Math776: Graph Theory (I)

Fall, 2017
Homework 2 solution
1. [page 31, #20 ] Show that a graph is 2-edge-connected if and only if it has
a strongly-connected orientation, one in which every vertex can be reached
from every other vertex by a directed path.
Solution:
⇒: Let G be a 2-edge-connected graph. Since G is 2-edge-connected, there
is a cycle C in G. We can orient the edges in the same direction so that C
is strongly connected. Suppose that the edges oriented forms a subgraph
H and it is a strongly-connected orientation.
If H is not an induced subgraph of G, orient any missing edges arbitrarily.
Without loss of generality, we can assume H is an induced subgraph of G.
If H 6= G, pick a vertex x 6∈ H. Since G is 2-edge-connected, we have two
edge-disjoint paths P1 and P2 from x to some vertices in H. We can orient
the edges in P1 toward x and the edges in P2 leaving x. Then repeat this
process until H = G.

(⇐) Let H be a strongly connected orientation of G. Suppose that G has

a bridge whose endpoints are x and y. Then there is only one directed
edge between x and y, and so H is not strongly connected.
2. [page 31, #21 ] Find a short inductive proof for the existence of normal
spanning trees in finite connected graphs.
Solution: The graph G of one vertex has a normal spanning tree by
definition, as does P2 . Let any graph G of size |G| = n − 1 have a normal
spanning tree. Then consider another vertex which may be connected to
an arbitrary number of vertices. Call this vertex u. Choose one of the
vertices v ∈ N (u). Let u be the parent of v. Let all other vertices in
N (u) be directly above v in the tree. We will now reform the tree in
the following way to make it normal. Consider all vertices which are in
N (u) \ v. We will let all vertices in N (N (u) \ v) vertices become parents
of the vertices in N (u) \ v unless they are already on uP r in T . We will
continue to do this process, until there are no more vertices which are
in the neighborhoods which are not on uP r in T . This tree is normal,
because certainly for u, all the vertices in can be related to are either on
uP r or above it in the tree T . It suffices to show that all the other vertices
above u in T still have the normality condition. All of these vertices must
either be connected to something above it, in which case we still have the
normality condition. Or if they are connected to something below it, say
w is connected to something lower, x in T . Then by our construction of
T , x must be on wP r. Therefore, the normality condition holds for all the
vertices which have been moved in the tree.

1
3. [page 31, #24 ] Show that every automorphism of a tree fixes a vertex or
an edge.
Proof: Let T be a tree. If |V (T )| = 1 then clearly the identity automor-
phism must fix the vertex. If |V (G)| = 2, there are two automorphisms
- the identity and the map that switches the two vertices. In the latter
case, the edge is fixed. Now suppose for all T with |V (T )| ≤ n, every
automorphism fixes either an edge or a vertex. Let |V (T )| = n + 1 and
let {v1 , . . . , vk } ∈ V (T ) be the set of vertices with degree one. We note
that this set is nonempty as T is a tree. Let φ be an automorphism of
T . Then, φ(vi ) ∈ {v1 , . . . , vk } for 1 ≤ i ≤ k as automorphisms preserve
degrees. Consider T 0 = T \{v1 , . . . , vk }. Then φ is also an automorphism
of T 0 , T 0 is also a tree, and |V (T 0 )| < n + 1. By the induction hypothesis,
φ fixes an edge or a vertex of T 0 . Thus, φ fixes an edge or a vertex of T .
4. [page 32, #27 ] Prove or disprove that a graph is bipartite if and only if
no two adjacent vertices have the same distance from any other vertices.
Solution: ⇒: Let G be a bipartite graph, and consider two adjacent
vertices x, y ∈ G. We may assume G is connected since a graph is bipartite
if and only if each of its connected components is bipartite. We may
partition G into two subgraphs X and Y such that E(X) = E(Y ) = ∅.
Without loss of generality, assume x ∈ X and y ∈ Y . Let zx ∈ X and
zy ∈ Y . Since there are no edges within X or Y , d(x, zx ) is even, d(y, zx ) is
odd, d(x, zy ) is odd, and d(y, zy ) is even. Hence, no two adjacent vertices
share distance to any other vertex.

⇐: Suppose that G is not bipartite. By Proposition 1.6.1, G contains an

odd cycle. Let C be an odd cycle of the smallest order. Observe that C
must be geodesic in the sense:

dC (u, v) = dG (u, v)

for any vertex u and v in C. Otherwise, we can construct another odd

cycle with smaller order!
Say |C| = 2k + 1 and C has vertices v0 , v1 , . . . , v2k+1 . Let x = v0 , y =
v2k+1 , and z = vk . We have dC (x, y) = 1 and dC (x, z) = dC (y, z) = k.
Since C is geodesic, we have

dG (x, y) = 1, and dG (x, z) = dG (y, z) = k.

5. [page 32, #28 ] Find a function f : N → N such that, for all k ∈ N, every
graph of average degree at least f (k) has a bipartite subgraph of minimum
degree at least k.
Solution: Let f (k) = 4k. The graph of G with an average degree of 4k
will have a subgraph H with minimum degree 2k. Now take the maximal
bipartite graph in H with the maximal number of edges, call it HB .
Claim: HB will have minimum degree ≥ k.

2
 Assume not. Then ∃ v ∈ HB such that d(v) < k. That means that v must
have lost over half of its neighbors from H → HB . So v is on the same
’side’ as over half of its neighbors from H in HB . But by moving v to
the other ’side’ of the partition, v will be able to connect to those vertices
0
and HB will have more edges than HB →← because HB was said to be
maximal with respect to edges. So HB must have minimal degree ≥ k.
6. [page 32, #30 ] Prove or Disprove that every connected graph contains a
walk that traverses each of its edges exactly once in each direction.
Proof: Let G be an connected graph. For each edge uv of G, we replace
it by two arcs uv and vu. Call the resulting directed graph D. It is clearly
−
D is strongly connected and for each vertex v d+ H (v) = dH (v). By Euler’s
theorem, there exists an Euler circuit of H, which traverses each of its
edges in G exactly once in each direction.