Electronic Journal of Differential Equations, Vol. 2007(2007), No. 92, pp. 1–9. ISSN: 1072-6691. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu ftp ejde.math.txstate.edu (login: ftp) CONTINUOUS DEPENDENCE FOR THE BRINKMAN EQUATIONS OF FLOW IN DOUBLE-DIFFUSIVE CONVECTION HONGLIANG TU, CHANGHAO LIN Abstract. This paper concerns the structural stability for convective motion in a fluid-saturated porous medium under the Brinkman scheme. Continuous dependence for the solutions on the gravity coefficients and the Soret coefficient are proved. First of all, an a priori bound in L2 norm is derived whereby we show the solution depends continuously in L2 norm on changes in the gravity coefficients and the Soret coefficient. This estimate also implies that the solutions decay exponentially. 1. Introduction Within the context of fluid flow in porous media, or simply within theory of fluid flow, there has been substantial recent interest in deriving stability estimates where changes in coefficient are allowed, or even the model (the equations themselves) changes. This type of stability has earned the name structural stability, and is different from continuous dependence on the initial data. The concept of structural stability in which the study of continuous dependence (or stability) is on changes in the model itself rather than the initial data. Thus structural stability constitutes a class of stability problem every bit important. Structural stability is focus of attention now, and for the relevant results, the reader is referred to [1, 2, 3, 5, 6, 7, 8, 9, 10]. The Brinkman model is believed accurate when the flow velocity is too large for Darcy’s law to be valid, and additionally the porosity is not too small. In this article, we are concerned with structural stability for the Brinkman equations modeling the double diffusive convection. The temperature field is non-constant and a solute is also diffused throughout the porous body. The Brinkman equations governing the flow of fluid in double-diffusive convection are ui,t = ν∆ui − aui − p,i + gi T + hi C T,t + ui T,i = ∆T (1.1) C,t + ui C,i = ∆C + ρ∆T in Ω × t > 0 ui,i = 0 2000 Mathematics Subject Classification. 35B30, 35K55, 35Q35. Key words and phrases. Continuous dependence; structural stability; gravity coefficients; Soret coefficient; Brinkman equations. c 2007 Texas State University - San Marcos. Submitted April 9, 2007. Published June 16, 2007. 