Elena V. Konstantinova
[email protected]Son En Gun
[email protected]Sobolev Institute of Mathematics, Ak. Koptyug av. 4, Novosibirsk, 630090, Russia
Novosibisk State University, Pirogova str. 2, Novosibirsk, 630090, Russia
Abstract
The Pancake graphs , are Cayley graphs over the symmetric group generated by prefix-reversals. There are six generating sets of prefix-reversals of cardinality three which give connected Cayley graphs over the symmetric group known as cubic Pancake graphs. In this paper we study the girth of the cubic Pancake graphs. It is proved that considered cubic Pancake graphs have the girths at most twelve.
The Pancake graph , is the Cayley graph over the symmetric group of permutations written as strings in one-line notation, where for any , with the generating set of all prefix–reversals inversing the order of any substring , of a permutation when multiplied on the right, i.e. . This graph is well known because of the open combinatorial Pancake problem of finding its diameter [3].
The graph is a connected vertex–transitive -regular graph of order with no loops and multiple edges. It is almost pancyclic [4, 9] since it contains cycles of length , , but doesn’t contain cycles of length or . Since the length of the shortest cycle contained in the graph is six, hence we have for any , where is the girth of . The girths of the burnt Pancake graphs over the hyperoctahedral group was considered in [2]. The (burnt) Pancake graphs are commonly used in computer science to represent interconnection networks [1, 10, 11].
Importance of fixed–degree Pancake graphs, in particular, cubic Pancake graphs as models of networks was shown in [1] by D. W. Bass and I. H. Sudborough. The authors have considered cubic Pancake graphs as induced subgraphs of the Pancake graph . The necessary conditions for a set of three prefix–reversals to generate the symmetric group were found. In particular, it was shown that the cubic Pancake graphs over the symmetric group , , are connected with the following generating sets:
The set is known as ’big-3’ flips, and the corresponding cubic Pancake graph generated by this set is called as the Big-3 Pancake network [10]. J. Sawada and A. Williams have conjectured in [10] that this graph has cyclic Gray codes.
In this paper we study cubic Pancake graphs and their girths. Our main result is obtained for the cubic Pancake graphs that are Cayley graphs over the symmetric group generated by the sets .
Theorem 1
(1)
(2)
(3)
For , the computational results on the girths of the cubic Pancake graph are given as follows:
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
6
8
10
12
12
16
16
16
20
20
20
24
24
24
28
For any , the computational results show that . One can conjecture that this is true for any .
The paper is organized as follows. The proof of Theorem 1 is based on the characterization of small cycles in the Pancake graphs. We present preliminary results with main definitions and notation in Section 2, where it is shown that for any there are no –cycles and –cycles in . Then we prove Theorem 1 in Section 3.
2 Preliminary results
The first results on a characterization of small cycles in the Pancake graph were obtained in [5] where the following cycle representation via a product of generating elements was used. A sequence of prefix–reversals , where , and for any , such that , where is called a form of a cycle of length . A cycle of length is also called an –cycle, and a vertex of is identified with the permutation which corresponds to this vertex. It is evident that any –cycle can be presented by forms (not necessarily different) with respect to a vertex and a direction. The canonical form of an –cycle is called a form with a lexicographically maximal sequence of indices . We shortly write for a cycle form , where , appears exactly times and for any . The form is canonical if . By using this description, the following results characterizing – and –cycles were obtained.
Theorem 2
[5]
The Pancake graph has independent –cycles of the canonical form
(4)
and distinct –cycles of the canonical form
(5)
where . Each of the vertices of belongs to exactly one –cycle and distinct –cycles.
The complete characterization of –cycles is given by the following theorem.
Theorem 3
[7] Each of vertices of belongs to distinct –cycles of the following canonical forms:
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
The complete characterization of –cycles in the Pancake graphs were obtained in [6].
In general, the complete characterization of small cycles in the Pancake graphs presented in [5, 6] is based on the hierarchical structure of the Pancake graphs. The graph is constructed from copies of such that each has the vertex set
where and the edge set
where Any two copies and are connected by edges presented as where Prefix-reversals defines internal edges in , and the prefix-reversal defines external edges between copies. Copies , or just when it is not important to specify the last element of permutations belonging to the copy, are also called copies.
The hierarchical structure of the Pancake graphs is used to prove the following two results.
Theorem 4
In the cubic Pancake graphs , there are no cycles of length . For there are -cycles of the canonical form .
Proof. Since the cubic Pancake graphs are induced subgraphs of the Pancake graph , then let us consider all possible cases for forming –cycles in the Pancake graphs with taking into account that the generating set of contains only three elements , where is odd. If then there are cycles of length of the canonical form (see Lemma 1 in [8]). If then due to the hierarchical structure of the Pancake graph , cycles of length could be formed from paths of length , belonging to different copies of .
Further, we consider all possible options for the distribution of vertices by copies. Without loss of generality we always put .
Case 1: 10-cycle within has vertices from two copies of
Suppose that a sought –cycle is formed on vertices from copies and , where . It was shown in [5, Lemma 2] that if two vertices and belonging to the same copy, are at the distance at most two, then their external neighbours and should belong to distinct copies. Hence, a sought cycle cannot occur in situations when its two (three) vertices belong to one copy and eight (seven) vertices belong to another one. Therefore, such a cycle must have at least four vertices in each of the two copies. Hence, there are two following cases.
Case Suppose that four vertices of a sought –cycle belong to one copy, and other six vertices belong to another copy. Let and . Since then and should belong to . Herewith, the four vertices of should form a path of length three whose endpoints should be adjacent to vertices from .
Consider all options for passing from to by internal edges in a copy . Since the generating set of consists of the elements and corresponding to internal edges then there are two ways to get paths of length five from to (see Figure 1).
The first path is presented as follows:
(14)
such that:
Since then we have:
(15)
Figure 1: Case 1: 10-cycle within has vertices from two copies of
Note that with for any . Hence, we immediately can conclude that given by (15) and belong to different copies of since for any odd . For we get and there is no path of length three between and , presented as or , since and . This gives a contradiction with an assumption that and belong to the same copy . Thus, a sought cycle cannot occur neither in nor in .
The second path is presented as follows:
(16)
such that
and hence
which means that this permutation belongs to . However, belongs to which gives a contradiction again. Thus, a sought –cycle cannot occur in this case.
Case Suppose that five vertices of a sought –cycle belong to a copy , and other five vertices belong to a copy , where , , and . Then the five vertices of should form a path of length four whose endpoints should be adjacent to the vertices from (see Figure 1).
There are two ways to get paths of length four from to .
The first way is given as follows:
more precisely, we have:
The second way is presented as follows:
and we have:
Since then we have either:
or
which gives a contradiction with an assumption that and belong to the copy for any odd . Thus, in this case a sought cycle cannot occur in , and hence in .
Case 2: 10-cycle within has vertices from three copies of
There are four possible situations in this case.
Case Suppose that two vertices of a sought –cycle belong to one copy, other two vertices belong to another copy, and remaining six vertices belong to the third copy. Let , , and , then and should belong to (see Figure 2).
It is evident that there are two ways to get paths of length one from to :
or
Since , we have either
or
Similar, there are two ways to get a path of length one from to such that the first way is given by , where we have either
or
and the second way is given by , where we have either
or
Figure 2: Case 2: 10-cycle within has vertices from three copies of
and since we get:
To get a sought -cycle there should be a path of length five between and , where . Let us check this. If the vertices , and belong to the copy , and there are two ways to get a path of length five from and . Namely, applying to we have either:
or
and applying to we have either:
or
Hence, a path of length five does not occur between and .
If the vertices and belong to the copy , and the following two cases are possible:
or
Again, a path of length five does not occur between and .
