TWMS J. App. Eng. Math. V.9, No.1, Special Issue, 2019, pp.

9-21

DIRECT AND INVERSE PROBLEMS FOR DIFFUSION OPERATOR

WITH DISCONTINUITY POINTS

A. ERGUN1 , R. KH. AMIROV2 , §

Abstract. In this study, the diffusion operator with discontinuity points has been con-
sidered. Under certain initial and jump conditions, integral equations have been derived
for solutions and integral representation have been presented. Some important spectral
properties of eigenvalue and eigenfunctions have been obtained. Reconstruction of the
diffusion operator with discontinuity points problem have been proved by Weyl function,
spectral datas and two sectra.

Keywords: Integral equation, Sturm-Liouville, Diffusion operator, inverse problems.

AMS Subject Classification: 34K08, 34L05, 34L10, 34E05

1. Introduction
Let’s define the following boundary value problem;
l (y) := −y 00 + [2λp (x) + q (x)] y = λ2 y, x ∈ (0, a1 ) ∪ (a1 , a2 ) ∪ (a2 , π) (1)
U (y) = y 0 (0) = 0, V (y) = y (π) = 0 (2)
y (a1 + 0) = α1 y (a1 − 0) (3)
y 0 (a1 + 0) = β1 y 0 (a1 − 0) + iλγ1 y (a1 − 0) (4)
y (a2 + 0) = α2 y (a2 − 0) (5)
0 0
y (a2 + 0) = β2 y (a2 − 0) + iλγ2 y (a2 − 0) (6)
where λ is a spectral parameter, q(x)∈ L2 [0, π], p(x) ∈W21 [0, π]
, a1 , a2 ∈ (0, π),a1 < a2 ,
2 2 2 2 1
|α1 − 1| + γ1 6= 0,|α2 − 1| + γ2 6= 0, βi = αi (i = 1, 2) .
The fundamental studies on the spectral theory of the Sturm-Liouville equations were
performed by Bernoulli, Euler, Sturm and Liouville. The first study, which is considered
to be the beginning of the inverse problem theory for the differential equations, was put
forth by Ambartsumyan. A lot of study were done about the inverse problem in [1 − 15].
1
Cumhuriyet University- Vocational School of Sivas, 58140 Sivas, Turkey.
e-mail: [email protected]; ORCID: https://orcid.org/0000-00002-2795-8097.
2
Cumhuriyet University- Faculty of Science and Arts-Department of Mathematics, 58140 Sivas, Turkey.
e-mail: [email protected]; ORCID: https://orcid.org/0000-0001-6754-2283,
§ Selected papers of International Conference on Life and Engineering Sciences (ICOLES 2018), Kyrenia,
Cyprus, 2-6 September, 2018.
TWMS Journal of Applied and Engineering Mathematics, Vol.9, No.1, 2019, Special Issue; c Işık
University, Department of Mathematics; all rights reserved.

9
10 TWMS J. APP. ENG. MATH. V.9, N.1, SPECIAL ISSUE, 2019

The inverse problem is called reconstruction of the operator, whose spectral characteristics
are given in sequences. Spectral theory has a wide usage and plays an important role in
engineering and geophysics and natural sciences.
Jaulent and Jean have investigated spectral and inverse problems of diffusion operators
in[3] . In [6] , Gasymov and Guseinov have focused on spectral theory of diffusion opera-
tors.
The inverse quasiperiodic problem for the diffusion operator was studied by Nabiev [8].
Inverse spectral problem for pencils of differential operators on a finite interval from the
Weyl function was studied Buterin and Yurko in [19] . In [20] , another considerable
paper have performed by Panakhov and Koyunbakan, the half inverse problem for diffusion
operators have been considered. Yang investigated the inverse nodal problem for diffusion
operators in [21] .

2. Preliminaries.
In this section, we will recall some basic definitions and concepts in order to fresh memories.
Let ϕ (x, λ) , ψ (x, λ) be solution of (1) respectively under the initial conditions
φ (0, λ) = 1, φ0 (0, λ) = 0 (7)
0
ψ (π, λ) = 0, ψ (π, λ) = 1 (8)
and discontinuity conditions (3) , (6) .
The function ϕ (x, λ) provides (9) , (10) , (11).
In this case, if 0 ≤ x < a1 , then we have
1 x
Z
φ (x, λ) = eiλx + sin λ (x − t) χ (t) y (t, λ) dt, (9)
λ 0
if a1 < x < a2 , we have
φ (x, λ) = 21 (α1 + β1 ) eiλx + 21 (α1 − β1 ) eiλ(2a1 −x) + γ21 eiλx − γ21 eiλ(2a1 −x)
Ra
+ 21 (α1 + β1 ) 0 1 sin λ(x−t)
λ χ (t) y (t, λ) dt
R a1 sin λ(x+t−2a 1)
Ra (10)
1
− 2 (α1 − β1 ) 0 λ χ (t) y (t, λ) dt − i γ21 0 1 cos λ(x−t)
λ χ (t) y (t, λ) dt
γ1 R a1 cos λ(x+t−2a1 ) R x sin λ(x−t)
+i 2 0 λ χ (t) y (t, λ) dt + a1 λ χ (t) y (t, λ) dt.
if a2 < x ≤ π, we can write
γ1 α+
φ (x, λ) = α1+ α2+ eiλx + α1− α2− eiλ(2a1 −2a2 +x) + α1+ α2− eiλ(2a2 −x) + α1− α2+ eiλ(2a1 −x) + 2 iλx
2 e
− − +
γ α γ α γ α
− 12 2 eiλ(2a1 −2a2 +x) + 12 2 eiλ(2a2 −x) − 12 2 eiλ(2a1 −x) − γ14γ2 eiλx − γ14γ2 eiλ(2a1 −2a2 +x)
−
γ α +
γ α γ α +
γ α−
− 22 1 eiλ(2a2 −x) − 22 1 eiλ(2a1 −x) + 22 1 eiλx + 22 1 eiλ(2a1 −2a2 +x) − γ14γ2 eiλ(2a2 −x)

