Straightening Billiard Trajectories in Flat Disks and Katok-Zemlyakov Construction
\.Ismail Sa\u{g}lam

TL;DR
This paper introduces a method to straighten billiard trajectories in flat disks, analyzes their closures, and generalizes the Katok-Zemlyakov construction to rational flat disks, providing new insights into their geometric structure.
Contribution
It presents a novel method for straightening billiard trajectories and extends the Katok-Zemlyakov construction to rational flat disks, including calculating associated Euler characteristics.
Findings
Trajectory closures contain a vertex for almost all directions.
A translation surface is constructed for each rational flat disk.
Euler characteristics of these surfaces are explicitly calculated.
Abstract
We provide a method to straighten each billiard trajectory in a flat disk. As an application, we show that for each point on the disk and for almost all directions, the closure of the corresponding billiard trajectory contains a vertex. We generalize Katok-Zemlyakov construction to the flat disks with rational angle data: for each rational flat disk we obtain a translation surface and a projection to the doubling of the disk. We calculate the Euler characteristics of this translation surface.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsMathematical Dynamics and Fractals · Geometric and Algebraic Topology · Geometric Analysis and Curvature Flows
Straightening Billiard Trajectories in Flat Disks and Katok-Zemlyakov Construction
İsmail Sağlam Electronic address: [email protected] Adana Alparslan Turkes Science and Technology University
Abstract
We provide a method to straighten each billiard trajectory in a flat disk. As an application, we show that for each point on the disk and for almost all directions, the closure of the corresponding billiard trajectory contains a vertex. We generalize Katok-Zemlyakov construction to the flat disks with rational angle data: for each rational flat disk we obtain a translation surface and a projection to the doubling of the disk. We calculate Euler characteristics of this translation surface.
1 Introduction
Billiard flows in polygons is a subject which is being studied extensively. Ergodicity and minimality of the flows, number of closed geodesics and generalized diagonals are among the most important topics in this area. See the surveys [4],[3],[8].
Polygons are special examples of the flat surfaces. A flat surface is a surface together with a metric obtained by gluing Euclidean triangles along their edges by isometries. See [9],[12],[13],[14],[5] for more information about flat surfaces. For each point in the surface, there exists a well defined notion of angle, . If an interior point has angle , then it is called non-singular. Otherwise, it is called singular. If a boundary point has angle , then it is called non-singular. Otherwise, it is called singular. Let be a compact flat surface. Let and be the boundary and the interior of , respectively. The following formula (Gauss-Bonnet) holds:
[TABLE]
where is the Euler characteristics of .
It is hard to say that the study of the billiard flow on any compact flat surface is possible. Nevertheless, there is a family of flat surfaces that has been studied extensively up to now. This family consists of the closed flat surfaces with trivial holonomy groups. Such a surface is called a translation surface. See [16], [7] for more information about the translation surfaces. The aim of this manuscript is to point out that there is another family of the flat surfaces for which it is not too much to hope that the billiard flow can be studied in a systematic manner. This family consists of flat disks.
A flat disk is a closed, oriented topological disk together with a flat metric on it. Polygons are simplest examples of the flat disks. Each flat disk may have finitely many singular interior and singular boundary points. Gauss-Bonnet formula becomes
[TABLE]
Now we construct a flat disk. See Figure 1. Note that the length of the edge is equal to the length of the edge , and the length of the edge is equal to the length of the edge . If we glue and , and , then we get a flat disk with two singular interior points and four singular boundary points. Indeed, any flat metric on a disk may be constructed from a flat disk in a similar manner. See Section 2.2.
There is a simple but quite useful trick used in the study of the polygonal billiards to unfold a billiard trajectory in a polygon. See Section 2.1. The idea of the trick is to reflect the billiard table instead of reflecting the billiard ball. We give a generalization of this trick for an arbitrary flat disk. See Section 2.2. As a result, we generalize a theorem obtained by Boldrighini, Keane and Marchetti [1]. See also [11]
In [1], it was proved that for any point in a polygon and for almost all directions, the closure of the billiard trajectory passing through this point having the same direction contains a vertex. See [2] for an improvement of this result. We show that for all points on a flat disk and almost all directions, the closure of the billiard trajectory contains a singular boundary point or a singular interior point. See Section 2.3.
