NUMERICAL SOLUTIONS OF THE SYSTEM OF FRACTIONAL DIFFERENTIAL
EQUATIONS FOR OBSERVING EPIDEMIC MODELS
A THESIS SUBMITTED TO THE GRADUATE SCHOOL OF APPLIED SCIENCES
OF
NEAR EAST UNIVERSITY
By
LAWIN DHAHIR HAYDER HAYDER
In Partial Fulfillment of the Requirements for the Degree of Master of Science
in
Mathematics
NICOSIA, 2019
Lawin Dhahir Hayder
NUMERICAL SOLUTIONS OF THE SYSTEM OF FRACTIONAL NEU
Hayder DIFFERENTIAL EQUATIONS FOR OBSERVING EPIDEMIC MODELS 2019
NUMERICAL SOLUTIONS OF THE SYSTEM OF FRACTIONAL DIFFERENTIAL EQUATIONS
FOR OBSERVING EPIDEMIC MODELS
A THESIS SUBMITTED TO THE GRADUATE SCHOOL OF APPLIED SCIENCES
OF
NEAR EAST UNIVERSITY
By
LAWIN DHAHIR HAYDER HAYDER
In Partial Fulfillment of the Requirements for the Degree of Master of Science
in
Mathematics
NICOSIA, 2019
Lawin Dhahir Hayder Hayder : NUMERICAL SOLUTIONS OF THE SYSTEM OF FRACTIONAL DIFFERENTIAL EQUATIONS FOR OBSERVING EPIDEMIC MODELS
Approval of Director of Graduate School of Applied Sciences
Prof. Dr. Nadire ÇAVUŞ
We certify this thesis is satisfactory for the award of the degree of Masters of Science in Mathematics Department
Examining Committee in Charge
Prof. Dr. Evren Hınçal Committee Chairman, Department of Mathematics, NEU.
Assist. Prof. Dr. Bilgen Kaymakamzade Supervisor, Department of Mathematics, NEU.
Prof. Dr. Allaberen Ashyralyev Co-Supervisor, Department of Mathematics, NEU.
Assoc. Prof. Dr. Murat Tezer Department of Primary Mathematics Teaching, NEU.
Assist. Prof. Dr. Firudin Muradov Department of Mathematics, NEU.
I hereby declare that all information in this document has been obtained and presented in accordance with academic rules and ethical conduct. I also declare that, as required by these rules and conduct, I have fully cited and referenced all material and results that are not original to this work.
Name, Last name: Lawin Hayder Signature:
Date:
ACKNOWLEDGMENTS
First and foremost, I would like to dedicate this thesis to my parents and to thank them from the bottom of my heart for their love, moral, material and spiritual support throughout my life. Thank you both for giving me the strength to reach for the top and achieve my dreams.
I would like to express my profound appreciation to my supervisors, Prof. Dr. Allaberen Ashyralyev and Assist. Prof. Dr. Bilgen Kaymakamzade for their support and professional guidance throughout this thesis project. I would like to also thank the Committee Chairman, Department of Mathematics, Prof. Dr. Evren Hınçal.
I would like to appreciate my lovely brother for being the most supportive and for
standing by me at all times. I would like to also thank my sisters. Lastly, to all my
colleagues, friends and everyone who has helped me. I wish for all of the success and
happiness.
To my family…
ABSTRACT
In this thesis, the system of fractional differential equations for observing epidemic models problems are investigated. Applying Fourier series, Laplace transform and Fourier transform methods, the solutions of six problems are obtained. First and second order of accuracy difference schemes are presented for the solution of the one-dimensional epidemic models problem and the numerical procedure for implementation of these schemes is discussed.