1 2 H. TU, C. LIN EJDE-2007/92 where ui , T , C and p represent fluid velocity, temperature, salt concentration and pressure, respectively. The quantities gi (x) and hi (x) are gravity vector terms, and positive constants ρ, a are known the Soret coefficient and the Darcy coefficient, respectively. Ω is a bounded domain of of R3 , and with sufficiently smooth boundary ∂Ω. ∆ is the Laplace operator, k · k and hu, vi denote the norm and inner product on L2 (Ω) . Associated with (1.1), we impose the boundary data ui = 0; T = 0; C = 0 on ∂Ω (1.2) and the initial data ui (x, 0) = zi (x); T (x, 0) = T0 (x); C(x, 0) = C0 (x); x ∈ Ω (1.3) In (1.1) and in the equations throughout, a comma is used to denote partial differ∂ui i entiation: For example ui,i denotes ∂u ∂xi and ui,t denotes ∂t , and we employ the convention of summing over repeated indices from 1 to 3 . 2. A priori bounds Multiplying (1.1)2 by T and integrating over Ω, we have 1 d kT k2 = −k∇T k2 (2.1) 2 dt Multiplying (1.1)3 by C and integrating over Ω, then using arithmetic-geometric mean inequality, we obtain 1 ρ2 1 d kCk2 + k∇Ck2 ≤ k∇T k2 (2.2) 2 dt 2 2 Multiplying (1.1)1 by ui and integrating over Ω, furthermore using CauchySchwarz, arithmetic-geometric mean, then using Poincare’s inequality, we find a 1 1 d kuk2 + νk∇uk2 + kuk2 ≤ [g 2 kT k2 + h2 kCk2 ] 2 dt 2 a (2.3) 1 2 [g k∇T k2 + h2 k∇Ck2 ] ≤ aλ1 where g 2 = max gi gi ; h2 = max hi hi . Ω Ω and λ1 is the first eigenvalue of the problem ∆φ + λφ = 0 in Ω φ = 0 on ∂Ω Multiplying (1.1)1 by ui,t and integrating over Ω , furthermore using CauchySchwarz and arithmetic-geometric mean inequality, then using Poincare’s Poincare’s inequality, we find a d ν d 1 kuk2 + k∇uk2 ≤ [g 2 kT k2 + h2 kCk2 ] 2 dt 2 dt 2 (2.4) 1 2 ≤ [g k∇T k2 + h2 k∇Ck2 ] 2λ1 Multiplying (2.1) by Γ11 , (2.2) by Γ12 and (2.3) by Γ13 , then adding all results to (2.4) leads to d Q1 (t) + G(t) ≤ 0 (2.5) dt EJDE-2007/92 CONTINUOUS DEPENDENCE 3 where Γ1i , (i = 1, 2, 3) are positive constants at our disposal, Γ11 Γ12 (Γ13 + a) ν kT k2 + kCk2 + kuk2 + k∇uk2 2 2 2 2 (2.6) (2Γ11 − ρ2 Γ12 ) 1 1 Γ13 ) aΓ13 − g2 ( + ]k∇T k2 + kuk2 2 λ1 2 a 2 Γ12 1 2 1 Γ13 ) +[ − h ( + ]k∇Ck2 + Γ13 νk∇uk2 2 λ1 1 2 a (2.7) Q1 (t) = and G(t) = [ We can select Γ1i , (i = 1, 2, 3) to secure that all the all the coefficients of (2.7) are positive, such as Γ13 = a ; 2 Γ12 = 4h2 ; λ1 Γ11 > (2h2 ρ2 + g 2 ) . λ1 Thus, with the help of Poincare’s inequality, we can show that 1 Γ13 ) (2Γ11 − ρ2 Γ12 ) − g2 ( + ]kT k2 + Γ13 νk∇uk2 2 2 a 2λ1 Γ12 1 Γ13 a +[ − h2 ( + ]kCk2 + Γ13 kuk2 2 2 a 2 G(t) ≥ [λ1 (2.8) We can easily show that κ1 Q1 (t) ≤ G(t) (2.9) where κ1 is a positive constant represented by κ1 = min 1 2Γ13 ) ], [λ1 (2Γ11 − ρ2 Γ12 ) − g 2 (1 + Γ11 a 1 aΓ13 2Γ13 ) ], , 2Γ13 [Γ12 λ1 − h2 (1 + Γ12 a Γ13 + a Thus, from (2.5), we can derive that d Q1 (t) + κ1 Q1 (t) ≤ 0 dt (2.10) Q1 (t) ≤ Q1 (0)e−κ1 t (2.11) Furthermore, we obtain Therefore, recalling the definition of Q1 (t) and combining (2.5), we can obtain Z t 2 2 2 2 kT k ≤ M1 ; kCk ≤ M1 ; kuk ≤ M1 ; k∇uk ≤ M1 ; kuk2 dη ≤ M1 ; 0 Z t Z t Z t k∇uk2 dη ≤ M1 ; k∇T k2 dη ≤ M1 ; k∇Ck2 dη ≤ M1 . 