Alternative preconditioning by utilization of a thin ﬁ lm module ’ s dark diode ﬁ ngerprint

. Although the continuously advancing silicon wafer-based modules dominate the commercial PV landscape, thin ﬁ lm technologies have not lost any of their attraction, especially in areas where their advantages count, like light weight, ﬂ exibility, and easy manufacturing. This has been the case for chalcogenides in the past and it will be for coming perovskite-based materials, whether as stand-alone, in multi-or heterojunction devices. Unfortunately, many thin ﬁ lm technologies suffer from metastability, i.e., their physical properties change temporarily with storage, transport or operating conditions, on time scales from hours to months. For this reason, preconditioning is crucial, before reliably evaluating such a module ’ s performance. Presently, the respective preconditioning standards are exclusively focused on illumination-induced stabilization of the module ’ s power at the maximum power point ( P MPP ). However, using P MPP as the only marker might not be the wisest choice. First, the P MPP is basically a black box, i.e., a module may show the same temporary power value at times, while being in very different condition if one looked closely on its device physics then. This may lead to false assumptions about the module ’ s quality. Second, aiming for the highest stable P MPP of a module might not always be the desired goal, e.g., in warranty cases where the actual ﬁ eld performance of a module is in question and not how it would behave in perfect state after standard preconditioning. To overcome these limitations of present preconditioning standards, an alternative additional approach is required. In this report, we give a brief view on the inevitable shortcomings of present methods for thin ﬁ lm modules and demonstrate how the dark current characteristic of a thin ﬁ lm module can be used like a ﬁ ngerprint instead, representing its device physics that de ﬁ ne its actual state. Whereas in PV research, dark IV curves are commonly analyzed in detail for hints on charge transport mechanisms, interface properties or semiconductor degradation in the device, such effort would be inconvenient and unnecessary for fast-track commercial module testing. Here, we suggest focusing merely on the effective device properties, which are re ﬂ ected quantitatively in the diode-parameters. The goal is to feed a recorded module dark current curve into an automated mathematical procedure, which ﬁ ts the data to the double-diode model, enabling the extraction of the diode parameter-set. With this as a marker, instead of using solely P MPP during preconditioning treatments, it is much more likely that the desired previous physical state of a module is really reinstated. Additionally, the described dark current approach is conveniently independent of a light source ’ s properties and insensitive to module soiling. The results presented here, give a ﬁ rst impression on the potential that such a method could have, showcasing effects of dark storage degradation and their recovery by illumination or bias-induced preconditioning on the dark current characteristics of individual CdTe and CIGS commercial PV-modules of different generations and manufacturers.


