Release Notes for NWChem version 3.2.1 NWChem Known Bugs in version 3.2.1 * Texas integrals do not function properly on the CRAY T3E. * Properties module does not work with spherical basis sets. Previous NWChem Known Bugs in version 3.2 * This is fixed in version 3.2.1. MP2 gradients: MP2 gradients are known to give wrong results. * Direct MP2 optimizations: MP2 optimizations that are performed with the direct option can have problems with CPHF convergence. In this case, the user is advised to use the default options which is a semi-direct calculation. It is always best to use the default algorithms and parameters in NWChem, until they do not work for your particular case. Machine specific changes to the defaults are continually being added to NWChem. * Texas integrals with high angular momentum and large exponents: This is fixed in version 3.2.1. There were multiple instabilities found in various recursions in the algorithms used in the Texas integral code (and in many other codes). All but one has been resolved. NWChem now checks for the possibility of this last instability and automatically turns the Texas integral code off when necessary. User's should not turn off the Texas integral code by default (except on the T3E). In 3.2, integrals that were calculated using the Texas integral package had errors when calculating integrals with high angular momentum and large exponents. This showed up, for example, when the cc-pVTZ basis set has been used. * Message mismatch or job hanging: These messages will occasionally be observed on NWMPP1 especially when running CCSD calculations. These are errors that happen sporadically and are being worked on. If you experience either of these problems, please send your bug report using the correct channel listed in the Reporting Problems with NWChem page. This will help us in the determination of the problem. Note that the message mismatch is generally found in the MPL version of NWChem on NWMPP1 and the job hanging is generally found in the LAPI version. If you experience a problem with one version, you might want to try the other version. On NWMPP1, the binaries are: /u/nwchem/bin/nwchem -> LAPI version /u/nwchem/bin/nwchem.mpl -> MPL version * start, restart, and continue in the title: When the words start, restart, and continue appear at the beginning of the title, NWChem erroneously picks them up when scanning for start, restart, and continue directives. The work around is to not use these words at the beginning of the title. * CCSD(T) energies are unstable on the IBMSP: CCSD(T) energies are unstable on the IBMSP using either the LAPI or the MPL version. Error that the user will see are ga_mismatch or jobs that just hang. We are working with IBM to solve this problem. NWChem Frequently Asked Questions * Where is the NWChem home page? http://www.emsl.pnl.gov:2080/docs/nwchem/nwchem.html * Where is the User's Manual? http://www.emsl.pnl.gov:2080/docs/nwchem/doc/user/index.html * Where is the Programmer's Manual? http://www.emsl.pnl.gov:2080/docs/nwchem/doc/prog/index.html * How do I submit a job on NWMPP1? -Check out the "Running NWChem" subsection in the LoadLeveler Description at MSCF page: http://www.emsl.pnl.gov:2080/mscf/hardware/hpcc/loadl_configuration.html -Use llnw on NWMPP1 -There is an nqenw for running on the NERSC T3E as well. * Is there any on-line training available? The slides from the NWChem / Ecce Tutorial workshop, March 23-25, 1998 are now available. http://www.emsl.pnl.gov:2080/mscf/training/tutorial/index.html * How do I get a better initial Hessian than the default diagonal or internal hessian for an optimization? The easiest method is to calculate a Hessian at a lower level of theory and basis set and then read that Hessian in for the geometry optimization. The key to doing this is to include INHESS 1 in DRIVER or STEPPER. An example of how to do this is: --- example start -------------------------------------------------------------- start h2o_optimization title; H2O optimization using a better initial Hessian geometry units angstroms O 0.0 0.0 0.0 H 0.0 0.0 1.0 H 0.0 1.0 0.0 end basis H library sto-3g O library sto-3g end scf print low end task scf freq basis H library 6-31g** O library 6-31g** end driver inhess 1 print medium end task scf optimize --- example end ---------------------------------------------------------------- * I am having problems getting my initial guess correct when I use ECPs. What can I do to get good starting orbitals? -The atomic initial guess currently does not work correctly in all cases with the presence of ECPs. A good method to use is independently converging fragments with ECPs. More details on how to do this can be found in the User's Manual. http://www.emsl.pnl.gov:2080/docs/nwchem/doc/user/index.html -Remember that you can (and SHOULD!) look at your initial orbitals by putting the line: print "initial vectors" into the SCF block of your input deck. If the order isn't correct, you can use the "swap" directive to change the order of the orbitals. For example: vectors atomic swap 173 175 174 176 output end.movecs will cause the orbitals 173 - 175 to be swapped to 175, 176, 173 and 174. Note that the swaps are pairwise: first 173 and 175 