Professional Documents
Culture Documents
Ebook Direct and Inverse Scattering For The Matrix Schrodinger Equation 1St Edition Tuncay Aktosun Ricardo Weder Online PDF All Chapter
Ebook Direct and Inverse Scattering For The Matrix Schrodinger Equation 1St Edition Tuncay Aktosun Ricardo Weder Online PDF All Chapter
https://ebookmeta.com/product/chess-explained-
the-c3-sicilian-1st-edition-sam-collins/
https://ebookmeta.com/product/starting-out-the-c3-sicilian-1st-
edition-john-emms/
https://ebookmeta.com/product/the-paper-issue-83-1st-edition-
origamiusa/
https://ebookmeta.com/product/structural-equation-modeling-for-
health-and-medicine-1st-edition-douglas-gunzler/
Direct Selling For Dummies Belinda Ellsworth
https://ebookmeta.com/product/direct-selling-for-dummies-belinda-
ellsworth/
https://ebookmeta.com/product/dilations-linear-matrix-
inequalities-the-matrix-cube-problem-and-beta-distributions-1st-
edition-j-william-helton/
https://ebookmeta.com/product/direct-selling-for-dummies-2nd-
edition-belinda-ellsworth/
https://ebookmeta.com/product/the-american-
revolution-1774-83-2nd-edition-daniel-marston/
https://ebookmeta.com/product/random-matrix-methods-for-machine-
learning-1st-edition-romain-couillet/
Applied Mathematical Sciences
Tuncay Aktosun
Ricardo Weder
Volume 203
Series Editors
Anthony Bloch, Department of Mathematics, University of Michigan, Ann Arbor,
MI, USA
abloch@umich.edu
C. L. Epstein, Department of Mathematics, University of Pennsylvania,
Philadelphia, PA, USA
cle@math.upenn.edu
Alain Goriely, Department of Mathematics, University of Oxford, Oxford, UK
goriely@maths.ox.ac.uk
Leslie Greengard, New York University, New York, NY, USA
Greengard@cims.nyu.edu
Advisory Editors
J. Bell, Center for Computational Sciences and Engineering, Lawrence Berkeley
National Laboratory, Berkeley, CA, USA
P. Constantin, Department of Mathematics, Princeton University, Princeton, NJ,
USA
R. Durrett, Department of Mathematics, Duke University, Durham, CA, USA
R. Kohn, Courant Institute of Mathematical Sciences, New York University,
New York, NY, USA
R. Pego, Department of Mathematical Sciences, Carnegie Mellon University,
Pittsburgh, PA, USA
L. Ryzhik, Department of Mathematics, Stanford University, Stanford, CA, USA
A. Singer, Department of Mathematics, Princeton University, Princeton, NJ, USA
A. Stevens, Department of Applied Mathematics, University of Münster, Münster,
Germany
S. Wright, Computer Sciences Department, University of Wisconsin, Madison, WI,
USA
Founding Editors
F. John, New York University, New York, NY, USA
J. P. LaSalle, Brown University, Providence, RI, USA
L. Sirovich, Brown University, Providence, RI, USA
The mathematization of all sciences, the fading of traditional scientific boundaries,
the impact of computer technology, the growing importance of computer modeling
and the necessity of scientific planning all create the need both in education and
research for books that are introductory to and abreast of these developments. The
purpose of this series is to provide such books, suitable for the user of mathematics,
the mathematician interested in applications, and the student scientist. In particular,
this series will provide an outlet for topics of immediate interest because of the nov-
elty of its treatment of an application or of mathematics being applied or lying close
to applications. These books should be accessible to readers versed in mathematics
or science and engineering, and will feature a lively tutorial style, a focus on topics
of current interest, and present clear exposition of broad appeal. A compliment to
the Applied Mathematical Sciences series is the Texts in Applied Mathematics se-
ries, which publishes textbooks suitable for advanced undergraduate and beginning
graduate courses.
This Springer imprint is published by the registered company Springer Nature Switzerland AG.
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Ricardo Weder dedicates this monograph to
Teresa, Natalie Danitza, Ricardo Eugenio,
Robert, Fernanda, Anya, Mariel, and Maxine.
Preface
vii
viii Preface
tegral equation in the class of bounded and integrable functions rather than in the
class of bounded and square-integrable functions. It may be more appropriate to
look for a solution to a homogenous Riemann–Hilbert problem in a more restricted
class rather than in a Hardy space. By having different but equivalent forms of the
characterization of the scattering data, we also understand better how different ap-
proaches and solution techniques are related to each other in solving the direct and
inverse problems.
The theory presented is supplemented by various explicitly solved examples.
Some of these examples illustrate the independence of the characterization condi-
tions by showing the failure of one characterization condition despite the satisfaction
of the remaining conditions. Some examples illustrate how Levinson’s theorem is
applied in one form of the characterization. Some other examples illustrate how the
boundary condition is reconstructed with the help of the high-energy analysis of the
scattering matrix or by using the unitarity of the generalized Fourier map.
We hope that our monograph will benefit a wide variety of readers. For exam-
ple, scientists and engineers, who are interested in using and applying the results
without worrying about the mathematical details or proofs, can benefit by reading
the summaries of the direct and inverse problems and seeing how various relevant
quantities are constructed, by viewing various characterization theorems provided,
and by looking at some of the explicitly solved examples. This can be done by
checking Sects. 2.6, 2.7, 2.8, 3.1, 5.2, and Chap. 6. Beginners in mathematics and
mathematical physics who are interested in the scattering and spectral theories of
the direct and inverse problems can try to read the entire monograph, at times by
consulting the mathematical notation and preliminaries provided in Appendix A.
Advanced researchers can at any order look at the new results presented, techniques
used, connections made among different techniques, and proofs provided.
One of the goals in our monograph is to generalize the characterization result pro-
vided by Agranovich and Marchenko [11] from the Dirichlet boundary condition to
the general boundary condition. Motivated by quantum physics, the scattering ma-
trix is traditionally normalized to one at high energies in the scalar case and to the
identity matrix in the matrix case. The same normalization in the non-Dirichlet case
becomes problematic and prevents formulating a characterization. Our monograph
clarifies this issue so that the scattering and inverse scattering problems can be ana-
lyzed properly regardless of the boundary condition used.
ix
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 An Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Notes on the Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3 Direct Scattering I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.1 Outline of the Solution to the Direct Problem . . . . . . . . . . . . . . . . . . 49
3.2 Special Solutions to the Schrödinger Equation . . . . . . . . . . . . . . . . . 51
3.3 The Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
3.4 Equivalence of the Formulations of the Boundary Condition . . . . . 94
3.5 The Quadratic Form of the Hamiltonian . . . . . . . . . . . . . . . . . . . . . . 103
3.6 Transformations of the Jost and Scattering Matrices . . . . . . . . . . . . 109
3.7 The Jost and Scattering Matrices with Zero Potential . . . . . . . . . . . 112
3.8 Low-Energy Analysis with Potentials in the Faddeev Class . . . . . . 114
3.9 Low-Energy Analysis with Potentials of Finite
Second Moment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
3.10 High-Energy Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
3.11 Bound States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
3.12 Levinson’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
3.13 Further Properties of the Scattering Data . . . . . . . . . . . . . . . . . . . . . . 226
3.14 The Marchenko Integral Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
xi
xii Contents
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 607
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 619
Abstract In this introductory chapter the goals of the monograph are described, the
contents of the remaining chapters and Appendix A are outlined, and the relevant
general references in the literature are mentioned. The direct and inverse scatter-
ing problems for the matrix Schrödinger equation on the half line with the general
self-adjoint boundary condition are viewed as two mappings, and it is indicated how
these two mappings become inverses of each other by specifying their domains ap-
propriately. The traditional definition of the scattering matrix in one way with the
Dirichlet boundary condition and in a different way with a non-Dirichlet boundary
condition is criticized, and it is indicated that such a practice makes it impossible
to formulate a well-posed inverse scattering problem unless the boundary condition
is already known and the Dirichlet and non-Dirichlet boundary conditions are not
mixed. It is emphasized that the recovery of the boundary condition should be a part
of the solution to the inverse scattering problem rather than a part of the inverse
scattering data.
1.1 An Overview
In this monograph we analyze the direct and inverse scattering problems for the
matrix Schrödinger equation on the half line with the general self-adjoint bound-
ary condition. In the direct scattering problem, given the potential and the boundary
condition we determine the scattering matrix and the bound-state information con-
sisting of the bound-state energies and the bound-state normalization matrices. In
the associated inverse problem, given the scattering matrix and the bound-state in-
formation, we determine the corresponding potential and the boundary condition.
We refer to the data set consisting of the potential and the boundary condition as the
input data set and denote it by D. We refer to the data set consisting of the scatter-
ing matrix and the bound-state information as the scattering data set and denote it
by S. Thus, we can mathematically view the direct scattering problem as the map
D : D → S with the domain D[D] and view the inverse scattering problem as the
map D : S → D with the domain D[D].
