• 沒有找到結果。

A note on cyclic m-cycle systems of K-r((m))

N/A
N/A
Protected

Academic year: 2021

Share "A note on cyclic m-cycle systems of K-r((m))"

Copied!
6
0
0

加載中.... (立即查看全文)

全文

(1)

Digital Object Identifier (DOI) 10.1007/s00373-006-0658-z

Graphs and

Combinatorics

©Springer-Verlag 2006

A Note on Cyclic m-cycle Systems of K

r(m)

Shung-Liang Wu1and Hung-Lin Fu2

1 National United University, Miaoli, Taiwan, R.O.C.

2Department of Applied Mathematics, National Chaio Tung University, Hsin Chu, Taiwan,

R.O.C.

Abstract. It was proved by Buratti and Del Fra that for each pair of odd integers r and m, there exists a cyclic m-cycle system of the balanced complete r-partite graph Kr(m)except for

the case when r = m = 3. In this note, we study the existence of a cyclic m-cycle system of

Kr(m)where r or m is even. Combining the work of Buratti and Del Fra, we prove that cyclic m-cycle systems of Kr(m)exist if and only if (a) Kr(m)is an even graph (b) (r, m) = (3, 3) and

(c) (r, m) ≡ (t , 2) (mod 4) where t ∈ {2, 3}.

The Main Result

An m-cycle system of a simple graph G is a set C of edge disjoint m-cycles which partition the edge set of G. The necessary conditions for the existence of an m-cycle system of a graph G are that the value of m is not exceeding the order of G, m divides the number of edges in G, and the degree of each vertex in G is even. An even graph is a graph with vertex degrees all even. If G is a complete graph Kvon v vertices,

then it is called an m-cycle system of order v.

Alspach and Gavlas [1] and ˇSajna [11] have completely settled the existence prob-lem of m-cycle systems of Kv and Kv− I, where I is a 1-factor. Moreover, there

have been many results in the literature concerning the existence problem of cyclic

m-cycle systems. The reader can refer to [2–10, 12, 13].

A graph G is said to be a complete r-partite graph (r > 1) if its vertex set V can be partitioned into r disjoint non-empty sets V1, . . . , Vr (called partite sets) such

that there exists exactly one edge between each pair of vertices from different partite sets. If|Vi| = nifor 1≤ i ≤ r, the complete r-partite graph is denoted by Kn1,···,nr. In particular, if n1= · · · = nr = k(> 1), it is called balanced and the graph will be

simply denoted by Kr(k).

Let C = (c0, . . . , cm−1) be an m-cycle. An m-cycle system of a graph G, C, is

said to be cyclic if V (G) = Zvand (c0+ 1, . . . , cm−1+ 1) ∈ C (mod v) whenever

(c0, . . . , cm−1) ∈ C.

The necessary conditions for the existence of a cyclic m-cycle system of a com-plete r-partite graph G are that G is balanced, say Kr(k), and any partite set in Kr(k)

(2)

rZk = {0, r, . . . , (k − 1)r}

of Zrkor its coset. For i = 0, . . . , r − 1, let Vi denote the ith partite set of Kr(k).

We may assume Vi = {i, i + r, i + 2r, . . . , i + (k − 1)r}. Note that the set of distinct

differences of edges in Kr(k)is Zrk\ ± {0, r, 2r, . . . , k/2 r}.

For any edge{a, b} in G with V (G) = Zv, we shall use± |a − b| to denote the

difference of the edge{a, b}. The number of distinct differences in a cycle C is called

the weight of C.

Let m = ab be a positive integer (> 2). An m-cycle C in Kr(k)with weight a has

index rk

b if for each edge{s, t} in C, the edges {s + i ·rkb, t + i ·rkb} ( mod rk) with

i ∈ Zbare also in C.

Proposition 1 ([14]). Let m = ab be a positive integer (> 2). Then there exists an

m-cycle C = (c0, . . . , cm−1) in Kr(k)with weight a and indexrkb if and only if each of

the following conditions is satisfied:

(1) for 0≤ i = j ≤ a − 1, ci ≡ cj(mod rk

b);

(2) the differences of the edges{ci, ci−1} (1 ≤ i ≤ a) are all distinct;

(3) ca= c0+ t ·rkb, where gcd (t, b) = 1; and

(4) cia+j = cj+ i · t · rkb where 0≤ j ≤ a − 1 and 0 ≤ i ≤ b − 1.

