We show that a bijection of the hyperbolic plane that sends horocycles to horocycles (respectively hypercycles to hypercycles) is an isometry. This extends a previous result of J. Jeffers on geodesics to all curves with constant curvature in . We go beyond by showing that every abstract automorphism of the geodesic graph (respectively horocycles and hypercycles graphs) is induced by an earthquake map (respectively an isometry) of . This shadowed the difference between the geometry of geodesics and that of horocycles/hypercycles.
1. Introduction
In this article, we prove two kinds of results. The first one is inspired by works of Jeffers [4] where he showed, among many other similar results, that geodesic-preserving bijections of the hyperbolic plane are isometries. The second one is about the automorphisms of the graph associated to objects on a given geometry: an Ivanov-like theorem.
Jeffers-type results:
In Lost theorem of geometry [4], J. Jeffers showed that geodesic-preserving bijections of (respectively ) are isometries (respectively affine map). This raises the question of finding the geometries that have this kind of results. Let be a group and be a set with a -action; we say that is a -geometry in the sense of Thurston. A set of object in stable under the action of is a set of -geometric object. For instance when and is the group of Möbius transformation, one can choose to be the set of all geodesics.
Question 1.1.
For which pairs , where is a -geometry and is a set of geometric objects on , we have a Jeffers-like result? In other words, every bijection preserving elements of is an element of .
Aside from well-known geometric objects on like geodesics and hyperbolic circles, we also have horocycles (respectively hypercycles) which correspond to orbit flows of parabolic isometries (respectively hyperbolic isometries); see [1].
In this article we first extend Jeffers result to the family of horocycles. Our theorem is:
Theorem A.
Let be a bijection that sends horocycles to horocycles. Then, is an isometry.
We also prove a similar result for hypercycles. We recall that by definition, points on a hypercycle are at the same distance to a given geodesic and this property characterizes hypercycles.
Theorem B.
Let be a bijection that sends hypercycles to hypercycles. Then, is an isometry.
Since geodesics (respectively circles, horocycles and hypercycles) have constant curvature (respectively , and ), our theorem together with J. Jeffers theorems answer 1.1 for all curves with constant curvature in . The main idea that is used to prove these results is to show that postcomposing a given geodesic-preserving bijection (respectively circle-preserving, horocycle-preserving, hypercycle-preserving) leads to the identity map. One of the major obstacles we overcame in our proof of A and B is that horocycles and hypercycles are not completly determined by their endpoints; which was central in the geodesic case.Another contrast between geodesics/circles and horocycles/hypercycles is that some of the results proved in [4] are not true for horocycles. For instance a bijection on the set of all circles in that preserves inclusion of circles, induces a bijection of that sends circles to circles; thus is induced by a Möbius transformation using Theorem 5.1 in [4]. It contrasts with the situation where we consider the set of all horocylces of endowed with the partial order induced by inclusion:
(1)
In particular implies that and have the same center in ; the center of a horocycle referred to the unique point of .Here is a counter-example of a map on the set of all horocycles that preserves inclusion without being an isometry. Let denotes the set of all horocycles centered at . We denote by the horocycle centered at with Euclidean radius equal (when , is a horizontal line and is the imaginary part of all its points). Now, let be the map defined by and for any . One can see that is a bijection that preserves the order while is not induced by an isometry since its restriction to is not continuous. In general, any bijection of induces a such map on .
Ivanov-like results:
These results fit into many others and follow the idea that geometry can be encoded by a group action. On the other hand, Jeffers result on geodesics together with our results have a stronger version which can be stated as a Ivanov-like theorem (see [5] for survey on Ivanov metaconjecture). To do so, let (respectively and ) be the graph of geodesics (respectively horocycles and hypercyles) in where the vertices correspond to geodesics (respectively horocycles and hypercycles) and the edges correspond to disjointness. We aim to describe the automorphism group of these graphs. Let , , and be four different points on considered as . The sets and are linked if and are in different components of . A map preserves links if it sends linked sets to linked sets. Since every is either orientation-preserving or reversing, it turns out that is link-preserving. Since a geodesic is completely defined by its endpoints, every element induces an automorphism of defined by where is the geodesic defined by and in . So, there is a natural injection . Unlike geodesics, horocycles and hypercycles are not determined by their endpoints and this implies different behaviors as shown below:
Theorem C.
