Results in Physics 24 (2021) 104124

Contents lists available at ScienceDirect

Results in Physics
journal homepage: www.elsevier.com/locate/rinp

Fractional stochastic sır model

Badr Saad T. Alkahtani a, Ilknur Koca b, *
a
Department of Mathematics, Riyadh, 11989, College of Science, King Saud University, P.O. Box 1142, Saudi Arabia
b
Department of Accounting and Financial Management, Seydikemer High School of Applied Sciences, Mugla Sıtkı Kocman University, Mugla, Turkey

A R T I C L E I N F O A B S T R A C T

Keywords: Stochastic and fractional differentiation have been developed independently to depicting processes following
Caputo randomness and power, a declining memory and passage from one process to another respectively. Very recently,
Atangana-Baleanu fractional stochastic differential equations were suggested with the aim to capture processes following at the
Caputo-Fabrizio differential operators
same time randomness and memory nonlocality. In this paper to further explore the applicability of this type of
Global derivative
Existence and uniqueness
differential equations, a SIR model was considered and analyzed analytically and numerically. Some numerical
Numerical approximations simulations are presented for different values of fractional orders and densities of randomness.

Introduction collected accurately, then clearly the mathematical models should be

questioned. Indeed the spread of infectious diseases among humans is a
In the last decades few mathematical tools and concepts have been serious worry and have reduced human population through death in the
utilized to predict real world problems, for example classical differential last past years. We can list few including Spanish flu that left many
and integral operators, fractional differential and integrals operators millions of death, HIV, syphilis, Ebola, malaria, dengue fever, Lassa
and stochastic-differential equations. They have been used for many fever, and covid-19 and its variants [19–25]. One main way to control
different purposes. For example classical differentiation and integration their spread is perhaps to suggest a mathematical model that will be able
have been used for modelling classical mechanical problems were to accurately replicate the spread such that its future can be predicted. Is
memory is not captured. Indeed with these two mathematical operators the predictions are accurate, one could take decisions on how to control
many real world have been modelled with some limitations. Stochastic the spread. Very recently, Atangana and Seda suggested the use of
differential equations, have been used to modelling complex real world fractional stochastic differential and integral equations [18]. This
problems following randomness. Fractional calculus was introduced as approach is very effective for modelling spread of infectious disease that
extension of classical derivative, they have been introduced to replicate follows processes with non-locality and randomness. In this paper, we
complex problems following power law processes, a declining memory analyse a mathematical model display the spread of some infectious
and a passage from one process to another steady and non-steady states. diseases with fractional stochastic approach.
In particular, these concepts have been applied to model epidemiolog­
ical problems. The spread of many infectious diseases have been Fractional differentiation
modelled and simulated, in some case the mathematical models have
been compared with experimental data [1–4,9–11]. In some situation, In this section, we remember some important definitions for frac­
the mathematical models were in good agreement with experimental tional derivative with local and non-local kernels [5–8,13].
data, clear indication that mathematical models were able to replicate
Definition 1. Caputo fractional derivative of order α > 0 of a function
accurately. However in many situation, modellers observed a clear
f : (0, ∞)→R, according to Caputo, the fractional derivative of a
disagreement between the observed data and the mathematical equa­
continuous and differentiable function f is given as:
tions. In these cases two questions could be asked the first question is to
know if the collected data were done accurately, and the second one will () ∫ t ( )
1 d
be is the mathematical model suitable to model this situation? Now if the C
Dαt f t = (t − x)− α
f x dx, 0 < α⩽1.1 (1)
Γ(1 − α) dx
answer of the first question is yes, meaning the experimental data were 0

* Corresponding author.
E-mail addresses: balqahtani1@ksu.edu.sa (B.S.T. Alkahtani), ilknurkoca@mu.edu.tr (I. Koca).

https://doi.org/10.1016/j.rinp.2021.104124
Received 20 February 2021; Received in revised form 21 March 2021; Accepted 22 March 2021
Available online 5 April 2021
2211-3797/© 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
B.S.T. Alkahtani and I. Koca Results in Physics 24 (2021) 104124

() () ∫t ( )
1− α α
f t = u t + u y (t − y)α− 1 dy.12 (12)
Definition 2. The Riemann–Liouville fractional integral of order α > 0 B(α) B(α)Γ(α) a
of a function f : (0, ∞)→R, according to Riemann–Liouville, the frac­
tional integral that is considered as anti-fractional derivative of a func­
tion f is: Definition 7. Let f(t) be continuous, g(t) be a non-constant increasing
() ∫t ( ) positive function. And also taking K(t) as kernel with singular or non-
1 singular versions. For 0 < α⩽1, a fractional global derivative of
I αt f t = (t − x)α− 1 f x dx, x > a.2 (2)
Γ(α) 0 Caputo sense is defined by
()
C α
0 Dg f (t) = Dg f (t)*K t 13 (13)
Definition 3. Let f ∈ H1 (a, b), b > a, 0 < α < 1 then, the new Caputo
derivative of fractional derivative is defined as: Also with Riemann–Liouville version, we have

1
∫ t ( ) [
(t − x)
] RL α
0 Dg f (t) = Dg (f (t)*K(t) ), 14 (14)
(3)
CF α ′
a Dt f (t) = f x exp − α dx.3
1− α 1− α
where * means the convolution operator.
a

and also if the function does not belongs to H1 (a, b) then, the derivative We have to note that definitions above given with global idea are
can be reformulated as obtained very early by Atangana. You can see detailed analysis for
∫ t( ( ) ( )) [ ] global derivative in his paper referenced with [5]. Let us see some
CF α
D f (t) =
α
f t − f x exp − α
(t − x)
dx.4 (4) versions of its below.
Global derivative with Caputo version is given by
a t
1− α a 1− α
() ( )
Theorem 1:Let 0 < α < 1 then the following time fractional ordi­ 1
∫t
nary differential equation
C α
D
0 g f t = Dg f x (t − x)− α dx, 0 < α⩽1.15 (15)
Γ(1 − α) 0
( )
CF α
0 Dt f (t) = u t , 5 (5) Global derivative with Riemann–Liouville version is given by
() ∫t ( )
has a unique solution with taking the inverse Laplace transform and RL α 1
f x (t − x)− α dx, 0 < α⩽1.16 (16)
0 Dg f t = Dg
using the convolution theorem below: Γ(1 − α) 0

()
2(1 − α)
()
2α
∫t () Global derivative with Caputo-Fabrizio version is given by
f t = u t + u s ds, t⩾0.6 (6) ∫t ( ) [ ]
(2 − α) (2 − α) 0 1 (t − x)
CF α
0 Dg f (t) = Dg f x exp − α dx.17 (17)
1− α 0 1− α

Definition 4. Let f ∈ H1 (a, b), b > a, α ∈ (0, 1) then, the definition of And finally global derivative with Atangana-Baleanu versions are
the new fractional derivative is given as: given by
∫ ( ) [ ]
( ) [ ] B(α) t (t − x)α
B(α)
∫ t
(t − x)α
ABC α
D f (t) = D g f x E α − α dx, 18 (18)
(7)
0 t
ABC α
Dt f (t) = f
′
x Eα − α dx, 7 1− α 0 1− α
a
1− α a 1− α
∫ ( ) [ ]
B(α) t
(t − x)α
where ABC
ABR α
(19)
a Dt is fractional operator with Mittag–Leffler kernel in the
α 0 Dg f (t) = Dg f x Eα − α dx.19
1− α a 1− α
Caputo sense with order αwith respect to t and
( ) In numerical proofs integral versions of those derivatives below are
α
B α = 1− α+ ,8 (8) used, so integral operators with global derivative with Riemann–Liou­
Γ(α)
ville version is given by
is a normalization function. () ∫t ( ) ( )
1
g x f x (t − x)α− 1 dx.20 (20)
α ′
0Ig f t =
Definition 5. Let f ∈ H1 (a, b), b > a, α ∈ (0, 1) and not differentiable Γ(α) 0
then, the definition of the new fractional derivative is given as:
Caputo-Fabrizio version is given by
∫ ( ) [ ] ( ) ( ) ∫t ( ) ( )
B(α) d t
(t − x)α 1− α ′ α
(9) (21)
ABR α CF α ′
a Dt f (t) = f x Eα − α dx.9 0 Ig f (t) = g x f x + g x f x dx.21
1 − α dt a 1− α M(α) M(α) 0
Atangana-Baleanu version is given by
() () () ∫t ( ) ( )
Definition 6. The fractional integral of order α ∈ (0, 1) of a new 1− α α
g x f x (t − x)α− 1 dx.22
AB α ′ ′
I f t = f t g t +
fractional derivative is defined as: 0 g
B(α) B(α)Γ(α) a
() () ∫t ( ) (22)
1− α α
AB α
a It f t = f t + f y (t − y)α− 1 dy.10 (10)
B(α) B(α)Γ(α) a Model derivation of stochastic SIR epidemic model
When αis zero, initial function is obtained and when αis 1, the or­
Although deterministic differential equations have been used inten­
dinary integral is obtained.
sively to replicate the spread of some infectious diseases, nevertheless, a
The following time fractional ordinary differential equation
( ) day by day collection of data showed that, their spread sometime follow
ABC α
0 Dt f (t) = u t , 11 (11) non-locality and randomness. A clear indication that neither fractional
differential equation nor stochastic differential equations cannot be used
has a unique solution with taking the inverse Laplace transform and to replicate such spread. Indeed if the spread follows randomness then
using the convolution theorem below: stochastic differential equations are suitable for modelling such prob

