7

Let $\bigodot O_{1},\bigodot O_{2}$ and $\bigodot O_{3}$ be internal tangents to $\bigodot O$ at $A, B$ and $C$, respectively, and mutually intersecting at $D,E,F$ respectively. (As shown in Figure)

Assume that $GH,IJ,KL$ are external common tangents of $\bigodot O_{1},\bigodot O_{2}$ and $\bigodot O_{3}$ respectively.

show that

$$\dfrac{S_{FLA}}{S_{AGE}}\cdot\dfrac{S_{EHC}}{S_{CID}}\cdot\dfrac{S_{DJB}}{S_{BKF}}=1$$

This result was found by my student, and this result fell nice, because it is almost similar to Ceva's theorem, only difference in this problem it is area, or can I say it is Ceva's theorem generalization!. I try by Ceva's theorem and area, Sine theorem, and so on, and can't solve it enter image description here

math110
  • 93,304
  • @user21820,Oh,Thank you ,I know I have post this result is "reslut'' My previous error habits! – math110 Jan 03 '15 at 15:09
  • 1
    Well there are a ton of other spelling mistakes in your question. It's not that important since we can guess the meaning but you may want to consider reading through your own question carefully before (or even after) posting. I know English isn't your first language so grammar would be harder for you, but I'm sure you can spell properly if you want! – user21820 Jan 03 '15 at 15:13
  • I propose the following attacking plan: for first, show that if we replace the circle with center $O_3$ with the circle having $OC$ as a diameter, the "cross-ratio" is unchanged. The configuration in which $O_1,O_2,O_3$ are the midpoints of $OA,OB,OC$ is easy to solve, since the diameter of the circle through $D,E,F$ equals $OA$ by Johnson's theorem. – Jack D'Aurizio Jan 03 '15 at 16:14
  • Nice,+1 @JackD'Aurizio,I fell you plan maybe encounter obstacles,I have try it this idea before. – math110 Jan 03 '15 at 16:31
  • Observation: $AG$ and $CH$ will intersect on the big circle, and if you connect that point with $E$, you may move $E$ along that line without changing the ratio of the incient triangles. You can do the same for $D$ and $F$, and the three lines will intersect in one point. Not sure whether this is of any use, but moving all three points out to the circle (where all areas become zero so you'd need some limit process) or in to that common point might be useful as one step in a proof. Not sure how, though. Not sure how to prove my observation either. But I thought I'd share it just in case. – MvG Jan 05 '15 at 22:51

1 Answers1

4

I'm going to assume for now that in the following figure three lines $BK$, $AL$ and $EE_1$ meet at one point $X$, and furthermore, $X$ lies on the circle $\Omega$. I'll prove it later.

Let $R_1$, $R_2$ and $R$ be the radii of $\Omega_1$, $\Omega_2$ and $\Omega$ respectively. I need them to introduce $k_1 = \frac{R}{R_1}$ and $k_2 = \frac{R}{R_2}$.

I'm going to prove now that $\frac{S_2}{S_1}$ depends only on $k_1$ and $k_2$ and nothing else.

image #1 here

$EE_1$ is radical axis of circles $\Omega_1$ and $\Omega_2$ so $pow(Y, \Omega_1) = pow(Y, \Omega_2) \Longrightarrow YK^2 = YL^2$. Therefore $XY$ is median in $\triangle XKL$ and $EY$ is median in $\triangle EKL$.

So $S_{XKY} = S_{XLY}$ and $S_{EKY} = S_{ELY}$. Hence $S_{XKE} = S_{XLE}$. I'll name it $S$. So $S = S_{XKE} = S_{XLE}$.

Returning to what we want: $$\frac{S_2}{S_1} = \frac{S_2}{S}\frac{S}{S_1} = \frac{BK}{KX}\cdot\frac{XL}{LA}.$$

Because of some basic similarities (or dilation, whichever you prefer) $$1 + \frac{XL}{LA} = \frac{XA}{LA} = \frac{R}{R_1} = k_1,$$ so $$\frac{XL}{LA} = k_1 - 1.$$

Similarly, $$\frac{XK}{KB} = k_2 - 1.$$

Finally, $$\frac{S_2}{S_1} = \frac{BK}{KX}\cdot\frac{XL}{LA} = \frac{k_1 - 1}{k_2 - 1}.$$

The only thing left to do is to multiply 3 such fractions together, one for every pair of interior circles: $$\frac{k_1 - 1}{k_2 - 1}\cdot\frac{k_2 - 1}{k_3 - 1}\cdot\frac{k_3 - 1}{k_1 - 1} = 1.$$


As promised, I'm going to show why $BK$, $AL$ and $EE_1$ meet on the big circle.

Consider the dilation with center A such that it transforms $\Omega_1 \rightarrow \Omega$. Under that dilation point $L$ becomes some point $L'$ on the big circle. A dilation always transforms a line into a parallel line so tangent at $L'$ to the big circle is parallel to the line $KL$.

The very same thing happens to point $K$ under the dilation that transforms $\Omega_2 \rightarrow \Omega$. It becomes some point $K'$ on $\Omega$ and tangent line at $K'$ is also parallel to $KL$.

Fortunately, the number of points on $\Omega$ with that property is very limited, namely 1 (actually 2 but we are interested only in those to the other side of $KL$ than $A$ and $B$). So considering that points inside each of the triples ($B$, $K$, $K'$) and ($A$, $L$, $L'$) are collinear, $BK$ and $AL$ meet on $\Omega$. Let's name that point $X$.

Great. One more line to go: $EE_1$. So we want to prove that radical axis of $\Omega_1$ and $\Omega_2$ passes through $X$. In other words we want to prove that $pow(X, \Omega_1) = pow(X, \Omega_2)$ or $XK\cdot XB = XL\cdot XA$. So points $B$, $K$, $L$ and $A$ should be concyclic.

And they are. It's a simple matter of spotting equal angles so I'll just leave the following figure as illustration of where to look for them.

enter image description here

ElThor
  • 912