There are four aspects in the direct and inverse problems. These are the exis-
tence, uniqueness, construction, and characterization. For the existence in the direct
problem, given D in a specified class D[D] we determine whether a corresponding S
exists in some specific class. The uniqueness is concerned with whether there exists
a unique S corresponding to a given D in the domain of D, or two or more distinct
sets S may correspond to the same D. The construction deals with the recovery of
S from D. Similarly, for the existence in the inverse problem, given S in a speci-
we determine whether a corresponding D exists in some specific
fied class D[D],
class. The uniqueness in the inverse problem is concerned with whether there exists
a unique D corresponding to a given S in the domain of D, or two or more distinct
sets D may correspond to the same S. The construction in the inverse problem deals
with the recovery of D from S. The characterization problem is the most delicate
one. It consists of finding a domain D[D] for the direct map D and a domain D[D]
for the inverse map D in such a way that we have
D D[D] = D[D], D]
DD[ = D[D],
i.e., the range of D[D] under D coincides with the domain of D and that the range of
D[D] under D coincides with the domain of D. If the characterization holds then we
have D = D −1 , i.e., the map D
coincides with the inverse of the map D. Informally,
we can describe the characterization problem as the identification of a class of input
data sets and a class of scattering data sets in such a way that there is a one-to-one
correspondence between those two classes.
Our goal in this monograph is to provide a complete solution to the direct and in-
verse scattering problems for the matrix Schrödinger equation on the half line with
the general self-adjoint boundary condition and also to present a complete character-
ization. We establish our characterization result in such a way that it certainly holds
in the scalar case, it holds for any self-adjoint boundary condition, it yields the con-
struction in the corresponding direct and inverse problems, and it reveals how the
individual conditions in the characterization affect the existence, uniqueness, and
construction. Through various comments and explicit examples, we clarify various
issues arising in our characterization so that our approach can be useful in establish-
ing characterizations for some other direct and inverse problems.
The only viable characterization in the literature for the matrix Schrödinger oper-
ator on the half line can be found in the seminal work by Agranovich and Marchenko
in [8, 9, 11]. However, the analysis in [8, 9, 11] is restricted to the Dirichlet boundary
condition, and hence our study can be viewed as a generalization of the characteri-
zation given in [8, 9, 11]. It is remarkable that a characterization in the scalar case
valid for a general self-adjoint boundary condition does not exist and cannot ex-
ist in the way a scattering matrix is defined traditionally. As indicated in Section 4
of [27], as a result of defining [30, 78, 186, 192] the scattering matrix in one way
with the Dirichlet boundary condition and in a different way with a non-Dirichlet
1.1 An Overview 3
lacks the small-k analysis of the scattering data, both of which are essential for the
analysis of the relevant inverse problem. Thus, our study can also be considered as
a complement to Harmer’s work in [137–139]. In our monograph we use the results
from previous work presented in [11, 23, 24, 32, 276–278].
The theory of matrix Schrödinger equations has origins in the beginning of quan-
tum mechanics. It was realized that the matrix equations could be used to take
into account internal properties such as spin and to consider systems of particles
such as a collection of atoms, molecules, and nuclei. A well-known example is
the Pauli equation, which is the formulation of the Schrödinger equation for half-
spin particles. For further applications on this issue, we refer the reader to [8–
11, 78, 102, 172, 180, 201, 204]. In this context the matrix Schrödinger equation
on the half line appears after applying the technique of separation of variables on
a Schrödinger system with a spherically symmetric potential. Once the variables
are separated one obtains a matrix equation that only depends on the radial vari-
able. In standard quantum mechanical applications the Dirichlet boundary condition
naturally arises. This is because the Dirichlet boundary condition is related to the
finiteness of the wavefunction at the origin.
The study of quantum wires has physical relevance in the design of elementary
gates in quantum computing and in nanotubes for microscopic electronic devices.
A quantum graph is an idealization of quantum wires with very small cross sectional
areas, where such wires meet at the vertices. The quantum graph is obtained in the
zero limit of the cross sectional area of quantum wires. With the advent of quantum
graphs, the interest in matrix Schrödinger equations has been expanded drastically
and the consideration of general boundary conditions has become crucial. The ma-
trix Schrödinger equation on the half line with the general self-adjoint boundary
condition has now many applications in the scattering theory on quantum graphs
[59–61, 65, 79, 97, 98, 124, 135, 145, 161, 163, 164, 174–178, 183].
A particular phenomenon governed by a matrix Schrödinger equation is a star
graph, i.e., a one-vertex graph with a finite number of semi-infinite edges. In this
case, a homogeneous boundary condition linear in the wavefunction and its deriva-
tive is imposed at the vertex, and the dynamics on each edge is governed by the
Schrödinger equation. Physically, a star graph represents a finite number of very
thin quantum wires connected at the vertex. For example, a string of atoms may
form a star graph.
The consideration of the general self-adjoint boundary condition at a vertex
rather than only the Dirichlet boundary condition is relevant. For quantum graphs it
is crucial that the boundary condition at each vertex links the values of the wave-
function and its derivative arriving from different edges. An important case is the
Kirchhoff boundary condition, which amounts to the continuity of the wavefunction
at a vertex and also that at that vertex the sum of the derivatives of the wavefunctions
from all the edges is zero. The boundary condition at each vertex in the quantum
graph depends on how the limit is taken, and we refer the reader to Section 7.5 of
[60] for a detailed discussion of the related extensive literature. Thus, it is relevant
to consider all possible self-adjoint boundary conditions to take into account various
limiting procedures in quantum graphs.
1.1 An Overview 5
both the solution to the matrix Schrödinger equation and the spatial derivative of
that solution. Consequently, the study of the corresponding inverse problem requires
the analysis of both (3.14.1) and the derivative Marchenko equation (3.14.6). The
Dirichlet boundary condition is singled out among all self-adjoint boundary con-
ditions as the only one such that the high-energy limit S∞ of the scattering matrix
appearing in (3.10.33) is equal to −I. The general self-adjoint boundary condition
is more complicated. As elaborated in Proposition 3.15.1 and Proposition 5.2.9, in
our characterization theorems in Sects. 2.6 and 2.7 we construct the boundary ma-
trices A and B by solving (3.15.2) with the help of the input from the first two terms
in the high-energy behavior of the scattering matrix. In the proof of our character-
ization theorem in Sect. 2.8 we use a different method exploiting the unitarity of
the generalized Fourier map. The latter method of characterization does not exist in
[11].
In Chap. 3 we present the solution to the direct scattering problem for the matrix
Schrödinger equation on the half line with the general self-adjoint boundary con-
dition. Using an input data set in the Faddeev class we establish the properties of
various quantities such as the Jost solution, the Jost matrix, the physical solution,
and the regular solution. Various equivalent formulations of the general self-adjoint
boundary condition are considered. We construct the self-adjoint realizations of the
corresponding matrix Schrödinger operator. We determine the low-energy and high-
energy behavior of the Jost matrix and the scattering matrix. We show how the gen-
eral self-adjoint boundary condition is related to the high-energy behavior of the
scattering matrix, and we provide an explicit construction of the boundary matrices
A and B from the scattering data set. We analyze the bound states, and we prove
Levinson’s theorem by showing how the phase of the determinant of the scattering
matrix is related to the total number of bound states including the multiplicities. We
also analyze the Marchenko integral equation and the derivative Marchenko integral
equation.
In Chap. 4 we continue our analysis of the direct scattering problem for the ma-
trix Schrödinger equation on the half line with the general self-adjoint boundary
condition. We describe the scattering process physically and mathematically, and
we provide the definition of the scattering operator in terms of the wave operators.
We introduce the limiting absorption principle, and we show how the Hamiltonian
and its resolvent are related to the potential and the two boundary matrices A and B.
We introduce the generalized Fourier maps associated with the absolutely continu-
ous spectrum and provide their basic properties. We express the wave operators and
the scattering operator in terms of the generalized Fourier maps and their adjoints.
We prove that the scattering matrix defined in (2.2.28) in terms of the Jost matrix
coincides with the scattering matrix derived from the scattering operator. We con-
sider various other topics such as the properties of the spectral shift function, trace
formulas of Buslaev–Faddeev type, and a Bargmann–Birman–Schwinger bound on
the number of bound states.
In Chap. 5 we provide the solution to the inverse scattering problem by construct-
ing the corresponding input data set in the Faddeev class from a scattering data set
in the Marchenko class. We discuss the nonuniqueness arising in the inverse scatter-
1.1 An Overview 7
ing problem if the scattering matrix is defined one way with the Dirichlet boundary
condition and in a different way with the non-Dirichlet boundary condition, as done
traditionally. We present the solution to the inverse scattering problem by analyzing
the solvability of the Marchenko integral equation and the derivative Marchenko
integral equation. We present a proof of the main characterization result presented
in Theorem 2.6.1. By establishing the equivalents among various characterization
properties, we prove the alternate characterization theorems stated in Sect. 2.7. We
study the inverse scattering problem when the input scattering data set lacks the
bound-state information. From the given scattering matrix we show how to construct
a scattering data set belonging to the Marchenko class so that the constructed scat-
tering data set can be used as input into a properly posed inverse scattering problem.
