Studies in Computational Intelligence 1109

Harendra Singh
Hemen Dutta Editors

Computational
Methods
for Biological
Models
Studies in Computational Intelligence

Volume 1109

Series Editor
Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland
The series “Studies in Computational Intelligence” (SCI) publishes new develop-
ments and advances in the various areas of computational intelligence—quickly and
with a high quality. The intent is to cover the theory, applications, and design methods
of computational intelligence, as embedded in the fields of engineering, computer
science, physics and life sciences, as well as the methodologies behind them. The
series contains monographs, lecture notes and edited volumes in computational
intelligence spanning the areas of neural networks, connectionist systems, genetic
algorithms, evolutionary computation, artificial intelligence, cellular automata, self-
organizing systems, soft computing, fuzzy systems, and hybrid intelligent systems.
Of particular value to both the contributors and the readership are the short publica-
tion timeframe and the world-wide distribution, which enable both wide and rapid
dissemination of research output.
Indexed by SCOPUS, DBLP, WTI Frankfurt eG, zbMATH, SCImago.
All books published in the series are submitted for consideration in Web of Science.
Harendra Singh · Hemen Dutta
Editors

Computational Methods
for Biological Models
Editors
Harendra Singh Hemen Dutta
Department of Mathematics Department of Mathematics
Post Graduate College Ghazipur Gauhati University
Ghazipur, Uttar Pradesh, India Guwahati, Assam, India

ISSN 1860-949X ISSN 1860-9503 (electronic)

Studies in Computational Intelligence
ISBN 978-981-99-5000-3 ISBN 978-981-99-5001-0 (eBook)
https://doi.org/10.1007/978-981-99-5001-0

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Singapore Pte Ltd. 2023

This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
Preface

This book is designed to include several topics in the areas of biological models and
their treatments using computational methods. Biological modeling is very useful to
understand real-life problems. The chapters include useful computational methods
for solving various types of biological models as well as their significance and rele-
vance in other scientific areas of study and research. There are ten chapters in this
book, and they are organized as follows.
Chapter “Exponential-Time-Differencing Method for the Solution of Diffusive
HIV-I Model” deals with time-dependent partial differential equation (PDE) prob-
lems. The modeling and application of some epidemiological diseases have been
the subject of activities. Most of these models mainly exist in the form of ordinary
differential equations (ODE), due to some of the challenges encountered in solving
systems with coupled partial differential equations (PDEs). At first, the problem is
discretized with respect to space in such that, the arising system of ODEs is solved
using the novel fourth-order ETD method of Runge-Kutta type. A typical example
is given to the HIV model in the diffusive case which could rarely be found in the
literature. Numerical results which reflect the physical behavior of the model under
consideration are given in one and two dimensions.
Chapter “An Effective Technique for Solving a Model Describing Biological
Species Living Together” presents a reliable scheme to obtain an analytical approx-
imation to the solution of a system of nonlinear integro-differential equations that
arise in biology. The dynamics of the populations of two distinct species are char-
acterized by this system. The suggested strategy involves two steps. Construct an
integral operator first, and then apply Halpern’s iteration scheme to the constructed
operator to develop the required iterative approach. This approach’s convergence
is also addressed. Numerous numerical examples are taken into consideration to
demonstrate the effectiveness of the strategy. The proposed approach has several
key advantages over existing approaches, including the ability to handle nonlin-
earity without the use of Adomian polynomials, the ability to solve the problem
without the use of the Lagrange multiplier or constrained variations, and the ability
to take into account both interval endpoints. In contrast to existing semi-analytical

v
vi Preface

methods, the suggested approach addresses the problems without requiring lineariza-
tion, discretization, or perturbation assumptions. Three test problems are taken into
account to confirm the method’s efficacy.
Chapter “Numerical Solution for a Tumor Cells Dynamics Within Their
Micro-environment” focuses to include a delay term to model key transformations
taking place within the dynamics modeling the interaction among transformed epithe-
lial cells (TECs), fibroblasts, myofibroblasts, transformed growth factor (TGF−β),
and epithelial growth factor (EGF), in silico, in a setup mimicking experiments in
a tumor chamber invasion assay. This consideration of such experiments resulted
into a derivation of system of nonlinear quasi time-depended delay parabolic partial
differential equations. The well-pose of the unique solution of the resulting system of
nonlinear quasi time-depended delay parabolic partial differential equations presents
that an extension of the existing Gronwall’s inequality for ordinary differential equa-
tions to Gronwall’s inequality for linear, delay ordinary differential equations is thus
derived. The asymptotic stability and Hopf bifurcation analysis are established. Based
on the asymptotic stability condition, a novel numerical method is derived, analyzed,
and implemented, and improved results are presented for discussion. Hence, based
on the obtained results, one strongly believes that the contribution in this chapter has
a great deal toward the process of crafting a drug that can slow and/or confine tumor
invasion.
Chapter “A Mathematical Model to Study Regulatory Properties and Dynamical
Behaviour of Glycolytic Pathway Using Bifurcation Analysis” deals with a dynam-
ical system based on mass balance equations and S-system representation to study
the regulatory properties, rate control distributions, and dynamical behavior of reac-
tions in this glycolytic pathway. This representation involves several parameters.
Few standard tools from the bifurcation analysis such as phase-portraits, time series
plots, the Lyapunov exponents, and d∞ plots have been employed to investigate
the stability of the pathway with respect to certain chosen parameters. The study is
further extended by eliminating the glycogen branch from the original pathway and
by adding an external perturbation into the system.
Chapter “On Solutions of Biological Models Using Reproducing Kernel Hilbert
Space Method” presents the reproducing kernel Hilbert space method (RKHSM) for
some systems that have significant applications in biology. The proposed method’s
error estimations and convergence analyses are discussed. The assessment of the
RKHSM is made by testing some illustrative applications. The results suggest that
the RKHSM is an efficient and highly convenient method to solve the fractional
systems arising in biology.
Chapter “A Study of the Fractional Tumour–Immune Unhealthy Diet Model Using
the Pseudo-operational Matrix Method” deals with the fractional tumor-immune
unhealthy diet (TIUNHD) model. The fractional-order Laguerre operational matrix-
based method is used to solve it. The fractional derivative of Caputo is used in this
chapter. In this method, first the pseudo-operational matrices of fractional, integer
order integration are constructed, and then used these pseudo-operational matrices to
approximate the unknown solutions of the given system of nonlinear fractional differ-
ential equations model. This results in an algebraic system of equations, which can be
Preface vii

easily solved by Newton’s iterative method. A numerical experiment demonstrates

the proposed algorithm’s applicability to solving the TIUNHD model.
Chapter “Analysis of a Fractional Stage-Structured Model With Crowley–Martin
Type Functional Response by Lagrange Polynomial Based Method” focuses on
the dynamics of a stage-structured predator-prey system that replicates interactions
between two densities of prey and predator populations. The adult predator popula-
tion and the juvenile predator are the two compartments that make up the predator
population in the model. The predator relies on both prey and juvenile predator,
which is another element of the paradigm that can be termed cannibalism. Crowley-
Martin type functional denotes the nature of the interaction between prey and adult
predators, while Holling type-I functional response denotes the nature of contact
between juvenile and adult predators. The concept of memory is introduced in the
form of the Caputo fractional derivative to reflect the complicated dynamics of inter-
action among the species. As a result, the model is able to incorporate all relevant
historical information about the occurrence, from its inception to the desired time,
into its calculations. The boundedness and existence and uniqueness of solutions to
the proposed model are also investigated. The numerical simulations are performed
by using the Lagrange polynomial-based method which is novel in the field of math-
ematical biology. Simulations have been accomplished to examine the significance
of parameters related to cannibalism, the conversion rate from prey to adult predator,
harvesting of an adult predator, and growth rate of juvenile predators on the overall
behavior of the system. The noteworthy performance of the fractional operator on
the anticipated predator-prey model’s dynamical behavior is well demonstrated by
numerical results.
Chapter “Qualitative Theory and Approximate Solution to Norovirus Model
Under Non Singular Kernel Type Derivatives” deals with a fractional-order nonlinear
dynamical system of Norovirus disease. The considered mathematical model consists
of susceptible, exposed, vaccinated, infected, and recovered classes. Firstly, the
Caputo Fabrizio fractional derivative (CFFD) is used to examine the qualitative
theory and approximate solutions. Some adequate results for the existence of approx-
imate solutions are given using fixed point theory. The Laplace Transform (LT) and
the Adomian decomposition (ADM) approach are used to get approximate results
for each compartment. The graphical presentation that corresponds to available real
facts for different fractional orders is given. Secondly, by using piecewise global
fractional derivatives in sense of singular and non-singular kernels, the aforemen-
tioned system is also investigated. Sufficient conditions for the existence and unique-
ness of the solution to the proposed model by using the fixed-point approach are
established. Numerical simulation is performed by extending Newton’s interpolation
formula for the considered model under the mentioned derivatives. Graphical presen-
tations for various compartments of the model are given against the available real
data for different fractional orders. The numerical findings are performed by using
MATLAB 16.
Chapter ”Study of the SIRI Model Utilizing the Caputo Derivative” presents
qualitative studies of the SIRI model described by a fractional operator. In the
investigations, the Caputo operator is utilized in the modeling of the epidemic.
viii Preface

Using the reproduction number the stability of the model is analyzed. A qualita-
tive property as the boundedness of the solutions of the SIRI model is focused. A
numerical scheme including the numerical scheme of the Caputo operator has been
utilized and explained rigorously. The results of the chapter have been illustrated by
the graphics of the solutions obtained using the proposed numerical scheme.
Chapter “Implementation of Vaccination in an Epidemic Model for COVID-19”
aims to establish the Susceptible, Vaccination, Infectious, and Recovered (SVIR)
pandemic paradigm. The proposed model’s general characteristics are first discussed.
Particularly, both theoretical and numerical analysis is done on the basic reproduc-
tion numbers. Reproduction numbers are used to describe both the local and global
stability of equilibrium points. Secondly, determine the stability of the equilibrium
points using the Lypanouv function and the Painocare-Bandixson condition. The
transmission of Covid-19 and its regulation are shown in simulation with the aid of
MATLAB according to genuine data in India. In the suggested model, the effects
of vaccination are investigated. This simulated result is consistent with the actual
parameters and indicates the necessity for booster doses to eradicate the virus from
the population.
The editors sincerely acknowledge the cooperation of contributors while dealing
with their works for possible publication in this book. Reviewers deserve deep grat-
itude for offering their kind help in selecting and finalizing the contents of the
book. The editors also thankfully acknowledge the support of the editorial staff
at Springer. The editors gratefully acknowledge the encouragement received from
many colleagues and friends for this book project.

Ghazipur, India Harendra Singh

Guwahati, India Hemen Dutta
Contents

Exponential-Time-Differencing Method for the Solution

of Diffusive HIV-I Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Kolade M. Owolabi, Edson Pindza, and Gulay Oguz
An Effective Technique for Solving a Model Describing Biological
Species Living Together . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Saurabh Tomar and Soniya Dhama
Numerical Solution for a Tumor Cells Dynamics Within Their
Micro-environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Kolade M. Owolabi, Albert Shikongo, and Edson Pindza
A Mathematical Model to Study Regulatory Properties
and Dynamical Behaviour of Glycolytic Pathway Using
Bifurcation Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
Shruti Tomar, Naresh M. Chadha, and Ankita Khanna
On Solutions of Biological Models Using Reproducing Kernel
Hilbert Space Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Nourhane Attia and Ali Akgül
A Study of the Fractional Tumour–Immune Unhealthy Diet Model
Using the Pseudo-operational Matrix Method . . . . . . . . . . . . . . . . . . . . . . . . 137
Saurabh Kumar and Vikas Gupta
Analysis of a Fractional Stage-Structured Model With
Crowley–Martin Type Functional Response by Lagrange
Polynomial Based Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Chandrali Baishya and P. Veeresha
Qualitative Theory and Approximate Solution to Norovirus Model
Under Non Singular Kernel Type Derivatives . . . . . . . . . . . . . . . . . . . . . . . . 181
Eiman, Waleed Ahmed, Kamal Shah, and Thabet Abdeljawad

ix
x Contents

Study of the SIRI Model Utilizing the Caputo Derivative . . . . . . . . . . . . . . 211

Ndolane Sene
Implementation of Vaccination in an Epidemic Model
for COVID-19 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
Yerra Shankar Rao
About the Editors

Dr. Harendra Singh is an assistant professor in the Department of Mathematics,

Post-Graduate College, Ghazipur-233001, Uttar Pradesh, India. He has listed in top
2% scientist list published by Stanford University. His area of interest is mathemat-
ical modelling, fractional differential equations, integral equations, calculus of vari-
ations, analytical and numerical methods. His works have been published in applied
mathematics and computations, applied mathematical modelling, applied numerical
mathematics, Chaos Solitons & Fractals, numerical methods for partial differential
equations, physica A, astrophysics and space science, Electronic Journal of Differen-
tial Equations, Few Body-system and several other peer-reviewed international jour-
nals. His 50 research papers have been published in various journals of repute with
h-index of 22 and i10 index 31. He has delivered lectures in a number of national and
international conferences and workshop. He has also attended short-terms programs
and workshops.

Dr. Hemen Dutta is serving Gauhati University, India, regularly as a faculty member
since 2010. Prior to joining Gauhati University, he served three other academic
institutions in different capacities. He is currently interested in applied analysis and
mathematical modelling. He has over 170 publications in the forms of research
papers, book chapters and proceedings papers and serving as an editor of several
SCI/SCIE and Scopus indexed journals as well as book series editor of top-class
publishers like Elsevier and Taylor & Francis. He has to his credit several authored
and edited books and conference proceedings published by reputed publishers like
Springer, Birkhauser, Wiley, Taylor & Francis, Elsevier, etc. He is a selected member
as well as life member of some academies and societies. He is the recipient of several
grants, research projects and holding some honorary positions at foreign and national
university/institution.

xi
Exponential-Time-Differencing Method
for the Solution of Diffusive HIV-I Model

Kolade M. Owolabi, Edson Pindza, and Gulay Oguz

Abstract In recent years, the modeling and application of some epidemiological

diseases have been the subject of activities. Most of these models mainly exist in the
form of ordinary differential equations (ODE), due to some the challenges encoun-
tered in solving systems with coupled partial differential equations (PDEs). This
work is extended to PDE, by formulating a robust numerical technique based on the
formulation of an exponential-time-differencing (ETD) scheme which could easily
circumvent the issue of stiffness inherent in coupled PDE system. Spectral method
is suitably applied to time-dependent PDE problems. At first, the problem is dis-
cretized with respect to space in such that, the arising system of ODEs is solved
using the novel fourth-order ETD method of Runge-Kutta type. A typical example
is given to the HIV model in the diffusive case which could rarely be found in the
literature. Numerical results which reflect the physical behaviour of the model under
consideration are given in one and two dimensions.

K. M. Owolabi (B)
Department of Mathematical Sciences, Federal University of Technology Akure,
PMB 704 Akure, Ondo State, Nigeria
e-mail: [email protected]
E. Pindza
Department of Mathematics and Applied Mathematics, University of Pretoria,
Pretoria 002, South Africa
Department of Mathematics and Statistics, Tshwane University of Technology, Pretoria West,
Pretoria 0183, South Africa
G. Oguz
Department of Mathematics, Harran Unıversity, Sanlıurfa, Turkey

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 1
H. Singh and H. Dutta (eds.), Computational Methods for Biological Models,
Studies in Computational Intelligence 1109,
https://doi.org/10.1007/978-981-99-5001-0_1
2 K. M. Owolabi et al.

1 Introduction

This chapter considers higher order numerical method based on the novel exponential-
time-differencing (ETD) method for the solving of an evolution equation of the
parabolic type
ut = DLu + N (t, u) (1.1)

where L stands for the linear operator representing diffusive term, and N represents
a nonlinear operator which accounts for the totality of local reactions. D is diagonal
matrix representing the diffusion constants of each of sub-population class or species.
The linear terms of most time-dependent PDEs are mainly responsible for the issue
of stiffness, with exponential decay property as in the case of dissipative PDE, or
spatiotemporal oscillation as peculiar to dispersive cases. Therefore, it is desirable
to use higher-order approximation techniques in both time and space for any time-
dependent parabolic PDE that adds a nonlinear low-order with a linear higher-order
terms. When one discretize (1.1) in space, it leads to a large (or coupled) system of
ODEs in the form
u t = DLu + N(t, u). (1.2)

Such diffusive systems often arise when wave equations such as the Navier-Stokes,
Schrödinger, Kuramoto, Kortweg-de Vries (KdV), and nonlinear Burgers’ equations
are discretized [1, 7]. The ETD methods are time integration solvers which when
combined with the special methods can be used efficiently to obtain good numerical
solutions for either stiff or moderately stiff PDE problems.
The authors in [5, 7, 13, 15] have successfully applied the ETD methods as effi-
cient time-solvers in conjunction with the spectral schemes to solve a range of linear
and nonlinear time-dependent PDEs. Also, the stability and convergence properties
for this class of schemes have been considered. Application of the ETD method for
solving the diffusive HIV model in one-dimensional space will be considered in
this work. A lot of time integrator algorithms are formulated to solve stiff problems
have been proposed. For instance, some of these formulations are either based on the
Runge-Kutta or Adams-Bashforth-Moulton methods. This class of numerical meth-
ods is most suitable to handle semi-linear equations which can be divided into a linear
part, which harbors the stiffest part of the dynamical problem, and a non-linear-part,
which often varies more slowly than the linear-part. Recall that many time-dependent
reaction diffusion PDEs such as (1.1) mainly combined low-order (nonlinear terms)
with higher-order (linear terms). Hence, the nature of such equations allows one to
apply any method of choice in time and space directions [7, 13].
Exponential time-differencing schemes have been re-examined many times in
different forms and under several authors. A case study is the exact linear part method
which was suggested in [3] for random order. In [4], the explicit formulation of the
linear part schemes of irregular order was reported, which was named the ETD
schemes. According to Cox and Matthews, the idea of ETD methods was earlier
introduced in the computational area of electrodynamics. In the year 2002, the duo
Exponential-Time-Differencing Method for the Solution of Diffusive HIV-I Model 3

in [4] further designed a higher-order scheme, and they come up with the new Runge-
Kutta (RK) types whose orders are ranging from first-order to fourth-order. Ever since
the inception of these schemes, many authors have developed some interests in the
application of the ETD time-solver schemes as a viable and effective approach to
address the stiffness challenge posed in the computation of time-dependent PDE
problems. This chapter presents and implements the novel Fourier spectral method
in space with the fourth-order ETD Runge-Kutta (ETDRK4) method for the solution
of HIV models. Hence, the remaining part of this work is arranged into different
sections in the following manner. The Fourier spectral approximation technique and
the novel exponential time differencing methods are introduced in Sect. 2. As a typical
example, the multicomponent HIV model is introduced in Sect. 3, its stability analysis
is also presented. Applicability and suitability of the numerical methods suggested is
demonstrated on HIV model with detail numerical experiment in Sect. 4. Conclusion
is given in the last part.

2 Numerical Techniques for Reaction-Diffusion Problems

Spectral methods can be categorized into different types, and one clear difference
often made is among the tau, Galerkin, and collocation (also known as the pseu-
dospectral) spectral algorithms [28, 29]. The Galerkin and tau methods perform well
using coefficients of the global expansion, while the collocation or pseudospectral
method work well with its values at points. As a result of the general notion of
the ansatz functions, spectral algorithms are commonly referred to as the general
methods, which means the value of a derivative at a point with regards to space often
depends on the solution at every other points in space, and off grid points. As a matter
of fact, spectral discretization schemes have a high-order level of accuracy, talking in
terms of spectral- convergence over any finite difference schemes. More importantly,
diffusion and dispersion properties of the spatial Laplacian operator are of impor-
tance when one compares with the known finite difference (FD) schemes. It is not
difficult to see that when the concept of a modified wave number is considered, spec-
tral methods often results to the exact derivative of the function, the major concern
should be the truncation error when a finite set of ansatz functions/coefficients is
prematurely truncated. Though spectral algorithms are flexible geometrically and
more cumbersome to compute than the FD methods, but spectral methods was
regarded among the first sets of methods to be used in practical simulation processes
[7, 21, 30].
Spectral method is most applicable when one desires high level of accuracy to
compute the solution of the time-dependent PDEs. This class of methods is suitable
for developing numerical schemes in most application areas of engineering, physics
and mathematics. Spectral method has capacity to accurately differentiate the lin-
ear operator. Also, this method can be used to compute the numerical solution of
ODEs and PDEs, based on the polynomial expansion of the solutions. The numer-
ical integration is primarily based on the variational of the continuous problem in
4 K. M. Owolabi et al.

its formulation which is computed as quadrature scheme. For other numerical tech-
niques which have been used to study models in mathematical biology, readers are
referred to [6, 25, 26, 31].

2.1 Spectral Method for Spatial Discretization

Define the Laplacian or linear operator L as a complete orthonormal set of eigenfunc-

tions {ψ j } which satisfies the extreme conditions on D ⊂ Rd , d ≤ 3, and eigenvalues
λ j , that is, (L)ψ j = λ j ψ j on D, it is reasonable to assume that
⎧ ⎫
⎨ ∞
 ∞
 ⎬
u := u= ū j ψ j , ū j = u, ψ j , |ū j |2 |λ|2 < ∞, , (2.3)
⎩ ⎭
j=0 j=0

for some u ∈ u, it shows that

∞

Lu = ū j λ2j ψ j . (2.4)
j=0

For notational convenience, one writes problem (1.1) in the form, taking u :=
(u, v, w)T , which implies

ut = DLu + N (u), in ζ L , (2.5)

∂u
u(x, 0) = u0 (x), x ∈ ζ, = 0 on ∂ζ × (0, L), (2.6)
∂ν
with ⎛ ⎞
f 1 (u, v, w)
N (u) ≡ ⎝ f 2 (u, v, w)⎠ (2.7)
f 3 (u, v, w)

and ⎛ ⎞
η1 0 0
D = ⎝ 0 η2 0 ⎠ . (2.8)
0 0 η3

With N = 0 in one dimension, (2.5) becomes a simple diffusion equation

u t = DLu, (2.9)

subject to the initial condition u(x, 0) = u 0 (x) and any boundary condition of choice
with x ∈ [0, L]. With (2.3) and Eq. (2.4), the exact solution of (2.9) is given as
Exponential-Time-Differencing Method for the Solution of Diffusive HIV-I Model 5

∞
 ∞

ū j (0)e−Dλ j t ψ j (x),
2
u(x, t) = ū j (t)ψ j (x) = (2.10)
j=0 j=0

with ū j (0) = u 0 (x), ψ j (x), for the case of homogeneous Dirichlet condition, we
  
( j + 1)π 2 2 ( j + 1)π(x − 0)
have λ j = and ψ = sin .
L L L
To demonstrate this approach using multi-component reaction-diffusion system
with the reaction kinetics, one uses the fast Fourier transform for the reaction-
diffusion problem to have
 
Ut (ζx , ζ y , t) = η1 ζx2 + ζ y2 U (ζx , ζ y , t) + N [ f 1 (u(x, y, t), v(x, y, t), w(x, y, t))],
 
α/2 α/2
Vt (ζx , ζ y , t) = η2 ζx + ζ y V (ζx , ζ y , t) + N [ f 2 (u(x, y, t), v(x, y, t), w(x, y, t))],
 
α/2 α/2
Wt (ζx , ζ y , t) = η2 ζx + ζ y W (ζx , ζ y , t) + N [ f 3 (u(x, y, t), v(x, y, t), w(x, y, t))]

where U , V and W are the double Fourier transforms which stand for the population
densities u(x, y, t), v(x, y, t) and w(x, y, t), respectively. In other words,
 ∞  ∞
N [u(x, y, t)] = U (ζx , ζ y , t) = u(x, y, t)e−i(ζx x+ζ y y) d xd y,
 0
∞  0
∞
N [v(x, y, t)] = V (ζx , ζ y , t) = v(x, y, t)e−i(ζx x+ζ y y) d xd y.
0
∞ 
0
∞
N [w(x, y, t)] = W (ζx , ζ y , t) = w(x, y, t)e−i(ζx x+ζ y y) d xd y.
0 0
 
With ζ 2 = ζx2 + ζ y2 , the issue of stiffness is explicitly removed in the linear piece
via the integrating factors, next we set

U = eη1 ζ t Ū , V = eη2 ζ t V̄ , W = eη3 ζ t W̄

2 2 2

in such that

∂t Ū = eη1 ζ t N [ f 1 (u, v, w)],

2

∂t V̄ = eη2 ζ t N [ f 2 (u, v, w)],

2
(2.11)
η3 ζ 2 t
∂t W̄ = e N [ f 3 (u, v, w)],

Again, one requires to discretize the equal domain size (square) by via the equi-
spaced points L x , L y along the directions x, y, respectively. We adopt the Matlab
discrete fast Fourier transform to change Eq. (2.11) to ODEs
6 K. M. Owolabi et al.

∂t Ūi, j = e K u ζi, j t N [ f 1 (u i, j , vi, j , wi, j )],

2

∂t V̄i, j = e K v ζi, j t N [ f 2 (u i, j , vi, j , wi, j )],

2
(2.12)
K W ζi,2 j t
∂t W̄i, j = e N [ f 3 (u i, j , vi, j , wi, j )],

α/2 α/2
with u i, j = u(xi , y j ), vi, j = v(xi , y j ), wi, j = w(xi , y j ) and ζi, j = χx (i)+

α/2
χ y ( j) . At this point we utilize the zero-flux boundary conditions clamped at
the two ends of the domain. Any explicit solver can be used to integrate in time at
this point. In order to gain a better accuracy, we employ the novel ETDRK4 method
as described below.

2.2 The ETDRK4 Method

To derive the time integration solver, the notations used in [4] is followed. Consider
the ordinary differential equation

dυ(t)
= Lυ(t) + N(υ(t), t), (2.13)
dt

where L denotes the stiffness parameter harbouring the diffusion, and N(υ(t), t)
represents nonlinear coupled source or forcing term that accounts for the totality of
local reactions. The formulation of the k-step ETD methods is followed directly in
[2, 4]. To start with, one multiplies (2.13) with the integrating factor e−Lt , we then
integrate it over a single time-step from t = tn and t = tn+1 = tn + ηt to have
 ηt
υ(tn+1 ) = υ(tn )eLηt + eLηt e−Lζ N(υ(tn + ζ), tn + τ )dτ . (2.14)
0

Different ETD schemes arise from the approximation of integral function in (2.14).
This scheme is called the exact method [4], and the next step is to derive approxi-
mations to the integral.
By using the Newton backward-difference scheme, and source term N(υ(t), t) at
the nth, and preceding time-steps, we approximate N(υ(tn + ζ), tn + ζ) as


s−1 
j −τ /ηt
N(υ(tn + τ ), tn + τ ) ≈ An (tn + τ ) = (−1) ∇ j An (tn ), (2.15)
j=0
j

where ∇ is the backward difference scheme expressed as

Exponential-Time-Differencing Method for the Solution of Diffusive HIV-I Model 7

j 
j
∇ j An (tn ) = (−1)k An−k (tn−k ),
k=0
k
j 
j
≈ (−1)k N(u(tn−k ), tn−k ), (2.16)
k=0
k

and 
−
j! = (− − 1) · · · ( − j + 1), for j = 1, . . . , s − 1,
j

where
 = ζ/ηt.
−
If one substitutes (2.15) (2.14), bear in mind that (0! 0
= 1), then


s−1  
1
−
υ(tn+1 ) − υ(tn )eLηt ≈ ηt (−1) j eLηt (1−) d∇ j An (tn ). (2.17)
j=0 0 j

At this point, the integral in (2.17) can be indicated as

 1
α j = (−1) j eLηt (1−) d, (2.18)
0

by obtaining in the generating function, one computes α j . For ω ∈ R, |ω| < 1, there-
fore, this can be executed by introducing the generating function as
∞

(ω) = αjω j, (2.19)
j=0

which is readily found to be

 ∞ 
1
−
(ω) = eLηt (1−) (−ω) j d,
0 j=0
j
 1
= eLηt (1−) (1 − ω)− d,
0
eLηt (1 − ω − e−Lηt )
= . (2.20)
(1 − ω)(Lηt + log(1 − ω))

On rearranging (2.20) as
8 K. M. Owolabi et al.

1
(Lηt + log(1 − ω))(ω) = eLηt − ,
1−ω

and then obtain the expansion in powers of ω as


ω2 ω3
Lηt − ω − − − ··· (α0 + α1 ω + α2 ω 2 + · · · ) = eL − 1 − ω − ω 2 − ω 2 − · · · ,
2 3

for j ≥ 0, we have the recurrence relation for α j in the manner that

Lηtα0 = eLηt − 1, (2.21)

1 1 1 
j
1
Lηtα j+1 + 1 = α j + α j−1 + α j−2 + · · · + α0 = αk .
2 3 j +1 k=0
j +1−k

On putting Eqs. (2.16), and (2.18) into (2.17), the general ETD method [4] of order
s is presented as


s−1 
j 
j
υn+1 = υn eLηt + ηt αj (−1)k Nn−k , (2.22)
j=0 k=0
k

where u n and N are numerical approximation to u(tn ) and N(u(tn ), tn ), respectively.

Note that the formulation and description of the ETD schemes are via the linear
multi-step method (LMM) idea. Therefore, their usage require kth previous evalua-
tions of the term N(υ(t), t) which make them difficult and more challenging to use
due to the requirement of starting values. This issue was the reason the authors in
[4] seeking for means to overcome the stiffness challenge, and finally opined that
the problem can only be avoided by applying the Runge-Kutta methods which are
self starter. More so, the importance of large area of stability (LAS) and small error
constants than multi-step methods. It is of great interest to develop the Runge-Kutta
based ETD schemes.
Direct application of the classical Runge-Kutta method of order four and intro-
duction of some necessary parameters results to the novel Cox and Matthews [4]
method tagged the ETDRK4:

υn+1 = υn eLηt + Nn [−4 − Lηt + eLηt (4 − 3Lηt + L2 ηt 2 )]

+2(N (an , tn + ηt/2) + N (bn , tn + ηt/2))[2 + Lηt + eLηt (−2 + Lηt)]
+N (cn , tn + ηt)[−4 − 3Lηt − L2 ηt 2 + eLηt (4 − Lηt)]/L3 ηt 2 , (2.23)

where

an = υn eLηt/2 + (eLηt/2 − I)Nn /L,

bn = υn eLηt/2 + (eLηt/2 − I)N (an , tn + ηt/2)/L,
cn = υn eLηt/2 + (eLηt/2 − I)(2N (bn , tn + ηt/2) − Nn )/L.
Exponential-Time-Differencing Method for the Solution of Diffusive HIV-I Model 9

The terms an and bn are used for approximation of u at tn + ηt/2, while the term cn
approximates u at tn + η [13, 22, 23, 27]. The method (2.23) serves as the quadrature
formula for (2.14) which one derives via the quadratic interpolation points tn , tn +
ηt/2 and tn + ηt, when an average value of N is applied at an and bn .

3 The HIV-I Model and Its Stability Analysis

A lot of efforts has been made in the mathematical modeling of HIV-I infection
model. Mathematical models of both fractional order and classical order have also
been derived to discuss the HIV-I infection dynamics in the bloodstream where the
cell-free-viral spread is the major route of viral infections [8–12]. The model which
describe a cell-to-cell dynamic of HIV-I infection in the tissue is given by
 t 
du u + v + −∞ g(t−ξ)u(ξ)v(ξ)
ϑ+u(ξ)
dξ uv
= φu 1 − −ϕ ,
dt σ ϑ+u
 t (3.1)
dv g(t − ξ)u(ξ)v(ξ)
=ψ dξ − ηv,
dt −∞ ϑ + u(ξ)

where u and v denote the population densities at time t of healthy and infected cells,
respectively, φ is the reproduction rate of class u, σ represents the carrying capacity,
and the maximum rate of infection is given by ϕ, the death rate infected population
η, the fraction of cells that survives the incubation period is denoted as ψ, while ρ is
half saturation constant. By following the discussion in [8], we take the delay kernel
g(ξ) as
g(ξ) = ρe−ρξ , ξ ≥ 0,

we let  t
u(ξ)v(ξ)
w(t) = ρe−ρ(t−ξ) dξ,
−∞ ϑ + u(ξ)

to finally arrive at equations


du u+v+w uv
= φu 1 − −ϕ ,
dt σ ϑ+u
dv
= ψw − ηv, (3.2)
dt
dw uv
=ρ − ρw.
dt ϑ+u

In ecological context, Segel and Jackson [24] called the attention of many
researchers to the fact that diffusion-driven or Turing instability can occur in an
ecological context. They conducted their research based on a predator-prey dynam-
10 K. M. Owolabi et al.

ics where the inclusion of diffusion term (random dispersal) leads to an instability
of the steady-state distribution to perturbation of a specific wavelength. In 2016,
the same idea was extended to model the spread of Hepatitis B virus (HBV) in the
presence of diffusion [15]. Further reports on diffusive instability in both ecological
and epidemiological contexts can be found in [16–20].
Now assume that the cells u, v, w are population densities, that is population per
unit volume, and depends on a spatial variable x and time t, that is u = u(x, t), v =
v(x, t) and w = w(x, t). Then a conservation law can be formulated in the form

u+v+w uv
u t + x = f 1 (u, v, w) = φu 1 − −ϕ ,
σ ϑ+u
vt + x = f 2 (u, v, w) = ψw − ηv, (3.3)
uv
wt + x = f 3 (u, v, w) = ρ − ρw,
ϑ+u

where  denotes the population flux. Assuming Fick’s law for the flux , we obtain
the following reaction-diffusion HIV-I system

u+v+w uv
u t − Du u x x = f 1 (u, v, w) = φu 1 − −ϕ ,
σ ϑ+u
vt − Dv vx x = f 2 (u, v, w) = ψw − ηv, (3.4)
uv
wt − D w w x x = f 3 (u, v, w) = ρ − ρw,
ϑ+u

where Du , Dv and Dw are diffusion coefficients. System of the form (3.4) become
one of the most essential classes of nonlinear phenomena due to their occurrence in
many chemical (combustion) and biological (pattern formation) processes.
In order to make the correct choice of parameters which will be used to simu-
late the full reaction-diffusion dynamics, it is necessary to examine the stability of
HIV-I model. To achieve this, we set f i (u, v, w) = 0, i = 1, 2, 3 in the absence of
diffusion (that is, Du = Dv = Dw = 0) as

u+v+w uv
0 = φu 1 − −ϕ ,
σ ϑ+u
0 = ψw − ηv, (3.5)
uv
0=ρ − ρw.
ϑ+u

It is obvious that the above algebraic system has three equilibrium states namely; the
washout state corresponding to E 0 = (0,  the healthy cell state E 1 = (σ, 0, 0)
 0, 0),
1 ψσ
with basic reproduction number R0 = η ϑ+σ > 1. The infected equilibrium is cal-
culated as E ∗ = (u ∗ , v ∗ , z ∗ ), where
Exponential-Time-Differencing Method for the Solution of Diffusive HIV-I Model 11

ηϑ
x∗ = ,
ψ−η
ψϑ(φηϑ − φσψ − φση)
y∗ = ,
(ψ − η)(ψφϑ + φϑη + ϕϑψ − ϕση)
ηϑ(φηϑ − φσψ − φση)
w∗ = .
(ψ − η)(ψφϑ + φϑη + ϕϑψ − ϕση)

The community matrix corresponding to the autonomous system is given by

⎛ φ(2u+v+w) ϕϑv ⎞
φ− σ
− (ϑ+u)2
− φu
σ
− ϕu
ϑ+u
− φu
σ
⎜ ⎟
⎜ ⎟
A(u,v,w) = ⎜
⎜ 0 −η ψ ⎟⎟ (3.6)
⎝ ⎠
ρϑv ρu
(ϑ+u)2 ϑ+u
−ρ

Theorem 3.1 If R0 < 1, then healthy class E 1 = (σ, 0, 0) is locally asymptotically

stable.

Proof At healthy equilibrium, The characteristic polynomial at point E 1 = (σ, 0, 0)

is given as
 ϕσ 
 τ + ψ φ + ϑ+σ r 
    

Ψ (τ ) = τ I − A E1  =  0 τ + η −ψ 
 
 0 σ
−ρ ϑ+σ λ + ρ
= τ 3 + (τ + ρ + η) τ 2 + (φρ + φη + ρη (1 − R0 )) τ + φηρ (1 − R0 ) .

For R0 < 1, it implies that all the coefficients of Ψ (τ ) are positive, and

(τ + ρ + η) (φρ + φη + ρη (1 − R0 )) > φηρ (1 − R0 )

is satisfied. By using the Routh-Hurwitz criteria, since all the roots of Ψ (τ ) have
negative real parts. This implies that the general condition | arg(τ )| > π2 holds, and
as a result the equilibrium point E 1 = (σ, 0, 0) is locally asymptotically stable. 

Next, the community matrix A at nontrivial point E ∗ = (u ∗ , v ∗ , w ∗ ) is calculated

as ⎛ ∗ ⎞
− p −q −φ uσ
AE ∗ = ⎝ 0 ∗ −η ψ ⎠
ρϑv
(ϑ+u ∗ )2
ρ ψη −ρ

∗ ∗
where p = −φ + σφ v ∗ + 2 σφ u ∗ + σφ w ∗ + (ϑ+u
ϕϑv η
∗ )2 and q = φ σ + ϕ ψ .
u

In attempt to gain insightful understanding, the following assumption was made

on the parameters of the dynamical system
12 K. M. Owolabi et al.

u∗ v∗ φ αϑv ∗
− φ − 2φ − φ − w∗ − < 0, (3.7)
σ σ σ (ϑ + u ∗ )2

we denote the discriminant by β and the trace by α as

∗ ∗ ∗
ϑv ϑψv
α = 2 pη + p 2 + pφ uσ (ϑ+u − q (ϑ+u
   ∗ )2∗ ∗
∗ )2

ϑv
β = α − 4 p η + pη
2 2 2
φ σ (ϑ+u ∗ )2 + p
u

Theorem 3.2 Assume that R0 > 1 and inequality in (3.7) is satisfied. (1) If either of
β < 0 or β = 0 and α > 0, then E ∗ is locally asymptotically stable for all ρ > 0; (2)
If β = 0 and α > 0, then there exists ρ0 > 0, such that E ∗ is locally asymptotically
stable for ρ > 0 and ρ = ρ0 ; (3) If β > 0, then the following cases arise: (i) α > 0,
then E ∗ is locally asymptotically stable for all ρ > 0; (ii) α < 0, then there exist
0 < ρ1 < ρ2 , such that E ∗ is locally asymptotically stable for ρ < ρ1 or ρ > ρ2 .

Proof At interior equilibrium point E ∗ = (u ∗ , v ∗ , w ∗ ) , the characteristic polyno-

mial takes the form
 ∗ 
 τ+p q φ uσ 
  
Ψ (τ ) = τ I − A(E ∗ )  =  0 τ + η −ψ 
 ρϑv ∗ 
− ∗ 2
−ρ βη τ + ρ 
(ϑ+u ) 
∗ ρϑv ∗ ∗ ρϑv ∗ ∗
= τ 3 + ( p + ρ + η) τ 2 + pρ + pη + φuσ τ + η φuσ + q ρψϑv∗ 2 .
(ϑ+u ∗ )2 (ϑ+u ∗ )2 (ϑ+u )
(3.8)

Since ( 3.7) is satisfied, then for p > 0 and all the coefficients of Ψ (τ ) are positive. By
adopting the Routh-Hurwitz Theorem to (3.8), bear in mind that E ∗ = (u ∗ , v ∗ , w ∗ )
is locally asymptotically stable if

φu ∗ ρϑv ∗ ρψϑv ∗ φu ∗ ρϑv ∗
Λ (ρ) = ( p + ρ + η) pη + pρ + −q −η > 0. (3.9)
σ (ϑ + u ∗ )2 (ϑ + u ∗ )2 σ (ϑ + u ∗ )2

We rewrite (3.9) into the following form


u ∗ ϑv ∗
Λ (ρ) = r + p ρ2 + αρ + pη 2 + p 2 η.
σ (ϑ + u ∗ )2

Finally, we arrive at the following conclusion:

If β < 0 or β = 0 and α > 0, then Λ (ρ) have no real-root or one negative real-
root. In such a situation Λ (ρ) > 0 and E ∗ is locally asymptotically stable for all
ρ > 0. If β = 0 and α > 0, then Λ (ρ) have one root ρ0 > 0. Thus, Λ (ρ) > 0 for
all ρ > 0 and ρ = ρ0 . Then, E ∗ is locally asymptotically stable for all ρ > 0 and
ρ = ρ0 . Now, let us assume that β > 0. Then, if α > 0, Λ (ρ) have two negative real
roots, so Λ (ρ) > 0. the interior point E ∗ which correspond to the existence of all
cells is locally asymptotically stable. In the case α < 0, Λ (ρ) have two positive real
Exponential-Time-Differencing Method for the Solution of Diffusive HIV-I Model 13

Fig. 1 Chaotic attractors and time evolution results of different cells

roots 0 < ρ1 < ρ2 such that A (ρ) > 0 for ρ ∈ (0, ρ1 ) ∪ (ρ2 , +∞). Thus we have
that E ∗ is locally asymptotically stable for ρ ∈ (0, ρ1 ) ∪ (ρ2 , +∞). 

To illustrate the behavior of the dynamical system with respect to stability, the
following parameters are used for the numerical simulations

φ = 0.65/day, σ = 2 × 106 /ml, ϕ = 11/day, η = 0.3/day, ϑ = 3.7 × 106 /ml,

ρ = (0, 1], ψ = 0.847, u 0 = 5 × 105 /ml, v0 = 5 × 102 /ml, w0 = 2 × 102 /ml.
(3.10)
14 K. M. Owolabi et al.

Fig. 2 Chaotic attractors and time evolution results of different cells

With above parameters, it is obvious that R0 = 0.9906 < 1 which satisfies the con-
dition on Theorem 3. The simulation results as shown in Figs. 1 and 2 show that the
healthy cells predominate over the infected class. Clearly the point E 1 is asymptoti-
cally stable. The coexistence of both healthy and infected cells are also demonstrated.
Each of the cells also enters into permanence, which in reality is true. No matter the
level of sensitization which individual and government will put in place to wipe out
infected or latent cells completely.
Exponential-Time-Differencing Method for the Solution of Diffusive HIV-I Model 15

4 Numerical Results and Discussions

In this segment, we employ the numerical techniques in Sect. 2 to solve the diffusive
HIV-I infection model as discussed in Sect. 3, to test the accuracy and efficiency
of the proposed method. The experimental results are given in one and two spa-
tial dimensions. All computations are carried out using the Matlab 2019a software
package.

4.1 One Dimensional Case

Consider the time-dependent reaction-diffusion HIV-I model (3.4), where

u(x, t), v(x, t) and w(x, t) represent the species densities at time t and position x.
The diffusion coefficients Du , Dv and Dw describe the intensity of spatial interaction
due to self-motion of the cells within a given habitat or domain. In the computation,
we experiment using Du = 0.07, Dv = 0.05, Dv = 0.02 and the homogenous (zero-
flux) boundary conditions subject to non-negative random initial condition taken in
the form

Fig. 3 One dimensional distribution of species with ϕ = 4 for t = 10

16 K. M. Owolabi et al.

Fig. 4 One dimensional distribution of species with ϕ = 4 for t = 10

u 0 = 0.07(ones(N , 1)), v0 = 0.01(ones(N , 1)), w0 = 0.02(ones(N , 1)),

(4.1)
where (u ∗ , v ∗ , w ∗ ) = (0.07, 0.01, 0.02) are the small perturbations. The effect of ϕ
which symbolizes the maximum rate of infection is shown in Figs. 3 and 4 which
correspond to ϕ = 4 and ϕ = 9, respectively at t = 10. It is obvious that all cells
coexist. With increasing time (t), and varying value of ρ as indicated in the caption,
we obtain different spatiotemporal distribution of dynamic system as displayed in
Fig. 5.

4.2 Two Dimensional Case

It is in higher dimensions that the numerical idea presented in Sect. 2 really become
of serious value. We choose to illustrate the numerical technique using the result of
nontrivial example which guarantee the existence of all cells as stated in Theorem 3.
Here, we assume that about 30% of the infected class survive incubation which when
calculated results to the value ϕ = 3.31, so that p = 0.0816, α = −0.0938 < 0,
R0 = 3.8596, β = 0.00515 > 0. The second condition in Theorem 3 holds. Then
the point E ∗ is locally asymptotically stable. Both infected and healthy cells do
Exponential-Time-Differencing Method for the Solution of Diffusive HIV-I Model 17

Fig. 5 one dimensional evolution of (3.4) for t = 50. Columns 1–3 correspond to ρ = 0.6, 0.8, 1.0,
respectively

coexist. For the simulation experiments in all 2D cases, the time step t = 0.05, the
space step h is kept to 1/4, that is x =  y = 0.25, and the homogeneous (zero-flux)
Neumann boundary conditions on equal domain size (L). The 2D experiments are
repeated with different initial conditions and some perturbed parameters.
In Figs. 6 and 7, we experiment with the initial condition set as

u 0 = u ∗ − 0.5(sin(π(x − μ)/(2μ)))m (sin(π(y − μ)/(2μ)))m ,

v0 = v ∗ (sin(π(x − μ)/(2μ)))m (sin(π(y − μ)/(2μ)))m , (4.2)
w0 = w ∗ − 0.25(sin(π(x − ν)/(2μ)))m (sin(π(y − μ)/(2μ)))m

to mimic spatiotemporal evolution of the dynamic model with contour and surface
plots, where (u ∗ , v ∗ , w ∗ ) = (1, 0.25, 3), m = 0.8 and μ = 0.5.
18 K. M. Owolabi et al.

Fig. 6 Two dimensional results for system (3.4) with initial condition (4.2) and ϕ = 5

Fig. 7 Two dimensional results for system (3.4) with initial condition (4.2) and ϕ = 3.5
Exponential-Time-Differencing Method for the Solution of Diffusive HIV-I Model 19

Fig. 8 Two dimensional chaotic evolution of HIV-I system with random initial data (4.3) and
ρ = 0.3. Simulation runs for t = 200

During the experiment, it was observed that variation of the initial conditions could
also induce some strange and chaotic behavior of the system under consideration. To
achieve this, we allow the initial condition to evolve naturally by using the computer
random points computed as

u 0 = randn(N , N )0.7, v0 = randn(N , N )0.5, w0 = randn(N , N )0.2, (4.3)

20 K. M. Owolabi et al.

Fig. 9 Two dimensional chaotic evolution of HIV-I system with random initial data (4.3) and
ρ = 0.6. Simulation runs for t = 200

with N = 200. This experiment is repeated for ρ = 0.3, 0.6, 1.0, to obtain the chaotic
(spiral) results in Figs. 8, 9 and 10, respectively. With the experimental results in 2D,
one could conclude that it is possible to obtain chaotic patterns in epidemiological
models if such dynamics are extended to diffusive cases.
Exponential-Time-Differencing Method for the Solution of Diffusive HIV-I Model 21

Fig. 10 Two dimensional chaotic evolution of HIV-I system with random initial data (4.3) and
ρ = 1.0. Simulation runs for t = 200

5 Conclusion

This work is mainly concerned with the solution of the time-dependent HIV-I infec-
tion model in a diffusive form. To overcome the stiffness issue inherent with the
diffusion term in the model, a Fourier spectral method is introduced for the spatial
discretization, and the resulting system of ODEs is advanced in time with the modified
ETDRK4 scheme. To explore the dynamic richness of the HIV-I model, numerical
experiments are conducted in one and two dimensions to address the points and
22 K. M. Owolabi et al.

queries that may naturally occur. It was also observed that the model under certain
parameter values and the choice of initial conditions can give rise to the evolution of
chaotic pattern formation which has applications in applied physics and engineering.
Extension of the numerical techniques to non-integer order models in sciences and
technology is left for future research.

References

1. S.J. Achar, C. Baishya, P. Veeresha, L. Akinyemi, Dynamics of fractional model of biological

pest control in tea plants under the influence of Beddington-DeAngelis functional response.
Fractal and Fract. 6(1) (2022). https://doi.org/10.3390/fractalfract6010001
2. G. Beylkin, J.M. Keiser, L. Vozovoi, A new class of time discretization schemes for the solution
of nonlinear PDEs. J. Comput. Phys. 147, 362–387 (1998)
3. M. Calvo, C. Palencia, A class of explicit multi-step exponential integrators for semi-linear
problems. Numer. Math. 102, 367–381 (2006)
4. S.M. Cox, P.C. Matthews, Exponential time differencing for stiff systems. J. Comput. Phys.
176, 430–455 (2002)
5. Q. Du, W. Zhu, Stability analysis and applications of the exponential time differencing schemes.
J. Comput. Appl. Math. 22, 200–209 (2004)
6. E. Ilhan, P. Veeresha, Haci M. Baskonus, Fractional approach for a mathematical model of
atmospheric dynamics of CO2 gas with an efficient method. Chaos, Solitons and Fractals 152,
111347 (2021) (2021)
7. A.K. Kassam, L.N. Trefethen, Fourth-order time-stepping for stiff PDEs. SIAM J. Sci. Computi.
26, 1214–1233 (2005)
8. C. Kou, Y. Yan, J. Liu, Stability analysis for fractional differential equations and their applica-
tions in the models of HIV-I infection. CMES 39, 301–317 (2009)
9. P. De Leenheer, H.L. Smith, Virus dynamics: a global analysis. J. SIAM Appl. Math. 63,
1313–1327 (2003)
10. P.A. Naik, K.M. Owolabi, M. Yavuz, J. Zu, Chaotic dynamics of fractional order HIV-1 model
involving AIDS-related cancer cells. Chaos, Solitons Fractals 140, 110272 (2020)
11. P.A. Naik, J. Zu, K.M. Owolabi, Global dynamics of a fractional order model for the transmis-
sion of HIV epidemic with optimal control. Chaos, Solitons Fractals 138, 109826 (2020)
12. P.A. Naik, J. Zu, K.M. Owolabi, Modelling the mechanics of viral kinetics under immune
control during primary infection of HIV-I with treatment in fractional order. Phys. A 545,
123816 (2020)
13. K.M. Owolabi, K.C. Patidar, Higher-order time-stepping methods for time-dependent reaction-
diffusion equations arising in biology. Appl. Math. Comput. 240, 30–50 (2014)
14. K.M. Owolabi, K.C. Patidar, Solution of pattern waves for diffusive Fisher-like non-linear
equations with adaptive methods. Int. J. Nonlinear Sci. Numer. Simul. 17, 291–304 (2016)
15. K.M. Owolabi, Numerical solution of diffusive HBV model in a fractional medium. Springer
Plus 5, 1643 (2016). https://doi.org/10.1186/s40064-016-3295-x
16. K.M. Owolabi, K.C. Patidar, Numerical simulations of multicomponent ecological models
with adaptive methods. Theor. Biol. Med. Model. 13, 1 (2016)
17. K.M. Owolabi, Mathematical study of multispecies dynamics modeling predator-prey spatial
interactions. J. Numer. Math. 25, 1–16 (2017)
18. K.M. Owolabi, A. Atangana, Spatiotemporal dynamics of fractional predator-prey system with
stage structure for the predator. Int. J. Appl. Comput. Math. 3, 903–924 (2017)
19. K.M. Owolabi, K.C. Patidar, A. Shikongo, Efficient numerical method for a model arising
in biological stoichiometry of tumour dynamics. Discret. Continuous Dyn. Syst. Ser. S 12,
591–613 (2019)
Exponential-Time-Differencing Method for the Solution of Diffusive HIV-I Model 23

20. K.M. Owolabi, High-dimensional spatial patterns in fractional reaction- diffusion system aris-
ing in biology. Chaos, Solitons Fractals 134, 109723 (2020)
21. E. Pindza, K.M. Owolabi, Fourier spectral method for higher order space fractional reaction-
diffusion equations. Commun. Nonlinear Sci. Numer. Simul. 40, 112–128 (2016)
22. Y. Saad, Analysis of some Krylov susbspace approximations to the matrix exponential operator.
SIAM J. Numer. Anal. 131, 209–228 (1992)
23. T. Schmelzer, L.N. Trefethen, Evaluating matrix functions for exponential integrators via
Carathéodory-Fejér approximation and contour integrals. Electron. Trans. Numer. Anal. 29,
1–18 (2007)
24. L.A. Segel, J.L. Jackson, Dissipative structure: An explanation and an ecological example. J.
Theor. Biol. 37, 545–559 (1972)
25. H. Singh, D. Baleanu, J. Singh, H. Dutta, Computational study of fractional order smoking
model. Chaos, Solitons Fractals 142, 110440 (2021)
26. H. Singh, H. Srivastava, D. Baleanu, Methods of Mathematical Modelling: Infectious Disease
(Academic, Cambridge, Massachusetts, 2022)
27. L.N. Trefethen, H.M. Gutknecht, The Carathéodory-Fejér method for real rational approxima-
tion. SIAM J. Numer. Anal. 20, 420–436 (1983)
28. L.N. Trefethen, M. Embere, Spectra and Pseudospectra: The Behavior of Nonnormal Matrices
and Operators (Princeton University Press, New Jersey, 2005)
29. L.N. Trefethen, Finite Difference and Spectral Methods for Ordinary and Partial Differential
Equations (Upson Hall Cornell University Ithaca, New York, 1996)
30. L.N. Trefethen, Spectral Methods in MATLAB (SIAM, Philadelphia, 2000)
31. P. Veeresha, L. Akinyemi, K. Oluwasegun, M. Şenol, B. Oduro, Numerical surfaces of fractional
Zika virus model with diffusion effect of mosquito-borne and sexually transmitted disease.
Math. Methods Appl. Sci. 45, 2994–3013 (2022)
An Effective Technique for Solving
a Model Describing Biological Species
Living Together

Saurabh Tomar and Soniya Dhama

Abstract In this work, a reliable scheme is suggested to obtain an analytical approx-

imation to the solution of a system of nonlinear integro-differential equations that
arises in biology. The dynamics of the populations of two distinct species are char-
acterized by this system. The suggested strategy involves two steps. Construct an
integral operator first, and then apply Halpern’s iteration scheme to the constructed
operator to develop the required iterative approach. This approach’s convergence
is also addressed. Numerous numerical examples are taken into consideration to
demonstrate the effectiveness of the strategy. The proposed approach has several
key advantages over existing approaches, including the ability to handle nonlinearity
without the use of Adomian polynomials, the ability to solve the problem without
the use of the Lagrange multiplier or constrained variations, and the ability to take
into account both interval endpoints. In contrast to existing semi-analytical methods,
the suggested approach addresses the problems without requiring linearization, dis-
cretization, or perturbation assumptions. Three test problems are taken into account
to confirm the method’s efficacy.

Keywords Mathematical biology · Integro-differential equation · The equilibrium

state of two species living together · Fixed point method

S. Tomar (B)
Department of Mathematics and Statistics, Indian Institute of Technology Kanpur,
Kanpur, UP 208016, India
e-mail: [email protected]
S. Dhama
Department of Mathematical Sciences, Rajiv Gandhi Institute of Petroleum Technology,
Jais Amethi 229304, UP, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 25
H. Singh and H. Dutta (eds.), Computational Methods for Biological Models,
Studies in Computational Intelligence 1109,
https://doi.org/10.1007/978-981-99-5001-0_2
26 S. Tomar and S. Dhama

1 Introduction

Differential equations, integral and delay integral equations, and their systems are
used to simulate many problems in engineering and research [1]. In general, func-
tional equations, such as integral and integro-differential equations, partial differen-
tial equations, stochastic equations, and others, are frequently the result of mathemati-
cal modelling of real-life problems[2–11]. Integro-differential equation has long been
considered one of the most important tools in practical mathematics. Integral equa-
tions are found in a variety of scientific and technical domains. Integro-differential
equations appear in many mathematical descriptions of physical processes. Fluid
dynamics, biological models, and chemical kinetics are all examples of this [12–29].
Integro-differential equations appear in a variety of physical situations [30, 31]. As
a result, computing the solutions to these equations is critical.
This study is concerned with the following system of nonlinear delay integro-
differential equations, which are associated with the dynamics of two interacting
species [32, 33]:
  t 
ds1
= s1 (t) κ1 − λ1 s2 (t) − F1 (t − τ )s2 (τ )dτ + G 1 (t), κ1 , λ1 > 0, 0 ≤ t ≤ T,
dt t−T0
  t 
ds2
= s2 (t) −κ2 + λ2 s1 (t) + F2 (t − τ )s1 (τ )dτ + G 2 (t), κ2 , λ2 > 0, 0 ≤ t ≤ T,
dt t−T0
(1.1)

subject to
s1 (0) = a1 , s2 (0) = a2 ,

where G 1 , G 2 , F1 and F2 are known functions while s1 and s2 are unknown.

Here, s1 and s2 are two different species’ numbers at time t, respectively, with
the profile being that the first species expands while the second contracts. If they
are together and we suppose that the second species consumes the first, then the
second species’ rate will increase at a rate of dsdt2 , which depends on both the current
populations of the second species and all previous values of the first species.
The replacement of a steady state condition between these two species is illustrated
by the following system
  t 
ds1
= s1 (t) κ1 − λ1 s2 (t) − F1 (t − τ )s2 (τ )dτ ,
dt t−T0
  t 
ds2
= s2 (t) −κ2 + λ2 s1 (t) + F2 (t − τ )s1 (τ )dτ , (1.2)
dt t−T0
An Effective Technique for Solving a Model Describing Biological … 27

where κ1 and −κ2 represent the first and second species’ rising and decreasing
coefficients, respectively. Here, T0 is considered as the finite heredity duration for
considered species s1 and s2 . Note that the above system (1.2) is a special case of
(1.1).
The evaluation of exact solutions for these problems is not an easy task, and
to be more precise, the exact solutions are available only in a few special cases.
Therefore, the evaluation of the solutions using numerical or semi-analytical schemes
is critical. This prompted the development of numerical or semi-analytical techniques
for finding solutions to problems. Over the past, some numerical as well as semi-
analytical techniques have been suggested to tackle the integro-differential equations.
For example, in [34], Babolian and Biazar used the Adomian decomposition method
to solve the problem. In [8], Shakeri and Dehghan applied the variational iteration
method to the problem and suggested a numerical scheme which is based on the
pseudospectral Legendre method to tackle this model. In [35], Yousefi developed
a numerical scheme to obtain the approximate solutions of this biological model
based on Legendre Multiwavelets. In [36], Sahu and Ray introduced the various
schemes to handle this system of integro-differential equations based on Legendre
spectral collocation. In [37], Shakourifar and Dehghan analysis the numerical issues
of a piecewise polynomial collocation scheme for this system. In [38], Tari used
the differential transform approach to solve this model. In [16], Yüzbaşı and Sezer
proposed a numerical methodology based on exponential polynomials to solve this
model. In [39], Ramezani et al. suggested a complex spline collocation approach to
handle such type of model. In [40], Hafez et al. have obtained the numerical solution
to this model via Bernoulli collocation scheme. In [12], Sekar and Murugesan have
suggested a single term Walsh series approach to solve this model. In [41], Thirumalai
at el. have presented a spectral collocation methodology to solve this model.
Fixed-point iterative methods have been studied for decades in order to better
understand many applications in applied mathematics, and numerous fixed-point
iterative strategies have been utilized to tackle a wide range of problems [42–52].
In this work, an efficient iterative methodology based on the Halpern fixed point
approach is designed to solve the considered biological model. In the construction
of the methodology, an equivalent integral operator form of the model is obtained,
and then Halpern’s approach is combined with this integral representation. As a
result, the methodology involves an unknown parameter that greatly accelerates the
convergence of the obtained solutions. A method based on residual minimization is
also provided to assess the unidentified parameter. The significance of this unknown
parameter in the scheme is that it greatly accelerates the convergence of the obtained
solutions. In the case of problems with strong nonlinearity or challenging integrable
functions, the integration becomes very challenging to evaluate for further iterations
or impossible to proceed with further iterations. To remedy this, Taylor’s series is
utilized to simplify the iterative process and to improve the computational efficacy
of the methodology. The convergence analysis of the methodology is also discussed.
To verify the method’s effectiveness, some numerical examples are considered.
28 S. Tomar and S. Dhama

This chapter is organized as follows. In Sect. 2, the proposed methodology to solve

(1.1) is presented. In Sect. 3, numerical simulations are given and, the conclusion is
summarized in Sect. 4.

2 Development of the Methodology

In this section, we formulate a methodology by constructing different integral opera-

tors to obtain the approximate analytical solutions for the model (1.1). Let us consider
the model
  t 
ds1
= s1 (t) κ1 − λ1 s2 (t) − F1 (t − τ )s2 (τ )dτ + G 1 (t),
dt t−T0
  t 
ds2
= s2 (t) −κ2 + λ2 s1 (t) + F2 (t − τ )s1 (τ )dτ + G 2 (t). (2.1)
dt t−T0

2.1 Type-I

Re-writing (2.1) in the following manner

  t 
ds1
1 = − s1 (t) κ1 − λ1 s2 (t) − F1 (t − τ )s2 (τ )dτ − G 1 (t) = 0,
dt t−T0
  t 
ds2
2 = − s2 (t) −κ2 + λ2 s1 (t) + F2 (t − τ )s1 (τ )dτ − G 2 (t) = 0,
dt t−T0
(2.2)

Now adding and subtracting dsdt1 and dsdt2 in the system (2.2), respectively as follows

ds1 ds1
+ 1 − = 0,
dt dt
ds2 ds2
+ 2 − = 0, (2.3)
dt dt
and re-write (2.3) as

ds1 ds1
= −1 + ,
dt dt
ds2 ds2
= −2 + . (2.4)
dt dt
An Effective Technique for Solving a Model Describing Biological … 29

Now, let us construct an equivalent integral form of the system (2.4) by integrating
(2.4) from 0 to t as
 t 
ds1 (u)
s1 (t) = s1 (0) + − 1 (u) du,
0 du
 t 
ds2 (u)
s2 (t) = s2 (0) + − 2 (u) du. (2.5)
0 du

Now using the following identities

 t 
ds1 (u)
du = s1 (t) − s1 (0),
0 du
 t 
ds2 (u)
du = s2 (t) − s2 (0), (2.6)
0 du

the system (2.5) resulting in the following system

 t
s1 (t) = s1 (t) − (1 (u)) du,
0
 t
s2 (t) = s2 (t) − (2 (u)) du. (2.7)
0

Now using (2.2) and (2.7), we get

 t   u 
ds1
s1 (t) = s1 (t) − − s1 (u) κ1 − λ1 s2 (u) − F1 (u − τ )s2 (τ )dτ − G 1 (u) du,
0 du u−T0
 t   u 
ds2
s2 (t) = s2 (t) − − s2 (u) −κ2 + λ2 s1 (u) + F2 (u − τ )s1 (τ )dτ − G 2 (u) du.
0 du u−T0
(2.8)

Now re-write (2.8) in the operator form

s1 (t) = T1 [s1 , s2 ],
s2 (t) = T2 [s1 , s2 ], (2.9)

where
 t   u 
ds1
T1 [s1 , s2 ] = s1 (t) − − s1 (u) κ1 − λ1 s2 (u) − F1 (u − τ )s2 (τ )dτ − G 1 (u) du,
0 du u−T0
 t   u 
ds2
T2 [s1 , s2 ] = s2 (t) − − s2 (u) −κ2 + λ2 s1 (u) + F2 (u − τ )s1 (τ )dτ − G 2 (u) du.
0 du u−T0
(2.10)
30 S. Tomar and S. Dhama

Now, in view of (2.10), the Halpern fixed-point scheme will be implemented, and
the Halpern’s approach [42–45] for system of equations is given as

s1,n+1 (t) = αn v + (1 − αn )T1 [s1,n , s2,n ], v = s1,0 ,

s2,n+1 (t) = αn v + (1 − αn )T2 [s1,n , s2,n ], v = s2,0 , n ≥ 0, (2.11)

From the combination of (2.10) and (2.11), we get the following iterative approach
 t
ds1,n
s1,n+1 (t) =αn s1,0 + (1 − αn )s1,n (t) − (1 − αn ) − s1,n (u) κ1 − λ1 s2,n (u)
0 du
 u   
− F1 (u − τ )s2,n (τ )dτ − G 1 (u) du , n ≥ 0, (2.12)
u−T0
 t
ds2,n
s2,n+1 (t) =αn s2,0 + (1 − αn )s2,n (t) − (1 − αn ) − s2,n (u) κ2 − λ2 s1,n (u)
0 du
 u   
− F2 (u − τ )s1,n (τ )dτ − G 2 (u) du , n ≥ 0. (2.13)
u−T0

Begin with the initial guess s1,0 = a1 and s2,0 = a2 , which satisfies the given initial
conditions for the system, and for the aforementioned problem, analytical iterative
approximations can be obtained. Now, by establishing the residual norm, we outline
the process for evaluating αn

1 
N 2
R1 (αn ) ≈ R1,r (t j ; αn ) , (2.14)
N i=1

1 
N 2
R2 (αn ) ≈ R2,r (t j ; αn ) , (2.15)
N i=1

where ti , 1 ≤ i ≤ N are the discrete grids of [0, T ] and minimizing R1 (αn ) and
R2 (αn ) for αn to evaluate αn , where R1,r (t; αn ) and R2,r (t; αn ) are the residual errors
defined by
  t 
ds1,n (t)
R1,r (t; αn ) ≡ − s1,n (t) κ1 − λ1 s2,n (t) − F1 (t − τ )s2,n (τ )dτ − G 1 (t),
dt t−T0

  t 
ds2,n (t)
R2,r (t; αn ) ≡ − s2,n (t) −κ2 + λ2 s1,n (t) + F2 (t − τ )s1,n (τ )dτ − G 2 (t).
dt t−T0

In cases of strong nonlinearity or complicated integrands, the iterative process of

the scheme requires more computational work and is challenging to compute further
iterations. To minimize computational work, Taylor’s series is used to simplify the
integrand in the iterative process.
An Effective Technique for Solving a Model Describing Biological … 31

2.2 Type-II

Now, let us construct a different integral operator to formulate the methodology.

