Results in Physics 19 (2020) 103551

Contents lists available at ScienceDirect

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

The Riemann–Liouville fractional derivative for Ambartsumian equation

E.R. El-Zahar a,b , A.M. Alotaibi c , A. Ebaid c , A.F. Aljohani c , J.F. Gómez Aguilar d ,∗
a
Department of Mathematics, Faculty of Sciences and Humanities, Prince Sattam bin Abdulaziz University, Alkharj 11942, Saudi Arabia
b
Department of Basic Engineering Science, Faculty of Engineering, Shebin El-Kom 32511, Menofia University, Egypt
c
Department of Mathematics, Faculty of Science, University of Tabuk, P.O. Box 741, Tabuk 71491, Saudi Arabia
d CONACyT-Tecnológico Nacional de México/CENIDET, Interior Internado Palmira S/N, Col. Palmira, C.P. 62490, Cuernavaca, Morelos, Mexico

ARTICLE INFO ABSTRACT

Keywords: The Ambartsumian equation, based on the modified Riemann–Liouville fractional derivative, is analyzed in this
Ambartsumian equation paper. The solution is expressed as a power series of arbitrary powers and its convergence has been proven.
Riemann–Liouville In addition, we show that the present solution reduces to the results in the literature when the fractional
Series solution
derivative tends to 1. Moreover, the behavior of the obtained solution is discussed through figures.

1. Introduction approaches were applied in the literature to solve various fractional

differential equations in the applied sciences, the power series approach
Based on the modified Riemann–Liouville fractional derivative (R– is more useful here as will be shown later in proving the convergence
L FD), we consider a fractional model of the Ambartsumian delay issue.
equation (ADE) [1] in the form:

𝐷𝑡𝛼 𝜙(𝑡) = 𝜙(𝑡) + 𝛾 𝜙 (𝛾 𝑡) , (1) 1.1. Definitions and properties

subjected to
The modified R–L FD is defined as [5–8]
𝜙(0) = 𝜔, (2)
⎧ 1 𝑡 −𝛼−1 (𝑓 (𝑡) − 𝑓 (0)) 𝑑𝜁 ,
where 𝐷𝑡𝛼 is the R–L FD, 𝛾 = 𝜇1 , 𝜇 > 1, and 𝜔 is a constant. In the ⎪ 𝛤 (−𝛼) ∫0 (𝑡 − 𝜁) if 𝛼 < 0,
⎪
special case 𝛼 → 1, the present model has been solved via several 𝐷𝑡𝛼 𝑓 (𝑡) = ⎨ 1 𝑑 ∫0𝑡 (𝑡 − 𝜁)−𝛼 (𝑓 (𝜁) − 𝑓 (0))𝑑𝜁 , if 0 < 𝛼 < 1,
𝛤 (1−𝛼) 𝑑𝑡
analytical approaches [2,3]. The authors [4] investigated the current ⎪
⎪(𝐷𝑡𝑛 𝑓 (𝑡))𝛼−𝑛 , if 𝑛 ≤ 𝛼 < 𝑛 + 1, 𝑛 ≥ 1,
problem in view of the Caputo’s fractional definition. The modified R– ⎩
L FD was applied to analyze several scientific models in Mathematical
(3)
Physics [5–8].
The model (1)–(2) can also be analyzed via various methods such so that 𝑓 : R → R, 𝑡 → 𝑓 (𝑡), denotes a continuous function and
as the ADM [9,10], and the HPM [11–13]. In the literature [14– 𝛤 represents the Gamma function. Some important properties of the
21], many applications of the fractional calculus in several fields such fractional derivative (2) are
as engineering, physics, biology, and medicine have been discussed.
𝛤 (1 + 𝑟) 𝑟−𝛼
Cattani and Pierro [14] reported on the fractal geometry of DNA by the 𝐷𝑡𝛼 𝑡𝑟 = 𝑡 , (4)
𝛤 (1 + 𝑟 − 𝛼)
binary image analysis. Kiymaz et al. [15] introduced an extension of Ca-
puto fractional derivative operator and they provided its applications. 𝐷𝑡𝛼 (𝑓 (𝑡)𝑔(𝑡)) = 𝑔(𝑡)𝐷𝑡𝛼 𝑓 (𝑡) + 𝑓 (𝑡)𝐷𝑡𝛼 𝑔(𝑡), (5)
Abdon [16] implemented fractional calculus to predict complex system.
𝑑𝑓 (𝑔(𝑡)) 𝛼
Jagdev et al. [17] analyzed the fractional epidemiological model for 𝐷𝑡𝛼 𝑓 (𝑔(𝑡)) = 𝐷𝑡 𝑔(𝑡). (6)
𝑑𝑔(𝑡)
computer viruses, while Yokus and Gülbahar [18] developed numerical
solutions with linearization techniques for the Fractional Harry Dym The modified R–L FD (3) is considered here because it possesses the
Equation. Other interesting recent applications and utilizations of the properties (4)–(6) which help in obtaining the solutions as a closed
fractional calculus can be found in Refs. [19–21]. Although several form.

