Physical Organic Chemistry and Data Analysis

Reaction Selectivity Analysis

Connect product and rate ratios to free-energy differences while treating conformers, transition states, and statistical models under explicit assumptions.

1. Rate ratios, product ratios, and free energies

When two pathways follow the same kinetic framework, their rate ratio is related to the activation-free-energy difference.

k1k2=exp-ΔΔGRT

Treat ee as a fraction and fix which product or reaction face defines the sign. Yield also reflects side reactions and recovery and is not automatically a rate ratio.

2. Curtin-Hammett conditions

If reactant conformers interconvert much faster than product formation, selectivity depends on the free energies of the transition states relative to the equilibrating ensemble. Slow interconversion requires an explicit kinetic model.

Check the assumptions

Do not map a product ratio to one barrier difference without considering equilibrium, reversibility, the rate-determining process, temperature, solvent, and catalyst speciation.

3. Conformer and transition-state ensembles

Multiple pathways leading to one outcome can be combined into an effective free energy.

Geff=-RTlniNexp-GiRT

Record conformer search, deduplication, energy window, theory level, and standard state, and test sensitivity to more than one nominal minimum.

4. Site and facial selectivity

Site selectivity compares reaction positions, facial selectivity compares faces at one center, and chemoselectivity compares functional groups or reaction modes. Define the complete competing pathway set and denominator.

5. Data-driven models

  1. Define the reaction unit and target.
  2. Generate conformers, transition states, and fields consistently.
  3. Fix descriptor units and coordinate orientation.
  4. Design grouped splits across substrates or reaction series.
  5. Control overfitting with regularization and nested validation.
  6. Compare interpretation with independent computation or experiment.

6. Reporting checklist

  • Ratio definition, sign, temperature, and units
  • Product assignment and measurement uncertainty
  • Conformer and pathway coverage
  • Electronic-structure and thermal-correction settings
  • Data split, preprocessing, and metrics
  • Applicability domain and external validation

Continue to regularized regression, PCA, and SHAP

Last reviewed: August 4, 2026