A polygon is rational if its angles are rational multiples of . From a rational polygon one can obtain a translation surface. This is known as Katok-Zemlyakov construction. See [15]. This surface is called invariant surface. By that way one can argue that a rational polygon contains closed billiard trajectories, and give upper and lower bounds for number of closed trajectories and generalized diagonals. See [6]. In this manuscript we do a parallel construction for rational flat disks.
A flat disk is called rational if the angles at its singular points are rational multiples of . This is equivalent that the holonomy group of the doubling of the flat disk is finite, where the the doubling of a flat surface with boundary is the surface obtained by gluing two copies of the surface along their boundaries. For such a flat disk, we can construct a translation surface and a projection from the translation surface to the disk or its doubling. We show that the translation surface that we obtain is unique; it does not depend on how we cut the disk to obtain this translation surface. See Sections 3.2, 3.3. Finally, we calculate the Euler characteristic of this translation surface. See Section 3.4.
2 Straightening Billiard Trajectories and a Generalization of the Theorem of Boldrighini, Keane and Marchetti
In this section we recall how to unfold billiard trajectories in the polygons. Then we extend the construction for the flat disks. Also, we extend the theorem of Boldrighini, Keane and Marchetti to the flat disks [1].
2.1 Unfolding Trajectories in Polygons
Let be a polygon in . Label its edges as . Let be a billiard trajectory which never hits the vertices . Assume that first hits the edge , then the edge , and in general let be the edge which hits at its -th intersection with the boundary . Therefore we have an infinite word
[TABLE]
Let and be the polygon obtained by reflecting along . In general, let , , be the copy of obtained by reflecting along its edge . See the Figure 2.
Let and , where we identify the edges labeled by in and . Then we can unfold at as a straight line segment. By that way, we can obtain a line segment in which hits the edge . Call this segment .
In general, we can form the polygon , where and are glued through the edges of and labeled by . Similarly, I can unfold at as a straight line segment intersecting . We call this line segment . This means that I can obtain an unbounded complete flat surface
with a non-singular interior and a straight half-line . Also, there is a projection sending to . It is clear that the restriction of this projection to each is an isometry between and .
2.2 Unfolding of Trajectories on Flat Disks
Before we proceed, let us make it clear what we mean by unfolding. In the case of polygons, we obtained a flat surface with non-singular interior and a line on it such that there is a projection sending to the billiard trajectory . For the case of the flat disks, we need to seek for such a surface and a projection map.
Given a flat disk , one my try to construct as follows. Let be the edges of . Let be a billiard trajectory which never hits the singular points. Let
[TABLE]
where is the edge on which the trajectory intersects the boundary at the -th time. Let be the copies of . Let and , where we identify the edges labeled in and . In general, let
[TABLE]
where we glue the edgs labeled in and . Similarly
[TABLE]
Indeed, is not good for our purposes. It has infinitely many singular points in its interior if interior of is singular.
Another important thing is that there may be trajectories in which never hit the boundary and the singular points. See the Figure 3. Take two copies of a square and glue the edges with , with , with to get a flat disk with two singular points on its interior. The geodesic shown in the figure is periodic and never hits the boundary. Note that this situation does not arise in the case where the angle at each interior point is greater than or equal to . In that case, a trajectory either hits the boundary or the singular points. This can be proved by a simple argument involving Gauss-Bonnet Formula.
Now we explain how to unfold the billiard trajectories in the flat disks. Let be a flat disk with singular interior points and singular boundary points . Let be the boundary segments or the edges of .
Let be not self-intersecting broken geodesics such that
joins and a singular boundary point , 2. 2.
, 3. 3.
.