Introduction
Metastability in commercial thin film PV modules are an ongoing issue and will continue with the newly upcoming technologies like perovskites and combined devices involving such technologies, like multijunctions or heterojunctions.
The reasons for metastability lie simply in the character of the respective absorber materials and their deposition.Polycrystalline composites are prone to bulk-lattice and crystal surface defects, which inevitably facilitate diffusion and ion/molecule migration along grain boundaries and therewith promote trap-assisted Shockley-Read-Hall (SRH) recombination, unwanted leakage pathways, energetic mismatch, interfacial barriers or changes in dopant type/ density [1,2].One phenomenon typically seen in chalcogen-ide photovoltaic devices is the "electronic doping" of CdS, i.e., the semiconductor changes its nature from p-type in the dark, to n-type under illumination and vice-versa, controlled by its deep defect levels [3].Therefore, the superposition principle, i.e., photocurrent being the sum of diode current and short-circuit current, is not valid for chalcogenide PV.Instead, bulk and interface recombination currents depend on light spectrum, intensity and voltage.Deducted diode saturation currents can differ by several orders of magnitude between illumination and dark, as do the resistances [4].The activation or promotion of metastability effects can be thermally or depending on illumination or bias-promoted during operation or initiated by darkness.As result, the behavior of metastable thin film devices might not be as expected as when dealing with classic wafer-based technologies.To compensate for those effects, preconditioning standards have been introduced as mandatory procedure, to temporarily stabilize the module's behavior, before it is characterized.The standard IEC 61215 addresses such stabilization procedures.The major accepted standard method is by illumination with natural/simulated sunlight, but other techniques (e.g., by bias) may be considered after validation.While minimum requirements on specs and irradiance for illumination stabilization are universal for all module types, the IEC 61215-1 also gives special requirements for the individual technologies, i.e., c-Si (part 1-1) [5], CdTe (1-2) [6], a-Si (1-3) [7] and CIGS (1)(2)(3)(4) [8] modules, which vary by illumination dosages, duration and stability criteria.These procedures aim exclusively for a stable maximum power value P MAX of the module, regardless of the state it was in before the stabilization or ever.However, there are situations, like in quality control or warranty cases, where it would be desirable not to obtain a module's best, but to reinstate a particular previous state of performance.For instance, to prove underperformance of modules of a PV plant, samples could be brought to a laboratory with a standard light source.But during transport and waiting time in undefined light, darkness or heat, the modules might be affected further, e.g., by dark degradation.If standard preconditioning before testing was applied in this case, the modules' performance might exceed that one observed in the field and any underperformance becomes unprovable.This example demonstrates the need for an alternative preconditioning concept, which allows to determine the state of a module before an altering event and use the difference as a guide to apply illumination/bias/heat to recover that previous state later.The P MPP is too crude for that purpose, especially for module technologies whose physics strongly depend on illumination, like chalcogenides.The solution to this problem lies in the evaluation of the dark diode behavior of the module instead.Dark current characteristics are a standard diagnostic tool in solar cell research, as they allow a very defined analysis of device failures and mechanisms, ideal to unveil issues of charge mobility, interface degradation, charge transfer, traps or dopant migration [9][10][11].For modules on the other hand, the utilization of dark-IVs has been rather sparse, i.e., mostly limited to module development and quality testing [12] or research outdoor studies [13], despite the valuable information they can provide.However, with advancing remote wireless monitoring and automated AI supported data analysis tools, dark current characterization of modules has been suggested recently as a valuable tool for every-day plant diagnostics [14].
To account for all kinds of "non-idealities", the dark current behavior of a thin film solar cell is usually described by the two-diode model, i.e., an equivalent circuit comprising a main diode D 1 , an additional "weak" diode D 2 , a parallel shunt-resistor R SH and a resistor in series R S .Its current-voltage response is represented by the corresponding two-diode equation: This formula includes each diode's saturation current (I 01 , I 02 ), the series resistance (R S ), the shunt resistance (R SH ) and each diode's ideality factor (n 1 , n 2 ) [15].The main diode D 1 is the "ideal" diode contribution, representing the diffusion current.The "weak" diode D 2 represents recombination processes in the depletion zone and tunneling processes.Ideally, the diode ideality factors of the two diodes would be n 1 = 1 and 1 < n 2 < 2. However, for CdTe and CIGS devices, it was observed to be n 2 >2, due to a large density of bulk traps from lattice defects and surface defects along grain boundaries, that both lead to SRH recombination via mid-gap and interface states [16][17][18].The series resistance R S in chalcogenide devices has many contributions from bulk resistance of the semiconductor (doping, defects, impurities), interface resistance between layers and the contact resistance at the metallic contacts, like between the CIGS semiconductor and the molybdenum electrode, caused by energetic mismatch [16].The shunt resistance R SH comes from parasitic leakage current pathways, which for CdTe and CIGS are caused by pin holes and unwanted conductivity along the grain boundaries e.g., by metal diffusion [18,19].These effects are affected by fluctuations during fabrication or treatment, so variation of R SH by 2-3 orders of magnitude in identical batches are typical [20].It should be noted that a few reports mention observed non-ohmic shunt contributions in the dark current characteristic of chalcogenide devices, which could not be fit to the model as is.In those cases the authors suspect space-charge limited current (SCLC), caused by imbalanced charge transport between holes and electrons in the device and altered their model by addition of another term, relating to film thickness, trap distribution and conductivity [18,19].However, in most cases the unaltered two-diode model is considered sufficient.Additional to the effects found in both chalcogenide thin film devices, the presence of copper in CdTe devices leads to several interesting exclusive effects in respective cells.Therein, Cu is required to form an ohmic contact between CdTe and the back contact (via Cu x Te 1-x phases).While moderate diffusion of Cu into CdTe has even shown to improve the device, fast movement along lattice defects, leads easily to its migration towards grain boundaries or layer interfaces, where Cu creates recombination centers or leakage pathways.Further, the consequential decreasing presence of Cu near the back contact leads to an energy barrier and thereby higher series resistance [21].
Photovoltaic modules are basically a superposition of multiple individual solar cells connected in series and in parallel.Due to the special way of thin film module manufacturing by alternating large-area deposition and lasering steps, the resulting structure consists of 100-300 narrow stripe-shaped individual cells.Depending on the desired operating voltage of the module, they are organized in just one or in two parallel strings (e.g., half and half) of cells in series.Like a solar cell, also a module can be represented by an equivalent circuit correspondingly to the two-diode model, as shown in Figure 1.
Assuming that all cells behave similarly, the resulting two-diode equation for modules can be written as follows, containing the number of cells in series N S and of parallel strings N P as additional factors: The two-diode equation, whether for cell or module, is implicit and therefore difficult to solve to extract the diode parameters.Plenty of literature is dedicated to that problem alone, divided into analytical, numerical deterministic and numerical stochastic approaches [22][23][24].While an analytical solution is only feasible with considerable simplifications and partial solutions for sampling points (e.g., @V OC , @V = 0 from photo-IVs), the numerical way is far more straight forward, but quite sensitive to the setting of parameter boundaries.For a qualitative evaluation of a dark current characteristics of a cell or module according to the two-diode model, the data are plotted semi-logarithmically and the curve visually divided into four different regions: (i) the symmetric pit of the shunt resistance dominated region at low bias (forward and reverse), (ii) the recombination dominated (weak diode) region at low intermediate forward bias, (iii) the diffusion dominated (ideal diode) region at higher intermediate forward bias and finally (iv) the series resistance dominated region at high forward bias [21].Occasionally, the visual distinction between the two different diode regions is not very clear.In those cases, the diodes' region is treated as one and discussed [25].