2
B.S.T. Alkahtani and I. Koca Results in Physics 24 (2021) 104124

lems. Some published papers involving the use of stochastic differential multiplier first. The multiplier is calculated as
equations can be found in [1–4]. In general the differentiation with ∫
randomness is given as below: λ = e μdt = eμt .30 (30)

dx = f1 (t, x, y)dt + f2 (t, x, y)dw, 23 (23) If we do necessary calculations then we have

() Λ
here wi = [w1 , …, wn ], for i = 1, …, n is an independent Wiener process. n t = + ce− μt .31 (31)
μ
In the SIR model, s(t) is susceptible, i (t) is infective and r(t) is removed
individuals in the population and given below If we consider n(0) = 0, we will have
() Λ
.s(t) = Λ − μs(t) − βs(t)i (t), (24)
n t = (1 − e− μt )32 (32)
μ
()
.i t = βs(t)i (t) − (μ + γ + ∊)i (t), For ∀t⩾0, we have that
.r(t) = γi (t) − μr(t). () Λ
n t ⩽ .33 (33)
The above Eq. (24) is the transform into stochastic, by adding μ
invironment of noise.
⎧ (s(t), i (t), r(t)) ∈ R3 : s( t)⩾0, i ( t)⩾0, r( t)⩾0, ⎫
ds(t) = [Λ − μs(t) − βs(t)i (t)]dt − w1 s(t)dB1 (t), ⎨ + ⎬
di (t) = [βs(t)i (t) − (μ + γ + ∊)i (t)]dt + w2 i (t)dB2 (t), γ= () () () () Λ ,
⎩ n t =s t +i t +r t ⩽ ⎭
dr(t) = [γi (t) − μr(t)]dt + w3 r(t)dB3 (t) μ

where (wi )i=1,2,3 are densities of randomness and Bi (t)i=1,2,3 are inviron­ region is positive invaryant for the system. Also for t→∞
ment of noise. ()
All parameters above are positive constants. Also Λis the birth rate, Λ
t→∞
limsupn t ⩽ .34 (34)
μis the death rate, βis the average number of contacts per infective per μ
day, γis the recovery rate and ∊is the death rate of infectives caused by So all the solutions of system is uniformly bounded in γ.
disease.
1. Equilibrium points
0.1. Equilibrium point and stability analysis SIR epidemic model
In this section, we derive the equilibrium points for both disease-free
( )
In this subsection we can analyze the dynamics of the stochastic and endemic The disease-free equilibrium is given as E0 = μ , 0, 0 , for
Λ

models with the help of the stability analysis of the deterministic i = 0. The endemic equilibrium is obtained by solving the following
equations. In fact, the solution to the deterministic model corresponds to system.
the mean of the stochastic model. İn this subsection, firstly we study on
deterministic SIR model to obtain equilibrium points and also we Λ − μs* − βs* i * = 0, (35)
analyze sufficient conditions under which the equilibrium points are
locally stable. Let us consider model below: βs* i * − (μ + γ + ∊)i * = 0,
() ()() γi * − μr* = 0.
ds(t)
= Λ − μs t − βs t i t , (25) Then we have
dt
(μ + γ + ∊) * βΛ − μ(μ + γ + ∊) * γ(βΛ − μ(μ + γ + ∊))
di (t) s* = ,i = ,r = .36
= βs(t)i (t) − (μ + γ + ∊)i (t), β (μ + γ + ∊) − μ(μ + γ + ∊)
dt (36)
dr(t)
= γi (t) − μr(t). So endemic equilibrium point
dt
( ) ((μ + γ + ∊) βΛ − μ(μ + γ + ∊) βΛ − μ(μ + γ + ∊))
Here if n(t) show total population size in dynamic, we calculate n(t) with E* s* , i * , r* = , , .
β (μ + γ + ∊) (μ + γ + ∊)
following calculation.
dn(t) ds(t) di (t) dr(t)
= + + .26 (26) 1.0.1. Local and global stability of the endemic equilibrium
dt dt dt dt
So we have First we consider Sır model as below
() () () () ( )
dn(t) ds(t)
= Λ − μs t − μi t − ∊ + t − μ r t , (27) = f1 s, i , r , (37)
dt dt

⩽ Λ − μs(t) − μi (t) − μr(t), di (t)

= f2 (s, i , r),
⩽ Λ − μn(t). dt
dr(t)
Then we will write = f3 (s, i , r).
() dt
dn(t)
+ μn t ⩽Λ.28 (28) Here
dt
Let us calculate inequality above with considering equality as f1 (s, i , r) = Λ − μs(t) − βs(t)i (t), (38)
()
dn(t) f2 (s, i , r) = βs(t)i (t) − (μ + γ + ∊)i (t),
+ μn t = Λ.29 (29)
dt f3 (s, i , r) = γi (t) − μr(t).
It’s a linear differential equation above so we have to find Lagrange We compute first the Jacobian matrix of the SIR model for endemic

3
B.S.T. Alkahtani and I. Koca Results in Physics 24 (2021) 104124

equilibrium
.L(t) < 0ifR0 < 1,
⎛ ⎞ ⎡ ⎤
− μ − βi * − βs* 0 .L(t) > 0ifR0 > 1.
J ⎝E* ⎠ = ⎣ βi * βs* − (μ + γ + ∊) 0 ⎦39 (39)
Theorem: If R0 ⩾1, the endemic point E* (s* , i * , r* ) is globally
0 γ − μ
asymptotically stable.
We now construct a characteristic equation associated to SIR model Proof:We prove this using the idea of a fractional Lyapunov function.
We start by defining the Lyapunov function associated the system as
k = det[J(E* ) − λI] = 0.40 (40)
below:
From the above, we obtain the following characteristic polynomial ( ) ( ( ) (
s* ) i*
( ) L s* , i * , r* = s − s* + s* log + i − i * + i * log + r − r*
k λ = λ3 + k1 λ2 + k2 λ + k3 , 41 (41) s i
r* )
where + r* log , 50 (50)
r
k1 = 3μ + γ + ∊ + βi * − βs* , (42) (s − s* ) ( )
i − i* (r − r* )
.L(t) = .s + .i + .r, (51)
( ) s i r
k2 = μ2 − 2μβs* + μβi * + μ + γ + ∊ (2μ + βi * ),
2 *
) 2 *
k3 = (μ + γ + ∊)(μ + μβi − μ βs . (s − s* )
= (Λ − μs(t) − βs(t)i (t))
Theorem: At the endemic equilibrium point E* (s* , i * , r* ) is locally s
( *)
asymptotic stable if all the eigenvalues λ1 , λ2 and λ3 satisfy the following i− i
condition
+ (βs(t)i (t) − (μ + γ + ∊)i (t))
i
απ (r − r * )
|argλi |〉 fori = 1, 2, 3.43 (43) + (γi (t) − μr(t)).
2 r
Proof: If all the coefficients in the characteristic equation are posi­ Then we write
tive (λi > 0 for i = 1, 2, 3) then the system is asymptotically stable. The
s* ( )
matrix H is Hurwitz matrix, that is, all the real parts of the eigen values .L(t) = Λ + μs* + βsi + βsi + μ + γ + ∊ i * + μr* + γi (52)
are in the left half plane so the above condition can be satisfied by s
Routh-Hurwitz criteria [14–17].The Hurwitz matrix for the character­ ( ( )
istic polynomial k(λ) is written as s* γi r*
− Λ + μs + βsi + μ + γ + ∊ i + βsi * + μr +
⎡ ⎤ s r
k1 k3 0 = Π1 − Π2 ,
H3 = 1 k2 0 ⎦44
⎣ (44)
0 k1 k3
where
Then we have ( )
s*
Π 1 = Λ + μs* + βsi + βsi + μ + γ + ∊ i * + μr* + γi , (53)
H1 = k1 > 0, H2 = k1 k2 − k3 > 0.45 (45) s

So the proof is completed. ( )

s* γi r*
Π2 = Λ + μs + βsi + μ + γ + ∊ i + βsi * + μr + .
s r
1.0.2. Global stability of the endemic equilibrium point Therefore if
Π 1 − Π2 > 0then.L(t) > 0, (54)
We present the global stability of the system

.s(t) = Λ − μs(t) − βs(t)i (t), (46) Π1 − Π2 = 0then.L(t) = 0,

Π1 − Π2 < 0then.L(t) < 0.
()
.i t = βs(t)i (t) − (μ + γ + ∊)i (t),
2. Analysis of SIR stochastic model
.r(t) = γi (t) − μr(t).
We define in the classical case In this section, we consider a general SIR stochastic model where the
classical time derivative is convertal to global derivative. Noting a global
1 βs(t)i (t) − l1 i (t)
.L = .i = , (47) derivative of a differentiable function f with respect to an increasing
l1 l1
non-negativecontinuous function g is defined
( ) ( ) t→t
(βs(t) − l1 )i (t) 1 f (t) − f (t )
1
= , l1 = μ+γ+∊ . Dg f t = lim .
l1 g(t) − g(t1 )

