The Concepts and Applications of Fractional Order Differential Calculus

in Modelling of Viscoelastic Systems: A primer

Mohammad Amirian Matlob1, Yousef Jamali1,2*

1
Biomathematics Laboratory, Department of Applied Mathematics, Tarbiat Modares University, Iran
2
Computational physical Sciences Research Laboratory, School of Nano-Science, Institute for Research in Fundamental Sciences (IPM), Tehran, Iran

Viscoelasticity and related phenomena are of great importance in the study of mechanical properties of material especially, biological
materials. Certain materials show some complex effects in mechanical tests, which cannot be described by standard linear equation
(SLE) mostly owing to shape memory effect during deformation. Recently, researchers have been applying fractional calculus in order
for probing viscoelasticity of such materials with a high precision. Fractional calculus is a powerful tool for modeling complex
phenomenon. In this tutorial based paper, we try present clear descriptions of the fractional calculus, its techniques and its
implementation. The intention is to keep the details to a minimum while still conveying a good idea of what and how can be done with
this powerful tool. We try to expose the reader to the basic techniques that are used to solve the fractional equations analytically
and/or numerically. More specifically, modeling the shape memory phenomena with this powerful tool are studied from different
perspectives, as well as presented some physical interpretation in this case. Moreover, in order to show the relationship between
fractional models and standard linear equations, a fractal system comprising spring and damper elements is considered, and
the constitutive equation is approximated with a fractional element. Finally, after a brief literature review, two fractional models are
utilized to investigate the viscoelasticity of the cell, and the comparison is made among them, experimental data, and previous models.
Verification results indicate that not only does the fractional model match the experimental data well, but it also can be a good
substitute for previously used models.

Keywords: Viscoelasticity, Fractional calculus, Mechanical properties, Cell biomechanics, Fractional modeling, Fractal system.

Author Summary
Fractional Calculus is a new powerful tool which has been recently employed to model complex biological systems with non-linear
behavior and long-term memory .In spite of its complicated mathematical background, fractional calculus came into being of some
simple questions which were related to the derivation concept; such questions as while the first order derivative represents the slope
of a function, what a half order derivative of a function reveal about it? Finding answers to such questions, scientists managed to open
a new window of opportunity to mathematical and real world, which has arisen many new questions and intriguing results. For
example, the fractional order derivative of a constant function, unlike the ordinary derivative, is not always zero. In this tutorial-
based paper it is sought to answer the aforementioned questions and to construct a comprehensive picture of what fractional calculus
is, and how it can be utilized for modelisation purpose. The focus of the research has been on viscoelastic materials. After an extensive
literature review of the concepts and application of this potent tool, a novel application of this tool is developed for simulating a
dynamic system in order to investigate the mechanical behavior of a cell.

*Corresponding author. Tel: +98-21-8288-4762

E-mail addresses: y.jamali@ipm.ir (Y.Jamali), m.amirianmatlob@modares.ac.ir (M.Amirian Matlob).
1
Contents

Introduction ............................................................................................................................................ 3
Methods .................................................................................................................................................. 3
Fractional calculus................................................................................................................................... 3
Definition of fractional calculus ......................................................................................................... 6
The interpretation of fractional calculus ................................................................................................ 9
Laplacean interpretation ...................................................................................................................... 9
Analytical and numerical methods ....................................................................................................... 10
Analytical methods ........................................................................................................................... 10
Numerical solution ............................................................................................................................ 13
Viscoelastic systems .............................................................................................................................. 14
Creep test .......................................................................................................................................... 15
Periodic test ....................................................................................................................................... 16
Fractional viscoelastic model ................................................................................................................ 17
Periodic strain ................................................................................................................................... 21
Relation between fractional equations and SLE ................................................................................... 21
Application ............................................................................................................................................ 23
Two-dimension fractional network model ........................................................................................... 24
Discussion ............................................................................................................................................. 27
Appendix ............................................................................................................................................... 28
Acknowledgment .................................................................................................................................. 33
References ............................................................................................................................................ 34

2
Introduction

The concept of derivative is the main idea of calculus. It shows the sensitivity to change of a function
i.e. the rate or slope of a quantity. The current definition of derivative was suggested by Newton in
1666. Newton with a physical viewpoint of derivative interpreted the instantaneous velocity [1]. The
intuition of researchers from derivative and integral is based on their geometrical or physical meaning,
e.g. the first and second order derivative of displacement is called velocity and acceleration respectively
(also jerk and jounce for 3th and 4th derivatives). As well, integral of a curve function means the area
under the curve. This form of classical calculus was developed extensively over four centuries. Today,
scientists are able to describe and model many physical phenomena with an ordinary differential
equation [2]. In many cases, however, the classical calculus is not able to describe exactly these complex
phenomena. Under a small deformation, for example, the relationship between the force and
displacement in an ideal spring (small deformations in elastic materials) is linear, i.e. force is related to
the zero derivative of displacement, while in an ideal damper, the force is proportional to the velocity
of extension or compression. In other words, force is related to the first derivative of deformation.
Which law, however, governs the materials with intermediate mechanical properties (i.e. between ideal
spring and ideal damper)? With this aim, many researchers try to answer the question and to model as
well as to analyse the mechanical behavior of these non-linear systems by means of fractional calculus
[3].
Fractional calculus is a generalization of ordinary differentiation and integration to arbitrary non-integer
order, but with this definition, many interesting questions will arise; for example, if the first derivative
of a function gives you the slope of the function, what is the geometrical meaning of half derivative? In
half order, which operator must be used twice to obtain the first derivative? The early history of this
questions back to the birth of fractional calculus in 1695 when Gottfried Wilhelm Leibniz suggested
the possibility of fractional derivatives for the first time]2[ .
In this article, we aim to introduce fractional calculus as a new tool for modeling the complex systems,
especially viscoelastic material. First, we briefly discuss the basic concepts of fractional calculus and
explain the essential steps of the fractionalization algorithm. Next, we present an interpretation of
fractional derivative and elaborate upon how fractional equations could be solved analytically. Then,
we briefly look at the modeling of viscoelastic systems by the help of this approach. Ultimately, after
overviewing some recent works, we present an application of the approach in modeling biomechanical
properties of a cell and indicate that the proposed model predicts the cell behavior much better than the
previous spring-dashpot models, as well as the model outputs are in good agreement with experimental
data. To sum, we are going to give the minimum need to get reader ”feet wet”, so that a reader can
quickly get into building a fractional calculus model for a complex system.

Methods

Fractional calculus
In mathematics, many complex concepts developed from simple concepts. For example, we can refer
to the extension of natural number to the real one in some mathematical formulae. Let's give an example
to clarify: the factorial of a non-negative integer n, denoted by n!, is the product of all positive integers
less than or equal to n. On the other hand, there is a concept which named Gama function and defined
as follows:
∞
Γ(𝑥) = ∫0 𝑡 𝑧−1 𝑒 −𝑡 𝑑𝑥 (1)

3
One property of the function for 𝑛 ∈ ℝ+ is
Γ(𝑛 + 1) = nΓ(𝑛) (2)
Hence, this function is equal to factorial for the integer numbers. As a result, the gamma function could
be considered as an extension of factorial function to real numbers. For instance, according to the above
formalism, a factorial of 1/2 can be obtained as follows:

1 3 3 1 3√𝜋
( )! = Γ( ) = Γ( ) = (3)
2 2 2 2 2
Indeed, according to Wikipedia, the gamma function can be seen as the solution to find a smooth curve
that connects the points (x, y) given by y = (x − 1)! at the positive integer values for x. Let’s use this
approach to extend the concept of derivative to non-integer order; consider 𝑛𝑡ℎ derivation of power
function 𝑔(𝑥):

𝑔(𝑥) = 𝑥 𝑘 𝑠. 𝑡 𝑥 ≥ 0 (4)
𝑑𝑛 𝑘! Γ(1 + 𝑘)
𝑛
𝑔(𝑥) = 𝑥 𝑘−𝑛 = 𝑥 𝑘−𝑛 (5)
𝑑𝑥 (𝑘 − 𝑛)! Γ(1 + 𝑘 − 𝑛)
which 𝑘 and 𝑛 are real integer number respectively, and 𝑘 ≥ 𝑛. To generalize the above equation, it
could be possible to extend the integer number 𝑛 to a real value which named 𝛼:
𝑑𝛼 Γ(1 + 𝑘)
𝛼
𝑔(𝑥) = 𝑥 𝑘−𝛼 (6)
𝑑𝑥 Γ(1 + 𝑘 − 𝛼)
Then for fractional derivative of an arbitrary function, expand the function in a power series of 𝑥 first,
and then by using 𝑒𝑞. 5, derivate the expansion. For example, for derivative 𝑓(𝑥) = 𝑒 𝑘𝑥 to 𝛼 order, we
rewrite 𝑓(𝑥) function as follows:

x2 x3
f (x )  1  x   
2! 3! (7)
Hence [4]:
d     , kx  
 sign  x   sign  x  k  e kx 1 
1 1 1 
f  x  x   x1  x 2  
dx
 1     2    3         
(8)
That Γ(−𝛼, 𝑘𝑥) is incomplete gamma function (discussed in Box 3).
Of course, this is an arbitrary way to define fractional derivative and not the only way; for example, it
is possible to use an exponential function 𝑓(𝑥) = 𝑒 𝑘𝑥 instead of a power function, i.e. we can
define
𝐷𝛼 𝑓(𝑥) = 𝑘 𝛼 𝑒 𝑘𝑥 (9)
The fractional derivative of the exponential function obtained by Liouville in 1832, and the fractional
derivative of power function got by Riemann in 1847 [4]. By comparing Eq. 8 with 9, it is noticed that
when the order is an integer, the results are same, but at the non-integer order they are different! In fact,
contrary to the integer-order derivative in which the definition and the output under these operators are
same and unique in the fractional derivatives, under different operators, the result is not the same and
not necessarily unique. In other words, there are multiple definitions for fractional derivative and all of
them are mathematically correct. From the physical perspective, each definition has its own application
and interpretation.

4
Fractional derivative has many interesting and counter-intuitive properties; for example, the Riemann
derivative (Eq. 5) of a constant is not zero! that is,
𝑑𝛼 𝑑𝛼 1
𝑐𝑜𝑛𝑠 = 𝑥0 = 𝑥 −𝛼 (10)
𝑑𝑥 𝛼 𝑑𝑥 𝛼 Γ(1−𝛼)

Putting it all together, there are many ways to define a fractional derivative, provided that each
definition approach to ordinary derivative in the integer order limit—this method known as
the fractionalization algorithm—i.e.:

lim  f (x )   n f (x ) , n  0,1,....
 n

Where Δ represents an arbitrary operator. As well, two more examples provided in Box 1 which have
also been used in the following sections.

Box 1

In this box, a basic definition of both the ordinary integral and the ordinary derivative is presented by
two examples, as well as its extension to fractional operators is proposed by the help of fractionalization
algorithm (ref section 2); resulting we obtain two important fractional functions.

Example 1. Assuming 𝑓(𝑥) ∈ [𝑎, 𝑏]. The first order derivative of function 𝑓(𝑥) is defined as follows:
df f (t )  f (t  h)
f (1) (t )   lim (B1.1)
dt h0 h
by applying derivative operator on the above equation, we get:
d2 f f (t )  2 f (t  h)  f (t  2h)
f (2)
(t )  2  lim (B1.2)
dt h  0 h2
by the help of induction and above procedure, we gain:
dn f 1 n
n
f ( n ) (t )  2
 lim n  (1) r
  f (t  rh) (B1.3)
dt h 0 h
r 0 r
𝑡−𝑎 𝑛 𝑛(𝑛−1)(𝑛−2)…(𝑛−𝑟+1)
Where 𝑎 ≤ 𝑡 ≤ 𝑏, ℎ = , and ( ) = .
𝑛 𝑟 𝑟!
Since the equation (B1.3) hold for all 𝑛 ∈ 𝑁, we expand 𝑛 order to 𝛼 ∈ 𝑅 by the fractionalization
algorithm; thereby achieving a really important equation known as fractional Grunwald-Letnikov
differintegral mentioned in section 2.
1 
 
 (1)
( )
f n (t )  lim r
  f (t  rh) (B1.4)
h  0 h
r 0 r
Example 2. Consider the following equation which is known as the integral operator of order n:
We can rewrite it as follow according to the Cauchy equation [6]:
xn xn1 x1

 D f 
n
  f ( x ) dx 0 0 dxn 1 (B1.5)
 
Since the above equation holds for all 𝑛 ∈ 𝑁, we apply the fractionalization algorithm to expand the
obtained equation to fractional equation. In other words, if n → α, we will able to gain an integral known
as Riemann-Liouville. Hence, it seems that fractional calculus is an expansion of integer calculations.
x
1
(n  1)! 
D
n
f  ( x   )n1 f ( )d (B1.6)

5
Definition of fractional calculus
As mentioned in the previous section, the fractional order derivative is not necessarily unique; in this
regard, there are some accepted and common definitions in the literature. In the following, we
mentioned a number of important ones.
Grünwald–Letnikov derivative
Grünwald–Letnikov derivative is a basic extension of the natural derivative to fractional one, which
derived in Box 1 (Eq. B1.4). It was introduced by Anton Karl Grünwald in 1867, and then by Aleksey
Vasilievich Letnikov in 1868. Hence, it is written as [5]:

1 
 
D f (t )  lim
h  0 h
 (1) m
  f (t mh) (11)
m0 m
where 𝑛 ∈ 𝑁, and the binomial coefficient is calculated by the help of the Gamma function.

