Existence of global smooth solutions for Euler equations with symmetry (II)

Nonlinear Analysis 41 (2000) 187 – 203

www.elsevier.nl/locate/na

Existence of global smooth solutions for Euler equations with symmetry (II) Tong Yang a;∗ , Changjiang Zhu a;b , Yongshu Zheng c a

Department of Mathematics, City University of Hong Kong, Tat Chee Avenue, Kowloon, Hong Kong b Wuhan Institute of Physics and Mathematics, Chinese Academy of Sciences, People’s Republic of China c Department of Mathematics, Huaqiao University, Fujian, People’s Republic of China Received 12 January 1998; accepted 11 April 1998

Keywords: A priori estimates; Euler equations; Characteristic method; Spherical symmetry; Global smooth solutions

1. Introduction Consider the compressible Euler equations governing the gas ow surrounding a solid ball with mass M and frictional damping in n dimensions, ˜ = 0; ˜t + ∇ · (u) ˜ ⊗ u) + ∇P() ˜ = −M ˜ (u) ˜ t + ∇ · (u

x − 2u; ˜ |x| n

(1.1)

where ; ˜ u; P and M are the density, velocity, pressure and mass of the gas, respectively, n ≥ 3 is the dimension of x, and ¿0 is the frictional constant. In the following discussion, we assume the pressure satis es the law and 1¡ ¡3, i.e. P = K 2 ˜ ; K is a positive constant. We will study the existence and non-existence of global smooth solutions for the initial boundary problem of Eq. (1.1). Firstly, we will show that regular solutions cannot be global if the initial density has compact support. This is a generalization of Theorem 2.1 in [7] to the case when there is an extra term due to the external force caused by the mass of the ball. In fact, the authors in [7] have noted that the Theorem 2.1 there is valid for any value of  by studying an ordinary dierential equation of H (t) as shown in Section 2 of this paper. ∗

Corresponding author. Tel.: +852-2788-8646; fax: +852-2788-8561. E-mail address: [email protected] (T. Yang)

0362-546X/00/$ - see front matter ? 2000 Elsevier Science Ltd. All rights reserved. PII: S 0 3 6 2 - 5 4 6 X ( 9 8 ) 0 0 2 7 3 - 9

188

T. Yang et al. / Nonlinear Analysis 41 (2000) 187 – 203

For Euler equations without damping in multi-dimensional space, Sideris [12] gave sucient conditions for non-global existence of C 1 solutions when inf  0 (x)¿0, where  0 (x) is the initial density. The non-existence of C 1 solutions in [12] is related to the shock formation. In Section 2, we will study the singularity of the solutions at the vacuum states when the solutions contain no shocks. Thus, the time when the regular solutions blow up in our discussion is before the time when shock forms. When inf  0 (x) = 0, Makino et al. [10] proved the non-global existence of regular solutions by assuming the initial data ( 0 (x); u 0 (x)) to be of compact support, where u 0 (x) is the initial velocity. For the Euler–Poisson equations governing gaseous stars, Makino and Perthame [9] proved the non-global existence of tame solutions under the condition of spherical symmetry. Local existence of tame or regular solutions for these two systems was proved by Makino et al. [8, 10] by using Kato’s [3] theory for quasilinear symmetric hyperbolic system. Secondly, we study the global existence of smooth solutions of the initial boundary value problem of Eq. (1.1) with spherical symmetry. Even though the global existence of regular solutions for Cauchy and initial boundary value problem of one-dimensional quasilinear hyperbolic systems has been extensively investigated (see [2,4–6,14]), much less is known for systems in high dimensions. To study the problem with spherical symmetry, where the system reduces to a one-dimensional system with singular source terms, is an initial stage to achieve this goal in high dimensions. The singularity is at r = 0 or ∞, where r is the radius. In Section 3, we study Eq. (1.1) with spherical symmetry and damping outside a core region of ball with mass M . Some sucient conditions are given for the global existence of smooth solutions when the frictional damping is suciently large. The study is a generalization of the one in [13] to the case when M 6= 0, which is based on technical estimation of the C 1 -norm of the solutions. 2. No global existence In this section, we consider the initial-boundary value problem of Eq. (1.1) with initial and boundary conditions given as follows: (x; ˜ 0) = ˜ 0 (x) ≥ 0; u · x = 0;

u(x; 0) = u 0 (x);

when |x| ≥ 1;

when |x| = 1:

