WorksMathematics
Complete Topological Invariant and Coding of Homoclinic Flows on the Two-Dimensional Sphere
Abstract
We study codimension one homoclinic flows on closed surfaces and introduce a complete topological invariant based on the structural graph of the flow and the distinguishing graphs of its atoms. This invariant completely determines the topological equivalence class of a homoclinic flow. For flows on the two-dimensional sphere, we construct a canonical coding algorithm that avoids the global renumbering required in the classical coding of Morse flows and substantially reduces the computational complexity. As an application, we derive recursive formulas for the number of topologically distinct homoclinic flows on the two-dimensional sphere and compute these numbers for flows with a small number of saddle singularities and closed orbits.
Introduction
Morse–Smale vector fields form the class of structurally stable dynamical systems and play a crucial role in the qualitative theory of dynamical systems. In 1970, Palis and Smale [7] proved that Morse–Smale vector fields form an open and dense subset in the space of vector fields on closed surfaces.
The simplest structurally unstable vector fields are the codimension one vector fields, which appear in generic one-parameter families of vector fields. They form the codimension one strata in the space of vector fields and belong to the boundary of the Morse–Smale vector fields set.
In this paper, we study codimension one flows on closed surfaces possessing exactly one homoclinic saddle loop. We are interested in their classification up to topological trajectory equivalence, which will simply be referred to as topological equivalence. Since merely the orbit structure is relevant, and we are considering smooth vector fields, we shall freely identify vector fields with the flows generated by them and use the terms vector field and flow interchangeably throughout the paper.
Complete topological invariants are important in the topological classification of Morse–Smale flows and their codimension one bifurcations. Such invariants are usually described by graphs equipped with additional combinatorial data. Classical examples include the separatrix diagram of Peixoto [8], embedded graphs encoded by rotation systems [1], and chord diagrams for Morse flows with one source [5]. This problem was also thoroughly studied in [6]. Flows with collective dynamics were researched from a similar point of view in [12]. Related problems concerning the topology of functions on surfaces and their generic deformations were studied in [2], where deformations in the general position of optimal functions on oriented surfaces with boundary were investigated.
In our recent paper [9], we constructed a complete topological invariant for gradient codimension one flows on closed surfaces, and in [11] for compact surfaces with boundary. Those flows contain neither periodic orbits nor homoclinic loops. The invariant extends the chord diagram of Morse flows by introducing a distinguished point, which provides a canonical numbering of the chords and, consequently, a canonical code of the flow.
The aim of the present paper is to construct a complete topological invariant for codimension one homoclinic flows on closed surfaces. In contrast to the gradient case, such flows possess a unique homoclinic saddle loop and may contain periodic orbits. These flows were researched in [3] and [4]. The proposed invariant is again based on a chord diagram equipped with additional combinatorial data adapted to the homoclinic loop. As an application, we derive explicit formulas for counting topologically distinct codimension one homoclinic flows on the sphere.
Throughout this paper, a graph is understood as a finite one-dimensional CW-complex (topological graph). In particular, loops and multiple edges are allowed.
Homoclinic Flows on Closed Surfaces
Throughout this paper we are assuming that is the two-dimensional sphere and
is a flow on a closed surface . Furthermore, we denote a set of all the sources of , a set of all the saddles of and a set of all the sinks of .
Definition 2.1. A set
is called a stable manifold of a singular point of the flow whereas a set
is called an unstable manifold of a singular point .
Definition 2.2. Let be a saddle point of . If there exists a trajectory of such that
then is called a homoclinic saddle, and is called a homoclinic trajectory.
We also state a neighbourhood of a subset of to be regular in case the singular points of belonging to are the only singular points belonging to and does not intersect any trajectory of more than twice.
Let be a homoclinic trajectory of a saddle . Put .
The closure of the component containing trajectories for which is either the -limit set or the -limit set is called the -component (where stands for hyperbolic, since this component contains one hyperbolic sector). The closure of the other component is called the -component (standing for double saddle, since it is a two-dimensional disk which is topologically equivalent to the one with a double saddle on its boundary).
In this paper, we study codimension-one homoclinic flows on the boundary of Morse flows on the two-dimensional sphere. Such flows naturally arise in one-parameter bifurcations of Morse vector fields and can be approximated by Morse vector fields. We introduce a complete topological invariant for these flows, define an equivalence relation on these invariants, and construct a finite code which determines the invariant, and hence the flow itself, up to topological equivalence.
Definition 2.3. A vector field on is called an -vector field (respectively, an -vector field) if it satisfies the following conditions:
It has finitely many singular points and periodic orbits all of which are hyperbolic.
For every trajectory contained in the -component, both its - and -limit sets are critical elements.
For every trajectory contained in the interior of the -component, its -limit set (respectively, -limit set) is a critical element, whereas its -limit set (respectively, -limit set) is either a critical element or the boundary of the -component.
The vector field contains a homoclinic trajectory .
(5) If
where , then
Definition 2.4. A flow on a closed surface is called an -flow (respectively, an -flow) if it is generated by an -vector field (respectively, an -vector field).
-flows and -flows will be referred to collectively as homoclinic flows. These are precisely the codimension one flows considered throughout this paper.
Complete Topological Invariant
Atoms and Distinguishing Graphs
Let
be an -flow on with a homoclinic saddle and homoclinic trajectory .
Denote by the disjoint union of arbitrarily fixed regular neighborhoods of all sources and repelling periodic orbits of , and by the disjoint union of arbitrarily fixed regular neighborhoods of all sinks and attracting periodic orbits of . Let be a fixed regular neighborhood of the set .
