Caregie Mellon Uni 484 Grah Theory
Caregie Mellon Uni 484 Grah Theory
Caregie Mellon Uni 484 Grah Theory
Diestel 1.2: Let d ∈ N and V := {0, 1}d ; thus, V is the set of all 0–1 sequences of length d. The
graph on V in which two such sequences form an edge if and only if they differ in exactly one position
is called the d-dimensional cube. Determine the average degree, number of edges, diameter, girth and
circumference of this graph.
(Hint for the circumference: induction on d.)
All vertices have degree d (one adjacent vertex for each position in the sequence), so the average degree
is d.
There are 2d vertices, each of degree d, so
1X 1
|E| = deg v = 2d d = 2d−1 d
2 2
v∈V
The diameter is d: 0d (d occurrences of the coordinate 0) and 1d have distance at least d, because each
edge changes one position, and d positions must be changed. For arbitrary x, y ∈ V that differ in c ≤ d
coordinates, let x0 = x, and xi+1 be xi with the next index that differs from y corrected. xc = y, and
x0 x1 · · · xc is a path of length at most d.
For d < 2, there are no cycles (since there are fewer than 3 vertices), so the girth is ∞. For d ≥ 2, there
is a 4-cycle holding all indices beyond the first two constant at 0: 00 · · · , 01 · · · , 11 · · · , and 10 · · · .
Note that there cannot be a 3-cycle as if v is adjacent to x1 and x2 then x1 and x2 disagree on two
indicies.
For d < 2, the circumference is 0, as above. Otherwise, the circumference is 2d , by induction. The base
case is by inspection. For d ≥ 3, a d-cube is the Cartesian product of a (d − 1)-cube with K2 (that is,
two (d − 1)-cubes with corresponding vertices adjacent). By the induction hypothesis, a d − 1 cube has
a 2d−1 -cycle, say x1 x2 · · · x2d−1 . The path x1 x2 · · · x2d−1 (but not the edge x2d−1 x1 in one d − 1 cube,
followed by the edge to the other cube, followed by the reversed path x2d−1 · · · x2 x1 , and then an edge
back to the first cube is a 2d cycle. The circumference cannot be greater because there are 2d vertices.
1
HW1 21-484 Graph Theory SOLUTIONS (hbovik)
Diestel 1.8: Show that graphs of girth at least 5 and order n have a minimum degree of o(n). In other
words, show that there is a function f : N → N such that f (n)/n → 0 as n → ∞ and δ(G) ≤ f (n) for
all such graphs G.
Let v be an arbitrary vertex of G. Since there are no 3-cycles in G, the neighborhoods of the neighbors
of v do not intersect with the neighbors of v. Since there are no 4-cycles in G, the neighborhoods of the
neighbors of v are pairwise disjoint, except for containing v. This yields n = |G| ≥ 1 + δ + δ(δ − 1) > δ 2 .
√ √
So δ ≤ n, and since n is o(n), we are done.
2
HW1 21-484 Graph Theory SOLUTIONS (hbovik)
Diestel 1.9: Show that every connected graph G contains a path of length at least min{2δ(G), |G|−1}.
Let G be a connected graph, let δ = δ(G), and let P = x0 x1 · · · xk be a path of maximal length. If
P is of length at least |G| − 1, we are done. Otherwise, the set O = V (G) \ V (P ) is nonempty, and
since the graph is connected, there must be a V (P )–O path, P 0 = y0 y1 · · · ym (and this path is clearly
non-trivial). If P is of length less than 2δ, we will prove that there is a cycle spanning V (P ). Deleting
an edge of this cycle incident with y0 allows us to extend the remaining path with P 0 , forming a path
longer than P , which is a contradiction.
It remains to prove that there is a cycle spanning V (P ). Recall we denote the set of neighbors of v
as N (v). First, observe that N (x0 ) ⊂ P and N (xm ) ⊂ P , because if either end of P is adjacent to
a vertex outside of P , then the path can be extended and is not maximal. If x0 xi+1 and xm xi are
both edges in P , then there is a cycle: x0 . . . xi xm . . . xi+1 x0 . So, it suffices to show by the Pigeonhole
Principal that such a pair of edges must occur. Note first that a special case of this is where there is
an edge x0 xm , since the other edge is given in the path. The vertex x0 has at least δ − 1 neighbors
out of {x2 · · · xm−1 } (because it is adjacent to x1 and not adjacent to xm ), and for each neighbor xi
there is a corresponding vertex xi−1 to which xm is not adjacent. So, xm must have at least δ − 1
neighbors out of {x1 · · · xm−2 }, and of those δ − 1 are forbidden. Since m < 2δ, there are fewer than
2δ − 2 possible neighbors, so by the Pigeonhole Principal one of the neighbors is forbidden, so there is
a cycle, and so the path is not maximal.
