(Graph Theory) Is Minimum Spanning Tree Unique?

The following problem with two parts explores a condition that always implies a unique minimum spanning tree.

Part A

Suppose $G$ is a simple connected graph with finitely many vertices, and suppose that $e$ is an edge in $G$ such that removing $e$ from $G$ results in a disconnected graph. Show that e is in every spanning tree of $G$.

Proof

• Firstly, we want to show $e$ is not in any cycle. Suppose $e$ is in some cycle, then, by the important exercise, removing $e$ from $G$ will still give us a connected graph as $G$ is connected in the first place. However, this is a contradiction because removing $e$ from $G$ results in a disconnected graph. Thus, $e$ is not in any cycle.
• Then, we suppose $T$ is an arbitrary spanning tree of $G$. Notice that we always have a spanning tree because every connected graph has a spanning tree. However, since this is an assumption, it is not necessary to state the existence (see vacuous truth).
• Suppose $e$ is incident on $v,w$. Then there is a path (as $T$ is connected) between $v,w$ in $T$. If the path is $v,e,w$, then we are done. If the path does not contain $e$, then the path, along with $e$, completes a cycle containing $v,w$, which is a contradiction. Therefore, $T$ must contain $e$.

Part B

Suppose that $G$ is a simple connected weighted graph with finitely many vertices, and that if distinct edges $e$ and $e\prime$ in $G$ have the same weight, then removing either $e$ or $e\prime$ from $G$ results in a disconnected graph. Show that $G$ has a unique minimal spanning tree.

Proof

• Claim: in a connected weighted graph $G$, if the weight of each edge is distinct, then the graph $G$ has a unique minimum spanning tree.
• Proof of the claim:
  • Suppose $G$ is such a graph.
  • Suppose $T,T^{\prime }$ are two mst. Suppose an edge $e$ is in $T^{\prime }$ but not in $T$. Adding $e$ to $T$ completes a cycle. Choose from this cycle (except for $e$) the edge with the smallest possible weight and remove it. Say we choose $e^{\prime }$, and we thus obtain a new spanning tree $T^{\prime \prime }$.
  • Because the weights are all distinct, it’s either $w(e^{\prime })>w(e)$ or $w(e^{\prime })<w(e)$.
  • If $w(e^{\prime })>w(e)$, then $w(T^{\prime \prime })<w(T)$, a contradiction since $w(T)$ is already an MST, which means $e$ is also in $T$.
  • If $w(e^{\prime })<w(e)$, then
    • If in this cycle, the vertex with the largest weight is not $e$, say it’s $k$, then we can always obtain a smaller total weight by including $e$ and not including $k$ in our mst, which means we get a contradiction on the weight of $T$.
    • If in this cycle, the vertex with the largest weight is $e$. Then observe that not all edges in the cycle we completed before are in $T^{\prime }$, or $T^{\prime }$ will contain that exact same cycle. Thus, we pick any one of those edges from the cycle that is not in $T^{\prime }$, say $l$, and $w(l)<w(e)$. Adding $l$ to $T^{\prime }$ completes a cycle (not necessarily the same one as before) in $T^{\prime }$. Therefore, in $T^{\prime }$, we can get a smaller total weight by replacing $e$ with $l$. Thus, we get a contradiction on the weight of $T^{\prime }$.
• Therefore, $T,T^{\prime }$ must not contain any different edge. We successfully proved the claim that in a connected weighted graph $G$, if the weight of each edge is distinct, then the graph $G$ has a unique minimum spanning tree.
• Now we can use the claim to prove Part B.
• Suppose we remove all (groups of) edges in $G$ with the same weight. We get at least one component, and within each component, each edge must have distinct weight.
• Apply the claim to every component, we see that each of them has its unique minimum spanning tree.
• By Part A, a minimum spanning tree of $G$ must contain all the edges with the same weight. Thus, minimizing the weight of the spanning tree of $T$ is equivalent to minimizing the sum of the spanning tree of the spanning tree(s) of the component(s). Since each component has a unique mst, so does $G$.