The periodic orbits of decompose the surface into connected regions. The closures of these regions are called atoms. The unique atom containing the set is called the -atom, while all remaining atoms are called -atoms.
The homoclinic trajectory divides the -atom into two connected components. The closure of the component whose interior is disjoint from is called the -atom; the closure of the other component is called the -atom.
Henceforth, if we speak of atoms, we will mean only -atoms, -atoms, or -atoms. Moreover, we agree that an orientation on each atom is defined by the direction of the homoclinic trajectory.
Also it is worth drawing one’s attention that we will define for a specific subset a flow
so that for any , we obtain provided the equation does not arise any contradictions.
We first define the distinguishing graph of a -atom.
Definition 3.1. Let be a -atom. The distinguishing graph of is the disconnected union
where
and the following conditions hold:
The vertices of are the points of
while the edges are the connected components obtained after removing the vertices.
The vertices of are colored blue, whereas the vertices of are colored red.
The edges from () are colored thick red (thick blue) whereas the other edges in () are colored thin red (thin blue).
Every edge of the distinguishing graph lying on the boundary of a neighborhood of a periodic orbit is oriented according to the direction of the flow.
The connected components of the union of all oriented edges of a distinguishing graph are called its oriented cycles. All the connected components of the union of all the thick red and thick blue edges of a distinguishing graph are called its boundary cycles.
Definition 3.2. Let be an -atom (or an -atom). The distinguishing graph of is a union of a disconnected union of graphs and and the set where
and the following conditions hold:
(1) The vertices of are the points of
while the edges are the connected components obtained after removing the vertices.
(2) The vertices of are colored blue, whereas the vertices of are colored red, and the set is colored black.
(3) The edges from () are colored thick red (thick blue) whereas the other edges in () are colored thin red (thin blue).
(4) Every edge of the distinguishing graph lying on the boundary of a neighborhood of a periodic orbit or a homoclinic trajectory is oriented according to the direction of the flow.
As in the definition before, the connected components of the union of all oriented edges of a distinguishing graph are called its oriented cycles. All the connected components of the union of all the thick red, thick blue and black edges of a distinguishing graph are called its boundary cycles.
We are now ready to define the distinguishing graph of the entire flow.
Definition 3.3. The distinguishing graph of an -flow on a closed surface is the collection consisting of the distinguishing graphs of the -atom, the -atom, and all -atoms of ; besides, following conditions hold true:
(1) The non-black oriented cycles occurring in these distinguishing graphs are labeled by the integers , where is the number of periodic orbits of whereas the black oriented cycle is label by .
(2) Two non-black oriented cycles receive the same label if and only if they belong to distinguishing graphs of different atoms and correspond to the same periodic orbit of the flow.
Definition 3.4. Two finite collections of graphs are said to be isomorphic if they contain the same number of graphs and there exists a bijection between the two collections such that corresponding graphs are isomorphic.
Theorem 3.5. Let and be two -flows on , each having periodic orbits. Then and are topologically equivalent if and only if their distinguishing graphs and satisfy the following conditions.
(1) The collections and are isomorphic, and the corresponding graph isomorphisms preserve the colors of the edges and the orientations of the oriented cycles.
(2) There exists a permutation
with its rest point such that if an oriented cycle of a graph in is labeled by , then the corresponding oriented cycle of the corresponding graph in is labeled by .
Structural Graph
Now let us construct a structural graph of an -flow on the 2-dimensional sphere .
Suppose that has -atoms, . We denote them as . Also we have an -atom and an -atom .
We fix in an arbitrary way points , , and , .
Definition 3.6. A structural graph of the flow is a topological graph for which following conditions hold true:
(1) are the vertices of this graph and there are a symbol ‘+’ near and a symbol ‘-’ near ;
(2) two vertices of this graph are connected with a single edge if and only if the corresponding atoms on are adjacent;
(3) if a vertex belongs to an -atom, then it is colored blue, otherwise it is colored black;
(4) if an edge connects two black vertices it is colored black and called an -edge;
(5) if an edge is not an H-edge then it is colored red if it connects two vertices corresponding to atoms divided by an attracting cycle and it is colored blue if it connects two vertices corresponding to atoms divided by a repelling cycle;
(6) edges which are not black are assigned with numbers from 1 to in a random way;
(7) there is a symbol “+“ or “−“ near each edge provided that “+“ is placed iff the appropriate closed or homoclinic trajectory is directed along the defined orientation.
A structural graph is actually a tree — a connected graph without cycles.
Theorem 3.7. Let be a distinguishing graph for an -flow on . Let also be a tree in which vertices and edges are colored either black or blue. Then is a structural graph for if there is a bijection mapping a set of vertices in into a set of distinguishing graphs in and there is a permutation for which next conditions hold true:
(1) blue vertices in are mapped into graphs for C-atoms for whereas the black vertex with the symbol “+“ is mapped into the graph for the -atom and the black vertex with the symbol “−“ is mapped into the graph for the -atom;
(2) two vertices in are connected with a black edge if and only if corresponding graphs in have black edges;
(3) two vertices in are connected with a red edge labeled by number if and only if corresponding graphs in have red oriented cycles having the same label ;
(4) two vertices in are connected with a blue edge labeled by number if and only if corresponding graphs in have blue oriented cycles having the same label ;
(5) there is always “+“ near the black edge in and any other edge labeled by number is marked by “+“ iff the corresponding graphs have oriented cycles having the same label which assign the direction of the appropriate closed trajectories with respect to the defined orientation on .