Then we write Indeed if g is differentiable then

( )()
βs(t) ()
(48)
′
.L(t) = − 1 i t , f (t)
l1 Dg f t = ′ .
g (t)
()
βs* Let us consider following stochastic SIR model,
= (R0 − 1)i t , R0 = .
l1 ds(t) = [Λ − μs(t) − βs(t)i (t)]dt − w1 s(t)dB1 (t), (55)
Therefore, for all three cases
di (t) = [βs(t)i (t) − (μ + γ + ∊)i (t)]dt + w2 i (t)dB2 (t),
.L(t) = 0ifR0 = 1, (49) dr(t) = [γi (t) − μr(t)]dt + w3 r(t)dB3 (t)
With this model above we can easily determine the extinction and

4
B.S.T. Alkahtani and I. Koca Results in Physics 24 (2021) 104124

persistence of the disease. s(t) and i (t) and r(t) are positive solutions for and also
t⩾0 of system above for any given initial value (s(0),i (0),r(0)) ∈ R3+ . Also
() () () Λ |g1 (t, i ) − g1 (t, i 1 )|2 ⩽ α2 |i − i 1 |2 ,
s t + i t + r t ⩽ , 56 (56) |g2 (t, i ) − g2 (t, i 1 )|2 ⩽ α2 |i − i 1 |2 ,
μ
|h1 (t, r) − h1 (t, r1 )|2 ⩽α3 |r − r1 |2 .
for t⩾0.
But now we consider system of stochastic differential equations with ii)
global derivative first ( )
( ) [ ( ) ( ) ( )] ( )( ) |f1 (t, s)|2 ⩽ β1 1 + |s|2 ,
( )
Dg s t = Λ − μs t − βs t i t − σs t i t , (57) |f2 (t, s)|2 ⩽ β1 1 + |s|2 ,
( 2)
( ) |g1 (t, i )|2 ⩽ β2 1 + |i | ,
( )
Dg i ( t) = [βs(t)i (t) − (μ + γ + ∊)i (t)] + σs(t)i (t), |g2 (t, i )|2 ⩽ β2 1 + |i |2 ,
Dg r t = [γi (t) − μr(t)],
also
s(0) = s0 , i (0) = i 0 and r(t) = r0 . ( )
Since g is differentiable, then we can write |h1 (t, r)|2 ⩽β3 1 + |r|2 .
′ ′
ds(t) = [Λ − μs(t) − βs(t)i (t)]g (t)dt − w1 s(t)g (t)dB1 (t), (58) Proof of Lipschitz condition for equations of SIR model: In this
part we reconsider stochastic model as below
′ ′
di (t) = [βs(t)i (t) − (μ + γ + ∊)i (t)]g (t)dt + w2 i (t)g (t)dB2 (t), ds(t) = f1 (t, s(t))dt + f2 (t, s(t))dB(t), (62)
′ ′
dr(t) = [γi (t) − μr(t)]g (t)dt + w3 r(t)g (t)dB3 (t)
Its worth noting that if the environmental noiser (wi )i=1,2,3 = 0 then di (t) = g1 (t, i (t))dt + g2 (t, i (t))dB(t),
dr(t) = h1 (t, r(t))dt + h2 (t, r(t))dB(t).
the model is simple shetor ministic.
Applying the integral on both sides, we have
Here
() ( ) ∫t ( )
(59) (63)
′
s t =s 0 + g τ (Λ − μs(τ) − βs(τ)i (τ))dτ f1 (t, s(t)) = Λ − μs(t) − βs(t)i (t),
0

∫ t ( ) f2 (t, s(t)) = − σs(t)i (t),

−
′
g τ (σs(τ)i (τ))dτ, g1 (t, i (t)) = βs(t)i (t) − (μ + γ + ∊)i (t),
0 g2 (t, i (t)) = σs(t)i (t),
∫t ( ) h1 (t, r(t)) = γi (t) − μr(t).
( ) ′
i (t) = i 0 + g τ (βs(τ)i (τ) − (μ + γ + ∊)i (τ))dτ Now we define the following norm
0
∫t ( )
t∈[0,T]
‖ϑ‖∞ = sup |ϑ|2 ,
′
+ g τ (σ s(τ)i (τ))dτ,
0

() ( ) ∫t ( ) ∫t ( ) ( ) ( ) then we have ∀s, s1 ∈ R2 and t ∈ [0, T]

′ ′
r t =r 0 + g τ (γi (τ) − μr(τ))dτ + w3 g τ r τ dB3 τ .
0 0 |f1 (t, s) − f1 (t, s1 )|2 = |( − μ − βi (t))(s − s1 ) |2 , (64)
{ 2 }
Now with the idea of Brownian motion, we have ⩽ 2μ + 2β2 |i (t)|2 |s − s1 |2 ,
() ( ) ∫t ( ) { t∈[0,T]
}
s t =s 0 +
′
g τ (Λ − μs(τ) − βs(τ)i (τ))dτ (60) ⩽ 2μ2 + 2β2 sup |i (t)|2 |s − s1 |2 ,
0 { 2 }
⩽ 2μ + 2β2 ‖i (t)‖2∞ |s − s1 |2 , ⩽α1 |s − s1 |2 .
∫ t ( )
′
−
0
g τ (σs(τ)i (τ))dB(τ), where α1 = {2μ2 + 2β2 ‖i (t)‖2∞ }. And

( )
∫t ( ) |f2 (t, s) − f2 (t, s1 )|2 = |( − σi (t))(s − s1 ) |2 ,
i (t) = i 0 +
′
g τ (βs(τ)i (τ) − (μ + γ + ∊)i (τ))dτ { 2 }
⩽ σ |i (t)|2 |s − s1 |2 ,
0
∫t ( ) { t∈[0,T] }
+
′
g τ (σ s(τ)i (τ))dB(τ), ⩽ σ2 sup |i (t)|2 |s − s1 |2 ,
0 { 2 }
⩽ σ ‖i (t)‖2∞ |s − s1 |2 , ⩽α1 |s − s1 |2 ,
() ( ) ∫t ( ) ∫t ( ) ( ) ( )
′ ′ { }
r t =r 0 + g τ (γi (τ) − μr(τ))dτ + w3 g τ r τ dB3 τ .
0 0 where α1 = σ2 ‖i (t)‖2∞ .Also we show that ∀i , i 1 ∈ R2 and t ∈ [0, T]
then
Thus, with classical global derivative we have the following
nonlinear stochastic equation. Let us present now the condition under |g1 (t, i ) − g1 (t, i 1 )|2 = |(βs(t) − (μ + γ + ∊))(i − i 1 ) |2 , (65)
which the nonlinear case has unique solution which is taking advantage
{ }
of Atangana’s paper referenced by [5,6]. ⩽ 2β2 |s(t)|2 + 2(μ + γ + ∊)2 |i − i 1 |2 ,
Theorem: Assume that there are four positive constants α1 { t∈[0,T] }
⩽ 2β2 sup |s(t)|2 + 2(μ + γ + ∊)2 |i − i 1 |2 ,
,α1 , α2 , α2 , α3 and β1 , β1 , β2 , β2 , β3 such that. { 2 }
i) ⩽ 2β ‖s(t)‖2∞ +2(μ + γ + ∊)2 |i − i 1 |2 , ⩽α2 |i − i 1 |2 ,

|f1 (t, s) − f1 (t, s1 )|2 ⩽α1 |s − s1 |2 , (61) { }

where α2 = 2β2 ‖s(t)‖2∞ +2(μ + γ + ∊)2 .
|f2 (t, s) − f2 (t, s1 )|2 ⩽α1 |s − s1 |2

