arXiv:2212.00049v1 [math.

GM] 1 Dec 2022

Fourier Series in Fractional Dimensional

Space

Ali Dorostkar1* and Ahmad Sabihi1

1* Fractal Group, Isfahan, Iran.

*Corresponding author(s). E-mail(s):

m110alidorostkar@gmail.com;

Abstract
In this paper, a Fourier series in fractional dimensional space is intro-
duced for an arbitrarily periodic function f (t; α). We call it fractional
Fourier series of the order α. Extending the basis functions of the lin-
ear space into fractional one, by rotation transformation, we define
a real and complex Fourier series and obtain their coefficients. It is
also shown that the fractional derivative of a periodic function can be
realized through (fractional) Fourier series with modified coefficients.

Keywords: Fractional Fourier series, Rotation transformation, Fractional

dimensional space

1 Introduction
Fourier series plays an important role in mathematics and applied sciences.
It is a well representation of periodic function given in linear space of integer
dimensions. Many people have attempted to generalize it to fractional domain.
There are two methods generalizing Fourier series and discrete-time Fourier
transform to fractional Fourier series and discrete-time fractional Fourier trans-
form, respectively [6].
Several definitions to solve some of partial fractional differential equations
[1] are developed introducing modified Riemann-Liouville fractional derivatives
[10] and for solving nonlinear equations using local fractional Fourier series
[2, 7–9]. Jumarie [3] in a paper introduces fractional sine and cosine functions
and using them, he makes a Fourier series of fractional order.

1
2 Fourier Series in Fractional Dimensional Space

However, the authors of Ref.[4] have proposed there is an invalidity since frac-
tional Fourier series (FFS) of the basis functions cos(nωt)α and sin(nωt)α
derived from Mittag-Leffler function might not hold. In this paper, we show
how a new definition for basis functions of fractional dimension can create the
new basis functions for FFS.

2 Preliminary and Definitions

As is well-known, Fourier series asserts that any arbitrarily periodic function
satisfying Dirichlet’s sufficient conditions [5] can be expressed by a summation
of cosine and sinus functions and their coefficients are realized from the pro-
jection of function on three sets of basis functions including 1, cos(nωt) and
sin(nωt) (n=0,1,..), where the basis functions are orthonormal. The Fourier
series based on three sets of basis functions of ϕ0 , ϕ1 and ϕ2 can be expressed
in the following where ϕ0 , ϕ1 and ϕ2 are 1, cos(nωt) and sin(nωt) (n=0,1,..),
respectively.
∞ ∞
A0 X X
f (t) = ϕ0 + An ϕ1 + Bn ϕ2 (1)
2 n=1 n=1

First of all, we describe a general idea for definition of the basis functions for
fractional domain by the realization of linear and fractional spaces. Based on
linear algebra, a schematic of basis functions can be represented by orthogo-
nal vectors as depicted by Fig.1 (here ϕ1 and ϕ2 ). For sake of simplicity, we
describe the scenario with the two basis functions ϕ1 and ϕ2 . Based upon a
concept of linear algebra, we can move from higher (lower) dimension to lower
(higher) one through the integer derivative (integral) operator. For instance,
differentiating (integrating) cosine function, one can reach the sinus one and
vice versa.
If suppose the fractional space is a space between two consecutive integer
dimensions, then the basis functions in the fractional space can be represented
as a rotation of basis functions of the linear space. Fig.1 shows that the two
basis functions ϕ1 and ϕ2 are orthogonal to each other. Applying the fractional
derivative to ϕ1 (ϕ2 ), yields ϕα α
1 (ϕ2 ). These new basis functions are orthonor-
mal, which make a fractional derivative to be normalized (see Definition 2).

Definition1
Fractional space is a whole space between and included linear integer spaces.

Definition2
Let ϕ0 , ϕ1 and ϕ2 be the basis functions of the linear space. Let ϕα α
0 , ϕ1 and
α
ϕ2 be the basis functions of the fractional space, then there are the following
 Fourier Series in Fractional Dimensional Space 3

Fig. 1 Schematic of basis functions for linear and fractional spaces

relations between the basis functions of the both spaces:

 
ϕα
0 =D 1α
(2)

 
1 πα
ϕα
1 = D α
cos(nωt) = cos(nωt + ) (3)
(nω)α 2
 
1 πα
ϕα
2 = D α
sin(nωt) = sin(nωt + ) (4)
(nω)α 2
1
The factor of (nω) α is used for normalization to ensure an orthonormal condi-

tion.
Lemma1
Let ϕα0 be a first basis function in the fractional space, then
 
πα
ϕα
0 = D α
1 = cos( ) (5)
2

Proof :
We consider the basis function of ϕ0 in integer space based on the following
expression:  
ϕ0 = lim cos(nωt) = 1 (6)
n→0