But it is important to highlight that actually Theorem 3.5 provides a full list of conditions which help identify whether two -flows are topologically equivalent or not. Furthermore, having a distinguishing graph for an -flow one has an opportunity to recover the flow according to its topological nature.
Nevertheless, it is not enough to recover completely such a flow having only its structural graph at disposal — one can only understand how periodic orbits of the flow are placed on the sphere, the behaviour of trajectories near them and which atom contains a homoclinic trajectory. It can be helpful in some cases but not when the overall topological characteristics of the flow is needed. Apart from that, a structural graph can be obtained from the distinguishing graph (but not on the contrary).
Encoding of Homoclinic Flows
Let
be an -flow with closed trajectories, , on the two-dimensional sphere .
After the speculations from the previous section, we now can easily construct a distinguishing graph for this flow having oriented cycles. Furthermore, we are able to construct a structural graph for .
For convenience we will make sure that an edge in a structural graph and a pair of oriented cycles which are corresponding to the same closed trajectory in are assigned with the same number . It appears that now each -atom is assigned unambiguously with a specific natural number.
Let be the set of blue vertices and be the set of black vertices in . We are about to construct a function for numbering the blue vertices in
We will require a helpful notation for making our speculations plainer. Let be a set of vertices in a tree which have degree 1. Denote the subgraph of which can be obtained through the exclusion of the vertices from and the edges incident on them.
Returning to , we define according to the following algorithm:
on the first step, we choose a non-black edge in with an endpoint of degree 1 which has the least assigned natural number (we denote it as ) — thus, we state that ; the corresponding edge has colour (red or blue);
on the -th step, we choose a non-black edge in with an endpoint of degree 1 which has the least assigned natural number (we denote it as ) — thus, we state that ; the corresponding edge has colour (red or blue);
the procedure is done until all the blue vertices have been numbered — afterwards we state that and provided that is a vertex belonging to the -atom and is a vertex belonging to the -atom.
We are about to define a structural part for the flow code. Its idea is directly derived from the concept of construction of a Prüfer sequence for standard trees but it is refined so it can be used for coding -flows in a proper way.
Now we construct two specific functions
To describe it less formally, our structural part of the code will have symbols in itself. These mappings will explain which exactly symbol will be placed in each position and which indices it will possess. So, for we claim that
where is the other endpoint of the edge in the graph incident to ; furthermore, the same edge in is assigned with the label .
Definition 4.1. A structural flow code for the flow is a sequence of symbols
where instead of dots the symbols assigned respectively with their lower right indices are consecutively written respectively in colours .
Remark 4.2. If the flow has no -atoms, then the structural flow code has no symbols inside round brackets.
What is left to do is to write a code for distinguishing graphs for all the atoms. But we must memorize all the objects and notations made in this chapter before.
Let’s fix an arbitrary atom . We construct a flow code for its distinguishing graph .
We denote the set of all the vertices of as and the set of all the oriented cycles as . Now, let
be a mapping which maps each cycle from into where is a vertex in corresponding to the atom adjacent to through exactly this cycle. Additionally, where
We also fix a permutation which maps an endpoint of a thin non-black edge in to the other endpoint.
Also note that, of course, it may appear that . We state at once that the flow code for this kind of atoms is an empty string. Hereafter, we assume has a non-empty set of vertices.
Definition 4.3. A sequence of symbols
from is called a -like code list if any two-digit number in it is separated by two black commas from both left and right side; furthermore, where but so that contains an opening bracket and contains a closing bracket.
Definition 4.4. A sequence of symbols
where , is called an HC-like code if each , , is an HC-like code list and the first inclusions of natural numbers in the code form an increasing sequence; moreover, so that contains an opening bracket and contains a closing bracket.
Definition 4.5. A cycle list in an HC-like code is a sequence of symbols
where , and , , , are fixed provided that:
each , , is a number considered with memorizing in which code list it is placed and where namely in this code list it is placed;
if , , then the next number in the list (if is the last one must be the first one) has its second location in the code — the corresponding code list is and ;
but for any ;
numbers among having more than one digit in this sequence are separated by commas from both left and right sides.
A set of cycle lists being a set of all cyclic permutations of the same cycle list is called a cycle.
Definition 4.6. A sequence of symbols
is called a cycle-like code list for an -like code if is a cycle list for this code and so that contains an opening bracket and contains a closing bracket.
Definition 4.7. A sequence of symbols
is called a cycle-like code for an -like code if is list of some mutually distinct cycle-like code lists for the -like code and so that contains an opening bracket and contains a closing bracket.
Now, we consider a subgraph of which contains all non-red edges and blue vertices.
Suppose that are boundary cycles of . Let also , , be some arbitrarily fixed point — we call it a distinguishing vertex. We have an ordered set
obtained after bypassing completely along the fixed orientation starting from the point .
We start bypassing each boundary cycle according to the defined motion, gradually increasing by 1. Once we meet any point , , we define in this point a value of a function
which tells us endpoints of how many thin blue edges we have met since the beginning of bypassing . Then we claim a value of a corresponding function
in the same point. corresponds to , , , if:
and ;
, , where corresponds to ;
, where corresponds to .