Experimental details 2.1 Modules under test
As showcase samples for this study, we chose 5 different module models of CdTe thin film and CIGS thin film cell types (see Tab. 1).The donor/acceptor structure of the modules' cells is usually not entirely disclosed by the manufacturer, therefore some details are presumed from sparse disclosed information and the known technical state-of-the-art [26][27][28][29][30][31].
Three different models of CdTe modules have been obtained from First Solar to account for technology jumps between generations and for manufacturing variations at different production sites.The most significant difference between the FS-2 (2006/2008) and FS-4 (2019) series is the replacement of CdS and the introduction of a graded Cd (Te,Se) layer to improve absorption and carrier life time.A comparison between two nominally identical old generation models from their US-based (FS-265 series) and their Malaysia-based (FS-262 series) production site was encouraged by the historic recall of First Solar's mid2008/mid2009 US-batch due to production issues.Regarding CIGS, two different models were obtained, an older generation type by Würth Solar (2011) and a new generation model by Solar Frontier (2019).Again, also in this case the new model differs to the old one by the established enhancements of the cell composition, such as replacement of the CdS layer and introduction of graded absorber layer, additionally the difference of Solar Frontiers deposition two-step sputtering and selenization technique to Würth Solar's one-step co-evaporation.For each model test series, 5 modules of consecutive serial numbers of the same production batch have been purchased where possible.While the younger modules were in pristine "out of the box" condition, the older types were second hand and therefore with unknown history.To account for unknown history, storage and transport conditions, all modules were kept in dark storage for three months before initial characterization.

Instrumentation and methods
Dark current characteristics of the modules were recorded with a Keithley 2612B system source meter.Voltage sweeps were carried out in forward direction with measurement delay of 100 ms between voltage setting and current measurement (typical duration of a voltage sweep with 15000 points = 150 s).Photocurrent characteristics were recorded with a Mencke&Tegtmeyer PV-KLA capacitive curve tracer (measurement accuracy ± 0.1%, typical duration of measurement with 15000 points = 2 s).For outdoor operation of the modules, they were mounted on a roof platform (16 m height) on a 37°tilted rack facing southward direction.The modules were kept at maximum power point, the power DC/AC converted and fed into the local power grid by one AE-Conversion micro -inverter each, fitted to the respective module's feed-in DC values (models INV350/60 and INV500/90).The microinverters were connected to a datalogger, continuously recording the modules' MPP power, current and voltage values.The setup with individual microinverters allowed to run a combination of totally different module types and numbers at the same time.Irradiation on the outdoor rack was logged in the module plane by a Mencke & Tegtmeyer reference cell model Si-mV-85-Pt1000-4L-E (measurement accuracy ±6%).The temperature of the modules was measured by Omega-Engineering Pt100 thermo sensors attached to the back glass (using heat conducting paste Noctua NT-H2) and recorded using a Graphtec midi-Logger B. Friedel and S. Winter: EPJ Photovoltaics 15, 3 (2024) GL840 (measurement accuracy ± 0.1%).For laboratory photocurrent and for EQE measurements, an LED-based Wavelabs module solar simulator model SINUS-2100p served as standard (AAA+) light source.During dark storage and accompanying characterization, the modules were kept in perfectly dark fitted panel flight cases.The white light preconditioning was either performed outdoors in sun light on the rack or on a ventilation-cooled lab stage with an LED based large area white light source (irradiance of 850 W/m 2 ).The latter setup was also used to perform thermal studies.Bias-based preconditioning was performed by repeated voltage sweeps (0 V to max. 120 V) with a Keithley 2612B system source meter.To track changes in dark current characteristics, the curves were subject to comparative visual analysis of their respective shunt resistance, diode/recombination and series resistance dominated regions.Additionally, dark current characteristics were fitted to two-diode model for PV modules to compare the quantitative diode parameters.Therefore, the implicit formula was implemented in Python and fitted numerically to the measured data using the trust-region-method.

