A NONHOMOGENEOUS BACKWARD HEAT PROBLEM: REGULARIZATION AND ERROR ESTIMATES

Similar documents
An Indian Journal FULL PAPER. Trade Science Inc. A stage-structured model of a single-species with density-dependent and birth pulses ABSTRACT

On the Derivatives of Bessel and Modified Bessel Functions with Respect to the Order and the Argument

Boyce/DiPrima 9 th ed, Ch 2.1: Linear Equations; Method of Integrating Factors

A Condition for Stability in an SIR Age Structured Disease Model with Decreasing Survival Rate

On Ψ-Conditional Asymptotic Stability of First Order Non-Linear Matrix Lyapunov Systems

Elementary Differential Equations and Boundary Value Problems

CSE 245: Computer Aided Circuit Simulation and Verification

Midterm exam 2, April 7, 2009 (solutions)

Lecture 1: Numerical Integration The Trapezoidal and Simpson s Rule

2.1. Differential Equations and Solutions #3, 4, 17, 20, 24, 35

Final Exam : Solutions

Chapter 5 The Laplace Transform. x(t) input y(t) output Dynamic System

On General Solutions of First-Order Nonlinear Matrix and Scalar Ordinary Differential Equations

Lecture 4: Laplace Transforms

Spring 2006 Process Dynamics, Operations, and Control Lesson 2: Mathematics Review

Logistic equation of Human population growth (generalization to the case of reactive environment).

4.1 The Uniform Distribution Def n: A c.r.v. X has a continuous uniform distribution on [a, b] when its pdf is = 1 a x b

Let s look again at the first order linear differential equation we are attempting to solve, in its standard form:

UNIT #5 EXPONENTIAL AND LOGARITHMIC FUNCTIONS

Impulsive Differential Equations. by using the Euler Method

Homework #2: CMPT-379 Distributed on Oct 2; due on Oct 16 Anoop Sarkar

CPSC 211 Data Structures & Implementations (c) Texas A&M University [ 259] B-Trees

General Article Application of differential equation in L-R and C-R circuit analysis by classical method. Abstract

Transfer function and the Laplace transformation

Chap.3 Laplace Transform

FIRST-ORDER SYSTEMS OF ORDINARY DIFFERENTIAL EQUATIONS I: Introduction and Linear Systems

Chapter 12 Introduction To The Laplace Transform

H is equal to the surface current J S

SUFFICIENT CONDITIONS FOR EXISTENCE SOLUTION OF LINEAR TWO-POINT BOUNDARY PROBLEM IN MINIMIZATION OF QUADRATIC FUNCTIONAL

On the Speed of Heat Wave. Mihály Makai

1. Inverse Matrix 4[(3 7) (02)] 1[(0 7) (3 2)] Recall that the inverse of A is equal to:

NEW APPLICATIONS OF THE ABEL-LIOUVILLE FORMULA

UNSTEADY FLOW OF A FLUID PARTICLE SUSPENSION BETWEEN TWO PARALLEL PLATES SUDDENLY SET IN MOTION WITH SAME SPEED

I) Title: Rational Expectations and Adaptive Learning. II) Contents: Introduction to Adaptive Learning

Phys463.nb Conductivity. Another equivalent definition of the Fermi velocity is

Ali Karimpour Associate Professor Ferdowsi University of Mashhad. Reference: System Identification Theory For The User Lennart Ljung

Cramér-Rao Inequality: Let f(x; θ) be a probability density function with continuous parameter

A MATHEMATICAL MODEL FOR NATURAL COOLING OF A CUP OF TEA

CHAPTER CHAPTER14. Expectations: The Basic Tools. Prepared by: Fernando Quijano and Yvonn Quijano

Asymptotic Solutions of Fifth Order Critically Damped Nonlinear Systems with Pair Wise Equal Eigenvalues and another is Distinct

Charging of capacitor through inductor and resistor

L p -L q -Time decay estimate for solution of the Cauchy problem for hyperbolic partial differential equations of linear thermoelasticity

Deift/Zhou Steepest descent, Part I

Numerical Simulation for the 2-D Heat Equation with Derivative Boundary Conditions

A THREE COMPARTMENT MATHEMATICAL MODEL OF LIVER

Lecture 2: Current in RC circuit D.K.Pandey

XV Exponential and Logarithmic Functions

whereby we can express the phase by any one of the formulas cos ( 3 whereby we can express the phase by any one of the formulas

Application of variational iteration method for solving the nonlinear generalized Ito system

Asymptotic instability of nonlinear differential equations

5. An object moving along an x-coordinate axis with its scale measured in meters has a velocity of 6t

Arturo R. Samana* in collaboration with Carlos Bertulani*, & FranjoKrmpotic(UNLP-Argentina) *Department of Physics Texas A&M University -Commerce 07/

