|
Normal Forms and Unfoldings for Local Dynamical
Systems
J. Murdock, Springer Monographs in Mathematics, Springer Verlag 2003,
494 p., price € 79,95, ISBN 0-387-95464-3. |
Reviewer: Henk Broer, University of Groningen, The
Netherlands. |
The theory of local dyamical systems studies neighborhoods of a given
equilibrium point, in particular the dynamical behavior that is
generically possible. In the book under review the dynamical systems
are autonomous (or occasionally time-periodic) systems of ordinary
differential equations. At the equilibrium point the first concern is
with the linear part, but next also the rest of the Taylor series
comes into view as well as unfoldings or deformations of these by
parameters. The interest then is with the dynamical behavior as a
function of these parameters, in particular with transitions or
bifurcations between different types of behavior. The number of
parameters needed to see a given type of equilibrium in a persistent
way, is called the codimension of the `singularity'. Usually models
are considered by truncating the Taylor series at a suitable
order. The question then is what properties of the truncated family of
systems are persistent. Such properties concern the bifurcation of
other equilibria and periodic solutions, their stable and unstable
manifold, homo- or heteroclinic connections, etc. The persistence
problem for families of systems can be restricted to a given universe,
like to the world of Hamiltonian, of volume preserving, of equivariant
or of reversible systems.
One major tool of this research is the theory of normal forms,
which at least goes back to Poincaré. It deals with the
adjustment of the Taylor series at the equilibrium by changes of
coordinates. The adjustment aims to be a simplification, which often
amounts to the elimination of as many as possible terms in the
series. Normal form theory in itself already covers a large area of
research, with historical contributors like Birkhoff, Sternberg,
Bruno, Gustafson, Takens and many others. The techniques at equilibria
largely translate to periodic and quasi-periodic orbits and also it
turns out that the preservation of a symplectic or volume form,
invariance (or equivariance) under a symmetry group can be preserved
by using a Lie-formalism. In the latter case also reversing of
symmetries can be taken into account. Moreover, in this same setting
the local normal form at a fixed point of a diffeomorphism (Takens
approximation by a flow) can be dealt with. This scope is what I
expected when seeing the general title of this almost 500 page work.
The book under review however does not address these general
issues, but quickly narrows down to four major normal form `styles',
among which the semisimple style, the `Elphick-Iooss' inner product
style and the `Cushman-Sanders' sl(2) style. All styles
refer to different solutions of the problem to characterize
`complementary' subspaces of homogeneously polynomial vector fields
that can not be eliminated by changes of coordinates. The treatment
has a definite algorithmic and algebraic nature, aiming at the
computation of normal forms in concrete examples with help of
computer-algebra packages. Background for this is the theory of
representations and invariants. An appendix of more than 60 pages
provides an algebraic background for this.
To my knowledge the monograph under review is the first succesful
attempt to deal with the `Elphick-Iooss' inner product style and the
`Cushman-Sanders' sl(2) style at a larger scale. The emphasis
on the computability of normal forms by computer-algebraic tools is
useful for applications in concrete examples. As the author has it in
the preface, the book ``... does not require any specific knowledge
other than the basic theories of advanced calculus, differential
equations and linear algebra.'' Moreover, the book is meant to be
self-contained. From this on the one hand a certain emphasis on the
algebraic aspects can be understood. On the other hand, however, my
general feeling is that the whole style is dominated too much by
algorithms and algebra.
To explain this further, let me elaborate briefly on the most
common normal form style, the semisimple style. Here the linear part
S of the vector field at the equilibrium is semisimple and
the corresponding ad-operator £ =
adS = [S, - ], defined on
the spaces of homogeneous polynomial
vector fields, is semisimple as well. In this case the image
im £ consists of terms that can be eliminated by
changes of coordinates. A natural choice for the `complementary'
space then is the kernel ker £. The elements of
this space are vector fields that commute with the linear part, which
gives rise to a toroidal symmetry (in the center manifold). This
includes the case of the 1-torus or circle. As a consequence, any
truncated normal form polynomial can be reduced in dimension by
dividing out this symmetry. This kind of reduction is important for
many applications, in particular for bifurcation theory, as can be
seen from a number of textbooks. In the Hamiltonian setting this
method is closely related to classical reductions in integrable
systems.
Moreover, when the linear part of the vector field at the
equilibrium point has a Jordan splitting A = S + N, with
S semisimple and N nilpotent, one still can take
care that the remaining terms all commute with S. This again
makes a toroidal reduction possible. However, now the nilpotent term
N gives the possibility of even further restricting the
complementary space. The option of toroidal reduction is one of the
reasons why many users of normal form theory restrict to the
semisimple style. In the nilpotent case the semisimple part vanishes
and the above considerations break down. This is one good reason to
resort to one of the other normal form styles.
In the present book, these more general issues of normal form
theory are not dealt with in any substantial way: there is no
mentioning at all of this toroidal symmetry. Indeed, this would have
required a more generally mathematical entrance to the theory, with a
good eye for applications. As a further illustration of this let me
mention that, when introducing universal unfoldings of matrices the
author uses a computation with series and not the more geometric
theory of `matrices depending on parameters' by Arnol'd. Terms as
`versal' and `transversal' do not occur in the book. A similar remark
holds for differential geometric aspects of the Lie theory approach
and general remarks on normal form approaches that automatically
preserve given structures, again see above. This narrowness of scope
also is reflected in the bibliography, where I found many items
missing. Among those I like to mention F. Takens, Singularities of
vector fields, Publ Math IHES 43 1974, which is
fundamental both for the semisimple and nilpotent case.
In summary, I like to raise the question for what readership the
book was written. The author has two disclaimers in the preface saying
that the text is not an `introduction' to the subject and that
``... it is is unlikely that a student will find this book to be
understandable without some previous exposure to the subject of
dynamical systems, ....'' I would agree to both statements. In
addition to this, I have doubts that any reader, not already
acquainted with an appropriate amount of algebra, would find the text
accessible, this notwithstanding the appendices. In that respect the
text really seems to address experts in the field. As said before, the
text succesfully addresses computer-algebraic aspects of certain
normal form computations that are useful for applications in concrete
examples.
Nevertheless, assuming a wider background in both algebra and
(differential) geometry, could have led to a different book with a
broader scope as sketched before. Such a strategy would have done more
justice to the present title and size of the book, which then also
might have been more interesting for an even larger group of experts.