Filippo De Angelis1 2

1, CNR-ISTM, Perugia, , Italy
2, Istituto Italiano di Tecnologia, Genova, , Italy

The astonishing performance of lead-halide perovskites in optoelectronic devices is due to a unique combination of highly efficient generation, transport and collection of charge carriers. The initial stage of charge generation, i.e. exciton dissociation, has attracted a huge interest. Experimental estimates of the exciton binding energy (Eb) in MAPbI3 lie in the 6-25 meV range, implying an efficient exciton dissociation at room temperature. Despite its pivotal importance, the physical mechanisms beyond efficient charge separation in lead-halide perovskites are still largely unknown and no quantitative models of how the various dielectric contributions (electronic, vibrational, pseudo-rotational) affect Eb in lead-halide perovskites have been reported.
Here, we investigate the factors which influence the exciton binding energy in a series of metal-halide perovskites using accurate first-principles calculations coupled to ab initio molecular dynamics simulations. By introducing a novel computational approach to self-consistently incorporate the effect of phonon screening into a full relativistic Bethe Salpeter equation framework, we effectively simulate the electronic and vibrational contributions affecting the exciton binding energy for the parent MAPbI3 compound. By calculating Eb for a few snapshots along an ab initio MD trajectory we verify that dynamic disorder does not significantly affect the exciton binding energy. We also use a phenomenological model to approximately account for exciton screening due to MA cations rotation, finding that this contribution is irrelevant for models delivering realistic values of the static dielectric constant. We show, on the other hand, that Eb is strongly modulated by screening from low-frequency phonons, which accounts for a factor 2 reduction in the converged Eb value. Our phonon-screened, full relativistic estimate of Eb for MAPbI3 amounts to 15 meV, in agreement with current experimental estimates. Based on this background, we then explore how different material combinations (e.g. Pb -> Pb:Sn-> Sn; I -> Br; MA -> FA -> Cs) can be optimized to tune the value of Eb.