Electronic Journal of Differential Equations, Vol. 2010(2010), No. 75, pp. 1–15. ISSN: 1072-6691. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu ftp ejde.math.txstate.edu REACTION-DIFFUSION SYSTEM OF EQUATIONS IN NON-STATIONARY MEDIUM AND ARBITRARY NON-SMOOTH DOMAINS SIKIRU ADIGUN SANNI Abstract. A system of non-linear partial differential equations describing one-step irreversible reaction, reactant to product, in a non-stationary medium and non-smooth domain is considered. After obtaining the necessary a priori estimates, the existence of a unique local strong solution to the system is proved using a fixed point theorem. 1. Introduction We consider the semilinear parabolic system of partial differential equations ∇.v̄ = 0 in ΩT (1.1) ∂v̄ 1 − ν∆v̄ = −∇.(v̄ ⊗ v̄) − ∇p in ΩT ∂t ρ ∂u − k∆u = −∇.(v̄u) + Qwf (u) in ΩT ∂t ∂w − d∆w = −∇.(v̄w) − wf (u) in ΩT ∂t v̄ = 0̄, u = w = 0 on ∂Ω × [0, T ) v̄(x, 0) = v̄0 (x), u(x, 0) = u0 (x), w(x, 0) = w0 (x) (1.2) (1.3) (1.4) (1.5) (1.6) where 0̄ is the zero vector in R3 , ⊗ is the matrix multiplication defined by the tensor v̄ ⊗ v̄ := vi vj (i, j = 1, 2, 3) and ΩT := Ω × [0, T ). Notice then that ∇.(v̄ ⊗ v̄) = ∂vj ∂ ∂ ∂xi (v̄i v̄j ) = ∂xj (v̄i v̄j ) = vi ∂xi = v̄.∇v̄ (using (1.1)). In applications, the system models a single-step irreversible reaction, reactant → product in non-stationary incompressible medium. v̄(x, t) is the velocity of the medium; ν and ρ are the kinematic viscosity and the density of the medium respectively. u(x, t) is the temperature in the reaction vessel, w(x, t) is the mass fraction of the reactant, 1−w(x, t) is the mass fraction of the product, k the positive thermal conductivity and d the reactant diffusivity. Qwf (u) and −wf (u) are the reaction kinetics, determined by a positive, uniformly bounded and differentiable function 2000 Mathematics Subject Classification. 35B40, 35K57, 80A25. Key words and phrases. Irreversible reaction; reactant diffusivity; thermal conductivity; a priori estimates; Banach’s fixed point theorem. c 2010 Texas State University - San Marcos. Submitted November 27, 2009. Published May 21, 2010. 1 2 S. A. SANNI EJDE-2010/75 f (u). Furthermore, f 0 (u) is assumed to be Lipschitz continuous. It is assumed that Ω is an open and bounded arbitrary non-smooth domain in R3 . Theoretically, the reactant decomposes at a rate which is proportional to w(x, t)f (u), where f (u) is the approximate number of molecules that have sufficient energy for the reaction to begin. In this paper, we shall assume that 0 ≤ f (u) ≤ B 0 0 |f (u)| ≤ B , 0 (1.7) 0 |f (u) − f (ũ)| ≤ L|u − ũ| (1.8) For further information on chemical kinetics and combustion, the reader is referred to Buckmaster[3], Buckmaster and Ludford [4], and Frank-Kamenetskii [9]. Several combustion models assumed some smoothness on the boundary vis-a-vis stationary media. Authors of these models include Avrin [1, 2], Daddiouaissa [5], De Oliviera et al [6], Fitzgibbon and Martin [8], Henry [10], Konach [11], Sanni [13], Sattinger [14], and some literature cited in them. In this paper, we establish the existence of a unique local-in-time strong solution to the system (1.1)-(1.6), in arbitrary non-smooth domains. Clearly, the inclusion of the Navier-Stokes equations in the system implies that the medium is non-stationary. Using Leray projector [15], the problem (1.1)-(1.6) can be reduced to that of finding only (v̄, u, w) by a variational formulation. We are thus motivated to define: