DOI 10.

1007/s10958-020-04707-2
Journal of Mathematical Sciences, Vol. 245, No. 4, March, 2020

ON A BOUNDARY-VALUE PROBLEM FOR A FOURTH-ORDER PARTIAL

INTEGRO-DIFFERENTIAL EQUATION WITH DEGENERATE KERNEL

T. K. Yuldashev UDC 517.968

Abstract. In this paper, the classical solvability of a nonlocal boundary-value problem for a three-
dimensional, homogeneous, fourth-order, pseudoelliptic integro-differential equation with degenerate
kernel is proved. The spectral Fourier method based on the separation of variables is used and a
countable system of algebraic equations is obtained. A solution is constructed explicitly in the form
of a Fourier series. The absolute and uniform convergence of the series obtained and the possibility
of termwise differentiation of the solution with respect to all variables are justified. A criterion of the
unique solvability of the problem considered is ascertained.
Keywords and phrases: pseudoelliptic equation, degenerate kernel, integral condition, unique solv-
ability, classical solution.
AMS Subject Classification: 35A02, 35M10, 35S05

1. Formulation of the problem. Mathematical modeling of many processes occurring in the real
world leads to the study of mixed and boundary-value problems for partial differential equations.
Therefore, the theory of mixed problems is currently one of the most important sections of the theory
of differential equations.
The study of many problems of gas dynamics, the theory of elasticity, and the theory of plates
and shells leads to the consideration of higher-order partial differential equations. From the point
of view of physical applications, fourth-order differential equations are also of interest (see [1, 5, 22,
26, 27]). Many publications are devoted to the study of boundary-value problems for differential and
integro-differential partial differential equations (see also [2–4, 6, 10, 19, 23]).
If the boundary of the domain where a physical process occurs is not available for measurements,
nonlocal conditions in integral form can serve as additional information that allows one to prove the
unique solvability of the problem. Such nonlocal problems were considered in the works of many
authors (see, e.g., [12, 14, 24, 25]).
In the study of partial differential and integro-differential equations, the method of separation of
variables is often applied in the works of many authors (see, e.g., [8, 13, 15–18, 20, 28, 29]).
In this paper, we study the existence and uniqueness of a classical solution to a nonlocal boundary-
value problem for a pseudo-elliptic fourth-order integro-differential equation with a degenerate kernel in
a three-dimensional domain. First, the corresponding pseudo-elliptic differential equation is solved for
given boundary conditions. The method of separation of variables is used. A solution is constructed in
the form of a Fourier series. We prove that this solution is the unique classical solution of the problem.
Further, the solution of the given integro-differential equation with the same boundary conditions is
constructed in the form of a Fourier series. A solution of the boundary-value problem is constructed
in the explicit form by the reduction to a countable system of algebraic equations. The absolute
and uniform convergence of the series obtained and the termwise differentiability of the solution with
respect to all three variables are proved. We also note that partial integral-differential equations with
degenerate kernels were examined by many authors (see, e.g., [7, 9, 21, 30, 31, 33]).

Translated from Itogi Nauki i Tekhniki, Seriya Sovremennaya Matematika i Ee Prilozheniya. Tematicheskie
Obzory, Vol. 145, Geometry and Mechanics, 2018.

508 1072–3374/20/2454–0508 
c 2020 Springer Science+Business Media, LLC
 In a three-dimensional domain ω = {(t, x, y)mid0 < t < β, 0 < x, y < l}, we consider the following
integro-differential equation:

Utt (t, x, y) − Uttxx (t, x, y) − Uttyy (t, x, y) + Uxx (t, x, y) + Uyy (t, x, y)
β
 
+ν K(t, s) Uxx (s, x, y) + Uyy (s, x, y) ds = 0, (1)
0

where β and l are given positive real numbers, ν is the real spectral parameter, and
k

K(t, s) = ai (t)bi (s), ai (t), bi (s) ∈ C 2 [0; β].
i=1

Here we assume that the functions ai (t) and bi (s) are linearly independent.
We consider Eq. (1) under the following conditions:
   
U (t, x, y) ∈ C ω ∩ C 1 ω ∪ {x = 0} ∪ {x = l} ∪ {y = 0} ∪ {y = l}
2+2+0 2+0+2
∩ C 2 (ω) ∩ Ct,x,y (ω) ∩ Ct,x,y (ω), (2)

U (0, x, y) = U (β, x, y), 0 ≤ x, y ≤ l, (3)

β
U (t, x, y)dt = ϕ(x, y), 0 ≤ x, y ≤ l, (4)
0
U (t, 0, y) = U (t, l, y) = U (t, x, 0) = U (t, x, l) = 0, 0 ≤ t ≤ β, (5)

where C r (ω) is the class of functions possessing continuous derivatives ∂ r /∂tr , ∂ r /∂xr , ∂ r /∂y r in the
r+s+0
domain ω, Ct,x,y (ω) is the class of functions possessing continuous derivatives ∂ r+s /∂tr ∂xs in the
r+0+s
domain ω, Ct,x,y is the class of functions possessing discontinuous derivatives ∂ r+s /∂tr ∂y s in the
domain ω, r = 1, r0 , s = 1, s0 , r ≤ r0 , s ≤ s0 are integers, ϕ(x, y) is a given sufficiently smooth
function, and

ϕ(0, y) = ϕ(l, y) = ϕ(x, 0) = ϕ(x, l) = 0,

ω = {(t, x, y)|0 ≤ t ≤ β, 0 ≤ x, y ≤ l}

First, we consider the following differential equation:

Utt (t, x, y) − Uttxx (t, x, y) − Uttyy (t, x, y) + Uxx (t, x, y) + Uyy (t, x, y) = 0. (6)

Problem 1. In the three-dimensional domain ω, find a function U (t, x, y) satisfying Eq. (6) and the
conditions (2)–(5).

