From 17b3efcfd1fd4ee4b565a0affc807e2f1c8afc0b Mon Sep 17 00:00:00 2001 From: Theresa Windus Date: Mon, 2 Aug 1999 23:56:33 +0000 Subject: [PATCH] Release notes for version 3.3 --- release.notes.3.3 | 272 ++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 272 insertions(+) create mode 100644 release.notes.3.3 diff --git a/release.notes.3.3 b/release.notes.3.3 new file mode 100644 index 0000000000..e6bf2dc1ed --- /dev/null +++ b/release.notes.3.3 @@ -0,0 +1,272 @@ +NWChem Version 3.3 Release Notes +(Also available at http://www.emsl.pnl.gov/pub/docs/nwchem/support/release.3.3.html) + +NOTE: These release notes are not a substitute for reading the User Manual! +They are meant to give a thumbnail sketch of the new capabilities and bug +fixes that are available in NWChem Version 3.3. When there is a conflict +between the release notes and the User Manual, the User Manual takes +precedence. + +Several new major capabilities are available in NWChem, including several new +modules. The new modules are: + + o Car-Parrinello simulations (the PSPW module), + o Python interface (the PYTHON module), and + o analysis of trajectory information during a molecular dynamics calculation + and of ESP calculations(the ANALYSIS module). + +For each of these capabilities, the user is referred to the User Manual for more +information. + +Additional major additions to the capabilities include: + + o GIAO NMR chemical shift capabilities for closed shells (GIAO in the + PROPERTY module), + o The DRIVER module has been greatly improved, with many new features, + o Several new DFT functionals have been added: + Exchange Perdew91, Gill96, and PBE96 functionals + Correlation PBE96 functional + Exchange-Correlation Becke97, Becke97-1, and HCTH functionals + o The ElectroStatic Potential (ESP) module has several new features, + o Automatic recognition of molecular symmetry (AUTOSYM in the + GEOMETRY module), + o Addition of Hondo 1e- energy and gradient integrals, + o Increased accuracy in the integral API. + +Again, for each of these capabilities, the user is referred to the User Manual for +more information. + +Listed below are the other major and many minor changes for each module +with significant changes. These descriptions are somewhat terse and more +information is generally available in the User Manual. + +Task level directives: + + o Unrecognized directives will now cause NWChem to stop and print an + error message + o Task analysis added for MD runs + o Task python added for python interface + o Task pspw added for pseudopotential plane-wave (Car-Parrinello) runs + +Initial orbital Guess: + + o New code from Paul Sherwood to generate reasonable initial atomic + guesses for heavy elements with ECPs. This generates guess densities for + those atoms that previously didn't work and produces much better guess + densities than the previous version. + o The atomic guess code now generates f orbital density fragments in the + correct order for both cartesian and spherical basis sets. + o Higher accuracy is now used in the initial guess to accomodate large basis + sets. If linear dependencies are found, even higher accuracy is used. + o Users are warned if initial guess can lead to symmetry problems. + o BUG FIX: UHF projections of small basis sets to large basis sets will now + work + +Basis Sets: + + o Basis set library has been updated to include all new basis sets at the EMSL + Gaussian Basis Set Order Form page of David Feller + (http://www.emsl.pnl.gov:2080/forms/basisform.html) + o BUG FIX: Some of the Stuttgart ECPs which had the coefficients and + exponents switched are now correct. + +Geometry: + + o Extended redundant internals to allow user constraints. + o Adjustable geometries. + o AUTOZ is now the default. + o A tolerance is now available in SYMMETRY directive of GEOMETRY + directive. + o AUTOSYM is now available (with a tolerance) on the GEOMETRY line + directive. + o Optional velocity input for Car Parinello code. + o More robust z-matrix code. + o C7 and D6 symmetries are properly set up. + +DFT: + + o Introduced X Perdew91, Gill96, and PB96 functionals. + o Introduced C PB96 functional. + o Introduced XC Becke97, Becke97-1, and HCTH. + o DFT is now able to use