Chapter 1-40
Chapter 1-40
Chapter 1-40
1
Plane figurate numbers 2
Similarly, by adding to a point three, five, seven etc. points and organizing
them in the form of a square, one can obtain the numbers 1, 4, 9, 16, 25, 36,
49, 64, 81, 100, ..., which are called square numbers.
Figurate Numbers Downloaded from www.worldscientific.com
by KAINAN UNIVERSITY on 02/08/15. For personal use only.
By adding to a point four, seven, ten etc. points and forming from them a
regular pentagon, one can construct pentagonal numbers
1, 5, 12, 22, 35, 51, 70, 92, 117, 145, . . ..
heptagonal numbers 1, 7, 18, 34, 55, 81, 112, 148, 189, 235, . . ., octagonal
numbers 1, 8, 21, 40, 65, 96, 133, 176, 225, 280, . .. , nonagonal
numbers 1, 9, 24, 46, 75, 111, 154, 204, 261, 325, . .. ,decagonal
numbers 1, 10, 27, 52, 85, 126, 175, 232, 297, 370, . ..,
Plane figurate numbers 3
hendecagonal numbers 1, 11, 30, 58, 95, 141, 196, 260, 333, 415, . .. ,
dodecagonal numbers 1, 12, 33, 64, 105, 156, 217, 288, 369, 460, . .. , etc.
So, we have constructed several simplest classes of polygonal numbers
— positive integers, corresponding to an arrangement of points on the
plane, which forms a regular polygon. One speaks about m-gonal numbers
if the arrangement forms a regular m-gon.
1.1.2. Polygonal numbers were a concern of Pythagorean Geometry,
since Pythagoras is credited with initiating them, and originating the
notion that these numbers are generated from a gnomon or basic unit. A
gnomon is a shape which, when added to a figure, yields another figure
similar to the original. So, in our case, a gnomon is the piece which needs to
be added to a polygonal number to transform it to the next bigger one.
The gnomon of a triangular number is the positive integer of the
general form 𝑛 + 1, 𝑛 = 1, 2, 3, . . .: starting with n-th triangular number,
one obtains (n + 1)-th triangular number adjoining the line with n + 1
elements. For instance, the 21-point triangle, composed of gnomons, looks
like this:
The gnomon of the square number is the odd number of the general form
2n + 1, n = 1, 2, 3, . . . : in order to get the (n + 1)-th square from n-th square,
we should adjoin 2n + 1 elements, one to the end of each column, one to the
end of each row, and a single one to the corner. The square of size 6 x 6,
composed of gnomons, looks like this:
Plane figurate numbers 4
The general rule for enlarging the regular polygon to the next size is to
extend two adjacent sides by one point and then to add the required extra
sides between those points. So, to transform n-th m-gonal number into the
(n + 1)-th m-gonal number, one adjoins (m — 2)n + 1 elements.
Therefore, the triangular numbers are obtained as consecutive sums of
elements of the arithmetic progression 1, 2, 3, 4, . . . , n , . . . ; namely, they
are 1 = 1, 3 = 1 + 2, 6 = 3 + 3, 10 = 6 +, … The square numbers are
obtained as consecutive sums of elements of the arithmetic progression
1,3, 5, 7, . . . , 2𝑛 + 1, . . . : 1 = 1, 4 = 1 + 3, 9 = 4 + 5, 16 = 9 + 7, … The
pentagonal numbers are obtained by summation of the elements of the
arithmetic progression 1, 4, 7,9 , . . . , 3𝑛 + 1 , . . . : 1 = 1 , 5 = 1 + 4 , 12 =
5 + 7 , 22 = 12 + 10 , . . .. The hexagonal numbers are obtained by
summation of the elements of the arithmetic progression
1, 5, 9, 13, . . . , 4𝑛 + 1, . . . : 1 = 1, 6 = 1 + 5, 15 = 6 + 9, 28 = 15 +
13, . . ., and so on.
Putting the points in one line, one can speak about linear numbers. In
fact, any positive integer is a linear number. Similarly to the above
construction, the linear numbers have, as a gnomon, the number 1, and are
obtained as consecutive sums of elements of the sequence
1, 1, 1, 1, . . . , 1, . . . : 1 = 1, 2 = 1 + 1, 3 = 2 + 1, 4 = 3 + 1, … .
In fact, the linear numbers are one-dimensional analogues of two-
dimensional polygonal numbers.
1.1.3. The first general definition of m-gonal numbers was given by
Hypsicles of Alexandria in Il-th century BC and was quoted by Diophantys
in his tract On polygonal numbers (see [Diop], [Heat10]): if there are as many
numbers as we please beginning with one and increasing by the same common
difference, then when the common difference is 1, the sum of all the terms is a
triangular number; when 2, a square; when 3, a pentagonal number; and the
number of the angles is called after the number exceeding the common difference by
2, and the side after the number of terms including 1.
In contemporary mathematical language, it has the following form: n-th
m-gonal number 𝑆𝑚 (𝑛) is the sum of the first n elements of the arithmetic
progression
1,1 + (𝑚 − 2), 1 + 2(𝑚 − 2), 1 + 3(𝑚 − 2), … , 𝑚 ≥ 3.
So, by definition, it holds
𝑆𝑚 (𝑛) = 1 + (1 + (𝑚 − 2)) + (1 + (𝑚 − 2)) + ⋯ + (1 + (𝑚 − 2)(𝑛 − 1)).
Plane figurate numbers 5
In particular, we get
𝑆3 (𝑛) = 1 + 2 + ⋯ + 𝑛, 𝑆4 (𝑛) = 1 + 3 + ⋯ + (2𝑛 − 1),
𝑆5 (𝑛) = 1 + 4 + ⋯ + (3𝑛 − 2)
𝑆6 (𝑛) = 1 + 5 + ⋯ + (4𝑛 − 3), 𝑆7 (𝑛) = 1 + 6 + ⋯ + (5𝑛 − 4),
𝑆8 (𝑛) = 1 + 7 + ⋯ + (6𝑛 − 5).
Above expression implies the following recurrent formula for m-gonal
numbers:
𝑆𝑚 (𝑛 + 1) = 𝑆𝑚 (𝑛) + (1 + (𝑚 − 2)𝑛), 𝑆𝑚 (1) = 1.
In particular, we get
𝑆3 (𝑛 + 1) = 𝑆3 (𝑛) + (𝑛 + 1), 𝑆4 (𝑛 + 1) = 𝑆4 (𝑛) + (2𝑛 + 1),
𝑆5 (𝑛 + 1) = 𝑆5 (𝑛) + (3𝑛 + 1),
𝑆6 (𝑛 + 1) = 𝑆5 (𝑛) + (4𝑛 + 1), 𝑆7 (𝑛 + 1) = 𝑆7 (𝑛) + (5𝑛 + 1),
𝑆8 (𝑛 + 1) = 𝑆8 (𝑛) + (6𝑛 + 1).
For many applications it is convenient to add the value 𝑆𝑚 (0) = 0to the
list.
Since the sum of the first n elements of an arithmetic progression
𝑎 +𝑎
𝑎1 , … , 𝑎𝑛 , … is equal to 1 𝑛 ∙ 𝑛, one obtains the following general formula
2
for n-th m-gonal number:
𝑛((𝑚 − 2)𝑛 − 𝑚 + 4) (𝑚 − 2) 2
𝑆𝑚 (𝑛) = = (𝑛 − 𝑛) + 𝑛
2 2
(𝑚 − 2)𝑛2 − (𝑚 − 4)𝑛 1
= = 𝑚(𝑛2 − 𝑛) − 𝑛2 + 2𝑛.
=2 2
On the other hand, this formula can be obtained by induction using the
fact, that the triangular number with the index n +1 is obtained from the
triangular number with the index n by addition of the number n + 1. For n
1∙(1+1)
= 1 one has 𝑆3 (1) = 1 = .Going from n to n + 1, one obtains 𝑆3 (𝑛 +
2
𝑛(𝑛+1) (𝑛+1)(𝑛+2)
1) = 𝑆3 (𝑛) + (𝑛 + 1) = + (𝑛 + 1) = .
2 2
Plane figurate numbers 7
𝑆𝑚 (𝑛 + 1) = 𝑆𝑚 (𝑛) + (1 + (𝑚 − 2)𝑛)
𝑛((𝑚 − 2)𝑛 − 𝑚 + 4)
= + (1 + (𝑚 − 2)𝑛)
2
𝑛2 (𝑚 − 2) + 𝑛(4 − 𝑚) + 2𝑛(𝑚 − 2) + 2
=
2
(𝑛2 + 𝑛 + 1)(𝑚 − 2) + 𝑛 + 1)(4 − 𝑚)
=
2
(𝑛 + 1)((𝑚 − 2)(𝑛 + 1) − 𝑚 + 4)
= .
2
The special summation1 has the form
2𝑆𝑚 (𝑛) =
1 + 1 + (𝑚 − 2) + … + 1 + (𝑚 − 2)(𝑛 − 1) +
1 + (𝑚 − 2)(𝑛 − 1) + 1 + (𝑚 − 2)(𝑛 − 2) + ⋯ + 1.
It gives 2𝑆𝑚 (𝑛) = (2 + (𝑚 − 2)(𝑛 − 1) + ⋯ + (2 + (𝑚 − 2)(𝑛 − 1)) =
𝑛((𝑚−2)𝑛−𝑚+4)
𝑛(2 + (𝑚 − 2)(𝑛 − 1)), and 𝑆𝑚 (𝑛) = .
2
Finally, another way to derive the general formula for 𝑆𝑚 (𝑛)is to use
the first three m-gonal numbers to find the coefficients A, B, and C of the
general 2-nd degree polynomial 𝐴𝑛2 + 𝐵𝑛 + 𝐶. For n = 1
1According to legend, at age 10 Gauss was told by his teacher to sum up all the numbers from
1 to 100. He reasoned that each number i could be paired up with 101 — i, to form a sum of
101, and if this was done 100 times, it would result in twice the actual sum, since each number
100∙101
would get used twice due to the pairing. Hence, the sum would be 1 + ⋯ + 100 = .
