Critical curve for weakly coupled system of semilinear Euler-Poisson-Darboux-Tricomi equations Yuequn Li, Fei Guo Abstract . This paper investigates a weakly coupled system of semilinear Euler-Poisson-Darboux-Tricomi equations (EPDTS) with power-type nonlinear terms. More precisely, in the case where the damping terms dominate over the mass terms, the critical curve in the p − q p-q plane that delineates the threshold between global existence and blow-up for the EPDTS is given by Γ m ​ ( n , p , q , β 1 , β 2 ) = 0 , \Gamma_{m}(n,p,q,\beta_{1},\beta_{2})=0, where Γ m \Gamma_{m} is defined by ( 1.16 ). Through the construction of new test functions, the blow-up problem is addressed when Γ m ​ ( n , p , q , β 1 , β 2 ) ≥ 0 \Gamma_{m}(n,p,q,\beta_{1},\beta_{2})\geq 0 . Based on the ( L 1 ∩ L 2 ) − L 2 (L^{1}\cap L^{2})-L^{2} estimates of the solution to the corresponding linear equation established in our previous work [ 17 ] , we derive the global existence of solutions with small initial data when Γ m ​ ( n , p , q , β 1 , β 2 ) < 0 \Gamma_{m}(n,p,q,\beta_{1},\beta_{2})<0 , provided that the damping terms prevail over the mass terms. Keywords : Semilinear Euler-Poisson-Darboux-Tricomi equations; Weakly coupled system; Critical curve; Test functions; (L^{1}\cap L^{2})-L^{2} estimates. AMS Subject Classification (2020): 35B44, 35G55, 35L52 1. Introduction This paper investigates the critical curve problem for the weakly coupled system of semilinear Euler-Poisson-Darboux-Tricomi equations \left\{\begin{aligned} &\partial_{t}^{2}u-t^{2m}\Delta u+\frac{\mu_{1}}{t}\partial_{t}u+\frac{\nu_{1}^{2}}{t^{2}}u=|v|^{p},&\quad&t>1,\ x\in\mathbb{R}^{n},\\ &\partial_{t}^{2}v-t^{2m}\Delta v+\frac{\mu_{2}}{t}\partial_{t}v+\frac{\nu_{2}^{2}}{t^{2}}v=|u|^{q},&\quad&t>1,\ x\in\mathbb{R}^{n},\\ &(u,\partial_{t}u,v,\partial_{t}v)(1,x)=(u_{0},u_{1},v_{0},v_{1})(x),&\quad&x\in\mathbb{R}^{n},\end{aligned}\right. (1.1) where m>-1 , n\geq 1 denotes the space dimension, \mu_{i},\nu_{i}^{2} , i=1,2 are nonnegative numbers and p,q>1 . We find that there exists a curve related to \mu_{i},\nu_{i} , i=1,2 in p-q plane, which can be used to distinguish the blow-up phenomenon and the global existence of solutions to ( 1.1 ). The critical exponent for the classical semilinear wave equation \partial_{t}^{2}u-\Delta u=|u|^{p} (1.2) is the well-known Strauss exponent p_{S}(n) , where n\geq 1 denotes the space dimension. Namely, local solutions with small initial data exist globally when p>p_{S}(n) and the blow-up phenomenon occurs if 10,\ x\in\mathbb{R}^{n},\\ &\partial_{t}^{2}v-\Delta v=|u|^{q},&\quad&t>0,\ x\in\mathbb{R}^{n},\\ &(u,\partial_{t}u,v,\partial_{t}v)(0,x)=(u_{0},u_{1},v_{0},v_{1})(x),&\quad&x\in\mathbb{R}^{n}\end{aligned}\right. (1.4) with p,q>1 . The critical curve for ( 1.4 ) in the p-q plane is described by \Gamma_{W}(n,p,q)=0 , where \Gamma_{W}(n,p,q):=\max\Big\{\frac{p+2+q^{-1}}{pq-1},\frac{q+2+p^{-1}}{pq-1}\Big\}-\frac{n-1}{2}. (1.5) Roughly speaking, the small data global solutions to ( 1.4 ) exist when \Gamma_{W}(n,p,q)<0 , while if \Gamma_{W}(n,p,q)\geq 0 , solutions blow up in finite time. Observe that \Gamma_{W}(n,p,p)=0 if and only if ( 1.3 ) holds, so the critical curve problem for ( 1.4 ) can be regarded as a generalization of the Strauss exponent problem. However, the study of coupled system ( 1.4 ) is not a simple generalization of the results of its single equation ( 1.2 ). On the contrary, the critical curve described by \Gamma_{W}(n,p,q)\geq 0 expands the blow-up range compared to the single case: 10 . Consider the weakly coupled system of semilinear damped wave equations \left\{\begin{aligned} &\partial_{t}^{2}u-\Delta u+b(t)\partial_{t}u=|v|^{p},&\quad&t>0,\ x\in\mathbb{R}^{n},\ p>1,\\ &\partial_{t}^{2}v-\Delta v+b(t)\partial_{t}v=|u|^{q},&\quad&t>0,\ x\in\mathbb{R}^{n},\ q>1,\\ &(u,v,\partial_{t}u,\partial_{t}v)(0,x)=(u_{0},u_{1},v_{0},v_{1})(x),&\quad&x\in\mathbb{R}^{n}.\end{aligned}\right. (1.8) If b(t)=1, in contrast to ( 1.4 ), its critical curve is described by \Gamma_{DW}(n,p,q)=0 , where \Gamma_{DW}(n,p,q):=\max\Big\{\frac{p+1}{pq-1},\frac{q+1}{pq-1}\Big\}-\frac{n}{2}, highlighting that the appearance of the damping terms \partial_{t}u,\partial_{t}v has a notable impact on the coupled system ( 1.8 ). About the details, one can refer to [ 20 , 21 , 23 , 24 , 25 , 32 ] . Note \Gamma_{DW}(n,p,p)=0\iff p=p_{F}(n) , where p_{F}(n):=1+\frac{2}{n} denotes the Fujita exponent, which is the critical exponent for the heat equation \partial_{t}u-\Delta u=|u|^{p} and the single damped wave equation \partial_{t}^{2}u-\Delta u+\partial_{t}u=|u|^{p}, (1.9) one can see [ 7 , 14 , 34 , 36 ] . Hence we can claim that the critical curve \Gamma_{DW}(n,p,q)=0 generalizes the critical exponent for ( 1.9 ). Furthermore, in the case of general effective damping, i.e., b(t)=(1+t)^{-\beta} , \beta\in(-1,1) , the authors [ 24 ] showed that the critical curve is still described by \Gamma_{DW}(n,p,q)=0 . This explains why the damping term is referred to as effective in this case. In addition, the authors [ 18 , 19 ] established relevant results of global solutions under different power nonlinearities or additional regularity assumptions on the initial data. For the weakly coupled system of semilinear wave equations with scale-invariant damping and mass \left\{\begin{aligned} &\partial_{t}^{2}u-\Delta u+\frac{\mu_{1}}{t}\partial_{t}u+\frac{\nu_{1}^{2}}{t^{2}}u=|v|^{p},&\quad&t>1,\ x\in\mathbb{R}^{n},\ p>1,\\ &\partial_{t}^{2}v-\Delta v+\frac{\mu_{2}}{t}\partial_{t}v+\frac{\nu_{2}^{2}}{t^{2}}v=|u|^{q},&\quad&t>1,\ x\in\mathbb{R}^{n},\ q>1,\\ &(u,v,\partial_{t}u,\partial_{t}v)(1,x)=(u_{0},u_{1},v_{0},v_{1})(x),&\quad&x\in\mathbb{R}^{n}.\end{aligned}\right. (1.10) Chen-Palmieri [ 2 ] showed that, in the case of \delta_{i}:=(\mu_{i}-1)^{2}-4\nu_{i}^{2}\geq(n+1)^{2},\ i=1,2, (1.11) the critical curve for ( 1.10 ) is given by \Gamma_{0}(n,p,q,\beta_{1},\beta_{2})=0 , where \Gamma_{0}(n,p,q,\beta_{1},\beta_{2}):=\max\Big\{\frac{p+1}{pq-1}-\frac{\beta_{1}-1}{2},\frac{q+1}{pq-1}-\frac{\beta_{2}-1}{2}\Big\}-\frac{n}{2}, (1.12) where \beta_{i}=\frac{\mu_{i}+1-\sqrt{\delta_{i}}}{2},i=1,2 . Clearly, the study of the critical curve for ( 1.10 ) is based on the corresponding single power-type semilinear wave equation \partial_{t}^{2}u-\Delta u+\frac{\mu}{t}\partial_{t}u+\frac{\nu^{2}}{t^{2}}u=|u|^{p}. (1.13) Since the linear equation of ( 1.13 ) is invariant under the transformation \tilde{u}(\lambda t,\lambda x) , the damping term \frac{\mu}{t}\partial_{t}u and the mass term \frac{\nu^{2}}{t^{2}}u can be viewed as having the same scaling. Following Palmieri’s analysis, this balance induces an interaction where the relative sizes of \mu and \nu^{2} play a decisive role in determining whether the damping term or the mass term prevails in ( 1.13 ). In fact, the magnitude of the number \delta:=(\mu-1)^{2}-4\nu^{2} can characterize this interaction. Specifically, Nascimento-Palmieri-Reissig [ 22 , 26 , 29 ] proved that for any \delta\geq 0, the local (in time) solutions to ( 1.13 ) blow up in finite time when 1p_{F}(n+\frac{\mu-1-\sqrt{\delta}}{2}) for any \delta\geq(n+2\sigma-1)^{2} , where \sigma is a positive number describing the regularity of the initial data, and p_{F} is Fujita exponent mentioned above. Roughly speaking, when the damping term dominates over the mass term, in other words, \delta is relatively large, ( 1.13 ) is parabolic-like as p_{F}(n+\frac{\mu-1-\sqrt{\delta}}{2}) is its critical exponent. In [ 30 ] , Palmieri-Tu also proved the blow-up phenomenon if 11,\ m>-1,\\ &u(1,x)=u_{0}(x),\ \partial_{t}u(1,x)=u_{1}(x),&\quad&x\in\mathbb{R}^{n}.\end{aligned}\right. (1.14) By applying Fourier analysis methods, the corresponding linear problem \partial_{t}^{2}u-t^{2m}\Delta u+\frac{\mu}{t}\partial_{t}u+\frac{\nu^{2}}{t^{2}}u=0 (1.15) associated with ( 1.14 ) is reduced to a Bessel equation in the frequency space. Using the properties of Bessel functions, we obtained the (L^{1}\cap L^{2})-L^{2} estimates of the solutions to ( 1.15 ). On this basis, we established the global existence of solutions to ( 1.14 ) by Duhamel’s principle and contraction mapping principle, provided that p>p_{F}((m+1)n+\frac{\mu-1-\sqrt{\delta}}{2}) in the case of \delta\geq(m+1)^{2}(n+2\sigma-1)^{2} , where \sigma>0 . Combining this with the blow-up result established by Palmieri [ 27 ] , namely, solutions blow up in finite time if 10 such that a\leq Cb . (2) |D|^{\sigma} denotes the pseudo-differential operators with symbol |\xi|^{\sigma} . H^{\sigma}_{p}{(\mathbb{R}^{n})} , \dot{H}^{\sigma}_{p}{(\mathbb{R}^{n})} mean the non-homogeneous Sobolev space and homogeneous Sobolev space respectively, equipped with the norm \|f\|_{H^{\sigma}_{p}{(\mathbb{R}^{n})}}=\|(1+|D|^{2})^{\frac{\sigma}{2}}f\|_{L^{p}(\mathbb{R}^{n})} , \|f\|_{\dot{H}^{\sigma}_{p}{(\mathbb{R}^{n})}}=\||D|^{\sigma}f\|_{L^{p}(\mathbb{R}^{n})} . For simplicity in writing, we omit \mathbb{R}^{n} . (3) \lceil\cdot\rceil is the ceiling function, i.e., \lceil x\rceil:=\min\{k\in\mathbb{Z}:x\leq k\} ; [\cdot]_{+} is the positive part function, i.e., [x]_{+}:=\max\{x,0\} . (4) B_{r} represents the ball centered at the origin with radius r , that is B_{r}=\{x\in\mathbb{R}^{n}:|x|\leq r\} . This paper is organized as follows. In Section 2 , we provide the definition of the energy solution to ( 1.1 ) and present the main results. Section 3 is devoted to give the blow-up result, that is Theorem 2.1 . In Section 4 , we list the results established in our recent work [ 17 ] , then we establish the global well-posedness with high regularity of initial data, i.e., Theorems 2.2 - 2.4 . And we establish the global well-posedness with low regularity of initial data in Section 5, i.e., Theorems 2.5 - 2.7 . In the appendix, we provide a brief outline of the proof of the local (in time) existence of solutions to ( 1.1 ), namely, Proposition 2.1 . 2. Main results In this section, we present the blow-up result and global existence results for ( 1.1 ), thereby determining its critical curve. We first give the definition of energy solution to ( 1.1 ). Let us introduce the space D^{\sigma}=\left\{\begin{aligned} &\bigl(L^{1}\cap H^{\sigma}\bigr)\times\bigl(L^{1}\cap H^{\sigma-1}\bigr),&\quad&\sigma\geq 1,\\ &\bigl(L^{1}\cap H^{\sigma}\bigr)\times\bigl(L^{1}\cap L^{2}\bigr),&\quad&\sigma\in(0,1)\end{aligned}\right. (2.1) with the norm \|(f,g)\|_{D^{\sigma}}=\left\{\begin{aligned} &\|f\|_{L^{1}}+\|f\|_{H^{\sigma}}+\|g\|_{L^{1}}+\|g\|_{H^{\sigma-1}},&\quad&\sigma\geq 1,\\ &\|f\|_{L^{1}}+\|f\|_{H^{\sigma}}+\|g\|_{L^{1}}+\|g\|_{L^{2}},&\quad&\sigma\in(0,1),\end{aligned}\right. (2.2) and D^{1}:=D. Definition 2.1 . Let (u_{0},u_{1},v_{0},v_{1})\in D\times D be compactly supported with \text{supp}(u_{0},u_{1},v_{0},v_{1})\subset B_{M}(0) (2.3) for some M>0 . We say that (u,v) is an energy solution to ( 1.1 ) on [1,T) if \displaystyle(u,v)\in\Big(C\bigl([1,T);H^{1}(\mathbb{R}^{n})\bigr)\cap C^{1}\bigl([1,T);L^{2}(\mathbb{R}^{n})\bigr)\Big)^{2}, (2.4) and it satisfies the support property \text{supp}(u,v)(t,\cdot)\subset B_{\phi_{m}(t)-\phi_{m}(1)+M},\ \phi_{m}(t)=\frac{t^{m+1}}{m+1},\ \text{for any}\ t\in(1,T) (2.5) as well as the following integral equalities \begin{split}&\iint_{[1,T)\times\mathbb{R}^{n}}u(t,x)\Bigl(\partial_{t}^{2}\varPhi_{1}(t,x)-t^{2m}\Delta\varPhi_{1}(t,x)-\partial_{t}\bigl(\frac{\mu_{1}}{t}\varPhi_{1}(t,x)\bigr)+\frac{\nu_{1}^{2}}{t^{2}}\varPhi_{1}(t,x)\Bigr)dxdt\\ &=\int_{\mathbb{R}^{n}}\Bigl(-u_{0}(x)\partial_{t}\varPhi_{1}(1,x)+\bigl(\mu_{1}u_{0}(x)+u_{1}(x)\bigr)\varPhi_{1}(1,x)\Bigr)dx\\ &\quad+\iint_{[1,T)\times\mathbb{R}^{n}}|v(t,x)|^{p}\varPhi_{1}(t,x)dxdt\end{split} (2.6) and \begin{split}&\iint_{[1,T)\times\mathbb{R}^{n}}v(t,x)\Bigl(\partial_{t}^{2}\varPhi_{2}(t,x)-t^{2m}\Delta\varPhi_{2}(t,x)-\partial_{t}\bigl(\frac{\mu_{2}}{t}\varPhi_{2}(t,x)\bigr)+\frac{\nu_{2}^{2}}{t^{2}}\varPhi_{2}(t,x)\Bigr)dxdt\\ &=\int_{\mathbb{R}^{n}}\Bigl(-v_{0}(x)\partial_{t}\varPhi_{2}(1,x)+\bigl(\mu_{2}v_{0}(x)+v_{1}(x)\bigr)\varPhi_{2}(1,x)\Bigr)dx\\ &\quad+\iint_{[1,T)\times\mathbb{R}^{n}}|u(t,x)|^{q}\varPhi_{2}(t,x)dxdt,\end{split} (2.7) for any \varPhi_{1}(t,x),\varPhi_{2}(t,x)\in C_{0}^{\infty}\bigl([1,T)\times\mathbb{R}^{n}\bigr) . Remark 2.1 . The support property ( 2.5 ) of the solution (u,v) reflects the finite propagation speed of the wave, which has been explained in [ 11 , 12 , 27 ] . Proposition 2.1 . (Local existence). Let n\geq 1,m>-1 , \mu_{i},\nu_{i}^{2}\geq 0 , i=1,2 such that \delta_{1},\delta_{2}>0 , where \delta_{i},i=1,2 are defined by ( 1.11 ). Suppose the initial data (u_{0},u_{1},v_{0},v_{1})\in D^{\sigma}\times D^{\sigma} with some \sigma>0 and satisfy the support condition ( 2.3 ). \text{If}\ \ \sigma\geq 1,\ \text{we assume}\ p,q>\lceil\sigma\rceil\ \text{and}\ \left\{\begin{aligned} &12\sigma,\\ &12\sigma,\\ &11 and a unique solution (u,v)\in\left\{\begin{aligned} &\Big(C\big([1,T);H^{\sigma}\big)\cap C^{1}\big([1,T);H^{\sigma-1}\big)\Big)^{2},\ &\sigma\geq 1,\\ &\Big(C\big([1,T);H^{\sigma}\big)\Big)^{2},\ &0<\sigma<1\end{aligned}\right. to ( 1.1 ). Moreover, (u,v) satisfies the support property ( 2.5 ) and the integral equalities ( 2.6 ), ( 2.7 ). Theorem 2.1 . (Blow-up). Let n\geq 1,m>-1 , \nu_{1}^{2},\nu_{2}^{2}\geq 0 and \left\{\begin{aligned} &\mu_{1},\mu_{2}>1,&\ &\text{if}\ -10,&\ &\text{if}\ m\geq 0\end{aligned}\right. such that \delta_{1},\delta_{2}>0 , where \delta_{1},\delta_{2} are defined by ( 1.11 ). Suppose that (u_{0},u_{1},v_{0},v_{1})\in D\times D are initial data satisfying the support condition ( 2.5 ) and \displaystyle\liminf\limits_{R\rightarrow+\infty}\int_{|x|0, (2.8) \displaystyle\liminf\limits_{R\rightarrow+\infty}\int_{|x|0. If p,q>1 satisfy \displaystyle p>\frac{1+\beta_{1}}{2m+1+\beta_{2}},\ q>\frac{1+\beta_{2}}{2m+1+\beta_{1}} (2.9) and \displaystyle\Gamma_{m}(n,p,q,\beta_{1},\beta_{2})\geq 0, (2.10) then the energy solution (u,v) to ( 1.1 ) blows up in finite time, where \Gamma_{m}(n,p,q,\beta_{1},\beta_{2}) is defined by ( 1.16 ). Remark 2.2 . \mu_{i}>1,i=1,2 when m\in(-1,0) are technical conditions introduced to ensure that 2m+1+\beta_{i}>0,i=1,2 in ( 2.9 ). We illustrate the non-emptiness of the conditions imposed in Theorem 2.1 with relevant examples. In other words, we claim that the solution to ( 1.1 ) blows up in the case of Examples 2.1 - 2.2 . Example 2.1 . In the case of m>0 , for example, m=0.1,n=2,\mu_{1}=3,\nu_{1}=0.25,\mu_{2}=0.5,\nu_{2}=0.125 , a straightforward calculation reveals \delta_{1}=\frac{15}{4} , \delta_{2}=\frac{3}{16} , \beta_{1}=2-\frac{\sqrt{15}}{4} , \beta_{2}=\frac{3}{4}-\frac{\sqrt{3}}{8} . Choosing