On the Solvability of a Problem Nonlocal in Time for a Semilinear Multidimensional Wave Equation

We study a nonlocal (in time) problem for semilinear multidimensional wave equations. The theorems on existence and uniqueness of solutions of this problem are proved. Вивчається нелокальна за часом задача для напівлінійних багатовимiрних хвильових рівнянь. Доведено теореми про існування та єдиність...

Повний опис

Збережено в:
Бібліографічні деталі
Опубліковано в: :Український математичний журнал
Дата:2015
Автори: Kharibegashvili, S., Midodashvili, B.
Формат: Стаття
Мова:English
Опубліковано: Інститут математики НАН України 2015
Теми:
Онлайн доступ:https://nasplib.isofts.kiev.ua/handle/123456789/165435
Теги: Додати тег
Немає тегів, Будьте першим, хто поставить тег для цього запису!
Назва журналу:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Цитувати:On the Solvability of a Problem Nonlocal in Time for a Semilinear Multidimensional Wave Equation / S. Kharibegashvili, B. Midodashvili // Український математичний журнал. — 2015. — Т. 67, № 1. — С. 88–105. — Бібліогр.: 23 назв. — англ.

Репозитарії

Digital Library of Periodicals of National Academy of Sciences of Ukraine
id nasplib_isofts_kiev_ua-123456789-165435
record_format dspace
spelling Kharibegashvili, S.
Midodashvili, B.
2020-02-13T12:58:21Z
2020-02-13T12:58:21Z
2015
On the Solvability of a Problem Nonlocal in Time for a Semilinear Multidimensional Wave Equation / S. Kharibegashvili, B. Midodashvili // Український математичний журнал. — 2015. — Т. 67, № 1. — С. 88–105. — Бібліогр.: 23 назв. — англ.
1027-3190
https://nasplib.isofts.kiev.ua/handle/123456789/165435
517.9
We study a nonlocal (in time) problem for semilinear multidimensional wave equations. The theorems on existence and uniqueness of solutions of this problem are proved.
Вивчається нелокальна за часом задача для напівлінійних багатовимiрних хвильових рівнянь. Доведено теореми про існування та єдиність розв'язків цієї задачі.
en
Інститут математики НАН України
Український математичний журнал
Статті
On the Solvability of a Problem Nonlocal in Time for a Semilinear Multidimensional Wave Equation
Про розв'язність нелокальної за часом задачі для напівлінійного багатовимірного хвильового рівняння
Article
published earlier
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
collection DSpace DC
title On the Solvability of a Problem Nonlocal in Time for a Semilinear Multidimensional Wave Equation
spellingShingle On the Solvability of a Problem Nonlocal in Time for a Semilinear Multidimensional Wave Equation
Kharibegashvili, S.
Midodashvili, B.
Статті
title_short On the Solvability of a Problem Nonlocal in Time for a Semilinear Multidimensional Wave Equation
title_full On the Solvability of a Problem Nonlocal in Time for a Semilinear Multidimensional Wave Equation
title_fullStr On the Solvability of a Problem Nonlocal in Time for a Semilinear Multidimensional Wave Equation
title_full_unstemmed On the Solvability of a Problem Nonlocal in Time for a Semilinear Multidimensional Wave Equation
title_sort on the solvability of a problem nonlocal in time for a semilinear multidimensional wave equation
author Kharibegashvili, S.
Midodashvili, B.
author_facet Kharibegashvili, S.
Midodashvili, B.
topic Статті
topic_facet Статті
publishDate 2015
language English
container_title Український математичний журнал
publisher Інститут математики НАН України
format Article
title_alt Про розв'язність нелокальної за часом задачі для напівлінійного багатовимірного хвильового рівняння
description We study a nonlocal (in time) problem for semilinear multidimensional wave equations. The theorems on existence and uniqueness of solutions of this problem are proved. Вивчається нелокальна за часом задача для напівлінійних багатовимiрних хвильових рівнянь. Доведено теореми про існування та єдиність розв'язків цієї задачі.
issn 1027-3190
url https://nasplib.isofts.kiev.ua/handle/123456789/165435
citation_txt On the Solvability of a Problem Nonlocal in Time for a Semilinear Multidimensional Wave Equation / S. Kharibegashvili, B. Midodashvili // Український математичний журнал. — 2015. — Т. 67, № 1. — С. 88–105. — Бібліогр.: 23 назв. — англ.
work_keys_str_mv AT kharibegashvilis onthesolvabilityofaproblemnonlocalintimeforasemilinearmultidimensionalwaveequation
AT midodashvilib onthesolvabilityofaproblemnonlocalintimeforasemilinearmultidimensionalwaveequation
AT kharibegashvilis prorozvâznístʹnelokalʹnoízačasomzadačídlânapívlíníinogobagatovimírnogohvilʹovogorívnânnâ
AT midodashvilib prorozvâznístʹnelokalʹnoízačasomzadačídlânapívlíníinogobagatovimírnogohvilʹovogorívnânnâ
first_indexed 2025-11-25T22:34:34Z
last_indexed 2025-11-25T22:34:34Z
_version_ 1850567818007805952
