REFLECTIONLESS POTENTIALS AND STOCHASTIC ANALYSIS. Setsuo TANIGUCHI. Faculty of Mathematics, Kyushu Univ.

Σχετικά έγγραφα
Brownian sheet and the reflectionless potential. Setsuo TANIGUCHI. Faculty of Math., Kyushu Univ. Fukuoka , JAPAN

D Alembert s Solution to the Wave Equation

SCHOOL OF MATHEMATICAL SCIENCES G11LMA Linear Mathematics Examination Solutions

Example Sheet 3 Solutions

SOLUTIONS TO MATH38181 EXTREME VALUES AND FINANCIAL RISK EXAM

On the Galois Group of Linear Difference-Differential Equations

Laplace Expansion. Peter McCullagh. WHOA-PSI, St Louis August, Department of Statistics University of Chicago

P AND P. P : actual probability. P : risk neutral probability. Realtionship: mutual absolute continuity P P. For example:

Homework 8 Model Solution Section

Solutions to Exercise Sheet 5


Gaussian related distributions

Teor imov r. ta matem. statist. Vip. 94, 2016, stor

12. Radon-Nikodym Theorem

ST5224: Advanced Statistical Theory II

Second Order Partial Differential Equations


Uniform Convergence of Fourier Series Michael Taylor

Parametrized Surfaces

Local Approximation with Kernels

On the summability of divergent power series solutions for certain first-order linear PDEs Masaki HIBINO (Meijo University)

Ax = b. 7x = 21. x = 21 7 = 3.

Solution Series 9. i=1 x i and i=1 x i.

Jesse Maassen and Mark Lundstrom Purdue University November 25, 2013

Partial Differential Equations in Biology The boundary element method. March 26, 2013

Empirical best prediction under area-level Poisson mixed models

Every set of first-order formulas is equivalent to an independent set

DiracDelta. Notations. Primary definition. Specific values. General characteristics. Traditional name. Traditional notation

M a t h e m a t i c a B a l k a n i c a. On Some Generalizations of Classical Integral Transforms. Nina Virchenko

Vol. 37 ( 2017 ) No. 3. J. of Math. (PRC) : A : (2017) k=1. ,, f. f + u = f φ, x 1. x n : ( ).

Statistical Inference I Locally most powerful tests

The circle theorem and related theorems for Gauss-type quadrature rules


Inverse trigonometric functions & General Solution of Trigonometric Equations

SOLUTIONS TO MATH38181 EXTREME VALUES AND FINANCIAL RISK EXAM

Appendix to On the stability of a compressible axisymmetric rotating flow in a pipe. By Z. Rusak & J. H. Lee

k A = [k, k]( )[a 1, a 2 ] = [ka 1,ka 2 ] 4For the division of two intervals of confidence in R +

Mock Exam 7. 1 Hong Kong Educational Publishing Company. Section A 1. Reference: HKDSE Math M Q2 (a) (1 + kx) n 1M + 1A = (1) =

Nowhere-zero flows Let be a digraph, Abelian group. A Γ-circulation in is a mapping : such that, where, and : tail in X, head in

Chapter 5, 6 Multiple Random Variables ENCS Probability and Stochastic Processes

Chapter 6: Systems of Linear Differential. be continuous functions on the interval

Graded Refractive-Index

ECE Spring Prof. David R. Jackson ECE Dept. Notes 2

Fractional Colorings and Zykov Products of graphs

THE SECOND ISOMORPHISM THEOREM ON ORDERED SET UNDER ANTIORDERS. Daniel A. Romano

ω ω ω ω ω ω+2 ω ω+2 + ω ω ω ω+2 + ω ω+1 ω ω+2 2 ω ω ω ω ω ω ω ω+1 ω ω2 ω ω2 + ω ω ω2 + ω ω ω ω2 + ω ω+1 ω ω2 + ω ω+1 + ω ω ω ω2 + ω

Π Ο Λ Ι Τ Ι Κ Α Κ Α Ι Σ Τ Ρ Α Τ Ι Ω Τ Ι Κ Α Γ Ε Γ Ο Ν Ο Τ Α

( y) Partial Differential Equations

MATHEMATICS. 1. If A and B are square matrices of order 3 such that A = -1, B =3, then 3AB = 1) -9 2) -27 3) -81 4) 81

Łs t r t rs tø r P r s tø PrØ rø rs tø P r s r t t r s t Ø t q s P r s tr. 2stŁ s q t q s t rt r s t s t ss s Ø r s t r t. Łs t r t t Ø t q s

