Georgian Mathematical Journal Volume 10 (2003), Number 3, 427–465 SOLVABILITY AND ASYMPTOTICS OF SOLUTIONS OF CRACK-TYPE BOUNDARY-CONTACT PROBLEMS OF THE COUPLE-STRESS ELASTICITY O. CHKADUA Abstract. Spatial boundary value problems of statics of couple-stress elasticity for anisotropic homogeneous media (with contact on a part of the boundary) with an open crack are studied supposing that one medium has a smooth boundary and the other one has an open crack. Using the method of the potential theory and the theory of pseudodifferential equations on manifolds with boundary, the existence and uniqueness theorems are proved in Besov and Bessel-potential spaces. The smoothness and a complete asymptotics of solutions near the contact boundaries and near crack edge are studied. Properties of exponents of the first terms of the asymptotic expansion of solutions are established. Classes of isotropic, transversally-isotropic and anisotropic bodies are found, where oscillation vanishes. 2000 Mathematics Subject Classification: 74B05, 74G70, 35B40, 47G30. Key words and phrases: Couple-stress elasticity, anisotropic homogeneous medium, asymptotic expansion, strongly elliptic pseudodifferential equations, potentials. Introduction The paper is dedicated to the study of solvability and asymptotics of solutions of spatial crack type boundary-contact problems of statics of couple-stress elasticity for anisotropic homogeneous media with contact on a part of the boundary. A vast number of works are devoted to the justification and axiomatization of elasticity and couple-stress elasticity. The fundamentals of the theory of couple-stress elasticity are included in the works by W. Voight [40], E. Cosserat, F. Cosserat [11], and developed later in the works by E. Aero and E. Kuvshinski [1], G. Grioli [21], R. Mindlin [31], W. Koiter [24], W. Nowacki [34], V. Kupradze, T. Gegelia, M. Basheleishvili, T. Burchuladze [26], T. Burchuladze and T. Gegelia [4] and others. It is well-known that solutions of elliptic boundary value problems in domains with corners, edges and conical points have singularities regardless the smoothness properties of given data. Among theoretical investigations the methods suggested and developed by V. Kondrat’ev [25], V, Maz’ya [27], V, Maz’ya and B. Plamenevsky [28]–[30], S. Nazarov and B. Plamenevsky [33], M. Dauge [13], P. Grisvard [22] and others c Heldermann Verlag www.heldermann.de ISSN 1072-947X / $8.00 / ° 428 O. CHKADUA attracted attention of many scientists. They used the Mellin transform which allows them to reduce the problem to the investigation of spectral properties of ordinary differential operators depending on the parameter. The method of the potential theory and the theory of pseudodifferential equations used in this paper makes it possible to obtain more precise asymptotic representations of solutions of the problems posed, which frequently have crucial importance in applications (e.g., in crack extension problems). For the development of this method see [18], [3], [10], [9] and other papers. The method of the potential theory was successfully applied to the classical problems of elasticity and couple-stress elasticity theory by V. Kupradze and his disciples. In the present paper we consider the contact of two media, one of which has a smooth boundary, while the other has a boundary containing a closed cuspidal edge (the corresponding dihedral angle is equal to 2π), i.e., an open crack. Theorems on the existence and uniqueness of solutions of these boundarycontact problems are obtained using the potential theory and the general theory of pseudodifferential equations on a manifold with boundary. Using the asymptotic expansion of solutions of strongly elliptic pseudodifferential equations obtained in [10] (see also [18], [3]) and also the asymptotic expansion of potential-type functions [9], we obtain a complete asymptotic expansion of solutions of boundary-contact problems near the contact boundaries and near the crack edge. Here it is worth noticing the effective formulae for calculating the exponent of the first terms of asymptotic expansion of solutions of these problems by means of the symbol of the corresponding boundary pseudodifferential equations. The properties of exponents of the first terms of the asymptotic expansion of solutions are established. Important classes of isotropic, transversally-isotropic and anisotropic bodies are found, where oscillation vanishes. These results are new even for the problems of elasticity. 1. Formulation of the Problems Let D1 be a finite domain, D2 be a domain that can be both finite or infinite in the Euclidean space R3 with compact boundaries ∂D1 , ∂D2 (∂D1 ∈ C ∞ ), and let there exist a surface S0 of the class C ∞ of dimension two, which divides (1) (2) (1) the domain D2 into two subdomains D2 and D2 with C ∞ boundaries ∂D2 (1) (2) (2) (1) (2) and ∂D2 (D2 ∩ D2 = ∅, D2 ∩ D2 = S 0 ). Then ∂S0 is the boundary of the surface S0 (∂S0 ⊂ ∂D2 ), representing one-dimensional closed cuspidal edge, where ∂S0 is the crack edge. Let the domains D1 and D2 have the contact on the two-dimensional mani(1) (2) (1) (2) folds S 0 and S 0 of the class C ∞ , i.e., ∂D1 ∩ ∂D2 = S 0 ∪ S 0 , D1 ∩ D2 = ∅, (1) (2) (1) (2) (1) (1) (1) S 0 ∩ S 0 = ∅, and S1 = ∂D1 \ (S 0 ∪ S 0 ). Then ∂D2 = S2 ∪ S 0 ∪ S 0 , (2) (2) (2) ∂D2 = S2 ∪ S 0 ∪ S 0 . Suppose that the domains Dq , q = 1, 2, are filled with anisotropic homogeneous elastic materials. SOLVABILITY AND ASYMPTOTICS OF SOLUTIONS 429 The basic static equations of couple-stress elasticity for anisotropic homogeneous media are written in terms of displacement and rotation components as (see [5], [19]) M(q) (∂x )U (q) + F (q) = 0 in Dq , q = 1, 2, (1.1) (q) (q) (q) where U (q) = (u(q) , ω (q) ), u(q) = (u1 , u2 , u3 ) is the displacement vector, ω (q) = (q) (q) (q) (q) (q) (ω1 , ω2 , ω3 ) is the rotation vector, F (q) = (F1 , . . . , F6 ) is the mass force applied to Dq , and M(q) (∂x ) is the matrix differential operator 1 2 (q) (q) M (∂x ) M (∂x ) , M(q) (∂x ) = 3 4 M (q) (∂x ) M (q) (∂x ) 6×6 l M (q) (∂x ) = k 1 (q) l (q) M jk (∂x )k3×3 , 2 (q) (q) l = 1, 4, q = 1, 2, (q) (q) M jk (∂x ) = bijlk ∂i ∂l − εlrk aijlr ∂i , M jk (∂x ) = aijlk ∂i ∂l , 3 (q) (q) (q) (1.2) M jk (∂x ) = blkij ∂i ∂l + εirj airlk ∂l , 4 (q) (q) (q) (q) (q) M jk (∂x ) = cijlk ∂i ∂l − blrij εlrk ∂i + εirj birlk ∂l − εipj εlrk aiplr ; (q) (q) (q) εikj is the Levi-Civita symbol, aijlk , bijlk , cijlk are the elastic constants satisfying the conditions (q) (q) (q) (q) aijlk = alkij , cijlk = clkij , q = 1, 2. In (1.2) and in what follows, under the repeated indices we understand the summation from 1 to 3. It is assumed that the quadratic forms (q) (q) (q) aijlk ξij ξlk + 2bijlk ξij ηlk + cijlk ηij ηlk , q = 1, 2, with respect to variables ξij , ηij are positive-definite, i.e., ∃M > 0 (q) (q) (q) aijlk ξij ξlk+2bijlk ξij ηlk+cijlk ηij ηlk ≥ M (ξij ξij+ηlk ηlk ) for all ξij , ηlk , q = 1, 2. (1.3) We introduce the differential stress operator 1 2 (q) (q) N (∂z , n(z)) N (∂z , n(z)) N (q) (∂z , n(z)) = 3 4 N (q) (∂z , n(z)) N (q) (∂z , n(z)) l N (q) (∂z , n(z)) = k l (q) N jk (∂z , n(z))k3×3 , , 6×6 l = 1, 4, q = 1, 2, 2 (q) (q) (q) 4 (q) (q) (q) 1 (q) (q) N jk (∂z , n(z)) = bijlk ni (z)∂l − aijlk εlrk ni (z), 3 (q) (q) N jk (∂z , n(z)) = cijlk ni (z)∂l − blrij εlrk ni (z), N jk (∂z , n(z)) = aijlk ni (z)∂l , N jk (∂z , n(z)) = blkij ni (z)∂l , where n(z) = (n1 (z), n2 (z), n3 (z)) is the unit normal of the manifold ∂D1 at a point z ∈ ∂D1 (external with respect to D1 ) and a point z ∈ ∂D2 (internal with respect to D2 ). 430 O. CHKADUA In what follows the stress operators are denoted by N (q) = N (q) (∂z , n(z)), q = 1, 2. ◦ Let M(q) (ξ) be the symbol of the differential operator M(q) (∂x ) and M (q) (ξ) be the principal homogeneous symbol of the differential operator M(q) (∂x ). We introduce the following notation for Besov and Bessel potential spaces (see [39]): s s s s es = B es = H ep,r ep,r e ps × H e ps . Bsp,r = Bp,r × Bp,r , B ×B , Hsp = Hps × Hps , H p,r p (q) (q) From the symmetry of the coefficients aijlk , cijlk and the positive-definiteness of the quadratic forms (1.3) it follows (see [19]) that the operators M(q) (∂x ), q = 1, 2, are strongly elliptic, formally self-adjoint differential operators and therefore for any real vector ξ ∈ R3 and any complex vector η ∈ C6 the relations ◦ ◦ (q) Re(M (q) (ξ)η, η) = (M (q) (ξ)η, η) ≥ P0 |ξ|2 |η|2 (q) are valid, where P0 ◦ = const > 0 depends only on the elastic constants. Thus the matrices M (q) (ξ) are positive-definite for ξ ∈ R3 \ {0}. Taking into account the property of the Levi-Civita symbol δil δir δik εipj εlrk = det δpl δpr δpk (δpl is the Kronecker symbol), δjl δjr δjk it is not difficult to observe that ◦ (M(q) (ξ)η, η) = (M (q) (ξ)η, η), q = 1, 2. Then we obtain that the matrices M(q) (ξ), q = 1, 2, are positive-definite for ξ ∈ R3 \ {0}. Since detM(q) (ξ) 6= 0 for ξ 6= 0. (1) 1 Let U (1) ∈ Wp1 (D1 ), U (2) ∈ Wp,loc (D2 ). Then r1 U (2) = rD(1) U (2) ∈ Wp1 (D2 ) (2) 2 1 and r2 U (2) = rD(2) U (2) ∈ Wp,loc (D2 ), where ri is the restriction operator on (i) 2 D2 , i = 1, 2. From the theorem on traces (see [39]) it follows that the trace of (i) the functions U (i) , ri U (2) , i = 1, 2, exists on ∂D1 , ∂D2 , i = 1, 2, and {U (1) }± ∈ 0 0 1/p 1/p (i) Bp,p (∂D1 ), {ri U (2) }± ∈ Bp,p (∂D2 ), i = 1, 2, p0 = p/(p − 1). Let U (1) ∈ 1 (D2 ) be such that M(1) (∂x )U (1) ∈ Lp (D1 ), M(2) (∂x )U (2) ∈ Wp1 (D1 ), U (2) ∈ Wp,loc Lp,comp (D2 ). Then {N (1) U (1) }± and {N (2) (ri U (2) )}± , i = 1, 2, are correctly defined by the equalities Z ¤ £ (1) (1) V M (∂x )U (1) + E (1) (U (1) , V (1) ) dx D1 (1) = ±h{N U (1) }± , {V (1) }± i∂D1 for all V (1) ∈ Wp10 (D1 ) (1.4) SOLVABILITY AND ASYMPTOTICS OF SOLUTIONS and Z £ (i) D2 431 (i) (i) ¤ V 2 M(2) (∂x )(ri U (2) ) + E (2) (ri U (2) , V2 ) dx (i) = ∓h{N (2) (ri U (2) )}± , {V2 }± i∂D(i) , 2 (1) (1) (2) i = 1, 2, (1.5) (2) 1 for all V2 ∈ Wp10 (D2 ) (V2 ∈ Wp,comp (D2 )), where (1) E (1) (U (1) , V (1) ) = aijlk ξij (U (1) )ξlk (V (1) + bijlk ξij (V (1) (1) (1) ) + bijlk ξij (U (1) )ηlk (V (1) (1) )ηlk (U (1) ) + cijlk ηij (U (1) )ηlk (V ) (1) ); here U (1) = (u(1) , ω (1) ), (1) ξij (U (1) ) = ∂j ui − εijk ωk is the deformation component, ηij (U (1) ) = ∂j ωi is the bending torsion component. (i) The quadratic form E (2) (ri U (2) , V2 ), i = 1, 2, is defined analogously. In the case of the infinite domain D2 , for solving equation (1.1) the condition U (2) (x) = o(1) for |x| → ∞ (1.6) is assumed to be fulfilled at infinity (see [4]). We can prove (see [4]) that ∂ µ U (2) (x) = O(|x|−1−|µ| ) for |x| → ∞ is valid for any solution of equation (1.1) satisfying (1.6) the relation. Let us consider the model problems M1 and M2 . We will study the solvability and asymptotics of solutions U (q) ∈ Wp1 (Dq ), q = 1 1, 2, (U (2) ∈ Wp,loc (D2 ) with condition (1.6) at infinity) the following boundarycontact problems of couple-stress elasticity: Problem M1 : M(q) (∂x )U (q) = 0 πS1 {U (1) }+ = ϕ1 πS (1) {N (2) (r1 U (2) )}+ = ϕ2 in Dq , q = 1, 2, on S1 , (1) on S2 , 2 (2) πS (2) {N (2) (r2 U (2) )}+ = ϕ3 on S2 , 2 (i) πS (i) {U (1) }+ − πS (i) {ri U (2) }+ = fi on S0 , 0 0 π (i) {N (1) U (1) }+ − π (i) {N (2) (r U (2) )}+ = h on S (i) , i = 1, 2, i i 0 S S 0 0 where 1/p0 −1/p (1) −1/p (2) ϕ1 ∈ Bp,p (S1 ), ϕ2 ∈ Bp,p (S2 ), ϕ3 ∈ Bp,p (S2 ), 1/p0 (i) fi ∈ Bp,p (S0 ), −1/p (i) hi ∈ Bp,p (S0 ), i = 1, 2, 1 < p < ∞, p0 = p/(p − 1). If D2 is a finite domain, then we have the following wedge-type problem: 432 O. CHKADUA Wedge-type Problem M2 : M(q) (∂x )U (q) = 0 πS1 {N (1) U (1) }+ = ϕ1 π (1) {N (2) (r1 U (2) )}+ = ϕ2 in Dq , q = 1, 2, on S1 , (1) on S2 , S2 (2) on S2 , πS (2) {N (2) (r2 U (2) )}+ = ϕ3 2 (i) (1) + π } − πS (i) {ri U (2) }+ = fi on S0 , (i) {U S0 0 π (i) {N (1) U (1) }+ − π (i) {N (2) (r U (2) )}+ = h on S (i) , i = 1, 2, i i 0 S S 0 0 where 1/p0 1/p0 (1) 1/p0 (2) ϕ1 ∈ Bp,p (S1 ), ϕ2 ∈ Bp,p (S2 ), ϕ3 ∈ Bp,p (S2 ), −1/p 1/p0 (i) (i) fi ∈ Bp,p (S0 ), hi ∈ Bp,p (S0 ), i = 1, 2, 1 < p < ∞. 2. Fundamental Solutions and Potentials Consider the fundamental matrix-functions µ Z ¶ ¡ (q) 0 ¢−1 −iτ x3 −1 (q) H (x) = Fξ0 →x0 ± M (iξ , iτ ) e dτ , q = 1, 2, L± where the sign “−” refers to the case x3 > 0 and R the sign “+” to the case 0 0 x3 < 0; x = (x1 , x2 , x3 ), x = (x1 , x2 ), ξ = (ξ1 , ξ2 ); L± denotes integration over the contour L± , where L+ (L− ) has the positive orientation and covers all roots of the polynomial detM(q) (iξ 0 , iτ ) with respect to τ in the upper (resp. lower) τ -half-plane. F −1 is the inverse Fourier transform. The simple-layer potentials are of the form Z (1) V (g1 )(x) = H(1) (x − y)g1 (y)dy S, x ∈ / ∂D1 , ∂D1 Z (1) (2) V (g2 )(x) = H(2) (x − y)g2 (y)dy S, x ∈ / ∂D2 , (1) ∂D Z 2 (2) V(3) (g3 )(x) = H(2) (x − y)g3 (y)dy S, x ∈ / ∂D2 . (2) ∂D2 For these potentials the theorems below are valid. Theorem 2.1. Let 1 < p < ∞, 1 ≤ r ≤ ∞. Then the operators V(i) , i = 1, 2, 3, admit extensions to the operators which are continuous in the following spaces: ¢ ¡ s+1+1/p (D1 ) , (D1 ) Bsp,p (∂D1 ) → Hs+1+1/p V(1) : Bsp,r (∂D1 ) → Bp,r p (1) ¢ (1) (1) ¡ (1) (D2 ) , (D2 ) Bsp,p (∂D2 ) → Hs+1+1/p V(2) : Bsp,r (∂D2 ) → Bs+1+1/p p p,r (2) s+1+1/p (2) ¢ (2) s+1+1/p (2) ¡ V(3) : Bsp,r (∂D2 ) → Bp,r,loc (D2 ) Bsp,p (∂D2 ) → Hp,loc (D2 ) . SOLVABILITY AND ASYMPTOTICS OF SOLUTIONS 433 Theorem 2.2. Let 1 < p < ∞, 1 ≤ r ≤ ∞, ε > 0, g1 ∈ B−1+ε p,r (∂D1 ), (1) (2) −1+ε −1+ε g2 ∈ Bp,r (∂D2 ), g3 ∈ Bp,r (∂D2 ). Then Z (1) ± {V (g1 )(z)} = H(1) (z − y)g1 (y)dy S, z ∈ ∂D1 , Z∂D1 (1) {V(2) (g2 )(z)}± = H(2) (z − y)g2 (y)dy S, z ∈ ∂D2 , (1) ∂D Z 2 (2) H(2) (z − y)g3 (y)dy S, z ∈ ∂D2 . {V(3) (g3 )(z)}± = (2) ∂D2 −1/p −1/p (1) Theorem 2.3. Let 1 < p < ∞, g1 ∈ Bp,p (∂D1 ), g2 ∈ Bp,p (∂D2 ), g3 ∈ −1/p (2) Bp,p (∂D2 ). Then © (1) (1) ª± 1 N V (g1 )(z) = ∓ g1 (z) 2 Z + N (1) (∂z , n(z))H(1) (z − y)g1 (y)dy S, z ∈ ∂D1 , © ∂D1 ª± 1 N (2) V(2) (g2 )(z) = ± g2 (z) 2 Z (1) + N (2) (∂z , n(z))H(2) (z − y)g2 (y)dy S, z ∈ ∂D2 , (1) © ∂D2 ª± 1 N (2) V(3) (g3 )(z) = ± g3 (z) 2 Z (2) (2) + N (∂z , n(z))H(2) (z − y)g3 (y)dy S, z ∈ ∂D2 . (2) ∂D2 Let us introduce the following notation: Z (1) V−1 (g1 )(z) = H(1) (z − y)g1 (y)dy S, z ∈ ∂D1 , Z∂D1 (2) (1) V−1 (g2 )(z) = H(2) (z − y)g2 (y)dy S, z ∈ ∂D2 , (1) ∂D Z 2 (3) (2) V−1 (g3 )(z) = H(2) (z − y)g3 (y)dy S, z ∈ ∂D2 , (2) ∂D Z 2 ∗ (1) N (1) (∂z , n(z))H(1) (z − y)g1 (y)dy S, z ∈ ∂D1 , V 0 (g1 )(z) = Z∂D1 ∗ (2) (1) N (2) (∂z , n(z))H(2) (z − y)g2 (y)dy S, z ∈ ∂D2 , V 0 (g2 )(z) = (1) ∂D Z 2 ∗ (3) (2) N (2) (∂z , n(z))H(2) (z − y)g3 (y)dy S, z ∈ ∂D2 . V 0 (g3 )(z) = (2) ∂D2 (i) Theorem 2.4. Let 1 < p < ∞, 1 ≤ r ≤ ∞. Then the operators V−1 , ∗ (i) V 0 , i = 1, 2, 3, admit extensions to the operators which are continuous in the 434 O. CHKADUA following spaces: (i) V−1 : Hsp (∂Ωi ) → Hs+1 p (∂Ωi ) ¡ s ¢ Bp,r (∂Ωi ) → Bs+1 (∂Ω ) , i = 1, 2, 3, i p,r ∗ (i) V0 : Hsp (∂Ωi ) → Hsp (∂Ωi ) ¡ s ¢ Bp,r (∂Ωi ) → Bsp,r (∂Ωi ) , i = 1, 2, 3, (1) (2) here Ω1 = D1 , Ω2 = D2 , Ω3 = D2 . 3. Uniqueness, Existence and Smoothness Theorems for Problem M1 From the ellipticity of the differential operator M(2) (∂x ) it follows that any generalized solution of the equation M(2) (∂x )U (2) = 0 in D2 is an analytic function in D2 (see [17]). Then we see that the equalities ( {r1 U (2) }+ − {r2 U (2) }+ = 0 on S0 , {N (2) (r1 U (2) )}+ + {N (2) (r2 U (2) )}+ = 0 on S0 (3.1) are valid on S0 . Let us study the uniqueness of a solution of the boundary-contact problem 1 M1 in the classes W21 (Dq ), q = 1, 2 (W2,loc (D2 ) with condition (1.6) at infinity). Lemma 3.1. A solution of the boundary-contact problem M1 is unique in the 1 classes W21 (Dq ), q = 1, 2 (W2,loc (D2 ) and satisfying condition (1.6) at infinity). Proof. Let U (q) , q = 1, 2, be a solution of the homogeneous problem M1 . (2) (1) We write the Green formulae (see (1.4), (1.5)) in the domains D1 , D2 , D2 for the vector-functions U (1) , r1 U (2) , r2 U (2) as Z E (1) (U (1) , U (1) )dx = h{N (1) U (1) }+ , {U (1) }+ i∂D1 , D1 Z (3.2) E (2) (ri U (2) , ri U (2) )dx = −h{N (2) (ri U (2) )}+ , {ri U (2) }+ i∂D(i) , i = 1, 2. (i) 2 D2 Taking into account the boundary and boundary-contact conditions, the Green formulas (3.2) can be rewritten as Z E (1) (U (1) , U (1) )dx = h{N (1) U (1) }+ , {U (1) }+ iS (1) ∪S (2) , 0 0 D1 Z (3.3) (2) (2) (2) (2) (2) + (2) + E (ri U , ri U )dx = −h{N (ri U )} , {ri U } iS (i) ∪S0 , i = 1, 2. (i) D2 0 Since equalities (3.1) are fulfilled for the function U (2) , we have h{N (2) (r1 U (2) )}+ , {r1 U (2) }+ iS0 = −h{N (2) (r2 U (2) )}+ , {r2 U (2) }+ iS0 . (3.4) SOLVABILITY AND ASYMPTOTICS OF SOLUTIONS 435 Summing now the Green formulas (3.3) and taking into account (3.4), we obtain Z Z (1) (1) (1) E (U , U )dx + E (2) (r1 U (2) , r1 U (2) )dx (1) D1 D2 Z + E (2) (r2 U (2) , r2 U (2) )dx = 0. (3.5) (2) D2 From equality (3.5) and the positive-definiteness of forms (1.3) we have ( (q) (q) ∂j ui − εijk ωk = 0, (q) ∂j ω i = 0, q = 1, 2. Therefore u(q) = [a(1) × x] + b(q) , ω (q) = a(q) , q = 1, 2, and ¡ ¢ U (q) = [a(q) × x] + b(q) , a(q) , q = 1, 2, where a(q) and b(q) , q = 1, 2, are arbitrary three-dimensional constant vectors. By virtue of the contact conditions it is clear that a(1) = a(2) and b(1) = b(2) . Since {U (1) }+ = 0 on S1 , we have U (q) (x) = 0, x ∈ Dq , ¤ q = 1, 2. 1/p0 Any extension Φ(1) ∈ Bp,p (∂D1 ) of the function ϕ1 onto the whole boundary ∂D1 has the form (1) (1) (1) Φ(1) = Φ0 + ϕ0 + ψ0 , 0 0 (2) (1) (1) (1) (1) e 1/p e 1/p ). ), ψ ∈ B where Φ is some fixed extension of ϕ1 and ϕ ∈ B p,p (S p,p (S 0 −1/p 0 0 (1) 0 0 Any extension Φ(2) ∈ Bp,p (∂D2 ) of the function ϕ2 onto the whole bound(1) ary ∂D2 has the form (2) (2) (2) Φ(2) = Φ0 + ϕ0 + ψ0 , (2) (2) (1) (2) e −1/p e −1/p where Φ0 is some fixed extension of ϕ2 and ϕ0 ∈ B p,p (S0 ), ψ0 ∈ Bp,p (S0 ). −1/p (2) Any extension Φ(3) ∈ Bp,p (∂D2 ) of the function ϕ3 onto the whole bound(2) ary ∂D2 has the form (3) (3) (3) Φ(3) = Φ0 + ϕ0 + ψ0 , e p,p (S ), ψ ∈ B e p,p (S0 ). where Φ0 is some fixed extension of ϕ3 and ϕ0 ∈ B 0 0 A solution of the boundary-contact problem M1 will be sought for in the form of the simple-layer potentials ( (1) V(2) g2 in D2 , U (1) = V(1) g1 in D1 , U (2) = (2) V(3) g3 in D2 . (3) (3) −1/p (2) (3) −1/p 436 O. CHKADUA Taking into account the boundary and boundary-contact conditions of problem M1 and equality (3.1) we obtain a system of equations with respect to (1) (1) (2) (2) (3) (3) (g1 , g2 , g3 , ϕ0 , ψ0 , ϕ0 , ψ0 , ϕ0 , ψ0 ): (1) (1) (1) (1) V−1 g1 − ϕ0 − ψ0 = Φ0 on ∂D1 , ∗ (2) (2) (2) (2) (1) on ∂D2 , ( 21 I+ V 0 )g2 − ϕ0 − ψ0 = Φ0 ∗ (3) (3) (3) (3) (2) 1 on ∂D2 , I+ ( V 0 )g3 − ϕ0 − ψ0 = Φ0 2 (2) (1) (1) (1) −πS (1) V−1 g2 + ϕ0 = f1 − πS (1) Φ0 on S0 , 0 0 ∗ (1) (2) (2) (1) (3.6) πS (1) (− 12 I+ V 0 )g1 − ϕ0 = h1 + πS (1) Φ0 on S0 , 0 0 (3) (1) (1) (2) −πS (2) V−1 g3 + ψ0 = f2 − πS (2) Φ0 on S0 , 0 0 ∗ (1) (3) (3) (2) 1 πS (2) (− 2 I+ V 0 )g1 − ϕ0 = h2 + πS (2) Φ0 on S0 , 0 0 (2) (3) π V g − π V g = 0 on S0 , S0 −1 2 S0 −1 3 (2) (3) (2) (3) ψ0 + ψ0 = −πS0 Φ0 − πS0 Φ0 on