Definition 1.1. We call a solution (v̄, u, w) of the system (1.1)-(1.6) a strong solution, provided (v̄, u, w) ∈ X 3 , where X is defined by X := L∞ [0, T ; H01 (Ω)] ∩ H 1 [0, T ; H01 (Ω)] ∩ W 1,∞ [0, T ; L2 (Ω)] (1.9) 2. A priori estimates We will need the following Sobolev embedding theorem, stated and proved in [7, pp. 265-266]. Theorem 2.1. Assume that Ω ⊂ Rn is open and bounded. Suppose U ∈ W01,p (Ω) for some 1 ≤ p < n. Then we have the estimate kU kLq (Ω) ≤ Ck∇U kLp (Ω) (2.1) for each q ∈ [1, p∗], the constant C depending only on p, q, n and Ω, where p∗ := is the Sobolev conjugate. np n−p Notice that the hypothesis of Theorem 2.1 requires no smoothness assumption on the boundary Ω. We now set out to obtain a priori estimates required to prove the existence of a unique local strong solution to the system (1.1)-(1.6). We first state and prove the following Lemmas. Lemma 2.2. Let u ∈ H 1 (Ω) and v, w, p ∈ H01 (Ω). Then Z (2.2) uwp dx ≤ kuk2L2 (Ω) + C(Ω)−1 kwk2H 1 (Ω) kpk2H 1 (Ω) 0 0 Ω Z uwp dx ≤ kuk2L2 (Ω) + kpk2H 1 (Ω) + C(Ω)−3 kwk4H 1 (Ω) kpk2L2 (Ω) (2.3) 0 0 Z Ω uwp dx ≤ kuk2L2 (Ω) + C(Ω)−1 kwk2H 1 (Ω) T −1/2 kpk2L2 (Ω) + T 1/2 kpk2H 1 (Ω) Ω 0 0 (2.4) EJDE-2010/75 REACTION-DIFFUSION SYSTEM 3 Z u (pw − p̃w̃) dx Ω h i (2.5) ≤ kuk2L2 (Ω) + C(Ω)−1 kpk2H 1 (Ω) kw − w̃k2H 1 (Ω) + kp − p̃k2H 1 (Ω) kw̃k2H 1 (Ω) 0 0 0 0 Z u (pw − p̃w̃) dx Ω h ≤ kuk2L2 (Ω) + C(Ω)−1 kpk2H 1 (Ω) T −1/2 kw − w̃k2L2 (Ω) 0 i + T 1/2 kw − w̃k2H 1 (Ω) + kp − p̃k2H 1 (Ω) T −1/2 kw̃k2L2 (Ω) + T 1/2 kw̃k2H 1 (Ω) 0 0 0 (2.6) Z uvwp dx ≤ k∇uk2L2 (Ω) + C(Ω)−1 kvk4H 1 (Ω) kuk2L2 (Ω) 0 Ω (2.7) + C(Ω) T −1/2 kpk2L2 (Ω) + T 1/2 kpk2H 1 (Ω) kwk2H 1 (Ω) 0 0 Z u(vwp − v̄˜w̃p̃)dx Ω ≤ C(Ω)−1 kvk4H 1 (Ω) + kw̃k4H 1 (Ω) kuk2L2 (Ω) 0 0 h + k∇uk2L2 (Ω) + C(Ω) kwk2H 1 (Ω) T −1/2 kp − p̃k2L2 (Ω) + T 1/2 kp − p̃k2H 1 (Ω) 0 0 i −1/2 2 1/2 2 2 2 + T kp̃kL2 (Ω) + T kp̃kH 1 (Ω) kw − w̃kH 1 (Ω) + kv − v̄˜kH 1 (Ω) 0 0 0 (2.8) Proof. 1. Proof of (2.2). By Hölder’s inequality, Z uwp dx ≤ kukL2 (Ω) kwkL4 (Ω) kpkL4 (Ω) Ω (2.9) ≤ C(Ω)kukL2 (Ω) kwkH01 (Ω) kpkH01 (Ω) by Sobolev embedding theorem. Then (2.2) follows easily from (2.9) by Cauchy’s inequality with . 2. Proof of (2.3) and (2.4). Z Z 1 p2 w2 dx (by Cauchy’s inequality with ) uwp dx ≤ kuk2L2 (Ω) + 4 Ω Ω 1 2 ≤ kukL2 (Ω) + kpkL2 (Ω) kpkL6 (Ω) kwk2L6 (Ω) (by Hölder’s inequality) 4 ≤ kuk2L2 (Ω) + C(Ω)(4)−1 kpkL2 (Ω) kpkH01 (Ω) kwk2H 1 (Ω) , 0 (2.10) by Sobolev embedding theorem. Then (2.3) and (2.4) follow by applying Cauchy’s inequality with 2 and T 1/2 , respectively, to the appropriate factors of the second term on the right side of (2.10). 3. Proof of (2.5) and (2.6). Z Z Z u(pw − p̃w̃)dx = up(w − w̃)dx + uw̃(p − p̃)dx. (2.11) Ω Ω Ω Then (2.5) and (2.6) follows by applying (2.2) and (2.4) to (2.11) respectively. 4 S. A. SANNI EJDE-2010/75 4. Proof of (2.7). By Young’s inequality and then by Hölder’s inequality, Z Z Z 1 1 uvwp dx ≤ u2 v 2 dx + w2 p2 dx 2 2 Ω Ω Ω 1 (2.12) ≤ kukL2 (Ω) kukL6 (Ω) kvk2L6 (Ω) + kwk2L6 (Ω) kpkL2 (Ω) kpkL6 (Ω) 2 ≤ kukL2 (Ω) kukH01 (Ω) kvk2H 1 (Ω) + kwk2H 1 (Ω) kpkL2 (Ω) kpkH01 (Ω) , 0 0 by By Sobolev embedding theorem. (2.7) follows by applying Cauchy’s inequalities with and T 1/2 to the first and second terms on the right side of (2.12) respectively. 