Problem 2. In the three-dimensional domain ω, find a function U (t, x, y) satisfying Eq. (1) and the
conditions (2)–(5).

Note that Eq. (6) is called a pseudo-elliptic equation (similarly to pseudo-parabolic and pseudo-
hyperbolic equations). As elliptic differential equations, pseudo-elliptic equations have numerous ap-
plications in various fields of science.

509
2. Particular solutions of problem 1. We search for a nontrivial partial solution of Eq. (6) in
a three-dimensional domain ω in the formU (t, x, y) = T (t)V (x, y) (see [11]). Then from Eq. (6) we
obtain

T  (t)V (x, y) − T  (t)Vxx (x, y) − T  (t)Vyy (x, y) = − T (t)Vxx (x, y) + T (t)Vyy (x, y) .
Now we divide both sides by T (t)V (x, y):
T  (t) T  (t) Vxx (x, y) Vyy (x, y) Vxx (x, y) Vyy (x, y)
− + + = + .
T (t) T (t) V (x, y) V (x, y) V (x, y) V (x, y)
As in [32], we set
Vxx (x, y) Vyy (x, y) T  (t) T  (t) Vxx (x, y) Vyy (x, y)
+ = −μ2 , − + + = −μ2 ,
V (x, y) V (x, y) T (t) T (t) V (x, y) V (x, y)
where μ2 is the separation constant, 0 < μ. Hence, taking into account the boundary conditions (3)
and (5), we obtain
Vxx (x, y) + Vyy (x, y) + μ2 V (x, y) = 0, 0 < x, y < l, (7)
V (0, y) = V (l, y) = V (x, 0) = V (x, l) = 0, (8)
 2
T (t) − λ T (t) = 0, 0 < t < β, (9)
T (0) = T (β), (10)
where λ2 = μ2 /(1 + μ2 ).
We solve the spectral problem (7), (8) also by the method of separation of variables:
V (x, y) = X(x) · Y (y). (11)
Substituting (11) into (7) and (8) and separating the variables, we get two eigenvalue problems
X  (x) + ω12 X(x) = 0, X(0) = X(l) = 0, (12)

Y (y) + ω22 Y (y) = 0, Y (0) = Y (l) = 0, (13)
where 0 < ωi are the separation constants, i = 1, 2, ω1 + ω2 = μ.
Solving the problems (12) and (13), we obtain
Xn (x) = An sin ω1n x, Ym (y) = Am sin ω2m y,
where An and Am are arbitrary constants, ω1n = πn/l, and ω2m = πm/l, n, m = 1, 2, 3, . . .. Then (11)
becomes
πn πm
Vn,m (x, y) = An,m sin x sin y,
l l
where An,m are unknown coefficients, which can be found from the following norm condition:
l l
2
Vn,m (x, y) dx dy = 1;
0 0
we have An,m = 2/l. Hence from (7) and (8) we obtain
Vn,m (x, y) = Xn (x) · Ym (y), (14)
where
2 πn 2 πm π(n + m)
Xn (x) = sin x, Ym (y) = sin y, μn,m = ; n, m = 1, 2, . . . .
l l l l l
Then the general solution of Eq. (9) has the form
Tn,m (t) = an,m eλn,m t + bn,m e−λn,m t , (15)

510
 μ2n,m
where an,m and bn,m are arbitrary constants and λn,m = .
1 + μ2n,m
From the condition (10) and (15) we find the relationship between the constants an,m and bn,m :
1 − e−λn,m β
an,m = bn,m . (16)
eλn,m β − 1
Substituting (16) in the function (15), we obtain
1 − e−λn,m β λn,m t
Tn,m (t) = bn,m ·e + e−λn,m t . (17)
eλn,m β − 1
3. Uniqueness of the solution. To establish the uniqueness of the solution, we show that for a
homogeneous integral condition
β
U (t, x, y)dt = 0, 0 ≤ x, y ≤ l,
0
the problem 1 has only the trivial solution. Taking into account (14), we search for a solution of the
problem 1 in the three-dimensional domain ω in the form of a Fourier series
∞
2  πn πm
U (t, x, y) = un,m (t) sin x sin y, (18)
l l l
n,m=1

where
l l
2 πn πm
un,m (t) = U (t, x, y) sin x sin y dx dy, n, m = 1, 2, . . . . (19)
l l l
0 0
The functions (19) satisfy Eq. (9) and the condition (10). Indeed, differentiating Eqs. (19) twice by t
and using Eq. (6), we obtain
l l
2 πn πm
un,m (t) = Utt (t, x, y) sin x sin y dx dy
l l l
0 0
l l  
2 πn πm
= Uttxx + Uttyy − Uxx − Uyy sin x sin y dx dy
l l l
0 0
l l
2   πn πm
= Uttxx − Uxx sin x sin y dx dy
l l l
0 0
l l
2   πn πm
+ Uttyy − Uyy sin x sin y dx dy. (20)
l l l
0 0
Integrating twice by parts by x in the first integral on the right-hand side of (20), then integrating
twice by parts by y in the second integral on the right-hand side of (20), and using the conditions (5),
we obtain
un,m (t) − λ2n,m un,m (t) = 0, (21)
where
μ2n,m π(n + m)
λ2n,m = , μn,m = .
1 + μ2n,m l