p=2.1>\frac{1+\beta_{1}}{2m+1+\beta_{2}}=\frac{3-\frac{\sqrt{15}}{4}}{1.95-\frac{\sqrt{3}}{8}}\approx 1.172,q=2.2>\frac{1+\beta_{2}}{2m+1+\beta_{1}}=\frac{1.75-\frac{\sqrt{3}}{8}}{3.2-\frac{\sqrt{15}}{4}}\approx 0.687 , we obtain \Gamma_{m}(n,p,q,\beta_{1},\beta_{2})\approx 0.017>0 . Example 2.2 . In the case of m\in(-1,0) , such as m=-0.6,n=3,\mu_{1}=2,\nu_{1}=0.125,\mu_{2}=1.8,\nu_{2}=0.25 , then \delta_{1}=\frac{15}{16},\delta_{2}=0.39,\beta_{1}=\frac{3}{2}-\frac{\sqrt{15}}{8},\beta_{2}=1.4-\frac{\sqrt{0.39}}{2} . Choosing p=2.3>\frac{1+\beta_{1}}{2m+1+\beta_{2}}\approx 2.271 and q=2.7>\frac{1+\beta_{2}}{2m+1+\beta_{1}}\approx 2.559 , we have \Gamma_{m}(n,p,q,\beta_{1},\beta_{2})\approx 0.066>0 . After establishing the blow-up result, we anticipate that \Gamma_{m}(n,p,q,\beta_{1},\beta_{2})=0 (2.11) is the critical curve for ( 1.1 ). Let \displaystyle\tilde{p}:=\tilde{p}(m,n,\beta_{1},\beta_{2})=\frac{(m+1)n+\beta_{1}+1}{(m+1)n+\beta_{2}-1}=1+\frac{2+\beta_{1}-\beta_{2}}{(m+1)n+\beta_{2}-1}, (2.12) \displaystyle\tilde{q}:=\tilde{q}(m,n,\beta_{1},\beta_{2})=\frac{(m+1)n+\beta_{2}+1}{(m+1)n+\beta_{1}-1}=1+\frac{2+\beta_{2}-\beta_{1}}{(m+1)n+\beta_{1}-1}. By a straightforward algebraic computation, we find that p\leq\tilde{p}\ \text{and}\ q\leq\tilde{q}\implies\Gamma_{m}(n,p,q,\beta_{1},\beta_{2})\geq 0, (2.13) Hence the necessary conditions for \Gamma_{m}(n,p,q,\beta_{1},\beta_{2})<0 are as follows \displaystyle p>\tilde{p}\ \ \text{and}\ \ q>\tilde{q}; (2.14) or, \displaystyle p\leq\tilde{p}\ \ \text{and}\ \ q>\tilde{q}; (2.15) or, \displaystyle p>\tilde{p}\ \ \text{and}\ \ q\leq\tilde{q}. (2.16) In other words, only the above three cases ( 2.14 )-( 2.16 ) can result in \Gamma_{m}(n,p,q,\beta_{1},\beta_{2})<0 . Next, we categorize and present the global existence results according to the different ranges of \sigma . For (u_{0},u_{1},v_{0},v_{1})\in D^{\sigma}\times D^{\sigma} with \sigma\geq 1 , we have the three global existence results. Theorem 2.2 . (Global existence for p>\tilde{p} and q>\tilde{q} (I) ). Let n\geq 1,\sigma\geq 1,m>-1 , \mu_{1}, \mu_{2}>1,\nu_{1}^{2},\nu_{2}^{2}\geq 0 such that both \delta_{1} and \delta_{2}\geq(m+1)^{2}(n+2\sigma-1)^{2} . Let p,q>1 with p,q>\lceil\sigma\rceil , and assume that if n>2\sigma , it holds that p,q\leq 1+\frac{2}{n-2\sigma} . For p>\tilde{p} and q>\tilde{q} , there exists a small constant \varepsilon>0 such that for any initial data (u_{0},u_{1},v_{0},v_{1})\in D^{\sigma}\times D^{\sigma} satisfying the support condition ( 2.3 ), if \|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\leq\varepsilon , then ( 1.1 ) has a unique global solution (u,v)\in\Big(C\bigl([1,\infty);H^{\sigma}(\mathbb{R}^{n})\bigr)\cap C^{1}\bigl([1,\infty);H^{\sigma-1}(\mathbb{R}^{n})\bigr)\Big)^{2}, and (u,v) satisfies the support property \text{supp}(u,v)(t,\cdot)\subset B_{\phi_{m}(t)-\phi_{m}(1)+R},\ \ \text{for any}\ t\in(1,\infty), (2.17) where \phi_{m}(t) is defined by ( 2.5 ) and \tilde{p},\tilde{q} are defined by ( 2.12 ). Furthermore, the solution ( u,v ) satisfies the decay estimates \displaystyle\|u(t,\cdot)\|_{L^{2}}\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\big(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\bigr), (2.18) \displaystyle\|u(t,\cdot)\|_{\dot{H}^{\sigma}}\lesssim t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\ell_{1}(t)\big(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\bigr), (2.19) \displaystyle\|\partial_{t}u(t,\cdot)\|_{L^{2}}\lesssim t^{m-(m+1)(1+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\big(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\bigr), (2.20) \displaystyle\|\partial_{t}u(t,\cdot)\|_{\dot{H}^{\sigma-1}}\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\ell_{1}(t)\big(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\bigr), (2.21) \displaystyle\|v(t,\cdot)\|_{L^{2}}\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\big(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\bigr), (2.22) \displaystyle\|v(t,\cdot)\|_{\dot{H}^{\sigma}}\lesssim t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\ell_{2}(t)\big(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\bigr), (2.23) \displaystyle\|\partial_{t}v(t,\cdot)\|_{L^{2}}\lesssim t^{m-(m+1)(1+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\big(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\bigr), (2.24) \displaystyle\|\partial_{t}v(t,\cdot)\|_{\dot{H}^{\sigma-1}}\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\ell_{2}(t)\big(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\bigr), (2.25) where \ell_{i}(t)=\left\{\begin{aligned} &1,\ &\text{if}\ \delta_{i}>\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\\ &\bigl(1+\log{t}\bigr)^{\frac{1}{2}},\ &\text{if}\ \delta_{i}=\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\end{aligned}\right.\ \ i=1,2. (2.26) Remark 2.3 . The assumptions \mu_{1},\mu_{2}>1 are technical conditions imposed to guarantee that the denominators (m+1)n+\beta_{1}-1 and (m+1)n+\beta_{2}-1 in the definitions of \tilde{p} and \tilde{q} remain positive, thereby ensuring the validity of the integral exponents ( 4.50 ) and ( 4.58 ). It should be emphasized that since our analysis focuses on the case where \delta_{1},\delta_{2} are relatively large, imposing \mu_{1},\mu_{2}>1 is naturally reasonable. We need to provide an example to demonstrate the validity of the conditions in Theorem 2.2 . Example 2.3 . For \sigma=1.1,m=0.4,n=3>2\sigma,\mu_{1}=15,\nu_{1}=6,\mu_{2}=8.7,\nu_{2}=0.25 , we calculate that \delta_{1}=52,\delta_{2}=59.04>34.5744=(m+1)^{2}(n+2\sigma-1)^{2},\beta_{1}=\frac{16-\sqrt{52}}{2},\beta_{2}=\frac{9.7-\sqrt{59.04}}{2},\tilde{p}=\frac{5.2+\frac{16-\sqrt{52}}{2}}{3.2+\frac{9.7-\sqrt{59.04}}{2}}\approx 2.28,\tilde{q}=\frac{5.2+\frac{9.7-\sqrt{59.04}}{2}}{3.2+\frac{16-\sqrt{52}}{2}}\approx 0.82,1+\frac{2}{n-2\sigma}=3.5 . Then for any 2=\lceil 1.1\rceil<\frac{5.2+\frac{16-\sqrt{52}}{2}}{3.2+\frac{9.7-\sqrt{59.04}}{2}}\tilde{q} (I)) . Let n\geq 1,\sigma\geq 1,m>-1, \mu_{1}, \mu_{2}>1,\nu_{1}^{2},\nu_{2}^{2}\geq 0 such that \delta_{1}\geq(m+1)^{2}(n+2\sigma-1)^{2},\delta_{2}>(m+1)^{2}(n+2\sigma-1)^{2} . Let p,q>1 with p,q>\lceil\sigma\rceil , and assume that if n>2\sigma , p,q\leq 1+\frac{2}{n-2\sigma} . For p\leq\tilde{p} and q>\tilde{q} satisfying \frac{q+1}{pq-1}<\frac{(m+1)n+\beta_{2}-1}{2}, (2.27) there exists a small constant \varepsilon>0 such that for any initial data (u_{0},u_{1},v_{0},v_{1})\in D^{\sigma}\times D^{\sigma} satisfying the support condition ( 2.3 ), and if \|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\leq\varepsilon , then ( 1.1 ) admits a unique global solution (u,v)\in\Big(C\bigl([1,\infty);H^{\sigma}(\mathbb{R}^{n})\bigr)\cap C^{1}\bigl([1,\infty);H^{\sigma-1}(\mathbb{R}^{n})\bigr)\Big)^{2}, and ( u,v ) satisfies the support property ( 2.17 ), where \tilde{p},\tilde{q} are defined by ( 2.12 ). Moreover, the solution v satisfies the decay estimates ( 2.22 )- ( 2.25 ), and u satisfies the estimates \displaystyle\|u(t,\cdot)\|_{L^{2}}\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}+\alpha_{1}}\big(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\bigr), (2.28) \displaystyle\|u(t,\cdot)\|_{\dot{H}^{\sigma}}\lesssim t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}+\alpha_{1}}\ell_{1}(t)\big(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\bigr), (2.29) \displaystyle\|\partial_{t}u(t,\cdot)\|_{L^{2}}\lesssim t^{m-(m+1)(1+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}+\alpha_{1}}\big(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\bigr), \displaystyle\|\partial_{t}u(t,\cdot)\|_{\dot{H}^{\sigma-1}}\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}+\alpha_{1}}\ell_{1}(t)\big(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\bigr), where \ell_{1}(t) is defined by ( 2.26 ) and \alpha_{1}=\left\{\begin{aligned} &\big((m+1)n+\beta_{2}-1\big)(\tilde{p}-p),&\quad&\text{if}\ \ p<\tilde{p},\\ &\epsilon,&\quad&\text{if}\ \ p=\tilde{p}\end{aligned}\right. (2.30) with \epsilon>0 being an sufficiently small constant. Remark 2.5 . In the proof of Theorem 2.3 shown in Section 4.3 , we observe that when p\leq\tilde{p} , the integral corresponding to index ( 4.115 ), i.e., \int_{1}^{t}\tau^{(m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p}d\tau , cannot be controlled by a convergent improper integral. Therefore, it is necessary to rely on the technical condition ( 2.31 ) to ensure the integrability of another improper integral ( 4.123 ). However, the cost of this approach is the emergence of a non-decaying factor t^{\alpha_{1}} in the estimate of u(t,\cdot) . It is still necessary to provide an example to illustrate the reasonableness of conditions stated in Theorem 2.3 . Example 2.4 . For \sigma=1.4 , m=0.3,n=4>2\sigma,\mu_{1}=20,\mu_{2}=9,\nu_{1}=8.5,\nu_{2}=1 , we calculate that \delta_{1}=72,\delta_{2}=60>(m+1)^{2}(n+2\sigma-1)^{2}=56.8516,\beta_{1}=\frac{21-\sqrt{72}}{2},\beta_{2}=\frac{10-\sqrt{60}}{2},\tilde{p}=\frac{6.2+\frac{21-\sqrt{72}}{2}}{4.2+\frac{10-\sqrt{60}}{2}}\approx 2.339,\tilde{q}=\frac{6.2+\frac{10-\sqrt{60}}{2}}{4.2+\frac{21-\sqrt{72}}{2}}\approx 0.7007 . Choosing p=2.1<\tilde{p},q=2.2>\tilde{q}, we have p,q>\lceil 1.4\rceil=2 , p,q<1+\frac{2}{n-2\sigma}\approx 2.667 and \frac{q+1}{pq-1}=\frac{3.2}{3.62}\approx 0.884<\frac{(m+1)n+\beta_{2}-1}{2}\approx 2.6635 . Hence the solution ( u,v ) to ( 1.1 ) exists globally for p=2.1,q=2.2 . Theorem 2.4 . (Global existence for p>\tilde{p} and q\leq\tilde{q} (I) ). Let n\geq 1,\sigma\geq 1,m>-1,\mu_{1}, \mu_{2}>1,\nu_{1},\nu_{2}\geq 0 such that \delta_{1}>(m+1)^{2}(n+2\sigma-1)^{2},\delta_{2}\geq(m+1)^{2}(n+2\sigma-1)^{2} . Let p,q>1 with p,q>\lceil\sigma\rceil , and assume that if n>2\sigma , p,q\leq 1+\frac{2}{n-2\sigma} . For p>\tilde{p} and q\leq\tilde{q} satisfying \frac{p+1}{pq-1}<\frac{(m+1)n+\beta_{1}-1}{2}, (2.31) there exists a small constant \varepsilon>0 such that for any initial data (u_{0},u_{1},v_{0},v_{1})\in D^{\sigma}\times D^{\sigma} satisfying the support condition ( 2.3 ), and if \|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\leq\varepsilon , then ( 1.1 ) admits a unique global solution (u,v)\in\Big(C\bigl([1,\infty);H^{\sigma}(\mathbb{R}^{n})\bigr)\cap C^{1}\bigl([1,\infty);H^{\sigma-1}(\mathbb{R}^{n})\bigr)\Big)^{2}, and ( u,v ) satisfies the support property ( 2.17 ), where \tilde{p},\tilde{q} are defined by ( 2.12 ). Moreover, the solution u satisfies the decay estimates ( 2.18 )- ( 2.21 ), and v satisfies the estimates \displaystyle\|v(t,\cdot)\|_{L^{2}}\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}+\alpha_{2}}\big(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\bigr) (2.32) \displaystyle\|v(t,\cdot)\|_{\dot{H}^{\sigma}}\lesssim t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}+\alpha_{2}}\ell_{2}(t)\big(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\bigr), (2.33) \displaystyle\|\partial_{t}v(t,\cdot)\|_{L^{2}}\lesssim t^{m-(m+1)(1+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}+\alpha_{2}}\big(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\bigr), \displaystyle\|\partial_{t}v(t,\cdot)\|_{\dot{H}^{\sigma-1}}\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}+\alpha_{2}}\ell_{2}(t)\big(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\bigr), where \ell_{2}(t) is defined by ( 2.26 ) and \alpha_{2}=\left\{\begin{aligned} &\big((m+1)n+\beta_{1}-1\big)(\tilde{q}-q),&\quad&\text{if}\ \ q<\tilde{q},\\ &\epsilon,&\quad&\text{if}\ \ q=\tilde{q}\end{aligned}\right. (2.34) with \epsilon>0 being an sufficiently small constant. Remark 2.6 . By swapping the positions of p and q in Theorem 2.3 , we can directly derive Theorem 2.4 . Therefore, we can also refer to Theorem 2.4 as being dual to Theorem 2.3 . Furthermore, it can be straightforward to provide examples to illustrate the reasonableness of the various conditions outlined in Theorem 2.4 , and we will not repeat them here. For (u_{0},u_{1},v_{0},v_{1})\in D^{\sigma}\times D^{\sigma} with 0<\sigma<1 , we also establish three global existence results. Theorem 2.5 . (Global existence for p>\tilde{p} and q>\tilde{q} (II) ). Let n\geq 1,\sigma\in(0,1) , m>-1 , \mu_{1},\mu_{2}>1,\nu_{1}^{2},\nu_{2}^{2}\geq 0 such that both \delta_{1} and \delta_{2}\geq(m+1)^{2}(n+2\sigma-1)^{2} . Let p,q>1 , and when n>2\sigma , we assume p,q\leq\frac{n}{n-2\sigma} . For p>\tilde{p} and q>\tilde{q} , there exists a small constant \varepsilon>0 such that for any initial data (u_{0},u_{1},v_{0},v_{1})\in D^{\sigma}\times D^{\sigma} satisfying the support condition ( 2.3 ), and if \|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\leq\varepsilon , then ( 1.1 ) admits a unique global solution (u,v)\in\Big(C\bigl([1,\infty);H^{\sigma}(\mathbb{R}^{n})\bigr)\Big)^{2} and ( u,v ) satisfies the support property ( 2.17 ). Furthermore, the solution ( u,v ) satisfies the decay estimates ( 2.18 ), ( 2.19 ), ( 2.22 ) and ( 2.23 ). Theorem 2.6 . (Global existence for p\leq\tilde{p} and q>\tilde{q} (II) ). Let n\geq 1,\sigma\in(0,1) , m>-1 , \mu_{1},\mu_{2}>1 , \nu_{1}^{2},\nu_{2}^{2}\geq 0 , m>-1 such that \delta_{1}\geq(m+1)^{2}(n+2\sigma-1)^{2} and \delta_{2}>(m+1)^{2}(n+2\sigma-1)^{2} . Let p,q>1 , and when n>2\sigma , we assume p,q\leq\frac{n}{n-2\sigma} . For p\leq\tilde{p} and q>\tilde{q} satisfying \frac{q+1}{pq-1}<\frac{(m+1)n+\beta_{2}-1}{2}, there exists a small constant \varepsilon>0 such that for any initial data (u_{0},u_{1},v_{0},v_{1})\in D^{\sigma}\times D^{\sigma} satisfying the support condition ( 2.3 ), and if \|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\leq\varepsilon , then ( 1.1 ) admits a unique global solution (u,v)\in\Big(C\bigl([1,\infty);H^{\sigma}(\mathbb{R}^{n})\bigr)\Big)^{2} and ( u,v ) satisfies the support property ( 2.17 ). Furthermore, the solution u satisfies the estimates ( 2.28 ), ( 2.29 ) and v satisfies the decay estimates ( 2.22 ), ( 2.23 ). Theorem 2.7 . (Global existence for p>\tilde{p} and q\leq\tilde{q} (II)). Let n\geq 1,\sigma\in(0,1) , m>-1 , \mu_{1},\mu_{2}>1 , \nu_{1}^{2},\nu_{2}^{2}\geq 0 , m>-1 such that \delta_{1}>(m+1)^{2}(n+2\sigma-1)^{2} and \delta_{2}\geq(m+1)^{2}(n+2\sigma-1)^{2} . Let p,q>1 , and when n>2\sigma , we assume p,q\leq\frac{n}{n-2\sigma} . For p>\tilde{p} and q\leq\tilde{q} satisfying \frac{p+1}{pq-1}<\frac{(m+1)n+\beta_{1}-1}{2}, there exists a small constant \varepsilon>0 such that for any initial data (u_{0},u_{1},v_{0},v_{1})\in D^{\sigma}\times D^{\sigma} satisfying the support condition ( 2.3 ), and if \|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\leq\varepsilon , then ( 1.1 ) admits a unique global solution (u,v)\in\Big(C\bigl([1,\infty);H^{\sigma}(\mathbb{R}^{n})\bigr)\Big)^{2} and ( u,v ) satisfies the support property ( 2.17 ). Furthermore, the solution u satisfies the decay estimates ( 2.18 ), ( 2.19 ) and v satisfies the estimates ( 2.32 ), ( 2.33 ). Remark 2.7 . Regarding the reasonableness of the conditions in Theorems 2.5 - 2.7 , some examples similar to Examples 2.3 - 2.4 can be provided; however, we will not elaborate on them here. Remark 2.8 . The regularity of initial data considered in Theorems 2.5 - 2.7 is lower than that in Theorems 2.2 - 2.4 . Inspired by [ 29 ] , we consider different function spaces for for the two cases of high and low regularity of the initial data, as shown in ( 4.12 ) and ( 5.1 ), respectively. This is precisely why the requirements for p,q in Theorems 2.5 - 2.7 differ from those in Theorems 2.2 - 2.4 . In terms of the proof process, the two cases are similar, therefore, we provide a relatively concise argument for the proofs of Theorems 2.5 - 2.7 in Section 5 . In the following remark, we emphasize that the results for the weakly coupled system ( 1.1 ) are not just a simple generalization of those for the corresponding single equation ( 1.14 ). Remark 2.9 . We summarize the results of Theorems 2.1 - 2.7 in the following relationship diagram \begin{aligned} &\bullet p\leq\tilde{p}\ \text{and}\ \ q\leq\tilde{q}\ \implies\ \ \Gamma_{m}(n,p,q,\beta_{1},\beta_{2})\geq 0\implies\text{Blow-up result},\\ &\bullet\Gamma_{m}(n,p,q,\beta_{1},\beta_{2})<0\implies\left.