Temperature-related effects on outdoor module testing
Photocurrent characteristics of PV modules À unlike dark current À are affected by a combination of temperature and illumination caused effects.A light source for PV operation or measurement is characterized by its spatial homogeneity, temporal stability and spectral match to the AM1.5G spectrum.Especially outdoor sunlight is subject to fast fluctuations in irradiance and spectrum caused by atmospheric phenomena.These fluctuations lead not only to a quick change in photocurrent, but also affect the module temperature by thermalization and non-radiative recombination processes [32].In order to get comparable data, calculated corrections towards standard test conditions (STC) are a usual approach, using the synchronously measured in-plane irradiance and module's (back-side) temperature.However, while fast fluctuations in irradiance are easily instantly monitored, the resulting measurable temperature changes are always delayed, which might compromise the corrections and lead to false results.This effect is expected to be especially pronounced for glass-glass modules, due to their relatively thin absorber layer (<10 mm) compared to the thick enclosing glass-sheets (3.3 mm).
To get an idea of the delay time-scales for our chalcogenide glass-glass test modules, we studied their thermal transient behavior.For that purpose, a module was equipped with three temperature sensors on the glass back-sheet (T1, T2, T3) and the transient thermal response was recorded upon different durations of white-light illumination at an irradiance of 850 W/m 2 .Figure 2 shows the thermal transient behavior for light exposures of 60 s, 10 min and 60 min for a CdTe module of the FS-4117 series.
For the shortest illumination period of 60 s (Fig. 2a), it can be seen that the module temperature rises linearly from room temperature by a few degrees and even continues to  From a linear approximation of the rise time (magnified onset in Fig. 2b), we obtain a heating rate of 2.4°/min.The module's cooling on the other hand is falling bi-exponentially with rate constants between 0.0003 s −1 and 0.002 s −1 , depending on the sensor position on the module.For longer illumination durations, it can be seen that the module temperature does not continue to increase linearly but starts curving after a few minutes (Fig. 2c).Eventually, starting to show for 60 min of illumination, the measured module temperature approaches a level of saturation (Fig. 2d).If that level was reached, it would mark the balance between light-induced heating and the cooling losses to the environment.Here again, the module temperature peak lies beyond the time where the light has been switched off, in this case by several minutes and the cooling is exponential.This experiment clearly demonstrates that irradiance fluctuations do lead to significantly delayed changes in externally measured module temperature.This delay is a result of heat conduction and heat dissipation within the module.While the irradiance change can be assumed to have immediate effect on the junction temperature, it takes time, before the heat is transferred from the thin semiconductor layer, through the back-electrode and the 3 mm glass sheet to reach the thermal sensor.Even then, the recorded temperature may not correspond to the true junction temperature anymore, because part of the heat has been dissipated in the large glass volume on either side of the thin film.This means the externally measured temperature response of a module upon changes of illumination, depends on module architecture and the environmental conditions and is therefore hard to predict.Exactly for this reason, a common workaround is the determination of the true junction temperature of a cell/ module from its momentary V OC and back-calculation via given label V OC and temperature coefficient b.However, this approach is only limitedly applicable for metastable modules, as their temperature coefficients and the V OC have been found to alter with the metastabilities, as well [33][34][35].Thus, there is no alternative to account for fast temperature fluctuations in thin film solar cells, which come with outdoor operation and monitoring.

