Baselines for Lifetime of Organic Solar Cells

The process of accurately gauging lifetime improvements in organic photovoltaics (OPVs) or other similar emerging technologies, such as perovskites solar cells is still a major challenge. The presented work is part of a larger effort of developing a worldwide database of lifetimes that can help establishing reference baselines of stability performance for OPVs and other emerging PV technologies, which can then be utilized for pass‐fail testing standards and predicting tools. The study constitutes scanning of literature articles related to stability data of OPVs, reported until mid‐2015 and collecting the reported data into a database. A generic lifetime marker is utilized for rating the stability of various reported devices. The collected data is combined with an earlier developed and reported database, which was based on articles reported until mid‐2013. The extended database is utilized for establishing the baselines of lifetime for OPVs tested under different conditions. The work also provides the recent progress in stability of unencapsulated OPVs with different architectures, as well as presents the updated diagram of the reported record lifetimes of OPVs. The presented work is another step forward towards the development of pass‐fail testing standards and lifetime prediction tools for emerging PV technologies.


Introduction
There exists a set of international test standards with pass-fail criteria (typically published by IEC and ASTM standards organizations) in the photovoltaic (PV) world and if the new manufactured (Si based) PV pass these tests, then there is a high probability that the product will survive 20-30 years in the end use environment. [1] Such a confidence however is mostly originating from the many years of experience in the field and the large set of tested Si based modules and therefore is very much customized for these types of technologies. Meanwhile, rapidly developing emerging PV technologies, such as organic photovoltaics (OPV), dye sensitized solar cells (DSSC), perovskite solar cells (PVSK) and others alike, due to their relatively young age still lack the field experience and generated reproducible data for building test procedures, which if passed, would allow to state with confidence the necessary reliability of these products in the end use environment. The lessons learned from the inorganic technologies cannot be utilized either, since the emerging PV technologies considerably differ in architecture from their inorganic counterparts [2] and due to their increased sensitivity towards the testing environments, [3][4][5][6] the common testing standards are not suitable for these technologies. [7] These challenges however have received significant attention in the recent years especially in the field of OPVs. In particular, at the sequence of International Summits on Organic solar cell Stability (ISOS) the issue of reliable testing of OPVs was thoroughly addressed, and in 2011 recommendations were published based on the consensus of a large number of renowned research groups in the field, that outlined guidelines for reliable stability testing of organic solar cells. [8] The guidelines set certain criteria on the test conditions and therefore allowed reproducibly recording the aging of the samples under specific controllable conditions, in both indoor and outdoor testing environments. While this very much helped in reducing the spread in the testing procedures among the different groups, as well as improving the reproducibility of the reported device lifetimes, [9] the question still remained: how to develop a methodology that would allow either conducting pass-fail tests and building confidence around the product durability or predicting its lifetime in end use environment, based on accelerated testing. Significant efforts are put today towards resolving this. In particular, the group in Technical University of Denmark has recently demonstrated an approach based on statistical analyses, where a large variety of OPV samples were tested under different ISOS tests and the average lifetime of the samples under each test condition was determined. [10,11] The values were then used to calculate the ratio between the accelerated and real outdoor tests, which could potentially be utilized for predicting device performance. However, despite the relatively large datasets the studies were limited to only a few architectural variations and while they well demonstrated the concept, the established values could not be regarded as sufficiently generic for application beyond the reported studies.
The works however continued and recently a manuscript was published by the same group, where the same statistical approach was utilized for analyzing the entire literature related to stability of OPVs. [12] In that study, the authors collected and analyzed all the articles reported until March 2013 covering Suren A. Gevorgyan is a senior scientist at the Technical University of Denmark. He was born in Armenia in 1982 and received his PhD from the Technical University of Denmark in 2010. From the early years of his career Suren has been working in the field of organic photovoltaics and specializing in characterization and stability improvements of the devices. His main research interests include standard testing, electro-optical characterization and device engineering of organic and perovskite solar cells.
Nieves Espinosa holds a position as researcher at Universidad Politécnica de Cartagena. The focus of her research has been on the eco-design of selected energy technologies, such as organic photovoltaics, products and processes investigating the life cycle impacts associated. Special attention is given to their implementation and integration in buildings or smart grids. Currently she is group leader in an EU COST Action StableNextSol devoted to unraveling the degradation mechanisms of organic solar cells.
Laura Ciammaruchi received her PhD degree from the University of Rome "Tor Vergata" in 2014, with a thesis titled "Studies on Stability and Degradation of Hybrid and Organic Solar Cells". Half of her doctoral activity was carried out at the University of Rochester, NY (USA) in Prof. Ching W. Tang's lab. Her expertise includes fabrication and characterization of various OPV technologies, with a specific focus on stability and degradation analyses.
stability studies of OPVs (total of 2500 articles). A generic lifetime marker was developed that allowed gauging and intercomparing the stability of the different OPV devices reported in these articles. The lifetime of the samples was categorized depending on the device type and architecture, as well as test conditions, which helped to better understand and elucidate the typical bottlenecks for the device stability. The study additionally helped to establish averages for the lifetimes of OPVs, (3 of 9) 1600910 wileyonlinelibrary.com tested under different test conditions. However, due to the limited amount of data for certain test conditions (especially for outdoor data) some averages lacked statistical significance and thus, could not be regarded as reliable baselines for device lifetime. The initiative therefore continued with the purpose of further enriching the lifetime database with both data reported in literature and experimental data and converting the database into a generic hub of baselines for the lifetime of OPVs and other emerging PV technologies alike, and utilizing the data for establishing test standards and prediction tools.
This work is a complementary to the aforementioned earlier reported study and is presenting the results of the follow-up literature analyses for an additional period from March-2013 until March-2015. The data analysis provides more solidified distributions of the lifetimes and allows drawing conclusions on the baselines for the OPV lifetimes tested under specific conditions. An updated version of the lifetime progress diagram is presented as well.

