3
$\begingroup$

While solving a book on ellipses, I came across the following property of an ellipse which was given without proof :-

If the normals be drawn at the extremities of a focal chord of an ellipse, a line through their point of intersection parallel to the major axis will bisect the chord.

Working with the standard ellipse, i.e., $$ \frac {x^2}{a^2}+\frac {y^2}{b^2}=1,$$we find that the normal at a point $(a\cos\theta,b\sin\theta)$ is given by $$ N(\theta)= a\, x \, \sec \theta- b\, y \csc \theta=a^2-b^2 $$

The midpoint of $2$ parametric points $\theta$ and $\phi$ is

$$ \left(\frac a2(\cos\theta+\cos\phi),\frac b2(\sin\theta+\sin\phi)\right)\equiv \left(a\cos \frac {\theta+\phi}2\cos \frac {\theta-\phi}2,b\sin \frac {\theta+\phi}2\cos \frac {\theta-\phi}2\right)$$

Since the chord is a focal chord, we have the relation that relates $\theta$ and $\phi$, i.e.,

$$\tan \frac {\theta}2\tan \frac {\phi}2=\frac {e-1}{e+1}$$

Hence the line through the midpoint parallel to the major axis is

$$y=b\sin \frac {\theta+\phi}2\cos \frac {\theta-\phi}2$$

But proceeding after this (finding the intersection of the normals at the parametric points) becomes very complicated. Is there a smarter and easier way to prove the given property except using brute force?

$\endgroup$
  • $\begingroup$ yeah,the edit is ok but some answers would be appreciated $\endgroup$ – Abhishek Bakshi Feb 27 '15 at 13:47
3
$\begingroup$

Unlike the one I gave in this, this is a "pure geometric solution".

enter image description here

The figure is pretty much self-explanatory but still I will spell it out:

  1. $P$ and $Q$ are points on the ellipse.
  2. $PR$ and $QR$ are the tangents at these points.
  3. $PC$ and $QC$ are the normals at these points.
  4. Line $AB$ is a directrix.
  5. $F_1$ is a focus.
  6. $CD$ is drawn parallel to the major axis.

The angles of same color are equal(excuse the right angles from this rule), but for those who are color-blind or otherwise incapable of distinguishing the colors, I will spell out which angles are equal(you can figure out why):

  1. $\angle PRA=\angle PCD=\alpha$
  2. $\angle QRB=\angle QCD=\beta$
  3. $\angle CPD=\angle PRF_1=\theta$
  4. $\angle CQD=\angle QRF_1=\gamma$

Since this is an ellipse, $\dfrac{PF_1}{PA}=\dfrac{QF_1}{QB}=e\tag{i}$

Using basic trigonometry, we can write $PF_1=PR\sin\theta$,$\,PA=PR\sin\alpha$,$\,QF_1=QR\sin\gamma$ and $QB=QR\sin\beta$. Now substituting these in (i):

$$\frac{\sin\theta}{\sin\alpha}=\frac{\sin\gamma}{\sin\beta}\tag{ii}$$

Applying sine rule in $\Delta CQD$: $\dfrac{CD}{QD}=\dfrac{\sin\gamma}{\sin\beta}\tag{iii}$

Applying sine rule in $\Delta CPD$: $\dfrac{CD}{PD}=\dfrac{\sin\theta}{\sin\alpha}\tag{iv}$

Dividing (iii) and (iv), the ratio $PD:QD=\dfrac{\sin\gamma\sin\alpha}{\sin\beta\sin\theta}$

Now what does statement (ii) tell you about this ratio?

$\endgroup$
2
$\begingroup$

This exercise can be understood as an application of a general result about perimeter bisectors of triangles.

Proposition. Given $\triangle ABC$ with incircle $\bigcirc I$ meeting the edges at $D$, $E$, $F$ as shown. If $F^\prime$ is the point opposite $F$ in $\bigcirc I$, and if $F^{\prime\prime}$ is the point where $\overleftrightarrow{CF^\prime}$ meets $\overline{AB}$, then $$|\overline{CA}|+|\overline{AF^{\prime\prime}}| = |\overline{CB}|+|\overline{BF^{\prime\prime}}| \tag{$\star$}$$ so that $\overline{CF^{\prime\prime}}$ is a perimeter bisector of $\triangle ABC$.

enter image description here

Proof of Proposition. Let the perpendicular to $\overline{FF^\prime}$ at $F^\prime$ meet the edges of the triangle at $A^\prime$ and $B^\prime$. By tangent properties of circles, we have $$\overline{CE}\cong\overline{CD} \qquad \overline{A^\prime E}\cong\overline{A^\prime F^\prime} \qquad \overline{B^\prime D}\cong\overline{B^\prime F^\prime}$$ Consequently, $|\overline{CA^\prime}| + |\overline{A^\prime F^\prime}| = |\overline{CB^\prime}| + |\overline{B^\prime F^\prime}|$, so that $\overline{CF}$ is a perimeter bisector of $\triangle A^\prime B^\prime C$. The Proposition holds by the similarity of $\triangle ABC$ and $\triangle A^\prime B^\prime C$. $\square$


The Proposition has a helpful corollary.

Corollary. Given $\triangle ABC$ with incenter $I$ and perimeter bisector $\overline{CF^{\prime\prime}}$, if $M$ is on $\overline{AB}$ such that $\overline{IM} \parallel \overline{CF^{\prime\prime}}$, then $M$ is the midpoint of $\overline{AB}$.

enter image description here

Proof of Corollary. The points of tangency of the triangle with its incircle separate the perimeter into three pairs of congruent segments, marked $a$, $b$, $c$. Thus, the semi-perimeter of $\triangle ABC$ is $a+b+c$, and since $|\overline{BC}| = b+c$, it follows that $|\overline{BF^{\prime\prime}}| = a = |\overline{AF}|$. Thus, $\overline{FF^{\prime\prime}}$ lies between congruent segments. In $\triangle FF^\prime F^{\prime\prime}$, segment $\overline{IM}$ passes through the midpoint of one side ($\overline{FF^\prime}$) and is parallel to another ($\overline{F^\prime F^{\prime\prime}}$); it necessarily meets the third side ($\overline{FF^{\prime\prime}}$) at its midpoint, which must also be the midpoint of $\overline{AB}$. $\square$


To solve the original problem, it basically suffices to embed the above triangle into an ellipse:

enter image description here

In the above, the ellipse's foci are $C$ and $F^{\prime\prime}$, and $\overline{AB}$ is a chord through the latter. The fundamental nature of ellipses implies that $(\star)$ holds; therefore, $\overline{CF^{\prime\prime}}$ is a perimeter bisector of $\triangle ABC$. Moreover, the reflection property of ellipses implies that normals at $A$ and $B$ bisect angles $\angle CAF^{\prime\prime}$ and $\angle CBF^{\prime\prime}$; therefore, the intersection of these normals is the incenter of $\triangle ABC$. The result follows by the Corollary. $\square$

$\endgroup$
1
$\begingroup$

Referring to G-man's sketch, $ \epsilon = eccentricity, RQB = \phi , F_1QR =\psi, then \frac {\cos\psi }{\cos\phi} = \epsilon $

The proof is very near from here.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.