Geometric transformation

Geometry Level 4

We define for all position vectors P R 2 { 0 } \vec P \in \mathbb {R} ^ 2 \setminus \{\vec 0 \} in the two-dimensional plane (except the zero vector) a coordinate transformation T : P P T: \vec P \mapsto \vec P ' with P = T ( P ) = 1 P 2 P \vec P' = T(\vec P) = \frac{1}{|\vec P|^2} \cdot \vec P Thus, the vectors P \vec P and P \vec P ' point in the same direction, but their lengths are inverse to each other, so that P = P 1 | \vec P' | = | \vec P | ^ {-1} . This transformation is similar to a reflection, since inner part and outer part of the unit circle around O O are interchanged. This mapping is bijective, so that each point P P in the plane is uniquely assigned a different point P P' .

We now consider an arbitrary circle C C and arbitrary straight line g g in the two-dimensional plane. We assume that both curves do not go through the origin O O .

What shape do the curves C C and g g take when subjected to the coordinate transformation T T ?

Circles remain circles and lines remain lines Both C C and g g are mapped to circles. g g is mapped to a circle and C C transfroms to a straight line Both C C and g g are mapped to lines.

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

1 solution

Theorem: T ( C ) T(C) and T ( g ) T(g) are both circles.

It is easy to exclude that neither C C nor g g is transformed into a straight line. For each curve γ \gamma we can define a distance to the origin by d γ = min P γ P d_\gamma = \text{min}_{\vec P \in \gamma}{| \vec P | } . Since the coordinate transformation inverts the lengths, the point closest to the origin is mapped to a point furthest from the origin. Thus, the image T ( γ ) T(\gamma) of curve γ \gamma must be within a circle of radius d γ 1 d_\gamma^{-1} , because P = P 1 d γ 1 |\vec P'| = |\vec P|^{-1} \leq d_\gamma^{-1} for each P γ \vec P \in \gamma . Since neither C C nor g g goes through the origin, their distances are strictly positive ( d C , d g > 0 d_C, d_g > 0 ), so that their images T ( C ) T(C) and T ( g ) T(g) are bounded. But a straight line continues to infinity, so that there is a contradiction here.

The proof that the curves are actually circles is much more complicated. In the proof here, which I have worked for myself, I use complex numbers to algebraically represent the transformation. But there may be other proofs that are much shorter or more elegant.

Proof:

We treat the vectors P = ( x , y ) R 2 \vec P = (x,y) \in \mathbb{R}^2 as complex numbers z = x + i y = r e i φ C z = x + iy = r e^{i \varphi} \in \mathbb{C} with real part x = r cos φ x = r \cos \varphi and imaginary part y = r sin φ y = r \sin \varphi . Thus, the transformation T T corresponds to a mapping f : z 1 z = 1 r e i φ f: z \mapsto \frac{1}{z^\ast} = \frac{1}{r} e^{i \varphi} Therefore, f ( z ) f(z) and z z have the same phase φ \varphi and point in the same direction. Furthermore, f ( z ) = r 1 = z 1 |f(z)| = r^{-1} = |z|^{-1} , so that the absolute values are inverse to each other.

A circle around point a a with radius R R corresponds to the points z = a + a a R e i φ = a ( 1 + ρ e i φ ) , ρ = R a , φ [ 0 , 2 π ] z = a + \frac{a}{|a|} R e^{i \varphi} = a \left(1 + \rho e^{i \varphi} \right), \quad \rho = \frac{R}{|a|}, \varphi \in [0,2\pi] We assume ρ 1 \rho \not= 1 (the origin is not intersected), so that f ( z ) = 1 a 1 1 + ρ e i φ = 1 a ( 1 1 ρ 2 1 1 ρ 2 + 1 1 + ρ e i φ ) = 1 a ( 1 1 ρ 2 ρ ( ρ + e i φ ) ( 1 ρ 2 ) ( 1 + ρ e i φ ) ) = 1 ( 1 ρ 2 ) a ( 1 ρ ρ + e i φ 1 + ρ e i φ ) = a + a a R e i φ where a = 1 ( 1 ρ 2 ) a = a a 2 R 2 , R = ρ ( 1 ρ 2 ) a = R a 2 R 2 , e i φ = ρ + e i φ 1 + ρ e i φ e i φ 2 = ρ + e i φ 1 + ρ e i φ ρ + e i φ 1 + ρ e i φ = ρ 2 + 1 + 2 ρ cos φ 1 + ρ 2 + 2 ρ cos φ = 1 \begin{aligned} f(z) &= \frac{1}{a^\ast} \frac{1}{1 + \rho e^{-i \varphi}} \\ &= \frac{1}{a^\ast} \left( \frac{1}{1 - \rho^2} - \frac{1}{1 - \rho^2} + \frac{1}{1 + \rho e^{-i \varphi}} \right)\\ &= \frac{1}{a^\ast} \left( \frac{1}{1 - \rho^2} - \frac{\rho (\rho + e^{-i \varphi})}{(1 - \rho^2)(1 + \rho e^{-i \varphi})} \right)\\ &= \frac{1}{(1 - \rho^2) a^\ast} \left( 1 - \rho \frac{\rho + e^{- i \varphi}}{1 + \rho e^{- i \varphi}} \right)\\ &= a' + \frac{a'}{|a'|} R' e^{i \varphi'} \\ \text{where} \quad a' &= \frac{1}{(1 - \rho^2) a^\ast} = \frac{a}{|a|^2 - R^2}, \\ R' &= \frac{\rho }{|(1 - \rho^2) a^\ast|} = \frac{R}{||a|^2 - R^2|}, \\ e^{i \varphi'} &= - \frac{\rho + e^{- i \varphi}}{1 + \rho e^{- i \varphi}} \\ |e^{i \varphi'}|^2 &= \frac{\rho + e^{- i \varphi}}{1 + \rho e^{- i \varphi}} \frac{\rho + e^{i \varphi}}{1 + \rho e^{ i \varphi}} = \frac{\rho^2 + 1 + 2 \rho \cos \varphi }{1 + \rho^2 + 2 \rho \cos \varphi} = 1 \end{aligned} Therefore φ R \varphi' \in \mathbb{R} and all points of the image are one a circle with center a a' and radius R R' . Since the map f f is bijective and continuous, the circle must also be completely traversed.

