We investigate theoretically light- and bias-induced metastabilities in (CIGS) based solar cells, suggesting the Se–Cu divacancy complex as the source of this hitherto puzzling phenomena. Due to its amphoteric nature, the complex is able to convert by persistent carrier capture or emission from a shallow donor into a shallow acceptor configuration, and vice versa, thereby changing in a metastable fashion the local net acceptor density inside the CIGS absorber of the solar cell, e.g., a CdS/CIGS heterojunction. In order to establish a comprehensive picture of metastability caused by the complex, we determine defect formation energies from first-principles calculations, employ numerical simulations of equilibrium defect thermodynamics, and develop a model for the transition dynamics after creation of a metastable nonequilibrium state. We find that the complex can account for the light-induced metastabilities, i.e., the “red” and “blue” illumination effects, as well as for the reverse-bias effect. Thus, our model implies that the different metastabilities observed in CIGS share a common origin. A defect state in the band gap caused by in the acceptor configuration creates a potentially detrimental recombination center and may contribute to the saturation of the open circuit voltage in larger-gap alloys with higher Ga content. Therefore, the presence of metastable defects should be regarded as a concern for solar cell performance.
REFERENCES
In principle, neither the experimental nor the LDA equilibrium lattice constant (usually smaller than the real lattice constant) represents an optimal situation. The choice of the LDA (equilibrium) lattice constant implies an overestimated overlap and, hence, interaction between defect and host orbitals, whereas the choice of the experimental (nonequilibrium) lattice constant implies the presence of some hydrostatic pressure acting on the lattice in the calculation. Fortunately, there are usually only minor differences in the defect formation and transition energies. Since the large lattice relaxation of the anion vacancies in II-VI compounds is particularly sensitive to the lattice constant (Ref. 20) the differences can be more pronounced, in the case of anion vacancies, however. For CIS and CGS, we confirmed by additional calculations that changes by not more than when using the LDA lattice constant.
We determined from fitting an effective-mass-like density of states (degeneracy factor of 2) to the numerical density of states, calculated in LDA including spin-orbit coupling. The obtained value is close to determined experimentally in Ref. 37.
In Ref. 20, we determined from the difference of the single-particle energy of the orbitals in the defect calculation relative to energy in the pure host. The present method appears to yield more consistent results and less uncontrolled scatter compared to the former method. Accordingly, the potential alignment for the complex determined here differs by up to from that determined in Ref. 20. The most significant change is that we now find a shallow state of the acceptor configuration of , similar to the shallow state of the isolated (see Table I). Before, we found a deeper state (Ref. 20) for the complex. Further, the deep acceptor levels of the isolated now appear as closely spaced, but separate and levels, whereas we found a weak negative- behavior and a transition before. (Ref. 20).
The energy barriers , , and are actually determined for the isolated . Since energies of the and defect levels and the mechanism of III-III bond formation/breakup that leads to the barriers is hardly affected upon complex formation with , the respective energies for are expected to be very similar.
For the electron capture, Eq. (2), the empty level needs to be activated to energies below the CBM. Similarly, for the hole capture, Eq. (3), the occupied level must be activated to energies above the VBM. This happens only for some fraction of the time needed for a full oscillation around the local minima in the CCD [Figs. 3(b) and 5(b)].
