ODEbox: A Toolbox for Ordinary Differential Equations

Sturm Liouville Example 3

The Truncated Hydrogen Differential Equation
Matthew Harker and Paul O’Leary
Institute for Automation
University of Leoben
A-8700 Leoben, Austria
URL: automation.unileoben.ac.at

Original: January 9, 2013

c 2013

Last Modified: February 27, 2013

Contents
1 The Mathieu Differential Equation 2

2 tidy up the Workspace 2

3 Define the Nodes and Compute the Basis Functions 2

3.1 Generate the Differentiating Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . 2
3.2 Define the Constraints and Compute the Admissible Functions . . . . . . . . . . . 2

4 Setup the Sturm-Liouville Matrix Linear Differential Operator 3

5 Solve the Eigenvector Problem 3

6 Known Eigenvalues 3

7 Plot the Results 4

8 Write the Figures to Disk 6

1
1 The Mathieu Differential Equation
This example addresses the solution of the truncated hydrogen differential equation, a Sturm-
Liouville equation. This is a boundary value problem with an LNP singular condition at y(0). A
series of eigenvalues are known for this problem, they will be used here to evaluate the quality of
the eigenvalues computed numerically.
 
2 1
−ÿ + − = λ y with y(0) = LN P and y(1000) = 0 (1)
x2 x

in the closed interval 0 ≤ x ≤ 1000.

2 tidy up the Workspace

1 clear all;
2 close all;
3 setUpGraphics;

3 Define the Nodes and Compute the Basis Functions

Define the number of nodes and the number of basis functions used.
4 noPts = 1000;
5 noBfs = round(noPts/2);
Compute the Chebyshev points, but scaled to form a open interval 0 < x < 1000.
6 x = dopNodes( noPts, ’cheby’ );
7 xMax = 1000;
8 x = xMax * (x + 1)/2;
Synthesize a complete set of basis functions
9 B = dop(x);

3.1 Generate the Differentiating Matrix

Synthesize the differentiating matrix with support length ls
10 ls = 13;
11 D = dopDiffLocal( x, ls, ls );

3.2 Define the Constraints and Compute the Admissible Functions

Define the constraints, in this example only one constraint.
12 C = zeros( noPts, 1 );
13 C(end,end) = 1;
Compute the set of constrained basis functions, i.e., admissible functions.

2
14 Bc = dopConstrain( C, B );
Truncate the constrained basis functions to avoide aliasing
15 Bc = Bc(:,1:noBfs);

4 Setup the Sturm-Liouville Matrix Linear Differential

Operator
Setup the Linear differential operator for the differential equation.
16 Gx = diag((1./x).*(2./x - 1) ) ;
17 L = Bc’ * (-D * D + Gx)* Bc;

5 Solve the Eigenvector Problem

Solve the eigenvalue problem
18 [Vec, Val] = eig( L );
19 vals = diag( Val );
and sort the eigenvalues
20 [vals, inds] = sort(vals);
21 Vec = Vec(:,inds);
Compute the eigenfunctions
22 sols = Bc * Vec;

6 Known Eigenvalues
Some of the eigenvalues are known for this differential equation [2, 1]
23 L0Known = -6.250000000000E-2;
24 L9Known = -2.066115702478E-3;
25 L17Known = -2.5757359232E-4;
26 L18Known = 2.873901310E-5;
Display the relative error in the numerically computed eigenvalues with respect to the known
eigenvalues.
27 disp( [’L_’,int2str(0),’ = ’,num2str(vals(1),’%1.10E’),’ ’,num2str(
L0Known,’%1.10E’),’ ’,num2str((L0Known - vals(1))/L0Known,’
%1.10E’) ] );
28 disp( [’L_’,int2str(9),’ = ’,num2str(vals(10),’%1.10E’),’ ’,num2str(
L9Known,’%1.10E’),’ ’,num2str((L9Known - vals(10))/L9Known,’
%1.10E’) ] );

3
29 disp( [’L_’,int2str(17),’ = ’,num2str(vals(18),’%1.10E’),’ ’,num2str(
L17Known,’%1.10E’),’ ’,num2str((L17Known - vals(18))/L17Known,’
%1.10E’) ] );
30 disp( [’L_’,int2str(18),’ = ’,num2str(vals(19),’%1.10E’),’ ’,num2str(
L18Known,’%1.10E’),’ ’,num2str((L18Known - vals(19))/L18Known,’
%1.10E’) ] );

