11

I'm currently researching a particular dynamical system that is very geometric in nature. As part of this, I need to prove the following results (the second obviously implies the first). They are "obviously" true - but the proof doesn't seem so obvious.

  1. Equilateral triangles $ABC$ and $DEF$ are concentric. Points $A,G$ and $H$ are collinear. Points $B,H$ and $I$ are collinear. Let $DG=FI=x$ and $EH=y$. Prove that $x=y$.

enter image description here

  1. Now suppose the internal triangle $DEF$ is rotated about its centre by some small angle so that points $A$ and $G$ are still on opposite sides of line $DF$. Prove that the result is still true.

enter image description here

I have an algebraic solution to the first problem above. But I'm trying to find a clean geometric proof. I've made no progress on the second generalisation of the first result. I've tried all the usual tricks (congruence, rule of sines and cosines etc.) but to no avail. Does anyone have any approach to such problems that might work?

Auslander
  • 1,319
  • Aw, darn! My statement is actually very incorrect in general. (It's true, though, that $\overleftrightarrow{CI}$ contains $G$ in your first and second scenarios.) I'm not sure how I failed to see the failure. Oh, well ... I'm going to delete my comments to hide my shame. :) – Blue Feb 23 '15 at 05:49
  • No problems. Thanks for thinking about the problem in any case! – Auslander Feb 23 '15 at 10:57
  • Projective geometry? – Narasimham Feb 25 '15 at 12:42
  • @MonkeysUncle, yes, that would imply it. But I don't think it's clear how to show this. – Auslander Feb 25 '15 at 12:49
  • @David Yes I realized you could show GIH is 60 degrees, but hard to say anything about the other two angles. Or rather, you can't even do that can you. – MonkeysUncle Feb 25 '15 at 12:50
  • @Narasimham, I know little about projective geometry but I'll have a read and see if it yields anything. – Auslander Feb 25 '15 at 12:52
  • i think i have a proof using euclidean geometry. i will post it later when i get sometime this evening. – abel Feb 25 '15 at 17:50
  • Would love to see it @abel. Thanks for having a go in any case. – Auslander Feb 25 '15 at 19:52
  • i have an incomplete answer. – abel Feb 25 '15 at 23:37
  • Can you tell us what the dynamical system is? Sounds interesting. – Howcan Feb 26 '15 at 00:31
  • @Howcan, consider a point $X_1$ on line $DE$ of interior triangle. Then project $X_1$ to $X_2$ on $EF$ by line $AX_1$. Then project $X_2$ to point $X_3$ on line $FD$ by line $BX_2$. Repeat ad infinitum. The point bounces from side to side of the interior triangle. I need to show that the path always converges to an equilateral triangle. That means excluding other limit cycles, hence the problem I've posed here. – Auslander Feb 26 '15 at 01:39
  • @Narasimham, I've tried to turn my conjecture into a statement that avoids lengths so that projective geometry might be of use. Does this look correct? Equilateral triangles $ABC$ and $DEF$ have the same centre. $A$ and $G$ are on opposite sides of $FD$. Then If $A,G,H$ are collinear and $B,H,I$ are collinear then $C,I,G$ are collinear. Does this look correct? And what would the dual of this theorem be? – Auslander Feb 26 '15 at 23:16

4 Answers4

3

Proof of conjecture 2:

Let $M=AG\cap DF, N=BH\cap DE$ By Menelaus $\frac{FM}{MD}\frac{DG}{GE}\frac{EH}{HF}=1$ and $\frac{DN}{NE}\frac{EH}{HF}\frac{FI}{ID}=1$ but $IF=DG$ so $ID=GE$ and so $\frac{FM}{MD}=\frac{DN}{NE}$ but now since $DF=DE$ we have $MD=EN$. Now since $A,M,D$ are not collinear we have that $A\not=M$ and $B\not=N$ but from $DM=NE$ we have since $\angle ADM=180-\angle ADF=180-\angle BED=\angle BEN$ and $BE=AD$ that $BEN\cong ADM$ so $\angle BNE=\angle AMD$ so now $EN=DM,\angle HNE=\angle GMD,\angle NEH=120=\angle GDM$ so $ENH\cong DMG$ thus $y=x$. So we managed to obtain this thanks to $A,D,F$ not being collinear and in fact this is the only requirement. The proof is the same in other cases that have different placements of points because we can always get $AMD\cong BNE$. When $A,D,F$ are collinear we always have $DG=FI$ which we can easily see from Menelaus used in the proof.

Leader47
  • 500
  • Sry its $180-\angle ADF$ a type I hope everything else is correct do tell me if there are any more typos I edited this one. – Leader47 Feb 27 '15 at 17:45
  • The next line should also say 180 instead of 80. I think this proof is exactly what I was looking for. I doubt I'll see anything more elementary than this. Thanks for your time and help! – Auslander Feb 28 '15 at 04:23
  • I want to be able to acknowledge your assistance with this part of my work. Please provide me with means to do this. – Auslander Feb 28 '15 at 10:58
  • You can contact me via email [email protected] for anything you need, my full name is Zarko Randjelovic. – Leader47 Feb 28 '15 at 13:12
3

I'll approach this problem by relating the congruence of $\overline{DG}$, $\overline{EH}$, $\overline{FI}$ to how (or whether) $\overleftrightarrow{AG}$ contains $H$, $\overleftrightarrow{BH}$ contains $I$, and/or $\overleftrightarrow{CI}$ contains $G$.

