# The Narayana Numbers

Contents

Narayana triangle

Meta-Narayana triangle

## Narayana Triangle

The narayana numbers are an arrangement of numbers similar to Pascal's triangle that count a sort of permutation:

1 sum: 1 1 1 sum: 2 1 3 1 sum: 5 1 6 6 1 sum: 14 1 10 20 10 1 sum: 42 1 15 50 50 15 1 sum: 132 1 21 105 175 105 21 1 sum: 429 1 28 196 490 490 196 28 1 sum: 1430 1 36 336 1176 1764 1176 336 36 1 sum: 4862 1 45 540 2520 5292 5292 2520 540 45 1 sum: 16796Here it is easy to recognize the triangular numbers, Sloane's A000217, in the second diagonal. They are a "two-dimensional figurate" sequence, because they count the number of objects in an arragement that is determined by a two-dimensional figure.

The third diagonal (1, 6, 20, 50, 105, 196, ...) is a similar type of sequence describing four-dimensional hyper-pyramid beginning with a square-pyramidal base (Sloane's A002415). In a manner similar to the discussion in my A006542 page, it is made by taking successive sums of another sequence (A000330) which itself is made by taking the sums of yet another sequence (the squares, A000290).

The fourth diagonal (1, 10, 50, 175, 490, 1176, ...) is sequence A006542, which corresponds to a six-dimensional figure like a pyramid starting with a five-sided base. That sequence is described more fully here.

So we have a pattern, going from a triangle to a square to a pentagon,
and going from 2 to 4 to 6 dimensions. However, if you check this more
closely, you might notice an irregularity: whereas the triangle
sequence (1, 3, 6, 10, ...) is adding (2, 3, 4, ...) each time, and
the square sequence (1, 4, 9, 16, ...) is adding (3, 5, 7, ...) each
time — the pentagon sequence (1, 6, 16, 31, ...) is adding (5, 10,
15, ...) As N is going from 2 to 4 to 6, the N^{th}-order
differences go from 1 to 2 to 5. If the pattern were consistent, it
should be something more like (4, 7, 10, ...) (with a second-order
difference of 3).

This irregularity becomes a bit more surprising when you look at the
next diagonal of the Narayana triangle, (1, 15, 105, 490, 1764, 5292,
...) which turns out to derive from an 8^{th}-order difference of 14:

The actual pattern here is that the 2N^{th}-order difference
is the N^{th} Catalan number.

Self-Definition by Symmetry Contraints

Suppose we want to find a minimal set of constraints needed to fix all
of the numbers in the Narayana triangls. Michael Somos pointed
out^{1} to me that the following "axioms" suffice:

- The number at the top is 1.
- The triangle is bilaterally symmetrical.
- The N
^{th}diagonal follows a order 2(N-1) progression (referring to polynomial order, i.e. a constant sequence, a quadratic sequence, a quartic sequence, etc.). (Add as many initial zeros as needed to uniquely determine the sequence)

As it turns out, these three are sufficient, and the
2(N-1)^{th}-order differences of the N^{th} diagonal always turn
out to be the Catalan numbers as discussed above, which are also the
row sums of the N^{th} rows. One can easily see that a very similar
set of axioms defines Pascal's triangle, and suggests extension to
higher orders. But the above are not the only set of axioms that work.

### Extending the Sequence to Higher-order Triangles

There are at least two ways to go beyond Pascal and Narayana, that are self-consistent and easily estensible. The method used here is probably better known is to compute rows using successive products of fractions of binomial coefficients, the product of fractions method (follow that link for larger tables of the triangles, and links to related OEIS sequences).

The other method uses successive differences and
produces triangles that differ starting with the 3^{rd} (i.e. the
first triangle after Narayana); somewhat smaller values and row sums.
(Again, follow that link for larger tables of the triangles and links
to related OEIS sequences).

## The "Meta-Narayana" or Hoggatt Triangle

Named for Verner Hoggatt, who worksed out many of its properties in 1977, this triangle is discussed in [2].

This triangle is discussed more fully on myA001181 page. The row sums (1, 2, 6, 22, 92, 422, ...) are the Baxter permutations (Sloane's A1181).

1 sum: 1 1 1 sum: 2 1 4 1 sum: 6 1 10 10 1 sum: 22 1 20 50 20 1 sum: 92 1 35 175 175 35 1 sum: 422 1 56 490 980 490 56 1 sum: 2074 1 84 1176 4116 4116 1176 84 1 sum: 10754 1 120 2520 14112 24696 14112 2520 120 1 sum: 58202 1 165 4950 41580 116424 116424 41580 4950 165 1 sum: 326240 1 220 9075 108900 457380 731808 457380 108900 9075 220 1 sum: 1882960
To generate a row of this triangle, we need the 2^{nd} diagonal: 1, 4,
10, 20, 35, 56, 84, 120, 165, 220, ... which are the
"tetrahedral" numbers. To get a row
N, we need the first N-1 of these numbers. For example, to get the
7^{th} row we use the numbers 1, 4, 10, 20, 35, 56. The first number
in a row is always 1. Then you iterate the process of multiplying by a
tetrahedral number and dividing by another one:

first term: 1

1×56/1 = 56

56×35/4 = 490

490×20/10 = 980

980×10/20 = 490

490×4/35 = 56

56×1/56 = 1

## Meta-Meta-Narayana Triangle

The row sums (1, 2, 7, 32, 177, ...) are the Hoggatt sequence (Sloane's A5362).

1 sum: 1 1 1 sum: 2 1 5 1 sum: 7 1 15 15 1 sum: 32 1 35 105 35 1 sum: 177 1 70 490 490 70 1 sum: 1122 1 126 1764 4116 1764 126 1 sum: 7898 1 210 5292 24696 24696 5292 210 1 sum: 60398 1 330 13860 116424 232848 116424 13860 330 1 sum: 494078 1 495 32670 457380 1646568 1646568 457380 32670 495 1 sum: 4274228
To generate a row of this triangle, we need the 2^{nd} diagonal: 1, 5,
15, 35, 70, 126, 210, 330, 495, ... which are the
"hyper-tetrahedral" numbers. To get a row
N, we need the first N-1 of these numbers. For example, to get the
7^{th} row we use the numbers 1, 5, 15, 35, 70, 126. The first number
in a row is always 1. Then you iterate the process of multiplying by a
hyper-tetrahedral number and dividing by another one:

first term: 1

1×126/1 = 126

126×70/5 = 1764

1764×35/15 = 4116

4116×15/35 = 1764

1764×5/70 = 126

126×1/126 = 1

References

**1 :**
Michael Somos, personal correspondence, 2022.

[2] Daniel C. Fielder and Cecil O. Alford, "On a conjecture by Hoggatt with extensions to Hoggatt sums and Hoggatt triangles." Fibonacci Quarterly #27(2) pp. 160-168 (1989).

Some other sequences are discussed here.

This page was written in the "embarrassingly readable" markup language RHTF, and was last updated on 2024 Mar 01. s.27