See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/381551935

ON INVERSE SOURCE PROBLEM FOR SOBOLEV EQUATION WITH MITTAG-

LEFFLER KERNEL IN L r SPACE

Article · January 2023

DOI: 10.54671/BMAA-2023-4-3

CITATIONS READS
0 38

4 authors, including:

Bui Duc Nam Le Xuan Dai

Ho Chi Minh City University of Industry and Trade Ho Chi Minh City University of Technology
9 PUBLICATIONS 8 CITATIONS 5 PUBLICATIONS 15 CITATIONS

SEE PROFILE SEE PROFILE

Le Dinh Long
Industrial University of Ho Chi Minh City
73 PUBLICATIONS 596 CITATIONS

SEE PROFILE

All content following this page was uploaded by Bui Duc Nam on 20 June 2024.

The user has requested enhancement of the downloaded file.

Bulletin of Mathematical Analysis and Applications
ISSN: 1821-1291, URL: http://www.bmathaa.org
Volume 15 Issue 4(2023), Pages 21-33
https://doi.org/10.54671/BMAA-2023-4-3

ON INVERSE SOURCE PROBLEM FOR SOBOLEV EQUATION

WITH MITTAG-LEFFLER KERNEL IN Lr SPACE

BUI DUC NAM, LE XUAN DAI, LE DINH LONG, NGUYEN HOANG TUAN

Abstract. In this paper, we consider a Sobolev equation with the Atangana-

Baleanu-Caputo fractional derivative. We give the explicit fomula of the source
term. Under the observations of tha data in Lr spaces, we provide a regularized
solution using Fourier truncated method. We give the error estimate between
the exact solution and the regularized solution. The main tool is of using some
embeddings.

1. Introduction
Let Ω ∈ RN (N ≥ 1) be a bounded domain with sufficiently smooth boundary
∂Ω. In this paper, we are interested to study time fractional diffusion equation
with fractional derivative as follows

ABC β


0
 Dt u(x, t) + mLu(x, t) + Lu(x, t) = g(t)f (x), x ∈ Ω, t ∈ (0, 1),
u(x, t) = 0, x ∈ ∂Ω, t ∈ (0, 1),

u(x, 1) = ρ(x), x ∈ Ω.

(1.1)
Here in the main equation as above, the Atangana - Baleanu fractional derivative
ABC β
0 Dt u(x, t) is defined by
Zt  −β(t − s)β 
ABC β M(β) ∂u(x, s)
0 Dt u(x, t) = Eβ,1 ds, (1.2)
1−β ∂s 1−β
0

where the normalization function M(β) can be any function satisfying the condi-
β
tions M(β) = 1 − β + Γ(β) , here M(0) = M(1) = 1 (see Definition 2.1 in [1]) and
Eβ,1 is the MittagLeffler function. Our main goal in this paper is of finding the
source term f (x) from the given data ρ and the measured data at the final time

2000 Mathematics Subject Classification. 35K05, 35K99, 47J06, 47H10x .

Key words and phrases. Source problem; Fractional pseudo-parabolic problem; Ill-posed
problem; Convergence estimates; Regularization, Mittag-Leffler kernel, Atagana-Baleanu-Caputo
derivative.
c 2023 Universiteti i Prishtinës, Prishtinë, Kosovë.
Submitted September 14, 2023. Published October 31, 2023.
Communicated by Douglas R. Anderson.
21
22 B.D. NAM, L.X. DAI, L.D. LONG, N.H. TUAN