S0 . It is almost obvious that system (3.6) has a solution if and only if the compatibility conditions on ∂S0 (2) (3) (2) (2) (3) e s−1 Φ0 ∈ Bs−1 p,r (∂D2 ) : πS0 Φ0 +πS0 Φ0 ∈ Bp,r (S0 ) (3.7) (1) ∃Φ0 ∈ Bs−1 p,r (∂D2 ), (1) (2) (2) (3) s−1 s−1 hold for ϕ2 ∈ Bs−1 p,r (S2 ), ϕ3 ∈ Bp,r (S2 ), πS0 Φ0 + πS0 Φ0 ∈ Bp,r (S0 ), 1 ≤ r ≤ ∞, 1 < p < ∞, 1/p − 1/2 < s < 1/p + 1/2. Note that these conditions hold automatically when 1/p − 1/2 < s < 1/p or 1/p < s < 1/p + 1/2 (see [39]). Denote by A the operator corresponding to system (3.6) and acting in the spaces (1) (2) (2) ¡ (1) s ¢ s A : H sp →H sp B p,r → B p,r , where (1) (1) (2) s s−1 s−1 s−1 e s (S (1) ) ⊕ H e s (S (2) ) H p = Hp (∂D1 ) ⊕ Hp (∂D2 ) ⊕ Hp (∂D2 ) ⊕ H 0 0 p p (1) (2) s−1 s−1 s−1 s−1 e e e e ⊕H (S ) ⊕ H (S0 ) ⊕ H (S ) ⊕ H (S0 ), p 0 (2) p p 0 p (1) (2) (1) (1) s−1 s s−1 s−1 s s H p = Hp (∂D1 ) ⊕ Hp (∂D2 ) ⊕ Hp (∂D2 ) ⊕ Hp (S0 ) ⊕ Hp (S0 ) (2) (2) s−1 s ⊕Hsp (S0 ) ⊕ Hs−1 p (S0 ) ⊕ Hp (S0 ) ⊕ Hp (S0 ), (1) (2) (1) s−1 s−1 s−1 s e s (S (2) ) e s (S (1) ) ⊕ B B p,r = Bp,r (∂D1 ) ⊕ Bp,r (∂D2 ) ⊕ Bp,r (∂D2 ) ⊕ B 0 0 p,r p,r (2) (1) e s−1 (S0 ), e s−1 (S ) ⊕ B e s−1 (S0 ) ⊕ B e s−1 (S ) ⊕ B ⊕B p,r 0 (2) p,r p,r (1) 0 (2) p,r (1) (1) s s s−1 s−1 s s−1 B p,r = Bp,r (∂D1 ) ⊕ Bp,r (∂D2 ) ⊕ Bp,r (∂D2 ) ⊕ Bp,r (S0 ) ⊕ Bp,r (S0 ) (2) (2) s s−1 ⊕Bsp,r (S0 ) ⊕ Bs−1 p,r (S0 ) ⊕ Bp,r (S0 ) ⊕ Bp,r (S0 ); SOLVABILITY AND ASYMPTOTICS OF SOLUTIONS 437 the symbol ⊕ denotes a direct sum of the spaces. Consider the composition of the operators D ◦ A, where D is the invertible operator of the form (2) (3) D = diag{I, −V−1 , −V−1 , I, . . . , I}54×54 . Consider now the operator AM = TM + D ◦ A, where M = 2, 3, . . . , (2) (3) TM = diag{0, (−V−1 )M , (−V−1 )M , 0, . . . , 0}54×54 . Since the operator AM differs from the operator D ◦ A in a compact operator, it is sufficient to investigate the operator AM acting in the spaces (1) (3) (3) ¡ (1) s ¢ s AM : H sp →H sp B p,r → B p,r , where (3) (1) (2) (1) (1) s s s s s s−1 H p = Hp (∂D1 ) ⊕ Hp (∂D2 ) ⊕ Hp (∂D2 ) ⊕ Hp (S0 ) ⊕ Hp (S0 ) (2) (2) s s−1 ⊕Hsp (S0 ) ⊕ Hs−1 p (S0 ) ⊕ Hp (S0 ) ⊕ Hp (S0 ), (3) (1) (2) (1) (1) s s s s s s−1 B p,r = Bp,r (∂D1 ) ⊕ Bp,r (∂D2 ) ⊕ Bp,r (∂D2 ) ⊕ Bp,r (S0 ) ⊕ Bp,r (S0 ) (2) (2) s s−1 ⊕Bsp,r (S0 ) ⊕ Bs−1 p,r (S0 ) ⊕ Bp,r (S0 ) ⊕ Bp,r (S0 ). Now consider the system of equations that corresponds to given by (1) (1) (1) (1) V−1 ge1 − ϕ e0 − ψe0 = Ψ0 ∗ (2) (2) (2) (2) (2) (2) (2) M (2) 1 I+ e0 +V−1 ψe0 = Ψ0 ) ( g2 +V−1 ϕ [(−V −V V 0 )]e −1 −1 2 ∗ (3) (3) M (3) 1 (3) (3) (3) (3) (3) I+ g3 +V−1 ϕ e0 +V−1 ψe0 = Ψ0 (−V ) −V ( V −1 −1 2 0 )e (2) (1) −πS (1) V−1 ge2 + ϕ e0 = G1 0 ∗ (1) (2) πS (1) (− 12 I+ V 0 )e g1 − ϕ e0 = G2 0 (3) (1) −πS (2) V−1 ge3 + ψe0 = F1 0 ∗ (1) (3) 1 π g1 − ϕ e0 = F2 (2) (− I+ V 0 )e 2 S 0 (2) (3) π e2 − πS0 V−1 ge3 = E1 S0 V−1 g e(2) e(3) ψ0 + ψ0 = E2 where (1) Ψ0 ∈ Hsp (∂D1 ) ¡ ¢ Bsp,r (∂D1 ) , (3) Ψ0 (2) (1) Ψ0 ∈ Hsp (∂D2 ) (2) ¢ (2) ¡ ∈ Hsp (∂D2 ) Bsp,r (∂D2 ) , ¡ the operator AM on ∂D1 , (1) on ∂D2 , (2) on ∂D2 , (1) on S0 , (1) on S0 , (2) on S0 , (2) on S0 , on S0 , on S0 , (1) ¢ Bsp,r (∂D2 ) , (3.8) 438 O. CHKADUA (1) ¢ (1) ¡ s−1 (1) ¢ Bsp,r (S0 ) , G2 ∈ Hs−1 Bp,r (S0 ) , p (S0 ) (2) ¡ (2) ¢ (2) ¡ s−1 (2) ¢ F1 ∈ Hsp (S0 ) Bsp,r (S0 ) , F2 ∈ Hs−1 Bp,r (S0 ) , p (S0 ) ¡ ¢ ¡ s−1 ¢ E1 ∈ Hsp (S0 ) Bsp,r (S0 ) , E2 ∈ Hs−1 Bp,r (S0 ) . p (S0 ) (1) G1 ∈ Hsp (S0 ) ¡ (1) ∗ (i) (i) The ΨDO −V−1 is positive and the operators −V−1 ( 12 I+ V 0 ), i = 2, 3, are nonnegative, i.e., (1) −1/2 h−V−1 ϕ, ϕi∂D1 > 0 for all ϕ ∈ H2 (∂D1 ), ϕ 6= 0, D ³ E ∗ (2) ´ (2) 1 −1/2 (1) − V−1 I+ V 0 ψ, ψ ≥ 0 for all ψ ∈ H2 (∂D2 ) (1) 2 ∂D2 and D (3) − V−1 ³1 ∗ (3) ´ I+ V 0 E −1/2 ψ, ψ (2) > 0 for all ψ ∈ H2 (∂D2 ), (2) 2 ∂D2 the equality being fulfilled only when ψ = ([a × x] + b, a), where a and b are arbitrary three-dimensional constant vectors. The proof of these inequalities follows from the Green formulae (see [32]). Then the ΨDOs ³ ∗ (i) ´ (i) (i) (i) 1 BM = (−V−1 )M − V−1 I+ V 0 , i = 2, 3, 2 are positive operators, i.e., (2) −1/2 hBM ϕ, ϕi∂D(1) > 0 for all ψ ∈ H2 2 (3) −1/2 hBM ψ, ψi∂D(2) > 0 for all ψ ∈ H2 2 (1) (1) (∂D2 ), ϕ 6= 0, (2) (∂D2 ), ψ 6= 0. (i) Hence the ΨDOs V−1 and BM , i = 2, 3, are invertible (which is proved as in [32], [7]). The first, second and third equations of system (3.8) imply ge1 = (V−1 )−1 ϕ e0 + (V−1 )−1 ψe0 + (V−1 )−1 Ψ0 , (i) (i) (i) (i) (i) (i) (i) (i) gei = −(BM )−1 V−1 ϕ e0 − (BM )−1 V−1 ψe0 + (BM )−1 Ψ0 , i = 2, 3. (1) (1) (1) (1) (1) (1) After substituting ge1 , ge2 , ge3 into the remaining equations of system (3.8), we obtain a system of equations whose corresponding operator has the form πS (1) A(x, D) 0 0 0 0 πS (2) B(x, D) 0 + T−∞ , P= 0 0 0 πS0 C(x, D) 36×36 where A(x, D) = B(x, D) = (2) (2) (2) πS (1) V−1 (BM )−1 V−1 0 πS (1) (− 12 I+ 0 −I (3) (3) I (3) ∗ (1) (1) V 0 )(V−1 )−1 πS (2) V−1 (BM )−1 V−1 I −I πS (2) (− 12 I+ V 0 )(V−1 )−1 0 ∗ (1) 0 , (1) , SOLVABILITY AND ASYMPTOTICS OF SOLUTIONS à C(x, D) = (2) (2) (2) (3) (3) 439 ! (3) −πS0 V−1 (BM )−1 V−1 πS0 V−1 (BM )−1 V−1 I I and T−∞ is the operator of order −∞. Further, after the localization, the operators A(x, D) and B(x, D) are reduced by means of lifting to the strongly elliptic ΨDOs of order 1, while the operator C(x, D) is reduced to positive-definite ΨDO. Indeed, let A(x0 , D0 ) and B(x0 , D0 ) be ΨDOs with the symbols σA (x0 , ξ 0 ) and σB (x0 , ξ 0 ) (ξ 0 = (ξ1 , ξ2 )) “frozen” at the points and written in terms of some (1) (2) local coordinate system of the manifolds S0 and S0 , respectively. Denote µ ¶ µ ¶ L− 0 L+ 0 0 0 0 0 R(x , D ) = ◦ A(x , D ) ◦ 0 I 0 I and µ 0 0 Q(x , D ) = L− 0 0 I ¶ µ 0 0 ◦ B(x , D ) ◦ L+ 0 0 I ¶ , where L+ = diagΛ+ , L− = diagπ+ Λ− ` are 6 × 6 matrix operators, Λ± is a ΨDO operator with the symbol Λ± (ξ 0 ) = ξ2 ± i ± i|ξ1 |, π+ denotes the operator of restriction onto R2+ , and ` is an extension operator. The operators µ ¶ L± 0 0 I are invertible in the respective spaces (see [39]). The principal homogeneous symbols of the ΨDOs R(x0 , D0 ) and Q(x0 , D0 ) are written as à ! 00 0 0 00 00 (ξ −i|ξ |)σ (x , ξ )(ξ +i|ξ |) (ξ −i|ξ |)I n−1 N n−1 n−1 2 (1) σR (x0 , ξ 0 ) = , x0 ∈ S 0 , 00 0 0 −(ξn−1 + i|ξ |)I σN1 (x , ξ ) à ! 00 0 0 00 00 (ξ −i|ξ |)σ (x , ξ )(ξ +i|ξ |) (ξ −i|ξ |)I n−1 N n−1 n−1 3 (2) σQ (x0 , ξ 0 ) = , x0 ∈ S 0 , 00 0 0 −(ξn−1 + i|ξ |)I σN1 (x , ξ ) where σN1 (x0 , ξ 0 ), σN2 (x0 , ξ 0 ) and σN3 (x0 , ξ 0 ) are the principal homogeneous symbols of the ΨDOs ³ 1 ∗ (1) ´ (1) (2) (2) (2) (3) (3) (3) N1 = − I+ V 0 (V−1 )−1 , N2 = V−1 (BM )−1 V−1 , N3 = V−1 (BM )−1 V−1 , 2 respectively, written in terms of a given local coordinate system, and I is the identity matrix. (k) Let λR , k = 1, . . . , 12, be the eigenvalues of the matrix ¡ ¢−1 (1) σR (x1 , 0, +1) σR (x1 , 0, −1), x1 ∈ ∂S0 , where µ σR (x1 , 0, −1) = σN2 (x1 , 0, −1) −I I σN1 (x1 , 0, −1) ¶ , 440 O. CHKADUA µ σR (x1 , 0, +1) = σN2 (x1 , 0, +1) I −I σN1 (x1 , 0, +1) ¶ , (k) and let λQ , k = 1, . . . , 12, be the eigenvalues of the matrix ¢−1 ¡ (2) σQ (x1 , 0, +1) σQ (x1 , 0, −1), x1 ∈ ∂S0 , where µ σQ (x1 , 0, −1) = µ σQ (x1 , 0, +1) = σN3 (x1 , 0, −1) −I I σN1 (x1 , 0, −1) σN3 (x1 , 0, +1) I −I σN1 (x1 , 0, +1) Introduce the notation ¯ ¯1 ¯ ¯ (j) δR = sup ¯ arg λR (x1 )¯, 1≤j≤12 2π δQ = (1) ¶ , ¶ ¯1 ¯ ¯ ¯ (j) sup ¯ arg λQ (x1 )¯, 1≤j≤12 2π (2) x1 ∈∂S0 x1 ∈∂S0 δ = max(δR , δQ ). Using the general theory of pseudodifferential operators (ΨDOs) (see [35], [36], the following propositions are valid. Lemma 3.2. Let 1 < p < ∞, 1 ≤ r ≤ ∞, 1/p − 1/2 + δ < s < 1/p + 1/2 − δ. Then the operators e s (R2 ) ⊕ H e s (R2 ) → Hs−1 (R2 ) ⊕ Hs−1 (R2 ) R(x0 , D0 ) : H p + p + p + p + ¢ ¡ s e (R2 ) ⊕ B e s (R2 ) → Bs−1 (R2 ) ⊕ Bs−1 (R2 ) , B p,r + p,r + p,r + p,r + e s (R2 ) ⊕ H e s (R2 ) → Hs−1 (R2 ) ⊕ Hs−1 (R2 ) Q(x0 , D0 ) : H p + p + p + p + ¡ s ¢ 2 s 2 s−1 2 s−1 e (R ) ⊕ B e (R ) → B (R ) ⊕ B (R2 ) B p,r + p,r + p,r + p,r + are Fredholm. Note that the ΨDOs R(x0 , D0 ) and Q(x0 , D0 ) are Fredholm in the anisotropic Bessel potential spaces with weight e (µ,s),k (R2 ) ⊕ H e (µ,s),k (R2 ) → H(µ,s−1),k (R2 ) ⊕ H(µ,s−1),k (R2 ) H p + p + p + p + for all µ ∈ R and k = 0, 1, . . . (see [10]). Since the operators πS (1) N1 , πS (1) N2 , πS (2) N1 and πS (2) N2 are positive-definite, 0 0 0 0 we obtain a strong Gårding inequality for the operators A(x, D) and B(x, D) (see [6, Lemma 3.3]). Hence, using the results obtained in [2], [23], we have Lemma 3.3. Let 1 < p < ∞, 1 ≤ r ≤ ∞, 1/p − 1/2 + δ < s < 1/p + 1/2 − δ. Then the operators e s−1 (S (1) ) ⊕ H e s (S (1) ) → Hs (S (1) ) ⊕ Hs−1 (S (1) ) πS (1) A(x, D) : H 0 0 0 0 p p p p 0 ¡ s−1 (1) ¢ (1) (1) e (S ) ⊕ B e s (S ) → Bs (S ) ⊕ Bs−1 (S (1) ) B 0 0 0 0 p,r p,r p,r p,r e s−1 (S (2) ) ⊕ H e s (S (2) ) → Hs (S (2) ) ⊕ Hs−1 (S (2) ) πS (2) A(x, D) : H 0 0 0 0 p p p p 0 SOLVABILITY AND ASYMPTOTICS OF SOLUTIONS ¡ 441 ¢ e s−1 (S (2) ) ⊕ B e s (S (2) ) → Bs (S (2) ) ⊕ Bs−1 (S (2) ) B 0 0 0 0 p,r p,r p,r p,r are invertible. Let us consider the operator C(x, D). The corresponding system à ! µ ¶ (2) e1 ϕ0 E πS0 C(x, D) = (3) E2 ϕ0 is reduced to pseudodifferential equation on the open manifold S0 (3) e1 , πS0 Nψe0 = E (2) (3) ψe0 = −ψe0 + E2 , where N = N2 + N3 . The ΨDO πS0 N is positive-definite and the following proposition holds for it. Lemma 3.4. Let 1 < p < ∞, 1 ≤ r ≤ ∞, 1/p − 1/2 + δ < s < 1/p + 1/2 − δ. Then the ΨDOs e s−1 (S0 ) → Hs (S0 ) πS0 N : H p p ¡ s−1 ¢ e (S0 ) → Bs (S0 ) B p,r p,r and e s−1 (S0 ) ⊕ H e s−1 (S0 ) → Hs (S0 ) ⊕ Hs (S0 ) πS0 C(x, D) : H p p p p ¢ ¡ s−1 s−1 s e (S0 ) ⊕ B e (S0 ) → B (S0 ) ⊕ Bs (S0 ) B p,r p,r p,r p,r are invertible. Note that the ΨDO πS0 N is invertible in anisotropic Bessel potential spaces (µ,s),k e (µ,s−1),k (S0 ) (see [10]). → Hp with weight H p Lemmas 3.3 and 3.4 imply the validity of the following proposition. Lemma 3.5. Let 1 < p < ∞, 1 ≤ r ≤ ∞, 1/p − 1/2 + δ < s < 1/p + 1/2 − δ. Then the operator (1) (1) e s−1 (S (1) ) ⊕ H e s (S (1) ) H Hsp (S0 ) ⊕ Hs−1 0 0 p p p (S0 ) ⊕ ⊕ e s−1 (S (2) ) ⊕ H e s (S (2) ) → Hs (S (2) ) ⊕ Hs−1 (S (2) ) P : H 0 0 p p 0 0 p p ⊕ ⊕ e s−1 (S0 ) e s−1 (S0 ) ⊕ H Hsp (S0 ) ⊕ Hsp (S0 ) H p p (1) e s (S (1) ) e s−1 (S (1) ) ⊕ B e s−1 (S (1) ) B Bsp,r (S0 ) ⊕ B 0 p,r p,r 0 ⊕ p,r 0 ⊕ e s−1 (2) e s (S (2) ) → Bs (S (2) ) ⊕ B e s−1 (S (2) ) Bp,r (S0 ) ⊕ B 0 p,r 0 0 p,r p,r ⊕ ⊕ e s−1 (S0 ) e s−1 (S0 ) ⊕ B Bsp,r (S0 ) ⊕ Bsp,r (S0 ) B p,r p,r is invertible. 