5
B.S.T. Alkahtani and I. Koca Results in Physics 24 (2021) 104124

|g2 (t, i ) − g2 (t, i 1 )|2 = |(σ s(t))(i − i 1 ) |2 ,

{ 2 }
⩽ σ |s(t)|2 |i − i 1 |2 , ⩽ 2γ2 |i (t)|2 +2μ2 |r(t)|2 ,
{ t∈[0,T] }
⩽ σ2 sup |s(t)|2 |i − i 1 |2 , t∈[0,T]
{ 2 } ⩽ 2γ2 sup |i (t)|2 +2μ2 |r(t)|2 ,
⩽ σ ‖s(t)‖2∞ |i − i 1 |2 , ⩽α2 |i − i 1 |2 , ( )
μ2 ( )
{ } ⩽ 2γ2 ‖i (t)‖2∞ +2μ2 |r(t)|2 ,⩽2γ2 ‖i (t)‖2∞ 1+ |r(t)|2 ,⩽β3 1+|r(t)|2 ,
where α2 = σ 2 ‖s(t)‖2∞ . γ 2 ‖i (t)‖2∞
Finally we show that ∀r, r1 ∈ R2 and t ∈ [0, T] then ( )〈
μ2
|h1 (t, r) − h1 (t, r1 )| ⩽ 2
| − μ(r − r1 ) | , 2
such that γ2 ‖i (t)‖2∞
1. Both two condition are verified. So according
2
= μ2 |r − r}
1| ,
⩽ {μ + ∊ |r − r1 |2 ,
2 to the above theorem, the si r system has a unique solution.
⩽ α3 |r − r1 |2 ,
Extinction
{ }
where α3 = μ2 + ∊ .
So condition (i) is satisfied. Now we will verify the second condition In this section, we present the extinction of speces classes. To do this,
below: we defined
Proof of linear growth condition for equations of SIR mod­ t→∞ 1
∫t ( )
el:∀(t, s) ∈ R2 ×[t0 , T] then we will show that 〈x(t)〉 = lim x τ dτ.71 (71)
t 0
|f1 (t, s)|2 = |Λ − (μ + βi (t))s(t)|2 , (66) We start with the class s(t). Applying the integral on both sides of .s(t)
yields
⩽ 2|Λ|2 + (μ + βi (t))2 |s(t)|2 , () ( ) ∫ t ∫ ( ) ( )
w1 t
( ( t∈[0,T]
) ) s t − s 0 = [Λ − μs(t) − βs(t)i (t)]d τ − s τ dB1 τ 72 (72)
⩽ 2|Λ|2 1 + 2 μ2 + β2 sup |i (t)|2 |s(t)|2 , 0 t 0
( ( ) ) Dividing on both sides by t gives
2 2μ2 2β2 ‖i (t)‖2∞ 2 ( )
∫ ∫ ( ) ∫t ( )( ) ∫ ( ) ( )
⩽ 2|Λ| 1+ 2
+ 2
|s(t)| , ⩽β1 1 + |s(t)|2 , s(t)− s(0) 1 t μ t β w1 t
|Λ| |Λ| = Λdτ − s τ dτ − s τ i τ dτ − s τ dB1 τ
t t 0 t 0 0 t t 0
( )〈
(73)
2μ2 2β2 ‖i (t)‖2∞
under condition |Λ|2
+ |Λ|2
1. ( ) ( )
∫ t
s(t) − s(0) w1
Also = Λ − μ〈s(t)〉 − β〈s(t)i (t)〉 − s τ dB1 τ .
t t 0

|f2 (t, s)|2 = | − σs(t)i (t)|2 , (67) {

t→∞ Λ t→∞ 1 s(t) − s(0)
lim〈s(t)〉 = − lim − β〈s(t)i (t)〉 +
( ) μ μt t
⩽ σ2 |i (t)|2 |s(t)|2 , ∫ t ( ) ( )}
( t∈[0,T] )( w1
) − s τ dB1 τ 74 (74)
⩽ σ2 sup |i (t)|2 1 + |s(t)|2 , t 0
( 2 )( ) ( )
⩽ σ ‖i (t)‖2∞ 1 + |s(t)|2 , ⩽β1 1 + |s(t)|2 , Then we have
() t→∞ Λ
where β1 = σ2 ‖i (t)‖2∞ . lim〈s(t)〉 = .75 (75)
μ
|g1 (t, i )|2 = |βs(t)i (t) − (μ + γ + ∊)i (t)|2 , (68) With the class i (t), we have
() ( ) ∫t ( )( ) ( )∫ t ( )
( )
⩽ 2β2 |s(t)|2 |i (t)|2 + 2(μ + γ + ∊)2 |i (t)|2 , i t − i 0 =β s τ i τ dτ − μ + γ + ∊ i τ dτ
( 2 ) 0 0
(76)
⩽ 2β |s(t)|2 + 2(μ + γ + ∊)2 |i (t)|2 , ∫t ( ) ( )
( t∈[0,T] )( ) + w2 i τ dB2 τ , 76
⩽ 2 β2 sup |s(t)|2 + (μ + γ + ∊)2 1 + |i (t)|2 , 0
( )( ) ( )
⩽ 2 β2 ‖s(t)‖2∞ +(μ + γ + ∊)2 1 + |i (t)|2 , ⩽β2 1 + |i (t)|2 , Then
( ) ∫ t ( ) ( )
( ) i (t) − i (0) w2
where β2 = β 2
‖s(t)‖2∞ +(μ + γ + ∊) . 2
t
= β〈s(t)i (t)〉 − μ + γ + ∊ 〈i (t)〉 +
t 0
i τ dB2 τ , 77

(77)
2
|g2 (t, i )| = |σ s(t)i (t)| , 2
(69)
⎧ ⎫
( ) ⎪ 〈s(t)i (t)〉 ⎪
⎨ ⎬
⩽ σ |s(t)| |i (t)|2 ,
2 2 t→∞
lim〈i(t)〉 = lim
t→∞β ∫t ( ) ( ) 78
( t∈[0,T] ) (μ + γ + ∊) ⎪
t→∞ w
2 i (0) − i (t) ⎪
( ) ⎩ +lim i τ dB2 τ + ⎭
⩽ σ 2 sup|s(t)|2 1 + |i (t)|2 , t 0 t
(78)
( )( ) ( )
⩽ σ2 ‖s(t)‖2∞ 1 + |i (t)|2 , ⩽β2 1 + |i (t)|2 ,
So we have
( )
σ2 ‖s(t)‖2∞
t→∞
where β2 = and finally lim〈i(t)〉 = 0.79 (79)
2
|h1 (t, r)| = |γi (t) − μr(t)| , 2
(70) For class r(t), we have

6
B.S.T. Alkahtani and I. Koca Results in Physics 24 (2021) 104124

∫ ( ) ( ) { ( ) ( ) }
r(t) − r(0) w3 t
= γ〈i (t)〉 − μ〈r(t)〉 + r τ dB3 τ 80 (80) +
3 1
Φ2 tn , xn Δt − Φ2 tn− 1 , xn− 1 Δt .
t t 0 2 2
Finally we have Replacing Φ1 (t, x) and Φ2 (t, x) by their values, we have the following
t→∞ scheme
lim 〈r(t)〉 = 0.81 (81) ( ) ( ) ∫ tn+1 ( ) ( ( ))
′
x tn+1 − x tn = g τ f1 τ, x τ dτ
Atangana Toufik scheme for stochastic equations with global tn
∫ tn+1 ( ) ( ( )) (93)
derivative +
′ ′
g τ f2 τ, x τ B (τ)dτ.93
tn

Let us consider stochastic equation with global derivative with Then,

following;
() ( ( )) ( ( )) x(tn+1 ) − x(tn ) (94)
α
0 Dg x t = f1 t, x t + f2 t, x t , (82)
{ ( ) ( ) ( ) ( ) }
3 ′ 1 ′
= g tn f1 tn , xn Δt − g tn− 1 f1 tn− 1 , xn− 1 Δt
x(t0 ) = x0 . 2 2
{ ( ) ( ) ( ) ( ) }
If g is positively increasing function differentiable then we can write 3 ′ ′ 1 ′ ′
+ g tn f2 tn , xn B (tn )Δt − g tn− 1 f2 tn− 1 , xn− 1 B (tn− 1 )Δt
( ) 2 2
′( ) ( ( )) ′( ) ( ( ))
α
0 Dt x t = g t f1 t, x t + g t f2 t, x t 83 (83)