B) 25y e. 5. Find the second partial f. 6. Find the second partials (including the mixed partials) of

Lecture 1: Growth and decay of current in RL circuit. Growth of current in LR Circuit. D.K.Pandey

Chapter 3: Fourier Representation of Signals and LTI Systems. Chih-Wei Liu

IMPROVED HYPERBOLIC FUNCTION METHOD AND EXACT SOLUTIONS FOR VARIABLE COEFFICIENT BENJAMIN-BONA-MAHONY-BURGERS EQUATION

Lagrangian for RLC circuits using analogy with the classical mechanics concepts

Homotopy perturbation technique

Positive continuous solution of a quadratic integral equation of fractional orders

Vanishing Viscosity Method. There are another instructive and perhaps more natural discontinuous solutions of the conservation law

MEM 355 Performance Enhancement of Dynamical Systems A First Control Problem - Cruise Control

SECTION where P (cos θ, sin θ) and Q(cos θ, sin θ) are polynomials in cos θ and sin θ, provided Q is never equal to zero.

Smoking Tobacco Experiencing with Induced Death

EXISTENCE OF NON-OSCILLATORY SOLUTIONS TO FIRST-ORDER NEUTRAL DIFFERENTIAL EQUATIONS

DEPARTMENT OF ELECTRICAL &ELECTRONICS ENGINEERING SIGNALS AND SYSTEMS. Assoc. Prof. Dr. Burak Kelleci. Spring 2018

2. The Laplace Transform

3+<6,&6([DP. September 29, SID (last 5 digits): --

The Matrix Exponential

Economics 302 (Sec. 001) Intermediate Macroeconomic Theory and Policy (Spring 2011) 3/28/2012. UW Madison

Boyce/DiPrima 9 th ed, Ch 7.8: Repeated Eigenvalues

Institute of Actuaries of India

LINEAR DELAY DIFFERENTIAL EQUATION WITH A POSITIVE AND A NEGATIVE TERM

Applied Statistics and Probability for Engineers, 6 th edition October 17, 2016

The Matrix Exponential

Poisson process Markov process

ERROR ANALYSIS A.J. Pintar and D. Caspary Department of Chemical Engineering Michigan Technological University Houghton, MI September, 2012

4.3 Design of Sections for Flexure (Part II)

Instructors Solution for Assignment 3 Chapter 3: Time Domain Analysis of LTIC Systems

Double Slits in Space and Time

Stability and Bifurcation in a Neural Network Model with Two Delays

Lattice Boltzmann method for relativistic hydrodynamics: Issues. on conservation law of particle number and discontinuities

Feedback Control and Synchronization of Chaos for the Coupled Dynamos Dynamical System *

Function Spaces. a x 3. (Letting x = 1 =)) a(0) + b + c (1) = 0. Row reducing the matrix. b 1. e 4 3. e 9. >: (x = 1 =)) a(0) + b + c (1) = 0

Continous system: differential equations

The Global and Pullback Attractors for a Strongly Damped Wave Equation with Delays *

AR(1) Process. The first-order autoregressive process, AR(1) is. where e t is WN(0, σ 2 )

Voltage v(z) ~ E(z)D. We can actually get to this wave behavior by using circuit theory, w/o going into details of the EM fields!

Ordinary Differential Equations

C From Faraday's Law, the induced voltage is, C The effect of electromagnetic induction in the coil itself is called selfinduction.

Inextensible flows of S s surfaces of biharmonic

fiziks Institute for NET/JRF, GATE, IIT JAM, JEST, TIFR and GRE in PHYSICAL SCIENCES MATEMATICAL PHYSICS SOLUTIONS are

Differential Equations

Heat/Di usion Equation. 2 = 0 k constant w(x; 0) = '(x) initial condition. ( w2 2 ) t (kww x ) x + k(w x ) 2 dx. (w x ) 2 dx 0.

Properties Of Solutions To A Generalized Liénard Equation With Forcing Term

An approximate solution for nonlinear backward parabolic equations

EE 434 Lecture 22. Bipolar Device Models

The transition:transversion rate ratio vs. the T-ratio.

