EUROPEAN JOURNAL OF PURE AND APPLIED MATHEMATICS Vol. 1, No. 2, 2008, (11-20) ISSN 1307-5543 – www.ejpam.com Multiple solutions of steady MHD flow of dilatant fluids Zakia Hammouch∗ LAMFA, CNRS UMR 6140, Faculté de Mathématiques et d’Informatique, Université de Picardie Jules Verne, 80039 Amiens, France Abstract. In this paper we consider the problem of a steady MHD flow of a non-Newtonian power- law and electrically conducting fluid in presence of an applied magnetic field. The boundary layer equations are solved in similarity form via the Lyapunov energy method, we show that this problem has an infinite number of positive global solutions. AMS subject classifications: 34B15, 34B40, 76D10, 76M55 Key words: Asymptotic solution, Boundary-layer, Degenerate differential equation, MHD flow, Power- law fluid, Similarity solution 1. Introduction The study of non-Newtonian fluid flows has considerable interests, this is primarily be- cause of the numerous applications in several engineering fields. Such processes are wire drawing, glass fiber and paper production, crystal growing, drawing of plastic sheets etc. For more details about the behavior in both steady and unsteady flow situations, together with mathematical models, we refer the reader to the books [1] by Astarita and Marucci, [2] by Bohme and the references therein. One particular non-Newtonian model which has been widely studied is the Ostwald-de Wael power-law model [3] [4], which relies the shear stress to the strain rate uy by the expression τx y = k|uy |n−1uy , (1.1) where k is a positive constant, and n > 0 is called the power-law index . The case n < 1 is referred to pseudo-plastic or shear-thinning fluid, the case n > 1 is known as dilatant or shear-thickening fluid. The Newtonian fluid is a special case where the power-law index n is equal to one. In the present work we shall restrict our study to the case n> 1. The magnetohydrodynamics (MHD) flow problems find also applications in a large variety of physical, geophysical and industrial fields [5]. It is also interesting to study the flow of non- Newtonian fluids with externally imposed magnetic fields. To the author knowledge MHD flow of non-Newtonian fluids was first studied by Sarpkaya [6]. In [7] Sapunkov derived the equations describing the similarity solutions for the non-Newtonian flow when the external ∗Corresponding author. Email address: zakia.hammouch@u-picardie.fr (Zakia Hammouch) http://www.ejpam.com 11 c© 2007 EJPAM All rights reserved. Zakia Hammouch / Eur. J. Pure Appl. Math, 1 (2008), (11-20) 12 applied magnetic field varies as x m−1 2 , in presence of a pressure gradient, he used the method of series expansion. Later, Djuvic [8] employed a Crocco’s variables to study the unsteady flow with exponentially external velocity (in time). Recently, Liao [9] introduced a powerful tech- nique (homotopy analysis) to give analytic solutions of MHD viscous flows of non-Newtonian fluids over a stretching sheet. In this paper, we reconsider the steady two-dimensional laminar flow of an incompressible viscous electrically conducting dilatant fluid over a stretching flat plate with a power-law ve- locity distribution in the presence of a perpendicular magnetic field. Our interest in this work has been motivated by the work of Chiam [10], who have considered the flow over an imper- meable flat plate, for which similarity solutions were found via the Crocco transformation. 