Regularity of large solutions for the compressible magnetohydrodynamic equations

  • Xin Liu1Email author,

    Affiliated with

    • Yuming Qin2 and

      Affiliated with

      • Xiaozhen Peng3

        Affiliated with

        Boundary Value Problems20112011:30

        DOI: 10.1186/1687-2770-2011-30

        Received: 10 June 2011

        Accepted: 10 October 2011

        Published: 10 October 2011

        Abstract

        In this paper, we consider the initial-boundary value problem of one-dimensional compressible magnetohydrodynamics flows. The existence and continuous dependence of global solutions in H1 have been established in Chen and Wang (Z Angew Math Phys 54, 608-632, 2003). We will obtain the regularity of global solutions under certain assumptions on the initial data by deriving some new a priori estimates.

        Keywords

        magnetohydrodynamics (MHD) global solutions regularity initial-boundary value problem

        1 Introduction

        Magnetohydrodynamics (MHD) is concerned with the flow of electrically conducting fluids in the presence of magnetic fields, either externally applied or generated within the fluid by inductive action. The application of magnetohydrodynamics covers a very wide range of physical areas from liquid metals to cosmic plasmas, for example, the intensely heated and ionized fluids in an electromagnetic field in astrophysics, geophysics, high-speed aerodynamics, and plasma physics. There is a complex interaction between the magnetic and fluid dynamic phenomena, and both hydrodynamic and electrodynamic effects have to be considered. For convenience, we consider the following plane magnetohydrodynamic equations in the Lagrangian coordinate system:
        v t - u y = 0 , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ1_HTML.gif
        (1.1)
        u t + ( p + 1 2 b 2 ) y = λ u y v y , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ2_HTML.gif
        (1.2)
        w t - b y = μ w y v y , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ3_HTML.gif
        (1.3)
        ( v b ) t - w y = ν b y v y , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ4_HTML.gif
        (1.4)
        E t + u ( p + 1 2 b 2 ) - w b y = λ u u y + μ w w y + ν b b y + κ θ y v y . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ5_HTML.gif
        (1.5)

        Here, v, u, w , b , θ, and p are the specific volume, the longitudinal velocity, the transverse velocity, the transverse magnetic filed, the absolute temperature, and the pressure, respectively; λ, μ, ν, and κ are the bulk viscosity coefficient, the shear viscosity coefficient, the magnetic diffusivity, and the heat conductivity, respectively.

        We consider problem (1.1)-(1.5) in the region {y ∈ Ω: = (0, 1), t ≥ 0} under the initial-boundary conditions
        ( v , u , w , b , θ ) | t = 0 = ( v 0 , u 0 , w 0 , b 0 , θ ) ( y ) , y Ω , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ6_HTML.gif
        (1.6)
        ( u , w , b , θ y ) | Ω = 0 . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ7_HTML.gif
        (1.7)
        In this paper, we focus on an initial-boundary problem for the magnetohydrodynamic flows of a perfect gas with following equations of state:
        p = R θ v , e = c v θ , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equa_HTML.gif
        where R is the gas constant and c v is the heat capacity of the gas at constant volume. For concreteness, we assume that λ, μ, and ν are constants, and κ depends on the temperature θ with C1κ(θ)/(1 + θ r ) ≤ C2 for some positive constants C1, C2 and, r ≥ 2. The growth condition assumed on κ is motivated by the physical fact: κθ5/2 for important physical regimes (see [1, 2]). The total energy of the magnetohydrodynamics flows is
        E = e + 1 2 ( u 2 + w 2 ) + 1 2 v b 2 . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equb_HTML.gif
        Before showing our main results, let us first recall the related results in the literature. For the one-dimensional ideal gas, i.e.,
        e = c v θ , σ = R θ v + μ u y v , Q = - κ θ y v , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ8_HTML.gif
        (1.8)
        with suitable positive constants c v , R. Kazhikhov and Shelukhin [35], Kawashima and Nishida [6] established the existence of global smooth solutions. Zheng and Qin [7] proved the existence of maximal attractors in H i (i = 1, 2). However, under very high temperatures and densities, constitutive relations (1.8) become inadequate. Thus, a more realistic model would be a linearly viscous gas (or Newtonian fluid)
        σ ( v , θ , u y ) = - p ( v , θ ) + μ ( v , θ ) v u y http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ9_HTML.gif
        (1.9)
        satisfying Fourier's law of heat flux
        Q ( v , θ , θ y ) = - κ ( v , θ ) v θ y http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ10_HTML.gif
        (1.10)

        whose internal energy e and pressure p are coupled by the standard thermodynamical relation (1.8). In this case, Kawohl [8] obtained the existence of global solutions with the exponents r ∈ [0, 1], q ≥ 2r + 2. Jiang [9] also established the global existence with basically same constitutive relations as those in [8] but with the exponents r ∈ [0, 1], qr + 1. When the exponents q, r satisfy the more general constitutive relations than those in [8, 9], Qin [10] established the regularity and asymptotic behavior of global solutions with arbitrary initial data for a one-dimensional viscous heat-conductive real gas.

        For the radiative and reactive gas, Ducomet [11] established the global existence and exponential decay in H1 of smooth solutions, and Umehara and Tani [12] proved the global existence of smooth solutions for a self-gravitating radiative and reactive gas.

        For the radiative magnetohydrodynamic equations with self-gravitation, Ducomet and Feireisl [13] proved the existence of global-in-time solutions of this problem with arbitrarily large initial data and conservative boundary conditions on a bounded spatial domain in ℝ3. Recently, under the technical condition that κ(ρ, θ) satisfies
        k 1 ( 1 + θ q ) κ ( ρ , θ ) k 2 ( 1 + θ q ) , k 1 ( 1 + θ q ) | κ ρ ( ρ , θ ) | k 2 ( 1 + θ q ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equc_HTML.gif

        for some q > 5 2 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_IEq1_HTML.gif, Zhang and Xie [14] investigated the existence of global smooth solutions.