511
 The differential equations (21) with λ = λn,m coincide with Eq. (9). Further, taking into account
the condition (3), from (19), we obtain

l l
2 πn πm
un,m (0) = U (0, x, y) sin x sin y dx dy
l l l
0 0
l l
2 πn πm
= U (β, x, y) sin x sin y dx dy = un,m (β). (22)
l l l
0 0

The condition (22) coincides with the condition (10). Then for the problem (21), (22), (17) we obtain
the formula
1 − e−λn,m β λn,m t
un,m (t) = bn,m ·e + e−λn,m t . (23)
eλn,m β − 1
To find the constants bn,m , we use the integral condition (4) and the formula (19):

β  l  l β
2 πn πm
un,m (t)dt = U (t, x, y)dt sin x sin y dx dy
l l l
0 0 0 0
l l
2 πn πm
= ϕ(x, y) sin x sin y dx dy = ϕn,m . (24)
l l l
0 0

Then from (23) and (24) we obtain the following relation:

β β
1 − e−λn,m β λn,m t
ϕn,m = un,m (t)dt = bn,m λ β
·e + e−λn,m t dt
e n,m −1
0 0
β
1 − e−λn,m β eλn,m t e−λn,m t 2bn,m
= bn,m · − = σ1n,m (β), (25)
eλn,m β − 1 λn,m λn,m 0 λn,m
where σ1n,m (β) = 1 − e−λn,m β > 0.
We find bn,m from (25),
λn,m
bn,m = ϕn,m . (26)
2σ1n,m (β)
Substituting the coefficients (26) in the formula (23), we obtain
λn,m eλn,m t e−λn,m t
un,m (t) = ϕn,m + , (27)
2 σ2n,m (β) σ1n,m (β)
where σ2n,m (β) = eλn,m β − 1 > 0.
Substituting the functions (27) in the series (18), we arrive at the formal solution of the problem 1:
∞
2  λn,m eλn,m t e−λn,m t πn πm
U (t, x, y) = ϕn,m + sin x sin y, (28)
l 2 σ2n,m (β) σ1n,m (β) l l
n,m=1

where
μ2n,m π(n + m)
λn,m = , μn,m = .
1 + μ2n,m l

512
 Now we assume that ϕ(x, y) ≡ 0. Then ϕn,m ≡ 0, and the formulas (19) and (28) imply
l l
πn πm
U (t, x, y) sin x sin y dx dy = 0, n, m = 1, 2, . . . .
l l
0 0
   
2 πn 2 πm
Hence, due to the completeness of the systems of eigenfunctions sin x and sin y
l l l l
in L2 [0, l], we conclude that U (t, x, y) ≡ 0 for all x, y ∈ [0, l] and t ∈ [0, β].
Therefore, if the problem 1 has a solution, then this solution is unique in the three-dimensional
domain ω.

4. Existence of a solution. We show that under certain conditions for the function ϕ(x, y), the
series (28) converges absolutely and uniformly. For any n, m and 0 < β, the following estimates are
valid:
       
un,m (t) ≤ C1 ϕn,m , un,m (t) ≤ C1 ϕn,m , (29)
where 0 < C1 = const.
Indeed, since 0 < λn,m < 1 and λn,m → 1 as n, m → ∞, using the formula (27) we obtain
  eβ 1  
un,m (t) ≤ + ϕn,m .
σ2n,m (β) σ1n,m (β)
Differentiating the expressions (27) twice, we similarly get
   eβ 1  
un,m (t) ≤ + ϕn,m .
σ2n,m (β) σ1n,m (β)
Hence the estimates (29) hold, where
eβ 1
C1 = . +
σ2n,m (β) σ1n,m (β)
 
Condition 1. Let a function ϕ(x, y) ∈ C 3 [0; l] × [0; l] on a segment [0; l] have piecewise-continuous
derivatives of the fourth order and let
ϕ(0, y) = ϕ(l, y) = ϕ(x, 0) = ϕ(x, l) = 0,
ϕxx (0, y) = ϕxx (l, y) = ϕxx (x, 0) = ϕxx (x, l) = 0,
ϕyy (0, y) = ϕyy (l, y) = ϕyy (x, 0) = ϕyy (x, l) = 0.
Let the conditions 1 be satisfied. Then, integrating by parts four times by the variable x in (24),
we have
l l
2 πn πm
ϕn,m = ϕ(x, y) sin x sin y dx dy;
l l l
0 0
we set  4
l pn,m
ϕn,m =− , (30)
π n4
where
∞
 l l
4  2
p2n,m ≤ 2 ϕxxxx (x, y) dx dy < ∞. (31)
n,m=1
l
0 0

513
Similarly, integrating by parts four times by the variable y, we obtain
 4
l pn,m
ϕn,m = − , (32)
π m4
where
∞
 l l
4  2
p2n,m ≤ 2 ϕyyyy (x, y) dx dy < ∞. (33)
l
n,m=1 0 0
Taking into account the estimates (29)–(31) and applying the Cauchy–Bunyakovsky inequality, for
the series (28) we obtain
  2  ∞    πn   πm  2C1  
∞ 
     