5. Proof of (2.8). Z u(vwp − v̄˜w̃p̃)dx Ω Z Z Z (2.13) = uvw(p − p̃)dx + uv p̃(w − w̃)dx + up̃w̃(v − v̄˜)dx . Ω Ω Ω Then (2.8) follows by applying (2.7) to the each term on the right side of (2.13). This concludes the proof of Lemma 2.2 Lemma 2.3. Let (1.1)-(1.4) hold. Suppose v̄0 , u0 , w0 ∈ H01 (Ω) ∩ H 2 (Ω), then k∂t v̄0 k2L2 (Ω) + k∂t u0 k2L2 (Ω) + k∂t w0 k2L2 (Ω) ≤ C(k∇v̄0 k1H (Ω)2 + k∇u0 k1H (Ω)2 + k∇w0 k1H (Ω)2 )(1 + kv̄0 k2H 1 (Ω) ), (2.14) 0 where C = C(ν, k, d, B, ρ, Ω, Q) Proof. Taking (1.1) and (1.3) on t = 0 and multiplying the corresponding equation to (1.3) by ∂t u0 , we estimate Z |∂t u0 |2 dx Ω Z Z Z =− ∂t u0 .v̄0 .∇u0 dx + k ∂t u0 ∆u0 dx + Q ∂t u0 w0 f (u0 )dx Ω Ω Ω Z Z Z 1 QB |w0 |2 dx + k |∆u0 |2 dx ≤ 2 |∂t u0 |2 dx + (2.15) 4 0 Ω Ω (Integrating by parts, using Cauchy’s inequality with and (1.7)) Z C(Q, B, k, Ω) ≤ 2 |∂t v̄0 |2 dx + Ω 2 1 × kv̄0 kH 1 (Ω) k∇u0 kH (Ω)2 + kw0 k2H 1 (Ω) + k∇2 u0 k2L2 (Ω) , 0 0 by Hölder and Poincare’s inequalities and using that k∆v̄0 kL2 (Ω) ≤ k∇2 v̄0 kL2 (Ω) . Choosing > 0 sufficiently small and simplifying, we deduce i h (2.16) k∂t u0 k2L2 (Ω) ≤ C(Q, B, k, Ω) (1 + kv̄0 k2H 1 (Ω) )k∇u0 k1H (Ω)2 + kw0 k2H 1 (Ω) 0 0 Evaluating (1.1), (1.2) and (1.4) at t = 0, we obtain analogous estimates to (2.16), viz: k∂t w0 k2L2 (Ω) k∂t v̄0 k2L2 (Ω) ≤ C(ν, Ω)(1 + kv̄0 k2H 1 (Ω) )k∇v̄0 k1H (Ω)2 0 i h 2 ≤ C(d, B, Ω) (1 + kv̄0 kH 1 (Ω) )k∇w0 k1H (Ω)2 + kw0 k2H 1 (Ω) 0 Combining (2.16), (2.17) and (2.18), we deduce (2.14). 0 (2.17) (2.18) EJDE-2010/75 REACTION-DIFFUSION SYSTEM 5 Theorem 2.4. Let v̄0 , u0 , w0 ∈ H01 (Ω) ∩ H 2 (Ω). Suppose (v̄, u, w) is a strong solution of the system (1.1)-(1.6). Then we have the estimate sup k∂t v̄k2L2 (Ω) + kv̄k2H 1 (Ω) + k∂t uk2L2 (Ω) + kuk2H 1 (Ω) + k∂t wk2L2 (Ω) + kwk2H 1 (Ω) 0 [0,T ] 0 0 + k∇ (∂t v̄) k2L2 [0,T ;L2 (Ω)] + k∇(∂t u)k2L2 [0,T ;L2 (Ω)] + k∇ (∂t w) k2L2 [0,T ;L2 (Ω)] CT [(1 + kv̄0 k2H 1 (Ω) )(G(v̄0 , u0 , w0 ) + 1)]3 0 ≤ {1 − 2CT [(1 + kv̄0 k2H 1 (Ω) )(G(v̄0 , u0 , w0 ) + 1)]2 }3/2 0 =: Σ = constant, (2.19) for T < {2C[(1 + kv̄0 k2H 1 (Ω) )(k∇v̄0 k1H (Ω)2 + k∇u0 k1H (Ω)2 + k∇w0 k1H (Ω)2 ) + 1]2 }−1 , 0 (2.20) where G(v̄0 , u0 , w0 ) = k∇v̄0 k1H (Ω)2 + k∇u0 k1H (Ω)2 + k∇w0 k1H (Ω)2 and C = C(ν, k, d, Q, B, B 0 , Ω). We will use (1.3) and the corresponding conditions in (1.5) and (1.6) to obtain estimates for u; and thereafter, for brevity, state analogous estimates for v̄ and w. Proof. 1. Multiplying (1.3) by ∂t u, integrating the ensuing equation by parts over Ω and using (1.1) and (1.5), we deduce Z k d 2 |∂t u| dx + |∇u| dx 2 dt Ω Ω Z Z = ∇(∂t u).(v̄u)dx + Q ∂t uwf (u)dx Z 2 Ω Ω (2.21) 1 ≤ k∇(∂t u)k2L2 (Ω) + C(Ω) kv̄k1H (Ω)2 kuk1H (Ω)2 + Q2 B 2 kwk2H 1 (Ω) 0 + k∂t uk2L2 (Ω) , using (2.2)) of Lemma 2.2. Simplifying, (2.21) yields d k kuk2H 1 (Ω) 0 dt 2 1 ≤ k∇(∂t u)k2L2 (Ω) + C(Ω) kv̄k1H (Ω)2 kuk1H (Ω)2 + Q2 B 2 kwk2H 1 (Ω) 0 (2.22) 2. Differentiating (1.3) with respect to t yields ∂ (∂t u) − k∆(∂t u) = −∂t v̄.∇u + v̄.∇(∂t u) + Q∂t w.f (u) + Qw∂t uf 0 (u) ∂t (2.23) 6 S. A. SANNI EJDE-2010/75 Multiply by ∂t u and integrating by parts over Ω and use (1.1), (1.5) to deduce: 1 d (k∂t uk2L2 (Ω) ) + kk∇(∂t u)k2L2 (Ω) 2 dtZ Z = ∇(∂t u).∂t v̄.udx + ∇(∂t u).v̄.∂t udx Ω Z Z 0 +Q ∂t u∂t uwf (u)dx + Q ∂t u∂t wf (u)dx Ω Ω (2.24) Ω ≤ k∇(∂t u)k2L2 (Ω) + k∇(∂t v̄)k2L2 (Ω) + C(Ω)−3 k∂t v̄k2L2 (Ω) kuk4H 1 (Ω) 0 k∇(∂t u)k2L2 (Ω) + QB + 0 + k∇(∂t u)k2L2 (Ω) [k∂t uk2L2 (Ω) QB(k∂t uk2L2 (Ω) + + −3 + C(Ω) k∇(∂t u)k2L2 (Ω) k∂t uk2L2 (Ω) kvk4H 1 (Ω) 0 + C(Ω)−3 k∂t uk2L2 (Ω) kwk4H 1 (Ω) ] 0 k∂t wk2L2 (Ω) ), using (2.3) of Lemma 2.2 and Young’s inequality. 3. Combining (2.22) and (2.24) we deduce d (k∂t uk2 2 + kkuk2H 1 (Ω) ) + 2kk∇(∂t u)k2L2 (Ω) 0 dt h L (Ω) ≤ C k∇(∂t u)k2L2 (Ω) + k∇(∂t v̄)k2L2 (Ω) + k∂t uk2L2 (Ω) + 1 + −1 + −3 1 + k∂t v̄k2L2 (Ω) kv̄k2H 1 (Ω) + k∂t uk2L2 (Ω) 0 3 i 2 2 2 + kukH 1 (Ω) + k∂t wkL2 (Ω) + kwkH 1 (Ω) 0 (2.25) 0 0 where C = C(Q, B, B , Ω). 