Journal of the Minnesota Academy of Science

Volume 58 Number 2 Article 4

1994

Some Conjectures Concerning Triangular Numbers

Bruce Brandt

Follow this and additional works at: https://digitalcommons.morris.umn.edu/jmas

Part of the Number Theory Commons

Recommended Citation
Brandt, B. (1994). Some Conjectures Concerning Triangular Numbers. Journal of the Minnesota Academy
of Science, Vol. 58 No.2, 21-25.
Retrieved from https://digitalcommons.morris.umn.edu/jmas/vol58/iss2/4

This Article is brought to you for free and open access by the Journals at University of Minnesota Morris Digital
Well. It has been accepted for inclusion in Journal of the Minnesota Academy of Science by an authorized editor of
University of Minnesota Morris Digital Well. For more information, please contact skulann@morris.umn.edu.
 Research Articles

BRUCE BRAND'li

ABsTRAcr
Strong empirical evidence supports conjectures that certain number panems always hold. These patterns
conce~ the func~on cr, defined by the equation cr(n) = n - m2, m 2 being the nearest square to n, on the
domam of the tnangular numbers. Triangular squares or triangular numbers of the form m2+m are also
~entioned in most of the conjectures. One of the conjectures, for example, is that the sum of cr over the
tnangular numbers up to a triangular square is 0. Some of these patterns can be described by strings of
symbols, such as "S" and "L," formed by first writing one symbol and then simultaneously substituting
"SLSLS" for "S" and "SLSLSLS" for "L," and continuing to make that substitution indefinitely.

Given positive n, let cr(n) be defined as n - m2 example, Beiler (1)); they are not given here because
where m 2 is the perfect square closest to n Although ~ they are only preliminary to this discussion.
seems too trivial to have any interesting properties, We now come to the patterns referred to at the
many patterns are found when we consider cr on the beginning of the paper:
triangular numbers, numbers of the form m(m +1)/2. {1) If n is a triangular square, the sum of cr(k) for
My conjectures (and one proven theorem), besides triangular k up to n equals 0.
involving cr and the triangular numbers, involve (2) If n is a triangular anti-square, the sum of cr(k)
triangular squares and "triangular anti-squares." The for triangular k up to n is 0, provided we count
former, of course, are triangular numbers of the form toward the sum half of o(n) itself.
m 2, and the latter are triangular numbers of the form 13} The sum of the negative values of cr over the
m 2 + m. Triangular anti-squares, in other words, are triangular numbers between two adjacent
triangular numbers which are also twice a triangular triangular squares is minus the triangular anti-
number. They are, roughly speaking, as far from a square between them.
square as numbers of that size can be. {4} The sum of the negative values of o over the
An interesting way of expressing these definitions is triangular numbers between two adjacent
that a triangular number is the sum of the first positive triangular anti-squares is minus the triangular
integers up to a certain point; a triangular square is, in square between them.
addition, the sum of the first positive odd integers up to 15} If n is a triangular square or a triangular anti-
a certain point; and a triangular anti-square (besides square, the number of positive values of o over
being a triangular number) is the sum of the first triangular numbers up to and including n
positive even integers up to a certain point. equals the number of negative values.
The formula for a triangular square reduces to a (In view of Ill and (21, 13} and (41 also imply
well-known Pell equation which has an infinite number analogous facts about the positive values of o . It is
of solutions. So there are an infinite number of somewhat easier to state (41 the way I did because if n
triangular squares. Similarly, there are an infinite is a triangular anti-square a(n) is positive.) You can
number of triangular anti-squares. In fact a number is a verify the first instances of ll}-{51 by looking at Table 1.
triangular anti-square if and only if it is the geometric I have verified {11 - {51 by computer for the first
mean of two adjacent triangular squares. The first few eight triangular squares (Table 2) and the first eight
triangular squares are 1, 36, 1225, 41616. The first few triangular anti-squares (Table 3).
triangular anti-squares are 6, 210, 7140. But this isn't all. We fmd more patterns when we
From the definition of a triangular anti-square, it is consider the "neighborhoods" of triangular squares and
easy to see that cr of a triangular anti-square is always triangular anti-squares . The neighborhood of a
positive. Another fact that we will need is that cr of triangular number is defined as the three triangular
such a number is always even. (Of course, cr on a numbers before it , itself, and the three triangular
triangular square is zero.) numbers after it. All statements about neighborhoods
The proofs of these statements should be easy to are only claimed to be true for positive triangular
anyone who has studied Pell equations (see, for squares and anti-squares.

