Review of technology specific degradation in crystalline silicon, cadmium telluride, copper indium gallium selenide, dye sensitised, organic and perovskite solar cells in photovoltaic modules: Understanding how reliability improvements in mature technologies can enhance emerging technologies

A comprehensive understanding of failure modes of solar photovoltaic (PV) modules is key to extending their operational lifetime in the field. In this review, first, specific failure modes associated with mature PV technologies, such as crystalline silicon (c‐Si), copper indium gallium selenide (CIGS) and cadmium telluride (CdTe), are framed by sources of specific failure modes, their development from the early‐developmental stages onwards and their impact upon long term performance of PV modules. These failure modes are sorted by both PV technology and location of occurrence in PV modules, such as substrate, encapsulant, front and rear electrode, absorber and interlayers. The second part of the review is focused on emerging PV technologies, such as perovskites solar cells, dye sensitised and organic PVs, where due to their low to medium technology readiness levels, specific long‐term degradation mechanisms have not fully emerged, and most mechanisms are only partially understood. However, an in‐depth summary of the known stability challenges associated with each emerging PV technology is presented. Finally, in this paper, lessons learned from mature PV technologies are reviewed, and considerations are given in to how these might be applied to the further development of emerging technologies. Namely, any emerging PV technology must eventually pass industry‐standard qualification tests, while warranties for the lifetime of modern c‐Si‐based modules might be extended beyond the existing warranted life of 25 years.


Funding information
European Cooperation in Science and Technology failure modes associated with mature PV technologies, such as crystalline silicon (c-Si), copper indium gallium selenide (CIGS) and cadmium telluride (CdTe), are framed by sources of specific failure modes, their development from the early-developmental stages onwards and their impact upon long term performance of PV modules. These failure modes are sorted by both PV technology and location of occurrence in PV modules, such as substrate, encapsulant, front and rear electrode, absorber and interlayers. The second part of the review is focused on emerging PV technologies, such as perovskites solar cells, dye sensitised and organic PVs, where due to their low to medium technology readiness levels, specific long-term degradation mechanisms have not fully emerged, and most mechanisms are only partially understood. However, an in-depth summary of the known stability challenges associated with each emerging PV technology is presented. Finally, in this paper, lessons learned from mature PV technologies are reviewed, and considerations are given in to how these might be applied to the further development of emerging technologies. Namely, any emerging PV technology must eventually pass industry-standard qualification tests, Failure modes are different between differing technologies due to material and process challenges, as shown in Table 1. As a result, this review has been partioned into three sections. The first part focuses on the technology specific degradation modes of mature PV technologies and introduces degradation modes found in silicon PV

| c-Si
Silicon, the second most abundant element on the earth's surface, is the most developed semiconductor material for PV applications and dominates the market. 6 Being one of the oldest PV technologies, its degradation mechanisms have been studied extensively. [7][8][9][10] In addition to common environmental and voltage stresses, the c-Si systems can also suffer from mechanical loads because silicon wafers are T A B L E 1 Summary of the specific degradation and failure mechanisms of the PV technologies discussed in this article (schematics of technologies are given in Figure 1

| Cracked cells
The reduction of silicon wafer thickness aims to decrease the cost of silicon-based PV cells and modules. Nevertheless, the smaller thicknesses decrease the robustness of solar cells against mechanical loads and may cause cell cracking. 13 Cells that crack during the production process can be detected and eliminated. However, it is not possible to entirely avoid the formation of microcracks on the PV cells; therefore, it is crucial to quantify their long-term effects on the performance of PV modules, which is beginning to get more research interest. Microcracks and imperfections increase the risk of breakage during the production cycle and can propagate further during the lifetime of PV modules although they initially show little or no power loss. 14,15 There is some debate in the community on cell cracks as not all cracks result in a loss in performance; however, these can become larger with time and eventually lead to disconnections in parts of the cells. [16][17][18] Microcracks may form in several stages, namely, during (1) ingot cutting, (2) production of cell and module, (3) transportation and installation and (4) operation of PV module due to environmental factors such as temperature cycles, wind, snow and hail. 15,19,20 Cracks interrupt the electrical conductivity in cell regions, which leads to reduction in the short-circuit current and the increase of the series resistance, resulting in output power reduction of PV modules 21,22 and can also increase PID. 23 The position, length and orientation of microcracks influence this power reduction. 22 Cracks and microcracks are distinguished based on their size: A crack with a width up to 30 μm is classified as a microcrack. 24 Cracks F I G U R E 1 Schematic diagrams of technologies discussed in Table 1. Note: Schematic diagrams of DSCs and MHPs are given in Sections 3.2 and 3.4, respectively. The OPV diagram shows a side perspective of a 2-cell monolithically connected module.
occur in different shapes and sizes. 25 Cracks in c-Si PV cells and modules are further classified using other criteria such as severity and position 21,26 (see Table 2). Star-shaped cracks consist of several line cracks originating from an induced point. Line-shaped cracks are also initiated due to laser-cutting of wafers. 27 Figure 2 shows the classification of cracks according to their orientations in silicon PV cells.  28 In other studies, 29,30 experimental results have proven that cracking alone reduced the fill factor (FF) and output power up to 4% and 3% respectively.

