HANDBOOK NUMERICAL ANALYSIS P. G. CIARLET and J. L. LIONS « Editors AUT its Finite Difference Methods (Part 1) Solution of Equations in R” (Part 1) NORTH-HOLLANDELSEVIER SCIENCE B.V. Sara Burgerhartstraat P.O, Box 211, 1000 AE Amsterdam, The Netherlands © 1900 Elsevier Science B.V. All rights reserved, This work is protected under copyright hy Elsevier Science, and the following terms and conditions apply 1 ts ase Photocopying Single phatocopics of single chapters tray be made for personal use as allowed by national copyright laws. Permission ofthe Publisher and payment of a fee is required forall other photocopying, including multiple or systematic copying, copying for advertising oF promotional purposes, resale, and all forms of document delivery, Spectal rates are available for educauional institutions that wish to make phoiocopics for non-profit educatonal classroom use Permissions may be sought directly from Ekevicr's Science & ‘Technology Rights Department in Oxford. UK: phone: (+44) 1865 843830. fax: (+44) 1865 853833. e-mail: permissions @esevier com, You nay alsocomplete your request om lin via the Elsevier Science homepage (htp:/www.elsevier com), by selecting ‘Cusiomer Support” and then ‘Obtaining Pemussions In the USA, users may clear permissions ant make payments through the Copyright Clearance Center Inc.,222 Rosewood Drive, Danvers, MA 01923, USA phone: (+1) (978) 7508400, fas (+1) (978) 7504744, and in the UK through the Copyright Liceasing Agency Rapid Clearance Service (CLARCS). 90 Tottenham Court Road, London WIP GLP. UK: phone: (+4) 207 (681 5555: fax: (#4) 207 631 5500. Other counties may have a local reprographic nghis agency for payments. Derivative Works ‘Tables of comtents may be reproduced for internal cirelation, bt permission of Elsevier Science is required for externalresale tor distribution of such material Permission of the Publisher is required forall other derivative works, inchading compilations and translations Electronic Storage or Usage ‘Permission ofthe Publisher is required wo store oruse clectroncally any material contained in this work, inclding any chapter fo part of a chapter Except asoutlined above, no part ofthis work may be reproduced. stored ina retrieval system or transmitted in any formar by any means, ekctrotic, mechanical, photocopying. recording or otherwise, without prior written permission of the Publisher ‘Addiess permissions requests to: Elsevier's Science & Technology Rights Depart noted above. Notice No responsibility is assurned by the Publisher for any injury and/or damage to persons oF propery asa matter of products labity, repligence or viersise, oF from any usc or operation of apy methods, products instructions or ideas commined in the nateral herein, Because of rapid advances in the medical sciences. in particular. independent ventication of diagnoses and drug dosages should be made, First edition 1990 Severn! impression 1992 Thin impression 2003 Liteary of Congress Cataloging in Publication Data ‘Handbook of numerical analysivgeneral editors, P.G, Claret, 1. ‘ons p. om Inclides bibliographical references Contents: . 4 Finite difference methods (pt 1): Solution of ‘xquations in R™ (pt. 1). ISBN 0-444-70306-7 (v1) | Numericalanalysis. 1. Ciarlet, Philippe G. Il Lions, Jacques Louis QAI7.IS7 1980 5194-40 80.23314 cw ISBN: 0.444 703667 © she oper used in hs publication mets he reurements of ANSUNISO 220 48-1992 (Remaner of Pap), Primed in The NeterlandsIntroduction Gil. Marchuk ‘The finite difference method isa universal and efficient numerical method for solving differential equations. Its intensive development, which began at the end of 19403 and the beginning of 1950s, was stimulated by the need to cope with a number of complex problems of science and technology. Powerful computers provided an impetus of paramount importance for the development and application of the finite difference method which in itself is sufficiently simple in utilization and can be conveniently realized using computers of different architecture. A large number of complicated multidimensional problems in electrodynamics, elasticity theory, fuid mechanics, gas dynamics, theory of particle and radiation transfer, atmosphere and ocean dynamics, and plasma physics were solved employing the finite difference techniques. ‘Numerous spectacular results have been obtained in the theory of finite difference methods during the last four decades. In ordinary differential equations, the stability of the main classical finite difference methods was investigated and the relevant accuracy estimates were constructed, a large number of new versions of these methods were constructed, and efficient algorithms were suggested for their realization in a wide field of applications-oriented problems. The nccds of electronics, kinetics, and catalysis stimulated the development of a broad class of methods for solving stiff systems of ‘equations, Problemsin control theory, biology, and medicine wereimportant for the progress in finite difference methods of solving delay ordinary differential equations In partial differential equations, the achievements of the finite difference method are even more impressive. Finite difference counterparts of the main differential ‘operators of mathematical physics were constructed, including those with conser- vation properties, that is, those obeying the discrete counterparts of the laws of conservation. An clegant theory of approximation, stability, and convergence of the finite difference method was constructed. HANDBOOK OF NUMERICAL ANALYSIS, YOL. 1 Finite Difference Methods (Par 1} Solution of Equations in 8° (Part 1) dlted by PCG, Cart and JL. Lions 1© 1990, Ebevier Science Publishers BV. (North-Holland) 34 Inoduetion The efforts of the specialists in differential equations and numerical mathematics yielded a convenient and efficient apparatus of the finite difference method, including spectral analysis, discrete maximum principle, and energy method. Considerable progress was achieved in the methods on a sequence of grids, including extrapolation methods. A flood of new results in the theory of the finite clement ‘method greatly stimulated new achievements in the theory of the finite difference method. ‘An important stage in the progress of finite difference methods was the development of the alternating direction implicit method, the fractional steps method, and the splitting method. The realization of these methods consists in solving a large number of one-dimensional problems. A considerable number of versions of this class of methods have been suggested, having high approximation accuracy and absolute stability. Numerous multidimensional problems in physics, ‘mechanics, and geophysical hydrodynamics have been solved using these methods. The theory of the finite difference method is far from having been completed, especially in the field of nonlinear partial differential equations. Life never ceases to offer new complex problems, and the method of finite differences remains a powerful approach to solving them. It would be impossible in the Handbook of Numerical Analysis to cover even ll the basic achievements of the theory. Nevertheless, I do not doubt that its publication will help making the finite difference method interesting to new people and will attract new researchers to solving its problemsFinite Difference Methods for Linear Parabolic Equations Vidar Thomée Department of Mathematics Chalmers University of Technology '$-41296 Gateborg, Sweden HANDBOOK OF NUMERICAL ANALYSIS. VOL 1 nite Diflerence Methods (Part 1) Solution of Equations in #* (Past 1) Eli F.C and JL Li {© 19%, Ehever Science Palishes BLY. (Nort Holland)Contents Prerace (CaarTen L Introduction 1, The pure inital value problem 2 The mized initial boundary value problem Cuarter Th The Pure Initial Value Problem Finite diference schemes 1, theory for finite difrence schemes with constant coefficients , Particular single-step schemes for constant coefcents problems ‘Sonie multistep difference schemes with constant coeficients John’s approach to maximam-norm estimates ‘Stability of difeence schemes for general parabolic equations ‘Convergence estimates ‘Coarten II, The Mixed Inti Boundary Value Problem 10. The energy method 11, Monotonicity and maximum principle ype arguments 12, Stabiity analysis by spectral methods 1, Various additional topics Rererences sr or Seo ‘Sunirer INpeX 109 0 us us 167 183 191 195Preface This article is devoted to the numerical solution of linear partial differential equations of parabolic type by means of finite difference methods. The emphasis in ‘our presentation will beon concepts and basic principles of the analysis, and weshall therefore often restrict our considerations to model problems in as much as the choice of the parabolic equation, the regularity of its coefficients, the underlying ‘geometry and the boundary conditions are concerned. In the beginning of the article proofs are provided for the principal results, but as the situation under investigation requires more technical machinery, the analysis will become more sketchy. The reader will then be referred to the literature quoted for fuller accounts of both results, and theoretical foundations. The article is divided into three chapters. The first of these is of an introductory nature and presents some of the basic problems and concepts of the theory for the ‘model heat equation in one space dimension. This chapter is subdivided into two sections, devoted to the pure initial value problem and the mixed initial boundary value problem, respectively. This division is then the basis for the plan of the rest of the article, where Chapters IT and III treat these two classes of problems in greater depth and generality. Whereas most of the theory in Chapter II depends on Fourier analysis, that of Chapter TI relies heavily on energy and monotonicity type arguments. The finite difference method for partial differential equations has a relatively short history. After the fundamental theoretical paper by COURANT, FRIEDRICHS and Lewy [1928] on the solution of the problems of mathematical physics by means of finite differences, the subject lay dormant tll the period of, and immediately following, the Second World War, when considerable theoretical progress was made, and large scale practical applications became possible with the aid of computers. In this ‘context a major role was played by the work of von Neumann, partly reported in O'Brian, Hyman and Kaptan [1951]. For parabolic equations a highlight of the early theory was the celebrated paper by Joun [1952], which had a great influence oon the subsequent research. The field then had its golden age during the 1950s and 1960s, and major contributions were given by Douglas, Kreiss, Lees, Samarskii, Widlund and others. At the end of this period the theory for the pure initial value problem had become reasonably well developed and complete, and this was essentially also true for mixed initial boundary value problems in one space dimension. For multidimensional problems in general domains the situation was less satisfactory, partly because the finite difference method employs the values ofthe solution at the points of a 910 Ve Thome ‘mesh, which does not necessarily fit the domain. This led the development into a difierent direction based on variational formulations of the boundary value problems, and using piecewise polynomial approximating functions on more flexible partitions of the domain. This approach, the finite element method, was better suited for complex geometries and many numerical analysts (including the author of this article) abandoned the classical finite difference method to work with finite elements. The papers on parabolic equations using finite differences after 1970 are few, particularly in the West. It should be said, however, that finite elements and finite differences have many points in common, and that it may be mote appropriate to think of the new development as a continuation of the established theory rather than a break away from it. In the Russian literature, for instance, finite element methods are often referred to as variational difference schemes, and variational thinking was, in fact, used already in the paper by Courant, Friedrichs and Lewy quoted above. The finite clement theory owes much ofits present level of development and sophistication to the foundation provided by the finite difference theory. However, in the present Handbook, the two subjects are separated, and we shall only very briefly touch upon their interrelation below. In our presentation here we shall not discuss techniques for solving the algebraic, lincar systems of equations that result from the discretization of the initial boundary value problems, but refer to other articles of this Handbook concerning such matters. Neither shall we treat the related area of alternating direction implicit ‘methods, or fractional step methods, which are designed to reduce the amount of ‘computation necded in multidimensional, particularly rectangular, domains, and to which a special article of this volume is devoted. Several textbooks exist which treat finite difference methods for parabolic problems, and we refer, in particular, to Ricirwyer and Mowrow [1967] and Samarskit and Guun [1973] for thorough accounts of the field, but also (in chronological order of publication) to Cotsarz [1955], Forsvrie and Wasow [1960], Ruanenk: and Fiuiprow [1960], Fox [1962], Gopuxov and Ryanenkit [1964], Savvev [1964], SwrrH [1965], Banuska, PrAceR and VirAsex [1966], Mrrcitett [1969], and Saansxit [1971]. In addition we would like to mention the survey papers by Douatas [1961a] and Tower [1969]. We have included in our list of references a large number of original papers, not all of which arc quoted in our text, For treatises on the theory of parabolic differential equations, covering existence, uniqueness and regularity results such as needed