\begin{array}[]{l}p>\tilde{p}\ \text{and}\ q>\tilde{q},\ \text{or}\\ p\leq\tilde{p}\ \text{and}\ q>\tilde{q},\ \text{or}\\ p>\tilde{p}\ \text{and}\ q\leq\tilde{q}\end{array}\right\}\small\overset{\text{\text{large} $\delta_{1,2}$ }}{\implies}\text{Global existence.}\end{aligned} Therefore, we can conclude that when the damping terms dominate over the mass terms, the critical curve of ( 1.1 ) is described by the curve \Gamma_{m}(n,p,q,\beta_{1},\beta_{2})=0 in the p-q plane. Moreover, from this relationship diagram we see \Gamma_{m}(n,p,q,\beta_{1},\beta_{2})\geq 0 can be equivalently expressed as p\leq\tilde{p},q\leq\tilde{q} , while \Gamma_{m}(n,p,q,\beta_{1},\beta_{2})<0 can be alternatively expressed as: p>\tilde{p},q>\tilde{q}, or p\leq\tilde{p},q>\tilde{q}, or p>\tilde{p},q\leq\tilde{q} . Next, we focus on analysing the blow-up result, i.e., Theorem 2.1 , to illustrate that the main results (Theorems 2.1 - 2.7 ) are not merely a straightforward extension of those for the single equation ( 1.14 ). For convenience, denote p_{F}^{1}:=p_{F}\big((m+1)n+\frac{\mu_{1}-1-\sqrt{\delta_{1}}}{2}\big),\ \ p_{F}^{2}:=p_{F}\big((m+1)n+\frac{\mu_{2}-1-\sqrt{\delta_{2}}}{2}\big). From ( 2.12 ) it is clear that \tilde{p},\tilde{q} can be regarded as "perturbations" of the Fujita index. Indeed, we see \tilde{p}=p_{F}^{2},\ \tilde{q}=p_{F}^{1} hold when \beta_{1}=\beta_{2} . By the analysis in Section 1 , we know when p\leq p_{F}((m+1)n+\frac{\mu-1-\sqrt{\delta}}{2}) , the solution to ( 1.14 ) will blow up. Consequently, if the result was directly extended, the blow-up region for ( 1.1 ) would be described by p\leq p_{F}^{1},q\leq p_{F}^{2} (or p\leq p_{F}^{2},q\leq p_{F}^{1} ), however, the blow-up condition \Gamma_{m}(n,p,q,\beta_{1},\beta_{2})\geq 0 does not coincide with this, except in the special case where \beta_{1}=\beta_{2} . We present the blow-up region of ( 1.1 ) in the form of a diagram (Diagram 1). 0 \max\{1,\frac{1+\beta_{1}}{2m+1+\beta_{2}}\} \tilde{p}=p_{F}^{2} \max\{1,\frac{1+\beta_{2}}{2m+1+\beta_{1}}\} \tilde{q}=p_{F}^{1} Case1: \beta_{1}=\beta_{2} \tilde{p} \tilde{q} Case2: \beta_{1}>\beta_{2} \tilde{p} \tilde{q} Case3: \beta_{1}<\beta_{2} p q Diagram 1 Diagram 1 clearly shows that the blow-up result for ( 1.1 ) can be regarded as a direct extension of its single equation ( 1.14 ) only when \beta_{1}=\beta_{2} . In contrast, if \beta_{1}\neq\beta_{2} , the blow-up region of ( 1.1 ) undergoes a "shift," which more prominently reflects the complex interplay of mutual constraints and balance between the two damping terms and the two mass terms. 3. Blow-up result The main tool for proving Theorem 2.1 is the test function. Referring to the methods in [ 2 , 22 ] , we find that the presence of m in Gellerstedt operator \partial_{t}^{2}-t^{2m}\Delta compels us to seek new test functions. As will be seen in the following proof, the test functions we use are closely related to m , highlighting the differences between the Gellerstedt operator and the standard wave operator \partial_{t}^{2}-\Delta . Let us assume by contradiction that (u,v) is a global energy solution to ( 1.1 ), that is T=\infty in ( 2.4 ). Note that h_{i}(t)=t^{\beta_{i}},\beta_{i}=\frac{\mu_{i}+1-\sqrt{\delta_{i}}}{2},i=1,2 satisfy h_{i}^{\prime\prime}(t)-\frac{\mu_{i}}{t}h_{i}^{\prime}(t)+\frac{\mu_{i}+\nu_{i}^{2}}{t^{2}}h_{i}(t)=0,\ i=1,2. (3.1) Multiplying both sides of the first equation in ( 1.1 ) by h_{1}(t) and the second by h_{2}(t) , we have the following equations in divergence form \left\{\begin{aligned} &\partial_{t}^{2}(h_{1}u)-\Delta(t^{2m}h_{1}u)+\partial_{t}(\frac{\mu_{1}}{t}h_{1}u-2h_{1}^{\prime}u)=h_{1}|v|^{p},\\ &\partial_{t}^{2}(h_{2}v)-\Delta(t^{2m}h_{2}v)+\partial_{t}(\frac{\mu_{2}}{t}h_{2}v-2h_{2}^{\prime}v)=h_{2}|u|^{q}.\end{aligned}\right. (3.2) Choose a non-increasing function \lambda(t)\in C_{0}^{\infty}\big([0,\infty)\big) , a radially symmetric function \phi(x)\in C_{0}^{\infty}(\mathbb{R}^{n}) such that \phi is non-increasing with respect to |x| , and satisfy \displaystyle\lambda(t)=1\ \text{on}\ [0,\frac{1}{2}],\ \text{supp}\lambda(t)\subset[0,1]; (3.3) \displaystyle\phi(x)=1\ \text{on}\ B_{\frac{1}{2}},\ \text{supp}\phi(x)\subset B_{1}; \displaystyle|\lambda^{\prime}(t)|,\ |\lambda^{\prime\prime}(t)|\lesssim\lambda(t)^{\frac{1}{r}},\ \ \ |\Delta\phi|\lesssim\phi^{\frac{1}{r}}\ \ \text{for any}\ r>1. Regarding the existences of \lambda and \phi , one can refer to [ 22 ] . For d,R\geq 1, let \psi_{d,R}(t,x)=\lambda\bigl(\frac{t-1}{d}\bigr)\phi\bigl(\frac{x}{R^{m+1}}\bigr) . Substituting \varPhi_{1}(t,x)=h_{1}(t)\psi_{d,R}(t,x) into ( 2.6 ) and using ( 3.1 ), we have \displaystyle I_{d,R}: \displaystyle=\int_{1}^{d+1}\int_{B_{R^{m+1}}}h_{1}(t)\psi_{d,R}(t,x)|v(t,x)|^{p}dxdt \displaystyle=-\underbrace{\int_{B_{R^{m+1}}}\bigl(\frac{\mu_{1}-1+\sqrt{\delta_{1}}}{2}u_{0}(x)+u_{1}(x)\bigr)\phi(\frac{x}{R^{m+1}})dx}_{K_{0}} \displaystyle+\underbrace{\int_{\frac{d}{2}+1}^{d+1}\int_{B_{R^{m+1}}}u(t,x)h_{1}(t)\partial_{t}^{2}\psi_{d,R}(t,x)dxdt}_{K_{1}} \displaystyle+\underbrace{\int_{\frac{d}{2}+1}^{d+1}\int_{B_{R^{m+1}}}u(t,x)\bigl(2h_{1}^{\prime}(t)-\frac{\mu_{1}}{t}h_{1}(t)\bigr)\partial_{t}\psi_{d,R}(t,x)dxdt}_{K_{2}} \displaystyle-\underbrace{\int_{1}^{d+1}\int_{B_{R^{m+1}}\textbackslash B_{\frac{R^{m+1}}{2}}}u(t,x)t^{2m}h_{1}(t)\Delta\psi_{d,R}(t,x)dxdt}_{K_{3}}. In view of ( 2.8 ), there exists a R_{1}>0 such that for any R\geq R_{1}, we have K_{0}>0 , so I_{d,R}-1 , i.e., \int_{1}^{d+1}t^{2mq^{\prime}+\beta_{1}q^{\prime}+\beta_{2}(1-q^{\prime})}dt\lesssim(d+1)^{2mq^{\prime}+\beta_{1}q^{\prime}+\beta_{2}(1-q^{\prime})+1}. Then \displaystyle|K_{3}| \displaystyle\lesssim R^{-2(m+1)}\Big(\int_{1}^{d+1}\int_{B_{R^{m+1}}\textbackslash B_{\frac{R^{m+1}}{2}}}t^{2mq^{\prime}}h_{1}^{q^{\prime}}(t)h_{2}^{1-q^{\prime}}(t)dxdt\Big)^{\frac{1}{q^{\prime}}} \displaystyle\ \ \times\Big(\int_{1}^{d+1}\int_{B_{R^{m+1}}\textbackslash B_{\frac{R^{m+1}}{2}}}h_{2}(t)|u(t,x)|^{q}\Big|\lambda\bigl(\frac{t-1}{d}\bigr)\Delta\phi\bigl(\frac{x}{R^{m+1}}\bigr)\Bigr|^{q}dxdt\Bigr)^{\frac{1}{q}} \displaystyle\lesssim(d+1)^{2m+\beta_{1}+\beta_{2}(\frac{1}{q^{\prime}}-1)+\frac{1}{q^{\prime}}}R^{-2(m+1)+\frac{(m+1)n}{q^{\prime}}}\widehat{J}^{\frac{1}{q}}_{d,R}, (3.7) where \widehat{J}_{d,R}=\int_{1}^{d+1}\int_{B_{R^{m+1}}\textbackslash B_{\frac{R^{m+1}}{2}}}h_{2}(t)|u(t,x)|^{q}\psi_{d,R}(t,x)dxdt. By ( 3.4 )-( 3.7 ), we get the estimate of I_{d,R} as \displaystyle I_{d,R}\lesssim \displaystyle(d+1)^{-2+\beta_{1}+\beta_{2}(\frac{1}{q^{\prime}}-1)+\frac{1}{q^{\prime}}}R^{\frac{(m+1)n}{q^{\prime}}}\widetilde{J}_{d,R}^{\frac{1}{q}} \displaystyle+(d+1)^{2m+\beta_{1}+\beta_{2}(\frac{1}{q^{\prime}}-1)+\frac{1}{q^{\prime}}}R^{-2(m+1)+\frac{(m+1)n}{q^{\prime}}}\widehat{J}^{\frac{1}{q}}_{d,R},\ \ \text{for any}\ R\geq R_{1}. (3.8) Substituting \varPhi_{2}(t,x)=h_{2}(t)\psi_{d,R}(t,x) into ( 2.7 ) and using ( 3.1 ) yields \displaystyle J_{d,R}: \displaystyle=\int_{1}^{d+1}\int_{B_{R^{m+1}}}h_{2}(t)\psi_{d,R}(t,x)|u(t,x)|^{q}dxdt \displaystyle=-\underbrace{\int_{B_{R^{m+1}}}\bigl(\frac{\mu_{2}-1+\sqrt{\delta_{2}}}{2}v_{0}(x)+v_{1}(x)\bigr)\phi(\frac{x}{R^{m+1}})dx}_{L_{0}} \displaystyle+\underbrace{\int_{\frac{d}{2}+1}^{d+1}\int_{B_{R^{m+1}}}v(t,x)h_{2}(t)\partial_{t}^{2}\psi_{d,R}(t,x)dxdt}_{L_{1}} \displaystyle+\underbrace{\int_{\frac{d}{2}+1}^{d+1}\int_{B_{R^{m+1}}}v(t,x)\bigl(2h_{2}^{\prime}(t)-\frac{\mu_{2}}{t}h_{2}(t)\bigr)\partial_{t}\psi_{d,R}(t,x)dxdt}_{L_{2}} \displaystyle-\underbrace{\int_{1}^{d+1}\int_{B_{R^{m+1}}\textbackslash B_{\frac{R^{m+1}}{2}}}v(t,x)t^{2m}h_{2}(t)\Delta\psi_{d,R}(t,x)dxdt}_{L_{3}}. By ( 2.8 ), there exists a R_{2}>0 such that L_{0}>0 holds for any R\geq R_{2}, so J_{d,R}-1 , which guarantees that \int_{1}^{d+1}t^{2mp^{\prime}+\beta_{2}p^{\prime}+\beta_{1}(1-p^{\prime})}dt\lesssim(d+1)^{2mp^{\prime}+\beta_{2}p^{\prime}+\beta_{1}(1-p^{\prime})+1}, then \displaystyle|L_{3}|\lesssim R^{-2(m+1)}\Big(\int_{1}^{d+1}\int_{B_{R^{m+1}}\textbackslash B_{\frac{R^{m+1}}{2}}}t^{2mp^{\prime}}h_{2}^{p^{\prime}}(t)h_{1}^{1-p^{\prime}}(t)dxdt\Big)^{\frac{1}{p^{\prime}}} \displaystyle\times\Big(\int_{1}^{d+1}\int_{B_{R^{m+1}}\textbackslash B_{\frac{R^{m+1}}{2}}}h_{1}(t)|v(t,x)|^{p}\Big|\lambda\bigl(\frac{t-1}{d}\bigr)\Delta\phi\bigl(\frac{x}{R^{m+1}}\bigr)\Bigr|^{p}dxdt\Bigr)^{\frac{1}{p}} \displaystyle\lesssim(d+1)^{2m+\beta_{1}(\frac{1}{p^{\prime}}-1)+\beta_{2}+\frac{1}{p^{\prime}}}R^{-2(m+1)+\frac{(m+1)n}{p^{\prime}}}\widehat{I}^{\frac{1}{p}}_{d,R}, (3.12) where \widehat{I}_{d,R}=\int_{1}^{d+1}\int_{B_{R^{m+1}}\textbackslash B_{\frac{R^{m+1}}{2}}}h_{1}(t)|v(t,x)|^{p}\psi_{d,R}(t,x)dxdt. Substituting the estimates ( 3.10 ), ( 3.11 ), ( 3.12 ) of L_{1} , L_{2} and L_{3} into ( 3.9 ) gives that \displaystyle J_{d,R}\lesssim \displaystyle(d+1)^{-2+\beta_{1}(\frac{1}{p^{\prime}}-1)+\beta_{2}+\frac{1}{p^{\prime}}}R^{\frac{(m+1)n}{p^{\prime}}}\widetilde{I}_{d,R}^{\frac{1}{p}} (3.13) \displaystyle+(d+1)^{2m+\beta_{1}(\frac{1}{p^{\prime}}-1)+\beta_{2}+\frac{1}{p^{\prime}}}R^{-2(m+1)+\frac{(m+1)n}{p^{\prime}}}\widehat{I}^{\frac{1}{p}}_{d,R} holds for any R\geq R_{2} . Let d=R\geq\max\{1,R_{1},R_{2}\} , from ( 3.8 ) and ( 3.13 ), we get \displaystyle I_{R,R}\lesssim(R+1)^{-2+\beta_{1}+\beta_{2}(\frac{1}{q^{\prime}}-1)+\frac{1}{q^{\prime}}+\frac{(m+1)n}{q^{\prime}}}\big(\widetilde{J}_{R,R}^{\frac{1}{q}}+\widehat{J}^{\frac{1}{q}}_{R,R}\big), (3.14) \displaystyle J_{R,R}\lesssim(R+1)^{-2+\beta_{2}+\beta_{1}(\frac{1}{p^{\prime}}-1)+\frac{1}{p^{\prime}}+\frac{(m+1)n}{p^{\prime}}}\big(\widetilde{I}_{R,R}^{\frac{1}{p}}+\widehat{I}^{\frac{1}{p}}_{R,R}\big). (3.15) Note that \widetilde{I}_{R,R},\widehat{I}_{R,R}\leq I_{R,R} and \widetilde{J}_{R,R},\widehat{J}_{R,R}\leq J_{R,R} . Using ( 3.14 ) and ( 3.15 ), we obtain \displaystyle I_{R,R}\lesssim(R+1)^{-2+\beta_{1}+\beta_{2}(\frac{1}{q^{\prime}}-1)+\frac{1}{q^{\prime}}+\frac{(m+1)n}{q^{\prime}}}J_{R,R}^{\frac{1}{q}} (3.16) \displaystyle\lesssim(R+1)^{-2+\beta_{1}+\beta_{2}(\frac{1}{q^{\prime}}-1)+\frac{1}{q^{\prime}}+\frac{(m+1)n}{q^{\prime}}+\frac{1}{q}\big(-2+\beta_{1}(\frac{1}{p^{\prime}}-1)+\beta_{2}+\frac{1}{p^{\prime}}+\frac{(m+1)n}{p^{\prime}}\big)}I_{R,R}^{\frac{1}{pq}} \displaystyle=(R+1)^{-2-\frac{2}{q}+(1-\frac{1}{pq})\big((m+1)n+\beta_{1}+1\big)}I_{R,R}^{\frac{1}{pq}} and \displaystyle J_{R,R}\lesssim(R+1)^{-2+\beta_{2}+\beta_{1}(\frac{1}{p^{\prime}}-1)+\frac{1}{p^{\prime}}+\frac{(m+1)n}{p^{\prime}}}I_{R,R}^{\frac{1}{p}} (3.17) \displaystyle\lesssim(R+1)^{-2+\beta_{2}+\beta_{1}(\frac{1}{p^{\prime}}-1)+\frac{1}{p^{\prime}}+\frac{(m+1)n}{p^{\prime}}+\frac{1}{p}\big(-2+\beta_{1}+\beta_{2}(\frac{1}{q^{\prime}}-1)+\frac{1}{q^{\prime}}+\frac{(m+1)n}{q^{\prime}}\big)}J_{R,R}^{\frac{1}{pq}} \displaystyle=(R+1)^{-2-\frac{2}{p}+(1-\frac{1}{pq})\big((m+1)n+\beta_{2}+1\big)}J_{R,R}^{\frac{1}{pq}}. In other words, \displaystyle I_{R,R}\lesssim(R+1)^{-2p\frac{q+1}{pq-1}+(m+1)n+\beta_{1}+1}, (3.18) \displaystyle J_{R,R}\lesssim(R+1)^{-2q\frac{p+1}{pq-1}+(m+1)n+\beta_{2}+1}. (3.19) If -2p\frac{q+1}{pq-1}+(m+1)n+\beta_{1}+1<0 , i.e., \frac{p+1}{pq-1}-\frac{\beta_{1}-1}{2}>\frac{(m+1)n}{2} , by ( 3.18 ), \lim\limits_{R\rightarrow+\infty}I_{R,R}=0 holds. Then \begin{aligned} \lim\limits_{R\rightarrow+\infty}\int_{1}^{R+1}\int_{B_{R^{m+1}}}h_{1}(t)|v(t,x)&|^{p}\psi_{R,R}(t,x)dxdt=\int_{1}^{\infty}\int_{R^{n}}h_{1}(t)|v(t,x)|^{p}dxdt=0,\end{aligned} (3.20) which yields v=0 almost everywhere. This contradicts the assumption ( 2.8 ) on ( v_{0},v_{1} ). If -2p\frac{q+1}{pq-1}+(m+1)n+\beta_{1}+1=0 , i.e., \frac{p+1}{pq-1}-\frac{\beta_{1}-1}{2}=\frac{(m+1)n}{2} , by ( 3.18 ), \lim\limits_{R\rightarrow+\infty}I_{R,R}\leq C holds. Then \displaystyle\lim\limits_{R\rightarrow+\infty}\int_{1}^{R+1}\int_{B_{R^{m+1}}}h_{1}(t)|v(t,x)|^{p}\psi_{R,R}(t,x)dxdt=\int_{1}^{\infty}\int_{R^{n}}h_{1}(t)|v(t,x)|^{p}dxdt\leq C, which means h_{1}(t)|v(t,x)|^{p}\in L^{1}([1,\infty)\times\mathbb{R}^{n}) . Consequently, employing the dominated convergence theorem and the definitions of \widetilde{I}_{R,R} , \widehat{I}_{R,R} gives that \lim\limits_{R\rightarrow+\infty}\widetilde{I}_{R,R}=\lim\limits_{R\rightarrow+\infty}\widehat{I}_{R,R}=0. From ( 3.14 ), we have \lim\limits_{R\rightarrow+\infty}I_{R,R}=0 , so ( 3.20 ) holds in this case and we derive the same contradiction. By applying ( 3.19 ) and the same argument above, we can get u=0 almost everywhere under the condition \frac{q+1}{pq-1}-\frac{\beta_{2}-1}{2}\geq\frac{(m+1)n}{2} , which contradicts the assumption ( 2.8 ) on ( u_{0},u_{1} ). In summary, assuming T=\infty leads to the contradiction under the condition \frac{p+1}{pq-1}-\frac{\beta_{1}-1}{2}\geq\frac{(m+1)n}{2}\ \ \ \text{or}\ \ \ \frac{q+1}{pq-1}-\frac{\beta_{2}-1}{2}\geq\frac{(m+1)n}{2}, (3.21) provided p,q fulfill ( 2.9 ), which means the energy solution (u,v) cannot exist globally. Since the relations ( 3.21 ) on p,q are equivalent \Gamma_{m}(n,p,q,\beta_{1},\beta_{2})\geq 0, the proof is completed. 