If , there is a contradiction with an assumption that and belong to the copy . Thus, a sought cycle cannot occur in this case.
Case Suppose that two vertices of a sought –cycle belong to one copy, other three vertices belong to another copy, and remaining five vertices belong to the third copy. Let , , and , then and should belong to
(see Figure 2). There are two ways to get a path of length two from to such that either
or
Similar, there are two ways to get a path of length two from to . The first one is presented as follows:
To get a sought -cycle there should be a path of length four between and , where . Let us check this. If the vertices and belong to the copy , and the following cases are possible:
or
Hence, a path of length four does not occur between and . However, let us note that in this case we have a cycle of length eight given by the canonical form obtained from (12) by putting .
If , there is a contradiction with an assumption that and belong to the copy . Thus, a sought cycle cannot occur in this case.
Case Suppose that two vertices of a sought –cycle belong to one copy, other four vertices belong to another copy, and remaining four vertices belong to the third copy (see Figure 3). Let , , and . Then both and should belong to either or , since either or . More precisely, we have:
(19)
On the other hand, there are two ways to get a path of length three from to . The first way is presented as follows:
(20)
such that we have:
Figure 3: Case 2: 10-cycle within has vertices from three copies of
To get a sought -cycle there should be a path of length three between and (see Figure 3, Subcase ). Let us check this.
If or , there is a contradiction with an assumption that both and belong to either or .
If then by (19) and (22) we have and , but there is no a path of length three between them, since we have either
or
If then by (19) and (22) we have and , but there is no a path of length three between them, since we have either
or
Thus, a sought cycle cannot occur in this case.
Case Suppose that three vertices of a sought –cycle belong to one copy, other three vertices belong to another copy, and remaining four vertices belong to the third copy (see Figure 3). Let , , and . Then both and should belong to either or , since either or . More precisely, we have:
(23)
On the other hand, since , then by (17) and (18) there are two ways to get a path of length two from to such that either
or
and since , then we have:
(24)
To get a sought -cycle there should be a path of length three between and . Let us check this. If then by (23) and (24) we have and , but there is no a path of length three between them, since we have either
or
If then by (23) and (24) there is a contradiction with an assumption that both and belong to either or . Hence, a sought cycle cannot occur in this case.
Case 3: 10-cycle within has vertices from four copies of
There are two possible situations in this case.
Case Suppose that two vertices of a sought –cycle belong to the first copy, two vertices belong to the second copy, two vertices belong to the third copy and remaining four vertices belong to the fourth copy (see Figure 4).
Figure 4: Case 3: 10-cycle within has vertices from four copies of
Let , , and , then both and should belong to either or , since either or . More precisely, we have:
(25)
On the other hand, similar to Case there are four ways to reach by a path of length four from such that we have:
(26)
To get a sought -cycle there should be a path of length three between and . Let us check this. If then by (25) we have and , but there is no a path of length three between them, since we have either
or
If then by (25) and (26) there is a contradiction with an assumption that both and belong to either or . Hence, a sought cycle cannot occur in this case.
Figure 5: Case 3: 10-cycle within has vertices from four copies of
Case There are two subcases due to a sequence of vertices from copies forming a cycle: 1) (2,3,3,2); 2) (3,2,3,2) (see Figure 5).
Subcase 1) Suppose that two vertices of a sought –cycle belong to the first copy, three vertices belong to the second copy, three vertices belong to the third copy and remaining two vertices belong to the fourth copy (see Figure 5, Subcase 1).
Let , , and , . Similar to Case there are four ways to reach by a path of length five from such that we have:
On the other hand, since and then , and there are two ways to get a path of length one from to such that either
or
and since then we have either
or
As one can see, vertices and belong to the same copy . Let us check whether there is a path of length two between these two vertices. Indeed, there are two ways to get a path of length two from to . The first way is presented as follows:
(27)
where
The second way is presented as follows:
(28)
where
Thus, a sought cycle cannot occur in this subcase.