U (t, x, y) ≤  n,m  
u (t) · sin x ·
  sin y  ≤  n,m 
ϕ
l n,m=1 l l l n,m=1
 
   ∞
1  
∞
   ∞ 1  
≤ γ1 ≤ γ1   p2n,m
4  n,m 
p 8
n,m=1
n n,m=1
n n,m=1
 
   l l
∞   
2γ1 
 1  2
≤ 8
 ϕxxxx (x, y) dx dy < ∞, (34)
l n,m=1
n
0 0

where
 4
2C1 l
γ1 = .
l π
Similarly, taking into account the estimates (29), (32), and (33), we obtain
 
 l l
  2γ   ∞
1 
   2
  1

U (t, x, y) ≤ 8
ϕyyyy (x, y) dx dy < ∞. (35)
l n,m=1
m
0 0

From (34) and (35) we see that the series (28) converges absolutely and uniformly in the three-
dimensional domain ω.
For the function (28), we show the continuity of all the derivatives involving in Eq. (6). First, we
formally calculate all required derivatives of the function (28):
∞
2  λ3n,m eλn,m t e−λn,m t πn πm
Utt (t, x, y) = ϕn,m + sin x sin y, (36)
l 2 σ2n,m (β) σ1n,m (β) l l
n,m=1

e−λn,m t  πn 2
∞

2 λn,m eλn,m t πn πm
Uxx (t, x, y) = − ϕn,m + sin x sin y, (37)
l 2 σ2n,m (β) σ1n,m (β) l l l
n,m=1

e−λn,m t  πm 2
∞
2 λn,m eλn,m t πn πm
Uyy (t, x, y) = − ϕn,m + sin x sin y, (38)
l 2 σ2n,m (β) σ1n,m (β) l l l
n,m=1

e−λn,m t  πn 2
∞
2 λ3n,m eλn,m t πn πm
Uttxx (t, x, y) = − ϕn,m + sin x sin y, (39)
l 2 σ2n,m (β) σ1n,m (β) l l l
n,m=1

e−λn,m t  πm 2
∞
2 λ3n,m eλn,m t πn πm
Uttyy (t, x, y) = − ϕn,m + sin x sin y. (40)
l 2 σ2n,m (β) σ1n,m (β) l l l
n,m=1

514
Taking into account the estimates (29)–(31), we obtain from (36)
  2  ∞    πn   πm  2C1  
∞ 
      
Utt (t, x, y) ≤ un,m (t) · sin x · sin y ≤ ϕn,m 
l n,m=1 l l l n,m=1
 
   l l
∞   
2γ1 
 1  2
≤ 8
 ϕxxxx (x, y) dx dy < ∞. (41)
l n
n,m=1 0 0

Taking into account the estimates (29), (32), and (33), we obtain
 
 l l
  2γ   ∞
1 
   2
  1

 tt
U (t, x, y) ≤ 8
ϕ yyyy (x, y) dx dy < ∞. (42)
l n,m=1
m
0 0

Now taking into account the estimates (29)–(31), we obtain from (37)
  2π 2 ∞   πn   πm  ∞  
  2  2 
Uxx (t, x, y) ≤ 3 n un,m (t) sin x sin y  ≤ γ2 n ϕn,m 
l n,m=1 l l n,m=1
 
 
1  
∞ ∞ ∞  
   1    2 
≤ γ2 p n,m  ≤ γ2   p n,m 
n,m=1
n2 n,m=1
n4 n,m=1
 
   l l
 ∞    2
2γ2  1 
≤ 4
 ϕ xxxx (x, y) dx dy < ∞, (43)
l n,m=1
n
0 0

where
2C1 π 2
γ2 = .
l3
Using (43), from (38), we obtain
  2π 2 ∞   πn   πm  ∞  
     
Uyy (t, x, y) ≤ 3 m2 un,m (t) sin x sin y  ≤ γ2 m2 ϕn,m 
l n,m=1 l l n,m=1
 
   ∞ 
1   
∞
   ∞ 1    2 
≤ γ2 pn,m  ≤ γ2   pn,m 
m2 m4
n,m=1 n,m=1 n,m=1
 
   l l
∞   
2γ2 
 1  2
≤ 4
 ϕyyyy (x, y) dx dy < ∞. (44)
l m
n,m=1 0 0

Similarly to (41)–(44), for (39) and (40) we can easily show that
   
   
Uttxx (t, x, y) < ∞, Uttyy (t, x, y) < ∞.
Hence, the function U (t, x, y) defined by the series (28) satisfies the condition (2). Thus, the following
theorem is proved.
Theorem 1. Let the condition 1 be satisfied. Then the problem 1 is uniquely solvable in the three-
dimensional domain ω. The solution is defined by the series (28). In this case, termwise differentiation
of the series (28) by all variables is possible and the resulting series converge absolutely and uniformly.

515
5. Integro-differential equation (1). Now we search for a nontrivial function U (t, x, y) in the
three-dimensional domain ω satisfying Eq. (1) and the conditions (2)–(5). We search for a nontrivial
solution of the problem 2 in the form of a Fourier series (18). Substituting the series (18) into Eq. (1),
we get
β k
 2 2
un,m (t) − λn,m un,m (t) = νλn,m ai (t)bi (s)un,m (s)ds, (45)
0 i=1

where the functions un,m (t) are defined by the formula (19) and
μ2n,m π(n + m)
λ2n,m = , μn,m = .
1 + μ2n,m l
Using the notation
β
τin,m = bi (s)un,m (s)ds, (46)
0
we rewrite Eqs. (45) in the form
k

un,m (t) − λ2n,m un,m (t) = νλ2n,m ai (t)τin,m . (47)
i=1

We solve the differential equations (47) by the method of variation of constants:

k
 t
λn,m t −λn,m t
un,m (t) = cn,m e + dn,m e + νλn,m τin,m sinh λn,m (t − s)ai (s)ds. (48)
i=1 0

