For Mark

Geometry Level 5

A B C \triangle ABC has integer sides. A second (non-degenerate) triangle is formed from the incenter I I , the orthocenter H H , and the fletcher point F l Fl of A B C \triangle ABC . If H I F l \triangle HIFl has an integral area, what is its smallest possible value?


The answer is 66.

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

1 solution

Mark Hennings
Jan 14, 2021

The incentre I I of the triangle has barycentric coordinates a : b : c a:b:c . The orthocentre H H of the triangle has barycentric coordinates sec A : sec B : sec C \sec A:\sec B :\sec C , or 1 b 2 + c 2 a 2 : 1 a 2 + c 2 b 2 : 1 a 2 + b 2 c 2 \tfrac{1}{b^2+c^2-a^2}:\tfrac{1}{a^2+c^2-b^2}:\tfrac{1}{a^2+b^2-c^2} . Of course, these coordinates are singular if the triangle is right-angled, but the normalized version of these coordinates is not singular when the triangle is right-angled, so there is no problem here.

The Fletcher point F l Fl has barycentric coordinates a ( 2 cos 2 1 2 B + cos 2 1 2 C cos 2 1 2 A ) : b ( 2 cos 2 1 2 A + cos 2 1 2 C cos 2 1 2 B ) : c ( 2 cos 2 1 2 A + cos 2 1 2 B cos 2 1 2 C ) a\left(2 - \tfrac{\cos^2\frac12B + \cos^2\frac12C}{\cos^2\frac12A}\right)\,:\,b\left(2 - \tfrac{\cos^2\frac12A + \cos^2\frac12C}{\cos^2\frac12B}\right) \;:\; c\left(2 - \tfrac{\cos^2\frac12A + \cos^2\frac12B}{\cos^2\frac12C}\right) After some substitution, these coordinates cam be written ( b c ) 2 + ( b + c ) a 2 a 2 b + c a : ( a c ) 2 + ( a + c ) b 2 b 2 a + c b : ( a b ) 2 + ( a + b ) c 2 c 2 a + b c \tfrac{(b-c)^2 + (b+c)a - 2a^2}{b+c-a} \;:\; \tfrac{(a-c)^2 + (a+c)b - 2b^2}{a+c-b} \;:\; \tfrac{(a-b)^2 + (a+b)c - 2c^2}{a+b-c} The area of H I F l HIFl is equal to the modulus of d e t M A B C \mathrm{det}M\,|ABC| , where A B C |ABC| is the area of A B C ABC and d e t M \mathrm{det}M is the determinant of the 3 × 3 3 \times 3 matrix whose rows are the normalized barycentric coordinates of H , I , F l H,\,I,\,Fl . Thus H I F l = ( a b ) ( a c ) ( b c ) ( a 2 + b 2 + c 2 2 a b 2 a c 2 b c ) 2 ( a + b + c ) ( a 4 + b 4 + c 4 ( a 3 b + a b 3 + a 3 c + a c 3 + b 3 c + b c 3 ) + a b c ( a + b + c ) ) A B C |HIFl| \; = \; \left| \frac{(a-b)(a-c)(b-c)(a^2+b^2+c^2-2ab-2ac-2bc)}{2(a+b+c)\big(a^4+b^4+c^4-(a^3b+ab^3+a^3c+ac^3+b^3c+bc^3)+abc(a+b+c)\big)}\right|\,|ABC| If we write a = v + w a=v+w , b = u + w b=u+w , c = u + v c = u+v , then we have H I F l = ( u v ) ( u w ) ( v w ) ( u v + u w + v w ) 2 ( u + v + w ) ( u 2 ( v w ) 2 + v 2 ( u w ) 2 + w 2 ( u v ) 2 ) A B C |HIFl| \; = \; \left|\frac{(u-v)(u-w)(v-w)(uv + uw + vw)}{2(u+v+w)(u^2(v-w)^2 + v^2(u-w)^2 + w^2(u-v)^2)}\right|\,|ABC| and hence H I F L HIFL is nondegenerate provided that u , v , w u,v,w are distinct, and hence provided that a , b , c a,b,c are distinct.

Let us call the triangle H I F l HIFl the FHI-triangle of A B C ABC . If a , b , c a,b,c are integers and H I F l |HIFl| is an integer, then A B C |ABC| is rational, and so we can find P N P \in\mathbb{N} such that P 2 A B C P^2|ABC| is an integer, and hence the triangle with sides P a , P b , P c Pa,Pb,Pc is Heronian. Thus we can find Q N Q \in \mathbb{N} such that the triangle Δ 0 \Delta_0 with sides P Q a , P Q b , P Q c \tfrac{P}{Q}a,\tfrac{P}{Q}b,\tfrac{P}{Q}c is a primitive Heronian triangle. If we write P Q = p q \tfrac{P}{Q} = \tfrac{p}{q} as a fraction in lowest terms, we see that any prime factor of p p will be a common factor of the sides of Δ 0 \Delta_0 . Since Δ 0 \Delta_0 is primitive, we deduce that p p has no prime factors, and hence that p = 1 p=1 . If we let Γ 0 \Gamma_0 FHI-triangle of the Heronian triangle Δ 0 \Delta_0 , then H I F l = q 2 Γ 0 |HIFl| = q^2|\Gamma_0| and we are interested in choosing q q to be as small as possible, since we want to minimize H I F l |HIFl| . Suppose that Γ 0 = m n |\Gamma_0| = \tfrac{m}{n} , where m , n N m,n \in \mathbb{N} are coprime. If the prime factorisation of n n is p 1 r 1 p 2 r 2 p N r N p_1^{r_1}p_2^{r_2} \cdots p_N^{r_N} , where p 1 , p 2 , , p N p_1,p_2,\ldots,p_N are distinct primes and r 1 , r 2 , , r N N r_1,r_2,\ldots,r_N \in \mathbb{N} , then q q must be equal to p 1 s 1 p 2 s 2 p N s N p_1^{s_1}p_2^{s_2} \cdots p_N^{s_N} , where s j = { 1 2 r j r j even 1 2 ( r j + 1 ) r j odd s_j \; =\; \left\{ \begin{array}{lll} \frac12r_j & \hspace{1cm} & r_j\;\mbox{even} \\ \frac12(r_j+1) & & r_j\; \mbox{odd} \end{array}\right. and hence we deduce that H I F l = ξ ( n ) m |HIFl| = \xi(n)m , where ξ ( n ) \xi(n) , sometimes called the core of n n , is the product of the primes that occur with odd multiplicity in the prime factorization of n n .