2
Plane figurate numbers 9
𝑥((𝑚 − 3)𝑥 + 1)
= 𝑆𝑚 (1)𝑥 + 𝑆𝑚 (2)𝑥 2
(1 − 𝑥)3
+𝑆𝑚 (3)𝑥 3 + ⋯ + 𝑆𝑚 (𝑛)𝑥 𝑛 + ⋯ , |𝑥| < 1.
In particular, one gets
𝑥
= 𝑥 + 3𝑥 2 + 6𝑥 3 + ⋯ + 𝑆3 (𝑛)𝑥 𝑛 + ⋯ , |𝑥| < 1;
(1 − 𝑥)3
𝑥(𝑥 + 1)
= 𝑥 + 4𝑥 2 + 9𝑥 3 + ⋯ + 𝑆4 (𝑛)𝑥 𝑛 + ⋯ , |𝑥| < 1;
(1 − 𝑥)3
𝑥(2𝑥 + 1)
= 𝑥 + 5𝑥 2 + 12𝑥 3 + ⋯ + 𝑆5 (𝑛)𝑥 𝑛 + ⋯ , |𝑥| < 1:
(1 − 𝑥)3
𝑥(3𝑥 + 1)
= 𝑥 + 6𝑥 2 + 15𝑥 3 + ⋯ + 𝑆6 (𝑛)𝑥 𝑛 + ⋯ , |𝑥| < 1.
(1 − 𝑥)3
In order to obtain the above formula, let us consider two polynomials
𝑓(𝑥)
In fact, one has the decomposition = 𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 + ⋯ + 𝑐𝑛 𝑥 𝑛 + ⋯ if
𝑔(𝑥)
|x|<r, and 𝑟 = min |𝑥𝑖 |, where 𝑥1 , … , 𝑥𝑛 are the roots of the polynomial
1≤𝑛
𝑓(𝑥)
𝑔(𝑥). It yields that the rational function is the generating function of the
𝑔(𝑥)
obtained sequence 𝑐0 , 𝑐1 , 𝑐2 , … , 𝑐𝑛 , … .Moreover, one gets 𝑓(𝑥) =
Plane figurate numbers 10
Now we can find the generating function for the sequence of m-gonal
numbers. Let us consider the recurrent equation Sm(n + 1) — Sm(n) + (1 + (m
— 2)n). Going from n to n + 1, one
Plane figurate numbers 11
S m( n + 3) — 3 S m( n + 2) + 3 S m( n + 1) — S m( n) = 0.
It is a linear recurrent equation of 3-rd order with coefficients b 0 = 1 , b i =
—3,b2 = 3,63 = —1. Its initial values are Sm(1) = 1, Sm(2) = m, Sm(3) = 3m — 3.
Denoting Sm(n + 1) by cn, one can rewrite the above equation as
cn+3 3cn+2 + 3cn+i cn — 0, c0 — 1, cl — m, c2 — 3m 3.
2
Sinceg (x ) = 1—3x+3x2 —x3 = (1—x)3 has three coinciding roots xi = x2 = x3 =
1, the generating function for the sequence of the m-gonal numbers obtains
the form
1 + (m — 3)x
Sm(1) + Sm(2)x
(1 — x)3
+ STO(3)x2 + ••• + Sm(n)xn i +n-i
••• , \x\ < 1.
1+ ⋯+𝑛 + 1+ 2 + ⋯+ 𝑛 +
3𝑆3 (𝑛) + 𝑆3 (𝑛 − 1) = 𝑛 + (𝑛 − 1) + ⋯ + 1 +
1 + ⋯ + (𝑛 − 1)
implies 3𝑆3 (𝑛) + 𝑆3 (𝑛 − 1) = 1 + ⋯ + 𝑛 + (𝑛 + 1) + ⋯ + 2𝑛 = 𝑆3 (2𝑛).
1.2.3. A triangular number with odd index can be constructed using
the following similar formula:
3𝑛(𝑛+1) (𝑛+1)(𝑛+2)
In fact, one has 3𝑆3 (𝑛) + 𝑆3 (𝑛 + 1 ) = + =
2 2
𝑛 2𝑛(2𝑛+1)
(4𝑛 + 2) = = 𝑆3 (2𝑛 + 1).
2 2
3𝑆3 (𝑛) + 𝑆3 (𝑛 + 1) =
1 + ⋯+ 𝑛 +1 + 2 + ⋯+ 𝑛 +
𝑛 + (𝑛 − 1) + ⋯ + 1 +
1 + ⋯ + (𝑛 − 1) + 𝑛 + (𝑛 + 1)
shows, that
𝑆3 (𝑛) + 𝑆3 (𝑛 + 2) = 𝑆3 (𝑛 − 1) + 𝑆3 (𝑛 + 1) + 𝑛 + (𝑛 + 2)
= 2𝑆3 (𝑛) + 1 + 2(𝑛 + 1) = 2(𝑆3 (𝑛) + (𝑛 + 1)) + 1
= 2𝑆3 (𝑛 + 1) + 1.
figurate numbers 18
1 + 2 + ⋯ + (𝑛 − 1) +
1 + ⋯ + (𝑛 − 1) +
𝑆3 (𝑛) + 3𝑆3 (𝑛 − 1) = 1 + ⋯ + (𝑛 − 1) +
1 + ⋯ + (𝑛 − 1) +
gives 𝑆3 (𝑛) + 𝑆3 (𝑛 − 1) = 1 + 5 + ⋯ + (4(𝑛 − 1) + 1) = 𝑆6 (𝑛).