To simplify, the m-cycle C = (c0, . . . , ca−1, c0+ t ·rkb, . . . , ca−1+ t ·rkb, . . . , c0+

(b − 1) · t ·rkb, . . . , ca−1+ (b − 1) · t ·rkb) will be denoted by C = [c0, . . . , ca−1]t·rk/b.

Note that if C is any cycle with weight a in a cyclic m-cycle system of Kr(k), then

C is precisely an m-cycle with indexrkb.

The following results are either known or easy to verify, we list them without the details of proof.

Theorem 2 ([2]). For each pair of odd integers r and m, there exists a cyclic m-cycle

system of Kr(m)with the exception that (r, m) = (3, 3).

Lemma 3 ([14]). If there is a cyclic m-cycle system of a graph G, then G is 2r-regular

for some positive integer r.

Proposition 4 ([14]). If there is a cyclic m-cycle system of Kr(m)with m even and m > 4, then (r, m) ≡ (t, 2) (mod 4) where t ∈ {2, 3}.

Note that if m is odd, then r must be odd since Kr(m)is an even graph.

For an m-cycle C = (c0, . . . , cm−1), we shall use ∂C to denote the set of distinct

differences{±(ci−ci−1)|i = 1, . . . , m} where cm= c0. Given a set D = {C1, . . . , Cp}

of m-cycles, the list of differences from D is defined as the union of the multisets

∂C1, . . . , ∂Cp, i.e., ∂D =

p

i=1∂Ci.

Theorem 5 ([14]). Let D be a set of m-cycles with vertices in Zrk such that ∂D = Zrk\ ± {0, r, 2r, . . . , k/2 r}. Then there exists a cyclic m-cycle system of Kr(k).

(3)

We are now ready for the main result. First, we will assume Ci = (vi,0, vi,1, . . . ,

vi,2s, vi,2s+1, vi,2s, vi,2s−1, . . . , vi,1) to be a (4s + 2)-cycle, and an m-cycle with

weight m is called full, otherwise short.

Theorem 6. A cyclic m-cycle system of Kr(m)exists if and only if (a) Kr(m)is an even

graph (b) (r, m) = (3, 3) and (c) (r, m) ≡ (t, 2) (mod 4) where t ∈ {2, 3}.

Proof. The necessary part follows by Theorem 2 and Proposition 4. Therefore, we

prove the sufficiency in what follows. The proof is split into 4 cases: (i) (r, m) ≡ (0, 2) (mod 4) (ii) (r, m) ≡ (1, 2) (mod 4) (iii) r ≡ 0 (mod 2) and m ≡ 0 (mod 4) (iv) r ≡ 1 (mod 2) and m ≡ 0 (mod 4). Note that if m is odd, then r must be odd and this case has been settled by Buratti and Del Fra in [2].

Case 1. (r, m) ≡ (0, 2) (mod 4).

Subcase 1.1. r ≡ 0 (mod 4) and m ≡ 2 (mod 8), say r = 4p and m = 8k + 2.

Let C= [c0, . . . , c4k]r(4k+1)be a short m-cycle defined as

ci =          2rj, if i = 2j with j = 0, . . . , 2k − 1; 4r − 1 + 8j r, if i = 4j + 1 with j = 0, . . . , k − 1; 7r − 1 + 8j r, if i = 4j + 3 with j = 0, . . . , k − 2; r(8k − 1) + 1, if i = 4k − 1; and 4rk + 2, if i = 4k.

It can be checked that all values in C∗are certainly pairwise distinct and the set of differences occurring in Cis ∂C= ±{r − 2, 2r − 1, 3r − 1, . . . , (4k + 1)r − 1}.

For i = 1, . . . , p, the full m-cycles Ciare defined as

vi,0 = 0; for j = 0, . . . , 2k − 1, vi,2j +1 = jr − 3 + 4i, vi,2j +1 = vi,2j +1+ 2;

for j = 1, . . . , 2k − 1, vi,2j = r(4k + 1 − j) − 6 + 8i, vi,2j = vi,2j + 3; vi,4k =

r(2k + 1) − 5 + 8i, vi,4k = vi,4k+ 14; and vi,4k+1= 2rk − 2 + 4i.

