Table Of ContentApplications of Lie group integrators
and exponential schemes
Workshop on Lie group methods and control
theory,
Edinburgh, 28.6-1.7, 2004,
Brynjulf Owren
Dept of Math Sci, NTNU
ApplicationsofLiegroupintegratorsandexponentialschemes –p.1/60
Report Documentation Page Form Approved
OMB No. 0704-0188
Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and
maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information,
including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington
VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if it
does not display a currently valid OMB control number.
1. REPORT DATE 2. REPORT TYPE 3. DATES COVERED
03 JAN 2005 N/A -
4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER
Applications of Lie group integrators and exponential schemes
5b. GRANT NUMBER
5c. PROGRAM ELEMENT NUMBER
6. AUTHOR(S) 5d. PROJECT NUMBER
5e. TASK NUMBER
5f. WORK UNIT NUMBER
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION
Dept of Math Sci, NTNU REPORT NUMBER
9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR’S ACRONYM(S)
11. SPONSOR/MONITOR’S REPORT
NUMBER(S)
12. DISTRIBUTION/AVAILABILITY STATEMENT
Approved for public release, distribution unlimited
13. SUPPLEMENTARY NOTES
See also ADM001749, Lie Group Methods And Control Theory Workshop Held on 28 June 2004 - 1 July
2004., The original document contains color images.
14. ABSTRACT
15. SUBJECT TERMS
16. SECURITY CLASSIFICATION OF: 17. LIMITATION OF 18. NUMBER 19a. NAME OF
ABSTRACT OF PAGES RESPONSIBLE PERSON
a. REPORT b. ABSTRACT c. THIS PAGE UU 60
unclassified unclassified unclassified
Standard Form 298 (Rev. 8-98)
Prescribed by ANSI Std Z39-18
Outline
PART I (Introductory)
(cid:15) Linear IVPs, Eigenvalue problems, linear PDEs
(cid:15) Manifolds (“stay on manifold” principle)
(cid:15) Classical problems (“curved path” principle)
PART II (Recent results on exp ints)
(cid:15) A unified approach to exponential integrators
(cid:15) Order theory
(cid:15) Bounds for dimensions of involved function spaces
ApplicationsofLiegroupintegratorsandexponentialschemes –p.2/60
I.1 Linear IVPs
One may for instance write
R Rn(cid:2)n
u_ = A(t) u; A : !
In literature, usually u 2 Rn.
LGI: Magnus series or related (Cayley etc)
When/Why use this scheme.
1. Highly oscillatory ODEs, large imaginary eigenvalues.
Iserles
2. PDEs, A(t) unbounded, classical example: Linear
Schrödinger equation (LSE)). Blanes & Moan,
Hochbruck & Lubich.
Recently also Landau-Lifschitz equation Sun, Qin, Ma
ApplicationsofLiegroupintegratorsandexponentialschemes –p.3/60
I.1 Magnus works on LSE!
du
i = H(t) u; H(t) unbounded, selfadjoint
dt
Z
d exp is not invertible for 2k(cid:25)i 2 (cid:27)(u); k 2 nf0g.
u
Truncated series is still unbounded at 1.
H & L find error bounds of the form
p p(cid:0)1
ku (cid:0) u(t )k = C h t max kD u(t)k
m m m
0(cid:20)t(cid:20)t
m
D is a “differentiation operator” related to the LSE.
ApplicationsofLiegroupintegratorsandexponentialschemes –p.4/60
Eigenvalue problems
Stability of travelling wave solutions to PDEs. Boils down
to eigenvalue problem
_
Y = A(t; (cid:21)) Y
where (cid:21) is a parameter.
Needs to be solved for several (cid:21).
Magnus integrators used with success by Malham, Oliver
and others.
Early work by Moan on such problems.
ApplicationsofLiegroupintegratorsandexponentialschemes –p.5/60
I.2 Problems on (nonlinear) manifolds
A large part of the applications I know involves the
orthogonal group which acts transitively on either of
(cid:15) The orthogonal group itself (or its tangent bundle).
(cid:15) Stiefel manifold. (n (cid:2) p matrices with orthonormal
columns)
(cid:15) The n (cid:0) 1-sphere. (Stiefel with p = 1)
ApplicationsofLiegroupintegratorsandexponentialschemes –p.6/60
I.2 Orthogonal group problems
Most used examples are on n = 3 (3D rotations): Free
rigid body, spinning top,. . .
Most LGIs work. RKMK, Crouch-Grossmann,. . .
Scheme.
combined with all possible “coordinates” exp, Cayley,
CCSK etc.
My evaluation
(cid:15) Most Lie group integrators do little else for you than
maintaining orthogonality.
(cid:15) Poor long-time behaviour.
(cid:15) Hard to get reversible / symplectic schemes.
(cid:15) There are exceptions (Lewis and Simo, Zanna et
al.) but these LGIs seem expensive.
ApplicationsofLiegroupintegratorsandexponentialschemes –p.7/60
I.2 Stiefel manifolds
Some applications which involve computation on Stiefel
manifolds
(cid:15) Computation of Lyapunov exponents
(cid:15) Multivariate data analysis (optimisation, gradient
flows)
(cid:15) Neural networks, Independent Component Analysis
Maintain orthonormality. Inexpensive stepping,
Demands.
cost O(np2) per step.
Most LGIs work. RKMK,
Schemes.
Crouch-Grossmann,. . . combined with all possible
“coordinates” exp, Cayley, CCSK etc. Most of them
can be implemented in O(np2) ops per step, but
special care must be taken.
ApplicationsofLiegroupintegratorsandexponentialschemes –p.8/60
I.2 Stiefel manifolds
My evaluation
(cid:15) Lie group integrators meet requirements specified
in literature
(cid:15) Long-time behaviour has not been an issue.
(cid:15) Overall judgement: Lie group integrators are
competitive, if not superior to classical integrators.
Sources
(cid:15) Dieci, Van Vleck [schemes, but also general
viewpoints, Lyapunov exponents]
(cid:15) Trendafilov. [Multivariate data analysis]
(cid:15) Celledoni, Fiori.[Neural nets, ICA]
(cid:15) LGIs for Stiefel, Krogstad, Celledoni + O
ApplicationsofLiegroupintegratorsandexponentialschemes –p.9/60