Literature Data
The data collection procedure is explained in detail elsewhere. [12] Briefly, the articles were identified using the search engine ScienceDirect and exploring expressions based on different combinations of words such as polymer, plastic, organic, solar cells, photovoltaics, stability, aging and lifetime. For the clarity, the articles, analyzed in the earlier study, mentioned in the introduction, will be referred to as "2013 dataset" (since it was limited to reports until year 2013). The new dataset, which is mostly based on articles published between March-2013 and March 2015 and a few papers published in conference proceedings dating before 2013, will be referred to as "2015 dataset". The new dataset was inserted into an online database for further analyses. The total number of articles in 2015 dataset was 2286, out of which 303 contained actual lifetime data, while the rest only discussed the theory behind the stability issue. The 303 articles presented aging curves for a total of 983 devices, which are called data points. For the comparison, for 2013 dataset a total of 2500 articles were scanned, which also revealed precisely 303 articles with reported experimental lifetime data. This shows the increasing interest towards the OPV stability in the recent years.

Lifetime Determination
The online database with the new articles (hosted at http://plasticphotovoltaics.org/) was shared among and analyzed by different groups from consortia of the COST Action project StableNextSol (http://www.stablenextsol.eu). The analysis involved scanning each article individually, identifying whether the article contains experimental lifetime data, registering the reported data by filling the database with the reported sample structures, encapsulation, testing conditions and determining the lifetime from the reported aging curves. The lifetime T 80 is defined as the time when the device is degraded by 20% from the initial value. Since the OPV devices however in most cases show rapid initial decay followed by slower degradation, in the ISOS guidelines it was proposed to identify two starting points, E 0 and E S where the second point is defined at a stage when the aging is changed to a slower pace. Accordingly, the two values T 80 and T S80 calculated from the two starting points define the lifetime of the device at the initial and at the stabilized stages. Figure 1a demonstrates a number of commonly observed aging curves for OPVs and the reason for such diversity in shapes is the multitude of aging mechanisms and competing annealing processes that take place at the same time in the device when the latter is exposed to different aging conditions. Figure 1a and Table 1 explain how for different shapes the two lifetime parameters are identified. The approach was applied for identifying the lifetime for all the reported aging curves in the literature. In case the T 80 value was not reached within the duration of the experiment, the time of the last measurement T final was chosen to represent the lifetime (the minimum possible lifetime). Figure 1b together with Table 1 explain how after identification of the two lifetime values T 80 and T S80 the best one is identified to represent the device by simply choosing the one where the most amount of energy is produced. Graphically this will correspond to the highlighted surface areas on the plot in Figure 1b, denoted as I and II. A more detailed explanation is provided in the previous study. [12] The established database is publicly available at http://plasticphotovoltaics.org/lifetime-predictor.html, where an online interface can be found that allows analyzing and reproducing the collected data and utilizing special filters for specific sorting of the data. An instruction video is additionally provided in the website for navigating though the tool and the database.