3
HW1 21-484 Graph Theory SOLUTIONS (hbovik)
Diestel 1.13: Determine κ(G) and λ(G) for G = P m , C n , K n , Km,n and the d-dimensional cube
(Exercise 2); d, m, n ≥ 3.
κ(P m ) = λ(P m ) = 1.
In C n , given two non-equal and arbitrary points, there exist two internally-disjoint paths linking
those points. Informally, these correspond to going around the cycle clockwise or counter-clockwise.
Formally, if we label the cycle C n = x0 x1 x2 · · · xn x0 and have points xi and xj with i < j, one path is
xi xi+1 · · · xj (simply xi xj in the case that they are adjacent) and the other is xj xj+1 · · · xn x0 x1 · · · xi
(omitting xj+1 when j = n and omitting x1 when i = 0). Therefore, if one edge or vertex is arbitrarily
deleted, it affects at most one of the two paths, so all points remain connected. There are at least 3
vertices, so, κ(C n ) ≥ 2 and λ(C n ) ≥ 2. Each vertex is adjacent to two other vertices, so δ(C n ) = 2,
and by Proposition 1.4.2 we conclude
κ(C n ) = λ(C n ) = 2.
(Note that a shorter proof is to say that deleting any vertex or edge gives a path!)
In K n , suppose j ≤ n − 1 vertices are deleted. Any two remaining vertices are adjacent, and therefore
connected. Because there are n vertices, we conclude κ(K n ) = n − 1. This also bounds λ(K n ) ≥ n − 1
by Proposition 1.4.2. Finally, each vertex has at most n − 1 neighbors, because there are only n − 1
other vertices, so δ(K n ) ≤ n − 1, and also by Proposition 1.4.2 this implies λ(K n ) ≤ n − 1. Therefore
κ(K n ) = λ(K n ) = n − 1
In Km,n assume m ≤ n (because otherwise we look at Kn,m which is the same graph). Since vertices
in one partition have exactly n neighbors in vertices in the other have exactly m, and both partitions
are non-empty (since m and n are positive), we know δ(Km,n ) = m. By Proposition 1.4.2 this means
κ(Km,n ) ≤ λ(Km,n ) ≤ m. Next, given two vertices in Km,n , u and v, we will show that there is a
path linking u and v even after deleting up to m − 1 vertices. If u and v are adjacent, this is clearly
true, since the edge uv will remain. Otherwise, u and v are in the same partition, and have the other
partition mutually-adjacent. The other partition contains at least m vertices, so if m − 1 vertices are
deleted, at least one vertex, w, remains in the other partition. uwv is a path linking u and v after
deleting the vertices, and since there are at least m vertices in one partition, there are at least m in
the graph, so we conclude κ(Km,n ) ≥ m. Therefore
κ(Km,n ) = λ(Km,n ) = m.
It is easy to bound the connectivities of the d-dimensional cube, Qd above by its minimum degree, d
(all vertices have degree d as shown in the first question). It remains to show κ(Qd ) ≥ d so that we
can conclude (using Proposition 1.4.2) that κ(Qd ) = λ(Qd ) = d. We go by induction on d proving
κ(Qd ) ≥ d with base case d = 2 solved earlier since Q2 = C 4 . For d > 2, Qd is composed of two
Qd−1 cubes (one with first coordinate 0 and the other with first coordinate 1) which we call A and
one B. Given a set X of vertices with |X| < d, we consider two cases. In the first, X is a cut-set of,
without loss of generality, A. So by the induction hypothesis X is a subset of the vertices of A (since
4
HW1 21-484 Graph Theory SOLUTIONS (hbovik)
we need all d − 1 vertices to disconnect a Qd−1 ), so B remains connected, and each remaining vertex
of A is adjacent to a vertex in B, so the whole graph is connected. Otherwise, A and B both remain
connected after removing X. In this case the graph is only disconnected if there is no path from A to
B. But there were 2d−1 edges that cross from A to B, so one vertex from each pair is deleted. Because
2d−1 ≥ d for d ≥ 3, X cannot disconnect A from B, completing the proof that
κ(Qd ) = λ(Qd ) = d.
