Benchmarking Isomerization Energies for C 5 $$ {\mathrm{C}}_5 $$ - C 7 $$ {\mathrm{C}}_7 $$ Hydrocarbons: The ISOC7...

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Materials Science

Background:

• Accurate benchmark databases are essential for developing robust and efficient electronic structure methods.
• Existing databases may not adequately cover the diversity of molecular structures needed for method validation.

Purpose of the Study:

• To introduce the ISOC7 database, a comprehensive collection of C5-C7 hydrocarbon constitutional isomers.
• To provide high-accuracy reference isomerization energies for these isomers.
• To benchmark the performance of various computational methods using this new database.

Main Methods:

• Generation of 1308 unique constitutional isomers of C5-C7 saturated and unsaturated hydrocarbons.
• Calculation of reference isomerization energies using the CCSD(T)/CBS level of theory via the W1-F12 protocol.
• Benchmarking of 40 Density Functional Theory (DFT) functionals, semiempirical methods, tight-binding methods, and machine learning potentials.

Main Results:

• The ISOC7 database spans 146 kcal/mol in isomerization energies.
• DFT performance generally improves with higher rungs of Jacob's Ladder, with ωB97M-D4 being the top performer (RMSD 1.62 kcal/mol).
• The g-xTB tight-binding method achieved an RMSD of 4.14 kcal/mol, while the AIMNet2 machine learning potential showed exceptional accuracy (RMSD 1.67 kcal/mol), rivaling DFT at lower cost.

Conclusions:

• The ISOC7 database serves as a rigorous benchmark for computational methods.
• Modern DFT functionals and machine learning potentials demonstrate significant accuracy for hydrocarbon isomer energies.
• The database facilitates the advancement and validation of both quantum chemical and machine learning approaches in electronic structure theory.