| Snail trails
Crystalline-Si PV modules in the field may develop local line-shape discolorations, so-called snail trails, over the cells after a period of months to a few years; see Figure 3. Closer inspection shows that the discoloration occurs on the silver paste only. Snail trails form in the presence of cell cracks and depend on the packaging polymers (e.g., ethylene-vinyl acetate [EVA] acetic acid formation). 31,32 Moisture ingress seems to be the cause of silver line corrosion. Chemical reactions at the silver paste-encapsulant interface may lead to the formation of silver-containing nanoparticles above the silver lines. 33 The line conductivity is only reduced to a limited extent. Optical transmission loss is also negligible as the discoloration happens above the silver lines; the encapsulant away from the silver line remains unaffected. 33 To observe snail trails in an accelerated test, a combination of mechanical load, UV exposure and temperature elevation is recommended. 34,35 While the snail trail itself does not have a direct or significant effect on cell or module performance, it is an indication of moisture ingress, commonly caused by mechanical stress induced loss of module hermeticity and cell cracking. These are performance risks. Consequently, some reports indicate that the impact of snail trails on PV performance is negligible, 36 but other studies conclude that output energy produced by PV modules can be reduced up to 20% due to snail trails. 37,38 It appears that in the latter case, positive correlations between the occurrence of snail trails and power loss were misinterpreted as causal, whereas instead they have a common origin (moisture ingress, leading to more problems than snail trails alone).

| Hot spots
A hot spot is a high temperature area on the PV module that may cause serious damage on the solar cells and other elements on the modules (see Figure 4). discharging may occur under these circumstances. PV modules are daisy-chained to improve power efficiency and lower the system cost.
In that case, the potential difference between cells and frame may reach $1000 V, depending on the module, combiner and inverter ratings. This may soon even go as high as 1500 V, as some manufacturers are pushing to reduce the balance of system (BoS) cost.
Degradation phenomena in PV modules related to this high voltage are termed PID and typically lead to losses up to 5% but sometimes significantly more. PID degradation is primarily caused by the potential bias but is also affected by humidity and temperature. This is related to the leakage current between the frame and the cell. Recent work, which subjected c-Si PVs to PID testing under IEC 62804-1 standards, estimated that annual degradation was 11.2% per annum in a high humidity environment such as Miami (as opposed to 6.9% in a lower humidity environment). 55 Discolouration, delamination, microcracks, shunts and even stack- LID can occur even at low light exposure at room temperature, and the formation of a boron-oxygen and iron-boron defects in the silicon wafer are the main degradation mechanisms. 61 To mitigate the effects of LID, it has been proposed to decrease the oxygen content or substitute boron by other dopants such as gallium. 62 LETID is a specific degradation type first observed on PERC-type multi-crystalline Si PVs in the field. 63 Later works showed that mono-Si cells also suffer from degradation under the combination of light and temperature stress. In comparison to LID, which occurs in a short period time of initial exposure to sunlight, LETID develops more slowly. The consequence is power loss up to several percent, though there can be a recovery over several years in warmer climates. 64 As in the case of LID, boron-oxygen complexes can be observed after LETID but are not the main root cause for degradation. 63 Hydrogen redistribution phenomena are currently considered responsible. 64 Preventive measures that can be taken for the mitigation of LETID include use of silicon wafers with low oxygen content, dielectrics with little hydrogen and low firing temperatures.

