## Available bug-fix 2014 binaries

Operating System MPI SVN revision # diffs Date Link

## Bug-fixes

### r49617: fixed a typo in the ReaxFF NaH forcefield file

The NaH forcefield contained a bug in the angle definitions. The Na-H-Na angle had the wrong atom numbers, causing them to overwrite the H-Na-Na angle. The correction is based on the text of the corresponding article.

### r49393: bugfix band and adfdos. Restart DOS option did not work properly

Band would stop with an error message. Also adfdos would show an empty window.

### r49134: Bugfix reading a Hessian from restart in geometry optimization

This fix should solve problems with non-converging transition state optimizations when using an exact Hessian matrix. The problem occurs only when the input and the internal order of atoms are different

### r49110: Bugfix restarting from a vels file

There was a problem reading atomic labels when restarting from a vels file which could lead to wrong bond energies for some atoms, for example due to the missing correction for double C-O bonds because it is done by checking element names.

Work around if you run into the bug: select all, copy, delete, paste, and continue

### r49072: Bugfix reaxff with non-orthogonal unit cells and non-reactive reax

Resolve an issue that may lead to loss of energy when non-reactive reaxff is used with a non-orthogonal unit cell.

### r49004: Fix a shared memory leak for asymmetric MPI configurations

This fix resolves an issue when shared memory is leaked in MPI jobs where some nodes have one and others more than one process allocated to them.

### r48739: bugfix dftb. Often in periodic calculations problems with diagonalization.

This could either happen during the SCC part, or at the very end when calculating the interpolated bands.

### r48659: bugfix Band: linear symmetry detected for non-linear systems

This would happen when the last atoms keyword contained only one atom. The results would simply be very wrong.

### r48258: Reaxff: Resolve perfomance issue for stretched unit cells where only one or two cell dimensions are small.

The sparse linear solver will now be used in such cases instead of the dense one

### r48248: bugfix band: crash in calculating critical bond path (QTAIM)

This release fixes documentation search and potentially slow startup on mac and linux (when domain lookups for the license fails).

### r48166: Fix documentation search in ADFinput

Mac and linux SVN rev r48101, Windows SVN rev r48093

### r48083: fix ADFinput Builder search error

IntelMPI 5.0 makes mpiexec.hydra.exe the default way of launching MPI jobs, which also adds the possibility to run with -bootstrap fork, removing the need for a daemon. However, the mpiexec.hydra.exe causes pmi_proxy.exe process to pop up in a commandline window. To prevent the popup we switched back to the old mpiexec.smpd.exe launcher.

### r48072: Fix AnalyticalFreq memory usage issue for very large molecules with frozen core

When running analytical frequency calculations ADF keeps some precalculated data in memory. For large molecules with frozen core this may lead to very high memory usage because of two arrays whose size scales as Gridsize*Natoms. With this fix, if ADF determines that these arrays together will occupy more than 1000 MB (100 MB for 32-bit platforms) it will switch to a slower algorithm that uses less memory. The threshold value can be changed using the B1SIZE key in the AnalyticalFreq block, which specifies the number of MB.

### r48015: Updated IntelMPI to 5.1 for 64bit Windows

This made it possible to change the mpiexec command on 64bit Windows to use -bootstrap fork, causing it no longer need the IntelMPI daemon.

This will hopefully solve the FP exception problem on old systems (SUSE 10 and alike)

### r47848: adfview now leaves atoms as they are

In periodic systems determination of the cross-cell bonds by ADFview would map atoms into some cell, sometimes different from the original cell.

### r47833: Bugfix in Densf with spin-unrestricted excitations

The fix that should allow visualization of properties using a t21 file coming from a spin-unrestricted excitations calculation. Visualization of transition densities is still not possible but other properties should work now.

### r47782: SCM_NODOMAINCHECK removed, SCM_DOMAINCHECK added

By default no domain check is performed. If you have a license that requires it you need to turn it on by explicitly setting SCM_DOMAINCHECK to yes. SCM_NODOMAINCHECK is not used any more, so it does not harm if it is defined.

