From:	jkl "-at-" ccl.net
Date:	Fri, 13 Mar 92 19:50:37 EST
Subject:	Summary of BIG AB INITIO

And this is a summary of responses on BIG AB INITIO.
Sorry for deletions, but it is big even without them.
Jan
jkl $#at#$ ccl.net
==================================================

My original(?) question:
Subject: The future of ab initio
Date: Tue, 03 Mar 92 14:50:56 EST
From: jkl %! at !% ccl.net

Dear netters,

We, in the Ohio Supercomputer Center, are trying to predict the trends
and needs in computational chemistry of the (near) future.
And as usual, there are more questions than answers.

I have a big favo(u)r to ask. Please, answer me the following question:

What would you do if you could ROUTINELY perform ab initio calculations
(let say at the level of HF or HF/MP2) with 500 contracted basis functions?
What kind of projects and applications would you run? Can you think about
any commercial R & D applications where ab initio methods are important
at this level.

Thank you very much in advance. Mail it to me (jkl-: at :-ccl.net), and I will
summarize the responses.

Jan Labanowski

====================================
>From GERSON ( ( at ) ) dfwvm04.vnet.ibm.com Tue Mar  3 16:43:35 1992
Date: Tue, 3 Mar 92 15:41:41 CST
From: "Dr. Dennis Gerson" 
Subject: Abinitio Futures

At IBM (even those of us in technical marketing support do applied research
in joint projects with IBM Research Division) we are looking at computational
tools to simulate "real world" behaviour of polymers under processing
conditions.  The polymers are used as adhesives, photoresits or media sub-
strates (like CDs, floppys or tape).  As such we need to calculate properties
such as thermal expansion coefficient, 2nd and 3rd order polarizabilities,
UV spectra, fugacity.  Most of our calculations are done at the MP2 level
using HF hamiltonians except when we are looking at excited states.
Most UV spectra are calculated using CNDO/S as a first approximation.
Most of the systems are 80-150 heavy atoms and require a 6-31G*basis for
properties and a 3-21G* for geometry.  Our dynamics and mechanics are done
using parameters developed from abinitio into the CHARMm and Discover force
fields.

I hope this helps....Regards, Dennis Gerson

====================================
>From cushing[ AT ]smoked.cse.ogi.edu Tue Mar  3 20:21:03 1992
From: Judy Bayard Cushing 
Subject: Re:  The future of ab initio

...as a computer scientist just beginning to study ab initio codes,
and as a database specialist, i would guess that anyone who routinely
performed lots of ab initio calculatins of the level you specified
would want to interface with a database system to hold his/her results
(at a minimum) and to peruse other's previous (successful) runs....
(--deleted--)
--judy
cushing
oregon graduate institute.
====================================
>From d3e353 \\at// minerva.pnl.gov Tue Mar  3 20:26:53 1992
Date: Tue, 03 Mar 92 17:23:17 -0800
From: d3e353 ( ( at ) ) minerva.pnl.gov
Subject: Re: The future of ab initio

Jan,

We are rather close to putting in place some of the large scale
computing capabilities that you describe, here at Pacific Northwest
Laboratory.  Massively parallel versions of direct SCF and direct MP2
are being finished up and tested by members of our Molecular Science
Software Group and their collaborators.  Our parallel machine is the
512 processor Intel Touchstone Delta Computer located at Caltech (and
the Intel iPSC/860 at PNL), but these codes can run on other
distributed memory or shared memory machines with few modifications (if
any).  Our present code development work specifically targets the
500-1000 basis function range on massively parallel machines.

At PNL, we need to be able to do calculations of this size (and larger)
to model molecular processes in the environment, dealing with
environmental restoration and waste remediation.  Most of this
chemistry is in solution phase, and  requires an array of computational
methods to study it.  Quantum chemistry is just one of the required
approaches.  For example, we also pursue large scale molecular dynamics
computations for enzyme redesign, and have developed a new massively
parallel code for large biopolymer simulations in the 20,000-100,000
atom range, with classical force fields.  Another prototype nearing
completion is a massively parallel Local Density Functional method for
periodic systems.