Keywords: Epidemic models; fractional differential equations; Fourier series method;
Laplace transform solution; difference scheme
ÖZET
Bu tez çalışmasında, epidemik model problemleri için kesirli türevli diferansiyel denklem sistemleri incelenmiştir. Fourier serileri, Laplace dönüşümü ve Fourier dönüşümü yöntemlerini uygulama, ile altı problemlerin çözümleri bülümüştür. Birinci ve ikinci dereceden doğruluk farkı şemaları tek boyutlu epidemik model probleminin çözümü için sunulmuş ve bu şemaların uygulanmasına yönelik sayısal prosedür ele alınmıştır.
Anahtar Kelimeler: Epidemik modeller; kesirli türevli diferansiyel denklemler; Fourier
serisi yöntemi; Laplace dönüşümü çözümü; fark şeması
TABLE OF CONTENTS
ACKNOWLEDGMENTS
………..….……….……..……..….……..………..……….… iiABSTRACT
………..….………..….……..……..….……..………..………..…...….……. ivÖZET
………..….………..….……..……..….……..………..………..….………..….…… vTABLE OF CONTENTS
………..….………..….……..……..….……..……….……….. viLIST OF TABLES
..………..….………..….……..……..….……..………..………..… viiCHAPTER 1: INTRODUCTION
………..….………..….……..……..….……..………..……….. 1CHAPTER 2: METHODS OF SOLUTIONS OF SYSTEM OF FRACTIONAL PARTIAL DIFFERENTIAL EQUATONS………..
72.1 Fourier Series Method ..………...……….……...…………..….……..
7
2.2 Laplace Transform method ………...…………..…….………
322.3 Fourier Transform Method ..……….………....……….………….
46CHAPTER 3: FINITE DIFFERENCE METHOD FOR THE SOLUTION OF
FRACTIONAL PARTIAL DIFFERNETIAL EQUATIONS
………..……….. 52CHAPTER 4: CONCLUSION
..…...……….………..….……....……..………. 67REFERENCES
…….………..….……..…………..………..……..…...……..………..……… 68APPENDICES
..……..………..………..…..….…………..………. 71LIST OF TABLES
Table 3.1: Error analysis
………..….……..……..………..….……..…………..…..………. 66Table 3.2: Error analysis
………..….……..……..………..….……..…………..…..………. 66CHAPTER 1 INTRODUCTION
Fractional differential equations take an important role in applied mathematics (A. Guner
& S. Bekir, 2017), engineering (Y. Xiao-Jun, 2011), physics (M. Bayram, A. Sec¸er, & A.
Adıg¨uzel, 2017), biology (V. Srivastava, K. Kumar, M. K. Awasthi, & B. K. Singh, 2014) and other fields of science. The system of a fractional differential equation is used in various fields in applied science.
Fractional differential equations are formed from fractional calculus which is a branch of mathematics that deals with the properties of integrals and derivatives of non-integer orders (W. F. Ames, 1999). The concept was first mentioned in a letter sent to L’Hopital by Leibniz in 1695, where the idea of semi-derivatives was suggested. Other famous mathematicians, Liouville, Grunwald, Riemann, etc. have proposed original approaches to improve fractional calculus over time (K. Shukla & P. Sapra, 2019). So that fractional differential equations can be applied in the various fields listed above, it is important to develop methods of solutions. Several methods already employed are perturbation techniques (O. Abdulaziz, I. Hashim, & S. Momani, 2008), variational iterative method (V.
Gejji & S. Bhalekar, 2007), decomposition methods (H. Jafari & V. Gejji, 2006), integral transform methods (B. Sontakke, G. Kamble, & S. Acharya, 2017), and numerical methods (Y. Yan, K. Pal, & N. Ford, 2014).
An epidemiological study is important to help understand the impact of infectious diseases in a community. The mathematical model is used to analyze data and study the spread and transmission of infectious diseases. With new ideas in epidemiology, we can investigate models by model building, perform estimation of parameters, check sensitivity of models by varying parameters, and compute their numerical simulations. Over the course of history, we have seen examples of epidemic outbreaks infecting large numbers of people.