With the help of Levinson’s theorem, we prove the two equivalent characterization
results stated in Theorems 2.7.9 and 2.7.10. We provide a proof of Parseval’s equal-
ity, expressing the completeness of the set consisting of the physical solution and
the bound-state solutions. We present the generalized Fourier map and establish its
properties. Using the generalized Fourier map we prove the characterization result
stated in Theorem 2.8.1. We also consider the characterization of the scattering data
when the potentials are restricted to the class L1p (R+ ) for p > 1 instead of only p = 1
in the Faddeev class. Finally, we formulate the characterization of the scattering data
in the special case of the purely Dirichlet boundary condition.
In Chap. 6 we illustrate the theory presented earlier via explicitly solved exam-
ples. We show how the Marchenko integral equation can yield explicitly solved ex-
amples when its kernel contains a matrix exponential and hence becomes separable.
We demonstrate the necessity of the integrability of the potential when a general
self-adjoint boundary condition is used rather than only the Dirichlet boundary con-
dition. We illustrate the characterization of the scattering data by various examples
where all the characterization conditions are satisfied or one or more of the con-
ditions are not satisfied. The examples where only one characterization condition
fails indicate the independence of the characterization conditions applied. In Exam-
ples 6.3.23, 6.3.24, 6.3.25, and 6.3.27 we consider potentials outside the Faddeev
class because they decay too slowly at infinity. These examples show that outside
the Faddeev class the uniqueness aspect of the inverse scattering problem fails when
we use as input a scattering data set in the Marchenko class. This proves that the
Faddeev class is optimal for the inverse scattering problem with the scattering data
sets in the Marchenko class. In Chap. 6 we also provide some examples to illus-
trate how a solution to the zero-energy Schrödinger equation is affected by various
restrictions on the scattering data. We illustrate the solution to the inverse scatter-
ing problem with various explicit examples, and we demonstrate how the potential,
boundary condition, and other relevant quantities are constructed from a given scat-
tering data set. We also illustrate the use of Levinson’s theorem and the generalized
Fourier map through some explicit examples.
Appendix A provides the mathematical preliminaries and it includes some ba-
sic definitions, notations, and results used in our monograph. The subject index and
the symbol index may be useful to determine the exact location of each prelim-
8 1 Introduction
inary result. The preliminaries include a brief description of Banach and Hilbert
spaces, Sobolev spaces, Hardy spaces, and some other functions spaces used in the
monograph, as well as distributions, the Fourier transform, basic inequalities, lin-
ear integral equations, and the Gronwall inequalities. A brief review is included
for particular linear operators such as those that are self-adjoint, compact, Hilbert–
Schmidt, unitary, or trace class. Also reviewed are various concepts from functional
analysis such as the resolvent, spectrum, spectral shift function, spectral theorem,
self-adjoint extensions, absolute continuity, quadratic forms, polar decomposition,
and Hermitian symplectic geometry. A glimpse of Appendix A reveals how the di-
rect and inverse scattering problems for the matrix Schrödinger equation on the half
line with the general self-adjoint boundary condition are related to a host of areas in
functional analysis and theory of ordinary differential equations.
The bibliography on the topics studied in this monograph and in related areas is vast
and we have no claim for completeness. Our aim is to provide the readers with a
general perspective of the literature so that they can see how the results presented in
this monograph are related to the direct and inverse scattering and spectral theories
and also appreciate how broad the subject is. We remark that the choice of references
listed is also influenced by our personal scientific interests.
We refer the reader to the excellent historical review by Roger Newton in the
foreword to [78]. The first attempts to obtain information on potentials from the
continuous spectrum, i.e., from the scattering data, appeared in the physics literature
in late 1930s. The scattering matrix was introduced by Wheeler [283] and Heisen-
berg [140–142]. The introduction of the scattering matrix gave the quantum me-
chanical inverse scattering problem a great impetus and importance, and the inverse
problem was started to be viewed as obtaining information about the interactions
from the scattering matrix. In [140–142] Heisenberg conjectured that all relevant
physical information was encoded in the scattering matrix, and he proposed to for-
mulate all physical theories in terms of the scattering matrix rather than, for exam-
ple, the Hamiltonian. Heisenberg’s proposal was intensively studied and it came to
be known as the S-matrix theory in high-energy physics. Heisenberg also discussed
the inverse scattering problem and made the conjecture that the uniqueness holds,
i.e., that the scattering matrix uniquely determines the potential and consequently
the scattering matrix uniquely determines the bound states. He also discussed the
construction of the potential from the scattering matrix and the characterization of
the scattering data.
The Jost solution, which plays a central role in scattering on the half line and on
the line, was introduced by Res Jost [155]. Fröberg [114] investigated the inverse
problem by starting from the radial Schrödinger equation with a central potential
and tried to construct the potential from the knowledge of one scattering phase shift
of a given angular momentum, without using the JWKB approximation. However,
1.2 Notes on the Bibliography 9
Bargmann [47, 48] proved that, in the presence of bound states, the potential in the
radial Schrödinger equation cannot uniquely be determined by the phase shift of one
angular momentum known at all positive energies. Levinson [184, 185] proved that
if there are no bound states then the uniqueness holds, and he also proved what is
now known as Levinson’s theorem, which makes the connection between the bound
states and the phase shift of a given angular momentum. Marchenko [189, 192]
considered the uniqueness question studied by Levinson and used the method of
transformation operators to investigate the relevant inverse problems.
The method of transformation operators continues to play an important role in the
inverse scattering and spectral theories. Gel’fand and Levitan [121, 122, 186, 192]
analyzed the inverse problem by using the spectral function as the input data. They
obtained a characterization of the spectral function for the Schrödinger equation
on a finite interval and on the half line by using the method of transformation op-
erators. This approach came to be known as the Gel’fand–Levitan method. Mark
Krein [165–170] used a different but related approach to study the inverse prob-
lem. Marchenko [186, 190, 192] solved the inverse scattering problem for the
Schrödinger equation on the half line via the method of transformation operators,
and he obtained a characterization of the scattering data. This approach to the solu-
tion of the inverse scattering problem came to be known as the Marchenko method.
We refer the reader to [156] for a related result and Krein’s work [171] for a different
approach. An important outcome of these general methods to solve the inverse prob-
lem is the construction of families of potentials for which the Schrödinger equation
with radially symmetric potentials has explicit solutions at a given angular momen-
tum. These are now known as Bargmann potentials [47, 48] and they correspond to
a scattering matrix S(k) which is a rational function of k.
The Gel’fand–Levitan method and the Marchenko method were soon general-
ized to the matrix cases, i.e., to analyze systems of differential equations. Newton
and Jost [204] extended the Gel’fand–Levitan method to the matrix Schrödinger
equation with the Dirichlet boundary condition, and we also mention Krein’s related
work [172]. Newton extended [201] this method to study particles with spin. The
Marchenko method was generalized to the matrix Schrödinger equation with the
Dirichlet boundary condition by Agranovich and Marchenko in [8, 9] and to parti-
cles with spin in [10]. For a detailed presentation of the results of Agranovich and
Marchenko, we refer the reader also to [11, 102]. For further results in the inverse
scattering theory for systems of differential equations on the half line, we refer the
reader to Section 3 of Chapter IX of the monograph by Chadan and Sabatier [78].
A characterization of the spectral function for the Dirac system with the Dirichlet
boundary condition was given by Gasymov and Levitan in [120], and a characteriza-
tion of the scattering data for the Dirac system with the Dirichlet boundary condition
was given by Gasymov in [117] and by Gasymov and Levitan in [119]. For the Dirac
system with a radially symmetric potential Blancarte, Grébert, and Weder obtained
[64] high-energy and low-energy estimates and proved the Parzen theorem for all
angular momenta.
Using the Marchenko method Faddeev proposed [103] a characterization of the
scattering data for the Schrödinger equation on the full line for real-valued poten-
10 1 Introduction
tials in the L11 (R) class, i.e., potentials that are integrable and that have a finite first
moment. This class of potentials is now known as the Faddeev class. However, the
validity of Faddeev’s characterization was questioned first by Chadan and Sabatier
in the 1977 edition of their monograph [77] and then by Deift and Trubowitz [83].
