Calculating the NMR Chemical Shielding of Large Molecules

Abstract

This thesis examines three approximations that significantly reduce the computational time of theoretical NMR shielding calculations for large molecules, whilst largely retaining the accuracy of the parent method: fragmentation, locally dense basis sets and composite methods.

For fragmentation it is established that Level 4 fragments reliably reproduce full molecule shieldings, when hydrogen bonds are treated as single bonds, and long range through space corrections are incorporated through the McConnell equation and background charges.

The pcS-n basis set family is demonstrated to converge more rapidly towards the basis set limit than all other examined families. Furthermore, it is established that this limit is consistent with convergence towards experimental values.

A systematic investigation of locally dense basis sets established that a group based partitioning of the pcS-4, pcS-2 and pcS-1 basis sets, augmented with through space allocations, allowed the shielding to be produced within chemical accuracy for a variety of compounds.

Finally, composite methods utilising a variety of levels of theory were systematically investigated, and it was found that a double composite method combining the HF, MP2 and CCSD(T) levels of theory and the pcS-4, pcS-2 and pcS-1 basis sets yielded NMR shieldings that were within chemical accuracy of CCSD(T)/pcS-4 calculations, themselves having converged closely to experimental values.

When considered in combination this work represents a significant step towards achieving chemical accuracy for protein NMR shielding calculations.