From chemistry-request@server.ccl.net Mon Jun 12 16:17:08 2000 Received: from bobino.sefmedia.com (bobino.sefmedia.com [207.164.6.2]) by server.ccl.net (8.8.7/8.8.7) with ESMTP id QAA12337 for ; Mon, 12 Jun 2000 16:17:02 -0400 Received: from ip252.montreal27.dialup.canada.psi.net (ip252.montreal27.dialup.canada.psi.net [154.5.1.252]) by bobino.sefmedia.com (NTMail 5.06.0014/NT6102.00.ff403942) with ESMTP id ejwgaaaa for chemistry@ccl.net; Mon, 12 Jun 2000 16:17:45 -0400 Message-ID: <021b01bfd4ab$57efe960$9a0101c0@chemcomp.com> From: "bill email" To: Subject: CCG - COMP division of ACS Excellence Awards Date: Mon, 12 Jun 2000 16:17:41 -0400 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 5.00.2615.200 X-Mimeole: Produced By Microsoft MimeOLE V5.00.2615.200 CHEMICAL COMPUTING GROUP ESTABLISHS STUDENT EXCELLENCE AWARDS IN CONJUNCTION WITH THE COMP DIVISION OF ACS June 14, 2000 (Montreal/Washington D.C.) An important contribution to graduate education in computational chemistry was made this week by the Chemical Computing Group (CCG). As part of an agreement between CCG and the ACS Division of Computers in Chemistry, CCG will provide $100,000 in merit-based awards for graduate students to allow them to attend ACS National meetings as well as provide them with software licenses. "We are very pleased to establish the CCG Excellence Award in conjunction with the COMP (Computers in Chemistry) division of ACS," stated Bill Hayden, Vice President - Sales & Marketing of Chemical Computing Group. "The fund rewards deserving students in the field of computational chemistry while at the same time allowing them to present their work at ACS national meetings where they can interact directly with leading scientists in their field. We see our contribution as an opportunity to invest in the future of scientific research and help promote new ideas and new researchers in the field" Dr. Philip Bowen, Chairman of the COMP division of ACS said, "We are very happy that CCG has set up the Excellence Awards. It is important for students to have the opportunity to attend a venue like the national meetings, to meet with fellow scientists and learn about new breakthroughs in the area of computational chemistry. Additionally, by supplying the deserving winners with their MOE software package, CCG provides them with a valuable state-of-the-art computational tool." Ten CCG Excellence Awards will be given out each year at the two ACS National Meetings. Awards will be made on the basis of the scientific merit of submitted abstracts, and on the established guidelines of COMP Division student travel awards. Chemical Computing Group (http://www.chemcomp.com) develops and markets MOE, the Molecular Operating Environment, the next generation of computer application for chemical researchers in the pharmaceutical and biotechnology fields. Its built-in applications cover the spectrum of drug discovery including: Protein/Homology modeling, High Throughput Screening, Combinatorial Library Design, Modeling/Simulations and Methods Development. MOE's unique architecture and platform independence allows it to be used corporate-wide by a variety of researchers from methods developers to computational experts to medicinal chemists. Its customers include a worldwide roster of leading pharmaceutical and biotechnology. For further information and/or to apply for the award please contact Professor Curt M. Breneman, COMP Treasurer and Fundraising Coordinator at the RPI Department of Chemistry: brenec@rpi.edu , (518) 276-2678 - phone / (518) 276-4045 - fax. Contact: Bill Hayden, Vice President 514 393 1055 - phone 514 874 9538 - fax http://www.chemcomp.com info@chemcomp.com William A. Hayden Vice President, Sales & Marketing CHEMICIAL COMPUTING GROUP hayden@chemcomp.com http://www.chemcomp.com 514 393 1055 - phone 514 874 9538 - fax From chemistry-request@server.ccl.net Mon Jun 12 18:11:33 2000 Received: from lrz.uni-muenchen.de (root@lsanca1-ar99-078-042.biz.dsl.gtei.net [4.3.78.42]) by server.ccl.net (8.8.7/8.8.7) with ESMTP id SAA12843 