Mark Dow

Geek art

Simple recursive systems and fractal patterns

Fibonacci word arithmetic subsequences

    Arithmetic sub-sequences of the morphism { aa --> a b, b --> a }, which forms the Fibonacci word, OEIS A003849. The Rabbit sequence is a closely related sequence.

Arithmetic subsequences of Fibonacci word, F( c*n - d ) 
Fibonacci word - Fibonacci number self-similarity

References

Arithmetic subsequences of Fibonacci words, F( c*n - d ) 

The Fibonacci word F(n), arithemetic subsequences (with black and white as symbols a and b) and their discrete Fourier transform ampitude spectra:


F( n )
diff( F(n) )
Fibonacci( n ) Fibonacci( n ) amplitude spectrum, link to
FSpectrum_1_0_65536.fig (MATLAB figure with full amplitude spectrum)

F( c*n - d )

Subsequence (with black and white as symbols a and b):

Amplitude spectrum of subsequence, link to:

F( 2*n - 0 )
F(  Fm(3)*n  )
diff( F(2*n) )
Fibonacci( 2* n - 0 ) Fibonacci( 2*n - 0 ) amplitude spectrum, link to
F( 2*n - 1 ) Fibonacci( 2* n - 1 ) FSpectrum_2_0_65536.fig (MATLAB figure with full amplitude spectrum)
Why are F( 2*n - 1 ) and F( 2*n - 1 ) near complements? What is their difference?
F( 3*n - 0 )
F(  Fm(4)*n  )
diff( F(3*n) )
Fibonacci( 3*n - 0 ) Fibonacci( 3*n - 0 ) amplitude spectrum, link to
F( 3*n - 1 ) Fibonacci( 3*n - 1 ) FSpectrum_3_0_65536.fig (MATLAB figure with full amplitude spectrum)
F( 3*n - 2 ) Fibonacci( 3*n - 2 )

F( 4*n - 0 )
diff( F(4*n) )
Fibonacci( 4*n - 0 ) Fibonacci( 4*n - 0 ) amplitude spectrum, link to
F( 4*n - 1 ) Fibonacci( 4*n - 1 ) FSpectrum_4_0_65536.fig (MATLAB figure with full data set)
F( 4*n - 2 ) Fibonacci( 4*n - 2 )
F( 4*n - 3 ) Fibonacci( 4*n - 3 )
F( 5*n - 0 )
F(  Fm(5)*n  ) diff( F(5*n) )
Fibonacci( 5*n - 0 ) Fibonacci( 5*n - 0 ) amplitude spectrum, link to
F( 5*n - 1 ) Fibonacci( 5*n - 1 ) FSpectrum_5_0_65536.fig (MATLAB figure with full amplitude spectrum)
F( 5*n - 2 ) Fibonacci( 5*n - 2 )
F( 5*n - 3 ) Fibonacci( 5*n - 3 )
F( 5*n - 4 ) Fibonacci( 5*n - 4 )
F( 6*n - 0 )
diff( F(6*n) )
Fibonacci( 6*n - 0 ) Fibonacci( 6*n - 0 ) amplitude spectrum, link to
F( 6*n - 1 ) Fibonacci( 6*n - 1 ) FSpectrum_6_0_65536.fig (MATLAB figure with full amplitude spectrum)
F( 6*n - 2 ) Fibonacci( 6*n - 2 )
F( 6*n - 3 ) Fibonacci( 6*n - 3 )
F( 6*n - 4 ) Fibonacci( 6*n - 4 )
F( 6*n - 5 ) Fibonacci( 6*n - 5 )
F( 7*n - 0 )
diff( F(7*n) )
Fibonacci( 7*n - 0 )
F( 8*n - 0 )
F(  Fm(6)*n  )
diff( F(8*n) )
Fibonacci( 8*n - 0 )
F( 9*n - 0 ) Fibonacci( 9*n - 0 )
F( 10*n - 0 ) Fibonacci( 10*n - 0 )
F( 11*n - 0 ) Fibonacci( 11*n - 0 )
F( 12*n - 0 ) Fibonacci( 12*n - 0 )
F( 13*n - 0 )
F(  Fm(7)*n  )
diff( F(13*n) )
Fibonacci( 13*n - 0 )
F( 14*n - 0 )
diff( F(14*n) )
Fibonacci( 14*n - 0 )
F( 15*n - 0 )
diff( F(15*n) )
Fibonacci( 15*n - 0 )
F( 16*n - 0 )
diff( F(16*n) )
Fibonacci( 16*n - 0 )
F( 17*n - 0 )
diff( F(17*n) )
Fibonacci( 17*n - 0 )
F( 18*n - 0 )
diff( F(18*n) )
Fibonacci( 18*n - 0 )
F( 19*n - 0 )
diff( F(19*n) )
Fibonacci( 19*n - 0 )
F( 20*n - 0 )
diff( F(20*n) )
Fibonacci( 20*n - 0 )
F( 21*n - 0 )
F(  Fm(8)*n  )
diff( F(21*n) )
Fibonacci( 21*n - 0 )
F( 22*n - 0 )
diff( F(21*n) )
Fibonacci( 22*n - 0 )
F( 23*n - 0 )
diff( F(21*n) )
Fibonacci( 23*n - 0 )
F( 24*n - 0 )
diff( F(21*n) )
Fibonacci( 24*n - 0 )
F( 34*n - 0 )
F(  Fm(9)*n  )
diff( F(21*n) )
Fibonacci( 34*n - 0 )
F( 55*n - 0 )
F(  Fm(10)*n  )
diff( F(21*n) )
Fibonacci( 55*n - 0 )
F( 89*n - 0 )
F(  Fm(11)*n  )
diff( F(21*n) )
Fibonacci( 89*n - 0 )
F( 144*n - 0 )
F(  Fm(12)*n  )
diff( F(21*n) )
Fibonacci( 55*n - 0 )