∗ Corresponding author.
E-mail address: [email protected] (J.F. Gómez Aguilar).

https://doi.org/10.1016/j.rinp.2020.103551
Received 16 May 2020; Received in revised form 17 September 2020; Accepted 25 October 2020
Available online 19 November 2020
2211-3797/© 2020 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/).
E.R. El-Zahar et al. Results in Physics 19 (2020) 103551

2. Power series solution

Assume that Eq. (1) has a fractional power series solution (FPSS) in
the following form:
∑
∞
𝜙(𝑡) = 𝑎𝑛 (𝛼, 𝛾) 𝑡𝛼𝑛 , (7)
𝑛=0

which yields,
∑∞
𝛤 [𝛼(𝑛 + 1) + 1]
𝐷𝑡𝛼 𝜙(𝑡) = 𝑎𝑛+1 (𝛼, 𝛾) 𝑡𝛼𝑛 . (8)
𝑛=0
𝛤 [𝛼(𝑛 + 1)]

On inserting (7)–(8) into (1), we have

∑∞
𝛤 [𝛼(𝑛 + 1) + 1] ∑∞ ∑∞
𝑎𝑛+1 (𝛼, 𝛾) 𝑡𝛼𝑛 = − 𝑎𝑛 (𝛼, 𝛾)𝑡𝛼𝑛 + 𝛾 𝛼𝑛+1 𝑎𝑛 (𝛼, 𝛾) 𝑡𝛼𝑛 ,
𝑛=0
𝛤 [𝛼(𝑛 + 1)] 𝑛=0 𝑛=0

(9) Fig. 1. Domain of convergence of the approximate solutions (17) at 𝜔 = 1 and 𝜇 = 1.4.

or
∑∞
𝛤 [𝛼(𝑛 + 1) + 1] ∑∞
( 𝛼𝑛+1 ) As 𝛼 → 1, we have from (17) that
𝑎𝑛+1 (𝛼, 𝛾) 𝑡𝛼𝑛 = 𝛾 − 1 𝑎𝑛 (𝛼, 𝛾)𝑡𝛼𝑛 , (10) [ ( 𝑛 ) ]
𝛤 [𝛼(𝑛 + 1)]
𝑛=0 𝑛=0 ∑∞ ∏( ) 𝑡𝑛
lim 𝜙(𝑡) = 𝜔 1 + 𝛾𝑘 − 1 ,
Collecting the like terms of same powers, yields 𝛼→1 𝑛!
𝑛=1 𝑘=1
∞ ( ) [ ( 𝑛 ( )) 𝑛 ]
∑ 𝛤 [𝛼(𝑛 + 1) + 1] ( ) ∑∞ ∏ 1 𝑡
𝑎𝑛+1 (𝛼, 𝛾) − 𝛾 𝛼𝑛+1 − 1 𝑎𝑛 (𝛼, 𝛾) 𝑡𝛼𝑛 = 0. (11) =𝜔 1+ − 1 , (18)
𝑛=0
𝛤 [𝛼(𝑛 + 1)] 𝜇𝑘 𝑛!
𝑛=1 𝑘=1

Therefore which is the same closed form solution obtained by [2] as 𝛼 → 1.