We consider the graph (respectively and ) of geodesics (respectively horocycles and hypercycles).
•
The natural injection is an isomorphism. In other words, every automorphisms of is induced by a link-preserving homeomorphism.
•
The natural injection is an isomorphism and the same result also happens for hypercycles.
C suggests that horocycles and hypercycles are more rigid than geodesics in some sense, since their automorphisms group are smaller than the one for geodesics. This comes from the fact that the (respectively ) is richer than since horocycles (respectively hypercycles) can be tangent or can intersect twice in . These facts endow and with much more properties. Every isometry induces a map on the boundary: a boundary map. Therefore, the group of boundary maps induced by isometries is a subgroup of . Moreover, by postcomposing an element by an element in , we get an orientation-preserving homeomorphism of ; and by Thurston earthquake theorem such homeomorphisms are induced by so called earthquake maps on (see [3] for more details on earthquake maps and Thurston’s earthquake theorem).
So, an earthquake map in induced automorphism on an but not on nor on . One can easily guess that an earthquake is more likely to break tangencies between horocycles (respectively hypercycles) than transverse intersections between geodesics (see Figure1 for an example of a simple earthquake and how it breaks tangencies on horocycles). And this what C tells us. The difference between automorphisms induced by earthquake maps and those induced by isometries is that the latter sends geodesics that intersect at a single point to geodesics with the same property. So every automorphism of , which sends geodesics intersecting at a single point to geodesics of the same type is induced by an isometry.
Overview
The proof of A relies on three facts. First we show that a bijection that sends horocycles to horocycles extends to the boundary of . And then, we prove that after postcomposing with an isometry, one can show that fixes pointwise and the set . The proof ends up by showing that if fixes pointwise and a horizontal line then whenever . For B we show that a hypercycle-preserving bijection has to send geodesic to geodesic and Jeffers theorem implies that it is an isometry.
Finally, we will show C. For the case of geodesics, we show that an automorphism of induced a boundary map . In the case of horocycles we use postcomposition techniques to show that a given automorphism comes from an isometry. We end the proof for hypercycles by reduction to the horocycles case namely by showing that a non trivial element induces a non trivial element .
Acknowledgments:
The athours was inspired by Benson Farb talk on reconstruction problems [2]. They would like to thank Dan Margalit for comments on this work. The authors are grateful to Katherine Williams Booth, Ryan Dickmann, Masseye Gaye, Abdou Aziz Diop and Amadou Sy for their comments and the interest shown to this work.
2. Proof of A
The proof of A relies on two big facts. First we show that a horocycle-preserving bijection extends to a bijection on the boundary of (see Proposition2.2). Then, we show that up to postcomposition by isometries, it fixes pointwise and (see Proposition2.5) which helps to end the proof. We recall some basic facts about horocycles that will be useful for what follows.
•
If and , there is a unique horocycle, denoted , centered at and passing through .
•
If and are two distinct points in , there are exactly two horocycles passing through and .
•
Let , and be three distinct points with and . There is at most one horocycle passing through , and .
Now, let be a bijection such that for all , . Two distinct horocycles intersect at most twice and the intersection pattern between horocycles is preserved by . We aim to show that is an isometry.
Extension of to .
Recall for every , denotes the set of all horocycles centered at . The partial order (1) on induces a total order on . The following equivalence gives another way to see that partial order in terms of intersections:
This follows from the Jordan theorem since separates into two components one of which contains . As a consequence, we have:
Lemma 2.1.
Let and be two elements of such that . Then and have the same center and the order is preserved on the images: .
Proof.
Assume that while and have different centers. Then, there exists such that and . Therefore and which is absurd. Using the same technique, we can show that the order is preserved.∎
Proposition 2.2.
