Applied Mathematics and Computation 219 (2012) 1093–1107 Contents lists available at SciVerse ScienceDirect Applied Mathematics and Computation journal homepage: www.elsevier.com/locate/amc On Bitsadze–Samarskii type nonlocal boundary value problems for elliptic differential and difference equations: Well-posedness Allaberen Ashyralyev a,⇑, Elif Ozturk b a b Department of Mathematics, Fatih University, 34500 Buyukcekmece, Istanbul, Turkey Department of Mathematics, Uludag University, 16340 Bursa, Turkey a r t i c l e i n f o a b s t r a c t The well-posedness of the Bitsadze–Samarskii type nonlocal boundary value problem in Hölder spaces with a weight is established. The coercivity inequalities for the solution of the nonlocal boundary value problem for elliptic equations are obtained. The stable second order of accuracy difference scheme for the approximate solution of the problem is presented. The well-posedness of this difference scheme in difference analogue of Hölder spaces is established. For applications, the almost coercivity and the coercivity estimates for solutions of difference schemes for elliptic equations are obtained. Ó 2012 Elsevier Inc. All rights reserved. Keywords: Elliptic equations Nonlocal boundary value problems Difference schemes Stability 1. Introduction The role played by coercive inequalities in the study of local boundary-value problems for elliptic and parabolic differential equations is well known (see, e.g., [1,2]). In the present paper, we consider the Bitsadze–Samarskii type nonlocal boundary value problem 8 2 d uðtÞ > > > dt2 þ AuðtÞ ¼ f ðtÞ ð0 < t < 1Þ; > > > > J X > > < uð0Þ ¼ u; uð1Þ ¼ aj uðkj Þ þ w; j¼1 > > > > J X > > > > jaj j 6 1 0 < k1 < k2 < < kJ < 1; > : ð1:1Þ j¼1 for the differential equation of elliptic type in a Hilbert space H with self-adjoint positive definite operator A with a closed domain DðAÞ H is considered. Here f(t) be a given abstract continuous function defined on ½0; 1 with values in H; u and w are elements of DðAÞ and kj are numbers from the set ½0; 1. It is known (see, for example, [3–7]) that various nonlocal boundary-value problems for elliptic equations can be reduced to the nonlocal boundary-value problem (1.1). The simply nonlocal boundary value problem was presented and investigated for the first time by Bitsadze and Samarskii [8]. Further in [9–27] the Bitsadze–Samarskii type nonlocal boundary value problems were investigated for the various differential and difference equations of elliptic type. Methods of the solutions of the abstract elliptic differential and difference equations have been studied extensively by many researchers (see [28–34] and the references therein). ⇑ Corresponding author. E-mail address: [email protected] (A. Ashyralyev). 0096-3003/$ - see front matter Ó 2012 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.amc.2012.07.016 1094 A. Ashyralyev, E. Ozturk / Applied Mathematics and Computation 219 (2012) 1093–1107 A function u(t) is called a solution of problem (1.1) if following conditions are satisfied: (i) u(t) is twice continuously differentiable on the segment ½0; 1. The derivatives at the endpoints of the segment are understood as the appropriate unilateral derivatives. (ii) The element u(t) belongs to DðAÞ for all t 2 ½0; 1 and the function AuðtÞ is continuous on the segment ½0; 1. (iii) u(t) satisfies the equation and nonlocal boundary conditions (1.1). A solution of problem (1.1) defined in this manner will from now on be referred to as a solution of problem (1.1) in the space Cð½0; 1; HÞ of all continuous functions uðtÞ defined on ½0; 1 with values in H, equipped with the norm kukCð½0;1;HÞ ¼ maxkuðtÞkH : 06t61 The well-posedness in Cð½0; 1; HÞ of the boundary value problem (1.1) means that coercive inequality h i ku00 kCð½0;1;HÞ þ kAukCð½0;1;HÞ 6 M kf kCð½0;1;HÞ þ kAukH þ kAwkH ð1:2Þ is true for its solution uðtÞ 2 Cð½0; 1; HÞ with some M, does not depend on u, w and f ðtÞ 2 Cð½0; 1; HÞ. In this paper, positive constants, which can differ in time (hence: not a subject of precision) will be indicated with an M. On the other hand Mða; b; . . .Þ is used to focus on the fact that the constant depends only on a; b; . . .. In fact, inequality (1.2) does not, generally speaking, hold in an arbitrary Hilbert space H and for the general unbounded self-adjoint positive definite operator A. Therefore, the problem (1.1) is not well-posed in Cð½0; 1; HÞ. The well-posedness of the boundary value problem (1.1) can be established if one considers this problem in certain spaces Fð½0; 1; HÞ of smooth Hvalued functions on ½0; 1. A function u(t) is said to be a solution of problem (1.1) in Fð½0; 1; HÞ if it is a solution of this problem in Cð½0; 1; HÞ and the functions u00 ðtÞ and AuðtÞ belong to Fð½0; 1; HÞ. As in the case of the space Cð½0; 1; HÞ, we say that the problem (1.1) is well-posed in Fð½0; 1; HÞ, if the following coercivity inequality is satisfied: h i