Olympiad Maths Prep

Track / Stage 9 / 38 of 80 #1918 of 2000

Problem 1918

IMO P2/P5; hard shortlist
Geometry Difficulty 9.1 Prove it USA IMO · United States

Let A=(0,0,0)A = (0,0,0) be the origin in the three dimensional coordinate space. The *weight* of a point is the sum of the absolute values of its coordinates. A point is a *primitive lattice point* if all its coordinates are integers with their greatest common divisor equal to 1. A square ABCDABCD is called a *unbalanced primitive integer square* if it has integer side length and the points BB and DD are primitive lattice points with different weights.
Show that there are infinitely many unbalanced primitive integer squares ABiCiDiAB_iC_iD_i such that the plane containing the squares are not parallel to each other.

This one wants a proof. Work it on paper, read the official solution, then mark yourself honestly — the ladder only means something if the record is true.

Official solution

First Solution: Let (a,b,c)(a, b, c) be a Pythagorean Triple, that is, a,b,ca, b, c are positive integers with a2+b2=c2a^2 + b^2 = c^2. The key facts are
(i) for any positive integers mm and nn such that m>nm > n, (m2n2,2mn,m2+n2)(m^2 - n^2, 2mn, m^2 + n^2) is a Pythagorean triple; and
(ii) the vector [ca,cb,a+bc][c-a, c-b, a+b-c] has integer length 2cab2c-a-b.

To establish (ii), we compute
(ca)2+(cb)2+(a+bc)2(2cab)2=2c2+a2+b22c(a+b)+(a+b)22c(a+b)+c24c2+4c(a+b)(a+b)2=a2+b2c2=0. \begin{align*} & (c-a)^2 + (c-b)^2 + (a+b-c)^2 - (2c-a-b)^2 \\ &= 2c^2 + a^2 + b^2 - 2c(a+b) + (a+b)^2 - 2c(a+b) + c^2 \\ & \quad -4c^2 + 4c(a+b) - (a+b)^2 \\ &= a^2 + b^2 - c^2 = 0. \end{align*}
Putting these two facts together shows that the vector [2n2,(mn)2,2n(mn)][2n^2, (m-n)^2, 2n(m-n)] has integer length. For i=1,2,i = 1, 2, \dots, let oi=2i+1o_i = 2i + 1. We set n=1n = 1 and mn=oim - n = o_i, and define vectors
ABi=ui=[2,2oi,oi2],ADi=vi=[2oi,oi22,2oi], \overrightarrow{AB_i} = \mathbf{u}_i = [2, 2o_i, o_i^2], \quad \overrightarrow{AD_i} = \mathbf{v}_i = [2o_i, o_i^2 - 2, -2o_i],
and
wi=[oi2,2oi,2]. \mathbf{w}_i = [-o_i^2, 2o_i, -2].
It is not difficult to check that 22+(2oi)2+(oi2)2=(oi2+2)22^2 + (2o_i)^2 + (o_i^2)^2 = (o_i^2 + 2)^2,
and that
(2oi)2+(oi22)2+(2oi)2=8oi2+oi44oi2+4=oi4+4oi2+4=(oi2+2)2, \begin{align*} & (2o_i)^2 + (o_i^2 - 2)^2 + (-2o_i)^2 \\ &= 8o_i^2 + o_i^4 - 4o_i^2 + 4 \\ &= o_i^4 + 4o_i^2 + 4 = (o_i^2 + 2)^2, \end{align*}
implying that the vectors ui\mathbf{u}_i, vi\mathbf{v}_i, wi\mathbf{w}_i have the same integer length. Because gcd(2,oi)=gcd(2oi,oi22)=1\gcd(2, o_i) = \gcd(2o_i, o_i^2 - 2) = 1, points Bi=(2,2oi,oi2)B_i = (2, 2o_i, o_i^2) and Di=(2oi,oi22,2oi)D_i = (2o_i, o_i^2 - 2, -2o_i) are primitive lattice points. It is also routine to check that the weights of BiB_i and DiD_i are distinct, because 2+2oi+oi24oi+oi222 + 2o_i + o_i^2 \neq 4o_i + o_i^2 - 2, that is, 2oi2 \neq o_i. Finally, we claim that ABiADiAB_i \perp AD_i, because the dot product
uivi=4oi+2oi(oi22)2oi3=4oi+2oi34oi2oi3=0. \begin{align*} \mathbf{u}_i \cdot \mathbf{v}_i &= 4o_i + 2o_i(o_i^2 - 2) - 2o_i^3 \\ &= 4o_i + 2o_i^3 - 4o_i - 2o_i^3 = 0. \end{align*}
It follows that there are infinitely many unbalanced primary integer squares ABiCiDiAB_iC_iD_i. (We can set Ci=Bi+viC_i = B_i + v_i.)

