Time to try a different approach. Starting with the old definition of 
,
and integrating by parts in x,
The first term is of course zero because both 
and
go to zero as 
.
So
Recalling equation 3.3,
Thus we can reduce this cumbersome notation to something a little simpler for those of us with overlap integrals all coded up already as a simple subroutine...
This I got to work. It is somewhat more appealing, since T
should be
a symmetric matrix, i.e. T
= T
. This is an obvious truth when
equation 3.13 is used to calculate T, but is not so from the
asymmetric form. There is no good reason for this -- Wesley Allen and I
proved all this quite rigorously, so there must be something wrong with my
coding of it.