ku00 kFð½0;1;HÞ þ kAukFð½0;1;HÞ 6 MðdÞ kf kFð½0;1;HÞ þ kAukH0 þ kAwkH0 ; where H0 H. Let us denote by C a0;1 ð½0; 1; HÞ, 0 < a < 1 the Banach spaces obtained by completion of the set of all smooth H-valued functions uðtÞ in the norm kukC a 0;1 ð½0;1;HÞ ¼ kukCð½0;1;HÞ þ sup kuðt þ sÞ uðtÞkH 0<t<tþs<1 sa ðt þ sÞa ð1 tÞa : In the present paper, the well-posedness of the nonlocal boundary-value problem (1.1) in C a0;1 ð½0; 1; HÞ spaces is established. The stable second order of accuracy difference scheme for the approximate solution of this problem is presented. The coercive inequalities for the solution of this difference scheme in difference analogue of C a0;1 ð½0; 1; HÞ spaces are established. For applications, the almost coercive stability and the coercive stability estimates for solutions of difference schemes for approximate solutions of nonlocal boundary value problems for elliptic equations are achieved. 2. The differential elliptic equation 1 1 In this section, we denote B ¼ A2 . Then it is clear that B is a self-adjoint positive definite operator and B P d2 I. We now state the following result, which will be called upon later in this work. Lemma 2.1. The following estimates hold [12]: kBa expðtBÞkH!H 6 t a kðI e2B Þ1 kH!H aa e 6 MðdÞ: ; 0 6 a 6 e; t > 0; Lemma 2.2. Let D¼ J X aj ðI e2B Þ1 ðeð1kj ÞB eð1þkj ÞB Þ: j¼1 Then the operator I D has an inverse T ¼ ðI DÞ1 ð2:1Þ ð2:2Þ 1095 A. Ashyralyev, E. Ozturk / Applied Mathematics and Computation 219 (2012) 1093–1107 and the following estimate is satisfied: kTkH!H 6 MðdÞ: ð2:3Þ Proof. The proof of estimate (2.3) is based on the estimate kI DkH!H P 1 J 1 X jaj jeð1kJ Þd2 : ð2:4Þ j¼1 Using the spectral representation of B, we get kðI e2B Þ1 ðeð1kj ÞB eð1þkj ÞB ÞkE!E 6 sup jð1 e2l Þ1 ðeð1kj Þl eð1þkj Þl Þj 1 d2 6l<1 1 1 6 eð1kj Þd2 sup ð1 e2l Þ1 ð1 e2kj l Þ 6 eð1kJ Þd2 : ð2:5Þ 1 d2 6 l<1 Then, applying the triangle inequality and estimate (2.5), we obtain (2.4). Thus, Lemma 2.2 is proved. h Now, we will obtain the formula for solution of problem (1.1). It is clear that (see [12]) the boundary value problem for the elliptic equation 2 d u dt 0 < t < 1; uð0Þ ¼ u; uð1Þ ¼ x þ AuðtÞ ¼ f ðtÞ; 2 ð2:6Þ has a unique solution Z 1 uðtÞ ¼ ðI e2B Þ1 ðetB eð2tÞB Þu þ ðeð1tÞB eð1þtÞB Þu1 ðeð1tÞB eð1þtÞB Þð2BÞ1 ðeð1sÞB eð1þsÞB Þf ðsÞ ds þ ð2BÞ1 Z 0 1 ðejtsjB eðtþsÞB Þf ðsÞ ds: ð2:7Þ 0 Using (2.7) and nonlocal boundary conditions, we get Z 1 J X x ¼ aj ðI e2B Þ1 ðekj B eð2kj ÞB Þu þ ðeð1kj ÞB eð1þkj ÞB Þx ðeð1kj ÞB eð1þkj ÞB Þð2BÞ1 ðeð1sÞB eð1þsÞB Þf ðsÞds 0 j¼1 þð2BÞ1 Z 1 ðejkj sjB eðkj þsÞB Þf ðsÞds þ w: 0 From Lemma 2.2 it follows that I e2B J X aj ðeð1kj ÞB eð1þkj ÞB Þ j¼1 has an inverse. Therefore " Z 1 J X 1 aj ðekj B eð2kj ÞB Þu ðeð1kj ÞB eð1þkj ÞB Þ B1 ðeð1sÞB eð1þsÞB Þf ðsÞ ds 2 0 j¼1 !)# Z kj Z 1 Z 1 eðkj sÞB f ðsÞ ds þ eðskj ÞB f ðsÞ ds eðkj þsÞB f ðsÞ ds : uð1Þ ¼ x ¼ P ðI e2B Þw þ 1 þðI e2B Þ B1 2 0 kj ð2:8Þ 0 Here, P¼ 2B Ie !1 J X ð1kj ÞB aj ðe ð1þkj ÞB e Þ : j¼1 Consequently, if the function f(t) continuously differentiable on ½0; 1, u; w 2 DðAÞ and formulas (2.7), (2.8) give a solution of the problem (1.1). Theorem 2.1. Suppose that u; w 2 DðAÞ. Then the boundary value problem (1.1) is well posed in a Hölder space C a0;1 ð½0; 1; HÞ and the following coercivity inequality holds: ku00 kC a 0;1 ð½0;1;HÞ þ kAukC a 0;1 ð½0;1;HÞ 6 MðdÞ½kAukH þ kAwkH þ Mðd; k1 ; kJ Þ kf kC a ð½0;1;HÞ : 0;1 að1 aÞ 1096 A. Ashyralyev, E. Ozturk / Applied Mathematics and Computation 219 (2012) 1093–1107 Proof. By [12], we have the following coercivity inequality ku00 kC a 0;1 ð½0;1;HÞ þ kAukC a 0;1 ð½0;1;HÞ 6 MðdÞ að1 aÞ kf kC a 0;1 ð½0;1;HÞ þ MðdÞ½kAuð0ÞkH þ kAuð1ÞkH ð2:9Þ for the solution of boundary value problem (2.6). Then, the proof of Theorem 2.1 is based on coercivity inequality (2.9) and on the estimate kAuð1ÞkH 6 Mðd; k1 ; kJ Þ kf kC a ð½0;1;HÞ þ MðdÞ½kAukH þ kAwkH : 0;1 að1 aÞ ð2:10Þ Therefore, we will prove (2.10). First, applying formula (2.8), we can write Auð1Þ ¼ P ðI e2B ÞAw þ J X 1 1 1 aj ðekj B eð2kj ÞB ÞAu eð1kj ÞB ðf ð1Þ f ðkj ÞÞ þ ekj B ðf ð0Þ f ðkj ÞÞ þ ðeð1þkj ÞB 2 2 2 j¼1 1 1 þeð2kj ÞB eð2þkj ÞB Þf ð1Þ þ ðeð2þkj ÞB þ eð3kj ÞB Þf ðkj Þ þ ðI eð1kj ÞB ÞðI þ eð1þkj ÞB Þf ð1Þ þ ðeð1þkj ÞB eð2kj ÞB 2 2 Z 1 Z 1 1 ð1þsÞB ð1sÞB ð3kj ÞB ð1kj ÞB ð1þkj ÞB 1 e Þf ð0Þ ðe e Þ Be ðf ðsÞ f ð1ÞÞ ds Be ðf ðsÞ f ð0ÞÞ ds 2 0 2 0 ! Z kj Z 1 Z 1 1 ðkj sÞB ðskj ÞB ðsþkj ÞB 2B Be ðf ðsÞ f ðkj ÞÞ ds þ Be ðf ðsÞ f ðkj ÞÞ ds Be ðf ðsÞ f ð0ÞÞds þ ðI e Þ 2 kj 0 0 þðI e2B Þðf ðkj Þ f ð1ÞÞ ¼ J 1 þ J 2 þ J 3 þ J 4 þ J 5 þ J 6 þ J 7 : Here J1 ¼ P ! J X aj ðekj B eð2kj ÞB ÞAu þ ðI e2B ÞAw ; j¼1 J X 1 aj e2B I þ ðI eð1kj ÞB ÞðI þ eð1þkj ÞB Þ ðeð1þkj ÞB þ eð1kj ÞB eð2kj ÞB þ eð2þkj ÞB Þ f ð1Þ 2 j¼1 1 1 þ I e2B þ ðeð1kj ÞB ekj B þ eð2þkj ÞB þ eð3kj ÞB Þ f ðkj Þ þ ðeð1þkj ÞB eð2kj ÞB eð3kj ÞB þ ekj B Þf ð0Þ ; 2 2 J2 ¼ P J 3 ¼ P J X aj ðeð1kj ÞB eð1þkj ÞB Þ j¼1 J 4 ¼ P J X ð1kj ÞB aj ðe e ð1þkj ÞB j¼1 J5 ¼ P J X 2B aj ðI e j¼1 J6 ¼ P J X aj ðI e2B Þ j¼1 J 7 ¼ P 1 Þ 2 J X aj ðI e 1 2 2B j¼1 Z 1 2 1 Þ 2 Z ! 