Open Access

Convergence rate of solutions toward stationary solutions to the bipolar Navier-Stokes-Poisson equations in a half line

Boundary Value Problems20132013:124

DOI: 10.1186/1687-2770-2013-124

Received: 2 November 2012

Accepted: 26 April 2013

Published: 14 May 2013

Abstract

In this paper, we show the convergence rate of a solution toward the stationary solution to the initial boundary value problem for the one-dimensional bipolar compressible Navier-Stokes-Poisson equations. For the supersonic flow at spatial infinity, if an initial perturbation decays with the algebraic or the exponential rate in the spatial asymptotic point, the solution converges to the corresponding stationary solution with the same rate in time as time tends to infinity. For the transonic flow at spatial infinity, the solution converges to the stationary solution in time with the lower rate than that of the initial perturbation in the spatial. These results are proved by the weighted energy method.

MSC:35M31, 35Q35.

Keywords

convergence rate Navier-Stokes-Poisson equation stationary wave weighted energy method

1 Introduction

In this paper, we are concerned with the following bipolar Navier-Stokes-Poisson equations:
{ t ρ 1 + x ( ρ 1 u 1 ) = 0 , t ( ρ 1 u 1 ) + x ( ρ 1 u 1 2 + P 1 ( ρ 1 ) ) = μ 1 u 1 x x + ρ 1 E , t ρ 2 + x ( ρ 2 u 2 ) = 0 , t ( ρ 2 u 2 ) + x ( ρ 2 u 2 2 + P 2 ( ρ 2 ) ) = μ 2 u 2 x x ρ 2 E , E x = ρ 1 ρ 2 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ1_HTML.gif
(1.1)

in a one-dimensional half space R + : = ( 0 , ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq1_HTML.gif. Here the unknown functions are the densities ρ i ( i = 1 , 2 ) > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq2_HTML.gif, the velocities u i https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq3_HTML.gif ( i = 1 , 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq4_HTML.gif), and the electron field E. P i ( ρ i ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq5_HTML.gif ( i = 1 , 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq4_HTML.gif) is the pressure depending only on the density. μ i https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq6_HTML.gif ( i = 1 , 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq4_HTML.gif) is viscosity coefficient. Throughout this paper, we assume that two fluids of electrons and ions have the same equation of state P 1 ( ) = P 2 ( ) = P ( ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq7_HTML.gif with P ( ρ ) = K ρ γ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq8_HTML.gif for K > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq9_HTML.gif and γ 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq10_HTML.gif, and also they have the same viscosity coefficients μ 1 = μ 2 = 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq11_HTML.gif. The bipolar Navier-Stokes-Poisson system is used to simulate the transport of charged particles (e.g., electrons and ions). It consists of the compressible Navier-Stokes equation of two-fluid under the influence of the electro-static potential force governed by the self-consisted Poisson equation. Note that when we only consider one particle in the fluids, we also have the unipolar Navier-Stokes-Poisson equations. For more details, we can refer to [14].

Recently, some important progress was made for the compressible unipolar Navier-Stokes-Poisson system. The local and/or global existence of a renormalized weak solution for the Cauchy problem of the multi-dimensional compressible Navier-Stokes-Poisson system were proved in [57]. Chan [8] also considered the nonexistence of global weak solutions to the Navier-Stokes-Poisson equations in R N https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq12_HTML.gif. Hao and Li [9] established the global strong solutions of the initial value problem for the multi-dimensional compressible Navier-Stokes-Poisson system in a Besov space. The global existence and L 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq13_HTML.gif-decay rate of the smooth solution of the initial value problem for the compressible Navier-Stokes-Poisson system in R 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq14_HTML.gif were achieved by Li and his collaborators in [10, 11]. The pointwise estimates of the smooth solutions for the three-dimensional isentropic compressible Navier-Stokes-Poisson equation were obtained in [12]. The quasineutral limit of the compressible Navier-Stokes-Poisson system was studied in [1315]. However, the results about the bipolar Navier-Stokes-Poisson equations are very few. Lastly, Li et al. [16] showed the global existence and asymptotic behavior of smooth solutions for the initial value problem of the bipolar Navier-Stokes-Poisson equations. Duan and Yang [17] studied the unique existence and asymptotic stability of a stationary solution for the initial boundary value problem, and they showed that the large-time behavior of solutions for the bipolar Navier-Stokes-Poisson equations coincided with the one for the single Navier-Stokes system in the absence of the electric field. The consistency is also observed and proved between the bipolar Euler-Poisson system and the single damped Euler equation; for example, see [1820] and the references therein.

In this paper, we are going to discuss the initial-boundary value problem for the one-dimensional bipolar Navier-Stokes-Poisson equations. Now we give the initial condition
( ρ 1 , u 1 , ρ 2 , u 2 ) ( x , 0 ) = ( ρ 10 , u 10 , ρ 20 , u 20 ) ( x ) ( ρ + , u + , ρ + , u + ) as  x , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ2_HTML.gif
(1.2)
and the boundary date
u 1 ( 0 , t ) = u 2 ( 0 , t ) = u b < 0 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ3_HTML.gif
(1.3)
Here, we suppose inf x R + ρ i 0 ( i = 1 , 2 ) > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq15_HTML.gif and further the compatibility condition u b = u 10 ( 0 ) = u 20 ( 0 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq16_HTML.gif. Moreover, for the unique existence, we also assume
E ( + , t ) = 0 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ4_HTML.gif
(1.4)
In [17], the authors showed that the solution to (1.1)-(1.4) converges to the corresponding stationary solution of the single Navier-Stokes system in the absence of the electric field
{ t ρ + x ( ρ u ) = 0 , t ( ρ u ) + x ( ρ u + P ( ρ ) ) = u x x , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ5_HTML.gif
(1.5)
as time tends to infinity. Then, let ( ρ ˜ , u ˜ ) ( x ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq17_HTML.gif be the stationary solution to the system (1.5). We know that the stationary solution ( ρ ˜ , u ˜ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq18_HTML.gif satisfies
{ ( ρ ˜ u ˜ ) x = 0 , ( ρ ˜ u ˜ 2 + P ( ρ ˜ ) ) x = μ ρ ˜ x x , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ6_HTML.gif
(1.6)
and the boundary and spatial asymptotic conditions
u ˜ ( 0 ) = u b , lim x ( ρ ˜ , u ˜ ) = ( ρ + , u + ) , inf x R + ρ ˜ ( x ) > 0 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ7_HTML.gif
(1.7)

In this paper, we are mainly concerned with the decay rate of solutions to (1.1)-(1.4) toward the stationary solution ( ρ ˜ , u ˜ , ρ ˜ , u ˜ , 0 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq19_HTML.gif. Now we state the main result in the following theorem.

Theorem 1.1 Suppose that M + 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq20_HTML.gif and u b < u https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq21_HTML.gif hold. The initial data ( ρ 10 , u 10 , ρ 20 , u 20 , E 0 ) ( x ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq22_HTML.gifis supposed to satisfy
( ρ i 0 , u i 0 ) ( i = 1 , 2 ) H 1 ( R + ) , E 0 ( x ) L 2 ( R + ) , inf x R + ( ρ 10 , ρ 20 ) > 0 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ8_HTML.gif
(1.8)
and there exists a positive constant ε 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq23_HTML.gif such that
( ρ 10 ρ ˜ , u 10 u ˜ , ρ 20 ρ ˜ , u 20 u ˜ ) 1 + E 0 + δ < ε 0 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ9_HTML.gif
(1.9)
  1. (i)
    When M + > 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq24_HTML.gif, in addition, the initial data also satisfies ( 1 + x ) α 2 ( ρ 10 ρ ˜ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq25_HTML.gif, ( 1 + x ) α 2 ( u 10 u ˜ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq26_HTML.gif, ( 1 + x ) α 2 ( ρ 20 ρ ˜ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq27_HTML.gif, ( 1 + x ) α 2 ( u 20 u ˜ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq28_HTML.gif, ( 1 + x ) α 2 E 0 L 2 ( R + ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq29_HTML.gif for a certain positive constant α, then the solution ( ρ 1 , u 1 , ρ 2 , u 2 , E ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq30_HTML.gif to (1.1)-(1.3) satisfies the decay estimate
    ( ρ 1 ρ ˜ , u 1 u ˜ , ρ 2 ρ ˜ , u 2 u ˜ , E ) L C ( 1 + x ) α 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ10_HTML.gif
    (1.10)
     
On the other hand, if the initial data satisfies e ζ 2 x ( ρ 10 ρ ˜ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq31_HTML.gif, e ζ 2 x ( u 10 u ˜ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq32_HTML.gif, e ζ 2 x ( ρ 20 ρ ˜ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq33_HTML.gif, e ζ 2 x ( u 20 u ˜ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq34_HTML.gif, e ζ 2 x E 0 L 2 ( R + ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq35_HTML.gif for a certain positive constant ζ, then there exists a positive constant α such that the solution ( ρ 1 , u 1 , ρ 2 , u 2 , E ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq30_HTML.gif to (1.1)-(1.3) satisfies
( ρ 1 ρ ˜ , u 1 u ˜ , ρ 2 ρ ˜ , u 2 u ˜ , E ) L C e α t . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ11_HTML.gif
(1.11)
  1. (ii)
    When M + = 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq36_HTML.gif, and there exists a positive constant ε 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq23_HTML.gif such that if the initial data also satisfies ( 1 + x ) α 2 ( ρ 10 ρ ˜ , u 10 u ˜ , ρ 20 ρ ˜ , u 20 u ˜ ) 1 + ( 1 + x ) α 2 E 0 < ε 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq37_HTML.gif for a certain constant α satisfying α [ 2 , α ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq38_HTML.gif, where α https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq39_HTML.gif is a constant defined by
    α ( α 2 ) = 4 γ + 1 and α > 0 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equa_HTML.gif
     
then the solution ( ρ 1 , u 1 , ρ 2 , u 2 , E ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq30_HTML.gif to (1.1)-(1.3) satisfies
( ρ 1 ρ ˜ , u 1 u ˜ , ρ 2 ρ ˜ , u 2 u ˜ , E ) L C ( 1 + t ) α 4 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ12_HTML.gif
(1.12)

where M + https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq40_HTML.gif, u https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq41_HTML.gif and δ are defined in Section  2, and E 0 ( x ) = x ( ρ 10 ρ 20 ) ( y ) d y https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq42_HTML.gif.