My tool of choice is my Extended Ceva's Theorem, introduced in this answer, but reformulated here:

Extended Ceva's Theorem. Lines $\overleftrightarrow{A^+B^{-}}$, $\overleftrightarrow{B^+C^-}$, $\overleftrightarrow{C^+A^-}$ through the edges of $\triangle ABC$ concur if and only if $$\alpha^+ \beta^+ \gamma^+ + \alpha^-\beta^-\gamma^- + \alpha^+\alpha^- + \beta^+\beta^- +\gamma^+\gamma^- = 1 \tag{$\star$}$$

enter image description here

where $$\alpha^+ := \frac{|BA^+|}{|A^+ C|} \qquad \beta^+ := \frac{|CB^+|}{|B^+A|} \qquad \gamma^+ := \frac{|AC^+|}{|C^+B|}$$ $$\alpha^- := \frac{|CA^-|}{|A^- B|} \qquad \beta^- := \frac{|AB^-|}{|B^-C|} \qquad \gamma^- := \frac{|BC^-|}{|C^-A|}$$


Now for the problem at hand.

enter image description here

Extend the edges of $\triangle DEF$ to meet the sides of $\triangle ABC$ at points $P_i$ and $Q_i$; by symmetry, there are $p$ and $q$ such that $$ \frac{|BP_a|}{|P_aC|} = \frac{|CP_b|}{|P_bA|} = \frac{|AP_c|}{|P_cB|} = p \qquad \frac{|CQ_a|}{|Q_aB|} = \frac{|AQ_b|}{|Q_bC|} = \frac{|BQ_c|}{|Q_cA|} = q $$

Moreover, extend $\overline{AG}$, $\overline{BH}$, $\overline{CI}$ to meet the edges of $\triangle ABC$ at $A^\prime$, $B^\prime$, $C^\prime$, and define $$\alpha := \frac{|CA^\prime|}{|A^\prime B|} \qquad \beta := \frac{|AB^\prime|}{|B^\prime C|} \qquad \gamma := \frac{|BC^\prime|}{|C^\prime A|}$$

By the ECT with substitutions

$$A^+ \to P_a \qquad A^-\to A^\prime \qquad B^+\to C \qquad B^-\to Q_b \qquad C^+\to A \qquad C^-\to C^\prime$$ so that $$\alpha^+ = p \qquad \beta^+ = \gamma^+ = 0 \qquad \alpha^- = \alpha \qquad \beta^- = q \qquad \gamma^- = \gamma$$ we have that $\overleftrightarrow{CI}$ contains $G$ (which is to say that $\overleftrightarrow{AG}$, $\overleftrightarrow{CI}$, $\overleftrightarrow{DE}$ concur at $G$) if and only if $$1 = \alpha\;(\; p + q \gamma \;) \tag{$1$}$$

Likewise, $\overleftrightarrow{AG}$ contains $H$ if and only if $$1 = \beta\;(\; p + q \alpha \;) \tag{$2$}$$ and $\overleftrightarrow{BH}$ contains $I$ if and only if $$1 = \gamma\;(\;p + q \beta \;) \tag{$3$}$$

Now, the problem asserts that $\overleftrightarrow{AG}$ and $\overleftrightarrow{BH}$ contain $H$ and $I$, respectively; therefore, $(2)$ and $(3)$ hold. The problem also asserts that $\overline{DG}\cong\overline{FI}$, so that $\overline{AC^\prime}\cong\overline{CB^\prime}$, which implies $\alpha = \gamma$. Provided that $p\neq 0$, we may conclude that $\alpha = \beta = \gamma$, that all three concurrencies hold, and that $\overline{DG}\cong\overline{EH}\cong\overline{FI}$.

Note. When $p=0$, the figure is such that $\overleftrightarrow{DF}$ contains $A$ (and so forth for other edge-lines and vertices). Here, we need not assume $\overline{DG}\cong\overline{FI}$, as the fact that $\alpha = \gamma$ (and therefore that congruence) follows from assuming $(1)$ and $(2)$ (with $\beta \neq 0$). There's no guarantee, though, that $(1)$ holds. @Ewan's answer gives a case in which it doesn't.

Note. When $q=0$, the figure is such that $\overleftrightarrow{DE}$ contains $A$ (and so forth). By the problem's construction, any point $G$ on $\overline{DE}$ causes $H$ and $E$ to coincide, whereupon $I$ and $F$ also coincide; to satisfy the condition $\overline{DG}\cong\overline{FI}$, we must have that $G$ and $D$ coincide, as well. This, obviously, satisfies all the concurrencies and gives that $\alpha = \beta = \gamma$.