7.4 QUANTUM MECHANICAL TREATMENT OF FLUCTUATIONS *

S.Y. B.Sc. (IT) : Sem. III. Applied Mathematics. Q.1 Attempt the following (any THREE) [15]

Laplace Transform. National Chiao Tung University Chun-Jen Tsai 10/19/2011

Transcription:

Elcronic Journal o Dirnial Equaions, Vol. 88, No. 33, pp. 1 14. ISSN: 17-6691. URL: hp://jd.mah.xsa.du or hp://jd.mah.un.du p jd.mah.xsa.du loin: p A NONHOMOGENEOUS BACKWARD HEA PROBLEM: REGULARIZAION AND ERROR ESIMAES DANG DUC RONG, NGUYEN HUY UAN Absrac. W considr h problm o indin h iniial mpraur, rom h inal mpraur, in h nonhomonous ha quaion u u xx = x,, x,, π,, u, = uπ, =, x,, π,. his problm is known as h backward ha problm and is svrly ill-posd. Our oal is o prsn a simpl and convnin rularizaion mhod, and sharp rror simas or is approxima soluions. W illusra our rsuls wih a numrical xampl. 1. Inroducion For a posiiv numbr, w considr h problm o indin h mpraur ux,, x,, π [, ], such ha u u xx = x,, x,, π,, 1.1 u, = uπ,, x,, π [, ], 1. ux, = x, x, π. 1.3 whr x, x, z ar ivn. h problm is calld h backward ha problm, h backward Cauchy problm, or h inal valu problm. As is known, h nonhomonous problm is svrly ill-posd; i.., soluions do no always xis, and in h cas o xisnc, hs do no dpnd coninuously on h ivn daa. In ac, rom small nois conaminad physical masurmns, h corrspondin soluions hav lar rrors. I maks diicul o do numrical calculaions. Hnc, a rularizaion is in ordr. Las and Lions, in [17], rularizd h problm by addin a corrcor o h main quaion. hy considrd h problm u + Au ɛa Au =, < <, u = ϕ. Mahmaics Subjc Classiicaion. 35K5, 35K99, 47J6, 47H1. Ky words and phrass. Backward ha problm; ill-posd problm; nonhomonous ha quaion; conracion principl. c 8 xas Sa Univrsiy - San Marcos. Submid Novmbr 9, 7. Publishd March 6, 8. Suppord by h Council or Naural Scincs o Vinam. 1

D. D. RONG, N. H. UAN EJDE-8/33 Gajwski and Zaccharias [1] considrd a similar problm. hir rror sima or h approxima soluions is u ɛ u u. No ha hs sima can no b usd a im =. In 1983, Showalr, prsnd a dirn mhod calld h quasiboundary valu QBV mhod o rulariz ha linar homonous problm which av a sabiliy sima br han h on o discussd mhod. h main idas o h mhod is o addin an appropria corrcor ino h inal daa. Usin h mhod, Clark and Oppnhimr, in [4], and Dnch-Bssila, vry rcnly in [5], rularizd h backward problm by rplacin h inal condiion by and u + ɛu = 1.4 u ɛu = 1.5 rspcivly. Alhouh hr ar many paprs on h linar homonous cas o h backward problm, w only ind a w paprs on h nonhomonous cas, such in [8, 9]. In 6, ron and uan [8], approximad h problm 1.1 1.3 by h quasi-rvrsibiliy mhod. Howvr, h sabiliy maniud o h mhod is o ordr ɛ. Morovr, h rror bwn h approxima problm and h xac soluion is 8 ɛ 4 u., + 4 x, x 4 L, ;L,π, s [5, pa 5] which is vry lar whn ɛ ixd and is small nd o zro. Vry rcnly, in [9], h auhors usd an improvd vrsion o QBV mhod o rulariz problm in on dimnsional o 1.1 1.3 in h nonlinar cas o uncion. Howvr, in [9], h auhors can only sima h rror in h cas which, h inal valu saisis h condiion k k < 1.6 k=1 s [9, pa 4]. h uncions saisyin his condiion ar qui scarc and so his mhod is no usul o considr many nonhomonous backward problm in h anohr cas o inal valu, which h condiion 1.6 is no saisid or uncions such as x = a, whr a is consan. W also no ha h rror bwn h approxima problm and h xac soluion is Cɛ, which is no nar o zro, i ɛ ixd and nd o zro. Hnc, h convrnc o h approxima soluion is vry slow whn is nar o h oriinal im. In h prsn papr, w shall rulariz his problm 1.1 1.3by prurbin h inal valu wih nw way, which is dirn h ways in 1.4and 1.5. W approxima problm by h ollowin problm

