Problem 5: Balkan MO 1996, Problem 3

by henderson, Feb 14, 2016, 4:56 PM

$$\bf\color{red}Problem \ 5 \  $$In a convex pentagon $ABCDE$, the points $M$, $N$, $P$, $Q$, $R$ are the midpoints of the sides $AB$, $BC$, $CD$, $DE$, $EA$, respectively. If the segments $AP$, $BQ$, $CR$ and $DM$ pass through a single point, prove that $EN$ contains that point as well.
$$\bf\color{red}Solution \ 1 \  $$Let the four lines concur at $ X$. Let $ EX$ meet $ BC$ at $ N'$. Let $ \angle BXM = \angle DXQ = m$, $ \angle MXA = \angle PXD = n$, $ \angle AXR = \angle CXP = p$, $ \angle RXE = \angle N'XC = q$, $ \angle EXQ = \angle BXN' = r$.
Let $ [ABC]$ denote the area of $ \triangle ABC$. Then,
$ [AXM] = [MXB]$ (since they have the same base and height) $ \implies \frac {1}{2} BX \cdot MX \cdot \sin m = \frac {1}{2}MX \cdot \sin n$, or
$ BX \sin m = \ AX \sin n$. Similarly,
$ AX \sin p = EX \sin q$
$ EX \sin r = DX \sin m$
$ DX \sin n = CX \sin p$
Multiplying these,
$ BX \sin r = CX \sin q$
Thus, $ [BXN'] = [CXN']$, so $ N' = N$ is the midpoint of $ BC$.

So the problem generalizes to any convex $n$-gon.

$$\bf\color{red}Solution \ 2 \  $$Here is my solution of the problem. It uses a generalization of the Ceva theorem indeed (by the way, I have never seen it being called Poncelet theorem, but who knows):

Theorem 1. Let n be a positive integer, and let $A_1A_2...A_{2n+1}$ be any (2n + 1)-gon. We work with indices cyclic modulo 2n + 1, so that $A_{2n+1+i}=A_i$ for any i. Let P be a point in the plane; for every index i, let the line $A_iP$ meet the line $A_{n+i}A_{n+1+i}$ at a point $X_i$. Then,

$\prod_{i=1}^{2n+1}\frac{A_{n+i}X_i}{X_iA_{n+1+i}}=1$.

Hereby, we work with directed segments.

Proof of Theorem 1. Let's use not only directed segments, but also directed areas, and let's denote the directed area of a triangle $P_1P_2P_3$ by $\left[P_1P_2P_3\right]$.

Let S and T be the orthogonal projections of the points $A_{n+i}$ and $A_{n+1+i}$ on the line $A_iP$. Then, the two lines $A_{n+i}S$ and $A_{n+1+i}T$ are parallel to each other (since they are both perpendicular to $A_iP$); if we direct these two lines in the same way, then Thales yields $\frac{A_{n+i}X_i}{A_{n+1+i}X_i}=\frac{A_{n+i}S}{A_{n+1+i}T}$. On the other hand, the segments $A_{n+i}S$ and $A_{n+1+i}T$ are the altitudes of the triangles $A_iPA_{n+i}$ and $A_iPA_{n+1+i}$ to the common side $A_iP$; since the area of a triangle equals $\pm\frac12\cdot\text{ side }\cdot\text{ corresponding altitude}$, where the $\pm$ sign is + or - depending on the orientation of the triangle, it follows that $\left[A_iPA_{n+i}\right]=\pm\frac12\cdot A_iP\cdot A_{n+i}S$ and $\left[A_iPA_{n+1+i}\right]=\pm\frac12\cdot A_iP\cdot A_{n+1+i}T$; note that the two $\pm$ signs are either both + or both - (since the lines $A_{n+i}S$ and $A_{n+1+i}T$ are directed in the same way). Hence, $\frac{\left[A_iPA_{n+i}\right]}{\left[A_iPA_{n+1+i}\right]}=\frac{A_{n+i}S}{A_{n+1+i}T}$. Comparing this with $\frac{A_{n+i}X_i}{A_{n+1+i}X_i}=\frac{A_{n+i}S}{A_{n+1+i}T}$, we obtain $\frac{A_{n+i}X_i}{A_{n+1+i}X_i}=\frac{\left[A_iPA_{n+i}\right]}{\left[A_iPA_{n+1+i}\right]}$. Thus,