1 Beð1sÞB ðf ðsÞ f ð1ÞÞ ds ; 0 Z ! 1 Be ð1þsÞB ðf ðsÞ f ð0ÞÞ ds ; 0 ! kj ðkj sÞB Be ðf ðsÞ f ðkj ÞÞ ds ; 0 Z 1 Þ 2 ! 1 Beðskj ÞB ðf ðsÞ f ðkj ÞÞ ds ; kj Z ! 1 Be ðsþkj ÞB ðf ðsÞ f ð0ÞÞ ds : 0 Let us estimate J k for k ¼ 1; . . . ; 7 separately. Using estimates (2.1) and (2.3), we obtain kJ 1 kH 6 kPkH!H J X ! jaj jkekj B eð2kj ÞB kH!H kAukH þ kI e2B kH!H kAwkH 6 M 1 ðdÞ½kAukH þ kAwkH : j¼1 Let us estimate J 2 . Using estimates (2.1), (2.3) and the definition of the norm of the space C a0;1 ð½0; 1; HÞ, we get kJ 2 kH 6 M 2 ðdÞmaxkf ðtÞkH 6 M 2 ðdÞkf kCa 06t61 0;1 ð½0;1;HÞ : Now we will estimate J 3 . Using estimates (2.1), (2.3) and the definition of the norm of space C a0;1 ð½0; 1; HÞ, we obtain 1097 A. Ashyralyev, E. Ozturk / Applied Mathematics and Computation 219 (2012) 1093–1107 kJ 3 kH 6 kPkH!H 6 Z 1 Z J X 1 1 ds jaj j kBðeð1kj ÞB eð1þkj B Þeð1sÞB kH!H kf ðsÞ f ð1ÞkH ds 6 M 3 ðdÞ kf kC a ð½0;1;HÞ 0;1 2 2 ðs þ kj Þ 0 0 j¼1 M 3 ðdÞ kf kC a ð½0;1;HÞ : 0;1 1 kJ In a similar manner, we can show that kJ 4 kH 6 kPkH!H Z Z 1 J X 1 1 ð1kj ÞB ds jaj j kB e eð1þkj ÞB eð1þsÞB kH!H kf ðsÞ f ð0ÞkH ds 6 M 4 ðdÞ kf kCa ð½0;1;HÞ 0;1 2 2 þ s kj 0 0 j¼1 6 M 4 ðdÞkf kC a 0;1 ð½0;1;HÞ : Using the estimates (2.1), (2.3) and the definition of the norm of space C a0;1 ð½0; 1; HÞ, we get kJ 5 kH 6 kPkH!H 6 M 5 ðdÞ Z kj J X 1 jaj j kI e2B kH!H kBeðkj sÞB kH!H kf ðsÞ f ðkj ÞkH ds 2 0 j¼1 Z J X jaj j 0 j¼1 6 M 5 ðdÞ kj 1 að1 kJ Þa ds a ð1 sÞ ðkj kf kC a 0;1 ð½0;1;HÞ sÞ1a kaj kf kC a 0;1 ð½0;1;HÞ 6 M 5 ðdÞ J X j¼1 jaj j ð1 kj Þa Z kj ds ðkj sÞ1a kaj 0 kf kC a 0;1 ð½0;1;HÞ : Now we will estimate J 6 . Using estimates (2.1), (2.3) and the definition of the norm of space C a0;1 ð½0; 1; HÞ, we get kJ 6 kH 6 kPkH!H 6 M 6 ðdÞ Z 1 J X 1 jaj j kI e2B kH!H kBeðskj ÞB kH!H dskf ðsÞ f ðkj ÞkH 2 kj j¼1 J X j¼1 6 jaj j ð1 kj Þa Z kj 1 ds ðs kj Þ1a sa kf kC a 0;1 ð½0;1;HÞ 6 M 6 ðdÞ J X j¼1 jaj j ð1 kj Þa kaj Z kj 1 ds ðs kj Þ1a kf kC a 0;1 ð½0;1;HÞ M 6 ðdÞ kf kC a ð½0;1;HÞ : 0;1 aka1 Finally, we will estimate J 7 . Using estimates (2.1), (2.3) and the definition of the norm of space C a0;1 ð½0; 1; HÞ, we obtain kJ 7 kH 6 kPkH!H 6 M 7 ðdÞ Z Z 1 J J X X 1 1 ds jaj j kBðI e2B Þeðsþkj ÞB kH!H dskf ðsÞ f ð0ÞkH 6 M 7 ðdÞ jaj j kf kC a ð½0;1;HÞ 0;1 2 0 0 s þ kj j¼1 j¼1 1 kf kC a ð½0;1;HÞ : 0;1 k1 Combining the estimates in H for J k , k ¼ 1; . . . ; 7, we get (2.10). Hence, Theorem 2.1 is proved. h Remark. We can obtain the well-posedness the boundary value problem (1.1) in a Hölder space C a0;1 ð½0; 1; EÞ with the positive operator A in an arbitrary Banach space E under the additional assumptions that the operator I J X aj ðI e2B Þ1 ðeð1kj ÞB eð1þkj ÞB Þ: j¼1 has an inverse T and the following estimate is satisfied: kTkE!E 6 MðdÞ: Now, we consider the applications of Theorem 2.1. First, we consider the nonlocal boundary value problem for the elliptic equation 8 utt ðaðxÞux Þx þ du ¼ f ðt; xÞ; 0 < t < 1; 0 < x < 1; > > > > J > X > > > uð0; xÞ ¼ u ðxÞ; uð1; xÞ ¼ aj uðkj ; xÞ þ wðxÞ; 0 6 x 6 1; > > < j¼1 J > X > > > jaj j 6 1; 0 < k1 < k2 < < kJ < 1; > > > > j¼1 > > : uðt; 0Þ ¼ uðt; 1Þ; ux ðt; 1Þ ¼ ux ðt; 0Þ; 0 6 t 6 1; ð2:11Þ 1098 A. Ashyralyev, E. Ozturk / Applied Mathematics and Computation 219 (2012) 1093–1107 where aðxÞ, uðxÞ, wðxÞ and f ðt; xÞ are given sufficiently smooth functions aðxÞ P a > 0, d ¼ const > 0. The problem has a unique solution uðt; xÞ. This allows us to reduce the nonlocal problem (2.11) to the nonlocal boundary value problem (1.1) in a Hilbert space H ¼ L2 ½0; 1 with a self-adjoint positive definite operator A defined by (2.11). Theorem 2.2. The solutions of nonlocal boundary value problem (2.11) satisfy the coercivity inequality kutt kC a 0;1 ð½0;1;L2 ½0;1Þ þ kukC a 2 0;1 ð½0;1;W 2 ½0;1Þ 6 Mðd; k1 ; kJ Þ kf kC a ð½0;1;L2 ½0;1Þ þ MðdÞ½kukW 2 ½0;1 þ kwkW 2 ½0;1 : 0;1 2 2 að1 aÞ The proof of Theorem 2.2 is based on the abstract Theorem 2.1 and the symmetry properties of the space operator A generated by problem (2.11). Second, let X be the unit open cube in Rn fx ¼ ðx1 ; . . . ; xn Þ : 0 < xk < 1; 1 6 k 6 ng with boundary S, X ¼ X [ S. In ½0; 1 X, the Dirichlet–Bitsadze–Samarskii type mixed boundary value problem for the multidimensional elliptic equation 8 n X > > > utt ðar ðxÞuxr Þxr ¼ f ðt; xÞ; x 2 X; 0 < t < 1; > > > r¼1 > > > > J > X > > < uð0; xÞ ¼ uðxÞ; uð1; xÞ ¼ aj uðkj ; xÞ þ wðxÞ; x 2 X; j¼1 > > > J > X > > > jaj j 6 1; 0 < k1 < < kJ < 1; > > > > j¼1 > > : uðt; xÞjxS ¼ 0; 0 6 t 6 1; ð2:12Þ x 2 X; is considered. The problem has unique smooth solution uðt; xÞ for the smooth ar ðxÞ P a > 0, f ðt; xÞ, uðxÞ and wðxÞ functions. This allows us to reduce the Dirichlet–Bitsadze–Samarskii type mixed boundary value problem to the nonlocal boundary problem in a Hilbert space H ¼ L2 ðXÞ of the all integrable functions defined on X, equipped with norm kf kL2 ðXÞ ¼ Z Z jf ðxÞj2 dx1 dxn 12 : x2X with a self adjoint positive