4. Following steps 1-3 in respect of (1.1), (1.2), (1.4) and the corresponding conditions in (1.5), we obtain analogous estimates to (2.25): d (k∂t v̄k2L2 (Ω) + νkv̄k2H 1 (Ω) ) + 2νk∇(∂t v̄)k2L2 (Ω) 0 dt h ≤ C(Ω) k∇(∂t v̄)k2L2 (Ω) + (−1 + −3 ) k∂t v̄k2L2 (Ω) + kv̄k2H 1 (Ω) i 0 (2.26) , d (k∂t wk2L2 (Ω) + dkwk2H 1 (Ω) ) + 2dk∇(∂t w)k2L2 (Ω) 0 dt h ≤ C k∇(∂t w)k2L2 (Ω) + k∇(∂t v̄)k2L2 (Ω) + k∂t uk2L2 (Ω) + (1 + −1 + −3 ) 1 + k∂t v̄k2L2 (Ω) + kv̄k2H 1 (Ω) + k∂t uk2L2 (Ω) 0 3 i 2 2 2 + kukH 1 (Ω) + k∂t wkL2 (Ω) + kwkH 1 (Ω) , 0 (2.27) 0 0 where C = C(B, B , Ω). 5. Combining (2.25)-(2.27), choosing > 0 sufficiently small and simplifying, we deduce d k∂t v̄k2L2 (Ω) + kv̄k2H 1 (Ω) + k∂t uk2L2 (Ω) + kuk2H 1 (Ω) + k∂t wk2L2 (Ω) 0 0 dt + kwk2H 1 (Ω) + k∇(∂t v̄)k2L2 (Ω) + k∇(∂t u)k2L2 (Ω) + k∇(∂t w)k2L2 (Ω) 0 ≤ C(ν, k, d, Q, B, B 0 , Ω) 1 + k∂t v̄k2L2 (Ω) + kv̄k2H 1 (Ω) + k∂t uk2L2 (Ω) 0 EJDE-2010/75 REACTION-DIFFUSION SYSTEM + kuk2H 1 (Ω) + k∂t wk2L2 (Ω) + kwk2H 1 (Ω) 0 7 3 0 Solving the above equation, maximizing the left and right sides of the ensuing inequalities and using Lemma 2.3 concludes the proof of the Theorem 2.1. 3. Existence of a Solution We prove the existence of a unique local strong solution to the system (1.1)-(1.6), in a subset K of the space X 3 equipped with the norm k(η, ξ, ζ)kX 3 h ≤ kηk2L∞ [0,T ;H 1 (Ω)] + k∂t ηk2L∞ [0,T ;L2 (Ω)] + kξk2L∞ [0,T ;H 1 (Ω)] 0 0 + k∂t ξk2L∞ [0,T ;L2 (Ω)] + kζk2L∞ [0,T ;H 1 (Ω)] + k∂t ζk2L∞ [0,T ;L2 (Ω)] 0 + k∇(∂t η)k2L2 [0,T ;L2 (Ω)] + k∇(∂t ξ)k2L2 [0,T ;L2 (Ω)] + k∇(∂t ζ)k2L2 [0,T ;L2 (Ω)] i 12 , (3.1) where X is defined by (1.9). Theorem 3.1. Let v̄0 , u0 and w0 ∈ H01 (Ω) ∩ H 2 (Ω). Then there exists a unique local strong solution to the system (1.1)-(1.6). Proof. 1. The fixed point arguments for the system (1.1)-(1.6) are ∇.Q̄ = 0 in ΩT (3.2) ∂ Q̄ 1 − ν∆Q̄ = −∇.(v̄ ⊗ v̄) − ∇Y in ΩT ∂t ρ ∂R − k∆R = −∇.(v̄u) + Qwf (u) in ΩT ∂t ∂S − d∆S = −∇.(v̄w) − wf (u) in ΩT ∂t Q̄ = 0̄, R = S = 0 on ∂Ω × [0, T ) Q̄(x, 0) = v¯0 (x), R(x, 0) = u0 (x), S(x, 0) = w0 (x), (3.3) (3.4) (3.5) (3.6) (3.7) where Y is the pressure distribution corresponding to the solution (Q̄, R, S). 2. We next define a mapping τ : X3 → X3 (3.8) by setting τ [(v̄, u, w)] = (Q̄, R, S), whenever (Q̄, R, S) is derived from (v̄, u, w) via (3.2)-(3.7). We will prove that for sufficiently small T > 0, τ is a contraction mapping. Choose (v̄, u, w), (v̄˜, ũ, w̃) ∈ X 3 and define τ [(v̄, u, w)] = (Q, R, S), ˜ , R̃, S̃). τ [(v̄˜, ũ, w̃)] = (Q̄ ˜ , R̃, S̃) of the system (3.2)-(3.7), we have Thus, for two solutions (Q, R, S), and (Q̄ ˜) = 0 ∇.(Q̄ − Q̄ in ΩT ∂ ˜ ) − ν∆(Q̄ − Q̄ ˜ ) = −∇.(v̄ ⊗ v̄ − v̄˜ ⊗ v̄˜) − 1 ∇(Y − Ỹ ) in Ω (Q̄ − Q̄ T ∂t ρ ∂ (R − R̃) − k∆(R − R̃) = −∇.(v̄u − v̄˜ũ) + Q(wf (u) − w̃f (ũ)) in ΩT ∂t (3.9) (3.10) (3.11) 8 S. A. SANNI EJDE-2010/75 ∂ (S − S̃) − d∆(S − S̃) = −∇.(v̄w − v̄˜w̃) − (wf (u) − w̃f (ũ)) in ΩT ∂t ˜ = 0̄, R − R̃ = S − S̃ = 0 on ∂Ω × [0, T ) Q̄ − Q̄ ˜ )(x, 0) = 0̄, (R − R̃)(x, 0) = 0, (S − S̃)(x, 0) = 0 (Q̄ − Q̄ (3.12) (3.13) (3.14) 3. Multiplying (3.11) by ∂t (R − R̃), integrating the ensuing equation by parts over Ω, using (3.13) and applying (2.5) of Lemma 2.2, we deduce k d k∇(∂t (R − R̃))k2L2 (Ω) k∂t (R − R̃)k2L2 (Ω) + 2 dt Z Z ˜ = ∇ ∂t (R − R̃) .