MATH20902: Discrete Maths, Solutions to Problem Set 1. These solutions, as well as the corresponding problems, are available at

Transcription

1 MATH20902: Discrete Maths, Solutions to Problem Set 1 These solutions, as well as the corresponding problems, are available at (1). The upper panel in the figure below shows the maze and the network of paths within it while the lower panel shows the network of paths as a graph with named vertices. Finish Start f g h d c Finish e b n m i j k l Start a Once one has the bottom panel of the figure above, it s easy to abstract it further, to a form in which the solution is obvious: Start f e n j i m a b d c g h k l Finish Notice that there is not a unique path from the start vertex to the finish. The cycle that includes vertices g, h and n encloses a segment of the maze s walls that are, in a natural sense, disconnected from the rest. 1

2 (2). The complete graph K 4 can t be a subgraph of the complete bipartite graph K 4,4. To see why, consider the diagram of K 4,4 below and note that edges run between all possible pairs of red and white vertices, but that no edges run red-to-red or white-to-white. This observation enables us to prove our result by contradiction. Suppose for contradiction that we can find a subset V consisting of four vertices from K 4,4 and that these, along with a suitable subset of the edges, form a subgraph isomorphic to K 4. K 4,4 K 4 As V contains four vertices and the diagram of K 4,4 above organises the vertices into only two groups (the ones from the definition of bipartite: see the first lecture) the Pigeonhole Principle tells us that there must be two vertices in V that belong to the same group. But then these two cannot be adjacent in any subgraph of K 4,4, as a bipartite graph has no edges running between vertices in the same group. And this contradicts the claim that V is the vertex set of a subgraph isomorphic to K 4, for in K 4 every vertex is adjacent to every other. (3). P 4 isn t an induced subgraph of K 4,4, though C 4 is. Let s say that the vertex set of P 4 is {v 1, v 2, v 3, v 4 }, with edges (v j, v j+1 ) for 1 j 3. Now suppose aiming at a proof by contradiction that we have a bijection α mapping these four vertices to a subset of those in K 4,4. Then α(v 1 ) and α(v 3 ) would have to lie in one of the bipartite graph s two subsets (see the previous answer for an illustration of these subsets) while α(v 2 ) and α(v 4 ) would lie in the other subset. But then the subgraph of K 4,4 induced by the vertices {α(v 1 ), α(v 2 ), α(v 3 ), α(v 4 )} would include the following four edges {(α(v 1 ), α(v 2 )), (α(v 2 ), α(v 3 )), (α(v 3 ), α(v 4 )), (α(v 4 ), α(v 1 ))} while P 4 has only three edges. Thus the induced subgraph cannot be isomorphic to P 4, which contradicts our initial assumption. (4). To do these problems, it s sufficient to exhibit a cycle of the requisite length that is contained in the corresponding cube graph. (a) The vertices of I 2 are labelled by ordered pairs of 1 s and 0 s and all possible labels occur. Two vertices are adjacent if their labels differ at only a single position. The following cycle is clearly isomorphic to C 4 as, (i) it s a cycle and (ii) it contains 4 distinct vertices. In fact, I 2 is isomorphic to C 4. (00, 10, 11, 01, 00). 2

3 (b) Here a suitable cycle is (000, 100, 110, 010, 011, 111, 101, 001, 000). (c) The cycle here is long, but simply related to the one from part (b): (0000, 1000, 1100, 0100, 0110, 1110, 1010, 0010, 0011, 1011, 1111, 0111, 0101, 1101, 1001, 0001, 0000 ). (5). If you followed the hint in the question, you should have ended up with diagrams like the two below, which are alternative representations of the graphs in the original illustration. u 8 u 6 u 2 v 1 v 5 v 6 v 3 v 7 v 8 u 1 u 4 v 4 v 2 u 5 u 7 u 3 It is clear that the two graphs are not isomorphic and one way to establish this rigorously is to focus on the vertices with degree three. In the graph at left these are u 2, u 4 and u 7 and two of them, u 4 and u 7 are adjacent. In the graph at right there are also three vertices with degree three v 3, v 7 and v 8 and two of them, v 3 and v 8, are adjacent. But v 3 and v 8 are also both adjacent to v 6 and there is no corresponding shared neighbour for u 4 and u 7 on the left. One can use these observations to prove that the graphs aren t isomorphic by contradiction. Suppose there is a suitable bijection α from the vertex set {u 1,..., u 8 } of the graph on the left to the vertex set {v 1,..., v 8 } on the right. Then α would have to map the subset {u 4, u 7 } to the subset {v 3, v 8 } as these are the adjacent vertices of degree three. But if this bijection really established a graph isomorphism, the existence of the edges (v 3, v 6 ) and (v 8, v 6 ) on the right would imply the existence of the edges (α 1 (v 3 ), α 1 (v 6 )) and (α 1 (v 8 ), α 1 (v 6 )) on the left and that would imply the existence of a vertex α 1 (v 6 ) that is adjacent to both u 4 and u 7. No such vertex exists, which contradicts the assumption that α establishes a graph isomorphism and so no such α can exist. The attentive reader will note that the argument above uses, implicitly, the following proposition, that appeared in lecture: Proposition. If G 1 (V 1, E 1 ) and G 2 (V 2, E 2 ) are isomorphic and α : V 1 V 2 is the bijection that establishes the isomorphism, then deg(v) = deg(α(v)) for every v V 1 and deg(u) = deg(α 1 (u)) for every u V 2. 3