2. Derivation of the model Consider a steady two-dimensional laminar flow of an incompressible dilatant and electri- cally conducting fluid of density ρ, past a semi-infinite flat plate. Let (x , y) be the Cartesian coordinates of any point in the flow domain, where x−axis is along the plate and y−axis is normal to it. Assume that a magnetic field H(x), is applied normally to the plate. The continuity and momentum equations can be simplified, within the boundary-layer ap- proximation, into the following equations (see [7] [10]) ux + vy = 0, (2.1) uux + vuy = ν(|uy |n−1uy)y + ueue x + σµ2H2 ρ (ue − u). (2.2) Accompanied by the boundary conditions u(x , 0) = Uw(x), v(x , 0) = Vw(x) and u(x , y)→ ue(x) as y →∞. (2.3) Where the functions u and v are the velocity components in the x and y directions respectively, ue(x) = U∞xm is the free-stream velocity. The parameters ν , n,µ,σ and H are the kinematic viscosity, the flow behavior index, the magnetic permeability, the electrical conductivity of the fluid, and the magnetic field intensity respectively. The functions Uw(x) = uw xm(uw > 0) and Vw(x) = vw x m(2n−1)−n n+1 are the stretching and the suction/injection velocities respectively. In term of the stream-function (ψ which satisfied u(x , y) = ψy(x , y) and v(x , y) = −ψx(x , y)), equations (2.1),(2.2) can be reduced to the single equation ψyψx y −ψxψy y = ν(|ψy y |n−1ψy y)y + ueue x + σµ2H2 ρ (ue −ψy), (2.4) subject to ψy(x , 0) = uw xm, ψx(x , 0) =−vw x m(2n−1)−n n+1 and ψy(x , y) = U∞xm as y →∞. (2.5) Zakia Hammouch / Eur. J. Pure Appl. Math, 1 (2008), (11-20) 13 According to Sapunkov [7], similarity solutions for problem (2.4),(2.5) exist only if the mag- netic field has the following form H(x)∼ x m−1 2 . To look for similarity solutions we define the following η := Ay x−a and ψ(x , y) := Bx b f (η), (2.6) where f is the transformed dimensionless stream function and η is the similarity variable. Thanks to (2.6), the function f satisfies the new boundary value problem    (| f ′′|n−1 f ′′)′+ a f f ′′+m(1− f ′2) +M(1− f ′) = 0, f (0) = α, f ′(0) = δ, f ′(∞) = 1, (2.7) if and only if a = 1+m(2n− 1) n+ 1 , b = 1+m(n− 2) n+ 1 , a− b = m, and the parameters A and B satisfy AB = u∞ and νBn−2A2(n−1) = 1. Where the primes denote differentiation with respect to η, the function f ′(η) denotes the normalized velocity and the parameters M = σµ2H2 0(n+ 1) u∞ρ , α=− (n+ 1)vw (m+ 1)(νu2n−1 ∞ ) 1 n+1 and δ = uw u∞ , are respectively: The Hartmann number, the suction/injection and the stretching parameters. Such problems have been investigated by several authors for example, Anderson et al. [11], Zhang et al. [12] and Kumari and Nath [13]. In the same context, Chiam [10] studied Problem (2.1)-(2.3). To look for similarity solutions, he solved the following boundary value problem    n| f ′′|n−1 f ′′′+ f f ′′+ β(1− f ′2) +M(1− f ′) = 0, f (0) = 0, f ′(0) = 0, f ′(∞) = 0. (2.8) Where β = m(n+1) (2n−1)m+1 . We aim here to stress that for n 6= 1, equation (2.7)1 can be degenerate at some point ηs for which f ′′(ηs) = 0 (for more details see [15]) and then any solution of (2.7) is not necessarily of C3(0,∞). Hence equations (2.7)1 and (2.8)1 are not equivalent. Let us notice that for the Newtonian case (n = 1), problem (2.7) reduces to the Falkner-Skan flow in Magnetohydrodynamics, which has been studied by Hildyard [17], Aly et al. [18] and Hoernel [19]. The case m = M = 0 leads to the generalized Blasius problem (see [20]). Zakia Hammouch / Eur. J. Pure Appl. Math, 1 (2008), (11-20) 14 While the case m = −M , by a suitable scaling, is referred to the mixed convection of a non- Newtonian fluid in a porous medium (see for example [21]). We note also that in absence of the magnetic field, problem (2.7) is simplified to the Falkner-Skan flow for non-Newtonian