Perovskite Photovoltaic Modules: Life Cycle Assessment of Pre-industrial Production Process.

Photovoltaic devices based on perovskite materials have a great potential to become an exceptional source of energy while preserving the environment. However, to enter the global market, they require further development to achieve the necessary performance requirements. The environmental performance of a pre-industrial process of production of a large-area carbon stack perovskite module is analyzed in this work through life cycle assessment (LCA). From the pre-industrial process an ideal process is simulated to establish a benchmark for pre-industrial and laboratory-scale processes. Perovskite is shown to be the most harmful layer of the carbon stack module because of the energy consumed in the preparation and annealing of the precursor solution, and not because of its Pb content. This work stresses the necessity of decreasing energy consumption during module preparation as the most effective way to reduce environmental impacts of perovskite solar cells.

Introduction

Photovoltaics (PV) represent a potential technology to mitigate the climate change and other pollution consequences while obtaining energy to power human activity (Chu et al., 2017). Nowadays, PV technologies based on halide perovskites have chiefly been developed at the laboratory scale, where it has raised much interest among the scientific community (Assadi et al., 2018). Its development is addressed in multiple ways: decreasing costs of production, enhancing its poor lifespan, guaranteeing safety despite its lead content or substituting it for another less toxic element, and producing them at industrial scale while maintaining high power conversion efficiency (PCE) (Chen et al., 2015, Fang et al., 2016, Zhang et al., 2016a). Thus far, there has been a fast progression in efficiencies that over 20% efficient perovskite solar cells (PSC) have been obtained in several laboratories around the world (Bi et al., 2016, Saliba et al., 2018, Saliba et al., 2016, Shin et al., 2017, Tan et al., 2017, Yang et al., 2017). Nevertheless, for bringing PSCs to commercialization and launching them into the global market, as several companies aim to do (Edis, 2015, Gifford, 2015, Peleg, 2015, Sherahilo, 2018), paramount parameters encompass low cost, large area, high throughput, high solar-to-energy PCE, reproducibility, cost performance, long lifetime, and low environmental impact (Qiu et al., 2018).

The mainstream architecture and deposition techniques used in laboratories cannot be easily translated to larger substrates. For example, spin-coating or anti-solvent deposition methods present a large waste of material and a difficult implementation in large scale (Baker et al., 2017b, Jiang et al., 2018), besides leading to an increase of environmental impacts (Alberola-Borràs et al., 2018b). On the other hand, some materials used in several laboratory configurations such as spiro-MeOTAD or Au should be avoided for their high cost, reduced stability, and high environmental burden (Alberola-Borràs et al., 2018a, Meroni et al., 2018). Consequently, new architectures have been investigated to overcome these limitations. Architectures in which the perovskite is deposited through slot die (Burkitt et al., 2018, Cotella et al., 2017, Schmidt et al., 2015), blade coating (Baker et al., 2017b, Di Giacomo et al., 2015, Matteocci et al., 2014), and solvent-free pressure processing (Chen et al., 2017) are discarded because they still need an evaporated metal contact to complete the device or have low efficiencies (<5%). At the same time, a laminated device with a metal grid poly(3,4-ethylenedioxythiophene)polystyrene (PEDOT:PSS) cathode has been reported with an efficiency over 10% (Bryant et al., 2014a, Di Giacomo et al., 2015, Hooper et al., 2015), but the lifetime of this has not yet been proven. On the other hand, a large-area module based on a fully printed mesoporous stack, using carbon as cathode, has been reported, exhibiting low cost, high throughput, and high stability (Baker et al., 2017a, Cai et al., 2017, Mei et al., 2014). In this configuration, the use of the expensive and unstable Spiro-MeOTAD and gold are avoided. As such, it is viewed as one of the closest to commercialization (Cai et al., 2017, Moulder et al., 1992). Perovskite is infiltrated into a semiconducting scaffold of mesoporous titania (m-TiO2), an insulating scaffold of mesoporous zirconia (m-ZrO2), and a cathode of carbon, whose porosity is crucial to control crystallization of the perovskite over a large area (Cotella et al., 2017). These layers are deposited through screen printing, which enables reproducibility in large-area substrates (Philip et al., 2016, Yasin et al., 2016). Despite the fact that infiltration of the precursor solution is usually conducted manually, recently an automated system to deposit the perovskite with a robotic dispenser and a mesh has demonstrated more homogeneous depositions on large areas (Meroni et al., 2018) and modules with active areas of up to 198 cm2 have been reported (De Rossi et al., 2018). Furthermore, this configuration with a proper encapsulation exhibits outstanding lifetimes beyond 1 year (Grancini et al., 2017). By using different perovskite compositions in this carbon stack an efficiency close to 16% has been reached (Zhang et al., 2017c). Yet, tuning the perovskite composition with formamidinium, cesium, methylammonium, iodide, and bromide ions has led to adverse environmental consequences due to an increased amount of reagents (Alberola-Borràs et al., 2018b). Another advantage of the process is the usage of an ultra-fast annealing process with near-infrared radiation technique (Hooper et al., 2014). However, fluorine-doped tin oxide (FTO) remains the most expensive material in the structure (Park et al., 2016).

