The relationship between order of

(m,
b nb)-Riemann-Liouville fractional integral and the
fractal dimensions of a fractal function
A.A.Navish∗, M.Priya, R.Uthayakumar
Department of Mathematics, The Gandhigram Rural Institute - Deemed to be University, Dindigul,
TamilNadu, India - 624 302.

Abstract
This paper mainly investigates the relationship between the fractal dimension and the
order of fractional integral of (m,
b nb)- Riemann-Liouville fractional integral of the famous
fractal function known as Weierstrass function. This artifact inspects many approaches
to the use of fractional calculus, in particular (m,
b nb)-Riemann-Liouville fractional in-
tegral in science and engineering. Our result is manifested by varying the order of
fractional integral. The related fractal dimension can be seen in the graphical repre-
sentation.
Keywords: (m, b n
b)-Riemann-Liouville fractional integral, Fractal dimension,
Weierstrass function, Fractional integral

1. Introduction
The analysis of integrals and derivatives of non-integer order is known as fractional
calculus theory, which has a large part of its use as a real-life problem. Over the last
two decades, too many mathematicians, such as Reimann, Liouville, Caputo argued
important findings related to fractional calculus, which is the perfect mathematical
method for applied sciences. The definition of fractal has attracted significant interest
in recent days and has been applied to a number of science and engineering issues. One
of the key uses of the order of fractional integro-differential equation is to relate the
fractal component of the graph of a fractal function. In 2004, K.Yao et.al studied the

∗
Corresponding author
Email addresses: [email protected] (A.A.Navish), [email protected]
(M.Priya), [email protected] (R.Uthayakumar)

Preprint submitted to Elsevier February 19, 2021

fractional calculus of the Weierstrass function. Riemann-Liouvillie fractional integro-
differential equation was used for their analysis. In this paper we were interested in
using the more generalized version of the Riemann-Liouvillie fractional integral defined
as (m,
b n
b)-Riemann-Liouvillie fractional integral equation.
To simplify discussions, we make the following symbols.
Notations Description
J The unit interval [0, 1]
φ Continuous function φ on J
G(φ, J) The graph of function φ on J
OSC(φ,J) = supx,y∈J |φ(x) − φ(y)| The oscillation of φ on interval J
So far in this present paper, by M symbolizing a positive constant that may have
varying values at different occasions, by Mi (b
α) denoting a positive constant based on
α
b.

2. Preparations
Definition 2.1. [2] Let φ be a continuous function on a finite interval J. Then (m,
b nb)
b ≥ 0 is defined by
Riemann-Liouville fractional integral of φ of order α
α pb
n + 1)1− mb
Z
(b
b
α
n
b α
pnb+1 − τ nb+1 ) mb −1 τ nb φ(τ )dτ, pb ∈ J
b
b J0 φ(x) = (b
b
m
mΓ
b mb (b
α) 0

where m > 0, n ∈ R − {−1}.

Remark 2.1. [3] The definition of fractional derivatives can also be defined by Defini-
tion 2.1. We are mainly interested in considering a fractal function, ie., the Weierstrass
function defined by
X
W(t) = λ−βj sin (λj t), 0 < βb < 1, λ > 1
b

j≥1

Let
α pb
n + 1)1− mb
Z
(b
b
α
−b
D α
pnb+1 − τ nb+1 ) mb −1 τ nb sinaτ dτ = Sbx (b
b
sinat = (b α, a)
mΓm (bα) 0

denote the (m,

b nb) Riemann-Liouville fractional integral of sin at, and let
α pb
n + 1)1− mb
Z
(b
b
α
−b
D α
pnb+1 − τ nb+1 ) mb −1 τ nb cosaτ dτ = C
b
cosat = (b bx (b
α, a)
mΓm (bα) 0

2
denote the (m,
b nb) Riemann–Liouville fractional integral of cosat. For λ > 1, 0 < β,
bα b<1
with 0 < β + α
b b < 1.
Define X b
g(t) := D−bα (W(t)) = λ−βj Sbx (b
α , λj )
j≥1

be the fractional integral of Weierstrass function W(t) of order α

b.

