I will use {a, b, c, d}, {x, y, z, w} as the two parts of K4,4. Since each point has degree 4 and each 2-factor of the factorization uses 2 edges, there must be two 2-factors in any 2-factorization of K4,4.
First remove the 2-factor (cycle) axbyczdwa now consider the graph that remains is the 2-factor (cycle) aydxcwbza. These two cycles form a 2-Factorization (in fact a Hamiltonian factorization) of K4,4.
Note that there is another answer in which each 2-factor of the 2-factorization consists of two 4 cycles.
This cannot have a 2-factorization as there arepoints of odd degree.
I will use {x, 0, 1, 2, 3, 4, 5, 6, 7} for the point set of K9. A method for finding 2-factorizations was given in the notes. x is the 'infinite' point, so x + 1 = x and x + 2 = x, all other arithmetic is mod 8. Start with the 2-cycle x01726354, and add 1 (mod 8) and 2 (mod 8). Thus the full 2-factorization is:
x01726354, x12037465, x23140576
Note that this was done "by inspection".
The two triangles of the bowtie each form a C3.
Note that this question asks for a decomposition, not a factor or factorization. This essentially means that we only need to C3's which decompose the edge set and they need not be disjoint. This graph does not have a factorization into C3's as that would require sets of disjoint C3's which cover the point set. Indeed this graph does not even have a 2-factor.
Decomposition into a C6 and a C4:
C6: abcfeda, C4: bdceb
Alternately, decomposition into a C4 and two C3:
C4: dcdeb, C3: abda, C3: fcef.
If we try to decompose into C3s, the points a and f must appear in the triples abda and fcef respectively. The remaining edges form a 4 cycle, and contain no more C3s.
First remove the 1-factor {000, 100}, {001, 101}, {111, 011}, {110, 010}. The remaining edges are now a 2-factor (two disjoint C4s).
This is the original Kirkman's Schoolgirl Problem:
Fifteen young ladies in a school walk out three abreast for seven days in
succession:
it is required to arrange them daily so that no two shall walk
twice abreast.
Labeling the schoolgirls a - o the following is a solution:
Monday | Tuesday | Wednesday | Thursday | Friday | Saturday | Sunday |
abc djn ehm fio gkl | ahi beg cmn dko fjl | ajk bmo cef dhl gin | ade bln cij fkm gho | afg bhj clo dim ekn | alm bik cdg ejo fhn | ano bdf chk eil gjm |
Note that there are 80 non-isomorphic solutions. And of course relabeling will produce a different looking solutions.
χ(G) = 3. χ'(G) = 4.
Note that this graph contains a K3, so χ(G) is at least 3. It is not hard to find a 3 colouring of the vertices.
Since the graph contains a vertex of degree 4, χ'(G) is at least 4. It is not hard to find a 4 colouring of the edges.
χ(G) = 3. χ'(G) = 4.
Note that this graph contains a K4, so χ(G) is at least 4. Here are the colour classes of a 4-colouring":
C1 = {b, f},
C2 = {a, c},
C3 = {e},
C4 = {d}.
Δ = 4, so χ' is at least 4. Here are the colour classes of a 4-edge colouring":
C1 = { bc, ad, cf },
C2 = { be, cd },
C3 = { bd, ce},
C4 = { ab, cf, de }.
Note that this graph contains a K4, so χ(G) is at least 4. In fact χ(G) = 4, here is a 4-colouring Ci is the coulour class i.
C1 = {e, a, i},
C2 = {b, f},
C3 = {c, g},
C4 = {d, h},
Note that the maximum degree Δ = 4, so χ'(G) is either 4 or 5 by Vizing's Theorem. In fact χ'(G) = 4, here is a 4-colouring Ci is the coulour class i.
C1 = { ef, gh, dc, ab},
C2 = { eg, fi, ac, bd},
C3 = { ec, ad, fg, ih},
C4 = { ed, bc, fh, gi},
This graph is bipartite, so χ'(G) = 2:
C1 = {000, 011, 101, 110},
C2 = {001, 010, 100, 111},
Note that this is exactly the partition of Lab 1 Question 1.
Since Q3 is 3-regular, Δ = 3, and so χ'(G) is either 3 or 4 by Vizing's Theorem. In fact it has a 1-factorization, this is a 3-colouring of the edges.
C1 = { {000 001}, {100, 101}, {110, 111}, 010, 011} }
C2 = { {000, 010}, {100, 110}, {101, 111}, {001, 011} },
C3 = { {000, 100}, { 001, 101}, {010, 110}, {111, 011} }.
Suppose not, that is suppose that there is a minimum colouring
such that every vertex in the class Ci
is not adjacent to a vertex of some other colour.
But if a vertex u in Ci is not adjacent to a vertex of
some colour j, then u can be re-coloured j.
Proceeding in this way for every vertex in Ci
will empty Ci and produce a smaller colouring.
This contradicts the assumption that the colouring was minimum already.
(⇒) Necessity was done in the lectures.
(⇐) Let G be a k-regular bipartite graph, and r, s are an integers
such that k = rs.
Since G is k-regular bipartite it has a matching M1 (Corollary of Hall's Theorem).
Now since each vertex of G appears with degree 1 in M1,
G - M1 is a k-1 regular graph.
Further deletion of edges cannot violate the bipartite property,
so G - M1 is bipartite.
So G - M1 is a k-1 regular bipartite graph and so has a matching
M2.
We proceed in this manner until we have r matchings
M1, . . ., Mr.
Now F1 = M1 ∪ . . . ∪ Mr
is an r-factor of G.
We have shown that any k regular bipartite graph G with k ≥ r has
an r-factor.
Removing the edges of this r-factor results in a new graph G1
which is k-r regular bipartite and so has a r-factor.
Continue removing s r-factors as above, k - sr = 0.
Peter Danziger, April 2008