forked from jeffhammond/nwchem-git-svn-deprecated
-
Notifications
You must be signed in to change notification settings - Fork 0
/
release.notes.3.3
313 lines (266 loc) · 14.1 KB
/
release.notes.3.3
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
NWChem Version 3.3 and 3.3.1 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 and 3.3.1. When there
is a conflictbetween the release notes and the User Manual, the User
Manual takes precedence.
The changes for version 3.3.1 are listed at the bottom of this file.
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
List of Release 3.3.1 changes in the source and binary download files:
o Fix for message passing with Linux. This fixes a known
performance problem.
o Modifications in the GA tools which were causing performance
problems between workstations.
o Fix for the passing of character strings in the PSPW code.
o Fix for the T3E makefile so that the make works cleanly.
o Fix tools makefiles to allow for larger than 2 GB files on the
SP and to fix a diagonalization problem which was causing problems
with the scalability of DFT.
o Work arounds to LAPI problems that have been identified on
the SP. The symptoms generally showed up as jobs that
would hang.
o The periodic boundary conditions for an MD run now checks
for images more than one cell away.
o Fix in the PREPARE module so that some solvents that were
not being written out correctly in the PDB file now are.
o Additional printing of information in the ecce.out file to
be compatible with Ecce 1.5.
o Slightly looser value of EPREC in DRIVER for DFT
calculations. (5e-6 instead of 1e-7)
o Improved screening on the grid for DFT which only effects large
calculations (usually those that have molecules that are "spread
out").
o Dramatically improved parallel efficiency of the Lagrange
multipliers and the non-local pseudopotentials on high-latency
parallel platforms in the PSPW code.
o Improved persistence of tolerance changes when linear
dependence is found. This had caused some convergence
problems.
o The TIGHT option has replaced AOTOL and MOTOL in MP2 (undocumented
feature of version 3.3).
o Fix in the CCSD calculation to add some synchronization which
provided more stability.
o The examples directory has been updated and the tests
directory has been deleted because the input files were
obsolete.