Progress in the understanding of light‐ and elevated temperature‐induced degradation in silicon solar cells: A review

At present, the commercially dominant and rapidly expanding PV‐device technology is based on the passivated emitter and rear cell (PERC) design developed at UNSW. However, this technology has been found to suffer from a carrier‐induced degradation commonly referred to as ‘light‐ and elevated temperature‐induced degradation’ (LeTID) and can result in up to 16% relative performance losses. LeTID was recently shown to occur in almost every type of silicon wafer, independent of the doping material. Even though the degradation mechanism is known to recover under normal operation conditions, it is a lengthy process that drastically affects the energy yield, stability and, ultimately, the levelized cost of electricity (LCOE) of installed systems. Despite the joint effort of many research groups, the root cause of the degradation is still unknown. Here, we provide an overview of the existing literature and describe key LeTID characteristics and how these have led to the development of various theories of the underlying mechanism. Further, given the continuously appearing and strong evidence of hydrogen involvement in LeTID, many mitigation methods concerning hydrogenation have been suggested. We discuss such reported methods, bearing in mind crucial consumer necessities in terms of sustained cell performance and minimised LCOE.


| INTRODUCTION
First reported in 2012, 1 light-and elevated temperature-induced degradation (LeTID) 2 was a new and unexpected degradation mechanism found to impact multicrystalline silicon (mc-Si) passivated emitter and rear cells (PERC) under typical solar cell operating conditions. With the industry set to transition production to mc-Si PERC at that time, this degradation raised significant concerns. Since then, LeTID has been a subject of extensive research, with the root cause of the degradation still yet to be fully understood. Prior to the development of mitigation treatments, LeTID was reported to reduce the performance of devices by 10% rel typically, but up to 16% rel , 3 and to occur over long timescales. Surprisingly, the degradation has been recently found to also occur in higher purity materials such as Czochralski-grown (Cz-Si) and float-zoned (FZ-Si) silicon. In a similar manner to the extensively studied boron-oxygen light-induced degradation (BO-LID) in Cz-Si, the degradation is followed by a recovery in performance at conditions reachable within modules in the field when exposed to sunlight and operated in open-circuit mode ($50 C-85 C). 4,5 Unlike BO-LID, however, the entire degradation and recovery process are expected to require decades under such conditions, causing a potentially huge loss in output power over a module's lifetime 6 and uncertainty of whether the module will even recover during this timeframe.
Various treatments now exist to reduce or eliminate the impact of LeTID on silicon solar cells. This paper reviews existing literature to discuss the current understanding of LeTID and collate what is known about the defect and its key characteristics and discuss them in the context of the various explanatory theories that have been proposed.

| AN INTRODUCTION TO LETID-FIRST OBSERVATIONS
The first observation of this unique degradation behaviour was reported by Ramspeck et al. 1 at Schott Solar. It was not until 3 years later, in 2015, that the degradation phenomenon was coined 'LeTID' by Kersten et al. 2,7 to reflect the conditions under which the degradation behaviour was observed and studied, that is, under illumination at elevated temperatures. However, LeTID can also be induced via applied bias, indicating that the degradation is in fact induced by excess carrier injection thus making it more accurately described as a form of carrier-induced degradation (CID). 7,8 Nevertheless, the term CID is commonly used to describe a range of other CID mechanisms in silicon including BO-LID, copper-induced light-induced degradation (Cu-LID) 9,10 and surface defects among others. In an attempt to distinguish LeTID from the other reported CID mechanisms, some groups later described the phenomenon as specific to mc-Si (e.g., mc-CID and mc-LID) [11][12][13][14][15] although the defect was subsequently found in various other forms of silicon. Another alternative identification that is perhaps more distinguishing, is 'hydrogen-induced degradation' (HID), as we proposed in Wenham et al. 16 This is due to strong indications and a consensus in recent literature that hydrogen is directly involved in the degradation process, either as a precursor for defect formation or as the LeTID defect itself. However, as the degradation under normal operating conditions will mostly occur via illumination at elevated temperatures, and there is confusion caused by the specification of the CID type, 'LeTID' has been adopted by the industry as the most standard and recognisable terminology. For the purpose of this review paper, we will be referring to the phenomenon as LeTID and the recombination active defect as the 'LeTID-related defect'.

