c} v 2 b = ∂ t w 2 b − χ [ 1 , 2 ] ( x ) ∂ t w 2 c defined on [ 0 , ∞ ) × K [0,\infty)\times\mathcal{K} [ 0 , ∞ ) × K is the solution of the IBVP
□ v 2 b = [ Δ , χ [ 1 , 2 ] ] ∂ t w 2 c , v 2 b ∣ ∂ K = 0 \Box v_{2}^{b}=[\Delta,\chi_{[1,2]}]\partial_{t}w_{2}^{c},\qquad v_{2}^{b}|_{\partial\mathcal{K}}=0 □ v 2 b = [ Δ , χ [ 1 , 2 ] ] ∂ t w 2 c , v 2 b ∣ ∂ K = 0
with ( v 2 b , ∂ t v 2 b ) ( 0 , x ) = ( 0 , 0 ) (v_2^b,\partial_t v_2^b)(0,x)=(0,0) ( v 2 b , ∂ t v 2 b ) ( 0 , x ) = ( 0 , 0 ) . Since [ Δ , χ [ 1 , 2 ] ] ∂ t w 2 c [\Delta,\chi_{[1,2]}]\partial_t w_2^c [ Δ , χ [ 1 , 2 ] ] ∂ t w 2 c has a compact support in the space variable, the estimate of v 2 b v_2^b v 2 b is analogous to that of w 1 b w_1^b w 1 b in (115) and it holds that
⟨ t ⟩ ∥ ∂ ≤ 1 v 2 b ( t , x ) ∥ L 2 ( K R ) ≲ sup s ∈ [ 0 , t ] ⟨ s ⟩ ∥ [ Δ , χ [ 1 , 2 ] ] ∂ t w 2 c ( s ) ∥ L 2 ( K ) \langle t\rangle\|\partial^{\leq1}v_2^b(t,x)\|_{L^2(K_R)}\lesssim\sup_{s\in[0,t]}\langle s\rangle\|[\Delta,\chi_{[1,2]}]\partial_t w_2^c(s)\|_{L^2(K)} ⟨ t ⟩ ∥ ∂ ≤ 1 v 2 b ( t , x ) ∥ L 2 ( K R ) ≲ s ∈ [ 0 , t ] sup ⟨ s ⟩ ∥ [ Δ , χ [ 1 , 2 ] ] ∂ t w 2 c ( s ) ∥ L 2 ( K )
≲ sup s ∈ [ 0 , t ] ⟨ s ⟩ ∥ ∂ x ≤ 1 ∂ t w 2 c ( s ) ∥ L ∞ ( ∣ x ∣ ≤ 2 ) . (117) \lesssim\sup_{s\in[0,t]}\langle s\rangle\|\partial_x^{\leq1}\partial_t w_2^c(s)\|_{L^\infty(|x|\leq2)}. \tag*{(117)} ≲ s ∈ [ 0 , t ] sup ⟨ s ⟩ ∥ ∂ x ≤ 1 ∂ t w 2 c ( s ) ∥ L ∞ ( ∣ x ∣ ≤ 2 ) . (117)
On the other hand, the solution of problem (116) can be decomposed into w 2 c = w 2 , 1 c + w 2 , 2 c w_2^c=w_{2,1}^c+w_{2,2}^c w 2 c = w 2 , 1 c + w 2 , 2 c , where w 2 , l c w_{2,l}^c w 2 , l c are the solutions of the problems □ w 2 , l c = χ [ 2 , 3 ] ( x ) F l ( t , x ) \Box w_{2,l}^c=\chi_{[2,3]}(x)F_l(t,x) □ w 2 , l c = χ [ 2 , 3 ] ( x ) F l ( t , x ) with ( w 2 , l c , ∂ t w 2 , l c ) ( 0 , x ) = ( 0 , 0 ) (w_{2,l}^c,\partial_t w_{2,l}^c)(0,x)=(0,0) ( w 2 , l c , ∂ t w 2 , l c ) ( 0 , x ) = ( 0 , 0 ) (l = 0 , 1 l=0,1 l = 0 , 1 ), respectively. Applying Lemma 4.2 and (43) to w 2 , 0 c w_{2,0}^c w 2 , 0 c and w 2 , 1 c w_{2,1}^c w 2 , 1 c , respectively, yields
⟨ t ⟩ ∥ ∂ x ≤ 1 ∂ t w 2 , 0 c ( t , x ) ∥ L ∞ ( ∣ x ∣ ≤ t / 10 ) ≲ ∑ ∣ a ∣ ≤ 5 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 , 1 + δ 0 ( s , y ) ∣ ∂ a F 0 ( s , y ) ∣ , \langle t\rangle\|\partial_x^{\leq1}\partial_t w_{2,0}^c(t,x)\|_{L^\infty(|x|\leq t/10)}
\lesssim\sum_{|a|\leq5}\sup_{\substack{(s,y)\in[0,t]\times\mathbb{R}^2\setminus K_2}}\langle y\rangle^{1/2}W_{1,1+\delta_0}(s,y)|\partial^a F_0(s,y)|, ⟨ t ⟩ ∥ ∂ x ≤ 1 ∂ t w 2 , 0 c ( t , x ) ∥ L ∞ ( ∣ x ∣ ≤ t /10 ) ≲ ∣ a ∣ ≤ 5 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 , 1 + δ 0 ( s , y ) ∣ ∂ a F 0 ( s , y ) ∣ ,
⟨ t − ∣ x ∣ ⟩ ∥ ∂ x ≤ 1 ∂ t w 2 , 1 c ( t , x ) ∥ ≲ ∑ ∣ a ∣ ≤ 2 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 + δ 0 , 1 ( s , y ) ∣ Z a F 1 ( s , y ) ∣ . (118) \langle t-|x|\rangle\|\partial_x^{\leq1}\partial_t w_{2,1}^c(t,x)\|
\lesssim\sum_{|a|\leq2}\sup_{\substack{(s,y)\in[0,t]\times\mathbb{R}^2\setminus K_2}}\langle y\rangle^{1/2}W_{1+\delta_0,1}(s,y)|Z^a F_1(s,y)|. \tag*{(118)} ⟨ t − ∣ x ∣ ⟩ ∥ ∂ x ≤ 1 ∂ t w 2 , 1 c ( t , x ) ∥ ≲ ∣ a ∣ ≤ 2 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 + δ 0 , 1 ( s , y ) ∣ Z a F 1 ( s , y ) ∣. (118)
For t ≥ max { 10 R , 20 } t\geq\max\{10R,20\} t ≥ max { 10 R , 20 } , (113) can be achieved by (115) , (117) and (118) . The proof of (113) for t ≤ max { 10 R , 20 } t\leq\max\{10R,20\} t ≤ max { 10 R , 20 } is much easier. Therefore, the proof of (113) is completed. ▫
Lemma 5.2. Suppose that the obstacle O ⊂ R 2 \mathcal{O}\subset\mathbb{R}^2 O ⊂ R 2 is star-shaped, K = R 2 ∖ O \mathcal{K}=\mathbb{R}^2\setminus\mathcal{O} K = R 2 ∖ O and let w w w be the solution of the IBVP
{ □ w = ∂ t F 0 ( t , x ) + F 1 ( t , x ) , ( t , x ) ∈ [ 0 , ∞ ) × K , w = 0 , ( t , x ) ∈ [ 0 , ∞ ) × ∂ K , ( w , ∂ t w ) ( 0 , x ) = ( w 0 , w 1 ) ( x ) , x ∈ K , \begin{cases}
\Box w=\partial_tF_0(t,x)+F_1(t,x), & (t,x)\in[0,\infty)\times\mathcal{K},\\
w=0, & (t,x)\in[0,\infty)\times\partial\mathcal{K},\\
(w,\partial_t w)(0,x)=(w_0,w_1)(x), & x\in\mathcal{K},
\end{cases} ⎩ ⎨ ⎧ □ w = ∂ t F 0 ( t , x ) + F 1 ( t , x ) , w = 0 , ( w , ∂ t w ) ( 0 , x ) = ( w 0 , w 1 ) ( x ) , ( t , x ) ∈ [ 0 , ∞ ) × K , ( t , x ) ∈ [ 0 , ∞ ) × ∂ K , x ∈ K , where ( w 0 , w 1 ) (w_0,w_1) ( w 0 , w 1 ) has compact support and supp x ( F 0 , F 1 ) ( t , x ) ⊂ { x : ∣ x ∣ ≤ t + M 0 } \operatorname{supp}_x(F_0,F_1)(t,x)\subset\{x:|x|\leq t+M_0\} supp x ( F 0 , F 1 ) ( t , x ) ⊂ { x : ∣ x ∣ ≤ t + M 0 } . Then one has that for any μ , ν > 0 \mu,\nu>0 μ , ν > 0 and μ + ν < 1 / 2 \mu+\nu<1/2 μ + ν < 1/2 ,
⟨ t + ∣ x ∣ ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ μ ∣ w ( t , x ) ∣ ≲ ∑ ∣ a ∣ ≤ 4 sup s ∈ [ 0 , t ] ⟨ s ⟩ 1 / 2 + μ + ν ∥ ∂ a ( ∂ t F 0 , F 1 ) ( s ) ∥ L 2 ( K 3 ) \langle t+|x|\rangle^{1/2}\langle t-|x|\rangle^\mu|w(t,x)|
\lesssim\sum_{|a|\leq4}\sup_{s\in[0,t]}\langle s\rangle^{1/2+\mu+\nu}\|\partial^a(\partial_tF_0,F_1)(s)\|_{L^2(K_3)} ⟨ t + ∣ x ∣ ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ μ ∣ w ( t , x ) ∣ ≲ ∣ a ∣ ≤ 4 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ 1/2 + μ + ν ∥ ∂ a ( ∂ t F 0 , F 1 ) ( s ) ∥ L 2 ( K 3 ) + ∥ ( w 0 , w 1 ) ∥ H 5 ( K ) + ln ( 2 + t + ∣ x ∣ ) ∑ ∣ a ∣ ≤ 6 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 , 1 ( s , y ) ∣ Z a F 0 ( s , y ) ∣ +\|(w_0,w_1)\|_{H^5(\mathcal{K})}+\ln(2+t+|x|)\sum_{|a|\leq6}\sup_{\substack{(s,y)\in[0,t]\times\mathbb{R}^2\setminus K_2}}\langle y\rangle^{1/2}W_{1,1}(s,y)|Z^aF_0(s,y)| + ∥ ( w 0 , w 1 ) ∥ H 5 ( K ) + ln ( 2 + t + ∣ x ∣ ) ∣ a ∣ ≤ 6 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 , 1 ( s , y ) ∣ Z a F 0 ( s , y ) ∣ + ∑ ∣ a ∣ ≤ 5 sup y ∈ K ∣ ∂ a ( F 0 , F 1 ) ( 0 , y ) ∣ + ∑ ∣ a ∣ ≤ 5 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 + μ + ν , 1 ( s , y ) ∣ ∂ a F 1 ( s , y ) ∣ . (119) +\sum_{|a|\leq5}\sup_{y\in\mathcal{K}}|\partial^a(F_0,F_1)(0,y)|+\sum_{|a|\leq5}\sup_{\substack{(s,y)\in[0,t]\times\mathbb{R}^2\setminus K_2}}\langle y\rangle^{1/2}W_{1+\mu+\nu,1}(s,y)|\partial^aF_1(s,y)|. \tag*{(119)} + ∣ a ∣ ≤ 5 ∑ y ∈ K sup ∣ ∂ a ( F 0 , F 1 ) ( 0 , y ) ∣ + ∣ a ∣ ≤ 5 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 + μ + ν , 1 ( s , y ) ∣ ∂ a F 1 ( s , y ) ∣. (119) Furthermore, if F 0 ( t , x ) ≡ 0 F_0(t,x)\equiv0 F 0 ( t , x ) ≡ 0 , then it holds that
⟨ t − ∣ x ∣ ⟩ 1 / 2 − μ ∣ ∂ w ( t , x ) ∣ ≲ ∥ ( w 0 , w 1 ) ∥ H 4 ( K ) + ∑ ∣ a ∣ ≤ 3 sup s ∈ [ 0 , t ] ⟨ s ⟩ 1 / 2 ∥ ∂ a F 1 ( s ) ∥ L 2 ( K 4 ) \langle t-|x|\rangle^{1/2-\mu}|\partial w(t,x)|
\lesssim\|(w_0,w_1)\|_{H^4(\mathcal{K})}+\sum_{|a|\leq3}\sup_{s\in[0,t]}\langle s\rangle^{1/2}\|\partial^aF_1(s)\|_{L^2(K_4)} ⟨ t − ∣ x ∣ ⟩ 1/2 − μ ∣ ∂ w ( t , x ) ∣ ≲ ∥ ( w 0 , w 1 ) ∥ H 4 ( K ) + ∣ a ∣ ≤ 3 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ 1/2 ∥ ∂ a F 1 ( s ) ∥ L 2 ( K 4 ) + ∑ ∣ a ∣ ≤ 3 sup y ∈ K ∣ ∂ a F 1 ( 0 , y ) ∣ + ∑ ∣ a ∣ ≤ 4 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 + μ + ν , 1 ( s , y ) ∣ Z a F 1 ( s , y ) ∣ . (120) +\sum_{|a|\leq3}\sup_{y\in\mathcal{K}}|\partial^aF_1(0,y)|+\sum_{|a|\leq4}\sup_{\substack{(s,y)\in[0,t]\times\mathbb{R}^2\setminus K_2}}\langle y\rangle^{1/2}W_{1+\mu+\nu,1}(s,y)|Z^aF_1(s,y)|. \tag*{(120)} + ∣ a ∣ ≤ 3 ∑ y ∈ K sup ∣ ∂ a F 1 ( 0 , y ) ∣ + ∣ a ∣ ≤ 4 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 + μ + ν , 1 ( s , y ) ∣ Z a F 1 ( s , y ) ∣. (120)
Proof. Since the estimate (120) is just (5.1) of [17 ] , we omit the related proof. Next, we turn to prove (119) . Let w 1 b , w 2 b w_1^b,w_2^b w 1 b , w 2 b be the solutions to
{ □ w 1 b = ( 1 − χ [ 2 , 3 ] ( x ) ) ( ∂ t F 0 + F 1 ) , □ w 2 b = χ [ 2 , 3 ] ( x ) ( ∂ t F 0 + F 1 ) , w 1 b ∣ ∂ K = w 2 b ∣ ∂ K = 0 , ( w 1 b , ∂ t w 1 b ) ( 0 , x ) = ( w 0 , w 1 ) ( x ) , ( w 2 b , ∂ t w 2 b ) ( 0 , x ) = ( 0 , 0 ) , (121) \begin{cases}
\Box w_1^b=(1-\chi_{[2,3]}(x))(\partial_tF_0+F_1),\\
\Box w_2^b=\chi_{[2,3]}(x)(\partial_tF_0+F_1),\\
w_1^b|_{\partial\mathcal{K}}=w_2^b|_{\partial\mathcal{K}}=0,\\
(w_1^b,\partial_t w_1^b)(0,x)=(w_0,w_1)(x),\qquad(w_2^b,\partial_t w_2^b)(0,x)=(0,0), \tag*{(121)}
\end{cases} ⎩ ⎨ ⎧ □ w 1 b = ( 1 − χ [ 2 , 3 ] ( x )) ( ∂ t F 0 + F 1 ) , □ w 2 b = χ [ 2 , 3 ] ( x ) ( ∂ t F 0 + F 1 ) , w 1 b ∣ ∂ K = w 2 b ∣ ∂ K = 0 , ( w 1 b , ∂ t w 1 b ) ( 0 , x ) = ( w 0 , w 1 ) ( x ) , ( w 2 b , ∂ t w 2 b ) ( 0 , x ) = ( 0 , 0 ) , (121)
respectively. Thus w = w 1 b + w 2 b w = w_1^b + w_2^b w = w 1 b + w 2 b holds. By (32) with m = ∣ a ∣ + 2 m = |a| + 2 m = ∣ a ∣ + 2 , one has
⟨ t ⟩ η ∥ ∂ a w 1 b ( t , x ) ∥ L ∞ ( K R ) ≲ ⟨ t ⟩ η ∥ ∂ a w 1 b ( t , x ) ∥ H 2 ( K R + 1 ) ≲ ∥ ( w 0 , w 1 ) ∥ H ∣ a ∣ + 2 ( K ) + ∑ ∣ b ∣ ≤ ∣ a ∣ + 1 sup s ∈ [ 0 , t ] ⟨ s ⟩ η ∥ ∂ b ( ∂ t F 0 , F 1 ) ( s ) ∥ L 2 ( K 3 ) . (122) \langle t\rangle^\eta\|\partial^a w_1^b(t,x)\|_{L^\infty(\mathcal{K}_R)} \lesssim\langle t\rangle^\eta\|\partial^a w_1^b(t,x)\|_{H^2(\mathcal{K}_{R+1})}
\lesssim\|(w_0,w_1)\|_{H^{|a|+2}(\mathcal{K})}+\sum_{|b|\leq|a|+1}\sup_{s\in[0,t]}\langle s\rangle^\eta\|\partial^b(\partial_tF_0,F_1)(s)\|_{L^2(\mathcal{K}_3)}. \tag*{(122)} ⟨ t ⟩ η ∥ ∂ a w 1 b ( t , x ) ∥ L ∞ ( K R ) ≲ ⟨ t ⟩ η ∥ ∂ a w 1 b ( t , x ) ∥ H 2 ( K R + 1 ) ≲ ∥ ( w 0 , w 1 ) ∥ H ∣ a ∣ + 2 ( K ) + ∣ b ∣ ≤ ∣ a ∣ + 1 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ η ∥ ∂ b ( ∂ t F 0 , F 1 ) ( s ) ∥ L 2 ( K 3 ) . (122)
where η ∈ ( 0 , 1 ) \eta\in(0,1) η ∈ ( 0 , 1 ) .
We now focus on the estimate of w 1 b w_1^b w 1 b in the region ∣ x ∣ ≥ R |x|\geq R ∣ x ∣ ≥ R . Let w 1 c = χ [ 3 , 4 ] ( x ) w 1 b w_1^c=\chi_{[3,4]}(x)w_1^b w 1 c = χ [ 3 , 4 ] ( x ) w 1 b be the solution of the Cauchy problem
{ □ w 1 c = − [ Δ , χ [ 3 , 4 ] ( x ) ] w 1 b , ( w 1 c , ∂ t w 1 c ) ( 0 , x ) = ( χ [ 3 , 4 ] w 0 , χ [ 3 , 4 ] w 1 ) ( x ) . \begin{cases}
\Box w_1^c=-[\Delta,\chi_{[3,4]}(x)]w_1^b,\\
(w_1^c,\partial_t w_1^c)(0,x)=(\chi_{[3,4]}w_0,\chi_{[3,4]}w_1)(x).
\end{cases} { □ w 1 c = − [ Δ , χ [ 3 , 4 ] ( x )] w 1 b , ( w 1 c , ∂ t w 1 c ) ( 0 , x ) = ( χ [ 3 , 4 ] w 0 , χ [ 3 , 4 ] w 1 ) ( x ) .
Note that [ Δ , χ [ 3 , 4 ] ( x ) ] w 1 b [\Delta,\chi_{[3,4]}(x)]w_1^b [ Δ , χ [ 3 , 4 ] ( x )] w 1 b is supported in { x ∈ R 2 , ∣ x ∣ ≤ 4 } \{x\in\mathbb{R}^2, |x|\leq4\} { x ∈ R 2 , ∣ x ∣ ≤ 4 } . Then it follows from (40) and (44) that
⟨ t + ∣ x ∣ ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ μ ∣ w 1 c ( t , x ) ∣ ≲ ∥ ( w 0 , w 1 ) ∥ W 1 , ∞ ( K ) + ∑ ∣ a ∣ ≤ 2 sup s ∈ [ 0 , t ] ⟨ s ⟩ 1 / 2 + μ + ν ∥ ∂ a [ Δ , χ [ 3 , 4 ] ] w 1 b ( s ) ∥ L ∞ \langle t+|x|\rangle^{1/2}\langle t-|x|\rangle^\mu|w_1^c(t,x)|
\lesssim\|(w_0,w_1)\|_{W^{1,\infty}(\mathcal{K})}
+\sum_{|a|\leq2}\sup_{s\in[0,t]}\langle s\rangle^{1/2+\mu+\nu}\|\partial^a[\Delta,\chi_{[3,4]}]w_1^b(s)\|_{L^\infty} ⟨ t + ∣ x ∣ ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ μ ∣ w 1 c ( t , x ) ∣ ≲ ∥ ( w 0 , w 1 ) ∥ W 1 , ∞ ( K ) + ∣ a ∣ ≤ 2 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ 1/2 + μ + ν ∥ ∂ a [ Δ , χ [ 3 , 4 ] ] w 1 b ( s ) ∥ L ∞
≲ ∥ ( w 0 , w 1 ) ∥ H 3 ( K ) + ∑ ∣ a ∣ ≤ 3 sup s ∈ [ 0 , t ] ⟨ s ⟩ 1 / 2 + μ + ν ∥ ∂ a w 1 b ( s ) ∥ L ∞ ( K 4 ) \lesssim\|(w_0,w_1)\|_{H^3(\mathcal{K})}
+\sum_{|a|\leq3}\sup_{s\in[0,t]}\langle s\rangle^{1/2+\mu+\nu}\|\partial^a w_1^b(s)\|_{L^\infty(\mathcal{K}_4)} ≲ ∥ ( w 0 , w 1 ) ∥ H 3 ( K ) + ∣ a ∣ ≤ 3 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ 1/2 + μ + ν ∥ ∂ a w 1 b ( s ) ∥ L ∞ ( K 4 )
≲ ∥ ( w 0 , w 1 ) ∥ H 5 ( K ) + ∑ ∣ a ∣ ≤ 4 sup s ∈ [ 0 , t ] ⟨ s ⟩ 1 / 2 + μ + ν ∥ ∂ a ( ∂ t F 0 , F 1 ) ( s ) ∥ L 2 ( K 3 ) . (123) \lesssim\|(w_0,w_1)\|_{H^5(\mathcal{K})}
+\sum_{|a|\leq4}\sup_{s\in[0,t]}\langle s\rangle^{1/2+\mu+\nu}\|\partial^a(\partial_tF_0,F_1)(s)\|_{L^2(\mathcal{K}_3)}. \tag*{(123)} ≲ ∥ ( w 0 , w 1 ) ∥ H 5 ( K ) + ∣ a ∣ ≤ 4 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ 1/2 + μ + ν ∥ ∂ a ( ∂ t F 0 , F 1 ) ( s ) ∥ L 2 ( K 3 ) . (123)
where (122) has been used for η = 1 / 2 + μ + ν \eta=1/2+\mu+\nu η = 1/2 + μ + ν .
Collecting (122) and (123) with w 1 c = χ [ 3 , 4 ] ( x ) w 1 b w_1^c=\chi_{[3,4]}(x)w_1^b w 1 c = χ [ 3 , 4 ] ( x ) w 1 b and R > 4 R>4 R > 4 yields
⟨ t + ∣ x ∣ ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ μ ∣ w 1 b ( t , x ) ∣ ≲ ∥ ( w 0 , w 1 ) ∥ H 5 ( K ) + ∑ ∣ a ∣ ≤ 4 sup s ∈ [ 0 , t ] ⟨ s ⟩ 1 / 2 + μ + ν ∥ ∂ a ( ∂ t F 0 , F 1 ) ( s ) ∥ L 2 ( K 3 ) . (124) \langle t+|x|\rangle^{1/2}\langle t-|x|\rangle^\mu|w_1^b(t,x)|
\lesssim\|(w_0,w_1)\|_{H^5(\mathcal{K})}
+\sum_{|a|\leq4}\sup_{s\in[0,t]}\langle s\rangle^{1/2+\mu+\nu}\|\partial^a(\partial_tF_0,F_1)(s)\|_{L^2(\mathcal{K}_3)}. \tag*{(124)} ⟨ t + ∣ x ∣ ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ μ ∣ w 1 b ( t , x ) ∣ ≲ ∥ ( w 0 , w 1 ) ∥ H 5 ( K ) + ∣ a ∣ ≤ 4 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ 1/2 + μ + ν ∥ ∂ a ( ∂ t F 0 , F 1 ) ( s ) ∥ L 2 ( K 3 ) . (124)
Next, we deal with w 2 b w_2^b w 2 b . Let w 2 , 0 c w_{2,0}^c w 2 , 0 c and w 2 , 1 c w_{2,1}^c w 2 , 1 c be the Cauchy problems of
□ w 2 , 0 c = ∂ t ( χ [ 2 , 3 ] ( x ) F 0 ) , □ w 2 , 1 c = χ [ 2 , 3 ] ( x ) F 1 , ( w 2 , l c , ∂ t w 2 , l c ) ( 0 , x ) = ( 0 , 0 ) , l = 0 , 1 , \Box w_{2,0}^c=\partial_t(\chi_{[2,3]}(x)F_0),\qquad\Box w_{2,1}^c=\chi_{[2,3]}(x)F_1,\qquad(w_{2,l}^c,\partial_t w_{2,l}^c)(0,x)=(0,0),\quad l=0,1, □ w 2 , 0 c = ∂ t ( χ [ 2 , 3 ] ( x ) F 0 ) , □ w 2 , 1 c = χ [ 2 , 3 ] ( x ) F 1 , ( w 2 , l c , ∂ t w 2 , l c ) ( 0 , x ) = ( 0 , 0 ) , l = 0 , 1 ,
respectively. Then v 2 b = w 2 b − χ [ 1 , 2 ] ( x ) ( w 2 , 0 c + w 2 , 1 c ) v_2^b=w_2^b-\chi_{[1,2]}(x)(w_{2,0}^c+w_{2,1}^c) v 2 b = w 2 b − χ [ 1 , 2 ] ( x ) ( w 2 , 0 c + w 2 , 1 c ) is the solution of the IBVP
□ v 2 b = [ Δ , χ [ 1 , 2 ] ( x ) ] ( w 2 , 0 c + w 2 , 1 c ) , v 2 b ∣ ∂ K = 0 , ( v 2 b , ∂ t v 2 b ) ( 0 , x ) = ( 0 , 0 ) . \Box v_2^b=[\Delta,\chi_{[1,2]}(x)](w_{2,0}^c+w_{2,1}^c),\qquad v_2^b|_{\partial\mathcal{K}}=0,\qquad(v_2^b,\partial_t v_2^b)(0,x)=(0,0). □ v 2 b = [ Δ , χ [ 1 , 2 ] ( x )] ( w 2 , 0 c + w 2 , 1 c ) , v 2 b ∣ ∂ K = 0 , ( v 2 b , ∂ t v 2 b ) ( 0 , x ) = ( 0 , 0 ) .
Since the estimate of v 2 b v_2^b v 2 b is similar to that of w 1 b w_1^b w 1 b in (124) , we then have
⟨ t + ∣ x ∣ ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ μ ∣ v 2 b ( t , x ) ∣ ≲ ∑ ∣ a ∣ ≤ 4 sup s ∈ [ 0 , t ] ⟨ s ⟩ 1 / 2 + μ + ν / 2 ∥ ∂ a [ Δ , χ [ 1 , 2 ] ] ( w 2 , 0 c + w 2 , 1 c ) ( s ) ∥ L 2 \langle t+|x|\rangle^{1/2}\langle t-|x|\rangle^\mu|v_2^b(t,x)|
\lesssim\sum_{|a|\leq4}\sup_{s\in[0,t]}\langle s\rangle^{1/2+\mu+\nu/2}\|\partial^a[\Delta,\chi_{[1,2]}](w_{2,0}^c+w_{2,1}^c)(s)\|_{L^2} ⟨ t + ∣ x ∣ ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ μ ∣ v 2 b ( t , x ) ∣ ≲ ∣ a ∣ ≤ 4 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ 1/2 + μ + ν /2 ∥ ∂ a [ Δ , χ [ 1 , 2 ] ] ( w 2 , 0 c + w 2 , 1 c ) ( s ) ∥ L 2
≲ ∑ ∣ a ∣ ≤ 5 sup s ∈ [ 0 , t ] ⟨ s ⟩ 1 / 2 + μ + ν / 2 ∥ ∂ a ( w 2 , 0 c + w 2 , 1 c ) ( s , y ) ∥ L ∞ ( ∣ y ∣ ≤ 2 ) . (125) \lesssim\sum_{|a|\leq5}\sup_{s\in[0,t]}\langle s\rangle^{1/2+\mu+\nu/2}\|\partial^a(w_{2,0}^c+w_{2,1}^c)(s,y)\|_{L^\infty(|y|\leq2)}. \tag*{(125)} ≲ ∣ a ∣ ≤ 5 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ 1/2 + μ + ν /2 ∥ ∂ a ( w 2 , 0 c + w 2 , 1 c ) ( s , y ) ∥ L ∞ ( ∣ y ∣ ≤ 2 ) . (125)
Applying (36) with η = 1 / 2 − μ − ν \eta=1/2-\mu-\nu η = 1/2 − μ − ν and (40) to ∂ a w 2 , 0 c \partial^a w^c_{2,0} ∂ a w 2 , 0 c and ∂ a w 2 , 1 c \partial^a w^c_{2,1} ∂ a w 2 , 1 c , respectively, yields that for any ∣ a ∣ ≤ 5 |a|\leq5 ∣ a ∣ ≤ 5 ,
⟨ t + ∣ x ∣ ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ μ + ν ∣ ∂ a w 2 , 0 c ( t , x ) ∣ ≲ ∑ ∣ b ∣ ≤ 5 sup y ∈ K ∣ ∂ b F 0 ( 0 , y ) ∣ + ln ( 2 + t + ∣ x ∣ ) ∑ ∣ b ∣ ≤ 6 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 , 1 ( s , y ) ∣ Z b F 0 ( s , y ) ∣ , ⟨ t + ∣ x ∣ ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ μ + ν / 2 ∣ ∂ a w 2 , 1 c ( t , x ) ∣ ≲ ∑ ∣ b ∣ ≤ 4 sup y ∈ K ∣ ∂ b F 1 ( 0 , y ) ∣ + ∑ ∣ b ∣ ≤ 5 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 + μ + ν , 1 ( s , y ) ∣ ∂ b F 1 ( s , y ) ∣ . (126) \begin{aligned}
\langle t+|x|\rangle^{1/2}\langle t-|x|\rangle^{\mu+\nu}|\partial^a w^c_{2,0}(t,x)|
&\lesssim\sum_{|b|\leq5}\sup_{y\in\mathcal K}|\partial^b F_0(0,y)|+\ln(2+t+|x|)\sum_{|b|\leq6}\sup_{(s,y)\in[0,t]\times\mathbb R^2\setminus\mathcal K_2}\langle y\rangle^{1/2}\mathcal W_{1,1}(s,y)|Z^bF_0(s,y)|,\\
\langle t+|x|\rangle^{1/2}\langle t-|x|\rangle^{\mu+\nu/2}|\partial^a w^c_{2,1}(t,x)|
&\lesssim\sum_{|b|\leq4}\sup_{y\in\mathcal K}|\partial^b F_1(0,y)|+\sum_{|b|\leq5}\sup_{(s,y)\in[0,t]\times\mathbb R^2\setminus\mathcal K_2}\langle y\rangle^{1/2}\mathcal W_{1+\mu+\nu,1}(s,y)|\partial^bF_1(s,y)|. \tag*{(126)}
\end{aligned} ⟨ t + ∣ x ∣ ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ μ + ν ∣ ∂ a w 2 , 0 c ( t , x ) ∣ ⟨ t + ∣ x ∣ ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ μ + ν /2 ∣ ∂ a w 2 , 1 c ( t , x ) ∣ ≲ ∣ b ∣ ≤ 5 ∑ y ∈ K sup ∣ ∂ b F 0 ( 0 , y ) ∣ + ln ( 2 + t + ∣ x ∣ ) ∣ b ∣ ≤ 6 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 , 1 ( s , y ) ∣ Z b F 0 ( s , y ) ∣ , ≲ ∣ b ∣ ≤ 4 ∑ y ∈ K sup ∣ ∂ b F 1 ( 0 , y ) ∣ + ∣ b ∣ ≤ 5 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 + μ + ν , 1 ( s , y ) ∣ ∂ b F 1 ( s , y ) ∣. (126)
where sup y ∈ K ∣ ∂ b F 0 ( 0 , y ) ∣ \sup_{y\in\mathcal K}|\partial^bF_0(0,y)| sup y ∈ K ∣ ∂ b F 0 ( 0 , y ) ∣ and sup y ∈ K ∣ ∂ b F 1 ( 0 , y ) ∣ \sup_{y\in\mathcal K}|\partial^bF_1(0,y)| sup y ∈ K ∣ ∂ b F 1 ( 0 , y ) ∣ come from the initial data. Hence, (119) can be obtained by (124) –(126) with w = w 1 b + w 2 b w=w^b_1+w^b_2 w = w 1 b + w 2 b and w 2 b = v 2 b + χ [ 1 , 2 ] ( x ) ( w 2 , 0 c + w 2 , 1 c ) w^b_2=v^b_2+\chi_{[1,2]}(x)(w^c_{2,0}+w^c_{2,1}) w 2 b = v 2 b + χ [ 1 , 2 ] ( x ) ( w 2 , 0 c + w 2 , 1 c ) .
Lemma 5.3. Suppose that the obstacle O ⊂ R 2 \mathcal O\subset\mathbb R^2 O ⊂ R 2 is star-shaped, K = R 2 ∖ O \mathcal K=\mathbb R^2\setminus\mathcal O K = R 2 ∖ O and let w w w be the solution of the IBVP
{ □ w = ∂ t F 0 ( t , x ) + ∑ Ξ = 0 , 1 , 2 , θ ∂ Ξ G Ξ ( t , x ) + G 1 r ( t , x ) + G 2 r ( t , x ) , ( t , x ) ∈ [ 0 , ∞ ) × K , w = 0 , ( t , x ) ∈ [ 0 , ∞ ) × ∂ K , ( w , ∂ t w ) ( 0 , x ) = ( w 0 , w 1 ) ( x ) , x ∈ K . \begin{cases}
\square w=\partial_tF^0(t,x)+\sum_{\Xi=0,1,2,\theta}\partial_\Xi G^\Xi(t,x)+G^r_1(t,x)+G^r_2(t,x), &(t,x)\in[0,\infty)\times\mathcal K,\\
w=0, &(t,x)\in[0,\infty)\times\partial\mathcal K,\\
(w,\partial_tw)(0,x)=(w_0,w_1)(x), &x\in\mathcal K.
\end{cases} ⎩ ⎨ ⎧ □ w = ∂ t F 0 ( t , x ) + ∑ Ξ = 0 , 1 , 2 , θ ∂ Ξ G Ξ ( t , x ) + G 1 r ( t , x ) + G 2 r ( t , x ) , w = 0 , ( w , ∂ t w ) ( 0 , x ) = ( w 0 , w 1 ) ( x ) , ( t , x ) ∈ [ 0 , ∞ ) × K , ( t , x ) ∈ [ 0 , ∞ ) × ∂ K , x ∈ K . where ( w 0 , w 1 ) (w_0,w_1) ( w 0 , w 1 ) has compact support, supp x F 0 ( t , x ) ⊂ { x : 9 t 10 ≤ ∣ x ∣ ≤ t + M 0 } \operatorname{supp}_x F^0(t,x)\subset\{x:\frac{9t}{10}\leq|x|\leq t+M_0\} supp x F 0 ( t , x ) ⊂ { x : 10 9 t ≤ ∣ x ∣ ≤ t + M 0 } and G 1 r G^r_1 G 1 r , G 2 r G^r_2 G 2 r , G Ξ G^\Xi G Ξ (Ξ = 0 , 1 , 2 , θ \Xi=0,1,2,\theta Ξ = 0 , 1 , 2 , θ ) are supported in { x : ∣ x ∣ ≤ t + M 0 } \{x:|x|\leq t+M_0\} { x : ∣ x ∣ ≤ t + M 0 } . Then one has that for any η , μ , ν > 0 \eta,\mu,\nu>0 η , μ , ν > 0 , η < 1 / 2 \eta<1/2 η < 1/2 and μ + ν < 1 / 2 \mu+\nu<1/2 μ + ν < 1/2 ,
⟨ t ⟩ ∥ ∂ w ( t , x ) ∥ L ∞ ( ∣ x ∣ ≤ t / 10 ) + m # ⟨ x ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ 1 − η ∣ ∂ w ( t , x ) ∣ ≲ ∥ ( w 0 , w 1 ) ∥ H 9 ( K ) + ∑ ∣ a ∣ ≤ 8 ∑ Ξ = 0 , 1 , 2 , θ sup y ∈ K ∣ ∂ a ( ∂ ≤ 1 F 0 , ∂ ≤ 1 G Ξ , G 1 r , G 2 r ) ( 0 , y ) ∣ + ∑ ∣ a ∣ ≤ 8 ∑ Ξ = 0 , 1 , 2 , θ sup s ∈ [ 0 , t ] ⟨ s ⟩ ∥ ∂ a ( ∂ t F 0 , ∂ ≤ 1 G Ξ , G 1 r , G 2 r ) ( s ) ∥ L 2 ( K 4 ) + ∑ i = 1 , 2 sup s ∈ [ 0 , t ] ⟨ s ⟩ 3 / 2 + μ ∥ G i r ( s ) ∥ L ∞ ( K 3 ) + sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 3 / 2 + μ , 1 ( s , y ) ∣ G 1 r ( s , y ) ∣ + ln ( 2 + t + ∣ x ∣ ) sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 3 / 2 , 1 + ν ( s , y ) ∣ G 2 r ( s , y ) ∣ + ∑ ∣ a ∣ ≤ 9 ∑ Ξ = 0 , 1 , 2 , θ sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 + μ + ν , 1 ( s , y ) ∣ Z a ( ∂ ≤ 1 G Ξ , G 1 r , G 2 r ) ( s , y ) ∣ + ( 1 + m # ln ( 2 + t + ∣ x ∣ ) ) ∑ ∣ a ∣ ≤ 13 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 , 1 + μ ( s , y ) ∣ Z a F 0 ( s , y ) ∣ . (127) \begin{aligned}
\langle t\rangle\|\partial w(t,x)\|_{L^\infty(|x|\leq t/10)}
+m^\#\langle x\rangle^{1/2}\langle t-|x|\rangle^{1-\eta}|\partial w(t,x)|
\lesssim{}&\|(w_0,w_1)\|_{H^9(\mathcal K)}
+\sum_{|a|\leq8}\sum_{\Xi=0,1,2,\theta}\sup_{y\in\mathcal K}|\partial^a(\partial^{\leq1}F^0,\partial^{\leq1}G^\Xi,G^r_1,G^r_2)(0,y)|\\
&+\sum_{|a|\leq8}\sum_{\Xi=0,1,2,\theta}\sup_{s\in[0,t]}\langle s\rangle\|\partial^a(\partial_tF^0,\partial^{\leq1}G^\Xi,G^r_1,G^r_2)(s)\|_{L^2(\mathcal K_4)}\\
&+\sum_{i=1,2}\sup_{s\in[0,t]}\langle s\rangle^{3/2+\mu}\|G^r_i(s)\|_{L^\infty(\mathcal K_3)}
+\sup_{(s,y)\in[0,t]\times\mathbb R^2\setminus\mathcal K_2}\langle y\rangle^{1/2}\mathcal W_{3/2+\mu,1}(s,y)|G^r_1(s,y)|\\
&+\ln(2+t+|x|)\sup_{(s,y)\in[0,t]\times\mathbb R^2\setminus\mathcal K_2}\langle y\rangle^{1/2}\mathcal W_{3/2,1+\nu}(s,y)|G^r_2(s,y)|\\
&+\sum_{|a|\leq9}\sum_{\Xi=0,1,2,\theta}\sup_{(s,y)\in[0,t]\times\mathbb R^2\setminus\mathcal K_2}\langle y\rangle^{1/2}\mathcal W_{1+\mu+\nu,1}(s,y)|Z^a(\partial^{\leq1}G^\Xi,G^r_1,G^r_2)(s,y)|\\
&+(1+m^\#\ln(2+t+|x|))\sum_{|a|\leq13}\sup_{(s,y)\in[0,t]\times\mathbb R^2\setminus\mathcal K_2}\langle y\rangle^{1/2}\mathcal W_{1,1+\mu}(s,y)|Z^aF^0(s,y)|. \tag*{(127)}
\end{aligned} ⟨ t ⟩ ∥ ∂ w ( t , x ) ∥ L ∞ ( ∣ x ∣ ≤ t /10 ) + m # ⟨ x ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ 1 − η ∣ ∂ w ( t , x ) ∣ ≲ ∥ ( w 0 , w 1 ) ∥ H 9 ( K ) + ∣ a ∣ ≤ 8 ∑ Ξ = 0 , 1 , 2 , θ ∑ y ∈ K sup ∣ ∂ a ( ∂ ≤ 1 F 0 , ∂ ≤ 1 G Ξ , G 1 r , G 2 r ) ( 0 , y ) ∣ + ∣ a ∣ ≤ 8 ∑ Ξ = 0 , 1 , 2 , θ ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ ∥ ∂ a ( ∂ t F 0 , ∂ ≤ 1 G Ξ , G 1 r , G 2 r ) ( s ) ∥ L 2 ( K 4 ) + i = 1 , 2 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ 3/2 + μ ∥ G i r ( s ) ∥ L ∞ ( K 3 ) + ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 3/2 + μ , 1 ( s , y ) ∣ G 1 r ( s , y ) ∣ + ln ( 2 + t + ∣ x ∣ ) ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 3/2 , 1 + ν ( s , y ) ∣ G 2 r ( s , y ) ∣ + ∣ a ∣ ≤ 9 ∑ Ξ = 0 , 1 , 2 , θ ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 + μ + ν , 1 ( s , y ) ∣ Z a ( ∂ ≤ 1 G Ξ , G 1 r , G 2 r ) ( s , y ) ∣ + ( 1 + m # ln ( 2 + t + ∣ x ∣ )) ∣ a ∣ ≤ 13 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 , 1 + μ ( s , y ) ∣ Z a F 0 ( s , y ) ∣. (127) where m # = 0 , 1 m^\#=0,1 m # = 0 , 1 .