        For the non-radiative and non self-gravitation magnetohydrodynamic flows, there have been a number of studies under various conditions by several authors (see, e.g., [2, 1522]). The existence and uniqueness of local smooth solutions were first obtained in [21]; moreover, the existence of global smooth solutions with small smooth initial data was shown in [20]. Chen and Wang [15] investigated a free boundary problem with general large initial data with exponents r ∈ [0, 1], q ≥ 2r + 2. Under the technical condition that κ(ρ, θ) satisfies
        C - 1 ( 1 + θ q ) κ ( ρ , θ ) C ( 1 + θ q ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equd_HTML.gif
        for q ≥ 2, Chen and Wang [16] also proved the existence and continuous dependence of global strong solutions with large initial data. Wang [22] established large solutions to the initial-boundary value problem for planar magnetohydrodynamics. Under the technical condition upon
        κ ( ρ , θ ) κ ( ρ ) > C ρ , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Eque_HTML.gif

        Fan et al. [18] investigated the uniqueness of the weak solutions of MHD with Lebesgue initial data. Fan et al. [19] also considered a one-dimensional plane compressible MHD flows and proved that as the shear viscosity goes to zero, global weak solutions converge to a solution of the original equations with zero shear viscosity. The uniqueness and continuous dependence of weak solutions for the Cauchy problem have been proved by Hoff and Tsyganov [17].

        As mentioned above, the global existence in H + i ( i =  2 , 4 ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_IEq2_HTML.gif of global solutions has never been studied for Equations (1.1)-(1.5) of the nonlinear one-dimensional compressible magnetohydrodynamics flows with initial-boundary conditions (1.6)-(1.7). The main aim of this paper is to prove the regularity of solutions in the subspace H + i http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_IEq3_HTML.gif of (H i [0, 1])7(i = 2, 4) for systems (1.1)-(1.7). In order to obtain higher regularity of global solutions, there are many complicated estimates on higher derivations of solutions to be involved, this is our main difficulty. To overcome this difficulty, we should use some proper embedding theorems, the interpolation techniques as well as many delicate estimates. This is the novelty of the paper.

        We define three spaces as follows:
        H + 1 = ( v , u , w , b , θ ) ( H 1 ( Ω ) ) 7 : v ( x ) > 0 , θ ( x ) > 0 , x Ω , u ( 0 ) = u ( 1 ) = 0 , w ( 0 ) = w ( 1 ) = b ( 0 ) = b ( 1 ) = 0 , H + i = ( v , u , w , b , θ ) ( H i ( Ω ) ) 7 : v ( x ) > 0 , θ ( x ) > 0 , x Ω , u ( 0 ) = u ( 1 ) = 0 , w ( 0 ) = w ( 1 ) = b ( 0 ) = b ( 1 ) = 0 , θ ( 0 ) = θ ( 1 ) = 0 , i = 2 , 4 . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equf_HTML.gif

        The notation in this paper will be stated as follows:

        L p , 1 ≤ p ≤ +∞, W m, p , mN, H1 = W1,2, H 0 1 = W 0 1 , 2 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_IEq4_HTML.gif denote the usual (Sobolev) spaces on Ω. In addition, ||·|| B denotes the norm in the space B, we also put = L 2 ( Ω ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_IEq5_HTML.gif. Constants C i (i = 1, 2, 3, 4) depend on the H + i http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_IEq3_HTML.gif norm of the initial data (v0, u0, w 0 , b0, θ0) and T > 0.

        Now we are in a position to state our main results.

        Theorem 1.1 Assume that the initial data ( v 0 , u 0 , w 0 , b 0 , θ 0 ) H + 2 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_IEq6_HTML.gifand e, p, and κ are C3functions. Then, the problem (1.1) -(1.7) admits a unique global solution ( v ( t ) , u ( t ) , w ( t ) , b ( t ) , θ ( t ) ) H + 2 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_IEq7_HTML.gifsuch that for any T > 0,
        v ( t ) v ¯ H 2 2 + u ( t ) H 2 2 + w ( t ) H 2 2 + b ( t ) H 2 2 + θ ( t ) θ ¯ H 2 2 + u t ( t ) 2 + w t ( t ) 2 + b t ( t ) 2 + θ t ( t ) 2 + 0 t ( v v ¯ H 2 2 + u H 3 2 + w H 3 2 + b H 3 2 + θ θ ¯ H 3 2 + u t y 2 + w t y 2 + b t y 2 + θ t y 2 ) ( s ) d s C 2 , t [ 0 , T ] , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ11_HTML.gif
        (1.11)
        where v ̄ = 0 1 v d y = 0 1 v 0 d y http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_IEq8_HTML.gif, constant θ ̄ > 0 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_IEq9_HTML.gif is determined by
        e ( v ̄ , θ ̄ ) = 0 1 1 2 ( u 0 2 + | w 0 | 2 + v 0 | b 0 | 2 ) + e ( v 0 , θ 0 ) ( y ) d y . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equg_HTML.gif
        Theorem 1.2 Assume that the initial data ( v 0 , u 0 , w 0 , b 0 , θ 0 ) H + 4 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_IEq10_HTML.gifand e, p, and κ are C5functions on 0 < v < +∞ and 0 ≤ θ < +∞. Then, the problem (1.1)-(1.7) admits a unique global solution ( v ( t ) , u ( t ) , w ( t ) , b ( t ) , θ ( t ) ) H + 4 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_IEq11_HTML.gifsuch that for any T > 0,
        v ( t ) v ¯ H 4 2 + u ( t ) H 4 2 + w ( t ) H 4 2 + b ( t ) H 4 2 + θ ( t ) θ ¯ H 4 2 + u t t ( t ) 2 + w t t ( t ) 2 + b t t ( t ) 2 + u t ( t ) H 2 2 + w t ( t ) H 2 2 + b t ( t ) H 2 2 + θ t ( t ) H 2 2 + θ t t ( t ) 2 + 0 t ( v v ¯ H 4 2 + u H 5 2 + w H 5 2 + b H 5 2 + θ θ ¯ H 5 2 + u t H 3 2 + w t H 3 2 + b t H 3 2 + θ t H 3 2 + u t t H 1 2 + w t t H 1 2 + b t t H 1 2 + θ t t H 1 2 ) ( s ) d s C 4 , t [ 0 , T ] . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ12_HTML.gif
        (1.12)