are swapped, then the the current orbitals 174 and 176 are swapped. More information on initial orbitals can be obtained in the User's Manual. Changes additions to functionality in NWChem (from the CVS log information) * More robust QA test cases (Courtesy of our users and in particular Dave Feller and John Nicholas; Thanks guys!) * Java Reporting tool developed by Danielle Farrar went into production. This tool is for use by all EMSL users/Collaborators and the specific point of contact at our distributed sites. It requires Java 1.1.x and allows users to browse for input, output, and error files that are attached to the email sent to the support staff. This eliminates "cutting" and "pasting" of this information. * "nqenw" a tool to submit jobs to NQE specific to NERSC. Operates like "llnw" on NWMPP1 at EMSL. * modifications were made to default DFT parameters for more robust/stable operation and energies/gradients. * new input style "lst" for geometries added. See http://www.emsl.pnl.gov:2080/docs/nwchem/doc/user/index.html * Several basis sets and ECPs (including large and small core Stuttgart) where added to the NWChem library. * bug in frozen core determination in MP2 was fixed. * an integrated preparation module "prepar" for the MD code was included. This eliminates the necessity for nwtop, nwrst, nwsgm in previous releases of NWChem. * Job Identification numbers for NQE and LoadLeveler are now reported in the NWChem output file. * The integral API was modified to pre-determine if the last Texas instability exists in a given basis set and turns Texas integrals off if necessary. The Texas integral interface now uses better logic for determining whether to use Texas 93 or Texas 95 routes. * Default Integral accuracy for radial cutoff in the API was tightened. (standard 10**-15, high 10**-30) * The integral API was extended to handle the required functionality for periodic DFT gradients. * default NWChem integral settings now handle spherical basis sets but not general contractions. * Fixed normalization for fitting style basis functions for DFT/RI methods. * Spin Orbit integrals and integral API routines were added. * Basis Set Object was modified to read Spin-Orbit specifications * Basis Set Object can now write basis set information to the Ecce output file. * LAPI and SP1 are now separate targets and builds for the IBMSP systems. * ChemIO updated to latest revision from Dr. J. Nieplocha * DFT weight derivatives bug was fixed. * ChangeLog for ESP module: > Changes to ESP since NWChem release 3.0 (TPS) > > Oct 06 Initial Electrostatic Potential (ESP) module > Current Features of the FIT: > a) Fit on grid points (cf CHELPG) > b) Constraints on total charge of atom groups > c) Restraints (cf RESP) > > Oct 09 Added USES_BLAS=esp_denmat > > Apr 21 1998 print under util_print contol > put fitted charges onto rtdb > > Apr 27 1998 defaults without esp input block > move probe/factor to esp_rtdbin > > Jun 15 1998 print probe radius and atom radius factor > Jun 16 1998 change default factor to 2.0 > > Aug 08 1998 destroy geometries on rtdb before storing new > > Aug 13 1998 bug fixes, extr printout * Geometry object knows how to print distances/angles/dihedral angles based on atomic radii. * Geometry object can now auto-detect and use symmetry when requested via the autosym keyword. This is still experimental but functional. * Accuracy of SCF Energies/Gradients now consistent. * integration of "NBO" directly into NWChem. You must have a license from the University of Wisconsin, Madison for NBO to use this functionality. * > Changes to nwARGOS since NWChem release 3.1 > > Dec 29 Include pme force calculation in time considered > in the load-balancing schemes. > > Jan 05 Include additional timer calls > > Jan 06 PME setup put in single subroutine > > Jan 06 PME self energy calculated only once > > Jan 06 Correct error in redetermination box-box list > > Jan 07 Bug fix in arg_listbb.F > > Jan 09 Moved zeroing pme charge grid, removing a sync > > Jan 15 Bug fix in arg_wrtraj.F/arg_hpsrti.F > Make calls to arg_listbb consistent > All changes copied to release tree > > Jan 23 Distance restraints > Duplicate call to forces removed from arg_stdesc > > Jan 26 Revised load balancing allowing nodes not involved in pme > reciprocal space calculations to continue with local > real space force calculations. This includes a different > order in box-box pairs, synchronization on selected nodes, > and delayed synchronization by use of flag/wait combination. > For now, this option is selected by 'develop 1' in the input file. > > Allow pair loadbalancing to continue with previous sub-box sizes. > > Jan 28 Add arg_forcon (was missing in repository) > > Feb 11 Modified load balancing scheme for PME (develop 1) > Modified atom name output in trajectory header > > Feb 11 Allow writing out trajectory and property files for minimizations > > Feb 13 Fix bug in steepest descent code (update pairlist and forces in > case solute shake is used) after recording. > Fix bug in resetting forces when a sd step is too large. > > Feb 15 Set velocities and kinetic energies to zero in stdesc runs > > Feb 17 Bug fix in pressure scaling array boxsiz > Use tim file to selectively