Remark 5.2. Note that when m # = 0 m^\#=0 m # = 0 , there is no ln ( 2 + t + ∣ x ∣ ) \ln(2+t+|x|) ln ( 2 + t + ∣ x ∣ ) factor in the last line of (127) for the estimate of ∥ ∂ w ( t , x ) ∥ L ∞ ( ∣ x ∣ ≤ t / 10 ) \|\partial w(t,x)\|_{L^\infty(|x|\leq t/10)} ∥ ∂ w ( t , x ) ∥ L ∞ ( ∣ x ∣ ≤ t /10 ) .
Proof. First, recall (5.4) of [17 ]
⟨ x ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ ∣ ∂ w d i v ( t , x ) ∣ ≲ ∥ ( w 0 , w 1 ) ∥ H 9 ( K ) + ∑ ∣ a ∣ ≤ 8 ∑ Ξ = 0 , 1 , 2 , θ sup y ∈ K ∣ ∂ a ( ∂ ≤ 1 G 1 Ξ , G 1 r ) ( 0 , y ) ∣ + ∑ ∣ a ∣ ≤ 8 ∑ Ξ = 0 , 1 , 2 , θ sup s ∈ [ 0 , t ] ⟨ s ⟩ ∥ ∂ a ( ∂ ≤ 1 G 1 Ξ , G 1 r ) ( s ) ∥ L 2 ( K 4 ) + ∑ i = 1 , 2 sup s ∈ [ 0 , t ] ⟨ s ⟩ 3 / 2 + μ ∥ G i r ( s ) ∥ L ∞ ( K 3 ) + sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 3 / 2 + μ , 1 ( s , y ) ∣ G 1 r ( s , y ) ∣ + ∑ ∣ a ∣ ≤ 9 ∑ Ξ = 0 , 1 , 2 , θ sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 + μ + ν , 1 ( s , y ) ∣ Z a ( G 1 r , Z ≤ 1 G Ξ ) ( s , y ) ∣ . (128) \begin{aligned}
\langle x\rangle^{1/2}\langle t-|x|\rangle|\partial w^{div}(t,x)| \lesssim{}& \|(w_0,w_1)\|_{H^9(\mathcal{K})}+\sum_{|a|\leq8}\sum_{\Xi=0,1,2,\theta}\sup_{y\in\mathcal{K}}|\partial^a(\partial^{\leq1}G_1^\Xi,G_1^r)(0,y)| \\
&+\sum_{|a|\leq8}\sum_{\Xi=0,1,2,\theta}\sup_{s\in[0,t]}\langle s\rangle\|\partial^a(\partial^{\leq1}G_1^\Xi,G_1^r)(s)\|_{L^2(\mathcal{K}_4)}+\sum_{i=1,2}\sup_{s\in[0,t]}\langle s\rangle^{3/2+\mu}\|G_i^r(s)\|_{L^\infty(\mathcal{K}_3)} \\
&+\sup_{(s,y)\in[0,t]\times\mathbb{R}^2\setminus\mathcal{K}_2}\langle y\rangle^{1/2}W_{3/2+\mu,1}(s,y)|G_1^r(s,y)| \\
&+\sum_{|a|\leq9}\sum_{\Xi=0,1,2,\theta}\sup_{(s,y)\in[0,t]\times\mathbb{R}^2\setminus\mathcal{K}_2}\langle y\rangle^{1/2}W_{1+\mu+\nu,1}(s,y)|Z^a(G_1^r,Z^{\leq1}G^\Xi)(s,y)|. \tag*{(128)}
\end{aligned} ⟨ x ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ ∣ ∂ w d i v ( t , x ) ∣ ≲ ∥ ( w 0 , w 1 ) ∥ H 9 ( K ) + ∣ a ∣ ≤ 8 ∑ Ξ = 0 , 1 , 2 , θ ∑ y ∈ K sup ∣ ∂ a ( ∂ ≤ 1 G 1 Ξ , G 1 r ) ( 0 , y ) ∣ + ∣ a ∣ ≤ 8 ∑ Ξ = 0 , 1 , 2 , θ ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ ∥ ∂ a ( ∂ ≤ 1 G 1 Ξ , G 1 r ) ( s ) ∥ L 2 ( K 4 ) + i = 1 , 2 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ 3/2 + μ ∥ G i r ( s ) ∥ L ∞ ( K 3 ) + ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 3/2 + μ , 1 ( s , y ) ∣ G 1 r ( s , y ) ∣ + ∣ a ∣ ≤ 9 ∑ Ξ = 0 , 1 , 2 , θ ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 + μ + ν , 1 ( s , y ) ∣ Z a ( G 1 r , Z ≤ 1 G Ξ ) ( s , y ) ∣. (128)
provided that w d i v w^{div} w d i v is the solution of
{ □ w d i v = ∑ Ξ = 0 , 1 , 2 , θ ∂ Ξ G Ξ ( t , x ) + G 1 r ( t , x ) , ( t , x ) ∈ [ 0 , ∞ ) × K , w d i v = 0 , ( t , x ) ∈ [ 0 , ∞ ) × ∂ K , ( w d i v , ∂ t w d i v ) ( 0 , x ) = ( w 0 , w 1 ) ( x ) , x ∈ K . \left\{
\begin{aligned}
\Box w^{div}&=\sum_{\Xi=0,1,2,\theta}\partial_\Xi G^\Xi(t,x)+G_1^r(t,x), && (t,x)\in[0,\infty)\times\mathcal{K},\\
w^{div}&=0, && (t,x)\in[0,\infty)\times\partial\mathcal{K},\\
(w^{div},\partial_t w^{div})(0,x)&=(w_0,w_1)(x), && x\in\mathcal{K}.
\end{aligned}
\right. ⎩ ⎨ ⎧ □ w d i v w d i v ( w d i v , ∂ t w d i v ) ( 0 , x ) = Ξ = 0 , 1 , 2 , θ ∑ ∂ Ξ G Ξ ( t , x ) + G 1 r ( t , x ) , = 0 , = ( w 0 , w 1 ) ( x ) , ( t , x ) ∈ [ 0 , ∞ ) × K , ( t , x ) ∈ [ 0 , ∞ ) × ∂ K , x ∈ K .
Thus, it suffices to treat the estimates of w ~ = w − w d i v \widetilde w=w-w^{div} w = w − w d i v , where w ~ \widetilde w w is the solution of
□ w ~ = ∂ t F 0 ( t , x ) + G 2 r ( t , x ) , w ~ ∣ ∂ K = 0 , ( w ~ , ∂ t w ~ ) ( 0 , x ) = ( 0 , 0 ) . \Box\widetilde w=\partial_tF^0(t,x)+G_2^r(t,x),\qquad\widetilde w|_{\partial\mathcal{K}}=0,\qquad(\widetilde w,\partial_t\widetilde w)(0,x)=(0,0). □ w = ∂ t F 0 ( t , x ) + G 2 r ( t , x ) , w ∣ ∂ K = 0 , ( w , ∂ t w ) ( 0 , x ) = ( 0 , 0 ) .
Let w 1 b w_1^b w 1 b and w 2 b w_2^b w 2 b be the solutions of
{ □ w 1 b = ( 1 − χ [ 2 , 3 ] ( x ) ) ( ∂ t F 0 + G 2 r ) , □ w 2 b = χ [ 2 , 3 ] ( x ) ( ∂ t F 0 + G 2 r ) , w 1 b ∣ ∂ K = w 2 b ∣ ∂ K = 0 , ( w 1 b , ∂ t w 1 b ) ( 0 , x ) = ( w 2 b , ∂ t w 2 b ) ( 0 , x ) = ( 0 , 0 ) , (129) \left\{
\begin{aligned}
\Box w_1^b&=(1-\chi_{[2,3]}(x))(\partial_tF^0+G_2^r),\\
\Box w_2^b&=\chi_{[2,3]}(x)(\partial_tF^0+G_2^r),\\
w_1^b|_{\partial\mathcal{K}}&=w_2^b|_{\partial\mathcal{K}}=0,\\
(w_1^b,\partial_t w_1^b)(0,x)&=(w_2^b,\partial_t w_2^b)(0,x)=(0,0),
\end{aligned}
\right. \tag*{(129)} ⎩ ⎨ ⎧ □ w 1 b □ w 2 b w 1 b ∣ ∂ K ( w 1 b , ∂ t w 1 b ) ( 0 , x ) = ( 1 − χ [ 2 , 3 ] ( x )) ( ∂ t F 0 + G 2 r ) , = χ [ 2 , 3 ] ( x ) ( ∂ t F 0 + G 2 r ) , = w 2 b ∣ ∂ K = 0 , = ( w 2 b , ∂ t w 2 b ) ( 0 , x ) = ( 0 , 0 ) , (129)
respectively. By (32) with m = ∣ a ∣ + 2 m=|a|+2 m = ∣ a ∣ + 2 , one has
⟨ t ⟩ ∥ ∂ a w 1 b ( t , x ) ∥ L ∞ ( K R ) ≲ ⟨ t ⟩ ∥ ∂ a w 1 b ( t , x ) ∥ H 2 ( K R + 1 ) ≲ ∑ ∣ b ∣ ≤ ∣ a ∣ + 1 sup s ∈ [ 0 , t ] ⟨ s ⟩ ∥ ∂ b ( ∂ t F 0 , G 2 r ) ( s ) ∥ L 2 ( K 3 ) . (130) \langle t\rangle\|\partial^a w_1^b(t,x)\|_{L^\infty(\mathcal{K}_R)}\lesssim\langle t\rangle\|\partial^a w_1^b(t,x)\|_{H^2(\mathcal{K}_{R+1})}
\lesssim\sum_{|b|\leq|a|+1}\sup_{s\in[0,t]}\langle s\rangle\|\partial^b(\partial_tF^0,G_2^r)(s)\|_{L^2(\mathcal{K}_3)}. \tag*{(130)} ⟨ t ⟩ ∥ ∂ a w 1 b ( t , x ) ∥ L ∞ ( K R ) ≲ ⟨ t ⟩ ∥ ∂ a w 1 b ( t , x ) ∥ H 2 ( K R + 1 ) ≲ ∣ b ∣ ≤ ∣ a ∣ + 1 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ ∥ ∂ b ( ∂ t F 0 , G 2 r ) ( s ) ∥ L 2 ( K 3 ) . (130)
Next, we turn to the estimate of w 1 b w_1^b w 1 b in the region ∣ x ∣ ≥ R |x|\geq R ∣ x ∣ ≥ R . It is easy to find that w 1 c = χ [ 3 , 4 ] ( x ) w 1 b w_1^c=\chi_{[3,4]}(x)w_1^b w 1 c = χ [ 3 , 4 ] ( x ) w 1 b is the solution of the Cauchy problem
□ w 1 c = − [ Δ , χ [ 3 , 4 ] ( x ) ] w 1 b , ( w 1 c , ∂ t w 1 c ) ( 0 , x ) = ( 0 , 0 ) . \Box w_1^c=-[\Delta,\chi_{[3,4]}(x)]w_1^b,\qquad(w_1^c,\partial_t w_1^c)(0,x)=(0,0). □ w 1 c = − [ Δ , χ [ 3 , 4 ] ( x )] w 1 b , ( w 1 c , ∂ t w 1 c ) ( 0 , x ) = ( 0 , 0 ) .
By applying (45) to □ w 1 c \Box w_1^c □ w 1 c together with (5.19), one obtains
⟨ x ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ ∣ ∂ w 1 c ( t , x ) ∣ ≲ ∑ ∣ b ∣ ≤ 6 sup s ∈ [ 0 , t ] ⟨ s ⟩ ∥ ∂ b [ Δ , χ [ 3 , 4 ] ] w 1 b ( s ) ∥ L ∞ ≲ ∑ ∣ b ∣ ≤ 7 sup s ∈ [ 0 , t ] ⟨ s ⟩ ∥ ∂ b w 1 b ( s ) ∥ L ∞ ( K 4 ) ≲ ∑ ∣ b ∣ ≤ 8 sup s ∈ [ 0 , t ] ⟨ s ⟩ ∥ ∂ b ( ∂ t F 0 , G 2 r ) ( s ) ∥ L 2 ( K 3 ) . (131) \begin{aligned}
\langle x\rangle^{1/2}\langle t-|x|\rangle|\partial w_1^c(t,x)|\lesssim{}&
\sum_{|b|\leq6}\sup_{s\in[0,t]}\langle s\rangle\|\partial^b[\Delta,\chi_{[3,4]}]w_1^b(s)\|_{L^\infty}\\
&\lesssim\sum_{|b|\leq7}\sup_{s\in[0,t]}\langle s\rangle\|\partial^b w_1^b(s)\|_{L^\infty(\mathcal{K}_4)}\\
&\lesssim\sum_{|b|\leq8}\sup_{s\in[0,t]}\langle s\rangle\|\partial^b(\partial_tF^0,G_2^r)(s)\|_{L^2(\mathcal{K}_3)}. \tag*{(131)}
\end{aligned} ⟨ x ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ ∣ ∂ w 1 c ( t , x ) ∣ ≲ ∣ b ∣ ≤ 6 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ ∥ ∂ b [ Δ , χ [ 3 , 4 ] ] w 1 b ( s ) ∥ L ∞ ≲ ∣ b ∣ ≤ 7 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ ∥ ∂ b w 1 b ( s ) ∥ L ∞ ( K 4 ) ≲ ∣ b ∣ ≤ 8 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ ∥ ∂ b ( ∂ t F 0 , G 2 r ) ( s ) ∥ L 2 ( K 3 ) . (131)
Note that ∂ w 1 b = ∂ w 1 c \partial w_1^b=\partial w_1^c ∂ w 1 b = ∂ w 1 c holds in the region ∣ x ∣ ≥ 4 |x|\ge4 ∣ x ∣ ≥ 4 . Then collecting (130) -(131) with R > 4 R>4 R > 4 leads to
⟨ x ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ ∣ ∂ w 1 b ( t , x ) ∣ ≲ ∑ ∣ b ∣ ≤ 8 sup s ∈ [ 0 , t ] ⟨ s ⟩ ∥ ∂ b ( ∂ t F 0 , G 2 r ) ( s ) ∥ L 2 ( K 3 ) . (132) \langle x\rangle^{1/2}\langle t-|x|\rangle|\partial w_1^b(t,x)|\lesssim\sum_{|b|\le8}\sup_{s\in[0,t]}\langle s\rangle\|\partial^b(\partial_tF^0,G_2^r)(s)\|_{L^2(\mathcal{K}_3)}. \tag*{(132)} ⟨ x ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ ∣ ∂ w 1 b ( t , x ) ∣ ≲ ∣ b ∣ ≤ 8 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ ∥ ∂ b ( ∂ t F 0 , G 2 r ) ( s ) ∥ L 2 ( K 3 ) . (132)
We next deal with ∂ t w 2 b \partial_t w_2^b ∂ t w 2 b and ∂ i w 2 b \partial_i w_2^b ∂ i w 2 b (i = 1 , 2 i=1,2 i = 1 , 2 ), separately. Let w 2 c w_2^c w 2 c and w r c w_r^c w r c be the solutions of the Cauchy problems
□ w 2 c = χ [ 2 , 3 ] ( x ) F 0 ( t , x ) , □ w r c = χ [ 2 , 3 ] ( x ) G 2 r ( t , x ) (133) \square w_2^c=\chi_{[2,3]}(x)F^0(t,x),\qquad\square w_r^c=\chi_{[2,3]}(x)G_2^r(t,x) \tag*{(133)} □ w 2 c = χ [ 2 , 3 ] ( x ) F 0 ( t , x ) , □ w r c = χ [ 2 , 3 ] ( x ) G 2 r ( t , x ) (133)
with ( w 2 c , ∂ t w 2 c ) ( 0 , x ) = ( 0 , 0 ) (w_2^c,\partial_t w_2^c)(0,x)=(0,0) ( w 2 c , ∂ t w 2 c ) ( 0 , x ) = ( 0 , 0 ) and ( w r c , ∂ t w r c ) ( 0 , x ) = ( 0 , 0 ) (w_r^c,\partial_t w_r^c)(0,x)=(0,0) ( w r c , ∂ t w r c ) ( 0 , x ) = ( 0 , 0 ) , respectively. Then, by (129) and (133) , v 2 b = w 2 b − χ [ 1 , 2 ] ( x ) ( ∂ t w 2 c + w r c ) v_2^b=w_2^b-\chi_{[1,2]}(x)(\partial_t w_2^c+w_r^c) v 2 b = w 2 b − χ [ 1 , 2 ] ( x ) ( ∂ t w 2 c + w r c ) is the solution of the IBVP
□ v 2 b = [ Δ , χ [ 1 , 2 ] ( x ) ] ( ∂ t w 2 c + w r c ) , v 2 b ∣ ∂ K = 0 , ( v 2 b , ∂ t v 2 b ) ( 0 , x ) = ( 0 , − χ [ 2 , 3 ] ( x ) F 0 ( 0 , x ) ) . \square v_2^b=[\Delta,\chi_{[1,2]}(x)](\partial_t w_2^c+w_r^c),\qquad v_2^b|_{\partial\mathcal{K}}=0,\qquad(v_2^b,\partial_t v_2^b)(0,x)=(0,-\chi_{[2,3]}(x)F^0(0,x)). □ v 2 b = [ Δ , χ [ 1 , 2 ] ( x )] ( ∂ t w 2 c + w r c ) , v 2 b ∣ ∂ K = 0 , ( v 2 b , ∂ t v 2 b ) ( 0 , x ) = ( 0 , − χ [ 2 , 3 ] ( x ) F 0 ( 0 , x )) .
Since the estimate of v 2 b v_2^b v 2 b is analogous to that of w 1 b w_1^b w 1 b in (132) , then it holds that
⟨ x ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ ∣ ∂ v 2 b ( t , x ) ∣ ≲ ∥ F 0 ( 0 , y ) ∥ H 8 ( K ) + ∑ ∣ b ∣ ≤ 9 sup s ∈ [ 0 , t ] ⟨ s ⟩ ∥ ∂ b ( ∂ t w 2 c , w r c ) ( s , y ) ∥ L ∞ ( ∣ y ∣ ≤ 2 ) . (134) \langle x\rangle^{1/2}\langle t-|x|\rangle|\partial v_2^b(t,x)|\lesssim\|F^0(0,y)\|_{H^8(\mathcal{K})}+\sum_{|b|\le9}\sup_{s\in[0,t]}\langle s\rangle\|\partial^b(\partial_t w_2^c,w_r^c)(s,y)\|_{L^\infty(|y|\le2)}. \tag*{(134)} ⟨ x ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ ∣ ∂ v 2 b ( t , x ) ∣ ≲ ∥ F 0 ( 0 , y ) ∥ H 8 ( K ) + ∣ b ∣ ≤ 9 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ ∥ ∂ b ( ∂ t w 2 c , w r c ) ( s , y ) ∥ L ∞ ( ∣ y ∣ ≤ 2 ) . (134)
For the estimates of w 2 c w_2^c w 2 c and w r c w_r^c w r c , applying Lemma 4.2, (2.7), (2.12), (2.13) to (133) , respectively, yields
∑ ∣ b ∣ ≤ 9 ⟨ t ⟩ ∣ ∂ b ∂ t w 2 c ( t , x ) ∣ ≲ ∑ ∣ a ∣ ≤ 13 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 , 1 + μ ( s , y ) ∣ Z a F 0 ( s , y ) ∣ + ∑ ∣ a ∣ ≤ 8 sup y ∈ K ∣ ∂ a F 0 ( 0 , y ) ∣ , ∣ x ∣ ≤ t / 10 or ∣ x ∣ ≤ 2 , \sum_{|b|\le9}\langle t\rangle|\partial^b\partial_t w_2^c(t,x)|\lesssim\sum_{|a|\le13}\sup_{(s,y)\in[0,t]\times\mathbb{R}^2\setminus\mathcal{K}_2}\langle y\rangle^{1/2}W_{1,1+\mu}(s,y)|Z^aF^0(s,y)|
+\sum_{|a|\le8}\sup_{y\in\mathcal{K}}|\partial^aF^0(0,y)|,\qquad|x|\le t/10\text{ or }|x|\le2, ∣ b ∣ ≤ 9 ∑ ⟨ t ⟩ ∣ ∂ b ∂ t w 2 c ( t , x ) ∣ ≲ ∣ a ∣ ≤ 13 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 , 1 + μ ( s , y ) ∣ Z a F 0 ( s , y ) ∣ + ∣ a ∣ ≤ 8 ∑ y ∈ K sup ∣ ∂ a F 0 ( 0 , y ) ∣ , ∣ x ∣ ≤ t /10 or ∣ x ∣ ≤ 2 ,
⟨ x ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ 1 − η ∣ ∂ t 2 w 2 c ( t , x ) ∣ ≲ ln ( 2 + t + ∣ x ∣ ) ∑ ∣ a ∣ ≤ 2 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 , 1 ( s , y ) ∣ Z a F 0 ( s , y ) ∣ , \langle x\rangle^{1/2}\langle t-|x|\rangle^{1-\eta}|\partial_t^2w_2^c(t,x)|\lesssim\ln(2+t+|x|)\sum_{|a|\le2}\sup_{(s,y)\in[0,t]\times\mathbb{R}^2\setminus\mathcal{K}_2}\langle y\rangle^{1/2}W_{1,1}(s,y)|Z^aF^0(s,y)|, ⟨ x ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ 1 − η ∣ ∂ t 2 w 2 c ( t , x ) ∣ ≲ ln ( 2 + t + ∣ x ∣ ) ∣ a ∣ ≤ 2 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 , 1 ( s , y ) ∣ Z a F 0 ( s , y ) ∣ ,
⟨ t + ∣ x ∣ ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ 1 / 2 ∣ w r c ( t , x ) ∣ ≲ ln ( 2 + t + ∣ x ∣ ) sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 3 / 2 , 1 + ν ( s , y ) ∣ G 2 r ( s , y ) ∣ , \langle t+|x|\rangle^{1/2}\langle t-|x|\rangle^{1/2}|w_r^c(t,x)|\lesssim\ln(2+t+|x|)\sup_{(s,y)\in[0,t]\times\mathbb{R}^2\setminus\mathcal{K}_2}\langle y\rangle^{1/2}W_{3/2,1+\nu}(s,y)|G_2^r(s,y)|, ⟨ t + ∣ x ∣ ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ 1/2 ∣ w r c ( t , x ) ∣ ≲ ln ( 2 + t + ∣ x ∣ ) ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 3/2 , 1 + ν ( s , y ) ∣ G 2 r ( s , y ) ∣ ,
∑ ∣ b ∣ ≤ 8 ⟨ x ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ ∣ ∂ b ∂ t w r c ( t , x ) ∣ ≲ ∑ ∣ a ∣ ≤ 9 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 + μ , 1 ( s , y ) ∣ Z a G 2 r ( s , y ) ∣ + ∑ ∣ a ∣ ≤ 7 sup y ∈ K ∣ ∂ a G 2 r ( 0 , y ) ∣ . (135) \sum_{|b|\le8}\langle x\rangle^{1/2}\langle t-|x|\rangle|\partial^b\partial_t w_r^c(t,x)|\lesssim\sum_{|a|\le9}\sup_{(s,y)\in[0,t]\times\mathbb{R}^2\setminus\mathcal{K}_2}\langle y\rangle^{1/2}W_{1+\mu,1}(s,y)|Z^aG_2^r(s,y)|
+\sum_{|a|\le7}\sup_{y\in\mathcal{K}}|\partial^aG_2^r(0,y)|. \tag*{(135)} ∣ b ∣ ≤ 8 ∑ ⟨ x ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ ∣ ∂ b ∂ t w r c ( t , x ) ∣ ≲ ∣ a ∣ ≤ 9 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 + μ , 1 ( s , y ) ∣ Z a G 2 r ( s , y ) ∣ + ∣ a ∣ ≤ 7 ∑ y ∈ K sup ∣ ∂ a G 2 r ( 0 , y ) ∣. (135)
Collecting (134) , (135) with ∂ t w 2 b = ∂ t v 2 b + χ [ 1 , 2 ] ( x ) ( ∂ t 2 w 2 c + ∂ t w r c ) \partial_t w_2^b=\partial_t v_2^b+\chi_{[1,2]}(x)(\partial_t^2w_2^c+\partial_t w_r^c) ∂ t w 2 b = ∂ t v 2 b + χ [ 1 , 2 ] ( x ) ( ∂ t 2 w 2 c + ∂ t w r c ) derives
⟨ x ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ 1 − η ∣ ∂ t w 2 b ( t , x ) ∣ ≲ ∑ ∣ a ∣ ≤ 8 sup y ∈ K ∣ ∂ a ( F 0 , G 2 r ) ( 0 , y ) ∣ + ln ( 2 + t + ∣ x ∣ ) ∑ ∣ a ∣ ≤ 13 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 , 1 + μ ( s , y ) ∣ Z a F 0 ( s , y ) ∣ + ln ( 2 + t + ∣ x ∣ ) sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 3 / 2 , 1 + ν ( s , y ) ∣ G 2 r ( s , y ) ∣ + ∑ ∣ a ∣ ≤ 9 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 + μ , 1 ( s , y ) ∣ Z a G 2 r ( s , y ) ∣ (136) \begin{aligned}
\langle x\rangle^{1/2}\langle t-|x|\rangle^{1-\eta}|\partial_t w_2^b(t,x)|
\lesssim{}&\sum_{|a|\le8}\sup_{y\in\mathcal{K}}|\partial^a(F^0,G_2^r)(0,y)|\\
&+\ln(2+t+|x|)\sum_{|a|\le13}\sup_{(s,y)\in[0,t]\times\mathbb{R}^2\setminus\mathcal{K}_2}\langle y\rangle^{1/2}W_{1,1+\mu}(s,y)|Z^aF^0(s,y)|\\
&+\ln(2+t+|x|)\sup_{(s,y)\in[0,t]\times\mathbb{R}^2\setminus\mathcal{K}_2}\langle y\rangle^{1/2}W_{3/2,1+\nu}(s,y)|G_2^r(s,y)|\\
&+\sum_{|a|\le9}\sup_{(s,y)\in[0,t]\times\mathbb{R}^2\setminus\mathcal{K}_2}\langle y\rangle^{1/2}W_{1+\mu,1}(s,y)|Z^aG_2^r(s,y)| \tag*{(136)}
\end{aligned} ⟨ x ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ 1 − η ∣ ∂ t w 2 b ( t , x ) ∣ ≲ ∣ a ∣ ≤ 8 ∑ y ∈ K sup ∣ ∂ a ( F 0 , G 2 r ) ( 0 , y ) ∣ + ln ( 2 + t + ∣ x ∣ ) ∣ a ∣ ≤ 13 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 , 1 + μ ( s , y ) ∣ Z a F 0 ( s , y ) ∣ + ln ( 2 + t + ∣ x ∣ ) ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 3/2 , 1 + ν ( s , y ) ∣ G 2 r ( s , y ) ∣ + ∣ a ∣ ≤ 9 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 + μ , 1 ( s , y ) ∣ Z a G 2 r ( s , y ) ∣ (136)
and for ∣ x ∣ ≤ t / 10 |x| \le t/10 ∣ x ∣ ≤ t /10
⟨ t ⟩ ∣ ∂ t w 2 b ( t , x ) ∣ ≲ ∑ ∣ a ∣ ≤ 8 sup y ∈ K ∣ ∂ a ( F 0 , G 2 r ) ( 0 , y ) ∣ + ∑ ∣ a ∣ ≤ 13 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 , 1 + μ ( s , y ) ∣ Z a F 0 ( s , y ) ∣ + ln ( 2 + t + ∣ x ∣ ) sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 3 / 2 , 1 + ν ( s , y ) ∣ G 2 r ( s , y ) ∣ + ∑ ∣ a ∣ ≤ 9 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 + μ , 1 ( s , y ) ∣ Z a G 2 r ( s , y ) ∣ . (137) \langle t\rangle|\partial_t w_2^b(t,x)| \lesssim\sum_{|a|\le8}\sup_{y\in\mathcal{K}}|\partial^a(F^0,G_2^r)(0,y)|
+ \sum_{|a|\le13}\sup_{(s,y)\in[0,t]\times\mathbb{R}^2\setminus\mathcal{K}_2}\langle y\rangle^{1/2}\mathcal{W}_{1,1+\mu}(s,y)|Z^aF^0(s,y)|
+ \ln(2+t+|x|)\sup_{(s,y)\in[0,t]\times\mathbb{R}^2\setminus\mathcal{K}_2}\langle y\rangle^{1/2}\mathcal{W}_{3/2,1+\nu}(s,y)|G_2^r(s,y)|
+ \sum_{|a|\le9}\sup_{(s,y)\in[0,t]\times\mathbb{R}^2\setminus\mathcal{K}_2}\langle y\rangle^{1/2}\mathcal{W}_{1+\mu,1}(s,y)|Z^aG_2^r(s,y)|. \tag*{(137)} ⟨ t ⟩ ∣ ∂ t w 2 b ( t , x ) ∣ ≲ ∣ a ∣ ≤ 8 ∑ y ∈ K sup ∣ ∂ a ( F 0 , G 2 r ) ( 0 , y ) ∣ + ∣ a ∣ ≤ 13 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 , 1 + μ ( s , y ) ∣ Z a F 0 ( s , y ) ∣ + ln ( 2 + t + ∣ x ∣ ) ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 3/2 , 1 + ν ( s , y ) ∣ G 2 r ( s , y ) ∣ + ∣ a ∣ ≤ 9 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 + μ , 1 ( s , y ) ∣ Z a G 2 r ( s , y ) ∣. (137)
Next, we turn to the estimate of ∂ x w 2 b \partial_xw_2^b ∂ x w 2 b . Let w ~ 2 c \widetilde{w}_2^c w 2 c be the Cauchy problem of
□ w ~ 2 c = ∂ t ( χ [ 2 , 3 ] ( x ) F 0 ( t , x ) ) , ( w ~ 2 c , ∂ t w ~ 2 c ) ( 0 , x ) = ( 0 , 0 ) . \Box\widetilde{w}_2^c=\partial_t\left(\chi_{[2,3]}(x)F^0(t,x)\right),\qquad(\widetilde{w}_2^c,\partial_t\widetilde{w}_2^c)(0,x)=(0,0). □ w 2 c = ∂ t ( χ [ 2 , 3 ] ( x ) F 0 ( t , x ) ) , ( w 2 c , ∂ t w 2 c ) ( 0 , x ) = ( 0 , 0 ) .
Therefore, v ~ 2 b = w 2 b − χ [ 1 , 2 ] ( x ) ( w ~ 2 c + w r c ) \widetilde{v}_2^b=w_2^b-\chi_{[1,2]}(x)(\widetilde{w}_2^c+w_r^c) v 2 b = w 2 b − χ [ 1 , 2 ] ( x ) ( w 2 c + w r c ) is the solution of the IBVP
□ v ~ 2 b = [ Δ , χ [ 1 , 2 ] ( x ) ] ( w ~ 2 c + w r c ) , v ~ 2 b ∣ ∂ K = 0 , ( v ~ 2 b , ∂ t v ~ 2 b ) ( 0 , x ) = ( 0 , 0 ) . \Box\widetilde{v}_2^b=[\Delta,\chi_{[1,2]}(x)](\widetilde{w}_2^c+w_r^c),\qquad\widetilde{v}_2^b|_{\partial\mathcal{K}}=0,\qquad(\widetilde{v}_2^b,\partial_t\widetilde{v}_2^b)(0,x)=(0,0). □ v 2 b = [ Δ , χ [ 1 , 2 ] ( x )] ( w 2 c + w r c ) , v 2 b ∣ ∂ K = 0 , ( v 2 b , ∂ t v 2 b ) ( 0 , x ) = ( 0 , 0 ) .