    (  1)(  2)...(  m  1)
  )12(
m m!

Riemann-Liouville fractional derivative

Riemann-Liouville fractional derivative acquiring by Riemann in 1847 is defined as follows.

d n
RL
a Dt f (t )  ( ) ( a Dt ( n  ) ) f (t )
dt
(13)
1 d t f ( x)
 ( )n  dx, (n  [ ]  1, t  a)
(n   ) dt a (t  x)  n 1

where 𝛼 > 0; this operator is an extension of Cauchy’s integral (Box 1) from the natural number to real
one. Based on the fractionalization algorithm, it seems logical to reach relation (13) by n order
derivative of the Eq. B1.6
In addition, according to the above relation, if 0 < 𝛼 < 1 then the Riemann-Liouville operator reduced
to
1 d t f ( x)
Dt f (t )   (14)
RL
dx
(1   ) dt (t  x)
a a

It is worth noting that this relation is the same of Eq. 5. [5], and also RL, 𝛼, 𝑎, and 𝑡 are the abbreviation
of Riemann-Liouville, fractional order, the lower and upper bound of the above integral respectively.

Caputo derivative
Since Riemann-Liouville fractional derivatives failed in the description and modeling of some complex
phenomena, Caputo derivative was introduced in 1967 [6]. The Caputo derivative of fractional order α
(𝑛 − 1 ≤ 𝛼 < 𝑛 ) of function 𝑓(𝑡) defined as

t D f ( )
n
1
(n   ) a (t   ) n 1

C
a Dt f (t )  d (15)

6
where 𝐷𝑛 is 𝑛𝑡ℎ derivative operation, and C represents the Caputo word. In other words, based on box
2, it can be found that the Caputo derivative is equal to the Riemann–Liouville integral of a
𝑛𝑡ℎ derivative of a function.
It is worth noting again that “the behavior of all of the fractional derivatives, when the order is integer
are same; it, however, could be different in the non-integer order; for example, in the non-integer order,
the Caputo derivative of a constant, unlike the Riemann–Liouville derivative is zero.
There are more fractional derivatives for more detailed information, the interested reader is referred to
[4, 5, 7]. Since we try to analyse and to investigate viscoelastic systems via the Riemann-Liouville and
Caputo fractional derivative, the main focus of the article is on these two types of derivative.
As previously mentioned, different definitions for fractional derivative with the different properties can
be proposed, which all of them are valid and mathematically acceptable. However, the main question
is “which relation should be applied in modeling of a specific phenomenon? In other words, which
definition would be more appropriate for a specific problem?” As a rule of thumb, since they tend to
interpret natural phenomena, the definition which is more consistent with the experimental results have
more privilege than the other fractional definitions.

BOX 2
In this box, some properties of fractional Riemann-Liouville differintegral, such as commutative
property, distributive laws, and so forth are investigated2.

Riemann–Liouville integral and derivative

To find a profound understanding of Riemann–Liouville integral and derivative, some of the most
crucial properties of this operator are mentioned in the following.
Lemma 1. Assuming arbitrary function 𝑓(𝑥) and 𝑚, 𝑛 ≥ 0 the following equations hold3.
1. Semi-group property:
I am I an f  I am n f
2. commutative property:
I am I an f (x )  I an I am f (x )
Lemma 2. Let 𝑓1 and 𝑓2 are two functions on [a,b] as well as 𝑐1 , 𝑐2 ∈ ℝ, 𝑛 > 0, and 𝑚 > 𝑛. Regarding
these, the following equations hold4:
1. Linearity rules:
RL
Dan (f 1  f 2 )  RL Dan f 1  RL Dan f 2 , RL
Dan (c1f 1 )  c1 RL Dan (f 1 )
2. Zero rule:
𝐷0 𝑓 = 𝑓
3. Product rule:
 q
 
RL
Dtq (fg )     RL Dt j (f ) RL Dtq  j ( g )
j 0  j 

4. In the general, semi-group property does not hold for Riemann-Liouville fractional derivative.
Indeed, the following equation is not always true.
D D f  RL D a b f
RL a RL b

2
All the mentioned properties in this box hold almost everywhere (a.e) on [a, b] and also, if (X, Σ, μ) is a measure space, a quality 𝑃 is said to hold almost
everywhere in 𝑋 if 𝜇({𝑥 ∈ 𝑋: 𝑃(𝑥)}) = 0. Also, integral operator on interval [𝑎, 𝑏] is defined as 𝐼𝑎𝑏 .
3 𝑎
𝑓 ∈ 𝐿1 [𝑎, 𝑏] where 𝐿1 ≔ {𝑓: [𝑎, 𝑏] → ℝ; 𝑓 𝑖𝑠 𝑚𝑒𝑎𝑠𝑢𝑟𝑎𝑏𝑙𝑒 𝑜𝑛 [𝑎, 𝑏] 𝑎𝑛𝑑 ∫𝑏 |𝑓(𝑥)|1 𝑑𝑥 < ∞ }.
4
Provided that 𝐷𝑎𝑛 𝑓1 and 𝐷𝑎𝑛 𝑓2 almost everywhere define.
7
NB: To prove the above lemmas see [8].

Caputo derivative
In this part, some fundamental properties related to Caputo operator are represented based on ref [5, 8].
Let 𝑓 is a enough differentiable function, 𝑐1 , 𝑐2 ∈ ℝ, and 𝑚 > 𝑛 ≥ 0. Considering these,
1. Caputo derivative is the left inverse of Riemann-Liouville integral.
C
Dan I an f  f
2.
m 1
D k f (a )
I an C Dan f (x )  f (x )   (x  a ) k
k 0 k!
3. Distributive law in Caputo derivative.
C
Dan c1f 1  c 2f 2   c1  C Dan f 1   c 2  C Dan f 2 
4. Leibniz equation5.
(x  a )  n 
n 
C
D an  fg  (x )  g (a )  f (x )  f (a )    C D an g (x )  f (x )      I ak  n g (x )  C D ak f (x )
(1  n ) k 1  k 

5. The semi-group property, the following equation, holds under some special condition6.
C
D a C D a f  C D a   f

NB: The part 5 of the above Lemma) does not hold for Riemann–Liouville derivative generally. To
prove this claim, assuming 𝒇 function with identity function 𝒇(𝒙) = 𝟏, as well as 𝑎 = 0, 𝑛 = 1, and
1
𝜀 = 1/2. Thus, if Riemann–Liouville operator is applied to the left side, we will obtain (𝐷02 𝑓 ́)(𝑥) =
1
𝐷02 0 = 0. On the other hand, by applying this operator to the right side, we will gain

1
D03/ 2 f ( x)  D 2 I 01/ 2 f ( x)  x 3/ 2
(1/ 2)

The relationship between fractional integral and fractional derivative

In this part, the relationship between fractional integral and fractional derivative is investigated. There
are always several properties in the classical integral and derivative area accepted as constant roles;
these properties, however, might not hold always in the fractional sense. The following equations, for
instance, are almost same in the classical calculation, but it has a major difference in the fractional
calculation (𝑚 > 𝛼).
dm and dm
D   D  m D m a I m  D   D m D  m  aI
m 

dx m dx m

Indeed, a constant and a variable term in calculation represent this difference, which the constant and
variable term are related to non-fractional and fractional sense respectively. In other words, assuming
function f: [a, b] → ℝ where −∞ < a < b < ∞; if the function 𝑓(𝑥) on closed interval [𝑎, 𝑏] has been
𝑛 order differentiable, then for all 𝑥 ∈ (𝑎, 𝑏), we have:

5
Assuming 0 < 𝑛 < 1, 𝑔 and 𝑓 are analytical on (𝑎 − ℎ, 𝑎 + ℎ) are essential to hold the equation.
6
Special condition for existing this property is 𝑓 function must belong to 𝐶 𝑘 [𝑎, 𝑏] where 𝐶 𝑘 [𝑎, 𝑏] ≔ {𝑓: [𝑎, 𝑏] → ℝ; 𝑓 ℎ𝑎𝑠 𝑎 𝑐𝑜𝑛𝑡𝑖𝑛𝑢𝑜𝑢𝑠 𝐾𝑡ℎ 𝑑𝑒𝑟𝑖𝑣𝑎𝑡𝑖𝑣𝑒} for
some 𝑎 < 𝑏 and some 𝑘 ∈ ℕ. Existing ℓ ∈ ℕ with ℓ ≤ 𝑘 and 𝑛, 𝑛 + 𝜀 ∈ [ℓ − 1, ℓ] is essential as well (𝑛, 𝜀 > 0).
NB: The existence of ℓ with the mentioned properties is essential. Otherwise, it is enough to assume that 𝑛 = 𝜀 = 7/10(𝑖. 𝑒 7/10 = 𝑛 < 1 < 𝑛 + 𝜀 = 7/5),
𝑎 = 0, and 𝑓(𝑥) = 𝑥, causing the right side of lemma become zero owing to C D 7/5f (x )  (I 3/5f )(x )  I 3/5 0  0 ; however, 𝑐𝐷7/10𝑓(𝑥)= 1 𝑥 13/10 . As a result,
0 0 0 0 Γ(13/10)
the left side of lemma will obtain as follow:
1
C
D 07/10 ( C D 07/10f )(x )  x 2/5
(3 / 5)
8
 ( C Da I a f )(x )  f (x )

In a simple word, if the fractional derivative with order 𝛼 ∈ ℝ is applied on the fractional integral with
the same order, the output of that will be 𝑓(𝑥). Thus, we can say, “Caputo derivative is the left inverse
of Riemann-Liouville integral.” Also
n 1
| x  a |j
(I a C D a f )(x )  f (x )   (f (j)
(x ) )
j! x a
j 0

NB: To prove the mentioned theorems see [9].

With regarding the Riemann-Liouville as a derivative operator, the following equation hold (𝒏 > 𝟎,
𝒎 > 𝒏):
m 1
(x  a )n k 1
I an RL Dan f (x )  f (x )   lim D m k 1I am n f (z )
k  0 ( n  k )
z a 

and particularly, for 0 < 𝑛 < 1, we have:

(x  a )n 1
I an RL Dan f (x )  f (x )  lim I a1n f (z )
(n ) z a 

The interpretation of fractional calculus

There is a dozen of suggestions for interpretation of fractional calculus [4, 7]. However, most of them
are a little bit abstract and give no physical intuition. Hence, we just present a most useful interpretation
of fractional calculus. The core of these interpretations is memory concept. In general, when the output
of a system at each time 𝑡 depends only on the input at time 𝑡, such systems are said to be memoryless
systems. On the other side, when the system has to remember previous values of the input in order to
determine the current value of the output, such systems are said non-memoryless systems, or memory
systems. For example, all of the Markov chain phenomena are memoryless [3, 4], and human decision
making or shape memory alloy are non-memoryless.

Laplacean interpretation
Suppose that 𝑌(𝑡) is a quantity whose value in terms of 𝑓(𝑡) can be achieved as follows:

(t   )( 1)
t
Y (t )   f ( )d (16)
0
( )

i.e. the output 𝑌(𝑡) can be viewed as a power-weighted sum which stores the previous input of function
𝑓(𝑡). Based on the above definition, such system is a non-memoryless system and in such systems,
memory decays at the rate of 𝑤(𝑡) = 𝑡 𝛼−1 /Γ(α).
Applying the Caputo derivative of order 𝛼 to the both sides of the last relation leads to
C
0 DtY (t )  f (t ) (17)

As a result, the differential equation governing the system memory 𝑌(𝑡) is described by a fractional
derivative. Therefore, the fractional derivative is a good candidate to explain the system with memory.
The nature of weighted function determines the type of fractional derivative which describes a system
memory. For example, If the weight function of a system is defined by 𝑡 1−𝛼 ∕ 𝛤(𝛼) , the Riemann-
9
Liouville elements, and by 𝑡 1−𝛼 𝜃(𝑥 − 𝑡)/Γ(𝛼) the Caputo elements are used--which θ is the Heaviside
function [4].
Since the convolution operation representing the integral of the product of the two functions after one
is reversed and shifted, so if one of the functions considered as a weighted function, it may be claimed
that the task of this function is collecting the system memory over time. For example, in the Riemann-
Liouville, "The relationship indicates that the information function 𝐹(𝑡) is memorized (storage) with a
power low rate function". In Ref [4] some other weight function is reviewed.
From the physical point of view, what the memory is and how it is defined in a system depends on a
deep understanding of the phenomena. Also, there is no certain rules and methods to select the
fractional type in a modeling. In other words, any definition that its result is more consistent with the
experimental data is the suitable definition to us.