        2 Proof of Theorem 1.1

        In this section, we study the global existence of problem (1.1)-(1.7) in H + 2 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_IEq12_HTML.gif by establishing a series of priori estimates. Without loss of generality, we take c v = R = 1. We begin with the following lemma.

        Lemma 2.1 Assume that the initial data ( v 0 , u 0 , w 0 , b 0 , θ 0 ) H + 1 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_IEq13_HTML.gifand e, p, and κ are C2functions on 0 < v < +∞ and 0 ≤ θ < +∞ and there exists a positive constant C0such that
        0 < C 0 - 1 v 0 ( y ) C 0 , 0 < C 0 - 1 θ 0 ( y ) C 0 . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equh_HTML.gif
        Then, for the initial data ( v 0 , u 0 , w 0 , b 0 , θ 0 ) H + 1 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_IEq13_HTML.gif, the problem (1.1) -(1.7) admits a unique global solution ( v ( t ) , u ( t ) , w ( t ) , b ( t ) , θ ( t ) ) H + 1 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_IEq14_HTML.gifsuch that for any T > 0
        0 < C 1 - 1 v ( y , t ) C 1 , 0 < C 1 - 1 θ ( y , t ) C 1 , ( y , t ) [ 0 , 1 ] × [ 0 , T ] http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ13_HTML.gif
        (2.1)
        and for any t ∈ [0, T],
        v ( t ) v ¯ H 1 2 + u ( t ) H 1 2 + w ( t ) H 1 2 + b ( t ) H 1 2 + θ ( t ) θ ¯ H 1 2 + 0 t ( v v ¯ H 1 2 + u H 2 2 + w H 2 2 + b H 2 2 + θ θ ¯ H 2 2 + u t 2 + w t 2 + b t 2 + θ t 2 ) ( s ) d s C 1 . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ14_HTML.gif
        (2.2)

        Proof. See, e.g., [16].

        Lemma 2.2 Under the assumptions in Theorem 1.1, the following estimate holds:
        u t ( t ) 2 + w t ( t ) 2 + b t ( t ) 2 + θ t ( t ) 2 + 0 t ( u t y 2 + w t y 2 + b t y 2 + θ t y 2 ) ( s ) d s C 2 , t [ 0 , T ] . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ15_HTML.gif
        (2.3)
        Proof. Differentiating (1.2) with respect to t, multiplying the resultant by u t , and then integrating the resulting equation over Q t : = Ω × [0, t], we infer
        u t ( t ) 2 + 0 t u t y ( s ) 2 d s ε 0 t u t y ( s ) 2 d s + C 2 0 t ( θ t ( s ) 2 + b b t ( s ) 2 + u y ( s ) L 4 4 ) d s ε 0 t u t y ( s ) 2 d s + C 2 0 t ( u y ( s ) L 2 + θ t ( s ) 2 + b t ( s ) 2 ) d s C 2 + ε 0 t u t y ( s ) 2 d s , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equi_HTML.gif
        which implies
        u t ( t ) 2 + 0 t u t y ( s ) 2 d s C 2 . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ16_HTML.gif
        (2.4)
        Analogously, we have
        w t ( t ) 2 + b t ( t ) 2 + θ t ( t ) 2 + 0 t ( w t y 2 + b t y 2 + θ t y 2 ) ( s ) d s C 2 . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ17_HTML.gif
        (2.5)

        Thus, (2.3) follows from (2.4)-(2.5).

        Lemma 2.3 Under the assumptions in Theorem 1.1, the following estimate holds:
        u y y ( t ) 2 + w y y ( t ) 2 + b y y ( t ) 2 + θ y y ( t ) 2 + 0 t ( u y y y 2 + w y y y 2 + b y y y 2 + θ y y y 2 ) ( s ) d s C 2 , t [ 0 , T ] . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ18_HTML.gif
        (2.6)
        Proof. Equation (1.2) can be rewritten as
        u t = - λ θ y v + λ θ v y v - λ u y v y v 2 + λ u y y v - b b y . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ19_HTML.gif
        (2.7)
        Using equation (2.7), Lemmas 2.1-2.2, Sobolev's embedding theorem and Young's inequality, we have
        u y y ( t ) C 2 ( | | u t ( t ) | | + | | θ y ( t ) | | + | | b b y ( t ) | | + | | θ v y ( t ) | | + | | v y u y ( t ) | | ) C 2 ( | | u t ( t ) | | + | | θ y ( t ) | | + θ ( t ) L | | v y ( t ) | | + | | b y ( t ) | | 2 + | | u y ( t ) | | L | | v y ( t ) | | ) ε | | u y y ( t ) | | + C 2 ( | | u t ( t ) | | + 1 ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equj_HTML.gif
        which leads to
        u y y ( t ) C 2 , 0 t u y y y ( s ) 2 d s C 2 0 t u t y ( s ) 2 d s C 2 . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ20_HTML.gif
        (2.8)
        Similarly, we derive
        w y y ( t ) + b y y ( t ) + θ y y ( t ) C 2 ( w t ( t ) + b t ( t ) + θ t ( t ) + 1 ) C 2 , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ21_HTML.gif
        (2.9)
        0 t ( w y y y 2 + b y y y 2 + θ y y y 2 ) ( s ) d s C 2 . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ22_HTML.gif
        (2.10)

        Thus, (2.6) follows from (2.8)-(2.10).