Examples of such are the 1918 Spanish flu outbreak which killed millions and more recently, we have had cases such as HIV/AIDs, SARS, and Ebola outbreaks. World Health Organization (WHO) reports indicating that an estimated 13 million people worldwide die
from infectious diseases (WHO, 2012). In light of this, the issue of developing realistic epidemic spreading models and controlling the outbreak and spread of infectious diseases should be considered paramount. The research of this kind helps to understand the ratio of disease spread in the population and to control their parameters (I. Abubakar et al., 2012; B.
T. Grenfell, 1992).
Various classical epidemic models have been proposed and studied such as SIR, SIS, SEIR, and SIRS. Kermack and McKendrick developed the first known mathematical and population-level model applicable in studying influenza outbreaks (N. Baca¨er, 2011). The model contains three groups: Susceptible (containing those individuals who have a high tendency of contracting the disease), infectious (who currently have the disease who can transmit it to the susceptible individuals), and recovered (this contains individuals that have previously contracted the infection and have now been removed from the epidemic either by recovery or by death). This model is known as the SIR model.
In the SIR model, an infected individual is brought right into a population in which all the individuals are all susceptible. Vertical transmission (transmission of a disease by an infected parent to their children) may be integrated into the SIR model if we consider that a portion of the children of the infected individuals is infected at birth (H. W. Hethcote, 2000;
S. Waziri, S. Massawe, & D. Makinde, 2012). This will come in handy when considering a case such as the HIV mother-to-child transmission (MTCT) epidemic. The dynamics of diseases like measles (A. Ahmad, 2018) and influenza (G. H. Li & Y. X. Zhang, 2017) have also been explained using the modified SIR model. This model may be extended to add a state of temporary immunity where individuals who have been removed are returned to the susceptible class after they’ve missed out on their immunity. This extension is called the SIRS model as has been stated earlier.
In the SIS model, no perennial immunity from the infection exists, individuals can be infected again and can return to the susceptible class. El-Saka studied the stability of equilibrium points for a fractional-order SIS epidemic model (H. A. A. El-Saka, 2014).
According to him using fractional differential equations can aid in reducing errors that arise from the neglected parameters in modeling real-life phenomena. The numerical solutions of
the models were given and he was able to verify the theoretical analysis using numerical simulations. In the paper, Prakash et al. (B. Prakash, A. Setia, & D. Alapatt, 2017) employed a fractional-order nonlinear SEIR model with a non-constant population mathematical model to model infectious diseases. They proposed a faster and simpler numerical methods based on Harr wavelets to solve the SEIR model deriving and validating the error bounds.
Jun-Jie Wang et al. (J. J. Wang, K. H. Reilly, H. Han, Z. H. Peng, & N. Wang, 2010) employed a deterministic transmission model for the Chinese HIV MTCT epidemic to demonstrate how it is affected by some key parameters. They presented a system of ordinary differential equations and their solutions were derived using this model. HIV positive children delivered by the infected mothers were taken as the susceptible group (S ), the transmission rate for HIV positive mothers was (β), and the screening proportion (α) was defined as the percentage of pregnant women who have tested HIV positive to the proportion of HIV positive pregnant cases. They found out that in China, these three factors have the biggest influence on the epidemic. This led them to conclude that proper testing for pregnant women, strengthening prevention of mother-to-child transmission (PMTCT) interventions, and reducing the amount of HIV positive occurrences in women of reproductive ages are steps that will aid in curbing the HIV MTCT epidemic in China.
