The following was posted by Stuart Margolis on my Facebook. I hope he doesn't mind me including it here lightly edited.
Ron Graham wrote a few papers in finite semigroups in the late 1960s that were known only to the Rhodes school of semigroups for many years. They have been rediscovered over the last few years and are as fresh and important today as they were half a century ago.
The paper:
"On Finite O-Simple Semigroups and Graph Theory", R. Graham, MATHEMATICAL SYSTEMS THEORY, Vol, 2, NO. 4, 325-339, 1968,
was the first paper to explicitly look at finite 0-simple semigroups as bipartite group labeled graphs (also called gain graphs, voltage graphs and other names). Among many results, it has the beautiful theorem that classifies finite 0-simple semigroups whose idempotent generated subsemigroup has only trivial subgroups and idempotent generated subsemigroups of finite 0-simple semigroups in general. By a result of Des FitzGerald this work can be extended to study idempotent generated subsemigroups for all finite semigroups.
The results were rediscovered later and given a more topological flavor by C.H. Houghton in the early 1970s. The so called Graham-Houghton graph of a 0-simple semigroup has been a tool of great import in a burgeoning literature on idempotent generated semigroups that has appeared over the last years.
A treatment of this work appears in in section 4.13 of the Rhodes-Steinberg book, "The Q-Theory of Finite Semigroups".
The paper:
Maximal Subsemigroups of Finite Semigroups* N. GRAHAM, R. GRAHAM, AND J. RHODES, JOURNAL OF COMBINATORIAL THEORY 4, 203-209 (1968) does just what its title says- describes maximal subsemigroups of finite semigroups.
The paper remained largely unknown for many years and gets rediscovered every so often. In the last few years the paper "Chains of subsemigroups" by Cameron, Gadouleau, Mitchell and Peresse uses these results to study the longest chain of subsemigroups of a finite semigroup.