Consensus statement for stability assessment and reporting for perovskite photovoltaics based on ISOS procedures

We report a consensus between researchers in the field of perovskite solar cells (PSC) on procedures for conducting and reporting PSC stability testing. The International Summit on Organic Photovoltaic Stability (ISOS) protocols are re-assessed and extended for PSCs. In particular, the updated protocols highlight testing for: (1) Redistribution of charged species under electric fields, (2) distinguishing between degradation induced by various stress factors, and (3) reversible degradation. The recommended protocols are not for replacing existing qualification standards (e.g., IEC 61215), but rather to contribute to developing an understanding of PSC degradation mechanisms on research devices. Acceptance of these protocols and sharing the suggested datasets will facilitate inter-laboratory coordination will aid accumulation of PSC stability data acquired under well-defined and comparable conditions. This will allow the application of advanced approaches to analyzing large data sets, such as machine learning methods, and will accelerate the development of stable PSC devices.


Introduction
To ensure economic feasibility and competitive levelized cost of electricity, a new photovoltaic (PV) technology must achieve long-term stability. Desired durability can range from a few months to 25 years depending on the application, and is linked to the lifetime of the product in which the PV device is integrated (for example, the application range spans from disposable electronics to long-term facade elements). For power plants, the expectation for a PV module is 20 to 25 years of operation to match the reliability of currently employed silicon wafer-based modules. Halide PSCs are one of the most promising emerging PV technologies, especially when employed in high-efficiency multijunction architectures. The power conversion efficiency (PCE) of these potentially inexpensive, solutionprocessable devices has exhibited tremendous growth over the last decade, reaching 24.2% in a single junction PSC and 28% in a perovskite-on-silicon tandem 1,2 . The next major challenge for PSC technology, along with large area processing and manufacturing upscaling, consists in improving their reliability.
The degradation of PSCs is affected by multiple parameters, including the exposure to visible 3 and ultra-violet (UV) 4 light, high temperature, [5][6][7] contamination from the ambient environment (oxygen, humidity) [8][9][10] and electrical bias [11][12][13] . A detailed understanding of the various failure modes occurring during in-field operation of the solar cells is key to minimizing or even eliminating performance losses. Together with field tests, accelerated life-time (ALT) tests are of fundamental 5 importance to reduce the time to market of a new PV technology. Ideally, this requires sufficient understanding and verification that ALT testing indeed reproduces and amplifies only the failure modes observed under real operational conditions. Moreover, the acceleration factor should in principle be derived from a physical model such as, for example, the activation energy (i.e. the Arrhenius factor) in thermal ALT tests.
In 2011, a broad consortium of researchers developed recommendations for stability evaluation of organic photovoltaics (OPV) 14 . These standardized aging experiments are recognized as the "ISOS protocols", and were established at the "International Summit on Organic PV Stability (ISOS)" (Roskilde, Denmark; 2010). The ISOS protocols outline a consensus between researchers in the OPV field on performing and reporting degradation studies in a controlled and reproducible way, with fewer, yet more comparable testing conditions than those previously considered by the OPV community. The well-classified ISOS testing protocols and reporting requirements have subsequently allowed direct comparison of results between different research laboratories working on different solar cell designs, thus enabling successful round robin experiments, [15][16][17] and a comprehensive understanding of degradation in those devices.