Analytical and numerical methods

Analytical methods
There are different methods to solve fractional differential equations analytically. One of the most
common and widely used methods is the Laplace transform. In the following, by an example, this
method is illustrated (for other methods see reference [7]).
Before proceeding, it is worth noting that in general, the number of initial conditions that are required
for a given ordinary differential equation will depend upon the order of the differential equation.
However, in the fractional differential equation, number of initial condition equals to the integer lower
bound of order value (𝛼) [10, 11].
Consider the following differential equation

 D x(t )  kx(t )  f (18)

Which 𝑥(𝑡) is displacement, 𝑘, and 𝜉 are constants, as well as the fractional derivative is also Caputo
and 0 < 𝛼 < 1--In what follows, it has been shown that this differential equation model is the dynamics
of a purely elastic spring and a viscoelastic element connecting in parallel with a body of mass m, which
a force f is applied on a body. In order to solve, the first step is to take the Laplace transform of both
sides of the original differential equation (the Laplace transformation is concisely explained in Box 3).
Based on the equation B3.2 in Box 3, we have
f
X (s)  (19)
 (s  k /  )


where 𝛼 and 𝑠 are fractional order and Laplace domain variable respectively. Also, it is supposed that
𝑥(0) = 0. To find the solution, all we need to do is to take the inverse transform (Box 3.1):
f k
x(t )  t  1 E , ( t ) (20)
 

Which 𝐸𝛼,𝛼 (𝑡) is Mittag-Leffler function [for more information see the Box 3 footnote]

10
In the last example, if the spring is ignored, the Eq. 18 will be reduced to
𝑓 = 𝜉𝐷𝛼 𝑥(𝑡) (21)
Then as above, by a) taking the Laplace transforms of both sides of the equation, b) simplifying
algebraically the result to solve the obtained equation in terms of 𝑠, and c) finally finding the inverse
transform, we have:
𝑥(𝑡) = 𝐾𝑡 𝛼 (22)

Which 𝐾 = 𝑓/𝜉Γ(1 + 𝛼). Although the Laplace transformation method is one of the simple and
practical methods for solving the fractional equations--same as the ordinary differential equations, most
of the fractional equation could not be solved analytically. In what follows, we present a numerical
technique to solve Caputo fractional differential equation.

BOX 3

In this box, the concept of Laplace Transform is presented for several fractional definitions. To begin
with, let us bring up some basic facts in this regard [7].
The function 𝐹(𝑠) of the complex variable 𝑠 is defined as the following equation—which is known as
the Laplace transform of the function 𝑓(𝑡):

F (s )  L {f (t )}   e st f (t )dt (B 3.1)
0

To exist integral (B3.1), the function 𝑓(𝑡) must have been an exponential order 𝛼. In other words, the
existence of two positive constants, such as 𝑀 and 𝑇 which satisfy the following condition is essential.
e t | f (t ) |  M for all t  T .
Indeed, when 𝑡 → ∞, the function 𝑓(𝑡) cannot grow faster than a certain exponential function.

With the help of the inverse Laplace transform, the original 𝑓(𝑡) can be gained from the Laplace
transform 𝐹(𝑠)
1 a  ib

2 i b  a ib
f (t )  L 1{F (s )}  lim e st F (s )ds (B 3.2)

where the integration is done along the vertical line 𝑅𝑒(𝑠) = 𝑎 in the complex plane such that 𝑎 is
greater than the real part of all singularities of 𝐹(𝑠), which guarantees that the contour path is in the
convergent region. If either all singularities are in the left half-plane or 𝐹(𝑠) is a smooth function on
−∞ < 𝑅𝑒(𝑠) < ∞ (i.e., no singularities), then 𝑎 can be set to zero and the above inverse integral
formula becomes identical to the inverse Fourier transform.

The direct calculation of the inverse Laplace transform (B3.2) is often sophisticated. Sometimes,
however, it gives useful information on the unknown original 𝑓(𝑡) behavior which we look for.

The following formula seems to be another useful property for the Laplace transform of the derivative
of an integer order 𝑛 of the function 𝑓(𝑡):
n 1
L{ f (n )
(t )}  s n F (s )  s n 1f (0)  s n 2f (0)   f ( n 1)
(0)  s n F (s )   s n k 1f (k )
(0) ( B 3.3)
k 0

which can be obtained from the definition (B3.1). To facilitate applying the Laplace transform, we
gather some formulae of vital functions in the table (B3.1) for both the integer mode and the non-integer
mode.

11
 Table B3.1. Laplace transform table of some basic fractional calculus.

Fractional order Integer order

𝑓(𝑡) = 𝐿−1 {𝐹(𝑠)} 𝐹(𝑠) = 𝐿{𝑓(𝑡)} 𝑓(𝑡) = 𝐿−1 {𝐹(𝑠)} 𝑭(𝒔) = 𝑳{𝒇(𝒕)}
1 t  1 1  (t )
s  ( )
1 t  1 t 1 t k 1
e , k  1, 2,3,...
(s  a ) ( ) sk ( k )
a 1
 ( , at ) 7
a F (s )  bG (s ) a f (t )  bg (t ); a , b 
s (s  a ) ( )
s E  ( at  ) 8
s n L {f (t )}  s n 1f (0) f (n ) (t ), n  0
s (s   a ) s n  2 f (0)  f ( n 1)
(0)
 1 
1 t E  , (at ) a sin(at )

s a s  a2 2

a 1  E (at  ) s cos(at )

s (s  a ) s  a2
2

s F ( ;1; at )
1 1
9
F (s  c ) e ct f (t )
s (s  a )
1 t  E1,1 (at ) (1)n F ( n ) (s ) t n f (t ), n  1, 2,3,...

s (s  a )
s t  E1,1 (at ), 0    1 F (s )G (s ) t

(s  a )
 f (t   ) g ( )d 
0

   1  
1
s t E  , (at )
 s
F (u )du f (t )
s a t
s   t  1 F (s ) t

(s  a ) (  )
1F1 ( ;  ; at )
s  f ( )d
0

The Laplace transform of fractional operators

In this part, the Laplace transform of fractional operators is represented, and some formulae of which
is summarized in table (B3.2).
We introduced Riemann- Liouville integral in Box 1 as follow:
t
1 1  1
( ) 0

I f  (t   ) 1 f ( )d   [t * f (t )] (B 3. 4)
( )
0 t

With applying the Laplace transform on the Eq. B3.4, we obtain (based on Table B3.1):
1
L {0 I  f }  L {t  1}L {f (t )}  s  F (s )
( )

7
incomplete gamma function

(a, z )   t a 1e t dt
z
8
𝐸𝛼,𝛽 is denoted to Mittage-Leffler function such that

xk
E  ,  (x )   , ,   0
k  0 ( k    )
Also, if 𝛼 = 𝛽 = 1, then this function can be equaled with exponential function.

9
Hypergeometric functions



(b j )
q
(ai  n ) z n
p

j 1
F ({a };{b }; z )  i 1

 (a ) 
p q i j
(b j  n ) n !
p q
n 0
i 1 i j 1

12
Similarly, other Laplace transform of fractional operators will be gained by utilizing the proposed
formulae in Table B3.1 [7].
Table B3.2. The Laplace transform of some fractional operators with order 𝛼.

Operator Fractional formula Laplace Transform of the fractional order

operators
Riemann- 1
t
s  F (s )
( ) 0
Liouville integral

0 I t f (t )  (t   ) 1 f ( )d 

Riemann- 1 d t f (x ) n 1

(n   ) dt a (t  x )  n 1 s  F (s )   s k  RL0 Dt k 1f (t )  , n  1    n

RL
0 Dt f (t )  ( )n dx
Liouville
derivative k 0 t 0

t D f ( )
Caputo derivative n n 1
s  F (s )   s   k 1f
1
Dt f (t )   d (0), n  1    n
C (k )

(n   ) (t   ) n 1
0 a
k 0
Griinwald- n
  s  F (s ), 0    1
Leitnikov 0 Dt f (t )  lim h   ( 1) m   f (t  mh )
h 0
fractional nh t a m 0 m 
derivative10

Numerical solution
Until now, various numerical methods have been proposed for solving fractional differential equations
which are based on discretizing the relationships [12-17]. In this section, a new algorithm which was
recently published in [18] is proposed discretizing the Caputo derivative.
Considering the Caputo derivative of function 𝑢(𝑡) with order 0 < 𝛼 < 1, the following equations hold
for discretizing 𝑢(𝑡) on [0,T] in a uniform way.
1 n
u ( s )

t j 1
t u (tn 1 ) 
(1   ) j 0 t j
(tn 1  s ) 
s
ds

j 1 )  u (t j ) t j 1
1 n u (t (23)
 
(1   ) j 0  t j (tn1  s) ds  r ,
 n 1

1 n u (tn 1 j )  u (tn  j )

 
(2   ) j 0
d , j

 rn,1

Where 𝑑𝛼,𝑗 = (𝑗 + 1)1−𝛼 − 𝑗1−𝛼 and 𝑗 = 0,1,2, … , 𝑛, as well as 𝑟𝛼,𝜏

𝑛+1
is local truncation error.

To utilize the above algorithm, equation (21), which used in the last section simulation, is considered.
The following equation will be obtained by applying the equation (23) on equation (21).
n
x(tn 1 )    (2   )( f  )  x(tn )   d , j ( x(tn 1 j )  x(tn  j )) (24)
j 1

Comparison between the numerical method and the analytical answer is shown in both the figure (1)
and the table (1)

10
The Laplace transform of the Griinwald- Letnikov fractional derivative of order 𝛼 > 1 does not exist in the classical sense.

13
 Fig 1. A schematic comparison of analytical and numerical answer with the represented mesh in Table 1.

Table 1. Comparison between the numerical results, equation 24, and the analytical answers of equation (25) with 𝑓 = 1100𝑝𝑁 stress (𝜉 and
𝛼 are chosen from table 4).
Mesh point (𝝉) Numerical result in 𝒕 = 𝟑 Analytical result in 𝒕 = 𝟑 Difference Percentage error
0.1 848.3334 849.9605 1.6271 0.1914
0.01 849.8002 849.9605 0.1603 0.0188
0.001 849.9445 849.9605 0.0160 0.0018

Viscoelastic systems

Elasticity is the ability of a material to resist a distorting or deforming force and return to its original
form when the force is removed. According to the classical theory in the infinitesimal deformation,
most elastic materials, based on Hooke's Law (Eq. 25), can be described by a linear relation between
the stress 𝜎 and strain 𝜀 [19].
𝜎(𝑡) = 𝐸𝜀(𝑡) (25)

where 𝐸 is a constant known as the elastic or Young's modulus.

The viscosity of a fluid is a quantity which describes its resistance to deformation under shear stress or
tensile stress. In the ideal viscose fluid, according to the Newtonian fluid (Eq. 26) the stress is
proportional to the local shear velocity (i.e. the rate of deformation over time):
𝑑𝜀(𝑡)
𝜎(𝑡) = 𝜂 (26)
𝑑𝑡
where 𝜂 is the shear modulus.
However, many liquids briefly respond like elastic solids when subjected to abrupt stress. On the
contrary, many solids will flow like liquids, albeit very slowly, under small stress. Such materials
possessing both elasticity (reaction to deformation) and viscosity (reaction to the rate of deformation)
are known as viscoelastic materials. When these materials subjected to sinusoidal stress, the
corresponding strain is neither in the same phase as the applied stress (like an ideal elastic material) nor
in the 𝜋/2 out of phase (like an ideal viscose material). In these materials, some part of input energy is
stored and recover in each cycle and the other part of the energy is dissipated as heat. Materials whose
behavior exhibits these characterized are called viscoelastic. If both strain and stress be not only
infinitesimal but dependent upon the time as well, strain-stress relation (constitutive equation) can be
described by a linear differential equation with constant coefficient i.e. the material which exhibits
linear viscoelastic behavior. One of the main consequent of this assumption is the stress (strain)
responses to successive strain (stress) stimuli are additive. In other words, in the creep experiment
consist of applying a step stress 𝜎0 , (a stress increment at time t = 0, which is kept constant for t>0),
and measuring the corresponding stain respond 𝜀(𝑡), the constitutive equation is 𝜀(𝑡) = 𝜎0 𝐽(𝑡), where
𝐽(𝑡)—which termed creep compliance—is the strain at time t owing to a unit stress increment at time
14
0. Suppose that various stress increments Δ𝜎 occur at the successive time interval Δ𝑡 based on the
superposition of effects, we get:

 (t )   0 J (t )   J (t  t ) (27)

After N steps, we have

N
  
 (t )   0 J (t )   J (t  nt )   t (28)
n 1  t t nt 
This relation recognized as Boltzmann superposition principle which states that the response of a
material to a given load is independent of the response of the material to any load, which is already on
the material [19].