4 (6). The basic strategy here is to use the Pigeonhole Principle: we ll show that there are only (n 1) possible values of deg(v) and, as there are n vertices, some value must be shared by at least two vertices. Now, if G(V, E) is a graph with V = n vertices then we know that 0 deg(v) (n 1) for all v V. That is, a vertex can have a minimum of zero neighbours and a maximum of (n 1). This seems to suggest that there are n possible values of deg(v), which would spoil our pigeonhole argument. But a moment s thought shows that the maximal and minimal degrees can t occur in the same graph. If some vertex has degree zero, then all the others have degree at most (n 2). Alternatively, if some vertex has degree (n 1) then it must be adjacent to all the others and so they all have degree at least one. Thus, exactly one of the following three things happens: deg(v) = 0 for two or more vertices. In this case the result we seek is clearly true: the entry 0 is repeated in the degree sequence. deg(v) = 0 for exactly one vertex and 1 deg(u) (n 2) for all the others. But then there are (n 1) vertices with non-zero degree and only (n 2) possible values for their degrees, so some non-zero value must be repeated in the degree sequence. deg(v) > 0 for all v V. In this case we have 1 deg(v) (n 1) for all n vertices and so, as above, some non-zero value must be repeated in the degree sequence. (7). It is possible to construct a proof by induction along the lines of the examples in the answer to Problem??. The key idea is that I d+1 is essentially two copies of I d, glued together. The figure below shows how this works for I 2 and I 3 and suggests a way to use a cycle in I d to construct a cycle in I d I 2 I Expressing the idea in words, we construct a cycle in I 3 by splitting its vertex labels into two groups: those ending in 0 and those ending in 1. Each group has a natural, one-to-one correspondence with the vertex labels in I 2 (just ignore the final digit in the labels from I 3 ). To get the desired cycle in I 3 we traverse the first group 4

5 of vertices in the order suggested by the cycle from I 2 and then jump over to the second group and traverse that in the opposite order. Now we ll develop a recursive algorithm that generates a sequence of vertex labels for a cycle in I d+1, given one for I d, but before we can do this we need a little notation. Let s say that w j,d, where 0 j 2 d, is the label of the j-th vertex in the cycle found in I d. From the answer to the previous question, we could say that w 0,2 = 00 w 1,2 = 01 w 2,2 = 11 w 3,2 = 10 w 4,2 = 00 As we want to build up the vertex labels recursively, we ll also need a notation to indicate concatenation of letters. We ll write w j,k 1 to mean append a 1 on to the end of the string w j,k. Similarly, w j,k 0 means append a 0. Thus, for example, w 1,2 0 = 01 0 = 010 w 1,2 1 = 01 1 = 011 Then define the rest recursively by w j,d 0 If 0 j < 2 d w j,d+1 = w 2 d+1 j 1,d 1 If 2 d j < 2 d+1 w 0,d 0 If j = 2 d+1 5

6 (8). A suitable set with four elements is {1, 2, 3, 4} and we can label the vertices of the graph T 4 with subsets of the form {j, k}. The vertex set is then and the adjacency matrix is {{1, 2}, {1, 3}, {1, 4}, {2, 3}, {2, 4}, {3, 4}} {1, 2} {1, 3} {1, 4} {2, 3} {2, 4} {3, 4} {1, 2} {1, 3} {1, 4} {2, 3} {2, 4} {3, 4} Similar considerations for T 5 lead to the adjacency matrix {1, 2} {1, 3} {1, 4} {1, 5} {2, 3} {2, 4} {2, 5} {3, 4} {3, 5} {4, 5} {1, 2} {1, 3} {1, 4} {1, 5} {2, 3} {2, 4} {2, 5} {3, 4} {3, 5} {4, 5} and to the figure below, which shows T 4 on the left and T 5 on the right. {2,4} {2,4} {1,2} {2,3} {2,5} {2,3} {1,2} {1,4} {1,3} {3,5} {1,5} {1,3} {3,4} {1,4} {4,5} {3,4} (a) The number of vertices in T N is the same as the number of ways to choose two distinct elements from a set of N, which is ( ) N N! N(N 1) = =. 2 2! (N 2)! 2 6

7 (b) Suppose the set and vertex labels are like those used above and consider the vertex {j, k}. There are (N 2) elements in the original set (that is, numbers from {1,..., N}) that are different from both j and k and each such number gives rise to a pair of vertices adjacent to {j, k}. That is, for each i such that i j and i k, we have two distinct adjacent vertices: one corresponding to the subset {i, j} and another corresponding to {i, k}. Thus there are 2(N 2) = 2N 4 vertices adjacent to {j, k} or, equivalently deg({j, k}) = 2N 4. (c) If two vertices are adjacent, their corresponding two-element subsets of {1,..., N} share a member. Let s say that x corresponds to the subset {j 1, k} while y corresponds to {j 2, k}, with j 1 j 2. Now consider those elements of the original set that differ from all three of j 1, j 2 and k and define U = {1,..., N}\{j 1, j 2, k}. There are (N 3) elements in U and each gives rise to a vertex that is adjacent to both x and y: the corresponding subsets are of the form {i, k}. In addition, the vertex corresponding to {j 1, j 2 } is adjacent to both x and y, giving a total of (N 3) + 1 = N 2 common neighbours. (d) If two vertices x and y are not adjacent, then their corresponding two-element subsets have no elements in common. Let s say that x corresponds to {j 1, k 1 } while y corresponds to {j 2, k 2 }, where all four of the numbers j 1, j 2, k 1 and k 2 are distinct. The only way a vertex can be adjacent to both x and y is if its two-element subset shares a member with each of the subsets corresponding to x and y and there are clearly only four ways this can happen: {j 1, j 2 }, {j 1, k 2 }, {k 1, j 2 } and {k 1, k 2 }. 7