Now we convert equation above to stochastic version as following: and

() ( ) ∫ t ( ) ( ( ))
′
∫ t ( ) ( ( ))
′
x(tn+1 ) − x(tn ) (95)
x t = x 0 + g τ f1 τ,x τ dτ + g τ f2 τ, x τ dB(τ).84
0 0 ( ) ( )
3 1
(84) = (g(tn ) − g(tn− 1 ))f1 tn , xn − (g(tn− 1 ) − g(tn− 2 ))f1 tn− 1 , xn− 1
2 2
Now we assume that B(t) is differentiable so we can write 3
( )
( ) ( ) ∫ t ( ) ( ( )) ∫ t ( ) ( ( )) + (g(tn ) − g(tn− 1 ))(B(tn ) − B(tn− 1 ))f2 tn , xn
′ ′ ′
2
x t = x 0 + g τ f1 τ,x τ dτ + g τ f2 τ,x τ B (τ)dτ.85 ( )
0 0 1
− (g(tn− 1 ) − g(tn− 2 ))(B(tn− 1 ) − B(tn− 2 ))f2 tn− 1 , xn− 1 .
(85) 2
We have at the point tn+1 = (n + 1)Δt,
( ) ( ) ∫ tn+1 ( ) ( ( )) Application of Atangana Toufik numerical scheme for fractional order SIR
x tn+1 − x 0 =
′
g τ f1 τ , x τ d τ stochastic model with global derivative
0
∫ tn+1 ( ) ( ( )) (86)
+
′ ′
g τ f2 τ, x τ B (τ)dτ.86 In this part, we show numerical scheme for solving for fractional
0 order SIR stochastic model with global derivative. In order to make a
And at the point tn = nΔt, more useful analysis, we will take kernels exponential decay, power-law
( ) ( ) ∫ tn ( ) ( ( )) ∫ tn ( ) ( ( )) and the Mittag–Leffler rule. While putting numerical scheme we will use
x tn − x 0 =
′
g τ f1 τ,x τ dτ +
′ ′
g τ f2 τ,x τ B (τ)dτ.87 Atangana Toufik numerical rules [12].
0 0 First of all let us write fractional order model (exponential kernel)
(87) with global derivative
() ( ( )) ( ( ))
Now, taking the difference of equations above we have following α
(96)
0 Dg s t = f1 t, s t + f2 t, s t ,
equation
( ) ( ) ∫ tn+1 ( ) ( ( )) ()
α
0 Dg i t = g1 (t, i (t)) + g2 (t, i (t)),
′
x tn+1 − x tn = g τ f1 τ, x τ dτ
()
(88)
tn
∫ tn+1 ( ) ( ( )) α
′ ′ 0 Dg r t = h1 (t, r(t)) + h2 (t, r(t)),
+ g τ f2 τ, x τ B (τ)dτ.88
tn
s(t0 ) = s0 , i (t0 ) = i 0 and r(t0 ) = r0 .
Let us do some simplifications Here
(89)
′
g (τ)f1 (τ, x(τ)) = Φ1 (τ, x(τ)), f1 (t, s(t)) = [Λ − μs(t) − βs(t)i (t)], (97)
′ ′
g (τ)f2 (τ, x(τ))B (τ) = Φ2 (τ, x(τ)). f2 (t, s(t)) = [ − σ s(t)i (t)],
g1 (t, i (t)) = [βs(t)i (t) − (μ + γ + ∊)i (t)],
So we write
g2 (t, i (t)) = [σs(t)i (t)],
( ) ( ) ∫ tn+1 ( ( )) ∫ tn+1 ( ( )) h1 (t, r(t)) = [γi (t) − μr(t)].
x tn+1 − x tn = Φ1 τ , x τ d τ + Φ2 τ, x τ dτ.90
tn tn If g(t) is positively increasing function differentiable then we have
(90) followings
Then consider the interpolation α
( ) ′( ) ( ( )) ′( ) ( ( ))
(98)
0 Dt s t = g t f1 t, s t + g t f2 t, s t ,
( ) ( ) ( )
τ − tn− 1 τ − tn
p1 τ = Φ1 tn , xn − Φ1 tn− 1 , xn− 1 .91 (91) α
( ) ′ ′
tn − tn− 1 tn − tn− 1 0 Dt i t = g (t)g1 (t, i (t)) + g (t)g2 (t, i (t)),
α
( ) ′ ′
( ) ( ) { ( ) ( ) } 0 Dt r t = g (t)h1 (t, r(t)) + g (t)h2 (t, r(t))
3 1
x tn+1 − x tn = Φ1 tn , xn Δt − Φ1 tn− 1 , xn− 1 Δt (92) So we can convert to above system to integral version with expo­
2 2
nential kernel below:

7
B.S.T. Alkahtani and I. Koca Results in Physics 24 (2021) 104124

() ( ) ∫ t ( ) ( ( )) ∫ t ( ) ( ( ))
s t =s 0 +
′
g τ f1 τ, s τ dτ +
′
g τ f2 τ, s τ dB(τ), i (tn+1 ) − i (tn ) (105)
0 0
( ) ( )
(99) 3 1
= (g(tn ) − g(tn− 1 ))g1 tn , i n − (g(tn− 1 ) − g(tn− 2 ))g1 tn− 1 , i n− 1
∫ ( ) ( ( )) ∫ t ( ) ( ( )) 2 2
( ) t
′ ′ ( )
i (t) = i 0 + g τ g1 τ, i τ dτ + g τ g2 τ, i τ dB(τ), 3
0 0 + (g(tn ) − g(tn− 1 ))(B(tn ) − B(tn− 1 ))g2 tn , i n
∫ t ( ) ( ( )) ∫ t ( ) ( ( )) 2
( ) ′ ′ ( )
r(t) = r 0 + g τ h1 τ , r τ d τ + g τ h2 τ, r τ dB(τ). 1
0 0 − (g(tn− 1 ) − g(tn− 2 ))(B(tn− 1 ) − B(tn− 2 ))g2 tn− 1 , i n− 1 ,
2
Here we assume that B(t) is differentiable then we can write
r(tn+1 ) − r(tn ) (106)
() ( ) ∫ t ( ) ( ( )) ∫ t ( ) ( ( ))
′ ′ ′
s t =s 0 + g τ f1 τ, s τ dτ + g τ f2 τ, s τ B (τ)dτ, ( ) ( )
3 1
0 0
= (g(tn ) − g(tn− 1 ))h1 tn , rn − (g(tn− 1 ) − g(tn− 2 ))h1 tn− 1 , rn− 1 .
(100) 2 2
∫ t ( ) ( ( )) ∫ t ( ) ( ( ))
( ) ′ ′ ′ Atangana-Toufik scheme for Caputo-Fabrizio order stochostic equation
i (t) = i 0 + g τ g1 τ, i τ dτ + g τ g2 τ, i τ B (τ)dτ,
0 0 with global derivative version
∫ t ( ) ( ( )) ∫ t ( ) ( ( ))
( ) ′ ′ ′
r(t) = r 0 + g τ h1 τ , r τ d τ + g τ h2 τ, r τ B (τ)dτ.
0 0 We start first with introducing the equation with Caputo-Fabrizio
version.
Now, taking the difference of equations above we have following ( ( )) ( ( ))
equation
CF α
Dg x(t) = f1 t, x t + f2 t, x t , (107)

( ) ( ) ∫ tn+1 ( ) ( ( )) ∫ tn+1 ( ) ( ( ))
(101)
′ ′ ′
s tn+1 − s tn = g τ f1 τ, s τ dτ + g τ f2 τ, s τ B (τ)dτ,
tn tn

∫ tn+1 ( ) ( ( )) ∫ tn+1 ( ) ( ( ))
′ ′ ′
i (tn+1 ) − i (tn ) = τ g1 τ , i τ d τ +
g g τ g2 τ, i τ B (τ)dτ,
tn tn
∫ tn+1 ( ) ( ( )) ∫ tn+1 ( ) ( ( ))
′ ′ ′
r(tn+1 ) − r(tn ) = g τ h1 τ , r τ d τ + g τ h2 τ, i τ B (τ)dτ
tn tn

x(t0 ) = x0 .
Rest of the proof we can apply Atangana Toufik scheme on its final
form, we have following numerical scheme If g is differentiable, then we will write
( ( )) ( ( ))
(102) (108)
CF α ′ ′
s(tn+1 ) − s(tn ) Dt x(t) = g (t)f1 t, x t + g (t)f2 t, x t 108
{ ( ) (
3 ′
)
1 ′
( ) ( ) } From the definition of the Caputo-Fabrizio integral, we can rewrite
=
2
g tn f1 tn , sn Δt − g tn−
2
1 f1 tn− 1 , sn− 1 Δt the above equation as;
{ ( ) ( ) ( ) ( ) } ( ( )) ∫t ( ( ))
3 ′ 1 ′ 1− α ′ α
g (τ)f1 τ, x τ dτ (109)
′ ′ ′
+ g tn f2 tn , sn B (tn )Δt − g tn− 1 f2 tn− 1 , sn− 1 B (tn− 1 )Δt , x(t) − x(0) = g (t)f1 t, x t +
2 2 M(α) M(α) 0
( ( )) ∫t ( ( ))
1− α ′ α
(103)
′
i (tn+1 ) − i (tn ) + g (t)f2 t, x t B(t) + g (τ)f2 τ, x τ dB(τ).
M(α) M(α) 0
{ ( ) ( ) ( ) ( ) }
=
3 ′ 1 ′
g tn g1 tn , i n Δt − g tn− g1 tn− 1 , i n− 1 Δt Now we assume that B(t) is differentiable then we can write
1
2 2 ( ( )) ∫t ( ( ))
{ ( ) ( ) ( ) ( ) } 1− α ′ α
g (τ)f1 τ, x τ dτ (110)
′
3 ′ ′ 1 ′ ′ x(t) − x(0) = g (t)f1 t, x t +
+ g tn g2 tn , i n B (tn )Δt − g tn− 1 g2 tn− 1 , i n− 1 B (tn− 1 )Δt M(α) M(α) 0
2 2
( ( )) ∫t ( ( )) ( )
1− α ′ α ′ ′
+ g (t)f2 t, x t B(t) + g (τ)f2 τ, x τ B τ dτ.
and M(α) M(α) 0
s(tn+1 ) − s(tn ) (104) We have at the point tn+1 = (n + 1)Δ(t),
( ) ( ) ( ( ))
1− α ′
=
3 1
(g(tn ) − g(tn− 1 ))f1 tn , sn − (g(tn− 1 ) − g(tn− 2 ))f1 tn− 1 , sn− 1 x(tn+1 ) − x(0) = g (tn+1 )f1 tn+1 , x tn+1 (111)
2 2 M(α)
( )
3
+ (g(tn ) − g(tn− 1 ))(B(tn ) − B(tn− 1 ))f2 tn , sn
2
( )
1
− (g(tn− 1 ) − g(tn− 2 ))(B(tn− 1 ) − B(tn− 2 ))f2 tn− 1 , sn− 1 ,
2

