ebook img

Fractional reaction-diffusion equations PDF

0.14 MB·
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Fractional reaction-diffusion equations

FRACTIONAL REACTION-DIFFUSION EQUATIONS 7 0 R.K. SAXENA 0 Department of Mathematics and Statistics, Jai Narain Vyas University 2 Jodhpur, 342005, India n a J A.M. MATHAI 3 Department of Mathematics and Statistics, McGill University ] Montreal, Canada H3A 2K6 A C H.J. HAUBOLD . h Office for Outer Space Affairs, United Nations t a P.O.Box 500, A1400 Vienna, Austria m [ 2 Abstract. In a series of papers, Saxena, Mathai, and Haubold (2002, 2004a, v 2004b) derived solutions of a number of fractional kinetic equations in terms 3 7 of generalized Mittag-Leffler functions which provide the extension of the 4 work of Haubold and Mathai (1995, 2000). The subject of the present paper 4 is to investigate the solution of a fractional reaction-diffusion equation. The 0 6 results derived are of general nature and include the results reported earlier 0 by many authors, notably by Jespersen, Metzler, and Fogedby (1999) for / h anomalousdiffusion anddel-Castillo-Negrete, Carreras, andLynch (2003)for t a reaction-diffusionsystems withL´evyflights. Thesolutionhasbeendeveloped m in terms of the H-function in a compact form with the help of Laplace and : v Fourier transforms. Most of the results obtained are in a form suitable for i X numerical computation. r a 1 Introduction Reaction-diffusion models have found numerous applications in pattern for- mation in biology, chemistry, and physics, see Murray (2003), Kuramoto (2003),WilhelmssonandLazzaro(2001),andHundsdorferandVerwer(2003). These systems indicate that diffusion can produce spontaneous formation of spatio-temporalpatterns. Fordetails, onecanrefertotheworkofNicolis and Prigogine (1977) and Haken (2004). A general model for reaction-diffusion systems is investigated by Henry and Wearne (2000, 2002) and Henry, Lang- lands, and Wearne (2005). The simplest reaction-diffusion models are of the form ∂N ∂2N = d +F(N),N = N(x,t), (1) ∂t ∂x2 where d is the diffusion coefficient and F(N) is a nonlinear function rep- resenting reaction kinetics. It is interesting to observe that for F(N) = γN(1 − N2), eq. (1) reduces to the Fisher-Kolmogorov equation and if we set F(N) = N(1 − N2), it gives rise to the real Ginsburg-Landau equa- tion. Del-Castillo-Negrete, Carreras, and Lynch (2002) studied the front propagation and segregation in a system of reaction-diffusion equations with cross-diffusion. Recently, del-Castillo-Negrete, Carreras, and Lynch (2003) discussed the dynamics in reaction-diffusion systems with non-Gaussian dif- fusion caused by asymmetric L´evy flights and solved the following model ∂N = ηDαN +F(N),N = N(x,t), (2) ∂t x with F = 0. In this paper we present a solution of a more general model of reaction- diffusion systems (2) in which ∂N has been replaced by ∂βN,β > 0. This ∂t ∂tβ new model extends the work of Jespersen, Metzler, and Fogedby (1999) and del-Castillo-Negrete, Carreras, and