Notations Throughout this paper, C > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq43_HTML.gif denotes the generic positive constant independent of time. L p ( R ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq44_HTML.gif ( 1 p < https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq45_HTML.gif) denotes the space of measurable functions with the finite norm L p = ( R | | p d x ) 1 p https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq46_HTML.gif, and L https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq47_HTML.gif is the space of bounded measurable functions on with the norm L = ess sup x | | https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq48_HTML.gif. We use https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq49_HTML.gif to denote the L 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq13_HTML.gif-norm. H k ( R ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq50_HTML.gif ( k 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq51_HTML.gif) stands for the space of L 2 ( R ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq52_HTML.gif-functions f whose derivatives (in the sense of distribution) D x l f https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq53_HTML.gif ( l k https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq54_HTML.gif) are also L 2 ( R ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq52_HTML.gif-functions with the norm k = ( l = 0 k D l 2 ) 1 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq55_HTML.gif. Moreover, C k ( [ 0 , T ] ; H l ( R ) ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq56_HTML.gif ( k , l 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq57_HTML.gif) denotes the space of the k-times continuously differentiable functions on the interval [ 0 , T ] https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq58_HTML.gif with values in H l ( R ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq59_HTML.gif.

The rest of the paper is organized as follows. In Section 2, we review the results of the stationary solution and the non-stationary solutions, then we reformulate our problem. Finally, we give the a priori estimates for the cases M + > 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq24_HTML.gif and M + = 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq36_HTML.gif in Section 3 and 4, respectively.

2 Stationary solution and global existence of non-stationary solution

In this section we mainly review the property of a stationary solution, and the unique existence and asymptotic behavior of non-stationary solutions for (1.1)-(1.3). To begin with, we recall the stationary equation
{ ( ρ ˜ u ˜ ) x = 0 , ( ρ ˜ u ˜ 2 + P ( ρ ˜ ) ) x = μ ρ ˜ x x https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ13_HTML.gif
(2.1)
with
u ˜ ( 0 ) = u b < 0 , lim x ( ρ ˜ , u ˜ ) = ( ρ + , u + ) , inf x R + ρ ˜ ( x ) > 0 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ14_HTML.gif
(2.2)
Integrating (2.1)1 over ( x , ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq60_HTML.gif yields ρ ˜ ( x ) = ρ + u + ( u ˜ ( x ) ) 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq61_HTML.gif, which implies by letting x 0 + https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq62_HTML.gif, ρ b = ρ ˜ ( 0 ) = ρ + u + ( u b ) 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq63_HTML.gif. Namely, u ˜ = u + v + v ˜ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq64_HTML.gif ( v ˜ = 1 ρ ˜ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq65_HTML.gif, v + = 1 ρ + https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq66_HTML.gif), which together with (2.1) implies
u + = v + v ( 0 ) u b < 0 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ15_HTML.gif
(2.3)
Thus, the condition u + < 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq67_HTML.gif has to be assumed whenever the outflow problem, i.e., the case u b < 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq68_HTML.gif, is consider. Moreover, let the strength of the boundary layer ( ρ ˜ , u ˜ ) ( x ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq17_HTML.gif be measured by δ : = | u + u b | https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq69_HTML.gif. Finally, we also define ( v , u ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq70_HTML.gif as follows:
u + = u + v + v ( v + v ) [ P ( 1 v ) P ( 1 v + ) ] , u = u + v + v , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ16_HTML.gif
(2.4)

and denote the Mach number at infinity M + = : | u + | P ( ρ + ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq71_HTML.gif. Then one has the following lemma.

Lemma 2.1 (see [21, 22])

Assume that the condition (2.3) holds. The boundary problem (2.1)-(2.2) has a smooth solution ( ρ ˜ , u ˜ ) ( x ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq17_HTML.gif, if and only if M + 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq20_HTML.gif and u b < u https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq21_HTML.gif. Moreover, if M + > 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq24_HTML.gif, there exist two positive constants λ and C such that the stationary solution ( ρ ˜ , u ˜ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq18_HTML.gif satisfies the estimate
| x k ( ρ ˜ ( x ) ρ + , u ˜ ( x ) u + ) | C δ e λ x for k = 0 , 1 , 2 , . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ17_HTML.gif
(2.5)
If M + = 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq36_HTML.gif, the stationary solution ( ρ ˜ , u ˜ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq18_HTML.gif satisfies
| x k ( ρ ˜ ( x ) ρ + , u ˜ u + ) | C δ k + 1 ( 1 + δ x ) k + 1 for k = 0 , 1 , 2 , . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ18_HTML.gif
(2.6)

As to the stability of the stationary solution of (1.1)-(1.4), Duan and Yang showed the following results in [17].

Lemma 2.2 (see [17])

Suppose that M + 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq20_HTML.gif and u b < u https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq21_HTML.gif hold. In addition, the initial data ( ρ 10 , u 10 , ρ 20 , u 20 , E 0 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq72_HTML.gif is supposed to satisfy
( ρ 10 ρ ˜ , u 10 u ˜ , ρ 20 ρ ˜ , u 20 u ˜ ) H 1 ( R + ) , E 0 L 2 ( R + ) , inf x R + ( ρ 10 , ρ 20 ) > 0 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equb_HTML.gif
Then there exists a positive constant ε 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq23_HTML.gif such that if
( ρ 10 ρ ˜ , u 10 u ˜ , ρ 20 ρ ˜ , u 20 u ˜ ) 1 + E 0 + δ < ε 0 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equc_HTML.gif
the initial boundary value problem (1.1)-(1.3) has a unique solution ( ρ 1 , u 1 , ρ 2 , u 2 , E ) X ( 0 , T ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq73_HTML.gif for arbitrary T > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq74_HTML.gif. Moreover, the solution ( ρ 1 , u 1 , ρ 2 , u 2 , E ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq30_HTML.gif converges to the stationary solution ( ρ ˜ , u ˜ , ρ ˜ , u ˜ , 0 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq19_HTML.gif as time tends to infinity:
lim t sup x R + | ρ 1 ρ ˜ , u 1 u ˜ , ρ 2 ρ ˜ , u 2 u ˜ , E | = 0 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equd_HTML.gif
Here the solution space X ( 0 , T ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq75_HTML.gif is defined by
X ( 0 , T ) = { ( ρ 1 , u 1 , ρ 2 , u 2 , E ) : ρ 1 ρ ˜ , u 1 u ˜ , ρ 2 ρ ˜ , u 2 u ˜ C ( 0 , T ; H 1 ) , ( ρ 1 ρ ˜ ) x , ( ρ 2 ρ ˜ ) x L 2 ( 0 , T ; L 2 ) , ( u 1 u ˜ ) x , ( u 2 u ˜ ) x L 2 ( 0 , T ; H 1 ) , E C ( 0 , T ; L 2 ) , ( u 1 u ˜ ) ( t , 0 ) = ( u 2 u ˜ ) ( t , 0 ) = 0 ( 0 t T ) } . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Eque_HTML.gif
Finally, to enclose this section, we reformulate the original problem in terms of the perturbed variables. Set ( φ 1 , ψ 1 , φ 2 , ψ 2 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq76_HTML.gif from the stationary solution as
φ i = ρ i ρ ˜ , ψ i = u i u ˜ , i = 1 , 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equf_HTML.gif
Due to (1.1) and (1.6), we have the system of equations for ( φ 1 , ψ 1 , φ 2 , ψ 2 , E ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq77_HTML.gif as
{ t φ 1 + u 1 x φ 1 + ρ 1 x ψ 1 = f 1 , ρ 1 t ψ 1 + P ( ρ 1 ) x φ 1 + ρ 1 u 1 x ψ 1 = ψ 1 x x g 1 + ρ 1 E , t φ 2 + u 2 x φ 2 + ρ 2 x ψ 2 = f 2 , ρ 2 t ψ 2 + P ( ρ 2 ) x φ 2 + ρ 2 u 2 x ψ 2 = ψ 2 x x g 2 ρ 2 E , E = x ( φ 1 φ 2 ) ( y , t ) d y , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ19_HTML.gif
(2.7)
where the nonlinear terms f i https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq78_HTML.gif ( i = 1 , 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq4_HTML.gif) and g i https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq79_HTML.gif ( i = 1 , 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq4_HTML.gif) are given by
f i = u ˜ x φ i + ρ ˜ x ψ i , g i = u ˜ x ( u ˜ φ i + ρ i ψ i ) + ρ ˜ x ( P ( ρ i ) P ( ρ ˜ ) ) . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equg_HTML.gif
The initial and boundary condition to (2.7) are derived from (1.2), (1.3) and (1.4) as follows:
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ20_HTML.gif
(2.8)
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ21_HTML.gif
(2.9)
The uniform bound of the solutions in the weighted Sobolev space is derived later in Sections 3 and 4. For this purpose, we introduce the function spaces X ω ( 0 , T ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq80_HTML.gif and X ω 1 ( 0 , T ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq81_HTML.gif defined by
X ω ( 0 , T ) = { ( φ 1 , ψ 1 , φ 2 , ψ 2 , E ) : ( ω φ 1 , ω ψ 1 , ω φ 2 , ω ψ 2 , ω E ) C ( 0 , T ; L 2 ( R + ) ) } https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equh_HTML.gif
and
X ω 1 ( 0 , T ) = { ( φ 1 , ψ 1 , φ 2 , ψ 2 , E ) : ( ω φ 1 , ω ψ 1 , ω φ 2 , ω ψ 2 ) C ( 0 , T ; H 1 ( R + ) ) ; ω E C ( 0 , T ; L 2 ( R + ) ) } . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equi_HTML.gif
Here the two types of weight functions are considered: ω ( x ) : = ( 1 + x ) α https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq82_HTML.gif, or ω ( x ) = e α x https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq83_HTML.gif. Also, we use the norms | | 2 , ω https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq84_HTML.gif, | | a , α https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq85_HTML.gif, and | | e , α https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq86_HTML.gif defined by
| f | 2 , ω : = ( 0 ω ( x ) f ( x ) 2 d x ) 1 2 , | f | a , α : = | f | 2 , ( 1 + x ) α , | f | e , α : = | f | 2 , e α x . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equj_HTML.gif