Figure 1: (a) Weierstrass function [W(t)] and (b) Fractional integral of W(t)

Lemma 2.1. [7] For Sbx (b

α, a) and C
bx (b
α, a), we have the following relations:
(i) Sbx (b
α, a) = aC
bx (b
α + 1, a);

(ii) d(Sbx (b α − 1, a);

α, a))/dx = Sbx (b

(iii) d(C
bx (b
α, a))/dx = C α − 1, a).
bx (b

Proposition 2.1. [8] Let φ be continuous on J, 1 < s0 ≤ 2, let P = {0 = t0 < t1 <

t2 < ... < tn = 1} be a partition of J, δi = [ti−1 , ti ), |P | = max |δi |. Then we have
1≤i≤n

0 0
X
K s (G(φ, J)) = lim+ inf OSC(φ, δi ) |δi |s −1
δ→0 |P |<δ P

0
where K s (G(φ, J)) denotes the K-measure of G(φ, J).
Proposition 2.2. [9] Let φ be continuous on J, M > 0, 0 ≤ s0 ≤ 1, suppose
0
p − qb|s
|φ(x) − φ(y)| ≤ M |b (0 ≤ pb, qb ≤ 1)

3
Then
dimB (G(φ, J)) ≤ 2 − s0 .

Proposition 2.3. [8] Let φ be continuous on J. Then

(i) dimK G(φ, J) ≤ dimB G(φ, J);

(ii) If the values of dimK G(φ, A) are equal for all open intervals of A of J. Then
dimK G(φ, J) ≤ dimP G(φ, J).

Proposition 2.4. Let 0 < α

b < 1, a > 1, J.

(i) Let x ∈ J. Then

n + 1)αb/m
(b
|Sbpb(b
α, a)| ≤ , |C α, a)| ≤ M a−bα
bpb(b
αbΓm
b (b
α)
  mαbb −2   mαbb −2 
4jπ 3π M h1 a−1 (8−3/4)π (8−1/4)π
(ii) Let b
pbj = a
,j = 2, 3, . . . , h = a
, M1 (b
α) = n
b+1 a
− a
≥
M a−1
α, a) − Sbbpb (b
Sbbpb +h1 (b α)a−bα
α, a) ≥ M1 (b
j j

 α
b −1

(4π) m α α
1 −1 −1
(iii) Let x > 0, M2 = + (2π) +π . Then
b b b
m m
α
b b
mΓ
b mb (b
α) b
m
b

|Sb0 pb(b α)a1−bα

α, a)| ≤ M2 (b

α
b
2(4π) m
(iv) Let x > 0, 0 < h < 1, M3 (b
α) = . Then
b
α
b.mΓmb (b
α)

|Spb+h (b
α, a) − Sbpb(b α)a−bα
α, a)| ≤ M3 (b

Proof. We prove the proposition one by one.

4
(i)
pb
n + 1)1−bα/m
Z
(b b
α
|Sbpb(b
α, a)| ≤ pnb+1 − τ nb+1 ) mb −1 τ nb sinaτ |dτ
|(b
b

mΓ
b mb (b
α) 0
1−b
α/m Z pbn
b +1
(b
n + 1) b
α du
≤ pnb+1 − u) mb −1 |
|(b
b

mΓ
b mb (b
α) 0 (b
n + 1)
α
b/m Z pbn
b +1
(b
n + 1) b
α
≤ pnb+1 − u) mb −1 du
b
(b
mΓ
b m b (b
α) 0
n + 1)αb/m
(b b
≤
mΓ
b m b (b
α)

(ii) We may verify that

Z pbj +h1
α
α, a) − Sbpbj (b pj + h1 )nb+1 − τ nb+1 ) mb −1 τ nb sinaτ dτ
b
Sbpbj +h1 (b α, a) = ((b
pbj
Z pbj
α α
pj + h1 )nb+1 − τ nb+1 ) mb −1 − (b
pn+
j −τ
n
) b −1 ]τ nb sinaτ dτ
b+1 m
b b
+ [((b
pbj −2h1
Z pbj −2h1
α α
pj + h1 )nb+1 − τ nb+1 ) mb −1 − (b
pn+
j −τ
n
) b −1 ]τ nb sinaτ dτ
b+1 m
b b
+ [((b
0
= I1 + I2 + I3