4. Global existence of solutions with high regularity of initial data In this section, we aim to establish the global existence of solutions with high regularity of initial data, i.e., Theorems 2.2 - 2.4 . The key tool is the estimates of the solution and its derivatives for the corresponding linear equation of ( 1.1 ), which were precisely established in our recent study [ 17 ] . These results will be outlined in Section 4.1 without repeating the proofs. Based on these estimates, we can construct the appropriate space and the associated solution operator, then apply Duhamel’s principle along with the contraction mapping principle to derive the global existence. 4.1. The estimates for the corresponding linear equation For the corresponding linear equation of ( 1.1 ) \left\{\begin{aligned} &\partial_{t}^{2}u-t^{2m}\Delta u+\frac{\mu}{t}\partial_{t}u+\frac{\nu^{2}}{t^{2}}u=0,\ &\quad&t>\tau\geq 1,\\ &u(\tau,x)=f(x),\ \partial_{t}u(\tau,x)=g(x),&\quad&x\in\mathbb{R}^{n},\end{aligned}\right. (4.1) we established the (L^{1}\cap L^{2})-L^{2} estimates in [ 17 ] and we list the results. Proposition 4.1 . Let \sigma>0 , \delta=(\mu-1)^{2}-4\nu^{2}>0 and (f,g)\in D^{\sigma} , then for any \kappa\in[0,\sigma] , the solution u to ( 4.1 ) with \tau=1 satisfies \displaystyle\| \displaystyle u(t,\cdot)\|_{\dot{H}^{\kappa}}\lesssim\|(f,g)\|_{D^{\kappa}}\left\{\begin{aligned} &t^{-\frac{\mu+m}{2}},\ &\text{if}\ \kappa>\frac{\sqrt{\delta}}{2(m+1)}+\frac{1}{2}-\frac{n}{2},\\ &t^{-\frac{\mu+m}{2}}\bigl(1+\log{t}\bigr)^{\frac{1}{2}},\ &\text{if}\ \kappa=\frac{\sqrt{\delta}}{2(m+1)}+\frac{1}{2}-\frac{n}{2},\\ &t^{-(m+1)(\kappa+\frac{n}{2})+\frac{\sqrt{\delta}-\mu+1}{2}},\ &\text{if}\ \kappa<\frac{\sqrt{\delta}}{2(m+1)}+\frac{1}{2}-\frac{n}{2}.\end{aligned}\right. (4.2) Moreover, for any \kappa\in[1,\sigma], \begin{aligned} \|&\partial_{t}u(t,\cdot)\|_{\dot{H}^{\kappa-1}}\lesssim\|(f,g)\|_{D^{\kappa}}\left\{\begin{aligned} &t^{m-\frac{\mu+m}{2}},\ &\text{if}\ \kappa>\frac{\sqrt{\delta}}{2(m+1)}+\frac{1}{2}-\frac{n}{2},\\ &t^{m-\frac{\mu+m}{2}}\bigl(1+\log{t}\bigr)^{\frac{1}{2}},\ &\text{if}\ \kappa=\frac{\sqrt{\delta}}{2(m+1)}+\frac{1}{2}-\frac{n}{2},\\ &t^{m-(m+1)(\kappa+\frac{n}{2})+\frac{\sqrt{\delta}-\mu+1}{2}},\ &\text{if}\ \kappa<\frac{\sqrt{\delta}}{2(m+1)}+\frac{1}{2}-\frac{n}{2}.\end{aligned}\right.\end{aligned} (4.3) Proposition 4.2 . Let \delta=(\mu-1)^{2}-4\nu^{2}>0 , f=0 , g\in H^{[\sigma-1]_{+}}\cap L^{1} with \sigma>0 , then for any \kappa\in[0,\sigma] , the solution u to ( 4.1 ) satisfies \displaystyle\|u(t,\cdot) \displaystyle\|_{\dot{H}^{\kappa}}\lesssim\bigl(\|g\|_{L^{1}}+\tau^{(m+1)(\frac{n}{2}+[\kappa-1]_{+})}\|g\|_{\dot{H}^{[\kappa-1]_{+}}}\bigr) (4.4) \displaystyle\times\left\{\begin{aligned} &t^{-\frac{\mu+m}{2}}\tau^{-(m+1)(\kappa+\frac{n}{2})+\frac{\mu+m}{2}+1},\ &\text{if}\ \kappa>\frac{\sqrt{\delta}}{2(m+1)}+\frac{1}{2}-\frac{n}{2},\\ &t^{-\frac{\mu+m}{2}}\tau^{-\frac{\sqrt{\delta}-\mu-1}{2}}\bigl(1+\log\frac{t}{\tau}\bigr)^{\frac{1}{2}},\ &\text{if}\ \kappa=\frac{\sqrt{\delta}}{2(m+1)}+\frac{1}{2}-\frac{n}{2},\\ &t^{-(m+1)(\kappa+\frac{n}{2})+\frac{\sqrt{\delta}-\mu+1}{2}}\tau^{-\frac{\sqrt{\delta}-\mu-1}{2}},\ &\text{if}\ \kappa<\frac{\sqrt{\delta}}{2(m+1)}+\frac{1}{2}-\frac{n}{2}.\\ \end{aligned}\right. Moreover, for any \kappa\in[1,\sigma] , \displaystyle\|\partial_{t}u(t,\cdot)\|_{\dot{H}^{\kappa-1}}\lesssim\bigl(\|g\|_{L^{1}}+\tau^{(m+1)(\frac{n}{2}+\kappa-1)}\|g\|_{\dot{H}^{\kappa-1}}+\tau^{(m+1)(\frac{n}{2}+[\kappa-2]_{+})}\|g\|_{\dot{H}^{[\kappa-2]_{+}}}\bigr) \displaystyle\quad\quad\times\left\{\begin{aligned} &t^{m-\frac{\mu+m}{2}}\tau^{-(m+1)(\kappa+\frac{n}{2})+\frac{\mu+m}{2}+1},\ &\text{if}\ \kappa>\frac{\sqrt{\delta}}{2(m+1)}+\frac{1}{2}-\frac{n}{2},\\ &t^{m-\frac{\mu+m}{2}}\tau^{-\frac{\sqrt{\delta}-\mu-1}{2}}\bigl(1+\log\frac{t}{\tau}\bigr)^{\frac{1}{2}},\ &\text{if}\ \kappa=\frac{\sqrt{\delta}}{2(m+1)}+\frac{1}{2}-\frac{n}{2},\\ &t^{m-(m+1)(\kappa+\frac{n}{2})+\frac{\sqrt{\delta}-\mu+1}{2}}\tau^{-\frac{\sqrt{\delta}-\mu-1}{2}},\ &\text{if}\ \kappa<\frac{\sqrt{\delta}}{2(m+1)}+\frac{1}{2}-\frac{n}{2}.\\ \end{aligned}\right. (4.5) 4.2. The proof of Theorem 2.2 Denote by E_{0}^{\ \mu,\nu}(t,\tau,x) and E_{1}^{\ \mu,\nu}(t,\tau,x) the fundamental solutions to ( 4.1 ) with the initial data (f,g)=(\delta_{0},0) and (0,\delta_{0}) , respectively, where \delta_{0} is the Dirac function. By Duhamel’s principle, the solution to ( 1.1 ) can be expressed as \begin{aligned} &u(t,x)=E_{0}^{\ \mu_{1},\nu_{1}}(t,1,x)\ast u_{0}(x)+E_{1}^{\ \mu_{1},\nu_{1}}(t,1,x)\ast u_{1}(x)+\int_{1}^{t}E_{1}^{\ \mu_{1},\nu_{1}}(t,\tau,x)\ast|v(\tau,x)|^{p}d\tau,\\ &v(t,x)=E_{0}^{\ \mu_{2},\nu_{2}}(t,1,x)\ast v_{0}(x)+E_{1}^{\ \mu_{2},\nu_{2}}(t,1,x)\ast v_{1}(x)+\int_{1}^{t}E_{1}^{\ \mu_{2},\nu_{2}}(t,\tau,x)\ast|u(\tau,x)|^{q}d\tau,\end{aligned} (4.6) where \ast denotes the convolution with respect to x . For convenience, let \displaystyle u^{l}(t,x)=E_{0}^{\ \mu_{1},\nu_{1}}(t,1,x)\ast u_{0}(x)+E_{1}^{\ \mu_{1},\nu_{1}}(t,1,x)\ast u_{1}(x), (4.7) \displaystyle v^{l}(t,x)=E_{0}^{\ \mu_{2},\nu_{2}}(t,1,x)\ast v_{0}(x)+E_{1}^{\ \mu_{2},\nu_{2}}(t,1,x)\ast v_{1}(x), \displaystyle H_{1}(v)(t,x)=\int_{1}^{t}E_{1}^{\ \mu_{1},\nu_{1}}(t,\tau,x)\ast|v(\tau,x)|^{p}d\tau, \displaystyle H_{2}(u)(t,x)=\int_{1}^{t}E_{1}^{\ \mu_{2},\nu_{2}}(t,\tau,x)\ast|u(\tau,x)|^{q}d\tau. For the sake of clarity, let us introduce some notations and expressions. \ell_{i}(t)\ (i=1,2) are shown as ( 2.26 ), if \sigma>1 , let \displaystyle M_{1}(t,u)= \displaystyle t^{-\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}+(m+1)\frac{n}{2}}\|u(t,\cdot)\|_{L^{2}} \displaystyle+t^{-\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}+(m+1)(\sigma+\frac{n}{2})}\ell_{1}^{-1}(t)\|u(t,\cdot)\|_{\dot{H}^{\sigma}} \displaystyle+t^{-m-\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}+(m+1)(1+\frac{n}{2})}\|\partial_{t}u(t,\cdot)\|_{L^{2}} (4.8) \displaystyle+t^{-m-\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}+(m+1)(\sigma+\frac{n}{2})}\ell_{1}^{-1}(t)\|\partial_{t}u(t,\cdot)\|_{\dot{H}^{\sigma-1}}, \displaystyle M_{2}(t,v)= \displaystyle t^{-\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}+(m+1)\frac{n}{2}}\|v(t,\cdot)\|_{L^{2}} \displaystyle+t^{-\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}+(m+1)(\sigma+\frac{n}{2})}\ell_{2}^{-1}(t)\|v(t,\cdot)\|_{\dot{H}^{\sigma}} \displaystyle+t^{-m-\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}+(m+1)(1+\frac{n}{2})}\|\partial_{t}v(t,\cdot)\|_{L^{2}} (4.9) \displaystyle+t^{-m-\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}+(m+1)(\sigma+\frac{n}{2})}\ell_{2}^{-1}(t)\|\partial_{t}v(t,\cdot)\|_{\dot{H}^{\sigma-1}}; while if \sigma=1 , \displaystyle M_{1}(t,u)= \displaystyle t^{-\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}+(m+1)\frac{n}{2}}\|u(t,\cdot)\|_{L^{2}} \displaystyle+t^{-\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}+(m+1)(\sigma+\frac{n}{2})}\ell_{1}^{-1}(t)\|u(t,\cdot)\|_{\dot{H}^{\sigma}} \displaystyle+t^{-m-\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}+(m+1)(\sigma+\frac{n}{2})}\ell_{1}^{-1}(t)\|\partial_{t}u(t,\cdot)\|_{\dot{H}^{\sigma-1}}, (4.10) \displaystyle M_{2}(t,v)= \displaystyle t^{-\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}+(m+1)\frac{n}{2}}\|v(t,\cdot)\|_{L^{2}} \displaystyle+t^{-\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}+(m+1)(\sigma+\frac{n}{2})}\ell_{2}^{-1}(t)\|v(t,\cdot)\|_{\dot{H}^{\sigma}} (4.11) \displaystyle+t^{-m-\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}+(m+1)(\sigma+\frac{n}{2})}\ell_{2}^{-1}(t)\|\partial_{t}v(t,\cdot)\|_{\dot{H}^{\sigma-1}}. For \sigma\geq 1 and T>1 , define the function space \displaystyle\mathcal{X}(T):=\Big\{(u,v)\in(C\bigl([1,T];H^{\sigma}\bigr)\cap C^{1}\bigl([1,T];H^{\sigma-1}\bigr)\Big)^{2} \displaystyle\text{such that}\ \ \text{supp}\big(u(t,\cdot),v(t,\cdot)\big)\subset B_{\phi_{m}(t)-\phi_{m}(1)+M}\Big\} equipped with the norm \|(u,v)\|_{\mathcal{X}(T)}=\sup\limits_{t\in[1,T]}\big(t^{-\alpha_{1}}M_{1}(t,u)+t^{-\alpha_{2}}M_{2}(t,v)\big), (4.12) where \phi_{m}(t) is defined by ( 2.5 ) and \alpha_{1},\alpha_{2} are given by \alpha_{1}=\left\{\begin{aligned} &\big((m+1)n+\beta_{2}-1\big)(\tilde{p}-p),&\quad&\text{if}\ \ p<\tilde{p},\\ &\epsilon,&\quad&\text{if}\ \ p=\tilde{p},\\ &0,&\quad&\text{if}\ \ p>\tilde{p},\end{aligned}\right. (4.13) \alpha_{2}=\left\{\begin{aligned} &\big((m+1)n+\beta_{1}-1\big)(\tilde{q}-q),&\quad&\text{if}\ \ q<\tilde{q},\\ &\epsilon,&\quad&\text{if}\ \ q=\tilde{q},\\ &0,&\quad&\text{if}\ \ q>\tilde{q},\end{aligned}\right. (4.14) where \epsilon>0 is sufficiently small. Specially, \alpha_{1}=\alpha_{2}=0 under the assumptions of Theorem 2.2 . Based on the representations ( 4.6 ) of the solution to ( 1.1 ), we define the operator \mathcal{N} by \mathcal{N}(u,v):=\big(u^{l}+H_{1}(v),v^{l}+H_{2}(u)\big) (4.15) and introduce the subset of \mathcal{X}(T) X(T,K)=\bigl\{(u,v)\in\mathcal{X}(T):\|(u,v)\|_{\mathcal{X}(T)}\leq K\bigr\}, where u^{l},v^{l},H_{1}(v),H_{2}(u) are defined in ( 4.7 ) and K is a positive constant to be determined. Then we have Proposition 4.3 . Under the conditions of Theorem 2.2 , there exists a constant C>0 such that for any T>1 and any (u,v),(\tilde{u},\tilde{v})\in\mathcal{X}(T) , \displaystyle\|\mathcal{N}(u,v)\|_{\mathcal{X}(T)}\leq C\bigl(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\big)+C\big(\|(u,v)\|_{\mathcal{X}{(T)}}^{p}+\|(u,v)\|_{\mathcal{X}{(T)}}^{q}\bigr), (4.16) \displaystyle\|\mathcal{N}(u,v)-\mathcal{N}(\tilde{u},\tilde{v})\|_{\mathcal{X}(T)}\leq C\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(T)} \displaystyle\quad\quad\quad\quad\quad\quad\quad\quad\times\bigl(\|(u,v)\|_{\mathcal{X}(T)}^{p-1}+\|(u,v)\|_{\mathcal{X}(T)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(T)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(T)}^{q-1}\bigr). (4.17) The proofs of ( 4.16 ) and ( 4.17 ) are rather lengthy, so we present them in detail in the following two sections. To conclude this section, we use Proposition 4.3 to establish the global existence of solutions for \sigma\geq 1 , thereby completing the proof of Theorem 2.2 . Proof of Theorem 2.2 . Choose K=3C\|(u_{0},u_{1})\|_{D^{\sigma}}+3C\|(v_{0},v_{1})\|_{D^{\sigma}} in X(T,K) , where C is the constant mentioned in Proposition 4.3 . Then for any (u,v),(\tilde{u},\tilde{v})\in\mathcal{X}(T,K) , it follows from ( 4.16 ) that \displaystyle\|\mathcal{N}(u,v)\|_{\mathcal{X}(T)}\leq C(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}) \displaystyle\times\Big(1+(3C)^{p}\big(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\big)^{p-1}+(3C)^{q}\big(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\big)^{q-1}\Big) \displaystyle\leq 3C(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}})=K (4.18) provided that \|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\leq\min\big\{\bigl(\frac{1}{3^{p}C^{p}}\bigr)^{\frac{1}{p-1}},\bigl(\frac{1}{3^{q}C^{q}}\bigr)^{\frac{1}{q-1}}\big\} . Moreover, by ( 4.17 ), we have \displaystyle\|\mathcal{N}(u,v)-\mathcal{N}(\tilde{u},\tilde{v})\|_{\mathcal{X}(T)}\leq\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(T)} \displaystyle\ \times 2C\Big(\big(3C(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}})\big)^{p-1}+\big(3C(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}})\big)^{q-1}\Big) \displaystyle\leq\frac{1}{2}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(T)} (4.19) as long as \|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\leq\min\big\{\frac{1}{8C^{p}3^{p-1}},\frac{1}{8C^{q}3^{q-1}}\big\} . Therefore, if \resizebox{24800161}{}{$\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\leq\varepsilon:=\min\big\{\bigl(\frac{1}{3^{p}C^{p}}\bigr)^{\frac{1}{p-1}},\bigl(\frac{1}{3^{q}C^{q}}\bigr)^{\frac{1}{q-1}},\frac{1}{8C^{p}3^{p-1}},\frac{1}{8C^{q}3^{q-1}}\big\}$}, from ( 4.18 ) and ( 4.19 ), we see that \mathcal{N} is a contraction mapping from X(T,K) into itself. According to the contraction mapping principle, there exists a unique (u,v)\in X(T,K) such that \mathcal{N}(u,v)=(u,v) , i.e., (u,v) is the unique solution to ( 1.1 ). Furthermore, due to the constant C does not depend on the choice of T , (u,v) is actually the global solution. The choice of K gives that the solution (u,v) satisfies \|(u,v)\|_{\mathcal{X}(t)}\leq 3C\big(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\big) for any t>1 , which yields the validness of the estimates shown in Theorem 2.2 . ∎ 4.2.1. The proof of ( 4.16 ) By the definition ( 4.15 ) of \mathcal{N}, we now provide the estimate for ( u^{l},v^{l} ). In the case of \delta_{1},\delta_{2}>(m+1)^{2}(n+2\sigma-1)^{2} with \sigma>1 , by Proposition 4.1 , we have \displaystyle\|u^{l}(t,\cdot)\|_{L^{2}}\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\|(u_{0},u_{1})\|_{L^{2}}, (4.20) \displaystyle\|u^{l}(t,\cdot)\|_{\dot{H}^{\sigma}}\lesssim t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\|(u_{0},u_{1})\|_{D^{\sigma}}, \displaystyle\|\partial_{t}u^{l}(t,\cdot)\|_{L^{2}}\lesssim t^{m-(m+1)(1+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\|(u_{0},u_{1})\|_{D^{1}}, \displaystyle\|\partial_{t}u^{l}(t,\cdot)\|_{\dot{H}^{\sigma-1}}\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\|(u_{0},u_{1})\|_{D^{\sigma}} and \displaystyle\|v^{l}(t,\cdot)\|_{L^{2}}\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(v_{0},v_{1})\|_{L^{2}}, (4.21) \displaystyle\|v^{l}(t,\cdot)\|_{\dot{H}^{\sigma}}\lesssim t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(v_{0},v_{1})\|_{D^{\sigma}}, \displaystyle\|\partial_{t}v^{l}(t,\cdot)\|_{L^{2}}\lesssim t^{m-(m+1)(1+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(v_{0},v_{1})\|_{D^{1}}, \displaystyle\|\partial_{t}v^{l}(t,\cdot)\|_{\dot{H}^{\sigma-1}}\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(v_{0},v_{1})\|_{D^{\sigma}}. By the definitions ( 4.8 ), ( 4.9 ) of M_{1}(t,u^{l}) , M_{2}(t,v^{l}) , we get M_{1}(t,u^{l})+M_{2}(t,v^{l})\lesssim\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}, which yields \|(u^{l},v^{l})\|_{\mathcal{X}(T)}\lesssim\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}} by ( 4.12 ). In the same way, we can deal with the remaining cases \displaystyle\text{Case 2:}\ \delta_{1}>\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \delta_{2}=\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \sigma>1, (4.22) \displaystyle\text{Case 3:}\ \delta_{1}=\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \delta_{2}>\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \sigma>1, (4.23) \displaystyle\text{Case 4:}\ \delta_{1}=\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \delta_{2}=\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \sigma>1, (4.24) \displaystyle\text{Case 