| CdTe
CdTe technology dominates the thin film PV market based on the relatively low cost for manufacturing, increases in module efficiency (18%), small temperature coefficient (0.25%/ o C) and large scale of manufacture. 49 The latter has been largely due to one manufacturer, First Solar, that now produces in excess of 5 GW per year. The history of CdTe module deployment is much shorter than that for c-Si so less is known about the long-term performance.
Wendlandt et al. 50 reported the range of measured degradation rates for CdTe modules to be 0.2-4%/year per year with a median value of 0.5%/year. The range can partly be attributed to variations in the manufacturing method and module sealing but is largely due to the non-linear development of CdTe module degradation with time.
The IEA-PVPS Report (2014) 34 mentions the main CdTe-specific failure mechanisms as being: • Front glass breakage that can cause short term failure; • Back contact degradation that causes longer-term loss of performance.
These are detailed in the following subsections. In addition, we will treat PID in CdTe modules.

| Front glass breakage
A consideration for CdTe PV technology is the constraint imposed on the manufacturing process by the superstrate configuration where the front glass is used as the substrate for depositing the thin films in the PV device. This prohibits hardening or tempering of the front glass (the superstrate) because it has to endure a series of temperature cycles during the deposition of the thin film coatings and heat treatment. The lack of hardening or tempering makes CdTe modules more susceptible to failure due to front impact (see Figure 5). 66,67

| Back contact degradation
The role of copper in CdTe modules has been an important factor for reducing back contact series resistance and doping the CdTe absorber layer. 68 However, too much copper applied to the back surface will result in diffusion to the front junction and cause an increase in carrier recombination and loss in V oc . Controlling the amount of copper applied to the back surface of CdTe is therefore crucial to obtain high efficiency devices. Artegiani et al. 69 present evidence that just 0.1 nm of Cu suffices. The drop in PV module performance over the first 2-3 years of deployment is attributed to copper diffusion to the front junction. As Perrenoud et al. 68 have pointed out, the solubility of copper in CdTe is low; Cu concentrates at the crystal grain boundaries, which provide a fast diffusion pathway to the front junction.
The temperature coefficient of CdTe is À0.25% per C temperature rise, half that of c-Si. That makes CdTe an attractive choice in warm climates; however, the high operating temperatures could enhance the diffusion of copper. Strevel et al. 70 have observed a 4-7% power loss over the first 1-2 years before a linear degradation factor of À0.7%/year is established. The initial drop goes faster at higher operating temperatures. Strevel et al. state that the initial power output of the module is underspecified to deal with this initial degradation.
In a series of controlled laboratory heat cycling tests on experimental CdTe solar cells, Bertoncello et al. 71 have attributed degradation to two different mechanisms. The first is copper diffusion, and the second is oxidation. One of the observations was an increase in series resistance, which was attributed to loss of copper from the back contact accompanied by a conversion of low resistance Cu 2 Te to high resistance CuTe. The diffusion of the excess copper through the CdTe to the CdS buffer layer causes loss of short wavelength external quantum efficiency. The oxygen ingress during accelerated heat testing caused the formation of TeO at the back contact, increasing resistance. However, in a paper on As doped CdTe, an air anneal resulted in enhancement of the V oc. 72 The oxidation degradation might occur specifically to Cu doped back contacts.
A radical solution to copper-related degradation is to replace Cu by another element, preferably one with a higher solubility, to obtain higher acceptor concentrations. With As doping, this concentration has been shown to exceed 1 Â 10 16 cm À3 . 73