Note that the estimate of v ~ 2 b \widetilde{v}_2^b v 2 b is similar to that of v 2 b v_2^b v 2 b in (134) . Then it holds that
⟨ x ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ ∣ ∂ a v ~ 2 b ( t , x ) ∣ ≲ ∑ ∣ b ∣ ≤ 9 sup s ∈ [ 0 , t ] ⟨ s ⟩ ∥ ∂ b ( w ~ 2 c + w r c ) ( s , y ) ∥ L ∞ ( ∣ y ∣ ≤ 2 ) . (138) \langle x\rangle^{1/2}\langle t-|x|\rangle|\partial^a\widetilde{v}_2^b(t,x)|\lesssim\sum_{|b|\le9}\sup_{s\in[0,t]}\langle s\rangle\|\partial^b(\widetilde{w}_2^c+w_r^c)(s,y)\|_{L^\infty(|y|\le2)}. \tag*{(138)} ⟨ x ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ ∣ ∂ a v 2 b ( t , x ) ∣ ≲ ∣ b ∣ ≤ 9 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ ∥ ∂ b ( w 2 c + w r c ) ( s , y ) ∥ L ∞ ( ∣ y ∣ ≤ 2 ) . (138)
By utilizing (35) to □ ∂ i w ~ 2 c = ∂ t ∂ i ( χ [ 2 , 3 ] ( x ) F 0 ( t , x ) ) \Box\partial_i\widetilde{w}_2^c=\partial_t\partial_i(\chi_{[2,3]}(x)F^0(t,x)) □ ∂ i w 2 c = ∂ t ∂ i ( χ [ 2 , 3 ] ( x ) F 0 ( t , x )) with ( w ~ 2 c , ∂ t w ~ 2 c ) ( 0 , x ) = ( 0 , 0 ) (\widetilde{w}_2^c,\partial_t\widetilde{w}_2^c)(0,x)=(0,0) ( w 2 c , ∂ t w 2 c ) ( 0 , x ) = ( 0 , 0 ) and Corollary 4.3 to □ ∂ b w ~ 2 c = ∂ t ∂ b ( χ [ 2 , 3 ] ( x ) F 0 ( t , x ) ) \Box\partial^b\widetilde{w}_2^c=\partial_t\partial^b(\chi_{[2,3]}(x)F^0(t,x)) □ ∂ b w 2 c = ∂ t ∂ b ( χ [ 2 , 3 ] ( x ) F 0 ( t , x )) , respectively, one has
∑ ∣ b ∣ ≤ 9 ⟨ t ⟩ ∣ ∂ b w ~ 2 c ( t , x ) ∣ ≲ ∑ ∣ a ∣ ≤ 13 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 , 1 + μ ( s , y ) ∣ Z a F 0 ( s , y ) ∣ + ∑ ∣ a ∣ ≤ 9 sup y ∈ K ∣ ∂ a F 0 ( 0 , y ) ∣ , ∣ x ∣ ≤ t / 10 or ∣ x ∣ ≤ 2 , \sum_{|b|\le9}\langle t\rangle|\partial^b\widetilde{w}_2^c(t,x)|\lesssim\sum_{|a|\le13}\sup_{(s,y)\in[0,t]\times\mathbb{R}^2\setminus\mathcal{K}_2}\langle y\rangle^{1/2}\mathcal{W}_{1,1+\mu}(s,y)|Z^aF^0(s,y)|
+\sum_{|a|\le9}\sup_{y\in\mathcal{K}}|\partial^aF^0(0,y)|,\qquad|x|\le t/10\ \text{or}\ |x|\le2, ∣ b ∣ ≤ 9 ∑ ⟨ t ⟩ ∣ ∂ b w 2 c ( t , x ) ∣ ≲ ∣ a ∣ ≤ 13 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 , 1 + μ ( s , y ) ∣ Z a F 0 ( s , y ) ∣ + ∣ a ∣ ≤ 9 ∑ y ∈ K sup ∣ ∂ a F 0 ( 0 , y ) ∣ , ∣ x ∣ ≤ t /10 or ∣ x ∣ ≤ 2 ,
⟨ x ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ 1 − η ∣ ∂ i w ~ 2 c ( t , x ) ∣ ≲ ln ( 2 + t + ∣ x ∣ ) ∑ ∣ a ∣ ≤ 2 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 , 1 ( s , y ) ∣ Z a F 0 ( s , y ) ∣ . (139) \langle x\rangle^{1/2}\langle t-|x|\rangle^{1-\eta}|\partial_i\widetilde{w}_2^c(t,x)|\lesssim\ln(2+t+|x|)\sum_{|a|\le2}\sup_{(s,y)\in[0,t]\times\mathbb{R}^2\setminus\mathcal{K}_2}\langle y\rangle^{1/2}\mathcal{W}_{1,1}(s,y)|Z^aF^0(s,y)|. \tag*{(139)} ⟨ x ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ 1 − η ∣ ∂ i w 2 c ( t , x ) ∣ ≲ ln ( 2 + t + ∣ x ∣ ) ∣ a ∣ ≤ 2 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 , 1 ( s , y ) ∣ Z a F 0 ( s , y ) ∣. (139)
It follows from (135) , (138) , (139) and ∂ i w 2 b = ∂ i v ~ 2 b + χ [ 1 , 2 ] ( x ) ∂ i ( w ~ 2 c + w r c ) + ( ∂ i χ [ 1 , 2 ] ( x ) ) ( w ~ 2 c + w r c ) \partial_iw_2^b=\partial_i\widetilde{v}_2^b+\chi_{[1,2]}(x)\partial_i(\widetilde{w}_2^c+w_r^c)+(\partial_i\chi_{[1,2]}(x))(\widetilde{w}_2^c+w_r^c) ∂ i w 2 b = ∂ i v 2 b + χ [ 1 , 2 ] ( x ) ∂ i ( w 2 c + w r c ) + ( ∂ i χ [ 1 , 2 ] ( x )) ( w 2 c + w r c ) that
⟨ x ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ 1 − η ∣ ∂ i w 2 b ( t , x ) ∣ ≲ ∑ ∣ a ∣ ≤ 8 sup y ∈ K ∣ ∂ a ( ∂ ≤ 1 F 0 , G 2 r ) ( 0 , y ) ∣ + ln ( 2 + t + ∣ x ∣ ) ∑ ∣ a ∣ ≤ 13 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 , 1 + μ ( s , y ) ∣ Z a F 0 ( s , y ) ∣ + ln ( 2 + t + ∣ x ∣ ) sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 3 / 2 , 1 + ν ( s , y ) ∣ G 2 r ( s , y ) ∣ + ∑ ∣ a ∣ ≤ 9 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 + μ , 1 ( s , y ) ∣ Z a G 2 r ( s , y ) ∣ . (140) \begin{aligned}
\langle x\rangle^{1/2}\langle t-|x|\rangle^{1-\eta}|\partial_iw_2^b(t,x)|
\lesssim{}& \sum_{|a|\le8}\sup_{y\in\mathcal{K}}|\partial^a(\partial^{\le1}F^0,G_2^r)(0,y)| \\
&+\ln(2+t+|x|)\sum_{|a|\le13}\sup_{(s,y)\in[0,t]\times\mathbb{R}^2\setminus\mathcal{K}_2}\langle y\rangle^{1/2}\mathcal{W}_{1,1+\mu}(s,y)|Z^aF^0(s,y)| \\
&+\ln(2+t+|x|)\sup_{(s,y)\in[0,t]\times\mathbb{R}^2\setminus\mathcal{K}_2}\langle y\rangle^{1/2}\mathcal{W}_{3/2,1+\nu}(s,y)|G_2^r(s,y)| \\
&+\sum_{|a|\le9}\sup_{(s,y)\in[0,t]\times\mathbb{R}^2\setminus\mathcal{K}_2}\langle y\rangle^{1/2}\mathcal{W}_{1+\mu,1}(s,y)|Z^aG_2^r(s,y)|. \tag*{(140)}
\end{aligned} ⟨ x ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ 1 − η ∣ ∂ i w 2 b ( t , x ) ∣ ≲ ∣ a ∣ ≤ 8 ∑ y ∈ K sup ∣ ∂ a ( ∂ ≤ 1 F 0 , G 2 r ) ( 0 , y ) ∣ + ln ( 2 + t + ∣ x ∣ ) ∣ a ∣ ≤ 13 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 , 1 + μ ( s , y ) ∣ Z a F 0 ( s , y ) ∣ + ln ( 2 + t + ∣ x ∣ ) ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 3/2 , 1 + ν ( s , y ) ∣ G 2 r ( s , y ) ∣ + ∣ a ∣ ≤ 9 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 + μ , 1 ( s , y ) ∣ Z a G 2 r ( s , y ) ∣. (140)
and for ∣ x ∣ ≤ t / 10 |x| \le t/10 ∣ x ∣ ≤ t /10
⟨ t ⟩ ∣ ∂ i w 2 b ( t , x ) ∣ ≲ ∑ ∣ a ∣ ≤ 8 sup y ∈ K ∣ ∂ a ( ∂ ≤ 1 F o , G 2 r ) ( 0 , y ) ∣ + ∑ ∣ a ∣ ≤ 13 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 w 1 , 1 + μ ( s , y ) ∣ Z a F o ( s , y ) ∣ + ln ( 2 + t + ∣ x ∣ ) sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 w 3 / 2 , 1 + ν ( s , y ) ∣ G 2 r ( s , y ) ∣ + ∑ ∣ a ∣ ≤ 9 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 w 1 + μ , 1 ( s , y ) ∣ Z a G 2 r ( s , y ) ∣ . (141) \langle t\rangle|\partial_i w_2^b(t,x)| \lesssim\sum_{|a|\le8}\sup_{y\in\mathcal{K}}|\partial^a(\partial^{\le1}F^o,G_2^r)(0,y)|
+\sum_{|a|\le13}\sup_{(s,y)\in[0,t]\times\mathbb{R}^2\setminus\mathcal{K}_2}\langle y\rangle^{1/2}w_{1,1+\mu}(s,y)|Z^aF^o(s,y)|
+\ln(2+t+|x|)\sup_{(s,y)\in[0,t]\times\mathbb{R}^2\setminus\mathcal{K}_2}\langle y\rangle^{1/2}w_{3/2,1+\nu}(s,y)|G_2^r(s,y)|
+\sum_{|a|\le9}\sup_{(s,y)\in[0,t]\times\mathbb{R}^2\setminus\mathcal{K}_2}\langle y\rangle^{1/2}w_{1+\mu,1}(s,y)|Z^aG_2^r(s,y)|. \tag*{(141)} ⟨ t ⟩ ∣ ∂ i w 2 b ( t , x ) ∣ ≲ ∣ a ∣ ≤ 8 ∑ y ∈ K sup ∣ ∂ a ( ∂ ≤ 1 F o , G 2 r ) ( 0 , y ) ∣ + ∣ a ∣ ≤ 13 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 w 1 , 1 + μ ( s , y ) ∣ Z a F o ( s , y ) ∣ + ln ( 2 + t + ∣ x ∣ ) ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 w 3/2 , 1 + ν ( s , y ) ∣ G 2 r ( s , y ) ∣ + ∣ a ∣ ≤ 9 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 w 1 + μ , 1 ( s , y ) ∣ Z a G 2 r ( s , y ) ∣. (141)
Therefore, (127) is achieved by (128) , (132) , (136) –(137) and (140) –(141) .
Energy estimates and local energy decay estimates
Bootstrap assumptions on the error solution
Let u u u and w h o m w_{\mathrm{hom}} w hom be the solutions to (1) and (7) , respectively, define the error solution
v ( t , x ) = u ( t , x ) − w h o m ( t , x ) . (142) v(t,x)=u(t,x)-w_{\mathrm{hom}}(t,x). \tag*{(142)} v ( t , x ) = u ( t , x ) − w hom ( t , x ) . (142)
We make the following bootstrap assumptions for all t ∈ [ 0 , T ε ] t\in[0,T_\varepsilon] t ∈ [ 0 , T ε ] with T ε = e κ / ε 2 T_\varepsilon=e^{\kappa/\varepsilon^2} T ε = e κ / ε 2 ,
∑ ∣ a ∣ ≤ N + 1 ∣ ∂ Z a v ∣ ≤ ε 1 ⟨ x ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ δ 0 − 1 , (143) \sum_{|a|\le N+1}|\partial Z^a v|\le\varepsilon_1\langle x\rangle^{-1/2}\langle t-|x|\rangle^{\delta_0-1}, \tag*{(143)} ∣ a ∣ ≤ N + 1 ∑ ∣ ∂ Z a v ∣ ≤ ε 1 ⟨ x ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ δ 0 − 1 , (143)
∑ ∣ a ∣ ≤ N + 1 ∣ ∂ ˉ Z a v ∣ ≤ ε 1 ⟨ x ⟩ − 1 / 2 ⟨ t + ∣ x ∣ ⟩ δ 0 − 1 , (144) \sum_{|a|\le N+1}|\bar{\partial}Z^a v|\le\varepsilon_1\langle x\rangle^{-1/2}\langle t+|x|\rangle^{\delta_0-1}, \tag*{(144)} ∣ a ∣ ≤ N + 1 ∑ ∣ ∂ ˉ Z a v ∣ ≤ ε 1 ⟨ x ⟩ − 1/2 ⟨ t + ∣ x ∣ ⟩ δ 0 − 1 , (144)
∑ ∣ a ∣ ≤ N + 1 ∣ Z a v ∣ ≤ ε 1 ⟨ t + ∣ x ∣ ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ 3 δ 0 − 1 / 2 , (145) \sum_{|a|\le N+1}|Z^a v|\le\varepsilon_1\langle t+|x|\rangle^{-1/2}\langle t-|x|\rangle^{3\delta_0-1/2}, \tag*{(145)} ∣ a ∣ ≤ N + 1 ∑ ∣ Z a v ∣ ≤ ε 1 ⟨ t + ∣ x ∣ ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ 3 δ 0 − 1/2 , (145)
∑ ∣ a ∣ ≤ N + 1 ∣ ∂ Z a v ∣ ≤ ε 2 2 ⟨ t ⟩ − 1 , ∣ x ∣ ≤ t / 10 , (146) \sum_{|a|\le N+1}|\partial Z^a v|\le\varepsilon_2^2\langle t\rangle^{-1},\qquad|x|\le t/10, \tag*{(146)} ∣ a ∣ ≤ N + 1 ∑ ∣ ∂ Z a v ∣ ≤ ε 2 2 ⟨ t ⟩ − 1 , ∣ x ∣ ≤ t /10 , (146)
where the small positive constants ε 1 \varepsilon_1 ε 1 and ε 2 \varepsilon_2 ε 2 will be determined later, ε ≤ ε 1 ≤ ε 2 ≤ 1 \varepsilon\le\varepsilon_1\le\varepsilon_2\le1 ε ≤ ε 1 ≤ ε 2 ≤ 1 and δ 0 = 10 − 3 \delta_0=10^{-3} δ 0 = 1 0 − 3 .
In addition, for convenience, the following assumption is imposed in the proof of Theorem 1.1
ε 1 2 ln ( 2 + t ) ≤ ε 2 2 ln ( 2 + t ) ≲ κ ε 2 2 ε 2 ≲ 1. (147) \varepsilon_1^2\ln(2+t)\le\varepsilon_2^2\ln(2+t)\lesssim\frac{\kappa\varepsilon_2^2}{\varepsilon^2}\lesssim1. \tag*{(147)} ε 1 2 ln ( 2 + t ) ≤ ε 2 2 ln ( 2 + t ) ≲ ε 2 κ ε 2 2 ≲ 1. (147)
Note that due to supp ( u 0 , u 1 ) ⊂ { x : ∣ x ∣ ≤ M 0 } \operatorname{supp}(u_0,u_1)\subset\{x:|x|\le M_0\} supp ( u 0 , u 1 ) ⊂ { x : ∣ x ∣ ≤ M 0 } , then the solution u u u and v v v are supported in { x ∈ K : ∣ x ∣ ≤ t + M 0 } \{x\in\mathcal{K}:|x|\le t+M_0\} { x ∈ K : ∣ x ∣ ≤ t + M 0 } .
Energy estimates
Lemma 6.1. Under the assumptions of Theorem 1.1 , let v v v be defined by (142) and suppose that (143) –(146) hold. Then we have that for t ∈ [ 0 , T ε ] t\in[0,T_\varepsilon] t ∈ [ 0 , T ε ] ,
∑ i ≤ 2 N ∥ ∂ ∂ t i v ( t ) ∥ L 2 ( K ) ≲ ε . (148) \sum_{i\le2N}\|\partial\partial_t^i v(t)\|_{L^2(\mathcal{K})}\lesssim\varepsilon. \tag*{(148)} i ≤ 2 N ∑ ∥ ∂ ∂ t i v ( t ) ∥ L 2 ( K ) ≲ ε . (148)
Proof. By (49) together with □ w h o m ≡ 0 \square w_{\mathrm{hom}}\equiv0 □ w hom ≡ 0 and (142) , one has
□ Z a v = Z a Q ( ∂ u , ∂ 2 u ) = Q α β γ ( ∂ α β 2 Z a v ∂ γ u + ∂ α β 2 Z a w h o m ∂ γ u ) + Q α β γ δ ( ∂ α β 2 Z a v ∂ γ u ∂ δ u + ∂ α β 2 Z a w h o m ∂ γ u ∂ δ u ) + ∑ b + c ≤ a , b < a Q a b c α β γ ∂ α β 2 Z b u ∂ γ Z c u + ∑ b + c + d ≤ a , b < a Q a b c d α β γ δ ∂ α β 2 Z b u ∂ γ Z c u ∂ δ Z d u . (149) \begin{aligned}
\square Z^{a}v
&=Z^{a}Q(\partial u,\partial^{2}u)\\
&=Q^{\alpha\beta\gamma}\left(\partial_{\alpha\beta}^{2}Z^{a}v\partial_{\gamma}u+\partial_{\alpha\beta}^{2}Z^{a}w_{\mathrm{hom}}\partial_{\gamma}u\right)+Q^{\alpha\beta\gamma\delta}\left(\partial_{\alpha\beta}^{2}Z^{a}v\partial_{\gamma}u\partial_{\delta}u+\partial_{\alpha\beta}^{2}Z^{a}w_{\mathrm{hom}}\partial_{\gamma}u\partial_{\delta}u\right)\\
&\quad+\sum_{\substack{b+c\leq a,b<a}}Q_{abc}^{\alpha\beta\gamma}\partial_{\alpha\beta}^{2}Z^{b}u\partial_{\gamma}Z^{c}u+\sum_{\substack{b+c+d\leq a,b<a}}Q_{abcd}^{\alpha\beta\gamma\delta}\partial_{\alpha\beta}^{2}Z^{b}u\partial_{\gamma}Z^{c}u\partial_{\delta}Z^{d}u. \tag*{(149)}
\end{aligned} □ Z a v = Z a Q ( ∂ u , ∂ 2 u ) = Q α β γ ( ∂ α β 2 Z a v ∂ γ u + ∂ α β 2 Z a w hom ∂ γ u ) + Q α β γ δ ( ∂ α β 2 Z a v ∂ γ u ∂ δ u + ∂ α β 2 Z a w hom ∂ γ u ∂ δ u ) + b + c ≤ a , b < a ∑ Q ab c α β γ ∂ α β 2 Z b u ∂ γ Z c u + b + c + d ≤ a , b < a ∑ Q ab c d α β γ δ ∂ α β 2 Z b u ∂ γ Z c u ∂ δ Z d u . (149)
where the summations ∑ α , β , γ = 0 2 \sum_{\alpha,\beta,\gamma=0}^{2} ∑ α , β , γ = 0 2 and ∑ α , β , γ , δ = 0 2 \sum_{\alpha,\beta,\gamma,\delta=0}^{2} ∑ α , β , γ , δ = 0 2 in (149) and the following are omitted. Multiplying (149) by e q ∂ t Z a v e^{q}\partial_{t}Z^{a}v e q ∂ t Z a v with the ghost weight q = q ( ∣ x ∣ − t ) q=q(|x|-t) q = q ( ∣ x ∣ − t ) and q ( s ) = ∫ − ∞ s d λ ⟨ λ ⟩ 3 / 2 q(s)=\int_{-\infty}^{s}\frac{d\lambda}{\langle\lambda\rangle^{3/2}} q ( s ) = ∫ − ∞ s ⟨ λ ⟩ 3/2 d λ introduced in [2 ] , yields
1 2 ∂ t [ e q ( ∣ ∂ t Z a v ∣ 2 + ∣ ∇ Z a v ∣ 2 ) ] − div ( e q ∂ t Z a v ∇ Z a v ) + e q 2 ⟨ t − ∣ x ∣ ⟩ 3 / 2 ∣ ∂ ˉ Z a v ∣ 2 = Q α β γ e q ∂ t Z a v ( ∂ α β 2 Z a v ∂ γ u + ∂ α β 2 Z a w h o m ∂ γ u ) + Q α β γ δ e q ∂ t Z a v ( ∂ α β 2 Z a v ∂ γ u ∂ δ u + ∂ α β 2 Z a w h o m ∂ γ u ∂ δ u ) + ∑ b + c ≤ a b < a Q a b c α β γ e q ∂ t Z a v ∂ α β 2 Z b u ∂ γ Z c u + ∑ b + c + d ≤ a b < a Q a b c d α β γ δ e q ∂ t Z a v ∂ α β 2 Z b u ∂ γ Z c u ∂ δ Z d u . (150) \begin{aligned}
\frac{1}{2}\partial_{t}\left[e^{q}\left(|\partial_{t}Z^{a}v|^{2}+|\nabla Z^{a}v|^{2}\right)\right]
&-\operatorname{div}\left(e^{q}\partial_{t}Z^{a}v\nabla Z^{a}v\right)+\frac{e^{q}}{2\langle t-|x|\rangle^{3/2}}|\bar{\partial}Z^{a}v|^{2}\\
&=Q^{\alpha\beta\gamma}e^{q}\partial_{t}Z^{a}v\left(\partial_{\alpha\beta}^{2}Z^{a}v\partial_{\gamma}u+\partial_{\alpha\beta}^{2}Z^{a}w_{\mathrm{hom}}\partial_{\gamma}u\right)\\
&\quad+Q^{\alpha\beta\gamma\delta}e^{q}\partial_{t}Z^{a}v\left(\partial_{\alpha\beta}^{2}Z^{a}v\partial_{\gamma}u\partial_{\delta}u+\partial_{\alpha\beta}^{2}Z^{a}w_{\mathrm{hom}}\partial_{\gamma}u\partial_{\delta}u\right)\\
&\quad+\sum_{\substack{b+c\leq a\\b<a}}Q_{abc}^{\alpha\beta\gamma}e^{q}\partial_{t}Z^{a}v\partial_{\alpha\beta}^{2}Z^{b}u\partial_{\gamma}Z^{c}u\\
&\quad+\sum_{\substack{b+c+d\leq a\\b<a}}Q_{abcd}^{\alpha\beta\gamma\delta}e^{q}\partial_{t}Z^{a}v\partial_{\alpha\beta}^{2}Z^{b}u\partial_{\gamma}Z^{c}u\partial_{\delta}Z^{d}u. \tag*{(150)}
\end{aligned} 2 1 ∂ t [ e q ( ∣ ∂ t Z a v ∣ 2 + ∣∇ Z a v ∣ 2 ) ] − div ( e q ∂ t Z a v ∇ Z a v ) + 2 ⟨ t − ∣ x ∣ ⟩ 3/2 e q ∣ ∂ ˉ Z a v ∣ 2 = Q α β γ e q ∂ t Z a v ( ∂ α β 2 Z a v ∂ γ u + ∂ α β 2 Z a w hom ∂ γ u ) + Q α β γ δ e q ∂ t Z a v ( ∂ α β 2 Z a v ∂ γ u ∂ δ u + ∂ α β 2 Z a w hom ∂ γ u ∂ δ u ) + b + c ≤ a b < a ∑ Q ab c α β γ e q ∂ t Z a v ∂ α β 2 Z b u ∂ γ Z c u + b + c + d ≤ a b < a ∑ Q ab c d α β γ δ e q ∂ t Z a v ∂ α β 2 Z b u ∂ γ Z c u ∂ δ Z d u . (150)
where ∣ ∂ ˉ f ∣ = ( ∣ ∂ ˉ 1 f ∣ 2 + ∣ ∂ ˉ 2 f ∣ 2 ) 1 / 2 |\bar{\partial}f|=(|\bar{\partial}_{1}f|^{2}+|\bar{\partial}_{2}f|^{2})^{1/2} ∣ ∂ ˉ f ∣ = ( ∣ ∂ ˉ 1 f ∣ 2 + ∣ ∂ ˉ 2 f ∣ 2 ) 1/2 . For the terms in the second and third lines of (150) , it follows from the symmetric conditions (3) and direct computations that
Q α β γ e q ∂ t Z a v ∂ α β 2 Z a v ∂ γ u + Q α β γ e q ∂ t Z a v ∂ α β 2 Z a v ∂ γ u ∂ δ u = ∂ α ( Q α β γ e q ∂ t Z a v ∂ β Z a v ∂ γ u ) − Q α β γ ∂ t Z a v ∂ β Z a v ∂ α ( e q ∂ γ u ) − 1 2 ∂ t ( Q α β γ e q ∂ α Z a v ∂ β Z a v ∂ γ u ) + 1 2 Q α β γ ∂ α Z a v ∂ β Z a v ∂ t ( e q ∂ γ u ) + ∂ α ( Q α β γ δ e q ∂ t Z a v ∂ β Z a v ∂ γ u ∂ δ u ) − Q α β γ δ ∂ t Z a v ∂ β Z a v ∂ α ( e q ∂ γ u ∂ δ u ) − 1 2 ∂ t ( Q α β γ δ e q ∂ α Z a v ∂ β Z a v ∂ γ u ∂ δ u ) + 1 2 Q α β γ δ ∂ α Z a v ∂ β Z a v ∂ t ( e q ∂ γ u ∂ δ u ) . (151) \begin{aligned}
&Q^{\alpha\beta\gamma}e^{q}\partial_{t}Z^{a}v\partial_{\alpha\beta}^{2}Z^{a}v\partial_{\gamma}u+Q^{\alpha\beta\gamma}e^{q}\partial_{t}Z^{a}v\partial_{\alpha\beta}^{2}Z^{a}v\partial_{\gamma}u\partial_{\delta}u\\
&=\partial_{\alpha}\left(Q^{\alpha\beta\gamma}e^{q}\partial_{t}Z^{a}v\partial_{\beta}Z^{a}v\partial_{\gamma}u\right)-Q^{\alpha\beta\gamma}\partial_{t}Z^{a}v\partial_{\beta}Z^{a}v\partial_{\alpha}\left(e^{q}\partial_{\gamma}u\right)\\
&\quad-\frac{1}{2}\partial_{t}\left(Q^{\alpha\beta\gamma}e^{q}\partial_{\alpha}Z^{a}v\partial_{\beta}Z^{a}v\partial_{\gamma}u\right)+\frac{1}{2}Q^{\alpha\beta\gamma}\partial_{\alpha}Z^{a}v\partial_{\beta}Z^{a}v\partial_{t}\left(e^{q}\partial_{\gamma}u\right)\\
&\quad+\partial_{\alpha}\left(Q^{\alpha\beta\gamma\delta}e^{q}\partial_{t}Z^{a}v\partial_{\beta}Z^{a}v\partial_{\gamma}u\partial_{\delta}u\right)-Q^{\alpha\beta\gamma\delta}\partial_{t}Z^{a}v\partial_{\beta}Z^{a}v\partial_{\alpha}\left(e^{q}\partial_{\gamma}u\partial_{\delta}u\right)\\
&\quad-\frac{1}{2}\partial_{t}\left(Q^{\alpha\beta\gamma\delta}e^{q}\partial_{\alpha}Z^{a}v\partial_{\beta}Z^{a}v\partial_{\gamma}u\partial_{\delta}u\right)+\frac{1}{2}Q^{\alpha\beta\gamma\delta}\partial_{\alpha}Z^{a}v\partial_{\beta}Z^{a}v\partial_{t}\left(e^{q}\partial_{\gamma}u\partial_{\delta}u\right). \tag*{(151)}
\end{aligned} Q α β γ e q ∂ t Z a v ∂ α β 2 Z a v ∂ γ u + Q α β γ e q ∂ t Z a v ∂ α β 2 Z a v ∂ γ u ∂ δ u = ∂ α ( Q α β γ e q ∂ t Z a v ∂ β Z a v ∂ γ u ) − Q α β γ ∂ t Z a v ∂ β Z a v ∂ α ( e q ∂ γ u ) − 2 1 ∂ t ( Q α β γ e q ∂ α Z a v ∂ β Z a v ∂ γ u ) + 2 1 Q α β γ ∂ α Z a v ∂ β Z a v ∂ t ( e q ∂ γ u ) + ∂ α ( Q α β γ δ e q ∂ t Z a v ∂ β Z a v ∂ γ u ∂ δ u ) − Q α β γ δ ∂ t Z a v ∂ β Z a v ∂ α ( e q ∂ γ u ∂ δ u ) − 2 1 ∂ t ( Q α β γ δ e q ∂ α Z a v ∂ β Z a v ∂ γ u ∂ δ u ) + 2 1 Q α β γ δ ∂ α Z a v ∂ β Z a v ∂ t ( e q ∂ γ u ∂ δ u ) . (151)
Integrating (150) and (151) over [ 0 , t ] × K [0,t]\times\mathcal{K} [ 0 , t ] × K with Z a = ∂ t i Z^{a}=\partial_{t}^{i} Z a = ∂ t i (i ≤ 2 N i\leq2N i ≤ 2 N ) and the boundary conditions ∂ t l u ∣ [ 0 , ∞ ) × ∂ K = 0 \partial_{t}^{l}u|_{[0,\infty)\times\partial\mathcal{K}}=0 ∂ t l u ∣ [ 0 , ∞ ) × ∂ K = 0 (l ≥ 0 l\geq0 l ≥ 0 ), implies
∥ ∂ ∂ t i v ( t ) ∥ L 2 ( K ) 2 + ∫ 0 t ∫ K ∣ ∂ ˉ ∂ t i v ∣ 2 ⟨ s − ∣ x ∣ ⟩ 3 / 2 d x d s ≲ ∥ ∂ ∂ t i v ( 0 ) ∥ L 2 ( K ) 2 ( 1 + ∥ ∂ u ( 0 ) ∥ L ∞ ( K ) ) + ∥ ∂ u ( t ) ∥ L ∞ ( K ) ∥ ∂ ∂ t i v ( t ) ∥ L 2 ( K ) + ∑ b + c ≤ a , b < a ∣ ∫ 0 t ∫ K I 1 a b c d x d s ∣ + ∑ b + c + d ≤ a , b < a ∣ ∫ 0 t ∫ K I 2 a b c d d x d s ∣ + ∫ 0 t ∫ K ∣ I a ∣ d x d s , (152) \begin{aligned}
\|\partial\partial_{t}^{i}v(t)\|_{L^{2}(\mathcal{K})}^{2}
+\int_{0}^{t}\int_{\mathcal{K}}\frac{|\bar{\partial}\partial_{t}^{i}v|^{2}}{\langle s-|x|\rangle^{3/2}}\,dx\,ds
&\lesssim\|\partial\partial_{t}^{i}v(0)\|_{L^{2}(\mathcal{K})}^{2}\left(1+\|\partial u(0)\|_{L^{\infty}(\mathcal{K})}\right)\\
&\quad+\|\partial u(t)\|_{L^{\infty}(\mathcal{K})}\|\partial\partial_{t}^{i}v(t)\|_{L^{2}(\mathcal{K})}\\
&\quad+\sum_{\substack{b+c\leq a,b<a}}\left|\int_{0}^{t}\int_{\mathcal{K}}I_{1}^{abc}\,dx\,ds\right|
+\sum_{\substack{b+c+d\leq a,b<a}}\left|\int_{0}^{t}\int_{\mathcal{K}}I_{2}^{abcd}\,dx\,ds\right|
+\int_{0}^{t}\int_{\mathcal{K}}|I^{a}|\,dx\,ds, \tag*{(152)}
\end{aligned} ∥ ∂ ∂ t i v ( t ) ∥ L 2 ( K ) 2 + ∫ 0 t ∫ K ⟨ s − ∣ x ∣ ⟩ 3/2 ∣ ∂ ˉ ∂ t i v ∣ 2 d x d s ≲ ∥ ∂ ∂ t i v ( 0 ) ∥ L 2 ( K ) 2 ( 1 + ∥ ∂ u ( 0 ) ∥ L ∞ ( K ) ) + ∥ ∂ u ( t ) ∥ L ∞ ( K ) ∥ ∂ ∂ t i v ( t ) ∥ L 2 ( K ) + b + c ≤ a , b < a ∑ ∫ 0 t ∫ K I 1 ab c d x d s + b + c + d ≤ a , b < a ∑ ∫ 0 t ∫ K I 2 ab c d d x d s + ∫ 0 t ∫ K ∣ I a ∣ d x d s , (152)
where
I 1 a b c = Q a b c α β γ e q ∂ t Z a v ∂ α β 2 Z b v ∂ γ Z c u , I 2 a b c d = Q a b c d α β γ δ e q ∂ t Z a v ∂ α β 2 Z b u ∂ γ Z c u ∂ δ Z d u , I a = − Q a b c α β γ e q ∂ t Z a v ∂ β Z a v ∂ α q ∂ γ u − Q a b c α β γ e q ∂ t Z a v ∂ β Z a v ∂ α γ 2 u + 1 2 Q a b c α β γ e q ∂ α Z a v ∂ β Z a v ∂ t ∂ γ u + 1 2 Q a b c α β γ e q ∂ α Z a v ∂ β Z a v ∂ γ 2 u − Q a b c α β γ δ ∂ t Z a v ∂ β Z a v ∂ α ( e q ∂ γ u ∂ δ u ) + 1 2 Q a b c α β γ δ ∂ α Z a v ∂ β Z a v ∂ t ( e q ∂ γ u ∂ δ u ) + Q a b c α β γ e q ∂ t Z a v ∂ α β 2 Z a w h o m ∂ γ u + Q a b c α β γ δ e q ∂ t Z a v ∂ α β 2 Z a w h o m ∂ γ u ∂ δ u . (153) \begin{aligned}
I_1^{abc} &= Q_{abc}^{\alpha\beta\gamma}e^q\partial_t Z^a v\partial_{\alpha\beta}^2 Z^b v\partial_\gamma Z^c u,\\
I_2^{abcd} &= Q_{abcd}^{\alpha\beta\gamma\delta}e^q\partial_t Z^a v\partial_{\alpha\beta}^2 Z^b u\partial_\gamma Z^c u\partial_\delta Z^d u,\\
I^a &= -Q_{abc}^{\alpha\beta\gamma}e^q\partial_t Z^a v\partial_\beta Z^a v\partial_\alpha q\partial_\gamma u-Q_{abc}^{\alpha\beta\gamma}e^q\partial_t Z^a v\partial_\beta Z^a v\partial_{\alpha\gamma}^2u\\
&\quad+\frac{1}{2}Q_{abc}^{\alpha\beta\gamma}e^q\partial_\alpha Z^a v\partial_\beta Z^a v\partial_t\partial_\gamma u+\frac{1}{2}Q_{abc}^{\alpha\beta\gamma}e^q\partial_\alpha Z^a v\partial_\beta Z^a v\partial_{\gamma}^2u\\
&\quad-Q_{abc}^{\alpha\beta\gamma\delta}\partial_t Z^a v\partial_\beta Z^a v\partial_\alpha(e^q\partial_\gamma u\partial_\delta u)+\frac{1}{2}Q_{abc}^{\alpha\beta\gamma\delta}\partial_\alpha Z^a v\partial_\beta Z^a v\partial_t(e^q\partial_\gamma u\partial_\delta u)\\
&\quad+Q_{abc}^{\alpha\beta\gamma}e^q\partial_t Z^a v\partial_{\alpha\beta}^2 Z^a w_{\mathrm{hom}}\partial_\gamma u+Q_{abc}^{\alpha\beta\gamma\delta}e^q\partial_t Z^a v\partial_{\alpha\beta}^2 Z^a w_{\mathrm{hom}}\partial_\gamma u\partial_\delta u. \tag*{(153)}
\end{aligned} I 1 ab c I 2 ab c d I a = Q ab c α β γ e q ∂ t Z a v ∂ α β 2 Z b v ∂ γ Z c u , = Q ab c d α β γ δ e q ∂ t Z a v ∂ α β 2 Z b u ∂ γ Z c u ∂ δ Z d u , = − Q ab c α β γ e q ∂ t Z a v ∂ β Z a v ∂ α q ∂ γ u − Q ab c α β γ e q ∂ t Z a v ∂ β Z a v ∂ α γ 2 u + 2 1 Q ab c α β γ e q ∂ α Z a v ∂ β Z a v ∂ t ∂ γ u + 2 1 Q ab c α β γ e q ∂ α Z a v ∂ β Z a v ∂ γ 2 u − Q ab c α β γ δ ∂ t Z a v ∂ β Z a v ∂ α ( e q ∂ γ u ∂ δ u ) + 2 1 Q ab c α β γ δ ∂ α Z a v ∂ β Z a v ∂ t ( e q ∂ γ u ∂ δ u ) + Q ab c α β γ e q ∂ t Z a v ∂ α β 2 Z a w hom ∂ γ u + Q ab c α β γ δ e q ∂ t Z a v ∂ α β 2 Z a w hom ∂ γ u ∂ δ u . (153)
Although Z a = ∂ t i Z^a=\partial_t^i Z a = ∂ t i has been selected, the estimates of I 1 a b c I_1^{abc} I 1 ab c , I 2 a b c d I_2^{abcd} I 2 ab c d , I a I^a I a also hold for the vector fields Z ∈ { ∂ , Ω } Z\in\{\partial,\Omega\} Z ∈ { ∂ , Ω } . By the virtue of definition (142) , one has
I 1 a b c = Q a b c α β γ e q ∂ t Z a v ∂ α β 2 Z b w h o m ∂ γ Z c w h o m + I 11 a b c , I 11 a b c = Q a b c α β γ e q ∂ t Z a v ( ∂ α β 2 Z b v ∂ γ Z c v + ∂ α β 2 Z b v ∂ γ Z c w h o m + ∂ α β 2 Z b w h o m ∂ γ Z c v ) . (154) \begin{aligned}
I_1^{abc}&=Q_{abc}^{\alpha\beta\gamma}e^q\partial_t Z^a v\partial_{\alpha\beta}^2 Z^b w_{\mathrm{hom}}\partial_\gamma Z^c w_{\mathrm{hom}}+I_{11}^{abc},\\
I_{11}^{abc}&=Q_{abc}^{\alpha\beta\gamma}e^q\partial_t Z^a v\left(\partial_{\alpha\beta}^2 Z^b v\partial_\gamma Z^c v+\partial_{\alpha\beta}^2 Z^b v\partial_\gamma Z^c w_{\mathrm{hom}}+\partial_{\alpha\beta}^2 Z^b w_{\mathrm{hom}}\partial_\gamma Z^c v\right). \tag*{(154)}
\end{aligned} I 1 ab c I 11 ab c = Q ab c α β γ e q ∂ t Z a v ∂ α β 2 Z b w hom ∂ γ Z c w hom + I 11 ab c , = Q ab c α β γ e q ∂ t Z a v ( ∂ α β 2 Z b v ∂ γ Z c v + ∂ α β 2 Z b v ∂ γ Z c w hom + ∂ α β 2 Z b w hom ∂ γ Z c v ) . (154)
For the first term on the right hand side of the first line in (154) , it follows from (47) , (50) and (91) that
∣ Q a b c α β γ ∂ α β 2 Z b w h o m ∂ γ Z c w h o m ∣ ≲ ∣ ∂ ‾ ∂ Z b w h o m ∣ ∣ ∂ Z c w h o m ∣ + ∣ ∂ 2 Z b w h o m ∣ ∣ ∂ ‾ Z c w h o m ∣ ≲ ε 2 ⟨ x ⟩ − 1 ⟨ s + ∣ x ∣ ⟩ − 1 ⟨ s − ∣ x ∣ ⟩ − 1 , \left|Q_{abc}^{\alpha\beta\gamma}\partial_{\alpha\beta}^2 Z^b w_{\mathrm{hom}}\partial_\gamma Z^c w_{\mathrm{hom}}\right|
\lesssim\left|\overline{\partial}\partial Z^b w_{\mathrm{hom}}\right|\left|\partial Z^c w_{\mathrm{hom}}\right|+\left|\partial^2 Z^b w_{\mathrm{hom}}\right|\left|\overline{\partial}Z^c w_{\mathrm{hom}}\right|
\lesssim\varepsilon^2\langle x\rangle^{-1}\langle s+|x|\rangle^{-1}\langle s-|x|\rangle^{-1}, Q ab c α β γ ∂ α β 2 Z b w hom ∂ γ Z c w hom ≲ ∂ ∂ Z b w hom ∣ ∂ Z c w hom ∣ + ∂ 2 Z b w hom ∂ Z c w hom ≲ ε 2 ⟨ x ⟩ − 1 ⟨ s + ∣ x ∣ ⟩ − 1 ⟨ s − ∣ x ∣ ⟩ − 1 ,
which yields
∥ Q a b c α β γ ∂ α β 2 Z b w h o m ∂ γ Z c w h o m ∥ L 2 ( K ) ≲ ε 2 ⟨ s ⟩ − 3 / 2 . (155) \left\|Q_{abc}^{\alpha\beta\gamma}\partial_{\alpha\beta}^2 Z^b w_{\mathrm{hom}}\partial_\gamma Z^c w_{\mathrm{hom}}\right\|_{L^2(K)}
\lesssim\varepsilon^2\langle s\rangle^{-3/2}. \tag*{(155)} Q ab c α β γ ∂ α β 2 Z b w hom ∂ γ Z c w hom L 2 ( K ) ≲ ε 2 ⟨ s ⟩ − 3/2 . (155)
When ∣ b ∣ ≤ N |b|\le N ∣ b ∣ ≤ N in the first term I 11 a b c I_{11}^{abc} I 11 ab c , it can be deduced from (47) , (50) , (143) and (144) in the region ∣ x ∣ ≥ t / 10 |x|\ge t/10 ∣ x ∣ ≥ t /10 that
∣ Q a b c α β γ ∂ α β 2 Z b v ∂ γ Z c v ∣ ≲ ∣ ∂ ‾ ∂ Z b v ∣ ∣ ∂ Z c v ∣ + ∣ ∂ 2 Z b v ∣ ∣ ∂ ‾ Z c v ∣ ≲ ε 1 ⟨ s ⟩ δ 0 − 3 / 2 ∣ ∂ Z c v ∣ + ε 1 ⟨ s ⟩ − 1 / 2 ⟨ s − ∣ x ∣ ⟩ δ 0 − 1 ∣ ∂ ‾ Z c v ∣ . (156) \left|Q_{abc}^{\alpha\beta\gamma}\partial_{\alpha\beta}^2 Z^b v\partial_\gamma Z^c v\right|
\lesssim\left|\overline{\partial}\partial Z^b v\right|\left|\partial Z^c v\right|+\left|\partial^2 Z^b v\right|\left|\overline{\partial}Z^c v\right|
\lesssim\varepsilon_1\langle s\rangle^{\delta_0-3/2}\left|\partial Z^c v\right|+\varepsilon_1\langle s\rangle^{-1/2}\langle s-|x|\rangle^{\delta_0-1}\left|\overline{\partial}Z^c v\right|. \tag*{(156)} Q ab c α β γ ∂ α β 2 Z b v ∂ γ Z c v ≲ ∂ ∂ Z b v ∣ ∂ Z c v ∣ + ∂ 2 Z b v ∂ Z c v ≲ ε 1 ⟨ s ⟩ δ 0 − 3/2 ∣ ∂ Z c v ∣ + ε 1 ⟨ s ⟩ − 1/2 ⟨ s − ∣ x ∣ ⟩ δ 0 − 1 ∂ Z c v . (156)
While in the region ∣ x ∣ ≤ s / 10 |x|\le s/10 ∣ x ∣ ≤ s /10 , (92) and (146) directly lead to
∥ ∂ α β 2 Z b v ∂ γ Z c v ∥ L 2 ( K ∩ { ∣ x ∣ ≤ s / 10 } ) ≲ ∑ j ≤ 2 N ∥ ∂ ∂ j v ( s ) ∥ L 2 ( K ) ( ε ⟨ s ⟩ − 1 ln − 2 ( 2 + s ) + ε 2 2 ⟨ s ⟩ − 1 ) . (157) \left\|\partial_{\alpha\beta}^2 Z^b v\partial_\gamma Z^c v\right\|_{L^2(K\cap\{|x|\le s/10\})}
\lesssim\sum_{j\le2N}\left\|\partial\partial^j v(s)\right\|_{L^2(K)}\left(\varepsilon\langle s\rangle^{-1}\ln^{-2}(2+s)+\varepsilon_2^2\langle s\rangle^{-1}\right). \tag*{(157)} ∂ α β 2 Z b v ∂ γ Z c v L 2 ( K ∩ { ∣ x ∣ ≤ s /10 }) ≲ j ≤ 2 N ∑ ∂ ∂ j v ( s ) L 2 ( K ) ( ε ⟨ s ⟩ − 1 ln − 2 ( 2 + s ) + ε 2 2 ⟨ s ⟩ − 1 ) . (157)
When ∣ b ∣ ≥ N + 1 |b|\ge N+1 ∣ b ∣ ≥ N + 1 in the first term I 11 a b c I_{11}^{abc} I 11 ab c , one can see that for ∣ c ∣ ≤ N − 1 |c|\le N-1 ∣ c ∣ ≤ N − 1 ,
Q a b c α β γ e q ∂ t Z a v ∂ α β 2 Z b v ∂ γ Z c v = ∂ α ( Q a b c α β γ e q ∂ t Z a v ∂ β Z b v ∂ γ Z c v ) − ∂ t ( Q a b c α β γ e q ∂ α Z a v ∂ β Z b v ∂ γ Z c v ) − Q a b c α β γ e q ∂ t Z a v ∂ α γ 2 Z b v ∂ β Z c v − Q a b c α β γ e q ∂ α ∂ t Z a v ∂ β Z b v ∂ γ Z c v + Q a b c α β γ ∂ α Z a v ∂ t ( e q ∂ β Z b v ∂ γ Z c v ) . \begin{aligned}
Q_{abc}^{\alpha\beta\gamma}e^q\partial_t Z^a v\partial_{\alpha\beta}^2 Z^b v\partial_\gamma Z^c v
={}&\partial_\alpha\left(Q_{abc}^{\alpha\beta\gamma}e^q\partial_t Z^a v\partial_\beta Z^b v\partial_\gamma Z^c v\right)
-\partial_t\left(Q_{abc}^{\alpha\beta\gamma}e^q\partial_\alpha Z^a v\partial_\beta Z^b v\partial_\gamma Z^c v\right)\\
&-Q_{abc}^{\alpha\beta\gamma}e^q\partial_t Z^a v\partial_{\alpha\gamma}^2 Z^b v\partial_\beta Z^c v
-Q_{abc}^{\alpha\beta\gamma}e^q\partial_\alpha\partial_t Z^a v\partial_\beta Z^b v\partial_\gamma Z^c v\\
&+Q_{abc}^{\alpha\beta\gamma}\partial_\alpha Z^a v\partial_t\left(e^q\partial_\beta Z^b v\partial_\gamma Z^c v\right).