| The impact of LeTID-Cells, modules and systems
Numerous reports of the impact of LeTID on a wide range of wafers, cells, modules and even some installed systems have been published. In Table 1, a list of studies on 'untreated' p-type mc-Si PERC cells and Fokuhl et al. 27 modules reported for LeTID throughout literature are compiled. The extent in apparent degradation varies significantly between different studies. Petter et al. 3 presented findings detailing degradation up to 16% rel on high-performance mc-Si PERC cells whereas Kersten et al. 6 presented field data of mc-Si PERC modules installed in both Germany and Cyprus. The modules in Cyprus were observed to have up to 7% rel power loss after 3 years of field operation and were likely to continue degrading for several years. Even in such a warm climate, the degradation alone could take a decade or more, with regeneration projected to occur on a much longer timescale. It could be inferred from the laboratory tests presented in that study that a duration of 15 years out in the field could be required before maximum degradation is reached within the specific climates, with regeneration expected to occur on a much longer time scale. As PV modules have a designated warranty of 25 to 30 years, this could pose significant issues for manufacturers, energy suppliers and consumers. Although not apparent in the table, the overall amount of reported LeTID has been reducing over time. Sen et al. 19 reported that cells made by one manufacturer from 2015 to 2019 has shown a >7.5% rel reduction in LeTID extent, with the majority of devices post-2018 achieving <5% rel degradation even without the employment of mitigation strategies (see Section 5), compared to 10% rel earlier. The authors note that the degree of degradation presented in Table 1 does not necessarily represent what can be optimised for in mass-production as samples are often chosen for their susceptibility to LeTID for the purpose of study.
One complication in assessing the degree of degradation and comparison between different studies, however, is the conditions used for testing. As seen in Table 1

| THE KEY BEHAVIOURS OF LETID
There are several defect characteristics of LeTID that differentiate the phenomenon from other forms of CID in silicon. In this section, these key characteristics are explored.

| A dependence on firing
One of the most well-known properties of LeTID is the dependence of the degradation extent on the peak temperature of the solar cell metallization or contact firing process. It was demonstrated in a number of works by Chan et al., 14 Nakayashiki et al., 26 Bredemeier et al. 34 and Eberle et al. 35 that LeTID is triggered by firing, with higher temperatures trending towards more severe degradation. This characteristic behaviour became a useful indicator for our later work identifying LeTID in other materials such as p-type Cz-Si wafers 36 and n-type silicon wafers. 37 Figure 1 shows the comparison between wafers fired at various temperatures between 400 C and 920 C. In several studies, LeTID was not found to occur during subsequent LeTID-stress testing if wafers were fired below approximately 650 C or not fired at all, but increased drastically with increasing temperatures above 650 C. From Figure 1A, depicting the temperature-degradation relationship for (up to 900 C) and subsequently light soaked at 75± 5 C ( Figure 2B).
The recombination properties of the defects in FZ-Si are compared with mc-Si and Cz-Si in Section 4. These studies on FZ-Si facilitates opportunities to study the defect using wafers with significantly lower concentrations of metallic impurities and crystallographic defects. 46 Although it is commonly held that n-type silicon wafers are less susceptible to various forms of CID, 50 more recent work has demonstrated the existence of LeTID in n-type materials as well. 22,37,45,51 Renevier et al. 52 investigated a form of bulk degradation in n-type Cz-Si wafers, formed on wafers coated with a SiO 2 /SiN x passivation scheme after rapid thermal annealing. However, this degradation was only observed when the wafers had at least one boron-diffused side.
Samples either passivated in Si-rich SiN x films or Al 2 O 3 or with dualsided phosphorus diffused layers did not show signs of degradation.
The authors concluded that the degradation may originate from the interaction between the SiO 2 layers and the boron-doped silicon; however, no exact root cause was determined. In 2018, we later confirmed similar findings using boron-or phosphorus diffused n-type Cz-Si wafers in addition to nondiffused controls. 37 It was found that degradation of the n-type bulk only occurred when a diffused layer There has also been further studies of LeTID in n-type mc-Si wafers, in particular the work of Sio et al. 22 and Vargas et al. 45 Using direct comparisons to p-type mc-Si, both authors concluded that the degradation rate under illumination of n-type mc-Si was slower than the rates observed in p-type mc-Si under similar conditions. Although there remains a possibility that the degradation observed on different wafer types is caused by different defects, there is strong evidence and many behavioural similarities suggesting that the recombination active defect in each of these materials is the same.

