previous up next contents
Previous: 2 Apexes and pencils Up: Viewing and Touching the Next: 4 Some constructions   Contents

3 Orthogonality

When two ordinary circles $ \pi_{1},\,\pi_{2}$ intersect at right angle, a direct application of Pythagoras theorem leads to $ \left\Vert \omega\omega_{2}\right\Vert ^{2}-r_{1}^{2}-r_{2}^{2}=0$. When trying to restate this relation into a $ \mathbb{P}_{\mathbb{R}}\mathbb{R}^{4}$ relation by using (2.1), we obtain the following :

Proposition 3.1   Define the Minkowski metric over $ \mathbb{R}^{4}$ by :

$\displaystyle mink\left(X_{1},\,X_{2}\right)=a_{1}a_{2}+b_{1}b_{2}+c_{1}c_{2}-d_{1}d_{2}=^{\mathbf{t}}\!X_{2}\,.\,Mink\,.\,X_{1}$ (3.1)

Then $ mink\left(X_{1},\,X_{2}\right)=0$ is a characteristic property of orthogonal cycles.

Proof. Applied to the various types of cycles, we have :
$\displaystyle \left(circle,\, circle\right)$ $\displaystyle :$ $\displaystyle -2\,mink\left(X_{1},\,X_{2}\right)=k_{1}k_{2}\left(\left\Vert \omega\omega_{2}\right\Vert ^{2}-r_{1}^{2}-r_{2}^{2}\right)$  
$\displaystyle \left(line,\, circle\right)$ $\displaystyle :$ $\displaystyle mink\left(X_{1},\,X_{2}\right)=k_{1}k_{2}\left(x_{2}\cos\tau_{1}+y_{2}\sin\tau_{1}+p_{1}\right)$  
$\displaystyle \left(line,\, line\right)$ $\displaystyle :$ $\displaystyle mink\left(X_{1},\,X_{2}\right)=k_{1}\, k_{2}\cos\left(\tau_{1}-\tau_{2}\right)$ (3.2)

where the normalizing factor $ k_{i}$ is $ c_{i}+d_{i}$ for a circle and $ \sqrt{a_{i}^{2}+b_{i}^{2}}$ for a line. $ \qedsymbol$

Notation 3.2   The light cone $ \mathbb{S}_{Mink}$ in Minkowski space $ \mathbb{P}_{\mathbb{R}}\mathbb{R}^{4}$ has equation $ a^{2}+b^{2}+c^{2}-d^{2}=0$. It doesnt contains any point at infinity. Therefore, it can be identified with the unit sphere $ \mathbb{S}_{Eucl}$ in $ \mathbb{R}^{3}\times\left\{ 1\right\} $ and $ \mathcal{S}_{j}\in\mathbb{S}_{Mink}$, $ S_{j}\in\mathbb{S}_{Eucl}$ will often be shortened into $ \mathcal{S}_{j}\in\mathbb{S}$, $ S_{j}\in\mathbb{S}$.

Theorem 3.3 (Orthogonality)   We have the following results :
  1. Point $ \omega_{1}$ belongs to cycle $ \pi_{0}$ iff the point-circle $ \left\{ \omega_{1}\right\} $ is orthogonal to $ \pi_{0}$.
  2. The bundle of all the cycles $ \pi_{j}$ orthogonal to a given cycle $ \pi_{0}$ is characterized by : $ \mathcal{A}_{j}$ belongs to the polar plane of $ \mathcal{A}_{0}$ relative to the light cone $ \mathbb{S}_{Mink}$. When nothing is at infinity, point $ A_{j}$ belongs to the polar plane of $ A_{0}$ relative to the unit sphere $ \mathbb{S}_{Eucl}$.
  3. The set $ shadow\left(\pi_{0}\right)=\left\{ shadow\left(\omega\right)\mid\omega\in\pi_{0}\right\} $ is the circle where the polar plane of $ \mathcal{A}_{0}$ intersects $ \mathbb{S}$. Conversely, as in Fig. 1, $ \mathcal{A}_{0}$ is the vertex of the circular cone that is tangent to $ \mathbb{S}$ along the circle $ shadow\left(\pi_{0}\right)$.
  4. Cycles $ \pi$ orthogonal to two different cycles $ \pi_{1}$ and $ \pi_{2}$ form a pencil whose apex line is the polar respective to $ \mathbb{S}$ of the apex line of pencil $ P\left(\pi_{1},\pi_{2}\right)$.