Now by Lemma 2.1

Z (4j+3)π
a α
pj + h1 )nb+1 − τ nb+1 ) mb −1 τ nb sinaτ dτ
b
I1 = ((b
4jπ
a
Z (4j+1)π
a α
pj + h1 )nb+1 − τ nb+1 ) mb −1 τ nb sinaτ dτ
b
= ((b
4jπ
a
Z (4j+2)π
a α α
pj + h1 )nb+1 − τ nb+1 ) mb −1 − ((b
pj + h1 )nb+1 − (τ + π/a)nb+1 ) mb −1 ]τ nb sinaτ dτ
b b
+ [((b
(4j+1)π
a
≥ 0.

5
 while by Lemma 2.1 again,
Z 4jπ
a α α
pj + h1 )nb+1 − τ nb+1 ) mb −1 − ((b
pj )nb+1 − τ nb+1 ) mb −1 ]τ nb sinaτ dτ
b b
I2 = [((b
(4j−6)π
a
2j−1 (2k+1)π
X Z a α α
pj + h1 )nb+1 − τ nb+1 ) mb −1 − ((b
pj + h1 )nb+1 − (τ + π/a)nb+1 ) mb −1 ]τ nb
b b
= [((b
2kπ
k=2j−3 a
α
pj + h1 )nb+1 − (τ + π/a)nb+1 ) mb −1 sinaτ dτ ≥ 0.
b
+((b

By considering (3) and using the way close to the estimation of J(b
p, h) in Lemma
2.1, we get an appropriate large λ.
2j−4 Z (2k+1)π
X a α
−1
I3 ≥ M h1 a pnjb+1 − τ nb+1 ) mb −1 τ nb sinaτ dτ
b
(b
2kπ
k=0 a
"  mαbb −2  mαbb −2 #
M h1 a−1

(8 − 3/4)π (8 − 1/4)π
≥ − ≥ M a−1
n
b+1 a a

(iii) As we know (by Lemma 2.1), Sb0 pb(b c0 pb(b

α, a) = aC α, a) = aC
bpb(b
α, a). Hence
Z pb+h
α
Spb0 +h (b p + h)nb+1 − τ nb+1 ) mb −1 τ nb cosaτ dτ
b
α, a) =a ((b
0

2N1 +1/2
Related to the facts in part (ii), by selecting N1 to satisfy a
π ≤ pb+η −2h0 <
2(N1 +1)+1/2
a
π and writing
Z pb+η Z pb+η
α α
−1
n
b+1
−τ n
b+1 m
p + η)nb+1 − τ nb+1 ) mb −1 τ m
b b
((b
p + η) ) m
b τ cosaτ dτ
b
= ((b b
cosaτ dτ
0 pb+η−2h0
Z pb+η−2h0
α
p + η)nb+1 − τ nb+1 ) mb −1 τ m
b
+ ((b b
cosaτ dτ
(2N1 +1/2)π
a
Z π
2a α
p + η)nb+1 − τ nb+1 ) mb −1 τ m
b
+ ((b b
cosaτ dτ
0
(2j+3/2)π
N1 −1 Z
X a α
p + η)nb+1 − τ nb+1 ) mb −1
b
+ ((b
(2j+1/2)π
j=0 a
α
p + η)nb+1 − (τ + π/a)nb+1 ) mb −1 τ m
−((b
b
b
cosaτ dτ

6
 we see that
N1 −1 Z
(2j+3/2)π   mαbb
X a α π/aτ
|Spb0 +η (b ≤ Ma n
b+1
−τ n
b+1 −1 m
1− 1−( )nb+1
b
α, a)| ((b
p + η) ) m
b τ b
dτ
j=0
(2j+1/2)π
a
pb + η − τ
α
b
+o(ah0mb )
α
b/m
≤ M3 ah0
b

by a similar argument to that of part(iii) of this theorem. Thus we have the

requisite outcome.