definite operator A defined by (2.12). Theorem 2.3. The solutions of nonlocal boundary value problem (2.12), satisfy the coercivity inequality kutt kC a 0;1 ð½0;1;L2 ðXÞÞ þ kukCa 2 0;1 ð½0;1;W 2 ðXÞÞ 6 Mðk1 ; kJ Þ þ M½kukW 2 ðXÞ þ kwkW 2 ðXÞ : kf k a 2 2 að1 aÞ C0;1 ð½0;1;L2 ðXÞÞ The proof of Theorem 2.3 is based on the abstract Theorem 2.1, the symmetry properties of the space operator A generated by problem (2.12) and the following theorem on the coercivity inequality for the solution of the elliptic differential problem in L2 ðXÞ. Theorem 2.4. For the solutions of the elliptic differential problem n X ðar ðxÞuxr Þxr ¼ xðxÞ; x 2 X; uðxÞ ¼ 0; x 2 S; r¼1 the following coercivity inequality holds [12]: kukW 2 ðXÞ 6 MkxkL2 ðXÞ : 2 3. The second order of accuracy difference scheme Let us associate to the nonlocal boundary value problem (1.1) the corresponding difference problem 8 2 s ðukþ1 2uk þ uk1 Þ þ Auk ¼ uk ; uk ¼ f ðtk Þ; > > > > < tk ¼ ks; 1 6 k 6 N 1; Ns ¼ 1; u0 ¼ u; J X > > > aj u½ksr þ ðu½ksr u½ksr 1 Þ ksr ksr þ w: > : uN ¼ j¼1 It is a well-known that B ¼ 12 sA þ ð3:1Þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 4A þ s2 A2 is self-adjoint positive definite operator and R ¼ ðI þ sBÞ1 which is de- fined on the whole space H is a bounded operator. Here, I is the identity operator. 1099 A. Ashyralyev, E. Ozturk / Applied Mathematics and Computation 219 (2012) 1093–1107 Now, let us give some lemmas that will be needed below. Lemma 3.1. The following estimates hold [12]: 8 1 s > ksA2 > Rk kH#H 6 MðdÞ ; > ks < ke k kðI R2N Þ1 kH!H 6 MðdÞ; k P 1; k k kR kH#H 6 MðdÞð1 þ dsÞ ; kskBR kH#H 6 MðdÞ; k P 1; d > 0; > > a > : kBb ðRkþr Rk Þk 6 MðdÞ ðrsaÞþb ; 1 6 k < k þ r 6 N; 0 6 a; b 6 1: H#H ð3:2Þ ðksÞ Lemma 3.2. The following estimate holds [12]: N1 X skBRj kH!H 6 MðdÞYðs; BÞ; ð3:3Þ j¼1 where 1 Yðs; BÞ ¼ min ln ; 1 þ j ln kBkH!H j : s Here and in future. Lemma 3.3. The operator I R2N J X aj RN kj s kj kj kj kj kj ðR IÞ RN s þ RNþ s 1 þ RNþ s þ s j¼1 s has an inverse Ts ¼ I R2N kj kj kj kj !1 J X k k aj RN s þ RNþ s þ j j ðR IÞ RN s þ RNþ s 1 s j¼1 s and the following estimate is satisfied: kT s kH#H 6 Mðd; k1 ; kJ Þ: ð3:4Þ Proof. We have that 1 T s T ¼ T s T e2A2 R2N þ k 1 kj kj kj ! kj 1 J kj X j k k : aj e 1 s s A2 RN s eð1þ s sÞA2 þ RNþ s þ j j ðR IÞ RN½ s þ RNþ s 1 s j¼1 s ð3:5Þ Using estimate (3.2), we obtain k k k k k 1 k 1 J X kj kj 2A12 2N N sj Nþ sj N sj Nþ sj 1 ð1 sj sÞA2 ð1þ sj sÞA2 ðR I Þ R R þ aj e R þR e þ þR e s s j¼1 6 M 1 ðdÞs: H!H ð3:6Þ Estimate (3.4) is derived from formula (3.5) and estimate (3.6). Therefore, Lemma 3.3 is proved. h Theorem 3.1. For any uk , 1 6 k 6 N 1 the solution of the problem (3.1) exists and the following formula holds ( 2N 1 uk ¼ ðI R Þ k ðR R 2Nk Þu þ ðR Nk R Nþk Nk ÞuN ðR R Nþk N1 X ÞS ðRN1i RN1þi Þui s ) i¼1 N1 X þ S ðRjkij1 Rkþi1 Þui s for k ¼ 0; . . . ; N 1; ð3:7Þ i¼1 8 0 0 kj 1 ½ s 1 J N 1 N1 N1 < kj X X ½kj1i X X X kj kj kj kj kj uN ¼ T s @ a j R½ s R2N½ s u RN½ s RNþ½ s S ðRN1i RN1þi Þui s þ ðI R2N ÞS@ Rs ui s þ Ri½ s 1 ui s R½ s þi1 ui sA : kj i¼1 j¼1 i¼1 i¼1 i¼½ s 8 0 kj 199 1 ½ s 1 < kj N 1 N1 N1 == X X X X kj kj kj kj kj kj kj kj ½ s 1 2N½ s N½ s Nþ½ s 1 Ni1 Nþi1 2N @ ½ s 2i i½ s 1 ½ s þi2 þ S ðR þR uþ R þR R Þui s þ SðI R Þ R ui s R ui s R ui sA ðR IÞ R þ ðI R2N ÞwA: : ; ; s s kj i¼1 i¼1 i¼1 i¼½ s 1100 A. Ashyralyev, E. Ozturk / Applied Mathematics and Computation 219 (2012) 1093–1107 Here S ¼ ð2I þ sBÞ1 B1 : Proof. The second order of accuracy difference scheme ( 1 6 k 6 N 1; Ns ¼ 1; s12 ½ukþ1 2uk þ uk1 þ Auk ¼ uk ; u0 ¼ u; ð3:8Þ uN ¼ x has a solution and the following formula holds: ( 2N 1 uk ¼ ðI R Þ k ðR R 2Nk Þu þ ðR Nk R Nþk Nk Þx ðR R Nþk N 1 X ÞS ðRN1i RN1þi Þui s i¼1 ) þS N1 X ðRjkij1 i¼1 Rkþi1 Þui s: ð3:9Þ Applying formula (3.9) and the nonlocal boundary condition J X x¼ j¼1 aj u½kjs þ kj s kj u kj u½kj1 þ w; s ½s s we obtain uN ¼ ( J X 2N 1 aj ðI R Þ ( kj ½s R R kj 2N½ s ) N1 X kj kj kj kj N½ s Nþ½ s N½ s Nþ½ s N1i N1þi uþ R R x R R R Þui s S ðR i¼1 j¼1 0 kj 1 ½ s 1 N 1 N 1 X X X ½kj1i i½kj 1 ½kj þi1 s s s @ R usþ R us R u sA þS i i i¼1 kj i i¼1 kj i¼½ s n kj n kj kj kj kj ðI R2N Þ1 ðR IÞ R½ s 1 þ R2N½ s u RN½ s þ RNþ½ s 1 x 0 kj 1 ½ s 1 N1 N1 X X ½kji2 X kj kj kj N½ s Nþ½ s 1 Ni1 Nþi1 i½ 1 þ R þR R ui þ S@ R s ðR IÞui s R s ðR IÞui sA S s R þ s s i¼1 i¼1 kj i¼½ s þ1 ) Su½kj s þ SR1 u½kj s S s s N1 X kj R½ s þi2 ðR IÞui s þ w: i¼1 Since the operator I R2N J kj kj kj kj X k k aj RN½ s þ RNþ½ s þ j ðR IÞ RN½ s þ RNþ½ s 1 s j¼1 s has an inverse T s , it follows that uN ¼ x ¼ T s J X aj ( kj kj R½ s R2N½ s kj kj X N 1 ðRNi1 RNþi1 Þui s u RN½ s RNþ½ s S j¼1 i¼1 1 0k k ½ sj 1 k N1 N1 k kj kj X ½ j i1 X X j j kj kj 2N B i½ s 1 ½ s þi1 s ðR IÞ R½ s 1 þ R2N½ s u þðI R ÞS@ R ui s þ R ui s R u i sC Aþ i¼1 i¼1 k i¼½ sj s s 99 0k 1 1 > X ½ sj 1 k => = N 1 N 1 N 1 k kj kj kj X X X j j B C C þ ðI R2N ÞwA: R½ s 2i ui s Ri½ s 1 ui sA SðI R2N Þ R½ s þi2 ui s þ RN½ s þ RNþ½ s 1 S ðRNi1 RNþi1 Þui s þ ðI R2N ÞS@ > ;> ; kj i¼1 i¼1 i¼1 i¼½ s Hence, Theorem 3.1 is established. h with values in the Hilbert space H. The Banach spaces Let Fð½0; 1s ; HÞ be the linear space of mesh functions us ¼ fuk gN1 1 Cð½0; 1s ; HÞ and C a0;1 ð½0; 1s ; HÞ, 0 < a < 1 have the norms kus kCð½0;1s ;HÞ ¼ max kuk kH ; 16k6N1 kus kC a 0;1 ð½0;1s ;HÞ ¼ kus kCð½0;1s ;HÞ þ ððN kÞsÞa ððk þ rÞsÞa kukþr uk kH : ðrsÞa 16k6kþr6N1 sup 1101 A. Ashyralyev, E. Ozturk / Applied Mathematics and Computation 219 (2012) 1093–1107 Theorem 3.2. The solution of difference scheme (3.1) in Cð½0; 1s ; HÞ obey the almost coercive inequality 2 N1 þ fAuk gN1 fs ðukþ1 2uk þ uk1 Þg1 1 Cð½0;1s ;HÞ Cð½0;1s ;HÞ h i 6 Mðd; k1 ; kJ Þ Yðs; BÞkus kCð½0;1s ;HÞ þ kAukH þ kAwkH : ð3:10Þ Proof. By [12], 2 N1 fs ðukþ1 2uk þ uk1 Þg1 Cð½0;1s ;HÞ þ fAuk gN1 1 Cð½0;1s ;HÞ h i 6 MðdÞ Yðs; BÞkus kCð½0;1s ;HÞ þ kAukH þ kAuN kH ð3:11Þ for the solution of difference scheme (3.8). Then, the proof of (3.10) is based on (3.11) and on the estimate kAuN kH 6 MðdÞ Yðs; BÞkus kCð½0;1s ;HÞ þ kAukH þ kAwkH ð3:12Þ for the solution of difference scheme (3.1). Applying formula (3.7) and A ¼ B2 R, we get AuN ¼ J 1 þ J 2 ; ð3:13Þ where J1 ¼ T s ! k k J kj kj X j j k k R½ s 1 þ R2N½ s ðR IÞAu þ ðI R2N ÞAw ; aj R½ s R2N½ s Au þ j j s j¼1 s ð3:14Þ 8 0k 1 > X ½ sj 1 J N1 N1 N1 < kj kj kj kj kj X X X X B C B RN½ s þ RNþ½ s S B2 ðRNi RNþi Þui s þ SðI R2N Þ@ B2 R½ s i ui s þ B2 Ri½ s ui s B2 R½ s þi ui sA J 2 ¼ T s @ aj > : k i¼1 i¼1 j¼1 i¼1 j i¼½ s 8 0k 1 j > X ½ s 1 N1 N1 < kj kj kj kj X X kj kj B C ðR IÞ RN½ s þ RNþ½ s 1 S B2 ðRNi RNþi Þui s þ ðI R2N ÞS@ B2 R½ s 1i ui s B2 Ri½ s ui sA þ > s s : kj i¼1 i¼1 i¼½ s )! N1 kj X : ð3:15Þ ðI R2N ÞS B2 R½ s þi1 ui s 0 i¼1 To this end it suffices to show that kJ 1 kH 6 Mðd; k1 ; kJ Þ½kAukH þ kAwkH ð3:16Þ kJ 2 kH 6 Mðd; k1 ; kJ ÞYðs; BÞkus kCð½0;1s ;HÞ : ð3:17Þ and Estimate (3.16) follows from formula (3.14), and estimates (3.2) and (3.4). Using formula (3.15) and estimates (3.2), (3.4), we obtain ( J N1 X X N½kj Nþ½kj 1 R s s kð2I þ sBÞ kH!H kJ2 kH 6 kT s kH!H ja j j þ kBRNi kH!H þ kBRNþi kH!H kui kH s R H!H j¼1 0 ½s BX ½ksj iþ1 þkSkH!H @ BR kj i¼1 H!H H!H H!H kj i¼½ s þ1 " kj N½kj kj þ kR IkH!H R s s s H!H 0k j ½ s 1 kj BX BR½ s i þkR IkH!H kSkH!H @ i¼1 i¼1 N 1 X i½kj þ1 BR s kui kH s þ Nþ½kj 1 s þ R H!H N1 X ½kj þiþ1 BR s kui kH s þ kð2I þ sBÞ1 kH!H H!H N1 X i½kj þ1 BR s þ k i¼½ sj H!H i¼1 N1 X 1 C kui kH sA kBRNi kH!H þ kBRNþi kH!H kui kH s i¼1 N1 X ½kj þi BR s H!H i¼1 391 > = N1 X C 7 C i Akui kH s5 A 6 Mðd; k1 ;kJ Þ skBR kH!H kui kH : > ; H!H i¼1 1 From the last estimate and estimate (3.3) follows estimate (3.17). Thus, Theorem 3.2 is proved. h Theorem 3.3. The difference scheme (3.1) is well posed in Hölder spaces C a0;1 ð½0; 1s ; HÞ and the following coercivity inequality holds: 1102 A. Ashyralyev, E. Ozturk / Applied Mathematics and Computation 219 (2012) 1093–1107 2 N1 fs ðukþ1 2uk þ uk1 Þg1 C a0;1 ð½0;1s ;HÞ 6 Mðd; k1 ; kJ Þ 1 kus kCa ð½0;1s ;HÞ þ kAukH þ kAwkH : 0;1 að1 aÞ ð3:18Þ Proof. By [12], 2 N1 fs ðukþ1 2uk þ uk1 Þg1 C a0;1 ð½0;1s ;HÞ 6 MðdÞ 1 kus kC a ð½0;1s ;HÞ þ kAukH þ kAuN kH 0;1 að1 aÞ ð3:19Þ for the solution of difference scheme (3.8). Then, the proof of (3.18) is based on (3.19) and on the estimate kAuN kH 6 Mðd; k1 ; kJ Þ 1 kus kC a ð½0;1s ;HÞ þ kAukH þ kAwkH : 0;1 að1 aÞ Applying the triangle inequality, formula (3.13) and estimate (3.16), we get kAuN kH 6 kJ 1 kH þ kJ 2 kH 6 kJ 2 kH þ Mðd; k1 ; kJ Þ½kAukH þ kAwkH : To this end it suffices to show that kJ 2 kH 6 Mðd; k1 ; kJ Þ 1 að1 aÞ kus kC a 0;1 ð½0;1s ;HÞ ð3:20Þ : Applying formula (3.15), we get ( J X kj kj X N 1 aj RN½ s RNþ½ s S J2 ¼ T s k N½ sj R kj kj X N1 i¼1 j¼1 sB2 RNi ðui uN1 Þ þ RN½ s RNþ½ s S k R Nþ½ sj BS I R N1 sB2 RNþi1 ðui u1 Þ i¼1 uN1 R k N½ sj k R Nþ½ sj kj ½X s 1 kj N 2N1 2N u1 þ I R S sB2 R½ s i ui u kj BS R R ½s i¼1 N1 X kj þ I R2N S sB2 Ri½ s ui u kj ½s k i¼½ sj N1 X kj kj kj þ I R2N BS I R½ s 1 þ R1 RN½ s 2 u kj I R2N S sB2 R½ s þi ðui u1 Þ ½s ( kj kj kj ðR I Þ I R2N BSR½ s I RN1 u1 þ s s i¼1 X N1 kj kj RN½ s þ RNþ½ s 1 S B2 RNi ðui uN1 Þs i¼1 X N1 kj kj kj kj kj kj RN½ s þ RNþ½ s 1 S B2 RNþi ðui u1 Þs þ RN½ s þ RNþ½ s 1 BS I RN1 uN1 RN½ s þ RNþ½ s 1 BS RN R2N1 u1 i¼1 N1 N1 ½X X X s 1 kj kj kj þ I R2N S B2 R½ s 1i ui u kj s I R2N S B2 Ri1½ s ui u kj s I R2N S B2 R½ s þi1 ðui u1 Þs kj ½s i¼1 kj þ I R2N BSR½ s I RN1 u1 ½s k i¼½ sj i¼1 k X 6 kj j J z2 ; þ SBsR R½ s 1 RN½ s u kj ¼ ½s z¼1 where J 12 ¼ T s J kj kj kj kj kj X aj RN½ s RNþ½ s BS I RN1 uN1 RN½ s RNþ½ s BS RN R2N1 u1 I R2N BS I R½ s 1 þ R1 j¼1 kj kj kj kj kj kj ðR IÞ RN½ s þ RNþ½ s 1 BS I RN1 uN1 RN½ s 2 u kj I R2N BSR½ s I RN1 u1 þ s ½s s k kj kj kj kj j ; RN½ s þ RNþ½ s 1 BS RN R2N1 u1 þ I R2N BSR½ s I RN1 u1 þ I R2N SB2 s R½ s 1 RN½ s u kj ½s J 22 ¼ Ts J X kj Nþ½ s aj R R kj N½ s þ j¼1 J 32 ¼ T s J X kj kj aj RN½ s RNþ½ s j¼1 kj s kj s ! X N2 kj kj kj N½ s Nþ½ s 1 2 Ni ðR I Þ R S sB R ðui uN1 Þ ; R s i¼1 X N1 kj kj kj ðR IÞ RN½ s þ RNþ½ s 