Atom Pair Contribution Method: Fast and General Procedure To Predict Molecular Formation Enthalpies.

Area of Science:

• Computational Chemistry
• Physical Chemistry
• Chemical Thermodynamics

Background:

• Predicting molecular formation enthalpies is crucial for understanding chemical reactions and stability.
• Existing methods like bond contribution (BC) and group contribution (GC) have limitations in accuracy and applicability.
• High-throughput screening requires computationally inexpensive yet reliable predictive models.

Purpose of the Study:

• To develop and validate a novel atom pair contribution (APC) model for predicting gas-phase formation enthalpies (Δ_fH°).
• To establish an APC model that is broadly applicable and requires minimal computational resources.
• To demonstrate the superior accuracy of the APC model compared to existing computational methods.

Main Methods:

• Developed an atom pair contribution (APC) model using increments for bonded and geminal atom pairs.
• Incorporated 15 specific structural correction terms into the APC model.
• Compiled and utilized a large dataset of 2671 experimental and high-level theoretical data points for model training and validation.

Main Results:

• The APC model demonstrates high reliability in predicting gas-phase formation enthalpies.
• Achieved wide applicability with only 68 adjustable parameters, outperforming GC methods.
• Outperformed quantitative structure-property relationship (QSPR) methods, semiempirical Hamiltonians, and low-level DFT approaches.

Conclusions:

• The developed APC model is a valuable tool for rapid screening of molecular properties.
• It provides a computationally efficient and accurate alternative for predicting formation enthalpies.
• The model's simplicity and reliability make it suitable for applications where chemical accuracy is not paramount.