#New Minimizers?
4 messages in this thread
I view one of the strongest features of Hyperchem to be the use of
minimally modified 'public domain' (academic) minimizers. This is because
there is a wealth of information available on the strengths and weaknesses
for each minimizer.
It is in this context that I inquire about future minimizers which may be
included in Hyperchem. The obvious class of minimizers which have not yet
been addressed are ab initio calculations. I personally would be
interested in carrying out some ab initio calculations on some simple
systems, particularly as computing power improves in the next few years
(SGI workstations, 686/200 Hz or whatever they chose to call it).
As I understand it, porting ab initio packages is no trivial matter and
those currently using ab initio packages may well prefer to use existing
large computer systems. I would be interested in hearing other opinions as
to what the market might be.
My selfish interest is to see such packages offered in an accessible and
user frinedly environment such as Hyperchem. As the tools of computational
chemists loose the requirement for the user to be both a mathmatician and a
computer programmer, organic chemists will begin to use these tools by
formulating questions and problems in the language of organic chemistry.
John,
I'm not familiar with ab initio stuff. Is it all proprietary commercial
stuff, or are there decent public (academic) programs in this class as
well? And how much computing power are people throwing at these problems;
is this "two days on a Cray" type of work?
Rich
Richard,
I don't think I can really say anything about the new types of
computational methds we will have in the coming releases: partly because
stuff for the next release has to wait for formal announcments and partly
because we haven't made the decisions yet for further down the road. I
can, however, add a few things about ab initio codes. The major programs
are university-written. The biggest player is Gaussian, (John Pople's
group at Carnegie Mellon) whose Gaussian92 (current version) is probably
the most sophisticated code around for ab initio calculations. It does a
remarkable range of different types of calculation. It is cheap (but not
free) to academics and I believe less cheap to industrial users. It is
text-input based, although some people have a graphical front end to it for
a large amount of money. GAMESS (pronounced Games) comes from north Dakota
State University, and is free but not public domain, I think. Functionality
is similar to Gaussian but not so extensive, and usability is harder.
CADPAC from Cambridge University has its own strengths and weaknesses
compared to Gaussian, but I don't know its pricing (if any). The only
commercially-written ab initio program I am aware of is Spartan, from
Wavefunction Inc.. It has a subset of the capabilities of Gaussian and co,
but does have a strong graphical component to it.
Some of the ab initio stuff does require Craydays. More and more
calculations, however, are becoming as routine as semi-empirical and not
too much more expensive in terms of time (but still some). It is a bit
like molecular dynamics: you can do useful stuff in short times on small
machines, but there will always be some people who want to do massive
calculations and spend massive amounts of money.
Tom
>> but there will always be some people who want to do massive calculations
and spend massive amounts of money
Tom,
And there will always be salesmen to help them <g>.
The only folks I was familiar with were the Gaussian people, and I have to
admit that the one contact I ever had with them was less than gratifying.
I had to call them for some reason, and whoever it was a talked to (he was
one of the higher-ranking geeks, supposedly) copped a truly superior 'tude,
so I could live without 'em…
So based on that 5 minute encounter (okay, so I don't believe in reasonable
sample sizes <g>), I hope y'all smoke 'em in a future release.
Rich