Subcase 2. Suppose that three vertices of a sought –cycle belong to the first copy, two vertices belong to the second copy, three vertices belong to the third copy and remaining two vertices belong to the fourth copy. Let , and , (see Figure 5, Subcase 2). There are two ways to get a path of length two from to . The first way is given as follows:
(29)
where
The second way is given as follows:
(30)
where
Since , then by (29) and (30) either or , and since either or we have:
such that with we obtain:
On the other hand, there are two ways to get a path of length three from to such that either
or
It is easy to see that vertices and belong to the copy . Let us check whether there is a path of length two between these two vertices. By (27) and (28), there are two ways to get a path of length two from to such that either
or
Hence, there is no a path of length two between these two vertices.
One can also see that vertices and belong to the copy . Let us check whether there is a path of length two between these two vertices. By (27) and (28), there are two ways to get a path of length two from to such as either
or
Hence, there is no a path of length two between these two vertices, and a sought cycle cannot occur in this subcase.
Case 4: 10-cycle within has vertices from five copies of
There is the only possible situation if each of five copies has two vertices.
Figure 6: Case 4: 10-cycle within has vertices from five copies of
Case Suppose that vertices of a sought –cycle belong to the first copy, vertices belong to the second copy, vertices belong to the third copy, vertices belong to the fourth copy and vertices belong to the fifth copy.
Let , , , and . Since can be reached from by either or , and the same we have for and , then there are four ways to get a path of length three from to (see Figure 6) such that:
and since we have:
(31)
On the other hand, there are two ways to get a path of length two from to such that either
or
and since we have:
and since either or we also have:
which gives us as follows:
(32)
To get a sought -cycle, either or should hold. Let us check this.
As one can see, there are two vertices and belonging to the same copy, and there is the only way to get a sought -cycle containing these vertices by the canonical form .
It is evident that neither nor could not be reached from neither by nor by . Thus, a sought -cycle can not occur in this case. By using similar arguments, one can check that for there is no a -cycle in the graph. For any odd , there is a contradiction with an assumption that both and belong to the same copy.
This complete the proof since all possible cases are considered. A -cycle in the graph occurs only in the case when and with the canonical form .
Theorem 5
In the cubic Pancake graphs there are no cycles of length .
Proof. To prove this theorem, we use the same arguments as we used to prove Theorem 4. Namely, we consider all possible cases for forming –cycles in the Pancake graphs with taking into account that the generating set of contains only three elements , where is odd. Due to the hierarchical structure of , cycles of length could be formed from paths of length , belonging to different copies of . Further, we consider all possible options for the distribution of vertices by copies.
Within the proof without loss of generality we always put .
Case 1: –cycle within has vertices from two copies of
Suppose that a sought –cycle is formed on vertices from two different copies of . By [5, Lemma 2], such a cycle cannot occur if its two (three) vertices belong to one copy and nine (eight) vertices belong to another one. Therefore, a sought cycle must have at least four vertices in each of the two copies. Hence, there are two following cases.
Case Suppose that four vertices of a sought –cycle belong to one copy, and other seven vertices belong to another copy. Let , , then and should belong to . Herewith, the four vertices of should form a path of length three whose endpoints should be adjacent to vertices from . On the other hand, it is obvious that there are only two ways to get path of length six from to , namely, either:
or
Since then we have either:
or
Note that with for any . Hence, we immediately can conclude that and belong to different copies of since for any odd . This gives a contradiction with an assumption that and belong to the same copy . Thus, a sought –cycle cannot occur in this case.
Case Suppose that five vertices of a sought –cycle belong to copy , and other six vertices belong to copy , where , and . Herewith, the five vertices of should form a path of length four whose endpoints should be adjacent to vertices from . By (14) and (16), there are two ways to get paths of length five from to . Moreover, since then we have either:
or
Since with for any , then we immediately can conclude that and belong to different copies of since for any odd . This gives a contradiction with an assumption that and belong to the same copy . Thus, a sought –cycle cannot occur in this case.