Ashyralyev et al. (A. Ashyralyev, E. Hincal, & B. Kaymakamzade, 2018) studied the stability of initial-boundary value problem for the system of partial differential equations for observing HIV mother to child transmission epidemic models. The study was aimed at helping to understand the estimation of the transmission rate from mathematical models representing the dynamics of the population of infectious diseases using numerical methods. In their paper, various initial-boundary-value problems for the system of partial differential equations they presented as the initial-value problem for the system of ordinary
differential equations
du1(t)
dt + αu1(t)+ Au1(t) = f1(t),
du2(t)
dt + βu2(t) − β1u1(t)+ cAu2(t)= f2(t),
du3(t)
dt + γu3(t) − γ1u1(t)+ eAu3(t) = f3(t),
du4(t)
dt + du4(t) − d1u3(t) − d2u2(t)+ lAu4(t)= f4(t),
0 < t < T , um(0)= ϕm, m = 1, 2, 3, 4
(1.1)
in a Hilbert space H with a self-adjoint positive definite operator A. They proved theorems on stability by applying the operator approach. Moreover, difference schemes for approximate solution of system (1.1) were presented and theorems on stability of these difference schemes were proved. Numerical result was given.
Ameera Masour (2018) in her master’s thesis obtained the solution of a system of partial differential equations by solving analytically using Fourier series, Laplace transform and Fourier transform methods. The first order of accuracy difference scheme for the numerical solution of the initial-boundary value problem for one-dimensional partial differential equations was presented. Numerical results were given.
In the present study, systems of fractional differential equations which an extension of partial differential equations are used to modify the system (1.1). We considered stable solution of the initial value problem for solving of fractional differential equations for observing epidemic models and we used classical methods to solve the initial value problem for the
system of one dimensional partial differential equation
∂u1(t,x)
∂t + αDt12u1(t, x) −∂2u∂x1(t,x)2 = f1(t, x),
∂u2(t,x)
∂t + βDt12u2(t, x) − β1u1(t, x) − −∂2u∂x2(t,x)2 = f2(t, x),
∂u3(t,x)
∂t + γDt12u3(t, x) − γ1u1(t, x) −∂2u∂x3(t,x)2 = f3(t, x),
∂u4(t,x)
∂t + dDt12u4(t, x) − d1u3(t, x) − d2u2(t, x) − ∂2u∂x4(t,x)2 = f4(t, x),
0 < t < T , 0 < x < π,
um(t, 0)= um(t, π)= 0, 0 6 t 6 T,
um(0, x)= ϕm(x), 06 x 6 π, m = 1, 2, 3, 4.
(1.2)
Here,
Dαt = Dα0+
is the standard Riemann-Liouville’s derivative of order α ∈ (0, 1). This system of fractional differential equations corresponding to the Basset problem (A. Ashyralyev, 2011). The present work aims to study numerical solutions of the initial value problem for the system of fractional differential equations observing the HIV mother-to-child transmission epidemic. The first and second-order of accuracy difference schemes for the numerical solution of the system of one-dimensional fractional partial differential equations are presented and the illustrative numerical results are provided.
The thesis organization is as follows. Chapter 1 is an introduction. The history of epidemiology problems with the system of fractional partial differential equations is presented. In Chapter 2, the methods of solution of the system of fractional partial differential equations by solving analytically using Fourier series, Laplace transform, and Fourier transform methods are presented. In chapter 3, the first and second-order of
accuracy single-step difference schemes for the approximate solutions of one-dimensional epidemiology problem for the system of fractional partial differential equations are presented. Numerical results are provided by the Gauss elimination method. Chapter 4 is a conclusion.
CHAPTER 2
METHODS OF SOLUTION OF SYSTEM OF FRACTIONAL PARTIAL DIFFERENTIAL EQUATIONS
It is known that the system of fractional partial differential equations can be solved analytically by Fourier series, Laplace transform and Fourier transform methods. In this section, three different analytical methods by examples are illustrated.
2.1 Fourier Series Method
First, we consider the Fourier series method for the solution of the mixed problems for the system of fractional partial differential equations.