Definition 4.8. Let , , be a boundary cycle in and be its distinguishing vertex. A sequence of symbols from
colored blue is called a code list for with a distinguishing vertex provided that:
if is a thick blue edge which is not an oriented cycle then and ;
if is a thick blue oriented cycle then and — there is a string in the right lower corner of the closing curly bracket in which denotes ;
if is a black oriented cycle then and ;
numbers having more than one digit in this sequence are separated by commas from both left and right sides.
Definition 4.9. Let be an arbitrary fixed vertex of . A sequence of symbols
is called a flow code for the graph with a distinguishing vertex if each , , is a code list for a boundary cycle with a distinguished vertex for the flow so that:
, ;
if we take the least non-zero number in for each included in it only once, it will represent some — then and is a boundary cycle containing ;
and — there is a string in the right lower corner of the closing round bracket in that defines where is the vertex in corresponding to the atom .
Code lists for boundary cycles of with distinguishing vertices are actually -like code lists while the flow code for with a distinguishing vertex is actually an -like code.
After collapsing all the boundary cycles in the interior of the atom into points we will get that all the thin blue edges in divide it into different 2-cells which we will call canonical 2-cells. Cycles, in fact, give us the boundaries of these 2-cells in atoms — a closure of each of them contains exactly one sink or one attracting closed trajectory or an attracting homoclinic trajectory if we speak of the -atom.
To be more precise, there are several cycle lists which may correspond to a single canonical 2-cell. So, we agree that we call a canonical cycle list such a cycle list for such a 2-cell which starts from the least possible natural number.
Definition 4.10. A sequence of symbols
colored red is called an advanced cycle if is a canonical cycle list for a specific canonical 2-cell with a red oriented cycle inside it, and — there is a string in the right lower corner of the closing curly bracket in which denotes .
The constructed code unambiguously defines a placement of stable manifolds of saddles inside of atoms. Moreover, during constructing the code we bypass different fragments of boundaries of these 2-cells — we will say in this case that we meet different cycles.
The only thing left to define by the code is where we have exactly closed trajectories and how the orientation of these trajectories behaves. Thus, we have just to write down specific codes for the appropriate cycles.
We define a linear order on the set of cycles in the atom — for instance, a cycle is considered to be not greater than a cycle provided that the first meeting of a number from which the cycle is formed happens before the first meeting of a number from which the cycle is formed.
Definition 4.11. Suppose that the atom has different cycles, . A sequence of symbols
is called a cycle code for the atom if:
unless is the -atom, is an empty string; otherwise it is an advanced cycle of corresponding to the homoclinic trajectory in its closure;
are all its advanced cycles corresponding to all the attracting closed trajectories belonging to the closure of arranged in ascending order;
and — there is a string in the right lower corner of the closing round bracket in that defines where is the vertex in corresponding to the atom .
We define a linear order on the set like that:
Now we are ready to write down two sections of a distinguishing flow code for according to the further definition.
Definition 4.12. A distinguishing flow code for is two strings of symbols
where the all non-empty flow codes for all the atoms containing blue symbols must replace the dots in the first string arranged in ascending order of their indices, and all the cycle codes for all the atoms of colored red and arranged in ascending order of their indices must replace the dots in the second string.
We also call a distinguishing flow code canonical if the numbers in its first string (not the indices) are placed in a lexicographic order according to the canonical order of natural numbers.
We can see a diagram for a specific -flow on the two-dimensional sphere in Figure 1. It is a rough sketch of the flow where its closed trajectories, singular points, the homoclinic trajectory, stable and unstable manifolds of all the saddles are highlighted.
Figure 1. Separatrix diagram of the -flow on the two-dimensional sphere
Its structural graph and distinguishing graphs are depicted below.
For the distinguishing graph describing the -atom the blue vertex on the black edge is claimed to be a distinguishing vertex.
For the distinguishing graph describing the -atom a blue distinguishing vertex is stated to be an upper blue vertex in oriented cycle 3.
Finally, for the distinguishing graph describing the atom corresponding to the vertex with label 1 in the structural graph the vertex on the blue oriented cycle is a distinguishing vertex.
The full flow code is as follows:
Figure 2. Structural graph of the -flow
Figure 3. Distinguishing graph of the -atom
Code as a Complete Topological Invariant
We are going to refine the aforementioned approach to constructing flow codes, research the properties of the new codes and ways to understand whether three sequences of symbols correctly define an -flow on the 2-dimensional sphere.
First, we are going to number the vertices of the structural flow code unambiguously. We leave labels “0 +“ and “0 −“ at their natural places. The algorithm is following.
Suppose that there is an observer in the point “0 +“ standing with his back to the black edge. He numbers all the endpoints of degree 1 from which one can find a path to the point with label “0 +“ without passing the point with label “0−“ (if there are any) consecutively moving from one point to another in a counterclockwise direction starting from the least unused yet natural number and gradually increasing it by 1.
Then, the observer shifts to the point with label “0 −“ and numbers all the endpoints of degree 1 from which one can find a path to the point with label “0 −“ without passing the point with label “0 +“ (if there are any) consecutively moving from one point to another in a counterclockwise direction starting from the least unused yet natural number and gradually increasing it by 1.
Figure 4. Distinguishing graph of the -atom
Figure 5. Distinguishing graph of the -atom adjacent to the -atom
(3) Afterwards, the observer excludes all the numbered vertices and the edges incident on them and repeats the operation until he runs out of unlabeled vertices.
The obtained numbering is called a canonical numbering. There is the only numbering of closed trajectories in an -flow for which the function on the set of the vertices defined in the previous chapter gives us the canonical numbering. Hereafter, we will presume that the vertices of the structural graph have always the canonical code and that the oriented cycles of the distinguishing graphs are labeled in a relevant way.