fluids. A complete study on this subject is given in [22] by Denier and Dabrowski. Very recently, Aly et al. [18] reported a theoretical and numerical investigations on the exis- tence of solutions to problem (2.7) for Newtonian fluids (n= 1), say    f ′′′+ m+1 2 f f ′′+m(1− f ′2) +M(1− f ′) = 0, f (0) = α≥ 0, f ′(0) = δ, f ′′(0) = γ. (2.9) They showed that problem (2.9) has multiple solutions for any δ ∈ (0,Γ) and γ ∈ R satisfying γ2 ≤ 2m 3 δ3+Mδ2− 2(M +m)δ, (2.10) where Γ = − 3M 4m  1+ r 1+ 16m 3M2 (m+M)   > 1. In the present work, we aim to extend their results to the non-Newtonian dilatant fluids (n > 1), by using a condition on γ which is different from (2.10) and without any restriction on the parameter δ. 3. Non-uniqueness of solutions Guided by the analysis of [14], [15] and [16], we aim to prove the existence of solutions to problem (2.7), for related values of the parameters m, M , n,α,δ and γ. This result will be established by mean of the so-called shooting method, the boundary value problem (2.7) is then converted into the following initial value problem    (| f ′′|n−1 f ′′)′+ a f f ′′+m(1− f ′2) +M(1− f ′) = 0, f (0) = α, f ′(0) = δ, f ′′(0) = γ. (3.1) Where the real number γ is the shooting parameter. The initial value problem (3.1) can be transformed into the equivalent first order ordinary differential system          f ′ = g, g ′ = |h| 1−n n h, h′ =−a f |h| −m(1− g2)−M(1− g), (3.2) with the conditions f (0) = α, g(0) = δ, h(0) = |γ|n−1γ. (3.3) Zakia Hammouch / Eur. J. Pure Appl. Math, 1 (2008), (11-20) 15 By the classical theory of ordinary differential equations, problem (3.2),(3.3) has a unique local (maximal) solution for every γ 6= 0. Let fγ denotes this solution and (0,ηγ), ηγ ≤ ∞, denotes its maximal interval of existence. The main task now is to show how existence of solutions depends on γ. The local solution fγ satisfies the following | f ′′γ| n−1 f ′′γ + a f ′γ fγ−M( fγ+α) = |γ|n−1γ+ aαδ− (M +m)η+(a+m) ∫ 0 η f ′γ 2(τ)dτ. (3.4) Equation (3.4) will be used for proving the main results. Definition 3.1. A function fγ is said to be a solution to (3.1) if f ∈ C2(0,∞), | f ′′γ | n−1 f ′′γ ∈ C1(0,∞) and satisfies lim η→∞ f ′γ(η) = 1 (i) and lim η→∞ f ′′γ (η) = 0 (ii) 3.1. Suction/Injection flows (α ∈ R) Theorem 3.1. Assume α ∈ R, δ > 0, M > 0 , n> 1 and − 1 3n < m<−M. For any γ satisfying |γ|n−1γ >−aαδ (?), problem (3.1) admits a global unbounded solution. Proof. From a physical point of view, it is more convenient to prove the result for the cases α≥ 0 (suction) α < 0 (injection) separately. We have to show that fγ is a positive monotonic increasing function on (0,ηγ), globally de- fined and going to infinity with η. For this sake we define the Lyapunov Energy function by V (η) = 1 n+ 1 | f ′′|n+1− m 3 f ′3− M 2 f ′2+ (M +m) f ′. (3.5) Which satisfies V ′(η) =−a f f ′′2. Then V is monotonic decreasing on (0,ηγ). On the other hand, from equation (3.4) and condition (?) we see that fγ ′ and fγ are positive on (0,ηγ) as long as fγ exists. Using the Lyapunov function V we see that fγ ′ and fγ ′′ are bounded, since V is bounded from below by 3M+4m 6 . If fγ were also bounded, say f →η∞L with L ∈ (0,∞) (since fγ is positive). Then fγ ′(η) →η∞ 0 which implies that f ′′γ (ηk) →k∞ 0, where (ηk)k≥0 is a sequence tending to infinity with k. Using again (3.4) to deduce | f ′′γ (ηk)|n−1 f ′′γ (ηk) + a f ′γ(ηk) fγ(ηk) =−M( fγ(ηk) +α) + |γ|n−1γ+ aαδ− (M +m)ηk + (a+m) ∫ 0 ηk f ′γ 2(τ)dτ. Zakia Hammouch / Eur. J. Pure Appl. Math, 1 (2008), (11-20) 16 Letting k→∞, the right hand side goes to zero while the left hand side goes to minus infinity, which is