8
B.S.T. Alkahtani and I. Koca Results in Physics 24 (2021) 104124

∫ tn+1 ( ( ))
α ′ { ( ( )) ( ( )) }
+ g (τ)f1 τ, x τ d τ 1 − α g(tn+1 ) − g(tn )
M(α) 0 = f1 tn+1 , x tn+1 − f2 tn+1 , x tn+1 B(tn+1 )
M(α) Δt
( ( )) { ( ( )) ( ( )) }
1− α ′ 1 − α g(tn ) − g(tn− 1 )
+ g (tn+1 )f2 tn+1 , x tn+1 B(tn+1 ) + f1 tn , x tn − f2 tn , x tn B(tn )
M(α) M(α) Δt
∫ tn+1 ( ( )) ( ) { }
α ′ ′ α 3 1 ( )
+ g (τ)f2 τ, x τ B τ dτ + ϕ1 (tn , xn )Δt − ϕ1 tn− 1 , xn− 1 Δt
M(α) 0 M(α) 2 2
{ }
α 3 1 ( )
and at the point tn = nΔ(t), − ϕ2 (tn , xn )Δt − ϕ2 tn− 1 , xn− 1 Δt ,
M(α) 2 2
( ( ))
1− α ′
x(tn ) − x(0) = g (tn )f1 tn , x tn (112) ( ) ( )
M(α) g(tn ) − g(tn− 1 )
(118)
′
ϕ1 (tn , xn ) = g (tn )f1 tn , xn = f1 tn , xn ,
∫ tn ( ( )) Δt
α ′
+ g (τ)f1 τ, x τ d τ ( ) ( )
M(α) 0 ( ) ′ g(tn− 1 ) − g(tn− 2 )
( ( )) ϕ1 tn− 1 , xn− 1 = g (tn− 1 )f1 tn− 1 , xn− 1 = f1 tn− 1 , xn− 1 ,
1− α ′ Δt
+ g (tn )f2 tn , x tn B(tn ) ′ ′
M(α) ϕ2 (tn , xn ) = g (tn )f2 (tn , xn )B (tn ),
∫ tn ( ( )) ( ) ( )
α g(tn ) − g(tn− 1 ) B(tn ) − B(tn− 1 )
+
′ ′
g (τ)f2 τ, x τ B τ dτ. = f2 tn , xn ,
M(α) 0 Δt Δt
( ) ′ ( ) ′
Let us put some simplicity again for equation above; ϕ2 tn− 1 , xn− 1 = g (tn− 1 )f2 tn− 1 , xn− 1 B (tn− 1 )
( )
g(tn− 1 ) − g(tn− 2 ) B(tn− 1 ) − B(tn− 2 )
(113)
′
g (t)f1 (t, x(t)) = ϕ1 (t, x(t)), = f2 tn− 1 , xn− 1 .
Δt Δt
′ ′
g (t)f2 (t, x(t))B (t) = ϕ2 (t, x(t)).
And
Putting simplifications above and then consider the interpolation;
x(tn+1 ) − x(tn ) (119)
τ − tn− ( ) τ − tn
(114)
1
P(τ) = ϕ1 (tn , xn ) − ϕ1 tn− 1 , xn− 1 ,
tn − tn− 1 tn − tn− 1

( ) τ − tn { ( ( )) ( ( )) }
τ − tn− 1 1− α g(tn+1 )− g(tn )
P(τ) = ϕ2 (tn , xn ) − ϕ2 tn− 1 , xn− 1 . = f1 tn+1 ,x tn+1 − f2 tn+1 ,x tn+1 B(tn+1 )
tn − tn− 1 tn − tn− 1 M(α) Δt
{ ( ( )) ( ( )) }
Now taking the difference of equations above we have followings; 1− α g(tn )− g(tn− 1 )
+ f1 tn ,x tn − f2 tn ,x tn B(tn )
M(α) Δt
x(tn+1 ) − x(tn ) (115) { ( ) ( ) }
α 3 g(tn )− g(tn− 1 ) 1 g(tn− 1 )− g(tn− 2 )
+ f1 tn ,xn Δt− f1 tn− 1 ,xn− 1 Δt
M(α) 2 Δt 2 Δt
{ ( ( )) ( ( )) } { ( )
1− α ′ α 3 g(tn )− g(tn− 1 ) B(tn )− B(tn− 1 )
= g (tn+1 ) f1 tn+1 ,x tn+1 − f2 tn+1 ,x tn+1 B(tn+1 ) − f2 tn ,xn Δt
M(α) M(α) 2 Δt Δt
( ) }
{ ( ( )) ( ( )) } 1 g(tn− 1 )− g(tn− 2 ) B(tn− 1 )− B(tn− 2 )
1− α ′ − f2 tn− 1 ,xn− 1 Δt .
+ g (tn ) f1 tn ,x tn − f2 tn ,x tn B(tn ) 2 Δt Δt
M(α)
∫ tn+1 ( ( )) ∫ tn+1 ( ( )) ( )
α ′ α ′ ′ If we arrange all operations then we have;
+ g (τ)f1 τ,x τ dτ − g (τ)f2 τ,x τ B τ dτ.
M(α) tn M(α) tn x(tn+1 ) − x(tn ) (120)

We consider here interpolation polynominals.

x(tn+1 ) − x(tn ) (116) 1− α g(tn+1 )− g(tn )
{ ( ( )) ( ( )) }
= f1 tn+1 ,x tn+1 − f2 tn+1 ,x tn+1 B(tn+1 )
{ ( ( )) ( ( )) } M(α) Δt
1 − α g(tn+1 ) − g(tn ) { ( ( )) ( ( )) }
= f1 tn+1 , x tn+1 − f2 tn+1 , x tn+1 B(tn+1 ) 1− α g(tn )− g(tn− 1 )
M(α) Δt + f1 tn ,x tn − f2 tn ,x tn B(tn )
{ ( ( )) ( ( )) } M(α) Δt
1 − α g(tn ) − g(tn− 1 ) { ( ) ( ) ( )}
+ f1 tn , x tn − f2 tn , x tn B(tn ) α 3 1
M(α) Δt + g(tn )− g(tn− 1 ) f1 tn ,xn − (g(tn− 1 )− g(tn− 2 ))f1 tn− 1 ,xn− 1
M(α) 2 2
∫ tn+1 ( ( )) ∫ tn+1 ( ( )) { ( )
α α α 3 g(tn )− g(tn− 1 )
+ ϕ1 τ , x τ d τ − ϕ2 τ, x τ dτ, − f2 tn ,xn (B(tn )− B(tn− 1 ))
M(α) tn M(α) tn M(α) 2 Δt
( ) }
1 g(tn− 1 )− g(tn− 2 )
x(tn+1 ) − x(tn ) (117) −
2 Δt
f2 tn− 1 ,xn− 1 (B(tn− 1 )− B(tn− 2 )) .

Now we can apply this (*) scheme on sır stochastic system.

s(tn+1 ) − s(tn ) (121)

9
B.S.T. Alkahtani and I. Koca Results in Physics 24 (2021) 104124

{ ( ) ( ) }
1− α g(tn+1 )− g(tn )
= f1 tn+1 ,sn+1 − f2 tn+1 ,sn+1 B(tn+1 ) x(t) − x(0) (127)
M(α) Δt
{ ( ) ( ) }
1− α g(tn )− g(tn− 1 )
+ f1 tn ,sn − f2 tn ,sn B(tn )
M(α) Δt ∫ t
{ ( ) ( ) ( )} 1− α ′ ( ( )) α ( ( ))
g (τ)f1 τ,x τ (t − τ)α− 1 dτ
′
α 3 1 = g (t)f1 t,x t +
+ g(tn )− g(tn− 1 ) f1 tn ,sn − (g(tn− 1 )− g(tn− 2 ))f1 tn− 1 ,sn− 1 B(α) B(α)Γ(α) 0
M(α) 2 2 ∫
( ( )) t ( ( ))
{ ( ) 1− α ′ α ′( )
g (τ)f2 τ,x τ (t − τ)α− 1 B τ dτ.
′
α 3 g(tn )− g(tn− 1 ) + g (t)f2 t,x t B(t) +
− f2 tn ,sn (B(tn )− B(tn− 1 )) B(α) B(α)Γ(α) 0
M(α) 2 Δt
( ) }
1 g(tn− 1 )− g(tn− 2 ) We have at the point tn+1 = (n + 1)Δ(t),
− f2 tn− 1 ,sn− 1 (B(tn− 1 )− B(tn− 2 )) ,
2 Δt ( ( ))
1− α ′
x(tn+1 ) − x(0) = g (tn+1 )f1 tn+1 , x tn+1 (128)
B(α)
i (tn+1 ) − i (tn ) (122)
∫ tn+1 ( ( ))
α
g (τ)f1 τ, x τ (tn+1 − τ)α− 1 dτ
′
+
B(α)Γ(α) 0
{ ( ) ( ) }
1− α g(tn+1 )− g(tn ) ( ( ))
= f1 tn+1 ,i n+1 − f2 tn+1 ,i n+1 B(tn+1 ) 1− α ′
M(α) Δt + g (tn+1 )f2 tn+1 , x tn+1 B(tn+1 )
{ ( ) ( ) } B(α)
1− α g(tn )− g(tn− 1 ) ∫ tn+1 ( ( )) ( )
+ f1 tn ,i n − f2 tn ,i n B(tn ) α
g (τ)f2 τ, x τ (tn+1 − τ)α− 1 B τ dτ.
′ ′
M(α) Δt +
{ ( ) ( ) ( )} B(α)Γ(α) 0
α 3 1
+ g(tn )− g(tn− 1 ) f1 tn ,i n − (g(tn− 1 )− g(tn− 2 ))f1 tn− 1 ,i n− 1 Let us put some simplicity again for equation above;
M(α) 2 2
{ ( )
α 3 g(tn )− g(tn− 1 ) (129)
′