(2.1)

We consider the regular solution of Eqs. (1.1) and (2.1) de ned as follows. Deÿnition 2.1. A solution of Eqs. (1.1) and (2.1) is called a regular solution in [0; T ) × , if (i) (; ˜ u) ∈ C 1 ([0; T ) × ); ˜ ≥ 0, (ii) ˜ −1 ∈ C 1 ([0; T ) × ), and Mx u t + u · ∇u = −u − |x| n holds on the exterior of the support of . ˜ Here = R n \B1 ; B1 = {x: |x|¡1}.

T. Yang et al. / Nonlinear Analysis 41 (2000) 187 – 203

189

Theorem 2.2. Let ((x; ˜ t); u(x; t)) be a regular solution of Eqs. (1.1) and (2.1) on 0 ≤ t¡T . If the support of the initial data ˜ 0 (x) has compact support and ˜ 0 (x) 6≡ 0 and the mass M is suciently small; then T is ÿnite. Remark 2.3. The following proof is similar to that of Makino et al. for the case when  = 0; M = 0 [10] and Liu and Yang for the case ¿0; M = 0 [7]. We generalize their proof to the case when ¿0; M ¿0. Proof. Let (t) = supp (x; ˜ t); S(t) = @ (t). By Eq. (1.1) and the de nition of regular solution, for any x ∈ S(t 0 ), there exists x 0 ∈ S(0) and a curve x(t) connecting x 0 and x such that dx(t) = u(x(t); t); dt

x(t) ∈ S(t); 0 ≤ t ≤ t 0 :

Let L = hx; xi ; I = hu; ui, where h·; ·i denotes inner product in R n . Then we have dL = 2 hu; xi; dt   dL 2M du d2 L ; x + 2 hu; ui = − − = 2 + 2I dt 2 dt dt |x| n−2 and dI =2 dt



du ;u dt



 = −2I − 2M

u;

 x : |x| n

(2.2)

(2.3)

It follows from Eqs. (2.2) and (2.3) that 2M dL dL 1 dI M d2 L − − = − − L−n=2 : dt 2 dt  dt |x| n−2  dt

(2.4)

Integrating Eq. (2.4), we have 1 2M dL ≤ −L − I + L1−n=2 + C 0 ; dt  (n − 2)

(2.5)

where C 0 is a constant. Thus 2M dL ≤ −L + L1−n=2 + C 0 : dt (n − 2)

(2.6)

By letting L1 = L n=2 , we have C 0 n 1−2=n Mn n dL1 + ≤ − L1 + L ; dt 2 2 1 (n − 2) which implies that there exists a constant L∗1 depending only on the initial data, such that 1 ≤ L1 ≤ L∗1 :

(2.7)

190

T. Yang et al. / Nonlinear Analysis 41 (2000) 187 – 203

Now as in [10], let Z 1 (x; ˜ t)|x|2 dx: H (t) = 2 (t)

(2.8)

From Eq. (1.1), we have Z 1 u ˜ · x dx H 0 (t) = 2 (t) and H 00 (t) =

Z

(t)

(|u| ˜ 2 + nP) dx − H 0 (t) +

Z

Z D0

P dS − M

(t)

˜ dx; |x| n−2

(2.9)

where D 0 =R{x: |x| = 1}. ˜ t) dx be the total mass. Then we have from the Holder inequality Let m = (t) (x; that Z

Z m=

(t)

(x; ˜ t) dx ≤



(t)

1=  Z

˜ dx

1= 0

(t)

dx

;

where 1= + 1= 0 = 1. Thus there exists a constant V ∗ ¿0 such that Z

(t)

˜ dx

1=

0

≥ m(V ∗ )−1= :

Thus, 0

H 00 (t) + H 0 (t) ≥ (nK 2 m −1 (V ∗ )− = − M )m: When M is suciently small, there exists a constant ¿0, such that H 00 (t) + H 0 (t) ≥ : Integrating the above inequality yields   1 −t 1 H 0 (0) −t  t+ e − − (e − 1): H (t) ≥ H (0) +     Thus, H (t) → ∞

as t → ∞;

which is a contradiction to Eqs. (2.7) and (2.8). Hence, T must be nite and this completes the proof of Theorem 2.2.

