Prospects for inorganic CsPbI₃ perovskite solar cells with commercially viable lifetimes

In the past decade, metal-halide perovskite semiconductors have generated enormous interest in the photovoltaic research community for their potential as efficient, inexpensive photovoltaic absorbers. The most rapid increases in power-conversion efficiency (PCE) in this class of solar cells have been driven by hybrid organic–inorganic perovskites, which have reached record efficiencies as high as 25.7%.¹ Inorganic CsPbI₃ has also generated considerable interest, given its ideal bandgap (1.73 eV) for integration into Si–perovskite tandem solar cells and its potential for improved stability with the absence of organic cations that can volatilize on extended annealing, such as methylammonium (MA⁺) and formamidinium (FA⁺). Unfortunately, the desirable broadband absorbing polymorphs of CsPbI₃ (β and γ) are metastable at and near room temperature; given time, they spontaneously transform to the yellow δ-CsPbI₃ polymorph (as shown in Fig. 1). The methods employed thus far to stabilize β- or γ-CsPbI₃ can be broken into two major categories: (1) compositional changes, such as alloying with a second halide, so that the β- or γ-phases of the new alloys are thermodynamically stable, and (2) engineering the film morphology, interfaces, and defect density to slow down the transition from the β- and γ-phases to the δ-phase.

To date, the most stable inorganic perovskite solar cell (PSC)² uses an all-inorganic device stack with passivated interfaces and a thermodynamically stable nanocrystalline γ-CsPbI₃ photoactive layer to realize an extrapolated t₈₀ lifetime of 51 000 h—equivalent to 5.8 years of continuous illumination or about 33 years of outdoor operation in Princeton, NJ, which receives ∼4.25 kWh/m²/day.³ This extrapolated lifetime rivals that of conventional inorganic solar cells, such as those made with Si. However, as is common among emerging solar cell technologies, optimizing the materials and architecture for stability often comes at the expense of PCE. While the CsPbI₃ cell referenced above holds a record lifetime, its PCE of 17.4% is only 60% of its Shockley–Queisser limit based on the bandgap of CsPbI₃. In contrast, the most efficient CsPbI₃ solar cells reported to date have an efficiency of 21% but a lifetime of only several hundred hours. In this perspective, we summarize the methods developed in recent years to improve CsPbI₃ PSC stability and discuss the prospects of developing efficient inorganic PSCs with commercially viable lifetimes.

The β- and γ-phases of CsPbI₃ are not thermodynamically stable at room temperature. They gradually convert to the δ-phase in an inert atmosphere; this transformation is accelerated in the presence of humidity.⁴ Thus far, alloying and grain size reduction have been the two main approaches to modify the free energy landscape of CsPbI₃ polymorphs such that γ-CsPbI₃ is thermodynamically preferred at standard solar cell operating temperatures. Replacing the B-site divalent cation or the X-halide in the perovskite ABX₃ chemical structure with elements that have a smaller atomic radius increases the Goldschmidt tolerance factor toward a cubic crystal structure and favors the more symmetric β- and γ-polymorphs over their δ-counterpart [depicted schematically in Fig. 2(a)].^5–9 Reducing the crystalline domain size to <100 nm increases the surface energy contribution^2,10 to the Gibbs free energy compared to the bulk contribution, which also stabilizes the γ-polymorph [Fig. 2(b)]. However, both alloying and grain size reduction have, thus far, come with material properties that compromise the efficiency of solar cells they comprise. Replacing 40% or more of the iodine in CsPbI₃ with bromine, for example, has been shown to prevent the spontaneous phase transition to the δ-phase at room temperature.⁶ However, this large degree of bromine incorporation raises the bandgap of CsPbI_XBr_3−X to >1.9 eV, significantly decreasing visible light absorption compared to CsPbI₃, thereby limiting the efficiency ceiling of both single junction and tandem solar cells. Similarly, additives to perovskite precursor solutions, such as water,¹⁰ polyvinylpyrrolidone (PVP),² and other surfactants,¹¹ have resulted in the formation of sub-100 nm nanocrystals of γ-CsPbI₃, effectively destabilizing δ-CsPbI₃ in favor of γ-CsPbI₃ at room temperature. The efficiency of solar cells incorporating these active layers, however, is reduced relative to those having similar active layers with larger grains, likely due to an increase in grain boundary density and the incorporation of non-conductive additives within the active layers.

