Determinant Formulas for Scattering Matrices of Schrödinger Operators with Finitely Many Concentric $δ$-Shells

Determinan t F orm ulas for Scattering Matrices of Sc hr¨ odinger Op erators with Finitely Man y Concen tric δ -Shells Masahiro Kaminaga Keyw ords: δ -in teraction; scattering matrix; resolv en t formula; partial- w av e decomp osition Mathematical Sub ject Classification (2020): 35J10, 35P25, 47A40, 81U20 Abstract W e study stationary scattering for Sc hr¨ odinger op erators in R 3 with finitely many concen tric δ –shell in teractions of constant real strengths. Starting from the self–adjoint realization and the b oundary resolv ent form ula for this mo del, w e sho w that, after partial–wa v e reduction, the same finite-dimensional boundary matrices that arise in the resolven t form ula also determine the channel scattering co efficien ts. More pre- cisely , for each angular momentum ℓ , the channel co efficien t S ℓ ( k ) sat- isfies S ℓ ( k ) = det K ℓ ( k 2 − i 0) / det K ℓ ( k 2 + i 0) for almost every k > 0, where K ℓ ( z ) = I N + m ℓ ( z )Θ is the ℓ –th reduced b oundary matrix. Th us, in eac h c hannel, the p ositiv e–energy scattering problem is re- duced to a finite-dimensional matrix problem, and the scattering phase is reco vered from det K ℓ ( k 2 + i 0). W e then study the first non trivial case of tw o concentric shells in the s –wa ve channel, where the interaction b et ween the shells pro duces non trivial threshold effects. W e derive an explicit formula for S 0 ( k ) and analyze its behavior as k ↓ 0. In the regular threshold regime, we obtain an explicit scattering length. W e further identify a threshold–critical configuration characterized by the existence of a nontrivial zero–energy radial solution, regular at the origin, whose exterior constan t term v anishes. In the corresp onding nondegenerate exceptional case, the usual finite scattering length breaks down, and instead S 0 ( k ) → − 1 as k ↓ 0. 1 In tro duction Sc hr¨ odinger op erators with singular in teractions are a classical source of explicit mo dels in sp ectral and scattering theory , as discussed in [1] and in the w ork of Brasche, Exner, Kuperin, and ˇ Seba [4]. Among them, concentric 1 δ –shell interactions provide a natural radial mo del. Since the interaction is supported on concentric spheres, the op erator is rotationally symmetric and the scattering problem admits a partial–w av e decomp osition. At the same time, when tw o or more shells are presen t, the interaction among the shells pro duces nontrivial scattering and threshold effects. This makes the mo del simple enough for explicit calculation, but still ric h enough to show in teresting phenomena. In this pap er we study stationary scattering for H = − ∆ + N X j =1 α j δ ( | x | − R j ) in L 2 ( R 3 ) , (1.1) where 0 < R 1 < · · · < R N and α 1 , . . . , α N ∈ R , and we compare H with the free op erator H 0 = − ∆. F or finitely many concen tric spherical δ –shells, Shabani [10] studied the mo del by self–adjoint extension metho ds and reduced it to radial one– dimensional equations with matc hing conditions in each partial wa v e. On the scattering side, Hounkonnou, Hounkp e, and Shabani [5] studied scat- tering theory for finitely many sphere interactions supp orted by concentric spheres. Related sp ectral and scattering prop erties for radially symmetric p enetrable wall mo dels were studied by Ikebe and Shimada [6]. These p en- etrable wall mo dels ma y b e view ed as a regular counterpart of spherical shell interactions. Th us, in the earlier literature, the scattering problem is treated after partial–wa v e reduction by solving radial equations and imp os- ing matc hing conditions at the shells in each c hannel. In particular, the existence of a partial–w av e description for this mo del is not new. Our starting p oin t is the b oundary resolven t framework obtained in [7]. F or finitely many concentric shells, that pap er giv es a self–adjoint realization of H and a resolv ent form ula in terms of a b oundary op erator. This frame- w ork is closely related to the general theory of hypersurface δ –interactions; see, for example, [2]. It is also related to abstract Kre ˘ ın–t yp e res olv en t form ulas for singular p erturbations; see [8]. In addition, the construction in [7] allows the surface strengths α j to b e b ounded real–v alued functions on the shells and do es not assume that they are constant. In the presen t pap er w e do not rep eat that construction. Instead, we restrict ourselves to the rotationally symmetric case of constant shell strengths and study the p ositiv e–energy scattering problem from the b oundary resolven t formula. F or the conv enience of the reader, in Section 2 w e recall only the precise input from [7] that is needed b elo w, namely the quadratic–form realization of H , the in terface description of D ( H ), the b oundary resolv ent form ula, the trace–class prop erty of the resolv ent difference, and the c haracterization of the p oin t sp ectrum by the b oundary matrix K N ( z ). The main p oin t of the present pap er is that, in the rotationally sym- metric case, the same reduced b oundary matrices that arise in the resolv ent 2 form ula also determine the channel scattering co efficien ts. F or each angular momen tum ℓ ≥ 0, let S ℓ ( k ) denote the channel scattering co efficien t. Our main result is the determinan t formula S ℓ ( k ) = det K ℓ ( k 2 − i 0) det K ℓ ( k 2 + i 0) , for almost ev ery k > 0 . (1.2) Here K ℓ ( z ) is the ℓ –th reduced b oundary matrix. The full b oundary op erator has the form K N ( z ) = I + m ( z )Θ , Θ = diag( α 1 R 2 1 , . . . , α N R 2 N ) , where m ( z ) is the b oundary op erator asso ciated with the free Green function and I is the iden tit y operator on the corresp onding b oundary space. Its ℓ –th partial–w av e comp onen t is K N ,ℓ ( z ) = I N + m ℓ ( z )Θ , where I N denotes the N × N iden tity matrix. T o simplify notation, w e write K ℓ ( z ) for K N ,ℓ ( z ). Thus (1.2) is not merely an explicit c hannel form ula. It sho ws that the p ositiv e–energy scattering data are enco ded by exactly the same reduced b oundary matrices that app ear in the resolven t form ula. T o the best of our kno wledge, the determinant represen tation (1.2), written explicitly in terms of the reduced b oundary matrix arising from the resolv en t form ula, do es not app ear in the earlier partial–wa v e literature on concentric δ –shells. As an application of this framework, w e study the first nontrivial case of t wo concen tric shells. In the s –wa v e c hannel w e deriv e an explicit form ula for the scattering co efficien t and analyze its low–energy b ehavior as k ↓ 0. In the regular threshold regime, the phase shift admits the standard expansion δ 0 ( k ) = − a s k + o ( k ) ( k ↓ 0) , (1.3) whic h defines the scattering length a s . W e obtain an explicit formula for a s and sho w that it agrees with the coefficient in the asymptotics of the zero–energy radial solution, normalized so that, for | x | > R 2 , u ( x ) = 1 − a | x | , where a = a s . W e also iden tify a threshold–critical configuration c harac- terized b y the existence of a nontrivial zero–energy radial solution which is regular at the origin and whose exterior constant term v anishes. In the corresp onding nondegenerate exceptional case, the usual finite scattering length do es not exist and one has S 0 ( k ) → − 1 ( k ↓ 0) . 