TURBOMOLE manual
v8.0

1.4 New defaults since Turbomole V8.0

The new version V8.0, which was introduced in 2025, has brought a whole series of old standards up to date. In recent years, the use of Turbomole has gradually evolved from a tool for experts who have dealt in detail with individual systems and various methods to a commonly used package for many applications, which are also carried out in large numbers by occasional quantum chemistry users.

In particular, the integration into workflows and scripting environments such as Python (e.g. using iPython or Jupyter Notebooks) has influenced the use of Turbomole in general.

Various defaults changed to avoid tedious additional entries according to more modern standards and use cases:

  • define generates a shorter control file with just the most important keywords and using the simplified assignment of basis sets without the need to specify atom numbers or element names

  • The def2-SV(P) basis set and accompanying ECPs are now the default for all elements up to Rn. Beyond Rn, the def-SV(P) basis set is still the default.

  • The default DFT functional is now PBE (BP86 for pre-V8.0 versions)

  • The default grid size is 3 instead of m3. Time spent for DFT functional evaluation is for most use cases not a time limiting factor any more and usage of m-grids, initially introduced to speed up calculations, are seen as not accurate enough for many tasks like NMR or EPR properties

  • The default method for the COSMO solvation model is changed to Gaussian Charge Model (GCM) using Gaussian charges on Lebedev grids, as those are more accurate for analytical gradients, for geometry optimizations and 2nd derivatives like vibrational frequencies. To use the former default, a new option for $cosmo was introduced, adding a line with the content klamt will switch back to the former cavity default and the usage of point charges.

  • Weight derivatives are included by default in DFT gradient calculations with grad and rdgrad as well as in evib calculations. This ensures consistency with aoforce.

  • The re-use of input settings using a simple and generic input template for the control file without the need to run define now eases the usage of Turbomole. The modules ridft, dscf and riper automatically import missing auxiliary basis sets form the basis set libraries that are shipped with Turbomole, and also generate start orbitals using either superposition of atomic densities ($atomdens) or an extended Hückel guess ($eht). If no orbitals or occupation are given, the superposition of atomic densities is used by default.

  • Hartree–Fock calculations in dscf and ridft apply the convergence criterion for the density matrix change ($denconv) to allow for subsequent MP2 and CC calculations without further adjustments.

  • One-component calculations in mpshift use the current-dependent generalization for meta-GGAs by default. This affects NMR, EPR, pNMR, and ZFS calculations.

  • RT-TDDFT calculations in riper also use the generalized kinetic-energy density for meta-GGAs by default.

  • A new data group ($relham) was introduced to simplify the input for relativistic all-electron calculations.

  • The generalized DIIS approach ($gdiis) is now activated by default in two-component calculations.

  • The keyword $rpaconv now controls the convergence threshold for response calculations in all DFT response modules (escf, egrad, aoforce, mpshift) which ensures consistency for NMR parameters such as coupling constants and chemical shifts. The module-specific keywords $forceconv and $shiftconv overwrite the general $rpaconv settings for aoforce and mpshift, respectively.

  • ridft calculations now longer require an explicit inclusion of $rij and $marij in the control file but set it automatically. Likewise, the jbas is assigned automatically for a given basis set.

  • The initial guess with a superposition of atomic density is carried out at the same level of theory as the SCF/DFT calculation. That is, the efficient approximations are applied and it is now the default guess for dscf, ridft, and riper. By default, the partial charges of the atoms in a molecule are obtained with the xTB model.

  • riper automatically determines the occupation and performs an initial guess if no orbital or bands input is given.

  • Geometry optimizations with relax consider the root mean square of the gradient as a further criterion for convergence. Default is \(10^{-4}\) in atomic units.