Dejean's theorem

Dejean's theorem (formerly Dejean's conjecture) is a statement about repetitions in infinite strings of symbols. It belongs to the field of combinatorics on words; it was conjectured in 1972 by Françoise Dejean and proven in 2009 by Currie and Rampersad and, independently, by Rao.[1]

Context

In the study of strings, concatenation is seen as analogous to multiplication of numbers. For instance, if s {\displaystyle s} is any string, then the concatenation s s {\displaystyle ss} of two copies of s {\displaystyle s} is called the square of s {\displaystyle s} , and denoted s 2 {\displaystyle s^{2}} . This exponential notation may also be extended to fractional powers: if s {\displaystyle s} has length {\displaystyle \ell } , and e {\displaystyle e} is a non-negative rational number of the form n / {\displaystyle n/\ell } , then s e {\displaystyle s^{e}} denotes the string formed by the first n {\displaystyle n} characters of the infinite repetition s s s s s {\displaystyle sssss\dots } .[1]

A square-free word is a string that does not contain any square as a substring. In particular, it avoids repeating the same symbol consecutively, repeating the same pair of symbols, etc. Axel Thue showed that there exists an infinite square-free word using a three-symbol alphabet, the sequence of differences between consecutive elements of the Thue–Morse sequence. However, it is not possible for an infinite two-symbol word (or even a two-symbol word of length greater than three) to be square-free.[1]

For alphabets of two symbols, however, there do exist infinite cube-free words, words with no substring of the form s s s {\displaystyle sss} . One such example is the Thue–Morse sequence itself; another is the Kolakoski sequence. More strongly, the Thue–Morse sequence contains no substring that is a power strictly greater than two.[1]

In 1972, Dejean investigated the problem of determining, for each possible alphabet size, the threshold between exponents e {\displaystyle e} for which there exists an infinite e {\displaystyle e} -power-free word, and the exponents for which no such word exists. The problem was solved for two-symbol alphabets by the Thue–Morse sequence, and Dejean solved it as well for three-symbol alphabets. She conjectured a precise formula for the threshold exponent for every larger alphabet size;[2] this formula is Dejean's conjecture, now a theorem.[1]

Statement

Let k {\displaystyle k} be the number of symbols in an alphabet. For every k {\displaystyle k} , define RT ( k ) {\displaystyle \operatorname {RT} (k)} , the repeat threshold, to be the infimum of exponents e {\displaystyle e} such that there exists an infinite e {\displaystyle e} -power-free word on a k {\displaystyle k} -symbol alphabet. Thus, for instance, the Thue–Morse sequence shows that RT ( 2 ) = 2 {\displaystyle \operatorname {RT} (2)=2} , and an argument based on the Lovász local lemma can be used to show that RT ( k ) {\displaystyle \operatorname {RT} (k)} is finite for all k {\displaystyle k} .[1]

Then Dejean's conjecture is that the repeat threshold can be calculated by the simple formula[1][2]

RT ( k ) = k k 1 {\displaystyle \operatorname {RT} (k)={\frac {k}{k-1}}}

except in two exceptional cases:

RT ( 3 ) = 7 4 {\displaystyle \operatorname {RT} (3)={\frac {7}{4}}}

and

RT ( 4 ) = 7 5 . {\displaystyle \operatorname {RT} (4)={\frac {7}{5}}.}

Progress and proof

Dejean herself proved the conjecture for k = 3 {\displaystyle k=3} .[2] The case k = 4 {\displaystyle k=4} was proven by Jean-Jacques Pansiot in 1984.[3] The next progress was by Moulin Ollagnier in 1992, who proved the conjecture for all alphabet sizes up to k 11 {\displaystyle k\leq 11} .[4] This analysis was extended up to k 14 {\displaystyle k\leq 14} in 2007 by Mohammad-Noori and Currie.[5]

In the other direction, also in 2007, Arturo Carpi showed the conjecture to be true for large alphabets, with k 33 {\displaystyle k\geq 33} .[6] This reduced the problem to a finite number of remaining cases, which were solved in 2009 and published in 2011 by Currie and Rampersad[7] and independently by Rao.[8]

Dejean words

An infinite string that meets Dejean's formula (having no repetitions of exponent above the repetition threshold) is called a Dejean word. Thus, for instance, the Thue–Morse sequence is a Dejean word.

References

  1. ^ a b c d e f g Rampersad, Narad; Shallit, Jeffrey (2016), "Repetitions in words", Combinatorics, words and symbolic dynamics, Encyclopedia Math. Appl., vol. 159, Cambridge Univ. Press, Cambridge, pp. 101–150, MR 3525483
  2. ^ a b c Dejean, Françoise (1972), "Sur un théorème de Thue", Journal of Combinatorial Theory, Series A, 13: 90–99, doi:10.1016/0097-3165(72)90011-8, MR 0300959
  3. ^ Pansiot, Jean-Jacques (1984), "À propos d'une conjecture de F. Dejean sur les répétitions dans les mots", Discrete Applied Mathematics, 7 (3): 297–311, doi:10.1016/0166-218x(84)90006-4, MR 0736893
  4. ^ Moulin Ollagnier, Jean (1992), "Proof of Dejean's conjecture for alphabets with 5, 6, 7, 8, 9, 10 and 11 letters", Theoretical Computer Science, 95 (2): 187–205, doi:10.1016/0304-3975(92)90264-G, MR 1156042
  5. ^ Mohammad-Noori, M.; Currie, James D. (2007), "Dejean's conjecture and Sturmian words", European Journal of Combinatorics, 28 (3): 876–890, doi:10.1016/j.ejc.2005.11.005, MR 2300768
  6. ^ Carpi, Arturo (2007), "On Dejean's conjecture over large alphabets", Theoretical Computer Science, 385 (1–3): 137–151, doi:10.1016/j.tcs.2007.06.001, MR 2356248
  7. ^ Currie, James; Rampersad, Narad (2011), "A proof of Dejean's conjecture", Mathematics of Computation, 80 (274): 1063–1070, arXiv:0905.1129, doi:10.1090/S0025-5718-2010-02407-X, MR 2772111
  8. ^ Rao, Michaël (2011), "Last cases of Dejean's conjecture", Theoretical Computer Science, 412 (27): 3010–3018, doi:10.1016/j.tcs.2010.06.020, MR 2830264