spherical basis functions as well as cartesian. + o DFT gradients can now use the Texas integrals. + o Linear scaling is acheived in the exchange contribution when charge fitting + basis sets are used with hybrid density functionals. + o If the vectors directive is used with SCF or DFT and a problem occurs, the + atomic guess will be used. + o Lebedev grids have been significantly modified. + o Delley weights are no longer allowed. + o Modified grid input: you can now add nrad nang accqrad. + o Better screening on the exchange correlation numerical grid, which + improves performance by at least 20%. + o Default convergence parameters was modified to produce more robust + results. + o Introduced check on integrated density: a warning message is printed if the + claimed accuracy criteria for the XC numerical grid are not met. + o Default finite difference step for nuclear hessian generation for DFT + changed to 0.01 au and for all other methods it is 0.001 au. + o BUG FIX: In cases in which the spin polarization is large in open shell + systems, wrong wave functions and gradients were obtained using + Perdew86 C functionals. This has been fixed. + o BUG FIX: If the user has a CD fitting basis set with a shell larger than any + shell in the AO basis set, the program would stop. This no longer applies. + o BUG FIX: All hybrid functionals were giving incorrect results when using + open shell wavefunctions because the total density was being used instead + of separate alpha and beta Hartree-Fock exchange contributions. This is + now fixed. (It was also fixed in later releases of 3.2.1.) + +SCF: + + o Eliminated or tightened screening options to make integral selection + threshold more consistent between file and direct routines. + o Improved precision in linear solve to stabilize direct SCF. + o If the vectors directive is used with SCF or DFT and a problem occurs, the + atomic guess will be used. + +Driver: + + o Revamped Driver capabilities. + o Ensure constraints are imposed after geometry step. + o Extended redundant internals to allow user constraints. + o AUTOZ is now the default. + +Properties: + + o Note that spherical functions do not work yet with properties. + o Input for the NBO program are produced. + o Addition of GIAO NMR chemical shift calculations. + o Added "all" keyword. + o BUG FIX:Property package will correctly skip property evaluation for + generally contracted basis sets and basis sets with pure sp functions (to get + the properties use the "segement" keyword when specifying a basis set). + +MP2: + + o TIGHT directive MP2 for higher precision energy and gradients + o BUG FIX: Non-abelian point groups are now handled properly in the MP2 + gradient code. + o Make RI-MP2 compute SCF reference if it is not already done (or done + with sufficient accuracy). + o BUG FIX: RIMP2 - Fix overflow memory problem for large jobs. + +Hessian and Frequencies: + + o Default finite difference step for nuclear hessian generation for DFT + changed to 0.01 au and for all other methods it is 0.001 au. + o BUG FIX: Numerical hessian restarts now work properly. + +ESP: + + o Use of fast routines for ESP calculations. + o Write name.xyz and name.plt file (plot with gOpenMol). + o Write name.er file to recover previous runs. + o Write name.q file with partial charges. + o Added Schwarz screening - screen option in ESP block. + o Added option to read in previous esp grid, if name.grid is present, it will + automatically get used unless you give a command recalculate. + o Print atom radii. + o Additional constraint options. + +Prepare module: + + o Changed topology and restart format printed out. + o New ALTLOC, MODEL, LINK, and CHAIN commands. + o Allow comment lines in sequence file - begin line with #. + o Always recalculate grid if ESP required on a fragment. + o Implement 'modify segment' and 'update lists' commands. + o Force PDB link cards. + o Add modify commands for bonded interaction parameters. + o Recode pre_mklist to allow and identify multiple node (greater than 3) + bonded interactions. + o Automatically identifies cysteine-sulphur bridges. + o Default HIS to HID for amber when reading PDB without hydrogens. + o Write out segment charges when writing topology