Fig 2. A graph versus time of (a) a step-plus increment stress applied to linear viscoelastic material and (b) the resulting strain.

in the limit Δ𝑡 → 0
t d (t )
 (t )   0 J (t )   J (t  t ) dt  (29)
0 dt 
This relation is termed hereditary or superposition integral. Similarly, with applying a step strain, we
get
t d  (t )
 (t )   0G (t )   G (t  t ) dt  (30)
0 dt 
Where G(t) is relaxation modulus.
In the mechanical science, to evaluate mechanical properties of materials, after sudden stress and
dynamic (periodic) experiments, researchers implement certain tests, such as creep tests. In the
following, we will briefly discuss these two tests.

Creep test
One of the most common tests on the viscoelastic system is creep test. In this test, a system is subjected
to constant tension and the amount of deformation is recorded over time. For viscoelastic materials
creep vs. time diagram shows power-law behavior (figure 3).

 (t )   J (t ) (31)

15
Where 𝐽(𝑡) = 1/𝐺(𝑡) is termed creep compliance.

Fig 3. The profile of a shear creep experimental.

Periodic test
In this common experiment, a system subjected to a cyclic applied stress. Therefore, as mentioned
above, for a viscoelastic system, strain in response to an imposed periodic stress of angular frequency
𝜔 is also periodic with the same frequency but in the different phase (figure 4). Based on relation (30),
if the periodic stress 𝜀 = 𝜀 0 sin(𝜔𝑡) applied, we expect that the stress follows a periodic response with
the same frequency and different phase:

   0 cos(t   )   0 cos  sin  t   0 sin  cos  t (32)

And by extension of relaxation modulus, we define two shear storage modulus G’, and shear loss
modulus G” as follows

G   ( 0 /  0 ) cos 
G   ( 0 /  0 )sin  (33)
G  / G   tan 

Fig 4. Geometry and time profile of a simple shear experiment with sinusoidally varying shear [19].

16
Fractional viscoelastic model

There are different models to evaluate and to predict the constitutive equation in the viscoelastic system,
these models commonly, composing of a different combination of springs and dampers elements [20,
21]. For example, two well-known models to name are Maxwell and Kelvin models (Table 2).
In this section, a model which is known as fractional order element (FOE) with equation (39) is
presented in order to investigate the mechanical response of viscoelastic systems from different points
of view. It is indicated that this fractional model can be a good substitute for modeling the viscoelastic
materials than SLE models. With this aim, let us take a closer look at Maxwell model; based on this
model, the constitutive equation will be as follows:
1 𝑑𝜎 𝜎 𝑑𝜀
+ = (34)
𝐸 𝑑𝑡 𝜂 𝑑𝑡
where E(0+ ) = η(0+ ). Also, 𝜎, 𝜀, 𝜂, and 𝐸 are stress, strain, shear modulus, and Young modulus
respectively. The following equation will be obtained by taking Laplace transformation and considering
a step strain function.
𝑡
−( )
𝐺(𝑡) = 𝐸𝑒 𝜏 (35)
Where 𝑡 > 0 and 𝐺(𝑡) is relaxation modulus.
In nature, the 𝐺(𝑡) value in viscoelastic materials does not usually follow a simple exponential behavior,
and a more realistic expression for its behavior is a power-low response [3]:
𝐸 𝑡 −𝛼
𝐺(𝑡) = ( ) (36)
Γ(1 − 𝛼) 𝜏

Table 2. Viscoelastic models, which 𝜎, 𝜀 𝑎𝑛𝑑 𝜏 are stress, strain and 𝜂/𝐸 respectively; also, 𝑋𝛼 = 𝐸1−𝛼 𝜂𝛼 and 0 < 𝛼 < 1.

Name Model Constitutive equation Relaxation modulus

(𝑮(𝒕))
Maxwell fluid  (t )   D  (t )   ED  (t ) Ee t /

Voigt fluid  (t )  E  (t )   D  (t ) E   (t )
 (t )  X  D   (t )

E t 
 
FOE (1   )   

As it mentioned in the previous section, the value of stress in a pure elastic system, an ideal spring, is
proportional to the derivative zero displacements. On the other hand, it is proportional to the first-order
derivative of displacement in the pure viscous system. It is expected that the viscoelastic systems have
a behavior between these two systems, an elastic and viscous material, resulting to conclude that stress
is proportional to a fractional derivative whose order is between zero and one (Fig 5). Hence, from the
mathematical point of view, this physical interpretation can be modelled as the following equation.
𝜎(𝑡) = 𝑋𝛼 𝐷𝛼 𝜀(𝑡) (37)

Where 0 ≤ 𝛼 ≤ 1 and 𝑋𝛼 are the constant coefficients to make an equal relation between stress and
strain—the 𝛼 index represents the dependence of the constant 𝑋 to expanded order.
17
 Fig 5. A schematic representation of viscoelastic regime and stress-strain relation from a physical point of view.

To obtain the coefficients, the essential condition of fractionalization algorithm (section 2) is applied
first. Thus, the equation (41) must contain both Hook’s law and Newton fluid in boundary points, i.e.
𝛼 = 0 𝑎𝑛𝑑 𝛼 = 1. In other words

lim X   E , lim X    (38)

 0  1

One of the specific cases which can satisfy the mentioned boundary condition is 𝑋𝛼 = 𝐸1−𝛼 𝜂𝛼 . In what
follows, it is shown that the proposed 𝑋𝛼 is a good suggestion. To make a sufficient condition in
fractionalization algorithm, the physical point of view must be considered and to this end, the equation
(37) with the assumed coefficients is rewritten.
𝑑 𝛼 𝜀(𝑡)
𝜎(𝑡) = 𝐸1−𝛼 𝜂𝛼 (39)
𝑑𝑡 𝛼
By applying the Laplace transform to the above FOE the following equation (40) will be obtained—
derivative is Caputo.

 ( s)  E1  s  (s) (40)

As a result, the relaxation modulus will be achieved by considering a unit step strain function.
𝐸 𝑡 −𝛼
𝐺(𝑡) = ( )
Γ(1 − 𝛼) 𝜏
Where 𝜏 = 𝐸/𝜂. Thus, we gained the same equation (40) by the help of physical interpretation and
mathematical methods which have better match with the experimental results.
In equation (43), 𝛼 and 𝑋𝛼 are quantities which represent the system property. In other words, it could
be possible to consider the 𝐸1−𝛼 𝜂𝛼 coefficient as an independent quantity which is introduced by
variable 𝑋𝛼 —it means that whatever determine the quantity and the structure of the system are just the
value of 𝛼 and 𝑋𝛼 ; indeed, it is possible that a system with 𝐸 ′ and 𝜂′ modulus has equal dynamical
behavior as another system which has different modulus as 𝐸 and 𝜂 i.e. 𝜂 ≠ 𝜂′ and 𝐸 ≠ 𝐸 ′ , but
𝐸 (1−𝛼) 𝜂𝛼 = 𝐸′(1−𝛼) 𝜂′𝛼 . Hence, the product of two moduli which expressed in equation (40) will be
just one independent quantity.
If 𝛼 → 0, the equation (40) will lead to the Hook’s law and if 𝛼 → 1, then the equation turn to Newton
fluid model. As a result, based on the physical interpretations, it looks logical to conclude that the
proposed model can be a proper tool for investigating the mechanical properties of viscoelastic
materials.
To take a more precisely look at the FOE equation in the viscoelastic systems, in what follows, the
formalism of inheritance integral for the fractional state is investigated. Equation (41) is obtained by
taking partial integral from equation (29).
t dJ (t  t )
 (t )   (t ) J (0 )    (t ) dt  (41)
0 d (t  t )

18
Also, the creeping state with unit step stress function will be as follow based on FOE equation.

1 (t /  )
J (t )  (42)
E (1   )
By taking derivative from equation (42) and putting the obtained result in equation (41), equation (43)
will be achieved.

1  1 t 
 ( ) 0 (t  t )  (t )dt 
 1
 (t )   (t ) J (0 )  (43)
E   
Which the equation inside the bracket is Riemann-Liouville integral; thus, by rewriting the above
equation, we have

1
 (t )   (t ) J (0  )  
0 I t  (t )
(44)
E 
According to the abovementioned subjects, in equation (44), the previous events are embedded in the
fractional kernel function 𝑡 𝛼−1 /Γ(𝛼); which the obtained result has match with the proposed
interpretation of section (3). Regarding those interpretation, the information of a system, which is
related to deformation, are storing with a power-law function which has fractional order 𝛼 from the
initial step (𝑡 + = 0) until the present moment (𝑡)—which the fractional order is the indication of saving
order information. According to the kernel function 𝑡 𝛼−1 /Γ(𝛼) in equation (44), 𝛼 = 0 and 𝛼 = 1 are
equal with a memory-less system and full-memory system in creeping state respectively owing to
Γ(0) = ∞. Indeed, in a pure viscose system, system’s position completely depend on this fact that how
stress applies on the system over time. In other words, the nature of stress and the amount of which has
a direct impact on the present position of the system. However, owing to this fact that the effect of
memory at different times has a unit weight function, the system is able to save all the prior history of
itself precisely and equally. In contrast, in the pure elastic system, the present stress is an effective
element on the amount of displacement, not previous stresses, because the weight function is zero.
Similarly, the above procedure can be followed for relaxation modulus. The following equation will be
obtained by putting equation (36) in (30).

E  t d  (t )
 (t )   0G(t )  
(1   ) 0
(t  t ) 
dt 
dt  (45)

As a result, by the help of equation (15) and rewriting the above equation, constitutive equation is gained
as follow.

 (t )   0G(t )  E  C0 Dt  (t )
Based on the equation (45), in the relaxation state, 𝛼 = 0 and 𝛼 = 1 represent a full-memory system
and a memory-less system respectively—Γ(0) = ∞. In other words, in the pure elastic systems, the
displacements of an object in prior times completely keep in the memory of object owing to Because
of storing displacement changes over time by an unit function.—known as weight function—over time,
and there is no differences in very far and very near time. As a result, all the previous displacements
will have been an effect on the value of stress. In the pure viscose system, however, such a memory as
the mentioned one does not exist, as well as the previous displacement will not have impacted on the
present stress of system; thus, the system is absolutely devoid of any memory in such condition. This
is a really intriguing, fascinating interpretation of FOE model with equation (39) because on the one,
19
hand the fractional order 𝛼can represent a system which is devoid of any memory, i.e. a memory-less
system, when the order is zero, but on the other hand, it can be the representation of a full-memory
system.
It is indicated that the fractional models are a valuable tool for describing the dynamic properties of real
materials, specifically, in polymers which use for controlling the sounds and vibrations [22-33]. Also,
the combination of fractional models with other linear elements can make intriguing structures; for
example, it can be referred to fractional Maxwell and Kelvin models which are known as spring-spot
models (Table 2). Since SLEs are not always precisely successful in investigating many viscoelastic
systems, researchers applied spring-pot models to address the problem, as well as by virtue of these
fractional models and the combination of them, they achieved the astonishing results [33-42]—to more
details see the “Application” section.

Table 3. Fractional order Voigt Models.

Model Constitutive equation Relaxation modulus (𝑮(𝒕))


E1  t 
 (t )  E11 D  (t )  E  (t )
    E
(1   )   1 


E1  t 
 (t )  E11 D  (t )  D  (t )
      (t )
(1   )   1 

 
E1  t  E1  t 
 (t )  E11 D  (t )  E 2 2 D  (t )
        
(1   )   1  (1   )   2 

Table 4. Fractional order Maxwell Models

Model Constitutive equation Relaxation modulus (𝑮(𝒕))

  t   
EE      
 (t )   D  (t )  E  D  (t )
   
   1  

 1
t   t  
E  E 1 ,1     
 (t )   1
D 1
 (t )  2 D  (t ) 
      


t    t    
E  E    ,1      
 (t )    
D  
 (t )  E  D  (t )
 
      

Assuming the fractional Maxwell model which made of two FOE models with (𝛼, 𝐸1 , 𝜏1 ) and (𝛽, 𝐸2 , 𝜏2 )
elements, where 0 < 𝛼, 𝛽 < 1. Regarding this, the constitutive equation, in relaxation state, will be as
follow [to prove, see Appendix]:

t   t    
G (t )  E   E   ,1      (46)
      

20
Where 𝐸𝛼,𝛽 is Mittag-Lefler function [to see the definition of this function, see Appendix]. Above
equation will be reduced to the relaxation modulus of classical Maxwell by assuming 𝛼 = 0 and 𝛽 = 1.
In other words, this equation includes not only the classical relaxation state but also further expansion
as compared to the SLEs (more fractional models is presented in the table 3 and 4).