Let be a horocycle-preserving bijection. Then extends to a well-defined bijection .
Let and be any horocycle centered at . We set where is the center of . By Lemma2.1, depends only on ; so it is a well-defined map. Moreover, if , then and have the same center and one can assume that . Since also preserves order, it follows that which implies that . Thus, is a bijection.∎
Postcomposition and fixed sets.
The action of on extends to and is -transitive on . In fact, if and are two distinct points in , is an isometry of and , . So, after postcomposing with an isometry, we can assume that and .
Lemma 2.3.
Let be a geodesic-preserving bijection. Then, there exists an isometry such that for all .
Proof.
Since the action of is 2-transitive, there exists an isometry such that and . Therefore, fixes and . Now, let be the horocycle centered at and passing through and be the one centered at and passing through . Since fixes and , and are centered at and respectively, and tangent at a point . It follows that and . Again, by postcomposing with we have that fixes , and and thereby we have and .Let be the sequence of horocycles where is centered at and passes through (see Figure 3).
Since is fixed by and then is either equal to or using the fact that is tangent to and . By postcomposing again with a reflection along we have that is fixed setwise by . Therefore is fixed setwise by for all since the tangency pattern between the horocylces is preserved by . So, for all ; where .∎
Now, we can assume without lost of generality that is a horocycle-preserving bijection such that fixes pointwise . Let be the sequence of subsets of defined by induction as follow:
•
,
•
if then .
Set ; is the set of dyadic numbers in and it is a dense set.
Lemma 2.4.
For all in , and the horocycle is fixed setwise by .
Proof.
First, for all and every horocycle centered at with diameter is fixed setwise.
Now, assume that is fixed pointwise by and every horocycle of diameter and centered at is fixed setwise. For every ,
Since is a fixed set of for all , we have for all . Therefore, the horizontal line passing through all is a fixed set. Let be the horocycle passing through and centered at (see Figure3). That horocycle is tangent to at . Since and , then is fixed setwise and this implies that . So, for and the proof follows by induction.
Let and . Then, and . Since is tangent to , is tangent to . It implies that and is a fixed set of .∎
Proposition 2.5.
Let be a horocycle-preserving bijection. Then, there exists an isometry such that , and fixes pointwise .
Proof.
Lemma2.4 implies that there exist such that for all . Now, assume that . Set and denotes the set of all horocycles centered at a point in and tangent to .By Lemma2.4, for all , is a fixed set of . So, since is pinched between elements of . Hence, .
On the other side, we already know that is fixed setwise by . Let be a point in . We have , and . It implies that and this achieve the proof.∎
Now, we give the proof of A.
Proof of A.
Let be a bijection which sends horocycles to horocycles. By Proposition2.2 extends to and by Proposition2.5 there exists such that fixes and .
Let such that (the case follows the same idea). Then . So , and are fixed point of . Hence, is a fixed set since a horocycle is completely determined by three points. Let such that and the horocylce passing through and : . It follows . If then the order on the horocycles and is not preserved under which is absurd. So, . Therefore, which achieve the proof.∎
3. Proof of B
Hypercylces intersect in four different ways and we first show that they are all preserved by . This implies that hypercycles with same endpoints are mapped to hypercycles with the same property. Finally, we show that a bijection that sends hypercycles to hypercycles also sends geodesics to geodesics which achieves the proof by Jeffers theorem.
A hypercycle is an arc of circle (with endpoints on ) not orthogonal to or a line in that is neither vertical nor horizontal. Unlike geodesics, two hypercycles and may intersect in four different types:
•
Type 1: and intersect at one point and they are tangent;
•
Type 2: and intersect at one point with one endpoint in common;
•
Type 3: and intersect at one point and have different endpoints;
•
Type 4: and intersect at two points.
For the reminder of this section, we set to be a bijection that sends hypercycles to hypercylces. It is not hard to see that preserves intersections of type 4; that is if two hypercycles intersect twice so will be their images under . We first start by showing that the other intersection types are also preserved.
