Garden of Eden
A Garden of Eden is a pattern that has no parents and thus can only occur in generation 0. The term was first used in connection with cellular automata by John W. Tukey, many years before Conway's Game of Life was conceived. It was known from the start that Gardens of Eden exist in Life because of a theorem by Edward Moore that guarantees their existence in a wide class of cellular automata.
Garden of Eden theorem
The Garden of Eden theorem was proved by Edward Moore and John Myhill pre-1970 and shows that a wide class of cellular automata must contain Garden of Eden patterns. Of particular interest is that Conway's Game of Life falls into this class, and thus Gardens of Eden were known to exist right from the day it was conceived.
Statement of the theorem
A finite pattern (or finite configuration) is a pattern with a finite number of bits. A cellular automaton is said to be injective over finite patterns if no two distinct finite patterns map into the same finite pattern. It is said to be surjective if every pattern is mapped to by some other pattern. Thus, by definition a cellular automaton contains Gardens of Eden if and only if it is not surjective.
The Garden of Eden theorem states that the class of surjective cellular automata and those which are injective over finite configurations coincide. In other words, a cellular automaton has a Garden of Eden if and only if it has two different finite configurations that evolve into the same configuration in one step.
As a corollary, every injective cellular automaton (i.e., one with one-to-one global mapping for both finite and infinite patterns) is surjective and hence bijective. However, surjective cellular automata do not need to be injective over infinite patterns (and thus need not be injective in general).
Application to Conway's Game of Life
The theorem applies to Conway's Game of Life because it is easy to find two different finite patterns that are mapped into the same configuration. The configuration in which every cell is dead, and the one in which exactly one cell is alive both lead to the one in which every cell is dead. The Garden of Eden theorem then implies that there must exist a Garden of Eden pattern.
A related concept to Gardens of Eden is that of orphans, which are finite patterns that can not occur as part of the evolution of another pattern. That is, they are Gardens of Eden that can be extended in any way to form other Gardens of Eden.
Several Gardens of Eden and orphans have been constructed, the first by Roger Banks et al. at MIT in 1971. It had a bounding box of size 33 × 9 and had 226 cells. Jean Hardouin-Duparc found the second and third orphans by computer search in 1973, which had bounding boxes of size 122 × 6 and 117 × 6. His goal was to find Gardens of Eden with minimal height, and it is believed that no Gardens of Eden exist with height less than 5.
Many smaller Gardens of Eden have been found in more recent years. Garden of Eden 2 was found by Achim Flammenkamp in 1991, contained 143 cells, and had a bounding box of size 14 × 14. Garden of Eden 3 was found by Achim Flammenkamp in 2004, contained 81 cells, and had a bounding box of size 13 × 12. The smallest known Garden of Eden for about five years was Garden of Eden 4, which was also found by Achim Flammenkamp in 2004. It contained 72 cells and had a bounding box of size 12 × 11. The current smallest-known example is Garden of Eden 6, which contains 56 live cells and a bounding box of 10 × 10. It was found on December 14, 2011. Computer searches have revealed that there are no Gardens of Eden contained within a 6 × 6 bounding box.
- Achim Flammenkamp (November 7, 2008). "Garden of Eden / Orphan". Retrieved on February 13, 2009.
- Achim Flammenkamp (November 7, 2008). "Garden of Eden / Orphan". Retrieved on February 14, 2009.
- Adam P. Goucher (January 14, 2012). "Gardens of Eden". Retrieved on June 18, 2012.
- Garden of Eden at the Life Lexicon