These figures and spectra were generated with the MATLAB progams Fibonacci_arithmetic_subsequence.m and Fibonacci_arithmetic_subsequence_set.m. See file headers for details.

Fibonacci word - Fibonacci number self-similarity

    It has been noted that the subsequences F( Fm*n ), where Fm is the mth Fibonacci number, are composed of long runs of with sequentially alternating lengths that get longer with Fm. This is easy to see in the diagrams above. Here several F( Fm*n ) sequences are stacked, each with a height Fm of the mth one. A self-similar pattern is clear. Coloring of contiguous regions (right) shows how this pattern relates to the original sequence F( n ), along the top row. This is very similar to the Thue-Morse self-similarity and its relationship to the relflected gray code.
Fibonacci word number self-similarity, link to Fibonacci word number self-similarity color, link to
Each row is an arithmetic sub-sequence of the Fibonacci word,  F( Fm*n ), where Fm is the mth Fibonacci number. The height of each set of rows is Fm. The top row is the beginning of the infinite Fibonacci word F( n ). Colored by region. The colored regions are related by a scale transformation related to the golden ratio, in particular the ratio of sequential Fibonacci numbers, Fm(m+1)/Fm(m).

    The even Fibonacci number sub-sequences ( F( Fm*n ) with m even ) are all self-similar, and the odd Fibonacci number sub-sequences ( F( Fm*n ) with m odd ) are all self-similar.

    What are the lengths, L, of the runs? It looks like the lengths grow about linearly with Fm and the first two or three runs have an additive equality:

L(n)(1) + L(n)(2)  + L(n)(3) = L(n+1)(1) + L(n+1)(2)    with n odd
L(n)(1) + L(n)(2)            = L(n+1)(1)                with n even





    Another approximate self-similarity has to do with just one of the symbols of the Fibonacci word, the second to occur:
       1 2 3 4 5 6 7 8 9 ... (index n)
F(n) = a b a a b a b a a ...
The second symbol, b, occurs at indices:
( n | F(n)=b ) = 2, 5, 7, 10, 13, 15, 18, 20, 23, 26, 28, 31, 34, 36, 39,  41, 44, 47, 49, 52, 54, 57, 60, 62, 65... (OEIS A001950 )

    If the indices n at which F(n) = b are multiplied by Fm, the sequences formed by the sequence values at these indices are correlated with the sequence itself:
F( Fm(m+2) ) * (n | F(n)=b) ) ~= F( Fm(m) * n )

e.g.:


F( 3*2 ), F( 3*5 ), F( 3*7 ), ...   ~= F( 1*n )
F( 5*2 ), F( 5*5 ), F( 5*7 ), ...   ~= F( 2*n )
F( 8*2 ), F( 8*5 ), F( 8*7 ), ...   ~= F( 3*n )
F(13*2 ), F(13*5 ), F(13*7 ), ...   ~= F( 5*n )
...