Similar to the situation for OPVs several years ago, stability studies for PSCs are drawing increasing attention, as reflected by the growing number of publications on the topic and the increasing shift in emphasis of research in the field toward stability-related issues. Despite a large number of publications (over 3000 papers related to PSC stability only in the last three years), it is difficult to compare available results, mostly due to differences in the control and reporting of parameters as well as the inconsistent application of statistics to PSC stability data 18 . The time is now ripe for the development of unified PSC stability evaluation procedures 11, [18][19][20][21][22][23][24][25][26] , that the research community can broadly adopt, similar to the procedures recently developed for PSC efficiency measurements [27][28][29][30] . The ISOS protocols are an excellent starting point for unification of PSC degradation and stability testing, provided the particularities of perovskite solar cells are also addressed in the new protocols. The aim is to develop a consensus on standardized testing that enables consolidation of a large volume of published data into a single database. This can, in turn, be used to reliably compare stability studies, to analyze the relative significance of various degradation factors, and to ultimately identify key failure mechanisms in the devices. If the research community adopts a unified set of protocols in their experimentation and, perhaps more critically, in their reporting, a broad PSC stability database for different perovskite compositions (bandgaps) as well as solar cell layer stacks and architectures can be accumulated over time, eventually allowing easier identification of common features. In turn, this can lead to the effective implementation of machine learning (ML) toward predicting lifetimes and failures 31 .
The present work presents and extends the outcomes of a Round ISOS stability protocols are most frequently applied at the cell level, but their application to neat materials, "half" cells (incomplete PV stacks) and mini-modules can also provide valuable information on degradation processes. These protocols are not intended to be a standard qualification test, nor are they suited for application by industry or insurance agencies. Unlike qualification standards, the solar cell cannot pass or fail ISOS stability tests. Instead, these are research guidelines aimed at ensuring the comparability of solar cell testing performed at different laboratories, and therefore assist in improving the quality and relevance of published data in the field. The existing qualification tests described in the IEC 61215 and 61730 standards 33,34 were designed to apply to the field performance of silicon wafer-based solar panels to screen for well-understood degradation modes generally associated with conventional solar module level issues.
These tests are unlikely to be well suited to emerging PV technologies (i.e., organic, dye-sensitized, halide perovskite, among others), due to their fundamentally different material properties and device architectures. In fact, today, various reports show that the stability of perovskite-based devices cannot be fully assessed by the procedures developed for conventional PV products. In particular, when testing PSC stability, special attention should be paid to 1) recovery processes after stress removal; 2) the presence of mobile charged species (ions); and 3) distinguishing the processes related to exposure to ambient atmosphere, from device-related, intrinsic factors. PSCs are also known to exhibit hysteresis in their current density-voltage (J-V) characteristics (i.e., a dependence of the J-V measurements on direction and rate of the voltage sweep 35,36 ), which imposes constraints on the cell performance and stability characterization methods, but not necessarily signifies less stable devices 37 . The update to the ISOS aging protocols presented here reflects all mentioned features. In addition, we suggest reporting guidelines to facilitate data 6 aggregation and comparisons. To this end, we also advocate the reporting of performance of suboptimized cells in addition to that of "champion" devices, as these data sets are critically important for training machine learning models.

Existing ISOS stability protocols
As noted by Christians et al., stability research should be performed at three connected hierarchical levels: material, solar cell, and solar module 21 . Many of the stress tests specific to encapsulated modules (such as hail test, PID, bypass diode stability etc. 33,34 ), as well as mechanical stability and special consideration for space applications, are outside the scope of this report, and indeed will likely be leveraged from the existing IEC 61215 standard.
ISOS protocols designed for OPV are grouped by the applied stresses (see Table 1) 14 . Each group of the original ISOS protocols has three levels of sophistication, which aim at covering different levels of laboratory infrastructure. The first level requires only basic equipment providing lower control over the stress factors. The second and third levels require more specialized tools, such as environmental chambers and maximum power point (MPP) trackers, but provide a higher level of confidence in the results reported and, in most cases, more stringent test conditions. Explicit values in Table 1 stand for the controlled (i.e. monitored and adjusted) parameters. Note, that suggested protocols allow working with both encapsulated and unencapsulated devices as long as it is clearly reported (see detailed discussion in section 3.3).