| A dependence on dielectrics
In addition to the strong relationship between peak firing temperature and LeTID extent, several reports have repeatedly shown the crucial role of the passivation layers on the degradation. Initially, the increased degradation observed on mc-Si PERC when compared to Al-BSF devices 1 led to the suggestion of aluminium (Al) gettering being the underlying prevention mechanism. 25,47 However, the duration of time above the melting point of Al during cofiring is typically quite short (<5 s), which would limit the gettering efficacy only to very fast-diffusing impurities or those close to the rear-surface, rendering it unlikely. 53 During the earlier stages, there were doubts as to whether LeTID was caused by a bulk-related defect or rather a degradation brought about by the AlO x films. 25 Although the dielectric layers play a significant role, numerous studies have demonstrated the

| Spatial dependence of degradation
Another distinguishing feature of LeTID is its spatial dependence.
Degradation at grain boundaries was shown to be weaker than that in the intragrain regions of a mc-Si cell. 11,60,61 It was hypothesised that structural defects present at the grain boundaries inhibit LeTID defect formation 11 or create an internal gettering effect. 60 Using μ-PL imaging, Jensen et al. observed both a slower degradation and regeneration rate within a grain boundary when compared to intra-grain regions ( Figure 3A). These 'denuded zones' of weaker degradation ( Figure 3B,C), were observed by Niewelt et al. 62 to be approximately 200-to -400-μm wide across a grain boundary. In the intragrain regions, degradation was shown to be homogeneous, 11,62 and no difference in LeTID defect density was observed between intragrain regions and dislocations. 60 Other studies have shown that spatial degradation properties of the observed LeTID can be extremely process dependent. A study by Lindroos et al. 63  states. The earliest models presented by Krauß et al. 17 and Sperber et al. 66 adapted the three-state model that resembled those historically used to describe BO-LID behaviour to demonstrate similar resemblances in defect evolution. 17,66 The most widely used version is presented in Figure 6. The defect evolves from a recombination inactive State A into a degraded State B and eventually into a recovered and stable State C (see Figure 6). The maximum extent of LeTID  The state diagram presented in Figure 6 is able to describe many, but not all, behaviours of LeTID. An expansion to the model was proposed by our group based on the impact of dark annealing on defect destabilization. 69 By subjecting wafers to a cyclic process of illuminated annealing followed by dark annealing treatments, Fung et al.
identified that recovery under light was not permanent. With each dark annealing step at a higher temperature of 232 C for 8 min, a subsequent illuminated anneal (conducted at 135 C, illumination intensity of 25.7 kW/m 2 ) would initiate another degradation and recovery cycle. However, after continued cycling, an exponential reduction in the maximum LeTID-related lifetime to a non-zero steady-state is observed ( Figure 7). It was thus concluded that the dark annealing process was producing additional defect precursors that remained in a state of inactivity in the dark but would later activate under

| Modulating the defect kinetics
The LeTID defect model tends to be more complex than similar models used to describe other forms of LID in silicon. There are multiple factors that influence the kinetics and the influence of these factors tend to convolute the results.

Excess carrier concentration
One of the earlier systematic studies on p-type mc-Si LeTID provided evidence of the prominent role of excess carrier concentration (Δn) on LeTID kinetics. 2  What is important to note, however, is that all these reports converge at demonstrating that the key factor readily impacting the LeTID kinetics is that of the excess carrier availability. This is independent of their generation origin, that is, photogenerated or as the result of applied current. Therefore, accurate LeTID kinetics studies become a challenging task due to the variations in structural 43