In addition to efficiency and cost issues, the environmental impact of the devices should be considered in the future implementation of this technology. The toxicity of lead embedded in perovskite remains one of the main concerns of PSCs since their early days (Park et al., 2016, Qiu et al., 2018, Rajagopal et al., 2018). Pb is notorious for its detrimental effects in the human body (Fewtrell et al., 2003, Qiu et al., 2018). Its damaging activity consists in the mimicry of the essential ions Ca, Zn, and Fe involved in biological processes (Babayigit et al., 2016b, Klaassen, 1980). Nonetheless, different studies have assured that its presence in PSCs should not pose a restrictive concern for its commercialization (Hailegnaw et al., 2015, Hauck et al., 2017). In fact, emissions of Pb stemming from other established applications are higher than those related to PSCs, such as lead-acid batteries, crystalline solar cell panels (during its production), and weather-proofing lead sheets on roofs (Gottesfeld and Pokhrel, 2011, Hauck et al., 2017). Still, PSCs embedded in consumer electronics or portable systems may find a barrier in the European market through the “RoHS Directive” (European Parliament, 2011), as it restricts the use of lead to 0.1% for homogeneous materials (Kadro and Hagfeldt, 2017). Several solutions to mitigate the detrimental effect of lead in PSCs have been proposed by the scientific community, such as designing safe production processes to prevent harmful consequences due to handling of Pb (Hauck et al., 2017) and efficient recycling processes for Pb as well as for the rest of the materials present in the solar cell (Chen et al., 2014, Kadro and Hagfeldt, 2017, Rajagopal et al., 2018, Zhang et al., 2016b). In parallel, Pb-free PSCs are under development using either Sn or Bi as substitutes (Abate, 2017, Jain et al., 2018, Liu et al., 2017), but their efficiencies are still quite low, and benefits in the environment, derived from the usage of these elements instead of Pb, are in doubt (Babayigit et al., 2016a, Serrano-Lujan et al., 2015).