Status of OPV Lifetime
The data collected from 2015 dataset was compared with the older dataset. The comparison revealed no significant difference in the data distribution between the two, but rather one complemented the other. The two datasets were therefore combined, which improved the intercomparison and baselining of the lifetime distributions under different test conditions. The best section is mathematically defined by the largest surface area among the two gray regions denoted with I and II. Reproduced with permission. [12] Copyright 2015, John Wiley and Sons.
wileyonlinelibrary.com Figure 2 shows the reported lifetime values plotted against each year, where the triangles show the data tested under light and the circles represent the dark tests. The solid line is showing the total number of reports for each year (corresponding to the right axes). The plot clearly demonstrates the significant increase in the device stability and quantity of the reported lifetime data in the recent years with the total number of data points reaching beyond 300 for the past few years, which corresponds to more than 100 articles per year (given that one article contains approximately 3 data points). This is a clear indication of the significantly growing importance of the issue of lifetime in the recent years.

Baseline for Lifetime
The combination of the two datasets significantly increased the total amount of data points and therefore improved the statistical significance of the lifetime distributions for the samples tested under different test conditions. This enabled the possibility for establishing baselines based on such distributions. In order to do so, the data were categorized according to four groups, similar to the earlier reported study [12] : Group 1 and Group 2 represented the unencapsulated samples (samples that did not contain extra packaging layers) tested under light and in dark, respectively, and Group 3 and Group 4 hosted the encapsulated samples (samples that were packaged with barrier materials) correspondingly tested under light and in dark. The tests under light were further distinguished by: • indoor soaking under light source with spectrum close to AM 1.5 and intensity close to 1 sun • indoor exposure to low UV or low intensity light • outdoor testing under real sun  Table 1. The list of steps for determining the lifetime marker. Reproduced with permission. [12] Copyright 2015, John Wiley and Sons.
Parameters Method *Determination of starting point E 0 &T 0 T 0 & E 0 pair is either chosen at the first measurement point or if the curve has an initial increase followed by a reduction (such as the curve 3 in Figure 1a) then T 0 & E 0 is set at the maximum point.
E 0 -initial performance T 0 -initial time

Determination of stabilized section E S & T S
If after a certain point the aging curve enters into a more stable phase (commonly observed during solar cell aging), then a second pair of starting values T S & E S is identified, typically chosen at a point from where the aging rate almost doesn't change anymore, as shown on curve 1 in Figure 1a). E S -performance at the start of stabilized section T S -starting time of stabilized section

Determination of T 80 and T S80
T 80 (or if applicable T S80 ) is determined by subtracting T 0 (or T S ) from the time when 80% of E 0 (or E S ) is reached. Figure 1b highlights the areas determined by T 80 and T S80 T 80 -time when performance reaches 80% of E 0 T S80 -time when performance reaches 80% of E S

Lifetime marker [E 0 ;T 80 ] or [E S ;T S80 ]
The largest area among I and II in Figure 1b (part of the curve where the sample produces the largest amount of energy) will then determine the pair that will describe the lifetime. The simple geometrical calculations reveal that the ratio of the areas of the trapezoids I and II are proportional to the ratio of the areas of the rectangles defined by the products of E 0 × T 80 and E S × T S80 .