The following lemma, concerning the existence of the solution locally in time, is proved by the standard iteration method. Hence we omit the proof.

Lemma 2.3 If the initial data satisfies (1.8) and ω φ 10 , ω ψ 10 , ω φ 20 , ω ψ 20 , ω E 0 L 2 ( R + ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq87_HTML.gif, there exists a positive constant T such that the initial boundary value problem (2.7)-(2.9) has a unique solution ( φ 1 , ψ 1 , φ 2 , ψ 2 , E ) X ω ( 0 , T ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq88_HTML.gif. Moreover, if the initial data satisfies (1.8), (1.9) and ω φ 10 , ω ψ 10 , ω φ 20 , ω ψ 20 H 1 ( R + ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq89_HTML.gif and ω E 0 L 2 ( R + ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq90_HTML.gif, there exists a unique solution ( φ 1 , ψ 1 , φ 2 , ψ 2 , E ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq77_HTML.gif in X ω 1 ( 0 , T ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq81_HTML.gif.

3 A priori estimates for M + > 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq24_HTML.gif

In this section, we derive the a priori estimates of the solution ( φ 1 , ψ 1 , φ 2 , ψ 2 , E ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq77_HTML.gif for the case that M + > 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq24_HTML.gif holds in some Sobolev space. To summarize the a priori estimate, we use the following notation (see [23]) for a weight function W ( x , t ) = χ ( t ) ω ( x ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq91_HTML.gif:
N ( t ) = sup 0 τ t ( φ 1 , ψ 1 , φ 2 , ψ 2 ) ( τ ) 1 , M ( t ) 2 = 0 t χ ( τ ) ( ( φ 1 x , φ 2 x , E x ) ( τ ) 2 + ( ψ 1 x , ψ 2 x ) ( τ ) 1 2 ) d τ + 0 t χ ( τ ) ( φ 1 ( τ , 0 ) 2 + φ 2 ( τ , 0 ) 2 + E ( τ , 0 ) 2 ) d τ , L ( t ) 2 = 0 t χ t ( τ ) ( | ( φ 1 , ψ 1 , φ 2 , ψ 2 , E ) ( τ ) | 2 , ω 2 + ( φ 1 x , ψ 1 x , φ 2 x , ψ 2 x ) ( τ ) 2 ) + χ ( τ ) ( | ( ψ 1 , ψ 2 ) ( τ ) | 2 , ω x x 2 + | ( φ 1 , ψ 1 , φ 2 , ψ 2 , E ) ( τ ) | 2 , | u ˜ x | ω 2 ) d τ . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equk_HTML.gif
Proposition 3.1 Suppose that the same assumptions as in Theorem  1.1 hold.
  1. (i)
    (Algebraic decay) Suppose that ( φ 1 , ψ 1 , φ 2 , ψ 2 , E ) X ( 1 + x ) α ( 0 , T ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq92_HTML.gif is a solution to (2.7)-(2.9) for certain positive constants α and T. Then there exist positive constants ε 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq23_HTML.gif and C such that if N ( T ) + δ ε 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq93_HTML.gif, then the solution ( φ 1 , ψ 1 , φ 2 , ψ 2 , E ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq77_HTML.gif satisfies the estimate
    ( 1 + t ) α + ε ( ( φ 1 , ψ 1 , φ 2 , ψ 2 ) ( t ) 1 2 + E ( t ) 2 ) + 0 t ( 1 + τ ) α + ε ( ( φ 1 x , φ 2 x , E x ) ( τ ) 2 + ( ψ 1 x , ψ 2 x ) ( τ ) 1 2 ) d τ + 0 t ( 1 + τ ) α + ε | ( φ 1 , φ 2 , E , φ 1 x , φ 2 x ) ( τ , 0 ) | 2 d τ C ( ( φ 10 , ψ 10 , φ 20 , ψ 20 ) 1 2 + ( φ 10 , ψ 10 , φ 20 , ψ 20 , E 0 ) a , α 2 ) ( 1 + t ) ε https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ22_HTML.gif
    (3.1)
     
for arbitrary t [ 0 , T ] https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq94_HTML.gif and ε > 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq95_HTML.gif.
  1. (ii)
    (Exponential decay) Suppose that ( φ 1 , ψ 1 , φ 2 , ψ 2 , E ) X e ζ x ( 0 , T ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq96_HTML.gif is a solution to (2.7)-(2.9) for certain positive constants ζ and T. Then there exist positive constants ε 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq23_HTML.gif, C, β (<ζ) and α satisfying α β https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq97_HTML.gif such that if N ( T ) + δ ε 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq93_HTML.gif, then the solution ( φ 1 , ψ 1 , φ 2 , ψ 2 , E ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq77_HTML.gif satisfies the estimate
    https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ23_HTML.gif
    (3.2)
     

For the sake of clarity, we divide the proof of Proposition 4.1 into the following lemmas. We first derive the basic energy estimate.