fulltext UDC 517.9 S. Kharibegashvili (I. Javakhishvili Tbilisi State Univ. and Georg. Techn. Univ., Georgia), B. Midodashvili (I. Javakhishvili Tbilisi State Univ. and Gori Teaching Univ., Georgia) ON THE SOLVABILITY OF A PROBLEM NONLOCAL IN TIME FOR A SEMILINEAR MULTIDIMENSIONAL WAVE EQUATION ПРО РОЗВ’ЯЗНIСТЬ НЕЛОКАЛЬНОЇ ЗА ЧАСОМ ЗАДАЧI ДЛЯ НАПIВЛIНIЙНОГО БАГАТОВИМIРНОГО ХВИЛЬОВОГО РIВНЯННЯ We study a nonlocal (in time) problem for semilinear multidimensional wave equations. The theorems on existence and uniqueness of solutions of this problem are proved. Вивчається нелокальна за часом задача для напiвлiнiйних багатовимiрних хвильових рiвнянь. Доведено теореми про iснування та єдинiсть розв’язкiв цiєї задачi. 1. Introduction. In the space Rn+1 of variables x = (x1, . . . , xn) and t, in the cylindrical domain DT = Ω × (0, T ), where Ω is a Lipschitz domain in Rn, consider a nonlocal problem of finding a solution u(x, t) of the following equation: Lλu := ∂2u ∂t2 − n∑ i=1 ∂2u ∂x2 i + λf(x, t, u) = F (x, t), (x, t) ∈ DT , (1.1) satisfying the homogeneous boundary condition on the part of the boundary Γ := ∂Ω× (0, T ) of the cylinder DT u ∣∣ Γ = 0, (1.2) the initial condition u(x, 0) = ϕ(x), x ∈ Ω, (1.3) and the nonlocal condition Kµut := ut(x, 0)− µut(x, T ) = ψ(x), x ∈ Ω, (1.4) where f, F, ϕ and ψ are given functions; λ and µ are given nonzero constants and n ≥ 2. To the study of nonlocal problems for partial differential equations there are devoted many papers. When a nonlocal problem is posed for abstract evolution equations and hyperbolic partial differential equations we would suggest the reader refer to works [1 – 15] and the references therein. Note that the problem (1.1) – (1.4) in the work [15] is studied in the class of continuous functions for the case of one spatial variable, i.e., for n = 1. The method of investigation given in the work [15], based on the integral representation of the solution of corresponding linear problem, is useless for multidimensional case, i.e., for n > 1. In this work the problem (1.1) – (1.4) in the multidimensional case is studied in the Sobolev space W 1 2 (DT ), basing on expansions of the functions from the space 0 W 1 2(Ω) in the basis, consisting of eigenfunctions of spectral problem ∆w = λ̃w, w |∂Ω = 0 and using embedding theorems in the Sobolev spaces. It must be noted also that if for n = 1 there is no need of any restriction on the behavior of function f(x, t, u) with respect to variable u when u→∞, c© S. KHARIBEGASHVILI, B. MIDODASHVILI, 2015 88 ISSN 1027-3190. Укр. мат. журн., 2015, т. 67, № 1 ON THE SOLVABILITY OF A PROBLEM NONLOCAL IN TIME . . . 89 while in the case n > 1, we require of function f(x, t, u) that for u → ∞ it must have a growth not exceeding polynomial. Moreover, for using the embedding theorems in the Sobolev spaces we additionally require that the order of polynomial growth must be less than a certain value, which depends of the dimension of the space. Below, on the function f = f(x, t, u) we impose the following requirements: f ∈ C(DT × R), |f(x, t, u)| ≤M1 +M2|u|α, (x, t, u) ∈ DT × R, (1.5) where 0 ≤ α = const < n+ 1 n− 1 . (1.6) Remark 1.1. The embedding operator I : W 1 2 (DT ) → Lq(DT ) represents a linear continuous compact operator for 1 < q < 2(n+ 1) n− 1 ,when n > 1 [16]. At the same time the Nemitski operatorN : Lq(DT )→ L2(DT ), acting by the formula Nu = f(x, t, u) due to (1.5) is continuous and bounded if q ≥ 2α [17]. Thus, since due to (1.6) we have 2α < 2(n+ 1) n− 1 , then there exists the number q such that 1 < q < 2(n+ 1) n− 1 and q ≥ 2α. Therefore, in this case the operator N0 = NI : 0 W 1 2(DT ,Γ)→ L2(DT ), (1.7) where 0 W 1 2(DT ,Γ) := {w ∈ W 1 2 (DT ) : w |Γ = 0}, will be continuous and compact. Besides, from u ∈ 0 W 1 2(DT ,Γ) it follows that f(x, t, u) ∈ L2(DT ) and, if um → u in the space 0 W 1 2(DT ,Γ), then f(x, t, um)→ f(x, t, u) in the space L2(DT ). Definition 1.1. Let function f satisfy conditions (1.5) and (1.6), F ∈ L2(DT ), ϕ ∈ 0 W 1 2(Ω) := := {v ∈ W 1 2 (Ω) : v |∂Ω = 0}, ψ ∈ L2(Ω). We call a function u a generalized solution of prob- lem (1.1) – (1.4), if u ∈ 0 W 1 2(DT ,Γ) and there exists a sequence of functions um ∈ 0 C 2(DT ,Γ) := := { w ∈ C2(DT ) : w |Γ = 0 } such that lim m→∞ ‖um − u‖ 0 W 1 2(DT ,Γ) = 0, lim m→∞ ‖Lλum − F‖L2(DT ) = 0, (1.8) lim m→∞ ‖um |t=0 − ϕ‖ 0 W 1 2(Ω) = 0, lim m→∞ ‖Kµumt − ψ‖L2(Ω) = 0. (1.9) It is obvious that a classical solution u ∈ C2(DT ) of problem (1.1) – (1.4) represents a generalized solution of this problem. It is easy to verify that a generalized solution of problem (1.1) – (1.4) is a solution of problem (1.1) in the sense of the theory of distributions. Indeed, let Fm := Lλum, ϕm := um |t=0, ψm := Kµumt. Multiplying the both sides of the equality Lλum = Fm by test function w ∈ V := { v ∈ 0 W 1 2(DT ,Γ) : v(x, T )−µv(x, 0) = 0, x ∈ Ω } and integrating in the domain DT , after simple transformations, connected with integration by parts and the equality w |Γ = 0, we get ∫ Ω [umt(x, T )w(x, T )− umt(x, 0)w(x, 0)]dx+ ISSN 1027-3190. Укр. мат. журн., 2015, т. 67, № 1 90 S. KHARIBEGASHVILI, B. MIDODASHVILI + ∫ DT [ −umtwt + n∑ i=1 umxiwxi + λf(x, t, um)w ] dx dt = ∫ DT Fmw dxdt ∀w ∈ V. (1.10) Due toKµumt = ψm(x) andw(x, T )−µw(x, 0) = 0, x ∈ Ω, it is easy to see that umt(x, T )w(x, T )− − umt(x, 0)w(x, 0) = umt(x, T )(w(x, T ) − µw(x, 0)) − ψm(x)w(x, 0) = −ψm(x)w(x, 0), x ∈ Ω. Therefore, equality (1.10) takes the form − ∫ Ω ψm(x)w(x, 0)dx+ + ∫ DT [ −umtwt + n∑ i=1 umxiwxi + λf(x, t, um)w ] dx dt = ∫ DT Fmw dxdt ∀w ∈ V. (1.11) In view of (1.5), (1.6) according to Remark 1.1 we have f(x, t, um) → f(x, t, u) in the space L2(DT ), when um → u in the space 0 W 1 2(DT ,Γ). Therefore, due to (1.8) and (1.9), passing to the limit in equality (1.11) for m→∞, we get − ∫ Ω ψw(x, 0)dx+ ∫ DT [ −utwt + n∑ i=1 uxiwxi + λf(x, t, u)w ] dx dt = ∫ DT Fwdx dt ∀w ∈ V. (1.12) Since C∞0 (DT ) ⊂ V, then from (1.12), integrating by parts, we have∫ DT [ u2w + λf(x, t, u)w ] dx dt = ∫ DT Fw dx dt ∀w ∈ C∞0 (DT ), (1.13) where 2 := ∂2/∂t2 − ∑n i=1 ∂2/∂x2 i , and C∞0 (DT ) is a space of finite infinitely differentiable functions in DT . Equality (1.13), which is valid for any w ∈ C∞0 (DT ), means that a generalized solution u of problem (1.1) – (1.4) is a solution of equation (1.1) in the sense of the theory of distributions, besides, since the trace operator u → u |t=0 is well defined in the space 0 W 1 2(DT ,Γ), and, particularly, is a continuous operator from the space 0 W 1 2(DT ,Γ) into the space L2(Ω×{t = 0}), then due to (1.8) and (1.9) we receive that the initial condition (1.3) is fulfilled in the sense of the trace theory, while the nonlocal condition (1.4) in the integral sense is taken into account in equality (1.12), which is valid for all w ∈ V. Note also that if a generalized solution u belongs to the class C2(DT ), then due to the standard reasoning, connected with the integral equality (1.12), which is valid for any w ∈ V [16], we have that u is a classical solution of problem (1.1) – (1.4), satisfying the equation (1.1), the boundary condition (1.2), the initial condition (1.3) and the nonlocal condition (1.4) pointwisely. Note that even in the linear case, i.e., for λ = 0, problem (1.1) – (1.4) is not always well-posed. For example, when λ = 0 and |µ| = 1, the corresponding to (1.1) – (1.4) homogeneous problem may have infinite number of linearly independent solutions (see Remark 3.2). The work is organized in the following way. In Section 2 we single out the class of semilinear equations (1.1), when for |µ| < 1 a priori estimate is valid for the generalized solution of prob- lem (1.1) – (1.4). In Section 3 on the basis of a priori estimate, received in the previous section, the ISSN 1027-3190. Укр. мат. журн., 2015, т. 67, № 1 ON THE SOLVABILITY OF A PROBLEM NONLOCAL IN TIME . . . 91 solvability of problem (1.1) – (1.4) is proved. Finally, in Section 4 we give the conditions imposed on the data of the problem, which provide the uniqueness of the solution of this problem. 2. A priori estimate of the solution of problem (1.1) – (1.4). Let g(x, t, u) = u∫ 0 f(x, t, s)ds, (x, t, u) ∈ DT × R. (2.1) Consider the following conditions imposed on function g = g(x, t, u): g(x, t, u) ≥ −M3, (x, t, u) ∈ DT × R, (2.2) gt ∈ C(DT × R), gt(x, t, u) ≤M4, (x, t, u) ∈ DT × R, (2.3) where Mi = const ≥ 0, i = 3, 4. Let us consider some classes of functions f = f(x, t, u) frequently encountered in applications and which satisfy conditions (1.5), (2.2) and (2.3) : 1. f(x, t, u) = f0(x, t)β(u), where f0, ∂ ∂t f0 ∈ C(DT ) and β ∈ C(R), |β(u)| ≤ M̃1 + M̃2|u|α, M̃i = const ≥ 0, α = const ≥ 0. In this case g(x, t, u) = f0(x, t) ∫ u 0 β(s)ds and when f0 ≥ 0, ∂ ∂t f0 ≤ 0, ∫ u 0 β(s)ds ≥ −M, M = const ≥ 0, conditions (1.5), (2.2) and (2.3) will be fulfilled. 2. f(x, t, u) = f0(x, t)|u|α signu, where f0, ∂ ∂t f0 ∈ C(DT ) and α > 1. In this case g(x, t, u) = = f0(x, t) |u|α+1 α+ 1 and when f0 ≥ 0, ∂ ∂t f0 ≤ 0, conditions (1.5), (2.2) and (2.3) will be also fulfilled. Lemma 2.1. Let λ > 0, |µ| < 1, F ∈ L2(DT ), ϕ ∈ 0 W 1 2(Ω), ψ ∈ L2(Ω) and conditions (1.5), (1.6), (2.2), (2.3) be fulfilled. Then for a generalized solution u of problem (1.1) – (1.4) the following a priori estimate: ‖u‖ 0 W 1 2(DT ,Γ) ≤ c1‖F‖L2(DT ) + c2‖ϕ‖ 0 W 1 2(Ω) + c3‖ψ‖L2(Ω) + c4‖ϕ‖ α+1 2 0 W 1 2(Ω) + c5 (2.4) is valid with nonnegative constants ci = ci(λ, µ,Ω, T,M1,M2,M3,M4) not depending on u, F, ϕ, ψ, and ci > 0 for i < 4, whereas in the linear case, i.e., when λ = 0 the constants c4 = c5 = 0 and due to (2.4) in this case we have the uniqueness of the solution of problem (1.1) – (1.4). Proof. Let u be a generalized solution of problem (1.1) – (1.4). In view of Definition 1.1 there exists a sequence of the functions um ∈ 0 C 2(DT ,Γ) such that limit equalities (1.8), (1.9) are fulfilled. Set Lλum = Fm, (x, t) ∈ DT , (2.5) um |Γ = 0, (2.6) um(x, 0) = ϕm(x), x ∈ Ω, (2.7) Kµumt = ψm(x), x ∈ Ω. (2.8) ISSN 1027-3190. Укр. мат. журн., 2015, т. 67, № 1 92 S. KHARIBEGASHVILI, B. MIDODASHVILI Multiplying both sides of equation (2.5) by 2umt and integrating in the domain Dτ := DT ∩{t < < τ}, 0 < τ ≤ T, due to (2.1), we obtain∫ Dτ ∂ ∂t (∂um ∂t )2 dx dt− 2 ∫ Dτ n∑ i=1 ∂2um ∂x2 i ∂um ∂t dx dt+ 2λ ∫ Dτ d dt ( g(x, t, um(x, t) ) dx dt− −2λ ∫ Dτ gt(x, t, um(x, t))dx dt = 2 ∫ Dτ Fm ∂um ∂t dx dt. (2.9) Let ωτ := {(x, t) ∈ DT : x ∈ Ω, t = τ}, 0 ≤ τ ≤ T. Denote by ν := (νx1 , νx2 , . . . , νxn , νt) the unit vector of the outer normal to ∂Dτ . Since νxi ∣∣ ωτ∪ω0 = 0, i = 1, . . . , n, νt ∣∣ Γτ=Γ∩{t≤τ} = 0, νt ∣∣ ωτ = 1, νt ∣∣ ω0 = −1, then, taking into account equalities (2.6) and integrating by parts, we have∫ Dτ ∂ ∂t (∂um ∂t )2 dx dt = ∫ ∂Dτ (∂um ∂t )2 νtds = ∫ ωτ u2 mtdx− ∫ ω0 u2 mtdx, (2.10) −2 ∫ Dτ ∂2um ∂x2 i ∂um ∂t dx dt = ∫ Dτ [(u2 mxi)t − 2(umxiumt)xi ]dx dt = = ∫ ωτ u2 mxidx− ∫ ω0 u2 mxidx, i = 1, . . . , n, (2.11) 2λ ∫ Dτ d dt ( g(x, t, um(x, t) ) dxdt = 2λ ∫ ∂Dτ g(x, t, um(x, t))νtds = = 2λ ∫ ωτ g(x, t, um(x, t))dx− 2λ ∫ ω0 g(x, t, um(x, t))dx. (2.12) In view of (2.10), (2.11), (2.12) from (2.9) we get∫ ωτ [ u2 mt + n∑ i=1 u2 mxi ] dx = ∫ ω0 [ u2 mt + n∑ i=1 u2 mxi ] dx− 2λ ∫ ωτ g(x, t, um(x, t))dx+ +2λ ∫ ω0 g(x, t, um(x, t))dx+ 2λ ∫ Dτ gt(x, t, um(x, t))dx dt+ 2 ∫ Dτ Fmumtdx dt. (2.13) Let wm(τ) := ∫ ωτ [ u2 mt + n∑ i=1 u2 mxi ] dx. (2.14) Since 2Fmumt ≤ ε−1F 2 m + εu2 mt for any ε = const > 0, then due to (2.2), (2.3) and (2.14) from (2.13) it follows that ISSN 1027-3190. Укр. мат. журн., 2015, т. 67, № 1 ON THE SOLVABILITY OF A PROBLEM NONLOCAL IN TIME . . . 93 wm(τ) ≤ wm(0) + 2λM3 mes Ω + 2λ ∫ ω0 |g(x, t, um(x, t))|dx+ +2λM4τ mes Ω + ε ∫ DT u2 mtdx dt+ ε−1 ∫ DT F 2 mdx dt. (2.15) Taking into account that ∫ Dτ u2 mtdx dt = τ∫ 0  ∫ ωs u2 mtdx ds ≤ τ∫ 0  ∫ ωs [ u2 mt + n∑ i=1 u2 mxi ] dx ds = τ∫ 0 wm(s)ds, from (2.15) we obtain wm(τ) ≤ ε τ∫ 0 wm(s)ds+ wm(0) + 2λ(M3 +M4τ) mes Ω+ +2λ ∫ ω0 |g(x, t, um(x, t))|dx+ ε−1 ∫ Dτ F 2 mdx dt, 0 < τ ≤ T. (2.16) Because of Dτ ⊂ DT , 0 < τ ≤ T, then according to the Gronwall’s lemma [18] from (2.16) it follows that wm(τ) ≤ [ wm(0) + 2λ(M3 +M4T ) mes Ω+ +2λ ∫ ω0 |g(x, t, um(x, t))|dx+ ε−1 ∫ DT F 2 mdx dt ] eετ , 0 < τ ≤ T. (2.17) Using obvious inequality |a+ b|2 = a2 + b2 + 2ab ≤ a2 + b2 + ε1a 2 + ε−1 1 b2 = (1 + ε1)a2 + (1 + ε−1 1 )b2, which is valid for any ε1 > 0, from (2.8) we have |umt(x, 0)|2 = |µumt(x, T ) + ψm(x)|2 ≤ |µ|2(1 + ε1)u2 mt(x, T ) + (1 + ε−1 1 )ψ2 m(x). (2.18) From (2.18) we obtain ∫ ω0 u2 mtdx = ∫ Ω |umt(x, 0)|2 dx ≤ ≤ |µ|2(1 + ε1) ∫ Ω u2 mt(x, T )dx+ (1 + ε−1 1 ) ∫ Ω ψ2 m(x) dx = ISSN 1027-3190. Укр. мат. журн., 2015, т. 67, № 1 94 S. KHARIBEGASHVILI, B. MIDODASHVILI = |µ|2(1 + ε1) ∫ ωT u2 mtdx+ (1 + ε−1 1 )‖ψm‖2L2(Ω). (2.19) In view of (2.7), (2.14) from (2.17) we get∫ ωT u2 mtdx ≤ wm(T ) ≤  ∫ ω0 n∑ i=1 ϕ2 mxidx+ ∫ ω0 u2 mtdx+M5 eεT , (2.20) where M5 = 2λ(M3 +M4T ) mes Ω + 2λ ∫ ω0 |g(x, t, um(x, t))|dx+ ε−1 ∫ DT F 2 m dx dt. (2.21) From (2.19) and (2.20) it follows that∫ ω0 u2 mtdx ≤ |µ|2(1 + ε1) ∫ ω0 n∑ i=1 ϕ2 mxidx+ ∫ ω0 u2 mtdx+M5 eεT + (1 + ε−1 1 )‖ψm‖2L2(Ω). (2.22) Because |µ| < 1, then positive constants ε and ε1 can be chosen so small that µ1 = |µ|2(1 + ε1)eεT < 1. (2.23) Due to (2.23) from (2.22) we obtain∫ ω0 u2 mtdx ≤ (1− µ1)−1 |µ|2(1 + ε1) ∫ ω0 n∑ i=1 ϕ2 mxidx+M5 eεT + (1 + ε−1 1 )‖ψm‖2L2(Ω)  ≤ ≤ (1− µ1)−1 [ |µ|2(1 + ε1) ( ‖ϕm‖20 W 1 2(Ω) +M5 ) eεT + (1 + ε−1 1 )‖ψm‖2L2(Ω) ] . (2.24) From (2.7), (2.14) and (2.24) it follows that wm(0) = ∫ ω0 [ u2 mt + n∑ i=1 ϕ2 mxi ] dx ≤ ‖ϕm‖20 W 1 2(Ω) + +(1− µ1)−1 [ |µ|2(1 + ε1) ( ‖ϕm‖20 W 1 2(Ω) +M5 ) eεT + (1 + ε−1 1 )‖ψm‖2L2(Ω) ] . (2.25) In view of (2.21), (2.25) from (2.17) we get wm(τ) ≤ ‖ϕm‖20W 1 2(Ω) + (1− µ1)−1 × × |µ|2(1 + ε1) ‖ϕm‖20 W 1 2(Ω) + 2λ(M3 +M4T ) mes Ω + ISSN 1027-3190. Укр. мат. журн., 2015, т. 67, № 1 ON THE SOLVABILITY OF A PROBLEM NONLOCAL IN TIME . . . 95 + 2λ ∫ ω0 |g(x, t, um(x, t))|dx+ ε−1 ∫ DT F 2 mdx dt eεT + (1 + ε−1 1 )‖ψm‖2L2(Ω + + 2λ(M3 +M4T ) mes Ω + 2λ ∫ ω0 |g(x, t, um(x, t))|dx+ ε−1 ∫ DT F 2 mdx dt  eεT = = γ̃1‖Fm‖2L2(DT ) + γ̃2‖ϕm‖20 W 1 2(Ω) + γ̃3‖ψm‖2L2(Ω) + γ̃4 ∫ ω0 |g(x, t, um(x, t))|dx+ γ̃5. (2.26) Here γ̃1 = ε−1eεT [ (1− µ1)−1(1 + ε1)eεT + 1 ] , γ̃2 = eεT [ 1 + (1− µ1)−1|µ|2(1 + ε1) ] , γ̃3 = (1− µ1)−1(1 + ε−1 1 )eεT , γ̃4 = 2λ[(1− µ1)−1|µ|2(1 + ε1) + 1]eεT , γ̃5 = 2λ(M3 +M4T ) mes Ω[(1− µ1)−1|µ|2(1 + ε1)eεT + 1]eεT . (2.27) Since for fixed τ the function um(x, τ) ∈ 0 W 1 2(Ω), then due to the Friedrichs inequality [16] we have ∫ ωτ [ u2 m + u2 mt + n∑ i=1 u2 mxi ] dx ≤ c0wm(τ) = c0 ∫ ωτ [ u2 mt + n∑ i=1 u2 mxi ] dx, (2.28) where positive constant c0 = c0(Ω) does not depend on um. From (2.26) and (2.28) it follows ‖um‖20 W 1 2(DT ,Γ) = T∫ 0  ∫ ωτ ( u2 m + u2 mt + n∑ i=1 u2 mxi ) dx dτ ≤ ≤ T∫ 0 c0wm(τ)dτ ≤ c0T γ̃1‖Fm‖2L2(DT ) + c0T γ̃2‖ϕm‖20 W 1 2(Ω) + c0T γ̃3‖ψm‖2L2(Ω)+ +c0T γ̃4 ∫ Ω ∣∣g(x, 0, um(x, 0)) ∣∣dx+ c0T γ̃5. (2.29) Due to (2.1), (1.5) we have |g(x, 0, s)| ≤M6 +M7|s|α+1, (2.30) ISSN 1027-3190. Укр. мат. журн., 2015, т. 67, № 1 96 S. KHARIBEGASHVILI, B. MIDODASHVILI where M6 and M7 are some nonnegative constants. Taking into account (2.30) from (2.29) we get ‖um‖20 W 1 2(DT ,Γ) ≤ c0T γ̃1‖Fm‖2L2(DT ) + c0T γ̃2‖ϕm‖20 W 1 2(Ω) + c0T γ̃3‖ψm‖2L2(Ω)+ +c0T γ̃4M6 mes Ω + c0T γ̃4M7 ∫ Ω |um(x, 0)|α+1dx+ c0T γ̃5. (2.31) Reasoning from Remark 1.1, concerning the space W 1 2 (Ω), in view of the equality dim Ω = = dimDT − 1 = n show that the embedding operator I : W 1 2 (Ω) → Lq(Ω) is a linear continuous compact operator for 1 < q < 2n n− 2 , when n > 2 and for any q > 1 when n = 2 [16]. At the same time the Nemitski operator N1 : Lq(Ω)→ L2(Ω), acting by the formula N1u = |u| α+1 2 is continuous and bounded if q ≥ 2 α+ 1 2 = α + 1 [17]. Thus, if α + 1 < 2n n− 2 , i.e., α < n+ 2 n− 2 , which is fulfilled due to (1.6) since n+ 1 n− 1 < n+ 2 n− 2 , then there exists number q such that 1 < q < 2n n− 2 and q ≥ α+ 1. Therefore, in this case the operator N2 = N1I : W 1 2 (Ω)→ L2(Ω) will be continuous and