Let $M$ and $N$ be the midpoints of the respective sides $AB$ and $AC$ of an acute-angled triangle $ABC$. Let $P$ be the foot of the perpendicular from $N$ onto $BC$ and let $A_1$ be the midpoint of $MP$. Points $B_1$ and $C_1$ are obtained similarly. If $AA_1$, $BB_1$ and $CC_1$ are concurrent, show that the triangle $ABC$ is isosceles. Mircea Becheanu
Problem
Source: Romanian TST 2002
Tags: ratio, geometry, rectangle, geometry proposed
19.03.2004 17:29
It's problem 3 from the first Romanian TST in 2002. I have a computational proof and I won't post it because it's really boring: we use Menelaus to find the ratio in which side BC is divided by AA<sub>1</sub>, then we write Ceva's condition for concurrence with the ratios we have just calculated and we get something like (tanA-tanB)(tanB-tanC)(tanC-tanA)=0, so 2 of them must be equal, so the triangle is isosceles. I remember slaving on this one for 45 minutes or so (during th contest).
19.03.2004 18:18
Valiowk wrote: Let M and N be the midpoints of the respective sides AB and AC of an acute-angled triangle ABC. Let P be the foot of the perpendicular from N onto BC and let A<sub>1</sub> be the midpoint of MP. Points B<sub>1</sub> and C<sub>1</sub> are obtained similarly. Well, this looks a bit strange since the definition of $A_1$ is not symmetric in $B$ and $C$. However, we can easily see that if we take the foot $Q$ of the perpendicular from $M$ onto $BC$, and define $A_1$ as the midpoint of $NQ$, then we get the same point $A_1$. In fact, the quadrilateral $MNQP$ is a rectangle. Valiowk wrote: If $AA_1$, $BB_1$ and $CC_1$ are concurrent, show that triangle $ABC$ is isosceles. Well, I have reduced this to $(b^2+3c^2-a^2)(c^2+3a^2-b^2)(a^2+3b^2-c^2)$ = $(c^2+3b^2-a^2)(a^2+3c^2-b^2)(b^2+3a^2-c^2)$ ==> triangle is isosceles. But this is algebra, again. Any nice proof? Darij
20.03.2004 03:34
I think I got here as well during that contest. We replace expressions of the type a<sup>2</sup>+b<sup>2</sup>-c<sup>2</sup> with 2ab*cosC, we divide both sides by 8 and by abc, and we get \pi{cyl}(b*cosA+c)=\pi{cyl}(c*cosA+b)=>\pi{cyl}(b*cosA+c)/(a*cosB+c)=1. b*cosA+a*cosB=c and b*cosA/a*cosB=tanB/tanA=>(b*cosA+c)/(a*cosB+c)=(2tanB+tanA)/(2tanA+tanB)=>\pi{cyl}(2tanB+tanA)/(2tanA+tanB)=1, and from here it's easy to derive \pi{cyl}(tanA-tanB)=0. P.S. {cyl} comes from cyclic.
05.01.2010 22:43
See the attached drawing, the triangles $ MNA_1$ and $ BCA_2$ are centrally similar, centered at $ A$. $ A_2$ is the center of the rectangle erected outwardly on $ BC$, of dimensions $ a, h_a$. If $ A' \equiv BC \cap AA_2$, we have then: $ \frac {BA'}{CA'} = \frac {S_{[ABA_2]}}{S_{[ACA_2]}}=\frac{AB \cdot BA_2 \cdot sin (\widehat {ABC}+\widehat {CBA_2})}{AC \cdot CA_2 \cdot sin (\widehat {ACB}+\widehat {BCA_2})}=\frac{c \cdot sin (\angle B + \angle \alpha)}{b \cdot sin (\angle C + \angle \alpha )}$ ( 1 ), where $ a, b, c$ and $ h_a, h_b, h_c$ are the sides and altitudes of the triangle, $ S_{[XYZ]}$ is the area of the $ \triangle XYZ$, $ \alpha$ being the half-angle of the diagonals of the rectangle constructed as explained and other two similar relations. If $ AA', BB', CC'$ are concurrent ($ B', C'$ are obtained similarly), then, by Ceva we get as the product of fractions (1) the unit and from this we get the same results as Darij. Best regards, sunken rock
Attachments:

06.12.2012 06:11