Lynch (2003). Most of the results are obtained in a compact form suitable for numerical computation. AgeneralizationoftheMittag-Lefflerfunction(Mittag-Leffler,1903,1905) ∞ zn E (z) := ,α ∈ C,Re(α) > 0, (3) α Γ(nα+1) n=0 X was introduced by Wiman (1905) in the generalized form ∞ zn E (z) := ,α,β ∈ C,Re(α) > 0. (4) α,β Γ(nα+β) n=0 X The main results of these functions are available in the handbook of Erd´elyi, Magnus, Oberhettinger, and Tricomi (1955, Section 18.1) and monographs by Dzherbashyan (1966, 1993). The H-function is defined by means of a Mellin-Barnes type integral in the following manner (Mathai and Saxena, 1978) Hm,n(z) = Hm,n z (ap,Ap) (5) p,q p,q (bq,Bq) h (cid:12) i 1 = Hm,n z(cid:12)(cid:12)(a1,A1),...,(ap,Ap) = Θ(ξ)z−ξdξ, p,q h (cid:12)(b1,B1),...,(Bq,Bq)i 2πi ZΩ (cid:12) where i = (−1)1/2, (cid:12) Πm Γ(b +B ξ) Πn Γ(1−a −A ξ) j=1 j j j=1 j j Θ(ξ) = , (6) Πhq Γ(1−b −Bihξ) Πp Γ(a +Aiξ) j=m+1 j j j=n+1 j j h ih i and an empty product is always interpreted as unity; m,n,p,q ∈ N with 0 0 ≤ n ≤ p,1 ≤ m ≤ q,A ,B ∈ R ,a ,b ∈ R or C(i = 1,...,p;j = 1,...,q) i j + i j such that A (b +k) 6= B (a −l−1),k,l ∈ N ;i = 1,...,n;j = 1,...,m, (7) i j j i 0 where we employ the usual notations:N = (0,1,2...),R = (−∞,∞),R = 0 + (0,∞) and C being the complex number field. The contour Ω is either L ,L , or L . These contours are defined in the monographs by Prud- −∞ +∞ iγ∞ nikov, Brychkov, and Marichev (1989), Mathai (1993), and Kilbas and Saigo (2004). However, the explicit definitions of these contours are given below to complete the definition of the H-function. (i) Ω = L is a left loop situated in a horizontal strip starting at the −∞ point −∞+iϕ−1 and terminating at the point −∞+iϕ with −∞ < ϕ < 2 1 ϕ < +∞; 2 (ii) Ω = L is a right loop situated in a horizontal strip starting at the +∞ point +∞+iϕ and terminating at the point +∞+iϕ with −∞ < ϕ < 1 2 1 ϕ < +∞. 2 (iii) Ω = L is a contour starting at the point γ −i∞ and terminating iγ∞ at the point γ +i∞, where γ ∈ R = (−∞,+∞). A detailed and comprehensive account of the H-function is available from the monograph by Mathai and Saxena (1978), Prudnikov, Brychkov, and Marichev (1989), and Kilbas and Saigo (2004). The relation connecting Ψ (z) and the H-function is given for the first time in the monograph by p q Mathai and Saxena (1978, p.11, Eq.1.7.8) as Ψ (a1,A1),...,(ap,Ap)|z = H1,p −z (1−a1,A1),...,(1−ap,Ap) , (8) p q (b1,B1),...,(bq,Bq) p,q+1 (0,1),(1−b1,B1),...,(1−bq,Bq) h i h (cid:12) i (cid:12) where pΨq(z) is Wright’s generalized hyperg(cid:12)eometric function (Wright, 1940); also see (Erd´elyi, Magnus, Oberhettinger, and Tricomi, 1953, Section 4.1), defined by means of the series representation in the form ∞ [Πp Γ(a +A r)]zr Ψ (z) = Ψ (ap,Ap)|z = j=1 j j , (9) p q p q (bq,Bq) [Πq Γ(b +B r)(r)!] h i rX=0 j=1 j j where z ∈ C,a ,b ∈ C,A ,B ∈ R = (−∞,∞),A ,B 6= 0 (i = 1,...,p;j = i j i j i j 1,...,q), q B − p A > −1;C being the set of complex numbers and j=1 j j=1 j Γ(z) is Euler’s gamma function. This function