3 This gives a concrete zero–energy interpretation of the threshold anomaly in the double–shell mo del. The pap er is organized as follows. In Section 2 we rec all the quadratic form realization and the resolv ent form ula from [7]. In Section 3 we pro ve the determinan t formula for the c hannel scattering matrices. In Section 4 w e sp ecialize to the double δ –shell case and deriv e an explicit form ula for the s – w av e scattering co efficien t. In Section 5 w e analyze the lo w–energy b eha vior, including the regular threshold regime and the nondegenerate exceptional threshold regime, and giv e a zero–energy interpretation of the latter. 2 The mo del and the resolv en t formula In this section w e recall, in the present concentric-shell setting, the precise ingredien ts from [7] that will b e used later. W e do not rep eat the pro ofs. 2.1 The op erator and its quadratic form F or j = 1 , . . . , N , let S j = { x ∈ R 3 : | x | = R j } . W e consider the Schr¨ odinger op erator H = − ∆ + N X j =1 α j δ ( | x | − R j ) in L 2 ( R 3 ), where α 1 , . . . , α N ∈ R . Schr¨ odinger operators with δ –interactions supp orted on hypersurfaces are well studied; see, for example, [1, 4, 2]. F or abstract Kre ˘ ın–t yp e resolven t form ulas for singular p erturbations, see also [8]. F or finitely many concentric spherical shells, a self–adjoin t realization and a b oundary integral resolven t formula were obtained in [7]. W e define the sesquilinear form h on the form domain D [ h ] = H 1 ( R 3 ) b y h [ u, v ] = Z R 3 ∇ u · ∇ v dx + N X j =1 α j Z S j uv dσ j , u, v ∈ H 1 ( R 3 ) . (2.1) Here D [ h ] denotes the form domain of h , and dσ j denotes the surface mea- sure on S j . Since eac h S j is a smo oth compact h yp ersurface, the trace map H 1 ( R 3 ) → L 2 ( S j ) is b ounded. Hence, by the trace inequality , the b oundary terms are form b ounded with relative b ound zero with resp ect to the Dirichlet form, and th us h is closed and low er semib ounded. 4 2.2 Boundary op erators and the resolv en t form ula Let H 0 = − ∆ denote the free Hamiltonian. Let z ∈ C \ [0 , ∞ ) and c ho ose √ z so that Im √ z > 0. W e write R 0 ( z ) = ( H 0 − z ) − 1 , G z ( x, y ) = e i √ z | x − y | 4 π | x − y | . Here R 0 ( z ) is the free resolv ent and G z ( x, y ) is the corresponding free Green function. F or each j = 1 , . . . , N , let τ j : H 1 ( R 3 ) → L 2 ( S 2 ) denote the trace on S j transp orted to the unit sphere by the parametrization x = R j ω , that is, ( τ j u )( ω ) = u ( R j ω ) , ω ∈ S 2 . Th us τ j u is the trace of u on S j , viewed as a function on S 2 . Accordingly , all surface integrals defining the b oundary op erators b elo w are written with resp ect to dω on S 2 rather than dσ j on S j . W e define the single–lay er op erators (Γ j ( z ) φ )( x ) = Z S 2 G z ( x, R j ω ) φ ( ω ) dω , j = 1 , . . . , N , and define Γ( z ) : N M j =1 L 2 ( S 2 ) → L 2 ( R 3 ) b y Γ( z )( φ 1 , . . . , φ N ) = N X j =1 Γ j ( z ) φ j . W e also introduce the op erator matrix m ( z ) =  m ij ( z )  N i,j =1 on L N j =1 L 2 ( S 2 ), where each en try m ij ( z ) is the op erator on L 2 ( S 2 ) giv en b y ( m ij ( z ) φ )( ω ) = Z S 2 G z ( R i ω , R j ω ′ ) φ ( ω ′ ) dω ′ . F ollowing [7], we set Θ = diag( α 1 R 2 1 , . . . , α N R 2 N ) , K N ( z ) = I + m ( z )Θ , (2.2) where I denotes the iden tit y op erator on L 2 ( S 2 ) ⊕ · · · ⊕ L 2 ( S 2 ). Thus K N ( z ) is the b oundary op erator matrix asso ciated with the shell interaction. The next theorem collects the only facts from [7] that will be used in the presen t pap er. 5 Theorem 2.1. The quadr atic form h in (2.1) is close d and lower semi- b ounde d on H 1 ( R 3 ) , and ther efor e defines a self–adjoint op er ator H in L 2 ( R 3 ) . Mor e over, a function u ∈ L 2 ( R 3 ) b elongs to the op er ator domain D ( H ) if and only if u is pie c ewise H 2 on the r e gions sep ar ate d by the shel ls, is c ontinuous acr oss e ach spher e r = R j , and satisfies ∂ r u ( R j + 0 , ω ) − ∂ r u ( R j − 0 , ω ) = α j u ( R j , ω ) , j = 1 , . . . , N , ω ∈ S 2 . (2.3) Her e ∂ r u ( R j ± 0 , ω ) denotes the r adial derivative taken fr om the exterior and interior sides of the spher e r = R j . L et z ∈ C \ [0 , ∞ ) . If K N ( z ) is invertible, then ( H − z ) − 1 = ( H 0 − z ) − 1 − Γ( z )Θ K N ( z ) − 1 Γ( z ) ∗ . (2.4) Her e Γ( z ) ∗ denotes the adjoint of Γ( z ) . Mor e over, ( H − z ) − 1 − ( H 0 − z ) − 1 is tr ac e class. Final ly, for z ∈ C \ [0 , ∞ ) , the op er ator K N ( z ) is noninvertible if and only if z ∈ σ p ( H ) , wher e σ p ( H ) denotes the p oint sp e ctrum of H . Theorem 2.1 is a sp ecialization of the self–adjoin t realization, the b ound- ary resolven t formula, and the sp ectral characterization of the b oundary op erator prov ed in [7]. In particular, since H is self–adjoin t, i / ∈ σ p ( H ), and hence K N ( i ) is in vertible. The trace–class resolven t difference in Theorem 2.1 will b e used in Section 3 to obtain the existence and completeness of the w av e op erators. 2.3 P artial–w av e reduction Since α 1 , . . . , α N are real constan ts, b oth H and the b oundary op erator m ( z ) are rotationally symmetric. Accordingly , on each cop y of L 2 ( S 2 ), the op era- tor m ( z ) is diagonal with resp ect to the spherical harmonic decomp osition, and the same is true for the direct sum L 2 ( S 2 ) ⊕ · · · ⊕ L 2 ( S 2 ) . Let { Y ℓm } ℓ ≥ 0 , − ℓ ≤ m ≤ ℓ b e an orthonormal basis of spherical harmonics on S 2 . F or each ℓ ≥ 0, the restriction of m ( z ) to the ℓ –th spherical harmonic sector is describ ed b y an N × N complex matrix, whic h we denote b y m ℓ ( z ). W e denote by K N ,ℓ ( z ) the ℓ –th partial–wa ve comp onent of the b oundary op erator matrix K N ( z ) introduced in (2.2). F or simplicity , we write K ℓ ( z ) for K N ,ℓ ( z ) in the rest of the pap er. 6 Lemma 2.2. L et z ∈ C \ [0 , ∞ ) and set k = √ z with Im k > 0 . Then m ℓ ( z ) is the N × N matrix whose entries ar e given by ( m ℓ ( z )) ij = ik j ℓ  k min { R i , R j }  h (1) ℓ  k max { R i , R j }  , 1 ≤ i, j ≤ N , (2.5) wher e j ℓ = j ℓ is the spheric al Bessel function and h (1) ℓ = h (1) ℓ is the outgoing spheric al Hankel function. We also denote by h (2) ℓ the inc oming spheric al Hankel function. In p articular, ( m ℓ ( z )) ij = ( m ℓ ( z )) j i . Mor e over, the ℓ –th p artial–wave c omp onent of K N ( z ) is K ℓ ( z ) = I N + m ℓ ( z )Θ , (2.6) wher e I N denotes the N × N identity matrix. Pr o of. Let r < := min {| x | , | y |} , r > := max {| x | , | y |} , and write b x = x | x | , b y = y | y | . The spherical harmonic expansion of the free Green function reads G z ( x, y ) = ik ∞ X n =0 n X q = − n j n ( k r < ) h (1) n ( k r > ) Y nq ( b x ) Y nq ( b y ) . Hence, for φ = Y ℓm , w e obtain ( m ij ( z ) Y ℓm )( ω ) = Z S 2 G z ( R i ω , R j ω ′ ) Y ℓm ( ω ′ ) dω ′ = ik j ℓ  k min { R i , R j }  h (1) ℓ  k max { R i , R j }  Y ℓm ( ω ) , b y orthonormality of the spherical harmonics. This prov es (2.5). Since Θ = diag ( α 1 R 2 1 , . . . , α N R 2 N ) acts only on the shell index and do es not mix spherical harmonics, the restriction of K N ( z ) = I + m ( z )Θ to the ℓ –th partial w av e is exactly K ℓ ( z ) = I N + m ℓ ( z )Θ . This completes the pro of. Th us, in each angular momentum channel, the b oundary op erator matrix K N ( z ) reduces to the finite matrix K ℓ ( z ). 