𝛤 [𝛼(𝑛 + 1)] ( 𝛼𝑛+1 )
𝑎𝑛+1 (𝛼, 𝛾) = 𝛾 − 1 𝑎𝑛 (𝛼, 𝛾), 𝑛 ≥ 0. (12) 3. Convergence analysis and remarks
𝛤 [𝛼(𝑛 + 1) + 1]
Accordingly,
To proving convergence of (17), we assume that
1 ( 𝑛 )
𝑎1 = (𝛾 − 1) 𝑎0 , 1 ∏( )
𝛤 [𝛼 + 1] 𝑐𝑛 = 𝛾 (𝑘−1)𝛼+1 − 1 . (19)
1 ( ) 𝛤 [𝛼𝑛 + 1] 𝑘=1
𝑎2 = (𝛾 − 1) 𝛾 𝛼+1 − 1 𝑎0 ,
𝛤 [2𝛼 + 1]
( )( ) Accordingly, the theorem below is resulted.
1
𝑎3 = (𝛾 − 1) 𝛾 𝛼+1 − 1 𝛾 2𝛼+1 − 1 𝑎0 ,
𝛤 [3𝛼 + 1]
( )( )( ) Theorem 1. The series (17) has an infinite radius of convergence for
1
𝑎4 = (𝛾 − 1) 𝛾 𝛼+1 − 1 𝛾 2𝛼+1 − 1 𝛾 3𝛼+1 − 1 𝑎0 , |𝜇| > 1∀ 𝛼 ∈ (0, 1].
𝛤 [4𝛼 + 1]
. (13) Proof. Assume that 𝜌 is the radius of convergence. Therefore, we have
. from (17) and the ratio test that
. 1 | 𝑐𝑛+1 |
= lim || |,
1 ( )( ) ( ) 𝜌 𝑛→∞ | 𝑐𝑛 ||
𝑎𝑛+1 = (𝛾 − 1) 𝛾 𝛼+1 − 1 𝛾 2𝛼+1 − 1 … 𝛾 𝛼𝑛+1 − 1 𝑎0 ,
𝛤 [𝛼(𝑛 + 1) + 1] | ∏𝑛+1 ( (𝑘−1)𝛼+1 )|
| 𝛤 [𝛼𝑛 + 1] 𝑘=1 𝛾 −1 |
= lim || ∏𝑛 ( )|,
and therefore 𝑛→∞ | 𝛤 [𝛼(𝑛 + 1) + 1] (𝑘−1)𝛼+1 − 1 ||
| 𝑘=1 𝛾 |
𝑎0 ∏𝑛
( 𝑘𝛼+1 ) | ( 𝛼𝑛+1 )|
𝑎𝑛+1 (𝛼, 𝛾) = 𝛾 −1 , 𝑛 ≥ 0. (14) | 𝑛 𝛤 [𝛼𝑛] |
𝛤 [𝛼(𝑛 + 1) + 1] 𝑘=0 = lim | 𝛾 − 1 |,
𝑛→∞ | (𝑛 + 1)𝛤 [𝛼(𝑛 + 1)] |
| 𝑛 | | 𝛤 [𝛼𝑛] | | 1 |
Hence = lim || | × lim | | × lim | − 1|| ,
𝑛→∞ | 𝑛 + 1 || 𝑛→∞ || 𝛤 [𝛼(𝑛 + 1)] || 𝑛→∞ || 𝜇 𝛼𝑛+1 |
∑
∞
= 1 × 0 × 1 ∀ 𝛼 ∈ (0, 1], where 𝜇 > 1,
𝜙(𝑡) = 𝑎0 + 𝑎𝑛+1 (𝛼, 𝛾) 𝑡𝛼(𝑛+1) ,
𝑛=0 = 0. (20)
∑
∞
𝑡𝛼(𝑛+1) ∏𝑛
( 𝑘𝛼+1 )
= 𝑎0 + 𝑎0 𝛾 −1 , 4. Discussion
𝑛=0
𝛤 [𝛼(𝑛 + 1) + 1] 𝑘=0
[ ( 𝑛 ) ]
∑∞ ∏( ) 𝑡𝛼(𝑛+1)
= 𝑎0 1 + 𝛾 𝑘𝛼+1
−1 . (15) From (17), the 𝑚-term approximate solution is
𝑛=0 𝑘=0
𝛤 [𝛼(𝑛 + 1) + 1] [ ( 𝑛 ) ]
∑ ∏
𝑚−1
( (𝑘−1)𝛼+1 ) 𝑡𝛼𝑛
On using the initial condition (2), we finally obtain 𝛹𝑚 (𝑡) = 𝜔 1 + 𝛾 −1 , 𝑚 ≥ 2. (21)
𝛤 [𝛼𝑛 + 1]
[ ( 𝑛 ) ] 𝑛=1 𝑘=1
∑∞ ∏( ) 𝑡𝛼(𝑛+1) Although the series (17) converges in the limit, it has a certain domain
𝜙(𝑡) = 𝜔 1 + 𝛾 𝑘𝛼+1 − 1 , (16)
𝑛=0 𝑘=0
𝛤 [𝛼(𝑛 + 1) + 1] of convergence as shown in Fig. 1 for 𝛹60 (𝑡), 𝛹70 (𝑡), 𝛹80 (𝑡), and 𝛹90 (𝑡).
Hence, these approximations are not valid in the whole domain 𝑡 ∈
which can also be written as
[ ( 𝑛 ) ] [0, ∞). However, the domain of convergence can be enhanced to include
∑
∞ ∏( ) 𝑡𝛼𝑛 the whole domain 𝑡 ∈ [0, ∞) using the Pade Approximants (PA).
𝜙(𝑡) = 𝜔 1 + 𝛾 (𝑘−1)𝛼+1 − 1 . (17)
𝑛=1 𝑘=1
𝛤 [𝛼𝑛 + 1] To achieve this goal, the diagonal PA [𝑁∕𝑁] for 𝑁 = 3, 4, 5, 6 are