As a remedy, Deift and Trubowitz [83] provided a modification of Faddeev’s char-
acterization for potentials in the smaller class of L21 (R), which consists of potentials
that are integrable and that have a finite second moment. In [83] Deift and Trubowitz
raised the uncertainty issue on the continuity of the scattering matrix at zero energy
under the L11 (R) condition. This turned out to be a rather delicate issue. We remark
that Marchenko’s important monograph [192] was originally published in 1977 in
Russian, but its English translation was first published in 1986. It turned out that in
[192] Marchenko had already given a characterization in the L11 (R) class, by care-
fully avoiding making an assumption about the continuity of the scattering matrix
at zero energy. Another characterization in the L11 (R) class avoiding the continuity
of the scattering matrix at zero energy was presented by Melin [196]. Finally, Klaus
in [162] and Aktosun and Klaus in [19] settled the issue by proving the continuity
of the scattering matrix at zero energy for potentials in the L11 (R) class. A charac-
terization for potentials in the L11 (R) class, by using the Marchenko method and the
continuity of the scattering matrix at zero energy, can also be found in [19].
For the matrix Schrödinger equation on the line Aktosun, Klaus, and van der Mee
proved in [22] the continuity of the scattering matrix at zero energy for potentials in
the L11 (R) class. For the matrix Schrödinger equation on the half line, the continuity
of the scattering matrix at zero energy for potentials in the class L11 (R+ ) was proved
by Agranovich and Marchenko [11] in the case of the Dirichlet boundary condition
and by Aktosun, Klaus, and Weder [23] in the case of the general self-adjoint bound-
ary condition. We remark that the continuity result by Agranovich and Marchenko
with the Dirichlet boundary condition can be found in Theorem 5.6.2 on p. 137
of [11].
The problem of the reconstruction of the entire potential on the full line from
some appropriate scattering data, when the potential is partially known has been
studied by various researchers such as Novikova and Markushevich [214]; Aktosun,
Klaus, and van der Mee [20]; Aktosun [13, 14]; Rundell and Sacks [227]; Grébert
and Weder [132]; Gesztesy and Simon [127]; Aktosun and Weder [29]; and Akto-
sun and Papanicolaou [26]. For the inverse resonance problem where the potential
is constructed from some data containing the resonances, we refer the reader to
the studies by Marletta, Shterenberg, and Weikard [194] and Matthew and Weikard
[195] and the references quoted in the aforementioned papers. For various further
results in nonhomogeneous media, energy-dependent potentials, and other related
problems on the inverse scattering on the line, we refer the reader to the monograph
by Chadan and Sabatier [78] and the review by Aktosun and Klaus [18]. For direct
and inverse scattering problems for systems of first-order differential equations on
the line, we refer the reader to the work by Beals and Coifman [53, 54]. For the
analysis of the inverse scattering for the Dirac system on the line the reader can
consult the study by Hinton, Jordan, Klaus, Shaw [143]. For the study of the direct
1.2 Notes on the Bibliography 11
and inverse scattering for higher-order equations on the line we refer the reader to
the studies by Beals [52] and by Beals, Deift, Tomei [55].
In 1967 Gardner, Greene, Kruskal, and Miura [116] found a remarkable and as-
tonishing application of the inverse scattering problem for the Schrödinger equation
on the line to the solution of the Cauchy problem for the Korteweg–de Vries equa-
tion. There is an extensive literature on this method, which is now known as the in-
verse scattering transform, and of its relation to the theory of solitons and the theory
of integrable nonlinear evolution equations. For a presentation of these results we
refer the reader to the works by Ablowitz and Clarkson [1]; Ablowitz, Kaup, Newell,
and Segur [2]; Ablowitz and Segur [3]; Aktosun [16]; Faddeev and Takhtajan [107];
Fokas [110]; Lax [181]; Marchenko [192]; Novikov, Manakov, Pitaevskii, and Za-
kharov [213], and many references contained in these works. For the application of
the inverse scattering transform to nonlinear evolution equations on star graphs the
reader can consult the paper by Caudrelier [76] and the references therein.
For a generalization of the Gel’fand–Levitan method and the Marchenko method
to discrete systems on the half line and on the line we refer the reader to many
references such as those by Case [71–73], Case and Chiu [74], and Case and Kac
[75].
A problem closely related to an inverse scattering problem is an inverse spectral
problem. In an inverse spectral problem one uses the spectrum of a related opera-
tor to obtain information on the coefficients in an equation and the corresponding
boundary conditions. We can refer to a coefficient to be recovered as the potential
by using the analogy with the Schrödinger equation. An inverse spectral problem
can be formulated in various ways depending on the input data available. One pos-
sible input data set to use is the spectral function, which is the case [121, 122] in
the Gel’fand–Levitan method. The method by Krein [165–170] is of this nature. In
inverse spectral problems involving only a discrete spectrum, as the input data set
one can use the entire spectrum and an appropriate set of norming constants or use
the spectra corresponding to two different boundary conditions. Another possibility
is to use a mixed input data set consisting of information related to a part of the
discrete or continuous spectrum, some norming constants, and partial knowledge of
the potential. The use of mixed input data may be appropriate if one is interested in
determining a part of the spectrum, the norming constants, or the boundary condi-
tions not known a priori. In using as input the spectra from two different boundary
conditions, one can instead use a mixed input data set by combining some infor-
mation from the two spectral data sets. There are other possibilities in the choice
of input data, such as using the Weyl–Titchmarsh m-function or the spectral shift
function.
The first inverse spectral problem was probably considered by Lord Rayleigh,
who briefly discussed the possibility of constructing the variable density of a string
from the frequencies of its vibration, for which we refer the reader to p. 172 of
Section 141 in Chapter VI of [238]. However, with the development of quantum
mechanics and specially with the introduction of the Schrödinger equation in 1925,
the spectra of differential equations acquired a much greater importance and a dif-
ferent physical significance. The equations used to describe mechanical vibrations
12 1 Introduction
were applied to the dynamics of atoms and molecules and the spectrum was related
to energy levels. It is remarkable that the first result in the inverse spectral problem
for the Schrödinger equation was obtained only four years after its creation by Vic-
tor Ambartsumian [34]. Ambartsumian proved that if the eigenvalues corresponding
to the Schrödinger operator in a finite interval with the Neumann boundary condi-
tions at both ends coincide with the eigenvalues when the potential is identically
zero, then it is necessary that the potential has to be identically zero. This is actu-
ally a very special case, and Borg [66] proved that in general one spectrum is not
enough to determine the corresponding Sturm–Liouville operator. Borg considered
two Schrödinger operators on a finite interval for which the boundary condition at
one end is the same and the boundary condition at the other end is different. He
proved that if the eigenvalues for the two Schrödinger operators coincide then the
potentials and the boundary conditions must be the same for both operators.
For the Schrödinger operator on the half line, Marchenko proved [191] that the
potential and the boundary condition are uniquely determined either by the spectral
function or by the Weyl–Titchmarsh m-function. Borg [67] and Marchenko [191]
studied the Schrödinger operator on the half line with a boundary condition at the
origin when there is no continuous spectrum. They proved that two sets of dis-
crete spectra associated with two different boundary conditions at the origin (and
a fixed boundary condition at infinity if any) uniquely determine the potential and
the boundary condition at the origin. Gesztesy and Simon obtained [126] a gener-
alization of the Borg–Marchenko theorem to the case with a continuous spectrum,
and they obtained a uniqueness result by using a spectral shift function as input;
in fact, their paper contained various other uniqueness results for inverse spectral
problems. Aktosun and Weder established [30, 31] another generalization of the
Borg–Marchenko theorem to the case with a continuous spectrum by using the in-
put data given in terms of the spectral measure, and they proved that the input data
set consisting of the discrete eigenvalues for one boundary condition at the origin,
the continuous part of the spectral measure for that boundary condition, and a subset
of the discrete eigenvalues for a different boundary condition uniquely determines
the potential and the boundary conditions at the origin. Levitan and Gasymov [187]
solved the characterization problem for two spectra by providing the necessary and
sufficient conditions for two sets of real numbers to be the spectra of a Schrödinger
operator in a finite interval with different boundary conditions. Gasymov and Dzha-
biev [118] solved the characterization problem using two spectra for the Dirac equa-
tion on a finite interval.
Hochstadt and Lieberman [144] proved that if two Schrödinger operators on the
interval [0, 1] satisfy the same boundary conditions and their potentials coincide in
the interval (1/2, 1) then their potentials must also coincide in the interval (0, 1/2).
The same result and its various generalizations were proved by Gesztesy and Simon
[130] by using another method. The reader is referred to [130] for various other
results and a detailed discussion of the literature on the Hochstadt–Lieberman the-
orem. Simon introduced a new method [235] for the inverse spectral theory, and its
elaborations were given by Gesztesy and Simon [129, 131] and Ramm and Simon
[220]. One of the new results [235] obtained with this method is a local version
1.2 Notes on the Bibliography 13
tion, we note that the nonuniqueness [47, 48] indicated by Bargmann occurs when
the phase shift is known at all energies but at one angular momentum value. Thus,
Bargmann’s nonuniqueness result does not bring a conclusive answer to Heisen-
berg’s question. The breakthrough came from Faddeev [100] who proved that in
three dimensions the high-energy limit of the scattering matrix uniquely determines
the potential and hence also the bound-state energies. Faddeev considered potentials
that were not necessarily spherically symmetric, but using his result in the particular
case of potentials with spherical symmetry one concludes that the high-energy limit
of the phase shifts known for all angular momenta uniquely determines the potential
and hence also the bound-state energies. Thus, Faddeev’s result [100] proved that
Heisenberg’s conjecture [140–142] was correct.
