Surjectivity and injectivity [of cellular automaton maps]

One can think of a cellular automaton rule as a mapping (endomorphism) from the space of possible states of the cellular automaton to itself. (See page 869.) Usually this mapping is contractive, so that not all the states which appear as input to the mapping can also appear as output. But in some cases, the mapping is surjective or onto, meaning that any state which appears as input can also appear as output. Among k=2, r=1 elementary cellular automata it turns out that this happens precisely for those 30 rules that are additive with respect to at least the first or last position on which they depend (see pages 601 and 1087); this includes both rules 90 and 150 and rules 30 and 45. With k=2, r=2 there are a total of 4,294,967,296 possible rules. Out of these 141,884 are onto—and 11,388 of these turn out not to be additive with respect to any position. The easiest way to test whether a particular rule is onto seems to be essentially just to construct the minimal finite automaton discussed on page 957. The onto k=2, r=2 rules were found in 1961 in a computer study by Gustav Hedlund and others; they later apparently provided input in the design of S-boxes for DES cryptography (see page 1085).

Even when a cellular automaton mapping is surjective, it is still often many-to-one, in the sense that several input states can yield the same output state. (Thus for example additive rules such as 90 and 150, as well as one-sided additive rules such as 30 and 45 are always 4-to-1.) But some surjective rules also have the property of being injective, so that different input states always yield different output states. And in such a case the cellular automaton mapping is one-to-one or bijective (an automorphism). This is equivalent to saying that the rule is reversible, as discussed on page 1017.

(In 2D such properties are in general undecidable; see page 1138.)