write out timings for timing analysis > Properly close global array ga_flg > Bug fix in solvent fix > Fix format > > Feb 18 Add header to tim file > Time array dimension as parameter mtimes > > Feb 22 Add initial PME load balancing scheme > > Feb 25 Add arg_select > > Mar 03 Revised load balancing and timing > Set CG search interval to 2 times last step size in SD > nwArgos output shared file on SGI's > > Mar 04 Bug fix in arg_pscals.F > > Mar 05 Bug fix in arg_mdstep.F > Bug fix in arg_reset.F > Write proper number of total steps to property file > > Mar 10 Change output format of diffusion constant > > Mar 11 Fix QM/MM bug in zone atom determination > > Mar 14 Fix QMD property problem > > Apr 22 Initial pmf stuff > Optional read system_calc from input > > May 15 Initial 2nd derivative stuff > > May 29 Remove debug print statements from arg_newton.F > Fix bug in arg_memor1 in calculation minimum sub-box size > Issue warning if pressure scaling disabled > > Jun 22 Fix memory leak and add memory check for fft > > Jul 16 Catch lost solute atoms caused by small mol on many nodes > Prevent pwtot to be nan > > Aug 07 Initilization bug fix in arg_sync * DFT added new perdew 1996 functional * prepar module ChangeLog > Apr 14 1998 Initial prepar module routines > Add routines for rtdb input, third and excluded lists > Make lprint integer print control > Apr 15 1998 Print control compliant with nwchem > Bug fix in atom type generation > Parameter checks > Apr 16 1998 Handle solvent > Apr 17 1998 Generalize ring identification to 3,4,5 and 6 membered rings > Modified code to retrieve improper dihedral parameters > Allow force field atom type file in dir_s, dir_x, dir_u and dir_t > Apr 20 1998 Consolidate par and typ files > Corrections in improper dihedral center determination > Apr 21 1998 Coordinates in nm > List missing parameters only once > Read coordinates from rtdb > Add pre_hnames > Call to scf and esp to get charges in pre_charge > Apr 22 1998 Driver routines for restart generation > Apr 23 1998 Fix several errors > Center solute routine > Write restart routine > Read geometry object routine > Apr 28 1998 Add remaining options > Apr 29 1998 Bug fixes and Id's in all files > No constraints in esp when intergal molecule > Make sgm format identical to previous format > Apply amber improper format to sgm file > Make current directory ffield_t if not specified > Apr 30 1998 Report missing impropers for amber in correct order > May 01 1998 Fix bug for pdb starting with residue number larger than 1 > Properly deallocate heap memory > Correction handling single atom first molecule > May 05 1998 Minor changes and bug fixes > May 15 1998 Bug fix in declaration > Jun 03 1998 Change pre_namamb to pre_rename > No partial charge estimation if atom types unresolved > Bug fix in atom type string comparison in pre_ffield > Jun 04 1998 Bug fix in atom typing (pre_atype) > Charge determination separate for each segment (pre_charge) > Charge constraint of zero on set of cap atoms (pre_charge) > Handle charmm improper dihedrals (pre_ffield) > List atoms not found (pre_misfit) > Rename residues (CYS,HIS) appropriately for force field > Bug fix in index to array cseq (pre_rename) > Jun 05 1998 Bug fix in residue renaming (pre_rename) > Jun 08 1998 Compliance with util_print > Jun 09 1998 Avoid geom_rtdb_load error message > Jun 11 1998 Add topology modification commands > Fix bugs in reading solvent from pdb file > Keep name file > Jun 15 1998 Bug fixes: pre_boxsiz, pre_orient, pre_wrtrst, pre_wrttop > Jun 24 1998 Fix memory problem occurring in solvent-only preparations > Jul 09 1998 Add positional fix commands > Slightly increase hydrogen radius > Jul 14 1998 Add routine to determine if a segment is amino acid (pre_protein) > Put modified residue name on sequence file > Number non chain link atoms from 3 up ( instead of all being 3) > Fix file name problem in pre_mkrst > Expand link array to include global atom number > Fix bug in solvation routine > Fix bug in bonded list overwrite in pre_mkfrg > More diagnostic messages added > Fix file name bug in explicit write rst command > Jul 15 1998 Remove persistency new_top new_seq and new_rst > Resize solvent arrays when needed to read rst file > Allow coordinate read before solvation > Avoid write of ccordinates if explicit command given > Bug fix in restart file name > Read modified restart format > Jul 16 1998 Change default value of xpndw to 0.2 nm > Aug 07 1998 Initilization bug fix in pre_misfit > Aug 07 1998 Initilization bug fix in pre_short > Aug 10 1998 Add missing initialization angle constraints > Aug 13 1998 Add POx in charge determination * Symmetry module now handles spherical functions and T, Th, Oh groups * Task Module now prints each task time by default. * Frequencies and IR Intensities now computed if module computes dipole moments by default. * Default step-size for finite-difference hessian set to 0.001 * task_num_gradient now has optional 5 point difference formula for higher precision gradient. * Added "make smallversions" instead of "make versions" to allow for smaller routines for the util_version.F information. * Vib Module: added KM/mol units and dipole derivatives in normal modes