Let and be two hypercycles tangent at a point . The point divided (respectively ) into two sub-arcs and (respectively and ).
Lemma 3.1.
For all in and for all in there exists a hypercycle passing through and and disjoint to . Moreover, this property holds only for pairs of hypercycles with intersection of Type 1.
Proof.
Let and be two points in and respectively. Set to be the Euclidean center of the hypercycle , it lies on the euclidean bisector of the segment . Let be a point in that bisector sufficiently close to such that its (Euclidean) distance to is greater than the distance between and . The circle centered at passing through and defined a hypercycle passing through and and disjoint from , since is a tangent point (see Figure6).
In Type 2 and Type 3 configurations, divided into two components. Moreover, and are in different components. So, every hypercycle from to crosses .∎
The following is a direct consequence of Lemma3.1:
Corollary 3.2.
The hypercycles and are tangent if and only if and are.
We describe a special intersection pattern between three horocycles.
Definition 3.3.
Let , and be three hypercycles tangent at a point . We say that is between of and if turning around on a small circle starting at gives as crossing pattern. Then, we say that is below relatively to .
Lemma 3.4.
If is between of and , then is between and .
Proof.
Assume that is between and . Then separates and and it follows that every hypercycle that intersect both and intersect . Since preserves tangency, , and are tangent at . If is not between and , then there exists a hypercycle that intersects and without intersecting . Therefore, is a hypercycle that intersects and and disjoint from ; which is a contradiction. So, is between and .∎
Lemma 3.5.
Let and be two hypercycles that intersect at a point of Type 2. Then, there exists a hypercycle tangent to at such that every hypercycle below relatively to intersects twice. Moreover, this property does not occur for intersections of Type 3.
Proof.
Let and be two hypercycles that intersect at point of Type 2. Let and be the endpoints of and respectively; with (since and have one endpoint in common). Then, let be a hypercycle tangent to at and intersecting transversally at (see Figure8).
Then, if is below relatively to its endpoints are in the interval that does not contained the endpoints of . It follows that and are not intertwined and intersects . So, they intersect twice.
If and intersect once on Type 3 point, one can see that for every tangent to , there exists below relatively to that intersect at a transverse point; this come essentially from the fact that and have no common endpoints (see Figure9).
∎
Corollary 3.6.
If and intersect at a point of Type 2 (respectively Type 3), so are and .
Now that we know preserves the intersection types, we show the following:
Proposition 3.7.
If and have the same endpoints, then and have also the same endpoints.
Proof.
Assume that and have the same endpoints. So every hypercycles that intersect once at point of Type 3 also intersect .
If and have at most one endpoint in common, there exists that intersect once at a point of Type 3 such that is disjoint from . By taking we obtain a hypercycle that intersect at a point of Type 3 and disjoint from , which is impossible. So, and have the same endpoints.∎
Now, we are able to give the proof of B.
Proof B.
Let us show that a hypercycle-preserving bijection maps geodesics to geodesics. Let be a geodesic in with endpoints and . The set of all points at distance from is a crescent defined by two hypercylces and with endpoints . Then and have the same endpoints by Proposition3.7, and let denotes the geodesic with same endpoints as and . Assume that there is a point such that . There exists a hypercycle passing through and with same endpoints as and . So, is a hypercycle with the same endpoints as and intersecting at , and this is impossible.
So, and it follows that sends geodesics to geodesics. Using Jeffers Theorem ([4]-Theorem 2), we conclude that is an isometry.∎
4. Proof of C
We will prove C for the goedesic case and we will end up with the horocycle and hypercycle cases.
We recall that Thurston earthquake theorem states that every element of is the boundary map of an earthquake of . We will use it in our proof together with the followings:
Lemma 4.1.
Let be an automorphism of ; and let , be two disjoint geodesics with no common endpoints. Then and have no endpoints in common. Equivalently, if , are two disjoint geodesics with a common endpoint, so are and .
Proof.