442 O. CHKADUA Theorem 3.6. Let 1 < p < ∞, 1 ≤ r ≤ ∞, 1/p−1/2+δ < s < 1/p+1/2−δ, M = 2, 3, . . . . Then the operator (1) (3) ¡ (1) s ¢ s B p,r → B p,r (3) AM : H sp →H sp is invertible. Lemma 2.1 and Theorem 3.6 imply that the following proposition is valid. Theorem 3.7. Let 1 < p < ∞, 1 ≤ r ≤ ∞, 1/p−1/2+δ < s < 1/p+1/2−δ. Then the operator (1) (2) A : H sp →H sp ¡ (1) B s p,r (2) →B s p,r ¢ is invertible. If we take s = 1/p0 in the condition for the operator A to be invertible (see Theorem 3.7), we conclude that p must satisfy the equality 4 4 <p< . 3 − 2δ 1 + 2δ Theorem 3.7 and the above reasoning imply the validity of the solution existence and uniqueness for problem M1 . Theorem 3.8. Let 4/(3−2δ) < p < 4/(1+2δ) and the compatibility condition (3.7) be fulfilled for s = 1 − 1/p. Then the boundary-contact problem M1 has 1 a unique solution in the classes Wp1 (Dq ), q = 1, 2, (Wp,loc (D2 ) with condition (1.6) at infinity), which is given by the formulae (1) in D2 , V(2) g2 (2) U (1) = V(1) g1 in D1 , U (2) = V(3) g3 in D2 , π V(2) g = π V(3) g on S , S0 −1 2 S0 −1 3 0 where gq , q = 1, 2, 3, are obtained from system (3.6). Theorems 2.1, 3.7 and the embedding theorems (see [39]) imply Theorem 3.9. Let 4/(3 − 2δ) < p < 4/(1 + 2δ), 1 ≤ t ≤ ∞, 1 < r < ∞, 1/r −1/2+δ < s < 1/r +1/2−δ, the compatibility condition (3.7) with t instead 1 (D2 ) with condition (1.6) of p be fulfilled, U (q) ∈ Wp1 (Dq ), q = 1, 2, (U (2) ∈ Wp,loc at infinity) be a solution of the boundary-contact problem M1 . Then: (i) (2) (1) s s−1 if ϕ1 ∈ Bsr,r (S1 ), ϕ2 ∈ Bs−1 r,r (S2 ), ϕ3 ∈ Br,r (S2 ), fi ∈ Br,r (S0 ), hi ∈ (i) s+1/r s+1/r (q) ∈ Hr (Dq ), q = 1, 2, (U (2) ∈ Hr,loc (D2 )); Bs−1 r,r (S0 ), i = 1, 2, we have U (1) (2) (i) s−1 s if ϕ1 ∈ Bsr,t (S1 ), ϕ2 ∈ Bs−1 r,t (S2 ), ϕ3 ∈ Br,t (S2 ), fi ∈ Br,t (S0 ), hi ∈ s+1/r (i) s+1/r (q) ∈ Br,t (Dq ), q = 1, 2, (U (2) ∈ Br,t,loc (D2 )). Bs−1 r,t (S0 ), i = 1, 2, we have U SOLVABILITY AND ASYMPTOTICS OF SOLUTIONS 443 4. Asymptotics of Solutions Now we will write the asymptotics of solutions of the boundary-contact problem M1 . The boundary and contact data are assumed to be sufficiently smooth, i.e., (1) (2) ),∞ ),∞ ϕ1 ∈ Hr(∞,s+2M +1),∞ (S1 ), ϕ2 ∈ H(∞,s+2M (S2 ), ϕ3 ∈ H(∞,s+2M (S2 ), r r (i) (i) ),∞ +1),∞ (S0 ), hi ∈ H(∞,s+2M (S0 ), i = 1, 2; fi ∈ H(∞,s+2M r r here the numbers r and s satisfy the conditions of Theorem 3.9. Let m1 , . . . , m2` be algebraic multiplicities of the eigenvalues λ1 , . . . , λ2` , P2` (j) j=1 mj = 12, where λj = λR , j = 1, . . . , 2`. We introduce the notation ¡ ¢−1 bR (x1 ) = σR (x1 , 0, +1) σR (x1 , 0, −1). Let (1) b0R (x1 ) = K−1 (x1 )bR (x1 )K(x1 ), x1 ∈ ∂S0 , be a quasi-diagonal form, where K is some nondegenerate matrix function detK(x1 ) 6= 0, and K ∈ C ∞ (see [20]). The asymptotics of solutions to a strongly elliptic pseudodifferential equation (see [10, Theorem 2.1]) implies the asymptotics of solutions of the pseudodifferential equation R(x0 , D0 )χ = F, ),∞ ),∞ F ∈ H(∞,s+M (R2+ ) × H(∞,s+M (R2+ ) r,comp r,comp (1) written in terms of some local coordinate system of the manifold S0 . Thus we obtain an asymptotic expansion of the solution χ = (χ1 , χ2 )> ´ ³ 1 1/2+∆(x1 ) 0 log x2,+ K−1 (x1 )c0 (x1 ) χ(x1 , x2,+ ) = K(x1 )x2,+ Bapr − 2πi M X 1/2+∆(x1 )+k + K(x1 )x2,+ Bk (x1 , log x2,+ ) + χM +1 (x1 , x2,+ ), (4.1) k=1 +1),∞ +1),∞ e (∞,s+M e (∞,s+M for all sufficiently small x2,+ > 0, χM +1 ∈ H (R2+ )× H (R2+ ); r,comp r,comp B0apr (t) is defined in [10]; the 12 × 12 matrix function Bk (x1 , t) is a polynomial of order νk = k(2m0 − 1) + m0 − 1, m0 = max{m1 , . . . , m2` } with respect to the variable t with 12-dimensional vector coefficients which depend on the variable x1 and ∆(x1 ) = (∆1 (x1 ), ∆2 (x1 )); here (j) (j) (j) (j) ∆j (x1 ) = (δ1 (x1 ), . . . , δ1 (x1 ), . . . , δ` (x1 ), . . . , δ` (x1 )) , j = 1, 2, | {z } | {z } m1 -times m` -times i 1 (1) arg λk (x1 ) − |λk (x1 )|, δk (x1 ) = 2π 2π 1 i (2) δk (x1 ) = − arg λk (x1 ) − |λk (x1 )|, k = 1, . . . , `. 2π 2π 444 O. CHKADUA Without loss of generality suppose that the matrix B0apr (t) has the form B0apr (t) = diag{Ba0pr (t), Ba0pr (t)}; here Ba0pr (t) is the upper triangular block-diagonal 6×6-matrix-function defined in [9]. Hence for the functions χ1 and χ2 we can write an asymptotic expansion. Indeed, let µ ¶ K11 (x1 ) K12 (x1 ) K(x1 ) = K21 (x1 ) K22 (x1 ) 12×12 and (1) (2) K−1 (x1 )c0 (x1 ) = (c0 (x1 ), c0 (x1 ))> , (4.2) (i) where Kij (x1 ), i, j = 1, 2, are 6 × 6-matrices, c0 , i = 1, 2, are six-dimensional vector functions. Then 2 ³ ´ X 1 1/2+∆j (x1 ) 0 (i) χi (x1 , x2,+ ) = log x2,+ c0 (x1 ) Kij (x1 )x2,+ Bapr − 2πi j=1 + 2 X M X 1/2+∆j (x1 )+k Kij (x1 )x2,+ (i) (i) Bkj (x1 , log x2,+ ) + χM +1 (x1 , x2,+ ), i = 1, 2, (4.3) j=1 k=1 (i) +1) (i) e (∞,s+M (R2+ ) and Bkj (x1 , t) is a polynomial of order νk = where χM +1 ∈ H r,comp k(2m0 − 1) + m0 − 1 with respect to the variable t with six-dimensional vector coefficients which depend on the variable x1 . We can also obtain an analogous asymptotic expansion of the solution χ e= (e χ1 , χ e2 ) of the strongly elliptic equation ),∞ ),∞ Q(x0 , D0 )e χ = Fe, Fe ∈ H(∞,s+M (R2+ ) × H(∞,s+M (R2+ ) r,comp r,comp (2) in terms of some local coordinate system on the manifold S0 . Indeed, we have χ ei (x1 , x2,+ ) = 2 X 1/2+∆j (x1 ) Kij (x1 )x2,+ j=1 + M 2 X X e j (x1 )+k 1/2+∆ Kij (x1 )x2,+ ³ ´ 1 (i) Ba0pr − log x2,+ b0 (x1 ) 2πi (i) (i) Bkj (x1 , log x2,+ ) + χ eM +1 (x1 , x2,+ ), i = 1, 2, (4.4) j=1 k=1 (i) +1),∞ e (∞,s+M e j , j = 1, 2, are defined as ∆j , j = 1, 2, by where χ eM +1 ∈ H (R2+ ), ∆ r,comp (k) means of the eigenvalues λQ , k = 1, . . . , 12, of the matrix bQ . Let us consider the pseudodifferential equation (3) e1 and ψe0(2) = −ψe0(3) + E2 , πS0 Nψe0 = E where (2) (2) (2) (3) (3) (3) N = V−1 (BM )−1 V−1 + V−1 (BM )−1 V−1 . SOLVABILITY AND ASYMPTOTICS OF SOLUTIONS 445 The following equalities hold for the principal homogeneous symbols of the ∗ (i) (i) operators V−1 and V 0 , i = 2, 3: σV(2) (x0 , ξ 0 ) = σV(3) (x0 , ξ 0 ) for x0 ∈ S 0 , −1 −1 (4.5) σ ∗(2) (x0 , ξ 0 ) = −σ ∗(3) (x0 , ξ 0 ) for x0 ∈ S 0 . V0 V0 In view of equality (4.5) we can write the symbol σN (x0 , ξ 0 ) of the ΨDO operator N as follows: h³ 1 ´¡ ¢−1 i−1 σN (x0 , ξ 0 ) = I + σ ∗(2) (x0 , ξ 0 ) σ−V(2) (x0 , ξ 0 ) V0 −1 2 h³ 1 ´¡ ¢ i−1 0 0 0 0 −1 ∗ + I − σ (2) (x , ξ ) σ−V(2) (x , ξ ) . V0 −1 2 Since the symbol σ ∗(2) (x0 , ξ 0 ) is an odd matrix function with respect to ξ 0 , while V0 the symbol σ−V(2) (x0 , ξ 0 ) is an even matrix function, one can easily ascertain that −1 the symbol σN (x0 , ξ 0 ) is even with respect to the variable ξ 0 , i.e., σN (x0 , −ξ 0 ) = σN (x0 , ξ 0 ), x0 ∈ S 0 , and all eigenvalues of the matrix ¡ ¢−1 σN (x1 , 0, +1) σN (x1 , 0, −1) = I, x1 ∈ ∂S0 , (j) are trivial λN = 1, j = 1, . . . , 6. Applying the result on strongly elliptic pseudodifferential equations (see [10, Theorem 2.1]), we obtain, in terms of some local coordinate system, the follow(i) ing result on asymptotic expansion of the functions ψ0 , i = 2, 3: (i) ψ0 (x1 , x2,+ ) i+1 = (−1) −1/2 c0 (x1 )x2,+ + M X −1/2+k (i) (i) dk (x1 )+ψM +1 (x1 , x2,+ ), x2,+ (4.6) k=1 (i) +1),∞ e (∞,s+M where ∈ C0∞ (R), and the remainder ψM +1 ∈ H (R2+ ), i = 2, 3, r,comp M ∈ N. As we can see from (4.6), due to the properties of the symbol σN (x0 , ξ 0 ) (see [12]) there are no logarithms in the entire asymptotic expansion. (1) (1) (2) (2) (3) (3) Let g = (g1 , g2 , g3 , ϕ0 , ψ0 ϕ0 , ψ0 , ϕ0 , ψ0 ) be a solution of system (3.6), i.e., Ag = Φ, where ¡ (1) (2) (3) (1) (2) (1) Φ = Φ0 , Φ0 , Φ0 , f1 − πS (1) Φ0 , h1 + πS (1) Φ0 , f2 − πS (2) Φ0 , 0 0 0 (3) (2) (3) ¢ h2 + πS (2) Φ0 , 0, −(πS0 Φ0 + πS0 Φ0 ) . (i) c0 , d k 0 Then D ◦ Ag = Ψ; (4.7) here ¡ (1) (2) (2) (3) (3) (1) (2) (1) Ψ = Φ0 , −V−1 Φ0 , −V−1 Φ0 , f1 − πS (1) Φ0 , h1 + πS (1) Φ0 , f2 − πS (2) Φ0 , 0 0 0 446 O. CHKADUA (3) (2) (3) ¢ h2 + πS (2) Φ0 , 0, −(πS0 Φ0 + πS0 Φ0 ) . 0 Now adding the expression © ª (2) (3) T2M +1 g = diag 0, −(V−1 )2M +1 , −(V−1 )2M +1 , 0, . . . , 0 g to both parts of system (4.7), we obtain the equality e A2M +1 g = Ψ, (4.8) where ¡ (2) (2) (2) 2M +1 (2) (3) (3) e = Φ(1) Ψ g2 , −V−1 Φ0 − (V−1 )2M +1 g3 , f1 − π 0 , −V−1 Φ0 − (V−1 ) (1) S0 (2) (1) (3) (2) (3) ¢ h1 + πS (1) Φ0 , f2 − πS (2) Φ0 , h2 + πS (2) Φ0 , 0, −(πS0 Φ0 + πS0 Φ0 ) . 