T. Yang et al. / Nonlinear Analysis 41 (2000) 187 – 203

191

3. Global existence with spherical symmetry Now, we are going to study the global existence of smooth solutions for Eq. (1.1) with spherical symmetry outside a core region. That is, we look for solutions of the form x · u(|x|; t): ˜ = (|x|; ˜ t); u= |x| By denoting r = |x|, Eq. (1.1) becomes ˜t +

1 r n−1

(r n−1 u) ˜ r = 0;

(u ˜ t + uur ) + Pr = −

(3.1)

M ˜ − 2u; ˜ r n−1

where P() ˜ = K 2 ˜ , 1¡ ¡3. We consider system (3:1) with the following initial and boundary conditions: (r; ˜ 0) = ˜ 0 (r);

u(r; 0) = u 0 (r);

u(1; t) = 0:

(3.2)

˜ then we can rewrite Eq. (3.1) as Let  = r n−1 , t + (u)r = 0; ut + uur +

K 2  −2 M K 2 (n − 1) −1  = − − 2u + : r r (n−1)( −1) r n−1 r n r (n−1)( −2)

In the following discussion we use the Lagrangian coordinates as follows: Z r (r; t) dr:  = t;  = 1

(3.3)

(3.4)

Then ¿0 as long as ¿0 for r¿1, and Eq. (3.3) becomes  + 2 u = 0; u +

K 2  −1 M K 2 (n − 1) −1  = − − 2u + :  r (n−1)( −1) r n−1 r n r (n−1)( −2)

(3.5)

Now we let  = −1 , and replace (; ) by (x; t). Eq. (3.5) can be rewritten as t − ux = 0; ut −

K 2 − −1 M K 2 (n − 1)1− x = − n−1 − 2u + (n−1)( −1) r r r n r (n−1)( −2)

(3.6)

with the following initial and boundary conditions: (x; 0) =  0 (x);

u(x; 0) = u 0 (x);

By transformation (3:4), we have Z x (s; t) ds; x ≥ 0: r=1 + 0

u(0; t) = 0:

(3.7)

192

T. Yang et al. / Nonlinear Analysis 41 (2000) 187 – 203

To symmetrize system (3:6), we use the following Riemann invariants: √ 2K −(1=2)(n−1)( −1) −( −1)=2 r  ; w=u +

−1 √ 2K −(1=2)(n−1)( −1) −( −1)=2 r  : z=u −

−1

(3.8)

By using rt = u and rx = , Eq. (3.6) becomes wt + wx = − zt + zx = −

M r n−1 M

r n−1

− (w + z) −

(n − 1)( − 1) (w − z)(w + z); 8r

(3.9)

(n − 1)( − 1) − (w + z) + (w − z)(w + z) 8r

with the corresponding initial and boundary conditions given by w(x; 0) = w 0 (x); z(x; 0) = z 0 (x); w(0; t) + z(0; t) = 0: √ Here − =  = K Ar n−1 (w − z)( +1)=( −1) are the two characteristic speeds and  A=

−1 √ 4K

( +1)=( −1) :

For later use, we list some equations as follows: 1

−1 dr = (w + z) + (w − z); d t 2 4 1

−1 dr = (w + z) − (w − z); d t 2 4 √ d(w − z) = − 2K Ar n−1 (w − z)( +1)=( −1) zx d t −

(n − 1)( − 1) (w − z)(w + z); 4r

√ d(w − z) = − 2K Ar n−1 (w − z)( +1)=( −1) wx d t −

(n − 1)( − 1) (w − z)(w + z); 4r

√ d(w + z) 2M = − n−1 − 2(w + z) + 2K Ar n−1 (w − z)( +1)=( −1) zx ; d t r √ 2M d(w + z) = − n−1 − 2(w + z) − 2K Ar n−1 (w − z)( +1)=( −1) wx ; d t r

(3.10)

where we have used the notation d=d t = @=@t + (@=@x) and d=d t = @=@t + (@=@x):