        Lemma 2.4 Under the assumptions in Theorem 1.1, the following estimate holds:
        v y y ( t ) 2 + 0 t v y y ( s ) 2 d s C 2 , t [ 0 , T ] . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ23_HTML.gif
        (2.11)
        Proof. Differentiating (1.2) with respect to y, we obtain
        λ d d t v y y v + θ v 2 v y y = u t y + E ( y , t ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ24_HTML.gif
        (2.12)
        where
        E ( y , t ) = θ y y v + 2 v y ( λ u y y - θ y ) v 2 + 2 v y 2 ( θ - λ u y ) v 3 + b b y y + b y 2 . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equk_HTML.gif
        Multiplying (2.12) by v y y v http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_IEq15_HTML.gif, integrating the resulting equation over Q t , and then using the Young inequality and interpolation theorem, we can conclude
        v y y v ( t ) 2 + C 1 1 0 t v y y v ( s ) 2 d s 1 4 C 1 0 t v y y v ( s ) 2 d s + C 2 0 t ( u t y 2 + θ y y 2 + v y u y y 2 + v y L 4 4 + u y v y 2 2 + b y L 4 4 + b L 2 b y y 2 ) ( s ) d s 1 2 C 1 0 t v y y v ( s ) 2 d s + C 2 0 t ( v y 2 + u t y 2 + u y y L 2 + θ y y 2 + b y y 2 ) ( s ) d s , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equl_HTML.gif

        which, together with Lemmas 2.1-2.3, yields (2.11).

        Proof of Theorem 1.1. By Lemmas 2.1-2.4, we complete the proof of Theorem 1.1.

        3 Proof of Theorem 1.2

        In this section, we study the global existence of problem (1.1)-(1.7) in H + 4 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_IEq16_HTML.gif by establishing a series of priori estimates. We begin with the following lemmas.

        Lemma 3.1 Under the assumptions in Theorem 1.2, the following estimates hold:
        u t y ( y , 0 ) + w t y ( y , 0 ) + w t y ( y , 0 ) + θ t y ( y , 0 ) C 3 , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ25_HTML.gif
        (3.1)
        u t t ( y , 0 ) + w t t ( y , 0 ) + b t t ( y , 0 ) + θ t t ( y , 0 ) + u t y y ( y , 0 ) + w t y y ( y , 0 ) + b t y y ( y , 0 ) + θ t y y ( y , 0 ) C 3 . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ26_HTML.gif
        (3.2)
        Proof. We easily infer from (1.2), Lemma 2.1 and Theorems 1.1 that
        u t ( t ) C 3 ( v y ( t ) + θ y ( t ) + u y y ( t ) + u y ( t ) L v y ( t ) + b ( t ) L b y ( t ) ) C 3 ( v y ( t ) + θ y ( t ) + u y y ( t ) + b y ( t ) ) . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equm_HTML.gif
        Differentiating (1.2) with respect to y, and using Theorem 1.1, we get
        u t y ( t ) C 3 ( v y ( t ) H 1 + θ y ( t ) H 1 + u y ( t ) H 2 + b y ( t ) H 1 ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ27_HTML.gif
        (3.3)
        or
        u y y y ( t ) C 3 ( v y ( t ) H 1 + θ y ( t ) H 1 + b y ( t ) H 1 + u t y ( t ) ) . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ28_HTML.gif
        (3.4)
        Differentiating (1.2) with respect to y twice, using the embedding theorem and Theorem 1.1, we conclude
        u t y y ( t ) C 3 ( v y ( t ) H 2 + θ y ( t ) H 2 + u y ( t ) H 3 + b y ( t ) H 2 ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ29_HTML.gif
        (3.5)
        or
        u y y y y ( t ) C 3 ( v y ( t ) H 2 + θ y ( t ) H 2 + b y ( t ) H 2 + u t y y ( t ) ) . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ30_HTML.gif
        (3.6)
        Similarly, we have
        w t ( t ) C 3 ( w y ( t ) H 1 + b y ( t ) + v y ( t ) ) , w t y ( t ) C 3 ( w y ( t ) H 2 + b y ( t ) H 1 + v y ( t ) H 1 ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ31_HTML.gif
        (3.7)
        or
        w y y y ( t ) C 3 ( b y ( t ) H 1 + v y ( t ) H 1 + b t y ( t ) ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ32_HTML.gif
        (3.8)
        w t y y ( t ) C 3 ( w y ( t ) H 3 + b y ( t ) H 2 + v y ( t ) H 2 ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ33_HTML.gif
        (3.9)
        or
        w y y y y ( t ) C 3 ( b y ( t ) H 2 + v y ( t ) H 2 + w t y y ( t ) ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ34_HTML.gif
        (3.10)
        b t C 3 ( b y H 1 + w y + v y ) , b t y ( t ) C 3 ( b y ( t ) H 2 + w y ( t ) H 1 + v y ( t ) H 1 ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ35_HTML.gif
        (3.11)
        or
        b y y y ( t ) C 3 ( w y ( t ) H 1 + v y ( t ) H 1 + b t y ( t ) ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ36_HTML.gif
        (3.12)
        b t y y ( t ) C 3 ( b y ( t ) H 3 + w y ( t ) H 2 + v y ( t ) H 2 ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ37_HTML.gif
        (3.13)
        or
        b y y y y ( t ) C 3 ( w y ( t ) H 2 + v y ( t ) H 2 + b t y y ( t ) ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ38_HTML.gif
        (3.14)
        θ t ( t ) C 3 ( u y ( t ) + v y ( t ) + θ y y ( t ) + u y ( t ) L u y ( t ) + w y ( t ) L w y ( t ) + b y ( t ) L b y ( t ) + θ y ( t ) L θ y ( t ) ) , C 3 ( θ y y ( t ) + u y y ( t ) + w y y ( t ) + b y y ( t ) ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ39_HTML.gif
        (3.15)
        θ t y ( t ) C 3 ( θ t ( t ) + θ y ( t ) H 2 + v y ( t ) H 1 + u y ( t ) H 1 + w y ( t ) H 1 + b y ( t ) H 1 ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ40_HTML.gif
        (3.16)
        or
        θ y y y ( t ) C 3 ( v y ( t ) H 1 + u y ( t ) H 1 + w y ( t ) H 1 + b y ( t ) H 1 + θ t y ( t ) ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ41_HTML.gif
        (3.17)
        θ t y y ( t ) C 3 ( θ y ( t ) H 3 + v y ( t ) H 2 + u y ( t ) H 2 + w y ( t ) H 2 + b y ( t ) H 2 ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ42_HTML.gif
        (3.18)
        or
        θ y y y y ( t ) C 3 ( v y ( t ) H 2 + u y ( t ) H 2 + w y ( t ) H 2 + b y ( t ) H 2 + θ t y y ( t ) ) . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ43_HTML.gif
        (3.19)
        Differentiating (1.2) with respect to t, and using Theorem 1.1, (3.3), (3.5), (3.11)-(3.12) and (3.16), we derive
        u t t ( t ) C 3 ( v y ( t ) H 2 + u y ( t ) H 3 + b y ( t ) H 2 + θ y ( t ) H 2 ) . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ44_HTML.gif
        (3.20)
        Similarly, we can conclude
        w t t ( t ) C 3 ( v y ( t ) H 2 + b y ( t ) H 2 + w y ( t ) H 3 ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ45_HTML.gif
        (3.21)
        b t t ( t ) C 3 ( v y ( t ) H 2 + b y ( t ) H 3 + w y ( t ) H 2 ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ46_HTML.gif
        (3.22)
        θ t t ( t ) C 3 ( v y ( t ) H 2 + u y ( t ) H 2 + b y ( t ) H 2 + w y ( t ) H 2 + θ y ( t ) H 3 ) . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ47_HTML.gif
        (3.23)