Blue
  • 75,673
1

let the two concentric equilateral triangles be $ABC, DEF$ are placed so that $DEF$ is inside $ABC.$ the three points $G, H, E$ are on the sides $DE, EF$ and $FD$ respectively. we also have the points $A, G, H$ collinear, a$B, H, I$ collinear and $DG = IF.$ let $O$ be the center of the two circumcircles.

claim 1: the triangle $GHI$ is an equilateral triangle.

proof: consider the triangles $DGO, FIO.$ they are congruent because $$\angle ODG = \angle OFI = 30^\circ, OD = OF = radius, DG = IF. \implies OG = OI, \angle OID + \angle OGD = 180^\circ$$

this also gives us $$\angle GOI = 120^\circ.$$ by looking at the isosceles triangle $OIG,$ and using the fact $\angle IOG = 120^\circ,$ we can see that $$\angle OIC = \angle OGI = 30^\circ,\, IG \parallel OF.$$

let $H'$ be the unique point on $EF$ such that $IGH'$is an equilateral triangle. the point $H'$ is unique because the base $GH$ of the equilateral triangle which $O$ as the circumcenter is fixed.

it seems clear, but i need to prove that this $H' = H.$

abel
  • 29,170
  • Hi @abel. Thanks for your attempt. There's a problem in your proof, unfortunately. Where you write that $\angle{OIC}=\angle{OGI}$, you've assumed that points $C,I$ and $G$ are collinear. Unfortunately we don't know that this is the case (until we've actually proved the result!). – Auslander Feb 26 '15 at 00:07
  • @David, i did not assume that. in fact you can forget about $ABC.$ start from the inside triangle $DEF$ and the fixed points $G, I$ equal distance from $D$ and $F.$ construct the $H'$. now extend $H'G, GI,$ and $IH'$ to meet the outer circle at $A, B$ and $C.$ – abel Feb 26 '15 at 00:14
  • Hi @abel. But how do you know that $H'G$ will intersect the outer circle at point $A$? You only know this if $H=H'$, and this is what we need to prove. – Auslander Feb 26 '15 at 01:54
  • what i am doing is, given the circumcenter $O$ of the equilateral triangle $DEF$ and two points $G, I$ on $DE$ and $DF$ such that $DG = IF,$ constructing an equilateral triangle $ABC$ concentric with $DEF$ and has the property that $A,G,H$ are collinear and $C,I,G$ and $B,H, I$ are collinear. – abel Feb 26 '15 at 02:01
  • Thanks for clarifying @abel. I get that. But then the triangle you've constructed doesn't necessarily agree with the triangle given. But I guess you're aware of that. – Auslander Feb 26 '15 at 02:07
  • @David, but the triangle (say) $A'B'C'$ i constructed is unique. so that means if we know the circle that $ABC$ is on, then we are done. do you see what i mean? – abel Feb 26 '15 at 02:09
  • Sorry if I'm being infuriating, @abel. I see that your triangle $A'B'C'$ is unique by construction. But I don't see that that means that it is equal to triangle $ABC$. They're on the same circle, yes. – Auslander Feb 26 '15 at 02:36
1

The generalized property is false (unfortunately). Here is a counterexample with simple coordinates :

$$ \begin{array}{llllll} A & (1,\frac{5}{\sqrt{3}}) & B & (2,-\frac{4}{\sqrt{3}}) & C & (-3,-\frac{1}{\sqrt{3}}) \\ D & (0,\frac{2}{\sqrt{3}}) & E & (1,-\frac{1}{\sqrt{3}}) & F & (-1,-\frac{1}{\sqrt{3}}) \\ G & (\frac{1}{3},\frac{1}{\sqrt{3}}) & H & (0,-\frac{1}{\sqrt{3}}) & I & (-\frac{2}{\sqrt{3}},0) \\ \end{array} $$

It is straightforward to check that with those values, $ABC$ and $DEF$ are both equilateral and centered at the origin, that $H$ is the middel of $[EF]$ and that

$$ \overrightarrow{HA}=3\overrightarrow{HG}, \overrightarrow{DE}=3\overrightarrow{DG}, \overrightarrow{BI}=\frac{4}{3}\overrightarrow{BH}, \overrightarrow{FD}=3\overrightarrow{FI}. $$

Here is the corresponding figure :enter image description here

Ewan Delanoy
  • 61,600
  • I forgot to mention a key assumption. Sorry, that is completely my fault, and I thank you for your efforts. Points $A$ and $G$ are on opposite sides of line $DF$. I've amended the question to reflect this. Sorry again! – Auslander Feb 26 '15 at 20:17
  • @David I just noticed that $D$ is the middle of $[AF]$ in my example – Ewan Delanoy Feb 26 '15 at 21:09
  • Right. What does this mean about the truth of my conjecture? Are you suggesting that it's still false unless I add further restrictions? – Auslander Feb 26 '15 at 21:19
  • I just did a little simulation of my own and it appears that if $A$ is anywhere on line $DF$, then if you change the position of $G$, $DG$ is always equal to $FI$. Otherwise the conjecture still looks true... – Auslander Feb 26 '15 at 22:19