It is easy to see that we just rearranged the four small triangles on the
previous picture into one big triangle.
Of course, one can prove it by induction: for n = 1, it holds 𝑆6 (1) = 1 =
𝑆3 (2 ∙ 1 − 1), and, going from n to n + 1, we obtain
𝑆6 (𝑛 + 1) = 𝑆6 (𝑛) + (4𝑛 + 1) = 𝑆3 (2𝑛 − 1) + (4𝑛 + 1)
(2𝑛 − 1)2𝑛 4𝑛2 + 6𝑛 + 2
= + (4𝑛 + 1) =
2 2
(2𝑛 + 1)(2𝑛 + 2)
= = 𝑆3 (2𝑛 + 1) = 𝑆3 (2(𝑛 + 1) − 1).
2
A special summation of the form
1 + 2 + 4 + ⋯ + (2𝑛 − 2) +
𝑆3 (2𝑛 − 1) =
3 + 5 + ⋯ + (2𝑛 − 1)
also gives 𝑆3 (2𝑛 − 1) = 1 + 5 + 9 + ⋯ + (4𝑛 − 3) = 𝑆6 (𝑛).
1.2.10. The octagonal number theorem shows a connection between
octagonal and triangular numbers:
𝑆8 (𝑛) = 6𝑆3 (𝑛 − 1) + 𝑛.
(𝑛−1)𝑛
In fact, one 6𝑆3 (𝑛 − 1) + 𝑛 = 6 + 𝑛 = (3𝑛2 − 3𝑛) + 𝑛 = 3𝑛2 − 2𝑛 =
2
6𝑛2 −4𝑛
= 𝑆8 (𝑛).
2
The geometrical interpretation of this property for n = 4 is given below:
3
It means any polygonal number is a linear combination of triangular numbers
with non-zero coefficients.
A geometrical illustration of this property for 𝑚 = 6 was given in the
proof of the formula 𝑆6 (𝑛) = 𝑆3 (𝑛) + 3𝑆3 (𝑛 − 1), which is an particular case
of the Bachet de Meziriac formula.
1.2.13. An additional useful equation has the following form:
(3 + 2√2)𝑛 + (3 − 2√2)𝑛 − 2
𝑢𝑛 = , 𝑎𝑛𝑑
4
(3 + 2√2)𝑛 − (3 − 2√2)𝑛
𝑣𝑛 =
4√2
1
√2 = 1 + = [1,2,2,2,2, … ] = [1, (2)],
1
2+
2+⋯
i.e., 𝑘 = 1, and all positive integer solutions of the equation 𝑥 2 − 2𝑦 2 =
1 have the form 𝑥 = 𝑃2𝑛−1 , 𝑦 = 𝑄2𝑛−1 , 𝑛 ∈ 𝑁
The numerators and denominators of the convergents of a continued
1
fraction 𝑎0 + 1 = [𝑎0 , 𝑎1 , … , 𝑎𝑛 , … ]have the following property (see
𝑎1 +
𝑎2 +⋯
[Buch09]):
𝑃𝑛 = 𝑎𝑛 𝑃𝑛−1 + 𝑃𝑛−2 , and 𝑄𝑛 = 𝑎𝑛 𝑄𝑛−1 + 𝑄𝑛−2
for any positive integer 𝑛, 𝑛 ≥ 2. In particular, for the decomposition
√2 = [1, (2)] of √2 into the continued fraction, we have:
𝑃0 1
𝑃0 = 1, 𝑄0 = 1 (since = 1 = );
𝑄0 1
𝑃1 1 3
𝑃1 = 3, 𝑄2 = 2 (since = 1+ = );
𝑄1 2 2
These formulas with the identities (1 ± √2)2 = 3 ± 2√2 yield now4 4 the
formulas for 𝑢𝑛 and 𝑣𝑛 :
𝑃2𝑛−1 − 1
𝑢𝑛 =
2
(1 + √2)2𝑛 + (1 − √2)2𝑛 − 2 (3 + 2√2)𝑛 + (3 − 2√2)𝑛 − 2
= = ,
4 4
𝑄2𝑛−1 (1 + √2)2𝑛 − (1 − √2)2𝑛
𝑣𝑛 = =
2 4√2
(3 + 2√2)𝑛 − (3 − 2√2)𝑛
= .
