From Laplacian Transport to Dirichlet-to-Neumann Gibbs) Semigroups

The paper gives a short account of some basic properties of Dirichlet-to-Neumann operators Λγ∂Ω including the corresponding semigroups motivated by the Laplacian transport in anisotropic media (γ ≠ I) and by elliptic systems with dynamical boundary conditions.

Saved in:
Bibliographic Details
Published in:Журнал математической физики, анализа, геометрии
Date:2008
Main Author: Zagrebnov, V.A.
Format: Article
Language:English
Published: Фізико-технічний інститут низьких температур ім. Б.І. Вєркіна НАН України 2008
Online Access:https://nasplib.isofts.kiev.ua/handle/123456789/106522
Tags: Add Tag
No Tags, Be the first to tag this record!
Journal Title:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Cite this:From Laplacian Transport to Dirichlet-to-Neumann Gibbs) Semigroups / V.A. Zagrebnov // Журнал математической физики, анализа, геометрии. — 2008. — Т. 4, № 4. — С. 551-568. — Бібліогр.: 23 назв. — англ.

Institution

Digital Library of Periodicals of National Academy of Sciences of Ukraine
_version_ 1860258628640964608
author Zagrebnov, V.A.
author_facet Zagrebnov, V.A.
citation_txt From Laplacian Transport to Dirichlet-to-Neumann Gibbs) Semigroups / V.A. Zagrebnov // Журнал математической физики, анализа, геометрии. — 2008. — Т. 4, № 4. — С. 551-568. — Бібліогр.: 23 назв. — англ.
collection DSpace DC
container_title Журнал математической физики, анализа, геометрии
description The paper gives a short account of some basic properties of Dirichlet-to-Neumann operators Λγ∂Ω including the corresponding semigroups motivated by the Laplacian transport in anisotropic media (γ ≠ I) and by elliptic systems with dynamical boundary conditions.
first_indexed 2025-12-07T18:52:37Z
format Article
fulltext Journal of Mathematical Physics, Analysis, Geometry 2008, vol. 4, No. 4, pp. 551�568 From Laplacian Transport to Dirichlet-to-Neumann (Gibbs) Semigroups V.A. Zagrebnov Centre de Physique Th�eorique - UMR 6207, Universit�e de la M�editerran�ee Luminy - Case 907, 13288 Marseille, Cedex 09, France E-mail:Valentin.Zagrebnov@cpt.univ-mrs.fr Received January 9, 2008 The paper gives a short account of some basic properties of Dirichlet- to-Neumann operators � ;@ including the corresponding semigroups moti- vated by the Laplacian transport in anisotropic media ( 6= I) and by elliptic systems with dynamical boundary conditions. To illustrate these notions and the properties we use the explicitly constructed Lax semigroups. We demon- strate that for a general smooth bounded convex domain � R d the cor- responding Dirichlet-to-Neumann semigroup � U(t) := e �t� ;@ t�0 in the Hilbert space L 2(@ ) belongs to the trace-norm von Neumann�Schatten ideal for any t > 0. This means that it is in fact an immediate Gibbs semi- group. Recently H. Emamirad and I. Laadnani have constructed a Trotter� Kato�Cherno� product-type approximating family f(V ;@ (t=n)) ng n�1 strongly converging to the semigroup U(t) for n ! 1. We conclude the paper by discussion of a conjecture about convergence of the Emamirad� Laadnani approximantes in the trace-norm topology. Key words: Laplacian transport, Dirichlet-to-Neumann operators, Lax semigroups, Dirichlet-to-Neumann semigroups, Gibbs semigroups. Mathematics Subject Classi�cation 2000: 47A55, 47D03, 81Q10. 1. Laplacian Transport and Dirichlet-to-Neumann Operators E x a m p l e 1.1. It is well known (see, e.g., [LeUl]) that the problem of determining a conductivity matrix �eld (x) = [ i;j(x)] d i;j=1, for x in a bounded open domain � R d , is related to "measuring" the elliptic Dirichlet-to-Neumann An extended version of the author's talk presented on the Lyapunov Memorial Conference, June 24-30, 2007 (Kharkiv National University, Ukraine), which is based on the common project with Prof. Hassan Emamirad (Laboratoire de Math�ematiques, Universit�e de Poitiers). c V.A. Zagrebnov, 2008 V.A. Zagrebnov map for associated conductivity equation. Notice that the solution of this problem has a lot of practical applications in various domains: geophysics, electrochemistry etc. It is also an important diagnostic tool in medicine, e.g., in the electrical impedance tomography ; the tissue in the human body is an example of highly anisotropic conductor [BaBr]. Under the assumption that there is no sources or sinks of current the potential v(x); x 2 ; for a given voltage f(!); ! 2 @ ; on the (smooth) boundary @ of is a solution of the Dirichlet problem:( div( rv) = 0 in ; vj@ = f on @ : (P1) Then the corresponding to (P1) Dirichlet-to-Neumann map (operator) � ;@ is de�ned by � ;@ : f 7! @vf=@� := � � rvf j@ : (1.1) Here � is the unit outer-normal vector to the boundary at ! 