Our scaling work initially focused on an inverted p–i–n structure with an ITO/NiO/Cs_0.17FA_0.83PbI₃/C₆₀/BCP/Ag layer stack. We have demonstrated rapid open-air processing of manufacturable device layers by developing a robust and stable spray coated NiO HTL (Scheideler et al., 2019) and perovskite active layer with RSPP. The top ETL and electrode were thermally evaporated in vacuum. A complete cost model for the full manufacturing process was developed to determine the annual costs for module production in a 9-step process were each calculated based for the entire process (Figure 2A) and amounted to $36.50/m² or $0.23/W. Technoeconomic analysis was previously reported for a pilot line for a 2.46 m² module produced in a factory generating 100 MW of annual production, which represents a significant investment on the order of tens of millions of dollars and was aimed at targeting utility-scale energy production.

Here, we extend the analysis for the identical module architecture using a pilot line that could be currently achieved in a lab setting, a more practical situation relevant to the current state-of-the-art facilities employed by academic and industrial perovskite processing. This line consists of 0.1 m² modules, with total annual production of either 4 or 19 MW of annual production, and the cost model estimates prices of $1.80/W or $1.65/W respectively, with the majority of the cost being driven by the ITO sputtering as well as the ETL and top electrode evaporation steps. These steps represent 58% of the total production cost per module, with their costs being elevated primarily due to low throughputs because of their reliance on vacuum conditions. Slower processing speeds necessitate more tools to keep pace with the other production steps, thus multiplying the labor, utility, and floorspace required to accommodate the tools. In addition, the cost of these three vacuum steps is disproportionately exacerbated by the scaling down of the production line; they only represent 44% of production costs in the 100 MW factory. Labor costs experience the largest relative growth when downsizing, as the number and cost of human workers do not adjust as naturally as other inputs such as precursor and utility usage. Surprisingly, the materials cost drops precipitously for the smaller pilot line, comprising just 4% of manufacturing costs compared to 44% for the 100 MW factory (Figure 2B). All labor costs were based on US manufacturing, and this trend illustrates the challenges that US-based perovskite start-ups often encounter since labor costs can be exorbitant in R&D facilities until reaching factory-scale production. This transition has not yet been reached for perovskites due to continued challenges regarding module reliability and operational lifetimes.

Our cost model has proven particularly useful because of the specificity of the material and equipment quotes we have garnered from industry suppliers. In addition, the model calculates cost as a function of throughput, meeting a user-input target of modules per hour by automatically adjusting the number of tools required, changing the other dependent variables such as materials and power usage accordingly. This greatly facilitated our analysis of production bottlenecks and optimal output levels. Moving forward, more exact direct labor data would help to increase the accuracy of our cost model; there is some uncertainty about labor intensity, the number of operators required per tool, and how much room there is for automation. Other recent cost models for perovskite modules were useful for comparison as we financed our own production line, and they also show disproportionate expenses from rate-limiting vacuum steps (Chang et al., 2017; Li et al., 2018). However, it is unclear based on these reports if our method of increasing tool numbers to meet desired throughput is the industry standard. Future work is therefore needed to address how to either reduce the adverse speeds and costs associated with vacuum steps or to sidestep them altogether. Increased transparency regarding low throughput steps would likely be a boon to the perovskite field.

Our analysis clearly indicates that the vacuum-based processes dominate the cost of production, with sputtered ITO, thermally evaporated electron transport layer (ETL) and the top electrode, and the vacuum processed encapsulation as the largest contributions. Clearly, vacuum deposition both poses a challenge for integrated linear production and dominates the overall cost of module production. Therefore, a significant opportunity exists to develop open-air spray-based capabilities for fully open-air manufactured perovskite modules with scalable methods.

Ironically, the costly ETL, top electrode, and encapsulation are also implicated in perovskite degradation mechanisms leading to metal diffusion/metal halide formation that currently dictates cell stability and lifetimes.