Ra
+ γ14γ2 eiλ(2a1 −x) + α1+ α2+ + γ14γ2 0 1 sin λ(x−t) λ χ (t) y (t, λ) dt
+ − γ1 γ2
R a1 sin λ(x+t−2a2 )
+ −α1 α2 + 4 0 λ χ (t) y (t, λ) dt
γ1 γ2
R a1 sin λ(x+t−2a1 ) Ra
− +
+ −α1 α2 − 4 0 λ χ (t) y (t, λ) dt − α2− a12 sin λ(x+t−2a
λ
2)
χ (t) y (t, λ) dt
− − γ1 γ2
R a1 sin λ(2a1 −2a2 +x−t) + a2 sin λ(x−t)
R
+ α1 α2 − 4 λ χ (t) y (t, λ) dt + α2 a1 λ χ (t) y (t, λ) dt
i + +

0R a1 cos λ(x−t)
− 2 γ1 α2 − γ2 α1 0 χ (t) y (t, λ) dt
i − −

R a1 cosλλ(x+t−2a2 )
+ 2 −γ1 α2 − γ2 α1 0 λ χ (t) y (t, λ) dt

R a1 cos λ(x+t−2a 1)
Ra
i +
+ 2 γ1 α2 + γ2 α2 0 −
λ χ (t) y (t, λ) dt + iγ22 a12 cos λ(x+t−2a
λ
2)
χ (t) y (t, λ) dt
i − −

R a1 cos λ(2a1 −2a2 +x−t) iγ2 R a2 cos λ(x−t)
+ 2 γ1 α2 + γ2 α2 0 λ χ (t) y (t, λ) dt − 2 a1 λ χ (t) y (t, λ) dt
R x sin λ(x−t)
+ a2 λ χ (t) y (t, λ) dt.
(11)
The function ψ (x, λ) provides (12) , (13) , (14).
 A. ERGUN, R. AMIROV: DIRECT AND INVERSE PROBLEMS FOR DIFFUSION... 11

if a2 < x ≤ π, we have
π
sin λ (x − π) sin λ (x − t)
Z
ψ (x, λ) = + χ (t) ψ (t, λ) dt. (12)
λ x λ
if a1 < x < a2 , we have
α+ α− Rπ
ψ (x, λ) = λ2 sin λ (x − π) − λ2 sin λ (2a2 − x − π) + α2+ a2 sin λ(x−t) λ χ (t) ψ (t, λ) dt
− π sin λ(x+t−2a2 ) iγ2 iγ2
R
−α2 a2 λ χ (t) ψ (t, λ) dt + 2λα2 β2 cos λ (x − π) − 2λα2 β2 cos λ (2a2 − x − π)
iγ2 R π cos λ(x−t) Rπ
+ 2α2 β2 a2 λ χ (t) ψ (t, λ) dt − 2αiγ22β2 a2 cos λ(x+t−2a
λ
2)
χ (t) ψ (t, λ) dt
R a2 sin λ(x−t)
+ x λ χ (t) ψ (t, λ) dt
(13)
if 0 ≤ x < a1 , then it has the form

ψ (x, λ) = α1+ α2+ sin λ(x−π)

λ − α1− α2+ sin λ(2aλ1 −x−π) + α1− α2− sin λ(2a2 −2a
λ
1 +x−π)

iα+1 γ2 cos λ(x−π) iα− γ

−α1+ α2− cos λ(2aλ2 −x−π) + 2α2 β2 λ − 2α12 β22 cos λ(2aλ1 −x−π)
iα+ γ iα− γ iα+2 γ2 cos λ(2a1 −x−π)
− 2α12 β22 cos λ(2aλ2 −x−π) + 2α12 β22 cos λ(2a2 −2aλ
1 +x−π)
− 2α1 β1 λ
iα+ γ iα−
2 γ2 cos λ(2a2 −x−π) iα−2 γ1 cos λ(2a2 −2a1 +x−π)
+ 2α21 β11 cos λ(x−π)
λ + 2α1 β1 λ − 2α1 β1 λ
sin λ(2a1 −x−π) sin λ(2a2 −x−π)
+ 4α1γβ11γα22 β2 sin λ(x−π)
λ − γ1 γ2
4α1 β1 α2 β2 λ + γ1 γ2
4α1β1 α2 β2
γ1 γ2 sin λ(2a2 −2a1 +x−π) + + γ1 γ2 R π sin λ(x−t)λ
− 4α1 β1 α2 β2 λ  + α1 α2 + 4α1 β1 α2 β2 a2 λ χ (t) ψ (t, λ) dt

π sin λ(x+t−2a )
+ α1− α2+ + 4α1γβ11γα22 β2 a2
R 1
λ χ (t) ψ (t, λ) dt
 R
π sin λ(x+t−2a )
− α1+ α2− + 4α1 β11 α22 β2 a2
γ γ
λ
2
χ (t) ψ (t, λ) dt
 R
− − γ1 γ2 π sin λ(2a2 −2a1 +x−π)
− α1 α2 + 4α1 β1 α2 β2 a2 λ χ (t) ψ (t, λ) dt

α+ α+
 
1 γ2 2 γ1
R π cos λ(x−t)
+i 2α2 β2 + 2α1 β1 λ χ (t) ψ (t, λ) dt
 −  Ra2
α1 γ2 α+ γ π
−i 2α2 β2 + 2α21 β11 a2 cos λ(x+t−2a λ
1)
χ (t) ψ (t, λ) dt
 + − 
α γ α γ π
−i 2α12 β22 + 2α21 β11 a2 cos λ(x+t−2a 2)
R
λ χ (t) ψ (t, λ) dt
 +
α γ −
α γ
 (14)
−i 2α12 β22 + 2α21 β11 a2 cos λ(2a2 −2a
R π 1 +x−π)
λ χ (t) ψ (t, λ) dt
a sin λ(x−t) a
+α1+ a12 χ (t) ψ (t, λ) dt + α1− a12 sin λ(x+t−2a 1)
R R
λ λ χ (t) ψ (t, λ) dt
γ1 R a2 cos λ(x−t) γ1 R a2 cos λ(x+t−2a1 )
+i 2α1 β1 a1 χ (t) ψ (t, λ) dt − i 2α1 β1 a1 χ (t) ψ (t, λ) dt
R a1 sin λ(x−t) λ λ
+ x λ χ (t) ψ (t, λ) dt.
Where χ (t) = 2λp (t) + q (t).
Theorem 2.1. If p (x) ∈ W21 (0, π) and q (x) ∈ L2 (0, π); yυ (x, λ) solution of the equa-
tions (1) , providing initial conditions (2) and discontinuity conditions (3) − (6) , then it
has the form
Z x
Kυ (x, λ) eiλt dt υ = 1, 3


