Fields, Local Reactivity and Sterics

Electronic and Steric Descriptors

Compare three-dimensional fields and local descriptors for reaction selectivity in terms of definition, input, interpretation, and limitations.

1. Choosing a descriptor

Select descriptors for the chemical hypothesis being tested, not simply because they are available. Steric shape, electrostatics, orbital interaction, density deformation, and local reactivity are distinct concepts.

DescriptorPrimary informationKey caution
Electron density ρ(r)Electron distribution and molecular boundaryNonnegative; distinct from orbital amplitude
ESPElectrostatic potential for a test chargeFix the evaluation surface, units, and sign
MO / LUMO fieldOrbital phase and local amplitudeMatch orbitals and global phase
Fukui / dual descriptorLocal response to electron-number changesDefine charge state, spin, and relaxation
RDG and sign(λ2)ρLow-density, low-gradient regionsNot an interaction energy
IGMH and IRIInterfragment, bonded, and nonbonded regionsDepends on fragments, isovalue, and grid
Sterimol and %VburDirectional steric occupancyDepends on axis, center, radius, and conformer

2. Electron density and electrostatic potential

Electron density describes electron distribution, while ESP is the potential produced by nuclei and electrons. Compare ESP on the same density isosurface or spatial points and fix units and color range.

3. Orbitals and frontier fields

Local HOMO or LUMO amplitude can suggest orbital-interaction regions, but orbital ordering may change between calculations. Match by energy, symmetry, and overlap and align the global phase before field comparison.

4. Fukui functions and the dual descriptor

Δf(r)=f+(r)-f-(r)

Finite-difference densities and frontier-orbital approximations are not identical. Keep geometry, spin, occupations, and grids consistent across N, N+1, and N-1 systems and state the sign convention.

5. NCI, IGMH, and IRI

These real-space functions visualize interaction regions. Do not convert surface presence, area, or color directly into an interaction energy; combine them with transition-state energies, EDA, substituent effects, or experiments.

6. Conversion to machine-learning features

  1. Align molecules in a common coordinate system.
  2. Use identical grid origin, axes, shape, and spacing.
  3. Mask or integrate chemically relevant regions.
  4. Preserve scale and sign.
  5. Control dimensionality with PCA, regularization, and grouped validation.

Continue to cube-file definitions

Last reviewed: August 4, 2026