5:}\ \delta_{1}>\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \delta_{2}>\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \sigma=1, (4.25) \displaystyle\text{Case 6:}\ \delta_{1}>\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \delta_{2}=\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \sigma=1, (4.26) \displaystyle\text{Case 7:}\ \delta_{1}=\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \delta_{2}>\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \sigma=1, (4.27) \displaystyle\text{Case 8:}\ \delta_{1}=\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \delta_{2}=\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \sigma=1. (4.28) Combining the above eight cases gives the estimates of (u^{l},v^{l}) as \|(u^{l},v^{l})\|_{\mathcal{X}(T)}\lesssim\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}. (4.29) In what follows, we will use Proposition 4.2 and Duhamel’s principle to establish the estimates for the nonlinear term, namely, \|\big(H_{1}(v),H_{2}(u)\big)\|_{\mathcal{X}(T)}\lesssim\|(u,v)\|_{\mathcal{X}{(T)}}^{p}+\|(u,v)\|_{\mathcal{X}{(T)}}^{q}. (4.30) From ( 4.29 ) and ( 4.30 ), we have essentially proved ( 4.16 ). Before proving ( 4.30 ), let us list some primary results that will be used throughout the whole proof. Lemma 4.1 ( [ 17 ] ) . For a function w\in\mathcal{X}(T) , T>1 and 1\leq\tau1\ \text{and}\ s>\lceil\sigma-1\rceil\ (\geq 1), (4.31) \displaystyle\||w(\tau,\cdot)|^{s}\|_{\dot{H}^{\sigma-2}}\lesssim\|w(\tau,\cdot)\|_{L^{r_{3}}}^{s-1}\||D|^{\sigma-2}w(\tau,\cdot)\|_{L^{r_{4}}},\ \sigma>2\ \text{and}\ s>\lceil\sigma-2\rceil\ (\geq 1), (4.32) where \frac{s-1}{r_{1}}+\frac{1}{r_{2}}=\frac{s-1}{r_{3}}+\frac{1}{r_{4}}=\frac{1}{2} . (2) For any r\geq 1 , it holds that \displaystyle\|w(\tau,\cdot)\|_{L^{r}}\lesssim\|w(\tau,\cdot)\|_{\dot{H}^{\sigma}}^{\theta_{1}(r)}\|w(\tau,\cdot)\|_{L^{2}}^{1-\theta_{1}(r)}, (4.33) \displaystyle\||D|^{\sigma-1}w(\tau,\cdot)\|_{L^{r}}\lesssim\|w(\tau,\cdot)\|_{\dot{H}^{\sigma}}^{\theta_{2}(r)}\|w(\tau,\cdot)\|_{L^{2}}^{1-\theta_{2}(r)}\ \text{for}\ \sigma>1, (4.34) \displaystyle\||D|^{\sigma-2}w(\tau,\cdot)\|_{L^{r}}\lesssim\|w(\tau,\cdot)\|_{\dot{H}^{\sigma}}^{\theta_{3}(r)}\|w(\tau,\cdot)\|_{L^{2}}^{1-\theta_{3}(r)}\ \text{for}\ \sigma>2, (4.35) where \displaystyle\theta_{1}(r)=(\frac{1}{2}-\frac{1}{r})\frac{n}{\sigma}\in[0,1], (4.36) \displaystyle\theta_{2}(r)=\bigl(\frac{1}{2}-\frac{1}{r}+\frac{\sigma-1}{n}\bigr)\frac{n}{\sigma}\in[\frac{\sigma-1}{\sigma},1], (4.37) \displaystyle\theta_{3}(r)=(\frac{1}{2}-\frac{1}{r}+\frac{\sigma-2}{n})\frac{n}{\sigma}\in[\frac{\sigma-2}{\sigma},1]. (4.38) Remark 4.1 . A detailed proof of Lemma 4.1 was given in Corollary 3.3 [ 17 ] , and therefore it will not be repeated here. Lemma 4.2 . In the case of \delta_{1},\delta_{2}\geq(m+1)^{2}(n+2\sigma-1)^{2} with \sigma\geq 1 , p,q>1 , for any 1\leq\tau\leq t\leq T and any (u,v),(\tilde{u},\tilde{v})\in\mathcal{X}(T) , we have \displaystyle\||v(\tau,\cdot)|^{p}\|_{L^{1}}\lesssim\tau^{(m+1)n+(-(m+1)n-\beta_{2}+1)p}\ell_{2}(\tau)^{p\theta_{1}(2p)}M_{2}(\tau,v)^{p}, (4.39) \displaystyle\||v(\tau,\cdot)|^{p}\|_{L^{2}}\lesssim\tau^{(m+1)\frac{n}{2}+(-(m+1)n-\beta_{2}+1)p}\ell_{2}(\tau)^{p\theta_{1}(2p)}M_{2}(\tau,v)^{p}, (4.40) \displaystyle\||v(\tau,\cdot)|^{p}\|_{\dot{H}^{\sigma-1}}\lesssim\tau^{(m+1)(\frac{n}{2}-\sigma+1)+(-(m+1)n-\beta_{2}+1)p}\ell_{2}(\tau)^{\theta_{1}(r_{1})(p-1)+\theta_{2}(r_{2})}M_{2}(\tau,v)^{p}, (4.41) and if \sigma>2 , \displaystyle\||v(\tau,\cdot)|^{p}\|_{\dot{H}^{\sigma-2}}\lesssim\tau^{(m+1)(\frac{n}{2}-\sigma+2)+(-(m+1)n-\beta_{2}+1)p}\ell_{2}(\tau)^{\theta_{1}(r_{3})(p-1)+\theta_{3}(r_{4})}M_{2}(\tau,v)^{p}, (4.42) where \frac{p-1}{r_{1}}+\frac{1}{r_{2}}=\frac{p-1}{r_{3}}+\frac{1}{r_{4}}=\frac{1}{2} , \beta_{2} is defined by ( 1.12 ). As for u(\tau,x) , we have \displaystyle\||u(\tau,\cdot)|^{q}\|_{L^{1}}\lesssim\tau^{(m+1)n+(-(m+1)n-\beta_{1}+1)q}\ell_{1}(\tau)^{q\theta_{1}(2q)}M_{1}(\tau,u)^{q}, (4.43) \displaystyle\||u(\tau,\cdot)|^{q}\|_{L^{2}}\lesssim\tau^{(m+1)\frac{n}{2}+(-(m+1)n-\beta_{1}+1)q}\ell_{1}(\tau)^{q\theta_{1}(2q)}M_{1}(\tau,u)^{q}, (4.44) \displaystyle\||u(\tau,\cdot)|^{q}\|_{\dot{H}^{\sigma-1}}\lesssim\tau^{(m+1)(\frac{n}{2}-\sigma+1)+(-(m+1)n-\beta_{1}+1)q}\ell_{1}(\tau)^{\theta_{1}(r_{5})(q-1)+\theta_{2}(r_{6})}M_{1}(\tau,u)^{q}, (4.45) and if \sigma>2 , \displaystyle\||u(\tau,\cdot)|^{q}\|_{\dot{H}^{\sigma-2}}\lesssim\tau^{(m+1)(\frac{n}{2}-\sigma+2)+(-(m+1)n-\beta_{1}+1)q}\ell_{1}(\tau)^{\theta_{1}(r_{7})(q-1)+\theta_{3}(r_{8})}M_{2}(\tau,u)^{q}. (4.46) where \frac{q-1}{r_{5}}+\frac{1}{r_{6}}=\frac{q-1}{r_{7}}+\frac{1}{r_{8}}=\frac{1}{2} , \beta_{1} is defined by ( 1.12 ). Remark 4.2 . We point out that the numbers r_{i}(i=1,\cdots,8) in Lemma 4.2 are well-defined under the assumptions of Theorem 2.2 . About p,r_{1},r_{2} , we need to ensure that \theta_{1}(2p)=(\frac{1}{2}-\frac{1}{2p})\frac{n}{\sigma}\in[0,1]\ \ \ \text{and}\ \ \left\{\begin{aligned} &\theta_{1}(r_{1})=(\frac{1}{2}-\frac{1}{r_{1}})\frac{n}{\sigma}\in[0,1],\\ &\theta_{2}(r_{2})=(\frac{1}{2}-\frac{1}{r_{2}}+\frac{\sigma-1}{n})\frac{n}{\sigma}\in[\frac{\sigma-1}{\sigma},1],\end{aligned}\right. which are equal to \left\{\begin{aligned} &p\geq 1,\ &n\leq 2\sigma,\\ &1\leq p\leq\frac{n}{n-2\sigma},\ &n>2\sigma,\end{aligned}\right. (4.47) \left\{\begin{aligned} &\frac{1}{r_{1}}\in(0,\frac{1}{2}],\ &n\leq 2\sigma,\\ &\frac{1}{r_{1}}\in[\frac{n-2\sigma}{2n},\frac{1}{2}],\ &n>2\sigma,\end{aligned}\right.\quad\text{and}\ \left\{\begin{aligned} &\frac{1}{r_{2}}\in(0,\frac{1}{2}],\ &n\leq 2,\\ &\frac{1}{r_{2}}\in[\frac{n-2}{2n},\frac{1}{2}],\ &n>2.\end{aligned}\right. (4.48) Clearly, ( 4.47 ) holds under the assumptions regarding p in Theorem 2.2 . Since \frac{p-1}{r_{1}}+\frac{1}{r_{2}}=\frac{1}{2} and \sigma\geq 1 in ( 4.41 ), in order to ensure the validity of ( 4.48 ), it suffices to guarantee \frac{1}{2}\in\left\{\begin{aligned} &(0,\frac{p}{2}],\ &\quad&n\leq 2,\\ &\big(\frac{1}{2}-\frac{1}{n},\frac{p}{2}\big],\ &\quad&22\sigma,\end{aligned}\right. the first two cases are automatically satisfied for p\geq 1 , and the condition p\leq 1+\frac{2}{n-2\sigma}\ (n>2\sigma) in Theorem 2.2 gives the validity of the third case. The existence of r_{5},r_{6} can be obtained through the same derivation process. As for r_{3},r_{4} , we need to ensure that \theta_{1}(r_{3})\in[0,1],\theta_{3}(r_{4})\in[0,1] , which are equivalent to ensuring \left\{\begin{aligned} &\frac{1}{r_{3}}\in(0,\frac{1}{2}],\ &n\leq 2\sigma,\\ &\frac{1}{r_{3}}\in[\frac{n-2\sigma}{2n},\frac{1}{2}],\ &n>2\sigma,\end{aligned}\right.\quad\text{and}\ \left\{\begin{aligned} &\frac{1}{r_{4}}\in(0,\frac{1}{2}],\ &n\leq 4,\\ &\frac{1}{r_{4}}\in[\frac{n-4}{2n},\frac{1}{2}],\ &n>4.\end{aligned}\right. (4.49) Note that \sigma\geq 2 in ( 4.42 ) and \frac{p-1}{r_{3}}+\frac{1}{r_{4}}=\frac{1}{2} , it suffices to verify that \frac{1}{2}\in\left\{\begin{aligned} &(0,\frac{p}{2}],\ &\quad&n\leq 4,\\ &\big(\frac{1}{2}-\frac{2}{n},\frac{p}{2}\big],\ &\quad&42\sigma.\end{aligned}\right. Hence, it is sufficient to guarantee p\leq 1+\frac{4}{n-2\sigma} (when n>2\sigma ). Obviously, the condition p\leq 1+\frac{2}{n-2\sigma}\ (n>2\sigma) in Theorem 2.2 can ensure this. The method of demonstrating the existence of r_{7},r_{8} is the same as that for r_{3},r_{4} . Proof of Lemma 4.2 . By ( 4.33 ) in Lemma 4.1 and the definition ( 4.9 ) of M_{2}(\tau,v) , we obtain \displaystyle\||v(\tau,\cdot)|^{p}\|_{L^{2}}=\|v(\tau,\cdot)\|_{L^{2p}}^{p}\lesssim\|v(\tau,\cdot)\|_{\dot{H}^{\sigma}}^{p\theta_{1}(2p)}\|v(\tau,\cdot)\|_{\dot{L}^{2}}^{p(1-\theta_{1}(2p))} \displaystyle\lesssim\big(\tau^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\ell_{2}(\tau)M_{2}(\tau,v)\big)^{p\theta_{1}(2p)}\big(\tau^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}M_{2}(\tau,v)\big)^{p(1-\theta_{1}(2p))} \displaystyle\quad=\tau^{\big(-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}\big)p-(m+1)\sigma p\theta_{1}(2p)}\ell_{2}(\tau)^{p\theta_{1}(2p)}M_{2}(\tau,v)^{p} \displaystyle\quad=\tau^{\frac{(m+1)n}{2}+(-(m+1)n-\beta_{2}+1)p}\ell_{2}(\tau)^{p\theta_{1}(2p)}M_{2}(\tau,v)^{p}. It follows from H \ddot{\text{o}} lder’s inequality that \displaystyle\||v(\tau,\cdot)|^{p}\|_{L^{1}}\leq\||v(\tau,\cdot)|^{p}\|_{L^{2}}\big(\int_{|x|\leq\phi_{m}(\tau)-\phi_{m}(1)+M}dx\big)^{\frac{1}{2}}\lesssim\tau^{\frac{(m+1)n}{2}}\||v(\tau,\cdot)|\|_{L^{2}} \displaystyle\lesssim\tau^{(m+1)n+(-(m+1)n-\beta_{2}+1)p}\ell_{2}(\tau)^{p\theta_{1}(2p)}M_{2}(\tau,v)^{p}. By ( 4.31 ), ( 4.33 )-( 4.34 ) in Lemma 4.1 , we have \displaystyle\||v(\tau,\cdot)|^{p}\|_{\dot{H}^{\sigma-1}}\lesssim\|v(\tau,\cdot)\|_{L^{r_{1}}}^{p-1}\||D|^{\sigma-1}v(\tau,\cdot)\|_{L^{r_{2}}} \displaystyle\lesssim\|v(\tau,\cdot)\|_{\dot{H}^{\sigma}}^{\theta_{1}(r_{1})(p-1)+\theta_{2}(r_{2})}\|v(\tau,\cdot)\|_{L^{2}}^{(1-\theta_{1}(r_{1}))(p-1)+1-\theta_{2}(r_{2})} \displaystyle\lesssim\tau^{\big(-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}\big)p-(m+1)\sigma(\theta_{1}(r_{1})(p-1)+\theta_{2}(r_{2}))}\ell_{2}(\tau)^{\theta_{1}(r_{1})(p-1)+\theta_{2}(r_{2}))}M_{2}(\tau,v)^{p} \displaystyle=\tau^{(-(m+1)n-\beta_{2}+1)p+(m+1)(\frac{n}{2}-\sigma+1)}\ell_{2}(\tau)^{\theta_{1}(r_{1})(p-1)+\theta_{2}(r_{2}))}M_{2}(\tau,v)^{p}; while if \sigma>2 , using ( 4.9 ), ( 4.32 )-( 4.33 ) and ( 4.35 ), we immediately obtain ( 4.42 ). Using the same argument, ( 4.43 )-( 4.46 ) can be derived from Lemma 4.1 and the definition ( 4.8 ) of M_{1}(\tau,u) , so we omit the details. ∎ With the preliminary results of Lemma 4.2 , we now proceed to prove ( 4.30 ). Proof of ( 4.30 ). Case 1: \delta_{1},\delta_{2}>\big((m+1)(n+2\sigma-1)\big)^{2} with \sigma>1 . The condition p>\tilde{p} yields \displaystyle(m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p<-1, (4.50) then \displaystyle\int_{1}^{t}\tau^{(m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p}d\tau\leq\int_{1}^{\infty}\tau^{(m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p}d\tau<\infty. (4.51) By ( 4.4 ) in Proposition 4.2 , ( 4.39 )-( 4.40 ), ( 4.51 ) and the Duhamel’s principle, we have \displaystyle\|H_{1}(v)(t,\cdot)\|_{L^{2}}\lesssim\int_{1}^{t}\|E_{1}^{\mu_{1},\nu_{1}}(t,\tau,\cdot)\ast|v(\tau,\cdot)|^{p}\|_{L^{2}}d\tau (4.52) \displaystyle\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\int_{1}^{t}\tau^{-\frac{\sqrt{\delta_{1}}-\mu_{1}-1}{2}}\big(\||v(\tau,\cdot)|^{p}\|_{L^{1}}+\tau^{\frac{(m+1)n}{2}}\||v(\tau,\cdot)|^{p}\|_{L^{2}}\big)d\tau \displaystyle\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\int_{1}^{t}\tau^{(m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p}M_{2}(\tau,v)^{p}d\tau \displaystyle\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{p}\int_{1}^{t}\tau^{(m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p}d\tau \displaystyle\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{p}. Using ( 4.5 ) and the same argument as in deriving ( 4.52 ), we get \|\partial_{t}H_{1}(v)(t,\cdot)\|_{L^{2}}\lesssim t^{m-(m+1)(1+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{p}. (4.53) Applying ( 4.4 ), ( 4.39 )-( 4.41 ) yields \displaystyle\|H_{1}(v)(t,\cdot)\|_{\dot{H}^{\sigma}}\lesssim\int_{1}^{t}\|E_{1}^{\mu_{1},\nu_{1}}(t,\tau,\cdot)\ast|v(t,\cdot)|^{p}\|_{\dot{H}^{\sigma}}d\tau \displaystyle\lesssim t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}-1}{2}}\int_{1}^{t}\tau^{-\frac{\sqrt{\delta_{1}}-\mu_{1}-1}{2}}\big(\||v(\tau,\cdot)|^{p}\|_{L^{1}}+\tau^{(m+1)(\frac{n}{2}+\sigma-1)}\||v(\tau,\cdot)|^{p}\|_{\dot{H}^{\sigma-1}}\big)d\tau \displaystyle\lesssim t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}-1}{2}}\int_{1}^{t}\tau^{(m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p}M_{2}(\tau,v)^{p}d\tau \displaystyle\lesssim t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}-1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{p}. (4.54) To estimate \|\partial_{t}H_{1}(v)(t,\cdot)\|_{\dot{H}^{\sigma-1}} , if \sigma>2 , then by ( 4.5 ), ( 4.39 )-( 4.42 ), we have \displaystyle\|\partial_{t}H_{1}(v)(t,\cdot)\|_{\dot{H}^{\sigma-1}}\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}-1}{2}}\int_{1}^{t}\tau^{-\frac{\sqrt{\delta_{1}}-\mu_{1}-1}{2}}\big(\||v(\tau,\cdot)|^{p}\|_{L^{1}} \displaystyle\quad\quad\quad\quad+\tau^{(m+1)(\frac{n}{2}+\sigma-1)}\||v(\tau,\cdot)|^{p}\|_{\dot{H}^{\sigma-1}}+\tau^{(m+1)(\frac{n}{2}+\sigma-2)}\||v(\tau,\cdot)|^{p}\|_{\dot{H}^{\sigma-2}}\big)d\tau \displaystyle\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}-1}{2}}\int_{1}^{t}\tau^{(m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p}M_{2}(\tau,v)^{p}d\tau \displaystyle\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}-1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{p}; (4.55) while for 1<\sigma\leq 2 , it follows from ( 4.5 ), ( 4.39 )- ( 4.41 ) that \displaystyle\|\partial_{t}H_{1}(v)(t,\cdot)\|_{\dot{H}^{\sigma-1}}\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}-1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{p}. (4.56) Combining ( 4.52 )-( 4.56 ) and the definition ( 4.8 ) of M_{1}(t,H_{1}(v)) , we get M_{1}(t,H_{1}(v))\lesssim\|(u,v)\|_{\mathcal{X}(t)}^{p}. (4.57) Since q>\tilde{q} , it can be inferred that \displaystyle(m+1)n+\beta_{2}+(-(m+1)n-\beta_{1}+1)q<-1, (4.58) then \int_{1}^{t}\tau^{(m+1)n+\beta_{2}+(-(m+1)n-\beta_{1}+1)q}d\tau\leq\int_{1}^{\infty}\tau^{(m+1)n+\beta_{2}+(-(m+1)n-\beta_{1}+1)p}d\tau<\infty. (4.59) From ( 4.4 ), ( 4.43 )-( 4.44 ) and ( 4.59 ), we conclude that \begin{aligned} &\|H_{2}(u)(t,\cdot)\|_{L^{2}}\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\int_{1}^{t}\tau^{-\frac{\sqrt{\delta_{2}}-\mu_{2}-1}{2}}\big(\||u(\tau,\cdot)|^{q}\|_{L^{1}}+\tau^{(m+1)\frac{n}{2}}\||u(\tau,\cdot)|^{q}\|_{L^{2}}\big)d\tau\end{aligned} \displaystyle\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\int_{1}^{t}\tau^{(m+1)n+\beta_{2}+(-(m+1)n-\beta_{1}+1)q}M_{1}(\tau,u)^{q}d\tau \displaystyle\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{q}\int_{1}^{t}\tau^{(m+1)n+\beta_{2}+(-(m+1)n-\beta_{1}+1)q}d\tau \displaystyle\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{q}. (4.60) The estimates for \|\partial_{t}H_{2}(u)(t,\cdot)\|_{L^{2}} , \|H_{2}(u)(t,\cdot)\|_{\dot{H}^{\sigma}} , \|\partial_{t}H_{2}(u)(t,\cdot)\|_{\dot{H}^{\sigma-1}} can be derived by using ( 4.4 )-( 4.5 ), ( 4.43 )-( 4.46 ) and ( 4.59 ), along with the same argument as applied in obtaining ( 4.53 )-( 4.56 ). Hence, we present the results as follows and omit the details \displaystyle\|\partial_{t}H_{2}(u)(t,\cdot)\|_{L^{2}}\lesssim