At a formal level, modulo an assumption about exceptional points, a character-
ization for the inverse scattering problem in two or more dimensions is given by
Faddeev [106]. In this characterization there are two key elements playing a central
role. One of them is a new Green’s function known as the Faddeev Green’s function,
and the other is a new class of exponentially growing solutions known as the Fad-
deev solutions, which were previously introduced by Faddeev in [104, 105]. Using
a nonlinear ∂¯ -equation, some related characterizations in two or more dimensions
were given by Novikov and Henkin in [212] and by Weder in [246]. The charac-
terization given by Novikov and Henkin utilizes some results from several complex
variables, and the characterization by Weder is based on a limiting absorption prin-
ciple for the Faddeev Green’s function [247].
Let us mention some uniqueness and nonuniqueness results for the multidimen-
sional inverse Schrödinger scattering at fixed energy. Novikov [210] and Eskin
and Ralston [95] proved that the potential is determined uniquely by the scatter-
ing matrix at a fixed energy if that potential decays exponentially at infinity. On
the other hand, if we consider general short-range potentials decaying at infinity as
C(1 + |x|)−1−ε or faster, where C and ε are some fixed positive parameters, it is
known that the scattering matrix at a fixed energy does not uniquely determine the
potential. In fact, some examples [78] of radially oscillating potentials in three di-
mensions are known in such a way that those potentials decay proportional to |x|−2
at infinity with the scattering matrix being equal to the identity at some positive en-
ergy. In two dimensions, there are also examples [133] of potentials decaying pro-
portional to |x|−2 at infinity with the scattering matrix being equal to the identity at
some positive energy. Weder [248] proved that if two potentials have the same scat-
tering amplitude at some positive energy and they coincide outside some ball then
they necessarily coincide everywhere. In fact, the same problem also appears in dif-
ferent settings. We refer the reader to the work by Weder [254] and the references
therein for the proof that the scattering matrix at a fixed energy uniquely determines
an exponentially decreasing perturbation in stratified media. Weder [268, 269] also
proved that the scattering matrix at a fixed quasi-energy uniquely determines time-
periodic potentials that decay exponentially at infinity.
It is a remarkable fact that the uniqueness in the inverse scattering problem at a
fixed energy is restored for short-range potentials having a regular behavior at infin-
ity. It was proved by Weder and Yafaev [279] that if a short-range potential is equal
1.2 Notes on the Bibliography 15
to a sum of homogeneous terms outside a ball then the scattering matrix at a fixed
energy uniquely determines the potential. In [279] a method for the reconstruction
of the potential was also given, and more generally it was proved that if a short-
range potential is asymptotic to an infinite sum of homogenous terms at infinity
then the scattering matrix at a fixed energy uniquely determines each one of the ho-
mogeneous terms. For similar results in the case of long-range potentials, including
the Coulomb potential, we refer the reader to the work by Weder and Yafaev [280].
Similar problems in the short-range case were studied using a different method by
Joshi [152] and by Joshi and Sá Barreto [153, 154].
For more information on multidimensional inverse scattering problems, we refer
the reader to [78, 203, 211, 245] and the references therein.
There are two main approaches in the analysis of scattering problems. One of
them is the stationary approach and the other is the time-dependent approach. All
the results so far mentioned in this section related to the inverse scattering theory
have been obtained by using methods in the stationary approach. These methods are
based on the study of time-periodic solutions to the evolution equations describing
the scattering process. In a scattering process, if the particles are far apart from each
other then the interaction between them is negligible, and if the particles are close
to each other then they interact with each other. The scattering process describes
the time evolution of particles that are initially, i.e., at large negative times, far apart
from each other in the region where the interaction is weak. As time evolves the
particles come close to each other in the region where the interaction is strong. Fi-
nally, at large positive times the particles are again far apart from each other in the
region where the interaction is weak. At any time, the particles are localized in a
well-defined region of space. These two properties, i.e., the propagation in time and
the localization in space, form an essential part of any scattering process. In the sta-
tionary approach, one uses time-periodic solutions that actually do not propagate
in time and hence such solutions are not localized in space. In other words, in the
stationary approach the time-propagation and space-localization properties are lost.
Of course, these two properties are recuperated by using the Fourier analysis after
the scattering problem is solved. The main point is that these two properties are not
used in the stationary approach in the analysis of scattering problems. Nevertheless,
the stationary methods make up a powerful tool, and a lot of information in the
scattering of waves and particles is obtained by using the stationary approach.
On the other hand, the applicability of stationary methods has its limitation in
situations where it is not practical to discard the two aforementioned properties. For
example, the use of stationary methods is not practical in the scattering of three or
more particles where the configuration space is large and at each time the solution
is localized in a small region of space. Another such example is when there are
topological effects in quantum mechanics. For example, in the magnetic Aharonov–
Bohm effect a charged particle, e.g., an electron, interacts with a toroidal magnet and
it is essential to know if the electron travels through the hole of the torus or outside of
the hole. For such scattering problems it is essential to use time-dependent methods
that consider the scattering of finite-energy wave packets so that we can analyze the
two aforementioned properties in a precise manner.
16 1 Introduction
In [91–93] Enss and Weder pioneered the use of time-dependent methods (scat-
tering of wave packets) in the inverse scattering theory, and in particular they ini-
tiated a time-dependent method to estimate the high-velocity (or high-energy) be-
havior of the scattering solutions and of the scattering matrix. In this method all
the physical properties of the scattering process including the time-propagation and
space-localization are utilized in the proofs. Consequently, this method has a lot of
flexibility, which makes it possible to solve many problems by using proofs that
closely follow the physical intuition. We now briefly review some of the results that
have been obtained by using this method.
Using this method Enss and Weder [91–93] analyzed the non-relativistic
Schrödinger equation for an N-body system with long-range potentials. The Stark
effect (scattering by a decaying electric potential and an electric field that is con-
stant in space) for an N-body system was considered by Weder [250], Nicoleau
[206], Adachi and Maehara [5], and Valencia and Weder [243]. The Stark effect
was analyzed by Nicoleau [207] with a time-periodic electric field and by Adachi,
Kamada, Kazuno, and Toratani [6] with electric fields that are asymptotically zero in
time. Enss and Weder [94] analyzed the inverse problem for two-cluster scattering.
Arians [37, 38] and Jung [159] considered two-body systems with decaying elec-
tromagnetic potentials in the entire space. Arians [39] considered the inverse prob-
lem with decaying electromagnetic potentials and a constant magnetic field. Weder
[252] considered an N-body system in the entire space with time-dependent poten-
tials. The relativistic case of the Dirac equation was studied by Jung [157, 158] with
time-independent potentials and by Ito [148, 149] with time-dependent potentials.
Weder analyzed the nonlinear Schrödinger equation in [253, 255, 258–262] and the
nonlinear Klein–Gordon equation in [256, 264]. Weder [271, 272] studied the forced
nonlinear Schrödinger equation on the half line. Sandoval Romero and Weder [229]
considered a nonlinear Schrödinger equation with a potential and a nonlocal nonlin-
earity. Nicoleau [209] considered the Schrödinger equation with a strong repulsive
potential, and Nicoleau [208] studied the Schrödinger equation in a semiclassical
process. Daudé and Nicoleau [81, 82] studied the scattering in a black-hole metric.
Nicoleau [205], Weder [265], and Ballesteros and Weder [41–46] considered the
magnetic Aharonov–Bohm effect and Weder [273] studied the electric Aharonov–
Bohm effect. We refer the reader to the review papers [251, 257, 263, 266, 270, 275]
for further information.
Another approach in the inverse scattering and spectral theories is the boundary
control method introduced by Belishev [56]. For reviews of this method we refer the
reader to [57, 58] by Belishev. For the relation between the boundary control method
and Simon’s method [235], the reader can consult the paper [228] by Rybkin. The
scattering theory for nonself-adjoint operators has been extensively studied, and we
refer the reader to the studies by Falconi, Faupin, Fröhlich, and Schubnel [108], by
Faupin and Fröhlich [109], and the references in these papers. For scattering on the
line with energy-dependent potentials, the reader can consult the work by Aktosun,
Klaus, van der Mee [21] and the references therein. There is also an extensive lit-
1.2 Notes on the Bibliography 17
erature on the direct and inverse problems for canonical systems, and we refer the
reader to [40] and the references quoted there. For some additional references on
the literature for the scattering process, the reader can consult Sect. 4.1.