here, we refer to Frispmaw [1964] and Layzenseoua, Sotownixov and Unatceva [1968] would like to take this opportunity to thank Chalmers University of Technology for granting me a reduction of my teaching load while writing this article, to Ann-Britt Karlsson and Yumi Karlsson for typing the manuscript, and to Mohammad Asadzadeh for proofreading the entire work.Cuarrer T Introduction In this first introductory chapter our purpose is to use the simplest possible model problems to present some basic concepts which are important for the understanding of the formulation and analysis of finite difference methods for parabolic partial differential equations. The chapter is subdivided into two sections corresponding to the two basic problems discussed in the rest of this article, namely the pure initial value problem and the mixed initial boundary value problem. Tin the first section we thus consider the pure initial value problem for the heat ‘equation in one space dimension. We begin with the simplest example of an explicit one-step, or two-level, finite difference approximation, discuss its stability with respect to the maximum norm and relate its formal accuracy to its rate of ‘convergence to the exact solution. We also present an example of the construction of ‘a more accurate explicit scheme. We then introduce the application of Fourier techniques in the analysis of both stability, now with respect to the Ly-norm, accuracy, and convergence. We finally touch upon the possibility of using more than two time levels in our approximations. Section 2 is devoted to the mixed initial boundary value problem for the same basic parabolic equation, with Dirichlet type boundary conditions at the endpoints of a finite interval in the space variable. Here we discuss the possibility and advantage of using implicit methods, requiring the solution of a linear system of equations at each time level. Stability and error analysis is carried out for the simplest such methods, the backward Euler method and the more accurate Crank-Nicolson method. Again both maximum-norm estimates based on positivity properties of the difference scheme and /,-norm estimates derived by Fourier analysis are treated. A brief mention is made of the possibility of extending some initial boundary value problems to periodic pure initial value problems. ‘The material in this chapter is standard and we refer to the basic textbooks quoted in our preface for further details and references. 1. The pure initial value problem In this first section of our introduction to the solution of parabolic problems by means of finite difference methods we shall discuss several such methods for the pure initial value problem for the simple homogeneous heat equation in one space dimension,a V.Thomde Cure We thus wish to find the solution of the pure initial value problem ou tor xeR, 130, ax (1 ux,0)=0(x) for xeR, where R denotes the real axis and v is a given smooth bounded function. I is well known that this problem admits a unique solution, many properties of which may be deduced, for instance, from the representation al ee ntnan, ela fe “ 'o( y) dy =(E(t)v) (x). Here we think of the right-hand side as defining the solution operator E(t) of (1.1).1In particular, we may note that this solution operator is bounded, or, more precisely, up| Eide) =suplucx,9)| 0. ay For the numerical solution ofthe problem (1.1) by finite differences we introduce 2 grid of mesh points (x,0)=(jh,mk) where hand k are mesh parameters which are small and thought of tending tozero, and where j and nare integers, n> 0. We then look for an approximate solution of (1.1) at these mesh points, which will be denoted bby Us, by solving a problem in which the derivatives in (1.1) have been replaced by finite difference quotients. Define thus for functions defined on the grid the forward and backward difference quotients 2,UJ=W MU}. —U9), 8. Uj=hMUj—UF-a), and similarly, for instance, ‘uy t—up. The simplest finite difference equation corresponding to (1.1) is then for —c04 the ‘method is unstable. To sec this, we may choose v,=(—1)/e where eis a small positive number so that || =e. Then [a 40 2a)(— a= Ye 1=41)(—1¥e, or, more generally, Up=(-4y(— ye, whence [|U"||=44—1)'e+a0 as non. We thus find that in this case, even though the initial data are very small, the discrete solution tends to infinity in norm as nro, that is, for instance, at t=nk= 1, when k=1/n+0. This may be interpreted to mean that very small perturbations of the initial data (for instance by roundoff errors) may cause such big changes in the discrete solution at later times that it becomes useless ‘We now restrict the considerations to the stable case, 2<4, and we'shall show that, provided the initial data and thus the exact solution of (.1) are smooth enough, the discrete solution converges to the exact solution as the mesh parameters tend to zero. In order to demonstrate this, we need to notice that the exact solution satisfies the diflerence equation with a small error, which tends to zero with hand k. In fact, setting uj =u(jh, nk) we have by Taylor series expansion t}=O,uj—2, B,uj=O(k + h*) = OU), ifA<4, ‘where the constants involved in the order relations will depend on upper bounds for“ ¥. Thame Cuaerer 1 uae? and Bw/ex*. The expression 7] is referred to as the truncation or local discretization error. We have now the following result. Here and below we denote by (C™ the set of functions of (x,t) with bounded derivatives of orders at most m and in with respect to x and t, respectively. ‘TueoreM 1.1. Assume that ueC*? and k{h? = 2<4. Then there is a C= C(u, T) such that IU" wh] 1. Take vj—e'*®, Then U}=L acl Mom EE, and hence, by repeated application, qu which proves the theorem. IE(E,)I"-+00 as n+00, The condition (1.8) is a special case of what is referred to as von Neumann's condition, and which was first proposed in O'BRIAN, HYMAN and Kartan [1951] We shall now see that in a slightly different setting this is also sufficient for stability. The symbol is particularly suited for investigating stability in the framework of Fourier analysis. It is then most convenient to use the I,-norm to measure the mesh functions. Let thus ¥={V,}*.., be a mesh function in the space variable and sot 2 Nw hs=(s, 5 wr) ‘The set of mesh functions thus normed will be denoted /;,, below. Let us also define, for such a mesh funtion its dierete Fourier afore PEAY Ve, 7 ‘where we asume that the sum is absolutely convergent. Recall the Parseval relation, 1 f oa in io a on s J [Pine)|? a. asSrenion 1 Introduction ” We may now define stability with respect to the norm |+J},.,0F in J, ,, 0 mean, analogously to above, EV lag SCIVllay for O — for A 1 if1> 3 the von Neumann condition is not satisfied then. ‘Thus, in particular, the present method is stable in !,, for 2<<3 and unstable in both J, and maximum norm if 4>3 ‘We note that if <3, we have strict inequality for ¢ #2nn, or |E@I<1 for 0<[éi0. Such a two-step or three-level scheme would formally be accurate of second order in both x and ¢. Although, as we shall see in Section 3 below, the particular scheme (1.14) tums out to be unstable for any combination of kand k, other multistep schemes are useful in applications. 2, The mixed initial boundary value problem In physical situations our above pure initial value model problem (1) is generally not adequate, and instead itis required to solve the parabolic equation on a finite interval with boundary values given at the endpoints ofthe interval for positive time. ‘We shall therefore have reason to consider the following model problem: eu_ ou a u0,)=u(1,=0, for e>0, en Mx,0=0% for xe[0,1] For the approximate solution we may again cover the domain with a grid of mesh points, this time by dividing [0,1] into subintervals of equal length h= 1/M, where M is a positive integer, and consider the mesh points (jh, nk) withj—0,...,M andn=0, 1,.... Letting as above U denote the approximate sotution at (jh, nk), the for xe [0,1], 130,2 ¥.Thomée Curren t natural explicit finite difference scheme is now 2,U}=0,0,U5, forj=l,...,M-I1, m0, Us= UL: for n>0, 22) UP=V sem, for f= +M, given, of, for Uj Up aAUF rt UP H=29U5, j UStt=U5i'=0, This scheme is referred to as the forward Euler scheme. This time we are looking for a sequence of vectors U*=(UG...., Ui, =0 which we first norm by ao M=1, aT with Ux = max |U FI. IO "hax=max U5 When 2=k/h? <4 we conclude, as before, that ION eSTO" cc or, defining the local solution operator E,, in the obvious way, VEt eller Sle 120. so that the scheme is maximum-norm stable for 2<4. Here, in order to sce that this condition is also necessary for stability, we modify ‘our above example so as to incorporate the boundary conditions and set UP=V,=(-1) sin mh, j=0,...,M. Then, by a simple calculation, 1-21-2eos AAV, j If A> 4 we have for h sufficiently small [1-24-22 c08 nhl >y>1 and hence, if nk= |. Say, VU" ber2 YOM 420 ash, k—0. In the presence of stability we may define the local discretization or truncation error ¢} and show convergence in exactly the same way as before, and obtain the following theorem: TurorEM 2.1. Assume that we C*? and that U* is the solution of the forward Euler scheme (2.2) with <4. Then there is a constant C= Clu, T) such that [Uw x SCH for nk 1, since then for some interior mesh point of [0,1] the equation uses mesh points outside this interval. In such a case the equation has to be modified near the endpoints, which significantly complicates the analysis, ‘The stability requirement k< 44? used in the forward Euler method above is quite restrictive in practice and it would be desirable to be able to use h and k, for instance, of the same order of magnitude. For this purpose one may define an approximate solution, instead as above, implicitly by the following equation, which is referred to as the backward Euler scheme and which was proposed first in Laasoven [1949]: BUF" =2,8,U5"%, j= lyM—1, 020, Us"! =UR'=0, 030, @3) U9 =V,=0(h), M. For U" given this defines U"*" by means of the linear system of equations (42UF AUF HUTT =U5, uy =! us In matrix notation it may be written as BO" to OF where O"* 1! and O* are thought of as vectors with (M — 1) components and Bis the diagonally dominant, symmetric, tridiagonal matrix 142-2 0 0 -2 14% -2 0 0 B=} 0 =a 1420 =. o =A 1422 ‘Clearly this system may easily be solved for O**'. Introducing the finite- dimensional space! of(M-+ 1}vectors {V,}¥ with Vo = Vy =Oand the operator By, on I? defined by BuxVj=(14 22K, AV 1+ Vins) Vj— ROY), JM Ma1, the above system may also be written By" =U",By ¥. Thomée Curren = Bai UN We shall show now that this method is stable in maximum-norm without any restrictions on k and h. In fact, with 4 arbitrary, we have IU La nSIU"Loay 220. ea For, with suitable jy, wed ye lUZ | FU AD SUG) WU" Loa tT aay" esr i+ = from which (2.4) follows at once. This may also be expressed by saying that MEV Looe 1, which is the interesting case it we want to be able to take h and k of the same order, one obtains instead CF DIO Ne SALUD $A DIOL o.ay which does not show maximum-norm stability, since 24—1 > 1. For A<1 we have immediately as before a O(k? +h?)=O(h?) convergence estimate. We shall return later to the question of maximum-norm stability and convergence for 4> 1 We turn now instead to an analysis in a real I,-type space. We introduce thus for vectors V =(Vo,..., Vy)" the inner product WW yahS VW, Bo ‘and the corresponding norm Wlaa=, nie=( = v1) We denote by /2., the space I? equipped with this inner product and norm and note that this space is spanned by the (M—1) vectors @. P=Iy...M—1, with components 45> ~/28in.xpjh j=0,....M, and that these form an orthonormal basis with respect to the above inner product, or {. if p=4, n= 5re= 0 tpg We also observe that the @, are eigenfunctions of the finite difference operators ~8,8; in the sense that = 0,8,0,=(2/h?\L—cosmplde,s, j= 1-..M—L We shall now use these notions to discuss the stability of the three differenceSecniox 2 Inroduction n schemes considered above. Let V be given initial data in [2,,. Then V= "Sep where ¢,—V, ope ‘The forward Euler method associates with these initial data U]= V+ kB V= e,(1—24(1—c0s mphiog,» =... M—1, Us=UL=0, oF, more generally, Uj= LT cgElaph)*o,,» j=0,..-.M, (2.6) where EK By Parsevals relation we have thus 2A+2i.c0s 1UtLay=(E) fetenhe) "0. Thus, the Fourier analysis ‘method here shows stability in 2, for any 2. The convergence follows again by the BE)2% ¥. Thomée Cuneren standard method and gives thus in this case the following: ‘TnworeM 2.3. Let we C** be the solution of 2.1) and U that of the Crank-Nicolson scheme (2.5). Then, with C=Clu, T) we have Uw, CUP +K) for nk0, 10,)= 24, =0 for t>0, —x,0)=u(x)_ in [0,1]. In this case we may extend » first to [0,2] by requiring x(x) =n(2—x)and then extend, this function to a 4-periodic odd function on R. Looking for a solution with these properties will then lead to a solution of our original problem. Again, the finite difference methods suggested by the above may be applied for the periodic pure initial value problem and yields an approximate solution of our boundary value problem, Let us therefore, more generally, consider the periodic initial value problem, which we normalize to have period 2x, xeR, 1>0, ux, 0)=000), oee+2m=Hx) for xeR, and let us also consider a finite difference approximation defined by Lb,U* (x—ph)=¥ a,U%(x—ph) for n20, 210) U(x) =060),0 ¥. Thomée Cuanren 1 where h=2n/M for some integer M and where the sums are finite. For the analysis ofthis situation itis again convenient to use Fouricr analysis and to work in the space L¥ of 2x-periodic functions with norm toh=( J wearer)” For such functions we have the Fourier series representation nx) = Dee, where & 7 and we recall Parseval's relation Wol3,.