To find the unknown coefficients cn,m and dn,m in (48), we use the condition (22):
σ2n,m (β) −λn,m t
un,m (t) = cn,m eλn,m t + e + ξn,m (t), (49)
σ1n,m (β)
where
ηn,m (β) −λn,m t
ξn,m(t) = e + ηn,m (t), σ1n,m (β) = 1 − e−λn,m β > 0,
σ1n,m (β)
k t
ηn,m (t) = νλn,m τin,m sinh λn,m (t − s)ai (s)ds, σ2n,m (β) = eλn,m β − 1 > 0.
i=1 0

To find cn,m in (49), we use the condition (24):

β β
σ2n,m (β) −λn,m t 2cn,m
ϕn,m = un,m (t)dt = cn,m eλn,m t + e dt + γn,m = σ2n,m (β) + γn,m , (50)
σ1n,m (β) λn,m
0 0

where
β
γn,m = ξn,m (t)dt.
0
From (50) we obtain
λn,m  
cn,m = ϕn,m − γn,m . (51)
2σ2n,m (β)

516
Substituting (51) into the formula (49), we obtain
k

un,m (t) = Dn,m (t) − ν τin,mEin,m (t), (52)
i=1

where
Dn,m (t) = ϕn,m Bn,m (t),
  β
e−λn,m t
Ein,m (t) = Bn,m (t) − δin,m (β) − δin,m (t) + Bn,m (t) δin,m (t)dt,
σ1n,m (β)
0
λn,m σ2n,m (β) −λn,m t
Bn,m (t) = eλn,m t + e ,
2σ2n,m (β) σ1n,m (β)
t
δin,m (t) = λn,m sinh λn,m (t − s)ai (s)ds.
0

Substituting (52) into (46), we obtain the following infinite (countable) system of algebraic equations:
k

τin,m + ν τjn,mHijn,m = Ψin,m , (53)
j=1

where
β β
Ψin,m = bi (s)Dn,m s)ds, Hijn,m = bi (s)Ejn,m (s)ds.
0 0
Note that the linear independence of the functions ai (t) and bi (s) implies that Hijn,m = 0. The
system (53) is uniquely solvable for any finite Ψin,m if the following condition is satisfied:
 
1 + νH11n,m νH12n,m ... νH1kn,m 

 νH21n,m 1 + νH22n,m . . . νH2kn,m 

δn,m (ν) =  .. .. .. ..  = 0. (54)
 . . . . 
 
 νHk1n,m νHk2n,m . . . 1 + νHkkn,m
The determinant δn,m (ν) in (54) is a polynomial of ν of degree at most k. The equation δn,m (ν) = 0
has at most k different roots. These roots are eigenvalues of the kernel of the integro-differential
equation (1). For other values of ν, the condition (54) is satisfied. For such values of ν, the system (53)
has a unique solution for any finite nonzero right-hand side. Therefore, if the condition (54) is fulfilled,
then the unique solvability of the nonlocal problem 2 is proved.
Then solutions of the system (53) are written in the form
δin,m (ν)
τin,m = , i = 1, k, (55)
δn,m (ν)
where
 
1 + νH11n,m . . . νH1(i−1)n,m Ψ1n,m νH1(i+1)n,m . . . νH1kn,m 

 νH21n,m . . . νH2(i−1)n,m Ψ2n,m νH2(i+1)n,m . . . νH2kn,m 

δin,m (ν) =  . . . . . . .. 
 .. .. .. .. .. .. . 
 
 νHk1n,m . . . νHk(i−1)n,m Ψkn,m νHk(i+1)n,m . . . 1 + νHkkn,m

517
Substituting (55) into (52), we obtain
k
 δin,m (ν)
un,m (t) = Dn,m (t) − ν Ein,m (t). (56)
δn,m (ν)
i=1

Now we substitute (56) into the Fourier series (18):

∞
 k

2   δin,m (ν) πn πm
U (t, x, y) = Dn,m (t) − ν Ein,m (t) sin x sin y. (57)
l δn,m (ν) l l
n,m=1 i=1

Now we prove the uniform convergence of the series (57). First, we consider the convergence of the
series
∞
πn πm
Dn,m (t) sin x sin y. (58)
n,m=1
l l
We take into account the relation Dn,m (t) = ϕn,m Bn,m (t). The smooth functions
λn,m σ2n,m (β) −λn,m t
Bn,m (t) = eλn,m t + e
2σ2n,m (β) σ1n,m (β)
are bounded together with their derivatives up to the second order. Therefore, the following estimates
similar to (29) hold:
        
Dn,m (t) ≤ C1 ϕn,m , Dn,m (t) ≤ C1 ϕn,m , (59)
where
eβ 1
0 < C1 = + .
σ2n,m (β) σ1n,m (β)
Taking into account the estimates (30)–(33) and (59), similarly to (34) and (35), we obtain
∞    πn   πm 
 ∞  
    