0 0 0 (2) (1) Thus (−L−1 + ϕ0 , ϕ0 ) satisfies, in some (1) fold S0 , the pseudodifferential equation µ 0 0 R(x , D ) (3) (1) Φ0 , local coordinate system of the mani- χ1 χ2 ¶ = F, (1) and (−L−1 + ψ0 , ϕ0 ) satisfies, in some local coordinate system of the manifold (2) S0 , the pseudodifferential equation µ ¶ χ e1 0 0 Q(x , D ) = Fe, χ e2 where F = (L− F1 , F2 ), (1) (2) (2) (2) (2) F1 = f1 − πS (1) Φ0 − πS (1) V−1 (B2M +1 )−1 V−1 Φ0 0 0 (2) (2) (2) (2) −πS (1) V−1 (B2M +1 )−1 V−1 ψ0 (2) (2) (2) − πS (1) V−1 (B2M +1 )−1 (V−1 )2M +1 g2 , 0 0 ³ 1 ∗ (1) ´ (2) (1) (1) F2 = h1 + πS (1) Φ0 − πS (1) − I+ V 0 (V−1 )−1 Φ0 0 0 2 ³ 1 ∗ (1) ´ (1) (1) −πS (1) − I+ V 0 (V−1 )−1 ψ0 0 2 and Fe = (L− Fe1 , Fe2 ), (1) (3) (3) (3) (3) Fe1 = f2 − πS (2) Φ0 + πS (2) V−1 (B2M +1 )−1 V−1 Φ0 0 0 (3) (3) (3) (3) −πS (2) V−1 (B2M +1 )−1 (V−1 )2M +1 ψ0 0 (3) (3) (3) − πS (2) V−1 (B2M +1 )−1 (V−1 )2M +1 g3 , 0 ³ 1 ∗ (1) ´ (1) (1) (3) Fe2 = h2 + πS (2) Φ0 − πS (2) − I+ V 0 (V−1 )−1 Φ0 0 0 2 ³ 1 ∗ (1) ´ (1) (1) −πS (2) − I+ V 0 (V−1 )−1 ϕ0 ; 0 2 here (∞,s+2M ),∞ Fi ∈ Hr,comp (R2+ ), ),∞ Fei ∈ H(∞,s+2M (R2+ ), i = 1, 2. r,comp SOLVABILITY AND ASYMPTOTICS OF SOLUTIONS (2) 447 (3) Further, we have that (ψ0 , ψ0 ) is a solution of the system à ! µ ¶ (2) 0 ψ0 πS0 C(x, D) = . (2) (3) (3) −(πS0 Φ0 + πS0 Φ0 ) ψ0 This system can be reduced to a pseudodifferential equation with the positivedefinite operator (3) (2) πS0 Nψ0 = E1 , where (2) (3) (2) (3) ψ0 = −ψ0 − (πS0 Φ0 + πS0 Φ0 ), (2) (2) (3) (3) (3) N = V−1 (B2M +1 )−1 V−1 + V−1 (B2M +1 )−1 V−1 and (2) (2) (2) (2) (2) (2) (2) E1 = −πS0 V−1 (B2M +1 )−1 V−1 Φ0 − πS0 V−1 (B2M +1 )−1 (V−1 )2M +1 g2 (2) (2) (2) (2) (2) (2) (2) (2) (3) + πS0 V−1 (B2M +1 )−1 V−1 ϕ0 − πS0 V−1 (B2M +1 )−1 V−1 (πS0 Φ0 + πS0 Φ0 ); here ),∞ E1 ∈ H(∞,s+2M (R2+ ). r,comp Hence we can obtain asymptotic expansions (4.3), (4.4) and (4.6) for the func(2) (2) (1) (3) (2) −1 (3) tions −L−1 + ϕ0 , ϕ0 , −L+ ψ0 , ϕ0 and ψ0 , ψ0 , respectively. We now define g1 , g2 and g3 by the first three equations of system (4.8) (1) (1) (1) (1) (1) (1) g1 = (V−1 )−1 ϕ0 + (V−1 )−1 ψ0 + (V−1 )−1 Φ0 , (i) (i) (i) gi = −(B2M +1 )−1 V−1 ϕ0 − (i) (i) + (B2M +1 )−1 Φ0 + Gi , (4.9) (i) (i) (i) (B2M +1 )−1 V−1 ψ0 i = 2, 3, (4.10) where (i) (i) Gi = (B2M +1 )−1 (−V−1 )2M +1 gi , i = 2, 3, (∞,s+2M ),∞ G2 ∈ Hr (1) (∞,s+2M ),∞ (∂D2 ), G3 ∈ Hr (2) (∂D2 ). Using expansions (4.9) and (4.10), we obtain the following representation, i.e., the solutions of the boundary-contact problem M1 are expressed by the potential-type functions (1) (1) (1) (1) U (1) = V(1) (V−1 )−1 ϕ0 + V(1) (V−1 )−1 ψ0 + R1 , (4.11) (2) (2) (2) (2) (2) (2) (3) (3) (3) (3) (3) (3) r1 U (2) = −V(2) (B2M +1 )−1 V−1 ϕ0 − V(2) (B2M +1 )−1 V−1 ψ0 + R2 , (4.12) r2 U (2) = −V(3) (B2M +1 )−1 V−1 ϕ0 − V(3) (B2M +1 )−1 V−1 ψ0 + R3 , (4.13) where (1) R1 ∈ C M +1 (D1 ), R2 ∈ C M +1 (D2 ), (1) (1) suppϕ0 ⊂ S 0 , (2) suppψ0 ⊂ S 0 , (1) (2) (3) (2) suppψ0 ⊂ S 0 , (2) R3 ∈ C M +1 (D2 ), (2) (1) suppϕ0 ⊂ S 0 , (3) suppϕ0 ⊂ S 0 , suppψ0 ⊂ S 0 . Thus, taking into account (4.11), (4.12), (4.13), using the asymptotic expan(2) (3) (1) (1) (2) −1 (3) and ψ0 , ψ0 (see (4.3), sions of the functions −L−1 + ϕ0 , ϕ0 , −L+ ψ0 , ϕ0 448 O. CHKADUA (4.4), (4.6)), the asymptotic expansion of potential-type functions (see [9, Theorems 2.2 and 2.3]) we derive the following asymptotic expansions of the solutions of the considered boundary-contact problem M1 in terms of some local (1) (2) coordinate systems of curves ∂S0 , ∂S0 , ∂S0 : (1) a) the asymptotic expansion near the contact boundary ∂S0 : U (q) (x1 , x2 , x3 ) = (u(q) , ω (q) )(x1 , x2 , x3 ) ½ nX `0 2 X s −1 h XX (q) (q) m = Re x3 dsjm (x1 , θ)(zs,θ )1/2+∆j (x1 )−m θ=±1 j=1 s=1 ³ m=0 ×Ba0pr − + M +2 MX +2−l X ´i 1 (q) (q) log zs,θ cjm (x1 ) 2πi ¾ (q) (q) 1/2+∆j (x1 )+k+p (q) (q) Bskmpj (x1 , log zs,ϑ ) xl2 xm 3 dslmpj (x1 , ϑ)(zs,ϑ ) k,l=0 p+m=0 k+l+p+m6=0 (q) +UM +1 (x1 , x2 , x3 ) for M > (q) (q) 2 − min{[s − 1], 0}, q = 1, 2, r (4.14) (q) with the coefficients dsjm (·, ±1), cjm , dslmpj (·, ±1) ∈ C0∞ (R) and the remainder (q) UM +1 ∈ C0M +1 (R3± ), q = 1, 2, where the signs “+” and “-” refer to the cases q = 1 and q = 2, respectively. Here (q) (q) (q) (q) zs,+1 = (−1)q [x2 + x3 τs,+1 ], zs,−1 = (−1)q+1 [x2 − x3 τs,−1 ], −π < Argzs,±1 < π, τs,±1 ∈ C0∞ (R), ◦ (q) 0 are all different roots of the polynomial det M (q) ((Jκ> (x1 , 0))−1 · {τs,±1 }`s=1 (0, ±1, τ )) of multiplicity ns , s = 1, . . . , `0 , in the complex lower half-plane (ns (q) and `0 depend on q). Bskmpj (x1 , t) is a polynomial of order νkmp = νk + p + m P̀ mj = 6) with respect (νk = k(2m0 − 1) + m0 − 1, m0 = max{m1 , . . . , m` }, j=1 to the variable t with vector coefficients which depend on the variable x1 . We write the following relation between the leading (first) coefficients of the asymptotic expansions (4.14) and (4.3) (see [9, Theorem 2.3]): 1 (s) 1 Gκ (x1 , 0) V −1,m (x1 , 0, +1)σ −1(1) (x1 , 0, +1)K2j (x1 ), V−1 2π 1 (s) 1 (1) dsjm (x1 , −1) = − Gκ (x1 , 0) V −1,m (x1 , 0, −1)σ −1(1) (x1 , 0, −1) V−1 2π iπ(−1/2−∆j (x1 )) × K2j (x1 )e , (1) dsjm (x1 , +1) = 2 (s) (−1)m+1 Gκ (x1 , 0) V −1,m (x1 , 0, +1)σ −1 ∗ (2) (x1 , 0, +1)K1j (x1 ), 1 I+V0 2π 2 2 (s) (−1)m+1 (2) dsjm (x1 , −1) = Gκ (x1 , 0) V −1,m (x1 , 0, −1)σ −1 ∗ (2) (x1 , 0, −1) 1 I+V0 2π 2 (2) dsjm (x1 , +1) = SOLVABILITY AND ASYMPTOTICS OF SOLUTIONS 449 × K1j (x1 )eiπ(−1/2−∆j (x1 )) , j = 1, 2, s = 1, . . . , `0 , m = 0, . . . , ns − 1; here Gκ is the square root from the Gram determinant of the diffeomorphism κ and q (s) im+1 dns −1−m (q) (x , 0, ±1) = (τ − τs,±1 )ns V −1,m 1 m!(ns − 1 − m)! dτ ns −1−m £ ◦ (q) ¤−1 ¯¯ > −1 × M ((Jκ (x1 , 0)) (0, ±1, τ )) ¯ (q) , q = 1, 2. τ =τs,±1 (q) The coefficients cjm (x1 ) in (4.14) are defined as follows: ³1 ´ (1) (2) cjm (x1 ) = ajm (x1 )Ba−pr + ∆j (x1 ) c0 (x1 ), 2 j = 1, 2, m = 0, . . . , ns − 1, ³1 ´ (2) (1) cjm (x1 ) = ajm (x1 )Ba−pr + ∆j (x1 ) c0 (x1 ), 2 where ml m1 Ba−pr (t) = diag{B− (t), . . . , B− )t)}, ´ ³ ´ ³ iπ(t+1 1 mr ∂t Γ(t + 1)e 2 , B− (t) = B mr − 2πi r B mr (t) = kbm kp (t)kmr ×mr , ³ ´ iπ(t+1) 1 ´p−k (−1)p+k dp−k ³ 2 Γ(t + 1)e , k ≤ p, r bm 2πi (p − k)! dtp−k kp (t) = 0, k > p, p = 0, . . . , mr − 1, r = 1, . . . , `. Further, (j) (j) ajm (x1 ) = diag{am1 (λ1 ), . . . , am` (λ` )}, j = 1, 2, 3 1 i λ(1) argλr (x1 ) + log |λr (x1 )| + m, r (x1 ) = − − 2 2π 2π 1 i 3 λ(2) argλr (x1 ) + log |λr (x1 )| + m, r (x1 ) = − + 2 2π 2π m = 0, 1, . . . , ns − 1; mr (j) amr (λ(j) r ) = kakp (λr kmr ×mr , where p (j) r X (−1)p+k (2πi)l−p bm (µr ) kl , m = 0, k ≤ p, −i (j) p−l+1 mr (j) (λ + 1) r l=k akp (λr ) = (j) r m = 1, 2, . . . , ns − 1, k ≤ p, (−1)p+k bm kp (λr ), 0, k > p; (j) (j) (j) (1) here λr = −1 + m + µr , 0 < Reµr < 1, j = 1, 2, r = 1, . . . , `, and c0 (x1 ), (2) c0 (x1 ) are defined using the first coefficients of the asymptotic expansion of (1) (2) the functions −L−1 + ϕ0 and ϕ0 , respectively (see (4.3)). 450 O. CHKADUA (2) b) the asymptotics of solutions near the contact boundary ∂S0 : U (q) (x1 , x2 , x3 ) = (u(q) , ω (q) )(x1 , x2 , x3 ) ½ nX `0 2 X s −1 h XX e (q) (q) m = Re x3 dsjm (x1 , θ)(zz,θ )1/2+∆j (x1 )−m ϑ=±1 j=1 s=1 ×Ba0pr M +2 MX +2−l X + ³ m=0 ´i 1 (q) (q) − log zs,θ cjm (x1 ) 2πi ¾ (q) (q) 1/2+∆j (x1 )+k+p (q) (q) Bskmpj (x1 , log zs,ϑ ) xl2 xm 3 dslmpj (x1 , ϑ)(zs,ϑ ) k,l=0 p+m=0 k+l+p+m6=0 (q) +UM +1 (x1 , x2 , x3 ) for M > (q) (q) 2 − min{[s − 1], 0}, q = 1, 2; r (q) (4.15) (q) here zs,+1 = (−1)q [x2 + x3 τs,+1 ], zs,−1 = (−1)q+1 [x2 − x3 τs,−1 ]. The coefficients (q) (q) dsjm (·, ±1) are calculated similarly as in a) and the coefficients bjm are defined (q) as in a) by using the first coefficients b0 (q = 1, 2) of the asymptotic expansion (4.4). c) the asymptotics of solutions near the cuspidal edge ∂S0 : (ri U (2) )(x1 , x2 , x3 ) = ri (u(2) , ω (2) )(x1 , x2 , x3 ) `0 XX = ¾ ½ nX +2−l M +1 MX s −1 X j 1/2−j (i) l j −1/2+p+k (i) dslkjp (x1 ) x2 x3 zs,ϑ Re x3 zs,θ dsj (x1 , θ) + ϑ=±1 s=1 k,l=0 j+p=1 l+k+j+p6=1 j=0 (i) +UM +1 (x1 , x2 , x3 ) for M > (i) 2 − min{[s], 0}, i = 1, 2, r (i) (4.16) (i) with the coefficients dsj (·, ±1), dslkjp ∈ C0∞ (R) and the remainder UM +1 ∈ C0M +1 (R3± ), i = 1, 2; here (2) (2) zs,+1 = −x2 −x3 τs,+1 , zs,−1 = x2 −x3 τs,−1 , −π < Argzs,±1 < π, τs,±1 ∈ C0∞ (R), ◦ (2) 0 {τs,±1 }`s=1 are all different roots of the polynomial det M (2) ((Jκ> (x1 , 0))−1 · (0, ±1, τ )) of multiplicity ns , s = 1, . . . , `0 , in the complex lower half-plane. (i) The coefficients dsj (x1 , ±1) have the form (see [9, Theorem 2.3]): (1) 2 (s) dsj (x1 , +1) = Gκ (x1 , 0) V −1,j (x1 , 0, +1)σ −1 1 ∗(2) I+V0 2 (1) (x1 , 0, +1)c(j) (x1 ), 2 (s) dsj (x1 , −1) = −iGκ (x1 , 0) V −1,j (x1 , 0, −1)σ −1 1 ∗(2) I+V0 2 (2) (x1 , 0, −1)c(j) (x1 ) 2 (s) dsj (x1 , +1) = (−1)j+1 Gκ (x1 , 0) V −1,j (x1 , 0, +1)σ −1 1 ∗(3) I+V0 2 (2) (x1 , 0, +1)c(j) (x1 ), 2 (s) dsj (x1 , −1) = (−1)j+1 iGκ (x1 , 0) V −1,j (x1 , 0, −1)σ −1 1 ∗(3) I+V0 2 (x1 , 0, −1)c(j) (x1 ), SOLVABILITY AND ASYMPTOTICS OF SOLUTIONS 451 s = 1, . . . , l0 , j = 0, . . . , ns − 1, where Gκ is the square root from the Gram determinant of the diffeomorphism κ, 2 (s) ij+1 dns −1−j (2) (τ − τs,±1 )ns n −1−j s j!(ns − 1 − j)! dτ ¡ ◦ (2) ¢−1 ¯¯ × M ((Jκ> (x1 , 0))−1 · (0, ±1, τ )) ¯ (2) , V −1,j (x1 , 0, ±1) = 1´ ij ³ (j) c0 (x1 ) c (x1 ) = √ Γ j − 2 2 π τ =τs,±1 and c0 (x1 ) is the first coefficient of the asymptotic expansion in (4.6). 