file. + o Add support for DNA and RNA with AMBER force field. + o Add ignore keyword so that bonded parameters between quantum atoms + do not need to be defined in the database. + o Make 1.0 default scaling for ESP fitted charges. + o BUG FIX:Multiple node bonded interaction error takes care of itself. + o BUG FIX:If last segment was a single atom, the atom would be lost. This + is now fixed. + o KNOWN BUG:In some systems, when converting a restart file to a PDB + file, the restart file gets corrupted. + +MD: + + o Changed topology format - cannot use old files with the new code!! + o Added ability to analyze the trajectory on the fly - see the ANALYSIS + module of User Manual. + o Implement periodic boundary conditions for bonded interactions. + o Improved load balancing. + o Dynamic assignment of subset of nodes for performing the FFT + calculation. + o Split PME calculation now default - see pme keyword in MD block. + o Improved periodic boundary condition scheme. + o Exclude constraints between atoms fixed in space - improves reliability of + Shake algorithm. + o Remove the restriction that bonded interactions can involve atoms on at + most two segments. + o Improved printing SHAKE dev-too-large error message identifying the + atoms that are involved. + o BUG FIX: Fix J-Box or I-Box is Illegal problem. Prevent dimension + problems if box index list from previous run exceeds current dimension. + +Integral changes: + + o Inclusion of energy and gradient one-electron integrals from HONDO. + o Turn off SP integral code due to accuracy problems. + o Several fixes to the integral code with respect to memory and bugs in + TEXAS when high angular momentum and large exponents are present in + the basis set. + o Fix of ECP integral bug when the ECP contained a zero r-exponent. + o Routines for relativistic integral code (include?) + o Generates relativistic one-electron integrals (include?) + +Tests, manuals and misc script changes: + + o Updated User's Manual to reflect new and modified functionality in + NWChem. + o QA tests have been updated to reflect the new functionality in NWChem. + o Updated scripts to run NWChem in LoadLeveller and NQE and to test + results. + o Addition of some scripts to create animated gifs based on using RASMOL + and files printed out by NWChem during optimizations and frequency + calculations. + o Updates in the Programmer's Manual (much more is still left to come!). + +Installation availability: + +The current tested platforms and O/S versions are: + + o IBM SP with P2SC nodes, AIX 4.2.1, and PSSP 2.3 + o IBM SP with silver nodes (SMP nodes with two 604e processors), AIX + 4.3.2, and PSSP 3.1. + o IBM RS6000 workstation, AIX 3.2, 4.1. and 4.3 + o Cray T3E, 2.0.4.61 UNICOSMK + o SGI R8000/10000, IRIX 6.2, 6.5 + o SGI R4000, IRIX 5.3 + o SUN workstations, SunOS 4.1.3 and Solaris 5.5 + o Compaq DEC alpha workstion (600 MHz EV6), Digital UNIX V4.0E Rev. + 1091, DEC C V5.8-009, Digital Fortran V5.2 + o Linux. Since there are at least 8 popular distributions of the Linux + operating system and numerous others in existence, including + downloading everything and building your own Linux OS, it is impossible + to test all possible versions of Linux with NWChem. NWChem Release 3.3 + has been tested on Slackware 3.4, 3.5, 4.0, RedHat 5.1, 5.2, and 6.0, + Mandrake based on RedHat 6.0, and RedHat 6.0 for the Power PC + Macintosh. These all use the EGCS compilers at different levels. Those + distributed from Slackware are somewhat different than those distributed + from RedHat but the code is configured to run on all of them. The Linux + Alpha version 5.2 from Red Hat fails to compile the code as well as the beta + release of the Digital Fortran compiler for Linux based Alpha systems. + +NOTE:In releases of NWChem prior to 3.3 additional processes had to be +created on workstation clusters to support remote access to shared memory. +This is no longer the case. The TCGMSG process group file just needs to refer +to processes running NWChem. + +A general FAQ is available at +http://www.emsl.pnl.gov:2080/docs/nwchem/support/NWChem_FAQ.html + +A list of Known Bugs is available at +http://www.emsl.pnl.gov:2080/docs/nwchem/support/known_bugs.3.3.html