u(x, 1) = ρ(x), ρ ∈ L2 (Ω) such that

kg − g kLr (0,1) + kρ − ρ kLr (Ω) ≤ . (1.3)
One of the narrow branches of fractional analysis is the theory of fractional diffusion
equations. Fractional-time diffusion equations are used to model complex phenom-
ena such as long-term memory or spatial interactions, non-local and local dynamics.
For details, please refer to documents [2, 3, 5, 6, 7, 8, 18]. One of the modern trends
in fractional analysis is the development of fractional operators with non-singular
kernels. The study of these fractional derivatives is important to satisfy the need
for modeling applications in various fields, such as fluids, mechanics, viscoelasticity,
biology, physics and engineering, see in [3, 4, 17, 9, 10, 11, 12, 13, 21, 22, 23, 24].
Several definitions of fractional derivatives have been given based on non-special
nuclei such as Atangana-Baleanu-Caputo fractions and derivatives. Regarding the
study of (1.1) problem with Atangana-Baleanu derivative, we list some previous
results as follows
• Under the case F = F (x, t), from [19], the kernels of the extended Mittag-
Leffler type functions are studied in this study using a partial differential
equation model with the new universal fractional derivatives. Analysis
and consideration are given to an initial boundary value problem for the
anomalous diffusion of fractional order. The Mittag-Leffler kernel fractional
derivative, also known as the Atangana-Babeanu fractional derivative in
time, is interpreted in the Caputo sense. They discovered findings on the
existence, uniqueness, and regularity of the solution.
• Under the case F = g(t)f (x), from [14, 16], the problem of determining in-
verse source problem for fractional diffusion equation containing Atangana-
Baleanu-Caputo fractional derivative. We first establish an explicit formula
of the source term from the average data of the function in the time variable.
We then show that the inverse source problem is ill-posed in the meaning
of Hadamard i.e., the source function is not stable according to the given
data. To overcome this instability, we propose a regularized method as
in the Fractional Landweber method. We also obtain the upper bounds
and find the convergence rate between the regularized solution and sought
source function. Estimates are also derived in two cases on selection rules,
a priori parameter, and a posterior parameter. Numerical examples are
given which illustrate the usefulness of our method.
• Under the case F = F (x, t, u(x, t)), in the paper [15], they investigated a
nonlinear time fraction Volterra equation with a Mittag-Leffler multiplier in
Hilbert space. By applying the properties of the Mittag-Leffler function and
the eigenvalue expansion, the existence of a light solution to our problem
has been proved. The main tool to prove our results is the use of some
Sobolev embeddings and some fixed point theorems.
As we know, the inverse issue for diffusion equation with Atangana-Babeanu frac-
tional derivative where the observed data is in the Lp (Ω) space with p 6= 2 is solved
for the first time in this study. One significant challenge is that because the data
is not in L2 (Ω), we cannot utilize Parseval equality directly. We get around these
issues by leveraging embedding between Lp (Ω) and Hs (Ω) scale-spaces. We have
the regularized solution

through the Fourier series truncation method with the ob-
served data ϕ , g ∈ Lr (0, T ) × Lr (Ω). After that, the error established between
 ON INVERSE SOURCE PROBLEM FOR SOBOLEV EQUATION 23

the regularized solution and the exact solution in the Theorem (3.1), by the main
analytical technique is to use some embeddings and some evaluations using Hlder
inequality.
The structure of our paper is as follows. The existence of mild solution u to
(1.1) in Section 1.In the Section 2, we have some preliminaries. The main results
in Section 3 is Theorem 3.1, our main tool is Sobolev embeddings.

2. Preliminaries
Let us recall that the spectral problem
(
−Lϕn (x) = ξn ϕn (x), in Ω,
(2.1)
ϕn (x) = 0, on ∂Ω,
admits a family of eigenvalues
0 < ξ1 ≤ ξ2 ≤ ξ3 ≤ ... ≤ ξn ≤ ... % ∞.
For all r ≥ 0, the operator Lr (here L = −∆) also possesses the following
representation:
X∞ Z 
r
L h= h(x)ϕn (x)dx ξnr ϕn ,
n=1 Ω
 ∞ Z
X 2 
r 2 2r
h ∈ H (Ω) = h ∈ L (Ω) : h(x)ϕn (x)dx ξn < ∞ . (2.2)
n=1 Ω
Consider on Hr the norm
∞ Z
X 2  21
khkHr (Ω) = h(x)ϕn (x)dx ξn2r , h ∈ H(Ar ).
n=1 Ω

Lemma 2.1. ([6]) Let 0 < β < 1, then there exist 0 < B1 , B2 , B3 such that
B1 B2 B3
≤ Eβ,1 (−y) ≤ , Eβ,α (−y) ≤ , for all y ≥ 0, α ∈ R. (2.3)
1+y 1+y 1+y
Lemma 2.2. For ξ > 0, β > 0, m ∈ N∗ , we have
dm
Eβ,1 (−ξtβ ) = −ξtβ−m Eβ,β−m+1 (−ξtβ ),
dtm
d
tEβ,2 (−ξtβ ) = Eβ,1 (−ξtβ ),


dt
d β−1
Eβ,β (−ξtβ ) = −tβ−2 Eβ,β−1 (−ξtβ ).


