Professional Documents
Culture Documents
N° Courant PDF
N° Courant PDF
Lax*
Abstract: The portion of the Courant-Friedrichs-Lwy paper [Math. Ann. 100, 32 (1928)] that was devoted to hyperbolic difference
equations is critically reviewedin terms of its basic contribution to the numerical solution of partial differential equations. Some subse-
quent developments are then discussed, including recent literature related to the von Neumann condition, some irreversible schemes,
generalizations of the energy method, some new difference schemes, and mixed initial boundary value problems.
At the time their paper was written, Courant, Friedrichs Here is a brief description of Part I1 of the paper, con-
and Lewy’ were interested in difference equations as a cerning hyperbolic equations: The authors start by point-
tool for proving the existence of solutions of partial dif- ing out that a scheme cannot be a convergent one if the
ferential equations; discretization was a way of construct- ratio of the time and space mesh is so large that thedomain
ing a sequence of finite dimensional problems whose of dependence of a point in the difference scheme does not
solutions tended in the limit to the solution of the con- contain all points of its domain of dependence in the dif-
tinuous problem.** For this reason the authors developed ferential equation; such a scheme ignores information
a method which was applicable to equations with variable which does influence the value of the solution of the dif-
coefficients; for the same reason they were not interested ferential equation. This fact is obvious to those who are
in studyingall possible difference schemes but were satisfied used to thinking in these terms but was very much un-
when they found one that worked. suspected by quite a large number of otherwise eminent
Today the methods of functional analysis-projection, scientists who, in their haste to use too large a value of
mollification-are so well developed (thanks in large part At, fell into the trapof using difference schemes that could
to the pioneering work of Friedrichs) that descent to the not converge and which in fact did diverge spectacularly.
finite dimensional case is not necessary for proving the The authors start with the centered difference scheme
existence of solutions of linear problems.+ On the other forthe one-dimensional wave equation where A x / A t
hand, the onset of the Second World War brought an equals the sound speed. The resulting difference equations
unprecedented (and unabated) pressure on the technologi- are extremely simple and can be solved explicitly. Then
cal centers to providenumericalsolutions of partial- the authors turn to the centered difference scheme for the
differential equations; here the ideas introduced by wave equation on an arbitrary grid and remark that they
C-F-L-the famous stability condition, energy inequalities, will not bother to write down an explicit representation of
and the leapfrog difference method-turned out to be the solution because “it is too complicated to yield a
basic. This is an outstanding instance of research under- limiting value easily as the meshwidth tends to zero;”
taken for purely theoretical purposes’ turning out to be of thereby they failed to discover the von Neumann stability
immense practical importance. criterion.*
Courant Institute of Mathematical Sciences. New York University. The convergence of the difference scheme is proved by
** Courant was inspiredpartlybyEuler’s use of discretization in the carrying over the energy method developed by Friedrichs
calculus of variations. Also. the authors were aware of the importance of dif-
ference methods for elliptic equations. and Lewy for thewave e q ~ a t i o nThe
. ~ crux of that method
t Quite to the contrary, the difference method may he the only effective
one for dealing with severely nonlinear problems, as indicated by the success of is a quadratic integral identity obtained as follows: multi-
Glimm’s recent work on shock waves.6
I Curiously, in the late twenties the authors were interested through the ply the wave equation by U t , write the product as a di-
work of Prandtl in hyperbolic problems of compressible gas dynamics. They
considered solving these equations numerically by the method of character- * The explicit representation was given by Lewy in a paper16 published 235
istics but not by difference schemes. twenty years later.
P. D. LAX
The accuracy of a difference scheme is measured by how sibledifferenceschemes for themcanbe quite useful.
closely solutions of the differential equation satisfy the Such a difference scheme was introduced by Friedrichs in
difference equation. A scheme is accurate of order m if Ref. 4. In some nonlinear casesthe slight dissipationwhich
for every smooth solution u of the differential equation makes the differenceschemeirreversibleisevenindis-
there is a constant k such that pensable.