Ray Bair
Group Leader, Molecular Science Software
Molecular Science Research Center
Pacific Northwest Laboratory
d3e353 -AatT- pnlg.pnl.gov

=======================================
Date: Tue, 3 Mar 92 18:38:13 EST
From: m10!frisch \\at// uunet.UU.NET (Michael Frisch)
Subject: Re: The future of ab initio

(--deleted--)
HF/6-31G* on C30H62 with no symmetry (574 basis functions) takes well under
a day in Gaussian 90 on an inexpensive workstation (RS/6000 model 530).
Shouldn't that question be "What applications do you do now that such
calculations have been possible for a couple of years?"

Michael Frisch
Gaussian, Inc.

==========================================
Date: Tue, 3 Mar 92 22:31:36 -0500
From: neal #*at*# vision.poly-eng.uakron.edu (Neal Neuburger)

(--deleted--)
Perhaps, determining the spectral properties of UV absorbers in/for
polymer additives. Or, determining the spectral properties,  of organic
pigments.

Neal Neuburger


=========================================
Date: Tue, 3 Mar 1992 22:41 EST
From: KANTERS-0at0-urvax.urich.edu
Subject: Re: The future of ab initio

Hello Jan,

I just saw you message regarding the future of ab inition calculations. I am
very new in all this. My background is mainly in Inorganic Chemistry,
more specifically gold-phosphine clusters. I did some EHMO calculations on
these type of compounds to rationalize structural changes upon chemical mo-
dification. For a lot of this kind of things I find extended Huckel satis-
factorily.
At the moment I am collaborating with another Inorganic Chemist who works on
Re-di-imine complexes. For this research we are interested in more quantita-
tive calculations in order to be able to predict what modifications in the
ligand system will give us the desired UV/VIS absorbtions. With that in mind
we want to run Argus (by Mark Thompson, when he releases it) or ZINDO on
the computing facilities we have here at Richmond.(--deleted--)

Rene Kanters
Chemistry Department
University of Richmond
KANTERS (- at -) URVAX.URICH.EDU

===============================================
Date:         Tue, 03 Mar 92 22:51:24 EST
From: Rami Osman 
Subject:      Re: The future of ab initio

Hi Jan,
  There are really two parts to the answer, but this is perhaps because of the
ambiguous definition of ROUTINELY. If Routinely means just being able to run
such a job and the results will arrive in a few days (or a few weeks?) it would
be of limited use to me. I would be interested in calculating a good quality
potential energy surface of a large molecule, say an isomerization of a tripep-
tide, within a reasonable time limit for example a few hours. Such potential
reactions, which is one of my interests.
  I guess at the end of the survey you WILL tell us where we can get such a
          COMPUTER.
                   Regards, Rami Osman

==============================================
Date: Tue, 3 Mar 92 23:17:52 -0500
From: fredvc()at()esvax.dnet.dupont.com
Subject: RE: the future of ab initio

(--deleted quotes of jkl-: at :-ccl.net and m10!frisch-: at :-uunet.UU.NET--)
We have certainly come a long way from the time when calculation of the
rotational barrier in ethane was worthy of publication as a JACS Communication.
Dr. Frisch certainly provides a good benchmark for where we are.  However,
one would really like to complete such a calculation in 30 minutes or less!!
Such a time scale, within a factor of 2-3, is what I believe to be the intended
frame of reference for the question.

FREDERIC A. VAN-CATLEDGE
fredvc ( ( at ) ) esvax.dnet.dupont.com
(--deleted--)
==============================================
>From pvrs()at()organik.uni-erlangen.de Wed Mar  4 06:49:57 1992
From: Paul Schleyer 
Subject: Re: The future of ab initio
Date: Wed, 4 Mar 92 12:43:58 MET

(--deleted jkl "-at-" ccl.net quote--)
Dear Dr. Labanowski,
Such capability would open up a very large area of chemistry, particulalarly
with regard to the combination of computations on middle large molecules with
experiment. HF and MP2 levels are very good for structures and the computation
of many properties (e.g., vibrational and nmr spectra) which would often be
decisive in helping the experimentalist.
    This applies both to pure and more applied research areas. I do not
think in terms of "commercial R & D applications" in my ivory tower.
    But do tell me when it will be possible to carry out such calculatins!
What is underway?
                                         Sincerely, Paul Schleyer
(--deleted--)