Let and be two disjoint geodesics with no common endpoints –let’s say and respectively– with the following cyclic order . Set and to be the geodesics defined by and respectively. It follows that the set has the following properties (see Figure10):
•
, and for ;
•
for every geodesic , ;
•
for every geodesic , .
Therefore, the set satisfy the same property. So, for if has at most one endpoint in common with , then there exists a geodesic intersecting but disjoint from ; which is a contradiction. We deduce that for , share one endpoint with and the other one with .
Now, if and has one endpoint in common, there exists a geodesic such that but which also leads to a contradiction. In conclusion, we proved that and have different endpoints.∎
Proof of C: Geodesic case.
We will show that an automorphism of the geodesic graph is induced by an earthquake maps after postcomposing by an isometry.
Let’s show that defined a map on the boundary. Consider ; and be two disjoint geodesics with as a common endpoint. So, and are disjoint. Moreover and have a common endpoint by Lemma4.1; set . The map is well-defined and is a bijection. Let be an increasing sequence with as limit. The sequence of geodesics with endpoints and is such that is between and , which mean every geodesic intersecting and also intersect . Moreover, there is no geodesic with one endpoint equal to such that every is between and . It follows that is a sequence of geodesics with a common endpoint such that is between and . This implies that the sequence is either increasing or decreasing; and let . If , then the geodesic with endpoint is such that is between and for all , which leads to a contradiction. So, and we have continuity for . Hence, is a homeomorphism.Now let’s show that after postcomposing with an isometry, is orientation-preserving. Since the action of isometries on is 3-transitive, after postcomposing with an isometry we can assume that fixes , and . Let , be two points on such (where ”” stands for the counterclockwise order on ) and assume that . Then and are disjoint but not and , which is absurd since is an automorphism. The other cases follow the same idea since there exists a segment which does not contained and . Thus and by Thurston earthquake theorem, is induced by an earthquake map.∎
Let’s introduce some terminology that are going to be useful for the proof in horocycle and hypercycle cases.
Definition 4.2.
A family of hypercycles (respectively horocycles) is continuous if:
•
is between and , i.e, every hypercycles (respectively horocycles) intersecting and intersects whenever ;
•
if is a hypercycle (respectively horocycle) between and , then either intersects uncountably many ’s or there exists such that (respectively there exists such that ). This means that there is no gap in the family .
A continuous family exhausts a continuous family if for every , there exists such that is between and for all .
By definition, elements in a continuous family of horocycles must have the same center. A horocycle can be seen as the limit set of a continuous family of hypercycles. Even more, we have:
Lemma 4.3.
A horocyle is disjoint from a hypercycle if and only if there exists a continuous family of hypercycles converging to and such that .
Proof.
Since isometries acts transitively on horocylces, we can assume that is centered at and passes through . The hypercyles is below and up to a translation, we can assume that intersects the -axis orthogonally at with . Let and be the endpoints of . We defined to be the hypercycle with endpoints and passing through for . The set is a continuous family of hypercycles that accumulate to the horizontal horocycle passing through .∎
Since continuous families of horocycles/hypercycles are defined using intersection properties, it follows that they are preserved under automorphisms, i.e, an automorphism sends a continuous family of horocycles/hypercycles to a continuous family. As consequence, we have:
Lemma 4.4.
Let be an automorphism of (respectively ):
(1)
If and are two tangent horocycles, then is tangent to .
(2)
If and are two hypercycles with the same endpoints, so are and .
Proof.
Let be an automorphim and assume that and are two tangent horocycles. So, there exists a continuous family such that and for all . The set is a continuous family where intersects and is disjoint from for all . Hence, is tangent to .
Two hypercycles and have the same endpoints if and only if there exists a hypercycle between and which is characterized by an intersection property. Since the property of having the same endpoints is characterized by intersection, it is preserved by automorphisms.∎
Lemma 4.5.
Let be a continuous family of hypercycles converging to a horocycle. Then, we have one of the following situations:
•
the union of all is equal to one of the components of ;
•
converges to a horocycle or a hypercycle/geodesic.
