An adaptive distance-based group contribution method for thermodynamic property prediction.

Area of Science:

• Computational Chemistry
• Physical Chemistry
• Chemical Thermodynamics

Background:

• Accurate prediction of thermodynamic properties, specifically standard enthalpy of formation (Hf,298K), is crucial for understanding chemical reactions and molecular behavior.
• Existing methods like conventional group additivity (GA) may lack accuracy for complex, large hydrocarbon molecules.

Purpose of the Study:

• To develop an automatic and adaptive distance-based group contribution (DBGC) method for predicting thermodynamic properties of large hydrocarbon molecules.
• To compare the accuracy and cost-effectiveness of the DBGC method against conventional group additivity (GA) methods.

Main Methods:

• Developed a novel distance-based group contribution (DBGC) method characterizing group interactions using an exponential decay function based on group-to-group distance (number of bonds).
• Constructed a database of molecular bonding information and standard enthalpy of formation (Hf,298K) for alkanes, alkenes, and radicals using M06-2X/def2-TZVP//B3LYP/6-31G(d) computational methods.
• Employed multiple linear regression (MLR) and artificial neural network (ANN) for fitting group contributions and interaction parameters.

Main Results:

• The DBGC method demonstrated higher accuracy in predicting Hf,298K for alkanes compared to the conventional GA method when using identical training datasets.
• DBGC showed reduced discrepancies with literature data for highly branched large hydrocarbons, outperforming the GA method.
• The method maintained satisfactory overall accuracy when applied to alkenes and radicals.

Conclusions:

• The developed automatic and adaptive DBGC method provides an accurate and potentially inexpensive approach for predicting the standard enthalpy of formation of large hydrocarbon molecules.
• DBGC offers a significant improvement over traditional GA methods, especially for complex molecular structures.
• This method holds promise for broader applications in computational chemistry and thermodynamic property prediction.