If p ≥ 2, then for i = 1, . . . , p − 1, the remaining full m -cycles Cp+iare given

by

vp+i,0 = 0; for j = 0, . . . , 2k − 1, vp+i,2j +1 = jr − 2 + 4i, vp+i,2j +1 =

vp+i,2j +1+ 2; for j = 1, . . . , 2k − 1, vp+i,2j = r(4k + 1 − j) − 3 + 8i, vp+i,2j =

vp+i,2j + 3; vp+i,4k = r(2k + 1) − 2 + 8i, vp+i,4k = vp+i,4k+ 1; and vp+i,4k+1 =

2rk − 1 + 4i.

By routine computation, we have that all values in each full m-cycle constructed above are also pairwise distinct and 2p−1i=1 ∂Ci = ±{1, 2, . . . , r − 3, r − 1} ∪

4k−1

i=0 ±{r + 1 + ir, r + 2 + ir, . . . , 2r − 2 + ir}.

Since ∂C∗∪2p−1i=1 ∂Ci = Zrm\ ± {0, r, 2r, . . . , rm/2}, it follows from Theorem

5 that there exists a cyclic m-cycle system of Kr(m).

(4)

If k = 0, then C= [0, 4r − 1, 3r − 2]3r is the short 6-cycle. For i = 1, . . . , p,

the full 6-cycles are Ci = (0, 4i − 3, 2r − 4 + 8i, r − 2 + 4i, 2r − 3 + 8i, 4i − 1) and,

if p ≥ 2, for i = 1, . . . , p − 1, the remaining full 6-cycles are Cp+i = (0, 4i, 2r +

1+ 8i, r + 1 + 4i, 2r + 2 + 8i, 4i + 2).

We then have that ∂C= ±{2, r + 1, 2r + 1} and2p−1i=1 ∂Ci = ±{1, 3, 4, . . . ,

r − 1, r + 2, r + 3, . . . , 2r − 1, 2r + 2, 2r + 3, . . . , 3r − 1}.

If k > 0, then the short m-cycle C= [c0, . . . , c4k+2]r(4k+3)is defined as

ci =          2j r, if i = 2j with j = 0, . . . , 2k; 3r + 1 + 8j r, if i = 4j + 1 with j = 0, . . . , k − 1; 6r + 1 + 8j r, if i = 4j + 3 with j = 0, . . . , k − 1; 4r(2k + 1) − 1, if i = 4k + 1; and r(4k + 3) − 2, if i = 4k + 2, and ∂C= ±{2, r + 1, 2r + 1, . . . , (4k + 2)r + 1}. For i = 1, . . . , p, the full m-cycles Ciare defined as

vi,0= 0; for j = 0, . . . , 2k, vi,2j +1= jr −3+4i, vi,2j +1 = vi,2j +1+2, vi,2j +2=

r(4k + 2 − j ) − 4 + 8i, and vi,2j +2 = vi,2j +2+ 1; and vi,4k+3 = r(2k + 1) − 2 + 4i.

For i = 1, . . . , p − 1, the rest of full m-cycles Cp+iare given by

vp+i,0 = 0; for j = 0, . . . , 2k, vp+i,2j +1= jr + 4i, vp+i,2j +1 = vp+i,2j +1+ 2,

vp+i,2j +2= r(4k + 2 − j) + 1 + 8i, and vp+i,2j +2 = vp+i,2j +2+ 1; and vp+i,4k+3=

r(2k + 1) + 1 + 4i.

An easy verification shows that2p−1i=1 ∂Ci = ±{1, 3, 4, . . . , r−1}∪

4k+1

i=0 ±{(r+

2+ ir, r + 3 + ir, . . . , 2r − 1 + ir}.

Case 2. r ≡ 1 ( mod 4) and m ≡ 2 (mod 4), say r = 4p + 1 and m = 4k + 2.

It suffices to consider the full m- cycles.

For i = 1, . . . , p, the full m-cycles Ciare defined as

vi,0 = 0; for j = 0, . . . , k − 1, vi,2j +1 = jr − 3 + 4i, vi,2j +1 = vi,2j +1+ 2;

for j = 1, . . . , k − 1 (if k ≥ 2), vi,2j = r(2k + 1 − j) − 6 + 8i, vi,2j = vi,2j+ 3;

vi,2k = r(k + 1) − 5 + 8i, vi,2k = vi,2k+ 1; and vi,2k+1= rk − 2 + 4i.