1 S sB2 RNþi ðui u1 Þ; s i¼2 1103 A. Ashyralyev, E. Ozturk / Applied Mathematics and Computation 219 (2012) 1093–1107 0 kj kj J ½X ½X s 1 s 1 kj kj X kj kj B 4 2N 2 ½ s i 2N IR J 2 ¼ T s aj @ I R S sB R ui u½kj þ sB2 ðR IÞR½ s 1i ui u½kj ; s s i¼1 j¼1 s s i¼1 0 0 J N1 N1 kj kj X X X k k B 2N B ¼ T s aj @ðI R ÞS@ sB2 Ri½ s ui u½kj þ j j ðI R2N ÞS sB2 ðI RÞR½ s 1i ui u½kj ; J 52 j¼1 s s k i¼½ sj þ1 s s k i¼½ sj þ1 ! X J N1 N1 X k k X kj kj 2N 2 ½ sj þi 2 ½ sj þi1 S sB R ¼ T s aj I R S sB R ðui u1 Þ ðR IÞðui u1 Þ : J 62 s i¼2 j¼1 Second let us estimate Jm 2 s i¼1 for any m ¼ 1; . . . ; 6 separately. We start with J 12 . Using estimates (3.2) and (3.4), we obtain kJ 12 kH 6 Mðd; k1 ; kJ Þ max kuk kH 6 Mðd; k1 ; kJ Þkus kC a 0;1 ð½0;1s ;HÞ 16k6N1 : For J 22 , applying (3.2), (3.4) and the definition of the norm of space C a0;1 ð½0; 1s ; HÞ, we obtain kJ 22 kH 6 kT s kH!H J X j¼1 2I þ sBÞ1 kH!H Nþ½kj s jaj j R N½kj s þ R H!H H!H ! N2 X skBRNi kH!H kui uN1 kH i¼1 kj N½kj kj R s kR Ik þ H!H s s H!H Nþ½kj 1 s þ R k H!H N2 Mðd; k1 ; kJ Þ X ððN 1Þs isÞa 6 s kus kCa ð½0;1s ;HÞ : a 0;1 ðN 1 iÞs ððN 1ÞsÞ i¼1 The sum N2 X ððN 1Þs isÞa s ðN 1 iÞs i¼1 is the lower Darboux integral sum for the integral Z 1 0 ds ð1 sÞ1a : It follows that 1 kJ 22 kH 6 M 1 ðd; k1 ; kJ Þ kus kC a a 0;1 ð½0;1s ;HÞ : Now let us estimate J 32 . Using estimates (3.2) and (3.4) and the definition of the norm of space C a0;1 ð½0; 1s ; HÞ, we get kJ 32 kH 6 kT s kH!H J X j¼1 þsBÞ1 kH!H N½kj s jaj j R H!H Nþ½kj s þ R H!H N½kj kj kj R s kR Ik þ H!H s ! N1 X skBRNþi kH!H kui u1 kH 6 Mðd; k1 ; kJ Þ s H!H N 1 X s i¼2 ðN þ iÞs i¼2 kus kC a 0;1 ð½0;1s ;HÞ Nþ½kj 1 s þ R kð2I H!H 6 M 1 ðd; k1 ; kJ Þkus kCa 0;1 ð½0;1s ;HÞ : Let us estimate J 42 . Using estimates (3.2), and (3.4) and the definition of the norm space C a0;1 ð½0; 1s ; HÞ, we obtain kJ 42 kH 6 kT s kH!H J X j¼1 kj X kj þ sBR½ s i ½ s 1 i¼1 H!H 0 kj ½X 1 kj kj s Bkj BðR IÞR½ s 1i jaj jkI R kH!H kð2I þ sBÞ kH!H @ s 2N 1 s s 1 kj ½ s 1 J X X C u u kj A 6 Mðd; k1 ; kJ Þ jaj j h i i ½ s kj H j¼1 i¼1 s H!H i¼1 s h i kj s s is s is s H a h i a kj u u kj i ½ s a s ð1 isÞ kus kC a 0;1 ð½0;1s ;HÞ : By the lower Darboux integral sum for the integral, it follows kJ 42 kH 6 Z J X M 1 ðd; k1 ; kJ Þ kj ds M 2 ðd; k1 ; kJ Þ s M 3 ðd; k1 ; kJ Þ s jaj j a kus kC a ð½0;1s ;HÞ 6 ku kCa ð½0;1s ;HÞ : a a ku kC a0;1 ð½0;1s ;HÞ 6 1a 0;1 0;1 a ð1 k Þ a ð1 k Þ k ðk sÞ 0 J j j j j¼1 Let us estimate J 52 . Estimates (3.2) and (3.4) and the definition of the norm space C a0;1 ð½0; 1s ; HÞ give 1104 A. Ashyralyev, E. Ozturk / Applied Mathematics and Computation 219 (2012) 1093–1107 0 1 h i a kj N1 s i s s X s B C kJ 52 kH 6 Mðd; k1 ; kJ Þm jaj j@ h i h i a Akus kCa0;1 ð½0;1s ;HÞ : kj kj a j¼1 kj i s s ð s Þ 1 s i s s i¼½ J s The sum enclosed in the right-hand side brackets is the lower Darboux integral sum for the integral Z 1 ds ð1 kj Þa sa ðs kj Þ1a kj : Since Z kj 1 ds ð1 kj Þa sa ðs k Þ1a 6 j 1 að1 kj Þa ; it follows that kJ 52 kH 6 M 1 ðd; k1 ; kJ Þ 1 að1 kJ Þa kus kC a 0;1 ð½0;1s ;HÞ : Lastly, let us estimate J 62 . In exactly the same manner using estimates (3.2) and (3.4) and the definition of the norm space C a0;1 ð½0; 1s ; HÞ, we obtain J X kj þ 1 s kJ 62 kH 6 Mðd; k1 ; kJ Þ jaj j ln ku kC a ð½0;1s ;HÞ : 0;1 kj j¼1 It follows that kJ 62 kH 6 M 1 ðd; k1 ; kJ Þkus kC a 0;1 ð½0;1s ;HÞ : Combining estimates in H for J m h 2 , m ¼ 1; . . . ; 6, we get estimate (3.20). Thus, Theorem 3.3 is proved. Now we consider applications of Theorems 3.1–3.3. First, the nonlocal boundary value problem (2.11) for the elliptic equation is considered. The discretization of problem (2.11) is carried out in two steps. In the first step, let us define the grid space ½0; 1h ¼ fx : xr ¼ rh; 0 6 r 6 K; Kh ¼ 1g: We introduce the Hilbert space L2h ¼ L2 ð½0; 1h Þ of the grid functions uh ðxÞ ¼ fur g1K1 defined on ½0; 1h , equipped with the norm kuh kL2h ¼ !1=2 K1 X juh ðxÞj2 h : r¼1 To the differential operator A generated by the problem (2.11), we assign the difference operator Axh by the formula Axh uh ðxÞ ¼ fðaðxÞuxx;r þ dur gM1 ; 1 h ð3:21Þ M1 r g1 acting in the space of grid functions u ðxÞ ¼ fu satisfying the conditions u0 ¼ uM , u1 u0 ¼ uM uM1 . With the help of Axh we arrive at the nonlocal boundary value problem 8 2 x h d uh ðt;xÞ h > > > dt2 þ Ah u ðt; xÞ ¼ f ðt; xÞ; 0 < t < 1; x 2 ½0; 1h ; < J X > > uh ð0; xÞ ¼ uh ðxÞ; uh ð1; xÞ ¼ aj uh ðkj ; xÞ þ wh ðxÞ; x 2 ½0; 1h > : ð3:22Þ j¼1 for an infinite system of ordinary differential equations. In the second step, we replace problem (3.22) by the difference scheme below 8 h u ðxÞ2uhk ðxÞþuhk1 ðxÞ > kþ1 þ Axh uhk ðxÞ ¼ fkh ðxÞ; fkh ðxÞ ¼ f ðt k ; xÞ; > > s2 > > < t ¼ ks; 1 6 k 6 N 1; Ns ¼ 1; uh0 ðxÞ ¼ uh ðxÞ; x 2 ½0; 1h : k J > h i X > > h > aj uh½kj =s ðxÞ þ ksj ksj uh½kj =sþ1 ðxÞ uh½kj =s ðxÞ þ wh ðxÞ; > : uN ðxÞ ¼ ð3:23Þ x 2 ½0; 1h : j¼1 Theorem 3.4. Let s and jhj be sufficiently small positive numbers. The solutions of the difference