Periodic strain
As previously detailed, by imposing a periodic strain upon a viscoelastic object, stress is neither exactly
in the same phase with strain nor 90o out of phase. In this section, the trueness of this fact is investigated
for a FOE model. Assuming a FOE is subjected to a sinusoidal strain, i.e, 𝜀 = 𝜀 0 sin(𝜔𝑡), then the stress
of which will be obtained as follow by the help of equation (39) and Caputo derivative (15).

  t 
 (t )  E 1   0
  (1   ) a
(t   )  cos( )d  
 )74(
As a result, the following equation hold [to prove the equation, see Appendix].
   
 (t )   0 E 1    i sin(t   ) exp  i (2  1) 
2  2  )74(
In equation (74), if 𝛼 = 1, stress will be 90o out of phase with strain as a pure viscous object, and if
𝛼 = 0, it will be exactly in the same phase with strain as a pure elastic object. Therefore, when 0 <
𝛼 < 1, then stress will neither exactly in the same phase with strain nor 90o out of phase. Hence, the
result of equation (74) thoroughly matches with physical interpretations.

Relation between fractional equations and SLE

In this section, with the help of a new fractal system which made of quite a few springs and dampers
(Fig 6), the relation between fractional equations and SLEs is investigated.
Increasing the number of elements in the SLEs is one approach to improving the accuracy of a model
in the viscoelastic systems. Calculation of and work with such sophisticated systems are not usually
easy. Thus, in order to investigate the constitutive equation in infinite systems, which made of many
both springs and dampers, different models were derived so far, and their outputs were reduced to
fractional equations [24,48,49]. Regarding this, a new fractal system is considered and after massive
mathematical calculation, it reveals that the constitutive equation is equal to a spring-pot11.
Consider the spring-damper circuit in Fig 6, so equation (74) will be obtained with the help of governing
equations on the stress/strain elements.
𝜎̅0𝑠 (𝑠) 1
= 𝜂́ 0 (1 − ) (74)
𝜀̅0𝑑 (𝑠) 𝐸 𝜂́
1+ ́ 0 (1+ 1 (1−
1
)
𝜂1 𝐸1 𝐸 𝜂́ 1
1+ 1 (1+ 2 (1− ⋮ )
𝜂́2 𝐸2
𝐸 𝜂́𝑛−1
1+ 𝑛−2 (1+ )
𝜂́𝑛−1 𝐸𝑛−1 +𝜂́𝑛

Where 𝜀̅(𝑠) = 𝐿{𝜀(𝑡)} , 𝜎̅(𝑠) = 𝐿{𝜎(𝑡)}.

11
From the mathematical point of view, by imposing a local force to the proposed system, an equivalent topology for investigating the relationship between stress
and strain over time is found based on the mathematical formulae. In other words, the govern equation in figure 9 is in R space with the Euclidean norm, while
the output of the assumed system is located in 𝐿1 space.
𝑎
𝐿1 ≔ {𝑓: [𝑎, 𝑏] → ℝ; 𝑓 𝑖𝑠 𝑚𝑒𝑎𝑠𝑢𝑟𝑎𝑏𝑙𝑒 𝑜𝑛 [𝑎, 𝑏] 𝑎𝑛𝑑 ∫ |𝑓(𝑥)|1 𝑑𝑥 < ∞ }
𝑏

21
Based on equation (05), after rewriting equation (74) to (05), the following equation is obtained12.
1.(0+𝛼) 1.(2−𝛼) 2.(1+𝛼) 2.(3−𝛼)
𝑥 (1−𝛼)𝑥 𝑥 2.3 𝑥 3.4 𝑥 4.5 𝑥
𝑥(𝑥 + 1)𝛼−1 = 1.2
… (05)
1+ 1+ 1+ 1+ 1+ 1+
(𝑛−1)(𝑛−𝛾)
𝜎̅0𝑠 (𝑠) 𝑐0/𝑠 (1−𝛾)𝑐0/𝑠 (2𝑛−1)(2𝑛−2)
𝑐0/𝑠
= 𝜂0́ (1 − ⋯ ) (02)
𝜀̅0𝑑 (𝑠) 1+ 1+ 1

Where 0 < 𝑥, 𝛼 < 1 and 𝑐0 = 𝐸0 /𝜂0 . The equation (05) will be reduced to the following equation by
assuming 𝑐0 ≪ 𝑠 when 𝑛 → ∞.
𝜎̅0𝑠 (𝑠)
≈ 𝑠𝜂0 (1 − (𝑐0 /𝑠)𝛾 )
𝜀̅0𝑑 (𝑠)

Fig. 6 Diagram of the finite mechanical arrangement applied to simulate the generalized damper.

By assuming 𝐸0 = 0 and rewriting the above equation, we have [to follow the proof in detail, see
Appendix]:
𝑑𝜀𝑑 (𝑡) 𝑑 1−2𝛾 𝜀𝑑 (𝑡)
𝜎 𝑑 (𝑡) = 2𝜂0 − 𝜂0 (07)
𝑑𝑡 𝑑𝑡 1−2𝛾

In other words, the output of the considered system is a combination of a FOE and damper connecting
1
in a series way when 0 < 𝛾 < . Also, based on the achieved result, derivative in the above equation is
2
Caputo.

12
Notations
𝑎1 𝑎1 𝑎2 𝑎𝑛
𝑎2 = ………
𝑏1 + 𝑏1 + 𝑏2 + 𝑏𝑛
𝑎3
𝑏2 + 𝑏3 + ⋱
𝑎𝑛
+
𝑏𝑛

22
Application

As the abovementioned sections, fractional calculus is a powerful tool for modeling complex systems,
specifically for viscoelastic materials. In this section, at first, a brief review of the application of
fractional calculus in the modelisation of viscoelastic materials in the variety fields are presented. In the
next step, a one-dimensional fractional model for investigating a cell deformation in creeping state is
proposed, and then a 2D fractional model for simulating the network of actin filaments on the cellular
surface is proposed. Owing to the perfect match with experimental data and simulated model, we
logically conclude that fractional models can be a great replacement for previous models, which are
made of spring and damper, in this regard.
Fractional calculus has been employed for modelling viscoelastic systems which cover many fields and
subjects. In ref [43], Djordjević, V.D., et al. utilized fractional calculation in four parametric model
with equation (05) to investigate a cell’s viscoelasticity in the range of 10−2 − 102 𝐻𝑧 for a periodic
test. The output of this model was compared with experimental data obtained from the magnetic
oscillatory cytometry method, and it revealed that the proposed model highly matched with
experimental data.

 (t )  a0 (t )  a1D  (t )  a2 (t ) )05(

It is noteworthy Riemann-Liouville (fractional) derivative was considered in equation (05).

Investigation of asphalt mixtures’ property during their service life is a real challenge owing to its
complexity and sensitivity to environmental and loading conditions. It has been shown that asphalt
mixtures behave as linear viscoelastic materials when it is subjected to loading conditions.
Traditionally, the viscoelastic material is modeled via creep/recovery functions by the help of spring
and damper models. The output of these models have shown an exponential behavior, in nature,
however, the mechanical response of these materials have a power-law behavior rather than an
exponential. Considering this fact, spring-pot models were proposed for predicting creep/recovery
behavior of asphalt mixtures. Finally, the result of proposed models were compared with experimental
data, showing that they have in good agreement with the data [39].
Scholars believe that some pulmonary diseases are related to lung viscoelasticity. Thus, profound
understanding of viscoelastic models in this field can bring about a new progress in lung pathology and
trauma. As a result, stress relaxation test was applied on a pig’s lung, and the output of this test showed
a power-law behavior than an exponential. Regarding this, two integer standard linear solid—
generalized Maxwell models—and a Fractional standard linear solid model were proposed for
predicting the mechanical behavior of a system. The result of this investigation indicated that the
fractional model by far better matches with experimental data than integer models [41].
To investigate the viscoelastic response of human breast tissue cells, the fractional Zener model was
recently proposed [44]. The authors demonstrated that the proposed fractional model has a better result
as compared to that of the non-fractional models showed for probing the mechanical response in
relaxation state.
According to the recent studies [42], fractional Maxwell and Kelvin model were proposed for analyzing
and interpreting the mechanical response of polymer materials, and the result of which have shown that
the fractional Maxwell model was not able to investigate the polymer behavior. On the other hand, the
fractional Kelvin model is a good proper model with this aim [42].
In 2008, a fractional model was utilized to investigate the arterial viscoelasticity. In that research, the
output of the fractional model and SEL model was pinpointed by the help of least-squire. The result
23
that validation showed that the fractional model is able to analyze the system mechanical response to
tensile perfectly [35].

Two-dimension fractional network model

In the last section of the paper, to investigate the mechanical response of a cell, two (new) fractional
models are suggested. In the first step, a one-dimension fractional model is proposed, and in the second,
a more realistic model is put forward in two dimensions by extending the first model.
We define a cell as the smallest vital unit of the configuration of any live existence. The cell is made of
different parts, such as cytoplasm, nucleus, cytoskeleton (CSK), etc. The CSK plays a remarkably
essential role in the intracellular force transmission, cellular contractility, as well as the structural
integrity of cells not only in the static but in dynamic states as well. The biological functions of cells,
such as growth, differentiation, and apoptosis are associated with changes in the cell shape, are related
to the mechanical behavior of the CSK [45-47]. The CSK structures are composed of three main
elements: actin filaments, intermediate filaments, and microtubules. The structural changes in the CSK,
including deformation and rearrangement, are related to changes in the tensile forces that are applied to
the actin filaments [45, 46, 48-50]; as a result, the actin filaments have mainly tension-bearing roles.
The viscoelastic properties of cells have been found to play an important role in distinguishing
diagnostic, stem cells and etc. Regarding this, several two-dimensional (2D) and three-dimensional
(3D) CSK network models have been proposed to calculate the mechanical response of cells [56,58-
61].
The viscoelasticity of a cell under some specific external forces was measured with the help of Magnetic
Bead Microrheometry methods in creeping state with 𝐹 = 1100𝑝𝑁 [51]. To investigate the gained
results, the author utilized a model which made of two springs and two dampers (Fig 7). The result of
this model is compared with the experimental data (Fig 8).

Fig. 7. Bausch model has been proposed to investigate a cell viscoelasticity under deformation [51].

The mechanical response of a living cell is quite intricate. The complex, heterogeneous characteristics
of cellular structure cause that simple linear models of viscoelasticity cannot predict the mechanical
properties of cells under the deformation quite well, especially under minuscule deformation. Regarding
this, fractional models are applied to address the point mentioned. In the first step, a FOE model with
Eq. 39 is proposed for modeling a cell deformation in the creeping state. To validate the proposed
model, the model result is compared with the experimental data which expressed in the ref [51]. In order
to make the same condition as experimental data, after solving equation (39) by regarding the creep
test, equation (22) is obtained. According to Fig 8, it is clear that the FOE model is a great agreement
with the experimental data. By the help of simulated annealing optimization methods, the best
coefficients—𝛼, 𝐸, 𝑎𝑛𝑑 𝜂—in equation (22) is gained as well (Table 5). Hence, it looks logical to
conclude that the fractional model can be a suitable replacement for modeling the mechanical properties
of a cell than the Bausch model.

24
 Fig 8. Comparison between the FOE model and the Bausch model. The red points and the black line represent the experimental and the
response of the FOE models respectively. Which the right side related to the FOE model and the left side represents the Bausch model.
Table 5. The optimized coefficients of the FOE model which obtained by annealing optimization methods.

Fractional order (𝛼) Young’s modulus (Pa) Shear modulus (𝑃𝑎. 𝑠 −𝛼 ) MSE

0.112889249504105 0.003456900326295 1.207297449819304e-6 69.1723

To investigate the models’ error with experimental data, normalized root-mean-square deviation
(NRMSD) method is applied.

MSE
NRMSD  100
y max  y max

Also, NRMSE was calculated 10.21% and 3.16% for Bausch model and FOE model by the help of
above formula respectively, showing the fractional model is highly accurate for probing and predicting
the mechanical properties of a subjected cell.

Fig 9. A schematic representation of the preposed network model. Which the black points are fixed and, the others are dynamic. Also, the
lines between two points represent the actin filaments connecting by cross-linking protein (in cell cortex).