A significant number of studies based on life cycle assessment (LCA) have been conducted to support PSCs on its way to commercialization. Some previous LCA studies, oriented toward the commercialization of perovskite PV modules, evaluated some techniques suitable for low-cost manufacturing. For instance, an LCA analyzing from cradle to gate two perovskite devices using spray and co-evaporation methods was reported (Celik et al., 2016). In contrast, the first LCA applied to PSCs compared two deposition methods, spin-coating and evaporation (Espinosa et al., 2015). Other LCAs likewise analyze laboratory-scale devices to find weak points and possible improvements from an early stage of PSC development (Alberola-Borràs et al., 2018a, Gong et al., 2015, Zhang et al., 2017a). Another LCA contrasting different configurations of Si/perovskite tandems concludes that the best configuration was free of spiro-MeOTAD and used Al instead of noble metals (Monteiro Lunardi et al., 2017). More analyses based on LCA contrast a handful of configurations of tandems with perovskite (Celik et al., 2017a, Celik et al., 2017b, Hauck et al., 2017, Itten and Stucki, 2017). LCA has been directly applied to the perovskite layer to contrast various compositions combining different cations and anions (Alberola-Borràs et al., 2018b). Similarly, different PSCs containing different perovskite compositions are compared in two studies (Ibn-Mohammed et al., 2017, Zhang et al., 2017b). The substitution of Pb for Sn in the perovskite layer is also analyzed in several studies (Babayigit et al., 2016a, Celik et al., 2017b, Serrano-Lujan et al., 2015). To the best of our knowledge, an LCA study directly applied to an industrial process of production of large-area PV modules based on perovskite has not been performed to date.

In this work, the environmental performance of a perovskite module, based on the carbon mesoporous stack architecture and produced with a pre-industrial process, is analyzed via LCA, from cradle to gate, to determine the major environmental impacts of each manufacturing step. This pre-industrial process is intended to be a preliminary step toward commercialization of perovskite modules. Remarkably, the energy consumptions of all equipment were directly measured and turned out to cause the most significant portion of environmental impact. The investigated process is based on a high-throughput process of production of a large-area module (hereafter referred to as pre-industrial module), reported in a previous work (De Rossi et al., 2018), to which some alterations are implemented (Baker et al., 2017a). Usage of data stemming from a pre-industrial process provides a good approach of the environmental impact that will generate a real process. In addition, an ideal industrial process of production, based on the pre-industrial one (hereafter referred to as the ideal module), is simulated (not directly measured) and environmentally assessed. In the ideal industrial process, energy consumption of some steps and usage of some materials are optimized with respect to those in the pre-industrial process, as it should be expected for the ideal implementation of a production line. We define an ideality coefficient that quantifies how close a given fabrication procedure is to the ideal process, in terms of environmental impacts. Finally, the progress attained by the large module produced via this pre-industrial process with respect to a small PSC produced by means of the most extended laboratory-scale process pertaining to a previous phase of development (hereafter referred to as laboratory-scale PSC) is illustrated via comparison of its ideality coefficient (Alberola-Borràs et al., 2018a).

Results and Discussion

LCA of Pre-industrial Module

Environment wise, the production of a perovskite PV module with a carbon stack architecture (pre-industrial module) is scrutinized to elucidate its main weaknesses. For this purpose, the impact of each of the layers of the module is estimated for all the categories considered for this study, which is shown in Figure 1. Impacts of each layer are divided by the overall impact of the module. To make them comparable, impacts of each layer are aggregated per category. More information about how the environmental impacts are obtained can be found in Transparent Methods, and Tables S1 and S4–S9 in the Supplemental Information.

Figure 1

Aggregated Impacts of Each Layer of the Carbon Stack Perovskite Module, Sorted by Impact Categories

Distribution of impacts in Figure 1 exposes that the perovskite layer presents the biggest impact among all layers. Its contribution is superior to that of the rest of the layers, except in photochemical oxidation (POP) category. For most of the categories, the contribution of the perovskite layer is superior to 50%. In contrast, for ozone layer depletion (ODP), POP, and acidification (AP) categories, the impact of perovskite layer is below 50%. Most of the impact of the perovskite layer stems from the use of energy flow, except for the abiotic depletion (ADP) category where it mostly stems from the materials flow. Both heating up and annealing processes involved in the perovskite deposition contribute similarly to the impact. Both high consumptions originate from a forced convection generated to assist perovskite crystallization during the annealing process and a process of heating of perovskite reagents carried out in a hot plate, which needs optimization. Thus, a reduction of its impact should be among the next goals to improve the sustainability of the pre-industrial process. For instance, the amount of precursor reagents could be reduced with an automatic deposition using a robot and a mesh, instead of depositing it manually (Meroni et al., 2018). A reduction of the energy required for heating up the precursor solution at 70°C and annealing the perovskite layer—e.g., via heat recovery and other methods for reducing the crystallization time—would also be necessary.

