Introduction Reflectionless potentials Main results
Inverse scattering for reflectionless Schrödingeroperators and generalized KdV solitons
Rostyslav Hryniv1 Yaroslav Mykytyuk2
1 Ukrainian Catholic University, Lviv, Ukraine
University of Rzeszów, Rzeszów, Poland
2 Lviv Franko National University, Lviv, Ukraine
Operator theory and Krein spaces
Vienna, 22 December 2019
Introduction Reflectionless potentials Main results
Solitary waves and Korteweg–de Vries equation
In August 1834, JOHN SCOTT RUSSEL, a Scottish civil engineer, navalarchitect and shipbuilder observed an unusual solitary wave in achannel that kept its form and velocity for a long time
He experimentally established some intriguing properties of these“waves of translation” (nowadays known as
:::::::Russel
:::::::solitary
:::::::waves)
In 1870-ies, Lord RAYLEIGH suggested theoretical background for thisphenomenon and in 1877 BOUSSINESQ derived the equation for thewave profile φ:
∂tφ+ ∂3xφ− 6φ∂xφ = 0;
it was rediscovered in 1895 by KORTEWEG and DE VRIES and isnowadays known as the Korteweg–de Vries (KdV) equation
The KdV equation is a nonlinear dispersive PDE possessing infinitelymany first integrals and many interesting properties, n-soliton solutionsbeing one of them
Introduction Reflectionless potentials Main results
Two-soliton solution
Introduction Reflectionless potentials Main results
KdV via IST
The interest in KdV was essentially revived after GARDNER, GREENE,KRUSKAL and MIURA found in 1967 that KdV can be solved by theinverse scattering transform (IST) technique:
with a solution φ(x , t) to KdV, associatea family St of Schrödinger operators on the line
St := − d2
dx2 + φ(·, t)
then the scattering data of St
SD(t) := (r(·, t), {−κ2j (t)}nj=1, {αj(t)}nj=1)
evolve along a 1st order linear flow:
r(k , t) = r(k ,0)e−8ik3t , κj(t) ≡ κj(0), αj(t) = αj(0)e4κ3j t
the initial value φ(·,0) gives SD(0)then determine SD(t) from the flow and solvethe inverse scattering problem with SD(t) to find φ(·, t)
Introduction Reflectionless potentials Main results
Solitons and IST
Solitons and Schrödinger operators:
Solitons for KdV correspond toSchrödinger operators with reflectionless potentials
Natural questions:how far the (classical) inverse scattering theory can be generalizedfor 1D Schrödinger operators with reflectionless potentials?what are the corresponding “soliton” solutions of the KdV?
Our results in a nutshell:Complete answers for integrable potentials ≡ integrable solitons
Introduction Reflectionless potentials Main results
Potential scattering for Schrödinger operators:
In quantum mechanics, the Hamiltonian of the total energy for a “light”particle (electron) in the field of the “heavy” particle (nucleus) is theSchrödinger operator
Sq := − d2
dx2 + q(x) in the space L2(R)
When the potential q is real-valued and of compact support, then theequation −ψ′′ + qψ = k2ψ at the energy k2, k ∈ R, has the Jost solution
e+(x , k) =
{eikx , x � 1a(k)eikx + b(k)e−ikx , x � −1
Introduction Reflectionless potentials Main results
The scattering solution
e+(x , k)
a(k)=
{t−(k)eikx , x � 1eikx + r−(k)e−ikx , x � −1
eikx
r−(k)e−ikx
t−(k)eikx
represents an incident wave eikx coming from −∞ which• partly reflects back to −∞ (term r−(k)e−ikx ) and• partly passes through to +∞ (term t−(k)eikx )
Introduction Reflectionless potentials Main results
Scattering data:Here r−(k) := b(k)/a(k) is the left reflection coefficient and
t−(k) := 1/a(k) is the left transmission coefficientSimilarly define right reflection r+ and transmission t+ coefficients;then get the scattering matrix
S(k) :=
(t−(k) r+(k)r−(k) t+(k)
)Properties of the scattering matrix:
unitary for real kt−(k) = t+(k) =: t(k) admits meromorphic cont. in C+
r±(−k) = r±(k), t(−k) = t(k)
S(k) uniquely determined by r+ or r− and the poles of t
Discrete spectrum data:
Eigenvalues: −κ21 < −κ2
2 < · · · < −κ2n ⇐⇒ a(iκj) = 0
Norming constants: α1, α2, . . . , αn, αj := ‖e+(·, iκj)‖
Introduction Reflectionless potentials Main results
Scattering on Faddeev–Marchenko potentials
Jost solutionsFor q ∈ L1(R, (1 + |x |)dx), the equation
−ψ′′ + qψ = k2ψ, k ∈ R,has Jost solutions e±(x , k) = e±ikx (1 + o(1)) as x → ±∞
Scattering coefficients
One then looks for a left scattering solution
ψ−(x , k) ∼{
eikx + r−(k) e−ikx if x → −∞,t−(k) eikx if x →∞
Discrete spectral dataSame as for q of compact support:
- finitely many eigenvalues −κ21 < −κ2
2 < · · · < −κ2n < 0
- corresponding norming constants α1, α2, . . . , αn
Introduction Reflectionless potentials Main results
Direct and inverse scattering for Sq
Scattering problems
Direct scattering: q 7→(
r+, (−κ2j )n
j=1, (αj)nj=1
)scattering data
Inverse scattering (ISP):(
r+, (−κ2j )n
j=1, (αj)nj=1
)7→ q
The inverse scattering problem was completely solved for q in theFaddeev–Marchenko (FM) class
i.e. real-valued q in L1(R, (1 + |x |)dx)
by MARCHENKO, GELFAND, LEVITAN, and KREIN in 1950-ies:characterized reflection coefficients;suggested an algorithm for determining q from SD
Introduction Reflectionless potentials Main results
Scattering for FM potentials
Jost solution: e+(x , k) = eikx(1 + o(1)), x →∞
transformation operator with kernel K+(x , t) s.t.
