a b s t r a c t In 1961, Paul Erdös posed the question whether abelian squares can be avoided in arbitrarily long words over a finite alphabet. An abelian square is a non-empty word uv, where u and v are permutations (anagrams) of each other. The case of the four letter alphabet Σ 4 = {a, b, c, d} turned out to be the most challenging and remained open until 1992 when the author presented an abelian square-free (a-2-free) endomorphism g 85 of Σ * 4 . The size of this g 85 , i.e., |g 85 (abcd)|, is equal to 4×85 (uniform modulus). Until recently, all known methods for constructing arbitrarily long a-2-free words on Σ 4 have been based on the structure of g 85 and on the endomorphism g 98 of Σ * 4 found in 2002. In this paper, a great many new a-2-free endomorphisms of Σ * 4 are reported. The sizes of these endomorphisms range from 4 × 102 to 4 × 115. Importantly, twelve of the new a-2-free endomorphisms, of modulus m = 109, can be used to construct an a-2-free (commutatively functional) substitution σ 109 of Σ * 4 with 12 image words for each letter. The properties of σ 109 lead to a considerable improvement for the lower bound of the exponential growth of c n , i.e., of the number of a-2-free words over 4 letters of length n. It is obtained that c n > β −50 β n with β = 12 1/m 1.02306. Originally, in 1998, Carpi established the exponential growth of c n by showing that c n > β −t β n with β = 2 19/t = 2 19/(85 3 −85)1.000021, where t = 85 3 − 85 is the modulus of the substitution that he constructs starting from g 85 .