| PID
Like c-Si modules, PID in CdTe modules is also strongest on the negative string end. 75  As with c-Si solar modules, the mechanisms governing PID in CdTe modules are not yet fully understood. Leakage currents develop due to the large potential difference between the grounded frame and the cells, particularly those at negative potential. The glass F I G U R E 5 A hot spot due to cracked front glass (not readily spotted by visual inspection) detected using aerial infrared thermography 66,67 superstrate for CdTe modules is typically a soda-lime glass, and sodium can migrate towards the cell and reach the junction region degrading cell performance by introduction of recombination centres.
This typically degrades the V oc and FF of the cell. The mechanism of 'Na transport' and the subsequent cell degradation depend on the moisture conditions. In a dry climate, the sodium migration described above will operate and can be reversed by subjecting the modules to a reverse bias. Moisture ingress however will result in irreversible degradation of the module as the reduced sodium will react with the moisture to produce atomic hydrogen, which will then react with the SnO 2 -based TCO. This is the mechanism that leads to visible TCO degradation near the edges of a module (see Figure 6).

| Copper indium gallium diselenide (CIGS)
CIGS is now one of the most mature thin-film PV technologies with rapid growth of installations and production capacity, which is due to its low fabrication costs, short energy payback time and most importantly due to their freedom of size, shape and flexibility, making CIGS suitable for integration in various infrastructures. 77 Jordan et al.
reported that CIGS modules installed in the 21st century demonstrated low median power degradation rates of 0.5% per year. 78 The majority of the modules showed rates between 0% and 1% per year, while some modules actually improved during outdoor operation. A small quantity of outlier modules showed worse field behaviour.
Degradation in CIGS PV systems can be induced by various stress loads including humidity, partial shading and biases. CIGS solar devices are formed as a multi-layered material stack and responses to such stress loads differ for each layer. Figure 7 demonstrates the cross-sectional schematic of the material stack in a typical CIGS monolithically interconnected device and illustrates the degradation mechanisms. P1, P2 and P3 denote the presence of so-called scribes (patterning lines), which are necessary for module formation. 79,80 This following subsections discuss these particular failure mechanisms in detail under following subsections: and external biases (PID) • Wormlike defect formation due to partial shading.

| Reduction of contact conductivity due to water ingress
In order to study the impact of humidity, many studies have looked at the performance loss of CIGS devices without any packaging under damp heat conditions (85 C/85% RH). In this way, the intrinsic stability of the solar cells and minimodules could be studied. A literature review 80 revealed strongly varying degradation rates under these damp heat conditions. The most impacted device parameters were the FF and the open circuit voltage. These devices were on the other hand mostly stable when exposed to dry heat conditions, which was also the case for packaged devices exposed to damp heat. This indicates that adequate packaging, both flexible and rigid, is sufficient to keep the devices stable. Nevertheless, in case of insufficient water protection, like a damaged edge seal or a broken front sheet, humidity could enter a CIGS device. This can have a negative impact on especially the conductivity of the front contact and the monolithic interconnection.
In CIGS devices, several types of transparent conductive oxides In the case of non-encapsulated CIGS solar cells, thus allowing water ingress, increased resistivity of ZnO:Al is often found to be a major cause for efficiency loss, and even a minor resistivity increase will directly impact the device performance due to series resistance increases. 80 Damp heat related resistivity increase of ZnO:Al is primarily caused by a decrease of carrier mobility due to grain boundary degradation. This is typically caused by the diffusion of 'foreign' species, like water and CO 2 , from the environment into the grain boundaries, 81,82 where the potential barrier can then increase. 85,86 The resistivity increase was reported to be largely reversible by annealing in vacuum 87 or in a reducing atmosphere at elevated temperatures. 88 The more expensive ITO is generally more stable than ZnO:Al in the presence of humidity and elevated temperatures. Degradation of ITO can be caused by the migration of water and alkaline species into the layer leading to electrochemical instability. Temperature-humidity stress of this material was further shown to cause recrystallisation and local concentrations of In and Sn. 89 Another effect that can occur in the presence of humidity is the degradation of the conductive molybdenum film, which functions as the back contact. This material can oxidise if directly exposed to (liquid) water and oxygen, especially under elevated temperatures. Oxidation can first lead to the formation of black and blue stains on the metallic molybdenum surface, which can contain molybdenum oxide (MoO 2 /MoO 3 , potentially with sodium or selenium 90 ). These materials can be badly conducting and/or poorly reflecting. 83,84 This oxidation can mainly affect the scribes of monolithically interconnected devices, while it will not likely occur in the covered molybdenum back contact in the bulk of the material, due to the lack of direct water. In the location of the second scribe (see Figure 7, location P2), where a Mo/ZnO:Al contact is responsible for the current transport between solar cells, increased resistance of the scribe has been observed in model systems exposed to damp heat conditions. 79,[91][92][93] Possible explanations are the introduction of an oxide layer at the Mo/ZnO:Al interface as well as increased resistivity of ZnO:Al in this scribe. 80 Moreover, oxidation of the P3 scribe was also observed, for example on positions that have been damaged by the scribing process. As long as some conductive molybdenum is present, this is not per se a problem: If the layer is only partly degraded, the current can still laterally bypass via a non-degraded part. 93 Figure 8). 76

