9

Inspired by this problem, I'd like to ask

Given a triangle $\triangle ABC$, what is the point $X$ on the plane so that $$[AEXF] = [BFXD] = [CDXE]$$ where $D, E, F$ are the orthogonal projections of $X$ onto $BC, CA, AB$, respectively.

By "what is the point" I mean to ask for its name (because knowing a point's name gives you power over it). I mean, triangle centers have been studied exhaustively so I'd be very surprised if such a simple setup hasn't come up in the past. It'd be also interesting to know how to construct it and its coordinates in terms of $A, B, C$.

triangle ABC

For starting some work, let's name the relevant lengths as in the image above. $$\begin{cases} \overline{XD} = x,\qquad\overline{BC} = a,\qquad\overline{CA} = b,\qquad\overline{AB} = c\\ \overline{XE} = y,\qquad\overline{CD} = i,\qquad\overline{AE} = k,\qquad\overline{BF} = m,\\ \overline{XF} = z,\qquad\overline{BD} = j,\qquad\overline{CE} = l, \qquad\overline{AF} = n\end{cases}$$

There are several ways to write the areas of the quadrilaterals, each one giving us a set of equivalent equations.


$$\begin{cases}~ [AEXF] &= \dfrac12(ky+nz)\\ [BFXD] &= \dfrac12(mz+jx)\\ [CDXE] &= \dfrac12(ix+ly) \end{cases}$$ $$ \implies ky+nz = mz+jx = ix+ly $$


$$\begin{cases}~ [AEXF] &= \dfrac12\sin A(kn+yz)\\ [BFXD] &= \dfrac12\sin B(mj+zx)\\ [CDXE] &= \dfrac12\sin C(il+xy) \end{cases}$$ $$\begin{aligned} &\implies \sin A(kn+yz) = \sin B(mj+zx) = \sin C(il+xy)\\ &\implies a(kn+yz) = b(mj+zx) = c(il+xy) \end{aligned}$$ The second implication above is a consequence of the law of sines.

Alma Arjuna
  • 7,581
  • Let AB=13, BC=6, CA=9. Find the numerical values of the coordinates of X and look for these here. Does this help? – Rosie F Dec 12 '25 at 15:08
  • @RosieF I wouldn't know how to find the numerical values – Alma Arjuna Dec 12 '25 at 17:00
  • 8
    It's Kimberling's X(5373) "Equiareality Center" in the Encyclopedia of Triangle Centers. (Acuteness may be a factor in existence and/or uniqueness. I haven't checked the references.) – Blue Dec 12 '25 at 18:15
  • To solve it by hand, you can take vertex $A$ as the origin, and shift vertices $B$ and $C$ accordingly. Then you can apply a linear transformation $T$ , such that $TC = (1, 0)$ and $TB = (0, 1)$. Since linear transformations preserve the area ratios of the figures they act on, you can now solve the problem on the resulting right triangle. It is much easier, and this is what I will do in the solution that I will provide soon. – Hosam Hajeer Dec 12 '25 at 20:27
  • 4
    @Hosam Hajeer : your proposal isn't possible because a linear transformation doesn't preserve right angles in general, unless it is an orthogonal transform. – Jean Marie Dec 12 '25 at 21:12
  • Ooooooh, what a fatal mistake. – Hosam Hajeer Dec 12 '25 at 21:23

2 Answers2

3
  1. Define the vectors $v_1 = AB, v_2 = BC, v_3 = CA$

  2. Normalize these vectors into $u_1, u_2, u_3$ respectively.

  3. Find the perpendicular vectors $w_1, w_2, w_3$, respectively, by rotating $u_i$ $90^\circ$ clockwise , $ i = 1, 2, 3 $.