(iv) The result for the case x ≤ 4h0 is trivial. Suppose x > 4h0 . Apparently, it
maintains that
Z pb+h
α
p + h)nb+1 − τ nb+1 ) mb −1 τ m sinaτ dτ = o(hαb/m
b
((b b b
) (1)
pb

Around the same moment, it is also evident that

Z pb Z pb
α α
−1 m
n
b+1
p+h) −τ ) n
b+1 m
τ sinaτ dτ − p+h)nb+1 −τ nb+1 ) mb −1 τ m sinaτ dτ = o(hαb/m
b b
((b b
b
((b b b
)
pb−2h0 pb−2h0
(2)
By setting
2N0 2(N0 + 1)
π ≤ x − 2h0 < π
a a
for some N0 . We conclude that
Z x−2h0
α α
p + h)nb+1 − τ nb+1 ) mb −1 − (b pnb+1 − τ nb+1 ) mb −1 ]τ nb sinaτ dτ + o(hαb/m
b b
J(bp, h) = [((b b
)
0
nb+1 ! mαbb −1

N −1 2(j+1)π
0 Z 
X λ α h
pnb+1 − τ nb+1 ) mb −1  1 +
b
= (b
j=0
2jπ
λ
pb − τ
nb+1 ! mαbb −1 nb+1 ! mαbb −1

 
h − π/a π/a α
b
 τ nb sinaτ dτ + o(h0mb )
− 1+ + 1−
pb − τ pb − τ

Since for small βb1 , βb2 < 1, we have

α α α
(1 − βb1 ) mb −1 − 1 − (1 + βb1 − βb2 ) mb −1 + (1 − βb2 ) mb −1 ≈ βb1 βb2
b b b
(3)

7
 It follows that
2(j+1)π
N0 −1 Z
X a α α
−1
pnb+1 − τ nb+1 ) mb −3 τ nb sinaτ dτ | + o(h0mb )
b
|J(b
p, h)| ≤ M ha |
b
(b
2jπ
j=0 a

nb+1 ! mαbb −2
 
−1

M ha α x − 2h0 α
pnb+1 ) mb −2  1 −
b
≤ − 1 + o(h0mb )
b

α
(b
(b
n + 1)( m
b
b
− 2) pb

Proposition 2.5. Let λ > 1, 0 < α

b, βb < 1, with βb + α
b < 1, and J = [0, 1]. Then g(t)
is continuous on J.
Proof. By using Proposition 2.4(i), we get

X X
|g(t)| = λ−βj Sbpb(b
α , λj ) ≤ λ−βj < ∞
b b

j≥1 j≥1

This shows the continuity of g(t).

3. Main Results
Theorem 3.1. Let 0 < βb < 1,b b α < 1, λ > 1 and J. Then dimB G(g, J) ≤
α < 1, with β+b
2 − βb − α
b.
−N
Proof. For any given 0 < h < 1, there exists integer N ≥ 0 such that h ∈ [λ(bn+1),λ ).
Then
N
X
|g(b
p + h) − g(b
p)| ≤ λ−βj Spb+h (b
α, λj ) − Sbpb(b
α , λj )
b

j=1
X∞
+ λ−βj Sbpb+h (b
α, λj ) − Sbpb(b
α, λj )
b

j=N
= 41 + 42 (4)

By the mean-value theorem and Proposition 2.4(iii), we get

N
X N
X
−βj
41 = λ h|Spb0 +h (b
α, λ)| ≤M λ−βj h(λj )1−bα ≤ M hβ+bα (5)
b b b

j=1 j=1

8
by Proposition 2.4(iv), we get
∞  β+b
b α
X
−βj j 1−b
α 1
42 ≤ M λ (λ ) ≤M ≤ M hβ+bα (6)
b b

j=N +1
λnb+1

Combine (4), (5) and (6), we have

|g(b p)| ≤ M hβ+bα

p + h) − g(b (7)
b

By (7) and Proposition 2.2 completing the proof of theorem 1.