EJDE-8/33 REGULARIZAION AND ERROR ESIMAES 3 u ɛ u ɛ xx = whr < ɛ < 1, ɛp + p sinpx, x,, π,, 1.7 u ɛ, = u ɛ π, = x,, π [, ] 1.8 u ɛ x, = ɛp + p sinpx, x, π 1.9 p = π x, sinpxdx, p = π x sinpxdx 1.1 and, is h innr produc in L, π. W shall prov ha, h uniqu soluion u ɛ o 1.7 1.9 saisis h ollowin qualiy u ɛ x, = ɛp + p s ɛp + p sds sinpx 1.11 whr. No ha our mhod iv a br approximaion han h quasi-rvrsibiliy mhod in [8], and h inal valu x is no ssnial o saisy h condiion *, which only in L, π. Espcially,h convrnc o h approxima soluion a = is also provd In [9], h rror u., u ɛ., is no ivn. his is an improvmn o many known rsuls in [1, 4, 5, 9, 1, 8, 9, 3, 31]. h rmaindr o h papr is dividd ino hr scions. In Scion 1, w shall show ha 1.7 1.9 is wll posd and ha h soluion u ɛ x, saisis 1.11. hn, in Scion, w sima h rror bwn an xac soluion u o Problm 1.1 1.3 and h approximaion soluion u ɛ. In ac, w shall prov ha u ɛ., u., whr is norm in L, π and C dpnds on u and. Finally, a numrical xprimn will b ivn in Scion 3. C 1 + ln ɛ 1.1. Wll-posdnss o Problm 1.7 1.9 In his scion, w shall sudy h xisnc, h uniqunss and h sabiliy o a wak soluion o Problm 1.7 1.9. horm.1. L x, L, ; L, π and x L, π. L a ivn ɛ,. hn 1.7 1.9 has a uniqu wak soluion u ɛ C[, ]; L, π L, ; H 1, π C 1, ; H 1, π saisyin 1.11. h soluion dpnds coninuously on in C[, ]; L, π. Proo. h proo is dividd ino wo sps. In Sp 1, w prov h xisnc and h uniqunss o a soluion o 1.7 1.9. In Sp, h sabiliy o h soluion is ivn. Sp 1. h xisnc and h uniqunss o a soluion o 1.7 1.9 W divid his sp ino wo pars.

4 D. D. RONG, N. H. UAN EJDE-8/33 Par A I u ɛ C[, ]; L, π L, ; H 1, π C 1, ; H 1, π saisis 1.1 hn u ɛ is soluion o 1.7 1.9. W hav u ɛ s x, = ɛp + p ɛp + p sds sinpx.1 or. W can vriy dircly ha u ɛ C[, ]; L, π C 1, ; H 1, π L, ; H 1, π. In ac, u ɛ C, ]; H 1, π. Morovr, on has u ɛ x, p = ɛp + p p s ɛp + p sds + ɛp + p sinpx = p u ɛ x,, sin px sinpx + π ɛp + p sinpx and = u ɛ xxx, + ɛp + p sinpx u ɛ x, = ɛp + p sinpx So u ɛ is h soluion o 1.7 1.9. Par B h Problm 1.7 1.9 has a mos on soluion C[, ]; H 1, π C 1, ; L, π. A proo o his samn can b ound in [3, horm 11]. Sinc Par A and Par B ar provd, w compl h proo o Sp 1. Sp. h soluion o h problm 1.7 1.9 dpnds coninuously on in L, π. L u and v b wo soluions o 1.7 1.9 corrspondin o h inal valus and h. From w hav s ux, = ɛp + p ɛp + p sds sinpx, vx, = whr h ɛp + p p = x sinpxdx, π For λ >, w din h uncion hn hλ =. s ɛp + p sds sinpx, h p = π 1 ɛλ + λ. hx sinpxdx. ln/ɛ hλ h = ɛ ɛ,. 1 + ln/ɛ.3

EJDE-8/33 REGULARIZAION AND ERROR ESIMAES 5 his ollows ha u., v., = π ɛp + p h p π p h p ɛ 1 + ln/ɛ = h. ɛ 1 + ln/ɛ Hnc u., v., h. ɛ 1 + ln/ɛ his compls h proo o Sp and h proo o our horm..4 Rmark.. In [9, 1, 8], h sabiliy maniud is ɛ s [5, horm.1], i is ɛ 1. On advana o his mhod o rularizaion is ha h ordr o h rror, inroducd by small chans in h inal valu, is lss han h ordr ivn in [8]. horm.3. For any x L, π, h approximaion u ɛ x, convrs o x in L, π as ɛ nds o zro. Proo. W hav x = p sinpx, whr p is dind in 1.1. L α >, choos som N or which π p=n+1 p < α/. W hav hn u ɛ x, x = π ɛ p 4 p ɛp +..5 N u ɛ x, x ɛ π p 4 p + α By akin ɛ such ha ɛ < α π N p4 p 1/, w u ɛ x, x < α which compls h proo. In h cas d dx L, π, w hav h rror sima u ɛ x, x = π 1 ɛp + p hn, w his compls h proo. = π π ɛ p 4 p ɛp + 1 + ln ɛ ux, x p 4 p = 1 + ln/ɛ xx 1 + ln/ɛ xx.