=2e Ys. we find, by (2.10), pte oS a Deveney, ni or, with a(2) =, a,¢-! and similarly for 2, LMin0y te 7 Assuming b(€)0 for all real ¢ we thus conclude Oj! = EGHO}, where E(E)=bE)‘a(e), and hence um ative Saute 7 By Parseval’s relation we find for the operator E, defined by U**! =E,U* that LEtola, «= (x Yleuwye! °) 1 ‘The approximation by finite differences may be explicit or implicit. In the latter case the algebraic system of equations to be solved at a specific time level generally results in a finite algorithm only when the problem is periodic in space. We shall allow the differential equations and the approximating difference schemes to have variable but smooth coeificients. Our main concerns are the stability ofthe finite difference scheme with respect to various norms and the rate of convergence of the approximate, or discrete, solution to the exact, or continuous, solution of the given problem. In Section 3 we introduce the basic concepts of the theory, such as explicit and implicit schemes, one-step and multistep schemes, truncation errors, orders of accuracy, and various notions of stability and convergence. We also give a brief account of the Lax-Richtmyer theory of the relation between stability and convergence. Section 4 is devoted to the Fourier method of analysis of one-step constant coefficient methods in the space L,(R) In this case, using Parseval’s relation, the stability may be translated to the boundedness of a family of powers of trigonometric rational functions, which are matrix-valued in the systems case. The well-known von Neumann necessary condition for stability then relates to the eigenvalues of these matrix-valued functions. A somewhat stronger condition, introduced in a special case in Jon [1952], is shown to be sufficient for stability and to imply regularity properties of the difference methods similar to those of the continuous problem. The Fourier method is also applied to derive estimates for the rate of convergence for parabolic diflerence schemes. In Section $ several examples of specific standard finite difference methods are investigated, in light of the theory, with respect to stability and accuracy. In Section 66 multistep schemes with constant coefficients are brought into the framework of one-step methods and particular examples are analyzed. Scetion 7 develops some of the material from the fundamental paper by John quoted above, in which Fourier methods, somewhat more refined than those of the u2 V. Thownée Cuneren 1 carlier sections, are used to derive stability bounds in the maximum-norm for single- step explicit schemes in the case ofa scalar second-order equation (which may have variable coefficients) in one space dimension, In Section 8 we describe generali- zations of these results, principally by Widlund, to explicit and implicit, one- and multistep schemes for parabolic systems of equations in an arbitrary number of space dimensions. In the final Section 9 we discuss the rate of convergence of finite difference schemes for parabolic equations and systems. Here more attention than before is paid to the relation between the regularity of the data of the continuous problem and the convergence properties of the approximating scheme. In order to describe this the Besov spaces B;“are introduced into the analysis. Direct results are presented which show that a specific degree of regularity together with a given order of accuracy and parabolicity of the method implies a certain order of convergence. Often this convergence is of lower order for nonregular data even when the method has high accuracy, but we also indicate how a preliminary smoothing of data can recover the ‘convergence rate caused by low regularity. We finally present examples of inverse results which make it possible to draw conclusions about the data from the observed rate of convergence. 3. Finite difference schemes In this section we shall, in a mote systematic and general way than in Chapter T, introduce the basic concepts relating to finite difference schemes for the numerical solution of the pure initial value problem mesos s = LPs sey) im xR, Gy ice x, 0)= (x) on RY where #=(2,,....0,) is a multi-index, |al=,-+----bay and Drom @/ox,)* --O/Ox,). Although our purpose is the solution of parabolic problems we shall not specify the type of the linear differential operator P(x, ,D) at present, but simply assume, to ‘begin with, that the inital value problem (3.1) has a sufficiently regular solution for ‘our purposes. We shall return in Sections 4 and § below to more precise descriptions of parabolic equations. We begin our discussion here by considering the case of the homogeneous equation so that f=0, and we first consider an explicit finite difference scheme *() =D aglx, nk, UB) 7 SALUD), n=0,1,....Senin 3 Pare intial salue problem 3 where U" stands or the approximate solution at time nk, where kis the time step and where h is the mesh size in space, f=(f,,....f..) is a multi-index with integer ‘components, and the summation is finite. The space varible x may range either over Ri or over the mesh points fh. We shall also allow the scheme to be implicit, ic. defined by the equation BiyU" SALAUY, n=0,1,...5 where Bt, is an operator of the same form as Af, but where we also assume that Bi, has a bounded inverse (in the space of functions under consideration) so that the latter equation may be solved for U** and written UE UY The solution of this discrete problem may then, in both cases, be written U8 Bem Enna Fano of, if we set U° =e, and introduce the product, ERE Bana Euan Ean G2 as UN= EE. Inmany cases we shall assume that k and h are tied together by a relation such as kjht=2=constant, most often with q=M. the order of the operator P(x, t,D). We shall then omit hin the notation and write, for stance, F,,, instead of E,4,.. When the coefficients are independent oft, we may write simply F, and the solution in such a case is simply U"= Ef For the nonhomogeneous equation we will usc difference approximations of the form BEgU"=ARAU+KME af, n=O.1,-.., where ME. isan operator which could be of the same type as 4f ,and Bf ,,or, integral average such as ty of Fd di. Min fd=! We find then, with 17 UN Egg UKM S, and the solution of the discrete problem takes the form Ura eRg Usk Bn" MLL 64) In thesimple case that the coefficients are independent of , that kand hare related asu V.Thomie Cuneren tt above, that Mif,, =f" and U=» this reduces to Uta Bek SBE iS ‘The difference equation is said to be consistent with the differential equation if, formally, for smooth solutions ofthe differential equation, the difference equation is satisfied with an error which is o(k) as k,i+0. More precisely, introducing the truncation error, Tat AO BE gut — AE gut) — Mi af, the scheme is said to be accurate of order win x and vin ¢ if, uniformly for O< nk 0 given, we may choose i such that le- dh 2. This formula may only be used for n>m—I and thus requires, in addition to the natural initial condition U® =e, that U',..., U"“! be prescribed. ‘A common way to analyze such a scheme is to reduce it to a two-level or one-step scheme of the form discussed above. For this purpose one introduces the compound vector-valued unknown function OF(USUE UF, and the matrices of operators Aba, Aba Ale ar Aaa Alay a Ot ante r 0 0 0 1 a Au-| 0 1 0 | ana : 0 1 0 0 10 With this notation the scheme can be written as En O*+kF" for n>m—1, (Mi f.0.....0 ‘Again we may introduce the discrete local solution operator of the homogeneous equation, Bana Bia) Ahn and define its stability and consistency in the obvious way. In particular, if is ‘a normed linear space in which we consider our functions, then O*is sought in the product space "x --- x 4 with m factors, and the stability is measured with respect to the corresponding norm. In the discussion of the accuracy of the scheme special attention has to be given the choice of the starting values U',...,U"~* Similarly as before, with a =(u',u*~',...,u°-""!) forthe exact solution we now have, for n>m—1, Oma Eee On ak YE tay 69) Jem where the truncation error only stems from the first components of the compound solutions. It follows that, in the presence of stability, if the truncation error is Of" +k", say, then the global error is of the same order provided O"~'—a"-*“0 V. Thome (Cowrren I matches this. Note that, since the initial error is a local error, a lower order approximation may be used in the initial steps, because an extra factor kis available in this term, which is not needed to compensate for the summation in (3.9), see the example below. As an illustration, consider for the solution of the model homogencous ‘one-dimensional heat equation the three-level equation UTC) UT Hx)_ UNG h)—2U%)4+ Ue C0) _ Piet EW) aig dak, U8) = AU" + A) —4AU(3) + AU" A) + U9, In system form this may be expressed in terms of O*=(U%, U"-4)F=(U", VF as U8 x)= 22 UM a) — ARUN) + 22UN( H+), (x)=). or OO) =. 090) 2 Od/z, —% Vane P22 OV png [2 SJousna[-* SJore[ Soren for n>, By the symmetry around (x, nk) the exact solution satisfies (3.10) with an error of O(a? +42), which translates into we[e | Ex i 1 -soue +4) as h,k0. If we assume 2 constant the order term reduces to O(4?) as k tends to 0. With U? =U! =p the initial accuracy is also O{k)= Oth?) for this case, but if k and h are independent, a more natural choice is UG) = 100) +40,5,000) =x, k)+O(k +kh?) as hk+0. Gly ‘The definition of stability is now interpreted to mean the boundedness of {* in terms of O'=(U!, U9)", and this would imply convergence of order O(h?) or OU? +k), respectively, in the two cases, for sufficiently smooth initial data. However, it tums out that the present scheme is unstable for any choice of 2. For ‘example, ifthe scheme is considered at the mesh points jh and we set U = U(jA) and. AW’ =[pn» WE May choose the initial values OB} =(— t= (— ay a,SecTion 3 Par intial wale problem 4 where « is a fixed 2-vector, and then find jeer 8 TP Oy=(-1) [ 1 Hi 4 For any 1 the 2x2 matrix entering here has two distinct real eigenvalues, one of which is less than — 1. If we choose for a the corresponding eigenvector, itis clear that O* is unbounded as n grows and thus the scheme is unstable. ‘As we shall seein a later chapter this scheme may be stabilized, for any constant 2, by replacing U" by the average 4(U"*! +U"~") so that the scheme becomes UT G)—U" Mx) _ UGH) —U" 0) U1) UA) 2k e . which was proposed in Du Forr and Franket [1953]. Asanother example we could consider the implicit three-level method derived by selecting a second-order accurate backward difference approximation to @ufat at (n+ Ik. This requirement defines the scheme uniquely as QU" (9) —2U%E) HAUT ok 0,5, U8, .13) and, as we shall see in Section 6 below, itis unconditionally stable. The truncation error here is O(h? +42) and, in order to match this order we may set U=e and choose U" by (3.11). Note that thus the first-order accurate forward Euler method is accurate enough to match the second-order method (3.13). Note also that no stability restriction needs to be imposed for ths first step because the formula is only used once. ‘We shall end this section by making an observation concerning the accuracy of the Du Fort-Frankel scheme (3.12) Let u be a sufficiently smooth function. With 8,8, as before and correspondingly for 0,8, and with &,, the symmetric difference ‘quotient, defined by Baul t= (ul, £8) — ul, WYK) = HG, +3,)ulx, 0), we have for the truncation error of (3.12) rane) B.1y) _ Wet Ayu x) tH) 40H —; = 8,ux, mk) —2,5 ux, mk) + (8/05, x, nk) u_ ew eu -[B-25]aunst Bn + O(F?)+ O60") + OLK/H" ‘Consistency with the heat equation therefore requires that k/htends to-2er0, wla ¥. Thome Cuarren I the case eg. if k/h? = i=constant. However, ifinstead k/h=1=constant, we obtain Su, Fu ‘a t6-[% Shae Bt] osmnye ove) as h0, a ‘which shows that the scheme is then consistent, not with the heat equation, but with the second-order hyperbolic equation pou, bu Oa ae ton ax? ‘We shall return in Section 6 to discuss the stability of the Du Fort~Frankel scheme. 4. L, theory for finite difference schemes with constant coefficients In this section we shall use the Fourier transform systematically to express and analyze the notions of consistency, stability and convergence for constant coefficient single-step finite difference methods applied to the pure inital value problem for ‘a homogeneous parabolic equation, or system of equations, in d space dimensions. ‘As we have seen earlier such material is also relevant to the study of initial boundary value problems when the boundary conditions may be interpreted as periodicity conditions. First, we define the Fourier transform oyer Rt! by Lxey i a= fervent, xgeR, x where, as always below, we asume that o is small enough for large |x| that the definitions and subsequent calculations are justified. We recall Fourier’s inversion formula, Hx) = On) 48(— x)= ane [acensas, and Parseval's relation, lo] =2ny-*7 hol, where, as for the rest of this section, -I| denotes the norm in L(t), We introduce as above, for a=(2,,..., %,)amultiindex and |x|=2, +--+ aq the mixed derivative of order |al, De= Gar, ) lx and recall that Dy O=GHAD=IMSAL, where § = Gi-- Ei 41) Consider now the initial value problemsecon 4 Pre inl eae polem 8 Sarpw= 5 Pde forxeRt, 130, u(x, 0}=x) in RY, where vis sufficiently smooth and small for large |x|, in a way that we shall not make precise at present. We shall focus our attention on equations which are parabolic, and begin by a definition of this concept which generalizes the heat equation (4.2) al which we have considered above, and the more general case when P(D) is a second-order elliptic operator, Ou PD = Eo mseae an +3 og +Pol, (43) where (pa) isa symmetric positive definite constant d x d matrix with real elements. For this purpose we introduce the characteristic polynomial of P(D), and say (admitting the possibility of complex-valued coefficients) that (4.2) is parabolic (of order M) if Re Piig)<—clgIM+C, c>0, FER‘, where [l= ba)”. ‘We recall that this isthe same as saying that — P(D) is strongly elliptic. Sometimes, we shall allow (4.2) to be an NV system. In this case the coefficients P, and