5. Investigation of the Properties of Exponents of the First Terms of an Asymptotic Expansion of Solutions of Problem M1 in the Neighbouhoods of Contact Boundaries We consider the properties of exponents of the first terms of an asymptotic expansion of solutions of the boundary-contact problem M1 in the neighbour(1) hood of the contact boundary ∂S0 . Analogous properties will be valid in the (2) neighbourhood of the contact boundary ∂S0 , too. For the sake of brevity, by γm and δm (m = 1, . . . , 6) we denote the real and the imaginary part of the first term exponent of the asymptotic expansions (4.14), and (4.15), i.e., ¯ 1 ¯¯ 1 1 ¯ γm (x1 ) = − ¯ argλm (x1 )¯, δm (x1 ) = − log |λm (x1 )|, 2 2π 2π (1) m = 1, . . . , 6, x1 ∈ ∂S0 , (m) where λm (x1 ) = λR (x1 ) m = 1, . . . , 12, are the eigenvalues of the matrix bR (x1 ) = (σR (x1 , 0, +1))−1 σR (x1 , 0, −1). Theorem 5.1. The real parts γm , m = 1, . . . , 6, of the exponents of the first terms of the asymptotic expansion of solutions of the boundary-contact problem (1) M1 near the contact boundary ∂S0 depend on the elastic constants and also on the geometry of the contact boundaries and may take any values from the interval ]0, 1/2[ , i.e., (a) if the elastic constants satisfy the limit conditions (2) aijlk (1) aijlk (2) → 0, then γm → 1/2 (m = 1, . . . , 6); (b) if for (2) aijlk → ∞, (1) aijlk bijlk (1) bijlk (2) → 0, (2) bijlk (1) bijlk cijlk (1) cijlk → 0, (2) → ∞, cijlk (1) cijlk →∞ 452 O. CHKADUA the limiting relations |ζ2 | |Imhζ1 , ζ2 i| |ζ1 | 1. M2+ + M1+ → 0; 2. → c (c > 0), |ζ2 | |ζ1 | |ζ1 | |ζ2 | are valid, then γm → 0 and δm → 0 (m = 1, . . . , 6), where ζ = (ζ1 , ζ2 ) is the eigenvector of the matrix bR , while M1+ and M2+ + + = are maximal eigenvalues of the matrices σN = σN1 (x1 , 0, +1) and σN 2 1 σN2 (x1 , 0, +1). (q) (q) (q) Here the limits of the coefficients aijlk , bijlk , cijlk are understood in the uniform sense with respect to the indices i, j, l, k. (1) (1) (1) Proof. Multiply the coefficients aijlk , bijlk , cijlk by α (α > 0), and the coefficients (2) (2) (2) aijlk , bijlk , cijlk by β (β > 0); i.e., we consider the differential equations with (1) (1) (1) (2) (2) (2) the elastic constants αaijlk , αbijlk , αcijlk and βaijlk , βbijlk , βcijlk . (a) Taking into consideration the estimate obtained in [6] we have ¯1 ¯ 1 ´ ³ √β 1 ¯ ¯ ¯ argλm (x1 )¯ ≤ arctg √ p + + , 2π π α m1 m2 + + −1 + where αm+ m2 are minimal eigenvalues of the matrices σN and σN , re1, β 1 2 spectively. Consequently, as β/α → 0 we get γm → 12 (m = 1, . . . , 6). (b) Since 1 1 2Imhζ1 , ζ2 i arg λm (x1 ) = arctg + + 2π 2π hσN2 ζ1 , ζ1 i + hσN ζ , ζ2 i 1 2 1 2Imhζ1 , ζ2 i + arctg − , − 2π hσN2 ζ1 , ζ1 i + hσN ζ , ζ2 i 1 2 we obtain the estimate 2|Imhζ1 ,ζ2 i| ¯1 ¯ 1 |ζ1 | |ζ2 | ¯ ¯ . ¯ arg λm (x1 )¯ ≥ arctg + |ζ1 | 2| −1 2π π β M2 |ζ2 | + αM1+ |ζ |ζ1 | (5.1) Further, since |λm (x1 )|2 = − − (hσN ζ , ζ1 i + hσN ζ , ζ2 i)2 + (2Imhζ1 , ζ2 i)2 2 1 1 2 , + + (hσN ζ , ζ1 i + hσN ζ , ζ2 i)2 + (2Imhζ1 , ζ2 i)2 2 1 1 2 we get 1 ≤ |λm (x1 )|2 ≤ b, b where (5.2) (β −1 M2+ |ζ1 |2 + αM1+ |ζ2 |2 )2 + (2Imhζ1 , ζ2 i)2 b= ; + 2 2 2 2 (β −1 m+ 2 |ζ1 | + αm1 |ζ2 | ) + (2Imhζ1 , ζ2 i) + + here αM1+ and β −1 M2+ are maximal eigenvalues of the matrices σN and σN 1 2 respectively, while the vectors ζ1 and ζ2 depend in general on α and β. SOLVABILITY AND ASYMPTOTICS OF SOLUTIONS 453 Thus, using the inequalities (5.1) and (5.2) and taking into consideration the limiting relations 1 and 2 from subsection (b), for β/α → ∞ we get γm → 0 and δm → 0 (m = 1, . . . , 6). ¤ Remark 5.2. In the centrally symmetric isotropic case the exponents of the first terms of the asymptotic expansion of solutions of the boundary-contact problem M1 γj + iδj (j = 1, . . . , 6) are calculated explicitly. Indeed, the differential operator of the couple-stress elasticity theory for a homogeneous isotropic centrally symmetric medium takes the form (see [26]): 1 2 (q) M (q) (∂x ) M (∂x ) M(q) (∂x ) = 3 , 4 (q) (q) M (∂x ) M (∂x ) 6×6 ° k (q) ° k (q) M (∂x ) = ° M ij (∂x )°3×3 , k = 1, 4, q = 1, 2, 1 (q) M ij (∂x ) = (µq + αq )δij ∆ + (λq + µq − αq )∂i ∂j , 3 X 2 (q) 3 (q) εijk ∂k , M ij (∂x ) = M ij (∂x ) = −2αq k=1 £ ¤ (q) M ij (∂x ) = δij (νq + βq )∆ − 4αq + (εq + νq − βq )∂i ∂j , 4 where δij and εijk are respectively the Kronecker and the Levi-Civita symbol. The coefficients λq , µq , αq , νq , βq , εq , q = 1, 2, are the elastic constants satisfying the conditions µq > 0, 3λq + 2µq > 0, αq > 0, νq > 0, 3εq + 2νq > 0, βq > 0. The stress operator of couple-stress elasticity is written as 1 2 (q) (q) N (∂z , n(z)) N (∂z , n(z)) , N (q) (∂z , n(z)) = 3 4 (q) (q) (∂ , n(z)) (∂ , n(z)) N N z z 6×6 ° ° k k (q) (q) N (∂z , n(z)) = ° N ij (∂z , n(z))°3×3 , k = 1, 4, q = 1, 2, 1 (q) ij (∂z , n(z)) N 2 (q) N ij (∂z , n(z)) = λq ni (z)∂j + (µq − αq )nj (z)∂i + (µq + αq )δij = −2αq 3 X 3 (q) ij (∂z , n(z)) εijk nk (z), N ∂ , ∂n(z) = 0, k=1 4 (q) N ij (∂z , n(z)) = εq ni (z)∂j + (νq − βq )ni (z)∂i + (νq + βq )δij ∂ . ∂n(z) ± The matrices σN = σNq (x1 , 0, ±1), q = 1, 2, have the following expressions q à 1 ! à 1 ! ± σ± σ 0 0 ± ± N1 N2 σN1 = , σN2 = , 2 2 ± σ± σ 0 0 N1 N 2 6×6 6×6 454 O. CHKADUA where 1 ± N1 σ µ1 + α 1 0 0 ν1 + β1 0 0 2 0 a1 ∓ib1 0 c1 ∓id1 = , σ± , N1 = 0 ±ib2 a1 0 ±id1 c1 3×3 3×3 1 0 0 µ2 + α2 a b 1 ± 2 2 0 ∓i σ N2 = , a22 − b22 a22 − b22 a2 b2 0 ±i 2 a2 − b22 a22 − b22 3×3 1 0 0 ν2 + β2 c2 d2 2 ± 0 ∓i σ N2 = , 2 2 2 2 c2 − d2 c2 − d2 d2 c2 0 ±i 2 c2 − d22 c22 − d22 3×3 here 2(λq + 2µq )(µq + αq ) aq = , λq + αq + 3µq 2(µq + αq )2 bq = , λq + αq + 3µq q = 1, 2, and 2(εq + 2νq )(νq + βq ) 2(νq + βq )2 cq = , dq = , q = 1, 2. εq + βq + 3νq εq + βq + 3νq Hence in the considered case the eigenvalues λj (j = 1, . . . , 12) of the matrix ¡ ¢−1 bR (x1 ) = σR (x1 , 0, +1) σR (x1 , 0, −1) are calculated in the manner as follows: λ1,2 λ3,4 λ5,6 λ7,8 λ9,10 p (µ1 + α1 ) − (µ2 + α2 ) ± 2i (µ1 + α1 )(µ2 + α2 ) = , µ1 + α1 + µ2 + α2 ( B±√B 2 −AC for B 2 − AC ≥ 0, A = √ B±i AC−B 2 for B 2 − AC < 0, A √ ( B± B 2 −AC for B 2 − AC ≥ 0, C = √ B±i AC−B 2 for B 2 − AC < 0, C p (ν1 + β1 ) − (ν2 + β2 ) ± 2i (ν1 + β1 )(ν2 + β2 ) , = ν1 + β1 + ν2 + β2 √ e e 2 −A eC e B B± e2 − A eC e ≥ 0, for B e A = √ e eC− e B e2 B±i A e2 − A eC e < 0, for B e A SOLVABILITY AND ASYMPTOTICS OF SOLUTIONS λ11,12 = √ e B± e B±i e 2 −A eC e B e C e2 − A eC e ≥ 0, for B eC− e B e2 A e C e2 − A eC e < 0, for B √ 455 where A = a2 a1 + b1 b2 + a22 − b22 + a2 b1 + b2 a1 , B = a2 a1 − b1 b2 − a22 + b22 , C = a2 a1 + b1 b2 + a22 − b22 − a2 b1 − b2 a1 , e = c2 c1 + d1 d2 + c22 − d22 + c2 d1 + d2 c1 , A e = c2 c1 − d1 d2 − c22 + d22 , B e = c2 c1 + d1 d2 + c2 − d2 − c2 d1 − d2 c1 . C 2 2 Hence we obtain √ 1 1 µ2 + α2 γ1 = − arctg √ , δ1 = 0, 2 π µ1 + α1 1 , B 2 − AC ≥ 0, 2 √ 1 1 AC−B 2 − arctg , B 2 − AC < 0, γ2,3 = 2 2π B √ 1 arctg AC−B 2 B 2 − AC < 0, 2π |B| √ 2 1 − 2π log B± BA −AC , B 2 − AC ≥ 0, ³ ´ δ2,3 = ∓ 1 log A , B 2 − AC < 0, 2π C √ 1 1 ν2 + β2 γ4 = − arctg √ , δ4 = 0, 2 π ν1 + β1 1 e2 − A eC e ≥ 0, , B 2 √ ee B e2 1 1 e2 − A eC e < 0, − 2π arctg AC− , B γ5,6 = e 2 B √ ee B e2 1 e2 − A eC e < 0, arctg A|C− , B e 2π B| √ e e 2 −A eC e B − 1 log B± e2 − A eC e ≥ 0, , B e 2π A ³ ´ δ5,6 = ∓ 1 log Ae , e2 − A eC e < 0. B e 2π C B > 0, B < 0, e > 0, B e < 0, B e2 − A eC e 6= 0, then B 0 (t) = I, i.e., the Remark 5.3. If B 2 − AC 6= 0 and B apr first terms of the asymptotic expansion of solutions of the boundary-contact (i) problem M1 contain no logarithms near the contact boundary ∂S0 , i = 1, 2. Remark 5.4. Note that in the isotropic case we obtain more exact limit relations (for the elasticity case see [6]): a) if B > 0, B 2 − AC ≥ 0 and µ2 + α2 → 0 or µ1 → ∞, then γ1 → 1/2, γ2 = γ3 = 1/2; b) if B > 0, B 2 − AC < 0 and µ1 → ∞, then γm → 1/2, m = 1, 2, 3; c) if B > 0, B 2 − AC < 0 and µ2 + α2 → 0, then γ1 → 1/2, γ2,3 → 1/4; d) if B < 0 and µ1 + α1 → 0 or µ2 → ∞, then γm → 0, m = 1, 2, 3; 456 O. CHKADUA e) if |µ1 − µ2 | → 0, |α1 − α2 | → 0, then γm → 1/4, m = 1, 2, 3, and e > 0, B e2 − A eC e ≥ 0 and ν2 + β2 → 0 or ν1 → ∞, then γ4 → 1/2, ã) if B γ5,6 = 1/2; e > 0, B e2 − A eC e < 0 and ν1 → ∞, then γm → 1/2, m = 4, 5, 6; b̃) if B e > 0, B e2 − A eC e < 0 and ν2 + β2 → 0, then γ4 → 1/2, γ5,6 → 1/4; c̃) if B e < 0, ν1 + β1 → 0 or ν2 → ∞, then γm → 0, m = 4, 5, 6; d̃) if B ẽ) if |ν1 − ν2 | → 0, |β1 − β2 | → 0, then γm → 1/4, m = 4, 5, 6. It is not difficult to see that if the conditions e2 − A eC e>0 B 2 − AC > 0 and B hold, then the exponent of the first term of the asymptotics of solutions of the boundary-contact problem M1 has the form n √µ + α √ν + β o 1 1 2 2 2 2 γ = − arctg max √ ,√ . 2 π µ1 + α 1 ν 1 + β1 Since γ < 12 , γ2,3 = γ5,6 = 12 , the oscillation of solutions vanishes in some neighbourhood of the contact boundary and therefore solutions describe the real physical process. Note that this class has been found only in the spatial case since in the plane case it is known that the oscillation does not vanish. In the general case we have found a class of anisotropic bodies when the oscillation in the asymptotic expansion vanishes and the real parts γj j = 1, . . . , 6, of exponents of the first terms of the asymptotic expansion are calculated by simpler formulas. + Let αj > 0 and βj > 0, j = 1, . . . , 6, be the eigenvalues of the matrices σN 1 + and σN , respectively. 2 Theorem If the¶conditions µ 5.5. + σN1 − αj I 1) rank < 6, j = 1, . . . , 6; + σN − βj I 2 (j) (j) 2) hζ1 , ζ 1 i 6= 0, j = 1, . . . , 6, (j) are satisfied, where ζ1 (j = 1, . . . , 6) is the common eigenvector of the matrices + + σN1 and σN2 , which correspond to the eigenvalues αj and βj , then the oscillation participanting in the asymptotic expansion of solutions of the boundary-contact problem M1 vanishes, i.e., δj = 0, j = 1, . . . , 6, and the real parts of the exponents of the first terms of the asymptotic expansion are calculated by a simpler formula 1 1 1 , j = 1, . . . , 6. γj = − arctg p 2 π αj βj Proof. From the condition 1) we obtain the existence of the common eigenvec(j) + + tors ζ1 , j = 1, . . . , 6, for the matrices σN and σN , i.e., 1 2 (j) (j) + σN ζ = αj ζ1 , 1 1 (j) (j) + σN ζ = βj ζ1 , j = 1, . . . , 6. 