tindependentcontribution
; 13 27th Avenue S.E., Minneapolis, MN 55414-3101.

Vol. 57, No. 2, 1993 21

Research Articles
(However, the elements of the neighborhood can
Table 1. The First Few Values of cr on the be zero. We are now defining the triangular number 1n
Triangular Numbers. as n(n+l)/2, for any integer n, and considering the
domain of cr to be all triangular numbers.) If 1n is a
triangular square, then:
k cr(k) cr(tn_ 1) + crCtn+l) = -1,
triangular square: 1 0 cr(tn_2) + crC1n+ 2) = 2,
3 -1 cr(tn-Y + crCtn+Y = 1.
triangular anti-square: 6 2 On the other hand, if 1n is a triangular anti-square
10 1 then:
15 -1
cr(tn-1) + cr(tn+1) = 0,
21 -4
28 cr(tn-2) + crC1n+2) = -1,
3
triangular square: 36 0 cr(tn-Y + crCtn+Y = -4.
45 -4 Furthermore, if 1n is a triangular square,
55 6 161 crC1n+1) + crC1n+2) = °Ctn+3),
66 2 cr(tn-1) + cr(tn-2) = cr(tn_y,
78 -3 and if 1n is a triangular anti-square,
91 -9
105 5 cr(tn) + 2cr(tn-1) = cr(tn-2)
Notice that the five equations concerning the
120 -1
neighborhoods of triangular squares are reduncbnt.
136 -8
I will prove {6}. Let the triangullr square be
153 9
expressible as x(x+ 1)/2 or as y2, where x and y are
171 2
positive. The next three triangular numbers are
190 -6
(x+l)(x+2)/2; (x+2)(x+3)/ 2; and (x•.3><x•4l 2. First I
triangular anti-square: 210 14
will show that the squares closest to those three
numbers are, respectively, (y+1>2, (y+11 2 anJ tv•2>2.
I must show:
Table 2. The First Eight Triangular Squares. (y+1) 2 - (y+l) < (x+l)(x+ 2). 2 ~ ( v• l , 2 • (y+ 1),
(y+1) 2 - (y+ 1) < (x+2)(x+3J:'2 .5. <v• l ,2 • ty+ 1),
1
(y+ 2)2 - (y+ 2) < (x+3)(x+4)/ 2 ~ <,· • 2 ,2 • <y+ 2).
36
1225 Because x(x+l)/2 = y 2 , those inequ3liues c3n be
41616 written as:
1413721 (y+ 1) 2 - (y+ 1) < y2 + (x+ 1) .5. (y+ 1 ,2 • ( y+ 1),
48024900 (y+1)2 - (y+l) < y 2 + (2x+3) .5. <y• l ,2 • ty+l),
1631432881 (y+2)2 - (y+2) < y2 + (3x+6) .5. ty•2 ,2 • (y+2).
55420693056
Algebraic manipulation yields:
y < x+l 5 3y+2,
y < 2x+3 5 3y+2,
3y+ 2 < 3x+6 5 5y+6.
First I prove the left-hand inequalities:
Table 3. The First Eight Triangular Anti-Squares. y > 0 ⇒ 2y 2: (y+ 1)
⇒ y•2y 2: y(y+l)
6
:::) X (x+ 1) ,2: y(y+ 1)
210
7140 ⇒ X 2: Y,

242556 from which they all follow.