First, re-writing (2.1) in the following manner
  t 
ds1
1 = − s1 (t)κ1 + s1 (t) −λ1 s2 (t) − F1 (t − τ )s2 (τ )dτ − G 1 (t) = 0,
dt t−T0
  t 
ds2
2 = + s2 (t)κ2 − s2 (t) λ2 s1 (t) + F2 (t − τ )s1 (τ )dτ − G 2 (t) = 0,
dt t−T0
(2.16)

Now adding and subtracting ds1

dt
− s1 (t)κ1 and ds2
dt
+ s2 (t)κ2 in system (2.16),
respectively as follows
 
ds1 ds1
− s1 (t)κ1 + 1 − − s1 (t)κ1 = 0,
dt dt
 
ds2 ds2
+ s2 (t)κ2 + 2 − + s2 (t)κ2 = 0, (2.17)
dt dt

and re-write (2.17) as

ds1 ds1
− s1 (t)κ1 = −1 + − s1 (t)κ1 ,
dt dt
ds2 ds2
+ s2 (t)κ2 = −2 + + s2 (t)κ2 . (2.18)
dt dt
Now, let us construct an equivalent integral form of the system (2.18) and for that
treating (2.18) by means of the method of variation of parameters as
  
t
ds1 (u)
−κ1 (u−t)
s1 (t) = s1 (0) + e − 1 (u) du,
0 du
 t  
ds2 (u)
s2 (t) = s2 (0) + eκ2 (u−t) − 2 (u) du. (2.19)
0 du

Now using the integrated by parts, (2.19) resulting in the following system
 t
s1 (t) = s1 (t) − e−κ1 (u−t) (1 (u)) du,
0
 t
s2 (t) = s2 (t) − eκ2 (u−t) (2 (u)) du. (2.20)
0
32 S. Tomar and S. Dhama

In view of (2.16) and (2.20), we get

 t   u 
ds1
s1 = s1 − e−κ1 (u−t) − s1 (u) κ1 − λ1 s2 (u) − F1 (u − τ )s2 (τ )dτ − G 1 (u) du,
0 du u−T0
 t   u 
ds2
s2 = s2 − eκ2 (u−t) − s2 (u) −κ2 + λ2 s1 (u) + F2 (u − τ )s1 (τ )dτ − G 2 (u) du.
0 du u−T0
(2.21)

Re-write (2.21) in the operator form

s1 (t) = T1 [s1 , s2 ],
s2 (t) = T2 [s1 , s2 ], (2.22)

where
 t
ds1
T1 [s1 , s2 ] = s1 (t) − e−κ1 (u−t) − s1 (u) κ1 − λ1 s2 (u)
du
 u 0
 
− F1 (u − τ )s2 (τ )dτ − G 1 (u) du,
u−T0

 t
ds2
T2 [s1 , s2 ] = s2 (t) − eκ2 (u−t) − s2 (u) − κ2 + λ2 s1 (u)
du
 u 0
 
+ F2 (u − τ )s1 (τ )dτ − G 2 (u) du.
u−T0

Now, in view of (2.22) and the Halpern fixed-point scheme will be implemented,
and from the combination of (2.22) and (2.11), we get the following iterative approach
 t
ds1,n
s1,n+1 (t) =αn s1,0 + (1 − αn )s1,n (t) − (1 − αn ) e−κ1 (u−t) − s1,n (u) κ1 − λ1 s2,n (u)
0 du
 u   
− F1 (u − τ )s2,n (τ )dτ − G 1 (u) du , n ≥ 0, (2.23)
u−T0

 t
ds2,n
s2,n+1 (t) =αn s2,0 + (1 − αn )s2,n (t) − (1 − αn ) eκ2 (u−t) − s2,n (u) κ2 − λ2 s1,n (u)
0 du
 u   
− F2 (u − τ )s1,n (τ )dτ − G 2 (u) du , n ≥ 0. (2.24)
u−T0

Begin with the initial guess s1,0 = a1 and s2,0 = a2 , which satisfies the given
initial conditions for the system, and for the aforementioned problem, analytical
iterative approximations can be obtained. The evaluation of αn can be achieved via
the residual norm (2.14) and (2.15).
An Effective Technique for Solving a Model Describing Biological … 33

In cases of strong nonlinearity or complicated integrands, the iterative process

of the scheme requires more computational work and is difficult to compute further
iterates. To minimize computational work, multivariate Taylor’s series is used to
simplify the integrand in the iterative process (2.23) and (2.24) as follows
 t 
s1,n+1 (t) =αn s1,0 + (1 − αn )s1,n (t) − (1 − αn ) Pr (t, u)du , (2.25)
0

 t 
s2,n+1 (t) =αn s2,0 + (1 − αn )s2,n (t) − (1 − αn ) Q r (t, u)du , (2.26)
0

where

 u  
ds1,n
e−κ1 (u−t) − s1,n (u) κ1 − λ1 s2,n (u) − F1 (u − τ )s2,n (τ )dτ − G 1 (u)
du u−T0
= Pr (t, u) + O(t r +1 ) + O(u r +1 ),

and
 u  
κ2 (u−t) ds2,n
e − s2,n (u) κ2 − λ2 s1,n (u) − F2 (u − τ )s1,n (τ )dτ − G 2 (u)
du u−T0
= Q r (t, u) + O(t r +1 ) + O(u r +1 ).

3 Implementation and Numerical Results

This section presents the numerical simulation of the proposed methodology. To

verify the applicability and efficacy of our approach, we consider some examples.
Example 3.1 First, consider the following system [8]
  t 
ds1 1
= s1 (t) 1 − s2 (t) − s2 (τ )dτ + G 1 (t),
dt 3 t− 21
  t 
ds2
= s2 (t) −2 + s1 (t) + (t − τ − 1)s1 (τ )dτ + G 1 (t), (3.1)
dt t− 21

where
5 49 17 23
G 1 (t) = − t 3 + t 2 + t − ,
2 12 12 6

15 3 1 2 3
G 2 (t) = t − t − t − 1,
8 4 8
34 S. Tomar and S. Dhama

with initial data s1 (0) = 1 and s2 (0) = 0. The true solution is s1 (t) = 1 − 3t and
s2 (t) = −t + t 2 .
The proposed methodology type-I formula for (3.1) is
t 
ds1,n 1
s1,n+1 (t) =αn s1,0 + (1 − αn )s1,n (t) − (1 − αn ) − s1,n (u) 1 − s2 (u)
du 3
 u   
0

− s2,n (τ )dτ − G 1 (u) du , n ≥ 0, (3.2)

u−1/2

 t
ds2,n
s2,n+1 (t) =αn s2,0 + (1 − αn )s2,n (t) − (1 − αn ) − s2,n (u) − 2 + s1,n (u)
0 du
 u   
− (u − τ − 1)s1,n (τ )dτ − G 2 (u) du , n ≥ 0. (3.3)
u−1/2

By selecting the initial approximation s1,0 (0) = 1 and s2,0 (0) = 0, we obtain the
following results

s1,1 (t) =1 − 2.833333333t − 0.6250000000t 4 + 1.361111111t 3 + 0.7083333333t 2

+ 2.833333α0 t + 0.6250000000α0 t 4 − 1.361111111α0 t 3 − 0.7083333333α0 t 2 ,

s2,1 (t) = − t + 0.4687500000t 4 − 0.08333333333t 3 + 0.1875000000t 2 + α0 t

− 0.4687500000α0 t 4 + 0.08333333333α0 t 3 − 0.1875000000α0 t 2 ,

and the corresponding values of α0 for s1,1 (t) and s2,1 (t) are identified via (2.14) and
(2.15) with N = 10, which are 0.212566491334 and 0.5159049808, respectively.
After substituting these values in the above approximations yield

s1,1 (t) =1.0 − 2.23106160762256t − 0.492145942915818t 4 + 1.07178449781807t 3

+ 0.557765401945012t 2 + 0.08333333333α0 t 3 − 0.1875000000α0 t 2 ,

s2,1 (t) = − 0.484095019159353t + 0.226919540230947t 4 − 0.0403412515949991t 3

+ 0.0907678160923787t 2 .

In a similar way, we get

s1,2 (t) =1 − 2.61929358205884t − 0.174965099041535t 4 + 1.04272781454564t 3

+ 0.00934064327203696t 9 − 0.0310576388745547t 8 + 0.0160985245420873t 7
+ 0.0190709989334935t 6 − 0.105091677912895t 5 − 0.0998384249658t 2 ,
An Effective Technique for Solving a Model Describing Biological … 35

s2,2 (t) = − 0.805746743313668t + 0.292932489183872t 4 + 0.0377332701136180t 3

− 0.00499910038799772t 9 + 0.00762362808881598t 8 + 0.0170794661567121t 7
− 0.0195895303212047t 6 − 0.0686099838784783t 5 + 0.498677762195t 2 ,

and

s1,3 (t) =1 − 2.92923543071108t + 0.00000202932247142605t 19 − 0.0000132807881680772t 18

+ 0.0000260745045911674t 17 + 4.1633476063582510−7 t 16 − 0.0000519695905629862t 15
− 0.000205884840310831t 14 + 0.00128990424794279t 13 − 0.00223452861649316t 12
+ 0.000363556855104520t 11 + 0.00510059589298002t 10 + 0.00369794905646681t 9
− 0.0333932807050378t 8 + 0.0141485215915262t 7 − 0.000803309330564085t 6
+ 0.0229932835874042t 5 − 0.0659363332384586t 4 + 0.458240373533745t 3
− 0.0343263814770440t 2 ,

s2,3 (t) = − 0.956632866055380t − 0.00000117552087300173t 19 + 0.00000322615172814328t 18

+ 0.0000127488126725305t 17 − 0.0000490257239538026t 16 + 0.00000826857412053931t 15
+ 0.000253183655953299t 14 − 0.000374437492100575t 13 − 0.00100124698471227t 12
+ 0.00351654888558956t 11 − 0.00313378894737590t 10 − 0.00606926377948803t 9
+ 0.000606410328391509t 8 + 0.0358975526509382t 7 + 0.00888009012000952t 6
− 0.149865172735764t 5 + 0.185460620094t 4 + 0.011425647744927t 3 + 0.875460413974t 2 .

The achieved solutions along with the true solutions are represented in Figs. 1,
2, 3 and 4. From figures, it is clear that as the obtained solutions via the proposed
methodology type-I is really well in line with the precise solutions as iterations
progress.
Now, let us implement the proposed methodology type-II formula for (3.1) which
is given by
 t
ds1,n 1
s1,n+1 (t) =αn s1,0 + (1 − αn )s1,n (t) − (1 − αn ) e−(u−t) − s1,n (u) 1 − s2 (u)
0 du 3
 u   
− s2,n (τ )dτ − G 1 (u) du , n ≥ 0, (3.4)
u−1/2

 t
ds2,n
s2,n+1 (t) =αn s2,0 + (1 − αn )s2,n (t) − (1 − αn ) e2(u−t) − s2,n (u) − 2 + s1,n (u)
0 du
 u   
− (u − τ − 1)s1,n (τ )dτ − G 2 (u) du , n ≥ 0. (3.5)
u−1/2
36 S. Tomar and S. Dhama

(a) s1,2 (t) with exact (b) s2,2 (t) with exact

Fig. 1 Graph of exact and approximate solutions for Example 3.1 for n = 1 (Type-I)

(a) s1,3 (t) with exact (b) s2,3 (t) with exact

Fig. 2 Example 3.1 for n = 2 (Type-I), graph of exact and approximate solutions

(a) s1,4 (t) with exact (b) s2,4 (t) with exact

Fig. 3 Example 3.1 for n = 3 (Type-I), graph of exact and approximate solutions
An Effective Technique for Solving a Model Describing Biological … 37

(a) s1,6 (t) with exact (b) s2,6 (t) with exact

Fig. 4 Example 3.1 for n = 5 (Type-I), graph of exact and approximate solutions

By selecting the initial approximation s1,0 (0) = 1 and s2,0 (0) = 0, we obtain the
following results

s1,1 (t) =1.0 − 2.87547562844537t − 6.4029772411110610−12 t 15 − 9.6025091459746710−11 t 14

− 1.3445503821511210−9 t 13 − 1.7479498591924610−8 t 12 − 2.0975272192356710−7 t 11
− 0.00000230729129406353t 10 − −0.000207656472353199t 8 − 0.00166125200690230t 7
− 0.0116287618645051t 6 − 0.0697725703916233t 5 − 0.348862852182235t 4
+ 1.14173297037919t 3 − 0.718868906857623t 2 − 4.00151034981323 + ...

s2,1 (t) = − 0.927879703429945t − 3.1606508862982310−8 t 15 + 2.3706248001542110−7 t 14

− 0.00000165935349750096t 13 + 0.0000107858071018801t 12 − 0.0000647148675836570t 11
+ 0.000355932674395691t 10 − 00320339421135040t 7 + 0112118797587480t 6
− 0336356392361385t + 0840890981449893t 4 − 0811894740454808t 3
5

+ 110185714782306t 2 + 39513081604731010−9 t 16 + ...

and the corresponding values of α0 for s1,1 (t) and s2,1 (t) are identified via (2.14) and
(2.15) with N = 10 which are −0.01487375133 and 0.072120296570, respectively.
The obtained approximate solutions along with the exact solutions are represented
in Figs. 5, 6, and 7. From figures, it is clear that as the obtained solutions via the
proposed methodology type-II is really well in line with the precise solutions as
iterations progress. As it is evident from figures that methodology type-II provides
better results compare to methodology type-I with less number of iterations.
The dynamics of the populations of two distinct species are depicted in Fig. 7,
which demonstrates that as time increases, the population of the first species decays
gradually throughout the domain, whereas the population of the second species
38 S. Tomar and S. Dhama

(a) s1,2 (t) with exact (b) s2,2 (t) with exact

Fig. 5 Example 3.1 for n = 1 (Type-II), graph of exact and approximate solutions

(a) s1,3 (t) with exact (b) s2,3 (t) with exact

Fig. 6 Example 3.1 for n = 2 (Type-II), graph of exact and approximate solutions

decays gradually upto t = 0.5 and grows gradually. Also, the absolute errors are
represented in Tables 1, 2, 3 and 4 to show the accuracy of the methodology type-I
and type-II.

Example 3.2 Next, consider the following system [8]

  t 
ds1
= s1 (t) 2 − s2 (t) − (2(t − τ ) − 3)s2 (τ )dτ + G 1 (t),
dt t− 13
  t 
ds2
= s2 (t) −2 + s1 (t) + (t − τ )s1 (τ )dτ + G 1 (t), (3.6)
dt t− 13
An Effective Technique for Solving a Model Describing Biological … 39

(a) s1,4 (t) with exact (b) s2,4 (t) with exact

Fig. 7 Example 3.1 for n = 3 (Type-II), graph of exact and approximate solutions

Table 1 Absolute errors for Example 3.1 of type-I

t |s1,3 (t) − s1 (t)| |s1,4 (t) − s1 (t)| |s1,8 (t) − s1 (t)|
0 0.0 0.0 0.0
0.1 0.00718507006898716 0.000891279284103597 0.000889494472576802
0.2 0.0163476879219852 0.00216035479162602 0.00112653138860758
0.3 0.0300346962119031 0.00454282171553791 0.00115465578042079
0.4 0.0506879676833916 0.00893980179413609 0.00111615352292521
0.5 0.0806577754170309 0.0164263431020639 0.00100324038881405
0.6 0.122187549122741 0.0282479343832137 0.000754709480734972
0.7 0.177350976276551 0.0457845059745663 0.000308203207529667
0.8 0.247924969052615 0.0704658826323428 0.000384502793038699
0.9 0.335189924569745 0.103630472896985 0.00137329894734139
1.0 0.439662305629278 0.146329608018994 0.00275032875682268

where  
−t 7 −t 13 1 −t 22 1 −t
G 1 (t) = t 2 − 3te − e + te
2 3 + e 3 − 2t,
2 6 9

1 −t  
G 2 (t) = e 342t 3 − 8t 2 + 325t + 324 ,
648

with initial data s1 (0) = 0 and s2 (0) = 0. The true solution is s1 (t) = −t 2 and s2 (t) =
1 −t
2
te .
40 S. Tomar and S. Dhama

Table 2 Absolute errors for Example 3.1 of type-I

t |s2,3 (t) − s2 (t)| |s2,4 (t) − s2 (t)| |s2,8 (t) − s2 (t)|
0 0.0 0.0 0.0
0.1 0.003119803062820975 0.000931011619598313 0.00107219324549537
0.2 0.004033054674715708 0.000698492761385477 0.000304756199101336
0.3 0.00326236043635339 0.0000859071352009855 0.00105146411550314
0.4 0.00145872782526993 0.000899328263183508 0.00219590295873545
0.5 0.000707024021335934 0.00130478506802512 0.00270368146535593
0.6 0.00260428631965243 0.000991188466249460 0.00245228524538504
0.7 0.00365090520793163 0.000149542530616359 0.00154968498192945
0.8 0.00327501981361736 0.00192046348971511 0.000264227288312852
0.9 0.000846028496138101 0.00371004855083354 0.00104453386956990
1.0 0.00439773375375940 0.00439574478085381 0.00199883823548036

Table 3 Absolute errors for Example 3.1 of type-II

t |s1,2 (t) − s1 (t)| |s1,3 (t) − s1 (t)| |s1,4 (t) − s1 (t)|
0 0.0 0.0 0.0
0.1 0.000891633770617073 0.000600464801671752 0.000798065720892693
0.2 0.00130775258405169 0.000964971366172018 0.000983812315787647
0.3 0.00157177926884652 0.00125164153468421 0.000987091330415968
0.4 0.00168513820878941 0.00149510935742414 0.000957606830722130
0.5 0.00159314894012341 0.00166281063303275 0.000915304446242882
0.6 0.00136337912438966 0.00170516304411017 0.000843768303506631
0.7 0.00128079560935612 0.00158368883206395 0.000739385128795567
0.8 0.00186567285102290 0.00126767490137714 0.000626173117829198
0.9 0.00381236263717399 0.000696185124954063 0.000544500688943739
1.0 0.00783021374842985 0.000289768042434080 0.000522043528108362

Table 4 Absolute errors for Example 3.1 of type-II

t |s2,2 (t) − s2 (t)| |s2,3 (t) − s2 (t)| |s2,4 (t) − s2 (t)|
0 0.0 0.0 0.0
0.1 0.000593366270215279 0.000541767646333416 0.000396238727569748
0.2 0.00211686352074594 0.000337971205008752 0.000604898093448963
0.3 0.00369578102524581 0.00154060099911887 0.00193566961903682
0.4 0.00479926918788698 0.00240675089359949 0.00293405391055046
0.5 0.00526171571140666 0.00263220868468192 0.00329165641498311
0.6 0.00524668917517690 0.00219593535811591 0.00299218410610402
0.7 0.00517653869971149 0.00129061657382865 0.00224806226289570
0.8 0.00563692347044462 0.000251270711093654 0.00143951209357457
0.9 0.00725087080170486 0.000516518877662742 0.00106003706498801
1.0 0.0105017080812198 0.000598803993307098 00.00167292149102660
An Effective Technique for Solving a Model Describing Biological … 41

The proposed methodology type-I formula for (3.6) is

t 
ds1,n
s1,n+1 (t) =αn s1,0 + (1 − αn )s1,n (t) − (1 − αn ) − s1,n (u) 2 − s2 (u)
du
 u 
0
 
− 2(u − τ ) − 3)s2,n (τ )dτ − G 1 (u) du , n ≥ 0, (3.7)
u−1/3

 t
ds2,n
s2,n+1 (t) =αn s2,0 + (1 − αn )s2,n (t) − (1 − αn ) − s2,n (u) − 2 + s1,n (u)
0 du
 u   
− (u − τ )s1,n (τ )dτ − G 2 (u) du , n ≥ 0. (3.8)
u−1/3

By selecting the initial approximation s1,0 (0) = 0 and s2,0 (0) = 0, we obtain the
following results

s1,1 (t) = − 0.268431862536286t 2 + 0.171035570038697t 3 + 0.00753822992804326t 4

− 0.00365488244947080t 5 + 0.00119290910790607t 6 − 0.000293694245766914t 7
+ 0.0000580588950322939t 8 − 0.00000958833992998826t 9 + 0.00000135968942726075t 10
− 1.6893094679747210−7 t 11 + 1.8674713141384310−8 t 12 − 1.8593996319187410−9 t 13
+ 1.6840736977851610−10 t 14 − 1.3988522581440510−11 t 15 + 1.0729827747934810−12 t 16 ,

s2,1 (t) =0.220389070542336t + 0.000340106590343112t 2 − 0.0387721512991148t 3

+ 0.0779694358361584t 4 − 0.0526031526397346t 5 + 0.0207663415620332t 6
− 0.00579071958938951t 7 + 0.00125021225647604t 8 − 0.000220506938170207t 9
+ 0.0000329056500230278t 10 − 0.00000425825093366597t 11 + 4.8667173422000310−7 t 12
− 4.9819359161595810−8 t 13 + 4.6192748435408510−9 t 14
− 3.9148526150224210−10 t 15 + 3.0556453036649410−11 t 16 ,

and the corresponding values of α0 for s1,1 (t) and s2,1 (t) are identified via (2.14) and
(2.15) with N = 10 which are 0.7315681374637 and 0.55922185891, respectively.
The achieved solutions along with the true solutions are represented in Figs. 8, 9,
10 and 11. From figures, it is clear that as the obtained solutions via the proposed
methodology type-I is really well in line with the precise solutions as iterations
progress. However, this approach requires more iterations to achieve a good agree-
ment with the exact solutions.
Now, let us implement the proposed methodology type-II formula for (3.6) which
is given by
 t
ds1,n
s1,n+1 (t) =αn s1,0 + (1 − αn )s1,n (t) − (1 − αn ) e−2(u−t) − s1,n (u) 2 − s2 (u)
0 du
 u   
− 2(u − τ ) − 3)s2,n (τ )dτ − G 1 (u) du , (3.9)
u−1/3
42 S. Tomar and S. Dhama

(a) s1,2 (t) with exact (b) s2,2 (t) with exact

Fig. 8 Example 3.2 for n = 1 (Type-I), graph of exact and approximate solutions

(a) s1,3 (t) with exact (b) s2,3 (t) with exact

Fig. 9 Example 3.2 for n = 2 (Type-I), graph of exact and approximate solutions

(a) s1,4 (t) with exact (b) s2,4 (t) with exact

Fig. 10 Example 3.2 for n = 5 (Type-I), graph of exact and approximate solutions
An Effective Technique for Solving a Model Describing Biological … 43

(a) s1,6 (t) with exact (b) s2,6 (t) with exact

Fig. 11 Example 3.2 for n = 7 (Type-I), graph of exact and approximate solutions

 t
ds2,n
s2,n+1 (t) =αn s2,0 + (1 − αn )s2,n (t) − (1 − αn ) e2(u−t) − s2,n (u) − 2 + s1,n (u)
0 du
 u   
− (u − τ )s1,n (τ )dτ − G 2 (u) du . (3.10)
u−1/3

By selecting the initial approximation s1,0 (0) = 0 and s2,0 (0) = 0, we obtain the
following results
1 1
s1,1 (t) =0.785569668081790e 3 t 3 + 0.325832873693015e 3 t 4
1 1
− 0.0517761826690271e 3 t 5 + 0.0913522303413294e 3 t 6
1 1
− 0.00960707494515600e t 7 + 0.00602301960748580e 3 t 8
3

1 1
− 0.000232029280570617e 3 t 9 + 0.000196959206102791e 3 t 10
1
+ 0.00000344581998910735e11 + 0.00000435043719156179e 3 t 12
1 1
+ 2.7645142997530610−7 e 3 t 13 + 7.6393392306879610−8 e 3 t 14
1
+ 7.0250300276984610−9 e 3 t 15 − 0.964108229009469t 2
− 1.12479293384438t 3 − 0.441882938296007t 4 + 0.0642738819339645t 5
− 0.125869685454014t 6 + 0.0128165181237371t 7 − 0.00834508362907489t 8
+ 0.000302877915859399t 9 − 0.000274184218567433t 10 − 0.00000528949470284348t 11
− 0.00000608452910460047t 12 − 3.9449808286074610−7 t 13 − 1.0725057208143610−7 t 14
1
− 9.9391280887061210−9 t 15 − 1.5859124474361510−9 t 16 + 1.1270339535231910−9 t 16 e 3
44 S. Tomar and S. Dhama

(a) s1,2 (t) with exact (b) s2,2 (t) with exact

Fig. 12 Example 3.2 for n = 1 (Type-II), graph of exact and approximate solutions

s2,1 (t) =0.433780558559770t − 0.433111144117548t 2 + 0.212427516331739t 3

+ 0.0472495027134935t 4 − 0.122435901482380t 5 + 0.0816852973121234t 6
− 0.0347362341756747t 7 + 0.0111447875156318t 8 − 0.00291063199931374t 9
+ 0.000646892911249343t 10
− 0.000125998190919646t 11 + 0.0000219575896206182t 12
− 0.00000347614761264751t 13 + 5.0568439899227710−7 t 14 − 6.8195126990510710−8 t 15
−9 16
+ 8.5845335806808510 t ,

and the corresponding values of α0 for s1,1 (t) and s2,1 (t) are identified via (2.14)
and (2.15) with N = 10 which are 0.0358917716425683 and 0.13243888254754,
respectively.
The obtained approximate solutions along with the exact solutions are represented
in Figs. 12, and 13. From figures, it is clear that as the obtained solutions via the
proposed methodology type-II is really well in line with the precise solutions as
iterations progress. As it is evident from figures that methodology type-II provides
better results compare to methodology type-I with less number of iterations. Even
the obtained solutions via type-II using second iteration are in good agreement with
true solutions.
The dynamics of the populations of two distinct species are depicted in Fig. 13,
which demonstrates that as time increases, the population of the first species decays
gradually, whereas the population of the second species grows gradually throughout
the domain. Also, the absolute errors are represented in Tables 5, 6, 7 and 8 to show
the accuracy of the methodology type-I and type-II.
An Effective Technique for Solving a Model Describing Biological … 45

(a) s1,3 (t) with exact (b) s2,3 (t) with exact

Fig. 13 Example 3.2 for n = 2 (Type-II), graph of exact and approximate solutions

Table 5 Absolute errors for Example 3.2 of type-I

t |s1,3 (t) − s1 (t)| |s1,4 (t) − s1 (t)| |s1,8 (t) − s1 (t)|
0 0.0 0.0 0.0
0.1 0.00480136916120316 0.00322069191865686 0.0000174249722959229
0.2 0.0200936701738036 0.0137960584587350 0.0000306391477422549
0.3 0.0473264952538380 0.0332561022468991 0.000285078738717887
0.4 0.0881090249113091 0.0633571248342677 0.000846715759702549
0.5 0.144215549956031 0.106101246634957 0.00172996705824874
0.6 0.217592173174226 0.163758817518988 0.00279929801952972
0.7 0.310364193552304 0.238892728265536 0.00368516823133852
0.8 0.42484391807577 0.334383942629798 0.00367579102989757
0.9 0.5635388115645 0.453457727331305 0.00158076815295083
1.0 0.72915999432191 0.599710105313309 0.00443750096521933

Table 6 Absolute errors for Example 3.2 of type-I

t |s2,3 (t) − s2 (t)| |s2,4 (t) − s2 (t)| |s2,8 (t) − s2 (t)|
0 0.0 0.0 0.0
0.1 0.0036819673240259185 0.00330246881953400 0.000130732210885096
0.2 0.005258511510903147 0.00536190186866586 0.000333249984681358
0.3 0.005439147037049161 0.00643120982770425 0.0005193557798276464
0.4 0.004631057495267987 0.00656948321984713 0.000626890117261719
0.5 0.003020017607675979 0.00571427145600914 0.000617553144770078
0.6 0.000627363524772967 0.00373688127726224 0.000478472927137291
0.7 0.00265059426164335 0.000486260472108258 0.000225326285604482
0.8 0.00701575289818374 0.00417528910537326 0.0000948103089639662
0.9 0.0127499738355621 0.0103430959730567 0.000402306736056013
1.0 0.0202020329006536 0.0180478121563033 0.000586362971222137
46 S. Tomar and S. Dhama

Table 7 Absolute errors for Example 3.2 of type-II

t |s1,2 (t) − s1 (t)| |s1,3 (t) − s1 (t)| |s1,4 (t) − s1 (t)|
0 0.0 0.0 0.0
0.1 0.0000970882166753675 0.000120193078859533 0.000123672826318245
0.2 0.000372857145899987 0.000479683597553831 0.000493708963232158
0.3 0.000786507026367282 0.00105181080713520 0.00108323941436941
0.4 0.00128044471014857 0.00178164106987519 0.00183612206692332
0.5 0.00179021027021825 0.00259113472445216 0.00267127296906178
0.6 0.00225415147462876 0.00338198583480465 0.00348519763166022
0.7 0.00262131623992051 0.00403506324793529 0.00415263796687698
0.8 0.00285690671206773 0.00440531479661943 0.00452541282204355
0.9 0.00294515883021340 0.00431100804699747 0.00442982283818383
1.0 0.00288993082563893 0.00351629365242401 0.00366364829838384

Table 8 Absolute errors for Example 3.2 of type-II

t |s2,2 (t) − s2 (t)| |s2,3 (t) − s2 (t)| |s2,4 (t) − s2 (t)|
0 0.0 0.0 0.0
0.1 0.000405054818801584 0.000217970742560261 0.000250564757945076
0.2 0.000711462381281988 0.000391988256563461 0.000451164032148690
0.3 0.000890379346648068 0.000523349782605056 0.000603959806075624
0.4 0.000900915503568023 0.000603062752819683 0.000699613712112496
0.5 0.000715845650909430 0.000620852639215519 0.000725268421813668
0.6 0.000343520163927885 0.000571135014163976 0.000670025330357010
0.7 0.000154136222035667 0.000455237391732782 0.000528313970022870
0.8 0.000648150240422335 0.000278797425358718 0.000301602171264376
0.9 0.000934087631095121 0.0000431798322647892 8.0888545300750210−7
1.0 0.000728922919670083 0.000270075922915725 0.000360588313551208

Example 3.3 Next, consider the following system [8]

  t 
ds1 1
= s1 (t) − 2s2 (t) − s2 (τ )dτ + G 1 (t),
dt 3 3
t− 10
  t 
ds2 1 −(t−τ )
= s2 (t) − + s1 (t) + e s1 (τ )dτ + G 1 (t), (3.11)
dt 2 t− 103
An Effective Technique for Solving a Model Describing Biological … 47

(a) s1,2 (t) with exact (b) s2,2 (t) with exact

Fig. 14 Example 3.3 for n = 1 (Type-I), graph of exact and approximate solutions

where
  
1 1 1 1 1 1 3
G 1 (t) = cos t − sin t + sin t − cos t + cos t − ,
4 4 3 2 4 4 10
     
1 1 1 3 1 1 3 3 3
G 2 (t) = − cos t + sin t − + sin t − cos t + e− 10 cos t − − sin t − ,
4 4 2 8 8 8 10 10

with initial data s1 (0) = 0 and s2 (0) = 0. The exact solution of this system is s1 (t) =
1
4
sin t and s2 (t) = − 41 sin t.
The proposed methodology type-I formula for (3.11) is
t 
ds1,n 1
s1,n+1 (t) =αn s1,0 + (1 − αn )s1,n (t) − (1 − αn ) − s1,n (u) − 2s2 (u)
du 3
 u   
0

− s2,n (τ )dτ − G 1 (u) du , n ≥ 0, (3.12)

u−3/10

 t
ds2,n 1
s2,n+1 (t) =αn s2,0 + (1 − αn )s2,n (t) − (1 − αn ) − s2,n (u) − + s1,n (u)
0 du 2
 u   
− e −(u−τ ) s1,n (τ )dτ − G 2 (u) du , n ≥ 0. (3.13)
u−3/10

By selecting the initial approximation s1,0 (0) = 0 and s2,0 (0) = 0, we obtain the
results and the obtained approximate solutions along with the exact solutions are
represented in Figs. 14, 15, and 16. From figures, it is clear that as the obtained
solutions via the proposed methodology type-I is really well in line with the precise
solutions as iterations progress. However, this approach requires more iterations to
achieve a good agreement with the exact solutions.
48 S. Tomar and S. Dhama

(a) s1,3 (t) with exact (b) s2,3 (t) with exact

Fig. 15 Example 3.3 for n = 2 (Type-I), graph of exact and approximate solutions

(a) s1,4 (t) with exact (b) s2,4 (t) with exact

Fig. 16 Example 3.3 for n = 3 (Type-I), graph of exact and approximate solutions

Now, let us implement the proposed methodology type-II formula for (3.11) which
is given by
 t
ds1,n 1
s1,n+1 (t) =αn s1,0 + (1 − αn )s1,n (t) − (1 − αn ) e−1/3(u−t) − s1,n (u) − 2s2 (u)
0 du 3
 u   
− s2,n (τ )dτ − G 1 (u) du , n ≥ 0, (3.14)
u−3/10

 t
ds2,n 1
s2,n+1 (t) =αn s2,0 + (1 − αn )s2,n (t) − (1 − αn ) e1/2(u−t) − s2,n (u) − + s1,n (u)
0 du 2
 u   
− e−(u−τ ) s1,n (τ )dτ − G 2 (u) du , n ≥ 0. (3.15)
u−3/10

By selecting the initial approximation s1,0 (0) = 0 and s2,0 (0) = 0, we obtain the
results and the obtained approximate solutions along with the exact solutions are
An Effective Technique for Solving a Model Describing Biological … 49

(a) s1,3 (t) with exact (b) s2,3 (t) with exact

Fig. 17 Example 3.3 for n = 1 (Type-II), graph of exact and approximate solutions

(a) s1,4 (t) with exact (b) s2,4 (t) with exact

Fig. 18 Example 3.3 for n = 2 (Type-II), graph of exact and approximate solutions

represented in Figs. 17, and 18. From figures, it is clear that as the obtained solutions
via the proposed methodology type-II is really well in line with the precise solutions
as iterations progress. As it is evident from figures that methodology type-II provides
better results compare to methodology type-I with less number of iterations. Even the
obtained solutions via type-II using second iteration are in good agreement with true
solutions. The dynamics of the populations of two distinct species are depicted in
Fig. 18, which demonstrates that as time increases, the population of the first species
decays gradually, whereas the population of the second species grows gradually
throughout the domain.
50 S. Tomar and S. Dhama

4 Conclusion

In this study, a practical methodology for solving a system of two nonlinear

integro-differential equations is proposed. The considered nonlinear delay integro-
differential equations are associated with the dynamics of two interacting species.
The methodology consists of two types of formulas, which are based on the different
integral operators. The suggested methodology gives consistent approximate solu-
tions to the problems. Some numerical test problems have been carried out to display
the efficiency of the approach. It is clear from the computational simulations that as
the iterations progress, it can be observed that the obtained solutions approach the
exact solution. The dynamics of the populations of two distinct species are depicted
in Figs. 7, 13 and 18 which demonstrate that as time increases, the population of dis-
tinct species decays or grows. It can be observed that the type-I approach is simple
but requires more iterations to produce a good accurate solution. From the results, it
is clear that type-II approach provides better results compared to type-I in less num-
ber of iterations. Also, the figures demonstrate that the proposed methodology is an
effective tool to solve nonlinear delay integro-differential equations. In addition, this
method can also be extended to other physical models in the future.

References

1. A.-M. Wazwaz, Linear and Nonlinear Integral Equations, vol. 639 (Springer, 2011)
2. H. Singh, H. Dutta, M.M. Cavalcanti, Topics in Integral and Integro-Differential Equations
(Springer, 2021)
3. R.K. Pandey, H. Singh, An efficient numerical algorithm to solve Volterra integral equation of
second kind (2021), pp. 215–228
4. H. Wang, H. Fu, H. Zhang, Z. Hu, A practical thermodynamic method to calculate the best
glass-forming composition for bulk metallic glasses. Int. J. Nonlinear Sci. Numer. Simul. 8(2),
171–178 (2007)
5. L. Xu, J.-H. He, Y. Liu, Electrospun nanoporous spheres with Chinese drug. Int. J. Nonlinear
Sci. Numer. Simul. 8(2), 199–202 (2007)
6. F. Sun, M. Gao, S. Lei, Y. Zhao, K. Wang, Y. Shi, N. Wang, The fractal dimension of the fractal
model of dropwise condensation and its experimental study. Int. J. Nonlinear Sci. Numer.
Simul. 8(2), 211–222 (2007)
7. T.-L. Bo, L. Xie, X.J. Zheng, Numerical approach to wind ripple in desert. Int. J. Nonlinear
Sci. Numer. Simul. 8(2), 223–228 (2007)
8. F. Shakeri, M. Dehghan, Solution of a model describing biological species living together using
the variational iteration method. Math. Comput. Model. 48(5–6), 685–699 (2008)
9. J. Biazar, H. Ghazvini, M. Eslami, He’s homotopy perturbation method for systems of integro-
differential equations. Chaos, Solitons & Fractals 39(3), 1253–1258 (2009)
10. M. Saeidy, M. Matinfar, J. Vahidi, Analytical solution of BVPs for fourth-order integro-
differential equations by using homotopy analysis method. Int. J. Nonlinear Sci. 9(4), 414–421
(2010)
11. J. Biazar, E. Babolian, R. Islam, Solution of the system of ordinary differential equations by a
domain decomposition method. Appl. Math. Comput. 147(3), 713–719 (2004)
12. R. Sekar, K. Murugesan, Single term Walsh series method for the system of nonlinear delay
Volterra integro-differential equations describing biological species living together. Int. J. Appl.
Comput. Math. 4(1), 1–13 (2018)
An Effective Technique for Solving a Model Describing Biological … 51

13. M. Marin, Generalized solutions in elasticity of micropolar bodies with voids, Revista de la
Academia Canaria de Ciencias:= Folia Canariensis Academiae Scientiarum 8(1), 101–106
(1996)
14. M. Marin, A temporally evolutionary equation in elasticity of micropolar bodies with voids.
Bull. Ser. Appl. Math. Phys 60(3)
15. M. Marin, A partition of energy in thermoelasticity of microstretch bodies. Nonlinear Anal.
Real World Appl. 11(4), 2436–2447 (2010)
16. Ş Yüzbaşı, M. Sezer, An exponential approach for the system of nonlinear delay integro-
differential equations describing biological species living together. Neural Comput. Appl.
27(3), 769–779 (2016)
17. J. Saberi-Nadjafi, M. Tamamgar, The variational iteration method: a highly promising method
for solving the system of integro-differential equations. Comput. Math. Appl. 56(2), 346–351
(2008)
18. N.H. Sweilam, Fourth order integro-differential equations using variational iteration method.
Comput. Math. Appl. 54(7–8), 1086–1091 (2007)
19. S.-Q. Wang, J.-H. He, Variational iteration method for solving integro-differential equations.
Phys. Lett. A 367(3), 188–191 (2007)
20. K. Maleknejad, M.T. Kajani, Solving linear integro-differential equation system by Galerkin
methods with hybrid functions. Appl. Math. Comput. 159(3), 603–612 (2004)
21. K. Maleknejad, B. Basirat, E. Hashemizadeh, A Bernstein operational matrix approach for
solving a system of high order linear Volterra-Fredholm integro-differential equations. Math.
Comput. Model. 55(3–4), 1363–1372 (2012)
22. J. Pour-Mahmoud, M.Y. Rahimi-Ardabili, S. Shahmorad, Numerical solution of the system
of Fredholm integro-differential equations by the tau method. Appl. Math. Comput. 168(1),
465–478 (2005)
23. A. Arikoglu, I. Ozkol, Solutions of integral and integro-differential equation systems by using
differential transform method. Comput. Math. Appl. 56(9), 2411–2417 (2008)
24. H. Singh, H.M. Srivastava, D. Baleanu, Methods of Mathematical Modelling: Infectious Dis-
eases (Academic, 2022)
25. H. Singh, D. Baleanu, J. Singh, H. Dutta, Computational study of fractional order smoking
model. Chaos, Solitons & Fractals 142, 110440 (2021)
26. H. Singh, Analysis of drug treatment of the fractional HIV infection model of cd4+ t-cells.
Chaos, Solitons & Fractals 146, 110868 (2021)
27. H. Singh, Analysis for fractional dynamics of Ebola virus model. Chaos, Solitons & Fractals
138, 109992 (2020)
28. S. Kumar, D. Zeidan, An efficient Mittag-Leffler kernel approach for time-fractional advection-
reaction-diffusion equation. Appl. Numer. Math. 170, 190–207 (2021)
29. S. Kumar, J.J. Nieto, B. Ahmad, Chebyshev spectral method for solving fuzzy fractional
Fredholm-Volterra integro-differential equation. Math. Comput. Simul. 192, 501–513 (2022)
30. M. Dehghan, M. Shahini, Rational pseudospectral approximation to the solution of a nonlinear
integro-differential equation arising in modeling of the population growth. Appl. Math. Model.
39(18), 5521–5530 (2015)
31. F. Fakhar-Izadi, M. Dehghan, An efficient pseudo-spectral Legendre-Galerkin method for solv-
ing a nonlinear partial integro-differential equation arising in population dynamics. Math.
Methods Appl. Sci. 36(12), 1485–1511 (2013)
32. A. Jerri, Introduction to Integral Equations with Applications (Wiley, 1999)
33. Y. Khan, H. Vazquez-Leal, Q. Wu, An efficient iterated method for mathematical biology
model. Neural Comput. Appl. 23(3), 677–682 (2013)
34. E. Babolian, J. Biazar, Solving the problem of biological species living together by a domain
decomposition method. Appl. Math. Comput. 129(2–3), 339–343 (2002)
35. S. Yousefi, Numerical solution of a model describing biological species living together by using
Legendre multiwavelet method. Int. J. Nonlinear Sci. 11(1), 109–113 (2011)
36. P.K. Sahu, S.S. Ray, Legendre spectral collocation method for the solution of the model describ-
ing biological species living together. J. Comput. Appl. Math. 296, 47–55 (2016)
52 S. Tomar and S. Dhama

37. M. Shakourifar, M. Dehghan, On the numerical solution of nonlinear systems of Volterra

integro-differential equations with delay arguments. Computing 82(4), 241–260 (2008)
38. A. Tari, The differential transform method for solving the model describing biological species
living together
39. M. Ramezani, H. Jafari, S.J. Johnston, D. Baleanu, Complex b-spline collocation method for
solving weakly singular Volterra integral equations of the second kind. Miskolc Math. Notes
16(2), 1091–1103 (2015)
40. R. Hafez, E. Doha, A. Bhrawy, D. Baleanu, Numerical solutions of two-dimensional mixed
Volterra-Fredholm integral equations via Bernoulli collocation method. Rom. J. Phys. 62(111),
1–11 (2017)
41. S. Thirumalai, R. Seshadri, S. Yuzbasi, Population dynamics between a prey and a predator
using spectral collocation method. Int. J. Biomath. 12(05), 1950049 (2019)
42. B. Halpern, Fixed points of nonexpanding maps. Bull. Am. Math. Soc. 73(6), 957–961 (1967)
43. S. He, T. Wu, Y.J. Cho, T.M. Rassias, Optimal parameter selections for a general Halpern
iteration. Numerical Algorithms 82(4), 1171–1188 (2019)
44. L. Leuştean, A. Nicolae, Effective results on compositions of nonexpansive mappings. J. Math.
Anal. Appl. 410(2), 902–907 (2014)
45. S. Tomar, An effective approach for solving a class of nonlinear singular boundary value
problems arising in different physical phenomena. Int. J. Comput. Math. 98(10), 2060–2077
(2021)
46. S. Tomar, A computationally efficient iterative scheme for solving fourth-order boundary value
problems. Int. J. Appl. Comput. Math. 6(4), 1–16 (2020)
47. S. Tomar, A rapid-converging analytical iterative scheme for solving singular initial value
problems of Lane-Emden type. Int. J. Appl. Comput. Math. 7(3), 1–17 (2021)
48. K. Muangchoo-in, K. Sitthithakerngkiet, P. Sa-Ngiamsunthorn, P. Kumam, Approximation
theorems of a solution of amperometric enzymatic reactions based on green’s fixed point
normal-s iteration. Adv. Differ. Equ. 2021(1), 1–13 (2021)
49. S.A. Khuri, A. Sayfy, Numerical solution of functional differential equations: a green’s
function-based iterative approach. Int. J. Comput. Math. 95(10), 1937–1949 (2018)
50. S.A. Khuri, A. Sayfy, A fixed point iteration method using green’s functions for the solution of
nonlinear boundary value problems over semi-infinite intervals. Int. J. Comput. Math. 97(6),
1303–1319 (2020)
51. R.K. Pandey, S. Tomar, An efficient analytical iterative technique for solving nonlinear differ-
ential equations. Comput. Appl. Math. 40(5), 1–16 (2021)
52. S. Tomar, A. Verma, K. Vajravelu, An effective method for solving singular boundary value
problems with some relevant physical applications. Comput. Appl. Math. 41(1), 1–20 (2022)
Numerical Solution for a Tumor Cells
Dynamics Within Their
Micro-environment

Kolade M. Owolabi, Albert Shikongo, and Edson Pindza

Abstract Many processes depend on recent history and recent history is mainly
modeled by delay differential equations. This implies that delay differential equa-
tions introduce more realistic dynamics in modeling processes, due to their novel
technique of capturing past and present processes simultaneously. Therefore, incor-
porating a delay term to model key transformations taking place within the dynam-
ics modeling the interaction among transformed epithelial cells (TECs), fibroblasts,
myofibroblasts, transformed growth factor (TGF−β), and epithelial growth factor
(EGF), in silico, in a setup mimicking experiments in a tumor chamber invasion assay
are thus, considered in this chapter. This consideration of such experiments resulted
into a derivation of system of nonlinear quasi time-depended delay parabolic partial
differential equations. The well-pose of the unique solution of the resulting sys-
tem of nonlinear quasi time-depended delay parabolic partial differential equations,
presents that an extension of the existing Gronwall’s inequality for ordinary differ-
ential equations to the Gronwall’s inequality for linear, delay ordinary differential
equations is thus, derived. The asymptotic stability and Hopf bifurcation analysis are
established. Based on the asymptotic stability condition, a novel numerical method
is derived, analyzed, implemented and improved results are presented for discussion.
Hence, based on the obtained results, one strongly believes that the contribution in
this chapter has a great deal toward the process of crafting a drug that can slow and/or
confine tumor invasion.

K. M. Owolabi (B)
Department of Mathematical Sciences, Federal University of Technology Akure, PMB 704
Akure, Ondo State, Nigeria
e-mail: [email protected]
A. Shikongo
Engineering Mathematics, School of Engineering and the Built Environment, JEDS Campus,
University of Namibia, Private Bag 13301, Windhoek, Namibia
E. Pindza
Department of Mathematics and Applied Mathematics, University of Pretoria, Pretoria 002,
South Africa

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 53
H. Singh and H. Dutta (eds.), Computational Methods for Biological Models,
Studies in Computational Intelligence 1109,
https://doi.org/10.1007/978-981-99-5001-0_3
54 K. M. Owolabi et al.

Keywords Tumor cells’ micro-environment · Time dependent quasi delay

parabolic partial differential equations · Extended Gronwall’s inequality · Hopf
bifurcation · Numerical solution

1 Introduction

Following the fact that tumor micro-environment remained consistent and pertinent in
a host due to active epithelial cells, fibroblasts, myofibroblasts, endothelial cells, and
inflammatory cells [31], then the presence of such cells within the micro-environment
implies that the surrounding extra-cellular matrix (ECM) plays a significant role in
confining and/or confronting cancer within its micro-environment. This seems to be a
fact because in [31, 32] it is deduced Matrix Metalloproteinase (MMP), human Ductal
Carcinoma in Situ (DCIS) and Matrix Metalloproteinase (MMP) can be modulated
as cell adhesion or blocking material. Consequently, presenting that several classes
of MMPs can be expressed in the form of periductal fibroblasts and myofibroblasts.
This presents an intense stromal involvement during early invasion. Therefore, it is
vital that the dynamics corresponding to tumor cells degrading the basal membrane
and invade into the stroma be considered, eventhough such invasion of transformed
epithelial cells (TECs) into stroma is a complex step toward metastasis [2].
Therefore, as in [41] Dn , D f , Dm , D E , DG , D P are defined to denote constant
diffusion coefficients for the density of transformed epithelial cells, density of fibrob-
lasts, density of myofibroblasts, concentration of epidermal growth factor, concen-
tration of transformed growth factor, and concentration of matrix metalloproteinase.
Thus, from the dynamical system point of view, the process of simplifying the com-
plex steps toward metastasis has led Kim and Friedman in [31] to derive the dynamics
of confining and/or confronting cancer within its micro-environment as

⎫
∂n = ∇ · (D ∇n) − ∇ · χ n  ∇E ∇ρ ⎪
− ∇χ1n Is n  ⎪
1+(|∇ρ|/λρ )2 ⎪
∂t n n
1+(|∇ E|/λ E )2 ⎪
⎪ ⎪
⎪
⎪
⎪
  ⎪
⎪
E4 ⎪
⎪
+a11 4 n 1 − n −an ρI , ⎪
⎪
k +E 4 ∗ 12 s ⎪
⎪
E ⎪
⎪
⎪
⎪
⎪
⎪
∂f
= ∇ · (D ∇ ) − + ⎪
⎪
∂t f f a 12 G f a 22 f, ⎪
⎪
⎪
⎪
  ⎪
⎪
⎪
⎪
∂m = ∇ · (D ∇m) − ∇ · χ m  ∇G ⎪
⎬
∂t m m + a 21 G f + a 31 m,
1+(|∇G|/λG )2 (1)
⎪
⎪
⎪
⎪
  ⎪
⎪
∂ρ ρ ⎪
⎪
= −a Pn + (a f + a m) 1 − ρ∗ , ⎪
⎪
∂t 41 42 43 ⎪
⎪
⎪
⎪
⎪
⎪
∂ E = ∇ · (D ∇ E) + I ⎪
⎪
∂t E − (a51 f + a52 m) − a53 E, ⎪
⎪
⎪
⎪
⎪
⎪
∂G = ∇ · (D ∇G) + a I
⎪
⎪
 n − a G, ⎪
⎪
∂t G 61 + 62 ⎪
⎪
⎪
⎪
⎪
⎪
∂ P = ∇ · (D ∇ P) + a I m − a P, ⎭
∂t p 71 − 72
Numerical Solution for a Tumor Cells Dynamics Within Their Micro-environment 55

where, ∀t > 0, xi ∈ [−L i , L i ], i := 1, 2, 3, L ∈ Z such that

 = {x = (x1 , x2 , x3 )}, + =  ∩ {x1 > 0}, − =  ∩ {x1 < 0}, ∗ = + ∪ − ,

S = {−L 0 < x1 < L 0 , such that x1  = 0, −L i < xi < L i , ∀ i = 2, 3}.

When 0 < L 0 < L 1 , then Kim and Friedman in [31] set the characteristic function of
a set A ∈ R to be denoted by I A . This implies that I A (x) = 1, if x ∈ A, A(x) = 0, if
x∈ / A. Furthermore, Kim and Friedman in [31] defined n(x, t) ∈ + to denotes the
density of transformed epithelial cells (TECs); f (x, t) ∈ − to denotes the density of
fibroblasts (f); m(x, t) ∈ − to denotes the density of myofibroblasts (m); ρ(x, t) ∈ S
to denotes the concentration of extracellular matrix (ECM); E(x, t) ∈ ∗ to denotes
the concentration of epidermal growth factor (EGF); G(x, t) ∈ ∗ to denotes the
concentration of transformed growth factor (TGF-β) and P(x, t) ∈ ∗ to denotes
the concentration of matrix metalloproteinase (MMP).
One can see that the experiment for the dynamics reported in Eq. (2) is carried
out in such a way that the dependent variables are separated by semi-permeable
membrane, which is defined for x1 = 0, ∀ i = 2, 3. But, the transmission conditions
at the semi-permeable membrane are given in [31] as

∂ E+ ∂ E− + ⎫
∂x 1 = ∂x 1 , − ∂∂xE 1 + γ(E + − E − ) = 0 on M, t > 0, ⎪
⎪
⎪
⎪
⎪
⎬
∂G + ∂G − +
∂x 1 = ∂x 1 , − ∂G + −
∂x 1 + γ(G − G ) = 0 on M, t > 0, (2)
⎪
⎪
⎪
⎪
∂ P+ ∂ P− + ⎪
⎭
∂x 1 = ∂x 1 , − ∂∂xP 1 + γ(P + − P − ) = 0 on M, t > 0,

such that
⎧
⎪ +
⎨ E (x, t), if x1 > 0,
E(x, t) =
⎪
⎩ E − (x, t) if x < 0,
1
⎧
⎪ +
⎨G (x, t), if x1 > 0,
G(x, t) =
⎪
⎩G − (x, t), if x < 0,
1
⎧
⎪ +
⎨ P (x, t), if x1 > 0,
P(x, t) =
⎪
⎩ P − (x, t), if x < 0,
1

where, γ denotes a positive parameter [17]. Since the dynamics reported in Eqs. (1)–
(2) constitute a system of nonlinear time-dependent parabolic equations, then the
boundary and initial conditions are required. But the γ presents that the membrane
is not permeable to cells [31]. Hence, it necessitated the no flux boundary conditions
to be imposed as
56 K. M. Owolabi et al.
  ⎫
∇E ∇ρ ⎪
⎪
Dn ∇n − χn n  − χ1n I S n  · ν = 0, on ∂+ , t > 0, ⎪
⎪
1+(|∇ E|/λ E )2 1+(|∇ρ|/λρ )2 ⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
D f ∇ f · ν = 0, on ∂− , t > 0, ⎬
  (3)
⎪
⎪
⎪
⎪
Dm ∇m − χm m  ∇G · ν = 0, on ∂− , t > 0, Dρ ∇ρ · ν = 0, ⎪
⎪
⎪
⎪
1+(|∇G|/λG )2 ⎪
⎪
⎪
⎪
⎪
⎪
⎭
D E ∇ E · ν = 0, DG ∇G · ν = 0, D P ∇ P · ν = 0, on ∂ − {x1 = 0, i = 2, 3}, t > 0,

in which ν denotes the outward normal vector. The initial conditions are
   ⎫
n 0 (x) = 21 1 + tanh − 1 (0.8 − x) ∈ + , ⎪
⎪
⎪
⎪
   ⎪
⎪
⎪
⎪
f 0 (x) = 0, 143 2 1 + tanh − (x − 0.2 ∈ − ,
1 1
⎬
(4)
⎪
m 0 (x) = 0.0 ∈ − , ρ0 (x) = 1.0 ∈ S, E 0 (x) = 1.0 ∈ ∗ , ⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎭
G 0 (x) = 1.0 ∈ ∗ , P0 (x) = 0.0 ∈ ∗ .

Due to the complexity of the dynamics within the micro-environment [31, 32], we
strongly recommend that the incorporation of time (τ ∈ N) required for some trans-
formations and/or productions to take place should be incorporated into the dynamics
in Eq. (1). The incorporation of time (τ ) is the recommended to be in corporated into
the following.
• transformation of fibroblasts into myofibroblasts.
• degradation of the extra-cellular matrix (ECM).
• production of fibroblasts and myofibroblasts by the transformed epithelial cells
(TEC).
• production of transformed epithelial cells by the transformed growth factor
(TGF−β).
• the production of fibroblasts by the Matrix Metalloproteinase (MMP).
Such incorporation of time (τ ) implies that the spatial domain for ρ should be
increased from S to the entire domain ∗ . Thus ignoring the vertical variables,
then the incorporation of time (τ ) implies that the dynamics in Eq. (1) becomes
⎫
∂n
− Dn n = −∇ · χn n √ ∇E
− ∇χ1n Is n √ ∇ρ ⎪
⎪
∂t

1+(|∇ E|/λ E )2  1+(|∇ρ|/λρ )2 ⎪
⎪
⎪
⎪
nE4
+ ka411+E 1 − n
∈ [0, L ], ⎪
⎪
4 n ∗ −a12 ρIs 1 ⎪
⎪
∂f
E ⎪
⎪
− D f  f = (−a12 G(x, ⎪
∂t  t − τ ) + a22 ) f (x,
 t − τ ) ∈ [−L 1 , 0), ⎪
⎪
⎪
⎪
⎪
∂m
− Dm m = −∇ · χm m √ ∇G ⎬
∂t 1+(|∇G|/λG )2 (5)
+a21 G(x, t − τ ) f (x, t − τ ) + a31 m ∈ [−L1 , 0),  ⎪
⎪
⎪
⎪
∂ρ ρ ⎪
⎪
∂t − Dρ ρ = −a41 P(x, t − τ )n(x, t − τ ) + (a42 f + a43 m) 1 − ρ∗ ∈ [−L 1 , L 1 ], ⎪ ⎪
⎪
∂E ⎪
⎪
− D E = I  (a f (x, t − τ ) + a m(x, t − τ )) − a E ∈ [−L , L ], ⎪
⎪
∂t
∂G
E − 51 52 53 1 1
⎪
⎪
− D G = a I
61 + n(x, t − τ ) − a G ∈ [−L , L ], ⎪
⎪
∂t
∂P
G 62 1 1 ⎪
⎭
∂t − D p P = a 71 I − m(x, t − τ ) − a 72 P ∈ [−L 1 , L 1 ],
Numerical Solution for a Tumor Cells Dynamics Within Their Micro-environment 57

for (x, t) ∈ [−L 1 , L 1 ] × [−τ , 0]. The dynamics in Eq. (5) is a system of discrete
delay reaction-diffusion equations. Delay differential equations (DDEs) are widely
used for analysis and predictions in various areas of life sciences see [41] and the
references therein. Hence, the first aim in this chapter, is to carry out mathematical
analysis, which leads to the investigation of how time delay τ affects the dynamics
in Eq. (5). By applying the Poincaré normal form and the center manifold theorem
as in [22], one finds conditions for the functions and derives formulas which deter-
mine the properties of Hopf bifurcation. More specifically, the chapter presents that
equilibrium point losses its stability and the dynamics exhibit Hopf bifurcation under
certain conditions.
The second aim is to present the improved numerical solution as compared to [41]
and for more numerical method related literature, we refer our readers to [41] and
the references therein.
The rest of the chapter is structured as follow. Mathematical analysis of the
extended model is presented in Sect. 2. A robust numerical scheme based on the fitted
finite difference technique is formulated, implemented and analysed for convergence
in Sect. 3. To justify the effectiveness of the novel numerical method numerical results
are presented in Sects. 4 and 5 concludes the chapter.

2 Mathematical Analysis

Let u = [n, f, m, ρ, E, G, P]T , D = [Dn , D f , Dm , Dρ , D E , DG , D P ]T , where T

denote a transpose. Then the extended dynamics in Eq. (5) can be rewritten as

∂u
− Du = F(u(x, t), ∇u(x, t), u(x, t − τ )), for t ∈ [t0 − τ , t0 ], (6)
∂t
where,
∇E ∇ρ ⎫
F1 (u(x, t), ∇u(x, t), u(x, t − τ )) = −∇ · χn n  − ∇χ1n Is n  ⎪
⎪
1+(|∇ E|/λ E )2 1+(|∇ρ|/λρ )2 ⎪
⎪
  ⎪
⎪
a11 n E 4 ⎪
⎪
+ 1 − n −an ρI ∈ [0, L 1 ], ⎪
⎪
k 4E +E 4 ∗ s ⎪
⎪
12
⎪
⎪
F2 (u(x, t), ∇u(x, t), u(x, t − τ )) = (−a12G(x, t − τ ) + a22 ) f (x, ⎪
 t − τ ) ∈ [−L 1 , 0), ⎪
⎪
⎪
⎪
⎪
⎪
F3 (u(x, t), ∇u(x, t), u(x, t − τ )) = −∇ · χm m  ∇G ⎪
⎪
⎪
⎬
1+(|∇G|/λG )2
+a21 G(x, t − τ ) f (x, t − τ ) + a31 m ∈ [L 1 , 0), ⎪
⎪
⎪
⎪
F4 (u(x, t), ∇u(x, t), u(x, t − τ )) = −a
 41 P(x, t − τ )n(x, t − τ ) ⎪
⎪
⎪
⎪
+(a42 f + a43 m) 1 − ρρ∗ ∈ [−L 1 , L 1 ], ⎪
⎪
⎪
⎪
⎪
F5 (u(x, t), ∇u(x, t), u(x, t − τ )) = I− (a51 f (x, t − τ ) ⎪
⎪
⎪
⎪
+a52 m(x, t − τ )) − a53 E ∈ [−L 1 , L 1 ], ⎪
⎪
⎪
⎪
F6 (u(x, t), ∇u(x, t), u(x, t − τ )) = a61 I+ n(x, t − τ ) − a62 G ∈ [−L 1 , L 1 ], ⎪
⎪
⎪
⎭
F7 (u(x, t), ∇u(x, t), u(x, t − τ )) = a71 I− m(x, t − τ ) − a72 P ∈ [−L 1 , L 1 ],

with all other terminal conditions remained unchanged, as given in (2) through to (4).
58 K. M. Owolabi et al.

2.1 Solution Continuously Depending on the Data

Let v(x, t, t − τ ), z(x, t, t − τ ) ∈ C21 [−L 1 , L 1 ] denote two solutions for the dynam-
ics in Eq. (5), such that v(x, t, t − τ ) − z(x, t, t − τ ) =: u(x, t, t − τ ), (where −τ <
t < 0), yields the following results.
Theorem 2.1 Let
∂u(x,t,t−ø)
⎫
∂t −Du(x, t,t − τ ) = 0; x∈ [−L 1 , L 1 ], t > 0, ⎪
⎪
⎪
⎪
1 1
u 1 (x, 0) = 2 1 + tanh − (0.8 − x) ∈ [0, L 1 ], ⎪
⎪
   ⎬
1 1
u 2 (x, 0) = 0, 143 2 1 + tanh − (x − 0.2 ∈ [−L 1 , 0), u 3 (x, 0) = 0.0 ∈ [L 1 , 0), ⎪
⎪ ⎪
ρ0 (x) = 1.0 ∈ [−L 1 , L 1 ], u 4 (x, 0) = 1.0 ∈ [−L 1 , L 1 ], u 5 (x, 0) = 1.0 ∈ [−L 1 , 0), ⎪
⎪
⎪
⎭
u 6 (x, 0)(x) = 0.0 ∈ [−L 1 , L 1 ].

Then u(x, t, t − τ ) is identically zero.

Proof Proceeding component-wise for elements of the vector u(x, t), one obtains,
⎫
u(·, t, t − τ )ut (·, t, t − τ ) = 21 ∂t u2 (·, t, t − τ ), ⎬
⎭
u(·, t, t − τ )ut−τ (·, t, t − τ ) = 21 ∂t−τ u2 (·, t, t − τ ).

Similarly, one also finds

u(·, t, t − τ )ux x (·, t, t − τ ) = ∂x (u(·, t, t − τ )ux (·, t, t − τ )) − u2x (·, t, t − τ ),

then in view of Theorem 2.1

 L  L
1 d 1 1 1
u2 (·, t, t − τ )d x = ∂t u2 (·, t, t − τ )d x
2 dt −L 1 −L 1 2
 L  L
1 1
= u(·, t, t − τ )ut (·, t, t − τ )d x = D u(·, t)ux x (·, t, t − τ )d x,
−L 1 −L 1
 L  L
1 1
=D ∂x (u(·, t, t − τ )ux (·, t, t − τ ))d x − u2x (·, t, t − τ )d x,
−L 1 −L 1
 L  L
L 1 1
= Du(·, t, t − τ )ux (·, t, t − τ )|−L
1 − u2x (·, t, t − τ )d x = −D u2x (·, t, t − τ )d x ≤ 0.
1 −L 1 −L 1

L
These implies that the function t → −L1 1 u2 (·, t, t − τ )d x is a non-increasing func-
tion. Hence
 L1  L1
0≤ u2 (·, t, t − τ )d x ≤ u2 (·, 0, 0)d x = 0, ∀x, t.
−L 1 −L 1

A similar results can be established that,

Numerical Solution for a Tumor Cells Dynamics Within Their Micro-environment 59
 L1  L1
1 d
u (·, t, t − τ ) = −D
2
u2x (·, t, t − τ )d x ≤ 0.
2 d(t − τ ) −L 1 −L 1

This proves uniqueness of solution to the nonlinear quasi time-depended delay

parabolic partial differential equations (DPPDEs) in Eq. (6).
Corollary 2.2 Let v, z ∈ C21 [−L 1 , L 1 ] denote solutions to the dynamics in Eq. (6)
with initial states v0 , z0 , such that v0 − z0 =: u0 and Fu , Fv denote the real-valued
functions bounding v and z, respectively. Then
 t
v−z 2
≤ exp(−ηt) v0 − z0 2
+ exp(−η(t − s)) Fu − Fv 2 ds,
0

and
 t
v − z 2 ≤ exp(−η(t − τ )) v0 − z0 2 + τ exp(−η(t − τ − s)) Fu − Fv 2 ds,
t−τ

for some −τ < t < 0, and , τ,η ∈ R+ .

In order to prove the Corollary 2.2, the following are preliminaries to the proof.
Definition 2.3 Let v, z ∈ C21 [−L 1 , L 1 ] then
 L1
(v, z) := v(·, t), z(·, t)d x,
−L 1

defines an in inner product [14].

Lemma 2.4 If v, z ∈ C21 [−L 1 , L 1 ] then

(vx x + σv, z) = σ(v, z) − (vx , zx ) = (v, zx x + σz),

where, σ ∈ R.

Proof Since

∂x [vx z] = vx x z + vx zx ,

then
L1
(∂x [vx , z]) = (vx x , z) + (vx , zx ), ⇒ (vx , z)|−L 1
= (vx x , z) + (vx , zx ),
⇒ (vx x , z) = −(vx , zx ). (7)

Similarly,

∂x [vzx ] = vzx x + vx zx ,
60 K. M. Owolabi et al.

then it follows from equation in (7) that

(v, zx x ) = −(vx , zx ). (8)

Combining equation in (7) with equation in (8) yields

(vx x , z) = (v, zx x ).

Hence, the results follows. 

Lemma 2.5 Let v ∈ C21 [−L 1 , L 1 ], then

 L1  21
1
v := (v, v) = 2 v (·, t)d x
2
,
−L 1

defines a norm.

Proof See [14].

Corollary 2.6 Let v, z ∈ C21 [−L 1 , L 1 ] then

|(v, v)| ≤ v v .

Proof Let [14]

P2 (σ) = v + σz 2
= v 2
+ 2σ(v, z) + σ 2 z 2 ,

denotes a polynomial of degree two. Then P2 (σ) ≥ 0, ∀σ ∈ R. Thus,

 P2 (σ) = 4(v, z)2 − 4 v 2

z 2
≤ 0, ⇒ (v, z)2 ≤ v 2
z 2,

which concludes the proof of Corollary 2.6.

Proof of Corollary 2.2: By means of Poincaré inequality [14], Schwarz inequality

in Lemma 2.5, Corollary 2.6 and in view of the prove to Theorem 2.1,

1 d 1 d 2
u(·, t, t − τ ) 2 + u(·, t, t − τ ) 2 + D u(·, t) 2
2 dt 2 d(t − τ ) c
≤ 2(F(·, t, t − τ ), u(·, t, t − τ )) ≤ 2 F u(·, t, t − τ ) ,

where, c > 0, F := Fu − Fv . Thus,

2  ⎫
d
dt
u(·, t, t − τ ) 2
+ c
D− u(·, t, t − τ ) 2
≤ 1
F 2
, ⎬
2  ⎭
d
d(t−τ )
u(·, t, t − τ ) 2
+ c
D− u(·, t, t − τ ) 2
≤ 1
F 2
,
Numerical Solution for a Tumor Cells Dynamics Within Their Micro-environment 61

for some arbitrary sufficiently small > 0. Let η = 2c D −  > 0 and  = 1/. Then,
Eq. (9) becomes
⎫
d
dt
u(·, t, t − τ ) 2
+ η u(·, t, t − τ ) 2
≤ F 2
, ⎬
(9)
⎭
d
d(t−τ )
u(·, t, t − τ ) 2
+ η τ u(·, t, t − τ ) 2
≤ τ F 2
.

Applying the Gronwall’s inequality [14] to equation in (9), one finds the following
results
Theorem 2.7 (Extended Gronwall’s inequality) Let I ≡ [−L 1 , L 1 ]. Let  tau and
u denote real-valued continuous functions defined on I. If u is differentiable in the
interior I◦ of I and satisfies the differential inequality

d
u(·, t, t − τ ) 2
+ η τ u(·, t, t − τ ) 2
≤ τ F 2
, t − τ ∈ I◦ ,
d(t − τ )

then u is bounded by the solution of the corresponding differential equation v  (t) =

β(t)v(t):
 t 
u(·, t, t − τ ) ≤ u(·, a0 , t − τ ) exp η τ ds , ∀t ∈ I.
a0

2.2 Local Stability

In [41] we have presented the following results.

Theorem 2.8 The equilibrium point for system in Eq. (5) is given by
⎛  ⎞
⎛ ⎞ 21
⎜ a62 a22  L1 + 1 + ⎟ L 21
E =⎜ ⎝ ⎠ , 0, a22 , 0⎟ .
⎝ a61 a12 , 0, 0, L 1 1 + (L 1 ) + ln 
2

1 + L2 − L a12 ⎠
1 1

To investigate the linearized stability of the system (5), we let

(n, f, m, ρ, E, G, P) = (n̄ + n ∗ , f¯ + f ∗ , m̄ + m ∗ , ρ̄ + ρ∗ , Ē + E ∗ , Ḡ + G ∗ , P̄ + P ∗ ).

