Martin McVay - Spirals, One-dimensional Waves and Other Non-homogeneous Attractors in a Two-dimensional Coupled Map Lattice


  • The Thesis (Zip Archive of LaTeX and Postscript Image Files - run LaTeX three times after unzipping)

  • Aims and Objectives

  • A Guide to this Thesis

  • Animated GIFs

  • The 2x2 Lattice Bifurcations


    Aims and Objectives

    We begin with a coupled map lattice designed to model insect population dynamics which, while undergoing diverging oscillations for zero coupling and almost all initial conditions, displays long-term non-homogeneous behaviour on the lattice for small positive values of the coupling parameter, notably spiral waves and spatial chaos. We would like to understand how this can be.

    However, the biological model's local map is highly asymmetric and defined on an infinite domain, so to understand the mechanisms at work we spend the greater part of the thesis investigating a more mathematically friendly system which is defined on the closed unit disc and commutes with rotations, while at the same time bearing more than a passing resemblance to the biological model and displaying the same sorts of behaviour when coupled in a lattice.

    We prove the existence of persistent one-dimensional waves and regular spiral configurations for the friendly equations, use circulant matrices results to prove stability of one-dimensional waves for small non-zero coupling and provide numerical evidence for the stability of the spirals. We also give an in-depth account of the special attractors observed on the 2x2 lattice and the bifurcations responsible for their creation, which, while not strictly relevant to the main problem, form an integral part of our system's full dynamics.

    By the end we will have completed a thorough investigation into our specially constructed `friendly' dynamical system, and while we are unable to carry the results across directly to the biological model there seems little doubt that the processes at work in the one must be analogous to the less symmetric processes taking place in the other. Our friendly set of equations does not appear to be at all limited in the richness of its behaviour by its simplicity - rather it seems that asymmetries are an uunnecessary complication when trying to understand such observed spatial dynamics.

    Lattice simulations for a range of variations on the friendly local map display similar patterns, suggesting that what we may be looking at is just one representative of a much larger family of dynamical systems, and we hope that the properties of the system studied here will be useful in understanding the dynamics of its relatives.


    A Guide to this Thesis

    Chapter 1 - We define two dynamical systems, both two-dimensional coupled map lattices. One describes an established biological model, the other a simpler, friendlier system which displays similar behaviour in computer simulations. Both systems can be decomposed into the local map and a coupling matrix which acts on the vector of ordered lattice variables. The coupling map is common to both systems, and usually takes the form of a block circulant matrix with circulant blocks. The properties of such matrices, which will be critical to many of the proofs in this thesis regarding the friendly system, are described, as are the orbits of the two local maps.

    Chapter 2 - We summarise the behaviour of the two lattices for a range of parameter values and lattice sizes, and note the similarities. This is where we leave the difficult biological model equations. The rest of the thesis will take the form of an investigation into our simpler system. The reader is left to judge how much of what we learn can be seen as relevant to the biological model.

    Chapter 3 - Restricting attention to the friendlier of the two systems, we begin by investigating the rich variety of behaviours observed in the 2x2 lattice. For small coupling orbits accumulate on a torus, and we display numerous trajectory plots and infer from them the bifurcations which lead to the formation of various different kinds of attractor, among them fixed points which do not correspond to a homogeneous lattice, and closed orbits which undergo length-doubling bifurcations as the torus breaks up.

    Chapter 4 - We then move on to larger lattices and make a definition of what it means to have a spiral centre present. We describe a number of symmetries preserved by the dynamics and the invariant linear subspaces which they generate. We also look at the series of bifurcations at the origin which take place as the coupling parameter is varied, which we will later associate with the creation of at least partial surfaces of points whose distances from the origin are unchanged by a single iteration of the dynamics, and which include one-dimensional lattice waves and regularly distributed spiral structures.

    Chapter 5 - We limit our attention to one-dimensional patterns on the lattice for non-zero coupling. Using properties of circulant matrices, we prove that the homogeneous lattice fixed points are always stable to perturbations, the checkerboard and one-dimensional waves of spatial periods 2, 3 and 4 always unstable in at least one direction (the 4 case provided coupling is small), and waves of spatial period greater than or equal to 5, and equal to the lattice size in the direction of travel, stable for small coupling and small values of the local map parameter.

    Chapter 6 - Finally we return to our investigation of regular spirals on an even-sized lattice and argue that in suitable parameter ranges such configurations exist as fixed points (up to uniform rotation of all lattice sites) close to their corresponding creation bifurcations at the origin. We show that regular configurations of four spiral centres on 2x2 and 4x4 lattices are unstable for small coupling, but numerical computations of eigenvalues for the 6x6 and 8x8 cases indicate that larger spirals are in fact stable in certain regions of parameter space.

    Chapter 7 - We take stock of what we have learnt, say a few additional words on the biological model and give a wish-list for future research on the more friendly system. We also mention other works which expand more fully on the topics of coupled map lattices, the Nicholson-Bailey system and spiral waves in continuous media.