includes many special func- P P tions. It is interesting to observe that for A = B = 1,∀i and j, eq. (9) i j reduces to a generalized hypergeometric function F (z) as p q Πp Γ(a ) Ψ (ap,1) z = j=1 j F (a ,...,a ;b ,...,b ;z), (10) p q (bq,1) Πq Γ(b ) p q 1 p 1 q h (cid:12) i j=1 j (cid:12) (cid:12) where a 6= −ν(j = 1,...,p and ν = 0,1,2,....);p < q or p = q,|z| < 1. j Prior to (9), Wright (1933) introduced a special case of (9) in the form ∞ 1 zr Φ(a,b;z) = Ψ − z = , 0 1 (b,a) Γ(ar +b)(r)! h (cid:12) i rX=0 (cid:12) (cid:12) which widely occurs in problems of fractional diffusion. It has been shown by Saxena, Mathai, and Haubold (2004b) that (1,1) E (z) = Ψ z (11) α,β 1 1 (β,α) = H1,1h−z (cid:12)(cid:12)(0,i1) . (12) 1,2 (cid:12)(0,1),(1−β,α) h (cid:12) i (cid:12) If we further take β = 1 in (11) and (12), w(cid:12)e find that (1,1) E (z) = E (z) = Ψ z (13) α,1 α 1 1 (1,α) = H1,1 −z (0,1) h (cid:12)(cid:12), i (14) 1,2 (0,1),(0,α) (cid:12) h (cid:12) i (cid:12) (cid:12) where Re(α) > 0,α ∈ C. FromMathai andSaxena (1978)and Prudnikov, Brychkov, and Marichev (1989, p.355, Eq.2.25.3), it follows that the Laplace transform of the H- function is given by L tρ−1Hm,n ztσ (ap,Ap) = s−ρHm,n+1 zs−σ (1−ρ,σ),(ap,Ap) , (15) p,q (bq,Bq) p+1,q (bq,Bq) n h (cid:12) io h (cid:12) i (cid:12) (cid:12) where σ > 0,Re(s) >(cid:12) 0,Re[ρ + σmin (bj )] > 0,(cid:12)|argz| < [π/2]θ,θ > 0; 1≤j≤m Bj θ = n A − p A + m B − q B . j=1 j j=n+1 j j=1 j j=m+1 j By virtue of the cancellation law for the H-function (Mathai and Saxena, P P P P 1978), it can be readily seen that L−1 s−ρHm,n zsσ (ap,Ap) = tρ−1Hm,n zt−σ (ap,Ap),(ρ,σ) , (16) p,q (bq,Bq) p+1,q (bq,Bq) n h (cid:12) io h (cid:12) i (cid:12) (cid:12) where σ > 0,Re(s) > 0,Re[(cid:12)ρ+σmax (1−aj)] > 0,|argz(cid:12)| < 1πθ ,θ > 0; 1≤j≤n Aj 2 1 1 θ = θ−a. Two interesting special cases of (16) are worth mentioning. If we employ the identity (Mathai and Saxena, 1978) H1,0 x = xαexp(−x), (17) 0,1 (α,1) h (cid:12) i (cid:12) we obtain (cid:12) L−1[s−ρexp(−zsσ)] = tρ−1H1,0 zt−σ (ρ,σ) , (18) 1,1 (0,1) h (cid:12) i where Re(s) > 0,σ > 0. (cid:12) (cid:12) Further if we use the identity (Mathai and Saxena, 1978) H2,0 x = 2K (2x1/2), (19) 0,2 (ν,1)(−ν,1) ν 2 2 h (cid:12) i (cid:12) eq.(16) yields (cid:12) z2t−2σ 2L−1[s−ρK (zsσ)] = tρ−1H2,0 (ρ,2σ) , (20) ν 1,2 " 4 (ν2,1),(−ν2,1)# (cid:12) (cid:12) (cid:12) where Re(ρ) > 0,Re(z2) > 0,Re(s) > 0, and K (.) is the Bessel function of ν the third kind . In view of the result of Saxena, Mathai, and Haubold (2004a, p.49), also see Prudnikov, Brychkov, and Marichev (1989, p.355, eq.