Therefore, ϕα
0 can be expressed as:

(nω)α
 
1 πα πα
ϕα
0 = lim α
D α
cos(nωt) = lim cos(nωt + ) = cos( ) (7)
n→0 (nω) n→0 (nω)α 2 2
4 Fourier Series in Fractional Dimensional Space

Definition3
Fractional Fourier series based upon the fractional basis functions can be
expressed as follows:
∞ ∞
πα X πα X πα
f (t; α) = a0 cos( )+ an cos(nωt + )+ bn sin(nωt + ) (8)
2 n=1
2 n=1
2

where in this notation f(t; α) denotes the function at fractional dimension α.

Letting α = 0 changes f (t; α) from fractional dimension α into the original
one in the linear space i.e. into f (t) as shown in Eq.(1). Expanding the cosine
and sinus terms and rearranging, we find:

∞  
πα X πα πα
f (t; α) = a0 cos( )+ an cos( ) + bn sin( ) cos(nωt)
2 n=1
2 2
∞  
X πα πα
+ − an sin( ) + bn cos( ) sin(nωt) (9)
n=1
2 2

For instance, whenever α is equal to 12 , the function can be expressed as follows:

√ √ ∞ √ ∞
1 2 2X 2X
f (t; α = )= a0 + (an + bn )cos(nωt) + (−an + bn )sin(nωt)
2 2 2 n=1 2 n=1
(10)
Definition4
Looking at Fig1 and the relation (9), one finds out that the FFS coefficients are
a re-scaling of FS ones. This means that when moving to fractional dimension
of the order α and projecting the function f (t) on basis functions in fractional
space, we really have re-scaled coefficients of FS. These re-scaled coefficients
are derived from the following clockwise rotation matrix:

cos( πα πα
    
An
= 2 ) sin( 2 ) an
(11)
Bn − sin( 2 ) cos( πα
πα
2 ) bn

The FFS based upon the basis functions of the linear space can be
generalized from Eq.(1) as:
∞ ∞
A0 X X
f (t; α) = + An cos(nωt) + Bn sin(nωt) (12)
2 n=1 n=1
 Fourier Series in Fractional Dimensional Space 5

and by a comparison of the coefficients with Eq.(9) the rotation matrix can be
realized as follows:

cos( πα ) sin( πα
 
πα 2 2 )
R(− ) = (13)
2 − sin( πα πα
2 ) cos( 2 )

Theorem1
Let the function f (t; α) defined on [0, ∞) × [0, ∞) −→ R as a fractional
dimension space be a continuously and integrably periodic function of the order
α defined on [0, ∞) −→ R satisfying Dirichlet’s sufficient conditions. If Fourier
series of the function f (t; α) is expressed by the relation (8), then, the new
fractional coefficients expressed by (12) are as follows:

 A0 = 2a0 cos( πα
2 )
Ak = ak cos( πα πα
2 ) + bk sin( 2 ) if k ≥ 1 (14)
Bk = bk cos( 2 ) − ak sin( πα
πα
2 ) if k ≥ 1


and
 RT
 a0 cos( πα

2 ) = T2 −2T f (t; α)dt
2

 RT
ak = T2 −2T f (t; α)cos(kωt + πα 2 )dt if k ≥ 1
(15)
 2
 T
 bk = 2 2T f (t; α)sin(kωt + πα )dt if k ≥ 1
 R
T − 2
2
Proof
The new coefficients in the Fourier series expansion of f (t; α) given by (12)
are as follows:
T T T ∞
2 2 2
Z Z Z
2 2 πα 2 X πα
A0 = f (t; α)dt = a0 cos( )dt + an cos(nωt + )dt
T − T2 T − T2 2 T − T2 n=1
2
T ∞
2
Z 2 X πα πα
bn sin(nωt + )dt = 2a0 cos( )
T − T2 n=1
2 2
(16)

Multiplying the both-hand sides of f (t; α) by cos(kωt) and integrating over

period T , we have
T T
2 2
Z Z
2 2 πα
Ak = f (t; α)cos(kωt)dt = a0 cos( )cos(kωt)dt +
T − T2 T − T2 2
T ∞
2
Z 2 X πα
an cos(nωt + )cos(kωt)dt +
T − T2 n=1
2
T ∞
1
Z 2 X πα
bn sin(nωt + )cos(kωt)dt = (17)
T − T2 n=1
2
6 Fourier Series in Fractional Dimensional Space