e+(x , t) = eikx +
∫ ∞x
K+(x , t)eikt dt
then K+ and
F+(s) :=1
2π
∫R
r+(k)eiks dk +∑
αje−κj s
are related via the Marchenko equation
K+(x , t) + F+(x + t) +
∫ ∞x
K+(x , s)F+(s + t) ds = 0, t > x
Introduction Reflectionless potentials Main results
Solution to the ISP for FM potentials:
Algorithm:1 given SD, construct F+
2 solve the Marchenko equation for K+
3 then q(x) = −2 ddx K+(x , x)
Tasks:justify the algorithmestablish uniquenesscharacterize scattering data (SD) for q ∈ (FM)
Completed for potentials in (FM) by V. A. Marchenkoessential contributions byL. Faddeev, I. Gelfand, B. Levitan, M. Krein, P. Deift, E. Trubowitz a.o.
Introduction Reflectionless potentials Main results
KdV and IST
finding solutions of the KdV equation
m
solving the inverse scattering problem for the Schrödinger oper.
A natural question:
How far can one generalize the IST beyond (FM)?
E.g., to include q = δ or other distributions;or to allow infinite discrete spectrum?
Introduction Reflectionless potentials Main results
Reflectionless potentials
The inverse scattering problem is exactly soluble for a class of
:::::::::::::reflectionless potentials (r± ≡ 0)
Examples of reflectionless potentials producing just one negativeeigenvalue were constructed in
V. Bargmann, On the connection between phase shifts andscattering potential, Rev. Mod. Phys. 21 (1949), 30–45.
Later in 1956, I. Kay and H. E. Moses
I. Kay and H. E. Moses, Reflectionless transmission throughdielectrics and scattering potentials, J. Appl. Phys. 27 (1956),no. 12, 1503–1508.
obtained a formula for all classical reflectionless FM potentials
Introduction Reflectionless potentials Main results
Reflectionless potentials ! soliton solutions of KdV
These reflectionless potentials have the form
q(x) = −2d2
dx2 log det(δjs + αjαs
e−(κj+κs)x
κj + κs
)1≤j,s≤n
, (1)
where (κj)nj=1 and (αj)
nj=1 are arbitrary sequences of positive numbers
the first of which is strictly decreasing
They generate the n-soliton solutions of KdV:
φ(x , t) = −2d2
dx2 log det(δjs + αjαs
e4(κ3j +κ
3s)t−(κj+κs)x
κj + κs
)1≤j,s≤n
(2)
Potentials in (1) are called classical reflectionless potentials anddenoted Qcl
Introduction Reflectionless potentials Main results
Generalized reflectionless potentials?
A natural question arises,
Can one enlarge the class Qcl to getgeneralized soliton solutions of KdV?
F. Gesztesy, W. Karwowski and Z. Zhao, Limits of soliton solutions,Duke Math. J. 68 (1992), no. 1, 101–150.
gave a certain class Q∗ of potentials, for which analogues offormula (1) for the classical reflectionless potentialsformula (2) of soliton solutions
hold true
Introduction Reflectionless potentials Main results
Gesztesy–Karwowsky–Zhao class
Namely, these potentials have the form
q(x) = −2d2
dx2 log det (I + C(x)), x ∈ R,
where C(x) is a trace class operator in `2 with matrix entries
cjs(x) := αjαse−(κj+κs)x
κj + κs, j , s ∈ N.