A straight line that does not go through the origin can be represented in the complex plane by the points z = u + i u λ , λ R z = u + i u \lambda, \quad \lambda \in \mathbb{R} where u u is the distance vector from the origin to the line and i u i u is a direction vector that is parallel to the line. The image of this line then gives f ( z ) = 1 u i u λ = 1 u i u λ u + i u λ u + i u λ = u + i u λ u 2 ( 1 + λ 2 ) = u 2 u 2 ( 2 1 + λ 2 + i 2 λ 1 + λ 2 ) = u 2 u 2 ( 1 1 + 2 1 + λ 2 + i 2 λ 1 + λ 2 ) = u 2 u 2 ( 1 + 1 λ 2 1 + λ 2 v + i 2 λ 1 + λ 2 w ) = u ( 1 + e i φ ) where u = u 2 u 2 , φ = arctan w v v 2 + w 2 = ( λ 4 2 λ 2 + 1 ) + 4 λ 2 λ 4 + 2 λ 2 + 1 = 1 \begin{aligned} f(z) &= \frac{1}{u^\ast - i u^\ast \lambda} \\ &= \frac{1}{u^\ast - i u^\ast \lambda} \frac{u + i u \lambda}{u + i u \lambda} \\ &= \frac{u + i u \lambda}{|u|^2 ( 1 + \lambda^2)} \\ &= \frac{u}{2|u|^2} \left(\frac{2}{1 + \lambda^2} + i \frac{2\lambda}{1 + \lambda^2} \right) \\ &= \frac{u}{2|u|^2} \left(1 - 1 + \frac{2}{1 + \lambda^2} + i \frac{2\lambda}{1 + \lambda^2} \right) \\ &= \frac{u}{2|u|^2} \Bigg(1 + \underbrace{\frac{1 - \lambda^2}{1 + \lambda^2}}_{v} + i \underbrace{\frac{2\lambda}{1 + \lambda^2}}_{w} \Bigg) \\ &= u' \left(1 + e^{i \varphi} \right)\\ \text{where} \quad u' &= \frac{u}{2|u|^2}, \\ \varphi &= \arctan \frac{w}{v} \\ v^2 + w^2 &= \frac{(\lambda^4 - 2 \lambda^2 + 1) + 4 \lambda^2 }{\lambda^4 + 2 \lambda^2 + 1} = 1 \end{aligned} Thus, all points lie on a circle around the point u u ' with the radius u | u' | . This circle goes in particular through the origin O O . However, the origin itself is reached only in the limit case λ ± \lambda \to \pm \infty , since lim λ ± v = 1 \lim_{\lambda \to \pm \infty} v = -1 und lim λ ± u = ± 0 \lim_{\lambda \to \pm \infty} u = \pm 0 . QED

More information about this so-called circle inversion can be found on Wikipedia.

This is actually the principle behind the Peaucellier-Lipkin Linkage which converts circular movements to straight-line motions. It inverts a circle that's touching O O into a line

Julian Poon - 3 years ago

Log in to reply

I have never actually heard of the Peaucellier–Lipkin linkage. However, it is very interesting that a mathematical mapping of lines on circles actually has a direct technical implementation for the construction of steam engines. Thanks to you, I've learned something new.

Markus Michelmann - 3 years ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...