=============================================
Date: 04 Mar 92  10:47:02 EDT
From: 
Subject: Large scale computations

E. M. EVLETH
Dynamique des Interactions Moleculaires
Universite Pierre et Marie Curie
4 Place Jussieu, Tour 22, Paris 75005
(1) 44 27 42 08
UDIM018 at FRORS31

Although Dr. Labanowski requested reponse his large scale computation
question to be sent to his own address, Mike Frisch's comment on
using G90 on this size permitted me to respond with a question.

First, a lot of people will be doing large scale calculations of
the 500 CGTO size in the future as well as presently, and routinely.
We are working on zeolite substructures and reactions occurring
on these models, and that gets big.  Certain transition states
are correlation sensitive and given the "ifyness" of using a
model system our protocols for the present will not go beyond
SCF DZP optimizations followed by a MP2 estimate. Currently
distributed codes like Gaussian90 can handle this kind of
calculation in direct.  Note that it is not so much the single
point calculation time which is en jeux but the geometry optimization
times.  Even for relatively small systems these can go on for
days or weeks in work station environments. It would also be
useful to have work station clusters capable of speeding up that
operation.

Now the question and remark.  The supercomputer environment is one
configuration but people are building up clusters of work stations.
In some types of problems (molecular dynamics, quantum Monte Carlo)
some people are running 20 or more RS/6000 on the same problem.
Many labs or Universities have unused units in the night, weekend
or vacation periods and those who have work to do groan when
they see a machine inactive.

Where are people going on this type of clustering with regard to
"standard" quantum chemistry codes, especially in direct SCF or MP2?

This is of current interest in France and plans are to go "cluster"
for the next few years.  Any information from the other side
of the Atlantic would be appreciated.

E. M. Evleth

================================
Date: Wed, 4 Mar 1992 05:55 CST
From: Andy Holder 
Subject: A gadfly forever.....

I haven't run this out yet, but I'm sure that AMPAC using the
  highly successful AM1 method could perform this calculation
  in a much shorter time than an ab initio calculation with
  reasonable results.  I think the discussion should refocus
  on "What can be done now that couldn't be done before?"  Semi-
  empirical methods can now do calculations on reasonably sized
  polymer chains or biomolecules efficiently.  Oh, well.


  Andy Holder
(--deleted--)

===================================
From: Dale Moberg 
Message-Id: <9203041441.AA00301 : at : axon.cis.ohio-state.edu>
Subject: comp chem tutorials in Columbus?

(--deleted--)
A molecular biologist and myself have considerable interest
in understanding comp chem potential for biopolymers,
especially surface contour potential maps along stretches
of DNA, RNA and polypeptides.
(--deleted--)

Dale Moberg

======================================
Date: Wed, 4 Mar 92 09:59:05 -0500
From: philk &$at$& hermann.polymer.uakron.edu (Phil Klunzinger)
Subject: ab initio wish list


   I am currently using semi-empirical calculations for modulus and normal mode
analysis on a series of polymers.  I would like to replicate a few of these
calculations using ab-initio calculations, if only 1) i knew more about them,
2) they were readily aviable, 3) i could do them so quickly my thesis would not
be delayed (unless of course the data is so good they make my thesis).

                                                Philip E. Klunzinger
(--deleted--)

=====================================
Date:    Wed, 04 Mar 92 17:24 FWT
From: "Roland Wiest . . . 88-41-61-28"     
         (Lab. Chimie Quantique,  BP 296,  F-67008 Strasbourg Cedex)
Subject: Re: The future of ab initio