the polynomial P() are N x N matrices and the condition (4.4) is replaced by APE) SelM +E, €>0, EER, 45) (44) where ALA)=max Re if (1,}f.4 are the eigenvalues of 4, Clearly (44) is satisfied with M2 for the second-order operator in (4.3) since then RePit}=— ¥ pagjb— ¥ (imppé,+Re po, ake ra <-déP HCl 4D<—4clgP +E, EERE Equations for which (4.4) or (4.5) are satisfied are referred to below as parabolic in the sense of PetRovskit [1937]“4 . Thome Cuarren tL Letting a(é,) denote the Fourier transform with respect to x of the solution at time t, we obtain by virtue of (4.1) the ordinary differential equation d gh PODME.) for e>0, HE, 0) =A), so that, formally, HE, )=explePLig ae), from which 1x1) may be found by means of the inverse Fourier transform. Using this together with Parseval's relation we find at once for the solution Ee of 4.2), LEQe < suplexp(-POSDI Ie, (46) and hence, by (44, HE Wel < Campers" pol 0, and by (4.20), IE(g0. (422) Writing EO —expinkPa) YEE (EE) —exptkPs exp kPUE), (4.23) im50 ¥. Thame ‘Cuaerer I we hence have LEAS)" —exp(nkP(E))| < CakhX(1 + |E1Me- < ChAL +E), AIG 0 we have Ae g|M He CAML ED GR OOM, Hence, EE) —exp(nkPOE)| 4 thi satisfied, and for @ <} the stability condition reads a ‘ 1. Returning to the one-dimensional equation (5.1) we shall show now that for any given positive integer v there exists a unique explicit method for (5.1) of accuracy ye= Dv using the (20+ 1) points x-+jh, j= —¥,...,. This method will be shown to be parabolic for J sufficiently small ‘We shall prove this by constructing an even trigonometric polynomial ce BE)=EG = F alaeosjé56 V. Thome Cunrren IL such that Bg, Aye" 4O(E"**) as 6-410, (83) and such that, for small 4, JEG AL<1 for 0<|él0, ‘and, by taking the (20th power and replacing = by 4é, =F basesintsy = Efusil eos with f,2)>0. m Hence if I0 for 0<|é| 4, -=2/(1 — 20)if0 <4 and we recognize our above con for aceuraey and stability for this method. This procedure for constructing a finite difference operator may be thought of, in the following way, as a discretization in the space variable, followed by a separate 0. 4 dent *AnEI0E.) for >0, 612) or, by integration between 1,=nk and t,,, =(n+ Dk, OE tay = empAE OLE, ta) 6.13) This may now be approximated, with r as above, by OF Ney H(AAUAE)ONE) = EOEJONE)- ‘The property (5.10)is referred to as ellipticity ofthe finite difference operator Ay. The use of special rational functions r=) to replace e* in (5.13) may be interpreted as the use of a Runge-Kutta type method for the integration of the ordinary differential equation (5.12) (of eg. Hewaicr [1962].Seeron 5 ‘Pure initial value problem 9 The discretization in space only, leading to the above system of ordinary differential equations with respect to time, with the values at the mesh points as the ‘unknowns, is sometimes referred to in the literature as the method of lines (cf. ¢-, Frankun [1969], Lets [1961], Varaa [1962). Discretization in time only, of the method of Rove [1931], can sometimes also be used as a preliminary step. ‘These considerations may be generalized to a scalar parabolic equation of the form ow a Po L PD, (5.14) boi where, for simplicity, the elliptic operator P{D) only includes derivatives of the highest order M and has real coefficients. We may then first approximate P(D) by a finite difference operator QAVix)I=W-M Sap VCx— Bh 6.15) 7 for which we introduce the symbol O(¢) satisfying QUE) = Lage = Pls) + OEM“) as 60, 7 where 1 is the order of accuracy. The finite diflerence operator (5.15) is said to be lliptic (cf. Tuomer [1964] if Q)<0 ford 2, Bi) =e + OU gM) mel OMA, The discussion of the stability is analogous to above. In particular, if we can construct an elliptic finite difference operator consistent with P(D), then the @-method with 0=0, 1 or + gives an explicit forward Euler, implicit backward Euler ot a Crank-Nicolson method for (5.14), which is L,-stable or parabolic under the appropriate assumption, Recalling that —P(D) is elliptic it PGE) =—(-1"?_-D Per>aet un for #0, c>0, it is natural to try to construct an elliptic finite difference operator consistent with P(D) simply by replacing the derivatives in P(D) by symmetric finite difference quotients, which corresponds to taking OG) =Plisin )=PIH+ OE?) as E+0, where sing =(sin€,,...,sin,). The polynomial (2) thus defined is nonpositive and may be used to construct an L,-stable finite difference operator by setting@ V. Thome (Cuaeren HL F{8)="(4Q{E) with r(2) as above. However, as Q{E)=0 if, for instance, am this operators not ellipticin the above sense, and the corresponding finite diference scheme for (5.14) thus not parabolic. In order to make Q, clliptic one may modify the definition by setting Qe)=Plising)— ¥ (cose, which does not change the consistency since the additional term is of order (||?) as £90. The operator thus defined uses other than the closest possible neighbors at the ‘mesh point x. For instance, for the second-order operator PD)= the term p,,0*4/0x} is replaced by ux+ 2he,) —2u(x) + x —2he,) Pas ai Another possible elliptic finite difference operator which does not have this disadvantage Or F p?adas ae with symbol QO= E pale —1)-—e-%), 4 which is elliptic as -QO= YL pyll—cosé, —ising, \ cos é, +isin &) on = Bratt —e08é,)(1 cos &)+ Zensin gsin >eLa ~eos EP sing) -20 $0 cos é;). ‘Another common choice is 2,¥e= Zrn., 5,,Vx)+ hr 8,34. V0), where 6,,=4,,+6,,), or ~HH=2Y py cos 8)+ F pysingysing,Section 5 Pare inital value problem 6 whi to iscasily seen to imply that Q, iselliptic. For d =2 the latter choice corresponds QVIx)= Y py(Vle+ he,)—2V(x)+ V(x— hey) /h? a + Pia(Voxt he, +he,)~ Vix he, —he,) V(x — he, + hez)+ Vix—he, —he,)y/h?, and thus uses the nine points x, x++he,,x:+he, ++he,. Let us remark about the simple model problem (5.1) that, by our earlier discussion, we know that there exists a trigonometric polynomial of order such that (6. (5.5) Ag) 2 =F Bradt eos sy? =E +O") as £0, and such that by (5.6) the corresponding difference operator, Oya F fash OEY. is elliptic. ‘Theabove makes it natural to ask for rational functions (2) which approximate e* near z=0 and which satisfy the appropriate boundedness conditions. The most commonly used functions of this type are the Padé approximants (cL. ¢2. VARGA [1961, 1962)) which are defined by n,.A2) == 7O) =e OG) a5 2-0, where d,,, and 1,, are polynomials of degree p and g, respectively. One may show that these polynomials are uniquely determined by & +4-i) Jo (PQ! 1,f2)= and (r+q—fipt Sudel= 2 ioseanilo ih For pq we recognize the rational functions re,Py,9 and r, corresponding to the forward and backward Euler method and Crank-Nicolson's method. tis known that rp dD<1 for 2<0, ifpaa )a V.Thomée ‘Cuarren I For pm 62) BUN SAUTE FAW! where m>2, together with the initial conditions Ulao! for j=0,....m=1, where the v’ are determined in some fashion from the initial data v. We shall assumeSecriow 6 Pare Inia value problem 6 that the 4,, and B, are finite difference operators with constant coefficients, Auth = Day —Bh) for 7 where the dj and hy are constant and the sums finite Introducing as in Section 3 the vectors O*=(U", U"-',...,U""" YF and the corresponding matrices 4, and B, equation (6.2) takes the form BO '=4,0" fornzm—t, or O50" forn>m—1, (63) where Bey Belg Be Ay d 0 0 =| 0 bos 0 0 1 0 We say that (62) is stable in L if (6.3) is stable in the product space L$ so that the powers of the operator £, are bounded on L3. Introducing the symbols or characteristic polynomials of the operators By and Ay, He) = Loos 7 and af8)= Lage’, 7 bessm GeRY, the symbol, or amplification matrix, of (6.3) is al’) anf) HE) He) PO 0 Bo=| 0 1 0 0 10 (We assume that B, is invertible in L, so that b(2)#0.) The Ly-stability of our ‘multistep scheme therefore reduces to the uniform boundedness in £of the powers of% ¥. Thomée Cuarren 11 the matrix E{g), The von Neumann condition is AEQ)0, AEN . In the former ease, we have immediately mt ya i” Ma (68) lnk and since the product of the roots of (6.7) equals 2d +20).<1, we have, with the proper ordering, ll0, ie. the Du Fort-Frankel scheme is parabolic. ‘We shall now show that the method is actually stable in Ly. We recall that for any (2x 2)matrix = (a,) there isa unitary matrix U which transforms A toa triangularo V.Phombe Curren I [° sh where j1,.2 are the eigenvalues of A and matrix, oF usau: m aM a a Hy +4 a Maa + daatla Mao» so that, in particular, Imix ¥ lash ron Applied to our matrix B(g) this shows x ieai|*!<4 ‘The Ly-stability of the method is now equivalent to the uniform boundedness of the powers of the triangular matrix [4 ee aoe where Imi<7 ens | rmbt) mS. 1 # and, since |y| <1 for j= 1,2, to the boundedness of my, it) In the case (69) we have ft») Sm z vemAl—y for eR, and, in the case (6.10), [mls H3)|0, By >6, it is clear by (6.11) that (4o—KBoQ,)”* exists on Ly, and we may solve (6.13) for U*** to obtain Unt — Yo —KBoQs)” May KAQ,U" For the investigation of the stability we consider the characteristic equation YD (Ap oe yunh= fo which is thus the characteristic equation of the one-step system formulation of 6.13) as well as ofthe scalar problem. For the Fourier transform of the solution we have Orb — Fo BoA QE May BADENO where A= kh™. ‘We shall consider some specific choices of linear multistep methods which are used for ordinary differential equations. Let us being with the Adams methods, They consist in writing (6.12) in the form Ug I= UE)+ j 2,008) ds, and then replacing U(s) in the integrand by a Lagrange interpolation polynomial. Using thus the polynomial determined by the values at ty ...st, 1m ON€ Obtains the m-step Adams-Bashforth method utcusk Se bg QU™* (6.14) where the coefficients b,,, may be found in e.g. Henrict [1962]. Ifinstead we use the7” Thome Coarse 1 polynomial interpolating also at ¢,,, we obtain the m-step Adams-Moulton method (1 Rb%0 QU" =U S68 Q,U I, where the coefficients bj, are also casy to determine. By their construction these methods are of order O(k) and O(k""), respectively, in time, to which the error ofthe discretization in space, O(F!), say, has to be added. In view of the relation kh™™ = A=constant the total order is thus ("4 and COgeminens 184 respectively. For m=2 we have the Adams-Bashforth method Ut =(143kQ gu" 1kQ,U" and the Adams-Moulton method (1 7kO,)U" = (1+ kU" with the characteristic equations (1+ HOE)as HOO (1 AQ? 14 HOE) + HAVE respectively. By direct computation it follows that von Neumann’s condition is satisfied for the ‘Adams-Bashforth method if and only if =AQG)<1 for FER! (6.17) For the standard approximation to the one-dimensional heat equation we have Q(e)=2{cosé— 1) and the condition (6.17)holdsifand only if <4. This requirement is thus more severe than for our previous methods and the method is therefore not competitive. Turning now to the Adams-Moulion formula (6.15) we easily see that this method, although implicit, cannot be unconditionally stable. In fact, this would ‘demand that both roots of (6.16) are in the unit disk forall values of p=AQ(E)<0. However as p-»—0 the equation becomes fue +3u- = 0, which has the root x= —K(4+/21) outside the unit disk. The method is thus unstable for large 4 and hence also inferior to our previous methods. We now turn to the method of backward differencing to construct a finite difference operator from (6.12). It consists in replacing the time derivativeiin (6.12) by the derivative of the interpolating polynomial, based on the values at t,.4, tapes slam eValuated at f,,,, t0 obtain an implicit scheme. This yields a method of the form ete £ ageneoue, HQ,U", (6.15) (6.16)Secnon 6 ‘Pure inital value probler ” or (qk) = FE yur! (6.18) i It would also be possible to construct in this way an explicit scheme by evaluati ty ie agur t= +hQ)U"+ g Baus (6.19) at or, at some earlier point t, _p J=2.--.4™ Since the order of accuracy in the replacement of the time derivative is O(k") we find that the total error, a5 in (6.14), is O(M#+K", or, with £=kA-™ constant, ‘Othman, For m=1 these schemes reduce to the backward and forward Euler methods discussed above, but since we are interested now in multistep methods we consider only m>2. We briefly discuss the stability of these schemes and begin with (6.18). The characteristic equation is, (= ADEE) ye — = Busi” 4=0, (6.20) 2 and m=3 we have, in particular, for (6.18), (@—kQ,JU"*!=2U"—4U"!, (621) (A-ROgUN t= 3u"— aU U2 For In order to decide whether the roots ofthe characteristic polynomials correspond- ing to these methods are in the unit disk we transform it to the let half-plane by setting =(1-+n)/(1—n) and apply the Hurwitz criterion: In order that the equation Efeo yymt'=0 with 7o>0 has all roots in the left half-plane if suffices that D,=71>0, We now apply this criterion to show that the method (6.21) always satisfies von Neumann's condition. In fact, setting p=AQ(@) we obtain for the transformed equation 4-py? +2—2p—p=0. (622)n V. Thome Conrres HL Here for p-<0 we have =2-2p>0, 2p 0 -p 4-pl and hence for 0<|£| 0, Both methods thus satisfy von Neumann's condition for any A and are parabolic. For m=2 and 3 the explicit schemes (6.19) are que Q,U+4UT! for n> 1, and ye For the first the characteristic equation is W—2pu—1=0 with the roots «= p+./p? +, one of which is always outside the unit disk if p <0. Hence this method is unstable for any choice of 1 it contains (6.5) asa special ease. For the second method the characteristic equation is +3 —pye—3e+4=0, and with p=0 the roots are = 1 and 4.=4(—5+/33) 0 that this method is also unconditionally unstable. ‘We shall now describe two higher-order three-level methods from Douctas and $+KOQUEUN!