Chapter 2
The Matrix Schrödinger Equation and
the Characterization of the Scattering
Data
Abstract In this chapter we describe the basic ingredients of the direct and inverse
scattering problems for the matrix Schrödinger equation on the half line with the
general self-adjoint boundary condition. We show how the analysis of star graphs
and the Schrödinger scattering problem on the full line can be reduced to the study of
the matrix Schrödinger equation on the half line with some appropriate self-adjoint
boundary conditions. To analyze the direct and inverse problems on the half line,
we introduce the input data set consisting of a potential and two constant boundary
matrices describing the boundary condition. We define the Faddeev class of input
data sets by imposing some appropriate restrictions on the input data sets. We in-
troduce the scattering data set consisting of a scattering matrix and the bound-state
information. We define the Marchenko class of scattering data sets by imposing
some appropriate restrictions on the scattering data sets. The unique solutions to the
direct and inverse scattering problems are achieved by establishing a one-to-one cor-
respondence between the Faddeev class of input data sets and the Marchenko class
of scattering data sets. Various equivalent descriptions of the Marchenko class are
introduced. Such equivalent descriptions allow us to present various different but
equivalent results for the characterization of the scattering data in the solution to the
inverse problem. For the reader’s convenience various equivalent characterization
theorems are stated but their proofs are postponed until Chap. 5.
In this chapter we present the matrix Schrödinger equation on the half line contain-
ing an n × n matrix-valued potential V (x). We also introduce the general self-adjoint
boundary condition at x = 0 in terms of a pair of constant n × n matrices denoted
by A and B, which are uniquely determined up to a postmultiplication by an invert-
ible matrix. An input data set D is identified, as in (2.5.1), as the set consisting of
the potential and the two boundary matrices. The Faddeev class of input data sets
is described in Definition 2.5.1 by imposing the restrictions (2.2.2) and (2.2.3) on
the potential V (x) and the restrictions (2.2.5) and (2.2.6) on the boundary matri-
ces A and B. Corresponding to each input data set in the Faddeev class, we then
have a self-adjoint Schrödinger operator on the half line. Associated with that self-
adjoint Schrödinger operator, we introduce various relevant quantities to analyze
the corresponding direct and inverse problems. Such quantities include the Jost so-
lution f (k, x) appearing in (2.2.26), the Jost matrix J(k) appearing in (2.2.27), the
scattering matrix S(k) appearing in (2.2.28), the physical solution Ψ (k, x) appear-
ing in (2.2.29), the bound-state energies −κ 2j , the multiplicity of each bound state
m j , the normalization matrix M j , and the normalized bound-state matrix solution
Ψj (x) appearing in (2.2.30). A scattering data set S is described, as in (2.5.2), as the
set consisting of the scattering matrix, the bound-state energies, the multiplicity of
each bound state, and the normalization matrix for each bound state. The Marchenko
class of scattering data sets is described in Definition 2.5.5 by introducing the appro-
priate restrictions on the scattering data sets. A first characterization of the scattering
data sets is achieved by establishing a one-to-one correspondence between the Fad-
deev class of input data sets and the Marchenko class of scattering data sets. Several
equivalent formulations of this characterization are presented, and this is done by
introducing various lists of properties that are satisfied by a scattering data set if and
only if it belongs to the Marchenko class. Moreover, we give a characterization of
the scattering data set that differs from the other characterization formulations in the
way in which the boundary condition is obtained from the scattering data set.
The chapter is organized as follows. In Sect. 2.2 we introduce the matrix
Schrödinger equation, the self-adjoint boundary condition (2.2.4), the boundary ma-
trices A and B, and other relevant quantities used in the analysis of the corresponding
direct and inverse problems. We also show that A and B are uniquely determined up
to a postmultiplication by an invertible matrix. In Sect. 2.3 we show that the di-
rect and inverse scattering problems on a star graph can be analyzed as a special
case of the corresponding problems for the half-line matrix Schrödinger operator. In
Sect. 2.4 it is shown that the full-line Schrödinger equation can also be analyzed as a
special case of the 2×2 matrix Schrödinger equation. In Sect. 2.5, the input data sets
and the scattering data sets are introduced and the Faddeev class of input data sets is
defined. Then, various properties identified by Arabic and Roman numerals are in-
troduced as (1), (2), (3), (4), and (I), (III), (V), (VI). In fact, there are two versions
of (3) denoted by (3a ) and (3b ); there are five versions of (4) denoted by (4a ), (4b ),
(4c ), (4d ), (4e ); there are three versions of (III) denoted by (IIIa ), (IIIb ), (IIIc );
and there are eight versions of (V) denoted by (Va ), (Vb ), (Vc ), (Vd ), (Ve ), (V f ),
(Vg ), (Vh ). Having defined various properties for a scattering data set S, we are able
to formulate various characterizations for S so that it corresponds to a unique input
data set D in the Faddeev class, where we recall that the uniqueness of the bound-
ary matrices is understood up to postmultiplication by an invertible matrix. Such a
typical characterization has the form (1, 2, 3, 4), by which we mean S satisfies the
properties (1), (2), either of the two properties (3a ) or (3b ), and any one of the five
properties listed as (4a ), (4b ), (4c ), (4d ), (4e ). Another typical characterization has
2.1 Outline of the Chapter 21
the form (1, 2, III + V, 4), by which we mean S satisfies (1, 2, 4), and that instead
of satisfying either of the two properties (3a ) or (3b ), it satisfies two other proper-
ties, the first of which being one of the three properties (IIIa ), (IIIb ), (IIIc ) and the
second being one of the eight properties (Va ), (Vb ), (Vc ), (Vd ), (Ve ), (V f ), (Vg ),
(Vh ). The choice of notation enables us to streamline all the characterizations of S.
In Sect. 2.6 we provide a formulation of the characterization of a scattering data set
in the Marchenko class in the form (1, 2, 3a , 4a ). We also indicate that, by replacing
(2.2.3) with (2.6.1) in the definition of the Faddeev class and by replacing (2.5.8)
with (2.6.2) in the property (2), we obtain a characterization of the scattering data
sets in a subclass of the Marchenko class. In Sect. 2.7 we provide several formula-
tions of the characterization in the form (1, 2, III + V, 4). We then present another
◦ ◦
form of the characterization in the form (1, 2, L + 5, 4). Here, the property (L) is
introduced in Definition 2.7.7 and it is related to Levinson’s theorem; the property
◦ ◦
(5) is described in Theorem 2.7.9; and the property (4) has three versions denoted
◦ ◦ ◦
by (4c ), (4d ), (4e ), which are all described in Definition 2.7.8. Finally, in Sect. 2.8
still another characterization is presented in the form (I + VI, 2, A + V, 4), where
the property (A) is described in Theorem 2.8.1. This final characterization is based
on the solution to the inverse problem via the method summarized in Sect. 5.11 and
utilizing the generalized Fourier map studied in Chap. 5.
We note that one of the characterization formulations, namely (1, 2, 3a , 4a ), en-
ables us to define the Marchenko class of scattering data sets, which is given in
Definition 2.5.5. Let us also remark on the construction of the boundary matrices A
and B in the solution to the inverse problem. Related to the characterization formula-
tions (1, 2, 3, 4) and (1, 2, III + V, 4), we mention that the matrices A and B are con-
structed by using the data related to the high-energy limit of the scattering matrix.
On the other hand, related to the characterization formulations (I + VI, 2, A + V, 4),
the boundary matrices A and B are constructed by exploiting the unitarity of the
generalized Fourier map, which is studied in Chap. 5.
Note that some of the characterization formulations may have certain advantages
over others. For example, the characterization formulations in the form (1, 2, 3, 4)
require that we have at hand the solution to the Marchenko integral equation (3.14.1)
in order to check if (3) is satisfied. On the other hand, the characterization formula-
tions in the form (1, 2, III+V, 4) and (I+VI, 2, A+V, 4) do not require that we have
at hand the solution to the Marchenko integral equation. Having various equivalent
formulations of the characterization allows us to use various different techniques to
solve the corresponding inverse scattering problem, and it also allows us to under-
stand how all these different techniques are related to each other. For example, in
some cases it may be advantageous to solve an integral equation rather than a corre-
sponding Riemann–Hilbert problem or vice versa. It may also be more convenient
to look for a solution to an integral equation in the class of bounded and integrable
functions rather than in the class of bounded and square-integrable functions. It may
be more convenient to look for a solution to a homogenous Riemann–Hilbert prob-
lem in a more restricted class rather than in a Hardy space. In Sects. 2.6, 2.7, and 2.8
we state several theorems, each presenting a characterization formulation using a
22 2 The Matrix Schrödinger Equation and the Characterization of the Scattering Data
specific set of conditions. The equivalent but different forms of the characterization
allow us to choose a specific set of conditions among many possibilities, and this of-
fers a great flexibility in choosing the solution method to solve the inverse problem
depending on the available data type.