| Partial shading
The impact of partial shading strongly depends on the design of the module. Commercial CIGS modules can be divided into two classes. propagates over the cell area. These long and winding defects have a width of tens of micrometres and can have a length of multiple centimetres (see Figure 9). In these defects, the CIGS absorber material has recrystallised and formed into a thick semi-porous and likely conductive structure. Due to the volume expansion, a 'ridge' of elevated material 100-102 is formed. At these positions, the front F I G U R E 8 Optical images of CIGS solar modules (A) before and (B) after PID test of 1043 h in a chamber of 85 C and 85% RH and À1000 V. TCO corrosion is shown by the blue arrows and (C) is an electroluminescence (EL) image of the state of (B). 76 F I G U R E 9 A close-up image of worm like defects in two interconnected cells of a CIGS module exposed to partial shading stress. Reprinted with permission of Bakker et al. 104 contact is still intact but is lifted from its original position. The appearance of wormlike defects leads to the formation of localised shunts in the devices, negatively affecting the module output.
Although the performance loss of one wormlike defect can be minor, repeated exposure to harsh partial shades will lead to multiplication of the losses. 103,104 Alongside to wormlike defects, also nonpermanent changes in device performance were observed due to (mild) reverse bias exposure. 105 Since these effects were often reversible, for example for small cells under illumination, 106

| DSCs
The DSC was first reported by O'Regan and Grätzel in 1991 113 and may still have a role to play in energy generation for emerging indoor applications. 114 The indoor efficiency of DSCs is very impressive (e.g. 28.9% at 1000 lux). 115 However, the AM1.5G certified performance is still only 12.25% (0.0963 cm 2 aperture area) and decreases down to 8.8% for a sub-module (398.8 cm 2 device area), 116 as a result of absorption limitations.

| Device architecture
The DSC has several components; the photon absorption occurs within a dye molecule and charge separation and collection are a result from interfacing materials. 117 The remaining components that make up a DSC are the anode and cathode substrates, the anode and cathode electrodes, the electrolyte, and the encapsulant (see There are a significant number of up-to-date reviews that describe in depth the DSC (Vlachopoulos and Hagfeldt 117 is a recent example-a list of reviews can also be found within). The stress factors for the DSC are essentially the same as other technologies although the added complexity is the impact of electrolyte solvent egress and electrolyte corrosiveness (see Figure 10). We next describe the degradation pathways each of the device components suffers from when exposed to stress factors.

| Photoanode
Dye stability is the limiting factor for the photoanode. 120  Polymer-based counter electrodes can be considered the most promising alternative because they can be low-cost, transparent and flexible, while still exhibiting equivalent or superior catalytic activity.
One can argue that the counter-electrode is the final piece of the jigsaw puzzle which is a DSC, and thus, its development direction will depend on the dye/electrolyte combination.

| Electrolyte
The electrolyte is composed of a solvent (organic or aqueous) and a  In 2011, consensus standards were developed by the OPV community to provide a common framework to comprehensibly assess stability, as the IEC standards were considered too harsh to be considered meaningful. 137 A series of interlaboratory studies have been conducted on the stability of OPV devices. These papers highlight the complex relationships between materials, technological steps, degradation protocols and PV properties.