C.S. 430 Assignment 6, Sample Solutions

Probability and Random Processes (Part II)

SCITECH Volume 13, Issue 2 RESEARCH ORGANISATION Published online: March 29, 2018

1. If log x 2 y 2 = a, then dy / dx = x 2 + y 2 1] xy 2] y / x. 3] x / y 4] none of these

ENGR 691/692 Section 66 (Fall 06): Machine Learning Assigned: August 30 Homework 1: Bayesian Decision Theory (solutions) Due: September 13

forms This gives Remark 1. How to remember the above formulas: Substituting these into the equation we obtain with

Mellin transforms and asymptotics: Harmonic sums

Κεφάλαιο 1 Πραγματικοί Αριθμοί 1.1 Σύνολα

6.1. Dirac Equation. Hamiltonian. Dirac Eq.

Homework for 1/27 Due 2/5

The Simply Typed Lambda Calculus

u = 0 u = ϕ t + Π) = 0 t + Π = C(t) C(t) C(t) = K K C(t) ϕ = ϕ 1 + C(t) dt Kt 2 ϕ = 0

Math 446 Homework 3 Solutions. (1). (i): Reverse triangle inequality for metrics: Let (X, d) be a metric space and let x, y, z X.

Credit Risk. Finance and Insurance - Stochastic Analysis and Practical Methods Spring School Jena, March 2009

Computable error bounds for asymptotic expansions formulas of distributions related to gamma functions

A Two-Sided Laplace Inversion Algorithm with Computable Error Bounds and Its Applications in Financial Engineering

Nonlinear Fourier transform for the conductivity equation. Visibility and Invisibility in Impedance Tomography

Solutions to Selected Homework Problems 1.26 Claim: α : S S which is 1-1 but not onto β : S S which is onto but not 1-1. y z = z y y, z S.


2 Composition. Invertible Mappings

de Rham Theorem May 10, 2016

Matrices and Determinants



Other Test Constructions: Likelihood Ratio & Bayes Tests

Fourier Series. MATH 211, Calculus II. J. Robert Buchanan. Spring Department of Mathematics

ANSWERSHEET (TOPIC = DIFFERENTIAL CALCULUS) COLLECTION #2. h 0 h h 0 h h 0 ( ) g k = g 0 + g 1 + g g 2009 =?

Math 5440 Problem Set 4 Solutions

SPECIAL FUNCTIONS and POLYNOMIALS

Differential equations

Divergence for log concave functions

w o = R 1 p. (1) R = p =. = 1

Τίτλος: Eνεργά δυναμικά στη θεωρία συναρτησιακών του πρώτου πίνακα πυκνότητας

ADVANCED STRUCTURAL MECHANICS

Προβολές και Μετασχηματισμοί Παρατήρησης

Chapter 6: Systems of Linear Differential. be continuous functions on the interval

f(w) f(z) = C f(z) = z z + h z h = h h h 0,h C f(z + h) f(z)

ΠΑΡΟΡΑΜΑΤΑ ΕΚΔΟΣΗ 12 ΜΑΡΤΙΟΥ 2018

Computing the Macdonald function for complex orders

MÉTHODES ET EXERCICES

Tridiagonal matrices. Gérard MEURANT. October, 2008

Overview. Transition Semantics. Configurations and the transition relation. Executions and computation

Homomorphism in Intuitionistic Fuzzy Automata

Exercise 1.1. Verify that if we apply GS to the coordinate basis Gauss form ds 2 = E(u, v)du 2 + 2F (u, v)dudv + G(u, v)dv 2

Takeaki Yamazaki (Toyo Univ.) 山崎丈明 ( 東洋大学 ) Oct. 24, RIMS

6. MAXIMUM LIKELIHOOD ESTIMATION

Reminders: linear functions

Aquinas College. Edexcel Mathematical formulae and statistics tables DO NOT WRITE ON THIS BOOKLET

Srednicki Chapter 55

6.642, Continuum Electromechanics, Fall 2004 Prof. Markus Zahn Lecture 8: Electrohydrodynamic and Ferrohydrodynamic Instabilities

Transcript:

REFLECTIONLESS POTENTIALS AND STOCHASTIC ANALYSIS Setsuo TANIGUCHI Faculty of Mathematics, Kyushu Univ. http://www.math.kyushu-u.ac.jp/~taniguch/ 0