(2.25.3.2)), the cosine transform of the H-function is given by ∞ tρ−1cos(kt)Hm,n atµ (ap,Ap) dt (21) Z0 p,q h (cid:12)(bq,Bq)i = π Hn+1,m kµ (1−bq,Bq),((cid:12)(cid:12)1+2ρ,µ2) , kρ q+1,p+2" a (cid:12)(ρ,µ),(1−ap,Ap),(1+2ρ,µ2)# (cid:12) (cid:12) where Re[ρ+µmin ] > 1,|arg a| < 1(cid:12)πθ;θ > 0,θ is defined with the result 1≤j≤m 2 eq. (15). The Riemann-Liouville fractional integral of order ν is defined by (Miller and Ross, 1993, p.45) 1 t D−νf(t) = (t−u)ν−1f(u)du, (22) 0 t Γ(ν) Z0 where Re(ν) > 0. Following Samko, Kilbas, and Marichev (1990, p.37), we define the frac- tional derivative for α > 0 in the form 1 dn t f(u)du Dαf(t) = , n = [α]+1, (23) 0 t Γ(n−α)dtn (t−u)α−n+1 Z0 where [α] means the integral part of the number α. In particular, if 0 < α < 1, d 1 t f(u)du Dαf(t) = , (24) 0 t dtΓ(1−α) (t−u)α Z0 and if α = n ∈ N = {1,2,...,}, then Dnf(t) = Dnf(t)(D = d/dt), (25) 0 t is the usual derivative of order n. From Erd´elyi, Magnus, Oberhettinger, and Tricomi (1954b, p. 182), we have L D−νf(t) = s−νF(s), (26) 0 t n o where ∞ F(s) = L{f(t);s} = f∗(s) = exp(−st)f(t)dt,Re(s) > 0. (27) Z0 The Laplace transform of the fractional derivative is given by Oldham and Spanier (1974, p.134, eq.(8.1.3)) n L{ Dαf(t)} = sαF(s)− sr−1 Dα−rf(t)| . (28) 0 t 0 t t=0 r=1 X In certain boundary-value problems, the following fractional derivative of order α > 0 is introduced by Caputo (1969) in the form 1 t f(m)(τ)dτ Dαf(t) = , (29) t Γ(m−α) (t−τ)α+1−m Z0 m−1 < α ≤ m,Re(α) > 0,m ∈ N. dmf = ,if α = m. (30) dtm Caputo (1969) has given the Laplace transform of this derivative as m−1 L{Dαf(t);s} = sαF(s)− sα−r−1fr(0+),m−1 < a ≤ m. (31) t r=0 X The above formula is very useful in deriving the solution of differ-integral equations of fractional order governing certain physical problems of reaction and diffusion. We also need the Weyl fractional operator defined by 1 dn t f(u)du Dµf(t) = , (32) −∞ x Γ(n−µ)dtn (t−u)µ−n+1 Z−∞ where n = [µ] is an integral part of µ > 0. Its Fourier transform is (Metzler and Klafter, 2000, p.59, A.11, 2004) F { Dµf(x)} = (ik)µf˜(k), (33) −∞ x where we define the Fourier transform as ∞ ˜ h(q) = h(x)exp(iqx)dx. (34) Z−∞ Following the convention initiated by Compte (1996), we suppress the imag- inary unit in Fourier space by adopting the slightly modified form of above result in our investigations (Metzler and Klafter, 2000, p.59, A.12, 2004) F { Dµf(x)} = −|k|µf˜(k) (35) −∞ x instead of (33). Finally we also need the following property of the H-function (Mathai and Saxena, 1978) 1 Hm,n xδ (ap,Ap) = Hm,n x (ap,Ap/δ) , (36) p.q (bq,Bq) δ p,q (bq,Bq/δ) h (cid:12) i h (cid:12) i (cid:12) (cid:12) where δ > 0. (cid:12) (cid:12) 2 The Fractional Reaction-Diffusion Equation In this section we will investigate the solution of the reaction-diffusion equa- tion (37). The result is given in the form of the following theorem. Theorem. Consider the following fractional reaction-diffusion model ∂βN(x,t) = η DαN(x,t)+ϕ(x,t);η,t > 0,x ∈ R,0 < β ≤ 2, (37) ∂tβ −∞ x with the initial condition N(x,0) = f(x),N (x,0) = g(x) for x ∈ R, (38) t where N (x,0) means the first partial derivative of N(x,t) with respect to t ϕ evaluated at t = 0,η is a diffusion constant and ϕ(x,t) is a nonlinear function belonging to the area of reaction-diffusion. Then for the solution of (37), subject to the initial conditions (38), there holds the formula 1 ∞ N(x,t) = f˜(k)E (−η|k|αtβ)exp(−ikx)dk (39) β,1 2π Z−∞ 1 ∞ + tg˜(k)E (η|k|αtβ)exp(−ikx)dk β,2 