The Square Pyramid Puzzle

Author(s): W. S. Anglin
Source: The American Mathematical Monthly, Vol. 97, No. 2 (Feb., 1990), pp. 120-124
Published by: Taylor & Francis, Ltd. on behalf of the Mathematical Association of America
Stable URL: https://www.jstor.org/stable/2323911
Accessed: 23-12-2024 05:00 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Taylor & Francis, Ltd., Mathematical Association of America are collaborating with
JSTOR to digitize, preserve and extend access to The American Mathematical Monthly

This content downloaded from 49.204.84.210 on Mon, 23 Dec 2024 05:00:09 UTC
All use subject to https://about.jstor.org/terms
The Square Pyramid Puzzle

W. S. ANGLIN, University of Notre Dame, Notre Dame, IN 46556

BILL ANGLIN received an interdisciplinary Ph.D. in Mathematics and Philos-

ophy at McGill University in 1987. He is currently a Postdoctoral Fellow in
Philosophy of Religion at Notre Dame University, and he is also continuing
his work in classical Number Theory.

In 1875 Edouard Lucas challenged the readers of the Nouvelles Annales de

Mathematiques to prove the following:
A square pyramid of cannon-balls contains a square number of cannon-balls
only when it has 24 cannon-balls along its base [4].

In other words, the only nontrivial, natural number solution of

12 + 22+ 32 + _.. +X2 = y2

is x = 24 and y = 70.
We begin this article by recalling the history of this Diophantine equation. We
then present a new, elementary solution of it.
In 1876 Moret-Blanc gave a "proof" of Lucas's assertion which, although
incomplete, at least gives a good framework for solving it [7]. Moret-Blanc divided
the problem into two cases, one with x even and one with x odd. This is exactly the
approach we take in this paper. In 1877, having pointed out that there was a gap in
Moret-Blanc's argument, Lucas published a proof which also contained a gap
[5]. Lucas could handle the case with x even but not the case with x odd. In 1918
G. N. Watson managed to fill Lucas's gap with about 14 pages of a specially
extended theory of Jacobian elliptic functions [9]. This was the first complete proof
of Lucas's assertion but it was very complicated. In 1952 Ljunggren used Skolem's
method to give a simpler, more arithmetic solution to the square pyramid problem
[3]. In 1966 Baker and Davenport used a theorem from transcendental number
theory to solve the simultaneous Diophantine equations 3X2 - 2 = y2 and 8x - 7
= Z2 [1]. Their method, recently made more practical by Waldschmidt, can easily be
adapted to demonstrate Lucas's proposition. In 1975 Kanagasabapathy and
Ponnudurai gave an elementary solution for the simultaneous Diophantine equa-
tions 3x2 - 2 = y2 and 8x2 - 7 = z2 [2], and, in 1985, De Gang Ma used the ideas
of Kanagasabapathy and Ponnudurai to arrive at an elementary solution to Lucas's
problem [6]. This was the first complete solution accessible at the undergraduate
level.
In this paper we draw on one of Ma's ideas to give a new proof of Lucas's
assertion. This proof is so short and simple that it is surprising that it escaped
mathematicians for over a hundred years.

120

This content downloaded from 49.204.84.210 on Mon, 23 Dec 2024 05:00:09 UTC
All use subject to https://about.jstor.org/terms
 THE SQUARE PYRAMID PUZZLE 121

Our equation is equivalent to x(x + 1)(2x + 1) = 6y2. In this section we solve it

assuming that x is even, and, in the next section, we solve it assuming that x is odd.
Suppose, then, that x is even. Under this assumption we can solve the equation with
the help of the following three lemmas which were probably all known to Fermat.

LEMMA 1. The area of a Pythagorean triangle is never a square.