T. Yang et al. / Nonlinear Analysis 41 (2000) 187 – 203

193

From now on, we assume the initial and boundary conditions satisfy w 0 (x);

z 0 (x) ∈ Cb1 [0; ∞);

(3.11)

(r 0 (x)) n=2−1 (w 0 (x) − z 0 (x)) ≥  and w 0 (0) + z 0 (0) = 0;

(3.12)

 0 (0)w 0x (0) +  0 (0)z 0x (0) = −2M;

where  is a positive constant, and  0 (x) and  0 (x) are the given characteristic speeds at t = 0. Remark 3.1. There exists a special solution of Eq. (3.9) satisfying Eqs. (3.11) and (3.12), i.e. w(x; t) = − z(x; t) = M1 (r(x; t))1−n=2 ; √ p where M1 = 2 M = (n − 2)( − 1). Now, we are ready to prove the following lemma on the C 0 -norm of the solutions inside the smoothness region of the solutions. Lemma 3.2. Under the conditions of Eqs. (3.11) and (3.12), if |(r 0 (x)) n=2−1 w 0 (x)|; |(r 0 (x)) n=2−1 z 0 (x)| ≤ M1 ;

(3.13)

then there exists a positive constant  0 depending only on M1 and  such that when ¿ 0 , we have |(r(x; t)) n=2−1 w(x; t)|; |(r(x; t)) n=2−1 z(x; t)| ≤ M1 ;

(3.14)

w(x; t) − z(x; t)¿0;

(3.15)

where (x; t) is inside the smoothness region of the solutions and M1 is the constant deÿned in the above Remark 3:1: Proof. We now prove Eq. (3.15) rst. By Eq. (3.10), we have d(w − z) = −D(x; t)(w − z); d t

(3.16)

where √ (n − 1)( − 1) (w + z): D(x; t) = 2K Ar n−1 (w − z)2=( −1) wx + 4r Integrating Eq. (3.16) along the -characteristic curve, we have  Z t  1−n=2 exp − D(x(); ) d ¿0; (w − z)(x; t) ≥ (r 0 (x)) 0

as long as the solution is smooth.

194

T. Yang et al. / Nonlinear Analysis 41 (2000) 187 – 203

To prove Eq. (3.14), we introduce the following transformations: w(x; t) = w(x;  t) + M1 (r(x; t))1−n=2 ; z(x; t) = z (x; t) − M1 (r(x; t))1−n=2 :

(3.17)

Then Eq. (3.9) becomes  w t + w x = −A(x; t)(w + z ) + br −n=2 w; z t + z x = −B(x; t)(w + z ) − br −n=2 z ;

(3.18)

where A(x; t) =  + +

(3 − )(n − 2) − 2(n − 1)( − 1) M1 r −n=2 ; 8

B(x; t) =  − − b = 12 Let

p

(n − 1)( − 1) (w − z ) 8r

(n − 1)( − 1) (w − z ) 8r

(3 − )(n − 2) − 2(n − 1)( − 1) M1 r −n=2 ; 8

(n − 2)( − 1)M :

 N t w = W + (x + Ce ) ebt ; L   N z = − Z + (x + Cet ) ebt ; L

(3.19)



(3.20)

where C and L are two positive constants to be determined later. A smooth solution exists at least locally under conditions (3:11) and (3:12), cf. [1]. We use N to denote the bound for |w(x; t)| and |z(x; t)| of the local solution. Using Eq. (3.20), the system (3:18) for (w(x;  t); z (x; t)) can be written as follows: Wt + Wx +


N N (Cet + ) + (x + Cet )b 1 − r −n=2 L L

= −A(x; t)(W − Z) − bW + br −n=2 W; Zt + Z x +

N N (Cet + ) + (x + Cet )b(1 + r −n=2 ) L L

= B(x; t)(W − Z) − bZ − br −n=2 Z:

(3.21)

T. Yang et al. / Nonlinear Analysis 41 (2000) 187 – 203

195

If we consider system (3:21) in the region [0; L] × R+ , then the initial and boundary conditions have the following properties: N (x + C)¡0; L N Z(x; 0) = −z (x; 0) − (x + C)¡0; L N −bt W (L; t) = w(L;  t)e − N − Cet ¡0; L N −bt Z(L; t) = −z (L; t)e − N − Cet ¡0: L

