Benchmark Test of Three-Layered ONIOM Method

Benchmark Test of Three-Layered ONIOM Method 

ONIOM combination3 Estimated Costb Atoms in the model system only All atoms in the real system  

One notes that the results of QM MM MM depend sensitively on the choice of the charges used in the Amber calculation. The use of the Mulliken charges, for instance, in QM MM MM increases the error from —47 kcal/mol with RESP charges to -86 kcal/mol. This implies that the results will also depend sensitively on how to arbitrarily scale the near-border charges for the QM-MM interaction, because the problematic QM-MM boundary is very close to the reaction center. 

To summarize, we have systematically tested all possible three- and two-layer ONIOM combinations of high-level QM (HQ=B3LYP/6-31G ), low-level QM (LQ=AM1), and MM (Amber) for the deprotonation energy and structure of a test molecule, an ionic form of a peptide. We find the errors introduced in the ONIOM approximation, in comparison with the target HQ (or HQ HQ HQ) calculation, generally increases in the order  

1 There is no well-defined way of calculating deprotonation energy with Amber. However, this does not matter as it totally cancels out in ONIOM by taking the difference between the real and model system. 

---