  4. Let the point we're looking for be $P = (x_1, y_1) $

  5. Referring to the figure in the OP, it follows that

$n = u_1 \cdot (P - A) $

$z = w_1 \cdot (P - A) $

$m = - u_1 \cdot (P - B)$

$j = u_2 \cdot (P - B) $

$x = w_2 \cdot (P - B) $

$i = - u_2 \cdot (P - C) $

$ l = u_3 \cdot (P - C) $

$ y = w_3 \cdot (P - C) $

$ k = - u_3 \cdot (P - A) $

These are all the distances we need

  1. Now, we have the following area expressions,

$[ABC] = \dfrac{1}{2} ( u_3 \cdot (A - C) )( w_3 \cdot (B - A) ) $

Using the transpose operator, this can be written as

$ [ABC] = \dfrac{1}{2} (A - C)^T u_3 w_3^T (B - A) $

  1. Similarly, and as listed in the OP, we find that

$[AEPF] = \dfrac{1}{2} ( k y + n z ) = \dfrac{1}{3} [ABC]$

Therefore, the first equation we have is,

$ \dfrac{1}{2} ( - (P - A)^T u_3 w_3^T (P - C) + (P - A) u_1 w_1^T (P - A) ) = \dfrac{1}{3} [ABC] $

  1. The second equation we have is, similarly,

$[BFPD] = \dfrac{1}{2} ( mz + jx) = \dfrac{1}{3}[ABC]$

Therefore,

$ \dfrac{1}{2} ( - (P - B)^T u_1 w_1 (P - A) + (P - B) u_2 w_2^T (P - B) ) = \dfrac{1}{3} [ABC] $

Both area expressions are quadratic equations in $x_1, y_1$, the coordinates of $P$.

  1. Now all we need is a two-variable quadratic system solver, and this I have.

  2. Example I took $A = (3, 2) , B = (8,12), C = (17, 5) $. The solver produced four real solutions but only one of them is valid with $P$ inside the triangle, all the other three solutions had $P$ outside the triangle. The valid solution had $P = (9.5097811 , 5.657915)$

enter image description here

Hosam Hajeer
  • 511
  • 1
  • 10
2

Observations and a conjecture:

In the figure below, $O$ is circumcenter of triangle $ABC$ and $K$ is centroid. $Q$ is the midpoint of $OK$, and hence lies on the Euler line.

Drawing $QF$, $QH$, $QJ$ perpendicular to $AC$, $BC$, $AB$, respectively, according to GeoGebra we have quadrilateral areas$$QJAF=\frac{QJ\cdot JA+AF\cdot FQ}{2}=170.94$$$$QFCH=\frac{QF\cdot FC+CH\cdot HQ}{2}=168.57$$$$QHBJ=\frac{QH\cdot HB+BJ\cdot JQ}{2}=166.84$$

4-als in. scalene

These numbers are not far from their arithmetic mean of $168.78$. If the discrepancies are due only to rounding errors, then perpendiculars from point $Q$, midway between circumcenter and centroid, determine three equal quadrilaterals.

Moreover, for $\angle A >\angle B>\angle C$, if perpendiculars are instead dropped from $O$, then it can be seen that $OJAF$ is greatest in area and $OFCH$ is least, while if perpendiculars are dropped from $K$, then $KJAF$ is least in area and $KFCH$ is greatest. So again, equality requires an intermediate point like $Q$ or one very near it.

  • 2
    The solution pt is not (in general) on the Euler line. Consider the right-triangle question that inspired this one; the solution pt is "visibly" distinct from the line. (The Encyclopedia of Triangle Centers doesn't list $X(5373)$ among the pts on the Euler line, either, but omissions could be oversights.) More formally, one can coordinatize, parameterize an arbitrary Euler-line pt as $O+k(K-O)$ (w/circumcenter $O$ and centroid $K$), calculate associated quad areas, and observe that no $k$ makes all three equal unless the triangle is isosceles. – Blue Dec 15 '25 at 18:57
  • 1
    @Blue—Yes, in my example the quadrilaterals are more nearly equal ($168.56$, $168.52$ and $168.43$) when perpendiculars are dropped from a point slightly to the left of point $Q$ and off the Euler line. So much for my conjecture. – Edward Porcella Dec 17 '25 at 19:03
  • nice answer ,.. +1 – TShiong Dec 31 '25 at 11:48