1: \begin{abstract}

2: Let $\AZD$ be the Cantor space of $\ZD$-indexed configurations in a

3: finite alphabet $\sA$, and let $\shift{}$ be the $\ZD$-action of

4: shifts on $\AZD$.  A {\dfn cellular automaton} is a continuous,

5: $\shift{}$-commuting self-map $\Phi$ of $\AZD$, and a {\dfn $\Phi$-invariant

6: subshift} is a closed, $(\Phi,\shift{})$-invariant subset

7: $\gA\subset\AZD$.  Suppose $\ba\in\AZD$ is $\gA$-admissible everywhere

8: except for some small region we call a {\dfn defect}. It has been

9: empirically observed that such defects persist under iteration of

10: $\Phi$, and often propagate like `particles' which coalesce or

11: annihilate on contact.  We construct algebraic invariants

12: for these defects, which explain their persistence under $\Phi$, and

13: partly explain the outcomes of their collisions.

14: Some invariants are based on the cocycles of multidimensional

15: subshifts; others arise from the higher-dimensional

16: (co)homology/homotopy groups for subshifts, obtained by generalizing

17: the Conway-Lagarias tiling groups and the Geller-Propp fundamental

18: group.

19: 

20: {\footnotesize

21: \breath

22: 

23: \begin{tabular}{rl}

24: {\bf MSC:}& 37B50 (primary), 37B15, 37A20 (secondary)\\

25: {\bf Keywords:}& Cellular automata, subshift, cocycle,

26: cohomology, tiling group, \\

27: & defect, kink, domain boundary.

28: \end{tabular}}

29: \end{abstract}

30: