CBH-BDC Enhanced Δ-ML for Predicting the Accurate Standard Enthalpy of Formation

Area of Science:

• Physical Chemistry
• Computational Chemistry
• Chemical Informatics

Background:

• Standard enthalpy of formation (Δ_fH°) is crucial for understanding physicochemical processes.
• Existing methods like CBH-BDC have limitations in high-accuracy electron energy calculations and parameter scope.
• Accurate prediction of Δ_fH° is vital for various chemical applications.

Purpose of the Study:

• To develop an enhanced delta machine learning (Δ-ML) approach for accurate prediction of Δ_fH°.
• To overcome the limitations of previous CBH-BDC methods.
• To enable high-accuracy Δ_fH° predictions without computationally expensive quantum calculations.

Main Methods:

• Introduced a CBH-BDC enhanced Δ-ML approach.
• Utilized effective and interpretable molecular descriptors derived from CBH and BDC.
• Validated the approach using 464 species with experimental Δ_fH° data.

Main Results:

• The Δ-ML approach accurately predicts Δ_fH°, bypassing high-level quantum calculations.
• Successfully extrapolated Δ_fH° from DFT to CCSD(T) accuracy for over 120,000 molecules.
• Demonstrated significant improvements in prediction accuracy compared to previous methods.

Conclusions:

• The CBH-BDC enhanced Δ-ML approach provides accurate and efficient prediction of Δ_fH°.
• This method facilitates the creation of high-quality Δ_fH° databases for chemical deep learning.
• The approach offers a valuable tool for computational chemistry and materials science.