2 @ and the function u := uf is the solution of the Dirichlet problem (P1). The Dirichlet-to-Neumann operator (1.1) is also called the voltage-to-current map, since the function � ;@ f gives the induced current �ux trough the boundary @ . The key (inverse) problem is whether one can determine the conductivity matrix by knowing electrical boundary measurements, i.e., the corresponding Dirichlet-to-Neumann operator? Unfortunately, this operator does not determine the matrix uniquely, see e.g. [GrUl] and references there. E x a m p l e 1.2. The problem of electrical current �ux in the form (P1) is an example of the so-called Laplacian transport. Besides the voltage-to-current problem the motivation to study this kind of transport comes for instance from the transfer across biological membranes, see e.g. [Sap], [GrFiSap]. Let some "species" of concentration C(x), x 2 R d , di�use in the isotropic bulk ( = I) from a (distant) source localized on the closed boundary @ 0 to- wards a semipermeable compact interface @ on which they disappear at a given rate W . Then the steady concentration �eld (Laplacian transport with a di�usion coe�cient D) obeys the set of equations8><>: �C = 0; x 2 0 n ; C(!0 2 @ 0) = C0; at the source; (�D) @�C(!) =W (C(!)� 0); on the interface ! 2 @ : (P2) Let C = C0(1 � u). Then �u = 0, x 2 . If we put � := D=W , then the boundary conditions on @ take the form: (I + �@�)u j@ (!) = 1 j@ (!), where (1 j@ )(!) = �@ (!) is a characteristic function of the set @ , and u(!0) = 0, !0 2 @ 0 on the source boundary. 552 Journal of Mathematical Physics, Analysis, Geometry, 2008, vol. 4, No. 4 From Laplacian Transport to Dirichlet-to-Neumann (Gibbs) Semigroups Consider now the following auxiliary Laplace�Dirichlet problem �u = 0; x 2 0 n ; u j@ (!) = f(! 2 @ ) and u j@ 0 (!) = 0; (1.2) with solution uf . Then similarly to (1.1), with the problem (1.2) we can associate a Dirihlet-to-Neumann operator � =I;@ : f 7! @�uf j@ (1.3) with the domain dom(�I;@ ), which belongs to a certain Sobolev space (Sect. 2). The advantage of this approach is that as soon as the operator (1.3) is de�ned it can be applied for studying the mixed boundary value problem (P2). This gives in particular the value of the particle �ux due to Laplacian transport across the membrane @ . Indeed, one obtains that (I + ��I;@ )u j@ = 1 j@ , and that the local (di�usive) particle �ux is de�ned as: � j@ := D C0(@nu) j@ = D C0(�I;@ (I + ��I;@ ) �11) j@ : (1.4) Then the corresponding total �ux across the membrane @ � := (�; 1)L2(@ ) = D C0(�(I + ��I;@ ) �11; 1)L2(@ ) (1.5) is experimentally measurable macroscopic response of the system expressed via transport parameters D;C0; � and geometry of @ . Here (�; �)L2(@ ) is a scalar product in the Hilbert space @H := L 2(@ ). The aim of this paper is twofold: (i) to give a short account of some standard results about Dirichlet-to-Neumann operators and related Dirichlet-to-Neumann semigroups that solve a certain class of elliptic systems with dynamical boundary conditions; (ii) to present some recent results concerning the approximation theory and the Gibbs character of the Dirichlet-to-Neumann semigroups for compact sets with smooth boundaries @ . To this end in the next Sect. 2 we recall some fundamental properties of the Dirichlet-to-Neumann operators and semigroups, we illustrate them by a few elementary examples, including the Lax semigroups [Lax]. In Section 3 we present the strong Emamirad�Laadnani approximations of the Dirichlet-to-Neumann semigroups inspired by the Cherno� theory and by its generalizations in [NeZag, CaZag2]. We show in Sect. 4 that for compact sets with smooth boundaries @ the Dirichlet-to-Neumann semigroups are in fact (immediate) Gibbs semigroups [Zag2]. Some recent results and conjectures about approximations of the Dirichlet-to- Neumann (Gibbs) semigroups in operator and trace-norm topologies are collected in the last Sect. 5. Journal of Mathematical Physics, Analysis, Geometry, 2008, vol. 4, No. 4 553 V.A. Zagrebnov 2. Dirichlet-to-Neumann Operators and Semigroups 2.1. Dirichlet-to-Neumann Operators Let be an open bounded domain in Rd with a smooth boundary @ . Let be a C1( ) matrix-valued function on , which we call the Laplacian transport matrix in domain . We suppose that the matrix-valued function (x) := [ i;j(x)] d i;j=1 satis�es the following hypotheses: (H1) The real coe�cients are symmetric and i;j(x) = j;i(x) 2 C 1( ). (H2) There exist two constants 0 < c1 � c2 <1 such that for all � 2 R d we have c1k�k 2 � nX i;j=1 �i�j i;j(x) � c2k�k 2 : (2.1) Then the Dirichlet-to-Neumann operator � ;@ associated with the Laplacian transport in is de�ned as follows. Let f 2 C(@ ), and denote by vf the unique solution (see, e.g., [GiTr, Th. 6.25]) of the Dirichlet problem( A ;@ v := div( rv) = 0 in ; v j@ = f on @ ; (P1) in the Banach space X := C( ). Here the operator A ;@ is de�ned on its maximal domain dom(A ;@ ) := fu 2 X : A ;@ u 2 Xg: (2.2) De�nition 2.1. The Dirichlet-to-Neumann operator is the map � ;@ : f 7! @vf=@� = � � rvf j@ ; (2.3) with the domain dom(� ;@ ) = ff 2 @C( R) : vf 2 Ker(A ;@ ) and j(� � rvf j@ )j <1g: (2.4) Here � denotes the unit outer-normal vector at ! 