Dark storage studies (ISOS-D) provide information on the tolerance of the solar cells to oxygen, moisture, other aggressive atmosphere components (e.g. CO2, NOx, H2S, etc.) and elevated temperatures. In other words, ISOS-D tests give an estimation of the cell shelf life under ambient conditions without light exposure. Ambient atmosphere is crucial for the lifetime of perovskite absorbers and some of the transport layers used in PSC architectures 9,38,39 . Particularly, interaction with ambient species can promote the formation of traps 40 or charge barriers 41 (as a result of increased density of mobile defects/ionic species and electronic traps within the active layer) as well as perovskite decomposition, which quickly deteriorate device performance 8 .
Atmospheric species were also shown to charge perovskite surface, affecting ions distribution across the device 42 . Elevated temperatures are used to study the thermal stability of the cell and to accelerate the degradation induced by other stressors 43 . Thermal degradation in the dark was observed in PSCs at elevated temperatures due to chemical and structural instabilities of the absorber materials 25,44 or transport layers 45,46 .
PV applications. 47,48 At the moment, the impact of phase transitions on the device lifetime is unclear, and so is the impact of different temperature regimes during (accelerated) aging. ISOS-D-1 tests onshelf stability, where the cell environment is only monitored but not explicitly controlled (it is assumed that room temperature (RT) in the laboratory is 23±4 °C). Monitoring and reporting ambient relative humidity level is critical in all the protocols, since dry (R.H. < 20%) and humid air represent dramatically different stress conditions for PSCs 39,49 . ISOS-D-2 is a thermal stability test performed at controlled (i.e. monitored and adjusted) elevated temperatures of 65 or 85 °C. ISOS-D-3 is a damp heat test requiring accurate control of both temperature and humidity with the same temperature setpoints of 65 or 85°C and the introduction of 85% relative humidity.
In outdoor stability studies (ISOS-O), aging is achieved under illumination by natural sunlight at ambient environment. Although these conditions are not necessarily reproducible (they depend on weather, location, season etc.), the results of outdoor testing are the most relevant to the device operation. Unlike other protocols, they can be directly applied for realistic device lifetime assessment, albeit specific to a given climate. Field tests also allow determining whether the list of failure modes identified in the lab is complete and adequate for understanding reliability of the solar cell under real operating conditions, and further, can provide a reference point for acceleration factor calculations. A similar approach was originally pursued to rapidly accelerate the stability of Si modules through the Flat-Plate Solar Array project (FSA) using five "Block Buys" that directly correlated outdoor tests with having passed various qualification tests that ultimately helped form the foundation of IEC 61215 50 . In this manner, in the FSA project from 1975-1985 outdoor module reliability was improved in a manner that reduced module failure rates from ~50% pre-Block V to ~1% for Block V 51,52 . To date, PSC outdoor stability studies are scarce [53][54][55][56] , but the community has gained some critical insights with ISOS-O experiments, such as the importance of light/dark cycling 54  The results obtained by J-V measurements and MPP tracking do not necessarily coincide in PSCs ( Fig. 1a) 22 , although they generally demonstrate similar trends 11,57 . Therefore, it is crucial for PSC characterization to rigorously describe the load and recovery time before J-V measurements. MPP tracking is encouraged, whenever possible, both as the most practical electrical bias condition for aging and as a reliable tool for PSC performance assessment (see more discussion in section 4.1). However, it is mandatory only at the third, most advanced level of ISOS protocols. It is also possible to use a fixed voltage bias near MPP (instead of active MPP tracking) at lower sophistication levels, as suggested by the original ISOS protocols 14 .