        Thus, (3.1) follows from (3.3), (3.7), (3.11) and (3.16), and (3.2) from (3.5), (3.9), (3.13), (3.18) and (3.20)-(3.23).

        Lemma 3.2 Under the assumptions in Theorem 1.2, the following estimates hold, for any t ∈ [0, T],
        u t t ( t ) 2 + 0 t u t t y ( s ) 2 d s C 3 + C 3 0 t ( b t y y 2 + θ t y y 2 ) ( s ) d s , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ48_HTML.gif
        (3.24)
        w t t ( t ) 2 + b t t ( t ) 2 + 0 t ( w t t y ( s ) 2 + b t t y ( s ) 2 ) d s C 3 + C 3 0 t ( b t y y ( s ) 2 + w t y y ( s ) 2 ) d s , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ49_HTML.gif
        (3.25)
        θ t t ( t ) 2 + 0 t θ t t y ( s ) 2 d s C 3 + C 2 ε - 1 0 t θ t y y ( s ) 2 d s + C 1 ε 0 t ( u t y y 2 + u t t y 2 + w t y y 2 + w t t y 2 + b t y y 2 + b t t y 2 ) ( s ) d s . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ50_HTML.gif
        (3.26)
        Proof. Differentiating (1.2) with respect to t twice, multiplying the resulting equation by u tt , performing an integration by parts, and using Lemma 2.1, we have
        1 2 d d t 0 1 u t t 2 ( y , t ) d y λ u t t y ( t ) 2 + C 4 ( θ t t ( t ) + u y y ( t ) + θ t u y ( t ) + b b t t ( t ) + b t 2 ( t ) 2 + u t y ( t ) ) u t t y ( t ) C 1 1 u t t y ( t ) 2 + C 4 ( θ t t ( t ) 2 + b t t ( t ) 2 + u t y ( t ) 2 + u y ( t ) 2 + θ t ( t ) 2 ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ51_HTML.gif
        (3.27)
        Thus, using Theorem 1.1 and Lemma 3.1, we get
        u t t ( t ) 2 + 0 t u t t y ( s ) 2 d s C 3 + C 3 0 t ( b t y y 2 + θ t y y 2 ) ( s ) d s . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equn_HTML.gif
        Analogously, we obtain
        w t t ( t ) 2 + b t t ( t ) 2 + 0 t ( w t t y 2 + b t t y 2 ) ( s ) d s C 3 + C 3 0 t ( w t y y 2 + b t y y 2 ) ( s ) d s . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equo_HTML.gif
        Equation (1.5) can be rewritten as
        ( c v θ ) t + p u y = κ θ y v y + λ u y 2 + μ | w y | 2 + ν | b y | 2 v . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ52_HTML.gif
        (3.28)
        Differentiating (3.28) with respect to t twice, multiplying the resulting equation by θ tt in L2 0[1] and integrating by parts, we arrive at
        1 2 d d t 0 1 c v θ t t ( y , t ) d y = - 0 1 κ θ y v t t θ t t y ( y , t ) d y - 0 1 p - λ u y v u t t y θ t t ( y , t ) d y + 0 1 μ w y v w t t y + ν b y v b t t y θ t t ( y , t ) d y - 2 0 1 p t - ( λ u y v ) t u t y θ t t ( y , t ) d y + 2 0 1 ( μ w y v ) t w t y + ( ν b y v ) t b t y θ t t ( y , t ) d y + 0 1 - p + λ u y v t t u y θ t t ( y , t ) d y + 0 1 ( μ w y v ) t t w y + ( ν b y v ) t t b y θ t t ( y , t ) d y = B 1 + B 2 + B 3 + B 4 + B 5 + B 6 + B 7 . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equp_HTML.gif
        By virtue of Theorem 1.1 and Lemmas 3.1-3.2, using the embedding theorem, we deduce for any ε ∈ (0, 1),