Case 2: 11-cycle within has vertices from three copies of
There are five different situations in this case.
Case Suppose that two vertices of a sought –cycle belong to one copy, other two vertices belong to another copy, and remaining seven vertices belong to the third copy. Let , , , then and should belong to . Using similar reasoning shown in the proof of Theorem 4, Case , one can conclude that there are four ways to reach by a path of length four from such that we have:
To get a sought -cycle there should be a path of length six between and , where . Let us check this. If then the vertices , and belong to the copy , and the following cases are possible:
or
and
or
Hence, a path of length five does not occur between and .
If the vertices and belong to the copy , and the following two cases are possible:
or
Again, a path of length five does not occur between and .
If , there is a contradiction with an assumption that and belong to the copy . Thus, a sought cycle cannot occur in this case.
Case Suppose that two vertices of a sought –cycle belong to the first copy, other three vertices belong to the second copy, and remaining six vertices belong to the third copy. Let , , , then and should belong to . Taking into account similar reasoning used in the proof of Theorem 4, Case , one can conclude that there are four ways to reach by a path of length five from such that we have:
To get a sought -cycle there should be a path of length five between and , where . Let us check this. If the vertices and belong to the copy , and the following cases are possible:
or
Hence, a path of length five does not occur between and .
If , there is a contradiction with an assumption that and belong to the copy . Thus, a sought cycle cannot occur in this case.
Case Suppose that two vertices of a sought –cycle belong to the first copy, other four vertices belong to the second copy, and remaining five vertices belong to the third copy. Let , , , then and should belong to either or , since either or , where:
(33)
By the same reasoning used in the proof of Theorem 4, Case , one can see that there are two ways to reach by a path of length five from such that we have:
(34)
To get a sought -cycle there should be a path of length four between and . Let us check this. If or , there is a contradiction with an assumption that both and belong to either or . If then by (33) and (34) we have and , but there is no a path of length four between them, since we have either
or
If then by (33) and (34) we have and , but there is no a path of length four between them, since we have either
or
Thus, a sought cycle cannot occur in this case.
Case Suppose that three vertices of a sought –cycle belong to the first copy, other three vertices belong to the second copy, and remaining five vertices belong to the third copy. Let , , , then and should belong to either or , since either or , where
(35)
Taking into account the same arguments as we used in the proof of Theorem 4, Case , one can conclude that there are two ways to reach by a path of length five from such that we have:
(36)
To get a sought -cycle there should be a path of length four between and . Let us check this. If then by (35) and (36) we have and , but there is no a path of length four between them, since we have either
or
Hence, a path of length four does not occur between and . However, let us note that in this case we have a cycle of length eight given by the canonical form obtained from (12) by putting .
If then by (35) and (36) there is a contradiction with an assumption that both and belong to either or . Hence, a sought cycle cannot occur in this case.
Case Suppose that three vertices of a sought –cycle belong to the first copy, other four vertices belong to the second copy, and remaining four vertices belong to the third copy. Let , , , then and should belong to either or , since either or , where
(37)
On the other hand, it is obvious that there are two ways to get a path of length five from to :
(38)
To get a sought -cycle there should be a path of length four between and . Let us check this. If then by (37) and (38) we have and , but there is no a path of length three between them, since we have either
or
Hence, a path of length three does not occur between and .
If then by (37) and (38) we have and , but there is no a path of length three between them, since we have either
or
Again, a path of length three does not occur between and .
If then there is a contradiction with an assumption that both and belong to either or . Hence, a sought cycle cannot occur in this case.
Case 3: 11-cycle within has vertices from four copies of
There are three possible situations in this case.