Proof Suppose, on the contrary, there are Pythagorean triangles with square
areas. Let w2 be the smallest area for which such a triangle exists. Let x and y be
the legs of a Pythagorean triangle with area w2. Then X2 + y2 = Z2 for some integer
z, and xy/2 = w 2. Since w is minimal, x and y are relatively prime and, without
loss of generality, we may take x odd and y even. (It follows by a congruence
consideration modulo 4 that x and y are not both odd.) By the well-known theorem
for Pythagorean triangles, there are relatively prime positive integers r and s of
different parity such that x = r2 _ 52 and y = 2rs. Hence (r2 - s2)rs = w2 and
s/4 < s < w2. Since r, s, r - s and r + s are pairwise relatively prime, and since
(r - s)(r + s)rs = W2, it follows that there are positive integers a, b, c and d such
that r = a2, s = b2, a _ b2 = r -S = c2 and a2 + b2 = r + s = d2. Note that c
and d are relatively prime since r - s and r + s are relatively prime. Noting also
that c and d are odd (because r and s have different parity), let X= (c + d)/2
and Y = (d - c)/2. Then X and Y are relatively prime and X2 + Y2 = a2. Hence
one of X and Y is even, and XY/2 = (d2 - c2)/8 = b2/4 = s/4 is a square
integer. Since the triangle with sides X, Y and a is a Pythagorean triangle with
square area s/4, it follows from the minimality of w2 that w2 < s/4. Contradic-
tion.

LEMMA 2. There are no positive integers x such that 2x4 + 1 is a square.

Proof. To obtain a contradiction, suppose that (x, y) is the least positive integer
solution of 2x4 + 1 = y2. Then for some positive integer s, y = 2s + 1 and
x4 = 2s(s + 1). If s is odd then s and 2(s + 1) are relatively prime and, for some
integers u and v, s u4 and 2(s + 1) = v4. This gives 2(u4 + 1) = v4 with u odd
and v even. Hence we have 2(1 + 1) 0 (mod 8). Since this is impossible, s cannot
be odd. Since s is even, 2s and s + 1 are relatively prime, and there are integers u
and v, both greater than 1, such that 2s = u4 and s + 1 = v4. Let w be the positive
integer such that u = 2w. Let a be the positive integer such that V2 = 2a + 1. Then
u4/2 + 1 = s +1 = v4 so that 2w4 = (v4 -1)/4 = a(a + 1). Since V2 = 2a + 1,
it follows from congruence considerations modulo 4 that a is even. Since 2w4=
a(a + 1), it follows that there are positive integers b and c such that a = 2b4 and
a + 1 = c4. However, this implies that 2b4 + 1 = (c2)2 and hence y < c2 (by the
minimality of (x, y)). On the other hand, c2 < a + 1 < V2 < S + 1 < y. Contradic-
tion.

LEMMA 3. There is exactly one positive integer x, namely, 1, such that 8x4 + 1 is a
square.

Proof Suppose 8x4 + 1 = (2s + 1)2. Then 2x4 = s(s + 1). If s is even then
there are integers u and v such that s = 2u4 and s + 1 = v4. In that case,
2u4 + 1 = s + 1 = v4 and, by Lemma 2, u = 0 and, hence, x = 0. If s is odd

This content downloaded from 49.204.84.210 on Mon, 23 Dec 2024 05:00:09 UTC
All use subject to https://about.jstor.org/terms
122 W. S. ANGLIN [February
then there are integers u and v such that s = u4 and s + 1 = 2v4. In
that case, u4 + 1 = 2 V4. Since u is odd, a congruence consideration modulo 4
shows that v is odd. Squaring both sides of U4 + 1 = 2V4, we obtain 4v8 - 4U4 =
U8 - 2U4 + 1 and hence (V4 - U2)(V4 + U2) = ((U4 - 1)/2)2, an integer square.
Since V4 and u2 are relatively prime, it follows that both (V4 - u2)/2 and
(V4 + u2)/2 are integer squares. Now (v2 - U)2 + (V2 + U)2 = 4(V4 + U2)/2 = A2
and (v2 - u)(V2 + u)/2 = (V4 - u2)/2 = B2. By Lemma 1, this is impossible unless
v2= +u. Since u4 + 1 = 2V4, we obtain u4 - 2u2 + 1 = 0 and u2 = 1. From this
it follows that s = 1 and x = ? 1.
With the above three lemmas, we are now in a position to solve x(x + 1)
(2x + 1) = 6y2 under the assumption that x is even. Suppose, then, that x is even.
Then x + 1 is odd. Since x, x + 1 and 2x + 1 are relatively prime in pairs, it
follows that x + 1 and 2x + 1 (both being odd) are either squares or triples of
squares. Thus x + 1 # 2 (mod 3) and 2x + 1 # 2 (mod 3). Hence x 0 (mod 3),
and for some nonnegative integers p, q and r, we have x = 6q2, x + 1 -p2 and
2x + 1 = r2. Thus 6q2 = (r - p)(r + p). Since p and r are both odd, 4 is a factor
of (r - p)(r + p) = 6q2 and thus q is even. Let q' be the integer such that q = 2q'.
We now have 6q'2 = ((r - p)/2)((r + p)/2) and, since (r - p)/2 and (r + p)/2
are relatively prime (because r2 and p2 are relatively prime), we obtain one of the
following two cases.