yυ (x, λ) = y0υ (x, λ) +
−x
where
 R0 (x) eiλx

; 0 ≤ x < a1
y0υ (x, λ) = R1 (x) eiλx + R2 (x) eiλ(2a1 −x) ; a1 < x < a 2

R3 (x) e iλ(2a1 −x) + R4 (x) e iλ(2a1 −2a 2 +x) ; a2 < x ≤ π
12 TWMS J. APP. ENG. MATH. V.9, N.1, SPECIAL ISSUE, 2019
Rx  γ1  −i Rax p(t)dt
R0 (x) = e−i 0 , R1 (x) = R0 (a1 ) α1+ +
p(x)dx
e 1 ,
2
 γ1 i a p(t)dt
 Rx
R2 (x) = R0 (a1 ) α1− − e 1 ,
2
 γ  γ1   Rx
2 i p(t)dt
R3 (x) = − α1+ + R0 (a1 ) − α2− R1 (a2 ) e a2 ,
2 2
γ  γ1  −i x p(t)dt
 R
2
R4 (x) = α1− + R0 (a1 ) + α2− R2 (a2 ) e a2
2 2
Rx
and σ (x) = 0 (2 |p (t)| + (x − t) |q (t)|) dt , where the function Kυ (x, t) satisfies the con-
dition Z x
|Kυ (x, λ)| dt ≤ ecυ σ(x) − 1
−x
with c1 = 1 ,c2 = α1+ + α1−  + γ21 + 2 c3 = α2−  + γ2 (α1 + 1) + α2+ + 1 .
 
  


Theorem 2.2. Let p (x) ∈ W21 (0, π) ,q (x) ∈ L2 (0, π). M (x, t) ,N (x, t)
are summable functions on [0, π] such that the representation for each x ∈ [0, π] / {a1 , a2 }
Z x Z x
φ (x, λ) = φ0 (x, λ) + M (x, t) cos λtdt + N (x, t) sin λtdt
0 0
is satisfied,
a1 < x < a 2

R x λ) = A4 cos [λx − βR(x)]

φ (x,
x
+ A5 cos [λ (2a1 − x) − β (x)]
(15)
+ 0 M (x, t) cos λtdt + 0 N (x, t) sin λtdt
where A4 = α1+ + γ21 , A5 = α1− − γ21 .

a2 < x ≤ π
h R i
x Ra
φ (x, λ) =A1 cos λ (2a1 − x) + a2 p (t) dt − 0 1 p (t) dt
h R i
x Ra
+A2 cos λ (2a1 − x) + a2 p (t) dt − 0 2 p (t) dt
(16)
h R i
x Ra
−A1 cos λ (2a1 − 2a2 + x) − a2 p (t) dt + 0 1 p (t) dt
 Rx 
+AR 3x cos λ (2a1 − 2a2 + Rx)x − 0 p (t) dt
+ 0 M (x, t) cos λtdt + 0 N (x, t) sin λtdt
where A1 = − 2 α1+ + γ21 , A2 = −α2− α1+ + γ21 , A3 = α2− α1− − γ21 .
γ2



Thus , following the relations hold;
A1 x
Z
q (t) + p2 (t) dt


cos β (x) · M (x, x) + sin β (x) · N (x, x) = (17)
2 0
2
p (x) = p (0) + [M (x, x) sin β (x) − N (x, x) cos β (x)] (18)
A1
A2 x
Z
t=2a1 −x+0
q (t) + p2 (t) dt


[M (x, t) cos β (x) − N (x, t) sin β (x)]t=2a1 −x−0 = (19)
2 0
t=2a1 −2a2 +x+0
[(A1 cos β3 (x) − A3 cos β4 (x)) M (x,
Rt) + (A1 sin β3 (x)

− A3 sin β4 (x)) N (x, t)]t=2a1 −2a2 +x−0
1 2 2 x 2
= 2 A1 + A3 − 2A1 A3 cos (β3 − β4 ) 0 q (t) + p (t) dt
(20)

∂M (x, t) 
N (x, 0) = = 0. (21)
∂t 
t=0
 A. ERGUN, R. AMIROV: DIRECT AND INVERSE PROBLEMS FOR DIFFUSION... 13

If p (x) ∈ W22 (0, π) , q (x) ∈ W21 (0, π), so system (22) is provided for M (x, t), N (x, t).
( 2 2
∂ M (x,t)
∂x2 − M (x, t) q (x) − 2p (x) ∂N∂t (x,t)
=∂ M (x,t)
∂t2
2
∂ N (x,t) 2 (22)
∂x2
− N (x, t) q (x) + 2p (x) ∂M∂t(x,t) = ∂ N∂t(x,t)
2

On the contrary , if M (x, t), N (x, t)are summable on [0, π] for each x ∈ [0, π] / {a1 , a2 }and
satisfy equalities (22) and (17) − (21) , then the function φ (x, λ) which is defined by
(15) − (16)is a solution of (1) − (6).
Definition 2.3. If y0 (x) a nontrivial solution of equation (1) with conditions (2) − (6),
then λ0 is called eigenvalue. Additionally, y0 (x) is called the eigenfunction of the problem
corresponding to the eigenvalue λ0 .
We suppose that the function q (x) provides the condition
Z πh i
y (x)2 + q (x) |y (x)|2 dx > 0
 0 
(23)
0

for all y (x) ∈ W22 [0, a1 ) ∪ (a1 , a2 ) ∪ (a2 , π] such that y (x) 6= 0 and
y 0 (0) · y (0) − y 0 (π) · y (π) = 0.
Lemma 2.4. Eigenvalues {λn } for the problem (1) − (6) are real.
Proof. We set l (y) := [−y 00 + q (x) y]. Then,
Rπ
(l (y) , y) = 0nl (y) · y (x)dx o
Rπ
= 0 |y 0 (x)|2 + |y (x)|2 q (x) dx

due to condition (23), follow this (l (y) , y) > 0.