t (2.4)
dt
Lemma 2.3. [6] For t > 0, and ξ > 0, and 0 < β < 1, then one has
∂tβ Eβ,1 (−ξtβ ) = − ξEβ,1 (−ξtβ ),
M(β)
Lemma 2.4. For β ∈ (0, 1), putting A3,n (m, β) =
, it
ξn M(β) + ξn σn (1 − β)
gives
1 mM(β)
A3,n (m, β) ≥  · (2.5)
ξn M(β) + (1 − β)
ξ1 σ1
24 B.D. NAM, L.X. DAI, L.D. LONG, N.H. TUAN

ξn 1 1
Proof. First of all, we notice that ξn σn = 1+mξn ≥ 1
+m
= ξ1−1 +m
, and we have
ξ1

M(β) M(β)
A3,n (m, β) =
≥ 2
 
ξn M(β) + ξn σn (1 − β) ξn M(β)
+ (1 − β)
1+mξn ξ n σn

mM(β) 1 mM(β)
≥  ≥  · (2.6)
ξn M(β)
+ (1 − β) ξn M(β)
+ (1 − β)
ξ n σn ξ1 σ1

Lemma 2.5. For M > 0, by Lemmas 2.1 and 2.2, we have

Z M Z M
tα−1 Eα,α (−λn tα ) dt = tα−1 Eα,α (−λn tα ) dt
0 0
Z η
1 d 1
=− Eα,1 (−λn tα ) dt = (1 − Eα,1 (−λn M α )) . (2.7)
λn 0 dt λn

Lemma 2.6. Let β ∈ (0, 1), we have estimate

β(ξ1−1 + m)−1
 
1 mM(β) 
  1 − Eβ,1 − −1
ξn M(β) + (1 − β) M(β) + ξ1−1 + m


ξ1 σ1
(1 − β)
Z1
σn βM(β)  βξn σn (1 − s)β  1
≤
2 Eβ,β − (1 − s)β−1 ds ≤ ·
M(β) + σn ξn (1 − β) M(β) + σn ξn (1 − β) ξn
0
(2.8)

Proof. For Eβ,β (−z) ≥ 0 for 0 < β < 1 and z ≥ 0, and using the Lemmas 2.4 and
2.6, we obtain

Z1
σn βM(β) βξn σn (1 − s)β  
a)
2 Eβ,β − (1 − s)β−1 ds
M(β) + σn ξn (1 − β) M(β) + σ n ξn (1 − β)
0
β(ξ1−1 + m)−1
 
1 mM(β) 
≥   1 − Eβ,1 − −1 . (2.9)
ξn M(β) + (1 − β) M(β) + ξ1−1 + m


ξ σ
(1 − β)
1 1

Z1
σn βM(β)  βξn σn (1 − s)β 
b)
2 Eβ,β − (1 − s)β−1 ds
M(β) + σn ξn (1 − β) M(β) + σn ξn (1 − β)
0
Z1
M(β) d βξn σn (1 − s)β

=−
Eβ,1 − dτ
ξn M(β) + ξn σn (1 − β) ds M(β) + σn ξn (1 − β)
0
1 βξn σn
 1
= 1 − Eβ,1 − ≤ · (2.10)
ξn M(β) + σn ξn (1 − β) ξn


 ON INVERSE SOURCE PROBLEM FOR SOBOLEV EQUATION 25

Lemma 2.7. [17] The following statement are true:

N 2N

Lr (Ω) ,→ Hs (Ω), if − < s ≤ 0, r≥ , 
N − 4s 

4
(2.11)
s r N 2N 
H (Ω) ,→ L (Ω), if 0≤s< , r≤ . 

4 N − 4s

3. The inverse source problem (1.1)

Let us first to review the initial value problem as follows

ABC β


0
 Dt u(x, t) + mLu(x, t) + Lu(x, t) = F (x, t), in Ω × (0, 1],
u(x, t) = 0, on ∂Ω × (0, 1], (3.1)

u(x, 0) = u0 (x), in Ω,


∞
X
where u0 and F are given functions. Let u(x, t) = un (t)ϕn (x) be the Fourier
Z j=1

series in L2 (Ω) with un (t) = u(x, t)ϕn (x)dx, then we have the fractional integro-
Ω
differential equation involving the Atangana-Baleanu fractional derivative in the
form

