5. Skill: Significantly reducing cost of IGMH visual analysis for interfragment interactions

(Note: The strategy described in this section is available since Multiwfn updated on 2025-Nov-24.)

Most people use the IGMH method primarily for visualization and analysis of interfragment interactions. In practice, this only requires calculating the \(\delta g^\text{inter}\) and \(\text{sign}(\lambda_2)\rho\) grid data in the overlapping regions between fragments; grid points in all non-overlapping regions are completely unnecessary. Taking this into account, a strategy is supported in Multiwfn to significantly reducing the cost of grid data calculation in IGMH (and mIGM, IGM). The implementation is as follows: A parameter named IGMvdwscl can be found in settings.ini file. The surface of each fragment is defined as the union of atomic spheres, where vdW radius of each atom is scaled by IGMvdwscl. During the grid data calculation step of IGMH, mIGM, or IGM analysis, a grid point is computed only if it lies within the overlapping region between the surfaces of two or more fragments; otherwise, it is directly skipped. With this strategy, the time required to compute IGMH grid data is often reduced by several times or more. The exact percentage of time saved depends on how large a portion of the user-defined grid box consists of ignorable grid points.

Obviously, after enabling this strategy, you can only plot isosurfaces of \(\delta g^\text{inter}\). Isosurfaces of \(\delta g\) or \(\delta g^\text{intra}\) can no longer be plotted, because only the \(\delta g^\text{inter}\) isosurfaces are guaranteed to appear exclusively in the interfragment overlap regions.

IGMvdwscl defaults to 0, which means the time-saving strategy is disabled. To enable this strategy, set IGMvdwscl to an appropriate positive value. The larger the value, the weaker the time-saving effect; the smaller the value, the higher the risk of truncating the \(\delta g^\text{inter}\) isosurfaces. A value of 2.0 is generally recommended: it is sufficiently safe while still providing a significant reduction in computational time.

This strategy does not affect cost and result of calculation of \(\delta G^\text{atom}\) and \(\delta G^\text{pair}\) indices.

This strategy is more detailedly described in my blog article http://sobereva.com/756, and an example is given there.