4√2
For small values of n we obtain the following results:
for n=1,one has 𝑢1 = 1, 𝑣1 = 1, and 𝑆4,3 (1) = 𝑆3 (1) = 𝑆4 (1) = 1;
for n=2, one has 𝑢2 = 8, 𝑣2 = 6, 𝑎𝑛𝑑 𝑆4,3 (2) = 𝑆3 (8) = 𝑆4 (6) = 36;
for n=3, one has 𝑢3 = 49, 𝑣3 = 35, 𝑎𝑛𝑑 𝑆4,3 (3) = 𝑆3 (49) = 𝑆4 (35) =
1225;
for n=4, one has 𝑢4 = 288, 𝑣4 = 204, 𝑎𝑛𝑑 𝑆4,3 (4) = 𝑆3 (288) = 𝑆4 (204) =
41616;
for n=5, one has 𝑢5 = 1681, 𝑣5 = 1189, 𝑎𝑛𝑑 𝑆4,3 (5) = 𝑆3 (1681) =
𝑆4 (1189) = 1413721;
for n=6, one has 𝑢6 = 9800, 𝑣6 = 6930, 𝑎𝑛𝑑 𝑆4,3 (6) = 𝑆3 (9800) =
𝑆4 (6930) = 48024900;
for 𝑛 = 7, one has 𝑢7 = 57121, 𝑣7 = 40391, and 𝑆4,3 (7) = 𝑆3 (57121) =
𝑆4 (40391) = 1631432881.
4of course, the same result can be obtained using the following well-known rule:
if(x0 , y0 )is the smallest positive integer solution of the Pell’s equation x 2 − Dy 2 = 1,then all
positive integer solutions (x,y)of the equation are given by x + y√D = ±(x0 + Y0 √D)n , n ∈ N;
(x0 +y0 √D)n +(x0 −y0 √D)n (x0 +y0 √D)n −(x0 −y0 √D)n
therefore, x = ,and , x = , n ∈ N, in our case D=2,
2 2
(2+3√2)n +(2−3√2)n (2+3√2)n −(2−3√2)n
(x0 , y0 ) = (3,2),and we have , x = ,y= .
2 2√2
1.3.3. Using the above formulas, one obtains now, that n-th square
triangular number𝑆4,3 (𝑛) has the following form:
1
𝑆4,3 (𝑛) = ((17 + 12√2)𝑛 + (17 − 12√2)𝑛 − 2).
32
In fact, 𝑆4,3 (𝑛) = 𝑣𝑛2 . Noting that (3 ± 2√2)2 = 17 ± 12√2, one gets
2
(3 + 2√2)𝑛 − (3 − 2√2)𝑛
𝑆4,3 (𝑛) = ( )
4√2
1
= ((17 + 12√2) + (17 − 12√2)𝑛 − 2).
𝑛
32
𝑢
1.3.4. As n becomes larger, the ratio 𝑛 approaches √2 = 1.4142 …:
𝑣𝑛
𝑢1 1 𝑢2 8 𝑢3 49
= = 1; = = 1.3333 … ; = = 1.4;
𝑣1 1 𝑣2 6 𝑣3 35
𝑢4 288 𝑢5 1681
= = 1.4117 … ; = = 1.4137 … ;
𝑣4 204 𝑣5 1189
𝑢6 9800 𝑢7 57121
= = 1.4141 … ; = = 1.4142 ….
𝑣6 6930 𝑣7 40391
𝑢𝑛 𝑃2𝑛−1 −2 𝑃2𝑛−1 2 𝑃1
In general, it holds = = − , since 𝑄𝑡 → ∞ and →
𝑣𝑛 𝑄2𝑛−1 𝑄2𝑛−1 𝑄2𝑛−1 𝑄𝑡
𝑢𝑛
√2 for t→ ∞ , we get 𝑙𝑖𝑚 = √2.
𝑛→∞ 𝑣𝑛
𝑆4,3 (𝑛+1)
Similarly, the ratio of successive square triangular numbers
𝑆4,3 (𝑛)
approaches 17+12√2 = 33.9705 … . (𝑠𝑒𝑒[𝑊𝑖𝑘𝑖11]):
36 1225 41616
= 36; = 34.0277 … ; = 33.9722 … ;
1 36 1225
1413721 48024900
= 33.9706 … ; = 33.9705 ….
41616 1413721
1.3.5. There is an another way to obtain the consecutive members of
the sequence of square triangular numbers (see [CoGu96]): n-th square
𝑃𝑛
triangular number 𝑆4,3 (𝑛) is equal to (𝑃𝑛 𝑄𝑛 )2 where is n-th convergent in the
𝑄𝑛
2 ∙ 41 + 17 99 2 ∙ 99 + 41 239
= , = .
2 ∙ 29 + 12 70 2 ∙ 70 + 29 169
The numerators and denominators of the obtained fractions permit
to construct the first few square triangular numbers:
The first equality is obvious, while the second one follows from the
equality 𝑢𝑛 (𝑢𝑛 + 1) = 2𝑆4,3 (𝑛) and the chain of simple inequalities:
Hence, one has 𝑢𝑛 < √2𝑆4,3 (𝑛) < 𝑢𝑛 + 1, i.e., 𝑢𝑛 is the greatest integer less
then or equal to √2𝑆4,3 (𝑛) .So, we get 𝑢𝑛 = ⌊√2𝑆4,3 (𝑛)⌋
1.3.7. The above result of Euler, giving all indices 𝑣𝑛 and 𝑢𝑛 of the
square and triangular numbers, which correspond to n-th square triangular
number 𝑆4,3 (𝑛) = 𝑆4 (𝑣𝑛 ) = 𝑆3 (𝑢𝑛 ), is very beautiful, but almost useless to
apply for big values of n. However, there are other ways for calculation
indices 𝑣𝑛 and 𝑢𝑛 .In particular, all positive integer solutions 𝑢𝑛 and 𝑣𝑛 of the
1
equation 𝑢(𝑢 + 1) = 𝑣 2 can be obtained by the following recurrent formulas:
2
𝑎 𝑏
𝑣𝑛−1 = − = 𝑎(3 − 2√2) − 𝑏(3 + 2√2)
3 + 2√2 3 − 2√2
= 3(𝑎 − 𝑏) − 2√2(𝑎 + 𝑏).