impossible. then fγ is a global unbounded solution to (3.1). From the above f ′γ and f ′′γ are bounded and f ′γ is monotonic increasing on (η1,∞), for η1 large enough. Then there exists l > 0 such that limη→∞ f ′γ(η) = l, and there exits a sequence (ζk)k, tending to infinity with k such that limk→∞ f ′′γ (ζk) = 0. Making recourse to the Lyapunov function V we get limη→∞ f ′′γ (ζk) = 0. Assume now that f ′′γ is not monotonic on any interval [η2,∞). Then, there exists a sequence (τk)k going to infinity with k such that: • (| f ′′γ | n−1 f ′′γ ) ′(τk) = 0, • | f ′′γ | n−1 f ′′γ (τ2k) is a local minimum, • | f ′′γ | n−1 f ′′γ (τ2k+1) is a local maximum. From (3.1)1, we have f ′′γ (τk) =− m(1− f ′2γ (τk)) +M(1− f ′γ(τk)) a fγ(τk) . Since f ′γ is bounded and fγ goes to infinity with ηk, we get easily from the above that f ′′γ goes to zero with η. Now we show that fγ satisfies (i). Recall that f ′γ is a positive bounded function then f ′γ → l with l ∈ (α,∞). At infinity we have fγ ∼ ηl and from identity (3.4) we get | f ′′γ | n−1 f ′′γ ∼ η[ml2+Ml − (M +m)] + o(1) as η approaches infinity, this leaves only the possibility that l is either 1 or −M m − 1, thanks to the positivity of f ′γ we deduce that l = 1. To finish we show the result for α < 0. In such case, the function fγ is negative on a small neighborhood of zero. According to (3.4) fγ cannot have a local maximum, then two possi- bilities arise: • Either fγ < 0 ∀η ∈ (0,ηγ) • Or ∃η? such that fγ < 0 on (0,η?), fγ(η?) = 0 and fγ > 0 ∀η > η?. Assume that the first assertion holds, then α > fγ(∞) ≤ 0 and f ′γ(∞) = 0, f ′ being positive we use again (3.4) to get that f ′′γ is positive. A contradiction. Then, fγ has exactly one zero η?. We define the shifted function h by : η 7−→ h(η) = fγ(η+η?), which satisfies h(0) = 0, h′(0) = δ and h′′(0) > 0, and we use the above analysis to conclude that h is an unbounded global solution to (3.1). 3.2. Reversed flows (δ < 0) Now we pay attention to the case of reversed flows (δ < 0). First, we show that the shooting parameter has to be positive. Zakia Hammouch / Eur. J. Pure Appl. Math, 1 (2008), (11-20) 17 Proposition 3.1. Let fγ be a solution to (3.1) with m ∈ (− 1 3n ,−M), α < 0, δ < 0 and γ ≤ 0, then the condition (i) is failed. Proof. Let δ < 0, if γ ≤ 0 then fγ ′′ is negative on some (0,η0), for η0 small, and equation (3.1) can be written as ( fγ ′′eF )′ =− eF n | f ′′γ | 1−n h m(1− f ′2γ ) +M(1− f ′γ) i , where F(η) = a n ∫ 0 η fγ| fγ′′|1−ndτ. From this we see that η 7−→ f ′′eF decreases and then f ′′γ (η)≤ 0 for all η ∈ (0,ηγ). It follows that f ′γ is decreasing on (0,ηγ) and then the condition (i) could not be satisfied. Theorem 3.2. Let δ < 0,α > 0 and m ∈ (− 1 3n ,−M). For any γ > 0 satisfying αγn− 1 2 δ2γn−1+ aα2δ− M 2 α2 > 0 (??), problem (3.1) has a global unbounded solution. Proof. Let fγ be the local solution of (3.1), define the auxiliary function G(η) = fγ fγ ′′| f ′′γ | n−1− 1 2 f ′2γ | f ′′ γ | n−1+ a f 2 γ f ′γ − M 2 f 2 γ , (3.6) which satisfies G′(η) =−(m+M) fγ+ � 2a+m+ (n− 1)a 2n � f ′2γ fγ+ n− 1 2n f ′2γ f ′′γ −1 h m(1− f ′γ 2) +M(1− f ′γ) i , (3.7) and G(0)> 0. Since f ′γ < 0, the function fγ is negative on a small neighborhood of zero. Assume that there exists η1 ∈ (0,∞) such that fγ(η)> 0, f ′γ < 0 ∀η ∈ [0,η1) and fγ(η1) = 0. Hence G is a monotonic nondecreasing function on (0,η1) and then G(η1) ≤ 0. Then G(η) ≤ 0 for all η ∈ (0,η1), in particular G(0) ≤ 0, which is a contradiction with (??). Therefore we have : • Either fγ > 0 and f ′γ ≤ 0 ∀η≥ 0 •Or ∃η2 > 0 : fγ > 0, f ′γ < 0 ∀η ∈ (0,η2), f ′γ(η2) = 0 and fγ(η2) is a local maximum. Assume that the