cp's OEIS Frontend

The array is read by downward antidiagonals: A(1,1), A(1,2), A(2,1), A(1,3), A(2,2), A(3,1), etc.

In Kimberling's terminology, this array is called the dispersion of sequence A255411 (when started from its first nonzero term, 4). The left column is the complement of that sequence, which is A256450.

To compute a(n) we shift its binary representation right as many steps k as necessary that the result were an odd number. Then one is added to that odd number, and the prime factorization of the resulting even number is shifted the same k number of steps towards larger primes, whose product is then decremented by one to get the final result.

In the essence, a(n) tells which number in array A246275 is at the same position where n is in the array A135764. As the topmost row in both arrays is A005408 (odd numbers), they are fixed, i.e. a(2n+1) = 2n+1 for all n.

Equally: a(n) tells which number in array A246273 is at the same position where n is in the array A054582, as they are the transposes of above two arrays.

T(n,k) is the number of paths from node 0 to odd k in a directed graph with 2n+1 vertices labeled 0, 1, ..., 2n+1 and edges leading from i to i+1 for all i, from i to i+2 for even i, and from i to i-2 for odd i. - Grace Work, Mar 01 2020

In the essence, a(n) tells which number in square array A249741 (the sieve of Eratosthenes minus 1) is at the same position where n is in array A135764, which is formed from odd numbers whose binary expansions are shifted successively leftwards on the successive rows. As the topmost row in both arrays is A005408 (odd numbers), they are fixed, i.e., a(2n+1) = 2n+1 for all n.

Equally: a(n) tells which number in array A114881 is at the same position where n is in the array A054582, as they are the transposes of above two arrays.

From Peter Munn, Apr 28 2025: (Start)

Each row in the Wythoff array, A035513, and this extended array satisfies the Fibonacci recurrence; that is each term after the first 2 is the sum of the preceding 2 terms.

We use F_i to denote the i-th Fibonacci term, A000045(i). In particular, we refer below to F_0 = 0, F_1 = 1 and F_2 = 1 several times. Note that to fully understand the description of the relationship between neighboring columns it is important to distinguish F_1 and F_2, although they have the same integer value. Similarly, the identity of an array term should be understood here as including its position in the array, not only its integer value.

The terms of this extended Wythoff array map 1:1 onto the nonempty finite subsets of Fibonacci terms (from F_0 onwards) that do not include both F_i and F_{i+1} for any i. With this map each term is the sum of its subset image. See the table in the examples.

Full description of the mapping with its relationship to A035513:

The (unextended) Wythoff array A035513 includes every positive integer exactly once. So, using the Zeckendorf representation (see link below), the array terms map 1:1 to nonempty finite subsets of the Fibonacci terms from F_2 onwards -- more precisely, onto those that do not include both F_i and F_{i+1} for any i. (Again, each array term is the sum of the Fibonacci numbers from the relevant subset.)

As shown in the Kimberling 1995 link, when we proceed from one term to the next in a row, the indices of the Fibonacci terms in the corresponding subset are incremented. When we proceed leftwards, the indices are decremented, with the subsets for the leftmost column being those that include F_2.

And when we add 2 columns on the left of the Wythoff array, the mapping continues to decrement the indices, so the corresponding extra subsets have F_0 (new leftmost column) or F_1 as their first Fibonacci term.

Thus the terms of this extended Wythoff array map 1:1 onto the nonempty finite subsets of Fibonacci terms (from F_0 onwards) that do not include both F_i and F_{i+1} for any i. The leftmost column is the nonnegative integers: if we were to remove F_0 (value 0) from the subset for an integer in this column, the subset would form the Zeckendorf representation of the integer, as subsets do in the unextended array.

(End)

Transpose of A141396.

Background discussion: Suppose that s is an increasing sequence of positive integers, that the complement t of s is infinite, and that t(1)=1. The dispersion of s is the array D whose n-th row is (t(n), s(t(n)), s(s(t(n))), s(s(s(t(n)))), ...). Every positive integer occurs exactly once in D, so that, as a sequence, D is a permutation of the positive integers. The sequence u given by u(n)=(number of the row of D that contains n) is a fractal sequence. Examples:

(1) s=A000040 (the primes), D=A114537, u=A114538.