### r47459: Old QMMM in ADF: The mm.restart file is now saved every print_cycles iterations also during a CG optimization.

Before this change the intermediate mm.restart file was saved only during quasi-Newton optimizations.

### r47139: The GUI now reads the lattice vectors from the History section if present on a rkf file

When your DFTB calculation did finish with an error, the lattice vectors were not updated in the final resuls on the RKF file. The history section does contain those vectors, so use these instead.

This will only make a difference when a periodic geometry optimization including optimization of the lattice runs into an error with DFTB

### r46955: bugfix band: program would stop optimizing a geometry with a meta GGA with this error

ERROR DETECTED: NucGradVOCEvaluatorModule: assumed fully allocated array, implement sparseness

### r46953: bugfix step size during geometry optimization

During geometry optimization it is sometimes possible that a relatively small (within the trust radius) step in delocalized coordinates results in large atomic displacements. This has been fixed now by checking the changes in Cartesian coordinates before taking a step.

### r46660: Suppress expired license messages for GUI and Utils

They were causing trouble if your license file contained some expired entries and you were using adfprep.

### r46604: Simplified interface for isotopic shift calculations in ADF

It is now possible to calculate an IR isotopic shift without modifying the atom type. The alternative mass is specified using the IsotopicShift keyword as follows:

IsotopicShift n1=m1 n2=m2 ...

Here, n1, n2, etc. are sequential numbers of atoms in the input geometry and m1, m2, etc. are their alternative masses. All frequencies and intensities are calculated as usual with the original atomic masses taken from the corresponding fragment files and the isotopic shifts are calculated and printed afterwards. The Hessian is not broken down into irreducible representations and the rigid motions are not projected out prior to the normal mode analysis. Normal modes (eigenvectors) corresponding to the original and the shifted frequencies are matched against each other based on their ovelap, which allows to calculate an isotopic shift for each frequency. This analysis may, however, fail for highly symmetric molecules with degenerate modes.

This feature currently works in conjunction with analytical frequencies only.

### r46545: ADF analytical frequencies with frozen core should use considerably less memory.

During analytical frequency calculations with frozen core the derivatives of the frozen core density w.r.t. atomic coordinates are precalculated and stored in shared memory as sparse grid function. For large calculations (> 150 atoms) that would lead to unacceptably large memory usage because a very safe sparsity cutoff threshold of 1e-15 was used. Tests show that using a 1e-8 cutoff yields essentially the same frequencies (abs deviation less than 1e-3 1/cm) while reducing the memory usage (for the test molecules) by factor 2 to 3. Thus the default value was changed to 1e-8 and a possibility was added to set this parameter in the input using a B1THRESH keyword in the AnalyticalFreq block.

The Gradients suboption AccurateNucGradKin did not work properly. By default it is off, so it does not affect standard calculations. The resulting gradients are more accurate when using a frozen core.

### r46459: bugfix band: prevent crash in DOS routine on windows

Especially on windows band would sometimes crash with a stack overflow problem. The stack usage has been reduced in the DOS routines.

### r46378: bugfix BANDstructure: click on band shows orbital in ADFview, but the wrong orbtal was shown

In ADFview the information shown is (and was) correct. However, bandstructure switched the K point and band index around, so the orbital shown is not were the user clicked in the band structure.

### r46365: upgrade default OpenMPI version to 1.8.x

Since the linking libraries have changed with for the OpenMPI 1.8 family, we now increase the default linking string to be 1.8 compatible for Intel builds on Linux. Note: if you run an older version of OpenMPI, see Install/x86_64_linux/intel/Makeflags for how to change the linking.

### r46264: bug fix for the CIF file reader

Potential problem after importing a CIF file: atoms too close, because too many atoms are read.

### r45774: fix job monitoring for windows

The ps -p command used on windows is not working, switch to ps -e.

### r45765: fix job monitoring for windows

The mon.cmd script was using /dev/null which does not exist on windows

### r45761: minor fixes for scmjobd (GUI job monitoring process)

It is more stable for remote jobs while still working for local jobs, avoiding displaying old local results when remote results are available.

