In some cases the atoms selected for geometry constraints, or a partial Hessian were incorrect. In the calculation a different selection was used.
The gui mixed up the tests for the A and G tensor, making it practically impossible to run either.
In a job with a lattice with a slightly asymmtric lattice the program failed to generetate the integration points for the Brillouin zone.
For big systems, the occupation key produced invalid inputs (due to lines that were too long), fixed.
In parallel calculations, small differences on different nodes could result in non-convergence, or incorrect results. These differences could occur in the construction of the basis vectors in the Davidson algorithm for the TDDFT calculation of excitation energies. So far only seen in case of hybrid functionals. This has been solved.
The option to hide atoms in ADFinput did not work any more, fixed.
The option to change the size and colors of atom info did not work any more, fixed.
The expert option to use the occupation key has been repaired.
When disabling the "Frequencies%useA1Displacments" key, the program could fail with a restart problem. Also this option should almost always be combined with the disabled "Frequencies%doA1Projection", which can now be set by the GUI.
The A-DIIS scheme has been extended to support relativistic Spin-Orbit SCF. Besides, it is now possible to set two threshold values for A-DIIS, see below.
SCF ADIIS THRESH1=a1 THRESH2=a2 END
Here, a1 and a2 (a1 > a2) correspond to values of the maximum element of the [F,P] commutator matrix, ErrMax. If ErrMax ≥ a1, only A-DIIS coefficients are used to determine the next Fock matrix. If ErrMax < a2 then only SDIIS coefficients are used. For ErrMax between a2 and a1 the total DIIS coefficients are calculated from SDIIS and A-DIIS values weighted proportionally according to the ErrMax value. Thus, the weight of A-DIIS coefficients decreases with the ErrMax value.
The default values for THRESH1 and THRESH2 are 0.01 and 0.0001, respectively.
For the dispersion correced PBEsol functional the scaling factor is 1.42. In the code this should include the conversion factor from Angstrom to bohr.
On Windows, when changing the overall text size in the system, the menu with the periodic table was not readable any more.
The A-DIIS scheme by X. Hu and W. Yang, J. Chem Phys 132, 054109 (2010) has been implemented. According to preliminary tests, this scheme performs much better than Energy-DIIS and it does not require calulation of the total energy. Therefore, it is also much faster than E-DIIS. To use the method just specify ADIIS in the SCF input block:
SCF ADIIS END
Possibly incorrect SCF density and spinors in case of spin-orbit coupling, especially in case of symmetry NOSYM. This has been fixed. The bug was in the densf program.
The gradients were wrong with relativistic calculations, due to an incorrect optimization.
Now GENNBO can handle up to 998 atoms and 9998 basis functions (up from 200 and 2000 respectively).
When OpenBabel was used to determine bonds it failed when you had atom types like C.1 and so on.
For example, ADFview would show strange bonds if the molecule had not been set up with ADFinput.
In case of Z-matrix input for the atoms the first 3 atoms in the input were not always given a correct COSMO radius.
The XC kernel could produce NaN's in case of zero α-density and non-zero β-density in an integration point. This is fixed.
This particular bug could happen for large systems with many symmetry operators. (This is what you can get when using a large super cell)
The RamanRange and ScanFreq features will now reset symmetry to AUTO before starting the frequency scan. Besides, the symmetry tolerance will be reduced to 10-6 so that small displacements are treated correctly. The later is especially important for large flat molecules. The symmetry tolerance value can be changed using a TOL= argument of the corresponding keyword, RAMANRANGE or SCANFREQ, respectively. Note an important difference with the TOL parameter of the SYMMETRY keyword: the latter is valid for the calculation preceding the frequency scan while the value used in the RAMANRANGE/SCANFREQ keyword is valid during the frequency scan.
The subopions RSCALE and S6 were ignored when specified by the user
Distance labels (from the geometry info popup) in ADFview and other GUI modules did not sometimes not work, fixed
The program tries when possible to plot from old runkf files. This is not always possible, and may fail.
With the latest change, it has become possible to optimize very large systems using delocalized coordinates.
The spin orbit implementation was wrong for Slater type functions with an s character. In principle this affects all spin-orbit calculations, but in practice this caused only (severe) errors in all electron calculations at the spin-orbit level.
A bug fix was fixed for the perturbative calculation of spin-orbit coupled excitation energies (keyword SOPERT) in case of the model exchange-correlation potential SAOP.
The nuclear gradient of the core kinetic energy density was calculated incorrectly. This has a small effect on the nuclear gradients of Meta GGAs.
When doing a geometry optimization with dispersion correction a huge amount of incr was printed
ADFspectra: MCD C term was not read in correctly in case of NODIRECT option, this is fixed.
There was a memory leak in MCD calculations, which showed up in large calculations.
For band it is hard to handle slightly asymmetric inputs. The GUI produced in some cases slightly inaccurate coordinates. Also the program BAND has been made a bit more tolerant for such a situation.
With ADF2009 gradients are stored differently on the result tape. Now adfreport uses Gradients_CART if available, and falls back to the Gradients variable.
This means that if you perform a single point run with gradients (using the Gradients key), adfreport can now find the gardients on the result file.
STCONTRIB can now be used to analyze spin-orbit coupled excitations in case of hybrids and TDA. STCONTRIB still can not be used for spin-orbit coupled excitations in case of hybrids if one does not use TDA.
The analysis of the temperature dependent C term contained a bug that is now fixed.
The scaling of the nuclear gradients has been much improved. In an example with 70 atoms in the unit cell the overall calculation was three times faster.
The routine that tries to find the high symmetry paths through the BZ sometimes crashed when the symmetry of the BZ was lower than that of the pure lattice (without atoms).
Due to a typo, you could not create excitation calculations with specified number of frequencies.
Numerical issue solved for TPSS in case of a very small spin-polarization.
When you renamed a job, the .results directory (if present) was not renamed. As the files in the .results directory are not used by the ADF-GUI this is only a problem when users are inspecting the files in that directory.
In some rare cases it is possible that the ADF_DSYEVX routine enters an infinite loop. This behavior has been fixed
Changed the default DIMIX parameter from 1.0 to 0.2, and NCYCLEDAMP from 5 to 1
Better handling of string constants and allow an extra data_global variable
An option has been introduced (and is now the default) to read and write temporary files in single precision. For most calculations this means that roughly half of the IO needs to be done. For true benchmark calculations, double precision can be enforced with the key
Programmer FlioSinglePrecision false End
In some occasions the program asked after a single point calculation to read in the new coordinates. In fact the order of the atoms was changed by band. This is now detected.
When saving a picture in BANDgui, the program was hanging.
When using adfview on the result of a (band) geometry optimiziation, the coordinates were wrong.
Search:
Snapshots:
