lda_x_erf is actually lda_x_erfc
Related to #360 (closed).
@fstein93 writes
I have carried out calculations (CP2K trunk) with the same functional with the implementation of P. Gori-Giorgi (CP2K 8.1). Unfortunately, I am not able to reproduce my numbers which I have obtained with the implementation for a short-ranged LDA functional by PGG. With LibXC, I employ LDA_X-LDA_X_ERF+LDA_C_PW-LDA_C_PMGB06 with the same values of the range-separation parameter. As a simple example, I try to calculate the BSSE-corrected dissociation energy of He2 with short-ranged LDA+long-ranged HF exchange+long-ranged MP2 (see 10.1103/PhysRevA.72.012510 ). Our implementation of HF and MP2 has not changed significantly. At an interatomic distance of 3.0 A, I obtain with an aug-cc-DZ basis set the following BSSE-corrected binding energies (in Hartree), short-ranged DFT contribution, HF+MP2 contribution before: -9.5e-06 | -2.63E-04 | -8.79E-05 now: -2.58e-04 | -3.78E-05 | -8.36E-05 The sum of other energy contributions deviate less than 5.0E-05 from each other. Their implementation is unchanged since then.
Thus, the large differences are related to the short-ranged DFT part. My question: Have I misunderstood the implementation of short-ranged LDA with LibXC?