Because altering the chemical composition of CsPbI₃ to increase its stability has, thus far been met with trade-offs in PSC efficiency, approaches that do not alter the material composition—and thus the thermodynamics—but, instead, increase the kinetic barrier for transition from any of the desired black phases to the undesirable δ-phase are potentially attractive. Defects and interfaces are known to lower the energy barrier for phase transitions [Fig. 2(c)]; thus, decreasing the density of defects in CsPbI₃ can synergistically improve the phase stability of PSC active layers while also increasing PCE. For example, the use of co-solvents,¹² the incorporation of volatile additives,^13,14 and controlling the formation of intermediate phases during crystal growth^15,16 have been deployed to access pinhole-free films with μm-scale grains. Point defects can also be passivated with surface treatments, such as the exposure of CsPbI₃ active layers to organic ammonium halide salts.^13,16–18

Perhaps, the most common and successful surface passivation strategy for organic–inorganic hybrid PSCs is the formation of a thin 2D perovskite overlayer atop the bulk 3D perovskite. The 2D perovskite overlayer passivates surface defects, reduces ion migration, and helps prevent water ingress.¹⁹ While such 2D overlayers are notoriously difficult to form on inorganic perovskites, Zhao et al. have recently demonstrated that the surface treatment of CsPbI₃ with CsCl can produce a 2D Cs₂PbI₂Cl₂ overlayer, the presence of which improves both the stability and efficiency of PSCs comprising this layer.² 

In addition to causing lattice defects, strain can alter the free energy difference between CsPbI₃ polymorphs. As calculated by Steele et al., strain in CsPbI₃ induced by adhering the film to a substrate can reduce the thermodynamic driving force for transformation from the γ-phase to the δ-phase.²⁰ On the other hand, the introduction of soft interfacial adlayers, such as alkyltrimethoxysilane derivatives,²¹ have been shown to relieve residual tensile stress from the thermal expansion coefficient mismatch between β-CsPbI₃ and its surrounding layers,²² improving phase stability, and consequently the lifetime and efficiency of the resulting PSCs. It remains unclear, however, why substrate-induced tensile strain favors the γ-phase of CsPbI₃ in one case and the δ-phase in the other. This, along with the possibility of matching thermal expansion coefficients of surrounding layers or using thermal processing routes to preferentially stabilize the β- or γ-polymorphs, should be further explored.

A potentially useful property of CsPbI₃ is its polymorphic reversibility, accessible through high temperature annealing. In a recent study by Liu et al.,¹⁴ the authors fabricated an all-inorganic PSC with the structure FTO/NiO_x/CsPbI₃/ZnO/ITO. Operating under simulated sunlight and maximum power point (MPP) tracking at 60 °C, the PCE of the device decreased to 90% of its original value after ∼300 h, primarily due to the transformation of the γ-phase to δ-CsPbI₃. Annealing the solar cell at 350 °C, however, fully restored the metastable γ-phase and the device recovered its initial performance characteristics. This performance loss and recovery cycle was demonstrated ten times, suggesting that it could be repeated indefinitely. While a 300 h lifetime is too short for a solar cell to be practical, if this recovery can be controllably achieved in a PSC with a lifetime of several years, this approach could open up the possibility of field-serviceable solar cells that can be periodically restored to like-new condition.

As increasingly stable PSCs are developed, accelerated aging will be required to study the lifetimes of these devices on reasonable laboratory timescales. For solar cells with a thermodynamically stable inorganic absorber layer, such as nanocrystalline γ-CsPbI₃, accelerated aging with temperature as the stressor has been shown to be highly effective.² However, for devices in which the active layers can undergo polymorphic transformation over the prescribed testing window, the extraction of an acceleration factor can be complicated by the fact that the transformation rate does not monotonically increase with temperature. Testing at low temperatures provides a large driving force for transformation, but the thermally activated atomic motion required for transformation is slow. As temperature increases, the driving force for transformation is reduced while atomic motion accelerates, leading to competition that produces a maximum transformation rate at some intermediate temperature (∼250 °C in CsPbI₃) with slower transitions above and below.

A compiled a list of notable inorganic PSCs is shown in Table I, including highly efficient and stable devices employing each phase stabilization strategy discussed above. The most stable device uses nanocrystalline γ-CsPbI₃, interface passivation, and an all-inorganic device stack to realize a t₈₀ lifetime of 51 000 h, with a PCE of 17.4%.² On the other hand, the highest efficiency devices that deploy CsPbI₃ have reported PCEs of 20%–21%, with t₈₀ lifetimes on the order of 100–1000 h.