Lemma 3.2 Suppose that the same assumptions as in Theorem  1.1 hold. Then there exists a positive constant ε 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq23_HTML.gif such that N ( T ) + δ < ε 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq98_HTML.gif, it holds that
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ24_HTML.gif
(3.3)
Proof From (2.7), a direct computation yields
( ρ 1 E 1 + ρ 2 E 2 + 1 2 E 2 ) t G 1 x + ψ 1 x 2 + ψ 2 x 2 = ( ψ 1 ψ 1 x + ψ 2 ψ 2 x ) x + R 1 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ25_HTML.gif
(3.4)
here
E i : = E ( ρ i , u i ) ( i = 1 , 2 ) , E ( ρ , u ) : = Φ ( ρ , ρ ˜ ) + 1 2 | u u ˜ | 2 , Φ ( ρ , ρ ˜ ) : = ρ ˜ ρ P ( s ) P ( ρ ˜ ) s 2 d s , G 1 = ρ 1 u 1 E 1 ρ 2 u 2 E 2 ( P ( ρ 1 ) P ( ρ ˜ ) ) ψ 1 ( P ( ρ 2 ) P ( ρ ˜ ) ) ψ 2 u ˜ 2 E 2 , R 1 = u ˜ x [ P ( ρ 1 ) P ( ρ ˜ ) P ( ρ ˜ ) φ 1 + ( ρ 1 u 1 ρ ˜ u ˜ ) ψ 1 + P ( ρ 2 ) P ( ρ ˜ ) P ( ρ ˜ ) φ 2 + ( ρ 2 u 2 ρ ˜ u ˜ ) ψ 2 1 2 E 2 ] 1 ρ ˜ P ( ρ ˜ ) x φ 1 ψ 1 1 ρ ˜ P ( ρ ˜ ) x φ 2 ψ 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equl_HTML.gif
Owing to Lemmas 2.1 and 2.2, we see that the energy form E ( ρ , u ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq99_HTML.gif is equivalent to | ( ρ ρ ˜ , u u ˜ ) | 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq100_HTML.gif. That is, there exist positive constants c and C such that
c ( φ i 2 + ψ i 2 ) E i C ( φ i 2 + ψ i 2 ) , i = 1 , 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ26_HTML.gif
(3.5)
We also have positive bounds of ρ i https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq101_HTML.gif ( i = 1 , 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq4_HTML.gif) as
0 < c ρ i ( i = 1 , 2 ) C , ( t , x ) [ 0 , T ] × R + . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ27_HTML.gif
(3.6)
Further, multiplying (3.4) by a weight function W ( t , x ) = χ ( t ) ω ( x ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq102_HTML.gif, we have
( W ρ 1 E 1 + W ρ 2 E 2 + 1 2 W E 2 ) t ( W G 1 ) x + W x G 1 + W ψ 1 x 2 + W ψ 2 x 2 = W t ( ρ 1 E 1 + ρ 2 E 2 + 1 2 E 2 ) + ( i = 1 2 W ψ i ψ i x 1 2 W x ψ i 2 ) x + 1 2 W x x ( ψ 1 2 + ψ 2 2 ) + W R 1 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ28_HTML.gif
(3.7)
Due to the boundary conditions (1.3) and (2.9), the integration of the second term on the left-hand side of (3.7) over R + https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq103_HTML.gif
R + ( W [ ρ 1 u 1 E 1 + ρ 2 u 2 E 2 + ( P ( ρ 1 ) P ( ρ ˜ ) ) ψ 1 + ( P ( ρ 1 ) P ( ρ ˜ ) ) ψ 2 + u ˜ 2 E 2 ] ) x d x = χ ( t ) ρ 1 ( t , 0 ) u b E 1 ( t , 0 ) χ ( t ) ρ 2 ( t , 0 ) u b E 2 ( t , 0 ) 1 2 u b χ ( t ) E ( t , 0 ) 2 C χ ( t ) ( φ 1 ( t , 0 ) 2 + φ 2 ( t , 0 ) 2 + E ( t , 0 ) 2 ) , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ29_HTML.gif
(3.8)
where we have used the estimates (3.5) and (3.6). Next, G 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq104_HTML.gif can be computed as
G 1 = G 11 + G 12 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ30_HTML.gif
(3.9)
with
G 11 = K γ ρ + γ 2 | u + | 2 ( φ 1 2 + φ 2 2 ) + ρ + | u + | 2 ( ψ 1 2 + ψ 2 2 ) K γ ρ + γ 1 ( φ 1 ψ 1 + φ 2 ψ 2 ) u + 2 E 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equm_HTML.gif
and
G 12 = i = 1 2 ( K γ ρ + u + 2 ρ i 2 ( ρ ˜ γ 1 ρ + γ 3 ρ i 2 ) φ i 2 + K ρ + u + ρ ˜ γ 1 [ Φ ( ρ ˜ ρ i ) γ 2 ( ρ ˜ ρ i 1 ) 2 ] + K γ ( ρ ˜ γ 1 ρ + γ 1 ) φ i ψ i + K ρ ˜ γ [ ( ρ i ρ ˜ ) γ 1 γ ( ρ i ρ ˜ 1 ) ] ψ i + ( ρ i u i ρ + u + ) E i ) 1 2 ( u ˜ u + ) E 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equn_HTML.gif
The conditions M + > 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq24_HTML.gif and u + < 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq67_HTML.gif yield that the quadratic form G 11 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq105_HTML.gif is positive definite since
G 11 = i = 1 2 [ ( P ( ρ + ) 3 2 2 ρ + φ i 2 + ρ + ρ + 2 ψ i 2 ) ( M + 1 ) + P ( ρ + ) 2 ρ + ( P ( ρ + ) φ i ρ + ψ i ) 2 ] 1 2 u + E 2 C ( φ 1 2 + ψ 1 2 + φ 2 2 + ψ 2 2 + E 2 ) , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equo_HTML.gif
which yields
0 t R + W x G 11 d x d τ C 0 t χ ( τ ) | ( φ 1 , ψ 1 , φ 2 , ψ 2 , E ) | 2 , ω x 2 d τ . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ31_HTML.gif
(3.10)
Using (2.5), (3.5) and the inequalities | Φ ( s ) r 2 ( s 1 ) 2 | C | s 1 | 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq106_HTML.gif, | s γ 1 γ ( s 1 ) | C | s 1 | 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq107_HTML.gif for | s 1 | 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq108_HTML.gif, we have the estimate for G 12 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq109_HTML.gif as
| G 12 | C ( N ( t ) + δ ) ( φ 1 2 + ψ 1 2 + φ 2 2 + ψ 2 2 + E 2 ) , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equp_HTML.gif
which implies
0 t R + W x G 12 d x d τ C ( N ( t ) + δ ) 0 t χ ( τ ) | ( φ 1 , ψ 1 , φ 2 , ψ 2 , E ) | 2 , ω x 2 d τ . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ32_HTML.gif
(3.11)
Moreover, the positive bound of ρ i https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq101_HTML.gif ( i = 1 , 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq4_HTML.gif), (3.6) and the Schwarz inequality yield the estimate for R 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq110_HTML.gif as
| R 1 | C | u ˜ x | ( φ 1 2 + ψ 1 2 + φ 2 2 + ψ 2 2 + E 2 ) , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equq_HTML.gif
then we have
0 t R + W R 1 d x d τ C 0 t χ ( t ) | ( φ 1 , ψ 1 , φ 2 , ψ 2 , E ) | 2 , | u ˜ x | ω 2 d τ . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ33_HTML.gif
(3.12)

Therefore, integrating (3.7) over R + × ( 0 , t ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq111_HTML.gif, substituting the above inequalities (3.8)-(3.12) into the resultant equality and then taking N ( T ) + δ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq112_HTML.gif suitably small, we obtain the desired estimate (3.3). □