ABC β


0 Dt 1 + mξn un (t) + ξn un (t) = Fn (t), (3.2)

in [20], the solution (3.1) can be represented as by Fourier series

∞ Z
X 
u(x, t) = u(x, t)ϕn (x)dx ϕn (x)
n=1 Ω

and then given by

ξn
M(β)  −β 1+mξ n
tβ 
un (t) = ξn

Eβ,1
ξn

u0,n
M(β) + 1+mξ n
1−β M(β) + 1+mξ n
1−β
 1  1−β
+ ξn

Fn (t)
1 + mξn M(β) + 1+mξ 1 − β
n
 1  βM(β)
+
1 + mξn M(β) + ξn 1 − β 2



1+mξn
Zt  ξn
−β 1+mξ (t − s)β 
× Eβ,β n
ξn

(t − s)β−1 Fn (τ )dτ. (3.3)
M(β) + 1+mξn 1−β
0
26 B.D. NAM, L.X. DAI, L.D. LONG, N.H. TUAN

Let us denote σn = (1 + mξn )−1 , this implies that

∞
X M(β)  −βξn σn tβ 
u(x, t) =
Eβ,1
u(0), ϕn ϕn (x),
n=1
M(β) + ξn σn 1 − β M(β) + ξn σn 1 − β
∞

X σn 1 − β
+
F (·, t), ϕn ϕn (x)
n=1
M(β) + ξn σn 1 − β
∞
X σn βM(β)
+

2
n=1 M(β) + ξn σn 1 − β
Zt  −βξn σn (t − τ )β 
× Eβ,β
(t − s)β−1 F (·, s), ϕn dsϕn (x)· (3.4)
M(β) + ξn σn 1 − β
0
Let us now return the problem of identifying the source term. Let t = 1, u(x, 0) =
0, F (x, t) = g(t)f (x), and Fn (1) = 0, we get
Z Z
σn βM(β)
ρ(x)ϕn (x)dx = f (x)ϕn (x)dx

2
M(β) + ξn σn 1 − β
Ω Ω
Z1  −βξn σn (1 − s)β 
× Eβ,β
(1 − s)β−1 g(s)ds. (3.5)
M(β) + ξn σn 1 − β
0
To make the formula even more compact, we put
σn βM(β)
A1,n (m, β) =
2 ,
M(β) + ξn σn (1 − β)
βξn σn
A2,n (m, β) = . (3.6)
M(β) + ξn σn (1 − β)
From (3.5) and (3.6), we receive
Z Z
ρ(x)ϕn (x)dx = f (x)ϕn (x)dx
Ω Ω
Z1
Eβ,β − A2,n (m, β)(1 − s)β (1 − s)β−1 g(s)ds.


× A1,n (m, β) (3.7)
0
From (3.7), it gives
R
Z ρ(x)ϕn (x)dx
Ω
f (x)ϕn (x)dx = ·
R1

Ω A1,n (m, β) Eβ,β − A2,n (m, β)(1 − s)β (1 − s)β−1 g(s)ds
0
(3.8)
Through some basic transformations, we get
R
X∞ ρ(x)ϕn (x)dx
Ω
f (x) = ϕn (x).
R1

n=1 A (m, β) E − A (m, β)(1 − s) β (1 − s)β−1 g(s)ds
1,n β,β 2,n
0
(3.9)
 ON INVERSE SOURCE PROBLEM FOR SOBOLEV EQUATION 27

 R1
−1
As n → ∞,i.e., A1,n (m, β) Eβ,β − A2,n (m, β)(1 − s)β (1 − s)β−1 g(s)ds
0
→ ∞, see in the Lemma 2.6. Thus, it can be concluded from formula (3.9) that the
small perturbation of ρε (x) will cause a great change of f (x). Thus our problem
(1.1) is ill-posed. Next, we will give the conditional stability results of the source
term f (x).
Theorem 3.1. Let us take (g , ρ ) ∈ Lr (0, 1) × Lr (Ω) such that g (t) > G2 > 0 for
any 0 ≤ t ≤ 1 for any β1 < r < 2 and condition

g − g Lr (0,1)
+ ρ − ρ Lr (Ω)
≤ . (3.10)

Assume that f ∈ H(Aj+k ) for k > 0 and 0 < j < N4 . With the Fourier truncation
method, we have
R
X ρ (x)ϕn (x)dx
Ω
fC (x) = ϕn (x).
R1

ξn ≤C A β
Eβ,β − A2,n (m, β)(1 − s) (1 − s) β−1
1,n (m, β) g (s)ds
0
(3.11)
Then we have
fC − f 2N . |C |−k f Hj+k (Ω)
+ C  A4 (B3 , r, β, G2 , m, ξ1 ) kf kHj (Ω)
L N −4j (Ω)
N
j+1+ 2r −N
+ A5 (ξ1 , σ1 , m, β) (C ) 4
, (3.12)
whereby C satisfies that
 N

j+1+ 2r −N
lim C  = lim (C ) 4
 = 0, lim C = +∞. (3.13)
→0 →0 →0

Remark. We can take C satisfying (3.13) as follows

s−1
N N
C =  j+1+ 2r − 4 , 0 < s < 1.