| OPVs
As with other technologies, the degradation of OPV solar cells is related to several stress factors that can be separated into both extrinsic and intrinsic factors. The intrinsic factors included the metastable morphology, stability of materials, the diffusion of the electrodes and buffer layers into the active material (see Table 1 for a fuller list); whereas the extrinsic factors include many of the same factors as other technologies (oxygen and water infiltration, irradiation, heating). [138][139][140] However, most OPV modules are made onto flexible substrates so bending is an additional mechanical stress factor. 141 There have been several review articles on OPV stability, and it is clear that degradation is also not due to a single intrinsic or extrinsic failure mechanism. 142 For example, this has been demonstrated by consideration of the combined effect of humidity and temperature on OPV degradation, leading to an interaction effect. When an OPV is stressed by both temperature and humidity, a greater degradation is observed than when each factor is increased individually. 143 Furthermore, by not considering the interaction effects in other reports, misleading conclusions can be reached due to the significant impact interactions have on the rate of degradation. 144 The effects of applying multiple stress factors on OPV modules simultaneously using a design of experiment approach were performed to demonstrate predictive ageing of OPVs based on multi-stress testing using a log-linear life model. 145 There has been a large body of research aimed at improving the stability; this has focused on active material design, device engineer- Despite these highly encouraging recent results, the next challenge is to scale this efficiency and stability to larger area modules.
One noticeable trait in the OPV community (as with the DSC and MHP ones) is that outdoor monitoring has been a sparingly adopted approach for testing the stability. By testing emerging PVs in outdoor conditions, multiple stress factors can be applied, and outdoor testing remains one of the best approaches to review stability as the PVs are subjected to multiple stress factors 151,152 including tests that aren't conducted in ISOS consensus standards such as the impact of condensation. 153  The operational stability of the first MHPs 159 was in the order of minutes rather than hours, and this has improved from minutes to days, to weeks, when the liquid electrolyte was replaced with a solidstate hole conductor, compositional engineering of perovskites, 163,164 passivation of perovskites and rational charge selective layers. In view of these rapid developments, it is not difficult to envisage that MHP will achieve a 25-year life time warranty, which is common for commercial silicon PV modules.
Typically, the efficient lead-based MHPs display a bandgap close to 1.5 eV, and lowering the bandgap further will be of importance. By substituting the Pb with Sn, a reduced bandgap can be obtained, and mixed Pb-Sn-based perovskites are potentially attractive to achieve the bandgap value of 1.2-1.3 eV. The energy gap is in the ideal range and is suited for perovskite-perovskite tandem application. 165    Indoor measurements do not necessarily replicate the stress conditions of real world outdoor operational conditions, with seasons, weather, day-night cycles and varying temperature load. To really assess the reliability of perovskite cells, we thus must leave the lab for the outdoor world. That has been done ( Figure 11C), but so far there are only a few reports on outdoor testing and the ones that exist describe data for very few cells. Exposure times longer than a month are available in only a few instances. The available data set for outdoor testing is currently too small to draw conclusions about MHP solar cell reliability in the field, but the increased stability seen with indoor testing indicates that within the next few years, we will see an increasing amount of field testing, which will be interesting to follow.
It is worth mentioning that most of the available data are on small area cells. Similar to other PV technologies, the efficiency tends to decrease when the area goes up due to local defects and material non-uniformity ( Figure 12D). Technology learning cycles are necessary to decrease the efficiency gap between small and larger areas.

| Failure modes and their mitigation
Before discussing about the failure modes of MHP, it is vital to know the structure of MHPs, as this has a bearing on the degradation routes. Typical MHPs consist of at least five layers with four interfaces (see Figure 12). Most MHPs consist of a TCO coated glass substrate as anode or cathode (depending on whether the configuration is pin or nip, respectively), a perovskite absorber layer, an electron selective n-type layer (e.g. SnO 2 , TiO 2 and PCBM) or an hole selective p-type layer (e.g. Spiro-OMeTAD, CuSCN and PTAA) and an anode/ cathode (Au, Cu or Ag), which is also dependent on cell architecture.
We can sub-divide MHPs into five classes depending on the placement and nature of the charge-transporting layer, namely, the planar n-i-p structure, planar p-i-n structure, the mesoscopic n-i-p structure, mesoscopic p-i-n structure and the triple mesoscopic structure  Atmosphere composition (moisture and oxygen) Perovskite layers are liable to degradation under exposure to moisture and air, to note here moisture, oxygen and UV radiation are indispensable for the degradation process. A number of engineering strategies have been adopted in order to improve the stability including interfacial engineering strategies with site-based substitution in the perovskite lattice, doping in charge transporting layers, passivation by using various materials (small molecules, polymers, ligands, perovskite quantum dots and lowdimensional perovskites) and a protective layer for vulnerable layers. 181 In the recent 'consensus statement for stability assessment', it was recommended that testing of devices was still to be conducted under nitrogen environments in order to limit the degradation due to atmospheric factors in order to study other degradation pathways. 182