2 @ , and vf is the solution of Dirichlet problem (P1). The solution vf := L@ f of the problem (P1) is called the -harmonic lifting of f , where L@ : C(@ ) 7! C 2( ) \ C( ) is called the lifting operator with domain dom(L@ ) = C(@ ). If T@ : C( ) 7! C(@ ) denotes the trace operator on the smooth boundary @ , i.e., v j@ = T@ v, then [Eng]: L@ = (T@ jKer(A ;@ ) )�1 and dom(� ;@ ) = T@ fKer(A ;@ )g: (2.5) 554 Journal of Mathematical Physics, Analysis, Geometry, 2008, vol. 4, No. 4 From Laplacian Transport to Dirichlet-to-Neumann (Gibbs) Semigroups Remark 2.2. Let @X := C(@ ). Then (2.5) implies T@ L@ u = u ; u 2 @X and L@ T@ w = w ; w 2 Ker(A ;@ ): (2.6) One also gets that the lifting operator is bounded: L@ 2 L(@X;X), whereas the Dirichlet-to-Neumann operator (2.3) is obviously not. Now let H be Hilbert space L2( ) and @H := L 2(@ ) denote the boundary space. In order that the problem (P1) admits a unique solution vf , one has to assume that f 2 W 1=2 2 (@ ), and then vf belongs the Sobolev space W 1 2 ( ), see e.g. [Tay, Ch.7]. So, we can de�ne Dirichlet-to-Neumann operator in the Hilbert space @H by (2.3) with the domain dom(� ;@ ) := ff 2W 1=2 2 (@ ) : � ;@ f 2 @H = L 2(@ )g: (2.7) Proposition 2.3. The Dirichlet-to-Neumann operator (2.3) with domain (2.7) in the Hilbert space @H is unbounded, nonnegative, selfadjoint, �rst-order elliptic pseudodi�erential operator with compact resolvent. The complete proof can be found, e.g., in [Tay, Ch. 7], [Tay1]. Therefore, we give here only some comments on these properties of the Dirichlet-to-Neumann operator (2.3) in @H = L 2(@ ). Remark 2.4. (a) By virtue of de�nition (2.3) for any f 2W 1=2 2 (@ ) one gets (f;� ;@ f)@H = Z @ d�(!) vf (!) � � (!)(rvf )(!) (2.8) = Z dx div(vf(x) ( rvf)(x)) = Z dx (rvf(x) � rvf)(x)) � 0; since the matrix veri�es (H2). Thus, operator � ;@ is nonnegative. (b) In fact to ensure the existence of the trace T@ (� � r(L@ f)) one has ini- tially to de�ne the operator � ;@ for f 2W 3=2 2 (@ ). Then Dirichlet-to-Neumann operator is a selfadjoint extension with domain (2.7) and moreover it is a bounded map � ;@ :W 1=2 2 (@ ) 7!W �1=2 2 (@ ). (c) By (2.8) and since derivatives of the �rst-order are involved in (2.3), one can conclude that this operator should be elliptic and pseudodi�erential. If (x) = I, then �I;@ is, roughly, the operator (��@ ) 1=2, where �@ is the Laplace� Beltrami operator on @ with corresponding induced metric [Tay, Ch.7], [Tay1]. (d) Compactness of the imbedding W 1=2 2 (@ ) ,! L 2(@ ) implies the compact- ness of the resolvent of � ;@ . Journal of Mathematical Physics, Analysis, Geometry, 2008, vol. 4, No. 4 555 V.A. Zagrebnov By (a) and (d) the spectrum �(� ;@ ) of the Dirichlet-to-Neumann operator is a set of nonnegative increasing eigenvalues f�kg 1 k=1. The rate of increasing is given by the Weyl asymptotic formula, see, e.g., [Hor, Tay]: Proposition 2.5. Let � ;@ (x; �), for (x; �) 2 T � @ , be the symbol of the �rst-order elliptic pseudodi�erential Dirichlet-to-Neumann operator � ;@ . Then the asymptotic behaviour of the corresponding eigenvalues as k !1 has the form �k � � k C(@ ;� ) � 1=(d�1) ; where C(@ ;� ) := 1 (2�)d�1 Z � ;@ (x;�)�1 dx d�: Another important result is due to Hislop and Lutzer [HiLu]. It concerns a localization (rapid decay) of the -harmonic lifting of the corresponding eigen- functions. Proposition 2.6. Let f�kg 1 k=1 be eigenfunctions of the Dirichlet-to-Neumann operator: � ;@ = �k�k with k�kkL2(@ ) = 1. Let v�k := L@ �k be the -harmonic lifting of �k to corresponding to the problem (P1). Then for any compact C � and x 2 C one gets the representation jv�k(x)j = (x; p; C)=�k p (2.9) with arbitrary large p > 0. Here (x; p; C) is a decreasing function of the distance dist(x; @ ). Since by the Weyl asymptotic formula we have �k = O(k1=(d�1)), the decay im- plied by the estimate (2.9) is algebraic. Conjecture 2.7. [HiLu]. In fact the order of decay instead of (x; p; C)=�k p is exponential: O(exp[� k dist(C; @ )]). 2.2. Example of a Dirichlet-to-Neumann Operator To illustrate the results mentioned above we consider a simple example which will be useful below for contraction of the Lax semigroups. Consider a homogeneous isotropic case: (x) = I, and let = R := fx 2 R d=3 : kxk < Rg. Then A ;@ R = �@ R and for the harmonic lifting of f(!) = X l;m f (R) l;m Yl;m(�; ') 2W 1=2 2 (@ R) 556 Journal of Mathematical Physics, Analysis, Geometry, 2008, vol. 4, No. 4 From Laplacian Transport to Dirichlet-to-Neumann (Gibbs) Semigroups we obtain vf (r; �; ') = X l;m � r R �l f (R) l;m Yl;m(�; '); (2.10) since the spherical functions fYl;mg 1 l=0;jmj�l form a complete orthonormal basis in the Hilbert space @H = L 2(@ R; d� sin � d'). De�nition (2.3) and (2.10) imply that nonnegative, selfadjoint, �rst-order elliptic pseudodi�erential Dirichlet-to-Neumann operator (�I;@ Rf)(! = (R; �; ')) = 1X l=0 m=lX m=�l � l R � f (R) l;m Yl;m(�; ') (2.11) has discrete spectrum �(�I;@ R) := f�l;m = l=Rg1 l=0;jmj�l with spherical eigen- functions (�I;@ RYl;m)(R; �; ') = � l R � Yl;m(�; ') (2.12) and multiplicity m. The operator (2.11) is obviously unbounded and it has a com- pact resolvent. Remark 2.8. Since by virtue of (2.10) the -harmonic lifting of the eigen- function Yl;m to the ball R is vYl;m(r; �; ') = � r R �l Yl;m(�; '); one can check the localization (Prop. 2.6) and Conjecture about the exponential decay explicitly. For distances 0 < dist(x; @ R) = R � r � R, one obtains jvYl;m(r; �; ')j = O(e�l(R�r)=R). 