Proof. Straightforward from(3.1) and (3.2). $ \qedsymbol$

Remark 3.4   Real cycle $ \pi_{0}$ is orthogonal to the imaginary circle $ \pi_{1}$ ( $ r_{1}^{2}<0$) when $ \pi_{0}$ cuts the real circle $ \pi_{2}=\pi\left(\omega_{1},\,\left\vert r_{1}\right\vert\right)$ along a diameter of $ \pi_{2}$.

Proposition 3.5   In $ \mathbb{R}^{4}$, the orthogonal projector onto the space whose projective is the polar plane of $ apex\left(\pi_{0}\right)$ is $ id-p$ while projector $ p$ onto $ Vect\left(X_{0}\right)$ is :

$\displaystyle p\left(X\right)=\left(mink\left(X,\,X_{0}\right)/mink\left(X_{0},\,X_{0}\right)\right)\, X_{0}$

Proof. Both $ p\circ p=p$ and $ mink\left(p\left(X\right),\,X-p\left(X\right)\right)=0$ are straightforward. $ \qedsymbol$

Remark 3.6   A matrix formulation of the preceding is

$\displaystyle \boxed{p}=\frac{1}{mink\left(X_{0},\,X_{0}\right)}\, X_{0}\,.\,^{\mathbf{t}}\!X_{0}\,.\,Mink$

where matrix $ Mink$ is defined in (1.3).

Definition 3.7   The Gram matrix $ \operatorname{G}_{p,q,\cdots,r}$ of $ X_{p},\, X_{q},\cdots,\, X_{r}\in\mathbb{R}^{4}$ is the matrix of all the Minkowski products. In this context, notation $ W_{pq}=mink\left(X_{p},\,X_{q}\right)$ and $ w_{p}^{2}=mink\left(X_{p},\,X_{q}\right)$ will be used, leading to 

$\displaystyle \operatorname{G}_{pq}=\left(\begin{array}{cc}
w_{p}^{2} & W_{pq}\\
W_{pq} & w_{q}^{2}\end{array}\right)$

Proposition 3.8   Two cycles $ \pi_{1},\pi_{2}$ are (respectively) secant, tangent or external when $ \operatorname{signum}\det\operatorname{G}_{12}$ is (respectively) $ +1$, 0 or $ -1$.

Proof. Necessary to understand Theorem 3.9. Will be proved as a Corollary. $ \qedsymbol$

Theorem 3.9 (Classification)   Quantity $ \sigma\left(P\right)\doteq\operatorname{signum}\det\operatorname{G}_{12}$ is the same for any choice of two cycles $ \pi_{1}\neq\pi_{2}$ in a given pencil P. According to this value, we have the following classification : When $ P$ is a tangent pencil, so is $ P^{\perp}$ (using $ \omega_{0}$ and $ \Delta_{1}^{\perp}$ orthogonal to $ \Delta_{1}$ at $ \omega_{0}$). When $ P$ is $ isoptic\left(\omega_{1},\omega_{2}\right)$ then $ P^{\perp}$ is $ isotomic\left(\omega_{1},\omega_{2}\right)$ and conversely.

Proof. To see that $ \sigma\left(P\right)$ is well defined, consider $ X_{3}=k_{13}X_{1}+k_{23}X_{2},\, X_{4}=k_{14}X_{1}+k_{24}X_{2}$. Then $ -G_{12}$ is the discriminant of $ mink\left(X_{3},\,X_{3}\right)$, i.e. the condition for the apex line of $ P$ cuts $ \mathbb{S}$ : this will not change when choosing two other points on this line. Another proof is $ \operatorname{G}_{34}=^{\mathbf{t}}\!\left(k\right)\,.\,\operatorname{G}_{12}\,.\,\left(k\right)$ while $ \pi_{3}\neq\pi_{4}$ implies $ \det\left(k\right)\neq0$.

When apex line of $ P$ cuts $ \mathbb{S}$ in two points $ S_{1},\, S_{2}$, then $ P$ is $ isotomic\left(\omega_{1},\omega_{2}\right)$ by the very definition of a pencil. When $ apex\left(P\right)$ is tangent to $ \mathbb{S}$ at $ S_{0}\neq S_{\infty}$, this line cuts the south plane at some $ \mathcal{A}_{1}=apex\left(\Delta_{1}\right)$, $ \omega_{0}\in\Delta_{1}$ and $ P$ is a tangent pencil. When $ apex\left(P\right)\cap\mathbb{S}=\emptyset$, the polar of $ apex\left(P\right)$ cuts $ \mathbb{S}$ in two points $ S_{1},\, S_{2}$ and $ P^{\perp}=isotomic\left(\omega_{1},\omega_{2}\right)$. For each $ \pi\in P$ we have $ \left\{ \omega_{i}\right\} \perp\pi$ for $ i=1,2$ so that $ \omega_{i}\in\pi$ while the converse is clear. $ \qedsymbol$

Remark 3.10   Pencils are sometimes described as "elliptic" or "hyperbolic". Topologically, any quadric is connected in a projective space and therefore elliptic. Moreover, pencils are linear, i.e. are straight lines and not quadrics. Using Apollonius as eponym for the $ isotomic$ circles (English use) is questionnable since quite all circles are Apollonian, while using Poncelet (French use) is questionnable since Poncelet's article is .