6 D. D. RONG, N. H. UAN EJDE-8/33 horm.4. L x, ɛ L, π b as in horm.3, and l xx b in L, ; L, π. I h squnc u ɛ x, convrs in L, π, hn h problm 1.1 1.3 has a uniqu soluion u. Furhrmor,w hn hav ha u ɛ x, convrs o u as ɛ nds o zro uniormly in. Proo. Assum ha lim ɛ u ɛ x, = u x xiss. L ux, = u p s p sds sinpx whr u p = π π u x sinpxdx. I is clar o s ha ux, saisis 1.1 1.. W hav h ormula o u ɛ x, whr u ɛ p = π w hav u ɛ x, = u ɛ p s ɛp + p sds sinpx uɛ x, sinpxdx. In viw o h inqualiy a + b a + b, u ɛ x, ux, π u ɛ p u p + π u ɛ x, u x + 1 + ln/ɛ 4 = u ɛ x, u x + 1 + ln/ɛ s ɛ p 4 ɛp + p ds p 4 p ds xx ds u ɛ x, u x + 1 + ln/ɛ xx L, ;L,π 4 Hnc, lim ɛ u ɛ x, = ux,. hus lim ɛ u ɛ x, = ux,. Usin horm.3, w hav ux, = x. Hnc, ux, is h uniqu soluion o h problm 1.1 1.3. W also s ha u ɛ x, convrs o ux, uniormly in. horm.5. L x,, x, ɛ b as horm.4. I h squnc u ɛ x, convrs in L, π, hn h problm 1.1 1.3 has a uniqu soluion u. Furhrmor,w hav ha u ɛ x, convrs o ux, as ɛ nds o zro in C 1, ; L, π. Proo. Assum ha lim ɛ u ɛ x, = vx in L, π. L vx = v p sinpx whr v p = π π vx sinpxdx. Dno by w p = vp p and wx = w p sinpx. I is asy o show ha h uncion ux, dind by ux, = w p s p ds sinpx is a soluion o h problm u x, u xx = x,, ux, = wx

EJDE-8/33 REGULARIZAION AND ERROR ESIMAES 7 Sinc u ɛ x, is h soluion o 1.7 1.9, w hav whr So ha u ɛ p = π u ɛ p = π By a dirc compuaion, Hnc u ɛ p = p u ɛ p + ɛp + p, u p = p u p + p u ɛ x, sinpxdx, u ɛ x, sinpxdx, u p = π u p = π u ɛ p u p = 1 p uɛ p u p + u ɛ., u., = π u ɛ p u p ux, sinpxdx u x, sinpxdx ɛ ɛp + p.6 π u ɛ p u p ɛ + π ɛp + p u ɛ x, u x, +., 1 + ln/ɛ u ɛ., u., u ɛ x, u x, +., 1 + ln/ɛ Usin lim ɛ u ɛ x, = vx = u x,, w lim ɛ u ɛ x, ux, =. On h ohr hand, w hav u ɛ x, = u ɛ s p + ɛp + p sds sinpx I ollows ha ux, = u p + u ɛ., u., u ɛ p u p + s p sds sinpx u ɛ., u., + 1 + ln/ɛ xx., ɛ p 4 ɛp + p sds Hnc, lim ɛ u ɛ x, ux, =. Usin h horm.3, w obain ux, = x. his implis ha ux, is h uniqu soluion o 1.1 1.3. horm.6. I hr xiss m, so ha m m p convrs, hn u ɛ C1 ɛ x, x m m

8 D. D. RONG, N. H. UAN EJDE-8/33 whr C 1 = 4 m m p. Proo. L m b in, such ha m m p convrs, and l n b in,. Fix a naural inr p, and din p ɛ = ɛ n ɛp +. n I can b shown ha p ɛ p ɛ, or all ɛ > whr ɛ = np. Furhrmor, rom.5, w hav I ollows ha u ɛ x, x = ɛ p 4 p ɛp + = ɛ n p 4 p p ɛ.7 u ɛ x, x ɛ n n n n p 4 n p n.8 I w choos n = m, w obain u ɛ x, x C 1 ɛ m m. horm.7. L L, ; L, π and L, π and ɛ,. Suppos ha Problm 1.1 1.3 has a uniqu soluion ux, in C[, ]; H 1, π C 1, ; L, π which saisis u xx., <. hn u., u ɛ., C 1 + ln/ɛ or vry [, ], whr C = sup [, ] u xx., and u ɛ is h uniqu soluion o 1.7 1.9. Proo. Suppos 1.1 1.3 has an xac soluion u in h spac C[, ]; H 1 I C 1, ; L I, w h ormula ux, = p From 1.11 and.9, w obain s p sds sinpx.9 u p u ɛ p = ɛp + p s p sds = ɛp ɛp + p s p sds p p p s p sds 1 + ln/ɛ.1

