A case study on accelerated light‐ and elevated temperature‐induced degradation testing of commercial multi‐crystalline silicon passivated emitter and rear cell modules

Light‐ and elevated temperature‐induced degradation (LeTID) can have significant and long‐lasting effects on silicon photovoltaic modules. Its behaviour is complex, showing highly variable degradation under different conditions or due to minor changes in device fabrication. Here, we show the large difference in LeTID kinetics and extents in multi‐crystalline passivated emitter and rear cell (multi‐PERC) modules from four different manufacturers. Varied accelerated testing conditions are found to impact the maximum extent of degradation in different ways for different manufacturers complicating the ability to develop a universal predictive model for field degradation. Relative changes in the open‐circuit voltage (VOC) have previously been used to assess extents of LeTID; however, due to the greater impact of the defect at lower injection, the VOC is shown to degrade less than half as much as the voltage at maximum power point (VMPP). The MPP current (IMPP) and fill factor (FF) also degrade significantly, having an even larger overall impact on the power output. These observations imply that currently employed methodologies for testing LeTID are inadequate, which limits the reliability of future predictive models. In light of this, the field must develop a more holistic approach to analysing LeTID‐impacted modules, which incorporates information about changes under MPP conditions. This will allow for a much clearer understanding of LeTID in the field, which will assist the performance of future PV systems.

gallium doped 6,7 and float-zone. [8][9][10][11][12][13] The exact defect causing the degradation is not yet known; however, there is a growing consensus that hydrogen is involved, due to the strong correlation with the hydrogen concentration. 14 The degradation can lead to >10% relative efficiency loss if untreated and take decades to degrade and recover in the field. 15 LeTID has a complex behaviour showing varied degradation and recovery behaviours based on wafer type, testing conditions, device structure 16 and even minor differences in prior thermal history. 17,18 In addition, module operating temperatures vary widely depending on location and type of installation. A roof-mounted module in Sydney can reach almost 100 C and a BIPV module in Tucson can spend over 1600 h above 50 C every year, while a field rack-mounted array in Hamburg does not even reach 60 C and spends only 17 h over 50 C in a year. 19 As a result, despite several years of investigation, the impact on commercial modules is not well understood.
A study by Kersten et al has shown much slower degradation of LeTID sensitive modules installed in a cool German climate when compared with a warm, sunny Greek climate. 20 Fokuhl et al 16 have shown that modules consisting of different cell structures show different degradation and regeneration behaviour and different responses to accelerated testing conditions. Relative to multi-PERC modules, in monocrystalline silicon PERC modules, the recovery process appeared to accelerate far more significantly than the degradation process, leading to a reduced extent of degradation. This was even more the case for silicon heterojunction (SHJ) modules, which only showed improvement under the further accelerated conditions. Deceglie et al have shown that even modules of the same type from the same manufacturer can degrade differently in the field, possibly attributed to different rear contacts. 21 Dupuis et al used comprehensive experimental data of bifacial module degradation combined with predictive modelling to highlight the challenges accounting for the large variability between cells in a module as well as other uncertainties including activation energies. 22 All of this complexity and uncertainty can be disastrous for warranty claims, investors and consumers relying on accurate projections and costing of future output, designers sizing systems and so forth. If left untreated, the highly variable degradation could cause havoc for IEC qualification and reliability testing. 23 Figure 1A. To enable a wide variation of accelerated testing conditions to be tested on a single module, the temperature was allowed to naturally rise due to the illumination with no diffuser, thereby causing varied illumination and temperatures across the module. The array was constructed separately for each module resulting in slight variations in angles and positioning for each, thereby causing differences in illumination intensity and temperature across each module. To enable meaningful results, the temperature and irradiance were measured at the centre of each cell of each module; temperature was measured using both an infrared gun on the front and a type K thermocouple (on the front and rear, showing typically 1 C-2 C higher on the illuminated front side) to gain confidence in the measurements; the irradiance was measured with a Thorlabs 40-W thermal detector. To demonstrate the non-uniformities, Figure 1B,C shows the illumination intensity and temperature maps, respectively, for the 120 half-cell module from manufacturer D. Temperatures ranged from $50 C to $115 C.
Illumination varied from $0.5 to 2 kW/m 2 in a similar distribution.
Since the temperature rise was due only to illumination, the two were linked (see Figure 1D), as would occur similarly in field conditions. A solid line shows a linear trend of all data points from all manufacturers. Some spread can be expected due to thermal conductivity between adjacent cells and increased thermal radiation from module edges.
LeTID kinetics is dependent on both temperature and the excess carrier density. 15,24 However, due to the logarithmic dependence of the excess carrier density on the illumination intensity, increasing the illumination intensity around this range has been shown to have minimal effect on the degradation kinetics, 25 as also confirmed in our data (not shown). In light of this and for simplicity, results in this paper will be reported according to the temperature variations. V MPP for each cell in each module was then calculated according to the following equation 28 : A sister module (same batch) from manufacturer A was also sent to DNV-GL (now PV Evolution Labs, known as PVEL) and tested similarly using a highly uniform Class A + A + A + Pasan SunSim 3b solar simulator. Periodic IV measurements and EL images were taken over 250 h; the module temperature naturally rose to $100 C ± 4 C.
After the 10,000 min of LeTID testing, the module from manufacturer C was dark annealed in a laminator for 10 h at 150 C. The purpose of this was to destabilise the regenerated defect in a manner similar to that shown by Fung et al who used 8 min at 232 C. 29 Since the laminated module could only be taken to 150 C without risking permanent damage, the anneal was performed for 10 h to account for the exponentially slower reactions at the lower temperature. The module was then LeTID tested again for 10,000 min.
A depiction of the various experimental conditions and process flow is shown in Figure 2.