(v̄u − v̄ ũ)dx + Q ∂t (R − R̃)(wf (u) − w̃f (ũ))dx Ω Ω ≤ k∇(∂t (R − R̃))k2L2 (Ω) + k∂t (R − R̃)k2L2 (Ω) + C(Ω)−1 kv̄k2H 1 (Ω) ku − ũk2H 1 (Ω) + kũk2H 1 (Ω) kv − v̄˜k2H 1 (Ω) 0 0 0 0 0 −1 2 2 + C(Q, B, B , Ω) kwkH 1 (Ω) ku − ũkH 1 (Ω) + kw − w̃k2H 1 (Ω) , 0 0 (3.15) 0 where we have used some bounds in (1.7) and (1.8). 4. Further, we differentiate (3.11) with respect t to get ∂ (∂t (R − R̃)) − k∆(∂t (R − R̃)) ∂t = −∇. (∂t v̄u − ∂t v̄˜ũ + v̄∂t u − v̄˜∂t ũ) Q ∂t wf (u) − ∂t w̃f (ũ) + w∂t uf 0 (u) − w̃∂t ũf 0 (ũ) (3.16) Multiplying (3.16) by ∂t (R − R̃), integrating by parts over Ω, and applying Young’s inequality with , (2.6) and (2.8) as appropriate, we deduce 1 d (k∂t (R − R̃)k2L2 (Ω) ) + kk∇(∂t (R − R̃))k2L2 (Ω) 2 dtZ (∂t v̄u − ∂t v̄˜ũ + v̄∂t u − v̄˜∂t ũ).∇(∂t (R − R̃))dx Z +Q ∂t (R − R̃).(∂t wf (u) − ∂t w̃f (ũ) + w∂t uf 0 (u) − w̃∂t ũf 0 (ũ))dx Ω ≤ 3k∇(∂t (R − R̃))k2L2 (Ω) + 2 + C(Ω)−1 kwk4H 1 (Ω) k∂t (R − R̃)k2L2 (Ω) 0 nh 0 −1 −1/2 2 + C(Ω, B, B , Q, L) T k∂t v̄˜kL2 (Ω) + k∂t ũk2L2 (Ω) + k∂t wk2L2 (Ω) + k∂t uk2L2 (Ω) + T 1/2 k∇(∂t v̄˜)k2L2 (Ω) + k∇(∂t ũ)k2L2 (Ω) + k∇(∂t w)k2L2 (Ω) i + k∇(∂t u)k2L2 (Ω) ku − ũk2H 1 (Ω) + kv̄ − v̄˜k2H 1 (Ω) + kw − w̃k2H 1 (Ω) 0 0 0 h −1/2 2 2 2 + T k∂t (v̄ − v̄˜)kL2 (Ω) + k∂t (u − ũ)kL2 (Ω) + k∂t (w − w̃)kL2 (Ω) + T 1/2 k∇(∂t (v̄ − v̄˜))k2L2 (Ω) + k∇(∂t (u − ũ))k2L2 (Ω) i o + k∇(∂t (w − w̃))k2L2 (Ω) 1 + kuk2H 1 (Ω) + kv̄k2H 1 (Ω) + kw̃k2H 1 (Ω) . = Ω 0 0 0 EJDE-2010/75 REACTION-DIFFUSION SYSTEM 9 5. Combining the above inequality with (3.15), Choosing > 0 sufficiently small, and simplifying, we deduce d k∂t (R − R̃)k2L2 (Ω) + kR − R̃k2H 1 (Ω) + k∇(∂t (R − R̃))k2L2 (Ω) 0 dt n h ≤ C (1 + kwk2H 1 (Ω) )2 k∂t (R − R̃)k2L2 (Ω) + kR − R̃k2H 1 (Ω) + 1 + kv̄k2H 1 (Ω) 0 0 0 2 2 −1/2 2 2 2 + kũkH 1 (Ω) + kwkH 1 (Ω) + T k∂t v̄˜kL2 (Ω) + k∂t ũkL2 (Ω) + k∂t wkL2 (Ω) 0 0 + k∂t uk2L2 (Ω) + T 1/2 k∇(∂t v̄˜)k2L2 (Ω) + k∇(∂t ũ)k2L2 (Ω) + k∇(∂t w))k2L2 (Ω) i ku − ũk2H 1 (Ω) + kv̄ − v̄˜k2H 1 (Ω) + kw − w̃k2H 1 (Ω) + k∇(∂t u)k2L2 (Ω) 0 0 0 h −1/2 2 2 2 ˜ k∂t (v̄ − v̄ )kL2 (Ω) + k∂t (u − ũ)kL2 (Ω) + k∂t (w − w̃)kL2 (Ω) + T + T 1/2 k∇(∂t (v̄ − v̄˜))k2L2 (Ω) + k∇(∂t (u − ũ))k2L2 (Ω) i o + k∇(∂t (w − w̃))k2L2 (Ω) 1 + kuk2H 1 (Ω) + kv̄k2H 1 (Ω) + kw̃k2H 1 (Ω) , 0 0 0 (3.17) where C = C(k, Ω, B, B 0 , Q, L). ˜ and S − S̃, which for 6. There exist analogous estimates to (3.17) for Q̄ − Q̄ brevity, we do not render here. If we combine these estimates with (3.17), we deduce, after an application of the differential form of the Gronwall’s inequality, the estimates: ˜ , R̃, S̃)k 3 k(Q̄, R, S) − (Q̄ X = kτ [(v̄, u, w)] − τ [(v̄˜, ũ, w̃)]kX 3 1/2 h i ≤ C T + T 1/2 exp 2−1 T (1 + kwk2H 1 (Ω) ) 0 1/2 × 1 + k(v̄, u, w)k2X 3 + k(v̄˜, ũ, w̃)k2X 3 k(v̄, u, w) − (v̄˜, ũ, w̃)kX 3 (3.18) where C = C(Q, B, B 0 , Ω, k, d, ν, L). 7. Notice that the bound in (3.18) is not uniform. Thus we need to prove the existence of a unique solution in a subset of X 3 . Define a convex set √ K := {(v̄, u, w)|(v̄, u, w) − (v̄0 , u0 , w0 ) ∈ X03 and k(v̄, u, w)kX 3 ≤ 2 Σ}, (3.19) where X03 is the set where the initial and the boundary values are zero; and Σ = constant is the bound in (2.19). We will show that, if T > 0 is sufficiently small, then τ [K] ⊆ K, kτ [(v̄, u, w)] − τ [(v̄˜, ũ, w̃)]kX 3 ≤ γk(v̄, u, w) − (v̄˜, ũ, w̃)kX 3 (3.20) for all (v̄, u, w), (v̄˜, ũ, w̃) ∈ K and some γ < 1. Using (2.19) and (3.7), we have kτ [(v̄0 , u0 , w0 )]kX 3 = k(Q̄(x, 0), R(x, 0), S(x, 0))kX 3 = k(v̄0 , u0 , w0 )kX 3 ≤ √ Σ (3.21) 10 S. A. SANNI EJDE-2010/75 Therefore, for (v̄, u, w) ∈ K, using (3.18) and (3.21), kτ [(v̄, u, w)]kX 3 ≤ kτ [(v̄0 , u0 , w0 )]kX 3 + kτ [(v̄, u, w)] − τ [(v̄0 , u0 , w0 )]kX 3 1/2 h i √ ≤ Σ + C T + T 1/2 exp 2−1 T (1 + kwk2H 1 (Ω) ) 0 1/2 (3.22) k(v̄, u, w) − (v̄0 , u0 , w0 )kX 3 × 1 + k(v̄, u, w)k2X 3 + k(v̄0 , u0 , w0 )k2X 3 1/2 √ √ ≤ Σ + C T + T 1/2 exp 2−1 T (1 + 4Σ) (1 + 5Σ)1/2 (4 Σ) √ ≤ 2 Σ, for T > 0 sufficiently small such that 1/2 4C T + T 1/2 exp[2−1 T (1 + 4Σ)](1 + 5Σ)1/2 ≤ 1 (3.23) Thus τ [(v̄, u, w)] ∈ K, and hence τ (K) ⊆ K for T > 0 sufficiently small. Furthermore, if T is chosen sufficiently small such that 1/2 C T + T 1/2 exp[2−1 T (1 + 4Σ)](1 + 5Σ)1/2 = γ < 1, (3.24) then, (3.18) implies kτ [(v̄, u, w)] − τ [(v̄˜, ũ, w̃)]kX 3 < γk(v̄, u, w) − (v̄˜, ũ, w̃)kX 3 (3.25) for all (v̄, u, w), (v̄˜, ũ, w̃) ∈ K. Thus, the mapping τ is a strict contraction for sufficiently small T > 0. 8. Given (v̄k , uk , wk ) (k = 0, 1, 2, . . . ), inductively define (Q, R, S) := (v̄k+1 , uk+1 , wk+1 ) ∈ K to be the unique weak solution of the linear initial boundary value problem ∇.v̄k+1 = 0 in ΩT ∂v̄k+1 1 − ν∆v̄k+1 = −∇.(v̄k ⊗ v̄k ) − ∇pk+1 in ΩT ∂t ρ ∂uk+1 − k∆uk+1 = −∇.(v̄k uk ) + Qwk f (uk ) in ΩT ∂t ∂wk+1 − d∆wk+1 = −∇.(v̄k wk ) − wk f (uk ) in ΩT ∂t v̄k+1 = 0̄, uk+1 = wk+1 = 0 on ∂Ω × [0, T ) v̄k+1 (x, 0) = v̄0 (x), uk+1 (x, 0) = u0 (x), wk+1 (x, 0) = w0 (x), (3.26) (3.27) (3.28) (3.29) (3.30) (3.31) where Y := pk+1 is the pressure distribution corresponding to (v̄k+1 , uk+1 , wk+1 ). By the definition of the mapping τ , we have (for k = 0, 1, 2, . . . ), using (3.26)(3.31) that (v̄k+1 , uk+1 , wk+1 ) = τ [(v̄k , uk , wk )]. (3.32) Consider the series (v̄1 , u1 , w1 ) + X r≥2 [(v̄r , ur , wr ) − (v̄r−1 , ur−1 , wr−1 )] (3.33) EJDE-2010/75 REACTION-DIFFUSION SYSTEM 11 The partial sum of the first k + 1 terms of the series (3.33) is (v̄1 , u1 , w1 ) + k+1 X [(v̄r , ur , wr ) − (v̄r−1 , ur−1 , wr−1 )] = (v̄k+1 , uk+1 , wk+1 ) (3.34) r=2 Now, using (3.25), we have k(v̄2 , u2 , w2 ) − (v̄1 , u1 , w1 )kX 3 = kτ [(v̄1 , u1 , w1 )] − τ [(v̄0 , u0 , w0 )]kX 3 < γk(v̄1 , u1 , w1 ) − (v̄0 , u0 , w0 )kX 3 k(v̄3 , u3 , w3 ) − (v̄2 , u2 , w2 )kX 3 = kτ [(v̄2 , u2 , w2 )] − τ [(v̄1 , u1 , w1 )]kX 3 < γ 2 k(v̄1 , u1 , w1 ) − (v̄0 , u0 , w0 )kX 3 (3.35) (3.36) By induction, k(v̄k+1 , uk+1 , wk+1 ) − (v̄k , uk , wk )kX 3 < γ k k(v̄1 , u1 , w1 ) − (v̄0 , u0 , w0 )kX 3 √ (3.37) < 4γ k Σ, since (v̄1 , u1 , w1 ), (v̄0 , u0 , w0 ) are in K, defined by (3.19). Hence the series (3.33) is absolutely convergent, since using (3.37), the series X √ 4γ k Σ, (3.38) k=0 which converges, dominates X k(v̄1 , u1 , w1 )kX 3 + k(v̄r , ur , wr ) − (v̄r−1 , ur−1 , wr−1 )kX 3 . (3.39) r≥2 This implies that the series (3.33) is convergent. Define lim (v̄k+1 , uk+1 , wk+1 ) := (v̄, u, w). k→∞ Thus (v̄k+1 , uk+1 , wk+1 ) → (v̄, u, w) uniformly in K. Thus lim (v̄k+1 , uk+1 , wk+1 ) = (v̄, u, w) = lim τ [(v̄k , uk , wk )] = τ [(v̄, u, w)] k→∞ k→∞ (3.40) By (3.40), (v̄, u, w) ∈ K is the unique fixed point of τ . 9. As in [15], define V := The closure of {ζ̄ ∈ Cc∞ (Ω) : ∇.ζ̄ = 0} in H01 (Ω). (3.41) In view of the previous steps of this section, we are motivated to give the following definition. Definition 3.2. The weak formulation of (1.1)-(1.6) is: For given (v̄0 , u0 , w0 ) ∈ [H01 (Ω) ∩ H 2 (Ω)]3 , find (v̄, u, w) ∈ K satisfying Z Z Z ∂t v̄.ζ̄dx + ν ∇v̄ : ∇ζ̄dx = − ∇.(v̄ ⊗ v̄).ζ̄dx (3.42) Ω Ω Ω Z Z Z Z ∂t uξdx + k ∇u.∇ξdx = − ∇.(v̄u)ξdx + Q wf (u)ξdx (3.43) Ω ZΩ ZΩ Z ZΩ ∂t wξdx + d ∇w.∇ξdx = − ∇.