EJDE-8/33 REGULARIZAION AND ERROR ESIMAES 9 I ollows ha u.,., u ɛ.,., = π u p u ɛ p π 1 + ln/ɛ uxx =., 1 + ln/ɛ p p p s p sds C 1 + ln/ɛ Hnc u., u ɛ., C 1 + ln/ɛ whr C = sup [, ] u xx.,. his compls h proo Rmark.8. No ha in [8, horm 3.3], h xac soluion u saisis h condiion ux, L, π, whil h condiion o is in his horm is u L, π. So, his also implis ha h inal valu in our horm is only in L, π, no saisyin h condiion * ivn in [9] s Inroducion. Furhr mor, w also hav h rror sima u., u ɛ., which is no ivn in [8, 9]. Hnc, his rsul is an improvmn o known rsul in [8, 9]. horm.9. L L, ; L, π and L, π and ɛ,. Suppos ha Problm 1.1 1.3 has a uniqu soluion ux, in C[, ]; H 1, π C 1, ; L, π which saisis u xxxx., <. hn or vry [, ], whr u., u ɛ., D 1 + ln/ɛ 1/ D = sup [, ] u xxxx., + xx., and u ɛ is h uniqu soluion o 1.7 1.9. Proo. In viw o.6, w hav u ɛ p u p = p u ɛ ɛp p u p ɛp + p = ɛp4 ɛp + p ɛp 4 ɛp = u ɛp + p ɛp + p ɛp = p u ɛp + p p ɛp s p sds ɛp + p

1 D. D. RONG, N. H. UAN EJDE-8/33 Hnc, w u., u ɛ., = π his compls h proo. u ɛ p u p ɛ π ɛp + = p 8 u p + p 4 p 1 + ln/ɛ u xxxxx, + xx x, In h cas o nonxac daa, on has h ollowin rsul. horm.1. L,, ɛ b as in horm.7. Assum ha h xac soluion u o 1.1 1.3 corrspondin o saisis u C[, ]; L, π L, ; H 1, π C 1, ; L, π, and u xx., <. L ɛ L, π b a masurd daa such ha hn hr xiss a uncion u ɛ saisyin ɛ ɛ. u., u ɛ., C + 1 + ln/ɛ or vry [, ] and C is dind in horm.7. Proo. L v ɛ b h soluion o problm 1.7 1.9 corrspondin o and l aain u ɛ b h soluion o problm 1.7 1.9 corrspondin o ɛ whr, ɛ ar in rih hand sid o 1.7. Usin horm.7 and Sp in horm.1, w u ɛ., u., u ɛ., v ɛ., + v ɛ., u., ɛ1 + ln/ɛ ɛ + 1 + ln/ɛ u xx., C + 1 + ln/ɛ or vry, and whr C is dind in horm.7. his compld h proo. W considr h xac soluion o his problm is 3. A numrical xampl u u xx = x, sin x, ux, 1 = x sin x. ux, = sin x 3.1 No ha ux, 1/ = sinx 1.6487171 sinx. L n b h masurd inal daa n x = sinx + 1 n sinnx.

EJDE-8/33 REGULARIZAION AND ERROR ESIMAES 11 So ha h daa rror, a h inal im, is 1 F n = n L,π = n sin nxdx = π n. h soluion o 3.4, corrspondin h inal valu n, is h rror a h oriinal im is u n x, = sinx + 1 n n 1 sinnx, On := u n., u., L,π = hn, w noic ha n n sin nx dx = n π n. lim F n = lim 1 π n L,π = lim =, n n n n 3. n π lim On = lim n n un., u., L,π = lim =. n n 3.3 From h wo qualiis abov, w s ha 3.1 is an ill-posd problm. Approximain h problm as in 1.1 1.3, h rularizd soluion is u ɛ 1 s 1 x, = ɛp p + ɛp p sds sinpx 3.4 + or 1. Hnc, w hav u ɛ x, = I ollows ha 1 1 ɛ + 1 sin x u ɛ x, 1 = 1/ ɛ + 1 1 1 s 1 n ɛ + 1 ds sin x + 1 sinnx. 3.5 n ɛn + n s 3 1 n ɛ + 1 ds sin x + 1 sinnx 3.6 n ɛn + n L a ɛ = u ɛ., 1 u., 1 b h rror bwn h rularizd soluion u ɛ and h xac soluion u in h im = 1. L n = 3 and ɛ = ɛ 1 = 1 π, ɛ = ɛ = 1 4 π, ɛ = ɛ 3 = 1 1 π, ɛ = ɛ 4 = 1 15 π. W no ha h nw mhod in his aricl iv a br approximaion han h prvious mhod in [8]. o prov his, w hav in viw o h rror abl in [8, p. 9]. Furhrmor, w coninu o approxima his problm by h mhod ivn in [8], which ivs rularizd soluion v ɛ x, = p p ɛ + 1 ɛ s p sds sinpx. + s