2
E.R. El-Zahar et al. Results in Physics 19 (2020) 103551

Table 1
Comparison between the present results and other methods in the literature.
𝑡 Daftarday-Gejji HATM [7] Present PA Exact [5]
and Jafari
technique [2]
[ ] [ ] [ ] [ ]
1 2 3 4
1
(𝑡) 2
(𝑡) 3
(𝑡) 4
(𝑡)

0.0 1 1 1 1 1 1 1
0.5 0.8727825992 0.8727825992 0.8727272727 0.8729413222 0.8729409259 0.8729409265 0.8729409265
1.0 0.7694328044 0.7694328044 0.7704918033 0.7717933884 0.7717847387 0.7717847887 0.7717847885
1.5 0.6788327993 0.6788327993 0.6865671642 0.6899802422 0.6899343729 0.6899349314 0.6899349261
2.0 0.5898647673 0.5898647673 0.6164383561 0.6228435550 0.6227056589 0.6227084454 0.6227083998

𝛹80 (𝑡), and 𝛹90 (𝑡) has its own domain of convergence as shown in Fig. 1.
Hence, these approximations were not valid in the whole domain of
the time-domain 𝑡 ∈ [0, ∞). It was also shown that such domains
of convergence were enhanced and therefore enlarged to include the
whole domain 𝑡 ∈ [0, ∞) through applying the Pade Approximants
(PA), especially, the diagonal PA [𝑁∕𝑁] for 𝑁 = 3, 4, 5, 6 as depicted
and shown in Fig. 2. The obtained results in Fig. 2 showed that the
domains of convergence were really enhanced even by using a low
order of the PA, where each diagonal PA converges in the whole
domain.