Proposition 3.11   Any $ \psi\in\mathbb{G}_{circ}$ transforms not only cycles into cycles (Proposition 1.8) but also orthogonal pencils into orthogonal pencils.

Proof. A pencil $ P$ is characterized by two cycles that belong to $ P^{\perp}$. But homographies are conformal by differentiability while the property is obvious for $ z\mapsto\overline{z}$. $ \qedsymbol$

Proposition 3.12   When $ P$ and $ P^{\perp}$ are tangent pencils, they are the images of the Cartesian coordinate lines under the transformation $ z\mapsto\omega_{0}+\exp i\left(\tau_{1}+\pi/2\right)/\overline{z}$ (notations of Theorem 3.9).

Theorem 3.13 (Bifocal coords)   Pencils $ isotomic\left(\omega_{1},\,\omega_{2}\right)$ and $ isoptic\left(\omega_{1},\,\omega_{2}\right)$ are the images, under a suitable homography $ \psi$, of the polar coordinates grid. More precisely, coordinate circle $ \left\vert\zeta\right\vert=\rho$ is transformed into the $ isotomic\left(\rho\right)$ cycle characterized by $ \left\Vert \omega\,\omega_{2}\right\Vert /\left\Vert \omega\,\omega_{1}\right\Vert =\rho$ while a coordinate ray, i.e. half a line from 0, characterized by $ \zeta/\left\vert\zeta\right\vert=\exp i\vartheta$, goes onto "half" an $ isoptic$ cycle. This $ isoptic\left(\vartheta\right)$ arc, bounded by $ \omega_{1}$ and $ \omega_{2}$, is characterized by $ \left(\overrightarrow{\omega\,\omega_{1}},\,\overrightarrow{\omega\,\omega_{2}}\right)=\vartheta$ .

Proof. Use $ \psi$ defined by $ 0\mapsto\omega_{2},\,\infty\mapsto\omega_{1},\,1\mapsto\infty$ and consider $ \omega_{0}\doteq\psi\left(\zeta\right)$ where $ \zeta_{0}$ is the ordinary point $ \zeta_{0}=\rho_{0}\,\exp i\vartheta_{0}$. By Proposition 1.8, we have :

$\displaystyle \zeta_{0}=\rho_{0}\exp i\vartheta_{0}=\gamma\left(0,\,\infty,\,\r...
...,\omega_{0},\,\infty\right)=\frac{\omega_{2}-\omega_{0}}{\omega_{1}-\omega_{0}}$ (3.3)

Let $ \omega$ be any point on the whole $ isoptic$ cycle through $ \omega_{1},\omega_{2}$ and $ \omega_{0}$ except from the basis points. By Proposition 1.7, $ \omega$ comes from a $ \zeta$ such that $ \rho\exp i\vartheta=\mu\,\rho_{0}\exp i\vartheta_{0}$ where $ \mu$ is real. Therefore, we have either $ \zeta=\rho\exp i\vartheta_{0}$ or $ \zeta=-\rho\exp i\vartheta_{0}=\rho\exp i\left(\vartheta_{0}+\pi\right)$. In the first case, $ \zeta$ and $ \zeta_{o}$ are on the same half line bounded by 0 and $ \infty$ and, by continuity $ \omega_{0}$ and $ \omega$ are on the same arc bounded by $ \omega_{1}$ and $ \omega_{2}$. From now on this arc will be referred as the $ isoptic\left(\vartheta\right)$ arc while the union of both arcs will be referred as the $ isoptic\left(\vartheta+k\pi\right)$ cycle.