Example 2.1. Consider the mixed problem for the system of fractional partial differential equations
∂u1(t,x)
∂t + αDt12u1(t, x) − ∂2u∂x1(t,x)2 = (2t + t2+ α83t√32π) sin x,
∂u2(t,x)
∂t + βDt12u2(t, x) − β1u1(t, x) − ∂2u∂x2(t,x)2
= (2t + β83t√32π −β1t2+ t2) sin x,
∂u3(t,x)
∂t + δDt12u3(t, x) − δ1u1(t, x) −∂2u∂x3(t,x)2
= (2t + δ83 t√32π −δ1t2+ t2) sin x,
∂u4(t,x)
∂t + dDt12u4(t, x) − d1u3(t, x) − d2u2(t, x) −∂2u∂x4(t,x)2
= (2t + d83 t√32π − d1t2− d2t2+ t2) sin x, 0 < t < 1, 0 < x < π,
u1(t, 0)= u2(t, 0)= u3(t, 0)= u4(t, 0)= 0, 0 ≤ t ≤ 1, u1(t, π)= u2(t, π)= u3(t, π)= u4(t, π)= 0, 0 ≤ t ≤ 1, u1(0, x)= u2(0, x)= u3(0, x) = u4(0, x)= 0, 0 ≤ x ≤ π.
(2.1)
Solution. To solve this problem, we consider the Sturm-Liouville problem
−u00(x) − λu(x)= 0, 0 < x < π, u(0) = u(π) = 0, (2.2) generated by the space operator of problem (2.1). It is clear to see that the solution of the Sturm-Liouville problem (2.2) is
λk = −k2, uk(x)= sin kx, k = 1, 2, ....
Then, with using Fourier series solution of problem (2.1) by formula
u1(t, x)= P∞
k=1Ak(t) sin kx,
u2(t, x)= P∞
k=1Bk(t) sin kx,
u3(t, x)= P∞
k=1Ck(t) sin kx,
u4(t, x)= P∞
k=1Dk(t) sin kx,
(2.3)
where Ak(t), Bk(t), Ck(t) and Dk(t) are unknown functions. Putting system (2.3) to the system (2.1), we obtain
∞
P
k=1A0k(t) sin kx+ αP∞
k=1D
1 2
t Ak(t) sin kx+ P∞
k=1k2Ak(t) sin kx
= (2t + t2+ α83 t√32π) sin x,
∞
P
k=1B0k(t) sin kx+ βP∞
k=1D
1 2
t Bk(t) sin kx − β1
∞
P
k=1Ak(t) sin kx+ P∞
k=1k2Bk(t) sin kx
= (2t + β83 t√32π −β1t2+ t2) sin x,
∞
P
k=1Ck0(t) sin kx+ δP∞
k=1D
1 2
t Ck(t) sin kx − δ1
∞
P
k=1Ak(t) sin kx+ P∞
k=1k2Ck(t) sin kx
= (2t + δ83t√32π −δ1t2+ t2) sin x,
∞
P
k=1D0k(t) sin kx+ dP∞
k=1D
1 2
t Dk(t) sin kx − d1
∞
P
k=1Ck(t) sin kx − d2
∞
P
k=1Bk(t) sin kx +P∞
k=1k2Dk(t) sin kx= (2t + d83 t√32π − d1t2− d2t2+ t2) sin x,
0 < t < 1, 0 < x < π.
Applying the initial conditions to the system (2.3), we can write
u1(0, x)= P∞
k=1Ak(0) sin kx= 0,
u2(0, x)= P∞
k=1Bk(0) sin kx= 0,
u3(0, x)= P∞
k=1Ck(0) sin kx= 0,
u4(0, x)= P∞
k=1Dk(0) sin kx= 0,
0 ≤ x ≤ π.