1 D. D. RONG, N. H. UAN EJDE-8/33 abl 1. ɛ u ɛ a ɛ ɛ 1 = 1 π 1.5944314355 sinx.687885585 +4.636337144 1 3993 sin3x ɛ = 1 4 π 1.64815976557 sinx.73741545 +4.636337144 1 3991 sin3x ɛ 3 = 1 1 π 1.64871713843 sin x 1.53314137 1 9 +4.636337144 1 3984 sin3x ɛ 4 = 1 16 π 1.648717711 sinx 5.81786885 1 3979 +4.636337144 1 3979 sin3x abl. ɛ u ɛ u u ɛ 1 π 1.643563444 sinx +.84366355 sin x.1465156 1 4 π 1.648617955 sinx +.16487171 sin 1x.6639156 1 1 π 1.6487171 sinx + 1 1 sin1 1 x.66365678 1 16 π 1.6487171 sinx + 1 16 sin1 16 x.66365678 1 8 1 3 π 1.6487171 sinx + 1 3 sin1 3 x.66365678 1 15 abl 3. ɛ v ɛ a ɛ ɛ 1 = 1 π 1.717146 sinx.664167946 +4.177348 1 3988 sin3x ɛ = 1 4 π 1.656775314 sinx.1944595 +4.177348 1 3986 sin3x ɛ 3 = 1 1 π 1.64874344 sin x.3851434344 +4.177348 1 398 sin3x ɛ 4 = 1 16 π 1.6487173 sinx.566875 1 9 +4.177348 1 3974 sin3x Hnc, w hav v ɛ x, = 1 1 ɛ + 1 sin x or 1. I ollows ha v ɛ x, 1 1 = 1/ ɛ + 1 sin x 1 s n ɛ s + s ds sin x + 1 sinnx 3.7 n ɛ + n s 1 1 n ɛ s + s ds sin x + 1 sinnx. 3.8 n ɛ + n Lookin a abls 1,,3, a comparison bwn h hr mhods, w can s h rror rsuls o in abl 1 ar smallr han h rrors in abls and 3. his shows ha our approach has a nic rularizin c and iv a br approximaion wih comparison o h prvious mhod in, or xampl [8, 9]. Acknowldmns. h auhors would lik o hank h rrs or hir valuabl criicisms ladin o h improvd vrsion o our papr.