(v̄w)ξdx − wf (u)ξdx (3.44) Ω Ω v̄(x, 0) = v̄0 (x), Ω u(x, 0) = u0 (x), for each ζ ∈ V and each ξ ∈ H01 (Ω). Ω w(x, 0) = w0 (x), (3.45) 12 S. A. SANNI EJDE-2010/75 10. Before verifying that (v̄, u, w) is weak solution of (1.1)-(1.6), we first prove the following Lemma. Lemma 3.3. If (v̄k , uk , wk ) ∈ K, ξ ∈ H01 (Ω) and ζ̄ ∈ V , then Z Z ∇.(v̄k ⊗ v̄k ).ζ̄dx → ∇.(v̄ ⊗ v̄).ζ̄dx Ω Z ZΩ ∇.(v̄k uk )ξdx → ∇.(v̄u)ξdx ZΩ ZΩ ∇.(v̄k wk )ξdx → ∇.(v̄w)ξdx Ω (3.46) (3.47) (3.48) Ω in L2 (Ω) Z wk f (uk )ξdx → wf (u)ξdx f (uk ) → f (u) Z Ω (3.49) (3.50) Ω Proof. (i). Proof of (3.46). Integrating by parts, we have Z Z | ∇.(v̄k ⊗ v̄k ).ζ̄dx| = | v̄k ⊗ v̄k : ∇ζ̄dx| ≤ kζkH01 (Ω) kuk k2H 1 (Ω) , Ω 0 Ω (3.51) by using (2.9) of Lemma 2.2. Equation (3.46) follows by taking limits on both sides of (3.51). Further, the proofs of (3.47) and (3.48) follow by similar calculations. (ii). Proof of (3.49). We have Z Z Z uk Z 2 0 0 B |uk | + f (0)2 dx |f (uk )|2 dx = f (r)dr + f (0) dx ≤ Ω Ω Ω 0 Z (3.52) 0 2 0 ≤ C1 (B , f (0)) (|uk | + 2|uk | + 1|)dx ≤ C2 (B , f (0), Ω)(kuk kL2 (Ω) + 1)2 Ω R where we have used the first inequality in (1.8) and the estimate Ω |uk |dx ≤ 1 |Ω| 2 kuk kL2 (Ω) ). Then (3.49) follows by taking limits on both sides of (3.52). (iii). Proof of (3.50). We estimate Z | wk f (uk )ξdx| ≤ kwk kH01 (Ω) kf (uk )kL2 (Ω) kξkH01 (Ω) , (3.53) Ω using (2.9) of Lemma 2.2. Hence, (3.50) follows by taking limits in (3.53). 11. We now verify that (v̄, u, w) ∈ K is a weak solution of (1.1). Fix ζ ∈ V and ξ ∈ H01 (Ω). Using (3.26)-(3.31), we have Z Z Z ∂t v̄k+1 .ζ̄dx + ν ∇v̄k+1 : ∇ζ̄dx = − ∇.(v̄k ⊗ v̄k ).ζ̄dx (3.54) Ω Ω Ω Z Z ∂t uk+1 ξdx + k ∇uk+1 .∇ξdx Ω Ω Z Z (3.55) =− ∇.(v̄k uk )ξdx + Q wk f (uk )ξdx Ω Ω Z Z ∂t wk+1 ξdx + d ∇wk+1 .∇ξdx Ω Ω Z Z (3.56) =− ∇.(v̄k wk )ξdx − wk f (uk )ξdx Ω v̄k+1 (x, 0) = v̄0 (x), uk+1 (x, 0) = u0 (x), Ω wk+1 (x, 0) = w0 (x) . (3.57) EJDE-2010/75 REACTION-DIFFUSION SYSTEM 13 Letting k → ∞ in (3.54)-(3.57) and using Lemma 2.3 to handle the nonlinear terms yield (3.42)-(3.45) as desired. 12. We next demonstrate how to obtain the pressure Y = pk+1 . First, we obtain the boundary condition on pressure by taking (1.2) on the boundary and using (1.5) to deduce 1 ∇p = ν∆v̄ on ∂Ω (3.58) ρ Following the steps in [12], we express (3.58) in terms of the standard normal derivatives as ∂ 2 v̄ 1 ∂p (3.59) = ν n̂. 2 ρ ∂n ∂n where n̂(x) is the inward normal at x on ∂Ω. Taking the divergence of (3.27) yields the equation satisfied by Y = pk+1 as ∆pk+1 = ρ∇.[∇.(v̄k ⊗ v̄k )], (3.60) which is a form of Poisson’s equation. Further, in sympathy with the boundary condition (3.59), we impose the the boundary condition on the pressure pk+1 as 1 ∂pk+1 ∂ 2 v̄ = ν n̂. 2 ρ ∂n ∂n Hence, the formal solution of (3.60) subject to the condition (3.61) is Z pk+1 (x, t) = −ρ G(x, y)∇.[∇.(v̄k (y, t) ⊗ v̄k (y, t))]dy Ω Z ∂ 2 v̄(y, t) dS(y) + ρν G(x, y)n̂. ∂n2 ∂Ω (3.61) (3.62) where G(x, y) is the Green’s function satisfying the Laplace’s equation in the form ∆G(x, y) = δ(x − y) (3.63) with the condition ∂G(x, y) = 0 (x on ∂Ω). ∂n where δ is the Dirac delta function. 1 , x, y ∈ Ω, we define For n = 3, G(x, y) = |x−y| Z 1 pk+1 (x, t) := −ρ ∇.[∇.(v̄k (y, t) ⊗ v̄k (y, t))]dy Ω |x − y| + Z 1 ∂ 2 v̄(y, t) + ρν n̂. dS(y), |x − y| + ∂n2 Z ∂Ω 1 p (x, t) := −ρ ∇.[∇.(v̄(y, t) ⊗ v̄(y, t))]dy Ω |x − y| + Z 1 ∂ 2 v̄(y, t) + ρν n̂. dS(y) |x − y| + ∂n2 Z∂Ω 1 p(x, t) := −ρ ∇.[∇.(v̄(y, t) ⊗ v̄(y, t))]dy |x − y| Ω Z ∂ 2 v̄(y, t) 1 + ρν n̂. dS(y) ∂n2 ∂Ω |x − y| (3.64) (3.65) (3.66) (3.67) 14 S. A. SANNI EJDE-2010/75 where > 0. Notice that lim pk+1 (x, t) = pk+1 (x, t), → 0 lim p (x, t) = p(x, t) → 0 Hence, integrating twice by parts, using (1.1) and (1.5), we have |pk+1 (x, t) − p (x, t)| Z 1 = |ρ ∇.