From owner-nwchem-users@emsl.pnl.gov Tue Mar 11 15:07:59 2008 Received: from odyssey.emsl.pnl.gov (localhost [127.0.0.1]) by odyssey.emsl.pnl.gov (8.14.1/8.14.1) with ESMTP id m2BM7wJV009616 for ; Tue, 11 Mar 2008 15:07:59 -0700 (PDT) Received: (from majordom@localhost) by odyssey.emsl.pnl.gov (8.14.1/8.14.1/Submit) id m2BM7wjj009615 for nwchem-users-outgoing-0915; Tue, 11 Mar 2008 15:07:58 -0700 (PDT) X-Authentication-Warning: odyssey.emsl.pnl.gov: majordom set sender to owner-nwchem-users@emsl.pnl.gov using -f X-IronPort-AV: E=Sophos;i="4.25,482,1199692800"; d="scan'208";a="45740392" X-MimeOLE: Produced By Microsoft Exchange V6.5 Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable Subject: RE: [NWCHEM] COSMO, speed, and odd results Date: Tue, 11 Mar 2008 15:07:46 -0700 Message-ID: <1692FA5FE0EBFD40B2B4097DAFEA887C0351B3F7@EMAIL02.pnl.gov> In-Reply-To: <47D6FBB6.5060504@northwestern.edu> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [NWCHEM] COSMO, speed, and odd results Thread-Index: AciDwetakjiXf8foTni816sNho6SYgAAROnw References: <47D6FBB6.5060504@northwestern.edu> From: "DeJong, Wibe A" To: "Joshua Vura-Weis" , X-OriginalArrivalTime: 11 Mar 2008 22:07:55.0394 (UTC) FILETIME=[5B9D8E20:01C883C4] Sender: owner-nwchem-users@emsl.pnl.gov Precedence: bulk Joshua, The current COSMO implementation in NWChem does not include the terms for the analytics gradient. As a result, numerical gradients are calculated. This means for each gradient, 26 atoms * 3 (xyz) * 5 (five-point formula) =3D 390 energies need to be calculated. This means: 1. Yes this will be expensive, even though each energy point can be done quickly.=20 2. Your grep actually includes all those solvation energies and you would not see a real trend. 3. You should do a grep on "@" to see the geometry optimization steps taken, and to see the energy convergence. The error message you see tell me that one of the 390 displacements does not converge properly, but you would not see this with "print low". You might want to increase the default number of iterations in the dft. =20 This calculation is parallelizable, not to hundreds of processors, but definitely to 16 or 32. Thanks, Bert de Jong, Ph.D. Chief Scientist Capability Steward NWChem High Performance Software Development Environmental Molecular Science Laboratory Pacific Northwest National Laboratory P.O. Box 999, Mail Stop K8-91 Richland, WA 99352 Phone: (509) 371-6476 Email: bert.dejong@pnl.gov Fax: (509) 371-6445 Room: EMSL 1371 Web: http://mscf.emsl.pnl.gov -----Original Message----- From: owner-nwchem-developers@emsl.pnl.gov [mailto:owner-nwchem-developers@emsl.pnl.gov] On Behalf Of Joshua Vura-Weis Sent: Tuesday, March 11, 2008 2:38 PM To: nwchem-users@emsl.pnl.gov Subject: [NWCHEM] COSMO, speed, and odd results Hi all, This is a 2-part question about COSMO. I am trying to calculate the optimized structure of pyromellitimide (http://www.sigmaaldrich.com/catalog/search/ProductDetail/ALDRICH/P73005 ) in THF. The calculation has been going for over a week on one processor (2-year old Xeon), and the energy does not seem to be converging or well-behaved. There are lots of spikes and sometimes the calculation dies with following error, though we can restart from the last good geometry: num_grad: !! displaced energy failed =20 ------------------------------------------------------------------------ driver: task_gradient failed 0 So, first question: am I doing something wrong with the calculation to make it so erratic? I have been grepping 'sol phase energy' and plotting the results, and it's definitely not the nice downward-staircase that I usually see from geometry optimizations (currently it's an upward staircase with many spikes). Second, how parallelizable should this calculation be, either using 'parallel' or MPI? The input file is below. I'm using a development version of NWChem 5, downloaded on March 2 2007, and compiled on RH Linux with IFC/ICC. Xeon 3.2 Ghz with 4GB RAM. scratch_dir /scr permanent_dir /h2/hsk384/FILES memory 1800 mb print low start PI_b3lyp_321g_cosmo_THF_opt title "PI 3-21G cosmo opt b3lyp geometry" echo charge 0 geometry units angstroms C -0.69982096 1.17417649 0.00141751 C -1.44288637 0.00000322 -0.00186793 C -0.69987852 -1.17419408 -0.00312106 C 0.69982096 -1.17417649 -0.00141751 C 1.44288637 -0.00000322 0.00186793 C 0.69987852 1.17419408 0.00312106 C -1.17499623 -2.59556434 0.00032946 N -0.00012863 -3.37239359 -0.00368894 C 1.17491383 -2.59569429 0.00174194 C 1.17499623 2.59556434 -0.00032946 N 0.00012863 3.37239359 0.00368894 C -1.17491383 2.59569429 -0.00174194 C 0.00045225 4.83705633 0.00408774 O 2.32227577 3.03442120 -0.01460665 O -2.32202893 3.03485526 -0.01572991 O 2.32202893 -3.03485526 0.01572991 O -2.32227577 -3.03442120 0.01460665 C -0.00045225 -4.83705633 -0.00408774 H -2.52579133 0.00019832 -0.00312645 H 2.52579133 -0.00019832 0.00312645 H -0.89114159 5.18123602 -0.52360267 H 0.90780369 5.17827926 -0.49789027 H -0.01375284 5.22640815 1.02698095 H 0.01375284 -5.22640815 -1.02698095 H -0.90780369 -5.17827926 0.49789027 H 0.89114159 -5.18123602 0.52360267 end basis * library 3-21G* end driver maxiter 40 end dft xc b3lyp direct end cosmo dielec 7.6 end task dft optimize