—$UN? for n>?Secnion 6 Pare initial ealue problem n Gunn [1963] for the model heat equation in d space dimensions (with d<4), eu Lge *ER, 120, (623) aw paw with initial data u(x, 0)= n(x). ‘These methods will have the special feature that they only use the immediate neighbors at a given mesh point, which is, of course, particularly desirable when applying them to a mixed initial boundary value problem. However, as we shall see, the stability results we shall present do not have exactly the same character as the ones discussed above. Let us first note that for v smooth, with axthe 2, 3, eats ted, and u(-)=u(-,nk), we have aw BO OL, Bape 7g ae tO EM FOU +H) as hk. =40, 0,0 tutu) Setting, as in Section 5, hye 4 of a eee) dD Shout He) (6.24) ‘Thus, if we can approximate the sum appearing as a coefficient of h? on the right to second order in h, then we will have produced a Ovi* +k?) finite difference approximation for (6.23). Note that, for the exact solution of (6.23), tye ttt ou aun 3 gat? S aregr tute" V. Thode Curren and hence, using the fact that A7u=t,, Dui ue? eR -2E 0,5,0,.5,u¢ +O +k) as hk+0, (625) & and also, since A*u=Au, 4 atu, (wwe Lae es “) 2 E000 Bau t OWA +HE/K) as hy k0. 626) ‘The relation (6.24) together with (6.25) and (6.26) thus suggest the following two difference analogues of (6.23), namely os 3 faut Une Ur) wR ay 08.0,8,0"), an and, with =k", utiiuet . — ((1 = 14(82)U" +O HUF IBAYUT V4 BHY 248 2,08,U% (6.28) i ‘The crucial matter is then the stability of these schemes. In this regard we have the following two results from Dovatas and Gunn [1963] (where they were expressed for a cubic domain and in discrete fy norm). ‘THEOREM 6.1. The difference scheme (6.21) isconditionally stable in L(R¢) for d<4 in the sense that if U°=0 and = k/h? is bounded away from and 2 (or bounded away from 0 if d<3) then, with I+ 1=Il-In.yxe> JO1 (629) TuroneM 6.2. Thedifference scheme (6.28)is unconditionally stable in L(R*) for d<3 in the sense that (6.29) holds if U°=0, ‘Wenote that these results do not show stability in the sense of boundedness forall U? and U!, but are restricted to data with U°=0. After application of the Fourier transform, equations (6.27) and (6.28) may both, be written in the form H{hE\O" "=a, he)O" +a,(he)O"-*,Section 6 Pare intial value problem 1s and the corresponding characteristic equation is Hew? a, (8 a308)=0, with roots = (2,11, 2. The solution then satisfies (assuming distinet roots; the case of coinciding roots has to be treated separately) ey (hey +e,(Eyun(hey, 22, e(@+ex(6)=0@)=0, e1(s(he) +e )ua(he)= OME), which implies he a(hiP or@atit = Hah) ‘The stability stated then requires the boundedness of the ratio on the right as he R*. One may show that for the method (6.27) this ratio is bounded by 641+ 1°! +6) if'd<4 and for (6.28) it is bounded by 4 if d<3. Although this stat ‘concept is more restrictive than carlier, it suffices nevertheless in order to derive convergence estimates if we choose U® = u°, as then the ir error 2° =U° —u°=0. We obtain for this choice, in both cases, if U' is chosen so that [Ut ut) 5>0, under the ini ux, O)=n(x) for xem, (7) and an explicit finite difference approximation of the form a(x, t, NU" — Ih) +A (8) for n20, (73) U%)=na), ER,Stenox 7 ‘Pare initial value problon n where (x, #)=(jh, nk) varies over the mesh points in Rx [0, T]. The summation is over a finite set of points, |l| vat ec, BN IO which proves the result in the case considered. The case of variable coefficients may be considered as a perturbation of the constant coefficient case. We sketch the proof of the stability in the case that the coefficients depend on x but not on r or h, so that Eye) = abet Ih). Letting U*= Efowe shall want to estimate U"(x). for x arbitrary, by the maximum ‘norm of v. We then fix the coefficients of E, at xo to obtain the representation Eyota) =¥ dyo6— Ih) + b(ooGe—M, where d)=4(x9) and by(x) = a(x) — a(x). Thus2 ¥. Thome Cuneree 1 Ur EQ=T GU eH) +E bIU— Ih) EU") + SQ), where the latter equality defines E, and f*, and hence UM) = Etta) + TELS 1 Mx0) (7:10) Setting EDO) =L Gypb(— ph, we may write ERS d= Pde ' (Xo ~ Ph) ay Sur bio — ph) U™'™(xo— (P+ Dh) =Z iniU oil, aan where Yuslod =D Bm los I. Setting now Byfso OE bea i+ lie", oom + [B00 TTA. DE AE. Iml%o) Using the consistency relations one may show, by arguments similar to the ones used to estimate the a, above, for A fixed, nk <7, that with C independent of x, = Irmo) 0. (7.412) This is relevant for the analysis of the convergence of difference quotients of the approximate solution to derivatives of the exact solution of (7.1) and was used in Jou [1952] in the study of smoothness properties of the exact solution. The solution U" of the difference scheme defined by (7.3) may also be written in the form UAE) =AY. og op68 Mole I+ ST dame PMC) ME mollis AS, Cbames sl Ps and the coefficients g,,»,;:(8) may be thought of as determining a discrete fundamental solution. For the particular case that the coefficients of (7.3) are independent of x, t and, we are in the situation discussed in detail in the proof of ‘Theorem 7.2 above, with Gam il pms ‘The maximum-norm stability may be expressed in terms of the fundamental solution as sup hk LldgmpAHN0 and p, is an even natural number, such that Ele, + 8) = EE 4) explizge —ByEM(1 + O(N) as E-+0. 7.4) ‘The proof of this result may be cartied out by the above technique of John, It follows, of course, in particular, that von Neumann's condition is necessary stability. For an operator E, which is consistent with the heat equation con (7.14) is satisfied for €,=0, with a,=0, fy=2 and 41,=2. If there are no further points &, in [é|0, a constant C=C(A) such that, in the maximum norm Ete 0 (for earlier related work, of, also Juncosa and Youne [1957] and Wasow [1958)). Somewhat more general results were obtained in Haxsens [1970] and NoRoMARK [1974], We shall now briefly describe Hakberg’s result. Consider the initial value problem for the parabolic equation ow 1% an ae (7.15) M=2m, xeR, 1>0, and consider a consistent finite difference operator of the form Eyv=R(kA,)o, (7.16) where R is a real rational function and Agtd=h™ YS dioe-jh, d_j=deR. ues% ¥.Thomée Cuarren th ‘The operator thus defined is consistent with (7.15) if Ri) and Al= — EM 4+ (EM) as E> where A() is the symbol of Ay, ‘The symbol of F, is then E(Q=RUAQ), where A= kh, ‘and we say that £ is uniformly maximum-norm stable if, with C independent of 4, ytoly) as yO, Cay (7.18) [Efe 1, Ol, 00. It is easy to see that the second quantity is independent of h so that the condition W, exactly ifsremow 7 Pare initial sale problem » reduces to MJECYISC for 220. (12) 11 follows from (7.20) and (7.21) that stability in maximum norm implies stability in L, for other p, We remark that it follows from our above discussion that the condition for stability in C(R) or 1. is the same as in W,, In order to use these notions to prove stability itis thus needed to have access to some method to bound an expression such as the left-hand side in (7.22). One may then first show that in order to estimate a 2n-petiodic multiplier a it suffices to estimate a where 1 is a function in Cj(R) which is identically 1 for |é)|4d, Mf 0, which is the desired result. ‘The technique just described may be applied to give a simple proof of the sufficiency of the conditions of Theorem 7.3 and are also useful to show convergence estimates, of Section 9 below. in —n,n], Because E, is parabolic we have0 ¥. Thome Cuneren TL 8. Stability of difference schemes for general parabolic equations: In this section we shall consider the stability of finite difference schemes for general parabolic equations and systems, We first discuss equations or systems which are parabolic in the sense of Petrovskii and difference schemes which are parabolic in ‘a sense generalizing the one introduced by John and described in Section 7 above. We also show analogues of some known smoothing properties of parabolic problems and touch upon the possibility of using such propertics as definitions of parabolicity of difference schemes, ‘We shall thus first consider equations of the form au aos YL Paani, 1) bie xeR', 1>0, where, with a=(y,...5), Diu (Q/x,)"--- O/exyru, 12! ‘The equation is considered, as usual, under the initial condition x,0)=0(x), xe. Such an equation i said to be parabolic if x,t, D)is elliptic, or more precisely, ifthe real part of its characteristic polynomial, Porid= LD Pyoonigr, blew is negative-definite, and uniformly parabolic in a domain in RY xR. if, for (x,t) in this domain, at tae Re Plx,tig)< —elgIM for GER! with c>0, 62) It is clear that in this case M has to be an even positive number. We shail allow below also that (8.1) isa system of NV equations in N unknowns which is included by letting u be an N-vector u=(u,..., ty)" and the P, (Nx N) matrices, which we always assume to be sufficiently smooth for our purposes. In the system case we replace condition (8.2) by a corresponding condition for the eigenvalues. Setting, as in Section 4, for any (N x N) matrix with eigenvalues {2,}¥, AlA)=max Rey, we thus say that (8.1) is parabolic in Petrovskii’s sense if AP (x, —e1EM, Exe RS, 120, where now E=(é)....é0%, @=€i Gt Tis known that associated with a system of the above kind there isa fundamentalSecnion & Pare intial value problem 3 solution F\x,t,y.s) defined for x, yelR4, 0 0, (8.3) The fundamental solution also has the property [DEDEP (x, t,x+2,5)| SC) HM exp elses), with the power of ¢—s on the right independent ofa. (In the constant coefficient case T depends only on —x and ¢—s and thus I(x, t, x+z,5) is independent of x.) It follows, in particular, from the above representation and (8.3) that the initial value problem in (8.1) is well posed in L,=L,(R*) and Wy-=W™(R), the Sobolev space defined by the norm yeh, ee Further, the solution has the regularity property corresponding to our previous result in L, for the constant coefficient case (cf (4.9)): With u(x, ) = Bltio we have, for 10. ro) The following important result as then proved by WiDLUND [1965, 1966] (cf. also ‘Anowson [1963b]). ‘TutoREM 8.2. Ifthe difference scheme is parabolic in the sense of John, then it is stable in L, for (0 (6 arbitrary in the explicit case) (onan) SCU(n— m+ 1k) ™expl(C|BIM(n—m-+ 1)k)—yh-f. This is done by first freezing the coefficients at an arbitrary point (xo.to) and obtaining estimates for the corresponding constant coefficient problem by the Fourier method, whereby the analyticity of the symbol is used to move the path of integration into the complex. After this the frozen coefficient fundamental solution is employed as a parametrix to obtain the final estimate by a perturbation argument. With suitable choice of 6 this yields estimates of the form e.g. (2am) oreo lok) CU(n—m + 1)k)- en -() ) wh < to f= < Clin m+ Ik) ™exp(—cl yl) yh fo me ‘The first of these are analogous to the corresponding estimates (8.3) for the continuous problem. The second estimate, which is not needed for explicit schemes as then unrestricted, shows that the cont itions for large j are exponentially smail, The above results may also be gener: to certain multistep schemes. Following Wrv.unp [1966] we thus consider schemes of the form BRUT = ARUT + + ABU, 89) where the operators A, j= A, Ok) =F + Yay 6051, hole — fh), 7 mare of the form Az Ay, {nk}, where and similarly Byls)= A, ltA%) =1+ Yay ghx,t, box — Bh), 7 with the sums extending only over finite sets of multi-indices f. We note that theoy V. Thome Charren 1 natural generalizations to variable coefficients of the multistep schemes discussed in Seztion 6 may be writen this form. As caer, in addition othe ita condition U°=0, itis necessaty here to supply methods for the determination of U', ..., U"~', ‘and we shall assume tha this is done in some appropriate way, in particular 0 thal these values are suitably bounded by the initial data o. Jn the same way as previously discussed such a scheme may be represented as a two-evel scheme for an associated system in the variable O*=(U%U"',...,U"- 8), namely BO"? = HO", (BY Aka BAR o o 0 a : | Loo 1 o J Similarly to above we may introduce the principal symbol of the operator Bxy= Esk), BU, 68)" 'Ag 5.2) ' ve Bix, t, = Bix, 1,87 A, = a ° where Ale HE m Ait +S, aah eS, 7 and correspondingly for Blx, 1, é). For a consistent scheme we have, in particular, ix, 0)=Al, where Farm hy 10 0 A ou 0 o 0 10 It follows that the boundedness of 4°, n=0, 1,..., which we shall refer as stabilitySecvion 8 Pure intial value problem 9s ofthe matrix A, is a necessary condition for stability of the scheme (8.9), or of FAS is well known, isa stable matrix if and only ifall eigenvalues of A are in the closed unit disk and all eigenvalues on the unit circle are simple. Recall also that these cigenvalues are the roots of the equation preaye 0. We say that the multistep scheme (89) is parabolic (in the sense of John) if pABlx,t, )<1—clgM for l&jsx, J=1.. uniformly in x and t. The following is now the main result of Win. [1966] about the stability ofthe ‘multistep parabolic scheme. TueoreM 84. Assume that the multistep scheme (8.9) is consistent with (8.1) and that Aisa stable matrix. Then the scheme is stable in L, and LRU", SAln—m+ I), 1SpSo0. In order to determine if a scheme satisfies the parabolicity condition, one may employ the following criterion, see WipLUND [1966]. ‘THEOREM 8.5, Assume that the difference scheme (8.9) is consistent with (8.1), which is parabolic in Petrovskit’s sense. Assume further that all eigenvalues of A except one are in the open unit disk and that for &%0, |f)|<7,j=1.....d, all eigenvalues of Etx,t,€) ‘are in the open unit disk. Then the scheme is parabolic. We note that the requirement on A here is stronger than the stability of this ‘matrix. It may be shown that the theorem does not hold in general when only the stability of 4 is assumed. We shall briefly consider a different parabolicity concept and take smoothing properties like (8.4) and (8.8) as our new definitions (ef. Twowre [1966]. We restrict the discussion to the case of constant coefficients and with the basic space as L,=L,(R*). We consider thus the pure initial value problem for the equation a PD eR, 120, 10) wie where the P, are (NV x N) constant matrices. We say that (8.10) is weakly parabolic if the initial value problem is weakly correctly posed, that is if AIPED)SC for eR, and if for any positive integer m and 00, we have [RU"—D'ut 1,20 as k-+0, mkt, For another concept of a parabolic difference operator, of single-step or multistep type, see Tonkin and MoxiN [1974] and Moxtw [1975]. This concept is associated with the use of a discrete Fourier transform in space and a discrete Laplace transform in time, and results in a priori estimates in discrete analogues of the Sobolev spaces W5. 