\end{aligned} Q ab c α β γ e q ∂ t Z a v ∂ α β 2 Z b v ∂ γ Z c v = ∂ α ( Q ab c α β γ e q ∂ t Z a v ∂ β Z b v ∂ γ Z c v ) − ∂ t ( Q ab c α β γ e q ∂ α Z a v ∂ β Z b v ∂ γ Z c v ) − Q ab c α β γ e q ∂ t Z a v ∂ α γ 2 Z b v ∂ β Z c v − Q ab c α β γ e q ∂ α ∂ t Z a v ∂ β Z b v ∂ γ Z c v + Q ab c α β γ ∂ α Z a v ∂ t ( e q ∂ β Z b v ∂ γ Z c v ) .
Note that the terms in the last two lines above can be treated as for (156) and (157) with ∂ ˉ q = 0 \bar{\partial}q=0 ∂ ˉ q = 0 . Then by the Cauchy-Schwarz inequality, we arrive at
∣ ∫ 0 t ∫ K Q a b c α β γ e α q ∂ t Z a ∂ β Z b ∂ γ Z c v d x d s ∣ ≲ δ ~ ∑ j ≤ 2 N ∫ 0 t ∫ K ∣ ∂ ∂ t j v ∣ 2 ⟨ s − ∣ x ∣ ⟩ 3 / 2 d x d s + δ ~ − 1 ∑ j ≤ 2 N ∫ 0 t ∥ ∂ ∂ t j v ∥ L 2 ( K ) 2 ( ε 2 2 ⟨ s ⟩ − 1 + ε 1 ⟨ s ⟩ − 1 ln − 2 ( 2 + s ) ) d s . (158) \begin{aligned}
\left|\int_{0}^{t}\int_{\mathcal{K}}Q^{\alpha\beta\gamma}_{abc}e^{q}_{\alpha}\partial_{t}Z^{a}\partial_{\beta}Z^{b}\partial_{\gamma}Z^{c}v\,dxds\right|\lesssim\widetilde{\delta}\sum_{j\leq2N}\int_{0}^{t}\int_{\mathcal{K}}\frac{|\partial\partial^{j}_{t}v|^{2}}{\langle s-|x|\rangle^{3/2}}\,dxds \\
&+\widetilde{\delta}^{-1}\sum_{j\leq2N}\int_{0}^{t}\|\partial\partial^{j}_{t}v\|_{L^{2}(\mathcal{K})}^{2}\left(\varepsilon_{2}^{2}\langle s\rangle^{-1}+\varepsilon_{1}\langle s\rangle^{-1}\ln^{-2}(2+s)\right)\,ds. \tag*{(158)}
\end{aligned} ∫ 0 t ∫ K Q ab c α β γ e α q ∂ t Z a ∂ β Z b ∂ γ Z c v d x d s ≲ δ j ≤ 2 N ∑ ∫ 0 t ∫ K ⟨ s − ∣ x ∣ ⟩ 3/2 ∣ ∂ ∂ t j v ∣ 2 d x d s + δ − 1 j ≤ 2 N ∑ ∫ 0 t ∥ ∂ ∂ t j v ∥ L 2 ( K ) 2 ( ε 2 2 ⟨ s ⟩ − 1 + ε 1 ⟨ s ⟩ − 1 ln − 2 ( 2 + s ) ) d s . (158)
where δ ~ > 0 \widetilde{\delta}>0 δ > 0 is a small constant. In addition, the estimates of the other terms in I 11 a b c I_{11}^{abc} I 11 ab c are analogous.
Therefore, collecting (154) –(158) yields
∑ b + c ≤ a , b < a ∣ ∫ 0 t ∫ K I 11 a b c d x d s ∣ \sum_{b+c\leq a,b<a}\left|\int_{0}^{t}\int_{\mathcal{K}}I_{11}^{abc}\,dxds\right| b + c ≤ a , b < a ∑ ∫ 0 t ∫ K I 11 ab c d x d s
≲ ∥ ( ∂ w h o m , ∂ v ) ( 0 ) ∥ L ∞ ( K ) ∥ ∂ ∂ t i v ( 0 ) ∥ L 2 ( K ) 2 + ∥ ( ∂ w h o m , ∂ v ) ( t ) ∥ L ∞ ( K ) ∥ ∂ ∂ t i v ( t ) ∥ L 2 ( K ) 2 + ε 2 2 ∑ j ≤ 2 N ∫ 0 t ∥ ∂ ∂ t j v ∥ L 2 ( K ) 2 ⟨ s ⟩ − 3 / 2 d s + δ ~ ∑ j ≤ 2 N ∫ 0 t ∫ K ∣ ∂ ∂ t j v ∣ 2 ⟨ s − ∣ x ∣ ⟩ 3 / 2 d x d s + δ ~ − 1 ∑ j ≤ 2 N ∫ 0 t ∥ ∂ ∂ t j v ∥ L 2 ( K ) 2 ( ε 1 ⟨ s ⟩ − 1 ln − 2 ( 2 + s ) + ε 2 2 ⟨ s ⟩ − 1 ) d s . (159) \begin{aligned}
\lesssim\|(\partial w_{\mathrm{hom}},\partial v)(0)\|_{L^{\infty}(\mathcal{K})}\|\partial\partial^{i}_{t}v(0)\|_{L^{2}(\mathcal{K})}^{2}+\|(\partial w_{\mathrm{hom}},\partial v)(t)\|_{L^{\infty}(\mathcal{K})}\|\partial\partial^{i}_{t}v(t)\|_{L^{2}(\mathcal{K})}^{2} \\
&+\varepsilon_{2}^{2}\sum_{j\leq2N}\int_{0}^{t}\|\partial\partial^{j}_{t}v\|_{L^{2}(\mathcal{K})}^{2}\langle s\rangle^{-3/2}\,ds+\widetilde{\delta}\sum_{j\leq2N}\int_{0}^{t}\int_{\mathcal{K}}\frac{|\partial\partial^{j}_{t}v|^{2}}{\langle s-|x|\rangle^{3/2}}\,dxds \\
&+\widetilde{\delta}^{-1}\sum_{j\leq2N}\int_{0}^{t}\|\partial\partial^{j}_{t}v\|_{L^{2}(\mathcal{K})}^{2}\left(\varepsilon_{1}\langle s\rangle^{-1}\ln^{-2}(2+s)+\varepsilon_{2}^{2}\langle s\rangle^{-1}\right)\,ds. \tag*{(159)}
\end{aligned} ≲ ∥ ( ∂ w hom , ∂ v ) ( 0 ) ∥ L ∞ ( K ) ∥ ∂ ∂ t i v ( 0 ) ∥ L 2 ( K ) 2 + ∥ ( ∂ w hom , ∂ v ) ( t ) ∥ L ∞ ( K ) ∥ ∂ ∂ t i v ( t ) ∥ L 2 ( K ) 2 + ε 2 2 j ≤ 2 N ∑ ∫ 0 t ∥ ∂ ∂ t j v ∥ L 2 ( K ) 2 ⟨ s ⟩ − 3/2 d s + δ j ≤ 2 N ∑ ∫ 0 t ∫ K ⟨ s − ∣ x ∣ ⟩ 3/2 ∣ ∂ ∂ t j v ∣ 2 d x d s + δ − 1 j ≤ 2 N ∑ ∫ 0 t ∥ ∂ ∂ t j v ∥ L 2 ( K ) 2 ( ε 1 ⟨ s ⟩ − 1 ln − 2 ( 2 + s ) + ε 2 2 ⟨ s ⟩ − 1 ) d s . (159)
Analogously for the treatment on the term I 2 a b c d I_{2}^{abcd} I 2 ab c d in (153) , it can be concluded from (91) and (143) that
∑ b + c + d ≤ a , b < a ∣ ∫ 0 t ∫ K I 2 a b c d d x d s ∣ \sum_{b+c+d\leq a,b<a}\left|\int_{0}^{t}\int_{\mathcal{K}}I_{2}^{abcd}\,dxds\right| b + c + d ≤ a , b < a ∑ ∫ 0 t ∫ K I 2 ab c d d x d s
≲ ε 2 3 ∑ j ≤ 2 N ∫ 0 t ∥ ∂ ∂ t j v ∥ L 2 ( K ) 2 ⟨ s ⟩ − 1 d s + ε 2 2 ∑ j ≤ 2 N ∫ 0 t ∥ ∂ ∂ t j v ∥ L 2 ( K ) 2 ⟨ s ⟩ − 1 d s . (160) \lesssim\varepsilon_{2}^{3}\sum_{j\leq2N}\int_{0}^{t}\|\partial\partial^{j}_{t}v\|_{L^{2}(\mathcal{K})}^{2}\langle s\rangle^{-1}\,ds+\varepsilon_{2}^{2}\sum_{j\leq2N}\int_{0}^{t}\|\partial\partial^{j}_{t}v\|_{L^{2}(\mathcal{K})}^{2}\langle s\rangle^{-1}\,ds. \tag*{(160)} ≲ ε 2 3 j ≤ 2 N ∑ ∫ 0 t ∥ ∂ ∂ t j v ∥ L 2 ( K ) 2 ⟨ s ⟩ − 1 d s + ε 2 2 j ≤ 2 N ∑ ∫ 0 t ∥ ∂ ∂ t j v ∥ L 2 ( K ) 2 ⟨ s ⟩ − 1 d s . (160)
In addition, the estimates of I 1 a I_{1}^{a} I 1 a and I 2 a I_{2}^{a} I 2 a are similar to (159) and (160) .
Substituting (91) , (143) , (159) and (160) into (152) with small ε 1 \varepsilon_{1} ε 1 and δ ~ \widetilde{\delta} δ derives
E ( t ) 2 ≲ E ( 0 ) 2 + ∫ 0 t E ( s ) ( ε 2 2 ⟨ s ⟩ − 3 / 2 + ε 2 3 ⟨ s ⟩ − 1 ) d s + ∫ 0 t E ( s ) 2 ( ε 1 ⟨ s ⟩ − 1 ln − 2 ( 2 + s ) + ε 2 2 ⟨ s ⟩ − 1 ) d s . E(t)^{2}\lesssim E(0)^{2}+\int_{0}^{t}E(s)\left(\varepsilon_{2}^{2}\langle s\rangle^{-3/2}+\varepsilon_{2}^{3}\langle s\rangle^{-1}\right)\,ds+\int_{0}^{t}E(s)^{2}\left(\varepsilon_{1}\langle s\rangle^{-1}\ln^{-2}(2+s)+\varepsilon_{2}^{2}\langle s\rangle^{-1}\right)\,ds. E ( t ) 2 ≲ E ( 0 ) 2 + ∫ 0 t E ( s ) ( ε 2 2 ⟨ s ⟩ − 3/2 + ε 2 3 ⟨ s ⟩ − 1 ) d s + ∫ 0 t E ( s ) 2 ( ε 1 ⟨ s ⟩ − 1 ln − 2 ( 2 + s ) + ε 2 2 ⟨ s ⟩ − 1 ) d s .
where E ( t ) = sup s ∈ [ 0 , t ] ∑ i ≤ 2 N ∥ ∂ ∂ t i v ( s ) ∥ L 2 ( K ) E(t)=\sup_{s\in[0,t]}\sum_{i\leq2N}\|\partial\partial^{i}_{t}v(s)\|_{L^{2}(\mathcal{K})} E ( t ) = sup s ∈ [ 0 , t ] ∑ i ≤ 2 N ∥ ∂ ∂ t i v ( s ) ∥ L 2 ( K ) . Thus, one has
E ( t ) ≲ E ( 0 ) + ε 2 2 + ε 2 3 ln ( 2 + t ) + ∫ 0 t E ( s ) ( ε 1 ⟨ s ⟩ − 1 ln − 2 ( 2 + s ) + ε 2 2 ⟨ s ⟩ − 1 ) d s . E(t)\lesssim E(0)+\varepsilon_{2}^{2}+\varepsilon_{2}^{3}\ln(2+t)+\int_{0}^{t}E(s)\left(\varepsilon_{1}\langle s\rangle^{-1}\ln^{-2}(2+s)+\varepsilon_{2}^{2}\langle s\rangle^{-1}\right)\,ds. E ( t ) ≲ E ( 0 ) + ε 2 2 + ε 2 3 ln ( 2 + t ) + ∫ 0 t E ( s ) ( ε 1 ⟨ s ⟩ − 1 ln − 2 ( 2 + s ) + ε 2 2 ⟨ s ⟩ − 1 ) d s .
This, together with (147) and Lemma 2.3, yields (148) .
Lemma 6.2. Under the assumptions of Theorem 1.1, let v v v be defined by (142) and suppose that (143) –(146) hold. Then we have that for t ∈ [ 0 , T ε ] t\in[0,T_{\varepsilon}] t ∈ [ 0 , T ε ] ,
∑ ∣ a ∣ ≤ 2 N ∥ ∂ ∂ t a v ( t ) ∥ L 2 ( K ) ≲ ε 1 . (161) \sum_{|a|\leq2N}\|\partial\partial^{a}_{t}v(t)\|_{L^{2}(\mathcal{K})}\lesssim\varepsilon_{1}. \tag*{(161)} ∣ a ∣ ≤ 2 N ∑ ∥ ∂ ∂ t a v ( t ) ∥ L 2 ( K ) ≲ ε 1 . (161)
Proof. Although the proof is somewhat analogous to that of Lemma 6.2 in [17 ] , we still provide the details for the sake of completeness.
Set E j ( t ) = ∑ k = 0 2 N − j ∥ ∂ t k v ( t ) ∥ H j ( K ) E_j(t)=\sum_{k=0}^{2N-j}\|\partial_t^k v(t)\|_{H^j(\mathcal{K})} E j ( t ) = ∑ k = 0 2 N − j ∥ ∂ t k v ( t ) ∥ H j ( K ) with 0 ≤ j ≤ 2 N 0\le j\le2N 0 ≤ j ≤ 2 N . Then one can find that for j ≥ 1 j\ge1 j ≥ 1 ,
E j ( t ) ≲ ∑ k = 0 2 N − j [ ∥ ∂ t k v ( t ) ∥ L 2 ( K ) + ∑ 1 ≤ ∣ a ∣ ≤ j ( ∥ ∂ t ∂ t k ∂ x a v ( t ) ∥ L 2 ( K ) + ∥ ∂ x ∂ t k ∂ x a v ( t ) ∥ L 2 ( K ) ) ] ≲ E 0 ( t ) + E j − 1 ( t ) + ∑ k = 0 2 N − j ∑ 2 ≤ ∣ a ∣ ≤ j + 1 ∥ ∂ t k ∂ x a v ( t ) ∥ L 2 ( K ) , (162) \begin{aligned}
E_j(t)&\lesssim\sum_{k=0}^{2N-j}\left[\|\partial_t^k v(t)\|_{L^2(\mathcal{K})}+\sum_{1\le|a|\le j}\left(\|\partial_t\partial_t^k\partial_x^a v(t)\|_{L^2(\mathcal{K})}+\|\partial_x\partial_t^k\partial_x^a v(t)\|_{L^2(\mathcal{K})}\right)\right]\\
&\lesssim E_0(t)+E_{j-1}(t)+\sum_{k=0}^{2N-j}\sum_{2\le|a|\le j+1}\|\partial_t^k\partial_x^a v(t)\|_{L^2(\mathcal{K})}, \tag*{(162)}
\end{aligned} E j ( t ) ≲ k = 0 ∑ 2 N − j ∥ ∂ t k v ( t ) ∥ L 2 ( K ) + 1 ≤ ∣ a ∣ ≤ j ∑ ( ∥ ∂ t ∂ t k ∂ x a v ( t ) ∥ L 2 ( K ) + ∥ ∂ x ∂ t k ∂ x a v ( t ) ∥ L 2 ( K ) ) ≲ E 0 ( t ) + E j − 1 ( t ) + k = 0 ∑ 2 N − j 2 ≤ ∣ a ∣ ≤ j + 1 ∑ ∥ ∂ t k ∂ x a v ( t ) ∥ L 2 ( K ) , (162)
where we have used the fact of
∑ k = 0 2 N − j ∑ 1 ≤ ∣ a ∣ ≤ j ∥ ∂ t ∂ t k ∂ x a v ( t ) ∥ L 2 ( K ) ≲ ∑ k = 0 2 N − j ∑ ∣ a ′ ∣ ≤ j − 1 ∥ ∂ x ∂ t k + 1 ∂ x a ′ v ( t ) ∥ L 2 ( K ) ≲ E j − 1 ( t ) . \sum_{k=0}^{2N-j}\sum_{1\le|a|\le j}\|\partial_t\partial_t^k\partial_x^a v(t)\|_{L^2(\mathcal{K})}\lesssim\sum_{k=0}^{2N-j}\sum_{|a'|\le j-1}\|\partial_x\partial_t^{k+1}\partial_x^{a'}v(t)\|_{L^2(\mathcal{K})}\lesssim E_{j-1}(t). k = 0 ∑ 2 N − j 1 ≤ ∣ a ∣ ≤ j ∑ ∥ ∂ t ∂ t k ∂ x a v ( t ) ∥ L 2 ( K ) ≲ k = 0 ∑ 2 N − j ∣ a ′ ∣ ≤ j − 1 ∑ ∥ ∂ x ∂ t k + 1 ∂ x a ′ v ( t ) ∥ L 2 ( K ) ≲ E j − 1 ( t ) .
For the last term in the second line of (162) , it can be deduced from the elliptic estimate (33) that
∥ ∂ t k ∂ x a v ∥ L 2 ( K ) ≲ ∥ Δ x ∂ t k v ∥ H ∣ a ∣ − 2 ( K ) + ∥ ∂ t k v ∥ H ∣ a ∣ − 1 ( K R + 1 ) ≲ ∥ ∂ t k + 2 v ∥ H ∣ a ∣ − 2 ( K ) + ∥ ∂ t k Q ( ∂ u , ∂ 2 u ) ∥ H ∣ a ∣ − 2 ( K ) + ∥ ∂ t k v ∥ L 2 ( K R + 1 ) + ∑ 1 ≤ ∣ b ∣ ≤ ∣ a ∣ − 1 ∥ ∂ t k ∂ x b v ∥ L 2 ( K R + 1 ) , (163) \begin{aligned}
\|\partial_t^k\partial_x^a v\|_{L^2(\mathcal{K})}
&\lesssim\|\Delta_x\partial_t^k v\|_{H^{|a|-2}(\mathcal{K})}+\|\partial_t^k v\|_{H^{|a|-1}(\mathcal{K}_{R+1})}\\
&\lesssim\|\partial_t^{k+2}v\|_{H^{|a|-2}(\mathcal{K})}+\|\partial_t^k Q(\partial u,\partial^2u)\|_{H^{|a|-2}(\mathcal{K})}\\
&\quad+\|\partial_t^k v\|_{L^2(\mathcal{K}_{R+1})}+\sum_{1\le|b|\le|a|-1}\|\partial_t^k\partial_x^b v\|_{L^2(\mathcal{K}_{R+1})}, \tag*{(163)}
\end{aligned} ∥ ∂ t k ∂ x a v ∥ L 2 ( K ) ≲ ∥ Δ x ∂ t k v ∥ H ∣ a ∣ − 2 ( K ) + ∥ ∂ t k v ∥ H ∣ a ∣ − 1 ( K R + 1 ) ≲ ∥ ∂ t k + 2 v ∥ H ∣ a ∣ − 2 ( K ) + ∥ ∂ t k Q ( ∂ u , ∂ 2 u ) ∥ H ∣ a ∣ − 2 ( K ) + ∥ ∂ t k v ∥ L 2 ( K R + 1 ) + 1 ≤ ∣ b ∣ ≤ ∣ a ∣ − 1 ∑ ∥ ∂ t k ∂ x b v ∥ L 2 ( K R + 1 ) , (163)
where Δ v = ∂ t 2 v − □ v = ∂ t 2 v − Q ( ∂ u , ∂ 2 u ) \Delta v=\partial_t^2v-\Box v=\partial_t^2v-Q(\partial u,\partial^2u) Δ v = ∂ t 2 v − □ v = ∂ t 2 v − Q ( ∂ u , ∂ 2 u ) has been used. In addition, by (91) , (143) with ∣ a ∣ ≤ j + 1 |a|\le j+1 ∣ a ∣ ≤ j + 1 , k ≤ 2 N − j k\le2N-j k ≤ 2 N − j , we have
∥ ∂ t k Q ( ∂ u , ∂ 2 u ) ∥ H ∣ a ∣ − 2 ( K ) ≲ ε 2 + ∑ l ≤ k ε 1 ( ∥ ∂ t l ∂ t v ∥ H ∣ a ∣ − 1 ( K ) + ∥ ∂ t l + 2 v ∥ H ∣ a ∣ − 2 ( K ) ) ≲ ε 2 + ε 1 [ E j ( t ) + E j − 1 ( t ) ] . (164) \begin{aligned}
\|\partial_t^kQ(\partial u,\partial^2u)\|_{H^{|a|-2}(\mathcal{K})}
&\lesssim\varepsilon^2+\sum_{l\le k}\varepsilon_1\left(\|\partial_t^l\partial_t v\|_{H^{|a|-1}(\mathcal{K})}+\|\partial_t^{l+2}v\|_{H^{|a|-2}(\mathcal{K})}\right)\\
&\lesssim\varepsilon^2+\varepsilon_1[E_j(t)+E_{j-1}(t)]. \tag*{(164)}
\end{aligned} ∥ ∂ t k Q ( ∂ u , ∂ 2 u ) ∥ H ∣ a ∣ − 2 ( K ) ≲ ε 2 + l ≤ k ∑ ε 1 ( ∥ ∂ t l ∂ t v ∥ H ∣ a ∣ − 1 ( K ) + ∥ ∂ t l + 2 v ∥ H ∣ a ∣ − 2 ( K ) ) ≲ ε 2 + ε 1 [ E j ( t ) + E j − 1 ( t )] . (164)
Collecting (162) -(164) yields that for j ≥ 1 j\ge1 j ≥ 1 ,
E j ( t ) ≲ ε 2 + E 0 ( t ) + E j − 1 ( t ) + ε 1 E j ( t ) + ∥ v ∥ L 2 ( K R + 1 ) . E_j(t)\lesssim\varepsilon^2+E_0(t)+E_{j-1}(t)+\varepsilon_1E_j(t)+\|v\|_{L^2(\mathcal{K}_{R+1})}. E j ( t ) ≲ ε 2 + E 0 ( t ) + E j − 1 ( t ) + ε 1 E j ( t ) + ∥ v ∥ L 2 ( K R + 1 ) .
Together with (145) , (148) and the smallness of ε 1 \varepsilon_1 ε 1 , this completes the proof of (161) .
Lemma 6.3. Under the assumptions of Theorem 1.1 , let v v v be defined by (142) and suppose that (143) -(146) hold. Then we it holds that for t ∈ [ 0 , T ε ] t\in[0,T_\varepsilon] t ∈ [ 0 , T ε ] ,
∑ ∣ a ∣ ≤ 2 N − 1 ∥ ∂ Z a v ( t ) ∥ L 2 ( K ) ≲ ε 1 ( 1 + t ) 1 / 2 . (165) \sum_{|a|\le2N-1}\|\partial Z^a v(t)\|_{L^2(\mathcal{K})}\lesssim\varepsilon_1(1+t)^{1/2}. \tag*{(165)} ∣ a ∣ ≤ 2 N − 1 ∑ ∥ ∂ Z a v ( t ) ∥ L 2 ( K ) ≲ ε 1 ( 1 + t ) 1/2 . (165)
Proof. Analogously to Lemma 6.1 , we can obtain that for any multi-index a ∈ N 0 4 a\in\mathbb{N}_0^4 a ∈ N 0 4 with ∣ a ∣ ≤ 2 N − 1 |a|\le2N-1 ∣ a ∣ ≤ 2 N − 1 ,
∑ ∣ a ∣ ≤ 2 N − 1 ∥ ∂ Z a v ( t ) ∥ L 2 ( K ) 2 ≲ ∑ ∣ a ∣ ≤ 2 N − 1 ∥ ∂ Z a v ( 0 ) ∥ L 2 ( K ) 2 + ∑ ∣ a ∣ ≤ 2 N − 1 ( ∣ B 1 a ∣ + ∣ B 2 a ∣ + ∣ B 3 a ∣ ) + ∑ ∣ a ∣ ≤ 2 N − 1 ∫ 0 t ∥ ∂ Z a v ( s ) ∥ L 2 ( K ) ( ε 2 ⟨ s ⟩ − 3 / 2 + ε 3 ⟨ s ⟩ − 1 ) d s + ∑ ∣ a ∣ ≤ 2 N − 1 ∫ 0 t ∥ ∂ Z a v ( s ) ∥ L 2 ( K ) ( ε 1 ⟨ s ⟩ − 1 ln − 2 ( 2 + s ) + ε 2 2 ⟨ s ⟩ − 1 ) d s , (166) \begin{aligned}
\sum_{|a|\le2N-1}\|\partial Z^a v(t)\|_{L^2(\mathcal{K})}^2
&\lesssim\sum_{|a|\le2N-1}\|\partial Z^a v(0)\|_{L^2(\mathcal{K})}^2+\sum_{|a|\le2N-1}(|B_1^a|+|B_2^a|+|B_3^a|)\\
&\quad+\sum_{|a|\le2N-1}\int_0^t\|\partial Z^a v(s)\|_{L^2(\mathcal{K})}(\varepsilon^2\langle s\rangle^{-3/2}+\varepsilon^3\langle s\rangle^{-1})\,ds\\
&\quad+\sum_{|a|\le2N-1}\int_0^t\|\partial Z^a v(s)\|_{L^2(\mathcal{K})}(\varepsilon_1\langle s\rangle^{-1}\ln^{-2}(2+s)+\varepsilon_2^2\langle s\rangle^{-1})\,ds, \tag*{(166)}
\end{aligned} ∣ a ∣ ≤ 2 N − 1 ∑ ∥ ∂ Z a v ( t ) ∥ L 2 ( K ) 2 ≲ ∣ a ∣ ≤ 2 N − 1 ∑ ∥ ∂ Z a v ( 0 ) ∥ L 2 ( K ) 2 + ∣ a ∣ ≤ 2 N − 1 ∑ ( ∣ B 1 a ∣ + ∣ B 2 a ∣ + ∣ B 3 a ∣ ) + ∣ a ∣ ≤ 2 N − 1 ∑ ∫ 0 t ∥ ∂ Z a v ( s ) ∥ L 2 ( K ) ( ε 2 ⟨ s ⟩ − 3/2 + ε 3 ⟨ s ⟩ − 1 ) d s + ∣ a ∣ ≤ 2 N − 1 ∑ ∫ 0 t ∥ ∂ Z a v ( s ) ∥ L 2 ( K ) ( ε 1 ⟨ s ⟩ − 1 ln − 2 ( 2 + s ) + ε 2 2 ⟨ s ⟩ − 1 ) d s , (166)
where
B 1 a = ∫ 0 t ∫ ∂ K e q ( n 1 ( x ) , n 2 ( x ) ) ⋅ ∇ Z a v ( s , x ) ∂ t Z a v ( s , x ) d σ d s \mathcal{B}_1^a=\int_0^t\int_{\partial K}e^q(n_1(x),n_2(x))\cdot\nabla Z^a v(s,x)\partial_t Z^a v(s,x)\,d\sigma ds B 1 a = ∫ 0 t ∫ ∂ K e q ( n 1 ( x ) , n 2 ( x )) ⋅ ∇ Z a v ( s , x ) ∂ t Z a v ( s , x ) d σ d s
B 2 a = ∫ 0 t ∫ ∂ K ∑ j = 1 2 ∑ β , γ = 0 2 Q j β γ n j ( x ) e q ∂ t Z a v ∂ β Z a v ∂ γ u d σ d s \mathcal{B}_2^a=\int_0^t\int_{\partial K}\sum_{j=1}^{2}\sum_{\beta,\gamma=0}^{2}Q^{j\beta\gamma}n_j(x)e^q\partial_t Z^a v\partial_\beta Z^a v\partial_\gamma u\,d\sigma ds B 2 a = ∫ 0 t ∫ ∂ K j = 1 ∑ 2 β , γ = 0 ∑ 2 Q j β γ n j ( x ) e q ∂ t Z a v ∂ β Z a v ∂ γ u d σ d s
B 3 a = ∫ 0 t ∫ ∂ K ∑ j = 1 2 ∑ β , γ , δ = 0 2 Q j β γ δ n j ( x ) e q ∂ t Z a v ∂ β Z a v ∂ γ u ∂ δ u d σ d s \mathcal{B}_3^a=\int_0^t\int_{\partial K}\sum_{j=1}^{2}\sum_{\beta,\gamma,\delta=0}^{2}Q^{j\beta\gamma\delta}n_j(x)e^q\partial_t Z^a v\partial_\beta Z^a v\partial_\gamma u\partial_\delta u\,d\sigma ds B 3 a = ∫ 0 t ∫ ∂ K j = 1 ∑ 2 β , γ , δ = 0 ∑ 2 Q j β γ δ n j ( x ) e q ∂ t Z a v ∂ β Z a v ∂ γ u ∂ δ u d σ d s
( n 1 ( x ) , n 2 ( x ) ) (n_1(x),n_2(x)) ( n 1 ( x ) , n 2 ( x )) is the unit outer normal of the boundary ∂ K \partial K ∂ K and d σ d\sigma d σ is the curve measure on the boundary ∂ K \partial K ∂ K . According to ∂ K ⊂ K ‾ 1 \partial K\subset\overline{K}_1 ∂ K ⊂ K 1 and the trace theorem, we arrive at
∣ B 1 a ∣ ≲ ∑ ∣ b ∣ ≤ ∣ a ∣ ∫ 0 t ∥ ( 1 − χ [ 1 , 2 ] ( x ) ) ∂ Z a v ∥ H 1 ( K ) 2 d s \left|\mathcal{B}_1^a\right|\lesssim\sum_{|b|\leq|a|}\int_0^t\left\|(1-\chi_{[1,2]}(x))\partial Z^a v\right\|_{H^1(\mathcal{K})}^2\,ds ∣ B 1 a ∣ ≲ ∣ b ∣ ≤ ∣ a ∣ ∑ ∫ 0 t ( 1 − χ [ 1 , 2 ] ( x )) ∂ Z a v H 1 ( K ) 2 d s
≲ ∑ ∣ b ∣ ≤ ∣ a ∣ + 1 ∫ 0 t ∥ ∂ ∂ b v ∥ L 2 ( K 2 ) 2 d s ≲ ε 1 2 ( 1 + t ) . (167) \lesssim\sum_{|b|\leq|a|+1}\int_0^t\left\|\partial\partial^b v\right\|_{L^2(\mathcal{K}_2)}^2\,ds\lesssim\varepsilon_1^2(1+t). \tag*{(167)} ≲ ∣ b ∣ ≤ ∣ a ∣ + 1 ∑ ∫ 0 t ∂ ∂ b v L 2 ( K 2 ) 2 d s ≲ ε 1 2 ( 1 + t ) . (167)
where (161) has been used. Similarly, B 2 a \mathcal{B}_2^a B 2 a and B 3 a \mathcal{B}_3^a B 3 a can be also treated. Substituting (167) into (143) yields
E ~ ( t ) 2 ≲ E ~ ( 0 ) 2 + ε 1 2 ( 1 + t ) + ∫ 0 t E ~ ( s ) ( ε 2 2 ⟨ s ⟩ − 3 / 2 + ε 3 3 ⟨ s ⟩ − 1 ) d s + ∫ 0 t E ~ ( s ) 2 ( ε 1 ⟨ s ⟩ − 1 ln − 2 ( 2 + s ) + ε 2 2 ⟨ s ⟩ − 1 ) d s . (168) \begin{aligned}
\widetilde{E}(t)^2\lesssim\widetilde{E}(0)^2+\varepsilon_1^2(1+t)+\int_0^t\widetilde{E}(s)(\varepsilon_2^2\langle s\rangle^{-3/2}+\varepsilon_3^3\langle s\rangle^{-1})\,ds \\
&+\int_0^t\widetilde{E}(s)^2(\varepsilon_1\langle s\rangle^{-1}\ln^{-2}(2+s)+\varepsilon_2^2\langle s\rangle^{-1})\,ds. \tag*{(168)}
\end{aligned} E ( t ) 2 ≲ E ( 0 ) 2 + ε 1 2 ( 1 + t ) + ∫ 0 t E ( s ) ( ε 2 2 ⟨ s ⟩ − 3/2 + ε 3 3 ⟨ s ⟩ − 1 ) d s + ∫ 0 t E ( s ) 2 ( ε 1 ⟨ s ⟩ − 1 ln − 2 ( 2 + s ) + ε 2 2 ⟨ s ⟩ − 1 ) d s . (168)
where E ~ ( t ) = ∑ ∣ a ∣ ≤ 2 N − 1 ∥ ∂ Z a v ( t ) ∥ L 2 ( K ) \widetilde{E}(t)=\sum_{|a|\leq2N-1}\|\partial Z^a v(t)\|_{L^2(\mathcal{K})} E ( t ) = ∑ ∣ a ∣ ≤ 2 N − 1 ∥ ∂ Z a v ( t ) ∥ L 2 ( K ) . If E ~ ( t ) ≥ ε 1 ( 1 + t ) 1 / 2 \widetilde{E}(t)\geq\varepsilon_1(1+t)^{1/2} E ( t ) ≥ ε 1 ( 1 + t ) 1/2 , otherwise (165) is directly obtained, (168) can be reduced to
E ~ ( t ) ≲ E ~ ( 0 ) + ε 1 ( 1 + t ) 1 / 2 + ∫ 0 t ( ε 2 2 ⟨ s ⟩ − 3 / 2 + ε 3 3 ⟨ s ⟩ − 1 ) d s + ∫ 0 t E ~ ( s ) ( ε 1 ⟨ s ⟩ − 1 ln − 2 ( 2 + s ) + ε 2 2 ⟨ s ⟩ − 1 ) d s \begin{aligned}
\widetilde{E}(t)\lesssim\widetilde{E}(0)+\varepsilon_1(1+t)^{1/2}+\int_0^t(\varepsilon_2^2\langle s\rangle^{-3/2}+\varepsilon_3^3\langle s\rangle^{-1})\,ds \\
&+\int_0^t\widetilde{E}(s)(\varepsilon_1\langle s\rangle^{-1}\ln^{-2}(2+s)+\varepsilon_2^2\langle s\rangle^{-1})\,ds
\end{aligned} E ( t ) ≲ E ( 0 ) + ε 1 ( 1 + t ) 1/2 + ∫ 0 t ( ε 2 2 ⟨ s ⟩ − 3/2 + ε 3 3 ⟨ s ⟩ − 1 ) d s + ∫ 0 t E ( s ) ( ε 1 ⟨ s ⟩ − 1 ln − 2 ( 2 + s ) + ε 2 2 ⟨ s ⟩ − 1 ) d s
≲ ε 1 ( 1 + t ) 1 / 2 + ∫ 0 t E ~ ( s ) ( ε 1 ⟨ s ⟩ − 1 ln − 2 ( 2 + s ) + ε 2 2 ⟨ s ⟩ − 1 ) d s . \lesssim\varepsilon_1(1+t)^{1/2}+\int_0^t\widetilde{E}(s)(\varepsilon_1\langle s\rangle^{-1}\ln^{-2}(2+s)+\varepsilon_2^2\langle s\rangle^{-1})\,ds. ≲ ε 1 ( 1 + t ) 1/2 + ∫ 0 t E ( s ) ( ε 1 ⟨ s ⟩ − 1 ln − 2 ( 2 + s ) + ε 2 2 ⟨ s ⟩ − 1 ) d s .