Substituting into (5), and retaining only the linear terms in n, f, m, ρ, E, G, P, one
finds
62 K. M. Owolabi et al.
  ⎫
∂ n̄ − D ∂ ∂ = 0, in ∈ [0, L 1 ], t > 0, ⎪
∂t n ∂x ∂x n̄ ⎪
⎪
⎪
⎪
⎪
⎪
  ⎪
⎪
∂ f¯ ∂ ∂ ¯ ¯ ⎪
⎪
∂t − D f ∂x ∂x f = a22 f (x, t − τ ) ∈ [−L 1 , 0], t > 0, ⎪
⎪
⎪
⎪
⎪
  ⎪
⎪
∂ m̄ − D ∂ ∂
⎪
⎪
= a31 m̄(x, t) ∈ [−L 1 , 0], t > 0, ⎪
⎪
∂t m ∂x ∂x m̄ ⎪
⎪
⎪
⎪
  ⎪
⎬
∂ ρ̄ ∂ ∂ ¯
∂t − Dρ ∂x ∂x ρ̄ = a42 f (x, t) + a43 m̄(x, t) ∈ [−S, S], t > 0, ⎪
⎪
⎪
  ⎪
⎪
⎪
⎪
∂ Ē − D ∂ ∂ Ē = −a Ē(x, t) ∈ [−L , L ], t > 0, ⎪
⎪
∂t E ∂x ∂x 53 1 1 ⎪
⎪
⎪
⎪
  ⎪
⎪
⎪
∂ Ḡ − D ∂ ∂ Ḡ = a n̄(x, t − τ ) + a Ḡ(x, t) + a62 a22 − a62 a22 ∈ [−L , L ], t > 0, ⎪
⎪
⎪
∂t G ∂x ∂x 61 62 a61 a12 a12 1 1 ⎪
⎪
⎪
⎪
⎪
⎪
  ⎪
⎪
∂ P̄ − D ∂ ∂ P̄ = a I ⎭
∂t P ∂x ∂x 71 − m̄(x, t − τ ) − a72 P̄(x, t) ∈ [−L 1 , L 1 ],

on (x, t) ∈ [−L 1¯, L 1 ] × [−τ , 0], where, the characteristic equation is

(λ − Dn )(λ − D f − a22 e−λt )(λ − Dm − a31 )(λ − Dρ )(λ − D E + a53 )

(λ − DG − a62 )(λ − D P + a72 ) = 0.

Hence, the following results.

Theorem 2.9 The dynamics in Eq. (5) are asymptotically stable if λ < Dn , λ <
D f + a22 e−λt , λ < Dm + a31 , λ < D E − a53 , λ < DG + a62 , λ < D P − a72 .

2.3 Hopf Bifurcation Analysis

√
When τ = 0, we assume that λ = iω for ω > 0 and i = −1. In view of the eigen-
values, we have

iω − D f − a22 exp(iωτ ) = iω − D f − a22 (cos(ωτ ) + i sin(ωτ )) = 0.

Separating real and imaginary parts we have

ω + a22 sin(ωτ ) = 0 and − D f − a22 cos(ωτ ) = 0, (10)

which yields
 
1 Df
τi = cos−1 + 2iπ , i = 0, 1, 2, 3, . . . .
ω0 a22

Squaring on both sides of equations in (10), we find

ω 2 + 2a22 ω sin(ωτ ) + a22

2
sin2 (ωτ ) = 0,
D 2f + 2D f a22 cos(ωτ ) + a22
2
cos2 (ωτ ) = 0. (11)
Numerical Solution for a Tumor Cells Dynamics Within Their Micro-environment 63

Adding the two equations in Eq. (11) one finds

ω 2 + D 2f + 2(D f cos(ωτ ) + ω sin(ωτ ))a22 + a22

2
= 0,

which simplifies to

ω 2 + D 2f − 2(D 2f + ω 2 ) + a22
2
= 0,

⇒ −ω 2 − D 2f + a22
2
= 0, ⇒ ω0 = ± a22 2
− D 2f .

Choosing τ0 = min{τi }, we need to show that

 
dλ 
R = 0.
dτ τ =τ0

From the eigenvalues we have

   
dλ d(D f + a22 (cos(ωτ ) + i sin(ωτ ))
R =R = −a22 ω sin(ω0 τ0 ),
dτ dτ

where ω0 τ0 = 0. By summarizing the above analysis, we arrive at the following

theorem.
Theorem 2.10 The equilibrium
⎛  ⎞
⎛ ⎞ 21
⎜ a62 a22  L1 + 1 + L 21
⎟
E =⎜ , 0, 0, L 1 + (L ) 2 + ln ⎝  ⎠ , 0, a22 , 0⎟ ,
⎝ a61 a12 1 1
a12 ⎠
1 + L2 − L 1 1

of the system (5) is asymptotically stable for τ ∈ [0, τ0 ) and it undergoes Hopf bifur-
cation at τ = τ0 .

2.4 Global Stability Analysis

Theorem 2.11 The dynamics in Eq. (5) is globally stable.

Proof The prove to this theorem has been carried out already in [41].

3 Derivation and Analysis of the Numerical Method

The derivation of the novel numerical method for discretising the dynamics (5) is as
follows. We determine an approximation to the derivatives for the functions
64 K. M. Owolabi et al.

n(t, x), f (x, t), m(x, t), ρ(x, t), E(x, t), G(x, t), P(x, t),

with respect to the spatial variable x.

Let Sx be a positive integer. Discretize the interval [−L/2, L/2] through the points

−L/2 = x0 < x1 < x2 < · · · < xs−1 < xs < xs+1 · · · < x Sx −2 < x Sx −1 < x Sx = L/2,

where, the step-size x = x j+1 − x j = 1/Sx , j = 0, 1, . . . , Sx . Let

N j (t), F j (t), M j (t), R j (t), E j (t), G j (t), P j (t), (12)

denote the numerical approximations for n(t, x), f (x, t), m(x, t), ρ(x, t), E(x, t),
G(x, t), P(x, t). Then the spatial derivatives in the system in Eq. (5) are approxi-
mated as follows
⎛ ⎞ ⎫
∂E ∂ρ ⎪
⎪
∂ ⎜ ∂n ∂x ∂x ⎟ N j+1 −2N j +N j−1 ⎪
⎪
− χ1n Is n  ⎪
∂x ⎝ Dn ∂x − χn n
 ⎠ (ti , x j ) ≈ Dn ⎪
⎪
1+( ∂ E /λ E )2 ∂ρ
/λρ )2
φ2
n ⎪
⎪
⎪
∂x 1+(
∂x ⎪
⎪
⎪
−
(Dx E j ) ⎪
⎪
− ⎪
⎪
−χn (Dx N j )   − 2 ⎪
⎪
 ⎪
⎪
1+ Dx E j ⎪
⎪
⎪
λE ⎪
⎪
⎪
⎪
⎪
Dx+ (Dx− E j ) ⎪
⎪
⎪
−χn N j ⎛  ⎪
2 ⎞3/2 , ⎪
⎪
⎪
Dx− E j ⎪
⎪
⎝1+ ⎠ ⎪
⎪
λE ⎪
⎪
⎪
⎪
⎪
⎪
(Dx− R j ) ⎪
⎪
−χn (Dx− N j ) 
  − 2
⎪
⎪
⎪
 ⎪
⎪
1+ Dx R j ⎪
⎪
⎪
⎪
λρ ⎪
⎪
⎪
⎪
Dx+ (Dx− R j ) ⎪
⎪
⎪
⎪
−χ1n Is N j ⎛  2 ⎞3/2 , ⎪
⎪
⎪
Dx− R j ⎪
⎪
⎪
⎝1+ ⎠ ⎪
⎪
λρ ⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
  F j+1 −2F j +F j−1 ⎪
⎪
⎬
∂
∂x D f ∂∂xf (ti , x j ) ≈ D f ,
φ2f
⎪
(13)
⎪
⎪
⎪
⎪
⎛ ⎞ ⎪
⎪
∂G ⎪
⎪
∂ ⎝ ∂m ∂x M j+1 −2M j +M j−1 ⎪
⎪
⎠ (ti , x j ) ≈ Dm ⎪
∂x Dm ∂x − χm m
 ⎪
1+( ∂G /λG )2 φ2 ⎪
⎪
∂x m ⎪
⎪
− ⎪
⎪
(Dx G j ) ⎪
⎪
−
−χm (Dx M j )  ⎪
⎪
  − 2 ⎪
⎪
 ⎪
⎪
1+ Dx G j ⎪
⎪
λG ⎪
⎪
⎪
⎪
⎪
⎪
Dx+ (Dx− G j ) ⎪
⎪
−χm M j ⎛  ⎪
2 ⎞3/2 , ⎪
⎪
⎪
Dx− G j ⎪
⎪
⎪
⎝1+ ⎠ ⎪
⎪
λG ⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
  R −2R j +R j−1 ⎪
⎪
∂ ∂ρ ⎪
⎪
∂x Dρ ∂x (ti , x j ) ≈ Dρ j+1 , ⎪
⎪
2 (t) ⎪
⎪
⎪
⎪
⎪
⎪
  E −2E +E j−1 ⎪
⎪
∂ D E ∂∂xE (ti , x j ) ≈ D E j+1 2 j , ⎪
⎪
∂x ⎪
⎪
φE ⎪
⎪
⎪
⎪
⎪
⎪
  ⎪
⎪
∂ G j+1 −2G j +G j−1 ⎪
⎪
∂x DG ∂G
∂x (ti , x j ) ≈ DG 2 , ⎪
⎭
φG
Numerical Solution for a Tumor Cells Dynamics Within Their Micro-environment 65

where,

(·) j+1 − (·) j (·) j − (·) j−1

D + (·) j := , D − (·) j := ,
x x
and the denominator functions
!   "
Dn x χn x 4  f x 2 a22
φ2n := exp( 2
) − 1 , φ f := 2 sin ,  f := ,
χn Dn f 2 Df
!   
Dm x χm x 4 e x 2 a53
φ2m := exp( ) − 1 , φ2E := 2 sinh , e := ,
χm Dm e 2 De
 2   2 
4 g x a62 2 4  p x a72
φ2G := 2 sinh , g := , φ P := 2 sinh ,  p := .
g 2 Dg p 2 Dp

Let St be a positive integer such that t = 1/St where 0 < t < St . Then discretizing
the time interval [0, T ], through the points

0 = t0 < t1 < · · · < t St = T, where ti+1 − ti = t, i = 0, 1, . . . , (t St − 1).

We approximate the time derivative at ti by

N j+1 −N ji F i+1
⎫
j+1 −F j Mi+1
j+1 −M j
i+1 i i
∂n
∂t (x j , ti ) ≈ t , ∂f
∂t (x j , ti ) ≈ ψf , ∂m
∂t (x j , ti ) ≈ t ,⎪
⎪
⎪
⎬
∂ρ Ri+1
j+1 −R j
i
∂E E i+1
j+1 −E j
i
∂G G i+1
j+1 −G j
i
(14)
∂t (x j , ti ) ≈ t , ∂t (x j , ti ) ≈ ψE , ∂t (x j , ti ) ≈ ψG , ⎪
⎪
P i+1 ⎪
⎭
j+1 −P j
i
∂P
∂t (x j , ti ) ≈ ψP ,

where,

ψ f = ψw (t) = (1 − exp(−a22 t))/a22 , ψ E = (1 − exp(−a53 t))/a43 ,

ψG = (1 − exp(−a62 t))/a62 , ψ P = (1 − exp(−a72 t))/a72 ,

where, one can see that

φn → x, φ f → x, ψ f → t, φm → x, φ E → x, ψ E → t, φG → x, ψG → t,

φ P → x, ψ P → t as (t, x) → (0, 0).

The denominator functions in equations (13) and (14) are used explicitly to remove
the inherent stiffness in the central finite derivatives parts and can be derived by
using the theory of nonstandard finite difference methods, see, e.g., [35, 44, 45] and
references therein.
Combining the Eq. (13) for the spatial derivatives with Eq. (14) for time deriva-
tives, we obtain
66 K. M. Owolabi et al.

N ji+1 −N ji N j+1
i+1
−2N ji+1 +N j−1
i+1
(Dx− E ij )
⎫
− Dn = −χn (Dx− n ij )  ⎪
⎪
t φ2n   − 2 ⎪
⎪

1+ D x E ij ⎪
⎪
⎪
⎪
λE
⎪
⎪
Dx+ (Dx− E ij )
⎪
⎪
⎪
⎪
−χn N ji ⎛  − 2 ⎞3/2 ⎪
⎪
D x E ij ⎪
⎪
⎝1+ ⎠ ⎪
⎪
λE ⎪
⎪
⎪
⎪
D + (D − Ri )
⎪
⎪
⎪
⎪
−χ1n Is N ji ⎛ x x j2 ⎞3/2 ⎪
⎪
D x− Rij ⎪
⎪
⎝1+ ⎠ ⎪
⎪
λρ ⎪
⎪
⎪
⎪
a11 (E 4 )ij N ji ⎪
⎪
+ (1 − ), x ∈ [xs , L/2], ⎪
⎪
k 4E +(E 4 )ij N ji n ∗ −a12 ρIs ⎪
⎪
⎪
⎪
F i+1 ⎪
⎪
j −F j F i+1
j+1 −2F j +F j−1
i i+1 i+1
− Df = −a21 (HG )ij (H f )ij ⎪
⎪
ψf φ2f ⎪
⎪
⎪
⎪
+a22 (H f )ij , x ∈ [− L2 , xs ], ⎪
⎪
⎪
⎬
Mi+1
j −M j
i Mi+1
j+1 −2M j +M j−1
i+1 i+1
(Dx− G ij )
t − Dm φ2m
= −χm (Dx− Mij )    − i 2 ⎪
(15)

1+ Dx G j
⎪
⎪
⎪
⎪
λG ⎪
⎪
⎪
⎪
D + (D − G i ) ⎪
⎪
−χm Mij ⎛ x x j2 ⎞3/2 ⎪
⎪
⎪
⎪
D x− G ij ⎪
⎪
⎝1+ ⎠ ⎪
⎪
λG
⎪
⎪
⎪
⎪
+a21 (HG )ij (H f )ij + a31 Mij , x ∈ [− L2 , xs ], ⎪
⎪
⎪
⎪
R j −R j
i+1 i R j+1 −2R j +R j−1
i+1 i+1 i+1 ⎪
⎪
− = −a ( H ) i
( H ) i ⎪
⎪
t D ρ φρ2 41 p j n j ⎪
⎪
⎪
⎪
Rj i ⎪
⎪
+(a42 F j + a43 M j )(1 − ρ∗ ), x ∈ [− 2 , 2 ],
i i L L ⎪
⎪
⎪
⎪
E i+1 − E i E i+1
−2 E i+1
+ E i+1
L L ⎪
⎪
j j
− DE j+1 j j−1
= I a ( H ) i
+ a ( H )i
− a E i
, x ∈ [− , ], ⎪
⎪
ψE φE2  − 51 f j 52 m j 53 j 2 2 ⎪
⎪
⎪
⎪
G j −G j
i+1 i G j+1 −2G j +G j−1
i+1 i+1 i+1
⎪
⎪
− D = I  a ( H ) i
− a G i
, x ∈ [− ,
L L
], ⎪
⎪
ψG G φG2 + 61 n j 62 j 2 2 ⎪
⎪
P j −P j
i+1 i P j+1 −2P j +P j−1
i+1 i+1 i+1 ⎪
⎪
−D = I a (H ) − a P , x ∈ [− , ],
i i L L ⎭
ψP P φ2P − 71 m j 72 j 2 2

with the terminal conditions

⎛ ⎞ ⎫
⎪
⎪
⎜ ⎟ ⎪
⎪
⎜ Exi s +1 −Exi s −1 ⎟ ⎪
⎪
Nxis −1 = Nxis +1 − 2xχn Nxis ⎜   2 ⎟
⎪
⎪
⎝ 
 ⎠ ⎪
⎪
2x 1+
Exi +1 −Exi −1 ⎪
⎪
s s
⎪
⎪
⎛
2xλ E
⎞ ⎪
⎪
⎪
⎪
⎪
⎪
⎜ ⎟ ⎪
⎪
⎜ Rixs +1 −Rixs −1 ⎟ ⎪
⎪
−2xχ1n Nxis ⎜ ⎪
 
 2 ⎟ , ⎪
⎪
⎪
⎝  Rix +1 −Rix −1 ⎠ ⎪
2x 1+ s s ⎪
⎪
2xλρ ⎪
⎪
⎛ ⎞ ⎬
⎪
(16)
⎜ ⎟ ⎪
⎜ G i−L −G i−L ⎟ ⎪
F i−L +1 = F i−L −1 , Mi−L −1 = Mi−L +1 − 2xχm Mi−L ⎜ 2 +1 2 −1 ⎟ ⎪ ⎪
⎪
⎞2 ⎟ , ⎪

⎜  ⎛ i ⎪
2 2 2 2 2 ⎝ 

G −L −G i−L ⎠ ⎪ ⎪
⎪
2x 1+⎝ 2 +1 2 −1 ⎠ ⎪
⎪
⎪
2xλG
⎪
⎪
⎪
⎪
Ri− L +1 = Ri− L −1 , E i−L −1 = (E − )i−L +1 (1 + 2xγ), ⎪
⎪
⎪
⎪
2 2 2 2 ⎪
⎪
G i−L −1 = (G − )i−L +1 (1 + 2xγ), P i−L −1 = (P − )i−L +1 (1 + 2xγ), ⎪
⎪
 2   2 2    ⎪
⎪
2
⎪
⎪
Nx0j = 21 1 + tanh − 1 (0.8 − x j ) , Fx0j = 0.143 21 1 + tanh − 1 (x − 0.2 , ⎪
⎪
⎪
⎭
R0x j = 1.0, M0x j = 0.00, Ex0j = Gx0 j = Px0j = 1.00,

where, the no-flux boundary conditions are discretised by means of the central finite
difference [10], j = −L/2, 2, . . . , L/2 − 1, i = 0, 1, . . . , T − 1 and
Numerical Solution for a Tumor Cells Dynamics Within Their Micro-environment 67

(Hn )ij ≈ N (ti − τ , x j ), (H f )ij ≈ F(ti − τ , x j ), (HG )ij ≈ G(ti − τ , x j ),

(Hm )ij ≈ M(ti − τ , x j ), (H P )ij ≈ P(ti − τ , x j ), (17)

are denoting the history functions corresponding to n, f, m, G, P. The system in

Eq. (15) can further be simplified as
  ⎫
−D n
N j−1
i+1
+ 1
+ 2Dn
N ji+1 −
Dn
N j+1
i+1 ⎪
⎪
φ2 t φ2n φ2n ⎪
⎪
n
−
(Dx E j )
i ⎪
⎪
= −χn (Dx− n ij )  ⎪
⎪
  − i 2 ⎪
⎪

1+ Dx E j ⎪
⎪
⎪
⎪
λE ⎪
⎪
⎪
⎪
D + (D − E i ) ⎪
⎪
−χn N ji ⎛ x x j2 ⎞3/2 ⎪
⎪
D x− E ij
⎪
⎪
⎝1+ ⎠ ⎪
⎪
λE ⎪
⎪
⎪
⎪
⎪
⎪
D + (D − Ri ) ⎪
⎪
−χ1n Is N ji ⎛ x x j2 ⎞3/2 ⎪
⎪
D x− Rij
⎪
⎪
⎝1+ ⎠ ⎪
⎪
λρ ⎪
⎪
⎪
⎪
⎪
⎪
(E 4 )ij Ni N ji ⎪
⎪
+a11 4 N ji (1 − n ∗ −a12j ρIs )+ t , ⎪
⎪
k E +(E 4 )ij ⎪
⎪
⎪
⎪
  ⎪
⎪
⎪
⎪
D 2D f Df ⎪
⎪
− φ2f F i+1
j−1 +
1
ψf + F i+1 − F i+1 ⎪
⎪
φ2f j φ2f j+1 ⎪
⎪
f
⎪
⎪
F ij ⎪
⎪
= −a21 (HG )ij (H f )ij + a22 (H f )ij + ψf , ⎪
⎪
⎪
⎪
⎪
⎪
  ⎪
⎪
⎪
⎪
−D m
Mi+1
j−1 +
1
t + 2Dm
Mi+1 − Dm
Mi+1 ⎪
⎪
φ2m φ2m j φ2m j+1 ⎪
⎪
(Dx− G ij ) ⎪
⎪
= −χm (Dx− Mij )  ⎪
⎪



D x− G ij
2 ⎬
1+ (18)
λG
⎪
⎪
D + (D − G i )
⎪
⎪
⎪
⎪
−χm Mij ⎛ x x j2 ⎞3/2 ⎪
⎪
D x− G ij ⎪
⎪
⎝1+ ⎠ ⎪
⎪
λG ⎪
⎪
⎪
⎪
⎪
⎪
M i
⎪
⎪
+a21 (HG )ij (H f )ij + a31 Mij + t j , ⎪
⎪
⎪
⎪
⎪
⎪
  ⎪
⎪
Dρ 2Dρ
− (x)2 R j−1 + (t) + (x)2 R j − (x)2 R j+1
i+1 1 i+1 Dρ i+1 ⎪
⎪
⎪
⎪
⎪
⎪
R i
⎪
= −a41 (H P ) j (Hn ) j + (a42 F j + a43 M j )(1 − ρ∗ ) + (t) , ⎪
i i i i R j
⎪
⎪
⎪
⎪
⎪
⎪
  ⎪
⎪
⎪
⎪
− φ2 E j−1 + ψ E + φ2 E j − φ2 E j+1
D E i+1 1 2D E i+1 D E i+1 ⎪
⎪
⎪
⎪
E E E ⎪
⎪
E i ⎪
⎪
= I− (a51 (H f )ij + a52 (Hm )ij ) − a62 E ij + ψ Ej , ⎪
⎪
⎪
⎪
⎪
⎪
  ⎪
⎪
⎪
⎪
− φ2 G j−1 + ψG + φ2
DG i+1 1 2DG
G j − φ2 G j+1
i+1 DG i+1 ⎪
⎪
⎪
⎪
G G G ⎪
⎪
Gji ⎪
⎪
= a61 I+ (Hn ) j − a62 G j + ψG ,
i i ⎪
⎪
  ⎪
⎪
⎪
⎪
⎪
⎪
− φ2 P j−1 + ψ P + φ2
D P i+1 1 2D P
P j − φ2 P j+1
i+1 D P i+1
⎪
⎪
⎪
⎪
P P P
⎪
⎪
Pj i
⎭
= a71 I− (Hm ) j − a72 P j + ψ P ,
i i
68 K. M. Owolabi et al.

which can be written as a tridiagonal system given by

(Dx− E ij ) Dx+ (Dx− E ij )

⎫
An N ji+1 = −χn (Dx− n ij )  − χn N ji ⎛  − 2 ⎞3/2
⎪
⎪


 − 2 ⎪
⎪
1+ D x E ij D x E ij ⎪
⎪
λE
⎝1+
λE
⎠ ⎪
⎪
⎪
⎪
i ⎪
⎪
j ⎪
+ (D − Ri ) (E 4 )i N i N
−χ1n Is N ji ⎛ 
D x x j
+ j
N i (1 − j
) + , ⎪
⎪
 ⎞ 3/2 a 11 k 4 +(E 4 )i j n −a ρI t ⎪
⎪
D x− Rij
2
E j
∗ 12 s
⎪
⎪
⎝1+ ⎠ ⎪
⎪
λρ ⎪
⎪
⎪
⎪
Fj i ⎪
⎪
A f F j = −a21 (HG ) j (H f ) j + a22 (H f ) j + ψ f ,
i+1 i i i ⎪
⎪
⎪
⎪
− ⎪
⎪
− (Dx G j ) i
⎬
Am M j = −χm (Dx M j ) 
i+1 i
  

1+ Dx G j
− i 2
⎪
(19)
λG ⎪
⎪
⎪
⎪
+ (D − G i ) ⎪
⎪
D M i
⎪
⎪
−χm Mij ⎛ x x j2 ⎞3/2 + a21 (HG )ij (H f )ij + a31 Mij + t j , ⎪
⎪
D x− G ij ⎪
⎪
⎝1+ ⎠ ⎪
⎪
λG ⎪
⎪
⎪
⎪
⎪
⎪
R R i
⎪
⎪
Aρ R j = −a41 (H P )ij (Hn )ij + (a42 F ij + a43 Mij )(1 − ρ∗ ) + (t) ,
i+1 j
⎪
⎪
⎪
⎪
Ej
i
⎪
⎪
A E E j = I− (a51 (H f ) j + a52 (Hm ) j ) − a62 E j + ψ E ,
i+1 i i i ⎪
⎪
⎪
⎪
G i P i ⎭
A G G i+1
j = a I
61 + ( H )
n j
i − a G i + j , A P i+1 = a I
62 j ψG P j 71 − ( H )
m j
i − a Pi +
72 j ψP
j
,

where,
    ⎫
An = Tri − D n
, 1 + 2D n
,− D n D 2D D
, A f = Tri − φ2f , ψ1f + φ2 f , − φ2f , ⎪
⎪
φ2n t ⎪
 ⎪
φ2n φ2n
   f f f ⎪
⎪
⎪
Am = Tri − φ2 , t + φ2 , − φ2 , Aρ = Tri − (x)2 , t + (x)2 , − (x)2 , ⎪
Dm 1 2Dm Dm Dρ 1 2Dρ Dρ
⎬
 m m m   
DE 2D E DE DG 2DG DG ⎪
(20)
A E = Tri − φ2 , ψ E + φ2 , − φ2 , A G = Tri − φ2 , ψG + φ2 , − φ2 ,
1 1
⎪
⎪
 E  ⎪
⎪
E E G G G
⎪
⎪
DP 2D P DP ⎪
⎭
A P = Tri − φ2 , ψ P + φ2 , − φ2 .
1
P P P

On the interval [0, τ ], the delayed arguments tn − τ belong to [−τ , 0], and therefore,
the delayed variables in Eq. (15) are evaluated directly from the history functions

n 0 (t, x), f 0 (t, x), m 0 (t, x), G 0 (t, x), P 0 (t, x),

as

(Hn )ij ≈ n 0 (ti − τ , x j ), (H f )ij ≈ f 0 (ti − τ , x j ), (Hm )ij ≈ m 0 (ti − τ , x j ),

(HG )ij ≈ G 0 (ti − τ , x j ), (H P )ij ≈ P 0 (ti − τ , x j ), (21)

and Eq. (19) becomes

Numerical Solution for a Tumor Cells Dynamics Within Their Micro-environment 69
⎫
(Dx− E ij ) Dx+ (Dx− E ij ) ⎪
An N ji+1 = −χn (Dx− n ij ) 
  − 2 − χn N ji ⎛  − 2 ⎞3/2 ⎪
⎪
 ⎪
⎪
1+ D x E ij
⎝1+
D x E ij
⎠ ⎪
⎪
λE λE ⎪
⎪
⎪
⎪
D + (D − Ri )
⎪
⎪
⎪
⎪
−χ1n Is N ji ⎛ x x j2 ⎞3/2 ⎪
⎪
D x− Rij ⎪
⎪
⎝1+ ⎠ ⎪
⎪
λρ ⎪
⎪
⎪
⎪
(E 4 )ij Ni N ji ⎪
⎪
+a11 4 N ji (1 − n ∗ −a12j ρIs )+ , ⎪
⎪
k E +(E 4 )ij t ⎪
⎪
⎪
⎪
F ij ⎪
⎪
A f F i+1 = −a21 G 0 (ti − τ , x j ) f 0 (ti − τ , x j ) + a22 f 0 (ti − τ , x j ) + , ⎪
⎪
j ψf ⎬
(Dx− G ij ) D + (D − G i ) (22)
Am Mi+1 = −χm (Dx− Mij ) 
  − i 2 − χm Mij ⎛ x x j2 ⎞3/2 ⎪
⎪
j
 D x− G ij ⎪
⎪
1+ Dx G j ⎝1+ ⎠ ⎪
⎪
λG λG ⎪
⎪
⎪
⎪
M ⎪
⎪
⎪
i
+a21 G (ti − τ , x j ) f (ti − τ , x j ) + a31 M j + t ,
0 0 i j ⎪
⎪
⎪
⎪
R i ⎪
⎪
Aρ Ri+1 = −a 0 (t − τ , x )n 0 (t − τ , x ) + (a F i + a Mi )(1 − R ) + , ⎪
(t) ⎪
j
41 P i j i j 42 j 43 ρ∗ ⎪
j j ⎪
⎪
Ej
i
⎪
⎪
A E E j = I− (a51 f (ti − τ , x j ) + a52 m (ti − τ , x j )) − a62 E j + ψ E ,
i+1 0 0 i ⎪
⎪
⎪
⎪
G i ⎪
⎪
A G G j = a61 I+ n (ti − τ , x j ) − a62 G j + ψG ,
i+1 0 i j ⎪
⎪
⎪
⎪
Pj i ⎪
⎭
A P P j = a71 I− m (ti − τ , x j ) − a72 P j + ψ P .
i+1 0 i

Let s denotes the largest integer such that τs ≤ τ . Then using the system in Eq. (22),
one can compute N ji , F ij , Mij , Rij , E ij , G ij , P ij , for 1 ≤ i ≤ s. Up to this stage, one
interpolates the data

(t0 , N j0 ), (t1 , N j1 ), . . . , (ts , N js ), (t0 , F 0j ), (t1 , F 1j ), . . . , (ts , F sj ),

(t0 , M0j ), (t1 , M1j ), . . . , (ts , Msj ), (t0 , R0j ), (t1 , R1j ), . . . , (ts , Rsj ),
(t0 , E 0j ), (t1 , E 1j ), . . . , (ts , E sj ), (t0 , G 0j ), (t1 , G 1j ), . . . , (ts , G sj ),
(t0 , P 0j ), (t1 , P 1j ), . . . , (ts , P sj ),

using an interpolating cubic Hermite spline ϕ j (t) ([10]). Then

N ji = ϕn (ti , x j ), F ij = ϕ f (ti , x j ), Mij = ϕm (ti , x j ), Rij = ϕρ (ti , x j ), E ij = ϕ E (ti , x j ),

G ij = ϕG (ti , x j ), P ij = ϕ P (ti , x j ),

for all i = 0, 1, . . . , s and j = −L/2, 2, . . . , L/2 − 1.

For i = s + 1, s + 2, . . . , T − 1, when we move from level i to level i + 1 we
extend the definitions of the cubic Hermite spline ϕ j (t) to the point

(ti + t, (Hn )ij , ti + t, (H f )ij , ti + t, (Hm )ij , ti + t, (HG )ij ), ti + t, (H P )ij ).

Then, the history terms (Hn )ij , (H f )ij , (Hm )ij , (HG )ij , (H P )ij can be approximated
by the functions (ϕn ) j (ti − τ ), (ϕ f ) j (ti − τ ), (ϕm ) j (ti − τ ), (ϕG ) j (ti − τ ), (ϕ P ) j
(ti − τ ) for i ≥ s. This implies that,
70 K. M. Owolabi et al.

(Hn )ij ≈ (ϕn ) j (ti − τ ), (H f )ij ≈ (ϕ f ) j (ti − τ ), (Hm )ij ≈ (ϕm ) j (ti − τ ),
(HG )ij ≈ (ϕG ) j (ti − τ ), (H P )ij ≈ (ϕ P ) j (ti − τ ), (23)

and Eq. (22) becomes

⎫
(Dx− E ij ) Dx+ (Dx− E ij ) ⎪
⎪
An N ji+1 = −χn (Dx− n ij ) 

i
 − i 2 − χn N j ⎛  − i 2 ⎞3/2
⎪
⎪
⎪
 ⎪
⎪
1+ Dx E j ⎝1+
Dx E j
⎠ ⎪
⎪
λE λE ⎪
⎪
⎪
⎪
⎪
⎪
Dx+ (Dx− Rij ) ⎪
⎪
−χ1n Is N ji ⎛  ⎪
⎪
2 ⎞3/2 ⎪
⎪
Dx− Rij ⎪
⎪
⎝1+ ⎠ ⎪
⎪
λρ ⎪
⎪
⎪
⎪
⎪
⎪
(E 4 )ij Ni Ni ⎪
⎪
+a11 4 N ji (1 − n −a j ρI ) + tj , ⎪
⎪
k E +(E 4 )ij ∗ ⎪
⎪
12 s ⎪
⎪
⎪
⎪
Fi ⎪
⎪
A f F i+1 = −a21 (ϕG ) j (ti − τ )(ϕ f ) j (ti − τ ) + a22 (ϕ f ) j (ti − τ ) + ψ j , ⎪
⎪
j f ⎬
(Dx− G ij ) Dx+ (Dx− G ij ) (24)
Am Mi+1
j = −χm (Dx− Mij ) 

i
 − i 2 − χm M j ⎛  − i 2 ⎞3/2
⎪
⎪
⎪
 ⎪
⎪
1+ Dx G j ⎝1+
Dx G j
⎠ ⎪
⎪
λ λ ⎪
⎪
G G ⎪
⎪
⎪
⎪
Mj i ⎪
⎪
i
+a21 (ϕG ) j (ti − τ )(ϕ f ) j (ti − τ ) + a31 M j + t , ⎪
⎪
⎪
⎪
⎪
⎪
j ⎪
R i
R ⎪
Aρ R j = −a41 (ϕ P ) j (ti − τ )(ϕn ) j (ti − τ ) + (a42 F j + a43 M j )(1 − ρ∗ ) + (t) , ⎪
i+1 i i
⎪
⎪
⎪
⎪
E i ⎪
⎪
i+1 i
A E E j = I− (a51 (ϕ f ) j (ti − τ ) + a52 (ϕm ) j (ti − τ )) − a62 E j + ψ ,j ⎪
⎪
⎪
⎪
E ⎪
⎪
G i
⎪
⎪
A G G i+1 = a I (ϕ )
61 + n j i (t − τ ) − a G i + j , ⎪
⎪
j 62 j ψG ⎪
⎪
P i ⎪
⎪
i+1 i j ⎪
⎭
A P P j = a71 I− (ϕm ) j (ti − τ ) − a72 P j + ψ ,
P

where,

ϕn (ti − τ ) = [(Hn )i1 , (Hn )i2 . . . , (Hn )iL −1 ] , ϕ f (ti − τ ) = [(H f )i−L , (H f )i−L +1 . . . , (H f )ix0 −1 ] ,
2 2 2

ϕm (ti − τ ) = [(Hm )i−L , (Hm )i−L +1 . . . , (Hm )ix0 −1 ] , ϕG (ti − τ ) = [G i−L , G i−L +1 . . . , G iL −1 ] ,
2 2 2 2 2

ϕ P (ti − τ ) = [(H P )i−L , (H P )i−L +1 . . . , (H P )iL −1 ] .

2 2 2

The FOFDM is then consists of equations (18)–(23). Rewriting the FOFDM as a

system of equations, we have
⎫
An N = Fn , A f F = F f , Am M = Fm , ⎬
Aρ R = Fρ , A E E = FE , A G G = FG , (25)
⎭
A P P = FP .

Let the functions

n(x, t), f (x, t), m(x, t), E(x, t), G(x, t), P(x, t),
Numerical Solution for a Tumor Cells Dynamics Within Their Micro-environment 71

and their partial derivatives with respect to both t and x be smooth such that they
satisfy
 i+ j   i+ j 
 ∂ n(t, x)   f (t, x) 
  ≤ ϒn ,  ∂
 ∂t i x j   ∂t i x j  ≤ ϒ f ,
 i+ j   i+ j 
 ∂ m(t, x)   
  ≤ ϒm ,  ∂ E(t, x)  ≤ ϒ E ,
 ∂t i x j   ∂t i x j 
 i+ j   i+ j 
 ∂ G(t, x)   
  ≤ ϒG ,  ∂ P(t, x)  ≤ ϒ P , ∀i, j ≥ 0, (26)
 ∂t i x j   ∂t i x j 

where,

ϒn , ϒ f , ϒm ,ϒ E , ϒG , ϒ P ,

are constant that are independent of the time and space step-sizes. Then in view of
the FOFDM one can see that the truncation errors ςn , ς f , ςm , ςρ , ς E , ςG , ς P , are given
by
⎫
(ςn )ij = (An n)ij − (Fn )ij = (An (n − N ))ij , ⎪
⎪
⎪
(ς f )ij = (A f f )ij − (F f )ij = (A f ( f − F))ij , ⎪ ⎪
⎪
i ⎪
⎪
(ςm ) j = (Am m) j − (Fm ) j = (Am (m − M)) j , ⎪
i i i
⎬
(ςρ )ij = (Aρ ρ)ij − (Fρ )ij = (Aρ (ρ − R))ij , (27)
⎪
(ς E )ij = (A E E)ij − (FE )ij = (A E (E − E))ij , ⎪ ⎪
⎪
i ⎪
(ςG ) j = (A G Gn) j − (FG ) j = (A G (G − G)) j , ⎪
i i i ⎪
⎪
⎪
(ς P )ij = (A P P)ij − (FP )ij = (A P (P − P))ij . ⎭

Therefore,
⎫
maxi, j |n ij − N ji | ≤ ||A−1
n || maxi, j |(ςn ) j |, ⎪
i
⎪
⎪
⎪
maxi, j | f ji − F ij | ≤ ||A−1
f || maxi, j |(ς f ) j |, ⎪
i
⎪
i ⎪
maxi, j |m j − M j | ≤ ||Am || maxi, j |(ςm ) j |, ⎪
i i −1 ⎪
⎬
−1
maxi, j |ρ j − R j | ≤ ||Aρ || maxi, j |(ςρ ) j |,
i i i
(28)
⎪
⎪
maxi, j |E ij − E ij | ≤ ||A−1
E || maxi, j |(ς E ) j |, ⎪
i
⎪
⎪
maxi, j |G j − G j | ≤ ||A G || maxi, j |(ςG ) j |, ⎪
i i −1 i ⎪
⎪
⎪
−1 ⎭
maxi, j |P j − P j | ≤ ||A P || maxi, j |(ς P ) j |,
i i i

where,
72 K. M. Owolabi et al.
⎫
(ςn )ij ≤ (t) |n tt (ξ)| − Dn (x) ⎪
2
|n x x x x (ζ)|, x ∈ [xs , L/2], ⎪
⎪
2
(t)
12
(x)2 ⎪
⎪
(ς f ) j ≤ 2 | f tt (ξ)| − D f 12 | f x x x x (ζ)|, x ∈ [− 2 , xs ],
i L ⎪
⎪
⎪
⎪
(t) (x) 2
(ςm ) j ≤ 2 |m tt (ξ)| − Dm 12 |m x x x x (ζ)|, x ∈ [− 2 , xs ],
i L ⎪
⎪
⎬
(t) (x) 2
L−x ∗ L−x ∗
(ςρ ) j ≤ 2 |ρtt (ξ)| − Dρ 12 |ρx x x x (ζ)|, x ∈ [− 2 , 2 ],
i (29)
⎪
⎪
(ς E )ij ≤ (t) |E tt (ξ)| − D E (x)
2
|E x x x x (ζ)|, x ∈ [− L2 , L2 ], ⎪
⎪
2 12 ⎪
⎪
(t) (x)2 ⎪
⎪
(ςG ) j ≤ 2 |G tt (ξ)| − DG 12 |G x x x x (ζ)|, x ∈ [− 2 , 2 ],
i L L
⎪
⎪
(t) (x) 2 ⎪
⎭
(ς P ) j ≤ 2 |Ptt (ξ)| − D P 12 |Px x x x (ζ)|, x ∈ [− 2 , 2 ],
i L L

for ti−1 ≤ ξ ≤ ti+1 and x j−1 ≤ ζ ≤ x j+1 . In view of inequalities in (26) we see that
the inequalities in (29) is equivalent to
  ⎫
(ςn )ij ≤ (t)
− Dn (x)
2
ϒ , x ∈ [xs , L/2], ⎪
⎪
 2 12  n ⎪
⎪
⎪
⎪
(t) (x) 2
(ς f ) j ≤ 2 − D f 12 ϒ f , x ∈ [− 2 , xs ],
i L ⎪
⎪
  ⎪
⎪
⎪
⎪
(t) (x) 2
(ςm ) j ≤ 2 − Dm 12 ϒm , x ∈ [− 2 , xs ],
i L ⎪
⎪
  ⎪
⎬
(t) (x) 2
L−x ∗ L−x ∗
(ςρ ) j ≤ 2 − Dρ 12 ϒρ , x ∈ [− 2 , 2 ],
i
(30)
  ⎪
⎪
⎪
⎪
(ς E )ij ≤ (t) − D E
(x)2
ϒ E , x ∈ [− ,
L L
], ⎪
⎪
 2 12  2 2 ⎪
⎪
(t) (x) 2 ⎪
⎪
(ςG ) j ≤ 2 − DG 12 ϒG , x ∈ [− 2 , 2 ],
i L L
⎪
⎪
  ⎪
⎪
(t) (x) 2 ⎪
⎭
(ς P ) j ≤ 2 − D P 12 ϒ P , x ∈ [− 2 , 2 ],
i L L

for ti−1 ≤ ξ ≤ ti+1 and x j−1 ≤ ζ ≤ x j+1 . Moreover, by a result in [51], we have
−1 −1
||A−1 −1 −1
n || ≤ n , ||A f || ≤  f , ||Am || ≤ m , ||Aρ || ≤ ρ , ||A E || ≤  E ,

||A−1 −1
G || ≤ G , ||A P || ≤  P . (31)

Using (30) and (31) in (28), we obtain the following results.

Theorem 3.1 Let

Fn (x, t), F f (x, t), Fm (x, t), Fρ (x, t), FE (x, t), FG (x, t), FP (x, t),

be sufficiently smooth functions so that n(x, t), f (x, t), m(x, t), ρ(x, t), E(x, t),
G(x, t), P(x, t) ∈ C ∞ ([−L , L] × [0, T ]). Let (N ji , F ij , Mij , Rij , E ij , G ij , P ij , ), j =
1, 2, . . . L , i = 1, 2, . . . T be the approximate solutions to (6), obtained using the
FOFDM with N j0 = n 0j , F 0j = f j0 , M0j = m 0j , R0j = ρ0j , E 0j = E 0j , G 0j = G 0j , P 0j =
P j0 ,. Then there exists n ,  f , m , ρ ,  E , G ,  P independent of the step sizes
t and x such that
Numerical Solution for a Tumor Cells Dynamics Within Their Micro-environment 73
⎫
maxi, j |n ij − N ji | ≤ n [ (t) − Dn (x) ⎪
2
]ϒn , ⎪
⎪
2 12
⎪
⎪
⎪
⎪
⎪
maxi, j | f j − F j | ≤  f [ 2 − D f 12 (ζ)]ϒ f , ⎪
(t) (x) 2
i i ⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
maxi, j |m j − M j | ≤ m [ 2 − Dm 12 ]ϒm , ⎪
(t) (x) 2
i i ⎪
⎪
⎪
⎪
⎪
⎬
(t) (x) 2
maxi, j |ρ j − R j | ≤ ρ [ 2 − Dρ 12 ]ϒρ ,
i i (32)
⎪
⎪
⎪
⎪
⎪
⎪
maxi, j |E j − E j | ≤  E [ 2 − D E 12 ]ϒ E , ⎪
(t) (x) 2
i i
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
maxi, j |G j − G j | ≤ G [ 2 − DG 12 ]ϒG , ⎪
i i (t) (x)2
⎪
⎪
⎪
⎪
⎪
⎪
⎪
maxi, j |P j − P j | ≤  P [ 2 − D P 12 ]ϒ P . ⎭
(t) (x) 2
i i

4 Numerical Results and Discussions

Let Sx = St = 60, times t = 20 and t = 30 and using the parameter values in Table 1,
for L = 5 and T = 1, the numerical results for the system in Eq. (15) without a delay
term (omitted) are the same as in [41]. Therefore, numerical results for discretised
system in Eq. (15) with a delay term are presented for times t = 20, & 30 and delay
terms τ = 5, & 10, in Figs. 1 and 2, respectively.
The profiles for the dynamics without-delay (not shown) present that the den-
sity of the transformed epithelia cells are steadily rising to their steady state within
their compartment. Similar phenomena can be seen on the behaviour of the den-
sity of fibroblasts, whereas, for the density of the myofobroblasts, a slight growth
of the density of the cells is noted, which suddenly increases near the end of their
prescribed compartment. Moreover, the concentration of the extracellular matrix,

Table 1 Dimensionless parameter values used for the invasion essay model [31]
Dn = 3.6 × 10−4 D f = 6.12 × 10−5 Dm = 6.12 × 10−4 Dρ = 5.12 × 10−4
De = 5.98 × 10−1 Dg = 3.6 × 10−1 D p = 3.6 × 10−1 χn = 3.6 × 10−8
λ E = λG = λρ ρ = a11 = 0.69 a22 = 1.58 × 10−2 a51 = 2.03 × 10−1
a12 = 1.00
k E = 3.32 κ = 2.88 × 103 B = 5.00 a53 = 2.89 × 10−2
a62 = 2.89 × 10−2 a71 = 3732 a72 = 0.259 a41 = 3732
a43 = 0.518 n  = 2.88 × 103 = 0.1 a21 = 2.61 × 10−2
a31 = 4.53 × 10−3 r f = 100.0 γ = 0.1 χ1n = 1.8 × 10−4
χm = 3.96 × 10−6 a52 = 2.89 × 10−2 a61 = 2.03 × 10−1 a42 = 0.259
74 K. M. Owolabi et al.

1.2 0.15

(a) (b)
1

0.8 0.1
n

0.6

f
0.4 0.05

0.2

0 0
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
x x

0.035 100

0.03
(c) 0

-100
0.025
-200

0.02 -300
Rho
m

0.015 -400

-500
0.01
-600

0.005
-700
(d)
0 -800
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
x x

35 1.08

30
(e) 1.06 (f)
1.04
25

1.02
20
G
E

1
15
0.98

10
0.96

5 0.94

0 0.92
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
x x

1.8 (h)
1.6

1.4

1.2
P

0.8

0.6

0.4

0.2

0
-5 -4 -3 -2 -1 0 1 2 3 4 5
x

Fig. 1 Numerical solution of the system in (5) with delay τ = 5 and at time (t) = 20. Plots a–h
correspond to the behaviors of the transformed epithelial cells (TECs), fibroblasts cells, myofibrob-
lasts cells, concentration of extracellular matrix (ECM), concentration of epidermal growth factor
molecules (EGF), concentration of transformed growth factor molecules (TGF-β), concentration
of matrix metallo proteinase (MMP), respectively
Numerical Solution for a Tumor Cells Dynamics Within Their Micro-environment 75

1.2 0.16

1
(a) 0.14
(b)
0.12

0.8
0.1
n

0.6 0.08

f
0.06
0.4

0.04

0.2
0.02

0 0
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
x x

0.05 200

0.045
(c) 0
0.04

0.035 -200

0.03
Rho -400
m

0.025
-600
0.02

0.015 -800

0.01
-1000
0.005 (d)
0 -1200
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
x x

70 1.08

60 (e) 1.06 (f)

1.04
50
1.02

40 1
G
E

30 0.98

0.96
20
0.94

10
0.92

0 0.9
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
x x

3.5

3 (h)
2.5

2
P

1.5

0.5

0
-5 -4 -3 -2 -1 0 1 2 3 4 5
x

Fig. 2 Numerical solution of the system in (5) with delay τ = 10 and at time (t) = 30. Plots a–h
correspond to the behaviors of the transformed epithelial cells (TECs), fibroblasts cells, myofibrob-
lasts cells, concentration of extracellular matrix (ECM), concentration of epidermal growth factor
molecules (EGF), concentration of transformed growth factor molecules (TGF-β), concentration
of matrix metallo proteinase (MMP), respectively
76 K. M. Owolabi et al.

possesses a slight growth of the concentration which is decreasing as the experi-

ment progresses. It can be seen that the behaviour of fibroblasts gives rise to the
behaviour of the density of the concentration of epidermal growth factor, which
they secretes, whereas, the density of the transformed fibroblasts cells are influenced
by the behaviour of the concentration of the matrix metalloproteinase to a certain
extend, which once more enhanced by the concentration of the epidermal growth
factor molecules. Another interesting phenomena is seen on the behaviour of the
concentration of the transformed growth factor molecules, which is attributed by the
density of the transformed epithelial cells. These phenomena are exactly the same
as in Fig. 1.
Interestingly, the sinusoidal behaviour for the density of transformed epithelial
cells at an initial stage is notable, just before the density rises to its steady state.
The effects of the delay term in the behaviour of the concentration of the trans-
formed growth factor is notable, whereas the degradation of the extracellular matrix,
behaviours of the concentration of epidermal growth factor and matrix metallopro-
teinase are presented in a manner, which one can deduce a relationship between the
concentration of epidermal growth factor with that of the concentration of matrix
metalloproteinase. Thus, the even though this experiment is in silico, the experimen-
tal results mimic that of real scenarios [2].

5 Conclusion

In this chapter, we improved the results for the derived system of nonlinear quasi
time-depended delay parabolic partial differential equations [31, 41]. Even though
the main aim is to improve the results obtained in [31, 41], one can indeed see
that the incorporation of a delay term for crucial transformations ought to take take
place during the interaction of the experiment proposed in [31] is a significant con-
tribution. The incorporation of some transformations, led the original model to be
transformed to a system of non-linear quasi time-depended delay parabolic partial
differential equations. The establishment of well-pose for existence of uniqueness
of solution led to the extension of Gronwall’s inequality for linear delay differential
equations. The analysis of the resulting system of non-linear quasi time-depended
delay parabolic partial differential equations, enabled us to establish the asymptotic
stability condition and the Hopf bifurcation. Consequently, the derived numerical
method based on a fitted operator for solving the modified system in Eq. (5), is ana-
lyzed and implemented iterative in MATLAB R2021. Vividly, the delay term can be
observed after some time of the transformations, which in turn enables one to see
the sensitivity of the density of transformed epithelial cells. Thus, these findings are
indeed essential toward the design of the drug that can slow and/or confine tumor
invasion, particularly when the analysis present that the Hopf bifurcation affects
the entire experiment through the density of fibroblasts. Hence, the future research
direction is to extend the experiment to the higher dimensional space alond the recent
developments reported in [3–5, 25, 37–40, 52–55].
Numerical Solution for a Tumor Cells Dynamics Within Their Micro-environment 77

References

1. W.M. Abd-Elhameed, Y.H. Youssri, Spectral Tau algorithm for certain coupled system of
fractional differential equations via generalized fibonacci polynomial sequence. Iran. J. Sci.
Technol. Trans. A: Sci. 1–12 (2017)
2. J.L. Albritton, J.S. Miller, 3D bioprinting: improving in vitro models of metastasis with het-
erogeneous tumor micro-environments. Dis. Model Mech. 10(1), 3–14 (2017)
3. A. Atangana, Non validity of index law in fractional calculus: a fractional differential operator
with Markovian and non-Markovian properties. Phys. A 505, 688–70 (2018)
4. Atangana, A., Góomez-Aguilar, J.F.: Fractional derivatives with no-index law property: appli-
cation to chaos and statistics. Chaos, Solitons and Fractals 114, 516–535 (2018)
5. A.U. Awan, A. Sharif, K.A. Abro, M. Ozair, T. Hussain, Dynamical aspects of smoking model
with cravings to smoke. Nonlinear Eng. 10(1), 91–108 (2021)
6. Banks, H.T., Banks, J.E., John, E., Bommarco, R., Laubmeier, A.N., Myers, N.J., Rundlöf, M.,
Tillman, K.: Modeling bumble bee population dynamics with delay differential equations 351,
14–23 (2017)
7. Bashier, E.B.M., Patidar, K.C.: Optimal control of an epidemiological model with multiple
time delays. J. Appl. Math. Comput. 292(C), 47–56 (2017)
8. F. Boccardo, G. Petti, A. Lunardi, A. Rubagotti, Enterolactone in breast cyst fluid: correlation
with EGF and breast cancer risk. Breast Cancer Res. 79(1), 17–23 (2003)
9. K. Bottger, H. Hatzikirou, A. Chauviere, A. Deutsch, Investigation of the migra-
tion/proliferation dichotomy and its impact on avascular glioma invasion. Math. Model. Nat.
Phenomena 7(1), 105–135 (2012)
10. R.L. Burden, J.D. Faires, Numerical Analysis (Brooks/Cole, USA, 2011)
11. J. Cheng, L. Weiner, Tumours and their micro-environments: tilling the soil Commentary re:
A.M. Scott et al., A Phase I dose-escalation study of sibrotuzumab in patients with advanced
or metastatic fibroblast activation protein-positive cancer. Cancer Res. 9(5), 1590–1595 (2003)
12. B. Dalal, P. Keown, A. Greenberg, Immunocytochemical localization of secreted transforming
growth factor-beta 1 to the advancing edges of primary tumours and to lymph node metastases
of human mammary carcinoma. Am. J. Pathol. 143(2), 381–389 (1993)
13. E.H. Doha, Y.H. Youssri, M.A. Zakyan, Spectral solutions for differential and integral equations
with varying coefficients using classical orthogonal polynomials. Bull. Iran. Math. Soc. 1–29
(2018)
14. G. Evans, J. Blackledge, P. Yardley, Analytic Methods for Partial Differential Equations
(Springer, Berlin, 1999)
15. U. Foryś, N.Z. Bielczyk, K. Piskała, M. Płomecka, J. Poleszczuk, Impact of time delay in
perceptual decision-making: neuronal population modeling approach. Hindawi Complex. 2017,
14 (2017)
16. A. Friedman, Cancer as multifaceted disease. Math. Model. Nat. Phenom. 7(1), 3–28 (2012)
17. A. Friedman, C. Huang, J. Yong, Effective permeability of the boundary of a domain. Commun.
Partial Differ. Equ. 20(1&2), 59–102 (1995)
18. A. Friedman, Y. Kim, Tumor Cells Proliferation and migration under the influence of their
micro-environment. Math. Biosci. Eng. 8(2), 371–383 (2011)
19. A. Friedman, G. Lolas, Analysis of a Mathematical Model of Tumour Lymphangiogenesis.
Math. Models Methods Appl. Sci. 15(1), 95–107 (2005)
20. S.A. Gourley, Y. Kuang, J.D. Nagy, Dynamics of a delay differential equation model of hepatitis
B virus infection. J. Biol. Dyn. 2(2), 140–153 (2008)
21. R.M. Hafez, Y.H. Youssri, Jacobi collocation scheme for variable-order fractional reaction-
subdiffusion equation. Comput. Appl. Math. 4, 5315–5333 (2018)
22. B.D. Hassard, N.D. Kazarinoff, Y.H. Wan, Theory and Applications of Hopf Bifurcation (Cam-
bridge University Press, Cambridge, UK, 1981)
23. H. Hatzikirou, L. Brusch, C. Schaller, M. Simon, A. Deutsch, Prediction of travelling front
behaviour in a lattice-gas cellular automaton model for tumor invasion. Comput. Math. Appl.
59, 2326–2339 (2010)
78 K. M. Owolabi et al.

24. A. Hooff, Stromal involvement in malignant growth, Advance. Can. Res. 50, 159–196 (1988)
25. T. Hussain, A. Awan, A. Ullah, K.A. Abro, M. Ozair, M. Manzoor, J. Gómez-Aguilar, A.M.
Galal, A passive verses active exposure of mathematical smoking model: a role for optimal
and dynamical control. Nonlinear Eng. 11(1), 507–521 (2022)
26. A. Jafarian, M. Mokhtarpour, D. Baleanu, Artificial neural network approach for a class of
fractional ordinary differential equation 28(4), 765–773 (2017)
27. P. Jones, Extracellular matrix and tenascin-C in pathogenesis of breast cancer. Lancet
357(9273), 1992–1994 (2001)
28. S.M. Kassa, Y. Hailu, The effect of time delay and negligence in the analysis of behav-
ioral change models of infectious disease dynamics for human population. Biomath Commun.
Suppl.: BIOMATH 2017 4(1) (2017)
29. D.I. Ketcheson, K.T. Mandli, A.J. Ahmadia, J. Aron, A. Alghamdi, M. Quezada de Luna,
M. Manuel, M. Parsani, M.G. Knepley, M. Emmett, Matthew, PyClaw: accessible, extensible,
scalable tools for wave propagation problems. SIAM J. Sci. Comput. 34(4), C210–C231 (2012)
30. E. Khain, L.M. Sander, Dynamics and pattern formation in invasive tumor growth. Phys. Rev.
Lett. 96, 188103 (2006)
31. Y. Kim, A. Friedman, Interaction of tumor with its micro-environment: a mathematical model.
Bull. Math. Biol. 72, 1029–1068 (2010)
32. Y. Kim, H. Jeon, H. Othmer, The role of the tumor microenvironment in glioblastoma: a
mathematical model. IEEE Trans. Biomed. Eng. 64(3), 519–527 (2016)
33. Y. Kim, J. Wallace, F. Li, M. Ostrowski, A. Friedman, Transformed epithelial cells and fibrob-
lasts/myofibroblasts interaction in breast tumor: a mathematical model and experiments. J.
Math. Biol. 61(3), 401–421 (2010)
34. H. Lee, A.S. Silva, S. Concilio, Y. Li, M. Slifker, R.A. Gatenby, J.D. Cheng, Evolution of
tumor invasiveness: the adaptive tumor micro-environment landscape model. Can. Res. 71,
6327–6337 (2011)
35. R.E. Mickens, Nonstandard Finite Difference Models of Differential Equations (World Scien-
tific, Singapore, 1994)
36. E. Milotti, S. Stella, R. Chignola, Pulsation-limited oxygen diffusion in the tumour microen-
vironment. Sci. Rep. 7, 39762 (2017)
37. K.M. Owolabi, A. Atangana, Robustness of fractional difference schemes via the Caputo
subdiffusion-reaction equations. Chaos, Solitons Fractals 111, 119–127 (2018)
38. K.M. Owolabi, A. Atangana, Chaotic behaviour in system of noninteger-order ordinary differ-
ential equations. Chaos, Solitons Fractals 115, 362–370 (2018)
39. K.M. Owolabi, Analysis and numerical simulation of multicomponent system with Atangana-
Baleanu fractional derivative. Chaos, Solitons Fractals 115, 127–134 (2018)
40. K.M. Owolabi, Numerical patterns in system of integer and non-integer order derivatives.
Chaos, Solitons Fractals 115, 143–153 (2018)
41. K.M. Owolabi, K.C. Patidar, A. Shikongo, A fitted operator method for a system of delay
model of tumor cells dynamics within their micro-environment. Appl. Math. Inf. Sci. 2013,
1–18 (2022)
42. C.V. Pao, Nonlinear Parabolic and Elliptic Equations (Plenum, New York, 1996)
43. C.V. Pao, Convergence of solutions of reaction-diffusion systems with time delays. Nonlinear
Anal. 48, 349–362 (2002)
44. K.C. Patidar, On the use of non-standard finite difference methods. J. Differ. Equ. Appl. 11,
735–758 (2005)
45. K.C. Patidar, Nonstandard finite difference methods: recent trends and further developments.
J. Differ. Equ. Appl. 22(6), 817–849 (2016)
46. F.A. Rihan, N.F. Rihan, Dynamics of cancer-immune system with external treatment and opti-
mal control. J. Cancer Sci. Therapy DOI, 8–10 (2016)
47. A. Sadlonova, Z. Novak, M. Johnson, D. Bowe, S. Gault, G. Page, J. Thottassery, D. Welch,
A. Frost, Breast fibroblasts modulate epithelial cell proliferation in three-dimensional in vitro
co-culture. Breast Cancer Res. 7(1), R46–R59 (2005)
Numerical Solution for a Tumor Cells Dynamics Within Their Micro-environment 79

48. M. Samoszuk, Z. Tan, G. Chorn, Clonogenic growth of human breast cancer cells co-cultured
in direct contact with serum-activated fibroblasts. Breast Cancer Res. 7, R274–R283 (2005)
49. K. Schmitt, Delay and Functional Differential Equations and Their Applications (Academic,
New York and London, 1972)
50. L. Schwarz, J. Wright, M. Gingras, P. Kondaia, D. Danielpour, M. Pimentel, M. Sporn, A.
Greenberg, Aberrant TGF-beta production and regulation in metastatic malignancy. Growth
Factors 3(2), 115–127 (1990)
51. P.N. Shivakumar, K. Ji, Upper and lower bounds for inverse elements of finite and infinite
tridiagonal matrices. Linear Algebra Appl. 247, 297–316 (1996)
52. H. Singh, Analysis for fractional dynamics of Ebola virus model. Chaos, Solitons & Fractals
138, 109992 (2020)
53. H. Singh, Analysis of drug treatment of the fractional HIV infection model of CD4+ T-cells.
Chaos, Solitons & Fractals 146, 110868 (2021)
54. H. Singh, H.M. Srivastava, D. Baleanu, Methods of Mathematical Modeling Infectious Diseases
(Academic, 2022)
55. H. Singh, D. Baleanu, J. Singh, H. Dutta, Computational study of fractional order smoking
model. Chaos, Solitons & Fractals 142, 110440 (2021)
56. A.M. Stein, T. Demuth, D. Mobley, M. Berens, L.M. Sander, A mathematical model of glioblas-
toma tumor spheroid invasion in a three-dimensional in vitro experiment. Biophys. J. 92, 356–
365 (2007)
57. A.G. Taylor, A.C. Hindmarsh, User Documentation for Kinsol, A Nonlinear Solver for Sequen-
tial and Paraller Computers Center for Applied Scientific Computing, USA (1998)
58. L. Wakefield, D. Smith, T. Masui, C. Harris, M. Sporn, Distribution and modulation of the
cellular receptor for transforming growth factor-beta. J. Cell Biol. 105(2), 965–975 (1987)
59. Y. Yang, O. Dukhanina, B. Tang, M. Mamura, J. Letterio, J. MacGregor, S. Patel, S. Khozin,
Z. Liu, J. Green, M. Anver, G. Merlino, L. Wakefield, Lifetime exposure to a soluble TGF-
β antagonist protects mice against metastasis without adverse side effects. J. Clin. Investig.
109(12), 1607–1615 (2002)
60. M. Yashiro, K. Ikeda, M. Tendo, T. Ishikawa, K. Hirakawa, Effect of organ-specific fibroblasts
on proliferation and differentiation of breast cancer cells. Breast Cancer Res. Treat. 90, 307–313
(2005)
61. Q.X. Ye, Z.Y. Li, Introduction to Reaction-Diffusion Equations (Science Press, Beijing, China,
1990)
A Mathematical Model to Study
Regulatory Properties and Dynamical
Behaviour of Glycolytic Pathway Using
Bifurcation Analysis

Shruti Tomar, Naresh M. Chadha, and Ankita Khanna

Abstract The glycolytic pathway is an almost universal central pathway of glucose

catabolism and is known to be highly conserved across various species. In some mam-
malian tissues and cell types (for example: erythrocytes, renal medulla, brain, and
sperm), the glycolytic breakdown of glucose is the sole source of metabolic energy.
In glycolysis, phosphofructokinase-1 (PFK-1) catalyzes the transfer of a phospho-
ryl group from ATP to fructose 6-phosphate to yield fructose 1, 6-bisphosphate. The
PFK-1 reaction is essentially an irreversible reaction under cellular conditions. Addi-
tionally, it is the first “committed” step in the glycolytic pathway because glucose
6-phosphate and fructose 6-phosphate have other possible fates, but fructose 1, 6-
bisphosphate is targeted for glycolysis. In this chapter, a dynamical system based on
mass balance equations and S-system representation has been formulated to study the
regulatory properties, rate control distributions, and dynamical behaviour of reac-
tions in this glycolytic pathway. This representation involves several parameters.
Few standard tools from the bifurcation analysis such as phase-portraits, time series
plots, the Lyapunov exponents and d∞ plots have been employed to investigate the
stability of the pathway with respect to certain chosen parameters. The study is fur-
ther extended by eliminating the glycogen branch from the original pathway and by
adding an external perturbation into the system.

1 Introduction

The beginning of the study of glycolysis started with Eduard Buchner’s discovery of
fermentation in broken extracts of yeast cells in 1897. Throughout 1930s, the reac-
tions of glycolysis in extracts of yeast and muscle were a major focus of biochemical
research which led to whole pathway elucidation in yeast by Otto Warburg and Hans

S. Tomar · N. M. Chadha (B) · A. Khanna

School of Physical Sciences, DIT University, Uttarakhand, India
e-mail: [email protected]
A. Khanna
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 81
H. Singh and H. Dutta (eds.), Computational Methods for Biological Models,
Studies in Computational Intelligence 1109,
https://doi.org/10.1007/978-981-99-5001-0_4
82 S. Tomar et al.

von Euler-Chelpin [50]. The whole glycolytic pathway was elucidated in muscles
by Gustav Embden and Otto Meyerhof [10, 25]. Glycolysis occurs in the cytosol
of the cell. It is a metabolic pathway that creates ATP without the use of oxygen
but can also occur in the presence of oxygen. In cells, during aerobic respiration, the
pyruvate formed from the pathway enters the citric acid cycle and goes through oxida-
tive phosphorylation to undergo oxidation into carbon dioxide and water. In highly
oxidative tissues, such as the heart, pyruvate production is essential for acetyl-CoA
synthesis and L-malate synthesis. It serves as a precursor to many molecules, such
as lactate, alanine, and oxaloacetate. Glycolysis precedes lactic acid fermentation
i.e., the pyruvate produced in glycolysis serves as pre-requisite for the lactate pro-
duced in the lactic acid fermentation. Skeletal muscles prefer to catalyze glucose
into carbon dioxide and water during heavy exercise. But when oxygen is inad-
equate, these muscles simultaneously undergo anaerobic glycolysis and oxidative
phosphorylation [19].
The chemistry of glycolysis has been completely conserved during evolution from
vertebrates to yeast to spinach. Across different species, glycolysis differs only in its
regulation and in subsequent metabolic fate of the pyruvate formed. The thermody-
namic principles and types of mechanisms regulating glycolysis are common to all
pathways of cell metabolism. Many authors have shown their interest in glycolysis
pathway in different kind of cells namely E. coli cells, mammalian cells, pancreatic
β-cells etc. [7, 8]. In a study by Maria et al. [22], some characteristics and cell com-
ponents that caused stable glycolytic oscillations in E. coli cells have been identified.
Mulukutla et al.[24] studied the diversity of steady states in glycolysis and shifts in
metabolic states in cultured mammalian cells with the help of mathematical mod-
elling. They reported bistable behaviour of glycolytic enzymes regulated by lactate
inhibition, including the AKT pathway, and provided a new technique to control the
metabolism of mammalian cells in fed-batch cultures.
There are many studies available in the literature where the authors have attemped
to study Glycolysis through experimental and mathematical means. Camacho et al.
[6] have developed a model consisting a system of non-linear differential equations
for those patients who were affected by retinitis pigmentosa, an inherited retinal
disease which leads to irreversible blindness. This technique is used to model cellular
and molecular interactions in a cone cell. Torres and Melendez-Hevia [42] studied
the glycolysis pathway in a rat liver and determined some coefficients of glucokinase,
glucose 6-phosphate isomers and phosphofructokinase with the help of shortening
and enzyme titration method that can control the concentration of glycolysis in rat
liver. Their study suggested that phosphofructokinase may play an important role in
control of liver glycolysis. Bakker et al. [2] constructed and tested a kinetic model
with experimental results that described the steady state behaviour of glycolysis in the
blood-stream form of Trypanosoma brucei quantitatively. Verburg [45] constructed
a dynamical system with the help of rate equations to control glycolysis in T. brucei
and observed the effect of included rate constants. Refer to [12, 20, 28–31, 36–39,
47, 51], where the authors have used mathematical tools to study various related
models.
A Mathematical Model to Study Regulatory Properties … 83

Fig. 1 Pathway of glycolysis and glycogenolysis in terms of biological names and model variables
[48]

In this study, we consider the Glycolytic-Glycogenolytic pathway in perfused

rat-liver from Case study-4 of [48], shown in Fig. 1.
Initial steps of glycolysis and glycogenolysis, in terms of biochemical names
and, in terms of model variables are shown in Fig. 1. It is important to note that
glycolysis and glycogenolysis are central steps of carbohydrate metabolism. They
yield common intermediate, glucose-6-phosphate, which is subsequently used by the
organism as a major source of energy. The concentration of glucose in both human and
animal bodies varies according to nutritional requirements, just as fasting for a few
days does not cause a noticeable decline in glucose levels. In these circumstances, the
body must turn to other energy sources. In case of animals, if they starve for one day
they loose almost all glycogen. Subsequent feeding of glycogenic substances lead
to rapid replenishment in glucose level. The main route of glycogen utilization is
glycogenolysis via glycogen phosphorolysis to glucose-1-phosphate and subsequent
conversion to glucose-6-phosphate via the phosphoglucomutase reaction. Glucose-
6-phosphate can be used for energy production or formation of glucose.
In the present study, the pathway depicted in Fig. 1 is represented as a dynamical
system of non-linear ordinary differential equations, referred to as S-system repre-
sentation, which is primarily based upon mass balance equations. As we shall further
see that this dynamical system has many variables and parameters, and variation in
any one of them may disturb the dynamics of the whole pathway. It is almost impos-
sible to consider all the variables and the parameters involved in the pathway at one
time, thus we restrict ourselves by choosing four parameters, namely h 33 , h 55 , h 5,10
and h 52 . These parameters are involved in the phosphofructokinase and the glucoki-
nase reactions, respectively. Furthermore, these parameters may directly or indirectly
influence the concentrations of glucose-6-P and fructose-6-P in the pathway.
84 S. Tomar et al.