For the POP category, the most adverse layer is the blocking layer, accounting for more than 90% of the total. For this category, the impact mostly stems from the emissions. The most harmful compound emitted is isopropanol. The impact of blocking layer is also significant for the rest of the categories alongside the anode + substrate layer, whose contribution is above 10% in most of them. Use of energy and materials are the main responsible flows of their impact. Moreover, impact of the cathode is also noticeable.

To assist the analysis of the pre-industrial module, the distribution of impacts for each of the impact flow is depicted in Figure 2 for all categories. Materials, use of energy, amount of transportation, and emissions flows are included in this analysis. Impacts of each flow are divided by the total impact to obtain the percentage of contribution.

Figure 2

Aggregated Impacts of the Carbon Stack Perovskite Module, Sorted by Impact Categories

When the impact of the four types of flows are compared in Figure 2, the use of energy is seen to be the most detrimental for abiotic depletion (fossil fuel) (ADPF), climate change (GWP), AP, eutrophication (EP), cumulative energy demand (CED), human toxicity (cancer effects) (HTC), human toxicity (non-cancer effects) (HTNC), and freshwater ecotoxicity (FET) categories, varying between 69.1% and 90.2%. As well as it happens in Figure 3, impacts of those categories mainly stems from the perovskite layer, in particular from the heating up of the precursor solution and annealing of the film.

Figure 3

Aggregated Impacts of Material Inputs of the Carbon Stack Perovskite Module, Sorted by Impact Categories

In contrast, for ADP and ODP the most harmful flow is materials, which ranges from 56.7% to 92.6%. For ADP category, lead iodide reagent for the perovskite production is the most harmful material: its impact is one order of magnitude higher than methylammonium iodide (MAI), two orders of magnitude higher than 5-ammonium valeric acid iodide (AVAI), and three orders of magnitude higher than the solvent γ-butyrolactone (GBL). Meanwhile, for ODP the main material responsible for the impact is not as clear, since all layers contribute roughly the same. For the POP category, the contribution of emissions flow is higher than 90%, due to the release of isopropanol used copiously as a carrier to enable the blocking layer deposition via spray.

As the materials chosen are an important concern for the production of PV devices and their impact is usually hidden by that of the use of energy flow, we focus on the materials used for the production of the pre-industrial module. The impact of each compound used is divided by the total impact of the materials flow and displayed in Figure 3, sorted by categories. As the impacts of some of the compounds depicted are too little to be appreciated in the chart, they are aggregated in a single group (others), which comprises TiAcAc, TiO2, ethylcellulose, 2-(2-butoxyethoxy) ethyl acetate, nitrocellulose, AVAI, polyethylene terephthalate, zirconia, and carbon. On the other hand, contributions of the impact of FTO, glass substrate, isopropanol, α-terpineol, PbI2, MAI, and GBL are shown individually.

The glass substrate, accounting for the largest fraction of the mass of the pre-industrial module, represents the most detrimental material. Its contribution is above 43% for all categories, except for ADP (slightly below 20%) and ODP. For the ADP category, the most harmful material, contributing nearly 60% to this category, is the PbI2 used as reagent for the perovskite synthesis. The reason behind such contribution lies in the fact that a large amount of it is used and its impact per kilogram is high. Isopropanol and α-terpineol solvents have a significant contribution to the overall impact. Terpineol is especially detrimental for ODP category, where it represents nearly 67% of the total. On the other hand, mass of isopropanol used per kWh is the highest of all materials, i.e., 0.0908 kg/kWh. Impact of MAI is relatively modest except for ADP category, where it represents more than 20%. Moreover, GBL impacts are appreciable for every category. In consequence, the aggregate of compounds involved in the synthesis of perovskite (PbI2, MAI, GBL) is higher than 10% for ADP, GWP, EP, CED HTC, HTNC, and FET. This fact reinforces the need for reducing the usage of reagents for the synthesis of perovskite as pointed out in the analysis in Figure 2.