Thus, (165) is achieved by Lemma 2.3 and (147) .
Local energy decay estimates
It is noted that the improvement on the energy estimate in (165) mainly depends on the estimate of the boundary term in (167) , which can be enhanced by the local energy decay estimates.
Lemma 6.4. Under the assumptions of Theorem 1.1, let v v v be defined by (142) and suppose that (143) –(146) hold. Then we have that for t ∈ [ 0 , T ε ] t\in[0,T_\varepsilon] t ∈ [ 0 , T ε ] ,
∥ ∂ x ≤ 1 v ( t ) ∥ L 2 ( K R ) ≲ ε 1 2 ⟨ t ⟩ − 1 . (169) \|\partial_x^{\leq1}v(t)\|_{L^2(\mathcal{K}_R)}\lesssim\varepsilon_1^2\langle t\rangle^{-1}. \tag*{(169)} ∥ ∂ x ≤ 1 v ( t ) ∥ L 2 ( K R ) ≲ ε 1 2 ⟨ t ⟩ − 1 . (169)
Proof. The proof of (169) is analogous to that of (113) with the good structure in (54) . As in [17 ] , we firstly eliminate the quadratic Q 0 Q_0 Q 0 term in (54) with a = 0 a=0 a = 0 by introducing the good known
V 0 = v − ∑ μ = 0 2 1 2 C 0 , μ ( ω ) ∂ μ u (170) V_0=v-\sum_{\mu=0}^{2}\frac{1}{2}C^{0,\mu}(\omega)\partial_\mu u \tag*{(170)} V 0 = v − μ = 0 ∑ 2 2 1 C 0 , μ ( ω ) ∂ μ u (170)
and by using the identity for any functions f , g f,g f , g
□ ( f g ) = f □ g + g □ f + 2 Q 0 ( f , g ) , (171) \Box(fg)=f\Box g+g\Box f+2Q_0(f,g), \tag*{(171)} □ ( f g ) = f □ g + g □ f + 2 Q 0 ( f , g ) , (171)
which can be verified by direct computation. Thus, one has that V 0 ∣ [ 0 , ∞ ) × ∂ K = 0 V_0|_{[0,\infty)\times\partial K}=0 V 0 ∣ [ 0 , ∞ ) × ∂ K = 0 and
□ V 0 = ∂ θ ( C ( ω ) u ∂ θ 2 u ∣ x ∣ ) + ∑ μ = 0 2 ∂ μ ( C ( ω ) u Z ∂ u ∣ x ∣ ) + C ( ω ) u Z ∂ u ∣ x ∣ 2 + ∑ μ = 0 2 [ Δ , C 0 , μ ( ω ) ] u ∂ μ u + ∑ μ = 0 2 C ( ω ) ( ∂ μ u □ u + u □ ∂ μ u ) + 1 3 ∂ t [ ∑ α , β , γ , δ = 0 2 Q α β γ δ ω α ω β ω γ ω δ ( ∂ t u ) 3 ] + ∑ α , β , γ , δ = 0 2 Q α β γ δ [ ∂ α β 2 u ∂ γ u ∂ δ u − ω α ω β ω γ ω δ ∂ t 2 u ( ∂ t u ) 2 ] . (172) \begin{aligned}
\Box V_0={}&\partial_\theta\left(\frac{C(\omega)u\partial_\theta^2u}{|x|}\right)+\sum_{\mu=0}^{2}\partial_\mu\left(\frac{C(\omega)uZ\partial u}{|x|}\right)+\frac{C(\omega)uZ\partial u}{|x|^2}+\sum_{\mu=0}^{2}[\Delta,C^{0,\mu}(\omega)]u\partial_\mu u\\
&+\sum_{\mu=0}^{2}C(\omega)(\partial_\mu u\Box u+u\Box\partial_\mu u)+\frac{1}{3}\partial_t\left[\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\omega_\alpha\omega_\beta\omega_\gamma\omega_\delta(\partial_tu)^3\right]\\
&+\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\left[\partial_{\alpha\beta}^2u\partial_\gamma u\partial_\delta u-\omega_\alpha\omega_\beta\omega_\gamma\omega_\delta\partial_t^2u(\partial_tu)^2\right]. \tag*{(172)}
\end{aligned} □ V 0 = ∂ θ ( ∣ x ∣ C ( ω ) u ∂ θ 2 u ) + μ = 0 ∑ 2 ∂ μ ( ∣ x ∣ C ( ω ) u Z ∂ u ) + ∣ x ∣ 2 C ( ω ) u Z ∂ u + μ = 0 ∑ 2 [ Δ , C 0 , μ ( ω )] u ∂ μ u + μ = 0 ∑ 2 C ( ω ) ( ∂ μ u □ u + u □ ∂ μ u ) + 3 1 ∂ t α , β , γ , δ = 0 ∑ 2 Q α β γ δ ω α ω β ω γ ω δ ( ∂ t u ) 3 + α , β , γ , δ = 0 ∑ 2 Q α β γ δ [ ∂ α β 2 u ∂ γ u ∂ δ u − ω α ω β ω γ ω δ ∂ t 2 u ( ∂ t u ) 2 ] . (172)
By substituting the expressions of □ u \Box u □ u and □ ∂ μ u \Box\partial_\mu u □ ∂ μ u into (172) , we have
□ V 0 = ∂ t F 0 + ∑ Ξ = 0 , 1 , 2 , 3 , θ ∂ θ G Ξ + G 1 r + G 2 r , \Box V_0=\partial_tF^0+\sum_{\Xi=0,1,2,3,\theta}\partial_\theta G^\Xi+G_1^r+G_2^r, □ V 0 = ∂ t F 0 + Ξ = 0 , 1 , 2 , 3 , θ ∑ ∂ θ G Ξ + G 1 r + G 2 r ,
F 0 ( t , x ) = 1 3 η ( t , x ) ∑ α , β , γ , δ = 0 2 Q α β γ δ ω α ω β ω γ ω δ ( ∂ t u ) 3 F^0(t,x)=\frac{1}{3}\eta(t,x)\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\omega_\alpha\omega_\beta\omega_\gamma\omega_\delta(\partial_tu)^3 F 0 ( t , x ) = 3 1 η ( t , x ) α , β , γ , δ = 0 ∑ 2 Q α β γ δ ω α ω β ω γ ω δ ( ∂ t u ) 3
G α ( t , x ) = C ( ω ) u Z ∂ u ∣ x ∣ , α = 0 , 1 , 2 , G θ ( t , x ) : = C ( ω ) u ∂ 2 u ∣ x ∣ , G^\alpha(t,x)=\frac{C(\omega)uZ\partial u}{|x|},\qquad\alpha=0,1,2,\qquad G^\theta(t,x):=\frac{C(\omega)u\partial^2u}{|x|}, G α ( t , x ) = ∣ x ∣ C ( ω ) u Z ∂ u , α = 0 , 1 , 2 , G θ ( t , x ) := ∣ x ∣ C ( ω ) u ∂ 2 u ,
G 1 r ( t , x ) = 1 3 ∂ t [ ( 1 − η ( t , x ) ) ∑ α , β , γ , δ = 0 2 Q α β γ δ ω α ω β ω γ ω δ ( ∂ t u ) 3 ] + C ( ω ) u Z ∂ u ∣ x ∣ 2 + ∑ α , β , γ , μ = 0 2 C ( ω ) Q α β γ ( ∂ μ u ∂ α β 2 u ∂ γ u + u ∂ μ ( ∂ α β 2 u ∂ γ u ) ) + ∑ α , β , γ , δ = 0 2 Q α β γ δ [ ∂ α β 2 u ∂ γ u ∂ δ u − ω α ω β ω γ ω δ ∂ t 2 u ( ∂ t u ) 2 ] + 1 2 ∑ μ = 0 2 [ Δ , C 0 , μ ( ω ) ] u ∂ μ u , (173) \begin{aligned}
G_1^r(t,x)={}&\frac{1}{3}\partial_t\left[(1-\eta(t,x))\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\omega_\alpha\omega_\beta\omega_\gamma\omega_\delta(\partial_tu)^3\right]\\
&+\frac{C(\omega)uZ\partial u}{|x|^2}+\sum_{\alpha,\beta,\gamma,\mu=0}^{2}C(\omega)Q^{\alpha\beta\gamma}\left(\partial_\mu u\partial_{\alpha\beta}^2u\partial_\gamma u+u\partial_\mu(\partial_{\alpha\beta}^2u\partial_\gamma u)\right)\\
&+\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\left[\partial_{\alpha\beta}^2u\partial_\gamma u\partial_\delta u-\omega_\alpha\omega_\beta\omega_\gamma\omega_\delta\partial_t^2u(\partial_tu)^2\right]+\frac{1}{2}\sum_{\mu=0}^{2}[\Delta,C^{0,\mu}(\omega)]u\partial_\mu u, \tag*{(173)}
\end{aligned} G 1 r ( t , x ) = 3 1 ∂ t ( 1 − η ( t , x )) α , β , γ , δ = 0 ∑ 2 Q α β γ δ ω α ω β ω γ ω δ ( ∂ t u ) 3 + ∣ x ∣ 2 C ( ω ) u Z ∂ u + α , β , γ , μ = 0 ∑ 2 C ( ω ) Q α β γ ( ∂ μ u ∂ α β 2 u ∂ γ u + u ∂ μ ( ∂ α β 2 u ∂ γ u ) ) + α , β , γ , δ = 0 ∑ 2 Q α β γ δ [ ∂ α β 2 u ∂ γ u ∂ δ u − ω α ω β ω γ ω δ ∂ t 2 u ( ∂ t u ) 2 ] + 2 1 μ = 0 ∑ 2 [ Δ , C 0 , μ ( ω )] u ∂ μ u , (173)
G 2 r ( t , x ) = ∑ α , β , γ , δ , μ = 0 2 C ( ω ) Q α β γ δ ( ∂ μ u ∂ α β 2 u ∂ γ u ∂ δ u + u ∂ μ ( ∂ α β 2 u ∂ γ u ∂ δ u ) ) , G_2^r(t,x)=\sum_{\alpha,\beta,\gamma,\delta,\mu=0}^{2}C(\omega)Q^{\alpha\beta\gamma\delta}\left(\partial_\mu u\partial_{\alpha\beta}^2u\partial_\gamma u\partial_\delta u+u\partial_\mu(\partial_{\alpha\beta}^2u\partial_\gamma u\partial_\delta u)\right), G 2 r ( t , x ) = α , β , γ , δ , μ = 0 ∑ 2 C ( ω ) Q α β γ δ ( ∂ μ u ∂ α β 2 u ∂ γ u ∂ δ u + u ∂ μ ( ∂ α β 2 u ∂ γ u ∂ δ u ) ) ,
where
η ( t , x ) = χ [ 9 10 , 19 20 ] ( ∣ x ∣ t ) (174) \eta(t,x)=\chi_{\left[\frac{9}{10},\frac{19}{20}\right]}\left(\frac{|x|}{t}\right) \tag*{(174)} η ( t , x ) = χ [ 10 9 , 20 19 ] ( t ∣ x ∣ ) (174)
is a conic cutoff function and is supported in { ∣ x ∣ ≥ 9 t 10 } \{ |x| \ge\frac{9t}{10} \} { ∣ x ∣ ≥ 10 9 t } . For the first term in the last line of G 1 r ( t , x ) G^r_1(t,x) G 1 r ( t , x ) in (173) , similarly to the proof of (47) , one has that for any ∣ a ∣ ≤ 1 |a| \le1 ∣ a ∣ ≤ 1 ,
∣ Z a ( ∂ α β 2 u ∂ γ u ∂ δ u − ω α ω β ω γ ω δ ∂ t 2 u ( ∂ t u ) 2 ) ∣ ≲ ∑ b + c + d ≤ a ( ∣ ∂ ˉ Z b u ∣ ∣ ∂ Z c u ∣ ∣ ∂ Z d u ∣ + ∣ ∂ 2 Z b u ∣ ∣ ∂ ˉ Z c u ∣ ∣ ∂ Z d u ∣ + 1 ∣ x ∣ ∣ ∂ 2 Z b u ∣ ∣ ∂ Z c u ∣ ∣ ∂ Z d u ∣ ) (175) \left|Z^a\left(\partial^2_{\alpha\beta}u\partial_\gamma u\partial_\delta u-\omega_\alpha\omega_\beta\omega_\gamma\omega_\delta\partial_t^2u(\partial_tu)^2\right)\right|
\lesssim\sum_{b+c+d\le a}\left(|\bar{\partial}Z^b u||\partial Z^c u||\partial Z^d u|+|\partial^2Z^b u||\bar{\partial}Z^c u||\partial Z^d u|+\frac{1}{|x|}|\partial^2Z^b u||\partial Z^c u||\partial Z^d u|\right) \tag*{(175)} Z a ( ∂ α β 2 u ∂ γ u ∂ δ u − ω α ω β ω γ ω δ ∂ t 2 u ( ∂ t u ) 2 ) ≲ b + c + d ≤ a ∑ ( ∣ ∂ ˉ Z b u ∣∣ ∂ Z c u ∣∣ ∂ Z d u ∣ + ∣ ∂ 2 Z b u ∣∣ ∂ ˉ Z c u ∣∣ ∂ Z d u ∣ + ∣ x ∣ 1 ∣ ∂ 2 Z b u ∣∣ ∂ Z c u ∣∣ ∂ Z d u ∣ ) (175)
≲ ε 1 3 ⟨ y ⟩ − 3 / 2 ⟨ s + ∣ y ∣ ⟩ δ 0 − 1 ⟨ s − ∣ y ∣ ⟩ 2 δ 0 − 2 , \lesssim\varepsilon_1^3\langle y\rangle^{-3/2}\langle s+|y|\rangle^{\delta_0-1}\langle s-|y|\rangle^{2\delta_0-2}, ≲ ε 1 3 ⟨ y ⟩ − 3/2 ⟨ s + ∣ y ∣ ⟩ δ 0 − 1 ⟨ s − ∣ y ∣ ⟩ 2 δ 0 − 2 ,
where (91) , (143) and (144) have been used. For the last term in G 1 r ( t , x ) G^r_1(t,x) G 1 r ( t , x ) , by Δ = ∂ r 2 + 1 r ∂ r + ∂ θ 2 r 2 \Delta=\partial_r^2+\frac{1}{r}\partial_r+\frac{\partial_\theta^2}{r^2} Δ = ∂ r 2 + r 1 ∂ r + r 2 ∂ θ 2 and ∂ r ω = 0 \partial_r\omega=0 ∂ r ω = 0 , we have that for any function f f f ,
∣ [ Δ , C 0 , μ ( ω ) ] f ∣ ≲ ∣ x ∣ − 2 ∣ Z ≤ 1 f ∣ . (176) |[\Delta,C^{0,\mu}(\omega)]f|\lesssim|x|^{-2}|Z^{\leq1}f|. \tag*{(176)} ∣ [ Δ , C 0 , μ ( ω )] f ∣ ≲ ∣ x ∣ − 2 ∣ Z ≤ 1 f ∣. (176)
On the other hand, it can be concluded from (4) , (47) , (91) , (143) –(145) with N ≥ 53 N\geq53 N ≥ 53 that
∑ ∣ a ∣ ≤ 5 ∣ ∂ a F 0 ( s , y ) ∣ ≲ ε 1 3 ⟨ s + ∣ y ∣ ⟩ − 3 / 2 ⟨ s − ∣ y ∣ ⟩ − 2 , \sum_{|a|\leq5}|\partial^a F^0(s,y)|\lesssim\varepsilon_1^3\langle s+|y|\rangle^{-3/2}\langle s-|y|\rangle^{-2}, ∣ a ∣ ≤ 5 ∑ ∣ ∂ a F 0 ( s , y ) ∣ ≲ ε 1 3 ⟨ s + ∣ y ∣ ⟩ − 3/2 ⟨ s − ∣ y ∣ ⟩ − 2 ,
∑ ∣ a ∣ ≤ 2 ∣ Z a G Ξ ( s , y ) ∣ ≲ ε 1 2 ⟨ y ⟩ − 3 / 2 ⟨ s + ∣ y ∣ ⟩ − 1 / 2 ⟨ s − ∣ y ∣ ⟩ 4 δ 0 − 3 / 2 , Ξ = 0 , 1 , 2 , θ , \sum_{|a|\leq2}|Z^a G^\Xi(s,y)|\lesssim\varepsilon_1^2\langle y\rangle^{-3/2}\langle s+|y|\rangle^{-1/2}\langle s-|y|\rangle^{4\delta_0-3/2},\qquad\Xi=0,1,2,\theta, ∣ a ∣ ≤ 2 ∑ ∣ Z a G Ξ ( s , y ) ∣ ≲ ε 1 2 ⟨ y ⟩ − 3/2 ⟨ s + ∣ y ∣ ⟩ − 1/2 ⟨ s − ∣ y ∣ ⟩ 4 δ 0 − 3/2 , Ξ = 0 , 1 , 2 , θ ,
∑ ∣ a ∣ ≤ 1 ∣ ∂ a G 1 r ( s , y ) ∣ ≲ ε 1 3 ⟨ y ⟩ − 3 / 2 ⟨ s + ∣ y ∣ ⟩ 3 δ 0 − 3 + ε 1 2 ⟨ y ⟩ − 5 / 2 ⟨ s + ∣ y ∣ ⟩ − 1 / 2 ⟨ s − ∣ y ∣ ⟩ 4 δ 0 − 3 / 2 + ε 1 3 ⟨ y ⟩ − 1 ⟨ s + ∣ y ∣ ⟩ δ 0 − 3 / 2 ⟨ s − ∣ y ∣ ⟩ 4 δ 0 − 3 / 2 , (177) \begin{aligned}
\sum_{|a|\leq1}|\partial^a G^r_1(s,y)|
&\lesssim\varepsilon_1^3\langle y\rangle^{-3/2}\langle s+|y|\rangle^{3\delta_0-3}
+\varepsilon_1^2\langle y\rangle^{-5/2}\langle s+|y|\rangle^{-1/2}\langle s-|y|\rangle^{4\delta_0-3/2}\\
&\quad+\varepsilon_1^3\langle y\rangle^{-1}\langle s+|y|\rangle^{\delta_0-3/2}\langle s-|y|\rangle^{4\delta_0-3/2}, \tag*{(177)}
\end{aligned} ∣ a ∣ ≤ 1 ∑ ∣ ∂ a G 1 r ( s , y ) ∣ ≲ ε 1 3 ⟨ y ⟩ − 3/2 ⟨ s + ∣ y ∣ ⟩ 3 δ 0 − 3 + ε 1 2 ⟨ y ⟩ − 5/2 ⟨ s + ∣ y ∣ ⟩ − 1/2 ⟨ s − ∣ y ∣ ⟩ 4 δ 0 − 3/2 + ε 1 3 ⟨ y ⟩ − 1 ⟨ s + ∣ y ∣ ⟩ δ 0 − 3/2 ⟨ s − ∣ y ∣ ⟩ 4 δ 0 − 3/2 , (177)
∑ ∣ a ∣ ≤ 1 ∣ ∂ a G 2 r ( s , y ) ∣ ≲ ε 1 4 ⟨ y ⟩ − 2 ⟨ s − ∣ y ∣ ⟩ 6 δ 0 − 7 / 2 . \sum_{|a|\leq1}|\partial^a G^r_2(s,y)|\lesssim\varepsilon_1^4\langle y\rangle^{-2}\langle s-|y|\rangle^{6\delta_0-7/2}. ∣ a ∣ ≤ 1 ∑ ∣ ∂ a G 2 r ( s , y ) ∣ ≲ ε 1 4 ⟨ y ⟩ − 2 ⟨ s − ∣ y ∣ ⟩ 6 δ 0 − 7/2 .
Let w 1 b w^b_1 w 1 b and w 2 b w^b_2 w 2 b be the solutions of
{ □ w 1 b = ( 1 − χ [ 2 , 3 ] ( x ) ) [ ∂ t F 0 + ∑ Ξ = 0 , 1 , 2 , θ ∂ Ξ G Ξ + G 1 r + G 2 r ] − ∑ i = 1 2 G i ( ∂ i χ [ 2 , 3 ] ( x ) ) , □ w 2 b = ∂ t ( χ [ 2 , 3 ] ( x ) F 0 ) + ∑ Ξ = 0 , 1 , 2 , θ ∂ Ξ ( χ [ 2 , 3 ] ( x ) G Ξ ) + χ [ 2 , 3 ] ( x ) ( G 1 r + G 2 r ) , w 1 b ∣ ∂ K = w 2 b ∣ ∂ K = 0 , ( w 1 b , ∂ t w 1 b ) ( 0 , x ) = ( V 0 , ∂ t V 0 ) ( 0 , x ) , ( w 2 b , ∂ t w 2 b ) ( 0 , x ) = ( 0 , 0 ) , (178) \begin{cases}
\Box w^b_1=(1-\chi_{[2,3]}(x))\left[\partial_tF^0+\sum_{\Xi=0,1,2,\theta}\partial_\Xi G^\Xi+G^r_1+G^r_2\right]-\sum_{i=1}^{2}G^i(\partial_i\chi_{[2,3]}(x)),\\
\Box w^b_2=\partial_t(\chi_{[2,3]}(x)F^0)+\sum_{\Xi=0,1,2,\theta}\partial_\Xi(\chi_{[2,3]}(x)G^\Xi)+\chi_{[2,3]}(x)(G^r_1+G^r_2),\\
w^b_1|_{\partial K}=w^b_2|_{\partial K}=0,\\
(w^b_1,\partial_tw^b_1)(0,x)=(V_0,\partial_tV_0)(0,x),\qquad(w^b_2,\partial_tw^b_2)(0,x)=(0,0), \tag*{(178)}
\end{cases} ⎩ ⎨ ⎧ □ w 1 b = ( 1 − χ [ 2 , 3 ] ( x )) [ ∂ t F 0 + ∑ Ξ = 0 , 1 , 2 , θ ∂ Ξ G Ξ + G 1 r + G 2 r ] − ∑ i = 1 2 G i ( ∂ i χ [ 2 , 3 ] ( x )) , □ w 2 b = ∂ t ( χ [ 2 , 3 ] ( x ) F 0 ) + ∑ Ξ = 0 , 1 , 2 , θ ∂ Ξ ( χ [ 2 , 3 ] ( x ) G Ξ ) + χ [ 2 , 3 ] ( x ) ( G 1 r + G 2 r ) , w 1 b ∣ ∂ K = w 2 b ∣ ∂ K = 0 , ( w 1 b , ∂ t w 1 b ) ( 0 , x ) = ( V 0 , ∂ t V 0 ) ( 0 , x ) , ( w 2 b , ∂ t w 2 b ) ( 0 , x ) = ( 0 , 0 ) , (178)
respectively. It follows from (173) and (178) that V 0 = w 1 b + w 2 b V_0=w^b_1+w^b_2 V 0 = w 1 b + w 2 b . By (32) with m = 1 m=1 m = 1 and η = 1 \eta=1 η = 1 , one has
⟨ t ⟩ ∥ ∂ x ≤ 1 w 1 b ( t , x ) ∥ L 2 ( K R ) ≲ ε 2 + sup s ∈ [ 0 , t ] ⟨ s ⟩ ∥ □ w 1 b ( s ) ∥ L 2 ( K 3 ) ≲ ε 1 2 , (179) \langle t\rangle\|\partial_x^{\leq1}w^b_1(t,x)\|_{L^2(K_R)}
\lesssim\varepsilon^2+\sup_{s\in[0,t]}\langle s\rangle\|\Box w^b_1(s)\|_{L^2(K_3)}
\lesssim\varepsilon_1^2, \tag*{(179)} ⟨ t ⟩ ∥ ∂ x ≤ 1 w 1 b ( t , x ) ∥ L 2 ( K R ) ≲ ε 2 + s ∈ [ 0 , t ] sup ⟨ s ⟩ ∥ □ w 1 b ( s ) ∥ L 2 ( K 3 ) ≲ ε 1 2 , (179)
where (8) , (170) are (177) are used.
Next, we treat the estimates of w 2 b w^b_2 w 2 b . Let w c w^c w c , w Ξ c w^c_\Xi w Ξ c (Ξ = 0 , 1 , 2 , θ \Xi=0,1,2,\theta Ξ = 0 , 1 , 2 , θ ), w r , 1 c w^c_{r,1} w r , 1 c , w r , 2 c w^c_{r,2} w r , 2 c be the solutions of the Cauchy problem
□ w c = χ [ 2 , 3 ] ( x ) F 0 , □ w Ξ c = χ [ 2 , 3 ] ( x ) G Ξ ( Ξ = 0 , 1 , 2 , θ ) , □ w r , l c = χ [ 2 , 3 ] ( x ) G l r , l = 1 , 2 , (180) \Box w^c=\chi_{[2,3]}(x)F^0,\qquad\Box w^c_\Xi=\chi_{[2,3]}(x)G^\Xi\quad(\Xi=0,1,2,\theta),\qquad\Box w^c_{r,l}=\chi_{[2,3]}(x)G^r_l,\ l=1,2, \tag*{(180)} □ w c = χ [ 2 , 3 ] ( x ) F 0 , □ w Ξ c = χ [ 2 , 3 ] ( x ) G Ξ ( Ξ = 0 , 1 , 2 , θ ) , □ w r , l c = χ [ 2 , 3 ] ( x ) G l r , l = 1 , 2 , (180)
with zero initial data, respectively. Then v 2 b = w 2 b − χ [ 1 , 2 ] ( x ) ( ∂ t w c + ∑ Ξ = 0 , 1 , 2 , θ ∂ Ξ w Ξ c + ∑ l = 1 2 w r , l c ) v_2^b=w_2^b-\chi_{[1,2]}(x)(\partial_t w^c+\sum_{\Xi=0,1,2,\theta}\partial_{\Xi}w_\Xi^c+\sum_{l=1}^{2}w_{r,l}^c) v 2 b = w 2 b − χ [ 1 , 2 ] ( x ) ( ∂ t w c + ∑ Ξ = 0 , 1 , 2 , θ ∂ Ξ w Ξ c + ∑ l = 1 2 w r , l c ) is the solution of the IBVP
{ □ v 2 b = [ Δ , χ [ 1 , 2 ] ] ( ∂ t w c + ∑ Ξ = 0 , 1 , 2 , θ ∂ Ξ w Ξ c + ∑ l = 1 2 w r , l c ) , v 2 b ∣ ∂ K = 0 , ( v 2 b , ∂ t v 2 b ) ( 0 , x ) = ( 0 , χ [ 2 , 3 ] ( x ) ( F 0 ( 0 , x ) + G 0 ( 0 , x ) ) ) . \begin{cases}
\square v_2^b=[\Delta,\chi_{[1,2]}](\partial_t w^c+\sum_{\Xi=0,1,2,\theta}\partial_{\Xi}w_\Xi^c+\sum_{l=1}^{2}w_{r,l}^c),\\
v_2^b|_{\partial K}=0,\\
(v_2^b,\partial_t v_2^b)(0,x)=(0,\chi_{[2,3]}(x)(F^0(0,x)+G^0(0,x))).
\end{cases} ⎩ ⎨ ⎧ □ v 2 b = [ Δ , χ [ 1 , 2 ] ] ( ∂ t w c + ∑ Ξ = 0 , 1 , 2 , θ ∂ Ξ w Ξ c + ∑ l = 1 2 w r , l c ) , v 2 b ∣ ∂ K = 0 , ( v 2 b , ∂ t v 2 b ) ( 0 , x ) = ( 0 , χ [ 2 , 3 ] ( x ) ( F 0 ( 0 , x ) + G 0 ( 0 , x ))) .
Since □ v 2 b \square v_2^b □ v 2 b has compact support in the space variable, and the estimate of v 2 b v_2^b v 2 b is analogous to that of w 1 b w_1^b w 1 b in (179) , one then has
⟨ t ⟩ ∥ ∂ x ≤ 1 v 2 b ( t , x ) ∥ L 2 ( K R ) ≲ ε 1 2 + sup s ∈ [ 0 , t ] ⟨ s ⟩ ( ∥ ∂ x ≤ 1 ∂ t w c ( s ) ∥ L ∞ ( ∣ x ∣ ≤ 2 ) + ∑ Ξ = 0 , 1 , 2 , θ ∥ ∂ x ≤ 1 ∂ w Ξ c ( s ) ∥ L ∞ ( ∣ x ∣ ≤ 2 ) + ∑ l = 1 2 ∥ ∂ x ≤ 1 w r , l c ( s ) ∥ L ∞ ( ∣ x ∣ ≤ 2 ) ) . (181) \langle t\rangle\|\partial_x^{\leq1}v_2^b(t,x)\|_{L^2(K_R)}\lesssim\varepsilon_1^2+\sup_{s\in[0,t]}\langle s\rangle\left(\|\partial_x^{\leq1}\partial_t w^c(s)\|_{L^\infty(|x|\leq2)}+\sum_{\Xi=0,1,2,\theta}\|\partial_x^{\leq1}\partial w_\Xi^c(s)\|_{L^\infty(|x|\leq2)}+\sum_{l=1}^{2}\|\partial_x^{\leq1}w_{r,l}^c(s)\|_{L^\infty(|x|\leq2)}\right). \tag*{(181)} ⟨ t ⟩ ∥ ∂ x ≤ 1 v 2 b ( t , x ) ∥ L 2 ( K R ) ≲ ε 1 2 + s ∈ [ 0 , t ] sup ⟨ s ⟩ ∥ ∂ x ≤ 1 ∂ t w c ( s ) ∥ L ∞ ( ∣ x ∣ ≤ 2 ) + Ξ = 0 , 1 , 2 , θ ∑ ∥ ∂ x ≤ 1 ∂ w Ξ c ( s ) ∥ L ∞ ( ∣ x ∣ ≤ 2 ) + l = 1 ∑ 2 ∥ ∂ x ≤ 1 w r , l c ( s ) ∥ L ∞ ( ∣ x ∣ ≤ 2 ) . (181)
Applying (41) , (42) , (2.13) and Lemma 4.2 with $\mu=\nu=\delta_0$ to $w_{r,1}^c$, $w_{r,2}^c$, $w_\Xi^c$ and $w^c$ in (6.39)`, respectively, yields
⟨ t + ∣ x ∣ ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ 1 / 2 ∣ ∂ x ≤ 1 w r , 1 c ( t , x ) ∣ ≲ ε 1 2 , \langle t+|x|\rangle^{1/2}\langle t-|x|\rangle^{1/2}|\partial_x^{\leq1}w_{r,1}^c(t,x)|\lesssim\varepsilon_1^2, ⟨ t + ∣ x ∣ ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ 1/2 ∣ ∂ x ≤ 1 w r , 1 c ( t , x ) ∣ ≲ ε 1 2 ,
⟨ t + ∣ x ∣ ⟩ 1 / 2 ⟨ t − ∣ x ∣ ⟩ 1 / 2 ∣ ∂ x ≤ 1 w r , 2 c ( t , x ) ∣ ≲ ε 1 2 ln ( 2 + t ) , \langle t+|x|\rangle^{1/2}\langle t-|x|\rangle^{1/2}|\partial_x^{\leq1}w_{r,2}^c(t,x)|\lesssim\varepsilon_1^2\ln(2+t), ⟨ t + ∣ x ∣ ⟩ 1/2 ⟨ t − ∣ x ∣ ⟩ 1/2 ∣ ∂ x ≤ 1 w r , 2 c ( t , x ) ∣ ≲ ε 1 2 ln ( 2 + t ) ,
⟨ t − ∣ x ∣ ⟩ 1 + δ 0 ∣ ∂ x ≤ 1 ∂ w Ξ c ( t , x ) ∣ ≲ ε 1 2 , Ξ = 0 , 1 , 2 , θ , \langle t-|x|\rangle^{1+\delta_0}|\partial_x^{\leq1}\partial w_\Xi^c(t,x)|\lesssim\varepsilon_1^2,\qquad\Xi=0,1,2,\theta, ⟨ t − ∣ x ∣ ⟩ 1 + δ 0 ∣ ∂ x ≤ 1 ∂ w Ξ c ( t , x ) ∣ ≲ ε 1 2 , Ξ = 0 , 1 , 2 , θ ,
⟨ t ⟩ ∥ ∂ x ≤ 1 ∂ t w c ( t , x ) ∥ L ∞ ( ∣ x ∣ ≤ t / 10 ) ≲ ε 1 3 . (182) \langle t\rangle\|\partial_x^{\leq1}\partial_t w^c(t,x)\|_{L^\infty(|x|\leq t/10)}\lesssim\varepsilon_1^3. \tag*{(182)} ⟨ t ⟩ ∥ ∂ x ≤ 1 ∂ t w c ( t , x ) ∥ L ∞ ( ∣ x ∣ ≤ t /10 ) ≲ ε 1 3 . (182)
where (177) is also used. Collecting (170) , (179) , (181) and (182) with V 0 = w 1 b + w 2 b V_0=w_1^b+w_2^b V 0 = w 1 b + w 2 b , w 2 b = v 2 b + χ [ 1 , 2 ] ( x ) ( ∂ t w c + ∑ Ξ = 0 , 1 , 2 , θ ∂ Ξ w Ξ c + ∑ l = 1 2 w r , l c ) w_2^b=v_2^b+\chi_{[1,2]}(x)(\partial_t w^c+\sum_{\Xi=0,1,2,\theta}\partial_{\Xi}w_\Xi^c+\sum_{l=1}^{2}w_{r,l}^c) w 2 b = v 2 b + χ [ 1 , 2 ] ( x ) ( ∂ t w c + ∑ Ξ = 0 , 1 , 2 , θ ∂ Ξ w Ξ c + ∑ l = 1 2 w r , l c ) , (91) , (145) and (147) finishes the proof of (169) .
Lemma 6.5. Under the assumptions of Theorem 1.1, let v v v be defined by (142) and suppose that (143) -(146) hold. Then we have that for t ∈ [ 0 , T ε ] t\in[0,T_\varepsilon] t ∈ [ 0 , T ε ] ,
∑ ∣ a ∣ ≤ 2 N − 5 ∥ ∂ a v ( t ) ∥ L 2 ( K R ) ≲ ε 1 2 ⟨ t ⟩ 4 δ 0 − 1 / 2 . (183) \sum_{|a|\leq2N-5}\|\partial^a v(t)\|_{L^2(K_R)}\lesssim\varepsilon_1^2\langle t\rangle^{4\delta_0-1/2}. \tag*{(183)} ∣ a ∣ ≤ 2 N − 5 ∑ ∥ ∂ a v ( t ) ∥ L 2 ( K R ) ≲ ε 1 2 ⟨ t ⟩ 4 δ 0 − 1/2 . (183)
Proof. Choosing Z a = ∂ t i Z^a=\partial_t^i Z a = ∂ t i in (53) yields
□ ∂ t i v = ∑ j + k ≤ i { ∑ μ , ν = 0 2 ∣ x ∣ − 1 ( C ( ω ) ∂ μ ν 2 ∂ t j u Ω ∂ t k u + C ( ω ) Ω ∂ t j u ∂ ν ∂ t k u ) + ∑ μ = 0 2 [ C i j k 12 , μ ( ω ) Q 12 ( ∂ μ ∂ t j u , ∂ t k u ) + C i j k 0 , μ ( ω ) Q 0 ( ∂ μ ∂ t j u , ∂ t k u ) + C ( ω ) □ ∂ t j u ∂ μ ∂ t k u ] } + ∂ t i [ ∑ α , β , γ , δ = 0 2 Q α β γ δ ∂ α β 2 u ∂ γ u ∂ δ u ] . \square\partial_t^i v=\sum_{j+k\leq i}\left\{\sum_{\mu,\nu=0}^{2}|x|^{-1}\left(C(\omega)\partial_{\mu\nu}^2\partial_t^j u\,\Omega\partial_t^k u+C(\omega)\Omega\partial_t^j u\,\partial_\nu\partial_t^k u\right)
+\sum_{\mu=0}^{2}\left[C_{ijk}^{12,\mu}(\omega)Q_{12}(\partial_\mu\partial_t^j u,\partial_t^k u)+C_{ijk}^{0,\mu}(\omega)Q_0(\partial_\mu\partial_t^j u,\partial_t^k u)+C(\omega)\square\partial_t^j u\,\partial_\mu\partial_t^k u\right]\right\}
+\partial_t^i\left[\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\partial_{\alpha\beta}^2u\partial_\gamma u\partial_\delta u\right]. □ ∂ t i v = j + k ≤ i ∑ { μ , ν = 0 ∑ 2 ∣ x ∣ − 1 ( C ( ω ) ∂ μν 2 ∂ t j u Ω ∂ t k u + C ( ω ) Ω ∂ t j u ∂ ν ∂ t k u ) + μ = 0 ∑ 2 [ C ij k 12 , μ ( ω ) Q 12 ( ∂ μ ∂ t j u , ∂ t k u ) + C ij k 0 , μ ( ω ) Q 0 ( ∂ μ ∂ t j u , ∂ t k u ) + C ( ω ) □ ∂ t j u ∂ μ ∂ t k u ] } + ∂ t i α , β , γ , δ = 0 ∑ 2 Q α β γ δ ∂ α β 2 u ∂ γ u ∂ δ u .
