next up previous contents
Next: Partition theoremcluster theorems Up: Information correlationcumulants, clusters, Previous: Summation property

 

The Ursell functions, partitions, and graphs

Let tex2html_wrap_inline13293 be the set of all containers of k variables from a set of variables X. Let tex2html_wrap_inline13299 be the set of all possible partitions of N variables from X. Let any element of tex2html_wrap_inline13299 be called a graph. Each graph in tex2html_wrap_inline13299 is then a set of containers tex2html_wrap_inline13309 with the sum of the m's being N. (In the graphical picture, the tex2html_wrap_inline13315 are called clusters.) To each tex2html_wrap_inline13315 associate a function of the variables it contains, the Ursell function tex2html_wrap_inline13319. Now, consider the graphs tex2html_wrap_inline13321. For each graph tex2html_wrap_inline13323 form the product of the functions that it represents, tex2html_wrap_inline13325, and call this the weight of the graph tex2html_wrap_inline13323. Now, sum over the graphs, and define this sum as tex2html_wrap_inline13329. Due to the fact of the partitioning used to create W, the Ursell development of section 3.5 immediately holds.

Examples of the decomposition above occur in the Mayer f-function expansion where each edge connecting two points of the graph represents a factor in a product representing the function [69], and in quantum field theory calculations using Feynman diagrams [55]. The key thing to note is that the Mayer f-function expansion is fundamentally a theorem about labeled partitions, rather than graphs. Similarly, the cumulant expansion is fundamentally a theorem about labeled partitions. In the next section we make explicit this fundamental idea, and rigorously show the connections between the counting of labeled partitions, the partition theorem, and the linked cluster theorem.



David Wolf
Tue Mar 25 08:11:49 CST 1997