compact. Thus due to (1.9), (2.7) it follows that lim m→∞ ∫ Ω |um(x, 0)|α+1dx = ∫ Ω |ϕ(x)|α+1dx, (2.32) and also [16] ∫ Ω |ϕ(x)|α+1dx ≤ C1‖ϕ‖α+1 0 W 1 2(Ω) (2.33) with positive constant C1, not dependent on ϕ ∈ 0 W 1 2(Ω). In view of (1.8), (1.9), (2.5) – (2.8), (2.32) and (2.33), passing in (2.31) to the limit for m → ∞ we obtain ‖u‖20 W 1 2(DT ,Γ) ≤ c0T γ̃1‖F‖2L2(DT ) + c0T γ̃2‖ϕ‖20 W 1 2(Ω) + c0T γ̃3‖ψ‖2L2(Ω)+ +c0T γ̃4M7C1‖ϕ‖α+1 0 W 1 2(Ω) + c0T (γ̃5 + γ̃4M6 mes Ω). (2.34) Taking the square root from the both sides of inequality (2.34) and using the obvious inequality(∑k i=1 a2 i )1/2 ≤ ∑k i=1 |ai| we finally get ‖u‖ 0 W 1 2(DT ,Γ) ≤ c1‖F‖L2(DT ) + c2‖ϕ‖ 0 W 1 2(Ω) + c3‖ψ‖L2(Ω) + c4‖ϕ‖ α+1 2 0 W 1 2(Ω) + c5. (2.35) Here ISSN 1027-3190. Укр. мат. журн., 2015, т. 67, № 1 ON THE SOLVABILITY OF A PROBLEM NONLOCAL IN TIME . . . 97 c1 = (c0T γ̃1)1/2, c2 = (c0T γ̃2)1/2, c3 = (c0T γ̃3)1/2, c4 = (c0T γ̃4M7C1)1/2, c5 = [c0T (γ̃5 + γ̃4M6 mes Ω)]1/2, (2.36) where γ̃i, 1 ≤ i ≤ 5, are defined in (2.27). In the linear case, i.e., for λ = 0, due to (2.27) the constants γ̃4 = γ̃5 = 0 and from (2.36) it follows that in estimate (2.4) the constants c4 = c5 = 0. Whence it follows the uniqueness of the solution of problem (1.1) – (1.4) in the linear case. Lemma 2.1 is proved. 3. The existence of the solution of problem (1.1) – (1.4). For the existence of the solution of problem (1.1) – (1.4) in the case |µ| < 1 we will use well known facts about the solvability of the following linear mixed problem [16]: L0u := ∂2u ∂t2 − n∑ i=1 ∂2u ∂x2 i = F (x, t), (x, t) ∈ DT , (3.1) u ∣∣ Γ = 0, u(x, 0) = ϕ(x), ut(x, 0) = ψ̃(x), x ∈ Ω, (3.2) where F, ϕ and ψ̃ are given functions. For F ∈ L2(DT ), ϕ ∈ 0 W 1 2(Ω), ψ̃ ∈ L2(Ω) the unique generalized solution u of problem (3.1), (3.2) (in the sense of equality (1.12) where f = 0, and the number µ = 0 in the definition of the space V and u |t=0 = ϕ) from the class E2,1(DT ) with the norm [16] ‖v‖2E2,1(DT ) = sup 0≤τ≤T ∫ ωτ [ u2 + u2 t + n∑ i=1 u2 xi ] dx is given by formula [16] u = ∞∑ k=1 ak cosµkt+ bk sinµkt+ 1 µk t∫ 0 Fk(τ) sinµk(t− τ)dτ ϕk(x), (3.3) where λ̃k = −µ2 k, 0 < µ1 ≤ µ2 ≤ . . . , limk→∞ µk = ∞ are the eigenvalues, while ϕk ∈ 0 W 1 2(Ω) are the corresponding eigenfunctions of the spectral problem ∆w = λ̃w, w |∂Ω = 0 in the domain Ω ( ∆ := ∑n i=1 ∂2 ∂x2 i ) , simultaneously forming orthonormal basis in L2(Ω) and orthogonal basis in 0 W 1 2(Ω) in the sense of scalar product (v, w) 0 W 1 2(Ω) = ∫ Ω ∑n i=1 vxiwxidx [16], i.e., (ϕk, ϕl)L2(Ω) = δlk, (ϕk, ϕl) 0 W 1 2(Ω) = −λkδlk, δlk = 1, l = k, 0, l 6= k. (3.4) Here ak = (ϕ,ϕk)L2(Ω), bk = µ−1 k (ψ̃, ϕk)L2(Ω), k = 1, 2, . . . , (3.5) ISSN 1027-3190. Укр. мат. журн., 2015, т. 67, № 1 98 S. KHARIBEGASHVILI, B. MIDODASHVILI F (x, t) = ∞∑ k=1 Fk(t)ϕk(x), Fk(t) = (F,ϕk)L2(ωt), ωτ := DT ∩ {t = τ}, (3.6) besides, for the solution u from (3.3) it is valid the following estimate [16, 19]: ‖u‖E2,1(DT ) ≤ γ(‖F‖L2(DT ) + ‖ϕ‖ 0 W 1 2(Ω) + ‖ψ̃‖L2(Ω)) (3.7) with positive constant γ, not dependent on F, ϕ and ψ̃. Let us consider the linear problem corresponding to (1.1) – (1.4), i.e., the case when λ = 0 : L0u := ∂2u ∂t2 − n∑ i=1 ∂2u ∂x2 i = F (x, t), (x, t) ∈ DT , (3.8) u|Γ = 0, u(x, 0) = ϕ(x), Kµut = ψ(x), x ∈ Ω. (3.9) Let us show that when |µ| < 1 for any F ∈ L2(DT ), ϕ ∈ 0 W 1 2(Ω) and ψ ∈ L2(Ω) there exists a unique generalized solution of problem (3.8), (3.9) in the sense of Definition 1.1 for λ = 0. Indeed, for ϕ ∈ 0 W 1 2(Ω) and ψ ∈ L2(Ω) there are valid the expansions ϕ = ∑∞ k=1 akϕk and ψ = ∑∞ k=1 dkϕk in the spaces 0 W 1 2(Ω) and L2(Ω), respectively, where ak = (ϕ,ϕk)L2(Ω) and dk = (ψ,ϕk)L2(Ω) [16]. Therefore, setting ϕm = m∑ k=1 akϕk, ψm = m∑ k=1 dkϕk, (3.10) we have lim m→∞ ‖ϕm − ϕ‖ 0 W 1 2(Ω) = 0, lim m→∞ ‖ψm − ψ‖L2(Ω) = 0. (3.11) Since the space of finite infinitely differentiable functions C∞0 (DT ) is dense in the space L2(DT ), then for F ∈ L2(DT ) and any natural number m there exists a function Fm ∈ C∞0 (DT ) such that ‖Fm − F‖L2(DT ) < 1 m . (3.12) On the other hand, for function Fm in the space L2(DT ) there is valid the following expansion [16]: Fm(x, t) = ∞∑ k=1 Fm,k(t)ϕk(x), Fm,k(t) = (Fm, ϕk)L2(Ω). (3.13) Therefore, there exists a natural number lm such that limm→∞ lm =∞ and for F̃m(x, t) = lm∑ k=1 Fm,k(t)ϕk(x) (3.14) the inequality ISSN 1027-3190. Укр. мат. журн., 2015, т. 67, № 1 ON THE SOLVABILITY OF A PROBLEM NONLOCAL IN TIME . . . 