Definition 5.1. An -flow on the two-dimensional sphere with closed trajectories, , is called distinguished if for each its atom some single non-homoclinic trajectory belonging to a stable manifold of some saddle point if it exists and its alpha-limit set are fixed.
Thus, the corresponding fixed trajectory in each atom is called distinguished as well.
In this way, we automatically fix in the distinguishing graph for each atom a distinguishing vertex.
Definition 5.2. Suppose that is the 2-dimensional sphere. Two distinguished -flows
are strongly topologically equivalent if there exists a homeomorphism which maps each trajectory of the first flow onto a trajectory of the second one without reversing orientations and maps each distinguished trajectory on into a distinguished trajectory on $M.
We fix beforehand a set
Let us consider any fixed finite sequence of symbols from enclosed in the string
We number them as , . Thus we have obtained a multiset where elements may repeat, therefore each of them should be considered with its position in the structural flow code. Then we define a function
on the set of these symbols according to the following principle — iff is the least natural number which is not met among the symbols and which is greater than , .
Theorem 5.3. A structural flow code for an -flow on the 2-dimensional sphere is a string
where there are symbols (not necessarily different), , from or just an empty string instead of dots provided that in the first case the following conditions are true:
all the symbols are colored either red or blue;
the last symbol of the sequence contains either “0 +” or “0 −”;
the sequence does not contain natural numbers greater than and less than 2;
if one excludes symbols “0 +” and “0 −” from the string then a string will be obtained where each natural number followed by either the same number or by this number increased by 1;
provided that one excludes all the indices from the string, if “0 +” or “0 −” in the code is followed by another but different symbol then it never appears in the code again;
the symbol containing a natural number must not appear later than in -th position in the code.
Proof. The first statement in the theorem is obvious due to the described algorithm of constructing the structural flow code.
As for the second statement, according to the algorithm, the last symbol deals with the only left non-black edge which is attached to the vertex with either a label “0 +” or a label “0 −”. Hence, the second statement is true as well.
One also has no difficulties in seeing a reason why the third statement is true — if the structural flow code has symbols then it has vertices numbered from 1 to and two vertices assigned with labels “0 +” and “0 −”. Besides, there is no way 1 can appear in the code what follows from the canonical numbering.
The fourth, the fifth and the sixth statements are true what follows from the way of constructing a canonical numbering of the structural graph.
Theorem 5.4. The first string in a distinguishing flow code for an -flow on the 2-dimensional sphere is a string
where there is a concatenation of multiple -like codes placed in ascending order of the indices near their closing brackets instead of dots with the following properties:
the concatenation certainly contains an -like code with an index 0+ and this code is the only one which contains angle brackets;
there is only one number inside angle brackets;
each symbol inside of each -like code list is colored blue;
each natural number present in an arbitrary fixed -like code is placed there twice;
in any -like code each pair of adjacent -like code lists in this sequence contains a pair of the same natural numbers placed in different code lists;
(6) if there is an -like code with its corner index which contains an -like code list in curly brackets with some symbol in its right lower corner then provided that there is another -like code with its corner index , contains a code list in curly brackets with the symbol in its right lower corner;
(7) if is a number of the -like code lists in an arbitrary -like code, is the greatest natural number in this very code and is a number of different cycles in this code then
Proof. The first three statements are quite comprehensible.
The fourth statement is true because each natural number present in the code corresponds to an intersection of a stable manifold of a saddle (or a non-homoclinic trajectory belonging to a stable manifold of a saddle if we speak of the homoclinic saddle) with a regular neighbourhood of a source or of a repelling cycle (or with a regular neighbourhood of the union of the homoclinic saddle and a homoclinic trajectory). These intersections become intersections of boundary cycles with thin blue edges. And the same number corresponds to multiple vertices iff they are endpoints of the same thin blue edge. Thus, each natural number present in the code can appear there only twice.
The fifth statement follows directly from connectivity of the distinguishing graphs for the atoms which in its turn follows from connectivity of the considered surface.
The first part of the sixth statement implies that the atom corresponding to the -like code is adjacent to the atom corresponding to the -like code . So, the statement about the indices is fair considering how the flow code is constructed.
As for the seventh statement, let us fix the distinguishing graph for an arbitrary atom which is not the -atom and collapse each thick blue circle (which depicts a boundary of a regular neighbourhood of a source or of a repelling cycle) and each black circle (which depicts a boundary of a regular neighbourhood of the union of the homoclinic trajectory and the homoclinic saddle) into a point. All these circles correspond to the set of the code lists in the appropriate -like code. All the present thick red circles are collapsed into points as well but afterwards they are eliminated. Then we obtain a sphere with a cellular division onto it where the obtained points are 0-cells, the thin blue edges are 1-cells and the interiors of the cycles are 2-cells. As the Euler characteristics of the sphere is 2, we obtain an equality in question.
If one deals with the -atom then the seventh statement remains true as the algorithm is the same but one must glue to the component of the boundary of this atom containing the homoclinic trajectory the 2-disk containing a source so that all the trajectories flow from it to and consider that exactly one code list here corresponds not to a source but to .