### r45752: reduced disk usage for MetaGGA in band

When using a meta gga the program would store all the basis functions on disk. Now it will no longer do that, just as is the case for a normal GGA.

### r45745: Bugfix in TD-DFTB

There was a bug in the TD-DFTB implementation that lead to incorrectly calculated excitations for molecules with more than 64 atoms. The problem has been fixed in r45745.

### r45732: script for monitoring running jobs updated for stability reasons

In ADF2014 the scripts are only used for remote job monitoring.

### r45702: fix for band: workaround to be able to set the old k-space method

Using the line kspace 5, would be ignored. The way to get this is now

kspace 7 &
Grid Symmetric
End


NB: the ampersand matters.

### r45680: fix minor license details

In some cases messages that a license is needed were incorrectly displayed. The GUI would continue to work just fine, when a valid license was available. The incorrect messages have been fixed.

If SCM_NODOMAINCHECK has been set via the GUI Preferences it is now honored in the GetLicense dialog, avoiding delays in some cases.

The timeout values contacting the SCM license server from the GetLicense dialog have been increased for backup servers.

Support for http proxy servers has been added to the GetLicense dialog: for windows it should pick it out of the registry, for other systems the environment variable http_proxy is used, optionally with http_proxy_user and http_proxy_pass.

### r45651: Increment timeout values for autolicense when switching to backup servers

The timeout starts at 10 seconds, and is doubled when switching to backup servers (twice). You can also adjust the timeout value by settings SCM_AUTOLICENSE_TIMEOUT to the number of milliseconds.

Set SCM_LOG_AUTOLICENSE in the environment to log connection issues of the autolicense script to the specified file.

### r45574: bugfix ADFinput for BAND Kspace1/2/3 parameters

When you set explicit K space values (K-Space 1/2/3), it is saved properly and the run script is correct. If you next open the .adf file again, the K-Space 1/2/3 values would incorrectly be cleared. After saving they are also no longer in the run script. This bug has been solved (now they stay as they are, and the run script is as it should be).

### r45548: bugfix band, stop with problem in ModifiedEulerianPath

In some cases the code to generate an automatic path through the Brillouin Zone fails. The program used to stop but now at least continues and uses in such cases a very primitive path.

### r45509: bugfix band NMR. Prevent crashes by reducing the memory and disk usage

When doing NMR calculations with band on larger systems, with thousands of basis functions, the program would crash in PREPARENMRHAM. The reason would be the use of too much memory or disk. To get the old code set NMR%UseSharedMemory to false.

### r45430: Change MOPAC start script to contain local libraries

Include some required local libraries of MOPAC explicitly

## r45386: 2014.05 released

### r45303: Fixed a bug in parallel COSMO.

In same rare occasions it is possible that the COSMO conjugate gradient solver diverges on different MPI ranks. Up to now the bug was only found to lead to a “solve_for_u: equations are singular” error during analytical frequencies calculations on the Intel Sandy Bridge-E processors but it could possibly also lead to erroneous results in other cases. The adf2013 and earlier versions are not affected by this issue.

### r45294: Enable ScaLAPACK on ifort versions 14 and 15.

When building from source using ifort versions 14 and 15 ScaLAPACK will be enabled just like for ifort v12 before.

### r45267: Bugfix ADF COSMO geometry optimization parallel

Calculation could hang after convergence.

### r45182: Remove XQuartz tests for older MacOSX versions

Seems like the tests were causing problems even if XQuartz was installed correctly

### r44974: Multiple reaxff improvements

Fix restart issue for non-reactive reaxff: the valence and torsion angles are now searched for at each step instead of just at the first step. This should make sure that restarting a non-reactive MD will produce the same result as continuation of the original run (save for round off errors). Note: non-reactive MD simulation may produce a different trajectory after the fix compared to the earlier versions.

Solve rare problem when reaxff would hang during a parallel MD simulation.

Pdb atom names (only when read from a pdb input file) and residue information is now also copied over to output.pdb file.

Added a FORCES section to the trainset.in file. See examples/reaxff/water_ffield for an example.