− f2 tn ,i n (B(tn )− B(tn− 1 )) g (t)f1 (t, x(t)) = ϕ1 (t, x(t)),

M(α) 2 Δt
( ) } ′ ′
1 g(tn− 1 )− g(tn− 2 ) g (t)f2 (t, x(t))B (t) = ϕ2 (t, x(t)).
− f2 tn− 1 ,i n− 1 (B(tn− 1 )− B(tn− 2 )) ,
2 Δt Putting simplifications above and then consider the interpolation;
( ) τ − tn− 1 ( ) τ − tn
1 − α g(tn+1 ) − g(tn ) P(τ) = ϕ1 (tn , xn ) − ϕ1 tn− 1 , xn− 1 , (130)
r(tn+1 ) − r(tn ) = h1 tn+1 , rn+1 (123) tn − tn− 1 tn − tn− 1
M(α) Δt
τ − tn− 1 ( ) τ − tn
( ) P(τ) = ϕ2 (tn , xn ) − ϕ2 tn− 1 , xn− 1 .
1 − α g(tn ) − g(tn− 1 ) tn − tn− 1 tn − tn− 1
+ h1 tn , rn
M(α) Δt
{ ( ) ( ) Now taking the difference of equations above we have followings;
α 3
+ g(tn ) − g(tn− 1 ) h1 tn , rn x(tn+1 ) (131)
M(α) 2
( )} ( ) { ( ( )) ( ( )) }
1 1− α ′
− (g(tn− 1 ) − g(tn− 2 ) )h1 tn− 1 , rn− 1 . = x 0 + g (tn+1 ) f1 tn+1 , x tn+1 + f2 tn+1 , x tn+1 B(tn+1 )
2 B(α)
∫ tn+1 ( ( ))
α
g (τ)f1 τ, x τ (tn+1 − τ)α− 1 dτ
′
+
Atangana-Toufik scheme for Atangana-Baleanu order stochostic equation B(α)Γ(α) 0
with global derivative version ∫ tn+1 ( ( )) ( )
α
g (τ)f2 τ, x τ (tn+1 − τ)α− 1 B τ dτ.
′ ′
+
B(α)Γ(α) 0
We start first with introducing the equation with Atangana-Baleanu
version. We consider here interpolation polynominals.
( ( )) ( ( ))
ABC α
Dg x(t) = f1 t, x t + f2 t, x t , (124) x(tn+1 ) − x(0) (132)
{ ( ( )) ( ( )) }
x(t0 ) = x0 . 1 − α g(tn+1 ) − g(tn )
= f1 tn+1 , x tn+1 + f2 tn+1 , x tn+1 B(tn+1 )
B(α) Δt
If g is differentiable, then we will write ∫ tn+1 ( ( ))
α
( ( )) ( ( )) + ϕ1 τ, x τ (tn+1 − τ)α− 1 dτ
(125)
′ ′
AB α
Dt x(t) = g (t)f1 t, x t + g (t)f2 t, x t 125 B(α)Γ(α) 0
∫ tn+1 ( ( ))
From the definition of the Caputo-Fabrizio integral, we can rewrite +
α
ϕ2 τ, x τ (tn+1 − τ)α− 1 dτ.
the above equation as; B(α)Γ(α) 0

x(t) − x(0) (126) Then we write

x(tn+1 ) − x(0) (133)
( ( )) ∫ t ( ( ))
1− α ′ α
τ,x τ (t− τ)α− 1 dτ
′
= g (t)f1 t,x t + g (τ)f1
B(α) B(α)Γ(α) 0
( ( )) ∫t ( ( ))
1− α ′ α
g (τ)f2 τ,x τ (t− τ)α− 1 dB(τ).
′
+ g (t)f2 t,x t B(t)+
B(α) B(α)Γ(α) 0

Now we assume that B(t) is differentiable then we can write

10
B.S.T. Alkahtani and I. Koca Results in Physics 24 (2021) 104124

Replacing by its Lagrange polynomial interpolation formula, we

obtain
1 − α g(tn+1 ) − g(tn )
xn+1 = x0 + (134)
B(α) Δt

.{f1 (tn+1 , x(tn+1 )) + f2 (tn+1 , x(tn+1 ))B(tn+1 )}

( )[ ]
α(Δt)α ∑n (n − j + 1)α (n − j + 2 + α)
j
+ ϕ tj , x
B(α)Γ(α + 2) j=0 1 − (n − j)α (n − j + 2 + 2α)
( )[ ]
α(Δt)α ∑n
j− 1
(n − j + 1)α+1
− ϕ tj− 1 , x
B(α)Γ(α + 2) j=0 1 − (n − j)α (n − j + 1 + α)

Fig. 1. Stochastic behavior of s(t). ( )[

(n − j + 1)α (n − j + 2 + α)
]
α(Δt)α ∑n
+ ϕ tj , xj
B(α)Γ(α + 2) j=0 2 − (n − j)α (n − j + 2 + 2α)
( )[ ]
α(Δt)α ∑n (n − j + 1)α+1
− ϕ tj− 1 , xj− 1
.
B(α)Γ(α + 2) j=0 2 − (n − j)α (n − j + 1 + α)

Replacing ϕ1 (t, x(t)) and ϕ2 (t, x(t)) by their values, then the above
equation is converted to

xn+1 (135)

{ ( ( )) ( ( )) }
1− α g(tn+1 )− g(tn )
= x0 + f1 tn+1 ,x tn+1 +f2 tn+1 ,x tn+1 B(tn+1 )
B(α) Δt
( ) [ α ]
α(Δt)α− 1 ∑ n
( ( ) ( )) (n− j+1) (n− j+2+ α)
+ f1 tj ,xj g tj+1 − g tj
Fig. 2. Stochastic behavior of ı(t).
B(α)Γ(α +2) j=0 − (n− j)α (n− j+2+2α)
( ) [ ]
α(Δt)α− 1 ∑n
( ( ) ( )) (n− j+1)α+1
− f1 tj− 1 ,xj− 1
g tj − g tj− 1
B(α)Γ(α +2) j=0 − (n− j)α (n− j+1+ α)
( )
α(Δt)α− 2 ∑n
( ( ) ( ))( ( ) ( ))
+ f2 tj ,xj g tj+1 − g tj B tj+1 − B tj
B(α)Γ(α +2) j=0
[ ]
(n− j+1)α (n− j+2+ α)
.
− (n− j)α (n− j+2+2α)
( )
α(Δt)α− 2 ∑ n
j− 1
( ( ) ( ))( ( ) ( ))
− f2 tj− 1 ,x g tj − g tj− 1 B tj − B tj− 1
B(α)Γ(α +2) j=0
[ ]
(n− j+1)α+1
. .
− (n− j)α (n− j+1+ α)

Now we can apply final scheme on sır stochastic system. (See

Figs. 1–3)
{ ( ( ))
Fig. 3. Stochastic behavior of r(t). 1 − α g(tn+1 ) − g(tn )
sn+1 = s0 + f1 tn+1 , s tn+1
B(α) Δt
( ( )) } (136)
+ f2 tn+1 , s tn+1 B(tn+1 )
{ ( ( )) ( ( )) }
1 − α g(tn+1 ) − g(tn )
= f1 tn+1 , x tn+1 + f2 tn+1 , x tn+1 B(tn+1 )
B(α) Δt
( ( )) ( ) [ ]
n ∫ tj+1 α
α ∑ α(Δt)α− 1 ∑n
( ( ) ( )) (n− j+1) (n− j+2+ α)
+ ϕ1 τ, x τ (tn+1 − τ)α− 1 dτ + f1 tj ,sj g tj+1 − g tj
B(α)Γ(α) j=2 tj B(α)Γ(α +2) j=0 − (n− j)α (n− j+2+2α)
( ( )) ( ) [ ]
α n ∫
∑ tj+1 α(Δt)α− 1 ∑n
( ( ) ( )) (n− j+1)α+1
+ ϕ2 τ, x τ α− 1
(tn+1 − τ) d τ. − f1 tj− 1 ,sj− 1
g tj − g tj− 1
B(α)Γ(α) tj
B(α)Γ(α +2) j=0 − (n− j)α (n− j+1+ α)
j=2