    Dear Jan,

    We CAN, already, ROUTINELY do calculations with 500 contracted GTO's
Our ASTERIX set of programs (see: i) Ernenwein,R. et al, Comput. Phys.
Comm. 58(1990)305 ; ii) Rohmer,M-M.et al, id,60(1990)127 ;  iii) Wiest, R.
et al, id, 62(1991)107 )  allows such a calculation like (V 10  O 28) anion
with 1404 GTO's and 578 CGTO's. (see Rohmer et al, IJQC 40(1991)723 )
   Other calculations,  most of them in the organometallic chemistry area
have been done or are in progress, in relation with problems encountered
by chemists in either homogeneous catalysis or crystallography (electron
deformation density or molecular electrostatic potential), photochemistry,
PE spectroscopy, excited states, reactivity, etc, etc..
   For more information, see recent papers in SCI or Chem Abs, under author
names like Benard,M., Rohmer,M.M., Dedieu, A., Veillard, A., Daniel, C. etc.
  All this is more or less  fondamental research, with helas, no direct
interaction with R. & D. industrial departments.

     The real question, in vue of problems arising from experimental chemistry,
is " How to do easily, and rapidly, calculations involving 1000 or 1500 or 2000
CGTO's, in order to give a realistic description in organometallic or other
"big" systems like polypeptides, aminoacids etc."

I hope this parcellar answer will help you to appreciate the magnitude of the
problems treated
   Friendly
   Roland

===========================================
Date: Wed, 4 Mar 1992 13:25:11 -0500
From: zheng' at \`retina.chem.psu.EDU (Ya-Jun Zheng)

 To those who are interested in the performance of molecular mechanics on
hydrogen bonding interactions. The following reference may be very interesting.

    J. Mol. Struct. 1992, 265, 179.

 For system like CH3COO-...3H2O, the molecular mechanics results can be
10-14 kcal/mol off the experimental value.

    Yajun Zheng
==========================================
Date: 03/05/92 08:14:34
From: "Joe Golab"  nap.amoco.com>

(--deleted--)
If I could routinely perform calculations of the type you describe, I would
want more immediately. (HAHA.) I am assuming that by routine, you mean that
the hardware/software combination is efficient, which for me implies I will
get the answer within 24 hours or less. Mainly, we would perform more
realistic simulations of reaction dynamics. We would survey a potential energy
surface to a higher degree of accuracy. We would use the method to
automatically incorporate correlation in a search for transition states (and
try to "firmly" answer whether this TS was unique). We would hope that
thermochemical properties thus obtained would be better than now.

We would probably also start to look - at - this new thing called "ab initio"
molecular dynamics for polymeric species. We would also most probably become
very interested in better metal surfaces, i.e. either "less effective" ECPs or
larger metal clusters or larger adsorbed/absorbed species or all of the above
or any combination.

Obviously, Jan, these opinions are mine and in no way reflect research I am
engaged in -8 at 8- Amoco Chemical Company, Amoco Corporation or any of its
subsidiaries.

Hope all is well.
:Joe

===============================
Date: Tue, 3 Mar 92 15:36:31 -0800
From: ross &$at$& zeno.mmwb.ucsf.EDU (Bill Ross)
Subject: Re:  The future of ab initio


I would do more nucleic acids and ion-nucleic complexes
for force field parameterization. Currently I'm doing
optimizations & charge calculations on various base pairs.

Bill Ross
UC San Francisco
================================
Date: Tue, 3 Mar 92 19:45:00 EST
From: chan "-at-" tristan.TN.CORNELL.EDU (Ernest Chan)
Subject: future of ab inito calculation

Although I have never done any ab initio calculations, I will love to use HF
method to study more complicated chemical reactions.

If your basis set is larger still, say 5000, quantum drug design seems like a
nice and useful application.

-Jack Chan
================================
Date: Tue, 3 Mar 1992 19:58 EST
From: "To help us serve you better, please take a number..."
 
Subject: the future of ab initio

(--deleted--)
	(2)  As to what would we do if we could routinely run very large
ab initio jobs...we are a group who both consumes and develops ab initio
code.  We have a joint NIH grant with an organic chemist to study the
structure and reactivity of taxol.  We are restricted to non-QM methods
for the most part and in practice are limited to semi-empircal methods even
for crude models of the system.  We'd be overjoyed (and back in our element)
if we could do ab initio calculations on taxol.  In general, we could have
more involved interactions with our organic colleagues if we had access
to ab initio code for very large systems.