Computational solution, although I'll post it since it seems sort of instructive on ``how to bash'' Not that there's anything terribly new here... Employ barycentric coordinates. Let $S_A = \frac{b^2+c^2-a^2}{2}$, et cetera. By taking the midpoint of the foot of the perpendicular $(0: S_C : S_B)$ with $C = (0 : 0 : a^2)$ we find that $P=(0 : S_C : S_B + a^2)$. With $M = (a^2 : a^2 : 0)$ we get $A_1 = \left( a^2 : S_C+a^2 : S_B+a^2 \right)$. Hence $AA_1$ divides $BC$ in a $S_C+a^2$ to $S_B+a^2$ ratio. Now by Ceva's Theorem, the condition is equivalent to \begin{align*} 0 &= (S_B+a^2)(S_C+b^2)(S_A+c^2)-(S_B+c^2)(S_C+a^2)(S_A+b^2) \\ &= \sum_{\text{cyc}} a^2(b^2-c^2)S_a + \sum_{\text{cyc}} (c^2-b^2)S_{BC} \\ &= \sum_{\text{cyc}} (b^2-c^2)(a^2S_A-S_{BC}) \\ &= \frac{1}{4} \sum_{\text{cyc}} (b^2-c^2)(2a^2(b^2+c^2-a^2)-(b^2+a^2-c^2)(c^2+a^2-b^2)) \\ &= \frac{1}{4} \sum_{\text{cyc}} (b^2-c^2)(2a^2b^2+2b^2c^2+2c^2a^2+b^4+c^4-3a^4) \\ &= \frac{1}{4} \sum_{\text{cyc}} (b^2-c^2)(b^4+c^4-3a^4) \\ &= \frac{1}{4} \sum_{\text{cyc}} (b^2-c^2)(a^4+b^4+c^4-4a^4) \\ &= - \sum_{\text{cyc}} a^4(b^2-c^2) \\ &= (a^2-b^2)(b^2-c^2)(c^2-a^2) \end{align*} which holds if and only if $ABC$ is isosceles.
24.07.2018 15:42
Valiowk wrote: Let $M$ and $N$ be the midpoints of the respective sides $AB$ and $AC$ of an acute-angled triangle $ABC$. Let $P$ be the foot of the perpendicular from $N$ onto $BC$ and let $A_1$ be the midpoint of $MP$. Points $B_1$ and $C_1$ are obtained similarly. If $AA_1$, $BB_1$ and $CC_1$ are concurrent, show that the triangle $ABC$ is isosceles. Mircea Becheanu Is there a full proof
24.07.2018 16:42
Can somebody help me
24.07.2018 19:09
Valiowk wrote: ...Points $B_1$ and $C_1$ are obtained similarly... Mircea Becheanu This is not a clear definition for $B_1, C_1.$
27.07.2023 10:36
Solved from LiOG 3.8 Let the foot from $M$ to $BC$ be $T$, so $A_1$ is the center of rectangle $MNPT$. Let $AA_1$ meet $MN$ at $R$ and $BC$ at $S$. Let $MR=d$. Due to symmetry, since $AA_1$ passes through the center of the rectangle, we have $SP=d$ as well. Furthermore, by homothety at $A$, $BS=2d$, so $BP=3d$. However, we also have $$BP=BT+TP=\frac{c}{2}\cos\beta+\frac{a}{2},$$so therefore, $$3d=\frac{c}{2}\cos\beta+\frac{a}{2}.$$Thus, $$d=\frac{c\cos\beta+a}{6},$$so $$\frac{BS}{CS}=\frac{\frac{c\cos\beta+a}{3}}{a-\frac{c\cos\beta+a}{3}}=\frac{a+c\cos\beta}{2a-c\cos\beta}=\frac{2a^2+2ac\cos\beta}{4a^2-2ac\cos\beta}.$$However, by Law of Cosines, this is equal to $$\frac{3a^2+c^2-b^2}{3a^2+b^2-c^2}.$$Thus, by Ceva's theorem, it suffices to show that $$\prod_{cyc}(3a^2+b^2-c^2)=\prod_{cyc}(3a^2-b^2+c^2)$$implies that $\triangle ABC$ is isosceles. Let $p=a^2,q=b^2,c=r^2$ so that this is $$\prod_{cyc}(3p+q-r)=\prod_{cyc}(3p-q+r).$$Let $x=q-r,y=r-p,z=p-q$ so that this is $$(3p+x)(3q+y)(3r+z)=(3p-x)(3q-y)(3r-z)$$Since terms with an even number of factors in $\{x,y,z\}$ cancel, this is just $$9(pqz+pry+qrx)+xyz=0.$$Substituting $x,y,z$ back in, this is $$9(\sum_{cyc}pq(p-q))+(p-q)(q-r)(r-p)=\sum_{cyc}8p^2q+p^2r-9pq^2,$$and evaluating the cyclic sum and setting it to 0, this is simply just $$8(p^2q+q^2r+r^2p)-8(p^2r+q^2p+r^2q),$$so the condition is just $$p^2q+q^2r+r^2p-(p^2r+q^2p+r^2q)=0.$$This factors as $$(p-q)(q-r)(r-p)=0,$$so two of $p,q,r$ are equal so $\triangle ABC$ is isosceles and we are finally done as we step back from this crazy bash.