We havepi=1∂Ci = ±{1, 3, . . . , r −2}∪

p−1

i=0

2k

j =1±{jr +1+4i, jr +2+4i}.

For i = 1, . . . , p, let Cp+ibe the rest of the full m-cycles given by

vp+i,0 = 0; for j = 0, . . . , k−1, vp+i,2j +1= jr−2+4i, vp+i,2j +1 = vp+i,2j +1+

2; for j = 1, . . . , k − 1, vp+i,2j = r(2k + 1 − j) − 3 + 8i, vp+i,2j = vp+i,2j + 3;

vp+i,2k= r(k + 1) − 2 + 8i, vp+i,2k = vp+i,2k+ 1; and vp+i,2k+1= rk − 1 + 4i.

It follows thatpi=1∂Cp+i = ±{2, 4, . . . , r −1}∪

p−1

i=0

2k

j =1±{jr +3+4i, jr +

4+ 4i}, and2pi=1∂Ci = Zrm\ ± {0, r, 2r, . . . , (2k + 1)r}.

Case 3. r ≡ 0 (mod 2) and m ≡ 0 (mod 4). Subcase 3.1. m ≡ 0 (mod 8), say m = 8k.

(5)

For i = 1, . . . , r − 1, the short m-cycles Ci= [ci,0, . . . , ci,4k−1]4rkare defined as ci,j =    2rs, if j = 2s with s = 0, . . . , 2k − 1; 2r + i + 8rs, if j = 4s + 1 with s = 0, . . . , k − 1; and 5r + i + 8rs, if j = 4s + 3 with s = 0, . . . , k − 1.

We have ∂Ci= ±{i, r + i, 2r + i, . . . , (4k − 1)r + i} andr−1i=1∂Ci = Zrm\ ±

{0, r, 2r, . . . , 4rk}.

Subcase 3.2. m ≡ 4 (mod 8), say m = 8k + 4.

For i = 1, . . . , r − 1, the short m-cycles Ci= [ci,0, . . . , ci,4k+1]r(4k+2)are given by

ci,j =    2rs, if j = 2s with s = 0, . . . , 2k; r + i + 8rs, if j = 4s + 1 with s = 0, . . . , k; and 6r + i + 8rs, if k ≥ 1 and j = 4s + 3 with s = 0, . . . , k − 1. ∂Ci= ±{r − i, r + i, 2r + i, 3r + i, . . . , (4k + 1)r + i} andr−1i=1∂Ci= Zrm\ ± {0, r, 2r, . . . , (4k + 2)r}. Case 4. (r, m) ≡ (t, 0) (mod 4), t ∈ {1, 3}. Subcase 4.1. m ≡ 4 (mod 8), say m = 8k + 4.

For i = 1, . . . , r − 1, the short m-cycles are Ci= [0, i]2r and

r−1

i=1∂Ci∗ =

±{1, 2, . . . , r − 1, r + 1, r + 2, . . . , 2r − 1}.

If k ≥ 1 then for i = 1, . . . , r −1, and j = 1, . . . , k , the remaining short m-cycles are Ci,j= [0, 4jr + i]2r and ˆCi,j= [0, (4j + 1)r + i]2r.

By routine computation, it follows thatr−1i=1kj =1∂Ci,j∗ =k−1s=0±{2r + 1 +

4sr, 2r + 2 + 4sr, . . . , 3r − 1 + 4sr, 4r + 1 + 4sr, 4r + 2 + 4sr, . . . , 5r − 1 + 4sr} and r−1 i=1 k j =1∂ ˆCi,j∗ = k−1 s=0±{3r + 1 + 4sr, 3r + 2 + 4sr, . . . , 4r − 1 + 4sr, 5r + 1 + 4sr, 5r + 2 + 4sr, . . . , 6r − 1 + 4sr}.

Subcase 4.2. m ≡ 0 (mod 8), say m = 8k.

For i = 1, . . . , r − 1, andj = 1, . . . , k, the short m -cycles are Ci,j= [0, (4j − 2)r + i]2rand ˆCi,j= [0, (4j − 1)r + i]2r.