Let be a billiard trajectory which never hits the singular points. will intersect the set
[TABLE]
infinitely many times. Also, since is oriented it makes sense to say hits from left or right. Let
[TABLE]
be the infinite word consisting of the letters
[TABLE]
such that
- •
if the trajectory hits from the left at the -th time,
- •
if the trajectory hits from the right of at the -th time,
- •
if the trajectory hits at at the -th time.
Cut through ’s to get a flat disk with a non-singular interior and boundary components. Clearly, we can label boundary components of with the symbols
[TABLE]
so that this labeling coincides with the labeling of .
Take countably many copies of . Let and where we identify the edges
labeled in and if , 2. 2.
labeled in and in if , 3. 3.
labeled in and in if .
Observe that we can unfold to a line segment hitting the edge in . We can define
[TABLE]
as follows. We identify the edges
labeled in and if , 2. 2.
labeled in and in if , 3. 3.
labeled in and in if .
We can unfold to a line segment in which hits the edge in . Call this segment . Clearly, there is a projection which sends to and is an isometry when restricted to the the interior of each , . Let
[TABLE]
is a flat surface which is unbounded, complete and has non-singular interior. Clearly, there is a projection
[TABLE]
and a straight half-line such that is . Clearly, induces an isometry between the interior of and .
Example:
Take an equilateral triangle of the edge length 1 and isosceles right triangle having edge lengths . Label their edges as in the Figure 4. Glue the edges having the same labels to get a flat disk with one singular interior point and two singular boundary points. Call this disk . Let is the polygon obtained by cutting along . It is shown in Figure 5. Let be a billiard trajectory in . We can unfold it as follows. Whenever the trajectory of hits (from left or right) We put a copy of in the plane which is obtained by rotating about by an angle of 150∘ . Whenever the trajectory of hits (or ), we put a copy of in the plane which is obtained by reflecting along (or ). See the Figure 5.
Remark 1**.**
It may be possible that the boundary of is non-singular. See Figure 3. In that case one can choose an arbitrary point on the boundary and join singular interior points and this point by some broken geodesics to obtain , and construct in a similar manner. Indeed, in general, the boundary points that we choose to specify ’s do not need to be singular. We choose singular boundary points since this makes the notation considerably simple.
2.3 The theorem of Boldrighini, Keane and Marchetti
In this section we prove that for all points on a flat disk and for almost all directions, the billiard trajectory passing through this point having the same direction contains a vertex in its closure. Note that by almost all directions we mean that a set of full measure with respect to Lebesque measure on the unit tangent space of a point in the flat disk.
Let be a flat disk. Let be a non-singular interior point.
[TABLE]
where is the tangent space at . We identify with the unit circle on and endow it with the Lebesque measure on . Let be the Lebesque measure on .
Theorem 1**.**
Let be arbitrary point. For almost all , the billiard trajectory initiating at and having direction contains a singular vertex in its closure.
Proof.
Let and . Let bet the trajectory passing through and having the direction . Let be the set of singular points of . Let
[TABLE]
We will show that . It is clear that this is enough to prove the theorem.
Assume that there exists such that . Therefore there exists such that
[TABLE]
In other words, is a density point of . Let and . Let and the trajectories having initial point and directions and .
Assume that has singular interior points . Let be the edges of . Let be broken not self-intersecting geodesics joining these singular points and boundary. They satisfy also the following conditions:
joins and a singular boundary point , 2. 2.
, 3. 3.
.
Let and be the word sequences that we obtain from and as in Section 2.2.
[TABLE]
where . Let be the disk obtained from by cutting it through . Assume that . Then there exists a flat surface obtained from the copies of and two half-lines on it making an angle of with each other. Call these lines and . This implies that contains a V-shaped region. See the Figure 6. But this is impossible since the diameter of is finite. Therefore .
Let be the smallest integer such that . Consider the polygon . This polygon has at least one vertex inside the region bounded by and . Let be a circle of radius with center .
From the Figure 7 it follows that each trajectory having direction is not in . But it is evident that there exists a constant such that
[TABLE]
Therefore
[TABLE]
which contradicts with the assumption that is a point of density.
