Sunday, May 21, 2017

Solving the Density Functional Conundrum: Elimination of Systematic Errors To Derive Accurate Reaction Enthalpies of Complex Organic Reactions


Highlighted by Jan Jensen



Sengupta and Raghavachari present a quick and efficient way to increase the accuracy of computed reaction energies (ΔE).  For example, it is difficult to compute the reaction energy for Rxn1 because the bonding changes a lot: in effect, two double bonds are changed to 4 single bonds. By the same logic, it should be much easier to compute an accurate reaction energy for Rxn2.  

ΔE(Rxn1) = ΔE(Rxn2) - ΔE(Rxn3) 

So one should be able to get a good estimate of ΔE(Rxn1) by computing ΔE(Rxn2) and ΔE(Rxn3) at a relatively low and high level of theory, respectively. The accuracy can be further increased by larger fragments, either in Rxn3 or in an additional reaction.

Sengupta and Raghavachari test a four-reaction approach for 25 different reactions and a large variety of methods (DFT, HF, MP2, and CCSD(T)) and show that the mean absolute error relative to G4 can be reduced to ca 2 kcal/mol or less using the 6-311++G(3df,2p) basis set. For M06-2X they also tested the effect of basis set and showed that the MAE only increases from 2.2 to 2.6 kcal/mol on to the 6-31G(d) basis set.

Of course the high level calculations on the small fragments only have to be done once and a relatively small number of different fragments will be needed to cover most organic reactions.