### r44941: Bugfix band run script: KSpace VeryGood should not contain a space between Very and Good

This problem only shows when you select a KSpace VeryGood with a NumericalQuality that is not VeryGood.

### r44837: bugfix bandinput, lattice constant of Perovskite wrong

The value was 5.0 Angstrom and should be 3.84.

### r44786: Bugfix: adfreport showing TOC with -h flag depended on position of -h flag

The only visible result is that the TOC of a file was missing if you asked for it via the -h flag: adfreport -h filename

### r44648: Bugfix for band. Gradients wrong for spin-orbit calculations

The gradients were wrong when using the spin-orbit option. This bug was not present in the fix2013 version, and caused by the performance improvements.

### r44653: bugfix ADF: ZlmFit icw Alternative elements J

ZlmFit now compatible with user-defined elements J, which are elements can have arbitrary nuclear charge.

Showing the Run Script window for a running UFF job did update the wrong window

### r44613: bugfix for band, 5f frozen cores not supported

For some very heavy elements such as Uut, the program would use a 5d frozen core, whereas there is also a 5f frozen core available.

### r44457: bugfix when running without network connection

When no network connection is present, you will get delays due to the license checking of your domain name. This check can be avoided by setting SCM_NODOMAINCHECK.

In the 2014 version the SCM_NODOMAINCHECK feature did not avoid all domain checks as intended, that has been fixed.

### r44448: bugfix for band NMR, sometimes NaN for tensors.

Sometimes the printed NMR shieldings were NaN.

### r44421: Non-reactive ReaxFF bug fixed.

Fixed an issue with non-reactive reaxff that would cause a molecule to break up if its atoms appear in different unit cells at the start of a calculation

### r44414: bugfix band. SCF errors with the TB-mBJ model potential.

Due to tiny negative values of the kinetic energy density the calculation would stop in the SCF with an error in the diagonalization routine.

### r44385: bugfix ADFmovie: handling SDF file with non-conformers would fail

WHen handling an SDF file with different molecules (thus not different conformations of the same molecule), the bonds were not updated correctly.

### r44276: bugfix ADFinput: running MOPAC in ADFinput for periodic systems would mess up the bonds

Currently the bonds are not updated with the mopac bond order results for periodic systems.

### r44258: improved database for Perovskites in bandinput

If the third element was not O then the elements would be wrong (two would by oxygen). Also added CsPbCl3 and CsPbI3.

### r44237: bugfix ADFinput TD-DFTB intensity selection

The XC pull-down menu on the XC details page for BAND in ADFinput was showing the ADF XC menu by mistake. The identical XC menu on the BAND main page was (and is) correct.

### r44199: bugfix ADFinput for ReaxFF restarts

Restart did only work when using a relative path for the restart file. Now it may also be an absolute path.

### r44192: bugfix for spin-unrestricted TB-mBJ calculations with band.

In the original paper spin was not discussed. In a subsequent paper a spin-polarized form was specified, different from what was implemented in band.

### r44179: Added the KTB-mBJ and JTS-mTB-mBJ model potentials to band.

The TB-mBJ model potential is good at predicting band gaps. It can be thought of as depending on three parameters. The newly added functionals use different values for the three parameters. The KTB-mBJ is a reparametrized version of the original functional, based on a larger fit set. The JTS-mTB-mBJ potential has been specifically optimized for the band gap of Perovskites. See the XC documentation for details.

### r44109: Bugfix for band, program might stop in subroutine expbz

For two dimensional systems band would sometimes stop in the subroutin expbz. This could happen after the SCF near BZSTRUCT.

### r44044: Bugfix for band. Gradients wrong for spin-unrestricted

The gradients were wrong when using the UNRESTRICTED keyword.

### r43985: Performance fix for TB-mBJ model potential calculations with band

When using the TB-mBJ model potential, the program would store the bloch basis and its gradient on disk, even when the directBas option was on.

### r43980: Bugfix for band. ElectronicTemperature did not work

The key Convergence%ElectronicTemperature was broken. Repaired, and now also works for 1D and 3D. It used to work only for 2D. It still does not work for molecules.

