deletions | additions       diff --git a/subsection_QM_NMR_calculations_on__.tex b/subsection_QM_NMR_calculations_on__.tex  index 02a6389..4a3b394 100644  --- a/subsection_QM_NMR_calculations_on__.tex  +++ b/subsection_QM_NMR_calculations_on__.tex 
...
\subsection{QM \subsection{DFT  NMR calculations and protein structures used}  In this paper we benchmark the NMR chemical shift predictions  on full proteins} ubiqutin and Protein G. The structures are geometry optimized using PM6-D3H+ using the PCM solvation model and the CHARMM22/CMAP force field using the xx solvation model using the 1UBQ and 2. The PM6-D3H+ optimizations are done using the GAMESS program and a convergence criterion of 5x10^{-4} atomic units, while the CHARMM22/CMAP optimizations  In the following sections ProCS14 is benchmarked on QM NMR calculations on full protein models. The proteins are a crystal structure of ubiquitin(1UBQ) with 76 residues and two models of Protein G, 2OED\cite{2oed} which is a crystal structure refined with dipolar couplings and 1IGD\cite{1igd} the crystal structure of protein G. The protonation states of LYS, ARG, HIS, GLU and ASP is if necessary changed to the same as in ProCS14. The QM NMR calculations is done on both the unmodified crystal structures and optimized versions. PM6 was used in order to get an optimized structure with approximately the same bond angles and bond lengths as  the tripeptides. The PM6-D3H+\cite{dh3plus} with a PCM model was used and PM6-DH+\cite{pm6dh} with the COSMO solvation model. Pure PM6 was not able to converge most likely because it lacks the dispersion and hydrogen bonding corrections of PM6-D3H+ and PM6-DH. In addition to PM6 the proteins was also optimized with a selection of forcefields, AMBER\cite{amber}, CHARMM22/CMAP\cite{CHARMM2009} and AMOEBAPRO13\cite{AMOEBA}. The NMR calculations were done at the GIAO OPBE/6-31G(d,p) PCM level of theory in Gaussian 09.