# Implementation¶

The basic philosophy of the ADF QM/MM implementation has been to treat the MM region as a perturbation to the QM model system. The atoms of the QM model system are controlled by ADF as they would in a regular pure QM run, whereas the MM atoms are treated ‘separately’. For example, when geometry optimization is performed, the Hessian matrix is only generated / updated for the QM model system. This is possible because at each geometry step the MM subsystem is fully optimized (with the QM and LI atoms frozen). One should be aware however that there are not necessarily six zero eigenvalues of the Hessian, as the QM system is coupled to the MM system (and therefore not free to rotate/translate !). When internal coordinates are used to define the structure of the complex, the atoms of the QM model system are optimized using the internal coordinate system, whereas the MM subsystem is fully optimized at each step within the Cartesian coordinate system. Figure 1-6 shows the flow control of the ADF QM/MM implementation.

The geometry optimization of the MM subsystem is controlled separately from the QM system. For example, the default convergence criteria for the MM subsystem are far stricter than that of the QM system. Furthermore, there are options to use global optimization techniques on the MM subsystem, see the detailed description of input options later in this manual.

Figure 1-6. Schematic representation of the ADF QM/MM implementation