Equating coefficients sin kx, k = 1, 2, ..., we get
A01(t)+ αDt12A1(t)+ A1(t)= 2t + t2+ α83√t32π,
B01(t)+ βDt12B1(t) − β1A1(t)+ B1(t)
= 2t + β83 t√32π −β1t2+ t2,
C10(t)+ δDt12C1(t) − δ1A1(t)+ C1(t)
= 2t + δ83 t√32π −δ1t2+ t2,
D01(t)+ dDt12D1(t) − d1C1(t) − d2B1(t)+ D1(t)
= 2t + d83t√32π − d1t2− d2t2+ t2,
0 < t < 1, A1(0)= B1(0)= C1(0)= D1(0)= 0
(2.4)
and for k , 1
A0k(t)+ αDt12Ak(t)+ k2Ak(t) = 0,
B0k(t)+ βDt12Bk(t) − β1Ak(t)+ k2Bk(t)= 0,
Ck0(t)+ δDt12Ck(t) − δ1Ak(t)+ k2Ck(t)= 0,
D0k(t)+ dDt12Dk(t) − d1Ck(t) − d2Bk(t)+ k2Dk(t)= 0,
0 < t < 1, Ak(0)= Bk(0)= Ck(0)= Dk(0)= 0.
(2.5)
So, we have initial value problems for the system of ordinary differential equations. For solving the systems the Laplace transform method is applied.
Here and in future we assume that
L{Ak(t)}= Ak(s),
L {Bk(t)}= Bk(s),
L {Ck(t)}= Ck(s),
L {Dk(t)}= Dk(s).
Taking Laplace transform of both sides of system of fractional partial differential equations in the systems (2.4), (2.5) and using the following conditions Ak(0) = Bk(0) = Ck(0) = Dk(0)= 0, k ≥ 1, we obtain the following systems of algebraic equations
sA1(s)+ αs12A1(s)+ A1(s)= s22 + s23 + α 2
s52, sB1(s)+ βs12B1(s) − β1A1(s)+ B1(s)= s22 + β 2
s52
−β1s23 + s23,
sC1(s)+ δs12C1(s) − δ1A1(s)+ C1(s)= s22 + δ2
s52
−δ1s23 + s23,
sD1(s)+ ds12D1(s) − d1C1(s) − d2B1(s)+ D1(s)
= s22 + d 2
s
5 2
− d12 s3 − d22
s3 + s23,
(2.6)
and for k , 1
(s+ αs12 + k2)Ak(s)= 0,
(s+ βs12 + k2)Bk(s) − β1Ak(s)= 0,
(s+ δs12 + k2)Ck(s) − δ1Ak(s)= 0,
(s+ ds12 + k2)Dk(s) − d1Ck(s) − d2Bk(s)= 0.
(2.7)
For k , 1 from system (2.7) it follows Ak(s) = Bk(s) = Ck(s) = Dk(s) = 0. Taking the inverse Laplace transform with respect to t, we get
Ak(t)= Bk(t) = Ck(t)= Dk(t)= 0.
For finding A1(t), B1(t), C1(t) and D1(t), we use the system (2.6). First, we obtain A1(s). We have that
(s+αs12 + 1)A1(s)= s22 + s23 + α 2
s52
. Therefore,
A1(s)= 2!
s3. (2.8)
Second, we obtain B1(s). Using formula (2.8) in the second equation, we get
(s+βs12 + 1)B1(s) − β12!s3 = 2!s3(s+ βs12 −β1+ 1).
Therefore, B1(s)= 2!
s3. (2.9)
Third, we obtain C1(s). Applying formula (2.8) in the third equation, we obtain
(s+δs12 + 1)C1(s) − δ12!s3 = 2!s3(s+ δs12 −δ1+ 1)
or
C1(s)= 2!
s3. (2.10)
Fourth, we obtain D1(s). Applying formula (2.9) and (2.10) in the last equation, we get (s+ ds12 + 1)D1(s) − d12
s3 − d22 s3 = 2
s3(s+ ds12 − d1− d2+ 1) or
D1(s)= 2!
s3. (2.11)
Finally, applying formulas (2.8), (2.9), (2.10) and (2.11) and taking the inverse Laplace transform with respect to t, we get
A1(t)= B1(t) = C1(t)= D1(t) = t2.
Therefore, the exact solution of the problem (2.1) is
u1(t, x)= A1(t) sin x= t2sin x,
u2(t, x)= B1(t) sin x= t2sin x,
u3(t, x)= C1(t) sin x= t2sin x,
u4(t, x)= D1(t) sin x= t2sin x.