Lemma 2.5. Eigenfunctions corresponding to different eigenvalues of problem (1) − (6)
are orthogonal in the sense of the equality
Z π Z π
(λn + λk ) y (x, λn ) y (x, λk ) dx − 2 p (x) y (x, λn ) y (x, λk ) dx = 0. (24)
0 0

3. Properties of the spectrum

ψ (x, λ) ,φ (x, λ) are any two solution functions of equation (1).
W [ψ (x, λ) , φ (x, λ)] = ψ (x, λ) φ0 (x, λ) − ψ 0 (x, λ) φ (x, λ)
One can see that Wronskian is independent by x,
dW [ψ,φ]
dx = 0 and
W [ψ (x, λ) , φ (x, λ)]a1 +0 = W [ψ (x, λ) , φ (x, λ)]a1 −0 ,
W [ψ (x, λ) , φ (x, λ)]a2 +0 = W [ψ (x, λ) , φ (x, λ)]a2 −0
.
Then it can be shown as W [ψ, ϕ] = ∆ (λ).
The ∆ (λ) function, which is defined as the characteristic function of (1) − (6), is entire
for λ. In that case, {λn } is countable set of zeros of characteristic function.
Lemma 3.1. Eigenvalues of the problem (1) − (6) and zeros of the characteristic function
are overlapped. ψ (x, λ0 ) ,φ (x, λ0 ) functions corresponding to eigenvalue λn ; then there is
such a (ηn ) sequence that provides
ψ (x, λn ) = ηn φ (x, λn ) , ηn 6= 0. (25)
The proof of the lemma is done as in [7].
14 TWMS J. APP. ENG. MATH. V.9, N.1, SPECIAL ISSUE, 2019

Let use denote

Z π Z π
1
αn = ϕ2 (x, λn ) dx − p (x) ϕ2 (x, λn ) dx, n = 1, 2, 3, ... (26)
0 λn 0

{αn } are the normalized numbers of problem (1) − (6).

• •
Lemma 3.2. The equality ∆ (λn ) = 2λn βn αn is held. Here ∆ = d∆ dλ .
Proof. Since φ (x, λ0 ), ψ (x, λ0 ) are the solutions of (1), the following equations are
satisfied;
−φ00 (x, λ) + [2λp (x) + q (x)] φ (x, λ) = λ2 φ (x, λ)
(27)
−ψ 00 (x, λ) + [2λp (x) + q (x)] ψ (x, λ) = λ2 ψ (x, λ)
If equations (27) are differentiated with respect to λ, we get
• • •
− φ00 (x, λ) + [2λp (x) + q (x)] φ (x, λ) = λ2 φ (x, λ) + φ (x, λ) {2λ − 2p (x)}
• • • (28)
− ψ 00 (x, λ) + [2λp (x) + q (x)] ψ (x, λ) = λ2 ψ (x, λ) + ψ (x, λ) {2λ − 2p (x)}

are held. Namely,

•
 
•
d 0 0
dx φ (x, λ) · ψ (x, λ) − φ (x, λ) · ψ (x, λ) = − [2λ − 2p (x)] φ (x, λ) ψ (x, λ)

• •
 (29)
d 0 0
dx φ (x, λ) · ψ (x, λ) − φ (x, λ) · ψ (x, λ) = [2λ − 2p (x)] φ (x, λ) ψ (x, λ)

are held. If equations (29) are integrated from 0 to π , respectively, we obtain

 •
π Z π
•
0 0
− φ (ξ, λ) · ψ (ξ, λ) − φ (ξ, λ) · ψ (ξ, λ) = [2λ − 2p (ξ)] φ (ξ, λ) ψ (ξ, λ) dξ (30)
x x

and
•
 
d • 0 0
φ (ξ, λ) · ψ (ξ, λ) − φ (ξ, λ) · ψ (ξ, λ)
dx
= [2λ − 2p (x)] φ (ξ, λ) ψ (ξ, λ)

Z x 
• •

d 0 0
φ (ξ, λ) · ψ (ξ, λ) − φ (ξ, λ) · ψ (ξ, λ) dξ
dx
Z0 x
= [2λ − 2p (ξ)] φ (ξ, λ) ψ (ξ, λ) dξ
0

•
 
R a1 −0 •
0
d
dxφ (ξ, λ) · − ψ 0 (ξ, λ) φ0 (ξ, λ)
· ψ (ξ, λ) dξ
•
 
R a2 −0 d • 0 0
+ a1 +0 dx φ (ξ, λ) · ψ (ξ, λ) − φ (ξ, λ) · ψ (ξ, λ) dξ
•
 
Rx •
d 0 0
+ a2 +0 dx φ (ξ, λ) · ψ (ξ, λ) − φ (ξ, λ) · ψ (ξ, λ) dξ
R a −0 R a −0
= R 0 1 [2λ − 2p (ξ)] φ (ξ, λ) ψ (ξ, λ) dξ + a12+0 [2λ − 2p (ξ)] φ (ξ, λ) ψ (ξ, λ) dξ
x
+ a2 +0 [2λ − 2p (ξ)] φ (ξ, λ) ψ (ξ, λ) dξ
 A. ERGUN, R. AMIROV: DIRECT AND INVERSE PROBLEMS FOR DIFFUSION... 15

 •
a1 −0
•
φ (ξ, λ) · ψ 0 (ξ, λ)− φ0 (ξ, λ)· ψ (ξ, λ)
0

• • a2 −0
+ φ (ξ, λ) · ψ 0 (ξ, λ) − φ0 (ξ, λ) · ψ (ξ, λ)
 •
ax1 +0
•
0 0
+ φ (ξ, λ) · ψ (ξ, λ) − φ (ξ, λ) · ψ (ξ, λ)
Rx a2 +0
= 0 [2λ − 2p (ξ)] φ (ξ, λ) ψ (ξ, λ) dξ
 •
x
•
0 0
φ (ξ, λ) · ψ (ξ, λ) − φ (ξ, λ) · ψ (ξ, λ) (31)
0
Z x
= [2λ − 2p (ξ)] φ (ξ, λ) ψ (ξ, λ) dξ.
0
By addition side by side, we obtain the equality
h . i h. i
W φ (ξ, λ) , ψ (ξ, λ) + W φ (ξ, λ) , ψ (ξ, λ)
Z π
.
= − ∆ (λ) = [2λ − 2p (ξ)] φ (ξ, λ) ψ (ξ, λ) dξ
0
for λ → λn . This yield
• Rπ
∆ (λn ) =R− 0 [2λ − 2p (ξ)] ηnRφ2 (ξ, λn ) dξ
π π
= 2λn ηn 0 φ2 (ξ, λn ) dξ − λ1n 0 p (ξ) φ2 (ξ, λn ) dξ = 2λn ηn αn .
  0
  Γn =0  λ : |λ| = λn + δ, δ > 0,
Let denote n = 0, 1, 2, ... ,
 