Next, we obtain the estimate for the first-order derivatives of the solution for (2.7)-(2.9). As the existence of the higher-order derivatives of the solution is not supposed, we need to use the difference quotient for the rigorous derivation of the higher-order estimates. Since the argument using the difference quotient is similar to that in the paper [21, 22], we omit the details and proceed with the proof as if it verifies
( φ 1 , ψ 1 , φ 2 , ψ 2 ) C ( [ 0 , T ] ; H 2 ( R + ) ) , φ 1 x , φ 2 x L 2 ( [ 0 , T ] ; H 1 ( R + ) ) , E x L 2 ( [ 0 , T ] ; L 2 ( R + ) ) , ψ 1 x , ψ 2 x L 2 ( [ 0 , T ] ; H 2 ( R + ) ) . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equr_HTML.gif
Lemma 3.3 There exists a positive constant ε 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq23_HTML.gif such that if N ( T ) + δ < ε 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq98_HTML.gif, then
χ ( t ) ( φ 1 x , φ 2 x ) 2 + 0 t χ ( τ ) ( ( φ 1 x , φ 2 x , E x ) 2 + φ 1 x ( τ , 0 ) 2 + φ 2 x ( τ , 0 ) 2 ) d τ C ( ( φ 10 , φ 20 ) x 2 + | ( φ 10 , φ 20 , ψ 10 , ψ 20 , E 0 ) | 2 , ω 2 ) + C L ( t ) 2 + C δ M ( t ) 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ34_HTML.gif
(3.13)
Proof By differentiating the first and third equations of (2.7) in x, and then multiplying them by φ 1 x ρ 1 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq113_HTML.gif and φ 2 x ρ 2 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq114_HTML.gif, respectively, one has for i = 1 , 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq4_HTML.gif,
( φ i x 2 2 ρ i 3 ) t + ( u i φ i x 2 2 ρ i 3 ) x u ˜ x φ i x 2 ρ i 3 + ρ ˜ x ρ i 3 φ i x ψ i x + 1 ρ i 2 φ i x ψ i x x = f i x φ i x ρ i 3 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equs_HTML.gif
which yields
( φ 1 x 2 2 ρ 1 3 + φ 2 x 2 2 ρ 2 3 ) t + ( u 1 φ 1 x 2 2 ρ 1 3 + u 2 φ 2 x 2 2 ρ 2 3 ) x + 1 ρ 1 2 φ 1 x ψ 1 x x + 1 ρ 2 2 φ 2 x ψ 2 x x = R 21 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ35_HTML.gif
(3.14)
with R 21 = u ˜ x φ 1 x 2 ρ 1 3 ρ ˜ x ρ 1 3 φ 1 x ψ 1 x + u ˜ x φ 2 x 2 ρ 2 3 ρ ˜ x ρ 2 3 φ 2 x ψ 2 x f 1 x φ 1 x ρ 1 3 f 2 x φ 2 x ρ 2 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq115_HTML.gif. On the other hand, multiplying the second and fourth equations of (2.7) by φ 1 x ρ 1 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq116_HTML.gif and φ 2 x ρ 2 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq117_HTML.gif, respectively, gives
( ψ i ρ i φ i x ) t ( ψ i ρ i φ i t + ρ ˜ x ψ i 2 ρ i ) x ψ i x 2 + P ( ρ i ) ρ i 2 φ i x 2 1 ρ i u ˜ x φ i ψ i x + ρ ˜ x x ρ i ψ i 2 + 2 ρ ˜ x ρ i ψ i ψ i x + u ˜ x ( ρ ˜ x φ i ρ ˜ φ i x ) ψ i ρ i 2 = g i φ i x ρ i 2 + ( 1 ) i 1 E ρ i φ i x + φ i x ψ i x x ρ i 2 , i = 1 , 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equt_HTML.gif
Further, we have
( ψ 1 ρ 1 φ 1 x + ψ 2 ρ 2 φ 2 x ) t ( ψ 1 ρ 1 φ 1 t + ρ ˜ x ψ 1 2 ρ 1 + ψ 2 ρ 2 φ 2 t + ρ ˜ x ψ 2 2 ρ 2 ) x ψ 1 x 2 ψ 2 x 2 + P ( ρ 1 ) ρ 1 2 φ 1 x 2 + P ( ρ 2 ) ρ 2 2 φ 2 x 2 = E ρ 1 φ 1 x E ρ 2 φ 2 x + R 22 + φ 1 x ψ 1 x x ρ 1 2 + φ 2 x ψ 2 x x ρ 2 2 , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ36_HTML.gif
(3.15)
here
R 22 = 1 ρ 1 u ˜ x φ 1 ψ 1 x ρ ˜ x x ρ 1 ψ 1 2 2 ρ ˜ x ρ 1 ψ 1 ψ 1 x u ˜ x ( ρ ˜ x φ 1 ρ ˜ φ 1 x ) ψ 1 ρ 1 2 g 1 φ 1 x ρ 1 2 + 1 ρ 2 u ˜ x φ 2 ψ 2 x ρ ˜ x x ρ 2 ψ 2 2 u ˜ x ( ρ ˜ x φ 2 ρ ˜ φ 2 x ) ψ 2 ρ 2 2 g 2 φ 2 x ρ 2 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equu_HTML.gif
Combining (3.14) and (3.15), we have
( φ 1 x 2 2 ρ 1 3 + φ 1 x ψ 1 ρ 1 + φ 2 x 2 2 ρ 2 3 + φ 2 x ψ 2 ρ 2 ) t + ( u 1 φ 1 x 2 2 ρ 1 3 φ 1 t ψ 1 ρ 1 ρ ˜ x ψ 1 2 ρ 1 + u 2 φ 2 x 2 2 ρ 2 3 φ 2 t ψ 2 ρ 2 ρ ˜ x ψ 2 2 ρ 2 ) x + P ( ρ 1 ) φ 1 x 2 ρ 1 2 + P ( ρ 2 ) φ 2 x 2 ρ 2 2 = ψ 1 x 2 + ψ 2 x 2 + E ( φ 1 x ρ 1 φ 2 x ρ 2 ) + R 21 + R 22 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ37_HTML.gif
(3.16)
The second term on the right-hand side of (3.16) can be rewritten as
E ( φ 1 x ρ 1 φ 2 x ρ 2 ) = [ ( ln ρ 1 ln ρ 2 ) x ] E ( ρ 1 ρ 2 ) E ρ ˜ x ρ 1 ρ 2 = [ ( ln ρ 1 ln ρ 2 ) E ] x ( ln ρ 1 ln ρ 2 ) E x + E x E ρ ˜ x ρ 1 ρ 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equv_HTML.gif
Under the assumption (3.6) on the densities, it holds that
( ln ρ i ln ρ e ) E x C E x 2 , | E x E ρ ˜ x ρ 1 ρ 2 | C | ρ ˜ x | ( E 2 + E x 2 ) . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ38_HTML.gif
(3.17)
Moreover, owing to the Schwarz inequality with the aid of (2.5), R 21 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq118_HTML.gif is estimated as
R 21 | u ˜ x | ( φ 1 x 2 + ψ 1 x 2 + φ 2 x 2 + ψ 2 x 2 ) + | ρ ˜ x | ( φ 1 x 2 + ψ 1 x 2 + φ 2 x 2 + ψ 2 x 2 ) + | ρ ˜ x x | ( ψ 1 2 + φ 1 x 2 + ψ 2 2 + φ 2 x 2 ) + | u ˜ x x | ( φ 1 2 + φ 1 x 2 + φ 2 2 + φ 2 x 2 ) C δ ( φ 1 2 + φ 2 2 + ψ 1 2 + ψ 2 2 + φ 1 x 2 + φ 2 x 2 + ψ 1 x 2 + ψ 2 x 2 ) . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ39_HTML.gif
(3.18)
Similarly, we have
R 22 ε ( φ 1 x 2 + φ 2 x 2 ) + C ε ( ψ 1 x 2 ψ 1 2 + ψ 1 4 + ψ 2 x 2 ψ 2 2 + ψ 2 4 ) + C δ ( φ 1 2 + φ 2 2 + ψ 1 2 + ψ 2 2 + φ 1 x 2 + φ 2 x 2 + ψ 1 x 2 + ψ 2 x 2 ) . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ40_HTML.gif
(3.19)
Multiplying (3.16) by a weight function χ ( t ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq119_HTML.gif, we get
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ41_HTML.gif
(3.20)
The boundary condition (2.9) gives
0 [ χ ( t ) i = 1 2 ( u i φ i x 2 2 ρ i 3 φ i t ψ i ρ i ρ ˜ ψ i 2 ρ i ) ] x d x C χ ( t ) ( φ 1 x ( t , 0 ) 2 + φ 2 x ( t , 0 ) 2 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ42_HTML.gif
(3.21)
and
0 χ [ ( ln ρ 1 ln ρ 2 ) E ] x d x C χ ( φ 1 ( t , 0 ) 2 + φ 2 ( t , 0 ) 2 + E ( t , 0 ) 2 ) . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ43_HTML.gif
(3.22)
Integrating (3.20) over [ 0 , t ] × R + https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq120_HTML.gif, substituting (3.17), (3.18), (3.19), (3.21), (3.22) and the estimate
| φ 1 ( t , x ) | + | φ 2 ( t , x ) | + | ψ 1 ( t , x ) | + | ψ 2 ( t , x ) | + | E ( t , x ) | | φ 1 ( t , 0 ) | + | φ 2 ( t , 0 ) | + | E ( t , 0 ) | + x ( φ 1 x , φ 2 x , ψ 1 x , ψ 2 x , E x ) ( t ) , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ44_HTML.gif
(3.23)

which is proved by the similar computation as in [2124], in the resultant equality, and take ε and δ suitably small. These computations together with (3.3) give the desired estimate (3.13). □

Lemma 3.4 There exists a positive constant ε 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq23_HTML.gif such that if N ( T ) + δ < ε 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq98_HTML.gif, then
χ ( t ) ( ψ 1 x , ψ 2 x ) 2 + 0 t χ ( τ ) ( ψ 1 x x , ψ 2 x x ) 2 d τ C ( ( ψ 10 , ψ 20 ) x 2 + | ( φ 10 , φ 20 , ψ 10 , ψ 20 , E 0 ) | 2 , ω 2 ) + C L ( t ) 2 + C ( N ( t ) + δ ) M ( t ) 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ45_HTML.gif
(3.24)
Proof Multiplying (2.7)2 by ψ 1 x x ρ 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq121_HTML.gif, and (2.7)4 by ψ 2 x x ρ 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq122_HTML.gif, respectively, we have for i = 1 , 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq4_HTML.gif,
( ψ i x 2 2 ) t ( ψ i t ψ i x + u i ψ i x 2 2 ) x + ψ i x x 2 ρ i = ψ i x 3 2 u ˜ x ψ i x 2 2 + P ( ρ i ) ρ i φ i x ψ i x x + g i ψ i x x ρ i + ( 1 ) i E ψ i x x , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equw_HTML.gif
which yields
( ψ 1 x 2 2 + ψ 2 x 2 2 ) t i = 1 2 ( ψ i t ψ i x + u i ψ i x 2 2 ) x + ψ 1 x x 2 ρ 1 + ψ 2 x x 2 ρ 2 = E ( ψ 1 x x ψ 2 x x ) + R 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ46_HTML.gif
(3.25)
with
R 3 = ψ 1 x 3 2 u ˜ x ψ 1 x 2 2 + P ( ρ 1 ) ρ 1 φ 1 x ψ 1 x x ψ 2 x 3 2 u ˜ x ψ 2 x 2 2 + P ( ρ 2 ) ρ 2 φ 2 x ψ 2 x x + g 1 ψ 1 x x ρ 1 + g 2 ψ 2 x x ρ 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equx_HTML.gif
Note that E ψ 1 x x + E ψ 2 x x = ( E ( ψ 1 ψ 2 ) x ) x + E x ( ψ 1 ψ 2 ) x https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq123_HTML.gif, and the function R 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq124_HTML.gif is estimated by using (2.5) and Schwarz inequality as
R 3 ε ( ψ 1 x x 2 + ψ 2 x x 2 ) + C ε ( ψ 1 x 2 + ψ 2 x 2 + φ 1 x 2 + φ 2 x 2 + ψ 1 x 4 + ψ 2 x 4 ) + C ε | u ˜ x | ( φ 1 2 + φ 2 2 + ψ 1 2 + ψ 2 2 + ψ 1 x 2 + ψ 2 x 2 ) + C ε | ρ ˜ x | ( φ 1 2 + φ 2 2 ) , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ47_HTML.gif
(3.26)
where ε is an arbitrary positive constant and C ε https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq125_HTML.gif is a positive constant depending on ε. Then, multiplying (3.25) by a weight function χ ( t ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq119_HTML.gif, we get
i = 1 2 [ ( χ ψ i x 2 2 ) t ( χ ( ψ i t ψ i x + u i ψ i x 2 2 ) ) x + χ ψ i x x 2 ρ i ] = χ t i = 1 2 ψ i x 2 4 χ [ E ( ψ 1 x x ψ 2 x x ) R 3 ] . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ48_HTML.gif
(3.27)
Integrate (3.27) over [ 0 , t ] × R + https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq120_HTML.gif, substitute (3.26) as well as the estimate
0 ( ψ 1 x 4 + ψ 2 x 4 ) d x C ( ψ 1 x 1 2 ψ 1 x 2 + ψ 2 x 1 2 ψ 2 x 2 ) C N ( t ) ( ψ 1 x , ψ 1 x x , ψ 2 x , ψ 2 x x ) 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equy_HTML.gif
and
0 t χ | E ( ψ 1 ψ 2 ) x ( 0 , τ ) | d τ 0 t χ | E ( 0 , τ ) | 2 d τ + 0 t χ | ( ψ 1 ψ 2 ) x ( 0 , τ ) | 2 d τ 0 t χ | E ( 0 , τ ) | 2 d τ + C ε 0 t χ ( ψ 1 , ψ 2 ) x 2 d τ + ε 0 t χ ( ψ 1 , ψ 2 ) x x 2 d τ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equz_HTML.gif

in the resultant equality, and take ε suitably small. These computations together with (3.3), (3.13) and (3.23) give the desired estimate (3.24). □