PDE and Stochastic Analysis 194 K.Itô; stoch. integral, Itô s formula R.Feynman; Feynman path integral ( d ) x dx 1944 R.Cameron-W.Martin; 1 1947 M.Kac; the Feynman-Kac formula Heat eq: u = u(x, t) ( L = 1 u t = Lxu, u(, 0) = f i,j aij x i x j + ) i bi x i + V u(x, t) = E [ f(x(t, x))e Φ(x;V )] ( X(t, x):l-diff.pr. ) Reflectionless potentials, generalized... n-solitons of the KdV eq. 1

Reflectionless potential us (s S); us(x) = (d/dx) log det(i + Gs(x)), where S = {{η j, m j } 1 j n n N, η j, m j > 0, η i η j } Gs(x) = ( m i m j e (η i+η j )x η i + η j ) 1 i,j n. Schrö Op. (d/dx) + u; Scattering data Ξ 0 = {us s S}, Ξ u µ > 0, un Ξ 0 s.t. Spec( (d/dx) + un) [ µ, ), n = 1,,... un u (unif on cpts)

Σ = {σ finite meas on R with cpt supp} W = {w : [0, ) R conti, w(0) = 0} X(x) : W R: X(x, w) = w(x), w W P σ : the prob meas on W under which {X(x)} is the centered Gaussian pr with cov fn e ζ(x+y) e ζ x y X(x)X(y)dP = σ(dζ). W R ζ a 1 X(x 1 ) +... a k X(x k ) N(0, ), (x j 0) G = {P σ σ Σ} (bij) Σ ) X(x) dp σ = W R e ζx 1 σ(dζ) ζ 3

ψ(p σ )(x) = 4 d dx W ( exp 1 x ) 0 X(y) dy dp, x 0 The Plan of talk G:Cen. Gauss P σ, σ Σ () G 0 : P σ, σ = j c j δ p j ψ (1) Ξ:gen. rl. pot u unif conv. on cpts Ξ 0 : rl pot us, s S Realization of P σ, Spelling out s S, Solitons 4

reflectionless potential and n-soliton Σ 0 = { σ = n j=1 c } j δ p j n N, p j R, p j p i, c j > 0 σ Σ 0. {b(x)} x 0 ; an n-dim B.m.on (Ω, F, P ) ξσ(x) = e xd σ x 0 e yd σdb(y) (Dσ =diag[p j ]) ξ i σ (x) = exp i x 0 e yp idb i (y), i = 1,..., n Xσ(x) = c, ξσ(x) (c = (c j )) P σ (A) = P (Xσ 1 (A)) Xσ : Ω ω Xσ(, ω) W σ Σ 0 ; m < n, 1 j(1) < < j(m) n s.t. p j p j+1, p j(l) > 0, p j(l)+1 = p j(l) #{ p 1,..., pn } = n m. 5

ψ : Σ 0 σ {η j, m j } S p 1 p n 1 {η 1 < < ηn} = {p j(1),..., p j(m), r 1,..., r n m } m i = (0 < r 1 < < r n m : n j=1 c j /(p j η i c j(l)+1 c j(l) η i k i k i η k + η i η k η i η k + η i η k η i n k=1 k j(l),j(l)+1 p k + η i p k η i, r) = 1) p k + η i p k η i, if i = j(l), otherwise. 6

Thm 1. Let P σ G 0 = {P σ σ Σ 0 }. Then ( ( 4 log exp 1 x W 0 X(y) dy )dp σ) ( ) = log det I + G (x) ψ(σ) ( ) + log det I + G (0) ψ(σ) x n (p i + η i ). i=1 In particular, ψ : G 0 Ξ 0 and ψ(p σ ) = u ψ(σ). Moreover, ψ : G 0 Ξ 0 is bijective. ψ(p σ ) = u ψ(σ) on [0, ); The real analyticity does the rest of job u ψ(σ) (x) = ψ(p µ )( x), x (, 0] (µ(a) = σ( A)) 7

The τ-function of the KdV hierarchy is τ (x, t) = 1 + p j=1 m ij η ij n p=1 1 i 1 < <i p n (η ij η ik ) [ p exp ζ η 1 j<k p ij +η ij ], ik j=1 where x R, t = (t j ) R N with #{t j 0} <, {η j, m j } Ξ 0, ζ j = ζ j (x, t) = xη j + α=1 tαη j α+1. If t = (t, 0,... ), then v(x, t) = x log τ (x, t) solves the KdV eq; v t = 3 vv x + 1 4 v xxx. 8