9. Convergence estimates In the preceding sections we have shown a number of convergence results (cf. e.g. Theorem 3.2) to the effect that a stable finite difference scheme which is accurate of order jis convergent to that same order provided the exact solution is smooth98 ¥. Thome Cuneren 1 enough. We also know by the Lax equivalence theorem (Theorem 3.3) that convergence follows from stability and consistency without any regularity hypo- theses on the data other than that they belong o the space of functions under consideration. Our purpose in this section isto give more precise information about the relation between the rate of convergence and the smoothness of the exact solution, Consider thus an it ial value problem eu a Plt Due = Y Pony xeR, 10, on lew ux, 0)= 0x), xe, where the equation is parabolic (in the sense of Petrovskii), and a consistent finite difference scheme of the form UN TSE. U"= Bet Ana", 30, a= kh, 92) where Ay. A\(nk) and By .= Bink) are finite difference operators of the form used in(8.5). Weshall also assume that the scheme isaccurate of order 1. In Sections Sand 6we have become acquainted with several examples of difference schemes of various ‘orders of accuracy and we also quoted a result in Kress [1959a] where it is shown that schemes of arbitrarily high order of accuracy exist in great generality. In order toexpress our results we need to introduce certain Banach spaces, known, as the Besov spaces B;*, Recall from Section 7 that W,,= WR) denotes L(t) for 10, we introduce the modulus of continuity in W, of order jj=1,2, by 0, ,0)= sup IT,— Helly, bet where T,x(x)=1(x + )) Let now s> and 13,, 0F 5, =53, 4,42, and By WB,” ifs isa natural number, and we may think of a function in By, as one with s derivatives in W,. Here inclusion stands for continuous embedding s0 that in each case a corresponding inequality between norms holds. The Besov spaces are also intermediate spaces between the Sobolev spaces in the sense of the theory of interpolation of Banach spaces. For instance, ifm is a natural number and 0 then v= This shows thus that if the rate of convergence in W, is Oth"), where s0 and assume that ve L.. is such that JU wi], SOW asnk=t, k-+0. ‘Then ve BY". ‘The result of Theorem 9.4 s still best possible, however, in the sense that for some ve Bt, the Oth") convergence rate is best possible. ‘THEOREM 9.6, Let the assumptions of Theorem 9.5 hold and let 00. Then there exists av € BY, such that lim sup h-*]U"—w,_>0. Be In spite of the above, it is, however, possible to attain a Oh") convergence rate in the maximum-norm under weaker assumptions than v€ BY,(s0, _ fx"), 20, rts= fy xe. (9.10) Clearly many functions of one variable occurring in applications are linear combinations of functions ofthis form. As is easily seen, we have 7, Bs if and only if 8 0, b{” and bi" coincide with multipliers on FL, for|¢| <26and |f|>6, respectively. Since the multipliers on ¥L, are simply the functions in L.., the above conditions are scen to be satisfied for p=2 if HE=1+OUEH) as 60, and, for any multi-index £40, GE)= ONE 2X) as 6=2Bx, uniformly in B. Special examples of smoothing operators of orders 1 and 2, respectively, in the case d=1 are Mpa) = © f eo-nay 15) che (9.16) and for general 1, a smoothing operator of order 1 can easily be constructed in the form MiP(x)=h-* f ¥A™*yyeb~y) dy, where y, isa function which is piecewise a polynomial of degree (1—~ 1) and which vanishes outside (~ +4, .—3) for 4 odd and (— y+ 1, — I) for even. In fact, MY,104 V. Thome Cuarree HL J=1,2, corresponds to and, generally, palsin $2) 60) where p, is the polynomial of lowest degree such that p,(sin d=" O(") as C-+0. For p=2, the operator M, corresponding to _ fh lisa, $0= {i log, with 0<6 x we have, by (2.14), JEEOCEY S CECEY|-Isin 4%, and since the right-hand side is periodic it suffices to estimate it for |é| Since le MSO) < Cem Cn- "for [ele R, ie we conclude that W6, respectively, so that the parameter 4. in (9.14) is replaced by y. Our previous smoothing operators are then of orders (1,2) and itis easy to see, for instance, that the simple operator (9.15) has orders (2,1). The following result then holds.106 ¥. Thomée Cuaeten th ‘TueoneM 9.10. Assume that (9.1) and (9.2) are parabolic in the sense of Petrouskii and John, respectively, and that (9.2) is accurate of order p. Let d ss. Then for U" the discrete solution with U°=Myp we have JU"—wi|, SC lvl, t” Foy. In the proofs of these estimates one observes that QU" ~Qutly, <||QUU"—w ip, + Qu~ OWL, 0.20) ‘The second term on the right is easily bounded, using (9.19) and the smoothness properties of u*, so that eg. MQ, Qe ag, SCH apg 9 SCH see or NQ,— Qh |g, SCH OM oa, depending on the case considered. For the first term in (9.20) it sulices to bound 24(U"—u"), which may be done using the appropriate estimate for the fundamental solutions. For the case of constant cocflicients and p=2, sec also the proof of ‘Theorem 43. Results of similar type may be derived in strongly parabolic situations. In addition to the work described above we would like to quote also some recent work by Lazarov [1982] and Wenserr, Lazanov and Sreett [1984]. In these papers convergence estimates are obtained for finite difference schemes for second-order parabolic equations with periodic multidimensional boundary conditions. The particular emphasis is on weak solutions with low regularity, and averages of the data arc used rather than point-valuesin ordet to extract suffcicnt information. The analysis is close to that associated with the finite element method. In Govev and Lazanov [1984] discrete Fourier and Laplace transforms are used to derive convergence estimates in L, for the same problem under weak assumptions.Cuapter ID The Mixed Initial Boundary Value Problem This chapter is concerned with the numerical solution of mixed initial boundary value problems for parabolic equations. The problem whose solution is sought is then required to satisfy the differential equation in a bounded domain @ in space, and is subjected to boundary conditions on the boundary 02 of @, for positive time, and also to assume given initial values. Here the theory of finite difference methods is less complete and satisfactory than for the pure initial value problem. One reason for this is that in the case of more than one space dimension only very special domains, such as unions of rectangles with sides parallel to the coordinate axes, may be well represented by mesh domains. Further, even in one space dimension the transition between the finite difference equations in the interior and the approximation of the boundary conditions is more complex both to define and to analyze. This is the reason why finite element methods, which are based on variational formulations of the boundary valuc problems, have been more successful and caused the develop- ment of the classical finite difference method to stagnate. It should be reiterated that the finite element method is, in some sense, a development and refinement of the finite difference method and that the latter has contributed to the former in a variety of ways. As an illustration of this fact wwe have mentioned earliet that in the Russian literature a class of finite clement methods is referred to as variational finite difference methods. However, in this article we restrict ourselves to classical finite diflerence methods and refer to the article of Fujita and Suzuki in Volume If for the development along the lines just described, We have divided this chapter into four sections out of which the three first are devoted to the three major approaches to methods of analysis of finite difference schemes for initial boundary value problems, namely energy methods, methods based on maximum principles ot monotonicity, and spectral methods, Inthe fourth section we collect some additional topics, not covered in the earlier sections. In Section 10 we thus discuss stability and convergence results derived by diserete analogues of the classical energy arguments for the differential equations. We begin by considering a one-dimensional heat equation with Dirichlet type boundary conditions and show stability and error estimates in discrete Ly-norms for the standard explicit and implicit finite difference methods. We permit variable 109110 V. Thomée (Coupren IL coefficients and discuss the effect of lower-order terms and proceed to treat boundary conditions which involve derivatives. We then consider the extension to several space dimensions, including the casc of domains with curved boundarics When the domain is such that its boundary falls on mesh planes we demonstrate stability in the discrete L,-norm and convergence to the natural orders for the standard finite difference equations with Dirichlet boundary conditions. In the case of a curved boundary we describe a crude variant of the backward Euler method which has a low-order convergence rate and we end the section with a brief discussion of the relation of standard finite difference methods to special cases of the finite element method In Section L1 we consider the same types of problems.as in Section 10, but now the methods of analysis are based on discrete analogues of the maximum principle or related monotonicity arguments, depending on positivity of the coefficients of the finite difference operators. This restricts the generality of the methods but gives somewhat more satisfactory results in cases when they apply. In Section 12 we describe a variety of spectral methods and begin by discussing concept of spectrum relating to a family of operators, which was introduced in GovuNoy and RvaneNki [1964] to deal with stability of finite difference schemes in the presence of boundary conditions. Their approach shows that the stability in the case of one space dimension is dependent on the stability of the interior finite difference equations, and the left-sided and right-sided boundary value problems separately, This makes it natural to discuss the stability of a quarter-plane problem, with the space variable ranging over the positive axis, and we describe the analysis of such a situation by means of Fourier analysis in the vein ofthe school of Kreiss. We close the section with some remarks concerning the stability of initial value type difference operators which are suited for the quarter-plane problem, and concerning the use of eigenfunctions to derive maximum-norm error estimates, Finally, in Section 13, we collect some material which has not naturally fallen into any of the earlier sections. We thus discuss some results relating to the possibility of usinga variable time step, we consider a parabolic problem with a singularity caused by transformation by means of polar coordinates of spherically symmetric problems, and touch upon the case of a discontinuous coefficient in the parabolic equation, We further examine the possibility of treating the initial boundary value problem as.a pure boundary value problem and then present some interior estimates for standard parabolic difference schemes which may be used to draw strong conclusions about convergence away from the boundary from global results. Finally, we present an example of the application of finite difference schemes in existence theory. 10. The energy method In this section we shall consider the application of discrete variants of the energy method, commonly used in the study of boundary value problems for partial differential equations, to derive stability and convergence estimates for finiteSeenow 10 Mixed iil boundary clue problem un difference schemes, In contrast to the Fourier method which was the basis for the analysis described, for instance, in Sections 7 and 8, such methods are suitable for equations with variable coefficients and also for problems with boundary condi- tions which are not of Dirichlet type. This approach was developed first by Lees [1959, 1960a,b], Krewss [1959a¢, 1960] and Samarskit [1961a,b, 1962a] (ct also eg. BaBusKs, PRAGER and VITASEK 11966). ‘We shall first consider the model problem of the one-dimensional heat equation in divergence form, ou. 8 a! 0,)=u(1,)=0 for 130, (0.1) ufx,0)=v(x) for O0, from which we conclude at once, for instance, BU? <0, Worse", whence, by summation, Waa 0, UI=V,= 0h for j= ‘The analysis is similar but (10.8) is replaced by 8,U5-Uj=$10(UI?—K@,UT? 1, and hence the energy identity (10.9) by 40,0" (a28,U%,B, U9 =4e1,U"P. 10.10) Since obviously 12,4 | <2h- "VI, (10.11) we have 12,071 = 12a. 280 <2h-" fa 20,0", and hence from (10,10) and using (10.2), with 1=k/h?, FO [UF + CQ )"78, U8]? S ZAK Ia )2)'78,U" |,us ¥ Thande Curren tL ‘Therefore if 4 is chosen small enough so that UK 0, and obt (Ur, 0) +a" 20,0" 12,8,U8*2)—0, ious) Here 0,08, U8) (Ut — UNV MUN UN) = 40,1, and (10.14) thus immediately yields a1UT?<0, which shows the unconditional stability in this case. For the @-method, (10.15) Unt 80" 1 ~2)U",Secni0n 10 Mixed inital Boundary eae problere us we have similarly @,U%, U8) + (ae Yfs8,U" 3,0 (10.16) Here @,U%, U"*) = 48, UN? + (O—3)ke|9,U" 7, (10.17) and hence, if @24, we obtain at once a Ut <0, which shows unconditional stability. For @<4 we may use the difference equation (10.15) and (10.11) to obtain J8,U"] <2h~*atf,8,U"*, so that, using also (10.2), BO, U1? + Gea) 78,081 <— OKI OU"? (10.18) <(1~202AK (a2 4fa)78,U 2, The stability now follows under the condi 241 20K <1. (10.19) Ir-we look at the prools above we see that, by being slightly more careful, we obtain for the backward Euler method Hw +k Y 18,U"l?