2.3. Dirichlet-to-Neumann Semigroups on @X To de�ne the Dirichlet-to-Neumann semigroups on the boundary Banach space @X = C(@ ) we can follow the line of reasoning of [Esc] or [Eng]. To this end consider in X = C( ) the following elliptic system with the dynamical boundary conditions 8><>: div( ru(t; �)) = 0 in (0;1) � ; @u(t; �)=@t + @u(t; �)=@� = 0 on (0;1) � @ ; u(0; �) = f on @ : (P2) Proposition 2.9. The problem (P2) has a unique solution uf (t; x) for any f 2 C(@ ). Its trace on the boundary @ has the form uf (t; !) := (T@ uf (t; �))(!) = (U(t)f)(!); (2.13) where the family of operators fU(t) = e �t� ;@ gt�0 is a C0-semigroup generated by the Dirichlet-to-Neumann operator of the problem (P1). Journal of Mathematical Physics, Analysis, Geometry, 2008, vol. 4, No. 4 557 V.A. Zagrebnov The following key result about the properties of the Dirichlet-to-Neumann semigroups on the boundary Banach space @X = C(@ ) is due to Escher�Engel [Esc, Eng] and Emamirad�Laadnani [EmLa]: Proposition 2.10. The semigroup fU(t) = e �t� ;@ gt�0 is analytic, compact, positive, irreducible and Markov C0-semigroup of contractions on C(@ ). Remark 2.11. The complete proof can be found in the papers quoted above. So, here we make only some comments and hints concerning Prop. 2.10. 2.4. Dirichlet-to-Neumann Semigroups on @H The Dirichlet-to-Neumann semigroup fU(t) = e �t� ;@ gt�0 on @H is de�ned by selfadjoint and nonnegative Dirichlet-to-Neumann generator � ;@ of Prop. 2.3. Proposition 2.12. The Dirichlet-to-Neumann semigroup fU(t) = e �t� ;@ gt on the Hilbert space @H is a holomorphic quasisectorial contraction with values in the trace-class C1(@H) for Re (t) > 0. Remark 2.13. The �rst part of the statement follows from Prop. 2.3. Since the generator � ;@ is selfadjoint and nonnegative, the semigroup fU(t)gt is holo- morphic and quasisectorial contraction for Re (t) > 0, see, e.g., [CaZag1, Zag1]. The compactness of the resolvent of � ;@ implies the compactness of fU(t)gt>0, but to prove the last part of the statement we need a supplementary argument about asymptotic behaviour of its eigenvalues given by the Weyl asymptotic for- mula (Prop. 2.5). This behaviour of eigenvalues implies the second part of Prop. 2.12: Lemma 2.14. The Dirichlet-to-Neumann semigroup U(t) has values in the trace-class C1(@H) for any t > 0. P r o f. Since the Dirichlet-to-Neumann operator � ;@ is selfadjoint, we have to prove that kU(t)k1 = X k�1 e�t�k <1 (2.14) for t > 0. Here k � k1 denotes the norm in the trace-class C1(@H). Then the Weyl asymptotic formula implies that there exists a bounded M and a function r(k) such that X k�1 e�t�k � X k�1 expf�t[(k=c) 1 d�1 + r(k)]g � etM X k�1 expf�t(k=c) 1 d�1 g: Here c := C(@ ;� ) and the last sum converges for any t > 0, which proves the equation (2.14). 558 Journal of Mathematical Physics, Analysis, Geometry, 2008, vol. 4, No. 4 From Laplacian Transport to Dirichlet-to-Neumann (Gibbs) Semigroups 2.5. Example: Lax Semigroups A beautiful example of explicit representation of the Dirichlet-to-Neumann semigroup (2.13) is due to Lax [Lax, Ch. 36]. Let (x) = I, and = R (see Sect. 2.2). Following [Lax] we de�ne the mapping K(t) : v(x) 7! v(e�t=R x) for any u 2 C( R); (2.15) which is a semigroup for the parameter t � 0 in the Banach space X = C( R): (K(�)K(t)v)(x) = v(e��=R e�t=R x) = v(e�(�+t)=R x) ; �; t � 0 ; x 2 R : (2.16) Remark 2.15. It is clear that if v(x) is ( = I)-harmonic in C( R), then the function: x 7! v(e�t=R x) is also harmonic. Therefore, uf (t; x) := vf (e �t=R x) = (K(t)L@ Rf)(x) = (L@ Rft)(x); x 2 R; (2.17) is the harmonic lifting of the function ft(!) := vf (e �t=R !) ; ! 2 @ R, where vf solves the problem (P1) for = I. Since in the spherical coordinates x = (r; �; ') one has @uf (t; x)=@t = �@rvf (e �t=R r; �; ')e�t=R (r=R) and @uf (t; R; �; ')=@�I = @rvf (e �t=R r; �; ')e�t=R; we get that @uf (t; !)=@t+@uf (t; !)=@�I = 0, i.e., the function (2.17) is a solution of the problem (P2). Hence, according to (2.13) and (2.17) the operator family S(t) := T@ RK(t)L@ R ; t � 0; (2.18) de�nes the Dirichlet-to-Neumann semigroup corresponding to the problem (P2) for (x) = I, and = R, which is known as the Lax semigroup. By virtue of (2.17) and (2.18) the action of this semigroup is known explicitly: (S(t)f)(!) = vf (e �t=R !); ! 2 @ R: (2.19) Notice that the semigroup relation S(�)S(t) = T@ RK(�)L@ RT@ RK(t)L@ R = S(� + t); (2.20) follows from the properties of lifting and trace operators (see Remark 2.2), from identity (2.16) and de�nition (2.18). One �nds the generator � =I;@ R of this semigroup from the limit 0 = lim t!0 sup !2@ R j 1 t (f � S(t)f)(!)