Declaration of competing interest

The authors declare that they have no known competing finan-

cial interests or personal relationships that could have appeared to
influence the work reported in this paper.
Fig. 2. Plots of different Pade Approximants versus t at 𝜔 = 1 and 𝜇 = 1.4.
Acknowledgments

José Francisco Gómez Aguilar acknowledges the support provided

calculated and depicted in Fig. 2 in the same domain of Fig. 1. As a
by CONACyT: cátedras CONACyT para jóvenes investigadores 2014,
result, Fig. 2 shows that the domain of convergence is really enhanced
Mexico and SNI-CONACyT, Mexico.
even by using a low order of the PA.
In order to indicate the effectiveness of the obtained results via
References
the PAs, we introduce in Table 1 several comparisons between the
methods in the relevant literature and the current one. In Table 1, the
[1] Ambartsumian VA. Doklady Akad Nauk USSR 1994;44:223.
numerical results obtained via several PAs [𝑁∕𝑁], 𝑁 = 1, 2, 3, 4 are [2] Patade J, Bhalekar S. Natl Acad Sci Lett 2017;40:291.
listed against the corresponding ones in the relevant literature. It can [3] Kato T, McLeod JB. Bull Amer Math Soc 1971;77:891.
be easily seen from this table that the present calculations agree with [4] Kumar D, Singh J, Baleanu D, Rathore S. Eur Phys J Plus 2018;133:259.
[5] Huang Q, Zhdanov R. Physica A 2014;409:110.
the exact results, especially, by increasing the degree of the constructed
[6] Prakash A, Kumar M, Sharma KK. Appl Math Comput 2015;260:314.
PA. For declaration, the numerical results obtained via the PA [4∕4] are [7] Guner O, Aksoy E, Bekir A, Cevikel AC. Comput Math Appl 2016;71:1259.
much accurate than the lower-order PAs. Furthermore, the present PA [8] Sahadevan R, Prakash P. Chaos Solitons Fractals 2017;104:107.
analysis is of more accuracy than those in the relevant literature [2] [9] Alshaery A, Ebaid A. Acta Astronaut 2017;140:27.
and [5]. [10] Bakodah HO, Ebaid A. Rom Rep Phys 2018;70:115.
[11] Patra A, Saha Ray S. Int J Heat Mass Transfer 2014;76:162.
[12] Ayati Z, Biazar J. J Egyptian Math Soc 2015;23:424.
5. Conclusion [13] Zhang Y, Cattani C, Yang X-J. Entropy 2015;17:6753.
[14] Cattani C, Pierro G. Bull Math Biol 2013;75:1544.
In view of the modified R–L FD, the ADE was introduced and [15] Kiymaz IO, Çetinkaya A, Agarwa P. J Nonlinear Sci Appl 2016;09:3611.
[16] Abdon A. Chaos Solitons Fractals 2017;102:396.
analytically solved via the power series approach. The obtained series
[17] Jagdev S, Devendra K, Zakia H, Abdon A. Appl Math Comput 2018;316:504.
reduced to the same series in the literature as the fractional derivative [18] Yokus A, Gülbahar S. Appl Math Nonlinear Sci 2019;4:35.
tends to 1. The convergence of the obtained series was checked and [19] Gao W, Veeresha P, Prakasha DG, Baskonus, Haci M. Biology 2020;9(5):107.
proven by the ratio test. The obtained explicit series solution converges [20] Ilhan E, Kiymaz IO. Appl Math Nonlinear Sci 2020;5(1):171.
in the limit, however, each of the approximate solutions 𝛹60 (𝑡), 𝛹70 (𝑡), [21] Zaid O, Dumitru B. Appl Numer Math 2020;156:94.