7 3 Scattering matrices for finitely man y concen tric shells 3.1 F ree sp ectral represen tation and partial–w a ve decomp o- sition T o describ e scattering, w e work in the standard sp ectral represe n tation of the free Hamiltonian H 0 = − ∆; see, for example, [9, Ch. XI] and [11]. A t a fixed energy k 2 > 0, let u in b e a fixed free solution of ( − ∆ − k 2 ) u in = 0 in R 3 . W e call a solution u of ( H − k 2 ) u = 0 an outgoing solution with inciden t part u in if u in is a free solution of ( − ∆ − k 2 ) u in = 0 and u − u in satisfies the Sommerfeld radiation condition lim r →∞ r ( ∂ r − ik )( u − u in )( x ) = 0 . In this case, u in is called the incident wa v e. Here r = | x | and ∂ r denotes the radial deriv ativ e. In the present partial–w av e setting, we will take the inciden t part to b e j ℓ ( k | x | ) Y ℓm ( b x ) , where b x = x/ | x | . Since the spherical harmonics form a complete system on S 2 and the problem is rotationally in v arian t, it suffices to consider incident w av es of this form. The corresp onding outgoing term will b e describ ed b y outgoing spherical Hank el functions. Let b f denote the F ourier transform of f ∈ L 2 ( R 3 ), b f ( ξ ) = (2 π ) − 3 / 2 Z R 3 e − ix · ξ f ( x ) dx. W e define ( F 0 f )( k , ω ) = b f ( k ω ) , k > 0 , ω ∈ S 2 , initially for sufficiently regular f . Then, b y Plancherel’s theorem and the spherical c hange of v ariables, F 0 extends to a unitary op erator from L 2 ( R 3 ) on to L 2  (0 , ∞ ) , k 2 dk ; L 2 ( S 2 , dω )  , where dω denotes the standard surface measure on the unit sphere S 2 . In this representation, H 0 = − ∆ is diagonalized as multiplication by k 2 , that is, ( F 0 H 0 f )( k , ω ) = k 2 ( F 0 f )( k , ω ) , f ∈ D ( H 0 ) . 8 F or each k > 0, the fib er space is L 2 ( S 2 ), and the spherical harmonic decomp osition gives L 2 ( S 2 ) = ∞ M ℓ =0 H ℓ , H ℓ = span { Y ℓm : − ℓ ≤ m ≤ ℓ } . Equiv alently , for each fixed k > 0, ( F 0 f )( k , ω ) = ∞ X ℓ =0 ℓ X m = − ℓ ( F 0 f )( k , ℓ, m ) Y ℓm ( ω ) , where ( F 0 f )( k , ℓ, m ) = Z S 2 ( F 0 f )( k , ω ) Y ℓm ( ω ) dω . Since the shell strengths α 1 , . . . , α N are constan ts, b oth H and H 0 com- m ute with the natural unitary action of the rotation group on L 2 ( R 3 ). As a consequence, the scattering op erator admits a partial–wa v e decomp osition in the free sp ectral representation. The next theorem records this partial–wa v e decomp osition. Theorem 3.1. The wave op er ators W ± ( H , H 0 ) = s-lim t →±∞ e itH e − itH 0 exist and ar e c omplete. The sc attering op er ator S = W + ( H , H 0 ) ∗ W − ( H , H 0 ) is unitary on L 2 ( R 3 ) . In the fr e e sp e ctr al r epr esentation F 0 , ther e exists a me asur able family of unitary op er ators S ( k 2 ) on L 2 ( S 2 ) such that ( F 0 S F − 1 0 g )( k , ω ) = S ( k 2 ) g ( k , ω ) for almost every k > 0 . Mor e over, for e ach ℓ ≥ 0 ther e exists a sc alar function S ℓ ( k ) , define d for almost every k > 0 , such that ( F 0 S F − 1 0 g )( k , ℓ, m ) = S ℓ ( k ) g ( k, ℓ, m ) . (3.1) Equivalently, S ( k 2 ) ↾ H ℓ = S ℓ ( k ) I H ℓ for almost every k > 0 . Pr o of. By Theorem 2.1, ( H − i ) − 1 − ( H 0 − i ) − 1 9 is trace class. Hence the wa v e op erators exist and are complete b y the Birman–Kuro da theorem; see [9, Ch. XI]. Since H 0 = − ∆ has purely abso- lutely con tinuous sp ectrum, the scattering op erator S = W + ( H , H 0 ) ∗ W − ( H , H 0 ) is unitary on L 2 ( R 3 ). Since both H and H 0 comm ute with the action of the rotation group, the w av e op erators and hence S comm ute with rotations. Moreov er, the inter- t wining prop ert y H W ± ( H , H 0 ) = W ± ( H , H 0 ) H 0 implies that S commutes with H 0 . Therefore, in the free sp ectral represen tation F 0 , the op erator F 0 S F − 1 0 acts fib erwise with resp ect to the energy parameter k 2 . Thus there exists a measurable family of unitary op erators S ( k 2 ) on L 2 ( S 2 ) suc h that ( F 0 S F − 1 0 g )( k , ω ) = S ( k 2 ) g ( k , ω ) for almost ev ery k > 0. Since S comm utes with rotations, each fib er op erator S ( k 2 ) comm utes with the rotation action on L 2 ( S 2 ) for almost every k > 0. Bec ause eac h spherical harmonic subspace H ℓ is irreducible under rotations, Sc h ur’s lemma implies that S ( k 2 ) ↾ H ℓ = S ℓ ( k ) I H ℓ for almost ev ery k > 0. This is equiv alen t to (3.1). 3.2 Outgoing solutions in a fixed partial w av e W e now fix a partial wa v e and construct the corresp onding outgoing solu- tions. Lemma 3.2. F or e ach ℓ ≥ 0 and e ach k > 0 , the b oundary values m ℓ ( k 2 ± i 0) , K ℓ ( k 2 ± i 0) exist. Mor e over, if m ± ℓ ( k ) := m ℓ ( k 2 ± i 0) , K ± ℓ ( k ) := K ℓ ( k 2 ± i 0) , then m − ℓ ( k ) = m + ℓ ( k ) ∗ , and det K − ℓ ( k ) = det K + ℓ ( k ) . Pr o of. By Lemma 2.2, the entries of m ℓ ( z ) are giv en explicitly in terms of spherical Bessel and Hankel functions. These expressions admit b oundary 10 v alues on (0 , ∞ ), so that m ℓ ( k 2 ± i 0) exist for every k > 0, and hence so do K ℓ ( k 2 ± i 0). F or the upp er b oundary v alue, Lemma 2.2 gives ( m + ℓ ( k )) ij = ik j ℓ  k min { R i , R j }  h (1) ℓ  k max { R i , R j }  . F or the low er b oundary v alue, we approac h the cut (0 , ∞ ) from the lo wer half-plane while keeping the branc h determined by Im √ z > 0 on C \ [0 , ∞ ). Hence p k 2 − i 0 = − k , p k 2 + i 0 = k , for k > 0. Using j ℓ ( − t ) = ( − 1) ℓ j ℓ ( t ) , h (1) ℓ ( − t ) = ( − 1) ℓ h (2) ℓ ( t ) , t > 0 , w e obtain ( m − ℓ ( k )) ij = − ik j ℓ  k min { R i , R j }  h (2) ℓ  k max { R i , R j }  . Since j ℓ ( t ) is real for t > 0 and h (1) ℓ ( t ) = h (2) ℓ ( t ) , t > 0 , it follo ws that m − ℓ ( k ) = m + ℓ ( k ) ∗ . By definition, K ± ℓ ( k ) = I N + m ± ℓ ( k )Θ . Since Θ = Θ ∗ is a real diagonal matrix, Sylvester’s determinan t identit y det( I + AB ) = det( I + B A ) yields det K − ℓ ( k ) = det  I N + m − ℓ ( k )Θ  = det  I N + Θ m − ℓ ( k )  = det  I N + Θ m + ℓ ( k ) ∗  = det  ( I N + m + ℓ ( k )Θ) ∗  = det K + ℓ ( k ) . This pro ves the claim. W e construct the outgoing solution in the ( ℓ, m ) c hannel corresp onding to a giv en incident wa v e. Lemma 3.3. Fix ℓ ≥ 0 , fix m with − ℓ ≤ m ≤ ℓ , and let k > 0 satisfy det K ℓ ( k 2 + i 0)  = 0 . 