You might also like

• Fear: Trump in the White House

  

  From Everand

  Fear: Trump in the White House

  Bob Woodward

  Rating: 3.5 out of 5 stars

  3.5/5 (738)
• A Man Called Ove: A Novel

  

  From Everand

  A Man Called Ove: A Novel

  Fredrik Backman

  Rating: 4.5 out of 5 stars

  4.5/5 (4609)
• The Sympathizer: A Novel (Pulitzer Prize for Fiction)

  

  From Everand

  The Sympathizer: A Novel (Pulitzer Prize for Fiction)

  Viet Thanh Nguyen

  Rating: 4.5 out of 5 stars

  4.5/5 (120)
• • Magazines
  • Podcasts
  • Sheet music
• Devil in the Grove: Thurgood Marshall, the Groveland Boys, and the Dawn of a New America

  

  From Everand

  Devil in the Grove: Thurgood Marshall, the Groveland Boys, and the Dawn of a New America

  Gilbert King

  Rating: 4.5 out of 5 stars

  4.5/5 (266)
• A Heartbreaking Work Of Staggering Genius: A Memoir Based on a True Story

  

  From Everand

  A Heartbreaking Work Of Staggering Genius: A Memoir Based on a True Story

  Dave Eggers

  Rating: 3.5 out of 5 stars

  3.5/5 (231)
• Grit: The Power of Passion and Perseverance

  

  From Everand

  Grit: The Power of Passion and Perseverance

  Angela Duckworth

  Rating: 4 out of 5 stars

  4/5 (588)
• The Little Book of Hygge: Danish Secrets to Happy Living

  