Irradiance dependence and module monitoring during long-term field operation
The modules were checked for pure irradiance-dependent effects on the photocurrent-related module parameters.For this purpose, module photocurrents were measured under controlled laboratory conditions with the solar simulator in flash-mode (5000 ms flash, 400 ms measurement delay for light stabilization) with intensities between 100 W/m 2 and 1100 W/m 2 and compared to measurements taken on sunny days while mounted on the outdoor rack.In both cases, the characteristics were GT-corrected to STC conditions via the modified IEC 60891-Procedure 1 [36].
Then the module parameters were extracted and plotted against the original irradiance G. Figure 3 shows the  These results, taken from carefully selected "ideal" outdoor measurement days, show the limits of outdoor PV monitoring.While laboratory studies with high accuracy light source and sensoring equipment clearly showed irradiance dependent changes to device behavior, those trends evened out in outdoor characterization, by fast fluctuations in irradiance and delayed detection of temperature changes.The effect of such fluctuation is even worse when module data are continuously recorded 24/7 despite weather and daytime.
Most (micro-)inverters these days come with MPP tracking and the possibility of remote monitoring of the module output via datalogger.But the accuracy and reliability of these systems and thus the meaningfulness of these data are questionable.Often, accuracies for the generated/recorded MPP data are not even provided by the manufacturer, because the inverters are not meant as high end characterization equipment.We tested the capability of such microinverters for that purpose.For metastable modules, the GT-corrected STC data should demonstrate their changes upon long periods of darkness or sunshine.To evaluate the possibility to visualize lightinduced changes of long-term dark-stored chalcogenide modules merely from the microinverter data, pristine modules of each model were mounted to the outdoor rack and their parameters P MPP , I MPP and V MPP logged with 1-min resolution over several months.To compensate irradiance and temperature variations, the module performance data were GT-corrected to STC with the "Standard Irradiance and Desired Temperature (SIDT) À Procedure" [36] and data additionally filtered to the midday hours.The resulting development of the P MPP during two selected outdoor months for one module of each model is shown in Figure 4.It can be seen that even with corrected and midday filtering of the data, P MPP is scattering a lot.Those few days with small scattering represent such with stable sunny weather at noon.In the chosen examples, March was a particularly good month, while May only had few of such days in the beginning and few days in the second half of the month.The obtained values scatter around the module's label STC P MPP for the CIGS modules SF-165 (165 W) and WSG0074063 (63 W) and for the CdTe module FS-4117A (117.5 W).For the old generation CdTe modules, the values all fall below the expected power with an average of 59 W for FS-265 (label 65 W) and 52 W for FS-262 (label 62.5 W).This behavior is unexpected, since the modules reached the label power when photocurrent characteristics were measured with the curve tracer.Therefore, it is suggested that the microinverters, which were particularly chosen because they allow the high voltages of these two models (40 V to 90 V, max.500 W) have problems with correct determination of the relatively low current and power, even if the module operates at full power.

Temperature dependence of dark current characteristics
It is apparent that dark current measurements have a significant advantage for the reliable physical analysis of a PV module if outside of any highly controlled laboratory environment, they only depend on the module temperature, which in this case is dominated by the relatively stable environmental temperature.Resistance-induced heating of the thin film device during bias sweeps has been found insignificant in comparison [32].However, since the described nature of chalcogenide solar cells is known to lead to unexpected temperature related effects, we studied the dark current behavior of all our CdTe and CIGS test modules between 20 °C and 70 °C for potential trends, to enable their distinction from other effects in later analysis of dark degradation and preconditioning data.Figure 5 shows the semi-logarithmic dark current characteristics for one representative module of each type of CdTe and CIGS, for module temperatures between 25 °C and 70 °C.
Independent from type, model or module, the dark current characteristics show a shift of the curve's higher forward region towards lower voltages with increasing temperature.This is an expected behavior for any regular solar cell, due to the typical temperature dependence of the intrinsic carrier concentration in p-n-junctions, which affects the value of the saturation current.The extend of the voltage shift is proportional to the temperature difference and similar in all models.Besides that, there is an up-shift of the current proportional to the temperature in the low-bias region (magnified region shown exemplarily for FS-262 in Fig. 5), which was found for several modules across all models, but to very different extends.Since the low bias region is the shunt-resistance dominated domain of the dark IV curve, this observation supports the theory that the omnipresent leakage pathways in chalcogenide devices vary significantly in their density even between supposedly identical modules.This means that the R SH of an individual module will scale with the density of shunt pathways and additionally with module temperature by its effect on the number of charge carriers in the device and increased tunneling.Further detail analysis of the curves in the higher bias range do not show any changes in the linear slope, indicating the character of the diode (diode ideality n 1 , n 2 ) does not change within the investigated temperature range for any of the test modules.This matches with observations in literature, where changes of the diode ideality in CIGS cells were only found for certain much lower temperature values (below À40 °C) and assigned to (de)activation of certain metastable states [19].