W (x; 0) = w(x;  0) −

(3.22)

From Eqs. (3.21) and (3.22), we claim that W (x; t)¡0;

Z(x; t)¡0;

(x; t) ∈ [0; L] × [0; T ]:

(3.23)

If not, we let t = supt {t: W (x; )¡0; Z(x; )¡0; ∀x ∈ [0; L];  ∈ (0; t)}, then 0¡t ≤ T ¡+∞. By the continuity of W (x; t) and Z(x; t), there exists (x;  t) with 0 ≤ x¡L,  such that one of the following cases holds: Case 1: When x ∈ (0; L), @W @W ≥ 0; = 0; W (x;  t) = 0; Z(x;  t) ≤ 0; @t (x; t) @x (x; t) or Z(x;  t) = 0;

W (x;  t) ≤ 0;

Case 2: When x = 0, W (x;  t) = Z(x;  t) = 0;

@Z ≥ 0; @t (x; t)

@Z ≥ 0; @t (x; t)

@Z = 0: @x (x; t)

@Z ≤ 0: @x (x; t)

For the above two cases, if we assume the following a priori estimate: |(r(x; t))n=2−1 w(x; t)|; |(r(x; t))n=2−1 z(x; t)| ≤ M1 ;

(3.24)

then for suciently large , we have A(x; t)¿0;

B(x; t)¿0:

By applying the “maximum principle”, cf. [13, 14], when C¿ sup  for all w; z under consideration, we have a contradiction. Therefore, Eq. (3.23) holds. Equality (3:20) and inequality (3:23) imply that   N t (x + Ce ) ebt ; w(x;  t)¡ L   N (x + Cet ) ebt : z (x; t)¿− L

196

T. Yang et al. / Nonlinear Analysis 41 (2000) 187 – 203

Since L can be arbitrary, letting L → ∞ yields w(x;  t) ≤ 0;

z (x; t) ≥ 0:

Hence by Eq. (3.17), we have w(x; t) ≤ M1 (r(x; t))1−n=2 ;

z(x; t) ≥ −M1 (r(x; t))1−n=2 :

(3.25)

It is easy to get Eq. (3.14) from Eqs. (3.15) and (3.25). This completes the Proof of Lemma 3.2. Now, we estimate the derivatives of w(x; t); z(x; t) with respect to x. Let P(x; t) = wx (x; t); Q(x; t) = zx (x; t). By dierentiating Eq. (3.9) with respect to x, we have the following system for P; Q: √ + 1 n−1 r (w − z)2=( −1) (P − Q)P Pt + Px = −(P + Q) − K A

−1 (n − 1)( − 1) (n − 1)( − 1) (w − z)P − (w − z)(P + Q) − 4r 8r −

(n − 1)M (n − 1)( − 1) (w + z)(P − Q) + B(w − z)−2=( −1) 8r r 2n−1

+

(n − 1)( − 1) B(w − z)−(3− )=( −1) (w + z); 8r n+1

√ + 1 n−1 r (w − z)2=( −1) (Q − P)Q Qt + Qx = −(P + Q) − K A

−1 +

(n − 1)( − 1) (n − 1)( − 1) (w − z)Q + (w − z)(P + Q) 4r 8r

+

(n − 1)M (n − 1)( − 1) (w + z)(P − Q) + B(w − z)−2=( −1) 8r r 2n−1

−

(n − 1)( − 1) B(w − z)−(3− )=( −1) (w + z); 8r n+1

(3.26)

where  B=

−1 √ 4K

−2=( −1) :

Let P = (w − z)l r P;

 Q = (w − z)l r Q;

(3.27)

T. Yang et al. / Nonlinear Analysis 41 (2000) 187 – 203

197

where l = −( + 1)=2( − 1)¡0 and is a constant to be determined later. Then system (3:26) becomes √ + 1 n+ −1 Pt + Px = −(P + Q) − K A (w − z)(3− )=2( −1) P 2 r

