Hammett σ from composition predicts SN2 barriers
A data-enhanced Hammett model maps SN2 barrier heights across substituent and leaving-group space using additive σ values that depend only on group identity and distance to the reacting carbon.
Source
Data enhanced Hammett-equation: reaction barriers in chemical space
What they did
The authors compressed quantum SN2 potential-energy changes into reaction constants ρ and substituent constants σ, tested additivity versus resonance, compared Theil–Sen averaging with the classical one-reference Hammett fit, and benchmarked against kernel ridge regression.
What they found
Substituent effects are largely additive without resonance; SN2 σ values follow composition and distance to the reaction center (6 parameters vs 20 in a dummy encoding). Their robust fit beats a single-reference Hammett model (errors ≤3.8 vs up to 5.2 kcal mol−1).
The limits
What it doesn't show
The linear model cannot capture strong resonance coupling of substituents, and barriers are electronic-structure estimates rather than experimental condensed-phase rates.
Key terms
- Hammett equation
- Linear free-energy relation log(K/K0)=ρσ separating reaction and substituent effects.
- σ
- Substituent constant encoding electron-donating or -withdrawing power.
- ρ
- Reaction constant measuring sensitivity of the reaction to substituents.
- Theil–Sen regressor
- Robust linear fit used to avoid overfitting a single reference reaction.
- KRR
- Kernel ridge regression ML baseline for the same SN2 energies.
Flashcards
Research intelligence for this paper
See its role on concept claims, tensions it is part of, placement history, and related discoveries.
Quiz yourself
Hammett σ in this SN2 space depends mainly on:
Common questions
How many SN2 reactions were treated?
12 combinations of nucleophile and leaving group.
What substituents sat at R1–R4?
–H, –NO2, –CN, –NH3, –CH3.
How many parameters describe molecular σ with distance decay?
NG+1 = 6.
Worst-case MAE vs original Hammett?
3.8 vs up to 5.2 kcal mol−1.
More on Computational chemistry