10

Is the following assertion true? How can it be proved? I'm interested in any proofs, but in particular interested if it can be proven synthetically (i.e. without coordinates).

Call the endpoints of the major axis $A$ and $A'$. For any point $P$ on the ellipse, let drop the perpendicular from $P$ to the major axis at $Q$. Then $$\frac{|PQ|^2}{|AQ||A'Q|}$$ is a constant (and indeed equal to the square of the ratio of the major axis to the minor axis).

Background: This fact is used in a synthetic proof that Archimedes' Trammel draws an ellipse. Archimedes proof is lost, but, if we can prove this fact, ideally using synthetic (Euclidean) geometry, we can reconstruct a proof.


For background, an analytic proof is simple. Given an ellipse with center at the origin $\frac {x^2}{a^2} + \frac{y^2}{b^2} = 1$, the claim is that $$x^2 = k \cdot (b-y)(b+y)$$ which follows immediately by setting $k := \frac{a^2}{b^2}$.

SRobertJames
  • 6,117
  • 1
  • 12
  • 40
  • 2
    Drew up a quick Desmos demo for the interested, and it seems to hold true for a fixed ellipse regardless of the point $P$ chosen on it. Not sure offhand how to prove it without coordinates, but I haven't given it much thought. (Indeed, I was initially doubtful in the first place.) – PrincessEev Jan 13 '25 at 02:46
  • 1
    What is the synthetic definition of an ellipse? – Pranay Jan 13 '25 at 05:36
  • 3
    This is true in particular case where the ellipse becomes a circle $a=b = R$ and we have $h^2=AH\times HB$ where $h$ is the altitude and $H$ is the foot of of the altitude in right angle triangle the diameter of the circle is its base.Usually the facts about the circles are true for the compressed circles which are ellipses. – sirous Jan 13 '25 at 07:51
  • 1
    @sirous In a circle, $|PQ|$ is the geometric mean of $|AQ|, |A'Q|$. So this seems to be saying that in an ellipse, $|PQ|$ is a fixed multiple of that geometric mean. – SRobertJames Jan 13 '25 at 18:47

3 Answers3

7

Let $a$ and $b$ be the major and minor semi-axis lengths of the ellipse. Construct a circle of radius $a$ about the center of the ellipse, so that $AA'$ is a diameter of the circle. Extend $QP$ to intersect the circle at $P'.$

Then $$\frac{\lvert PQ\rvert}{\lvert P'Q\rvert} = \frac ba$$ and $$\lvert P'Q\rvert^2 = \lvert AQ\rvert\lvert A'Q\rvert$$

from which it follows that

$$\frac{\lvert PQ\rvert^2}{\lvert AQ\rvert\lvert A'Q\rvert} = \frac{b^2}{a^2}.$$

David K
  • 108,155
  • 1
    What is the synthetic definition of an ellipse? Is it defined by the first equation in your answer? – Pranay Jan 13 '25 at 06:32
  • 2
    @Pranay I wouldn't expect an ellipse to be defined that way. But it's well known to be a stretched circle, so I'm assuming a synthetic proof of that. Ultimately we could go all the way back to the axioms; how far is far enough? – David K Jan 13 '25 at 08:57
  • 2
    There are some useful hints about synthetic geometry of an ellipse, in particular why it is a stretched circle, in https://math.stackexchange.com/a/3740889/139123 – David K Jan 13 '25 at 09:05
  • @Pranay - There are two common synthetic definitions for the ellipse. If you are willing to involve 3D geometry, an ellipse is the intersection of a cone by a plane, where the angle between the plane and the cone centerline is greater than the apex angle. Using just plane geometry, given a line (the "directrix" $\ell$) and a point not on line (the "focus" $F$), and a real number $0 < e < 1$ (the "eccentricity"), an ellipse is the locus of all points $P$ such that the ratio of the distance from $P$ to $F$ to the distance from $P$ to $\ell$ is $e$. – Paul Sinclair Jan 13 '25 at 17:47
  • @PaulSinclair thanks! I also know that it’s the locus of all points whose sum of distances from the two foci is a fixed constant. I was wondering if all these different definitions can be shown to be equivalent using only synthetic geometry. I’m guessing this is true but I don’t know how. – Pranay Jan 13 '25 at 17:55
  • 3
    @Pranay - All of these were known to the ancient Greeks. But in fact, analytic geometry can be (and originally was) developed from synthetic geometry, coming from a sickly schoolboy named Descartes watching a fly crawl around the ceiling and realizing that its location could be precisely determined by its distance from two of the walls. So yes, anything provable by analytic geometry is also provable by synthetic geometry (and vice versa). – Paul Sinclair Jan 13 '25 at 18:03
  • 2
    @Pranay - As for how, there is a beautiful 3Blue1Brown video relating the two-foci and conic section definitions of an ellipse by fitting maximal spheres in the cone above and below the cutting plane, and tangent to it. The points of tangency to the plane turn out to be the foci, and the constancy of the sum of distances an artifact of the tangency of the spheres to the cone (which are both circles at fixed distances from the apex). I am not able to look up the video link right now. – Paul Sinclair Jan 13 '25 at 18:15
  • How do you know the first equation? (The second comes from the fact that $PQ \perp AA'$ and the Intersecting Chords Theorem.) – SRobertJames Jan 13 '25 at 19:15
  • @PaulSinclair (et al.): Those maximal spheres are known as Dandelin spheres. – Blue Jan 13 '25 at 19:17
  • 1
    @Blue - Thank you. I knew there was a name for it, but couldn't recall it. – Paul Sinclair Jan 13 '25 at 19:40
  • 1
    The first equation comes from viewing the ellipse as a "stretched circle," in this case actually compressed rather than stretched in the direction parallel to the minor axis. The semi-minor axis itself is one of the possible half-chords $PQ$ and this sets the ratio of $PQ$ to $P'Q.$ – David K Jan 14 '25 at 00:12
6

