L“APR=735 14:11:48, 24
Date: 1 APR 1975 14
Fron: CARHART
Subject: CONGEN comments; A4 problem

To: Lederberg

cc: Carhart

Unfortunately, your interest in CONGEN has taken off just when

my pushdown stack overfloweth. I (and Dennis and Sue) have seen your
note containing reactions to CONGEN, and we are in the process

of digesting and discussing. I hope to compose a reply before I
leave for Philadelphia on Sunday.

You will find that the gt40-draw is working somewhat more acurately
now, though there are still problems. During the week Peter and

I will be polishing this part of the program. By the way,

I think it might be helpful if I mentioned the method by which

the draw program (the tty draw) works ~ it may clarify my hesitance

at doing away with crossing bonds. The draw program contains a very
simple model builder (whose output you see on the GT40 screen) and a
“flattener" to spread the molecule out in the plane. The subsequent
code extracts a teletype idealization from the flattened model,

and of course crossing bonds can occur. We are already running several
seconds on moderate~sized molecules, and I feel that any

additional jockeying to get the drawing into a planar representation
will lead to excessive compute times. Our program contains the Feldmann
drawing program, and when that one works, the modelling phase is
skipped. Frequently, though, FELDRW fails but my part of the program
finds a way to “idealize" the failure. Those strangely

twisted large rings result, and sometimes spiro compounds are
unbelievably twisted. I am currently going over that section of

the draw program,and will see what I can do to detect these situations.

As for A4&, Harold is back and we have had a good talk about proups
and graphs. I think 22 nodes is about the limit - start with four
of the following units

vy
aa
C Cc~
- |
= / =
C-~C

{—

Cc

where the node in the center bears the free valence (the alternating
double bonds are your idea). now if we bond four of these together
as you were doing, we have a solution in 28 nodes. But we can save

6 nodes by merging, rather than bonding, the attachment points. this
gives an octahedron with each edge containing C=C~C, and with

four of the faces containind the fv-bearing carbons, If the
trivalent ~C= atoms in the drawing above are used for merging,

then they come out as hexavalent nodes in the final graph.

Ray