Tn = λ : λ − λn  ≥ δ, δ > 0, n = 0, 1, 2, ... ,
where δ > 0.
Lemma 3.3. For enough large values of n ,
Cδ |τ |(2a1 −2a2 +π)
|∆ (λ) − ∆0 (λ)| < e , λ ∈ Γn (32)
2
is provided.
Proof. As it is shown in [26] , |∆0 (λ)| ≥ Cδ e|Imλ|π for all λ ∈ T̄δ , where Cδ > 0 is
constant. On the other hand, since
lim e−|Imλ|π (∆ (λ) − ∆0 (λ)) (33)
|λ|→∞
Z π Z π 
−|Imλ|π
= lim e M̃ (π, t) cos λtdt + Ñ (π, t) sin λtdt = 0
|λ|→∞ 0 0
for enough large values of n (see[1]) we get (3) . The lemma is proved.
Lemma 3.4. The boundary value problem (1) − (6) has a countable number of eigenval-
ues, that grow unlimitedly, when that are ordered according to their absolute value. In
addition, eigenvalues can also be shown asymptotically as the following.
sn tn
λn = λ0n + 0 + 0 , n → ∞
λn λn
nπ
λ0n = + ψ1 (n) ; sup |ψ1 (n)| = c < +∞
2a1 − 2a2 + π n
where tn ∈ l2 and sn is a bounded sequence.
Proof. By the lemma 3.3., if λ ∈ Γn , |∆0 (λ)| ≥ Cδ e|Imλ|π > C2δ e|Imλ|π > |∆ (λ) − ∆0 (λ)|
is true. Applying Rouche’s theorem inside the Γn , the functions ∆ (λ) = ∆0 (λ) +
{∆ (λ) − ∆0 (λ)} and ∆0 (λ) have the same number of zeros. Namely , λ0 , λ1 , λ2 , ..., λn are
16 TWMS J. APP. ENG. MATH. V.9, N.1, SPECIAL ISSUE, 2019

zeros. Similarly, according

  to the Rouche’s theorem the function ∆ (λ) has a unique zero
inside each circle λ − λ0k  < δ for sufficiently large values of k. Since δ > 0, λn = λ0n + εn
. where lim εn = 0 . If ∆ (λn ) = 0, we have
n→∞
Z π Z π
0 0
N (π, t) sin λ0n + εn tdt = 0 (34)




∆ 0 λ n + εn + M (π, t) cos λn + εn tdt +
0 0
h R i
x Ra
∆0 λ0n + εn = A1 cos λ0n + εn (2a1 − x) + a2 p (t) dt − 0 1 p (t) dt



h R i
x Ra
+A2 cos λ0n + εn (2a1 − x) + a2 p (t) dt − 0 2 p (t) dt


h R i (35)
x Ra
−A1 cos λ0n + εn (2a1 − 2a2 + x) − a2 p (t) dt + 0 1 p (t) dt


Rx
+A3 cos λ0n + εn (2a1 − 2a2 + x) − 0 p (t) dt = 0


 
.
0 0



∆0 λn + εn = ∆ λn + o (1) εn , n → ∞
0

(35) takes the form of

h . i R 2a −π−0
0 M (π, t) cos λ0n + εn tdt



∆0 λn + o (1) εn + 0 1
R 2a −2a +π−0
M (π, t) cos λ0n + εn tdt


+ 2a11−π+02
Rπ R 2a −π−0
+ 2a1 −2a2 +π+0 M (π, t) cos λ0n + εn tdt + 0 1 N (π, t) sin λ0n + εn tdt



R 2a −2a +π−0 Rπ
N (π, t) sin λ0n + εn tdt + 2a1 −2a2 +π+0 N (π, t) sin λ0n + εn tdt = 0



+ 2a11−π+02

İt is easy to see that the function ∆0 (λ) = 0 is type of [17] , so there is a ηδ > 0, such
 . 0


that ∆0 λn  ≥ ηδ > 0 is satisfied for all n. We also have
nπ
λ0n = + ψ1 (n) (36)
2a1 − 2a2 + π
where sup |ψ1 (n)| < M for some constant M > 0 [18]. In addition to, when we replace
n
(36) into (35) and calculate certain transformations, we reach εn ∈ l2 . We can obtain
exactly
εn =
1 
− (M (π, 2a1 − π + 0) − M (π, 2a1 − π − 0)) sin λ0n + εn (2a1 − π)


− .
∆0 (λ0n )+o(1) (λ0n +εn )
− M (π, 2a1 − 2a2 + π
− 0)) sin λ0n + εn (2a1 − 2a2 + π)


− (M (π, 2a1 − 2a2 + π
+ 0)
π
+M (π, π) sin λ0n + εn − 0 Mt (π, t) sin λ0n + εn tdt

R

+ (N (π, 2a1 − π + 0) − N (π, 2a1 − π − 0)) cos λ0n + εn (2a1 − π)

+ (N (π, 2a1 − 2a2 + π
+ 0) − 0
0
R πN (π, 2a1 − 2a2 0+ π −
0)) cos
 λn + εn (2a1 − 2a2 + π) + N (π, 0)
−N (π, π) cos λn + εn π + 0 Nt (π, t) cos λn + εn tdt
Rπ Rπ
Since 0 Mt (π, t) sin λ0n + εn tdt ∈ l2 and 0 Nt (π, t) cos λ0n + εn tdt ∈ l2 , we have

εn = 2λ0 ∆10 (λ0 ) −A5 sin β4 (π) + λ0n (2a1 − π) + A1 sin β3 (π) − λ0n (2a1 − 2a2 + π)
    
n n  R π
0 (2a − 2a + π) − A sin β (π) − λ0 π 2 (t) dt
 

+A 3 sin β4 (π) + λn 1 2 4 4 n 0 q (t) + p
+ A5 cos β4 (π) + λ0n (2a1 − π) − A1 cos β3 (π) − λ0n (2a1 −2a2 + π)
    