The main contributions of this study are highlighted here: we have studied the
effect of range of the parameters h 33 , h 55 , h 5,10 and h 52 on the concentration of glucose
in glycolytic pathway with the help of standard tools from the bifurcation analysis
such as phase-portraits, time series plots, Lyapunov exponents and d∞ plots, [23].
In glycolytic pathway, glycolysis and glycogenolysis form central steps to produce
glucose-6-P which can be further used to generate glucose. Since both the branches
play an important role in this pathway, it was intriguing for us to study how this
pathway behaves when we remove any one of these branches. We study the pathway
by removing the right branch (glycogen branch) of the glycolytic pathway in rat
liver. In our study, we have perturbed the glycolytic system by adding a forcing
term of the form f 0 cos(ωt). The main observation of the study is that the change
in the parameters h 33 , h 55 , and h 5,10 , have moderate effects on the dynamics of the
pathway but a positive value of h 52 may convert a stable system into an unstable one.
We have also observed that if we remove the right branch of the pathway then the
converted system does not show any specific changes in the concentration of glucose
but external perturbations such as f 0 cos(ωt) can reverse the process.
The organization of chapter is as follows: In Sect. 2, the pathway is represented as
a dynamical system using the mass balance equations and a S-system representation.
In Sect. 3, we have discussed the interrelation between the parameters which are
included in the dynamical system. The standard tools from bifurcation analysis have
been employed to study the pathway in Sect. 4. In Sect. 5, one of the branches of
pathway is removed and external perturbation is introduced in the dynamical system
and the standard tools have been used to study the stability of the dynamics of the
pathway. Results and future work have been discussed in Sects. 6 and 7, respectively.
Few intermediate steps to calculate the rate constants, and certain important steps
for the steady state analysis of the pathway are presented in Appendices 1 and 2.

2 Glycolytic Pathway, Mass Balance Equations,

and S-System Representation

To represent the glycolytic pathway as a system of differential equations, the first

step is to identify the independent and dependent variables. The steady state con-
centrations of the dependent and independent variable have been taken from Case
study-4 [48].

(1) Steady state concentrations and steady state fluxes:

For the model, the primary data consist of steady-state concentrations and steady-
state fluxes.

V1+ = 1.068 µmol/min per gram fresh liver weight

V3− = 1.714 µmol/min per gram fresh liver weight
V5− = 0.646 µmol/min per gram fresh liver weight
A Mathematical Model to Study Regulatory Properties … 85

(2) Dependent Variables:

[G1P] = X 1s = 0.067 mM
[G6P] = X 2s = 0.465 mM
[F6P] = X 3s = 0.150 mM

(3) Independent variables:

[Pi ] = X 4 = 10.00 mM
[Glucose] = X 5 = 5.00 mM
[Phosphorylase a] = X 6 = 3.00 µmol/min/gm fresh liver weight
[Phosphoglucomutase] = X 7 = 40.00 µmol/min/gm fresh liver weight
[Phosphoglucose iso.] = X 8 = 136.00 µmol/min/gm fresh liver weight
[Phosphofructokinase] = X 9 = 2.86 µmol/min/gm fresh liver weight
[Glucokinase] = X 10 = 4.00 µmol/min/gm fresh liver weight
[Glycogen] = X 11 = 50.00 µmol/min/gm fresh liver weight

(4) Steady-state Fluxes:

V1+ = V1− = 1.068 µmol/min/gm fresh liver weight

V2+ = V2− = 1.714 µmol/min/gm fresh liver weight
V3+ = V3− = 1.714 µmol/min/gm fresh liver weight
V5− = 0.646 µmol/min/gm fresh liver weight

Since X 1 , X 2 , X 3 are three dependent variables, so we have three equations with

other variables. The mass-balance equation can represented as

d X1
= V1+ (X 4 , X 6 , X 11 ) − V1− (X 1 , X 2 , X 7 ),
dt
d X2
= V2+ (X 1 , X 2 , X 5 , X 7 , X 10 ) − V2− (X 2 , X 3 , X 8 ),
dt
d X3
= V3+ (X 2 , X 3 , X 8 ) − V3− (X 3 , X 9 ). (1)
dt

The rate of change in concentration with respect to time is represented by ddtX 1 , ddtX 2 ,
and ddtX 3 , whereas the function Vi+,− s are functions of concentrations. This system
is concerned with both the generation and depletion of X i ’s. A positive sign (+)
indicates a change in product, while a negative sign (−) indicates a change in the
substrate.
These equations given by Eq. (1) have been converted into a system of non-linear
ordinary differential equations, commonly referred to as S-system representation.
Each flux term’s variable is represented by a product of power-law functions with a
kinetic order parameter gi j or h i j , whose two indices correspond to the number of
affected and effecting variables, respectively. Furthermore, each flux term is assigned
86 S. Tomar et al.

Table 1 Kinetic order values and their representative reactions

Kinetic orders Values Reaction Reference
g14 , g16 0.66, 1 Phosphorylase a [11, 34]
reaction
h 11 , h 12 , h 17 1.53, −0.59, 1 Phosphoglucomutase [11, 32, 35]
reaction
h 55 , h 52 , h 5,10 0.84, −0.1, 1 Glucokinase reaction [3, 13, 26, 34, 44]
h 22 , h 23 , h 28 3.97, −3.06, 1 Phosphoglucose [3, 26, 40]
isomerase reaction
h 33 , h 39 0.3, 1 Phosphofructokinase [13, 26, 33, 43, 44]
reaction

a rate constant αi or βi , whose index corresponds to the number of equation. The

resulting S-system representation of the dynamical system is

d X1 g g
= α1 X 4 14 X 6 16 − β1 X 1h 11 X 2h 12 X 7h 17 ,
dt
d X2 g g g g g
= α2 X 1 21 X 2 22 X 5 25 X 7 27 X 102,10 − β2 X 2h 22 X 3h 23 X 8h 28 ,
dt
d X3 g g g
= α3 X 2 32 X 3 33 X 8 38 − β3 X 3h 33 X 9h 39 . (2)
dt
In this system Eq. (2), we have total 28 exponents and rate constant parameters,
the values of some of them are taken from literature, shown in Table 1. The values of
rest of the parameters viz. g21 , g22 , g25 , g27 , g32 , g33 , g38 and g2,10 are computed from
the branch point constraints, refer to Appendix 1. These values are g21 = 0.95, g22 =
−0.41, g25 = 0.32, g27 = 0.62 and g2,10 = 0.38 and g32 = h 22 = 3.97, g33 = h 23 =
−3.06, g38 = h 28 = 1.
At steady state, the rate constants are calculated by equating the measured fluxes
with the relevant power-law terms [48].
The main emphasis in this study is on the parameters h 33 , h 55 , h 52 and h 5,10
which are involved in mathematical representation of the phosphofructokinase and
the glucokinase reactions, respectively. Therefore, any change in the value of these
parameters will affect the concentrations of X 2 i.e. Glucose-6-P and X 3 i.e. Fructose-
6-P.
The dynamical system Eq. (2) consist many parameters which appear as exponents
gi j , h i j , and the rate constants, αi ’s and βi ’s. In many practical applications, it may
not be easy to control a parameter or a set of particular parameters in order to obtain
a desired outcome. Thus, it is instructive to explore the inter-dependency of these
parameters. For this, we formulate a set of transcendental equations obtained by
considering the steady state of the dynamical system Eq. (2). The solution set of
these transcendental equations will give the interrelations between certain important
parameters. This is the subject matter of Sect. 3.
A Mathematical Model to Study Regulatory Properties … 87

3 Interrelation Between Certain Parameters

By using certain precursor-product relations and the branch point restrictions (refer
to Appendix 1), we have

1.068 0.646 1.608

g21 = h 11 , g25 = h 55 , g27 = h 17 ,
1.741 1.741 1.714
0.646 1.068 0.646
g2,10 = h 5,10 , g22 = h 12 + h 52 . (3)
1.714 1.741 1.741
And by using the precursor-product relation, we have

g1,11 = 0, g22 = h 22 , g33 = h 23 , g38 = h 28 , α3 = β2 . (4)

From the dynamical system Eq. (2) at steady state, in conjunction with relations in
Eqs. (3) and (4), we have the following set of transcendental equations

α1 10g14 3g16 − β1 (0.067)h 11 (0.46)h 12 40h 17 = 0,

α2 X 10.6314h 11 X 20.6314h 11 +0.3769h 52 50.3769h 55 400.6314h 17 40.3769h 5,10 − β2 X 2h 22 X 3h 23 136h 28 = 0,
β2 X 2h 22 X 3h 23 136h 28 − β3 X 3h 33 2.86h 39 = 0. (5)

By using symbolic calculation in MAPLE software, the solution set of the tran-
scendental Eq. (5) is

β1 0.067h 11 0.46h 12 40h 17

α1 = ,
3g16 10g14
α2 = α2 , β1 = β1 ,
α2 0.0670.6134405514h 11 0.465(0.6134405514h 12 +0.3768961494h 52 ) 50.3710511200h 55
β2 =
0.465h 22
0.9381563594h 17 0.3768961494h 5,10
40 4
,
0.15h 23 136h 28
(1.591199772e − 18)α2 0.0670.6134405514h 11 0.4650.6134405514h 12 +0.3768961494h 52
β3 =
1
50.3710511200h 55 400.9381563594h 17 40.3768961494h 5,10
. (6)
0.15h 33
As mentioned earlier, in the present study, we have considered the parameters
h 52 , h 33 , h 5,10 and h 55 to analyse the behaviour of the dynamical system Eq. (2). By
setting the values of other parameters from Tables 1 and 2, the solution set Eq. (6)
reduces to the following form
88 S. Tomar et al.

Table 2 Rate constants and their values

Rate constant Values Rate constant Values
α1 0.077884314 α2 0.585012402
α3 0.000793456 β1 1.062708258
β2 0.000793456 β3 1.05880847
β5 0.038706421

Fig. 2 Behaviour of the rate constant parameter β2 , and β3 for range of parameter h 5,10 ∈
(0.001, 1.5), h 55 ∈ (0.001, 1.5), h 33 ∈ (0.001, 1), and h 52 ∈ (−0.06, −0.001). The values of
parameters are g14 = 0.66, g16 = 1, h 11 = 1.53, h 12 = −0.59, h 17 = 1, h 39 = 1, h 22 = 3.97,
h 23 = −3.06, h 28 = 1, and rate constants values are α2 = 0.58501242, β1 = 1.062708258

α1 = 14.7388,
β2 = (2.3451e + 06)0.465(0.3769h 52 −0.3619) 50.3711h 55 h 5,10 ,
(2.3446e − 18)0.465(0.3769h 52 −0.3619) 40.3769h 5,10 50.3711h 55
β3 = . (7)
0.15h 33
The relations between the parameters β2 , β3 and h 52 , h 33 , h 5,10 and h 55 given
by Eq. (7) are depicted in Fig. 2. In Fig. 2a, b, the surface plots for β2 with respect
to the parameters h 55 , h 52 and h 5,10 are plotted. For the range of parameters h 55 ∈
(0.001, 1.5) and h 5,10 ∈ (0.001, 1.5), β2 increases but for h 52 ∈ (−0.6, −0.001), a
slight decrease in value of β2 is observed. For the same range of the parameters used
to Fig. 2a, b, β3 show a slightly different trend as compared to β2 which is shown in
Fig. 2c, d and e. It is important to note that the scale of the surface plots for β2 and β3
for the same range of the parameters h 55 , h 52 and h 5,10 is very different. Furthermore,
A Mathematical Model to Study Regulatory Properties … 89

a change in the values of these parameters can affect the concentration of X 2 and X 3
i.e., Glucose-6-P and Fructose-6-P in the Glycolytic pathway.
In the next section Sect. 4, we study the dynamical system Eq. (2) to investigate
the behaviour of the pathway with respect to these chosen parameters by using the
standard tools from the Bifurcation Analysis. The commonly used tools are phase-
portraits, Lyapunov exponents, d∞ parameters plots and time series plots [41].

4 Bifurcation Analysis of the Glycolytic Pathway

Bifurcation analysis is a mathematical study of qualitative changes in system dynam-

ics which is commonly used to investigate the unexpected appearance, disappearance,
or change in the stability of equilibrium points with respect to certain parameters or
perturbations. Bifurcation analysis has long been used to investigate dynamical sys-
tems emerging from varied real world problems, refer to [15, 41, 49] and references
therein.
For the bifurcation analysis also, we shall continue with our chosen parameters for
the study, h 55 , h 33 ,h 52 and h 5,10 . We substitute the values of the independent variables
X 4 to X 11 in the dynamical system Eq. (2) to obtain the following dynamical system

d X1
= α1 10g14 3g16 − β1 X 1h 11 X 2h 12 40h 17 ,
dt
d X2
= α2 X 10.6314h 11 X 20.6314h 11 +0.3769h 52 50.3769h 55 400.6314h 17 40.3769h 5,10 − β2 X 2h 22 X 3h 23 136h 28 ,
dt
d X3
= β2 X 2h 22 X 3h 23 136h 28 − β3 X 3h 33 2.86h 39 . (8)
dt

To investigate the stability of the dynamical system, we first calculate the Jacobian
matrix [5, 21] of the dynamical system Eq. (8) which is given below
⎡ ⎤
A B 0
J = ⎣C D E ⎦, (9)
0 F G

where

64.8752X 10.53 25.016X 11.53 17.4848(40.38h 5,10 )(50.37h 55 )X 2(0.38h 52 −0.3619)

A=− , B= , C= ,
X 20.59 X 21.59 X 10.0614
1  
(0.38h 52 −1.3619)
D= 18.672(50.37h 55 )(4(0.38h 5,10 ) X 10.9386 X 2 0.38h 52 − 0.3619
X 33.06
1  
− 3.06 0.3382(0.465)0.37h 52 −0.354 (40.37h 5,10 )X 22.97 ,
X3
90 S. Tomar et al.

0.3382(0.465)0.37h 52 −0.354 40.37h 5,10 50.37h 55 X 23.97 1170.38(40.38h 5,10 )(50.37h 55 )X 22.97
E= , F= ,
X 34.06 X 33.06
−0.2607(0.465)0.37h 52 −0.354 40.37h 5,10 50.37h 55 X 23.97
G=
X 34.06
(8.1112e − 18)0.4650.38h 52 −0.3599 40.376h 5,10 )50.375h 55 )h h33−1
33 X 3 )
− .
0.15h 33

For obtaining the equilibrium point(s), we have solved the dynamical sys-
tem Eq. (8) at steady state in MAPLE. Considering the number of parameters
involved in the dynamical system, the expressions obtained for the equilibrium
points are very complex. Thus, we consider the equilibrium point as (X 1 , X 2 , X 3 ) =
(0.067, 0.465, 0.150) which is the concentration of dependent variables at the steady
state reported in case study-4 [48]. For this equilibrium point, determinant of the
Jacobian matrix Eq. (9) is

|J | = (6.3627e − 115)((2.1808e + 117)0.4650.75h 52 40.75h 5,10 50.74h 55 h 52

− (4.1260e + 100)0.4650.75h 52 40.746h 5,10 50.755h 55 h 33
+ (4.2913 + 99)0.4650.75h 52 40.756h 5,10 50.755h 55 h 33 h 52 ). (10)

It is well known that the nature of an equilibrium point can be determined by the
Jacobian matrix corresponding to a dynamical system. If the value of determinant of
the Jacobian matrix is positive, i.e. |J | > 0, then, equilibrium point will be a stable
point and if it is negative i.e. |J | < 0, then, the equilibrium point will be an unsta-
ble point or saddle point. For this dynamical system Eq. (8), determinant Eq. (10)
for the range of the set value of (h 33 , h 52 ) ∈ (0.001, 1) × (−0.6, −0.001) and
(h 55 , h 5,10 ) ∈ (0.001, 1.5) × (0.001, 1.5) is negative. Hence, the equilibrium point
(X 1 , X 2 , X 3 ) = (0.067, 0.465, 0.150) is a saddle point which is shown in Fig. 3a, b.
But when we choose positive value of inhibition parameter h 52 ∈ (0.001, 0.6), then
the determinant value becomes positive. Hence, for this parameter range, the equi-
librium point (X 1 , X 2 , X 3 ) = (0.067, 0.465, 0.150) will behave as a stable point,
Fig. 4a.
Characteristic polynomial of Jacobian matrix Eq. (9) is

aλ3 + bλ2 + cλ + d = 0, (11)

A Mathematical Model to Study Regulatory Properties … 91

Fig. 3 Negative value of determinant of Jacobian matrix Eq. (9) in range of parameters h 55 ∈
(0, 1.5), h 5,10 ∈ (0, 1.5), h 52 ∈ (−0.6, −0.001) and h 33 ∈ (0.001, 1) shown in (a), (b)

Fig. 4 The positive value of

determinant in the positive
range of parameter
h 52 ∈ (0.001, 0.6). Rest
parameter values are h 33 =
0.3, h 5,10 = 1, h 55 = 0.84

where

a = (1.2745e + 118)0.15h 33 ,
b = 0.15h 33 (0.465)0.37h 52 50.37h 55 ((6.5444e + 119)40.37h 5,10 − (2.0082e + 118)40.38h 5,10 h 52
+ (9.0783e + 101)40.38h 5,10 h 33 + (1.9284e + 118)40.38h 5,10 ) + (3.1003e + 119)(3/20)h 33 ),
c = (0.15)h 33 (0.465)0.37h 52 40.37h 510 50.37h 55 ((1.5920e + 121) + (2.2084e + 103)h 33
− (4.8852e + 119)h 52 ) + (0.15)h 33 (0.465)0.75h 52 40.76h 510 50.74h 55 ((1.3754e + 103)h 33 +
(1.3737e + 102)h 33 ) − (7.2695e + 119)h 52 − (1.4305e + 102)h 33 h 52 + (6.9809e + 119)),
d = (0.15)h 33 (0.465)0.75h 52 40.76h5,10 50.74h 55 ((3.3458e + 104)h 33 − (1.7684e + 121)h 52
− (3.4798e + 103)h 33 h 52 ).

Next, we investigate the nature of eigenvalues of the characteristic polynomial

Eq. (11). According to lemma [9], if the coefficients of a polynomial are positive
then, the roots of the polynomial will be negative. Here, in Fig. 5, we have shown
92 S. Tomar et al.

Fig. 5 Mesh plots of the coefficients of characteristic polynomial Eq. (11) for range of h 33 ∈
(0.001, 1), h 55 ∈ (0.001, 1.5), h 52 ∈ (−0.6, −0.001) and h 5,10 ∈ (0.001, 1.5) shown in (a), (b),
(d), (e), (g) and (h). For positive range of parameter h 52 ∈ (0.001, 0.6) shown in (c), (f), (i)

that for a chosen range of parameters h 33 , h 55 , h 52 and h 5,10 , the value of coefficients
will be positive. Hence, roots of the characteristic polynomial will be negative. On
this basis, for the range of these parameters, one can predict that the the dynamical
system Eq. (8) will remain stable.
The phase portrait of the dynamical system Eq. (8) for h 33 ∈ (0.001, 1) and related
time series plots that illustrate the rate of rise in concentration of X 1 , X 2 , and X 3 with
respect to these parameters are shown in Fig. 6. With increasing value of parameter
h 33 , the concentration of X 3 increases faster than X 2 and X 1 ; for h 33 = 1, system
reaches to a break point, refer to Fig. 6 (i). Figure 7 shows similar behaviour for other
parameters h 55 , h 52 and h 5,10 with respect to time. In Fig. 8, we have noticed that the
rate of change in X 1 is greater than X 2 and X 3 during short time periods (0 − 0.05).
However, after a period of time, the rate of change in X 1 gradually decreases, and
the concentration of X 3 gradually increases. This is due to the activation of feedback
mechanism in a cell catalysed by Phosphoglucomutase and Phosphoglucoisomerase.
A Mathematical Model to Study Regulatory Properties … 93

Fig. 6 Phase portrait and time series plots of the dynamical system Eq. (8) for the range of parameter
h 33 ∈ (0.001, 1), time step is [0, 50] and initial value are [0.067, 0.465, 0.150]

It is well known that a dynamical system can exhibit both regular and chaotic mode
of evolution. The exponentially fast divergence or convergence of near orbits in the
phase portrait can be defined by Lyapunov exponents which provide a qualitative
and quantitative characterisation of the dynamical behaviour of a system [1, 21].
If any one value of the Lyapunov exponent becomes positive then the system is
found to be chaotic. Figure 9 shows the dynamics of the Lyapunov exponent for
two set of values for h 33 . Figure 9a shows all negative values of Lyapunov exponent
94 S. Tomar et al.

Fig. 7 Time series plots of the dynamical system Eq. (8) for the range of parameter
h 55 , h 5,10 ∈ (0.001, 1.5) and h 52 ∈ (−0.6, −0.001),. Time step is [0, 50] and initial value are
[0.067, 0.465, 0.150]

(LE) for h 33 = 0.005, h 5,10 = 1, h 55 = 0.84, i.e. system may be stable but h 33 = 0.3
keeping all other parameters same as for in Fig. 9a, one L E ∗ (Lyapunov exponent
for h 33 = 0.3) becomes positive with the progression in time, shown in Fig. 9b.
Corresponding values of Lyapunov Exponents are shown in Table 3.
The overall behaviour of dynamical system is decided by the largest Lyapunov
exponent which is called the d∞ parameter. For this, we have considered the equation
as
d˙z = λ1 dz − γ dz2 , (12)
A Mathematical Model to Study Regulatory Properties … 95

Fig. 8 Time series plots of the dynamical system Eq. (8) for the range of parameter h 33 ∈ (0.001, 1),
h 55 , h 5,10 ∈ (0.001, 1.5) and h 52 ∈ (−0.6, −0.001), time step is [0, 0.05] and initial value are
[0.067, 0.465, 0.150]
96 S. Tomar et al.

Fig. 9 Lyapunov exponent graph of the dynamical system Eq. (8) for the parameter h 33 = 0.005
and h 33 = 0.3, h 55 = 0.84, h 5,10 = 1, h 52 and time interval is [0 : 50] and initial values are
[0.067, 0.465, 0.150]

Table 3 Lyapunov Exponents values with respect to time t = [0 : 50] when h 33 = 0.005 and 0.3,
h 55 = 0.84, h 5,10 = 1. Here L E and L E ∗ represents the Lyapunov exponent for h 33 = 0.005 and
h 33 = 0.3 respectively
t L E1 L E2 L E3 L E 1∗ L E 2∗ L E 3∗
5 −0.180282 −14.917085 −32.487695 −0.049546 −11.507682 −29.511196
10 −0.090572 −9.582886 −25.784965 −0.010262 −6.976805 −22.750673
15 −0.060607 −7.321299 −22.590820 −0.000617 −5.141802 −19.570416
20 −0.045592 −6.025724 −20.593075 0.003043 −4.120645 −17.595340
25 −0.036566 −5.171144 −19.178657 0.004698 −3.461569 −16.204213
30 −0.030538 −4.558621 −18.102693 0.005504 −2.997413 −15.150440
35 −0.026226 −4.094747 −17.244812 0.005898 −2.651086 −14.313309
40 −0.022988 −3.729381 −16.537723 0.006074 −2.381807 −13.625556
45 −0.020465 −3.433004 −15.940341 0.006127 −2.165862 −13.046217
50 −0.018445 −3.187015 −15.425905 0.006110 −1.988470 −12.548667

where distance between the trajectories is defined by dz from two different val-
ues of initial conditions, λ1 is the largest Lyapunov exponent and γ is a parame-
ter accounting for the folding phenomena. For a chaotic system, the graph of dz
increases with a slope λ1 until it reaches a stationary value around which it normally
fluctuates. Figures 10 and 11 show that the trajectories starting from (X 1 , X 2 , X 3 )
and (X 1 + δ, X 2 + δ, X 3 + δ), (δ = 1e − 10) show fluctuating trend even for a very
long time interval. Here initial value considered is (X 1 = 0.065, X 2 = 0.465, X 3 =
0.150)). This implies that the system is unpredictable which may lead to chaos. These
figures correspond to range of h 33 ∈ (0.001, 1), h 55 ∈ (0.001, 1.5), refer to Fig. 10
and h 5,10 ∈ (0.001, 1.5) refer to Fig. 11 with time interval [0 : 500].
A Mathematical Model to Study Regulatory Properties … 97

Fig. 10 d∞ plots of the dynamical system Eq. (8) for the range of parameter h 33 ∈ (0.001, 1),
h 55 ∈ (0.001, 1.5), time step is [0, 500] and initial value are [0.067, 0.465, 0.150]

5 Second Pathway

In this section, we consider a pathway after removing the right branch of glycolysis
pathway shown in Fig. 1. In addition to this, we introduce a perturbation term in the
corresponding dynamical system. The perturbation has been considered in the form
of a periodic function f 0 cos(ωt). In terms of biochemistry, a perturbation refers to
addition or removal of any of the substrates/enzymes/metabolites in the pathway.
Various studies have been carried out using perturbations in mathematical modelling
to predict the behaviour of various biochemical systems. For instance, Vervekyo et
al. [46] and Kar et al. [18] studied the nonlinear dynamics of a glycolysis pathway.
Vervekyo et al. characterized the influence of periodic influx on the glycolytic oscil-
lations within the forced Selkov system and its chaotic behaviour [4]. Brechmann et
al. presented certain facts about the Selkov oscillator model by studying the proper-
ties of dynamics of various solutions of the model. Refer to [16, 17, 27] for similar
studies.
The dynamical system Eq. (2) takes the following form for this pathway.
98 S. Tomar et al.

Fig. 11 d∞ plots of the dynamical system Eq. (8) for the range of parameter h 5,10 ∈ (0.001, 1.5),
and h 52 ∈ (−0.6, −0.001), time step is [0, 500] and initial value are [0.067, 0.465, 0.150]

0.646h 5,10
d X2 0.646h 52 0.646h 55

= α2 X 2 1.714 X 5 1.741 X 10 1.714 − β2 X 2h 22 X 3h 23 X 8h 28 + f 0 cos(ωt),

dt
d X3
= α3 X 2h 22 X 3h 23 X 8h 28 − β3 X 3h 33 X 9h 39 . (13)
dt
By substituting the values of the desired parameters given in Tables 1 and 2, and the
independent variables given in Sect. 2, we have the following dynamical system from
Eq. (13).

d X2
= 0.585012402X 20.38h 52 50.37h 55 40.38h 5,10 − 0.107910016X 23.97 X 3−3.06 + f 0 cos(ωt),
dt
d X3
= 136X 23.97 X 3−3.06 − 3.0281922242X 3h 33 . (14)
dt

We further divide our investigation into two cases, namely when (a) f 0 = 0, and (b)
f 0 = 0. These two cases are separately investigated in Sects. 5.1 and 5.2, respectively.
A Mathematical Model to Study Regulatory Properties … 99

5.1 When Forcing Coefficient is Equal to Zero i.e. f0 = 0

Let us consider f 0 = 0, then, reduced dynamical system will be

d X2
= 0.585012402X 20.38h 52 50.37h 55 40.38h 5,10 − 0.107910016X 23.97 X 3−3.06 ,
dt
d X3
= 136X 23.97 X 3−3.06 − 3.0281922242X 3h 33 . (15)
dt

At steady state, we get two sets of equilibrium points (0, 0) and (E 1 , E 2 ) where
 1
0.38h 52 −3.97
136β2
E1 = ∗
α2 50.37h 55 40.38h 52
 3.97
3.06
0.38h 33 h 52 −3.97h 33 +1.1628h 52
2.86β3 α2 50.37h 55 40.38h 5,10 ) 0.38h52 −3.97
, (16)
136β2 136β2
 3.97
0.38h 52 −3.97
1
h 33 +3.06+3.06 0.38h3.97−3.97
136β2 136β2
E2 = .
52
(17)
2.86β3 α2 50.37h 55 40.38h 52

The second equilibrium point is further dependent on the four parameters h 55 , h 33 ,

h 5,10 and h 52 .
At the first equilibrium point, no noticeable feature worth reporting was observed
in the corresponding phase portrait. Hence, we consider (0.001, 0.001) which is very
close to (0, 0) for further investigation. According to the theory due to bifurcation
analysis, if two trajectories are moving towards each other then, equilibrium point
will be stable. But in case, if they show repulsive behaviour towards each other then,
equilibrium point will be unstable or a saddle point. Here, at point (0.001, 0.001),
trajectories are moving towards this point which means this point will be stable and
system is stable in the neighbourhood of this point. Similar behaviour is observed
for the second equilibrium point, shown in Fig. 12.
The Jacobian matrix corresponding to the dynamical system Eq. (15) is

A B
J= , (18)
C D

where

0.431936X 22.97 0.332928X 23.97

A = 0.2223(40.38h 5,10 )50.37h 55 h 52 X 20.38h 52 −1 − 3.06
, B= ,
X3 X 34.06
0.431936X 22.97 0.332928X 23.97
C= , D = − − 3.0316h 33 X 3h 33 −1 .
X 32.06 X 34.06
100 S. Tomar et al.

Fig. 12 Phase-portrait of
Dynamical system Eq. (13)
corresponding parameter
values are h 33 = 0.3,
h 55 = 0.84, h 52 = −0.1 and
h 5,10 = 1

Determinant of Jacobian matrix Eq. (18) is

1
|J | = − ((2.08e − 08)(3558168(40.38h 5,10 )(50.37h 55 )h 52 X 20.38h 52 X 23.97
X 2 X 34.06
− 62954672h 33 X 23.97 X 3h 33 + 32400225(40.38h 5,10 )(50.37h 55 )h 33 h 52 X 20.38h 52 X 3h 33 X 33.06 )). (19)

General form of the characteristic polynomial is

λ2 + (3.0037X 23.97 X 3−3.06 + 2.3152X 22.97 X 3−3.06 + 0.3299h 33 X 3h33−1 − 4.4984(40.38h 5,10 )
(0.38h 52 −1) (0.38h 52 −1)
(50.37h 55 )h 52 X 2 λ − 1.4838(40.38h 5,10 )(50.37h 55 )h 33 h 52 X 2 X 3h33−1
0.38h
+ 0.7637h 33 X 22.97 X 3h33−4.06 − 13.5117(40.38h 5,10 )(50.37h 55 )h 52 X 2 52 X 22.97 = 0, (20)

Stability of the system at the point (0.001, 0.001).

With respect to point i.e. (0.001, 0.001), the determinant Eq. (19) is

|J | = (2.4383e + 03)(0.001)h 33 h 33 − 137.8129(0.001)0.38h 52 40.38h 5,10 50.37h 55 h 52

− 6.7392 + 05(0.001)0.38h 52 (0.001)h 33 4.0.38h 5,10 50.37h 55 h 33 h 52 . (21)

In Fig. 13a, b, c, for range of parameters h 55 ∈ (0.001, 1.5), h 5,10 ∈ (0.001, 1.5),
h 33 ∈ (0.001, 1) and h 52 ∈ (−0.6, −0.001), it is shown that the determinant Eq. (21)
is positive i.e. |J | > 0. Therefore, the point (0.001, 0.001) is a stable point. In system
Eq. (13), if we take positive values of this parameter i.e. h 52 ∈ (0.001, 0.6), then
determinant will become negative implicating that the point (0.001, 0.001) will be
a saddle point for positive values of h 52 , refer to Fig. 13d.
Next, we investigate the behaviour of the dynamical system by using the eigenval-
ues of the Jacobian matrix Eq. (18). By putting the point (0.001, 0.001) in the general
A Mathematical Model to Study Regulatory Properties … 101

Fig. 13 Determinant of the Jacobian matrix Eq. (21) for range of h 33 ∈ (0.001, 1), h 55 ∈
(0.001, 1.5), h 52 ∈ (−0.6, −0.001) and h 5,10 ∈ (0.001, 1.5) with respect to initial value
(X 2 , X 3 ) = (0.001, 0.001) and for positive range of h 52 ∈ (0.001, 0.6) shows negative determi-
nant

form of characteristic polynomial Eq. (20), we get the characteristic polynomial as

follows

λ2 + (1.4242 + (3.0316e + 03)(0.001)h 33 h 33 − 222.3000(0.001)0.38h 52 40.38h 5,10 50.37h 55 )

h 52 )λ − ((6.7392e + 05)(0.001)0.38h 52 (0.001)h 33 40.38h 5,10 5).37h 55 h 33 h 52 + (2.4383e + 03)
(0.001)h 33 h 33 − 137.8129(0.001)0.38h 52 40.38h 5,10 50.37h 55 h 52 ) = 0, (22)

and corresponding eigenvalues are

1
λ1 = (111.1500(0.001)0.38h 52 40.38h 5,10 50.37h 55 )h 52 − 1.5158e + 03(0.001)h 33 h 33 −
2
(9.1906e + 06(0.001)2h 33 h 233 − (1.1178e + 03(0.01)h 33 h 33 − 81.9671(0.001)0.38h 52 40.38h 5,10
50.37h 55 h 52 + 4.9417e + 04(0.001)0.76h 52 40.76h 5,10 50.74h 55 h 252 + 1.3478e + 06(0.001)0.38h 52
(0.001)h 33 40.38h 5,10 50.37h 55 h 33 h 52 + 2.0285)1/2 − 0.7121, (23)
1
λ2 = (9.1906e + 06(0.001) h 33 − 1.1178e + 03(0.001) h 33 − 81.9671(0.001)
2h 33 2 h 33 0.38h 52
2
40.38h 5,10 50.37h 55 h 52 + 4.9417e + 04(0.001)0.76h 52 40.76h 5,10 50.74h 55 h 252 + 1.3478e + 06
102 S. Tomar et al.

(0.001)0.38h 52 (0.001)h 33 40.38h 5,10 50.37h 55 h 33 h 52 + 2.0285)1/2 − (1.5158e + 03(0.001)h 33 h 33

+ 111.1500(0.001)0.38h 52 40.38h 5,10 50.37h 55 h 52 − 0.7121. (24)

These two eigenvalues are further dependent upon the three major parameters
h 33 , h 55 and h 5,10 and inhibition parameter h 52 . If (λ1 , λ2 ) = (−, −) then, system
will be stable and if (λ1 , λ2 ) = (+, −) i.e. if any one eigenvalue is positive then
system will become unstable. For the range of h 33 ∈ (0.001, 1), h 55 ∈ (0.001, 1.5),
h 52 ∈ (−0.6, −0.001) and h 5,10 ∈ (0.001, 1.5), both eigenvalues are negative which
is shown in Fig. 14a–g in three dimensional graphs. But for h 52 ∈ (0.001, 0.6), one
eigenvalue is positive. Hence, system will become unstable for positive value of h 52
as shown in Fig. 14h.
Stability of the system at the point (E 1 , E 2 ).
Next, we investigate the behaviour of the dynamical system Eq. (15) at the second
point (E 1 , E 2 ). Recall that the equilibrium point (E 1 , E 2 ) is given by Eq. (16).
 0.38h 52 −3.97
1
136β2
E1 =
α2 50.37h 55 4 0.38h 52

 3.97
3.06
0.38h 33 h 52 −3.97h 33 +1.1628h 52
2.86β3 α2 50.37h 55 40.38h 5,10 ) 0.38h 52 −3.97

,
136β2 136β2
 3.97
0.38h 52 −3.97
1
h 33 +3.06+3.06 0.38h3.97−3.97
136β2 136β2
E2 = .
52

2.86β3 α2 50.37h 55 40.38h 52

Figure 15a–d show the surface plots for E 1 and E 2 for h 33 ∈ (0.001, 1), h 55 ∈
(0.001, 1.5), h 52 ∈ (−0.6, −0.001) and h 5,10 ∈ (0.001, 1.5). The equilibrium points
E 1 and E 2 for these range of these parameters remain positive.
With respect to this equilibrium point (E 1 , E 2 ), the Jacobian matrix for the dynam-
ical system Eq. (15) is evaluated as

|J | = −(46256184(40.38h 5,10 )50.37h 55 h 52 T E R M1 ∗ (T E R M2)0.38h 52

3.97
T E R M1 ∗ (T E R M2)3.97 − 818410736h 33 (((272(68/625) 0.38h 52 −3.97 )
397 1
/(7579((117(40.38h 52 )(50.37h 55 )/200) 38h 52 −397 ))) (h 33 +60741/(5000((0.38h 52 )−3.97))+153/50) ))h 33 T E R M1
(T E R M2)3.97 + 421202925(40.38h 5,10 )50.37h 55 h 33 h 52
397
(((272(68/625)397/100(0.38h 52 −3.97) )/(7579((117(40.38h 52 )50.37h 55 )/200) 38h 52 −397 )))
1 397
h 33 +60741/5000(0.38h 52 −3.97)+153/50
)153/50 (((272(68/625) 100(0.38h 52 )−3.97 )/(7579((117(40.38h 52 )50.37h 55 )
397 1
/200) (38h 52 −397) ))) (h 33 +60741/(5000((0.38h 52 )−3.97))+153./50) )h 33 T E R M1 ∗ (T E R M2)((0.38h 52 )) )/
397
(625000000(((272(68/625) 38h 52 −397 )/(7579((117(40.38h 52
A Mathematical Model to Study Regulatory Properties … 103

Fig. 14 Behaviour of eigenvalues for range of h 33 ∈ (0.001, 1), h 55 ∈ (0.001, 1.5), h 52 ∈

(−0.6, −0.001) and h 5,10 ∈ (0.001, 1.5) with respect to initial value (X 2 , X 3 ) = (0.001, 0.001)
and for positive range of h 52 ∈ (0.001, 0.6)
104 S. Tomar et al.

Fig. 15 Second equilibrium point (E 1 , E 2 ) always remains in positive quadrant for h 33 ∈

(0.001, 1), h 55 ∈ (0.001, 1.5), h 52 ∈ (−0.6, −0.001) and h 5,10 ∈ (0.001, 1.5)

397 1
50.37h 55 )/200) 38h 52 −397 )) (h 33 +60741/(1900h 52 −19850)+153./50) )(203/50) T E R M1(544/(2925
(40.38h 52 )50.37h 55 ))(1/((0.38h 52 )−3.97)) ), (25)

where
   3.97  1.1628h52 −3.97h
3.06
33 +0.38h 33 h 52
T E R M1 = 27.864 5.377(40.38h 5,10 )(50.37h 55 ) (0.38h52 )−3.97 ,
  1
T E R M2 = 0.1850/40.38h 52 50.37h 55 (0.38h52 )−3.97 .

In this case, we found that for the range of h 55 ∈ (0.001, 1.5) and h 33 ∈ (0.001, 1),
determinant of Jacobian matrix Eq. (18) remains positive. Hence, equilibrium point
is stable point which is also shown in Fig. 16a, b. For the range of h 5,10 and h 52 , the
determinant is positive. Hence, for these parameters equilibrium point is stable. But,
for sufficiently big positive values of the inhibition parameter, h 52 and the equilib-
rium point becomes saddle, which is shown in Fig. 16c, d. As shown in Fig. 16d, the
positive value of h 52 leads to instability in the system. It means that the pathway
becomes activating instead of inhibiting. This further implies that, when G-6-P con-
centration is high, it will positively regulate (increase) its concentration instead of
A Mathematical Model to Study Regulatory Properties … 105

Fig. 16 Behaviour of determinant of Jacobian matrix Eq. (18) with respect to second equilibrium
point (E 1 , E 2 ), parameters range are h 55 ∈ (0.001, 1.5), h 33 ∈ (0.001, 1), h 52 ∈ (−0.6, −0.001)
and h 5,10 ∈ (0.001, 1.5). Figure c, d shows the behaviour for the negative range of h 52 and positive
range of inhibition parameter h 52 , (0.001, 0.6)

negatively regulating (decrease) itself. This is a deviation of the cell or system from
the homoeostatic behaviour and will lead to instability.
We further investigate the behaviour of dynamical system Eq. (15) by using the
eigenvalues of the Jacobian matrix Eq. 18. For the defined range of parameters,
Tables 4 and 5, show negative eigenvalues which implies that this system will be
stable. These eigenvalues are plotted in Fig. 17a–d.

5.2 When Forcing Coefficient is Not Equal to Zero i.e. f0  = 0

In this section, we consider the dynamical system Eq. (14), when f 0 = 0 i.e. pertur-
bation f 0 cos(ωt) is introduced in this system. Our motivation here is to investigate
the effect of this external perturbation onto the dynamical system, separately for
the forcing coefficient f 0 and the frequency ω. Hence, we consider only one point
(0.001, 0.001) for further investigation in this section.
106 S. Tomar et al.

Table 4 Eigenvalues of Jacobian matrix Eq. (18) for second equilibrium point (E 1 , E 2 ), with
respect to h 33 and h 55 parameters
h 33 λ1 λ2 h 55 λ1 λ2
0.1 −5.9148 −3.2721e + 0.001 −2.5297 −7.4918e +
10 04
0.2 −1.6451 −1.3425e + 0.1 −2.2920 −6.6279e +
06 04
0.3 −1.0944 −2.6565e + 0.5 −1.5376 −4.0420e +
04 04
0.45 −0.8788 −1.6614e + 0.8 −1.1391 −2.7908e +
03 04
0.6 −0.8237 −399.9017 1 −0.9324 −2.1808e +
04
0.8 −0.8207 −136.202 1.2 −0.7631 −1.7044e +
04
1 −0.8470 −71.4937 1.5 −0.5647 −1.1782e +
04

Table 5 Eigenvalues of Jacobian matrix Eq. (18) for second equilibrium point (E 1 , E 2 ) with respect
to h 52 and h 5,10 parameters
h 52 λ1 λ2 h 5,10 λ1 λ2
−0.6 −1.6905 −1.497e + 04 0.001 −2.8466 −267.3779
−0.5 −1.5647 −1.6434e + 0.1 −2.6290 −414.3657
04
−0.4 −1.4427 −1.8193e + 0.5 −1.8208 −2.5656e +
04 03
−0.3 −1.3241 −2.0350e + 0.8 −1.3469 −1.0379e +
04 04
−0.2 −1.2083 −2.3059e + 1 −1.0944 −2.6565e +
04 04
−0.1 −1.0944 −2.6565e + 1.2 −0.8863 −6.8302e +
04 04
−0.001 −0.9825 −3.1216e + 1.5 −0.6433 −2.8342e +
04 05

For the range of forcing coefficient f 0 ∈ (0.001, 0.8), the system shows the peri-
odic behaviour for a fixed value of other parameters ω = 0.4, h 33 = 0.3, h 5,10 = 1,
h 52 = −0.1 and h 55 = 0.84. The periodic behaviour is shown in Fig. 18a–d. Corre-
sponding time series plots are shown in Fig. 18e–h. For ω, the system shows oscillat-
ing behaviour for ω ∈ (0.001, 0.8) and f 0 = 0.4 (rest of the parameters are same as
in Fig. 18), behaviour and time series plots are shown in Fig. 19. Periodic behaviour
(scaled and zoomed) with respect to both parameters f 0 , and ω has been shown as
Fig. 20. The impact of the presence of the perturbation term in a dynamical system
A Mathematical Model to Study Regulatory Properties … 107

Fig. 17 Eigenvalues of Jacobian matrix Eq. (18) with respect to second equilibrium point
(E 1 , E 2 ), parameters range are h 55 ∈ (0.001, 1.5), h 33 ∈ (0.001, 1), h 52 ∈ (−0.6, −0.001) and
h 5,10 ∈ (0.001, 1.5)

is shown in Fig. 21a and b with the help of Lyapunov exponents for a chosen values
of the parameters.
By perturbing the system through mathematical modelling, we have observed that
the concentrations of X 2 (Glucose-6-P) and X 3 (Fructose-6-P) show oscillations with
respect to time, i.e the concentration of both the metabolites first increases and then
decreases. It has already been shown that Phosphofructokinase (PFK) is the control
enzyme responsible for the oscillations. In [14], the authors have demonstrated that
oscillations can only be achieved by infusing one of those metabolites pre-cursing
PFK.

6 Conclusion

In the present study, few initial committed steps of the glycolytic pathway were
formulated into a dynamical system with the help of mass balance equations and
S-system formulation. This dynamical system consists various parameters which
involve rate constants, exponents and dependent and independent variables. We stud-
108 S. Tomar et al.

Fig. 18 Effect of forcing term f 0 on dynamical system Eq. (15) with f 0 = [0.001, 0.1, 0.4, 0.8]
and rest parameters values are h 33 = 0.3, h 55 = 0.84, h 5,10 = 1, ω = 0.4 and h 52 = −0.1. Time
interval is 0 : 300 and initial values are (0.001, 0.001)

ied the interrelation between few of the important parameters by setting the system
at steady state which eventually lead to solving a set of transcendental equations. We
identified four parameters, namely, h 33 , h 55 , h 5,10 , and h 52 which relate the dynam-
ical system to the phosphofructokinase and glucokinase reactions, respectively. We
investigated the stability of this dynamical system with the help of the bifurcation
tools such as the phase portraits, the Jacobian stability, Lyapunov exponents, d∞ plots
and the time series plots for the parameters h 33 , h 55 , h 5,10 , and h 52 . The equilibrium
point (0.067, 0.465, 0.150) was considered for this investigation. These values are
the steady state concentration of the dependent variable X 1 , X 2 , and X 3 . We found
that h 33 is the only parameter which may disturb the stability of the system with
respect to time and rest three parameters can contribute to an increase in the concen-
tration of the fructose 1,6-bi-phosphate.
The second pathway considered in this study is by deleting the right-side branch
of the original pathway, or by shrugging off the contribution of glucose-1-P from the
pathway. We further added an external perturbation term f 0 cos(ωt) in it to study the
effect of forcing coefficient f 0 , and the frequency ω. We have considered two cases
when f 0 = 0, and f 0 = 0.
A Mathematical Model to Study Regulatory Properties … 109

Fig. 19 Effect of frequency ω on the dynamical system Eq. (15) ω = [0.001, 0.01, 0.2, 0.8] and
rest parameters are h 55 = 0.84, h 33 = 0.3, h 5,10 = 1, f 0 = 0.4 and h 52 = −0.1, time interval is
[0 : 300] and initial values are (0.001, 0.001)

Fig. 20 Scaled and Zoomed view of effect of frequency ω = 0.2, forcing coefficient f 0 = 0.4 and
rest parameters are h 55 = 0.84, h 33 = 0.3, h 5,10 = 1 and h 52 = −0.1, time interval is [0 : 300] and
initial values are (0.001, 0.001)

• For the case of f 0 = 0, we investigated the stability of the system around two points
(0.001, 0.001), and (E 1 , E 2 ). For both the points, the system is found to be stable
for the defined range of the parameters. However, we found that the parameter
h 52 can significantly affect the stability of the system. In fact, beyond a critical
value of h 52 system became unstable and rather than inhibiting the production of
glucose-6-P, it triggers activation.
110 S. Tomar et al.

Fig. 21 Comparison of Lyapunov exponents when f 0 = 0 and f 0 = 0, rest parameters are h 55 =

0.84, h 33 = 0.3, h 5,10 = 1, f 0 = 0.4, ω = 0.8 and h 52 = −0.1, time interval is 0 : 0.5 : 100 and
initial values are (0.001, 0.001)

• For the second case, when f 0 = 0, the behaviour corresponding to rest of the
parameters h 33 , h 55 , h 5,10 , and h 52 was observed similar to the first case. But in
presence of the forcing term with forcing coefficient f 0 , and the frequency ω. The
system showed periodic behaviour with oscillations. The results for both cases
were also verified with the help of the Lyapunov exponents.

7 Future Work

• This glycolytic pathway has various independent variables, and parameters, there-
fore, it is impossible to see the effect of range of all these parameters. In future,
another set of parameters can be chosen based on the experimental justification and
interest, and the effect of these parameters can be investigated on the glycolytic
pathway.
• The values of the parameters and independent variables can be perturbed and/or
additional perturbation term can be added in this system to study the behaviour of
system with respect to these perturbation terms.
• In the second pathway, we have removed only right branch of the glycolytic path-
way. The glycolytic pathway can be investigated by adding or removing more
branches based on theory in future studies.
• In this study, we have studied first few committed of the complete glycolytic
pathway. This study can be expanded for a complete pathway or one can study the
dynamical behaviour of other metabolic pathways by using standard tools from
bifurcation tools as demonstrated in this study.
A Mathematical Model to Study Regulatory Properties … 111

Appendix 1: S-System Study of Pathway

The mass-balance equation can be translated into S-system,

d X1 g g g
= α1 X 4 14 X 6 16 X 111,11 − β1 X 1h 11 X 2h 12 X 7h 17 ,
dt
d X2 g g g g g
= α2 X 1 21 X 2 22 X 5 25 X 7 27 X 102,10 − β2 X 2h 22 X 3h 23 X 8h 28 ,
dt
d X3 g g g
= α3 X 2 32 X 3 33 X 8 38 − β3 X 3h 33 X 9h 39 . (26)
dt
There is one rate constant for each pool’s production and one rate constant for
its degradation in this S-system. These are α1 , α2 , α3 and β1 , β2 , β3 , respectively.
For each X i s, one subscript is used for each power. Power is represented by g in the
production term and h in the degradation term. The parameters αi and gi j always refer
to production or gain, whereas the parameters βi and h i j always refer to degradation
or loss, i.e. the greater αi s indicate more production and the larger βi s imply a loss
in production.
With the help of the precursor-product relationship between X 2 & X 3 and branch
point constraint at X 2 , we can reduce the number of unknowns.
Because V2− and V3+ reflect the same process, the precursor-product relationship
implies that the values of the constraints’ parameters be equal.

α3 = β2 , g32 = h 22 , g33 = h 23 , g38 = h 28 . (27)

According to the branch point restriction at X 2 , the total of the two fluxes entering
X 2 from X 1 and X 5 , namely V1− and V5− , must equal the influx V2+ . The constraint
on branch points can be written as

V2+ = V1− + V5− , i.e.

g g g g g h
α2 X 1 21 X 2 22 X 5 25 X 7 27 X 102,10 = β1 X 1h 11 X 2h 12 X 7h 17 + β5 X 2h 52 X 5h 55 X 105,10 . (28)

These expressions can be understood as a power-law representation in this case.

We may use the partial differentiation of the left-hand side of the equation to get the
power terms g21 , g22 , g25 , g27 , and g2,10 .

∂ V2+ X 7
g27 =
∂ X 7 V2+
h
∂(β1 X 1h 11 X 2h 12 X 7h 17 + β5 X 2h 52 X 5h 55 X 105,10 ) X 7
= . (29)
∂ X7 V2+
112 S. Tomar et al.

By doing partial differentiation, the first term is

V−
g27 = β1 h 17 X 1h 11 X 2h 12 X 7h 17 −1 VX+7 , which is equivalent to h 17 V1+ and the second term of
2 2
the equation will become zero. Thus,
V−
g27 = h 17 V1+ . Similarly, we can compute the other ones i.e.
2
V− V− V−
g21 = h 11 V1+ , g25 = h 55 V5+ , g2,10 = h 5,10 V5+ . The last one is g22 whose derivation
2 2 2
may be difficult because of both the terms V5− and V1− depend upon X 2 .

∂ V2+ X 2
g22 = ,
∂ X 2 V2+
h
∂(β1 X 1h 11 X 2h 12 X 7h 17 + β5 X 2h 52 X 5h 55 X 105,10 ) X 2
= ,
∂ X2 V2+
X7
= β1 h 12 X 1h 11 X 2h 12 −1 X 7h 17 + β5 h 52 X 2h 52 −1 X 5h 55 X 105,10
h
,
V2+
h
β1 X 1h 11 X 2h 12 X 7h 17 β5 X 2h 52 X 5h 55 X 105,10
= h 12 + h ,
V2+ V2+
52

V1− V5−
= h 12 + h 52 + . (30)
V2+ V2

Appendix 2: Steady State Analysis of the Dynamical System

Eq. (8)

We require equations that characterise the steady state in an explicit algebraic form for
more theoretical investigations, such as a general analysis of logarithmic sensitivities.
The structure and characteristics of these equations can be investigated. To find the
steady state, divide each equation by its rate constant αi and set the three symbolic
equations to zero. The result is

g14 g16 β1 h 11 h 12 h 17
X4 X6 −X X X = 0,
α1 1 2 7
g g g g β2 h22 h23 h28
X 1 21 X 2 22 X 5 25 X 7 27 X 1 0g2,10 − X X 3 X 8 = 0,
α2 2
g g g β3 h33 h39
X 2 32 X 3 33 X 8 38 − X X 9 = 0. (31)
α3 3
 
Next, we define yi = log X i for i = 1, 2, . . . , 10 and bi = log αβii for i = 1, 2, 3.
g
Power term can be expressed as X i i j = exp(gi j log X i ), and taking log to all terms,
A Mathematical Model to Study Regulatory Properties … 113

we obtain

b1 + h 11 y1 + h 12 y2 + h 17 y7 − g14 y4 + g16 y6 = 0,
b2 + h 22 y2 + h 23 y3 + h 28 y8 − g21 y4 − g22 y2 − g25 y5 − g27 y7 − g2,10 y10 ,
b3 + h 33 y3 + h 39 y9 − g32 y2 − g33 y3 − g38 y8 = 0. (32)

By solving the system of linear equations (32), we will get two solution sets. One
is with respect to the parameters h 55 , h 52 , h 33 , h 5,10 , αi and βi for i = 1, 2, 3. While
putting the value of the parameters from Sect. 2, Table 1. The value of the variables
yi will be log(X i ), all X i are from Sect. 2, Table 2. Corresponding solutions set for
this system Eq. (32) is

h 33 = h 33 , h 52 = h 52 , h 55 = h 55 , h 5,10 = h 5,10 ,
α1 = 0.8333333333e − 2 exp(−1.530000000y1 + 0.5900000000y2 + 1.519706161)β1 ,
α2 = exp(0.5129289136h 5,10 + 0.5954920276h 55 − 0.3700000000y2 h 52 + 0.9486000000y1
− 3.604200000y2 + 3.060000000y3 − 2.625549624)β2 ,
α3 = α3 , β1 = β1 , β2 = β2 ,
β3 = 0.7352941176e − 2 exp(h 33 y3 − 3.970000000y2 + 3.060000000y3 + 1.050821625)α3 . (33)

The second solution set is generated only for Y1 , Y2 , Y3 , in which only αi , βi and
h 33 , h 55 , h 52 and h 5,10 parameters are unknown. Solution set is

1
Y1 = − ((2.042483660e + 06)
(46250.h 52 h 33 + (4.04800e + 05)h 33 + (1.41525e + 05)h 52 − (2.79837e + 05))
((−1.405361185e + 10)h 33 h 55 − (1.210512236e + 10)h 33 h 5,10 + (1.441680000e + 11)
α α1
log 1 h 33 + (4.836322660e + 10)h 52 h 33 − (2.355910293e + 11) + (4.528800000e + 10)h 52 log
β1 β1
α α α3
+ (1.480000000e + 10) log 1 h 33 h 52 − (4.477392000e + 10) log 1 + (7.221600000e + 10) log +
β1 β1 β3
α α
(7.221600000e + 10) log 2 + (5.330730828e + 11)h 33 + (2.360000000e + 10)h 33 log 2
β2 β2
+ (1.479914734e + 11)h 52 − (4.300405227e + 10)h 55 − (3.704167443e + 10)h 5,10 )), (34)

1
Y2 = − ((3.125000000e + 04)
(46250h 52 h 33 + (4.04800e + 05)h 33 + (1.41525e + 05)h 52 − (2.79837e + 05))
((−2.381968110e + 08)h 33 h 55 − (2.051715654e + 08)h 33 h 5,10 + (1.860630674e + 09)h 33 +
(α2 )
(4.00000000e + 08)h 33 log − (7.288822418e + 08)h 55 − (6.2782499033 + 08)h 5,10 +
β2
α1 α2
(2.48000000e + 08) log h 33 + (1.224000000e + 09) log + (9.666459509e + 08)+
β1 β2
α1 α3
(7.58880000e + 08) log + (1.224000000e + 09) log )), (35)
β1 β3
114 S. Tomar et al.

1
Y3 = (0.6250000000e − 5
(46250h 52 h 33 + (4.04800e + 05)h 33 + (1.41525e + 05)h 52 − (2.79837e + 05))
((2.857756613e + 10)h 52 + (4.728206699e + 10)h 55 + (4.072655574e + 10)h 5,10 − (7.400000000e + 09)
(α3 α2
h 52 log − (1.192119721e + 11) − (7.940000000e + 10) log − (4.922800000e + 10)
β3 β2
α1 α3
log − (6.476800000e + 10) log )). (36)
β1 β3

References

1. H.D.I. Abarbanel, R. Brown, M.B. Kennel, Lyapunov exponents in chaotic systems: their
importance and their evaluation using observed data. Int. J. Modern Phys. B 5(09), 1347–1375
(1991)
2. B.M. Bakker, P.A.M. Michels, F.R. Opperdoes, H.V. Westerhoff, Glycolysis in bloodstream
form Trypanosoma brucei can be understood in terms of the kinetics of the glycolytic enzymes.
J. Biol. Chem. 272(6), 3207–3215 (1997)
3. J.A. Bassham, G.H. Krause, Free energy changes and metabolic regulation in steady-state
photosynthetic carbon reduction. Biochimica et Biophysica Acta (BBA)-Bioenergetics 189(2),
207–221 (1969)
4. P. Brechmann, A.D. Rendall, Dynamics of the Selkov oscillator. Math. Biosci. 306, 152–159
(2018)
5. A. Buscarino, L. Fortuna, M. Frasca, Essentials of Nonlinear Circuit Dynamics with MATLAB®
and Laboratory Experiments (CRC Press, 2017)
6. E.T. Camacho, D. Brager, G. Elachouri, T. Korneyeva, G. Millet-Puel, J.-A. Sahel, T. Léveillard,
A mathematical analysis of aerobic glycolysis triggered by glucose uptake in cones. Sci. Rep.
9(1), 1–18 (2019)
7. M. Cascante, N.V. Torres, R. Franco, E. Meléndez-Hevia, E.I. Canela, Control analysis of
transition times. Mol. Cell. Biochem. 101(1), 83–91 (1991)
8. R. Chaudhry, M. Varacallo, Biochemistry, glycolysis (2018)
9. D.R. Curtiss, Recent extentions of Descartes’ rule of signs. Ann. Math. 251–278 (1918)
10. H. Euler, E. Adler, Uber die Komponenten der Dehydrasesysteme. IV. Hoppe-Seyler’s
Zeitschrift fur Physiologische Chemie 235, 122–173 (1935)
11. A.K. Groen, R. Van Der Meer, H.V. Westerhoff, R.J.A. Wanders, T.P.M. Akerboom, J.M. Tager,
H. Sies, Metabolic Compartmentation (Academic, New York, 1982), p. 9
12. R. Heinrich, S.M. Rapoport, T.A. Rapoport, Metabolic regulation and mathematical models.
Progress Biophys. Mol. Biol. 1–82 (1978)
13. H.G. Hers, L. Hue, Gluconeogenesis and related aspects of glycolysis. Ann. Rev. Biochem.
52(1), 617–653 (1983)
14. B. Hess, A. Boiteux, J. Krüger, Cooperation of glycolytic enzymes. Adv. Enzyme Regul. 7,
149–167 (1969). PMID: 4244004. https://doi.org/10.1016/0065-2571(69)90016-8
15. R.C. Hilborn, Chaos and Nonlinear Dynamics: An Introduction for Scientists and Engineers
(Oxford University Press, 2000)
16. A. Jeong, G. Fiorito, P. Keski-Rahkonen, M. Imboden, A. Kiss, N. Robinot, H. Gmuender, J.
Vlaanderen, R. Vermeulen, S. Kyrtopoulos, Z. Herceg, A. Ghantous, G. Lovison, C. Galassi, A.
Ranzi, V. Krogh, S. Grioni, C. Agnoli, C. Sacerdote, N. Mostafavi, A. Naccarati, A. Scalbert,
P. Vineis, N. Probst-Hensch; EXPOsOMICS Consortium, Perturbation of metabolic pathways
mediates the association of air pollutants with asthma and cardiovascular diseases. Environ.
Int. 119, 334–345 (2018). (PMID: 29990954)
17. C. Jin, X. Zhu, H. Wu, Y. Wang, X. Hu, Perturbation of phosphoglycerate kinase 1 (PGK1)
only marginally affects glycolysis in cancer cells. J. Biol. Chem. 295(19), 6425–6446 (2020)
A Mathematical Model to Study Regulatory Properties … 115

18. S. Kar, D.S. Ray, Nonlinear dynamics of glycolysis. Modern Phys. Lett. B 18(14), 653–678
(2004)
19. S. Lenzen, A fresh view of glycolysis and glucokinase regulation: history and current status.
J. Biol. Chem. 289(18), 12189–94 (2014)
20. J.C. Liao, Modelling and analysis of metabolic pathways. Curr. Opin. Biotechnol. 4(2), 211–
216 (1993)
21. S. Lynch, Dynamical Systems with Applications Using MATLAB (Birkhäuser, Boston, 2004)
22. G. Maria, In silico determination of some conditions leading to glycolytic oscillationsand their
interference with some other processes in E. coli cells. Front. Chem. 977 (2020)
23. C. McCann, Bifurcation Analysis of Non-linear Differential Equations (University of LiverPool,
2013)
24. B.C. Mulukutla, A. Yongky, S. Grimm, P. Daoutidis, W.-S. Hu, Multiplicity of steady states in
glycolysis and shift of metabolic state in cultured mammalian cells. PLoS One 10(3), e0121561
(2015)
25. D.L. Nelson, M.M. Cox, Lehninger Principles of Biochemistry, 6th edn
26. E.A. Newsholme, Regulation in Metabolism (Wiley, 1973)
27. G. Overal, B. Teusink, F.J. Bruggeman, J. Hulshof, R. Planqué, Bifurcation analysis of
metabolic pathways: an illustration from yeast glycolysis. bioRxiv, 163600 (2017)
28. M.A. Savageau, E.O. Voit, Recasting nonlinear differential equations as S-systems: a canonical
nonlinear form. Math. Biosci. 87(1), 83–115 (1987)
29. M.A. Savageau, Power-law formalism: a canonical nonlinear approach to modeling and analy-
sis, in Proceedings of the First World Congress on World Congress of Nonlinear Analysts’ 92,
vol. IV (1995), pp. 3323–3334
30. M.A. Savageau, Allometric morphogenesis of complex systems: derivation of the basic equa-
tions from first principles. Proc. Natl. Acad. Sci. 76(12), 6023–6025 (1979)
31. M.A. Savageau, Growth of complex systems can be related to the properties of their underlying
determinants. Proc. Natl. Acad. Sci. 76(11), 5413–5417 (1979)
32. W. Schwartz, H.R. Mahler, E.H. Cordes, Biological chemistry. XV und 872 S., 236 Abb., 96
Tab. Evanston-London 1966: Harper and Row Ltd. 90 s (1969), pp. 410–411
33. E. van Schaftingen, D.R. Davies, H.G. Hers, Fructose-2,6-bisphosphatase from rat liver. Eur.
J. Biochem. 124(1), 143–9 (1982)
34. M.C. Scrutton, M.F. Utter, The regulation of glycolysis and gluconeogenesis in animal tissues.
Annu. Rev. Biochem. 37, 249–302 (1968)
35. E.E. Selkov, J.G. Reich, Energy Metabolism of the Cell (Academic, 1981)
36. H. Singh, H. Srivastava, D. Baleanu, Methods of Mathematical Modelling: Infectious Disease
(Academic, 2022)
37. H. Singh, D. Baleanu, J. Singh, H. Dutta, Computational study of fractional order smoking
model. Chaos Solitons & Fractals 142, 110440 (2021)
38. H. Singh, Analysis of drug treatment of the fractional HIV infection model of CD4+ T-cells.
Chaos Solitons & Fractals 146, 110868 (2021)
39. H. Singh, Analysis for fractional dynamics of Ebola virus model. Chaos Solitons & Fractals
138, 109992 (2020)
40. M.W. Slein, G.T. Cori, C.F. Cori, A comparative study of hexokinase from yeast and animal
tissues. J. Biol. Chem. 186(2), 763–780 (1950)
41. S.H. Strogatz, Nonlinear Dynamics and Chaos: With Applications to Physics, Biology, Chem-
istry, and Engineering (CRC Press, 2018)
42. N.V. Torres, F. Mateo, J.M. Riol-Cimas, E. Meléndez-Hevia, Control of glycolysis in rat liver by
glucokinase and phosphofructokinase: influence of glucose concentration. Mol. Cell. Biochem.
93(1), 21–26 (1990)
43. N.V. Torres, F. Mateo, E. Melendez-Hevia, H. Kacser, Kinetics of metabolic pathways. A
system in vitro to study the control of flux. Biochem. J. 234(1), 169–174 (1986)
44. N.V. Torres, Modelization and experimental studies on the control of the glycolytic-
glycogenolytic pathway in rat liver. Mol. Cell. Biochem. 132, 117–126 (1994)
116 S. Tomar et al.

45. L.W.M. Verburg, Modeling and control, of glycolysis in trypanosoma brucei. Dissertation
Master Thesis, Department of Mathematics, Vrije Universiteit, Amsterdam, 2006
46. D.V. Verveyko, A.Y. Verisokin, E.B. Postnikov, Mathematical model of chaotic oscillations
and oscillatory entrainment in glycolysis originated from periodic substrate supply. Chaos:
Interdiscip. J. Nonlinear Sci. 27(8), 083104 (2017)
47. E.O. Voit, The best models of metabolism. Wiley Interdiscip. Rev.: Syst. Biol. Med. 9(6), e1391
(2017)
48. E.O. Voit, (Case Study-4) Computational Analysis of Biochemical Systems: A Practical Guide
for Biochemists and Molecular Biologists (Cambridge University Press, 2000)
49. X. Wang, Bifurcation analysis of two biological systems: a tritrophic food chain model and an
oscillating networks model. Doctoral dissertation, The University of Western Ontario, Canada
2018
50. O. Warburg, W. Christian, Uber ein neues Oxyda-tionsferment und sein Absorptionsspektrum.
Biochem. Z. 254, 438–458 (1932)
51. P.O. Westermark, A. Lansner, A model of phosphofructokinase and glycolytic oscillations in
the pancreatic β-cell. Biophys. J. 85(1), 126–139 (2003)
On Solutions of Biological Models Using
Reproducing Kernel Hilbert Space
Method

Nourhane Attia and Ali Akgül

Abstract Differential equations (DEs, for short) are becoming more and more indis-
pensable for modeling real-life problems. Modeling and then analyzing these DEs
help scientists to understand and make predictions about the system that they want
to analyze. And this is possible only in one case when their solutions are available.
However, the majority of fractional differential equations lack exact solutions. This
chapter’s goal is to introduce the reproducing kernel Hilbert space method (RKHSM)
for some systems that have significant applications in biology. The proposed method’s
error estimations and convergence analyses are discussed. The assessment of the
RKHSM is made by testing some illustrative applications. The results suggest that
the RKHSM is an efficient and highly convenient method to solve the fractional
systems arising in biology.