Dn,m (t) · sin x · sin y  ≤ C1 ϕn,m 
l l
n,m=1 n,m=1
 
   
1  
∞ ∞
   1  ∞
≤ γ3  p n,m  ≤ γ3   p2n,m
n4 n8
n,m=1 n,m=1 n,m=1
 
   l l
 ∞    2
2γ3  1 
≤  ϕ xxxx (x, y) dx dy < ∞, (60)
l n8
n,m=1 0 0

∞    πn   πm 
 ∞  
    
Dn,m (t) · sin x · sin y  ≤ C1 ϕn,m 
n,m=1
l l n,m=1
 
   l l
∞   
2γ3 
 1  2
≤  ϕyyyy (x, y) dx dy < ∞, (61)
l m8
n,m=1 0 0
 4
where γ3 = C1 l/π . Hence we conclude that the series (58) converges uniformly in the domain ω.
Now we consider the convergence of the series
∞ 
 k
δin,m (ν) πn πm
Ein,m (t) sin x sin y. (62)
δn,m (ν) l l
n,m=1 i=1

518
Since
  β
e−λn,m t
Ein,m (t) = Bn,m (t) − δin,m (β) − δin,m (t) + Bn,m (t) δin,m (t)dt,
σ1n,m (β)
0
t
δin,m (t) = λn,m sinh λn,m (t − s)ai (s)ds,
0

where 0 < λn,m < 1 and ai (t) ∈ C 2 [0; β],

the smoothness of these functions implies
   
    
Ein,m (t) ≤ C2 , Ein,m (t) ≤ C2 , (63)
 
 
where i = 1, k and 0 < C2 = const. The condition (54) implies that δn,m (ν) > 0. The determinants
δin,m (ν) contain the column
β
Ψin,m = bi (s)Dn,m (s)ds,
0

where bi (t) ∈ C 2 [0; β].

Taking into account the estimates (59)–(61) and (63) and the properties of the
determinants, we obtain
∞ 
 k 
 
 δ in,m (ν) πn πm 
 Ein,m (t) sin x sin y
 δn,m (ν) l l 
n,m=1 i=1
   
 ∞  k  ∞ 
 k 
    
1
≤ C2  δin,m (ν) ≤ C C
2 3  δin,m (ν)
δn,m (ν)    
n,m=1 i=1 n,m=1 i=1
k   1  
∞
 ∞
     
≤ C1 C2 C3 ϕn,m  δ in,m (ν) ≤ γ4 p n,m 
n4
n,m=1 i=1 n,m=1
   
   ∞    l l
 ∞ 1   ∞   
  2γ4 
 1  2
≤ γ4 8
p2n,m ≤ 8
 ϕxxxx (x, y) dx dy < ∞, (64)
n,m=1
n n,m=1
l n,m=1
n
0 0

 
∞ 
πm 
 k
 δin,m (ν) πn
 Ein,m (t) sin x sin y
 δn,m (ν) l l 
n,m=1 i=1
 
k      
1  
∞ ∞ ∞
        1  ∞
≤ C1 C2 C3 ϕn,m  δ in,m (ν) ≤ γ4 p  ≤ γ   p2n,m
m 4 n,m 4
m 8
n,m=1 i=1 n,m=1 n,m=1 n,m=1
 
   l l
 ∞    2
2γ4  1 
≤ 8
 ϕ yyyy (x, y) dx dy < ∞, (65)
l n,m=1
m
0 0

where
 4  −1 k
  
l    
γ4 = C1 C2 C3 C4 , C3 = lim δn,m (ν) , C4 = lim δ n,m (ν),
π n,m→∞ n,m→∞
i=1

519
  
1 + νH11n,m . . . νH1(i−1)n,m ψ1 νH1(i+1)n,m . . . νH1kn,m 

 νH21n,m . . . νH2(i−1)n,m ψ2 νH2(i+1)n,m . . . νH2kn,m 

δ in,m (ν) =  .. .. .. .. .. .. .. ,
 . . . . . . . 
 
 νHk1n,m . . . νHk(i−1)n,m ψk νHk(i+1)n,m . . . 1 + νHkkn,m
where
β
ψi = bi (t)dt.
0
From (64) and (65) we conclude that the series (62) converges uniformly in the domain ω. The
convergence of the series (58) and (62) implies the convergence of the series (57).
To establish the uniqueness of the solution of problem 2, we assume that ϕ(x, y) ≡ 0. Then
ϕn,m ≡ 0. Hence Dn,m (t) = ϕn,m Bn,m (t) ≡ 0 and
β
Ψin,m = bi (s)Dn,m (s)ds ≡ 0, δin,m (ν) ≡ 0.
0
Hence, from the formula (57) we have
l l
πn πm
U (t, x, y) sin x sin y dx dy = 0, n, m = 1, 2, . . . .
l l
0 0
   
2 πn 2 πm
Hence, due to the completeness of the systems of eigenfunctions sin x , sin y
l l l l
in L2 [0, l], we conclude that U (t, x, y) ≡ 0 for all x, y ∈ [0, l] and t ∈ [0, β]. Therefore, the solution of
the problem 2 is unique in the domain ω.
We prove the continuity of all derivatives of the function (57) involving in Eq. (1). We find all
required formal derivatives of the function (57):
∞
 k

2  
 δin,m (ν)  πn πm
Utt (t, x, y) = Dn,m (t) − ν Ein,m (t) sin x sin y, (66)
l δ (ν) l l
n,m=1 i=1 n,m
 
2 
∞ k
δin,m (ν)  πn 2 πn πm
Uxx (t, x, y) = − Dn,m (t) − ν Ein,m (t) sin x sin y, (67)
l n,m=1 δ (ν) l l l
i=1 n,m
 
2 
∞ k
δin,m (ν)  πm 2 πn πm
Uyy (t, x, y) = − Dn,m (t) − ν Ein,m (t) sin x sin y, (68)
l δ (ν) l l l
n,m=1 i=1 n,m
 