EJDE-8/33 REGULARIZAION AND ERROR ESIMAES 13 Rrncs [1] K. A. Ams, L. E. Payn; Coninuous dpndnc on modlin or som wll-posd prurbaions o h backward ha quaion, J. Inqual. Appl., Vol. 3 1999, 51-64. [] K. A. Ams, R. J. Huhs; Srucural Sabiliy or Ill-Posd Problms in Banach Spac, Smiroup Forum, Vol. 7 5, N 1, 17-145. [3] A. S. Carraso; Loarihmic convxiy and h slow voluion consrain in ill-posd iniial valu problms, SIAM J. Mah. Anal., Vol. 3 1999, No. 3, 479-496. [4] G. W. Clark, S. F. Oppnhimr; Quasirvrsibiliy mhods or non-wll posd problms, Elc. J. Di. Eqns., 1994 1994 no. 8, 1-9. [5] Dnch,M. and Bssila,K., A modiid quasi-boundary valu mhod or ill-posd problms, J.Mah.Anal.Appl, Vol.31, 5, pp.419-46. [6] H. W. Enl, W. Rundl, ds., Invrs problms in diusion procsss, SIAM, Philadlphia, 1995. [7] V.I. Gorbachuk; Spass o ininily dirniabl vcors o a nonnaiv sladjoin opraor, Ukr. Mah. J., 35 1983, 531-535. [8] Lawrnc C. Evans; Parial Dirnial Equaion, Amrican Mahmaical Sociy, Providnc, Rhod Island, Volum 19, 1997 [9] R.E. Ewin; h approximaion o crain parabolic quaions backward in im by Sobolv quaions, SIAM J. Mah. Anal., Vol. 6 1975, No., 83-94. [1] H. Gajwski, K. Zaccharias; Zur rularisirun inr klass nichkorrkr problm bi voluionlichunn, J. Mah. Anal. Appl., Vol. 38 197, 784-789. [11] J. Hadamard; Lcur no on Cauchy s problm in linar parial dirnial quaions,yal Uni Prss, Nw Havn, 193. [1] A. Hassanov, J.L. Mullr; A numrical mhod or backward parabolic problms wih nonsladjoin llipic opraor, Applid Numrical Mahmaics, 37 1, 55-78. [13] Y. Huan, Q. Zhn; Rularizaion or ill-posd Cauchy problms associad wih nraors o analyic smiroups, J. Dirnial Equaions, Vol. 3 4, No. 1, 38-54. [14] Y. Huan, Q. Zhn; Rularizaion or a class o ill-posd Cauchy problms, Proc. Amr. Mah. Soc. 133 5, 35-31. [15] V. K. Ivanov, I. V. Ml nikova, and F. M. Filinkov; Dirnial-Opraor Equaions and Ill-Posd problms, Nauka, Moscow, 1995 Russian. [16] F. John; Coninuous dpndnc on daa or soluions o parial dirnial quaions wih a prscribd bound, Comm. Pur Appl. Mah, 13 196, 551-585. [17] R. Laès, J.-L. Lions; Méhod d Quasi-révrsibilié Applicaions, Dunod, Paris, 1967. [18] K. Millr; Sabilizd quasi-rvrsibiliy and ohr narly-bs-possibl mhods or non-wll posd problms, Symposium on Non-Wll Posd Problms and Loarihmic Convxiy, Lcur Nos in Mahmaics, 316 1973, Sprinr-Vrla, Brlin, 161-176. [19] I. V. Ml nikova, S. V. Bochkarva; C-smiroups and rularizaion o an ill-posd Cauchy problm, Dok. Akad. Nauk., 39 1993, 7-73. [] I. V. Ml nikova, A. I. Filinkov; h Cauchy problm. hr approachs, Monoraph and Survys in Pur and Applid Mahmaics, 1, London-Nw York: Chapman & Hall, 1. [1] I. V. Ml nikova, Q. Zhn and J. Zhn; Rularizaion o wakly ill-posd Cauchy problm, J. Inv. Ill-posd Problms, Vol. 1, No. 5, 385-393. [] L. E. Payn; Som nral rmarks on improprly posd problms or parial dirnial quaions, Symposium on Non-Wll Posd Problms and Loarihmic Convxiy, Lcur Nos in Mahmaics, 316 1973, Sprinr-Vrla, Brlin, 1-3. [3] L. E. Payn; Imprprly Posd Problms in Parial Dirnial Equaions, SIAM, Philadlphia, PA, 1975. [4] A. Pazy; Smiroups o linar opraors and applicaion o parial dirnial quaions, Sprinr-Vrla, 1983. [5] Quan, P. H. and ron, D. D., A nonlinarly backward ha problm: uniqunss, rularizaion and rror sima, Applicabl Analysis, Vol. 85, Nos. 6-7, Jun-July 6, pp. 641-657. [6] R.E. Showalr; h inal valu problm or voluion quaions, J. Mah. Anal. Appl, 47 1974, 563-57. [7] R. E. Showalr; Cauchy problm or hypr-parabolic parial dirnial quaions, in rnds in h hory and Pracic o Non-Linar Analysis, Elsvir 1983.

14 D. D. RONG, N. H. UAN EJDE-8/33 [8] ron, D. D. and uan,n. H., Rularizaion and rror simas or nonhomonous backward ha problms, Elcron. J. Di. Eqns., Vol. 6, No. 4, 6, pp. 1-1. [9] ron, D. D., Quan, P. H., Khanh,.V. and uan,n.h., A nonlinar cas o h 1-D backward ha problm: Rularizaion and rror sima, Zischri Analysis und ihr Anwndunn, Volum 6, Issu, 7, pp. 31-45. [3] B. Yildiz, M. Ozdmir, Sabiliy o h soluion o backwrad ha quaion on a wak conpacum, Appl. Mah. Compu. 111 1-6. [31] B.Yildiz, H. Yis, A.Svr, A sabiliy sima on h rularizd soluion o h backward ha problm, Appl. Mah. Compu. 1353 561-567. [3] Campbll Hrick, Bh M. and Huhs, Rhonda J., Coninuous dpndnc rsuls or inhomonous ill-posd problms in Banach spac, J. Mah. Anal. Appl. 331 7, no. 1, 34357. Dan Duc ron Dparmn o Mahmaics and Compur Scincs, Hochiminh Ciy Naional Univrsiy, 7 Nuyn Van Cu, Hochiminh Ciy, Vinam E-mail addrss: ddron@mahdp.hcmuns.du.vn Nuyn Huy uan Dparmn o Inormaion chnoloy and Applid Mahmaics, on Duc han Univrsiy, 98 No a o, Hochiminh Ciy, Vinam E-mail addrss: uanhuy bs@yahoo.com