Theorem 5.5. The second string in a distinguishing flow code for an -flow on the 2-dimensional sphere is a string
where there is a concatenation of multiple cycle-like codes for some -like codes from the first string placed in ascending order of the indices near their closing brackets instead of dots with the following properties:
(1) each symbol inside of each cycle-like code list is colored red;
(2) a cycle-like code with an index “0 − “ contains a cycle-like code list with an index “0 + “;
(3) each cycle-like code for an -like code from the first string has an index in the right lower corner which is assigned to this very -like code in the first string;
(4) each cycle list from a cycle-like code with an index represents a cycle from the corresponding -like code; among all the cycle lists representing this cycle the list starts from the least possible natural number among those not included in round brackets;
(5) cycle-like code lists in each cycle code are placed in ascending order of corresponding cycles;
(6) if there is a cycle-like code with its corner index which contains a cycle-like code list in curly brackets with some symbol in its right lower corner then provided that there is another cycle-like code with its corner index , contains a code list in curly brackets with the symbol in its right lower corner.
All the statements are obviously reasonable due to the construction of the second string in the distinguishing flow code.
There also have to be some specific properties for both flow codes which have to generate a structural and distinguishing graphs for -flows without arising any contradictions.
Theorem 5.6. Let the sequence of symbols in the round brackets in the structural code for an -flow on the 2-dimensional sphere form a multiset . For the first string in a distinguishing flow code and a structural flow code for next conditions are equivalent provided that the string contains an HC-like code with an index :
(1) the code contains a code list with curly brackets and a corner index ;
(2) () is an element of the multiset having corner index “+” or “−” and in one of its positions it is colored blue and mapped into () by .
Theorem 5.7. Let the sequence of symbols in the round brackets in the structural code for an -flow on the 2-dimensional sphere form a multiset . For the second string in a distinguishing flow code and a structural flow code for $g next conditions are equivalent provided that the string contains a cycle-like code with an index and that so that either “0 −” or “0 +”:
(1) the code contains a code list with curly brackets and a corner index ;
(2) () is an element of the multiset having corner index “+” or “−” and in one of its positions it is colored red and mapped into () by .
Theorem 5.8. Let the sequence of symbols in the round brackets in the structural code for an -flow on the 2-dimensional sphere form a multiset . Suppose that the first string in a distinguishing flow code for does not contain an HC-like code with its index . Then if is present in the structural flow code only one of three following options is possible:
(1) “0 +” and it appears in the structural flow code only once colored blue;
(2) depicts some natural number not greater than and it appears in the structural flow code only once; besides, another symbol having the other colour exists in the code so that maps it into .
Definition 5.9. The ordered set of the three strings of symbols are called an -defining triple if they are enclosed in the following templates:
The symbols placed instead of the dots in each string are called a kernel of the appropriate string.
Theorem 5.10 (Realisation theorem). Suppose that there is an -defining triple. Furthermore, for the first string the list of conditions from Theorem 5.3 is true, for the second string the list of conditions from Theorem 5.4 is true, and for the third string the list of conditions from Theorem 5.5 is true. Besides, the statements from Theorem 5.6, Theorem 5.7 and Theorem 5.8 are true as well.
Then, these three strings form a structural and a distinguishing flow code for a specific distinguished -flow on the 2-dimensional sphere which is unique with regard to strong topological equivalence of -flows.
Proof. Three codes give us unambiguously the structural graph for the flow and all the distinguishing graphs for all its atoms. If code lists for some atoms are absent, it means these atoms contain no saddles and the structure of their distinguishing graphs is implied from what is written in the structural flow code. Furthermore, we can restore all the labels assigned to the oriented cycles on the graphs and tell that two oriented cycles are equivalent iff they are assigned by the same number.
Now, for each pair of equivalent oriented cycles if they depict an attracting closed trajectory we take a system of differential equations on the plane dealing with polar coordinates:
Then, we take a regular neighbourhood of the closed trajectory placed along the curve . The sign in the second equation depends on which direction the movement on this trajectory is moving. If we are interested in a repelling closed trajectory, we take a bit different system of differential equations:
The structural graph gives us a configuration of these trajectories relatively to each other. Also we are taking a regular neighbourhood of a union of a homoclinic trajectory and a homoclinic saddle of an appropriate dynamical system on the Euclidean plane — it is depicted by a pair of black oriented cycles.
Afterwards, in each atom we replace a non-oriented blue boundary cycle with a 2-disk
with the specified flow generated by the vector field
on it. Each thin blue edge is replaced by a region
with the specified flow generated by the vector field
on it. Finally, each canonical 2-cell without a red oriented cycle inside it is replaced by a 2-disk but with the specified flow inverse to the defined one on it before. Each empty region is filled by a 2-disk but there is another specified flow on it and it is generated by the vector field
so that the trajectories were directed towards attracting cycles. We got a phase portrait of a smooth dynamical system on the 2-dimensional sphere after consecutive smoothing of the vector firlds right before gluing.
Theorem 5.11 (Strong topological equivalence criterion). Two distinguished -flows on the 2-dimensional sphere are strongly topologically equivalent iff their triples of structural flow codes and parts of distinguishing flow codes are identical.
Now, let us research a phenomenon of ordinary topological equivalence of such kind of flows. We will assume that the sets have the same definition as they did in the previous chapter. Also, for an ordered multiset we denote its cyclic permutation starting from where is fixed number from . Analogously if one has an -like code list
with non-numerical symbols and and perhaps some other non-numerical symbols whose quantity is less than it will be said without arising any contradictions that
where forms a multiset (as earlier, two-digit numbers must be separated with commas from each side).
Let us presume that we have a sequence of code lists
We will define a listing linear order on the set of numerical symbols included in this list so that for two numerical symbols and from this sequence iff the first apparition of happens earlier in this sequence than the first apparition of .