99 ‖F̃m − Fm‖L2(DT ) < 1 m (3.15) is valid. From (3.12) and (3.15) it follows lim m→∞ ‖F̃m − F‖L2(DT ) = 0. (3.16) The solution u = um of problem (3.1), (3.2) for ϕ = ϕlm , ψ̃ = ∑lm k=1 d̃kϕk and F = F̃m, where ϕlm and F̃m are defined in (3.10) and (3.14), is given by formula (3.3) which, due to (3.4) – (3.6), takes the form um = lm∑ k=1 ak cosµkt+ d̃k µk sinµkt+ 1 µk t∫ 0 Fm,k(τ) sinµk(t− τ)dτ ϕk(x). (3.17) For determination of the coefficients d̃k let us substitute the right-hand side of expression (3.17) into the equality Kµumt = ψlm(x), where ψlm is defined in (3.10). Consequently, taking into account that the system of functions {ϕk(x)} represents a basis in L2(Ω) and 1− µ cosµkT 6= 0 for |µ| < 1, we obtain the following formulas: d̃k = 1 1− µ cosµkT (ϕlm , ϕk)L2(Ω) − akµµk sinµkT + + µ T∫ 0 Fm,k(τ) cosµk(T − τ)dτ , k = 1, . . . , lm. (3.18) Below we assume that the Lipschitz domain Ω is such that eigenfunctions ϕk ∈ C2(Ω), k ≥ 1. For example, this will take place if ∂Ω ∈ C [n/2]+3 [19]. This fact will also take place in the case of a piece- wisely smooth Lipschitz domain, e.g., for the parallelepiped Ω = {x ∈ Rn : |xi| < ai, i = 1, . . . , n} the correspondent eigenfunctions ϕk ∈ C∞(Ω) [20]. Therefore, since Fm ∈ C∞0 (DT ), then due to (3.13) the function Fm,k ∈ C2([0, T ]), and consequently the function um from (3.17) belongs to the space C2(DT ). Further, since ϕk|∂Ω = 0, then due to (3.17) we have um |Γ = 0, and thereby um ∈ 0 C 2(DT ,Γ), m = 1, 2, . . . . According to the construction the function um from (3.17) satisfies the following equalities: um |Γ = 0, L0um = F̃m, um(x, 0) = ϕlm(x), Kµumt = ψlm(x), x ∈ Ω, (3.19) and thereby (um − uk)|Γ = 0, L0(um − uk) = F̃m − F̃k, (um − uk)(x, 0) = (ϕlm − ϕlk)(x), Kµ(umt − ukt) = (ψlm − ψlk)(x), x ∈ Ω. Therefore, from a priori estimate (2.4), where for λ = 0 the coefficients c4 = c5 = 0, we obtain ‖um − uk‖ 0 W 1 2(DT ,Γ) ≤ c1‖F̃m − F̃k‖L2(DT ) + c2‖ϕlm − ϕlk‖ 0 W 1 2(Ω) + c3‖ψlm − ψlk‖L2(Ω). (3.20) ISSN 1027-3190. Укр. мат. журн., 2015, т. 67, № 1 100 S. KHARIBEGASHVILI, B. MIDODASHVILI In view of (3.11) and (3.16) from (3.20) it follows that the sequence um ∈ 0 C 2(DT ,Γ) is fundamental in the complete space 0 W 1 2(DT ,Γ). Therefore, there exists a function u ∈ 0 W 1 2(DT ,Γ) such that due to (3.11), (3.16) and (3.19) there are valid the limit equalities (1.8), (1.9) for λ = 0. The last means that the function u is a generalized solution of problem (3.8), (3.9). The uniqueness of this solution follows from a priori estimate (2.4) where the constants c4 = c5 = 0 for λ = 0. Therefore, for the solution u of problem (3.8), (3.9) we have u = L−1 0 (F,ϕ, ψ), where L−1 0 : L2(DT )× 0 W 1 2(Ω)× L2(Ω)→ 0 W 1 2(DT ,Γ), which norm due to (2.4) can be estimated as follows: ‖L−1 0 ‖ L2(DT )× 0 W 1 2(Ω)×L2(Ω)→ 0 W 1 2(DT ,Γ) ≤ γ0 = max(c1, c2, c3). (3.21) Due to the linearity of the operator L−1 0 : L2(DT )× 0 W 1 2(Ω)× L2(Ω)→ 0 W 1 2(DT ,Γ) we have a representation L−1 0 (F,ϕ, ψ) = L−1 0 (F, 0, 0) + L−1 0 (0, ϕ, 0) + L−1 0 (0, 0, ψ) = L−1 01 (F ) + L−1 02 (ϕ) + L−1 03 (ψ), (3.22) where L−1 01 : L2(DT )→ 0 W 1 2(DT ,Γ), L−1 02 : 0 W 1 2(Ω)→ 0 W 1 2(DT ,Γ) and L−1 03 : L2(Ω)→ 0 W 1 2(DT ,Γ) are linear continuous operators, besides, according to (3.21) ‖L−1 01 ‖ L2(DT )→ 0 W 1 2(DT ,Γ) ≤ γ0, ‖L−1 02 ‖ 0 W 1 2(Ω)→ 0 W 1 2(DT ,Γ) ≤ γ0, ‖L−1 03 ‖ L2(Ω)→ 0 W 1 2(DT ,Γ) ≤ γ0. (3.23) Remark 3.1. Note, that for F ∈ L2(DT ), ϕ ∈ 0 W 1 2(Ω), ψ ∈ L2(Ω), due to (1.5), (1.6), (3.21) – (3.23) and Remark 1.1 the function u ∈ 0 W 1 2(DT ,Γ) is a generalized solution of problem (1.1) – (1.4) if and only if, when u is a solution of the functional equation u = L−1 01 (−λf(x, t, u)) + L−1 01 (F ) + L−1 02 (ϕ) + L−1 03 (ψ) (3.24) in the space 0 W 1 2(DT ,Γ). Rewrite equation (3.24) in the form u = A0u := −λL−1 01 (N0u) + L−1 01 (F ) + L−1 02 (ϕ) + L−1 03 (ψ), (3.25) where the operator N0 : 0 W 1 2(DT ,Γ) → L2(DT ) from (1.7), according to Remark 1.1 is continuous and compact operator. Therefore, due to (3.23) the operator A0 : 0 W 1 2(DT ,Γ) → 0 W 1 2(DT ,Γ) from (3.25) is also continuous and compact. At the same time, according to Lemma 2.1 and (2.36) for any parameter τ ∈ [0, 1] and for any solution u of the equation u = τA0u with the parameter τ it is valid the same a priori estimate (2.4) with nonnegative constants ci, not dependent on u, F, ϕ, ψ and τ. Therefore, due to the Schaefer’s fixed point theorem [21], equation (3.25), and therefore, due to Remark 3.1 problem (1.1) – (1.4) has at least one solution u ∈ 0 W 1 2(DT ,Γ). Thus, we have proved the following theorem. ISSN 1027-3190. Укр. мат. журн., 2015, т. 67, № 1 ON THE SOLVABILITY OF A PROBLEM NONLOCAL IN TIME . . . 101 Theorem 3.1. Let λ > 0, |µ| < 1, F ∈ L2(DT ), ϕ ∈ 0 W 1 2(Ω), ψ ∈ L2(Ω), conditions (1.5), (1.6), (2.2), (2.3) be fulfilled. Then problem (1.1) – (1.4) has at least one generalized solution. Remark 3.2. Note that for |µ| = 1, even in the linear case, i.e., for f = 0, the homogeneous problem corresponding to (1.1) – (1.4) may have finite or even infinite number of linearly independent solutions. Indeed, in the case µ = 1 denote by Λ(1) the set of points µk from (3.3), for which the ratio µkT 2π is a natural number, i.e., Λ(1) = { µk : µkT 2π ∈ N } . If we search for a solution of problem (3.8), (3.9) in the form of representation (3.3), then for determination of unknown coefficients bk,contained in it, let us substitute the right-hand side of this representation into the equality Kµut = ψ(x). As a result we have µk(1− µ cosµkT )bk = (ψ,ϕk)L2(Ω) − akµk sinµkT + T∫ 0 Fk(τ) cosµk(T − τ)dτ. (3.26) It is obvious, that when Λ(1) 6= ∅ and µk ∈ Λ(1), µ = 1 we have 1− cosµkT = 0 and for F = 0, ϕ = ψ = 0 and thereby for ak = 0, Fk(τ) = 0 equality (3.26) will be satisfied by any number bk. Therefore, in accordance with (3.3) the function uk(x, t) = C sinµktϕk(x), C = const 6= 0, satisfies the homogeneous problem corresponding to (3.8), (3.9). Analogously, in the case µ = −1 denote by Λ(−1) the set of points µk from (3.3) for which the ratio µkT π is odd integer number. In this case 1 − µ cosµkT = 0 for µk ∈ Λ(−1), µ = −1 and the function uk(x, t) = C sinµktϕk(x), C = const 6= 0, is a nontrivial solution of the homogeneous problem corresponding to (3.8), (3.9). For example, when n = 2, Ω = (0, 1) × (0, 1) the eigenvalues and eigenfunctions of the Laplace operator ∆ are [20] λk = −π2(k2 1 + k2 2), ϕk(x1, x2) = sin k1π1x1 sin k2πx2, k = (k1, k2), i.e., µk = π √ k2 1 + k2 2. For k1 = p2−q2, k2 = 2pq, where p and q are any integer numbers we obtain µk = π(p2 + q2) [22]. In this case for T 2 ∈ N we have µkT 2π = (p2 + q2) T 2 ∈ N and according to the said above, when µ = 1 the homogeneous problem corresponding to (3.8), (3.9) has infinite number of linearly independent solutions up,q(x, t) = sinπ(p2 + q2)t sinπ(p2 − q2)x1 sin 2πpqx2 ∀p, q ∈ N. (3.27) Analogously, when µ = −1 the solutions of the homogeneous problem corresponding to (3.8), (3.9) in the case when p is an even number, while q and T odd numbers, are the functions from (3.27). 4. The uniqueness of the solution of problem (1.1) – (1.4). On the function f in equation (1.1) let us impose the following additional requirements: f, f ′u ∈ C(DT × R), |f ′u(x, t, u)| ≤ a+ b|u|γ , (x, t, u) ∈ DT × R, (4.1) where a, b, γ = const ≥ 0. It is obvious that from (4.1) we have condition (1.5) for α = γ + 1 and when γ < 2 n− 1 we get α = γ + 1 < n+ 1 n− 1 and, therefore, condition (1.6) is fulfilled. ISSN 1027-3190. Укр. мат. журн., 2015, т. 67, № 1 102 S. KHARIBEGASHVILI, B. MIDODASHVILI Theorem 4.1. Let |µ| < 1, F ∈ L2(DT ), ϕ ∈ 0 W 1 2(Ω), ψ ∈ L2(Ω) and condition (4.1) be fulfilled for γ < 2 n− 1 , and also hold conditions (2.2), (2.3). Then there exists a positive number λ0 = λ0(F, f, ϕ, ψ, µ,DT ) such that for 0 < λ < λ0 problem (1.1) – (1.4) can not have more than one generalized solution. Proof. Indeed, suppose that problem (1.1) – (1.4) has two different generalized solutions u1 and u2. According to Definition 1.1 there exist sequences of functions ujk ∈ 0 C 2(DT ,Γ), j = 1, 2, such that lim k→∞ ‖ujk − uj‖ 0 W 1 2(DT ,Γ) = 0, lim k→∞ ‖Lλujk − F‖L2(DT ) = 0, (4.2) lim k→∞ ‖ujk|t=0 − ϕ‖ 0 W 1 2(Ω) = 0, lim k→∞ ‖Kµujkt − ψ‖L2(Ω) = 0, j = 1, 2. (4.3) Let w := u2 − u1, wk := u2k − u1k, Fk := Lλu2k − Lλu1k, (4.4) gk := λ(f(x, t, u1k)− f(x, t, u2k)). (4.5) In view of (4.2), (4.3) and (4.4) it is easy to see that lim k→∞ ‖wk − w‖ 0 W 1 2(DT ,Γ) = 0, lim k→∞ ‖Fk‖L2(DT ) = 0, (4.6) lim k→∞ ‖wk|t=0‖ 0 W 1 2(Ω) = 0, lim k→∞ ‖Kµwkt‖L2(Ω) = 0. (4.7) In view of (4.4), (4.5) the function wk ∈ 0 C 2(DT ,Γ) satisfies the following equalities: ∂2wk ∂t2 − n∑ i=1 ∂2wk ∂x2 i = ( Fk + gk ) (x, t), (x, t) ∈ DT , (4.8) wk |Γ = 0, (4.9) wk(x, 0) = ϕ̃k(x), x ∈ Ω, (4.10) Kµwkt := wkt(x, 0)− µwkt(x, T ) = ψ̃k(x), x ∈ Ω, (4.11) where ϕ̃k(x) := u2k(x, 0)− u1k(x, 0), ψ̃k(x) := Kµu2kt −Kµu1kt. First let us estimate the function gk from (4.5). Taking into account the obvious inequality |d1 + d2|γ ≤ 2γ max(|d1|γ , |d2|γ) ≤ 2γ(|d1|γ + |d2|γ) for γ ≥ 0, due to (4.1) we have |f(x, t, u2k)− f(x, t, u1k)| = ∣∣∣∣∣∣(u2k − u1k) 1∫ 0 f ′u(x, t, u1k + τ(u2k − u1k))dτ ∣∣∣∣∣∣ ≤ ISSN 1027-3190. Укр. мат. журн., 2015, т. 67, № 1 ON THE SOLVABILITY OF A PROBLEM NONLOCAL IN TIME . . . 103 ≤ |u2k − u1k| 1∫ 0 (a+ b|(1− τ)u1k + τu2k|γ)dτ ≤ a|u2k − u1k|+ +2γb|u2k − u1k|(|u1k|γ + |u2k|γ) = a|wk|+ 2γb|wk|(|u1k|γ + |u2k|γ). (4.12) In view of (4.5) from (4.12) we obtain ‖gk‖L2(DT ) ≤ λa‖wk‖L2(DT ) + λ2γb‖ |wk|(|u1k|γ + |u2k|γ)‖L2(DT ) ≤ ≤ λa‖wk‖L2(DT ) + λ2γb‖wk‖Lp(DT )‖(|u1k|γ + |u2k|γ)‖Lq(DT ). (4.13) Here we used the Hölder’s inequality [23] ‖v1v2‖Lr(DT ) ≤ ‖v1‖Lp(DT )‖v2‖Lq(DT ), where 1 p + 1 q = 1 r and in the capacity of p, q and r we take p = 2 n+ 1 n− 1 , q = n+ 1, r = 2. (4.14) Since dimDT = n + 1, then according to the Sobolev embedding theorem [17] for 1 ≤ p ≤ ≤ 2(n+ 1) n− 1 we get ‖v‖Lp(DT ) ≤ Cp‖v‖W 1 2 (DT ) ∀v ∈W 1 2 (DT ) (4.15) with positive constant Cp, not dependent on v ∈W 1 2 (DT ). Due to the condition of the theorem γ < 2 n− 1 and, therefore, γ(n+ 1) < 2(n+ 1) n− 1 . Thus, due to (4.14) from (4.15) we have ‖wk‖Lp(DT ) ≤ Cp‖wk‖W 1 2 (DT ), p = 2(n+ 1) n− 1 , k ≥ 1, (4.16) ‖(|u1k|γ + |u2k|γ)‖Lq(DT ) ≤ ‖ |u1k|γ‖Lq(DT ) + ‖ |u2k|γ‖Lq(DT ) = = ‖u1k‖γLγ(n+1)(DT ) + ‖u2k‖γLγ(n+1)(DT ) ≤ C γ γ(n+1) ( ‖u1k‖γW 1 2 (DT ) + ‖u2k‖γW 1 2 (DT ) ) . (4.17) In view of the first equality of (4.2) there exists a natural number k0 such that for k ≥ k0 we obtain ‖uik‖γW 1 2 (DT ) ≤ ‖ui‖γW 1 2 (DT ) + 1, i = 1, 2, k ≥ k0. (4.18) Further, in view of (4.16), (4.17) and (4.18) from (4.13) we get ‖gk‖L2(DT ) ≤ λa‖wk‖L2(DT ) + λ2γbCpC γ γ(n+1)(‖u1‖γW 1 2 (DT ) + +‖u2‖γW 1 2 (DT ) + 2)‖wk‖W 1 2 (DT ) ≤ λM8‖wk‖W 1 2 (DT ), (4.19) ISSN 1027-3190. Укр. мат. журн., 2015, т. 67, № 1 104 S. KHARIBEGASHVILI, B. MIDODASHVILI where we have used the inequality ‖wk‖L2(DT ) ≤ ‖wk‖W 1 2 (DT ), M8 = a+ 2γbCpC γ γ(n+1) ( ‖u1‖γW 1 2 (DT ) + ‖u2‖γW 1 2 (DT ) + 2 ) , p = 2 n+ 1 n− 1 . (4.20) Since a priori estimate (2.4) is valid for λ = 0, then due to (2.27) and (2.36) in this estimate c4 = c5 = 0, and thereby for the solution