t^{m-(m+1)(1+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{q}, (4.61) \displaystyle\|H_{2}(u)(t,\cdot)\|_{\dot{H}^{\sigma}}\lesssim t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{q}, (4.62) \displaystyle\|\partial_{t}H_{2}(u)(t,\cdot)\|_{\dot{H}^{\sigma-1}}\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{q}. (4.63) It follows ( 4.60 )-( 4.63 ) that M_{2}(t,H_{2}(u))\lesssim\|(u,v)\|_{\mathcal{X}(t)}^{q}. (4.64) Therefore, we can immediately derive ( 4.30 ) by combining ( 4.57 ) and ( 4.64 ). Case 2: \delta_{1}>\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \delta_{2}=\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \sigma>1 . Clearly, \ell_{1}(\tau)=1 and \ell_{2}(\tau)=(1+\log\tau)^{\frac{1}{2}} , so by Proposition 4.2 and Lemma 4.2 , we have \displaystyle\|H_{1}(v)(t,\cdot)\|_{L^{2}}\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}} \displaystyle\quad\quad\quad\quad\times\int_{1}^{t}\tau^{(m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p}(1+\log\tau)^{\frac{p\theta_{1}(2p)}{2}}M_{2}(\tau,u)^{p}d\tau, (4.65) \displaystyle\|\partial_{t}H_{1}(v)(t,\cdot)\|_{L^{2}}\lesssim t^{m-(m+1)(1+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}} \displaystyle\quad\quad\quad\quad\times\int_{1}^{t}\tau^{(m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p}(1+\log\tau)^{\frac{p\theta_{1}(2p)}{2}}M_{2}(\tau,v)^{p}d\tau, (4.66) \displaystyle\|H_{1}(v)(t,\cdot)\|_{\dot{H}^{\sigma}}\lesssim t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}} \displaystyle\ \ \times\int_{1}^{t}\tau^{(m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p}(1+\log\tau)^{\frac{p\theta_{1}(2p)+\theta_{1}(r_{1})(p-1)+\theta_{2}(r_{2})}{2}}M_{2}(\tau,v)^{p}d\tau, (4.67) if \sigma>2 , \displaystyle\|\partial_{t}H_{1}(v)(t,\cdot)\|_{\dot{H}^{\sigma-1}}\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\int_{1}^{t}\tau^{(m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p} (4.68) \displaystyle\times(1+\log\tau)^{\frac{p\theta_{1}(p)+\theta_{1}(r_{1})(p-1)+\theta_{2}(r_{2})+\theta_{1}(r_{3})(p-1)+\theta_{3}(r_{4})}{2}}M_{2}(\tau,v)^{p}d\tau, while for 1<\sigma\leq 2 , \displaystyle\|\partial_{t}H_{1}(v)(t,\cdot)\|_{\dot{H}^{\sigma-1}}\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\int_{1}^{t}\tau^{(m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p} (4.69) \displaystyle\times(1+\log\tau)^{\frac{p\theta_{1}(2p)+\theta_{2}(r_{2})+\theta_{1}(r_{1})(p-1)}{2}}M_{2}(\tau,v)^{p}d\tau. Since (m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p<-1, we can set (m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p=-1-\gamma . Then \tau^{-\frac{\gamma}{2}}(1+\log{\tau})^{\frac{p\theta_{1}(2p)}{2}}\lesssim 1 holds for large \tau , so \displaystyle\int_{1}^{t}\tau^{(m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p}(1+\log\tau)^{\frac{\theta_{1}(p)+\theta_{1}(2p)}{2}p}d\tau\lesssim\int_{1}^{\infty}\tau^{-1-\frac{\gamma}{2}}d\tau<\infty. (4.70) Thus from ( 4.65 ), we infer that \|H_{1}(v)(t,\cdot)\|_{L^{2}}\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{p}. (4.71) Similarly, by ( 4.66 )-( 4.69 ), we obtain \displaystyle\|\partial_{t}H_{1}(v)(t,\cdot)\|_{L^{2}}\lesssim t^{m-(m+1)(1+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{p}, (4.72) \displaystyle\|H_{1}(v)(t,\cdot)\|_{\dot{H}^{\sigma}}\lesssim t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{p}, (4.73) \displaystyle\|\partial_{t}H_{1}(v)(t,\cdot)\|_{\dot{H}^{\sigma-1}}\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{p}. (4.74) Hence, ( 4.71 )-( 4.74 ) show that M_{1}(t,H_{1}(v))\lesssim\|(u,v)\|_{\mathcal{X}(t)}^{p}. (4.75) In view of \delta_{2}=(m+1)^{2}(n+2\sigma-1)^{2} and (m+1)n+\beta_{2}+(-(m+1)n-\beta_{1}+1)q<-1, we see that \frac{\sqrt{\delta_{2}}}{2(m+1)}+\frac{1}{2}-\frac{n}{2}=\sigma>1 and \int_{1}^{t}\tau^{(m+1)n+\beta_{2}+(-(m+1)n-\beta_{1}+1)q}d\tau<\infty , then by Proposition 4.1 and Lemma 4.2 , we have \displaystyle\|H_{2}(u)(t,\cdot)\|_{L^{2}}\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\int_{1}^{t}\tau^{(m+1)n+\beta_{2}+(-(m+1)n-\beta_{1}+1)q}M_{1}(\tau,u)^{q}d\tau \displaystyle\quad\quad\quad\quad\quad\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{q}, (4.76) \displaystyle\|\partial_{t}H_{2}(u)(t,\cdot)\|_{L^{2}}\lesssim t^{m-(m+1)(1+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{q}, (4.77) \displaystyle\|H_{2}(u)(t,\cdot)\|_{\dot{H}^{\sigma}}\lesssim t^{-\frac{\mu_{2}+m}{2}}\int_{1}^{t}(1+\log\frac{t}{\tau})^{\frac{1}{2}}\tau^{(m+1)n+\beta_{2}+(-(m+1)n-\beta_{1}+1)q}M_{1}(\tau,u)^{q}d\tau \displaystyle\quad\lesssim t^{-\frac{\mu_{2}+m}{2}}(1+\log t)^{\frac{1}{2}}\int_{1}^{t}\tau^{(m+1)n+\beta_{2}+(-(m+1)n-\beta_{1}+1)q}M_{1}(\tau,u)^{q}d\tau \displaystyle\quad=t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}(1+\log t)^{\frac{1}{2}}\int_{1}^{t}\tau^{(m+1)n+\beta_{2}+(-(m+1)n-\beta_{1}+1)q}M_{1}(\tau,u)^{q}d\tau \displaystyle\quad\lesssim t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}(1+\log t)^{\frac{1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{q} (4.78) and \displaystyle\|\partial_{t}H_{2}(u)(t,\cdot)\|_{\dot{H}^{\sigma-1}}\lesssim t^{m-\frac{\mu_{2}+m}{2}}(1+\log t)^{\frac{1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{q} (4.79) \displaystyle=t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}(1+\log t)^{\frac{1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{q}. It follows from ( 4.76 )-( 4.79 ) that M_{2}(t,H_{2}(u))\lesssim\|(u,v)\|_{\mathcal{X}(t)}^{q}. (4.80) Consequently, we conclude that ( 4.30 ) holds in this case by ( 4.75 ) and ( 4.80 ). The remaining cases Case 3- Case 8 presented in ( 4.23 )-( 4.28 ) can be handled in the same way as in Case 1 and Case 2. It is worth noting that when \sigma=1 , we should use the definitions ( 4.10 ) and ( 4.11 ) of M_{1}(t,H_{1}(v)) and M_{2}(t,H_{2}(u)) in place of ( 4.8 ) and ( 4.9 ), and we will not elaborate further. ∎ 4.2.2. The proof of ( 4.17 ) In Lemma 3.5 [ 17 ] , we established the following preliminary results, which will be used throughout the whole proof of ( 4.17 ). Lemma 4.3 ( [ 17 ] ) . For any T>1 , 1\leq\tau\leq t\leq T and any w,\tilde{w}\in\mathcal{X}(T) , we have (1) For any s,r\geq 1 , \begin{aligned} &\||w(\tau,\cdot)|^{s}-|\tilde{w}(\tau,\cdot)|^{s}\|_{L^{r}}\lesssim\|w(\tau,\cdot)-\tilde{w}(\tau,\cdot)\|_{\dot{H}^{\sigma}}^{\theta_{1}(rs)}\|w(\tau,\cdot)-\tilde{w}(\tau,\cdot)\|_{L^{2}}^{1-\theta_{1}(rs)}\\ &\times\Big(\|w(\tau,\cdot)\|_{\dot{H}^{\sigma}}^{\theta_{1}(rs)(s-1)}\|w(\tau,\cdot)\|_{L^{2}}^{(1-\theta_{1}(rs))(s-1)}\|\tilde{w}(\tau,\cdot)\|_{\dot{H}^{\sigma}}^{\theta_{1}(rs)(s-1)}\|\tilde{w}(\tau,\cdot)\|_{L^{2}}^{(1-\theta_{1}(rs))(s-1)}\Bigr).\end{aligned} (4.81) (2) If \sigma>1 and s>\lceil\sigma-1\rceil+1\ (=\lceil\sigma\rceil\geq 2) , \displaystyle\||w(\tau,\cdot)|^{s}-|\tilde{w}(\tau,\cdot)|^{s}\|_{\dot{H}^{\sigma-1}}\lesssim\|w(\tau,\cdot)-\tilde{w}(\tau,\cdot)\|_{\dot{H}^{\sigma}}^{\theta_{2}(b_{1})}\|w(\tau,\cdot)-\tilde{w}(\tau,\cdot)\|_{L^{2}}^{1-\theta_{2}(b_{1})} \displaystyle\times\Big(\|w(\tau,\cdot)\|_{\dot{H}^{\sigma}}^{(s-1)\theta_{1}((s-1)b_{2})}\|w(\tau,\cdot)\|_{L^{2}}^{(s-1)(1-\theta_{1}((s-1)b_{2}))} \displaystyle\quad\quad\quad+\|\tilde{w}(\tau,\cdot)\|_{\dot{H}^{\sigma}}^{(s-1)\theta_{1}((s-1)b_{2})}\|\tilde{w}(\tau,\cdot)\|_{L^{2}}^{(s-1)(1-\theta_{1}((s-1)b_{2}))}\Bigr) \displaystyle+\|w(\tau,\cdot)-\tilde{w}(\tau,\cdot)\|_{\dot{H}^{\sigma}}^{\theta_{1}(b_{3})}\|w(\tau,\cdot)-\tilde{w}(\tau,\cdot)\|_{L^{2}}^{1-\theta_{1}(b_{3})}\bigl(\|w(\tau,\cdot)\|_{\dot{H}^{\sigma}}+\|\tilde{w}(\tau,\cdot)\|_{\dot{H}^{\sigma}}\bigr)^{(s-2)\theta_{1}(b_{5})+\theta_{2}(b_{6})} \displaystyle\quad\quad\times\bigl(\|w(\tau,\cdot)\|_{L^{2}}+\|\tilde{w}(\tau,\cdot)\|_{L^{2}}\bigr)^{(s-2)(1-\theta_{1}(b_{5}))+1-\theta_{2}(b_{6})}. (4.82) (3) If \sigma>2 and s>\lceil\sigma-2\rceil+1\ (=\lceil\sigma-1\rceil\geq 2) , \displaystyle\||w(\tau,\cdot)|^{s}-|\tilde{w}(\tau,\cdot)|^{s}\|_{\dot{H}^{\sigma-2}}\lesssim\|w(\tau,\cdot)-\tilde{w}(\tau,\cdot)\|_{\dot{H}^{\sigma}}^{\theta_{3}(c_{1})}\|w(\tau,\cdot)-\tilde{w}(\tau,\cdot)\|_{L^{2}}^{1-\theta_{3}(c_{1})} \displaystyle\quad\quad\times\Big(\|w(\tau,\cdot)\|_{\dot{H}^{\sigma}}^{(s-1)\theta_{1}((s-1)c_{2})}\|w(\tau,\cdot)\|_{L^{2}}^{(s-1)(1-\theta_{1}((s-1)c_{2}))} \displaystyle\quad\quad\quad\quad\quad\quad+\|\tilde{w}(\tau,\cdot)\|_{\dot{H}^{\sigma}}^{(s-1)\theta_{1}((s-1)c_{2})}\|\tilde{w}(\tau,\cdot)\|_{L^{2}}^{(s-1)(1-\theta_{1}((s-1)c_{2}))}\Bigr) \displaystyle+\|w(\tau,\cdot)-\tilde{w}(\tau,\cdot)\|_{\dot{H}^{\sigma}}^{\theta_{1}(c_{3})}\|w(\tau,\cdot)-\tilde{w}(\tau,\cdot)\|_{L^{2}}^{1-\theta_{1}(c_{3})}\times\bigl(\|w\|_{\dot{H}^{\sigma}}+\|\tilde{w}\|_{\dot{H}^{\sigma}}\bigr)^{(s-2)\theta_{1}(c_{5})+\theta_{3}(c_{6})} \displaystyle\quad\quad\quad\times\bigl(\|w\|_{L^{2}}+\|\tilde{w}\|_{L^{2}}\bigr)^{(s-2)(1-\theta_{1}(c_{5}))+1-\theta_{3}(c_{6})}, (4.83) where \theta_{1},\theta_{2},\theta_{3} are defined in Lemma 4.1 , and b_{i},c_{i},i=1,2,3,5,6 satisfy \displaystyle\frac{1}{b_{1}}+\frac{1}{b_{2}}=\frac{1}{2},\ \ \frac{1}{2}-\frac{1}{b_{3}}=\frac{s-2}{b_{5}}+\frac{1}{b_{6}}, (4.84) \displaystyle\frac{1}{c_{1}}+\frac{1}{c_{2}}=\frac{1}{2},\ \ \frac{1}{2}-\frac{1}{c_{3}}=\frac{s-2}{c_{5}}+\frac{1}{c_{6}}. Lemma 4.4 . In the case of \delta_{1},\delta_{2}\geq(m+1)^{2}(n+2\sigma-1)^{2} with \sigma\geq 1 , for any 1\leq\tau\leq t\leq T and any (u,v),(\tilde{u},\tilde{v})\in\mathcal{X}(T) , we have \displaystyle\||v(\tau,\cdot)|^{p}-|\tilde{v}(\tau,\cdot)|^{p}\|_{L^{1}} \displaystyle\lesssim\tau^{(m+1)n+(-(m+1)n-\beta_{2}+1)p}\ell_{2}(\tau)^{p\theta_{1}(2p)}M_{2}(\tau,v-\tilde{v})\big(M_{2}(\tau,v)^{p-1}+M_{2}(\tau,\tilde{v})^{p-1}\big), (4.85) \displaystyle\||v(\tau,\cdot)|^{p}-|\tilde{v}(\tau,\cdot)|^{p}\|_{L^{2}} \displaystyle\lesssim\tau^{(m+1)\frac{n}{2}+(-(m+1)n-\beta_{2}+1)p}\ell_{2}(\tau)^{p\theta_{1}(2p)}M_{2}(\tau,v-\tilde{v})\big(M_{2}(\tau,v)^{p-1}+M_{2}(\tau,\tilde{v})^{p-1}\big); (4.86) For p>\lceil\sigma\rceil\ (\geq 2) , \displaystyle\||v(\tau,\cdot)|^{p}-|\tilde{v}(\tau,\cdot)|^{p}\|_{\dot{H}^{\sigma-1}} \displaystyle\lesssim\tau^{(m+1)(\frac{n}{2}-\sigma+1)+(-(m+1)n-\beta_{2}+1)p}\ell_{2}(\tau)^{\theta_{2}(b_{1})+(p-1)\theta_{1}((p-1)b_{2})+\theta_{1}(b_{3})+(p-2)\theta_{1}(b_{5})+\theta_{2}(b_{6})} \displaystyle\quad\times M_{2}(\tau,v-\tilde{v})\big(M_{2}(\tau,v)^{p-1}+M_{2}(\tau,\tilde{v})^{p-1}\big); (4.87) If \sigma>2 and for p>\lceil\sigma-1\rceil\ (\geq 2) , \displaystyle\||v(\tau,\cdot)|^{p}-|\tilde{v}(\tau,\cdot)|^{p}\|_{\dot{H}^{\sigma-2}} \displaystyle\lesssim\tau^{(m+1)(\frac{n}{2}-\sigma+2)+(-(m+1)n-\beta_{2}+1)p}\ell_{2}(\tau)^{\theta_{3}(c_{1})+(p-1)\theta_{1}((p-1)c_{2})+\theta_{1}(c_{3})+(p-2)\theta_{1}(c_{5})+\theta_{3}(c_{6})} \displaystyle\quad\times M_{2}(\tau,v-\tilde{v})\big(M_{2}(\tau,v)^{p-1}+M_{2}(\tau,\tilde{v})^{p-1}\big), (4.88) where \frac{1}{b_{1}}+\frac{1}{b_{2}}=\frac{1}{2},\ \ \frac{1}{2}-\frac{1}{b_{3}}=\frac{p-2}{b_{5}}+\frac{1}{b_{6}} , \frac{1}{c_{1}}+\frac{1}{c_{2}}=\frac{1}{2},\ \ \frac{1}{2}-\frac{1}{c_{3}}=\frac{p-2}{c_{5}}+\frac{1}{c_{6}} . \displaystyle\||u(\tau,\cdot)|^{q}-|\tilde{u}(\tau,\cdot)|^{q}\|_{L^{1}} \displaystyle\lesssim\tau^{(m+1)n+(-(m+1)n-\beta_{1}+1)q}\ell_{1}(\tau)^{q\theta_{1}(2q)}M_{1}(\tau,u-\tilde{u})\big(M_{1}(\tau,u)^{q-1}+M_{1}(\tau,\tilde{u})^{q-1}\big), (4.89) \displaystyle\||u(\tau,\cdot)|^{q}-|\tilde{u}(\tau,\cdot)|^{q}\|_{L^{2}} \displaystyle\lesssim\tau^{(m+1)\frac{n}{2}+(-(m+1)n-\beta_{1}+1)q}\ell_{1}(\tau)^{q\theta_{1}(2q)}M_{1}(\tau,u-\tilde{u})\big(M_{1}(\tau,u)^{q-1}+M_{1}(\tau,\tilde{u})^{q-1}\big); (4.90) For q>\lceil\sigma\rceil\ (\geq 2) , \displaystyle\||u(\tau,\cdot)|^{q}-|\tilde{u}(\tau,\cdot)|^{q}\|_{\dot{H}^{\sigma-1}} \displaystyle\lesssim\tau^{(m+1)(\frac{n}{2}-\sigma+1)+(-(m+1)n-\beta_{1}+1)q}\ell_{1}(\tau)^{\theta_{2}(d_{1})+(q-1)\theta_{1}((q-1)d_{2})+\theta_{1}(d_{3})+(q-2)\theta_{1}(d_{5})+\theta_{2}(d_{6})} \displaystyle\quad\times M_{1}(\tau,u-\tilde{u})\big(M_{1}(\tau,u)^{q-1}+M_{2}(\tau,\tilde{u})^{q-1}\big); (4.91) If \sigma>2 and for q>\lceil\sigma-1\rceil\ (\geq 2) , \displaystyle\||u(\tau,\cdot)|^{q}-|\tilde{u}(\tau,\cdot)|^{q}\|_{\dot{H}^{\sigma-2}} \displaystyle\lesssim\tau^{(m+1)(\frac{n}{2}-\sigma+2)+(-(m+1)n-\beta_{1}+1)q}\ell_{1}(\tau)^{\theta_{3}(e_{1})+(q-1)\theta_{1}((q-1)e_{2})+\theta_{1}(e_{3})+(q-2)\theta_{1}(e_{5})+\theta_{3}(e_{6})} \displaystyle\quad\times M_{1}(\tau,u-\tilde{u})\big(M_{1}(\tau,u)^{q-1}+M_{2}(\tau,\tilde{u})^{q-1}\big), (4.92) where \frac{1}{d_{1}}+\frac{1}{d_{2}}=\frac{1}{2},\ \ \frac{1}{2}-\frac{1}{d_{3}}=\frac{q-2}{d_{5}}+\frac{1}{d_{6}} , \frac{1}{e_{1}}+\frac{1}{e_{2}}=\frac{1}{2},\ \ \frac{1}{2}-\frac{1}{e_{3}}=\frac{q-2}{e_{5}}+\frac{1}{e_{6}} . Remark 4.3 . Before proving Lemma 4.4 , we need to show that, under the assumptions of Theorem 2.2 , the coefficients b_{i},c_{i},d_{i},e_{i} ( i=1,\cdots,6 ) exist. First, the parameter p needs to satisfy \theta_{1}(2p)\in[0,1], which is exactly the same as ( 4.47 ). The parameters b_{1},b_{2} need to satisfy \theta_{1}((p-1)b_{2})\in[0,1],\theta_{2}(b_{1})\in[\frac{\sigma-1}{\sigma},1] , respectively, which is equivalent to \frac{1}{b_{1}}\in\left\{\begin{aligned} &(0,\frac{1}{2}],\ &n\leq 2,\\ &[\frac{n-2}{2n},\frac{1}{2}],\ &n>2\end{aligned}\right.\ \text{and}\ \frac{1}{b_{2}}\in\left\{\begin{aligned} &(0,\frac{p-1}{2}],\ &n\leq 2\sigma,\\ &[\frac{(p-1)(n-2\sigma)}{2n},\frac{p-1}{2}],\ &n>2\sigma.\end{aligned}\right. Since \frac{1}{b_{1}}+\frac{1}{b_{2}}=\frac{1}{2} and \sigma\geq 1 , it suffices to ensure that \frac{1}{2}\in\left\{\begin{aligned} &(0,\frac{p}{2}],\ &n\leq 2,\\ &(\frac{n-2}{2n},\frac{p}{2}],\ &22\sigma.\end{aligned}\right. The condition p\leq 1+\frac{2}{n-2\sigma} (if n>2\sigma ) in Theorem 2.2 ensures that the above expressions are satisfied. Regarding b_{3},b_{5},b_{6} , it is necessary to ensure that \theta_{1}(b_{3}),\theta_{1}(b_{5})\in[0,1],\theta_{2}(b_{6})\in[\frac{\sigma-1}{\sigma},1] . This is equivalent to simply ensuring that \frac{1}{b_{3}},\frac{1}{b_{5}}\in\left\{\begin{aligned} &(0,\frac{1}{2}],\ &n\leq 2\sigma,\\ &[\frac{n-2\sigma}{2n},\frac{1}{2}],\ &n>2\sigma\end{aligned}\right.\ \ \text{and}\ \ \frac{1}{b_{6}}\in\left\{\begin{aligned} &(0,\frac{1}{2}],\ &n\leq 2,\\ &[\frac{n-2}{2n},\frac{1}{2}],\ &n>2.\end{aligned}\right. (4.93) For convenience, let \frac{1}{b_{4}}:=\frac{p-2}{b_{5}}+\frac{1}{b_{6}} , and note that \sigma\geq 1 , the restrictions ( 4.93 ) on b_{i}(i=3,5,6) turn into \frac{1}{b_{4}}\in\left\{\begin{aligned} &(0,\frac{p-1}{2}],\ &n\leq 2,\\ &(\frac{n-2}{2n},\frac{p-1}{2}],\ &22\sigma.