Then the error order fC − f 2N is of order

L N −4j (Ω)

N −N s+j+ N − N
n k(j+1+ 2r 2s 4 o
4 N −N s
max  s−1 , j+1+
2s 4 , ·

Proof. Using the triangle inequality, we have

fC − f Hj (Ω)
≤ F2, − f Hj (Ω)
+ F1, − F2, Hj (Ω)
+ F1, − f Hj (Ω)
,
| {z } | {z } | {z }
(3.14)
where we denote some following functions
R
X ρ(x)ϕn (x)dx
Ω
F1, (x) = ϕn (x),
R1

ξn ≤C A (m, β) E − A (m, β)(1 − s)β (1 − s)β−1 g (s)ds
1,n β,β 2,n 
0
(3.15)
28 B.D. NAM, L.X. DAI, L.D. LONG, N.H. TUAN

and
R
X ρ(x)ϕn (x)dx
Ω
F2, (x) = ϕn (x).
R1

ξn ≤C A1,n (m, β) Eβ,β − A2,n (m, β)(1 − s)β (1 − s)β−1 g(s)ds
0
(3.16)

Let us next consider some terms on the right hand side of (3.14).
Step 1. Estimate of F2, − f Hj (Ω) .
Let us recall the function f as follows.
R
∞
X ρ(x)ϕn (x)dx
Ω
f (x) = ϕn (x).
R1

n=1 A1,n (m, β) Eβ,β − A2,n (m, β)(1 − s)β (1 − s)β−1 g(s)ds
0

This expression together with the fomula (3.16) gives us the claim of the following
difference

f (x) − F2, (x)

R
X ρ(x)ϕn (x)dx
Ω
= ϕn (x)
R1

A1,n (mβ) Eβ,β − A2,n (m, β)(1 − s)β (1 − s)β−1 g(s)ds
ξn ≥C
0
X Z 
= f (x)ϕn (x)dx ϕn (x). (3.17)
ξn ≥C Ω

The norm on Hj (Ω) of (3.17) is calculated through the Parseval equality as follows
Z 2
2 X
f − F2, Hj (Ω)
= ξn2j f (x)ϕn (x)dx
ξn ≥C Ω
X Z 2
= ξn−2k ξn2j+2k f (x)ϕn (x)dx .
ξn ≥C Ω

It is easy to see that ξn−2k ≤ |C |−2k if ξn > C and k > 0. Hence, we have
Z 2
2 X
f − F2, Hj (Ω)
≤ |C |−2k ξn2j+2k f (x)ϕn (x)dx
ξn ≥C Ω

−2k 2
= C f Hj+k (Ω)
, (3.18)

It gives that

f − F2, Hj (Ω)
≤ |C |−k f Hj+k (Ω)
. (3.19)
 ON INVERSE SOURCE PROBLEM FOR SOBOLEV EQUATION 29

Step 2. Estimate of F1, − F2, Hj (Ω) .

Based on two formulas (3.15) and (3.16), we have

F1, (x) − F2, (x)

R1


X Eβ,β − A2,n (m, β)(1 − s)β (1 − s)β−1 g (s) − g(s) ds
0
=
R1

ξn ≤C Eβ,β − A2,n (m, β)(1 − s)β (1 − s)β−1 g (s)ds
0
R
ρ(x)ϕn (x)dx
Ω
× ϕn (x)· (3.20)
R1

A1,n (m, β) Eβ,β − A2,n (m, β)(1 − s)β (1 − s)β−1 g(s)ds
0

We follows from (3.20) that

F1, (x) − F2, (x)

R1


X Eβ,β − A2,n (m, β)(1 − s)β (1 − s)β−1 g (s) − g(s) ds
0
=
R1

ξn ≤C Eβ,β − A2,n (m, β)(1 − s)β (1 − s)β−1 g (s)ds
0
Z 
× f (x)ϕn (x)dx ϕn (x). (3.21)
Ω