Theorem 3.2. Let J, λ > 1, 0 < β,bαb < 1, with βb + α

b < 1, k > 0. Then for λ > λ0 , it
holds that dimK G(g, J) ≥ 2 − β − α
b b, where

λ0 = max{12π, λ1 , λ2 }
α) + 1)1/(β+bα/m)
λ1 = (2M5 (b b b

α) + 3)1/(β+bα/m)
λ2 = (2M5 (b b b

M5 (b
α) = 2M4 (b
α)/M1 (b
α)
α) = max{M20 (b
M4 (b α)}, M20 (b
α), M3 (b α) = 3πM2 (b
α). (8)

Proof. Let δ < 1/λ, we consider any partition P of J, for anyh given interval
 δi of P ,
2
there exists an integer N > 0, such that |δi | ∈ λN −1 , λN1−2 = λλN , λλN , let R = {b
 1

pj :
pbj = 4jπ
λN
, j = 2, 3, . . .}, h = λ3πN . Because λ > 12π, then, there exists at least one point
pbj of δi , such that pbi ∈ R, and (b pi , xi+1 ) ⊂ δi . Due to Proposition 2.4(ii), we get
α
α, λN ) − Sbpbj (b α)λ− mb N
α, λN ) ≥ M1 (b
b
Sbpbj +h (b (9)

In the other side, we can see that

|g(b p) − λ−βN
p + h) − g(b
b

N
X −1
α, λN ) − Sbpb(b
(Sbpb+h (b α, λN ))| ≤ λ−βj |Sbpb+h (b
α, λj ) − Sbpb(b
α, λj )|
b

j=1
∞
X
+ α, λj ) − Sbpb(b
|Sbpb+h (b α, λj )|
j=N +1
= 41 + 42 (10)

9
by mean-value theorem and Proposition 2.4(iii) and (iv), we get
N
X −1 N
X
−βj
41 = λ α, λj )| ≤
|Sb0 pb(b λ−βj hM2 (b
α)(λj )1−bα
b b

j=1 j=1

λ(β+bα)−1
b
≤ M20 (b
α) λ−(β+bα)N
b

1− λ(β+b
b α)−1
∞
λ−(β+bα)
X b
j −βb
42 = βj
λ M3 (b
α)(λ ) ≤ M3 (b
α). λ−(β+bα)N
b b

j=N +1
1− λ−(β+b
b α)

Hence !
λ(β+bα)−1 λ−(β+bα)
b b
41 +42 ≤ M4 (b
α) + λ(β+b
b α)N (11)
1 − λ(β+b
b α)−1
1 − λ−(β+b
b α)

Let λ(β+bα)−1 = y, it is not difficult to find that

b

λ(β+bα)−1 λ−(β+bα) λbq2 + 1

b b
+ ≤ (12)
1 − λ(β+b
b α)−1
1 − λ−(β+b
b α) (1 − qb)(λb
q − 1)

Since λ > λ0 , we have

1 2M5 (b
α) + 2
0 < qb < . < 1, (13)
2M5 (b
α) + 1 q−1
λb

then
λbq2 + 1 1
≤ (14)
(1 − qb)(λb
q − 1) M5 (b
α)
combine (10) and (14), we get

|g(b p) − λ−βN (Sbpb+h (b

p + h) − g(b α, λN ) − Sbpb(b α)λ−(β+bα)N
α, λN ))| ≤ 1/2M1 (b (15)
b b

pi , pbi + h) ⊂ δi , such that

By combining (9) and (15), we know that there exist (b

M1 (b
α) −(β+b
α)λ−(β+bα)N −
b α)N
|g(b
pi + h) − g(b
pi )| ≥ M1 (b λ ≥ M |δi |β+bα
b b
2
Hence
OSC(g, δi ) ≥ M |δi |β+bα (16)
b

10
By noting Proposition 2.1 and (16), we get
X
M 2−β−bα (G(g, J)) = lim+ inf OSC(g, δi )|δi |1−β−bα ≥ M 0 (17)
b b
δ→0 |P |<δ P

Hence dimK G(g, J) ≥ 2 − βb − α

b.
Combining Theorems 3.1 and 3.2, we can get our ultimate results.
Theorem 3.3. Let g(t) be the fractional integral of Weierstrass function W(t), 0 <
β,
bα b < 1, with βb + α
b < 1. Then, for λ > λ0 , it holds that

dimB G(g, J) = dimP G(g, J) = dimK G(g, J) = 2 − βb − α

b

Theorem 3.4. Let g(t) be the fractional integral of Weierstrass function W(t), 0 <
β,
bα b < 1, with βb + α
b < 1. Then, for λ > λ0 , it holds that

dimB G(g, J) = dimP G(g, J) = dimK G(g, J) = 2 − βb − α

b

For simplicity, if the box dimension, K-dimension, and the packing dimension of
the fractal function φ are equal on J, let dim G(φ, J) signify the three dimensions.
From Theorem 3.3, we have

dim G(g, J) = 2 − βb − α
b (18)

On the other hands, we know from References [8, 9, 10] that

dim G(W, J) = 2 − βb (19)

A combination of (18) and (19) leads to

dim G(g, J) = dim G(W, J) − α

b

this presents the linear relation between the order of the fractional calculus and the
fractal dimensions of the Weierstrass graphs.