11 Write K + ℓ ( k ) := K ℓ ( k 2 + i 0) , b ℓ ( k ) = t  j ℓ ( k R 1 ) , . . . , j ℓ ( k R N )  , and define c ℓ ( k ) := K + ℓ ( k ) − 1 b ℓ ( k ) ∈ C N . Then ther e exists an outgoing solution in the ( ℓ, m ) channel whose incident p art is j ℓ ( k | x | ) Y ℓm ( b x ) . F or | x | = r > R N , this solution has the form u + ℓm ( x, k ) = 1 2  h (2) ℓ ( k r ) + σ ℓ ( k ) h (1) ℓ ( k r )  Y ℓm ( b x ) , (3.2) wher e σ ℓ ( k ) = 1 − 2 ik t b ℓ ( k )Θ K + ℓ ( k ) − 1 b ℓ ( k ) . (3.3) Pr o of. By assumption, the matrix K + ℓ ( k ) = K ℓ ( k 2 + i 0) is in vertible, so c ℓ ( k ) is well defined. Let m + ℓ ( k ) := m ℓ ( k 2 + i 0) , and let G + k ( x, y ) := e ik | x − y | 4 π | x − y | . F or j = 1 , . . . , N , define Φ j ( x, k ) := Z S 2 G + k ( x, R j ω ′ ) Y ℓm ( ω ′ ) dω ′ . Equiv alently , Φ j ( x, k ) = R − 2 j Z S j G + k ( x, y ) Y ℓm ( y /R j ) dσ j ( y ) , where dσ j denotes the surface measure on S j . Th us Φ j ( · , k ) is smo oth on R 3 \ S j , satisfies ( − ∆ − k 2 )Φ j ( · , k ) = 0 in R 3 \ S j , and is con tinuous across each sphere. F or i  = j , the function Φ j ( · , k ) is smo oth across S i , hence ∂ r Φ j ( R i + 0 , ω , k ) − ∂ r Φ j ( R i − 0 , ω , k ) = 0 . 12 F or i = j , using the explicit form ulas for Φ j inside and outside S j , which will b e derived b elo w in (3.6) and (3.7), we obtain, at r = R j , ∂ r Φ j ( R j + 0 , ω , k ) − ∂ r Φ j ( R j − 0 , ω , k ) = ik 2  j ℓ ( k R j )  h (1) ℓ  ′ ( k R j ) − h (1) ℓ ( k R j ) j ′ ℓ ( k R j )  Y ℓm ( ω ) . Using the W ronskian identit y j ℓ ( t )  h (1) ℓ  ′ ( t ) − j ′ ℓ ( t ) h (1) ℓ ( t ) = i t 2 , w e obtain ∂ r Φ j ( R j + 0 , ω , k ) − ∂ r Φ j ( R j − 0 , ω , k ) = − R − 2 j Y ℓm ( ω ) . Therefore ∂ r Φ j ( R i + 0 , ω , k ) − ∂ r Φ j ( R i − 0 , ω , k ) = − δ ij R − 2 i Y ℓm ( ω ) . (3.4) W e next record tw o formulas for Φ j . First, ev aluating Φ j on S i and using Lemma 2.2, we obtain Φ j ( R i ω , k ) = ( m + ℓ ( k )) ij Y ℓm ( ω ) , 1 ≤ i, j ≤ N . (3.5) Second, for | x | = r > R j , the standard partial-wa v e expansion of the free Green function yields G + k ( x, R j ω ′ ) = ik ∞ X n =0 n X q = − n j n ( k R j ) h (1) n ( k r ) Y nq ( b x ) Y nq ( ω ′ ) . Multiplying b y Y ℓm ( ω ′ ) and integrating ov er S 2 , orthonormalit y of the spher- ical harmonics giv es Φ j ( x, k ) = ik j ℓ ( k R j ) h (1) ℓ ( k r ) Y ℓm ( b x ) , | x | = r > R j . (3.6) Similarly , for | x | = r < R j , the same partial-w av e expansion giv es Φ j ( x, k ) = ik h (1) ℓ ( k R j ) j ℓ ( k r ) Y ℓm ( b x ) , | x | = r < R j . (3.7) Hence Φ j ( · , k ) lies in the fixed ( ℓ, m ) channel on each annulus. No w set u + ℓm ( x, k ) = j ℓ ( k | x | ) Y ℓm ( b x ) − N X j =1 θ j c ℓ,j ( k ) Φ j ( x, k ) , (3.8) where θ j = α j R 2 j , j = 1 , . . . , N , 13 so that Θ = diag( θ 1 , . . . , θ N ) . By (3.6) and (3.7), the function u + ℓm ( · , k ) also lies in the fixed ( ℓ, m ) channel on eac h annulus. Since j ℓ ( k | x | ) Y ℓm ( b x ) is regular at the origin and each Φ j ( · , k ) is smo oth near the origin, the function u + ℓm ( · , k ) is regular at x = 0. Moreov er, ( − ∆ − k 2 ) u + ℓm ( · , k ) = 0 in R 3 \ N [ j =1 S j . W e v erify the interface conditions. Ev aluating (3.8) on S i and using (3.5), w e get u + ℓm ( R i ω , k ) =  j ℓ ( k R i ) − N X j =1 ( m + ℓ ( k )Θ) ij c ℓ,j ( k )  Y ℓm ( ω ) . Since K + ℓ ( k ) c ℓ ( k ) = b ℓ ( k ) , this b ecomes u + ℓm ( R i ω , k ) = c ℓ,i ( k ) Y ℓm ( ω ) . On the other hand, (3.4) gives ∂ r u + ℓm ( R i + 0 , ω , k ) − ∂ r u + ℓm ( R i − 0 , ω , k ) = − N X j =1 θ j c ℓ,j ( k )  ∂ r Φ j ( R i + 0 , ω , k ) − ∂ r Φ j ( R i − 0 , ω , k )  = θ i c ℓ,i ( k ) R − 2 i Y ℓm ( ω ) = α i c ℓ,i ( k ) Y ℓm ( ω ) = α i u + ℓm ( R i ω , k ) . Th us u + ℓm ( · , k ) satisfies the transmission conditions for the δ -shell interaction and solv es ( H − k 2 ) u + ℓm ( · , k ) = 0 in the sense of distribution. W e next determine its exterior form. F or | x | = r > R N , (3.6) yields u + ℓm ( x, k ) =  j ℓ ( k r ) − ik h (1) ℓ ( k r ) t b ℓ ( k )Θ c ℓ ( k )  Y ℓm ( b x ) . Using j ℓ ( k r ) = 1 2  h (1) ℓ ( k r ) + h (2) ℓ ( k r )  , w e obtain u + ℓm ( x, k ) = 1 2  h (2) ℓ ( k r ) + σ ℓ ( k ) h (1) ℓ ( k r )  Y ℓm ( b x ) , r > R N , 14 where σ ℓ ( k ) = 1 − 2 ik t b ℓ ( k )Θ c ℓ ( k ) . Since c ℓ ( k ) = K + ℓ ( k ) − 1 b ℓ ( k ) , this is exactly (3.3). By construction, the exterior part in volv es only h (1) ℓ ( k r ), hence the solution is outgoing in the sense of the Sommerfeld radiation con- dition. W e show that the outgoing solution in the ( ℓ, m ) channel is unique. Lemma 3.4. Fix ℓ ≥ 0 , fix m with − ℓ ≤ m ≤ ℓ , and let k > 0 . L et w b e a function in the ( ℓ, m ) channel such that: (i) w is r e gular at the origin, (ii) ( − ∆ − k 2 ) w = 0 in R 3 \ N [ j =1 S j , (iii) w is c ontinuous acr oss e ach spher e S j and satisfies ∂ r w ( R j + 0 , ω ) − ∂ r w ( R j − 0 , ω ) = α j w ( R j , ω ) , j = 1 , . . . , N , (iv) w is outgoing and has zer o incident p art, that is, for r > R N one has w ( x ) = γ h (1) ℓ ( k r ) Y ℓm ( b x ) for some γ ∈ C . Then w ≡ 0 on R 3 . Pr o of. By assumption, for r > R N , w ( x ) = γ h (1) ℓ ( k r ) Y ℓm ( b x ) for some γ ∈ C . W e recall that, as r → ∞ , h (1) ℓ ( k r ) = ( − i ) ℓ +1 e ikr k r  1 + O ( r − 1 )  , so that h (1) ℓ represen ts an outgoing spherical wa v e. 15 W e show that γ = 0. Fix r > R N and decomp ose the ball {| x | < r } in to the annuli determined by the shells. Applying Green’s iden tity on each ann ulus and summing ov er all ann uli, we obtain Z | x | = r w ∂ r w dσ = Z | x | R N . Since w lies in the fixed ( ℓ, m ) channel, we ma y write w ( x ) = g ( r ) r Y ℓm ( b x ) on eac h interv al ( R j , R j +1 ), where R 0 := 0 , R N +1 := ∞ . Then g satisfies g ′′ ( r ) +  k 2 − ℓ ( ℓ + 1) r 2  g ( r ) = 0 16 on eac h such interv al. F rom w = 0 on ( R N , ∞ ) w e obtain g ( R N + 0) = 0 , g ′ ( R N + 0) = 0 . Since w is contin uous across r = R N , w e also hav e g ( R N − 0) = g ( R N + 0) = 0 . Moreo ver, the jump condition ∂ r w ( R N + 0) − ∂ r w ( R N − 0) = α N w ( R N ) implies g ′ ( R N + 0) − g ′ ( R N − 0) = α N g ( R N ) = 0 , and hence g ′ ( R N − 0) = 0 . Therefore the Cauc hy data of g v anish at R N on ( R N − 1 , R N ), so uniqueness for the ODE implies g ≡ 0 there. Rep eating the same argument across the shells, w e conclude that w ≡ 0 on R 3 . Corollary 3.5. Fix ℓ ≥ 0 , fix m with − ℓ ≤ m ≤ ℓ , and let k > 0 . Then ther e exists at most one outgoing solution in the ( ℓ, m ) channel whose incident p art is j ℓ ( k | x | ) Y ℓm ( b x ) . Pr o of. Supp ose that u and e u are t wo outgoing solutions in the ( ℓ, m ) c hannel with the same inciden t part j ℓ ( k | x | ) Y ℓm ( b x ) . Then w := u − e u is regular at the origin, satisfies ( − ∆ − k 2 ) w = 0 in R 3 \ N [ j =1 S j , ob eys the same in terface conditions, and has zero incident part. Hence w satisfies all assumptions of Lemma 3.4. Therefore w ≡ 0 , and so u = e u. Th us the outgoing solution is unique. 