By taking the norm of (3.21) in space Hj (Ω) and using Parseval’ s equality, we
provide that
2
F1,ε − F2,ε Hj (Ω)
R1


" Eβ,β − A2,n (m, β)(1 − s)β (1 − s)β−1 g (s) − g(s) ds #2
X 0
=
R1

ξn ≤C Eβ,β − A2,n (m, β)(1 − s)β (1 − s)β−1 g (s)ds
0
Z 2
× ξn2j f (x)ϕn (x)dx . (3.22)
Ω

1
From (3.22), noting that r > β −1 and r∗ = 1 + , using Hölder inequality and
r−1
Lemma 3.10, we have
Z 1
Eβ,β − A2,n (m, β)(1 − s)β (1 − s)β−1 g (τ ) − g(τ ) dτ



0
Z 1  r1  Z 1  r1∗
∗
r β

β−1 r
≤ g (s) − g(s) ds Eβ,β − A2,n (1 − s) (1 − s) ds
0 0
r−1  r − 1  r−1
 r
r−1  r r
≤ g − g Lr (0,1)
B3r−1 = kg − gkLr (0,1) B3 · (3.23)
βr − 1 βr − 1
30 B.D. NAM, L.X. DAI, L.D. LONG, N.H. TUAN

This implies that

Z 1  r − 1  r−1
r
Eβ,β − A2,n (m, β)(1 − s)β (1 − s)β−1 g (s) − g(s) ds ≤ B3



.
0 βr − 1
(3.24)

It is easy to see that

1 M(β) + ξn σn (1 − β) M(β) 1  M(β) 

= ≥ ≥ · (3.25)
A2,n (m, β) βξn σn βξn σn ξn β

Next, the function g ≥ G2 , and using the Lemma 2.6, we have

Z1
Eβ,β − A2,n (m, β)(1 − s)β (1 − s)β−1 g (s)ds

0
Z1
Eβ,β − A2,n (1 − s)β (1 − s)β−1 ds


≥ G2
0
G2 M(β) h  β(ξ1−1 + m)−1 i
= 1 − Eβ,1 − −1 . (3.26)
ξn β M(β) + ξ1−1 + m


(1 − β)

From (3.24) and (3.26), we assert that

The right hand side of (3.22)

 −1
" #
 r − 1  r−1
r β  β(ξ1−1 + m)−1
≤ ξn  B3 1 − Eβ,1 −
−1 .
βr − 1 G2 M(β) M(β) + ξ1−1 + m (1 − β)
| {z }
A4 (B3 ,r,β,G2 ,m,ξ1 )
(3.27)

Combining (3.22) and (3.27), we find that

2 2
F1, − F2, Hj (Ω)
≤ A4 (B3 , r, β, G2 , m, ξ1 ) 2
X Z 2
2j+2
× ξn f (x)ϕn (x)dx . (3.28)
ξn ≤C Ω

The finite sum on the right above can be bounded as follows

X Z 2 X Z 2
2j+2 2 2j
ξn f (x)ϕn (x)dx ≤ |C | ξn f (x)ϕn (x)dx
ξn ≤C Ω ξn ≤C Ω
2
≤ |C |2 f Hj (Ω)
. (3.29)

Therefore, we follows from (3.28) that

F1, − F2, Hj (Ω)

≤ C  A4 (B3 , r, β, G2 , m, ξ1 ) f Hj (Ω)
· (3.30)
 ON INVERSE SOURCE PROBLEM FOR SOBOLEV EQUATION 31

Step 3. Estimate of F1 − f Hj (Ω)

. We derive that

F1, (x) − f (x)

X  Z1 −1
β β−1


= A1,n (m, β) Eβ,β − A2,n (m, β)(1 − s) (1 − s) g (s)ds
ξn ≤C 0
Z

× ρ (x) − ρ(x) ϕn (x)dx ϕn (x). (3.31)
Ω

By taking the norm of both sides of the above expression in space Hj (Ω), and using
Parseval’ s equality, we obtain that
2
F1, − f Hj (Ω)