{∇.[v̄k (y, t) ⊗ v̄k (y, t) − v̄(y, t) ⊗ v̄(y, t)]}dy| |x − y| + ZΩ 1 = |ρ ∇ ∇[ ] : [v̄k (y, t) ⊗ v̄k (y, t) − v̄(y, t) ⊗ v̄(y, t)]dy| |x − y| + Ω (3.68) which tends to 0 as k → ∞. Therefore lim pk+1 (x, t) = p (x, t), k→∞ (3.69) From whence sending to 0, we obtain lim pk+1 (x, t) = p(x, t), k→ ∞ (3.70) where, p(x, t) given by (3.67), is the pressure corresponding to the solution (v̄, u, w). Indeed, (3.67) is the formal solution for the pressure p(x, t) satisfying ∆p = ρ∇.[∇.(v̄ ⊗ v̄)], in terms of G(x, y), as obtained in [12]. (3.71) 4. Regularity The Analysis so far carried out requires no smoothness assumption on the boundary. However, for smooth solution up to the boundary, one requires the boundary ∂Ω to be C ∞ . The lengthy proofs of the associated regularity theorems are currently being established by an analysis of certain difference quotients in another paper. Acknowledgments. The author would like to thank the anonymous referee whose thoughtful comments improved the original version of this manuscript. References [1] Avrin, J. D.; Asymptotic behavior of one-step combustion models with multiple reactants on bounded domains, SIAM J. Math. Anal., 24 (1993), 290-298. [2] Avrin, J. D.; Qualitative theory of the Cauchy problem for a one-step reaction model on bounded domains, SIAM J. Math. Anal., 22 (1991), 379-391. [3] Buckmaster J.; The Mathematics of Combustion, SIAM, Philadelphia (1985). [4] Buckmaster, J.; Ludford, G.; Theory of Laminar Flames, Cambridge University Press, Cambridge (1982). [5] Daddiouaissa, E.; Existence of global solutions for a system of equations having a triangular matrix, Electron. Journal of Diff. Equations, 2009(2009), No. 09, 1-7. [6] De Oliveira, L. F.; Pereira, A. L.; Pereira, M. C.; Continuity of attractors for a reactiondiffusion problem with respect to variations of the domain, Electron. Journal of Diff. Equations, 2005(2005), No. 100, 1-18. [7] Evans, L. C.; Partial Differential Equations, American Mathematical Society, Providence, Rhode Island,(1998). [8] Fitzgibbon, W. E.; Martin, C. B.; The longtime behavior of solutions to a quasilinear combustion model, Nonlinear Analysis, Theory, Methods & Applications, 19(1992), No. 10, 947-961. [9] Frank-Kamenetski, D.; Diffusion and Heat Transfer in Chemical Kinetics. Plenum Press, New York (1969). EJDE-2010/75 REACTION-DIFFUSION SYSTEM 15 [10] Henry, D.; Geometric Theory of Semilinear Parabolic Equations, Lecture Notes in Math. 840, Springer Verlag, Berlin, New York, 1981. [11] Kouach, S., Existence of global solutions to reaction-diffusion systems with nonhomogeneous boundary conditions via Lyapunov functional, Electron. Journal of Diff. Equations, 2002(2002), No. 88, 1-13. [12] McCOMB, W. D.; The Physics of Fluid Turbulence, Clarendon Press, Oxford, (1990). [13] Sanni, S. A.; A Combustion model with unbounded thermal conductivity and reactant diffusivity in non-smooth domains, Electron. Journal of Diff. Equations, 2009(2009), No. 60, 1-14. [14] Sattinger, D.; A nonlinear parabolic system in the theory of combustion, Quart. Appl. math., 33(1975), 47-61. [15] Temam, R.; Navier-Stokes Equations, North-Holland Publishing Company, Oxford, (1977). Sikiru Adigun Sanni Department of Mathematics, Statistics and Computer Science, University of Uyo, Uyo, Akwa Ibom State, Nigeria E-mail address: sikirusanni@yahoo.com