for ; Mon, 12 Jun 2000 18:11:32 -0400 Received: (from eugene.leitl@localhost) by lrz.uni-muenchen.de (8.8.8/8.8.8) id OAA22566 for ; Mon, 12 Jun 2000 14:11:20 -0700 Resent-From: Eugene Leitl Resent-Message-ID: <14661.20984.177516.632937@lrz.uni-muenchen.de> Resent-Date: Mon, 12 Jun 2000 14:11:20 -0700 (PDT) Resent-To: X-Authentication-Warning: verdun.ks.uiuc.edu: jim owned process doing -bs In-Reply-To: <200006071839.e57Idtx18234@axe1.med.upenn.edu> Message-Id: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Errors-To: beowulf-admin@beowulf.org X-Mailman-Version: 1.1 Precedence: bulk List-Id: Discussion of topics related to Beowulf clusters X-BeenThere: beowulf@beowulf.org From: Jim Phillips To: axelsen@axe1.med.upenn.edu cc: beowulf@beowulf.org Subject: Re: Scalability of CHARMM on various architectures Date: Mon, 12 Jun 2000 17:06:45 -0500 (CDT) Hi, Depending on the features you require, NAMD is compatible with CHARMM and should scale better, especially on clusters. New version out soon. http://www.ks.uiuc.edu/Research/namd/ -Jim On Wed, 7 Jun 2000 axelsen@axe1.med.upenn.edu wrote: > > > > > We are designing a cluster for which the most important code will be the > computational chemistry program, CHARMM. In our preliminary tests on > an existing cluster, we have confirmed our expectation that interthread > communications will be our first bottleneck. A test run on a typical > problem did not scale beyond 4 nodes on a cluster composed of P2-450 > processors and fast ethernet interconnections. In contrast, the same > code scales almost perfectly up to at least 12 processors on an SGI-Unix > SMP machine. CHARMM uses PVM. > > I would appreciate contact from anyone who has run CHARMM on a cluster > and has considered the best way to make this code scale better. From > anyone, I would appreciate general guidance on several issues: > > * If we go with pentium II/III processors, how far is Myrinet likely to > permit us to scale these calculations? > > * With Myrinet, would alpha nodes tend to scale any better than pentium nodes? > > (a single 500 MHz alpha processor is about 1.6-fold faster than a > single 450 MHz P2 on a typical problem, but it is about 3-fold the > cost. With this question, I am looking for any additional advantages > of alpha to justify this cost) > > * Are there differences between different pentium chip sets that will impact > this problem? > > * Is there any advantage to buying dual-processor machines, either alpha > or pentium, with respect to scaling? Any advantage to dedicating one > processor in each dual-box to communications? If we dedicate processors > in this way, would both processors have to have the same clock speed? > > > > > ------------- axe@pharm.med.upenn.edu ----------------- > > Paul H. Axelsen .... .... . . . . > Department of Pharmacology . . . .. . .. . > University of Pennsylvania .... ... . . . . . . > 3620 Hamilton Walk . . . .. . .. > Philadelphia, PA 19104-6084 . .... . . . . > > ------------------------------------------------------- > > _______________________________________________ > Beowulf mailing list > Beowulf@beowulf.org > http://www.beowulf.org/mailman/listinfo/beowulf > _______________________________________________ Beowulf mailing list Beowulf@beowulf.org http://www.beowulf.org/mailman/listinfo/beowulf From chemistry-request@server.ccl.net Tue Jun 13 03:33:55 2000 Received: from linux2.ipc.pku.edu.cn (mao@linux2.ipc.pku.edu.cn [162.105.177.40]) by server.ccl.net (8.8.7/8.8.7) with ESMTP id DAA14648 for ; Tue, 13 Jun 2000 03:33:50 -0400 Received: from localhost (mao@localhost) by linux2.ipc.pku.edu.cn (8.9.3/8.9.3) with ESMTP id PAA11989 for ; Tue, 13 Jun 2000 15:35:59 +0800 Date: Tue, 13 Jun 2000 15:35:59 +0800 (CST) From: Fenglou Mao To: "CHEMISTRY@www.ccl.net" Subject: PentiumIII and linux 2.2.14-5 error? Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII hi, all, Someone declared that he found linux2.2.14-5 have "nan" error, I met it, too, he said in 2.2.14-12 will be OK, I also found in Athlon, I did not meet this kind of error. DId anyone else meet this, too? By the way, who cal tell me where I can find the new 2.2.14-12 kernal? Sincerely Yours, FengLou Mao ******************************* ADD:Mr. FengLou Mao Institute of Physical Chemistry Peking University BeiJing P.R.China Tel:86-10-62756833 Fax:86-10-62751725 From chemistry-request@server.ccl.net Tue Jun 13 09:41:22 2000 Received: from ant.chemie.fu-berlin.de (root@ant.chemie.fu-berlin.de [160.45.20.55]) by server.ccl.net (8.8.7/8.8.7) with ESMTP id JAA16803 for ; Tue, 13 Jun 2000 09:41:20 -0400 Received: (from leti@localhost) by ant.chemie.fu-berlin.de (8.9.3/8.9.3) id PAA32451; Tue, 13 Jun 2000 15:41:02 +0200 Date: Tue, 13 Jun 2000 15:41:02 +0200 (CEST) From: Leticia Gonzalez Herrero To: CHEMISTRY@server.ccl.net Subject: SUMMARY:HF/DFT and dissociation Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Dear CCl members, Some days ago, I posted a question about HF and DFT suitable for describing dissociation. Many many thanks to all who replied. Please find my original question and replies on the subject topic: >If I want to describe the rupture of a covalent bond, it is known >that RHF will lead to wrong dissociation energies while UHF will cure >the energy problem but converges to an unphysical mixture of singlet >and triplet states. Therefore, in general some other post-HF methods are >used to describe dissociation. >And what happens if I use DFT? Is DFT, restricted or unrestricted able to >describe the dissocation of a covalent bond? -----------------From Marcel Swart, there is an interesting article by Baerends/Gritsenko about this subject, where they prove that DFT does give the correct dissociation. I think you could find it in: J.Phys.Chem.A, Vol 101(30), 1997, p. 5383-5403. -----------------From Keith Refson, > If I want to describe the rupture of a covalent bond, it is known > that RHF will lead to wrong dissociation energies while UHF will cure > the energy problem but converges to an unphysical mixture of singlet > and triplet states. Therefore, in general some other post-HF methods > are used to describe dissociation. There's a nice discussion of this on p.60 of: "Bonding and Structure of Molecules and Solids" by David Pettifor (Oxford Science, ISBN 0-19-851786-6) > And what happens if I use DFT? Is DFT, restricted or unrestricted able > to describe the dissocation of a covalent bond? Yes, at least LSDA gives the correct dissociation limit. (LSDA is the spin-polarized LDA -- "unrestricted" in chemistry terminology). ---------------From Stefan Grimme For the dissociation of a covalent bond RDFT and UDFT will give the same failures as HF because the ansatz is the same in both methods (one-determinatal). However, in general the spin-contamination problem at intermediate distances is not so severe in DFT and the asymtotic energy with RDFT is usually better than with RHF because of the inherent self-interaction error of commonly used functionals (from another point of view one can argue that the RHF error for e.g. H2 at r=infinity is 0.5*(IP+EA) of the H-Atom, a quantity which is much smaller in DFT). -------------From Antonio M. Marquez Con DFT tienes el mismo problema. Al estar basado en un solo determinate de Slater no describiras bien la disociacion y obtendras mezclas singlete/triplete. Para u na descripcion correcta del problema necesitas una funcion multiconfiguracional tip o CASSCF al menos. Si se trata de un enlace covalente mas o menos normalito el calculo es sencillo. Si tienes metales de transicion implicados se complica pues puedes tener varios estados proximos en energia y cruces entre ellos en la superficie de energia potencial. ------------From Klaus Stark DFT based on a restricted determinant gives energies too high for large bond distances, i.e. goes to an incorrect dissociation limit. UHF references remedy for that. However, LDA/UHF dissociation energies might be too large. Gradient corrected and hybrid methods follow the "true" (e.g. CASSCF) potential curve much more closely. DFT methods based on UHF references do not have