Thus, for any nondegenerate primitive Heronian triangle Δ \Delta , calculate the area Γ |\Gamma| of its FHI-triangle Γ \Gamma , which will be a rational m n \tfrac{m}{n} in lowest terms. Then Ξ ( Δ ) = ξ ( n ) m \Xi(\Delta) = \xi(n)m will be the smallest possible area of the FHI-triangle of a triangle with integer sides, similar to Δ \Delta , whose FHI-triangle has integral area.

The Mathematica code

amax = 1000;
data = {};
For[a = 3, a <= amax, a++,
    For[b = Floor[a/2] + 1, b <  a, b++,
        For[c = a - b + 1, c < b, c++,
            If[GCD[GCD[a,b],c] > 1,,
                If[IntegerQ[Sqrt[(a + b + c) (a + b - c) (a + c - b) (b + c - a)/16]],AppendTo[data, {a, b, c}],
                           ]]]]]

calculates the 5853 5853 nondegenerate primitive Heronian triangles with largest side of 1000 1000 or less, and it is easy to apply the above theory to find the smallest triangle of the desired type that is similar to each of these. Of all these triangles, the smallest area is 66 \boxed{66} , coming from the triangle with sides 29 , 65 , 68 29,65,68 . The next smallest area is 286 286 (coming from a triangle similar to the 3 , 4 , 5 3,4,5 triangle), and the one after that is 2160 2160 .

In a separate calculation, I also checked all triangles with u , v , w 600 u,v,w \le 600 (where either u , v , w u,v,w are all integers or all half-integers) and found no smaller area than 66 66 either.

The diagram shows the optimal solution. The Soddy line through the incentre I and the Gergonne point is shown in green, and the Gergonne line is shown in light blue (only two of the Nobbs points are visible on the diagram, but two are enough to define the line). The Fletcher point is the intersection of these two perpendicular lines, and the FHI-triangle is shown in orange.

Nice problem. Thanks!

Mark Hennings - 4 months, 4 weeks ago

Exquisite exposition. To my mind, reading it is a bit like waltzing (galloping?) through mathematics. Thank you so much for taking the time to write it up.

It surprised me even one such triangle exists. Now that you point out there is a sequence, I think there are more than you name. For example a brute force search of integer triangles with sides less than 1000, yields FHI-triangle areas of 66, 264, 286, 594, 1056, ... Can you find any of these? The second smallest has a reference triangle with sides 136, 130, 58.

I've updated the diagram to replace it with a real FHI-triangle. :)

Fletcher Mattox - 4 months, 4 weeks ago

Log in to reply

There are an infinite number for every nondegenerate primitive Heronian triangle (since the fundamental, namely smallest, one can be scaled up by any integer factor), so I calculated 5853 5853 fundamental ones!

Your list of areas is

  • 66 66 , from the primitive (68,65,29) triangle, which needs no scale-up factor,
  • 264 = 4 × 66 264 = 4\times66 , from the above triangle, scaled up by a factor of 2,
  • 286 286 , from the (130,104,78) triangle, which is the primitive (5,4,3) triangle scaled up by a factor of 26,
  • 594 = 9 × 66 594 =9\times66 , from the primitive (68,65,29) triangle, scaled up by a factor of 3,
  • 1056 = 16 × 66 1056 = 16\times66 , from the primitive (68,65,29) triangle, scaled up by a factor of 4

Only 66 66 and 286 286 , therefore, appear in my list. My list of fundamental areas coming from primitive triangles goes 66 , 286 , 2160 , 2701 , 4662 , 8736 , 41625 . . . 66,\,286,\,2160,\,2701,\,4662,\,8736,\,41625\,... . By the 50th area in the list, we have reached an area of 10050236. By the 100th area in the list, we have reached 82882800.

Mark Hennings - 4 months, 4 weeks ago

Doh! Sorry about that. Shall I submit to OEIS? :)

Fletcher Mattox - 4 months, 4 weeks ago

Log in to reply

The only problem is that we are not 100% certain that there isn’t a smaller area! It is conceivably possible that a very large Heronian triangle might have a minuscule FHI-triangle, though I doubt it. I have not completed the analysis. If we submitted to OEIS it would be a list ordered by (in my case) the side-lengths of the underlying primitive Heronian triangles. It might be better for me to reorder that list by the dimensions of the actual triangles. Even better would be to prove the non-existence of the boojum massive Heronians with minuscule FHIs...

Mark Hennings - 4 months, 4 weeks ago

My first reaction was that my comment was somewhat tongue-in-cheek and that this sequence is a bit too obscure. Ha ha! Too obscure for OEIS? Give me a break. Yes, by all means submit it. I think I'll leave the proof to you!

Fletcher Mattox - 4 months, 4 weeks ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...