Arie Girshson 306236399 Exercise 1: Perfect Matching Notations: 𝑜(𝐺) – Number of odd components of graph G. Theorem: G has a perfect matching if and only if: 𝑜(𝐺 − 𝑆) ≤ |𝑆| 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑆 ⊂ 𝑉(𝐺). Proof: Suppose 𝐺 has a perfect matching 𝑀. Let 𝑆 ⊂ 𝑉(𝐺), and 𝐺1 , 𝐺2 , … , 𝐺𝑛 be the odd components of 𝐺 − 𝑆. Because 𝐺𝑖 is odd, there must be some vertex 𝑢𝑖 of 𝐺𝑖 matched inside 𝑀 with some vertex 𝑣𝑖 of 𝑆 (Because G has perfect matching). 𝑮𝟏 𝒖𝟏 𝑮𝟐 𝒖𝟐 ..… 𝑮𝒏 ..… 𝒖𝒏 𝑬𝒗𝒆𝒏 𝒄𝒐𝒎𝒑𝒐𝒏𝒆𝒏𝒕𝒔 𝒐𝒇 𝑮 − 𝑺 𝑶𝒅𝒅 𝒄𝒐𝒎𝒑𝒐𝒏𝒆𝒏𝒕𝒔 𝒐𝒇 𝑮 − 𝑺 𝑮 𝒗𝟏 𝒗𝟐 Since {𝑣1 , 𝑣2 , … , 𝑣𝑛 } ⊆ 𝑆: ..… 𝒗𝒏 ..… 𝑺 𝑜(𝐺 − 𝑆) = 𝑛 = |{𝑣1 , 𝑣2 , … , 𝑣𝑛 }| ≤ |𝑆|. Conversely, suppose 𝐺 satisfies the schematic above, but has no perfect matching. In this case, 𝐺 is a spanning sub-graph of a maximal graph 𝐺 ∗ having no perfect matching. Since 𝐺 − 𝑆 is a spanning sub-graph of 𝐺 ∗ − 𝑆 (as denoted above) we get : 𝑜(𝐺 ∗ − 𝑆) ≤ 𝑜(𝐺 − 𝑆) and 𝑜(𝐺 ∗ − 𝑆) ≤ |𝑆| 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑆 ⊂ 𝑉(𝐺 ∗ ). Reminder: Setting 𝑆 = 0 → 𝑜(𝐺 ∗ ) = 0 → 𝑣(𝐺 ∗ ) is even. Arie Girshson 306236399 Denote by 𝑈 the set of vertices of degree 𝑣 − 1 in 𝐺 ∗ . If 𝑈 = 𝑉, 𝐺 ∗ has a perfect matching, therefore, assume 𝑈 ≠ 𝑉 and show that 𝐺 ∗ − 𝑈 is a disjoint union of complete graphs. Suppose, to the contrary, that some component of 𝐺 ∗ − 𝑈 is not complete, meaning in this component there are vertices 𝑥, 𝑦, 𝑧 that 𝑥𝑦 ∈ 𝐸(𝐺 ∗ ), 𝑦𝑧 ∈ 𝐸(𝐺 ∗ ), 𝑥𝑧 ∉ 𝐸(𝐺 ∗ ). Moreover, since 𝑦 ∉ 𝑈, there is a vertex 𝑤 in 𝐺 ∗ − 𝑈 such that 𝑦𝑤 ∉ 𝐸(𝐺 ∗ ). 𝒚 𝒘 𝒙 𝒛 Since 𝐺 ∗ is a maximal graph (but not perfect matching), 𝐺 ∗ + 𝑒 has a perfect matching for all 𝑒 ∉ 𝐸(𝐺 ∗ ). Let 𝑀1 and 𝑀2 be perfect matching each, in 𝐺 ∗ + 𝑥𝑧 and 𝐺 ∗ + 𝑦𝑤, respectively. Denote by 𝐻 the sub-graph of 𝐺 ∗ ∪ {𝑥𝑧, 𝑦𝑤} induced by 𝑀1 ∆𝑀2 . Since each vertex of 𝐻 has degree two, 𝐻 is a disjoint union of even cycles, since edges of 𝑀1 alternate with edges of 𝑀2 around them. Distinguish two cases: Case 1: 𝑥𝑧 and 𝑦𝑤 are in different components of 𝐻. Then if 𝑦𝑤 is in cycle 𝐶 of 𝐻, the edges of 𝑀1 in 𝐶 with the edges of 𝑀2 not in 𝐶 constitute a perfect matching in 𝐺 ∗ , contradicting the definition of 𝐺 ∗ . Case 1 Illustration: 𝒚 𝒙 𝒘 𝒛 𝑀1 − 𝑅𝑒𝑑 𝐹𝑢𝑙𝑙 𝐿𝑖𝑛𝑒 𝑀2 − 𝐺𝑟𝑒𝑒𝑛 𝐷𝑎𝑠ℎ𝑒𝑑 𝐿𝑖𝑛𝑒 Arie Girshson 306236399 Case 2: 𝑥𝑧 and 𝑦𝑤 are in the same component 𝐶 of 𝐻. By symmetry of 𝑥 and 𝑧, assume that the vertices 𝑥, 𝑦, 𝑤 and 𝑧 occur in that order on 𝐶. Then the edges of 𝑀1 in the section 𝑦𝑤 … 𝑧 of 𝐶, together with the edge 𝑦𝑧 and the edges of 𝑀2 not in the section 𝑦𝑤 … 𝑧 of 𝐶 constitute a perfect matching in 𝐺 ∗ , and again contradicting the definition of 𝐺 ∗ . Case 2 Illustration: 𝒘 𝒚 𝑀1 − 𝑅𝑒𝑑 𝐹𝑢𝑙𝑙 𝐿𝑖𝑛𝑒 𝑀2 − 𝐺𝑟𝑒𝑒𝑛 𝐷𝑎𝑠ℎ𝑒𝑑 𝐿𝑖𝑛𝑒 𝒛 𝒙 According to the contradictions of cases 1 and 2, it follows that 𝐺 ∗ − 𝑈 is a disjoint union of complete graphs, and according to the following schematic (already shown above): 𝒚 𝒙 𝒘 𝒛 𝑜(𝐺 ∗ − 𝑈) ≤ 𝑈. Therefore, at most |𝑈| of the components of 𝐺 ∗ − 𝑈 are odd, then 𝐺 ∗ has a perfect matching – one vertex in each odd component of 𝐺 ∗ − 𝑈 is matched with a vertex of 𝑈, while the remaining vertices in 𝑈, and in components of 𝐺 ∗ − 𝑈, are matched as follows: .. …𝑮∗ − 𝑼 𝑶𝒅𝒅 𝒄𝒐𝒎𝒑𝒐𝒏𝒆𝒏𝒕𝒔 𝒐𝒇 .. … 𝒐𝒇 𝑮∗ − 𝑼 𝑬𝒗𝒆𝒏 𝒄𝒐𝒎𝒑𝒐𝒏𝒆𝒏𝒕𝒔 𝑮 ….. … 𝑼 Arie Girshson 306236399 Since 𝐺 ∗ was assumed to have no perfect matching, the desired contradiction is obtained. Therefore, 𝐺 have perfect matching.