				Sincerely,

					Michelle M. Francl
					Bryn Mawr College

				Internet: mfrancl-: at :-cc.brynmawr.edu
============================
Date: Tue, 3 Mar 92 16:54:32 -0800
From: mecolv "at@at" snll-arpagw.llnl.gov (colvin michael e)
Subject: Large ab initio calculations.
  
  Jan,
   On a semi-routine basis I am running 500 basis function single
point SCF and MP2 calculations (a 656 basis functions, 1968
primitives, no-sym direct SCF is currently queued up) as well as 250
basis function SCF optimizations and 120 basis function MP4 single
points.  I would certainly like to be able to run these on a truly
routine basis, that is, on my workstation.  I am involved in a number
of projects working with medicinal chemists to help understand the
biochemistry of pharmaceutical drugs.  To the medicinal chemists any
molecule with less than a hundred or so atoms is "small", so it's been
a challenge to find systems small enough to fit on our Y-MP.  Of
course there are lots of other complications in modeling biochemical
systems, such as aqueous solvation effects, which will make these
systems even more computationally demanding.
    For example one project involves studying the anti-cancer drug
cyclophosphamide and its active metabolites.  Since these compounds
contain several exotic moeties such as phosphorodiamides,
semiempirical methods don't work all that well, so we're forced to use
ab initio methods.  So far this has involved trying to determine the
aqueous-phase protonation state of the metabolites, looking at the
reaction enthalpies of deactivating reactions and exploring plausible
reaction pathways.  What we'd really like to is get a handle on the
DNA cross-linking which will require much larger calculations.  We did
a single point sto-3g SCF calculation on 6 base pairs of DNA
involving 1473 basis functions, but we can only afford this sort of
calculation once in a long while.
     To try to make such calculations more routine, we have been
working for several years to create a parallelized SCF, SCF gradient,
and MP2 package.  We have a "production" version of the first two
parts running on an NCUBE/2 and IPSC 860.  The performance is pretty
good, but the machines are sufficently cantancerous to make "routine"
runs difficult.
     If you'd like more info on the big calculations just send me some
mail at mecolv.,at,.sandia.llnl.gov.  If you're intersted in more info on
the parallel QC code send questions to Curt Janssen
(cljanss (+ at +) sandia.llnl.gov).
                                          --Mike Colvin
                                            Sandia National Laboratory
                                            Livermore, CA
===========================================
Date: Wed, 4 Mar 92 09:40:36 EST
From: states <-at-> ncbi.nlm.nih.GOV (David States)
Subject: Re: A gadfly forever..... (How far can you push semi-empirical
 methods?)

|>  Andy Holder writes:
(--deleted--)
To me this is a real opportunity.  Empirical energy simulations of
biological macromolecules have improved incrementally over the past
decade, but many significant terms in the potential energy function are
ignored, particularly polarization effects (charge induced dipole
terms, cooperativity in hydrogen bonding etc.).  These terms are
significant compared to kT.  It does not make alot of sense to me to
run longer and longer simulations when you know the underlying
potential function is flawed.  Semi-empirical methods offer a real
hope of addressing some of these issues.

So what are the largest peptide systems that have been evaluated
with semi-empirical methods?

Have any been done with Monte Carlo averaging over nuclear conformations?

Has anyone been bold enough to attempt dynamical simulations based
on semi-empirical PE functions and derivatives?

David States
National Center for Biotechnology Information / National Library of Medicine

=======================================
Date: Wed, 4 Mar 92 08:35:06 PST
From: case ( ( at ) ) scripps.EDU (David Case)
Subject: Re: A gadfly forever..... (How far can you push semi-empirical
 methods?)

David States wrote:
(--deleted--)

I too would enjoy hearing from people that know about this.  When I
wrote a review a few year ago [in "Conformational Analysis of
Medium-Sized Heterocycles", edited by Richard Glass] semiempirical
methods seemed to have very poor performance for non-bonded
interactions and hydrogen bonds -- things that are crucial to most
peptide and protein simulations.  Errors in barriers to rotation about
bonds, or in ring conformations (e.g. planar cylcopentane) were also
common.  But at that time, not much had been published with AM1,
although that appeared to give improved results in these areas.  Are
semi-empirical methods now meeting Dave States' desires?