As in the definition (170) , introducing the good unknown
V i = ∂ t i v − 1 2 ∑ μ = 0 2 ∑ j + k ≤ i C i j k 0 , μ ( ω ) ∂ μ ∂ t j u ∂ t k u . (184) V_i=\partial_t^i v-\frac{1}{2}\sum_{\mu=0}^{2}\sum_{j+k\leq i}C_{ijk}^{0,\mu}(\omega)\partial_\mu\partial_t^j u\partial_t^k u. \tag*{(184)} V i = ∂ t i v − 2 1 μ = 0 ∑ 2 j + k ≤ i ∑ C ij k 0 , μ ( ω ) ∂ μ ∂ t j u ∂ t k u . (184)
Then one has V i ∣ [ 0 , ∞ ) × ∂ K = 0 V_i|_{[0,\infty)\times\partial K}=0 V i ∣ [ 0 , ∞ ) × ∂ K = 0 and
□ V i = ∑ j + k ≤ i { ∑ μ , ν = 0 2 ∣ x ∣ − 1 ( C ( ω ) ∂ μ 2 ∂ t j u Ω ∂ t k u + C ( ω ) Ω ∂ μ ∂ t j u ∂ ν ∂ t k u ) + ∑ μ = 0 2 [ 1 2 ∑ μ = 0 2 [ Δ , C i j k 0 , μ ( ω ) ] ∂ μ ∂ t j u ∂ t k u + C i j k 12 , μ Q 12 ( ∂ μ ∂ t j u , ∂ t k u ) ] + C ( ω ) ∂ μ ∂ t j u □ ∂ t k u + C ( ω ) ∂ t k u □ ∂ μ ∂ t j u } + ∂ t i [ ∑ α , β , γ , δ = 0 2 Q α β γ δ ∂ α β 2 u ∂ γ u ∂ δ u ] . (185) \begin{aligned}
\square V_i={}&\sum_{j+k\leq i}\Biggl\{\sum_{\mu,\nu=0}^{2}|x|^{-1}\Bigl(C(\omega)\partial_\mu^2\partial_t^j u\Omega\partial_t^k u+C(\omega)\Omega\partial_\mu\partial_t^j u\partial_\nu\partial_t^k u\Bigr)\\
&+\sum_{\mu=0}^{2}\Biggl[\frac{1}{2}\sum_{\mu=0}^{2}[\Delta,C_{ijk}^{0,\mu}(\omega)]\partial_\mu\partial_t^j u\partial_t^k u+C_{ijk}^{12,\mu}Q_{12}(\partial_\mu\partial_t^j u,\partial_t^k u)\Biggr]\\
&+C(\omega)\partial_\mu\partial_t^j u\square\partial_t^k u+C(\omega)\partial_t^k u\square\partial_\mu\partial_t^j u\Biggr\}+\partial_t^i\Biggl[\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\partial_{\alpha\beta}^2u\partial_\gamma u\partial_\delta u\Biggr]. \tag*{(185)}
\end{aligned} □ V i = j + k ≤ i ∑ { μ , ν = 0 ∑ 2 ∣ x ∣ − 1 ( C ( ω ) ∂ μ 2 ∂ t j u Ω ∂ t k u + C ( ω ) Ω ∂ μ ∂ t j u ∂ ν ∂ t k u ) + μ = 0 ∑ 2 [ 2 1 μ = 0 ∑ 2 [ Δ , C ij k 0 , μ ( ω )] ∂ μ ∂ t j u ∂ t k u + C ij k 12 , μ Q 12 ( ∂ μ ∂ t j u , ∂ t k u ) ] + C ( ω ) ∂ μ ∂ t j u □ ∂ t k u + C ( ω ) ∂ t k u □ ∂ μ ∂ t j u } + ∂ t i [ α , β , γ , δ = 0 ∑ 2 Q α β γ δ ∂ α β 2 u ∂ γ u ∂ δ u ] . (185)
Applying (113) to (185) with F 1 = □ V i F_1=\square V_i F 1 = □ V i , F 0 ≡ 0 F_0\equiv0 F 0 ≡ 0 , i ≤ 2 N − 7 i\leq2N-7 i ≤ 2 N − 7 , μ = δ 0 \mu=\delta_0 μ = δ 0 and (8) yields
⟨ t ⟩ ∥ ∂ x ≤ 1 ∂ t V i ∥ L 2 ( K R 1 ) ≲ ε 2 + ∑ ∣ a ∣ ≤ 1 sup s ∈ [ 0 , t ] ⟨ s ⟩ ∥ ∂ a F 1 ( s ) ∥ L 2 ( K 3 ) + ∑ ∣ a ∣ ≤ 2 sup ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 ⟨ y ⟩ 1 / 2 W 1 + δ 0 , 1 ( s , y ) ∣ Z a F 1 ( s , y ) ∣ . (186) \langle t\rangle\|\partial_x^{\leq1}\partial_tV_i\|_{L^2(K_{R_1})}\lesssim\varepsilon^2+\sum_{|a|\leq1}\sup_{s\in[0,t]}\langle s\rangle\|\partial^aF_1(s)\|_{L^2(K_3)}
+\sum_{|a|\leq2}\sup_{(s,y)\in[0,t]\times\mathbb{R}^2\setminus K_2}\langle y\rangle^{1/2}\mathcal{W}_{1+\delta_0,1}(s,y)|Z^aF_1(s,y)|. \tag*{(186)} ⟨ t ⟩ ∥ ∂ x ≤ 1 ∂ t V i ∥ L 2 ( K R 1 ) ≲ ε 2 + ∣ a ∣ ≤ 1 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ ∥ ∂ a F 1 ( s ) ∥ L 2 ( K 3 ) + ∣ a ∣ ≤ 2 ∑ ( s , y ) ∈ [ 0 , t ] × R 2 ∖ K 2 sup ⟨ y ⟩ 1/2 W 1 + δ 0 , 1 ( s , y ) ∣ Z a F 1 ( s , y ) ∣. (186)
Note that (34) and (165) lead to
∑ ∣ a ∣ ≤ 2 N − 3 ∣ ∂ Z a v ∣ ≲ ε 1 ⟨ x ⟩ − 1 / 2 ( 1 + t ) 1 / 2 . (187) \sum_{|a|\leq2N-3}|\partial Z^a v|\lesssim\varepsilon_1\langle x\rangle^{-1/2}(1+t)^{1/2}. \tag*{(187)} ∣ a ∣ ≤ 2 N − 3 ∑ ∣ ∂ Z a v ∣ ≲ ε 1 ⟨ x ⟩ − 1/2 ( 1 + t ) 1/2 . (187)
It follows from (48) , (91) , (143) , (145) , (176) and (187) that
∑ ∣ a ∣ ≤ 2 ∣ Z a F 1 ( s , y ) ∣ ≲ ε 1 2 ⟨ y ⟩ − 3 / 2 ⟨ s + ∣ y ∣ ⟩ − 1 / 2 ⟨ s − ∣ y ∣ ⟩ 3 δ 0 − 1 / 2 ( 1 + s ) 1 / 2 + ε 1 3 ⟨ y ⟩ − 1 ⟨ s + ∣ y ∣ ⟩ − 1 / 2 ⟨ s − ∣ y ∣ ⟩ 4 δ 0 − 3 / 2 ( 1 + s ) 1 / 2 . (188) \sum_{|a|\leq2}|Z^aF_1(s,y)|\lesssim\varepsilon_1^2\langle y\rangle^{-3/2}\langle s+|y|\rangle^{-1/2}\langle s-|y|\rangle^{3\delta_0-1/2}(1+s)^{1/2}
+\varepsilon_1^3\langle y\rangle^{-1}\langle s+|y|\rangle^{-1/2}\langle s-|y|\rangle^{4\delta_0-3/2}(1+s)^{1/2}. \tag*{(188)} ∣ a ∣ ≤ 2 ∑ ∣ Z a F 1 ( s , y ) ∣ ≲ ε 1 2 ⟨ y ⟩ − 3/2 ⟨ s + ∣ y ∣ ⟩ − 1/2 ⟨ s − ∣ y ∣ ⟩ 3 δ 0 − 1/2 ( 1 + s ) 1/2 + ε 1 3 ⟨ y ⟩ − 1 ⟨ s + ∣ y ∣ ⟩ − 1/2 ⟨ s − ∣ y ∣ ⟩ 4 δ 0 − 3/2 ( 1 + s ) 1/2 . (188)
On the other hand, by (91) , (145) and (161) , one has
∑ ∣ a ∣ ≤ 1 ∥ ∂ a F 1 ( s ) ∥ L 2 ( K 3 ) ≲ ε 1 2 ⟨ s ⟩ 3 δ 0 − 1 . (189) \sum_{|a|\leq1}\|\partial^aF_1(s)\|_{L^2(K_3)}\lesssim\varepsilon_1^2\langle s\rangle^{3\delta_0-1}. \tag*{(189)} ∣ a ∣ ≤ 1 ∑ ∥ ∂ a F 1 ( s ) ∥ L 2 ( K 3 ) ≲ ε 1 2 ⟨ s ⟩ 3 δ 0 − 1 . (189)
Therefore, collecting (91) , (145) , (161) , (169) , (186) , (188) and (189) with the definition (184) yields
∑ i ≤ 2 N − 7 ∥ ∂ ≤ 1 ∂ t i v ∥ L 2 ( K R 1 ) ≲ ε 1 2 ⟨ t ⟩ 4 δ 0 − 1 / 2 . (190) \sum_{i\leq2N-7}\|\partial^{\leq1}\partial_t^i v\|_{L^2(K_{R_1})}\lesssim\varepsilon_1^2\langle t\rangle^{4\delta_0-1/2}. \tag*{(190)} i ≤ 2 N − 7 ∑ ∥ ∂ ≤ 1 ∂ t i v ∥ L 2 ( K R 1 ) ≲ ε 1 2 ⟨ t ⟩ 4 δ 0 − 1/2 . (190)
The remaining proof is similar to that of Lemma 6.6 in [17 ] . For j = 0 , 1 , 2 , ⋯ , 2 N − 5 j=0,1,2,\cdots,2N-5 j = 0 , 1 , 2 , ⋯ , 2 N − 5 , denote
E j l o c ( t ) = ∑ k ≤ j ∥ ∂ t k ∂ x 2 N − 5 − j v ∥ L 2 ( K R + j ) . E_j^{\mathrm{loc}}(t)=\sum_{k\leq j}\|\partial_t^k\partial_x^{2N-5-j}v\|_{L^2(K_{R+j})}. E j loc ( t ) = k ≤ j ∑ ∥ ∂ t k ∂ x 2 N − 5 − j v ∥ L 2 ( K R + j ) .
Then (169) and (190) imply
E 2 N − 5 l o c ( t ) + E 2 N − 6 l o c ( t ) ≲ ε 1 2 ⟨ t ⟩ 4 δ 0 − 1 / 2 . (191) E_{2N-5}^{\mathrm{loc}}(t)+E_{2N-6}^{\mathrm{loc}}(t)\lesssim\varepsilon_1^2\langle t\rangle^{4\delta_0-1/2}. \tag*{(191)} E 2 N − 5 loc ( t ) + E 2 N − 6 loc ( t ) ≲ ε 1 2 ⟨ t ⟩ 4 δ 0 − 1/2 . (191)
When j ≤ 2 N − 7 j \le2N-7 j ≤ 2 N − 7 , we have ∂ x 2 N − 5 − j = ∂ x 2 ∂ x 2 N − 7 − j \partial_x^{2N-5-j}=\partial_x^2\partial_x^{2N-7-j} ∂ x 2 N − 5 − j = ∂ x 2 ∂ x 2 N − 7 − j . Thus, one can apply the elliptic estimate (33) to ( 1 − χ [ R + j , R + j + 1 ] ) ∂ t k v (1-\chi_{[R+j,R+j+1]})\partial_t^k v ( 1 − χ [ R + j , R + j + 1 ] ) ∂ t k v to obtain
E j l o c ( t ) ≲ ∑ k ≤ j ∥ ∂ x 2 N − 5 − j ( 1 − χ [ R + j , R + j + 1 ] ) ∂ t k v ∥ L 2 ( K ) ≲ ∑ k ≤ j [ ∥ Δ ( 1 − χ [ R + j , R + j + 1 ] ) ∂ t k v ∥ H 2 N − 7 − j ( K ) + ∥ ( 1 − χ [ R + j , R + j + 1 ] ) ∂ t k v ∥ H 2 N − 6 − j ( K ) ] ≲ ∑ k ≤ j ∥ ( 1 − χ [ R + j , R + j + 1 ] ) Δ ∂ t k v ∥ H 2 N − 7 − j ( K ) + ∑ j + 1 ≤ l ≤ 2 N − 5 E l l o c ( t ) ≲ ∑ k ≤ j [ ∥ □ ∂ t k v ∥ H 2 N − 7 − j ( K R + j + 1 ) + ∥ ∂ t k + 2 v ∥ H 2 N − 7 − j ( K R + j + 1 ) ] + ∑ j + 1 ≤ l ≤ 2 N − 5 E l l o c ( t ) , (192) \begin{aligned}
E_j^{\mathrm{loc}}(t) &\lesssim\sum_{k\le j}\|\partial_x^{2N-5-j}(1-\chi_{[R+j,R+j+1]})\partial_t^k v\|_{L^2(\mathcal{K})}\\
&\lesssim\sum_{k\le j}\bigl[\|\Delta(1-\chi_{[R+j,R+j+1]})\partial_t^k v\|_{H^{2N-7-j}(\mathcal{K})}\\
&\qquad+\|(1-\chi_{[R+j,R+j+1]})\partial_t^k v\|_{H^{2N-6-j}(\mathcal{K})}\bigr]\\
&\lesssim\sum_{k\le j}\|(1-\chi_{[R+j,R+j+1]})\Delta\partial_t^k v\|_{H^{2N-7-j}(\mathcal{K})}+\sum_{j+1\le l\le2N-5}E_l^{\mathrm{loc}}(t)\\
&\lesssim\sum_{k\le j}\bigl[\|\square\partial_t^k v\|_{H^{2N-7-j}(\mathcal{K}_{R+j+1})}+\|\partial_t^{k+2}v\|_{H^{2N-7-j}(\mathcal{K}_{R+j+1})}\bigr]\\
&\qquad+\sum_{j+1\le l\le2N-5}E_l^{\mathrm{loc}}(t), \tag*{(192)}
\end{aligned} E j loc ( t ) ≲ k ≤ j ∑ ∥ ∂ x 2 N − 5 − j ( 1 − χ [ R + j , R + j + 1 ] ) ∂ t k v ∥ L 2 ( K ) ≲ k ≤ j ∑ [ ∥Δ ( 1 − χ [ R + j , R + j + 1 ] ) ∂ t k v ∥ H 2 N − 7 − j ( K ) + ∥ ( 1 − χ [ R + j , R + j + 1 ] ) ∂ t k v ∥ H 2 N − 6 − j ( K ) ] ≲ k ≤ j ∑ ∥ ( 1 − χ [ R + j , R + j + 1 ] ) Δ ∂ t k v ∥ H 2 N − 7 − j ( K ) + j + 1 ≤ l ≤ 2 N − 5 ∑ E l loc ( t ) ≲ k ≤ j ∑ [ ∥ □ ∂ t k v ∥ H 2 N − 7 − j ( K R + j + 1 ) + ∥ ∂ t k + 2 v ∥ H 2 N − 7 − j ( K R + j + 1 ) ] + j + 1 ≤ l ≤ 2 N − 5 ∑ E l loc ( t ) , (192)
where the fact of Δ = ∂ t 2 − □ \Delta=\partial_t^2-\square Δ = ∂ t 2 − □ has been used. Analogously to (189) , it can be concluded from u = v + w h o m u=v+w_{\mathrm{hom}} u = v + w hom , (91) , (6.2) and (6.20) that for j ≤ 2 N − 7 j\le2N-7 j ≤ 2 N − 7 ,
∑ k ≤ j ∥ ∂ t k v ∥ H 2 N − 7 − j ( K R + j + 1 ) ≲ ∑ k ≤ j ∥ ∂ t k Q ( ∂ u , ∂ 2 u ) ∥ H 2 N − 7 − j ( K R + j + 1 ) ≲ ε 1 2 ( 1 + t ) δ 0 − 1 . (193) \sum_{k\le j}\|\partial_t^k v\|_{H^{2N-7-j}(\mathcal{K}_{R+j+1})}\lesssim\sum_{k\le j}\|\partial_t^k Q(\partial u,\partial^2u)\|_{H^{2N-7-j}(\mathcal{K}_{R+j+1})}\lesssim\varepsilon_1^2(1+t)^{\delta_0-1}. \tag*{(193)} k ≤ j ∑ ∥ ∂ t k v ∥ H 2 N − 7 − j ( K R + j + 1 ) ≲ k ≤ j ∑ ∥ ∂ t k Q ( ∂ u , ∂ 2 u ) ∥ H 2 N − 7 − j ( K R + j + 1 ) ≲ ε 1 2 ( 1 + t ) δ 0 − 1 . (193)
Thus, (192) and (193) leads to that for j ≤ 2 N − 7 j\le2N-7 j ≤ 2 N − 7 ,
E j l o c ( t ) ≲ ε 1 2 ( 1 + t ) δ 0 − 1 + ∑ j + 1 ≤ l ≤ 2 N − 5 E l l o c ( t ) . E_j^{\mathrm{loc}}(t)\lesssim\varepsilon_1^2(1+t)^{\delta_0-1}+\sum_{j+1\le l\le2N-5}E_l^{\mathrm{loc}}(t). E j loc ( t ) ≲ ε 1 2 ( 1 + t ) δ 0 − 1 + j + 1 ≤ l ≤ 2 N − 5 ∑ E l loc ( t ) .
This, together with (191) , yields (183) .
Improved energy estimates
With the local energy decay estimate (183) , the energy estimates including all the vector fields in (165) will be improved.
Lemma 6.6. Under the assumptions of Theorem 1.1 , let v v v be defined by (142) and suppose that (143) –(146) hold. Then we have that for t ∈ [ 0 , T ε ] t\in[0,T_\varepsilon] t ∈ [ 0 , T ε ] ,
∑ ∣ a ∣ ≤ 2 N − 7 ∥ ∂ Z a v ( t ) ∥ L 2 ( K ) ≲ ε 1 ( 1 + t ) 4 δ 0 . (194) \sum_{|a|\le2N-7}\|\partial Z^a v(t)\|_{L^2(\mathcal{K})}\lesssim\varepsilon_1(1+t)^{4\delta_0}. \tag*{(194)} ∣ a ∣ ≤ 2 N − 7 ∑ ∥ ∂ Z a v ( t ) ∥ L 2 ( K ) ≲ ε 1 ( 1 + t ) 4 δ 0 . (194) Proof. Analogously to Lemma 6.3, one can obtain the following better estimates on the boundary terms B 1 a , B 2 a , B 3 a B_1^a,B_2^a,B_3^a B 1 a , B 2 a , B 3 a for ∣ a ∣ ≤ 2 N − 7 |a|\le2N-7 ∣ a ∣ ≤ 2 N − 7 ,
∑ l = 1 3 ∣ B l a ∣ ≲ ∑ ∣ b ∣ ≤ ∣ a ∣ + 1 ∫ 0 t ∥ ∂ ∂ b v ∥ L 2 ( K 2 ) 2 d s ≲ ε 1 2 ∫ 0 t ( 1 + s ) 8 δ 0 − 1 d s ≲ ε 1 2 ( 1 + t ) 8 δ 0 , (195) \sum_{l=1}^{3}|B_l^a|\lesssim\sum_{|b|\le|a|+1}\int_0^t\|\partial\partial^b v\|_{L^2(\mathcal{K}_2)}^2\,ds\lesssim\varepsilon_1^2\int_0^t(1+s)^{8\delta_0-1}\,ds\lesssim\varepsilon_1^2(1+t)^{8\delta_0}, \tag*{(195)} l = 1 ∑ 3 ∣ B l a ∣ ≲ ∣ b ∣ ≤ ∣ a ∣ + 1 ∑ ∫ 0 t ∥ ∂ ∂ b v ∥ L 2 ( K 2 ) 2 d s ≲ ε 1 2 ∫ 0 t ( 1 + s ) 8 δ 0 − 1 d s ≲ ε 1 2 ( 1 + t ) 8 δ 0 , (195) where (183) has been used. Then (194) is achieved.
In addition, the local energy decay estimate (183) can be further improved as follows.
Lemma 6.7. Under the assumptions of Theorem 1.1 , let v v v be defined by (142) and suppose that (143) –(146) hold. Then it holds that for t ∈ [ 0 , T ε ] t \in[0,T_\varepsilon] t ∈ [ 0 , T ε ] ,
∑ ∣ a ∣ ≤ 2 N − 11 ∥ ∂ Z a v ( t ) ∥ L 2 ( K R ) ≲ ε 1 2 ⟨ t ⟩ 8 δ 0 − 1 . (196) \sum_{|a|\leq2N-11}\|\partial Z^a v(t)\|_{L^2(K_R)} \lesssim\varepsilon_1^2\langle t\rangle^{8\delta_0-1}. \tag*{(196)} ∣ a ∣ ≤ 2 N − 11 ∑ ∥ ∂ Z a v ( t ) ∥ L 2 ( K R ) ≲ ε 1 2 ⟨ t ⟩ 8 δ 0 − 1 . (196)
Proof. The proof is analogous to that of (183) . At first, (34) and (194) imply
∑ ∣ a ∣ ≤ 2 N − 9 ∣ ∂ Z a v ∣ ≲ ε 1 ⟨ x ⟩ − 1 / 2 ( 1 + t ) 4 δ 0 . (197) \sum_{|a|\leq2N-9}|\partial Z^a v| \lesssim\varepsilon_1\langle x\rangle^{-1/2}(1+t)^{4\delta_0}. \tag*{(197)} ∣ a ∣ ≤ 2 N − 9 ∑ ∣ ∂ Z a v ∣ ≲ ε 1 ⟨ x ⟩ − 1/2 ( 1 + t ) 4 δ 0 . (197)
Then, with (197) instead of (187) , (188) can be improved to
∑ ∣ a ∣ ≤ 2 ∣ Z a F 1 ( s , y ) ∣ ≲ ε 1 2 ⟨ y ⟩ − 3 / 2 ⟨ s + ∣ y ∣ ⟩ − 1 / 2 ⟨ s − ∣ y ∣ ⟩ 3 δ 0 − 1 / 2 ( 1 + s ) 4 δ 0 + ε 1 3 ⟨ y ⟩ − 1 ⟨ s + ∣ y ∣ ⟩ − 1 / 2 ⟨ s − ∣ y ∣ ⟩ 4 δ 0 − 3 / 2 ( 1 + s ) 4 δ 0 . (198) \begin{aligned}
\sum_{|a|\leq2}|Z^aF_1(s,y)| &\lesssim\varepsilon_1^2\langle y\rangle^{-3/2}\langle s+|y|\rangle^{-1/2}\langle s-|y|\rangle^{3\delta_0-1/2}(1+s)^{4\delta_0}\\
&\quad+\varepsilon_1^3\langle y\rangle^{-1}\langle s+|y|\rangle^{-1/2}\langle s-|y|\rangle^{4\delta_0-3/2}(1+s)^{4\delta_0}. \tag*{(198)}
\end{aligned} ∣ a ∣ ≤ 2 ∑ ∣ Z a F 1 ( s , y ) ∣ ≲ ε 1 2 ⟨ y ⟩ − 3/2 ⟨ s + ∣ y ∣ ⟩ − 1/2 ⟨ s − ∣ y ∣ ⟩ 3 δ 0 − 1/2 ( 1 + s ) 4 δ 0 + ε 1 3 ⟨ y ⟩ − 1 ⟨ s + ∣ y ∣ ⟩ − 1/2 ⟨ s − ∣ y ∣ ⟩ 4 δ 0 − 3/2 ( 1 + s ) 4 δ 0 . (198)
Substituting (189) and (198) into (186) together with (91) , (145) , (183) , (184) and i ≤ 2 N − 13 i\leq2N-13 i ≤ 2 N − 13 , yields
∑ i ≤ 2 N − 13 ∥ ∂ ≤ 1 ∂ t i v ∥ L 2 ( K R 1 ) ≲ ε 1 2 ⟨ t ⟩ 8 δ 0 − 1 . (199) \sum_{i\leq2N-13}\|\partial^{\leq1}\partial_t^i v\|_{L^2(K_{R_1})}\lesssim\varepsilon_1^2\langle t\rangle^{8\delta_0-1}. \tag*{(199)} i ≤ 2 N − 13 ∑ ∥ ∂ ≤ 1 ∂ t i v ∥ L 2 ( K R 1 ) ≲ ε 1 2 ⟨ t ⟩ 8 δ 0 − 1 . (199)
Therefore, (196) is achieved by (199) with an analogous argument as in the proof of (183) . □
Lemma 6.8. Under the assumptions of Theorem 1.1 , let v v v be defined by (142) and suppose that (143) –(146) hold. Then we have that for t ∈ [ 0 , T ε ] t\in[0,T_\varepsilon] t ∈ [ 0 , T ε ] ,
∑ ∣ a ∣ ≤ 2 N − 13 ∥ ∂ Z a v ( t ) ∥ L 2 ( K ) ≲ ε 1 . (200) \sum_{|a|\leq2N-13}\|\partial Z^a v(t)\|_{L^2(K)}\lesssim\varepsilon_1. \tag*{(200)} ∣ a ∣ ≤ 2 N − 13 ∑ ∥ ∂ Z a v ( t ) ∥ L 2 ( K ) ≲ ε 1 . (200)
Proof. Since the proof is similar to that of (194) with (196) instead of (183) , we omit the details here. □
Improved pointwise estimates and proof of Theorem 1.1
In this section, we will improve the pointwise estimates (143) –(146) and close the bootstrap assumptions in Section 6. Although these estimates are analogous to the ones in Section 7 of [17 ] with some minor modifications, we still give the details for readers’ convenience.
Decay estimates on the good derivatives of solutions
Lemma 7.1. Under the assumptions of Theorem 1.1 , let v v v be defined by (142) and suppose that (143) –(146) hold. Then it holds that for t ∈ [ 0 , T ε ] t\in[0,T_\varepsilon] t ∈ [ 0 , T ε ] ,
∑ ∣ a ∣ ≤ 2 N − 16 ∣ Z a v ∣ ≲ ε 1 ⟨ t + ∣ x ∣ ⟩ 6 δ 0 . (201) \sum_{|a|\leq2N-16}|Z^a v|\lesssim\varepsilon_1\langle t+|x|\rangle^{6\delta_0}. \tag*{(201)} ∣ a ∣ ≤ 2 N − 16 ∑ ∣ Z a v ∣ ≲ ε 1 ⟨ t + ∣ x ∣ ⟩ 6 δ 0 . (201)
Proof. Until now, we have not obtained the estimates of Z a u Z^{a}u Z a u with higher order derivatives by use of (53) . To achieve this, we will employ the primitive equation (49) . In order to preserve the zero boundary conditions, set
Z ~ = χ [ 1 / 2 , 1 ] ( x ) Z , Z ∈ { ∂ , Ω } . (202) \widetilde{Z}=\chi_{[1/2,1]}(x)Z,\qquad Z\in\{\partial,\Omega\}. \tag*{(202)} Z = χ [ 1/2 , 1 ] ( x ) Z , Z ∈ { ∂ , Ω } . (202)
Applying (119) to □ Z ~ a v = □ ( Z ~ a − Z a ) v + □ Z a v \Box\widetilde{Z}^{a}v=\Box(\widetilde{Z}^{a}-Z^{a})v+\Box Z^{a}v □ Z a v = □ ( Z a − Z a ) v + □ Z a v with F 0 ( t , x ) ≡ 0 F_{0}(t,x)\equiv0 F 0 ( t , x ) ≡ 0 , μ = ν = δ 0 / 2 \mu=\nu=\delta_{0}/2 μ = ν = δ 0 /2 and ∣ a ∣ ≤ 2 N − 15 |a|\leq2N-15 ∣ a ∣ ≤ 2 N − 15 yields
⟨ t + ∣ x ∣ ⟩ 1 / 2 ∣ Z ~ a v ∣ ≲ ε 1 2 + ∑ ∣ b ∣ ≤ 4 sup s ∈ [ 0 , t ] ⟨ s ⟩ 1 / 2 + δ 0 ∥ ∂ b □ ( Z ~ a − Z a ) v ( s ) ∥ L 2 ( K 4 ) + ∑ ∣ b ∣ ≤ 5 sup ( s , y ) ∈ [ 0 , t ] × K ⟨ y ⟩ 1 / 2 W 1 + δ 0 , 1 ( s , y ) ∣ ∂ b □ Z a v ( s , y ) ∣ . (203) \begin{aligned}
\langle t+|x|\rangle^{1/2}|\widetilde{Z}^{a}v|
&\lesssim\varepsilon_{1}^{2}+\sum_{|b|\leq4}\sup_{s\in[0,t]}\langle s\rangle^{1/2+\delta_{0}}\|\partial^{b}\Box(\widetilde{Z}^{a}-Z^{a})v(s)\|_{L^{2}(K_{4})}\\
&\quad+\sum_{|b|\leq5}\sup_{(s,y)\in[0,t]\times K}\langle y\rangle^{1/2}\mathcal{W}_{1+\delta_{0},1}(s,y)|\partial^{b}\Box Z^{a}v(s,y)|. \tag*{(203)}
\end{aligned} ⟨ t + ∣ x ∣ ⟩ 1/2 ∣ Z a v ∣ ≲ ε 1 2 + ∣ b ∣ ≤ 4 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ 1/2 + δ 0 ∥ ∂ b □ ( Z a − Z a ) v ( s ) ∥ L 2 ( K 4 ) + ∣ b ∣ ≤ 5 ∑ ( s , y ) ∈ [ 0 , t ] × K sup ⟨ y ⟩ 1/2 W 1 + δ 0 , 1 ( s , y ) ∣ ∂ b □ Z a v ( s , y ) ∣. (203)
where we have used (8) and the fact of Z ~ = Z \widetilde{Z}=Z Z = Z in the region ∣ x ∣ ≥ 1 |x|\geq1 ∣ x ∣ ≥ 1 . In addition, it can be deduced from (183) that
∑ ∣ a ∣ ≤ 2 N − 15 ∑ ∣ b ∣ ≤ 4 sup s ∈ [ 0 , t ] ⟨ s ⟩ 1 / 2 + δ 0 ∥ ∂ b □ ( Z ~ a − Z a ) v ( s ) ∥ L 2 ( K 4 ) ≲ ε 1 2 ⟨ t ⟩ 5 δ 0 . (204) \sum_{|a|\leq2N-15}\sum_{|b|\leq4}\sup_{s\in[0,t]}\langle s\rangle^{1/2+\delta_{0}}\|\partial^{b}\Box(\widetilde{Z}^{a}-Z^{a})v(s)\|_{L^{2}(K_{4})}\lesssim\varepsilon_{1}^{2}\langle t\rangle^{5\delta_{0}}. \tag*{(204)} ∣ a ∣ ≤ 2 N − 15 ∑ ∣ b ∣ ≤ 4 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ 1/2 + δ 0 ∥ ∂ b □ ( Z a − Z a ) v ( s ) ∥ L 2 ( K 4 ) ≲ ε 1 2 ⟨ t ⟩ 5 δ 0 . (204)
For the second line of (203) , one can conclude from (49) , (91) , (143) and (197) that
∑ ∣ a ∣ ≤ 2 N − 15 ∑ ∣ b ∣ ≤ 5 ∣ Z b □ Z a v ( s , y ) ∣ ≲ ε 1 2 ⟨ y ⟩ − 1 ⟨ s − ∣ y ∣ ⟩ δ 0 − 1 ( 1 + s ) 4 δ 0 . (205) \sum_{|a|\leq2N-15}\sum_{|b|\leq5}|Z^{b}\Box Z^{a}v(s,y)|\lesssim\varepsilon_{1}^{2}\langle y\rangle^{-1}\langle s-|y|\rangle^{\delta_{0}-1}(1+s)^{4\delta_{0}}. \tag*{(205)} ∣ a ∣ ≤ 2 N − 15 ∑ ∣ b ∣ ≤ 5 ∑ ∣ Z b □ Z a v ( s , y ) ∣ ≲ ε 1 2 ⟨ y ⟩ − 1 ⟨ s − ∣ y ∣ ⟩ δ 0 − 1 ( 1 + s ) 4 δ 0 . (205)
Substituting (204) and (205) into (203) leads to
∑ ∣ a ∣ ≤ 2 N − 15 ∣ Z ~ a v ∣ ≲ ε 1 ⟨ t + ∣ x ∣ ⟩ 6 δ 0 . \sum_{|a|\leq2N-15}|\widetilde{Z}^{a}v|\lesssim\varepsilon_{1}\langle t+|x|\rangle^{6\delta_{0}}. ∣ a ∣ ≤ 2 N − 15 ∑ ∣ Z a v ∣ ≲ ε 1 ⟨ t + ∣ x ∣ ⟩ 6 δ 0 .
This, together with (183) and the Sobolev embedding, completes the proof of (201) .
Lemma 7.2. Under the assumptions of Theorem 1.1 , let v v v be defined by (142) and suppose that (143) -*(146) hold. Then we have that for t ∈ [ 0 , T ε ] t\in[0,T_{\varepsilon}] t ∈ [ 0 , T ε ] and ∣ x ∣ ≥ 1 + t / 2 |x|\geq1+t/2 ∣ x ∣ ≥ 1 + t /2 ,
∑ ∣ a ∣ ≤ 2 N − 17 ∣ ∂ Z ‾ a v ∣ ≲ ε 1 ⟨ t + ∣ x ∣ ⟩ 6 δ 0 − 1 . (206) \sum_{|a|\leq2N-17}|\partial\overline{Z}^{a}v|\lesssim\varepsilon_{1}\langle t+|x|\rangle^{6\delta_{0}-1}. \tag*{(206)} ∣ a ∣ ≤ 2 N − 17 ∑ ∣ ∂ Z a v ∣ ≲ ε 1 ⟨ t + ∣ x ∣ ⟩ 6 δ 0 − 1 . (206)
Proof. First, it follows from (205) that for ∣ a ∣ ≤ 2 N − 18 |a|\leq2N-18 ∣ a ∣ ≤ 2 N − 18 and ∣ y ∣ ≥ 1 + s / 2 |y|\geq1+s/2 ∣ y ∣ ≥ 1 + s /2 ,
∣ □ Z a v ( s , y ) ∣ ≲ ε 1 2 ⟨ s + ∣ y ∣ ⟩ 4 δ 0 − 1 ⟨ s − ∣ y ∣ ⟩ δ 0 − 1 . (207) |\Box Z^{a}v(s,y)|\lesssim\varepsilon_{1}^{2}\langle s+|y|\rangle^{4\delta_{0}-1}\langle s-|y|\rangle^{\delta_{0}-1}. \tag*{(207)} ∣ □ Z a v ( s , y ) ∣ ≲ ε 1 2 ⟨ s + ∣ y ∣ ⟩ 4 δ 0 − 1 ⟨ s − ∣ y ∣ ⟩ δ 0 − 1 . (207)
By plugging (8) , (201) and (207) into (95) with w = Z a v w=Z^{a}v w = Z a v and ∣ a ∣ ≤ 2 N − 17 |a|\leq2N-17 ∣ a ∣ ≤ 2 N − 17 , one has that for ∣ x ∣ ≥ 1 + t / 2 |x|\geq1+t/2 ∣ x ∣ ≥ 1 + t /2 ,
∣ ∂ + ( r 1 / 2 Z a v ) ( t , x ) ∣ ≲ ε 1 2 ⟨ x ⟩ − 1 + ε 1 2 ( r + t ) 4 δ 0 − 1 / 2 ∫ 0 t ( 1 + ∣ r + t − 2 s ∣ ) δ 0 − 1 d s + ε 1 ∫ 0 t ( r + t − s ) − 3 / 2 ⟨ r + t ⟩ 6 δ 0 d s ≲ ε 1 ⟨ t + ∣ x ∣ ⟩ 6 δ 0 − 1 / 2 , \begin{aligned}
|\partial_{+}(r^{1/2}Z^{a}v)(t,x)|
&\lesssim\varepsilon_{1}^{2}\langle x\rangle^{-1}+\varepsilon_{1}^{2}(r+t)^{4\delta_{0}-1/2}\int_{0}^{t}(1+|r+t-2s|)^{\delta_{0}-1}\,ds\\
&\quad+\varepsilon_{1}\int_{0}^{t}(r+t-s)^{-3/2}\langle r+t\rangle^{6\delta_{0}}\,ds\\
&\lesssim\varepsilon_{1}\langle t+|x|\rangle^{6\delta_{0}-1/2},
\end{aligned} ∣ ∂ + ( r 1/2 Z a v ) ( t , x ) ∣ ≲ ε 1 2 ⟨ x ⟩ − 1 + ε 1 2 ( r + t ) 4 δ 0 − 1/2 ∫ 0 t ( 1 + ∣ r + t − 2 s ∣ ) δ 0 − 1 d s + ε 1 ∫ 0 t ( r + t − s ) − 3/2 ⟨ r + t ⟩ 6 δ 0 d s ≲ ε 1 ⟨ t + ∣ x ∣ ⟩ 6 δ 0 − 1/2 ,
which further implies
∣ ∂ + Z a v ∣ ≲ r − 1 / 2 ∣ ∂ + ( r 1 / 2 Z a v ) ( t , x ) ∣ + r − 1 ∣ Z a v ( t , x ) ∣ ≲ ε 1 ⟨ t + ∣ x ∣ ⟩ 6 δ 0 − 1 . |\partial_+Z^av|\lesssim r^{-1/2}|\partial_+(r^{1/2}Z^av)(t,x)|+r^{-1}|Z^av(t,x)|
\lesssim\varepsilon_1\langle t+|x|\rangle^{6\delta_0-1}. ∣ ∂ + Z a v ∣ ≲ r − 1/2 ∣ ∂ + ( r 1/2 Z a v ) ( t , x ) ∣ + r − 1 ∣ Z a v ( t , x ) ∣ ≲ ε 1 ⟨ t + ∣ x ∣ ⟩ 6 δ 0 − 1 .
Together with (95) and (201) , this finishes the proof of (206) .