If is a non-singular boundary point, then we can argue similarly. Of course, the distance between and any trajectory originating from a singular point is zero.
∎
3 Katok-Zemlyakov Construction for Flat Disks
In this section we first recall Katok-Zemlyakov construction [15] for the polygons and then make a similar construction for the flat disks. We show that the translation surface that we construct is unique; it does not depend on the way that we cut the flat disk. Also, we calculate the genus of this translation surface.
3.1 Polygons
Here we summarize how we obtain a translation surface out of a rational polygon. Let be a polygon. Let be the vertices and be the edges of . Assume that the angle at , , is equal to , where . Let . Let denote the linear part of the group generated by reflections across the edges . Then has elements. Indeed, is isometric to the dihedral group of order .
For each , let be a polygon such that
is obtained by applying a translation of to , 2. 2.
if .
Label the edges of by the symbols . The reflection across gives an element of . Let denote this element. Consider the polygon
[TABLE]
Glue the edge of with the edge of for each . By that way we get a flat surface without boundary. and are glued by translations, therefore the surface that we get is a translation surface.
Example:
We can obtain a flat torus through from a square in . We can glue four copies of the square to get the torus. See Figure 8.
Call the surface that we obtain by the above construction . Let be the doubling of : the surface that we obtain by gluing two copies of along their boundaries. Then naturally gives a branched covering of which is local isometry outside of the ramification points.
3.2 Flat Disks
Now assume that we have a flat disk . Let be the singular interior points of . Assume that for each , where . Let be the boundary components of and be the singular boundary points. Assume that for each , where Let be the broken not self-intersecting geodesics such that
joins and a singular boundary point , 2. 2.
, 3. 3.
.
Cut through ’s to get a flat disk which has edges and a non-singular interior. For each , there exist two boundary components of which are labeled by . Since is oriented we can label one of them by and the other by , where and refer to left and right, respectively.
Since the interior of is non-singular there is a developing map
[TABLE]
which is a local isometry outside of the singular points of . Without loss of generality, we assume that this map is injective. That is, it is an isometry between and its image. Thus we assume that is a polygon in the plane.
Let be the linear part of the group of isometries of generated by the rotations of an angle about , , and the reflections across the edges of . Let . Then has order . Indeed, is isometric to the dihedral group of order .
For each let be a polygon obtained by a translation of such that whenever . Label the edges of by the symbols
[TABLE]
For each , let be the linear part of the reflection across to . Glue and along the edges and . It is clear that we can do this by a translation.
Let be the counterclockwise rotation of angle about the origin. Glue the polygons and through the edges and . Note that this gluing can be given by a translation of . By that way we obtain a translation surface . Clearly, there is a branched covering which is a local isometry except at the ramification points. We call invariant surface of .
Example:
Assume that is the flat disk obtained by gluing two edges of an equilateral triangle. Thus is a flat disk with one singular interior point and one singular boundary point with angles and . See the Figure 9. Then is obtained by gluing each edge of the hexagon on the middle by the opposite edge of the hexagon on the right. has genus two.
Example:
Assume that is the flat disk obtained by gluing two edges of an isosceles right triangle. Thus is a flat disk with one singular interior point and one singular boundary point with angles and . See the Figure 10. Then is obtained by gluing each edge of the square on the middle by the opposite edge of the square on the right. has genus one.
Remark 2**.**
If the boundary of is non-singular we can take any point in and some broken geodesics joining the singular points and . Repeating the construction above, we get a translation surface and a cover .
3.3 Uniqueness of Invariant Surface
In this section we prove that the surface which was found in the previous section does not depend on the chosen paths . First we introduce a notion of covering for the flat surfaces. Then we state some results found in [10]. From a topological point of view such a covering is nothing else than a branched covering. But, we also require that it respects the flat metrics.
Given a branched covering between two surfaces, we denote the set of the branched points by and the set of the ramification points by .
Let and be two flat surfaces.
Definition**.**
A map is called a flat (covering) map if it is a branched covering and
[TABLE]
is a local isometry.