scheme (3.23) satisfy the following almost coercivity estimates: A. Ashyralyev, E. Ozturk / Applied Mathematics and Computation 219 (2012) 1093–1107 max ks2 uhkþ1 2uhk þ uhk1 kL2h þ max kuhk kW 2 6 Mðd; k1 ; kJ Þ ln 16k6N1 16k6N1 2h 1105 1 max kfkh kL2h þ kuh kW 2 þ kwh kW 2 : 2h 2h s þ jhj 16k6N1 The proof of Theorem 3.4 is based on the abstract Theorem 3.2 on the estimate 1 min ln ; 1 þ j ln kBxh kH!H j 6 M ln s 1 s þ jhj and the symmetry properties of the difference operator Axh defined by (3.21) in L2h . Theorem 3.5. Let s and jhj be sufficiently small positive numbers. The solutions of the difference scheme (3.23) satisfy the following coercivity estimates: N1 k s2 uhkþ1 2uhk þ uhk1 1 kC a 0;1 ð½0;1s ;L2h Þ 6 Mðd; k1 ; kJ Þ kuh kW 2 þ kwh kW 2 þ 2h 2h N1 þ uhk 1 C a0;1 ð½0;1s ;W 22h Þ N1 1 k fkh 1 kCa ð½0;1 L2h Þ : s 0;1 að1 aÞ The proof of Theorem 3.5 is based on the abstract Theorem 3.3 and the symmetry properties of the difference operator Axh defined by (3.21) in L2h . Second, the nonlocal boundary value problem (2.11) is considered. The discretization of problem (3.22) is carried out in two steps. In the first step, let us define the grid sets e h ¼ fx ¼ xm ¼ ðh1 m1 ; . . . ; hm mm Þ; m ¼ ðm1 ; . . . ; mm Þ; 0 6 ms 6 Ns ; hs Ns ¼ 1; j ¼ 1; . . . ; mg; X e h \ X; Xh ¼ X e h \ S: Sh ¼ X e h Þ of the grid functions uh ðxÞ ¼ fuðh1 m1 ; . . . ; hh mm Þg defined on X e k , equipped with We introduce the Hilbert space L2h ¼ L2 ð X the norm 0 h ku kL2h ¼@ X 11=2 2 ju ðxÞj h1 hn A h : x2Xh To the differential operator A generated by problem (3.22), we assign the difference operator Axh by the formula n X Axh uh ¼ ar ðxÞuhxr x ;j r¼1 ð3:24Þ r r acting in the space of grid functions uh ðxÞ, satisfying the conditions uh ðxÞ ¼ 0 for all x 2 Sh . It is known that Axh is a self-adjoint positive definite operator in L2h . With the help of Axh we arrive at the nonlocal boundary value problem for an infinite system of ordinary differential equations 8 2 h > e h; > d udtðt;xÞ þ Axh uh ðt; xÞ ¼ f h ðt; xÞ; 0 6 t 6 1; x 2 X 2 > < J X > > uh ð0; xÞ ¼ uh ðxÞ; uh ð1; xÞ ¼ aj uh kj ; x þ wh ðxÞ; > : ð3:25Þ e h: x2X j¼1 In the second step, (3.25) is replaced by the difference scheme below, we get second order of accuracy difference scheme 8 uh ðxÞ2uh ðxÞþuh ðxÞ k k1 > kþ1 þ Axh uhk ðxÞ ¼ fkh ðxÞ; fkh ðxÞ ¼ f h ðt k ; xÞ; x 2 Xh ; > s2 > > > < e h; uh0 ðxÞ ¼ uh ðxÞ; x 2 X t k ¼ ks; 1 6 k 6 N 1; Ns ¼ 1; J > h i X > > h > > aj uh½kj =s ðxÞ þ ksj ksj uh½kj =sþ1 ðxÞ uh½kj =s ðxÞ þ wh ðxÞ; : uN ðxÞ ¼ ð3:26Þ e h: x2X j¼1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 2 Theorem 3.6. Let s and jhj ¼ h1 þ þ hn be sufficiently small positive numbers. The solutions of the difference scheme (3.26) satisfy the following almost coercivity estimates: max ks2 uhkþ1 2uhk þ uhk1 kL2h þ max kuhk kW 2 6 Mðd; k1 ; kJ Þ ln 16k6N1 16k6N1 2h 1 max kfkh kL2h þ kuh kW 2 þ kwh kW 2 : 2h 2h s þ jhj 16k6N1 The proof of Theorem 3.6 is based on the abstract Theorem 3.2 on the estimate 1106 A. Ashyralyev, E. Ozturk / Applied Mathematics and Computation 219 (2012) 1093–1107 1 min ln ; 1 þ j ln kBxh kH!H j 6 M ln s 1 s þ jhj and the symmetry properties of the difference operator Axh defined by (3.24) in L2h and the following theorem on the coercivity inequality for the solution of the elliptic difference problem in L2h . Theorem 3.7. For the solutions of the elliptic difference problem Axh uh ðxÞ ¼ xh ðxÞ; uh ðxÞ ¼ 0; x 2 Xh ; ð3:27Þ x 2 Sh the following coercivity inequality holds [34]: kuh kW 2 6 MðdÞkxh kL2h : 2h Theorem 3.8. Let s and jhj be a sufficiently small positive numbers. Then the solution of difference scheme (3.26) satisfy the following coercivity stability estimate: N1 N1 k s2 uhkþ1 2uhk þ uhk1 1 kC a ð½0;1 ;L2h Þ þ k uhk 1 kCa ð½0;1 ;W 2 Þ s s 2h 0;1 0;1 1 N1 k fkh 1 kCa ð½0;1s L2h Þ : 6 Mðd; k1 ; kJ Þ kuh kW 2 þ kwh kW 2 þ 0;1 2h 2h að1 aÞ The proof of Theorem 3.8 is based on the abstract Theorem 3.3 and the symmetry properties of the difference operator Axh defined by the formula (3.24) and on Theorem 3.7 on the coercivity inequality for the solution of the elliptic difference Eq. (3.1) in L2h . Acknowledgement The authors would like to thank Prof. P.E. Sobolevskii for his helpful suggestions to the improvement of this paper. References [1] O.A. Ladyzhenskaya, N.N. Ural’tseva, Linear and Quasilinear Equations of Elliptic Type, Nauka, Moscow, 1973 (Russian). [2] M.L. Vishik, A.D. Myshkis, O.A. Oleinik, Partial differential equations, in: Mathematics in USSR in the Last 40 Years, 1917–1957, Fizmatgiz, Moscow, 1959, pp. 563–599 (Russian). [3] P. Grisvard, Elliptic Problems in Nonsmooth Domains, Pitman Advanced Publishing Program, London, 1985, pp.1986–2109. [4] S. Agmon, Lectures on Elliptic Boundary Value Problems, D. Van Nostrand, Princeton, New Jersey, 1965. [5] S.G. Krein, Linear Differential Equations in Banach Space, Nauka, Moscow, 1966 (Russian). [6] A.L. Skubachevskii, Elliptic