Dark degradation after long-term field operation
After 6 months of field operation (March-August), modules were dismounted and transferred into dark storage, where their dark current characteristics were measured regularly to monitor potential effects of dark degradation on their device physics.Figure 6 shows an overview of these changes exemplarily for each model.
It can be seen that dark degradation affects each of the models in a different way, i.e., to different extends in particular regimes.Except for the old CIGS model WSG0074, all test modules show dark current characteristics well-fitted to the two-diode model, i.e., the symmetrical low bias shunt-dominated region (except for modules with bypass diode blocking the reverse bias part), followed by two linear regimes of the "weak" diode and the main diode with their different slopes in the intermediate bias region, and finally the series resistance dominated region at high bias visible from the change of curvature.The latter is slightly out of range for those models with higher currents due to the current limit of the source meter.It should be noted that while effects in the intermediate and high bias region occurred for all modules of a model set in similar fashion and extent, the changes in the low bias region varied even from module to module due to their earlier mentioned variation in shunt density.The new generation CdTe model FS-4117A (Fig. 6a) shows dark-storage induced decrease of the current in the shunt region, i.e., an increase in R SH , and further a shift of the curve in the main diode regime towards lower voltages, indicating an increase in its saturation current I 01 , but no change in slope.For the two old generation CdTe models FS-262 (Fig. 6b) and FS-265 (Fig. 6c), the most prominent common reaction to dark storage is the increase in series resistance R S , visible from the strong bending in the high bias regime.In the upper intermediate bias on the other hand, FS-262 shows a prominent change of the slope in the linear regime, which indicates a change in the ideality factor of the main diode n 1 , while FS-265 shows a voltage shift of the linear regime without change of slope instead, indicating a change in the saturation current I 01 , but not in the ideality factor n 1 .A more detailed look on the intermediate and high bias region (Fig. 6f) for FS-262 reveals that there is also a voltage shift of the curve overlapping the change of slope.Fitting of the FS-262 dark current characteristic to the twodiode equation shows an increase of the effective series resistance R S of the module from 1839 V for 1 h dark storage to 2828 V for 1632 h dark storage (note: "effective" = module values, i.e., not recalculated cell values with regard to N p and N s ).The fit also confirms the visual observations for the main diode regime, giving the extracted effective saturation current I 01 and the ideality factor n 1 , which were indeed found to decrease from 3.02 nA to 0.86 nA and to increase from 280 to 294 with longer dark storage, respectively.In the shunt region FS-265 modules show surprisingly no reaction to dark storage, while FS-262 reacts with an increase of R SH , like FS-4117A.Different to the CdTe modules, the new generation CIGS model SF-165 (Fig. 6d) reacts to dark storage by a decrease of the saturation current of the weak diode I 02 , but no change in slope, indicating merely a depopulation of recombination sites.Further, a subtle increase in I 01 (without change of slope) and an increase in shunt resistance R SH is observed.The old generation CIGS model WSG0074 (Fig. 6e), which seems to show only one diode regime, reacts to dark storage with a homogeneous shift of the curve in that region towards lower voltage, indicating also here a decrease in saturation current I 0 , again without change of slope, i.e., ideality factor n remains constant.Also here, this overlaps with the decreasing current in the shunt region due to an increased R SH with dark storage time.

Illumination-induced recovery effects on dark current
With the knowledge how the chalcogenide modules' characteristics change with dark storage, reinstating of the earlier characteristic À before dark storage À by means of preconditioning processes is the next step.For that purpose, dark stored modules have been exposed to 850 W/m 2 white light for 1 h.The module dark current characteristics was recorded before and then next after 1 h of cooling, followed by repeated measurements every hour, up to 24 h, to also monitor the stability of the light induced effects.Figure 7  As shown earlier, the model FS-262 modules (Fig. 7a) reacted to dark storage with a substantial increase of R S , an increase of n 1 , decrease of I 01 and increase of R SH .Looking now at those regimes after illumination, we found that all these observed dark degradation effects could be reversed, i.e., the shunt resistance and the series resistance decreasing, and also the main diode's ideality factor decreasing.However, the preconditioned state did persist for long, i.e., the effects started to regress within hours.After 24 h the values were halfway back towards those of the original pristine dark module.For the CIGS model SF-165 (Fig. 7b), dark storage had led to an increase of R SH , a decrease of I 02 and a subtle increase of I 01 .In this case, 1 h illumination resulted in the recovery of the shunt resistance and the weak diode's saturation current, by decreasing and increasing them, respectively.However, the saturation current of the main diode could not be restored, it did not decrease with light, but was found to increase instead.Also, for the CIGS model, the preconditioned state is short-lived, regressing on a similar timescale as the CdTe model.

