From owner-nwchem-users@emsl.pnl.gov Wed Dec 13 08:46:01 2006 Received: from odyssey.emsl.pnl.gov (localhost [127.0.0.1]) by odyssey.emsl.pnl.gov (8.13.8/8.13.8) with ESMTP id kBDGk0uF023325 for ; Wed, 13 Dec 2006 08:46:01 -0800 (PST) Received: (from majordom@localhost) by odyssey.emsl.pnl.gov (8.13.8/8.13.8/Submit) id kBDGk0we023324 for nwchem-users-outgoing-0915; Wed, 13 Dec 2006 08:46:00 -0800 (PST) X-Authentication-Warning: odyssey.emsl.pnl.gov: majordom set sender to owner-nwchem-users@emsl.pnl.gov using -f X-Ironport-SG: No_SBRS X-Ironport-SBRS: None X-IronPort-Anti-Spam-Filtered: true X-IronPort-Anti-Spam-Result: Ao8CANq8f0WA/C7d/2dsb2JhbAA X-IronPort-AV: i="4.12,164,1165219200"; d="scan'208"; a="14292889:sNHT28934339" X-Authentication-Warning: globe.wustl.edu: gelb owned process doing -bs Date: Wed, 13 Dec 2006 10:45:56 -0600 (CST) From: Lev Gelb X-X-Sender: gelb@globe.wustl.edu To: nwchem-users@emsl.pnl.gov Subject: [NWCHEM] - converging band calculations on lithium Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed Sender: owner-nwchem-users@emsl.pnl.gov Precedence: bulk Dear NWChem-ers, I'm trying to converge a band calculation on solid bcc lithium, and am running into some strange behavior when using very small unit cells. All of the following calculations are done using the same cutoff (11.644, corresponding to 10 grid points in a 6.51 au cubic cell) I'm using a in-house compilation of NWChem 5.0. (Which seems to work well on the SiC example from the Users Manual.) Here are the results: 128 atoms in the cell (4x4x4 bcc unit cells) 1 BZ point (Gamma point only): energy: -0.26441E+00/ion 8 BZ points (monkhorst-pack 2 2 2) -0.26562E+00/ion 16 atoms in the cell (2x2x2 bcc unit cells) 1 BZ point (Gamma point) -0.26172E+00/ion 8 BZ points (monkhorst-pack 2 2 2) -0.26470E+00/ion 64 BZ points (monkhorst-pack 4 4 4) -0.26535E+00/ion 125 BZ points(monkhorst-pack 5 5 5) -0.26516E+00/ion 2 atoms in the cell (one bcc unit cell) 1 BZ point (Gamma point) -0.34430E+00/ion 8 BZ points (monkhorst-pack 2 2 2) -0.28012E+00/ion 64 BZ points (monkhorst-pack 4 4 4) -0.26476E+00/ion 125 BZ points(monkhorst-pack 5 5 5) -0.26305E+00/ion 216 BZ points(monkhorst-pack 6 6 6) -0.26216E+00/ion 512 BZ points(monkhorst-pack 8 8 8) -0.26130E+00/ion The 16-atom, 64-point calculation agrees pretty well with the 128-atom, 8-point calculation (they should agree exactly, I believe). Likewise, the 16-atom, 8 point calc. agrees well with the 2-atom, 64-point calculation AND the 128-atom, 1-point calc. (again, as expected.) However, as I keep increasing the number of BZ points in the 2-atom calculation, the results start to diverge from those in the larger cells, even though, at least to my understanding of this method, the 2-atom cell with 512 BZ points should agree well with the 16/64 and 128/8 calculations. I have repeated this series with two different pseudopotentials and two different energy cutoffs, and observed basically the same trends. I have fiddled with the ewald cutoffs and tolerances, which did not have any real effect. Can anybody explain this behavior? Cheers, Lev Gelb Note 1 - Developers - there is an output formatting error in NWChem 5.0 - if you use more than 100 BZ points, the two-digit labels in the point-by-point output section just read '**' Note 2 - I'm also having some problems with unit cell optimization, but have not finished my test runs on that, will post later. Note 3 - Here is an example of the smallest-cell input file: #------------------------------------------------- title "Lithium crystal optimization" start Lithium_1xtal_bcc geometry units au nocenter noautosym noautoz print system crystal lat_a 6.51d0 lat_b 6.51d0 lat_c 6.51d0 alpha 90.0d0 beta 90.0d0 gamma 90.0d0 end Li 0.000000 0.000000 0.000000 Li 0.500000 0.500000 0.500000 end #***** setup the nwpw gamma point code **** nwpw simulation_cell ngrid 10 10 10 end tolerances 1.0e-9 1.0e-9 ewald_ncut 20 ewald_rcut 10.0 monkhorst-pack 1 1 1 end set nwpw:minimizer 1 set nwpw:psi_nolattice .true. # turns of unit cell checking for wavefunctions driver clear maxiter 100 end # was set, but shouldn't matter? set includestress .true. task band energy #-------------------------------------------------- -------------------------------------------------------------------------- Lev Gelb http://www.chemistry.wustl.edu/~gelb gelb@wustl.edu Department of Chemistry, Washington University, Campus Box 1134 St. Louis, MO 63130 USA ph:(314)935-5026 fax:(314)935-4481 -------------------------------------------------------------------------- "71.3841674321 percent of statistics are made up on the spot." --- J. A. Paulos