17 Prop osition 3.6. F or e ach ℓ ≥ 0 and e ach k > 0 , the matrix K ℓ ( k 2 + i 0) is invertible. Equivalently, Z ℓ := { k > 0 : det K ℓ ( k 2 + i 0) = 0 } = ∅ . Pr o of. Fix ℓ ≥ 0, fix k > 0, and choose m with − ℓ ≤ m ≤ ℓ . W rite Y := Y ℓm . Assume, for contradiction, that K ℓ ( k 2 + i 0) is not inv ertible. Then there exists a nonzero v ector c = t ( c 1 , . . . , c N ) ∈ C N suc h that K ℓ ( k 2 + i 0) c = 0 . Let G + k ( x, y ) := e ik | x − y | 4 π | x − y | , and for j = 1 , . . . , N define Φ j ( x, k ) := Z S 2 G + k ( x, R j ω ′ ) Y ( ω ′ ) dω ′ . As in the pro of of Lemma 3.3, each Φ j ( · , k ) lies in the fixed ( ℓ, m ) c hannel, is regular at the origin, satisfies ( − ∆ − k 2 )Φ j ( · , k ) = 0 in R 3 \ S j , is con tinuous across each sphere, and ob eys Φ j ( R i ω , k ) = ( m + ℓ ( k )) ij Y ( ω ) , 1 ≤ i, j ≤ N , together with ∂ r Φ j ( R i + 0 , ω , k ) − ∂ r Φ j ( R i − 0 , ω , k ) = − δ ij R − 2 i Y ( ω ) . Moreo ver, for r > R N one has Φ j ( x, k ) = ik j ℓ ( k R j ) h (1) ℓ ( k r ) Y ( b x ) , | x | = r > R N , since R j < R N < r . No w define u ( x ) = − N X j =1 (Θ c ) j Φ j ( x, k ) . 18 Then u lies in the fixed ( ℓ, m ) channel, is regular at the origin, satisfies ( − ∆ − k 2 ) u = 0 in R 3 \ N [ j =1 S j , and is outgoing with zero inciden t part, b ecause for r > R N it is a linear com bination of h (1) ℓ ( k r ) Y ( b x ) only . W e chec k the interface conditions. F or 1 ≤ i ≤ N , u ( R i ω ) = − N X j =1 ( m + ℓ ( k )) ij (Θ c ) j Y ( ω ) = − ( m + ℓ ( k )Θ c ) i Y ( ω ) . Since K ℓ ( k 2 + i 0) c = ( I N + m + ℓ ( k )Θ) c = 0 , w e hav e m + ℓ ( k )Θ c = − c, and therefore u ( R i ω ) = c i Y ( ω ) . Similarly , ∂ r u ( R i + 0 , ω ) − ∂ r u ( R i − 0 , ω ) = − N X j =1 (Θ c ) j  ∂ r Φ j ( R i + 0 , ω , k ) − ∂ r Φ j ( R i − 0 , ω , k )  = (Θ c ) i R − 2 i Y ( ω ) = α i c i Y ( ω ) = α i u ( R i ω ) . Th us u satisfies the transmission conditions. Since c  = 0, there exists some i such that c i  = 0, and hence u ( R i ω ) = c i Y ( ω ) ≡ 0 . Therefore u ≡ 0. W e ha ve thus constructed a nontrivial outgoing solution in the ( ℓ, m ) c hannel with zero incident part. This contradicts Lemma 3.4. Hence K ℓ ( k 2 + i 0) m ust b e in vertible. The final statemen t follows immediately . 3.3 Determination of the c hannel scattering co efficien t W e no w iden tify the scattering co efficien t in each c hannel and derive the determinan t formula. 19 Prop osition 3.7. Fix ℓ ≥ 0 and m with − ℓ ≤ m ≤ ℓ . Then, for almost every k > 0 , the stationary sc attering solution in the ( ℓ, m ) channel whose incident p art is j ℓ ( k | x | ) Y ℓm ( b x ) = 1 2  h (1) ℓ ( k | x | ) + h (2) ℓ ( k | x | )  Y ℓm ( b x ) has exterior form u ( x, k ) = 1 2  h (2) ℓ ( k r ) + S ℓ ( k ) h (1) ℓ ( k r )  Y ℓm ( b x ) , r > R N . (3.10) Equivalently, if an outgoing solution in the ( ℓ, m ) channel with the same incident p art has exterior form 1 2  h (2) ℓ ( k r ) + β h (1) ℓ ( k r )  Y ℓm ( b x ) , r > R N , then β = S ℓ ( k ) . Pr o of. W e use the standard stationary interpretation of the fib er scattering matrix in the rotationally symmetric setting. F or almost ev ery k > 0, the fib er op erator S ( k 2 ) maps the incoming amplitude at energy k 2 to the out- going amplitude of the corresp onding stationary scattering solution, see [9, Ch. XI, Sects. 8A–C] and [11]. No w j ℓ ( k | x | ) Y ℓm ( b x ) = 1 2 h (2) ℓ ( k | x | ) Y ℓm ( b x ) + 1 2 h (1) ℓ ( k | x | ) Y ℓm ( b x ) , so, in the standard incoming/outgoing normalization for spherical wa v es, the free channel wa v e has incoming amplitude 1 2 Y ℓm . Therefore the corre- sp onding stationary scattering solution has exterior form 1 2 h (2) ℓ ( k r ) Y ℓm ( b x ) + 1 2 h (1) ℓ ( k r )( S ( k 2 ) Y ℓm )( b x ) , r > R N . By Theorem 3.1, S ( k 2 ) ↾ H ℓ = S ℓ ( k ) I H ℓ for almost ev ery k > 0, hence S ( k 2 ) Y ℓm = S ℓ ( k ) Y ℓm . Substituting this giv es the stated formula. The final assertion follo ws from this form ula together with Corollary 3.5. W e now state the main result of this section, which gives a determinant form ula for S ℓ ( k ). 20 Theorem 3.8. F or e ach ℓ ≥ 0 and for almost every k > 0 , S ℓ ( k ) = det K ℓ ( k 2 − i 0) det K ℓ ( k 2 + i 0) . (3.11) Mor e over, by Pr op osition 3.6, the right-hand side of (3.11) is wel l define d for every k > 0 . In p articular, | S ℓ ( k ) | = 1 for almost every k > 0 , so one may cho ose a r e al-value d phase shift δ ℓ ( k ) , define d for almost every k > 0 , such that S ℓ ( k ) = e 2 iδ ℓ ( k ) . If J ⊂ (0 , ∞ ) is an interval, then, for any c ontinuous br anch of arg det K ℓ ( k 2 + i 0) on J , one may cho ose δ ℓ on J so that δ ℓ ( k ) = − arg det K ℓ ( k 2 + i 0) for almost every k ∈ J. (3.12) Pr o of. It suffices to prov e (3.11). Fix ℓ ≥ 0, and let k > 0 b e such that S ℓ ( k ) is defined. By Theorem 3.1, this holds for almost ev ery k > 0. By Prop osition 3.6, det K ℓ ( k 2 + i 0)  = 0 . Fix m with − ℓ ≤ m ≤ ℓ . Hence Lemma 3.3 applies, and there exists an outgoing solution in the ( ℓ, m ) channel with inciden t part j ℓ ( k | x | ) Y ℓm ( b x ) , and its exterior form is 1 2  h (2) ℓ ( k r ) + σ ℓ ( k ) h (1) ℓ ( k r )  Y ℓm ( b x ) , r > R N , where σ ℓ ( k ) = 1 − 2 ik t b ℓ ( k )Θ K ℓ ( k 2 + i 0) − 1 b ℓ ( k ) . By Prop osition 3.7, the outgoing co efficien t in this normalization is exactly S ℓ ( k ). Comparing (3.2) with (3.10), we obtain S ℓ ( k ) = σ ℓ ( k ) . Therefore S ℓ ( k ) = 1 − 2 ik t b ℓ ( k )Θ K ℓ ( k 2 + i 0) − 1 b ℓ ( k ) . (3.13) Moreo ver, using h (1) ℓ ( t ) + h (2) ℓ ( t ) = 2 j ℓ ( t ) , 21 w e obtain ( m ℓ ( k 2 − i 0) − m ℓ ( k 2 + i 0)) ij = − 2 ik j ℓ ( k R i ) j ℓ ( k R j ) , 1 ≤ i, j ≤ N . Equiv alently , m ℓ ( k 2 − i 0) − m ℓ ( k 2 + i 0) = − 2 ik b ℓ ( k ) t b ℓ ( k ) . Hence K ℓ ( k 2 − i 0) = K ℓ ( k 2 + i 0) − 2 ik b ℓ ( k ) t b ℓ ( k ) Θ . Applying the matrix determinan t lemma det( A + u t v ) = det( A )  1 + t v A − 1 u  with A = K ℓ ( k 2 + i 0) , u = − 2 ik b ℓ ( k ) , t v = t b ℓ ( k )Θ , w e obtain det K ℓ ( k 2 − i 0) = det K ℓ ( k 2 + i 0)  1 − 2 ik t b ℓ ( k )Θ K ℓ ( k 2 + i 0) − 1 b ℓ ( k )  = det K ℓ ( k 2 + i 0) S ℓ ( k ) b y (3.13). Therefore S ℓ ( k ) = det K ℓ ( k 2 − i 0) det K ℓ ( k 2 + i 0) . This prov es (3.11) for ev ery k > 0 suc h that S ℓ ( k ) is defined, hence for almost ev ery k > 0. By Proposition 3.6, the denominator in (3.11) is nonzero for ev ery k > 0, so the righ t-hand side is well defined for every k > 0. By Lemma 3.2, det K ℓ ( k 2 − i 0) = det K ℓ ( k 2 + i 0) . Com bining this with (3.11), we obtain | S ℓ ( k ) | = 1 for almost every k > 0 . Hence one may choose a real-v alued phase shift δ ℓ ( k ), defined for almost ev ery k > 0, such that S ℓ ( k ) = e 2 iδ ℓ ( k ) . No w let J ⊂ (0 , ∞ ) b e an interv al, and set D ℓ ( k ) := det K ℓ ( k 2 + i 0) . Since the en tries of K ℓ ( k 2 + i 0) dep end con tinuously on k > 0, the function D ℓ is con tinuous on (0 , ∞ ). By Prop osition 3.6, D ℓ ( k )  = 0 ( k > 0) . 