Hence, one gets 𝑣𝑛−1 + 𝑢𝑛+1 = 6(𝑎 − 𝑏) = 6𝑣𝑛. In other words, 𝑣𝑛+1 =
6𝑣𝑛 − 𝑣𝑛−1 holds.
𝑛 𝑛
(3+√2) +(3−2√2) −2 1
Similarly, one has 𝑢𝑛 = . Let 𝑢𝑛 = 𝑐 + 𝑑 − , where
4 2,
𝑛
(3+√2) (3−2√2) 𝑛
𝑐= , and d = . It implies 𝑢𝑛+1 = 𝑐(3 + 2√2) + 𝑑(3 − 2√2) −
4 4
1 1
= 3(𝑐 + 𝑑) + 2√2(𝑐 − 𝑑) − , and
2 2
𝑐 𝑑 1
𝑢𝑛−1 = + −
3 + 2√2 3 − 2√2 2
1 1
= 𝑐(3 − 2√2) + 𝑑(3 + 2√2) − = 3(𝑐 + 𝑑) − 2√2(𝑐 − 𝑑) − .
2 2
1
Hence, one gets 𝑢𝑛+1 + 𝑢𝑛−1 = 6(𝑐 + 𝑑) − 1 = 6 (𝑐 + 𝑑 − ) + 2 = 6𝑢𝑛 +
2
2. In other words, 𝑢𝑛+1 = 6𝑢𝑛 − 𝑢𝑛−1 + 2 holds (see [Weisll]).
So, starting from 𝑣1 = 1 and 𝑣2 = 6 , one obtains 𝑣3 = 35, 𝑣4 = 204,
𝑣5 = 1189, 𝑣6 = 6930, 𝑣7 = 40391, etc. Starting from 𝑢1 = 1 and 𝑢2 = 8,
one obtains 𝑢3 = 49, 𝑢4 = 288, 𝑢5 = 1681, 𝑢6 = 9800, 𝑢7 = 57121, etc.
1.3.8. Now we can prove that the sequence of square triangular
numbers can be obtained by the following recurrent equation:
𝑆4,3 (𝑛) = 34𝑆4,3 (𝑛 − 1) − 𝑆4,3 (𝑛 − 2) + 2, 𝑆4,3 (1) = 1, 𝑆4,3 (2) = 36.
In fact, we 𝑆4,3 (𝑛) = 𝑣𝑛2 = (𝑎 − 𝑏)2 , 𝑆4,3 (𝑛 − 1) = 𝑣𝑛−1
2
= (3(𝑎 − 𝑏) −
2√2(𝑎 + 𝑏)) , 𝑆4,3 (𝑛 + 1) = 𝑣𝑛+1 = (3(𝑎 − 𝑏) + 2√2(𝑎 + 𝑏))2 . since
2 3
(3+2√2)𝑛 (3−2√3) 𝑛 1
𝑎𝑏 = , = , we obtain
4√2 4√2 32
1
In fact, let 𝑆3 (𝑢) = 𝑢(𝑢 + 1) = 𝑣 2 = 𝑆4 (𝑣). Then one gets
2
= 𝑧 2 𝑥 2 + (𝑥 + 1)2 = 𝑧 2 .
In fact, these pairs give all positive integer solutions of the equation
2
𝑥 + (𝑥 + 1) = 𝑧 2 . In order to prove it, let us consider a
2
solution (𝑥, 𝑧) of our equation with x > 3 and check that the pair (3𝑥 − 2𝑧 +
1,3𝑧 − 4𝑥 − 2) also gives a positive integer solution of this equation, but in
this case it holds 3𝑧 − 4𝑥 − 2 < 𝑧. So, we should check the inequalities
3𝑥 − 2𝑧 + 1 > 0 and 0 < 3𝑧 − 4𝑥 − 2 < 𝑧, or, which is the same, the
inequalities 2𝑧 < 3𝑥 + 1, 3𝑧 > 4𝑥 + 2, and 2z < 4x + 2. As 3𝑥 + 1 <
4𝑥 + 2 𝑓𝑜𝑟 𝑥 > 3, one should check just first two above inequalities.
Since 𝑧 2 = 2𝑥 2 + 2𝑥 + 1, so 4𝑧 2 = 8𝑥 2 + 8𝑥 + 1 < 9 𝑥 2 + 6𝑥 + 1 = 3𝑥 + 1)2 ,
and 2z < 3x + 1. Similarly, 9z2 = 18x2 + 18x + 9 > 16x2 + 16x + 4 = (4x +
2 ) 2 , and 3z > 4x+2. So, using the operation 𝑔(𝑥, 𝑧) = (3𝑥 − 2𝑧 + 1, 3𝑧 −
4𝑥 − 2), which delimitates the value z, we necessary come to the pair (3, 5);
it means that for some positive integer n, one gets g n (x , z) = (3, 5).