8239770 Next, the right-hand inequalities:
279909630 x > 0 ⇒ x 2 < x (x+ 1) = 2y2
9508687656
:::) X < 'i2.y
323015470680
⇒ 2x+3 < 2 'i2.y + 3
< 3y + 3
⇒ 2x+3 5 3y+2.

22 Journal of the Minnesota Academy of Science

 Research Articles

The fact that x < ...fi.y quickly yields the other rn:o m2+m, m ~ 0, then 171 equals ak except when it equals
right-hand inequalities as well. sk+2m.
Now that I have found the nearest squares to the By considering the way the definitions of sk and ak
three triangular numbers, I will prove the equation ought to be extrapolated for k!S0, one finds that so=0
involving o. We are trying to show but ao is undefined, and of course s_k = sk and a_k=ak-
(x+l) (x+2)/2 - (y+1) 2 + (x+2) (x+3)/2 - (y+1)2 Thus when 1n is a triangular square k = 0 satisfies {8},
= (x+3) (x+4 )/2 - (y+2 )2. and is not an "exception," but we will consider k = O
an exception to 19) when 1n is a triangular anti-square.
The equation reduces algebraically to (x2+x)/2 = y2,
Given tn, the values of k which are exceptions to {8}
which is part of our hypotheses.
or 191, depending on whether 1n is a triangular square
Not all equations concerning neighborhoods of
or a triangular anti-square, form an interesting pattern.
triangular numbers are linear. We also have, for tn
Here is a longer list of the exceptions for anti-square
triangular square:
tn=0:
O<tn-2) O<tn+2) = 2O <tn-1) O<tn+1) 0,3, 7, 10, 14, 17,20,24,27,31,34,38,41, 44, 48,51, 55,58, ...
= -4(next lower triangular anti-square), The sequence of "short" gaps between adjacent
O(tn_3) o(tn+3) = -2 (next lower triangular exceptions (gaps of 3) and "long" gaps (gaps of 4) is:
square). SLSLS, SLSLSLS, Sl.Sl.S, SLSI.SLS, Sl.SLS;
SLSLS, SLSLSLS, Sl.Sl.S, SLSI.Sl.S, Sl.SLS, SLSLSI.S, SLSLS;
If 1n is a triangular anti-square, we have:
SLSLS, SLSLSLS, Sl.SLS, SLSLSLS, Sl.SLS;
o<tn-1) OCtn+1) = (next lower triangular square), SLSLS, SLSLSLS, SLSLS, SLSI.SLS, Sl.SLS, SLSLSI.S, SLSLS;
o<tn-2) OCtn+2) = (next lower triangular anti- SLSLS, SLSLSLS, SLSLS, SLSLSLS, SLSLS; ...
square). This sequence of short and long gaps is the same
But that's not all. Not only for k = 1,2,3, but for all for any triangular anti-square (although the length of
k - with certain exceptions I will make precise - the the short and long gaps depends on the triangular anti-
expression square). To derive the pattern, first write an "S." Then
171 o(tn_IJ + oCtn+k) simultaneously replace every "S" with an "SLSLS" and
is fixed, for any triangular square tn, or for any every "L" with an "SLSLSLS". Then repeat this
triangular anti-square ln· replacement indefinitely. Similarly, for triangular
For k~l, define· sk as follows: Pick a triangular squares, such as tn=l, there are short and long gaps
