See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/301985928

Ordinary Differential Equations

Chapter · December 2014

DOI: 10.1007/978-3-642-39386-0_5

CITATIONS READS

0 291

1 author:

Wolfgang Hackbusch
Max Planck Institute for Mathematics in the Sciences
501 PUBLICATIONS 21,752 CITATIONS

SEE PROFILE

All content following this page was uploaded by Wolfgang Hackbusch on 13 June 2016.

The user has requested enhancement of the downloaded file.

Springer Series in Computational Mathematics 45

Wolfgang Hackbusch

The Concept
of Stability
in Numerical
Mathematics
x Contents

3.4.7 Uniform Boundedness Theorem . . . . . . . . . . . . . . . . . . . . . . . . 36

3.4.8 Necessity of the Stability Condition, Equivalence Theorem . 38
3.4.9 Modified Definitions for Consistency and Convergence . . . . 39
3.5 Further Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.5.1 General Intervals and Product Quadrature . . . . . . . . . . . . . . . . 41
3.5.2 Consistency Versus Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.5.3 Perturbations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.5.4 Arbitrary Slow Convergence Versus Quantitative
Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

4 Interpolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.1 Interpolation Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.2 Convergence and Consistency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.3 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.4 Equivalence Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.5 Instability of Polynomial Interpolation . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.6 Is Stability Important for Practical Computations? . . . . . . . . . . . . . . . 53
4.7 Tensor Product Interpolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.8 Stability of Piecewise Polynomial Interpolation . . . . . . . . . . . . . . . . . 56
4.8.1 Case of Local Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.8.2 Spline Interpolation as an Example for Global Support . . . . . 57
4.9 From point-wise Convergence to Operator-Norm Convergence . . . . . 59
4.10 Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

5 Ordinary Differential Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

5.1 Initial-Value Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.1.1 Setting of the Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.1.2 One-Step Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.1.3 Multistep Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.2 Fixed-Point Theorem and Recursive Inequalities . . . . . . . . . . . . . . . . 66
5.3 Well-Conditioning of the Initial-Value Problem . . . . . . . . . . . . . . . . . 68
5.4 Analysis of One-Step Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.4.1 Implicit Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.4.2 Lipschitz Continuity of φ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.4.3 Consistency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.4.4 Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.4.5 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.5 Analysis of Multistep Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
5.5.1 Local Discretisation Error, Consistency . . . . . . . . . . . . . . . . . . 75
5.5.2 Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.5.3 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.5.4 Difference Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.5.5 Stability and Convergence Theorems . . . . . . . . . . . . . . . . . . . . 83
Contents xi

5.5.6 Construction of Optimal Multistep Methods . . . . . . . . . . . . . . 85

5.5.7 Further Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
5.5.8 Other Stability Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

6 Instationary Partial Differential Equations . . . . . . . . . . . . . . . . . . . . . . . . 93

6.1 Introduction and Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.1.1 Notation, Problem Setting, Function Spaces . . . . . . . . . . . . . . 93
6.1.2 The Hyperbolic Case A = a∂/∂x . . . . . . . . . . . . . . . . . . . . . . 95
6.1.3 The Parabolic Case A = ∂ 2 /∂x2 . . . . . . . . . . . . . . . . . . . . . . . 96
6.2 Semigroup of Solution Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6.3 Discretisation of the Partial Differential Equation . . . . . . . . . . . . . . . . 100
6.3.1 Notations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
6.3.2 Transfer Operators r, p . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
6.3.3 Difference Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
6.4 Consistency, Convergence, and Stability . . . . . . . . . . . . . . . . . . . . . . . 105
6.4.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
6.4.2 Convergence, Stability and Equivalence Theorems . . . . . . . . 106
6.4.3 Other Norms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.5 Sufficient and Necessary Conditions for Stability . . . . . . . . . . . . . . . . 109
6.5.1 First Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
6.5.2 Fourier Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
6.5.3 Further Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
6.5.4 Implicit Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
6.5.5 Vector-Valued Grid Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 124
6.5.6 Generalisations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
6.5.7 Dissipativity for Parabolic Discretisations . . . . . . . . . . . . . . . . 135
6.6 Consistency Versus Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

7 Stability for Discretisations of Elliptic Problems . . . . . . . . . . . . . . . . . . . 139

7.1 Elliptic Differential Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
7.2 Discretisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
7.3 General Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
7.3.1 Consistency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
7.3.2 Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
7.3.3 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
7.4 Application to Difference Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
7.4.1 Classical Choice of Norms . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
7.4.2 Bijectivity of L . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
7.5 Finite Element Discretisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
7.5.1 Variational Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
7.5.2 Galerkin Discretisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
7.5.3 Consistency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
7.5.4 Convergence and Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
 Chapter 5
Ordinary Differential Equations

The numerical treatment of ordinary differential equations is a field whose scope

has broadened quite a bit over the last fifty years. In particular, a whole spectrum
of different stability conditions has developed. Since this chapter is not the place to
present all details, we concentrate on the most basic concept of stability. As a side-
product, it will lead us to the power bounded matrices, which is a class of matrices
with certain stability properties. More details about ordinary differential equations
can be found, e.g., in [20], [23], [4, 3], [5, §§5-6], and in the two volumes [8], [9].

5.1 Initial-Value Problem

5.1.1 Setting of the Problem

Let f : R × R → R be a continuous function.1 In what follows we are looking

for continuously differentiable functions y(x) satisfying the ordinary differential
equation
y 0 (x) = f (x, y(x)). (5.1a)
The initial-value problem requires finding a solution y of (5.1a) which, in addition,
satisfies
y(x0 ) = y0 (5.1b)
for a given ‘initial value’ y0 .
Usually one is looking for the solution y at x ≥ x0 , either in a finite2 interval
I := [x0 , xE ] or in the unbounded interval I := [x0 , ∞). Correspondingly, f needs
to be defined on I × R.

1
In the case of a system of differential equations, f is defined in R × Rn and the solution y ∈
C 1 (R, Rn ) is vector-valued. For our considerations it is sufficient to study the scalar case n = 1.
2
However, it may happen that the solution exists only on a smaller interval [x0 , xS ) ⊂ [x0 , xE ].

W. Hackbusch, The Concept of Stability in Numerical Mathematics, 63

Springer Series in Computational Mathematics 45, DOI 10.1007/978-3-642-39386-0_5,
© Springer-Verlag Berlin Heidelberg 2014

View publication stats