Input file for CPL: TAPE21

In order to run the CPL code, you need the general ADF output file TAPE21 being present in the directory where CPL is running. Most of the computation's specific settings will be taken from TAPE21, such as the integration accuracy, the basis set, the density functional being employed, nuclear coordinates, and so on. That also means that nearly all of the aspects that affect the quality of CPL's results are already determined in the input for the ADF run. Four aspects are of particular importance here:

1. The numerical integration accuracy: the perturbation operators are large in the vicinity of the nuclei. Therefore, you have to make sure that the integration grid is fine enough in the atomic core regions. We have found that INTEGRATION 6 in the ADF input yields high enough integration accuracy for the CPL code in most cases. In case you can not afford such a high integration accuracy throughout, we suggest the use of the INTEGRATION block key to assure that the integration parameter equals 6 at least in the atomic core regions:

INTEGRATION
 ACCINT 4.0 :: or higher
 ACCSPH 6.0
END

This should yield a reasonable integration precision for many applications. We do not recommend to use an ACCINT parameter smaller than 4 for obtaining meaningful results and encourage to use higher settings whenever possible.

2.The basis set: NSSCCs are sensitive chemical probes, and therefore flexible basis sets have to be employed in order to yield a valid description of the MOs that determine the NSSCCs. We have found that it is imperative to use at least basis set TZ2P (V) from the ADF basis set database. Additional polarization functions in the valence shell may be necessary. Furthermore, the FC perturbation usually requires additional steep 1s functions (i.e. with exponents much higher than the nuclear charge) for a proper description. In the relativistic heavy element case, the use of additional steep basis functions as compared to the ZORA/TZ2P basis is mandatory. The use of steep functions is only of high importance for those nuclei, for which the NSSCC is to be evaluated.
Currently, no standard basis set suitable for NSSCCs has been added to the ADF basis set directory. We suggest to use basis TZ2P as a starting point and to add some 1s basis functions (and appropriate fit functions) with higher exponents in order to improve the accuracy of the FC term. For some of the heavy NMR nuclei (W, Pt, Hg, Tl, Pb), basis sets suitable for the computation of NSSCCs will be added to the ADF basis set directory in the near future. See Refs. .
For the nuclei for which NSSCCs are to be evaluated, it is necessary to use all-electron basis sets. This is not a restriction due to the implementation, but we have found that, with the available frozen core basis sets, the flexibility of the basis in the vicinity of the nuclei is not sufficient. It is possible to use frozen core basis sets if you add enough basis functions in the core region such that the basis approaches the flexibility of at least a double-zeta all-electron basis there [1]. In that sense, the savings in computational time due to usage of a frozen core basis are not as pronounced as in standard ADF computations. Unless reliable frozen-core basis sets for the NSSCC computation are available we strongly discourage the use of frozen core basis sets with the CPL program!

3.The density functional: the results of the CPL code depend mostly on the shape of the MOs that have been determined by ADF, and their orbital energies. Both, in turn, depend on the density functional or Kohn-Sham potential that has been chosen for the ADF run (and the basis set quality). It is difficult to give a general advice here concerning the NSSCCs. So far we have found that the use of GGAs improves the NSSCCs with respect to experiment in most cases in comparison to LDA. Different GGAs often yield very similar results. Further, in particular for those cases for which the OP term is large or even dominant, both standard LDAs and GGAs sometimes do not provide an accurate enough description of the orbitals, and deviation of the CPL results as compared to experiment can be substantial. Future developments of density functionals might be able to cure these problems. For the time being, we recommend that you base the CPL run on different choices of density functionals in the ADF run, and investigate the convergence of the result with respect to basis set and integration accuracy. Note that CPL itself uses the VWN functional by default to determine the first-order perturbed MOs. There are enough indications to believe that this is a reasonable approximation for NMR purposes.

4.Modeling the experimental setup: computing such sensitive numbers as NMR chemical shifts and in particular NSSCCs can result in substantial deviations from experimental data. The simple reason might be that the isolated system that has been computed at zero temperature is not at all a good approximation to the system that has been studied experimentally. We [3,4] and other authors have found that in particular solvent effects can contribute very substantially to the NSSCC. In case you are comparing CPL results to experimental data obtained in strongly coordinating solvents we suggest that you consider solvent effects as a major influence. We have found that even weakly coordinating solvents can cause sizeable effects on the NSSCCs for coordinatively unsaturated metal complexes. Other sources of errors can be the neglect of vibrational corrections to the NSSCCs (usually in the range of a few percent).

If the parameters of the underlying ADF computation are carefully chosen and the density functional is able to provide an accurate description of the molecule under investigation, it is possible to compute NSSCCs by means of DFT with very satisfactory accuracy (please note that for properties as sensitive as NSSCCs, agreement with experimental results within about 10% error can be regarded as quite good). Further, chemical trends will be correctly reproduced for a related series of molecules in most cases. However, due to the inherent approximate character of the density functionals currently available with ADF, and necessary basis set limitations, great care should be taken that the results are reliable. CPL assumes Aufbau configurations. Please make sure that there are no empty orbitals with energies below the highest occupied MO (HOMO). In addition, the SYMMETRY NOSYM key has to be used in the ADF computation. It is currently not possible to use dummy atoms in the ADF input if the TAPE21 is intended to be used for a subsequent CPL computation.