� (� =I;@ Rf)(!)j (2.21) = lim t!0 sup !2@ R j 1 t (vf (R; �; ')� vf (e �t=R R; �; '))� (� =I;@ Rf)(R; �; ')j: Journal of Mathematical Physics, Analysis, Geometry, 2008, vol. 4, No. 4 559 V.A. Zagrebnov Then the operator (� =I;@ Rf)(R; �; ') = @rvf (r = R; �; ') (2.22) for any function f from the domain dom(�I;@ R) = ff 2 @C( R) : vf 2 Ker(AI;@ R) and j(@rvf ) j@ R j <1g (2.23) is identical to (2.4) for the case = I and @ = @ R. Therefore, the gene- rator (2.22) of the Lax semigroup is the Dirichlet-to-Neumann operator in this particular case of the Banach space @X = C(@ R). Similarly, we can consider the Lax semigroup (2.18) in the Hilbert space @H = L 2(@ R; d� sin � d'). Since the generator of this semigroup is a particular case of the Dirichlet-to-Neumann operator (2.11), by (2.12) and (2.10) we again obtain the corresponding action in the explicit form (S(t)f)(!) (2.24) = (e�t�I;@ R f)(!)) = 1X l=0 m=lX m=�l 1X s=0 (�t)s s! � l R �s f (R) l;m Yl;m(�; ') = 1X l=0 m=lX m=�l (e� t=R)l f (R) l;m Yl;m(�; ') = vf (e �t=R !); ! 2 @ R; which coincides with (2.19). Notice that for t > 0 the Lax semigroups have their values in the trace-class C1(@H). This explicitly follows from (2.12), i.e., from the fact that the spectrum of the semigroup generator �(�I;@ R) := f�l;m = l=Rg1 l=0;jmj�l is discrete and TrS(t) = 1X l=0 (2l + 1) e�tl=R <1: (2.25) The last is proven in the whole generality in Th. 2.14. 3. Product Approximations of Dirichlet-to-Neumann Semigroups 3.1. Approximating Family Since in contrast to the Lax semigroup ( = I) the action of the general Dirichlet-to-Neumann semigroup for 6= I is known only implicitly (2.13), it is useful to construct converging approximations, which are simpler for calculations and analysis. 560 Journal of Mathematical Physics, Analysis, Geometry, 2008, vol. 4, No. 4 From Laplacian Transport to Dirichlet-to-Neumann (Gibbs) Semigroups One of them is the Emamirad�Laadnani approximation [EmLa], which is mo- tivated by the explicit action (2.19), (2.24) of the Lax semigroup (S(t)f)(!) = (T@ RKR(t)L@ Rf)(!) = vf (e �t=R !); ! 2 @ R; (3.1) KR(t) : v(x) 7! v(e�t=R x) for any v 2 C( R) (orH( R)): The suggestion of [EmLa] consists in substituting the family fKR(t)gt�0 by the �deformed operator family K ;R(t) : v(x) 7! v(e�(t=R) (x) x) for any v 2 C( R) (orH( R)): (3.2) De�nition 3.1. For the ball R the Emamirad�Laadnani approximating family fV ;R(t) := V ;@ R(t)gt�0 is de�ned by (V ;R(t)f)(!) := (T@ RK ;R(t)L@ Rf)(!) = vf (e �(t=R) (!) !); ! 2 @ R: (3.3) Remark 3.2. (a) Notice that the approximating family (3.3) is not a semigroup (V ;R(t)V ;R(s)f)(!) = (T@ RK ;R(t)L@ R ef(s))(!) (3.4) = v ef(s) (e�(t=R) (!) !) 6= vf (e �((t+s)=R) (!) !) = (V ;R(t+ s)f)(!): (b) This family is strongly continuous at t = 0: lim t&0 V ;R(t)f = f for any f 2 @X (or @H): (3.5) (c) By de�nition (3.3) this family has the derivative at t = +0: (@tV ;R(t)f)(!) jt=0= ��(!) � (!)(rvf )(!) = �(� ;@ Rf)(!); (3.6) which for any f 2 dom(� ;@ R) coincides with the (minus) Dirichlet-to-Neumann operator (2.3). 3.2. Strong Approximation of the Dirichlet-to-Neumann Semigroups By virtue of Remark 3.2 the Emamirad�Laadnani approximation family veri- �es the conditions of the Cherno� approximation theorem ([Che, Th. 1.1]): Proposition 3.3. Let f�(s)gs�0 be a family of the linear contractions on a Banach space B and let X0 be the generator of a C0-contraction semigroup. De�ne X(s) := s �1(I � �(s)), s > 0. Then for s ! +0 the family fX(s)gs>0 converges strongly in the resolvent sense to the operator X0 if and only if the sequence f�(t=n)ngn�1, t > 0, converges strongly to e�tX0 as n ! 1 uniformly on any compact t-intervals in R1 + . Journal of Mathematical Physics, Analysis, Geometry, 2008, vol. 4, No. 4 561 V.A. Zagrebnov Notice that fV ;R(t)gt�0 in the Banach space @X is the family of contractions because of the maximum principle for the -harmonic functions vf . Since the Dirichlet-to-Neumann operator (2.3) is densely de�ned and closed, Remark 3.2 (c) implies that the family X(s) := s �1(I � V ;R(s)) converges for s ! +0 to X0 = � ;@ R in the strong resolvent sense. Similar arguments are valid for the case of the Hilbert space @H. By virtue of Remark 2.4 the Dirichlet-to-Neumann operator � ;@ is nonnegative and self- adjoint. This implies again that (3.3) is the family of contractions in @H and that by Remark 3.2 (c) the family X(s) := s �1(I � V ;R(s)) converges for s! +0 to X0 = � ;@ R in the strong resolvent sense. Resuming the above observations we obtain the strong approximation of the Dirichlet-to-Neumann semigroup U(t) Corollary 3.4. [EmLa] lim n!1 (V ;R(t=n)) n f = U(t)f; for every f 2 @X or @H; (3.7) uniformly on any compact t-intervals in (0;1). The Emamirad�Laadnani approximation theorem (Cor. 3.4) has the following important extension to more general geometry than the ball [EmLa]. De�nition 3.5. We say that a bounded smooth domain in Rd has the pro- perty of the interior ball if for any ! 