Examining the device architecture column of Table I reveals that many solar cells were constructed with variants of the architecture FTO/TiO₂/perovskite/spiro-OMeTAD/Au. However, the presence of Li-doped spiro-OMeTAD as a hole transport layer, which is known to degrade thermally and via diffusion of Li ions,²³ makes it difficult to evaluate the intrinsic stability of CsPbI₃ in these device structures. We thus advocate for the use of a stable “baseline” architecture with, for example, CuSCN as the hole-transport layer and passivated transport layer interfaces² so that the stability of CsPbI₃ can be assessed in isolation. Tremendous progress has been made over the past decade to develop charge-transport materials and accompanying dopants that are stable against thermal annealing and extended operation.^24–26 However, device optimization is not a simple task and more work must be done to tailor these transport layers and interface passivation strategies for CsPbI₃ solar cells.

Another notable trend from Table I is the efficiency reduction in solar cells in which the active layers have been alloyed with a second halide or B-site cation, or the active layers employ nanocrystalline CsPbI₃. Halide alloying with chlorine or bromine inevitably increases the bandgap beyond 1.7 eV, which is detrimental to PCE and Si-tandem compatibility of the resulting solar cells. However, B-site alloying can potentially stabilize the γ-polymorph while reducing the bandgap, increasing visible absorption to produce single junction solar cells that have higher theoretical efficiencies. This strategy has yet to experimentally demonstrate the improved efficiency of solar cells or prove the thermodynamic stabilization of the desired polymorph in the active layer. Thus, we recommend further exploration of this area, including analyses to verify the inclusion of B-site dopants into the perovskite lattice and x-ray diffraction analysis of the active layers of aged solar cells to determine whether the δ-phase transition has been effectively suppressed. By far, the most effective CsPbI₃ stabilization approach has been realizing a nanocrystalline morphology by employing polymers and surfactants that limit grain growth. Further development should be devoted to more effectively removing or functionalizing these additives to minimize their hindrance on charge transport.

Inorganic perovskite CsPbI₃ is a promising PSC active layer alternative to hybrid organic–inorganic perovskites due to its high thermal stability. Devices comprising this material, however, have thus far struggled to achieve high operational stability because CsPbI₃ has a tendency to undergo spontaneous phase transition to the non-perovskite δ-phase at and near room temperature. In this perspective, we have discussed several methods to improve the phase stability of perovskite CsPbI₃, including alloying or reducing grain size to create a thermodynamically stable materials system, strain engineering to reduce the thermodynamic driving force for phase transition, and fabricating defect-free films with passivated surfaces to increase the kinetic barrier for phase transition. While employing nanocrystalline CsPbI₃ to achieve thermodynamic stability has yielded PSC lifetimes that are commensurate with commercial requirements, strain engineering to reduce the driving force for transformation and surface passivation to increase the kinetic barrier for transformation appear more promising methods to achieve commercially viable device efficiencies because they do not alter the bandgap or hinder charge transport. Fortunately, these stabilization methods need not be mutually exclusive, and a solar cell employing an appropriately strain-engineered, surface-passivated CsPbI₃ active layer with large grains coupled with the use of intrinsically stable transport layers is an exciting prospect for PSCs with commercially viable lifetimes and efficiencies.

Thus far, the field of perovskite photovoltaics has primarily focused on improving PCE. This paradigm has yielded impressive results, including the 25.7% record PCE recently reported for single-junction hybrid PSCs, but has largely failed to deliver PSCs with commercially viable lifetimes. We advocate for a different paradigm, where baseline devices that are highly stable (e.g., interface-stabilized nanocrystalline γ-CsPbI₃ with inorganic transport layers) are engineered toward higher efficiencies.

This material is based upon the work supported by the National Science Foundation Graduate Research Fellowship Program under Grant No. DGE-2039656. Q.C.B. thanks the Arnold and Mabel Beckman Foundation for funding this work through an Arnold O. Beckman Postdoctoral Fellowship in Chemical Sciences.

The authors have no conflicts to disclose.

Alan B. Kaplan: Conceptualization (equal); Visualization (lead); Writing – original draft (lead). Quinn C. Burlingame: Conceptualization (equal); Writing – original draft (supporting); Writing – review & editing (equal). Rudolph Holley III: Visualization (supporting); Writing – original draft (supporting). Yueh-Lin Loo: Supervision (lead); Writing – review & editing (equal).

Data sharing is not applicable to this article as no new data were created or analyzed in this study.