An ellipse $ABP$ can be defined as the intersection between a right circular cone and a plane. The plane $VAB$ through the axis of the cone and perpendicular to the plane of the ellipse intersects the ellipse at major axis $AB=2a$ (see figure below).

Consider then a plane through $P$ parallel to the base of the cone (and thus perpendicular to plane $VAB$), intersecting the cone at circle $A'B'P$, where $AA'$ and $BB'$ are two generatrices of the cone, and intersecting $AB$ at $H$. Line $PH$ is perpendicular to plane $VAB$ and is thus perpendicular to both $AB$ and $A'B'$.

By the intersecting chords theorem we know that $PH^2=A'H\cdot B'H$. But, on the other hand, we have by similitude: $$ A'H:AH=BC:AB, \quad\hbox{that is:}\quad A'H={BC\over AB}\cdot AH; $$ and: $$ B'H:BH=AD:AB, \quad\hbox{that is:}\quad B'H={AD\over AB}\cdot BH. $$

Plugging these into the previous formula we obtain:

$$ \tag1 PH^2={AD\cdot BC\over 4a^2}\,AH\cdot BH. $$

In particular, if we choose plane $A'B'P$ midway between $A$ and $B$, so that $H$ is the midpoint of $AB$ and $AH=BH=a$, while $PH=b$ is the minor axis of the ellipse, we get:

$$ b^2={AD\cdot BC\over 4a^2}\,a^2, \quad\implies\quad AD\cdot BC=4b^2. $$

We can then rewrite $(1)$ as:

$$ \tag2 PH^2={b^2\over a^2}\,AH\cdot BH. $$

enter image description here

  • Applying the Intersecting Chords Theorem requires knowing that $PH \perp A'B'$. How do you know that is so? – SRobertJames Jan 13 '25 at 18:58
  • I'm having trouble "seeing" point $H$. It's clear that the parallel plane must intersect $AB$ within the cone, and we can call that point $H$, so I don't doubt the correctness of what you write. But I'm nonetheless having trouble "seeing" how the parallel plane will intersect $AB$ at $H$, or "finding" this $H$ visually. Can you help? – SRobertJames Jan 13 '25 at 18:59
  • $VAB$ is the plane passing through the axis of the cone and perpendicular to the plane $ABP$ of the ellipse. Plane $A'B'P$ is also perpendicular to plane $VAB$. Hence the intersection $PH$ of planes $ABP$ and $A'B'P$ is perpendicular to plane $VAB$, and in particular it is perpendicular to both lines $AB$ and $A'B'$. – Intelligenti pauca Jan 13 '25 at 19:12
  • @SRobertJames I edited my answer adding the requested explanations. – Intelligenti pauca Jan 14 '25 at 16:20
1

enter image description here

In circle $c_1$ we have:

$IQ^2=AQ\cdot A'Q\space\space\space\space\space\space(1)$

Triangles SIQ and SRO are similar,Also $\triangle SPQ\sim\triangle SFO$ so we have:

$\frac{PQ}{FO}=\frac{SQ}{SO}$

$\frac{IQ}{RO}=\frac{SQ}{SO}$

$\Rightarrow \frac {PQ}{FO}=\frac {IQ}{RO}\space\space\space\space\space\space(2)$

$\frac {PQ}{IQ}=\frac{FO}{RO}\space\space\space\space\space\space(3)$

$RO=a$

$FO=b$

$\Rightarrow \frac{PQ^2}{b^2}=\frac{IQ^2}{a^2}\space\space\space\space\space\space(4)$

$\Rightarrow PQ^2=\frac{IQ^2}{a^2}\cdot b^2\space\space\space\space\space\space(5)$

From (1) and (5) we get:

$\frac {PQ^2}{AQ\cdot A'Q}=\frac{b^2}{a^2}$

sirous
  • 12,694
  • Should the first line read $A'Q$, not $AQ'$? And the second $SRO$, not $SRQ$. I can't figure out what $(2)$ is supposed to be. As is, how does $FQ$ connect to the similar triangles? You haven't even mentioned $F$. – SRobertJames Jan 14 '25 at 13:16
  • @SRobertJames, Sorry about numerous errors.This is because:1- I am old. 2- I have trouble with my eyes. 3-I was too tired for this job. Anyway I corrected the error. – sirous Jan 14 '25 at 16:16
  • Thanks. But you still haven't explained (2) or how you reason from similar triangles to point $F$. Is there a typo in (2)? – SRobertJames Jan 14 '25 at 16:56
  • @SRobertJames, I explained the reason for (2). – sirous Jan 14 '25 at 20:38