Exceptions
Exceptions are made in the following cases: • If E S is less than half of E 0 , in which case the sample is considered to have degraded before stabilization (see curve 2 in Figure 1a), then [E 0 ;T 80 ] is chosen by default to represent the lifetime.
• If the measurements has been stopped prior to reaching the 80% threshold then "T final -T 0 " or "T final -T S ", where T final is the point of last measurement (see curve 4 in Figure 1a) is chosen instead to represent the minimum possible lifetime.
(5 of 9) 1600910 wileyonlinelibrary.com data points. Each test category is also associated with, but not limited to the ISOS testing procedures shown in the legends. For each data distribution peak, "median" defines the region with the most commonly reported lifetimes and "maximum" defines the region with the highest lifetime values, which are highlighted by red and green bands, respectively. The corresponding time-ranges for the median and maximum are also listed in the table on the right lower corner of Figure 3.
The established baselines can serve as references for the performance under given test conditions for any newly produced sample: • If the sample outperforms the median, then the sample has an improved stability. • If the performance is in the maximum region or beyond, then the sample has an outstanding or record lifetime, respectively.
Group 2 of unencapsulated samples tested in dark (Figure 3 (e)) contains two medians associated with normal and inverted device layouts, which are discussed in the next section. From the right plots in Figure 3 it becomes obvious that while the unencapsulated samples tested in dark may show impressive stability reaching beyond a few months, the large majority of the devices fail within minutes when tested under light. Therefore, it is highly advisable to use both condition or as a minimum, the light tests in order to truly assess the potential of the manufactured device in terms of stability. For the encapsulated samples when looking at the outdoor tests, the most commonly reported lifetimes are within a few months and only a few are Adv. Energy Mater. 2016, 6, 1600910 www.MaterialsViews.com www.advenergymat.de  reported to reach a few years. Instead, in the case of indoor light soaking a few weeks appear to be the most common duration of the lifetime and only in a few cases samples lasting longer than a couple of seasons were reported. As a word of precaution, an attempt to predict the lifetime of the sample in outdoor test conditions, based on the ratios of the indoor light soaking and outdoor tests may lead to erroneous results, since one is not the acceleration of the other. For simulation of the outdoor tests a more complicated set of accelerated tests are required, such as a combination of a number of ISOS test procedures. Unfortunately, at this stage the database presented here does not contain sufficient data for each individual ISOS test procedure for in depth analyses. However, with the gradual increase of the database with time the intercomparison of the data for ISOS will also become possible, enabling the development of such a prediction tool.
Consequently, the presented baselines should mainly be regarded as generic reference points for lifetime of organic photovoltaics under given test conditions according to the aforementioned grouping.

Normal vs Inverted Structures
In Figure 3e the unencapsulated samples tested in the dark show two distinct peaks. These correspond to devices with normal (also known as conventional) and inverted architectures. The former typically employs aluminum back electrodes, while the latter has silver or gold based electrodes. Figure 4 similarly to Figure 3 utilizes the time blocks of the logarithmic scale with base 4 to break down the data distribution for conventional and inverted devices tested in dark. The upper plot corresponds to the samples with normal configuration and the plot in the bottom represents the inverted devices. The squares and circles correspond to unencapsulated and encapsulated devices, respectively. From the figure it is apparent that there is a significant difference in the stability between the two layouts for unencapsulated samples, which is less pronounced in the case of encapsulated samples. It has been established earlier that the normal structures are significantly less resistant towards moisture due to the high sensitivity of the aluminum, [13][14][15][16] therefore they show inferior stability when tested in the dark. When encapsulated, the sample becomes well protected from the humid environment and therefore the reaction of the electrode with moisture is significantly reduced. As a result, the encapsulation of the normal structure devices has a major impact on the stability for normal devices, while in the case of inverted structures, since the samples are already significantly stable in the dark without extra protection, the role of encapsulation does not seem to be significant, as can be seen in Figure 4. The plot demonstrates how the device geometry may significantly alter the stability and therefore precautions are necessary when comparing the lifetime of devices with different structures and while reporting stability it is always useful to report the full structure and composition of the devices. In the case of indoor light tests (not shown here), there is no obvious difference in the stability of the two structures for samples both with and without encapsulation, which is possibly because the heat produced by the light source creates a rather dry environment around the sample diminishing the effect of humidity.