Ideality Analysis

Minimization of material and energy consumption establishes the ideal scenario to decrease the environmental impacts caused by a device fabrication. Here, we define an ideality coefficient that quantifies how close a given fabrication procedure is to the ideal process, in percentage. Note that a technology requires an ideal coefficient as high as possible to reduce as much as possible the environmental impacts. However, no technological process can reach 100% ideality coefficient, as no technology can produce zero waste material and consume just the thermodynamic limit energy. This coefficient is depicted in percentages for both the pre-industrial module and the laboratory-scale PSC in Figure 4. Its value is the result of dividing the impact of the ideal process by the impact of the process to compare. For this analysis, the most fundamental categories are only used to ease its performance and thus its comprehension. As previously discussed and according to the data reported in Figure 1, the most concerning layers of the pre-industrial module are the anode + substrate, the blocking layer, and the perovskite, so these are the only layers included in the analysis. To assess in great detail these layers, this analysis is combined with the relative impacts—sorted by type of flow—of both the carbon stack module produced at pre-industrial scale and the PSC produced at the laboratory scale. The impact of each flow type pertaining to each layer is resized and aggregated to fit in the corresponding percentage of ideality coefficient. Results are sorted by device, by category, and finally by layer, where those of the anode + substrate layer (Figure 4A) are depicted from 0% to 100% and those of the blocking and perovskite layers (Figure 4B) are depicted from 0% to 1%. Further information about how these outcomes are obtained can be found in Tables S1–S9 and the Transparent Methods section in the Supplemental Information.

Figure 4

Ideality Coefficient for the Carbon Stack Module Produced with a Pre-industrial Process and the PSC Produced with a Process in the Laboratory Environment

The ideality coefficient quantifies how close a given fabrication procedure is to the ideal process, in percentage; its value is the result of dividing the impact of the ideal process by the impact of the process to compare. Relative impacts from pre-industrial module, PSC at laboratory scale, and ideal process, sorted by impact categories and layers: (A) anode + substrate, (B) blocking and perovskite layer.

See also Figures S1 and S2 in the Supplemental Information.

From the results in Figure 4, it is observed that the pre-industrial module reduces significantly all impacts, with ideality coefficients reaching values as high as 89.5% for ADP. In the anode + substrate layer, the pre-industrial process reaches the highest ideality coefficients among the three layers analyzed, where the pre-industrial device ranges from 55.8% to 89.5%. It is important to bear in mind that the cleaning step has been removed for the anode + substrate with respect to the process performed in the laboratory, which is the cause of the reduction in the impact of this layer. Remarkably, impacts derived from anode + substrate layer of both pre-industrial and ideal processes are almost alike, indicating that further optimization of the pre-industrial process should focus on the blocking layer and especially the perovskite layer. Theoretical optimization of materials and energy in the ideal process is the reason why ideality coefficient of the pre-industrial process is not closer to 100%. The materials flow is the most responsible for the impact of the anode + substrate, followed by the energy for all categories except ADP. The high values of the ideality coefficient of the pre-industrial process contrast with those of the laboratory-scale PSC, which does not surpass 3.0% (about 30-fold less), which reinforces the progression made by the pre-industrial process for the anode + substrate layer.