2 
∞ k
δin,m (ν)   πn 2 πn πm

Uttxx (t, x, y) = − Dn,m (t) − ν Ein,m (t) sin x sin y, (69)
l δ (ν) l l l
n,m=1 i=1 n,m
 
2 
∞ k
δin,m (ν)  πm 2 πn πm
Uttyy (t, x, y) = − Dn,m (t) − ν Ein,m (t) sin x sin y, (70)
l δn,m (ν) l l l
n,m=1 i=1

where

Dn,m (t) = λ2n,m Dn,m (t) = λ2n,m ϕn,m Bn,m (t),

 λ3n,m σ2n,m (β) −λn,m t

Bn,m (t) = λ2n,m Bn,m (t) = eλn,m t + e ,
2σ2n,m (β) σ1n,m (β)

520
and
 
 e−λn,m t
Ein,m (t) = λ2n,m Bn,m (t) − 
δin,m (β) − δin,m (t)
σ1n,m (β)
β β β
+ λ2n,m Bn,m(t) δin,m (t)dt 
+ 2Bn,m (t) 
δin,m (t)dt + Bn,m (t) 
δin,m (t)dt,
0 0 0

t

δin,m (t) = λ2n,m ai (t) + λ3n,m sinh λn,m (t − s)ai (s)ds.
0
In the proof of the absolute and uniform convergence of series (66)–(70), we use the method applied
for obtaining the estimates (60), (61), (64), and (65). Taking into account the formulas (30)–(33),
(59), and (63), similarly to the estimates (60), (61), (64), and (65), we obtain
  2  ∞    πn   πm 
     
Utt (t, x, y) ≤ Dn,m (t) · sin x · sin y
l l l
n,m=1
 
2ν   δin,m (ν)    πn   πm 
∞ k
 
+  E (t) · sin x · sin y
l  δn,m (ν) in,m 
n,m=1 i=1
l l
∞ ∞ k  
2C1    2ν

 


  
≤ ϕn,m + C1 C2 C3 ϕn,m δ in,m (ν)
l l
n,m=1 n,m=1 i=1
 
  
l l
 ∞ 1    2
≤ γ5   ϕ xxxx (x, y) dx dy < ∞, (71)
n,m=1
n8
0 0

 
   l l
   ∞    2
  1 
Utt (t, x, y) ≤ γ5  8
 ϕyyyy (x, y) dx dy < ∞, (72)
n,m=1
m
0 0

  2π 2  ∞    πn   πm 
    
Uxx (t, x, y) ≤ 3 n2 Dn,m (t) · sin x · sin y
l l l
n,m=1
 
2νπ 2  2  δin,m (ν)   πn   πm 
∞ k
 
+ 3 n  Ein,m (t) · sin x · sin y
l  δn,m (ν)  l l
n,m=1 i=1
∞ ∞ k  
2π 2 C1  2   2νπ 2

 
2

  
≤ n ϕn,m + C C C
1 2 3 n ϕn,m δ in,m (ν)
l3 l3
n,m=1 n,m=1 i=1
 
  
l l
 ∞ 1    2
≤ γ6  4
 ϕxxxx (x, y) dx dy < ∞, (73)
n,m=1
n
0 0

 
  l l
 



   1 
∞
  2
Uyy (t, x, y) ≤ γ6   ϕyyyy (x, y) dx dy < ∞, (74)
m 4
n,m=1 0 0

521
where
2(γ3 + νγ4 2π 2 (γ3 + νγ4
γ5 = , γ6 = .
l l3
Similarly to (71)–(74), for (69) and (70) we have
   
   
Uttxx (t, x, y) < ∞, Uttyy (t, x, y) < ∞.
Thus, we have proved the following theorem.
Theorem 2. Let the conditions of Theorem 1 be satisfied. If the condition (45) is fulfilled, then the
problem 1 is uniquely solvable in the three-dimensional domain ω and its solution is defined by the
series (57). Moreover, termwise differentiation of the series (57) by all variables is possible, and the
resulting series converge absolutely and uniformly.