Here,(κj)j∈N is an arbitrary bounded sequence of pairwise distinctpositive numbers;(αj)j∈N is an arbitrary sequence of positive numbersthe trace-class condition
∑∞j=1 α
2j /κj <∞ assumed to hold
Introduction Reflectionless potentials Main results
Marchenko class
Earlier in 1985 and 1991, D. S. Lundina and V. A. Marchenko
D. S. Lundina, Teoriya Funkts., Funkts. Analiz i ikh Prilozh. 44 (1985),57–66.
V. A. Marchenko, in What is integrability?, Springer Ser. NonlinearDynam., Springer, Berlin, 1991, pp. 273–318.
studied the properties of the closure B(κ) of
B(κ) := {q ∈ Qcl | σ(Sq) ⊂ [−κ2,∞)}, κ > 0,
in the topology of uniform convergence on compact subsets of RThe elements of the set
Q :=⋃κ>0
B(κ)
are called generalized reflectionless potentials; observe that Q∗ ⊂ Q.
Introduction Reflectionless potentials Main results
Classes Qp
Set Qp := Q∩ Lp(R), 1 ≤ p ≤ ∞; then it can be shown thatQp is closed in Lp(R)
for q ∈ Q∞ \ Qcl, the negative spectrum of Sq coincides with{−κ2
j }j∈N with κj → 0
for an eigenvalue −κ2j , a norming constant αj can be defined
WLOG, take κj strictly decreasing, introduce
Q∞ \ Qcl 3 q 7→ κ(q) := (κj)j∈N,
Q∞ \ Qcl 3 q 7→ α(q) := (αj)j∈N,
and define the scattering map Υp via
Qp \ Qcl 3 q 7→ Υp(q) :=(κ(q),α(q)
)Denote by `+p the set of all positive strictly decreasing sequences in `p
Introduction Reflectionless potentials Main results
Interesting questionsIn view of the Lieb–Thirring inequality, for q ∈ Lr (R) with r ∈ [1,∞) onehas κ(q) ∈ `p with p := 2r − 1 and
‖κ(q)‖`p ≤ Cr‖q−‖Lr ,
with q− := max{−q,0} and an absolute constant Cr .For p = 1 and negative q ∈ L1(R) this can be sharpened (T. WEIDL’96,D. HUNDERTMARK A.O.’98) to
‖κ(q)‖`1 ≤12‖q‖L1
The following questions seem of importance:1 Is it true that κ(Qr ) = `+p ? If so, is there cr > 0 s.t.
cr‖q−‖Lr ≤ ‖κ(q)‖`p?2 Can one describe the isospectral set
{α(q) | q ∈ Qr , κ(q) = κ(q0)}3 Is Υp injective? If so, can one reconstruct q ∈ Qr \ Qcl from(
κ(q),α(q))?
Introduction Reflectionless potentials Main results
Main results
For p = 1, Que. 1 was answered in the paper by Gesztesy a.o.
Our aim is to give complete answers to Que. 2 and 3 for p = 1.
Theorem (Scattering map is one-to-one)
The mapping Υ1 is injective and onto `+1 × RN+.
Proofessentially uses the first FZ trace formula and relies on the followingobjects:
1 K is a positive diagonal operator in `2 with simple spectrumκ = (κj)j∈N ∈ `+1 , (i.e., K is of trace class);
2 There is a nonzero ρ = (ρj) ∈ `2 and a positive operator G : `2 → `2s.t.
KG + GK = ( · |ρ)ρ;
then ρ ∈ dom G−1/2 and ‖G−1/2ρ‖ = 2 tr K
Introduction Reflectionless potentials Main results
Generalized reflectionless potentials
Theorem (Generalized reflectionless potentials)
Under the above assumptions, let A = diag{aj} be any positive diagonaloperator in `2. Then
1 the nonnegative decreasing function
ϕ(x) := ‖AexK (A2e2xK + G)−1/2ρ‖2, x ∈ R, (3)
can be analytically continued into the strip
Π := {z = x + iy | x , y ∈ R, |y | < π2‖K‖}; (4)
2 all components ρj of ρ are nonzero;3 the function q := 2φ′ is a reflectionless potential in Q1 satisfying
κ(q) = κ, while the sequence of norming constants α(q) = (αj)satisfy αj = aj |ρj |
Introduction Reflectionless potentials Main results
Generalized soliton solutions to KdV equation
Using the above construction of generalized reflectionless potentials,one can construct generalized soliton solutions to the KdV equation, viz.
Theorem (Generalized soliton solutions)
In the above constructions, take A(t) := eK 3tA in place of A and define
ϕ(x , t) := ‖A(t)exK (A(t)2e2xK + G)−1/2ρ‖2, x ∈ R, (5)
and u(x , t) := 2∂xϕ(x , t). Then u(x , t) is a generalized soliton solutionto KdV equation s.t. u(·, t) ∈ Q1 for all t ≥ 0.
Introduction Reflectionless potentials Main results
Thank you for your attention!