Examples:
If is a finite group of isometries of then is a flat covering. 2. 2.
The map obtained in Section 3.2 is flat. 3. 3.
If is a branched covering then the induced map on is flat, and this makes a flat covering.
Definition**.**
A closed, orientable flat surface is called really flat if it has finite holonomy group. 2. 2.
An orientable compact flat surface with boundary is called really flat if its doubling is really flat. 3. 3.
A non- orientable compact surface is really flat if its orientable double cover is really flat.
Thus the really flat surfaces are somewhere in between the translation surfaces and arbitrary flat surfaces. Note that a flat disk is really flat if and only if the angles at singular points are rational multiples of .
Theorem 2**.**
Let be a compact, oriented, closed really flat surface.
There exists a translation surface and a flat covering . Also, this map corresponds to the kernel of the holonomy represntation of . 2. 2.
* is a cyclic Galois covering.* 3. 3.
If has angle , , then the ramification index at each point in the fiber of is . 4. 4.
The degree of is the order of the holonomy group.
We call canonical cover. Now let be a rational flat sphere. As before, let be the singular interior points and be the singular boundary points. Assume that the angle at is and the angle at is , where and are coprime integers, and so are and . Let be the canonical covering of .
Theorem 3**.**
Let be the leaast common multiple of . Then
the degree of is , 2. 2.
the ramification index at each point in the fiber of or its mirror image is , 3. 3.
the ramification index of each point in the fiber of is ,
Now we prove that the surfaces and are same.
Theorem 4**.**
There is an isometry making the following diagram commutative:
[TABLE]
Proof.
has trivial holonomy group, so it corresponds to a subgroup of the kernel of the holonomy representation of . But, the degree of is , which is the degree of the holomomy group. Thus corresponds to the kernel of the holonomy group. But, also corresponds to the kernel of the holonomy group. Therefore and are same as flat coverings. ∎
3.4 Calculation of Euler Characteristics
We first give Riemann-Hurwitz formula. Let and be two compact orientable surfaces and be a branched covering. The following formula is called Riemann-Hurwitz formula.
[TABLE]
where is the degree of and is the ramification divisor:
[TABLE]
Here is the ramification index of at .
As before, let be a flat disk with singular interior points and singular boundary points . Let and . Consider the cover .
It has degree , where is the least common multiple of . 2. 2.
The ramification index at a point or its mirror image is . 3. 3.
The ramification index at a point over is . 4. 4.
The set consists of points. 5. 5.
If is the mirror image of , then the set consists of points. 6. 6.
The set consists of points.
It follows that
[TABLE]
Therefore
[TABLE]
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[1] C. Boldrighini, M. Keane, F. Marchetti, (1978). Billiards in polygons. The Annals of Probability, 6(4), 532-540.
- 2[2] G. Galperin, T. Krüger, and S. Troubetzkoy. ”Local instability of orbits in polygonal and polyhedral billiards.” Communications in mathematical physics 169.3 (1995): 463-473.
- 3[3] E. Gutkin, Billiards in polygons: survey of recent results, in ”Journal of statistical physics” 83.1 (1996): 7-26.
- 4[4] E. Gutkin, Billiard dynamics: a survey with the emphasis on open problems, in ” Regular and chaotic dynamics” 8.1 (2003): 1-13.
- 5[5] D. Hulin, and M. Troyanov, Prescribing curvature on open surfaces, in ”Mathematische Annalen” 293.1 (1992): 277-315.
- 6[6] H. Masur, (1988). Lower bounds for the number of saddle connections and closed trajectories of a quadratic differential. In Holomorphic Functions and Moduli I (pp. 215-228). Springer, New York, NY.
- 7[7] H. Masur, Ergodic theory of translation surfaces, in ”Handbook of dynamical systems 1” (2006): 527-547.
- 8[8] H. Masur and S. Tabachnikov, (2002). Rational billiards and flat structures, in ” Handbook of dynamical systems 1” (Vol. 1, pp. 1015-1089). Elsevier Science.