For the convenience of the reader, with the exception of the proof of Proposi-
tion 2.2.1, the details of the proofs in this chapter are presented in later chapters.
In this section we introduce the matrix Schrödinger equation (2.2.1), the matrix po-
tential V (x), and the boundary matrices A and B used to describe the self-adjoint
boundary condition at x = 0. In specifying the self-adjoint boundary condition,
we show that the boundary matrices A and B are uniquely determined modulo a
postmultiplication by an invertible matrix. We also introduce other main quanti-
ties associated with the corresponding direct scattering problem, i.e., the Jost solu-
tion f (k, x), the Jost matrix J(k), the scattering matrix S(k), the physical solution
Ψ (k, x), the bound-state normalization matrices M j , and the bound-state matrix so-
lutions Ψj (x).
Consider the matrix Schrödinger equation on the half line given by
We note that R+ := (0, +∞), the prime denotes the derivative with respect to the
spatial coordinate x, the scalar k2 is the complex-valued spectral parameter, the po-
tential V (x) is an n × n self-adjoint matrix-valued function of x and belongs to class
L11 (R+ ), and n is any positive integer. We assume that n is fixed and is known. The
self-adjointness of V (x) is expressed as
where the dagger denotes the matrix adjoint (complex conjugate and matrix trans-
pose). We equivalently say Hermitian to describe a self-adjoint matrix. We remark
that, unless we are in the scalar case, i.e., unless n = 1, the potential is not necessar-
ily real valued. The condition V ∈ L11 (R+ ) means that each entry of the matrix V (x)
is Lebesgue measurable in R+ and
∞
dx (1 + x) |V (x)| < +∞, (2.2.3)
0
where |V (x)| denotes the operator matrix norm. Clearly, a matrix-valued function
belongs to L11 (R+ ) if and only if each entry of that matrix belongs to L11 (R+ ).
The wavefunction ψ (k, x) appearing in (2.2.1) may be an n × n matrix-valued
function or it may be a column vector with n components. We are interested in
studying (2.2.1) with an n × n self-adjoint potential V in L11 (R+ ) under the gen-
2.2 The Matrix Schrödinger Equation on the Half Line 23
Let
1 := (Z1 , Z2 ) ∈ C2n : −B†1 Z1 + A†1 Z2 = 0 , (2.2.8)
2 := (Z1 , Z2 ) ∈ C2n : −B†2 Z1 + A†2 Z2 = 0 . (2.2.9)
A2 = A1 T, B2 = B1 T. (2.2.10)
Proof. If (2.2.10) holds, then using (2.2.10) in (2.2.8) and (2.2.9) we conclude that
1 = 2 . Let us now prove that (2.2.10) holds if 1 = 2 . Let D1 and D2 be the
operators from C2n into Cn defined as
where Ker denotes the kernel. Let us use Ran to denote the range. In order to prove
that (2.2.10) is satisfied, we first prove that Ran[D1 ] is equal to Cn . Suppose that W
is a vector in Cn orthogonal to Ran [D1 ], i.e.,
Using (2.2.7) in (2.2.16), we conclude that W = 0. Thus, we have shown that the
only vector in Cn that is orthogonal to Ran [D1 ] is the zero vector. Because Cn is the
direct sum of Ran [D1 ] and its orthogonal complement, we conclude that Ran [D1 ] =
Cn . Since the range of D1 is Cn , for any U ∈ Cn there exist Z1 and Z2 in Cn such
that
U = −B†1 Z1 + A†1 Z2 . (2.2.17)
Let us use E to denote the operator from Cn into Cn defined as
Comparing (2.2.18) and (2.2.22) we observe that E U is well defined. Let us also use
E to denote the n × n matrix, in the canonical basis, representing the linear operator
2.2 The Matrix Schrödinger Equation on the Half Line 25
E appearing in (2.2.18). The reader is referred to Sect. A.13 for the matrix repre-
sentation of a linear operator. We would like to show that the matrix E is invertible.
For this, it is enough to prove that the only solution to the linear system E U = 0 is
the zero vector in Cn , i.e., U = 0. Using the representation of U as in (2.2.17), from
(2.2.18) and E U = 0 we get
E U = −B†2 Z1 + A†2 Z2 = 0,
which indicates that (Z1 , Z2 ) ∈ Ker [D2 ]. On the other hand, as a result of (2.2.13)
we have (Z1 , Z2 ) ∈ Ker [D1 ]. Thus, with the help of (2.2.11), (2.2.12), and (2.2.17)
we obtain
U = −B†1 Z1 + A†1 Z2 = D1 (Z1 , Z2 ) = 0. (2.2.23)
As seen from (2.2.23) we have U = 0, and thus the matrix E is invertible. Letting
Z2 = 0 in (2.2.17) we get
E U = E (−B†1 Z1 ) = −B†2 Z1 , Z1 ∈ Cn ,
which implies
E B†1 = B†2 . (2.2.24)
In a similar way, letting Z1 = 0 in (2.2.17), we obtain
Next, we introduce the main quantities associated with the direct scattering prob-
lem for the matrix Schrödinger equation on the half line. The elaborations on such
quantities are provided in Chap. 3. We assume that the potential V (x) satisfies (2.2.2)
and (2.2.3) and that the boundary matrices A and B satisfy (2.2.5) and (2.2.6).
(a) Regardless of the boundary matrices A and B, the matrix Schrödinger equation
(2.2.1) possesses the n × n matrix-valued Jost solution f (k, x) satisfying the
asymptotic condition
where we recall that I is the n × n identity matrix. The existence and uniqueness
of f (k, x) as well as its relevant properties are summarized in Proposition 3.2.1.
(b) In terms of the boundary matrices A and B and the Jost solution f (k, x), the
corresponding n × n Jost matrix J(k) is given by
(c) In terms of the Jost matrix J(k) given in (2.2.27), the n × n scattering matrix
S(k) is given by
S(k) := −J(−k) J(k)−1 , k ∈ R. (2.2.28)
(d) In terms of the Jost solution f (k, x) appearing in (2.2.26) and the scattering
matrix S(k) appearing in (2.2.28), the n × n matrix-valued physical solution
Ψ (k, x) is given by
In this section we show that the half-line matrix Schrödinger equation with a
diagonal-matrix potential is unitarily equivalent to the Schrödinger equation on a
star graph. Consequently, the results that we obtain in this monograph for the direct
and inverse scattering problems on the half-line matrix Schrödinger equation also
apply to the corresponding direct and inverse problems on star graphs.
A star graph is a quantum graph with only one vertex and a finite number, say n,
of semi-infinite edges. The Hilbert space relevant to a star graph is given by
2.3 Star Graphs 27
n
H := L2 (R+ , C),
j=1
and the scalar product of two vectors Y and W in H is expressed in terms of their
sequence elements as
n
(Y, W) := ∑ (Y j ,W j )L2 (R+ ,C) . (2.3.2)
j=1
where
(LY) (x) := {−Y1 (x) +V1 (x)Y1 (x), · · · , −Yn (x) +Vn (x)Yn (x)},
with the potentials V j (x) for 1 ≤ j ≤ n being real-valued functions satisfying (2.2.3).
The relevant boundary condition for the star graph is given by
where the n × n matrices A and B satisfy (2.2.5) and (2.2.6). As in the case of the
matrix Schrödinger equation, this is the general self-adjoint boundary condition for
a star graph. For example, the Kirchhoff boundary condition corresponds to a special
case of (2.3.4) We refer the reader to [23, 32, 137–139, 163, 164] and Sect. 3.4 for
further information on the boundary condition for star graphs.
Let us now prove that a star graph is unitarily equivalent to a half-line matrix
Schrödinger operator with the diagonal-matrix potential given by
and with the boundary condition (2.2.4) involving the boundary matrices A and B
appearing in (2.3.4). We define the unitary operator U from H onto L2 (R+ ) via
⎡ ⎤
Y1 (x)
⎢ ⎥
U Y(x) := ⎣ ... ⎦ . (2.3.6)
Yn (x)
28 2 The Matrix Schrödinger Equation and the Characterization of the Scattering Data
With the help of (2.3.2), using the appropriate norms induced by the corresponding
scalar products, we get
and we stress that U is onto L2 (R+ ). Let us use ψ (x) to denote column vector
appearing on the right-hand side of (2.3.6). With the help of (2.3.1) and (2.3.5) we
observe that Y is a solution to (2.3.3) if and only if ψ is a solution to (2.2.1). In
a similar way, we see that Y satisfies the boundary condition (2.3.4) if and only if
ψ satisfies the boundary condition (2.2.4) with the same boundary matrices A and
B appearing in (2.3.4). Thus, we conclude the unitary equivalence between a star
graph and a half-line Schrödinger operator.