Bias-induced recovery effects on dark current
To study how solely bias-induced preconditioning affects the chalcogenide modules' dark current characteristic, dark stored modules were exposed to repeated forward bias sweeps in the dark, in a range from 0 to 120 V for all models, but WSG0074 with 0 to 45 V due to its lower operation voltage.Duration of each sweep was 5 min.Figure 8 shows the bias-induced effects on the dark current characteristics, again, exemplarily for the CdTe model FS-262 and the CIGS model SF-165.The domains of each characteristic, which were noted earlier to be subject to dark degradation, are marked and the direction of the change indicated, where one occurred.
In general, it was found that the effect of dark bias sweeps on the dark current characteristics is much weaker than that of 1 h of illumination.Further, it apparently does not affect the modules in the same way as light does.For the FS-262 model (Fig. 8a) for instance, it can be seen that although dark degradation had led to an increase of n 1 , decrease of I 01 and an increase of both resistances, R S and R SH before, the forward bias treatment was merely able to recover the main diode behavior, but had little effect on R S and even less on R SH .A possible explanation would be that certain reactions in the device, which led to changing charge transport in the first place, need certain photoactivation to reverse them.The effect could be related to the electronic doping phenomenon of CdS (or their new replacement wide-band gap semiconductors), described earlier.For SF-165, the response of the module characteristic to bias treatments is even more selected.While earlier dark degradation caused the R SH to rise, I 01 to increase and I 02 to decrease, the voltage sweeps affected neither the shunt resistance, nor the weak diode.The curve merely recuperated in the region of the main diode, i.e., its saturation current was found to decrease by the treatment.

Conclusion and outlook
The present preconditioning standards are focused on the stabilization of a module's power at its maximum power point, initiated by its exposure to defined dosages of simulated or natural sun light.Metastable PV technologies, such as thin film chalcogenides, tend to challenge users and laboratories, when it comes to preconditioning and the reliable determination of the performance of a module, despite the standard procedures that have been developed specifically for each cell type.Fact is that thin film devices behave very differently to wafer-based technologies, because their device physics are dominated by volatile effects at surfaces, interfaces, defects and grain boundaries.They lead to sensitivity to light and temperature fluctuations, which causes a challenge, especially for data obtained under outdoor conditions.With CdTe and CIGS showcase modules, we recalled how temperature and irradiance fluctuations in (outdoor) PV module characterization or preconditioning, can cause a real problem in combination with metastable thin film modules.A great deal of outdoor, but also laboratory measurements rely on GT-corrections.However, metastability of the temperature coefficients themselves and lack of synchronicity of fluctuating G and T records with module measurements cause scattering and false data, as reflected in our comparison of irradiance dependent module performance under controlled laboratory and sunny outdoor conditions.Further, it is a fact that P MPP values simply do not deliver any information on the state, a respective module is in.
We suggested an alternative and showed its capability: using dark current characteristics and/or their two-diode model fitting parameters instead, basically as a "fingerprint" of the module and its condition, different to a sole power value.We demonstrated that there is a basic dark-IV fingerprint, whose shape is typical for any (intact) module of a certain model/type.Temperature changes shift it as a whole and not selected regions.We showed further metastability-related effects on the module can be identified at specific regimes in the dark current characteristic, which were subject to change upon a module's environment/treatment.For instance, while dark degradation generally led to an increase of both module resistances, rather model specific alterations were found for in those regimes of the curve, representing their diffusion or recombination diode character.Then we used these "degradation fingerprints" to show if the module condition could be reversed the same way by using different preconditioning treatments.Thereby we successfully demonstrated merely by using dark current characteristics that illumination was able to reverse all dark degradation attributes from the dark current characteristic.In contrast, forward bias treatment in the dark was not able to reverse increased resistance, but only the altered diode characteristics.The results are promising, proving the ability of dark current characteristics, or the parameter sets extracted thereof, to define metastable module's individual condition and use it to monitor its recuperation, warn about unwanted changes and estimate the dynamics of progressing or regressing towards or from a stabilized state.Naturally, this dark-current fingerprint method can be used on any module regardless of its state.But should be kept in mind that the module two-diode model for parameter extraction assumes that all cells behave similarly, i.e., whatever effects took place in the module are evenly distributed.If individual cells are deviating or failing entirely, extracted parameter sets cannot be used to recalculate true cell values and observed parameter changes might lead to false assumptions on the underlying cause (e.g., one failing cell on the module might increase the module's R S ).However, modules with such permanent damage are usually identified at arrival in the testing lab by preceding thermography/electroluminescence imaging and thus normally not even processed further.The darkcurrent fingerprint method could also help with the development of much debated possible climate dependent label values for PV modules [35].And it is not restricted to typical metastable modules.We found in preliminary experiments that analogous to reported temporary performance losses on silicon heterojunction modules [37], alterations in their dark IVs could be found with proportional dimension and equal time scale.