Second Solution: We maintain the same notation as in the first solution. For i=1,2,i = 1, 2, \dots, let ei=2ie_i = 2i. We set n=1n = 1 and mn=eim - n = e_i. Then [2n2,(mn)2,2n(mn)]=[2,4i2,4i]=2[1,2i2,2i][2n^2, (m-n)^2, 2n(m-n)] = [2, 4i^2, 4i] = 2[1, 2i^2, 2i]. We define vectors
ABi=ui=[1,2i,2i2],ADi=vi=[2i,12i2,2i], \overrightarrow{AB_i} = \mathbf{u}_i' = [1, 2i, 2i^2], \quad \overrightarrow{AD_i} = \mathbf{v}_i' = [-2i, 1 - 2i^2, 2i],
and
wi=[2i2,2i,1]. \mathbf{w}_i' = [-2i^2, 2i, -1].
For each ii, the ordered triples of vectors (ui,vi,wi)(\mathbf{u}_i, \mathbf{v}_i, \mathbf{w}_i) and (ui,vi,wi)(\mathbf{u}_i', \mathbf{v}_i', \mathbf{w}_i') are dual to each other. It is straightforward to check that pairs of vectors (ui,vi)(\mathbf{u}_i', \mathbf{v}_i') provides another family of infinitely many unbalanced primitive integer squares ABiCiDiAB_iC_iD_i such that the planes containing the squares are not parallel to each other.

Third Solution: Let (a,b,c)(a, b, c) be a primitive Pythagorean triple, that is, a,b,ca, b, c are positive integers with a<b<ca < b < c, a2+b2=c2a^2 + b^2 = c^2 and gcd(a,b,c)=1\text{gcd}(a, b, c) = 1. It is not difficult to show that gcd(a,b)=gcd(c,b)=gcd(c,a)=1\text{gcd}(a, b) = \text{gcd}(c, b) = \text{gcd}(c, a) = 1. In order to find a vector with integer length, we consider c4=c2(a2+b2)=c2a2+c2b2=(a2+b2)a2+c2b2=c2a2+(a2+b2)b2c^4 = c^2(a^2 + b^2) = c^2a^2 + c^2b^2 = (a^2 + b^2)a^2 + c^2b^2 = c^2a^2 + (a^2 + b^2)b^2. Set
AB=u=[ab,b2,ac]andAD=v=[a2,ab,bc]. \overrightarrow{AB} = \mathbf{u} = [-ab, b^2, ac] \quad \text{and} \quad \overrightarrow{AD} = \mathbf{v} = [a^2, -ab, bc].
Hence C=(a2ab,b2ab,c(a+b))C = (a^2 - ab, b^2 - ab, c(a+b)).
By construction, AB2=AD2=(a2+b2)c2=c4AB^2 = AD^2 = (a^2 + b^2)c^2 = c^4 and
ABAD=uv=ab(a2b2+c2)=0. \overrightarrow{AB} \cdot \overrightarrow{AD} = \mathbf{u} \cdot \mathbf{v} = ab(-a^2 - b^2 + c^2) = 0.
We also note that the difference of the weights of BB and DD is equal to
(b2+ab+ac)(a2+ab+bc)=b2a2+acbc=(ba)(a+b)+c(ab)=(ba)(a+bc), \begin{aligned} & (b^2 + ab + ac) - (a^2 + ab + bc) \\ &= b^2 - a^2 + ac - bc \\ &= (b-a)(a+b) + c(a-b) \\ &= (b-a)(a+b-c), \end{aligned}
which is nonzero because gcd(a,b)=1\gcd(a, b) = 1 and a,ba, b, and cc are sides of a triangle. Because gcd(a,b)=gcd(c,b)=gcd(c,a)=1\gcd(a, b) = \gcd(c, b) = \gcd(c, a) = 1, gcd(b2,ac)=gcd(a2,bc)=1\gcd(b^2, ac) = \gcd(a^2, bc) = 1. Therefore, ABCDABCD is a unbalanced primitive integer square (associated to this primitive Pythagorean triple).
We note that vector w=[b,a,0]\mathbf{w} = [b, a, 0] is perpendicular to the plane containing the square ABCDABCD. Indeed,
uw=ab2+ab2=0=a2ba2b=vw. \mathbf{u} \cdot \mathbf{w} = -ab^2 + ab^2 = 0 = a^2b - a^2b = \mathbf{v} \cdot \mathbf{w}.
Finally, it is not difficult to see that for distinct primitive Pythagorean triples (a,b,c)(a, b, c) and (a,b,c)(a', b', c'), the vectors w=[b,a,0]\mathbf{w} = [b, a, 0] and w=[b,a,0]\mathbf{w}' = [b', a', 0] are not parallel to each other. Hence, the unbalanced primitive integer squares associated with distinct primitive Pythagorean triples lie in different planes. It is well known that there are infinitely many primitive Pythagorean triples, so our proof is complete.

Source: MathNet, licensed CC-BY-4.0. Statement and solution reproduced as published; topic, difficulty and ordering added by this site.