Proof of Proposition 3.1 Summing up the estimates (3.3), (3.13) and (3.24) and taking N ( T ) + δ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq112_HTML.gif suitably small, we have
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ49_HTML.gif
(3.28)
First, we prove the estimate (3.1). Noting the Poincaré-type inequality (3.23), and substituting ω ( x ) = ( 1 + x ) β https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq126_HTML.gif and χ ( t ) = ( 1 + t ) ξ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq127_HTML.gif in (3.28) for β [ 0 , α ] https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq128_HTML.gif and ξ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq129_HTML.gif gives
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ50_HTML.gif
(3.29)

Therefore, applying an induction to (3.29) gives the desired estimate (3.1). Since this computation is similar those in [23, 25], we omit the details.

Next, we prove the estimate (3.2). Substitute ω ( x ) = e β x https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq130_HTML.gif and χ ( t ) = e α t https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq131_HTML.gif in (3.28) for β < λ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq132_HTML.gif to obtain
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ51_HTML.gif
(3.30)

Here, we have used the Poincaré-type inequality (3.23) again. Thus, taking δ, β and α suitably small, we obtain the desired a priori estimate (3.2). □

4 A priori estimate for M + = 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq36_HTML.gif

In the section we proceed to consider the transonic case M + = 1 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq36_HTML.gif. To state the a priori estimate of the solution precisely, here we use the notations:
N 1 ( t ) = sup 0 τ t ( 1 + x ) α / 2 ( φ 1 , ψ 1 , φ 2 , ψ 2 ) ( τ ) 1 , M 1 ( t ) 2 = 0 t ( 1 + t ) ξ ( φ 1 x , φ 2 x , E x , ψ 1 x , ψ 2 x , ψ 1 x x , ψ 2 x x ) ( τ ) a , β 2 d τ . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equaa_HTML.gif
Proposition 4.1 Suppose that the same assumption as in Theorem  1.1 holds. Suppose that ( φ 1 , ψ 1 , φ 2 , ψ 2 , E ) X ( 1 + x ) α 1 ( 0 , T ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq133_HTML.gif is a solution to (2.7)-(2.9) for certain positive constants α and T. Then there exist positive constants ε 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq23_HTML.gif and C such that if N 1 ( T ) + δ ε 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq134_HTML.gif, then the solution ( φ 1 , ψ 1 , φ 2 , ψ 2 , E ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq77_HTML.gif satisfies the estimate
( 1 + t ) α / 2 + ε ( ( φ 1 , ψ 1 , φ 2 , ψ 2 ) ( t ) 1 2 + E ( t ) 2 ) + 0 t ( 1 + τ ) α / 2 + ε | ( φ 1 , φ 2 , φ 1 x , φ 2 x , E ) ( τ , 0 ) | 2 d τ + 0 t ( 1 + τ ) α / 2 + ε ( ( φ 1 x , φ 2 x , E x ) ( τ ) 2 + ( ψ 1 x , ψ 2 x ) ( τ ) 1 2 ) d τ C ( ( φ 10 , ψ 10 , φ 20 , ψ 20 ) x 2 + ( φ 10 , ψ 10 , φ 20 , ψ 20 , E 0 ) a , α 2 ) ( 1 + t ) ε . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ52_HTML.gif
(4.1)

In order to prove Proposition 4.1, we need to get a lower estimate for u ˜ x https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq135_HTML.gif and the Mach number M ˜ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq136_HTML.gif on the stationary solution ( ρ ˜ , u ˜ ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq18_HTML.gif defined by M ˜ ( x ) : = | u ˜ ( x ) | P ( ρ ˜ ( x ) ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq137_HTML.gif.

Lemma 4.2 (see [23])

The stationary solution u ˜ ( x ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq138_HTML.gif satisfies
u ˜ x ( x ) A ( u + u b ) γ + 2 δ 2 ( 1 + B x ) 2 , A : = ( γ + 1 ) ρ + 2 , B : = δ A https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equab_HTML.gif
for x ( 0 , ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq139_HTML.gif. Moreover, there exists a positive constant C such that
γ + 1 2 | u + | δ 1 + B x C δ 2 ( 1 + B x ) 2 M ˜ ( x ) 1 C δ 1 + B x . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equac_HTML.gif

Based on Lemma 4.2, we obtain the weighted L 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq13_HTML.gif estimate of ( φ 1 , ψ 1 , φ 2 , ψ 2 , E ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq77_HTML.gif.

Lemma 4.3 There exists a positive constant ε 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq23_HTML.gif such that if N 1 ( T ) + δ < ε 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq140_HTML.gif, then
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ53_HTML.gif
(4.2)

for β [ 0 , α ] https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq128_HTML.gif and ξ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq129_HTML.gif.