−1 −

(3n + 2 − 3)( − 1) (n − 1)( − 1) (w − z)P − (w − z)Q 8r 8r

−

(n − 1)( − 1) (n − 1) + 2 (w + z)P + (w + z)Q 4r 8r

+

(n − 1)M B(w − z)−(3− )=2( −1) r 2n+ −1

+

(n − 1)( − 1) B(w − z)(3 −5)=2( −1) (w + z); 8r n+ +1

√ + 1 n+ −1 r Qt + Qx = −(P + Q) − K A (w − z)(3− )=2( −1) Q2

−1 (n − 1)( − 1) (3n + 2 − 3)( − 1) (w − z)P + (w − z)Q 8r 8r (n − 1) + 2 (n − 1)( − 1) (w + z)P − (w + z)Q + 8r 4r +

+

(n − 1)M B(w − z)−(3− )=2( −1) r 2n+ −1

−

(n − 1)( − 1) B(w − z)(3 −5)=2( −1) (w + z): 8r n+ +1

(3.28)

Let P = F + a1 r − −n+1 (w − z)m + a2 r − −n (w − z)m+1 + a3 r − −n (w − z)m (w + z) Q = G + a1 r − −n+1 (w − z)m − a2 r − −n (w − z)m+1 + a3 r − −n (w − z)m (w + z); where m = −(3 − )=2( − 1)¡0. Then we have Ft + Fx √ + 1 n+ −1 = −K A (w − z)(3− )=2( −1) P 2 − P r

−1 −

(3n + 2 − 3)( − 1) (w − z)P 8r

−

√ (n − 1) + 2 (w + z)P + (2K Ama1 − )Q 4r

(3.29)

198

T. Yang et al. / Nonlinear Analysis 41 (2000) 187 – 203

  √ √ (n − 1)( − 1) 1 (w − z)Q + 2K A(m + 1)a2 − 2K Aa3 − 8 r   √ (n − 1)( − 1) 1 (w + z)Q + 2K Ama3 + 8 r (n − 1)( − 1) ma1 r − −n (w − z)−(3− )=2( −1) (w + z) 4   1 ( + n − 1)a1 + 2a3 r − −n (w − z)−(3− )=2( −1) (w + z) + 2 +

+

−1 ( + n − 1)a1 r − −n (w − z)(3 −5)=2( −1) 4

+

−1 ( + n)a2 r − −n−1 (w − z)(5 −7)=2( −1) 4

(n − 1)( − 1) (m + 1)a2 r − −n−1 (w − z)(3 −5)=2( −1) (w + z) 4  

−1 1 ( + n)a2 + ( + n)a3 r − −n−1 (w − z)(3 −5)=2( −1) (w + z) + 2 4 +

+ ((n − 1)B + 2a3 )Mr − −2n+1 (w − z)−(3− )=2( −1) (n − 1)( − 1) − −n−1 Br (w − z)(3 −5)=2( −1) (w + z) 8   1 (n − 1)( − 1) m + ( + n) a3 r − −n−1 (w − z)−(3− )=2( −1) (w + z)2 ; + 4 2 +

Gt + Gx √ + 1 n+ −1 r = −K A (w − z)(3− )=2( −1) Q2 − Q

−1 +

(3n + 2 − 3)( − 1) (w − z)Q 8r

√ (n − 1) + 2 (w + z)Q + (2K Ama1 − )P 4r   √ √ (n − 1)( − 1) 1 (w − z)P − 2K A(m + 1)a2 − 2K Aa3 − 8 r   √ (n − 1)( − 1) 1 (w + z)P + 2K Ama3 + 8 r −

+

(n − 1)( − 1) ma1 r − −n (w − z)−(3− )=2( −1) (w + z) 4

T. Yang et al. / Nonlinear Analysis 41 (2000) 187 – 203

199



 1 ( + n − 1)a1 + 2a3 r − −n (w − z)−(3− )=2( −1) (w + z) 2

−1 ( + n − 1)a1 r − −n (w − z)(3 −5)=2( −1) − 4

−1 ( + n)a2 r − −n−1 (w − z)(5 −7)=2( −1) + 4 +

(n − 1)( − 1) (m + 1)a2 r − −n−1 (w − z)(3 −5)=2( −1) (w + z) 4  

−1 1 ( + n)a2 + ( + n)a3 r − −n−1 (w − z)(3 −5)=2( −1) (w + z) − 2 4 −

+ ((n − 1)B + 2a3 )Mr − −2n+1 (w − z)−(3− )=2( −1) (n − 1)( − 1) − −n−1 (w − z)(3 −5)2( −1) (w + z) Br 8   1 (n − 1)( − 1) m + ( + n) a3 r − −n−1 (w − z)−(3− )=2( −1) (w + z)2 : + 4 2 (3.30) −

