Remnant PbI₂, an unforeseen necessity in high-efficiency hybrid perovskite-based solar cells?^a)

With the global growth in energy demand and with compelling climate-related environmental concerns, alternatives to the use of non-renewable and noxious fossil fuels are needed.¹ One such alternative energy resource, and arguably the only legitimate long-term solution, is solar energy. Photovoltaic devices which are capable of converting the photon flux to electricity are one such device.² Over the last 2 years, halide hybrid perovskite-based solar cells with high efficiency have engendered enormous interest in the photovoltaic community.^3,4 Among the perovskite choices, methylammonium lead iodide (MAPbI₃) has become the archetypal light absorber. Recently, however, Sn-based perovskites have been successfully implemented in functional solar cells.^5,6 MAPbI₃ is an attractive light absorber due to its extraordinary absorption coefficient of 1.5 × 10⁴ cm⁻¹ at 550 nm;⁷ it would take roughly 1 μm of material to absorb 99% of the flux at 550 nm. Furthermore, with a band gap of 1.55 eV (800 nm), assuming an external quantum efficiency of 90%, a maximum current density of ca. 23 mA/cm² is attainable with MAPbI₃.

Recent reports have commented on the variability in device performance as a function of perovskite layer fabrication.⁸ In our laboratory, we too have observed that seemingly identical films have markedly different device performance. For example, when our films of PbI₂ are exposed to MAI for several seconds (ca. 60 s), then a light brown colored film is obtained rather than the black color commonly observed for bulk MAPbI₃ (see Sec. S2 of the supplementary material for the optical band gap of bulk MAPbI₃).²³ This brown color suggests only partial conversion to MAPbI₃ and yields solar cells exhibiting a J_sc of 13.4 mA/cm² and a V_oc of 960 mV; these values are significantly below the 21.3 mA/cm² and 1000 mV obtained by others.⁴ Under the hypothesis that fully converted films will achieve optimal light harvesting efficiency, we increased the conversion time from seconds to 2 h. Unexpectedly, the 2-h dipping device did not show an improved photovoltaic response (J_sc = 9.7 mA/cm², V_oc = 846 mV) even though conversion to MAPbI₃ appeared to be complete. With the only obvious difference between these two devices being the dipping time, we hypothesized that the degree of conversion of PbI₂ to the MAPbI₃ perovskite is an important parameter in obtaining optimal device performance. We thus set out to understand the correlation between the method of fabrication of the MAPbI₃ layer, the precise chemical compositions, and both the physical and photo-physical properties of the film. We report here that remnant PbI₂ is crucial in forming a barrier layer to electron interception/recombination leading to optimized J_sc and V_oc in these hybrid perovskite-based solar cells.

We constructed perovskite-containing devices using a two-step deposition method according to a reported procedure with some modifications.⁴ (see Sec. S1 of the supplementary material for the experimental details).²³ MAPbI₃-containing photo-anodes were made by varying the dipping time of the PbI₂-coated photo-anode in MAI solution. In order to minimize the effects from unforeseen variables, care was taken to ensure that all films were prepared in an identical manner. The compositions of final MAPbI₃-containing films were monitored by X-ray diffraction (XRD). Independently of the dipping times, only the β-phase of the MAPbI₃ is formed (Figure 1).⁹ However, in addition to the β-phase, all films also showed the presence of unconverted PbI₂ (Figure 1, marked with *) which can be most easily observed via the (001) and (003) reflections at 2θ = 12.56° and 38.54° respectively. As the dipping time is increased, the intensities of PbI₂ reflections decrease with a concomitant increase in the MAPbI₃ intensities. In addition to the decrease in peak intensities of PbI₂, the peak width increases as the dipping time increases indicating that the size of the PbI₂ crystallites is decreasing, as expected, and the converse is observed for the MAPbI₃ reflections. This observation suggests that the conversion process begins from the surface of the PbI₂ crystallites and proceeds toward the center where the crystallite domain size of the MAPbI₃ phase increases and that of PbI₂ diminishes. Interestingly, the remnant PbI₂ phase can be seen in the data of other reports, but has not been identified as a primary source of variability in cell performance.^8,10

Considering that the perovskite is the primary light absorber within the device, we wanted to further investigate how the optical absorption of the film changes with increasing dipping time (Figure 2).^11,12 The pure PbI₂ film shows a band gap of 2.40 eV, consistent with the yellow color of PbI₂. As the PbI₂ film is gradually converted to the perovskite, the band gap is progressively shifted toward 1.60 eV. The deviation of MAPbI₃'s band gap (1.60 eV) from that of the bulk MAPbI₃ material (1.55 eV) could be explained by quantum confinement effects related with the sizes of TiO₂ and MAPbI₃ crystallites and their interfacial interaction.^13,14 Interestingly, we also noticed the presence of a second absorption in the light absorber layer, in which the gap gradually red shifts from 1.90 eV to 1.50 eV as the PbI₂ concentration is decreased from 9.5% to 0.3% (Figure 2—blue arrow).

Having established the chemical compositions and optical properties of the light absorber films, we proceeded to examine the photo-physical responses of the corresponding functional devices in order to determine how the remnant PbI₂ affects device performance. The pure PbI₂ based device remarkably achieved a 0.4% efficiency with a J_sc of 2.1 mA/cm² and a V_oc of 564 mV (Figure 3(a)). Upon progressive conversion of the PbI₂ layer to MAPbI₃, we observe two different regions (Figure 4, Table I). In the first region, the expected behavior is observed; as more PbI₂ is converted to MAPbI₃, the trend is toward higher photovoltaic efficiency, due both to J_sc and V_oc, until 1.7% PbI₂ is reached. The increase in J_sc is attributable, at least in part, to increasing absorption of light by the perovskite. We speculate that progressive elimination of PbI₂, present as a layer between TiO₂ and the perovskite, also leads to higher net yields for electron injection into TiO₂ and therefore, higher J values. For a sufficiently thick PbI₂ spacer layer, electron injection would occur in stepwise fashion, i.e., perovskite →PbI₂ →TiO₂. Finally, the photovoltage increase is attributable to the positive shift in TiO₂'s quasi-Fermi level as the population of photo-injected electrons is higher with increased concentration of MAPbI₃.