For σ Σ 0, let ψ(σ) = {η j, m j } Ξ 0 and Iσ(x, t) = Ω [ exp 1 x 0 X σ(y) dy + 1 ] (β t D σ)ξσ(x), ξσ(x) dp, where β t = ( ( xφ)φ 1) (0, t), ζ = diag[ζ j ], φ(x, t) = U { cosh(ζ) sinh(ζ)r 1 U 1 DσU } U 1, U O(n); D σ +c c = UR U 1 (R =diag[η j ]) Thm (i) log ( Iσ(x, t) ) = (1/) log τ (x, t) +(1/) log τ (0, t) (x/) n i=1 (p i + η i ) (ii) If t = (t, 0,... ), then vσ(x, t) = 4 x log( Iσ ) is an n-soliton of the KdV eq. 9

Change of variables formulae on W; Prop 1. Let φ(y) R n n be a sol of φ Eσφ = 0, where Eσ = D σ + c c. Let x > 0 and assume (A.1) det φ(y) 0, (A.) β(y) = (φ φ 1 )(y) is symm (0 y x). Then ( Ω exp 1 x 0 X σ(y) dy + 1 ) (β(0) D σ)ξσ(x), ξσ(x) dp = ( det φ(0) ) 1/ ( e xtrdσ det φ(x) ) 1/. 10

Pf of Thm 1: φ Eσφ = 0, φ(0) = I, φ (0) = Dσ; φ(y) = cosh(yeσ 1/ ) E 1/ sinh(yeσ 1/ )Dσ (Case1) p i < p i+1, i = 1,..., n 1. φ(y) = (1/)UV R 1 B { I + G ψ(σ) (y) } e yr B 1 XC, V = diag [ (D σ r j I) 1 c 1], R = diag[η j ], a(i) = sgn [ nβ=1 (p β η i ) ], { α i b(i) = a(i) η (η α η i ) } 1/, i nβ=1 (p β η i ) B = diag[b(j)], X ij = ( p j + η i ) 1. (Case) p ε j = p j ε m i=1 δ j,l(i)+1, ε 0. 11

Pf of Thm : φ(y) = φ(y, t); (A.1),(A.) are fulfilled (Case1) p i < p i+1, i = 1,..., n 1. φ(y, t) = 1 UR 1 V B{I + A(y, t)}e ζ(y,t) B 1 XC, where A(y, t) = ( mi m j e {ζ ) i(y,t)+ζ j (y,t)}. η i + η j 1 i,j n (Case) p ε j = p j ε m i=1 δ j,l(i)+1, ε 0. 1

ch of var formulae Dw; the Lebesgue meas on W n 1 (W n = {w : [0, ) R n conti, w(0) = 0}) ( fdp = f(w) exp 1 x W n W n 0 w ) Dw ( = f w ) ( Dw exp 1 x W n 0 w Dw ) Dw = e trd/ W n f (ch var w w Dw; Volterra) ( w ) Dw Itô s formula: Dw(x), w(x) = exp( 1 Dw(x), w(x) 1 x ) 0 Dw dp (w) x 0 Dw, w + trd 13

ξσ(x, w) = e xd σ x 0 e yd σdw(y) = w(y) + e xd σ x 0 e yd σdσw(y)dy ξσ(, w Dσw) = w Apply 1 with f(w) = g(ξσ( ; w)) ( exp 1 x W n 0 X σ(y) dy + 1 ) (γ(x) D σ)ξσ(x), ξσ(x) dp = e trd σ/ ( exp 1 x W n 0 E σw, w dy + 1 )dp γ(x)w(x), w(x) 14

( 3 1 = exp 1 x W n 0 w ) Dw γ : [0, x] R n n ( = exp 1 x W n 0 w γw ) Dw. x x x γ(x)w(x), w(x) = 0 γ w, w + 0 γw, w + trγ. ( 0 1 = exp 1 x W n 0 (γ + γ )w, w + 1 ) γ(x)w(x), w(x) ( exp 1 x 0 w ) ( Dw exp 1 x ) 0 trγ ( x ) ( exp 1 0 trγ = exp 1 x W n 0 (γ + γ )w, w + 1 )dp γ(x)w(x), w(x) 4 γ + γ = Eσ, γ(x) = CH tr: γ(y) = (φ φ 1 )(x y), φ Eσφ = 0 15