Lemma 7.3. Under the assumptions of Theorem 1.1 , let v v v be defined by (142) and suppose that (143) –(146) hold. Then one has
∑ ∣ a ∣ ≤ 2 N − 21 ∣ Z a v ∣ ≲ ε 1 ⟨ t + ∣ x ∣ ⟩ 8 δ 0 − 1 / 2 , \sum_{|a|\leq2N-21}|Z^av|\lesssim\varepsilon_1\langle t+|x|\rangle^{8\delta_0-1/2}, ∣ a ∣ ≤ 2 N − 21 ∑ ∣ Z a v ∣ ≲ ε 1 ⟨ t + ∣ x ∣ ⟩ 8 δ 0 − 1/2 , ∑ ∣ a ∣ ≤ 2 N − 21 ∣ ∂ Z a v ∣ ≲ ε 1 ⟨ x ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ − 1 / 2 ⟨ t + ∣ x ∣ ⟩ 8 δ 0 . (208) \sum_{|a|\leq2N-21}|\partial Z^av|\lesssim\varepsilon_1\langle x\rangle^{-1/2}\langle t-|x|\rangle^{-1/2}\langle t+|x|\rangle^{8\delta_0}. \tag*{(208)} ∣ a ∣ ≤ 2 N − 21 ∑ ∣ ∂ Z a v ∣ ≲ ε 1 ⟨ x ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ − 1/2 ⟨ t + ∣ x ∣ ⟩ 8 δ 0 . (208)
Proof. With the estimate (201) , we will use the equation (53) to derive (208) . First, in order to cancel out the quadratic Q 0 Q_0 Q 0 term in (53) while preserve the zero boundary conditions, introducing the good unknown
V a = Z ~ a v − 1 2 ∑ b + c ≤ a ∑ μ = 0 2 C a b c 0 , μ ( ω ) χ [ 1 , 2 ] ( x ) ∂ μ Z b u Z c v . (209) V^a=\widetilde Z^av-\frac{1}{2}\sum_{b+c\leq a}\sum_{\mu=0}^{2}C_{abc}^{0,\mu}(\omega)\chi_{[1,2]}(x)\partial_\mu Z^buZ^cv. \tag*{(209)} V a = Z a v − 2 1 b + c ≤ a ∑ μ = 0 ∑ 2 C ab c 0 , μ ( ω ) χ [ 1 , 2 ] ( x ) ∂ μ Z b u Z c v . (209)
This, together with (53) and (171) , yields V a ∣ [ 0 , ∞ ) × ∂ K = 0 V^a|_{[0,\infty)\times\partial K}=0 V a ∣ [ 0 , ∞ ) × ∂ K = 0 and
□ V a = □ ( Z ~ a − Z a ) v + F 1 a ( t , x ) , \Box V^a=\Box(\widetilde Z^a-Z^a)v+F_1^a(t,x), □ V a = □ ( Z a − Z a ) v + F 1 a ( t , x ) ,
F 1 a ( t , x ) = ∑ b + c ≤ a { ∑ μ , ν = 0 2 ∣ x ∣ − 1 ( C ( ω ) ∂ μ ν 2 Z b u Ω Z c v + C ( ω ) Ω ∂ μ Z b u ∂ ν Z c v ) + ∑ μ = 0 2 [ C a b c 12 , μ Q 12 ( ∂ μ Z b u , Z c v ) + C ( ω ) □ Z b u ∂ μ Z c v + 1 2 [ Δ , C a b c 0 , μ ( ω ) χ [ 1 , 2 ] ( x ) ] ( ∂ μ Z b u Z c v ) + C ( ω ) χ [ 1 , 2 ] ( x ) ( □ ∂ μ Z b u Z c v + ∂ μ Z b u □ Z c v ) + 1 2 C a b c 0 , μ ( ω ) ( 1 − χ [ 1 , 2 ] ( x ) ) Q 0 ( ∂ μ Z b u , Z c v ) ] } + Z a [ ∑ α , β , γ , δ = 0 2 Q α β γ δ ∂ α β 2 u ∂ γ u ∂ δ u ] . (210) \begin{aligned}
F_1^a(t,x)={}&\sum_{b+c\leq a}\left\{\sum_{\mu,\nu=0}^{2}|x|^{-1}\left(C(\omega)\partial_{\mu\nu}^2Z^bu\Omega Z^cv+C(\omega)\Omega\partial_\mu Z^bu\partial_\nu Z^cv\right)\right.\\
&+\sum_{\mu=0}^{2}\left[C_{abc}^{12,\mu}Q_{12}(\partial_\mu Z^bu,Z^cv)+C(\omega)\Box Z^bu\partial_\mu Z^cv\right.\\
&\left.\left.+\frac{1}{2}[\Delta,C_{abc}^{0,\mu}(\omega)\chi_{[1,2]}(x)](\partial_\mu Z^buZ^cv)+C(\omega)\chi_{[1,2]}(x)(\Box\partial_\mu Z^buZ^cv+\partial_\mu Z^bu\Box Z^cv)\right.\right.\\
&\left.\left.+\frac{1}{2}C_{abc}^{0,\mu}(\omega)(1-\chi_{[1,2]}(x))Q_0(\partial_\mu Z^bu,Z^cv)\right]\right\}+Z^a\left[\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\partial_{\alpha\beta}^2u\partial_\gamma u\partial_\delta u\right]. \tag*{(210)}
\end{aligned} F 1 a ( t , x ) = b + c ≤ a ∑ { μ , ν = 0 ∑ 2 ∣ x ∣ − 1 ( C ( ω ) ∂ μν 2 Z b u Ω Z c v + C ( ω ) Ω ∂ μ Z b u ∂ ν Z c v ) + μ = 0 ∑ 2 [ C ab c 12 , μ Q 12 ( ∂ μ Z b u , Z c v ) + C ( ω ) □ Z b u ∂ μ Z c v + 2 1 [ Δ , C ab c 0 , μ ( ω ) χ [ 1 , 2 ] ( x )] ( ∂ μ Z b u Z c v ) + C ( ω ) χ [ 1 , 2 ] ( x ) ( □ ∂ μ Z b u Z c v + ∂ μ Z b u □ Z c v ) + 2 1 C ab c 0 , μ ( ω ) ( 1 − χ [ 1 , 2 ] ( x )) Q 0 ( ∂ μ Z b u , Z c v ) ] } + Z a α , β , γ , δ = 0 ∑ 2 Q α β γ δ ∂ α β 2 u ∂ γ u ∂ δ u . (210)
For the first term on the fourth line of (210) , recall (47) of [17 ]
∣ [ Δ , C a b c 0 , μ ( ω ) χ [ 1 , 2 ] ( x ) ] ( ∂ μ Z b u Z c v ) ∣ ≲ ∣ x ∣ − 2 ∣ Z ≤ 1 ( ∂ μ Z b u Z c v ) ∣ . |[\Delta,C_{abc}^{0,\mu}(\omega)\chi_{[1,2]}(x)](\partial_\mu Z^buZ^cv)|\lesssim|x|^{-2}|Z^{\leq1}(\partial_\mu Z^buZ^cv)|. ∣ [ Δ , C ab c 0 , μ ( ω ) χ [ 1 , 2 ] ( x )] ( ∂ μ Z b u Z c v ) ∣ ≲ ∣ x ∣ − 2 ∣ Z ≤ 1 ( ∂ μ Z b u Z c v ) ∣.
Then it can be deduced from (48) , (91) , (143) , (145) , (197) and (201) that
∑ ∣ a ∣ ≤ 2 N − 21 ∑ ∣ b ∣ ≤ 5 ∣ Z b F 1 a ( s , y ) ∣ ≲ ε 1 2 ⟨ y ⟩ − 3 / 2 ( 1 + s ) 7 δ 0 − 1 , ∣ y ∣ ≤ 1 + s / 2 , \sum_{|a|\leq2N-21}\sum_{|b|\leq5}|Z^bF_1^a(s,y)|\lesssim\varepsilon_1^2\langle y\rangle^{-3/2}(1+s)^{7\delta_0-1},\qquad|y|\leq1+s/2, ∣ a ∣ ≤ 2 N − 21 ∑ ∣ b ∣ ≤ 5 ∑ ∣ Z b F 1 a ( s , y ) ∣ ≲ ε 1 2 ⟨ y ⟩ − 3/2 ( 1 + s ) 7 δ 0 − 1 , ∣ y ∣ ≤ 1 + s /2 ,
∑ ∣ a ∣ ≤ 2 N − 21 ∑ ∣ b ∣ ≤ 5 ∣ Z b F 1 a ( s , y ) ∣ ≲ ε 1 2 ⟨ s + ∣ y ∣ ⟩ 7 δ 0 − 3 / 2 ⟨ s − ∣ y ∣ ⟩ − 1 , ∣ y ∣ ≥ 1 + s / 2. (211) \sum_{|a|\leq2N-21}\sum_{|b|\leq5}|Z^bF_1^a(s,y)|\lesssim\varepsilon_1^2\langle s+|y|\rangle^{7\delta_0-3/2}\langle s-|y|\rangle^{-1},\qquad|y|\geq1+s/2. \tag*{(211)} ∣ a ∣ ≤ 2 N − 21 ∑ ∣ b ∣ ≤ 5 ∑ ∣ Z b F 1 a ( s , y ) ∣ ≲ ε 1 2 ⟨ s + ∣ y ∣ ⟩ 7 δ 0 − 3/2 ⟨ s − ∣ y ∣ ⟩ − 1 , ∣ y ∣ ≥ 1 + s /2. (211)
Applying (119) and (120) to □ V a \Box V^a □ V a in (210) with (204) , (211) , F 0 ≡ 0 F_0 \equiv0 F 0 ≡ 0 , μ = ν = δ 0 / 3 \mu=\nu=\delta_0/3 μ = ν = δ 0 /3 and ∣ a ∣ ≤ 2 N − 21 |a|\leq2N-21 ∣ a ∣ ≤ 2 N − 21 derives
∑ ∣ a ∣ ≤ 2 N − 21 ∣ V a ∣ ≲ ε 1 2 ⟨ t + ∣ x ∣ ⟩ 8 δ 0 − 1 / 2 . \sum_{|a|\leq2N-21}|V^a|\lesssim\varepsilon_1^2\langle t+|x|\rangle^{8\delta_0-1/2}. ∣ a ∣ ≤ 2 N − 21 ∑ ∣ V a ∣ ≲ ε 1 2 ⟨ t + ∣ x ∣ ⟩ 8 δ 0 − 1/2 .
∑ ∣ a ∣ ≤ 2 N − 21 ∣ ∂ V a ∣ ≲ ε 1 2 ⟨ x ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ − 1 / 2 ⟨ t + ∣ x ∣ ⟩ 8 δ 0 . \sum_{|a|\leq2N-21}|\partial V^a|\lesssim\varepsilon_1^2\langle x\rangle^{-1/2}\langle t-|x|\rangle^{-1/2}\langle t+|x|\rangle^{8\delta_0}. ∣ a ∣ ≤ 2 N − 21 ∑ ∣ ∂ V a ∣ ≲ ε 1 2 ⟨ x ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ − 1/2 ⟨ t + ∣ x ∣ ⟩ 8 δ 0 .
This, together with (91) , (145) , (196) , (197) and (209) , yields (208) .
By the second line of (208) , the estimate of the first line in (208) can be further improved as follows.
Lemma 7.4. Under the assumptions of Theorem 1.1 , let v v v be defined by (142) and suppose that (143) -(146) hold. Then one has
∑ ∣ a ∣ ≤ 2 N − 27 ∣ Z a v ∣ ≲ ( ε 1 2 + ε 1 κ ) ⟨ t + ∣ x ∣ ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ − 0.4 . (212) \sum_{|a|\leq2N-27}|Z^a v|\lesssim(\varepsilon_1^2+\varepsilon_1\kappa)\langle t+|x|\rangle^{-1/2}\langle t-|x|\rangle^{-0.4}. \tag*{(212)} ∣ a ∣ ≤ 2 N − 27 ∑ ∣ Z a v ∣ ≲ ( ε 1 2 + ε 1 κ ) ⟨ t + ∣ x ∣ ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ − 0.4 . (212)
Proof. Instead of (210) , we use the following equation
□ V a = □ ( Z ~ a − Z a ) v + ∂ t F 0 a ( t , x ) + F ~ 1 a ( t , x ) , \Box V^a=\Box(\widetilde Z^a-Z^a)v+\partial_tF_0^a(t,x)+\widetilde F_1^a(t,x), □ V a = □ ( Z a − Z a ) v + ∂ t F 0 a ( t , x ) + F 1 a ( t , x ) ,
F 0 a = 1 3 Z a [ ∑ α , β , γ , δ = 0 2 Q α β γ δ ω α ω β ω γ ω δ ( ∂ t u ) 3 ] . (213) F_0^a=\frac{1}{3}Z^a\left[\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\omega_\alpha\omega_\beta\omega_\gamma\omega_\delta(\partial_tu)^3\right]. \tag*{(213)} F 0 a = 3 1 Z a α , β , γ , δ = 0 ∑ 2 Q α β γ δ ω α ω β ω γ ω δ ( ∂ t u ) 3 . (213)
and
F ~ 1 a ( t , x ) = ∑ b + c ≤ a { ∑ μ , ν = 0 2 ∣ x ∣ − 1 ( C ( ω ) ∂ μ ν 2 Z b u Ω Z c u + C ( ω ) Ω ∂ μ Z b u ∂ ν Z c u ) + ∑ μ = 0 2 [ C a b c 12 , μ ( ω ) Q 12 ( ∂ μ Z b u , Z c u ) + C ( ω ) □ Z b u ∂ μ Z c u + 1 2 C a b c 0 , μ ( ω ) ( 1 − χ [ 1 , 2 ] ( x ) ) Q 0 ( ∂ μ Z b u , Z c u ) + 1 2 [ Δ , C a b c 0 , μ ( ω ) χ [ 1 , 2 ] ( x ) ] ( ∂ μ Z b u Z c u ) + C ( ω ) χ [ 1 , 2 ] ( x ) ( □ ∂ μ Z b u Z c u + ∂ μ Z b u □ Z c u ) ] } + Z a [ ∑ α , β , γ , δ = 0 2 Q α β γ δ ( ∂ α β 2 u ∂ γ u ∂ δ u − ω α ω β ω γ ω δ ∂ t 2 u ( ∂ t u ) 2 ) ] . \begin{aligned}
\widetilde F_1^a(t,x)={}&\sum_{b+c\leq a}\left\{\sum_{\mu,\nu=0}^{2}|x|^{-1}\left(C(\omega)\partial_{\mu\nu}^2Z^bu\Omega Z^cu+C(\omega)\Omega\partial_\mu Z^bu\partial_\nu Z^cu\right)\right.\\
&+\sum_{\mu=0}^{2}\left[C_{abc}^{12,\mu}(\omega)Q_{12}(\partial_\mu Z^bu,Z^cu)+C(\omega)\Box Z^bu\partial_\mu Z^cu+\frac{1}{2}C_{abc}^{0,\mu}(\omega)(1-\chi_{[1,2]}(x))Q_0(\partial_\mu Z^bu,Z^cu)\right.\\
&\left.\left.+\frac{1}{2}[\Delta,C_{abc}^{0,\mu}(\omega)\chi_{[1,2]}(x)](\partial_\mu Z^buZ^cu)+C(\omega)\chi_{[1,2]}(x)(\Box\partial_\mu Z^buZ^cu+\partial_\mu Z^bu\Box Z^cu)\right]\right\}\\
&+Z^a\left[\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\left(\partial_{\alpha\beta}^2u\partial_\gamma u\partial_\delta u-\omega_\alpha\omega_\beta\omega_\gamma\omega_\delta\partial_t^2u(\partial_tu)^2\right)\right].
\end{aligned} F 1 a ( t , x ) = b + c ≤ a ∑ { μ , ν = 0 ∑ 2 ∣ x ∣ − 1 ( C ( ω ) ∂ μν 2 Z b u Ω Z c u + C ( ω ) Ω ∂ μ Z b u ∂ ν Z c u ) + μ = 0 ∑ 2 [ C ab c 12 , μ ( ω ) Q 12 ( ∂ μ Z b u , Z c u ) + C ( ω ) □ Z b u ∂ μ Z c u + 2 1 C ab c 0 , μ ( ω ) ( 1 − χ [ 1 , 2 ] ( x )) Q 0 ( ∂ μ Z b u , Z c u ) + 2 1 [ Δ , C ab c 0 , μ ( ω ) χ [ 1 , 2 ] ( x )] ( ∂ μ Z b u Z c u ) + C ( ω ) χ [ 1 , 2 ] ( x ) ( □ ∂ μ Z b u Z c u + ∂ μ Z b u □ Z c u ) ] } + Z a α , β , γ , δ = 0 ∑ 2 Q α β γ δ ( ∂ α β 2 u ∂ γ u ∂ δ u − ω α ω β ω γ ω δ ∂ t 2 u ( ∂ t u ) 2 ) .
On the other hand, (34) and (200) imply
∑ ∣ a ∣ ≤ 2 N − 15 ∣ ∂ Z a v ( t , x ) ∣ ≲ ε 1 ⟨ x ⟩ − 1 / 2 . (214) \sum_{|a|\leq2N-15}|\partial Z^av(t,x)|\lesssim\varepsilon_1\langle x\rangle^{-1/2}. \tag*{(214)} ∣ a ∣ ≤ 2 N − 15 ∑ ∣ ∂ Z a v ( t , x ) ∣ ≲ ε 1 ⟨ x ⟩ − 1/2 . (214)
Together with (91) and (143) , this yields
∑ ∣ a ∣ ≤ 2 N − 27 ∑ ∣ b ∣ ≤ 6 ∣ Z b F 0 a ( s , y ) ∣ ≲ ε 1 3 ⟨ y ⟩ − 3 / 2 ⟨ s − ∣ y ∣ ⟩ 2 δ 0 − 2 . (215) \sum_{|a|\leq2N-27}\sum_{|b|\leq6}|Z^bF_0^a(s,y)|\lesssim\varepsilon_1^3\langle y\rangle^{-3/2}\langle s-|y|\rangle^{2\delta_0-2}. \tag*{(215)} ∣ a ∣ ≤ 2 N − 27 ∑ ∣ b ∣ ≤ 6 ∑ ∣ Z b F 0 a ( s , y ) ∣ ≲ ε 1 3 ⟨ y ⟩ − 3/2 ⟨ s − ∣ y ∣ ⟩ 2 δ 0 − 2 . (215)
Analogously to the estimates (175) , (177) and (211) , based on (47) , (48) , (50) , (206) , (208) and (214) , one arrives at
∑ ∣ a ∣ ≤ 2 N − 27 ∑ ∣ b ∣ ≤ 5 ∣ Z b F ~ 1 a ( s , y ) ∣ ≲ ε 1 2 ⟨ y ⟩ − 3 / 2 ( 1 + s ) 11 δ 0 − 3 / 2 , ∣ y ∣ ≤ 1 + s / 2 , ∑ ∣ a ∣ ≤ 2 N − 27 ∑ ∣ b ∣ ≤ 5 ∣ Z b F ~ 1 a ( s , y ) ∣ ≲ ε 1 2 ⟨ s + ∣ y ∣ ⟩ 11 δ 0 − 2 ⟨ s − ∣ y ∣ ⟩ − 1 , ∣ y ∣ ≥ 1 + s / 2. (216) \sum_{|a|\leq2N-27}\sum_{|b|\leq5}|Z^{b}\widetilde{F}^{a}_{1}(s,y)|\lesssim\varepsilon_{1}^{2}\langle y\rangle^{-3/2}(1+s)^{11\delta_{0}-3/2},\qquad|y|\leq1+s/2,
\quad
\sum_{|a|\leq2N-27}\sum_{|b|\leq5}|Z^{b}\widetilde{F}^{a}_{1}(s,y)|\lesssim\varepsilon_{1}^{2}\langle s+|y|\rangle^{11\delta_{0}-2}\langle s-|y|\rangle^{-1},\qquad|y|\geq1+s/2. \tag*{(216)} ∣ a ∣ ≤ 2 N − 27 ∑ ∣ b ∣ ≤ 5 ∑ ∣ Z b F 1 a ( s , y ) ∣ ≲ ε 1 2 ⟨ y ⟩ − 3/2 ( 1 + s ) 11 δ 0 − 3/2 , ∣ y ∣ ≤ 1 + s /2 , ∣ a ∣ ≤ 2 N − 27 ∑ ∣ b ∣ ≤ 5 ∑ ∣ Z b F 1 a ( s , y ) ∣ ≲ ε 1 2 ⟨ s + ∣ y ∣ ⟩ 11 δ 0 − 2 ⟨ s − ∣ y ∣ ⟩ − 1 , ∣ y ∣ ≥ 1 + s /2. (216)
With (196) instead of (183) , the estimate (204) can be improved to
∑ ∣ a ∣ ≤ 2 N − 27 ∑ ∣ b ∣ ≤ 4 sup s ∈ [ 0 , t ] ⟨ s ⟩ 1 − 8 δ 0 ∥ ∂ b □ ( Z ~ a − Z a ) v ( s ) ∥ L 2 ( K 4 ) ≲ ε 1 2 . (217) \sum_{|a|\leq2N-27}\sum_{|b|\leq4}\sup_{s\in[0,t]} \langle s\rangle^{1-8\delta_{0}}\|\partial^{b}\Box(\widetilde{Z}^{a}-Z^{a})v(s)\|_{L^{2}(\mathcal{K}_{4})}\lesssim\varepsilon_{1}^{2}. \tag*{(217)} ∣ a ∣ ≤ 2 N − 27 ∑ ∣ b ∣ ≤ 4 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ 1 − 8 δ 0 ∥ ∂ b □ ( Z a − Z a ) v ( s ) ∥ L 2 ( K 4 ) ≲ ε 1 2 . (217)
By using (119) to □ V a \Box V^{a} □ V a in (213) together with (147) , (215) , (216) , (217) , μ = 0.4 \mu=0.4 μ = 0.4 , ν = δ 0 \nu=\delta_{0} ν = δ 0 , ∣ a ∣ ≤ 2 N − 27 |a|\leq2N-27 ∣ a ∣ ≤ 2 N − 27 , F 0 = F 0 a F_{0}=F^{a}_{0} F 0 = F 0 a and F 1 = □ ( Z ~ a − Z a ) v + F ~ 1 a F_{1}=\Box(\widetilde{Z}^{a}-Z^{a})v+\widetilde{F}^{a}_{1} F 1 = □ ( Z a − Z a ) v + F 1 a , we can achieve
∑ ∣ a ∣ ≤ 2 N − 27 ∣ V a ∣ ≲ ( ε 1 2 + ε 1 κ ) ⟨ t + ∣ x ∣ ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ − 0.4 . \sum_{|a|\leq2N-27}|V^{a}|\lesssim(\varepsilon_{1}^{2}+\varepsilon_{1}\kappa)\langle t+|x|\rangle^{-1/2}\langle t-|x|\rangle^{-0.4}. ∣ a ∣ ≤ 2 N − 27 ∑ ∣ V a ∣ ≲ ( ε 1 2 + ε 1 κ ) ⟨ t + ∣ x ∣ ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ − 0.4 .
Combining this with (91) , (145) , (196) , (208) and (209) yields (212) .
Based on (208) , the estimate (206) of the good derivatives can be also refined.
Lemma 7.5. Under the assumptions of Theorem 1.1 , let v v v be defined by (142) and suppose that (143) –(146) hold. Then it holds that for t ∈ [ 0 , T ε ] t\in[0,T_{\varepsilon}] t ∈ [ 0 , T ε ] and ∣ x ∣ ≥ 1 + t / 2 |x|\geq1+t/2 ∣ x ∣ ≥ 1 + t /2 ,
∑ ∣ a ∣ ≤ 2 N − 23 ∣ ∂ ˉ Z a v ∣ ≲ ε 1 ⟨ t + ∣ x ∣ ⟩ 9 δ 0 − 3 / 2 . (218) \sum_{|a|\leq2N-23}|\bar{\partial}Z^{a}v|\lesssim\varepsilon_{1}\langle t+|x|\rangle^{9\delta_{0}-3/2}. \tag*{(218)} ∣ a ∣ ≤ 2 N − 23 ∑ ∣ ∂ ˉ Z a v ∣ ≲ ε 1 ⟨ t + ∣ x ∣ ⟩ 9 δ 0 − 3/2 . (218)
Proof. First, it follows from (47) , (49) , (50) , (91) , (143) , (144) , (206) , (208) and (214) that for ∣ x ∣ ≥ 1 + t / 2 |x|\geq1+t/2 ∣ x ∣ ≥ 1 + t /2 and ∣ a ∣ ≤ 2 N − 23 |a|\leq2N-23 ∣ a ∣ ≤ 2 N − 23 ,
∣ Z a Q ( ∂ u , ∂ 2 u ) ∣ ≲ ∑ b + c ≤ a ( ∣ ∂ ˉ Z b u ∣ ∣ ∂ Z c u ∣ + ∣ ∂ 2 Z b u ∣ ∣ ∂ ˉ Z c u ∣ ) + ∑ b + c + d ≤ a ∣ ∂ 2 Z b u ∣ ∣ ∂ Z c u ∣ ∣ ∂ Z d u ∣ |Z^{a}Q(\partial u,\partial^{2}u)|\lesssim\sum_{b+c\leq a}\left(|\bar{\partial}Z^{b}u||\partial Z^{c}u|+|\partial^{2}Z^{b}u||\bar{\partial}Z^{c}u|\right)+\sum_{b+c+d\leq a}|\partial^{2}Z^{b}u||\partial Z^{c}u||\partial Z^{d}u| ∣ Z a Q ( ∂ u , ∂ 2 u ) ∣ ≲ b + c ≤ a ∑ ( ∣ ∂ ˉ Z b u ∣∣ ∂ Z c u ∣ + ∣ ∂ 2 Z b u ∣∣ ∂ ˉ Z c u ∣ ) + b + c + d ≤ a ∑ ∣ ∂ 2 Z b u ∣∣ ∂ Z c u ∣∣ ∂ Z d u ∣
≲ ε 1 2 ⟨ t + ∣ x ∣ ⟩ 6 δ 0 − 3 / 2 ⟨ t − ∣ x ∣ ⟩ δ 0 − 1 + ε 1 2 ⟨ t + ∣ x ∣ ⟩ 9 δ 0 − 2 ⟨ t − ∣ x ∣ ⟩ − 1 / 2 + ε 1 3 ⟨ t + ∣ x ∣ ⟩ − 3 / 2 ⟨ t − ∣ x ∣ ⟩ 2 δ 0 − 2 . (219) \lesssim\varepsilon_{1}^{2}\langle t+|x|\rangle^{6\delta_{0}-3/2}\langle t-|x|\rangle^{\delta_{0}-1}+\varepsilon_{1}^{2}\langle t+|x|\rangle^{9\delta_{0}-2}\langle t-|x|\rangle^{-1/2}+\varepsilon_{1}^{3}\langle t+|x|\rangle^{-3/2}\langle t-|x|\rangle^{2\delta_{0}-2}. \tag*{(219)} ≲ ε 1 2 ⟨ t + ∣ x ∣ ⟩ 6 δ 0 − 3/2 ⟨ t − ∣ x ∣ ⟩ δ 0 − 1 + ε 1 2 ⟨ t + ∣ x ∣ ⟩ 9 δ 0 − 2 ⟨ t − ∣ x ∣ ⟩ − 1/2 + ε 1 3 ⟨ t + ∣ x ∣ ⟩ − 3/2 ⟨ t − ∣ x ∣ ⟩ 2 δ 0 − 2 . (219)
Similarly to Lemma 7.2, by using (95) to □ Z a v = Z a Q ( ∂ u , ∂ 2 u ) \Box Z^{a}v=Z^{a}Q(\partial u,\partial^{2}u) □ Z a v = Z a Q ( ∂ u , ∂ 2 u ) in (49) with ∣ a ∣ ≤ 2 N − 23 |a|\leq2N-23 ∣ a ∣ ≤ 2 N − 23 and (219) , one has that for ∣ x ∣ ≥ 1 + t / 2 |x|\geq1+t/2 ∣ x ∣ ≥ 1 + t /2 ,
∣ ∂ + ( r 1 / 2 Z a v ) ( t , x ) ∣ ≲ ε 2 ⟨ x ⟩ − 1 + ε 1 2 ( r + t ) 8 δ 0 − 1 ∫ 0 t ( 1 + ∣ r + t − 2 s ∣ ) δ 0 − 1 d s |\partial_{+}(r^{1/2}Z^{a}v)(t,x)|\lesssim\varepsilon^{2}\langle x\rangle^{-1}+\varepsilon_{1}^{2}(r+t)^{8\delta_{0}-1}\int_{0}^{t}(1+|r+t-2s|)^{\delta_{0}-1}\,ds ∣ ∂ + ( r 1/2 Z a v ) ( t , x ) ∣ ≲ ε 2 ⟨ x ⟩ − 1 + ε 1 2 ( r + t ) 8 δ 0 − 1 ∫ 0 t ( 1 + ∣ r + t − 2 s ∣ ) δ 0 − 1 d s
+ ε 1 ∫ 0 t ( r + t − s ) − 3 / 2 ⟨ r + t ⟩ 8 δ 0 − 1 / 2 d s ≲ ε 1 ⟨ t + ∣ x ∣ ⟩ 9 δ 0 − 1 , \qquad+\varepsilon_{1}\int_{0}^{t}(r+t-s)^{-3/2}\langle r+t\rangle^{8\delta_{0}-1/2}\,ds
\lesssim\varepsilon_{1}\langle t+|x|\rangle^{9\delta_{0}-1}, + ε 1 ∫ 0 t ( r + t − s ) − 3/2 ⟨ r + t ⟩ 8 δ 0 − 1/2 d s ≲ ε 1 ⟨ t + ∣ x ∣ ⟩ 9 δ 0 − 1 ,
where (208) is used again. This, together with (95) and (208) , leads to (218) .
:::
Improved pointwise estimates
Before improving (143) , the local energy decay estimates (196) will be enhanced.
Lemma 7.6. Under the assumptions of Theorem 1.1 , let v v v be defined by (142) and suppose that (143) –(146) hold. Then one has
∑ ∣ a ∣ ≤ 2 N − 25 ∥ ∂ a v ∥ L 2 ( K R ) ≲ ε 1 2 ⟨ t ⟩ − 1 . (220) \sum_{|a|\leq2N-25}\|\partial^a v\|_{L^2(K_R)}\lesssim\varepsilon_1^2\langle t\rangle^{-1}. \tag*{(220)} ∣ a ∣ ≤ 2 N − 25 ∑ ∥ ∂ a v ∥ L 2 ( K R ) ≲ ε 1 2 ⟨ t ⟩ − 1 . (220)
Proof. Instead of the equation (185) , we have
□ V i = F 0 i ( t , x ) + F 1 i ( t , x ) , (221) \square V_i=F_0^i(t,x)+F_1^i(t,x), \tag*{(221)} □ V i = F 0 i ( t , x ) + F 1 i ( t , x ) , (221)
where
F 0 i ( t , x ) = 1 3 η ( t , x ) ∂ t i [ ∑ α , β , γ , δ = 0 2 Q α β γ δ ω α ω β ω γ ω δ ( ∂ t u ) 3 ] , F_0^i(t,x)=\frac{1}{3}\eta(t,x)\partial_t^i\left[\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\omega_\alpha\omega_\beta\omega_\gamma\omega_\delta(\partial_tu)^3\right], F 0 i ( t , x ) = 3 1 η ( t , x ) ∂ t i α , β , γ , δ = 0 ∑ 2 Q α β γ δ ω α ω β ω γ ω δ ( ∂ t u ) 3 ,
F 1 i ( t , x ) = 1 3 ( 1 − η ( t , x ) ) ∂ t i [ ∑ α , β , γ , δ = 0 2 Q α β γ δ ω α ω β ω γ ω δ ( ∂ t u ) 3 ] + ∂ t i [ ∑ α , β , γ , δ = 0 2 Q α β γ δ ( ∂ α β 2 u ∂ γ u ∂ δ u − ω α ω β ω γ ω δ ∂ t 2 u ( ∂ t u ) 2 ) ] + ∑ j + k ≤ i { ∑ μ , ν = 0 2 ∣ x ∣ − 1 ( C ( ω ) ∂ μ 2 ∂ t j u Ω ∂ t k u + C ( ω ) Ω ∂ μ ∂ t j u ∂ ν ∂ t k u ) + ∑ μ = 0 2 [ 1 2 ∑ μ = 0 2 [ Δ , C i j k 0 , μ ( ω ) ] ( ∂ μ ∂ t j u ∂ t k u ) + C i j k 12 , μ Q 12 ( ∂ μ ∂ t j u , ∂ t k u ) + C ( ω ) ∂ μ ∂ t j u □ ∂ t k u + C ( ω ) ∂ t k u □ ∂ μ ∂ t j u ] } . \begin{aligned}
F_1^i(t,x)={}&\frac{1}{3}(1-\eta(t,x))\partial_t^i\left[\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\omega_\alpha\omega_\beta\omega_\gamma\omega_\delta(\partial_tu)^3\right]\\
&+\partial_t^i\left[\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\left(\partial_{\alpha\beta}^2u\partial_\gamma u\partial_\delta u-\omega_\alpha\omega_\beta\omega_\gamma\omega_\delta\partial_t^2u(\partial_tu)^2\right)\right]\\
&+\sum_{j+k\leq i}\left\{\sum_{\mu,\nu=0}^{2}|x|^{-1}\left(C(\omega)\partial_{\mu}^{2}\partial_t^j u\Omega\partial_t^k u+C(\omega)\Omega\partial_\mu\partial_t^j u\partial_\nu\partial_t^k u\right)\right.\\
&\qquad+\sum_{\mu=0}^{2}\left[\frac{1}{2}\sum_{\mu=0}^{2}[\Delta,C_{ijk}^{0,\mu}(\omega)](\partial_\mu\partial_t^j u\partial_t^k u)+C_{ijk}^{12,\mu}Q_{12}(\partial_\mu\partial_t^j u,\partial_t^k u)\right.\\
&\qquad\qquad\left.\left.+C(\omega)\partial_\mu\partial_t^j u\square\partial_t^k u+C(\omega)\partial_t^k u\square\partial_\mu\partial_t^j u\right]\right\}.
\end{aligned} F 1 i ( t , x ) = 3 1 ( 1 − η ( t , x )) ∂ t i α , β , γ , δ = 0 ∑ 2 Q α β γ δ ω α ω β ω γ ω δ ( ∂ t u ) 3 + ∂ t i α , β , γ , δ = 0 ∑ 2 Q α β γ δ ( ∂ α β 2 u ∂ γ u ∂ δ u − ω α ω β ω γ ω δ ∂ t 2 u ( ∂ t u ) 2 ) + j + k ≤ i ∑ { μ , ν = 0 ∑ 2 ∣ x ∣ − 1 ( C ( ω ) ∂ μ 2 ∂ t j u Ω ∂ t k u + C ( ω ) Ω ∂ μ ∂ t j u ∂ ν ∂ t k u ) + μ = 0 ∑ 2 [ 2 1 μ = 0 ∑ 2 [ Δ , C ij k 0 , μ ( ω )] ( ∂ μ ∂ t j u ∂ t k u ) + C ij k 12 , μ Q 12 ( ∂ μ ∂ t j u , ∂ t k u ) + C ( ω ) ∂ μ ∂ t j u □ ∂ t k u + C ( ω ) ∂ t k u □ ∂ μ ∂ t j u ] } .
Analogously to (215) , (216) , one has
∑ i ≤ 2 N − 28 ∑ ∣ a ∣ ≤ 5 ∣ ∂ a F 0 i ( s , y ) ∣ ≲ ε 1 3 ⟨ y ⟩ − 3 / 2 ⟨ s − ∣ y ∣ ⟩ 2 δ 0 − 2 . (222) \sum_{i\leq2N-28}\sum_{|a|\leq5}|\partial^aF_0^i(s,y)|\lesssim\varepsilon_1^3\langle y\rangle^{-3/2}\langle s-|y|\rangle^{2\delta_0-2}. \tag*{(222)} i ≤ 2 N − 28 ∑ ∣ a ∣ ≤ 5 ∑ ∣ ∂ a F 0 i ( s , y ) ∣ ≲ ε 1 3 ⟨ y ⟩ − 3/2 ⟨ s − ∣ y ∣ ⟩ 2 δ 0 − 2 . (222)
and
∑ i ≤ 2 N − 27 ∑ ∣ a ∣ ≤ 2 ∣ Z a F 1 i ( s , y ) ∣ ≲ ε 1 2 ⟨ y ⟩ − 3 / 2 ⟨ s ⟩ 11 δ 0 − 3 / 2 , ∣ y ∣ ≤ 1 + s / 2 , or ∣ y ∣ ≤ 3 , \sum_{i\leq2N-27}\sum_{|a|\leq2}|Z^aF_1^i(s,y)|\lesssim\varepsilon_1^2\langle y\rangle^{-3/2}\langle s\rangle^{11\delta_0-3/2},\qquad|y|\leq1+s/2,\ \text{or }|y|\leq3, i ≤ 2 N − 27 ∑ ∣ a ∣ ≤ 2 ∑ ∣ Z a F 1 i ( s , y ) ∣ ≲ ε 1 2 ⟨ y ⟩ − 3/2 ⟨ s ⟩ 11 δ 0 − 3/2 , ∣ y ∣ ≤ 1 + s /2 , or ∣ y ∣ ≤ 3 ,
∑ i ≤ 2 N − 27 ∑ ∣ a ∣ ≤ 2 ∣ Z a F 1 i ( s , y ) ∣ ≲ ε 1 2 ⟨ s + ∣ y ∣ ⟩ 11 δ 0 − 2 ⟨ s − ∣ y ∣ ⟩ − 1 , ∣ y ∣ ≥ 1 + s / 2. (223) \sum_{i\leq2N-27}\sum_{|a|\leq2}|Z^aF_1^i(s,y)|\lesssim\varepsilon_1^2\langle s+|y|\rangle^{11\delta_0-2}\langle s-|y|\rangle^{-1},\qquad|y|\geq1+s/2. \tag*{(223)} i ≤ 2 N − 27 ∑ ∣ a ∣ ≤ 2 ∑ ∣ Z a F 1 i ( s , y ) ∣ ≲ ε 1 2 ⟨ s + ∣ y ∣ ⟩ 11 δ 0 − 2 ⟨ s − ∣ y ∣ ⟩ − 1 , ∣ y ∣ ≥ 1 + s /2. (223)
Applying (113) to (221) together with μ = δ 0 \mu=\delta_0 μ = δ 0 , the definition (184) , (222) and (223) yields
∑ i ≤ 2 N − 27 ∥ ∂ ≤ 1 ∂ t ∂ t i v ∥ L 2 ( K R 1 ) ≲ ε 1 2 ⟨ t ⟩ − 1 . \sum_{i\leq2N-27}\|\partial^{\leq1}\partial_t\partial_t^i v\|_{L^2(K_{R_1})}\lesssim\varepsilon_1^2\langle t\rangle^{-1}. i ≤ 2 N − 27 ∑ ∥ ∂ ≤ 1 ∂ t ∂ t i v ∥ L 2 ( K R 1 ) ≲ ε 1 2 ⟨ t ⟩ − 1 .
This, together with (169) and the argument in the proof of (183) , derives (220) .