Fig. 1 .
Fig. 1.Equivalent circuit two-diode model for a photovoltaic module in the dark.

Table 1 .
Models and specifications of show-case modules selected for dark degradation and preconditioning experiments.Frontier SF-165 CIS (Cu-In-Se/S)/Zn(S,O) 2019 / Japan 165.0 1.93 85.5 Würth Solar WSG0074063 CIGS (Cu-In-Ga-Se/S)/CdS 2011 / Germany 63.0 2.0 31.5 do so until it peaks 20 s after the light has been switched off.

Fig. 2 .
Fig. 2. Temperature transient behavior of an FS-4117A CdTe glass-glass module upon short-term simulated solar irradiation at 850 W/m 2 for durations of 60 s (a,b), 10 min (c) and 60 min (d).Temperature was recorded in three positions on the backsheet (T1, T2, T3), further the room temperature in 1 m distance to the setup (RT).

Fig. 3 .
Fig. 3. G-dependence of illuminated module parameters, as extracted from STC-corrected indoor and outdoor photocurrent measurements for one CdTe module of FS-262 (a) and one module of FS-4117 (b).Upper row shows the for V OC , V MPP and P MPP data, the lower row shows the currents for I SC and I MPP .Levels of the label values are noted and additionally indicated by the dotted lines as guide to the eye.

Fig. 4 .
Fig. 4. Development of STC-corrected daily P MPP values for all CIGS and CdTe test models during two selected months of outdoor operation.The STC-label level for each of the models is indicated by the broken line as guide to the eye.

Fig. 5 .
Fig. 5. Temperature dependent dark current characteristics of CdTe (a-c) and CIGS (d-e) thin film test modules.The dashed lines indicate the different slopes of the D1 and D2 regimes.The magnified shunt-region for the FS-262 module is also shown (f).(Note: the CIGS modules incorporate a bypass diode, which inhibits continuation of the dark IV curve in the reverse bias direction).
shows the results of light-induced preconditioning on two examples, the CdTe model FS-262 and the CIGS model SF-165.Thereby the first diagram gives an overview of the full range dark current characteristic.The second and third one present the magnified regimes of the shunt region at low bias and the diode/series resistance dominated region at intermediate/ high bias, respectively.The illumination-induced changes to the characteristics are indicated in the respective sections.

Fig. 6 .
Fig. 6.Development of the dark current characteristics for all CdTe (a-c) and CIGS (d-e) test modules during 1 h to 1632 h of dark storage.The high-bias region for the FS-262 module (c) is shown in detail with fitted diode parameters (f).(Note: the variation in storage hours between different models occurred by experimental circumstances and has no further meaning).

Fig. 7 .
Fig. 7. Illumination effects on different parts of the dark current characteristics for exemplary modules of CdTe with model FS-262 (a) and of CIGS with model SF-165 (b).Shown is the full range characteristic and the magnified low and high bias region with the indicated light induced changes on the curve.

Fig. 8 .
Fig. 8. Effects on preconditioning with high voltage sweeps on the dark current characteristics of exemplary dark stored modules of CdTe model FS-262 (a) and CIGS model SF-165 (b).Shown is the full range characteristic and the magnified low and high bias region with the indicated bias-induced changes on the curve or missing of the same.