| LeTID with varied temperature
Example PL_LS images taken using the custom-built-line-scanning module PL tool are shown in Figure 3. These images were taken of the manufacturer A module as received and after 2000 min of LeTID testing. The differences in degradation due to the nonuniform illumination and temperature can be seen, as well as general variations due to differences between cells. This is to be expected for the reasons mentioned prior, and the fact that even cells manufactured on the same line do not degrade the same under the same conditions, as seen by the checkerboard variation in EL images of degraded modules such as that shown in, 15 and later in Figure 8B. conditions. 15 Roughly 300 h at 75 C in MPP is equivalent to $1 year in Greece. 20 The $75 C curve in each plot would therefore correspond to $5.5 years installed in Greece. These long timescales are concerning, as even in a warm sunny climate, with the exception of manufacturer C, the full extent of degradation is not realised within this timeframe, let alone the slower recovery process. The impact of this degradation is therefore likely to impact power output for most or all of a module's working life, even with the increasingly longer warranties. 30

| Temperature dependence of LeTID extent
Increasing the accelerated testing temperature is reported to cause a decrease in the maximum extent of LeTID observed. 15,16,25 With independent and competing degradation and recovery processes, increased acceleration is generally seen to increase the recovery process more significantly, thereby causing it to dominate earlier and reduce the maximum extent of degradation. However, we also know that even minor differences in prior thermal history, doping and wafer position can drastically change the kinetics of the degradation.  Table 1, averaged in 10 C increments for each manufacturer.

| LeTID temperature dependence following destabilisation
The 10-h dark anneal at 150 C was successful in destabilising the defect allowing a second LeTID cycle to be observed in the module from manufacturer C. The maximum degradation extent plotted with respect to temperature for the second LeTID testing cycle is shown in  It is not uncommon for V OC to be used as a metric for assessing LeTID due to the relationship between V OC and device recombination. 18,25,33,34 It is therefore interesting to note that the V MPP degrades 2.54%; more than twice as much as the relative drop in V OC .
This can be explained by the injection level dependence of the LeTID defect, having a more pronounced effect and increased recombination at lower injection levels/voltages. 4 In a similar way, the I SC is also observed to degrade more relatively than the V OC . Measuring at V OC is therefore not able to capture the full impact of the degradation. Figure 9 shows the ΔV MPP (A) with time and (B) maximum extent versus temperature of the manufacturer A sister module that was degraded at UNSW and ΔV MPP was calculated from EL_LS images taken with low current injection (= 0.95 A). 27,35 In hindsight, the injection current could have been modified to more accurately represent MPP conditions as a result of the changes due to degradation. Fortunately, due to the log dependency, the constant current used to create MPP conditions is a reasonable approximation. 36 In fact, the slightly larger current places the degraded cells at a slightly higher injection level than MPP and therefore a higher V MPP is measured. This therefore slightly underestimates the ΔV MPP meaning that the true impact would be even more pronounced than that shown. Comparing with Figures 4A and 5A, it can be seen that here V MPP also degrades roughly twice as much as V OC . These data are included in the maximum degradation data of Table 1. It is also interesting that even though the V OC shows a strong trend of reducing degradation extent at higher testing temperatures, the V MPP does not show quite as strong trend. As such, although still limited to only one aspect of module performance, the V MPP is a better representation of LeTID on cell or module performance than the V OC .
T A B L E 1 Maximum change in V OC averaged for all cells that had reached maximum degradation within 10,000 min, averaged in 10 C increments example, hydrogen bound to dopants, traps or in molecules. 29 This latter point may help explain the different behaviour seen after a module is dark annealed, as the dark anneal is known to alter concentrations of such states. 29 In any case, each of these possibilities mean that not only is the maximum extent of degradation under accelerated conditions not predictable from one manufacturer to another but also the longer term impact of degradation during the recovery phase is also unpredictable without more extensive testing data. We saw earlier (Figure 4)   showed that during LeTID testing, a larger impact was seen in the V MPP , rather than the commonly used metric of V OC , which was confirmed on the non-uniformly degraded sister module using EL_LS at $MPP conditions. Changes in I MPP and FF were also observed to contribute to reduced power output during LeTID conditions. These results combined indicate that current existing methodologies used to assess LeTID are inadequate. The vast variability in the degradation behaviour between manufacturers and the seemingly narrow view provided by only assessing V OC conditions or only testing the degradation phase indicates that future testing regimes should be expanded to accurately describe LeTID. A better understanding of this degradation in the field will have a profound impact on costing, system design and warranties.