functional differential equations and applications, in: Operator Theory – Advances and Applications, Birkhauser Verlag, Basel, Boston, Berlin, 1997. [7] V.L. Gorbachuk, M.L. Gorbachuk, Boundary Value Problems for Differential-Operator Equations, Naukova Dumka, Kiev, 1984 (Russian). [8] A.V. Bitsadze, A.A. Samarskii, On some simplest generalizations of linear elliptic problems, Doklady Akademii Nauk SSSR 185 (1969) 69–74. [9] P.E. Sobolevskii, The coercive solvability of difference equations, Doklady Akademii Nauk SSSR 201 (1971) 1063–1066 (Russian). [10] P.E. Sobolevskii, Well-posedness of difference elliptic equations, Discrete Dynamics in Nature and Society 1 (1997) 219–231. c [11] A. Ashyralyev, Well-posedness of the difference schemes for elliptic equations in C b; s ðEÞ spaces, Applied Mathematics Letter 22 (2009) 390–395. [12] A. Ashyralyev, P.E. Sobolevskii, New Difference Schemes for Partial Differential Equations, Birkhäuser Verlag, Basel, Boston, Berlin, 2004. [13] S. Agmon, S.A. Douglis, L. Nirenberg, Estimates near the boundary for solutions of elliptic partial differential equations satisfying general boundary conditions II, Communications on Pure and Applied Mathematics 17 (1964) 35–92. [14] A. Ashyralyev, Nonlocal boundary value problems for partial differential equations: well-posedness, AIP Conference Proceedings Global Analysis and Applied Mathematics: International Workshop on Global Analysis 729 (2004) 325–331. [15] A. Ashyralyev, P.E. Sobolevskii, Well-posedness of the difference schemes of the high order of accuracy for elliptic equations, Discrete Dynamics in Nature and Society 2006 (2006) 1–12. [16] A. Aibeche, A. Favini, Coerciveness estimate for Ventcel boundary value problem for a differential equation, Semigroup Forum 70 (2005) 269–277. [17] R. Agarwal, M. Bohner, V.B. Shakhmurov, Maximal regular boundary value problems in Banach-valued weighted spaces, Boundary Value Problems 1 (2005) 9–42. [18] Yu. A. Smirnitskii, P.E. Sobolevskii, Positivity of multidimensional difference operators in the C-norm, Uspekhi Matematicheskikh Nauk 36 (1981) 202– 203 (Russian). [19] A. Ashyralyev, C. Cuevas, S. Piskarev, On well-posedness of difference schemes for abstract elliptic problems in Lp ð½0; 1; EÞ spaces, Numerical Functional Analysis and Optimization 29 (2008) 43–65. [20] D. Orlovsky, S. Piskarev, On approximation of inverse problems for abstract elliptic problems, Journal of Inverse and Ill-Posed Problems 17 (2009) 765– 782. [21] A. Ashyralyev, On well-posedness of the nonlocal boundary value problem for elliptic equations, Numerical Functional Analysis and Optimization 24 (2003) 1–15. [22] G. Berikelashvili, On a nonlocal boundary value problem for a two-dimensional elliptic equation, Computational Methods in Applied Mathematics 3 (2003) 35–44. [23] D.G. Gordeziani, On a method of resolution of Bitsadze–Samarskii boundary value problem, Abstracts of Reports of Institute of Applied Mathematics, Tbilisi State University 2 (1970) 38–40. [24] D.G. Gordeziani, On Methods of Resolution of a Class of Nonlocal Boundary Value Problems, Tbilisi University Press, Tbilisi, 1981. [25] D.V. Kapanadze, On the Bitsadze–Samarskii nonlocal boundary value problem, Journal of Differential Equations 23 (1987) 543–545. [26] A.L. Skubachevskii, Nonlocal elliptic problems and multidimensional diffusion processes, Journal of Mathematical Physics 3 (1995) 327–360. A. Ashyralyev, E. Ozturk / Applied Mathematics and Computation 219 (2012) 1093–1107 1107 [27] V.A. Il’in, E.I. Moiseev, Two-dimensional nonlocal boundary value problems for Poisson’s operator in differential and difference variants, Mathematical Modelling 2 (1990) 139–159. [28] M.P. Sapagovas, A difference method of increased order of accuracy for the Poisson equation with nonlocal conditions, Differentsial’nye Uravneniya 44 (2008) 988–998. [29] A. Ashyralyev, A note on the Bitsadze–Samarskii type nonlocal boundary value problem in a Banach space, Journal of Mathematical Analysis and Applications 344 (2008) 557–573. [30] A.P. Soldatov, Problem of Bitsadze–Samarskii type for second order elliptic systems on the plane, Doklady Akademii Nauk 410 (2006) 607–611 (Russian). [31] A.A. Kilbas, O.A. Repin, An analog of the Bitsadze–Samarskii problem for a mixed type equation with a fractional derivative, Journal of Differential Equations 39 (2003) 674–680. [32] I.A. Gurbanov, A.A. Dosiev, On the numerical solution of nonlocal boundary problems for quasilinear elliptic equations, in: Approximate Methods for Operator Equations, Azerb. Gos. Univ., Baku, 1984, pp. 64–74. [33] A. Ashyralyev, E. Ozturk, Numerical solutions of Bitsadze–Samarskii problem for elliptic equation, Proceeding of 6-th International ISAAC Congress 10 (2007) 89–99. [34] P.E. Sobolevskii, Difference Methods for the Approximate Solution of Differential Equations, Izdat. Voronezh. Gosud. Univ., Voronezh, 1975 (Russian).
© Copyright 2024