Multilevel Fragment-Based Approach (MFBA): A Novel Hybrid Computational Method for the Study of Large Molecules.

We present a novel method for the calculation of large molecules and systems, the multilevel fragment-based approach. It is based on dividing the system into small fragments followed by separate calculations of these fragments and the interactions between them. Unlike previous fragmentation-based methods, we use multiple computational methods for the individual calculations. Using an accurate method only to calculate local interactions and more approximate methods for interactions over larger distances, it is possible to achieve results very close to a more demanding fragmented calculation using the higher level method only. The number of calculations performed at the higher level scales linearly with the size of the system, which significantly improves the efficiency and allows this scheme to be used for very large systems. In this work, we have combined density functional theory with the more approximate density functional tight binding method and applied this method to the calculation of model peptides. Formulation of first derivatives of the total energy within this fragmentation scheme is also presented and tested.