Bijectivity Let u = us Ξ 0 (s S), and e + (x; ζ) be the right Jost sol of L = (d/dx) + us; Le + ( ; ζ) = ζ e + ( ; ζ), e + (x; ζ) e iζx (x ) Then λ j C (R; R), 1 j n, s.t. e + (x; ζ) = e 1ζx j Define κ : Ξ 0 Σ 0 by Then κ(s) = j ( λ j (0))δ λ j (0). ψ(κ(s)) = s, κ(ψ(σ)) = σ. ζ 1λ j (x) ζ+ 1η j. 16

generalized reflectionless potentials u Ξ µ > 0, un Ξ 0 s.t. Φσ(x) = Spec ( ( d dx ) + u n ) [ µ, ), n = 1,,... un u (unif on cpts) W ( exp 1 x ) 0 X(y) dy dp σ ψ(p σ ) = 4 ( d dx ) Φ σ G G 0 P σ ψ(p σ ) Ξ 0 Ξ : bijective Question: P σ n P σ un u, ψ(g) Ξ 17

Thm 3 (i) Φσ C ([0, )), σ Σ (ii) Let σn Σ 0, σ Σ. Suppose n N suppσ n [ β, β] ( β > 0), σn σ (vag). Then Φ (j) σ n Φ (j) σ (unif on cpts) (j = 0, 1, ). (iii) Let σ Σ and suppσ [ β, β]. Define σn(dξ) = n j= n { σ ([ jβn, (j+1)β n )) + n } δ jβ/n. Then n 0 N, λ 0 > 0 st Spec ( ( d dx ) + ψ(p σ n )) [ λ 0, ), n n 0. (iv) P σ G, u Ξ st ψ(p σ ) = u on [0, ). Conversely, u Ξ, P σ G st... u(x) = ψ(p µ )( x), x (, 0] (µ(a) = σ( A)) 18

Brownian sheet {W (p, x)} (p,x) [0, ) ; a 1 W (p 1, x 1 ) + + a k W (p k, x k ) N(0, ) W (p, x)w (q, y)dp = min{p, q} min{x, y} Ω Wiener integral s.t. L ([0, ) ) h [0, ) hdw L (P ) (isom) χ [0, ) [a,b) [c,d) dw = W (b, d) W (a, d) W (b, c) + W (a, c) d + c + a b 19

For σ = j c j δ p j and a > 0, a b < p 1, set q 0 = b + a, q k = q 0 + k j=1 p j p j 1, p 0 = b, ( W (qj,y) W (q Wq 0...q n (y) = j 1,y) ) {q j q j 1 } 1/ 1 j n. ξ a,b,σ (y) = e yd σ y 0 e zd σdwq 0...q n (z) X a,b,σ (y) = c, ξ a,b,σ (y) Xσ It holds that where X a,b,σ (y) = h(q, z; y) = n j=1 [0, ) h( ; y)dw, e (y z)pj c j χ [qj 1,q qj q j ) [0,y)(q, z). j 1 0

Gaussian filtering theory Let u Ξ and take P σ so that u = ψ(p σ ). {b(x)} x 0 : a 1-dim Bm indep of {Xσ(x)} x 0 Put Y (x) = x 0 Xσ(y)dy + b(x), Gx = σ{y (y) y x} ˆXσ(x) = E[Xσ(x) Gx] γ(x) = (X σ(x) ˆXσ(x)) dp Ω exp( 1 x Ω 0 X ) ( σ dp = exp 1 x 0 γ). K(x, y); K + 0 K(, z)r σ(z, )dz = Rσ γ(x) = K(x, x). u(x) = 4(d/dx) log ( exp( 1 x Ω 0 X ) ) σ dp = dx d ( K(x, x)). Marchenko s formula 1

quadratic Wiener functional For σ Σ 0 and symm β R n n, define qσ,x = 1 x 0 X σ(y) dy + 1 rl pot; β 0 KdV; β = β t Dσ. βξ σ(x), ξσ(x). Let A = qσ,x and Bσ(a) = Dσ + ac c. eζqσ,x dp = { j=1 (1 ζa Ω j ) } 1/ ( aj ;ev of A ) = exp ( R ( e ζx 1 ) fa (x)dx) ( f A (x) = 1 x = { e xtrd σ det [ cosh(xbσ(ζ) 1/ ) ) n;a n x>0 exp( x/a n) (ζβ + Dσ)Bσ(ζ) 1/ sinh(xbσ(ζ) 1/ ) ]} 1/.