0 0 0 (2.12) where M1 is a generic positive constant depending on the coefficients of (1.1) and the initial data terms in (1.3). It also follows from inequality (2.11) that k∇uk2 , kT k2 and kCk2 decay exponentially as t tends to ∞. 4 H. TU, C. LIN EJDE-2007/92 3. Continuous dependence on the gravity coefficients Let (ui , T, C, P ) and (u∗i , T ∗ , C ∗ , P ∗ ) be solutions to (1.1) with the same boundary and initial data (1.2), (1.3), but with different gravity coefficients (gi , hi ) and (gi∗ , h∗i ), respectively. Namely, ui,t = ν∆ui − aui − p,i + gi T + hi C T,t + ui T,i = ∆T C,t + ui C,i = ∆C + ρ∆T in Ω × t > 0 ui,i = 0 and u∗i,t = ν∆u∗i − au∗i − p∗,i + gi∗ T ∗ + h∗i C ∗ T,t∗ + u∗i T,i∗ = ∆T ∗ C,t∗ + u∗i C,i∗ = ∆C ∗ + ρ∆T ∗ in Ω × t > 0 u∗i,i = 0 . Now set wi = ui − u∗i , Π = p − p∗ , S = T − T ∗ , γi = gi − gi∗ , Σ = C − C∗ µi = hi − h∗i (3.1) Clearly, (wi , Π, S, Σ) satisfies the equations wi,t = ν∆wi − awi − Π,i + γi T + gi∗ S + µi C + h∗i Σ S,t + wi T,i + u∗i S,i = ∆S Σ,t + wi C,i + u∗i Σ,i = ∆Σ + ρ∆S wi,i = 0 with the boundary-initial data wi = S = Σ = 0 (3.2) in Ω × {t > 0} on ∂Ω × {t > 0} wi (x, 0) = S(x, 0) = Σ(x, 0) = 0 (3.3) in Ω Multiplying (3.2)1 by wi and integrating over Ω, then using Cauchy-Schwarz’s inequality, and the arithmetic-geometric mean inequality, we obtain a 1 d 2 νk∇wk2 + kwk2 + kwk2 ≤ [(g ∗ )2 kSk2 +(h∗ )2 kΣk2 +µ2 kCk2 +γ 2 kT k2 ], (3.4) 2 2 dt a where γ 2 = max γi γi ; Ω µ2 = max ui ; Ω (g ∗ )2 = max gi∗ gi∗ ; Ω (h∗ )2 = max h∗i h∗i . Ω Multiplying (3.2)1 by wi,t and integrating over Ω, then using Cauchy-Schwarz inequality, and the arithmetic-geometric mean inequality, we have 1 d a d ν k∇wk2 + kwk2 ≤ (g ∗ )2 kSk2 + (h∗ )2 kΣk2 + µ2 kCk2 + γ 2 kT k2 (3.5) 2 dt 2 dt Multiplying (3.2)2 by S and integrating over Ω, then using Cauchy-Schwarz, the arithmetic-geometric mean inequality, the Sobolev inequality ,which holds for all ϕ ∈ C01 (ϕ), kϕk4 ≤ Λkϕ,i k 4 where k · k4 is the norm on L (Ω). EJDE-2007/92 CONTINUOUS DEPENDENCE 5 we derive d kSk2 = −2k∇Sk2 + 2hT,i , wi Si; dt therefore d kSk2 ≤ −2k∇Sk2 + 2k∇T k · kwk4 kSk4 dt 1 ≤ −2k∇Sk2 + 2Λ 2 k∇T k · k∇wk · k∇Sk Λ ≤ −2k∇Sk2 + 2α11 k∇Sk2 + k∇wk2 k · k∇T k2 2α11 (3.6) where α11 is a positive constant at our disposal. Using similar method as in (3.6), from equation (3.2)3 , we find that Λ ρ2 d kΣk2 ≤ (−2 + α12 + α13 )k∇Σk2 + k∇wk2 · k∇Ck2 + k∇Sk2 dt α13 α12 (3.7) where α1i , i = 2, 3 are positive constants at our disposal. Multiplying (3.6) by Γ21 and adding (3.7) leads to d (kΣk2 + Γ21 kSk2 ) dt Γ21 Λ 2 ≤ k∇wk2 ( k∇T k2 + k∇Ck2 ) 2 α11 α13 ρ2 + 2[(α11 − 1) + ]Γ21 k∇Sk2 + (−2 + α12 + α13 )k∇Σk2 2α12 Γ21 (3.8) where Γ21 is a positive constant at our disposal. In (3.8), we choose α11 = 1 ; 4 α12 = α13 = 1 ; 2 Γ21 = 4ρ2 . it follows that d (kΣk2 + 4ρ2 kSk2 ) + k∇Σk2 + 4ρ2 k∇Sk2 ≤ 2Λk∇wk2 (4ρ2 k∇T k2 + k∇Ck2 ) (3.9) dt Multiplying (3.4) by Γ22 and (3.5) by Γ23 , then adding all the results to (3.9), then using Poincare’s inequality, we have d Γ23 (Γ22 + aΓ23 ) νk∇wk2 + kwk2 + (kΣk2 + 4ρ2 kSk2 ) dt 2 2 2Γ22 ∗ 2 + λ1 − ( + Γ23 )(g ) kΣk2 a (3.10) 2 2Γ22 aΓ22 + 4ρ λ1 − ( + Γ23 )(h∗ )2 kSk2 + Γ22 νk∇wk2 + kwk2 a 2 2 2 Γ22 2 2 2 2 2 ≤ wk 4ρ k∇T k + kC,i k + ( + Γ23 ) µ kCk + γ kT k2 a where Γ2i , i = 2, 3 are positive constants at our disposal. We can choose Γ22 and Γ23 sufficient small to make sure that all the coefficients in (3.10) are positive, such as Γ22 = min{ aλ1 ρ2 aλ1 , }; 8(g ∗ )2 2(h∗ )2 Γ23 = min{ λ1 ρ2 λ 1 , }. 4(g ∗ )2 (h∗ )2 6 H. TU, C. LIN EJDE-2007/92 Let Q2 (t) = { Γ23 (Γ22 + aΓ23 ) νk∇wk2 + kwk2 + kΣk2 + 4ρ2 kSk2 } 2 2 and κ2 = min 2Γ22 ν aΓ22 2Γ22 ; ; (λ1 − ( + Γ23 )(g ∗ )2 ); Γ23 ν (Γ22 + aΓ23 ) a 1 2Γ22 [4ρ2 λ1 − ( + Γ23 )(h∗ )2 ] 2 4ρ a (3.11) It follows from (3.10) that Γ22 d Q2 (t)+κ2 Q2 (t) ≤ ( +Γ23 )[µ2 kCk2 +γ 2 kT k2 ]+2Λk∇wk2 [4ρ2 k∇T k2 +k∇Ck2 ] dt a (3.12) It is easy to show that 2Λk∇wk2 [4ρ2 k∇T k2 + k∇Ck2 ] ≤ Q2 (t)f2 (t) (3.13) where f2 (t) = τ2 (k∇T k2 + k∇Ck2 ), τ2 = 4Λ (4ρ2 + 1). Γ23 ν Thus, combining (3.12) and (3.13), we obtain d Q2 (t) + κ2 Q2 (t) ≤ M2 (µ2 + γ 2 ) + f2 (t)Q2 (t) dt (3.14) where M2 = 2( Γa22 + Γ23 )M1 . Now, multiplying both sides of (3.14) by eκ2 t and integrating from 0 and t, we get Z t M2 2 (µ + γ 2 ) + Q2 (t) ≤ f2 (η)Q2 (η)dη (3.15) κ2 0 Hence, applying Gronwall’s lemma and using (2.12), we obtain Q2 (t) ≤ M2 R t f2 (η)dη M2 2τ2 M1 e0 · (µ2 + γ 2 ) ≤ e · (µ2 + γ 2 ), ∀t > 0 κ2 κ2 (3.16) Consequently, from inequality (3.16), we can see that Q2 (t) → 0, as gi → gi∗ and hi → h∗i . Recalling the definition of Q2 (t), so (wi , Π, S, Σ) → 0 and (ui , T, C) → (u∗i , T ∗ , C ∗ ). So, continuous dependence on the gravity coefficients is proved. 4. Continuous dependence on the Soret coefficient Let (ui , T, C, P ) and (u∗i , T ∗ , C ∗ , P ∗ ) be solutions to (1.1 ), with the same boundary and initial data (1.2), (1.3), but with different Soret coefficient ρ and ρ∗ , respectively: ui,t = ν∆ui − aui − p,i + gi T + hi C T,t + ui T,i = ∆T C,t + ui C,i = ∆C + ρ∆T in Ω × t > 0 ui,i = 0 EJDE-2007/92 CONTINUOUS DEPENDENCE 7 and u∗i,t = ν∆u∗i − au∗i − p∗,i + gi T ∗ + hi C ∗ T,t∗ + u∗i T,i∗ = ∆T ∗ C,t∗ + u∗i C,i∗ = ∆C ∗ + ρ∗ ∆T ∗ u∗i,i in Ω × t > 0 =0 Now set wi = ui − u∗i , Π = p − p∗ , S = T − T ∗, Σ = C − C∗ (4.1) Clearly, (wi , Π, S, Σ) satisfies the equations wi,t = ν∆wi − awi − Π,i + +gi S + +hi Σ S,t + wi T,i + u∗i S,i = ∆S Σ,t + wi C,i + u∗i Σ,i = ∆Σ + (ρ − ρ∗ )∆T + ρ∗ ∆S wi,i = 0 in Ω × {t > 0} (4.2) with the boundary-initial data wi = S = Σ = 0 on ∂Ω × {t > 0} wi (x, 0) = S(x, 0) = Σ(x, 0) = 0 in Ω (4.3) Multiplying (4.2)1 by wi and integrating over Ω, furthermore using the CauchySchwarz’s inequality, and the arithmetic-geometric mean inequality, we