a problem with spin contamination. Indeed there's a paper by Pople, Gill and Handy showing that spin contamination is not well defined in DFT (Int. J. Quantum Chem. 56 (1995) 303). -----------------From Reinaldo Pis Diez Yes, you can use unrestricted dft in broken symmetry mode to study dissociation of covalent bonds. Broken symmetry is very important in such cases since it is the only way to force different electronic configurations for the isolated atoms, i.e, you can think of H2 dissociating into two H atoms, one having an spin-up electron and the other having an spin-down one. If you don't use broken symmetry the isolated atoms will be the same. That can be accomplished by using guess(mix) in gaussian 98 and using the keywords modifystartpotential and fragoccupations if you are dealing with adf99. ------------Richard P. Muller > If I want to describe the rupture of a covalent bond, it is known > that RHF will lead to wrong dissociation energies while UHF will cure > the energy problem but converges to an unphysical mixture of singlet > and triplet states. Therefore, in general some other post-HF methods are > used to describe dissociation. Correct, although you can also use a generalized valence bond (essentially a two configuration MCSCF, but one that doesn't require many additional 2e integrals beyond those required for HF) to get a correct dissociation energy and proper spin states. The Jaguar program suite (see http://www.schrodinger.com) has a nice GVB implementation that I programmed a few years back. > And what happens if I use DFT? Is DFT, restricted or unrestricted able > to describe the dissocation of a covalent bond? You can do either spin restricted or spin unrestricted DFT, although unrestricted is more common. Jaguar has implementations of both. The RDFT will give a better dissociation energy that RHF, but one that in many cases will go to the wrong limit. UDFT is more reliable for broken bonds. It would probably be possible to do a hybrid GVB-DFT that would combine the best elements of both theories, but I've never gotten around to doing it. Jaguar has something like this (or at least it once did), but it isn't quite what I had in mind. -----------------From Sergio Emanuel Galembeck The best methods to describe the rupture of a bond are multireference methods, as CAS, because the problem involves more then one configuration. This is the reason why DFT or other single-reference methods don't work well in dissociation. -------------Ernest Chamot : As the atoms connected by a bond separate, the electrons that initially formed that bond change from being paired to being independent: ie from being a singlet to being two separated doublets, or a triplet. And since electron reorganization will take place much faster than atomic movement, they can't be restricted to a specific multiplicity in the calculation of the PES as you break the bond. So the only choice, really, is some sort of UHF calculation. But as you point out, this leaves the calculation open to spin contamination. In the case of ab initio HF calculations, this can be extreme, and give artificially high activation energies. You can monitor the spin contamination in your calculation by plotting the vs. atom separation. There are some very nice illustrations of this, and the resulting effect creating an artificial "activation" energy for bond dissociation, in the figures in the paper by Chen and Schlegel. (It also interferes with convergence.): W. Chen & H. B. Schlegel, J. Chem. Phys. 101(7), 5957-68 (1994). They studied various post-HF methods, and in particular the value of spin projection to overcome this problem in MP calculations, for various H atom dissociations. The problem of spin contamination in UHF calculations is variable, however. In fact, for some calculations involving dissociation of 2 heavy atoms (as usual, H seems to be a particularly difficult problem), plotting shows that spin contamination is minimal, and doesn't necessarily create an artificial "transition state". I have seen this with AM1 in some