Given an un-directed graph G=⟨V,E⟩G = \langle V, E \rangle without self loops and duplicates,

Connectivity. The connectivity of an edge e=(u,v)e = (u, v), denoted as λe\lambda_e, is the minimum number of edges to remove to disconnect uu from vv.

Strength. The strength of an edge e=(u,v)e = (u, v), denoted as ses_e, is the maximum positive integer kk, s.t., it belongs to a kk-connected vertex induced sub-graph. A sub-graph is called kk-connected if its min-cut is at least kk.

Theorem
  1. ∑e∈E1/se≤n−1\sum_{e \in E} 1 / s_e \le n - 1, where n=∣V∣n = |V|.
  2. ∑e∈E1/λe≤n−1\sum_{e \in E} 1 / \lambda_e \le n - 1

Proof. We prove claim 1 by induction on the number of vertices nn. When n=2n = 2, ∣E∣=1|E| = 1, the strength of the only edge is 11. Therefore ∑e∈E1/se=1/1=1\sum_{e \in E} 1 / s_e = 1 / 1 = 1. Suppose claim 1 holds for all graphs with n−1n - 1 vertices. For a graph with nn vertices, consider a particular min-cut (S,Sˉ)∈V×V(S, \bar{S}) \in V \times V. If an edge e=(u,v)e = (u, v) is in the min-cut, then its strength is at most λ=G(S,Sˉ)\lambda = G(S, \bar{S}) (the min-cut value), since we need to remove at most λ\lambda edges to disconnect uu from vv in arbitrary vertex induced sub-graph that contains ee. But the original graph GG contains ee and has connectivity λ\lambda. It follows that the strength of ee is λ\lambda and

∑e∈E1se=∑e∈GS1se+∑e∈GSˉ1se+∑e∈(S,Sˉ)1se≤∣S∣−1+∣Sˉ∣−1+λ1λ≤n−1 \begin{aligned} \sum_{e \in E} \frac{1}{s_e} &= \sum_{e \in G_S} \frac{1}{s_e} + \sum_{e \in G_{\bar{S} } } \frac{1}{s_e} + \sum_{e \in (S, \bar{S}) } \frac{1}{s_e} \\ &\le |S| - 1 + |\bar{S}| - 1 + \lambda \frac{1}{ \lambda } \\ &\le n - 1 \end{aligned}

Finally, similar argument applies to claim 2.