first assertion holds, then fγ has a finit limit at infinity, say L ∈ (0,∞) and there exists a sequence (χk)k ≥ 0 tending to infinity with k such that f ′γ(χk) goes to zero at infinity. If f ′γ is monotonic (resp. non-monotonic on any interval (η,∞)) we get f ′γ goes to zero at infinity and then f ′′γ (δk) goes to zero at infinity for a sequence (δk)k≥0 going to Zakia Hammouch / Eur. J. Pure Appl. Math, 1 (2008), (11-20) 18 infinity with k (resp. f ′′γ (δk) = 0 and f ′γ(δk) goes to zero at infinity). Because G(0) < G(δk), we obtain a contradiction by taking the limit as k goes to infinity. Now, we claim that the function fγ cannot have a local maximum after η2. Actually, assume there exists η3 > η2 such that fγ(η3) is a local maximum. At this point the function G takes a negative value and satisfies G(η3)≥ G(0) a contradiction. Since fγ is monotonic increasing after η2 we deduce as the above that is a global solution. Next, we argue as in the proof of Theorem. 3.1 to show that fγ is unbounded at infinity and satisfies (i) and (ii). 3.3. Flow with large initial velocity (δ� 1) In this subsection, we construct asymptotic solutions to problem (3.1) when the real δ is very large. Adopting the method used in [23] by Aly et al., we assume that such solutions can be written under the following form f (η) = η+ ξr g(t), where t = ξsη, ξ= δ− 1 and r, s ∈ R. Then problem (3.1) reads    ξ(r+2s)(n−2)(|g ′′|n−1 g ′′)′+ aηξ−r g ′′+ ag g ′′− (2m+M)ξ−(r+s)−mg ′2 = 0, g(0) = αξ−r , g ′(0) = ξ1−(r+s), g ′(∞) = 0. (3.8) Setting r = 2n−1 n+1 and s = 2−n n+1 , ensures that the highest derivative remains present in the resulting problem. As ξ goes to infinity, we deduce    (|g ′′|n−1 g ′′)′+ ag g ′′+mg ′2 = 0, g(0) = 0, g ′(0) = 1, g ′(∞) = 0. (3.9) Problem (3.9) describes the steady free convection flow of a non-Newtonian power-law fluid over a stretching flat plate embedded in a porous medium. In [15], it was shown that for m ∈ (− 1 3n , 0) any local solution g, whith positive values of τ (τ = g ′′(0)), is global and satisfies the following asymptotic behaviour g(t)∼ t 1+m(2n−1) 1+m(n−2) , as t →∞. Consequently, a solution f for positive γ and large δ (if it exists), may have the following large η-behaviour f (η)∼ η � 1+ (δ− 1) 1 1+m(n−2)η m(n−1)+2 1+m(n−2) � . REFERENCES 19 4. Concluding remarks Based on the similarity transformation approach, the boundary layer equations for flows of purely viscous non-Newtonian dilatant and electrically conducting fluids are investigated. Using a shooting argument, it is shown that the relevant problem admits an infinite number of solutions ( [24] [25] [26] and [27]), this is due to the arbitrariness of the shooting parameter γ. From a physical point of view, we underline that γ = f ′′(0) originates from the local skin friction coefficient C fx , and the local Reynolds number Rex = (uw xm)2−n xn νk via the the formula C fx Rex 1 n+1 = 2 �a n � 1 n+1 |γ|n−1γ. In conclusion, we may expect that the solutions determined above are physically acceptable. However, only experiments are able to prove their physical existence. Acknowledgements I would like to thank the anonymous referee for his constructive suggestions, which have improved the earlier version of this work. References [1] ASTARITA G., and MARRUCCI G., Principles of Non–Newtonian Fluid Mechanics, McGraw-Hill, 1974. [2] BOHME H., Non–Newtonian Fluid Mechanics, North–Holland Series in Applied Mathematics and Mechanics, 1987. [3] ACRIVOS A., SHAH M.J., PETERSEN E.E, Momentum and heat transfer in laminar boundary-layer flows Equations of non-Newtonian fluids past external surfaces, A.I.Ch.E