22 Hence, on J one may c ho ose a contin uous branch of arg D ℓ ( k ), and for almost ev ery k ∈ J one has S ℓ ( k ) = D ℓ ( k ) D ℓ ( k ) = e − 2 i arg D ℓ ( k ) . Therefore one ma y choose δ ℓ on J so that δ ℓ ( k ) = − arg D ℓ ( k ) = − arg det K ℓ ( k 2 + i 0) for almost every k ∈ J, whic h prov es (3.12). R emark 3.9 . By Prop osition 3.6, the determinant ratio in (3.11) is w ell defined for ev ery k > 0. The phrase “for almost every k > 0” refers only to the measurable represen tative of the channel scattering co efficien t arising from the abstract scattering op erator. 4 The double δ –shell case W e restrict attention to the s –w av e c hannel ( ℓ = 0), whic h captures the leading con tribution in the low–energy regime. In this channel the formulas reduce to elementary functions and the threshold b eha vior can b e analyzed in a completely explicit form. F or ℓ ≥ 1, the same determinant formula remains v alid. In the present pap er, how ev er, w e restrict the threshold analysis to the s –wa v e chan- nel, which already captures the phenomenon of interest in the double–shell mo del. W e now sp ecialize to the case N = 2. Thus 0 < R 1 < R 2 , H = − ∆ + α 1 δ ( | x | − R 1 ) + α 2 δ ( | x | − R 2 ) . W e write θ j = α j R 2 j , j = 1 , 2 . In this case the general determinant form ula from the previous section b e- comes completely explicit. W e restrict atten tion to the s –wa v e channel, where all relev ant quantities can b e written in elementary functions. 4.1 The s –w a ve c hannel W e consider the case ℓ = 0. Then j 0 ( x ) = sin x x , h (1) 0 ( x ) = − ie ix x . F or conv enience, w e also set s j = sin( k R j ) , c j = cos( k R j ) , j = 1 , 2 . The follo wing lemma gives the b oundary matrix in this channel. 23 Lemma 4.1. F or k > 0 , the s –wave b oundary matrix m 0 ( k 2 + i 0) is given by m 0 ( k 2 + i 0) =      s 1 e ikR 1 k R 2 1 s 1 e ikR 2 k R 1 R 2 s 1 e ikR 2 k R 1 R 2 s 2 e ikR 2 k R 2 2      . A c c or dingly, K 0 ( k 2 + i 0) = I 2 + m 0 ( k 2 + i 0)Θ , Θ = diag ( θ 1 , θ 2 ) . Pr o of. This follows immediately from Lemma 2.2 with ℓ = 0, using j 0 ( t ) = sin t t , h (1) 0 ( t ) = − i e it t . The formula for K 0 ( k 2 + i 0) is the definition of K ℓ ( z ) sp ecialized to ℓ = 0. W e now derive an explicit form ula for S 0 ( k ) in the s –w av e c hannel, equiv- alen tly for the determinant det K 0 ( k 2 + i 0). Theorem 4.2. Define r e al functions A 0 ( k ) and B 0 ( k ) by A 0 ( k ) = R 2 1 R 2 2 k 2 + R 2 1 k c 2 s 2 θ 2 + R 2 2 k c 1 s 1 θ 1 + θ 1 θ 2  c 1 c 2 s 1 s 2 − c 2 2 s 2 1  , (4.1) and B 0 ( k ) = R 2 1 k s 2 2 θ 2 + R 2 2 k s 2 1 θ 1 + θ 1 θ 2 s 1 s 2  c 1 s 2 − c 2 s 1  . (4.2) Then k 2 R 2 1 R 2 2 det K 0 ( k 2 + i 0) = A 0 ( k ) + iB 0 ( k ) . (4.3) Conse quently, for almost every k > 0 , S 0 ( k ) = A 0 ( k ) − iB 0 ( k ) A 0 ( k ) + iB 0 ( k ) . (4.4) Mor e over, for any interval J ⊂ (0 , ∞ ) and any c ontinuous br anch of arg  A 0 ( k )+ iB 0 ( k )  on J , one may cho ose the phase shift δ 0 on J so that δ 0 ( k ) = − arg  A 0 ( k ) + iB 0 ( k )  , for almost every k ∈ J. (4.5) Pr o of. By Lemma 4.1, K 0 ( k 2 + i 0) =      1 + θ 1 s 1 e ikR 1 k R 2 1 θ 2 s 1 e ikR 2 k R 1 R 2 θ 1 s 1 e ikR 2 k R 1 R 2 1 + θ 2 s 2 e ikR 2 k R 2 2      . 24 Hence det K 0 ( k 2 + i 0) =  1 + θ 1 s 1 e ikR 1 k R 2 1  1 + θ 2 s 2 e ikR 2 k R 2 2  − θ 1 θ 2 s 2 1 e 2 ikR 2 k 2 R 2 1 R 2 2 . (4.6) Multiplying b y k 2 R 2 1 R 2 2 , w e obtain k 2 R 2 1 R 2 2 det K 0 ( k 2 + i 0) = k 2 R 2 1 R 2 2 + k R 2 2 θ 1 s 1 e ikR 1 + k R 2 1 θ 2 s 2 e ikR 2 + θ 1 θ 2 s 1 s 2 e ik ( R 1 + R 2 ) − θ 1 θ 2 s 2 1 e 2 ikR 2 . (4.7) Using e ikR j = c j + is j , e ik ( R 1 + R 2 ) = ( c 1 + is 1 )( c 2 + is 2 ) , e 2 ikR 2 = ( c 2 + is 2 ) 2 , w e separate the real and imaginary parts. A direct computation yields k 2 R 2 1 R 2 2 det K 0 ( k 2 + i 0) = A 0 ( k ) + iB 0 ( k ) , (4.8) where A 0 ( k ) and B 0 ( k ) are given by (4.1) and (4.2). This pro ves (4.3). The form ula for S 0 ( k ) now follows from Theorem 3.8. Indeed, by (4.3), det K 0 ( k 2 + i 0) = A 0 ( k ) + iB 0 ( k ) k 2 R 2 1 R 2 2 , and, b y Lemma 3.2, det K 0 ( k 2 − i 0) = det K 0 ( k 2 + i 0) = A 0 ( k ) − iB 0 ( k ) k 2 R 2 1 R 2 2 . Substituting these expressions into (3.11), w e obtain (4.4) for almost every k > 0. No w let J ⊂ (0 , ∞ ) b e an interv al. By Prop osition 3.6, det K 0 ( k 2 + i 0)  = 0 ( k > 0) . Hence, b y (4.3), A 0 ( k ) + iB 0 ( k )  = 0 ( k > 0) . Therefore, for an y contin uous branc h of arg  A 0 ( k ) + iB 0 ( k )  on J , one may c ho ose δ 0 on J so that δ 0 ( k ) = − arg det K 0 ( k 2 + i 0) = − arg  A 0 ( k )+ iB 0 ( k )  , for almost ev ery k ∈ J, whic h is exactly (4.5). R emark 4.3 . F orm ula (4.4) giv es a completely explicit expression for the s –w av e scattering matrix in the double δ –shell case. 25 5 Lo w–energy b eha vior in the double δ –shell case W e contin ue to assume that N = 2 and restrict atten tion to the s –wa v e c hannel ℓ = 0. Using the explicit form ula in Theorem 4.2, we analyze the b eha vior of the scattering matrix near the threshold k = 0. W e first derive the regular threshold expansion and, in particular, an explicit formula for the scattering length. Theorem 5.1. Set C 0 = R 2 1 R 2 2 + R 2 1 R 2 θ 2 + R 1 R 2 2 θ 1 + θ 1 θ 2 R 1 ( R 2 − R 1 ) . (5.1) Assume that C 0  = 0 . Then one may cho ose the phase shift δ 0 ( k ) for al l sufficiently smal l k > 0 so that δ 0 ( k ) → 0 ( k ↓ 0) . F or this choic e, as k ↓ 0 , δ 0 ( k ) = − a s k + O ( k 3 ) , (5.2) wher e the sc attering length is given by a s = R 2 1 R 2 2 ( θ 1 + θ 2 ) + θ 1 θ 2 R 1 R 2 ( R 2 − R 1 ) R 2 1 R 2 2 + R 2 1 R 2 θ 2 + R 1 R 2 2 θ 1 + θ 1 θ 2 R 1 ( R 2 − R 1 ) . (5.3) Equivalently, S 0 ( k ) = 1 − 2 ia s k + O ( k 2 ) , k ↓ 0 . (5.4) Pr o of. W e first sho w that A 0 ( k ) + iB 0 ( k )  = 0 for all sufficiently small k > 0. Once this is established, Theorem 4.2 allows us to c ho ose the phase shift so that δ 0 ( k ) = − arg  A 0 ( k ) + iB 0 ( k )  for all sufficiently small k > 0. W e b egin by expanding A 0 ( k ) and B 0 ( k ) as k ↓ 0. Using sin( k R j ) = k R j + O ( k 3 ) , cos( k R j ) = 1 + O ( k 2 ) , j = 1 , 2 , form ula (4.1) gives A 0 ( k ) = R 2 1 R 2 2 k 2 + R 2 1 k θ 2  k R 2 + O ( k 3 )  + R 2 2 k θ 1  k R 1 + O ( k 3 )  + θ 1 θ 2  R 1 R 2 − R 2 1  k 2 + O ( k 4 ) . (5.5) 26 Hence A 0 ( k ) = C 0 k 2 + O ( k 4 ) . (5.6) Similarly , (4.2) yields B 0 ( k ) = R 2 1 k θ 2  k 2 R 2 2 + O ( k 4 )  + R 2 2 k θ 1  k 2 R 2 1 + O ( k 4 )  + θ 1 θ 2 ( k R 1 )( k R 2 )  k R 2 − k R 1  + O ( k 5 ) , (5.7) and therefore B 0 ( k ) = Γ 0 k 3 + O ( k 5 ) , (5.8) where Γ 0 = R 2 1 R 2 2 ( θ 1 + θ 2 ) + θ 1 θ 2 R 1 R 2 ( R 2 − R 1 ) . (5.9) Since C 0  = 0, relation (5.6) sho ws that A 0 ( k )  = 0 for all sufficien tly small k > 0. Hence also A 0 ( k ) + iB 0 ( k )  = 0 for all sufficien tly small k > 0. Moreov er, b y (4.4), S 0 ( k ) = 1 − i B 0 ( k ) / A 0 ( k ) 1 + i B 0 ( k ) / A 0 ( k ) . Since B 