(2) s=A022343 (without initial 0), D=A035513 (Wythoff array), u=A003603.

(3) s=A007067, D=A035506 (Stolarsky array), u=A133299.

More recent examples of dispersions: A191426-A191455.

In the essence, a(n) tells which number in the array A135764 is at the same position where n is in the array A249741, the sieve of Eratosthenes minus 1. As the topmost row in both arrays is A005408 (odd numbers), they are fixed, i.e., a(2n+1) = 2n+1 for all n.

Equally: a(n) tells which number in array A054582 is at the same position where n is in the array A114881, as they are the transposes of above two arrays.

Row 1: A000225 (-1+2^n)

Row 2: A000079 (2^n)

Row 3: A055010

Row 4: 3*A000079

Row 5: A153894

Row 6: 5*A000079

Row 7: A086224

Row 8: A005009

Row 9: A052996

For a background discussion of dispersions, see A191426.

...

Each of the sequences (4n, n>2), (4n+1, n>0), (3n+2, n>=0), generates a dispersion. Each complement (beginning with its first term >1) also generates a dispersion. The six sequences and dispersions are listed here:

...

A191663=dispersion of A042948 (0 or 1 mod 4 and >1)

A054582=dispersion of A005843 (0 or 2 mod 4 and >1; evens)

A191664=dispersion of A014601 (0 or 3 mod 4 and >1)

A191665=dispersion of A042963 (1 or 2 mod 4 and >1)

A191448=dispersion of A005408 (1 or 3 mod 4 and >1, odds)

A191666=dispersion of A042964 (2 or 3 mod 4)

...

EXCEPT for at most 2 initial terms (so that column 1 always starts with 1):

A191663 has 1st col A042964, all else A042948

A054582 has 1st col A005408, all else A005843

A191664 has 1st col A042963, all else A014601

A191665 has 1st col A014601, all else A042963

A191448 has 1st col A005843, all else A005408

A191666 has 1st col A042948, all else A042964

...

There is a formula for sequences of the type "(a or b mod m)", (as in the Mathematica program below):

If f(n)=(n mod 2), then (a,b,a,b,a,b,...) is given by

a*f(n+1)+b*f(n), so that "(a or b mod m)" is given by

a*f(n+1)+b*f(n)+m*floor((n-1)/2)), for n>=1.

This sequence is a permutation of the natural numbers. - L. Edson Jeffery, Aug 13 2014

Row 1: A000975

Row 2: A081254

Row 3: A081253

Row 4: A052997

For a background discussion of dispersions, see A191426.

...

Each of the sequences (4n, n>2), (4n+1, n>0), (3n+2, n>=0), generates a dispersion. Each complement (beginning with its first term >1) also generates a dispersion. The six sequences and dispersions are listed here:

...

A191663=dispersion of A042948 (0 or 1 mod 4 and >1)

A054582=dispersion of A005843 (0 or 2 mod 4 and >1; evens)

A191664=dispersion of A014601 (0 or 3 mod 4 and >1)

A191665=dispersion of A042963 (1 or 2 mod 4 and >1)

A191448=dispersion of A005408 (1 or 3 mod 4 and >1, odds)

A191666=dispersion of A042964 (2 or 3 mod 4)

...

EXCEPT for at most 2 initial terms (so that column 1 always starts with 1):

A191663 has 1st col A042964, all else A042948

A054582 has 1st col A005408, all else A005843

A191664 has 1st col A042963, all else A014601

A191665 has 1st col A014601, all else A042963

A191448 has 1st col A005843, all else A005408

A191666 has 1st col A042948, all else A042964

...

There is a formula for sequences of the type "(a or b mod m)", (as in the Mathematica program below):

If f(n)=(n mod 2), then (a,b,a,b,a,b,...) is given by

a*f(n+1)+b*f(n), so that "(a or b mod m)" is given by

a*f(n+1)+b*f(n)+m*floor((n-1)/2)), for n>=1.