REFERENCES
1. A. M. Akhtyamov and A. P. Ayupova, “On the solution of the problem of diagnosing defects in
a small cavity in a rod,” Zh. Srednevolzh. Mat. Obshch., 12, No. 3, 37–42 (2010).
2. R. N. Antontsev, A. V. Kazhikhov, and V. N. Monahov, Boundary-Value Problems in Mechanics
of Inhomogeneous Liquids [in Russian], Nauka, Novosibirsk (1983).
3. Yu. P. Apakov, “On the solution of the boundary-value problem for a third-order equation with
multiple characteristics,” Ukr. Mat. Zh., 64, No. 1, 1–11 (2012).
4. A. T. Asanova, “On a nonlocal boundary-value problem for systems of hyperbolic equations with
impulsive effects,” Ukr. Mat. Zh., 65, No. 3, 315–328 (2013).
5. D. J. Benney and J. C. Luke, “Interactions of permanent waves of finite amplitude,” J. Math.
Phys., 43, 309–313 (1964).
6. M. Kh. Beshtokov, “A numerical method for solving a nonlocal boundary-value problem for a
third-order equation of hyperbolic type,” Zh. Vychisl. Mat. Mat. Fiz., 54, No. 9, 1497–1514
(2014).
7. A. A. Boychuk and A. P. Strakh, “Noether boundary-value problems for systems of linear integro-
dynamical equations with degenerate kernel on a time scale,” Nelin. Kolebaniya, 17, No. 1, 32–38
(2014).
8. V. A. Chernyatin, Fourier Method in Mixed Problems for Partial Differential Equations [in Rus-
sian], Moscow State Univ. (1991).
9. D. S. Dzhumabaev and E. A. Bakirova, “On the unique solvability of the boundary-value problem
for systems of Fredholm integro-differential equations with degenerate kernels,” Nelin. Kolebaniya,
18, No. 4, 489–506 (2015).
10. T. D. Dzhuraev and A. Sopuev, Theory of Fourth-Order Partial Differential Equations [in Rus-
sian], Fan, Tashkent (2000).
11. U. Egamberdiev and Yu. P. Apakov, “On the Dirichlet problem for a mixed elliptic-hyperbolic
equation in a three-dimensional domain,” Izv. Akad. Nauk Uzbek. SSR. Ser. Fiz. Mat., 3, 51–56
(1989).
12. D. G. Gordzeani and G. A. Avalishvili, “Solutions of nonlocal problems for one-dimensional
oscillations of the medium,” Mat. Model., 12, No. 1, 94–103 (2000).
13. V. A. Ilyin, “On the solvability of mixed problems for the hyperbolic and parabolic equations,”
Usp. Mat. Nauk, 15, No. 2 (92), 97–154 (1960).
14. N. I. Ivanchov, “Boundary-value problems for a parabolic equation with an integral condition,”
Differ. Uravn., 40, No. 4, 547–564 (2004).
15. N. Lazhetich, “On the existence of a classical solution of a mixed problem for a one-dimensional,
second-order hyperbolic equation,” Differ. Uravn., 34, No. 5, 682–694 (1998).

522
16. N. V. Martemyanova, “The Dirichlet problem for an equation of mixed elliptic-hyperbolic type
with variable potential,” Izv. Vyssh. Ucheb. Zaved. Mat., 11, 44–53 (2015).
17. E. I. Moiseev, “On the solution by a spectral method of a nonlocal boundary-value problem,”
Differ. Uravn., 35, No. 8, 1094–1100 (1999).
18. L. S. Pulkina, “Nonlocal problem for a hyperbolic equation with integral conditions of the first
kind with time-dependent kernels,” Izv. Vyssh. Ucheb. Zaved. Mat., 10, 32–44 (2012).
19. O. A. Repin, “On a problem with two nonlocal boundary conditions for a mixed-type equation,”
Dokl. Ross. Akad. Nauk, 365, No. 5, 593–595 (1999).
20. K. B. Sabitov, “A nonlocal problem for a parabolic-hyperbolic equation in a rectangular domain,”
Mat. Zametki, 89, No. 4, 596–602 (2011).
21. A. M. Samoilenko, A. A. Boichuk, and S. A. Krivosheya, “Boundary-value problems for systems
of integro-differential equations with degenerate kernels,” Ukr. Mat. Zh., 48, No. 11, 1785–1789
(1996).
22. R. A. Shabrov, “On estimates of the influence function for a fourth-order mathematical model,”
Izv. Voronezh Univ. Ser. Fiz. Mat., 2, 168–179 (2015).
23. I. V. Skrypnik, Methods for Investigating Nonlinear Elliptic Boundary-Value Problems [in Rus-
sian], Nauka, Moscow (1990).
24. R. K. Tagiev and V. M. Gabibov, “On a problem of optimal control for the heat equation with an
integral boundary condition,” Vestn. Samar. Gos. Univ. Ser. Fiz. Mat., 20, No. 1, 54–64 (2016).
25. I. V. Tikhonov, “Uniqueness theorems in linear nonlocal problems for abstract differential equa-
tions,” Izv. Ross. Akad. Nauk. Ser. Mat., 67, No. 2, 133–166 (2003).
26. M. V. Turbin, “Initial-boundary-value problem for the Herschel–Balkley model of the motion of
a fluid,” Izv. Voronezh Univ. Ser. Fiz. Mat., 2, 246–257 (2013).
27. G. B. Whitham, Linear and Nonlinear Waves, Wiley (1974).
28. T. K. Yuldashev, “Mixed problems for a fourth-order nonlinear differential equation with a small
parameter multiplied by a parabolic operator,” Zh. Vychisl. Mat. Mat. Fiz., 51, No. 9, 1703–1711
(2011).
29. T. K. Yuldashev, “Mixed problems for a nonlinear integro-differential equation with higher-order
parabolic operators,” Zh. Vychisl. Mat. Mat. Fiz., 52, No. 1, 112–123 (2012).
30. T. K. Yuldashev, “On a third-order partial integro-differential Fredholm equation,” Izv. Vyssh.
Ucheb. Zaved. Mat., 9, 74–79 (2015).
31. T. K. Yuldashev, “Nonlocal mixed problem for an integro-differential equation of Boussinesq type
with degenerate kernel,” Ukr. Mat. Zh., 68, No. 8, 1115–1131 (2016).
32. T. K. Yuldashev, “On a fourth-order mixed differential equation,” Izv. Inst. Mat. Mekh. Udmurt.
Gos. Univ., 47, No. 1, 119–128 (2016).
33. T. K. Yuldashev, “Mixed problem for a pseudoparabolic integro-differential equation with a de-
generate kernel,” Differ. Uravn., 53, No. 1, 101–110 (2017).

T. K. Yuldashev
M. F. Reshetnev Siberian State University of Science and Technologies, Krasnoyarsk, Russia
E-mail: tursun.k.yuldashev@gmail.com

523