Case (i). One of (r - p)/2 and (r + p)/2 has the form 6A2 and the other has
the form B2 (where A and B are nonnegative integers). Then p = ?(6A2 - B2)
and q = 2AB. Since 6q2 + 1 = x + 1 = p2, we have 24A2B2 + 1 (6A2 -B2)2 or
(6A2 - 3B2)2 - 8B4 = 1. By Lemma 3, B = 0 or 1 and, hence, x 6q2 -0 or 24.
The only nontrivial solution is thus with x = 24.

Case (ii). One of (r - p)/2 and (r + p)/2 has the form 3A2 and the other has
the form 2B2 (where A and B are nonnegative integers). Then p = (3A2 - 2B2)
and q = 2AB. This gives 24A2B2 + 1 = (3A2 - 2B2)2 and hence (3A2 - 6B2)2
2(2B)4 = 1. By Lemma 2, B = 0 and hence x = 6q2 = 0.
Thus when x is even, the only solution to Lucas's puzzle is x = 24 cannonballs
along the base of the square pyramid.

II

We have solved Lucas's problem under the assumption that x is even. In this
section we solve it under the assumption that x is odd. To do this, we first
investigate the solutions of the Diophantine equation X2 - 3y2 = 1.
Let a = 2 + VY and b = 2 - V3. Note that ab = 1. Where n is any nonnegative
integer, let un = (a' + bh)/2 and v, = (a' - b')/(2V3). Then u,, and vu are
integers, and it is a well-known result (from the theory of the Pell equation) that
when n > 1, (u,,, v") is the nth positive integer solution of X2 - 3y2 = 1. (See [8,
pp. 209-211].) Of course, when n = 0, we have the solution X = 1 and Y = 0.
In order to show that x = 1 is the only odd positive integer such that
x(x + 1)(2x + 1) has the form 6y2, we use the following lemmas.

LEMMA 4. Where m and n are nonnegative integers, Unn = UmUn + 3Vmvn and
Vm+n = UmVn + UnVn. Also if m-n > 0, then U.-n = UmUn -3vmvn and vm__
-UmVn + UnUm-

This content downloaded from 49.204.84.210 on Mon, 23 Dec 2024 05:00:09 UTC
All use subject to https://about.jstor.org/terms
1990] THE SQUARE PYRAMID PUZZLE 123

Proof. This follows by straight calculation from the definitions of u, and v".

Using Lemma 4, it is not hard to obtain the following result.

LEMMA 5. Where m is a nonnegative integer, um+2 = 2Uium+l - Um and Vm+2 =

2 u,v,, +1- v,n*

Using the fact that (um, vm) is a solution of X2 - 3y2 = 1, we also have the
following.

LEMMA 6. Where m is a nonnegative integer, U2r= 2u2 - = 6v2 + 1 and

V2m = 2u n,Vn,

LEMMA 7. Let m, n and r be nonnegative integers such that 2rm - n is nonnega-

tive. Then u2rn+ ? = ( -l) rU, (mod urm).

Proof. Using mathematical induction on r together with Lemma 4, we can show

that u(2r+1)ni 0 (mod um) and V2rm 0 (mod um). Since, by Lemma 6, U2rm -
2urm - 1 = 6vr2n, + 1, it follows that U2rmr(1)r (mod Ur). Thus U2rrnn -
U2rmu,, ? 3V2rmV,, (=1)rUn (mod um) (using Lemma 4).

Let us consider the first few values of un. Starting with n = 0, we have 1, 2, 7, 26,
97, 362, and so on. If we consider these values modulo 5, we have 1, 2, 2, 1, 2, 25 ....
By Lemma 5 it follows that this sequence is periodic. If we consider the values of un
modulo 8, we obtain 1, 2, 7, 2, 1, 2,.... By Lemma 5, this is a purely periodic
sequence with period length 4. Note that when n is even, un is odd. Using the laws
of quadratic reciprocity, the above comments lead us to the following two lemmas.