X  Z1 −2
A1,n (m, β) Eβ,β − A2,n (m, β)(1 − s)β (1 − s)β−1 g (s)ds


=
ξn ≤C 0
Z 2
ξn2j


× ρ (x) − ρ(x) ϕn (x)dx . (3.32)
Ω

By looking back the inequality (3.26), we get

2
F1, − f Hj (Ω)

i −1
!
mM(β) h β(ξ1−1 + m)−1

≤ M(β)
1 − Eβ,1 −
−1
ξ1 σ1 + (1 − β) M(β) + ξ1−1 + m (1 − β)
X Z  2
ξn2j+2


× ρ (x) − ρ(x) ϕn (x)dx . (3.33)
ξn ≤C Ω

Form (3.33), one has

X Z 2
2j+2


ξn ρ (x) − ρ(x) ϕn (x)dx
ξn ≤C Ω
N r−2N Z 2
2j+2+ N N
r − 2
X

= ξn ξn 2r ρ (x) − ρ(x) ϕn (x)dx
ξn ≤C Ω
N r−2N Z 2
2j+2+ N N
r − 2
X

≤ (C ) ξn 2r
ρ (x) − ρ(x) ϕn (x)dx
ξn ≤C Ω

2j+2+ N −N 2
= (C ) r 2
ρ − ρ H
N r−2N
4r (Ω)
. (3.34)
N r−2N
Since 1 < r < 2, with Lr (Ω) ,→ H 4r (Ω). Therefore, we get

ρ − ρ N r−2N ≤ C1 (N, r) ρ − ρ Lr (Ω)

≤ C1 (N, r). (3.35)
H 4r (Ω)

By summarizing all three evaluations (3.33), (3.34) and (3.35), we derive that
N
j+1+ 2r −N
F1, − f Hj (Ω)
≤ A5 (ξ1 , σ1 , m, β) (C ) 4
, (3.36)
32 B.D. NAM, L.X. DAI, L.D. LONG, N.H. TUAN

whereby
A5 (ξ1 , σ1 , m, β)
!−1
mM(β) h  β(ξ1−1 + m)−1 i
= M(β)
1 − Eβ,1 − −1 · (3.37)
M(β) + ξ1−1 + m


ξ1 σ1 + (1 − β) (1 − β)
From (3.19) to (3.37), we can conclude that
fC − f Hj (Ω)
≤ |C |−k f Hj+k (Ω)
+ C  A4 (B3 , r, β, G2 , m, ξ1 ) kf kHj (Ω)
N
j+1+ 2r −N
+ A5 (ξ1 , σ1 , m, β) (C ) 4
. (3.38)
2N
N
By using Lemma 2.7, since 0 < j < 4, we know that Hj (Ω) ,→ L N −4j (Ω), which
yields to the desired result (3.38). 

Acknowledgments. The authors would like to thank the anonymous referee for
his/her comments that helped us improve this article.

References
[1] G.M. Bahaa, A. Hamiaz, Optimality conditions for fractional differential inclusions with
nonsingular MittagLeffler kernel. Adv Differ Equ (2018), 257.
[2] Paulo Mendes de Carvalho-Neto and Gabriela Planas, Mild solutions to the time
fractionalNavier-Stokes equations in RN , J. Differential Equations 259, no. 7, (2015),
29482980.
[3] F. Mainardi, Fractional calculus and waves in linear viscoelasticity, Imperial CollegePress,
London, 2010. An introduction to mathematical models.
[4] M.M. Khader and K. M. Saad, On the numerical evaluation for studying the fractional
KdV,KdV-Burgers, and Burgers equations, Eur. Phys. J. Plus 133 (2018), 113.
[5] P. Chen and L. Y. Xiang, Existence of mild solutions for fractional evolution equations with
mixed monotone nonlocal conditions, Z. Angew. Math. Phys. 65, no. 4, (2014), 711728.
[6] Igor Podlubny, Fractional differential equations, Mathematics in Science and Engineering,
Vol. 198, Academic Press, Inc., San Diego, CA, 1999.
[7] N.H. Luc, L.N. Huynh, D. Baleanu, & N.H. Can. Identifying the space source term problem
for a generalization of the fractional diffusion equation with hyper-Bessel operator. Advances
in Difference Equations, 2020, 1-23.
[8] N.H. Can, O. Nikan, M.N. Rasoulizadeh, H. Jafari, & Y.S. Gasimov, Y. S., Numerical
computation of the time non-linear fractional generalized equal width model arising in shallow
water channel. Thermal Science, (2020), 49-58.
[9] N. H. Tuan, A. T. Nguyen, and C. Yang, Global well-posedness for fractional Sobolev-Galpern
type equations, DiscreteContin. Dyn. Syst. 42, no. 6, (2022), 26372665.
[10] N. A. Tuan, N. H. Tuan, and C. Yang, On Cauchy problem for fractional parabolic-elliptic
Keller-Segel model, Adv. Nonlinear Anal. 12 (2023), no. 1, 97116.
[11] N. H. Tuan, V. V. Au, and A. T. Nguyen, Mild solutions to a time-fractional Cauchy problem
with nonlocal nonlinearity inBesov spaces, Arch. Math. 118, no. 3, (2022), 305314.
[12] N. H. Tuan, M. Foondun, T. N. Thach, and R. Wang, On backward problems for stochastic
fractional reaction equations withstandard and fractional Brownian motion, Bull. Sci. Math.
179, no. 103-158, (2022) 58 pp.
[13] N. T. Bao, T. Caraballo, N. H. Tuan, and Y. Zhou, Existence and regularity results for termi-
nal value problem for nonlinear fractional wave equations, Nonlinearity 34 (2021), 14481503.
[14] N.D. Phuong, H.T.K. Van, N.H. Luc, L.D. Long. Determine the unknown source term for a
fractional order parabolic equation containing the Mittag-Leffler kernel. Journal of Nonlinear
and Convex Analysis, 23 (8), (2022), 1577-1600.
[15] T. Caraballo, T.B. Ngoc, N.H. Tuan, & R. Wang. On a nonlinear Volterra integrodifferen-
tial equation involving fractional derivative with Mittag-Leffler kernel, Proceedings of the
American Mathematical Society, 149(08), (2021), 3317-3334.
 ON INVERSE SOURCE PROBLEM FOR SOBOLEV EQUATION 33