Proof First, from (2.7), similar as (3.4), we also have
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ54_HTML.gif
(4.3)
Notice that from the second and fourth equations of (1.1), one has
2 E = t ( u 1 u 2 ) + 1 2 x ( u 1 2 u 2 2 ) + [ P ( ρ 1 ) x ρ 1 ρ 1 P ( ρ 2 ) x ρ 2 ρ 2 ] [ u 1 x x ρ 1 u 2 x x ρ 2 ] , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equad_HTML.gif
which implies
u ˜ x 2 E 2 = ( u ˜ x 4 ( u 1 u 2 ) E ) t u ˜ x 4 ( u 1 u 2 ) E t + u ˜ x 8 E x ( u 1 2 u 2 2 ) + A γ u ¯ x 4 ( γ 1 ) ( ρ 1 γ 1 ρ 2 γ 1 ) x E + u ˜ x 4 E [ u 1 x x ρ 1 u 2 x x ρ 2 ] = ( u ˜ x 4 ( u 1 u 2 ) E ) t + ( u ˜ u ˜ x 4 E ( ψ 1 ψ 2 ) ) x + u ˜ 8 ( ρ 1 + ρ 2 ) ( ψ 1 ψ 2 ) 2 u ˜ u ˜ x x 4 E ( ψ 1 ψ 2 ) + u ˜ x 8 E x ( ψ 1 2 ψ 2 2 ) + A γ u ¯ x 4 ( γ 1 ) ( ρ 1 γ 1 ρ 2 γ 1 ) x E + u ˜ x 4 E [ ( ψ 1 2 2 ψ 2 2 2 ) x + ( u 1 x x ρ 1 u 2 x x ρ 2 ) ] . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equae_HTML.gif
Plugging the above equality into (4.3), we arrive at
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ55_HTML.gif
(4.4)
where
R 4 = u ˜ u ˜ x x 4 E ( ψ 1 ψ 2 ) + u ˜ x 8 E x ( ψ 1 2 ψ 2 2 ) + u ˜ x 4 E ( ψ 1 ψ 1 x ψ 2 ψ 2 x ) + A γ u ˜ x 4 ( γ 1 ) ( ρ 1 γ 1 ρ 2 γ 1 ) x E + u ˜ x 4 [ u 1 x x ρ 1 u 2 x x ρ 2 ] E = : R 41 + R 42 + R 43 + R 44 + R 45 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equaf_HTML.gif
Further, multiplying (4.4) by a weight function W ( t , x ) : = ( 1 + B x ) β ( 1 + t ) ξ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq141_HTML.gif, we have
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ56_HTML.gif
(4.5)
Where G 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq142_HTML.gif and G 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq143_HTML.gif are defined
G 2 = ρ 1 u 1 E 1 ρ 2 u 2 E 2 ( P ( ρ 1 ) P ( ρ ˜ ) ) ψ 1 ( P ( ρ 2 ) P ( ρ ˜ ) ) ψ 2 u ˜ u ˜ x 4 ( ψ 1 ψ 2 ) E u ˜ 2 E 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equag_HTML.gif
and
G 3 : = W u ˜ x [ P ( ρ 1 ) P ( ρ ¯ ) P ( ρ ˜ ) φ 1 + ρ 1 ψ 1 2 ρ 1 8 ( ψ 1 ψ 2 ) 2 + P ( ρ 2 ) P ( ρ ¯ ) P ( ρ ˜ ) φ 2 + ρ 2 ψ 2 2 ρ 2 8 ( ψ 1 ψ 2 ) 2 ] 1 2 W x x ψ 1 2 1 2 W x x ψ 2 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equah_HTML.gif
By the same computation as in deriving (3.9), we rewrite the terms G 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq142_HTML.gif and G 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq143_HTML.gif to G 2 = G 21 + G 22 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq144_HTML.gif, G 3 = G 31 + G 32 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq145_HTML.gif with
G 21 = i = 1 2 [ ( P ( ρ + ) 3 / 2 2 ρ + φ i 2 + ρ + P ( ρ + ) 2 ψ i 2 ) ( M ˜ 1 ) + P ( ρ ˜ ) 2 ρ ˜ ( P ( ρ ˜ ) φ i ρ ˜ ψ i ) 2 ] u ˜ u ˜ x 4 ( ψ 1 ψ 2 ) E 1 2 u + E 2 , G 22 = i = 1 2 [ ρ ˜ P ( ρ ˜ ) u ˜ 2 ( 1 ρ i 2 1 ρ ˜ 2 ) φ i 2 + P ( ρ ˜ ) u ˜ [ Φ ( ρ ˜ ρ i ) γ 2 ( ρ ˜ ρ i 1 ) 2 ] + ( ρ i u i ρ ˜ u ˜ ) E i + ( P ( ρ i ) P ( ρ ˜ ) P ( ρ ˜ ) φ i ) ψ i [ ( P ( ρ ˜ ) 3 / 2 2 ρ ˜ P ( ρ + ) 3 / 2 2 ρ + ) φ i 2 + ( ρ ˜ P ( ρ ˜ ) 2 ρ + P ( ρ + ) 2 ) ψ i 2 ] ( M ˜ 1 ) ] 1 2 ( u ˜ u + ) E 2 , G 31 = W u ˜ x ( ρ + ψ 1 2 + 1 2 P ( ρ + ) φ 1 2 ) 1 2 W x x ψ 1 2 + W u ˜ x ( ρ + ψ 2 2 + 1 2 P ( ρ + ) φ 2 2 ) 1 2 W x x ψ 2 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equai_HTML.gif
and
G 32 = W u ˜ x i = 1 2 [ ( ρ i ρ + ) ψ i 2 + 1 2 ( P ( ρ ˜ ) P ( ρ + ) ) φ i 2 + P ( ρ i ) P ( ρ ˜ ) P ( ρ ˜ ) φ i 1 2 P ( ρ ˜ ) φ i 2 ] . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equaj_HTML.gif
By utilizing Lemma 4.2 with the aid of the fact that β < α https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq146_HTML.gif and u + < 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq67_HTML.gif, we obtain the lower estimate of W x G 21 + G 31 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq147_HTML.gif as
W x G 21 + G 31 i = 1 2 [ K γ ρ + γ 2 A 4 [ ( γ + 1 ) β + 2 ( u + u b ) γ + 2 ( γ 1 ) ] δ 2 ( 1 + t ) ξ ( 1 + B x ) β 2 φ i 2 + ρ + A 4 [ 4 ( u + u b ) γ + 2 ( γ + 1 ) β ( β 2 ) ] δ 2 ( 1 + t ) ξ ( 1 + B x ) β 2 ψ i 2 ] C β δ 3 ( 1 + t ) ξ ( 1 + B x ) β 3 ( φ 1 2 + ψ 1 2 + φ 2 2 + ψ 2 2 ) C δ 2 ( 1 C δ ) ( 1 + t ) ξ ( 1 + B x ) β 2 ( φ 1 2 + ψ 1 2 + φ 2 2 + ψ 2 2 ) + C β ( 1 + t ) ξ ( 1 + B x ) β 1 E 2 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ57_HTML.gif
(4.6)
for β ( 0 , α ] https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq148_HTML.gif. On the other hand, the estimates (2.6), (3.5) and (3.6) yield
| W x G 22 + G 32 | C ( N 1 ( t ) + δ 2 ) δ ( 1 + t ) ξ ( 1 + B x ) β 2 ( φ 1 2 + ψ 1 2 + φ 2 2 + ψ 2 2 ) + C δ β ( 1 + t ) ξ ( 1 + B x ) β 1 E 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ58_HTML.gif
(4.7)
For the first term on the right-hand side of (4.5), we estimate it as
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ59_HTML.gif
(4.8)
Similarly, we get
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ60_HTML.gif
(4.9)
In the same way, we estimate R 42 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq149_HTML.gif and R 43 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq150_HTML.gif as follows:
| 0 t R + W R 42 d x d τ | C δ 0 t ( 1 + τ ) ξ R + ( 1 + B x ) β ( ψ 1 x 2 + ψ 2 x 2 + E x 2 ) d x d τ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ61_HTML.gif
(4.10)
and
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ62_HTML.gif
(4.11)
Since ( ρ 1 γ 1 ρ 2 γ 1 ) x = ( γ 1 ) [ ( ρ 1 γ 2 ρ 2 γ 2 ) ρ ˜ x + ρ 1 γ 2 φ 1 x ρ 2 γ 2 φ 2 x ] https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq151_HTML.gif, we have
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ63_HTML.gif
(4.12)
Moreover, it is easy to compute R 45 = u ˜ x ( E ψ 1 x x ρ 1 E ψ 2 x x ρ 2 ) + u ˜ x x u ˜ x ρ 1 ρ 2 E ( φ 1 φ 2 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq152_HTML.gif, which implies
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ64_HTML.gif
(4.13)

Finally, integrate (4.5) over R + × ( 0 , t ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq111_HTML.gif, substitute (4.6)-(4.13) in the resultant equality, and take N 1 ( t ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq153_HTML.gif and δ suitably small. This procedure yields the desired estimate (4.2) for β ( 0 , α ] https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq148_HTML.gif.

Next, we prove (4.2) for β = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq154_HTML.gif. Substituting W = ( 1 + t ) ξ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq155_HTML.gif in (4.5) and integrating the resultant equality over R + × ( 0 , t ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq111_HTML.gif, we get
( 1 + t ) ξ ( φ 1 , ψ 1 , φ 2 , ψ 2 , E ) ( t ) 2 + 0 t ( 1 + τ ) ξ ( i = 1 2 φ i ( τ , 0 ) 2 + E ( τ , 0 ) 2 + ( ψ 1 x , ψ 2 x ) 2 ) d τ C ( φ 10 , ψ 10 , φ 10 , ψ 10 , E 0 ) 2 + C ξ 0 t ( 1 + τ ) ξ 1 ( φ 1 , ψ 1 , φ 2 , ψ 2 , E ) 2 d τ + C δ 0 t ( 1 + τ ) ξ ( φ 1 x , φ 2 x , E x , ψ 1 x x , ψ 2 x x ) 2 d x d τ . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equak_HTML.gif

Here, we have used the fact that G 2 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq156_HTML.gif holds. Therefore, we obtain the estimate (4.2) for the case of β = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq154_HTML.gif. □

In order to complete the proof of Proposition 4.1, we need to obtain the weighted estimate of ( φ 1 x , ψ 1 x , φ 2 x , ψ 2 x ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq157_HTML.gif.

Lemma 4.4 There exists a positive constant ε 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq23_HTML.gif such that if N 1 ( T ) + δ < ε 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq140_HTML.gif, then for β [ 0 , α ] https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq128_HTML.gif and ξ 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq129_HTML.gif,
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ65_HTML.gif
(4.14)
Proof Since the derivation of the estimate (4.14) is similar to that of (3.13) and (3.24), we only give the outline of the proof. Multiplying (3.16) by W = ( 1 + B x ) β ( 1 + t ) ξ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq158_HTML.gif, we have
i = 1 2 [ [ W ( φ i x 2 2 ρ i 3 + φ i x ψ i ρ i ) ] t + [ W ( u i φ i x 2 2 ρ i 3 φ i t ψ i ρ i ρ ˜ x ψ i 2 ρ i ) ] x + W P ( ρ i ) φ i x 2 ρ i 2 ] = i = 1 2 [ W t ( φ i x 2 2 ρ i 3 + φ i x ψ i ρ i ) + W ( ψ i x 2 + ( 1 ) i + 1 E φ i x ρ i + R 21 + R 22 ) W x ( u i φ i x 2 2 ρ i 3 φ i t ψ i ρ i ρ ˜ x ψ i 2 ρ i ) ] . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ66_HTML.gif
(4.15)
Integrating (4.15) over R + × ( 0 , t ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq111_HTML.gif and substituting (4.2) gives the estimate for φ 1 x https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq159_HTML.gif and φ 2 x https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq160_HTML.gif as
https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ67_HTML.gif
(4.16)
Here, we have used the inequalities
( ( 1 + B x ) β ) x | u 1 φ 1 x 2 2 ρ 1 3 φ 1 t ψ 1 ρ 1 ρ ˜ x ψ 1 2 ρ 1 + u 2 φ 2 x 2 2 ρ 2 3 φ 2 t ψ 2 ρ 2 ρ ˜ x ψ 2 2 ρ 2 | C δ ( 1 + B x ) β ( φ 1 x 2 + φ 2 x 2 ) + C ( 1 + B x ) β ( ψ 1 x 2 + ψ 2 x 2 ) + C β ( 1 + B x ) β 2 ( φ 1 2 + ψ 1 2 + φ 2 2 + ψ 2 2 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equal_HTML.gif
and
( 1 + B x ) β | R 21 + R 22 | ( ε + C δ ) ( 1 + B x ) β ( φ 1 x 2 + φ 2 x 2 ) + C ε ( 1 + B x ) β ( ψ 1 x 2 + ψ 2 x 2 ) + C δ ( 1 + B x ) β ( φ 1 2 + ψ 1 2 + φ 2 2 + ψ 2 2 ) , https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equam_HTML.gif

where ε is an arbitrary positive constant. We note that the third term on the right-hand side of the above inequality is estimated by applying the Poincaré-type inequality (3.23) for the case of β = 0 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq154_HTML.gif.