Case Suppose that two vertices of a sought –cycle belong to the first copy, two vertices belong to the second copy, two vertices belong to the third copy and remaining five vertices belong to the fourth copy. Let , , and , then and should belong to either or , since either or where
(39)
Taking into account the same arguments as we used in the proof of Theorem 4, Case , one can conclude that there are four ways to reach by a path of length four from such that we have:
(40)
To get a sought -cycle there should be a path of length four between and . Let us check this. If then by (39) we have , and , but there is no a path of length four between them, since we have either
or
If then by (39) and (40) there is a contradiction with an assumption that both and belong to either or . Hence, a sought cycle cannot occur in this case.
Case Suppose that three vertices of a sought –cycle belong to the first copy, three vertices belong to the second copy, three vertices belong to the third copy and remaining two vertices belong to the fourth copy. Let , , and . Taking into account the same arguments as we used in the proof of Theorem 4, Case , one can conclude that there are two ways to reach by a path of length three from such that we have:
(41)
On the other hand, by (29) and (30), and since , there are two ways to get paths of length three from to such that either
or
Then there are two ways to get paths of length two from to such that the first way is presented as follows:
Namely, we get either
or
The second way is presented as follows:
Namely, we get
or
Since , we get
(42)
To get a sought -cycle vertices and should be adjacent by an internal edge. However, if then by (42) and (41), we have two non-adjacent vertices and . If then by (41) and (42) there is a contradiction with an assumption that and belong to the same copy. Hence, a sought cycle cannot occur in this case.
Case There are two subcases due to a sequence of vertex from the copies forming a cycle: 1) (2,3,4,2); 2) (3,2,4,2).
Subcase 1. Suppose that four vertices of a sought –cycle belong to the first copy, two vertices belong to the second copy, three vertices belong to the third copy and remaining two vertices belong to the fourth copy. Taking into account the same arguments as we used in the proof of Theorem 4, Case , one can conclude that there are four ways to reach by a path of length five from :
On the other hand, there are two ways to reach by a path of length three from such that we have:
As one can see, vertices and belong to the same copy . Let us check whether there is a path of length three between these two vertices. Indeed, there are two ways to get a path of length three from to . The first way is presented as follows:
(43)
where
The second way is presented as follows:
(44)
where
Thus, a sought cycle cannot occur in this subcase.
Subcase 2. Suppose that three vertices of a sought –cycle belong to the first copy, two vertices belong to the second copy, four vertices belong to the third copy and remaining two vertices belong to the fourth copy. Let , and , . Taking into account the same arguments as we used in the proof of Theorem 4, Case , one can conclude that there are four ways to reach by a path of length five from :
On the other hand, there are two ways to reach by a path of length three from such that we have:
It is easy to see that vertices and belong to the copy . However, there is no a path of length three between these two vertices. Indeed, by (43) and (44), there are two ways to get a path of length three from to such that either
or
Hence, there is no a path of length three between these two vertices.
One can also see that vertices and belong to the copy . Let us check whether there is a path of length three between these two vertices. By (43) and (44), there are two ways to get a path of length three from to such as either
or
Thus, a sought cycle cannot occur in this subcase.
Case 4: 11-cycle within has vertices from five copies of
There is the only possible situation in this case.
Case Suppose that two vertices of a sought –cycle belong to the first copy, two vertices belong to the second copy, two vertices belong to the third copy, three vertices belong to the fourth copy and remaining two vertices belong to the fifth copy. Let , , , and . Taking into account the same arguments as we used in the proof of Theorem 4, Case , one can conclude that there are four ways to reach by a path of length four from :
(45)
On the other hand, there are four ways to reach by a path of length five from such that we have:
(46)
To get a sought -cycle, either or should hold. Let us check this. If , then by (46) and (45) we have:
As one can see, there are no a path of length two between and . Thus, a sought -cycle can not occur in this case. If , then by (46) and (45) we have:
Again, a sought -cycle can not occur in this case, since there is no a path of length two between and . By using similar arguments, one can check that for there is no a -cycle in the graph. For any odd , there is a contradiction with an assumption that both and belong to the same copy. This complete the proof of the last case of Theorem 5.