Using similar procedure we can get the solution of the following initial boundary value problem
∂u1(t,x)
∂t + α Dt12u1(t, x) −
n
P
r=1ar∂2u1(t,x)
∂x2r = f1(t, x),
∂u2(t,x)
∂t + β Dt12u2(t, x) − β1u1(t, x) −
n
P
r=1ar∂2u2(t,x)
∂x2r = f2(t, x),
∂u3(t,x)
∂t + δDt12 u3(t, x) − δ1u1(t, x) −
n
P
r=1ar
∂2u3(t,x)
∂x2r = f3(t, x),
∂u4(t,x)
∂t + dDt12u4(t, x) − d1u3(t, x) − d2u2(t, x) −
n
P
r=1ar∂2u4(t,x)
∂xr2
= f4(t, x),
x= (x1, ..., xn) ∈Ω, 0 < t < T,
u1(0, x)= ϕ(x), u2(0, x)= ψ (x) , u3(0, x)= ξ(x), u4(0, x)= λ(x), x= (x1, ..., xn) ∈Ω,
u1(t, x)= u2(t, x)= u3(t, x)= u4(t, x)= 0, x ∈ S, 0 ≤ t ≤ T
(2.12)
for the system of multidimensional fractional partial differential equations. Note that ar >
a0 > 0 and fk(t, x) , k = 1, 2, 3, 4
t ∈(0, T ) , x ∈Ω , ϕ(x), ψ (x) , ξ(x), λ(x), x ∈ Ω are given smooth functions. Here and in future Ω is the unit open cube in the n−dimensional Euclidean space Rn(0 < xk < 1, 1 ≤ k ≤ n) with the boundary
S,Ω = Ω ∪ S.
Note that the Fourier series method described in solving (2.12) can be used only in the case when (2.12) has constant coefficients.
Example 2.2. Consider the mixed problem for the system of fractional partial differential equations
∂u1(t,x)
∂t + αDt12u1(t, x) − ∂2u∂x1(t,x)2 = (2t + t2+ α83t√32π) cos x,
∂u2(t,x)
∂t + βDt12u2(t, x) − β1u1(t, x) − ∂2u∂x2(t,x)2
= (2t + β83t√32π −β1t2+ t2) cos x,
∂u3(t,x)
∂t + δDt12u3(t, x) − δ1u1(t, x) −∂2u∂x3(t,x)2
= (2t + δ83 t√32π −δ1t2+ t2) cos x,
∂u4(t,x)
∂t + dDt12u4(t, x) − d1u3(t, x) − d2u2(t, x) −∂2u∂x4(t,x)2
= (2t + d83 t√32π − d1t2− d2t2+ t2) cos x,
0 < t < 1, 0 < x < π,
u1(0, x)= u2(0, x)= u3(0, x) = u4(0, x)= cos x, 0 ≤ x ≤ π,
u1x(t, 0)= u2x(t, 0)= u3x(t, 0)= u4x(t, 0)= 0, 0 ≤ t ≤ 1,
u1x(t, π)= u2x(t, π)= u3x(t, π)= u4x(t, π)= 0, 0 ≤ t ≤ 1.
(2.13)
Solution. To solve the mixed problem, we consider the Sturm-Liouville problem
−u00(x) − λu(x)= 0, 0 < x < π, ux(0)= ux(π) = 0, (2.14) generated by the space operator of problem (2.13). It is clear to see that the solution of this Sturm-Liouville problem (2.14) is
λk = −k2, uk(x)= cos kx, k = 0, 1, ....