+A3 cos β4 (π) − λ0n (2a1 − 2a2 + π) − A4 cos β4 (π) + λ0n π [p (π) − p (0)] + λtn0

n

where tn ∈ l2 . So we obtain
sn tn
λn = λ0n + 0
+ 0
λn λn
 A. ERGUN, R. AMIROV: DIRECT AND INVERSE PROBLEMS FOR DIFFUSION... 17

where
sn = 2∆01(λ0 ) −A5 sin β4 (π) + λ0n (2a1 − π) + A1 sin β3 (π) − λ0n (2a1 − 2a2 + π)
    
n  R π
0 0 2
  

+A 3 sin β4 (π) + λn 0(2a1 − 2a2 + π) − A4 sin β4 (π)0− λn π 0 q (t)+ p (t) dt
+ A5 cos β4 (π) + λn (2a1 − π) − A1 cos β3 (π)  − λn (2a1 −2a2 + π)
+A3 cos β4 (π) − λ0n (2a1 − 2a2 + π) − A4 cos β4 (π) + λ0n π [p (π) − p (0)]

sn is a bounded sequence. This completes the proof. u

t

4. Inverse problem
Let’s show problem (1) − (6) as L (α, a1 , a2 ).

−y 00 + [2λp̃ (x) + q̃ (x)] y = λ2 y, x ∈ (0, π)



U (y) = y 0 (0) = 0, V (y) = y (π) = 0





 y (ã + 0) = α̃ y (ã − 0)
1 1 1
L̃ (α, a1 , a2 ) := 0 (ã + 0) = β̃ y 0 (ã − 0) + iλγ̃ y (ã − 0)

 y 1 1 1 1 1
y (ã + 0) = α̃ y (ã − 0)

2 2 2



 0
y (ã2 + 0) = β̃2 y 0 (ã2 − 0) + iλγ̃2 y (ã2 − 0)

we consider the boundary value problem L̃ (α, a1 , a2 ). L̃ (α, a1 , a2 ) has the same form
L (α, a1 , a2 ), but only its coefficients (q̃, p̃, α̃1 , α̃2 , ã1 , ã2 , γ̃1 , γ̃2 ) different. We suppose to the
coefficient (q̃, p̃, α̃1 , α̃2 , ã1 , ã2 , γ̃1 , γ̃2 ) provide conditions of problem (1) − (6). Let Φ (x, λ)
is a solution of equation (1) with the conditions U (Φ) = 1, V (Φ) = 0 and jump contitions
(3) − (6). We define M (λ) := Φ (0, λ).
The functions Φ (x, λ)and M (λ) are called the Weyl solution and Weyl function for the
boundary value problem (1) − (6).
Using the solution φ (x, λ) defined in the previous sections one has

ψ (x, λ) ψ (0, λ)
Φ (x, λ) := − = S (x, λ) + M (λ) · φ (x, λ) , M (λ) = − (37)
∆ (λ) ∆ (λ)

where ψ (x, λ) is a solution of (1) with the conditon ψ (π, λ) = 0, ψ 0 (π, λ) = −1 and
(3) − (6). S (x, λ) is defined from the equality

ψ (x, λ) = ψ (0, λ) · φ (x, λ) − ∆ (λ) · S (x, λ) (38)

then,
hΦ (x, λ) , φ (x, λ)i ≡ 1
hΦ (x, λ) , φ (x, λ)i ≡ −∆ (λ) for x 6= a1 , a2 .
The following theorem states the significance of the Weyl function.

Theorem 4.1. If M (λ) = M f (λ), then L (α, a1 , a2 ) = L̃ (α, a1 , a2 ). Thus, the boundary
value problem L (α, a1 , a2 )is uniquely defined by the Weyl function.

Proof. Let us define the matrix P (x, λ) = [Pj,k (x, λ)] , (j, k = 1, 2)
!  
φe (x, λ) Φe (x, λ) φ (x, λ) Φ (x, λ)
P (x, λ) · 0
e 0 (x, λ) = 0 0
φe (x, λ) Φ φ (x, λ) Φ (x, λ)
18 TWMS J. APP. ENG. MATH. V.9, N.1, SPECIAL ISSUE, 2019

So, we can write

e0

0
P11 (x, λ) = −φ (x, λ) ψe(x,λ) + φe (x, λ) ψ(x,λ)
∆(λ)


∆(λ) 


P12 (x, λ) = −φe (x, λ) ψ(x,λ) ψ(x,λ)
e 
∆(λ) + φ (x, λ) e


∆(λ)
e0 (x,λ) 0 0 (x,λ)
(39)
ψ
P21 (x, λ) = 0
−φ (x, λ) e − φ (x, λ) ψ∆(λ)
e 

∆(λ) 

0 (x,λ)

−φe (x, λ) ψ∆(λ) + φ0 (x, λ) ψ(x,λ)
e 
P22 (x, λ) =


∆(λ)
e

then we have
h 0 i
P11 (x, λ) = φ (x, λ) Se (x, λ) + M f (λ) · φe0 (x, λ)
0
−φe (x, λ) [S (x, λ) + M (λ) · φ (x, λ)] h i
0 0
= φ (x, λ) Se (x, λ) − φe (x, λ) S (x, λ) + M f (λ) − M (λ) φ (x, λ) φe0 (x, λ)

h i
P12 (x, λ) = φe (x, λ) [S (x, λ) + M (λ) · φ (x, λ)] − φ (x, λ) Se (x, λ) + M
f (λ) · φe (x, λ)
h i
= φe (x, λ) S (x, λ) − φ (x, λ) Se (x, λ) + M (λ) − M f (λ) φ (x, λ) φe (x, λ)

h 0 i
P21 (x, λ) = φ0 (x, λ) Se (x, λ) + M f (λ) · φe0 (x, λ)
0
−φe (x, λ) [S 0 (x, λ) + M (λ) · φ0 (x, λ)] h i
0 0
= φ0 (x, λ) Se (x, λ) − φe (x, λ) S 0 (x, λ) + M f (λ) − M (λ) φ0 (x, λ) φe0 (x, λ)

h i
P22 (x, λ) = φe (x, λ) [S 0 (x, λ) + M (λ) · φ0 (x, λ)] + φ0 (x, λ) Se (x, λ) + M
f (λ) · φe (x, λ)
0
h i
= φ0 (x, λ) S (x, λ) − φ0 (x, λ) Se (x, λ) + M (λ) − M f (λ) φ0 (x, λ) φe (x, λ) .