$\frac{A_{n+i}X_i}{X_iA_{n+1+i}}=-\frac{A_{n+i}X_i}{A_{n+1+i}X_i}=-\frac{\left[A_iPA_{n+i}\right]}{\left[A_iPA_{n+1+i}\right]}=\frac{\left[A_iPA_{n+i}\right]}{\left[A_{n+1+i}PA_i\right]}=\frac{\left[A_iPA_{n+i}\right]}{\left[A_{n+1+i}PA_{n+\left(n+1+i\right)}\right]}$,

where in the last step, we have used the fact that $A_{n+\left(n+1+i\right)}=A_{2n+1+i}=A_{i}$ (since indices are cyclic modulo 2n + 1). This identity holds for every index i; multiplying all such identities, we get

$\prod_{i=1}^{2n+1}\frac{A_{n+i}X_i}{X_iA_{n+1+i}}=\prod_{i=1}^{2n+1}\frac{\left[A_iPA_{n+i}\right]}{\left[A_{n+1+i}PA_{n+\left(n+1+i\right)}\right]}=\prod_{i=1}^{2n+1}\left[A_iPA_{n+i}\right]\left/\prod_{i=1}^{2n+1}\left[A_{n+1+i}PA_{n+\left(n+1+i\right)}\right]\right.$.

But actually, $\prod_{i=1}^{2n+1}\left[A_iPA_{n+i}\right]$ and $\prod_{i=1}^{2n+1}\left[A_{n+1+i}PA_{n+\left(n+1+i\right)}\right]$ are the same product, just with a different order of factors. Hence,

$\prod_{i=1}^{2n+1}\frac{A_{n+i}X_i}{X_iA_{n+1+i}}=\prod_{i=1}^{2n+1}\left[A_iPA_{n+i}\right]\left/\prod_{i=1}^{2n+1}\left[A_{n+1+i}PA_{n+\left(n+1+i\right)}\right]\right.=1$,

and Theorem 1 is proven.

Now, let's reformulate the problem (note that the pentagon does not need to be convex):

Theorem 2. Let $A_1A_2A_3A_4A_5$ be an arbitrary pentagon, and let $M_1$, $M_2$, $M_3$, $M_4$, $M_5$ be the midpoints of its sides $A_3A_4$, $A_4A_5$, $A_5A_1$, $A_1A_2$, $A_2A_3$, respectively. If the lines $A_1M_1$, $A_2M_2$, $A_3M_3$ and $A_4M_4$ have a common point O, then the line $A_5M_5$ also passes through this point O.

Proof of Theorem 2. Since the lines $A_1M_1$, $A_2M_2$, $A_3M_3$ and $A_4M_4$ pass through the point O, the points $M_1$, $M_2$, $M_3$, $M_4$ are the points where the lines $A_1O$, $A_2O$, $A_3O$, $A_4O$ meet the lines $A_3A_4$, $A_4A_5$, $A_5A_1$, $A_1A_2$, respectively. Let the line $A_5O$ meet the line $A_2A_3$ at a point $N_5$. Then, by Theorem 1, we have

$\frac{A_3M_1}{M_1A_4}\cdot\frac{A_4M_2}{M_2A_5}\cdot\frac{A_5M_3}{M_3A_1}\cdot\frac{A_1M_4}{M_4A_2}\cdot\frac{A_2N_5}{N_5A_3}=1$.

But since $M_1$, $M_2$, $M_3$, $M_4$ are the midpoints of the sides $A_3A_4$, $A_4A_5$, $A_5A_1$, $A_1A_2$, we have $\frac{A_3M_1}{M_1A_4}=1$, $\frac{A_4M_2}{M_2A_5}=1$, $\frac{A_5M_3}{M_3A_1}=1$, $\frac{A_1M_4}{M_4A_2}=1$. Thus, we also have $\frac{A_2N_5}{N_5A_3}=1$, so the point $N_5$ must be the midpoint of the side $A_2A_3$. Thus, $N_5=M_5$. Now, from the definition of the point $N_5$, it is clear that the line $A_5N_5$ passes through the point O; hence, the line $A_5M_5$ passes through the point O, and Theorem 2 is proven.

Darij
This post has been edited 5 times. Last edited by henderson, Sep 12, 2016, 2:42 PM

Comment

0 Comments

"Do not worry too much about your difficulties in mathematics, I can assure you that mine are still greater." - Albert Einstein

avatar

henderson
Archives
Shouts
Submit
7 shouts
Tags
About Owner
  • Posts: 312
  • Joined: Mar 10, 2015
Blog Stats
  • Blog created: Feb 11, 2016
  • Total entries: 77
  • Total visits: 20931
  • Total comments: 32
Search Blog
a