Solar visible and UV illumination
MHPs suffer from photo-induced degradation. The origin is not particularly well understood but likely related to a number of degradation pathways. The mechanism of UV degradation is distinctive under different environments. Although the degradation under illumination at low temperatures in perovskite has shown to be insignificant, 183,184 it increases significantly in the presence of H 2 O and O 2 or in contact with other materials. When MHPs are exposed to light in the presence of oxygen only, the photo-generated electrons react with the O 2 to form superoxide (O 2 À ). 185 This superoxide oxidises the perovskite to PbI 2 , I 2 and CH 3 NH 2 À (Figure 13). The I 2 further oxidises the perovskite. Another unique aspect is that in the absence of H 2 O and O 2 , the degradation of MHPs by UV radiation is partly reversible under 1-sun illumination. 186 The UV degradation also impacts other layers in the device. 187 Specifically, the most common electron-transport layer, TiO 2 , is a typical photo-catalyst for oxidising organic materials 188 with a band-gap of 3.20 eV ($400-nm wavelength). It can photocatalyze the decomposition of hybrid perovskite at their interface. 189 The degradation mechanism in the interface of perovskite/TiO 2 consists of two stages 190 (see again Figure 13). Moreover, charge generation under F I G U R E 1 2 Annotated diagram of a MHP with four device configurations: mesoscopic structure, planar structure, triple mesoscopic structure, and tandem structure with a lower-bandgap subcell. In the mesoporous structure, a thin mesoporous scaffold (typically TiO 2 or Al 2 O 3 ) infiltrated with absorber material is present between a charge extraction layer and the polycrystalline absorber layer. 177 light illumination and subsequent trapping on the surface of perovskite has been shown to initiate moisture-induced irreversible degradation to PbI 2, CH 3 NH 2 and HI vapours. 191 Literature reports indicate that the organic cation could become loosely bound to PbI 6 4À octahedra after light exposure. 192,193 TiO 2 itself is susceptible to degradation under UV and compromises the durability of MHPs. 194  were reported for thermal degradation at variable temperatures. 203 The thermal degradation of CH 3 NH 3 PbI 3 can occur even at lower temperatures (80 C) under an inert atmosphere if exposed for extended time (>60 min), 204 and the decomposition reaction leads to the formation of ammonia (NH 3 ) and methyl iodide (CH 3 I) gas and lead iodide (PbI 2 ). 205 It also undergoes a tetragonal-to-cubic phase transition around 56 C. 203 The structure and PV performance of an operational CH 3   unlikely to be used in scalable production, the mechanism of metal diffusion from front and rear electrodes might need further evaluation in the future, 209 as it has also been witnessed in other technologies.

Mechanical stability
MHPs exhibit poor resistance to fracture and are considered extremely fragile in the presence of applied loads. 208 Table 1 has been prepared to highlight the key specific degradation and failure mechanisms for each PV technology in order to provide a summary for the reader. And the presentation in Table 2  and so on can pave the way for production with long lifetime. In addition, mature technologies such as silicon became very reliable once the process was more fully developed. For example, the reduction in sodium and potassium by cleanroom processing and the surface passivation to reduce recombination rates had a major effect.
Additionally, the encapsulation processes were optimised, which removed damage due to water ingress and mechanical loading. So, modules or mini-modules should be used for ageing tests to identify module level failure mechanisms in order to design possible mitigation steps. The maturation of this technology is reflected by the fact the community has devised standard testing protocols. In this direction, a close collaboration of inter-laboratory device testing will be of significant importance to understand and resolve the degradation mechanisms.
Aside from module packaging, which is similar between c-Si and thin film, the mature thin film technologies (CdTe, CIGS) provide more lessons for emerging technologies. Any cell degradation modes commonly found in thin film cells (such as pinholes, reverse bias, shunting, TCO corrosion) would be the main topics of concern for an emerging thin film PV technology.