0 if n 6= k
ak cos πα
2 + b k sin πα
2 if n = k
where k ≥ 1
Multiplying the both-hand sides of f (t; α) by sin(kωt) and integrating over
period T , we have
T Z T2
2 2
Z 2 πα
Bk = f (t; α)sin(kωt)dt = a0 cos( )sin(kωt)dt +
T − T2 T − T2 2
Z T2 X
∞
2 πα
an cos(nωt + )sin(kωt)dt +
T − T2 n=1 2
Z T2 X∞
2 πα
bn sin(nωt + )sin(kωt)dt = (18)
T − T2 n=1 2

0 if n 6= k
bk cos πα πα
2 − ak sin 2 if n = k
where k ≥ 1
Note that the coefficients ak and bk are obtained from (8) when using fractional
πα πα
basis functions ϕα α
1 = cos(nωt + 2 ) and ϕ2 = sin(nωt + 2 ) of the fractional
dimension space and can be realized as follows:
−1 
cos( πα πα
   
an
= 2 ) sin( 2 ) An
(19)
bn − sin( 2 ) cos( πα
πα
2 ) Bn

3 Complex Formulation
In continue, the FFS is generalized to complex plane.
Lemma2
(inωt)
Let φαn be the basis function of the fractional space and φn = e be the
basis function of the linear space in the complex domain, then
 
1 πα
φα
n = D α
φn = ei(nωt+ 2 ) (20)
(nω)α

Proof
The proof is easily made as follows. φα
n for n > 0 and n < 0 can be calculated
as:
   
1 1 (inωt) 1
α
φn = α
α
D φn = α
α
D e = (inω)α e(inωt)
(nω) (nω) (nω)α
(21)
1 α α (inωt) α (inωt) ( iπ ) α (inωt) i(nωt+ πα )
(nω) (i )e = (i )e = (e 2 ) e =e 2
(nω)α
 Fourier Series in Fractional Dimensional Space 7

and for φα
0 , we have:
   
πα
+inωt
φα
0 = D α
1 = D α
lim e = e+i 2 (22)
n→0

Based on Eqs. (3), (4) and (5) the basis functions ϕα α

i (i = 0, 1, 2) and φn in
real and complex domains respectively can be related to each other as follows:

φα α ∗
0 + (φ0 )
ϕα
0 = = ℜ{φα
0} (23)
2

φα α ∗
n + (φn )
ϕα
1 = = ℜ{φα
n} (24)
2

φα α ∗
n − (φn )
ϕα
2 = = ℑ{φα
n} (25)
2i
where ℜ and ℑ are real and imaginary part, respectively. Therefore the general
form of basis function can be realized as:
πα
φα i(nωt+ 2 )
n = e (26)

It can be observed that in-spite of validity properties of ϕ−n = (ϕn )∗ in

conventional FS ,however; this type of properties in fractional space is invalid.
πα
φα i(−nωt+ 2 ) 6= (φα ∗
−n = e n) (27)

Theorem2 P+∞ i(nωt+ πα

2 ) satisfy Dirichlet’s
Let the complex function f (t; α) = n=−∞ cα ne
sufficient conditions and be defined on [0, ∞) × [0, ∞) −→ C, then implies that

+∞ ∞
X πα πα X πα
f (t; α) = cα i(nωt+ 2 ) = a0 cos( )+ an cos(nωt + )+
ne
n=−∞
2 n=1
2
∞
X πα
bn sin(nωt + ) (28)
n=1
2

where the complex coefficients are:

πα − iπα

 cα
0 = a0 cos( 2 )e
2 n=0
α 1 (29)
cn = 2 (an − ibn ) if n ≥ 1
c−n = 21 (a−n + ib−n )e−iπα if n ≤ −1
 α
8 Fourier Series in Fractional Dimensional Space

Proof
The proof is easily made as follows:
For all n ∈ Z, we expand the right-hand side of (28) and show that it is
identical to the left one. Expanding the right-hand side gives us
∞ ∞
πα X πα X πα
f (t; α) = a0 cos( )+ an cos(nωt + )+ bn sin(nωt + )=
2 n=1
2 n=1
2
∞  i(nωt+ πα ) πα 
πα X e 2 + e−i(nωt+ 2 )
a0 cos( ) + an +
2 n=1
2
∞  i(nωt+ πα ) πα 
X e 2 − e−i(nωt+ 2 )
bn =
n=1
2i
∞ ∞
πα X an bn i(nωt+ πα ) X an bn πα
a0 cos( ) + ( + )e 2 + ( − )e−i(nωt+ 2 ) (30)
2 n=1
2 2i n=1
2 2i