        B 1 C 1 θ t t y ( t ) 2 + C 2 θ t y ( t ) L ( u y ( t ) 2 + θ t ( t ) 2 ) θ t t y ( t ) + C 2 ( κ v ) t t ( t ) 2 θ y ( t ) L θ t t y ( t ) 2 C 1 θ t t y ( t ) 2 + C 2 ( u y ( t ) 2 + θ t ( t ) 2 + u t y ( t ) 2 + θ t y ( t ) 2 + θ t t ( t ) 2 + θ t y y ( t ) 2 ) , B 2 ε u t t y ( t ) 2 + C 2 ε 1 θ t t ( t ) 2 , B 3 ε ( w t t y ( t ) 2 + b t t y ( t ) 2 ) + C 2 ε 1 θ t t ( t ) 2 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equq_HTML.gif
        and
        B 4 C 1 0 1 ( | θ t | + | u y | + | u t y | + | u y | 2 ) | θ t t u t y | ( y , t ) d y C 2 u t y ( t ) | | 1 2 u t y y ( t ) | | 1 2 ( θ t ( t ) + u y ( t ) + u t y ( t ) ) θ t t ( t ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equr_HTML.gif
        which implies
        0 t B 4 d s C 2 sup 0 s t θ t t ( s ) 0 t u t y ( s ) | | 2 d s 1 4 0 t u t y y ( s ) | | 2 d s 1 4 × 0 t ( u t y | | 2 + θ t | | 2 + u y | | 2 ) ( s ) d s 1 2 ε sup 0 s t θ t t ( s ) | | 2 + 0 t u t y y ( s ) | | 2 d s + C 3 ε - 3 . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equs_HTML.gif
        B 5 C 1 0 1 ( | w y | 2 + | w t y | ) | w t y | + ( | w y | 2 + | w t y | ) | w t y | | θ t t | ( y , t ) d y C 2 w t y ( t ) | | L ( w y ( t ) | | 2 + w t y ( t ) ) θ t t ( t ) + C 2 w t y ( t ) | | L ( b y ( t ) | | 2 + b t y ( t ) ) θ t t ( t ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equt_HTML.gif
        which implies
        0 t B 5 d s ε ( sup 0 s t θ t t ( s ) 2 + 0 t ( w t y y 2 + w t y y 2 ( s ) d s ) + C 3 ε 3 , B 6 C 2 u y ( t ) | | L θ t t ( t ) [ ( θ t ( t ) | | L + u y ( t ) | | L ) ( θ t ( t ) + u y ( t ) ) + θ t t ( t ) + u t y ( t ) + u y ( t ) + u t t y ( t ) ] C 2 θ t t ( t ) ( θ t ( t ) + u y ( t ) | | H 1 + θ t y ( t ) + θ t t ( t ) + u t y ( t ) + u t t y ( t ) ) ε u t t y ( t ) 2 + C 2 ε 1 ( θ t ( t ) 2 + u y ( t ) | | H 1 2 + θ t y ( t ) 2 + θ t t ( t ) 2 + u t y ( t ) 2 ) , B 7 C 2 w y ( t ) | | L θ t t ( t ) ( w t t y ( t ) + w t y ( t ) + w y ( t ) ) + C 2 b y ( t ) | | L θ t t ( t ) × ( b t t y ( t ) + b t y ( t ) + b y ( t ) ) C 2 θ t t ( t ) ( w t y ( t ) + w y ( t ) | | H 1 + w t t y ( t ) + b t y ( t ) + b y ( t ) | | H 1 + b t t y ( t ) ) ε ( w t t y ( t ) 2 + b t t y ( t ) 2 ) + C 2 ε 1 ( θ t t ( t ) 2 + w y ( t ) | | H 1 2 + w t y ( t ) 2 + b y ( t ) | | H 1 2 + b t y ( t ) 2 ) . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equu_HTML.gif
        Thus, for ε ∈ (0, 1) small enough, we derive from above estimates
        θ t t ( t ) 2 + 0 t θ t t y ( s ) 2 d s C 2 ε 1 0 t ( θ t y y ( s ) 2 + θ t t ( s ) 2 ) d s + C 3 ε 3 + C 1 ε [ sup 0 s t θ t t ( s ) 2 + 0 t ( u t y y 2 + u t t y 2 + w t y y 2 + w t t y 2 + b t y y 2 + b t t y 2 ) ( s ) d s ] . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ53_HTML.gif
        (3.29)

        Thus, taking supremum in t on the left-hand side of (3.29), picking ε ∈ (0, 1) small enough, and using (3.23), we can derive estimate (3.26).