[16] N.H. Can, N.H. Luc, D. Baleanu, Y. Zhou Y. Inverse source problem for time fractional
diffusion equation with Mittag-Leffler kernel. Advances in Difference Equations, 1, (2020)
1-18.
[17] N.H. Tuan, T. Caraballo. On initial and terminal value problems for fractional nonclassical
diffusion equations. Proceedings of the American Mathematical Society, 149(1),(2021), 143-
161.
[18] N.H. Tuan, T.N. Thach, N.H. Can, & D. O’Regan. Regularization of a multidimensional
diffusion equation with conformable time derivative and discrete data. Mathematical Methods
in the Applied Sciences, 44(4), (2021), 2879-2891.
[19] T.B. Ngoc, D. Baleanu, L.T.M. D., & N.H.Tuan, Regularity results for fractional diffusion
equations involving fractional derivative with MittagLeffler kernel. Mathematical Methods in
the Applied Sciences, 43(12), (2020), 7208-7226.
[20] F. S. A. Musalhi, S. A. S. Nasser and K. Erkinjon, Initial and boundary value problems
for fractional differential equations involving Atangana-Baleanu derivative, SQU J. Sci. 23
(2018),137146
[21] M.Q. Vinh, E. Nane, D. O’Regan, N.H. Tuan, Terminal value problem for a nonlinear para-
bolic equation with Gaussian white noise, Electronic Research Archive, Volume 30, Issue 4,
(2020) 1374-1413.
[22] N.H. Tuan, T. Caraballo, P.T.K. Van, V.V. Au. On a terminal value problem for parabolic
reactiondiffusion systems with nonlocal coupled diffusivity terms, Communications in Non-
linear Science and Numerical Simulation, (2022), 106248
[23] N.H. Tuan, D. Lesnic, T.N. Thach, & T.B. Ngoc. Regularization of the backward stochastic
heat conduction problem. Journal of Inverse and Ill-posed Problems, 30(3), (2022), 351-362.
[24] N.H. Tuan, N.M. Hai, T.N. Thach, & N.H. Can. On stochastic elliptic equations driven
by Wiener process with non-local condition. Discrete and Continuous Dynamical Systems-S,
(2022).

Bui Duc Nam

Ho Chi Minh City University of Industry and Trade
E-mail address: [email protected]

Le Xuan Dai
Faculty of applied science, Ho Chi Minh City University of Technology, Ho Chi Minh
City, Viet Nam
E-mail address: [email protected]

Le Dinh Long
Falculty of Maths, FPT University HCM, Saigon Hi-tech Park, Ho Chi Minh City, Viet
Nam
E-mail address: [email protected]

Nguyen Hoang Tuan∗

Faculty of applied science, Ho Chi Minh City University of Technology, Ho Chi Minh
City, Viet Nam
E-mail address: Corresponding author : [email protected]

View publication stats