Minimum spanning trees across dense cities

Abstract

Consider~\(n\) nodes distributed independently across~\(N\) cities contained with the unit square~\(S\) according to a distribution~\(f.\) Each city is modelled as an~\(r_n \times r_n\) square contained within~\(S\) and~\(MSTC_n\) denotes the length of the minimum spanning tree containing all the~\(n\) nodes. We use approximation methods to obtain variance estimates for~\(MSTC_n\) and prove that if the cities are well-connected and densely populated in a certain sense, then~\(MSTC_n\) appropriately centred and scaled converges to zero in probability. Using the proof techniques, we alternately derive corresponding results for the length~\(MST_n\) of the minimum spanning tree for the usual case when the nodes are independently distributed throughout the unit square~\(S.\) In particular, we obtain that the variance of~\(MST_n\) grows at most as a power of the logarithm of~\(n\) and use a subsequence argument to get almost sure convergence of~\(MST_n\) appropriately centred and scaled.