2 @ there exists a tangent to @ at ! plane T!, and such that one can construct a ball tangent to T! at !, which is totally included in . If has this property, then with any point ! 2 @ , one can associate a unique point x!, which is the center of the biggest ball B(x!; r!) of radius r! included in . For any 0 < r � r!, we can construct the approximating family Vr(t) related to the ball B(xr;!; r) := fx 2 : jx � xr;!j � rg of radius r, which is centered on the line perpendicular to T! at the point ! 2 @ , i.e., xr;! = (r=r!)x! + (1� r=r!)!. Then we de�ne (V ;r(t)f)(!) := T@ vf � xr;! + e�(t=r) (!)(r �!) � : (3.8) Here �! is the outer-normal vector at !, the function vf = L@ f is the -harmonic lifting of the boundary condition f on @ , and T@ is the trace operator T@ : H1( ) 3 v 7�! v j@ 2 H 1=2(@ ): (3.9) Remark 3.6. Notice that: (a) since �! = (!�xr;!)=r, one gets (V ;r(t = 0)f)(!) := (T@ vf )(!) = f(!); 562 Journal of Mathematical Physics, Analysis, Geometry, 2008, vol. 4, No. 4 From Laplacian Transport to Dirichlet-to-Neumann (Gibbs) Semigroups (b) by virtue of (3.8) the strong derivative at t = 0 has the form (@tV ;r(t = 0)f)(!) = � (!)�! � (rvf )(!) = �(� ;@ f)(!); see (3.6). Proposition 3.7. [EmLa]. Let has the property of interior ball, and let inf !2@ fr > 0 : B(x!; r!) � g > 0; sup !2@ fr > 0 : B(x!; r!) � g <1: For any 0 < s � 1 we de�ne V ;sr! , i.e., V ;sr!f(!) = vf � xs;! + e�(t=(sr!)) (!)(sr! �!) � ; (3.10) where xs;! = sx! + (1� s)!. Then for any 0 < s � 1 lim n!1 (V ;sr!(t=n)) n f = U(t)f; for every f 2 @X or @H; (3.11) uniformly on any compact t-intervals in (0;1). Remark 3.8. By De�nition 3.1 for the ball R and the constant matrix-valued function (x) = I one obviously has V =I;R(t) = S(t) = U(t). On the other hand, for a general smooth domain with geometry verifying the conditions of Prop. 3.7, one is obliged to consider the family of approximations V ;sr! even for the homogeneous case = I. 4. Dirichlet-to-Neumann Gibbs Semigroups 4.1. Gibbs Semigroups Since by Lemma 2.14 for any Dirichlet-to-Neumann semigroup we obtain U(t > 0) 2 C1(@H), then one can check that it is in fact a Gibbs semigroup. To this end we recall the main de�nitions and some results that we need for the proof (see, e.g., [Zag2]). Let H be a separable, in�nite-dimensional complex Hilbert space. We denote by L(H) the algebra of all bounded operators on H and by C1(H) � L(H) the subspace of all compact operators. The C1(H) is a �-ideal in L(H), that is: if A 2 C1(H), then A� 2 C1(H) and if A 2 C1(H) and B 2 L(H), then AB 2 C1(H) and BA 2 C1(H). We say that a compact operator A 2 C1(H) belongs to the von Neumann�Schatten �-ideal Cp(H) for a certain 1 � p <1, if the norm kAkp := 0@X n�1 sn(A) p 1A1=p <1; (4.1) Journal of Mathematical Physics, Analysis, Geometry, 2008, vol. 4, No. 4 563 V.A. Zagrebnov where sn(A) := p �n(A�A) are the singular values of A de�ned by the eigenvalues f�n(�)gn�1 of nonnegative selfadjoint operator A�A. Since the norm kAkp is a nonincreasing function of p > 0, one gets kAk1 � kAkp � kAkq > kAk1(= kAk) (4.2) for 1 � p � q <1. Then for the von Neumann�Schatten ideals this implies the inclusions C1(H) � Cp(H) � Cq(H) � C1(H): (4.3) Let p�1 = q �1+r�1. Then, by virtue of the H�older inequality applied to (4.1), one gets kABkp � kAkqkBkr, if A 2 Cq(H) and B 2 Cr(H). Consequently, we obtain Lemma 4.1. The operator A belongs to the trace-class C1(H) if and only if there exist two (Hilbert�Schmidt) operators K1, K2 2 C2(H), such that A = K1 K2. Similarly, if K 2 Cp(H), then K p 2 C1(H). Let K be an integral operator in the Hilbert space L2(D;�). It is a Hilbert� Schmidt operator if and only if its kernel k(x; y) 2 L 2(D �D;� � �), and then one gets the estimate kKk2 � kkkL2(D�D;���). The proof is quite straightforward and can be found in, e.g., [Kat, Sim]. De�nition 4.2. [Zag2]. Let fG(t)gt�0 be a C0-semigroup on H with fG(t)gt>0 � C1(H). It is called the immediate Gibbs semigroup if G(t) 2 C1(H) for any t > 0, and it is called the eventually Gibbs semigroup if there is t0 > 0 such that G(t) 2 C1(H) for any t � t0. Remark 4.3. (a) Notice that by Lem. 4.1 any C0-semigroup such that one has fG(t)gt>0 � Cp(H) for some p <1 is an immediate Gibbs semigroup. (b) Since compact C0-semigroups are normcontinuous for any t > 0, the immediate Gibbs semigroups are k � k1-norm continuous for t > 0. For more details on the Gibbs semigroups properties we refer to the book [Zag2]. Corollary 4.4. By virtue of Prop. 2.12, Def. 4.2 and Remark 4.3 the Dirichlet-to-Neumann semigroup fU(t) = e �t� ;@ gt on the Hilbert space @H is a k � k1-holomorphic quasisectorial immediate Gibbs for Re (t) > 0. 4.2. Compact and Tr-norm Approximating Family Proposition 4.5. [EmLa] For the ball R the Emamirad�Laadnani appro- ximating family fV ;R(t)gt�0 consists of compact operators on the Banach space @X = C(@ R) for any t > 0. 