L_0 = -6.2499999978E-02 -6.2500000000E-02 3.4874503285E-10

L_9 = -2.0661156136E-03 -2.0661157025E-03 4.3009091823E-08
L_17 = -2.5757218232E-04 -2.5757359232E-04 5.4741446402E-06
L_18 = 2.8740937561E-05 2.8739013100E-05 -6.6963370220E-05

7 Plot the Results

32 setUpGraphics(10)
33 FigureSize=[1 1 10 10];
34 set(0,’DefaultFigureUnits’,’centimeters’);
35 set(0,’DefaultFigurePosition’,FigureSize);
36 set(0,’DefaultFigurePaperUnits’,’centimeters’);
37 set(0,’DefaultFigurePaperPosition’,FigureSize);
38 MyAxesPosition=[0.16 0.17 0.8 0.8];
39 set(0,’DefaultaxesPosition’,MyAxesPosition);
40 %
41 fig1 = figure;
42 P1 = [0.18 0.1 0.8 0.17];
43 A1 = axes( ’position’, P1, ’color’, ’w’ );
44 plot( x, sols(:,1), ’k’);
45 grid on;
46 xlabel(’ x ’);
47 ylabel(’y(x)_{0}’);
48 axis([0,xMax,-0.01,0.2]);
49 hold on;
50 %
51 P2 = [0.18 0.33 0.8 0.17];
52 A2 = axes( ’position’, P2, ’color’, ’w’ );
53 plot( x, sols(:,10), ’k’);
54 grid on;
55 ylabel(’y(x)_{9}’);
56 hold on;
57 %
58 P3 = [0.18 0.56 0.8 0.17];
59 A3 = axes( ’position’, P3, ’color’, ’w’ );
60 plot( x, sols(:,18), ’k’);
61 grid on;
62 ylabel(’y(x)_{17}’);
63 hold on;
64 %

4
65 P4 = [0.18 0.79 0.8 0.17];
66 A4 = axes( ’position’, P4, ’color’, ’w’ );
67 plot( x, sols(:,19), ’k’);
68 grid on;
69 grid on;
70 ylabel(’ y(x) ’);
71 ylabel(’y(x)_{18}’);
72 hold on;

y(x)18 0.1

−0.1
0 200 400 600 800 1000
0.1
y(x)17

−0.1
0 200 400 600 800 1000
0.1
y(x)9

−0.1
0 200 400 600 800 1000
0.2
y(x)0

0.1

0
0 200 400 600 800 1000
x

Figure 1: The eigenfunctions corresponding to the 4 known eigenvalues.

73 fig2 = figure;
74 P1 = [0.16 0.4 0.8 0.55] ;
75 A = axes(’position’,P1);
76 imagesc( log10(abs(Vec) ));
77 colorbar;
78 ylabel([’Basis function number m, n_b = ’,int2str(noBfs),’’]);
79 P2 = [0.16 0.1 0.615 0.25] ;
80 A = axes(’position’,P2);
81 plot(vals , ’k’);
82 range = axis;
83 axis([0,noBfs,0,5]);
84 grid on;
85 xlabel(’Eigenvector number n’);
86 ylabel(’\% rel. error in \lambda’);

5
 Basis function number m, nb = 500
−2
100
−4

200 −6

−8
300
−10

400 −12

−14
500
100 200 300 400 500
% rel. error in λ

0
0 100 200 300 400 500
Eigenvector number n

Figure 2: The Rayleigh-Ritz spectrum of the eigenfunctions with respect to the admissible func-
tions.

8 Write the Figures to Disk

87 fileType = ’eps’;
88 printFigure( fig1, [’SLEx5EigFnsNb’,int2str(noBfs)], fileType);
89 printFigure( fig2, [’SLEx5SpectrumNb’,int2str(noBfs)], fileType);

References
[1] Pierluigi Amodio and Giuseppina Settanni. A matrix method for the solution of sturm-liouville
problems. Journal of Numerical Analysis, Industrial and Applied Mathematics (JNAIAM),
6(1-2):1–13, 2011.

[2] J. D. Pryce. A test package for sturm-liouville solvers. ACM Trans. Math. Softw., 25(1):21–57,
March 1999.