Definition 5.12. Let
be an -like code including -like code lists and non-numerical symbols and . Then a new -like code
with -like code lists and is called an equivalent -like code regarding to the symbol where and are fixed, , if it is created according to the next algorithm:
first, , and for , , , iff due to the listing linear order defined on the sequence of code lists there exists the least numerical symbol which appears in this sequence only once but its second copy appears in ;
(2) second, is formed from through writing it down but replacing numerical symbols in the following way — each copy of a number is replaced by a copy of a number which means that the first apparition of in has number among all first apparition of the symbols.
Now, let us assume that a cycle-like code list for the -like code
is fixed. Suppose that the relevant cycle list is starting with some numerical symbol , where is fixed, . After the actions done in the items (1) and (2) in the Definition 5.12, this symbol was moved into some position in some fixed -like code list , . Then, the third action turned this symbol into another numerical symbol belonging to . Thus, the cycle-like code list relevant to the cycle list starting with will be called relevant to the cycle-like code list through the equivalence mentioned in Definition 5.12.
Definition 5.13. Suppose that and are cycle-like codes where and are some non-numerical symbols. Then the second cycle-like code is called relevant to the first cycle-like code through some equivalence if for each and are relevant through the same equivalence.
It is worth indicating that if one has a distinguishing graph for any atom in an -flow on the 2-dimensional sphere and has constructed the code
mentioned in Definition 5.12 just as the flow code for with some distinguished vertex then the shift to the equivalent -like code from this definition is merely constructing another flow code for but with another distinguishing vertex represented in the initial flow code through the symbol ! Moreover, the shift to the relevant cycle-like code list through this very equivalence is just a reencoding some canonical cycle in the atom using new flow code.
Thus, it is correct to speak of two equivalent -like codes if any of them is equivalent to the other one regarding to the specific numerical symbol in the specific -like code list.
Example 5.14. Let us consider the -like code
The set of all -like codes equivalent to it looks like this:
Theorem 5.15 (Topological equivalence criterion). Suppose that
are two -flows defined on the 2-dimensional sphere having their structural and distinguishing flow codes expressed as -defining triples , . Then these flows are topologically equivalent iff three following statements are true:
the first strings in and are identical;
kernels of the second strings in and form respectively concatenations of -like codes and so that for each is equivalent to ;
kernels of the third strings in and form respectively concatenations of cycle-like codes and , where is the number indicated in the previous item and for each , is relevant to through the equivalence mentioned in the previous item.
Two -defining triples whose properties described in items (1) – (3) in this theorem are called equivalent.
Examples
Let us introduce another definition.
Definition 6.1. Two -flows defined on the 2-dimensional sphere are called structurally equivalent if they have identical structural flow codes.
It is pretty obvious that from strong or ordinary topological equivalence of such flows follows their structural equivalence. Notwithstanding, their structural equivalence does not yet implies any kind of their topological equivalence.
It is possible to distinguish how many -flows up to their structural equivalence one can construct provided these flows have at most 6 closed trajectories. Of course, tangible results will be obtained by means of properties of structural flow codes researched in the previous section (first of all, Theorem 5.3 is needed). So, let us realize the magnitude of the problem.
The flow may have no trajectories at all, and then the kernel of its structural graph is empty. There is only 1 such flow up to structural equivalence.
If the flow has 1 closed trajectory, the kernel of its structural flow code may contain “0 + “ or “0 − “ and no natural numbers at all. This symbol has to be colored either red or blue. As a result — 8 flows.
If the flow has 2 closed trajectories, the kernel of its structural flow code does not contain natural numbers distinct from 2. Overall, we have 96 flows.
If the flow has closed trajectories, , the kernels of its structural flow code does not contain natural numbers greater than and less than 2. Here one may need more combinatorial methods to calculate the number of these flows.
Having symbols in the structural flow code, we are about to count the number of ways to place “0 − “ and “0 + “ in this code. This number is obviously calculated like this without considering so far that each symbol is supposed to have a specific colour and a specific index:
After we placed symbols “0 + “ and “0 − “ in any way in some places of the kernel of the structural flow code some places remain vacant but we remember their order number in the flow code. Henceforth, we introduce a symbol
where and denote natural numbers, , and denote consecutive natural numbers not greater than (mind you, the set may certainly be empty). It denotes a number of ways to write a complete form of the kernel of the structural flow code with symbols where last symbols and the symbols written in positions from to are filled with “0 + “ and “0 − “.
Thus, the number of all the ways of writing the structural flow code with symbols is equal to
Note that for an arbitrary fixed natural number and for any
The problem of direct calculating of these structural flow codes requires more advanced methods so for simplicity we consider small enough numbers of closed trajectories. We must calculate all the appropriate values of so as to evaluate the overall number of the flows for a specific .
. We need to analyze the following values:
denotes a number of ways to fill the second position in a three-position structural flow code if all the other positions are already filled. The option is unique — we can put only 3. So,
Analogously,
Thus,
N=4. Now we are interested in the following values:
Using the aforementioned formula, we are getting the following result:
So, due to structural equivalence of -flows, you can generate 1 such flow with no closed trajectories, 8 flows with 1 closed trajectory, 96 flows with 2 closed trajectories, 2560 flows with 3 closed trajectories and 32768 flows with 4 closed trajectories.
Now it is worth researching a number of homoclinic flows on the 2-dimensional sphere due to topological equivalence.