Winning Structures
A section in the earlier publication of 2013 dataset analyses [12] listed a number of device architectures with reported best stabilities without utilization of external packaging. It is important to clarify here that this category of samples may still contain certain protective layers, such as a substrate from one side and a solid metal electrode from the other and this is actually the case for the most stable samples reported for this category in the previous study. Nevertheless, the difference from the other categories is that no external packaging is utilized. In addition, for the same reason the term "intrinsic stability" used for describing the stability of such unencapsulated samples in the previous study will not be used here in order not to mislead the reader. Similar to the previous study, in this work a number of reports with unencapsulated samples of outstanding stability were registered, which are listed in Table 2. The table highlights the structures of the reported samples tested under light or in dark and their corresponding lifetime and efficiency values. The most impressive report corresponds to the sample tested under light showing an impressive lifetime of 96 days. [17] Unfortunately, the details of the top electrode configuration were not reported, but it was stated that it contained a combination of different metals. It is also worth mentioning that one of the samples tested in dark and showing an outstanding stability of 120 days, was produced in a roll-to-roll compatible process utilizing coating and printing techniques. [10] Nevertheless, despite a number of reports of impressive stability, producing encapsulation free samples in a roll-to-roll compatible process with sufficient stability under light test presents a serious challenge that still needs to be addressed. [18] Adv. Energy Mater. 2016, 6, 1600910 www.MaterialsViews.com www.advenergymat.de

Plot of the Record Lifetimes
In the previous report of 2013 dataset, [12] a so-called "lifetime progress diagram" was presented, which highlighted the best reported lifetimes of organic solar cells tested under different conditions. This diagram has been updated by additions from the new dataset and is now presented in Figure 5. The references to the reports are provided in the table below the diagram. The data is differentiated according to four test conditions: dark, indoor light soaking with AM1.5G spectrum light, light soaking with low UV light source and outdoor exposure. The conditions are differentiated by the colors in the plot. The arrows show the data points, which were defined by the last measurement of the experiment (T final ), since the device did not reach the T 80 lifetime during the experiment. The data points are connected with dashed lines in order to highlight Adv. Energy Mater. 2016, 6, 1600910 www.MaterialsViews.com www.advenergymat.de

Conclusions
This article has presented the results of the analysis of published literature related to the stability of organic solar cells reported in the recent years. The progress in the number of reports per year dealing with the lifetime of OPVs has been shown, which asserts the ever increasing interest towards resolving the stability issue of this technology. From the extended dataset, the distribution of the reported lifetime data on time axes was established for OPVs tested under five test conditions, such as shelf life, high humidity test, outdoor exposure, indoor light soaking with full spectrum and low UV spectrum. The peaks of the distributions were utilized for establishing baselines for OPV lifetime under different aging conditions, which can serve as a reference for determining whether a newly reported data has an improved or record lifetime compared to commonly reported values. It was revealed that the most commonly reported lifetime values of OPVs are laying in the vicinity of months in outdoor tests and only weeks for indoor light soaking. A few reports however demonstrated lifetimes of years under outdoor conditions. It was also observed that the unencapsulated samples tested in dark revealed two distinct peaks for lifetime distribution, which was shown to be caused by the difference between the conventional and inverted devices structures. This highlighted the importance of considering the device structure when comparing and reporting device stability. In addition, a list of devices without external packaging with outstanding stability has been listed together with the detailed analysis of their structures. The most stable unencapsulated sample was shown to last up to 96 days under light exposure with multilayer back electrode protection. The updated version of the diagram of the record stabilities has been presented as well showing the significant improvements in the lifetime of OPVs in recent years and highlighting the fact that OPVs are rapidly approaching the durability level necessary for the commercialization of the technology. This work constitutes a step forward towards the development of pass-fail testing standards and prediction tools for reliable assessment of the sample durability. The major challenge is still the significant lack of experimental data for each individual ISOS testing condition and in particular for the outdoor tests. Therefore, this work is another call to the community of emerging PV technologies for joining the forces and further developing the database into an international data hub with extended datasets.