| Module external stress factor effects
Certain stress factors are not well studied in emerging PV, such as soiling, chemical pollutants and, to some extent, mechanical loading.
The latter is particularly interesting; next generation PVs modules are often on flexible substrates so the types of mechanical stress are different to one would expect from a mature module. While it is commonplace to see tests on repeated bending, prolonged flexing under load such as wind might trigger a different failure mechanism. Indeed, the mounting format is likely to play a role in the mechanical stability.
Longer-term soiling and chemical pollution might have an influence.
The mismatch in thermal expansion coefficients and low fracture energy of layers in MHPs raise a concern as to whether devices can withstand mechanical stresses from temperature fluctuations. Large mismatches in CTE between adjacent materials could build up stress and lead to delamination during temperature cycling, which presents a direct path for moisture ingress to the solar cells. In addition, the metal oxide barrier layers used with flexible substrates might degrade under sustained chemical pollutant exposure.
Encapsulants need to be chosen carefully to be optically transparent, flexible enough to absorb any fluctuation in strain energy during temperature cycling, electrically insulating to mitigate PID, to have a reasonably low water vapour transmission rate and to not release byproducts that would be harmful to the electrical contacts and solar cell absorber (e.g. acetic acid and EVA).
One specific challenge faced with next generation modules is the production, which eventually could break the absorber materials. Stability of the encapsulants and edge sealants (if applicable) is required to minimise this. There is very little work in next generation PVs on edge sealants, and this is also an area that needs more research.
By simply considering internal stability issues of next generation technologies, it is clear that major improvements in encapsulation and module packaging are required. Many encapsulation strategies in literature are at low TRLs or too expensive for scalable use. Moving production to scale will add quality-engineering issues that are presently unknown; the EVA issues stipulated in Section 2.1.2 show how a stable material can give problems as companies move to mass manufacture. As the drive to low cost, mass manufacture starts for emerging technologies, issues with packaging will become more commonplace.

| CONCLUSION
The aim of the review paper was to describe technology specfic deg- Aside from module packaging, which is similar between c-Si and thin film, the mature thin film technologies (CdTe, CIGS) most probably will provide more lessons for emerging technologies. Understanding the long-term materials processes such as dopant or impurity diffusion will be common to all thin film technologies. In the case of CdTe, degradation associated with diffusion of the commonly used copper dopant is seen. Replacing copper with the much slower diffusing arsenic dopant has shown significantly reduced degradation. Any cell degradation modes commonly found in thin film cells (such as pinholes, reverse bias, shunting, TCO corrosion) would be the main topics of concern for an emerging thin film PV technology.
Understanding why failures happen is key for improvement in reliability. A detailed understanding of the various failure modes occurring during in-field operation of the solar cells is key to minimising or eliminating performance losses. One of the impediments for understanding the long-term behaviour of emerging solar cell technologies has been the issue of scalability. The vast majority of disseminated MHP devices were small sized laboratory scale devices with >94% having an active area of 0.2 cm 2 or less. There is no straightforward means of linking performance of small-scale devices to that of full-size modules, especially as edge effects become less significant.
Also, certain stress factors are not well studied in emerging PV, such as soiling, chemical pollutants and, to some extent, mechanical loading. The latter is particularly interesting as next generation PVs modules are often on flexible substrates, and there, a different mechanical stress distribution is to be expected.
Summarised, it is clear for the commercial viability of any PV technology that reliability is key, and it should not be separated from upscaling and other product development aspects. Therefore, especially for emerging lab scale PV technologies, it is important to explore how reliability can be optimised already under laboratory conditions.

ACKNOWLEDGEMENT
This review article is based upon work from COST Action CA16235 PEARL PV, WG2, supported by COST (European Cooperation in Science and Technology), a funding agency for research and innovation networks.

DATA AVAILABILITY STATEMENT
Data sharing is not applicable to this article as no new data were created or analyzed in this study.