564 Journal of Mathematical Physics, Analysis, Geometry, 2008, vol. 4, No. 4 From Laplacian Transport to Dirichlet-to-Neumann (Gibbs) Semigroups The proof follows from Def. 3.1 by Arzela�Ascoli criterium of compactness, since representation (3.3) and conditions on imply the uniform bound and equicontinuity of the sets fV ;R(t)(@X)gt for any t > 0. For the case of Hilbert space we recall the following useful condition for cha- racterization of the Tr-class operators [Zag2]. Proposition 4.6. If A 2 L(H) and P 1 j=1 kAejk < 1 for an orthonormal basis fejg 1 j=1 of H, then A 2 C1(H). Theorem 4.7. On the Hilbert space @H = L 2(@ R) the approximating family fV ;R(t)gt>0 � C1(@H). P r o f. Since the eigenfunctions f�kg 1 k=1 of the selfadjoint Dirichlet-to- Neumann operator � ;@ R form an orthonormal basis in L 2(@ R), we apply Prop. 4.6 for this basis. Let @ t; ;R := fx!(t) := e �(t=R) (!) !g!2@ R . By representation (3.3) and by estimate (2.9) one obtains kV ;R(t)�kk 2 = Z @ R d�(!)jv�k (x!)j 2 � j@ Rj sup !2@ R (x!; p; @ t; ;R) 2 =k 2p=(d�1) : (4.4) Then, by hypothesis (H2) on the matrix for the norm of the vector x! in R d one gets the estimate kx!k � ke �(t=R) k R � e �c1(t=R) R: Hence, for any t > 0 the dist(x!; @ R) � (1 � e �c1(t=R))R > 0, which for the estimates in (2.9) and in (4.4) implies that 0 < inf !2@ R (x!; p; @ t>0; ;R) � sup !2@ R (x!; p; @ t>0; ;R): Then, for 2p=(d � 1) > 1 the estimate (4.4) ensures the convergence of the series in the inequality kV ;R(t)k1 � 1X k=1 kV ;R(t)�kk; which �nishes the proof. Journal of Mathematical Physics, Analysis, Geometry, 2008, vol. 4, No. 4 565 V.A. Zagrebnov 5. Concluding Remarks: Trace-Norm Approximations The strong Emamirad�Laadnani approximation theorem (Cor. 3.4) and the re- sults of Sect. 4.2 proving that Dirichlet-to-Neumann semigroup U(t) and appro- ximants V ;@ (t=n) n belong to C1(@H), for all n � 1 and t > 0, motivate the following conjecture: Conjecture 5.1. [EmZa]. The Emamirad�Laadnani approximation theorem is valid in the Tr-norm topology of C1(@H). Remark 5.2. Notice that the strong approximation of the Dirichlet-to- Neumann Gibbs semigroup U(t) by the Tr-class family (V ;@ (t=n)) n does not lift automatically the topology of convergence to, e.g., operator-norm approximation [Zag2]. Therefore, to prove Conjecture 5.1 one needs additional arguments similar to those of [CaZag2]. To this end we put the di�erence in question �n(t) := (V ;@ (t=n)) n � U(t) in the following form: �n(t) = f(V ;@ (t=n)) kn � (U(t=n))kng(V ;R(t=n)) mn (5.1) + (U(t=n))knf(V ;@ (t=n)) mn � (U(t=n))mng: Here for any n > 1, we de�ne two variables kn = [n=2] and mn = [(n + 1)=2], where [x] denotes the integer part of x � 0, i.e., n = kn + mn. Then, for the estimate of �n(t) in the C1(@H)-topology one gets k�n(t)k1 � k(V ;@ (t=n)) kn � (U(t=n))knk k(V ;@ (t=n)) mnk1 (5.2) + k(U(t=n))knk1 k(V ;@ (t=n)) mn � (U(t=n))mnk: In spite of Remark 5.2, the explicit representation of approximants f(V ;@ (t=n)) ng n�1 allows to prove the corresponding operator-norm estimate. Theorem 5.3. [EmZa]. Let V ;@ R(t) be de�ned by (3.3). Then one gets the estimate k(V ;@ R(t=n)) n � U(t)k � "(n); lim n!1 "(n) = 0; (5.3) uniformly for any t-compact in R1 + . To establish (5.3) we use the "telescopic" representation (V ;@ R(t=n)) n � U(t) (5.4) = n�1X s=0 (V ;@ R(t=n)) (n�s�1) fV ;@ R(t=n)� U(t=n)g(U(t=n))s; and the operator-norm estimate of fV ;@ R(t=n)� U(t=n)g for large n. 566 Journal of Mathematical Physics, Analysis, Geometry, 2008, vol. 4, No. 4 From Laplacian Transport to Dirichlet-to-Neumann (Gibbs) Semigroups The next auxiliary result establishes a relation between the family of operators V ;@ R(t) and the Dirichlet-to-Neumann semigroup U(t). Lemma 5.4. [EmZa]. There exists a bounded operator W ;@ R(t) on @H such that V ;@ R(t) =W ;@ R(t)U(t) (5.5) for any t � 0. Now we return to the main inequality (5.2). To estimate the �rst term in the right-hand side of (5.2) we need Th. 5.3 and the Ginibre�Gruber inequality [CaZag2] k(V ;@ (t=n)) mnk1 � C U(mnt=n): To establish the latter we use representation (5.5) given by Lem. 5.4. To estimate the second term one needs only the result of Th. 5.3. All together this gives a proof of Conjecture 5.1 at least for the ball R. Acknowledgements. This paper is based on the lecture given by the author at the Lyapunov Memorial Conference, June 24-30, 2007 (Kharkiv National Uni- versity, Ukraine). I would like to express my gratitude to the organizers and in particular to Prof. Leonid A. Pastur for invitation and for support. The confe- rence talk, as well as the present account, is a part of the common project with Prof. Hassan Emamirad, whom I would like to thank for fruitful and pleasant collaboration. References [BaBr] D.C. Barber and B.H. Brown, Applied Potential Tomography. � J. Phys. E 17 (1984), 723�733. [CaZag1] V. Cachia and V.A. Zagrebnov, Operator-Norm Approximation of Semigroups by Quasisectorial Contractions. � J. Funct. Anal. 180 (2001), 176�194. [CaZag2] V. Cachia and V.A. Zagrebnov, Trotter Product Formula for Nonselfadjoint Gibbs Semigroups. � J. London Math. Soc. 64 (2001), 436�444. [Che] P.R. Cherno�, Product Formulas, Nonlinear Semigroups and Addition of Unbounded Operators. � Mem. Am. Math. Soc. 140 (1974), 1�121. [EmLa] H. Emamirad and I. Laadnani, An Approximating Family for the Dirichelet- to-Neumann Semigroup. � Adv. Di�. Eq. 11 (2006), 241�257. [EmZa] H. Emamirad and V.A. Zagrebnov, On the Gibbs Character of Dirichelet-to- Neumann Semigroups. (In preparation) [Eng] K.