At right are these sequences in rows, with row heights Fm(m+2). It is nearly identical to the "Fibonacci word - Fibonacci number self-similarity" above. The occasional differences are shown below, with differences shaded red.

Differences between these sequences and F( Fm ) apparently only occur at boundaries of the long runs. For larger Fm, the first occurances are farther along the sequences. How can this approximate self-similarity relationship be made exact -- what is the source of the discrepancies? Are the sequences related to a parallel cut line? Are they Sturmian?






Fibonacci word kb number self-similarity color, link to

F( Fm*n ), Fm increasing down, n increasing to right.

F( Fm(m+2) * ( n | F(n)=b ) ), m increasing down, n increasing to right.

The rows of this diagrams are generated with the MATLAB program Fibonacci_arithmetic_subsequence_kb_similarity.m. See header for details.

Notes

    The second symbol of the Fibonacci word, b, occurs at indices:
n = 2, 5, 7, 10, 13, 15, 18, 20, 23, 26, 28, 31, 34, 36, 39,  41, 44, 47, 49, 52, 54, 57, 60, 62, 65...
(OEIS A001950, the Upper Wythoff sequence (a Beatty sequence ): a(n) = floor(n*phi^2), where phi = (1+sqrt(5))/2 )
   = 2, 5, 7, 2*5, 13, 3*5, 2*3*3, 2*2*5, 23, 2*13, 2*2*7, 31, 2*17, 2*2*3*3, 3*13, 41, 2*2*11, 47, 7*7, 2*2*13, 2*27, 3*19, 2*2*3*5, 2*31, 5*13, ...
parity of number of 2's in factorization: 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, ... 
parity of number of factors:              1, 1, 1, 0, 1, 0, 1, 1, 1, 0, 1, 1, 0, 0, 0, 1, 1, 1, 0, 1, 0, 0, 0, 0, ...

References

The Fibonacci Word Fractal
Alexis Monnerot-Dumaine, February 8, 2009

The Fibonacci Word Fractal is a self-similar fractal curve based on the Fibonacci word through a simple and interesting drawing rule. This fractal reveals three types of patterns and a great number of self-similarities. We show a strong link with the Fibonacci numbers, prove several properties and conjecture others, we calculate its Hausdorff Dimension. Among various modes of construction, we define a word over a 3-letter alphabet that can generate a whole family of curves converging to the Fibonacci Word Fractal. We investigate the sturmian words that produce variants of such a pattern. We describe an interesting dynamical process that, also, creates that pattern. Finally, we generalize to any angle.


Infinite self-similar words
Grytczuk J.
Discrete Mathematics, Volume 161, Number 1, 5 December 1996 , pp. 133-141(9)

The words we consider in this paper are defined by some self-similarity conditions which in particular are satisfied by the well-known Fibonacci word f = 1011010110110 ... . We discuss structural as well as asymptotical properties of these words.


Lyndon words and singular factors of Sturmian words
Guy Melançon
Theoretical Computer Science, Volume 218, Issue 1, 28 April 1999, Pages 41-59

Two different factorizations of the Fibonacci infinite word were given independently in Wen and Wen (1994) and Melançon (1996). In a certain sense, these factorizations reveal a selfsimilarity property of the Fibonacci word. We first describe the intimate links between these two factorizations. We then propose a generalization to characteristic sturmian words.


------------------------------------
 Public Domain DedicationThis work is dedicated to the Public Domain.
There are no restrictions on use of the images or code on this page. Claiming to be the originator of the material, explicitly or implicitly, is bad
karma. A link (if appropriate), a note to dow[at]uoregon.edu, and credit are appreciated but not required.

Comments are welcome (dow[at]uoregon.edu).