Ideality coefficients for the blocking layer are significantly lower, ranging from 0.51% to 0.83% for the pre-industrial process, where the highest ideality coefficient pertains to the HTNC category. However, it is far from the ideal process, mostly due to the use of energy flow for the blocking layer. For the blocking layer, a significant optimization of both materials and use of energy flows is recommended to improve ideality. For instance, depositing this layer by screen printing would result in the optimization of TiAcAc solution and a decrease in the usage of energy, as it happens for the mesoporous layers in the pre-industrial process. Furthermore, a reduction in the thickness of the blocking layer to 8 nm is feasible, via electrophoretic deposition method, with a subsequent reduction in materials (Li et al., 2015). Other deposition methods such as spray-cast and semi-automatic spray pyrolysis might pose an alternative for the industrial manufacture of the carbon-stack perovskite module (Bishop et al., 2017, Krýsová et al., 2018). For the laboratory-scale process, ideality coefficients fluctuate between 0.0043% and 0.0143%, which are well below those of the pre-industrial one (about 58-fold less). Therefore, for the blocking layer the pre-industrial process is less harmful, which highlights the advancement it has achieved.

In addition, results reveal that the process of deposition of perovskite is the least optimized, narrowly followed by the blocking layer. Its ideality coefficients vary between 0.06% for the ADP category and 0.09% for GWP, CED, HTC, and HTNC categories. The highest amount of energy consumed for the pre-industrial process, to prepare the solution and to anneal the deposited layer, is responsible for these striking results because use of energy is the most detrimental flow for this layer. Therefore, finding an alternative, such as heating with near-infrared radiation (Baker et al., 2017a, Hooper et al., 2014), especially for the annealing step, should be fundamental to reduce the impact of this process to that of the ideal process and get it off the ground. Alternatively, using other heating techniques needing shorter operational times such as near-infrared radiation (Bryant et al., 2014b, Troughton et al., 2015), photonic flash-annealing (Troughton et al., 2016), and high-temperature, short-time annealing processes (Kim et al., 2017) could optimize the environmental performance of the perovskite layer. When compared with those of the laboratory-scale PSC, ideality coefficient values of the pre-industrial device are notably higher (about 8-fold higher), despite the fact that both processes are far from ideality, pointing out the improvement already achieved. The overall outcomes of the study are provided in Tables S10–S12 of the Supplemental Information.

Conclusions

A cradle-to-gate LCA of a pre-industrial process of production of a large-area perovskite module based on a carbon stack architecture is assessed. An ideality coefficient is obtained to evaluate the level of optimization of the pre-industrial module, which shows overall encouraging results. This ideality coefficient of the pre-industrial process is compared with that of a mesoporous structured PSC produced in the laboratory environment and with an extrapolated ideal situation in which material and energy consumption is minimized.

The perovskite layer is found to be the layer with the greatest impact on the pre-industrial module, mainly due to the energy consumed in the preparation and annealing of the precursor solution, rather than the Pb content, which raises a greater concern. This step is highly amenable to optimization.

Ideality coefficients of the pre-industrial process show a significant improvement regarding environmental impacts for the most relevant layers, namely, the FTO-glass substrate, the compact TiO2 blocking layer, and the perovskite. The first one generates low impacts, and is already close to optimal, whereas the energy consumptions of the perovskite and blocking layers are still too high and must be reduced.

Limitation of the Study

This study presents the environmental impacts of a large-area perovskite PV module produced with a pre-industrial process, which are obtained by a LCA. Due to the novel state of development of this process, encapsulation and contacts of the resulting PV module are not definite to date. Therefore, they were not included in the system analyzed herein. In parallel, some other assumptions were taken, such as the substrate and the anode are deposited right before the module deposition, and thus there is no need of applying a cleaning process onto them. The stability and efficiency of the PV devices based on perovskite can be significantly improved in the near future. However, in this study, their current empirical values are utilized. The LCA is performed from cradle to gate, therefore usage and end-of-life phases are not included in this work. Finally, among all the existing processes to produce large-scale perovskite PV modules, this study tackles the closest process to commercialization.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.