In this section we first show that a 2 × 2 matrix Schrödinger equation on the half line
is unitarily equivalent to a scalar Schrödinger equation on the full line with a point
interaction at x = 0. After that, we generalize this result to systems and we prove
that a 2n × 2n matrix Schrödinger equation on the half line is unitarily equivalent
to an n × n matrix Schrödinger equation on the full line with a point interaction at
x = 0. Hence, the results in this monograph apply, in particular, to the scalar and
matrix Schrödinger equations on the full line.
The relevant Hilbert space for the scalar Schrödinger equation on the full line
is L2 (R, C), where by L2 (R, C) we denote the Hilbert space of square-integrable
functions on R taking values in C. In terms of the scalar quantities ψ1 (x) and ψ2 (x),
let us define the column vector-valued quantity ψ (x) with two components as
with the superscript T denoting the matrix transpose. Next, we define the unitary
operator U from L2 (R+ ) onto L2 (R, C) as
ψ1 (x), x ≥ 0,
U ψ (x) := (2.4.1)
ψ2 (−x), x < 0.
We then have
U ψ L2 (R,C) = ψ L2 (R+ ) , ψ ∈ L2 (R+ ),
and U is onto L2 (R, C). Suppose that the matrix potential V in (2.2.1) is diagonal,
i.e.,
V (x) := diag{V1 (x),V2 (x)},
for some scalar-valued potentials V1 (x) and V2 (x). Let us use Y (x) to denote the
right-hand side of (2.4.1), i.e., we let
2.4 The Schrödinger Equation on the Full Line 29
ψ1 (x), x ≥ 0,
Y (x) :=
ψ2 (−x), x < 0,
so that we have
ψ1 (x), x > 0,
Y (x) :=
−ψ2 (−x), x < 0.
We remark that
Y (0+ ) = ψ1 (0), Y (0− ) = ψ2 (0), Y (0+ ) = ψ1 (0), Y (0− ) = −ψ2 (0).
We conclude that ψ (x) satisfies the Schrödinger equation (2.2.1) if and only Y (x)
satisfies the Schrödinger equation on the full line given by
Moreover, ψ satisfies the boundary condition (2.2.4) if and only if Y satisfies the
point-interaction condition
⎧
⎨−[B† ]11 Y (0+ ) − [B† ]12 Y (0− ) + [A† ]11 Y (0+ ) − [A† ]12 Y (0− ) = 0,
(2.4.2)
⎩
−[B† ]21 Y (0+ ) − [B† ]22 Y (0− ) + [A† ]21 Y (0+ ) − [A† ]22 Y (0− ) = 0,
where [A† ]i j and [B† ]i j denote the (i, j) entries of the matrices A† and B† , respec-
tively. For example, suppose that ψ satisfies the so-called δ -type boundary condition
where c is a real parameter, for which the special case c = 0 corresponds to the
Kirchhoff boundary condition. In this case, the matrices appearing in (2.4.2) are
given by
0 1 −1 c
A= , B= . (2.4.3)
0 1 1 0
The boundary conditions (2.4.2) corresponding to the matrices A and B given in
(2.4.3) are
Y (0− ) = Y (0+ ), Y (0+ ) −Y (0− ) = cY (0+ ),
which is related to the Schrödinger equation on the line with a δ -point interaction
with the coupling constant c, which occurs when the potential contains the compo-
nent c δ (x). In particular, when c = 0 we obtain the Schrödinger equation on the full
line without point interaction at x = 0.
30 2 The Matrix Schrödinger Equation and the Characterization of the Scattering Data
where V1 (x) and V2 (x) are some self-adjoint n × n matrices. As in the scalar case,
let us use Y (x) to denote the right-hand side of (2.4.4). Then, we see that ψ satisfies
the Schrödinger equation (2.2.1) if and only if Y satisfies the matrix Schrödinger
equation on the line given by
where Y (0) denote the common value in the first equality in (2.4.7). The point-
interaction condition in (2.4.7) corresponds to a δ -type interaction at x = 0 with the
coupling matrix Λ . If Λ = 0, which is analogous to having c = 0 in (2.4.3), then Y (x)
and Y (x) are continuous at x = 0, and in that case we obtain the matrix Schrödinger
equation on the full line without a point interaction at x = 0.
each M j is nonnegative Hermitian and has rank m j for some integer between 1 and
n. Thus, we use N to denote the number of bound states without counting the mul-
tiplicities. We remark that the possibility N = 0 is included in our consideration, in
which case the scattering data set consists of the scattering matrix alone. The finite-
ness of N is guaranteed by Theorem 3.11.1(g). The integer m j corresponds to the
multiplicity of the bound state associated with k = iκ j , and hence the integer N
defined as the sum of the ranks of the matrices M j , i.e.,
N
N := ∑ m j, (2.5.3)
j=1
We remark that the constant matrix S∞ can be obtained from the scattering
matrix S(k) via
S∞ := lim S(k). (2.5.7)
k→±∞
We also remark that, as a result of (2.5.5), the matrix S(k) − S∞ is square inte-
grable in k ∈ R and consequently Fs (y) belongs to L2 (R).
(2) For the matrix Fs (y) defined in (2.5.6), the derivative Fs (y) satisfies
∞
dy (1 + y) |Fs (y)| < +∞, (2.5.8)
0
where we recall that the norm used in (2.5.8) is the operator norm of a matrix.
We remark that for this property to hold the matrix S∞ defined in (2.5.7) should
exist and the quantity S(k) − S∞ should be square integrable in k ∈ R.
(3a ) The physical solution Ψ (k, x) to the Schrödinger equation (2.2.1) satisfies the
boundary condition (2.2.4), i.e.,
We clarify this property as follows. The scattering data set S yields a particular
n × n matrix-valued solution Ψ (k, x) to (2.2.1) known as the physical solution
given in (2.2.29) and also yields a pair of matrices A and B (modulo an invert-
ible matrix) satisfying (2.2.5) and (2.2.6). Our statement (3a ) is equivalent to
saying that (2.2.4) is satisfied if we use in (2.2.4) the quantities Ψ (k, x), A, and
B constructed from S.
(3b ) The Jost matrix J(k) satisfies
We clarify this property as follows: The scattering data set S yields a Jost matrix
J(k) constructed as in (2.2.27), unique up to a postmultiplication by an invert-
ible matrix. Using the scattering matrix S(k) given in S and the Jost matrix
constructed from S we must have (2.5.10) satisfied.
(4a ) The Marchenko equation (3.14.1) at x = 0 given by
∞
K(0, y) + F(y) + dz K(0, z) F(z + y) = 0, y ∈ R+ , (2.5.11)
0
has a unique solution K(0, y) in L1 (R+ ). Here, F(y) is the n × n matrix related
to Fs (y) given in (2.5.6) as
N
F(y) := Fs (y) + ∑ M 2j e−κ j y , y ∈ R+ . (2.5.12)
j=1
34 2 The Matrix Schrödinger Equation and the Characterization of the Scattering Data
In the case where N = 0 the sum on the right-hand side of (2.5.12) is absent and
F(y) = Fs (y) for y ∈ R+ . We remark that, in order for this property to hold, the
matrix S∞ defined in (2.5.7) should exist and that the quantity S(k) − S∞ should
be square integrable in k ∈ R.
(4b ) The only solution in L1 (R+ ) to the integral equation
∞
K(0, y) + dz K(0, z) F(z + y) = 0, y ∈ R+ , (2.5.13)
0
is the trivial solution X(y) ≡ 0. Note that F(y) appearing in (2.5.14) is the n × n
matrix defined in (2.5.12). We remark that, in order for this property to hold,
the matrix S∞ defined in (2.5.7) should exist and that the quantity S(k) − S∞
should be square integrable in k ∈ R.
(4d ) The only solution X̂(k) to the system
⎧
⎨X̂(iκ j ) M j = 0, 1 ≤ j ≤ N,
(2.5.15)
⎩
X̂(−k) + X̂(k) S(k) = 0, k ∈ R,
where X̂(k) is a row vector with n components belonging to the class L̂1 (C+ ),
is the trivial solution X̂(k) ≡ 0. For the definition of L̂1 (C+ ) we refer the reader
to Sect. A.11.
(4e ) The only solution h(k) to the system
⎧
⎨M j h(iκ j ) = 0, 1 ≤ j ≤ N,
(2.5.16)
⎩
h(−k) + S(k) h(k) = 0, k ∈ R,
Let us make some comments on the property (1) of Definition 2.5.2 above. The
property (2.5.5) implies that S(k)−S∞ is square integrable in k ∈ R, and hence it also
implies that Fs (y) belongs to L2 (R). In fact, (2.5.5) contains even more information.
The quantity G1 and hence (2.5.5) itself are used to construct the boundary matrices
A and B appearing in (2.2.4)–(2.2.6). On the other hand, for the construction of the
Another random document with
no related content on Scribd:
CHAPITRE IV
LES HAUTS PLATEAUX