Assessment of the performance of tuned range-separated hybrid density functionals in predicting accurate quasiparticle spectra

Abstract

Long-range corrected hybrid functionals that employ a non-empirically tuned range-separation parameter have been demonstrated to yield accurate ionization potentials and fundamental gaps for a wide range of finite systems. Here, we address the question of whether this high level of accuracy is limited to the highest occupied / lowest unoccupied energy levels to which the range-separation parameter is tuned, or whether it is retained for the entire valence spectrum. We examine several π−conjugated molecules and find that orbitals of a different character and symmetry require significantly different range-separation parameters and fractions of exact exchange. This imbalanced treatment of orbitals of a different nature biases the resulting eigenvalue spectra. Thus, the existing schemes for the tuning of range-separated hybrid functionals, while providing for good agreement between the highest occupied energy level and the first ionization potential, do not achieve accuracy comparable to reliable G 0 W 0 computations for the entire quasi-particle spectrum.3

Length dependence of ionization potentials of transacetylenes: Internally consistent DFT/GWapproach

“…Thus local or semilocal DFT functionals that produce a too small gap compared to the real electronic gap would overestimate screening, whereas HF that produces a too large gap would underestimate it. As shown recently by Bruneval and Marques [40], input orbitals and energies derived from hybrid functionals with a high fraction of exact-exchange yield G 0 W 0 IPs that agree well with experiment for small organic molecules, whereas for larger molecules the fraction of exact exchange has to be considerably lower [41][42][43]. In other words, there is a well-known starting-point dependence of the G 0 W 0 approach [43][44][45], and the best DFT starting point is usually also system-dependent.…”

Fully self-consistentGWand quasiparticle self-consistentGWfor molecules

“…Regarding the applicability of self-consistent GW methods: On the one hand, our results could not prove that any of the explored self-consistent GW approaches is clearly superior to one-shot G 0 W 0 calculations using an appropriate starting point (e.g., Hartree-Fock and certain hybrid functionals have been shown to provide an excellent starting point for one-shot GW calculations 17,19,32,33,75,76 ); On the other hand, at least for the IPs of the set of atoms and molecules considered here, the self-consistent results seems to improve, although slightly, the G 0 W 0 -HF and we did not observe any clear signature that the self-consistent GW results were pathological. This is interesting because there are situation where one would like to improve the one-particle DFT spectra using a charge or energy conserving scheme.…”

Charge Delocalization in Oligomers of Poly(2,5-bis(3-alkylthiophene-2-yl)thieno[3,2-b]thiophene) (PBTTT)

“…32 The minimization proceeds in the following way: A first ω value is obtained based on the initial geometry; the geometry is then reoptimized with this ω value and a new ω value is optimized from the new geometry; this process is repeated until the ω value converges (within a threshold of 0.0001 Bohr -1 ), which usually occurs after 3-4 iteration steps. We note that earlier works [33][34][35] sometimes used geometries …”