\end{aligned}\right. (4.94) Moreover, due to \frac{p-2}{b_{5}}+\frac{1}{b_{6}}=\frac{1}{2}-\frac{1}{b_{3}} , together with ( 4.93 ), we see b_{4} also needs to satisfy \frac{1}{b_{4}}\in\left\{\begin{aligned} &[0,\frac{1}{2}),\ &n\leq 2\sigma,\\ &[0,\frac{\sigma}{n}],\ &n>2\sigma.\end{aligned}\right. (4.95) Hence, the existence for b_{i},i=3,5,6 reduce to ensuring that both ( 4.94 ) and ( 4.95 ) hold simultaneously, which means that we need to guarantee \left\{\begin{aligned} &[0,\frac{1}{2})\cap(0,\frac{p-1}{2}]\neq\emptyset,\ &n\leq 2,\\ &[0,\frac{1}{2})\cap(\frac{n-2}{2n},\frac{p-1}{2}]\neq\emptyset,\ &22\sigma.\end{aligned}\right. (4.96) The condition p\leq 1+\frac{2}{n-2\sigma} (when n>2\sigma ) ensures the validity of ( 4.96 ). So far, we have completed the examination of the existence of b_{i},i=1,2,3,5,6 . Under the assumption \sigma\geq 2 , we need to guarantee \theta_{1}((p-1)c_{2}),\theta_{1}(c_{3}),\theta_{1}(c_{5})\in[0,1] , \theta_{3}(c_{1})\in[\frac{\sigma-2}{\sigma},1] and \theta_{3}(c_{6})\in[\frac{\sigma-2}{\sigma},1] . Repeating the argument for the existence of b_{i} , one can similarly deduce that the condition p\leq 1+\frac{4}{n-2\sigma}\ (\text{when}\ n>2\sigma) guarantees the existence of c_{i},i=1,2,3,5,6 . The verification of the existence of d_{i},e_{i},i=1,2,3,5,6 follows the same process as above and will not be repeated here. Proof of Lemma 4.4 . By ( 4.81 ) in Lemma 4.3 and the definitions of M_{2}(\tau,v),M_{2}(\tau,\tilde{v}),M_{2}(\tau,v-\tilde{v}) , we have \begin{aligned} &\||v(\tau,\cdot)|^{p}-|\tilde{v}(\tau,\cdot)|^{p}\|_{L^{2}}\lesssim\|v(\tau,\cdot)-\tilde{v}(\tau,\cdot)\|_{\dot{H}^{\sigma}}^{\theta_{1}(2p)}\|v(\tau,\cdot)-\tilde{v}(\tau,\cdot)\|_{L^{2}}^{1-\theta_{1}(2p)}\\ &\times\Bigl(\|v(\tau,\cdot)\|_{\dot{H}^{\sigma}}^{\theta_{1}(2p)(p-1)}\|v(\tau,\cdot)\|_{L^{2}}^{(1-\theta_{1}(2p))(p-1)}+\|\tilde{v}(\tau,\cdot)\|_{\dot{H}^{\sigma}}^{\theta_{1}(2p)(p-1)}\|\tilde{v}(\tau,\cdot)\|_{L^{2}}^{(1-\theta_{1}(2p))(p-1)}\Bigr)\\ &\lesssim\tau^{(-\frac{(m+1)n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2})p-(m+1)\sigma p\theta_{1}(2p)}\ell_{2}(\tau)^{p\theta_{1}(2p)}M_{2}(\tau,v-\tilde{v})\big(M_{2}(\tau,v)^{p-1}+M_{2}(\tau,\tilde{v})^{p-1}\big)\\ &=\tau^{\frac{(m+1)n}{2}+(-(m+1)n-\beta_{2}+1)p}\ell_{2}(\tau)^{p\theta_{1}(2p)}M_{2}(\tau,v-\tilde{v})\big(M_{2}(\tau,v)^{p-1}+M_{2}(\tau,\tilde{v})^{p-1}\big).\end{aligned} By H \ddot{\text{o}} lder’s inequality, we have \displaystyle\||v(\tau,\cdot)|^{p}-|\tilde{v}(\tau,\cdot)|^{p}\|_{L^{1}}\leq\||v(\tau,\cdot)|^{p}-|\tilde{v}(\tau,\cdot)|^{p}\|_{L^{2}}\big(\int_{|x|\leq\phi_{m}(\tau)-\phi_{m}(1)+M}dx\big)^{\frac{1}{2}} \displaystyle\lesssim\tau^{\frac{(m+1)n}{2}}\||v(\tau,\cdot)|^{p}-|\tilde{v}(\tau,\cdot)|^{p}\|_{L^{2}} \displaystyle\lesssim\tau^{(m+1)n+(-(m+1)n-\beta_{2}+1)p}\ell_{2}(\tau)^{p\theta_{1}(2p)}M_{2}(\tau,v-\tilde{v})\big(M_{2}(\tau,v)^{p-1}+M_{2}(\tau,\tilde{v})^{p-1}\big). The proofs of ( 4.87 )-( 4.92 ) follow a similar approach so the details are omitted. ∎ Subsequently, we use Lemma 4.4 along with Proposition 4.2 to establish the proof of ( 4.17 ). Proof of ( 4.17 ). Case 1: \delta_{1},\delta_{2}>(m+1)^{2}(n+2\sigma-1)^{2} with \sigma>1 . In this case, \frac{\sqrt{\delta_{1}}}{2(m+1)}+\frac{1}{2}-\frac{n}{2}>\sigma>1 , using ( 4.4 ) in Proposition 4.2 , ( 4.85 )-( 4.86 ) in Lemma 4.4 and the integrability condition ( 4.51 ), we obtain \displaystyle\|H_{1}(v)(t,\cdot)-H_{1}(\tilde{v})(t,\cdot)\|_{L^{2}}\lesssim\int_{1}^{t}\|E_{1}^{\mu_{1},\nu_{1}}(t,\tau,x)\ast\big(|v(\tau,x)|^{p}-|\tilde{v}(\tau,x)|^{p}\big)\|_{L^{2}}d\tau \displaystyle\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\int_{1}^{t}\tau^{-\frac{\sqrt{\delta_{1}}-\mu_{1}-1}{2}}\big(\||v(\tau,\cdot)|^{p}-|\tilde{v}(\tau,\cdot)|^{p}\|_{L^{1}}d\tau \displaystyle\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt+\tau^{(m+1)\frac{n}{2}}\||v(\tau,\cdot)|^{p}-|\tilde{v}(\tau,\cdot)|^{p}\|_{L^{2}}\big)d\tau \displaystyle\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\int_{1}^{t}\tau^{(m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p}M_{2}(\tau,v-\tilde{v})\big(M_{2}(\tau,v)^{p-1}+M_{2}(\tau,\tilde{v})^{p-1}\big)d\tau \displaystyle\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{p-1}\bigr). (4.97) In a similar manner, \displaystyle\|\partial_{t}\big(H_{1}(v)(t,\cdot)-H_{1}(\tilde{v})(t,\cdot)\big)\|_{L^{2}} \displaystyle\lesssim t^{m-(m+1)(1+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{p-1}\bigr). (4.98) By ( 4.4 ), ( 4.85 ) and ( 4.87 ), we get \displaystyle\|H_{1}(v)(t,\cdot)-H_{1}(\tilde{v})(t,\cdot)\|_{\dot{H}^{\sigma}}\lesssim t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\int_{1}^{t}\tau^{-\frac{\sqrt{\delta_{1}}-\mu_{1}-1}{2}}\||v(\tau,\cdot)|^{p}-|\tilde{v}(\tau,\cdot)|^{p}\|_{L^{1}}d\tau \displaystyle\hskip 9.24994pt\hskip 9.24994pt+t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\int_{1}^{t}\tau^{-\frac{\sqrt{\delta_{1}}-\mu_{1}-1}{2}+(m+1)(\frac{n}{2}+\sigma-1)}\||v(\tau,\cdot)|^{p}-|\tilde{v}(\tau,\cdot)|^{p}\|_{\dot{H}^{\sigma-1}}d\tau \displaystyle\lesssim t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\int_{1}^{t}\tau^{(m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p}M_{2}(\tau,v-\tilde{v})\big(M_{2}(\tau,v)^{p-1} \displaystyle\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt+M_{2}(\tau,\tilde{v})^{p-1}\big)d\tau \displaystyle\lesssim t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{p-1}\bigr). (4.99) If \sigma>2 , using ( 4.5 ), ( 4.85 ), ( 4.87 )-( 4.88 ) yields \displaystyle\|\partial_{t}(H_{1}(v)(t,\cdot)-H_{1}(\tilde{v})(t,\cdot))\|_{\dot{H}^{\sigma-1}} \displaystyle\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\int_{1}^{t}\tau^{-\frac{\sqrt{\delta_{1}}-\mu_{1}-1}{2}}\big(\||v(\tau,\cdot)|^{p}-|\tilde{v}(\tau,\cdot)|^{p}\|_{L^{1}}d\tau \displaystyle\hskip 9.24994pt+\tau^{(m+1)(\frac{n}{2}+\sigma-1)}\||v(\tau,\cdot)|^{p}-|\tilde{v}(\tau,\cdot)|^{p}\|_{\dot{H}^{\sigma-1}}+\tau^{(m+1)(\frac{n}{2}+\sigma-2)}\||v(\tau,\cdot)|^{p}-|\tilde{v}(\tau,\cdot)|^{p}\|_{\dot{H}^{\sigma-2}}\big)d\tau \displaystyle\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{p-1}\bigr), (4.100) while if 1<\sigma<2 , by ( 4.5 ), ( 4.85 )-( 4.87 ), we can also get \begin{aligned} &\|\partial_{t}(H_{1}(v)(t,\cdot)-H_{1}(\tilde{v})(t,\cdot))\|_{\dot{H}^{\sigma-1}}\\ &\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{p-1}\bigr).\end{aligned} (4.101) Hence, from ( 4.97 )-( 4.101 ), we infer that M_{1}(t,H_{1}(v)-H_{1}(\tilde{v}))\lesssim\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{p-1}\bigr). (4.102) By ( 4.4 ), ( 4.89 )-( 4.90 ) and the integrability condition ( 4.59 ), we get \displaystyle\|H_{2}(u)(t,\cdot)-H_{2}(\tilde{u})(t,\cdot)\|_{L^{2}} \displaystyle\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\int_{1}^{t}\tau^{-\frac{\sqrt{\delta_{2}}-\mu_{2}-1}{2}}\big(\||u(\tau,\cdot)|^{q}-|\tilde{u}(\tau,\cdot)|^{q}\|_{L^{1}}d\tau \displaystyle\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt+\tau^{(m+1)\frac{n}{2}}\||u(\tau,\cdot)|^{q}-|\tilde{u}(\tau,\cdot)|^{q}\|_{L^{2}}\big)d\tau \displaystyle\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\int_{1}^{t}\tau^{(m+1)n+\beta_{2}+(-(m+1)n-\beta_{1}+1)q}M_{1}(\tau,u-\tilde{u})\big(M_{1}(\tau,u)^{q-1}+M_{1}(\tau,\tilde{u})^{q-1}\big)d\tau \displaystyle\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{q-1}\bigr). (4.103) Following the same procedure, we arrive at the following estimates \begin{aligned} &\|\partial_{t}\big(H_{2}(u)(t,\cdot)-H_{2}(\tilde{u})(t,\cdot)\big)\|_{L^{2}}\lesssim t^{m-(m+1)(1+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{q-1}\bigr),\\ &\|H_{2}(u)(t,\cdot)-H_{2}(\tilde{u})(t,\cdot)\|_{\dot{H}^{\sigma}}\lesssim t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{q-1}\bigr),\\ &\|\partial_{t}\big(H_{2}(u)(t,\cdot)-H_{2}(\tilde{u})(t,\cdot)\big)\|_{\dot{H}^{\sigma-1}}\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{q-1}\bigr).\end{aligned} Then we derive M_{2}(t,H_{2}(u)-H_{2}(\tilde{u}))\lesssim\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{q-1}\bigr). (4.104) Consequently, combining ( 4.102 ) and ( 4.104 ) gives \begin{aligned} &\|\mathcal{N}(u,v)-\mathcal{N}(\tilde{u},\tilde{v})\|_{\mathcal{X}(T)}=\big\|\big(H_{1}(v)(t,\cdot)-H_{1}(\tilde{v})(t,\cdot),H_{2}(u)(t,\cdot)-H_{2}(\tilde{u})(t,\cdot)\big)\big\|_{\mathcal{X}(T)}\\ &\lesssim\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(T)}\bigl(\|(u,v)\|_{\mathcal{X}(T)}^{p-1}+\|(u,v)\|_{\mathcal{X}(T)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(T)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(T)}^{q-1}\bigr).\end{aligned} Case 2: \delta_{1}>(m+1)^{2}(n+2\sigma-1)^{2} , \delta_{2}=(m+1)^{2}(n+2\sigma-1)^{2} , \sigma>1 . In this case, \mathcal{l}_{1}(\tau)=1 and \mathcal{l}_{2}(\tau)=(1+\log\tau)^{\frac{1}{2}} . By Proposition 4.2 and Lemma 4.4 , we get \begin{aligned} &\|H_{1}(v)(t,\cdot)-H_{1}(\tilde{v})(t,\cdot)\|_{L^{2}}\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\int_{1}^{t}\tau^{(m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p}\\ &\hskip 9.24994pt\times(1+\log\tau)^{\frac{p\theta_{1}(2p)}{2}}M_{2}(\tau,v-\tilde{v})\big(M_{2}(\tau,v)^{p-1}+M_{2}(\tau,\tilde{v})^{p-1}\big)d\tau\\ &\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{p-1}\bigr),\end{aligned} here we used \int_{1}^{t}\tau^{(m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p}(1+\log\tau)^{\frac{p\theta_{1}(2p)}{2}}<\infty , one can see ( 4.70 ). With the same method, we conclude \begin{aligned} &\|\partial_{t}\big(H_{1}(v)(t,\cdot)-H_{1}(\tilde{v})(t,\cdot)\big)\|_{L^{2}}\lesssim t^{m-(m+1)(1+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{p-1}\bigr),\\ &\|H_{1}(v)(t,\cdot)-H_{1}(\tilde{v})(t,\cdot)\|_{\dot{H}^{\sigma}}\lesssim t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{p-1}\bigr),\\ &\|\partial_{t}(H_{1}(v)(t,\cdot)-H_{1}(\tilde{v})(t,\cdot))\|_{\dot{H}^{\sigma-1}}\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{p-1}\bigr).\end{aligned} Thus M_{1}(t,H_{1}(v)-H_{1}(\tilde{v}))\lesssim\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{p-1}\bigr). (4.105) Note that \frac{\sqrt{\delta_{2}}}{2(m+1)}+\frac{1}{2}-\frac{n}{2}=\sigma>1 , by Proposition 4.2 , Lemma 4.4 and the integrability condition ( 4.59 ), we have \displaystyle\|H_{2}(u)(t,\cdot)-H_{2}(\tilde{u})(t,\cdot)\|_{L^{2}}\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}} \displaystyle\hskip 9.24994pt\times\int_{1}^{t}\tau^{(m+1)n+\beta_{2}+(-(m+1)n-\beta_{1}+1)q}M_{1}(\tau,u-\tilde{u})\big(M_{1}(\tau,u)^{q-1}+M_{1}(\tau,\tilde{u})^{q-1}\big)d\tau \displaystyle\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{q-1}\bigr), (4.106) \displaystyle\|\partial_{t}\big(H_{2}(u)(t,\cdot)-H_{2}(\tilde{u})(t,\cdot)\big)\|_{L^{2}} \displaystyle\lesssim t^{m-(m+1)(1+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{q-1}\bigr), (4.107) \displaystyle\|H_{2}(u)(t,\cdot)-H_{2}(\tilde{u})(t,\cdot)\|_{\dot{H}^{\sigma}}\lesssim t^{-\frac{\mu_{2}+m}{2}}\int_{1}^{t}\tau^{(m+1)n+\beta_{2}+(-(m+1)n-\beta_{1}+1)q}(1+\log\frac{t}{\tau})^{\frac{1}{2}} \displaystyle\hskip 9.24994pt\hskip 9.24994pt\hskip 9.24994pt\times M_{1}(\tau,u-\tilde{u})\big(M_{1}(\tau,u)^{q-1}+M_{1}(\tau,\tilde{u})^{q-1}\big)d\tau \displaystyle\lesssim t^{-\frac{\mu_{2}+m}{2}}(1+\log t)^{\frac{1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{q-1}\bigr) \displaystyle=t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}(1+\log t)^{\frac{1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{q-1}\bigr) (4.108) and \displaystyle\|\partial_{t}\big(H_{2}(u)(t,\cdot)-H_{2}(\tilde{u})(t,\cdot)\big)\|_{\dot{H}^{\sigma-1}} \displaystyle\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}(1+\log t)^{\frac{1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{q-1}\bigr). (4.109) Combining ( 4.106 )-( 4.109 ) yields M_{2}(t,H_{2}(u)-H_{2}(\tilde{u}))\lesssim\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\bigl(\|(u,v)\|_{\mathcal{X}(t)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{q-1}\bigr). (4.110) Therefore, it follows from ( 4.105 ) and ( 4.110 ) that \displaystyle\|\mathcal{N}(u,v)-\mathcal{N}(\tilde{u},\tilde{v})\|_{\mathcal{X}(T)} \displaystyle\lesssim\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(T)}\bigl(\|(u,v)\|_{\mathcal{X}(T)}^{p-1}+\|(u,v)\|_{\mathcal{X}(T)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(T)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(T)}^{q-1}\bigr). For the remaining cases Case 3- Case 8 listed in ( 4.23 )- ( 4.28 ), we can proceed in a similar way and will not provide further details. ∎ 4.3. The proof of Theorem 2.3 In this section, we aim to prove the existence of global solutions for p\leq\tilde{p} and q>\tilde{q} , corresponding to Theorem 2.3 . The proof relies on Propositions 4.1 - 4.2 , Lemma 4.2 and Lemma 4.4 , so we will not repeatedly highlight these elements in the discussion. Based on the integrability analysis in Section 4.2 , specifically ( 4.51 ) and ( 4.70 ), it is clear that for different values of \delta_{1},\delta_{2} , the presence of (1+\log\tau)^{\frac{1}{2}} does not affect the integrability. However, when p\leq\tilde{p} , the situation changes, which is why we exclude the case of \delta_{2}=(m+1)^{2}(n+2\sigma-1)^{2} . We clarify the specific reasons for this at the end of this section, i.e., Remark 4.4 . By the definition ( 4.12 ) of \mathcal{X}(T) , we see \|(u,v)\|_{\mathcal{X}(T)}=\sup\limits_{t\in[1,T]}\big(t^{-\alpha_{1}}M_{1}(t,u)+M_{2}(t,v)\big), (4.111) where \alpha_{1} is given by ( 4.13 ), i.e., \alpha_{1}=\left\{\begin{aligned} &\big((m+1)n+\beta_{2}-1\big)(\tilde{p}-p),&\quad&\text{if}\ \ p<\tilde{p},\\ &\epsilon,&\quad&\text{if}\ \ p=\tilde{p}.