Light-soaking stability ("Laboratory weathering" in the original ISOS protocols 14 ) experiments (ISOS-L) in PSCs have been found to promote ion and defect migration 58-63 as well as phase segregation 64,65 in the perovskite photoactive layer causing efficiency degradation. Additionally, light can catalyze/accelerate harmful chemical reactions, which lead to perovskite decomposition 66 or defect formation 67 . Detrimental changes in organic charge extraction layers, material intermixing at the interfaces, and ions exchange with adjacent solar cell layers can also be caused by cell illumination 61,68,69 . Similar to OPV 70 and DSSC 71 , special attention should be paid to the spectral composition of the light source when studying PSC stability, particularly in the UV range. UV light was shown to assist perovskite decomposition 72 and increase the non-radiative recombination rate in PSCs based on mesoporous TiO2, 73 which thus may require UV blocking layers to become more stable. PSCs with novel transport layers were shown to be tolerant towards UV irradiation 74,75 . Light from UV-A and UV-B spectral ranges may have different effects on PSC stability 76 .
Thermal cycling (ISOS-T) in the dark and light-humidity-thermal cycling ("Solar thermal humidity cycling" in the original ISOS protocols 14 ) (ISOS-LT) are more sophisticated protocols that evaluate the damage to photovoltaic devices caused by diurnal and seasonal variations of the weather in terms of solar radiation, temperature, and humidity. These tests are relevant to PSCs as for any other outdoor dedicated PV technology since they i) simulate realistic conditions, ii) stimulate failure mechanisms related to delamination of layers/contacts 77 Table 1. Overview of reported 14 and suggested ISOS protocols. The latter are printed in bold.

Suggested new ISOS protocols relevant for PSCs
Recently some aging protocols for PSC were suggested 25,27 . Mostly, they feature subsets of the original ISOS protocols described above, although some additional ideas, particularly concerning electrical bias application 25 , have also been introduced. Below we suggest extensions of the aging procedures that particularly pertain to the unusual properties of halide perovskite PV. While a major part of the original purpose behind the ISOS protocols was to limit conditions for each level to facilitate more ready comparison, there is limited data on the following protocols. We propose a reporting framework, some example conditions, and a discussion as to why these might be relevant. A similar framework is likely to be useful for OPV and other emerging PV technologies. If these protocols become widespread over the next few years, the community can adopt a more informed decision on a limited number of consensus conditions.

Light/dark cycling (ISOS-LC)
Some PSC degradation modes have repeatedly been shown to be reversible (entirely or partly) in the dark (often referred to as metastability) 6 (Fig. 1c) and reversible photo-induced PCE increase with its subsequent decrease in the dark 54,82 ( Fig. 1b). In particular, reversible performance losses are attributed to cations redistribution 60 , metastable defects formation 67 or reversible chemical reactions 66 . The effects of PCE improvement under illumination after storage in the dark are commonly attributed to the neutralization of interfacial defects by photogenerated charge carriers or changes in the built-in electric field due to ion migration 84 . The PCE dynamics during the cycle were shown to change with status of the cell degradation 81 ; for example, the term "fatigue" was introduced for slowing down of PCE restoration with each consecutive cycle (Fig. 1b) 82 . Such metastability is attributed to the migration of ions, which is known to be pronounced in metal halide perovskites 83 . Reversible and irreversible degradation mechanisms may co-exist in a given PSC 79,81,85 .
ISOS protocols revised for perovskite PV should, therefore, include a group of light/dark cycling protocols to account for the recovery phenomena (ISOS-LC in Table 1

Electrical bias in the dark (ISOS-V)
Electrical bias is shown to cause PSC degradation (which is also affected by the presence of other stress factors) 12,13,86,87 . The degradation, in this case, is commonly initiated by ion migration 13 or charge carriers accumulation resulting in detrimental electrochemical reactions 9 .
Electric field is also shown to assist moisture-initiated perovskite degradation 87,88 since moisture ingress can result in the formation of hydrated perovskite phase containing mobile ions, whose drift accelerates the degradation 89 . Both positive (Fig. 1d) and negative (Fig. 1e) biases were shown to be potentially harmful 13,86 and might be realized during the solar panel operation. In our view, ISOS protocols revisited for perovskite photovoltaics should include ISOS-V group of testing in which the behavior of the cell is analyzed under exposure to certain electric bias in the dark (see Table 1).