Keywords Computational biology · Reproducing kernel Hilbert space method ·

Numerical solution

AMS (2010) Subject Classifications 46E22 · 34A08

N. Attia
National High School for Marine Sciences and Coastal (ENSSMAL), Dely Ibrahim University
Campus, Bois des Cars, B.P. 19, 16320 Algiers, Algeria
e-mail: [email protected]
A. Akgül (B)
Department of Computer Science and Mathematics, Lebanese American University, Beirut,
Lebanon
e-mail: [email protected]
Department of Mathematics, Arts and Science Faculty, Siirt University, 56100 Siirt, Turkey
Department of Mathematics, Mathematics Research Center, Near East University, Near East
Boulevard, PC: 99138 Nicosia/Mersin 10, Turkey

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 117
H. Singh and H. Dutta (eds.), Computational Methods for Biological Models,
Studies in Computational Intelligence 1109,
https://doi.org/10.1007/978-981-99-5001-0_5
118 N. Attia and A. Akgül

1 Introduction

PDEs serve as a crucial link between mathematics and its applications. Numerous
biological phenomena have been described using PDEs, and their solutions are essen-
tial to comprehending nature. However, especially for the non-linear equation, it is
quite challenging to find exact solutions. One of the main tasks for researchers now
is to create analytical and numerical approaches to address challenging nonlinear
problems [15–18].
For nonlinear PDEs, only a few basic numerical methods are employed. This offers
us the inspiration to propose a method for these problems. The primary contribution
of this research is the development of the RKHSM in order to numerically solve the
nonlinear Newell–Whitehead–Segel equations (NWSE) [20] that appear in biology.
The RKHSM is employed in this chapter to create numerical solutions for NWSEs.
The dynamical behavior near the bifurcation point for the Reyleigh–Benard convec-
tion of binary fluid mixtures is described by these equations that can be found in
mathematical biology.
An efficient numerical approach for difficult nonlinear problems without dis-
cretization is the RKHSM, which was first presented in 1908 [2].
The RKHSM, which was proposed in 1908 [2] has been used to resolve a vari-
ety of equation forms, including Riccati differential equations [7], forced Duffing
equations [10], generalized Schamel equation [3], fuzzy fractional IVPs [13], nonlo-
cal fractional boundary value problems [9], second-order PDEs [5], and many other
interesting studies [4, 8, 14, 19, 21].
The main benefits of the RKHSM are:
1. how simple it is to apply, especially since it is mesh-free.
2. its capacity to solve a variety of difficult differential equations.
The section that follows contains some fundamental RK theory definitions and the-
orems. In our general section (Sect. 3), The RKHSM’s explanation and application
to the suggested problem are presented. Error estimates are provided in the fourth
section. In the fifth section, two applications are used to validate the RKHSM’s effi-
cacy. The sixth section contains a discussion, graphics, and tables. The conclusion
is presented in the end.

2 Useful Mathematical Concepts

The RK theory contains the following basic definitions and theorems. In order to
start, we list a few RKHSs that are crucial for our study and are found in numerous
articles (see, e.g. [1, 12]). The reproducing kernel function (RKF) expressions are
then derived. For “completely continuous” and “absolutely continuous function(s),”
we used the acronyms “CC” and “ACF(s)” throughout the text.
On Solutions of Biological Models Using Reproducing … 119

Definition 2.1 If
1. (·, ϑ) ∈ J , ∀ ϑ ∈ ℵ,
2. f, (·, ϑ) = f(ϑ), ∀ ∈ J and ∀ ϑ ∈ ℵ,
a function  : ℵ × ℵ → C is said to be a reproducing kernel of the space H given
with J over ℵ = ∅ is a Hilbert space.

Remarks
• A Hilbert space that has a RK is a RKHS.
• The second assertion in the previous definition is a reproducing property (for short,
RP).

Definition 2.2 ([11]) The space of functions f where f(j ) , j = 0, 1, . . . , m − 1 are

ACFs on [a, L] and f(m) ∈ L 2 [a, L] is known as the function space G m 2 [a, L].

2 [a, L] are
Definition 2.3 ([11]) The inner product and norm of G m


m−1    L m m
dı f  dı g  d f d g
f, gG m =   + m dκ m dx, f, g ∈ G m
2 [a, L],
2 dκ ı κ=a dκ ı κ=a a dκ
ı=0

and

f Gm
2
= f, fG m2 , f ∈ Gm
2 [a, L].

As examples,
1. The space N32 [a, L] :

N32 [a, L] = {f| f, f and f are ACFs on [a, L], f(3) ∈ L 2 , and f(a) = f(L) = 0}.

2. The space W22 [a, L] :

W22 [a, L] = {f| f and f are ACFs on [a, L], f ∈ L 2 , and f(a) = 0}.

3. The space W12 [a, L] :

W12 [a, L] = {f| f is ACF on [a, L] and f ∈ L 2 }.

We use I = [a, L] × [0, T] ⊂ R × R+ throughout the entire research.

Definition 2.4 ([11]) The space of functions Z(κ, ϑ) where ∂κ∂m−1 ∂ϑn−1 Z(κ, ϑ) is
m+n−2

CC in I and ∂κ∂ m ∂ϑn Z(κ, ϑ) ∈ L 2 (I) is known as the space W(3,2)

m+n
2 (I) .
120 N. Attia and A. Akgül

Definition 2.5 ([11]) The inner product and norm of W(3,2)

2 (I) are

  T  ∂n ∂ı
m−1
∂n ∂ı
   ∂ı
n−1
∂ı
Z, Q (3,2) = Z(a, ϑ) Q(a, ϑ) dϑ + ı Z(κ, 0), ∂ϑı Q(κ, 0) m
W2 ∂ϑ n ∂κ ı ∂ϑ n ∂κ ı ∂ϑ G2
ı=0 0 ı=0
 T L m
∂ ∂n ∂m ∂n (3,2)
+ m ∂ϑn Z(κ, ϑ) ∂κ m ∂ϑn Q(κ, ϑ)dκdϑ, Z, Q ∈ W2 (I) ,
0 a ∂x

and

Z W(3,2) = Z, ZW(3,2) , Z ∈ W(3,2)
2 (I) .
2 2

As examples,
1. The space W(3,2)
2 (I) :

∂3 ∂5
W(3,2) (I) = {Z(κ, ϑ)| Z(κ, ϑ) is CC in I, Z(κ, ϑ) ∈ L 2 (I),
2
∂κ 2 ∂ϑ ∂κ 3 ∂ϑ2
and Z(a, ϑ) = Z(L, ϑ) = Z(κ, 0) = 0}.

2. The space W(1,1)

2 (I) :

∂2
W(1,1) (I) = {Z(κ, ϑ)| Z(κ, ϑ) is CC in I and Z(κ, ϑ) ∈ L 2 (I)}.
2
∂κ∂ϑ

Theorem 2.1 From the space W12 [a, L], we derive the RK function Mμ (κ) as [11]:

−a + 1 + κ , κ ≤ μ,
Mμ (κ) = (1)
μ − a + 1 , κ > μ.

Proof We must prove

f, Mμ W12
= f(μ).

We have
 L
f, Mμ W12
= f(a)Mμ (a) + f (κ)Mμ (κ) dκ.
a

We have to calculate Mμ (a) :

Mμ (a) = 1.
On Solutions of Biological Models Using Reproducing … 121

By applying the aforementioned equation, we get:

 L
f, Mμ W12
= f(a) + f (κ)Mμ (κ) dκ.
a

We have

1 , κ ≤ μ,
Mμ (κ) =
0 , κ > μ.

Thus

 μ  L
f, Mμ W12
= f(a) + f (κ)Mμ (κ) dκ + f (κ)Mμ (κ) dκ
a μ

so, we deduce

f, Mμ W12
= f(μ).

Theorem 2.2 From the space W22 [a, L], we derive the RK function Rη (ϑ) as [11]:

−1/3 a3 + 1/2 a2 η + a2 − η a + 1/2 ϑ a2 − ϑ η a − ϑ a + ϑ η + 1/2 η ϑ2 − 1/6 ϑ3 , ϑ ≤ η,

Rη (ϑ) =
−1/3 a3 + 1/2 a2 η − 1/6 η 3 + a2 − η a + 1/2 ϑ a2 − ϑ η a + 1/2 ϑ η 2 − ϑ a + ϑ η , ϑ > η.
(2)

Proof We must demonstrate

f, Rη W22
= f(η).

We have
 L
f, Rη W22
= f(a)Rη (a) + f (a)Rη (a) + f (ϑ)Rη (ϑ) dϑ.
a

Integrating by parts yields the following results:

  L
(3)
f, Rη W2 = f(a)Rη (a) + f (a) Rη (a) − Rη (a) + f (L)Rη (L) − f (t)Rη (t)dt.
2 a

Since f(t) ∈ W22 [a, L], f(a) = 0 follows. Then,


  L
f, Rη W22
= f (a) Rη (a) − Rη (a) + f (L)Rη (L) − f (ϑ)R(3)
η (ϑ)dϑ.
a
122 N. Attia and A. Akgül

We have to calculate the following:

Rη (a) = Rη (a) = −a + η,

Rη (L) = 0.

By applying the aforementioned equations, we get:

 L
f, Rη W22
=− f (ϑ)R(3)
η (ϑ)dϑ. (3)
a

We have

−1 , ϑ ≤ η,
R(3)
η (ϑ) = 0 , ϑ > η.

As a result,

 η  L
f, Rη W22
=− f (ϑ)R(3)
η (ϑ)dϑ − f (ϑ)R(3)
η (ϑ)dϑ
a η
 η
= f (ϑ) dϑ
a

and, since f(ϑ) ∈ W22 [a, L], we infer

f, Rη W22
= f(η).

Remarks The reproduction kernel function Zμ (κ) of N32 [a, L] is expressed here, but
due to its length, we have chosen not to include it. For additional information on this
point, the interested reader is directed to [12].

The reproducing kernel function of the space W(3,2)

2 (I)
1. The RK function O(μ,η) (κ, ϑ) of the space W(3,2)
2 (I) is produced by multiplying
the RKFs of the spaces N32 [a, L] and W22 [0, T]—i.e.:

O(μ,η) (κ, ϑ) = Zμ (κ) Rη (ϑ).

2. The RK function V(μ,η) (κ, ϑ) of W(1,1)

2 (I) is produced by multiplying two RKFs
of the same space W12 [a, L]—i.e.:

V(μ,η) (κ, ϑ) = Mμ (κ) Mη (ϑ).

On Solutions of Biological Models Using Reproducing … 123

3 The RKHS Approach

Our primary objective in this part is to outline the steps to apply the suggested
approach for solving the NWSE, which has the following form [20]:

2 [Z (κ, ϑ)] + A Z(κ, ϑ) + A Zq (κ, ϑ), 0 ≤ κ ≤ L, 0 ≤ ϑ ≤ T,

Lϑ [Z(κ, ϑ)] = A1 Lκ 2 3
(4)
accompanied by the following boundary conditions (BCs)

Z (0, ϑ) = φ1 (t), and Z (L, ϑ) = φ2 (ϑ), (5)

and initial conditions (IC)

Z (κ, 0) = Z0 (κ), (6)
∂
2
∂
where Lκ 2
= ∂κ 2 and Lϑ = ∂ϑ . A1 , A2 , and A3 are real constants. q is a positive inte-

ger. The function Z(κ, ϑ) will be determined and it could be seen as the temperature
distribution of temperature in an infinitely thin and long rod or as the flow velocity
of a fluid in an infinitely long pipe with small diameter. The functions φ1 (ϑ), φ2 (ϑ),
and Z0 (κ) are known.
One clever way to investigate problem (4) is to homogenize (5) and (6) using an
appropriate transformation. Thus, let’s take:

Q(κ, ϑ) = Z(κ, ϑ) + P(κ, ϑ),

with
 
1 1
P(κ, ϑ) = (φ1 (0) − φ1 (ϑ)) 1 − κ + (φ2 (0) − φ2 (t))κ − Z0 (κ).
L L

Then, (4)–(6) become

2 [Q(κ, ϑ)] + R [Q(κ, ϑ)] = (κ, ϑ), 0 ≤ κ ≤ L, 0 ≤ ϑ ≤ T, (7)

Lϑ [Q(κ, ϑ)] − A1 Lκ

with BCs
Q(0, ϑ) = Q(L, ϑ) = 0, (8)

and IC
Q(κ, 0) = 0, (9)

the remaining linear terms are denoted by the term R [Q(κ, ϑ)] , and the nonlinear
terms are expressed by the term (κ, ϑ).
Here, we present the linear operator : W(3,2)
2 (I) → W(1,1)
2 (I) , as

Q(κ, ϑ) = Lϑ [Q(κ, ϑ)] − A1 Lκ

2
[Q(κ, ϑ)] + R [Q(κ, ϑ)] . (10)
124 N. Attia and A. Akgül

Lemma 3.1 The operator : W(3,2)

2 (I) → W(1,1)
2 (I) is bounded and linear.
Proof The linearity of can be easily verified. We may demonstrate the boundedness
of directly. And, showing that

Q W(1,1) ≤C Q W(3,2) , with C > 0,

2 2

is necessary to prove this.

From Definition 2.5, we get

Q(κ, ϑ) 2 (1,1) =  Q(κ, ϑ), Q(κ, ϑ) (1,1)

W2 W2
 T 2
∂
= Q(a, ϑ) dϑ +  Q(κ, 0), Q(κ, 0)W1
0 ∂ϑ 2
 T L 2
∂ ∂
+ Q(κ, ϑ) dκdϑ
0 0 ∂κ ∂ϑ
 L 2  d 2
∂ ∂
= [ Q(0, 0)]2 + Q(κ, 0) dκ + Q(a, ϑ) dϑ
0 ∂κ 0 ∂ϑ
 T L 2
∂ ∂
+ Q(κ, ϑ) dκdϑ.
0 0 ∂κ ∂ϑ

The RP has allowed us to have

Q(κ, ϑ) = Q(•, ∗), O(κ,ϑ) (•, ∗) W(3,2) .

2

In the same way,


∂ı ∂j ∂ı ∂j
Q(κ, ϑ) = Q(•, ∗), O(κ,ϑ) (•, ∗) , ı, j = 0, 1.
∂κ ı ∂ϑj ∂κ ı ∂ϑj W(3,2)
2

The Schwarz inequality yields

 ı   

 ∂ ∂j   ∂ı ∂j 
 Q(κ, ϑ)  =  Q(•, ∗), O (•, ∗) 
 ∂κ ı ∂ϑj   ∂κ ı ∂ϑj
(κ,ϑ)

W(3,2)
 ı  2

 ∂ ∂ j 
≤ Q W(3,2)  
 ∂κ ı ∂ϑj O(κ,ϑ) (•, ∗) (3,2) .
2
W2

Thanks to the continuity of O(κ,ϑ) (•, ∗), we deduce

 ı 
 ∂ ∂j 
 Q(κ, ϑ)  ≤ Cı,j Q (3,2) , j, ı = 0, 1.
 ∂κ ı ∂ϑj  W2

Hence
On Solutions of Biological Models Using Reproducing … 125
 L  T
Q(κ, ϑ) 2
W(1,1)
≤C20,0 Q 2
W(3,2)
+ C21,0 Q 2
W(3,2)
dκ + C20,1 Q 2
W(3,2)
dϑ
2 2 0 2 0 2
 T L
+ C21,1 Q 2W(3,2) dκdϑ
0 0 2
 
≤ C20,0 + C21,0 L + C20,1 T + C21,1 L T Q 2
W(3,2)
.
2

By taking C = C20,0 + C21,0 L + C20,1 T + C21,1 L T, we get the desired result. 

Applying (10) allows us to rewrite (7) and (9) as follows

⎧
⎨ Q(κ, ϑ) =  (κ, ϑ) , 0 ≤ κ ≤ L, 0 ≤ ϑ ≤ T,
Q (0, ϑ) = Q (L, ϑ) = 0, (11)
⎩
Q (κ, 0) = 0.

Building the orthogonal function system of W(3,2)

2 (I) is necessary now. Let’s start
with some helpful functions:
∗
ρı (κ, ϑ) = V(κı ,ϑı ) (κ, ϑ) and ψı (κ, ϑ) = ρı (κ, ϑ),

where
∞
• {(κı , ϑı )}ı=1 is a dense countable set in I.
• The RKF connected to W(1,1) 2 (I) is represented by V(κı ,ϑı ) (κ, ϑ).
• The formal neighbor of is ∗ .
∞
Using the Gram–Schmidt procedure, we obtain the orthonormal system {ψ̄ı }ı=1 in
(3,2)
W2 (I) shown below:


ı
ψ̄ı (κ, ϑ) = ςık ψk (κ, ϑ), ςıı > 0, ı = 1, 2, . . . .
k=1

∞
Here, {ψı }ı=1 designates a function system in W(3,2)
2 whose expression can be found
using the formula:
∗ ∗
ψı (κ, ϑ) = ρı (κ, ϑ) = ρı (μ, η), O(κ,ϑ) (μ, η) W(3,2)
2

= V(μı ,ηı ) (μ, η), (μ,η) O(κ,ϑ) (μ, η) W(1,1)

2

= (μ,η) O(κ,ϑ) (μ, η)|(μ,η)=(κı ,ϑı )

= (μ,η) O(μ,η) (κ, ϑ)|(μ,η)=(κı ,ϑı ) .

Applying to (μ, η) variables is shown by the operator (μ,η) . However, the orthog-
onalization coefficients ςık are provided by:
126 N. Attia and A. Akgül
⎧
⎪
⎨ ψ1
1
, for ı = j = 1,
ςıj = 1
, for ı = j = 1, (12)
⎪ eı
⎩− 1 ı−1
eı k=j C ık ςkj , for ı > j,

 ı−1
where eı = ψı 2
− k=1
2
Cık , Cık = ψı , ψ̄k W(3,2)
.
2

∞ ∞
Theorem 3.1 If {(κı , ϑı )}ı=1 is dense on I, then {ψı }ı=1 represents the complete
(3,2)
system of W2 (I).

Proof It is evident that ψı ∈ W(3,2)

2 (I). Consequently, for each Q ∈ W(3,2)
2 , we write

Q(κ, ϑ), ψı (κ, ϑ)W(3,2) = 0, ı = 1, 2, . . . .

2

Since

Q(κ, ϑ), ψı (κ, ϑ)W(3,2) =  Q(κ, ϑ), ρı (κ, ϑ)W(1,1) = Q(κı , ϑı ) = 0.

2 2

∞
As a result of the density of {(κı , ϑı )}ı=1 in I,

Q(κ, ϑ) = 0. (13)

−1
And the fact that exists implies that

Q(κ, ϑ) = 0.

∞
Theorem 3.2 Assuming that {(κı , ϑı )}ı=1 is dense on I and (11) has a unique
(3,2)
solution in W2 (I), then this solution takes the form:

∞ 
 ı
Q(κ, ϑ) = ςık (κk , ϑk )ψ̄ı (κ, ϑ), (14)
ı=1 k=1

and the solution of (4)–(6) is

∞ 
 ı
Z(κ, ϑ) = ςık (κk , ϑk )ψ̄ı (κ, ϑ) − P(κ, ϑ). (15)
ı=1 k=1

 ∞
Proof The completeness of ψ̄ı (κ, ϑ) ı=1 in W(3,2)
2 (I) allows us to compute:
On Solutions of Biological Models Using Reproducing … 127

∞

Q(κ, ϑ) = Q(κ, ϑ), ψ̄ı (κ, ϑ) W(3,2) ψ̄ı (κ, ϑ)
2
ı=1
∞ ı
∗
= ςık Q(κ, ϑ), ρk (κ, ϑ) W(3,2) ψ̄ı (κ, ϑ)
2
ı=1 k=1
∞  ı
= ςık  Q(κ, ϑ), ρk (κ, ϑ)W(1,1) ψ̄ı (κ, ϑ)
2
ı=1 k=1
∞  ı
= ςık Q(κ, ϑ), V(κk ,ϑk ) (κ, ϑ) W(1,1) ψ̄ı (κ, ϑ)
2
ı=1 k=1
∞  ı
= ςık (κk , ϑk )ψ̄ı (κ, ϑ),
ı=1 k=1

with (κk , ϑk ) = Q(κk , ϑk ).

Additionally, we mention that it is simple to obtain (15) from Z(κ, ϑ) =
Q(κ, ϑ) − P(κ, ϑ) and (15). 

Remarks In this case, the approximate solution is obtained if we take a finite many
terms in (14):
p
ı
Q p (κ, ϑ) = ςık (κk , ϑk )ψ̄ı (κ, ϑ).
ı=1 k=1

W(3,2)
2 (I) is a Hilbert space, hence it is obvious that

∞ 
 ı
ςık (κk , ϑk )ψ̄ı (κ, ϑ) < ∞.
ı=1 k=1

We come to the conclusion that Q p (κ, ϑ) is convergent in the norm.

4 Analysis of Convergence

The following is the approximate answer for problem (14) provided by the RKHSM:


p
Q p (κ, ϑ) = ı ψ̄ı (κ, ϑ), (16)
ı=1

in which

ı
ı = ςık (κk , ϑk ). (17)
k=1
128 N. Attia and A. Akgül

Here, by letting (κ1 , ϑ1 ) = (0, L), the values of Q(κ1 , ϑ1 ) will be known from the
IC and BCs. And, Q0 (κ1 , ϑ1 ) = Q(κ1 , ϑ1 ).
∞
  4.1 Assume that {(κı , ϑı )}ı=1 is dense on I, (16) has a unique solution,
Theorem
and Q p W(3,2) is bounded. Then,
2

1. Q p (κ, ϑ) converges to Q(κ, ϑ).

2.

p
Q p (κ, ϑ) = ı ψ̄ı (κ, ϑ),
ı=1

where ı is given by (17).

Proof 1. From (16), we write

Q p+1 (κ, ϑ) = Q p (κ, ϑ) +  p+1 ψ̄ p+1 (κ, ϑ),

 ∞
since ψ̄ı (κ, ϑ) ı=1 is orthogonal,
   
Q p+1 2 (3,2) = Q p 2 (3,2) + 2
W2 W2 p+1
 2
= Q p−1 W(3,2) + 2p + 2p+1
2

..
.

p+1
= Q0 2
W(3,2)
+ ı2 , (18)
2
ı=1

and so    
Q p  (3,2) ≤ Q p+1  (3,2) .
W 2 W 2

 
The fact that Q p  is bounded makes its convergence obvious. Consequently,
W(3,2)
2
there is a positive constant  in which
∞

ı2 = .
ı=1

Consequently  2 ∞
ı ı=1 ∈ 2 .
   
Since Qq (κ, ϑ) − Qq−1 (κ, ϑ) ⊥ · · · ⊥ Q p+1 (κ, ϑ) − Q p (κ, ϑ) , for q > p
we have
On Solutions of Biological Models Using Reproducing … 129

   
Qq − Q p 2 (3,2) = Qq − Qq−1 + Qq−1 − · · · + Q p+1 − Q p 2 (3,2)
W2 W 2
 2  2  2
= Qq − Qq−1  (3,2) + Qq−1 − Qq−2  (3,2) + · · · + Q p+1 − Q p  (3,2) .
W 2 W 2 W 2
(19)

Furthermore,  
Qq − Qq−1 2 (3,2) = 2 .
W q2

Consequently, we have

  
q
Qq − Q p 2 (3,2) = 2 → 0, as n → ∞.
W 2
= p+1

Since W(3,2)
2 (I) is complete, we arrive at: Qn → Q̂ as n → ∞.
2. We begin by taking the limits in (16)
∞

Q̃(κ, ϑ) = ı ψ̄ı (κ, ϑ). (20)
ı=1

When we apply the linear operator to (20), we discover

∞

Q̃(κ, ϑ) = ı ψ̄ı (κ, ϑ),
ı=1

it follows that
∞

Q̃(x , t ) = ı ψ̄ı (κ, ϑ), ρ (κ, ϑ) W(1,1)
2
ı=1
∞
∗
= ı ψ̄ı (κ, ϑ), ρ (κ, ϑ) W(3,2)
2
ı=1
∞
= ı ψ̄ı (κ, ϑ), ψ (κ, ϑ) W(3,2) . (21)
2
ı=1

j
Multiplying (21) by ςj  and taking =1 to find

j ∞
 j

  
ςj  Q̃(x , t ) = ı ψ̄ı (κ, ϑ), ςj  ψ (κ, ϑ)
=1 ı=1 =1 W(3,2)
2
∞

= ı ψ̄ı (κ, ϑ), ψ̄j (κ, ϑ) W(3,2)
2
ı=1
= j .
130 N. Attia and A. Akgül

Using (17), we discover

Q̃(κ , ϑ ) = (κ , ϑ ).

∀(μ, η) ∈ I, ∃{(κ pj , ϑ pj )}∞

j =1 such that

(κ pj , ϑ pj ) → (μ, η), as j → ∞.

∞
That is a result of the fact that {(κı , ϑı )}ı=1 is dense on I.
Also we are aware of the following:

Q̃(κ pj , ϑ pj ) = (κ pj , ϑ pj ).

Using j → ∞ and the continuity of , we can determine the desired result.



5 Numerical Experiments

The theoretical foundation is tested using two applications of biological problems.

Now, the following algorithm can be used to apply the RKHSM:
1st Step : Fix p =  × q;
2nd Step : Set ψı (κı , ϑj ) = (μ,η) O(κ,ϑ) (μ, η)|(μ,η)=(κı ,ϑı ) ;
3rd Step : Calculate ςıj using
 (12);
4th Step : Set ψ̄ı (κ, ϑ) = ık=1 ςık ψk (κ, ϑ), ςıı > 0, ı ∈ {1, . . . , p};
5th Step : Choose an initial guess Q0 (κ1 , ϑ1 );
6th Step : Set ı = 1;
7th Step : Set ı = ık=1  ςpık (κk , tk );
8th Step : Qı (κı , ϑı ) = k=1 k ψ̄k (κı , ϑı );
9th Step : If n > ı, set ı = ı + 1. Go to step 7. Else stop.

Example 5.1 Considering the following NWSE:

⎧
⎪
⎪ ∂ϑ Z(κ, ϑ) = ∂ 2 2 Z(κ, ϑ) + 3Z(κ, ϑ) − 4Z3 (κ, ϑ), 0 ≤ κ, ϑ ≤ 1,
⎪
⎪ 
κ
⎪  √
⎪
⎪ e 6κ√
⎨ Z(κ, 0) = 3
4 √
,
6
e 6κ +e 2 κ ! "
⎪
⎪ √ √
⎪
⎪
⎪
⎪ Z(0, ϑ) =  3  , and Z
(1, ϑ) = 1 3 1− 1 .
⎪
⎩ 9ϑ 2 9ϑ + 3
2 e− 2 +1 e 2 2 +1

(22)
On Solutions of Biological Models Using Reproducing … 131

The aforementioned problem has the following exact solution:

#  √
3 e 6κ
Z(κ, ϑ) = √ √ .
4 e 6κ + e 26 κ− 29 ϑ

Example 5.2 Considering the NWSE as:

⎧
⎨ ∂t Z(κ, ϑ) = ∂κ 2 Z(κ, ϑ) + 2Z(κ, ϑ) − 3Z (κ, ϑ), 0 ≤ κ, ϑ ≤ 1,
2 2

Z(κ, 0) = λ, (23)
⎩ λe2ϑ λe2ϑ
Z(0, ϑ) = 3λeϑ sinh(ϑ)+1 , and Z (1, ϑ) = 3λeϑ sinh(ϑ)+1 .

The aforementioned problem has the following exact solution:

2λ exp(2ϑ)
Z(κ, ϑ) = −  .
3 λ(− exp(2ϑ)) + λ − 23

6 Computing Outcomes and Analysis

As we previously explained, the RKHSM is used for all cases. Considering p collo-
cation points with κı = ı , ı = 1, 2 . . . ,  and ϑj = qj , j = 1, 2 . . . , q with  × q =
p. By computing the absolute error values defined as |Z − Z p | with (κ, ϑ) ∈ [0, 1]2 ,
the numerical solution to each problem is determined and compared with the exact
solution.
In Example 5.1, we used 10 collocation points ( p = 10). Figures 1 and 2 show
the approximate and exact solutions for Example 5.1. These graphs demonstrate
how similar the behavior of the graphs is. To illustrate more contrasts between
the RKHSM’s and exact solutions, as well as other approaches already in use
[6] (Extended cubic uniform B-spline, Trigonometric cubic B-spline, and Uniform
Cubic B-spline; for short ECBS, TCBS, and UCBS, respectively), We calculated
and displayed the absolute errors in Table 1. The accuracy of the suggested method
is confirmed by the good agreement between the results from the RKHSM and true
solutions. Furthermore, we may assert that the RKHSM is more accurate than the
alternative approaches.
In Example 5.2, we used 10 collocation points( p = 10). Figures 3 and 4 show
the approximate and exact solutions for Example 5.2. These graphs demonstrate
how similar the behavior of the graphs is. To illustrate more contrasts between the
RKHSM’s and exact solutions, as well as other approaches already in use [6] (UCBS,
TCBS, and ECBS), We calculated and displayed the absolute errors in Table 2. The
accuracy of the suggested method is confirmed by the good agreement between the
results from the RKHSM and true solutions. Furthermore, we may assert that the
RKHSM is more accurate than the alternative approaches.
132 N. Attia and A. Akgül

Fig. 1 Comparison of the RKHSM solution to Example 5.1 at ϑ = 0.3 and the exact solution

Fig. 2 Comparison of the RKHSM solution to Example 5.2 at κ = 0.6 and the exact solution
On Solutions of Biological Models Using Reproducing … 133

Table 1 Results’ comparison for Example 5.1

κ ϑ UCBS [6] TCBS [6] ECBS [6] RKHSM
0.2 0.2 6.129 × 10−2 6.134 × 10−2 5.166 × 10−2 8.206 × 10−4
0.4 7.120 × 10−2 7.129 × 10−2 5.402× 10−2 4.215 × 10−4
0.6 4.312 × 10−2 6.430 × 10−2 4.511 × 10−2 2.344 × 10−4
0.8 4.661 × 10−2 4.670 × 10−2 2.860 × 10−2 3.560 × 10−5
1 1.560 × 10−2 1.563 × 10−2 1.518 × 10−3 3.300 × 10−4
0.4 0.2 8.711 × 10−2 8.718 × 10−2 7.224 × 10−2 7.496 × 10−4
0.4 1.005 × 10−1 1.006 × 10−1 7.421 × 10−2 4.584 × 10−5
0.6 8.771 × 10−2 8.786 × 10−2 5.888 × 10−2 3.047 × 10−5
0.8 6.001 × 10−2 6.015 × 10−2 3.321 × 10−2 8.582 × 10−6
1 1.640 × 10−2 1.650 × 10−2 4.323 × 10−3 6.559 × 10−4
0.6 0.2 8.030 × 10−2 8.036 × 10−2 6.461 × 10−2 4.713 × 10−4
0.4 9.072 × 10−2 9.085 × 10−2 6.456 × 10−2 2.153 × 10−4
0.6 7.637 × 10−2 7.651 × 10−2 4.822 × 10−2 5.270 × 10−7
0.8 4.899 × 10−2 4.911 × 10−2 2.328 × 10−2 2.042 × 10−4
1 9.878 × 10−3 9.969 × 10−3 9.679 × 10−3 1.020 × 10−3
0.8 0.2 4.857 × 10−2 4.862 × 10−2 3.721 × 10−2 2.060 × 10−5
0.4 5.319 × 10−2 5.328 × 10−2 3.588 × 10−2 3.690 × 10−4
0.6 6.420 × 10−2 4.321 × 10−2 2.500 × 10−2 3.494 × 10−6
0.8 2.582 × 10−2 2.590 × 10−2 9.644 × 10−3 3.464 × 10−4
1 2.862 × 10−3 2.915 × 10−3 9.118 × 10−3 1.067 × 10−3

Fig. 3 Comparison of the RKHSM solution to Example 5.2 at κ = 0.6 and the exact solution
134 N. Attia and A. Akgül

Fig. 4 Absolute error of the RKHSM for Example 5.2 at ϑ = 0.6

Table 2 Results’ comparison for Example 5.2

κ ϑ UCBS [6] TCBS [6] ECBS [6] RKHSM
0.2 0.2 8.323 × 10−4 8.295 × 10−4 6.068 × 10−4 1 × 10−10
0.4 1.011 × 10−3 1.007 × 10−3 6.800× 10−4 0
0.6 8.581 × 10−4 8.514 × 10−4 5.476 × 10−4 1 × 10−10
0.8 4.745 × 10−4 4.651 × 10−4 2.746 × 10−4 1 × 10−10
1 1.991 × 10−5 3.239 × 10−5 6.339 × 10−5 1 × 10−10
0.4 0.2 1.226 × 10−3 9.013 × 10−4 1.222 × 10−3 1 × 10−10
0.4 1.520 × 10−3 1.023 × 10−3 1.514 × 10−3 0
0.6 1.302 × 10−3 8.289 × 10−4 1.292 × 10−3 1 × 10−10
0.8 7.334 × 10−4 4.221 × 10−4 7.194 × 10−4 1 × 10−10
1 4.932 × 10−6 8.366 × 10−5 2.342 × 10−5 1 × 10−10
0.6 0.2 1.226 × 10−3 1.222 × 10−3 9.013 × 10−4 1 × 10−10
0.4 1.520 × 10−3 1.514 × 10−3 1.023 × 10−3 0
0.6 1.302 × 10−3 1.292 × 10−3 8.289 × 10−4 1 × 10−10
0.8 7.334 × 10−4 7.194 × 10−4 4.221 × 10−4 1 × 10−10
1 4.932 × 10−6 2.342 × 10−5 8.366 × 10−5 1 × 10−10
0.8 0.2 8.323 × 10−4 8.295 × 10−4 6.068 × 10−4 1 × 10−10
0.4 1.011 × 10−3 1.007 × 10−3 6.800 × 10−4 0
0.6 8.581 × 10−4 8.514 × 10−4 5.476 × 10−4 1 × 10−10
0.8 4.745 × 10−4 4.651 × 10−4 2.746 × 10−4 1 × 10−10
1 1.991 × 10−5 3.239 × 10−5 6.339 × 10−5 1 × 10−10
On Solutions of Biological Models Using Reproducing … 135

7 Conclusion

The NWSEs, which describe the dynamical behavior near the bifurcation point, were
effectively solved numerically in the current work. The two primary steps in using
the RKHSM are creating an orthonormal function system and defining a suitable
bounded linear operator. It is demonstrated that the suggested approach has good
convergence. To demonstrate the RKHSM’s capability and reliability, two examples
were used. When exact results are compared to our acquired results, it is seen that they
are in strong agreement. Figure 1 displays the variations of Z(κ, ϑ) in Example 5.1
over a fixed time ϑ with respect to position κ, demonstrating that Z(κ, ϑ) increases
as x increases. The variations of Z(κ, ϑ) in Example 5.2 are shown in Figs. 2 and 3.
This graph demonstrates that Z(κ, ϑ) increases as t increases. The absolute error of
RKHSM in Example 5.2 is shown in Fig. 4. The accuracy of the RKHSM is truly
demonstrated by this figure. Finally, it can be said that the suggested approach is very
convincing and effective for solving a wide range of nonlinear equations that describe
real biological systems. As part of our goal, we intend to make a novel contribution
to the literature by applying a novel method to partial differential equations with
multiple variables.

References

1. A. Akgül, A novel method for a fractional derivative with non-local and non-singular kernel.
Chaos Solitons Fractals 114, 478–482 (2018)
2. S. Zaremba, Sur le calcul numérique des fonctions demandées dans le problème de Dirichlet et
le problème hydrodynamique. Bulletin International de l’Académie des Sciences de Cracovie
68, 125–195 (1908)
3. B. Ghanbari, A. Akgül, Abundant new analytical and approximate solutions to the generalized
Schamel equation. Phys. Scr. 95, 075201 (2020)
4. E. Babolian, S. Javadi, E. Moradi, RKM for solving Bratu-type differential equations of frac-
tional order. Math. Methods Appl. Sci. 39, 1548–1557 (2016)
5. M. Modanli M, A. Akgül, On solutions to the second-order partial differential equations by
two accurate methods. Numer. Methods Part. Diff. Equ. 34, 1678–1692 (2018)
6. W.K. Zahra, W.A. Ouf, M.S. El-Azab, Cubic B-spline collocation algorithm for the numerical
solution of Newell Whitehead Segel type equations. Electr. J. Math. Anal. Appl. 2, 81–100
(2014)
7. M.G. Sakar, Iterative reproducing kernel Hilbert spaces method for Riccati differential equation.
J. Comput. Appl. Math. 309, 163–174 (2017)
8. W. Jiang, T. Tian, Numerical solution of nonlinear Volterra integro-differential equations of
fractional order by the reproducing kernel method. Appl. Math. Model. 39, 4871–4876 (2015)
9. F. Geng, M. Cui, A reproducing kernel method for solving nonlocal fractional boundary value
problems. Appl. Math. Lett. 25, 818–823 (2012)
10. F. Geng, M. Cui, New method based on the HPM and RKHSM for solving forced Duffing
equations with integral boundary conditions. J. Comput. Appl. Math. 233, 165–172 (2009)
11. M. Cui, Y. Lin, Nonlinear Numerical Analysis in the Reproducing Kernel Space (Nova Science
Publishers, Inc, New York, 2009)
12. N. Attia et al., On solutions of time-fractional advection-diffusion equation. Numer. Methods
Part. Diff. Equ. (2020). https://doi.org/10.1002/num.22621
136 N. Attia and A. Akgül

13. O. Abu Arqub, J. Singh, B. Maayah, M. Alhodaly, Reproducing kernel approach for numerical
solutions of fuzzy fractional initial value problems under the Mittag-Leffler kernel differential
operator. Math. Methods Appl. Sci. 1–22 (2021). https://doi.org/10.1002/mma.7305
14. S. Momani, N. Djeddi, M. Al-Smadi, S. Al-Omari, Numerical investigation for Caputo-Fabrizio
fractional Riccati and Bernoulli equations using iterative reproducing kernel method. Appl.
Numer. Math. 170, 418–434 (2021)
15. H. Singh, Analysis for fractional dynamics of Ebola virus model. Chaos Solitons & Fractals
138, 109992 (2020)
16. H. Singh, Analysis of drug treatment of the fractional HIV infection model of CD4+ T-cells.
Chaos Solitons & Fractals 146, 110868 (2021)
17. H. Singh, D. Baleanu, J. Singh, H. Dutta, Computational study of fractional order smoking
model. Chaos Solitons & Fractals 142, 110440 (2021)
18. H. Singh, H.M. Srivastava, D. Baleanu (Eds.), Methods of Mathematical Modeling: Infectious
Diseases (Elsevier Science, 2022). https://doi.org/10.1016/C2021-0-00445-4
19. T. Allahviranloo, H. Sahihi, Reproducing kernel method to solve fractional delay differential
equations. Appl. Math. Comput. 400, 126095 (2021)
20. D. Kumar, R.P. Sharma, Numerical approximation of Newell-Whitehead-Segel equation of
fractional order. Nonlinear Eng. 5(2), 81–86 (2016)
21. Y. Chellouf, B. Maayah, S. Momani, A. Alawneh, S. Alnabulsi, Numerical solution of fractional
differential equations with temporal two-point BVPs using reproducing kernal Hilbert space
method. AIMS Math. 6(4), 3465–3485 (2021)
A Study of the Fractional
Tumour–Immune Unhealthy Diet Model
Using the Pseudo-operational Matrix
Method

Saurabh Kumar and Vikas Gupta

Abstract One of the main causes of death in the world is cancer. Cancer is a col-
lection of illnesses characterized by the formation of tumours, malignant cells, or
cancer cells capable of becoming cancerous. Several studies have been conducted
on cancer and related disorders in recent years. In this chapter, we study the frac-
tional tumour-immune unhealthy diet (TIUNHD) model, and to solve it; we use the
fractional-order Laguerre operational matrix-based method. The fractional derivative
of Caputo is used in this chapter. In this method, first, we constructed the pseudo-
operational matrices of fractional, integer order integration and then used these
pseudo-operational matrices to approximate the unknown solutions of the given sys-
tem of non-linear fractional differential equations model. This results in an algebraic
system of equations, which can be easily solved by Newton’s iterative method. A
numerical experiment demonstrates the proposed algorithm’s applicability to solving
the TIUNHD model.

Keywords Caputo fractional derivative · Fractional-order Laguerre polynomials ·

Pseudo-operational matrix of integration · Fractional TIUNHD model

1 Introduction

An abnormal cell that divides uncontrollably anywhere on the body is defined as

cancer. It is typically used in epidemiology studies to determine the extent of infec-
tion and what might be causing it. Several studies show that new dietary habits,
climatic conditions, and other factors have increased cancer survivorship rates [1].
During the Leadership for the Twenty-First Century conference, Venter said, “human

S. Kumar · V. Gupta (B)

Department of Mathematics, Centre for Mathematical and Financial Computing, The LNM
Institute of Information Technology, Jaipur 302031, India
e-mail: [email protected]; [email protected]
S. Kumar
e-mail: [email protected]; [email protected]

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 137
H. Singh and H. Dutta (eds.), Computational Methods for Biological Models,
Studies in Computational Intelligence 1109,
https://doi.org/10.1007/978-981-99-5001-0_6
138 S. Kumar and V. Gupta

biology is considerably more complex than we thought.” Biological issues are not
only influenced by genetic factors, but also by the proteins and cells affected by
environmental factors [2]. Everyone is indeed born with some genetic characteris-
tics; however, these genes have minimal impact on how some people will turn out in
life. Although genes communicate essential information about high infection risk,
they rarely play a direct role in determining a disease’s actual occurrence or cause.
The relationship between diet (such as vitamins E, A, D, C, and B) and immunity
has gained importance in recent years. According to studies [3–5], these vitamins
are essential in maintaining immune system activity to protect tissues from injury.
For cancer studies, researchers have proposed various mathematical models using
ordinary differential equations (ODEs), delay differential equations, and partial dif-
ferential equations to interrogate the effects of tumour evolution on the dynamics
of other cells [6–10]. It has been found that obesity, hormones, and diet are among
the primary causes of cancer. However, most of the components remain unknown.
The Michaelis–Menten function has been used to study tumour-immune interac-
tions [11, 12]. Rambely and Alharbi proposed a healthy-immune model to illustrate
how the immune system resists tumour development dynamically [13]. According
to the analysis, their model compares results with those of a weakened immune sys-
tem. A model explaining how obesity and cancer interact to boost tumour response
to chemotherapy has been developed by Roberto et al. [14]. As a means of explor-
ing the effects of vitamin interventions on tumours and immune cells, Alharbi and
Rambely developed the tumour-immune-vitamin system via ODEs in [15]. A math-
ematical model based on ODEs was developed in 2012 by Mufudza and others [16]
on the influence of oestrogen on breast cancer dynamics. A mathematical model of
pancreatic cancer is considered by [17]. Arciero and others [18] presented a mathe-
matical model that examined tumour growth, immune escape, and siRNA treatment
of tumours. In this study, an unhealthy diet has been identified as one of the main
risks for increased cancer cases. According to a numerical simulation, vitamins can
boost immunity when taken regularly. The immune system is boosted by 16% per
day, with abnormal cells appearing in the tissues. The TIUNHD model dynamics
might assist in understanding how vitamin intake affects immune and tumour cell
dynamics. The TIUNHD model consists of a system of non-linear ODEs.
Numerous fractional operators have been described in the literature, including
Riemann–Liouville, Hadamard, Grunwald–Letnikov, Caputo derivatives, and other
adaptive derivatives [19]. Here are some references on how to solve fractional models
that govern real-life applications such as [20–24]. Because we seek a highly accurate,
fully discrete numerical scheme for solving the fractional TIUNHD model, we wish
to evaluate the applicability of the operational matrix approach based on fractional
order Laguerre polynomials. According to the proposed technique, we first develop
the fractional, integer order pseudo-operational integration matrices for fractional-
order Laguerre polynomials. The fractional Taylor’s functions are used to define
the fractional-order Laguerre polynomials. These pseudo-operational matrices are
used to approximate the integer and fractional order derivatives contained in the
TIUNHD model, leading to an algebraic system of equations. After collocating this
system of equations at the Newton–Cotes nodal points, we used Newton’s iterative
A Study of the Fractional Tumour–Immune Unhealthy Diet … 139

approach to solve the resulting system of algebraic equations in order to determine

the unknown coefficients. The fractional order Laguerre polynomials are easy to
work with and offer a highly accurate approximation solution, which is an advantage
of the suggested operational matrix approach. Contrary to previous discretization
methods [25–27] the proposed method converts the non-linear system into algebraic
equations without requiring any linearization. Some collocation methods based on
Laguerre and other polynomials to solve fractional differential equations are given
in [28–31]. The proposed pseudo-operational matrix (POM) technique can be con-
sidered as an alternative approach to the adaptive finite difference, spline collocation
techniques [32–35]. No article has yet discussed the numerical method based on
fractional order Laguerre polynomials for solving the fractional TIUNHD model. So
we were motivated to solve the fractional TIUNHD model using the fractional order
Laguerre operational matrix.
The rest of the chapter is structured as follows. Preliminaries will be discussed in
the next section. The fractional TIUNHD model is introduced in Sect. 3. In Sect. 5,
we discuss the function approximation. Section 6 presents the pseudo-operational
integration matrix for integer and fractional orders. In Sects. 7 and 8, we discussed
the numerical method and discussion of results, respectively. In Sect. 9, the conclusion
is discussed.

2 Preliminaries

Several definitions of fractional calculus are presented in this section, which is essen-
tial to the rest of the chapter.
Definition 2.1 The Riemann–Liouville integral of the function f (x) ∈ L 1 [a, b] in
the interval (a, b) with fractional order μ ≥ 0 is defined as follows [36]:
 x
μ
1
(μ)
f (t)(x − t)μ−1 dt, μ > 0,
J f (x) = a
(2.1)
f (x), μ = 0,

where, (μ) is widely known as the gamma function.

For μ, η > 0, f (x) ∈ L 1 [a, b] and ρ > −1, the Riemann–Liouville integral
exhibits the following properties.
(i) J μ J η f (x) = J η J μ f (x).
(ρ+1)
(ii) J μ x ρ = (ρ+1+μ) x (μ+ρ) .
μ η μ+η
(iii) J J f (x) = J f (x).

Definition 2.2 The Riemann–Liouville fractional-order derivative of the function

f (x) with order μ ∈ (0, 1), is defined as follows [36]:
 x
μ 1 d
D f (x) = f (t)(x − t)−μ dt, (2.2)
(1 − μ) d x a
140 S. Kumar and V. Gupta

where, D μ denotes the fractional derivative with order μ.

Definition 2.3 The Caputo fractional derivative of f (x) of order μ ∈ (r − 1, r ) can

be defined as follows [37, 38]:

 x
1 dr
D μ f (x) = (x − t)r −μ−1 f (t) dt, (2.3)
(r − μ) 0 dt r

where r is the smallest positive integer such that r − 1 < μ < r. When μ ∈ N, the
differential operator treat as same as the integer order derivative operator.
For r − 1 < μ ≤ r , and f (x) ∈ L 1 [a, b], following properties can be determined
by Caputo’s derivative definition:
(i) In case of constant functions, the Caputo derivative is zero when its order μ is
greater than zero.
(ii) (D μ J μ f 
)(x) = f (x).
(ρ+1) ρ−μ
x , if ρ > μ − 1, r − 1 < μ < r, ρ ∈ R
(iii) D μ x ρ = 1+ρ−μ .
0, if ρ ≤ μ − 1, r − 1 < μ < r, ρ ∈ {−1, −2, . . .}
 −1
(iv) (J μ D μ f )(x) = f (x) − rm=0 f m (0+ ) (m+1)
xm
.

Definition 2.4 (The Kronecker delta function [39]) Mathematically, the Kronecker-
delta function is a function of two variables; when the two variables are equal, the
function returns 1; otherwise, it returns 0, i.e.,

1, if i = j,
δi j = . (2.4)
0, if i = j.

Definition 2.5 (Fractional-order Laguerre functions) The analytic form of the frac-
tional order Laguerre functions in the interval (0 ≤ x < ∞) is defined as follows [40]:


m
(−1)s (m + 1)
L μm (t) = t μs . (2.5)
s=0
(s + 1)(s + 1)(m − s + 1)

In the form of recurrence relation, the fractional order Laguerre polynomials are
defined as follows:

L μm+1 (t) = (2m + 1 − t μ )L μm (t) − m 2 L μm−1 (t), m = 1, 2, . . . , (2.6)

μ μ
where L 0 (t) = 1 and L 1 = 1 − t μ .
μ
Theorem 2.1 In terms of the weight function wμ (t) = t μ−1 e−t in interval [0, ∞),
the fractional order Laguerre polynomials are orthogonal with the following orthog-
onal property:  ∞
μ μ δi j
wμ (t)L i (t)L j (t)dt = , (2.7)
0 μ
A Study of the Fractional Tumour–Immune Unhealthy Diet … 141

where δi j is the Kronecker delta function defined in Eq. (2.4).

μ 1
Proof By substituting y = t μ and using the property L m (z μ ) = L m (z) of the frac-
tional order Laguerre polynomials, we obtained,
 ∞  ∞  ∞
μ μ μ 1 μ μ
wμ (t)L i (t)L j (t)dt= t μ−1 e−t L i (t μ )L j (t μ )dt =
L i (z)L j (z)dz.
0 0 0 μ
(2.8)
By orthogonal property of fractional order Laguerre polynomials, we obtained the
desired result. 

3 Fractional TIUNHD Model

In this section, we explain how the TIUNHD model works. This model is based on
a nonlinear system of ODEs. Further, we hypothesized that tumour cells emerge in
tissue when the immune system cannot eliminate or suppress abnormal cells. The
primary role of the immune system is to protect the body from cancer. There are two
main populations in the TIUNHD model that is based on non-linear ODEs. First is
the tumour cells population and it is denoted by the symbol T. The cells of this type
may divide rapidly in the presence of a weak immune system while invading other
tissues or countering them [41]. Following is an ordinary differential equation that
represents tumour cell growth:

dT
= c1 T (1 − c2 T ) − c3 T I, (3.1)
dt
where the parameter c2 denotes the tumour reduction due to dietary metabolization
of the garbled tumour, and c1 represents the growth limit of tumour cells. Thirdly,
parameter c3 determines the rate at which tumour cells are eliminated or suppressed
by the immune system under the assumption that a poor diet compromises the immune
system’s response. The immune system plays a crucial role in preventing the devel-
opment of cancer. When tumours are present in the tissue, several immune cells
are produced. In some cases, these cells are ineffective at delaying or eliminating
tumour cells. I represents the immune response behaviour governed by the following
non-linear ordinary differential equation:

dI δ3 I T
= δ1 − δ2 I + − δ5 T I. (3.2)
dt δ4 + T

Parameters are described below in details

(i) δ1 represents the constant production of immune cells by the body on a regular
basis.
(ii) δ2 represents natural death percentage of the immune cells.
142 S. Kumar and V. Gupta

(iii) Symbol δ3 represents the percentage of immune cells that naturally die.
(iv) δ4 represents the threshold percentage for the immune system.
(v) δ5 represents the percentage of immune cells that are suppressed.
A TIUNHD model is defined as a system of ODEs as follows:

dT
dt
= c1 T (1 − c2 T ) − c3 T I,
(3.3)
dI
dt
= δ1 − δ2 I + δδ43+T
IT
− δ5 T I.

In this chapter our aim to consider the fractional order TIUNHD model which is
defined as follows:  β
d T
dt β
= c1 T (1 − c2 T ) − c3 T I,
dβ I (3.4)
dt β
= δ1 − δ2 I + δδ43+T
IT
− δ5 T I,

where 0 < β ≤ 1.
The initial conditions are choose as T (0) = 1, I (0) = 1.22 [13].

4 Stability

For the stability analysis of the TIUNHD model, we refer to [15].

5 Function Approximation

A function g(t) in the interval [0, ∞) can be expanded in terms of fractional order
Laguerre polynomials as the following formula:
∞

g(t) = am L μm (t), (5.1)
m=0

where {a0 , a1 , a2 , . . .} are the unknowns coefficients, and derived by the following
formula:  ∞
am = μ wμ (t)g(t)L μm dt, m = 0, 1, 2, 3, . . . . (5.2)
0

Function approximation in terms of M number of fractional order Laguerre polyno-

mials are considered in particular applications as follows

g(t) = AL μ (t), (5.3)

A Study of the Fractional Tumour–Immune Unhealthy Diet … 143

where A is the row vector and L μ (t) is the column vector defined as A = [a0 a1
μ μ μ
a2 · · · a M ] and L μ (t) = [L 0 (t) L 1 (t) · · · L M (t)]T respectively, here T denotes
the transpose of the vector.

6 Pseudo-operational Matrix

The fractional Taylor functions are used to calculate the POM of integer and fractional
order, and they are defined as follows [42]:
μ
Ti = t iμ . (6.1)

Using Eq. (2.5), we may express L μ (t) in the form of Eq. (6.2) as follows:

L μ (t) = D1 Tμ (t), (6.2)

where D1 = [di j ]((M+1)×(M+1)) , and entries di j are defined as follows:


(−1) j (i+1)
(i− j+1)(( j+1))2
, if i ≥ j,
di j = i, j = 0, 1, 2, . . . , M. (6.3)
0, otherwise,
t
Now, for finding the value of 0 L μ ( p)dp, we follow the following steps:
 t  t  t
L μ ( p)dp = D1 Tμ ( p)dp = D1 Tμ ( p)dp
0 0 0 (6.4)
= t D1 H1 Tμ (t) = t D1 H1 D1−1 L μ (t).

If D1 H1 D −1 = Q 1 , then Eq. (6.4) implies

 t
L μ ( p)dp = t Q 1 L μ (t).
0

Here Q 1 is called the POM of integration with integer order, and entries h i j of the
matrix H1 = [h i j ] is defined as follows:

1
, if i = j,
hi j = iμ+1
i, j = 0, 1, 2, . . . , M. (6.5)
0, otherwise,

Similarly, we can find the POM of integration with fractional order, which is deter-
mined by using the following procedure:
144 S. Kumar and V. Gupta

J η (L μ (t)) = J η (D1 Tμ (t)) = D1 J η (Tμ (t))

(6.6)
= t η D1 R1 Tμ (t) = t η D1 R1 D1−1 L μ (t),

suppose D1 R1 D −1 = Q 2 . So, Eq. (6.6) implies

J η (L μ (t)) = t η Q 2 L μ (t).

In this case, Q 2 is called the POM of integration with fractional order, and entries
ri j of the matrix R1 = [ri j ] is defined as follows:

(iμ+1)
(iμ+1+η)
, if i = j,
ri j = i, j = 0, 1, 2, . . . , M. (6.7)
0, otherwise,

7 Numerical Method to Solve the Fractional TIUNHD

Model with Caputo’s Derivative

This section focuses on the solution of the fractional order TIUNHD model defined
in Eq. (3.4) by using the fractional order Laguerre polynomials.
First, we approximate the highest-order derivatives of the tumour cell and immune
system in terms of the fractional order Laguerre polynomials. So, we obtained the
following
dT dI
= AL μ (t), and = B L μ (t), (7.1)
dt dt
where,
⎡
⎤T ⎡
⎤T ⎡ ⎤
a0 b0 L 0 (t)
⎢ a1 ⎥ ⎢ b1 ⎥ ⎢ L 1 (t) ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
A=⎢ . ⎥ , B=⎢ . ⎥ , L μ (t) = ⎢ . ⎥ . (7.2)
⎣ .. ⎦ ⎣ .. ⎦ ⎣ .. ⎦
am bm L m (t)

Based on the initial conditions and integration with respect to variable t of Eq. (7.1),
we arrive at the following approximations:

T (t) = t AQ 1 Lμ(t) + 1, (7.3)

and
I (t) = t B Q 1 Lμ(t) + 1.22. (7.4)
dβ T dβ I
To find the values of dt
, and dt
, we have
A Study of the Fractional Tumour–Immune Unhealthy Diet … 145

dβ T  
1−β dT 1−β
= Jt = Jt AL μ (t)
dt dt (7.5)
1−β
= A Jt L μ (t) = t 1−β Q 2 L μ (t).
β
Similarly, we can find the value of ddtI .
β β
Substituting the values of ddtT , ddtI , T (t), and I (t) into the Eq. (3.4), we obtained
the system of equations; then, we collocate this system of equations at the Newton–
Cotes nodal points defined as the ti = (M+1)2
2i−1
. And, finally we obtained the unknowns
A and B, by using Newton’s iterative method. So, we obtained the solutions T (t)
and I (t).

8 Discussion of Numerical Results

Numerical results derived using the proposed numerical technique are discussed in
this section. Using the symbolic and algebraic software package Mathematica, we
demonstrate the numerical solution of the fractional TIUNHD model. This analysis
assumes some arbitrarily chosen orders, so the system’s complexity under considera-
tion can be illustrated. The derivatives of arbitrary orders are more accessible in terms
of degree, allowing for a wider range of geometry. Table 1 contains the parameter
values, and primary conditions used in numerical simulations are given as T (0) = 1
and I (0) = 1.22. Figure 1 represents the number of tumour and immune cell pop-
ulations with different time sequences. The behavior of the tumour and immune
cells are shown for the various fractional order derivatives in Fig. 2a–b. Our graphs
demonstrate that enhancing immune cells can reduce tumour cell production. If the
fractional order is smaller, the decay or growth process is faster. So, the fractional
order operators provide a better depiction of the dynamics of the relationship between
the growth and decay of Immune cells and the growth or decay of tumours.

Table 1 The values of Parameters Cost

parameters
c1 0.4426
c2 0.4
c3 0.1469
δ1 0.7
δ2 0.57
δ3 0.7829
δ4 0.8620
δ5 0.3634
146 S. Kumar and V. Gupta

1.35

1.30

1.25

1.20
Population

1.15

1.10

1.05 Tumour Cells

Immune Cells
1.00

0 1 2 3 4 5 6
t

Fig. 1 The behaviour of the Tumour and Immune cells during different timing in Days and the
given values of μ = 21 , M = 12, β = 1, c1 = 0.4426, c2 = 0.4, c3 = 0.1469, δ1 = 0.7, δ2 = 0.57,
δ3 = 0.7829, δ4 = 0.8620 and δ5 = 0.3634

Tumour Cells Immune Cells

1.29
1.14
1.28
1.12

1.10 1.27

1.08 1.26

1.06 1.25

1.04 1.24

1.02 1.23

t t
0.5 1.0 1.5 2.0 0.5 1.0 1.5 2.0

(a) (b)

Fig. 2 For β = {0.75, 0.90, 1}, μ = 21 , and M = 12 a The behaviour of Tumour cells concerning
with the time t (Days), b The behaviour of Immune cells concerning with the time t (Days)

9 Conclusion

This chapter generalizes the fractional TIUNHD model, which is governed by

the system of non-linear differential equations. We solved the fractional TIUNHD
model using the fractional order Laguerre polynomials POM technique. Additionally,
the fractional-order curves approximate integer-order when β → 1. The proposed
method is simple to implement and efficiently solve the mathematical non-linear
TIUNHD model converted to non-linear algebraic equations. The proposed method
may be extended to approximate new fractional and adaptive derivatives used to
model complicated events in nutrition, food science, and biological science. We will
likely achieve better accuracy in future communications by using other classes of
polynomials also.
A Study of the Fractional Tumour–Immune Unhealthy Diet … 147

References

1. P. Rawla, T. Sunkara, V. Gaduputi, Epidemiology of pancreatic cancer: global trends, etiology

and risk factors. World J. Oncol. 10(1), 10 (2019)
2. P. Anand, A.B. Kunnumakara, C. Sundaram, K.B. Harikumar, S.T. Tharakan, O.S. Lai, B.
Sung, B.B. Aggarwal, Cancer is a preventable disease that requires major lifestyle changes.
Pharm. Res. 25(9), 2097–2116 (2008)
3. H. Khan, F. Hussain, A. Samad, Cure and prevention of diseases with vitamin c into perspective:
an overview. J. Crit. Rev 7(4), 289–293 (2019)
4. D. Jafari, A. Esmaeilzadeh, M. Mohammadi-Kordkhayli, N. Rezaei, Vitamin c and the immune
system, in Nutrition and Immunity (Springer, 2019), pp. 81–102
5. W.K. Jhang, D.H. Kim, S.J. Park, W. Dong, X. Liu, S. Zhu, D. Lu, K. Cai, R. Cai, Q. Li et al.,
Development of the anti-cancer food scoring system 2.0: Validation and nutritional analyses
of quantitative anti-cancer food scoring model. Nutr. Res. Pract. 14(1), 32–44 (2020)
6. A.E. Glick, A. Mastroberardino, An optimal control approach for the treatment of solid tumors
with angiogenesis inhibitors. Mathematics 5(4), 49 (2017)
7. M. Chipo, W. Sorofa, E. Chiyaka, Assessing the effects of estrogen on the dynamics of breast
cancer. Comput. Math. Methods Med.
8. S. Khajanchi, M. Perc, D. Ghosh, The influence of time delay in a chaotic cancer model. Chaos:
Interdiscip. J. Nonlinear Sci. 28(10), 103101 (2018)
9. S. Kumar, S. Das, S.-H. Ong, Analysis of tumor cells in the absence and presence of chemother-
apeutic treatment: the case of caputo-fabrizio time fractional derivative. Math. Comput. Simul.
190, 1–14 (2021)
10. M.A. Alqudah, Cancer treatment by stem cells and chemotherapy as a mathematical model
with numerical simulations. Alex. Eng. J. 59(4), 1953–1957 (2020)
11. V.A. Kuznetsov, I.A. Makalkin, M.A. Taylor, A.S. Perelson, Nonlinear dynamics of immuno-
genic tumors: parameter estimation and global bifurcation analysis. Bull. Math. Biol. 56(2),
295–321 (1994)
12. D. Kirschner, J.C. Panetta, Modeling immunotherapy of the tumor-immune interaction. J. Math.
Biol. 37(3), 235–252 (1998)
13. S.A. Alharbi, A.S. Rambely, A dynamic simulation of the immune system response to inhibit
and eliminate abnormal cells. Symmetry 11(4), 572 (2019)
14. R.A. Ku-Carrillo, S.E. Delgadillo, B. Chen-Charpentier, A mathematical model for the effect
of obesity on cancer growth and on the immune system response. Appl. Math. Model. 40(7–8),
4908–4920 (2016)
15. S.A. Alharbi et al., The effectiveness of cryotherapy in the management of sports injuries.
Saudi J. Sports Med. 20(1), 1 (2020)
16. C. Mufudza, W. Sorofa, E.T. Chiyaka, Assessing the effects of estrogen on the dynamics of
breast cancer. Comput. Math. Methods Med. (2012)
17. Y. Louzoun, C. Xue, G.B. Lesinski, A. Friedman, A mathematical model for pancreatic cancer
growth and treatments. J. Theor. Biol. 351, 74–82 (2014)
18. J. Arciero, T. Jackson, D. Kirschner, A mathematical model of tumor-immune evasion and
sirna treatment. Discrete & Contin. Dyn. Syst.-B 4(1), 39 (2004)
19. X. Yang, Y. Yang, M.N. Skandari, E. Tohidi, S. Shateyi, A new local non-integer derivative
and its application to optimal control problems. AIMS Math. 7(9), 16692–16705 (2022)
20. H. Singh, Analysis for fractional dynamics of ebola virus model. Chaos Solitons & Fractals
138, 109992 (2020)
21. H. Singh, Analysis of drug treatment of the fractional hiv infection model of cd4+ t-cells. Chaos
Solitons & Fractals 146, 110868 (2021)
22. H. Singh, D. Baleanu, J. Singh, H. Dutta, Computational study of fractional order smoking
model. Chaos Solitons & Fractals 142, 110440 (2021)
23. H. Singh, H. Srivastava, D. Baleanu, Methods of Mathematical Modeling: Infectious Disease
(Academic, London, 2022)
148 S. Kumar and V. Gupta

24. S. Kumar, V. Gupta, An application of variational iteration method for solving fuzzy time-
fractional diffusion equations. Neural Comput. Appl. 33, 17659–17668 (2021)
25. M.K. Kadalbajoo, V. Gupta, Hybrid finite difference methods for solving modified Burgers
and Burgers-Huxley equations. Neural Parallel Sci. Comput. 18(3–4), 409–422 (2010)
26. V. Gupta, M.K. Kadalbajoo, A singular perturbation approach to solve Burgers-Huxley equa-
tion via monotone finite difference scheme on layer-adaptive mesh. Commun. Nonlinear Sci.
Numer. Simul. 16(4), 1825–1844 (2011)
27. V. Gupta, M.K. Kadalbajoo, Qualitative analysis and numerical solution of Burgers’ equation
via B-spline collocation with implicit Euler method on piecewise uniform mesh. J. Numer.
Math. 24(2), 73–94 (2016)
28. Z. Tang, E. Tohidi, F. He, Generalized mapped nodal laguerre spectral collocation method for
volterra delay integro-differential equations with noncompact kernels. Comput. Appl. Math.
39(4), 1–22 (2020)
29. P. Xiaobing, X. Yang, M.H.N. Skandari, E. Tohidi, S. Shateyi, A new high accurate approximate
approach to solve optimal control problems of fractional order via efficient basis functions.
Alex. Eng. J. 61(8), 5805–5818 (2022)
30. F.A. Ghassabzadeh, E. Tohidi, H. Singh, S. Shateyi, Rbf collocation approach to calculate
numerically the solution of the nonlinear system of qfdes. J. King Saud Univ.-Sci. 33(2),
101288 (2021)
31. H.M. Srivastava, K.M. Saad, W.M. Hamanah, Certain new models of the multi-space fractal-
fractional kuramoto-sivashinsky and korteweg-de vries equations. Mathematics 10(7), 1089
(2022)
32. S. Sahoo, V. Gupta, Second-order parameter-uniform finite difference scheme for singularly
perturbed parabolic problem with a boundary turning point. J. Differ. Equ. Appl. 27(2), 223–240
(2021)
33. M. Kadalbajoo, V. Gupta, A parameter uniform b-spline collocation method for solving sin-
gularly perturbed turning point problem having twin boundary layers. Int. J. Comput. Math.
87(14), 3218–3235 (2010)
34. R. Dubey, V. Gupta, A mesh refinement algorithm for singularly perturbed boundary and
interior layer problems. Int. J. Comput. Methods 17(7), 1950024 (2020)
35. S. Sahoo, V. Gupta, Higher order robust numerical computation for singularly perturbed prob-
lem involving discontinuous convective and source term. Math. Meth. Appl. Sci. 45(8), 4876–
4898 (2022)
36. J. Munkhammar, Riemann-Liouville Fractional Derivatives and the Taylor-riemann Series
(2004)
37. S. Kumar, V. Gupta, An approach based on fractional-order lagrange polynomials for the
numerical approximation of fractional order non-linear volterra-fredholm integro-differential
equations. J. Appl. Math. Comput. 1–22 (2022)
38. S. Kumar, V. Gupta, J. Gómez-Aguilar, An efficient operational matrix technique to solve the
fractional order non-local boundary value problems. J. Math. Chem. 1–17 (2022)
39. J. He, P. Shang, Multidimensional scaling analysis of financial stocks based on kronecker-delta
dissimilarity. Commun. Nonlinear Sci. Numer. Simul. 63, 186–201 (2018)
40. H. Dehestani, Y. Ordokhani, M. Razzaghi, Fractional-order legendre-laguerre functions and
their applications in fractional partial differential equations. Appl. Math. Comput. 336, 433–453
(2018)
41. K. Abernathy, Z. Abernathy, A. Baxter, M. Stevens, Global dynamics of a breast cancer com-
petition model. Diff. Equ. Dyn. Syst. 28(4), 791–805 (2020)
42. V.S. Krishnasamy, S. Mashayekhi, M. Razzaghi, Numerical solutions of fractional differential
equations by using fractional taylor basis. IEEE/CAA J. Autom. Sinica 4(1), 98–106 (2017)
Analysis of a Fractional Stage-Structured
Model With Crowley–Martin Type
Functional Response by Lagrange
Polynomial Based Method

Chandrali Baishya and P. Veeresha

Abstract The dynamics of a stage-structured predator-prey system that replicates

interactions between two densities of prey and predator populations were investi-
gated in this work. The adult predator population and the juvenile predator are the
two compartments that make up the predator population in the model. The predator
relies on both prey and juvenile predator, which is another element of the paradigm
that can be termed cannibalism. Crowley–Martin type functional denotes the nature
of the interaction between prey and adult predators, while Holling type-I functional
response denotes the nature of contact between juvenile and adult predators. The con-
cept of memory is introduced in the form of the Caputo fractional derivative to reflect
the complicated dynamics of interaction among the species. As a result, the model
is able to incorporate all relevant historical information about the occurrence, from
its inception to the desired time, into its calculations. We have also investigated the
boundedness and existence and uniqueness of solutions to the proposed model. The
condition of existence and stability of various points of equilibrium are investigated.
The numerical simulations are performed by using the Lagrange polynomial-based
method which is novel in the field of mathematical biology. Simulations have been
accomplished to examine the significance of parameters related to cannibalism, the
conversion rate from prey to adult predator, harvesting of an adult predator, and
growth rate of juvenile predators on the overall behavior of the system. The notewor-
thy performance of the fractional operator on the anticipated predator-prey model’s
dynamical behavior is well demonstrated by numerical results.