4. Numerical Examples
This section provides a numerical example to emphasize the linear relationship
between the fractional integral order and the fractal dimensions of graphs of the Weier-
strass function.

11
 Here the following values are fixed to illustrate the graphical effects. Let λ = 2,
n = 0, m = 1, βb = 1/2 and α
b differ as 1/4, 1/8, 1/16, 1/32.

Table 1 displays the connection between the order of fractional integral and fractal
dimension of graph of W(t).

α
b dim G(g, J)
1/4 1.25
1/8 1.375
1/16 1.4375
1/32 1.46875
0 1.5
Table 1: Order of fractional integral and fractal dimensions of G(g, J)

Figure 2: Graphical representation of Table 1

5. Applications
? The theme of fractional calculus has a wide variety of applications in ubiquitous
fields such as fluid mechanics, electromagnetics, electrochemistry, viscoelasticity,
biological population models, optics and signal processing.
? Dynamic systems involving accurate daming modelling are addressed by fractional
differential equations for accurate modelling.
? In biological systems, the bioheat equation is continually used as a first order
heat transfer model. During laser surgery, the measurement of the heat flux at

12
 the boundary for various periodic or on-off boundary conditions that closely reflect
the heating and cooling of the skin surface is regulated by a fractional differential
equation.

? To obtain information about the electrical properties of many different materials,

in particular, of liquids the most widely used technique is the electrical spec-
troscopy impedance.

6. Conclusion
The order of the fractional integral and the fractal dimension of (m, b nb)-Riemann-
Liouville fractional integral of the Weierstrass function is connected by a linear rela-
tionship. In particular, the box dimension, the packing dimension and the K-dimension
of the fractional integral graph of the Weierstrass function are calculated. A numerical
illustration for a different values of α
b (order of fractional integral) is used to explain
our finding.

13
References
[1] Sun, H., Zhang, Y., Baleanu, D., Chen, W., & Chen, Y. (2018). A new collec-
tion of real world applications of fractional calculus in science and engineering.
Communications in Nonlinear Science and Numerical Simulation, 64, 213-231.

[2] Sarikaya, M. Z., Dahmani, Z., KIRIS, M. E., & Ahmad, F. (2016). (k, s)-Riemann-
Liouville fractional integral and applications. Hacettepe Journal of Mathematics
and Statistics, 45(1), 77-89.

[3] Yao, K., Su, W. Y., & Zhou, S. P. (2005). On the connection between the order
of fractional calculus and the dimensions of a fractal function. Chaos, Solitons &
Fractals, 23(2), 621-629. Fractals, 23(2), 621-629.

[4] Zhou, S. P., He, G. L., & Xie, T. F. (2004). On a class of fractals: the constructive
structure. Chaos, Solitons & Fractals, 19(5), 1099-1104.

[5] Yao, K., Su, W. Y., & Zhou, S. P. (2004). On the fractional calculus of a type of
Weierstrass function. Chinese Annals of Mathematics, 25(6), 711-716.

[6] Tatom, F. B. (1995). The relationship between fractional calculus and fractals.
Fractals, 3(01), 217-229.

[7] Miller KS, Ross B. An introduction to the fractional calculus and fractional differ-
ential equations. New York: John WileySons Inc. 1993.

[8] Hu TY, Lau KS. Fractal dimensions and singularities of the Weierstrass type func-
tions. Trans Amer Math Soc 1993;335(2):649–65.

[9] Falconer J. Fractal geometry: mathematical foundations and applications. New

York: John Wiley Sons Inc.; 1990.

[10] Rezakhanlou F. The packing measure of the graphs and level sets of certain con-
tinuous functions. Math Proc Camb Philos Soc 1988;104:347–60.

14