Related Experiment Videos
PDDG/PM3 and PDDG/MNDO: improved semiempirical methods.
Matthew P Repasky1, Jayaraman Chandrasekhar, William L Jorgensen
1Department of Chemistry, Yale University, 225 Prospect St., New Haven, Connecticut 06520-8107, USA.
Journal of Computational Chemistry
|October 24, 2002
Summary
Two new computational chemistry methods, PDDG/PM3 and PDDG/MNDO, significantly improve heat of formation accuracy for organic molecules. These Pairwise Distance Directed Gaussian modifications offer better results than standard NDDO schemes and density functional theory.
Area of Science:
- Computational Chemistry
- Quantum Chemistry
- Method Development
Background:
- Standard semiempirical methods like NDDO schemes (PM5, PM3, AM1, MNDO) have limitations in accurately predicting heats of formation for organic compounds.
- Existing methods often struggle with specific chemical systems, including hydrocarbon isomers, small rings, and molecules with multiple heteroatoms.
Purpose of the Study:
- To develop and introduce two novel semiempirical methods, PDDG/PM3 and PDDG/MNDO, incorporating a Pairwise Distance Directed Gaussian modification.
- To enhance the accuracy of calculated heats of formation for molecules containing carbon, hydrogen, nitrogen, and oxygen atoms.
- To overcome limitations of existing semiempirical methods and provide more reliable energetic predictions.
Main Methods:
- Developed PDDG/PM3 and PDDG/MNDO by adding a Gaussian modification to existing core repulsion functions.
- Reparameterized semiempirical parameters and modified the computation of atomic energies of formation.
- Introduced functional group information via pairwise atomic interactions using atom-based parameters.
Main Results:
- Achieved significant improvements in the accuracy of heats of formation for 622 diverse molecules.
- Reduced mean absolute errors from 4.4 to 3.2 kcal/mol for PM3 and from 8.4 to 5.2 kcal/mol for MNDO.
- Demonstrated superior performance over standard NDDO schemes, PM5, and density functional theory (DFT) with large basis sets for heats of formation.
- Showed substantial improvement in PDDG isomerization energies compared to B3LYP/6-31G* results.
- Overcame specific challenges related to hydrocarbon isomer stability and energetics of small rings and heteroatom-rich molecules.
Conclusions:
- The PDDG modification offers a straightforward enhancement to existing semiempirical software.
- PDDG/PM3 and PDDG/MNDO provide more accurate and internally consistent energetic predictions, including heats of formation and isomerization energies.
- These new methods represent a significant advancement in semiempirical quantum chemistry for organic systems.