    Appendices - Here we work through the algorithm for inverting the local map of the more friendly system, should it be required for future research (perhaps to compute stable manifolds of saddle points), provide an alternative proof of the stability of one-dimensional lattice waves of length 8, and give a rough idea of the types of dominant behaviour observed on square lattices of various sizes besides the 30x30 case discussed in the main text. Finally we display a number of lattices generated by variations on the local map which we have been studying, and after the bibliography the reader will find a larger version of the 2x2 lattice bifurcation diagram discussed in Chapter 3.


    Animated GIFs

    We look at the behaviour of a two-dimensional coupled map lattice where each site is associated with a complex number lying in the closed unit disc. We vary two parameters, δ, which affects the local map, and ε, the diffusive coupling strength. Initial conditions are random unless otherwise stated. The local map is described in detail in Chapter 1.

    Local state space (the unit disc):

    disc - Martin McVay 2001

  • `Spatial chaos' - 100x100 lattice, δ=0.5, ε=0.01, time=1100, 1200, ..., 2000.

    chaso01 - Martin McVay 2001 chaso02 - Martin McVay 2001 chaso03 - Martin McVay 2001 chaso04 - Martin McVay 2001 chaso05 - Martin McVay 2001 chaso06 - Martin McVay 2001 chaso07 - Martin McVay 2001 chaso08 - Martin McVay 2001 chaso09 - Martin McVay 2001 chaso10 - Martin McVay 2001

  • Spiral wave formation - 100x100 lattice, δ=0.1, ε=0.1, time=100, 200, ..., 5000.

    spiral1a - Martin McVay 2001 spiral1b - Martin McVay 2001 spiral1c - Martin McVay 2001 spiral1d - Martin McVay 2001 spiral1e - Martin McVay 2001 spiral1f - Martin McVay 2001 spiral1g - Martin McVay 2001 spiral1h - Martin McVay 2001 spiral1i - Martin McVay 2001 spiral1j - Martin McVay 2001

  • Periodic spiral waves - 100x100 lattice, δ=0.1, ε=0.1, time=15100, 15200, ..., 16500. (Of course, the period isn't really 1500, so you may detect a slight discontinuity.)

    spiral2a - Martin McVay 2001 spiral2b - Martin McVay 2001 spiral2c - Martin McVay 2001 spiral2d - Martin McVay 2001 spiral2e - Martin McVay 2001 spiral2f - Martin McVay 2001 spiral2g - Martin McVay 2001 spiral2h - Martin McVay 2001 spiral2i - Martin McVay 2001 spiral2j - Martin McVay 2001

  • Regularly distributed spiral waves - 100x100 lattice, δ=0.1, ε=0.1, special initial conditions described in Chapter 4, step size 100.

    spiral3a - Martin McVay 2001 spiral3b - Martin McVay 2001 spiral3c - Martin McVay 2001 spiral3d - Martin McVay 2001 spiral3e - Martin McVay 2001 spiral3f - Martin McVay 2001 spiral3g - Martin McVay 2001 spiral3h - Martin McVay 2001 spiral3i - Martin McVay 2001 spiral3j - Martin McVay 2001

  • One-dimensional waves - a typical one-dimensional wave of spatial period equal to the lattice size.

    wave01 - Martin McVay 2001 wave02 - Martin McVay 2001 wave03 - Martin McVay 2001 wave04 - Martin McVay 2001 wave05 - Martin McVay 2001 wave06 - Martin McVay 2001 wave07 - Martin McVay 2001 wave08 - Martin McVay 2001 wave09 - Martin McVay 2001 wave10 - Martin McVay 2001

    The random lattice simulations were performed on the C program thing.c which outputs an ASCII representation to both the screen and a plain text file, which may then be fed into the QBasic program graphic.bas to generate an image that may be captured by appropriate screen capture software and converted into animated GIF format. The regularly distributed spirals were created using regular.c.

    The C program's screen output shows, in addition to each site's current quadrant, the location of every clockwise and anticlockwise spiral centre, and the modulus at each site. Local minima of the moduli coincide with spiral centres, as do the points on the lattice where all four colours come together.

    A 100x100 lattice bitmap can be generated directly from Unix by using the program bitmap.c, provided the bare bones bitmap thing.bmp is replaced before each iteration.

    For screen output only, take any of these C programs and delete all lines containing the fprintf command.


    The 2x2 Lattice Bifurcations

    Click here and here to see the special bifurcation sequences described in Chapter 3 for the 2x2 lattice.

    sequence1a - Martin McVay 2001 sequence1b - Martin McVay 2001 sequence1c - Martin McVay 2001 sequence1d - Martin McVay 2001 sequence1e - Martin McVay 2001 sequence1f - Martin McVay 2001

    sequence2a - Martin McVay 2001 sequence2b - Martin McVay 2001 sequence2c - Martin McVay 2001 sequence2d - Martin McVay 2001 sequence2e - Martin McVay 2001 sequence2f - Martin McVay 2001 sequence2g - Martin McVay 2001 sequence2h - Martin McVay 2001

    If you have any questions, please e-mail martinmcvay@yahoo.co.uk. Thank you!