It is not hard to show that the von Neumann condition is
Ilu(t + 6) - Ss~(t)lII k6"". (2) necessary for the stability of schemes with variable coef-
ficients as well. Von Neumann conjectured that it is also
The method of C-F-L shows that if a scheme is stable in sufficient; something like thishas been demonstrated by a
the sense of (l), and accurate of order m in the sense of surprisingly elaborate extension of the energymethod.
(2), then for smooth data the overall error is O(6"): In the hands of Courant, Friedrichs, and Lewy, energy
inequalities were derived from energy identities, obtained
from difference analogs of formulas for the differentiation
Ilu(t) - S;u(O)[l I peat a", t = n6. (3)
of products. This hasbeen generalized and systematized by
Lees'5 who has devised a large number of inequalities in
Somewhat surprisingly, Strang proved in Ref. 20 that the this way. It was observed in Ref. 11 that all such energy
same error estimate (3) holds alsofor weakly stable schemes inequalities can be obtained by the purely algebraic method
which satisfy merely (l)N, provided that the solution u is of expressingcertainnon-negativesymbols as sums of
sufficiently differentiable. squares; according to Hilbert'stheory of higher order
Weak stability is very hard to verify for operators with positive forms,this is not always possibleand so the classi-
variable coefficients since it is an unstable concept, in the cal energy method has inherent limitations. Nonetheless,
+
following sense:If Sssatisfy (I),,,, an operator Si = Ss M it is possible to derive energy inequalities without energy
where the IIM([5 const. X 6, need not satisfy ( l ) N .In identities as was done, for instance, in Ref. 13 with the
contrast, the class of strongly stable operators satisfying aid of a certain amount of dissipation, and in Ref. 12 with
(1) is stable in this sense. the aid of nothing at all except sufficient differentiability
In Ref. 23 ThomCemakes a precise and surprisingly of the coefficients. The theorem of Lax and Nirenberg
delicate study of the degree of stability of a class of opera- reads: If the symbol isa sufficiently differentiable function
tors in the maximum norm. of x , E and if IS(x,()I 5 1 for all x and E, then the operator
The symbol (amplification matrix)of the operator S6is Sssatisfies the inequality.
For operators with coefficients independentof x it follows Obviously the expression of (4) implies stability in any
immediately from the isometric character of Fourier trans- finite time interval. The more delicate problemof proving
formation that necessary and sufficient for stability is the stability when the symbol is not bounded by inequality (1)
uniform boundedness of all powers of the symbol. This in norm but merely satisfies the von Neumann condition
reduces a stabilityquestion to a purematrix problem; has been handledby Kreiss.' He found that if in addition to
this matrixproblem,however, turns out to betrickier the von Neumann condition one requires dissipation in a
than it seems at first glance. An obvious necessary condi- definite sense then it is possible to introduce a new norm
tion, the famousonedue to vonNeumann,is for the equivalent to the L2 norm for which inequality (4) is true.
spectrum of such a matrix to lie in the unit disc in the The proof of the surprisinglydelicatematrixtheorems
complexplane. An obvioussufficientconditionis for needed to do this have been simplifiedby Parlett."
the norm of this matrix to be 5 1. A non-obvious suf- As already mentioned, earlier practical needs have called
ficient condition is that the numerical range of the matrix for a whole armyof novel difference schemes. Among these
be in the unit disc.' 3 ' 4 Necessary and sufficient conditions one might single out the crude but useful method intro-
were given by Kreiss' and Buchanan;' see also M ~ r t o n ' ~ duced by Friedrichs; the more accurate methods devised
and Morton and Schechter." by Du Fort-Frankel: by Lax-Wendr~ff,'~and the high-
If the difference scheme is reversible, i.e., can be used order schemesstudiedby Strang?l and manyothers.
to compute solutions both forward and backward in time, As more and more complicated schemes are being tried
then both positive and negative powersof the symbol have for more and more complicated systems of equations in
to be bounded. In this case the von Neumann condition more and morevariables, the verification of the von
requires that the spectrum be located on the unit circle. Neumann condition, although merely algebraic, becomes
One of the new developments has been the observation quite difficult and is often done numerically on the ma-
that although hyperbolic problems are reversible, irrever- chine. 237
238
P. D. LAX