square 1n such that n-k is positive; we have between adjacent exceptions. In this case it is best to
181 sk = o(tn_IJ + oCtn+k). look at negative as well as positive values of k:
Then sk is well-defined! (The exceptions I spoke of ... ,-53,-46,-36,-29,-22,-12,-5,5, 12,22,29,36,46,53, ...
occur when n-k is not positive.) Here the gaps of 7 and 10 form this pattern:
In the same way, let ak be defined as ... ; SLSLS, SLSI.SLS, Sl.SLS, Sl.SLSI.S, Sl.SLS;
191 ak = o(tn-k) + oCtn+k) SLSLS, SLSLSLS, Sl.SLS, SLSI.SLS, Sl.SLS, SLSI.SLS, SLSLS;
SLSLS, SLSLSLS, SLSLS, SLSI.SLS, Sl.SLS;
where tn is a triangular anti-square and k~l and
SLSLS, SLSLSLS, Sl.SLS, SLS(L)SLS, SLSLS, SLSLSLS, SLSLS;
n-k > 0.
SLSI.S, SLSI.SLS, Sl.SLS, SLSLSLS, Sl.Sl.S;
Even when n-k is nonpositive, 181 and 19) usually
SLSLS, SLSLSLS, Sl.SLS, SLSLSI.S, Sl.SLS, Sl.SLSI.S, SLSLS;
hold. For example, when tn=l (so n=l, and 18) is never
SLSLS, Sl.SLSLS, Sl.Sl.S, SLSLSLS, Sl.Sl.S; ...
claimed to be true), there are only 243 values of k from
1 to 2000 where 18) doesn't hold. In all those cases the The "L" in parentheses is the gap between -5 and 5.
For both squares and anti-squares, long gaps are of
expression 17) equals ak - 2. '
even length and short gaps are of odd length. A
What happens when ~ = 0? Then, being of the form
number is in the middle of a long gap between
m2 and of the form m2 + m, it would like to satisfy
exceptions for a triangular square (anti-square) if and
both 18) and 19}. Since we never have ak = sk (in fact ak
only if it is an exception for the next higher triangular
and sk are always of opposite parity), that is
anti-square (square). (We should consider the triangular
impossible, but for each k it is always one or the other.
anti-square 0 to be "higher" than the triangular square
The majority of the time, 191 holds; but for
0, which makes sense anyway, in order for triangular
k=3,7,10,14,17,20, ... , and in general, every third or
squares and triangular anti-squares to alternate and so
fourth value, 18) holds. (Notice that whether we
that triangular anti-squares are always the geometric
consider these k, or the complementary set, to be
mean between the "next higher" and "next lower"
"exceptions" depends on whether we consider 0 to be
triangular squares.)
a triangular square or a triangular anti-square.)
By applying this fact twice, one finds that a number
When ln = 6, {7} equals ak, exceRt when it equals
is an exception for a triangular square (anti-square),
sk+4. In general, if 1n is of the form m 2, m~, 17) equals
surrounded by short gaps, if and only if it is an
sk except when it equals ak-2m. If 1n is of the form