Keywords Stage-structured model · Caputo fractional derivative ·

Crowley–Martin type functional response · Holling type · Lagrange polynomial
based method

C. Baishya
Department of Studies and Research in Mathematics, Tumkur University, Tumkur 572103, India
P. Veeresha (B)
Department of Mathematics, Center for Mathematical Needs, CHRIST (Deemed to be
University), Bengaluru 560029, India
e-mail: [email protected]; [email protected]

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 149
H. Singh and H. Dutta (eds.), Computational Methods for Biological Models,
Studies in Computational Intelligence 1109,
https://doi.org/10.1007/978-981-99-5001-0_7
150 C. Baishya and P. Veeresha

2020 Subject Classification 34A34 · 70K20 · 92D40 · 26A33

1 Introduction

The study of the dynamics of species in the form of a dynamical system is one
of the most captivating issues in all of academia. It has gone beyond the realm of
species, and these systems have been used to represent a variety of natural phenomena
such as hydrodynamics [1], chemical reactions [2], plasma physics; social science
and economics issues such as business, finance, defense, criminology, epidemiol-
ogy, and others [3, 4]. These are the addition to novel applications on population
dynamics. In the literature of species dynamics stage-structured models have signif-
icant importance because every species naturally have physical growth in stages. So,
many mathematical biologists have contributed to the evaluation of stage-structured
predator-prey models [5–10]. The factors like predator harvesting, prey refuge, dis-
ease in prey/predator, delay, etc. have an impact on stage-structured models just like
they do on other predator-prey models. [11, 12].
The practice of eating another member of the same species as food is known as
cannibalism. Cannibalism comes in a variety of forms, including ritualistic, sacrifi-
cial, and survival cannibalism. Cannibalism is a widely prevalent ecological interface
that has been seen in more than 1,500 species of animals. In ecology, cannibalism
acts as a controller of the structure of a population. In [13], authors included forty
experimental research on cannibalism in crows, copepods, beetles, and fish species,
several scorpions, and others, taking cannibalism as a density control parameter. In
the field of mathematical biology, cannibalism has gotten a lot of attention. When
population densities are high and resources are few, cannibalism appears to occur in
many animals [14]. Depending on the system under examination, it might have desta-
bilizing or stabilizing impacts on the species. In [15], cannibalism is depicted as a tool
of stabilization in a prey-predator model. For growth, in a system of stage-structured
species, young species cohabit alongside adult species while neither reproducing nor
predating. Only adult predators can attack prey and breed, and almost all species in
the environment have an adult and juvenile stage structure. The pressure of preda-
tion on prey lessens as giant con-specifics consume small predators, which has a
substantial impact on organism dynamics. Cannibalism occurs within a species, and
its systems differ structurally from those of other populations [13, 16]. In [17], the
conditions that led to the evolutionary emergence of cannibalism are created, as well
as the criteria for cannibalism’s persistence in organisms are studied. Including can-
nibalism in predator species, author in [18] investigates a prey-predator system. A
cannibalistic predator-prey model has been proposed and examined in [19] where the
predator population is infected with a transmissible disease. In their research, Deng
et al. [20] have demonstrated that cannibalism affects the stability of the system in
both favorable and unfavorable ways. In [21], authors have analyzed that cannibalism
has significant impact on the evolution of predator species. On the other hand, the
Analysis of a Fractional Stage-Structured Model … 151

dynamic growth of a population is significantly impacted by harvesting. Population

harvesting is a natural part of wildlife management, fishing, and other industries.
Constant harvesting [22–24], proportional harvesting [25] and nonlinear harvesting
[26] are the most common harvesting components included in mathematical models.
An excellent harvesting strategy incorporating stage structure in predator species
was investigated in [27]. Including cannibalism in the predator, authors in [28] con-
structed a stage-structured predator-prey system which was later analyzed further by
Zhang et al. [29].
The functional response (FR), on the other hand, is a unique and noteworthy com-
ponent that plays a key role in all predator-prey interactions [30]. Some of the notable
FRs are Holling type I, II, III, and IV FR [31], Bedington-DeAngelis FR [32–34],
Leslie-Gower FR[35], square root FR [36], etc. When both prey and predator popu-
lations are large, the Crowley–Martin FR (CMFR) is used as a predator interference
function. The predator-prey model involving CMFR was originally put forward by
Crowley and Martin [37]. This FR is comparable to the Beddington–Deangelis func-
tional reaction. This varies only by one extra term, which describes how predators
interfere with one another when their prey is densely packed. Assuming that the
larger density of the predator population is what causes the depletion in the feeding
rate of predators [38], in this formulation the feeding rate per capita is:
γ uv
g(u, v) = ,
(1 + au)(1 + bv)

where u and v are the population density of the prey and predator respectively [38]. γ
is the rate of capture effect, a denotes the holding time and b indicates the magnitude
of interference among predators. In [39, 40], authors studied the global stability and
persistence in a model with CMFR where the prey population has two growth stages.
More research related to the CMFR can be found in [27, 37, 38, 41, 42]. Other than
the predator-prey dynamics, involvement of CMFR can also be noticed in the works
related to Human Immunodeficiency Virus (HIV), viral dynamics models, and other
infectious [43–46]. In [47], authors have used CMFR to observe a phytoplankton-
zooplankton system with delay. In [48, 49], for a prey-predator model with CMFR,
non-autonomous stochastic analysis is performed. Further to analyze bifurcation and
chaos control involving two delay times, discrete-time CMFR is used in [50].
Although fractional differentiation is as old as the classical order derivative, the
concept has gotten a lot of attention and has been studied extensively since 1967,
when it was first modified as part of the research in Caputo’s work [51]. Recently,
for modeling numerous real-world phenomena, rigorous applications of fractional
calculus can be noticed in several fields such as chemistry, image and signal process-
ing, control theory, electricity, biology, mechanics, and economics [52–54] by using
fractional derivatives (FDs), involving Caputo, Riemann–Liouville, Riesz, Weyl,
Grünwald–Letnikov, Marchaud and Hifler, Atangana–Baleanu, and Caputo–Fabrizio
operators. The most crucial subjects include history-dependent process, power law,
Levy statistics, anomalous diffusion, vibration and control, random walk, porous
media, special functions, chaos and groundwater problems, Riesz potential, singu-
152 C. Baishya and P. Veeresha

larities analysis and integral representations, fractional Brownian motion, computa-

tional fractional differential equations, non-Fourier heat conduction, FD and frac-
tals, nonlocal phenomena, biomedical engineering, fractional variational principles,
relaxation, fractional predator-prey system, rheology, soft matter mechanics, frac-
tional transforms, fractional wavelet, fractional signal and image processing, acous-
tic dissipation, geophysics, fluid dynamics, viscoelasticity. Supporting documenta-
tion of the above-mentioned works can be found in [55–58]. Riemann–Liouville
and Caputo FD received special attention from the research community over the
decades. Continuity and differentiability of an arbitrary function are not mandatory
with Riemann–Liouville FD at the origin. Benefits of the Caputo FD include the
ability to formulate models using standard initial and boundary conditions [51]. Fur-
thermore, under the Caputo FD, the derivative of a constant is zero. Although these
FDs have a lot of advantages, they cannot be used in every case. The fact that a con-
stant’s Riemann–Liouville derivative is not zero gives rise to various switches to the
modeling presumptions, and thus necessitates the generalization of the initial con-
ditions with fractional order. Additionally, when any arbitrary function is a constant
at the origin, the FD results in the singularity at it. Mittag–Leffler and exponen-
tial functions are examples of this. Only differentiable functions are defined for the
Caputo derivative. In the Riemann–Liouville sense, a function may nevertheless have
FDs of all orders lower than one even when they lack the first-order derivative. Due
to the singularity of the power-law-based FDs, the frequently employed fractional
differential operators (FDOs) have considerable limits when it comes to reproduc-
ing real-world issues. Caputo and Fabrizio brought in a novel fractional operator
to address this issue. They emphasized the fact that many physical phenomena are
non-singular and that utilizing singular operators to model such events could pro-
duce incorrect outcomes. An FDO with an exponential function as its kernel, known
as the Caputo–Fabrizio FD, has been developed to tackle this issue. With classical
derivative, exponential functions generate an evolution partial differential equation
and they are eigenfunctions of the decay differential ordinary equations [59]. Atan-
gana and Baleanu in 2016, to incorporate time non-locality into the system with a
non-singular kernel, introduced the generalized Mittag–Leffler function.
Numerous academics have investigated stage structure predator-prey models
incorporating FDs [60–70]. Recent literature contains some investigations on the
fractional cannibalistic model [71, 72]. In the present study, with Holling type I and
CMFR with the Caputo FD, we have examined a stage structure predator-prey model.
Even though some works on the fractional-order stage structure model can be found
in the literature on predator-prey dynamics, however, incorporation of cannibalism
with CMFR under the influence of FD is a novel contribution in this regard. Another
important aspect that we have integrated into the proposed model is that juvenile
predators do not reproduce even though they capture prey.
The structure of the remaining part of this chapter is presented here. In Sect. 2,
some definitions concerning the Caputo fractional-order systems are given. The pro-
posed fractional-order stage structure cannibalistic model is presented in Sect. 3. In
Sect. 5, the existence and uniqueness of the solution of the proposed system are ana-
lyzed. Section 6 concerns the boundedness of solutions. The dynamics of the system
Analysis of a Fractional Stage-Structured Model … 153

are presented in Sect. 6. Numerical simulations are performed in Sect. 7 to verify the
obtained analytical results. Section 8 contains the conclusion of the present work.

2 Preliminaries

In the present work, we have used the Caputo FDs because it supports the integer
order initial condition. In this section, we have presented certain theorems that have
been applied to determine the theoretical results corresponding to the solution of the
projected model.

Definition 2.1 ([56] Caputo FD) Suppose g(t) is k times continuously differentiable
function and g (k) (t) is integrable in [t0 , T ]. The FD of the order α for g(t) in the
Caputo sense is
 t
α 1 g (k) (τ )
Dt g(t) = dτ, (1)
(k − α) t0 (t − τ )α+1−k

where k is a positive integer with the property k − 1 < α < k and t > a. Here,  (·)
refers to Gamma function.

Lemma 2.2 ([73]) Consider the system

Dtα v(t) = g(t, v), t > t0 , (2)

choosing the initial condition as v(t0 ), where 0 < α ≤ 1 and g : [t0 , ∞) ×  →

Rn ,  ∈ Rn . Equation (2) has a unique solution on [t0 , ∞) × , when g(t, v) holds
the locally Lipchitz conditions concerning to v.

Lemma 2.3 ([74]) g(t) is a continuous function on [t0 , +∞) satisfying

Dtα g(t) ≤ −g(t) + ξ, g(t0 ) = g0 ,

where t0 ≥ 0 is the initial time, 0 < α ≤ 1,  = 0, (, ξ ) ∈ R2 . Then,

 
ξ ξ
g(t) ≤ g(t0 ) − E α [−(t − t0 )α ] + . (3)
 

3 Mathematical Model Formulation

In this chapter, a stage-structured cannibalism model has been analyzed intercon-

necting three species. We have considered two stages of the predator population viz.
juvenile and adult predators. Here u denotes the prey population and v and w denote
juvenile and adult predator populations respectively. Preys are the victim of an attack
154 C. Baishya and P. Veeresha

by both juvenile and adult predators. However, juvenile predators do not reproduce.
Predator-prey interaction is represented by CMFR whereas cannibalism is denoted
by Holling type I FR. The proposed work is observed within the frame of the Caputo
FD. To the best of our knowledge incorporation of FDs in a cannibalistic model with
CMFR is novel in the literature.
 u γ uw
Dtα0 u = r u 1 − − βuv − ,
k (au + 1)(bw + 1)
Dtα0 v = w − d1 v − μv − σ vw, (4)
α qγ uw
Dt0 w = μv − d2 w + pσ vw + − ξ w.
(au + 1)(bw + 1)

Here, k denotes the carrying capacity and r denotes the specific growth rate of the
prey. positive parameters a, b, and γ are used to analyze the impact of handling
time, the level of predator interference, and the proportion of prey captured by adult
predators respectively. The parameter ξ is the harvesting effort of the adult predator
to protect the prey population, q is the conversion factor of the adult predator, and
d1 and d2 are the natural death rate of juvenile and adult predators respectively. The
parameter μ represents the proportionality constant of the conversion of juvenile
to adult predators,  represents the birth rate of the juvenile predator, and β is the
capture rate of prey by juvenile predators. The catchability rate of young predators
by adults is represented by sigma, and the associated conversion factor is denoted
by p.

4 Boundedness

Theorem 4.1 All the non-negative solutions of (4) that started in are uniformly
bounded.
Proof Let (t) = u(t) + v(t) + q1 w(t). Then
 
1
Dtα (t) = Dtα u(t) + Dtα v(t) + Dtα w(t)
q
 
u μ d2 pσ ξ
= ru 1 − − βuv + w − μv − d1 v − σ vw + − w + vw − w.
k q q q q

We have
  u   μ

1
Dtα (t) + L (t) =u r 1 − + L − μ + d1 − − L v − (ξ + d2 − q − L)
k q q
− βuv − σ ( p − q)).
Analysis of a Fractional Stage-Structured Model … 155
 
Let L be a positive constant. If L = min (μ + d1 − μq ), (d2 + ξ − q) , then

 
1 k(r + L) 2 k(r + L)2
Dtα (t) + L (t) ≤ − u− + ,
k 2 4

which implies that,

k(r + L)2
Dtα (t) + L (t) ≤ .
4
By Lemma 2.3,
  
k(r + L)2 k(r + L)2
0< (t) ≤ + (t0 ) − E α −L(t − t0 )α , t ≥ t0 .
4L 4L

We have,

k(r + L)2 k(r + L)2 k(r + L)2

+ (t0 ) − E α −L(t − t0 )α → , t → ∞.
4L 4L 4L

This implies that

k(r + L)2
(t) ≤ , t → ∞.
4L
Hence, all the solution associated for the system (4) that starting in limited in

k(r + L)2
= (u, v, w) ∈ + : u(t) + v(t) + w(t) ≤ + η, η > 0 ,
4L

for η being sufficiently small.

5 Existence and Uniqueness of Solutions of the Model

Here, we examine the uniqueness and existence of the solution of the fractional
differential Eq. (4) by employing the Banach fixed-point theorem.

Dtα [u(t)] = ϕ1 (t, u),

Dtα [v(t)] = ϕ2 (t, v),
Dtα [w(t)] = ϕ3 (t, w). (5)
156 C. Baishya and P. Veeresha

Then, Eq. (5) can be presented in Volterra type integral equation

 t
1
u(t) − u(0) = ϕ1 (τ, u(τ ))(t − τ )α−1 dτ,
(α) 0
 t
1
v(t) − v(0) = ϕ2 (τ, v(τ ))(t − τ )α−1 dτ,
(α) 0
 t
1
w(t) − w(0) = ϕ3 (τ, w(τ ))(t − τ )α−1 dτ. (6)
(α) 0

Theorem 5.1 In the region × [t0 , T ], where

 
= (u, v, w) ∈ R3 : max {|u|, |v|, |w|} ≤ M ,

and T < +∞, the Lipschitz condition holds and contraction occurs by the kernel
ϕ1 if
  +L)2
0 ≤ r + 2rkM + β M + γ M < 1, Where M = k(r4L .

Proof Consider the function u(t) and ū(t) such as:

  
 u γ uw ū γ ūw 
||ϕ1 (t, u) − ϕ1 (t, ū)|| =  
r u(1 − k ) − βuv − ((1 + au)(1 + bw)) − r ū(1 − k ) − β ūv − ((1 + a ū)(1 + bw)) 
 
2r M γw
≤ r+ + β M ||u − ū|| + ||u − ū||
k (1 + bw)(1 + au)(1 + a ū)
 
2r M
≤ r+ + β M ||u − ū|| + γ w||u − ū||
k
 
2r M
≤ r + βM + + γ M ||u − ū||
k
= ψ1 ||u − ū||, (7)
 
where ψ1 = r + β M + 2rkM + γ M . The Lipschitz condition true for ϕ1 and if
0 ≤ ψ1 < 1, then ϕ1 follows contraction. Identically, it can be proved for ϕ2 and ϕ3

||ϕ2 (t, v) − ϕ2 (t, v̄)|| ≤ ψ2 ||v − v̄||,

||ϕ3 (t, w) − ϕ3 (t, w̄)|| ≤ ψ3 ||w − w̄||, (8)

where ψ2 = (μ + d1 + σ M) and ψ3 = (d2 + 2M + pσ M + ξ + qγ M). Contrac-

tions occurs if 0 < ψi < 1, i = 2, 3.

Theorem 5.2 The solution of the system (4) exists and will be unique, if we attains
some tα such that
1
ψi tα < 1, for i = 1, 2, 3.
(α)
Analysis of a Fractional Stage-Structured Model … 157

Proof This proof is shown in three steps:

1. Using the system (4), the recursive form can be written as
 t
1
ζ1,n (t) = u n (t) − u n−1 (t) = (ϕ1 (τ, u n−1 ) − ϕ1 (τ, u n−2 ))(t − τ )α−1 dτ,
(α) 0
 t
1
ζ2,n (t) = vn (t) − vn−1 (t) = (ϕ2 (τ, vn−1 ) − ϕ2 (τ, vn−2 ))(t − τ )α−1 dτ,
(α) 0
 t
1
ζ3,n (t) = wn (t) − wn−1 (t) = (ϕ3 (τ, wn−1 ) − ϕ3 (τ, wn−2 ))(t − τ )α−1 dτ. (9)
(α) 0

The initial conditions are u 0 (t) = u(0), v0 (t)v(0), w0 (t) = w(0). When the first
equation is subjected to the norm, we get

||ζ1,n (t)|| = ||u n (t) − u n−1 (t)||

  t 
 1 
= (ϕ1 (τ, u n−1 ) − ϕ1 (τ, u n−2 ))(t − τ ) dτ 
α−1
(α) 0 
 t
1
≤ ||(ϕ1 (τ, u n−1 ) − ϕ1 (τ, u n−2 ))(t − τ )α−1 dτ ||. (10)
(α) 0

Utilizing the Lipschitz condition Eq. (7), we get:

 t
1
||ζ1,n (t)|| ≤ ψ1 ||ζ1,n−1 (τ )dτ ||. (11)
(α) 0

Correspondingly,
 t
1
||ζ1,n || ≤ ψ2 ||ζ2,n−1 (τ )dτ ||,
(α) 0
 t
1
||ζ3,n || ≤ ψ3 ||ζ3,n−1 (τ )dτ ||. (12)
(α) 0

From the result, we can write:


n 
n 
n
u n (t) = ζ1,i , vn (t) = ζ2,i , wn (t) = ζ3,i .
i=1 i=1 i=1

Employing Eqs. (11) and (12) recursively, one can get

 n
1
||ζ1,i || ≤ ||u n (0)|| ψ1 t ,
(α)
 n
1
||ζ2,i || ≤ ||vn (0)|| ψ2 t ,
(α)
 n
1
||ζ3,i || ≤ ||wn (0)|| ψ3 t . (13)
(α)
158 C. Baishya and P. Veeresha

As a result, the existence and continuity are illustrated.

2. To examine that the correlation (13) composed is the solution for the system (4),
we have imagine the following:

u(t) − u(0) = u n (t) − ϑ1n ,

v(t) − v(0) = vn (t) − ϑ2n ,
w(t) − w(0) = wn (t) − ϑ3n . (14)

We set object to prove required results, as

  t 
 1 

||ϑ1n (t)|| =  (ϕ1 (τ, u) − ϕ1 (τ, u n−1 ))dτ 
(α) . (15)
0

This yields,

1
||ϑ1n (t)|| ≤ ψ1 ||u − u n−1 ||. (16)
(α)

By recursively repeating the same technique, we get

 n+1
1
||ϑ2n (t)|| ≤ ψ1 t M. (17)
(α)

At some tα , we have
 n+1
1
||ϑ2n (t)|| ≤ ψ1 tα M. (18)
(α)

Equation (18) shows that, as n → ∞, ||ϑ1n (t)|| → 0 if (α) 1

tα < 1. In the same
manner, it can be shown that ||ϑ2n (t)|| and ||ϑ3n (t)|| trend to zero.
3. We shall now exhibit the system’s uniqueness in terms of solution (4). Assume
that the system (4) has a new set of solutions, namely û, v̂, and ŵ. Then, using the
first equation in Eq. (5) as a starting point, we write:
 t
1
u(t) − û(t) = (ϕ1 (τ, u) − ϕ1 (τ, û))dτ. (19)
(α) 0

The following equation emerges after employing the norm

 t
1
||u(t) − û(t)|| = ||(ϕ1 (τ, u) − ϕ1 (τ, û))dτ ||. (20)
(α) 0

We acquire the Lipschitz condition by applying it.

Analysis of a Fractional Stage-Structured Model … 159

1
||u(t) − û(t)|| ≤ ψ1 t||u − û||. (21)
(α)

This results yields, at some tα

 
1
||u(t) − û(t)|| 1 − ψ1 tα ≤ 0. (22)
(α)
 
Hence, 1 − 1
ψt
(α) 1 α
> 0, which implies ||u(t) − û(t)|| = 0. Such that u(t) =
û(t).

6 Existence and Stability of the Points of Equilibrium of

the System (4)

Here, we shall evaluate the existence and stability of the points of equilibrium of
the (4).

6.1 Existence of Points of Equilibrium

Now, by the help of the system, we have (4).

 u γ uw
ru 1 − − βuv − = 0,
k (au + 1)(bw + 1)
w − d1 v − μv − σ vw = 0,
qγ uw
μv − d2 w + pσ vw + − ξ w = 0. (23)
(au + 1)(bw + 1)

The significant non-negative equilibrium points of the system (4) are the trivial equi-
librium Q 0 (0, 0, 0), Prey free equilibrium Q 2 (0, v2 , w2 ), Predator free equilibrium
Q 1 (u 1 , 0, 0), and co-existence equilibrium Q 3 (u ∗ , v ∗ , w ∗ ).
1. Predator free equilibrium point Q1 (u 1 , 0, 0) = (k, 0, 0) always exists. 
2. Prey free equilibrium point is Q 2 0, (dσ1 +μ)(d 2 +ξ )−μ (d1 +μ)(d2 +ξ )−μ
( p(d1 +μ)−μ)
, σ ( p−(d2 +ξ ))
.
Q 2 exists if (d1 + μ) (d2 + ξ ) < μ, p (d1 + μ) < μ, and p < d2 + ξ. Second
and third conditions imply the first condition. Therefore, Q 2 exists if R1 < 1
μ
where R1 = (d1 +μ)(d 2 +ξ )
. The quantity R1 means the mutual growth of juvenile
and adult predators.
3. Coexistence equilibrium point Q 3 (u ∗ , v ∗ , w ∗ ) shall be obtained from the follow-
ing isocline:
160 C. Baishya and P. Veeresha
 u γw
r 1− − βv − = 0, (24)
k (au + 1)(bw + 1)
w − d1 v − μv − σ vw = 0, (25)
qγ uw
μv − d2 w + pσ vw + − ξ w = 0. (26)
(au + 1)(bw + 1)

We shall verify the existence of u ∗ and w ∗ from isoclines (24) and (26). This
will lead to the stability of v ∗ from isocline (24). From Eq. (25), we obtain v =
w
d1 +μ+σ w
. From Eq. (24), we obtain the following:
√
(a) For u = 0, we get w = bd1 r +bμr −β−γ μ−γ d1 +r σ + (d1 (γ −br )−bμr +β+γ μ−r σ )2 +4r (d1 +μ)(bβ−br σ +γ σ )
2(bβ−br σ +γ σ ) =
w1 .
w1 > 0 if
bβ − br σ + γ σ > 0. (27)

(b) For w =0, we get u = k >0.

γ β(d1 +μ)
(c) dw
du
= 1 rk − (1+au)awγ
2 (1+bw) , where  = − (1+au)2 (1+bw) − (μ+σ w+d1 )2
< 0.
dw
du
< 0, if
r awγ
> . (28)
k (1 + au)2 (1 + bw)

Then, the isocline (24) passes through the points (0, w1 ) and (k, 0) and w
decreases as u increases under the conditions (27) and (28).
From the Eq. (26), we obtain the following
(a) For u = 0, we get w = (dσ1 +μ)(d 2 +ξ )−μ
(−d2 −ξ + p)
= w2 . w2 > 0 due to the condition of
existence of prey free equilibrium point.
(b) dw = 1¯ (1+au)qwr ¯ qγ u
2 (1+bw) > 0., where  = (1+au)2 (1+bw) +
(μ+d1 )(μ+2 pσ w)+ pσ 2 w 2
du (μ+σ w+d1 )2
− (ξ − d2 ).
Clearly, the isocline (25) passes through the points (0, w2 ) and w increases as u
increases under the conditions.
The analysis performed above interpret that the two isoclines (24) and (26) inter-
sect at a unique point (u ∗ , v ∗ ) if

w2 < w1 . (29)

Theorem 6.1 Coexistence equilibrium point Q 3 (u ∗ , v ∗ , w ∗ ) exists and is unique if

the conditions (27), (28), and (29) are satisfied.
Analysis of a Fractional Stage-Structured Model … 161

6.2 Stability Analysis

Jacobian matrix corresponding to the system (4) is

− (au+1)γ2w(bw+1) − 2r u
k + r − βv β(−u) γu
− (au+1)(bw+1) 2

J (u, v, w) = 0 −d1 − μ − σ w  − σv .
γ qw γ qu
(au+1)2 (bw+1)
μ + pσ w (au+1)(bw+1)2
− d2 − ξ + pσ v
If all the eigenvalues λi , i = 1, 2, . . . , n, of the Jacobian matrix J (E), E being the
point of equilibrium, satisfy the condition
απ
|arg(eig(J (E)))| = |arg(λi )| > , i = 1, 2, . . . , n. (30)
2
Then the E is a stable point of equilibrium. To evaluate the eigenvalues we solve the
characteristic equation |J (E) − λi I | = 0.
Lemma 6.1 ([75]) Define the following characteristic equation

P(λ) = λn + A1 λn−1 + A2 λn−2 + · · · + An = 0. (31)

All the roots of the characteristic Eq. (31) satisfy the Eq. (30) if we have
1. For n = 1, the condition is A1 > 0. √
4 A2 −A21
2. For n = 2, the conditions are A1 > 0, 4 A2 > A21 , |tan −1 A1
| > απ
2
.
3. For n = 3, if the discriminant of the polynomial P(λ) is positive then necessary
and sufficient conditions to satisfy the Eq. (30) are

A1 > 0, A2 > 0, A1 A2 > A3 .

If the discriminant of the polynomial P( ) is negative then necessary and suffi-

cient conditions to satisfy the Eq. (30) are

A1 > 0, A2 > 0, A1 A2 = A3 .

4. For general n, An > 0 is the necessary condition for Eq. (30) to be admit.
• At the trivial equilibrium Q 0 (0, 0, 0) the eigenvalues of the Jacobian matrix are
λ0 = r,  
1
λ1 = 2 −d1 − d2 − μ − ξ − (d1 + d2 + μ + ξ ) 2 − 4 ((d1 + μ) (d2 + ξ ) − μ) ,
  
λ2 = 21 −d1 − d2 − μ − ξ + (d1 + d2 + μ + ξ ) 2 − 4 ((d1 + μ) (d2 + ξ ) − μ) .
The eigenvalue corresponding to the u-direction is the positive number r. Since
product of the eigen value is a negative quantity (d1 + μ) (d2 + ξ ) − μ, eigenval-
ues corresponding to the v and w direction are of opposite signs. Trivial equilibrium
Q 0 (0, 0, 0) is a saddle point.
• At predator free equilibrium Q 1 (k, 0, 0), the eigenvalue corresponding to u-
direction is the negative number −r. The product of the eigenvalues correspond-
ing to the v and w direction is P = − γ kq(d 1 +μ)
ak+1
+ (μ + d1 ) (d2 + ξ ) − μ. Since
162 C. Baishya and P. Veeresha

(d1 + μ) (d2 + ξ ) < μ, clearly P is a negative quantity. Therefore, Q 1 is a saddle

point.
• At the prey-free equilibrium point Q 2 , the characteristic equation of the Jacobian
matrix is
P1 (λ) = λ3 + A1 λ2 + A2 λ + A3 , (32)

wher e
γw
A1 = + d1 + d2 + μ + ξ − pσ v − r + σ w,
bw + 1
   
γw γw
A2 = r − (−d2 − ξ + pσ v) + r − (−d1 − μ − σ w)
bw + 1 bw + 1
γ qw( − σ v)
− + (−d1 − μ − σ w) (−d2 − ξ + pσ v) ,
bw + 1
   
γw γw
A3 = ( − σ v) r − (μ + pσ w) − r − (−d1 − μ − σ w) (−d2 − ξ + pσ v) .
bw + 1 bw + 1

The equilibrium point Q 2 is stable if the third condition of the Lemma 6.1 is
satisfied.
• The characteristic equation corresponding to the Jacobian matrix at Q 3 (u ∗ , v ∗ , w ∗ )
is
P2 (λ) = λ3 + B1 λ2 + B2 λ + B3 , (33)

where

B1 = −( 11 + 22 + 33 ),
B2 = 11 22 + 22 33 + 11 33 − 12 21 − 23 32 − 13 31 ,
B3 = 12 21 33 + 23 32 11 + 13 31 22 − 11 22 33 − 12 23 31 − 13 21 32 .

Here, i j , i, j = 1, 2, 3 are entries of the Jacobian matrix. The equilibrium point

Q 3 is stable if the third condition of the Lemma 6.1 is satisfied.

7 Numerical Simulation

The Lagrange interpolation polynomial has been used as the basis for a large number
of effective numerical techniques for the solution of linear and nonlinear DEs. In the
setting of classical nonlinear systems, the Adams–Bashforth method, which makes
use of the Lagrange interpolation polynomial, has been recognized as an effective
numerical technique for solving complex system. The Adams–Bashforth has been
expanded due to its accuracy and effectiveness within the context of fractional inte-
gration and differentiation. But the majority of these studies employ polynomials
with linear Lagrange interpolation. In this section, we have presented the Lagranges
polynomial-based numerical method described in [76, 77] to solve the system (4),
Analysis of a Fractional Stage-Structured Model … 163

which is a modified Adams–Bashforth and take into account the approximation at

two successive places.
We consider a general nonlinear Cauchy’s problem
C
Dtα x(t) = φ(t, x(t)), 0 ≤ t ≤ T, x(0) = x0 . (34)

Corresponding to (34), the Volterra integral equation is as follows:

 t
1
x(t) = x0 + (t − s)α−1 φ(s, x(s))ds, x(0) = x0 . (35)
(α) 0

Considering a time interval divided into (n − 1) subintervals such that at a given

time t = tn , Eq. (35) becomes
 tn
1
x(tn ) = x0 + φ(s, x(s))(tn − s)α−1 ds. (36)
(α) 0

Equation (36) can be transformed into a sum as


1 
n−1 tn+1
x(tn ) = x0 + φ(s, x(s))(tn − s)α−1 ds. (37)
(α) k=0 tn

The function f (s, x(s)) is approximated by linear Langrange polynomial. On inte-

gration from Eq. (37) we get

hα 
n−1
hα 
n−1
x(tn ) = x0 + φ(sk , x(sk ))ck,n − φ(sk−1 , x(sk−1 ))τk,n ,
α(α + 2) α(α + 2)
k=1 k=1
(38)
where

ck,n = (n − k)α (2α + n − k − 1) − (n − k − 1)α (n − k + 1 + 2α), (39)

α+1 α
τk,n = (n − k) − (n − k − 1) (n − k + α). (40)

Again, subdivide the time interval into n subintervals such that at a given time t =
tn+1 , Eq. (38) becomes
 tn+1
1
x(tn+1 ) = x0 + φ(s, x(s))(tn+1 − s)α−1 ds. (41)
(α) 0
164 C. Baishya and P. Veeresha

From Eq. (41) we get,


1 
n tn+1
x(tn+1 ) = x0 + φ(s, x(s))(tn+1 − s)α−1 ds. (42)
(α) k=0 tn

Equation (42) yields

hα 
n
hα 
n
x(tn+1 ) = x0 + φ(sk , x(sk ))δk,n − φ(sk−1 , x(sk−1 ))μk,n ,
α(α + 2) α(α + 2)
k=1 k=1
(43)
where

δk,n = (n − k)α (2α + n − k − 1) − (n − k − 1)α (n − k + 2 + 2α), (44)

α+1 α
μk,n = (n − k − 1) − (n − k) (n − k + 1 + α). (45)

Subtracting (38) from (43) we obtain the formula to solve the Eq. (34) as follows:

hα n 
n−1
x(tn+1 ) = x(tn ) + φ(sk , x(sk ))δk,n − φ(sk , x(sk ))ck,n
α(α + 2) k=1 k=1

hα n 
n−1
− φ(sk−1 , x(sk−1 ))μk,n − φ(sk−1 , x(sk−1 ))τk,n .
α(α + 2) k=1 k=1
(46)

Using the method described in Eq. (46), the formula corresponding to the system (4)
becomes

n    
hα ui γ u i wi
u n+1 = u n + δi,n r u i 1 − − β − βu i vi −
α(α + 2) k (au i + 1)(bwi + 1)
i=1

n−1    
ui γ u i wi
− ci,n r u i 1 − − β − βu i vi −
k (au i + 1)(bwi + 1)
i=1
n    
u i−1 γ u i−1 wi−1
− δi,n r u i−1 1 − − βu i−1 vi−1 −
k (au i−1 + 1)(bwi−1 + 1)
i=1

n−1    
u i−1 γ u i−1 wi−1
+ c1,n r u i−1 1 − − βu i−1 vi−1 − ,
k (au i−1 + 1)(bwi−1 + 1)
1=1
(47)
Analysis of a Fractional Stage-Structured Model … 165

n
hα
vn+1 = vn + δi,n (wi − d1 vi − μvi − σ vi wi )
α(α + 2) i=1

n−1
− ci,n (wi − d1 vi − μvi − σ vi wi )
i=1
(48)
n
− δi,n (wi−1 − d1 vi−1 − μvi−1 − σ vi−1 wi−1 )
i=1


n−1 
+ c1,n (wi−1 − d1 vi−1 − μvi−1 − σ vi−1 wi−1 ) ,
1=1

  
hα
n
qγ u i wi
wn+1 = wn + δi,n μvi − d2 wi + pσ vi wi + − ξ wi
α(α + 2) (au i + 1)(bwi + 1)
i=1

n−1  
qγ u i wi
− ci,n μvi − d2 wi + pσ vi wi + − ξ wi
(au i + 1)(bwi + 1)
i=1

n  
qγ u i−1 wi−1
− δi,n μvi−1 − d2 wi−1 + pσ vi−1 wi−1 + − ξ wi−1
(au i−1 + 1)(bwi−1 + 1)
i=1

n−1  
qγ u i−1 wi−1
+ c1,n μvi−1 − d2 wi−1 + pσ vi−1 wi−1 + − ξ wi−1 .
(au i−1 + 1)(bwi−1 + 1)
1=1
(49)
For numerical simulation, we considered r = 3.5, k = 70, β = 0.01, γ = 5,  =
2.5, p = 0.5, q = 0.2, μ = 0.25, d1 = 1.2, d2 = 0.2, σ = 0.2, ξ = 1,
a = 0.01, b = 1.2. Initial populations are taken as u 0 = 1, v0 = 1, w0 = 1. For
these values, the points of equilibrium are Q 0 (0, 0, 0), Q 1 (70, 0, 0), Q 2
(0, 11.7368, 111.5), Q 3 (2.77162, 4.26873, 3.75984). That means the complex
dynamic of the system features the existence of two positive equilibria. Eigenvalues
of the Jacobian matrix are respectively {3.5, −2.12539, −0.524609},
{−3.5, 3.05634, −1.58869}, {−23.8231, −0.635757, 0.0468033}, {−2.68984,
−0.101439 + 0.47488i, −0.101439 − 0.47488i}. This indicate that coexistance
point of equilibrium Q 3 is stable spiral and Q 2 is unstable. In Fig. 1, we can view
the convergence to the coexistence points of equilibrium. Clearly, as the value of the
FD α decreases, the system attains stability faster than before.
In this section, we also looked at the impact of one of the parameters on the
dynamics of the system while maintaining the other parameters constant. In Figs. 2,
3, 4, we have presented the effect of the rate of cannibalism on the solution profile for
α = 1, α = 0.95, α = 0.9 respectively. It is observed that a stable system in Fig. 1
converted to unstable limit cycle in Fig. 2 with the increase in the rate if cannibalism.
This means an increase in the cannibalism rate destabilizes the system. Figure 2 also
indicates that loss of stability by a Hopf bifurcation will take place as the level of
cannibalism increases. But the presence of FD stabilizes the system which can be
observed in Figs. 3 and 4.
166 C. Baishya and P. Veeresha

Fig. 1 Nature of the system (4) for distinct α

Analysis of a Fractional Stage-Structured Model … 167

Fig. 2 Impact of cannibalism rate on the population of the system (4) for α = 1
168 C. Baishya and P. Veeresha

Fig. 3 Impact of cannibalism rate on the population of the system (4) for α = 0.95
Analysis of a Fractional Stage-Structured Model … 169

Fig. 4 Impact of cannibalism rate on the population of the system (4) for α = 0.9
170 C. Baishya and P. Veeresha

Figures 5, 6, and 7 represent the effect of change in the conversion rate of prey
to adult predators. It is noticed that an increase in this conversion rate leads to a
decrease in the prey population and an increase in the juvenile population. This
indicates that the adult predators’ need for food is satisfied with catching prey and
their attentions are diverted from juvenile predators. However, juvenile predator
increase is comparatively slow due to the cannibalism factor. Again FD has been
observed as a stabilizing tool here.
In Figs. 8 and 9 we have seen the influence of the harvesting rate on the population.
An increase in the harvesting rate of adult predator influence in increasing the prey
population and decrease the juvenile predator population. Fractional derivative helps
in achieving stability faster. Figures 10 and 11 help us understand the impact of
growth rate of juvenile predators on the population. It is observed that this growth
rate affects the prey population negatively.

8 Conclusion

The analysis of a stage structure predator-prey model presented in this chapter in the
context of the Caputo FD. Combination of the Caputo FD, CMFR, Cannibalism, and
non-reproductivity of the juvenile predator after consuming prey make the proposed
model unique from the available stage structure models in the literature. The existence
and stability of trivial, predator-free, prey-free, and co-existence points of equilibrium
points are examined. We have observed that, the solution of the system shows a switch
from stable spiral behavior to stable focus in some cases and unstable to stable in other
cases as the order of the FD reduces. One of the significantly noticeable points is that
the increase in the rate of cannibalism destabilizes the system. Moreover, as the level
of cannibalism increases, loss of stability takes place by a Hopf bifurcation. But the
presence of FDs reduces the destabilization. It can also be noticed that the conversion
factor is directly proportional to the growth of juvenile predator. The effects of
capturing these observations are effectively validated by numerical simulations in
this chapter using the Lagrange polynomial-based method. The simulated results
show the efficiency of the numerical technique. As a future direction of studies, one
can always look for chaotic dynamics which may occur in a system as a result of
cannibalism and then find a way to control them.
Analysis of a Fractional Stage-Structured Model … 171

Fig. 5 Impact of conversion rate of prey to adult predator on the population for α = 1
172 C. Baishya and P. Veeresha

Fig. 6 Impact of conversion rate of prey to adult predator on the population for α = 0.95
Analysis of a Fractional Stage-Structured Model … 173

Fig. 7 Impact of conversion rate of prey to adult predator on the population for α = 0.9
174 C. Baishya and P. Veeresha

Fig. 8 Effect of harvesting rate of adult predator on the population for α = 0.95

Fig. 9 Effect of harvesting rate of adult predator on the population for α = 0.9
Analysis of a Fractional Stage-Structured Model … 175

Fig. 10 Effect of growth rate of juvenile predator on the population for α = 0.95
176 C. Baishya and P. Veeresha

Fig. 11 Effect of growth rate of juvenile predator on the population for α = 0.95
Analysis of a Fractional Stage-Structured Model … 177

References

1. F.H. Busse, Transition to turbulence via the statistical limit cycle route, in Chaos and Order in
Nature, ed. by H. Haken. Springer Series in Synergetics (Springer, 1981), pp. 36–44
2. R.H. Hering, Oscillations in Lotka-Volterra systems of chemical reactions. J. Math. Chem.
5(2), 197–202 (1990)
3. S. Solomon, P. Richmond, Stable power laws in variable economies; Lotka-Volterra implies
pareto-zipf. Eur. Phys. J. B - Condens. Matter Complex Syst. 27(2), 257–261 (2002)
4. J. Sooknanan, B. Bhatt, D. Comissiong, A modified predator-prey model for the interaction of
police and gangs. R. Soc. Open Sci. 3(9) (2016)
5. W. Wang, L. Chen, A predator-prey system with stage-structure for predator. Comput. & Math.
Appl. 33, 83–91 (1997)
6. J. Cui, L. Chen, W. Wang, The effect of dispersal on population growth with stage-structure.
Comput. & Math. Appl. 39(1), 91–102 (2000)
7. S. Liu, E. Beretta, A stage-structured predator-prey model of Beddington-DeAngelis type.
SIAM J. Appl. Math. 66(4), 1101–1129 (2006)
8. l. You Hong, R. Yuan, A stage-structured predator-prey model with two delays due to juvenile
maturation. Acta Mathematicae Applicatae Sinica, English Series (2011)
9. S.G. Mortoja, P. Panja, S.K. Mondal, Dynamics of a predator-prey model with stage-structure
on both species and anti-predator behavior. Inf. Med. Unlocked 10, 50–57 (2018)
10. R.K. Naji, S.J. Majeed, The dynamical analysis of a prey-predator model with a refuge-stage
structure prey population. Int. J. Diff. Equ. 2016 (2016)
11. X. Zhang, L. Chen, A. Neumann, The stage-structured predator-prey model and optimal har-
vesting policy. Math. Biosci. 168(2), 201–210 (2000)
12. G. Tang, S. Tang, R. Cheke, Global analysis of a Holling type-II predator-prey model with a
constant prey refuge. Nonlinear Dyn. 76, 635–647 (2014)
13. G.A. Polis, The evolution and dynamics of intraspecific predation. Annu. Rev. Ecol. Syst.
12(1), 225–251 (1981)
14. L. Pizzatto, R. Shine, The behavioral ecology of cannibalism in cane toads (bufo marinus).
Behav. Ecol. Sociobiol. 63(1), 123–133 (2008)
15. C. Kohlmeier, W. Ebenhoh, The stabilizing role of cannibalism in a predator-prey system. Bull.
Math. Biol. 57(3), 401–411 (1995)
16. L. Fox, Cannibalism in natural populations. Ann. Rev. Ecol. Syst. 6, 87–106 (2003)
17. P.H. Getto, O. Diekmann, A.M. Roos, On the (dis) advantages of cannibalism. J. Math. Biol.
51(6), 695–712 (2005)
18. R. Marik, L. Pribylova, An age-structured model of cannibalism. Electr. J. Diff. Equ. (2006)
19. B. Santosh, S. Sudip, C. Joydev, Cannibalistic predator–prey model with disease in predator -
a delay model. Int. J. Bifurc. Chaos 25(10) (2015)
20. D. Hang, C. Fengde, Z. Zhenliang, L. Zhong, Dynamic behaviors of Lotka-Volterra predator–
prey model incorporating predator cannibalism. Adv. Diff. Equ. 359 (2019)
21. P. Mishra, S.N. Raw, B. Tiwari, On a cannibalistic predator-prey model with prey defense and
diffusion. Appl. Math. Model. 90, 165–190 (2022)
22. J. Luo, Y. Zhao, Stability and bifurcation analysis in a predator–prey system with constant
harvesting and prey group defense. Int. J. Bifurc. Chaos 27 (2017)
23. L. Zhang, W. Wang, Y. Xue, Spatiotemporal complexity of a predator-prey system with constant
harvest rate. Chaos Solitons & Fractals 41(1), 38–46 (2009)
24. C.R. Zhu, K.Q. Lan, Phase portraits, hopf bifurcations and limit cycles of Leslie-Gower
predator-prey systems with harvesting rates. Discrete Contin. Dyn. Syst. - B 14(1) (2010)
25. P. Lenzini, J. Rebaza, Nonconstant predator harvesting on ratio-dependent predator-prey mod-
els. Appl. Math. Sci. (2010)
26. R.P. Gupta, M. Banerjee, P. Chandra, Bifurcation analysis and control of Leslie-Gower predator-
prey model with Michaelis-Menten type prey-harvesting. Diff. Equ. Dyn. Syst. 20(3), 339–366
(2012)
178 C. Baishya and P. Veeresha

27. B. Dubey, S. Agarwal, A. Kumar, Optimal harvesting policy of a prey-predator model with
Crowley-Martin-type functional response and stage structure in the predator. Nonlinear Anal.:
Modell. Control 23(4), 493–514 (2018)
28. K.G. Magnusson, Destabilizing effect of cannibalism on a structured predator-prey system.
Math. Biosci. 155(1), 61–75 (1999)
29. F. Zhang, Y. Chen, J. Li, Dynamical analysis of a stage-structured predator-prey model with
cannibalism. Math. Biosci. 307, 33–41 (2019)
30. S. Zhang, S. Yuan, T. Zhang, A predator-prey model with different response functions to
juvenile and adult prey in deterministic and stochastic environments. Appl. Math. Comput.
413 (2022)
31. S.G. Mortoja, P. Panja, S.K. Mondal, Dynamics of a predator-prey model with nonlinear inci-
dence rate, Crowley-Martin type functional response and disease in prey population. Ecol.
Genet. Genomics 10 (2019)
32. J.R. Beddington, Mutual interference between parasites or predators and its effect on searching
efficiency. J. Anim. Ecol. 331–340 (1975)
33. D.L. DeAngelis, R.A. Goldstein, R.V. O’Neill, A model for tropic interaction. Ecology 56(4),
881–892 (1975)
34. S. Khajanchi, Dynamic behavior of a Beddington-DeAngelis type stage structured predator-
prey model. Appl. Math. Comput. 244, 344–360 (2014)
35. M.A. Aziz-Alaoui, M.D. Okiye, Boundedness and global stability for a predator-prey model
with modified Leslie-Gower and Holling-type II schemes. Appl. Math. Lett. 16(7), 1069–1075
(2003)
36. P.A. Braza, Predator-prey dynamics with square root functional responses. Nonlinear Anal.
Real World Appl. 13(4), 1837–1843 (2012)
37. P.H. Crowley, E.K. Martin, Functional responses and interference within and between year
classes of a dragonfly population. J. N. Amer. Benthol. Soc. 8(3), 211–221 (1989)
38. R.K. Upadhyay, R.K. Naji, Dynamics of a three species food chain model with Crowley-Martin
type functional response. Chaos Solitons & Fractals 42(3), 1337–1346 (2009)
39. Y. Meng, Xin, H.F. Huo, H. Xiang, Q.y. Yin, Stability in a predator–prey model with Crowley–
Martin function and stage structure for prey. Appl. Math. Comput. 232, 810–819 (2014)
40. A.P. Maiti, B. Dubey, A. Chakraborty, Global analysis of a delayed stage structure prey-predator
model with Crowley-Martin type functional response. Math. Comput. Simul. 162, 58–84 (2019)
41. J.P. Tripathi, S. Tyagi, S. Abbas, Global analysis of a delayed density dependent predator-prey
model with Crowley-Martin functional response. Commun. Nonlinear Sci. Numer. Simul.
30(1), 45–69 (2016)
42. B. Ghanbari, H. Gunerhan, H.M. Srivastava, An application of the Atangana-Baleanu fractional
derivative in mathematical biology: a three-species predator-prey model. Chaos Solitons &
Fractals 138 (2020)
43. X. Zhou, J. Cui, Global stability of the viral dynamics with Crowley-Martin functional response.
Bull. Korean Math. Soc. 48(3), 555–574 (2011)
44. S. Xu, Global stability of the virus dynamics model with Crowley-Martin functional response.
Electron. J. Qual. Theory Differ. Equ. 2012(9), 1–10 (2012)
45. C. Kang, H. Miao, X. Chen, J. Xu, D. Huang, Global stability of a diffusive and delayed virus
dynamics model with Crowley-Martin incidence function and CTL immune response. Adv.
Diff. Equ. 2017(1), 1–16 (2017)
46. N. Kumari, N. Mohan, Cross diffusion induced turing patterns in a tritrophic food chain model
with Crowley-Martin functional response. Mathematics 7(3) (2019)
47. T. Liao, H. Yu, M. Zhao, Dynamics of a delayed phytoplankton-zooplankton system with
Crowley-Martin functional response. Adv. Diff. Equ. 2017(1), 1–30 (2017)
48. Y. Zhang, S. Chen, S. Gao, Analysis of a nonautonomous stochastic predator-prey model with
Crowley-Martin functional response. Adv. Diff. Equ. 2016(1), 1–28 (2016)
49. C. Xu, G. Ren, Y. Yu, Extinction analysis of stochastic predator–prey system with stage structure
and Crowley–Martin functional response. Entropy 21(3) (2019)
Analysis of a Fractional Stage-Structured Model … 179

50. C. Liu, S. Li, Y. Yan, Hopf bifurcation analysis of a density predator-prey model with Crowley-
Martin functional response and two time delays. J. Appl. Anal. & Comput. 9(4), 1589–1605
(2019)
51. M. Caputo, Linear models of dissipation whose q is almost frequency independent-II. Geophys.
J. Int. 13(5), 529–539 (1967)
52. S.J. Achar, C. Baishya, P. Veeresha, L. Akinyemi, Dynamics of fractional model of biological
pest control in tea plants with Beddington–DeAngelis functional response. Fractal Fract. 6(1)
(2021)
53. C. Baishya, S.J. Achar, P. Veeresha, D.G. Prakasha, Dynamics of a fractional epidemiological
model with disease infection in both the populations. Chaos: Interdiscip. J. Nonlinear Sci. 31(4)
(2021)
54. C. Baishya, Dynamics of fractional holling type-ii predator-prey model with prey refuge and
additional food to predator. J. Appl. Nonlinear Dyn. 10(2), 315–328 (2020)
55. A.A. Kilbas, H.M. Srivastava, J. Trujillo, Theory and Applications of Fractional Differential
Equations, vol. 204 (Elsevier, North-Holland Mathematics Studies, 2006)
56. I. Podlubny, Fractional Differential Equations (Academic, 1999)
57. B. Ross, A brief history and exposition of the fundamental theory of fractional calculus, in
Fractional Calculus and Its Applications: Proceedings of the International Conference Held
at the University of New Haven, June 1974 (Springer, 1975), pp. 1–36
58. R.N. Premakumari, C. Baishya, M.K.A. Kaabar, Dynamics of a fractional plankton-fish model
under the influence of toxicity, refuge, and combine-harvesting efforts. J. Inequal. Appl. 137
(2022)
59. M. Caputo, M. Fabrizio, A new definition of fractional derivative without singular kernel. Prog.
Fract. Diff. Appl. 1(2), 73–85 (2015)
60. M. Sambath, R. Perumal, K. Balachandran, Stability analysis of fractional-order stage structure
prey-predator model. Indian J. Ind. Appl. Math. 11, 208–221 (2020)
61. C. Baishya, Dynamics of fractional stage structured predator prey model with prey refuge.
Indian J. Ecol. 47(4), 1118–1124 (2020)
62. L. Dai, J. Wang, Y. Ni, B. Xu, Dynamical analysis of a new fractional-order predator-prey
system with Holling type-III functional. Adv. Diff. Equ. 2021(1), 76 (2021)
63. I.U. Haq, Y. Mehmet, A. Nigar, A. Akgul, A SARS-CoV-2 fractional-order mathematical model
via the modified Euler method. Math. Comput. Appl. 5, 82 (2022)
64. O. Fatma, Y. Mehmet, S.M, Tamer, H. Rafla, Fractional order modelling of omicron SARS-
CoV-2 variant containing heart attack effect using real data from the United Kingdom. Chaos
Solitons & Fractals 157 (2022)
65. S. Harendra, Analysis for fractional dynamics of Ebola virus model. Chaos Solitons & Fractals
138(109992) (2020)
66. P. Veeresha, E. Ilhan, H.M. Baskonus, Fractional approach for analysis of the model describing
wind-influenced projectile motion. Physica Scripta 96(7), 075209 (2021)
67. S. Harendra, Analysis of drug treatment of the fractional HIV infection model of CD4+ T-cells.
Chaos Solitons & Fractals 164(110868) (2021)
68. S. Harendra, Computational study of fractional order smoking model. Chaos Solitons & Fractals
142(110440) (2021)
69. S. Harendra, S. Hari, B. Dumitru, Methods of Mathematical Modelling, 1st ed. (Elsevier, 2022)
70. R.N. Premakumari, C. Baishya, P. Veeresha, A fractional atmospheric circulation system under
the influence of a sliding mode controller. Symmetry 14(12) (2022)
71. L. Ning, Y. Mengting, Bifurcation control of a delayed fractional-order prey-predator model
with cannibalism and disease. Phys. A: Stat. Mech. Appl. 600(127600) (2022)
72. R. Maya, S. Agus, K. Muharini Wuryansari, D. Isnani, Dynamical analysis of a predator-prey
model incorporating predator cannibalism and refuge. Axioms 11(3) (2022)
73. B. Wang, L.-Q. Chen, Asymptotic stability analysis with numerical confirmation of an axially
accelerating beam constituted by the standard linear solid model. J. Sound Vib. 328(4–5),
456–466 (2009)
180 C. Baishya and P. Veeresha

74. H.-L. Li, L. Zhang, C. Hu, Y.-L. Jiang, Z. Teng, Dynamical analysis of a fractional-order
predator-prey model incorporating a prey refuge. J. Appl. Math. Comput. 54(1), 435–449
(2017)
75. E. Ahmed, A.S. Elgazzar, On fractional order differential equations model for nonlocal epi-
demics. Phys. A 379(2), 607–614 (2007)
76. S.S. Alzaid, B.S.T. Alkahtani, Modified numerical methods for fractional differential equations.
Alex. Eng. J. 58, 1439–1447 (2019)
77. T. Mekkaoui, A. Abdon, New numerical approximation of fractional derivative with non-local
and non-singular kernel: application to chaotic models. Eur. Phys. J. - Plus 132, 444 (2017)
Qualitative Theory and Approximate
Solution to Norovirus Model Under Non
Singular Kernel Type Derivatives

Eiman, Waleed Ahmed, Kamal Shah, and Thabet Abdeljawad

Abstract We study a fractional order nonlinear dynamical system of Norovirus dis-

ease. The considered mathematical model consists of susceptible, exposed, vacci-
nated, infected, and recovered classes. Firstly, the Caputo-Fabrizio fractional deriva-
tive (CFFD) is used to examine the qualitative theory and approximate solutions.
Some adequate results for the existence of approximate solutions are given using
fixed point theory. We use the Laplace Transform (LT) and the Adomian decom-
position (ADM) approach to get approximate results for each compartment. The
graphical presentation that corresponds to available real facts for different fractional
orders is given. Secondly, by using piecewise global fractional derivatives in sense
of singular and non-singular kernels, the aforementioned system is also investigated.
We establish sufficient conditions for the existence and uniqueness of the solution
to the proposed model by using the fixed-point approach. Numerical simulation is
performed by extending Newton’s interpolation formula for the considered model
under the mentioned derivatives. Graphical presentations for various compartments
of the model are given against the available real data for different fractional orders.
The numerical findings are performed by using Matlab 16.

Keywords Nonlinear dynamical system · Norovirus · Approximate solution ·

ADM · CFFD · Crossover behavior · Mathematical biology · Piecewise global
fractional derivatives · Newton interpolation formula

Eiman · W. Ahmed · K. Shah

Department of Mathematics, University of Malakand, Chakdara Dir(L) 18000 Khyber
Pakhtunkhwa, Pakistan
K. Shah · T. Abdeljawad (B)
Department of Mathematics and Sciences, Prince Sultan University,
P.O. Box 66833, 11586 Riyadh, Saudi Arabia
e-mail: [email protected]
T. Abdeljawad
Department of Medical Research, China Medical University, Taichung 40402, Taiwan

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 181
H. Singh and H. Dutta (eds.), Computational Methods for Biological Models,
Studies in Computational Intelligence 1109,
https://doi.org/10.1007/978-981-99-5001-0_8
182 Eiman et al.

1 Introduction

Nonlinear dynamical systems are useful for modeling a variety of real-world phe-
nomena and processes. Algebraic, difference and differential/integral equations are
widely used to represent dynamic systems. Nonlinear equations have been used to
model the majority of real-world processes and phenomena. Engineering, physics,
chemistry, control theory, and mathematical biology are among most of the fields
where the indicated area has applicability. Many diseases have affected human pop-
ulation throughout the history but the most dangerous of which are viral diseases.
Diseases like Measles, TB, malaria, HBV, HCV, Dengue fever, malignant malig-
nancies, Spanish flu etc that primarily impact human populations have resulted in
millions of deaths. As a result, they have developed a variety of tactics for limiting
and decreasing the rate of infections in their communities. Most epidemics have
been limited by the discovery of treatments and cures but still, there are some dis-
eases that have no proper medications or having no treatment i.e incurable. Among
these diseases, one of the infectious diseases is Norovirus disease. For more than 75
years, outbreaks of non-bacterial gastroenteritis have been linked to viruses. Because
of the increased in incidence during the winter months, Zahorsky [1] during 1929,
suggested the name “winter vomiting disease” to investigate outbreak. It should be
kept in mind that the initial outbreak was discovered in a village Norwalk of USA
in 1968. The virus was first reported in children and workers at a local elementary
school. It was named Norwalk virus at that time [2]. Later on the name Norovirus
was approved by an international committee of taxonomy in 2002.
Norovirus is considered the main reason for a fifth of all acute gastroenteritis
(AGE) cases each year worldwide [3–5]. And is linked to a majority of morbidity,
mortality, and healthcare costs. The transmission dynamics of this relatively prevalent
virus at the community level are still poorly known, in part due to underreporting of
sickness. According to serological studies, the first Norovirus infection occurs early
in infancy [6, 7]. AGE is responsible for over a quarter of all fatalities in children
under the age of five in Africa and Southeast Asia [8]. Norovirus is responsible for
the bulk of non-bacterial outbreaks and an estimated 18 percent of all endemic AGE
[9, 10]. In the United Kingdom alone, Norovirus is predicted to cost 81 million
pounds per year to healthcare systems and people [11]. Norovirus is spread mostly
through fecal, oral transmission, which might be aided by vomiting episodes [1].
Other factors, on the other hand, are more frequently linked to outbreaks spread
by the environment [12, 13]. This can happen through the contamination of food,
fluid, and surfaces [14, 15]. Vomiting, diarrhea, muscle aches, low-grade fever or
chills, headache, and stomach cramping are some of the symptoms of an infected
person. Symptoms normally appear 12 to 48 hours after the virus is consumed and
last 1 to 3 days [16, 17]. But vomiting and diarrhoea can seriously dehydrate people,
especially elderly people, young children, and people who already have medical con-
ditions. Therefore, necessary medical attention is required. There is a dose-response
association, meaning that people who are exposed to more virus particles have a
Qualitative Theory and Approximate Solution … 183

greater attack rate [18–20]. However, the infectious dose is modest, with only a 49
percent chance of infection from a single particle, implying that a dosage response
is improbable.
Many variables can aggravate Norovirus transmission and the resulting sickness.
According to researchers , the virus has a high seasonality, with over half of infections
occurring during the cold season. Seasonal variation is attributed to both environ-
mental factors and population behavior. For example, Norovirus thrives in lower
temperatures and may be aided by increased rainfall. Other factors in the population
can also influence the intensity of Norovirus outbreaks. The virus affects people of
all ages, but it is most common in children under the age of five. Because Norovirus
is a virus, it cannot be treated with antibiotics. To prevent the problem one has to
manage dehydration induced by vomiting and diarrhea. Also, hand washing with
soap and water is an excellent way to reduce Norovirus infection transmission. Here,
we refer some more details on the mentioned viral disease as [21–24, 26–29, 34].
It is remarkable that researchers are progressively using new tools and analy-
sis to investigate epidemiological diseases. The mathematical formulation of the
aforementioned phenomena is a strong tool. Mathematical models are becoming a
powerful technique for making predictions and analyzing the transmission dynamics
of infectious illnesses. Usually daily problems, such as infectious diseases are non-
linear in nature. This has led to a long standing interest in nonlinear mathematical
models that can be used to describe many real-world problems. As a result, multiple
mathematical models have been developed to describe various diseases. Similarly,
nonlinear differential equations have been used by numerous authors to formulate
the Norovirus infection. Since the majority of research regarding Norovirus infec-
tion has been based on mathematical models using classical derivatives. Classical
derivatives are local in nature and lack a higher degree of freedom. As a result, most
dynamic problems cannot be stated accurately and efficiently. As a result, mathe-
maticians have determined that utilizing non-integer order derivatives is superior to
traditional ones for describing a variety of real-world phenomena and processes. It
should be noted that fractional order integral and differential operators are nonlocal
in nature and have the ability to provide global dynamics which include the tradi-
tional order dynamics as a special case (see some detail in [30, 31]). As a result,
the researchers were drawn to examine numerous dynamical issues in context of
fractional calculus as can be seen in [32–34]. The said differential operators can be
divided into two kinds of classes on the basis of their kernels. Riemann-Liouville
and Caputo operators have singular kernel-type operators so placed in one of the
classes. On the other hand, Caputo-Fabrizio and Atangana-Baleanu, etc are placed
in the non-singular kernel type class. Both kinds have a variety of applications in the
mathematical modeling of various real-world problems (see details in [35–39]). For
more recent contribution, we refer some work as [40–43].
184 Eiman et al.

2 Formulation of Proposed Model

A model is designed that further categories all of the population members as:
H denotes susceptible, V vaccinated, exposed people are denoted by U, infected
people are represented by A,and recovered class C. A schematic dia gram of the
model is given in Fig. 1.
The classical form of the model we need to study is recalled from [44] as
⎧
⎪ ηH(t)A(t)
⎪
⎪ Ḣ(t) = ∧ − − (ρ + d)H(t),
⎪
⎪ N
⎪
⎪
⎪
⎪ (1 − τ )ηA(t)V(t)
⎪
⎪ V̇(t) = ρH(t) −
⎪
⎪ N
− dV(t),
⎨
ηH(t)A(t) (1 − τ )ηA(t)V(t)
⎪ U̇(t) = + − (β + d)U(t), (1)
⎪
⎪ N N
⎪
⎪
⎪
⎪ Ȧ(t) = βU(t) − (δ + d)A(t),
⎪
⎪
⎪
⎪ Ċ(t) = δA(t) − dC(t),
⎪
⎪
⎩
H(0) = H0 , V(0) = V0 , U(0) = U0 , A(0) = A0 , C(0) = C0 ,

where H0 , V0 , U0 , A0 , C0 ≥ 0. Here, ∧ is the recruitment rate into the susceptible

class, η is the effective control rate while ρ is the vaccination coverage rate, and d
is the natural morality rate. The rate of recovery is given by δ, β denotes developing
clinical symptoms and τ is vaccine effectiveness. Where N is total population in t.
N = H + V + U + A + C.

Fig. 1 A flow chart of the proposed model (1)

Qualitative Theory and Approximate Solution … 185

Fractional calculus is nowadays the most popular branch of calculus among

researchers. But dealing with FODEs, in particular, has received a lot more attention.
Researchers have examined fractional differential operators from a variety of per-
spectives, including existence theory, stability, and numerical methods (see details
[45, 46]). They developed a number of strategies, methodologies, and theories for
computing precise or numerical FODE solutions. For the desired results, researchers
employ well-known techniques such as HPM, ADM and a variety of other numerical
methods see [47–49]. Motivated by this work, in our project we handle the model
(1) under CFFD as
⎧
⎪ CF σ ηHA
⎪
⎪ Dt H(t) = ∧ − − (ρ + d)H,
⎪
⎪ N
⎪
⎪
⎪
⎪ (1 − τ )ηVA
⎪
⎪
CF σ
Dt V(t) = ρH − − dV,
⎪
⎪ N
⎨
CF σ ηHA (1 − τ )ηVA
⎪ Dt U(t) = + − (β + d)U, (2)
⎪
⎪ N N
⎪
⎪ CF σ
⎪
⎪ Dt A(t) = βU − (δ + d)A,
⎪
⎪
⎪
⎪ CF σ
Dt C(t) = δA − dC,
⎪
⎪
⎩
H(0) = H0 , V(0) = V0 , U(0) = U0 , A(0) = A0 , C(0) = C0 ,

where H0 , V0 , U0 , A0 , C0 ≥ 0 and σ ∈ (0, 1]. Firstly some qualitative results, such

as the existence and uniqueness of the solution that corresponds to the given model,
are proven. The fixed point theory of Krassnoselskii and Banach are employed to
obtain these results. To analyze some approximate analytical results, the LT and
ADM tools are also applied. The said method already has been used in many papers.
For instance see [50, 51]. Finally, graphical representations of approximation results
are produced using Matlab. For some more frequent results, we refer readers to see
[52–59].
Here we remark that we have studied the model (1) from different aspects. Our
analysis is devoted to the existence theory through a fixed point approach and ana-
lytical as well as the numerical study of the proposed model. The model has been
studied under the different concepts of fractional calculus. Some updated form of
fractional differential operators has been considered here in the form of piecewise
derivative. Both analytical and numerical results have been presented graphically by
using Matlab 16.

3 Preliminaries

Some basic definitions and results of FODEs are needed to obtain all these results.
Which are given as
Definition 1 ([60]) Let m ∈ H1 (0,  ),  > 0, σ ∈ (0, 1), then the CFFD is given
by
186 Eiman et al.
 t  
(M)(σ ) t−
CF
Dσt (m(t)) = 
m ( ) exp − σ d.
1−σ 0 1−σ

However if m does not belong to H1 (0,  ), the derivative is given as

 t  
(M)(σ ) t−
CF
Dσt (m(t)) = (m(t) − m( )) exp − σ d.
1−σ 0 1−σ

Definition 2 ([60]) For m ∈ H1 (0,  ),  > 0, the CF integral is given as

 t
(1 − σ ) σ
CF
Iσt [m(t)] = m(t) + m( )d, σ ∈ (0, 1].
(M)(σ ) (M)(σ ) 0

Definition 3 ([60]) If M(σ ) = 1, then the transform due to Laplace of C F Dσt m(t)
can be defined as

sL [m(t)] − m(0)
L [C F Dσt m(t)] = , s ≥ 0, σ ∈ (0, 1].
s + σ (1 − s)

If J = [0,  ],  < ∞, then Z = C ([0,  ] × R2 , R) is Banach space under

the norm (x, y) = supt∈J [|x(t)| + |y(t)|].

Theorem 1 ([61]) Y ⊂ Z and there exist operators P1 , P2 such that

1. P1 v + P2 v ∈ Y for every v ∈ Y;
2. P1 is contraction;
3. P2 is completely continuous,
then P1 v + P2 v = v has at least one solution.

Definition 4 ([62, 63]) Let U ∈ H 1 (0, τ ), then

  σ
σ ABC(σ ) t
d −σ
0 D t (U (t)) = U (α)E σ t −α
ABC
dα. (3)
1−σ 0 dα 1−σ

Definition 5 ([64]) We define


σ (1 − σ )U (t) t
σ (t − α)σ −1
0 I t U (t) = + U (α)dα.
ABC
(4)
ABC(σ ) 0 ABC(σ ) (σ )

Definition 6 ([65]) Let U , then the arbitrary order derivative in Caputo sense is
defined by
 
C σ
t
(t − α)n−σ −1 [U (α)]
0 Dt U (t) = dα.
0 (1 − σ )
Qualitative Theory and Approximate Solution … 187

Definition 7 For U , fractional integral is defined as

 t
C σ 1
0 It U (t) = (t − α)σ −1 dα, σ > 0.
(σ ) 0

Definition 8 ([66]) Consider U and g as differentiable and increasing functions

respectively, one has
⎧ t
⎪
⎪ U (τ )dτ, 0 < t ≤ t1 ,
⎨
0
0 It U (t) =  t
PG
⎪
⎪
⎩ U (τ )g  (τ )d(τ ), t1 < t ≤ T.
t1

Definition 9 If U is differentiable, one can define

PC σ
U  (t), 0 < t ≤ t1 ,
0 Dt U (t) = C σ
0 Dt U (t), t1 < t ≤ T.

Definition 10 Let U be usual differentiable, we define

⎧ t
⎪
⎪ U (τ )dτ, 0 < t ≤ t1 ,
⎨
0
PC
I t U (t) =  t
0
⎪
⎪ 1
⎩ (t − α)σ −1 U (α)d(α), t1 < t ≤ T.
σ t1

Definition 11 For differentiable function U in ABC sense, we have

C σ
PC ABC σ 0 Dt U (t), 0 < t ≤ t1 ,
0 Dt U (t) = ABC σ
0 Dt U (t), t1 < t ≤ T.