Furthermore, the operation 𝑓 (𝑥, 𝑧) = (3𝑥 + 2𝑧 + 1, 4𝑥 + 3𝑧 + 2) is
connected with the operation g by the formula 𝑓 ∙ 𝑔(𝑥, 𝑧) = 𝑓(3𝑥 − 2𝑧 +
1,3𝑧 − 4𝑥 − 2) = (𝑥, 𝑧), and, hence, we obtain, that, for any positive
integer k, it holds 𝑓 𝑘 ∙ 𝑔𝑘 (𝑥, 𝑧) = (𝑥, 𝑧). So, it is proven, that for any
positive integer solution (x , z ) of the equation
𝑥 2 + (𝑥 + 1)2 = 𝑧 2 . there exists some positive integer n (for which
𝑔¼ (𝑥, 𝑧) = (3, 5)), such that 𝑓 𝑛 (3, 5) = (𝑥, 𝑧). In other words, all positive
integer solutions of our equation belong to the following tree with the root
(3, 5):
(3,5) → (20,29) → (119,169) → (696,985) → ⋯
As it was proven above, any solution (𝑥𝑛 , 𝑧𝑛) of the equation 𝑥 2 +
1
(𝑥 + 1)2 = 𝑧 2 corresponds to a solution (𝑢𝑛 , 𝑣𝑛 ) of the equation 𝑢(𝑢 +
2
1
1) = 𝑣 2 with 𝑢𝑛 = 𝑧𝑛 − 𝑥𝑛 − 1 and 𝑣𝑛 = 2
(2𝑥𝑛 + 1 − 𝑧𝑛 ). So, 𝑢1 = 𝑧1 −
1 1
𝑥1 − 1 = 5 − 3 − 1 = 1, 𝑣1 = (2𝑥1 + 1 − 𝑧1 ) = (2 ∙ 3 + 1 − 5) = 1 , and
2 2
𝑢𝑛+1 = 𝑧𝑛+1 − 𝑥𝑛+1 − 1 = (4𝑥𝑛 + 3𝑧𝑛 + 2) − (3𝑥𝑛 + 2𝑧𝑛 + 1) = 𝑥𝑛 + 𝑧𝑛 +
1 1
1, 𝑣𝑛+1 = (2𝑥𝑛+1 + 1 − 𝑧𝑛 + 1) = (2(3𝑥𝑛 + 2𝑧𝑛 + 1) + 1 − (4𝑥𝑛 + 2𝑧𝑛 +
2 2
𝑧𝑛+1
2)) = 𝑥𝑛 + , Since the sequences 𝑥𝑛 and 𝑧𝑛 are increasing, so are the
,2
sequences 𝑢𝑛 and 𝑣𝑛 , and, therefore, we can get all positive integer
1
solutions of the equation 𝑢(𝑢 + 1) = 𝑣 2 in the form of an infinite sequence
2
(𝑢𝑛 , 𝑣𝑛 ), 𝑛 ∈ 𝑁 , depending on sequence (𝑥𝑛 , 𝑦𝑛 ), 𝑛 ∈ 𝑁:
𝑧𝑛 + 1
𝑢1 = 1 , 𝑣1 = 1, and 𝑢𝑛+1 = 𝑥𝑛 + 𝑧𝑛 , 𝑣𝑛+1 = 𝑥𝑛 + ,
2
Where
𝑥1 = 3, 𝑧1 = 5, and 𝑥𝑛 = 3𝑥𝑛−1 + 2𝑧𝑛−1 + 1,
𝑧𝑛 = 4𝑥𝑛−1 + 3𝑧𝑛−1 + 2 for 𝑛 ≥ 2.
So, we have proved, that there exist infinitely many triangular numbers
𝑆3 (𝑢𝑛 ), which are simultaneously the square numbers 𝑆4 (𝑣𝑛 ). Moreover,
we get a method to find indices 𝑢𝑛 and 𝑣𝑛 of such numbers, using the tree
(3,5) → (20,29) → (119,169) → (696,985) → ⋯ of all positive integer
solutions of the Diophantine equation 𝑥 2 + (𝑥 + 1)2 = 𝑧 2 . So, 𝑢1 = 1 and
𝑣1 = 1. Using the pair (𝑥1 , 𝑧1 ) = (3, 5), we obtain 𝑢2 = 3 + 5 = 8, and 𝑣2 =
5+1
3+ = 6.Using the pair (𝑥2 , 𝑧2 ) = (20, 29), we obtain𝑢3 = 20 + 29 =
2 29+1
49, and 𝑣3 = 20 + = 35, . Using the pair (𝑥3 , 𝑧3 ) = (119,169) we get
2 169+1
𝑢4 = 119 + 169 = 288, and 𝑣4 = 119 + = 3204. Using the pair
2
(𝑥4 , 𝑧4 ) = (696, 985), we get 𝑢 5 = 696 + 985 = 1681, and 𝑣5 = 696 +
985+1
= 1189, and so on (see also 4.3.1.).
2
1.3.11. The generating function for the sequence of the square tri-
𝑥(1+𝑥)
angular numbers has the form 𝑓(𝑥) = (1−𝑥)(1−34𝑥+𝑥 2 (see [SlPl95]).