Let
be a Morse flow on a 2-dimensional sphere . Suppose that is an arbitrary source of this flow and is its regular neighbourhood. Then, considering that trajectories which come out of and flowing into saddles of divide into sectors. One may construct a -flow following these steps:
First, we have to construct the -atom. We take and an arbitrary sector near it. It defines unambiguously a sink of this flow, or, more precisely, its cellular neighbourhood whose boundary is formed by trajectories belonging to stable manifolds of saddles (or the saddle if it is unique) of . We transform so that there is a new homoclinic saddle inside this neighbourhood with its homoclinic trajectory, the trajectory coming out of into and the trajectory coming out of into .
Second, we construct the -atom. We have to fill in the space inside the homoclinic trajectory. In order to do that, we should take an arbitrary Morse flow on a 2-dimensional sphere and to transform an arbitrary sink in into a hole in the sphere which contains an attracting closed trajectory. Thus, we glue the aforementioned -atom and the newly constructed flow on the sphere with 1 hole.
If one intends to add some -atoms, he has to do the second step for any source or sink in keeping in mind that a source must be replaced by a 2-disk having a repelling closed trajectory and a sink must be replaced by a 2-disk having an attracting closed trajectory.
Let denote the number of choices of Morse flows on the 2-dimensional sphere with a marked oriented sector near a source (due to topological equivalence) and saddles and markings of sources or sinks in them. It is equivalent to calculation of the number of all possible -atoms with non-homoclinic saddles and closed trajectories.
Let also denote the analogous number of choices of Morse flows on the 2-dimensional sphere with a marked source together with a given orientation at the boundary of its regular neighbourhood. It is, in its turn, equivalent to calculation of the number of all possible ways to construct -atoms and -atoms with non-homoclinic saddles and closed trajectories.
Besides, we introduce notations , .
The following formula holds true:
Let be a number of homoclinic flows on the 2-dimensional sphere with saddles and closed trajectories.
Theorem 6.2. The following formulas hold true:
After obtaining these formulas it is more convenient to go on with further calculations which help us to comprehend the number of homoclinic flows for some small numbers of saddles and closed trajectories in it. By means of the program code [10] we complete calculations for small numbers of and thus obtaining the following table of results:
| 1 | 2 | 8 | 24 | 64 |
| 2 | 8 | 60 | 288 | 1072 |
| 3 | 38 | 420 | 2752 | 12896 |
| 4 | 220 | 3132 | 25600 | 142048 |
| 5 | 1484 | 25488 | 246160 | 1558048 |
| 6 | 11124 | 221920 | 2453952 | 17352576 |
| 7 | 90098 | 2035500 | 25233504 | 196745568 |
| 8 | 769108 | 19382868 | 265686960 | 2264046112 |
| 9 | 6839220 | 189962696 | 2849218640 | 26355749664 |
Table 1.
Declaration on the use of AI tools
No generative AI tools were used in the preparation of this paper. The mathematical ideas, construction of the topological invariant, proofs of the main results, computational algorithms, numerical calculations, and figures are entirely the work of the authors. Both authors take full responsibility for the mathematical content and all results presented in the paper.
References
- [1]J. Gross and T. Tucker. Topological graph theory. Wiley, 1987.
- [2]B. I. Hladysh and O. O. Prishlyak. Deformations in the general position of the optimal functions on oriented surfaces with boundary. Ukrainian Mathematical Journal, 71(8):1173–1185, 2019.DOI
- [3]Ale Jan Homburg, Maria Kellner, and Jürgen Knobloch. Construction of codimension one homoclinic cycles. Dynamical Systems, 29(1):133–151, November 2013.DOI
- [4]Vladislav Kybalko and Tomoo Yokoyama. Topological characterizations of Morse–Smale flows on surfaces and generic non-Morse–Smale flows. Discrete and Continuous Dynamical Systems, 42(10):4787, 2022.
- [5]Z. Kybalko, A. Prishlyak, and R. Shchurko. Trajectory equivalence of optimal Morse flows on closed surfaces. Proc. Int. Geom. Cent., 11(1):12–26, 2018.
- [6]A. A. Oshemkov and V. V. Sharko. Classification of Morse-Smale flows on two-dimensional manifolds. Matem. Sbornik, 189(9):93–140, 1998.DOI
- [7]J. Palis and S. Smale. Structural stability theorems. Global Analysis (Proc. Sympos. Pure Math., Vol. XIV, Berkeley, Calif., 1968), 14:223–231, 1970.
- [8]M.M. Peixoto. On the classification of flows of 2-manifolds. Dynamical Systems (Proc. Symp. Univ. of Bahia, Salvador, Brasil, 1971), pages 389–419, 1973.
- [9]A. Prishlyak and I. Ovtsynov. Optimal codimension one gradient flows on closed surfaces. Proceedings of the International Geometry Center, 18(1):68–101, 2025.DOI
- [10]Alexandr Prishlyak. Codes of Rotation Systems of a Cell-Embedded Graph in an Oriented Surface, 2026.DOI
- [11]Alexandr Prishlyak and Ilia Ovtsynov. The structure of Morse and codimension one gradient flows on the sphere with holes. Memoirs on Differential Equations and Mathematical Physics, 97:137–158, 2026.arxiv.org/abs/2601.06926
- [12]Andrii Prus, Alexandr Prishlyak, and Sophija Huraka. Flows with collective dynamics on a sphere. Proceedings of the International Geometry Center, 14(1):60–79, May 2021.DOI