In the next step, a two-dimension fractional model, so-called 2D fractional network model, is proposed
to investigate the cell viscoelasticity. The cell can deal with the external and intercellular forces imposed
on it by virtue of CSK. Regarding this, these protein filaments, CSK, is considered as a started point of
making a 2D model.
In most cells, actin filaments are intensely concentrated in a layer just beneath the plasma membrane,
even though they are found throughout the cytoplasm of a eukaryotic cell. In this region, which is called

25
the cell cortex, these filaments are linked by actin-binding proteins into a meshwork that gives
mechanical strength to cells and supports the outer surface of them as well. As a simplification, without
loss of generality, imaging the whole mechanical properties of a cell is related to actin filaments [52].
Considering this, we can propose a dynamic network model, shown in Fig. 8, for probing the cell
mechanical behavior.
In 2003, a model as the proposed network model in this paper was suggested predicting the mechanical
response of a cell [53]. The mechanical properties of actin filaments were approximated with pure
elastic materials based on an experimental paper [54]; however, according to the experimental data,
these filaments have shown an elastic response in millisecond scale under deformation. Neglecting this
scale as a simplification may cause a remarkable error in long time simulation. In other words, the
deformation of any material under minuscular stress is divided into three phases which the first one is
related to the elastic limit. Also, the mechanical response of an actin filament was reported in this phase,
and by overgeneralizing this point, a pure elastic network model—which every actin was approximated
with a spring—was proposed in the scale of sec for probing the whole mechanical behavior of a cell
[53]. Consequently, the elastic network model, so-called pre-stressed cable network model, was failed
to predict the experimental data from bead micrometry method [52]. Regarding this, a new model with
FOE elements is suggested in this paper. Two significant merits of which, connected to the system
memory and the viscosity term in arbitrary scale time, are embedded in proposed model; resulting the
fractional model is able to highly match with the experimental data. The fractional network model is
based on the principle that actin filaments, according to ref [52, 55], have shown viscoelastic behavior.
Therefore, the FOE model can be a proper model to investigate the mechanical response of each actin
filament. By expanding the FOE model into a 2D model with the same order to each element, the
fractional network model, which depicts in the Fig. 9, will be obtained.

Fig 10. A schematic representation of 2D dynamic fractional simulated network.

Since the direct observation of a 𝑙0 (the length of an actin filament) is not exist from a living cell, 𝑙0 is
approximated from CSK pore (𝐷𝑝 ) in an adhere cell based on equation 𝐷𝑝 = √3𝑙0 /3 as paper [53]. In
other words, based on the observations, the value of 𝐷𝑝 is considered in the middle of experimental
data—that is Dp = 100nm; resulting l0 = 173.2nm. The space of network is approximated 4.5464 ×
106 nm2 based on 𝑙0 , which is almost same with the reported value in the ref [53].
To validate the simulation result and to compare it with the previous model, we simulate the FOE and
spring network, Fig 11 show the behavior of these two model under the same stress. Considering the
26
results, the fractional model is better matched with the experimental data. Hence, it seems that we could
conclude that 2D fractional network model is a good candidate for modeling and predicting CSK in
cells.

Fig 11. A comparison between the 2D dynamic fractional simulated network, the left side figure, and the elastic model which represented in
ref [53], the right side figure. Experimental data is obtained from ref [51].

Discussion

The fractional calculus is a powerful tool for describing the complex physical systems which have long-
term memory and long-term spatial interactions. In this paper, different aspects of this powerful tool
are introduced through a concise literature review to provide the reader with a picture of what fractional
calculus is. Then, with the help of developed concepts, it is sought to find a relation between fractional
and viscoelastic systems and to employ the obtained relation for investigating the mechanical property
of a cell.
In the first step, to investigate the mechanical response of elastic, viscous, and viscoelastic materials,
scholars typically apply spring, damper models and some combination of them respectively. In nature,
pure elastic or viscous systems can rarely found and moreover, the mechanical behavior of them are
modelled by the help of the Hook’s law and Newtonian fluid. Regarding a minuscule deformation, the
response of a material is modelled by these laws, thereby making some errors in both the simulation
and the calculation of theoretical models—specifically, bio-material such as cells and tissues. In spite
of neglecting the arisen errors in some cases, scholars are usually making a highly complex model for
improving the accuracy and reducing the error in minute systems. Considering this fact, proposing a
new method to cope well with the sophisticated problem can be a worthwhile tool. Based on the
investigation which is done in this paper, it is indicated that fractional calculus is a suitable candidate
to this end. To prove this claim, a fractal system, made of the combination of quite a few springs and
dampers on a large scale, was considered, as well as the output of the proposed system was
approximated only with the simple fractional model.
In the next step, a version of fractional models, known as FOE, was employed to model the mechanical
response of a cell in one-dimension. By virtue of the achieved result, a 2D model so-called 2D fractional
network model was proposed to investigate the mechanical response of actin filaments, exist in CKS
and cover a cellular outer layer. The creeping state was considered in both models, and the result of
which was validated by not only the experimental date but previous models as well. The results revealed
that in the one-dimensional model, data variance relative to a previous model presented in ref [51] was
about 11%, while in the FOE model, it was estimated 3%. In other words, the FOE model precision has
evaluated 97% in comparison with experimental data. Furthermore, the simulation of two-dimension
elastic lattice, represented in ref [53], for actin filaments on the cell surface, is totally invalid in

27
Magnetic Bead Microrheometry method. Meanwhile, the 2D fractional network model has able to
analyze same experimental date precisely and to consider the viscosity term of actin filament as well.
The fractional order 𝛼 in both models is about 0.11. Hence, not only does the investigated model shows
more elasticity behavior, but it also has high memory. As a result of done research the paper and many
other studies, applying the fractional tool to investigate and to model the mechanical response of
phenomena which shows power-law behavior is more efficient than the SLE models.

Appendix

To facilitate the context for readers, the proof of some equations presenting in the text is presented in
this part.

The proof of Equation 46:

According to equation (39)

d    1  d    2 
1    and 2   
dt   E 1 1  dt    E 2 2 
where the initial value of 𝜀1 (𝑡) and 𝜀2 (𝑡) is assumed zero. Considering 𝜀 = 𝜀1 + 𝜀2 based on Maxwell
model, we have
1 d   1 1 d  2
   (t )
E 1 1 dt  E 2 2 dt  
In addition, 𝜎 = 𝜎1 = 𝜎2 and 0 < 𝛼, 𝛽 < 1 then
E   d     2 (t )  d  (t )

 1  1 1  E 
E 2 2 dt    dt 
1 1

𝛽
Now let 𝜏 𝛼−𝛽 = 𝐸1 𝜏1𝛼 /𝐸2 𝜏2 and E  E 1 ( 1 /  ) , so
d     (t ) d   (t )
 (t )       
 E 
dt dt 
Taking the Laplace transform and applying the unit step strain function on the above equation, we
finally gain the following equation

E   s  1  
g (s ) 
s     (1/  )  
Hence, the following equation by utilizing the inverse Laplace on the above equation is achieved.
  
t   t  
G (t )  E   E    ,1      
      

The proof of Equation 48:

By taking inverse Laplace from equation (47) and by the help of convolution operation (see table
B3.1), the following equation is obtained
 0
 (s )  E 1  F (s )G (s ) s .t F (s )  L {t  },G (s )  L {cos(t )}
(1   )

according to the table (B3.1)

s
 (s )   0 E 1 
s  2
2

By taking inverse Laplace from the above equation, we have

28
 s 1 
 (t )   0 E  L { 2 } 1
(54)
s  2

In order to solve the above equation, it is enough to calculate L1{ 2 s 2 } . Thus, according to the
s 
(B3.2) equation
s 1 c i  s
L1{
s  2
2
}
2 i  c i 
(
s  2
2
)e st ds

By assuming ℎ(𝑧) = 𝑧 𝛼 𝑒 𝑧𝑡 /(𝑧 2 + 𝜔2 ) and Cauchy theorem [56]

s
L 1{ 2 }  h (z 1 )  h (z 2 )
s  2
Where 𝑧1 = 𝑧2 = ±𝑖𝜔 are non-analytical points of function ℎ(𝑧), resulting
(i ) it (i ) it
h( z1 )  h( z2 )  e  e
2i 2i
(i ) it
 (e  (1) e  it )
2i
  1 i 2 ( 1) it i  it
 e (e e e )
2
 
  1 i (t  ( 1)) i ( t  (3 1))
 (e 2
e 2
) (55)
2
And also, we have
 
i (t  ( 1)) i ( t  (3 1))
(e 2
e 2
)
   
 cos(t ) cos( (  1))  sin(t ) sin( (  1))  cos(t ) cos( (3  1))  sin(t ) sin( (3  1))
2 2 2 2
     
i sin(t ) cos( (  1))  cos(t ) sin( (  1))  (cos(t ) sin( (3  1))  sin(t ) cos( (3  1))) 
 2 2 2 2 
       
 cos(t )  cos( (  1))  cos( (3  1))   sin(t )  sin( (  1))  sin( (3  1)) 
 2 2   2 2 
        
i sin(t )  cos( (  1))  cos( (3  1))   cos(t )  sin( (  1))  sin( (3  1))  
  2 2   2 2 
       
 2 cos(t )  sin( (2  1)) sin    2sin(t )  sin( (2  1)) cos  
 2 2   2 2 
        
i  2sin(t )  cos( (2  1)) cos    2 cos(t )  sin  cos( (2  1))  
  2 2   2 2 
    
 2sin( (2  1)) cos(t ) sin   sin(t ) cos  
2  2 2 
    
i 2 cos( (2  1)) sin(t ) cos   cos(t ) sin  
2  2 2 
     
 2sin( (2  1)) sin(   t )  i  2 cos( (2  1)) sin(t   ) 
2 2  2 2 
    
 2sin(t   )   sin( (2  1))  i cos( (2  1)) 
2  2 2 
   
 2i sin(t   ) exp  i (2  1)  (56)
2  2 

According to (50), (56), and equations (57)

   
 (t )   0 E 1    i sin(t   ) exp  i (2  1) 
2  2 

29
The proof of Equation 52:

According to figure (6) the following equations hold.

𝜎𝑘𝑑 = 𝜎𝑘+1
𝑑
+ 𝜎𝑘𝑠 ∶ 𝑘 = 0,1, … . , 𝑛 − 1 (57)
𝑠 𝑠 𝑑
𝜀𝑘 = 𝜀𝑘+1 + 𝜀𝑘+1 ∶ 𝑘 = 0,1, … . , 𝑛 − 2 (58)
𝜀 = 𝜀0𝑠 + 𝜀0𝑑 (54)
𝑠
𝜀𝑛−1 = 𝜀𝑛𝑑 (65)
𝜎 = 𝜎0𝑑 (65)
Regarding Hook’s law and Newtonian fluid, we have
𝑑𝜀𝑑 (𝑡)
𝜎 𝑑 (𝑡) = 𝜂 (62)
𝑑𝑡
𝑠 (𝑡) 𝑠 (𝑡)
𝜎 = 𝐸𝜀 (65)
where 𝜎 𝑑 ،𝜀 𝑑 ،𝜎 𝑠 and 𝜀 𝑠 related to stress and strain in damper as well as stress and strain in spring
respectively. Accordingly, the following equations is obtained by taking Laplace transform from
equations (62) and (65).
𝜎̅ 𝑑 (𝑠) = 𝜂𝑠𝜀̅𝑑 (𝑠) (67)
𝜎̅ 𝑠 (𝑠) = 𝐸𝜀̅𝑠 (𝑠) (60)

Based on equation (54)

𝜎𝑘𝑑 (𝑡) = 𝜎𝑘+1𝑑 (𝑡)

+ 𝜎𝑘𝑠 (𝑡)
𝑑 𝑑 (𝑡)
(6) 𝑑𝜀𝑘 (𝑡) 𝑑𝜀𝑘+1
→ 𝜂𝑘 = 𝜂𝑘+1 + 𝜎𝑘𝑠 (𝑡)
𝑑𝑡 𝑑𝑡
(9)
→ 𝜂𝑘 𝑠𝜀̅𝑘𝑑 (𝑠) = 𝜂𝑘+1 𝜀̅𝑘+1 𝑑 (𝑠)
+ 𝜎̅𝑘𝑠 (𝑠)
𝑑 (𝑠) (9)
𝜎̅𝑘𝑠 (𝑠) 𝜀̅𝑘+1 𝑑 (𝑠)
𝜎̅𝑘+1
= 𝜂 𝑘 𝑠 − 𝜂 𝑘+1 𝑠 =
⏞ 𝜂 𝑘 𝑠 −
𝜀̅𝑘𝑑 (𝑠) 𝜀̅𝑘𝑑 (𝑠) 1
( ) 𝜎̅ 𝑑 (𝑠)
𝜂𝑘 𝑠 𝑘
1
= 𝜂𝑘 𝑠 −
1 𝜎̅ 𝑑 (𝑠)
( ) 𝑑𝑘
𝜂𝑘 𝑠 𝜎̅𝑘+1 (𝑠)
Thus, we have
(1)
̅𝑘𝑠 (𝑠)
𝜎 1 1
= 𝜂𝑘 𝑠(1 − )=
⏞ 𝜂𝑘 𝑠(1 − ̅ 𝑠 (𝑠)
) (66)
𝜀̅𝑘𝑑 (𝑠) 𝜎̅𝑑
𝑘 (𝑠)
𝜎
1+ 𝑑𝑘
𝑑
̅ 𝑘+1 (𝑠)
𝜎 ̅ 𝑘+1 (𝑠)
𝜎