Usually, the solar cell is kept near its MPP (i.e., positively biased with voltage < VOC), however, disconnected cells under illumination would be biased at open circuit voltages (typically, ~ 1 V for iodine single junction PSCs). We, therefore, suggest applying voltage equal to VMPP or VOC (as measured under AM1.5G illumination condition on fresh device) as a positive bias condition. Since bias-induced effects may have a threshold behavior 13 , we recommend voltages below Eg/q to avoid unnatural overstressing. Furthermore, a new protocol for assessing light-induced degradation (LID) in silicon modules will be introduced to IEC 61215. This will be applied to modules, which fail to stabilize in the efficiency measurement. However, instead of exposing the modules to light, they will be held under forward bias, such that the dark current density matches JSC. This condition was shown to mimic MPP under full sun exposure for silicon modules. While there is no similar data for PSCs, such stability test might be also useful.
In Three sophistication levels differ by the level of control over the sample temperature and atmosphere and required equipment, which is similar to that in corresponding ISOS-D protocols (Table 1) 14 .
Electrical bias application can cause a redistribution of charged species across the PSC, which might be reversible after the stress removal 13   intrinsic stability under a specific stressor conducted in an inert atmosphere (nitrogen, argon etc.). Index "I" is used at the end of the protocol name (see Table 2) to indicate the change of atmosphere in the corresponding test to the inert one, while other parameters are kept the same. perovskites, but only at a very low partial pressure. 105 Therefore, encapsulated PSCs may have a superior lifetime, compared to the unencapsulated samples, even for degradation experiments conducted in an inert atmosphere. Additionally, the environment in which the encapsulation is performed may also play a role. Thus, reporting the presence/absence and details of the encapsulation is mandatory also in "I" protocols.

Checklist for PSC stability studies
For comparison and reproducibility of results, it is crucial to report sufficient information about the aging experiments, in addition to a detailed description of the device preparation 18 . Table 3 is a suggested checklist for stability data reporting, in accordance with that required by 'Nature' journals for reporting PSC PV performance data 106 . We stress that even if a parameter is not controlled during the aging experiment (for example, temperature or relative humidity (stating if this is the value in the laboratory at room temperature, or in the aging apparatus) on the first level of ISOS protocols), it is still important to monitor and report the parameters listed in Table 3.
We recommend that researchers should specify the number of samples studied in each aging condition. According to critical analysis on the quality of PSC stability studies reported by Tiihonen and co-workers 18 , nearly half of the studies consider only 1 sample of each kind, which is particularly worrisome for PSCs typically characterized by relatively low reproducibility.
Ideally, statistics should be provided to account for sample-to-sample and batch-to-batch variations. The same work provides estimations of the desired sample size 18 .
Stability data are often reported as normalized parameter evolution with aging time, while only specifying the performance of a representative fresh device (champion or average).
Thus, the reported stabilities and efficiencies may be measured on different devices and, therefore, cannot be directly related. Any plot with normalized parameter variation should include the value to which it is normalized 23 .
Due to the ongoing development of best practices for J-V and efficiency measurements on PSCs [27][28][29][30]107 , the procedure for periodic measurements during the aging test should be clearly described. Typically, measurements of the J-V curve (or part thereof) are taken with certain periodicity depending on the characteristic degradation timescale of particular devices. Since J-V hysteresis is common in PSCs, steps should be taken to ensure the measurements are taken under (quasi) steady-state conditions. This can usually be achieved using a dynamic J-V approach 28 For every light source an irradiance in the range 800-1000 W/m² should ideally be applied, and the exact value of irradiance reported. Another important factor to be considered during aging experiments is the type and spectrum of the light source used. Table S1 shows a collection of the most stable reported PSCs to date that withstand more than 1000 h of lightsoaking while losing less than 15% of their initial PCE. As follows from Table S1, five main  Initial cell characterization: Current-voltage (J-V) curves of fresh devices, including voltage scan conditions (scan speed, direction, dwelling time, the number of power line cycles (NPLC), preconditioning etc.); stabilized photocurrent at MPP or MPP tracking data of fresh device; EQE/IPCE spectra (indicating the lock-in frequency and light bias if used, and if monochromatic light is smaller than active area, or larger and optical mask applied) and its comparison to JSC obtained from J-V data. Encapsulation: Wiring (materials, processing conditions, addition of a protective sealant); front and back side encapsulation layer(s) (materials (reference or composition, thickness), processing conditions (environment, temperature, duration)); edge sealant (materials (reference, thickness, width), processing conditions); geometry (rim (minimum distance between encapsulation edge and active area edge); device active area; picture or a scheme of the device).