We haver−1i=1kj =1∂Ci,j∗ =k−1s=0±{1 + 4sr, 2 + 4sr, . . . , r − 1 + 4sr, 2r + 1 +

4sr, 2r + 2 + 4sr, . . . , 3r − 1 + 4sr} andr−1i=1kj =1∂ ˆCi,j∗ =k−1s=0±{r +1+4sr, r +

2+ 4sr, . . . , 2r − 1 + 4sr, 3r + 1 + 4sr, 3r + 2 + 4sr, . . . , 4r − 1 + 4sr}.  We end this note with a conclusion. Assume m to be even (> 2) and Km− I

to be the complete graph with 1-factor I removed. Observing the consequence of Theorem 6, it is clear that if there exists a cyclic m -cycle system of Km− I, then

a cyclic m-cycle system of Krm− I is given. Unfortunately, there does not exist a

(6)

References

1. Alspach, B., Gavlas, H.: Cycle decompositions of Knand Kn– I , J . Combin. Theory

Ser. B81, 77–99 (2001)

2. Buratti, M., Del Fra, A.: Existence of cyclic k-cycle systems of the complete graph. Discrete Math.261, 113–125 (2003)

3. Buratti, M., Del Fra, A.: Cyclic Hamiltonian cycle systems of the complete graph. Dis-crete Math.279, 107–119 (2004)

4. Bryant, D., Gavlas, H., Ling, A.: Skolem-type difference sets for cycle systems. The Electronic Journal of Combinatorics10, 1–12 (2003)

5. Fu, H.L., Wu, S.L.: Cyclically decomposing the complete graph into cycles. Discrete Math.282, 267–273 (2004)

6. Kotzig, A: Decompositions of a complete graph into 4k-gons. (Russian) Mat. Fyz. Caso-pis Sloven. Akad. Vied15, 229–233 (1965)

7. Peltesohn, R.: Eine L ¨osung der beiden Heffterschen Differenzenprobleme. Compositio Math.6, 251–257 (1938)

8. Rosa, A.: On cyclic decompositions of the complete graph into (4m + 2)-gons. Mat. Fyz. Casopis Sloven. Akad. Vied16, 349–352 (1966)

9. Rosa, A.: On cyclic decompositions of the complete graph into polygons with odd num-ber of edges (Slovak). ˇCasopis P˘est. Mat.91, 53–63 (1966)

10. Rosa, A.: On decompositions of the complete graph into 4k-gons. (Russian) Mat. Fyz. Casopis Sloven. Akad. Vied17, 242–246 (1967)

11. ˇSajna, M.: Cycle decompositions III: Complete graphs and fixed length cycles. J. Combin. Des.10, 27–78 (2002)

12. Vietri, A.: Cyclic k-cycle system of order 2km + k; a solution of the last open cases. J. Combin. Des.12, 299–310 (2004)

13. Wu, S.L., Fu, H.L.: Cyclic m-cycle systems with m ≤ 32 or m = 2qwith q a prime power.

J. Combin. Des.14, 66–81 (2006)

14. Wu, S.L., Fu, H.L.: Maximum cyclic 4-cycle packings of the complete multipartite graph, (in preprint)

Received: December 16, 2004

參考文獻

相關文件

The function f (m, n) is introduced as the minimum number of lolis required in a loli field problem. We also obtained a detailed specific result of some numbers and the upper bound of

You are given the wavelength and total energy of a light pulse and asked to find the number of photons it

好了既然 Z[x] 中的 ideal 不一定是 principle ideal 那麼我們就不能學 Proposition 7.2.11 的方法得到 Z[x] 中的 irreducible element 就是 prime element 了..

volume suppressed mass: (TeV) 2 /M P ∼ 10 −4 eV → mm range can be experimentally tested for any number of extra dimensions - Light U(1) gauge bosons: no derivative couplings. =>

(a) The magnitude of the gravitational force exerted by the planet on an object of mass m at its surface is given by F = GmM / R 2 , where M is the mass of the planet and R is

At least one can show that such operators  has real eigenvalues for W 0 .   Æ OK. we  did it... For the Virasoro

incapable to extract any quantities from QCD, nor to tackle the most interesting physics, namely, the spontaneously chiral symmetry breaking and the color confinement.. 

• Formation of massive primordial stars as origin of objects in the early universe. • Supernova explosions might be visible to the most