        Lemma 3.3 Under the assumptions in Theorem 1.2, the following estimates hold, for any t ∈ [0,T],
        u t y ( t ) 2 + 0 t u t y y ( s ) 2 d s C 3 ε - 6 + C 2 ε 2 0 t ( b t y y 2 + θ t y y 2 + u t t y 2 ) ( s ) d s , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ54_HTML.gif
        (3.30)
        w t y ( t ) 2 + b t y ( t ) 2 + 0 t ( w t y y 2 + b t y y 2 ) ( s ) d s C 3 ε - 6 + C 2 ε 2 0 t ( w t y y 2 + b t t y 2 ) ( s ) d s , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ55_HTML.gif
        (3.31)
        θ t y ( t ) 2 + 0 t θ t y y ( s ) 2 d s C 3 ε - 6 + C 2 ε 2 0 t ( b t y y 2 + u t y y 2 + w t y y 2 + θ t t y 2 + θ y y y 2 θ t y 2 ) ( s ) d s . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ56_HTML.gif
        (3.32)
        Proof. Differentiating (1.2) with respect to y and t, multiplying the resulting equation by u ty , and integrating by parts, we arrive at
        1 2 d d t u t y ( t ) 2 = D 0 ( y , t ) + D 1 ( t ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ57_HTML.gif
        (3.33)
        where
        D 0 ( y , t ) = σ t y u t y | y = 0 y = 1 , D 1 ( t ) = - 0 1 σ t y u t y y d y , σ = - p + 1 2 | b | 2 - λ u y v . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equv_HTML.gif
        We use Theorem 1.1, Lemma 2.1, the interpolation inequality and Poincaré's inequality to obtain
        D 0 C 1 [ ( u y ( t ) L + θ t ( t ) L ) ( v y ( t ) L + θ y ( t ) L ) + b t ( t ) L b y ( t ) L + b t y ( t ) L b ( t ) L + θ t y ( t ) L + u y ( t ) L 2 + u y ( t ) L u y y ( t ) L + u t y ( t ) L v y ( t ) L + u t y y ( t ) L ] u t y ( t ) L C 3 ( D 01 + D 01 ) u t y ( t ) 1 2 u t y y ( t ) 1 2 , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ58_HTML.gif
        (3.34)
        where
        D 0 1 = u y ( t ) H 2 + θ t ( t ) + θ t y ( t ) + b t ( t ) + b t y ( t ) , D 0 2 = θ t y ( t ) 1 2 θ t y y ( t ) 1 2 + u t y y ( t ) 1 2 u t y y y ( t ) 1 2 + u t y y ( t ) + u t y ( t ) 1 2 u t y y ( t ) 1 2 + b t y ( t ) 1 2 b t y y ( t ) 1 2 . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equw_HTML.gif
        Using the Young inequality several times, we derive
        C 3 D 0 1 u t y ( t ) 1 2 u t y y ( t ) 1 2 ε 2 2 u t y y ( t ) 2 + C 3 ε - 2 3 ( u t y ( t ) 2 + u y ( t ) H 2 2 + θ t ( t ) 2 + θ t y ( t ) 2 + b t ( t ) 2 + b t y ( t ) 2 ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ59_HTML.gif
        (3.35)
        and
        C 3 D 0 2 u t y ( t ) 1 2 u t y y ( t ) 1 2 ε 2 2 u t y y ( t ) 2 + ε 2 ( u t y y y ( t ) 2 + b t y y ( t ) 2 + θ t y y ( t ) 2 ) + C 3 ε - 6 ( u t y ( t ) 2 + θ t y ( t ) 2 + b t y ( t ) 2 ) . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ60_HTML.gif
        (3.36)
        Thus, we infer from (3.34)-(3.36) that
        D 0 ε 2 ( u t y y y ( t ) 2 + u t y y ( t ) 2 + b t y y ( t ) 2 + θ t y y ( t ) 2 ) + C 3 ε - 6 ( u t y ( t ) 2 + θ t y ( t ) 2 + b t y ( t ) 2 + θ t ( t ) 2 + u y ( t ) H 2 2 + b t ( t ) 2 ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equx_HTML.gif
        which, together with Theorem 1.1, Lemma 2.1, and Lemmas 3.1-3.2, yields
        0 t D 0 d s ε 2 0 t ( u t y y y 2 + u t y y 2 + b t y y 2 + θ t y y 2 ) ( s ) d s + C 3 ε - 6 . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ61_HTML.gif
        (3.37)
        Similarly, by Theorem 1.1, Lemma 2.1, and Lemmas 3.1-3.2 and the embedding theorem, we have
        D 1 ( 2 C 3 ) - 1 u t y y ( t ) 2 + C 3 ( u t y ( t ) 2 + b t y ( t ) 2 + θ t ( t ) H 1 2 + u y ( t ) H 1 2 ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ62_HTML.gif
        (3.38)
        which, combined with (3.33), (3.37)-(3.38), Theorem 1.1, Lemma 2.1, and Lemmas 3.1-3.2, gives that for ε ∈ (0, 1) small enough,
        u t y ( t ) 2 + 0 t u t y y ( s ) 2 d s C 3 ε - 6 + C 2 ε 2 0 t ( b t y y 2 + θ t y y 2 + u t y y y 2 ) ( s ) d s . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ63_HTML.gif
        (3.39)
        On the other hand, differentiating (1.2) with respect to x and t, using Theorem 1.1 and Lemmas 3.1-3.2, we have
        u t y y y ( t ) C 1 u t t y ( t ) + C 2 ( u y y ( t ) H 2 2 + θ y ( t ) H 1 2 + v y ( t ) H 1 2 + b y ( t ) H 1 2 + θ t ( t ) H 2 2 + b t ( t ) H 2 2 ) . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ64_HTML.gif
        (3.40)

        Thus, inserting (3.40) into (3.39) implies estimate (3.30).