Aging conditions:
The light source used in the aging experiment (light source type, intensity, spectrum, filters applied, calibration); light cycling (if applicable

Stability figures of merit and acceleration factors
The time corresponding to efficiency drop to below 80% of its initial magnitude is commonly referred to as T80. T80 often serves as a figure of merit (FOM) of solar cell stability; it, therefore, would be optimal as the minimum aging test time. Despite the apparent simplicity, there are several approaches 126,127 to determine device T80 (see Fig. 2 and Table S2 for a summary). It is, therefore, of vital importance to describe in details metrics used when reporting stability studies. In particular, the original ISOS protocols 14 suggested the use of stabilized TS80 time, the time during which the PCE decreases by 20% of its magnitude after an arbitrarily defined stabilization time (blue color lines in Fig. 2). This suggestion is based on the widely known shape of 'PCE versus time' curve in OPVs with rapid initial degradation ("burn-in") 127 followed by a stabilized region. While similar dynamics have been observed in some PSCs 11,128,129 , it is not a universal trend for these devices (Fig. 3) and, thus, should be The T80 of the most stable PSCs have been demonstrated to exceed 1,000 hours (42 days) or even 10,000 hours (>1 year) for 2D-3D perovskites 131 under certain stress conditions including illumination 58,128,132 . Considered recent advancements in PSCs stability (see also table S1), we strongly recommend that reviewers and journal editors discourage the use of word "stable" in the title of scientific papers if they don´t match the state of art in terms of efficiency losses and harshness of aging conditions (1000 h under 1 sun illumination with PCE decrease less than 20% as a rule of thumb). Notably, such long exposure times are challenging to realize.
If T80 is not reached, it is difficult to predict the lifetime based on the observed 'PCE versus time' trend, due to the variety of curve shapes observed (Fig. 3). In this case, we suggest performing aging for at least 1000 hours and using the PCE after 1000 hours of stress (η1000, as a percentage of the initial PCE) as stability FOM. If authors choose to apply any kind of 25 extrapolation to determine T80 or TS80, it must be clearly distinguished from the measured data.
In this case, it is recommended to limit the extrapolation times to below one order of magnitude larger than the actual aging time. As the stability of PSCs improves, it may become common to quote T95 values, the lifetime to degrade to 95% of the starting efficiency, for the more stable cells. This will also be in keeping with the IEC procedures, where the pass-criteria is for the modules to operate at >95% of their starting performance after the stress exposures.
The presence of reversible degradation in PSCs complicates their stability assessment.
There are no broadly accepted FOMs accounting for partial reversibility at the moment, however, some procedures are currently debated 22,54 . Recovery effects can be studied in two types of experiments: 1) Continuous aging followed by the performance tracing after the stress removal, or 2) cycled stress experiments. In the former case, it was suggested to correct T80 accounting for the amount of restoration occurring during the rest period 22 . In cycled experiments (such as ISOS-LC), an analogue of T80 metric might be introduced for the energy output per cycle 54 .  with OPVs under concentrated sunlight (both polymer-141 and small molecules-based 134,142 ) revealed that increasing the light concentration from 1 sun to 100 suns did not change the degradation mechanisms. Similar results were obtained using high intensity simulated sunlight 142,143 with OPV lifetimes of tens of years predicted in each case.