If M (λ) ≡ M
f (λ) ,then
0 0
P11 (x, λ) = φ (x, λ) Se (x, λ) − φe (x, λ) S (x, λ)

P12 (x, λ) = φe (x, λ) S (x, λ) − φ (x, λ) Se (x, λ)

0 0
P21 (x, λ) = φ0 (x, λ) Se (x, λ) − φe (x, λ) S 0 (x, λ)
0
P22 (x, λ) = φ0 (x, λ) S (x, λ) − φ0 (x, λ) Se (x, λ) .
If M (λ) ≡ M f (λ) , the functions Pj,k (x, λ) , (j, k = 1, 2)are entire for λ . On the other
hand, the function φ (x, λ) has asymtotic expression due to (9) , (10) , (11) and the function
ψ (x, λ) has asymtotic expression due to (12) , (13) , (14)
 O |λ| e|σ|x



; 0 < x < a1
φ0 (x, λ) = O |λ| e|σ|(2a1 −x)
; a1 < x < a2 ,

O |λ| e |σ|(2a 1 −2a2 +x) ; a2 < x < π
|λ|(π−x)


 O |λ| e
; 0 < x < a1
ψ 0 (x, λ) = O |λ| e|λ|α1 (π−x)
; a1 < x < a2 .

O |λ| e |λ|α2 (π−x) ; a2 < x < π
We replace (9) , (10) , (11) into (39) for ∀x ∈ [0, π] / {a1 , a2 , e a2 }
a1 , e
|P11 (x, λ)| ≤ cδ and |P12 (x, λ)| ≤ Cδ
 A. ERGUN, R. AMIROV: DIRECT AND INVERSE PROBLEMS FOR DIFFUSION... 19

is held. cδ , Cδ can be shown presence of constants. We calculate certain transformation

of Liouville theorem, then
P11 (x, λ) ≡ A (x) and P12 (x, λ) ≡ 0 are held. Now by using (39), we obtain
0
φ (x, λ) · Φ0 (x, λ) − φe (x, λ) · Φ (x, λ) = A (x)
φe (x, λ) · Φ (x, λ) − φ (x, λ) · Φ
e (x, λ) = 0
and )
φ (x, λ) = φe (x, λ) · A (x)
e (x, λ) · A (x) (40)
Φ (x, λ) = Φ
therefore with (37) and (38)
h i
W [φ, Φ] = W φ (x, λ) , − ψ(x,λ)
∆(λ)
1
= ∆(λ) W [φ (x, λ) , −ψ (0, λ) φ (x, λ) + ∆ (λ) S (x, λ)]
= − ψ(0,λ)
∆(λ) W [φ (x, λ) , φ (x, λ)] + W [φ (x, λ) , S (x, λ)]
=1
h i h i
and similarly W φ,
eΦe = 1. According to (40)1 = W [φ (x, λ) , Φ (x, λ)] = W A (x) φe (x, λ) , A (x) Φ
e (x, λ)
h i
= A2 (x) W φe (x, λ) , Φ
e (x, λ) = A2 (x)

Functionφ (x, λ) has the following the representation in D = {λ : arg λ ∈ [ε, π − ε]}
for |λ| → ∞,
if a1 < x < a2 ;   
k 1
ϕ (x, λ) = exp (−i (λx − β (x))) 1 + O (41)
2 λ
k = 1 for x < a1 , k = A4 for x > a1 .
If a2 < x ≤ π;
  
k 1
ϕ (x, λ) = exp (−i (λ (2a1 − 2a2 + π) − (β1 (x) + β2 (x)))) 1 + O (42)
2 λ
k = A1 . Because of (40) , (41) , (42) ;a1 = ã1 ,a2 = ã2 are true. In addition to A (x) = 1
,A1 = Ã1 ve A4 = Ã4 are true. Because of (42), φ (x, λ) ≡ φ̃ (x, λ) , Φ (x, λ) ≡ Φ̃ (x, λ) is
held.
Furthermore , when we replace ϕ (x, λ) ≡ ϕ̃ (x, λ) and operation are calculated, we
reachp (x) = p̃ (x), q (x) = q̃ (x). Thus , for each λ,
(α1 − α̃1 )ϕ (a1 − 0, λ) = 0

β1 − β̃1 ϕ0 (a1 − 0, λ) + (γ1 − γ̃1 ) ϕ (a1 − 0, λ) = 0

 2 − α̃2 )ϕ (a2 − 0, λ) = 0

(α
β2 − β̃2 ϕ0 (a2 − 0, λ) + (γ2 − γ̃2 ) ϕ (a2 − 0, λ) = 0
α1 = α̃1 , α2 = α̃2 , γ1 = γ̃1 ve γ2 = γ̃2 are held. Consequently L (α, a1 , a2 ) = L̃ (α, a1 , a2 ).
The theorem is proved. 

Theorem 4.2. If µn = µ
en , λn = λ
en , ; n = 0, 1, 2, ..., then L (α, a1 , a2 ) = L̃ (α, a1 , a2 ).

Proof. Obviously, the characteristic function ∆ (λ) and ψ (0, λ) are uniquely identified by
the sequences λ2n and µ2n . If λn = λ
 
en and µn = µen ,(n = 0, 1, 2, ...), then ∆ (λ) ≡ ∆
e (λ)
and ψ (0, λ) ≡ ψe (0, λ). İt follows from (38) that M (λ) ≡ M f (λ). Therefore, applying
Theorem 10, we conclude that L (α, a1 , a2 ) = L̃ (α, a1 , a2 ). The theorem is proved. 
20 TWMS J. APP. ENG. MATH. V.9, N.1, SPECIAL ISSUE, 2019

Theorem 4.3. If αn = α en , λn = λ
en , ; n = 0, 1, 2, ..., then L (α, a1 , a2 ) = L̃ (α, a1 , a2 ). So,
spectral data {λn , αn } uniquely determines the problem L (α, a1 , a2 ).
Proof. The Weyl function M (λ) is meromorphic with simple poles at points λ2n . The
expression Z π Z π
∆ (λ) = ∆0 (λ) + A (π, λ) cos λtdt + B (π, λ) sin λtdt
0 0
•
and equalities 2λn βn αn = − ∆ (λn ) , ψ (x, λn ) = βn ϕ (x, λn ), we have
ψ (0, λn ) βn 1
Res M (λ) = − • =− • = .
λ=λn 2λn αn
∆ (λn ) ∆ (λn )
because the M (λ)is regular for λ ∈ Γn by Rouche’s theorem. So,
Z
1 M (µ)
M (λ) = dµ , λ ∈ Γn .
2πi Γn1 µ − λ
It follows from (38) that |M (λ)| ≤ Cδ |λ|−1 , λ ∈ Cδ . Where Γn1 = λ : |λ| = λ0n  , n = 1, 2, ... .
  