  From Everand

  The Little Book of Hygge: Danish Secrets to Happy Living

  Meik Wiking

  Rating: 3.5 out of 5 stars

  3.5/5 (399)
• Never Split the Difference: Negotiating As If Your Life Depended On It

  

  From Everand

  Never Split the Difference: Negotiating As If Your Life Depended On It

  Chris Voss

  Rating: 4.5 out of 5 stars

  4.5/5 (838)
• The Subtle Art of Not Giving a F*ck: A Counterintuitive Approach to Living a Good Life

  

  From Everand

  The Subtle Art of Not Giving a F*ck: A Counterintuitive Approach to Living a Good Life

  Mark Manson

  Rating: 4 out of 5 stars

  4/5 (5794)
• The Emperor of All Maladies: A Biography of Cancer

  

  From Everand

  The Emperor of All Maladies: A Biography of Cancer

  Siddhartha Mukherjee

  Rating: 4.5 out of 5 stars

  4.5/5 (271)
• The World Is Flat 3.0: A Brief History of the Twenty-first Century

  

  From Everand

  The World Is Flat 3.0: A Brief History of the Twenty-first Century

  Thomas L. Friedman

  Rating: 3.5 out of 5 stars

  3.5/5 (2219)
• Yes Please

  

  From Everand

  Yes Please

  Amy Poehler

  Rating: 4 out of 5 stars

  4/5 (1891)
• Shoe Dog: A Memoir by the Creator of Nike

  

  From Everand

  Shoe Dog: A Memoir by the Creator of Nike

  Phil Knight

  Rating: 4.5 out of 5 stars

  4.5/5 (537)
• Rise of ISIS: A Threat We Can't Ignore

  

  From Everand

  Rise of ISIS: A Threat We Can't Ignore

  Jay Sekulow

  Rating: 3.5 out of 5 stars

  3.5/5 (137)
• Principles: Life and Work

  

  From Everand

  Principles: Life and Work

  Ray Dalio

  Rating: 4 out of 5 stars

  4/5 (599)
• Team of Rivals: The Political Genius of Abraham Lincoln

  

  From Everand

  Team of Rivals: The Political Genius of Abraham Lincoln

  Doris Kearns Goodwin

  Rating: 4.5 out of 5 stars

  4.5/5 (234)
• The Hard Thing About Hard Things: Building a Business When There Are No Easy Answers

  

  From Everand

  The Hard Thing About Hard Things: Building a Business When There Are No Easy Answers

  Ben Horowitz

  Rating: 4.5 out of 5 stars

  4.5/5 (344)
• The Glass Castle: A Memoir

  

  From Everand

  The Glass Castle: A Memoir

  Jeannette Walls

  Rating: 4.5 out of 5 stars

  4.5/5 (1712)
• John Adams

  

  From Everand

  John Adams

  David McCullough

  Rating: 4.5 out of 5 stars

  4.5/5 (2409)
• The Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You Are

  

  From Everand

  The Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You Are

  Brene Brown

  Rating: 4 out of 5 stars

  4/5 (1090)
• The Perks of Being a Wallflower

  

  From Everand

  The Perks of Being a Wallflower

  Stephen Chbosky

  Rating: 4.5 out of 5 stars

  4.5/5 (2100)
• Angela's Ashes: A Memoir

  

  From Everand

  Angela's Ashes: A Memoir

  Frank McCourt

  Rating: 4.5 out of 5 stars

  4.5/5 (440)
• The Woman in Cabin 10

  

  From Everand

  The Woman in Cabin 10

  Ruth Ware

  Rating: 3.5 out of 5 stars

  3.5/5 (2322)
• Wolf Hall: A Novel

  

  From Everand

  Wolf Hall: A Novel

  Hilary Mantel

  Rating: 4 out of 5 stars

  4/5 (3811)
• Her Body and Other Parties: Stories

  

  From Everand

  Her Body and Other Parties: Stories

  Carmen Maria Machado

  Rating: 4 out of 5 stars

  4/5 (821)
• A Tree Grows in Brooklyn

  

  From Everand

  A Tree Grows in Brooklyn

  Betty Smith

  Rating: 4.5 out of 5 stars

  4.5/5 (1929)
• Sing, Unburied, Sing: A Novel

  

  From Everand

  Sing, Unburied, Sing: A Novel

  Jesmyn Ward

  Rating: 4 out of 5 stars

  4/5 (1103)
• Elon Musk: Tesla, SpaceX, and the Quest for a Fantastic Future

  

  From Everand

  Elon Musk: Tesla, SpaceX, and the Quest for a Fantastic Future

  Ashlee Vance

  Rating: 4.5 out of 5 stars

  4.5/5 (474)
• The Art of Racing in the Rain: A Novel

  

  From Everand

  The Art of Racing in the Rain: A Novel

  Garth Stein

  Rating: 4 out of 5 stars

  4/5 (4200)
• Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race

  

  From Everand

  Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race

  Margot Lee Shetterly

  Rating: 4 out of 5 stars

  4/5 (895)
• The Unwinding: An Inner History of the New America

  

  From Everand

  The Unwinding: An Inner History of the New America

  George Packer

  Rating: 4 out of 5 stars

  4/5 (45)
• The Outsider: A Novel

  

  From Everand

  The Outsider: A Novel

  Stephen King

  Rating: 4 out of 5 stars

  4/5 (1839)
• The Yellow House: A Memoir (2019 National Book Award Winner)

  

  From Everand

  The Yellow House: A Memoir (2019 National Book Award Winner)

  Sarah M. Broom

  Rating: 4 out of 5 stars

  4/5 (98)
• The Constant Gardener: A Novel

  

  From Everand

  The Constant Gardener: A Novel

  John le Carré

  Rating: 3.5 out of 5 stars

  3.5/5 (104)
• On Fire: The (Burning) Case for a Green New Deal

  

  From Everand

  On Fire: The (Burning) Case for a Green New Deal

  Naomi Klein

  Rating: 4 out of 5 stars

  4/5 (73)
• Tamil Computer Book - Web Design

  

  Document33 pages

  Tamil Computer Book - Web Design

  Sakthivel

  100% (28)
• Brooklyn: A Novel

  

  From Everand

  Brooklyn: A Novel

  Colm Toibin

  Rating: 3.5 out of 5 stars

  3.5/5 (1937)
• The Light Between Oceans: A Novel

  

  From Everand

  The Light Between Oceans: A Novel

  M.L. Stedman

  Rating: 4.5 out of 5 stars

  4.5/5 (789)
• Little Women

  

  From Everand

  Little Women

  Louisa May Alcott

  Rating: 4 out of 5 stars

  4/5 (104)
• Bad Feminist: Essays

  

  From Everand

  Bad Feminist: Essays

  Roxane Gay

  Rating: 4 out of 5 stars

  4/5 (1015)
• Manhattan Beach: A Novel

  

  From Everand

  Manhattan Beach: A Novel

  Jennifer Egan

  Rating: 3.5 out of 5 stars

  3.5/5 (792)
• Steve Jobs

  

  From Everand

  Steve Jobs

  Walter Isaacson

  Rating: 4.5 out of 5 stars

  4.5/5 (806)
• Siemens Basics of AC Drives

  

  Document116 pages

  Siemens Basics of AC Drives

  Sleman Shackoor

  No ratings yet
• Euro Benefits

  

  Document2 pages

  Euro Benefits

  Karthiganesbe85

  No ratings yet
• 6 Exchanges

  

  Document1 page

  6 Exchanges

  Karthiganesbe85

  No ratings yet
• Averaging

  

  Document3 pages

  Averaging

  Karthiganesbe85

  No ratings yet
• How To Study AND Take Exams: The Competition Master

  

  Document85 pages

  How To Study AND Take Exams: The Competition Master

  Karthiganesbe85

  No ratings yet
• Gloss

  

  Document1 page

  Gloss

  Karthiganesbe85

  No ratings yet
• Smartphone

  

  Document28 pages

  Smartphone

  Karthiganesbe85

  No ratings yet
• Auth

  

  Document1 page

  Auth

  Karthiganesbe85

  No ratings yet
• Introduction To Cryptanalyst: Part One

  

  Document3 pages

  Introduction To Cryptanalyst: Part One

  Karthiganesbe85

  No ratings yet
• Satellite TV

  

  Document11 pages

  Satellite TV

  Karthiganesbe85

  No ratings yet
• CH 11

  

  Document7 pages

  CH 11

  Karthiganesbe85

  No ratings yet
• Satellite Radio

  

  Document20 pages

  Satellite Radio

  Karthiganesbe85

  No ratings yet
• Ac S P

  

  Document75 pages

  Ac S P

  Anonymous FigYuONxuu

  No ratings yet
• How SMS Works

  

  Document8 pages

  How SMS Works

  Karthiganesbe85

  No ratings yet
• (Parenting) How To Help Your Child Succeed in School (2005 Edition)

  

  Document46 pages

  (Parenting) How To Help Your Child Succeed in School (2005 Edition)

  dorama81

  No ratings yet
• Camcoder

  

  Document21 pages

  Camcoder

  Karthiganesbe85

  No ratings yet
• Tamil Stories Siru Kathaigal Ranja Kumar

  

  Document60 pages

  Tamil Stories Siru Kathaigal Ranja Kumar

  Karthiganesbe85

  No ratings yet
• Current Affairs 2012-2014 Study Material

  

  Document434 pages

  Current Affairs 2012-2014 Study Material

  Karthiganesbe85

  No ratings yet
• Current Affairs January 2014 Bit Bank

  

  Document4 pages

  Current Affairs January 2014 Bit Bank

  Karthiganesbe85

  No ratings yet
• 5807 Simulink Tutorial

  

  Document31 pages

  5807 Simulink Tutorial

  Suresh

  No ratings yet
• Bogdan M. Wilamowski, J. David Irwin Power Electronics and Motor Drives, 2nd Edition The Industrial Electronics Handbook 2011

  

  Document1 page

  Bogdan M. Wilamowski, J. David Irwin Power Electronics and Motor Drives, 2nd Edition The Industrial Electronics Handbook 2011

  Karthiganesbe85

  No ratings yet
• File List

  

  Document5 pages

  File List

  Habiburrahman Vicky Fis

  No ratings yet
• Current Affairs February 2014 Bit Bank

  

  Document4 pages

  Current Affairs February 2014 Bit Bank

  Karthiganesbe85

  No ratings yet
• மனசே ரிலக் ப்ளீ மனசே ரிலாக்ஸ் ப்ளிஸ் பாகம்-2

  

  Document100 pages

  மனசே ரிலக் ப்ளீ மனசே ரிலாக்ஸ் ப்ளிஸ் பாகம்-2

  Sathiyaraj

  100% (3)
• Sheet 1 of 18: SAR Synthetic Aperture Radar Why Use Microwave Frequencies?

  

  Document18 pages

  Sheet 1 of 18: SAR Synthetic Aperture Radar Why Use Microwave Frequencies?

  Kai Hee Goh

  No ratings yet
• Delaunay Triangulation and Meshing - George & Borouchaki

  

  Document431 pages

  Delaunay Triangulation and Meshing - George & Borouchaki

  Karthiganesbe85

  No ratings yet
• Boundary Elements An Introductory Course - Brebbia and Dominguez

  

  Document325 pages

  Boundary Elements An Introductory Course - Brebbia and Dominguez

  Karthiganesbe85

  100% (3)
• RF Amp Tutorial

  

  Document17 pages

  RF Amp Tutorial

  Mihail

  No ratings yet