2π Z−∞ 1 t ∞ + ξβ−1 ϕ˜(k,t−ξ)E (−η|k|αξβ)exp(−ikx)dkdξ. β,β 2π Z0 Z−∞ Proof. If we apply the Laplace transform with respect to the time variable t and Fourier transform with respect to space variable x and use the initial conditions and (32), then the given equation transforms into the form f˜(k)sβ−1 g˜(k)sβ−2 ϕ˜∗(k) N˜∗(k,s) = + + . (40) sβ +η|k|α sβ +η|k|α sβ +η|k|α On taking the inverse Laplace transform of (40) and applying the result sa−1 L−1 = tβ−αE (−atβ), (41) (a+sβ) β,β−α+1 where Re(s) > 0,Re(β −α+1) > 0, it is seen that N˜(k,t) = f˜(k)E (−η|k|αtβ)+g˜(k)tE (−η|k|αtβ) (42) β,1 β,2 t + ϕ˜(k,t−ξ)ξβ−1E (−η|k|αξβ)dξ. β,β Z0 The required solution (39) is now obtained by taking the inverse Fourier transform of (38). Thus, we have 1 ∞ N(x,t) = f˜(k)E (−η|k|αtβ)exp(−ikx)dk β,1 2π Z−∞ 1 ∞ + tg˜(k)E (−η|k|αtβ)exp(−ikx)dk β,2 2π Z∞ 1 t ∞ + ξβ−1 ϕ∗(k,t−ξ)E (−η|k|αξβ)exp(−ikx)dkdξ. β,β 2π Z0 Z−∞ This completes the proof of the theorem. If we set α = 2, we obtain the result given by Debnath (2003). Note. By virtue of the identity (12), the solution (39) can be expressed in terms of the H-function as can be seen from the solutions given in the special cases of the theorem in the next section. 3 Special Cases When g(x) = 0, then applying the convolution theorem of the Fourier trans- form to the solution (35), the theorem yields Corollary 1.1. The solution of fractional reaction-diffusion equation ∂β N(x,t)−η DαN(x,t) = ϕ(x,t), x ∈ R,t > 0,η > 0, (43) ∂tβ −∞ x subject to the initial conditions N(x,0) = f(x),N (x,0) = 0 for x ∈ R,1 < β ≤ 2, (44) t where η is a diffusion constant and ϕ(x,t) is a nonlinear function belonging to the area of reaction-diffusion, is given by ∞ N(x,t) = G (x−τ,t)f(τ)dτ (45) 1 Z−∞ t x + (t−ξ)β−1 G (x−τ,t−ξ)ϕ(τ,ξ)dτdξ, 2 Z0 Z0 where 1 ∞ G (x,t) = exp(−ikx)E (−η|k|αtβ)dk (46) 1 β,1 2π Z−∞ 1 ∞ = cos(kx)H1,1 kη1/αtβ/α (0,1/α) dk πα 1,2 (0,1/α),(0,β/α) Z0 h (cid:12) i 1 |x| (cid:12) = H2,1 (1,1/α),(1,β/α)(cid:12),(1,1/2) ,α > 0, α|x| 3,3 "η1/αtβ/α (1,1),(1,1/α),(1,1/2) # (cid:12) (cid:12) (cid:12) 1 ∞ G (x,t) = exp(−ikx)E (−η|k|αtβ)dk (47) 2 β,β 2π Z−∞ 1 ∞ = cos(kx)H1,1 kη1/αtβ/α (0,1/α) dk πα 1,2 (0,1/α),(1−β,β/α) Z0 h (cid:12) i 1 |x| (cid:12) = H2,1 (1,1/α),(β,β/α(cid:12)),(1,1/2) ,α > 0. α|x| 3,3 "η1/αtβ/α (1,1),(1,1/α),(1,1/2) # (cid:12) (cid:12) (cid:12) If we set f(x) = δ(x),ϕ ≡ 0,g(x) = 0, where δ(x) is the Dirac-delta function, then we arrive at the following Corollary 1.2. Consider the following reaction-diffusion model ∂βN(x,t) = η DαN(x,t),η > 0,x ∈ R,0 < β ≤ 1, (48) ∂tβ −∞ x with the initial condition N(x,t = 0) = δ(x), where η is a diffusion constant and δ(x) is the Dirac-delta function. Then the solution of (44) is given by 1 |x| N(x,t) = H2,1 (1,1/α),(1,β/α),(1,1/2) . (49) α|x| 3,3 "(ηtβ)1/α (1,1),(1,1/α),(1,1/2) # (cid:12) (cid:12) (cid:12) In the case β = 1, then in view of the cancellation law for the H-function (Mathai and Saxena, 1978), (49) gives rise to the following result given by

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.