\end{aligned}\right. As for the operator \mathcal{N} defined by ( 4.15 ), our goal remains to establish Proposition 4.4 . Under the conditions of Theorem 2.3 , there exists a constant C>0 such that for any T>1 and any (u,v),(\tilde{u},\tilde{v})\in\mathcal{X}(T) , \displaystyle\|\mathcal{N}(u,v)\|_{\mathcal{X}(T)}\leq C(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}})+C(\|(u,v)\|_{\mathcal{X}{(T)}}^{p}+\|(u,v)\|_{\mathcal{X}{(T)}}^{q}), (4.112) \displaystyle\|\mathcal{N}(u,v)-\mathcal{N}(\tilde{u},\tilde{v})\|_{\mathcal{X}(T)}\leq C\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(T)} \displaystyle\quad\quad\quad\times\bigl(\|(u,v)\|_{\mathcal{X}(T)}^{p-1}+\|(u,v)\|_{\mathcal{X}(T)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(T)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(T)}^{q-1}\bigr). (4.113) After establishing Proposition 4.4 , Theorem 2.3 can be immediately obtained by following the proof method of Theorem 2.2 in Section 4.2 . Therefore, we omit the details here. It is worth mentioning that, since the proof of Proposition 4.4 is quite similar to that of Proposition 4.3 , we will focus on outlining the main differences. Thus, we only present the details on Case 1: \delta_{1},\delta_{2}>(m+1)^{2}(n+2\sigma-1)^{2},\sigma>1 . The other three cases \displaystyle\text{Case 2:}\ \delta_{1}=\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \delta_{2}>\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \sigma>1, \displaystyle\text{Case 3:}\ \delta_{1}>\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \delta_{2}>\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \sigma=1, \displaystyle\text{Case 4:}\ \delta_{1}=\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \delta_{2}>\bigl((m+1)(n+2\sigma-1)\bigr)^{2},\ \sigma=1 can be addressed analogously. Proof of Proposition 4.4 . The estimates for u^{l}(t,\cdot),v^{l}(t,\cdot) are exactly the same as ( 4.20 ) and ( 4.21 ), then \displaystyle\|(u^{l},v^{l})\|_{\mathcal{X}(T)} \displaystyle\lesssim\sup\limits_{t\in[1,T]}\big(t^{-\alpha_{1}}M_{1}(t,u^{l})+M_{2}(t,v^{l})\big) (4.114) \displaystyle\lesssim\sup\limits_{t\in[1,T]}\big(t^{-\alpha_{1}}\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\big) \displaystyle\lesssim\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}. In view of p\leq\tilde{p} , we get (m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p\geq-1. (4.115) By ( 4.111 ), M_{2}(\tau,v)\lesssim\|(u,v)\|_{\mathcal{X}(t)}(1\leq\tau0 sufficiently small such that ( 4.122 ) holds. Based on the analysis above, we conclude that if p\leq\tilde{p} , then \begin{aligned} \int_{1}^{t}\tau^{(m+1)n+\beta_{2}+(-(m+1)n-\beta_{1}+1+\alpha_{1})q}d\tau\leq\int_{1}^{\infty}\tau^{(m+1)n+\beta_{2}+(-(m+1)n-\beta_{1}+1+\alpha_{1})q}d\tau<\infty.\end{aligned} (4.123) Due to M_{1}(\tau,u)\lesssim\tau^{\alpha_{1}}\|(u,v)\|_{\mathcal{X}(t)}(1\leq\tau\leq t) , we get \displaystyle\|H_{2}(u)(t,\cdot)\|_{L^{2}}\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\int_{1}^{t}\tau^{(m+1)n+\beta_{2}+(-(m+1)n-\beta_{1}+1)q}M_{1}(\tau,u)^{q}d\tau (4.124) \displaystyle\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{q}\int_{1}^{t}\tau^{(m+1)n+\beta_{2}+(-(m+1)n-\beta_{1}+1+\alpha_{1})q}d\tau \displaystyle\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{q}. Analogously, we derive \displaystyle\|\partial_{t}H_{2}(u)(t,\cdot)\|_{L^{2}}\lesssim t^{m-(m+1)(1+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{q}, (4.125) \displaystyle\|H_{2}(u)(t,\cdot)\|_{\dot{H}^{\sigma}}\lesssim t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{q}, (4.126) \displaystyle\|\partial_{t}H_{2}(u)(t,\cdot)\|_{\dot{H}^{\sigma-1}}\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{q}. (4.127) As a result, ( 4.124 )-( 4.127 ) give M_{2}(t,H_{2}(u))\lesssim\|(u,v)\|_{\mathcal{X}(t)}^{p}. (4.128) Combining ( 4.121 ) and ( 4.128 ) yields \|\big(H_{1}(v),H_{2}(u)\big)\|_{\mathcal{X}(T)}\lesssim\|(u,v)\|_{\mathcal{X}(T)}^{p}+\|(u,v)\|_{\mathcal{X}(T)}^{q}, (4.129) together with ( 4.114 ), we get the validness of ( 4.112 ). In view of the definition of the norm of \mathcal{X}(t) , we have M_{2}(\tau,v)\lesssim\|(u,v)\|_{\mathcal{X}(t)} , M_{2}(\tau,\tilde{v})\lesssim\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)} and M_{2}(\tau,v-\tilde{v})\lesssim\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)} . Following the analysis of ( 4.116 ) and ( 4.117 ), we obtain \begin{aligned} &\|H_{1}(v)(t,\cdot)-H_{1}(\tilde{v})(t,\cdot)\|_{L^{2}}\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\\ &\quad\times\int_{1}^{t}\tau^{(m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p}M_{2}(\tau,v-\tilde{v})\big(M_{2}(\tau,v)^{p-1}+M_{2}(\tau,\tilde{v})^{p-1}\big)d\tau\\ &\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}+\alpha_{1}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\big(\|(u,v)\|_{\mathcal{X}(t)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{p-1}\big),\end{aligned} (4.130) \displaystyle\|\partial_{t}\big(H_{1}(v)(t,\cdot)-H_{1}(\tilde{v})(t,\cdot)\big)\|_{L^{2}} \displaystyle\lesssim t^{m-(m+1)(1+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}+\alpha_{1}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\big(\|(u,v)\|_{\mathcal{X}(t)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{p-1}\big), (4.131) \displaystyle\|H_{1}(v)(t,\cdot)-H_{1}(\tilde{v})(t,\cdot)\|_{\dot{H}^{\sigma}} \displaystyle\lesssim t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}+\alpha_{1}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\big(\|(u,v)\|_{\mathcal{X}(t)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{p-1}\big), (4.132) \displaystyle\|\partial_{t}\big(H_{1}(v)(t,\cdot)-H_{1}(\tilde{v})(t,\cdot)\big)\|_{\dot{H}^{\sigma-1}} \displaystyle\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}+\alpha_{1}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\big(\|(u,v)\|_{\mathcal{X}(t)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{p-1}\big). (4.133) Consequently, ( 4.130 )-( 4.133 ) yield M_{1}(t,H_{1}(v)-H_{1}(\tilde{v}))\lesssim t^{\alpha_{1}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\big(\|(u,v)\|_{\mathcal{X}(t)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{p-1}\big). (4.134) Note that \displaystyle M_{1}(\tau,u-\tilde{u})\lesssim\tau^{\alpha_{1}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}, \displaystyle M_{1}(\tau,u)\lesssim\tau^{\alpha_{1}}\|(u,v)\|_{\mathcal{X}(t)}, \displaystyle M_{1}(\tau,\tilde{u})\lesssim\tau^{\alpha_{1}}\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}, and by the analysis of ( 4.123 ), we have \displaystyle\|H_{2}(u)(t,\cdot)-H_{2}(\tilde{u})(t,\cdot)\|_{L^{2}}\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}-1}{2}}\int_{1}^{t}\tau^{(m+1)n+\beta_{2}+(-(m+1)n-\beta_{1}+1)q} \displaystyle\quad\quad\quad\quad\quad\quad\times M_{1}(\tau,u-\tilde{u})\big(M_{1}(\tau,u)^{q-1}+M_{1}(\tau,\tilde{u})^{q-1}\big)d\tau \displaystyle\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}-1}{2}}\int_{1}^{t}\tau^{(m+1)n+\beta_{2}+(-(m+1)n-\beta_{1}+1+\alpha_{1})q}d\tau \displaystyle\quad\quad\quad\quad\quad\quad\quad\quad\times\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\big(\|(u,v)\|_{\mathcal{X}(t)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{q-1}\big) \displaystyle\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{2}}-\mu_{2}-1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\big(\|(u,v)\|_{\mathcal{X}(t)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{q-1}\big), (4.135) \displaystyle\|\partial_{t}\big(H_{2}(u)(t,\cdot)-H_{2}(\tilde{u})(t,\cdot)\big)\|_{L^{2}} \displaystyle\lesssim t^{m-(m+1)(1+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\big(\|(u,v)\|_{\mathcal{X}(t)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{q-1}\big), (4.136) \displaystyle\|H_{2}(u)(t,\cdot)-H_{2}(\tilde{u})(t,\cdot)\|_{\dot{H}^{\sigma}} \displaystyle\lesssim t^{-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}-1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\big(\|(u,v)\|_{\mathcal{X}(t)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{q-1}\big), (4.137) \displaystyle\|\partial_{t}\big(H_{2}(u)(t,\cdot)-H_{2}(\tilde{u})(t,\cdot)\big)\|_{\dot{H}^{\sigma-1}} \displaystyle\lesssim t^{m-(m+1)(\sigma+\frac{n}{2})+\frac{\sqrt{\delta_{2}}-\mu_{2}-1}{2}}\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\big(\|(u,v)\|_{\mathcal{X}(t)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{q-1}\big). (4.138) Then ( 4.135 )-( 4.138 ) show that M_{2}(t,H_{2}(u)-H_{2}(\tilde{u}))\lesssim\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}\big(\|(u,v)\|_{\mathcal{X}(t)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{X}(t)}^{q-1}\big). (4.139) It follows from ( 4.134 ) and ( 4.139 ) that ( 4.113 ) holds. ∎ Remark 4.4 . In the case of \delta_{2}=(m+1)^{2}(n+2\sigma-1)^{2} , we see that \mathcal{l}_{2}(\tau)=(1+\log\tau)^{\frac{1}{2}} . Following the above process, we get \|H_{1}(v)(t,\cdot)\|_{L^{2}}\lesssim t^{-(m+1)\frac{n}{2}+\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}}\|(u,v)\|_{\mathcal{X}(t)}^{p}\times(I), where (I):=\int_{1}^{t}\tau^{(m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p}(1+\log\tau)^{\frac{p\theta_{1}(2p)}{2}}d\tau. (4.140) In view of (m+1)n+\beta_{1}+(-(m+1)n-\beta_{2}+1)p\geq-1 under the condition p\leq\tilde{p} , the occurrence of (1+\log\tau)^{\frac{p\theta_{1}(2p)}{2}} in (I) prevents us from handling the integral in the manner outlined in ( 4.116 ). In the estimates of \|H_{1}(v)(t,\cdot)\|_{\dot{H}^{\sigma}} , \|\partial_{t}H_{1}(v)(t,\cdot)\|_{L^{2}} , \|\partial_{t}H_{1}(v)(t,\cdot)\|_{\dot{H}^{\sigma-1}} , \|H_{1}(v)(t,\cdot)-H_{1}(\tilde{v})(t,\cdot)\|_{L^{2}} , \|H_{1}(v)(t,\cdot)-H_{1}(\tilde{v})(t,\cdot)\|_{\dot{H}^{\sigma}} , \|\partial_{t}\big(H_{1}(v)(t,\cdot)-H_{1}(\tilde{v})(t,\cdot)\big)\|_{L^{2}} , \|\partial_{t}\big(H_{1}(v)(t,\cdot)-H_{1}(\tilde{v})(t,\cdot)\big)\|_{\dot{H}^{\sigma-1}} , the factor (1+\log\tau) will still appear, which is beyond the current technical capabilities of this paper. Therefore, in Theorem 2.3 , we do not consider the case of \delta_{2}=(m+1)^{2}(n+2\sigma-1)^{2} . 4.4. The proof of Theorem 2.4 By exchanging the positions of p and q in the proof process of Theorem 2.3 , we can immediately obtain Theorem 2.4 , so we will not elaborate on the proof of Theorem 2.4 any further. 5. Global existence of solutions with low regularity of initial data In this section, we establish the global existence of the solution with low regularity of initial data, specifically Theorems 2.5 - 2.7 . The proofs are similar to those of Theorems 2.2 - 2.4 . Here, we briefly outline the differences. For specific details, one can refer to Sections 4.2 - 4.4 , according to the correspondence between the three groups of theorems as in Table 1 . Theorem 2.2 \longleftrightarrow Theorem 2.5 Theorem 2.3 \longleftrightarrow Theorem 2.6 Theorem 2.4 \longleftrightarrow Theorem 2.7 Table 1. We need to make the corresponding modifications to the function space and its norm defined in Section 4.2 . For \sigma\in(0,1) , let \displaystyle W_{1}(t,u)= \displaystyle t^{-\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}+(m+1)\frac{n}{2}}\|u(t,\cdot)\|_{L^{2}}+t^{-\frac{\sqrt{\delta_{1}}-\mu_{1}+1}{2}+(m+1)(\sigma+\frac{n}{2})}\ell_{1}^{-1}(t)\|u(t,\cdot)\|_{\dot{H}^{\sigma}}, \displaystyle W_{2}(t,v)= \displaystyle t^{-\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}+(m+1)\frac{n}{2}}\|v(t,\cdot)\|_{L^{2}}+t^{-\frac{\sqrt{\delta_{2}}-\mu_{2}+1}{2}+(m+1)(\sigma+\frac{n}{2})}\ell_{2}^{-1}(t)\|v(t,\cdot)\|_{\dot{H}^{\sigma}}, where \ell_{i}(t)(i=1,2) are shown in ( 2.26 ). Define the function space \mathcal{Y}(T):=\Big\{(u,v)\in\Big(C\bigl([1,T];H^{\sigma}\bigr)\Big)^{2}\ \ \text{such that}\ \ \text{supp}\big(u(t,\cdot),v(t,\cdot)\big)\subset B_{\phi_{m}(t)-\phi_{m}(1)+M}\Big\} equipped with the norm \|(u,v)\|_{\mathcal{Y}(T)}=\sup\limits_{t\in[1,T]}\big(t^{-\alpha_{1}}W_{1}(t,u)+t^{-\alpha_{2}}W_{2}(t,v)\big), (5.1) here M>0 , \alpha_{i},i=1,2 are the same as ( 4.13 ) and ( 4.14 ) and \phi_{m}(t) is defined by ( 2.5 ). By following the proof of Lemma 4.2 and Lemma 4.4 , we obtain Lemma 5.1 . In the case of \delta_{1},\delta_{2}\geq(m+1)^{2}(n+2\sigma-1)^{2} , p,q>1 , for any 1\leq\tau2\sigma\end{aligned}\right. (5.2) in Theorems 2.5 - 2.7 can ensure that \theta_{1}(2p),\theta_{1}(2q)\in[0,1] . Following Lemma 5.1 and the proof techniques of Propositions 4.3 - 4.4 , we can demonstrate that the following conclusions hold under the conditions of Theorems 2.5 - 2.7 , respectively. Proposition 5.1 . There exists a constant C>0 such that for any T>1 and any (u,v),(\tilde{u},\tilde{v})\in\mathcal{Y}(T) , we have \displaystyle\|\mathcal{N}(u,v)\|_{\mathcal{Y}(T)}\leq C\bigl(\|(u_{0},u_{1})\|_{D^{\sigma}}+\|(v_{0},v_{1})\|_{D^{\sigma}}\big)+C\big(\|(u,v)\|_{\mathcal{Y}{(T)}}^{p}+\|(\tilde{u},\tilde{v})\|_{\mathcal{Y}{(T)}}^{q}\bigr), (5.3) \displaystyle\|\mathcal{N}(u,v)-\mathcal{N}(\tilde{u},\tilde{v})\|_{\mathcal{Y}(T)}\leq C\|(u,v)-(\tilde{u},\tilde{v})\|_{\mathcal{Y}(T)} \displaystyle\quad\quad\quad\times\bigl(\|(u,v)\|_{\mathcal{Y}(T)}^{p-1}+\|(u,v)\|_{\mathcal{Y}(T)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{Y}(T)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{\mathcal{Y}(T)}^{q-1}\bigr), (5.4) where \mathcal{N} is given by ( 4.15 ). With proposition 5.1 , we can use the contraction mapping principle to give the global existence. Appendix The proof framework and methods for the local solutions, namely, Proposition 2.1 , are comparable to those for the global solutions, so we give a brief outline. For \sigma>1 , let \displaystyle X(T):=\Big\{(u,v)\in(C\bigl([1,T];H^{\sigma}\bigr)\cap C^{1}\bigl([1,T];H^{\sigma-1}\bigr)\Big)^{2} \displaystyle\text{such that}\ \ \text{supp}\big(u(t,\cdot),v(t,\cdot)\big)\subset B_{\phi_{m}(t)-\phi_{m}(1)+M}\Big\}, and for 0<\sigma<1 , X(T):=\Big\{(u,v)\in\Big(C\bigl([1,T];H^{\sigma}\bigr)\Big)^{2}\ \ \text{such that}\ \ \text{supp}\big(u(t,\cdot),v(t,\cdot)\big)\subset B_{\phi_{m}(t)-\phi_{m}(1)+M}\Big\} with the norm \|(u,v)\|_{X(T)}=\max\limits_{t\in[1,T]}\big(Q[u](t)+Q[v](t)\big) , where Q[w](t)=\left\{\resizebox{22450760}{}{$\begin{aligned} &\|w(t,\cdot)\|_{L^{2}}+\|w(t,\cdot)\|_{\dot{H}^{\sigma}}+\|\partial_{t}w(t,\cdot)\|_{L^{2}}+\|\partial_{t}w(t,\cdot)\|_{\dot{H}^{\sigma-1}},&\ &\sigma\geq 1,\\ &\|w(t,\cdot)\|_{L^{2}}+\|w(t,\cdot)\|_{\dot{H}^{\sigma}},&\ &\sigma\in(0,1).\end{aligned}$}\right. For T>1 , introduce the space X(T,K)=\{(u,v)\in X(T):\|(u,v)\|_{X(T)}\leq K\} and the operator \mathcal{N} as in ( 4.15 ), where and K>0 is determined later. Following the proof of Proposition 4.3 , we can prove Lemma A.1 . For any (u,v) and (\tilde{u},\tilde{v})\in X(T,K) , the following estimates hold \displaystyle\|\mathcal{N}(u,v)\|_{X(T)}\leq C_{1T}\big(\|u_{0},u_{1}\|_{D^{\sigma}}+\|v_{0},v_{1}\|_{D^{\sigma}}\big)+C_{2T}\big(\|(u,v)\|_{X(T)}^{p}+\|(u,v)\|_{X(T)}^{q}\big), (A.1) \displaystyle\|\mathcal{N}(u,v)-\mathcal{N}(\tilde{u},\tilde{v})\|_{X(T)}\leq C_{3T}\|(u,v)-(\tilde{u},\tilde{v})\|_{X(T)} \displaystyle\quad\quad\quad\quad\times\bigl(\|(u,v)\|_{X(T)}^{p-1}+\|(\tilde{u},\tilde{v})\|_{X(T)}^{p-1}+\|(u,v)\|_{X(T)}^{q-1}+\|(\tilde{u},\tilde{v})\|_{X(T)}^{q-1}\bigr), (A.2) where C_{1T} is bounded, and C_{2T},C_{3T}\rightarrow 0 , as T\rightarrow 1^{+} . Proof of Proposition 2.1 . By ( A.1 ) and ( A.2 ), for sufficiently large K , we choose T sufficiently close to 1^{+} such that \mathcal{N} maps X(T,K) into itself and \mathcal{N} is a contraction mapping. Thus, we establish the local existence and uniqueness of the solution to ( 1.1 ) in X(T) . ∎