David A. Case
(--deleted--)

================================
Date: Wed, 4 Mar 92 08:35:38 PST
From: d3f012(-(at)-)gator.pnl.GOV
Subject: large scale semi-empirical calculations


Andy Holder writes:(deleted)
David States writes:(deleted)
 We routinely do INDO/s SCF/CI calculations on the chromophores of
 bacterial photosynthetic reaction centers that involve ~600+ atoms,
 1500+ electrons, and a few thousand configurations (single-excited)
 in the CI.  These calculations also have included portions of the
 surrounding protein.

 We have been doing Molecular Dynamics with forces from a
 semi-empirical Hartree Fock solution.

Mark A. Thompson
(deleted)
d3f012(-(at)-)pnlg.pnl.gov

================================
Date: Fri, 06 Mar 92 08:36:37 +0900
From: kddlab!vega.rc.m-kasei.co.jp!mei (+ at +) uunet.UU.NET (Murakami Akinori)
Subject: Re: The future of ab initio

Thanks to Gaussian inc. and IBM
>>     What would you do if you could ROUTINELY perform ab initio
(--deleted--)
>> HF/6-31G* on C30H62 with no symmetry (574 basis functions) takes well under
(--deleted--)
>> Michael Frisch

We have following timing data of Gaussian90 on a IBM RS6000/550 which
is little bit expensive. IBM official Mflops is 25.
					unit[minute]
Molecule	Basis	SCF+grad
C10H8S2		132	28
C14H12		150	29
C18H12S2	212	109

We are running 200 basis( 20~30 atom) calculation routinely.

But we have to use about 300~500 basis on Metal complex calculation in
our catalyst research.
We want to optimize 500 basis molecule, it will take about 1000 hour
on 25 Mflops WS. One month for one calculation is not possible in our
research center.

					Akinori MURAKAMI
					Mitusbishi Kasei Corpration
					Yokohama, JAPAN
=====================================
Date: Sun, 8 Mar 1992 23:13 EST
From: "DOUGLAS A. SMITH" 
Subject: burning time on supercomputers

In response to Jan Labanowski's question, we have one area which still
seems to defy current resources, or at least pushes them to the extreme.
We have been studying hydrogen bonding in a series of di- and triamides
which involve from 8 to 20+ heavy atoms, using molecular mechanics to
probe the potential energy surface (you guessed it- the multiple minimum
problem).

We are now at the stage of beginning to take our tens of minima for each
compound, found using the AMBER force field in MacroModel, and submitting
each for ab initio single point and geometry optimization.  Hydrogen
bonding requires massive basis sets for molecules of this size - a recent
paper by Pople suggested that RHF/6-31+G** was a minimum requirement,
although they did their optimizations at RHF/6-31G* single points at
MP2/6-31+G**.  We want to also do RPAC calculations of NMR spectra and
chemical shift tensors, so the wavefunction must be a good one.  These
last few requirements are not trivial in and of themselves, but when
coupled with 8-10 compounds with from 5-40 minima each, the need for
resources just explodes.

Regardless, we are trying.

Doug Smith
Assistant Professor of Chemistry
The University of Toledo
Toledo, OH  43606-3390

voice   419-537-2116
fax     419-537-4033
email   fax0236 #*at*# uoft02.utoledo.edu



Similar Messages
08/01/1996:  Re: CCL:M:Heat of formation calculation using MOPAC.
08/01/1995:  Spin contamination, effect on energy and structure.
10/04/1994:  EPR Summary
06/08/1993:  undergrad computational chem
10/17/1994:  summary iron calculation
08/03/1995:  ACS Chicago - CINF Abstracts    - 29 pages document -
10/01/1993:  torsion of conjugated systems -- summary
03/07/1997:  Chem Topic: Excited States
05/01/1997:  calculation of chemical shifts (summary)
12/16/1994:  Spin contamination & AM1 "ROHF" versus UHF


Raw Message Text