5
HW1 21-484 Graph Theory SOLUTIONS (hbovik)
Diestel 1.21: Show that a tree without a vertex of degree 2 has more leaves than other vertices. Can
you find a very short proof that does not use induction?
X
Let T be a tree and consider the average degree of T . By the Handshaking Lemma, we know d(v) =
v∈V
2|E|. Thus, by Proposition 1.3.2, the average degree of T is
P
d(v)
v∈V 2|E| 2(|V | − 1)
= = < 2.
|V | |V | |V |
Let L be the set of leaves in V , and O be the set of other vertices. L and O partition V , so
P
d(v)
v∈V
2>
|V |
P P
d(v) + d(v)
v∈L v∈O
=
|V |
P P
1+ 3
v∈L v∈O
≥ Leaves have degree 1; d(v) ≥ 3 for each v ∈ O
|V |
|L| + 3|O|
=
|V |
2|O|
=1+ Noting |L| + |O| = |V |
|V |
|V |
|O| < ,
2
and since
|L| = |V | − |O|,
we conclude
|V |
|L| > > |O|.
2
6
HW1 21-484 Graph Theory SOLUTIONS (hbovik)
Diestel 1.28: Show that every automorphism of a tree fixes a vertex or an edge.
Case 1: Suppose v1 is not on both the x–x0 path and the y–y 0 path. Without loss of generality, v1 is
not on the y–y 0 path. We expect dist(x, y 0 ) ≤ rad G by the definition of central, however the
x–y 0 path goes through v1 and y and thus has length at least rad G + 2: a contradiction.
Case 2: Therefore, suppose v1 is on both the x–x0 path and the y–y 0 path. We have three internally
disjoint paths: the first from x0 to y, the second from y to x through v1 , and the last from x
to y 0 . Then dist(v1 , y 0 ) < dist(y, y 0 ) and dist(v1 , x0 ) < dist(x, x0 ). Furthermore, for any vertex
w, at least one of the x–w and y–w paths must pass through v1 as it lies on the x–y path, so
for every vertex v1 has smaller distance than at least one of the proposed centers. But then
v1 is at distance less than rad G from every vertex: a contradiction.
7
HW1 21-484 Graph Theory SOLUTIONS (hbovik)
1 |V (G)|
(|V (G)|)(|V (G)|−1)
from which we conclude |E(G)| = 2 2 = 4 .
Note that at least one of |V (G)|, |V (G)| − 1 must be odd, and thus the other must be a multiple of 4
in order for |E(G)| to be integral. Thus either
or
4 |V (G)| − 1 ⇒ |V (G)| ≡ 1 mod 4.
Let f be an isomorphism from G to G. Suppose v ∈ V (G) has degree `. Then f (v) has degree 4k − `.
By the definition of graph complement, there must be u ∈ V (G) with degree 4k − `. When ` 6= 2k, the
pair of vertices has distinct degree and must represent two different vertices. As the graph has an odd
number of vertices, not every pair can be distinct vertices. Thus there must be v ∈ V (G) such that
f (v) = v, in which case we necessarily have d(v) = 2k.
8
HW1 21-484 Graph Theory SOLUTIONS (hbovik)
From question 1.18, there must be more leaves than other vertices, so there are at most 4 non-leaves.
Further, if we delete all of the leaves, the non-leaves remain and form a tree, so we can consider the
number and structure of the non-leaves.
When there is 1 non-leaf, it must be that all other vertices are leaves adjacent to this vertex:
If there are two non-leaves, they must be adjacent to each other, and each must have at least two leaves
since it cannot be degree 2. The remaining 4 leaves can be distributed in any way: half and half:
9
HW1 21-484 Graph Theory SOLUTIONS (hbovik)
When there are 3 non-leaves, they must induce a P 2 , and to assure that each has degree at least 3 the
ends must have two leaves and the middle must have 1. Two leaves remain to be distributed in any
way: all in the middle:
Finally, if there are 4 non-leaves, those form either a P 3 or a star with 3 leaves. In both cases, the
remaining vertices are forced:
10