Dark annealing and thermal history
The same LeTID-defect activated during illumination has been shown to be activated whilst annealing in the dark at 175 C. 36 44 It was shown that the so-called 'stage' of the dark degradation and recovery cycle, which the sample was at when the pre-DA was stopped (i.e., before any degradation, during the degradation or during or after the regeneration), modulated the subsequent illuminated degradation kinetics to a similar extent, as previously shown in Chan et al. 73 The results demonstrated that there is a significant influence of even short dura-

| EVALUATION OF THE ROOT CAUSE
There have been numerous studies attempting to uncover the defect responsible for LeTID, however, identifying a specific defect in silicon out of thousands can sometimes be tricky. Nevertheless, the list of suspects can be greatly narrowed down through a range of characterisation techniques. One of these extensively used techniques is lifetime spectroscopy, namely, IDLS and the more advanced, temperature-and injection-dependent lifetime spectroscopy (TIDLS). 74 With these techniques, a fit of the SRH-related lifetime, isolated from the effects of intrinsic-and surface-related lifetime components, can then yield the capture-cross-section ratio (k-value) pertaining to a specific defect. This k-value is defined by a ratio between the electron and hole capture cross-sections (σ n and σ p , respectively) given by 75 where v th,h and v th,e are the thermal velocities of holes and electrons, respectively, τ n0 and τ p0 are the electron and hole capture time constants, respectively. In Table 2 into the ingot as depicted in Figure 10. During firing at high tempera-

| Hydrogen: A likely candidate
The overall dependence on the firing conditions (Section 3.1) and the need for a hydrogen containing dielectric to be present during firing The case of hydrogen being responsible for both degradation and F I G U R E 1 1 Evolution in normalised τ eff of mc-Si samples contaminated with various concentrations of metal impurities (Fe, Cr, Ni, Ti, Cu) as a function of laser-treatment (45 kW/m 2 , 140 C) duration. Legend contains corresponding extracted degradation and regeneration rates. This figure is adapted using data from Vargas 92 regeneration was also considered in Chan et al., 73 whereby one hydrogen atom could activate a defect, whereas a second atom could subsequently passivate the defect.

| A direct correlation with hydrogen
Most studies investigating the influence of hydrogen on LeTID have done so in two primary ways: by either (1)  Jensen et al. 98 concluded that the presence of a hydrogen source was essential for degradation, however, the time-temperature firing profile was not a necessity. This confirms that on samples with an alternative source of hydrogen, such as a hydrogenated dielectric layer, the firing process would only provide a means of introducing the hydrogen into the bulk, rather than directly activating LeTID. A similar form of LeTID may have also been activated using a shielded hydrogen plasma process on p-type FZ-Si wafers in a study by Bourret-Sicotte et al. 99 These results would further explain LeTID-related degradation occurring in pre-hydrogenated wafers where the dielectric layers were subsequently removed after firing. 100 Interestingly, a unique example of LeTID formation in the absence of firing is presented in the work of Sperber et al. 56 In this example, unfired and undiffused boron-doped p-type FZ-Si wafers exhibited LeTID-related degradation directly after the direct-PECVD deposition of dielectric layers, almost to the same extent as wafers fired at high temperature (850 C). It was postulated that the 400 C PECVD deposition step may possibly allow enough hydrogen into the silicon bulk to cause LeTID, even though commonly applied PECVD deposition processes for passivating layers use similar temperatures without LeTID activation in the absence of a firing step. One possible explanation could be the difference in deposition tools, that is, direct-PECVD compared to remote-PECVD, for which the former is known to be able to lead to more H-rich films. 101 The direct exposure of silicon to hydrogen containing plasma in a direct-PECVD reactor is also well known for introducing hydrogen into the silicon bulk, usually for defect passivation. 102 Figure 13B. There, it was assumed that a higher volume of hydrogen within the passivation layers would lead to higher bulk hydrogen concentration. 111 The effect of SiN x :H thickness on LeTID extent was confirmed by Bredemeier et al., 109 where, in addition, a saturation in the LeTID extent was found for film thicknesses above 105 nm. Nevertheless, authors of the latter work pointed out that caution needs to be taken with some of the assumptions made in their work and that of Vargas et al., because using bulk boron-hydrogen (B-H) pair densities to calculate the total influx of hydrogen during firing showed that films of higher hydrogen content did not necessarily give rise to a greater bulk in-diffusion of hydrogen, suggesting that other film properties of the SiN x :H layers may be the determinant of how much hydrogen is introduced to the silicon wafer. 107 Indeed, as shown by Bredemeier et al., 109 the SiN x :H film atomic density, rather than the hydrogen concentration alone, greatly influences the hydrogen incorporation and the subsequent LeTID degradation extent. It may be that because N-rich films are expected to have a dense structure in contrast to Si-rich dielectrics with lower Si-N bond densities that atomic hydrogen in the latter would move more freely, either through effusion or out-diffusion processes. However, at lower atomic densities, hydrogen tends to form dimers, which again restrict bulk in-diffusion, resulting in a drop in bulk hydrogen concentration.