)
More exactly, it holds
𝑥(1 + 𝑥)
= 𝑆4,3 (1)𝑥 + 𝑆4,3 (2)𝑥 2 + 𝑆4,3 (3)𝑥 3
(1 − 𝑥)(1 − 34𝑥 + 𝑥 2 )
So, the generating function for the sequence of the square triangular
numbers has the form
𝑓(𝑥) 𝑎0 + 𝑎1 𝑥 + 𝑎2 𝑥 2
= ,
𝑔(𝑥) 𝑏0 + 𝑏1 𝑥 + 𝑏2 𝑥 2 + 𝑏3 𝑥 3
where 𝑏0 = 1, 𝑏1 = −35, 𝑏2 = 35, 𝑏3 = −1 𝑎𝑛𝑑 𝑎0 = 𝑏0 𝑐0 = 1, 𝑎1 = 𝑏0 𝑐1 +
𝑏1 𝑐0 = 1.36 + (−35). 1 = 1, 𝑎2 = 𝑏0 𝑐2 + 𝑏1 𝑐1 + 𝑏2 𝑐0 = 1.1225 + (−35). 36 +
35.1 = 0.
The polynomial 𝑔(𝑥) = 1 − 35𝑥 + 35𝑥 2 − 𝑥 3 = (1 − 𝑥)(1 − 34𝑥 +
𝑥 2 ) has three real roots 𝑥1 = 1, 𝑥2 = 17 + 4√18, and 𝑥3 = 17 − 4√18. So,
one has min {|𝑥1 |, |𝑥2 |, |𝑥3 |} = 4√18 − 17 = 0.029 ….Hence, the generating
function for the sequence of the square triangular numbers has the form
1+𝑥
= 𝑆3,4 (1) + 𝑆3,4 (2)𝑥 + 𝑆3,4 (3)𝑥 3
(1 − 𝑥)(1 − 34𝑥 + 𝑥 2 )
+ ⋯ + 𝑆4,3 (𝑛 + 1)𝑥 𝑛 + ⋯ , |𝑥| < 4√18 − 17
1.4 Other highly polygonal numbers
Square triangular numbers are the most known highly polygonal
numbers, i.e., positive integers, which are polygonal in two or more ways.
However, there are many similar classes of such polygonal numbers (see
[Weis11]).
1.4.1. A pentagonal triangular number is a number which is
simultaneously pentagonal and triangular. Such numbers correspond
to the positive integer solutions of the Diophantine equation
1 1
𝑢(3𝑢 − 1) = 𝑣(𝑣 + 1).
2 2
Completing the square gives (6𝑢 − 1)2 − 3(2𝑣 + 1)2 = −2 . Substituting
𝑥 = 6𝑢 − 1 and 𝑦 = 2𝑣 + 1 gives the Pell-like Diophantine equation 𝑥 2 −
3𝑦 2 = −2 , which has positive integer solutions (𝑥, 𝑦) = (5,3), (19,11),
(71,41), (265,153), . .. (see, for example, [Nage51]). In terms of (𝑢, 𝑣), these
10 133
solutions give (𝑢, 𝑣) = (1,1), ( , 5), (12,20), ( , 76), (165,285), . . ., of
3 3
which the integer solutions are
(𝑢, 𝑣) = (1,1), (12, 20), (165, 285), (2296, 3976), (31977, 55385), . . .
(Sloane’s A046174 and A046175). They correspond to the pentagonal
triangular numbers (Sloane’s A014979)
1
𝑢(3𝑢 − 1) = 𝑣 2 .
2
1 1
𝑢(5𝑢 − 3) = 𝑣(𝑣 + 1).
2 2
Completing the square and rearranging gives (10𝑢 − 3)2 − 5(2𝑢 + 1)2 =
4. Substituting 𝑥 = 10u − 3 and 𝑦 = 2𝑣 + 1 gives the Pell-like equation
𝑥 2 − 5𝑦 2 = 4. It has the positive integer solutions (𝑥, 𝑦) = (3,1), (7, 3),
(18, 8), (47, 21), (322,144), ….The integer solutions in u and v are given by
1,55,121771, 5720653,12625478965, . . ..
1.4.8. An heptagonal square number is a number which is
simultaneously heptagonal and square. Such numbers correspond to the
positive integer solutions of the Diophantine equation
1
𝑢(5𝑢 − 3) = 𝑣 2 .
2
Completing the square and rearranging gives (10𝑢 − 3)2 − 40𝑣 2 = 9.
Substituting 𝑥 = 10𝑢 − 3 and 𝑦 = 2𝑣 gives the Pell-like equation 𝑥 2 −
10𝑦 2 = 9 . It has positive integer solutions (𝑥, 𝑦) =
(7, 2), (13, 4), (57,18), (253, 80), (487,154), . . ..The integer solutions in u and
v are given then by
(𝑢, 𝑣) = (1,1), (6, 9), (49, 77), (961,1519), (8214,12987), . . .
(Sloane’s A046195 and A046196). They correspond to the heptagonal
square numbers (Sloane’s A036354)
1, 81, 5929, 2307361, 168662169,…..