11
B.S.T. Alkahtani and I. Koca Results in Physics 24 (2021) 104124

( ) α [ ]
α(Δt)α− 2 ∑ n
( ( ) ( ))( ( ) ( )) (n − j + 1) (n − j + 2 + α)
+ f2 tj , sj g tj+1 − g tj B tj+1 − B tj
B(α)Γ(α + 2) j=0 − (n − j)α (n − j + 2 + 2α)
( ) [ ]
α(Δt)α− 2 ∑ n
( ( ) ( ))( ( ) ( )) (n − j + 1)α+1
− f2 tj− 1 , sj− 1
g tj − g tj− 1 B tj − B tj− 1 ,
B(α)Γ(α + 2) j=0 − (n − j)α (n − j + 1 + α)
{ ( ( )) ( ( )) }
1 − α g(tn+1 ) − g(tn )
i n+1 = i 0 + g1 tn+1 , i tn+1 + g2 tn+1 , i tn+1 B(tn+1 ) (137)
B(α) Δt

( ) [α ]
α(Δt)α− 1 ∑ n
( ( ) ( )) (n − j + 1) (n − j + 2 + α)
+ g1 tj , i j g tj+1 − g tj
B(α)Γ(α + 2) j=0 α
− (n − j) (n − j + 2 + 2α)
( ) [ ]
α(Δt)α− 1 ∑ n
( ( ) ( )) (n − j + 1)α+1
− g1 tj− 1 , i j− 1
g tj − g tj− 1
B(α)Γ(α + 2) j=0 − (n − j)α (n − j + 1 + α)
( ) α [ ]
α(Δt)α− 2 ∑ n
( ( ) ( ))( ( ) ( )) (n − j + 1) (n − j + 2 + α)
+ g2 tj , i j g tj+1 − g tj B tj+1 − B tj
B(α)Γ(α + 2) j=0 − (n − j)α (n − j + 2 + 2α)
( ) [ ]
α(Δt)α− 2 ∑ n
( ( ) ( ))( ( ) ( )) (n − j + 1)α+1
− g2 tj− 1 , i j− 1
g tj − g tj− 1 B tj − B tj− 1 .
B(α)Γ(α + 2) j=0 − (n − j)α (n − j + 1 + α)
{ ( ( ))}
1 − α g(tn+1 ) − g(tn )
rn+1 = r0 + h1 tn+1 , r tn+1
B(α) Δt
( ) [ α ]
α(Δt)α− 1 ∑ n
( ( ) ( )) (n − j + 1) (n − j + 2 + α)
+ h1 tj , rj g tj+1 − g tj
B(α)Γ(α + 2) j=0 − (n − j)α (n − j + 2 + 2α)
( ) [ ]
α(Δt)α− 1 ∑ n
( ( ) ( )) (n − j + 1)α+1
− h1 tj− 1 , rj− 1
g tj − g tj− 1 .
B(α)Γ(α + 2) j=0 − (n − j)α (n − j + 1 + α)

Numerical simulation

In this section, we depict numerical simulation of the chosen system

of fractional stochastic differential equations. We have made use of the
model with the Caputo differential operator and the numerical scheme
that was suggested by Atangana and Toufik where the Lagrange poly­
nomial interpolation is used. The numerical simulation are performed
for different values of fractional orders. The used initial conditions are
1000, 10 and 1 respectively. The first 3 figures are depicted for the
following values of densities of randomness 0.001, 0.006 and 0.006

Fig. 4. Stochastic behavior of ı(t). Fig. 5. Stochastic behavior of s(t).

12
B.S.T. Alkahtani and I. Koca Results in Physics 24 (2021) 104124

Fig. 10. Stochastic behavior of s(t).

Fig. 6. Stochastic behavior of r(t).

Fig. 11. Stochastic behavior of ı(t).

Fig. 7. Stochastic behavior of ı(t).

Fig. 12. Stochastic behavior of r(t).

respectively. To obtain Fig. 4, 5 the densities of randomness used are

given as 0.0016, 0.007 and 0.008, Figs. 7–9 0.0018, 0.01, and 0.01
Fig. 8. Stochastic behavior of s(t). finally figure Fig. 10–12, the randomness was removed. (See Fig. 6).

Conclusion

A simple SIR model was considered in this work. First we presented a

detailed analysis of stability using existing technics such as Lyapunov
function, the Ruth criteria. We presented the condition under which the
Lyapunov function was positive, zero and negative. The system of or­
dinary differential equation was converted to fractional stochastic dif­
ferential equation with the aim to include into the mathematical model
the effect of nonlocality and randomness. The existence and uniqueness
has been presented. A numerical scheme based on the Lagrange inter­
polation was used to solve numerically. Some simulations are presented.

Declaration of Competing Interest

Fig. 9. Stochastic behavior of r(t). The authors declare that they have no known competing financial

13
B.S.T. Alkahtani and I. Koca Results in Physics 24 (2021) 104124

interests or personal relationships that could have appeared to influence [12] Toufik M, Atangana A. New numerical approximation of fractional derivative with
non-local and non-singular kernel: application to chaotic models. Eur Phys J Plus
the work reported in this paper.
2017;132(10):444.
[13] Podlubny I. Fractional Differential Equations. New York: Academic Press; 1999.
Acknowledgements [14] Odibat ZM, Momani S. Int J Nonlinear Sci Numer Simul 2006;7:27–34.
[15] Ahmed E, El-Sayed AMA, El-Saka HAA. Equilibrium points, stability and numerical
solutions of fractional-order predator-prey and rabies models. J Math Anal Appl
The authors extend their appreciation to the Deputyship for Research 2007;325:542–53. https://doi.org/10.1016/j.jmaa.2006.01.087.
& Innovation, “Ministry of Education” in Saudi Arabia for funding this [16] Matignon D. Stability results for fractional differential equations with applications
research work through the project number IFKSURG-1437-017. to control processing. 2. Computational Eng. in Sys. Appl. 1996, 963. Lille, France
1996.
[17] Ahmed E, El-Sayed AMA, El-Saka HAA. On some Routh-Hurwitz, conditions for
References fractional order differential equations and their applications in Lorenz, R ossler,
Chua and Chen systems. Phys Lett A 2006;358:1–4.
[1] Tornatore E, Buccellato SM, Vetro P. Stability of a stochastic SIR system. Phys A [18] Atangana A, Araz S. Nonlinear equations with global differential and integral
2005;354:111–26. operators: existence, uniqueness with application to epidemiology. Results Phys
[2] Iannelli M. Mathematical problems in the description of age structured 103593; 2020.
populations. Math Biol Med 19–32 (Bari); 1983. [19] Khan MA, Atangana A. Modeling the dynamics of novel coronavirus (2019-nCov)
[3] Anderson RM, May RM. Population biology of infectious diseases, part I. Nature with fractional derivative. Alex Eng J 2020;59(4):2379–89.
1979;280:361–7. [20] Koutou O, Traoré B, Sangaré B. Mathematical modeling of malaria transmission
[4] Gray A, Greenhalgh D, Hu L, Mao X, Pan J. A stochastic differential equation SIS global dynamics: taking into account the immature stages of the vectors. Adv Differ
epidemic model. SIAM J Appl Math 2011;71:876–902. Equ 2018;2018:220. https://doi.org/10.1186/s13662-018-1671-2.
[5] Atangana A. Extension of rate of change concept: from local to nonlocal operators [21] Takeuchi S, Kuroda Y. Predicting spread of new pandemic swine-origin influenza A
with applications. Results Phys 2020. (H1N1) in localmid-size city: evaluation of hospital bed shortage and effectiveness
[6] Atangana A. Modelling the spread of Covid-19 with new fractal-fractional of vaccination. Nippon Eiseigaku Zasshi 2010;65(1):48–52.
operators: can the lockdown save mankind before vaccination? Chaos Solitons [22] Tracht S, Del Valle S, Hyman J. Mathematical modeling of the effectiveness of
Fractals 2020;136:109860. facemasks in reducing the spread of novel influenza A (H1N1). PLoS One 2010;5
[7] Atangana A, Baleanu D. New fractional derivatives with non-local and non-singular (2):e9018.
kernel. Theory and Application to Heat Transfer Model, Thermal Science 2016;20 [23] Ding Y, Wang Z, Ye H. Optimal control of a fractional-order HIV-immune system
(2):763–9. with memory. Control Systems Technology. IEEE Trans 2011; PP(99):1–7.
[8] Caputo M, Fabrizio M. A new deÖnition of fractional derivative without singular [24] Pinto CM, Machado JT. Fractional model for malaria transmission under control
kernel. Prog Fractional Di Serentiation Appl 2015;1(2):73–85. strategies. Comput Math Appl 2013;66(5):908–16.
[9] Uçar S. Analysis of a basic SEIRA model with Atangana-Baleanu derivative. AIMS [25] Sambo Dachollom, Chinwendu Emilian Madubueze. Mathematical Model of the
Math 5(2): 1411–1424. Transmission Dynamics of Lassa Fever Infection with Controls, Mathematical
[10] Özdemir N, Uçar E. Investigating of an immune system-cancer mathematical model Modelling and Applications. Vol. 5, No. 2, 2020, pp. 65-86. doi: 10.11648/j.
with Mittag-Leffler kernel. AIMS Math 5(2): 1519–1531. mma.20200502.13.
[11] Uçar S, Özdemir N, Koca I, Altun E. Novel analysis of the fractional glucose–insulin
regulatory system with non-singular kernel derivative. Eur Phys J Plus 135(6): 414.

14