In addition, we have
(10) (2) 𝑠 𝑑 (𝑠)
̅𝑘𝑠 (𝑠)
𝜎 𝐸𝑘 𝜀̅𝑘𝑠 (𝑠) 𝜀̅𝑘+1 (𝑠)+𝜀̅𝑘+1
𝑑 (𝑠) =
⏞ 𝑑 (𝑠) ⏞ 𝐸𝑘 (
= 𝑑 (𝑠) )
̅𝑘+1
𝜎 ̅𝑘+1
𝜎 ̅𝑘+1
𝜎
(9) 𝑠 𝑑 (𝑠)
𝐸𝑘 𝜀̅𝑘+1 (𝑠)+𝜀̅𝑘+1
=
⏞ ( 𝑑 (𝑠) )
𝜂𝑘+1 𝑠 𝜀̅𝑘+1
Therefore, the following equation holds
(10) 𝑠 (𝑠)
̅𝑘𝑠 (𝑠)
𝜎 𝐸𝑘 1 ̅𝑘+1
𝜎
𝑑 (𝑠) =
⏞ (1 + 𝑑 (𝑠) ) (64)
̅𝑘+1
𝜎 𝜂𝑘+1 𝑠 𝐸𝑘+1 𝜀̅𝑘+1
If we put equation (64) into equation (66), we will obtain
̅𝑘𝑠 (𝑠)
𝜎 1
= 𝜂𝑘 𝑠(1 − ̅ 𝑠𝑘+1 (𝑠)
) (64)
𝜀̅𝑘𝑑 (𝑠) 𝐸𝑘 1 𝜎
1+ (1+ )
𝜂𝑘+1 𝑠 𝐸𝑘+1 𝜀̅𝑑 (𝑠)
𝑘+1
In the step 𝑛𝑡ℎ, the following equation will be got

30
 𝑠 (𝑠) (9) (1)
𝜎̅𝑛−1 1 1
𝑑 (𝑠) =⏞ 𝜂𝑛−1 𝑠 𝑑 (𝑠) =
⏞ 𝜂𝑛−1 𝑠
𝜀̅𝑛−1 𝜎̅𝑛−1 𝜎̅ 𝑑 (𝑠)
1 + 𝑠𝑛
𝜎̅ 𝑠 (𝑠) ( 𝜎̅𝑛−1 (𝑠))
( 𝑛−1 )
(9,10) (4)
1 𝜂𝑛−1 𝑠
=
⏞ 𝜂𝑛−1 𝑠 𝑑 (𝑠) =⏞ 𝜂 𝑠
𝜂 𝑠 𝜀̅ (1 + 𝑛 )
1 + 𝑛 𝑠𝑛 𝐸
( 𝐸𝑛−1 𝜀̅𝑛−1 (𝑠)) 𝑛−1
Thus
𝑠 (𝑠)
̅𝑛−1
𝜎 𝜂𝑛−1 𝑠
𝑑 = 𝜂 𝑠 (45)
𝜀̅𝑛−1 (𝑠) (1+ 𝑛 )
𝐸𝑛−1
Assuming 𝜂𝑘 𝑠 = 𝜂𝑘́ to simplification, so based on equation (64) and (64), in step (𝑛 − 1)𝑡ℎ we will
have
𝜎 𝑠 (𝑠)
̅𝑛−2 1
𝑑 = 𝜂 ́ 𝑛−2 (1 − 𝐸 𝜂́𝑛−1
) (41)
𝜀̅𝑛−2 (𝑠) 1+ ́ 𝑛−2 (1+ )
𝜂𝑛−1 𝜂́
𝐸𝑛−1 ( 𝑛 +1)
𝐸𝑛−1
By rewriting the above equation
𝑠 (𝑠)
̅𝑛−2
𝜎 1
𝑑 = 𝜂 ́ 𝑛−2 (1 − 𝐸 𝜂́
) (42)
𝜀̅𝑛−2 (𝑠) 1+ ́ 𝑛−2 (1+ 𝑛−1 ́ )
𝜂𝑛−1 𝐸𝑛−1 +𝜂𝑛
In similar manner, we obtain
𝑠 (𝑠)
̅𝑛−3
𝜎 1
𝑑 (𝑠) = 𝜂́ 𝑛−3 (1 − 𝐸 𝜂́ 1
) (43)
𝜀̅𝑛−3 1+ ́ 𝑛−3 (1+ 𝑛−2 (1− )
𝜂𝑛−2 𝐸𝑛−2 𝐸 𝜂́𝑛−1
1+ 𝑛−2 (1+ )
𝜂́𝑛−1 𝐸𝑛−1 +𝜂́𝑛
The following equation is achieved in step (𝑛 − 4)𝑡ℎ by the help of above approach

𝜎̅0𝑠 (𝑠) 1
= 𝜂́ 0 (1 − ) (44)
𝜀̅0𝑑 (𝑠) 𝐸 𝜂́
1+ ́ 0 (1+ 1 (1−
1
)
𝜂1 𝐸1
𝐸 𝜂́ 1
1+ 1 1+ 2 (1− ⋮ )
𝜂́2 𝐸2
𝐸𝑛−2 𝜂́𝑛−1
1+ (1+ )
( 𝜂́𝑛−1 𝐸𝑛−1 +𝜂́𝑛 )
NB: Donating the following symbols to simplification are considered
𝑎1 𝑎 𝑎 𝑎
𝑎2 = 1 2 ……… 𝑛 (45)
𝑏1 + 𝑎3 𝑏1 + 𝑏2 + 𝑏𝑛
𝑏2 +
𝑏3 + ⋱
𝑎
+ 𝑛
𝑏𝑛
𝐸𝑖−1 𝜂́ 𝑖
= 𝑎𝑖 , = 𝑏𝑖 (46)
𝜂́ 𝑖 𝐸𝑖
1
𝐴= 1
1+𝑎1 (1+𝑏1 (1− 1 )
1+𝑎2 (1+𝑏2 (1− ⋱ )
𝜂́𝑛−1
1+𝑎𝑛−1 (1+ )
𝐸𝑛 +𝜂́𝑛
Thus
n−1 ή
1 a1 (1+b1 ) a (1+bn−2 ) an−1 (1+ En +ήn )
A= [1+a1 (1+b1 )]− [1+a2 (1+b2 )]−
… n−2 (47)
1+ 1

Let us to consider 𝑎𝑖 (1 + 𝑏𝑖 ) = 𝑑𝑖 , by rewriting equation (46), we obtain

n−1 ή
1 𝑑1 𝑑𝑛−2 𝑎𝑛−1 (1+ 𝐸𝑛 +ήn ) [1+𝑑1 ]−1 𝑑1 (−[1+𝑑2 ])−1 𝑑𝑛−2 (1+𝑎𝑛−1 )−1
𝐴 = [1+𝑑 … = … =
1 ]− [1+𝑑2 ]− 1+ 1 1+ 1+ 1+
𝑎𝑛−1 ή n−1 (𝐸𝑛 +ή n )−1
(78)
1
To take factor form term 1/𝑠, assuming

31
 𝐸𝑖−1 𝐸𝑖−1
𝑎́ 𝑖 = , ℎ𝑖 = (79)
𝜂𝑖 𝐸𝑖
Thus
1 1
𝑑𝑖 = 𝑎́ 𝑖 + ℎ𝑖 , 𝑎𝑖 = 𝑎́ 𝑖 (80)
𝑠 𝑠
Based on equation (76), the equation (44) is rewritten as follow
−1 −1
1 1 1
[1 + ℎ1 + 𝑎́ 1 ] (ℎ1 + 𝑎́ 1 ) (− [1 + ℎ1 + 𝑎́ 2 ])
𝐴= 𝑠 𝑠 𝑠 …
1+ 1+
−1
1 1
(ℎ𝑛−2 + 𝑎́ 𝑛−2 ) (1 + 𝑎𝑛−1 ) 𝑎́ 𝑛−1 𝜂𝑛−1 (𝐸𝑛 + 𝑠𝜂𝑛 )−1
𝑠 𝑠
1+ 1
−1
1 −1 1 1
[𝑠(1 + ℎ1 ) + 𝑎́ 1 ] (𝑠ℎ1 + 𝑎́ 1 ) (− [1 + ℎ1 + 𝑎́ 2 ])
=𝑠 𝑠 𝑠 …
1+ 1+
−1 −1
1 1 1 1
(𝑠ℎ𝑛−2 + 𝑎́ 𝑛−2 ) (1 + 𝑎𝑛−1 ) 𝑎́ 𝑛−1 𝜂𝑛−1 ( 𝐸𝑛 + 𝜂𝑛 )
𝑠 𝑠 𝑠 𝑠 (81)
1+ 1
The following expansion holds according to ref [57].
1.(0+𝛾)𝑥
𝑥 (1−𝛾)𝑥
𝑥(𝑥 + 1)𝛾−1 = 1.2
… (82)
1+ 1+ 1+
The coefficient of equation (81) is chosen in a way that the equation (82) holds.

[𝑠(1 + ℎ1 ) + 𝑎́ 1 ]−1 = 𝑐0 ,
1 −1
(𝑠ℎ1 + 𝑎́ 1 ) (− [1 + ℎ1 + 𝑎́ 2 ]) = (1 − 𝛾)𝑐0 , 𝑒𝑡𝑐 (83)
𝑠
Based on equation (63)
(𝑛 − 1)(𝑛 − 𝛾)
𝜎̅0𝑠 (𝑠) 𝑐0 /𝑠 (1 − 𝛾)𝑐0 /𝑠 (2𝑛 − 1)(2𝑛 − 2) 𝑐0 /𝑠
= 𝜂́ 0 (1 − … )
𝜀̅0𝑑 (𝑠) 1+ 1+ 1

Now let 𝑛 → ∞ then

𝜎̅0𝑠 (𝑠) 𝑐0 𝑐0 𝛾−1
= 𝜂́ 0 (1 − ( ) ( + 1) )
𝜀̅0𝑑 (𝑠) 𝑠 𝑠

Therefore
𝜎̅0𝑠 (𝑠) 𝑐 𝛾
≈ 𝑠𝜂0 (1 − ( 0 ) ) (84)
𝜀̅0𝑑 (𝑠) 𝑠
Before applying Laplace transform, the left side of the equation (84) must depend on only a damper.
Thus, the following calculation is done to this end.
(1) 𝑑
𝜎̅0𝑠 (𝑠) ̅ (𝑠)
𝜎 ̅1𝑑 (𝑠)
𝜎
=⏞ ̅ 0𝑑 (𝑠) − (85)
𝜀̅0𝑑 (𝑠) 𝜀0 𝜀̅0𝑑 (𝑠)
Also

𝜎̅1𝑑 (𝑠) (9) 1 (1)

1 (10)
1
=⏞ 𝑠𝜂0 =⏞ 𝑠𝜂 0 𝑠 (𝑠) =
⏞ 𝑠𝜂0 ( )
𝜀̅0𝑑 (𝑠) 𝑑 (𝑠)
𝜎̅0 𝜎̅0 𝐸0
1+ 𝑑 1 +
𝑑 𝜎̅1 (𝑠) 𝜎̅1𝑑 (𝑠)
( 𝜎̅1 (𝑠) ) ( )
𝜀̅0𝑑 (𝑠)
̅1𝑑 (𝑠)
𝜎
Considering =𝑦
𝜀̅0𝑑 (𝑠)
1
𝑦 = 𝑠𝜂0 ( )
𝐸0
1+
𝑦
As a result
32
 𝑦 = 𝑠𝜂0 − 𝐸0 (86)
By putting equation (86) into equation (85), we will obtain
𝜎̅0𝑑 (𝑠) 𝜎̅0𝑠 (𝑠)
= + 𝑠𝜂0 − 𝐸0
𝜀̅0𝑑 (𝑠) 𝜀̅0𝑑 (𝑠)
Hence, based on equation (84) we will have
𝜎̅0𝑑 (𝑠) 𝑐0 𝛾
= 𝑠𝜂 0 (1 − ( ) ) + 𝑠𝜂0 − 𝐸0
𝜀̅0𝑑 (𝑠) 𝑠
By rewriting the above equation
̅0𝑑 (𝑠)
𝜎
= 2𝑠𝜂0 − 𝑠𝜂0 (𝑐0 /𝑠)𝛾 − 𝐸0 ‫ز‬ (87)
𝜀̅0𝑑 (𝑠)
By the help of equation (83), we obtain
𝜎̅0𝑑 (𝑠)
= 2𝑠𝜂0 − 𝑠𝜂0 (𝑠(𝑠(1 + ℎ1 ) + 𝑎́ 1 ))−𝛾 − 𝐸0
𝜀̅0𝑑 (𝑠)
In other words
̅ 𝑑 (𝑠)
𝜎
= 2𝑠𝜂0 − 𝐸0 − 𝜂0 𝑠1−𝛾 (𝑠(1 + ℎ1 ) + 𝑎́ 1 )−𝛾 (88)
𝜀̅𝑑 (𝑠)
By the use of binomial distribution function for every arbitrary 𝛽, the following equation holds.
∞
1 𝑘+𝛽 𝑘
= ∑( )𝑍
(1 − 𝑧)𝛽+1 𝑘
𝑘=0
Thus
1 1 𝑘+𝛾 𝑎́ 𝑘
(𝑠(1 + ℎ1 ) + 𝑎́ 1 )−𝛾 = 𝛾 𝛾 = 𝛾 ∑∞
𝑘=0 ( ) (−1)𝑘 ( 1 ))
(𝑠(1+ℎ1 )) +(1−(−𝑎́ 1 /𝑠(1+ℎ1 )) (𝑠(1+ℎ1 )) 𝑘 𝑠(1+ℎ1
(89)
The following equation will have achieved by putting equation (89) into equation (88), Therefor