Now choose ai ; i = 1; 2; 3; as follows: √ 2K Ama1 −  = 0; √ √ (n − 1)( − 1) = 0; 2K A(m + 1)a2 − 2K Aa3 − 8 √ (n − 1)( − 1) = 0; 2K Ama3 + 8 i.e. ( − 1) ¡0; a1 = − √ K A(3 − ) (n − 1)( − 1)2 ( + 1) ; a2 = √ 8K A(3 − 5)(3 − ) (n − 1)( − 1)2 ¿0; a3 = √ 8K A(3 − ) where 1¡ ¡ 53 or 53 ¡ ¡3. Then the system for F and G becomes √ + 1 n+ −1 Ar Ft + Fx = −K (w − z)(3− )=2( −1) F 2

−1   3 − 1  + 1 F + 3−   22 ( − 1)2 − −n+1 +  (w − z)−(3− )=2( −1) ; − √ 1 r K A(3 − )2

(3.31)

200

T. Yang et al. / Nonlinear Analysis 41 (2000) 187 – 203

√ + 1 n+ −1 Gt + Gx = −K (w − z)(3− )=2( −1) G 2 Ar

−1   3 − 1  + 2 G + 3−   22 ( − 1)2 + 2 r − −n+1 (w − z)−(3− )=2( −1) ; − √ K A(3 − )2

(3.32)

where i ; i are continuous and bounded functions of the corresponding variables, i = 1; 2. Based on Lemma 3.2 and system (3:32), we are now ready to prove the following theorem. Theorem 3.3. Under the conditions of Lemma 3.2, if 1¡ ¡ 53 or 53 ¡ ¡3, and  is suciently large, then there exists a constant M2 depending only on M1 and , such that if |M1 − |1; F(x; 0); G(x; 0) ≥ M2 ; then there exists a global smooth solution for system (3:1) and (3:2). Proof. Firstly, we assume a priori estimate (r(x; t))n=2−1 (w(x; t) − z(x; t)) ≥ 1 ¿0;

(3.33)

where 1 is a positive constant depending only on M1 ; M2 and . By Lemma 3.2, we have |i |; |i |¡M3 ;

i = 1; 2;

where M3 is a positive constant depending only on M1 ; M2 and . When  is large enough and 1¡ ¡3, we know that the equation   √ + 1 n+ −1 3 − 1 Ar  + i y (w − z)(3− )=2( −1) y2 + 0 = −K

−1 3−   2 2 2 ( − 1) + i r − −n+1 (w − z)−(3− )=2( −1) ; − √ K A(3 − )2 has two positive roots given by   2( − 1)2 1 1 1=2  + i  r −n− +1 (w − z)−(3− )=2( −1) ; yi = √ K A( + 1)(3 − )  

−1 yi2 =  + i2 1=2 r −n− +1 (w − z)−(3− )=2( −1) ; i = 1; 2; √ K A(3 − )

T. Yang et al. / Nonlinear Analysis 41 (2000) 187 – 203

201

with yi1 ¡yi2 ; i = 1; 2, where |i1 | and |i2 | have an upper bound depending only on M1 ; M2 and . Therefore, when 1¡ ¡ 53 or 53 ¡ ¡3, if we choose =

(n − 2)(3 − ) − n + 1; 4( − 1)

then sup yi1 ¡inf yi2 , provided 1 −M1 is suciently small. Thus, there exists a constant M2 = max{sup y11 ; sup y21 } depending only on M1 and , such that if F(x; 0); G(x; 0) ≥ M2 ; we have M2 ≤ F(x; t);

G(x; t) ≤ sup{F(x; 0); G(x; 0)}:

(3.34)

Combining Lemma 3.2 and Eq. (3.34) yields the global existence of a smooth solution for Eqs. (3.1) and (3.2). Now, it remains to verify the a priori estimate (3:33). To do this, let H (x; t) = (r(x; t))n=2−1 (w(x; t) − z(x; t)):

(3.35)