LEMMA 8. If n is even then un is an odd nonmultiple of 5 and 2- = 1 iff n is a

multiple of 3.

LEMMA 9. If n is even then un is odd and ( I)=1 iff n is a multiple of 4.

The following and final lemma was first proved by Ma.

LEMMA 10. Where n is a nonnegative integer, un has the form 4M2 + 3 only when
un = 7.

Proof. Suppose un = 4M2 + 3. Then un 3 or 7 (mod 8) and, from the se-

quence of values of u,, modulo 8, it follows that n has the form 8k + 2. Suppose
n # 2 (and hence un 0 7). Then we can write n in the form 2k'2s + 2 where k' is
odd and s > 2. By Lemma 7, un = U2k'2s + 2 -(-)k'U2 (mod u2.). Since k' is odd
and u2 = 7, it follows that 4M2 = u - 3 - 10 (mod u2s) and, hence,

( Z21)() = (410)2= (4M2) 1

By Lemma 9 it follows that the first factor on the left is 1. By Lemma 8 it follows
that the second factor on the left is - 1. Since this is impossible, we may conclude
that n = 2 and hence u,, = 7.

This content downloaded from 49.204.84.210 on Mon, 23 Dec 2024 05:00:09 UTC
All use subject to https://about.jstor.org/terms
124 W. S. ANGLIN
Suppose now that x is an odd positive integer and x(x + 1)(2x + 1) = 6y2 for
some integer y. Since x, x + 1 and 2x + 1 are relatively prime in pairs, it follows
that x is either a square or a triple of a square, and hence x X 2 (mod 3). Moreover,
x + 1, being even, is either double a square or six times a square, and, hence,
x + 1 # 1 (mod 3). Thus x 1 (mod 3) and hence x + 1 2 (mod 3) and 2x + 1
-0 (mod 3). Thus for some nonnegative integers u, v and w, we have x =
u2, x += 2v2 and 2x+= 3w2. From this we obtain 6w2 +1 =4x+ 3=
4u2 + 3. Also (6w2 + 1)2 - 3(4vw)2 = 12w2(3w2 + 1- 4v2) + 1 = 12w2(2x + 1
+? - 2(x + 1)) + 1 = 1. Hence, by Lemma 10, 6w2 + 1 = 7. Thus w = 1 and
x = 1. This gives us the trivial 1 cannonball solution to Lucas's problem.
We may conclude that if a square number of cannonballs are stacked in a square
pyramid then there are exactly 4900 of them.

I would like to thank Professor J. Lambek for his kind help and encouragement.

REFERENCES

1. A. Baker and H. Davenport, The Equations 3X2 - 2 = y2 and 8x2 - 7 = Z2, Quarterly Journal of
Mathematics, ser. 2, 20 (1969) 129-37.
2. P. Kanagasabapathy and T. Ponnudurai, The simultaneous Diophantine equations y2 - 3x2 = -2
and z2 - 8x2 = - 7, Quarterly Journal of Mathematics, ser. 2, 26 (1975) 275-78.
3. W. Ljunggren, New solution of a problem proposed by E. Lucas, Norsk Mat. Tidsskrift, 34 (1952)
65-72.
4. Edouard Lucas, Question 1180, Nouvelles Annales de MathImatiques, ser. 2, 14 (1875) 336.
5. - , Question 1180, Nouvelles Annales de MathImatiques, ser. 2, 16 (1877) 429-32.
6. De Gang Ma, An elementary proof of the solution to the Diophantine equation 6y2 = x(x + 1)
(2x + 1), Sichuan Daxue Xuebao, 4 (1985) 107-16.
7. M. Moret-Blanc, Question 1180, Nouvelles Annales de MatheImatiques, ser. 2, 15 (1876) 46-48.
8. I. Niven and H. S. Zuckerman, An Introduction to the Theory of Numbers, 4th ed., John Wiley and
Sons, New York, 1980.
9. G. N. Watson, The problem of the square pyramid, Messenger of Mathematics, 48 (1918-19) 1-22.

This content downloaded from 49.204.84.210 on Mon, 23 Dec 2024 05:00:09 UTC
All use subject to https://about.jstor.org/terms