̅ 𝑑 (𝑠)
𝜎 𝑘+𝛾 𝑎́ 1𝑘
= 2𝑠𝜂0 − 𝐸0 − 𝜂0 𝑠1−2𝛾 (1+ℎ ) (−1)𝑘 ( 1 )) ∑∞
𝑘=0 ( (90)
𝜀̅𝑑 (𝑠) 1 𝑘 𝑠(1+ℎ1 )𝛾
By the use of Laplace transform from the above equation, we obtain
𝑑𝜀𝑑 (𝑡) 1 𝑑 1−2𝛾 𝜀𝑑 (𝑡)
𝜎 𝑑 (𝑡) = 2𝜂0 ( ) − 𝐸0 𝜀 𝑑 (𝑡) − 𝜂0 (1+ℎ )𝛾
+
𝑑𝑡 1 𝑑𝑡 1−2𝛾
1 𝑘+𝛾 𝑎́ 𝑘
𝜂0 (1+ℎ ∑∞
𝑘=0 ( ) (−1)𝑘 ( 1 ) 𝐿−1 {𝑠1−2𝛾−𝑘 𝜀̅𝑑 (𝑠)}
1 )𝛾 𝑘 1+ℎ1
(91)
where 0 < 𝛾 < 1.
Now let us assume that 𝐸0 = 0 in equation (91), resulting in we gain a combination of a damper and
FOE connected to each other in a series way.
𝑑𝜀 𝑑 (𝑡) 𝑑1−2𝛾 𝜀 𝑑 (𝑡)
𝜎 𝑑 (𝑡) = 2𝜂0 ( ) − 𝜂0
𝑑𝑡 𝑑𝑡 1−2𝛾

Acknowledgment

We would also like to show our gratitude to the Dr. Mahdieh Tahmasebi for sharing their pearls of
wisdom with us during the course of this research, and we thank Miss Mina Moeini for comments that
greatly improved the mathematical proofs in the paper.

33
References

1. Boyer, C.B. and U.C. Merzbach, A history of mathematics. 2011: John Wiley & Sons.
2. Dalir, M. and M. Bashour, Applications of fractional calculus. Applied Mathematical Sciences,
2010. 4(21): p. 1021-1032.
3. Magin, R.L., Fractional calculus in bioengineering. 2006: Begell House Redding.
4. Herrmann, R., Fractional calculus: An introduction for physicists. 2011: World Scientific.
5. Diethelm, K., The analysis of fractional differential equations: An application-oriented
exposition using differential operators of Caputo type. 2010: Springer.
6. Caputo, M., Linear models of dissipation whose Q is almost frequency independent—II.
Geophysical Journal International, 1967. 13(5): p. 529-539.
7. Podlubny, I., Fractional differential equations: an introduction to fractional derivatives,
fractional differential equations, to methods of their solution and some of their applications.
Vol. 198. 1998: Academic press.
8. Kilbas, A.A., H.M. Srivastava, and J.J. Trujillo, Preface. 2006, Elsevier.
9. Grigoletto, E.C. and E.C. de Oliveira, Fractional versions of the fundamental theorem of
calculus. Applied Mathematics, 2013. 4(07): p. 23.
10. Diethelm, K., Efficient solution of multi-term fractional differential equations using P (EC) m E
methods. Computing, 2003. 71(4): p. 305-319.
11. Diethelm, K., N.J. Ford, and A.D. Freed, Detailed error analysis for a fractional Adams
method. Numerical algorithms, 2004. 36(1): p. 31-52.
12. Meerschaert, M.M. and C. Tadjeran, Finite difference approximations for two-sided space-
fractional partial differential equations. Applied numerical mathematics, 2006. 56(1): p. 80-
90.
13. Jafari, H., C.M. Khalique, and M. Nazari, An algorithm for the numerical solution of nonlinear
fractional-order Van der Pol oscillator equation. Mathematical and Computer Modelling,
2012. 55(5): p. 1782-1786.
14. Jiang, H., et al., Analytical solutions for the multi-term time–space Caputo–Riesz fractional
advection–diffusion equations on a finite domain. Journal of Mathematical Analysis and
Applications, 2012. 389(2): p. 1117-1127.
15. Gülsu, M., Y. Öztürk, and A. Anapalı, Numerical approach for solving fractional relaxation–
oscillation equation. Applied Mathematical Modelling, 2013. 37(8): p. 5927-5937.
16. Bu, W., Y. Tang, and J. Yang, Galerkin finite element method for two-dimensional Riesz space
fractional diffusion equations. Journal of Computational Physics, 2014. 276: p. 26-38.
17. Bu, W., et al., Finite difference/finite element method for two-dimensional space and time
fractional Bloch–Torrey equations. Journal of Computational Physics, 2015. 293: p. 264-279.
18. Jin, B., et al., The Galerkin finite element method for a multi-term time-fractional diffusion
equation. Journal of Computational Physics, 2015. 281: p. 825-843.
19. Markovitz, H., Viscoelastic properties of polymers: By John D. Ferry, Wiley, New York, 1980,
xxiv+ 641 pp., $49.50. 1981, Academic Press.
20. McCrum, N.G., C. Buckley, and C.B. Bucknall, Principles of polymer engineering. 1997: Oxford
University Press, USA.
21. Tanner, R.I., Engineering rheology. Vol. 52. 2000: OUP Oxford.
22. Caputo, M. and F. Mainardi, A new dissipation model based on memory mechanism. Pure
and Applied Geophysics, 1971. 91(1): p. 134-147.
23. Bagley, R.L. and J. TORVIK, Fractional calculus-a different approach to the analysis of
viscoelastically damped structures. AIAA journal, 1983. 21(5): p. 741-748.

34
24. Rogers, L., Operators and fractional derivatives for viscoelastic constitutive equations.
Journal of Rheology, 1983. 27(4): p. 351-372.
25. Bagley, R.L. and P.J. Torvik, On the fractional calculus model of viscoelastic behavior. Journal
of Rheology, 1986. 30(1): p. 133-155.
26. Koh, C.G. and J.M. Kelly, Application of fractional derivatives to seismic analysis of base‐
isolated models. Earthquake engineering & structural dynamics, 1990. 19(2): p. 229-241.
27. Makris, N. and M. Constantinou, Fractional-derivative Maxwell model for viscous dampers.
Journal of Structural Engineering, 1991. 117(9): p. 2708-2724.
28. Friedrich, C. and H. Braun, Generalized Cole-Cole behavior and its rheological relevance.
Rheologica Acta, 1992. 31(4): p. 309-322.
29. Fenander, A., Modal synthesis when modeling damping by use of fractional derivatives. AIAA
journal, 1996. 34(5): p. 1051-1058.
30. Pritz, T., Analysis of four-parameter fractional derivative model of real solid materials.
Journal of Sound and Vibration, 1996. 195(1): p. 103-115.
31. Shimizu, N. and W. Zhang, Fractional calculus approach to dynamic problems of viscoelastic
materials. JSME International Journal Series C Mechanical Systems, Machine Elements and
Manufacturing, 1999. 42(4): p. 825-837.
32. Rossikhin, Y.A. and M. Shitikova, Analysis of dynamic behaviour of viscoelastic rods whose
rheological models contain fractional derivatives of two different orders. ZAMM‐Journal of
Applied Mathematics and Mechanics/Zeitschrift für Angewandte Mathematik und Mechanik,
2001. 81(6): p. 363-376.
33. Mainardi, F. and G. Spada, Creep, relaxation and viscosity properties for basic fractional
models in rheology. The European Physical Journal-Special Topics, 2011. 193(1): p. 133-160.
34. Barpi, F. and S. Valente, Creep and fracture in concrete: a fractional order rate approach.
Engineering Fracture Mechanics, 2002. 70(5): p. 611-623.
35. Craiem, D., et al., Fractional calculus applied to model arterial viscoelasticity. Latin American
applied research, 2008. 38(2): p. 141-145.
36. Lewandowski, R. and B. Chorążyczewski, Identification of the parameters of the Kelvin–Voigt
and the Maxwell fractional models, used to modeling of viscoelastic dampers. Computers &
structures, 2010. 88(1): p. 1-17.
37. Meral, F., T. Royston, and R. Magin, Fractional calculus in viscoelasticity: an experimental
study. Communications in Nonlinear Science and Numerical Simulation, 2010. 15(4): p. 939-
945.
38. Lewandowski, R. and Z. Pawlak, Dynamic analysis of frames with viscoelastic dampers
modelled by rheological models with fractionalderivatives. Journal of sound and Vibration,
2011. 330(5): p. 923-936.
39. Celauro, C., et al., Experimental validation of a fractional model for creep/recovery testing of
asphalt mixtures. Construction and Building Materials, 2012. 36: p. 458-466.
40. Grzesikiewicz, W., A. Wakulicz, and A. Zbiciak, Non-linear problems of fractional calculus in
modeling of mechanical systems. International Journal of Mechanical Sciences, 2013. 70: p.
90-98.
41. Dai, Z., et al., A model of lung parenchyma stress relaxation using fractional viscoelasticity.
Medical engineering & physics, 2015. 37(8): p. 752-758.
42. Jóźwiak, B., M. Orczykowska, and M. Dziubiński, Fractional generalizations of maxwell and
Kelvin-Voigt models for biopolymer characterization. PloS one, 2015. 10(11): p. e0143090.
43. Djordjević, V.D., et al., Fractional derivatives embody essential features of cell rheological
behavior. Annals of biomedical engineering, 2003. 31(6): p. 692-699.
44. Carmichael, B., et al., The fractional viscoelastic response of human breast tissue cells.
Physical biology, 2015. 12(4): p. 046001.

35
45. Ingber, D.E., et al., Cellular tensegrity: exploring how mechanical changes in the cytoskeleton
regulate cell growth, migration, and tissue pattern during morphogenesis. International
review of cytology, 1994. 150: p. 173-224.
46. Chicurel, M.E., C.S. Chen, and D.E. Ingber, Cellular control lies in the balance of forces.
Current opinion in cell biology, 1998. 10(2): p. 232-239.
47. Zemel, A., R. De, and S.A. Safran, Mechanical consequences of cellular force generation.
Current Opinion in Solid State and Materials Science, 2011. 15(5): p. 169-176.
48. Alberts, B., et al., Essential cell biology. 2013: Garland Science.
49. STAMENOVIĆ, D. and M.F. COUGHLIN, The role of prestress and architecture of the
cytoskeleton and deformability of cytoskeletal filaments in mechanics of adherent cells: a
quantitative analysis. Journal of Theoretical Biology, 1999. 201(1): p. 63-74.
50. Maurin, B., et al., Mechanical model of cytoskeleton structuration during cell adhesion and
spreading. Journal of biomechanics, 2008. 41(9): p. 2036-2041.
51. Bausch, A.R., et al., Local measurements of viscoelastic parameters of adherent cell surfaces
by magnetic bead microrheometry. Biophysical journal, 1998. 75(4): p. 2038-2049.
52. Chen, T.-J., C.-C. Wu, and F.-C. Su, Mechanical models of the cellular cytoskeletal network for
the analysis of intracellular mechanical properties and force distributions: A review. Medical
engineering & physics, 2012. 34(10): p. 1375-1386.
53. Coughlin, M.F. and D. Stamenović, A prestressed cable network model of the adherent cell
cytoskeleton. Biophysical journal, 2003. 84(2): p. 1328-1336.
54. Kojima, H., A. Ishijima, and T. Yanagida, Direct measurement of stiffness of single actin
filaments with and without tropomyosin by in vitro nanomanipulation. Proceedings of the
National Academy of Sciences, 1994. 91(26): p. 12962-12966.
55. Käs, J., et al., F-actin, a model polymer for semiflexible chains in dilute, semidilute, and liquid
crystalline solutions. Biophysical journal, 1996. 70(2): p. 609-625.
56. Brown, J.W., Complex variables and applications. 2009.
57. Jones, W.B. and W.T.C. Fractions, Analytic Theory and Applications. Encyclopedia of
Mathematics and its Applications, 1980.

36