Definition 12 For differentiable function U , we have

⎧  t
⎪ 1
⎪
⎪
⎨ σ (t − α)σ −1 U (α)d(α), 0 < t ≤ t1 ,
PC ABC I U (t) = t1
t  t ,
0 ⎪
⎪ 1−σ σ
⎪
⎩ U (t) + (t − α)σ −1 U (α)d(α) t1 < t ≤ T.
ABCσ ABCσ σ t1

[65]

Lemma 3.13.1 The solution of piecewise problem

PC ABC σ
0 Dt U (t) = G(t, U (t)), 0 < σ ≤ 1
188 Eiman et al.

such that right sides vanish at t = 0 is given by

⎧  t
⎪
⎪ 1
⎨ U0 +
⎪ G(α, U (α))(t − α)σ −1 dα, 0 < t ≤ t1
(σ ) 0
U (t) = 
⎪
⎪
⎪ 1−σ t σ (t − α)σ −1
⎩ U (t1 ) + G(t, U (t)) + G(α U (α))d(α) t1 < t ≤ T.
ABC(σ ) t1 ABC(σ ) (σ )

4 Theoretical Results by Using CF Concept

To elaborate the existence result, applying the integral operator C F Iσt on both sides
of
CF σ
Dt U (t) = ϕ(t, U (t)),
(5)
U (0) = U0 ,

we get  t
(1 − σ )ϕ(t, U (t)) σ ϕ(, U ( ))
U (t) = U0 + + d, (6)
(M)(σ ) 0 (M)(σ )

where
⎧ ⎧ ⎧
⎪
⎪ H(t) ⎪
⎪ H0 ⎪
⎪ ϕ1 (t, H, V, U, A, C),
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪ ⎪ ⎪
⎪
⎨ V(t) ⎪ V
⎨ 0 ⎨ ϕ2 (t, H, V, U, A, C),
⎪
U (t) = U(t) , U0 = U0 , ϕ(t, U (t)) = ϕ3 (t, H, V, U, A, C), (7)
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ A(t) ⎪
⎪ A0 ⎪
⎪ ϕ4 (t, H, V, U, A, C),
⎪
⎪ ⎪
⎪ ⎪
⎪
⎩ ⎩ ⎩
C(t) C0 ϕ5 (t, H, V, U, A, C).

For further analysis, the following assumptions are taken into consideration.

(D1) If Lϕ > 0, and U, U¯ ∈ Z , one have

|ϕ(t, U(t)) − ϕ(t, U¯ (t))| ≤ Lϕ [|U − U¯ |].

(D2) Let Cϕ , Cϕ > 0 and Mϕ > 0, such that

|ϕ(t, U (t))| ≤ Cϕ |U | + Mϕ .

Using (6) and (7), the two operators are defined as:

(1 − σ )ϕ(t, U (t))
P1 (U ) = U0 (t) + ,
(M)(σ )
 t
σ ϕ(, U ( ))
P2 (U ) = d. (8)
0 (M)(σ )
Qualitative Theory and Approximate Solution … 189

Lϕ
Theorem 2 Problem (6) has at least one solution if (M)(σ )
< 1.

Proof Consider Y = {U ∈ Z : U  ≤ ρ, ρ > 0} be closed convex set of Z . If

U − U¯ ∈ Y, one has

(1 − σ )
P1 U − P1 U¯  = sup ϕ(t, U (t)) − (ϕ(t, U¯ (t))
(M)(σ ) t∈J
(1 − σ )
≤ Lϕ sup |U (t) − U¯ (t)|
(M)(σ ) t∈J
Lϕ
≤ U − U¯ .
(M)(σ )

Hence P1 is contraction. Now to show P2 is completely continuous, let for any

U ∈ Y, one has
 t
σ
P2 (U ) = sup ϕ(, U ( )) d
t∈J 0 (M)(σ )

≤ [Cϕ ρ + Mϕ ] := . (9)
(M)(σ )

From (9) we conclude that P2 is bounded. Continuity of σ implies continuity of P2 .

Similarly, it can be proved that P2 is equi-continues by setting t1 < t2 ∈ J. 
(1+ )L
Theorem 3 The problem (6) has a unique solution if (M)(σ )
< 1.

Proof Let define P : A → A by

 t
(1 − σ )ϕ(t, U (t)) σ ϕ(, U ( ))
P(U ) = U0 + + d.
(M)(σ ) 0 (M)(σ )

If U , U¯ ∈ A, one has

(1 − σ )
P(U ) − P(U¯ ) ≤ sup ϕ(t, U (t)) − ϕ(t, U¯ (t))
t∈J (M)(σ )
 t
σ
+ sup |ϕ(, U ( )) − ϕ(, U¯ ( ))|d
(M)(σ ) t∈J 0
(1 +  )L
≤ U − U¯ . (10)
(M)(σ )

Hence the result. 

190 Eiman et al.

5 Computational Results of (2)

Setting (M)(σ ) = 1, and applying LT on both sides of (2), and using (σ, s) =
1
s
[s + σ (1 − s)], one has
⎧
⎪ ηHA
⎪
⎪ L [H(t)] = s −1 H0 + (σ, s)L [∧ − − (ρ + d)H],
⎪
⎪ N
⎪
⎪
⎪
⎪ (1 − τ )ηVA
⎪ −1
⎨ L [V(t)] = s V0 + (σ, s)L [ρH −
⎪
N
− dV],

⎪ ηHA (1 − τ )ηVA (11)

⎪
⎪ L [U(t)] = s −1 U0 + (σ, s)L [ + − (β + d)U],
⎪
⎪ N N
⎪
⎪
⎪
⎪ L [A(t)] = s −1 A0 + (σ, s)L [βU − (δ + d)A],
⎪
⎪
⎩
L [C(t)] = s −1 C0 + (σ, s)L [δA − dC].

Assume
∞ ∞
H(t) = Hq (t), V(t) = Vq (t),
q=0 q=0
∞ ∞
U(t) = Uq (t), A(t) = Aq (t), (12)
q=0 q=0
∞
C(t) = Cq (t),
q=0

and
∞
H(t)A(t) = Aq (H, A),
q=0
∞
(13)
V(t)A(t) = Bq (V, A),
q=0

such that the “Adomian polynomial” Aq (H, A) can be defined as

 q q 
1 dq
Aq (H, A) = λ H j (t)
j
λ A j (t)
j
.
q! dλq j=0 j=0 λ=0

and the “Adomian polynomial” Bq (V, A) can be defined as

 q q 
1 dq
Bq (V, A) = λ V j (t)
j
λ A j (t)
j
.
q! dλq j=0 j=0 λ=0
Qualitative Theory and Approximate Solution … 191

Reference to (12) and (13), the system (5) becomes

⎧  ∞   ∞ ∞ 
⎪
⎪ H(0) Aq (H, A)
⎪
⎪ L Hq (t) = + (σ, s)L ∧ −η − (ρ + d) Hq (t) ,
⎪
⎪ s N
⎪
⎪ q=0 q=0 q=0
⎪
⎪
⎪
⎪  ∞   ∞ ∞ ∞ 
⎪
⎪ V(0) Bq (V, A)
⎪
⎪ L Vq (t) = + (σ, s)L ρ Hq (t) − (1 − τ )η −d Vq (t) ,
⎪
⎪
⎪
⎪
s N
⎪
⎪
q=0 q=0 q=0 q=0
⎪
⎪  ∞   ∞ ∞ ∞ 
⎨ U(0) Aq (H, A) Bq (V, A)
L Uq (t) = + (σ, s)L η + (1 − τ )η − (β + d) Uq (t) ,
⎪
⎪ s N N
⎪
⎪ q=0 q=0 q=0 q=0
⎪
⎪  ∞   
⎪
⎪ ∞ ∞
⎪
⎪ A(0)
⎪
⎪L Aq (t) = + (σ, s)L β Uq (t) − (δ + d) Aq (t) ,
⎪
⎪ s
⎪
⎪ q=0 q=0 q=0
⎪
⎪
⎪
⎪  ∞   ∞ ∞ 
⎪
⎪ C(0)
⎪ Cq (t) = + (σ, s)L δ Aq (t) − d Cq (t) .
⎩L
⎪
s
q=0 q=0 q=0
(14)

From (14), using initial conditions, we equate terms as

⎧
⎪ H0
⎪
⎪ L [H0 (t)] = + (σ, s)L (∧),
⎪
⎪ s
⎪
⎪
⎪
⎪ V0
⎪
⎪ L [V0 (t)] = ,
⎪
⎪ s
⎪
⎨
U0
L [U0 (t)] = , (15)
⎪
⎪ s
⎪
⎪
⎪
⎪ A0
⎪
⎪ L [A0 (t)] = ,
⎪
⎪ s
⎪
⎪
⎪
⎪ C0
⎩ L [C0 (t)] = .
s
⎧  
⎪ ηA0 (H, A)
⎪
⎪ L [H1 (t)] = (σ, s)L − − (ρ + d)H0 (t) ,
⎪
⎪ N
⎪
⎪  
⎪
⎪ (1 − τ )ηB0 (V, A)
⎪
⎪ L [V1 (t)] = (σ, s)L ρH0 (t) − − dV0 (t) ,
⎪
⎪
⎪
⎪ N
⎪
⎨  
ηA0 (H, A) (1 − τ )ηB0 (V, A)
L [U1 (t)] = (σ, s)L + − (β + d)U(t) ,
⎪
⎪ N N
⎪
⎪  
⎪
⎪
⎪
⎪ L [A1 (t)] = (σ, s)L βU0 (t) − (δ + d)A0 (t) ,
⎪
⎪
⎪
⎪
⎪
⎪  
⎪
⎪
⎩ L [C1 (t)] = (σ, s)L δA0 (t) − dC0 (t) .
(16)
192 Eiman et al.
⎧  
⎪
⎪ ηA1 (H, A)
⎪
⎪ L [H 2 (t)] = (σ, s)L − − (ρ + d)H1 (t) ,
⎪
⎪ N
⎪
⎪  
⎪
⎪ (1 − τ )ηB1 (V, A)
⎪
⎪ L [V (t)] = (σ, ρH (t) − − (t) ,
⎪
⎪ 2 s)L 1 dV 1
⎪
⎪ N
⎪
⎪  
⎪
⎪ ηA1 (S, I ) (1 − τ )ηB1 (V, A)
⎪
⎪ L [U (t)] = (σ, s)L + − (β + d)U (t) ,
⎪
⎨
2
N N
1
 
⎪
⎪ L [A (t)] = (σ, s)L βU (t) − (δ + d)A (t) ,
⎪
⎪ 2 1 1
⎪
⎪  
⎪
⎪
⎪
⎪
⎪
⎪ L [C (t)] = (σ, s)L δA (t) − dC (t) .
⎪
⎪
2 1 1
⎪
⎪
⎪
⎪ ..
⎪
⎪
⎪
⎪ .
⎪
⎩

⎧   (17)
⎪ ηAq (S, I )
⎪
⎪ L [Hq+1 (t)] = (σ, s)L − − (ρ + d)Hq (t) ,
⎪
⎪ N
⎪
⎪  
⎪
⎪ (1 − τ )ηBq (V, A)
⎪
⎪ L [V (t)] = (σ, ρH (t) − − (t) ,
⎪
⎪ q+1 s)L q dV q
⎪
⎪ N
⎪
⎨  
ηAq (S, I ) (1 − τ )ηBq (V, A)
L [Uq+1 (t)] = (σ, s)L + − (β + d)Uq (t) ,
⎪
⎪ N N
⎪
⎪  
⎪
⎪
⎪
⎪ L [A (t)] = (σ, βU (t) − (δ + (t) ,
⎪
⎪ q+1 s)L q d)A q
⎪
⎪
⎪
⎪  
⎪
⎪
⎩ L [Cq+1 (t)] = (σ, s)L δAq (t) − dCq (t) .
(18)
After performing simplification, we get
⎧
⎪
⎪ H0 (t) = H0 + ∧ 1 + (t − 1)σ ,
⎪
⎪
⎪
⎪
⎪
⎪
⎨ V0 (t) = V0 ,
(19)
⎪
⎪ U0 (t) = U0 ,
⎪
⎪
⎪
⎪ A0 (t) = A0 ,
⎪
⎪
⎩
C0 (t) = C0 ,
Qualitative Theory and Approximate Solution … 193
⎧
⎪ η
⎪
⎪ H1 (t) = H0 − A0 − (ρ + d) 1 + (t − 1)σ
⎪
⎪ N
⎪
⎪
⎪
⎪
⎪
⎪ η t2
⎪
⎨ + ∧ 1 − N A0 1 + 2σ (t − 1) + σ 1 − 2t + 2! ,
2

(20)
⎪
⎪ η(1 − τ )
⎪
⎪ V1 (t) = ρH0 − A0 V0 − dV0 1 + (t − 1)σ
⎪
⎪ N
⎪
⎪
⎪
⎪
⎪
⎪ t2
⎩ + ρ ∧ 1 + 2(t − 1)σ + 1 − 2t + σ2 ,
2!
⎧
⎪
⎪ η η(1 − τ )
⎪
⎪ U1 (t) = A0 H0 + V0 A0 − (β + d)U0 1 + (t − 1)σ
⎪
⎪ N N
⎪
⎪
⎪
⎪ η ∧ A0 t2
⎪
⎪ + + − + − + σ2 ,
⎨ N
1 2(t 1)σ 1 2t
2!
(21)
⎪
⎪
⎪
⎪ A1 (t) = βU0 − (δ + d)A0 1 + (t − 1)σ ,
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎩ C1 (t) = (δA0 − dC0 ) 1 + (t − 1)σ .

⎧ 
⎪
⎪ −η −η
⎪
⎪ H2 (t) = H0 βU0 − (δ + d)A0 + A0 A0 H0 − (ρ + d)H0
⎪
⎪ N N
⎪
⎪
⎪
⎪ −η t2
⎪
⎪ − (ρ + − (ρ + + − + − + σ2
⎪
⎪ d) A 0 H0 d)H 0 1 2(t 1)σ 1 2t
⎪
⎪ N 2!
⎪
⎨ η η
+ ∧ βU0 − (δ + d)A0 + A0 (1 − A0 ) − (ρ + d) 1 − A0
⎪
⎪ N N
⎪
⎪
⎪
⎪ t2
⎪
⎪ 1 + 3(t − 1)σ + 3 1 − 2t + σ2
⎪
⎪
⎪
⎪ 2!
⎪
⎪ 
⎪
⎪ t2 t3
⎪
⎩ − 1 − 3t + 3 − σ 3
,
2! 3!
 (22)
⎧ −η η(1 − τ ) η(1 − τ )
⎪
⎪ V2 (t) = ρ A0 H0 − (ρ + d)H0 − d + A0 ρH0 − V0 A0 − dV0
⎪
⎪ N N N
⎪
⎪
⎪
⎪ 
⎪
⎪ t2
⎪
⎨ + V0 βU0 − (δ + d)A0 )
⎪ 1 + 2(t − 1)σ + 1 − 2t +
2!
σ2

⎪
  (23)
⎪
⎪ η η(1 − τ )
⎪
⎪ + ρ ∧ 1 − A0 − A0 − d
⎪
⎪ N N
⎪
⎪
⎪
⎪ 2 2 3
⎪
⎩ 1 + 3(t − 1)σ + 3 1 − 2t + t σ 2 − 1 − 3t + 3 t − t σ 3 ,
2! 2! 3!
194 Eiman et al.
⎧ 
⎪
⎪ η −η
⎪
⎪ U2 (t) = H0 βU0 − (δ + d)A0 ) + A0 ( A0 H0 − (ρ + d)H0
⎪
⎪ N N
⎪
⎪ 
⎪
⎪ η(1 − τ ) η(1 − τ )
⎪
⎪ + ρH − − + βU − (δ +
⎪
⎪
A 0 0 V A
0 0 dV 0 V 0 0 d)A0
⎪
⎪ N N
⎪
⎨ t2
1 + 2(t − 1)σ + 1 − 2t + σ2
⎪
⎪ 2!
⎪ 
⎪ 
⎪
⎪ η η
⎪
⎪ + ∧ βU − (δ + + (1 − ) + (1 − τ )ρA − (β +
⎪
⎪ 0 d)A 0 A0 A0 0 d)A 0
⎪
⎪ N N
⎪
⎪
⎪
⎪ t2 t2 t3
⎪
⎩ 1 + 3(t − 1)σ + 3 1 − 2t + σ 2 − 1 − 3t + 3 − σ3 ,
2! 2! 3!
  (24)
⎧ η η(1 − τ )
⎪
⎪ A2 (t) = β A0 H0 + A0 V0 − (β + d)U0 − (δ + d) βU0 − (δ + d)A0
⎪
⎪
⎪
⎪
N N
⎪
⎪
⎪
⎪ t2
⎪
⎪ 1 + 2(t − 1)σ + 1 − 2t + σ2
⎪
⎪
⎪
⎪ 2!
⎪
⎪
⎨ βη∧ t2
+ A0 1 + 3(t − 1)σ + 3 1 − 2t + σ2 (25)
⎪
⎪ N 2!
⎪
⎪
⎪
⎪ t2 t3
⎪
⎪ − 1 − 3t + 3 − σ3 ,
⎪
⎪
⎪
⎪ 2! 3!
⎪
⎪
⎪
⎪   .
⎪
⎪ t2 .
⎩ C2 (t) = δ βU0 − (δ + d)A0 − d δA0 − dC0 1 + 2(t − 1)σ + 1 − 2t + σ 2 .,
2!

and so on. In this way, the other terms are computed, hence one has
⎧
⎪
⎪ H(t) = H0 + H1 (t) + H2 (t) + H3 (t) + . . . .
⎪
⎪
⎪
⎨ V(t) = V0 + V1 (t) + V2 (t) + V3 (t) + . . . .
⎪
U(t) = U0 + U1 (t) + U2 (t) + U3 (t) + . . . . (26)
⎪
⎪
⎪
⎪ A(t) = A0 + A1 (t) + A2 (t) + A3 (t) + . . . .
⎪
⎪
⎩
C(t) = C0 + C1 (t) + C2 (t) + C3 (t) + . . . .

6 Results and Discussion

The solutions, up to few terms, of (26) is plotted by using MATLAB. The numerical
values for parameters are used from Table 1. Against various fractional orders, the
solutions are shown as Figs. 2, 3, 4, 5, and 6 respectively.
Qualitative Theory and Approximate Solution … 195

Table 1 Nomenclature and their values.

Nomenclature Stands for Value A Value B
∧ Rate of birth 0.90 5.0
η Rate of effective contact 0.020 0.020
ρ Rate of vaccination coverage 0.030 0.020
d Rate of mortality due to nature 0.020 0.020
δ Rate of recovery 0.050 0.50
β Rate of appearing of clinical symptoms 0.020 0.20
τ Rate of vaccine effectiveness 1.0 1.0
H0 75.0 75.0
V0 20.0 20.0
U0 55.0 55.0
A0 30.0 30.0
C0 20.0 20.0

80
0.90
70 0.92
0.94
0.96
60
0.98
1.00
H

50

40

30

20
0 10 20 30 40 50 60
Time (Days)

Fig. 2 Dynamical behavior of approximate values of H at several fractional order

45

40

35 0.90
0.92
V

0.94
30 0.96
0.98
1.00
25

20
0 10 20 30 40 50 60
Time (Days)

Fig. 3 Dynamical behavior of approximate values of V at several fractional order

196 Eiman et al.

60
0.90
50 0.92
0.94
0.96
40
0.98
1.00
U

30

20

10

0
0 10 20 30 40 50 60
Time (Days)

Fig. 4 Dynamical behavior of approximate values of U at several fractional order

0.90
25 0.92
0.94
20 0.96
0.98
15 1.00

10

0
0 10 20 30 40 50 60
Time (Days)

Fig. 5 Dynamical behavior of approximate values of V at several fractional order

32

30

28
0.90
C

26 0.92
0.94
24 0.96
0.98
1.00
22

20
0 10 20 30 40 50 60
Time (Days)

Fig. 6 Dynamical behavior of approximate values of C at several fractional order

Qualitative Theory and Approximate Solution … 197

7 Setting of Model (2) in Caputo and Atangana-Baleanu

Piecewise Equations Fractional Derivative

Now in this section, we investigate the mathematical model given in (1) by using the
ABC piecewise differential operators. For recent work on using the said operators,
we refer some frequent work like [67, 68]. We formulated the proposed model in the
aforementioned operators form with 0 < σ ≤ 1, t ∈ [0, T ], as

C ABC σ ηHA
0 Dt (H(t)) =∧− − (ρ + d)H,
N
C ABC σ (1 − τ )ηVA
0 Dt (V(t)) = ρH − − dV,
N
C ABC σ ηHA (1 − τ )ηVA
0 Dt (U(t)) = + − (β + d)U,
N N
C ABC σ (27)
0 Dt (A(t)) = βU − (δ + d)A,
C ABC σ
0 Dt (C(t))
= δA − dC,
H(0) = H0 ≥ 0, V(0) = V0 ≥ 0,
U(0) = U0 ≥ 0, A(0) = A0 ≥ 0
C(0) = C0 ≥ 0,

In more explicit form, the model (27) can also be written as

C σ
0 Dt (H(t)) = G 1 (H, V, U, A, C, t), 0 < t ≤ t1 ,
C
C ABC σ
0 Dt (H(t)) = ABC σ
0 Dt (H(t)) = ABC G 1 (H, V, U, A, C, t), t1 < t ≤ T,
C σ
0 Dt (V(t)) = G 2 (H, V, U, A, C, t), 0 < t ≤ t1 ,
C
C ABC σ
0 Dt (V(t)) = ABC σ
0 Dt (V(t)) = ABC G 2 (H, V, U, A, C, t), t1 < t ≤ T,
C σ
0 Dt (U(t)) = G 3 (H, V, U, A, C, t), 0 < t ≤ t1 ,
C
C ABC σ
0 Dt (U(t)) = ABC σ
0 Dt (U(t)) = ABC G 3 (H, V, U, A, C, t), t1 < t ≤ T,
C σ
0 Dt (A(t)) = G 4 (H, V, U, A, C, t), 0 < t ≤ t1 ,
C
C ABC σ
0 Dt (A(t)) = ABC σ
0 Dt (A(t)) = ABC G 4 (H, V, U, A, C, t), t1 < t ≤ T,
C σ
0 Dt (C(t)) = G 5 (H, V, U, A, C, t), 0 < t ≤ t1 ,
C
C ABC σ
0 Dt (C(t)) = ABC σ
(28)
0 Dt (C(t)) = ABC G 5 (H, V, U, A, C, t), t1 < t ≤ T.

where C0 Dtσ and 0ABC Dtσ are Caputo and ABC derivative respectively. We investi-
gate the existence and uniqueness of at least one approximate solution. It should be
noted that fixed point theory plays important role in investigating existence theory
for dynamical problems. Further, Newton’s interpolation formula will be used for
numerical investigation. In the end, all the numerical findings are displayed graphi-
cally by using some real data.
198 Eiman et al.

8 Qualitative Analysis of (2) Using Piecewise Derivative

We can re-write the system (28) as in the Lemma 3.13.1 to develop some results
about analysis as
PC ABC σ
0 Dt U (t) = G(t, U (t)), σ ∈ (0, 1],

is
⎧  t
⎪
⎪ G(ζ, U (ζ ))(t − ζ )σ −1
⎪
⎨ U0 + dζ, 0 < t ≤ t1 ,
0 (σ )
U (t) =  (29)
⎪ (1 − σ )G(t, U (t)) t σ (t − ζ )σ −1
⎪
⎪
⎩ U (t1 ) + + G(ζ U (ζ ))d(ζ ), t1 < t ≤ T,
ABC(σ ) t1 ABC(σ ) σ

where
⎧ ⎧
⎪ ⎨ C G 1 (H, V, U, A, C, t)
⎪
⎪
⎪ G1 =
⎪
⎪
⎪ ⎩ ABC G (H, V, U, A, C, t),
⎪
⎪ 1
⎪
⎪ ⎧
⎪
⎪
⎪
⎪ ⎨ C G (H, V, U, A, C, t)
⎧ ⎪
⎪ 2
⎧ ⎧ Ht1 ⎪
⎪ G2 =
⎪ H(t) ⎪ H0 ⎪
⎪ ⎪
⎪ ⎩ ABC G (H, V, U, A, C, t),
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ 2
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎧
⎪
⎪ V(t) ⎪
⎪ V ⎪ V
⎪ t1 ⎪
⎪
⎨ ⎨ 0 ⎨ ⎨ ⎨ C G 3 (H, V, U, A, C, t)
U (t) = U(t)
⎪
U 0 = U0
⎪
U t1 = Ut1
⎪
G(t, W(t)) = G 3 =
⎪ ⎩ ABC G (H, V, U, A, C, t),
(30)
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ A(t) ⎪
⎪ A0 ⎪
⎪ ⎪
⎪
3
⎪
⎪ ⎪
⎪ ⎪
⎪
At ⎪
⎪ ⎧
⎩ ⎩ ⎪ 1
⎩ ⎪
⎪ ⎨ C G (H, V, U, A, C, t)
C(t), C0 , ⎪
⎪ 4
Ct1 , ⎪
⎪ G 4 = ⎩ ABC
⎪
⎪
⎪
⎪ G 4 (H, V, U, A, C, t),
⎪
⎪
⎪
⎪ ⎧
⎪
⎪ ⎨ C G (H, V, U, A, C, t)
⎪
⎪ 5
⎪
⎩ G 5 = ⎩ ABC
⎪
G 5 (H, V, U, A, C, t).

Here, we described some assumptions as

(C1) There exists constants L U > 0; for every G, U¯ ∈ Z , as

|G(t, U ) − G(t, U¯ )| ≤ L G |U − U¯ |.

(C2) Let one has constants C G > 0 and MG > 0, such that

|G(t, U (t))| ≤ C G |U | + MG

holds.

Theorem 4 If G be a piecewise continuous operator on the sub interval 0 < t ≤ t1 ,

and t1 < t ≤ T on [0, T ] together with hypothesis (C2), then problem (28) has at
least one solution on each sub interval.
Qualitative Theory and Approximate Solution … 199

Proof Let define a closed subset by

B = {U ∈ Z : U  ≤ R1,2 , R1,2 > 0},

and define T : B → B as
⎧  t
⎪ 1 G(ζ, U (ζ ))(t − ζ )σ −1
⎪
⎪ U + dζ, 0 < t ≤ t1
⎨ 0 (σ )
0
T (U ) =  t (31)
⎪
⎪
⎪ (1 − σ )G(t, U (t)) σ (t − ζ )σ −1
⎩ U (t1 ) + + G(ζ U (ζ ))d(ζ ), t1 < t ≤ T.
ABC(σ ) (σ ) t1 ABC(σ )

On any U ∈ B, we get
⎧  t
⎪ 1 (t − ζ )σ −1
⎪
⎨ |U 0 | + 0
⎪
(σ )
|G(ζ, U (ζ ))|dζ,
|T (U )(t)| ≤  t
⎪
⎪
⎪ 1−σ 1 σ (t − ζ )σ −1
⎩ |U (t ) | + |G(t, U (t))| + |G(ζ U (ζ ))|d(ζ ),
1 ABC(σ ) ABC(σ ) t1 (σ )
⎧  t1
⎪ 1 σ −1 [C |U | + M ]dζ,
⎪
⎨ |U 0 | + (σ ) 0 (t − ζ )
⎪ G G
≤  t
⎪
⎪
⎪ 1−σ σ (t − ζ )σ −1
⎩ |U (t1 ) | + [C |U| + MG ] + [C |U | + MG ]d(ζ ),
ABC(σ ) G t1 ABC(σ ) (σ ) G
⎧
⎪ Tσ
⎨ |U 0 | + (σ + 1) [C G |U | + MG ] = R1 , 0 < t ≤ t1 ,
⎪
≤ σ
⎪
⎩ |U (t ) | + 1 − σ [C G |U | + MG ] + σ (T − T)
⎪
[C |U | + MG ]d(ζ ) = R2 , t1 < t ≤ T,
1 ABC(σ ) ABC(σ ) σ + 1 G
R1 , 0 < t ≤ t1 ,
T (U ) ≤
R2 , t1 < t ≤ T.

Fixing R1,2 = max{R1 , R2 }, we see that T (U ) ≤ R1,2 which demonstrates that T

is bounded and also T (U ) ∈ B. Thus T (B) ⊂ B. Further, for completely continuity,
we take ti < t j ∈ [0, t1 ] in the Caputo sense, such that
 t
 j (t j − ζ )σ −1
|T (U )(t j ) − T (U )(ti )| = 
 G(ζ, U (ζ ))dζ,
0 (σ )
 ti
1
− (ti − ζ )σ −1 G(ζ, U (ζ ))dζ
(σ ) 0
 ti
1
≤ [(ti − ζ )σ −1 − (t j − ζ )σ −1 ]|G(ζ, U (ζ )) dζ
(σ ) 0
 tj
1
+ (t j − ζ )σ −1 |G(ζ, U (ζ ))|dζ
(σ ) ti
 t
1 i
≤ [(ti − ζ )σ −1 − (t j − ζ )σ −1 ]dζ
(σ ) 0
 tj 
+ (t j − ζ )σ −1 dζ (C G |U | + MG )
ti
(C G U + MG ) σ
≤ [t j − tiσ + 2(t j − ti )σ ]. (32)
(σ + 1)
200 Eiman et al.

From (32), we see that as ti → t j , then

|T (U )(t j ) − T (U )(ti )| → 0, as i → j.

Since T is bounded on [0, t1 ], so is uniformly continuous also over [0, t1 ]. Hence,

we concluded that T over the said interval is equi-continuous also.
In addition let ti , t j ∈ [t1 , T ], here use 1 (σ ) = ABC(σ
1−σ
)
, and 2 (σ ) = ABC(σσ ) (σ )
 tj
|T (U )(t j ) − T (U )(ti )| = 1 (σ )G(t, U (t)) + 2 (σ ) (t j − ζ )σ −1 G(ζ, U (ζ ))dζ,
t1

(σ ) ti
− 1 (σ )G(t, U (t)) + (ti − ζ )σ −1 G(ζ, U (ζ ))dζ
ABC(σ ) (σ ) t1
 ti
≤ 2 (σ ) [(ti − ζ )σ −1 − (t j − ζ )σ −1 ]|G(ζ, U (ζ ))|dζ
t1
 tj
+ 2 (σ ) (t j − ζ )σ −1 |G(ζ, U (ζ ))|dζ
ti
 ti
≤ 2 (σ ) [(ti − ζ )σ −1 − (t j − ζ )σ −1 ]dζ
t1
 tj 
+ (t j − ζ )σ −1 dζ (C G |U | + MG )
ti
σ (C G U + MG ) σ
≤ [t − tiσ + 2(t j − ti )σ ]. (33)
ABC(σ ) (σ + 1) j

Also from (33), we see that

|T (U )(t j ) − T (U )(ti )| → 0, as ti → t j .

Due to the uniform continuity of T over the said interval, we conclude that T
is equi- continuous in [t1 , T ] interval. Therefore, T is equi- continuous mapping.
Using Arzelá-Ascoli Theorem, operator T is completely continuous and also uniform
continuous and has bounds. Hence, problem (28) has at least one solution on each
subinterval. 
Theorem 5 In view of hypothesis (C1) together with the condition that
  
Tσ σ Tσ
max L G , L G 1 (σ ) + < 1,
(σ + 1) ABC(σ ) (σ + 1)

the proposed model has a unique solution.

Proof We define here operator T : E → E and take U and U¯ ∈ E on [0, t1 ] in
Caputo sense as

 t  t
(t − ζ )σ −1 (t − ζ )σ −1 G(ζ, W̄(ζ ))
T (U ) − T (U¯ ) = max G(ζ, U (ζ ))dζ − dζ
t∈[0,t1 ] 0 (σ ) 0 (σ )
Tσ
≤ L G U − U¯ . (34)
(σ + 1)
Qualitative Theory and Approximate Solution … 201

Equation (34) yields

Tσ L G
T (U ) − T (U¯ ) ≤ U − U¯ . (35)
(σ + 1)

So T is contraction. Therefore in the given sub-interval, in sense of the Banach

contraction theorem, considered problem has a unique solution. Next, in the sense
of ABC derivative, for the other interval t ∈ [t1 , T ], we have

σ (T − T σ )L G
T (U ) − T (U¯ ) ≤ 1 (σ )L G U − U¯  + U − U¯ . (36)
ABC(σ ) (σ + 1)
or
 
σ (T − T)σ
T (U ) − T (U¯ ) ≤ L G 1 (σ ) + U − U¯ . (37)
ABC(σ ) (σ + 1)

Hence, we get the required result. 

9 Approximate Results Using Piecewise Derivative Concept

Here following the procedure [66, 67], we recall the scheme as

⎧  t
⎪ 1 1
⎪
⎨ H0 +
⎪ (t − α)σ −1C G 1 (α)dα, 0 < t ≤ t1 ,
(σ ) 0
H(t)) =  t ,
⎪
⎪
⎪
⎩ H(t1 ) + 1 (σ )
ABC G (t) +  (σ )
1 2 (t − α)σ −1 ABC G 1 (α)dα, t1 < t ≤ T,
t1
⎧  t
⎪ 1 1
⎪
⎨ V0 +
⎪ (t − α)σ −1C G 2 (α)dα, 0 < t ≤ t1 ,
(σ ) 0
V(t)) =  t ,
⎪
⎪
⎪
⎩ V(t1 ) + 1 (σ )
ABC G (t) +  (σ )
2 2 (t − α)σ −1 ABC G 2 (α)dα, t1 < t ≤ T,
t1
⎧  t
⎪ 1 1
⎪
⎪ U0 +
⎨ (t − α)σ −1C G 3 (α)dα, 0 < t ≤ t1 ,
(σ ) 0
U(t)) =  t ,
⎪
⎪ σ −1 ABC G (α)dα, t < t ≤ T,
⎪
⎩ U(t1 ) + 1 (σ ) ABC G (t) +  (σ )
3 2 (t − α) 3 1
t1
⎧  t
⎪ 1 1
⎪
⎪ A0 +
⎨ (t − α)σ −1C G 4 (α)dα, 0 < t ≤ t1 ,
(σ ) 0
A(t)) =  t ,
⎪
⎪ σ −1 ABC G (α)dα, t < t ≤ T,
⎪
⎩ A(t1 ) + 1 (σ ) ABC G (t) +  (σ )
4 2 (t − α) 4 1
t1
⎧  t
⎪ 1 1
⎪
⎪ C0 +
⎨ (t − α)σ −1C G 5 (α)dα, 0 < t ≤ t1 ,
(σ ) 0
C(t)) =  t , (38)
⎪
⎪ σ −1 ABC G (α)dα, t < t ≤ T,
⎪
⎩ C(t1 ) + 1 (σ )
ABC G (t) +  (σ )
5 2 (t − α) 5 1
t1
202 Eiman et al.

where C G i (t) =C G i (H, V, U, A, C, t) and ABC G i (t) =iABC G(H, V, U, A, C, t)

are the left hand side of equation (27) for i = 1, 2, 3, 4, also given in equation (28).
We derive the scheme for the first equation of system (38) and the same procedure
can be repeated for other compartments as well.
At t = tn+1 , one has
⎧  t1
⎪ 1
⎪
⎪ H0 + (t − α)σ −1C G 1 (H, V, U, A, C, α)dα,
⎪
⎪ (σ ) 0
⎨
H(tn+1 )) = H(t1 ) + 1 (σ ) ABC G 1 (H, V, U, A, C, tn ) (39)
⎪
⎪  tn+1
⎪
⎪
⎪
⎩ + 2 (σ ) (t − α)σ −1 ABC G 1 (α)dα, t1 < t ≤ T.
t1

Following [37], (39) implies

⎧ ⎧ i  
⎪
⎪
⎪ ⎪
⎪
C ( t)σ −1 
⎪
⎪ ⎪
⎪ G 1 (Hk−2 , Vk−2 , Uk−2 , Ak−2 , Ck−2 , t)
⎪
⎪ ⎪
⎪ (σ + 1)
⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪
k=2
⎪
⎪ ⎪
⎪ 
⎪
⎪ ⎪
⎪ i C
⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ + G 1 (Hk−1 , Vk−1 , Uk−1 , Ak−1 , Ck−1 , t)
⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ k=2
⎪
⎪ ⎪
⎪ 
⎪
⎪ ⎪
⎪ σ −1 ∇
⎪
⎪ ⎨ C k−2 , Vk−2 , Uk−2 , Ak−2 , Ck−2 , t) ( t)
⎪
⎪
⎪ H0 + − G 1 (H (σ + 2)
⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ C
⎪
⎪ ⎪
⎪ i
⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ + G 1 (Hk , Vk , Uk , Ak , Ck , tk )
⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ k=2
⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ − 2C G 1 (Hk−1 , Vk−1 , Uk−1 , Ak−1 , Ck−1 , t)
⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ 
⎪
⎪ ⎪
⎪ σ ( t)σ −1
⎪
⎪ ⎩ +C G 1 (Hk−2 , Vk−2 , Uk−2 , Ak−2 , Ck−2 , t)
⎪
⎪ (σ + 3)
⎪
⎨ 2
⎧ ⎫
H(tn+1 ) =
⎪ ⎪
⎪ 1 (σ ) ABC G ((Hn , Vn , Un , An , Cn , t )
1 n ⎪
⎪
(40)
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪ ⎪
⎪   ⎪
⎪ (δt)σ −1 ⎪
⎪ ⎪ n ABC ⎪
⎪
⎪ ⎪
⎪ σ ⎪
⎪
⎪ ⎪+
⎪ G 1 ((Hk−2 , Vk−2 , Uk−2 , Ak−2 , Ck−2 , tk−2 ) ⎪
⎪
⎪
⎪ ⎪
⎪ ABC(σ ) (σ + 1) ⎪ ⎪
⎪
⎪
⎪ ⎪ k=i+3
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ n  ABC ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ + G (H k−1 , V k−1 , U k−1 , A k−1 , C k−1 , t) ⎪
⎪
⎪
⎪ ⎪
⎪ 1 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪
k=i+3 ⎪
⎪
⎪
⎪ ⎨  ⎬
⎪
⎪ ∇σ (δt) σ −1
⎪
⎪ H(t1 ) + + ABC G ((Hk−2 , Vk−2 , Uk−2 , Ak−2 , Ck−2 , t )
⎪
⎪ ⎪
⎪ 1 k−2
ABC(σ ) (σ + 2) ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ n  ABC ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪+
⎪ G ((H k , V k , U k , Ak , C k , t ) ⎪
⎪
⎪
⎪ ⎪
⎪
1 k ⎪
⎪
⎪
⎪ ⎪ k=i+3
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ − 2 ABC G (Hk−1 , Vk−1 , Uk−1 , Ak−1 , Ck−1 , t) ⎪
⎪
⎪
⎪ ⎪
⎪ 1 ⎪
⎪
⎪
⎪ ⎪
⎪  ⎪
⎪
⎪
⎪ ⎪
⎪ σ σ (δt) σ −1 ⎪
⎪
⎪
⎩ ⎪
⎩+ ABC G 1 ((H k−2 ,V k−2 ,U k−2 ,A k−2 ,C k−2 , tk−2 ) . ⎪
⎭
ABC(σ ) (σ + 3)

Other compartments can also be expressed as

Qualitative Theory and Approximate Solution … 203
⎧ ⎧ i C  ⎫
⎪
⎪
⎪ ⎪
⎪ ( t)σ −1 ⎪ ⎪
⎪
⎪ ⎪
⎪ G (H k−2
, V k−2
, U k−2
, A k−2
, C k−2
, t) ⎪
⎪
⎪
⎪ ⎪
⎪
1
(σ + 1) ⎪⎪
⎪
⎪ ⎪
⎪ k=2 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ i C ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ + (H k−1
, k−1
, k−1
, k−1
, k−1
, ⎪
⎪
⎪
⎪ ⎪
⎪
G 1 V U A C t) ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ k=2 ⎪
⎪
⎪
⎪ ⎪
⎪  ⎪
⎪
⎪
⎪ ⎪
⎨ ∇( t) σ −1 ⎪
⎬
⎪
⎪ − G 1 (H , V , U , A , C , t)
C k−2 k−2 k−2 k−2 k−2
⎪
⎪ V0 + (σ + 2) .
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪ ⎪
⎪ ⎪
⎪ i C ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ + G 1 (H k , V k , U k , Ak , C k , tk ) ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪
k=2
⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ − 2 C
G (H k−1
, V k−1
, U k−1
, A k−1
, C k−1
, t) ⎪
⎪
⎪
⎪ ⎪
⎪
1
⎪
⎪
⎪
⎪ ⎪
⎪  ⎪
⎪
⎪
⎪ ⎪
⎪ σ ( t)σ −1 ⎪
⎪
⎪
⎪ ⎩ + C
G (H k−2
, V k−2
, U k−2
, A k−2
, C k−2
, t) ⎭
⎪
⎪ 1
(σ +
⎪
⎨ 2 3)
⎧ ⎫
V(tn+1 )) =
⎪ ⎪ 1 (σ )
⎪
ABC
G 1 ((Hn , Vn , Un , An , Cn , tn ) ⎪
⎪
(41)
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ n   σ −1 ⎪
⎪
⎪
⎪ ⎪
⎪
ABC
σ (δt) ⎪
⎪
⎪
⎪ ⎪
⎪ + G 1 ((H , V , U , A , C , tk−2 )
k−2 k−2 k−2 k−2 k−2 ⎪
⎪
⎪
⎪ ⎪
⎪ ABC(σ ) (σ + 1) ⎪ ⎪
⎪
⎪ ⎪
⎪ k=i+3 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪  ⎪
⎪
⎪
⎪ ⎪
⎪
n ABC ⎪
⎪
⎪
⎪ ⎪
⎪ + G 1 (H , V , U , A , C , t)
k−1 k−1 k−1 k−1 k−1 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪
k=i+3 ⎪
⎪
⎪
⎪ ⎨  ⎬
⎪
⎪ V(t1 ) + ∇σ (δt)σ −1
⎪
⎪ + ABC G ((H k−2
, V k−2
, U k−2
, A k−2
, C k−2
, t ) .
⎪
⎪ ⎪
⎪ 1 k−2
ABC(σ ) (σ + 2) ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪ ⎪
⎪ n  ABC ⎪
⎪
⎪
⎪ ⎪ ⎪
⎪
⎪ ⎪
⎪ + G 1 ((Hk , Vk , Uk , Ak , Ck , tk ) ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪
k=i+3 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ −2 ABC
G 1 (H , V , U , A , C , t)
k−1 k−1 k−1 k−1 k−1 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪  ⎪
⎪
⎪
⎪ ⎪
⎪ ABC σ σ (δt)σ −1 ⎪
⎪
⎪
⎩ ⎪
⎩+ G 1 ((H , V , U , A , C , tk−2 )
k−2 k−2 k−2 k−2 k−2
. ⎪
⎭
ABC(σ ) (σ + 3)

⎧ ⎧ i C  ⎫
⎪
⎪
⎪ ⎪
⎪ ( t)σ −1 ⎪
⎪
⎪
⎪ ⎪
⎪ G (H k−2
, V k−2
, U k−2
, A k−2
, C k−2
, t) ⎪
⎪
⎪
⎪ ⎪
⎪
1
(σ + 1) ⎪
⎪
⎪
⎪ ⎪
⎪ k=2 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ i C ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ + (H k−1
, k−1
, k−1
, k−1
, k−1
, ⎪
⎪
⎪
⎪ ⎪
⎪
G 1 V U A C t) ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ k=2 ⎪
⎪
⎪
⎪ ⎪
⎪  ⎪
⎪
⎪
⎪ ⎪
⎨ ∇( t) σ −1 ⎪
⎬
⎪
⎪ − C
G (H k−2
, V k−2
, U k−2
, A k−2
, C k−2
, t)
⎪
⎪ U0 + 1
(σ + 2) .
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪ ⎪
⎪ ⎪
⎪ i C ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ + G 1 ((Hk , Vk , Uk , Ak , Ck , tk ) ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪
k=2
⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ − 2 C
G (H k−1
, V k−1
, U k−1
, A k−1
, C k−1
, t) ⎪
⎪
⎪
⎪ ⎪
⎪
1
⎪
⎪
⎪
⎪ ⎪
⎪  σ −1 ⎪
⎪
⎪
⎪ ⎪
⎪ σ ( t) ⎪
⎪
⎪
⎪ ⎩ + C
G ((H k−2
, V k−2
, U k−2
, A k−2
, C k−2
, t ) ⎭
⎪
⎪ 1 k−2
(σ +
⎪
⎨ 2 3)
⎧ ⎫
U(tn+1 )) =
⎪ ⎪ 1 (σ )
⎪
ABC
H1 ((Hn , Vn , Un , An , Cn , tn ) ⎪
⎪
(42)
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ n   σ −1 ⎪
⎪
⎪
⎪ ⎪
⎪
ABC
σ (δt) ⎪
⎪
⎪
⎪ ⎪
⎪ + G 1 (H , V , U , A , C , tk−2 )
k−2 k−2 k−2 k−2 k−2 ⎪
⎪
⎪
⎪ ⎪
⎪ ABC(σ ) (σ + 1) ⎪ ⎪
⎪
⎪ ⎪
⎪ k=i+3 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪  ⎪
⎪
⎪
⎪ ⎪
⎪
n ABC ⎪
⎪
⎪
⎪ ⎪
⎪ + G 1 (H , V , U , A , C , t)
k−1 k−1 k−1 k−1 k−1 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪
k=i+3 ⎪
⎪
⎪
⎪ ⎨  ⎬
⎪
⎪ ∇σ (δt)σ −1
⎪
⎪ U(t 1 ) + + ABC G 1 (H , V , U , A , C , tk−2 )
k−2 k−2 k−2 k−2 k−2 .
⎪
⎪ ⎪
⎪ ABC(σ ) (σ + 2) ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪ ⎪
⎪ n  ABC ⎪
⎪
⎪
⎪ ⎪ ⎪
⎪
⎪ ⎪
⎪+ G 1 ((Hk , Vk , Uk , Ak , Ck , tk ) ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪
k=i+3 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ −2 ABC
G 1 (H , V , U , A , C , t)
k−1 k−1 k−1 k−1 k−1 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪  ⎪
⎪
⎪
⎪ ⎪
⎪ σ σ (δt) σ −1 ⎪
⎪
⎪
⎩ ⎪
⎩+ ABC
G 1 (H , V , U , A , C , tk−2 )
k−2 k−2 k−2 k−2 k−2
. ⎪
⎭
ABC(σ ) (σ + 3)
204 Eiman et al.
⎧ ⎧ i C  ⎫
⎪
⎪
⎪ ⎪
⎪ ( t)σ −1 ⎪ ⎪
⎪
⎪ ⎪
⎪ G (H k−2
, V k−2
, U k−2
, A k−2
, C k−2
, t) ⎪
⎪
⎪
⎪ ⎪
⎪
1
(σ + 1) ⎪ ⎪
⎪
⎪ ⎪
⎪ k=2 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ i C ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ + (H k−1
, k−1
, k−1
, k−1
, k−1
, ⎪
⎪
⎪
⎪ ⎪
⎪
G 1 V U A C t) ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ k=2 ⎪
⎪
⎪
⎪ ⎪
⎪  ⎪
⎪
⎪
⎪ ⎪
⎨ ∇( t) σ −1 ⎪
⎬
⎪
⎪ − G 1 (H , V , U , A , C , t)
C k−2 k−2 k−2 k−2 k−2
⎪
⎪ A0 + (σ + 2) .
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪ ⎪
⎪ ⎪
⎪ i C ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ + G 1 ((Hk , Vk , Uk , Ak , Ck , tk ) ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪
k=2
⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ − 2 C
G (H k−1
, V k−1
, U k−1
, A k−1
, C k−1
, t) ⎪
⎪
⎪
⎪ ⎪
⎪
1
⎪
⎪
⎪
⎪ ⎪
⎪  ⎪
⎪
⎪
⎪ ⎪
⎪ σ ( t) σ −1 ⎪
⎪
⎪
⎪ ⎩ + C
G (H k−2
, V k−2
, U k−2
, A k−2
, C k−2
, t) ⎭
⎪
⎪ 1
(σ +
⎪
⎨ 2 3)
⎧ ⎫
A(tn+1 )) =
⎪ ⎪ 1 (σ )
⎪
ABC
G 1 (Hn , Vn , Un , An , Cn , tn ) ⎪
⎪
(43)
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ n   σ −1 ⎪
⎪
⎪
⎪ ⎪
⎪
ABC
σ (δt) ⎪
⎪
⎪
⎪ ⎪
⎪ + G 1 (H , V , U , A , C , tk−2 )
k−2 k−2 k−2 k−2 k−2 ⎪
⎪
⎪
⎪ ⎪
⎪ ABC(σ ) (σ + 1) ⎪ ⎪
⎪
⎪ ⎪
⎪ k=i+3 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪  ⎪
⎪
⎪
⎪ ⎪
⎪
n ABC ⎪
⎪
⎪
⎪ ⎪
⎪ + G 1 (H , V , U , A , C , t)
k−1 k−1 k−1 k−1 k−1 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪
k=i+3 ⎪
⎪
⎪
⎪ ⎨  ⎬
⎪
⎪ A(t1 ) + ∇σ (δt)σ −1
⎪
⎪ + ABC G (H k−2
, V k−2
, U k−2
, A k−2
, C k−2
, t ) .
⎪
⎪ ⎪
⎪ 1 k−2
ABC(σ ) (σ + 2) ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪ ⎪
⎪ n  ABC ⎪
⎪
⎪
⎪ ⎪ ⎪
⎪
⎪ ⎪
⎪ + G 1 ((Hk , Vk , Uk , Ak , Ck , tk ) ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪
k=i+3 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ −2 ABC
G 1 (H , V , U , A , C , t)
k−1 k−1 k−1 k−1 k−1 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪  ⎪
⎪
⎪
⎪ ⎪
⎪ ABC σ σ (δt)σ −1 ⎪
⎪
⎪
⎩ ⎪
⎩+ G 1 ((H , V , U , A , C , tk−2 )
k−2 k−2 k−2 k−2 k−2
. ⎪
⎭
ABC(σ ) (σ + 3)

⎧ ⎧ i C  ⎫
⎪
⎪
⎪ ⎪
⎪ ( t)σ −1 ⎪ ⎪
⎪
⎪ ⎪
⎪ G (H k−2
, V k−2
, U k−2
, A k−2
, C k−2
, t) ⎪
⎪
⎪
⎪ ⎪
⎪
1
(σ + 1) ⎪ ⎪
⎪
⎪ ⎪
⎪ k=2 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ i C ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ + (H k−1
, k−1
, k−1
, k−1
, k−1
, ⎪
⎪
⎪
⎪ ⎪
⎪
G 1 V U A C t) ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ k=2 ⎪
⎪
⎪
⎪ ⎪
⎪  ⎪
⎪
⎪
⎪ ⎪
⎨ ∇( t) σ −1 ⎪
⎬
⎪
⎪ − C
G (H k−2
, V k−2
, U k−2
, A k−2
, C k−2
, t)
⎪
⎪ C0 + 1
(σ + 2) .
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪ ⎪
⎪ ⎪
⎪ i C ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ + G 1 ((Hk , Vk , Uk , Ak , Ck , tk ) ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪
k=2
⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ − 2 C
G (H k−1
, V k−1
, U k−1
, A k−1
, C k−1
, t) ⎪
⎪
⎪
⎪ ⎪
⎪
1
⎪
⎪
⎪
⎪ ⎪
⎪  σ −1 ⎪⎪
⎪
⎪ ⎪
⎪ σ ( t) ⎪
⎪
⎪
⎪ ⎩ + C
G (H k−2
, V k−2
, U k−2
, A k−2
, C k−2
, t) ⎭
⎪
⎪ 1
(σ +
⎪
⎨ 2 3)
⎧ ⎫
C(tn+1 )) =
⎪ ⎪ 1 (σ )
⎪
ABC
G 1 (Hn , Vn , Un , An , Cn , tn ) ⎪
⎪
(44)
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ n   σ −1 ⎪
⎪
⎪
⎪ ⎪
⎪
ABC
σ (δt) ⎪
⎪
⎪
⎪ ⎪
⎪ + G 1 (H , V , U , A , C , tk−2 )
k−2 k−2 k−2 k−2 k−2 ⎪
⎪
⎪
⎪ ⎪
⎪ ABC(σ ) (σ + 1) ⎪ ⎪
⎪
⎪ ⎪
⎪ k=i+3 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪  ⎪
⎪
⎪
⎪ ⎪
⎪
n ABC ⎪
⎪
⎪
⎪ ⎪
⎪ + G 1 (H , V , U , A , C , t)
k−1 k−1 k−1 k−1 k−1 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪
k=i+3 ⎪
⎪
⎪
⎪ ⎨  ⎬
⎪
⎪ ∇σ (δt)σ −1
⎪
⎪ C(t1 ) + + ABC G 1 (H , V , U , A , C , tk−2 )
k−2 k−2 k−2 k−2 k−2 ,
⎪
⎪ ⎪
⎪ ABC(σ ) (σ + 2) ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪ ⎪
⎪ n  ABC ⎪
⎪
⎪
⎪ ⎪ ⎪
⎪
⎪ ⎪
⎪+ G 1 ((Hk , Vk , Uk , Ak , Ck , tk ) ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪
k=i+3 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ −2 ABC
G 1 (H , V , U , A , C , t)
k−1 k−1 k−1 k−1 k−1 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪  ⎪
⎪
⎪
⎪ ⎪
⎪ σ σ (δt)σ −1 ⎪
⎪
⎪
⎩ ⎪
⎩+ ABC
G 1 ((H , V , U , A , C , tk−2 )
k−2 k−2 k−2 k−2 k−2
, ⎪
⎭
ABC(σ ) (σ + 3)
Qualitative Theory and Approximate Solution … 205

where
⎡ ⎤
σ σ σ 2 σ σ
⎢ 2(n − k) (1 + n − k) + (1 + n − k) (3σ + 10)(n − k) + 2(1 + n − k) σ + 9(1 + n − k) σ + 12(1 + n − k)
2
⎥
⎢ ⎥
=⎢ ⎥,
⎣ ⎦
− 2(n − k)(n − k)2 + (n − k)(5σ + 10)(−k + n) + 6(n − k)σ 2 + 18(n − k)σ + 12(n − k)

⎡ ⎤
3 + 2σ − k + n (1 + n − k)σ
⎢ ⎥
∇=⎢
⎣
⎥,
⎦
− n − k + 3σ + 3 (n − k)

 
= (1 + n − k)σ − (n − k)σ .

10 Numerical Simulation and Discussion

In this section, we present our proposed numerical scheme to simulate the considered
model numerically and graphically under the piecewise equation using ABC type
derivative. We use numerical values given in Table 1 and simulate the model in Figs.
7, 8, 9, 10 and 11 respectively. Further, the domain [0, 60] is divided into two sub
intervals [0, 20], [20, 60] respectively.

Fig. 7 Approximate results 80

presentation using piecewise 0.90
concept of H at several 0.94
Suscceptible class

0.96
fractional order in the model 60
1.00
(27)

40

20
0 10 20 30 40 50 60
Time (Days)

Fig. 8 Approximate results 45

presentation using piecewise
concept of V at several 40
Vacinnated class

fractional order in the model

35
(27) 0.90
0.94
30 0.96
1.00
25

20
0 10 20 30 40 50 60
Time (Days)
206 Eiman et al.

Fig. 9 Approximate results 60

presentation using piecewise 0.90
50 0.94
concept of U at several
0.96

Exposed class
fractional order in the model 40 1.00
(27)
30

20

10

0
0 10 20 30 40 50 60
Time (Days)

Fig. 10 Approximate results 30

presentation using piecewise 0.90
25 0.94
concept of A at several
0.96
Infected class

fractional order in the model 1.00

(27) 20

15

10

5
0 10 20 30 40 50 60
Time (Days)

Fig. 11 Approximate results 500

presentation using piecewise
concept of C at several 400
Recovered class

fractional order in the model 0.90

300 0.94
(27) 0.96
1.00
200

100

0
0 10 20 30 40 50 60
Time (Days)

11 Suberization of This Work

We have investigated a five-compartmental fractional order mathematical model of

Norovirus. The five compartments are of susceptible, vaccinated, exposed, infected,
and recovered individuals. Also by Laplace transform coupled with ADM some
approximate analytical results have been determined. From the graphs, it has been
concluded that the solution tending to the classical (integer) order solution when
σ → 1. It has concluded that by taking a few terms of the series solutions, the system
under investigation can be demonstrated properly. We also check the investigated
proposed model under piecewise equations by using ABC fractional order derivative.
Theoretical results by using fixed point theory and numerical analysis via the Adam
Qualitative Theory and Approximate Solution … 207

Bash-forth method have been used to obtain the required results. Several graphical
results have been presented to understand the crossover behavior of the proposed
model. The obtained analysis can be extended in the future for a more general model
of epidemiology.

Acknowledgements “Authors are thankful to Prince Sultan University for support through TAS
research lab.”

Conflict of Interest “The authors declare that they have no competing interests.”

References

1. J.L. Adler, R. Zickl, Winter vomiting disease. J. Infect. Dis. 119(6), 668–673 (1969)
2. R.L. Atmar, A.R. Opekun, M.A. Gilger, M.K. Estes, S.E. Crawford, F.H. Neill, S. Ramani, H.
Hill, J. Ferreira, D.Y. Graham, Determination of the 50% human infectious dose for Norwalk
virus. J. Infect. Dis. 209(7), 1016–1022 (2014)
3. U. Desselberger, I. Goodfellow, Noroviruses, a global cause of acute gastroenteritis. Lancet.
Infect. Dis 14(8), 664–665 (2014)
4. S.M. Ahmed, A.J. Hall, A.E. Robinson, L. Verhoef, P. Premkumar, U.D. Parashar, M. Koop-
mans, B.A. Lopman, Global prevalence of norovirus in cases of gastroenteritis: a systematic
review and meta-analysis. Lancet Infect. Dis. 14(8), 725–730 (2014)
5. B. Lopman, P. Gastanaduy, G.W. Park, A.J. Hall, U.D. Parashar, J. Vinjé, Environmental trans-
mission of norovirus gastroenteritis. Curr. Opin. Virol. 2(1), 96–102 (2012)
6. J.G. Castilho, V. Munford, H.R. Resque, U. Fagundes-Neto, J. Vinjé, M.L. Rácz, Genetic
diversity of norovirus among children with gastroenteritis in Sao Paulo State, Brazil. J. Clin.
Microb. 44(11), 3947–3953 (2006)
7. J. Yi, K. Wahl, B.K. Sederdahl, R.R. Jerris, C.S. Kraft, C. McCracken, S. Gillespie, E.J.
Anderson, A.E. Kirby, A.L. Shane, C.L. Moe, Molecular epidemiology of Norovirus in children
and the elderly in Atlanta, Georgia, United States. J. Med. Vir. 88(6), 961–970 (2016)
8. T. Murata, N. Katsushima, K. Mizuta, Y. Muraki, S. Hongo, Y. Matsuzaki, Prolonged Norovirus
shedding in infants 6 months of age with gastroenteritis. J. Pediatr. Infect. Dis. 26(1), 46–49
(2007)
9. R.L. Atmar, A.R. Opekun, M.A. Gilger, M.K. Estes, S.E. Crawford, F.H. Neill, D.Y. Graham,
Norwalk virus shedding after experimental human infection. Emerg. Infect. Dis. 14(10), 1553
(2008)
10. M.D. Graaf, J. Van Beek, M.P. Koopmans, Human norovirus transmission and evolution in a
changing world. Nat. Rev. Microb. 14(7), 421–433 (2016)
11. C.C. Tam, S.J. Brien, Economic cost of Campylobacter, norovirus and rotavirus disease in the
United Kingdom. PlOS ONE 11(2), e0138526 (2016)
12. S.M. Ahmed, B.A. Lopman, K. Levy, A systematic review and meta-analysis of the global
seasonality of norovirus. PlOS ONE 8(10), e75922 (2013)
13. J. Rohayem, Norovirus seasonality and the potential impact of climate change. Clin. Microbio.
Infe. 15(6), 524–527 (2009)
14. N. Carmona-Vicente, M. Fernández-Jiménez, J.M., Ribes, C.J. Téllez-Castillo, P. Khodayar-
Pardo, J. Rodríguez-Diaz, J. Buesa, Norovirus infections and seroprevalence of genotype gii.
4-specific antibodies in a Spanish population. J. Med. Viro. 87(4), 675–682 (2015)
15. K. Ozawa, T. Oka, N. Takeda, G.S. Hansman, Norovirus infections in symptomatic and asymp-
tomatic food handlers in Japan. J. Clin. Microbiol. 45(12), 3996–4005 (2007)
16. K. Simmons, M. Gambhir, J. Leon, B. Lopman, Duration of immunity to norovirus gastroen-
teritis. Emerg. Infect. Dis. 19(8), 1260 (2013)
208 Eiman et al.

17. M.O. Milbrath et al., Heterogeneity in norovirus shedding duration affects community risk.
Emerg. Infect. Dis. 141(8), 1572–1584 (2013)
18. E.T. Isakbaeva, M.A. Widdowson, R.S. Beard, S.N. Bulens, J. Mullins, S.S. Monroe, J. Bresee,
P. Sassano, E.H. Cramer, R.I. Glass, Norovirus transmission on cruise ship. Emerg. Infect. Dis.
11(1), 154 (2005)
19. C.N. Thornley, N.A. Emslie, T.W. Sprott, G.E. Greening, J.P. Rapana, Recurring norovirus
transmission on an airplane. Clin. Infect. Dis. 53(6), 515–520 (2011)
20. P.F. Teunis, C.L. Moe, P. Liu, S.E. Miller, L. Lindesmith, R.S. Baric, J. Le Pendu, R.L. Calderon,
Norwalk virus: how infectious is it? J. Med. Viro. 80(8), 1468–1476 (2008)
21. S. Honma, S. Nakata, K. Numata, K. Kogawa, T. Yamashita, M. Oseto, X. Jiang, S. Chiba,
Epidemiological study of prevalence of genogroup II human calicivirus (Mexico virus) infec-
tions in Japan and Southeast Asia as determined by enzyme-linked immunosorbent assays. J.
Clin. Microbiol. 36(9), 2481–2484 (1998)
22. A.J. Hall, B.A. Lopman, D.C. Payne, M.M. Patel, P.A. Gastanaduy, J. Vinjé, U.D. Parashar,
Norovirus disease in the United States. Emerg. Infect. Dis. 19(8), 1198–1205 (2013)
23. C.C. Lai, Y.H. Wang, C.Y. Wu, C.H. Hung, D.D.S. Jiang, F.T. Wu, A Norovirus outbreak in
a nursing home, Norovirus shedding time associated with age. J. Clin. Virol. 56(2), 96–101
(2013)
24. J. Le Pendu, E. Kindberg, L. Svensson, Mendelian resistance to human Norovirus infections.
Seminars Immunol. 18(6), 375–386 (2006). Academic Press
25. H. Singh, H. Srivastava, D. Baleanu, Methods of Mathematical Modeling: Infectious Diseases
Presents Computational Methods Related to Biological Systems and Their Numerical Treatment
via Mathematical Tools and Techniques (Elsevier Science, 2021). (ISBN: 9780323998888)
26. L. Verhoef, E. Depoortere, I. Boxman, E. Duizer, Y. Van Duynhoven, J. Harris, C. Johnsen,
A. Kroneman, S. Le Guyader, W. Lim, L. Maunula, Food Borne Viruses in Europe Network,
Emergence of new norovirus variants on spring cruise ships and prediction of winter epidemics.
Emerg. Infect. Dis. 14(2), 238 (2008)
27. E. Pelosi, P.R. Lambden et al., The seroepidemiology of genogroup 1 and genogroup 2 Norwalk-
like viruses in Italy. J. Med. Virol. 58(1), 93–99 (1999)
28. T. Li, Y. Wang, W. Pan, Parameter estimation for the one–term (Multiterm) fractional–order
SEIAR models of norovirus outbreak. Adv. Math. Phy. 2021, 14 (2021)
29. J.A. Marshall, L.D. Bruggink, The dynamics of norovirus outbreak epidemics: recent insights.
Int. J. Envir. Res. Publ. Health 8(4), 1141–1149 (2011)
30. Y. Nawaz, M.S. Arif, W. Shatanawi, A new numerical scheme for time fractional diffusive
SEAIR model with non-linear incidence rate: an application to computational biology. Fractal
Fract. 6(2), 78 (2022)
31. H. Singh, Analysis for fractional dynamics of Ebola virus model. Chaos, Solitons Fractals 138,
109992 (2020)
32. H. Singh, Analysis of drug treatment of the fractional HIV infection model of CD4+ T-cells.
Chaos, Solitons Fractals 146, 110868 (2021)
33. H. Singh, D. Baleanu, J. Singh, H. Dutta, Computational study of fractional order smoking
model. Chaos, Solitons Fractals 142, 110440 (2021)
34. S. Ahmad, A. Ullah, M. Partohaghighi, S. Saifullah, A. Akgül, F. Jarad, Oscillatory and complex
behaviour of Caputo-Fabrizio fractional order HIV-1 infection model. AIMS Math. 7(3), 4778–
4792 (2022)
35. H. Khan, J.F. Gómez-Aguilar, A. Alkhazzan, A. Khan, A fractional order HIV-TB coinfec-
tion model with nonsingular Mittag-Leffler Law. Math. Methods Appl. Sci. 43(6), 3786–3806
(2020)
36. S.K. Kabunga, E.F.D. Goufo, V.H. Tuong, Analysis and simulation of a mathematical model
of tuberculosis transmission in democratic Republic of the Congo. Adv. Differ. Equ. (1), 1–19
(2020)
37. S. Saifullah, A. Ali, M. Irfan, K. Shah, Time-fractional Klein-Gordon equation with soli-
tary/shock waves solutions. Math. Prob. Eng. 2021, 14, Article ID: 6858592 (2021)
Qualitative Theory and Approximate Solution … 209

38. H. Khan, J. Alzabut, A. Shah, S. Etemad, S. Rezapour, C. Park, A study on the fractal-fractional
tobacco smoking model. Aims Math. 7, 13887–13909 (2022)
39. S. Ahmad, A. Ullah, A Akgddotul. Investigating the complex behaviour of multi-scroll chaotic
system with Caputo fractal-fractional operator. Chaos Solitons Fractals 146, 110900 (2021)
40. K. Shah, M. Arfan, A. Ullah, Q. Al-Mdallal, K.J. Ansari, T. Abdeljawad, Computational study
on the dynamics of fractional order differential equations with applications. Chaos, Solitons
Fractals 157, 111955 (2022)
41. M.K. Bansal, K.S. Nisar, J. Choi, D. Kumar, Generalized fractional operators and their image
formulas. J. Comput. Anal. Appl. 30(2), 222–235 (2022)
42. M. Kumar Bansal, D. Kumar, P. Harjule, J. Singh, Fractional kinetic equations associated with
incomplete I-functions. Fractal Fract. 4(2), 19 (2020)
43. R.P. Agarwal, V. Lakshmikantham, J.J. Nieto, On the concept of solution for fractional differ-
ential equations with uncertainty. Nonl. Anal: Theo., Method. Appl. 72(6), 2859–2862 (2010)
44. T. Cui, P. Liu, A. Din, Fractal fractional and stochastic analysis of norovirus transmission
epidemic model with vaccination effects. Sci. Rep. 11(1), 1–25 (2021)
45. S. Yang, A. Xiao, H. Su, Convergence of the variational iteration method for solving multi-order
fractional differential equations. Comput. Math. Appl. 60(10), 2871–2879 (2010)
46. K.S. Miller, B. Ross, An Introduction to the Fractional Calculus and Fractional Differential
Equations (Wiley, New York, 1993)
47. Z. Odibat, D. Baleanu, Numerical simulation of initial value problems with generalized Caputo-
type fractional derivatives. Appl. Numer. Math. 156, 94–105 (2020)
48. P. Veeresha et al., Regarding new numerical solution of fractional Schistosomiasis disease
arising in biological phenomena. Chaos, Solitons Fractals 133, 109661 (2020)
49. A. Yokus, S. Gulbahar, Numerical solutions with linearization techniques of the fractional
Harry Dym equation. Appl. Math. Nonl. Sci. 4(1), 35–42 (2019)
50. A. Naghipour, J. Manafian, Application of the Laplace Adomian decomposition method and
implicit methods for solving Burgers equations. TWMS J. Pure Appl. Math. 6(1), 68–77 (2015)
51. J. Biazar, Solution of the epidemic model by Adomian decomposition method. Appl. Math.
Comput. 173(2), 1101–1106 (2006)
52. F. Haq, K. Shah, G. Rahman, M. Shahzad, Numerical solution of fractional order smoking
model via Laplace Adomian decomposition method. Alex. Eng. J. 57(2), 1061–1069 (2018)
53. A. Ali, K. Shah, R.A. Khan, Numerical treatment for traveling wave solutions of fractional
Whitham-Broer-Kaup equations. Alex. Eng. J. 57(3), 1991–1998 (2018)
54. O. Kiymaz, An Algorithm for solving initial value problems using Laplace Adomian Decom-
position Method. Appl. Math. Sci. 3(30), 1453–1459 (2009)
55. M. Caputo, M. Fabrizio, A new definition of fractional derivative without singular kernel. Progr.
Fract. Differ. Appl. 1(2), 1–13 (2015)
56. K. Shah, M.A. Alqudah, F. Jarad, T. Abdeljawad, Semi-analytical study of Pine Wilt Disease
model with convex rate under Caputo-Fabrizio fractional order derivative. Chaos, Solitons
Fractals 135, 109754 (2020)
57. G. Nazir, K. Shah, A. Debbouche, R.A. Khan, Study of HIV mathematical model under non-
singular kernel type derivative of fractional order. Chaos, Solitons Fractals 139, 11009 (2020)
58. A.S. Shaikh, K.S. Nisar, Transmission dynamics of fractional order Typhoid fever model using
Caputo-Fabrizio operator. Chaos, Solitons Fractals 128, 355–365 (2019)
59. M.A. Khan, S. Ullah, M. Farhan, The dynamics of Zika virus with Caputo fractional derivative.
AIMS, Math. 4(1), 134–146 (2019)
60. E.F.D. Goufo, Application of the Caputo-Fabrizio fractional derivative without singular kernel
to Korteweg-de Vries-Burgers equation. Math. Model. Anal. 21(2), 188–98 (2016)
61. M. Arfan, H. Alrabaiah, M.U. Rahman, Y.L. Sun, A.S. Hashim, B.A. Pansera, A. Ahmadian,
S. Salahshour, Investigation of fractal-fractional order model of COVID-19 in Pakistan under
Atangana-Baleanu Caputo (ABC) derivative. Results Phys. 24, 104046 (2021)
62. A. Atangana, D. Baleanu, New fractional derivatives with non-local and non-singular kernel.
Theo. Appl. Heat Transf Model Therm. Sci. 20(2), 763–769 (2016)
210 Eiman et al.