of my own work, and there is also an instance shown in: J. Baker, M. Muir, & J. Andzelm, J. Chem. Phys., 102(5), 2063-79 (1995). Of course, without correlation, HF methods (including AM1) systematically underpredict the dissociation energies. So you will want a correlated method anyway, and the high level post-HF methods can be very compute intensive. Which brings us to the use of DFT, and I think you are on exactly the right track: >And what happens if I use DFT? Is DFT, restricted or unrestricted able to >describe the dissocation of a covalent bond? Several papers now have shown that DFT methods (unrestricted calculations) seem to be much less prone to spin contamination than HF or post-HF methods. Baker & Andzelm reported the advantage of BLYP and ACM in minimizing spin contamination for a series of reactions, and observe, "under the right circumstances DFT wave functions can exhibit minimal contamination over an entire energy surface, including dissociation." Jungkamp and Seinfeld found B3LYP particularly effective in minimizing spin contamination problems in calculating BDE's and transition state barriers for a series of reactions: T. Jungkamp & J. Seinfeld, J. Chem. Phys., 107(5), 1513-21 (1997). All three of these papers, though, show that there are tradeoffs in accuracy of energies (overpredicting, underpredicting, difficulty converging, etc.), however, so you may still prefer a high level post-HF calculation with a large basis set. But in general, DFT looks like the best way to go. ------------------------------- =-=-=-=-=-=-=-==-=-=-=-=-=-=-=-=-=-=-==-=-=-=-=-=-=-=-=-=-=-=-=-= Dr. Leticia Gonzalez Tel: ()49 30 838 520 97 Freie Universitaet Berlin Fax: ()49 30 838 547 92 Institut fuer Chemie Room: 35.17 Takustrasse 3 leti@chemie.fu-berlin.de D-14195 Berlin http://userpage.chemie.fu-berlin.de/~leti =-=-=-=-=-=-=-==-=-=-=-=-=-=-=-=-=-=-==-=-=-=-=-=-=-=-=-=-=-=-=-= If you find a solution and become attached to it, the solution may become your next problem. From chemistry-request@server.ccl.net Tue Jun 13 10:25:28 2000 Received: from Lilly.com (carltn.d48.lilly.com [40.33.1.27]) by server.ccl.net (8.8.7/8.8.7) with ESMTP id KAA17234 for ; Tue, 13 Jun 2000 10:25:28 -0400 Received: from LILLY.COM ([40.59.1.39]) by CARLTN.D48.LILLY.COM (PMDF V5.2-31 #34229) with ESMTP id <01JQJS6EFITI00BL9M@CARLTN.D48.LILLY.COM> for chemistry@ccl.net; Tue, 13 Jun 2000 09:17:55 EST Received: from aammta1.d51.lilly.com (aammta1.d51.lilly.com [40.1.129.162]) by NEWMAN.D52.LILLY.COM (PMDF V5.2-31 #34229) with ESMTP id <01JQJSD02P26009RIC@NEWMAN.D52.LILLY.COM> for chemistry@ccl.net; Tue, 13 Jun 2000 09:23:15 +0000 (GMT) Date: Tue, 13 Jun 2000 09:23:09 -0500 From: HARPER_RICHARD_W@Lilly.com Subject: Re: CCL:PentiumIII and linux 2.2.14-5 error? To: chemistry@ccl.net Message-id: MIME-version: 1.0 Content-type: text/plain; charset=us-ascii X-MIMETrack: Serialize by Router on AAMMTA1/AM/LLY(Release 5.0.3 (Intl)|21 March 2000) at 06/13/2000 09:23:15 AM The latest stable version of the kernel is 2.2.16 - available at http://www.kernel.org. Regards, Dick Harper Fenglou Mao @ccl.net> on 06/13/2000 02:35:59 AM Sent by: Computational Chemistry List To: "CHEMISTRY@www.ccl.net" cc: Subject: CCL:PentiumIII and linux 2.2.14-5 error? hi, all, Someone declared that he found linux2.2.14-5 have "nan" error, I met it, too, he said in 2.2.14-12 will be OK, I also found in Athlon, I did not meet this kind of error. DId anyone else meet this, too? By the way, who cal tell me where I can find the new 2.2.14-12 kernal? Sincerely Yours, FengLou Mao ******************************* ADD:Mr. FengLou Mao Institute of Physical Chemistry Peking University BeiJing P.R.China Tel:86-10-62756833 Fax:86-10-62751725 -= This is automatically added to each message by mailing script =- CHEMISTRY@ccl.net -- To Everybody | CHEMISTRY-REQUEST@ccl.net -- To Admins MAILSERV@ccl.net -- HELP CHEMISTRY or HELP SEARCH CHEMISTRY-SEARCH@ccl.net -- archive search | Gopher: gopher.ccl.net 70 Ftp: ftp.ccl.net | WWW: http://www.ccl.net/chemistry/ | Jan: jkl@ccl.net