Next, we prove the estimate for ( ψ 1 x , ψ 2 x ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq161_HTML.gif. Multiply (3.25) by W = ( 1 + t ) ξ ( 1 + B x ) β https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq162_HTML.gif to get
( W ψ 1 x 2 2 + W ψ 2 x 2 2 ) t i = 1 2 ( W ψ i t ψ i x + W u i ψ i x 2 2 ) x + W i = 1 2 ψ i x x 2 ρ i = 1 2 W t i = 1 2 ψ i x 2 2 W x i = 1 2 ( ψ i t ψ i x + u i ψ i x 2 2 ) E W ψ 1 x x + E W ψ 2 x x + W R 3 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ68_HTML.gif
(4.17)
Integrate (4.17) in R + × ( 0 , t ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq111_HTML.gif and substitute (4.2) and (4.16) in the resultant equality with the inequalities
( ( 1 + B x ) β ) x | ψ 1 t ψ 1 x + u 1 ψ 1 x 2 2 + ψ 2 t ψ 2 x + u 2 ψ 2 x 2 2 | + ( 1 + B x ) β | R 3 | ε ( 1 + B x ) β i = 1 2 ψ i x x 2 + C ε ( 1 + B x ) β i = 1 2 ( φ i x 2 + ψ i x 2 + ψ i x 4 ) + C ε δ ( 1 + B x ) β 4 i = 1 2 ( φ i 2 + ψ i 2 ) https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equan_HTML.gif
and
0 ( 1 + B x ) β ( ψ 1 x 4 + ψ 2 x 4 ) d x C N 1 ( t ) ( ψ 1 x , ψ 1 x x , ψ 2 x , ψ 2 x x ) a , β 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equao_HTML.gif
This procedure yields
( 1 + t ) ξ ( ψ 1 x , ψ 2 x ) a , β 2 + 0 t ( 1 + τ ) ξ ( ψ 1 x x , ψ 2 x x ) a , β 2 d τ C ( ( φ 10 , ψ 10 , φ 20 , ψ 20 , E 0 ) a , β 2 + ( φ 10 x , ψ 10 x , φ 20 x , ψ 20 x ) a , β 2 ) + C ξ 0 t ( 1 + τ ) ξ 1 i = 1 2 ( φ i , ψ i , φ i x , ψ i x ) a , β 2 d τ + C ( N 1 ( t ) + δ ) M 1 ( t ) 2 . https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_Equ69_HTML.gif
(4.18)

Finally, adding (4.16) to (4.18) and taking N 1 ( t ) + δ https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq163_HTML.gif suitably small give the desired estimate (4.14). □

By the same inductive argument as in deriving (4.1), we can prove Proposition 4.1, which immediately yields the decay estimate (1.12).

Declarations

Acknowledgements

The research of Li is partially supported by the National Science Foundation of China (Grant No. 11171223) and the Innovation Program of Shanghai Municipal Education Commission (Grant No. 13ZZ109).

Authors’ Affiliations

(1)
Department of Mathematics, Hubei University of Science and Technology
(2)
Department of Mathematics, Shanghai Normal University

References

  1. Besse C, Claudel J, Degond P, et al.: A model hierarchy for ionospheric plasma modeling. Math. Models Methods Appl. Sci. 2004, 14: 393–415. 10.1142/S0218202504003283MathSciNetView Article
  2. Degond P: Mathematical modelling of microelectronics semiconductor devices. AMS/IP Stud. Adv. Math. 15. In Some Current Topics on Nonlinear Conservation Laws. Am. Math. Soc., Providence; 2000:77–110.
  3. Jüngel A Progress in Nonlinear Differential Equations and Their Applications. In Quasi-hydrodynamic Semiconductor Equations. Birkhäuser, Basel; 2001.View Article
  4. Sitnko A, Malnev V: Plasma Physics Theory. Chapman & Hall, London; 1995.
  5. Ducomet B: A remark about global existence for the Navier-Stokes-Poisson system. Appl. Math. Lett. 1999, 12: 31–37.MathSciNetView Article
  6. Ducomet B: Local and global existence for the coupled Navier-Stokes-Poisson problem. Q. Appl. Math. 2003, 61: 345–361.
  7. Zhang Y-H, Tan Z: On the existence of solutions to the Navier-Stokes-Poisson equations of a two-dimensional compressible flow. Math. Methods Appl. Sci. 2007, 30: 305–329. 10.1002/mma.786MathSciNetView Article
  8. Chan D:On the nonexistence of global weak solutions to the Navier-Stokes-Poisson equations in R N https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq12_HTML.gif. Commun. Partial Differ. Equ. 2010, 35: 535–557. 10.1080/03605300903473418View Article
  9. Hao C-C, Li H-L: Global existence for compressible Navier-Stokes-Poisson equations in three and higher dimensions. J. Differ. Equ. 2009, 246: 4791–4812. 10.1016/j.jde.2008.11.019MathSciNetView Article
  10. Li H-L, Matsumura A, Zhang G-J:Optimal decay rate of the compressible Navier-Stokes-Poisson system in R 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq14_HTML.gif. Arch. Ration. Mech. Anal. 2010, 196: 681–713. 10.1007/s00205-009-0255-4MathSciNetView Article
  11. Zhang G-J, Li H-L, Zhu C-J:Optimal decay rate of the non-isentropic compressible Navier-Stokes-Poisson system in R 3 https://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-124/MediaObjects/13661_2012_Article_376_IEq14_HTML.gif. J. Differ. Equ. 2011, 250: 866–891. 10.1016/j.jde.2010.07.035MathSciNetView Article
  12. Wang W-K, Wu Z-G: Pointwise estimates of solution for the Navier-Stokes-Poisson equations in multi-dimensions. J. Differ. Equ. 2010, 248: 1617–1636. 10.1016/j.jde.2010.01.003View Article
  13. Donatelli D, Marcati P: A quasineutral type limit for the Navier-Stokes-Poisson system with large data. Nonlinearity 2008, 21: 135–148. 10.1088/0951-7715/21/1/008MathSciNetView Article
  14. Ju Q-C, Li F-C, Li Y, Wang S: Rate of convergence from the Navier-Stokes-Poisson system to the incompressible Euler equations. J. Math. Phys. 2009., 50: Article ID 013533
  15. Wang S, Jiang S: The convergence of the Navier-Stokes-Poisson system to the incompressible Euler equations. Commun. Partial Differ. Equ. 2006, 31: 571–591. 10.1080/03605300500361487View Article
  16. Li H-L, Yang T, Zou C: Time asymptotic behavior of the bipolar Navier-Stokes-Poisson system. Acta Math. Sci., Ser. B 2009, 29: 1721–1736. 10.1016/S0252-9602(10)60013-6MathSciNetView Article
  17. Duan R-J, Yang X-F: Stability of rarefaction wave and boundary layer for outflow problem on the two-fluid Navier-Stokes-Poisson equations. Commun. Pure Appl. Anal. 2013, 12: 985–1014.MathSciNetView Article
  18. Gasser I, Hsiao L, Li H-L: Large time behavior of solutions of the bipolar hydrodynamical model for semiconductors. J. Differ. Equ. 2003, 192: 326–359. 10.1016/S0022-0396(03)00122-0MathSciNetView Article
  19. Huang F-M, Li Y-P: Large time behavior and quasineutral limit of solutions to a bipolar hydrodynamic model with large data and vacuum. Discrete Contin. Dyn. Syst. 2009, 24: 455–470.MathSciNetView Article
  20. Huang F-M, Mei M, Wang Y: Large time behavior of solution to n -dimensional bipolar hydrodynamic model for semiconductors. SIAM J. Math. Anal. 2011, 43: 1595–1630. 10.1137/100810228MathSciNetView Article
  21. Kawashima S, Nishibata S, Zhu P: Asymptotic stability of the stationary solution to the compressible Navier-Stokes-Poisson equations in half space. Commun. Math. Phys. 2003, 240: 483–500.MathSciNetView Article
  22. Kawashima S, Zhu P: Asymptotic stability of nonlinear wave for the compressible Navier-Stokes-Poisson equations in half space. J. Differ. Equ. 2008, 224: 3151–3179.MathSciNetView Article
  23. Nakamura T, Nishibata S, Yuge T: Convergence rate of solutions toward stationary solutions to the compressible Navier-Stokes equation in a half line. J. Differ. Equ. 2007, 241: 94–111. 10.1016/j.jde.2007.06.016MathSciNetView Article
  24. Nikkuni Y, Kawashima S: Stability of stationary solutions to the half-space problem for the discrete Boltzmann equation with multiple collisions. Kyushu J. Math. 2000, 54: 233–255. 10.2206/kyushujm.54.233MathSciNetView Article
  25. Matsumura A, Nishihara K: Asymptotic stability of traveling waves for scalar viscous conservation laws with non-convex nonlinearity. Commun. Math. Phys. 1994, 165: 83–96. 10.1007/BF02099739MathSciNetView Article

Copyright

© Zhou and Li; licensee Springer 2013

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://​creativecommons.​org/​licenses/​by/​2.​0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.