You may want to consult:
"Denotational Semantics-A Methodology for Language Development"
by David A. Schmidt, ISBN 0-697-06849-8.
I think that the Y combinator is:
Y = lambda h.(lambda x. h(x x))
(lambda x. h(x x)) .
I think that this does what it should is pretty patent from it, but
as for deriving it, this involves insight that I don't yet have.
john