J., 6 312 (1960). [4] ECE M.C. , BÜYÜK E., Similarity solutions for free convection to power-law fluids from a heated vertical plate, App. Math. Lett., 15 1-5 (2002). [5] PAVLOV K. B., Magnetohydrodynamic flow of an incompressible viscous fluid caused by deforma- tion of a surface, Magnitnaya Gidrodinamika, 4 146-147 (1974). [6] SARPKAYA T., Flow of non-Newtonian fluids in a magnetic field, AIChE J. 7 324-328 (1961). [7] SAPUNKOV S.Y., Self-similar solutions of non-Newtonian fluid boundary in MHD, Mekhanika Zhid- kosti I Gaza 2 77-82 (1967). [8] DJUVIC D. S., Hiemenz magnetic flow of power-law fluids, ASME J. Appl. Mech., 41 822-823 (1974). [9] LIAO S. J., On the analytic solution of magnetohydrodynamic flows of non-Newtonian fluids over a stretching sheet, J. Fluid Mech., 488 189-212 (2003). [10] CHIAM T. C., Solutions for the flow of a conducting power-law fluid in a transverse magnetic field and with a pressure gradient using Crocco variables, Acta Mech., Vol. 137 225 - 235 (1999). [11] ANDERSON H. I., BACH K.H., DANDAPAT B.S., Magnetohydrodynamic flow of a power-law fluid over a stretching sheet, Int. J. Non-Linear Mech., 27 929-939 (1992). [12] ZHANG Z., WANG J., SHI W., A boundary layer problem arising in gravity driven laminar film of power-law fluids along vertical walls, ZAMP. (J. Appl. Math. Phy.), 55 769-780 (2004). REFERENCES 20 [13] KUMARI M., NATH G., MHD boundary layer flow of a non-Newtonian fluid over a continuously moving surface with a parallel free stream, Acta. Mech., 146 139-150 (2001). [14] BRIGHI B., On a similarity boundary layer equation, ZAA(Zeitschrift für Analysis und ihre An- wendungen), 21 931-948 (2002). [15] HAMMOUCH Z., Étude mathématique et numérique de quelques problèmes issus de la dynamique des fluides, PhD Dissertation, Université de Picardie Jules Verne, Octobre 2006. [16] GUEDDA M., Similarity solutions of differential equations for boundary layer approximations in porous media, ZAMP. (J. Appl. Math. Phy.), 56 749-762 (2005). [17] HILDYARD L., Falkner-Skan Problems in Magnetohydrodynamics, The Phys. of Fluids, Vol.15 782– 793 (1972). [18] ALY E. H., BENLAHSEN M., GUEDDA M., Similarity solutions of a MHD boundary-layer flow past a continuous moving surface, Int. J. Engng. Sci. 45 486-503 (2007). [19] HOERNEL J-D., On the similarity solutions for a steady MHD equation, Com. Nonlin. Sci. Num. Sim., 13 1353-1360 (2008) [20] NACHMAN A., TALIAFERRO S., Mass transfer into boundary-layers for power-law fluids, Proc. R. Soc. Lon. A, 365 313-326 (1979). [21] POP I., Mixed convection to power-law type non-Newtonian fluids from a vetical wall, Pol. Plast. Tech and Eng., 30 47-65 (1991). [22] DENIER J.P., DABROWSKI P.P., On the boundary-layer equations for power-law fluids, Proc. R. Soc. Lond. A, 460 3143-3158 (2004). [23] ALY E.H, ELLIOTT L., INGHAM D.B., Mixed convection boundary-layer flow over a vertical surface embedded in a porous medium, Eur. J. Mech. B Fluids, 22 529-543 (2003). [24] MAGYARI E., ALY E.H., Mechanical and thermal characteristics of a mixed convection boundary- layer flow in a saturated porous mediuM Int. J. Heat Mass Trans., 49 3855-3865 (2006). [25] LIAO S.J., A new branch of solutions of boundary layer flows over a permeable stretching plate Int. J. Nonlinear. Mech., 42 819–830 (2007). [26] RILEY N., WEIDMAN P. D., Multiple Solutions of the Falkner-Skan Equation for Flow Past a Stretch- ing Boundary, SIAM J. Appl. Math., 49 1350-1358 (1989). [27] ROBINSON W.A., The existence of multiple solutions for the laminar flow in a uniformly porous channel with suction at both walls, J. Engng. Maths., 10 23-40 (1976).