A key innovation with broader application to the perovskite community is a comprehensive module investigation strategy that focuses on high-throughput degradation monitoring of devices with a focus on targeting interfaces for improving module stability. There is a dearth of fundamental understanding for statistical analyses of degradation modes in perovskite modules under operational conditions. Our group employs a full suite of optoelectronic, structural, and mechanical characterization including color-normalized images, large-area optical microscopy, X-ray diffraction, time resolved photoluminescence, and detailed photoluminescence mapping at length scales from nm to cm to understand the fundamental degradation pathways due to each individual stressor and combinations of stressors. This holistic approach including multiple synergistic environmental stressors and in-situ characterization capabilities represents a significant advance over existing stability studies.

Detecting early-onset of degradation streamlines the design iteration process toward achieving stable perovskite cells Comprehensive studies are necessary to understand and mitigate degradation mechanisms in perovskite solar cells and modules. These need to be validated by adherence to consensus IEC standards and ISOS protocols (Khenkin et al., 2020) for accelerated aging and characterized by tracking device performance under specific combinations of heat, light, humidity, and electrical bias. Including multiple synergistic stressors and in-situ characterization capabilities over a wide range of length scales represents a significant advance over existing stability studies and will allow for determination of the spatial distribution of surface defects, grain boundaries, phase segregation, and interfacial reactions. Each individual stressor and combinations of stressors for both unencapsulated (intrinsic material stability) and fully encapsulated (extrinsic device stability) devices informs our iterative design process. Additionally, there is a critical need to establish relationships between optoelectronic properties and device performance together with film morphology/structure and stability. The ability to predict degradation and evaluate stability with rapid characterization such as in-situ imaging and will streamline the design iteration process toward the critical requirement of stable perovskite solar cells and modules.

The role of mechanical properties has been largely overlooked until recently. Perovskites have been identified as a mechanically fragile PV technology (Rolston et al., 2016), and several recent reports have explored strain engineering as a concept to improve optoelectronic properties (Zhu et al., 2019; Chen et al., 2020b; Kim et al., 2020) and reduce perovskite degradation (Zheng et al., 2016; Xiao et al., 2017; Zhao et al., 2017; Rolston et al., 2018; Li et al., 2019). The prospects of this effort are impactful since typical thermal processing generates large residual tensile perovskites film strain from the CTE mismatch compared to the typically stiff substrates used (i.e. glass and silicon). Approaches for reduction of such mismatch strains from processing and during operation involve the use of higher CTE substrates and lower temperature processing. This includes open-air plasma processing which has been demonstrated to result in significantly lower perovskite film stress (Hovish et al., 2020).

Recent studies have also examined more exotic methods such as compositional tailoring through the replacement of ions in the perovskite lattice and interfacial management through the insertion of thin layers either above or below the perovskite to apparently modulate the perovskite film strain. However, there are some misunderstandings of the fundamental mechanics of these approaches in the literature that have led to incorrect conclusions. As an initial matter, there are conflicting reports of the substitution of smaller (Saidaminov et al., 2018) and larger cations (Lee et al., 2018b) to reduce strain. Additionally, the concept of interfacial management is unfounded as a mechanism to reduce CTE mismatch film strains since it is the CTE mismatch between the film and the substrate that dominates the film strains, neither the interface nor the insertion of intermediate films has a role on film stresses. This is particularly true regarding work that claims that a higher CTE HTL can offset the residual tensile strain in the perovskite (Xue et al., 2020). There are also similar claims of other methods for reducing tensile strain (Wu et al., 2019) and minimizing CTE mismatch with interlayers or 2D perovskite layers (Wang et al., 2019). This is simply not founded in well-established thin-film mechanics where intermediate or top layers play no role on CTE elastic stresses (unless very close to the film edge). This topic needs further elucidation in the field to avoid incorrect conclusions with undue attention being placed on the role of contact/buffer layers and interfacial management to relieve CTE mismatch stress.

In addition to educating the broader community on the importance of mechanics and its application to thin film PV devices, the primary goal of our efforts is to develop fully open-air manufactured perovskite cells with scalable processing that eliminates the remaining vacuum-based processes (Figure 4) and concomitantly improves cell stability. In addressing this remaining grand challenge, we focus on robust materials integration with a fundamental understanding of degradation mechanisms to produce commercially competitive and stable perovskite solar modules.