It is obvious that any cycle of the cubic Pancake graphs does belong to the Pancake graph , and should be described by one of the canonical formulas from Theorem 2 and Theorem 3. Let us check which cycles appear in for each
Case (1.1) Any cycle of , is formed by prefix–reversals from the set . If then by Theorem 2 the prefix–reversals and give -cycles of the form (4). For , by (5) there are no –cycles in , but there are –cycles of the form (7) when and there are –cycles of the form (12) for if we put , , . The canonical form in the last case is given by for any . Hence, the formula (1) holds.
Using similar arguments, one can see that any cycle of , is presented by prefix–reversals from the set . If then obviously has –cycles. If then by (5) there are no –cycles in , however by Theorem 3 there are –cycles of the canonical form (6). Indeed, if we put , , in (6) then we have the sequence . Thus, the formula (1) holds in this case.
The same arguments appear for the graph is even, whose generating set contains prefix–reversals , and . It has –cycles of the form (4) and –cycles of the form (13) if , but it does not have –cycles for . Moreover, for any even in there are no –cycles, but there are –cycles of the form (12) if we put , , . The canonical form of –cycles in this case is given by for any even . Thus, the formula (1) holds.
Case (1.2) In the case of the graph is odd, its generating elements , and give –cycles of the canonical form if we put , and in the form (12). Obviously, there are no –cycles in the graph since does not belong to the generating set for any . Hence, the formula (2) holds for any odd .
Case (1.3) The generating set of the graph , where is odd, gives the canonical form of –cycles if we put , and in the form (12). By Theorem 2 there are no –cycles in the graph for any . By Theorem 3 and by the characterization of -cycles in the Pancake graph [6, Theorem 4] there are no – and –cycles in for any . By Theorem 4 and Theorem 5 for any there are no –cycles and –cycles in , and the smallest cycle in is –cycle of the canonical form . This complete the proof of Theorem 1.
4 Acknowledgements
The first author is supported by the project No. FWNF-2022-0017 (the state contract of the Sobolev Institute of Mathematics). The second author was partially supported by Mathematical Center in Akademgorodok under agreement No. 075-15-2019-1613 with the Ministry of Science and Higher Education of the Russian Federation.
References
[1]D. W. Bass, I. H. Sudborough, Pancake problems with restricted prefix reversals and some corresponding Cayley networks, Journal of Parallel and Distributed Computing, 63 (2003) 327–336.
[2]P. E. C. Compeau, Girth of pancake graphs, Discrete Applied Mathematics, 159 (2011) 1641–1645.
[3]H. Dweighter, E 2569 in: Elementary problems and solutions, Amer. Math. Monthly, 82 (1975) 1010.
[4]A. Kanevsky, C. Feng, On the embedding of cycles in pancake graphs, Parallel Computing, 21 (1995) 923–936.
[5]E. V. Konstantinova, A. N. Medvedev, Cycles of length seven in the Pancake graph, Diskretnyi Analiz i Issledovanie Operatsii, 17 (2010) 46–55. (in Russian)
[6]E. V. Konstantinova, A. N. Medvedev, Cycles of length nine in the Pancake graph, Diskretnyi Analiz i Issledovanie Operatsii, 18 (2011) 33–60. (in Russian)
[7]E. Konstantinova, A. Medvedev, Small cycles in the Pancake graph, Ars Mathematica Contemporanea, 7 (2014) 237–246.
[8]E. Konstantinova, A. Medvedev, Independent even cycles in the Pancake graph and greedy Prefix-reversal Gray codes, Graphs and Combinatorics32 (2016) 1965–1978.
[9]J. J. Sheu, J. J. M. Tan, K. T. Chu, Cycle embedding in pancake interconnection networks, In.: Proc. 23rd Workshop on Combinatorial Mathematics and Computation Theory, Taiwan, 2006, 85–92.
[10]J. Sawada, A. Williams, Successor rules for flipping pancakes and burnt pancakes, Theoretical Computer Science, 609 (2016) 60–75.