-J. Engel, The Laplacian on C( ) with Generalized Wenzell Boundary Conditions. � Arch. Math. 81 (2003), 548�558. Journal of Mathematical Physics, Analysis, Geometry, 2008, vol. 4, No. 4 567 V.A. Zagrebnov [Esc] J. Escher, The the Dirichelet-to-Neumann Operator on Continuous Functions. � Ann. della Scuola Norm. Sup. Pisa 21 (1994), 235�266. [GiTr] D. Gilbarg and N.S. Trudinger, Partial Di�erential Equations of Second Order. Springer-Verlag, Heidelberg, Berlin, 1983. [GrFiSap] D.S. Grebenkov, M.Filoche, and B. Sapoval, Mathematical Basis for a General Theory of Laplacian Transport Towards Irregular Interfaces. � Phys. Rev. E 73 (2006), 021103-9. [GrUl] A. Greenleaf and G. Uhlmann, Local Uniqueness for the Dirichelet-to- Neumann Map via the Two Plane Transform. � Duke Math. J. 108 (2001), 559�617. [HiLu] P. Hislop and C. Lutzer, Spectral Asymptotics of the Dirichelet-to-Neumann Map on Multiply Connected Domains in Rn . � Inverse Problems 17 (2001), 1717�1741. [Hor] L. H�ormander, The Analysis of Linear Partial Di�erential Equations IV. Springer-Verlag, New York, 1985. [Kat] T. Kato, Perturbation Theory for Linear Operators. Springer-Verlag, Heidel- berg, Berlin, 1966. [Lax] P. Lax, Functional Analysis. Wiley Interscience, New York, 2002. [LeUl] J. Lee and G. Uhlmann, Determining Anisotropic Real-Analytic Conductivi- ties by Boundary Measurement. � Comm. Pure Appl. Math. 42 (1989), 1097- 1112. [NeZag] H. Neidhardt and V.A. Zagrebnov, Trotter-Kato Product Formula and Operator-Norm Convergence. � Commun. Math. Phys. 205 (1999), 129�159. [Sap] B. Sapoval, General Formulation of Laplacian Transfer Across Irregular Su- faces. � Phys. Rev. Lett. 73 (1994), 3314�3316. [Sim] B. Simon, Trace Ideals and Their Applications. Amer. Math. Soc., Providence, RI, 2005. [Tay] M.E. Taylor, Partial Di�erential Equations II: Qualitative Studies of Linear Equations. Springer-Verlag, New York, 1996. [Tay1] M.E. Taylor, Pseudodi�erential Operators. Princeton Univ. Press, New Jersey, 1996. [Zag1] V.A. Zagrebnov, Quasi-Sectorial Contractions. � J. Funct. Anal. 254 (2008), 2503�2511. [Zag2] V.A. Zagrebnov, Gibbs Semigroups. Leuven Notes in Math. and Theoret. Physics. 10. Leuven Univ. Press, Leuven, 2002. 568 Journal of Mathematical Physics, Analysis, Geometry, 2008, vol. 4, No. 4
id nasplib_isofts_kiev_ua-123456789-106522
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
issn 1812-9471
language English
last_indexed 2025-12-07T18:52:37Z
publishDate 2008
publisher Фізико-технічний інститут низьких температур ім. Б.І. Вєркіна НАН України
record_format dspace
spelling Zagrebnov, V.A.
2016-09-29T20:26:09Z
2016-09-29T20:26:09Z
2008
From Laplacian Transport to Dirichlet-to-Neumann Gibbs) Semigroups / V.A. Zagrebnov // Журнал математической физики, анализа, геометрии. — 2008. — Т. 4, № 4. — С. 551-568. — Бібліогр.: 23 назв. — англ.
1812-9471
https://nasplib.isofts.kiev.ua/handle/123456789/106522
The paper gives a short account of some basic properties of Dirichlet-to-Neumann operators Λγ∂Ω including the corresponding semigroups motivated by the Laplacian transport in anisotropic media (γ ≠ I) and by elliptic systems with dynamical boundary conditions.
This paper is based on the lecture given by the author at the Lyapunov Memorial Conference, June 24-30, 2007 (Kharkiv National University, Ukraine). I would like to express my gratitude to the organizers and in particular to Prof. Leonid A. Pastur for invitation and for support. The conference talk, as well as the present account, is a part of the common project with Prof. Hassan Emamirad, whom I would like to thank for fruitful and pleasant collaboration.
en
Фізико-технічний інститут низьких температур ім. Б.І. Вєркіна НАН України
Журнал математической физики, анализа, геометрии
From Laplacian Transport to Dirichlet-to-Neumann Gibbs) Semigroups
Article
published earlier
spellingShingle From Laplacian Transport to Dirichlet-to-Neumann Gibbs) Semigroups
Zagrebnov, V.A.
title From Laplacian Transport to Dirichlet-to-Neumann Gibbs) Semigroups
title_full From Laplacian Transport to Dirichlet-to-Neumann Gibbs) Semigroups
title_fullStr From Laplacian Transport to Dirichlet-to-Neumann Gibbs) Semigroups
title_full_unstemmed From Laplacian Transport to Dirichlet-to-Neumann Gibbs) Semigroups
title_short From Laplacian Transport to Dirichlet-to-Neumann Gibbs) Semigroups
title_sort from laplacian transport to dirichlet-to-neumann gibbs) semigroups
url https://nasplib.isofts.kiev.ua/handle/123456789/106522
work_keys_str_mv AT zagrebnovva fromlaplaciantransporttodirichlettoneumanngibbssemigroups