H. H. P. M. P. Cole wrote: October 27th, 2024, 6:30 am
[...]
small one-per-generation patterns which use
at most one infinite growth mechanism (most likely a wickstretcher engine). [...]
By "at most one infinite growth mechanism", do you really mean at most one unique growth mechanism (although as many instances of that mechanism as desired), or at most one engine?
As far as I can tell, "at most one" can be simplified to "one".
One question is: what is the lowest starting term X of a population sequence (X, X+1, X+2, ...)
In other words, what is the lowest X such that there exists a growing pattern with population(T) = X + T, for T = 0, 1, 2, ...?
In a two-dimensional two-state isotropic cellular automaton, X must be at least 2.
(X = 0 is impossible, because a single alive cell cannot spontaneously form in empty space.
X = 1 is impossible, because there is no way for a single alive cell to evolve into a two-bit pattern one tick later.
For example, assuming the square tiling and the Moore neighbourhood, the only possibilities for population 1 tick later are 0, 1, 4, 5, 8, 9.
Assuming the hexagonal tiling and the honeycomb neighbourhood, the only possibilities for population 1 tick later are 0, 1, 6, 7.)
In a 2D two-state isotropic CA
on the triangular tiling, there
might be some way to have X = 2, because there exist asymmetric two-bit patterns on the triangular tiling:
Code:
Select all
x = 12, y = 6, rule = //3L
2.B7.B$.A9.B$6.A3$5.B.B!
#C [[ GRID VIEWONLY ]]
In a 2D two-state isotropic CA on the square tiling, or on the hexagonal tiling, X = 2 is impossible, because every two-bit pattern remains unchanged when rotated 180 degrees, and therefore any growing pattern of this kind will necessarily grow in at least two directions at once, contradicting "one bit per tick".
Is X = 3 possible?
Is X = 3 possible with range-1 neighbourhoods (the Moore neighbourhood on the square tiling; the honeycomb neighbourhood on the hexagonal tiling)?
Is X = 2 indeed possible on the triangular tiling?