A090822 Gijswijt's sequence: a(1) = 1; for n>1, a(n) = largest integer k such that the word a(1)a(2)...a(n-1) is of the form xy^k for words x and y (where y has positive length), i.e., the maximal number of repeating blocks at the end of the sequence so far.

1, 1, 2, 1, 1, 2, 2, 2, 3, 1, 1, 2, 1, 1, 2, 2, 2, 3, 2, 1, 1, 2, 1, 1, 2, 2, 2, 3, 1, 1, 2, 1, 1, 2, 2, 2, 3, 2, 2, 2, 3, 2, 2, 2, 3, 3, 2, 1, 1, 2, 1, 1, 2, 2, 2, 3, 1, 1, 2, 1, 1, 2, 2, 2, 3, 2, 1, 1, 2, 1, 1, 2, 2, 2, 3, 1, 1, 2, 1, 1, 2, 2, 2, 3, 2, 2, 2, 3, 2, 2, 2, 3, 3, 2, 2, 2, 3, 2, 1

Offset: 1

Author

Dion Gijswijt, Feb 27 2004

Keywords

• nonn
• nice

Comments

Here xy^k means the concatenation of the words x and k copies of y.

The name "Gijswijt's sequence" is due to N. J. A. Sloane, not the author!

Fix n and suppose a(n) = k. Let len_y(n) = length of shortest y for this k and let len_x = n-1 - k*len_y(n) = corresponding length of x. A091407 and A091408 give len_y and len_x. For the subsequence when len_x = 0 see A091410 and A091411.

The first 4 occurs at a(220) (see A091409).

The first 5 appears around term 10^(10^23).

We believe that for all N >= 6, the first time N appears is at about position 2^(2^(3^(4^(5^...^(N-1))))). - N. J. A. Sloane and Allan Wilks, Mar 14 2004

For a similar formula, see p. 6 of Levi van de Pol article. - Levi van de Pol, Feb 06 2023

In the first 100000 terms the fraction of [1's, 2's, 3's, 4's] seems to converge, to about [.287, .530, .179, .005] respectively. - Allan Wilks, Mar 04 2004

When k=12 is reached, say, it is treated as the number 12, not as 1,2. This is not a base-dependent sequence.

Does this sequence have a finite average? Does anyone know the exact value? - Franklin T. Adams-Watters, Jan 23 2008

Answer: Given that "...the fraction of [1's, 2's, 3's, 4's] seems to converge, to about [.287, .530, .179, .005]..." that average should be the dot product of these vectors, i.e., about 1.904. - M. F. Hasler, Jan 24 2008

Second answer: The asymptotic densities of the numbers exist, and the average is the dot product. See pp. 56-59 of Levi van de Pol article. - Levi van de Pol, Feb 06 2023

Which is the first step with two consecutive 4's? Or the shortest run found so far between two 4's? - Sergio Pimentel, Oct 10 2016

Answer: The first x such that the x-th and (x+1)-th element are 4, is 255895648634818208370064452304769558261700170817472823... ...398081655524438021806620809813295008281436789493636144. See p. 55 of Levi van de Pol article. - Levi van de Pol, Feb 06 2023

References

• N. J. A. Sloane, Seven Staggering Sequences, in Homage to a Pied Puzzler, E. Pegg Jr., A. H. Schoen and T. Rodgers (editors), A. K. Peters, Wellesley, MA, 2009, pp. 93-110.

Links

• N. J. A. Sloane, Table of n, a(n) for n = 1..24000
• Andreas Abel and Andres Loeh, Haskell program
• Fokko J. van de Bult, Dion C. Gijswijt, John P. Linderman, N. J. A. Sloane and Allan R. Wilks, A Slow-Growing Sequence Defined by an Unusual Recurrence, J. Integer Sequences, Vol. 10 (2007), #07.1.2.
• Benjamin Chaffin, John P. Linderman, N. J. A. Sloane, and Allan R. Wilks, On Curling Numbers of Integer Sequences, arXiv:1212.6102 [math.CO], Dec 25 2012.
• Benjamin Chaffin, John P. Linderman, N. J. A. Sloane, and Allan R. Wilks, On Curling Numbers of Integer Sequences, Journal of Integer Sequences, Vol. 16 (2013), Article 13.4.3.
• Benjamin Chaffin and N. J. A. Sloane, The Curling Number Conjecture, preprint.
• Dion C. Gijswijt, Krulgetallen, Pythagoras, 55ste Jaargang, Nummer 3, Jan 2016, (Article in Dutch about this sequence, see pages 10-13, cover and back cover).
• Levi van de Pol, The first occurrence of a number in Gijswijt's sequence, arXiv:2209.04657 [math.CO], 2022.
• Levi van de Pol, The Growth Rate of Gijswijt's Sequence, J. Int. Seq. (2025) Vol. 28, Art. No. 25.4.6. See p. 2.
• N. J. A. Sloane, Seven Staggering Sequences.
• N. J. A. Sloane, Confessions of a Sequence Addict (AofA2017), slides of invited talk given at AofA 2017, Jun 19 2017, Princeton. Mentions this sequence.
• Index entries for sequences related to Gijswijt's sequence
• Index entries for sequences related to curling numbers

Crossrefs

Cf. A091407, A091408, A091409 (records), A091410, A091411, A091579, A091586, A091970, A093955-A093958.

A091412 gives lengths of runs. A091413 gives partial sums.

Cf. also A094004, A160766, A217206, A217207.

Generalizations: A094781, A091975, A091976, A092331-A092335.

Programs

Haskell

-- See link.

Maple

K:= proc(L)
local n,m,k,i,b;
m:= 0;
n:= nops(L);
for k from 1 do
  if k*(m+1) > n then return(m) fi;
  b:= L[-k..-1];
  for i from 1 while i*k <= n and L[-i*k .. -(i-1)*k-1] = b do od:
  m:= max(m, i-1);
od:
end proc:
A[1]:= 1:
for i from 2 to 220 do
  A[i]:= K([seq(A[j],j=1..i-1)])
od:
seq(A[i],i=1..220); # Robert Israel, Jul 02 2015

Mathematica

ClearAll[a]; reversed = {a[2]=1, a[1]=1}; blocs[len_] := Module[{bloc1, par, pos}, bloc1 = Take[reversed, len]; par = Partition[ reversed, len]; pos = Position[par, bloc_ /; bloc != bloc1, 1, 1]; If[pos == {}, Length[par], pos[[1, 1]] - 1]]; a[n_] := a[n] = Module[{an}, an = Table[{blocs[len], len}, {len, 1, Quotient[n-1, 2]}] // Sort // Last // First; PrependTo[ reversed, an]; an]; A090822 = Table[a[n], {n, 1, 99}] (* Jean-François Alcover, Aug 13 2012 *)

PARI

A090822(n,A=[])={while(#Ak||break; k=m);A=concat(A,k));A} \\ M. F. Hasler, Aug 08 2018

Python

def k(s):
    maxk = 1
    for m in range(1, len(s)+1):
        i, y, kk = 1, s[-m:], len(s)//m
        if kk <= maxk: return maxk
        while s[-(i+1)*m:-i*m] == y: i += 1
        maxk = max(maxk, i)
def aupton(terms):
    alst = [1]
    for n in range(2, terms+1):
        alst.append(k(alst))
    return alst
print(aupton(99)) # Michael S. Branicky, Mar 28 2022