### r43972: Bugfix for band. Frozen core too large while calculating the band structure (BZSTRUCT)

For systems where the frozen core error was close to the criterion, the program could stop when it is almost finished in the part where it calculates the band structure. Now the program will not check the frozen core in this part of the code.

### r43948: 3ob-2-1 parameter set added to DFTB

DFTB now contains the 3ob-2-1 parameter set, which allows DFTB3 calculations for organic compounds containing sulfur and phosphorus.

### r43933: Bugfix for band. Dispersion energy wrong

For periodic systems the dispersion energy was wrong. The gradients however were correct. So when using dispersion correction the bonding energy would be incorrect, the geometry would be fine.

### r43924: Improved support for dispersion corrections in DFTB

We have improved support for dispersion corrections in DFTB: All dispersion corrections now support both molecular and periodic systems and the ULG dispersion correction developed by Kim et al. has been added. Note that Grimme’s D3 with zero damping is no longer supported, as it was not fully parametrized for DFTB, and that the default parameters for the D3-BJ dispersion correction have been updated, which will produce slightly different results. See below for an updated documentation of DFTB’s dispersion keyword.

Dispersion

This key allows to specify options for the London dispersion correction. If it does not exist, no dispersion correction will be included.

method UFF|ULG|D2|D3-BJ

This subkey is used to specify a dispersion model. Please refer to the literature for details on the different methods:

• UFF: L. Zhechkov et al., J. Chem. Theory Comput., 2005, 1 (5), pp 841-847
• ULG: H. Kim et al., J. Phys. Chem. Lett., 2012, 3 (3), pp 360-363
• D2: S. Grimme, J. Comput. Chem., 2006, 27: 1787-1799
• D3-BJ: S. Grimme et al., J. Comput. Chem., 2011, 32: 1456-1465

If no method specified in the Dispersion block, a default dispersion correction defined in the metainfo.yaml file (depends on “ResourcesDir”, see metainfo.yaml) will be used. Note that not all dispersion models are supported by all parameter sets. A list of supported models can be found in the parameter’s metainfo.yaml file.

d3parameters s6=S6Param s8=S8Param a1=A1Param a2=A2Param

User-customized D3-BJ parameters can be specified with this subkey. Otherwise, the default parameters defined in metainfo.yaml are used. Note that all four parameters have to be specified if the d3parameters keyword is used.

### r43870: Fix impi.dll issue in the makelicinfo

When the empirical UFF dispersion was used, the gradients were incorrect. Typically one would see at the end of a geometry optimization the energy going up. Does not affect single point calculations.

### r43847: Bugfix ADFinput angle slider

Changing the angle between atoms by selecting three atoms and using the slider sometimes failed.

Reading a new molecule via the -m flag from an .adf file using adfprep was broken after r43728. This has been fixed.

With the latest update of MacOSX (10.9.5) Apple changed the requirements to get a properly signed application. The ADF-GUI.app has been updated to match these requirements.

The dmg file on the main Download and SnapShot pages will be signed properly (starting with ADF2014.02).

### r43736: bugfix band: nr. of symmetry equivalant atom sets wrong

For some very specific systems (with lots of symmetry) the program would incorrectly determine the nr. of sets of symmetry equivalent atoms. As a result the calculation would go very wrong or simply crash.

In some cases atom references that occur in the input (for example the linear transit atoms, or atoms for which to calculate NMR shifts, etc) could be wrong after saving. They would also be wrong in the run script. This would only occur after making changes to your molecule (reordering or deleting atoms).

As the fix changes the format of the .adf file a new version has been introduced (2014.02).

### r43722: ADFinput bugfix: Dihedral LT run script could be wrong

The atoms defining the dihedral to be used in a LT run could be wrong, and as a result the wrong dihedral would be used.

Now the Quality setting also adjusts the KSpace value and the (Soft)-Confinement setting. At the Quality details page you can adjust them individually as you like.

### r43673: bugfix band: printed band gap wrong when using spin-orbit coupling

The program would always print -no band gap-, even in case that there was a gap.