have 1 d 1 a kwk2 ≤ [g 2 kSk2 + h2 kΣk2 ] νk∇wk2 + kwk2 + 2 2 dt a (4.4) where g 2 = max gi gi , h2 = max hi hi . Ω Ω Multiplying (4.2)1 by wi,t and integrating over Ω, then using the Cauchy-Schwarz inequality, and the arithmetic-geometric mean inequality, we get 1 d a d 1 ν k∇wk2 + kwk2 ≤ [g 2 kSk2 + h2 kΣk2 ] 2 dt 2 dt 2 (4.5) Multiplying (4.2)2 by S and integrating over Ω, then using Cauchy-Schwarz, the arithmetic-geometric mean inequality, and using the Sobolev inequality, we can derive d Λ kSk2 dη ≤ −2k∇Sk2 + 2α21 k∇Sk2 + k∇wk2 k∇T k2 (4.6) dt 2α21 for α21 is a positive constant at our disposal. Using similar method as in (4.6), from equation (4.2)3 , we can find that d Λ kΣk2 ≤ (−2 + α22 + α23 + α24 )k∇Σk2 + k∇wk2 k∇Ck2 dt α23 ρ∗2 (ρ − ρ∗ )2 + k∇Sk2 + k∇T k2 α22 α24 for α2i , (i = 2, 3, 4) are positive constants at our disposal. (4.7) 8 H. TU, C. LIN EJDE-2007/92 Multiplying (4.6) by Γ31 and adding (4.7), leads to d (kΣk2 + Γ31 kSk2 ) dt ρ∗2 Γ31 k∇Sk2 2α22 Γ31 Γ31 (ρ − ρ∗ )2 Λ 2 + k∇T k2 + k∇wk2 k∇T k2 + k∇Ck2 α24 2 α21 α23 ≤ (−2 + α22 + α23 + α24 )k∇Σk2 + 2 (α21 − 1) + (4.8) where Γ31 is a positive constant at our disposal. In (4.8), we choose α21 = 1/2, α22 = 1, α23 = α24 = 1/4, It follows that d (kΣk2 + Γ31 kSk2 ) + k∇Σk2 + (Γ31 − ρ∗2 )k∇Sk2 dt ≤ Λk∇wk2 (Γ31 k∇T k2 + 4k∇Ck2 ) + 4(ρ − ρ∗ )2 k∇T k2 (4.9) Multiplying (4.4) by Γ32 and (4.5) by Γ33 , furthermore adding all the results to (4.9), we get d Γ33 (Γ32 + aΓ33 ) [ νk∇wk2 + kwk2 + (kΣk2 + Γ4 kSk2 )] dt 2 2 aΓ32 + Γ32 νk∇wk2 + kwk2 + k∇Σk2 + (Γ31 − ρ∗2 )k∇Sk2 2 2Γ32 ≤( + Γ33 )[g 2 kΣk2 + h2 kSk2 ] + 2Λk∇wk2 [Γ31 k∇T k2 a + 4k∇Ck2 ] + 4(ρ − ρ∗ )2 k∇T k2 (4.10) where Γ3i , (i = 2, 3) are positive constants at our disposal. We can choose Γ31 to make sure that Γ31 > ρ∗2 , therefore, by the Poincare’s inequality from (4.10), we have (Γ32 + aΓ33 ) d Γ33 νk∇wk2 + kwk2 + (kΣk2 + Γ31 kSk2 ) dt 2 2 aΓ32 Γ32 + Γ33 )h2 kSk2 + kwk2 + λ1 (Γ31 − ρ∗2 ) − (2 a 2 2Γ32 + Γ32 νk∇wk2 + λ1 − ( + Γ33 )g 2 kΣk2 a ≤ Λk∇wk2 [Γ31 k∇T k2 + 4k∇Ck2 ] + 4(ρ − ρ∗ )2 k∇T k2 (4.11) We can choose Γ32 and Γ33 sufficient small to make sure that all the coefficients in (4.11) are positive, such as we may choose Γ32 = min{ aλ1 aλ1 (Γ31 − ρ∗2 ) , }; 8g 2 8h2 Γ33 = min{ λ1 λ1 (Γ31 − ρ∗2 ) , }. 4g 2 4h2 Let Q3 (t) = { Γ33 (Γ32 + aΓ33 ) νk∇wk2 + kwk2 + kΣk2 + Γ31 kSk2 } 2 2 and κ3 = min{ 2Γ32 aΓ32 2Γ32 ; ; (λ1 − ( + Γ33 )g 2 ); Γ33 (Γ32 + aΓ33 ) a 1 2Γ32 [λ1 (Γ31 − ρ∗2 ) − ( + Γ33 )h2 ]} Γ31 a (4.12) EJDE-2007/92 CONTINUOUS DEPENDENCE Then, it follows from (4.11) that d Q3 (t) + κ3 Q3 (t) ≤ Λkwk2 [Γ31 k∇T k2 + 4kC,i k2 ] + 4(ρ − ρ∗ )2 k∇T k2 dt Also, it can be easily shown that Λk∇wk2 [Γ31 k∇T k2 + 4k∇Ck2 ] ≤ f3 (t)Q3 (t) where f3 (t) = τ3 (k∇T k2 + k∇Ck2 ), τ3 = 9 (4.13) (4.14) 2Λ (Γ31 + 4). Γ33 ν Thus, combing (4.13) and (4.14), we