The second region yields a notably different trend; surprisingly, below a concentration of 2% PbI₂, J_sc, V_oc, and ultimately η decrease. Considering that the light-harvesting efficiency would increase when the remaining 2% PbI₂ is converted to MAPbI₃ (albeit to only a small degree), then the remnant PbI₂ must have some other role. We posit that remnant PbI₂ serves to inhibit detrimental electron-transfer processes (Figure 5). Two such processes are back electron transfer from TiO₂ to holes in the valence band of the perovskite (charge-recombination) or to the holes in the HOMO of the HTM (charge-interception). This retardation of electron interception/recombination observation is reminiscent of the behavior of atomic layer deposited Al₂O₃/ZrO₂ layers that have been employed in dye-sensitized solar cells.^15–18 

It is conceivable that the conversion of PbI₂ to MAPbI₃ occurs from the solution interface toward the TiO₂/PbI₂ interface and thus would leave sandwiched between TiO₂ and MAPbI₃ a blocking layer of PbI₂ that inhibits charge-interception/recombination. For this hypothesis to be correct, it is crucial that the conduction-band-edge energy (E_cb) of the PbI₂ be higher than the E_cb of the TiO₂.^19–21 The work function of PbI₂ was measured by ultraviolet photoelectron spectroscopy (UPS) and was observed to be at 6.35 eV vs. vacuum level, which is 0.9 eV lower than the valence-band-edge energy (E_vb) of MAPbI₃ (see Sec. S7 of the supplementary material²³ for the work function of PbI₂); the E_cb (4.05 eV) was calculated by subtracting the work function from the band gap (2.30 eV). The E_cb of PbI₂ is 0.26 eV higher than the E_cb of TiO₂ and thus PbI₂ satisfies the conditions of a charge-recombination/interception barrier layer.

In order to probe the hypothesis that PbI₂ acts as a charge-interception barrier, dark current measurements, in which electrons flow from TiO₂ to the HOMO of the HTM, were made. Consistent with our hypothesis, Figure 6 illustrates that the onset of the dark current occurs at lower potentials as the PbI₂ concentration decreases. In the absence of other effects, the increasing dark current with increasing fraction of perovskite (and decreasing fraction of PbI₂) should result in progressively lower open-circuit photovoltages. Instead, the photocurrent density and the open-circuit photovoltage both increase, at least until to PbI₂ fraction reaches 1.7%. As discussed above, thinning of a PbI₂-based sandwich layer should lead to higher net injection yields, but excessive thinning would diminish the effectiveness of PbI₂ as a barrier layer for back electron transfer reactions.

Given the surprising role of remnant PbI₂ in these devices, we further probed the two-step conversion process by using scanning-electron microscopy (SEM) (Figure 7). Two domains of lead-containing materials (PbI₂ and MAPbI₃) are present. The first domain is sited within the mesoporous TiO₂ network (area 1) while the second grows on top of the network (area 2). Area 2 initially contains 200 nm crystals. As the dipping time is increased, the crystals show marked changes in size and morphology. The formation of bigger perovskite crystals is likely the result of the thermodynamically driven Ostwald ripening process, i.e., smaller perovskite crystals dissolves and re-deposits onto larger perovskite crystals.²² The rate of charge-interception, as measured via dark current, is proportional to the contact area between the perovskite and the HTM. Thus, the eventual formation of large, high-aspect-ratio crystals, as shown in Figure 7, may well lead to increases in contact area and thereby contributes to the dark-current in Figure 6. Regardless, we found that the formation of large perovskite crystals greatly decreased our success rate in constructing high-functioning, non-shorting solar cells.

In summary, residual PbI₂ appears to play an important role in boosting overall efficiencies for CH₃NH₃PbI₃-containing photovoltaics. PbI₂'s role appears to be that of a TiO₂-supported blocking layer, thereby slowing rates of electron(TiO₂)/hole(perovskite) recombination, as well as decreasing rates of electron interception by the hole-transporting material. Optimal performance for energy conversion is observed when ca. 98% of the initially present PbI₂ has been converted to the perovskite. Conversion to this extent requires about 15 min. Pushing beyond 98% (and beyond 15 min of reaction time) diminishes cell performance and diminishes the success rate in constructing non-shorting cells. The latter problem is evidently a consequence of conversion of small and more-or-less uniformly packed perovskite crystallites to larger, poorly packed crystallites of varying shape and size. Finally, the essential, but previously unrecognized, role played by remnant PbI₂ provides an additional explanation for why cells prepared dissolving and then depositing pre-formed CH₃NH₃PbI₃ generally under-perform those prepared via the intermediacy of PbI₂.

We thank Prof. Tobin Marks for use of the solar simulator and EQE measurement system. Electron microscopy was done at the Electron Probe Instrumentation Center (EPIC) at Northwestern University. Ultraviolet Photoemission Spectroscopy was done at the Keck Interdisciplinary Surface Science facility (Keck-II) at Northwestern University. This research was supported as part of the ANSER Center, an Energy Frontier Research Center funded by the U.S Department of Energy, Office of Science, Office of Basic Energy Sciences, under Award No. DE-SC0001059.