The significance of consensus for further research and data analyses
As ISOS protocols are intended for research purposes, a variety of testing procedures is suggested (see Table 1) concerning the four major degradation factors: atmosphere, temperature, electrical bias, and light. Table 4 shows a graphical representation of the corresponding aging test details, which can be used to study the effect of these factors and their combinations.  Table 4 schematically depict these relationships.
The many blank spaces in Table 4 can be filled with corresponding aging procedures (by analogy with additional ISOS tests suggested in section 3) in a straightforward manner. Other protocols can be constructed by varying "fixed" parameters (e.g. temperature, light intensity, R.H.). We do not aim to cover all the possibilities with ISOS protocols nor discredit studies with a systematic variation of a particular stressor. Nevertheless, these investigations would benefit from having common "reference points" with other studies conducted at different laboratories, device architectures, perovskite materials, and even different research questions in mind.
Consensual conditions, like ISOS protocols, may serve as such references.
Having unified procedures of stability studies with an easily machine-readable reporting structure could also lead to the creation of a large database that allows for more statistically reliable comparisons. Such a database is necessary to identify patterns in the data, and deduce 28 rules and heuristics for future work. Although the analysis of such large data sets will not be easy with the use of traditional approaches, statistical machine learning methods can enable important trends to be discovered 31,144,145 . These methods will be facilitated if abundant and uniformlyreported data sets are available. As indicated in Fig. 1 144 . However, this data subset resulted in better machine learning (ML) models compared to the entire dataset. The situation seems to be worse in stability related publications: in addition to unique measurement protocols, the storage conditions of the samples (which are also vital to assess the degradation process) are also not uniform, and not properly reported in a significant number of papers. These considerations make the requirement for consensus protocols even stricter for stability studies.
Determining the role of each stressor on device performance through ML methods could enable a direct comparison between results from different research laboratories around the world.
Briefly, by using information from aging measurements under the relevant stressors, as discussed above, one can optimize the steps needed for supervised learning algorithms. The PCE values obtained from aging tests are ideal training data and can be utilized for supervised learning. A so-called feature vector is implemented based on this data, and then the derived values are used as input for an artificial neural network or other machine learning tools. Ultimately, the reliability of ML strongly relies on having sufficient information (quantity and diversity) from aging tests that can provide an accurate prediction of device performance and degradation. Thus, the ML algorithms must be trained with extensive laboratory data, where prediction values can be compared to actual performance measurements. Once working, ML tools should provide knowledge extraction without the need to perform all tests presented in

Summary and Conclusions
This paper reports a consensus on procedures for perovskite solar cell stability studies discussed at the 11th International Summit on Organic and Hybrid Photovoltaics Stability (ISOS-11). Suggested protocols primarily rely on the original ISOS standards developed for OPV 14 , which have been shown to be highly relevant for uncovering various degradation pathways in PSCs. We further suggest extending the set of protocols in accordance with specific stability features of PSCs, including: 1) light/dark cycling (ISOS-LC) mimicking the diurnal cycle; 2) study of the cells behavior under continuously applied bias in the dark; 3) protocols for intrinsic solar cells stability studies (indexed with "I"). These tests have already proven their usefulness in understanding PSCs failure modes. We also propose a checklist for uniformly reporting results of PSC stability studies. This list ensures that the research can be reproduced and compared with results from other laboratories. Procedures and good practices for stability tests are discussed.
We expect that the guidelines for conducting and reporting stability studies described in this paper will improve comparability between data from different laboratories and device architectures. The set of procedures and practices suggested here serve as an intermediate stage in perovskite solar cells technology maturing, aimed at the identification of degradation pathways and the prospects for their mitigation. If broadly accepted, these quasi-standards would significantly speed up stability data accumulation, which can be utilized for predictive machine learning to further facilitate the development of stable and reliable PV devices.