Then, from Eqs. (3.10), (3.29) and (3.34), we have −

1 √

2K A

H

−(3− )=2( −1)−1

d H+ d t



 ( − 1) + 3 H −(3− )=2( −1) ≤ M4 ; √ K A(3 − ) (3.36)

where 3 ∈ Cb1 and M4 is a positive constant depending only on M1 ; M2 and . From Eq. (3.36), we have d d t



Z t  H −(3− )=2( −1) exp ( + 4 ()) d 0

Z t  ( + 4 ()) d ; ≤ M5 exp 0

where 4 () =

√ K A(3 − ) 3 ();

−1

and M5 is a positive constant depending only on M1 ; M2 and . When  is suciently large, we have   + 4 ()¿ : 2

(3.37)

202

T. Yang et al. / Nonlinear Analysis 41 (2000) 187 – 203

Hence, Z

t 0

Z s  exp ( + 4 ()) d ds

2 ≤  =

2 

0

Z

t 0

Z s  ( + 4 (s)) exp ( + 4 ()) d ds 0



Z t   exp ( + 4 ()) d − 1 : 0

(3.38)

Integrating Eq. (3.37) and using Eq. (3.38), we have H −(3− )=2( −1) ≤ M6 ; where M6 is a positive constant depending only on M1 ; M2 and . Therefore, when 1¡ ¡ 53 or 53 ¡ ¡3, we have (r(x; t))n=2−1 (w(x; t) − z(x; t)) ≥ 1 ¿0: This completes the proof of Theorem 3.3. Remark 3.4. When = 53 , the conclusion of Theorem 3.3 still holds. Acknowledgements The research of the rst author was supported in part by the RGC Competitive Earmarked Research Grant # 9040190. The research of the second author was supported in part by the grant of the National Youth Science Foundation of China #19301038. References [1] A. Douglis, Existence theorems for hyperbolic systems, Comm. Pure Appl. Math. 5 (1952) 119–154. [2] F. John, Formation of singularities in one-dimensional nonlinear wave propagation, Comm. Pure Appl. Math. 27 (1974) 377– 405. [3] T. Kato, The Cauchy problem for quasi-linear symmetric hyperbolic systems, Arch. Rat. Mech. Anal. 58 (1945) 181–205. [4] P.D. Lax, Development of singularities of solutions of nonlinear hyperbolic partial dierential equations, J. Math. Phys. 5 (1964) 611– 613. [5] Lin Longwei, Yang Tong, Existence and nonexistence of global smooth solutions for damped p-system with “really large” initial data, J. Part. Di. Eqs. 4 (1991) 45–51. [6] T.-P. Liu, Development of singularities in the nonlinear waves for quasilinear hyperbolic partial dierential equations, J. Di. Eqs. 33 (1979) 92–111. [7] T.-P. Liu, T. Yang, Compressible Euler equations with vacuum, J. Di. Eqs. 137 (1997). [8] T. Makino, Blowing up solutions of the Euler–Poisson equation for the evolution of gaseous stars, Transport Theory Statist. Phys. 21 (1992) 615– 624. [9] T. Makino, B. Perthame, Sur les solutions a symmetric spherique de l’equation d’Euler–Poisson pour l’evolution d’etoiles gazeuses, Japan J. Appl. Math. 7 (1990) 165–170.

T. Yang et al. / Nonlinear Analysis 41 (2000) 187 – 203

203

[10] T. Makino, S. Ukai, S. Kawashima, Sur la solution a support compact de l’equation d’Euler compressible, Japan J. Appl. Math. 3 (1986) 249–257. [11] T. Nishida, Nonlinear hyperbolic equations and related topics in uid dynamics, Publications Mathematiques D’ocsay 78.02, Department de Mathematique, Paris-Sud, 1978. [12] T. Sideris, Formation of singularities in three-dimensional compressible uids, Comm. Math. Phys. 101 (1985) 475– 485. [13] L.A. Ying, T. Yang, C.J. Zhu, Existence of global smooth solutions for Euler equations with symmetry, Comm. Partial Di. Eqs. 22 (1997) 1361–1387. [14] C.J. Zhu, Global resolvability for a viscoelastic model with relaxation, Proc. Roy. Soc. Edinburgh 125A (1995) 1277–1285.