| A concept of hydrogen-induced recombination
The prospect of hydrogen as a root cause has led to many theories regarding the recombination centre itself. A hypothesis was presented in Wenham et al. 16  Results that would appear to contradict the theory that hydrogen itself causes LeTID were presented by Jensen et al. 98 (refer back to Section 4.2.1). In that study, it was reported that the firing process itself does not cause LeTID and hydrogen is required for LeTID to occur. The finding that samples fired (in the absence of hydrogen containing films) before hydrogen introduction (via MIRHP) did not degrade, led to the conclusion that two reactants are requiredhydrogen and one or more defects that can be modified separately by firing. In addition, the solubility of interstitial hydrogen in silicon makes it very unlikely for the conditions required for complete dopant compensation-usually requiring hydrogen concentrations far exceeding the dopant concentration-to exist within the bulk at low temperature. It was concluded that the defects responsible for the degradation were likely formed near structural defects in the bulk of the silicon and have an intrinsic origin; however, the exact species could not be determined. From these studies, it would be useful to consider future DLTS analysis to be carried out on LeTID-susceptible FZ-Si wafers where the impact of crystallographic defects and metallic impurities may be largely excluded. These will be described here in the order in which they could be applied during the cell fabrication process. From a starting material perspective, LeTID extent reduces with wafer thickness, 83 and thus, using thinner wafers could be one approach to reducing LeTID in solar cells. Reduction of wafer thickness, however, potentially raises yield and handling concerns, particularly for screen-printed solar cells. 118 It has also been shown that wafer selection can play an important role in the LeTID extent of solar cells. 26 Wafers from different ingots, or even the same ingot 3,7 can show vastly different LeTID characteristics, and thus selecting wafers with lower LeTID potential is one method for reducing LeTID in finished solar cells. Nonetheless, in reference to p-type mc-Si wafers, it was noted in a study by Hanwha Q Cells that there were 'no LeTID free wafers found on the market'. 3 It is also desirable for a solar cell manufacturer to be able to use the entire ingot for solar cell production and to not have to screen for LeTID at a wafer level.