Because of residue theorem,
M (λ) = ∞ 1
P
−∞ 2λn αn (λ−λn ) is held. At last, from equalityM (λ) ≡ M (λ) by Theorem 10,
f
we conclude that L (α, a1 , a2 ) = L̃ (α, a1 , a2 ). The theorem is proved. 

5. Conclusions
In this study diffusion operator with discontinuity points is considered. Firstly , impor-
tant definitions and theorems which are used frequently in spectral theory of differential
operators are given. Integral equations for solition which satisfy certain initial and jump
condition of given equation has been obtained and useful integral represantation some im-
portant properties of eigenvalues and eigenfunction have been investigated. Finally, consist
of the inverse problem. It has been proven that the coefficients of the given problem are
uniquely determined by the Weyl function and spectral data.

References
[1] Marchenko, V. A., (1986), Sturm-Liouville Operators and Applications, AMS Chelsea Publishing,
Basel.
[2] Levitan, B. M. and Gasymov, M. G., (1964), Determination of a differential equation by two spectra.
Uspekhi Mathematicheskikh Nauk, 19, 2, 3-63.
[3] Jaulent, M . and Jean, C., (1976), The Inverse Problem for the one-dimensional Schrödinger equation
with an energy-dependent potential. Annales de L’Institut Henri Poincare A, 25, 2, 105-137.
[4] Yurko, V. A., (2000), Introduction to the Theory of Inverse Spectral Problems, Fizmatlit, MOscow,
Russia, 2007, English translation, Inverse Spectral Problems dor Differantial Operators and Their
Applications, Gordon and Breach, Amsterdam, The Netherlands.
[5] Levitan, B. M., (1978), Inverse Sturm-Liouville Problems, Nauka, Moscow, Russia.
[6] Gasymov M. G. and Guseinov, G. S., (1981), Determination of a diffusion operator from spectral
data, Akademiya Nauk Azerbaidzhanskoi SSR. Doklady, 37, 2, 19-23.
[7] Yurko, V. A., (2000), An inverse problem for differential operator pencils, Methematicheskikh Sbornik,
191, 10, 137-160.
[8] Nabiev, M. I., (2004), The inverse spectral problem for he diffusion operator on an interval. Mathe-
maticheskaya Fizika, Analiz, Geometriya, 11, 3, 302-313,
[9] Guseinov, G. S., (1986), Inverse spectral problems for a quadratic pencil of Sturm-Liouville operators
on a finite interval. Spectral Theory of Operators and Its Applications, 51-101, Elm, Baku, Azerbaijan.
[10] Freiling G. and Yurko V., (2001), Inverse Sturm-Liouville problems and their applications, Nova
Science, Huntington, NY, USA.
[11] Yurko, V. A., (2000), Inverse spectral problems for differential operators and their applications, vol.
2 of Analytical Methods and Special Functions, Gordon and Breach, Amsterdam, The Netherlands.
 A. ERGUN, R. AMIROV: DIRECT AND INVERSE PROBLEMS FOR DIFFUSION... 21

[12] Hald, O. H., (1984), Discontinuous inverse eigenvalue problems, Communications on Pure and Applied
Mathematics, 37, 5, 539–577.
[13] Amirov, R. Kh., (2006), On Sturm-Liouville operators with discontinuity conditions inside an inter-
val.Journal of Mathematical Analysis and Applications, 317, 1, 163-176.
[14] AmirovR. Kh. and Yurko, V. A., (2001), On differential operators with a singularity and discontinuity
conditions inside an interval.Ukrainian Mathematical Journal, 53,11,1443–1457.
[15] FreilingG.and Yurko, V., (2002), Inverse spectral problems for singular non-selfadjoint differential
operators with discontinuities in an interior point. Inverse Problems, 18, 3, 757-773.
[16] Levin, B. Ya., (1971), Entire Functions, MGV, Moscow, Russia.
[17] Jdanovich, B. F. (1960), Formulae for the zeros of Drichlet polynomials and quasi-polynomials. Dok-
lady Akademii Nauk SSSR, 135, 8, 1046-1049.
[18] Kreyn M. and Levin, B. Ya., (1949), On entire almost periodic functions of exponential type, Doklady
Akademii Nauk SSSR, 64, 285-287.
[19] Buterin S. A., Yurko V.A., (2006), Inverse spectral problem for pencils of differential operators on a
finite interval. Vesn. Bashkir. Univ, 4, 8-12.
[20] Koyunbakan H., Panakhov E. S., (2007), Half-inverse problem for diffusion operators on the finite
interval.Journal of Mathematical Analysis and Applications, 326, 1024-1030.
[21] Yang C. F.,(2010), Reconstruction of the diffusion operator from nodal data Zeitschrift fur Natur-
forschung, 65, 100-106.
[22] Amirov R. Kh., Ergun A., (2015), İki noktada süreksizlik kosullarına sahip difuzyon denkleminin
cozumleri icin integral gosterilim, Cumhuriyet Universty Science Journal, doi:10.17776/csj.06773.

Abdullah ERGN was born in Sivas, Turkey in 1978. He obtained his B.Sc. degree
in Mathematics from Cumhuriyet University in 2001. In advance, Dr. ERGN got
his Ph.D. degree in Mathematics from Tokat Gazi Osmanpaa University in 2015. He
has several refereed publications in yields of Applied Mathematics such as solutions
of partial differential equations and diffusion operator.

Rauf Amirov has been working at Department of Mathematics, Cumhuriyet Uni-

versity since 1994. He is currently working as a Full professor in the Department of
Mathematics in Cumhuriyet University . He has several refereed publications in yields
of Applied Mathematics such as solutions of partial differential equations, regular or
singular differential operators.