        Analogously, we can obtain estimates (3.31)-(3.32). □

        Lemma 3.4 Under the assumptions in Theorem 1.2, the following estimates hold for any t ∈ [0,T],
        u t t ( t ) 2 + u t y ( t ) 2 + w t t ( t ) 2 + w t y ( t ) 2 + b t t ( t ) 2 + b t y ( t ) 2 + θ t t ( t ) 2 + θ t y ( t ) 2 + 0 t ( u t t y 2 + u t y y 2 + w t t y 2 + w t y y 2 + b t t y 2 + b t y y 2 + θ t t y 2 + θ t y y 2 ) ( s ) d s C 4 , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ65_HTML.gif
        (3.41)
        v y y y ( t ) H 1 2 + v y y ( t ) W 1 , 2 + 0 t ( v y y y H 1 2 + v y y W 1 , 2 ) ( s ) d s C 4 , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ66_HTML.gif
        (3.42)
        u y y y ( t ) H 1 2 + u y y ( t ) W 1, 2 + w y y y ( t ) H 1 2 + w y y ( t ) W 1, 2 + b y y y ( t ) H 1 2 + b y y ( t ) W 1, 2 + θ y y y ( t ) H 1 2 + θ y y ( t ) W 1, 2 + v t y y y ( t ) 2 + u t y y ( t ) 2 + w t y y ( t ) 2 + b t y y ( t ) 2 + θ t y y ( t ) 2 + 0 t ( u t t 2 + w t t 2 + b t t 2 + θ t t 2 + u y y W 2 , 2 + w y y W 2 , 2 + b y y W 2 , 2 + θ y y W 2 , 2 + θ t y y H 1 2 + u t y y H 1 2 + w t y y H 1 2 + b t y y H 1 2 + θ t y W 1, 2 + u t y W 1, 2 + w t y W 1, 2 + b t y W 1, 2 + v t y y y H 1 2 ) ( s ) d s C 4 , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ67_HTML.gif
        (3.43)
        0 t ( u y y y y H 1 2 + w y y y y H 1 2 + b y y y y H 1 2 + θ y y y y H 1 2 ) ( s ) d s C 3 . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ68_HTML.gif
        (3.44)
        Proof. Adding up (3.30)-(3.32), picking ε ∈ (0, 1) enough small, by Lemmas 3.1-3.3, and Gronwall's inequality, we get
        u t y ( t ) 2 + w t y ( t ) 2 + b t y ( t ) 2 + θ t y ( t ) 2 + 0 t ( u t y y 2 + w t y y 2 + b t y y 2 + θ t y y 2 ) ( s ) d s C 3 ε 6 + C 2 ε 2 0 t ( u t t y 2 + w t t y 2 + b t t y 2 + θ t t y 2 + θ t y 2 θ y y y 2 ) ( s ) d s . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ69_HTML.gif
        (3.45)
        Now multiplying (3.24)-(3.26) by ε, ε, and ε 3 2 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_IEq17_HTML.gif, adding the resultant to (3.45), and choosing ε ∈ (0, 1) small enough, we obtain
        u t t ( t ) 2 + u t y ( t ) 2 + w t t ( t ) 2 + w t y ( t ) 2 + b t t ( t ) 2 + b t y ( t ) 2 + θ t t ( t ) 2 + θ t y ( t ) 2 + 0 t ( u t t y 2 + u t y y 2 + w t t y 2 + w t y y 2 + b t t y 2 + b t y y 2 + θ t t y 2 + θ t y y 2 ) ( s ) d s C 4 ε 6 + C 2 ε 2 0 t θ t y ( s ) 2 θ y y y ( s ) 2 d s , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equy_HTML.gif

        which, by Gronwall's inequality, gives the estimate (3.41).

        Differentiating (2.12) with respect to y, and using v tyy = u yyy , we obtain
        λ t v y y y v + θ v 2 v y y y = E 1 ( y , t ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ70_HTML.gif
        (3.46)
        with
        E 1 ( y , t ) = E y ( y , t ) + u t y y + θ v 2 y v y y + λ v y v y y v t . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equz_HTML.gif
        Obviously, we can infer from Lemmas 3.1-3.3 that
        E 1 ( t ) C 2 ( u t y y ( t ) + θ y ( t ) H 2 + u y ( t ) H 2 + b y ( t ) H 2 + v y ( t ) H 1 ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ71_HTML.gif
        (3.47)
        leading to
        0 t E 1 ( s ) 2 d s C 3 . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ72_HTML.gif
        (3.48)
        Multiplying (3.46) by v y y y v http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_IEq18_HTML.gif, we get
        d d t v y y y v ( t ) 2 + C 1 - 1 v y y y v ( t ) 2 C 1 E 1 ( t ) 2 , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ73_HTML.gif
        (3.49)
        which, combined with (3.48), gives
        v y y y ( t ) 2 + 0 t v y y y ( s ) 2 d s C 3 . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2011-30/MediaObjects/13661_2011_Article_76_Equ74_HTML.gif
        (3.50)
        By (3.4), (3.6), (3.8), (3.10), (3.12), (3.14), (3.17), (3.19), (3.41), (3.50), and Lemmas 3.1-3.3, and using the embedding theorem, we have
        u y y y ( t ) 2 + u y y ( t ) L 2 + w y y y ( t ) 2 + w y y ( t ) L 2