| MITIGATION OF LETID
Based on the observations that a phosphorus gettering process can reduce LeTID, 93,94 it is likely that optimisation of the emitter diffusion process during cell production can help to suppress LeTID. As also discussed in Section 4.2.2, dielectric layer composition, 54,82,119 thickness 83,111 and stoichiometry 109 has been shown to have a significant impact on LeTID, likely due to the modulation of hydrogen diffusion into the silicon bulk associated with the changes in the films.
Bredemeier et al. 109 identified that a SiN x :H refractive index (RI) of below 1.9 or above 3 resulted in reduced LeTID. Tuning of SiN x :H to avoid LeTID will have impacts on the solar cell electrical and optical performance that need to be considered. Changes in film hardness can alter the penetration depth of the Ag fingers during metallization, leading to shunting or poor ohmic contact, thus cofiring would need to be reoptimized with this approach. Changes in dielectric thickness will reduce the anti-reflective behaviours for certain wavelengths of light. Furthermore, although silicon-rich SiN X :H films (high RI) can provide better surface passivation they are usually absorptive in the short and medium wavelength range leading to parasitic losses. 120  demonstrated that the LeTID defect could be suppressed even at higher peak firing temperatures if the cooling rate were slowed. In that study, using a rapid thermal processor (RTP) with a slower cooling rate, LeTID was effectively suppressed even at a high peak firing temperature of 800 C. This study was carried out on lifetime samples, and when applied to actual solar cells, would need to be carefully approach similar to that proposed for rapid mitigation of the BO defect in the past. [123][124][125] As LeTID is a CID mechanism, an equivalent process can also be performed in the dark while injecting current into the cell, however the carrier densities achievable during this process are significantly lower than what can be achieved by high intensity illumination, thus a current injection LeTID mitigation process is generally slower. Annealing a finished cell in a belt firing furnace or 're-firing' is another approach that has been shown to effectively suppress LeTID, although this method can be detrimental to device series resistances (R s ) if not carefully optimised. 14,20,126 Peral et al. attributed this observed increase in R s with extended contact firing to a thickening of the glass layer surrounding the silver crystallites within the metal contact, thus inducing resistance against current flow. 126 Annealing in the dark has also been shown to suppress LeTID. 41,73,88 Similar to the approach in the belt-furnace, dark annealing has been shown to result in a drop in device FF for longer durations and higher temperatures. 122 To the avoid the drop in device FF, Sen et al. 127 proposed a possible alternative method whereby the anneal is performed prior to metal contact firing. This approach has shown to be effective on lifetime test structures but has yet to be demonstrated on cells, and the effect of changes in density of the SiN x :H layer after annealing on the cofiring process may need to be considered. In an alternative approach to avoid the FF increase, Hamer et al. 122 proposed annealing the cell in the dark with a reverse bias voltage applied. In this approach, it is hypothesised that the electric field repels hydrogen from the metal/Si contact interface that would otherwise be responsible for the drop in FF. A summary of the various mitigation strategies presented throughout the literature is presented in Table 3.
Several commercial tools to treat LeTID have emerged into the market. These include illuminated annealing tools: laser based ( Figure 14A), 133 LED-based, modified metallization furnace-based tools ( Figure 14B) 129 and dark annealing tools involving the injection of current ( Figure 14C). 121 The latter has seen wide adoption by the industry due to the increased throughput of the batch 'coin stack' technique that is compatible with this process in addition to a typically smaller production line footprint.
T A B L E 3 Summary of various suppression or mitigation strategies at a wafer or cell level as highlighted throughout the literature Reduction of the cooling rate during firing using an RTP furnace Chan et al. 14 Secondary post metallization firing process (480 C-660 C) + laser processing (200 C, >40 suns) Sen et al. 78 Reduction of the cooling rate during firing using an RTP furnace Bredemeier et al. 129 Post-firing anneals in a commercial in-line furnacebased tool (c.REG) Sen et al. 127 Low temperature annealing ($650 C) prior to metallization firing to suppress the issues associated with increased FF Sen et al. 20 Application of a two-step moderate temperature annealing process post-metallization firing in a belt furnace.

Sharma et al. 41
Optimisation of firing conditions with reductions in peak firing temperature and modification of belt speed. Use of post metallization dark annealing at moderate temperatures (300 C-550 C) Sharma et al. 130 Incorporation of postfiring forming gas annealing (FGA) Yli-koski et al. 131 Application of a low-temperature (200 C-300 C) long duration (>18.5 h) anneal to suppress subsequent LeTID formation Varshney et al. 111 Reductions in the thickness of SiN x films to reduce in-diffused hydrogen content Varshney et al. 132 Implementation of ALD deposited Al 2 O 3 films as a blocking layer for hydrogen in-diffusion during firing Bredemeier et al. 109 Tuning of SiN x films to high (towards 3) or low (towards 1.9) refractive indexes to reduce hydrogen mobility or density, respectively Payne et al. 8,15 High intensity laser (44.8 kW/m 2 ) and high temperature process (140 C) to accelerate defect formation and recovery rates Wang et al. 121 Use of single or double current injection annealing (CIA) processes (260 C, 14.5 A) on finished solar cells It may seem that from the many LeTID mitigation methods discussed above that there can be a tendency, in one way or the other, to opt for the reduction in total hydrogen concentrations within the bulk of the solar cells. 134 However, it is important to highlight the detrimental effects that this approach may represent. Hydrogen is recognised as being essential for the passivation of bulk defects, particularly in mc-Si wafers for grain boundaries and other structural defects, and in Cz-Si wafers for the passivation of BO defects as discussed in previous works. 135,136 It is therefore crucial to strike a balance between removing excess hydrogen that may lead to LeTID, while maintaining sufficient hydrogen concentrations for the passivation of other defects including the one responsible for BO-LID.

| CONCLUSION
Although there has been sustained research into LeTID for nearly a decade, the specific recombination-active defect remains unknown.
Nevertheless, there has been significant progress in understanding the properties and behaviour of the defect and perhaps most impor-