0 ( k ) A 0 ( k ) = O ( k ) ( k ↓ 0) , it follo ws that S 0 ( k ) = 1 + O ( k ) ( k ↓ 0) . Therefore, one may c ho ose the phase shift δ 0 ( k ) for all sufficiently small k > 0 so that δ 0 ( k ) → 0 ( k ↓ 0) . Since phase shifts are determined only mo dulo π , for this choice we hav e δ 0 ( k ) = − arctan B 0 ( k ) A 0 ( k ) (5.10) for all sufficien tly small k > 0. Moreo ver, B 0 ( k ) A 0 ( k ) = O ( k ) ( k ↓ 0) . Using arctan t = t + O ( t 3 ) ( t → 0) , 27 w e obtain from (5.10) that δ 0 ( k ) = − B 0 ( k ) A 0 ( k ) + O ( k 3 ) . (5.11) Com bining (5.6) and (5.8), we obtain δ 0 ( k ) = − Γ 0 C 0 k + O ( k 3 ) . (5.12) Th us (5.2) holds with a s = Γ 0 C 0 , whic h is exactly (5.3). Finally , since S 0 ( k ) = e 2 iδ 0 ( k ) , equation (5.2) implies S 0 ( k ) = 1 + 2 iδ 0 ( k ) + O ( k 2 ) = 1 − 2 ia s k + O ( k 2 ) , whic h prov es (5.4). R emark 5.2 . The condition C 0  = 0 c haracterizes the regular threshold regime. The complemen tary case C 0 = 0 corresp onds to a threshold–critical configuration. Under the standing assumption 0 < R 1 < R 2 , the condition C 0 = 0 already implies Γ 0  = 0 , as will b e sho wn in the pro of of Theorem 5.3. More precisely , the next theorem treats the nondegenerate exceptional case in which C 0 = 0 , C 2  = 0 , where C 2 is defined in Theorem 5.3. F urther degenerate situations, such as C 0 = 0 and C 2 = 0, require a separate higher–order analysis. W e next consider the nondegenerate exceptional threshold regime in the double δ –shell cas e. Theorem 5.3. Assume that N = 2 and ℓ = 0 , and define C 0 and Γ 0 by (5.1) and (5.9) . Supp ose that C 0 = 0 , C 2  = 0 , wher e C 2 = − 2 3 R 2 1 R 3 2 θ 2 − 2 3 R 3 1 R 2 2 θ 1 + θ 1 θ 2  − 2 3 ( R 3 1 R 2 + R 1 R 3 2 ) + R 2 1 R 2 2 + 1 3 R 4 1  . (5.13) 28 Then, as k ↓ 0 , A 0 ( k ) = C 2 k 4 + O ( k 6 ) , (5.14) and B 0 ( k ) = Γ 0 k 3 + O ( k 5 ) . (5.15) In p articular, B 0 ( k ) A 0 ( k ) = Γ 0 C 2 1 k + O ( k ) , k ↓ 0 , (5.16) and henc e S 0 ( k ) → − 1 ( k ↓ 0) . (5.17) Equivalently, one may cho ose the phase shift δ 0 ( k ) on (0 , ε ) , for some ε > 0 , so that δ 0 ( k ) → ± π 2 ( k ↓ 0) . (5.18) Pr o of. W e first note that, under the standing assumption 0 < R 1 < R 2 , the condition C 0 = 0 already implies Γ 0  = 0 . Indeed, if Γ 0 = 0 as w ell, then 0 = C 0 R 1 − Γ 0 R 1 R 2 = R 2  R 1 R 2 + ( R 2 − R 1 ) θ 1  , and hence R 1 R 2 + ( R 2 − R 1 ) θ 1 = 0 . Substituting this in to Γ 0 = R 1 R 2  R 1 R 2 ( θ 1 + θ 2 )+( R 2 − R 1 ) θ 1 θ 2  = R 1 R 2  θ 2  R 1 R 2 +( R 2 − R 1 ) θ 1  + R 1 R 2 θ 1  , w e obtain Γ 0 = R 2 1 R 2 2 θ 1 = − R 3 1 R 3 2 R 2 − R 1  = 0 , a con tradiction. W e next expand A 0 ( k ) to higher order as k ↓ 0. Using sin( k R j ) = k R j − ( k R j ) 3 6 + O ( k 5 ) , cos( k R j ) = 1 − ( k R j ) 2 2 + O ( k 4 ) , j = 1 , 2 , w e first compute c j s j = 1 − k 2 R 2 j 2 + O ( k 4 ) ! k R j − k 3 R 3 j 6 + O ( k 5 ) ! = k R j − 2 3 k 3 R 3 j + O ( k 5 ) . 29 Hence R 2 1 k c 2 s 2 θ 2 = R 2 1 R 2 θ 2 k 2 − 2 3 R 2 1 R 3 2 θ 2 k 4 + O ( k 6 ) , R 2 2 k c 1 s 1 θ 1 = R 1 R 2 2 θ 1 k 2 − 2 3 R 3 1 R 2 2 θ 1 k 4 + O ( k 6 ) . Next, w e expand the last term in (4.1). W e hav e s 1 s 2 = k 2 R 1 R 2 − k 4 6 ( R 3 1 R 2 + R 1 R 3 2 ) + O ( k 6 ) , c 1 c 2 = 1 − k 2 2 ( R 2 1 + R 2 2 ) + O ( k 4 ) , and therefore c 1 c 2 s 1 s 2 = k 2 R 1 R 2 − 2 3 k 4 ( R 3 1 R 2 + R 1 R 3 2 ) + O ( k 6 ) . Also, s 2 1 = k 2 R 2 1 − 1 3 k 4 R 4 1 + O ( k 6 ) , c 2 2 = 1 − k 2 R 2 2 + O ( k 4 ) , so c 2 2 s 2 1 = k 2 R 2 1 − k 4  R 2 1 R 2 2 + 1 3 R 4 1  + O ( k 6 ) . Th us c 1 c 2 s 1 s 2 − c 2 2 s 2 1 = k 2 R 1 ( R 2 − R 1 ) + k 4  − 2 3 ( R 3 1 R 2 + R 1 R 3 2 ) + R 2 1 R 2 2 + 1 3 R 4 1  + O ( k 6 ) . Substituting these expansions in to (4.1), we obtain A 0 ( k ) = C 0 k 2 + C 2 k 4 + O ( k 6 ) . Since C 0 = 0, this giv es (5.14). On the other hand, b y (5.8), B 0 ( k ) = Γ 0 k 3 + O ( k 5 ) , whic h prov es (5.15). Since C 2  = 0, w e may divide (5.15) by (5.14) and obtain B 0 ( k ) A 0 ( k ) = Γ 0 C 2 1 k + O ( k ) , k ↓ 0 . 30 Because C 0 = 0 implies Γ 0  = 0, the co efficien t Γ 0 /C 2 is nonzero. Hence     B 0 ( k ) A 0 ( k )     → ∞ ( k ↓ 0) . Using (4.4), we write S 0 ( k ) = 1 − i B 0 ( k ) / A 0 ( k ) 1 + i B 0 ( k ) / A 0 ( k ) . Since B 0 ( k ) / A 0 ( k ) → ±∞ , it follows that S 0 ( k ) → − 1 ( k ↓ 0) , whic h prov es (5.17). Since C 0 = 0 implies Γ 0  = 0, relation (5.15) sho ws that B 0 ( k )  = 0 for all sufficien tly small k > 0. Hence A 0 ( k ) + iB 0 ( k )  = 0 (0 < k < ε ) for some ε > 0. Since A 0 ( k ) + iB 0 ( k ) is con tin uous and nonv anishing on (0 , ε ), one may choose a contin uous branch of arg  A 0 ( k ) + iB 0 ( k )  there. Therefore, b y Theorem 4.2, one may choose the phase shift on (0 , ε ) so that δ 0 ( k ) = − arg  A 0 ( k ) + iB 0 ( k )  . In view of (5.17), this choice satisfies δ 0 ( k ) → ± π 2 ( k ↓ 0) , whic h prov es (5.18). W e sho w that the condition C 0 = 0 is equiv alen t to the existence of a non trivial zero–energy radial solution whose exterior constant term v anishes. Prop osition 5.4. Assume that N = 2 and ℓ = 0 . Then the fol lowing ar e e quivalent: (i) C 0 = 0 . (ii) Ther e exists a nontrivial r adial function u , pie c ewise harmonic away fr om the shel ls, c ontinuous acr oss the shel ls, and satisfying the δ –shel l jump c onditions, such that H u = 0 in the distributional sense, u is r e gular at the origin, and u ( x ) = O ( | x | − 1 ) ( | x | → ∞ ) . 31 Mor e pr e cisely, every r adial zer o–ener gy solution r e gular at the origin is of the form u ( x ) = f ( | x | ) , wher e f ( r ) =            a, 0 < r < R 1 , b + c r , R 1 < r < R 2 , d + e r , r > R 2 , with c onstants satisfying c = − θ 1 a, b = a + θ 1 R 1 a, and d = C 0 R 2 1 R 2 2 a. Henc e the exterior c onstant term vanishes if and only if C 0 = 0 . Pr o of. Let u ( x ) = f ( r ) b e a radial solution of H u = 0, where r = | x | . Aw ay from the shells, the equation reduces to − ∆ u = 0 . F or an s –wa v e radial function in three dimensions, the general harmonic solution is of the form f ( r ) = A + B r . Regularit y at the origin forces B = 0 in the region 0 < r < R 1 . Hence f ( r ) =            a, 0 < r < R 1 , b + c r , R 1 < r < R 2 , d + e r , r > R 2 , for suitable constan ts a, b, c, d, e . W e imp ose contin uit y and the δ –shell jump conditions. A t r = R 1 , con tinuit y giv es a = b + c R 1 . Since f ′ ( r ) = 0 (0 < r < R 1 ) , f ′ ( r ) = − c r 2 ( R 1 < r < R 2 ) , 32 the jump condition at r = R 1 yields − c R 2 1 = α 1 a. Using θ 1 = α 1 R 2 1 , w e obtain c = − θ 1 a, b = a + θ 1 R 1 a. A t r = R 2 , con tinuit y giv es b + c R 2 = d + e R 2 . Moreo ver, f ′ ( r ) = − c r 2 ( R 1 < r < R 2 ) , f ′ ( r ) = − e r 2 ( r > R 2 ) , so the jump condition at r = R 2 b ecomes − e R 2 2 + c R 2 2 = α 2  b + c R 2  . Equiv alently , e = c − θ 2  b + c R 2  . Substituting the expressions for b and c , we obtain e = − θ 1 a − θ 2  a + θ 1 R 1 a − θ 1 R 2 a  . It remains to compute the exterior constan t term d . F rom contin