By Proposition 3.11, each cycle $ \pi$ of the $ isotomic\left(\omega_{1},\,\omega_{2}\right)$ pencil is the image by $ \psi$ of a circle $ \left\vert\zeta\right\vert=\rho$, and thus characterized by the value of $ \rho$, while $ \left\Vert \omega\,\omega_{2}\right\Vert /\left\Vert \omega\,\omega_{1}\right\...
...rt\left(\omega_{2}-\omega\right)/\left(\omega_{1}-\omega\right)\right\vert=\rho$ comes from (3.3). $ \qedsymbol$

Fig. 2: Isoptic arc and isotomic cycle
\includegraphics[clip,height=40mm]{figures/apollonius03}

Remark 3.14   Using $ isoptic$ cycles, characterized only by $ \tan\vartheta$, together with $ isotomic$ cycles don't provide a system of coordinates since two cycles, one in each pencil, intersect in two points. The required uniqueness is only obtained when using $ isoptic$ arcs (characterized by $ \exp i\vartheta$).

Proposition 3.15   Transformation $ \psi$ defined by $ isoptic\left(\vartheta\right)\,\cap\,isotomic\left(\rho\right)\mapsto isoptic\left(\vartheta+\pi\right)\,\cap\,isotomic\left(\rho\right)$ belongs to $ \mathbb{G}_{circ}^{+}$. Conversely, any involutary element of $ \mathbb{G}_{circ}^{+}$ is either such a transformation or the identity.

Proof. An homography with exactly one fixed point cannot be involutary. Otherwise, both affirmations are equivalent to :

$\displaystyle \left(\omega_{2}-\psi\left(\omega\right)\right)\div\left(\omega_{...
...ight)=-\left(\omega_{2}-\omega\right)\div\left(\omega_{1}-\omega\right)\qedhere$

$ \qedsymbol$

Proposition 3.16   For three cycles $ \pi_{u},\,\pi_{v},\,\pi_{w}$, define $ X_{\perp}=X_{u}\wedge X_{v}\wedge X_{w}$ by the coefficients of $ a_{0},b_{0},c_{0},-d_{0}$ in $ \det\left(X_{0},X_{u,}X_{v}X_{w}\right)$ -generalized cross product- and $ Gram_{u,v,w}$ as in Definition 3.7. The three cycles are from the same pencil if and only if $ X_{\perp}=\left[0,0,0,0\right]$. Otherwise $ \pi_{\perp}$is the unique cycle orthogonal to the given three, and is real according to the sign of $ mink\left(X_{\perp},\,X_{\perp}\right)=-Gram_{u,v,w}$.

Proof. When the three cycles are from the same pencil, everything vanishes. When $ X_{\perp}\neq\left[0,0,0,0\right]$, the four points form a basis of $ \mathbb{R}^{4}$ and $ X_{u},\, X_{v},\, X_{w}$ define a plane whose polar relative to $ \mathbb{S}$ is $ X_{\perp}$. Cycle $ \pi_{\perp}$ is real when its apex is outside $ \mathbb{S}$ while $ mink\left(X_{\perp},\,X_{\perp}\right)=-Gram_{u,v,w}$ generalizes the usual cross product formula... and is easy to verify. $ \qedsymbol$

Remark 3.17   Let $ \omega_{1},\omega_{2}$ be two ordinary points, $ 2\, r=\left\Vert \omega_{1}\,\omega_{2}\right\Vert \neq0$, $ \omega_{0}$ the middle of $ \left[\omega_{1,}\omega_{2}\right]$. Choose $ X_{\infty}=\left[0,0,-1,1\right]$ and $ X_{1},X_{2}$ such that $ c_{i}+d_{i}=2$ (normalized form). Define :

$\displaystyle X_{p},\, X_{m}=\frac{1}{2}\left(X_{1}\pm X_{2}\right)\:;\quad X_{...
...}\wedge X_{s}\:;\quad X_{4}=\frac{1}{4\, r^{2}}\, X_{p}\wedge X_{m}\wedge X_{3}$

Then $ \pi_{3}$ is line $ \left(\omega_{1},\,\omega_{2}\right)$, $ \pi_{m}$ its perpendicular through $ \omega_{0}$, $ \pi_{4}$ is circle $ \left(\omega_{0},\, r\right)$ and $ \pi_{p}$ is circle $ \left(\omega_{0},\, i\, r\right)$. Pencil $ isoptic\left(\omega_{1},\,\omega_{2}\right)$ is $ P\left(\pi_{3},\,\pi_{4}\right)$ while $ isotomic\left(\omega_{1},\,\omega_{2}\right)$ is $ P\left(\pi_{m},\,\pi_{p}\right)$. This decomposition is useful since $ \boxed{\operatorname{G}_{34mp}}=4\, r^{2}\, Mink$.


previous up next contents
Previous: 2 Apexes and pencils Up: Viewing and Touching the Next: 4 Some constructions   Contents


douillet@ensait.fr
2009-03-13