Lemma 7.7. Under the assumptions of Theorem 1.1 , let v v v be defined by (142) and suppose that (143) -(146) hold. Then it holds that
∑ ∣ a ∣ ≤ 2 N − 46 ∣ ∂ Z a v ∣ ≲ ( ε 1 2 + ε 1 κ ) ⟨ x ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ δ 0 − 1 , ∑ ∣ a ∣ ≤ 2 N − 46 ∣ ∂ Z a v ∣ ≲ ε 1 2 ⟨ t ⟩ − 1 , ∣ x ∣ ≤ t / 10. (224) \sum_{|a|\leq2N-46} \lvert\partial Z^{a}v\rvert\lesssim\left(\varepsilon_{1}^{2}+\varepsilon_{1}\kappa\right)\langle x\rangle^{-1/2}\langle t-\lvert x\rvert\rangle^{\delta_{0}-1},
\qquad
\sum_{|a|\leq2N-46} \lvert\partial Z^{a}v\rvert\lesssim\varepsilon_{1}^{2}\langle t\rangle^{-1},\quad\lvert x\rvert\leq t/10. \tag*{(224)} ∣ a ∣ ≤ 2 N − 46 ∑ ∣ ∂ Z a v ∣ ≲ ( ε 1 2 + ε 1 κ ) ⟨ x ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ δ 0 − 1 , ∣ a ∣ ≤ 2 N − 46 ∑ ∣ ∂ Z a v ∣ ≲ ε 1 2 ⟨ t ⟩ − 1 , ∣ x ∣ ≤ t /10. (224)
Proof. Let V a V^{a} V a be defined by (209) . It follows from (54) instead of (53) that
□ V a = □ ( Z ~ a − Z a ) v + ∂ t F ~ 0 a ( t , x ) + ∑ Ξ = 0 , 1 , 2 , θ ∂ Ξ G a Ξ ( t , x ) + G a , 1 r ( t , x ) + G a , 2 r ( t , x ) , (225) \square V^{a}=\square(\widetilde{Z}^{a}-Z^{a})v+\partial_{t}\widetilde{F}_{0}^{a}(t,x)+\sum_{\Xi=0,1,2,\theta}\partial_{\Xi}G_{a}^{\Xi}(t,x)+G_{a,1}^{r}(t,x)+G_{a,2}^{r}(t,x), \tag*{(225)} □ V a = □ ( Z a − Z a ) v + ∂ t F 0 a ( t , x ) + Ξ = 0 , 1 , 2 , θ ∑ ∂ Ξ G a Ξ ( t , x ) + G a , 1 r ( t , x ) + G a , 2 r ( t , x ) , (225)
where
G a , 1 r ( t , x ) = ∑ b + c ≤ a { C ( ω ) Z c u Z ∂ Z b u ∣ x ∣ 2 + ∑ μ = 0 2 [ C ( ω ) ( 1 − χ [ 1 , 2 ] ( x ) ) Q 0 ( ∂ μ Z b u , Z c u ) + C ( ω ) ∂ μ Z c u ∑ b ′ + c ′ ≤ b α , β , γ = 0 2 Q b b ′ c ′ α β γ ∂ α β 2 Z b ′ u ∂ γ Z c ′ u + 1 2 [ Δ , C a b c 0 , μ ( ω ) χ [ 1 , 2 ] ( x ) ] ∂ μ Z b u Z c u ] + C ( ω ) χ [ 1 , 2 ] ( x ) Z c u ∑ b ′ + c ′ ≤ b α , β , γ = 0 2 Q b b ′ c ′ α β γ ∂ μ ( ∂ α β 2 Z b ′ u ∂ γ Z c ′ u ) + C ( ω ) χ [ 1 , 2 ] ( x ) ∂ μ Z b u ∑ b ′ + c ′ ≤ c α , β , γ = 0 2 Q c b ′ c ′ α β γ ∂ α β 2 Z b ′ u ∂ γ Z c ′ u } + 1 3 ∂ t Z a [ ( 1 − η ( t , x ) ) ∑ α , β , γ , δ = 0 2 Q α β γ δ ω α ω β ω γ ω δ ( ∂ t u ) 3 ] + Z a [ ∑ α , β , γ , δ = 0 2 Q α β γ δ ( ∂ α β 2 u ∂ γ u ∂ δ u − ω α ω β ω γ ω δ ∂ t 2 u ( ∂ t u ) 2 ) ] . (226) \begin{aligned}
G_{a,1}^{r}(t,x)={}&\sum_{b+c\leq a}\left\{\frac{C(\omega)Z^{c}uZ\partial Z^{b}u}{\lvert x\rvert^{2}}+\sum_{\mu=0}^{2}\left[C(\omega)(1-\chi_{[1,2]}(x))Q_{0}(\partial_{\mu}Z^{b}u,Z^{c}u)\right.\right.\\
&\left.\left.+C(\omega)\partial_{\mu}Z^{c}u\sum_{\substack{b'+c'\leq b\\ \alpha,\beta,\gamma=0}}^{2}Q_{bb'c'}^{\alpha\beta\gamma}\partial_{\alpha\beta}^{2}Z^{b'}u\partial_{\gamma}Z^{c'}u+\frac{1}{2}\left[\Delta,C_{abc}^{0,\mu}(\omega)\chi_{[1,2]}(x)\right]\partial_{\mu}Z^{b}uZ^{c}u\right]\right.\\
&\left.+C(\omega)\chi_{[1,2]}(x)Z^{c}u\sum_{\substack{b'+c'\leq b\\ \alpha,\beta,\gamma=0}}^{2}Q_{bb'c'}^{\alpha\beta\gamma}\partial_{\mu}\left(\partial_{\alpha\beta}^{2}Z^{b'}u\partial_{\gamma}Z^{c'}u\right)\right.\\
&\left.+C(\omega)\chi_{[1,2]}(x)\partial_{\mu}Z^{b}u\sum_{\substack{b'+c'\leq c\\ \alpha,\beta,\gamma=0}}^{2}Q_{cb'c'}^{\alpha\beta\gamma}\partial_{\alpha\beta}^{2}Z^{b'}u\partial_{\gamma}Z^{c'}u\right\}\\
&+\frac{1}{3}\partial_{t}Z^{a}\left[(1-\eta(t,x))\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\omega_{\alpha}\omega_{\beta}\omega_{\gamma}\omega_{\delta}(\partial_{t}u)^{3}\right]\\
&+Z^{a}\left[\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\left(\partial_{\alpha\beta}^{2}u\partial_{\gamma}u\partial_{\delta}u-\omega_{\alpha}\omega_{\beta}\omega_{\gamma}\omega_{\delta}\partial_{t}^{2}u(\partial_{t}u)^{2}\right)\right]. \tag*{(226)}
\end{aligned} G a , 1 r ( t , x ) = b + c ≤ a ∑ { ∣ x ∣ 2 C ( ω ) Z c u Z ∂ Z b u + μ = 0 ∑ 2 [ C ( ω ) ( 1 − χ [ 1 , 2 ] ( x )) Q 0 ( ∂ μ Z b u , Z c u ) + C ( ω ) ∂ μ Z c u b ′ + c ′ ≤ b α , β , γ = 0 ∑ 2 Q b b ′ c ′ α β γ ∂ α β 2 Z b ′ u ∂ γ Z c ′ u + 2 1 [ Δ , C ab c 0 , μ ( ω ) χ [ 1 , 2 ] ( x ) ] ∂ μ Z b u Z c u + C ( ω ) χ [ 1 , 2 ] ( x ) Z c u b ′ + c ′ ≤ b α , β , γ = 0 ∑ 2 Q b b ′ c ′ α β γ ∂ μ ( ∂ α β 2 Z b ′ u ∂ γ Z c ′ u ) + C ( ω ) χ [ 1 , 2 ] ( x ) ∂ μ Z b u b ′ + c ′ ≤ c α , β , γ = 0 ∑ 2 Q c b ′ c ′ α β γ ∂ α β 2 Z b ′ u ∂ γ Z c ′ u ⎭ ⎬ ⎫ + 3 1 ∂ t Z a ( 1 − η ( t , x )) α , β , γ , δ = 0 ∑ 2 Q α β γ δ ω α ω β ω γ ω δ ( ∂ t u ) 3 + Z a α , β , γ , δ = 0 ∑ 2 Q α β γ δ ( ∂ α β 2 u ∂ γ u ∂ δ u − ω α ω β ω γ ω δ ∂ t 2 u ( ∂ t u ) 2 ) . (226)
and
F ~ 0 a ( t , x ) = 1 3 Z a [ η ( t , x ) ∑ α , β , γ , δ = 0 2 Q α β γ δ ω α ω β ω γ ω δ ( ∂ t u ) 3 ] , \widetilde{F}_{0}^{a}(t,x)=\frac{1}{3}Z^{a}\left[\eta(t,x)\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\omega_{\alpha}\omega_{\beta}\omega_{\gamma}\omega_{\delta}(\partial_{t}u)^{3}\right], F 0 a ( t , x ) = 3 1 Z a η ( t , x ) α , β , γ , δ = 0 ∑ 2 Q α β γ δ ω α ω β ω γ ω δ ( ∂ t u ) 3 ,
G a θ ( t , x ) = ∑ b + c ≤ a C ( ω ) Z c u ∂ 2 Z b u ∣ x ∣ , G_{a}^{\theta}(t,x)=\sum_{b+c\leq a}\frac{C(\omega)Z^{c}u\partial^{2}Z^{b}u}{\lvert x\rvert}, G a θ ( t , x ) = b + c ≤ a ∑ ∣ x ∣ C ( ω ) Z c u ∂ 2 Z b u ,
G a α ( t , x ) = ∑ b + c ≤ a C ( ω ) Z c u Z ∂ Z b u ∣ x ∣ , α = 0 , 1 , 2 , G_{a}^{\alpha}(t,x)=\sum_{b+c\leq a}\frac{C(\omega)Z^{c}uZ\partial Z^{b}u}{\lvert x\rvert},\qquad\alpha=0,1,2, G a α ( t , x ) = b + c ≤ a ∑ ∣ x ∣ C ( ω ) Z c u Z ∂ Z b u , α = 0 , 1 , 2 ,
G a , 2 r ( t , x ) = ∑ b + c ≤ a ∑ α , β , γ , δ , μ = 0 2 { C ( ω ) ∂ μ Z c u ∑ b ′ + c ′ + d ≤ b Q b b ′ c ′ d α β γ δ ∂ α β 2 Z b ′ u ∂ γ Z c ′ u ∂ δ Z d u + C ( ω ) χ [ 1 , 2 ] ( x ) Z c u ∑ b ′ + c ′ + d ≤ b Q b b ′ c ′ d α β γ δ ∂ μ ( ∂ α β 2 Z b ′ u ∂ γ Z c ′ u ∂ δ Z d u ) + C ( ω ) χ [ 1 , 2 ] ( x ) ∂ μ Z b u ∑ b ′ + c ′ + d ≤ c Q c b ′ c ′ d α β γ δ ∂ α β 2 Z b ′ u ∂ γ Z c ′ u ∂ δ Z d u } . \begin{aligned}
G_{a,2}^{r}(t,x)={}&\sum_{b+c\leq a}\sum_{\alpha,\beta,\gamma,\delta,\mu=0}^{2}\left\{C(\omega)\partial_{\mu}Z^{c}u\sum_{b'+c'+d\leq b}Q_{bb'c'd}^{\alpha\beta\gamma\delta}\partial_{\alpha\beta}^{2}Z^{b'}u\partial_{\gamma}Z^{c'}u\partial_{\delta}Z^{d}u\right.\\
&+C(\omega)\chi_{[1,2]}(x)Z^{c}u\sum_{b'+c'+d\leq b}Q_{bb'c'd}^{\alpha\beta\gamma\delta}\partial_{\mu}\left(\partial_{\alpha\beta}^{2}Z^{b'}u\partial_{\gamma}Z^{c'}u\partial_{\delta}Z^{d}u\right)\\
&\left.+C(\omega)\chi_{[1,2]}(x)\partial_{\mu}Z^{b}u\sum_{b'+c'+d\leq c}Q_{cb'c'd}^{\alpha\beta\gamma\delta}\partial_{\alpha\beta}^{2}Z^{b'}u\partial_{\gamma}Z^{c'}u\partial_{\delta}Z^{d}u\right\}.
\end{aligned} G a , 2 r ( t , x ) = b + c ≤ a ∑ α , β , γ , δ , μ = 0 ∑ 2 { C ( ω ) ∂ μ Z c u b ′ + c ′ + d ≤ b ∑ Q b b ′ c ′ d α β γ δ ∂ α β 2 Z b ′ u ∂ γ Z c ′ u ∂ δ Z d u + C ( ω ) χ [ 1 , 2 ] ( x ) Z c u b ′ + c ′ + d ≤ b ∑ Q b b ′ c ′ d α β γ δ ∂ μ ( ∂ α β 2 Z b ′ u ∂ γ Z c ′ u ∂ δ Z d u ) + C ( ω ) χ [ 1 , 2 ] ( x ) ∂ μ Z b u b ′ + c ′ + d ≤ c ∑ Q c b ′ c ′ d α β γ δ ∂ α β 2 Z b ′ u ∂ γ Z c ′ u ∂ δ Z d u } .
For the estimates of F ~ a 0 \widetilde{F}_a^0 F a 0 , G a Ξ G_a^\Xi G a Ξ (Ξ = 0 , 1 , 2 , θ \Xi= 0,1,2,\theta Ξ = 0 , 1 , 2 , θ ) and G a , 2 r G_{a,2}^r G a , 2 r , it follows from (91) , (143) , (145) , (208) , (212) and (214) that
∑ ∣ a ∣ ≤ 2 N − 36 ∑ ∣ b ∣ ≤ 13 ∣ Z b F ~ a 0 ( s , y ) ∣ ≲ ε 1 3 ⟨ y ⟩ − 3 / 2 ⟨ s − ∣ y ∣ ⟩ 2 δ 0 − 2 , ∑ ∣ a ∣ ≤ 2 N − 36 ∑ ∣ b ∣ ≤ 10 ∣ Z b G a Ξ ( s , y ) ∣ ≲ ε 1 2 ⟨ y ⟩ − 3 / 2 ⟨ s + ∣ y ∣ ⟩ 8 δ 0 − 1 / 2 ⟨ s − ∣ y ∣ ⟩ 3 δ 0 − 1 , ∑ ∣ a ∣ ≤ 2 N − 36 ∑ ∣ b ∣ ≤ 9 ∣ Z b G a , 2 r ( s , y ) ∣ ≲ ε 1 4 ⟨ y ⟩ − 3 / 2 ⟨ s + ∣ y ∣ ⟩ − 1 / 2 ⟨ s − ∣ y ∣ ⟩ − 2 . (227) \begin{aligned}
\sum_{|a|\leq2N-36}\sum_{|b|\leq13}|Z^b\widetilde{F}_a^0(s,y)|&\lesssim\varepsilon_1^3\langle y\rangle^{-3/2}\langle s-|y|\rangle^{2\delta_0-2},\\
\sum_{|a|\leq2N-36}\sum_{|b|\leq10}|Z^bG_a^\Xi(s,y)|&\lesssim\varepsilon_1^2\langle y\rangle^{-3/2}\langle s+|y|\rangle^{8\delta_0-1/2}\langle s-|y|\rangle^{3\delta_0-1},\\
\sum_{|a|\leq2N-36}\sum_{|b|\leq9}|Z^bG_{a,2}^r(s,y)|&\lesssim\varepsilon_1^4\langle y\rangle^{-3/2}\langle s+|y|\rangle^{-1/2}\langle s-|y|\rangle^{-2}. \tag*{(227)}
\end{aligned} ∣ a ∣ ≤ 2 N − 36 ∑ ∣ b ∣ ≤ 13 ∑ ∣ Z b F a 0 ( s , y ) ∣ ∣ a ∣ ≤ 2 N − 36 ∑ ∣ b ∣ ≤ 10 ∑ ∣ Z b G a Ξ ( s , y ) ∣ ∣ a ∣ ≤ 2 N − 36 ∑ ∣ b ∣ ≤ 9 ∑ ∣ Z b G a , 2 r ( s , y ) ∣ ≲ ε 1 3 ⟨ y ⟩ − 3/2 ⟨ s − ∣ y ∣ ⟩ 2 δ 0 − 2 , ≲ ε 1 2 ⟨ y ⟩ − 3/2 ⟨ s + ∣ y ∣ ⟩ 8 δ 0 − 1/2 ⟨ s − ∣ y ∣ ⟩ 3 δ 0 − 1 , ≲ ε 1 4 ⟨ y ⟩ − 3/2 ⟨ s + ∣ y ∣ ⟩ − 1/2 ⟨ s − ∣ y ∣ ⟩ − 2 . (227)
Next, we focus on the estimate of G a , 1 r G_{a,1}^r G a , 1 r in (226) . In the region ∣ y ∣ ≤ 1 + s / 2 |y|\leq1+s/2 ∣ y ∣ ≤ 1 + s /2 , by using (91) , (143) , (145) and (208) , one obtains
∑ ∣ a ∣ ≤ 2 N − 36 ∑ ∣ b ∣ ≤ 9 ∣ Z b G a , 1 r ( s , y ) ∣ ≲ ε 1 2 ⟨ y ⟩ − 5 / 2 ⟨ s + ∣ y ∣ ⟩ 8 δ 0 − 1 / 2 ⟨ s − ∣ y ∣ ⟩ 3 δ 0 − 1 + ε 1 3 ⟨ y ⟩ − 1 ⟨ s + ∣ y ∣ ⟩ 8 δ 0 − 1 / 2 ⟨ s − ∣ y ∣ ⟩ 4 δ 0 − 2 . (228) \sum_{|a|\leq2N-36}\sum_{|b|\leq9}|Z^bG_{a,1}^r(s,y)|\lesssim\varepsilon_1^2\langle y\rangle^{-5/2}\langle s+|y|\rangle^{8\delta_0-1/2}\langle s-|y|\rangle^{3\delta_0-1}
+\varepsilon_1^3\langle y\rangle^{-1}\langle s+|y|\rangle^{8\delta_0-1/2}\langle s-|y|\rangle^{4\delta_0-2}. \tag*{(228)} ∣ a ∣ ≤ 2 N − 36 ∑ ∣ b ∣ ≤ 9 ∑ ∣ Z b G a , 1 r ( s , y ) ∣ ≲ ε 1 2 ⟨ y ⟩ − 5/2 ⟨ s + ∣ y ∣ ⟩ 8 δ 0 − 1/2 ⟨ s − ∣ y ∣ ⟩ 3 δ 0 − 1 + ε 1 3 ⟨ y ⟩ − 1 ⟨ s + ∣ y ∣ ⟩ 8 δ 0 − 1/2 ⟨ s − ∣ y ∣ ⟩ 4 δ 0 − 2 . (228)
In the region ∣ y ∣ ≥ 1 + s / 2 |y|\geq1+s/2 ∣ y ∣ ≥ 1 + s /2 , since the treatment on the last term in G a , 1 r G_{a,1}^r G a , 1 r is similar to that of (175) , then it follows from the first line of (228) together with (47) , (91) , (143) –(145) , (208) and (218) that
∑ ∣ a ∣ ≤ 2 N − 36 ∑ ∣ b ∣ ≤ 9 ∣ Z b G a , 1 r ( s , y ) ∣ ≲ ε 1 2 ⟨ y ⟩ − 5 / 2 ⟨ s + ∣ y ∣ ⟩ 8 δ 0 − 1 / 2 ⟨ s − ∣ y ∣ ⟩ 3 δ 0 − 1 + ε 1 3 ⟨ y ⟩ − 1 / 2 ⟨ s + ∣ y ∣ ⟩ 9 δ 0 − 2 ⟨ s − ∣ y ∣ ⟩ 3 δ 0 − 1 . (229) \sum_{|a|\leq2N-36}\sum_{|b|\leq9}|Z^bG_{a,1}^r(s,y)|\lesssim\varepsilon_1^2\langle y\rangle^{-5/2}\langle s+|y|\rangle^{8\delta_0-1/2}\langle s-|y|\rangle^{3\delta_0-1}
+\varepsilon_1^3\langle y\rangle^{-1/2}\langle s+|y|\rangle^{9\delta_0-2}\langle s-|y|\rangle^{3\delta_0-1}. \tag*{(229)} ∣ a ∣ ≤ 2 N − 36 ∑ ∣ b ∣ ≤ 9 ∑ ∣ Z b G a , 1 r ( s , y ) ∣ ≲ ε 1 2 ⟨ y ⟩ − 5/2 ⟨ s + ∣ y ∣ ⟩ 8 δ 0 − 1/2 ⟨ s − ∣ y ∣ ⟩ 3 δ 0 − 1 + ε 1 3 ⟨ y ⟩ − 1/2 ⟨ s + ∣ y ∣ ⟩ 9 δ 0 − 2 ⟨ s − ∣ y ∣ ⟩ 3 δ 0 − 1 . (229)
On the other hand, by using (220) , the estimate (217) can be further improved to
∑ ∣ a ∣ ≤ 2 N − 36 ∑ ∣ b ∣ ≤ 8 sup s ∈ [ 0 , t ] ⟨ s ⟩ ∥ ∂ b □ ( Z ~ a − Z a ) v ( s ) ∥ L 2 ( K 4 ) ≲ ε 1 2 . (230) \sum_{|a|\leq2N-36}\sum_{|b|\leq8}\sup_{s\in[0,t]}\langle s\rangle\|\partial^b\Box(\widetilde{Z}^a-Z^a)v(s)\|_{L^2(\mathcal{K}_4)}\lesssim\varepsilon_1^2. \tag*{(230)} ∣ a ∣ ≤ 2 N − 36 ∑ ∣ b ∣ ≤ 8 ∑ s ∈ [ 0 , t ] sup ⟨ s ⟩ ∥ ∂ b □ ( Z a − Z a ) v ( s ) ∥ L 2 ( K 4 ) ≲ ε 1 2 . (230)
Collecting (225) –(230) with (127) , m # = 1 m^\#=1 m # = 1 , ∣ a ∣ ≤ 2 N − 36 |a|\leq2N-36 ∣ a ∣ ≤ 2 N − 36 , η = μ = ν = δ 0 / 3 \eta=\mu=\nu=\delta_0/3 η = μ = ν = δ 0 /3 and G 1 r = □ ( Z ~ a − Z a ) v + G a , 1 r G_1^r=\Box(\widetilde{Z}^a-Z^a)v+G_{a,1}^r G 1 r = □ ( Z a − Z a ) v + G a , 1 r leads to
∑ ∣ a ∣ ≤ 2 N − 36 ∣ ∂ Z a v ( t , x ) ∣ ≲ ε 1 ⟨ x ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ δ 0 − 1 ⟨ t ⟩ 12 δ 0 . (231) \sum_{|a|\leq2N-36}|\partial Z^a v(t,x)|\lesssim\varepsilon_1\langle x\rangle^{-1/2}\langle t-|x|\rangle^{\delta_0-1}\langle t\rangle^{12\delta_0}. \tag*{(231)} ∣ a ∣ ≤ 2 N − 36 ∑ ∣ ∂ Z a v ( t , x ) ∣ ≲ ε 1 ⟨ x ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ δ 0 − 1 ⟨ t ⟩ 12 δ 0 . (231)
where we have used the definition (209) , (91) , (145) , (147) , (212) and (220) . To obtain (224) , one needs to improve ∥ G a i ( s ) ∥ L ∞ ( K 3 ) \|G_a^i(s)\|_{L^\infty(\mathcal{K}_3)} ∥ G a i ( s ) ∥ L ∞ ( K 3 ) (i = 1 , 2 i=1,2 i = 1 , 2 ). It follows from the first line of (228) with (231) instead of (208)
∑ ∣ a ∣ ≤ 2 N − 46 ∑ ∣ b ∣ ≤ 9 ∣ Z b G a , 1 r ( s , y ) ∣ ≲ ε 1 2 ⟨ y ⟩ − 5 / 2 ⟨ s + ∣ y ∣ ⟩ 12 δ 0 − 1 / 2 ⟨ s − ∣ y ∣ ⟩ − 1.4 + ε 1 3 ⟨ y ⟩ − 1 ⟨ s + ∣ y ∣ ⟩ 8 δ 0 − 1 / 2 ⟨ s − ∣ y ∣ ⟩ 4 δ 0 − 2 , ∑ ∣ a ∣ ≤ 2 N − 46 ∑ i = 1 , 2 ∥ G a i ( s ) ∥ L ∞ ( K 3 ) ≲ ε 1 2 ⟨ s ⟩ − 2 . (232) \begin{aligned}
\sum_{|a|\leq2N-46}\sum_{|b|\leq9}|Z^bG_{a,1}^r(s,y)|&\lesssim\varepsilon_1^2\langle y\rangle^{-5/2}\langle s+|y|\rangle^{12\delta_0-1/2}\langle s-|y|\rangle^{-1.4}\\
&\quad+\varepsilon_1^3\langle y\rangle^{-1}\langle s+|y|\rangle^{8\delta_0-1/2}\langle s-|y|\rangle^{4\delta_0-2},\\
\sum_{|a|\leq2N-46}\sum_{i=1,2}\|G_a^i(s)\|_{L^\infty(\mathcal{K}_3)}&\lesssim\varepsilon_1^2\langle s\rangle^{-2}. \tag*{(232)}
\end{aligned} ∣ a ∣ ≤ 2 N − 46 ∑ ∣ b ∣ ≤ 9 ∑ ∣ Z b G a , 1 r ( s , y ) ∣ ∣ a ∣ ≤ 2 N − 46 ∑ i = 1 , 2 ∑ ∥ G a i ( s ) ∥ L ∞ ( K 3 ) ≲ ε 1 2 ⟨ y ⟩ − 5/2 ⟨ s + ∣ y ∣ ⟩ 12 δ 0 − 1/2 ⟨ s − ∣ y ∣ ⟩ − 1.4 + ε 1 3 ⟨ y ⟩ − 1 ⟨ s + ∣ y ∣ ⟩ 8 δ 0 − 1/2 ⟨ s − ∣ y ∣ ⟩ 4 δ 0 − 2 , ≲ ε 1 2 ⟨ s ⟩ − 2 . (232)
where (220) is used. Therefore, analogously to (231) , applying (127) with m # = 0 , 1 m^\#=0,1 m # = 0 , 1 , ∣ a ∣ ≤ 2 N − 46 |a|\leq2N-46 ∣ a ∣ ≤ 2 N − 46 , (225) –(227) , (229) , (230) and (232) yields (224) . This completes the proof of (224) .
Lemma 7.8. Under the assumptions of Theorem 1.1 , let v v v be defined by (142) and suppose that (143) –(146) hold. Then we have
∑ ∣ a ∣ ≤ 2 N − 46 ∣ ∂ ˉ Z a v ∣ ≲ ( ε 1 2 + ε 1 κ ) ⟨ x ⟩ − 1 / 2 ⟨ t + ∣ x ∣ ⟩ δ 0 − 1 (233) \sum_{|a|\leq2N-46} |\bar{\partial}Z^a v| \lesssim(\varepsilon_1^2+\varepsilon_1\kappa)\langle x\rangle^{-1/2}\langle t+|x|\rangle^{\delta_0-1} \tag*{(233)} ∣ a ∣ ≤ 2 N − 46 ∑ ∣ ∂ ˉ Z a v ∣ ≲ ( ε 1 2 + ε 1 κ ) ⟨ x ⟩ − 1/2 ⟨ t + ∣ x ∣ ⟩ δ 0 − 1 (233)
and
∑ ∣ a ∣ ≤ 2 N − 52 ∣ Z a v ∣ ≲ ( ε 1 2 + ε 1 κ ) ⟨ t + ∣ x ∣ ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ 3 δ 0 − 1 / 2 . (234) \sum_{|a|\leq2N-52} |Z^a v| \lesssim(\varepsilon_1^2+\varepsilon_1\kappa)\langle t+|x|\rangle^{-1/2}\langle t-|x|\rangle^{3\delta_0-1/2}. \tag*{(234)} ∣ a ∣ ≤ 2 N − 52 ∑ ∣ Z a v ∣ ≲ ( ε 1 2 + ε 1 κ ) ⟨ t + ∣ x ∣ ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ 3 δ 0 − 1/2 . (234)
Proof. The proof is easier than that of (212) and (218) . First, we prove (233) in the region ∣ x ∣ ≥ 1 + t / 2 |x|\geq1+t/2 ∣ x ∣ ≥ 1 + t /2 since (233) can be achieved directly in the region ∣ x ∣ ≤ 1 + t / 2 |x|\leq1+t/2 ∣ x ∣ ≤ 1 + t /2 by (224) . Note that the estimate (219) may be improved with (218) , (231) instead of (206) , (208) , respectively,
∑ ∣ a ∣ ≤ 2 N − 46 ∣ Z a Q ( ∂ u , ∂ 2 u ) ∣ ≲ ε 1 2 ⟨ t + ∣ x ∣ ⟩ 13 δ 0 − 2 ⟨ t − ∣ x ∣ ⟩ δ 0 − 1 + ε 1 3 ⟨ t + ∣ x ∣ ⟩ − 3 / 2 ⟨ t − ∣ x ∣ ⟩ 2 δ 0 − 2 , ∣ x ∣ ≥ 1 + t / 2. \begin{aligned}
\sum_{|a|\leq2N-46}|Z^aQ(\partial u,\partial^2u)|
&\lesssim\varepsilon_1^2\langle t+|x|\rangle^{13\delta_0-2}\langle t-|x|\rangle^{\delta_0-1}\\
&\quad+\varepsilon_1^3\langle t+|x|\rangle^{-3/2}\langle t-|x|\rangle^{2\delta_0-2},\qquad|x|\geq1+t/2.
\end{aligned} ∣ a ∣ ≤ 2 N − 46 ∑ ∣ Z a Q ( ∂ u , ∂ 2 u ) ∣ ≲ ε 1 2 ⟨ t + ∣ x ∣ ⟩ 13 δ 0 − 2 ⟨ t − ∣ x ∣ ⟩ δ 0 − 1 + ε 1 3 ⟨ t + ∣ x ∣ ⟩ − 3/2 ⟨ t − ∣ x ∣ ⟩ 2 δ 0 − 2 , ∣ x ∣ ≥ 1 + t /2.
Thus, (233) is obtained by (212) and the same argument in the proof of (218) .
Due to
□ Z ~ a v = □ ( Z ~ a − Z a ) v + ∂ t F 0 a + F ~ 1 a , \Box\widetilde{Z}^a v=\Box(\widetilde{Z}^a-Z^a)v+\partial_tF_0^a+\widetilde{F}_1^a, □ Z a v = □ ( Z a − Z a ) v + ∂ t F 0 a + F 1 a ,
F ~ 1 a = ∑ α , β , γ = 0 2 ∑ b + c ≤ a Q a b c α β γ ∂ α β 2 Z b u ∂ γ Z c u + Z a [ ∑ α , β , γ , δ = 0 2 Q α β γ δ ∂ α β 2 u ∂ γ u ∂ δ u − ω α ω β ω γ ω δ ∂ t 2 u ( ∂ t u ) 2 ] (235) \widetilde{F}_1^a=\sum_{\alpha,\beta,\gamma=0}^{2}\sum_{b+c\leq a}Q^{\alpha\beta\gamma}_{abc}\partial_{\alpha\beta}^2Z^b u\partial_\gamma Z^c u
+Z^a\left[\sum_{\alpha,\beta,\gamma,\delta=0}^{2}Q^{\alpha\beta\gamma\delta}\partial_{\alpha\beta}^2u\partial_\gamma u\partial_\delta u-\omega_\alpha\omega_\beta\omega_\gamma\omega_\delta\partial_t^2u(\partial_tu)^2\right] \tag*{(235)} F 1 a = α , β , γ = 0 ∑ 2 b + c ≤ a ∑ Q ab c α β γ ∂ α β 2 Z b u ∂ γ Z c u + Z a α , β , γ , δ = 0 ∑ 2 Q α β γ δ ∂ α β 2 u ∂ γ u ∂ δ u − ω α ω β ω γ ω δ ∂ t 2 u ( ∂ t u ) 2 (235)
where F 0 a F_0^a F 0 a is defined in (213) , then analogously to (175) , it follows from (91) , (224) and (233) that
∑ ∣ a ∣ ≤ 2 N − 52 ∑ ∣ b ∣ ≤ 5 ∣ Z b F ~ 1 a ∣ ≲ ε 1 2 ⟨ x ⟩ − 1 ⟨ t + ∣ x ∣ ⟩ δ 0 − 1 ⟨ t − ∣ x ∣ ⟩ δ 0 − 1 . (236) \sum_{|a|\leq2N-52}\sum_{|b|\leq5}|Z^b\widetilde{F}_1^a|
\lesssim\varepsilon_1^2\langle x\rangle^{-1}\langle t+|x|\rangle^{\delta_0-1}\langle t-|x|\rangle^{\delta_0-1}. \tag*{(236)} ∣ a ∣ ≤ 2 N − 52 ∑ ∣ b ∣ ≤ 5 ∑ ∣ Z b F 1 a ∣ ≲ ε 1 2 ⟨ x ⟩ − 1 ⟨ t + ∣ x ∣ ⟩ δ 0 − 1 ⟨ t − ∣ x ∣ ⟩ δ 0 − 1 . (236)
As in the proof of (212) , combining (119) with □ Z a v = □ Z a u \Box Z^a v=\Box Z^a u □ Z a v = □ Z a u , ∣ a ∣ ≤ 2 N − 52 |a|\leq2N-52 ∣ a ∣ ≤ 2 N − 52 , μ = 1 / 2 − 3 δ 0 \mu=1/2-3\delta_0 μ = 1/2 − 3 δ 0 , ν = δ 0 \nu=\delta_0 ν = δ 0 , (147) , (215) , (220) and (235) derives (234) . □
Proof of Theorem 1.1
Proof of Theorem 1.1. Collecting (224) , (233) and (234) yields that there is a constant C 1 ≥ 1 C_1\geq1 C 1 ≥ 1 such that for t ∈ [ 0 , e κ / ε 2 ) t\in[0,e^{\kappa/\varepsilon^2}) t ∈ [ 0 , e κ / ε 2 ) ,
∑ ∣ a ∣ ≤ 2 N − 46 ∣ ∂ Z a v ∣ ≤ C 1 ( ε 1 2 + ε 1 κ ) ⟨ x ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ δ 0 − 1 , \sum_{|a|\leq2N-46}|\partial Z^a v|\leq C_1(\varepsilon_1^2+\varepsilon_1\kappa)\langle x\rangle^{-1/2}\langle t-|x|\rangle^{\delta_0-1}, ∣ a ∣ ≤ 2 N − 46 ∑ ∣ ∂ Z a v ∣ ≤ C 1 ( ε 1 2 + ε 1 κ ) ⟨ x ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ δ 0 − 1 ,
∑ ∣ a ∣ ≤ 2 N − 46 ∣ ∂ Z a v ∣ ≤ C 1 ε 1 2 ⟨ t ⟩ − 1 , ∣ x ∣ ≤ t / 10 , \sum_{|a|\leq2N-46}|\partial Z^a v|\leq C_1\varepsilon_1^2\langle t\rangle^{-1},\qquad|x|\leq t/10, ∣ a ∣ ≤ 2 N − 46 ∑ ∣ ∂ Z a v ∣ ≤ C 1 ε 1 2 ⟨ t ⟩ − 1 , ∣ x ∣ ≤ t /10 ,
∑ ∣ a ∣ ≤ 2 N − 46 ∣ ∂ ˉ Z a v ∣ ≤ C 1 ( ε 1 2 + ε 1 κ ) ⟨ x ⟩ − 1 / 2 ⟨ t + ∣ x ∣ ⟩ δ 0 − 1 , \sum_{|a|\leq2N-46}|\bar{\partial}Z^a v|\leq C_1(\varepsilon_1^2+\varepsilon_1\kappa)\langle x\rangle^{-1/2}\langle t+|x|\rangle^{\delta_0-1}, ∣ a ∣ ≤ 2 N − 46 ∑ ∣ ∂ ˉ Z a v ∣ ≤ C 1 ( ε 1 2 + ε 1 κ ) ⟨ x ⟩ − 1/2 ⟨ t + ∣ x ∣ ⟩ δ 0 − 1 ,
∑ ∣ a ∣ ≤ 2 N − 52 ∣ Z a v ∣ ≤ C 1 ( ε 1 2 + ε 1 κ ) ⟨ t + ∣ x ∣ ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ 3 δ 0 − 1 / 2 . \sum_{|a|\leq2N-52}|Z^a v|\leq C_1(\varepsilon_1^2+\varepsilon_1\kappa)\langle t+|x|\rangle^{-1/2}\langle t-|x|\rangle^{3\delta_0-1/2}. ∣ a ∣ ≤ 2 N − 52 ∑ ∣ Z a v ∣ ≤ C 1 ( ε 1 2 + ε 1 κ ) ⟨ t + ∣ x ∣ ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ 3 δ 0 − 1/2 .
Choosing ε 1 = ε \varepsilon_1=\varepsilon ε 1 = ε , ε 2 = 2 C 1 ε 1 \varepsilon_2=\sqrt{2C_1}\varepsilon_1 ε 2 = 2 C 1 ε 1 and ε 0 = κ 0 = 1 4 C 1 \varepsilon_0=\kappa_0=\frac{1}{4C_1} ε 0 = κ 0 = 4 C 1 1 . Then for N ≥ 53 N\ge53 N ≥ 53 , (147) is verified and (143) -(146) can be improved to
∑ ∣ a ∣ ≤ N + 1 ∣ ∂ Z a v ∣ ≤ 1 2 ε 1 ⟨ x ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ δ 0 − 1 . \sum_{|a|\le N+1}|\partial Za_v|\le\frac{1}{2}\varepsilon_1\langle x\rangle^{-1/2}\langle t-|x|\rangle^{\delta_0-1}. ∣ a ∣ ≤ N + 1 ∑ ∣ ∂ Z a v ∣ ≤ 2 1 ε 1 ⟨ x ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ δ 0 − 1 .
∑ ∣ a ∣ ≤ N + 1 ∣ ∂ Z a v ∣ ≤ 1 2 ε 2 2 ⟨ t ⟩ − 1 , ∣ x ∣ ≤ t / 10. \sum_{|a|\le N+1}|\partial Za_v|\le\frac{1}{2}\varepsilon_2^2\langle t\rangle^{-1},\qquad|x|\le t/10. ∣ a ∣ ≤ N + 1 ∑ ∣ ∂ Z a v ∣ ≤ 2 1 ε 2 2 ⟨ t ⟩ − 1 , ∣ x ∣ ≤ t /10.
∑ ∣ a ∣ ≤ N + 1 ∣ ∂ ‾ Z a v ∣ ≤ 1 2 ε 1 ⟨ x ⟩ − 1 / 2 ⟨ t + ∣ x ∣ ⟩ δ 0 − 1 . \sum_{|a|\le N+1}|\overline{\partial}Za_v|\le\frac{1}{2}\varepsilon_1\langle x\rangle^{-1/2}\langle t+|x|\rangle^{\delta_0-1}. ∣ a ∣ ≤ N + 1 ∑ ∣ ∂ Z a v ∣ ≤ 2 1 ε 1 ⟨ x ⟩ − 1/2 ⟨ t + ∣ x ∣ ⟩ δ 0 − 1 .
∑ ∣ a ∣ ≤ N + 1 ∣ Z a v ∣ ≤ 1 2 ε 1 ⟨ t + ∣ x ∣ ⟩ − 1 / 2 ⟨ t − ∣ x ∣ ⟩ 3 δ 0 − 1 / 2 . \sum_{|a|\le N+1}|Za_v|\le\frac{1}{2}\varepsilon_1\langle t+|x|\rangle^{-1/2}\langle t-|x|\rangle^{3\delta_0-1/2}. ∣ a ∣ ≤ N + 1 ∑ ∣ Z a v ∣ ≤ 2 1 ε 1 ⟨ t + ∣ x ∣ ⟩ − 1/2 ⟨ t − ∣ x ∣ ⟩ 3 δ 0 − 1/2 .
This, together with the local existence of classical solution to the initial boundary value problem of the hyperbolic equation under the compatibility conditions of initial boundary values, yields that problem (1) has a solution u ∈ ⋂ j = 0 2 N + 1 C j ( [ 0 , e κ / ε 2 ) , H 2 N + 1 − j ( K ) ) u\in\bigcap_{j=0}^{2N+1}C^j([0,e^{\kappa/\varepsilon^2}),H^{2N+1-j}(\mathcal{K})) u ∈ ⋂ j = 0 2 N + 1 C j ([ 0 , e κ / ε 2 ) , H 2 N + 1 − j ( K )) . In addition, (9) , (10) , (11) , (12) , (13) and (14) can be obtained by (224) , (233) , (234) and (220) with (91) , (93) , respectively. □
Conflict of Interest Statement:
The authors declare that there is no conflict of interest in relation to this article.
Data availability statement:
Data sharing is not applicable to this article as no data sets are generated during the current study.