uity at r = R 2 , d = b + c R 2 − e R 2 . Substituting the form ulas ab o ve and simplifying, we find d =  1 + θ 1 R 1 + θ 2 R 2 + θ 1 θ 2  1 R 1 R 2 − 1 R 2 2  a. Multiplying b y R 2 1 R 2 2 , this b ecomes R 2 1 R 2 2 d =  R 2 1 R 2 2 + R 1 R 2 2 θ 1 + R 2 1 R 2 θ 2 + θ 1 θ 2 R 1 ( R 2 − R 1 )  a = C 0 a. Hence d = C 0 R 2 1 R 2 2 a. 33 Therefore the exterior constan t term v anishes if and only if d = 0, that is, if and only if C 0 = 0. Con versely , any piecewise harmonic radial function that is contin uous across r = R 1 , R 2 and satisfies the ab o v e jump conditions is a distributional solution of H u = 0. Moreo v er, if a = 0, then the form ulas obtained ab o ve giv e b = c = d = e = 0 , so the solution is trivial. Hence a non trivial radial distributional solution, regular at the origin and satisfying u ( x ) = O ( | x | − 1 ) ( | x | → ∞ ) , exists if and only if a  = 0 and d = 0, which is equiv alen t to C 0 = 0. W e show that the scattering length is determined by the exterior b ehav- ior of the zero–energy radial solution. Prop osition 5.5. Assume that N = 2 , ℓ = 0 , and C 0  = 0 . L et u b e a nontrivial r adial solution of H u = 0 which is r e gular at the origin, and normalize it so that u ( x ) = 1 − a | x | ( | x | > R 2 ) . Then a = a s , wher e a s is given by (5.3) . Pr o of. In the notation of Prop osition 5.4, the exterior part of a radial zero– energy solution is d + e r , and the computation in the pro of of Prop osition 5.4 gives d = C 0 R 2 1 R 2 2 a 0 , e = − Γ 0 R 2 1 R 2 2 a 0 , where a 0 denotes the in terior constant on (0 , R 1 ). Since C 0  = 0, w e may normalize b y d = 1. Then u ( x ) = 1 + e | x | = 1 − Γ 0 C 0 1 | x | , | x | > R 2 . By (5.3), one has a s = Γ 0 C 0 . Therefore a = a s . 34 Prop osition 5.4 giv es a concrete in terpretation of the exceptional thresh- old condition in Theorem 5.3. In the regular case C 0  = 0, a zero–energy radial solution regular at the origin has a nonzero constant term at infinity , and the standard scattering–length description applies. By contrast, the condition C 0 = 0 means that the exterior constant term v anishes, leaving a deca ying tail of order r − 1 . Th us the exceptional regime corresp onds to a threshold–critical configu- ration in which the contributions of the tw o shells cancel at zero energy . In this case, S 0 ( k ) → − 1 ( k ↓ 0) as sho wn in Theorem 5.3. This behavior reflects the presence of a zero– energy s –wa ve solution with v anishing exterior constant term and explains the breakdo wn of the finite scattering–length picture. R emark 5.6 . A further degenerate situation ma y o ccur if C 0 = 0 and C 2 = 0. In that case, higher–order terms in the threshold expansion b ecome relev ant and require a separate analysis. Concluding Remarks In this pap er we derived a determinant formula for the c hannel scattering co efficien ts of Sc hr¨ odinger op erators with finitely man y concentric δ –shell in teractions. The main result shows that, after partial–wa v e reduction, the scattering problem is reduced to a finite-dimensional matrix problem gov- erned b y the same b oundary operator that app ears in the resolven t form ula. As an application, w e analyzed in detail the double–shell mo del in the s –w av e c hannel. W e obtained explicit form ulas for the scattering matrix and the scattering length, and we gav e a zero–energy characterization of a threshold–critical configuration. In particular, we sho wed that the condi- tion C 0 = 0 is equiv alent to the existence of a zero–energy radial solution with v anishing exterior constant term. In the corresp onding nondegener- ate exceptional case, this threshold–critical configuration yields the limiting b eha vior S 0 ( k ) → − 1 ( k ↓ 0) . The determinan t formula also admits a natural structural in terpretation. Whenev er ∥ m ℓ ( z )Θ ∥ < 1 , where ∥ · ∥ denotes the op erator norm of matrices on C N , the in verse of the reduced b oundary matrix is given by the conv ergen t Neumann series K ℓ ( z ) − 1 = ( I N + m ℓ ( z )Θ) − 1 = I N − m ℓ ( z )Θ + ( m ℓ ( z )Θ) 2 − · · · . Eac h term in this series represen ts one additional step in the multiple scat- tering pro cess b et ween the shells. Thus K ℓ ( z ) pro vides a finite-dimensional 35 description of this pro cess, and the determinant in the scattering form ula ma y b e viewed as enco ding the cum ulative effect of these rep eated in terac- tions. This interpretation is therefore not merely formal in regimes where the ab o ve Neumann series conv erges, for example for sufficien tly small cou- pling strengths. Sev eral natural problems remain for further study . One imp ortan t prob- lem is to relate the determinant formula to the sp ectral shift function and the Birman–Kre ˘ ın formula [3]. Another is to analyze more degenerate threshold situations, where higher-order terms in the expansion b ecome relev an t. It w ould also b e in teresting to extend the present approach to more general h yp ersurface in teractions. References [1] S. Alb ev erio, F. Gesztesy , R. Høegh–Krohn, and H. Holden, Solv able Mo dels in Quantum Mechanics, 2nd ed., AMS Chelsea Publishing, Providence, RI, 2005. DOI: 10.1090/c hel/350 [2] J. Behrndt, M. Langer, and V. Lotoreichik, Schr¨ odinger op erators with δ and δ ′ -p oten tials supp orted on h yp ersurfaces, Ann. Henri P oincar´ e 14 (2013), 385– 423. DOI: 10.1007/s00023-012-0189-5 [3] M. Sh. Birman and M. G. Kre ˘ ın, On the theory of wa ve operators and scatter- ing op erators, Dokl. Ak ad. Nauk SSSR 144 (1962), 475–478; English transl.: So viet Math. Dokl. 3 (1962), 740–744. [4] J. F. Brasc he, P . Exner, Y u. A. Kup erin, and P . ˇ Seba, Schr¨ odinger operators with singular in teractions, J. Math. Anal. Appl. 184 (1994), 112–139. DOI: 10.1006/jmaa.1994.1188 [5] M. N. Hounk onnou, M. Hounkp e, and J. Shabani, Scattering theory for finitely man y sphere interactions supported by concen tric spheres, J. Math. Phys. 38 (1997), 2832–2850. DOI: 10.1063/1.532022 [6] T. Ikebe and S. Shimada, Sp ectral and scattering theory for the Schr¨ odinger op erators with p enetrable wall p oten tials, J. Math. Ky oto Univ. 31 (1991), 219–258. DOI: 10.1215/kjm/1250519902 [7] M. Kaminaga, Schr¨ odinger op erators with concentric δ –shell interactions, to app ear in Analysis and Mathematical Ph ysics; [8] A. Posilicano, A Krein-like formula for singular p erturbations of self-adjoint op erators and applications, J. F unct. Anal. 183 (2001), 109–147. DOI: 10.1006/jfan.2000.3730 [9] M. Reed and B. Simon, Metho ds of Mo dern Mathematical Physics. I II: Scat- tering Theory , Academic Press, New Y ork, 1979. [10] J. Shabani, Finitely many δ interactions with supp orts on concentric spheres, J. Math. Ph ys. 29 (1988), 660–664. DOI: 10.1063/1.528005 36 [11] D. R. Y afaev, Mathematical Scattering Theory: General Theory , T ranslations of Mathematical Monographs, V ol. 105, American Mathematical Society , Pro v- idence, RI, 1992. DOI: 10.1090/mmono/105 37