Then, with using Fourier series solution of problem (2.13) by formula
u1(t, x)= P∞
k=0Ak(t) cos kx,
u2(t, x)= P∞
k=0Bk(t) cos kx,
u3(t, x)= P∞
k=0Ck(t) cos kx,
u4(t, x)= P∞
k=0Dk(t) cos kx,
(2.15)
where Ak(t), Bk(t), Ck(t) and Dk(t) are unknown functions. Putting system (2.15) to the system (2.13), we obtain
∞
P
k=0A0k(t) cos kx+ αP∞
k=0D
1 2
t Ak(t) cos kx+ P∞
k=0k2Ak(t) cos kx = (2t + t2+ α83t√32π) cos x,
∞
P
k=0B0k(t) cos kx+ β P∞
k=0D
1 2
t Bk(t) cos kx − β1
∞
P
k=0Ak(t) cos kx+ P∞
k=0k2Bk(t) cos kx
= (2t + β83 t√32π −β1t2+ t2) cos x,
∞
P
k=0C0k(t) cos kx+ δP∞
k=0D
1 2
tCk(t) cos kx − δ1
∞
P
k=0Ak(t) cos kx+ P∞
k=0k2Ck(t) cos kx
= (2t + δ83 t√32π −δ1t2+ t2) cos x,
∞
P
k=0D0k(t) cos kx+ d P∞
k=0D
1 2
t Dk(t) cos kx − d1
∞
P
k=0Ck(t) cos kx − d2
∞
P
k=0Bk(t) cos kx + P∞
k=0k2Dk(t) cos kx= (2t + d83t√32π − d1t2− d2t2+ t2) cos x,
0 < t < 1, 0 < x < π.
Applying the initial conditions to the system (2.15), we can write
u1(0, x)= P∞
k=0Ak(0) cos kx= cos x,
u2(0, x)= P∞
k=0Bk(0) cos kx= cos x,
u3(0, x)= P∞
k=0Ck(0) cos kx= cos x,
u4(0, x)= P∞
k=0Dk(0) cos kx= cos x,
0 ≤ x ≤ π.
Equating coefficients cos kx, k = 0, 1, ..., we get
A01(t)+ αDt12A1(t)+ A1(t)= 2t + t2+ α83√t32π,
B01(t)+ βDt12B1(t) − β1A1(t)+ B1(t)
= 2t + β83 t√32π −β1t2+ t2,
C10(t)+ δDt12C1(t) − δ1A1(t)+ C1(t)
= 2t + δ83 t√32π −δ1t2+ t2,
D01(t)+ dDt12D1(t) − d1C1(t) − d2B1(t)+ D1(t)
= 2t + d83t√32π − d1t2− d2t2+ t2,
0 < t < 1, A1(0)= B1(0)= C1(0)= D1(0)= 0
(2.16)
and
for k , 1
A0k(t)+ αDt12Ak(t)+ k2Ak(t) = 0,
B0k(t)+ βDt12Bk(t) − β1Ak(t)+ k2Bk(t)= 0,
Ck0(t)+ δDt12Ck(t) − δ1Ak(t)+ k2Ck(t)= 0,
D0k(t)+ dDt12Dk(t) − d1Ck(t) − d2Bk(t)+ k2Dk(t)= 0,
0 < t < 1, Ak(0)= Bk(0)= Ck(0)= Dk(0)= 0.
(2.17)
Taking Laplace transform of both sides of system fractional partial differential equations in the systems (2.16) and (2.17) and using the following conditions Ak(0) = Bk(0) = Ck(0) = Dk(0)= 0, k ≥ 1, we obtain the following systems of algebraic equations
sA1(s)+ αs12A1(s)+ A1(s)= s22 + s23 + α 2
s52
,
sB1(s)+ βs12B1(s) − β1A1(s)+ B1(s)= s22 + β 2
s52
−β1s23 + s23,
sC1(s)+ δs12C1(s) − δ1A1(s)+ C1(s)= s22 + δ2
s52
−δ1s23 + s23,
sD1(s)+ ds12D1(s) − d1C1(s) − d2B1(s)+ D1(s)
= s22 + d 2
s
5 2
− d1s23 − d2s23 + s23
(2.18)