63. E.F.D. Goufo, A bio mathematical view on the fractional dynamics of cellulose degradation.
Fract. Calc. Appl. Anal. 18(3), 554–64 (2015)
64. A. Atangana, Extension of rate of change concept: from local to nonlocal operators with
applications. Results Phys. 19, 103515 (2020)
65. A. Atangana, S.I. Araz, Nonlinear equations with global differential and integral operators:
existence, uniqueness with application to epidemiology. Results Phys. 20, 103593 (2021)
66. A. Atangana, S.I. Araz, New concept in calculus: piecewise differential and integral operators.
Chaos Soliton Fractals 145, 110638 (2021)
67. K. Shah, T. Abdeljawad, Study of a mathematical model of COVID-19 outbreak using some
advanced analysis. Waves Random Complex Media 2022, 1–18 (2022)
68. K. Shah, T. Abdeljawad, B. Abdalla, M.S. Abualrub, Utilizing fixed point approach to inves-
tigate piecewise equations with non-singular type derivative. AIMS Math. 7(8), 14614–14630
(2022)
Study of the SIRI Model Utilizing
the Caputo Derivative

Ndolane Sene

Abstract In this article, we make qualitative studies of the SIRI model described
by a fractional operator. In the investigations, the Caputo operator is utilized in the
modeling of the epidemic. We have used the reproduction number and we have
analyzed the stability of the model. A qualitative property as the boundedness of the
solutions of the SIRI model is focused. A numerical scheme including the numerical
scheme of the Caputo operator has been utilized and explained rigorously. The results
of the paper have been illustrated by the graphics of the solutions obtained using the
proposed numerical scheme in the paper.

Keywords Fractional SIRI model · Frcational derivative · Stability analysis ·

Reproduction number

1 Introduction

Modeling world problems via fractional operators has attracted many authors these
last times. The researchers find many applications of fractional operators in frac-
tional calculus fields. The operators have applications in physics [11, 15], they have
applications in biological models [3, 10, 18, 20, 27–29], they have applications
in mathematical physics [13, 16, 26], they have applications in chaos theory [21]
notably findings new types of attractors with the variations of the orders of the frac-
tional derivatives. The pandemic started in 2019 and had many consequences on our
life and generated interest in the epidemical models. The novel problems with frac-
tional operators consist to model the disease models and finding optimal control to
stop the spread of the different diseases. And it is stipulated that fractional operators
can bring advancements in modeling problems. Many operators are introduced in
fractional calculus, we can cite the old fractional operators: the Caputo type and the

N. Sene (B)
Section mathematics and statistics, Institut des Politiques Publiques, Cheikh Anta Diop
University, Dakar Fann, Senegal
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 211
H. Singh and H. Dutta (eds.), Computational Methods for Biological Models,
Studies in Computational Intelligence 1109,
https://doi.org/10.1007/978-981-99-5001-0_9
212 N. Sene

Riemann-Liouville type of derivatives [17, 19]. These two derivatives are the most
used in the fractional operators and recently the new fractional operators with new
types of kernels began to emerge, we have operators from Atangana and Baleanu
[12] and the operator from Caputo and Fabrizio [14]. These new operators grow
many interests this decade with many papers appearing to justify their applicability
in modeling problems.
We review the literature on the epidemic models which have used fractional oper-
ators. In [3], the authors propose the SIR model via the Atangana-Baleanu derivative.
The qualitative properties such as the stability and the existence of the solutions and
the graphical representations of the dynamics have been proposed in this work. In [4],
the authors proposed biological modeling via stochastic perturbation, the bounded-
ness, the permanency, the reproduction number of the model and the graphics of the
stochastic dynamics have been provided in this paper. In [5], for a model against the
novel coronavirus, the present paper gives exciting findings which can serve to reduce
the evolution of the coronavirus. In [6], the authors have proposed the analytical solu-
tions of the biological model using the Natural Transform. In [2], the authors have
proposed the ergodicity property, the extinction of disease, and the density function
of a SIRI model with nonlinear incidence rate and high-order stochastic perturba-
tions. In [1], the authors have described the SIRI model via stochastic perturbation.
In [7], the authors have investigated on SIRS reaction-diffusion epidemic model. In
[8], the problem of the stability study of the equilibriums in epidemic modeling has
attracted many researchers this decade, in their paper, the authors have presented
investigations on the stability of the equilibrium points of some epidemic models.
In [9], the authors investigated the HIV infection model via the fractional operators,
and they have used it in their modeling of the drug therapy effect. In [10], the authors
proposed investigations related to the modeling of the SIR model using the fractional
model. The model presented in this paper has a particularity to be described by the
nonlinear incidence rate. See also in [22, 25].
In the present paper, we address the SIRI model. The objective is to model using
the fractional operator and to study the global stability and the local stability using
the reproduction number. Another important finding of this paper is that we prove
via the mean value theorem that there exists an endemic equilibrium point for the
SIRI model, in other words, we do not need to get its form, we have just to prove
this point exists. The numerical scheme will be proposed and applied to the Caputo
derivative in the way to get the form of the behaviors of the solutions. We will use
the numerical scheme to propose the graphics of the solutions of the considered SIRI
model.

2 Fractional Calculus Operators

Let’s recall the operators used in this chapter. We will use the fractional integral named
the Riemann-Liouville operator integral. We have utilized the Caputo operator in the
investigations. We will recall it too. The function and transformation necessary for
Study of the SIRI Model Utilizing the Caputo Derivative 213

our investigations as the Simon Laplace transformation and the function known as
the Mittag-Leffler function will be recalled as well.

Definition 1 ([17, 19, 23]) Assume that the function w : [0, +∞[−→ R. Let’s the
Liouville-Riemann integral of the function w as the following form
 t
1
I α w (t) = (t − s)α−1 w(s)ds, (1)
(α) 0

where t > 0, α ∈ (0, 1) represents the order of the integral operator, and the function
gamma denotes (...).

Definition 2 ([17, 19, 24]) Assume that the function w : [0, +∞[−→ R. Let’s the
Liouville-Caputo operator of the function w as the form

  1 t
Dcα w (t) = (t − s)−α w (s)ds, (2)
(1 − α) 0

where t > 0, α ∈ (0, 1) represents the order of the integral operator, and the function
gamma denotes (...).

The Laplace transform and the Mittag-Leffler function will be used in the solutions
procedures concerning the fractional equations and then we recall them as well. For
the Laplace transform of the Caputo operator, we obtain the following form[17, 19]
 α  
L Dc w (t) = sα L {w(t)} − sα−1 w(0). (3)

with the condition α ∈ (0, 1). The Mittag-Leffler function is used to describe the form
of the solutions of the differential equations. Here recall this important function for
our investigations
∞
zn
Eα,β (z) = , (4)
n=0
(αn + β)

with α > 0, β ∈ R and the variable t is into the set C.

We have many other fractional operators, we have the CF derivative proposed in
the literature in [14] in 2015. We have another fractional operator known as the frac-
tional derivative with the ML function proposed in the investigation [12]. These two
new operators have had many successes in these recent years, notably in biological
modeling.
214 N. Sene

3 Fractional SIRI Model Under Caputo Operator

In this part, we will focus on the SIRI model under the Caputo derivative. Before
introducing the model we first recall the original model described by the integer-
order operator. We have many classes of SIRI models presented in the research,
here referring to the investigation in the literature, we have the following differential
equations

dS ϑSI
=  − μS − , (5)
dt 1 + νS
dI ϑSI
= − (μ + c + γ) I + δR, (6)
dt 1 + νS
dR
= γI − (μ + δ) R, (7)
dt
The initial condition considered for our present problem can be described in the
following form, that is in the form that

S (0) = S0 , I (0) = I0 , R (0) = R0 , (8)

The signification of the parameters utilized in the SIRI model is assigned in the
following Table, for more clarity, we have that (Table 1).
In the present SIRI model, the incidence function is represented by the function
ϑSI
that f (S, I ) = 1+νS which is considered to be Lipchitz continuous function. The sec-
ond step of the modeling will be to introduce the memory effect in the differential
equations (5)–(7), to arrive at our end we use the Caputo fractional operator which is
known to be an operator with singularity. Therefore, under Caputo fractional deriva-
tive, the fractional differential under consideration will be defined in the following
equations

Table 1 Parameters of the Parameters Descriptions

model
 The recruitment rate
μ The natural death rate
γ The rate of recovery from infection
ϑ The transmission rate
δ The relapse rate
c The death rate generated by infection
Study of the SIRI Model Utilizing the Caputo Derivative 215

ϑSI
Dcα S =  − μS − , (9)
1 + νS
ϑSI
Dcα I = − (μ + c + γ) I + δR, (10)
1 + νS
α
Dc R = γI − (μ + δ) R, (11)

The initial conditions of the model (5)–(7) are maintained, in other words, we con-
serve the same initial conditions, they are defined by the following equation

S (0) = S0 , I (0) = I0 , R (0) = R0 , (12)

We first prove that the total population variable is well-bounded. To arrives at this
end we proceed in the following form by calculating the sum of the three variables
and then we have that P = S + I + R, thus

Dcα P(t) =  − μP(t) − cI (t). (13)

We notice that Eq. (13) is a simple linear differential equation and then we can
determine its exact analytical solution by applying the Laplace transformation, and
then we get that


sα P̄(s) − sα−1 P(0) = − μP̄(s) − cĪ (s),
s

 1 sα−1 sα−1 P(0) cĪ (s)
P̄(s) = − α + α − α . (14)
μ s s +μ s +μ s +μ

The inversion of the Laplace transformation applied to Eq. (14) gives the exact
solution which in particular includes the Mittag-Leffler function, we obtain the form


P(t) ≤ 1 − Eα (−μ (t)α ) + P(0)Eα (−μ (t)α ) . (15)
μ

We conclude that the solutions of our present model are well-bounded because the
total population is bounded. And then, our SIRI model is biologically well definite.
An important remark is it follows from the convergence of the time to infinity, we
get the bound represented in the following relation


P(t) ≤ . (16)
μ

Equation (16) will play a role because it will permit us sometimes to write our model
with two variables and to study the stability analysis.
Let’s analyze the reproduction number which has an important role in the analysis
of the epidemics model, in general. The method to obtain is not new and can be found
216 N. Sene

in the literature, the method utilized in this paper is similar to the technique used in.
The first step is that Eqs. (5)–(7) can be written in the form described by

Dcα x = V − W, (17)

in the previous equation (17) the variable x = (S, I , R), and furthermore the matrix
V and W are described by the form that
⎛ ϑSI
⎞ ⎛ ⎞
1+νS
(μ + c + γ) I − δR
V =⎝ 0 ⎠ and W = ⎝ −γI + (μ + δ) R ⎠ . (18)
ϑSI
0 1+νS
−  + μS
 
The next step consists to apply the Jacobian at the equilibrium point P0 = μ , 0, 0
to the previous matrices V and W , the method to get the free equilibrium is illustrated
in the next section, we get the following
⎛ ϑ ⎞ ⎛ ⎞
μ+ν
0 0 μ + c + γ − 0
JV (E0 ) = ⎝ 0 0 0⎠ and J W (E0 ) = ⎝ −γ μ + δ 0⎠.
ϑ
0 0 0 μ+ν
0 γ
 ϑ   
μ+ν
0 μ + c + γ −
F= and V = . (19)
0 0 −γ μ+δ

To get the reproduction number for our original model in Eqs. (5)–(7), we apply the
definition of the spectral radius on the matrix defined by FV −1 , we have that

  ϑ (μ + δ)
R0 = ρ FV −1 = . (20)
(μ + β) [(μ + c + γ) (μ + δ) − γδ]

The study in this paper will be based on the reproduction number found in Eq. (20).
Notably, the local and global stability will be analyzed according to the reproduction
number R0 .

4 Stability Investigations

In this sub-section, we investigate the stability local of the equilibriums for our model
described in Eqs. (5)–(7). The first part before will be to establish the points where
the equations of our present model are zero. They are obtained by letting that
Study of the SIRI Model Utilizing the Caputo Derivative 217

ϑSI
 − μS − = 0, (21)
1 + νS
ϑSI
− (μ + c + γ) I + δR = 0, (22)
1 + νI
γI − (μ + δ) R = 0, (23)

After calculation, we get as a trivial equilibrium point the point given by the relation
that  

P0 = , 0, 0 . (24)
μ

The nontrivial equilibrium point known in the literature as the endemic point can be
represented by the following relationship that

ϑS ∗ I ∗
0 =  − μS ∗ − , (25)
1 + νS ∗
∗ ∗
ϑS I
0= − (μ + c + γ) I ∗ + δR∗ , (26)
1 + νS ∗
0 = γI ∗ − (μ + δ) R∗ . (27)

The endemic point P1 = (S ∗ , I ∗ , R∗ ) required to prove this point really exist because
it values at this stage is not trivial. To prove the existence of this point we proceed
as follows. We consider that the function I ∗ exists and we have to write the other
comportment depending on this variable I ∗ and we combine it with the use of the
classical mean value Theorem. We have the following for the last variable
γ ∗
R∗ = I . (28)
μ+δ

With the sum between the first and the second equation of our model (5)–(7), we get
the following representation

 − μS ∗ − (μ + c + γ) I ∗ + δR∗ = 0. (29)

Solving Eq. (29) previously mentioned function, we finally arrive at the end that

δγ
 μ+c+γ− μ+δ
S∗ = − I ∗. (30)
μ μ

It proves the existence and the unicity of the variable S ∗ because it depends on the
variable I ∗ . We now continue with the last by proving the variable I ∗ exists and is
unique, to this end we use the mean value Theorem, the sketch is to consider the
function that
218 N. Sene
 δγ

 μ+c+γ− μ+δ
ϑ μ
− μ
i
γ
f (i) =   − (μ + c + γ) + . (31)

δγ
μ+c+γ− μ+δ μ+δ
1+ν μ
− μ
i

Let us calculate the following values including the reproduction number in its expres-
sion. We suppose that a = R0 (μ+ν)
ϑ
and b = 0, we have the forms defined by

γ
f (a) = − (μ + c + γ) + < 0. (32)
μ+δ
 
ϑ γ ϑ 1
f (b) = − (μ + c + γ) + = 1− > 0. (33)
μ + ν μ+δ μ + ν F0

Using the fact that the function f is a bijective function, we arrive at that there exists a
unique I ∗ which is the solution of equation f (i) = 0 and which verifies the expression
defined by the form that
R0 (μ + ν)
0 ≤ I∗ ≤ . (34)
ϑ
We conclude that the endemic point Pe = (S ∗ , I ∗ , R∗ ) exists and is unique.
We consider the endemic point, and due to the complexity, of the calculations, we
try to establish the global stability of the endemic equilibrium point using the Lya-
punov direct method. To finish, we take the Lyapunov function which is represented
by
L (S, I , R) = L1 (S, I , R) + L2 (S, I , R) + L3 (S, I , R) . (35)

We calculate step by step the Caputo derivative operator applied to the previous
function represented in the previous Eq. (35). Note that we suppose that the function
L1 is represented by
   
S
∗ ∗ ∗ ∗ I
L1 (S, I , R) = S − S − S ln + I − I − I ln ∗ , (36)
S∗ I

Let’s calculate the Caputo operator to the previous function respect to the trajectories
of the system represented in Eqs. (5)–(7), we obtain the following relationships
   
S∗ I∗
Dcα V1 ≤ 1 − Dcα S + 1 − Dcα I ,
S I
  
S∗ ϑSI
≤ 1−  − μS − ,
S 1 + νS
  
I∗ ϑSI
+ 1− − (μ + c + γ) I + δR . (37)
I 1 + νS
Study of the SIRI Model Utilizing the Caputo Derivative 219

Using the fact the endemic point satisfied the following conditions, in other words
applying the conditions that

ϑS ∗ I ∗
 = μS ∗ + , (38)
1 + νS ∗
ϑS ∗ I ∗
= (μ + c + γ) I ∗ − δR∗ , (39)
1 + νS ∗
γI ∗ = (μ + δ) R∗ . (40)

Applying Eqs. (38)–(41), the calculations can be expressed in the following forms,
they are given by the form
  
(S − S ∗ )2 S∗ αSI ϑS ∗ I ∗
Dα L1 ≤ −μ − 1− − ,
S S 1 + νS 1 + νS ∗
  
I∗ ϑSI ϑS ∗ I ∗
+ 1− − ,
I 1 + βS 1 + νS ∗
   
I∗ I∗
− (μ + c + γ) 1 − I − I∗ + δ 1 − R − R∗ . (41)
I I

We suppose the second Lyapunov function described by the function given in the
next line and we calculate it Caputo derivatization
α,ρ
L2 (S, I , R) = κIt I (t − τ ) − I ∗ , (42)

We apply the Caputo derivative to the previous function and we consider the trajec-
tories of our model we get the following forms, which are described by the forms

Dα L2 = κDα Itα I − I ∗
2
,
∗ 2
= κ I −I . (43)

We end by considering the last Lyapunov function and we calculate its Caputo oper-
ator derivation by considering our model (5)–(7). The Lyapunov function considered
in this last section is represented by the form that
 
∗ R ∗
L3 (S, I , R) = R − R − R ln . (44)
R∗

As a previous procedure, we consider the trajectories of the Eqs. (5)–(7) and we

apply the Caputo derivative on the previous function described in the previous Eq.
(44), we have that
220 N. Sene
 
R∗
Dα,ρ L3 ≤ 1 − Dα,ρ R,
R
 
R∗
≤ 1− [γI − (μ + δ) R] . (45)
R

Here, we consider the endemic equilibrium point, for the simplification of the cal-
culations, and then we obtain the following relations
   
α R∗ ∗ R∗
D L3 ≤ γ 1 − I − I − (μ + δ) 1 − R − R∗ . (46)
R R

Note that the condition that R0 ≤ 1 ensures the existence of the endemic equilibrium
point as previously proved in this section. Summing Eqs. (41), (43), and (47), we get
the following condition

Dα L = Dα L1 + Dα L1 + Dα L3 ≤ 0. (47)

And then using the Lyapunov direct method characterization, and considering the
condition Eq. (47), we summarize that the endemic point is global asymptotic stable
when the condition that R0 > 1 is held.

5 Results and Discussions

This section, we present the numerical procedure and analyze the behaviors of the
dynamics of the SIRI epidemic model described by the Caputo operator. We notably
focus on the impact of the order of the derivative operator in the dynamics of the
model. The first sub-section addresses the numerical procedure used to get the graph-
ics.

5.1 Solution Procedure

This section describes the numerical procedure adopted in the present paper. This
section is significant in the computational epidemic model because it will permit
drawing the dynamics of the considered epidemic model. We begin by determining
the form of the solution by the utilization of the fractional integral, we get the forms

S(t) = S (0) + I α u (t, x) , (48)

α
I (t) = I (0) + I v (t, x) , (49)
R(t) = R (0) + I α w (t, x) , (50)
Study of the SIRI Model Utilizing the Caputo Derivative 221

where x = (S, I , R) and the functions u, v, and w are represented in our context by
the elementary functions defined by the forms that

ϑSI
u (t, x) =  − μS − , (51)
1 + νS
ϑSI
v (t, x) = − (μ + c + γ) I + δR, (52)
1 + νI
w (t, x) = γI − (μ + δ) R, (53)

The next part consists to find the numerical scheme of the integral used in the previous
solutions (51)–(53). And then at points (tn , xn ), we get the following form in the
procedure of the discretization


s
  
n
(α)  
α α α
I u (tn , xn ) = h n,j u tj , xj + h n−j u tj , xj , (54)
j=0 j=0


s
  
n
(α)  
I α v (tn , xn ) = hα n,j v t ,
j jx + hα
n−j v tj , xj , (55)
j=0 j=0


s
  
n
(α)  
I α w (tn , xn ) = hα n,j w t ,
j jx + hα
n−j w tj , xj . (56)
j=0 j=0

here we consider that h is the step size and the starting weight is denoted in our
previous equation by the symbol that n, j. Furthermore, the starting weight satisfies
the condition expressed in the following relation, for more explanation about this
formula see in [30]


s 
s
 (γ + 1)
γ γ
n,j j =− n−j j + nγ+α (57)
j=0 j=0
 (γ + 1 + α)

where the assumption on the parameter γ can be found in [30], note that for n =
0, 1, 3, ... we obtain the following relationships

(α)  (1 − α)
= (−1)n (58)
n
 (n + 1)  (1 − α − n)

Utilizing the numerical discretization described in Eqs. (54)–(56) for the Riemann
Liouville integral we get as numerical discretization for our model the relationships
defined by the form that
222 N. Sene


s
  
n
(α)  
x(tn ) = x (0) + hα n,j u tj , xj + hα n−j u tj , xj , (59)
j=0 j=0


s
  
n
(α)  
y(tn ) = y (0) + hα n,j v tj , xj + hα n−j v tj , xj , (60)
j=0 j=0


s
  
n
(α)  
α α
z(tn ) = z (0) + h n,j w tj , xj + h n−j w tj , xj . (61)
j=0 j=0

The numerical procedure of the functions utilized in Eqs. (5)–(7), in other words,
used in our epidemic model are given in the forms given as

  ϑSj Ij
u tj , xj =  − μSj − , (62)
1 + νSj
  ϑSj Ij
v tj , xj = − (μ + c + γ) Ij + δRj , (63)
1 + νSj
 
w tj , xj = γIj − (μ + δ) Rj , (64)

In this new part, we give the graphics and analyze the influence of the fractional
operator used in our modeling. In this section, we suppose that S(0) = 5, I (0) = 10
and R(0) = 0. let that  = 20, μ = 0.3, ϑ = 0.375, ν = 10, c = 0.4, γ = 0.2, and
δ = 0.02. In the first graphic, we utilize that the order of the fractional operator is
considered as α = 0.95. We obtain the following graphics (Figs. 1 and 2).
To see, the impact or the influence of Caputo derivative operator, we now consider
the order fixed at α = 0.85 and we try to observe the behaviors of the dynamics of
the considered SIRI model. We have the following graphics (Figs. 3 and 4).
We notice that the impact or the influence of the order of the derivative opera-
tor exists, we can notice that the graphics of the susceptible increase more rapidly

60 10

9
50
8

7
40
6

30 5

4
20
3

2
10
1

0 0
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70

(a) (b)

Fig. 1 Susceptible individuals versus time (a). Infective individuals versus time (b)
Study of the SIRI Model Utilizing the Caputo Derivative 223

1.4
1.5
1.2

1
1

0.8 0.5

0.6
0
0
0.4

20
0.2

40
0 0
0 10 20 30 40 50 60 70 2
4
60 6
8
10

(a) (b)

Fig. 2 Recovered individuals versus time (a). SIRI in three dimensional (b)

70 10

9
60
8

50 7

6
40
5
30
4

20 3

2
10
1

0 0
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70

(a) (b)

Fig. 3 Susceptible individuals versus time (a). Infective individuals versus time (b)

1.4
1.5
1.2

1
1

0.8 0.5

0.6
0
0
0.4
20

0.2 40

0 60
0
0 10 20 30 40 50 60 70 2
4
80 6
8
10

(a) (b)

Fig. 4 Recovered individuals versus time (a). SIRI in three dimensional (b)
224 N. Sene

when the order α = 0.85 than when the order of the operator is α = 0.95. The same
behavior can be noticed with the recovery individuals, where the pic of the recovered
is attended more rapidly by the order α = 0.85 than when the order of the derivative
operator is α = 0.95. The conclusion is that the order of the operator plays a deceler-
ation effect in the case of modeling the SIRI model. Thus the order of the fractional
operator can be used to control the considered disease.

6 Conclusion

This paper has discussed the SIRI model via the Caputo operator. We have introduced
the fractional operator in modeling the biological model. The equilibrium points of
the model have been provided and their stability analysis using the reproduction
number has been studied. The proof of the existence of the endemic equilibrium
via the mean value theorem is an interesting result in this paper. The second part
of this proposition is the numerical discretization used to obtain the solutions of the
proposed epidemic model via the Caputo derivative. The optimal control studies are
not proposed in this chapter and it can be interesting to focus in this direction in
future investigations.

Conflict of Interest The authors declare that they have no conflict of interest.

References

1. Tomás Caraballo, Mohamed El Fatini, Roger Pettersson, and Regragui Taki, A stochastic SIRI
epidemic model with relapse and media coverage, Discrete & Continuous Dynamical Systems
- Series S, 23(8), 2018
2. B. Zhou, D. Jiang, Y. Dai, T. Hayat, Ergodic property, extinction, and density function of a
SIRI epidemic model with nonlinear incidence rate and high-order stochastic perturbations.
Math. Methods Appl. Sci. 45(3) (2022)
3. N. Sene, SIR epidemic model with Mittag-Leffler fractional derivative. Chaos, Solitons Fractals
137, 109833 (2020)
4. N. Sene, Analysis of the stochastic model for predicting the novel coronavirus disease. Adv.
Differ. Equ. 2020, 568 (2020)
5. A. Omame, N. Sene, I. Nometa, C.I. Nwakanma, E.U. Nwafor, N.O. Iheonu, D. Okuonghae,
Analysis of COVID-19 and comorbidity co-infection model with optimal control. Opt. Control
Appl. Methods 42(6), 1568–1590 (2021)
6. N. Ali, R. Nawaz, L. Zada, A. Mouldi, S.M. Bouzgarrou, N. Sene, Analytical approximate
solution of the fractional order biological population model by using natural transform. J.
Nanomater. 2022, Article ID 6703086, 12 (2022)
7. B. Li, Q. Bie, Long-time dynamics of an SIRS reaction-diffusion epidemic model J (Math. Anal,
Appl, 2019)
8. D. Rostamy, E. Mottaghi, Stability analysis of a fractional-order epidemics model with multiple
equilibriums. Adv. Differ. Equ. 2016, 170 (2016)
Study of the SIRI Model Utilizing the Caputo Derivative 225

9. A.A.M. Arafa, S.Z. Rida, M. Khalil, A fractional-order model of HIV infection with drug
therapy effect (J. Egyptian Math, Soc, 2013)
10. A. Mouaouine, A. Boukhouima, K. Hattaf, N. Yousfi, A fractional order SIR epidemic model
with nonlinear incidence rate. Adv. Differ. Equ. 2018, 160 (2018)
11. A. Atangana, T. Mekkaoui, Trinition the complex number with two imaginary parts: Fractal,
chaos and fractional calculus. Chaos, Solitons Fractals 128, 366–381 (2019)
12. A. Atangana, D. Baleanu, New fractional derivatives with nonlocal and non-singular kernel:
theory and application to heat transfer model. Thermal Sci. 20(2), 763–769 (2016)
13. N.A. Shah, I. Khan, Heat transfer analysis in a second-grade fluid over and oscillating vertical
plate using fractional Caputo-Fabrizio derivatives. Eur. Phys. J. C 76, 362 (2016)
14. M. Caputo, M. Fabrizio, A new definition of fractional derivative without singular kernel. Progr.
Fract. Differ. Appl 1(2), 1–15 (2015)
15. D. Avci, M. Yavuz, N. Ozdemir, Fundamental Solutions to the Cauchy and Dirichlet Problems
for a Heat Conduction Equation Equipped with the Caputo-Fabrizio Differentiation (Nova
Science Publishers, 2019), pp. 95–107
16. D. Avci, N. Ozdemir, M. Yavuz, Fractional Optimal Control of Diffusive Transport Acting on
a Spherical Region. Methods of Mathematical Modelling: Fractional Differential Equations
(CRC Press, 2019), pp. 63–82
17. A.A. Kilbas, H.M. Srivastava, J.J. Trujillo, Theory and Applications of Fractional Differential
Equations. North-Holland Mathematics Studies (Elsevier, Amsterdam, 2006), p. 204
18. A. Jajarmi, A. Yusuf, D. Baleanu, M. Inc, A new fractional HRSV model and its optimal
control: A non-singular operator approach. Phys. A: Stat. Mech. Its Appl. 547, 123860 (2020)
19. I. Podlubny, Fractional Differential Equations. Mathematics in Science and Engineering (Aca-
demic, New York, 1999), p. 198
20. X. Wang, Z. Wang, Dynamic analysis of a delayed fractional-order SIR model with saturated
incidence and treatment function. Int. J. Bifurcat. Chaos 28(14), 1850180 (2018)
21. N. Sene, Analysis of a four-dimensional hyperchaotic system described by the Caputo-Liouville
fractional derivative. Complexity 2020, Article ID 8889831, 20 (2020). https://doi.org/10.1155/
2020/8889831
22. A. Elazzouzi, A.L. Alaoui, M. Tilioua, D.F.M. Torres, Analysis of a SIRI epidemic model with
distributed delay and relapse, Stat. Opt. Inform. Comput., 7 (2019)
23. J. Fahd, T. Abdeljawad, D. Baleanu, On the generalized fractional derivatives and their Caputo
modification. J. Nonlinear Sci. Appl. 10, 2607–2619 (2017)
24. J. Fahd, T. Abdeljawad, A modified Laplace transform for certain generalized fractional oper-
ators. Results Nonlinear Anal. 2, 88–98 (2018)
25. A. Lahrouz, L. Omari, D. Kiouach, Global analysis of a deterministic and stochastic nonlinear
SIRS epidemic model. Nonlinear Anal.: Model. Control 16(1), 59–76 (2011)
26. M. Yavuz, Characterization of two different fractional operators without singular kernel. Math.
Model. Nat. Phen. 14(3), 302 (2019)
27. H. Singh, Analysis of drug treatment of the fractional HIV infection model of CD4 T-cells.
Chaos, Solitons Fractals 146, 110868 (2021)
28. H. Singh, Computational study of fractional order smoking model. Chaos, Solitons Fractals
142, 110440 (2021)
29. H. Singh, Analysis for fractional dynamics of Ebola virus model. Chaos, Solitons Fractals 138,
109992 (2020)
30. R. Garrappa, Numerical solution of fractional differential equations: a survey and a software
tutorial mathematics 6(2), 16 (2018)
Implementation of Vaccination
in an Epidemic Model for COVID-19

Yerra Shankar Rao

Abstract The vaccination campaign against the pandemic diseases started in India
in 1st week of January 2021, and by 10th December 2021, there are 13, 123, 75,088
people have already been vaccinated. It is observed that people who were vacci-
nated but again infected with the disease COVID-19. This article establishes the
SVIR (Susceptible, Vaccination, Infectious and Recovered) pandemic paradigm. The
proposed model’s general characteristics are first discussed. Particularly, both theo-
retical and numerical analysis is done on the basic reproduction numbers. Reproduc-
tion numbers are used to describe both the local and global stability of equilibrium
points. Secondly, determine the stability of the equilibrium points using the Lypanouv
function and the Painocare-Bandixson condition. The transmission of Covid-19 and
its regulation are shown in simulation with the aid of MATLAB according to genuine
data in India. In the suggested model, the effects of vaccination are investigated. This
simulated result is consistent with the actual parameters and indicates the necessity
for booster doses to eradicate the virus from the population.

Keywords Vaccinations · Epidemic model · Lyponouv function · Stability

Nomenclature

Notations Descriptions
N Total number of populations
S Number of the susceptible population at the time t
I Number of the infected population at the time of t
V Number of vaccinated populations
R Number of recovered population after infected at the time t
A Influx rate that represented the migrant population

Y. S. Rao (B)
Department of Mathematics, NIST Institute of Science and Technology (Autonomous), Institute
Park, Pallur Hills, Berhampur, Odisha, India
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 227
H. Singh and H. Dutta (eds.), Computational Methods for Biological Models,
Studies in Computational Intelligence 1109,
https://doi.org/10.1007/978-981-99-5001-0_10
228 Y. S. Rao

β The infected rate at time t in the Corena transmission model

γ Reassembly rate of vaccination at time t
α Recover rate of the infected population at time t
ε Vaccinated population getting the infected rate at time t
d1 Natural death i.e. death rate other than COVID-19
d2 The death rate due to COVID-19

1 Introduction

The ongoing pandemic disease is Coronavirus (COVID-19) caused by severe acute

respiratory syndrome coronavirus-2(SARS-COV-2). This disease spreads rapidly
all over the world and its transmission rate is very high. The outbreak of SARS-
CoV-19 first appears in Wuhan city of China, in the last of December 2019[1].
Then it was spread worldwide across the globe. Therefore due to the imbalance
situation in the world on March 11th, 2020, the world health organization (WHO)
declared it a pandemic. There are more than 269,468,456 people infected, having
death 53, 11,895, across 169 countries as on 10th December 2021. But in India,
there were 34, 68, 2736 people infected and the death due to covid-19 was 47,
4479 people as on 10th December 2021. So the impact of covid-19 in various fields
like socio-economic imbalance, educational organizations, small-scale and large-
scale industries, unemployment, etc. Initially, the main symptoms of this disease are
mild cold with cough, high fever, aches, tiredness, loss of taste, headache, diarrhea,
chest pain, and shortness of breath whereas in complicated cases may include severe
pneumonia in lungs. Primarily this disease spreads with close contact with infected
persons, during coughing, sneezing, and talking. [2] Studied the non-integer order
smoking model using the memory principle. This method is used to that model to
minimize smoking in society. Infectious disease modeling in real-world problems like
COVID-19, Cancer, diabetes, and other diseases was discussed. The author mainly
discussed the basic concept of modeling and how it was implemented in a real-world
problem. The fixed point theory used for [3] analysis of COVID-19 transmission
and its mitigations in the sense of fractional order Caputo operator. This operator
provides deeper grasp of the epidemic diseases.
At an early stage, the government had taken initiatives like complete lock-down,
contact tracing and implementation of strict social distancing measures, and quar-
antine processes either home or hospitalization [4] studied the SIQR model for
preventing the covid-19. They mainly emphasize that people remain isolated to
prevent or control the disease. [5] adopted three quarantine processes according
to their stages. They were mainly to see the transmission behavior of the virus and
its preventive measures stages. [6] proposed a mathematical model and suggested
that the hospitalized quarantine is the best policy for curb the Covid-19. The study
conducted by [[7] designed an epidemic model that the double quarantine comes
close to preventing diseases. [8] carried out to investigation on the movement of
Implementation of Vaccination in an Epidemic Model for COVID-19 229

migrant People in Orissa during the lockdown period. The government of Odisha
adopted the quarantine method in schools or local areas for suppressing COVID-19.
In this paper, the authors [9] perform a trend analysis of the fatality rate, impact of
lockdown, and infection rate with valid data. They take the example of Bhilwara,
a city in Rajasthan, India which successfully controls diseases by use of tracing,
testing, and treatment technique. Due to the implementation of the above preventa-
tive measures the rate of active cases somehow decreased. The infected population
trajectory for India may have shown positive signs of slowing down. [10] In this paper,
the authors proposed an improved SEIR model to predict the covid-19 trends in Italy,
the United State of America, and New York. However, continuing this process leads
to chaos in society. Vaccination is the main weapon to control or prevent the spread of
the disease. Implementation of blocking measures, admitting mild patients into the
hospital, and tracking of the covid-19 cases controlled by Singapore, and other coun-
tries studied in this paper [11]. So all over the globe desperately started vaccination
campaigns as a marathon journey to eradicate or reduce the COVID-19 Pandemic. In
2020 there were more than 50 companies developed a vaccine to fight against SARS-
COV-2. The U.K. government first approved the vaccine on 2nd Dec 2020 named as
Pfizer-BioNTech BNT162b26. Then followed by the Oxford-AstraZeneca vaccine7
on 20th December 2020, and subsequently, the pharmaceutical company Moderna
TX, Inc named Moderna COVID-19 vaccine started on 8th Jan 2021, whereas only
the two companies vaccines Pfizer-BioNTech and Oxford-AstraZeneca were widely
provided to the population of U.K. by February 2021.
Now there are different types of vaccines available in the globe like a whole virus,
protein subunit, viral vector, and nucleic acid (mRNA, DNA). But the Indian govern-
ment approved three vaccines, out of which Covishield manufactured by Serum
Institute of India, Pune in collaboration with Oxford University-AstraZeneca, and
Covaxin manufactured by Bharat Biotech International Limited Hyderabad in collab-
oration with ICMR India are being used extensively for the vaccination program. In
the first phase, the vaccination program was started on January 16th, 2021, covering
the health care workers like doctors and nurses, then it was extended to frontline
workers In early February, and the vaccination program was expanded to include
frontline workers on 2nd February 2021. On March 1st, 2021 the government of
India started 2nd phase of vaccination campaigns for people of senior citizens above
60 years and people above 45 years of age with comorbidity. The country launched
vaccination for all people aged more than 45 years on April 1st , 2021. Subsequently,
the government then decided to expand its vaccination drive by allowing everyone
above 18 years to be vaccinated from 1st May 2021. As a result, more than 13,
123, 75,088 doses are given to the people of India from 16th Jan 2021 to 10th
December 2021. Till now there is no vaccine approved by the government of India
for the age group below 18 years. Governments plan to implement the vaccination
program for below 18 years people and kids not to increase the infection rate of
the above age groups. An epidemic model is a tool used to study the mechanism
through which the virus transmits, predict the future of the epidemic, and iden-
tify possible preventive measures for the virus. Some of the researchers studied the
different mathematical modeling of different diseases like the Ebola virus, Cancer
230 Y. S. Rao

treatment, HIV, etc. [12] analyzing the Ebola virus model in fractional modeling.
In that paper, he was shown the mathematical stability of the virus with the help of
CPU time computations. Analysis of drug treatment of HIV infection model CD4+
T cell was discussed by [2]. He was used to show the efficiency of the treatment of
HIV infected population. He was using the memory principle and decentralization of
the said diseases [13]. In this book, the authors describe the diseases modeling and
computational modeling of different diseases in real-world problems and their reme-
dies by different approaches. Many researchers have studied a different mathematical
model for covid19 and related issues. Reference [14] mainly focusing on the funda-
mental solution for preventing COVID-19 and efficient vaccines through a mathe-
matical model. The herd immunity concept and its implications for Covid-19 were
conducted by [15]. Reference [16] discussed the effectiveness and complicatedness
of the vaccine in real-world conditions for Pfizer-BioNTech’s and Moderna’s mRNA
vaccines. [17] proposed a mathematical model that explains vaccination strategy
is a major role in reducing social distance and minimize the covid-19 infections.
Reference [18] In this paper, the authors have discussed the age structure model
and suggested that by providing the vaccine to the elder age group, the mortality
rate can be decreased. Also, they suggest an effective vaccine given to all people of
Korea to reduce the incidence and mortality rate. Reference [19] focus on the optimal
duration of the vaccine campaign properly for eliminating the infection at Marocco.
They were taking a statistical method named Bootstrap for estimating the param-
eter and characterized the optimal function of the Pontryagins maximal principle for
optimal control of the diseases. This paper [17] suggested the high rate of vaccina-
tion given to people covid-19 cases should be resurgence. Reference [20] studied the
process of preparing vaccines after dealing with multiple morbidities and fatalities
and also said that personal health education is required for the destruction of the
virus. Reference [21] the 1st year covid-19 vaccine program carried out by Oliver
et al. They studied the global impact of vaccination programs to determine lives lost
if no vaccine was distributed. The above-mentioned work on mathematical modeling
of vaccination supply and the infection rate is constant. Comparing the vaccination
policy in different counties like Italy, South Africa, and India that leads a positive
impact and reduces the mortality rate The vaccination policy and their implementa-
tion effect in India, Italy, and the UK comparison that a positive impact. Receiving
of vaccine doses of [22] Delta variant and alpha variant virus and the effectiveness of
the doses supporting maximize vaccine and minimize the infection discussed in the
country UK. The existing work has been done on the variable of infection rate. The
sack of that work leads to human congestion and trouble for some people infected.
Thus we present the transmission model with vaccinated people getting infected and
susceptible, motivated by the work and reports.
Recently there is a new mutant or variant of COVID-19 was detected first in
South Africa on 24th November 2021. It is observed by scientists and medical
researchers that the virus spreads very rapidly across all countries including Australia
and European countries. On 26th November 2021, WHO declared the name of the
new Covid-19 virus as Omicron. In India, on 2nd December the first case of Omicron
was confirmed in Karnataka. As of 12th December 2021, there were 38 people were
Implementation of Vaccination in an Epidemic Model for COVID-19 231

detected with the omicron virus in India. The U.K. declared the first death due to the
omicron variant of covid -19 on 13th December 2021.
Fully vaccinated people should require the booster dose for preventing Covid-
19 virus transmission. The contribution of the study is that stability analysis is
performed in both analytical and graphical ways by using the data available in the
world meter. The behavior of the SVIR model is explained in mathematically using
a nonlinear system of ordinary differential equations and its solutions are performed
by MATLAB simulations. The rate of transmission, fatality, rate of vaccinations, and
rate of recovery from the susceptible people are considered in this model. Our results
were compared with other countries’ results like the USA, and the U.K. the model
constructed in this chapter is the extension of standard SIR models including a new
compartment for vaccination. That is essential for SARS COVID-19 transmission
dynamics. Our model uses continuous vaccination.
In this paper, we try to understand the transmission behavior of COVID-19 in the
populations with a briefing assessment of the transmission rate fatality rate vaccina-
tion rate, and the recovery rate in the bad effect of vaccination in India as well as
the world. This study determines the potential anti corena virus vaccine. This shows
eradication success depends on the vaccine and its coverage used.
The remainder of this paper is organized as follows in Sect. 2 a mathematical
approach is described briefly. In Sect. 3 we analyze the stability of equilibriums and
their behaviors. Mathematical results are explained by numerical simulations and
control techniques are given in Sect. 4. Lastly, we summarise the work and propose
the future scope.

2 Mathematical Hypothesis

To explain the mathematical model for the transmission of diseases we consider a

deterministic compartmental modeling approach. We consider the total number of
population N can be divided into four sub-populations or groups susceptible, infected,
vaccinated, and recovered populations. Because of COVID-19 Vaccine is available
in the hospital so the transition of the vaccinated and susceptible population and
the infected population is given at the rate γ and ε respectable. Since the efficacy
of the vaccine is not a hundred percent so the population was infected after taking
the vaccine. The influxes of the population are at the rate A getting susceptible. The
susceptible population was in contact with the infected population get infectious
at the rate β. We assume that the rate of infection is very less after getting the
vaccination. Those people who were infected were treated with hospitalization or
isolation getting recovery at the rate α. The death rate considers as d2 due to COVID-
19. The people were suffering from severe diseases like sugar, heart, blood pressure,
and low immunity power their death/ natural death rate consider as d1. We consider
the recovery population died at the rate of d1 and the susceptible at the rate δ. The
assumption of the model we represented a system of nonlinear systems of the ordinary
differential as in Fig. 1 and its equation.
232 Y. S. Rao

Fig. 1 Diagrammatic δR
representation of the model

A S βSI αI
I R

d1S (d1+d2)I d1R

γV εV
V

(d1+d2)V

3 Mathematical Model for Vaccination Epidemic Model

3.1 The Mathematical Equation of the Model

dS
= A − d1 S − β S I + γ V + δ R
dt
dI
= β S I + εV − (α + d1 + d2 )I
dt (1)
dV
= −(γ + ε + d1 + d2 )V
dt
dR
= α I + (δ + d1 )R
dt

3.2 Mathematical Analysis

Throughout the paper we assume that the initial conditions of the system (1) are
nonnegative:

S(0) = S0 ≥ 0, I (0) = I0 ≥ 0, V (0) = V0 ≥ 0, R(0) = R0 ≥ 0

3.3 Positivity and Boundedness of the Solutions

In this section, we shall define a region within which the solution of the model is
uniformly bounded as the set  ∈ 4+ for the populations. The total population is
defined as
Implementation of Vaccination in an Epidemic Model for COVID-19 233

N (t) = S(t) + I (t) + V (t) + R(t)

Hence dN
dt
= dS
dt
+ dI
dt
+ dV
dt
+ dR
dt

dN
= A − d1 N − d2 (I + V )
dt
dN
⇒ ≤ A − d1 N
dt
dN
⇒ ≤ dt
A − d1 N

Integrating both sides we have

 
dN
≤ dt
A − d1 N
ln(A − d1 N )
⇒− ≤t +C
d1

where C is a constant of Integrations

ln(A − d1 N ) ≥ −(d1 t + C)
Thus .
⇒ (A − d1 N ) ≥ C1 e−d1 t
Where C1 is a constant of integrations.
Since N (0) = N0 then (A − d1 N0 ) ≥ C1 .
Accordingly,

(A − d1 N ) ≥ (A − d1 N0 )e−d1 t
A (A − d1 N0 )e−d1 t
⇒ N (t) ≤ −
d1 d1
A
⇒ N (t) = ; t → ∞
d1
 
Thus N (t) ∈ 0, dA1 .
Hence the feasible set of the
 solution of the model equations
 enter and remains
in the invariant regions:  = (S, I, V, R) ∈ + : N (t) ≤ d1 .
4 A

4 Basic Reproduction Number

In an emerging epidemic, the basic reproduction number is usually useful as a

measure of the strength of the control measure needed to break the epidemic. This
number is the spectral radius of the next-generation matrix. According to the next-
generation matrix, one first identifies the infected subsystem of the model. These are
234 Y. S. Rao

the set of equations that describe the new infection and change in the population of
the system. The infected population is linearised about the infection-free equilibrium
forming the Jacobian Matrix. This Jacobian matrix is decomposed as J = T − K.
Where T is the rate of appearance of new infection in the population and K represents
the rate of transformation of the population into the population. The matrix T and K
are obtained as
⎛ ⎞
β S0 0 0
T = ⎝ 0 0 0⎠
0 00
⎛ ⎞
(α + d1 + d2 ) −ε 0
K =⎝ 0 (ε + d1 + d2 ) 0 ⎠
−α 0 (d1 + δ))
⎛ 1 −ε
⎞
(α+d1 +d2 ) (α+d1 +d2 )(ε+d1 +d2 )
0
⎜ ⎟
Lastly K −1 = ⎝ 0 1
(ε+d1 +d2 )
0 ⎠.
−α αε
−α (α+d1 +d 1
2 )(d1 +δ) (α+d1 +d2 )(ε+d1 +d2 )(d1 +δ) (d1 +δ)
Thus the next-generation matrix is
⎛ ⎞
β S0 −β S0 ε
(α+d1 +d2 ) (α+d1 +d2 )(ε+d1 +d2 )
0
⎜ ⎟
T K −1 = ⎝ 0 0 0⎠
0 0 0

The eigenvalues of the next-generation matrix are


β S0
ρ= (α+d1 +d2 )
00

Hence the spectral radius is

β S0 βA
ρ(T K −1 ) = R0 = = .
(α + d1 + d2 ) d1 (α + d1 + d2 )

5 Stability Analysis

In this section, we can discuss two equilibrium points of the epidemic model. i.e.
Corena free equilibrium point and endemic equilibrium point. These two equilibrium
points are based on the existence of the Corena virus in the population over continuous
time. The Corena-free equilibrium point is the point at which the corena virus does
Implementation of Vaccination in an Epidemic Model for COVID-19 235

not spread in an area because the infected population equal to zero during the time.
While the endemic equilibrium point is the point at which the virus must spread in
a defined area.
The Corena Free equilibrium (CFE) of the model is given by

E 0 = (S0 , I0 , V0 , R0 ) = (S0 , 0, 0, 0)

The endemic equilibrium of the model is given by

Aε(d1 + δ) + γ (d1 + δ)(d1 + d2 + α)(1 − R0 )I ∗ + εδα I ∗

S∗ =
ε(d1 + δ)(β I ∗ + d1 )
(d1 + d2 + α)(1 − R0 )I ∗
V∗ =
ε
∗
α I
R∗ =
(d1 + δ)

Theorem 1 The CFE is locally asymptotical stable if the reproduction number, R0

< 1, and endemic equilibrium if it exists, is unstable.

Proof According to the Next generation Matrix Principle

⎛ ⎞
−d1 −β S0 γ
JC F E = J (S0 , 0, 0) = ⎝ 0 β S0 − (d1 + d2 + α) ε ⎠
0 0 −(ε + d1 + d2 )

The eigenvalues of the Jacobian matrix J are

λ1 = −d1
λ2 = −(ε + d1 + d2 )
λ3 = β S0 − (d1 + d2 + α) < 0
β S0
⇒ λ3 = <1
(d1 + d2 + α)
β S0
λ3 = = R0 < 1
(d1 + d2 + α)

All eigenvalues of the Jacobian matrix are strictly negative provided less than
unity, then CFE is locally asymptotically stable, and Endemic equilibrium is unstable.
Hence the Corena disease invades the population. However, the existence of endemic
equilibrium implies that the Corena disease may persist in the population.
236 Y. S. Rao

Theorem 2 (Local Asymptotic stability of endemic Equilibrium) The endemic

equilibrium E* of the model (1) is locally asymptotically stable for R0 > 1.

Proof In order proof the endemic equilibrium of the system we can linearization the
Eq. (1) and the Jacobian matrix is given by

E E∗ E = E ∗ (S ∗ , I ∗ , V ∗ , R ∗ )
⎛ ⎞
−d1 − β S ∗ −β S ∗ γ δ
⎜ βI∗ β S ∗ − (d1 + d2 + α) ε 0 ⎟
=⎜⎝
⎟
⎠
0 0 −(d1 + d2 + ε) 0
0 α 0 −(d1 + δ)

The eigenvalues of E* are obtained by solving the equation det(E*-λI) = 0.

Thus we have that χ1 = −(d1 + d2 + ε) which has negative real parts and the
other three can be obtained by solving the cubic equation χ 3 + aχ 2 + bχ + c = 0.
Where

a = (d1 + β S ∗ − β S ∗ + (d1 + d2 + α) + (d1 + δ))

 
(d1 + β S ∗ )(d1 + δ) − (d1 + δ)(β S ∗ − (d1 + d2 + α))
b=
−(d1 + β S ∗ )(β S ∗ − (d1 + d2 + α) + αε − β 2 I ∗ S ∗ .
 
−(d1 + β S ∗ )(d1 + δ)(β S ∗ − (d1 + d2 + α)
c=
+(d1 + β S ∗ )αε + β 2 I ∗ S ∗ (d1 + δ) − β I ∗ αδ

Clearly a.b > c.

So by the Raurth-Hurtwz condition, endemic equilibrium is locally asymptotic
stable.

6 Global Stability for Infection-Free Equilibrium

4
Theorem 3 The disease-free equilibrium is global asymptotical stable in R+ , if
R0 < 1.

Proof Consider the Lyapunov function of the form

Implementation of Vaccination in an Epidemic Model for COVID-19 237

1
M= I
(α + d1 + d)
dM 1 dI
⇒ = ( )
dt (α + d1 + d2 ) dt
dM 1
⇒ = (β S I − (α + d1 + d2 )I + εV )
dt (α + d1 + d2 )
dM βSI (α + d1 + d2 )I εV
⇒ = − +
dt (α + d1 + d2 ) (α + d1 + d2 ) (α + d1 + d2 )
dM βS εV
⇒ =( − 1)I +
dt (α + d1 + d2 ) (α + d1 + d2 )
dM εV
⇒ = (R0 − 1)I +
dt (α + d1 + d2 )
dM εV
⇒ ≤ (R0 − 1)I +
dt (α + d1 + d2 )
dM
⇒ ≤0
dt

If R0 < 1 then ddtM ≤ 0. Hence by [23, 24] Maximum invariant principle, the E0
is globally asymptotically stable in the region.

7 Convergence Theorem for Lyapunov Stability

Theequilibriumpoint xeq ∈ Rn is Lyapunov

 stability if for all xeq ∈ ε > 0,∃δ >
0 : x(t0 ) − xeq  ≤ δ⇒ x(t) − xeq  ≤ ε, ∀t ≥ t0 ≥ 0. If the solutions start close
to xeq it, they will remain close to it forever and ε can make arbitrarily small by
choosing δ sufficiently small.
This theorem established the global asymptotically stable if it is Lyapunov stable
and for every initial state the solution exist [0, ∞) s on and x(t) → xeq as t → ∞.

8 Global Stability for Endemic Equilibrium

In this part, we will prove the global stability of the endemic equilibrium using the
[25] Pinocare Bendixson principle. For proofing the Bendixson principle we can
assume some assumptions. Consider a function f (x) : ⊆ R n → R n , where is
an open set and its differential ddtx = f (x) determines the solution for every value of
x(t) with its initial conditions x(t) = × 0. Therefore the equilibrium points satisfy
some conditions:
238 Y. S. Rao

1. is simply connected.
2. This open set exists a compact absorbing sub-set K of .
3. This equilibrium point in the open set is globally stable when it satisfies the
− t
condition given by q2 = lim sup q < 0, where q = (Z (x(s, x0 )))ds, such
t→∞ x ∈k 0
0

that the Bandixson matrix Z = (C f C −1 + C J [2] C −1 ) and (Z ) ≤ 0.

[2]
Here represent Loziniskii Measure and J denotes the second additive
compound matrix J.

Theorem 4 The system of Eq. (1) is globally stable in a given region when R0 > 1.

Proof For finding the global stability at the unique endemic equilibrium points (S*,
I*, V*) of the system (1) leaving the Recovery class the jacobian matrix can be
represented as
⎛ ⎞
−(d1 + β I ) −β S γ
J =⎝ βI β S − (α + d1 + d2 ) ε ⎠
0 0 −(ε + γ + d1 + d2 )

Since J ∈ R3X 3 so its second additive matrix J [2] becomes

J [2] =
⎛ ⎞
β S − (α + 2d1 + d2 + β I ) ε −γ
⎜ ⎟
⎝ 0 −(2d1 + d2 + γ + ε + β I ) −β S ⎠
0 −β I β S − (2d1 + d2 + γ + ε + α)

For obtaining the Bandixson matrix Z we can represent a diagonal matrix C =

⎛ ⎞
1 0 0
C(S, I, V ) = ⎝ 0 VI 0 ⎠ = Diag(1, VI , VI )
0 0 VI

In the system the vector field f then

        
−1 I V I V
CfC = Diaog 0, ,
V f I V f I

where
     
I V I/ V/
= −
V f I I V
⎛ ⎞
0 0  0
⎜ I/ / ⎟
Then C f C −1 = ⎜ − ⎟
V
0 0
⎝ I V  ⎠.
I/ /
0 0 I
− V
V
Implementation of Vaccination in an Epidemic Model for COVID-19 239

So the Bandixson matrix is defined as

 
Z 11 Z 12
Z = (C f C −1 + C J [2] C −1 ) =
Z 21 Z 22
⎛ ⎞
β S − (α + 2d1 + d2 + β I ) ε VI γ VI
⎜ / / ⎟
Z =⎜
⎝ 0 β S − (α + d1 + d2 ) + II − VV −β S ⎟
⎠
/ /
0 −β I β S − (2d1 + d2 + γ + ε + α) + II − VV

where

Z 11 = (β S − (α + 2d1 + d2 + β I ))
 
Z 12 = ε VI γ VI
 
0
Z 21 =
0
 / /

β S − (α + d1 + d2 ) + II − VV −β S
Z 22 = / /
−β I β S − (2d1 + d2 + γ + ε + α) + II − VV

The matrix Z can be calculated as (Z ) ≤ sup{y1 , y2 }.

Again above two values are

V
y1 = (Z 11 ) + |Z 12 | = β S − (α + 2d1 + d2 + β I ) + γ
I
I/ V/
y2 = |Z 21 | + (Z 22 ) = β S − (2d1 + d2 + γ + ε + α) + −
I V
Hence

V I/
y1 = β S − (α + 2d1 + d2 + β I ) + γ ≤ − d1
I I
I/ V/ I/
y2 = β S − (2d1 + d2 + γ + ε + α) + − ≤ − d1
I V I
Thus.
I/
(Z ) ≤ sup{y1 , y2 } ≤ I
− d1 .
Therefore

t
(Z )dt ≤ ln I (t) − d1 t
0

Finally, we obtain as
240 Y. S. Rao

t
(Z )dt
− 0 ln I (t)
q2 = ≤ − d1 < 0.
t t
This satisfies the Benxdison principle. These results show the global stability of
the endemic equilibrium.

9 Simulation and Discussion of Results

This figure indicates the susceptible populations recovered after giving the vacci-
nation doses. The rate of susceptibility decreases and the recovered rate increases
simultaneously concerning time (Figs. 2 and 3).
It explains that susceptible populations get infectious with low immunity in the
body. The infected population increases due to not taking the vaccine and also having
other diseases.
Figure 4 explains the vaccination program during the initial stage of the epidemic
goes steadily. Initially, the infected population was very high during the 1st phase
of vaccination programs. It indicates the effective vaccination given to the people to
control the epidemic diseases (Fig. 5).
This figure shows the dynamic behavior of the population. It is clear that when the
reproduction number is less than unity, it is locally stable and shows the asymptotical
behavior of the population observed (Fig. 6).
The optimal control has been shown in the above figure. The more vaccine is less
infected concerning time is observed in the real parameter. It reveals the level of
infection the population reaches after a certain time the vaccination used as a control
measures (Fig. 7).

Fig. 2 Susceptible population versus recovered population in the CFE point

Implementation of Vaccination in an Epidemic Model for COVID-19 241

Fig. 3 Susceptible population versus infected population

Fig. 4 Infected population versus vaccinated population

This figure indicates at the endemic equilibrium point the vaccination program
has been considered at the early time the infected population would have been signif-
icantly lower and the outbreak of covid-19 would be extinguished. That leads to the
death rate would decrease (Fig. 8).
The recovery of the population is very high and the vaccination program should
implement during this period. When people get the vaccination and are free from
other diseases the infected population getting more recovery from COVID-19. This
242 Y. S. Rao

Fig. 5 Behaviour of the populations

Fig. 6 Infected population versus vaccination concerning the time

figure shows the recovered population increases when more people get the vaccine
in two doses (Tables 1 and 2).
Implementation of Vaccination in an Epidemic Model for COVID-19 243

Fig. 7 Susceptible population versus recovered population

Fig. 8 Vaccination population versus recovered populations concerning the time

Table 1 CPU time for

t n Times (Days)
different values t and n
0.1 1000 0.123
0.01 10,000 0.912
0.001 100,000 17.132

Table 2 CPU time for

t n Times (Days)
different values t and n
0.1 2000 0.923
0.1 20,000 16.612
0.1 200,000 27.132
0.1 20,000,000 34.234
244 Y. S. Rao

10 Limitations of the Model

There were some drawbacks to this investigation. Unreported cases were not taken
into account during the vaccination period. A booster dose is not mentioned in the
vaccination. The reality of today’s seniors does not show any evidence of disease
control. Due to some confounding variables, the amount of covid-19 vaccination
coverage cannot be calculated. The uncertainty could not be taken into account
when estimating the outcomes. inherent effects of vaccination. As a result, the model
creates the best current method for available doses and its consideration of controlling
infectious diseases.

11 Conclusion and Future Scope

Here explore the SVIR model with variable infection rates for investigating the action
of virus propagation. Based on this model a control parameter basic reproduction
number R0 that determines the local and global dynamic behaviour of the disease
propagation has been obtained by mathematical analysis. The virus may die out in the
population when basic reproduction is below unity and they will be prevalent other-
wise. Also using the Bendixson’s principle and Lyapunov function to obtain global
stability for infection-free and endemic equilibrium. Further numerical simulations
verify the correctness of the theoretical analysis. More importantly, investigate the
effect of infection rate after the vaccinated population gets prevalence the peoples.
When providing some booster dose or a powerful effective vaccine according to simu-
lating results. This analysis can provide some insight into covid-19 countermeasures.
In the real world, the result can help vaccine companies or related organizations to
make a powerful effective vaccine or booster dose countermeasure to work well. In
the future time delay, the existence of an invariant distribution and erotic property
problem will study for preventing the diseases.

References

1. Z. Zhao et al., Applications of robotics, AI, and digital technologies during COVID-19: a
review. Disaster Med. Public Health Prep. Figure 1, 1–11 (2021). https://doi.org/10.1017/dmp.
2021.9
2. H. Singh, D. Baleanu, J. Singh, H. Dutta, Computational study of fractional order smoking
model. Chaos, Solitons Fractals 142, 110868 (2021). https://doi.org/10.1016/j.chaos.2021.
110868
3. B.S.T. Alkahtani, Mathematical modeling of COVID-19 transmission using a fractional order
derivative 46(7) (2023)
4. A.K. Rauta, Y.S. Rao, J. Behera, B. Dihudi, T.C. Panda, SIQRS epidemic modelling and
stability analysis of COVID-19 35–50 (2021). https://doi.org/10.1007/978-981-33-4236-1_3
5. B. Kumar, A. Kumar, Y. Shankar, B. Kumar, COVID-19 created chaos across the globe: Three
novel quarantine epidemic models. Chaos, Solitons Fractals 138 (2020)
Implementation of Vaccination in an Epidemic Model for COVID-19 245

6. B. Dihudi, Y.S. Rao, S.K. Rout, T.C. Panda, Transmission modelling on COVID-19 pandemic
and its challenges, in Sustainability Measures for COVID-19 Pandemic (2021), pp. 75–90.
https://doi.org/10.1007/978-981-16-3227-3
7. J. Behera, A.K. Rauta, Y.S. Rao, S. Patnaik, Mathematical modeling on double quarantine
process in the spread and stability of COVID-19, in Sustainability Measures for COVID-19
Pandemic (2021), pp. 37–57. https://doi.org/10.1007/978-981-16-3227-3_3
8. A.K. Rauta, Y.S. Rao, J. Behera, Spread of COVID-19 in Odisha (India) due to influx of migrants
and stability analysis using mathematical modeling, in Artificial Intelligence for COVOD-19,
vol. 358 (2021), pp. 295–309. https://doi.org/10.1007/978-3-030-69744-0_17
9. B.K. Mishra, A.K. Keshri, D.K. Saini, S. Ayesha, B.K. Mishra, Y.S. Rao, Mathematical model,
forecast and analysis on the spread of COVID-19. Chaos, Solitons Fractals 147, 110995 (2021)
10. J. Jiang, L. Jiang, G. Li, J. Luo, M. Wang, H. Xu, Prediction of the epidemic trends of COVID-
19 by the improved dynamic SEIR model, in Research Square (2020). http://www.epistemon
ikos.org/documents/f25211f313a49c7dce0e7ceb3877c9f503136066
11. X. Wang, L. Shi, Y. Zhang, H. Chen, G. Sun, Policy disparities in fighting COVID-19. Int. J.
Equity Health 20(33), 1–11 (2021)
12. H. Singh, Analysis for fractional dynamics of Ebola virus model, in Chaos, Solitons and
Fractals, vol. 138 (2020)
13. D.B. Harendra Singh, H. Srivastava, Methods of Mathematical Modelling Infectious Diseases,
1st edn. (Elsevier Sciences, 2022). https://www.elsevier.com/books/methods-of-mathematical-
modelling/singh/978-0-323-99888-8
14. M. Makhoul et al., Epidemiological impact of sars-cov-2 vaccination: mathematical modeling
analyses. Vaccines 8(4), 1–16 (2020). https://doi.org/10.3390/vaccines8040668
15. H.E. Randolph, L.B. Barreiro, Herd immunity: Understanding Covid-19 by Haley etal.
Immunity 52, 737–741 (2020)
16. M.G. Thompson et al., Interim estimates of vaccine effectiveness of BNT162b2 and mRNA-
1273 COVID-19 vaccines in preventing SARS-CoV-2 infection among health care personnel,
first responders, and other essential and frontline workers — Eight U.S. Locations, December
2020–March. MMWR Surveill. Summ. 70(13), 495–500 (2021). https://doi.org/10.15585/
mmwr.mm7013e3
17. D. Aldila, B. M. Samiadji, G. M. Simorangkir, S. H. A. Khosnaw, and M. Shahzad, “Impact
of early detection and vaccination strategy in COVID-19 eradication program in Jakarta,
Indonesia,” BMC Res. Notes, vol. 14, no. 1, pp. 1–7, 2021, doi: https://doi.org/10.1186/s13
104-021-05540-9
18. Y. Choi, J.S. Kim, J.E. Kim, H. Choi, C.H. Lee, Vaccination prioritization strategies for covid-
19 in korea: a mathematical modeling approach. Int. J. Environ. Res. Public Health 18(8), 1–19
(2021). https://doi.org/10.3390/ijerph18084240
19. M. Elhia, K. Chokri, M. Alkama, Optimal control and free optimal time problem for a covid-
19 model with saturated vaccination function. Commun. Math. Biol. Neurosci. 2021 (2021).
https://doi.org/10.28919/cmbn/5632
20. M. Kamraju, M.A. Ali, Mohd, A. Ali, A study on Covid-19 vaccination drive in India. BRICS
J. Educ. Res. 11(2), 76–79 (2021). https://www.researchgate.net/publication/352903714
21. O.J. Watson, G. Barnsley, J. Toor, A.B. Hogan, P. Winskill, A.C. Ghani, Global impact of the
first year of COVID-19 vaccination: a mathematical modelling study. Lancet Infect. Dis. 22(9),
1293–1302 (2022). https://doi.org/10.1016/S1473-3099(22)00320-6
22. J. Lopez Bernal et al., Effectiveness of Covid-19 vaccines against the B.1.617.2 (Delta) variant.
N. Engl. J. Med. 385(7), 585–594 (2021). https://doi.org/10.1056/nejmoa2108891
23. J.P. Lasalle, The stability of dynamical systems 25(3) (1976). https://doi.org/10.1137/1021079
24. M.Y. Li, J.R. Graef, L. Wang, J. Karsai, Global dynamics of a SEIR model with varying
total population size. Math. Biosci. 160(2), 191–213 (1999). https://doi.org/10.1016/S0025-
5564(99)00030-9
25. Y. Li, J.S. Muldowney, On bendixson s criterion. J. Differ. Equ. 106(1), 27–39 (1993). https://
doi.org/10.1006/jdeq.1993.1097