Moreover, an automorphism of preserves the convergence type of a family .
Proof.
Assume that does not foliate any component of . Let and be the set of endpoints of the family , i.e, are the endpoints of where . By defintion of a continuous family, the set is bounded from above by any element and is bounded from below by any ; set and . There are two cases depending on whether or not. If , there exists a horocycle centered at and disjoint from for all , since foliates any component of . We denote by the set of all horocycles centered at and disjoint from for all and let be the one with maximal radius. Since has maximal radius, it follows that converges to . The case follows the same idea by taking the set of all hypercycles disjoint from the ’s with endpoints and .
Now, assume that is a continuous family of hypercycles and that foliates one of the components of . Then, for every hypercycle in the foliated component with the same endpoints as there exist such that is between and . On the other hand, if is a continuous family admitting a hypercycle as a limit, is a hypercycle in the component containing the ’s and is disjoint from for all . Finally, if converges to a horocycle there exists a hypercycle with the same endpoints as and intersecting all ( take among the ones that intersects the horocycle ) and there is no hypercycle disjoint from all in the connected component of containing all . Therefore, the three types of convergence are characterized by different types of intersection patterns, so they are preserved by automorphisms.∎
Proof of C: Horocycles/Hypercycles cases.
Let be in ; the horocycle centered at and passing through and the horizontal horocycle passing through . Therefore, and are tangent, thanks to Lemma4.4. Since is 3-transitive on , we can assume that and have the same center like and , after postcomposing with an isometry. So, and are tangent at and again by postcomposing with we have and . From here, we can use the same argument like in the proof of Lemma2.4 to show that every horocycle centered at and passing through is fixed by . Now if is a given horocycle, there exists two horocycles and tangent to and such that and are tangent to (see Figure13).
Since fixes and , it follows that fixes . Thus, after postcomposing with an isometry acts trivially on and this implies that is induced by an isometry.
Now, let be an automorphism of . Our goal is to show that induced an automorphism of . Let be an horocycle and be a continuous family of hypercycles with limit . By Lemma4.5 is a continuous family which converges to a horocycle ; we defined . We claim that is an automorphism. Consider and two disjoint horocycles with and two continuous family of hypercycles converging to and respectively. Then, there exists such for , . And this implies that is disjoint from , and thus is an automorphism of . Let defined by ; is injective. Assume that while . Then, there exists a hypercycle such that ; and a horocycle disjoint form such that . Since is disjoint from , by Lemma4.3 there exists a continuous family containing such that is its limit set. It follows that is a continuous family converging to ; which is absurd since contains which is not disjoint form . Hence, is injective and this implies that every element of is induced by an isometry, as can be seen as a subgroup of .∎
References
[1]Alan Beardon, Geometry of discrete groups, Graduate Texts in Mathematics, vol. 91, Springer-Verlag, 1995.
[3]J. Hun, Earthquake on the hyperbolic plane in: Handbook of Teichmuller theory. Vol III: European Mathematical Society. IRMA Lectures in Mathematics and Theoretical Physics 17, (2012)
[4]J. Jeffers, Lost theorems of geometry, Amer. Math. Monthly, 108 (2000), 800-812.
[5] A. Papadopoulos, Rigid actions of the mapping class group, European Mathematical and Theoretical House, 2014
cheikh21.lo@ucad.edu.sn, University Cheikh Anta Diop, Dakar, Senegal abdoulkarim3.sane@ucad.edu.sn, University Cheikh Anta Diop, Dakar, Senegal
Address: Suite 609 315 Lupita Unions, Ronnieburgh, MI 62697
Phone: +2424755286529
Job: District Education Designer
Hobby: Yoga, Gunsmithing, Singing, 3D printing, Nordic skating, Soapmaking, Juggling
Introduction: My name is Moshe Kshlerin, I am a gleaming, attractive, outstanding, pleasant, delightful, outstanding, famous person who loves writing and wants to share my knowledge and understanding with you.
We notice you're using an ad blocker
Without advertising income, we can't keep making this site awesome for you.