wk of problem (4.8) – (4.11) the following estimate: ‖wk‖ 0 W 1 2(DT ,Γ) ≤ c0 1‖Fk + gk‖L2(DT ) + c0 2‖ϕ̃k‖ 0 W 1 2(Ω) + c0 3‖ψ̃k‖L2(Ω) (4.21) is valid, where the constants c0 1, c 0 2, c 0 3 do not depend on λ. Because of ‖wk‖ 0 W 1 2(DT ,Γ) = ‖wk‖W 1 2 (DT ) and due to (4.19) from (4.21) we have ‖wk‖ 0 W 1 2(DT ,Γ) ≤ c0 1‖Fk‖L2(DT ) + λc0 1M8‖wk‖ 0 W 1 2(DT ,Γ) + c0 2‖ϕ̃k‖ 0 W 1 2(Ω) + c0 3‖ψ̃k‖L2(Ω). (4.22) Note that since for u1 and u2 it is valid a priori estimate (2.4), then the constant M8 from (4.20) will depend on λ, F, f, ϕ, ψ, DT , besides, due to (2.27) and (2.36) the value of M8 continuously depends on λ for λ ≥ 0 and 0 ≤ lim λ→0+ M8 = M0 8 < +∞. (4.23) Due to (4.23) there exists a positive number λ0 = λ0(F, f, ϕ, ψ, µ,DT ) such that for 0 < λ < λ0 (4.24) we obtain λc0 1M8 < 1. Indeed, let us fix arbitrarily a positive number ε1. Then, due to (4.23), there exists a positive number λ1, such that 0 ≤ M8 < M0 8 + ε1 for 0 ≤ λ < λ1. It is obvious that for λ0 = min ( λ1, (c 0 1(M0 8 + ε1))−1 ) the condition λc0 1M8 < 1 will be fulfilled. Therefore, in the case (4.24) from (4.22) we get ‖wk‖ 0 W 1 2(DT ,Γ) ≤ (1− λc0 1M8)−1 [ c0 1‖Fk‖L2(DT ) + c0 2‖ϕ̃k‖ 0 W 1 2(Ω) + c0 3‖ψ̃k‖L2(Ω) ] , k ≥ k0. (4.25) From (4.2) and (4.4) it follows that limk→∞ ‖wk‖ 0 W 1 2(DT ,Γ) = ‖u2 − u1‖ 0 W 1 2(DT ,Γ) . On the other hand due to (4.6), (4.7) and (4.10), (4.11) from (4.25) we have limk→∞ ‖wk‖ 0 W 1 2(DT ,Γ) = 0. Thus ‖u2 − u1‖ 0 W 1 2(DT ,Γ) = 0, i.e., u2 = u1, which leads to contradiction. Theorem 4.1 is proved. 1. Byszewski L., Lakshmikantham V. Theorem about the existence and uniqueness of a solution of a nonlocal abstract Cauchy problem in a Banach space // Appl. Anal. – 1991. – 4, № 1. – P. 11 – 19. 2. Kiguradze T. Some boundary value problems for systems of linear partial differential equations of hyperbolic type // Mem. Different. Equat. Math. Phys. – 1994. – 1. – P. 1 – 144. 3. Kiguradze T. I. Some nonlocal problems for linear hyperbolic systems (in Russian) // Dokl. Akad. Nauk. – 1995. – 345, № 3. – P. 300 – 302 (English transl.: Dokl. Math. – 1995. – 52, № 3. – P. 376 – 378). 4. Aizicovici S., McKibben M. Existence results for a class of abstract nonlocal Cauchy problems // Nonlinear Anal. – 2000. – 39, № 5. – P. 649 – 668. ISSN 1027-3190. Укр. мат. журн., 2015, т. 67, № 1 ON THE SOLVABILITY OF A PROBLEM NONLOCAL IN TIME . . . 105 5. Gordeziani D. G., Avalishvili G. A. Investigation of the nonlocal initial boundary value problems for some hyperbolic equations // Hirosima Math. J. – 2001. – 31. – P. 345 – 366. 6. Avalishvili G. A. Nonlocal in time problems for evolution equations of second order // J. Appl. Anal. – 2002. – 8, № 2. – P. 245 – 259. 7. Midodashvili B. A nonlocal problem for fourth order hyperbolic equations with multiple characteristics // Electron. J. Different. Equat. – 2002. – 2002, № 85. – P. 1 – 7. 8. Bouziani A. On a class of nonclassical hyperbolic equations with nonlocal conditions // J. Appl. Math. Stochast. Anal. – 2002. – 15, № 2. – P. 135 – 153. 9. Kharibegashvili S. S. On the well-posedness of some nonlocal problems for the wave equation (in Russian) // Differents. Uravneniya. – 2003. – 39, № 4. – P. 539 – 553 (English transl.: Different. Equat. – 2003. – 39, № 4. – P. 577 – 592). 10. Bogveradze G., Kharibegashvili S. On some nonlocal problems for a hyperbolic equation of second order on a plane // Proc. A. Razmadze Math. Inst. – 2004. – 136. – P. 1 – 36. 11. Xue X. Existence of solutions for semilinear nonlocal Cauchy problems in Banach spaces // Electron. J. Different. Equat. – 2005. – 2005, № 64. – P. 1 – 7. 12. Beilin S. A. On a mixed nonlocal problem for a wave equation // Electron. J. Different. Equat. – 2006. – 2006, № 103. – P. 1 – 10. 13. Hernández E. Existence of solutions for an abstract second-order differential equation with nonlocal conditions // Electron. J. Different. Equat. – 2009. – 2009, № 96. – P. 1 – 10. 14. Kharibegashvili S., Midodashvili B. Some nonlocal problems for second order strictly hyperbolic systems on the plane // Georg. Math. J. – 2010. – 17, № 2. – P. 287 – 303. 15. Kharibegashvili S., Midodashvili B. Solvability of nonlocal problems for semilinear one-dimensional wave equations // Electron. J. Different. Equat. – 2012. – 2012, № 28. – P. 1 – 16. 16. Ladyzhenskaya O. A. The boundary value problems of mathematical physics. – New York: Springer-Verlag, 1985. 17. Kufner A., Fučik S. Nonlinear differential equations. – Amsterdam; New York: Elsevier, 1980. 18. Beckenbach E., Bellman R. Inequalities. – Berlin: Springer-Verlag, 1961. 19. Mikhailov V. P. Partial differential equations. – Moscow: Mir, 1978. 20. Mikhlin S. G. Mathematical physics. An advanced course. – Amsterdam: North-Holland, 1970. 21. Evans L. C. Partial differential equations // Grad. Stud. Math. – Providence, RI: Amer. Math. Soc., 1998. – 19. 22. Jones G. A., Jones J. M. Elementary number theory. – Springer, 1998. 23. Reed M., Simon B. Methods of modern mathematical physics. II: Fourier analysis. Selfadjointness. – New York etc.: Acad. Press, 1975. Received 20.11.13, after revision — 16.08.14 ISSN 1027-3190. Укр. мат. журн., 2015, т. 67, № 1