2 1 (5.3) SOLVABILITY AND ASYMPTOTICS OF SOLUTIONS 457 − + − Since σ + N1 = σN1 and σ N2 = σN2 , we have (j) (j) (j) − ζ = αj ζ 1 , σN 1 1 (j) − σN ζ = βj ζ 1 , j = 1, . . . , 6. 2 1 (5.4) Further, we look for an eigenvector ζ (j) of the matrix ¡ ¢−1 σR (x1 , 0, +1) σR (x1 , 0, −1) (j) (j) in the form ζ (j) = (ζ1 , γ ej ζ1 ). We obtain ¡ ¢−1 σR (x1 , 0, +1) σR (x1 , 0, −1)ζ (j) = λj ζ (j) . Taking into consideration the expression of the matrices σR (x1 , 0, ±1), we get ( (j) (j) + (j) − (j) ej ζ1 = λj σN ζ + λj γ ej ζ1 , σN ζ −γ 2 1 2 1 (5.5) (j) (j) − (j) + (j) ej σN = −λ ζ + λ γ e σ . ζ1 + γ ζ ζ j j j 1 1 1 N1 1 (j) Now, by the scalar multiplication of both equations (5.5) by the vector ζ we have ( (j) (j) (j) (j) (j) (j) + (j) − (j) ej hζ1 , ζ 1 i = λj hσN ζ , ζ 1 i + λj γ ej hζ1 , ζ 1 i, hσN ζ , ζ1 i − γ 2 1 2 1 (5.6) (j) (j) (j) (j) (j) (j) − (j) + (j) hζ1 , ζ 1 i + γ ej hσN ζ , ζ i = −λ hζ , ζ i + λ γ e hσ ζ , ζ i. j j j 1 1 1 1 1 1 N 1 1 Substituting (5.3) and (5.4) into (5.6), we get (j) (j) (j) (j) (j) (j) βj hζ (j) , ζ (j) i − γ ej hζ1 , ζ 1 i = λj βj hζ1 , ζ 1 i + λj γ ej hζ1 , ζ 1 i, 1 1 (j) (j) (j) (j) (j) (j) hζ (j) , ζ (j) i + γ e α hζ , ζ i = −λ hζ , ζ i + λ γ e α hζ , ζ i. 1 1 j j 1 1 j 1 1 (j) (j) j j j 1 1 Taking into account the condition 2), i.e., hζ1 , ζ 1 i 6= 0, we obtain the following system of equations with respect to λj and γ ej (j = 1, . . . , 6): ( βj − γ ej = λj βj + λj γ ej , 1+γ ej αj = −λj + λj γ ej αj . Further, p λj = p αj βj ∓ i αj βj ± i s , γ ej = ±i βj , j = 1, . . . , 6. αj Clearly, |λj | = 1, j = 1, . . . , 6. Hence δj = 0, j = 1, . . . , 6. Calculating γj (j = 1, . . . , 6), we get γj = 1 1 1 − arctg p , j = 1, . . . , 6. 2 π αj βj ¤ 458 O. CHKADUA If under the conditions of Theorem 5.5 αi βi 6= αj βj for all i 6= j, i, j = 1, . . . , 6, then Ba0pr (t) = I, i.e., the first terms of the asymptotic expansions (4.14), (4.15) contain no logarithms. Remark 5.6. Let us assume that in the neighborhood of the boundary the (1) (2) contact surfaces S0 and S0 are parallel to the isotropic plane. We consider (q) (q) (q) the case, where the coefficients bijlk = 0 and, instead of aijlk and cijlk , we have (q) the elastic constants of transversally-isotropic elastic bodies, i.e., instead of aijlk we have (q) (q) (q) (q) (q) a11 , a33 , a13 , a55 , a66 , q = 1, 2, (q) and, instead of cijlk , we have (q) (q) (q) (q) (q) c11 , c33 , c13 , c55 , c66 , q = 1, 2, which satisfy the following conditions: (q) (q) (q) (q) (q) (q) (q) (q) a11 − a66 > 0, a55 > 0, a66 > 0, c11 − c66 > 0, c55 > 0, c66 > 0, (q) (q) (a13 )2 (c13 )2 (q) (q) , q = 1, 2, , q = 1, 2. a33 > (q) c33 > (q) (q) (q) a11 − a66 c11 − c66 It is not difficult to see that in this case the eigenvalues αj and βj , j = 1, . . . , 6, + + of the matrices σN and σN are calculated explicitly. 1 2 (q) (q) (q) (q) Let a11 6= a33 and c11 6= c33 (q = 1, 2); then the conditions 1)–2) of Theorem 5.5 have the form e1 − D e1 e1 C1 − D1 C B B1 = , = , e2 − C e2 e2 D2 − C2 B2 D B where ¶ µq µq q q ¶ (q) (q) (q) (q) (q) (q) (q) (q) a11 a33 − a13 a11 a55 a2 + a3 a55 1 q q q q Bq = , Cq = , 2 (q) (q) (q) (q) (q) (q) a55 + a11 a33 a55 + a11 a33 µq q q q ¶ (q) (q) (q) (q) (q) a33 a11 a55 a2 + a3 1 q q Dq = , q = 1, 2, 2 (q) (q) (q) a55 + a11 a33 (q) (q) a2 and a3 are the roots of the equation (see [26]) £ (q) (q) (q) (q) (q) (q) (q) (q) (q) ¤ a11 a55 a2 + (a13 + a55 )2 − a11 a33 − (a55 )2 a + a33 a55 = 0, and µq eq = B (q) c55 ¶ q (q) c11 (q) c33 µq (q) c13 − q q (q) (q) (q) c55 + c11 c33 , eq = 1 C 2 q = 1, 2, q ¶ (q) + c3 q q , (q) (q) (q) c55 + c11 c33 (q) (q) c11 c55 (q) c2 SOLVABILITY AND ASYMPTOTICS OF SOLUTIONS q (q) eq = 1 D 2 (q) 459 µq q ¶ (q) (q) (q) (q) c11 c55 c2 + c3 q q , q = 1, 2, (q) (q) (q) c55 + c11 c33 q c33 (q) where c2 and c3 are the roots of the equation £ (q) (q) (q) (q) (q) (q) (q) ¤ (q) (q) c11 c55 c2 + (c13 + c55 )2 − c11 c33 − (c55 )2 c + c33 c55 = 0, q = 1, 2. (q) (q) (q) (q) If we assume that a13 = −a55 and c13 = −c55 , q = 1, 2, then the conditions 1)–2) of Theorem 5.5 can be rewritten in a simpler form: q q q q q q (1) (1) (1) (1) (1) (1) a11 − a33 a55 c11 − c33 c55 q q q =q and q =q . (2) (2) (2) (2) (2) (2) a33 − a11 a55 c33 − c11 c55 Now let us consider the case, where the domains Dq , q = 1, 2, are filled with the same material. The following theorem is fulfilled (for analogous results in the case of elasticity theory see [7], [8]; see also [14]). Theorem 5.7. If the domains Dq , q = 1, 2, are filled with the same material, then the asymptotic expansion of solutions of the boundary-contact problem M1 (1) near the contact boundary ∂S0 takes the form U (q) (x1 , x2 , x3 ) = (u(q) , ω (q) )(x1 , x2 , x3 ) ½ nX l0 2 X s −1 h i XX 1 (q) (q) +∆j (x1 )−m (q) 4 Re xm d (x , θ)(z ) = cjm (x1 ) 3 sjm 1 s,theta ϑ=±1 j=1 s=1 + +2−l M +2 MX X m=0 o (q) (q) 1 +∆j (x1 )+k+p (q) (q) 4 xl2 xm d (x , ϑ)(z ) B (x , log z ) 1 1 3 slmpj s,ϑ skmpj s,ϑ k,l=0 p+m=0 k+l+p+m6=0 (q) +UM +1 (x1 , x2 , x3 ) for M > (q) (q) 2 − min{[s − 1], 1}, q = 1, 2, r (5.7) (q) with the coefficients dsjm (· , ±1), cjm , dslmpj (· , ±1) ∈ C0∞ (R) and the remainder (q) 3 UM +1 ∈ C0M +1 (R± ), q = 1, 2, (q) (q) (q) (q) (q) zs,+1 = (−1)q (x2 + x3 τs,+1 ), zs,−1 = (−1)q+1 (x2 − x3 τs,−1 ), −π < Argτs,±1 < π. In this case the parameters ∆j are calculated by ¡ (j) ¢ (j) ∆j (x1 ) = δ1 (x1 ), . . . , δ6 (x1 ) , j = 1, 2, where i 1 i (2) log |λk (x1 )|, δk (x1 ) = − log |λk (x1 )|, k = 1, . . . , 6. 2π 2 2π Proof. When the domains Dq , q = 1, 2, are filled with the same material, we can show, like in [7], that the eigenvalues λk (x1 ), k = 1, . . . , 12, of the matrix ¡ ¢−1 (1) σR (x1 , 0, +1) σR (x1 , 0, −1), x2 ∈ ∂S0 , (1) δk (x1 ) = − 460 O. CHKADUA are calculated by means of the eigenvalues βk (x1 ), k = 1, 2, 3, of the matrix σ ∗ , V0 i.e., s 1 − 2βk (x1 ) if k = 1, . . . , 6, i 1 + 2β (x ) , k 1 s λk (x1 ) = 1 − 2βk−6 (x1 ) , if k = 7, . . . , 12, −i 1 + 2βk−6 (x1 ) where −1/2 < βk < 1/2, k = 1, . . . , 6. Hence the exponent of the first term of the asymptotic expansion of solutions has the form 1 1 i ∓ − log |λk (x1 )|, k = 1, . . . , 6. ¤ 2 4 2π Note that in this case the asymptotic expansion has step equal to one half (see [8]). The same asymptotic expansion can also be obtained near the contact bound(2) ary ∂S0 . 6. Solvability and Asymptotics of Solutions of the Boundary-Contact Wedge-Type Problem M2 We will formulate theorems of the uniqueness and existence of solutions of problem M2 . Assume that the following compatibility conditions are fulfilled on the curves (1) (2) ∂S0 , ∂S0 ∂S0 : (i) (i) (2) (1) (3) (2) s−1 s−1 ∃Φ0 , i = 1, 2, 3, Φ0 ∈ Bs−1 p,r (∂D1 ), Φ0 ∈ Bp,r (∂D2 ), Φ0 ∈ Bp,r (∂D2 ), (1) (2) e s−1 (S (1) ), h1 − πS (1) Φ0 + πS (1) Φ0 ∈ B 0 p,r 0 0 (1) (3) e s−1 (S (2) ) h2 − πS (2) Φ0 + πS (2) Φ0 ∈ B 0 p,r 0 0 (2) (3) s−1 e πS0 Φ0 + πS0 Φ0 ∈ Bp,r (S0 ) (1) (6.1) (2) s−1 s−1 hold for ϕ1 ∈ Bs−1 p,r (∂D1 ), ϕ2 ∈ Bp,r (∂D2 ), ϕ3 ∈ Bp,r (∂D2 ), 1 < p < ∞, 1 ≤ r ≤ ∞, 1/p − 1/2 < s < 1/p + 1/2. (i) Here Φ0 , i = 1, 2, 3, are some fixed extensions of the functions ϕi , i = 1, 2, 3, (1) (2) on ∂D1 , ∂D2 , ∂D2 , respectively. Theorem 6.1. Let 4/3 < p < 4 and the compatibility conditions (6.1) be (1) (2) fulfilled on ∂S0 , ∂S0 ∂S0 for s = 1 − 1/p. Then the boundary-contact wedgetype problem M2 has solutions of the classes Wp1 (Dq ), q = 1, 2, if and only if the condition Z Z Z ϕ3 ([a × z] + b, a) ds ϕ2 ([a × z] + b, a) ds − ϕ1 ([a × z] + b, a) ds − (2) (1) S2 S1 S2 Z Z h2 ([a × z] + b, a) ds = 0 h1 ([a × z] + b, a) ds + + (1) S0 (2) S0 SOLVABILITY AND ASYMPTOTICS OF SOLUTIONS 461 is fulfilled, where a and b are arbitrary three-dimensional constant vectors. Solutions of the boundary-contact problem M2 are given by the potential-type functions (1) (1) (1) (2) (2) (2) (3) (3) (3) (1) (1) (1) (2) (2) (2) (2) (3) (3) (3) (3) U (1) = V(1) (V−1 )−1 (B2M +1 )−1 ϕ0 + V(1) (V−1 )−1 (B2M +1 )−1 ψ0 + R1 , (6.2) r1 U (2) = V(2) (V−1 )−1 (B2M +1 )−1 ϕ0 + V−1 (V−1 )−1 (B2M +1 )−1 ψ0 + R2 , (6.3) r2 U (2) = V(3) (V−1 )−1 (B2M +1 )−1 ϕ0 + V−1 (V−1 )−1 (B2M +1 )−1 ψ0 + R3 , (6.4) (1) (2) with R1 ∈ C M +1 (D1 ), R2 ∈ C M +1 (D2 ), R3 ∈ C M +1 (D2 ), while the functions (1) (2) (1) (3) (2) (3) (ϕ0 , ϕ0 ), (ψ0 , ϕ0 ) and (ψ0 , ψ0 ) are solutions of the systems ( ( ( (2) (1) (1) πS0 N3 ψ0 = Ψ3 , πS (1) N1 ϕ0 = Ψ1 , πS (2) N2 ψ0 = Ψ2 , 0 0 (2) ϕ0 = (1) ϕ0 + G1 , (3) ϕ0 (1) = ψ0 + G2 , (3) (2) ψ0 = −ψ0 + G3 , respectively. Here (1) (2) (1) (3) N1 = (B2M +1 )−1 − (B2M +1 )−1 , N2 = (B2M +1 )−1 − (B2M +1 )−1 , (2) (3) N3 = −(B2M +1 )−1 − (B2M +1 )−1 , where ³ 1 ∗ (1) ´ (1) (1) (1) B2M +1 = −(V−1 )2M +1 + − I+ V 0 (V−1 )−1 , 2 ³1 ∗ (i) ´ (i) (i) (i) I+ V 0 (V−1 )−1 , i = 2, 3. B2M +1 = (V−1 )2M +1 + 2 The operators Nj , j = 1, 2, 3, are positive-definite ΨDOs. Let ¡ (1) ¢ (1) (`) (`) ∆Nj (x1 ) = δj (x1 ), . . . , δj (x1 ), . . . , δj (x1 ), . . . , δj (x1 ) , | {z } | {z } m1 -times m` -times i (k) (k) δj (x1 ) = − log |λNj (x1 )|, j = 1, 2, 3, k = 1, . . . , ` 2π (generally speaking, mk and ` depend on j), (k) where λNj , j = 1, 2, k = 1, . . . , `, are the eigenvalues of the matrix ¡ ¢−1 bNj (x1 ) = σNj (x1 , 0, +1 σNj (x1 , 0, −1) of multiplicity mk , k = 1, . . . , `, 6 P k=1 mk = 6; here mk , k = 1, . . . , `, and ` (k) depend on j, and the eigenvalues of the matrix bN3 = I are trivial, λN3 = 1, k = 1, . . . , 6. Note that the boundary data of the problem M2 are assumed to be suf(∞,S+2M ),∞ (∞,S+2M ),∞ (1) ficiently smooth, i.e., ϕ1 ∈ Hr (S1 ), ϕ2 ∈ Hr (S2 ), ϕ3 ∈ (∞,S+2M ),∞ (2) (∞,S+2M +1),∞ (i) (∞,S+2M ),∞ (i) Hr (S2 ), fi ∈ Hr (S0 ), hi ∈ Hr (S0 ), i = 1, 2, 1 − 12 < s < 1r + 21 . r 462 O. CHKADUA Recalling that solutions of the problem M2 are represented by potential-type functions (see (6.2), (6.3), (6.4)) and using the asymptotic expansion of such functions (see [9, Theorems 2.2 and 2.3]), we obtain the following asymptotic expansions of problem M2 in terms of some local coordinate systems of curves (1) (2) ∂S0 , ∂S0 , ∂S0 : (j) a) Asymptotic expansion near the contact boundaries ∂S0 , j = 1, 2: U (q) (x1 , x2 , x3 ) = (u(q) , ω (q) )(x1 , x2 , x3 ) ½ nX `0 s −1 h i XX (q) (q) 1 −j+i∆(x1 ) (j) j 2 x3 dsj (x1 , θ)(zs,θ ) = Re c (x1 ) ϑ=±1 s=1 + j=0 M +1 MX +2−l X ¾ − 1 +i∆(x1 )+p+k (q) (q) (q) xl2 xj3 dsljp (x1 , ϑ)zs,ϑ2 Bskjp (x1 , zs,ϑ ) l,k=0 p+j=1 k+l+p+j6=1 (q) UM +1 (x1 , x2 , x3 ) for M > 2 − min{[s], 0}, q = 1, 2, r (6.5) 3 (q) with UM +1 ∈ C0M +1 (R± ), q = 1, 2. Here ∆ = ∆Nj , j = 1, 2, and £ £ (q) (q) ¤ (q) (q) ¤ zs,+1 = (−1)q x2 + x3 τs,+1 , zs,−1 = (−1)q x2 − x3 τs,−1 , −π < Argzs,±1 < π. (q) The first coefficients dsj (·, ±1), c(j) are calculated as in(4.14) (see [9]). b) Asymptotic expansions near the cuspidal edge ∂S0 (see (4.16)): ri U (2) (x1 , x2 , x3 ) = ri (u(2) , ω (2) )(x1 , x2 , x3 ) = `0 XX Re ½ nX s −1 ϑ=±1 s=1 1 −j (i) 2 xj3 zs,θ dsj (x1 , θ) + +2−l M +1 MX X ¾ − 1 +p+k (i) xl2 xj3 zs,ϑ2 dslkjp (x1 ) l,k=0 j+p=1 l+k+j+p6=1 j=0 (i) +UM +1 (x1 , x2 , x3 ) for M > 2 − min{[s], 0}, i = 1, 2, r (6.6) 3 (i) with UM +1 ∈ C0M +1 (R± ), i = 1, 2. Here (2) (2) zs,+1 = −x2 − x3 τs,+1 , zs,−1 = −x2 − x3 τs,−1 , −π < Argzs,±1 < π. (i) The coefficients dsj (x1 , ±1) are calculated as in (4.16). Remark 6.2. Note that if we consider the boundary-contact problem M2 for the nonhomogeneous equations M(q) (∂x )U (q) + F (q) = 0 in Dq , q = 1, 2, and use the boundary and boundary-contact data ϕ1 = 0, ϕ2 = 0, ϕ3 = 0, fi = hi = 0, i = 1, 2, (1) then the compatibility conditions (6.1) are fulfilled automatically on ∂S0 , (2) ∂S0 , ∂S0 . SOLVABILITY AND ASYMPTOTICS OF SOLUTIONS 463 Remark 6.3. It is easy to see that if the matrices bNj , j = 1, 2, are unitary (b∗Nj = b−1 Nj ), then the oscillation in the asymptotic expansion (42) vanishes, i.e., ∆j = 0, j = 1, 2. In the case of transversally isotropic bodies we obtain the following necessary and sufficient conditions under which the oscillation vanishes: e1 D e1 − B e2 e1 B1 4C1 D1 − B12 B 4C 1 = = . (6.7) , e2 e2 D e2 − B e22 B2 4C2 D2 − B22 B 4C For the centrally symmetric isotropic case condition (6.7) tales the form λ1 + µ1 − α1 = λ2 + µ2 − α2 , ε1 + ν1 − β1 = ε2 + ν2 − β2 . (6.8) Condition (6.8) written in terms of the Poisson constants was found in [16], when investigating the asymptotic properties of solutions of boundary positivedefinite pseudodifferential equations of crack-type problems of elasticity (see also [38]). References 1. E. Aero and E. Kuvshinski, Fundamental equations of the theory of elastic media with rotationally interacting particles. (Russian) Fiz. Tverd. Tela 2(1960), 1399–1409; English transl.:Soviet Physics Solid State 2(1960), 1272–1281. 2. M. Agranovich, Elliptic singular integrodifferential operators. (Russian) Uspekhi Mat. Nauk 20 (1965), No. 5, 3–120. 3. J. Bennish, Asymptotics for elliptic boundary value problems for systems of pseudodifferential equations. J. Math. Anal. Appl. 179(1993), 417–445. 4. T. Buchukuri and T. Gegelia, On uniqueness of solutions of the principal problems of elasticity theory for infinity domains. (Russian) Diffrentsial’nye Uravneniya 25(1989), No. 9, 1556–1565. 5. T. Burchuladze and T. Gegelia, Development of the method of a potential in the theory of elasticity. (Russian) Metsniereba, Tbilisi, 1985. 6. O. Chkadua, Some boundary-contact problems of elasticity theory with mixed boundary conditions outside the contact surface. Math. Nachr. 188(1997), 23–48. 7. O. Chkadua, The Dirichlet, Neumann and mixed boundary value problems of the theory of elasticity in n-dimensional domains with boundaries containing closed cuspidul edges. Math. Nachr. 189(1998), 61–105. 8. O. Chkadua and R. Duduchava, Asymptotics of solutions to some boundary value problems of elasticity for bodies with cuspidal edges. Mem. Differential Equations Math. Phys. 15(1998), 29–58. 9. O. Chkadua and R. Duduchava, Asymptotics of functions represented by potentials. Russian J. Math. Phys. 7(2000), No. 1, 15–47. 10. O. Chkadua and R. Duduchava, Pseudodifferential equations on manifolds with boundary: Fredholm property and asymptotics. Math. Nachr. 222(2001), 79–139. 11. E. Cosserat and F. Cosserat, Sur les equations de la theorie de l’elasticite. C.R. Acad. Sci. Paris, 1898, No. 126, 1129–1132. 12. M. Costabel, M. Dauge, and R. Duduchava, Asymptotics without logarithmic terms for crack problems. Preprint 01–50, 2001, University of Rennes 1, France. 464 O. CHKADUA 13. M. Dauge, Elliptic boundary value problems in corner domains. Smoothness and Asymptotics of Solutions, Lecture Notes in Math. 1341, Springer-Verlag, Berlin, 1988. 14. R. Duduchava and D. Natroshvili, Mixed crack type problem in anisotropic elasticity. Math. Nachr. 191(1998), 83–107. 15. R. Duduchava, D. Natroshvili, and E. Shargorodsky, Boundary value problems of the mathematical theory of cracks. Trudy Inst. Prikl. Mat. I. N. Vekua, 1990, 63–84. 16. R. Duduchava, A. Sändig, and W. Wendland, Interface cracks in anisotropic composites. Math. Methods Appl. Sci. 22(1999), No. 16, 1413–1446. 17. Yu. Egorov, Linear differential equations of the principal type. (Russian) Nauka, Moscow, 1984. 18. G. Eskin, Boundary value problems for elliptic pseudodifferential equations. (Translated from Russian) —em Translations of Mathematical Monographs, 52. American Mathematical Society, Providence, R.I., 1981. 19. R. Gachechiladze and O. Maisaia, A study of some fundamental contact boundary problems of statics of elasticity theory and the moment theory of elasticity. (Russian) Tbiliss. Gos. Univ. Inst. Prikl. Mat. Trudy 5/6(1978), 68–92. 20. F. Gantmakher, Theory of matrices. (Russian) Nauka, Moscow, 1967. 21. G. Grioli, Mathematical theory of elastic equilibrium. Ergebnisse der angewandten Mathematik, Heft 7. Academic Press, Inc., Publishers, New York; Springer-Verlag, Berlin-Gttingen-Heidelberg, 1962. 22. P. Grisvard, Elliptic Problems in Non-smooth Domains. Pitman, London, Boston, 1985. 23. R. V. Kapanadze, On some properties of singular operators in normed spaces. (Russian) Trudy Tbilis. Gos. Univ. 129 (1968), 17–26. 24. W. Koiter, Couple-stresses in the theory of elasticity. I, II. Nederl. Akad. Wetensch. Proc. Ser. B 67(1964), 17-29; ibid. 67(1964), 30–44. 25. V. Kondrat’ev, Boundary problems for elliptic equations in domains with conical or angular points. (Russian) Trudy Moskov. Mat. Obshch. 16(1967), 209–292. 26. V. Kupradze, T. Gegelia, M. Basheleishvili, and T. Burchuladze Threedimensional problems of the mathematical theory of elasticity and thermoelasticity. Translated from the second Russian edition. Edited by V. D. Kupradze. North-Holland Series in Applied Mathematics and Mechanics, 25. North-Holland Publishing Co., Amsterdam-New York, 1979. 27. V. Maz’ya, Boundary integral equations. Analysis, IV, 127–222, Encyclopaedia Math. Sci., 27, Springer, Berlin, 1991. 28. V. Maz’ya and B. Plamenevskiı̆, Elliptic boundary value problems in a domain with a piecewise smooth boundary. (Russian) Proceedings of the Symposium on Continuum Mechanics and Related Problems of Analysis (Tbilisi, 1971), Vol. 1 (Russian), 171–181. Mecniereba, Tbilisi, 1973. 29. V. Maz’ya and B. Plamenevskiı̆, Lp -estimates of solutions in domains with edges. (Russian) Trudy Moskov. Mat. Obshch. 37(1978), 49–93, 270. 30. V. Maz’ya and B. Plamenevskiı̆, The first boundary value problem for classical equations of mathematical physics in domains with piecewise smooth boundaries. I, II. Z. Anal. Anwendungen 2(1983), No. 4, 335–359; 2(1983), No. 6, 523–551. 31. R. Mindlin, Influence of counpe-stress on stress concentrations. Experiment. Mech. 1963, No. 3, 1–7. 32. D. Natroshvili, O. Chkadua, and E. Shargorodsky, Mixed problems for homogeneous anisotropic elastic media. (Russian) Tbiliss. Gos. Univ. Inst. Prikl. Mat. Trudy 1990, 133–181. SOLVABILITY AND ASYMPTOTICS OF SOLUTIONS 465 33. S. Nazarov and B. Plamenevskiı̆, Elliptic problems in domains with piecewisesmooth boundaries. de Gruyter Expositions in Mathematics, 13. Walter de Gruyter & Co., Berlin, 1994; Russian original: Nauka, Moscow, 1991. 34. W. Nowacki, Teoria niesymetryczney sprezystosci. PWN, Warszawa, 1971. 35. E. Shargorodsky, Boundary value problems for elliptic pseudodifferential operators: the halfspace case. Proc. A. Razmadze Math. Inst. 99(1993), 44–80. 36. E. Shargorodsky, Boundary value problems for elliptic pseudodifferential operators on a manifolds. Proc. A. Razmadze Math. Inst. 104(1993), 108–132. 37. E. Shargorodsky, An Lp -analogue of the Vishik–Eskin theory. Mem. Differential Equations Math. Phys. 2(1994), 41–146. 38. T. C. T. Ting, Explixit solution and invariance of the singularities at an interface crack in anisotropic components. Internat. J. Solids Structures 22(1986), No. 9, 965–983. 39. H. Triebel, Interpolation theory, functional spaces, differential operators. NorthHolland, Amsterdam, 1968. 40. W. Voigt, Theoretische studien über Elasticitätsyerhälnisse der Kristable. Abn. der Königl, Ges., Wiss., Göttingen, 1887, No. 34. (Received 27.05.2003) Author’s address: A. Razmadze Mathematical Institute Georgian Academy of Sciences 1, M. Aleksidze St., Tbilisi 0193 Georgia E-mail: chkadua@rmi.acnet.ge