Changing variable n by −N at the right term of the right-hand side of (30),

we have
∞
πα X an bn πα
f (t; α) = a0 cos( )+ ( + )ei(nωt+ 2 ) +
2 n=1
2 2i
=−1
NX
a−N b−N −i(−N ωt+ πα )
( − )e 2 =
2 2i
N =−∞
∞ n=−1
πα X an bn πα
X a−n b−n πα
a0 cos( ) + ( + )ei(nωt+ 2 ) + ( − )e−iπα ei(nωt+ 2 ) =
2 n=1
2 2i n=−∞
2 2i
+∞
πα
X
cα i(nωt+ 2 ) (31)
ne
n=−∞

iπα
Getting a0 cos( πα α 2
2 ) = c0 e for n = 0, the right-hand side of (31) would imply
the left hand-side of (28).
Comparing the both sides of (31) together, we find the coefficients given by
(29).

4 The relations between Fractional Derivative,

FS and FFS
If we take a fractional derivative from Eq.(1) , then we have
∞ ∞
A0 α  X X
Dα f (t) = An Dα ϕ1 + Bn Dα ϕ2
 
D ϕ0 + (32)
2 n=1 n=1
 Fourier Series in Fractional Dimensional Space 9

and consequently by applying the direct fractional derivative of basis functions

of cosine and sinus functions, we have:
∞ ∞
A0 α  X π X π
Dα f (t) = f (t; α) = D 1 + An cos(nωt + α) + Bn sin(nωt + α)
2 n=1
2 n=1
2
∞ ∞
A0 π X π X π
= cos( α) + An cos(nωt + α) + Bn sin(nωt + α) (33)
2 2 n=1
2 n=1
2

then, based on the theorem 1 it can be figured out in terms of a conventional

FS with new modified coefficients.
∞ ∞
A′0 X ′ X
Dα f (t) = f (t; α) = + An cos(nωt) + Bn′ sin(nωt)
2 n=1 n=1
(34)

where  ′
 A0 = A0 cos( πα
2 )
A′k = An cos( πα πα
2 ) + Bn sin( 2 ) if n ≥ 1 (35)
Bk = Bn cos( 2 ) − An sin( πα
πα
 ′
2 ) if n ≥ 1

References
[1] M.A. Hammad and R.R. Khalil,”Fractional Fourier Series with Applica-
tions”, Amer.J.Comput. Appl. Math., 4 (6), 187-191, 2014.

[2] M.-S. Hu, R. P. Agarwal, and X.-J. Yang, “Local fractional Fourier series
with Application to Wave Equation in Fractal Vibrating String,” Abs.
Appl. Analysis, 2012, Article ID 567401, 15 pages, 2012.

[3] G. Jumarie, ”Fourier’s transform of fractional order via Mittag-Leffler

function and modified Riemann-Liouville derivative,” J. Appl. Math.
Informatics, 26 (5–6), 1101-1121, 2008.

[4] P. R. Massopust and Ahmed I. Zayed. ”On the Invalidity of Fourier series
Expansions of Fractional Order.” Fract. Calc. and Appl. Analysis, 18
(6),1507-1517, 2015.

[5] A.V. Oppenheim, and A.S. Willsky, Signals and Systems, Prentice Hall,
p.198, 1997

[6] S.C. Pei, M.H. Yeh, and T.L. Luo, ”Fractional Fourier Series Expansion
for Finite Signals and Dual Extension to Discrete-Time Fractional Fourier
Transform”, IEEE Trans. Sig. Process. 47 (10), 2883-2888, 1999.
10 Fourier Series in Fractional Dimensional Space

[7] H. M. Srivastava, A. K. Golmankhaneh, D. Baleanu, and X.-J.Yang,

“Local Fractional Sumudu Transform with Application to IVPs on Can-
tor Sets,” Abs. and Appl. Analysis, vol. 2014, Article ID 620529, 7 pages,
2014.

[8] S.-Q. Wang, Y.-J. Yang, and H. K. Jassim, “Local Fractional Function
Decomposition Method for Solving Inhomogeneous Wave Equations with
Local Fractional Derivative,” Abs. and Appl. Analysis, 2014, Article ID
176395, 7 pages, 2014.

[9] Y.J. Yang and S.Q. Wang ”Local Fractional Fourier Series Method for
Solving Nonlinear Equations with Local Fractional Operators”, Math.
Prob. Eng. 2015, Article ID: 481905, 9 pages, 2015.

[10] C.Yu, ”Fractional Fourier Series and its Applications”, J. Physics: con-
ference series, 1976, 2021.