obtain d Q3 (t) + κ3 Q3 (t) ≤ 4(ρ − ρ∗ )2 k∇T k2 + f3 (t)Q3 (t) (4.15) dt Now, multiplying both sides of (4.15) by eκ3 t and integrating over [0, t], we obtain Q3 (t) ≤ M3 (ρ − ρ∗ )2 + κ3 Z t f3 (η)Q3 (η)dη (4.16) 0 where M3 = 4M1 . Hence, applying Gronwall’s lemma and using (2.12), we obtain M3 2τ3 M1 M3 R t f3 (η)dη · (ρ − ρ∗ )2 ≤ Q3 (t) ≤ e0 e · (ρ − ρ∗ )2 , ∀t > 0 (4.17) κ3 κ3 As a result, from inequality (4.17), we can see that Q3 (t) → 0, as ρ → ρ∗ . Recalling the definition of Q3 (t), so (wi , Π, S, Σ) → 0 and (ui , T, C) → (u∗i , T ∗ , C ∗ ). Consequently, continuous dependence on the Soret coefficient is proved. Acknowledgments. The authors would like to express their gratitude to Professor Yan Liu for his help in writing parts of this article. References [1] K. A. Ames, B. Straughan; Non-standard and improperly posed problems. Mathematics in Science and Engineering Series, Vol. 194, San Diego Academic Press, 1997. [2] F. Franchi, B. Straughan; Continuous dependence and decay for the Forchheimer equations. Proc.Roy.Soc.London A, 459 (2003), 3195–3202. [3] J. Guo, Y. Qin; Steady nonlinear double-diffusive convection in a porous medium base upon the Brinkman-Forchheimer model. Journal of Mathematical Analysis and Applications, 204 (1996), 138–155. [4] P. N. Kaloni, Y. Qin; Spatial decay estimates for plane flow in the Darcy-Forchheimer model. Quarterly of Applied Mathematics, 56 (1998), 71–87. [5] L. E. Payne, B. Straughan; Convergence and continuous dependence for the BrinkForchheimer equations. Studies in Applied Mathematics, 102 (1999), 419–439. [6] L. E. Payne, B. Straughan; Stability in the initial-time geometry problem for Brinkman and Darcy equations of flow in a porous media. Journal de Mathematics Pureset Appliquees, 75 (1996), 225–271. [7] L. E. Payne, B. Straughan; Structural stability for the Darcy equations of flow in porous media. Proceedings of the royal Society London.Series A, 454 (1998), 1691–1698. [8] L. E. Payne, J. C. Song, B. Straughan; Continuous dependence and convergence results for Brinkman and Forchheimer models with variable viscosity. Proceedings of the royal Society London. Series A, 455 (1999), 2173–2190. [9] L. E. Payne, J. C. Song, B. Straughan; Effect of anisotropic permeability on Darcy’Law. Math. Meth. Appl. Sci, 24 (2001), 427–438. [10] B. Straughan, K. Hutter; A priori bounds and structural stability for double diffusive convection incorporating the Soret effect. Proceedings of the Royal Society London.Series A, 455 (1999), 767–777. 10 H. TU, C. LIN EJDE-2007/92 Hongliang Tu Department of Applied Mathematics, Zhuhai college of Jilin University, 519041, China E-mail address: tuhongliang00@126.com Changhao Lin School of Mathematical Sciences, South China Normal University, 510631, China E-mail address: linchh@scnu.edu.cn