Vol. 57, No. 2, 1993 23

Research Articles

exception for the next higher triangular square (anti- and crC1n+1) > 0. What are the values of crCtn+l) - cr(tn)
s quare). Some final patterns with regard to the for such n, in order? They are 3,7,10,14,17,20, ...
exceptions to {81 or 191: Two exceptions for a positive Since cr(tn+1)-cr(tn) equals, for such n, 1n+l - tn
triangular square, added or subtracted, result in an (remember the definition of a), which equals n+l,
exception for the next lower triangular anti-square. An perhaps all we have done is find a pattern in the
exception for a positive triangular anti-square, plus or sequence of positive and negative values of a.
minus an exception for the next lower triangular But there is more to it than this; look at those tn
square, yields an exception for that triangular square. such that cr(tn) ~ 0 but crCtn+ 1) < 0. The values of cr(tn)
(One must say "positive"; this second rule breaks down - crC1n+1), in order, are then 1,2,4,5,6,8,9,11,12, ... , the
when both the triangular anti-square and the triangular complementary set.
square are 0.) Nevertheless, there is a pattern in the signs of a.
The sequences sk and ak are interesting in Here are the signs of the values of a between 1 and 36:
themselves too: -++--+
If we write an isolated sign as "l" and two of the
same sign together as "2", we get
Table 4: The first values of sk and ak
1221.
Similarly, the signs of cr between 36 and 1225 can
Sk ak be written
1 -1 0 122122122212212221221221.
2 2 -1 What is the pattern? This algorithm will give the 1-2
3 1 -4 pattern between the nth positive triangular square and
4 -2 3 the (n+ 1)-st:
5 -7 0 1. Start with a "l".
6 4 -5 2. Repeat n-1 times the sequence of first 2a then
7 -1 8 2b:
8 -8 3 a. Simultaneously replace every "l" with a
9 9 -4 "21" and every "2" with a "211." Then add
10 2 -13 a final "2."
11 -7 8 b. Simultaneously replace every "l" with a
12 16 -1 "21" and every "2" with a "211." Then
13 7 -12 delete the initial "2".
14 -4 15 3. Simultaneously replace every "l" with a "122"
15 -17 4 and every "2" with a "1222". Then add a final
16 14 -9 "l ".
17 1 -24 In the case of the signs from 36 to 1225, the
18 -14 11 algorithm yields at each step:
19 23 -4 1. 1
20 8 -21 2a. 2 1 2
21 -9 20 2b. 1 1 2 1 2 1 1
22 -28 3 3. 122122122212212221221221
I have verified the validity of this algorithm for the
Notice that sk always has the same parity as k, next higher case too, the signs of a between 1225 and
whereas ak always has the opposite parity. If you use a 41616.
computer to generate a more extensive table, you will So far we have been considering the triangular
find many patterns, especially if you are familiar with numbers and the nearest squares to them. What
Pell equations. (Take the numbers 1 1·2 3·5 7· 1
happens if we consider the squares and the nearest
12, 17;29,41. The first member of each pair is th'e ~u~• of triangular numbers to them? At first glance, this will not
the previous two numbers; the second member work because a number can be exactly in between two
completes a three-number arithmetic sequence. Now adjacent triangular numbers. But such a number must
look at the solutions to sk = ±.1, ak = O or ak = -1. With be of the form m 2/2, which cannot be a square.
a larger table you also see patterns in the squares or In other words, define 't(n) as n - m(m+ 1)/2, where
products of the numbers in the list of pairs.) m(m+ 1)/2 is the nearest triangular number to n. Table 5
There are more patterns concerning cr on the lists the first few values of t on the domain of the
triangula: numbers. Look, on Table 1, at those tn, not squares.
necessanly squares or anti-squares, such that cr(tn) < o First of all, for those n such that 't(n 2) > 0 and
't((n+1) 2) < 0, what are the values of 't(n2) - 't((n+l)2)?

24
Journal of the Mtnnesota Academy of Sctence
 Research Articles
Secondly, and more interestingly, what panem can
Table 5: The first values of -r on the Squares be found in the parity of -r(n2)? Considering now all
integers n, not just positive n, and writing "O" for odd
n n2 -r(n2) parity and "E" for even parity, we have,
1 1 0 ... EEEEOOOEEEEOOOEEEOOOOE(E)EOOOOEEE
2 4 1 OOOEEEEOOOEEEE .. . ,
3 9 -1 where the "E" in parentheses corresponds to n=0.
4 16 1 The even and odd parities come in clumps of 3 or
5 25 -3 4. Writing "3" for a clump of three "O" 's or "E" 's, and
6 36 0 "4" for a clump of four, and continuing to write the
7 49 4 structure centered around 0 in parentheses, we have,
8 64 -2 ... 434334(3)433434 .. .
9 81 3 I have verified for a much larger number of values
10 100 -5 of -r(n 2) that this pattern can be generated in the
11 121 1 following way: First write the central "3" . Then
12 144 8 simultaneously replace each "3" with "34343" and each
13 169 -2 "4" with "3434343". (At each stage the "3" in the exact
14 196 6 middle corresponds to n=0.) Continue the process
15 225 -6 indefinitely.
16 256 3
17 289 -11 REFERENCE
18 324 -1 1. Beiler, Alben H. 1964. Recreations in the Theory of
19 361 10 Numbers: The Queen of Mathematics Entertains. Dover
20 400 -6 Publications, Inc. New York, NY 10014. 349 pp.
21 441 6
22 484 -12

They are, in order, 2,4,6,8, 10, 12,14, 16, 18, . .. ; this set is,
of course, the positive even numbers.

Vol. 57, No. 2, 1993 25